01-S - Antoniou - Lie Symmetries, Differential Equations and Applications

Lie Symmetries,
Differential Equations
and Applications
A Review of Theory and Computational Methods
Solomon M. Antoniou
SKEMSYS
Scientific Knowledge Engineering

and Management Systems
Corinthos 20100, Greece
July 2023
solomon antoniou@yahoo.com
Abstract
We review the Lie Symmetry Method in solving Differential Equa-
tions and Systems of Differential Equations.
Keywords: Lie Symmetries, Ordinary Differential Equations, Partial

Differential Equations, Systems of Differential Equations, Exact Solutions,
Conservation Laws.
1
Contents
Introduction
Chapter I. General Considerations

1. Definition of a Group
2. One-parameter Lie Group
3. Transformation of Differential Equations
4. Infinitesimal Transformations
5. Vector Fields and Flows
6. The Lie Derivative
7. The Lie Invariance Condition.
The General Algorithm
Chapter II. Ordinary Differential Equations

8. First Order Ordinary Differential Equations
9. Second Order Ordinary Differential Equations
10. Integrating Factors and Linearization
11. Third Order Ordinary Differential Equations.
Linearization. Higher Order ODEs
Chapter III. Partial Differential Equations

12. Partial Differential Equations I.
First Order PDEs
13. Partial Differential Equations II.
Second Order PDEs
14. Boundary Value Problems for PDEs
Chapter IV. Higher Order Partial Differential

Equations
15. Preliminaries
16. The KdV Equation
17. The Boussinesq Equation
18. The BBN Equation
19. Partial Differential Equations of Higher Dimensions
2
Chapter V. Non-Classical Symmetries
20. Non-Classical Symmetries
Chapter VI. Fundamental Solutions

21. Fundamental Solutions through Lie Symmetries
Chapter VII. The Direct Method

22. The Direct Method of Clarkson and Kruskal
Chapter VIII. Systems of Differential Equations

23. First Order Systems of Ordinary Differential Equations
24. Second Order Systems of Ordinary Differential Equations
25. Systems of Partial Differential Equations of the First Order
26. Systems of Partial Differential Equations of the Second Order
Chapter IX. The Optimal System

27. The Optimal System of Lie Algebras
Chapter X. Potential Symmetries

28. Potential Symmetries
Chapter XI. Generalized Symmetries

29. Lie-Bäcklund Symmetries
30. Lie-Bäcklund Symmetries for the Burgers Equation
in Potential Form
31. Lie-Bäcklund Symmetries for the Burgers Equation
32. Lie-Bäcklund Symmetries for the Fokker-Planck
Equation
33. Lie-Bäcklund Symmetries for a Nonlinear
Wave Equation
Chapter XII. Recursion Operators

34. Recursion Operators
Chapter XIII. Conditional Symmetries

35. Theory of Conditional Symmetries
36. Examples of Conditional Symmetries
3
Chapter XIV. Contact Symmetries
37. Contact Symmetries. Theory and an Example
Chapter XV. Conservation Laws

38. Variational Symmetries
39. The Inverse Lagrangian Problem
40. Hamilton’s Variational Principle
41. Some Auxiliary Operators and their Identities
42. Euler-Lagrange Equations in Several Variables
43. Noether’s Theorem
44. Adjoint Equations
45. The Main Conservation Theorem
46. The Double Reduction I
47. The Double Reduction II
Chapter XVI. Ricci Flow

48. Two-Dimensional Ricci Flow
Chapter XVII. Applications in Quantum Mechanics

49. The Schrödinger Equation
50. A Nonlinear Schrödinger Equation
Chapter XVIII. Applications in Epidemiology

51. The SIR Model
52. HIV Epidemiology
Chapter XIX. Some Special Equations

53. The Thomas Equation
54. The Fokker-Planck Equation
55. The Fisher Equation
56. The Fitzhugh-Nagumo Equation
57. A Nonlinear Diffusion Equation I
58. A Nonlinear Diffusion Equation and its Potential Form
59. A Nonlinear Diffusion Equation II
60. The Axially Symmetric Wave Equation
61. The Telegraph Equation
62. The Laplace Equation
4
Chapter XX. The Physics of Fluids
63. The Shallow Water Wave Equation
64. The Euler Equations
65. The Navier-Stokes Equations
Chapter XXI. General Relativity

66. The Einstein Equations for Axially Symmetric
Gravitational Fields
67. The Einstein-Maxwell Equations for Axially Symmetric
Gravitational Fields
68. Einstein Equations with Axial Symmetry and
Rotation
Chapter XXII. Differential Geometry

69. The Minimal Surface Equation
Chapter XXIII. Financial Applications

70. The Black-Scholes Equation
71. Portfolio Optimization and Stochastic Control
72. A General Solution to the Stochastic Equity Volatility
Equation (The BCG Equation)
73. Asian Options. The Geometric Average Case
74. Asian Options. The Arithmetic Average Case
Chapter XXIV. Applications in Cosmology

75. The Companeets Equation
Chapter XXV. Additional Equations

76. The mKdV-KP Equation
77. The Zabolotskaya-Khokhlov Equation
78. The Kuramoto-Sivashinsky Equation
79. The Kadomtsev-Petviashvili Equation
80. The Rosenau Equation
Chapter XXVI. Fractional Differential Equations

81. Lie Symmetries of Fractional Differential Equations
5
Chapter XXVII. Special Solution Methods of
Nonlinear PDEs
82. The Tanh-Coth Method
83. The Sech Method
84. The Auxiliary Equation Method I
85. The Exp-function Method
86. The (G0 /G) Method
87. The Sine-Cosine Method
88. The Projective Riccati Equation Method
89. The Riccati Equation Method with Variable
Expansion Coefficients
90. The Abel Equation Method with Variable
Expansion Coefficients
91. The Jacobi Elliptic Equation Method I
92. The Jacobi Elliptic Equation Method II
93. The Weierstrass Equation Method
94. The Homotopy Method
95. The F-Expansion Method
96. The Adomian Decomposition Method
97. The Simple Equation Method
98. The Auxiliary Equation Method II
99. The Variational Iteration Method
100. Painlevé Analysis
101. The Inverse Scattering Method. The Lax Pair.
The Darboux Method
102. The Hirota Direct Method
103. The Algebro-Geometric Method
Appendix A Appendix F
Appendix B Appendix G
Appendix C Appendix H
Appendix D Appendix I
Appendix E Appendix J
General References
6
Introduction
The Symmetry Method of solving differential equations was introduced by
Sophus Lie, a Norwegian mathematician at the end of 19th century. Lie’s
method has been applied in many areas of Science and Engineering and also
in Finance.
This is a comprehensive review explaining the essentials of the method. It
focuses mainly to problem-solving techniques to finding explicit solutions to
Ordinary and Partial Differential Equations and to Systems of Differential
Equations. It includes many examples from a number of areas, including the
Mechanics of Fluids, General Relativity and Finance.
A number of examples have been solved using some available software like
MAPLE and Mathematica.
We have tried to be as explicit as we could. After reading this text, the
reader hopefully, will be able to start her/his own research.
Chapter I
General Considerations
1 Definition of a Group
Let G be a set of elements a, b, c, ... and a composition law φ(a, b),
(a, b) → φ(a, b) (1.1)
such that
1. If a, b ∈ G then φ(a, b) ∈ G. This is the closure property.
2. If a, b, c ∈ G then φ(a, φ(b, c)) = φ(φ(a, b), c). This is the associative
property.
3. If a ∈ G, then there exists an e ∈ G such that φ(a, e) = φ(e, a) = a for
every a. The element e is called the identity.
4. If a ∈ G, then there exists a unique element a−1 ∈ G such that φ(a, a−1 ) =
φ(a−1 , a) = e. The element a−1 is called the inverse of the element a.
The set G together with the composition law φ(a, b) satisfying the above
axioms 1-4, is called a group.
7
Identity is a unique element of the group. To each element of the group,
there corresponds a unique inverse.
2 One-parameter Lie group

A transformation Rn →Rn
(x1 , x2 , ..., xn ) → (x̂1 , x̂2 , ..., x̂n )
written also as
x→x̂, where x = (x1 , x2 , ..., xn ) and x̂ = (x̂1 , x̂2 , ..., x̂n )
with
x̂i = f i (x1 , x2 , ..., xn ; ), or x̂i = f i (x; ), i = 1, 2, ..., n
is called a one-parameter local Lie group if

1. f i are smooth functions of the variables x1 , x2 , ..., xn
2. f i are analytic functions in the parameter , that is a function with
convergent Taylor series in .
3. = 0 can always be chosen to correspond to the identity element e.
4. There is a composition law between the f i :
f i (f (x, a), b) = f i (x, φ(a, b)), with f i (x; 0) = xi (2.1)
where φ(a, b) is a given composition law, for example φ(a, b) = a + b.

Example. The transformation
xy
x̂ = , ŷ = y − (2.2)
y−
is an one-parameter Lie group.
Proof. For
xy
x̂ = , ŷ = y − a (2.3)
y−a
and
xy
ˆ
x̃ = , ỹ = ŷ − b (2.4)
ŷ − b
8
we have
xy
(y − a)
y−a xy
x̃ = = and ỹ = y − (a + b) (2.5)
y−a−b y − (a + b)
We thus see that the closure relation is being satisfied with composition law
φ(a, b) = a + b. Since φ(a, b) = a + b, we have
φ(a, φ(b, c)) = a + (b + c) = (a + b) + c = φ(φ(a, b), c) (2.6)
which means that associativity is being satisfied too.

The identity element is e = 0, since φ(a, 0) = φ(0, a) = 0 + a = a for every a.
Finally the inverse is a−1 = −, since φ(, −) = + (−) = 0.
3 Invariance of Differential Equations

In this section we consider a number of differential equations which are in-
variant under a local Lie group of transformations.
I. Ordinary Differential Equations.
Example 1. We consider the ordinary differential equation
dy
x2 = x2 y 2 − xy − 1 (3.1)
dx
We introduce a scale transformation (x, y)→(x̂, ŷ) with
x = λα x̂, y = λβ ŷ (3.2)
where λ is a scale parameter. Since

dy dŷ
= λβ−α (3.3)
dx dx̂
equation (3.1) transforms into
dŷ
λα+β x̂2 = λ2(α+β) x̂2 ŷ 2 − λα+β x̂ŷ − 1 (3.4)
dx̂
Scale invariance requires α + β = 0, i.e. β = −α. Therefore scale invariance
of equation (3.1) is implemented by the transformation
x = λα x̂, y = λ−α ŷ (3.5)
9
We thus see that (3.1) is invariant under the local transformation group
x = e x̂, y = e− ŷ (3.6)
where is a parameter.
Going back, we could equally well consider the transformation
x = eλ x̂, y = eµ ŷ (3.7)
and require invariance of (3.1). We will then obtain λ + µ = 0, i.e. recover

transformation (3.6).
Example 2. We consider next the ordinary differential equation
d2 y α dy dy 2
y + − + βy 3 = 0 (3.8)
dx2 x dx dx
Let us now consider invariance of (3.8) under the transformation
x = eλ x̂, y = eµ ŷ (3.9)
Since
dy dŷ d2 y 2
(µ−2λ) d ŷ
= e(µ−λ) and = e (3.10)
dx dx̂ dx2 dx̂2
equation (3.8) is being transformed into the equation
µ
d2 ŷ
(µ−2λ) α (µ−λ) dŷ (µ−λ) dŷ 2
e ŷ e + e − e
dx̂2 eλ x̂ dx̂ dx̂
+ β(eµ ŷ)3 = 0
The above equation is equivalent to

d2 ŷ α dŷ dŷ 2
e2(µ−λ) ŷ + − e2(µ−λ)
dx̂2 x̂ dx̂ dx̂
3µ 3
+ βe ŷ = 0
which can also be written as

d2 ŷ α dŷ dŷ 2
ŷ + − + βe(µ+2λ) ŷ 3 = 0 (3.11)
dx̂2 x̂ dx̂ dx̂
10
Invariance requires µ + 2λ = 0, i.e. µ = −2λ. Therefore equation (3.8) is
invariant under the transformation
x = e x̂, y = e−2 ŷ (3.12)
where is a parameter.
II. Partial Differential Equations.
Example 1. We consider the heat equation
∂u ∂ 2u
= , u = u(x, t) (3.13)
∂t ∂x2
We introduce a scale transformation (x, y)→(x̂, ŷ) with
x = λα x̂, t = λβ t̂, u = λγ û (3.14)
where λ is a scale parameter. Since
∂u ∂ û ∂ 2u 2
γ−2α ∂ û
= λγ−β and = λ (3.15)
∂t ∂ t̂ ∂x2 ∂ x̂2
∂ û ∂ 2 û
λγ−β = λγ−2α 2
∂ t̂ ∂ x̂
The above equation can also be written as
∂ û ∂ 2 û
= λβ−2α 2 (3.16)
∂ t̂ ∂ x̂
Invariance requires β − 2α = 0, i.e. β = 2α. Therefore the heat equation is
invariant under the scale transformation
x = λα x̂, t = λ2α t̂, u = λγ û (3.17)
We shall show in the next sections that there are also different types of
invariance of differential equations which will serve to solve a number of
diffrential equations not solveble by obvious transformations.
11
4 Infinitesimal Transformations
Consider the local one-parameter transformation
x̂ = f (x, y; ), ŷ = g(x, y; ) (4.1)
with
f (x, y; 0) = x and g(x, y; 0) = y (4.2)
We expand x̂ and ŷ in a Taylor series around = 0:

∂f
x̂ = f (x, y; 0) + + O(2 ) (4.3)
∂ =0
∂g
ŷ = g(x, y; 0) + + O(2 ) (4.4)
∂ =0
If we let
∂f ∂g
= X(x, y) and = Y (x, y) (4.5)
∂ =0 ∂ =0
and taking into account (4.3), (4.4) and the conditions (4.2), we have
x̂ = x + X(x, y) + O(2 )

(4.6)
ŷ = y + Y (x, y) + O(2 )
Expressions (4.6) are referred to as infinitesimal transformations while

X and Y simply as infinitesimals.
Theorem 1. If the infinitesimals X and Y are known, we can recover the
Lie group by solving the system
dx̂
= X(x̂, ŷ), x̂|=0 = x
d (4.7)
dŷ
= Y (x̂, ŷ), ŷ|=0 = y
d
Proof. We shall prove a more general Theorem (see Theorem 2 below) just
after the Example.
Example. Consider the infinitesimal transformation
x̂ = x + x2 + O(2 ), ŷ = y + xy + O(2 ) (4.8)
12
In this case we have
X = x2 and Y = xy (4.9)
The Lie equations are given by

dx̂
= x̂2 , x̂|=0 = x
d (4.10)
dŷ
= x̂ŷ, ŷ|=0 = y
d
The first equation of the above system takes on the form by separation of
variables
dx̂
= d
x̂2
and by integration
Z Z
dx̂ 1
= d =⇒ − = + C1
x̂2 x̂
and thus
1
x̂ = − (4.11)
+ C1
The second equation of the system (4.10) takes the form by separation of
variables and taking into account (4.11)
dŷ dŷ 1
= x̂d =⇒ =− d
ŷ ŷ + C1
and by integration
ln(ŷ) = − ln( + C1 ) + C =⇒ ln(ŷ) = − ln( + C1 ) + ln(C2 )
and then
C2
ŷ = (4.12)
+ C1
13
Let us now take into account the initial conditions in order to determine the
”constants” C1 and C2 . We have, taking into account the expressions (4.11)
and (4.12), we have found for x̂ and ŷ :
1 1 1
x̂|=0 = x⇐⇒ − = x =⇒ − = x =⇒ C1 = −
+ C1 =0 C1 x
(4.13)
C2 C2 y
ŷ|=0 = y⇐⇒ = y =⇒ = y =⇒ C2 = −
+ C1 =0 C1 x
We can now find the explicit expressions for x̂ and ŷ:

1 1 x
x̂ = − =− = (4.14)
+ C1 1 1 − x
−
x
y
C2 −
ŷ = = x = y (4.15)
+ C1 1 1 − x
−
x
Conclusion: The one-parameter Lie group of transformation (4.8) is given by
x y
x̂ = , ŷ = (4.16)
1 − x 1 − x
It is obvious that upon expanding in Taylor series the right hand side of the
above expressions, we recover the infinitesimal transformation (4.8).
Let G be a local group of transformations defined by
x̂i = f i (x; ) (4.17)
where f i satisfy
f i (f (x, a); ) = f i (x; a + ), with f i (x; 0) = xi (4.18)
We can prove that
xi = f i (x̂; −) (4.19)
Upon expanding x̂i = f i (x; ) to first order in we get
x̂i = xi + ξ i (x) (4.20)
14
where
∂
ξ i (x) = f i (x; ) (4.21)
∂ =0
Let us call ξ i (x) the infinitesimal transformation of the group G. From

the geometrical point of view, equation (4.21) defines the tangent vector
ξ(x) = (ξ 1 (x), ξ 2 (x), · · · , ξ n (x)) (4.22)
at the point x. Therefore ξ is called the tangent vector field of group G.

The tangent vector field is associated to the first order differential operator
∂
X = ξi (summation over i) (4.23)
∂xi
which is called the generator of the group G.
Theorem 2. Let ξ i (x) be the infinitesimal transformation of the group G.
Then x̂i = f i (x; ) is the solution of the system
dx̂i
= ξ i (x̂), x̂i |=0 = xi (i = 1, · · · , n) (4.24)
d
This is a system of first order ordinary differential equations. It is usually
called the Lie equations. The solution of the system provides the one-
parameter local group once the infinitesimal transfromation is known. In
other words, the transformations x̂i = f i (x; ) are determined by integrating
the Lie equations.
Proof. We consider f i (f (x, a); ) = f i (x; a + ) and we Taylor expand the
two members. Expanding the LHS we obtain

i i ∂
f (f (x, a); ) = f (x; a) + f (f (x, a); ) · + O(2 )
i
(4.25)
∂ =0
Expanding the RHS, we obtain

i i ∂
f (x; a + ) = f (x; a) + f (x; ) · + O(2 )
i
(4.26)
∂ =0
Upon comparing the two expansions, we get

∂ i ∂ i
f (f (x, a); ) = f (x; ) (4.27)
∂ =0 ∂ =0
15
which is equivalent to
dx̂i
= ξ i (x̂) (4.28)
d
This completes the proof of the Theorem.
Let us now pose two questions:

Question 1. ”Why do we need the infinitesimals?”
Answer. In general, solving a differential equation, we need a transfor-
mation rule in order to reduce the equation into a simpler one which can
be solved using known methods. For example using a transformation (from
Cartesian to Canonical coordinates) r = r(x, y), s = s(x, y) one can convert
ds
a given DE into another one of the form = F (r) which can be solved by
dr
separation of variables.
Question 2. ”How one can determine the infinitesimals from a given differ-
ential equation?”
Answer. Given a DE (either ODE or PDE) we don’t know the Lie group
which leaves the DE invariant. However the infinitesimals X and Y can be
determined using Lie’s Invariance Condition.
5 Vector Fields and Flows

Let M be a differentiable manifold of dimension n.
A vector field V on the manifold M is a C ∞ −section of T M , i.e. it is a
C ∞ −map from M to T M that assigns to each point x∈M a vector in Tx M .
Here Tx M is the tangent plane of the manifold M at the point x. In a local
coordinate system x = (x1 , x2 , · · · , xn )∈M , the vector V can be expressed
as
n
X ∂
V = ξi (x) (5.1)
i=1
∂xi
where ξi (x)∈C ∞ (M ), i = 1, 2, · · · , n.
A curve γ at a point x∈M is a differentiable map γ : I−→M where I is
a subinterval of the set of real numbers containing zero, i.e. 0∈I, such that
γ(0) = x. An integral curve of the vector field at the curve γ at point
16
x∈M is a tangent vector γ̇(t) of the curve γ(t) which coincides with V at
the point x = γ(t) such that γ̇(t) = V |γ(t) for each t∈I. In a local system
of coordinates, the components of the curve satisfy the following system of
ODEs
dγi (t)
= ξi (γ(t)), i = 1, 2, · · · , n (5.2)
dt
The above system admits a unique solution for which γ(0) = x0 , according to
the existence and uniqueness theorem of ODEs. Therefore there is a unique
maximal integral curve γ(t) passing through the point x0 = γ(0)∈M . Such a
maximal integral curve is called the flow of V at the point x = γ(t) denoted
by Φ(t, x). We then impose the conditions
d
Φ(0, x) = x, Φ(s, Φ(t, x)) = Φ(t + s, x), Φ(t, x) = V |Φ(t,x) , x∈M (5.3)
dt
for all t, s∈M . It is supposed that parameters t, s are sufficiently small.
A more convenient notation for the flow is
Φ(t, x) = exp(tV x), Φ(0, x) = exp(tV x)|t=0 = x (5.4)
It is obvious that this parameterization of the flow satisfies the exponential
rules. On the other hand, using the above properties we get
−1
−1
Φ(−t, x) = Φ (t, x) or exp(tV ) (x) = exp(−tV )x (5.5)
The flow can also be expressed in infinitesimal form as

exp(tV )x = x + t V |x + O(t2 ) (5.6)
upon Taylor expanding and using the (initial) condition exp(tV )x|t=0 = x.
The flow exp(tV )x generated by the vector field V is also called a one-
parameter group of transformations since it is a Lie group acting on
the manifold M .
Given a flow satisfying the first two properties, we can reconstruct its vector
field V by differentiating the flow:
d
V |x = exp(tV )x, x∈M (5.7)
dt t=0
The inverse process of constructing the flow is usually called exponentiation

of V . Rectification of a vector field V in a neighborhood of a regular point
is always possible.
17
6 The Lie Derivative
Let V be a vector field in local coordinates
n
X ∂
V = ξi (x)
i=1
∂xi
Application of a vector field V to a smooth function f : M −→R also referred

to as the Lie derivative of f with respect to the vector field, determines
the infinitesimal rate of change of f as the parameter t of the induced flow
exp(tV )x of V varies:
n
X ∂f d
V (f (x)) = ξi (x) = f exp(tV )x (6.1)
i=1
∂xi dt t=0
7 The Lie Invariance Condition.

The General Algorithm
7.1. Preliminaries and Notation. Let us consider a system of n−th order
differential equations
Eν (x, u, u(1) , · · · , u(n) ) = 0, n = 1, 2, · · · , N (7.1)
Here
x = (x1 , x2 , · · · , xp )∈Rp , u = (u1 , u2 , · · · , uq )∈Rq (7.2)
are the independent and the dependent variables on the space E,
E = X×U 'Rp ×Rq (7.3)
The symbol u(k) denotes the partial derivative of order k of the components
uα of u. We shall also use the notation uα,J for the partial derivatives
∂ J uα
uα,J = (7.4)
∂xJ
where J is a symmetric multi-index of order k
J = (j1 , j2 , · · · , jk ), 1≤jν ≤p, k = j1 + j2 + · · · + jk (7.5)
18
The n−th order jet space J n (x, u(n) ) = J n E provides a set of local coor-
dinates (x, u(n) ). If p = 1 and q = 1 we have a single scalar differential
equation
E(x, u, u1 , · · · , un ) = 0 (7.6)
where
u1 = ux , u2 = uxx , · · · , un = u(n) (7.7)
The system Eν = 0 defined by a collection of smooth functions

E = (E1 , E2 , · · · , EN ) can be identified with a variety
SE = {(x, u(n) ) : E = 0} (7.8)
contained in the n−th order jet space J n with local coordinates (x, u(n) ).
A classical symmetry group of the system Eν = 0 is a local group G of
point transformations
Φ : E−→E (7.9)
Φ is an invertible map, mapping solutions to solutions
Φ : (x̂, û) = g·(x, u) (7.10)
Therefore a local Lie group of point transformations G is called a symmetry

group of the system of differential equations Eν = 0 if fˆ = g·f is a solution
whenever f is.
To find the symmetry group, the prolonged transformation
P r(n) Φ : J n −→J n (7.11)
is required to preserve the differential structure of the manifold SE .

Let Φ = exp(V ) be a one-parameter subgroup of the group G and let
n q
X ∂ X ∂
V = ξi (x) + φα (x, u) (7.12)
i=1
∂xi α=1 ∂uα
be the infinitesimal generator of Φ . The infinitesimal generator of the pro-

longed one-parameter subgroup P r(n) Φ is defined to be the prolongation
of the vector field V .
19
Definition. The n−th prolongation P r(n) V of V is a vector field on the
n−th jet space J n defined by
d
P r(n) V = P r(n) Φ (x, un ) (7.13)
(x,u(n) ) d =0
for every (x, un )∈J n .

If we integrate P r(n) V we find the prolongation of the group action P r(n) Φ
on the space J n . The prolonged vector field P r(n) V has the form
p q
(n)
X
i ∂ XX ∂
Pr V = ξ i
+ φJα (7.14)
i=1
∂x α=1 J
∂uα,J
where the coefficients φJα are given by the formula

p p
i ∂uα ∂uα,J
X X
J
φα = DJ φα − ξ i
+ ξi (7.15)
i=1
∂x i=1
∂xi
where DJ is the total derivative operator, i.e. DJ = Dj1 · · ·Djk , 1≤jν ≤p and
Di is the total derivative operator defined by
q
∂ XX ∂
Di = i
+ uα,J,i (7.16)
∂x α=1 J
∂uα,J
It is obvious that Di involves only finite terms of order J≤n where n is the
highest order derivative the function u appears.
7.2. Conclusion and General Algorithm. The determination of symme-
tries of ODEs or PDEs (or even systems of ODEs or PDEs) is based on two
Theorems given in Olver (Olver [39]). These two Theorems is all we really
need to determine symmetries. However these Theorems will be taking a
specific form depending on the given equation(s).
Theorem 1. Let
p q
X ∂ X ∂
V = ξ i (x, u) i + φα (x, u) α (7.17)
i=1
∂x α=1
∂u
be a vector field defined on an open subset M ⊂X×U (X is the space of
independent variables and U is the space of dependent variables). The n−th
prolongation of V is the vector field
q
(n)
X X ∂
Pr V =V + φJα (x, u(n) ) (7.18)
α J
∂uαJ
20
defined over the J = (j1 , j2 , · · · , jk ) with
p p
X X
φJα = DJ φα − ξ i uαi + ξ i uαJ,i (7.19)
i=1 i=1
where
∂uα ∂uαJ
uαi = and uαJ,i = (7.20)
∂xi ∂xi
is the sum over all J 0 s of order 0≤]J≤n, where n is the derivative of the
P
J
highest order.
Theorem 2. Suppose that
∆ν (x, u(n) ) = 0, ν = 1, 2, · · · , s (7.21)
is a system of differential equations. If G is a local group of transformations

acting on M ⊂X×U and
P r(n) V [∆ν (x, u(n) )] = 0, ν = 1, 2, · · · , s (7.22)
whenever ∆ν (x, u(n) ) = 0 for every infinitesimal generator V of G, then G is

a symmetry group of the system.
From (7.6), expanding and equating the coefficients of the derivatives of the
unknown function(s) to zero, we obtain a system of PDEs. This is called the
determining system of equations. Solving this system, we determine the
functions ξ 1 , ξ 2 , · · · and φ1 , φ2 , · · · , we thus find the vector V (or the vector
Γ in an alternative notation)
∂ ∂ ∂ ∂
V = ξ1 1
+ ξ 2 2 + · · · + φ1 + φ2 + ··· (7.23)
∂x ∂x ∂u1 ∂u2
or
∂ ∂ ∂ ∂
Γ=X +Y + ··· + U +Y + ··· (7.24)
∂x ∂y ∂u ∂y
The functions ξ i and φα contain a number of parameters α1 , α2 , · · · . The

vector Γ is then expressed as a linear combination
Γ = α1 Γ(1) + α2 Γ(2) + · · · (7.25)
21
where now the vectors Γ(1) , Γ(2) , · · · constitute the generators of the Lie Al-
gebra of symmetries of the given DE or the system of DEs.
Using Γ(1) , Γ(2) , · · · , we can reduce the original equation(s) into another equa-
tion(s) with less variables which can be solved using normal coordinates
or the invariant surface condition.
Example 1. Let the equation ut = F (u, ux , uxx , · · · ) admits the symmetry
generator Γ
∂ ∂ ∂
Γ = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u) (7.26)
∂x ∂t ∂u
We also suppose that Γ can be split into a combination of X (i) , i = 1, 2, 3
where
∂ ∂ ∂ ∂
X (1) = , X (2) = , X (3) = t − (u + 1) (7.27)
∂x ∂t ∂t ∂u
In the next we shall consider a number of generators Γ(i) , i = 1, 2, · · · , 6 where
Γ(1) = X (1) , Γ(2) = X (2) , Γ(3) = aX (1) + X (2) , Γ(4) = X (3) , Γ(5) = X (1) + X (3) ,
Γ(6) = X (2) + X (3) , i.e.
∂ ∂ ∂ ∂
Γ(1) = , Γ(2) = , Γ(3) = a + ,
∂x ∂t ∂x ∂t
∂ ∂ ∂ ∂ ∂
Γ(4) = t − (u + 1) , Γ(5) = + t − (u + 1) , (7.28)
∂t ∂u ∂x ∂t ∂u
∂ ∂
Γ(6) = (t + 1) − (u + 1)
∂t ∂u
We shall determine the group action Gi corresponding to each one of Γ(i)
using the Theorem of Section 5:
dx̂
= ξ(x̂, t̂, û), x̂|=0 = x
d
dt̂
= τ (x̂, t̂, û), t̂|=0 = t
d (7.29)
dû
= φ(x̂, t̂, û), û|=0 = u
d
∂
(i) For the generator Γ(1) = we have
∂x
ξ = 1, τ = 0, φ=0 (7.30)
22
We thus have to solve the system
dx̂ dt̂ dû

= 1, = 0, =0 (7.31)
d d d
Solving the above system we find
x̂ = + C1 , t̂ = C2 , û = C3 (7.32)
We have further, taking into account the initial conditions
x̂|=0 = x =⇒ ( + C1 )|=0 = x =⇒ C1 = x
t̂|=0 = t =⇒ (C2 )|=0 = t =⇒ C2 = t (7.33)
û|=0 = u =⇒ (C3 )|=0 = u =⇒ C3 = u
We thus get from (7.32), taking into account the above results
x̂ = + x, t̂ = t, û = u (7.34)
Therefore the group action G1 corresponding to Γ(1) is given by
G1 : (x, t, u)−→(x + , t, u) (7.35)

∂
(ii) For the generator Γ(2) = we have
∂t
ξ = 0, τ = 1, φ=0 (7.36)
dx̂ dt̂ dû

= 0, = 1, =0 (7.37)
d d d
x̂ = C1 , t̂ = + C2 , û = C3 (7.38)
x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

t̂|=0 = t =⇒ ( + C2 )|=0 = t =⇒ C2 = t (7.39)
û|=0 = u =⇒ (C3 )|=0 = u =⇒ C3 = u
23
x̂ = x, t̂ = + t, û = u (7.40)
G2 : (x, t, u)−→(x, t + , u) (7.41)

∂ ∂
(iii) For the generator Γ(3) = a + we have
∂x ∂t
ξ = a, τ = 1, φ=0 (7.42)

dx̂ dt̂ dû
= a, = 1, =0 (7.43)
d d d
x̂ = a + C1 , t̂ = + C2 , û = C3 (7.44)

x̂|=0 = x =⇒ (a + C1 )|=0 = x =⇒ C1 = x
t̂|=0 = t =⇒ ( + C2 )|=0 = t =⇒ C2 = t (7.45)
û|=0 = u =⇒ (C3 )|=0 = u =⇒ C3 = u
x̂ = a + x, t̂ = + t, û = u (7.46)
G3 : (x, t, u)−→(a + x, t + , u) (7.47)

∂ ∂
(iv) For the generator Γ(4) = t − (u + 1) we have
∂t ∂u
ξ = 0, τ = t, φ = −(u + 1) (7.48)

dx̂ dt̂ dû
= 0, = t̂, = −(û + 1) (7.49)
d d d
24
x̂ = C1 , t̂ = C2 ·e , 1 + û = C3 ·e− (7.50)
x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

t̂|=0 = t =⇒ (C2 ·e )|=0 = t =⇒ C2 = t (7.51)
û|=0 = u =⇒ (C3 ·e− − 1)|=0 = u =⇒ C3 − 1 = u
x̂ = x, t̂ = t·e , û = (u + 1)·e− − 1 (7.52)
G4 : (x, t, u)−→(x, t·e , (u + 1)·e− − 1) (7.53)
∂ ∂ ∂
(v) For the generator Γ(5) = + t − (u + 1) we have
∂x ∂t ∂u
ξ = 1, τ = t, φ = −(u + 1) (7.54)
dx̂ dt̂ dû

= 1, = t̂, = −(û + 1) (7.55)
d d d
x̂ = + C1 , t̂ = C2 ·e , 1 + û = C3 ·e− (7.56)
x̂|=0 = x =⇒ ( + C1 )|=0 = x =⇒ C1 = x
t̂|=0 = t =⇒ (C2 ·e )|=0 = t =⇒ C2 = t (7.57)
û|=0 = u =⇒ (C3 ·e− − 1)|=0 = u =⇒ C3 − 1 = u
x̂ = + x, t̂ = t·e , û = (u + 1)·e− − 1 (7.58)
25
G5 : (x, t, u)−→( + x, t·e , (u + 1)·e− − 1) (7.59)
∂ ∂
(vi) For the generator Γ(6) = (t + 1) − (u + 1) we have
∂t ∂u
ξ = 0, τ = 1 + t, φ = −(u + 1) (7.60)
dx̂ dt̂ dû

= 0, = 1 + t̂, = −(û + 1) (7.61)
d d d
x̂ = C1 , 1 + t̂ = C2 ·e , 1 + û = C3 ·e− (7.62)
x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

t̂|=0 = t =⇒ (C2 ·e − 1)|=0 = t =⇒ C2 − 1 = t (7.63)
û|=0 = u =⇒ (C3 ·e− − 1)|=0 = u =⇒ C3 − 1 = u
x̂ = x, t̂ = (t + 1)·e − 1, û = (u + 1)·e− − 1 (7.64)
G6 : (x, t, u)−→(x, (t + 1)·e − 1, (u + 1)·e− − 1) (7.65)
Example 2. For a first order ordinary differential equation y 0 = F (x, y) we

have
∂ ∂
V = ξ1 +φ (7.66)
∂x ∂y
or in an alternative notation
∂ ∂
Γ=X +Y (7.67)
∂x ∂y
26
We also have to consider the first prolongation
∂
P r(1) V = V + φx where φx = Dx (φ) − Dx (ξ 1 )y 0 (7.68)
∂y 0
or in another notation
∂
Γ(1) V = V + Y [x] where Y [x] = Dx (Y ) − Dx (X)y 0 (7.69)
∂y 0
Invariance of the differential equation y 0 = F (x, y) is being expressed by the
condition
Γ(1) V (∆) = 0 where ∆≡y 0 − F (x, y) (7.70)

∆=0
Chapter II
Ordinary Differential
Equations
8 First Order
Ordinary Differential Equations
8.1 The Lie Invariance Condition
In this section we shall find the invariance condition which determines the
infinitesimals X and Y of an ordinary differential equation (ODE) of the first
order. We consider the transformation
x̂ = x + X(x, y) + O(2 ), ŷ = y + Y (x, y) + O(2 ) (8.1)
and require invariance of the equation

dy
= F (x, y) (8.2)
dx
27
under (8.1), i.e.
dŷ
= F (x̂, ŷ) (8.3)
dx̂
Lemma 1. Considering the transformation (8.1), we have
dŷ dy
= + (Yx + Yy y 0 ) + O(2 ) (8.4)
dx dx
and
dx̂
= 1 + (Xx + Xy y 0 ) + O(2 ) (8.5)
dx
Proof. Under the transformation (8.1), we have
dŷ d
= (y + Y (x, y) + O(2 ))
dx dx
dy d
= + Y (x, y) + O(2 )
dx dx
dy ∂Y ∂Y dy
= + + + O(2 )
dx ∂x ∂y dx
dy
= + (Yx + Yy y 0 ) + O(2 )
dx
and
dx̂ d
= (x + X(x, y) + O(2 ))
dx dx
d d
= (x) + X(x, y) + O(2 )
dx dx
∂X ∂X dy
=1+ + + O(2 )
∂x ∂y dx
= 1 + (Xx + Xy y 0 ) + O(2 )
Lemma 2. Under the transformation (8.1), we have
F (x̂, ŷ) = F (x, y) + (XFx + Y Fy ) + O(2 ) (8.6)
Proof. Expanding in Taylor series, we have

F (x̂, ŷ) = F (x + X(x, y) + O(2 ), y + Y (x, y) + O(2 ))
= F (x, y) + (XFx + Y Fy ) + O(2 )
28
Lemma 3. Under the transformation (8.1), we have
dŷ dy
= + Yx + (Yy − Xx )y 0 − Xy (y 0 )2 + O(2 ) (8.7)
dx̂ dx
Proof. Under the transformation (8.1), and taking into account Lemma 1,
we obtain
dŷ d
dŷ (y + Y (x, y) + O(2 ))
= dx = dx
dx̂ dx̂ d
(x + X(x, y) + O(2 ))
dx dx
dy
+ (Yx + Yy y 0 ) + O(2 ))
= dx
1 + (Xx + Xy y 0 ) + O(2 )
dy
= + (Yx + Yy y 0 ) + O(2 )) 1 − (Xx + Xy y 0 ) + O(2 )
dx
dy
= + Yx + (Yy − Xx )y 0 − Xy (y 0 )2 + O(2 )
dx
dy
Theorem. Invariance of the differential equation = F (x, y) under the
dx
transformation (8.1), leads to the Invariance Condition
Yx + (Yy − Xx )F − Xy F 2 = XFx + Y Fy (8.8)
dy
Proof. Invariance of the differential equation = F (x, y) under the trans-
dx
dŷ
formation (8.1), means that = F (x̂, ŷ). Using (8.7) and (8.6) we see that
dx̂
invariance of the equation is equivalent to
dy
+ Yx + (Yy − Xx )y 0 − Xy (y 0 )2 + O(2 ) =
dx
= F (x, y) + (XFx + Y Fy ) + O(2 )
dy
Since = F (x, y), we obtain from the above equation, to order O(2 ), the
dx
condition
Yx + (Yy − Xx )F − Xy F 2 = XFx + Y Fy
29
Equation (8.8) is also called Lie’s Invariance Condition. From this equa-
tion we determine the infinitesimals X and Y of the differential equation
(8.2). However, we usually consider a solution ansatz in (8.8), since we have
to find two functions out of a single equation.
8.1a. First Ansatz. We first consider the ansatz
Y (x, y) = X(x, y)·F + θ(x, y) (8.9)
This choice will convert the invariance condition (8.8) into a PDE with only
one unknown function, the function θ(x, y). In fact, we have upon differen-
tiation of (8.9) with respect to x and y respectively
∂Y ∂X ∂F ∂θ
Yx ≡ = F +X + (8.10)
∂x ∂x ∂x ∂x
and
∂Y ∂X ∂F ∂θ
Yy ≡ = F +X + (8.11)
∂y ∂y ∂y ∂y
Equation (8.8), upon substitution of Yx and Yy taking into account expres-

sions (8.10) and (8.11) respectively, takes on the form
∂X ∂F ∂θ ∂X ∂F ∂θ
F +X + + F +X + − Xx F − X y F 2
∂x ∂x ∂x ∂y ∂y ∂y
= XFx + Y Fy
The above equation is equivalent to (after canceling the opposite terms)
∂θ ∂F ∂θ ∂F
+ XF + F =Y (8.12)
∂x ∂y ∂y ∂y
Since, as it follows from (8.9) XF = Y − θ, we obtain from equation (8.12)
∂θ ∂F ∂θ ∂F
+ (Y − θ) + F =Y
∂x ∂y ∂y ∂y
The previous equation is equivalent to
∂θ ∂θ ∂F
+ F =θ (8.13)
∂x ∂y ∂y
30
We thus have obtained a PDE with only one unknown function θ = θ(x, y).
This function can be determined by finding two independent solutions of the
auxiliary system
dx dy dθ
= = (8.14)
1 F ∂F
θ
∂y
Once we determine the function θ(x, y), we can find X(x, y) or Y (x, y) (or
both) considering equation (8.9).
Example. Find the generator Γ
∂ ∂
Γ=X +Y (8.15)
∂x ∂y
of symmetries of the differential equation
dy y + 2x
=− (8.16)
dx x
Solution. Since
y + 2x ∂F 1
F =− and =− (8.17)
x ∂x x
equation (8.13) determining θ becomes
∂θ y + 2x ∂θ 1
+ − = − θ (8.18)
∂x x ∂y x
The auxiliary system corresponding to the above PDE is
dx dy dθ
= =
1 y + 2x θ
− −
x x
which can be written as
dx xdy xdθ
=− =− (8.19)
1 y + 2x θ
From the first and last ratios of (8.19) we obtain the equation
dx dθ
=− (8.20)
x θ
31
which upon integration gives ln(x) = − ln θ + A from which we obtain
x·θ = C1 (8.21)
From the first and second ratios of (8.19) we obtain

x
dx = − dy
y + 2x
dy y
=− −2 (8.22)
dx x
The above is a homogeneous first order differential equation and can be solved
by the substitution
y
u= (8.23)
x
where u = u(x) is a new function. Since y = xu and
dy du
=u+x
dx dx
equation (8.22) becomes
du du du dx
u+x = −u − 2 ⇔ x = −2(u + 1) ⇔ = −2
dx dx u+1 x
Integrating the last equation we get
ln(u + 1) = −2 ln(x) + B
u + 1 = C2 x−2 (8.24)
Going back to the original variables, we have

y
+ 1 = C2 x−2
x
or
xy + x2 = C2 (8.25)
32
We thus see that the general solution of the auxiliary system is given by
F (C1 , C2 ) = 0, i.e. F (x·θ, xy + x2 ) = 0 from which we get x·θ = f (xy + x2 )
and then
1
θ= f (xy + x2 ) (8.26)
x
where f is an arbitrary function. We can take
1
θ= (8.27)
x
The general ansatz (8.9) becomes
y + 2x 1
Y (x, y) = X(x, y)· − + (8.28)
x x
Considering the choice Y (x, y) = 0, we obtain
1
X(x, y) = (8.29)
y + 2x
Therefore
1 ∂
Γ= (8.30)
y + 2x ∂x
8.1b. Second Ansatz. In this case we consider the choice
X(x, y) = 0 (8.31)
and thus the Invariance Condition becomes
Yx + Yy F = Y Fy (8.32)
Example. Determine the symmetries of the differential equation

dy tan(x − y)
=1+ (8.33)
dx x
Solution. We shall consider the case X = 0 and thus we obtain from
equation (8.32)
∂Y tan(x − y) ∂Y 1
+ 1+ = − Y (8.34)
∂x x ∂y x·cos2 (x − y)
33
It is proved that a solution of the above equation is given by
1
Y = (8.35)
x·cos(x − y)
Therefore
1 ∂
Γ= (8.36)
x·cos(x − y) ∂y
8.1c. Third Ansatz. In this case we consider the choice
Y (x, y) = 0 (8.37)
and thus the Invariance Condition becomes
Xx F − Xy F 2 = XFx (8.38)
8.1d. Fourth Ansatz. In this case we consider the choice
X(x, y) = P (x) and Y (x, y) = x·Q(y) (8.39)
and the Invariance Condition becomes
Q(y) + {x·Q0 (y) − P 0 (x)}F = P (x)·Fx + xQ(y)·Fy (8.40)
Example. Solution of a Riccati equation. Find the infinitesimals of the

equation
dy 1 1
= y2 − y − 2 (8.41)
dx x x
and then solve the equation using canonical coordinates.
Solution. Lie’s invariance condition (8.8) becomes,
1 1 1 1 2
Yx + (Yy − Xx ) y 2 − y − 2 − Xy y 2 − y − 2 =
x x x x (8.42)
y 2 1
= X 2 + 3 + Y 2y −
x x x
taking into account that
1 1
F = y2 − y − 2
x x
34
Considering the choice X = x, we get from (8.42)
1 1
Yx + (Yy − 1) y 2 − y − 2 =
x x
y 2 1
= x 2 + 3 + Y 2y −
x x x

2 1 1 2 3 1
Yx + y − y − 2 Yy = y + 2 + 2y − Y (8.43)
x x x x
The obvious choice Y = −y satisfies the above DE. We thus see that two
infinitesimals are
X = x, Y = −y (8.44)
It is obvious that there is an infinity of other choices.
In order to find the solution, we first determine the canonical coordinates
(r, s) by solving the system of PDEs (the system of equations (8.95) for
X = x, Y = −y)
∂r ∂r
·x − ·y = 0
∂x ∂y
(8.45)
∂s ∂s
·x − ·y = 1
∂x ∂y
One solution of the above system is (see Note 1 at the end of this Section
for details)
x = es , y = r·e−s (8.46)
Under the above transformation, we have
ds
dy e−s (dr − rds) 1−r
= = e−2s dr (8.47)
dx es ds ds
dr
Therefore the original equation (8.41) becomes
ds
1−r
e−2s dr = r2 e−2s − re−2s − e−2s (8.48)
ds
dr
35
which can be written as (after canceling the common factor e−2s )
ds
1−r
dr = r2 − r − 1 (8.49)
ds
dr
from which we derive
ds 1
= 2 (8.50)
dr r −1
The above is a separable differential equation with general solution
s = −arctanh(r) + C (8.51)
Solving with respect to r, we get
r = tanh(−s + C) (8.52)
Since
r
y= and s = ln(x) (8.53)
x
we get the following general solution of equation (8.41):
tanh(− ln(x) + C)
y= (8.54)
x
Note 1. The system (8.45) can be solved as follows. We consider the La-
grange equations
dx dy dr
= = (8.55)
x −y 0
and
dx dy ds
= = (8.56)
x −y 1
corresponding to the system (8.45).
Let us consider the first equation of (8.55). We find by integration ln(x) =
− ln(y) + a or ln(xy) = a and then xy = ea ≡C1 . From the last ratio we get
36
r = C2 . Therefore the general solution of (8.55) is given by F (r, xy) = 0
where F is an arbitrary function. Therefore we can consider r = f (xy). One
simple choice is r = xy.
We next turn to (8.56). From the first equation we get xy = C1 . The
second equation gives by integration − ln(y) = s + b or y = e−s C2 and then
yes = C2 . Therefore the general solution of (8.56) is G(yes , xy) = 0 where
G is an arbitrary function. Therefore we can take yes = g(xy). One simple
choice is yes = xy, i.e. x = es .
From the two solutions, i.e. r = xy and x = es , we obtain x = es and
y = re−s which is the solution (8.46) we have considered.
Note 2.The Riccati differential equation can be solved by transforming it
to a second order linear differential equation which can be solved in general
using a variety of methods. In fact, considering the general for of Riccati’s
equation
dy
= P (x) + Q(x)y + R(x)y 2 (8.57)
dx
the substitution
1 u0
y=− (8.58)
Ru
converts the equation into
d2 u R0 du
− Q + + P Ru = 0 (8.59)
dx2 R dx
It would be an interesting exercise to check if the solution found using Lie
symmetries coincides to the solution corresponding to that of the second
order differential equation.
Note 3. The infinitesimals X = x, Y = −y can also be found by exploiting
the scale invariance of the equation. In fact, using the transformation
ŷ = λy, x̂ = µx (8.60)
we find
dŷ λ dy
= (8.61)
dx̂ µ dx
and
1 1 λy 1 1
ŷ 2 − ŷ − 2 = λ2 y 2 − − 2 2 (8.62)
x̂ x̂ µx µ x
37
Therefore equation
dŷ 1 1
= ŷ 2 − ŷ − 2 (8.63)
dx̂ x̂ x̂
takes on the form
λ dy λy 1 1
= λ2 y 2 − − 2 2
µ dx µx µ x
from which we obtain
dy 1 1 1
= (λµ)y 2 − y − (8.64)
dx x (λµ) x2
Scale invariance implies that λµ = 1. Setting
λ = e− , µ = e (8.65)
we find upon Taylor expansion
x̂ = e x = x + x + O(2 ), ŷ = e− y = y + (−y) + O(2 ) (8.66)
From the above we can read off the infinitesimals: X = x and Y = −y.
Notice that we could equally well obtain X = −x and Y = y, by reversing
the values of λ and µ.
We should also mention that not all equations are scale invariant and if they
are, then scale invariance is not the only symmetry. We shall find below an
algorithmic procedure for revealing all (or almost all) the symmetries of a
given differential equation.
8.2 The Symmetry Generator.

The Extended Operator
We define the infinitesimal operator Γ by
∂ ∂
Γ=X +Y (8.67)
∂x ∂y
Since
F (x̂, ŷ) = F (x + X(x, y) + O(2 ), y + Y (x, y) + O(2 ))
(8.68)
= F (x, y) + (XFx + Y Fy ) + O(2 )
38
using the operator Γ we get
F (x̂, ŷ) = F (x, y) + ΓF + O(2 ) (8.69)
where we have obviously

∂
∂
ΓF = X +Y F = XFx + Y Fy (8.70)
∂x ∂y
The invariance condition for the differential equation y 0 = F (x, y) reads
(see equation (8.8))
Yx + (Yy − Xx )y 0 − Xy (y 0 )2 = XFx + Y Fy (8.71)
We now introduce the extended operator P r(1) Γ by

∂
P r(1) Γ = Γ + Y [x] (8.72)
∂y 0
i.e.
∂ ∂ ∂
P r(1) Γ = X +Y + Y [x] 0 (8.73)
∂x ∂y ∂y
and define ∆ by
∆ = y 0 − F (x, y) (8.74)
We then have
P r(1) Γ(∆) = 0
∂ ∂ ∂
⇔ P r(1) Γ(∆) = X +Y + Y [x] 0 (y 0 − F ) = 0 (8.75)
∂x ∂y ∂y
(1) [x]
⇔ P r Γ(∆) = −XFx − Y Fy + Y = 0
Comparing the above expression to the invariance condition (8.71), we find
Y [x] = Yx + (Yy − Xx )y 0 − Xy (y 0 )2 (8.76)
When we substitute y 0 = F (x, y) into (8.71), i.e. into
P r(1) Γ(∆) = 0
39
we get Lie’s invariance condition (8.75). We have thus obtained that
P r(1) Γ(∆)|∆=0 = 0 (8.77)
is a compact expression of the Invariance Condition. The operator Γ is

thus the generator of symmetries of the differential equation ∆ = 0.
We have further
Y [x] = Yx + (Yy − Xx )y 0 − Xy (y 0 )2 = Yx + Yy y 0 − (Xx + Xy y 0 )y 0 (8.78)
If we define the total differential operator by

∂ ∂
Dx = + y0 (8.79)
∂x ∂y
then Y [x] , as given by (8.76), is being expressed by
Y [x] = Dx (Y ) − Dx (X)y 0 (8.80)
Y [x] is called the extended infinitesimal.

Lemma 1. We have the following expressions
dY dX
= Dx (Y ) and = Dx (X) (8.81)
dx dx
Proof. Since
∂Y ∂Y
dY = dx + dy (8.82)
∂x ∂y
we have
dY ∂Y ∂Y dy
= + (8.83)
dx ∂x ∂y dx
or
dY
= Yx + Yy ·y 0 (8.84)
dx
dY
= Dx (Y ) (8.85)
dx
40
Similar proof for the second relation.
Theorem 1. We have the following expression
dŷ dy
= + Y [x] + O(2 ) (8.86)
dx̂ dx
Proof.
dŷ d dy
dŷ (y + Y (x, y) + O(2 )) + ·Dx (Y ) + O(2 )
= dx = dx = dx
dx̂ dx̂ d 2 1 + ·Dx (X) + O(2 )
(x + X(x, y) + O( ))
dx dx
dy
= + ·Dx (Y ) + O(2 ) 1 − ·Dx (X) + O(2 ) (8.87)
dx
dy
= + Dx (Y ) − Dx (X)y 0 + O(2 )
dx
dy
= + Y [x] + O(2 )
dx
The importance of this Theorem lies to the fact that it is suggestive of some
similar relations to higher order equations:
d2 ŷ d2 y
= + Y [xx] + O(2 ) (8.88)
dx̂2 dx2
d3 ŷ d3 y
= + Y [xxx] + O(2 ) (8.89)
dx̂3 dx3
with similar to Y [x] definitions for Y [xx] , Y [xxx] , etc.
8.2. The Canonical Coordinates. In order to simplify the solution pro-
cedure of any differential equation or system of differential equations (after
determining the symmetries of equation), we introduce a special system of
coordinates, called canonical coordinates. The introduction of the canon-
ical coordinates is based on the following Lemma:
Lemma. The equation
ds
= F (r) (8.90)
dr
is invariant under the transformation
r̂ = r, ŝ = s + (8.91)
41
Using the above Lemma, we can determine a new system of coordinates,
the canonical coordinates which transform a given DE into a DE with
separable variables (once we know the infinitesimals X and Y ).
In fact, we consider the transformation
r = r(x, y), s = s(x, y) (8.92)
and require (8.92) to be invariant, that is
r̂ = r(x̂, ŷ), ŝ = s(x̂, ŷ) (8.93)
Differentiation of the above relations with respect to , we obtain

∂r̂ ∂r ∂ x̂ ∂r ∂ ŷ
= · + ·
∂ ∂ x̂ ∂ ∂ ŷ ∂
(8.94)
∂ŝ ∂s ∂ x̂ ∂s ∂ ŷ
= · + ·
∂ ∂ x̂ ∂ ∂ ŷ ∂
Since, because of (8.91),
∂r̂ ∂ŝ
= 0 and =1
∂ ∂
and after setting = 0 in (8.94), we obtain
∂r ∂r
0= ·X(x, y) + ·Y (x, y)
∂x ∂y
∂s ∂s
1= ·X(x, y) + ·Y (x, y)
∂x ∂y
From the above system of PDEs we obtain r(x, y) and s(x, y). Writing the
system in the standard form
∂r ∂r
·X(x, y) + ·Y (x, y) = 0
∂x ∂y
(8.95)
∂s ∂s
·X(x, y) + ·Y (x, y) = 1
∂x ∂y
we can solve it using either Lagrange’s equations
dx dy dr dx dy ds
= = and = =
X Y 0 X Y 1
42
or the method of characteristics.
We are now going to derive the same system of equations (8.95) using a
different approach. We consider the operator Γ given by (8.67)
∂ ∂
Γ=X +Y
∂x ∂y
We then transform the partial derivatives considering the transformation
(x, y) → (r, s). We have
∂ ∂r ∂ ∂s ∂
= +
∂x ∂x ∂r ∂x ∂s
and
∂ ∂r ∂ ∂s ∂
= +
∂y ∂y ∂r ∂y ∂s
Therefore the operator Γ becomes
∂ ∂
Γ=X +Y
∂x ∂y
∂r ∂ ∂s ∂ ∂r ∂ ∂s ∂
=X + +Y +
∂x ∂r ∂x ∂s ∂y ∂r ∂y ∂s
∂ ∂ ∂ ∂
= X rx + sx + Y ry + sy
∂r ∂s ∂r ∂s
∂ ∂ ∂ ∂
Γ=X +Y = (Xrx + Y ry ) + (Xsx + Y sy ) (8.96)
∂x ∂y ∂r ∂s
Since we want to obtain an equation with separated variables, we require
operator Γ to be independent of s, i.e. to have the form
∂
Γ=
∂s
Therefore we should have
Xrx + Y ry = 0 and Xsx + Y sy = 1 (8.97)
which is the system of equations (8.95).
The above system can also be written as
{Γr = 0, Γs = 1} (8.98)
43
where Γ is the symmetry operator (8.67).
8.3. Another Method of Solution of First Order ODEs. We consider
in this subsection another method of solution of the equation y 0 = F (x, y)
based on the following two Lemmata:
Lemma 2. If φ is a solution of the differential equation y 0 = F (x, y) then
∂ ∂
Aφ≡ +F φ=0 (8.99)
∂x ∂y
Proof. A first integral of the given equation is a function φ = φ(x, y, y 0 )

that is constant along the solutions of y 0 = F (x, y), which means that the
expression
dφ ∂φ ∂φ dy ∂φ ∂φ
= + = + y0 (8.100)
dx ∂x ∂y dx ∂x ∂y
is zero if y 0 is substituted by F . Therefore introducing the operator A by
∂ ∂
A= +F (8.101)
∂x ∂y
we see that if φ is a solution of the equation y 0 = F (x, y) then Aφ = 0, i.e.
∂φ ∂φ
+F =0 (8.102)
∂x ∂y
Lemma 3. If φ is a solution of y 0 = F (x, y), then
Γφ≡Xφx + Y φy = 1 (8.103)
where
∂ ∂
Γ=X +Y (8.104)
∂x ∂y
Proof. Since Γ is a symmetry, this symmetry maps solutions to other so-
lutions and since all the solutions of Af = 0 are functions of ψ, we have
Γψ = F (ψ). Another solution φ considered as a function of ψ, i.e. φ = G(ψ),
should also satisfy the equation. But in this case
dG dG
Xφx + Y φy = (Xψx + Y ψy ) = H(ψ) (8.105)
dψ dψ
44
Since we can choose G whatever we like, we can for example take
Z
dψ
G(ψ) = (8.106)
H(ψ)
But in this case the right hand side of (8.105) becomes equal to 1, i.e.
Xφx + Y φy = 1 (8.107)
Collecting the above two Lemmata, we arrive at the following Theorem:

Theorem 2. The differential equation y 0 = F (x, y) admits a general solution
which is the solution of the system
Γφ = Xφx + Y φy = 1
(8.108)
Aφ = φx + F φy = 0
Let us now invent a method for solving the differential equation y 0 = F (x, y).
Solving the above system, we find
F 1
φx = − and φy = (8.109)
Y − XF Y − XF
Since
F 1
dφ = φx dx + φy dy = − dx + dy
Y − XF Y − XF
(8.110)
dy − F dx
=
Y − XF
we see that the solution φ is determined by the line integral
dy − F (x, y)dx
Z
φ(x, y) = (8.111)
Y (x, y) − X(x, y)F (x, y)
This is a general solution of the differential equation Aφ = 0 and therefore
a first integral of the differential equation y 0 = F (x, y). We thus have the
following
Theorem 3. If the differential equation y 0 = F (x, y) admits a symmetry
with generator
∂ ∂
Γ=X +Y (8.112)
∂x ∂y
45
then its solution is given by the line integral
dy − F (x, y)dx
Z
φ(x, y) = = φ0 = constant (8.113)
Y (x, y) − X(x, y)F (x, y)
Example. Solve the differential equation
dy tan(x − y)
=1+ (8.114)
dx x
Solution. We have already found that the generator of symmetries of the
given equation is
1 ∂
Γ= (8.115)
x·cos(x − y) ∂y
and thus
1
X = 0 and Y = (8.116)
x·cos(x − y)
Since
tan(x − y)
F =1+ (8.117)
x
the general solution formula (8.113) yields
Z dy − 1 + tan(x − y) dx

φ(x, y) = x
1
x·cos(x − y)
The above formula is equivalent to
Z
φ(x, y) = [x·cos(x − y)]dy − [x·cos(x − y) + sin(x − y)]dx
(8.118)
= −x·sin(x − y)
and from φ = φ0 = constant we have −x·sin(x − y) = φ0 and thus the

general solution of the differential equation is given by
φ0
y = x + arcsin (8.119)
x
46
8.3a. Integrating Factor. The integrating factor is a function M (x, y)
dy
that makes the differential equation = F (x, y), i.e. dy − F (x, y)dx = 0
dx
the differential of a function φ, i.e.
dφ = M (dy − F dx) (8.120)
Since, because of (8.110)
dy − F dx
dφ = (8.121)
Y − XF
upon comparing the above two relations we see that the integrating factor
M is given by
1
M= (8.122)
Y − XF
On the other hand, with
dφ = −M F dx + M dy (8.123)
the integrability condition has the form
(−M F )y = Mx (8.124)
Let us now write in full detail the above condition expressed in terms of X,
Y and F . Since (using (8.122))
F (Yy − Xy F − XFy ) Fy
(M F )y = − 2
+ (8.125)
(Y − XF ) Y − XF
and
Yx − Xx F − XFx
Mx = − (8.126)
(Y − XF )2
we obtain that the integrability condition is equivalent to
(M F )y + Mx = 0
F (Yy − Xy F − XFy ) Fy Yx − Xx F − XFx
⇔− 2
+ − =0
(Y − XF ) Y − XF (Y − XF )2
(8.127)
⇔ − F (Yy − Xy F − XFy ) + Fy (Y − XF )
− (Yx − Xx F − XFx ) = 0
⇔ Yx + (Yy − Xx )F − Xy F 2 = XFx + Y Fy
47
We thus see that the integrability condition is equivalent to Lie’s Invariance
Condition (8.8).
Example. Solve the equation
(2x4 y + y 4 )dx + (x5 − 2xy 3 )dy = 0 (8.128)
using an appropriate integrating factor.

Solution. First of all we shall find a tansformation law which leaves invariant
the equation. Under the transformation
x = eλ x̂, y = eµ ŷ (8.129)
the given equation transforms into
[2e(5λ+µ) (x̂)4 ŷ + e(λ+4µ) (ŷ)4 ]dx̂

(8.130)
+ [e(5λ+µ) (x̂)4 − 2e(λ+4µ) x̂(ŷ)3 ]dŷ = 0
Therefore we obtain invariance if the exponents become equal (because in

this case we have a cancellation of exponential factors)
5λ + µ = λ + 4µ ⇔ 4λ = 3µ (8.131)
The last equation is satisfied for the simple choice λ = 3 and µ = 4. Therefore
the transformation becomes
x = e3 x̂, y = e4 ŷ (8.132)
or by inversion
x̂ = e−3 x, ŷ = e−4 y (8.133)
Upon expansion in Taylor series, we get
x̂ = x + (−3x) + . . ., ŷ = y + (−4y) + . . . (8.134)
and thus the infinitesimal coefficients X and Y are
X = −3x, Y = −4y or even X = 3x, Y = 4y (8.135)
The original equation can be written as

dy 2x4 y + y 4
= F (x, y) with F = − 5 (8.136)
dx x − 2xy 3
48
The integrating factor is thus given by
1 1 x4 − 2y 3
M= = = (8.137)
Y − XF 2x4 y + y 4 5y(2x4 − y 3 )
4y + 3x 5
x − 2xy 3
Therefore if φ is a solution of the equation, then (we omit the factor 5 in the
denominator)
x4 − 2y 3 2x4 y + y 4
dφ = M (dy − F dx) = dy + dx
y(2x4 − y 3 ) x5 − 2xy 3
or
2x4 + y 3 x4 − 2y 3
dφ = dx + dy (8.138)
x(2x4 − y 3 ) y(2x4 − y 3 )
Since with
2x4 + y 3 x4 − 2y 3
P = , Q= (8.139)
x(2x4 − y 3 ) y(2x4 − y 3 )
we have
12x3 y 2
Py = Qx = (8.140)
(2x4 − y 3 )2
and thus we have an exact differential equation. Since φ is supposed to be a
solution, we should have
∂φ 2x4 + y 3 ∂φ x4 − 2y 3
= and = (8.141)
∂x x(2x4 − y 3 ) ∂y y(2x4 − y 3 )
Integrating with respect to x the first of the above equations, we obtain
1
φ = − ln(x) + ln(2x4 − y 3 ) + f (y) (8.142)
2
Differentiation of the above relation with respect to y and comparing to the
second of (8.141) we get
3y 2 0 x4 − 2y 3
− + f (y) =
2(2x4 − y 3 ) y(2x4 − y 3 )
49
from which we have
1
f 0 (y) =
2y
Integration of the previous equation with respect to y yields
1
f (y) = ln(y) + C (8.143)
2
We thus have the following expression for the solution φ:
1 1
φ = − ln(x) + ln(2x4 − y 3 ) + ln(y) + C (8.144)
2 2
and equating the above function to φ0 we have finally the general solution
y (2x4 − y 3 ) = Kx2 (8.145)
where K is an arbitrary constant.

8.4. Using Software. We are going to use two known Computer Algebra
Systems, MAPLE and Mathematica, to solve differential equations using Lie
symmetry methods. This is a semi-automatic procedure in solving differen-
tial equations. We prefer this step-by-step programming in order to keep
track of the details of revealing the underlying symmetries which will lead us
eventually to the solution method. We should stress the fact that this way
of exploring solutions to Differential Equations might also be appropriate for
experimentation. The reader is supposed to be familiar to some elementary
features of both MAPLE and Mathematica.
8.4.1 Using MAPLE. The calculations are performed within MAPLE it-
self, no extra software is needeed. It only suffices to call the DEtools and
PDEtools within MAPLE. Using MAPLE’s command > symgen, we can
determine the symmetries for first order ODEs. For higher order ODEs we
may use some alternative commands. More examples appear in Appendix C.
Example 1. Solve the differential equation
dy 3y x5
= + (8.146)
dx x 2y + x3
Solution. We use the following steps:
Step 1. We call the DEtools and PDEtools within MAPLE.
>restart;
50
>with(DEtools):
>with(PDEtools):
Step 2. We write down the ODE we consider.
3 ∗ y(x) x5
> ode := diff(y(x), x) = + ; (8.147)
x 2 ∗ y(x) + x3
The program responds
d 3 y(x) x5
y(x) = + (8.148)
dx x 2 y(x) + x3
Step 3. We find the symmetries (the infinitesimals X and Y ) using

>symgen(ode, HINT=[f (x), g(y)]);
[− ξ = x, − η = 3y] (8.149)
That means in our notation X = x and Y = 3y.
Step 4. We now determine the canonical transformation using the system
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
We thus consider the two equations
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ 3 ∗ y = 0 : (8.150)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ 3 ∗ y = 1 : (8.151)
The solution follows using the command
> pdsolve({eq1, eq2}, {r(x, y), s(x, y)}); (8.152)

n y y o
r(x, y) =− F 1 3 , s(x, y) = ln(x) +− F 2 3 (8.153)
x x
The reader easily recognizes that − F 1 and − F 2 are arbitrary functions. We
take
y
r = 3 s = ln(x) (8.154)
x
51
Step 5. Inverting the last transformation, we enter the command
> itr := {x = exp(s(r)), y(x) = r ∗ exp(3 ∗ s(r))}; (8.155)
The system responds
x = es(r) , y(x) = r e3 s(r)

(8.156)
Step 6. We make change of variables (from x, y(x) to r, s(r)) to the original

ode using the command (this single command is a substitute of Appendix A)
> dchange(itr, ode, simplif y); (8.157)
The system responds

d
e2s(r) 1 + 3r s(r) e2s(r) (6r2 + 3r + 1)
dr = (8.158)
d 2r + 1
s(r)
dr
d
Step 7. From the above equation, solving with respect to s(r) we consider
ds
the equation
> eqn := diff(s(r), r) = 2 ∗ r + 1 (8.159)
the system responds

d
s(r) = 2 r + 1 (8.160)
ds
Solving the above equation
> dsolve(eqn); (8.161)
the system responds
s(r) = r2 + r +− C1 (8.162)
Step 8. Using the initial variables given by (8.154), we transform (8.162) into
y 2 y
ln(x) = 3
+ 3 +C (8.163)
x x
52
which is the general solution of the original differential equation. This equa-
tion can also be written as
x6 · ln(x) = y 2 + x3 y + Cx6 (8.164)
which is the general solution of the equation (8.146) in implicit form. Ex-
ample 2. Solve the differential equation
dy y + y3
= (8.165)
dx x + (x + 1)y 2

>restart;
>with(DEtools):
>with(PDEtools):
y(x) + y(x)3
> ode := diff(y(x), x) = ; (8.166)
x + (x + 1) ∗ y(x)2
d y(x) + y(x)3
y(x) = (8.167)
dx x + (x + 1) y(x)2

>symgen(ode, HINT=[f (y), g(x)]);
[− ξ = y, − η = 0] (8.168)
That means in our notation X = y and Y = 0.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ y + diff(r(x, y), y) ∗ 0 = 0 : (8.169)
53
> eq2 := diff(s(x, y), x) ∗ y + diff(s(x, y), y) ∗ 0 = 1 : (8.170)

x
r(x, y) =− F 1(y), s(x, y) = +− F 2(y) (8.172)
y

take
x
r=y s= (8.173)
y
> itr := {x = r ∗ s(r), y(x) = r}; (8.174)
The system responds
{x = r s(r), y(x) = r} (8.175)

The system responds
1 1 + r2
d = (8.177)
s(r) + r2 s(r) + r

s(r) + r s(r)
dr
d
Step 7. From the above equation, solving with respect to s(r) we consider
ds
the equation
1
> eqn := diff(s(r), r) = (8.178)
1 + r2
54
the system responds
d 1
s(r) = (8.179)
ds 1 + r2
the program responds
s(r) = arctan(r) +− C1 (8.181)
Step 8. Using the initial variables given by (8.173), we transform (8.181) into
x
= arctan(y) + C (8.182)
y
x = y·arctan(y) + C·y (8.183)
8.5. Table of First Order Differential Equations admitting known

Symmetries. We list below some first-order differential equations with their
corresponding symmetry generators. The reader is advised to check, if not
all, at least some of them.
No. Differential Equation Symmetry Generator

∂
1 y 0 = F (y) X=
∂x
∂
2 y 0 = F (x) X=
∂y
∂ ∂
3 y 0 = F (αx + βy) X=β −α
∂x ∂y
p
y + x F ( x2 + y 2 ) ∂ ∂
4 y0 = p X=y −x
x − y F ( x2 + y 2 ) ∂x ∂y
55
y ∂ ∂
5 y0 = F X=x +y
x ∂x ∂y
y ∂ ∂
6 y 0 = xm−1 F m X=x + my
x ∂x ∂y
∂ ∂
7 xy 0 = F (xe−y ) X=x +
∂x ∂y
∂ ∂
8 y 0 = yF (ye−x ) X= +y
∂x ∂y
y y ∂ y ∂
9 y0 = + xF X= +
x x ∂x x ∂y
y ∂ ∂
10 xy 0 = y + F X = x2 + xy
x ∂x ∂y
y ∂ ∂
11 y0 = y X = xy + y2
x+F ∂x ∂y
x
0 y ∂
12 y = X=y
x + F (y) ∂x
∂
13 xy 0 = y + F (x) X=x
∂y
y ∂
14 xy 0 = X = xy
ln(x) + F (y) ∂x
∂
15 xy 0 = y [ln(y) + F (x)] X = xy
∂y
R ∂
16 y 0 = P (x)y + Q(x) X = exp P (x)dx
∂y
∂
y 0 = P (x)y + Q(x)y n
R
17 X = y n exp (1 − n) P (x)dx
∂y
∂
18 y 0 = P (x)y X=y
∂y
56
9 Second Order Ordinary
In this Section we consider infinitesimals and invariance of second order or-
dinary differential equations.
9.1. Second-Order Infinitesimals.
Using the identity
d dŷ d2 ŷ dx̂
= 2· (9.1)
dx dx̂ dx̂ dx
we obtain
d dŷ
d2 ŷ
= dx dx̂ (9.2)
dx̂2 dx̂
dx
Therefore
d dŷ d dy [x] 2

d2 ŷ + Y + O( )
= dx dx̂ = dx dx
dx̂ 2 dx̂ d
x + X· + O(2 )
dx dx
2
dy
2
+ Dx (Y [x] ) + O(2 )
= dx = (9.3)
1 + ·Dx (X) + O(2 )
d2 y
[x] 2 2
= + D x (Y ) + O( ) 1 − ·D x (X) + O( )
dx2
d2 y
= 2 + · Dx (Y [x] ) − Dx (X)y 00 + O(2 )
dx
Comparing the above expression to
d2 ŷ d2 y
= + ·Y [xx] + O(2 ) (9.4)
dx̂2 dx2
we derive the following expression for Y [xx] :
Y [xx] = Dx (Y [x] ) − Dx (X)y 00 (9.5)
57
Since Y [x] contains x, y and y 0 , we have to extend Dx to
∂ ∂ ∂
Dx = + y0 + y 00 0 (9.6)
∂x ∂y ∂y
At this stage we can find the explicit expression of Y [xx] :
Y [xx] = Dx (Y [x] ) − y 00 Dx (X)

= Dx Yx + (Yy − Xx )y − Xy (y ) − y 00 Dx (X)
0 0 2
∂
0 ∂ 00 ∂

= +y +y [Yx + (Yy − Xx )y 0 − Xy (y 0 )2 ]
∂x ∂y ∂y 0
∂
0 ∂ 00 ∂

00
−y +y +y X
∂x ∂y ∂y 0
= Yxx + (Yyx − Xxx )y 0 − Xyx (y 0 )2
+ y 0 [Yxy + (Yyy − Xxy )y 0 − Xyy (y 0 )2 ]
+ y 00 [(Yy − Xx ) − 2Xy y 0 ] − y 00 (Xx + y 0 Xy )
Collecting terms and rearranging, we obtain from the above the following
explicit expression for Y [xx] :
Y [xx] = Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2
(9.7)
− Xyy (y 0 )3 + (Yy − 2Xx )y 00 − 3Xy y 0 y 00
9.2. Invariance of Second-Order Equations.
Let us now consider the invariance of second-order equations. We consider
the equation
d2 ŷ dŷ
= F x̂, ŷ, (9.8)
dx̂2 dx̂
The above equation can be written as
d2 y
2
+ ·Y [xx] + O(2 )
dx (9.9)
2 dy

2 [x] 2
= F x + X + O( ), y + Y + O( ) + Y + O( )
dx
and upon expanding in Taylor series the right hand side,
d2 y
+ ·Y [xx] + O(2 )
dx2 (9.10)
dy
= F x, y, + (XFx + Y Fy + Y [x] Fy0 ) + O(2 )
dx
58
For
d2 y dy
= F x, y, (9.11)
dx2 dx
we obtain from (9.10)
Y [xx] = XFx + Y Fy + Y [x] Fy0 (9.12)
The previous equation can be written as
Y [xx] = P r(1) ΓF (9.13)
We then define the second extension to Γ:

∂ ∂ ∂ ∂
P r(2) Γ = X +Y + Y [x] 0 + Y [xx] 00 (9.14)
∂x ∂y ∂y ∂y
After defining ∆ by
∆ = y 00 − F (x, y, y 0 ) (9.15)
the invariance condition becomes
P r(2) Γ(∆)|∆=0 = 0 (9.16)
9.3. Example. The modified Emden equation. Find the point sym-
metries of the differential equation
y 00 + 3yy 0 + y 3 = 0, y = y(x) (9.17)
and then determine the general solution(s).

Solution. This is a second order differential equation. We define ∆ by
∆ = y 00 + 3yy 0 + y 3 (9.18)
We thus consider the invariance condition

P r(2) Γ(∆) = 0
∂ ∂ [x] ∂ [xx] ∂

⇔ X +Y +Y 0
+Y 00
(y 00 + 3yy 0 + y 3 ) = 0 (9.19)
∂x ∂y ∂y ∂y
0 2 [x] [xx]
⇔ Y (3y + 3y ) + Y (3y) + Y =0
59
Upon substituting Y [x] and Y [xx] given by (8.76) and (9.7) respectively, we
obtain from (9.19)
3(y 0 + y 2 )Y + 3y{Yx + (Yy − Xx )y 0 − Xy (y 0 )2 }

+ Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3 (9.20)
+ (Yy − 2Xx )y 00 − 3Xy y 0 y 00 = 0
We now take into account the condition ∆ = 0, which means that we have
to substitute y 00 by −3yy 0 − y 3 in (9.20). We thus have
3(y 0 + y 2 )Y + 3y{Yx + (Yy − Xx )y 0 − Xy (y 0 )2 }

+ (Yy − 2Xx )(−3yy 0 − y 3 ) − 3Xy y 0 (−3yy 0 − y 3 ) = 0
The above equation can be written as (arranged in powers of y 0 , (y 0 )2 and

(y 0 )3 )
(3Y + 3yXx + 3y 3 Xy − Xxx + 2Yxy )y 0

+ (6yXy + Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3 (9.22)
+ Yxx + 3yYx + 3y 2 Y − y 3 (Yy − 2Xx ) = 0
We then equate to zero the coefficients of the different powers of the deriva-
tives y 0 , (y 0 )2 and (y 0 )3 and we obtain the following system of equations
3Y + 3yXx + 3y 3 Xy − Xxx + 2Yxy = 0 (9.23)
6yXy + Yyy − 2Xxy = 0 (9.24)

Xyy = 0 (9.25)
Yxx + 3yYx + 3y 2 Y − y 3 (Yy − 2Xx ) = 0 (9.26)
We are going to solve the system of equations (9.23)-(9.26).
From equation (9.25) we obtain
X = α(x)y + β(x) (9.27)
Using the above expression for X, we obtain from (9.24)
Yyy = 2α0 (x) − 6α(x)y (9.28)
60
Integrating the above equation twice with respect to y, we find
Y = −α(x)y 3 + α0 (x)y 2 + γ(x)y + δ(x) (9.29)
Equation (9.23), taking into account (9.27) and (9.29), takes on the form
3{α00 (x) + β 0 (x) + γ(x)}y + {2γ 0 (x) − β 00 (x) + 3δ(x)} = 0 (9.30)
From the above equation we obtain the following two equations
α00 (x) + β 0 (x) + γ(x) = 0 (9.31)
2γ 0 (x) − β 00 (x) + 3δ(x) = 0 (9.32)

Finally, equation (9.26) becomes
2{α00 (x) + β 0 (x) + γ(x)}y 3 + {α000 (x) + 3δ(x) + 3γ 0 (x)}y 2

(9.33)
+ {γ 00 (x) + 3δ 0 (x)}y + δ 00 (x) = 0
From the above equation we obtain the following four equations
α00 (x) + β 0 (x) + γ(x) = 0 (9.34)
α000 (x) + 3δ(x) + 3γ 0 (x) = 0 (9.35)

γ 00 (x) + 3δ 0 (x) = 0 (9.36)
δ 00 (x) = 0 (9.37)
From the last equation we obtain
δ(x) = a1 x + a2 (9.38)
Equation (9.36) then gives γ 00 (x) = −3a1 and integrating twice, we have
3
γ(x) = − a1 x2 + a3 x + a4 (9.39)
2
From (9.32), using the expressions (9.38) and (9.39), we obtain β 00 (x) =
−3a1 x + 2a3 + 3a2 and integrating twice we obtain
1 1
β(x) = − a1 x3 + (2a3 + 3a2 )x2 + a5 x + a6 (9.40)
2 2
61
From (9.34) we obtain α00 (x) = 3a1 x2 −3a3 x−3a2 x−a4 −a5 and after double
integration, we have
1 1 1
α(x) = a1 x4 − (a2 + a3 )x3 − (a4 + a5 )x2 + a7 x + a8 (9.41)
4 2 2
Equation (9.35) is being satisfied automatically for the above found expres-
sions of α(x), δ(x) and γ(x).
From (9.27), using the expressions found for α(x) and β(x) given by (9.41)
and (9.40) respectively, we have the following expression for X:
1 1 1
4 3 2
X= a1 x − (a2 + a3 )x − (a4 + a5 )x + a7 x + a8 y
4 2 2 (9.42)
1 1
3 2
+ − a1 x + (2a3 + 3a2 )x + a5 x + a6
2 2
Using (9.29) and the expressions for α(x), γ(x) and δ(x) given by (9.41),
(9.39) and (9.38) respectively, we get the following expression for Y :
1 1 1
Y = − a1 x − (a2 + a3 )x − (a4 + a5 )x + a7 x + a8 y 3
4 3 2
4 2 2
3
3
+ a1 x − (a2 + a3 )x − (a4 + a5 )x + a7 y 2
2
(9.43)
2
3
2
+ − a1 x + a3 x + a4 y + (a1 x + a2 )
2
Upon renaming the various constants according to
a1 = 4c5 , a2 = 4c4 + 2c8 , a3 = −6c4 − 2c8

(9.44)
a4 = −c7 − 2c3 , a5 = c7 , a6 = c6 , a7 = c2 , a8 = c1
we obtain the following expressions for X and Y :
X = (c1 + c2 x + c3 x2 + c4 x3 + c5 x4 )y − 2c5 x3
(9.45)
+ c8 x 2 + c7 x + c6
Y = −(c1 + c2 x + c3 x2 + c4 x3 + c5 x4 )y 3
+ (4c5 x3 + 3c4 x2 + 2c3 x + c2 )y 2
(9.46)
+ (−6c5 x2 + (−6c4 − 2c8 )x − c7 − 2c3 )y
+ 4c5 x + 4c4 + 2c8
62
We have further
∂ ∂
Γ=X +Y =
∂x ∂y

2 3 4 3 2 ∂
= (c1 + c2 x + c3 x + c4 x + c5 x )y − 2c5 x + c8 x + c7 x + c6
∂x

+ − (c1 + c2 x + c3 x2 + c4 x3 + c5 x4 )y 3
(9.47)
3 2 2
+ 4c5 x + 3c4 x + 2c3 x + c2 y

+ −6c5 x2 + (−6c4 − 2c8 )x − c7 − 2c3 y

∂
+ 4c5 x + 4c4 + 2c8
∂y
The above expression can be written as

∂ 3 ∂
Γ = c1 y −y
∂x ∂y

∂ 3 2 ∂
+ c2 xy + (−xy + y )
∂x ∂y

2 ∂ 2 3 2 ∂
+ c3 x y + (−x y + 2xy − 2y)
∂x ∂y
(9.48)
3 ∂ 3 3 2 2 ∂
+ c4 x y + (−x y + 3x y − 6xy + 4)
∂x ∂y

4 3 ∂ 4 3 3 2 2 ∂
+ c5 (x y − 2x ) + (−x y + 4x y − 6x y + 4x)
∂x ∂y

∂ ∂ ∂ 2 ∂ ∂
+ c6 + c7 x −y + c8 x + (−2xy + 2)
∂x ∂x ∂y ∂x ∂y
63
We thus see that the Lie algebra of symmetries for equation (9.17) is being
spanned by the eight generators
∂ ∂
Γ(1) = y − y3
∂x ∂y
∂ ∂
Γ(2) = xy + (−xy 3 + y 2 )
∂x ∂y
∂ ∂
Γ(3) = x2 y + (−x2 y 3 + 2xy 2 − 2y)
∂x ∂y
∂ ∂
Γ(4) = x3 y + (−x3 y 3 + 3x2 y 2 − 6xy + 4)
∂x ∂y
(9.49)
(5) ∂ ∂
Γ = (x4 y − 2x3 ) + (−x4 y 3 + 4x3 y 2 − 6x2 y + 4x)
∂x ∂y
∂
Γ(6) =
∂x
∂ ∂
Γ(7) =x −y
∂x ∂y
∂ ∂
Γ(8) = x2 + (−2xy + 2)
∂x ∂y
(i) We shall find next a general solution to the equation (9.17) using the
generator
∂ ∂
Γ(1) = y − y3 (9.50)
∂x ∂y
In this case the system (8.95)
∂r ∂r
·X(x, y) + ·Y (x, y) = 0
∂x ∂y
∂s ∂s
·X(x, y) + ·Y (x, y) = 1
∂x ∂y
or its equivalent {Γ1 r = 0, Γ1 s = 1} becomes
yrx − y 3 ry = 0, ysx − y 3 sy = 1
or after simplification
rx − y 2 ry = 0, ysx − y 3 sy = 1 (9.51)
64
and can be solved using Lagrange’s equations
dx dy dr dx dy ds
= 2
= and = 3
= (9.52)
1 −y 0 y −y 1
We obtain from the first system of (9.52) by integration
1
x= + C1 and r = C2 (9.53)
y
Therefore the general solution of the first system in (9.52) is
1 1
F x − , r = 0 or r = f x − (9.54)
y y
A simple choice is
1
r =x− (9.55)
y
From the second system of (9.52) we obtain by integration
1
x= + C1 (9.56)
y
Equation
dx ds
= (9.57)
y 1
since y = 1/(x − C1 ), becomes (x − C1 )dx = ds, and by integration
x2
− C 1 x = s + C2 (9.58)
2
and since C1 = x − 1/y, equation (9.58) takes on the form
x2 1 x2 x
− x− x = s + C2 or − + = s + C2 (9.59)
2 y 2 y
Therefore the general solution of the second system in (9.52) is given by
1 x2 x x2 x 1
G x − , − + − s = 0 or s = − + + g x − (9.60)
y 2 y 2 y y
65
A simple choice is
x2 x
s=− + (9.61)
2 y
The system of (9.55) and (9.61),
1 x2 x
r =x− , s=− + (9.62)
y 2 y
can be solved with respect to {x, y}, one solution being
√ 1
x=r+ r2 + 2s, y=√ (9.63)
r2 + 2s
Under the transformation
√ 1
x = r + r2 + 2s≡α(r, s), y=√ ≡β(r, s) (9.64)
r2 + 2s
and using Appendix A, the original equation (9.17) takes on the form
srr
− √ 3 = 0 (9.65)
2s + r2 + r + sr
from which we conclude that srr = 0 and thus
s = C1 r + C2 (9.66)
Using s and r given by (9.62), the above equation becomes

x2 x 1
− + = C1 x − + C2 (9.67)
2 y y
from which we get
x + C1
y= (9.68)
1 2
x + C1 x + C 2
2
The above is a general solution of the equation (9.17).
(ii) We shall now find another general solution using the generator
∂ ∂
Γ(8) = x2 + (−2xy + 2) (9.69)
∂x ∂y
66
Since X = x2 , Y = −2xy + 2, the system (8.95) takes on the form
∂r 2 ∂r
·x + ·(−2xy + 2) = 0
∂x ∂y
∂s 2 ∂s
·x + ·(−2xy + 2) = 1
∂x ∂y
The above system admits the general solution
1
r = F ((−2 + xy)x), s = − + G((−2 + xy)x) (9.70)
x
where F, G are arbitrary functions. We take
1
r = (−2 + xy)x and s = − (9.71)
x
Inverting the above relations (solving with respect to x and y), we consider
the transformation
1
x=− , y(x) = −2 s(r) + r·s2 (r) (9.72)
s(r)
Using the above transformation and Appendix A, equation (9.17) takes on

the form
d2 3 d
3 d 2
s(r) − r s(r) − 3r s(r) =0 (9.73)
dr2 dr dr
The above equation under the substitution
d
s(r) = U (r) (9.74)
dr
transforms into a first kind Abel equation
dU
− r3 U 3 − 3rU 2 = 0, U = U (r) (9.75)
dr
Under the substitution
1
U= , Y = Y (r) (9.76)
Y
67
equation (9.75) gets transformed into a second kind Abel equation
dY
Y + 3rY + r3 = 0 (9.77)
dr
The above equation has been solved in Appendix C, Example 14. The solu-
tion is given by (under an obvious change of variables)
√
Y (r) = −r2 + C 2 ±C −r2 + C 2 , C > 0 (9.78)
Integrating (9.74) and using (9.76) and (9.78), we find

Z
dr
s(r) = √ (9.79)
−r + C ±C −r2 + C 2
2 2
Since
√
−r2 + C 2
Z
dr 1
√ = ∓
−r2 + C 2 ±C −r2 + C 2 r Cr
going back to the original variables given by (9.71), we obtain from (9.79)
the general solution of (9.17) in implicit form
√
1 1 −r2 + C 2
− = ∓ (9.80)
x r Cr r=(−2+xy)x
9.4. Using software. We are now going to solve second order ordinary
differential equations using MAPLE and Mathematica.
9.4.1. Using MAPLE.
Example. Solve the equation
y 00 + 3yy 0 + y 3 = 0, y = y(x) (9.81)

>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x$2) + 3 ∗ y(x) ∗ diff(y(x), x) + y(x)3 = 0; (9.82)
68
d2 d
2
y(x) + 3 y(x) y(x) + y(x)3 = 0 (9.83)
dx dx
>symgen(ode);
[− ξ = 1, − η = 0], [− ξ = −y, − η = y 3 ], [− ξ = −x, − η = y]
(9.84)
[− ξ = yx, − η = −y 3 x + y 2 ], [− ξ = −x2 /2, − η = yx − 1]
We make the choice X = y and Y = −y 3 using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ y − diff(r(x, y), y) ∗ y 3 = 0 : (9.85)

> eq2 := diff(s(x, y), x) ∗ y − diff(s(x, y), y) ∗ y 3 = 1 : (9.86)

yx − 1 yx − 1
1 2 x
r(x, y) =− F 1 − , s(x, y) = − x + +− F 2 −
y 2 y y
(9.88)
The reader easily recognizes that − F 1 and − F 2 are two arbitrary functions.
We take
yx − 1 1 x
r= s = − x2 + (9.89)
y 2 y
1
> itr := {x = r + sqrt(r2 + 2 ∗ s(r)), y(x) = }; (9.90)
sqrt(r2 + 2 ∗ s(r))
69
The system responds
( )
p 1
x = r + 2s(r) + r2 , y(x) = p (9.91)
2s(r) + r2

The system responds

d2
s(r)
− p dr2
d 3 = 0 (9.93)
2
2s(r) + r + s(r) + r
ds
d2
Step 7. From the above equation it follows that s(r) = 0. We thus
dr2
consider the equation
> eqn := diff(s(r), r$2) = 0 : (9.94)
the system responds
s(r) =− C1r +− C2 (9.96)
Step 8. Using the initial variables, according to (9.89), we transform (9.96)

into
1 x yx − 1
− x2 + = C 1 + C2 (9.97)
2 y y
C1 + x
y= (9.98)
1 2
x + C1 x + C 2
2
70
This is the solution we have already found (with much less effort!).
Note. We are going to determine the symmetries of the equation (we used
before) using different commands of MAPLE. Instead of using the >symgen
command we shall use the >DeterminingPDE command:
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x$2) + 3 ∗ y(x) ∗ diff(y(x), x) + y(x)3 = 0;
d2 d
2
y(x) + 3 y(x) y(x) + y(x)3 = 0
dx dx
>DetSys:=DeterminingPDE(ode); The program responds
∂4 ∂3 ∂4
{ 4 − ξ1 (x, y) = −2 − ξ1 (x, y) y + 2 − ξ1 (x, y) y2,
∂x ∂x3 ∂y∂x3
∂2 ∂
3 ∂2
− ξ1 (x, y) = 0, − η1 (x, y) = − − ξ1 (x, y) y + − ξ1 (x, y)
∂y 2 ∂y ∂x2
∂ ∂2 2 ∂4
2
− − ξ1 (x, y) y + 2 − ξ1 (x, y) y + − ξ1 (x, y)
∂x ∂y∂x 3 ∂y∂x3
∂3
−2 − ξ1 (x, y) y}
∂y∂x2
The above system can be solved using the command >pdsolve. In fact using
>Inf1:=pdsolve(DetSys), the program responds
1 1 1
{ − ξ1 (x, y) = y − C4x4 + (4 − C5y − 2 − C4)x3 + (12 − C1 + 12y − C6
24 24 24
1 1
− 6 − C5)x2 + (24 − C2 + 24 − C7y − 12 − C6)x + y − C8 + − C3 − − C7,
24 2
1
− η1 (x, y) = (− − C4x4 − 4 − C5x3 − 12 − C6x2 − 24 − C7x − 24 − C8)y 3
24
1 1
+ (4 − C4x3 + 12 − C5x2 + 24 − C7 + 24 − C6x)y 2 + (−6 − C4x2
24 24
1 1
+ (−24 − C1 − 12 − C5)x − 12 − C6 − 24 − C2)y + − C4x + − C1 + − C5}
6 6
71
We could find the infinitesimals using a single command >Infinitesimals. In
fact, using
Inf2:=Infinitesimals(ode), the program responds
[− ξ1 (x, y) = 1, − η1 (x, y) = 0], [− ξ1 (x, y) = x, − η1 (x, y) = −y],

1 1
[− ξ1 (x, y) = x2 , − η1 (x, y) = 1 − xy], [− ξ1 (x, y) = − + xy, − η1 (x, y) = −xy 3
2 2
1 3 1 2 1
+ y ], [− ξ1 (x, y) = x y − x , − η1 (x, y) = − (−1 + xy)3 ],
2
6 4 6
1 2 1 1
[− ξ1 (x, y) = x y − x, − η1 (x, y) = − y(−1 + xy)2 ],
2 2 2
1 4 1 3 1 1 1 1
[− ξ1 (x, y) = yx − x , − η1 (x, y) = − x4 y 3 + x3 y 2 − x2 y + x]
24 12 24 6 4 6
We thus see that either using successively DeterminingP DE() and pdsolve()
or Infinitesimals() as above, we gain much more symmetries (in fact the cor-
rect number of symmetry generators) compared to the command symgen().
9.4.2. Using Mathematica. We next demonstrate on how to use the
Mathematica package called MathLie in solving the second order equation.
The MathLie is an AddOn package to Mathematica (Reference [4]). We use
the following steps:
Step 1. We call the package MathLie
In[1]:=<<MathLie.m
Please do not forget to press Shift+Enter after each command.
Step 2. We enter the second order ODE using the command
In[2] := ode = ∂{x,2} y[x] + 3 ∗ y[x] ∗ ∂x y[x] + y[x]3 == 0; ode//LT F
where LTF means ”Lie Traditional Form”. The program responds
Out[2]//DisplayF orm =
y 3 + 3yyx + yx,x == 0
Step 3. We calculate the second prolongation, entering the command
In[3] := pro2 = P rolongation[ode, {y}, {x}]; pro2//LT F
72
3y 2 φ1 + 3yx φ1 − 3yyx (ξ1 )x − 3yyx2 (ξ1 )y + 3y(φ1 )x + 3yyx (φ1 )y
− 2(ξ1 )x yx,x − 3yx (ξ1 )y yx,x + (φ1 )y yx,x
− yx (ξ1 )x,x − 2yx2 (ξ1 )x,y − yx3 (ξ1 )y,y + (φ1 )x,x
+ 2yx (φ1 )x,y + yx2 (φ1 )y,y == 0
Step 4. We now have to take into account the condition ∆ = 0, i.e. we

substitute in the above expression yx,x by −3yyx − y 3 using the command
In[4] := prolong2 = pro2/.∂{x,2} y[x]→−3∗y[x]∗∂x y[x]−y[x]3 ; prolong2//LT F

3y 2 φ1 + 3yx φ1 − 3yyx (ξ1 )x − 2(−y 3 − 3yyx )(ξ1 )x − 3yyx2 (ξ1 )y
− 3yx (−y 3 − 3yyx )(ξ1 )y + 3y(φ1 )x + 3yyx (φ1 )y + (−y 3 − 3yyx )(φ1 )y
− yx (ξ1 )x,x − 2yx2 (ξ1 )x,y − yx3 (ξ1 )y,y + (φ1 )x,x
+ 2yx (φ1 )x,y + yx2 (φ1 )y,y == 0
Step 5. We collect together terms arranged with respect to yx using the

command
In[5] := cprolong2 = Collect[prolong2, {∂x y[x]}]; cprolong2//LT F
Out[5] :=
3y 2 φ1 + 2y 3 (ξ1 )x + 3y(φ1 )x − y 3 (φ1 )y − yx3 (ξ1 )y,y + (φ1 )x,x
+ yx (3φ1 + 3y(ξ1 )x + 3y 3 (ξ1 )y − (ξ1 )x,x + 2(φ1 )x,y )
+ yx2 (6y(ξ1 )x − 2(ξ1 )x,y + (φ1 )y,y ) == 0
From the above equation we can easily read off the determining equations.
However we can find the determining equations with a single command
In[6] := detode = DeterminingEquations[ode, {y}, {x}, {∂{x,2} y[x]}];

detode//LT F
73
(ξ1 )y,y == 0
3φ1 + 3y(ξ1 )x + 3y 3 (ξ1 )y − (ξ1 )x,x + 2(φ1 )x,y == 0
3y 2 φ1 + 2y 3 (ξ1 )x + 3y(φ1 )x − y 3 (φ1 )y − yx3 (ξ1 )y,y + (φ1 )x,x == 0
6y(ξ1 )x − 2(ξ1 )x,y + (φ1 )y,y == 0
It is obvious that the above system can be solved using the command DSolve[]
and thus to determine the infinitesimals of the heat equation.
Step 6. The infinitesimals can also be determined using the single command
In[6] := infode = Infinitesimals[ode, {y}, {x}]; infode//LT F
φ1 == 2(k7 − 2 k8 x) − (k2 − 6k8x2 )y

− (k5 + x(2k6 + 3k7x + 4k8x2 ))y 2
+ (k4 + x(k5 + x(k6 + x(k7 + k8x))))y 3 + k1(2 − 2xy)
ξ1 == k3 + k5 + k2x + 2k6x + k1x2 + 3k7x2 + 2k8x3
− (k4 + x(k5 + x(k6 + x(k7 + k8x))))y
Step 7. Using the above infinitesimals, we can find the invariants and then
to determine some solutions of the equation. The choice k1 = k2 = k3 =
k5 = k6 = k7 = k8 = 0 and k4 = 1 gives
xi[1][x, y] = −y;
phi[1][x, y] = y 3 ;
Let us now consider the canonical equations in its general from
In[7] := canonicalEquations = {ξ[x, y] ∗ ∂x s[x, y] + φ[x, y] ∗ ∂y s[x, y] == 1,

ξ[x, y] ∗ ∂x r[x, y] + φ[x, y] ∗ ∂y r[x, y] == 0};
canonicalEquations//LT F//T ableF orm
Out[7] :=
− 1 + ξsx + φsy == 0
ξrx + φry == 0
74
Step 8. Introduce the functions (i.e. the substitutions ξ→y, φ→ − y 3 )
In[8] := veccan =
{ξ→F unction[{x, y}, y], φ→F unction[{x, y}, −y 3 ]};
and then substitute them into the canonical equations
In[9] := caneq =
canonicalEquations/.veccan; caneq//LT F//T ableF orm
Out[9] :=
− 1 + ysx − y 3 sy == 0
yrx − y 3 ry == 0
Step 9. We now solve the above system of equations
In[10] : solc1 = DSolve[caneq[1], s[x, y], {x, y}]
with response

−1 + xy 

2

 2x − x y + 2yC[1] 
 y 
s[x, y]→

 2y 

 
and
In[11] : solc2 = DSolve[caneq[2], r[x, y], {x, y}]
with response

−1 + xy
r[x, y]→C[1]
y
Step 10. Under the choice
2x − x2 y −1 + xy
s= , r=
2y y
75
i.e.
√ 1
x=r+ r2 + 2s, y=√
r2 + 2s
and using Appendix B, we transform the original equation y 00 + 3yy 0 + y 3 = 0
into the equation srr = 0 from which we get s = C1 + C2 r and then going to
the original variables, we find the general solution of the equation.
MathLie is a really powerfull program for performing Lie symmetry analysis
of DEs or systems of DEs following all the intermediate steps as above. If in
addition the reader wanted the explicit expressions for the coefficients Y [x]
and Y [xx] , here are the commands:
firstExtensionYx = F rechetP rolong[{∂x y[x]}, {y}, {x},

{xi[1][x, y[x]]}, {phi[1][x, y[x]]}]; firstExtensionYx//LT F
and
secondExtensionYxx = F rechetP rolong[{∂{x,2} y[x]}, {y}, {x},
{xi[1][x, y[x]]}, {phi[1][x, y[x]]}]; secondExtensionYxx//LT F
respectively.
9.5. Determination of the equation with known symmetry
generators.
In this paragraph we shall determine the form of a second order differential
equation
y 00 = F (x, y, y 0 ) (9.99)
admitting the known symmetries

∂ ∂ ∂ ∂ ∂ ∂
Γ(1) = + , Γ(2) = x + y , Γ(3) = x2 + y2 (9.100)
∂x ∂y ∂x ∂y ∂x ∂y
In other words we shall determine the most general form of the function
F (x, y, y 0 ). We consider first the Invariance Condition
Γ(1) F = 0 (9.101)
Since in this case we have X = 1 and Y = 1, the invariance condition becomes
Fx + Fy = 0 (9.102)
76
from which we get that F should have the form
F = F (x − y, y 0 ) (9.103)
Considering Γ(2) , the invariance condition takes on the form
x·Fx + y·Fy + F = 0 (9.104)
which may also be written as (since Fy = −Fx )
x·Fx + y·(−Fx ) + F = 0 (9.105)
or even
(x − y)·Fx + F = 0 (9.106)
From the above equation we get that

G(y 0 )
F = (9.107)
x−y
Considering finally Γ(3) , we have X = x2 and Y = y 2 and thus the Invariance

Condition becomes with F given by (9.107)
x2 y2

0 0 2
− 2y + 2(y ) + − G(y 0 )
(x − y)2 (x − y)2
G(y 0 ) dG(y 0 )
+ 2(y − 2x) + 2y 0 =0
x−y dy 0
dG(y 0 )
0 0
0 2
2y 0
− 3G(y ) + 2 (y ) − y 0 = 0 (9.108)
dy
The general solution of the above equation is given by
G(y 0 ) = −2 y 0 + C·(y 0 )3/2 + (y 0 )2

(9.109)
Taking into account (9.107), (9.109) and (9.99), we see that the differential
equation has the form
−2 y 0 + C·(y 0 )3/2 + (y 0 )2

00
y = (9.110)
x−y
77
or even
2 0
y 00 + y + C·(y 0 )3/2 + (y 0 )2 = 0 (9.111)
x−y
9.6. More Examples on 2nd Order Differential Equations.
Let us now solve some more second order differential equations.
These differential equations will be solved using two methods:
(I) The Differential Invariants Method and
(II) The Normal Coordinates Method.
Given a second order ODE
y 00 = F (x, y, y 0 )
a differential invariant or a first integral or even a conservation law is
a function
I = I(x, y, y 0 )
which satisfies the equation
Dx I = 0, ∆ ≡ y 00 − F
∆=0
where Dx is the total differential operator. The function I usually has the
general form
I = α(x, y)(y 0 )2 + β(x, y)y 0 + γ(x, y)
An example is provided in subsection 9.9.
Alternatively, if
∂ ∂
Γ=X +Y
∂x ∂y
is a symmetry generator of the given equation, a differential invariant I0
satisfies the equation
∂I0 ∂I0
X +Y =0
∂x ∂y
The first prolongation
∂ ∂ ∂
P r(1) Γ(∆) = X +Y + Y [x] 0
∂x ∂y ∂y
78
provides the invariant I1 ,
I1 = Y [x]
evaluated for the given X and Y . The second order invariant I2 is evaluated
by
dI1
I2 =
dI0
Upon eliminating y, y 0 and y 00 using the original equation and the expressions
for I0 , I1 and I2 , we obtain a first order equation of the form
dI1
= G(I0 , I1 )
dI0
from which we find I1 = I1 (I0 ). We then determine the solution of the
original equation.
Example 1. Solve the Emden-Fowler differential equation
d2 y
2
= Ax−6 y 3 (9.112)
dx
Solution. This equation admits the infinitesimal generators
[X = x, Y = 2y] and [X = x2 , Y = xy] (9.113)
Let us take X = x2 , Y = xy. In this case we have
∂ ∂
Γ = x2 + xy (9.114)
∂x ∂y
The differential invariant I0 satisfies the equation
∂I0 ∂I0
x2 + xy =0 (9.115)
∂x ∂y
The above differential equation admits the obvious solution
y
I0 = (9.116)
x
The differential invariant I1 can be evaluated by the first prolongation
∂ ∂ ∂ ∂ 2
I1 = Dx (Y ) − y 0 Dx (X) = + y0 (xy) − y 0 + y0 (x )
∂x ∂y ∂x ∂y (9.117)
0 0 0
= (y + y x) − y (2x) = y − xy
79
The differential invariant I2 is given by
dI1
dI1
I2 = = dx (9.118)
dI0 dI0
dx
and since
dI1 dI0 xy 0 − y
= xy 00 , = (9.119)
dx dx x2
we get
dI1 x3 y 00
I2 = = 0 (9.120)
dI0 xy − y
Since
I1 dI1
y = xI0 and y 00 = (9.121)
x3 dI0
the original equation y 00 = Ax−6 y 3 takes on the form
I1 dI1
3
= Ax−6 (xI0 )3
x dI0
dI1
I1 = AI03 (9.122)
dI0
Introducing the traditional notation
u = I0 and w = I1 (9.123)
we transform the original differential equation into

dw
w = Au3 (9.124)
du
which is a first order separable differential equation. The general solution of
this last equation is given by
A 4
w2 = u +C
2
80
from which we get
r
A 4
w=± u +C (9.125)
2
which can be written, using the initial variables, as
r
A y 4
y − xy 0 = ± +C (9.126)
2 x
The above is a first order ordinary differential equation. For A = 2 and
C = 0 for example, we obtain the solutions
x2
y= , B is a constant (9.127)
∓1 + B·x
d2 y α dy dy 2
y + − + βy 3 = 0 (9.128)
dx2 x dx dx
Solution. Considering the scale transformation
ŷ = λy, x̂ = µx (9.129)
since
dŷ λ dy d2 ŷ λ d2 y
= and = (9.130)
dx̂ µ dx dx̂2 µ2 dx2
the original equation transforms into
d2 y
α dy dy 2
y + − + λµ2 βy 3 = 0 (9.131)
dx2 x dx dx
Scale invariance requires
λµ2 = 1 (9.132)
Therefore we can take
λ = e−2 and µ = e (9.133)
81
We thus see that the original equation is scale invariant under the transfor-
mation
ŷ = e−2 y, x̂ = e x (9.134)
Upon Taylor expanding in , we have

ŷ = e−2 y = (1 − 2 + O(2 ))y = 1 + (−2y) + O(2 )
(9.135)
x̂ = e x = (1 + + O(2 ))y = 1 + x + O(2 )
From the above expansions, we can read off the infinitesimal coefficients:
X = x Y = −2y (9.136)
Therefore we have
∂ ∂
Γ=x − 2y (9.137)
∂x ∂y
Performing a Lie symmetry analysis we obtain exactly the same infinitesimal
coefficients.
Let us now try to find some general solutions of the equation. In this case
the system
∂r ∂r
·X(x, y) + ·Y (x, y) = 0
∂x ∂y
∂s ∂s
·X(x, y) + ·Y (x, y) = 1
∂x ∂y
becomes
x rx − 2y ry = 0, x sx − 2y sy = 1
r = F (x2 y) and s = ln(x) + G(x2 y) (9.138)
A simple choice is
r = x2 y, s = ln(x) (9.139)
Solving the above system with respect to {x, y} we get
x = es , y = r·e−2s (9.140)
82
Using the above transformation and Appendix A, the original equation (9.128)
takes on the form
d2 d 3
s(r) + r[2(α − 1) − βr] s(r)
dr2 dr (9.141)
d 2 1 d
− (α − 1) s(r) + s(r) = 0
dr r dr
The substitution
d
s(r) = w(r) (9.142)
dr
converts equation (9.141) into
dw 1
+ r[2(α − 1) − βr]w3 − (α − 1)w2 + w = 0 (9.143)
dr r
The transformation
1
w= (9.144)
r·u(r)
du 2(α − 1)
u = −(α − 1)u + −β (9.145)
dr r
Under the further transformation
z
r=− (α6=1) (9.146)
α−1
equation (9.145) gets the form
du 2(α − 1) β
u −u= + (α6=1) (9.147)
dz z α−1
The above equation is accessible by Panayotounakos algorithm.
Let us now suppose we know u(r) thanks to Panayotounakos algorithm.
Integrating (9.142) and using (9.144), we find
Z
dr
s(r) = (9.148)
r·u(r)
83
and then going back to original variables through (9.139), we obtain the
solution of the original equation in implicit form
Z
dr
ln(x) = (9.149)
r·u(r) r=x2 y
Note 1. For α = 1 equation (9.145) becomes
du
u = −β (9.150)
dr
whose general solution is given by
p
u = ± −2βr + C (9.151)
Note 2. For α = 0 equation (9.145) becomes
du 2
u −u=− −β (9.152)
dr r
whose general solution can be found using Panayotounakos algorithm.
However for α = 0 the original equation has the form
d2 y dy 2
y − + βy 3 = 0 (9.153)
dx2 dx
and since does not contain the independent variable explicitly, it can be
solved by the substitution
dy d2 y dp
p= , 2
=p (9.154)
dx dx dy
Under the above substitution, equation (9.153) becomes
dp 1 1
− p = −βy 2 (9.155)
dy y p
which is a Bernoulli differential equation and can be solved by the substitu-
tion
w = p2 (9.156)
84
Under the above substitution, equation (9.155) takes on the form
dw 2
− = −2βy 2 (9.157)
dy y
which is a linear first order differential equation with general solution
w = C1 y 2 − 2βy 3 (9.158)
Since
p dy
p = ± C1 y 2 − 2βy 3 and p = (9.159)
dx
we arrive at the first order ordinary differential equation
dy p
= ± C1 y 2 − 2βy 3 (9.160)
dx
The above is a separable differential equation with general solution given by
2 r −2βy + C
1
x + C2 = ± √ arctanh (9.161)
C1 C1
d2 y dy 2 dy
x2 2 + x −y =0 (9.162)
dx dx dx
Solution. This equation is scale invariant, i.e. it is invariant under
x̂ = e x, ŷ = e y (9.163)
and thus admits the infinitesimal coefficients
X = x and Y = y (9.164)
Therefore
∂ ∂
Γ=x +y (9.165)
∂x ∂y
Let I0 be a solution of the equation
∂I0 ∂I0
x +y =0
∂x ∂y
85
and I1 the first invariant which in our case is zero. We may then consider
the invariants
y
I0 = and I1 = y 0 (9.166)
x
The invariant I2 is evaluated by
dI1 dI1 /dx y 00 x2 y 00
I2 = = = 0 = 0 (9.167)
dI0 dI0 /dx y y xy − y
− 2
x x
Using the substitutions
y
u = I0 = and w = I1 = y 0 (9.168)
x
we have
dw x2 y 00
I2 = = 0 (9.169)
du xy − y
Let us now eliminate y 0 and y 00 using w, I2 and then u. From (9.169) we
obtain
dw
x2 y 00 = (xy 0 − y)
du
dw
x2 y 00 = (xw − y) (9.170)
du
We also have
x(y 0 )2 = xw2 and yy 0 = yw (9.171)
Therefore equation (9.162) becomes because of (9.170) and (9.171)

dw
(xw − y) + xw2 − yw = 0 (9.172)
du
Dividing the above equation by x and using the definition of u, i.e. u = y/x,
we obtain the equation
dw
(w − u) + w2 − uw = 0 (9.173)
du
86
Writing the above equation as
dw dw
w2 + −u w−u =0 (9.174)
du du
and solving the quadratic equation with respect to w, we obtain the solutions
dw
w = u and w = − (9.175)
du
Equation w = u is equivalent to
y
y0 = (9.176)
x
which admits the general solution
y = C·x (9.177)
The second equation of (9.175) is equivalent to
dw
− = du (9.178)
w
and by integration
u + C1 = − ln(w) (9.179)
and then
w = C·e−u (9.180)
The above differential equation, written as
dy y
= C·exp − (9.181)
dx x
admits the following solution
Z
dr
ln(x) + C2 = (9.182)
−r + C1 ·e−r r=y/x
Let us now solve the equation using normal coordinates. Since X = x, Y =
y, the system
∂r ∂r
·X(x, y) + ·Y (x, y) = 0
∂x ∂y
∂s ∂s
·X(x, y) + ·Y (x, y) = 1
∂x ∂y
87
becomes
x rx + y ry = 0, x sx + y s y = 1
r = F (y/x) and s = ln(x) + G(y/x) (9.183)
A simple choice is
y
r= , s = ln(x) (9.184)
x
Solving the above system with respect to {x, y} we get
x = es , y = r·es (9.185)
Using the above transformation and Appendix A, the original equation (9.162)
takes on the form
d2 d 2 d
2
s(r) − (r + 1) s(r) − s(r) = 0 (9.186)
dr dr dr
The substitution
d
s(r) = w(r) (9.187)
dr
dw
− w = (r + 1)w2 (9.188)
dr
which is a Bernoulli differential equation with general solution
1
w= (9.189)
−r + C·e−r
Upon integration of (9.187) and taking into account (9.189) we have
Z
dr
s(r) = (9.190)
−r + C·e−r
Going back to the original variables through (9.184), we obtain
Z
dr
ln(x) + C2 = (9.191)
−r + C1 ·e−r r=y/x
88
The above is the general solution of the given equation in implicit form.
9.7. General Examples on 2nd Order ODEs.
Example 1. We consider the equation
y 00 = 0 (9.192)
Invariance with respect to infinitesimal point transformations is being ex-

pressed by the condition
P r(2) Γ(∆)|∆=0 = 0, ∆≡y 00 (9.193)
where P r(2) Γ is the operator

∂ ∂ ∂ ∂
P r(2) Γ = X +Y + Y [x] 0 + Y [xx] 00 (9.194)
∂x ∂y ∂y ∂y
The infinitesimals are evaluated by the formulas
Y [x] = Dx (Y ) − y 0 Dx (X) (9.195)
Y [xx] = Dx (Y [x] ) − y 00 Dx (X) (9.196)

where the total derivative operator Dx is defined by
∂ ∂ ∂
Dx = + y0 + y 00 0 (9.197)
∂x ∂y ∂y
We have the following explicit expressions for the infinitesimals:
Y [x] = Yx + (Yy − Xx )y 0 − Xy (y 0 )2 (9.198)
and
Y [xx] = Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3
(9.199)
+ (Yy − 2Xx )y 00 − 3Xy y 0 y 00
We obtain in our case
P r(2) Γ(∆) = Y [xx] (9.200)
and then
P r(2) Γ(∆)|∆=0 = 0 (9.201)
89
is equivalent to
Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3 = 0 (9.202)
taking into account (9.200), (9.199) and taking ∆ = y 00 = 0.

Equating to zero all the coefficients of y 0 , we obtain the system
Yxx = 0 (9.203)
2Yxy − Xxx = 0 (9.204)

Yyy − 2Xxy = 0 (9.205)
Xyy = 0 (9.206)
From (9.206) and (9.203) we obtain
X = α(x)y + β(x) (9.207)
and
Y = γ(y)x + δ(y) (9.208)
respectively. Equations (9.204) and (9.205) then become
2γ 0 (y) − α00 (x)y − β 00 (x) = 0 (9.209)
and
γ 00 (y)x + δ 00 (y) − 2α0 (x) = 0 (9.210)
respectively. Differentiation of (9.209) with respect to y yields
2γ 00 (y) − α00 (x) = 0 (9.211)
from which we get the system
γ 00 (y) = 0, α00 (x) = 0 (9.212)
Therefore we have
α(x) = a1 x + a2 and γ(y) = a3 y + a4 (9.213)
90
Equation (9.209) then yields β 00 (x) = 2a3 and thus
β(x) = a3 x2 + a5 x + a6 (9.214)
Similarly equation (9.210) becomes δ 00 (y) = 2a1 and thus
δ(y) = a1 y 2 + a7 y + a8 (9.215)
Therefore we have the following expressions for X and Y :
X = (a1 x + a2 )y + a3 x2 + a5 x + a6 (9.216)
and
Y = (a3 y + a4 )x + a1 y 2 + a7 y + a8 (9.217)
The symmetry generator then yields

∂ ∂
Γ=X +Y
∂x ∂y
∂
= [(a1 x + a2 )y + a3 x2 + a5 x + a6 ]
∂x
∂
+ [(a3 y + a4 )x + a1 y 2 + a7 y + a8 ] (9.218)
∂y
∂ ∂ ∂ ∂ ∂
= a1 xy + y2 + a2 y + a3 x 2 + xy
∂x ∂y ∂x ∂x ∂y
∂ ∂ ∂ ∂ ∂
+ a4 x + a5 x + a6 + a7 y + a8
∂y ∂x ∂x ∂y ∂y
We thus obtain the following generators
∂ ∂ ∂ ∂ ∂
Γ(1) = xy + y 2 , Γ(2) = y , Γ(3) = x2 + xy ,
∂x ∂y ∂x ∂x ∂y
(9.219)
(4) ∂ ∂ ∂ ∂ ∂
Γ = x , Γ(5) = x , Γ(6) = , Γ(7) = y , Γ(8) =
∂y ∂x ∂x ∂y ∂y
Equation y 00 (x) = 0 despite its simplicity, admits eight symmetry generatos.

Example 2. We consider next the equation
y 00 = F (x, y) (9.220)
91
P r(2) Γ(∆)|∆=0 = 0, ∆≡y 00 − F (x, y) (9.221)

∂ ∂ ∂ ∂
P r(2) Γ = X +Y + Y [x] 0 + Y [xx] 00 (9.222)
∂x ∂y ∂y ∂y
The infinitesimals are evaluated by the formulas (9.198) and (9.199) respec-
tively. We have in our case
P r(2) Γ(∆) = X(−Fx ) + Y (−Fy ) + Y [xx] (9.223)
and in expanded form
P r(2) Γ(∆) = −XFx − Y Fy

+ (Yy − 2Xx )y 00 − 3Xy y 0 y 00
The condition P r(2) Γ(∆)|∆=0 = 0 is derived from the above expression upon
substituting y 00 by F :
P r(2) Γ(∆)|∆=0 = 0 ⇐⇒ − XFx − Y Fy

+ (Yy − 2Xx )F − 3Xy F y 0 = 0
Equating to zero the coefficients of y 0 we obtain the system
2Yxy − Xxx − 3Xy F = 0 (9.226)
Yyy − 2Xxy = 0 (9.227)

Xyy = 0 (9.228)
−XFx − Y Fy + Yxx + (Yy − 2Xx )F = 0 (9.229)
From equation (9.228) we get
X = α(x)y + β(x) (9.230)
92
Equation (9.227) then yields Yyy − 2α0 (x) = 0 and by integration
Y = α0 (x)y 2 + γ(x)y + δ(x) (9.231)
Because of (9.230) and (9.231), equations (9.226) and (9.229) become
3α00 (x)y + 2γ 0 (x) − β 00 (x) − 3α(x)F = 0 (9.232)
and
− [α(x)y + β(x)]Fx − [α0 (x)y 2 + γ(x)y + δ(x)]Fy
(9.233)
+ α000 (x)y 2 + γ 00 (x)y + δ 00 (x) + [γ(x) − 2β 0 (x)]F = 0
respectively. At this stage we have to make a number of assumptions about

the general or the explicit form of the function F and then find the general
solution of equations (9.232) and (9.233).
Application. The Emden-Fowler equation. We consider the function
F (x, y) = Axn y m , m6=0 and m6=1 (9.234)
which corresponds to the general form of the Emden-Fowler equation
y 00 = Axn y m (9.235)
In this case, since Fx = Anxn−1 y m and Fy = Anxn y m−1 , equations (9.232)

and (9.233) take the form
3α00 (x)y + 2γ 0 (x) − β 00 (x) − 3Aa(x)xn y m = 0 (9.236)
and
− [α(x)y + β(x)]Anxn−1 y m − [α0 (x)y 2 + γ(x)y + δ(x)]Amxn y m−1
(9.237)
+ α000 (x)y 2 + γ 00 (x)y + δ 00 (x) + [γ(x) − 2β 0 (x)]Axn y m = 0
respectively. We distinguish the following cases:

(I) Let m6=2. In this case we obtain from (9.236) that
α(x) = 0 (9.238)
and 2γ 0 (x) − β 00 (x) = 0 and by integration

1
γ(x) = β 0 (x) + a1 (9.239)
2
93
Equation(9.237) then yields, because of (9.238) and (9.239)

n−1 m 1 0
− β(x)Anx y − β (x) + a1 y + δ(x) Amxn y m−1
2

1 000 00 3 0
+ β (x)y + δ (x) + − β (x) + a1 Axn y m = 0
2 2

1 0
A −n β(x) + − (m + 3)β (x) + (1 − m)a1 x xn−1 y m
2 (9.240)
1
− Am δ(x)xn y m−1 + β 000 (x)y + δ 00 (x) = 0
2
From the above equation we obtain the following system

1 0
− n β(x) + − (m + 3)β (x) + (1 − m)a1 x = 0
2 (9.241)
δ(x) = 0, β (x) = 0, δ 00 (x) = 0
000
Therefore
δ(x) = 0 (9.242)
and
β(x) = a2 x2 + a3 x + a4 (9.243)
From (9.239) we can determine γ(x):

1
γ(x) = a2 x + a3 + a1 (9.244)
2
The first of (9.241) then yields

2 1
−n(a2 x + a3 x + a4 ) + − (m + 3)(2a2 x + a3 ) + (1 − m)a1 x = 0
2
1
− (n + m + 3)a2 x2 + − (m + 2n + 3)a3 + (1 − m)a1 x
2 (9.245)
− na4 = 0
94
We thus obtain the system
(n + m + 3)a2 = 0
1
− (m + 2n + 3)a3 + (1 − m)a1 = 0 (9.246)
2
na4 = 0
Let us now list the values of the coefficients X and Y taking into account
(9.230), (9.231) and the various expressions of the functions α(x), β(x), γ(x)
and δ(x) we have found previously:
X = a2 x 2 + a3 x + a4
1 (9.247)
Y = a2 x + a3 + a1 y
2
We continue next by considering the system (9.246).
(Ia) If n + m + 3 = 0 then the second of (9.246) yields the equation
1 2(1 − m)
− na3 + (1 − m)a1 = 0. If n6=0, then we shall have a3 = a1 . In this
2 n
case we may take a2 6=0 while a4 = 0. Therefore in this case, using (9.247),
2(1 − m) 2(n + 4)
X = a2 x 2 + a1 x = a2 x 2 + a1 x
n n (9.248)
1 − m + n 2n + 4
Y = a2 x + a1 y = a2 x + a1 y
n n
The symmetry generator is
∂ ∂
Γ=X +Y
∂x ∂y
2(n + 4) ∂ 2n + 4 ∂
= a2 x 2 + a1 x + a2 x + a1 y (9.249)
n ∂x n ∂y
2 ∂ ∂ ∂ ∂
= a1 (n + 4)x + (n + 2)y + a2 x 2 + xy
n ∂x ∂y ∂x ∂y
and thus
∂ ∂ ∂ ∂
Γ(1) = (n + 4)x + (n + 2)y , Γ(2) = x2 + xy (9.250)
∂x ∂y ∂x ∂y
If n = 0, we may take a3 6=0, a4 6=0 and a2 6=0. On the other hand in this
case we will have a1 = 0 which is a consequence of (1 − m)a1 = 0 for m6=1.
95
Therefore we get from (9.247)
X = a2 x 2 + a3 x + a4
1 (9.251)
Y = a2 x + a3 y
2
∂ ∂
Γ=X +Y
∂x ∂y
∂ 1 ∂
= (a2 x2 + a3 x + a4 ) + a2 x + a3 y (9.252)
∂x 2 ∂y
∂ ∂ ∂ 1 ∂ ∂
= a2 x2 + xy + a3 x + y + a4
∂x ∂y ∂x 2 ∂y ∂x
and thus
∂ ∂ ∂ 1 ∂ ∂
Γ(1) = x2 + xy , Γ(2) = x + y , Γ(3) = (9.253)
∂x ∂y ∂x 2 ∂y ∂x
(Ib) Let n = 0. In this case we may take a4 6=0. We then obtain from (9.246)
(m + 3)a2 = 0
1 (9.254)
− (m + 3)a3 + (1 − m)a1 = 0
2
2(1 − m)
(Ib1) If m + 36=0 then a2 = 0 and a3 = a1 . Therefore we get from
m+3
(9.247)
2(1 − m)
X = a1 x + a4
m+3 (9.255)
4
Y = a1 y
m+3
∂ ∂
Γ=X +Y
∂x ∂y
2(1 − m) ∂ 4 ∂
= a1 x + a4 + a1 y (9.256)
m+3 ∂x m + 3 ∂y
2 ∂ ∂ ∂
= a1 (1 − m)x + 2y + a4
m+3 ∂x ∂y ∂x
96
and thus
∂ ∂ ∂
Γ(1) = (1 − m)x + 2y , Γ(2) = (9.257)
∂x ∂y ∂x
(Ib2) If m + 3 = 0, we may take a2 6=0 and a3 6=0. On the other hand we get
a1 = 0. Therefore we get from (9.247)
X = a2 x 2 + a3 x + a4
1 (9.258)
Y = a2 x + a3 y
2
which is similar to the case (Ia) and the generators are given by (9.253).
(Ic) Let all the coefficients of (9.246) be different than zero, i.e. n + m + 36=0,
m + 2n + 36=0 and n6=0. In this case we have a2 = 0, a4 = 0 and a3 =
2(1 − m)
a1 . Therefore we get from (9.247)
m + 2n + 3
2(1 − m)
X = a1 x
m + 2n + 3 (9.259)
2n + 4
Y = a1 y
m + 2n + 3
∂ ∂
Γ=X +Y
∂x ∂y
2(1 − m) ∂ 2n + 4 ∂
= a1 x + a1 y (9.260)
m + 2n + 3 ∂x m + 2n + 3 ∂y
2 ∂ ∂
= a1 (1 − m)x + (n + 2)y
m + 2n + 3 ∂x ∂y
and thus
∂ ∂
Γ(1) = (1 − m)x + (n + 2)y (9.261)
∂x ∂y
(II) Let m = 2. In this case we obtain from (9.236) that
α(x) = 0 (9.262)
and 2γ 0 (x) − β 00 (x) = 0 and by integration
1
γ(x) = β 0 (x) + a1 (9.263)
2
97
Equation(9.237) then yields, because of (9.262) and (9.263) (for m = 2)

n−1 2 1 0
− nAβ(x)x y − 2A β (x) + a1 y + δ(x) xn y
2

1 000 00 3 0
+ β (x)y + δ (x) + A − β (x) + a1 xn y 2 = 0
2 2

5 0
− A n β(x) + β (x) + a1 x xn−1 y 2
2
(9.264)
1 000
+ − 2A δ(x)x + β (x) y + δ 00 (x) = 0
n
2
From the above equation we obtain the system

5 0
n β(x) + β (x) + a1 x = 0,
2 (9.265)
1
− 2A δ(x)xn + β 000 (x) = 0, δ 00 (x) = 0
2
Therefore
δ(x) = a2 x + a3 (9.266)
and then the second equation of the system (9.265) yields
1
−2A (a2 x + a3 )xn + β 000 (x) = 0 (9.267)
2
The general solution of the above equation is
4A
β(x) = a2 xn+4
(n + 2)(n + 3)(n + 4)
(9.268)
4A 1
+ a3 xn+3 + a4 x2 + a5 x + a6
(n + 1)(n + 2)(n + 3) 2
The first of (9.265), because of (9.268) becomes
28 28
A 16 + n A 12 + n
a2 5 xn+3 + a3 5 xn+2
(n + 2)(n + 3)(n + 4) (n + 1)(n + 2)(n + 3) (9.269)
n 2n 1 2n 2
+ a4 + 1 x + a6 + + 1 a5 + a1 = 0
5 5 x 5 5
98
Using (9.263), we can determine γ(x):
2A
γ(x) = a2 xn+3
(n + 2)(n + 3)
(9.270)
2A 1 1
+ a3 xn+2 + a4 x + a5 + a1
(n + 1)(n + 2) 2 2
Let us now list the values of the coefficients X and Y taking into account
(9.230), (9.231) and the various expressions of the functions α(x), β(x), γ(x)
and δ(x) we have found previously:
4A
X = a2 xn+4
(n + 2)(n + 3)(n + 4)
4A 1
+ a3 xn+3 + a4 x2 + a5 x + a6
(n + 1)(n + 2)(n + 3) 2
2A (9.271)
Y = a2 xn+3
(n + 2)(n + 3)
2A n+2 1 1
+ a3 x + a4 x + a5 + a1 y
(n + 1)(n + 2) 2 2
+ a2 x + a3
We continue next by considering the consistency condition (9.269). In the
general case, provided all the denominators are non-zero ( n6=−1, −2, −3, −4)
we get
5
a2 = 0, a3 = 0, a4 = 0, a6 = 0 and a1 = − n + a5 (9.272)
2
We thus obtain in this case
X = a5 x
1 (9.273)
Y = a5 + a1 y = −a5 (n + 2)y
2
∂ ∂ ∂ ∂
Γ=X +Y = a5 x − a5 (n + 2)y (9.274)
∂x ∂y ∂x ∂y
and thus
∂ ∂
Γ(1) = x − (n + 2)y (9.275)
∂x ∂y
99
We may also consider the special cases
28 28 n 2n
16 + n = 0, 12 + n = 0, + 1 = 0, + 1 = 0, n = 0 (9.276)
5 5 5 5
as it is evident from (9.269). In all these cases we may take the corresponding
coefficient appearing in (9.269) different than zero.
28 20
Case 1. For 16 + n = 0, i.e. n = − we may take a2 6=0, a3 = a4 = a6 = 0
5 7
5
and a1 = a5 . We thus have in this case form (9.271)
14
343A 8/7
X = −a2 x + a5 x
12 (9.277)
49A 1/7 39
Y = −a2 x + a5 y + a2 x
3 70
∂ ∂
Γ=X +Y
∂x ∂y
343A 8/7 ∂
= −a2 x + a5 x
12 ∂x
49A 1/7 39 ∂
+ −a2 x + a5 y + a2 x
3 70 ∂y
343A ∂ 49A 1/7 ∂
= a2 − x8/7 − x y+x
12 ∂x 3 ∂y
∂ 39 ∂
+ a5 x + y
∂x 70 ∂y
and thus the generators of y 00 = Ax−20/7 y 2 are given by
343A 8/7 ∂ 49A 1/7
Γ(1) = − x − x y+x
12 ∂x 3
(9.278)
(2) ∂ 39 ∂
Γ =x + y
∂x 70 ∂y
28 15
Case 2. For 12 + n = 0, i.e. n = − we may take a3 6=0, a2 = a4 = a6 = 0
5 7
5
and a1 = − a5 . We thus have in this case form (9.271)
14
3423 6/7
X = a3 x + a5 x
12 (9.279)
49A 11
−1/7
Y = a3 x − a5 y + a3
4 70
100
∂ ∂
Γ=X +Y
∂x ∂y
343A ∂
6/7
= a3 x + a5 x
12 ∂x
49A 11 ∂
+ a3 x−1/7 − a5 y + a3
4 70 ∂y
343A
6/7 ∂ 49A −1/7 ∂
= a3 x + x y+1
12 ∂x 4 ∂y
∂ 11 ∂
+ a5 x − y
∂x 70 ∂y
and thus the generators of y 00 = Ax−15/7 y 2 are given by
343A 6/7 ∂ 49A −1/7
Γ(1) = x + x y+1
12 ∂x 4
(9.280)
∂ 11 ∂
Γ(2) = x − y
∂x 70 ∂y
n
Case 3. For + 1 = 0, i.e. n = −5, we may take a4 6=0, a2 = a3 = a6 = 0
5
5
and a1 = a5 . We thus have in this case form (9.271)
2
1
X = a4 x 2 + a5 x
2 (9.281)
1
Y = a4 x + 3a5 y
2
∂ ∂
Γ=X +Y
∂x ∂y
1 ∂ 1 ∂
= a4 x 2 + a5 x + a4 x + 3a5 y
2 ∂x 2 ∂y
1 ∂ ∂ ∂ ∂
= a4 x2 + xy + a5 x + 3y
2 ∂x ∂y ∂x ∂y
and thus the generators of y 00 = Ax−5 y 2 are given by
∂ ∂ ∂ ∂
Γ(1) = x2 + xy , Γ(2) = x + 3y (9.282)
∂x ∂y ∂x ∂y
101
5
Case 4. For n = 0 we may take a6 6=0, a2 = a3 = a4 = 0 and a1 = − a5 . We
2
thus have in this case form (9.271)
X = a5 x + a6
(9.283)
Y = −2a5 y
∂ ∂ ∂ ∂
Γ=X +Y = (a5 x + a6 ) + (−2a5 y)
∂x ∂y ∂x ∂y
∂ ∂ ∂
= a5 x − 2y + a6
∂x ∂y ∂x
and thus the generators of y 00 = Ay 2 are given by
∂ ∂ ∂
Γ(1) = x − 2y , Γ(2) = (9.284)
∂x ∂y ∂x
Example 3. We consider finally the equation
y 00 = F (x, y, y 0 ) (9.285)

P r(2) Γ(∆)|∆=0 = 0, ∆≡y 00 − F (x, y, y 0 ) (9.286)

∂ ∂ ∂ ∂
P r(2) Γ = X +Y + Y [x] 0 + Y [xx] 00 (9.287)
∂x ∂y ∂y ∂y
The infinitesimals are evaluated by the formulas (9.198) and (9.199) respec-
tively. We have in our case
P r(2) Γ(∆) = X(−Fx ) + Y (−Fy ) − Y [x] Fy0 + Y [xx] (9.288)
and in expanded form
P r(2) Γ(∆) = −XFx − Y Fy − [Yx + (Yy − Xx )y 0 − Xy (y 0 )2 ]Fy0

+ (Yy − 2Xx )y 00 − 3Xy y 0 y 00
102
The condition P r(2) Γ(∆)|∆=0 = 0 is derived from the above expression upon
substituting y 00 by F :
P r(2) Γ(∆)|∆=0 = 0 ⇐⇒ − XFx − Y Fy

− [Yx + (Yy − Xx )y 0 − Xy (y 0 )2 ]Fy0
(9.290)
+ Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3
+ (Yy − 2Xx )F − 3Xy F y 0 = 0
Equating to zero the coefficients of y 0 we obtain a system of equations from

which we can determine the infinitesimal coefficients X and Y . However
we cannot do that in this case since we don’t know the explicit form of the
function F (x, y, y 0 ).
9.9. Calculation of the First Integral I. The equation of oscillations.
In this Example we shall find the first integral of the equation for oscillations
u00 + u = 0, u = u(x) (9.291)
using the direct method. The first integral is a function
I = I(x, u, u0 ) (9.292)
which satisfies the equation
Dx I = 0, ∆ = u00 + u (9.293)
∆=0
where
∂ ∂ ∂
Dx = + u0 + u00 0 (9.294)
∂x ∂u ∂u
We have the equation
∂I ∂I ∂I
Dx I = + u0 + (−u) 0 = 0 (9.295)
∆=0 ∂x ∂u ∂u
where we have substituted u00 by −u to account for the condition ∆ = 0.

We make the ansatz
1
I = A(x, u)(u0 )2 + B(x, u)u0 + C(x, u) (9.296)
2
103
where the coefficients A(x, u), B(x, u) and C(x, u) will be determined by the
condition (9.295). Condition (9.295), because of (9.296) becomes

1 0 2 0 1
Ax (u ) + Bx u + Cx + Au (u ) + Bu u + Cu u0
0 2 0
2 2
0
+ (−u)(Au + B) = 0

1 0 3
1
Au (u ) + Ax + Bu (u0 )2
2 2 (9.297)
+ (Bu + Cu − Au)u0 + (Cx − Bu) = 0
Equating to zero the coefficients of the various powers of u0 we obtain the

following system of equations
Au = 0 (9.298)
1
Ax + Bu = 0 (9.299)
2
Bu + Cu − Au = 0 (9.300)
Cx − Bu = 0 (9.301)
Since Au = 0, we get that
A = A(x) (9.302)
1
and then equation (9.299) takes on the form Bu = − A0 (x) and by integra-
2
tion
1
B = − A0 (x)u + θ(x) (9.303)
2
Equations (9.300) and (9.301) then become

1 00
Cu = A (x) + A(x) u − θ0 (x) (9.304)
2
and
1
Cx = θ(x)u − A0 (x)u2 (9.305)
2
104
respectively. Differentiation of (9.304) with respect to x and (9.305) with
respect to u and equating the two derivatives, we get the equation

1 000
A (x) + A (x) u − θ00 (x) = θ(x) − A0 (x)u
0
(9.306)
2
From the above equation we obtain the two equations upon equating the
powers of u
1 000
A (x) + A0 (x) = −A0 (x) (9.307)
2
and
−θ00 (x) = θ(x) (9.308)
The last two equations admit the general solutions
A(x) = C1 + C2 sin(2x) + C3 cos(2x) (9.309)
and
θ(x) = C4 sin(x) + C5 cos(x) (9.310)
respectively. Equation (9.305) then takes on the form
Cx = [C4 sin(x) + C5 cos(x)]u − [C2 cos(2x) − C3 sin(2x)]u2
and by integration with respect to x
C(x, u) = [−C4 cos(x) + C5 sin(x)]u
1 (9.311)
− [C2 sin(2x) + C3 cos(2x)]u2 + η(u)
2
From equation (9.303), taking into account (9.309) and (9.310), we obtain
the following expression of B(x, u):
B(x, u) = [−C2 cos(2x) + C3 sin(2x)]u + C4 sin(x) + C5 cos(x) (9.312)
The expression (9.296) for I we have considered, because of (9.309), (9.312)
and (9.311) takes on the form
1
I = [C1 + C2 sin(2x) + C3 cos(2x)](u0 )2
2
+ {[−C2 cos(2x) + C3 sin(2x)]u + C4 sin(x) + C5 cos(x)}u0
(9.313)
+ [−C4 cos(x) + C5 sin(x)]u
1
− [C2 sin(2x) + C3 cos(2x)]u2 + η(u)
2
105
The above expression can also be written as
1
I = C1 (u0 )2 + η(u)
2
1 0 2 0 1 2
+ C2 (u ) sin(2x) − uu cos(2x) − u sin(2x)
2 2 (9.314)

1 0 2 0 1 2
+ C3 (u ) cos(2x) + uu sin(2x) − u cos(2x)
2 2
0 0
+ C4 [u sin(x) − ucos(x)] + C5 [u cos(x) + usin(x)]
The function η(u) appearing at the top of (9.314) is still undetermined. How-
1
ever the quantity ∆1 = C1 (u0 )2 + η(u) should satisfy the condition
2
Dx (∆1 ) =0 (9.315)
∆=0
i.e.
∂
0 ∂ ∂
+u + (−u) 0 ∆1 = 0 (9.316)
∂x ∂u ∂u
The above equation is equivalent to η 0 (u) = C1 u and by integration
1
η(u) = C1 u2 (9.317)
2
where we have neglected an irrelevant constant. Therefore the expression
(9.314) takes on the final form
1 1
I = C1 (u0 )2 + C1 u2
2 2
1 0 2 0 1 2
+ C2 (u ) sin(2x) − uu cos(2x) − u sin(2x)
2 2 (9.318)

1 0 2 0 1 2
+ C3 (u ) cos(2x) + uu sin(2x) − u cos(2x)
2 2
0 0
+ C4 [u sin(x) − ucos(x)] + C5 [u cos(x) + usin(x)]
The above expression gives rise to the following first integrals

1 1
I1 = (u0 )2 + u2 (9.319)
2 2
106
1 1 2
I2 = (u0 )2 sin(2x) − uu0 cos(2x) − u sin(2x) (9.320)
2 2
1 1 2
I3 = (u0 )2 cos(2x) + uu0 sin(2x) − u cos(2x) (9.321)
2 2
I4 = u0 sin(x) − u cos(x) (9.322)
I5 = u0 cos(x) + u sin(x) (9.323)
We should notice in turn that since the first integrals satisfies the equations
Dx Ii = 0, i = 1, 2, 3, 4, 5 (9.324)
we should have
Ii = Ci , i = 1, 2, 3, 4, 5 (9.325)
from which we can find the general solutions of the oscillation equation.
For example we can find a general solution of the equation u00 + u = 0 by
considering the equation I2 = C, i.e.
1 0 2 1
(u ) sin(2x) − uu0 cos(2x) − u2 sin(2x) = C (9.326)
2 2
9.10. Calculation of the First Integral II. The equation y 00 = 0.
We consider now the equation
y 00 = 0, y = y(x) (9.327)
This equation admits a first integral I = I(x, y, y 0 ) if
Dx (I) = 0, ∆ = y 00 (9.328)
∆=0
i.e.
∂ ∂
+ y0 I=0 (9.329)
∂x ∂y
Considering the ansatz
I = A(x, y)(y 0 )2 + B(x, y)y 0 + C(x, y) (9.330)
107
equation (9.329) takes on the form
Ax (y 0 )2 + Bx y 0 + Cx + y 0 {Ay (y 0 )2 + By y 0 + Cy } = 0 (9.331)
The above equation can be also written as
Ay (y 0 )3 + (Ax + By )(y 0 )2 + (Bx + Cy )y 0 + Cx = 0 (9.332)
from which we derive the system
Ay = 0, Ax + By = 0, Bx + Cy = 0, Cx = 0 (9.333)
Therefore
A = A(x), C = C(y) (9.334)
and thus
A0 (x) + By = 0, Bx + C 0 (y) = 0 (9.335)
Differentiation of By = −A0 (x) with respect to x and Bx = −C 0 (y) with

respect to y and equating the two derivatives, we obtain the equation A00 (x) =
C 00 (y) and thus A00 (x) = 0 and C 00 (y) = 0, i.e.
A(x) = a1 x + a2 and C(y) = c1 y + c2 (9.336)
Therefore
By = −a1 and Bx = −c1 (9.337)
and by integration
B = −a1 y + (x) and B = −c1 x + ζ(y) (9.338)
Consistency requires −a1 y + (x) = −c1 x + ζ(y) and thus (x) = −c1 x and
ζ(y) = −a1 y. Therefore
B = −c1 x − a1 y (9.339)
Using the expressions for A, B and C and taking into account (9.330), we
obtain
I = (a1 x + a2 )(y 0 )2 + (−c1 x − a1 y)y 0 + c1 y + c2 (9.340)
108
The above expression can be also written as
I = a1 {x(y 0 )2 − yy 0 } + a2 (y 0 )2 + c1 (y − xy 0 ) + c2 (9.341)
or even
y 0 2
I = a1 x − 0 (y ) + a2 (y 0 )(y 0 ) + c1 (y − xy 0 ) (9.342)
y
ignoring the irrelevant constant c2 . We thus get the following first integrals
y
I1 = x − , I2 = y 0 , I3 = y − xy 0 (9.343)
y0
9.11. Table of Differential Equations admitting known Symmetries.
We list below some second-order differential equations with their correspond-
ing symmetry generators. The reader is advised to check, if not all, at least
some of them.

∂
1 y 00 = F (y, y 0 ) X=
∂x
∂
2 y 00 = F (x, y 0 ) X=
∂y
∂ ∂
3 y 00 = F (αx + βy, y 0 ) X=β −α
∂x ∂y
p y − xy 0 ∂ ∂
4 y 00 = [1 + (y 0 )2 ]3/2 F x2 + y 2 , X=y −x
x + yy 0 ∂x ∂y
y − xy 0 ∂
5 y 00 = F y, X=y
y0 ∂x
∂
6 y 00 = F (x, y − xy 0 ) X=x
∂y
y ∂ ∂
7 xy 00 = F , y0 X=x +y
x ∂x ∂y
y0 ∂
8 y 00 = yF x, X=y
y ∂x
109
y0 ∂ ∂
9 y 00 = yF ye−x , X= +y
y ∂x ∂y
y ∂ ∂
10 x3 y 00 = F , y − xy 0 X = x2 + xy
x ∂x ∂y
y y − xy 0 ∂ ∂
11 x3 y 00 = (y 0 )3 F , X = xy + y2
x y0 ∂x ∂y
y ∂
00 0 2 0 3
12 xy + y = x (y ) F y, 0 − ln(x) X = xy
xy ∂x
xy 0 ∂
13 yy 00 = (y 0 )2 + y 2 F x, − ln(y) X = xy
y ∂y
∂ ∂
14 y 00 = xm−2 F (x−m y, x1−m y 0 ) X=x + my
∂x ∂y
p0 (y) 0 2 y0 ∂
15 y 00 = (y ) + p(y)F x, X = p(y)
p(y) p(y) ∂y
∂
16 p(x)y 00 − p00 (x)y = F (x, p(x)y 0 − p0 (x)y) X = p(x)
∂y
∂
17 p2 (x)y 00 + p(x)p0 (x)y 0 = F (y, p(x)y 0 ) X = p(x)
∂x
p00 (y) 1 p0 (y) ∂
18 y 00 + x(y 0 )3 = (y 0 )3 F y, 0 − x X = p(y)
p(y) y p(y) ∂x
10 Integrating Factors and Linearization

It is known from the integration methods of first order differential equations,
that in most cases we are looking for an integrating factor in order to convert
the equation into an exact one. This is also the case for second order ordi-
nary differential equations. We first introduce the concept of differential
functions.
10.1. The Space A of Differential Functions. Let z be the sequence
z = (x, u, u(1) , u(2) , · · · )
with elements z ν (ν≥1) where
z i = xi (1≤i≤n), z n+α = uα (1≤α≤m)
110
and the remaining elements represent the derivatives of u. Finite subse-
quences of z will be denoted by [z].
A differential function f is a locally analytic function f ([z]) of a finite
number of variables, i.e. locally expendable in a Taylor series with respect
to all arguments.
The highest order of derivatives appearing in a differential function f is called
the order of f and will be denoted by ord(f ). Thus ord(f ) = s means that
f = f (x, u, u(1) , · · · , u(s) )
The set of all differential functions of finite order will be denoted by A. The
set A is a vector space endowed with the usual multiplication of functions.
In other words, if
f ([z])∈A, g([z])∈A
and if α, β are any constants, then
αf + βg ∈A, ord(αf + βg)≤max{ord(f ), ord(g)}

f g ∈A, ord(f g) = max{ord(f ), ord(g)}
The space A is also closed under the total derivation, i.e.
if f ∈A, then Di (f )∈A, ord(Di (f )) = ord(f ) + 1
Lemma 1. Let f (x, y, y 0 , · · · , y (s) ) ∈A. If Dx (f ) = 0 identically for all

x, y, y 0 , · · · , y (s) , then f = C = constant. Likewise, if f (x, u, u(1) , · · · , u(s) )
is a differential function with one independent variable x and several depen-
dent variables u = (u1 , u2 , · · · , um ), the equation Dx (f ) = 0 implies that
f = C = constant.
Lemma 2. A differential function f (x, u, u(1) , · · · , u(s) )∈A with one inde-
pendent variable x is a total derivative
f = Dx (g), g(x, u, u(1) , · · · , u(s−1) ) ∈A
if and only if the equations

δf
= 0, α = 1, 2, · · · , m
δuα
111
hold identically in x, u, u(1) , · · · .
δ
Here is the variational derivative (the so-called Euler-Lagrange op-
δuα
erator) defined by
δ ∂ ∂ ∂
α
= α
− Di α + Di Dj α + · · · α = 1, 2, · · · , m
δu ∂u ∂ui ∂uij
10.2. Second Order ODEs. In this Section we shall consider second order
ordinary differential equations of the form
A(x, y, y 0 )y 00 + B(x, y, y 0 ) = 0 (10.1)
where x is the independent variable and y is the unknown function to be
determined. The prime denotes the derivative of the unknown function with
respect to x. The double prime is the second order derivative with respect
to x:
0 dy d2 y
00
y = y(x), y = , y = 2 (10.2)
dx dx
An integrating factor of equation (10.1) is any function µ(x, y, y 0 ) such
that
µA(x, y, y 0 )y 00 + µB(x, y, y 0 ) = Dx Φ(x, y, y 0 ) = 0 (10.3)
for some function Φ(x, y, y 0 ), called first integral. The symbol Dx is the
generalized derivative
∂ ∂ ∂
Dx = + y0 + y 00 0 + · · · (10.4)
∂x ∂y ∂y
Once we determine an integrating factor, we transform the equation into the
equation Dx Φ(x, y, y 0 ) = 0 from which we determine a first integral
Φ(x, y, y 0 ) = C (10.5)
We first consider an s−order ordinary differential equation of the form
A y (s) + B = 0 (10.6)
where
A = A(x, y, y 0 , y 00 , · · ·, y (s−1) )
(10.7)
B = B(x, y, y 0 , y 00 , · · ·, y (s−1) )
112
We are looking for an integrating factor µ with
µ = µ(x, y, y 0 , y 00 , · · ·, y (s−1) ) (10.8)
such that
µA y (s) + µB = Dx Φ = 0 (10.9)
with
Φ = Φ(x, y, y 0 , y 00 , · · ·, y (s−1) ) (10.10)
We have the following

Theorem 1. The integrating factors of equation (10.6) are determind by
the equation
δ
(µAy (s) + µB) = 0 (10.11)
δy
δ
where is the variational derivative defined by
δy
δ ∂ ∂ ∂ ∂
= − Dx 0 + Dx2 00 − Dx3 00 + · · · (10.12)
δy ∂y ∂y ∂y ∂y
Proof. First of all, equation (10.11) is a consequence of Lemma 2. The

δ
highest derivative that appears after the variational differentiation has
δu
the order 2s − 1. It occurs in the terms

s s s−1 s−1 (s) ∂(µA)
(−1) Dx (µA) and (−1) Dx y (10.13)
∂y (s−1)
We have, dropping the terms which do not contain y (2s−1) :

s s s−1 s−1 (s) ∂(µA)
(−1) Dx (µA) = −(−1) Dx y (10.14)
∂y (s−1)
We thus see that the terms containing y (2s−1) annihilate each other and hence
equation (10.11) involves only the variables x, y, y 0 , · · ·, y (2s−2) .
For the second order ODEs we have the following
113
Theorem 2. The integrating factors of equation (10.1) are determind by
the equation
δ
(µAy 00 + µB) = 0 (10.15)
δy
δ
where is the variational derivative defined in (10.12).
δy
Using the above Theorem 2, we arrive at the following
Theorem 3: The integrating factors µ(x, y, y 0 ) for the second order equation
(10.1) are determined by the system of the two equations
y 0 (µA)yy0 + (µA)xy0 + 2(µA)y − (µB)y0 y0 = 0 (10.16)
and
(y 0 )2 (µA)yy + 2y 0 (µA)xy + (µA)xx − y 0 (µB)yy0
(10.17)
− (µB)xy0 + (µB)y = 0
The above is an overdetermined system of second order PDEs which should

be compatible in order to determine the integrating factors. However this
system may have no solutions at all.
Proof. We have that
δ
(µAy 00 + µB)
δy
∂ ∂ ∂ (10.18)
= − Dx 0 + Dx2 00 (µAy 00 + µB)
∂y ∂y ∂y
= y (µA)y + (µB)y − Dx [y 00 (µA)y0 + (µB)y0 ] + Dx2 (µA)
00
Let us now calculate the total derivatives appearing in the above expression.
We obtain
∂
0 ∂ 00 ∂ 000 ∂

00
Dx [y (µA)y0 ] = +y +y 0
+y 00
(y 00 (µA)y0 )
∂x ∂y ∂y ∂y (10.19)
00 0 00 00 2 000
= y (µA)xy0 + y y (µA)yy0 + (y ) (µA)y0 y0 + y (µA)y0
∂ ∂ ∂
Dx [(µB)y0 ] = + y0 + y 00 0 ((µB)y0 )
∂x ∂y ∂y (10.20)
= (µB)xy0 + y (µB)yy0 + y 00 (µB)y0 y0
0
114
In order to calculate Dx2 (µA) we calculate Dx (µA) first. We have
∂
0 ∂ 00 ∂

Dx (µA) = +y +y (µA)
∂x ∂y ∂y 0 (10.21)
= (µA)x + y 0 (µA)y + y 00 (µA)y0
and then
Dx2 (µA) = Dx (Dx (µA))
∂
0 ∂ 00 ∂ 000 ∂

0 00
= +y +y +y (µA)x + y (µA)y + y (µA)y0
∂x ∂y ∂y 0 ∂y 00
= (µA)xx + y 0 (µA)xy + y 00 (µA)xy0
(10.22)
+ y 0 {(µA)xy + y 0 (µA)yy + y 00 (µA)yy0 }
+ y 00 {(µA)xy0 + y 0 (µA)yy0 + (µA)y + y 00 (µA)y0 y0 } + y 000 (µA)y0
= (µA)xx + 2y 0 (µA)xy + (y 0 )2 (µA)yy + y 00 (µA)y + 2y 00 (µA)xy0
+ 2y 0 y 00 (µA)yy0 + (y 00 )2 (µA)y0 y0 + y 000 (µA)y0
Collecting everything together, we obtain
δ
(µAy 00 + µB)
δy
= y 00 {y 0 (µA)yy0 + (µA)xy0 + 2(µA)y − (µB)y0 y0 } (10.23)
+ (y 0 )2 (µA)yy + 2y 0 (µA)xy + (µA)xx − y 0 (µB)yy0
− (µB)xy0 + (µB)y = 0
The above expression should vanish for every value of x, y, y 0 and y 00 . We
thus arrive at (10.16) and (10.17) respectively.
Example. Find the general solution of the equation
(y 0 )2 y0
y 00 + +3 =0 (10.24)
y x
by determining the integrating factors first.
Solution. In our case we have
(y 0 )2 y0
A = 1 and B = +3 (10.25)
y x
We are looking for integrating factors of the form µ(x, y). In this case equa-
tion (10.16) takes on the form
∂µ µ
− =0 (10.26)
∂y y
115
The above solution admits a general solution given by
µ = f (x)y (10.27)
Upon substitution of the above in equation (10.17) we obtain the equation
x2 f 00 − 3xf 0 + 3f = 0 (10.28)
The above is an Euler type of equation with two linearly independent so-
lutions f1 = x and f2 = x3 . We thus have the following two integrating
factors
µ1 = xy and µ2 = x3 y (10.29)
for the given equation. Multiplying the original equation by µ1 = xy we get

the equation
xyy 00 + x(y 0 )2 + 2yy 0 = 0 (10.30)
Since xyy 00 + x(y 0 )2 + 2yy 0 = Dx (xyy 0 + y 2 ), we obtain that Dx (xyy 0 + y 2 ) = 0

and thus
xyy 0 + y 2 = C1 (10.31)
Multiplying equation (10.24) by the second integrating factor and integrating

of the resulting equation, we get
x3 yy 0 = C2 (10.32)
Upon elimination of y 0 between equations (10.31) and (10.32), we obtain the

C2
equation 2 + y 2 = C1 and then solving with respect to y, we obtain the
x
general solution of the equation (10.24):
r
C2
y = ± C1 − 2 (10.33)
x
References
[1] N. H. Ibragimov: ”Integrating Factors, Adjoint Equations and
Lagrangians”. J. Math. Anal. Appl. 318 (2006) 742-757
[2] N. H. Ibragimov: ”Elementary Lie Group Analysis and Ordinary
116
Differential Equations”. Wiley 1999
10.3. Linearization
Linearization is the process of converting a second order ODE
y 00 = F (x, y, y 0 ) (10.34)
to the equation
d2 u
u00 ≡ =0 (10.35)
dt2
which is called the Forsyth-Laguerre form. Let us suppose that there is
a transformation
t = φ(x, y)
(10.36)
u = ψ(x, y)
which transforms (10.35) into (10.34). The form of the equation (10.34)
obtained from (10.35) by the transformation (10.36) was derived for the first
time by S. Lie (Lie [1]) and is given by
y 00 + A(x, y)(y 0 )3 + B(x, y)(y 0 )2 + C(x, y)y 0 + D(x, y) = 0 (10.37)
where the coefficients A(x, y), B(x, y), C(x, y) and D(x, y) are expressed in
terms of the functions φ(x, y) and ψ(x, y). Therefore the 2nd order ODE
under linearization should be at most of third order in y 0 .
The first derivative u0 transforms under (10.36) into
du Dx ψ ψx + y 0 ψy
u0 = = = = g(x, y, y 0 ) (10.38)
dt Dx φ φx + y 0 φy
The second derivative u00 transforms in a similar fashion

d2 u Dx g gx + y 0 gy + y 00 gy0
u00 = = = = P (x, y, y 0 , y 00 ) (10.39)
dt2 Dx φ φx + y 0 φy
Using the definition of g in (10.38), we get
(φx + y 0 φy )(ψxx + y 0 ψxy ) − (ψx + y 0 ψy )(φxx + y 0 φxy )

gx = (10.40)
(φx + y 0 φy )2
117
(φx + y 0 φy )(ψxy + y 0 ψyy ) − (ψx + y 0 ψy )(φxy + y 0 φxy )
gy = (10.41)
(φx + y 0 φy )2
and
(φx + y 0 φy )ψy − (ψx + y 0 ψy )φy
gy 0 = (10.42)
(φx + y 0 φy )2
Therefore we have
gx + y 0 gy + y 00 gy0
(φx + y 0 φy )(ψxx + y 0 ψxy ) − (ψx + y 0 ψy )(φxx + y 0 φxy )
=
(φx + y 0 φy )2
(φx + y 0 φy )(ψxy + y 0 ψyy ) − (ψx + y 0 ψy )(φxy + y 0 φxy ) (10.43)
+ y0
(φx + y 0 φy )2
(φx + y 0 φy )ψy − (ψx + y 0 ψy )φy
+ y 00
(φx + y 0 φy )2
Expanding the numerators we obtain
gx + y 0 gy + y 00 gy0 = (φx + y 0 φy )−2 {(φx ψy − φy ψx )y 00

+ (φy ψyy − φyy ψy )(y 0 )3 + [φx ψyy − φyy ψx + 2(φy ψxy − φxy ψy )](y 0 )2 (10.44)
+ [φy ψxx − φxx ψy + 2(φx ψxy − φxy ψx )]y 0 + (φx ψxx − φxx ψx )}
Factoring out the term (the coefficient of y 00 )
∆≡φx ψy − φy ψx , ∆6=0 (10.45)
which is the Jacobian of the transformation (10.36), we can write (10.44) as
gx + y 0 gy + y 00 gy0 = (φx + y 0 φy )−2 ∆{y 00 + A(x, y)(y 0 )3

(10.46)
+ B(x, y)(y 0 )2 + C(x, y)y 0 + D(x, y)}
where we have introduced the notation

φy ψyy − φyy ψy
A(x, y) = (10.47)
∆
φx ψyy − φyy ψx + 2(φy ψxy − φxy ψy )
B(x, y) = (10.48)
∆
φy ψxx − φxx ψy + 2(φx ψxy − φxy ψx )
C(x, y) = (10.49)
∆
118
φx ψxx − φxx ψx
D(x, y) = (10.50)
∆
Taking into account (10.39) and (10.46), we see that
u00 = (φx + y 0 φy )−3 ∆{y 00 + A(x, y)(y 0 )3

(10.51)
+ B(x, y)(y 0 )2 + C(x, y)y 0 + D(x, y)}
and thus u00 = 0 gets transformed into
y 00 + A(x, y)(y 0 )3 + B(x, y)(y 0 )2 + C(x, y)y 0 + D(x, y) = 0 (10.52)
The problem we have to face now is the following: How we could possibly
determine the functions φ and ψ of the transformation (10.36) once we know
the equation (10.52). We shall distinguish two cases:
Case I. Let φy = 0. This is called a fiber-preserving point transformation.
In this case we obtain from (10.47)-(10.50)
A=0
ψyy = ψy B
(10.53)
2ψxy = (φx )−1 ψy φxx + C
ψxx = (φx )−1 ψx φxx + ψy D
Comparing the mixed derivatives (ψxy )y = (ψyy )x and (ψxy )x = (ψxx )y we

get
Cy = 2Bx
(10.54)
(φx )−2 (2φx φxxx − 3φ2xx ) = 4(Dy + BD) − (2Cx + C 2 )
Since φy = 0, differentiation of the last equation with respect to y yields
Dyy − Bxx − Bx C + By D + Dy B = 0 (10.55)
We thus see that a second order ODE of the form (10.52) gets linearised
by the transformation t = φ(x), u = ψ(x, y) if the coefficients satisfy the
conditions
A=0
Cy = 2Bx (10.56)
Dyy − Bxx − Bx C + By D + Dy B = 0
119
Provided that the above conditions are satisfied, the functions φ(x) and
ψ(x, y) are determined by solving the system (10.53)-(10.54).
In this case we also have the following (Suksern and Sawatdithep [3])
Theorem 1. Equation (10.52) is linearizable under the point transformation
t = φ(x), u = ψ(x, y) (10.57)
if and only if its coefficients satisfy the conditions
A = 0, Cy = 2Bx , Dyy − Bxx − Bx C + By D + Dy B = 0 (10.58)
If conditions (10.58) are satisfied, the linearizing transformation (10.57) can

be obtained by solving the system
φy = 0, ψyy = ψy B (10.59)
The coefficients of the linearized equation
u00 + α(t)u0 + β(t)u + γ(t) = 0 (10.60)
are evaluated using the formulas

φxx ψy − 2φx ψxy + φx ψy C
α=
φ2x ψy
−φ2x ψy γ + 2ψxy ψx − ψxxψy − ψx ψy C + ψy2 D
β=
φ2x ψy ψ
γ1 + γ2 (10.61)
γ= 2 2
φx ψy
γ1 = −Dy ψy2 ψ − 2ψxy
2
ψ + 2ψxy ψx ψy + ψxy ψy ψC
γ2 = ψxxyψy ψ − ψxx ψy2 − ψx ψy2 C + ψy3 D − ψy2 ψBD
Case II. Let φy 6=0. In this case we get the following from (10.47)-(10.50)
φy ψyy = φyy ψy + A∆
2φ2y ψyy = 2φxy φy ψy − φyy ∆ − (Aφx − Bφy )∆
φ2y ψxx = 2φxy φy ψx − φx φxy ψx − φ2x ψx A + φx φy ψx B (10.62)
+ φ2y (ψy D − ψx C)
φ2y φxx = 2φxy φx φy − φ2x φyy − φ3x A + φ2x φy B − φ2y φx C + φ3y D
120
Using the relations (ψxy )y = (ψyy )x and (ψxy )x = (ψxx )y we get
2φy φyyy = 3(φ2yy − 2φxy φy A + 2φx φyy A + φ2x A2 )

(10.63)
− 2φx φy (Ay + AB) + φ2y (2By − 4Ax + 4AC − B 2 )
and
6φ2y φxyy = 3(4φxy φyy φy − φx φ2yy + 2φx φyy φy B − 2φxy φ2y B)
+ 3φ3x A2 + 3φx φ2y (−2Ax + 2AC − B 2 ) (10.64)
+ 2φ3y (−Bx + 2Cy + 3AD)
As a result of the relations between the mixed derivatives (φxyy )y = (φyyy )x

and (φxx )yy = (φxyy )x we get
H ≡ 3Axx − 2Bxy + Cyy − 3Ax C + 3Ay D + 2Bx B

(10.65)
− 3Cx A − Cy B + 6Dy A = 0
and
K ≡ Bxx − 2Cxy + 3Dyy − 6Ax D + Bx C + 3By D
(10.66)
− 2Cy C − 3Dx A + 3Dy B = 0
respectively.
We also have the following (Suksern and Sawatdithep [3])
Theorem 2. Equation (10.52) is linearizable under the point transformation
t = φ(x, y), u = ψ(x, y) (10.67)
if and only if its coefficients satisfy the conditions
H = 0, K=0 (10.68)
If conditions (10.68) are satisfied, the linearizing transformation (10.67) can

be obtained by solving the system (10.62). The coefficients of the linearized
equation
u00 + α(t)u0 + β(t)u + γ(t) = 0 (10.69)
121
are evaluated using the formulas
−2∆y φy + 3φx A∆ + 3φyy ∆ − φy B∆
α=
φ2y ∆
1
β = 3 2 {−Ax φy ∆2 + AAy φx φy ∆3 ∆x − ∆yy φy ∆ + 2∆2y φy
φy ∆
(10.70)
− 2∆y φx A∆ − 2∆y φyy ∆ + ∆y φy B∆
+ 3φxy A∆2 + φyyy ∆2 − φyy B∆2 }
φyy ψy − φ3y βψ − φ2y ψy α − φy ψyy + A∆
γ=
φ3y
Example 1. We shall solve the equation
y 00 + y(y 0 )2 = 0 (10.71)
using the linearization process. In this equation we have
A = 0, B = y, C = 0, D=0 (10.72)
comparing the given equation to (10.52). We can verify that the conditions
(10.56) are satisfied. Therefore the system (10.59) yields
φy = 0, ψyy = ψy y (10.73)
One solution of the above system may be taken as

Z
2
φ = x, ψ = ey /2 dy (10.74)
We thus see that

Z
2 /2
t = x, u= ey dy (10.75)
On the othe hand we find, using (10.61), that
α = 0, β = 0, γ=0 (10.76)
Therefore equation (10.71) under the transformation (10.75) transforms into

the equation u00 = 0 with general solution
u = C1 t + C2 (10.77)
122
from which we get, using (10.74)
Z
2
ey /2 dy = C1 x + C2 (10.78)

y
erfi √ = C1 x + C2 (10.79)
2
where erfi(z) is the error function of imaginary argument defined by
Zz
2 2
erfi(z) = √ ey dy (10.80)
π 0
Example 2. We solve next the equation
y 00 − 2(y 0 )2 − e2y = 0 (10.81)
A = 0, B = −2, C = 0, D = −e2y (10.82)
comparing the given equation to (10.52). We can verify that the conditions
(10.56) are satisfied. Therefore the system (10.59) yields
φy = 0, ψyy = −2ψy (10.83)

1
φ = x, ψ = − e−2y (10.84)
2
Using (10.61), we find
α = 0, β = 0, γ = −1 (10.85)

the equation u00 − 1 = 0 with general solution
1
u = t2 + C1 t + C2 (10.86)
2
123
which, because of (10.84), takes on the form
1 1
− e−2y = x2 + C1 x + C2 (10.87)
2 2
e−2y + x2 = C1 x + C2 (10.88)
Example 3. We solve finally the equation
y 00 − 2y(y 0 )3 = 0 (10.89)
A = −2y, B = 0, C = 0, D=0 (10.90)
comparing the given equation to (10.52). For this equation we expect φy 6=0
since A = −2y6=0. We can verify that conditions (10.65) and (10.66) are
satisfied. Then the functions φ and ψ are determined as solutions of the
system of the following four equations (compare to the system (10.62))
φx ψy − ∆
ψx = (10.91)
φy
φ2y ∆x = 3∆(φy φxy + 2yφ2x − φx φyy ) (10.92)

φ2y φxx
2φxy φy + 2φ2x y − φx φyy = (10.93)
φx
φ2y φxyy = −6yφxy φx φy + 2φxy φyy φy − 4y 2 φ3x + 6yφ2x φxy
(10.94)
− 2φ2x φy + φx φy φyyy − 2φx φ2yy
φ = y, ψ = −x (10.95)
which means that the linearizing transformation is
t = y, u = −x (10.96)
Using (10.70), we find
α = 0, β = 0, γ = −2t (10.97)
124
the equation u00 − 2t = 0 with general solution
1
u = t3 + C1 t + C2 (10.98)
3
which, because of (10.96), takes on the form
1
−x = y 3 + C1 y + C2 (10.99)
3
3x + y 3 + C1 y + C2 = 0 (10.100)
References
[1] S. Lie: ”Klassifikation und integration von gewöhnlichen differential-
gleichungen zwischen x, y die eine gruppe von transformationen gestatten
III”. Archiv for Matematik og Naturvidenskab 8(4) (1883) 371-427
[2] S. V. Meleshko: ”Methods of Constructing Exact Solutions of Partial
Differential Equations”. Springer 2005
[3] S. Suksern and C. Sawatdithep: ”Reduction of Second-Order
Ordinary Differential Equations into General Linear Equations via Point
Transformations and its Applications”.
Appl. Math. Inf. Sci. 14 (2020) 249-259.
11 Third Order Ordinary Differential

Equations. Linearization.
Higher Order ODEs
In this Section we consider third and higher order infinitesimals and invari-
ance of third and higher order equations.
11.1. Invariance and Infinitesimals for Third Order ODEs.
Invariance of the third order differential equation
∆≡F (x, y, y 0 , y 00 , y 000 ) = 0 (11.1)
by infinitesimal point symmetries
x̂ = x + X + O(2 ), ŷ = y + Y + O(2 ) (11.2)
125
is being expressed by
P r(3) Γ(∆)|∆=0 = 0 (11.3)
∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (11.4)
∂x ∂y ∂y ∂y ∂y
The infinitesimals are evaluated by the formulas
Y [x] = Dx (Y ) − y 0 Dx (X) (11.5)
Y [xx] = Dx (Y [x] ) − y 00 Dx (X) (11.6)
Y [xxx] = Dx (Y [xx] ) − y 000 Dx (X) (11.7)
where the total derivative operator Dx is defined by
∂ ∂ ∂ ∂
Dx = + y0 + y 00 0 + y 000 00 (11.8)
∂x ∂y ∂y ∂y
It is obvious that we have the following dependence of X and Y
X = X(x, y) and Y = Y (x, y) (11.9)
We can derive the following explicit expressions for the infinitesimals:
∂ ∂ ∂ ∂
Y [x] = Dx (Y ) − y 0 Dx (X) = + y0 Y − y0 + y0 X
∂x ∂y ∂x ∂y (11.10)
0 0 0 0 0 2
= Yx + y Yy − y (Xx + y Xy ) = Yx + (Yy − Xx )y − Xy (y )
For the infinitesimal Y [xx] we obtain
Y [xx] = Dx (Y [x] ) − y 00 Dx (X)
∂
0 ∂ 00 ∂

0 0 2
= +y +y Yx + (Yy − Xx )y − Xy (y )
∂x ∂y ∂y 0
∂ ∂
− y 00 + y0 X
∂x ∂y

= Yxx + (Yxy − Xxx )y 0 − Xxy (y 0 )2 (11.11)

0 0 0 2
+ y Yxy + (Yyy − Xxy )y − Xyy (y )

+ y 00 (Yy − Xx ) − 2Xy y 0 − y 00 (Xx + y 0 Xy )
= Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2 − Xyy (y 0 )3
+ (Yy − 2Xx )y 00 − 3Xy y 0 y 00
126
We have finally for the infinitesimal Y [xxx] :
Y [xxx] = Dx (Y [xx] ) − y 000 Dx (X)
∂
0 ∂ 00 ∂ 000 ∂

= +y +y 0
+y 00
Yxx + (2Yxy − Xxx )y 0
∂x ∂y ∂y ∂y

0 2 0 3 00 0 00
+ (Yyy − 2Xxy )(y ) − Xyy (y ) + (Yy − 2Xx )y − 3Xy y y
∂
0 ∂

000
−y +y X
∂x ∂y
= Yxxx + (2Yxxy − Xxxx )y 0 + (Yxyy − 2Xxxy )(y 0 )2 − Xxyy (y 0 )3
+ (Yxy − 2Xxx )y 00 − 3Xxy y 0 y 00

+ y 0 Yxxy + (2Yxyy − Xxxy )y 0 + (Yyyy − 2Xxyy )(y 0 )2

− Xyyy (y 0 )3 + (Yyy − 2Xxy )y 00 − 3Xyy y 0 y 00

00 0 0 2 00
+ y (2Yxy − Xxx ) + 2(Yyy − 2Xxy )y − 3Xyy (y ) − 3Xy y

+ y 000 (Yy − 2Xx ) − 3Xy y 0 − y 000 (Xy + y 0 Xy )
After collecting everything together and rearranging, we obtain

Y [xxx] = Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2
+ (Yyyy − 3Xxyy )(y 0 )3 − Xyyy (y 0 )4 + 3(Yxy − Xxx )y 00
(11.12)
+ 3(Yyy − 3Xxy )y 0 y 00 − 6Xyy (y 0 )2 y 00 − 3Xy (y 00 )2
+ (Yy − 3Xx )y 000 − 4Xy y 0 y 000
11.2. Example 1. The Blasius equation. Explore the symmetries of the
differential equation
1
y 000 + yy 00 = 0 (11.13)
2
and find the general solutions.
Solution. Let us denote by ∆ the quantity
1
∆≡y 000 + yy 00 (11.14)
2
If P r(3) Γ is the operator
∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (11.15)
∂x ∂y ∂y ∂y ∂y
127
then invariance of the equation is being expressed by
P r(3) Γ(∆) = 0
∂ ∂ [x] ∂ [xx] ∂ [xxx] ∂

⇔ X +Y +Y +Y +Y (∆) = 0 (11.16)
∂x ∂y ∂y 0 ∂y 00 ∂y 000
1 1
⇔ Y y 00 + Y [xx] y + Y [xxx] = 0
2 2
Upon substituting the infinitesimals Y [xx] and Y [xxx] into the above equation
we derive the equation
1 00 1
Y y + y Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2
2 2
− Xyy (y 0 )3 + (Yy − 2Xx )y 00 − 3Xy y 0 y 00
+ Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2 (11.17)
+ (Yy − 3Xx )y 000 − 4Xy y 0 y 000 = 0
We now have to take into account the condition ∆ = 0. We thus substitute

1
y 000 by − yy 00 in (11.17). We then obtain the equation
2
1 00 1
Y y + y Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2
2 2
− Xyy (y 0 )3 + (Yy − 2Xx )y 00 − 3Xy y 0 y 00
+ Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2 (11.18)
1 1
+ (Yy − 3Xx )(− yy 00 ) − 4Xy y 0 (− yy 00 ) = 0
2 2
128
The last equation can be written as
1
− Xyyy (y 0 )4 + (Yyyy − 3Xxyy − yXyy )(y 0 )3
2
1
+ [ y(Yyy − 2Xxy ) + 3(Yxyy − Xxxy )](y 0 )2
2
1
+ [ y(2Yxy − Xxx ) + (3Yxxy − Xxxx )]y 0 − 3Xy (y 00 )2
2 (11.19)
1 1
+ [3(Yxy − Xxx ) + yXx + Y ]y 00
2 2
1 0 00
+ [ yXy + 3(Yyy − 3Xxy )]y y
2
1
− 6Xyy (y 0 )2 y 00 + ( yYxx + Yxxx ) = 0
2
Equating the coefficients of the derivatives of the above equation to zero, we
obtain the following system of the determining equations
−Xyyy = 0 (11.20)
1
Yyyy − 3Xxyy − yXyy = 0 (11.21)
2
1
y(Yyy − 2Xxy ) + 3(Yxyy − Xxxy ) = 0 (11.22)
2
1
y(2Yxy − Xxx ) + (3Yxxy − Xxxx ) = 0 (11.23)
2
−3Xy = 0 (11.24)
1 1
3(Yxy − Xxx ) + yXx + Y = 0 (11.25)
2 2
1
yXy + 3(Yyy − 3Xxy ) = 0 (11.26)
2
−6Xyy = 0 (11.27)
1
yYxx + Yxxx = 0 (11.28)
2
We are going to solve the above system (11.20)-(11.28) of the determining
equations. From (11.24) we see that X is independent of y. Therefore
X = c(x) (11.29)
129
and thus the system is equivalent to
Yyyy = 0 (11.30)
1
yYyy + 3Yxyy = 0 (11.31)
2
1
y(2Yxy − c00 (x)) + (3Yxxy − c000 (x)) = 0 (11.32)
2
1 1
3(Yxy − c00 (x)) + yc0 (x) + Y = 0 (11.33)
2 2
Yyy = 0 (11.34)
1
yYxx + Yxxx = 0 (11.35)
2
Since Yyy = 0, we have
Y = a(x)y + b(x) (11.36)
Because of that, equations (11.30) and (11.31) are satisfied automatically.

Equation (11.35), because of (11.36) can be written as
1 00
y[a (x)y + b00 (x)] + a000 (x)y + b000 (x) = 0
2
1 00 1
a (x)y 2 + [ b00 (x) + a000 (x)]y + b000 (x) = 0 (11.37)
2 2
1 00
a00 (x) = 0, b (x) + a000 (x) = 0, b000 (x) = 0
2
a00 (x) = 0, b00 (x) = 0 (11.38)
Equation (11.32) gives

1
y[2a0 (x) − c00 (x)] + 3a00 (x) − c000 (x) = 0
2
130
from which we get the system
2a0 (x) − c00 (x) = 0, c000 (x) = 0 (11.39)
taking also into account the first of (11.38).

1 1
3[a0 (x) − c00 (x)] + yc0 (x) + [a(x)y + b(x)]y = 0
2 2
from which we obtain the system
1
3[a0 (x) − c00 (x)] + b(x) = 0, c0 (x) + a(x) = 0 (11.40)
2
From (11.38) we have
a(x) = a1 x + a2 , b(x) = a3 x + a4 (11.41)
From (11.39) we have c00 (x) = 2a1 and by integration
c(x) = a1 x2 + a5 x + a6 (11.42)
From c0 (x) + a(x) = 0 we get c0 (x) = −a1 x − a2 and by integration

1
c(x) = − a1 x2 − a2 x + a7 (11.43)
2
Upon comparing (11.42) and (11.43) we have a1 = 0, a5 = −a2 and a6 = a7 .
Therefore
X = c(x) = −a2 x + a6 and a(x) = a2 (11.44)
From the first of (11.40), i.e. 3[a0 (x) − c00 (x)] + 12 b(x) = 0, because of (11.44),
we have b(x) = 0. Therefore
Y = a2 y (11.45)
After changing names −a2 →c1 and a6 →c2 , we have the following values for
the infinitesimals X and Y :
X = c1 x + c2 , Y = −c1 y (11.46)
131
We thus have the following expression for the operator Γ:
∂ ∂ ∂ ∂
Γ=X +Y = (c1 x + c2 ) − c1 y (11.47)
∂x ∂y ∂x ∂y
or
∂ ∂ ∂ ∂ ∂
Γ=X +Y = c1 x −y + c2 (11.48)
∂x ∂y ∂x ∂y ∂x
Therefore the original equation admits two generators of symmetry
∂ ∂ ∂
Γ1 = x −y , Γ2 = (11.49)
∂x ∂y ∂x
Upon changing variables (x, y)→(r, s) using the system (Γ2 r = 0, Γ2 s = 1),
we have to solve the system of PDEs
1·rx + 0·ry = 0, 1·sx + 0·sy = 1 (11.50)
The general solution of the first equation in (11.50) is given by F (y, r) = 0

or r = f (y). The general solution of the second equation in (11.50) is given
by G(y, s − x) = 0 or s = x + g(y). We thus may take
r = y, s=x (11.51)
Considering the transformation
x≡a(r, s) = s, y≡b(r, s) = r (11.52)
and Appendices A and B, Blasius equation (11.13), transforms into
ds d3 s ds 2 d2 s d2 s 2
−2 − r + 6 =0 (11.53)
dr dr3 dr dr2 dr2
The substitution
d
U= s(r) (11.54)
dr
transforms (11.53) into
d2 U dU 2 dU
2U 2
− 6 − rU 2 =0 (11.55)
dr dr dr
132
The above equation has been solved in Appendix E, Example 13.
Integrating (11.54) and using the original variables (11.51), we obtain the
solution of the Blasius equation (11.13) in implicit form
Z
x = U (r)dr (11.56)
r=y
In Appendix E, Example 13, we have found that the general solution of

equation (11.55) (under the obvious change of variables y→U and r→x) is
given, in implicit form, by
Z
dp
−2 ln(r) = (11.57)
p3 u(p) p=r2 U
where u(r) is a solution of the differential equation (E.234), (E.238) solved by

Panayotounakos algorithm. We thus have arrived at the following Theorem:
Theorem. The general solution of the Blasius equation
1
y 000 + yy 00 = 0 (11.58)
2
is given by
Z
x = U (r)dr (11.59)
r=y
The function U (r) is evaluated in implicit form by

Z
dp
−2 ln(r) = 3
(11.60)
p u(p) p=r2 U
where u(r) is a solution of the differential equation (E.234), (E.238) (Abel’s

second kind differential equation yy 0 − y = f (x)) solved by Panayotounakos
algorithm.
11.2.2. Second method of solution for Blasius equation.
Since Blasius equation does not contain the independent variable x explicitly,
we can use the substitution
dy
p= (11.61)
dx
133
The above substitution leads to the identities
d2 y dp d3 y 2
2d p
dp 2
=p , =p +p (11.62)
dx2 dy dx3 dy 2 dy
Because of the above identities, Blasius equation is being converted into
d2 p dp 2 1 dp
p 2+ + y =0 (11.63)
dy dy 2 dy
We thus see that Blasius equation has been converted into a second order
ordinary differential equation where p is the unknown function and y is the
dependent variable. Equation (11.63) is a highly nonlinear equation and has
been solved in Appendix E, Example 14. The solution is given by (E.258),
where under the obvious change of variables x→y and y→p, reads in our case
Z
r
ln(y) = dr (11.64)
w(r) r=p/y2
where w(r) satisfies either (E.256) or (E.260) and can be solved thanks to
Panayotounakos algorithm.
Equation (11.64) provides an equation of the form p = p(y) which, when
combined to (11.61), gives the equation
dy
= p(y) (11.65)
dx
The above equation, when solved, gives y as a function of x: y = y(x).
11.2.3. Third method of solution for Blasius equation.
Blasius equation can also be solved by Nucci’s reduction algorithm.
We introduce the variables
w1 = y(x), w2 = y 0 (x), w3 = y 00 (x) (11.66)
We then have
1
w10 = w2 , w20 = w3 , w30 + w1 w3 = 0 (11.67)
2
Our task is to convert Blasius equation into a second order equation with
unknown function w2 depending on the variable w1 : w2 = w2 (w1 ). The above
equations (11.67) can also be written as
dw1 dw2 dw3 1
= w2 , = w3 , = − w1 w3 (11.68)
dx dx dx 2
134
Upon dividing the last two equations by the first one, we obtain the equations
dw2 w3 dw3 1 w1 w3
= and =− (11.69)
dw1 w2 dw1 2 w2
Solving the first of the above equations with respect to w3 , i.e.
dw2
w3 = w2 (11.70)
dw1
and substitution into the second, we get the equation
d dw2 1 dw2
w2 w2 = − w1 ·w2
dw1 dw1 2 dw1
The above equation is equivalent to (after canceling w2 and performing the
differentiation)
d2 w2 dw2 2 1 dw2
w2 + + w1 =0 (11.71)
dw12 dw1 2 dw1
Remarkably enough, the above equation is exactly the same to (11.63). We

thus see that we have found an equation with unknown function w2 (w1 ). This
equation, under the obvious changes w1 →x, w2 →y was solved in Appendix
E (Example 12). The solution is given by (E.214) which in our case is
Z
r
ln(w1 ) = dr (11.72)
w(r) r=w2 /w2
1
The function w(r) satisfies (E.212) and can be transformed into (E.217)
which can be solved using the Panayotounakos Algorithm.
We thus suppose that we have found w2 = w2 (w1 ), thanks to (11.72). Using
the first of (11.68), i.e.
dw1
= w2 (w1 ) (11.73)
dx
we obtain w1 by solving the above differential equation. In other words, we
can determine w1 and thus y(x):
w1 = y(x) (11.74)
135
11.3. Example 2. The Chazy equation. Find the general solution of
the equation
d3 y d2 y dy 2
= 2y 2 − 3 (11.75)
dx3 dx dx
Solution. Let us denote by ∆ the quantity
∆≡y 000 − 2yy 00 + 3(y 0 )2 (11.76)
If P r(3) Γ is the operator

∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (11.77)
∂x ∂y ∂y ∂y ∂y
then invariance of the equation is being expressed by
P r(3) Γ(∆) = 0
∂ ∂ ∂ ∂ ∂
⇔ X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (∆) = 0 (11.78)
∂x ∂y ∂y ∂y ∂y
00 [x] 0 [xx] [xxx]
⇔ Y (−2y ) + Y (6y ) + Y (−2y) + Y =0
Upon substituting the infinitesimals Y [x] , Y [xx] and Y [xxx] into the above
equation, we derive the equation

− 2Y y 00 + 6y 0 Yx + (Yy − Xx )y 0 − Xy (y 0 )2

− 2y Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2

0 3 00 0 00
− Xyy (y ) + (Yy − 2Xx )y − 3Xy y y
(11.79)
+ Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2
+ (Yy − 3Xx )y 000 − 4Xy y 0 y 000 = 0
136
y 000 by 2yy 00 − 3(y 0 )2 in (11.79). We then obtain the equation

− 2Y y 00 + 6y 0 Yx + (Yy − Xx )y 0 − Xy (y 0 )2

− 2y Yxx + (2Yxy − Xxx )y 0 + (Yyy − 2Xxy )(y 0 )2

0 3 00 0 00
− Xyy (y ) + (Yy − 2Xx )y − 3Xy y y
(11.80)
+ Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2
+ (Yy − 3Xx )(2yy 00 − 3(y 0 )2 ) − 4Xy y 0 (2yy 00 − 3(y 0 )2 ) = 0
− Xyyy (y 0 )4 + (6Xy + 2yXyy + Yyyy − 3Xxyy )(y 0 )3

+ [3(Yy + Xx ) − 2y(Yyy − 2Xxy ) + 3(Yxyy − Xxxy )](y 0 )2
+ [6Yx − 2y(2Yxy − Xxx ) + (3Yxxy − Xxxx )]y 0 − 3Xy (y 00 )2
(11.81)
+ [3(Yxy − Xxx ) − 2yXx − 2Y ]y 00
+ [−2yXy + 3(Yyy − 3Xxy )]y 0 y 00
− 6Xyy (y 0 )2 y 00 + (−2yYxx + Yxxx ) = 0
Equating the coefficients of the derivatives of the above equation to zero, we

obtain the following system of the determining equations
−Xyyy = 0 (11.82)
6Xy + 2yXyy + Yyyy − 3Xxyy = 0 (11.83)

3(Yy + Xx ) − 2y(Yyy − 2Xxy ) + 3(Yxyy − Xxxy ) = 0 (11.84)
6Yx − 2y(2Yxy − Xxx ) + (3Yxxy − Xxxx ) = 0 (11.85)
−3Xy = 0 (11.86)
3(Yxy − Xxx ) − 2yXx − 2Y = 0 (11.87)
−2yXy + 3(Yyy − 3Xxy ) = 0 (11.88)
−6Xyy = 0 (11.89)
137
−2yYxx + Yxxx = 0 (11.90)
The solution of the above system (11.82)-(11.90) of the determining equations
is given by
1
X = C 1 x2 + C 2 x + C 3 , Y = −C1 xy − C2 y − 3C1 (11.91)
2
and thus
∂ ∂
Γ=X +Y
∂x ∂y
1 ∂ ∂
= C 1 x2 + C 2 x + C 3 + (−C1 xy − C2 y − 3C1 ) (11.92)
2 ∂x ∂y
1 ∂ ∂ ∂ ∂ ∂
= C 1 x2 − (xy + 3) + C2 x −y + C3
2 ∂x ∂y ∂x ∂y ∂x
Therefore the generators of the Lie algebra are given by
∂ ∂ ∂ ∂ ∂
Γ(1) = x2 − 2(xy + 3) , Γ(2) = x −y , Γ(3) = (11.93)
∂x ∂y ∂x ∂y ∂x
Let us now try to find some general solutions of the Chazy equation taking
into account the above generators.
Case 1. In this case we take the generator Γ(3) . In other words we consider
X = 1, Y =0 (11.94)
The normal coordinates corresponding to this generator are given by (Ap-

penix G, equation (G.15))
r = y, s=x+y (11.95)
We then consider the transformation
x = s(r) − r, y(x) = r (11.96)
Under this substitutions, the Chazy equation takes on the form (G.19). The
substitution
d
s(r) = U (r) (11.97)
dr
138
converts (G.19) into (G.21). This last equation was solved in Appendix E,
Example 15. Integrating (11.97), we get
Z
s(r) = U (r)dr (11.98)
and using the initial variables (11.95),

Z
x + y = U (r)dr (11.99)
r=y
The solution of (G.21) is given by (E.280). Under the obvious change of

variables x→r, y→U , solution (E.280) has the form
Z
dp
−2 ln(r) = 3
(11.100)
p w(p) p=(−1+U )r2
From the above equation the function U (r) follows. The function w(r) fol-
lows either from (E.278) or (E.282) and can be solved by Panayotounakos
algorithm. We thus have proved the Theorem:
Theorem 1. The Chazy equation
d3 y d2 y dy 2
= 2y 2 − 3 (11.101)
dx3 dx dx
admits a general solution given by
Z
x + y = U (r)dr (11.102)
r=y
where U (r) is given implicitly by

Z
dp
−2 ln(r) = 3
(11.103)
p w(p) p=(−1+U )r2
and the function w(r) satisfies either (E.278) or (E.282) and can be solved
by Panayotounakos algorithm.
Case 2. In the second case we consider the generator Γ(1) . In other words
we take
X = x2 , Y = −6 − 2xy (11.104)
139
The normal coordinates corresponding to this generator are given by
1
r = (6 + xy)x, s=− (11.105)
x
We then consider the transformation
1
x=− , y = r·s(r)2 + 6·s(r) (11.106)
s(r)
Under the above transformation, Chazy equation takes on the form (G.44).
The substitution
d
s(r) = Y (r) (11.107)
dr
converts (G.44) into (G.46). Integrating (11.107),
Z
s(r) = Y (r)dr
and taking into account the transformation (11.105), we obtain

Z
1
− = Y (r)dr (11.108)
x r=(6+xy)x
Equation (G.46) was solved in Appendix E, Example 16. This solution is

given by (E.302). Under the obvious change of variables x→r, y→Y we
obtain from (E.302)
Z
dp
ln(r) = 3
(11.109)
p w(p) p=r2 Y
From above equation the function Y (r) follows. The function w(r) satisfies
either (E.300) or (E.304) and can be solved by Panayotounakos algorithm.
We have thus arrived at the following Theorem:
Theorem 2. The Chazy equation
d3 y d2 y dy 2
= 2y − 3 (11.110)
dx3 dx2 dx
Z
1
− = Y (r)dr (11.111)
x r=y
140
where Y (r) is given implicitly by
Z
dp
ln(r) = 3
(11.112)
p w(p) p=r2 Y
and the function w(r) satisfies either (E.300) or (E.304) and can be solved
by Panayotounakos algorithm.
11.4. Example 3. The equation y 000 = 0.
In this example we shall determine the generators of the Lie algebra of sym-
metries for the equation y 000 = 0. Invariance of the given equation is being
expressed by the condition
P r(3) Γ(∆) = 0, ∆ = y 000 (11.113)

∆=0

∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (11.114)
∂x ∂y ∂y ∂y ∂y
We have in this case
P r(3) Γ(∆)
∂ ∂ ∂ ∂ ∂
= X +Y + Y [x] 0 + Y [xx] 00 + Y [xxx] 000 (∆) (11.115)
∂x ∂y ∂y ∂y ∂y
[xxx]
=Y
Upon substituting the infinitesimal Y [xxx] into the above equation we derive
the equation
P r(3) Γ(∆)
= Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2
+ (Yyyy − 3Xxyy )(y 0 )3 − Xyyy (y 0 )4 + 3(Yxy − Xxx )y 00 (11.116)
+ (Yy − 3Xx )y 000 − 4Xy y 0 y 000

y 000 by zero in (11.116). We then obtain that the condition P r(3) Γ(∆) =0
∆=0
141
is equivalent to the equation
Yxxx + (3Yxxy − Xxxx )y 0 + 3(Yxyy − Xxxy )(y 0 )2

+ (Yyyy − 3Xxyy )(y 0 )3 − Xyyy (y 0 )4 + 3(Yxy − Xxx )y 00 (11.117)
+ 3(Yyy − 3Xxy )y 0 y 00 − 6Xyy (y 0 )2 y 00 − 3Xy (y 00 )2 = 0
Equating to zero all the coefficients of y 0 , we obtain the system of equations
Yxxx = 0 (11.118)
3Yxxy − Xxxx = 0 (11.119)

Yxyy − Xxxy = 0 (11.120)
Yyyy − 3Xxyy = 0 (11.121)
Xyyy = 0 (11.122)
Yxy − Xxx = 0 (11.123)
Yyy − 3Xxy = 0 (11.124)
Xyy = 0 (11.125)
Xy = 0 (11.126)
From the last equation above we get Xy = 0 and then
X = X(x) (11.127)
Equations (11.125) and (11.122) are satisfied identically. From equation

(11.124) we get Yyy = 0 and thus
Y = A(x)y + B(x) (11.128)
Equations (11.120) and (11.121) are satisfied identically. Equation (11.118),

because of (11.128) becomes A000 (x)y + B 000 (x) = 0 and thus A000 (x) = 0 and
B 000 (x) = 0, i.e.
A(x) = a1 + a2 x + a3 x2 (11.129)
and
B(x) = a4 + a5 x + a6 x2 (11.130)
142
From equation (11.123) we get A0 (x)−X 00 (x) = 0 and using (11.129) we have
X 00 (x) = a2 + 2a3 x and by integration
1 1
X = a3 x 3 + a2 x 2 + a7 x + a8 (11.131)
3 2
Equation (11.119) takes on the form 3A00 (x) − X 000 (x) = 0, which because of
(11.129) and (11.131) yields 3·2a3 − 2a3 = 0 and thus a3 = 0. Therefore
A(x) = a1 + a2 x
(11.132)
B(x) = a4 + a5 x + a6 x2
and
1
X = a2 x2 + a7 x + a8 , Y = (a1 + a2 x)y + a4 + a5 x + a6 x2 (11.133)
2
The generator is thus given by
∂ ∂
Γ=X +Y
∂x ∂y
1 ∂ (11.134)
2 2 ∂
= a2 x + a7 x + a8 + [(a1 + a2 x)y + a4 + a5 x + a6 x ]
2 ∂x ∂y
∂ 1 ∂
2 ∂ ∂ ∂
Γ = a1 y + a2 x + xy + a4 + a5 x
∂y 2 ∂x ∂y ∂y ∂y
(11.135)
∂ ∂ ∂
+ a6 x 2 + a7 x + a8
∂y ∂x ∂x
Therefore the equation y 000 = 0 admits the following symmetry generators
∂ 1 ∂ ∂ ∂
Γ(1) = y , Γ(2) = x2 + xy , Γ(3) = ,
∂y 2 ∂x ∂y ∂y
(11.136)
(4) ∂ ∂ ∂ ∂
Γ = x , Γ(5) = x2 , Γ(6) = x , Γ(7) =
∂y ∂y ∂x ∂x
11.5. Example 4. The first integrals of the equation y 000 = 0.
Equation y 000 = 0 admits a first integral
I = I(x, y, y 0 , y 00 ) (11.137)
143
if
Dx I = 0, ∆ = y 000 (11.138)
∆=0
where Dx is the total derivative operator

∂ ∂ ∂ ∂
Dx = + y0 + y 00 0 + y 000 00 (11.139)
∂x ∂y ∂y ∂y
We have
∂I ∂I ∂I
Dx I = 0 ⇐⇒ + y0 + y 00 0 = 0 (11.140)
∆=0 ∂x ∂y ∂y
Considering the ansatz
I = A(x, y)y 00 + B(x, y)y 0 + C(x, y) (11.141)
the defining condition (11.140) yields
Ax y 00 + Bx y 0 + Cx + y 0 (Ay y 00 + By y 0 + Cy ) + y 00 B = 0 (11.142)
(Ax + B)y 00 + (Bx + Cy )y 0 + Ay y 0 y 00 + By (y 0 )2 + Cx = 0 (11.143)
Ax + B = 0 (11.144)
Bx + Cy = 0 (11.145)
Ay = 0, By = 0, Cx = 0 (11.146)
From the last three equations we get
A = A(x), B = B(x), C = C(y) (11.147)
From (11.144) we get
B = −A0 (x) (11.148)
144
Equation (11.145) then becomes
−A00 (x) + C 0 (y) = 0 (11.149)
The above equation to be true, each member should be equal to a constant:
A00 (x) = a1 and C 0 (y) = a1 (11.150)
From the above two equations we obtain

1
A(x) = a1 x2 + a2 x + a3 and C = a1 y + a4 (11.151)
2
We thus get from (11.148) that
B = −(a1 x + a2 ) (11.152)
Therefore equation (11.141), because of (11.151) and (11.152) yields

1
I= a1 x2 + a2 x + a3 y 00 − (a1 x + a2 )y 0 + a1 y + a4 (11.153)
2
1
I = a1 x2 y 00 − xy 0 + y + a2 (xy 00 − y 0 ) + a3 y 00 + a4 (11.154)
2
We thus get the following first integrals for the equation y 000 = 0:
1
I1 = x2 y 00 − xy 0 + y, I2 = xy 00 − y 0 , I3 = y 00 (11.155)
2
11.6. Linearization. Linearization is the process of converting a third
order ODE
y 000 = F (x, y, y 0 , y 00 ) (11.156)
into an equation of the form (Forsyth-Laguerre)

d3 u
u000 (t)≡ = 0 or u000 (t) + α(t)u(t) = 0 (11.157)
dt3
Let us suppose that there is a transformation
t = φ(x, y)
(11.158)
u = ψ(x, y)
145
which transforms (11.157) into (11.156). We have already found how the first
and second derivatives of u transform under (11.158) in section 10.3. We also
have proved that u00 = P (x, y, y 0 , y 00 ) where
∆
P = 0 3
{y 00 + A(x, y)(y 0 )3
(φx + y φy ) (11.159)
+ B(x, y)(y 0 )2 + C(x, y)y 0 + D(x, y)}
upon comparing (10.39) and (10.46), where A, B, C and D are given by

(10.47)-(10.50) respectively.
The transformation of the third derivative u000 is then given by
Dx P Px + y 0 Py + y 00 Py0 + y 000 Py00
u000 = = (11.160)
Dx φ φx + y 0 φy
Calculating Px , Py , Py0 and Py00 with P given by (11.159), we can find the
explicit expression of u000 in terms of the functions φ and ψ as we did in the
case of the second derivative u00 . The explicit results are listed in Ibragimov
and Meleshko (Ibragimov and Meleshko [2]).
Case I. If φy = 0 then equation u000 +αu = 0 yields (after expanding (11.160))
y 000 + (A1 y 0 + A0 )y 00 + B3 (y 0 )3 + B2 (y 0 )2 + B1 y 0 + B0 = 0 (11.161)
where
A1 = 3(ψy )−1 ψyy , A0 = 3(φx ψy )−1 (φx ψxy − ψy φxx ) (11.162)
B3 = (ψy )−1 ψyyy , B2 = 3(φx ψy )−1 (φx ψxyy − ψyy φxx ) (11.163)
B1 = (φ2x ψy )−1 (3φ2xx ψy − φxxx φx ψy − 6φxx φx ψxy + 3φ2x ψxxy ) (11.164)
B0 = (φ2x ψy )−1 (3φ2xx ψy − φxxx φx ψx − 3φxx φx ψxx
(11.165)
+ φ2x ψxxx + αψφ5x )
We also have the following (Ibragimov and Meleshkov [2])
Theorem 1. Equation (11.161) is linearizable if and only if its coefficients
satisfy the conditions
(A0 )y − (A1 )x = 0, (3B1 − A20 − 3(A0 )x )x = 0,
3(A1 )x + A0 A1 − 3B2 = 0, 3(A1 )y + A21 − 9B3 = 0,
(11.166)
(9B1 − 6(A0 )x − 2A20 )(A1 )x + 9((B1 )x − A1 B0 )y
+ 3A0 (B1 )y − 27(B0 )yy = 0
146
If conditions (11.166) are satisfied, then the function φ is determined from
equation
dχ φxx
6 − 3χ2 = 3B1 − A20 − 3(A0 )x , χ= (11.167)
dx φx
The function ψ is determined by the following system
3ψyy = A1 ψy , 3ψxy = (3χ + A0 )ψy (11.168)

1
ψxxx = 3χψxx + B0 ψy − (3(A0 )x + A20 − 3B1 + 9χ2 )ψx
6 (11.169)
− Ωψ
with Ω given by
1
Ω= {9(A0 )xx + 18(A0 )x A0 + 54(B0 )y − 27(B1 )x
54 (11.170)
+ 4A30 − 18A0 B1 + 18A1 B0 }
The coefficient α appearing in u000 + αu = 0 is given by
α = Ω(φx )−3 (11.171)
Example. We shall apply the Theorem of this subsection to linearize the

6 3 00 6 6 6 6y
y 000 − y0 + y + 2 (y 0 )3 + (y 0 )2 + 2 y 0 + 3 = 0 (11.172)
y x y xy x x
For this equation we have the coefficients
6 3 6 6 6 6y
A1 = − , A0 = − , B3 = 2 , B2 = , B1 = 2 , B0 = 3 (11.173)
y x y xy x x
We can verify that the conditions (11.166) are satisfied.
Since 3B1 − A20 − 3(A0 )x = 0, equation (11.167) yields
dχ
6 − 3χ2 = 0 (11.174)
dx
An obvious solution of this equation is χ = 0 and thus
φ=x (11.175)
147
Equations (11.168) take on the form
∂ 2 ∂ 1
ln |ψy | = − , ln |ψy | = − (11.176)
∂y y ∂x x
K
with solution ψy = . Therefore
xy 2
K
ψ=− + f (x) (11.177)
xy
Taking for simplicity K = −1 and f (x) = 0, we obtain
1
ψ= (11.178)
xy
Therefore the linearizing transformation is given by
1
φ = x, ψ= (11.179)
xy
In other words,
1
t = x, u= (11.180)
xy
Since Ω = 0 from (11.170), we get α = 0 using (11.171). Therefore the
transformation (11.180) maps (11.172) into
u000 = 0 (11.181)
u = a2 t2 + a1 t + a0 (11.182)
and using (11.180),

1
= a2 x 2 + a1 x + a0 (11.183)
xy
and finally
1
= (a2 x2 + a1 x + a0 )x (11.184)
y
148
Case II. If φy 6=0 then equation u000 +αu = 0 yields (after expanding (11.160))
1
y 000 + {−3(y 00 )2 + (C2 (y 0 )2 + C1 y 0 + C0 )y 00 + D5 (y 0 )5
y0 +r (11.185)
0 4 0 3 0 2 0
+ D4 (y ) + D3 (y ) + D2 (y ) + D1 y + D0 } = 0
where
φx
r(x, y) = (11.186)
φy
The various coefficients C2 , C1 , C0 and D5 , D4 , D3 , D2 , D1 , D0 are complicated
expressions, expressed in terms of the derivatives of the functions φ and ψ
and are given in Ibragimov and Meleshkov (Ref [2]).
There is a similar Theorem in this case, however because of the complexity
of the calculations involved, we refer to Ref. [2].
Example. Equation
1
y 000 + {−3(y 00 )2 − x(y 0 )5 } = 0 (11.187)
y0
can be linearized by the transformation (see details in Ref.[2])
φ = y, ψ=x (11.188)
i.e. by
t = y, u=x (11.189)
In this case α = 1 and thus (11.187) maps into the equation
u000 + u = 0 (11.190)

√3 t √3 t
( )
u=e t/2
C1 cos + C2 sin + C3 e−t (11.191)
2 2
and because of (11.189),

√3 y √3 y
( )
x = ey/2 C1 cos + C2 sin + C3 e−y (11.192)
2 2
149
References
[2] N. H. Ibragimov and S. V. Meleshko: ”Linearization of third-order
Ordinary Differential Equations by Point and Contact Transformations”.
J. Math. Anal. Appl. 308 (2005) 266-289
[3] P. S. Laplace: ”Recherches sur le calcul intégral aux différences
partielles”, Mémoires de l’Académie Royale des Sciences de Paris, 1773/77,
341-402. Reprinted in Oeuvres Complètes de Laplace, Vol. 9, Gauthier-
Villars, Paris, 1893, 5-68, English transl., New York, 1966.
[4] E. Cotton: ”Sur les invariants différentiels de quelques équations
lineaires aux dérivées partielles du second ordre”. Annales Scientifique de
l’École Normale Supérieure 17, 1900, 211-244.
[5] J. Cockle: ”On linear differential equations of the third order”.
Quarterly Journal of Mathematics 15, 1876, 340-353.
[6] E. Laguerre: ”(i) Sur les équations différentielles linéaires du troisième
ordre”. Comptes Rendus de l’Académie des Sciences Paris 88, 1879, 116-119.
(ii) Sur quelques invariants des équations différentielles linéaires”. Comptes
Rendus de l’Académie des Sciences Paris 88, 1879, 224-227.
[7] G. H. Halphen: ”Sur les invariants des équations différentielles linéaires
du quatrième ordre”, Acta Mathematica 3, 1883, 325-380.
Reprinted in Oeuvres d’Halphen, Vol. 3, 1921.
[8] A. R. Forsyth: ”Invariants, covariants, and quotient-derivatives asso-
ciated with linear differential equations”. Philosophical Transactions of the
Royal Society of London 179, 1888, 377-489.
[9] S. Lie: ”Zur allgemeinen Theorie der partiellen Differentialgleichungen
beliebiger Ordnung”. Leipziger Berichte 1, 1895, 53-128.
Reprinted in Lie’s Gesammelte Abhandlundgen, Vol. 4, 1929, Paper IX.
[10] N. H. Ibragimov: ”Infinitesimal method in the theory of invariants
of algebraic and differential equations”, Notices of the South African Math-
ematical Society 29, 1997, 61-70.
[11] N. H. Ibragimov: ”Laplace type invariants for parabolic equations,
Nonlinear Dynamics”, to appear.
[12] N. H. Ibragimov: ”Elementary Lie Group Analysis and Ordinary Dif-
ferential Equations”, Wiley, Chichester, 1999.
[13] S. Lie: ”Klassifikation und Integration von gewöhnlichen Differential-
gleichungen zwischen x, y, die eine Gruppe von Transformationen gestatten”
(III), Archive for Mathematical Logic 8 (1883), 371-458.
150
Reprinted in Lie’s Gesammelte Abhandlundgen, Vol. 5, 1924, Paper XIV.
11.7. Higher Order Ordinary Differential Equations.

Invariance of an n-th order Ordinary Differential Equation
∆≡F (x, y, y 0 , y 00 , ..., y (n) ) = 0 (11.193)
under infinitesimal point transformation (11.2) is being expressed by the

condition
P r(n) Γ(∆)|∆=0 = 0 (11.194)
where P r(n) Γ is the operator

∂ ∂ ∂ ∂ ∂
P r(n) Γ = X +Y + Y [x] 0 + Y [xx] 00 + · · · + Y [(n)x] (n) (11.195)
∂x ∂y ∂y ∂y ∂y
The infinitesimals are evaluated recursively by the formulas
Y [(n)x] = Dx (Y [(n−1)x] ) − y (n) Dx (X) (11.196)
The total derivative operator Dx is defined by

∂ ∂ ∂ ∂ ∂
Dx = + y0 + y 00 0 + y 000 00 + · · · + y (n) (n−1)y (11.197)
∂x ∂y ∂y ∂y ∂y
Example 1. We first consider the 4th order equation
x3 ·y (iv) + 2x2 ·y 000 − xy 00 + y 0 − a4 x3 y = 0 (11.198)
This equation admits the following infinitesimals

[X = 0, Y = y], [X = 0, Y = I0 (ax)], [X = 0, Y = J0 (ax)],
(11.199)
[X = 0, Y = K0 (ax)], [X = 0, Y = Y0 (ax)]
where I0 (x), J0 (x), K0 (x), Y0 (x) are the Bessel functions. Solving the normal
coordinates system for [X = 0, Y = y], we find [r = F (x), s = ln(y) + G(x)].
We take for simplicity
r = x, s = ln(y) (11.200)
Therefore we consider the subsitution
x = r, y(x) = es(r) (11.201)
151
Under this subsitution, the original equation (11.198) takes on the form
4 3 2
3 d 3 d
d
3 d
2
r s(r) + 4r s(r) s(r) + 3r s(r)
dr4 dr3 dr dr2
d2 d 3
3 d 2 d
2 4
+ 6r3 s(r) s(r) + r s(r) + 2r s(r)
dr2 dr dr dr3 (11.202)
d2 d d 3 d2
2 2
+ 6r s(r) s(r) + 2r s(r) − r s(r)
dr2 dr dr dr2
d 2 d
−r s(r) + s(r) − a4 r3 = 0
dr dr
The above equation might seem more complicated than the original one.
However it admits an exact solution expressed in terms of Bessel’s functions.
Example 2. As our second example, we consider another 4th order equation
x·y (iv) + 4y 000 − a4 xy = 0 (11.203)
This equation admits the following infinitesimals
[X = 0, Y = y], [X = 1, Y = −y/x], [X = 0, Y = eax /x],
[X = 0, Y = e−ax /x]], [X = 0, Y = sin(ax)/x], (11.204)
[X = 0, Y = cos(ax)/x]
Taking [X = 0, Y = eax /x] and solving the normal coordinates system, we
get the solution r = F (x), s = xye−ax + G(x). Considering for simplicity
r = x, s = xye−ax and the accompanying transformation
s(r) ar
x = r, y(x) = e (11.205)
r
equation (11.203) transforms into a 4th order ODE with constant coefficients
d4 d3 2
2 d

3 d

s(r) + 4a s(r) + 6a s(r) + 4a s(r) = 0 (11.206)
dr4 dr3 dr2 dr
s(r) = C1 + C2 e−2ar + e−ar {C3 sin(ar) + C4 cos(ar)} (11.207)
Going back to the original variables, we get the general solution of the equa-
tion:
1
y = [C1 + C2 e−2ax + e−ax {C3 sin(ax) + C4 cos(ax)}]eax (11.208)
x
152
Chapter III
Partial Differential Equations
12 Partial Differential Equations I
First Order PDEs
In this Section we consider Lie symmetries of first order partial differential
equations of functions with two variables.
12.1. The Invariance Condition.
We consider a first order partial differential equation
F (t, x, u, ut , ux ) = 0 (12.1)
where u(t, x) is the unknown function and ut , ux are, as usual, the partial
derivatives with respect to t and x respectively.
We require invariance of (12.1) with respect to one-parameter transforma-
tions
t̂ = t + ·T (t, x, u) + O(2 ) (12.2)
x̂ = x + ·X(t, x, u) + O(2 ) (12.3)

û = u + ·U (t, x, u) + O(2 ) (12.4)
In analogy to transformation rules in ODEs, we expect the transformation
rules
ût̂ = ut + ·U [t] + O(2 ) (12.5)
ûx̂ = ux + ·U [x] + O(2 ) (12.6)

We also introduce the total differential operators Dt and Dx by
∂ ∂ ∂ ∂
Dt = + ut + utt + utx + ··· (12.7)
∂t ∂u ∂ut ∂ux
∂ ∂ ∂ ∂
Dx = + ux + utx + uxx + ··· (12.8)
∂x ∂u ∂ut ∂ux
153
Let us now find the explicit expressions for U [t] and U [x] .
We have the obvious expression
∂(u, x) u u u u
ut = = t x = t x = ut (12.9)
∂(t, x) x t x x 0 1
∂(u, x)
where is the Jacobian. Therefore
∂(t, x)
∂(û, x̂)
ût̂ = (12.10)
∂(t̂, x̂)
and thus
∂(û, x̂) ût ûx
∂(û, x̂) ∂(t, x) x̂t x̂x
ût̂ = = =
∂(t̂, x̂) ∂(t̂, x̂) t̂t t̂x
∂(t, x) x̂t x̂x
ut + ·Dt (U ) + O(2 ) ux + ·Dx (U ) + O(2 )
·Dt (X) + O(2 ) 1 + ·Dx (X) + O(2 )
= =
1 + ·Dt (T ) + O(2 ) ·Dx (T ) + O(2 ) (12.11)
·Dt (X) + O(2 ) 1 + ·Dx (X) + O(2 )
ut + ·[Dt (U ) + ut Dx (X) − ux Dt (X)] + O(2 )
=
1 + ·[Dt (T ) + Dx (X)] + O(2 )
= {ut + ·[Dt (U ) + ut Dx (X) − ux Dt (X)] + O(2 )}
·{1 − ·[Dt (T ) + Dx (X)] + O(2 )} =
= ut + ·[Dt (U ) − ut Dt (T ) − ux Dt (X)] + O(2 )
Therefore
U [t] = Dt (U ) − ut Dt (T ) − ux Dt (X) (12.12)
We have similarly
U [x] = Dx (U ) − ut Dx (T ) − ux Dx (X) (12.13)
We are now able to determine the explicit expressions for U [t] and U [x] .
We have
U [t] = Dt (U ) − ut Dt (T ) − ux Dt (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ut U − ut + ut T − ux + ut X
∂t ∂u ∂t ∂u ∂t ∂u
154
and upon expending the right hand side, we obtain
U [t] = Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )

(12.14)
= Ut + (Uu − Tt )ut − Xt ux − Tu (ut )2 − Xu ut ux
Similarly we have
U [x] = Dx (U ) − ut Dx (T ) − ux Dx (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ux U − ut + ux T − ux + ux X
∂x ∂u ∂x ∂u ∂x ∂u
and upon expending the right hand side, we obtain
U [x] = Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu )

(12.15)
= Ux + (Uu − Xx )ux − Tx ut − Xu (ux )2 − Tu ux ut
Introducing the infinitesimal operator Γ as

∂ ∂ ∂
Γ=T +X +U (12.16)
∂t ∂x ∂u
its first extension is
∂ ∂
P r(1) Γ = Γ + U [t] + U [x] (12.17)
∂ut ∂ux
If we denote by ∆ the quantity F ,
∆≡F (t, x, u, ut , ux ) (12.18)
then Lie’s Invariance Condition becomes
P r(1) Γ(∆)|∆=0 = 0 (12.19)
12.2. Example 1. Determine solutions to the differential equation
ut = (ux )2 , u = u(t, x) (12.20)
using Lie point symmetries.

Solution. This is a first order partial differential equation. We define ∆ by
∆ = ut − (ux )2 (12.21)
155
P r(1) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x] ut − (ux )2 = 0 (12.22)
∂t ∂x ∂u ∂ut ∂ux
[t] [x]
⇔ U − 2U ux = 0
Substituting (12.14) and (12.15) into (12.22), we obtain the equation
Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )
(12.23)
− 2ux [Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu )] = 0
We now have to take into account the condition ∆ = 0. This means that we
have to substitute ut by (ux )2 . We then obtain form (12.23) the equation
Ut + (ux )2 Uu − (ux )2 (Tt + (ux )2 Tu ) − ux (Xt + (ux )2 Xu )

(12.24)
− 2ux [Ux + ux Uu − (ux )2 (Tx + ux Tu ) − ux (Xx + ux Xu )] = 0
The previous equation can be written as (arranged in powers of ux )
Ut − (Xt + 2Ux )ux + (2Xx − Tt − Uu )(ux )2

(12.25)
+ (Xu + 2Tx )(ux )3 + Tu (ux )4 = 0
Equating the different powers of the derivative ux to zero, we obtain the

following system of determining equations
Ut = 0 (12.26)
Xt + 2Ux = 0 (12.27)
2Xx − Tt − Uu = 0 (12.28)
Xu + 2Tx = 0 (12.29)
Tu = 0 (12.30)
Let us now try to solve the above system and thus determine the unknown
functions T , X and U .
From equation (12.30) we find
T = α(t, x) (12.31)
156
and from equation (12.26) that
U = U (x, u) (12.32)
Therefore equation (12.29) becomes Xu + 2αx = 0 and by integration with

respect to u,
X = −2αx ·u + β(t, x) (12.33)
Equation (12.28) can be written as
2(−2αxx ·u + βx ) − αt − Uu = 0 (12.34)
and thus
Uu = −4αxx ·u + 2βx − αt (12.35)
Integrating the previous equation with respect to u, we obtain
U = −2αxx ·u2 + (2βx − αt )u + γ(t, x) (12.36)
Using the expressions (12.33) and (12.36) for X and U respectively, equation
(12.27) takes on the form
(−2αxt ·u + βt ) + 2[−2αxxx ·u2 + (2βxx − αtx )u + γx ] = 0
−4αxxx ·u2 + 4(βxx − αtx )u + βt + 2γx = 0 (12.37)
From the above equation we derive the system
αxxx = 0, βxx − αtx = 0, βt + 2γx = 0 (12.38)
On the other hand, equation (12.26), because of (12.36) becomes
−2αxxt ·u2 + (2βxt − αtt )u + γt = 0 (12.39)
from the above equation we obtain the system
αxxt = 0, 2βxt − αtt = 0, γt = 0 (12.40)
157
From equation γt = 0 we get
γ = δ(x) (12.41)
Equation βt + 2γx = 0 of the system (12.38) becomes
βt = −2δ 0 (x)
and by integration with respect to t we obtain
β = −2δ 0 (x)t + (x) (12.42)
Since βxt = −2δ 00 (x), equation 2βxt − αtt = 0 of the system (12.40) becomes
αtt = −4 δ 00 (x)
Integrating twice the above equation with respect to t, we get
α = −2δ 00 (x)t2 + ζ(x)t + η(x) (12.43)
Equation αxxt = 0 of the system (12.40) becomes
−4δ (iv) (x)t + ζ 00 (x) = 0
δ (iv) (x) = 0, ζ 00 (x) = 0 (12.44)
Equation αxxx = 0 of the system (12.38) becomes
−2δ (v) (x)t2 + ζ 000 (x)t + η 000 (x) = 0
From the above equation we get the system
δ (v) (x) = 0, ζ 000 (x) = 0, η 000 (x) = 0 (12.45)
Equation βxx − αtx = 0 of the system (12.38) takes on the form
2δ 000 (x)t + 00 (x) − ζ 0 (x) = 0
δ 000 (x) = 0, 00 (x) − ζ 0 (x) = 0 (12.46)
158
From the systems (12.44), (12.45) and (12.46) we obtain
δ 000 (x) = 0, ζ 00 (x) = 0, η 000 (x) = 0, 00 (x) − ζ 0 (x) = 0 (12.47)
From the first three equations of the above system we find
δ(x) = a3 x2 + a2 x + a1 (12.48)
ζ(x) = a4 x + a5 (12.49)
η(x) = a8 x2 + a7 x + a6 (12.50)
The fourth equation of (12.47) becomes 00 (x) = a4 and by integration
1
(x) = a4 x2 + a9 x + a10 (12.51)
2
Let us collect everything together:
α = −2δ 00 (x)t2 + ζ(x)t + η(x)

(12.52)
= −4a3 t2 + (a4 x + a5 )t + a8 x2 + a7 x + a6
β = −2δ 0 (x)t + (x)

1 (12.53)
= −2(2a3 x + a2 )t + a4 x2 + a9 x + a10
2
2
γ = δ(x) = a3 x + a2 x + a1 (12.54)
and
T = α(t, x) = −4a3 t2 + (a4 x + a5 )t + a8 x2 + a7 x + a6 (12.55)
X = −2αx ·u + β(t, x)
1 (12.56)
= −2(a4 t + 2a8 x + a7 )u − 2(2a3 x + a2 )t + a4 x2 + a9 x + a10
2
2
U = −2αxx ·u + (2βx − αt )u + γ(t, x)
(12.57)
= −4a8 u2 + (a4 x − a5 + 2a9 )u + a3 x2 + a2 x + a1
Using the explicit expressions for T , X and U , we are able to find the explicit
expression of the symmetry operator
∂ ∂ ∂
Γ=T +X +U (12.58)
∂t ∂x ∂u
159
We find that
∂
Γ = [−4a3 t2 + (a4 x + a5 )t + a8 x2 + a7 x + a6 ] +
∂t
+ − 2(a4 t + 2a8 x + a7 )u − 2(2a3 x + a2 )t
(12.59)
1 2 ∂
+ a4 x + a9 x + a10 +
2 ∂x
∂
+ [−4a8 u2 + (a4 x − a5 + 2a9 )u + a3 x2 + a2 x + a1 ]
∂u
∂ ∂ ∂ ∂ ∂ ∂
Γ = a1 + a2 −2t +x + a3 −4t2 − 4xt + x2
∂u ∂x ∂u ∂t ∂x ∂u
∂ 1 2 ∂ ∂ ∂ ∂
+ a4 xt + −2tu + x + a5 t − u + a6
∂t 2 ∂x ∂t ∂u ∂t
∂ (12.60)
∂
2 ∂ ∂ 2 ∂

+ a7 x − 2u + a8 x − 4xu − 4u
∂t ∂x ∂t ∂x ∂u
∂ ∂ ∂
+ a9 x + 2u + a10
∂x ∂u ∂x
from which we can read off the generators of the Lie symmetry algebra for
the given equation (10−dimensional algebra):
∂ ∂ ∂ ∂ ∂ ∂
Γ1 = , Γ2 = −2t + x , Γ3 = −4t2 − 4xt + x2
∂u ∂x ∂u ∂t ∂x ∂u
∂ 1 ∂ ∂ ∂ ∂
Γ4 = xt + −2tu + x2 , Γ5 = t − u , Γ6 =
∂t 2 ∂x ∂t ∂u ∂t (12.61)
7 ∂ ∂ 8 2 ∂ ∂ 2 ∂
Γ = x − 2u , Γ = x − 4xu − 4u
∂t ∂x ∂t ∂x ∂u
∂ ∂ ∂
Γ9 = x + 2u , Γ10 =
∂x ∂u ∂x
We can now find solutions to the equation by considering one generator or
a linear combination of generators. The solutions can then be found by con-
sidering either Canonical Coordinates or the Invariant Surface Condition.
These two methods are explained below.
12.3. The Canonical Coordinates. We consider the transformation
(t, x, u) → (r, s, v) and require the resulting partial differential equation to
160
be transformed into an ordinary differential equation not containing the vari-
able s. In other words we require the operator Γ (defined in (12.16)) to take
the form
∂
Γ= (12.62)
∂s
under the transformation
(t, x, u) → (r, s, v),
(12.63)
r = r(t, x, u), s = s(t, x, u), v = v(t, x, u)
The partial derivatives transform as

∂ ∂r ∂ ∂s ∂ ∂v ∂
= + + (12.64)
∂t ∂t ∂r ∂t ∂s ∂t ∂v
∂ ∂r ∂ ∂s ∂ ∂v ∂
= + + (12.65)
∂x ∂x ∂r ∂x ∂s ∂x ∂v
∂ ∂r ∂ ∂s ∂ ∂v ∂
= + + (12.66)
∂u ∂u ∂r ∂u ∂s ∂u ∂v
Therefore Γ takes on the form
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂r ∂ ∂s ∂ ∂v ∂ ∂r ∂ ∂s ∂ ∂v ∂
=T + + +X + + + (12.67)
∂t ∂r ∂t ∂s ∂t ∂v ∂x ∂r ∂x ∂s ∂x ∂v
∂r ∂ ∂s ∂ ∂v ∂
+U + +
∂u ∂r ∂u ∂s ∂u ∂v
∂ ∂
Γ = (T rt + Xrx + U ru ) + (T st + Xsx + U su ) +
∂r ∂s (12.68)
∂
+ (T vt + Xvx + U vu )
∂v
∂
Since we require Γ = , we should have
∂s
T rt + Xrx + U ru = 0 (12.69)
T st + Xsx + U su = 1 (12.70)
161
T vt + Xvx + U vu = 0 (12.71)
The above three equations constitute a system of first order PDEs, which
once solved, provide the transformation of the original PDE into an ordinary
differential equation with separable variables.
12.4. The Invariant Surface Condition.
We consider the transformation
t̂ = t + ·T (t, x, u) + O(2 ) (12.72)
x̂ = x + ·X(t, x, u) + O(2 ) (12.73)

û = u + ·U (t, x, u) + O(2 ) (12.74)
and require invariance of
u = u(t, x)→û = u(t̂, x̂) (12.75)
Since
û = u(t̂, x̂)⇔
u + U + O(2 ) = u(t + ·T (t, x, u) + O(2 ), (12.76)
x + ·X(t, x, u) + O(2 ))
upon expanding the right hand side, we have
u + U + O(2 ) = u(t, x) + (T ut + Xux ) + O(2 ) (12.77)
If u = u(t, x), then to order we obtain
T ut + Xux = U (12.78)
The above equation is called the Invariant Surface Condition.

From this equation we see that once we have determined T , X and U using
symmetry analysis, we can find the function u by solving the system
dt dx du
= = (12.79)
T X U
12.5. Solutions of the equation ut = (ux )2 .
We consider the sum of the generators Γ5 and Γ9 and then use one of the
two methods. Since
∂ ∂ ∂
Γ5 + Γ9 = t +x +u (12.80)
∂t ∂x ∂u
162
we see that
T = t, X = x, U =u (12.81)
(a) We first use the Invariant Surface Condition (12.78). This condition
becomes in our case
tut + xux = u (12.82)
In solving (12.82), we consider the system

dt dx du
= = (12.83)
t x u
From the first equation we obtain by integration
ln(t) = ln(x) + A, or ln(x/t) = B (12.84)
and thus
x
= C1 (12.85)
t
Integrating similarly the second equation, dt/t = du/u, we find similarly
u
= C2 (12.86)
t
Therefore the general solution of (12.82) is given by
x u u
F , = 0, or = f (x/t) (12.87)
t t t
and thus
x
u = t·f (12.88)
t
We introduce a new variable ξ by
x
ξ= (12.89)
t
and then
u = t·f (ξ) (12.90)
163
We shall find a differential equation by converting the original PDE into an
ODE satisfied by f (ξ). Since
x
ut = f (ξ) + t − 2 f 0 (x) = f (ξ) − ξf 0 (ξ) (12.91)
t
and

1 0
ux = t f (ξ) = f 0 (ξ) (12.92)
t
the original equation ut = (ux )2 becomes

2
0 0
f (ξ) − ξf (ξ) = f (ξ)

2
0 0
f (ξ) = ξf (ξ) + f (ξ) (12.93)
The above equation is a Clairaut-type of equation, the general form being

y = xy 0 + g(y 0 ). The Clairaut equation y = xy 0 + (y 0 )2 admits the general
solution
1
y = cx + c2 , y = − x2 (12.94)
4
The first is the general solution and the second is the singular solution.
Therefore in case of (12.93) we have the solution
1 x2
u = cx + c2 t, u=− (12.95)
4 t
which is a solution of the original equation (12.20).
(b) We next use Canonical Coordinates in solving the equation ut = (ux )2 .
The system of equations (12.69)-(12.70) takes in this case the form
trt + xrx + uru = 0

tst + xsx + usu = 1 (12.96)
tvt + xvx + uvu = 0
164
The first equation of the system (12.96) can be solved considering Lagrange’s
equations
dt dx du dr
= = = (12.97)
t x u 0
The above system admits three independent solutions
x u
r = C1 , = C2 , = C3 (12.98)
t t
Therefore the general solution of the first of (12.96) is given by
x u x u
F r, , = 0 or r = f , (12.99)
t t t t
A simple choice is
x
r= (12.100)
t
The second equation of the system (12.96) can be solved considering La-
grange’s equations
dt dx du ds
= = = (12.101)
t x u 1
x u
s = ln(t) + C1 , = C2 , = C3 (12.102)
t t
Therefore the general solution of the second equation of (12.96) is given by
x u x u
G s − ln(t), , = 0 or s − ln(t) = g , (12.103)
t t t t
A simple choice is
s = ln(t) (12.104)
The third equation of the system (12.96) can be solved considering Lagrange’s
equations
dt dx du dv
= = = (12.105)
t x u 0
165
x u
v = C1 , = C2 , = C3 (12.106)
t t
Therefore the general solution of the third of (12.96) is given by
x u x u
H v, , = 0 or v = h , (12.107)
t t t t
A simple choice is
u
v= (12.108)
t
Using the above choices (12.100), (12.104) and (12.108), we consider the
substitution
x
u = t·v(r, s), r = , s = ln(t) (12.109)
t
and transform the original equation ut = (ux )2 into another one in canonical
coordinates. Since
∂v ∂r ∂v ∂s
ut = v(r, s) + t +
∂r ∂t ∂s∂t
x 1 x
(12.110)
= v + t vr − 2 + vs · = v + − vr + vs
t t t
= v − r·vr + vs
and
∂v ∂r ∂v ∂s
1
ux = t + = t vr · + vs ·0 = vr (12.111)
∂r ∂x ∂s ∂x t
the original equation ut = (ux )2 becomes
v − r·vr + vs = (vr )2 (12.112)
For vs = 0 we have from (12.112)
v = r·vr + (vr )2 (12.113)
This is a Clairaut equation with solutions

1
v = cr + c2 , v = − r2 (12.114)
4
166
Therefore
2 1 x2
u = cx + c t, u=− (12.115)
4 t
are two general solutions of the equation ut = (ux )2 .
(c) We continue to finding solutions by considering the generator Γ8 :
∂ ∂ ∂
Γ8 = x2 − 4xu − 4u2 (12.116)
∂t ∂x ∂u
T = x2 , X = −4xu, U = −4u2 (12.117)
and thus the invariant surface condition (12.78) becomes
x2 ut − 4xuux = −4u2 (12.118)
In solving the above equation, we consider the system

dt dx du
2
= = (12.119)
x −4xu −4u2
The second equation of the above system after simplification becomes
dx du
=
x u
and after integration
u
= C2 (12.120)
x
The first equation of the system (12.119) after simplification, takes on the
form
dt dx
=
x −4u
and since u = C2 x,
dt dx 1
= , dt = − dx
x −4C2 x 4C2
167
and by integration
x x
t=− + C1 or t = − u + C1
4C2 4·
x
and thus
x2
t=− + C1 (12.121)
4u
We thus have obtained two independent solutions of the system (12.119)
given by (12.121) and (12.120) respectively. Therefore the general solution
of (12.118) is given by
x2 u
F t+ , =0 (12.122)
4u x
or
x2 u
t+ =f (12.123)
4u x
From the above equation we obtain
x2 = 4u·f (ξ) − 4tu (12.124)
where
u
ξ= (12.125)
x
Differentiation of (12.124) with respect to t and x, we obtain
u
ut = (12.126)
f (ξ) + ξf 0 (ξ) − t
and
x
ξ 2 f 0 (ξ) +
ux = 2 (12.127)
f (ξ) + ξf 0 (ξ) − t
respectively. Therefore the original equation ut = (ux )2 becomes
x
u ξ 2 f 0 (ξ) + 2
= 2
f (ξ) + ξf 0 (ξ) − t f (ξ) + ξf 0 (ξ) − t
168
0

2 0 x 2
u[f (ξ) + ξf (ξ) − t] = ξ f (ξ) +
2 2
x2

x (12.128)
⇔u + ξf 0 (ξ) = [ξ 2 f 0 (ξ)]2 + xξ 2 f 0 (ξ) +
4u 4
2 0 2
⇔[ξ f (ξ)] = 0
and thus f 0 (ξ) = 0. We then obtain
f (ξ) = C (12.129)
Equation (12.123) gives
x2
t+ =C
4u
and solving with respect to u we obtain
x2
u= (12.130)
4(C − t)
which is another general solution of the equation ut = (ux )2 .

(d) Finally we consider a linear combination of generators Γ6 and Γ2 : Γ6 +3Γ2 .
T = 1, X = −6t, U = 3x (12.131)
since
∂ ∂ ∂
Γ6 + 3Γ2 = − 6t + 3x (12.132)
∂t ∂x ∂u
The Invariant Surface Condition (12.78) takes on the form
ut − 6tux = 3x (12.133)
The corresponding Lagrange System is

dt dx du
= = (12.134)
1 −6t 3x
169
From the first equation of the above system we get −6tdt = dx and by
integration
x = −3t2 + C1 (12.135)
Equation dt = dx/3x because of (12.135) becomes

dt
dt = (12.136)
3(−3t2 + C1 )
and by integration
u = −3t3 + 3C1 t + C2 or u = −3t3 + 3(x + 3t2 )t + C2
and thus
u = 6t3 + 3xt + C2 (12.137)
Since we have two independent solutions of the system (12.134) given by

(12.135) and (12.137), the general solution of (12.133) is given by
F (x + 3t2 , u − 6t3 − 3xt) = 0 (12.138)
or
u − 6t3 − 3xt = f (x + 3t2 ) (12.139)
Introducing the auxiliary variable ξ by
ξ = x + 3t2 (12.140)
u = 6t3 + 3xt + f (ξ) (12.141)
From the above equation we calculate the partial derivatives ut and ux :
ut = 18t2 + 3x + 6tf 0 (ξ) and ux = 3t + f 0 (ξ) (12.142)
Therefore equation ut = (ux )2 becomes
3(3t2 + x) = [f 0 (ξ)]2
170
and thus
[f 0 (ξ)]2 = 3ξ (12.143)
√
We may take one solution f 0 (ξ) = 3ξ of the above equation, which when
integrated, yields
√
3 1
f (ξ) = √ +C (12.144)
2 ξ
We thus obtain using (12.141) and (12.144)

√
3 3 1
u = 6t + 3xt + √ +C (12.145)
2 x + 3t2
The above is one more general solution of the original equation ut = (ux )2 .
12.6. Using MAPLE. We shall solve the same equation ut = (ux )2 using
MAPLE. We use the following steps:
>restart;
>with(DEtools):
>with(PDEtools):
Step 2. We declare the dependent variable
> declare(u(x,t));
u(x, t) will now be displayed as u

We next write down the PDE we consider.
> P DE := diff(u(x, t), t) = (diff(u(x, t), x))2 ; (12.146)
ut = u2x (12.147)
Step 3. We find the determining equations using the command

> DetSys:=DeterminingPDE(P DE, integrabilityconditions=false);
[− ξ2u , − η1t , − − ξ1t − 2 − η1x , − ξ1u + 2 − ξ2x , − η1u + 2 − ξ1x −− ξ2t ] (12.148)
171
Step 4. We integrate the above system using the pdsolve command
> Inf:=pdsolve(DetSys);
1 2 1
− ξ1 (x, t, u) = − C4x + (− C1t +− C9 +− C2 −− C6u)x
2 2
1
+ (−4 − C7 − 4 − C4u)t − 2u− C5 +− C10
2
1 2 1
− ξ2 (x, t, u) = − C6x + (2 − C5 + 2 − C4t)x
2 2 (12.149)
1 2
+− C2t +− C3 + − C1t
2
1 2 1
− η1 (x, t, u) = − − C1x + (8 − C7 + 8 − C4u)x
8 8
− 2 − C6u2 +− C8 + u− C9
The infinitesimals however can be determined under a single command:
> Inf1:=Infinitesimals(P DE);
[− ξ1 (x, t, u) = 1, − ξ2 (x, t, u) = 0, − η1 (x, t, u) = 0],

[− ξ1 (x, t, u) = 0, − ξ2 (x, t, u) = 0, − η1 (x, t, u) = 1],
[− ξ1 (x, t, u) = 0, − ξ2 (x, t, u) = 1, − η1 (x, t, u) = 0],
[− ξ1 (x, t, u) = −2t, − ξ2 (x, t, u) = 0, − η1 (x, t, u) = x],
[− ξ1 (x, t, u) = −2u, − ξ2 (x, t, u) = x, − η1 (x, t, u) = 0],
1
[− ξ1 (x, t, u) = x, − ξ2 (x, t, u) = t, − η1 (x, t, u) = 0],
2 (12.150)
1
[− ξ1 (x, t, u) = x, − ξ2 (x, t, u) = 0, − η1 (x, t, u) = u],
2
1
[− ξ1 (x, t, u) = −2tx, − ξ2 (x, t, u) = −2t2 , − η1 (x, t, u) = x2 ],
2
1 1 2 1
[− ξ1 (x, t, u) = ux, − ξ2 (x, t, u) = − x , − η1 (x, t, u) = u2 ],
2 8 2
1 2
[− ξ1 (x, t, u) = x − 2tu, − ξ2 (x, t, u) = tx, − η1 (x, t, u) = ux]
2
The above is the set of ten generators of the Lie algebra of symmetries of the
given equation ut = u2x .
Step 5. We can consider now any generator or a linear combination of gen-
erators to determine some general solutions. Considering for example the
172
generator (ninth line above)
1 1 1
[− ξ1 (x, t, u) = ux, − ξ2 (x, t, u) = − x2 , − η1 (x, t, u) = u2 ]
2 8 2
we can associate the equation (the invariant surface condition)
−4xuux + x2 ut = −4u2 (12.151)
We enter the command

> eqn1:=−4 ∗ x ∗ u(x, t) ∗ diff(u(x, t), x) + x2 ∗ diff(u(x, t), t) = −4 ∗ u(x, t)2
−4xuux + x2 ut = −4u2
which can be solved using the pdsolve command. See case (c) above where
the generator used is essentially the Γ8 generator. The solution is given by
(12.130).
Step 5a. Considering next the linear combination of generators (third line+3·fourth
line above), i.e.
[− ξ1 (x, t, u) = −6t, − ξ2 (x, t, u) = 1, − η1 (x, t, u) = 3x]
we can associate the equation (the invariant surface condition)
−6tux + ut = 3x (12.152)
which is equation (12.133) and thus the solution of the original equation is
given in this case by (12.145).
Using Mathematica. We next demonstrate on how to use the Mathematica
package called MathLie to the equation ut = (ux )2 . The MathLie is an
AddOn package to Mathematica (Reference [4]). We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the given equation using the command
In[2] := eqn = ∂t U [x, t] − (∂x U [x, t])2 == 0; eqn//LT F

Ut − Ux2 = 0
In[3] := pro2 = P rolongation[eqn, {U }, {x, t}]; pro2//LT F
173
− Ux ξ1t − Ut Ux ξ1U + 2Ux3 ξ1U + 2Ux2 ξ1x − Ut ξ2t
− Ut2 ξ2U + 2Ut Ux2 ξ2U + 2Ut Ux ξ2x + φ1t + Ut φ1U
− 2Ux2 φ1U − 2Ux φ1x == 0
substitute in the above expression Ut by (Ux )2 using the command
In[4] := prolong2 = pro2/. ∂t U [x, t] → (∂x U [x, t])2 ; prolong2//LT F

− Ux ξ1t + Ux3 ξ1U + 2Ux2 ξ1x − Ux2 ξ2t + Ux4 ξ2U
+ 2Ux3 ξ2x + φ1t − Ux2 φ1U − 2Ux φ1x == 0
Step 5. We collect together terms arranged with respect to Ux , using the
command
In[5] := cprolong2 = Collect[prolong2, {∂x U [x, t]}];
cprolong2//LT F
Out[5] :=
Ux4 ξ2U + Ux3 (ξ1U + 2ξ2x ) + φ1t + Ux2 (2ξ1x − ξ2t − φ1U )
+ Ux (−ξ1t − φ1x ) == 0
In[6] := deteqn = DeterminingEquations[eqn, {U }, {x, t}, {∂t U [x, t]}];
deteqn//LT F
ξ2U == 0
φ1t == 0
ξ1U + 2ξ2x == 0
2ξ1x − ξ2t − φ1U == 0
− ξ1t − 2φ1x == 0
174
and thus to determine the infinitesimals of the equation.
In[6] := infeqn = Infinitesimals[eqn, {U }, {x, t}]; infeqn//LT F
φ1 == k5 − 4 k4 U 2 + U (k1 + 2 k10 x) + x (k6 + k7 x)

ξ1 == k8 + k9 x + k10 x2 − 2U (k3 + 2 k10 t + 2 k4 x)
− 2t(k6 + 2 k7 x)
ξ2 == k2 − 4 k7 t2 + x(k3 + k4 x) + t(−k1 + 2(k9 + k10 x))
to determine some solutions of the given equation. We consider the following
infinitesimals (for k2 = 1, k6 = 3 and all the rest k’s zero)
xi[1][x, t, U ] = −6 ∗ t;
xi[2][x, t, U ] = 1;
phi[1][x, t, U ] = 3 ∗ x;
We then determine the invariants corresponding to the above infinitesimals
In[7] := invar = xi[1]∗∂x F [x, t, U ]+xi[2]∗∂t F [x, t, U ]+φ[1]∗∂U F [x, t, U ] == 0
Out[7] := Ft + 3xFU − 6tFx == 0
Step 8. Solving the above equation,
In[8] := solution = DSolve[invar, F [x, t, U ], {x, t, U }]
we find

1 2 3
Out[8] : F (x, t, U )→C[1] (3t + x), −6t + U − 3tx (12.153)
6
From the above result it is obvious that we have two invariants:
ξ = 3t2 + x and F = −6t3 − 3tx + U
175
Step 9. We now have to substitute x by ξ − 3t2 and U by 6t3 + 3tx + F (ξ)
in order to find a solution of the equation

In[9] : rules = {x→ξ − 3 ∗ t , U →F unction {x, t}, 6 ∗ t3 + 3 ∗ t ∗ x
2

2
+ F [x + 3 ∗ t ] }

2 3 2
Out[9] := x→ − 3t + ξ, U →F unction {x, t}, 6t + 3tx + F [x + 3t ]
Step 10. We reduce next the equation under the above substitutions:
In[10] : reqn = eqn/.rules; reqn//LT F
The program responds with the following equation
Out[10] :=
18t2 + 3(−3t2 + ξ) + 6tFξ − (3t + Fξ )2 == 0
The above equation can be solved using the DSolve[] command. The reader
can see that this equation is equivalent to (Fξ )2 = 3ξ with one solution given
by
√
3 1
F (ξ) = √ +C
2 ξ
Taking into account the above solution and going back to the original vari-
ables, we arrive at the solution given by (12.145).
12.6. Example 2. Determine the symmetry generator of the differential
equation
ut = u u x , u = u(t, x) (12.154)
Solution. This is a first order partial differential equation. We define ∆ by
∆ = ut − uux (12.155)
176
P r(1) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x] ut − uux = 0 (12.156)
[t] [x]
⇔ U (−ux ) + U + U (−u) = 0
Substituting (12.14) and (12.15) into (12.156), we obtain the equation
− U ux + Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )
(12.157)
− u(Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu )) = 0
We now have to take into account the condition ∆ = 0. This means that we
have to substitute ut by uux . We then obtain
− U ux + Ut + uux Uu − uux (Tt + uux Tu ) − ux (Xt + uux Xu )

(12.158)
− u(Ux + ux Uu − uux (Tx + ux Tu ) − ux (Xx + ux Xu )) = 0
The previous equation can be written as, arranged in powers of ux
(Ut − uUx ) + (−U − Xt − uTt + uXx + u2 Tx )ux (12.159)
Equating the different powers of the derivative ux to zero, we obtain the

Ut − u Ux = 0 (12.160)
and
−U − Xt − u Tt + uXx + u2 Tx = 0 (12.161)
Equation Ut − u Ux = 0 admits the general solution

x + tu
U = F u, (12.162)
u
where F is an arbitrary function.
On the other hand, equation −U − Xt − u Tt + uXx + u2 Tx = 0 can be written
as
u(Xx + uTx ) − (Xt + uTt ) = U (12.163)
177
and using the substitution
G = X + uT (12.164)
we have to solve the equation
x + tu
uGx − Gt = F u, (12.165)
u
The above equation admits the general solution
x x + tu x + tu
G = ·F u, + H u, (12.166)
u u u
where H is another arbitrary function.
We thus see that we have the following expressions for the infinitesimals
X, T, U
x x + tu x + tu
X(x, t, u) = ·F u, + H u, − u T (x, t, u)
u u u
T (x, t, u) = T (x, t, u) (12.167)
x + tu
U (x, t, u) = F u,
u
The generator, because of the above expressions, takes on the form
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂ x x + tu x + tu ∂
=T + ·F u, + H u, −uT (12.168)
∂t u u u ∂x
x + tu ∂
+ F u,
u ∂u
where T = T (x, t, u).
13 Partial Differential Equations II

Second Order PDEs
In this Section we consider second order Partial Differential Equations
of functions with two variables.
13.1. The Invariance Condition. We first define the operator Γ by
∂ ∂ ∂
Γ=T +X +U (13.1)
∂t ∂x ∂u
178
The first and second extensions are given by
∂ ∂
P r(1) Γ = Γ + U [t] + U [x] (13.2)
∂ut ∂ux
and
∂ ∂ ∂
P r(2) Γ = P r(1) Γ + U [tt] + U [tx] + U [xx] (13.3)
∂utt ∂utx ∂uxx
respectively.
If we denote a second order partial differential equation of the function u(t, x)
by ∆,
∆≡F (t, x, u, ux , ut , utt , utx , uxx ) (13.4)
then the Invariance Condition reads
P r(2) Γ(∆)|∆=0 = 0 (13.5)
The extended transformations coefficients are given by

U [tt] = Dt (U [t] ) − utt Dt (T ) − utx Dt (X) (13.8)
U [tx] = Dt (U [x] ) − utx Dt (T ) − uxx Dt (X)
(13.9)
= Dx (U [t] ) − utt Dx (T ) − utx Dx (X)
U [xx] = Dx (U [x] ) − utx Dx (T ) − uxx Dx (X) (13.10)
where Dt and Dx are given by
∂ ∂ ∂ ∂
Dt = + ut + utt + utx + ··· (13.11)
and
∂ ∂ ∂ ∂
Dx = + ux + utx + uxx + ··· (13.12)
179
respectively.
We are now able to calculate the explicit expression for U [xx] :
U [xx] = Dx (U [x] ) − utx Dx (T ) − uxx Dx (X)

∂ ∂ ∂ ∂
= + ux + utx + uxx [Ux + (Uu − Xx )ux
− Xu (ux )2 − Tx ut − Tu ux ut ]
∂ ∂ ∂ ∂
− utx + ux T − uxx + ux X
∂x ∂u ∂x ∂u
= Uxx + (Uux − Xxx )ux − Xux (ux )2 − Txx ut − Tux ux ut
+ ux [Uxu + (Uuu − Xxu )ux − Xuu (ux )2 − Txu ut − Tuu ux ut ]
+ utx (−Tx − Tu ux ) + uxx {(Uu − Xx ) − Tu ut − 2Xu ux }
− utx (Tx + ux Tu ) − uxx (Xx + ux Xu )
Collecting terms together and rearranging, we have the following explicit

expression for U [xx] :
U [xx] = Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux

+ (Uuu − 2Xxu )(ux )2 − Tuu ut (ux )2 − Xuu (ux )3
(13.13)
− 2Tx utx + (Uu − 2Xx )uxx − 2Tu ux utx
− Tu ut uxx − 3Xu ux uxx
Let us now calculate U [tt] :

U [tt] = Dt (U [t] ) − utt Dt (T ) − utx Dt (X)
∂ ∂ ∂ ∂
= + ut + utt + utx [Ut + (Uu − Tt )ut
− Tu (ut )2 − Xt ux − Xu ut ux ]
∂ ∂ ∂ ∂
− utt + ut T − utx + ut X
∂t ∂u ∂t ∂u
= Utt + (Uut − Ttt )ut − Tut (ut )2 − Xtt ux − Xut ut ux
+ ut [Utu + (Uuu − Ttu )ut − Tuu (ut )2 − Xtu ux − Xuu ut ux ]
+ utt {(Uu − Tt ) − 2Tu ut − Xu ux } + utx (−Xt − Xu ut )
− utt (Tt + ut Tu ) − utx (Xt + ut Xu )
180
expression for U [tt] :
U [tt] = Utt − Xtt ux + (2Utu − Ttt )ut − 2Xtu ux ut

+ (Uuu − 2Ttu )(ut )2 − Xuu ux (ut )2 − Tuu (ut )3
(13.14)
− 2Xt utx + (Uu − 2Tt )utt − 2Xu ut utx
− Xu ux utt − 3Tu ut utt
We finally calculate U [tx] :
U [tx] = Dt (U [x] ) − utx Dt (T ) − uxx Dt (X)

∂ ∂ ∂ ∂
= + ut + utt + utx [Ux + (Uu − Xx )ux
− Xu (ux )2 − Tx ut − Tu ux ut ]
∂ ∂ ∂ ∂
− utx + ut T − uxx + ut X
∂t ∂u ∂t ∂u
= Utx + (Uut − Xxt )ux − Xut (ux )2 − Txt ut − Tut ux ut
+ ut [Uxu + (Uuu − Xxu )ux − Xuu (ux )2 − Txu ut − Tuu ux ut ]
+ utt (−Tx − Tu ux ) + utx {(Uu − Xx ) − 2Xu ux − Tu ut }
− utx (Tt + ut Tu ) − uxx (Xt + ut Xu )
expression for U [tx] :
U [tx] = Utx + (Uxu − Txt )ut + (Utu − Xxt )ux

+ (Uuu − Ttu − Xxu )ux ut + (Uu − Xx − Tt )uxt
− (Xu ux + Tu ut )uxt − Xt uxx − Tx utt (13.15)
− Xuu ut (ux )2 − Tuu ux (ut )2 − Tu ux utt
− Txu (ut )2 − Xtu (ux )2 − Xu ut uxx
13.2. Example 1. The Burgers Equation. Analyse the symmetries of
the Burgers equation
ut + uux = uxx (13.16)
Solution. We denote by ∆ the expression
∆≡ut + uux − uxx (13.17)
181
We have for the second extension P r(2) Γ
P r(2) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x]
(13.18)
[tt] ∂ [tx] ∂ [xx] ∂

+U +U +U ut + uux − uxx = 0
⇔ U ux + U [t] + U [x] u − U [xx] = 0
Using the known expressions of U [t] , U [x] and U [xx] , the previous equation
becomes

U ux + Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )

+ Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu ) u

− Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux (13.19)

− Tu ut uxx − 3Xu ux uxx = 0

ut by −uux + uxx . We then obtain from the above equation
U ux

2
+ Ut + (Uu − Tt − ux Xu )(−uux + uxx ) − Tu (−uux + uxx ) − Xt ux

+ Ux + ux Uu − (Tx + ux Tu )(−uux + uxx ) − ux (Xx + ux Xu ) u

− Uxx + (2Uxu − Xxx )ux − (Txx + 2Txu ux )(−uux + uxx )
+ (Uuu − 2Xxu )(ux )2 − Tuu (−uux + uxx )(ux )2 − Xuu (ux )3


− Tu (−uux + uxx )uxx − 3Xu ux uxx = 0
182
(13.20)
We have to equate to zero the coefficients of the partial derivatives of the

unknown function u(x, t). We can however simplify the whole procedure by
observing that the coefficients of utx and ux utx are −2Tx and −2Tu respec-
tively. Therefore since Tx = Tu = 0, we see that T does not depend on x and
u and thus
T = T (t) (13.21)
Therefore equation (13.20) becomes

U ux + Ut + (Uu − Tt − ux Xu )(−uux + uxx ) − Xt ux

+ Ux + ux Uu − ux (Xx + ux Xu ) u
(13.22)
− Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3

+ (Uu − 2Xx )uxx − 3Xu ux uxx = 0
The coefficient of ux uxx is 2Xu and since Xu = 0, we see that
X = X(x, t) (13.23)
Therefore equation (13.22) takes on the form

U ux + Ut + (Uu − Tt )(−uux + uxx ) − Xt ux

+ Ux + ux Uu − Xx ux u (13.24)

2
− Uxx + (2Uxu − Xxx )ux + Uuu (ux ) + (Uu − 2Xx )uxx = 0
Finally we get from the above equation that

The coefficient of (ux )2 is
Uuu = 0 (13.25)
183
The coefficient of uxx is
2Xx − Tt = 0 (13.26)
The coefficient of ux is
U + u(Tt − Xx ) − Xt − (2Uxu − Xxx ) = 0 (13.27)
The constant term is
Ut + uUx − Uxx = 0 (13.28)
From equation (13.25) we get that U is a linear function with respect to u:
U = α(x, t)u + β(x, t) (13.29)
where α(x, t) and β(x, t) are functions to be determined. Equation (13.26)

upon integration with respect to x gives
1
X = T 0 (t)x + γ(t) (13.30)
2
Equation (13.28), because of (13.29) becomes
αx u2 + (βx + αt − αxx )u + βt − βxx = 0 (13.31)
From the above equation, equating to zero the coefficients of u, we get the
system
αx = 0, βx + αt − αxx = 0, βt − βxx = 0
α = δ(t), βx = −δ 0 (t), βt − βxx = 0 (13.32)
We get further from the above system by integrating the second equation
with respect to x and substituting to the last equation:
α = δ(t), β = −δ 0 (t)x + (t), −δ 00 (t)x + 0 (t) = 0 (13.33)
Equating to zero the coefficients of x to the last equation, we get δ 00 (t) = 0

and 0 (t) = 0 and then δ(t) = a1 t + a2 , (t) = a3 . Therefore
α = a1 t + a2 and β = −a1 x + a3 (13.34)
184
We thus see that the function U is given by
U = (a1 t + a2 )u − a1 x + a3 (13.35)
taking into account (13.29) and (13.34).

Let us now turn into equation (13.27). This equation takes on the form,
taking into account the previous relations (13.30) and (13.35):
1 1
T 0 (t) + a1 t + a2 u − T 00 (t) + a1 x + a3 − γ 0 (t) = 0 (13.36)
2 2
Equating to zero the coefficients, we obtain the system
1 0 1 00
T (t) + a1 t + a2 = 0, T (t) + a1 = 0, a3 − γ 0 (t) = 0 (13.37)
2 2
Integration of the above equations yields
T (t) = −a1 t2 − 2a2 t + a4 , γ(t) = a3 t + a5 (13.38)
Taking into account the above expressions we obtain from (13.30) the explicit
expression of X:
X = −(a1 t + a2 )x + a3 t + a5 (13.39)
We then have the following expression for the generator

∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂ ∂
= (−a1 t2 − 2a2 t + a4 ) + [−(a1 t + a2 )x + a3 t + a5 ] (13.40)
∂t ∂x
∂
+ [(a1 t + a2 )u − a1 x + a3 ]
∂u
The above can also be written as
2 ∂ ∂ ∂

Γ = a1 −t − tx + (tu − x)
∂t ∂x ∂u
∂ ∂ ∂ ∂ ∂
+ a2 −2t − x +u + a3 t + (13.41)
∂t ∂x ∂u ∂x ∂u
∂ ∂
+ a4 + a5
∂t ∂x
185
We thus obtain the following generators of the Lie algebra of symmetries for
the Burgers equation
∂ ∂ ∂ ∂ ∂ ∂
Γ1 = −t2 − tx + (tu − x) , Γ2 = −2t − x +u ,
∂t ∂x ∂u ∂t ∂x ∂u (13.42)
3 ∂ ∂ 4 ∂ 5 ∂
Γ =t + , Γ = , Γ =
∂x ∂u ∂t ∂x
(i) We consider first the generator Γ1 and the associated invariant surface
condition T ut + Xux = U :
−t2 ut − txux = tu − x (13.43)
Two independent solutions of the auxiliary Lagrange system

dt dx du
2
= =
−t −tx tu − x
are given by
x
= C1 , tu − x = C2
t
Therefore the general solution of the invariant surface condition (13.43) is
given by F (x/t, tu − x) = 0, i.e.
x
tu − x = H
t
or solving with respect to u,
x 1 x
u = + ·H (13.44)
t t t
Using the above expression into the original equation ut + uux = uxx , we
obtain the differential equation the function H should satisfy
H 00 (ξ) − H(ξ)H 0 (ξ) = 0 (13.45)
where ξ is the similarity variable given by

x
ξ=
t
186
Once we determine H(ξ) from (13.45), we could determine a general solution
of Burgers equation using (13.44). The general solution of this equation is
given by
√
2 ξ + C
2
H(ξ) = ·tan √
C1 2 C1
Therefore the general solution of Burgers equation is given by
√
x 2 1 x + C t
2
u= + · ·tan √ (13.46)
t C1 t 2 C1 t
combining (13.44) and the solution of (13.45).
(ii) We consider next the generator Γ2 and the associated invariant surface
condition T ut + Xux = U :
−2t ut − x ux = u (13.47)
Two independent solutions of the auxiliary Lagrange system

dt dx du
= =
2t x −u
are given by
x √
√ = C1 , u t = C2
t
Therefore the √
general
√ solution of the invariant surface condition (13.47) is
given by F (x/ t, u t) = 0, i.e.
√ x
u t=G √
t
or solving with respect to u,
1 x
u= √ G √ (13.48)
t t
Using the above expression into the original equation ut + uux = uxx , we
obtain the differential equation the function G should satisfy
2G00 (η) + ηG0 (η) − 2G(η)G0 (η) + G(η) = 0 (13.49)
187
where η is the similarity variable given by
x
η=√
t
Once we determine G(η) from (13.49), we could determine a general solution
of Burgers equation using (13.48).
Equation (13.49) can be written as 2G00 + (ηG)0 − (G2 )0 = 0 which upon
integration gives
2G0 + ηG − G2 = C
The above is a Riccati type of equation which can be transformed into a
second order linear differential equation
2w00 + ηw0 + Cw = 0
w0
under the substitution G = −2 , where prime denotes differentiation with
w
respect to η. The general solution of the above 2nd order ODE is being
expressed in terms of Kummer’s M and U functions:
2 − C 3 η2 2 − C 3 η 2 η2
w(η) = η C1 ·M , ; + C2 ·U , ; exp −
2 2 4 2 2 4 4
Calculating w0 we find that G is given by
2 Z(C, η) + C 0 W (C, η)
G(η) = (13.50)
η M (C, η) + C 0 U (C, η)
where Z and W are given by
2 − C 3 η2 C 3 η2
Z(C, η) = CM , ; − (1 + C)M − , ;
2 2 4 2 2 4
and
2 − C 3 η2 C 3 η2
W (C, η) = CU , ; − (1 + C)U − , ;
2 2 4 2 2 4
with C 0 = C2 /C1 . √
Using (13.48) and (13.50) and taking into account that η = x/ t, we find
the general solution of Burgers equation to have the form
√ √
2 Z(C, x/ t) + C 0 W (C, x/ t)
u(x, t) = √ √ (13.51)
x M (C, x/ t) + C 0 U (C, x/ t)
188
where M and U are the original Kummer’s functions appearing in the ex-
pression of w(η).
13.3. Example 2. The potential Burgers Equation. Analyse the
symmetries of the potential Burgers equation
1
ut + (ux )2 = uxx (13.52)
2
1
∆≡ut + (ux )2 − uxx (13.53)
2
We have for the second extension Γ(2)
Γ(2) (∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x]
(13.54)
[tt] ∂ [tx] ∂ [xx] ∂ 1

2
+U +U +U ut + (ux ) − uxx = 0
∂utt ∂utx ∂uxx 2
[t] [x] [xx]
⇔ U + U ux − U =0
Using the known expressions for U [t] , U [x] and U [xx] , the above equation takes
on the form

+ Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu ) ux

− Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux
(13.55)
2 2 3
+ (Uuu − 2Xxu )(ux ) − Tuu ut (ux ) − Xuu (ux )


1
ut by − (ux )2 + uxx . We then obtain from equation (13.55)
2
189
1
Ut + (Uu − Tt − Xu ux ) − (ux )2 + uxx
2
1 2
− Tu − (ux )2 + uxx − Xt ux
2
1
2
+ Ux + ux Uu − (Tx + ux Tu ) − (ux ) + uxx
2

− ux (Xx + ux Xu ) ux
(13.56)
1
− Uxx + (2Uxu − Xxx )ux − (Txx + 2Txu ux ) − (ux )2 + uxx
2
1
2
+ (Uuu − 2Xxu )(ux ) − Tuu − (ux ) + uxx (ux )2
2
2
− Xuu (ux )3 − 2Tx utx + (Uu − 2Xx )uxx − 2Tu ux utx
1
2
− Tu − (ux ) + uxx uxx − 3Xu ux uxx = 0
2
We now have to equate to zero the coefficients of the partial derivatives of
the unknown function u(x, t). However we can simplify calculations once we
observe that the coefficient of utx and ux utx are −2Tx and −2Tu respectively
and since Tx = Tu = 0, we see that T depends on t only:
T = T (t) (13.57)

1
Ut + (Uu − Tt − Xu ux ) − (ux )2 + uxx − Xt ux
2
+ Ux + ux Uu − ux (Xx + ux Xu ) ux
(13.58)

The coefficient of ux uxx is 2Xu and since Xu = 0, we see that
X = X(x, t) (13.59)
190
1
Ut + (Uu − Tt ) − (ux )2 + uxx − Xt ux
2
+ Ux + (Uu − Xx )ux ux (13.60)

2
From the above equation we obtain that

1
− (Uu − Tt ) + (Uu − Xx ) − Uuu = 0 (13.61)
2
2Xx − Tt = 0 (13.62)
−Xt + Ux − (2Uxu − Xxx ) = 0 (13.63)
Ut − Uxx = 0 (13.64)
Upon combining (13.61) with (13.62) we obtain the equation
Uu − 2Uuu = 0 (13.65)
U = α(x, t) + β(x, t)eu/2 (13.66)
Integrating (13.62) with respect to x we find

1
X = T 0 (t)x + γ(t) (13.67)
2
From (13.63), using (13.67) and (13.66) we get
1
αx = T 00 (t)x + γ 0 (t) (13.68)
2
191
Integrating the above equation with respect to x we find
1
α(x, t) = T 00 (t)x2 + γ 0 (t)x + δ(t) (13.69)
4
From equation (13.64) upon substituting (13.66), we get the two equations
αt − αxx = 0 (13.70)
and
βt − βxx = 0 (13.71)
Equation (13.70), because of (13.69) takes on the form

1 000 1
T (t)x2 + γ 00 (t)x + δ 0 (t) − T 00 (t) = 0 (13.72)
4 2
Equating the coefficients of x to zero, we obtain from the above equation the
system
1
T 000 (t) = 0, γ 00 (t) = 0, δ 0 (t) − T 00 (t) = 0 (13.73)
2
The above system, upon integration gives
T (t) = a1 t2 + a2 t + a3 , γ(t) = a4 t + a5 , δ(t) = a1 t + a6 (13.74)
Using the above expressions, we get from (13.67) and (13.69) that
1
X = (2a1 t + a2 )x + a4 t + a5 (13.75)
2
and
1
α(x, t) = a1 x2 + a4 x + a1 t + a6 (13.76)
2
respectively.
Finally, we get from (13.66) that
1
U= a1 x2 + a4 x + a1 t + a6 + β(x, t)eu/2 (13.77)
2
192
where β(x, t) satisfies the heat equation (13.71): βx = βxx .
The generator of the Lie algebra of symmetries is then given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
2 ∂ 1 ∂
= (a1 t + a2 t + a3 ) + (2a1 t + a2 )x + a4 t + a5 (13.78)
∂t 2 ∂x
1 ∂
+ a1 x2 + a4 x + a1 t + a6 + β(x, t)eu/2
2 ∂u
The previous expression can be written as

∂ ∂ 1 ∂ ∂ 1 ∂
Γ = a1 t2 + tx + x2 + t + a2 t + x
∂u ∂x 2 ∂u ∂t 2 ∂x (13.79)
∂ ∂ ∂ ∂ ∂ u/2 ∂
+ a3 + a4 t +x + a5 + a6 + β(x, t)e
∂t ∂x ∂u ∂x ∂u ∂u
Therefore the Lie algebra of symmetries is being spanned by the six genera-
tors
2 ∂ ∂ 1 ∂ ∂ 1 ∂
1 2
Γ =t + tx + x +t , Γ2 = t + x
∂u ∂x 2 ∂u ∂t 2 ∂x (13.80)
3 ∂ 4 ∂ ∂ 5 ∂ 6 ∂
Γ = , Γ =t +x , Γ = , Γ =
∂t ∂x ∂u ∂x ∂u
and the generator
∂
Γ∞ = β(x, t)eu/2 (13.81)
∂u
which spans an infinite dimensional subalgebra, where β(x, t) is any solution
of the heat equation βt = βxx .
(i) We consider first the generator Γ1 and the associated invariant surface
condition
1
t2 ut + txux = x2 + t (13.82)
2
The corresponding Lagrange system
dt dx 2du
2
= = 2 (13.83)
t tx x + 2t
193
admits two independent solutions
x x2
= C1 and u − − ln(t) = C2 (13.84)
t 2t
x x2
F , u− − ln(t) = 0
t 2t
from which we get
x2 x
u− − ln(t) = H
2t t
and thus
x2 x
u= + ln(t) + H (13.85)
2t t
Upon substitution of the above expression into the original equation
1
ut + (ux )2 = uxx , we find the following differential equation which the
2
function H should satisfy
2H 00 (ξ) − (H 0 (ξ))2 = 0 (13.86)
where ξ = x/t is the similarity variable. The solution of the above equation
is given by
H(ξ) = −2 ln(C1 ξ + C2 ) (13.87)
Using (13.85) and (13.87) we see that the general solution of equation ut +
1
(ux )2 = uxx is given by
2
x2 x
u= + ln(t) − 2 ln C1 + C2 (13.88)
2t t
(ii) We consider next the generator Γ2 and the associated invariant surface
condition
1
tut + xux = 0 (13.89)
2
194
Two independent solutions of Lagrange’s system
dt 2dx du
= = (13.90)
t x 0
are given by
x
√ = C1 , u = C2 (13.91)
t
√
Therefore the equation (13.) admits the general solution F (x/ t, u) = 0 or
x
u=G √ (13.92)
t
Upon substitution of the above expression into the original equation
1
ut + (ux )2 = uxx we find the following differential equation which the func-
2
tion G should satisfy
2G00 (η) − (G0 (η))2 + η G0 (η) = 0 (13.93)

√
where η = x/ t is the similarity variable. The solution of the above equation
is given by (erf(x) is the error function)
η
G(η) = −2 ln C1 erf + C2 (13.94)
2
Using (13.92) and (13.94), we find the general solution of potential Burgers
1
equation ut + (ux )2 = uxx to have the form
2
x
u(x, t) = −2 ln C1 erf √ + C2 (13.95)
2 t
13.3. Example 3. The Heat Equation. Analyse the symmetries of the
heat equation
ut = uxx (13.96)
∆ = ut − uxx (13.97)
195
We determine the second extension P r(2) Γ. We have
P r(2) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x]
(13.98)
∂ ∂ ∂
+ U [tt] + U [tx] + U [xx] ut − uxx = 0
[t] [xx]
⇔ U −U =0
Using the explicit expressions for U [t] and U [xx] given by (12.14) and (13.13)
respectively, we write (13.98) in its equivalent form
− Uxx + Txx ut − (2Uxu − Xxx )ux + 2Txu ut ux
− (Uuu − 2Xxu )(ux )2 + Tuu ut (ux )2 + Xuu (ux )3 (13.99)
+ 2Tx utx − (Uu − 2Xx )uxx + 2Tu ux utx
+ Tu ut uxx + 3Xu ux uxx = 0
We now have to take into account the condition ∆ = 0, which means that
we have to substitute ut by uxx in (13.99). We thus obtain the equation
Ut + uxx Uu − uxx (Tt + uxx Tu ) − ux (Xt + uxx Xu )
− Uxx + Txx uxx − (2Uxu − Xxx )ux + 2Txu uxx ux
− (Uuu − 2Xxu )(ux )2 + Tuu uxx (ux )2 + Xuu (ux )3
+ Tu (uxx )2 + 3Xu ux uxx = 0
The above equation can be arranged in powers of the partial derivatives of
the function u as
(Xxx − Xt − 2Uxu )ux + (2Xx − Tt + Txx )uxx
+ 2(Xu + Txu )ux uxx + (2Xxu − Uuu )(ux )2
(13.100)
+ Xuu (ux )3 + Tuu (ux )2 uxx + 2Tx utx + 2Tu ux utx
+ Ut − Uxx = 0
We now equate to zero the coefficients of the derivatives of the function u.
Coefficient of ux :
Xxx − Xt − 2Uxu = 0 (13.101)
196
Coefficient of uxx :
2Xx − Tt + Txx = 0 (13.102)
Coefficient of ux uxx :
Xu + Txu = 0 (13.103)
Coefficient of (ux )2 :
2Xxu − Uuu = 0 (13.104)
Xuu = 0 (13.105)
Coefficient of (ux )2 uxx :
Tuu = 0 (13.106)
Coefficient of utx :
Tx = 0 (13.107)
Coefficient of ux utx :
Tu = 0 (13.108)
Final term with no derivatives of u:
Ut − Uxx = 0 (13.109)
Let us now try to solve the above system of equations (13.101)-(13.109). We

use the following steps:
Step (a). Equation Tu = 0 implies Tuu = 0. Equation Tx = 0 implies Txu = 0
and thus equation (13.103) implies Xu = 0 and thus Xuu = 0. On the other
hand Xxu = 0 and then Uuu = 0 as follows from (13.104).
Therefore equations (13.103)-(13.108) imply that
Tx = 0, Tu = 0, Xu = 0 and Uuu = 0 (13.110)
197
Step (b). Equations (13.110) imply that
T = α(t), X = X(t, x) and U = β(t, x)u + γ(t, x) (13.111)
Since Txx = 0, we obtain from (13.102) that 2Xx = α0 (t) and by integration
1
X = α0 (t)x + δ(t) (13.112)
2
Step (c). We obtain from (13.112) that Xxx = 0 and
1
Xt = α00 (t)x + δ 0 (t) (13.113)
2
We also obtain from the last of (13.111) that Uu = β(t, x) and thus Uxu =
βx (t, x). Therefore we obtain from (13.101) that
1
− α00 (t)x − δ 0 (t) − 2βx = 0
2
or solving with respect to βx ,
1 1
βx = − α00 (t)x − δ 0 (t) (13.114)
4 2
Integrating the previous equation with respect to x, we find
1 1
β(t, x) = − α00 (t)x2 − δ 0 (t)x + (t) (13.115)
8 2
Step (d). We obtain from (13.115) and (13.114)
1 1
βt = − α000 (t)x2 − δ 00 (t)x + 0 (t) (13.116)
8 2
and
1
βxx = − α00 (t) (13.117)
4
respectively. We thus have from (13.109), using the last of (13.111) that
(βt u + γt ) − (βxx u + γxx ) = 0
198
and then using (13.116) and (13.117),
1 1
{− α000 (t)x2 − δ 00 (t)x + 0 (t)}u + γt
8 2
1 00
− {− α (t)u + γxx } = 0
4
1 1 1
− α000 (t)x2 u − δ 00 (t)xu + { α00 (t) + 0 (t)}u
8 2 4 (13.118)
+ (γt − γxx ) = 0
Equating to zero all the coefficients of the above equation, we obtain the
system
1 00
α000 (t) = 0, δ 00 (t) = 0, α (t) + 0 (t) = 0, γt − γxx = 0 (13.119)
4
We thus obtain from the first two equations that
α(t) = a3 t2 + a2 t + a1 (13.120)
and
δ(t) = a4 t + a5 (13.121)
respectively.
We further obtain from the third of (13.119) that
1
0 (t) = − a3
2
and by integration
1
(t) = − a3 t + a6 (13.122)
2
The last of (13.119) implies that the function γ is an arbitrary solution of
the original heat equation.
Step (e). From equation (13.115), using (13.120), (13.121) and (13.122), we
obtain the explicit expression of β(t, x):
1 1 1
β(t, x) = − x2 − t a3 − xa4 + a6 (13.123)
4 2 2
199
Therefore we obtain from the last of (13.111)

1 2 1 1
U = − x − t a3 − xa4 + a6 u + γ(t, x) (13.124)
4 2 2
Using (13.120), we obtain using the first of (13.111)
T = a3 t2 + a2 t + a1 (13.125)
Finally we obtain from (13.112), using (13.120) and (13.121) that

1
X = (2a3 t + a2 )x + (a4 t + a5 ) (13.126)
2
Using the explicit expressions for T , X and U , we are able to find the explicit
expression of the operator
∂ ∂ ∂
Γ=T +X +U (13.127)
∂t ∂x ∂u
We find that
∂
Γ = (a3 t2 + a2 t + a1 )
∂t
1 ∂
+ (2a3 t + a2 )x + (a4 t + a5 ) (13.128)
2 ∂x

1 2 1 1 ∂
+ − x − t a3 − xa4 + a6 u + γ(t, x)
4 2 2 ∂u

∂ ∂ 1 ∂
Γ = a1 + a2 t + x
∂t ∂t 2 ∂x
2 ∂ ∂ 1
2 1 ∂
+ a3 t + tx − x + t u
∂t ∂x 4 2 ∂u
∂ (13.129)
1 ∂ ∂ ∂
+ a4 t − xu + a5 + a6 u
∂x 2 ∂u ∂x ∂u
∂
+ γ(t, x)
∂u
200
From the previous expression we can read off the generators of the Lie sym-
metry algebra for the heat equation (six-dimensional algebra):
∂ ∂ 1 ∂
Γ1 = , Γ2 = t + x
∂t ∂t 2 ∂x
∂ ∂ 1 1 ∂
Γ3 = t2 + tx − x2 + t u (13.130)
∂t ∂x 4 2 ∂u
∂ 1 ∂ ∂ ∂
Γ4 = t − xu , Γ5 = , Γ6 = u
∂x 2 ∂u ∂x ∂u
We also have the infinite-dimensional subalgebra spanned by the generator
Γ∞ ,
∂
Γ∞ = γ(t, x) (13.131)
∂u
where γ(t, x) is an arbitrary solution of the heat equation.
13.2a. Solutions of the Heat Equation. We can now determine solutions
to the heat equation by considering one generator or a linear combination of
the above generators:
Solution (a). We consider the generator Γ2 . In this case we have T = t,
X = x/2 and U = 0. The Invariant Surface Condition becomes
1
tut + xux = 0 or 2tut + xux = 0 (13.132)
2
Lagrange’s equations become in this case
dt dx du
= = (13.133)
2t x 0
We have two independent solutions of the above system
x
u = C1 and √ = C2 (13.134)
t
x
F u, √ = 0 (13.135)
t
or
x
u=f √ (13.136)
t
201
Introducing the new variable
x
ξ=√ (13.137)
t
we find for the partial derivatives
1 x 0 1
ut = − √ f (ξ) and uxx = f 00 (ξ) (13.138)
2t t t
Therefore equation ut = uxx becomes
1 x 1
− √ f 0 (ξ) = f 00 (ξ) or f 00 (ξ) = − ξf 0 (ξ) (13.139)
2 t 2
Integrating the last equation, we obtain
ξ2
f 0 (ξ) = A× exp − (13.140)
4
and by a second integration we get the following solution
f (ξ) = A·erf(ξ) + B (13.141)
where erf(x) is the error function.

Going back to the original variables, we obtain the following solution of the
heat equation
x
u(x, t) = A·erf √ + B (13.142)
t
Solution (b). We consider next the generator Γ3 . In this case we have
1 1
T = t2 , X = tx, U = − x2 + t u (13.143)
4 2
To the associated Invariant Surface Condition
1 1
t2 ut + txux = − x2 + t u (13.144)
4 2
there corresponds the following Lagrange system
dt dx du
= = − (13.145)
t2 tx
1 1
− x2 + t u
4 2
202
From the first two ratios of (13.145) we get the solution
x
= C1 (13.146)
t
The first and third ratios, taking into account that x = C1 t, take on the form

du 2 2
−4 = (C1 ) + dt (13.147)
u t
and by integration
−4 ln(u) = (C1 )2 t + 2 ln(t) + C2
or
x2 √
ln(u) = − − ln( t) + C2 (13.148)
4t
√ x2
u· t = C2 ·exp − (13.149)
4t
Since we have found two independent solutions of the Lagrange equations
given by (13.146) and (13.149), the general solution of the Invariant Surface
Condition (13.144) is given by
√ x2 x
F u· t·exp , =0 (13.150)
4t t
or
√ x2 x
u· t·exp =f (13.151)
4t t
1 x2 x
u = √ exp − f (13.152)
t 4t t
Introducing the auxiliary parameter ξ by
x
ξ= (13.153)
t
203
one can calculate the partial derivatives ut and uxx from (13.152):
1 1 x2 x2

x 0
ut = √ − 2 f (ξ) − 2 √ f (ξ) exp − (13.154)
4t t t t t 4t
and
1 1 x2 x2

1 00 0
uxx = √ − 2 f (ξ) + 2 √ f (ξ) − xf (ξ) exp − (13.155)
4t t t t t 4t
The heat equation ut = uxx , because of (13.154) and (13.155) takes on the
form (after simplifications)
f 00 (ξ) = 0 (13.156)
Therefore
x
f (ξ) = C1 ξ + C2 = C1 + C2 (13.157)
t
We thus obtain from (13.152) that
x 1 x2
u = C1 + C2 √ exp − (13.158)
t t 4t
which is a general solution of the heat equation.
Solution (c). We now find the solutions of the heat equation considering the
generator
∂ ∂ ∂
Γ2 + aΓ6 = x + 2t + 2au (13.159)
∂x ∂t ∂u
We determine the invariants of the symmetry transformation. The system
corresponding to the above generator is given by
dx dt du
= = (13.160)
x 2t 2au
dx dt 1
From equation = we obtain by integration ln(x) = ln(t) + A or
√ x 2t 2
x = C1 t and thus one invariant is given by
x
y=√ (13.161)
t
204
dt du
Integrating the equation = we obtain a· ln(t) = ln(u)+B or ta = C2 ·u
2t 2au
and thus a second invariant is given by
v = t−a ·u (13.162)
We are now able to transform the heat equation into an ordinary differential
equation. The new function will be v and will contain y as an independent
variable. We have that
∂ a dv ∂y
ut = (t ·v) = ata−1 ·v + ta
∂t dy ∂t
1 x 1
= ata−1 ·v + ta vy − √ = ata−1 ·v − ta−1 yvy (13.163)
2 t t 2
1
= ta−1 av − yvy
2
and
∂ a dv ∂y 1
ux = (t v) = ta = ta vy √ (13.164)
∂x dy ∂x t
∂ a 1 ta d ∂y
uxx = t vy √ = √ (vy )
∂x t t dy ∂x
a (13.165)
t 1
= √ vyy √ = ta−1 vyy
t t
The heat equation ut = uxx then transforms into
a−1
1
t av − yvy = ta−1 vyy
2
from which we get the equation
1
vyy + yvy − av = 0 (13.166)
2
The substitution
1
2
w(y) = v·exp y (13.167)
8
transforms (13.166) into the equation

1 1 2
wyy = a + + y ·w (13.168)
4 16
205
The substitution y = λz gives
1 1
wy = wz and wyy = wzz (13.169)
λ λ2
and thus equation (13.168) transforms into

2 1 1 4 2
wzz = λ a + + λ z ·w (13.170)
4 16
√
The choice λ = 2 transforms the above equation into Weber’s differential
equation

1 1 2
wzz = 2a + + z ·w (13.171)
2 4
The general solution of the above equation is

1 1
w(z) = A·U 2a + , z + B·V 2a + , z (13.172)
2 2
where U and V are the parabolic cylindrical functions (Abramowitz and
Stegun §19.1) and A, B are arbitrary constants. We thus find that the general
solution of (13.168) is given by
1 y 1 y
w(z) = A·U 2a + , √ + B·V 2a + , √ (13.173)
2 2 2 2
Going back to the original variables, we determine the general solution of the
heat equation:

a
1 x 1 x
u(x, t) = t A·U 2a + , √ + B·V 2a + , √
2 2t 2 2t
x2 (13.174)
×exp −
8t
Solution (d). We find next the solutions of the heat equation considering the
generator
∂ ∂ ∂
Γ1 − Γ4 = − 2t + xu (13.175)
∂t ∂x ∂u
206
We consider first the auxiliary system
dt dx du
= = (13.176)
1 −2t xu
dt dx
associated to the symmetry (13.175). From equation = we obtain by
1 −2t
integration −t2 = x + C1 and thus one invariant of the system is given by
y = x + t2 (13.177)
dt du
Equation = is being written, taking into account (13.177) (y − t2 )dt =
1 xu
du
and by integration we get
u
t3 t3
yt − = ln(u) + A⇔(x + t2 )t − = ln(u) + A
3 3
2t3
⇔xt + = ln(u) + A
3
and thus
2t3
u = C2 ·exp xt + (13.178)
3
Therefore another invariant of the equation is given by
2t3
v = u·exp −xt − (13.179)
3
We are now able to transform the heat equation into an ordinary differential
equation. The new function v will contain y as independent variable.
Let us now express the partial derivatives ut and uxx in terms of the new
variables. We have
2t3

∂
ut = v·exp xt +
∂t 3
dv ∂y
= exp(...) + (x + 2t2 )v·exp(...) (13.180)
dy ∂t
= exp(...)vy ·2t + (x + 2t2 )v·exp(...)
= [2tvy + (x + 2t2 )v]exp(...)
207
2t3

∂
ux = v·exp xt +
∂x 3
dv ∂y
= exp(...) + tv·exp(...) (13.181)
dy ∂x
= exp(...)vy + tv·exp(...)
= (vy + tv)exp(...)
and
2t3

∂
uxx = (vy + tv)·exp xt +
∂x 3
∂
= (vy + tv)·exp(...) + t(vy + tv)·exp(...)
∂x (13.182)
d ∂y dv ∂y
= (vy )· +t ·exp(...) + t(vy + tv)·exp(...)
dy ∂x dy ∂x
= (vyy + tvy )·exp(...) + t(vy + tv)·exp(...)
= (vyy + 2tvy + t2 v)·exp(...)
Therefore the heat equation is being transformed to the equation
2tvy + (x + 2t2 )v = vyy + 2tvy + t2 v (13.183)
The above equation is equivalent to vyy = (x + t2 )v, i.e.
vyy = y·v (13.184)
which is the Airy equation (Abramowitz and Stegun §10.4) admitting the
general solution
v(y) = K1 ·Ai(y) + K2 ·Bi(y) (13.185)
where Ai(y) and Bi(y) are Airy’s first and second kind functions.
Therefore, going back to the original variables, the general solution of the
heat equation ut = uxx is given by
2 2
2t3
u(t, x) = {K1 ·Ai(x + t ) + K2 ·Bi(x + t )}·exp xt + (13.186)
3
Solution (e). We finally find the solutions of the heat equation considering
the generator
∂ ∂ ∂
Γ1 + Γ3 + aΓ6 = 4tx + (4t2 + 1) − (x2 + 2t − a)u (13.187)
∂x ∂t ∂u
208
The auxiliary system associated to the above generator is given by
dx dt du
= 2 = 2
(13.188)
4tx 4t + 1 −(x + 2t − a)u
dx dt
Let us now determine the invariants of the equation. Equation = 2
4tx 4t + 1
dx 1 8t dt
written as = gives after integration
x 2 4t2 + 1
1 √
ln(x) = ln(4t2 + 1) + A or x = C1 · 4t2 + 1 (13.189)
2
Therefore one invariant of the equation is given by
x
y=√ (13.190)
4t2 + 1
dt du
From equation = taking into account (13.189) we
4t2 +1 2
−(x + 2t − a)u
obtain
dt du
=− 2 2
4t2 +1 [y (4t + 1) + 2t − a]u

[y 2 (4t2 + 1) + 2t − a]dt du
2
=−
4t + 1 u
or

2 2t a du
y + 2 − 2 dt = −
4t + 1 4t + 1 u
Integrating the above equation we get
1 a
y2t + ln(4t2 + 1) − arctan(2t) = − ln(u) + B
4 2
and thus
12 2 a
u = C2 ·exp −y t − ln(4t + 1) + arctan(2t) (13.191)
4 2
209
or even
1 tx2 a
u = C2 · √
4
·exp − 2 + arctan(2t) (13.192)
4t2 + 1 4t + 1 2
Therefore the second invariant of the equation is given by

√
4
tx2 a
v = u· 4t2 + 1·exp − arctan(2t) (13.193)
4t2 + 1 2
Let us now express the partial derivatives ut and uxx in terms of the new
variables. Since
1 tx2 a
u = v· √
4
·exp − 2 + arctan(2t) (13.194)
4t2 + 1 4t + 1 2
we find
tx2

∂ 1 a
ut = v· √
4
·exp − 2 + arctan(2t)
∂t 4t2 + 1 4t + 1 2
1
= √ {−4ty·vy − 2t·v − (1 − 4t2 )y 2 ·v + a·v}·exp(...)
(4t + 1) 4 4t2 + 1
2
(13.195)
tx2

∂ 1 a
ux = v· √4
·exp − 2 + arctan(2t)
∂x 4t2 + 1 4t + 1 2
√ (13.196)
1 4

= √ · 4t 2 + 1·v − 2xt·v ·exp(...)
y
(4t2 + 1) 4 4t2 + 1
and
∂ √

1 4

uxx = √ 2
4t + 1·vy − 2xt·v ·exp(...)
(4t2 + 1) 4 4t2 + 1 ∂x
(13.197)
1 2 2
= √ [vyy − 4ty·vy − 2t·v + 4y t ·v]·exp(...)
(4t2 + 1) 4 4t2 + 1
Therefore the heat equation ut = uxx transforms into
− 4ty·vy − 2t·v − (1 − 4t2 )y 2 ·v + a·v

(13.198)
= vyy − 4ty·vy − 2t·v + 4y 2 t2 ·v
210
vyy + (−a + y 2 )·v = 0 (13.199)
The substitution y = λz, since

1 1
vy = vz and vyy = vzz (13.200)
λ λ2
converts (13.199) into
vzz = (aλ2 − λ4 z 2 )·v (13.201)

1
The choice λ = √ gives us the equation
2
1 1
vzz = a − z 2 ·v (13.202)
2 4
which admits general solution
a a
v(z) = A·W , z + B·W , −z (13.203)
2 2
where W is the parabolic cylindrical function of a complex variable
(Abramowitz and Stegun §19.17).
Therefore
a √ a √
v(y) = A·W , y 2 + B·W , −y 2 (13.204)
2 2
The heat equation ut = uxx admits a general solution given by
√ √
1 a 2x a 2x
u(t, x) = √4
A·W , √ + B·W , − √
4t2 + 1 2 4t2 + 1 2 4t2 + 1 (13.205)
tx2 a
×exp − 2 + arctan(2t)
4t + 1 2
package called MathLie to the heat equation. The MathLie is an AddOn
package to Mathematica (Reference [4]). We use the following steps:
In[1]:=<<MathLie.m
211
Step 2. We enter the heat equation using the command
In[2] := heat = ∂t U [x, t] − ∂{x,2} U [x, t] == 0; heat//LT F
Ut − Ux,x = 0
In[3] := pro2 = P rolongation[heat, {U }, {x, t}]; pro2//LT F
− Ux ξ1t − Ut Ux ξ1U − Ut ξ2t − Ut2 ξ2U + φ1t + Ut φ1U + 2Ux ξ2U Ux,t
+ 2ξ2x Ux,t + 3Ux ξ1U Ux,x + 2ξ1x Ux,x + Ut ξ2U Ux,x − φ1U Ux,x
+ Ux3 ξ1U,U + 2Ux2 ξ1x,U + Ux ξ1x,x + Ut Ux2 ξ2U,U + 2Ut Ux ξ2x,U
+ Ut ξ2x,x − Ux2 φ1U,U − 2Ux φ1x,U − φ1x,x == 0

substitute in the above expression Ut by Uxx using the command
In[4] := prolong2 = pro2/. ∂t U [x, t] → ∂xx U [x, t]; prolong2//LT F

− Ux ξ1t + φ1t + 2Ux ξ2U Ux,t + 2ξ2x Ux,t + 2Ux ξ1U Ux,x + 2ξ1x Uxx
− ξ2t Ux,x + Ux3 ξ1U,U + 2Ux2 ξ1x,U + Ux ξ1x,x + Ux2 Ux,x ξ2U,U
+ 2Ux Ux,x ξ2U,U + Ux,x ξ2x,x − Ux2 φ1U,U − 2Ux φ1x,U − φ1x,x == 0
Step 5. We collect together terms arranged with respect to Ux , Uxx and Uxt
using the command
In[5] := cprolong2 = Collect[prolong2, {∂t U [x, t], ∂xx U [x, t], ∂x,t U [x, t]}];
cprolong2//LT F
212
Out[5] :=
φ1t + 2ξ2x Ux,t + Ux3 ξ1U,U + Ux,x (2ξ1x − ξ2t + ξ2x,x )
+ Ux2 (2ξ1x,U + Ux,x ξ2U,U − φ1U,U )
+ Ux (−ξ1t + 2ξ2U Ux,t + ξ1x,x + Ux,x (2ξ1U + 2ξ2x,U ) − φ1x,U )
− φ1x,x
In[6] := detheat = DeterminingEquations[heat, {U }, {x, t}, {∂t U [x, t]}];

detheat//LT F
ξ1U == 0
ξ2U == 0
φ1U,U == 0
ξ2x == 0
− ξ1t + ξ1x,x − 2φ1x,U == 0
φ1t − φ1x,x == 0
2ξ1x − ξ2t == 0
In[6] := inf heat = Infinitesimals[heat, {U }, {x, t}]; inf heat//LT F
− F1t + F1x,x = 0
ξ1 = k5 − 2k2 t + k6 x + k4 tx
ξ2 = k3 + t(2k6 + k4 t)
k4 t k4 x2
φ1 = U k1 − + k2 x − + F1
2 4
213
to determine some solutions of the heat equation. We consider the following
infinitesimals
xi[1][x, t, U ] = x;
xi[2][x, t, U ] = 2 ∗ t;
phi[1][x, t, U ] = a ∗ U ;
In[7] := invar = xi[1]∗∂x F [x, t, U ]+xi[2]∗∂t F [x, t, U ]+φ[1]∗∂U F [x, t, U ] == 0
Out[7] := 2tFt + aU FU + xFx == 0
Step 8. Solving the above equation,
In[8] := solution = DSolve[invar, F [x, t, U ], {x, t, U }]
we find

t
Out[8] : F (x, t, U )→C[1] 2 , U x−a (13.206)
x
From the above result it is obvious that we have two invariants:
t
ζ= and F = U x−a
x2
Step 9. We now have to substitute t by ζx2 and U by F xa in order to find a
solution of the heat equation

2 a t
In[9] : rules = t→ζ ∗ x , U →F unction {x, t}, x ∗ F 2
x

2 a t
Out[9] := t→x ζ, U →F unction {x, t}, x F 2
x
Step 10. We reduce next the heat equation under the above substitutions:
In[10] : rheat = heat/.rules; rheat//LT F
214
The program responds with an equation, which when divided by x−2+a takes
on the form
−a(−1 + a)F [ζ] + (2aζ + 2(a + 1)ζ − 8ζ + 1)Fζ [ζ] − 4ζ 2 Fζζ [ζ] == 0
The above equation can be solved using the DSolve[] command. The result
is
1 − a
−1+a 1 1
2 C[1]Hypergeometric1F 1 − , , − a 3 1
F [ζ]→2
ζ 2 2 2 4ζ
−a
a 1 a 1
2 C[1]Hypergeometric1F 1 − , , − 1
+2
ζ 2 2 4ζ
t
We can go to the original variables by the substitution ζ = 2 .
x
However we can reach more easily to the above results, using a single com-
mand

In[12] : reduction = LieReduction {∂t U [x, t] − ∂x,x U [x, t]},

t
{x, t}, {x, 2 ∗ t}, {a ∗ U } ; LT F [F latten[reduction]]//. ζ→ 2
x
Out[12]//DisplayF orm
t
− zeta1 == 0
x2
− U x−a + F1 == 0

a 2
x (−1 + a)aF1 + (−1 + 6zeta1 − 4azeta1)F1zeta1 + 4zeta1 F1zeta1,zeta1
== 0
13.3. Example 4. We shall now analyse the Lie Symmetries of the equation
ut = a2 uxx + bux + cu (13.207)
where a, b, c are constants different than zero. We shall also convert this
equation to the Heat Equation examined in the previous Example.
∆ = ut − a2 uxx − bux − cu (13.208)
215
We determine the second extension P r(2) Γ. We have
P r(2) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x]
∂ ∂ ∂
(13.209)
+ U [tt] + U [tx] + U [xx]

ut − a2 uxx − bux − cu = 0
⇔ U [t] − a2 U [xx] − bU [x] − cU = 0
Using the explicit expressions for U [t] , U [x] and U [xx] given by (12.14), (12.15)
and (13.13) respectively, we write (13.209) in its equivalent form

2
− a Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux

− 2Tx utx + (Uu − 2Xx )uxx − 2Tu ux utx (13.210)


2
− b Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut − cU = 0
We now have to take into account the condition ∆ = 0, which means that
we have to substitute ut by a2 uxx + bux + cu in (13.209). We thus obtain the
equation
Ut + (a2 uxx + bux + cu)[Uu − (Tt + (a2 uxx + bux + cu)Tu )]
− ux (Xt + (a2 uxx + bux + cu)Xu )

− a Uxx − Txx (a2 uxx + bux + cu) + (2Uxu − Xxx )ux
2
− 2Txu (a2 uxx + bux + cu)ux + (Uuu − 2Xxu )(ux )2

− Tuu (a2 uxx + bux + cu)(ux )2 − Xuu (ux )3

2
− Tu (a uxx + bux + cu)uxx − 3Xu ux uxx
216

− b Ux + (Uu − Xx )ux − Xu (ux )2
(13.211)
2
− (Tx + Tu ux )(a uxx + bux + cu) − cU = 0
The above equation has to be arranged in powers of the partial derivatives

of the function u and then equate to zero all the coefficients of the partial
derivatives. However we can read-off easily some coefficients of the following
partial derivatives:
Coefficient of utx is 2a2 Tx and coefficient of ux utx is 2a2 Tu . Therefore Tx =
Tu = 0 and because of that, we get that the coefficient of ux uxx is 2a2 Xu and
then Xu = 0. We thus have
T = α(t) and X = X(t, x) (13.212)
Therefore equation (13.211) gets simplified into
Ut + (a2 uxx + bux + cu)(Uu − Tt ) − ux Xt

2
− a Uxx + (2Uxu − Xxx )ux + Uuu (ux )2
(13.213)

+ (Uu − 2Xx )uxx

− b Ux + (Uu − Xx )ux − cU = 0
We now equate to zero the coefficients of the derivatives of the function u.

−a2 Uuu = 0 (13.214)
a2 (2Xx − Tt ) = 0 (13.215)
Coefficient of ux :
b(−Tt + Xx ) − Xt − a2 (2Uxu − Xxx ) = 0 (13.216)
Coefficient of zero order derivative of u:
Ut + cu(Uu − Tt ) − a2 Uxx − bUx − cU = 0 (13.217)
217
Let us now try to solve the above system of equations (13.214)-(13.217). We
use the following steps:
Step (a). Equation (13.214) implies that Uuu = 0 and thus
U = β(t, x)u + γ(t, x) (13.218)
Step (b). Equation (13.215) implies that 2Xx = α0 (t) and by integration
1
X = α0 (t)x + δ(t) (13.219)
2
Step (c). Equation (13.216) takes on the form because of the previously
found expressions for U and X
b 0 1 00 1 0
βx (t, x) = − α (t) − α (t)x − δ (t) (13.220)
4a2 4a2 2a2
b 0 1 1
β(t, x) = − 2
α (t)x − 2 α00 (t)x2 − 2 δ 0 (t)x + θ(t) (13.221)
4a 8a 2a
Step (d). We finally obtain from equation (13.217) using all the previously
found expressions
(βt u+γt )+cu(β−α0 (t))−a2 (βxx u+γxx )−b(βx u+γx )−c(βu+γ) = 0 (13.222)
The above equation can be arranged in powers of u as
[βt − cα0 (t) − a2 βxx − bβx ]u + (γt − a2 γxx − bγx − cγ) = 0 (13.223)
From the above expression we get the following two equations
βt − cα0 (t) − a2 βxx − bβx = 0 (13.224)
and
γt − a2 γxx − bγx − cγ = 0 (13.225)
The last equation expresses the fact that γ(t, x) is a solution of the original
equation.
218
Using the expression for β(t, x) given in (13.221), equation (13.224) becomes

b 00 1 000 2 1 00 0
− 2 α (t)x − 2 α (t)x − 2 δ (t)x + θ (t)
4a 8a 2a
1
− cα0 (t) − a2 − 2 α00 (t) (13.226)
4a
b 1 1
− b − 2 α0 (t) − 2 α00 (t)x − 2 δ 0 (t) = 0
4a 4a 2a
The above equation can be arranged in terms of powers of x as

1 000 2 1 00
− 2
α (t)x − 2
δ (t)x
8a 2a
b2 (13.227)
0

0 1 00 b 0
+ θ (t) + − c α (t) + α (t) + 2 δ (t) = 0
4a2 4 2a
Equating to zero the coefficients of x we obtain the following three equations
α000 (t) = 0 (13.228)
δ 00 (t) = 0 (13.229)
b2 1 b
θ0 (t) + 2
− c α0 (t) + α00 (t) + 2 δ 0 (t) = 0 (13.230)
4a 4 2a
From (13.228) and (13.229) we obtain
α(t) = a1 + a2 t + a3 t2 ≡T (13.231)
and
δ(t) = a4 + a5 t (13.232)
respectively. We then have from (13.230) that

b2 1 b
θ0 (t) = − 2
− c (a2 + 2a3 t) − 2a3 − 2 a5 (13.233)
4a 4 2a
and by integration with respect to t we obtain
b2 1 b
θ(t) = − 2
− c (a2 t + a3 t2 ) − a3 t − 2 a5 t + a6 (13.234)
4a 2 2a
219
We are now able to find the final expression of β(t, x) from (13.221) using
the above expressions:
b 1 1
β(t, x) = − 2
(a2 + 2a3 t)x − 2 a3 x2 − 2 a5 x
4a 4a 2a (13.235)
b2 1 b
− 2
− c (a2 t + a3 t2 ) − a3 t − 2 a5 t + a6
4a 2 2a
From (13.219) we determine X
1
X = (a2 + 2a3 t)x + a4 + a5 t (13.236)
2
and from (13.218) we are able to determine U
b 1 1
U (t, x, u) = − 2
(a2 + 2a3 t)xu − 2 a3 x2 u − 2 a5 xu
4a 4a 2a
b2 1
− − c (a2 t + a3 t2 )u − a3 tu (13.237)
4a2 2
b
− 2 a5 tu + a6 u + γ(t, x)
2a
where γ(t, x) is an arbitrary solution of the original equation.
We thus see that the generator of the Lie Symmetries of the equation (13.207)
is given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
2 ∂ 1 ∂
= (a1 + a2 t + a3 t ) + (a2 + 2a3 t)x + a4 + a5 t
∂t 2 ∂x

b 1 1
+ − 2 (a2 + 2a3 t)xu − 2 a3 x2 u − 2 a5 xu
4a 4a 2a
b2 1 b

∂
2
− 2
− c (a2 t + a3 t )u − a3 tu − 2 a5 tu + a6 u + γ(t, x)
4a 2 2a ∂u
∂

∂ 1 ∂

b b2 ∂
Γ = a1 + a2 t + x − x+ −c t u
∂t ∂t 2 ∂x 4a2 4a2 ∂u
2

2 ∂ ∂ b 1 2 b 1 ∂

2
+ a3 t + tx − tx + 2 x + −c t + t u (13.238)
∂t ∂x 2a2 4a 4a2 2 ∂u

∂ ∂ 1 b ∂ ∂ ∂
+ a4 + a5 t − 2
x + 2t u + a6 u + γ(t, x)
∂x ∂x 2a 2a ∂u ∂u ∂u
220
We thus see that the Lie Algebra of the infinitesimal transformations of the
given equation is being spanned by the six vector fields
∂ ∂ 1 ∂

b b2 ∂
(1) (2)
Γ = , Γ =t + x − x+ −c t u
∂t ∂t 2 ∂x 4a2 4a2 ∂u
2

2 ∂ ∂ b 1 2 b 1 ∂ (13.239)

(3) 2
Γ =t + tx − 2
tx + 2 x + 2
−c t + t u
∂t ∂x 2a 4a 4a 2 ∂u
∂ ∂ 1 b ∂ ∂
Γ(4) = , Γ(5) = t − 2
x + 2 t u , Γ(6) = u
∂x ∂x 2a 2a ∂u ∂u
and the infinite dimensional sub-algebra
∂
Γ(∞) = γ(t, x) (13.240)
∂u
13.3a. Conversion of the equation into the Heat equation.
We shall now convert the given equation into the Heat equation by consider-
ing a linear combination of the generators Γ(1) , Γ(4) and Γ(6) . In other words
we consider the generator
1 ∂ 1 ∂ ∂
X= + +u (13.241)
λ ∂t µ ∂x ∂u
In order to determine the invariants corresponding to the above generator,
we consider the system associated to the above generator
dt dx du
λ =µ = (13.242)
1 1 u
dt dx
From the equation λ = µ we obtain by integration λt + C1 = µx or
1 1
x = νt + C̃1 and thus one invariant of the equation is
y = x − νt (13.243)
dt du
From the equation λ = we obtain by integration λt = ln(C2 t) or
1 u
eλt = C2 u and thus another invariant is given by
v = u·e−λt (13.244)
and then
u = v·eλt , v = v(y) (13.245)
221
Using (13.243) and (13.245) we find that
ut = (vt − νvy + λv)eλt

ux = vy eλt (13.246)
uxx = vyy eλt
Therefore equation ut = a2 uxx + bux + cu is being converted to the equation

(after canceling the factor eλt )
vt − νvy + λv = a2 vyy + bvy + cv (13.247)
vt = a2 vyy + (ν + b)vy + (c − λ)v (13.248)
We determine next the λ and ν so as the coefficients of vy and v to become

zero:
ν + b = 0, c−λ=0 (13.249)
i.e.
ν = −b and λ = c (13.250)
Therefore the two invariants become
y = x + bt and v = e−ct ·u (13.251)
We thus see that the substitution
v = e−ct ·u(t, x = y − bt) (13.252)
converts equation ut = a2 uxx + bux + cu into the heat equation vt = a2 vyy .

13.4. Example 5. A Nonlinear Wave Equation. We consider the
nonlinear wave equation
utt = uuxx , u = u(x, t) (13.253)
Invariance is being expressed by
P r(2) Γ(∆)|∆=0 = 0 (13.254)
222
where
∆≡utt − uuxx (13.255)
For the second prolongation we consider the operator

∂ ∂ ∂
P r(2) Γ = U + U [tt] + U [xx] (13.256)
∂u ∂utt ∂uxx
where we have omitted some terms which, when applied to ∆, give zero.
We thus have
P r(2) Γ(∆) = 0
∂
[tt] ∂ [xx] ∂

⇐⇒ U +U +U (utt − uuxx ) = 0 (13.257)
∂u ∂utt ∂uxx
⇐⇒ − U uxx + U [tt] − u U [xx] = 0
Since U [tt] and U [xx] are given by
U [tt] = Utt − Xtt ux + (2Utu − Ttt )ut − 2Xtu ux ut

(13.258)
and
(13.259)
223
respectively, equation (13.257) takes on the form
− U uxx
+ Utt − Xtt ux + (2Utu − Ttt )ut − 2Xtu ux ut
(13.260)

− u Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux



utt by uuxx . We then obtain from (13.260) the equation
− U uxx
− 2Xt utx + u(Uu − 2Tt )uxx − 2Xu ut utx
− uXu ux uxx − 3uTu ut uxx
(13.261)

− u Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux


We now put to zero the coefficients of all the partial derivatives of the function
u. However we see that the coefficient of ut utx is −2Xu , the coefficient of
ut uxx is −2uTu , the coefficient of ux utx is 2uTu , and the coefficient of ux uxx
is 2uXu . We thus see that
Xu = 0 and Tu = 0 (13.262)
224
i.e. the functions X and T are independent of u. Because of that, equation
(13.261) gets simplified considerably:
− U uxx
+ Utt − Xtt ux + (2Utu − Ttt )ut + Uuu (ut )2
− 2Xt utx + u(Uu − 2Tt )uxx
(13.263)

− u Uxx − Txx ut + (2Uxu − Xxx )ux + Uuu (ux )2

− 2Tx utx + (Uu − 2Xx )uxx = 0
From the above equation we see that the coefficients of (ut )2 and (ux )2 are
Uuu and −uUuu respectively and thus
Uuu = 0 (13.264)
We have further that

The coefficient of ut is
2Utu − Ttt + uTxx = 0 (13.265)
−Xtt − u(2Uxu − Xxx ) = 0 (13.266)
The coefficient of utx is
−2Xt + 2uTx = 0 (13.267)
2u(Xx − Tt ) − U = 0 (13.268)
Finally the constant term is equal to
Utt − uUxx = 0 (13.269)
We now have to solve the system of equations (13.262) and (13.264)-(13.269).

From equation (13.267) we have uTx = Xt and since both T and X are
independent of u, we should have
Tx = 0 and Xt = 0 (13.270)
225
i.e.
T = T (t) and X = X(x) (13.271)
U = α(x, t)u + β(x, t) (13.272)
Therefore we obtain from (13.265) and (13.266) that
T 00 (t) = 2αt (x, t) (13.273)
and
X 00 (x) = 2αx (x, t) (13.274)
Equation (13.269) takes on the form
−αxx u2 + (αtt − βxx )u + βtt = 0 (13.275)
From the above equation we get
αxx = 0, αtt − βxx = 0, βtt = 0 (13.276)
Since αxx = 0 and βtt = 0 we should have
α(x, t) = a1 (t)x + a2 (t) (13.277)
and
β(x, t) = b1 (x)t + b2 (x) (13.278)
respectively.
Combining (13.273) with (13.277) we obtain
T 00 (t) = 2{a01 (t)x + a02 (t)} (13.279)
and since the left hand side does not depend on x, we should have a01 (t) = 0,
i.e. a1 (t) is a constant:
a1 (t) = C1 (13.280)
226
Therefore
α = C1 x + a2 (t) , X 00 (x) = 2C1 and T 00 (t) = 2a02 (t) (13.281)
and thus
X(x) = C1 x2 + C2 x + C3 (13.282)
Equation αtt − βxx = 0 becomes a002 (t) − b001 (x)t − b002 (x) = 0 and then
a002 (t) = 0, b001 (x) = 0, b002 (x) = 0 (13.283)
Therefore a2 (t) = C4 t + C5 , b1 (x) = b11 x + b12 , b2 (x) = b21 x + b22 . Since

T 00 (t) = 2C4 , we get
T (t) = C4 t2 + C5 t + C6 (13.284)
We also have
α = C1 x + C4 t + C5 (13.285)
and
β = (b11 x + b12 )t + b21 x + b22 (13.286)
Since, as follows from (13.268),
U = 2u(Xx − Tt ) (13.287)
we get, using (13.282) and (13.284),
U = 2u(2C1 x + C2 − 2C4 t − C5 ) (13.288)
On the other hand, U is given by (13.272) where α(x, t) and β(x, t) are given
by (13.285) and (13.286) respectively:
U = (C1 x + C4 t + C5 )u + (b11 x + b12 )t + b21 x + b22 (13.289)
The two expressions for U , i.e. (13.288) and (13.289), should be identical.
Therefore, upon comparing (13.288) and (13.289) we obtain
4C1 = C1 , −4C4 = C4 , 2(C2 − C5 ) = C5 (13.290)
227
i.e.
C1 = 0, C4 = 0, 2C2 = 3C5 (13.291)
and
b11 = b12 = b21 = b22 = 0 (13.292)
Because of the above, we obtain from (13.282) and (13.284)
X(x) = C2 x + C3 , T (t) = C5 t + C6 (13.293)
and from (13.287),
U = 2(C2 − C5 )u (13.294)
Introducing the assignments
C 2 → a1 , C3 → a2 , C5 → a3 , C6 → a4 (13.295)
we get the following expressions
X(x) = a1 x + a2 , T (t) = a3 t + a4 , U = 2(a1 − a3 )u (13.296)
The generator Γ of symmetries for the equation utt = uuxx is the given by
∂ ∂ ∂
Γ=X +T +U
∂x ∂t ∂u (13.297)
∂ ∂ ∂
= (a1 x + a2 ) + (a3 t + a4 ) + 2((a1 − a3 )u
∂x ∂t ∂u
or
∂ ∂ ∂ ∂ ∂ ∂
Γ = a1 x + 2u + a2 + a3 t − 2u + a4 (13.298)
∂x ∂u ∂x ∂t ∂u ∂t
Therefore the Lie algebra of symmetries for the equation utt = uuxx is
spanned by the four generators
∂ ∂ ∂
X1 = x + 2u , X 2 =
∂x ∂u ∂x (13.299)
∂ ∂ ∂
X 3 = t − 2u , X 4 =
∂t ∂u ∂t
228
14 Boundary Value Problems for
Partial Differential Equations
We shall consider a boundary value problem for an equation which occurs
in the Mechanics of Fluids. It is a differential equation that models mass
transfer from a horizontal flat plate. This equation has the form
√
x uyy = α y ux (14.1)
subject to conditions
√
x
u(x, y) = u0 when y = (14.2)
β
and
u(x, 0) = u1 (14.3)
Considering the symmetry operator

∂ ∂ ∂
Γ=X +Y +U (14.4)
∂x ∂y ∂u
the first and second extensions are given by
∂ ∂
P r(1) Γ = Γ + U [x] + U [y] (14.5)
∂ux ∂uy
and
∂ ∂ ∂
P r(2) Γ = P r(1) Γ + U [xx] + U [xy] + U [yy] (14.6)
∂uxx ∂uxy ∂uyy
respectively.
Invariance of equation is being expressed by the condition
P r(2) Γ(∆)|∆=0 = 0 (14.7)
where
√
∆≡ xuyy − αyux (14.8)
229
We have further
P r(2) Γ(∆) = 0
1 √ (14.9)
⇔ √ Xuyy − αY ux − αyU [x] + xU [yy] = 0
2 x
The coefficients U [x] and U [yy] are easily evaluated and they are given
explicitly by
U [x] = Ux + (Uu − Xx )ux − Yx uy − Yu ux uy − Xu (ux )2 (14.10)
and
U [yy] = Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3
+ (Uu − 2Yy )uyy − (3uy Yu + ux Xu )uyy − Xyy ux (14.11)
− 2Xyu ux uy − Xuu ux (uy )2 − 2Xu uy uxy − 2Xy uxy
respectively.
Therefore equation (14.9) can be written as
1
√ Xuyy − αY ux
2 x

2
− αy Ux + (Uu − Xx )ux − Yx uy − Yu ux uy − Xu (ux )
√

(14.12)
+ x Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3
+ (Uu − 2Yy )uyy − (3uy Yu + ux Xu )uyy − Xyy ux

2
− 2Xyu ux uy − Xuu ux (uy ) − 2Xu uy uxy − 2Xy uxy = 0
230
We now have to take into√ account the condition ∆ = 0. For this purpose we
substitute uyy by αyux / x in (14.12):
1 αyu
√ X √ x − αY ux
2 x x

2
− αy Ux + (Uu − Xx )ux − Yx uy − Yu ux uy − Xu (ux )
√

+ x Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3 (14.13)
αyux αyux
+ (Uu − 2Yy ) √ − (3uy Yu + ux Xu ) √ − Xyy ux
x x

2
− 2Xyu ux uy − Xuu ux (uy ) − 2Xu uy uxy − 2Xy uxy = 0

the function u(x, y). We obtain that the coefficients of uxy and uy uxy are
given by −2Xy and −2Xu respectively. We thus have Xy = 0 and Xu = 0.
Therefore
X = X(x) (14.14)
Because of that, equation (14.13) becomes

αy
X − αY ux
2x
− αy Ux + (Uu − Xx )ux − Yx uy − Yu ux uy
√
(14.15)
+ x Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2

3 αyux αyux
− Yuu (uy ) + (Uu − 2Yy ) √ − (3uy Yu ) √ =0
x x
The coefficient of ux uy is −2αyYu and thus Yu = 0. Therefore
Y = Y (x, y) (14.16)
231
Equation (14.15) becomes
αy
X − αY ux
2x
− αy Ux + (Uu − Xx )ux − Yx uy
√
(14.17)
+ x Uyy + (2Uyu − Yyy )uy + Uuu (uy )2

αy
+ √ (Uu − 2Yy )ux = 0
x
We finally obtain: √
Coefficient of (uy )2 is x Uuu and thus
Uuu = 0 (14.18)
Coefficient of uy
√
αyYx + x(2Uyu − Yyy ) = 0 (14.19)
Coefficient of ux
αy
αy(Xx − 2Yy ) + X − αY = 0 (14.20)
2x
Zeroth order derivative
√
−αyUx + xUyy = 0 (14.21)
We next solve the system of differential equations (14.18)-(14.21) taking into

account (14.14) and (14.16).
U = A(x, y)u + B(x, y) (14.22)

√
−αy(Ax u + Bx ) + x(Ayy u + Byy ) = 0 (14.23)

√ √
(−αyAx + xAyy )u + (−αyBx + xByy ) = 0 (14.24)
232
from which we obtain the following two equations
√
−αyAx + xAyy = 0 (14.25)
and
√
−αyBx + xByy = 0 (14.26)
Equation (14.20) is equivalent to

1 1
Xx + X = 2Yy + Y (14.27)
2x y
and since X = X(x), each member of the above equation should be equal to
a function say f (x):
1
Xx + X = f (x) (14.28)
2x
1
2Yy + Y = f (x) (14.29)
y
Upon integration of (14.28) and (14.29) we find
√
Z
1
X(x) = √ xf (x)dx (14.30)
x
and
1 g(x)
Y (x, y) = yf (x) + √ (14.31)
3 y
respectively.
Equation (14.19), because of (14.22) and (14.31) takes on the form
√ 3√

1 0 1 0
αy yf (x) + √ g (x) + 2 xAy − xg(x)y −5/2 = 0 (14.32)
3 y 4
and upon integration with respect to y we obtain that
1 g 0 (x) 1 1 f 0 (x)
A(x, y) = − α √ y 3/2 − g(x)y −3/2 − α √ y 3 + h(x) (14.33)
3 x 4 18 x
233
Equation (14.25), because of the above expression for A(x, y), takes on the
form

2 00 0 15/2 00 0
− 4α 2xf (x) − f (x) y − 24α 2xg (x) − g (x) y 6
2
(14.34)
3/2 0 0 9/2 2
+ 48αx f (x) + 3h (x) y + 135x g(x) = 0
The above equation should be true for every x and y. Therefore equating to
zero the coefficients of y we obtain
2xf 00 (x)−f 0 (x) = 0, 2xg 00 (x)−g 0 (x) = 0, f 0 (x)+3h0 (x) = 0, g(x) = 0
The above system admits solutions

1
f (x) = a1 + a2 x3/2 , g(x) = 0, h(x) = − a2 x3/2 + a3 (14.35)
3
We then obtain from (14.30) and (14.31) that
2 1
X(x) = a1 x + a2 x5/2 + a4 x−1/2 (14.36)
3 3
and
1
Y (x, y) = y(a1 + a2 x3/2 ) (14.37)
3
respectively.
We finally obtain from (14.33) that
1 1
A(x, y) = − a2 αy 3 − a2 x3/2 + a3 (14.38)
12 3
The function B(x, y) remains undetermined and satisfies the (original) equa-
tion (14.26). Therefore
2 1
X(x) = a1 x + a2 x5/2 + a4 x−1/2
3 3
1
Y (x, y) = y(a1 + a2 x3/2 ) (14.39)
3
1 1
U = − a2 αy 3 − a2 x3/2 + a3 u + B(x, y)
12 3
234
We thus obtain the following expression for the generator Γ of the Lie Sym-
metries
∂ ∂ ∂
Γ=X +Y +U
∂x ∂y ∂u
2 1 ∂ 1 ∂
= a1 x + a2 x5/2 + a4 x−1/2 + y(a1 + a2 x3/2 ) (14.40)
3 3 ∂x 3 ∂y

1 1 ∂
+ − a2 αy 3 − a2 x3/2 + a3 u + B(x, y)
12 3 ∂u
1 ∂ ∂
Γ = a1 2x +y
3 ∂x ∂y

1 5/2 ∂ 3/2 ∂ 1 3 3/2 ∂
+ a2 x + yx + − αy − x u (14.41)
3 ∂x ∂y 4 ∂u
∂ ∂ ∂
+ a3 u + a4 x−1/2 + B(x, y)
∂u ∂x ∂u
From the above expression we obtain the following four generators of the Lie
Algebra symmetries
∂ ∂
Γ(1) = 2x +y
∂x ∂y

∂ 3/2 ∂ 1 3 ∂
Γ(2) =x 5/2
+ yx + − αy − x 3/2
u
∂x ∂y 4 ∂u (14.42)
(3) ∂
Γ =u
∂u
∂
Γ(4) = x−1/2
∂x
and also the infinite subalgebra spanned by the generator
∂
Γ(∞) = B(x, y) (14.43)
∂u
The function B(x, y) is any solution of the original equation.
In order to face the BVP (14.1)-(14.3), we introduce an operator X, which
is a linear combination of the generators Γ(1) , Γ(2) , Γ(3) , and Γ(4) , by
X = Γ(1) + λΓ(2) + µΓ(3) + νΓ(4) (14.44)
235
where λ, µ, ν are parameters to be determined. We require invariance of the
BVP conditions by the operator X
√
X u(x, y) − u0 = 0, y = x/β = 0, X u(x, 0) − u1 = 0 = 0 (14.45)
The above two conditions lead us to the equation

α
µ − λ 1 + 3 x3/2 = 0 (14.46)
4β
µ = 0 and λ = 0 (14.47)
Therefore the solution of the BVP leads us to consider the invariance under
the operator
(1) (4) ∂ ∂ −1/2 ∂

ν ∂ ∂
Γ + νΓ ≡2x +y + νx = 2x + √ +y (14.48)
∂x ∂y ∂x x ∂x ∂y
The Invariant Surface Condition (Lagrange’s system) corresponding to the
above operator reads
dx dy du
ν = y = 0 (14.49)
2x + √
x
Integrating the first equation, we obtain
1
ln(2x3/2 + ν) = ln(y) + K ⇒ ln(2x3/2 + ν) = ln(y 3 ) + K1
3
and thus an independent solution of Lagrange’s auxiliary system is
2x3/2 + ν = C1 ·y 3 (14.50)
Another independent solution of the same system is given by
u = C2 (14.51)
Therefore the general solution of Lagrange’s system is given by

y3
Φ(C1 , C2 ) = 0 or Φ ,u = 0 (14.52)
2x3/2 + ν
236
From the above equation we derive
y3
u=F (14.53)
2x3/2 + ν
Introducing the similarity variable
y3
ξ= (14.54)
2x3/2 + ν
equation (14.1) under the substitution u = F (ξ) takes on the form
3ξF 00 (ξ) + (2 + αξ)F 0 (ξ) = 0 (14.55)
f (ξ) = F 0 (ξ) (14.56)
3ξf 0 (ξ) + (2 + αξ)f (ξ) = 0 (14.57)

1
−3/2
f (ξ) = K1 ξ exp − αξ (14.58)
3
Therefore
Z 1
F (ξ) = K2 + K1 ξ −3/2 exp − αξ dξ (14.59)
3
and after performing the integration we obtain
F (ξ) = K2 + K1 G(ξ) (14.60)
where
1 3 3 1/6 1 2 αξ 1
1/6 1/6
G(ξ) = 3ξ ξ exp − αξ + M , ; exp − αξ (14.61)
6 4 α 6 3 3 6
The function M is the Whittaker M −function which admits the following
Taylor expansion
1 2 αξ αξ 7/6 1 11 41

2 3
M , ; = 1 − (αξ) + (αξ) − (αξ) + · · ·
6 3 3 3 42 2520 589680
237
(14.62)
For y = 0 we have ξ = 0 and G(0) = 0 and thus u(x, 0) = u1 is equivalent to
K1 = u1 (14.63)
√
For y = x/β we obtain
x3/2
ξ= (14.64)
β 3 (2x3/2 + ν)
and G(ξ) is a constant for the above value of ξ, only if ν = 0. In this case
1
ξ0 = (14.65)
2β 3
√
Therefore u(x, y) = u0 for y = x/β is equivalent to
K1 + K2 G(ξ0 ) = u0 (14.66)
Equations (14.63) and (14.66) can be solved with respect to K1 and K2

yielding
u0 − u1
K1 = u1 and K2 = (14.67)
G(ξ0 )
Therefore the solution to the BVP (14.1)-(14.3) is given by

u0 − u1
u(x, y) = u1 + G(ξ) (14.68)
G(ξ0 )
Chapter IV
Higher Order Partial
238
15 Preliminaries
In this Section we are going to consider some higher order partial differential
equations like the Korteweg-DeVries (KdV) Equation, the Boussinesq Equa-
tion and the BBM Equation.
The KdV and Boussinesq equations are given by
ut + uux + uxxx = 0, utt + uuxx + (ux )2 + uxxxx = 0 (15.1)
respectively, while the BBM equation (after Bona, Burgers and Mahony) is
given by
ut + ux + uux − uxxt = 0 (15.2)
We shall need the formulas
[tt] [t]
U = Dt (U ) − utt Dt (T ) − utx Dt (X) (15.5)
U [xx] = Dx (U [x] ) − utx Dx (T ) − uxx Dx (X) (15.6)
U [xxx] = Dx (U [xx] ) − utxx Dx (T ) − uxxx Dx (X) (15.7)
[txx] [xx]
U = Dt (U ) − uttx Dt (T ) − utxx Dt (X) (15.8)
[xxxx] [xxx]
U = Dx (U ) − utxxx Dx (T ) − uxxxx Dx (X) (15.9)
in order to find the explicit expressions for U [tt] , U [xxx] , U [txx] and U [xxxx] .
The operators Dt and Dx are given by
∂ ∂ ∂ ∂
Dt = + ut + utt + utx + ··· (15.10)
and
∂ ∂ ∂ ∂
Dx = + ux + uxt + uxx + ··· (15.11)
respectively. The expressions U [t] and U [xx] are given by (12.14) and (13.13)
respectively.
We find further, using the above formulas (15.3)-(15.9)
U [tt] = Utt + (Uu − 2Tt )utt + (2Utu − Ttt )ut − Xu ux utt
− 3Tu ut utt + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3 − Xuu ux (ut )2 (15.12)
− Xtt ux − 2Xt uxt − 2Xu ut uxt − 2Xtu ut ux
239
U [xxx] = Uxxx − Xuuu (ux )4 + (Uuuu − 3Xxuu )(ux )3
+ 3(Uxuu − Xxxu )(ux )2 + (3Uxxu − Xxxx )ux
+ 3(Uxu − Xxx )uxx + 3(Uuu − 3Xxu )ux uxx
+ (Uu − 3Xx )uxxx − (4Xu ux + Tu ut )uxxx
(15.13)
− Txxx ut − 3Txxu ut ux − 3Txuu ut (ux )2 − Tuuu ut (ux )3
− 3Tu ux uxxt − 3Tx uxxt − 3Txx uxt − 6Txu ux uxt
− 3Txu ut uxx − 3Tuu ut ux uxx − 3Tuu (ux )2 uxt
− 3Tu uxt uxx − 6Xuu (ux )2 uxx − 3Xu (uxx )2
U [xxt] = Uxxt + (Uuuu − 2Xxuu − Ttuu )ut (ux )2
+ (Utu − 2Xxt )uxx + (2Uxu − 2Txt − Xxx )uxt
+ (Uxxu − Txxt )ut + (2Uxtu − Xxxt )ux
+ (Uu − 2Xx − Tt )uxxt − (3Xu ux + 2Tu ut )uxxt
+ (2Uuu − 4Xxu − 2Ttu )ux uxt − Xuuu ut (ux )3
+ (2Uxuu − Xxxu − 2Txtu )ut ux + (Utuu − 2Xxtu )(ux )2
(15.14)
+ (Uuu − 2Xxu − Ttu )ut uxx − 2Tu (uxt )2 − Txx utt
− 3Xtu ux uxx − 3Xu uxt uxx − Tu utt uxx − Txxu (ut )2
− 2Tx uxtt − 2Tu ux uxtt − Xtuu (ux )3 − Xt uxxx
− Xu ut uxxx − 2Txuu ux (ut )2 − Tuuu (ux )2 (ut )2
− 2Txu ux utt − 4Txu ut uxt − 3Xuu (ux )2 uxt − 3Xuu ut ux uxx
− Tuu (ux )2 utt − 4Tuu ux ut uxt − Tuu (ut )2 uxx
240
and
U [xxxx] = Uxxxx + (4Uxxxu − Xxxxx )ux + (6Uxxu − 4Xxxx )uxx

+ (4Uxu − 6Xxx )uxxx + 4(Uuu − 4Xxu )ux uxxx
+ (Uu − 4Xx )uxxxx − (Tu ut + 5Xu ux )uxxxx
− Xuuuu (ux )5 + (Uuuuu − 4Xxuuu )(ux )4 + (4Uxuuu − 6Xxxuu )(ux )3
+ (6Uxxuu − 4Xxxxu )(ux )2 − Txxxx ut − 4Txxxu ut ux − 6Txxuu ut (ux )2
− 4Txuuu ut (ux )3 − Tuuuu ut (ux )4 + (3Uuu − 12Xxu )(uxx )2
+ (6Uuuu − 24Xxuu )(ux )2 uxx − 6Txx uxxt − 12Txu ux uxxt
− 6Tuu (ux )2 uxxt − 4Tx uxxxt − 4Tu ux uxxxt − 4Txxx uxt
+ (12Uxuu − 18Xxxu )ux uxx − 12Txu uxt uxx
− 15Xuu ux (uxx )2 − 12Tuu ux uxt uxx − 3Tuu ut (uxx )2
− 6Tu uxx uxxt − 4Tu uxt uxxx − 10Xu uxx uxxx − 10Xuu (ux )2 uxxx
− 4Tuu ut ux uxxx − 10Xuuu (ux )3 uxx − 4Txu ut uxxx
− 4Tuuu (ux )3 uxt − 6Tuuu ut (ux )2 uxx − 12Txuu (ux )2 uxt
− 12Txuu ut ux uxx − 12Txxu ux uxt − 6Txxu ut uxx
(15.15)
16 The KdV Equation

We consider first the KdV equation
ut + uux + uxxx = 0 (16.1)
Considering the third prolongation,

∂ ∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +T +U + U [t] + U [x] + U [xxx] (16.2)
∂x ∂t ∂u ∂ut ∂ux ∂uxxx
and denoting by ∆ the quantity
∆ = ut + uux + uxxx (16.3)
241
we have
P r(3) Γ(∆) = 0
∂ ∂ ∂ ∂ ∂ ∂
⇔ X +T +U + U [t] + U [x] + U [xxx]
∂x ∂t ∂u ∂ut ∂ux ∂uxxx (16.4)
(ut + uux + uxxx ) = 0
⇔ U ·ux + U [t] + U [x] u + U [xxx] = 0
The infinitesimals U [t] , U [x] and U [xxx] are given by (12.14), (12.15) and
(15.13) respectively. Upon substituting the infinitesimals in the above equa-
tion, we derive the equation
U ·ux + Ut + (Uu − Tt )ut − Xt ux − Tu (ut )2 − Xu ut ux
+ [Ux + (Uu − Xx )ux − Tx ut − Xu (ux )2 − Tu ux ut ]u
+ Uxxx − Xuuu (ux )4 + (Uuuu − 3Xxuu )(ux )3
(16.5)
+ (Uu − 3Xx )uxxx − (4Xu ux + Tu ut )uxxx
− 3Tu uxt uxx − 6Xuu (ux )2 uxx − 3Xu (uxx )2 = 0
We now have to take into account the condition ∆ = 0. We substitute uxxx
by −ut − uux in the previous equation. We then obtain the equation
+ [Ux + (Uu − Xx )ux − Tx ut − Xu (ux )2 − Tu ux ut ]u
+ Uxxx − Xuuu (ux )4 + (Uuuu − 3Xxuu )(ux )3
(16.6)
− (Uu − 3Xx )(ut + uux ) + (4Xu ux + Tu ut )(ut + uux )
− 3Tu uxt uxx − 6Xuu (ux )2 uxx − 3Xu (uxx )2 = 0
242
We equate to zero all the coefficients of the partial derivatives of the unknown
function u.
We first see that the coefficient of (uxx )2 is −3Xu , the coefficient of ux uxxt is
−3Tu and the coefficient of uxxt is −3Tx . We thus have the three equations
Xu = 0, Tu = 0, Tx = 0 (16.7)
U ·ux + Ut + (Uu − Tt )ut − Xt ux

+ [Ux + (Uu − Xx )ux ]u
+ Uxxx + Uuuu (ux )3
(16.8)
+ 3Uxuu (ux )2 + (3Uxxu − Xxxx )ux
+ 3(Uxu − Xxx )uxx + 3Uuu ux uxx
− (Uu − 3Xx )(ut + uux ) = 0
The coefficient of ut is equal to −Tt + 3Xx , the coefficient of ux uxx is equal

to 3Uuu , the coefficient of (ux )2 is equal to 3Uxuu and the coefficient of (ux )3
is equal to Uuuu . We thus have the following equations
−Tt + 3Xx = 0 (16.9)
and
Uuu = 0 (16.10)
Equations Uxuu = 0 and Uuuu = 0 are implied by Uuu = 0. Therefore equation

U ·ux + Ut − Xt ux + [Ux + (Uu − Xx )ux ]u

+ Uxxx + (3Uxxu − Xxxx )ux + 3(Uxu − Xxx )uxx (16.11)
− (Uu − 3Xx )uux = 0
We finally equate to zero the coefficients of uxx , ux and the constant term
and obtain the equations
Uxu − Xxx = 0 (16.12)
U − Xt + 2uXx + 3Uxxu − Xxxx = 0 (16.13)
243
and
Ut + uUx + Uxxx = 0 (16.14)
respectively.
We are going to solve the system of equations (16.7), (16.9)-(16.10) and
(16.12)-(16.14).
From equations (16.7) and (16.10) we obtain the equations
T = T (t), X = X(t, x) (16.15)
and
U = a(t, x)u + b(t, x) (16.16)
ax u2 + (at + bx + axxx )u + bt + bxxx = 0 (16.17)
from which we derive the system
ax = 0, at + bx + axxx , bt + bxxx = 0
The above system is equivalent to
a = a(t), a0 (t) + bx = 0, bt + bxxx = 0 (16.18)
Equation (16.12), using (16.16) takes on the form
ax − Xxx = 0
and since ax = 0, we get Xxx = 0 and thus
X = c(t)x + d(t) (16.19)
From (16.13) we obtain, taking into account (16.16), ax = 0, and Xxx = 0,

that
[a(t) + 2c(t)]u + b − c0 (t)x − d0 (t) = 0
from which we get
a(t) + 2c(t) = 0, b − c0 (t)x − d0 (t) = 0 (16.20)
244
From equation a0 (t)+bx = 0 we have bx = −a0 (t) and integrating with respect
to x, we obtain
b = −a0 (t)x + e(t) (16.21)
Because of the above expression for b, equation bt + bxxx = 0 becomes

−a00 (t)x + e0 (t) = 0 and thus a00 (t) = 0 and e0 (t) = 0. Therefore
a(t) = a1 t + a2 and e(t) = a3 (16.22)
and also
b = −a1 x + a3 (16.23)
From equation a(t) + 2c(t) = 0, because of a(t) = a1 t + a2 , we have

1 1
c(t) = − a1 t − a2 (16.24)
2 2
From equation b − c0 (t)x − d0 (t) = 0, because of (16.23) and (16.249) we
obtain
1
− a1 x + [a3 − d0 (t)] = 0
2
and thus a1 = 0 and d0 (t) = a3 , i.e.
a1 = 0, d(t) = a3 t + a4 (16.25)
Therefore
1
a(t) = a2 , b = a3 , c(t) = − a2 (16.26)
2
From equation (16.9) we obtain T 0 (t) = 3c(t) = −(3/2)a2 and by integration
3
T (t) = − a2 t + a5 (16.27)
2
Since X(t) = c(t)x + d(t) and U = au + b, we have the following expressions
1
X = − a2 x + a3 t + a4 , U = a2 u + a3 (16.28)
2
245
Taking into account the above expressions for the functions X, T and U , we
have the following expression for the operator Γ:
∂ ∂ ∂
Γ=X +T +U
∂x ∂t ∂u
1 ∂ 3 ∂ ∂
= − a2 x + a3 t + a4 + − a2 t + a5 + (a2 u + a3 )
2 ∂x 2 ∂t ∂u (16.29)
1 ∂ ∂ ∂ ∂ ∂
= − a2 x + 3t − 2u + a3 t +
2 ∂x ∂t ∂u ∂x ∂u
∂ ∂
+ a4 + a5
∂x ∂t
Making the changes −(1/2)a2 →c1 , a3 →c2 , a5 →c3 and a4 →c4 , we write the
operator Γ as
∂ ∂ ∂
Γ=X +T +U
∂x ∂t ∂u (16.30)
∂ ∂ ∂ ∂ ∂ ∂ ∂
= c1 x + 3t − 2u + c2 t + + c3 + c4
∂x ∂t ∂u ∂x ∂u ∂t ∂x
Therefore the Lie algebra of symmetries for the KdV equation is being
spanned by the following four generators
∂ ∂ ∂ ∂ ∂
Γ1 = x + 3t − 2u , Γ2 = t +
∂x ∂t ∂u ∂x ∂u (16.31)
∂ ∂
Γ3 = , Γ4 =
∂t ∂x
(a) We consider the generator Γ3 + cΓ4 , i.e.
∂ ∂
Γ3 + cΓ4 = +c (16.32)
∂t ∂x
In this case the Invariant Surface Condition reads
ut + cux = 0 (16.33)
u = f (x − ct) (16.34)
which means that the combination Γ3 + cΓ4 corresponds to traveling wave
solutions. Introducing the similarity variable ξ by
ξ = x − ct (16.35)
246
the original KdV equation ut + uux + uxxx = 0 becomes
f 000 (ξ) + f (ξ)f 0 (ξ) − c·f 0 (ξ) = 0 (16.36)
Integrating once the above equation we obtain

1
f 00 (ξ) + f 2 (ξ) − c·f (ξ) = 0 (16.37)
2
where we have set to zero the integration constant. For c = 4k 2 , one special
solution of this equation is given by
12k 2
f (ξ) = (16.38)
cosh2 (kξ)
and thus one solution of the KdV equation is given by
12k 2
u(x, t) = = 12k 2 ×sech2 (k(x − 4k 2 t)) (16.39)
cosh2 (k(x − 4k 2 t))
which is the one-soliton solution of the KdV equation.

(b) We next consider the generator Γ1 , i.e.
∂ ∂ ∂
Γ1 = x + 3t − 2u (16.40)
∂x ∂t ∂u
In this case the Invariant Surface Condition reads
3tut + xux = −2u (16.41)

x
−2/3
u=t f 1/3 (16.42)
t
Introducing the similarity variable η by
x
η= (16.43)
t1/3
we see that the original equation takes on the form
1 2
f 000 (η) + f (η)f 0 (η) − ηf 0 (η) − f (η) = 0 (16.44)
3 3
247
Solving the above equation, we can find some other general solutions of the
KdV equation combining (16.42) and (16.43).
Let us however proceed further. Using Miura’s transformation (R. M. Miuta:
J. Math. Phys. 9 (1968) 1202-1204)
dw 1 2
f (η) = − w , w = w(η) (16.45)
dη 6

1 1 1 1 1
wηηηη − wwηηη − wwη2 − w2 wηη + w3 wη − ηwηη
3 3 6 18 3 (16.46)
1 2 1 2
+ ηwwη − wη + w = 0
9 3 9
d 1 1 1 1
− w wηηη − w2 wη − ηwη − w = 0 (16.47)
dη 3 6 3 3
from which we obtain the equation
1 1 1
wηηη − w2 wη − ηwη − w = 0 (16.48)
6 3 3
The above equation, integrated once, gives rise to the second Painlevé
transcendent
1 3 1
wηη = w + ηw + a (16.49)
18 3
Details about the Painlevé transcendents can be found in
E. L. Ince: ”Ordinary Differential Equations”, London 1926.
Using MAPLE. We are going to determine the symmetries of the KdV
equation using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
Step 2. We define the function and the independent variables.
> u := U (x, t);
248
The system responds by U (x, t)
Step 3. We write down the PDE we consider.
> KdV := diff(u, t) + u ∗ diff(u, x) + diff(u, x$3) = 0;
∂ ∂ ∂3
U (x, t) + U (x, t) U (x, t) + 3 U (x, t) = 0
∂t ∂x ∂x
Step 4. We find the symmetries using
>DetSys:=DeterminingPDE(KdV ); The program responds
∂ ∂ 1∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = − ξ2 (x, t, U ),
∂U ∂x 3 ∂t
∂2 ∂ ∂
2 − ξ1 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂t ∂U ∂x
∂2 2 ∂ ∂
ξ
− 2 (x, t, U ) = 0, η
− 1 (x, t, U ) = − U − 2ξ (x, t, U ) + − ξ1 (x, t, U )}
∂t2 3 ∂t ∂t
1
{ − ξ1 (x, t, U ) = − C1x + − C3t + − C4, − ξ2 (x, t, U ) = − C1t + − C2,
3
2
− η1 (x, t, U ) = − U − C1 + − C3}
3
fact, using
Inf2:=Infinitesimals(KdV ), the program responds
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],

[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = t,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 1],
1 2
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = − U ]
3 3
Using Mathematica. We shall next list the infinitesimals of the following
KdV-like PDEs. These infinitesimals have been calculated using the Math-
Lie program. We shall demonstrate the procedure for the first equation only.
249
We use the following steps:
In[1]:=<<MathLie.m
Step 2. We find the determining equations of the KdV equation using the
command
In[2] := detEqnsKdV = DeterminingEquations
[{∂t u[x, t] + u[x, t] ∗ ∂x u[x, t] + ∂x,x,x u[x, t]}, {u}, {x, t}, {∂t u[x, t]}];
detEqnsKdV//LT F

Out[2] :=
(ξ1 )u == 0
(ξ2 )u == 0
(φ1 )u,u == 0
(ξ2 )x == 0
φ1 − (ξ1 )t − u(ξ1 )x + u(ξ2 )t − (ξ1 )x,x,x + 3(φ1 )x,x,u == 0
(φ1 )t + u(φ1 )x + (φ1 )x,x,x == 0
− (ξ1 )x,x + (φ1 )x,u == 0
− 3(ξ1 )x + (ξ2 )t == 0
Step 3. The infinitesimals can be determined using the command

In[3] := infinitesimalsKdV = PDESolve[detEqnsKdV, {u}, {x, t}];
infinitesimalsKdV//LT F

k4 x
ξ1 == k1 + k2 t +
3
ξ2 == k3 + k4 t
2k4 u
φ1 == k2 −
3
Equation 1. Equation
ut + uux + uxxx = 0 (16.50)
250
admits the infinitesimals
ξ 1 = k3 + k2 t + k4 x
ξ 2 = k1 + 3k4 t (16.51)
1
φ1 = k2 − 2k4 u
6
ut + uux − ν uxx + µ uxxx = 0 (16.52)

ξ 1 = k2 + k3 t
ξ 2 = k1 (16.53)
φ1 = k3
ut − 6u2 ux + 6λ ux + uxxx = 0 (16.54)

ξ 1 = k2 + k3 (x + 12λ t)
ξ 2 = k1 + 3k3 t (16.55)
φ1 = −k3 u
ut + β uα ux + uxxx = 0 (16.56)

1
ξ 1 = k1 + k3 x
3
ξ 2 = k2 + k3 t (16.57)
2
φ1 = − k3 u
3α
u
+ ut + 6uux + uxxx = 0 (16.58)
t
251
ξ 1 = k1 + k4 + k2 x + k3 ln(t)
ξ 2 = 3k2 t (16.59)
k3
φ1 = − 2k2 u
6t
u
+ ut + 6uux + uxxx = 0 (16.60)
2t
√ 1 √
ξ 1 = k4 + 2k3 t + (2k1 + 3k2 t) x
√ 6
2
ξ = (k1 + k2 t)t (16.61)
√
1 4k3 − 16 k1 t u − 24k2 tu + k2 x
φ = √
24 t
u2x + utx + uuxx + σ uyy + µ uxxxx = 0 (16.62)

6σF 3 + 2σx(F 1)t − y(3(F 2)t + y(F 1)tt )
ξ1 =
6σ
2 2
ξ = F 2 + y(F 1)t
3 (16.63)
ξ3 = F 1
4σu(F 1)t − 6σ(F 3)t − 2σx(F 1)tt + 3y(F 2)tt + y 2 (F 1)ttt
φ1 = −
6σ
ut + 6uux + uxxx
(16.64)
+ (−αu2 ux + β uuxx + γ ux uxx + δ uxxxxx ) = 0
ξ 1 = k1 , ξ 2 = k2 , φ1 = 0 (16.65)
252
ut + αuux + βuxxx = 0 (16.66)

Z t
ξ = k2 + k1 α[t0 ]dt0
1
0
2 (16.67)
ξ =0
φ1 = k1
ut + k1 ux − h(u + xux ) + k0 (6u2 ux + uxxx ) = 0 (16.68)

x(F 1(k0 )t + k0 (F 1)t )
ξ1 = F 2 +
3k0
ξ2 = F 1
u(F 1(k0 )t + k0 (F 1)t )
φ1 = − (16.69)
3k0
3hk0 F 2 + k1 (F 1(k0 )t − 2k0 (F 1)t ) + 3k0 (−F 1(k1 )t + (F 2)t ) = 0
− F 1(k0 )2t + k0 ((k0 )t (F 1)t + F 1(k0 )tt )
+ k02 (3ht F 1 + 3h(F 1)t + (F 1)tt ) = 0
Equation 11. The system of equations
ut + 3αu2 ux + α(v 2 ux + 2uvvx ) + βuxxx = 0

(16.70)
vt + 3αv 2 vx + α(2uvux + u2 vx ) + βvxxx = 0

1
ξ 1 = k2 + k4 x
3
2
ξ = k3 + k4 t
1 (16.71)
φ1 = − k4 u + k1 v
3
1
φ2 = (−3k1 u − k4 v)
3
253
17 The Boussinesq Equation
We consider next the Boussinesq equation
utt + uuxx + (ux )2 + uxxxx = 0, u = u(t, x) (17.1)
Considering the fourth prolongation,

∂ ∂ ∂ ∂ ∂
P r(4) Γ = U + U [x] + U [tt] + U [xx] + U [xxxx] (17.2)
∂u ∂ux ∂utt ∂uxx ∂uxxxx
and denoting by ∆ the quantity
∆ = utt + uuxx + (ux )2 + uxxxx (17.3)
we have
P r(4) Γ(∆) = 0
∂
[x] ∂ [tt] ∂ [xx] ∂ [xxxx] ∂
⇔ U +U +U +U +U
∂u ∂ux ∂utt ∂uxx ∂uxxxx (17.4)
2
(utt + uuxx + (ux ) + uxxxx ) = 0
⇔ U ·uxx + 2U [x] ux + U [tt] + U [xx] u + U [xxxx] = 0
254
Upon substituting the infinitesimals in the above equation, we derive the
equation
U ·uxx + 2[Ux + Uu ux − ut (Tx + ux Tu ) − ux (Xx + ux Xu )]ux

+ Utt + (Uu − 2Tt )utt + (2Utu − Ttt )ut − Xu ux utt
− 3Tu ut utt + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3 − Xuu ux (ut )2
+ [Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux
− Tu ut uxx − 3Xu ux uxx ]u
+ Uxxxx + (4Uxxxu − Xxxxx )ux + (6Uxxu − 4Xxxx )uxx
+ (Uu − 4Xx )uxxxx − (Tu ut + 5Xu ux )uxxxx
(17.5)
− Xuuuu (ux )5 + (Uuuuu − 4Xxuuu )(ux )4 + (4Uxuuu − 6Xxxuu )(ux )3
− 12Txuu ut ux uxx − 12Txxu ux uxt − 6Txxu ut uxx = 0
255
uxxxx by −utt − uux − (ux )2 and obtain the equation
U ·uxx + 2[Ux + Uu ux − ut (Tx + ux Tu ) − ux (Xx + ux Xu )]ux

+ Utt + (Uu − 2Tt )utt + (2Utu − Ttt )ut − Xu ux utt
− 3Tu ut utt + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3 − Xuu ux (ut )2
+ [Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux
− Tu ut uxx − 3Xu ux uxx ]u
+ (Uu − 4Xx )(−utt − uuxx − (ux )2 )
− (Tu ut + 5Xu ux )(−utt − uuxx − (ux )2 ) (17.6)
5 4 3
− Xuuuu (ux ) + (Uuuuu − 4Xxuuu )(ux ) + (4Uxuuu − 6Xxxuu )(ux )
− 12Txuu ut ux uxx − 12Txxu ux uxt − 6Txxu ut uxx = 0
We now have to arrange the above equation in terms of the powers of partial
derivatives of the function u(t, x) and then equate to zero all the coefficients.
However, in order to simplify calculations, since we are doing everything by
hand, it is worth of simplifying further equation (17.6) by looking at some
special coefficients of the partial derivatives. For example we see that
the coefficient of uxxxt is −4Tx , the coefficient of ut uxt is −2Xu , the coefficient
of ut utt is −2Tu and the coefficient of uxt is −2Xt − 2uTx . We obtain further
256
the coefficient of ux uxxx is 4(Uuu − 4Xxu ). Therefore, equating to zero all
these coefficients, we obtain
Tx = 0, Tu = 0 (17.7)
Xt = 0, Xu = 0 (17.8)
Uuu = 0 (17.9)
and then equation (17.6) becomes
U ·uxx + 2(Ux + Uu ux − Xx ux )ux
+ Utt + (Uu − 2Tt )utt + (2Utu − Ttt )ut
+ [Uxx + (2Uxu − Xxx )ux + (Uu − 2Xx )uxx ]u
(17.10)
+ (4Uxu − 6Xxx )uxxx
− (Uu − 4Xx )(utt + uuxx + (ux )2 )
From the above equation we see that the coefficient of utt is given by
(Uu − 2Tt ) − (Uu − 4Xx ) = 2(2Xx − Tt ) and thus we obtain the equation
2Xx − Tt = 0 (17.11)
The coefficient of ut is 2Utu − Ttt and thus we obtain the equation
2Utu − Ttt = 0 (17.12)
The coefficient of uxxx is given by 4Uxu −6Xxx and then we have the equation
2Uxu − 3Xxx = 0 (17.13)

U ·uxx + 2(Ux + Uu ux − Xx ux )ux + Utt
+ [Uxx + (2Uxu − Xxx )ux + (Uu − 2Xx )uxx ]u
(17.14)
− (Uu − 4Xx )(uuxx + (ux )2 )
Equating the coefficient of uxx and the constant term to zero we have the
equations
U + u(Uu − 2Xx ) + (6Uxxu − 4Xxxx ) − u(Uu − 4Xx ) = 0 (17.15)
257
and
Utt + uUxx + Uxxxx = 0 (17.16)
From (17.14) then we obtain
2(Ux + Uu ux − Xx ux )ux + [(2Uxu − Xxx )ux ]u

(17.17)
+ (4Uxxxu − Xxxxx )ux − (Uu − 4Xx )(ux )2
The coefficient of (ux )2 is 2(Uu − Xx ) − (Uu − 4Xx ) = Uu + 2Xx and thus we

have the equation
Uu + 2Xx = 0 (17.18)
We finally equate to zero the coefficient of ux and we obtain the equation
2Ux + u(2Uxu − Xxx ) + (4Uxxxu − Xxxxx ) = 0 (17.19)
Therefore we have obtained the determining equations of the symmetries for

the Boussinesq equation.
We are going to solve the system of equations (17.7)-(17.9), (17.11)-(17.13),
(17.15)-(17.16) and (17.18)-(17.19).
From (17.7), (17.8) and (17.9) we obtain
T = T (t), X = X(x) and U = a(t, x)u + b(t, x) (17.20)
respectively.
From (17.11) we have 2X 0 (x) = T 0 (t) and since each member depends on
different variable, each one should be a constant: X 0 (x) = a1 and T 0 (t) = 2a1 .
Therefore
X(x) = a1 x + a2 and T (t) = 2a1 t + a3 (17.21)
From (17.12) we obtain at (t, x) = 0 and thus
a = a(x) (17.22)
From (17.13) we obtain 2a0 (x) − 3X 00 (x) = 0 and thus a0 (x) = 0 from which
we have
a(x) = a4 (17.23)
258
Let us now collect everything we have found so far:
X(x) = a1 x + a2 , T (t) = 2a1 t + a3 and U = a4 u + b(t, x) (17.24)
From (17.18) we have a4 + 2a1 = 0 and thus a4 = −2a1 . Therefore
U = −2a1 u + b(t, x) (17.25)
Equation (17.15) gives b(t, x) = 0 and thus
U = −2a1 u (17.26)
Equations (17.16) and (17.19) are satisfied identically.

The operator Γ has the form
∂ ∂ ∂
Γ=X +T +U
∂x ∂t ∂u
∂ ∂ ∂
= (a1 x + a2 ) + (2a1 t + a3 ) − 2a1 u (17.27)
∂x ∂t ∂u
∂ ∂ ∂ ∂ ∂
= a1 x + 2t − 2u + a2 + a3
∂x ∂t ∂u ∂x ∂t
Therefore the Lie algebra is being spanned by three generators given by
∂ ∂ ∂ ∂ ∂
Γ1 = x + 2t − 2u , Γ2 = , Γ3 = (17.28)
∂x ∂t ∂u ∂x ∂t
We shall determine a solution of the Boussinesq equation associated to the
generator Γ1 . The Invariant Surface Condition reads
dx dt du
= = (17.29)
x 2t −2u
The above system admits two independent solutions
x2 = C1 t and ut = C2 (17.30)
We thus consider the substitution

1
u = f (ξ) (17.31)
t
where
x2
ξ= (17.32)
t
259
Under the considered substitution, since
2x 0 4x2 00 2
ux = 2
f (ξ), uxx = 3
f (ξ) + 2 f 0 (ξ)
t t t
4 2
16x 48x 12
uxxxx = 5 f (iv) (ξ) + 4 f 000 (ξ) + 3 f 00 (ξ)
t t t
the original Boussinesq equation (17.1) can be cast to the form
t 0 x2 00 t
2 2
f (ξ) + 4f (ξ) + f (ξ) + 4f (ξ)f 00 (ξ) + 2 2 f (ξ)f 0 (ξ)+
x t x
2
2 x t
+ 4 f 0 (ξ) + 16 f (iv) (ξ) + 48f 000 (ξ) + 12 2 f 00 (ξ) = 0
t x
which, taking into account (17.32), is equivalent to the equation
2f (ξ) + 4ξf 0 (ξ) + ξ 2 f 00 (ξ) + 4ξf (ξ)f 00 (ξ) + 2f (ξ)f 0 (ξ)
(17.33)
+ 4ξ(f 0 (ξ))2 + 16ξ 2 f (iv) (ξ) + 48ξf 000 (ξ) + 12f 00 (ξ) = 0
Upon finding the general solution of the above equation, we can determine
some general solutions of the Boussinesq equation, combining (17.31) and
(17.32).
Using MAPLE. We are going to determine the symmetries of some equa-
tions which are also called Boussinesq equations in literature, using MAPLE.
Bussinesq 1. Reference: P.A. Clarkson and M. D. Kruskal: ”New Similarity
reductions of the Boussinesq equation”. J. Math. Phys. 30 (1989) 2201-2213.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> Buss := diff(u, t$2)+a∗diff(u, x$2)+b∗diff(u2 , x$2)+c∗diff(u, x$4) = 0;
∂2 ∂2 ∂ 2 ∂2
U (x, t) + a U (x, t) + b 2 U (x, t) + 2U (x, t) U (x, t)
∂t2 ∂x2 ∂x ∂x2
∂4
+c U (x, t) =0
∂x4
260
>DetSys:=DeterminingPDE(Buss); The program responds
∂ ∂ ∂ 1∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = − ξ2 (x, t, U ),
∂U ∂t ∂x 2 ∂t
∂ ∂ ∂2
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
1 ∂t − ξ2 (x, t, U ) (a + 2bU )
η
− 1 (x, t, U ) = − }
2 b
1
{ − ξ1 (x, t, U ) = − C1x + − C3, − ξ2 (x, t, U ) = − C1t + − C2,
2
1 − C1(a + 2bU )
− η1 (x, t, U ) = − }
2 b
fact, using
Inf2:=Infinitesimals(Buss), the program responds
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],

[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
1 1 −a − 2bU
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = ]
2 2 b
Bussinesq 2. Reference: P.A. Clarkson and M. D. Kruskal: ”New Similarity
reductions of the Boussinesq equation”. J. Math. Phys. 30 (1989) 2201-2213.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
1
> Buss := diff(u, t$2) + ∗ diff(u2 , x$2) + diff(u, x$4) = 0;
2
261
∂2 ∂ 2 ∂2 ∂4
U (x, t) + U (x, t) + U (x, t) U (x, t) + U (x, t) = 0
∂t2 ∂x ∂x2 ∂x4
∂ ∂ ∂ 1∂
∂U ∂t ∂x 2 ∂t
∂ ∂ ∂2
ξ
− 2 (x, t, U ) = 0, ξ
− 2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
− η1 (x, t, U ) = −U − ξ2 (x, t, U ) }
∂t
1
{ − ξ1 (x, t, U ) = − C1x + − C3, − ξ2 (x, t, U ) = − C1t + − C2,
2
− η1 (x, t, U ) = −U − C1}

fact, using
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
1
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = −U ]
2
Bussinesq 3. Reference: T. Nishitani and M. Tajiri: ”On Similarity Solu-
tions of the Boussinesq equation”. Phys. Lett. A 89 (1982) 379-380.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> Buss := diff(u, t$2) − diff(u, x$2) + diff(u2 , x$2) + diff(u, x$4) = 0;
262
∂2 ∂2 ∂ 2 ∂2
U (x, t) − U (x, t) + 2 U (x, t) + 2U (x, t) U (x, t)
∂t2 ∂x2 ∂x ∂x2
4
∂
+ 4 U (x, t) = 0
∂x
∂ ∂ ∂ 1∂
∂U ∂t ∂x 2 ∂t
∂ ∂ ∂2
ξ
− 2 (x, t, U ) = 0, ξ
− 2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
1 ∂
η
− 1 (x, t, U ) = − ξ
− 2 (x, t, U ) (−1 + 2U )}
2 ∂t
1
{ − ξ1 (x, t, U ) = − C1x + − C3, − ξ2 (x, t, U ) = − C1t + − C2,
2
1
− η1 (x, t, U ) = − C1 − − C1U }
2
fact, using
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],

[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
1 1
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = − U ]
2 2
Bussinesq 4. Reference: T. Nishitani and M. Tajiri: ”On Similarity Solu-
tions of the Boussinesq equation”. Phys. Lett. A 89 (1982) 379-380.
>restart;
>with(DEtools):
>with(PDEtools):
263
> u := U (x, t);
> Buss := diff(u, t$2) − diff(u, x$2) − diff(u2 , x$2) − diff(u, x$4) = 0;

∂2 ∂2 ∂ 2 ∂2
U (x, t) − U (x, t) − 2 U (x, t) − 2U (x, t) U (x, t)
∂t2 ∂x2 ∂x ∂x2
∂4
− U (x, t) = 0
∂x4
∂ ∂ ∂ 1∂
∂U ∂t ∂x 2 ∂t
∂ ∂ ∂2
ξ
− 2 (x, t, U ) = 0, ξ
− 2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
1 ∂
η
− 1 (x, t, U ) = − ξ
− 2 (x, t, U ) (1 + 2U )}
2 ∂t
1
{ − ξ1 (x, t, U ) = − C1x + − C3, − ξ2 (x, t, U ) = − C1t + − C2,
2
1
− η1 (x, t, U ) = − − C1 − − C1U }
2
fact, using
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],

[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
1 1
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = − − U ]
2 2
Bussinesq 5. Reference: P. Rosenau and J. L. Schwarzmeier: ”On Similarity
Solutions of the Boussinesq-type equations”. Phys. Lett. A115 (1986) 75-77.
264
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> Buss := diff(u, t$2) = diff(a ∗ u + b ∗ u2 , x$2) + c ∗ diff(u, x$4);
∂2 ∂2 ∂ 2 ∂2
U (x, t) = a U (x, t) + 2b U (x, t) + 2bU (x, t) U (x, t)
∂t2 ∂x2 ∂x ∂x2
∂4
+c U (x, t)
∂x4
∂ ∂ ∂ 1∂
∂U ∂t ∂x 2 ∂t
∂ ∂ ∂2
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
1 ∂t − ξ2 (x, t, U ) (a + 2bU )
− η1 (x, t, U ) = − }
2 b
1
{ − ξ1 (x, t, U ) = − C1x + − C3, − ξ2 (x, t, U ) = − C1t + − C2,
2
1 − C1(a + 2bU )
− η1 (x, t, U ) = − }
2 b
fact, using
265
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
1 1 −a − 2bU
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = ]
2 2 b
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> Buss := diff(u, t$2) = diff(a ∗ u + b ∗ u2 , x$2) + c ∗ diff(u, x$2, t$2);
∂2 ∂2 ∂ 2 ∂2
U (x, t) = a U (x, t) + 2b U (x, t) + 2bU (x, t) U (x, t)
∂t2 ∂x2 ∂x ∂x2
∂4
+c U (x, t)
∂x2 ∂t2
∂ ∂ ∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂t ∂x
∂ ∂ ∂2
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
− ξ2 (x, t, U ) (a + 2bU )
∂t
− η1 (x, t, U ) = − }
b
{ − ξ1 (x, t, U ) = − C3, − ξ2 (x, t, U )
= − C1t + − C2,
1 (−a − 2bU ) − C1
− η1 (x, t, U ) = − }
2 b
266
fact, using
[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],
−a − 2bU
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = ]
b
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> Buss := diff(u, x$2) + b ∗ diff(u2 , x, t) + c ∗ diff(u, x$2, t$2) = 0;
∂2 ∂ ∂ ∂2
a U (x, t) + b 2 U (x, t) U (x, t) + 2U (x, t) U (x, t)
∂x2 ∂t ∂x ∂x∂t
∂4
+c U (x, t)
∂x2 ∂t2
∂ ∂ ∂2
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂t ∂x2
∂ ∂ ∂
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂t ∂x
∂
η
− 1 (x, t, U ) = − ξ
− 1 (x, t, U ) U}
∂x
{ − ξ1 (x, t, U ) = − C1x + − C2, − ξ2 (x, t, U ) = − C3,
− η1 (x, t, U ) = − − C1U }
267
fact, using
[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],

[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = x,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = −U ]
18 The BBM Equation

We consider next the BBM equation (after Bona, Burgers and Mahony)
ut + ux + uux − uxxt = 0 (18.1)
Invariance of the BBM equation under infinitesimal point transformations is

being expressed by the condition
P r(3) Γ(∆) =0 (18.2)

∆=0
where
∆ = ut + ux + uux − uxxt (18.3)
and P r(3) Γ is the third prolongation given by

∂ ∂ ∂ ∂ ∂ ∂
P r(3) Γ = X +T +U + U [t] + U [x] + U [xxt] (18.4)
∂x ∂t ∂u ∂ut ∂ux ∂uxxt
We then have
P r(3) Γ(∆)
∂ ∂ ∂ ∂ ∂ ∂
= X +T +U + U [t] + U [x] + U [xxt]
∂x ∂t ∂u ∂ut ∂ux ∂uxxt (18.5)
(ut + ux + uux − uxxt )
⇔ U ·ux + U [t] + uU [x] − U [xxt]
268
Upon substituting U [t] , U [x] and U [xxt] , we find that P r(3) Γ(∆) is equivalent
to
+ u[Ux + (Uu − Xx )ux − Tx ut − Xu (ux )2 − Tu ux ut ]

− Uxxt + (Uuuu − 2Xxuu − Ttuu )ut (ux )2

+ (Uxxu − Txxt )ut + (2Uxtu − Xxxt )ux
+ (Uu − 2Xx − Tt )uxxt − (3Xu ux + 2Tu ut )uxxt
+ (2Uuu − 4Xxu − 2Ttu )ux uxt − Xuuu ut (ux )3
(18.6)
+ (2Uxuu − Xxxu − 2Txtu )ut ux + (Utuu − 2Xxtu )(ux )2
+ (Uuu − 2Xxu − Ttu )ut uxx − 2Tu (uxt )2 − Txx utt
− 3Xtu ux uxx − 3Xu uxt uxx − Tu utt uxx − Txxu (ut )2
− 2Tx uxtt − 2Tu ux uxtt − Xtuu (ux )3 − Xt uxxx
− Xu ut uxxx − 2Txuu ux (ut )2 − Tuuu (ux )2 (ut )2
− 2Txu ux utt − 4Txu ut uxt − 3Xuu (ux )2 uxt

2 2
− 3Xuu ut ux uxx − Tuu (ux ) utt − 4Tuu ux ut uxt − Tuu (ut ) uxx
269
In the above expression we substitute ut by −ux − uux + uxxt to account for
the condition ∆ = 0. We thus see that
P r(3) Γ(∆) =0
∆=0
⇔ U ·ux + Ut + (Uu − Tt − Xu ux )(−ux − uux + uxxt )
− Tu (−ux − uux + uxxt )2 − Xt ux
+ u [ Ux + (Uu − Xx )ux − Xu (ux )2
− (Tx + Tu ux )(−ux − uux + uxxt ) ]

− Uxxt + (Uuuu − 2Xxuu − Ttuu )(−ux − uux + uxxt )(ux )2

+ (Uxxu − Txxt )(−ux − uux + uxxt ) + (2Uxtu − Xxxt )ux
+ (Uu − 2Xx − Tt )uxxt − (3Xu ux + 2Tu (−ux − uux + uxxt ))uxxt
+ (2Uuu − 4Xxu − 2Ttu )ux uxt − Xuuu (−ux − uux + uxxt )(ux )3
+ (2Uxuu − Xxxu − 2Txtu )(−ux − uux + uxxt )ux
+ (Utuu − 2Xxtu )(ux )2
+ (Uuu − 2Xxu − Ttu )(−ux − uux + uxxt )uxx
− 2Tu (uxt )2 − Txx utt − 3Xtu ux uxx − 3Xu uxt uxx − Tu utt uxx
− Txxu (−ux − uux + uxxt )2 − 2Tx uxtt − 2Tu ux uxtt
− Xtuu (ux )3 − Xt uxxx − Xu (−ux − uux + uxxt )uxxx
− 2Txuu ux (−ux − uux + uxxt )2 − Tuuu (ux )2 (−ux − uux + uxxt )2
− 2Txu ux utt − 4Txu (−ux − uux + uxxt )uxt − 3Xuu (ux )2 uxt
− 3Xuu (−ux − uux + uxxt )ux uxx
− Tuu (ux )2 utt − 4Tuu ux (−ux − uux + uxxt )uxt

2
− Tuu (−ux − uux + uxxt ) uxx = 0
(18.7)
We have next to equate to zero the coefficients of the partial derivatives of

the function u. However, just for the economy of the calculations, we can
equate to zero the coefficients of uxt uxx , utt uxx , and uxtt first, since these
coefficients can easily be identified. Equating to zero all these coefficients,
270
we find that
Xu = 0, Tu = 0, Tx = 0 (18.8)
and thus
X = X(x, t) and T = T (t) (18.9)
Beacuase of the above, equation (18.7) becomes
P r(3) Γ(∆) =0
∆=0
⇔ U ·ux + Ut + (Uu − Tt )(−ux − uux + uxxt ) − Xt ux
+ u [ Ux + (Uu − Xx )ux ]

− Uxxt + Uuuu (ux )2 + (Utu − 2Xxt )uxx + (2Uxu − Xxx )uxt
(18.10)
+ Uxxu (−ux − uux + uxxt ) + (2Uxtu − Xxxt )ux
+ (Uu − 2Xx − Tt )uxxt + 2Uuu ux uxt
+ 2Uxuu (−ux − uux + uxxt )ux + Utuu (ux )2

+ Uuu (−ux − uux + uxxt )uxx − Xt uxxx = 0
From the above equation we see that the coefficient of uxxx is Xt and thus
Xt = 0 which means that X = X(x). On the other hand, the terms uxx uxxt
and ux uxt have coefficients Uuu and 2Uuu respectively. Therefore Uuu = 0.
Since
X = X(x) and Uuu = 0 (18.11)
U ·ux + Ut + (Uu − Tt )(−ux − uux + uxxt )

+ u [ Ux + (Uu − Xx )ux ]

− Uxxt + Utu uxx + (2Uxu − Xxx )uxt
(18.12)
+ Uxxu (−ux − uux + uxxt ) + 2Uxtu ux

+ (Uu − 2Xx − Tt )uxxt = 0
271
Equating to zero the coefficient of uxxt we get the equation
2Xx − Uxxu = 0 (18.13)
Equating to zero the coefficient of uxt we obtain the equation
2Uxu − Xxx = 0 (18.14)
Equating to zero the coefficient of uxx we find
Utu = 0 (18.15)
The zeroth order dervative term provides the equation
Ut + uUx − Uxxt = 0 (18.16)
Finally, equating to zero the coefficient of ux , we get the equation
U − (1 + u)(Uu − Tt ) + u (Uu − Xx )
(18.17)
+ (1 + u)Uxxu − 2Uxtu = 0
The system of equations Uuu = 0 and Utu = 0 admits the general solution
U = α(x)u + β(x, t) (18.18)
Because of the above expression of U , we obtain from (18.13) and (18.14)

the system
2Xx − α00 (x) = 0, 2α0 (x) − Xxx = 0 (18.19)
βt + u[α0 (x)u + βx ] − βxxt = 0 (18.20)
from which, equating to zero the coefficients of u, we get the system
α0 (x) = 0, βx = 0, βt − βxxt = 0 (18.21)
The above system gives
α = a1 and β = a2 (18.22)
272
X = a3 and U = a1 u + a2 (18.23)
Equation (18.17) then takes the form
a1 u + a2 + (1 + u)Tt = 0 (18.24)
Equating to zero the coefficients of u, we obtain the system
a1 + Tt = 0, a2 + Tt = 0 (18.25)
From the above system we obtain
T = −a1 t + a4 and a2 = a1 (18.26)
Therefore
X = a3 , T = −a1 t + a4 , U = a1 u + a1 (18.27)
The generator of the Lie symmetries of the BBM equation is thus given by
∂ ∂ ∂
X = a3 + (−a1 t + a4 ) + a1 (u + 1) (18.28)
∂x ∂t ∂u

∂ ∂ ∂ ∂
X = a1 − t + (u + 1) + a3 + a4 (18.29)
∂t ∂u ∂x ∂t
Therefore the generators of the symmetries of the BBM equation are given
by
∂ ∂ ∂ ∂
X 1 = −t + (u + 1) , X2 = , X3 = (18.30)
∂t ∂u ∂x ∂t
Considering the generator aX 2 + X 3 , i.e.
∂ ∂
a + (18.31)
∂x ∂t
we obtain the invariants ξ and F given by
ξ = at − x and u = F (ξ) (18.32)
273
Taking into account the above invariants, the original BBM equation takes
on the form
aF 000 (ξ) + (a − 1)F 0 (ξ) − F (ξ)F 0 (ξ) = 0 (18.33)
The above equation can be integrated once, giving

1−a 1
F 00 = C + F + F2 (18.34)
a 2a
The above equation is the differential equation of the undamped quadratic
anharmonic oscillator and admits a general solution expressed in terms of
Jacobi’s sn(ζ, m) elliptic function. In fact, multiplying by F 0 both members
of (18.34), we get the equation
1 d 0 2 d 1−a d 2 1 d 3
(F ) = C F + F + F (18.35)
2 dξ dξ 2a dξ 6a dξ
which by integration gives (setting the integration constant equal to zero)

1−a 2 1
(F 0 )2 = 2CF + F + F3 (18.36)
a 3a
If α1 , α2 and α3 are the three real roots of the cubic polynomial at the right
hand side of (18.36) (with α1 < α2 < α3 ), this equation becomes
dF 2 1
= (F − α1 )(F − α2 )(F − α3 ) (18.37)
dξ 3a
The substitution
F − α3
Z2 = , i.e F = a3 + (α2 − α3 )Z 2 (18.38)
α2 − α3
converts (18.37) into the equation
dZ 2
= λ2 (1 − Z 2 )(1 − m2 Z 2 ) (18.39)
dξ
where
α3 − α2 1
m2 = , λ2 = − (α3 − α1 ) (18.40)
α3 − α1 12a
274
Equation (18.39) admits the general solution
Z = sn(λξ, m) (18.41)
where sn(z, m) is Jacobi’s elliptic function. Using the second of (18.38), we
obtain the general solution of (18.34):
F (ξ) = α3 + (α2 − α3 )sn2 (λξ, m) (18.42)
Another choice would be the linear combination aX 2 + X 1 , however this case
does not lead to a solvable equation.
Using MAPLE. We are going to determine the symmetries of the BBM
equation using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> BBM := diff(u, t) + diff(u, x) + u ∗ diff(u, x) − diff(u, x$2, t) = 0;
∂ ∂ ∂ ∂3
U (x, t) + U (x, t) + U (x, t) U (x, t) − U (x, t) =0
∂t ∂x ∂x ∂x2 ∂t
>DetSys:=DeterminingPDE(BBM ); The program responds
∂ ∂ ∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂t ∂x
∂ ∂ ∂2
ξ
− 2 (x, t, U ) = 0, ξ
− 2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
η
− 1 (x, t, U ) = − − 2ξ (x, t, U ) (U + 1)}
∂t
{ − ξ1 (x, t, U ) = − C3, − ξ2 (x, t, U ) = − C1t + − C2,
− η1 (x, t, U ) = − − C1U − − C1}
275
fact, using
Inf2:=Infinitesimals(BBM ), the program responds
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = 1, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = 1,− ξ2 (x, t, U ) = 0, − η1 (x, t, U ) = 0],
[− ξ1 (x, t, U ) = 0,− ξ2 (x, t, U ) = t, − η1 (x, t, U ) = −U − 1]
The above exactly match the generators given by (18.30).
19 Partial Differential Equations of

Higher Dimensions
19.1. The Invariance Condition. In this Section we consider equations
of the form
F (t, x, y, u, ut , ux , uy , utx , uxx , ..., uyy ) = 0 (19.1)
Invariance of the above equation is being expressed by the condition
P r(2) Γ(F )|F =0 = 0 (19.2)
where the operator Γ is defined by

∂ ∂ ∂ ∂
Γ=T +X +Y +U (19.3)
∂t ∂x ∂y ∂u
and P r(1) Γ, P r(2) Γ are the first and second prolongations respectively:
∂ ∂ ∂
P r(1) Γ = Γ + U [t] + U [x] + U [y] (19.4)
∂ut ∂ux ∂uy
∂ ∂ ∂
P r(2) Γ = P r(1) Γ + U [tt] + U [tx] + U [ty]
∂utt ∂utx ∂uty
(19.5)
[xx] ∂ ∂ ∂
+U + U [xy] + U [yy]
It is obvious that
T = T (t, x, y, u), X = X(t, x, y, u),
(19.6)
Y = Y (t, x, y, u), U = U (t, x, y, u)
276
We also have the following expressions for the extended infinitesimals
U [t] = Dt (U ) − ut Dt (T ) − ux Dt (X) − uy Dt (Y ) (19.7)
U [x] = Dx (U ) − ut Dx (T ) − ux Dx (X) − uy Dx (Y ) (19.8)
[y]
U = Dy (U ) − ut Dy (T ) − ux Dy (X) − uy Dy (Y ) (19.9)
and
U [tt] = Dt (U [t] ) − utt Dt (T ) − utx Dt (X) − uty Dt (Y ) (19.10)
U [tx] = Dx (U [t] ) − utt Dx (T ) − utx Dx (X) − uty Dx (Y ) (19.11)
U [ty] = Dy (U [t] ) − utt Dy (T ) − utx Dy (X) − uty Dy (Y ) (19.12)
U [xx] = Dx (U [x] ) − utx Dx (T ) − uxx Dx (X) − uxy Dx (Y ) (19.13)
U [xy] = Dx (U [y] ) − uty Dx (T ) − uxy Dx (X) − uyy Dx (Y ) (19.14)
U [yy] = Dy (U [y] ) − uty Dy (T ) − uxy Dy (X) − uyy Dy (Y ) (19.15)
The total differential operators Dt , Dx and Dy are defined by
∂ ∂ ∂ ∂ ∂ ∂
Dt = + ut + utt + utx + uty + uttt + · · · (19.16)
∂t ∂u ∂ut ∂ux ∂uy ∂utt
∂ ∂ ∂ ∂ ∂ ∂
Dx = + ux + uxt + uxx + uxy + uxtt + · · · (19.17)
∂x ∂u ∂ut ∂ux ∂uy ∂utt
∂ ∂ ∂ ∂ ∂ ∂
Dy = + uy + uyt + uyx + uyy + uytt + · · · (19.18)
∂y ∂u ∂ut ∂ux ∂uy ∂utt
The Invariant Surface Condition reads in this case
T ut + Xux + Y uy = U (19.19)
We shall calculate explicitly some of the coefficients U [...] , V [...] .
We have for example
U [t] = Dt (U ) − ut Dt (T ) − ux Dt (X) − uy Dt (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + ut U − ut + ut T − ux + ut X
∂t ∂u ∂t ∂u ∂t ∂u
∂ ∂ (19.20)
− uy + ut Y
∂t ∂u
= Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu ) − uy (Yt + ut Yu )
= Ut + (Uu − Tt )ut − Tu (ut )2 − Xt ux − Xu ux ut − Yt uy − Yu uy ut
277
U [x] = Dx (U ) − ut Dx (T ) − ux Dx (X) − uy Dx (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + ux U − ut + ux T − ux + ux X
∂x ∂u ∂x ∂u ∂x ∂u
∂ ∂ (19.21)
− uy + ux Y
∂x ∂u
= Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu ) − uy (Yx + ux Yu )
= Ux + (Uu − Xx )ux − Xu (ux )2 − Tx ut − Tu ux ut − Yx uy − Yu ux uy
U [y] = Dy (U ) − ut Dy (T ) − ux Dy (X) − uy Dy (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + uy U − ut + uy T − ux + uy X
∂y ∂u ∂y ∂u ∂y ∂u
∂ ∂ (19.22)
− uy + uy Y
∂y ∂u
= Uy + uy Uu − ut (Ty + uy Tu ) − ux (Xy + uy Xu ) − uy (Yy + uy Yu )
= Uy + (Uu − Yy )uy − Yu (uy )2 − Ty ut − Tu uy ut − Xy ux − Xu ux uy
U [xx] = Dx (U [x] ) − utx Dx (T ) − uxx Dx (X) − uxy Dx (Y )
∂ ∂ ∂ ∂ ∂ [x]
= + ux + utx + uxx + uxy U
∂x ∂u ∂ut ∂ux ∂uy
∂ ∂ ∂ ∂ ∂ ∂
− utx + ux T − uxx + ux X − uxy + ux Y
∂x ∂u ∂x ∂u ∂x ∂u
∂ ∂ ∂ ∂ ∂
= + ux + utx + uxx + uxy
∂x ∂u ∂ut ∂ux ∂uy

Ux + (Uu − Xx )ux − Xu (ux )2 − Tx ut − Tu ux ut − Yx uy − Yu ux uy
− utx (Tx + ux Tu ) − uxx (Xx + ux Xu ) − uxy (Yx + ux Yu )

= Uxx + (Uux − Xxx )ux − Xux (ux )2 − Txx ut − Tux ux ut − Yxx uy − Yux ux uy

2
+ ux Uxu + (Uuu − Xxu )ux − Xuu (ux ) − Txu ut − Tuu ux ut − Yxu uy − Yuu ux uy
+ utx (−Tx − Tu ux ) + uxx (Uu − Xx − 2Xu ux − Tu ut − Yu uy ) + uxy (−Yx − Yu ux )
− utx (Tx + ux Tu ) − uxx (Xx + ux Xu ) − uxy (Yx + ux Yu )
278
After collecting everything together and rearranging, we get the following
expression for the coefficient U [xx] :
U [xx] = Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2

− Xuu (ux )3 − Yxx uy − [Txx + 2Txu ux + Tuu (ux )2 + Tu uxx ]ut
(19.23)
− 2Tx utx − 2Tu ux utx − 2Yxu ux uy − Yuu (ux )2 uy
+ (Uu − 2Xx )uxx − 3Xu ux uxx − Yu uy uxx − 2Yx uxy − 2Yu ux uxy
U [yy] = Dy (U [y] ) − uty Dy (T ) − uxy Dy (X) − uyy Dy (Y )

∂ ∂ ∂ ∂ ∂ [y]
= + uy + uty + uxy + uyy U
∂y ∂u ∂ut ∂ux ∂uy
∂ ∂ ∂ ∂ ∂ ∂
− uty + uy T − uxy + uy X − uyy + uy Y
∂y ∂u ∂y ∂u ∂y ∂u
∂ ∂ ∂ ∂ ∂
= + uy + uty + uxy + uyy
∂y ∂u ∂ut ∂ux ∂uy

Uy + (Uu − Yy )uy − Yu (uy )2 − Ty ut − Tu uy ut − Xy ux − Xu ux uy
− uty (Ty + uy Tu ) − uxy (Xy + uy Xu ) − uyy (Yy + uy Yu )

= Uyy + (Uuy − Yyy )uy − Yuy (uy )2 − Tyy ut − Tuy uy ut − Xyy ux − Xuy ux uy

2
+ uy Uyu + (Uuu − Yyu )uy − Yuu (uy ) − Tyu ut − Tuu uy ut − Xyu ux − Xuu ux uy
+ uty (−Ty − Tu uy ) + uxy (−Xy − Xu uy ) + uyy (Uu − Yy − 2Yu uy − Tu ut − Xu ux )
− uty (Ty + uy Tu ) − uxy (Xy + uy Xu ) − uyy (Yy + uy Yu )
expression for the coefficient U [yy] :
− Xyy ux − [Tyy + 2Tyu uy + Tuu (uy )2 + Tu uyy ]ut
(19.24)
− 2Ty uty − 2Tu uy uty − 2Xyu ux uy − Xuu (uy )2 ux
+ (Uu − 2Yy )uyy − 3Yu uy uyy − Xu ux uyy − 2Xy uxy − 2Xu uy uxy
279
We also have for U [tt] :
U [tt] = Dt (U [t] ) − utt Dt (T ) − utx Dt (X) − uty Dt (Y )
∂ ∂ ∂ ∂ ∂ [t]
= + ut + utt + utx + uty U
∂t ∂u ∂ut ∂ux ∂uy
∂ ∂ ∂ ∂ ∂ ∂
− utt + ut T − utx + ut X − uty + ut Y
∂t ∂u ∂t ∂u ∂t ∂u
∂ ∂ ∂ ∂ ∂
= + ut + utt + utx + uty
∂t ∂u ∂ut ∂ux ∂uy

2
Ut + (Uu − Tt )ut − Tu (ut ) − Xt ux − Xu ux ut − Yt uy − Yu ut uy
− utt (Tt + ut Tu ) − utx (Xt + ut Xu ) − uty (Yt + ut Yu )

2
= Utt + (Uut − Ttt )ut − Tut (ut ) − Xtt ux − Xut ux ut − Ytt uy − Yut ut uy

+ ut Utu + (Uuu − Ttu )ux − Tuu (ut )2 − Xtu ux − Xuu ux ut − Ytu uy − Yuu ut uy
+ utt (Uu − Tt − 2Tu ut − Xu ux − Yu uy ) + utx (−Xt − Xu ut ) + uty (−Yt − Yu ut )
− utt (Tt + ut Tu ) − utx (Xt + ut Xu ) − uty (Yt + ut Yu )

expression for the coefficient U [tt] :
U [tt] = Utt + (2Utu − Ttt )ut + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3
− Ytt uy − [Xtt + 2Xtu ut + Xuu (ut )2 + Xu utt ]ux
(19.25)
− 2Xt uxt − 2Xu ut uxt − 2Ytu ut uy − Yuu (ut )2 uy
+ (Uu − 2Tt )utt − 3Tu ut utt − Yu uy utt − 2Yt uyt − 2Yu ut uyt
For the coefficient U [xy] we find, using (19.14)
U [xy] = Uxy + (Uyu − Xxy )ux + (Uxu − Yxy )uy

+ (Uu − Xx − Yy )uxy + (Uuu − Xxu − Yyu )ux uy
− Xy uxx − Yx uyy − Txu ut uy − Tyu ut ux
− Xuu (ux )2 uy − Yuu ux (uy )2 − Xyu (ux )2 − Yxu (uy )2 (19.26)
− 2Xu ux uxy − 2Yu uy uxy − Xu uy uxx − Yu ux uyy
− Tu uy uxt − Tu ux uyt − Tu ut uxy − Txy ut
− Tx uyt − Ty uxt − Tuu ut ux uy
280
19.2. Example. A Nonlinear Diffusion Equation. We consider the
nonlinear diffusion equation
ut = (uux )x + (uuy )y (19.27)
ut − uuxx − (ux )2 − uuyy − (uy )2 = 0 (19.28)
P r(2) Γ(∆)|∆=0 = 0 (19.29)
where
∆≡ut − uuxx − (ux )2 − uuyy − (uy )2 (19.30)

∂ ∂ ∂ ∂ ∂ ∂
P r(2) Γ = U +U [t] +U [x] +U [y] +U [xx] +U [yy] (19.31)
∂u ∂ut ∂ux ∂uy ∂uxx ∂uyy
We have
P r(2) Γ(∆) = 0
∂ ∂ ∂ ∂ ∂ ∂
⇔ U + U [t] + U [x] + U [y] + U [xx] + U [yy] ∆=0
∂u ∂ut ∂ux ∂uy ∂uxx ∂uyy
⇔ − U uxx − U uyy + U [t] − 2U [x] ux − 2U [y] uy − U [xx] u − U [yy] u = 0
⇔ (uxx + uyy )U + 2U [x] ux + 2U [y] uy + (U [xx] + U [yy] )u − U [t] = 0
281
Upon substituting into the above equation the known expressions of the
infinitesimals U [t] , U [x] , U [y] and U [xx] , U [yy] , we obtain the equation
(uxx + uyy )U
+ 2[Ux + (Uu − Xx )ux − Xu (ux )2 − Tx ut − Tu ux ut − Yx uy − Yu ux uy ]ux
+ 2[Uy + (Uu − Yy )uy − Yu (uy )2 − Ty ut − Tu uy ut − Xy ux − Xu ux uy ]uy

+ Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3 − Yxx uy
− [Txx + 2Txu ux + Tuu (ux )2 + Tu uxx ]ut

− 2Tx utx − 2Tu ux utx − 2Yxu ux uy − Yuu (ux )2 uy + (Uu − 2Xx )uxx
− 3Xu ux uxx − Yu uy uxx − 2Yx uxy − 2Yu ux uxy
+ Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3 − Xyy ux
− [Tyy + 2Tyu uy + Tuu (uy )2 + Tu uyy ]ut
− 2Ty uty − 2Tu uy uty − 2Xyu ux uy − Xuu (uy )2 ux + (Uu − 2Yy )uyy

− 3Yu uy uyy − Xu ux uyy − 2Xy uxy − 2Xu uy uxy u
− [Ut + (Uu − Tt )ut − Tu (ut )2 − Xt ux − Xu ux ut − Yt uy − Yu uy ut ] = 0

(19.32)
282
We next take into account the condition ∆ = 0. We thus substitute ut by
uuxx + (ux )2 + uuyy + (uy )2 and we obtain the equation
(uxx + uyy )U
+ 2[Ux + (Uu − Xx )ux − Xu (ux )2
− (Tx + Tu ux )(uuxx + (ux )2 + uuyy + (uy )2 ) − Yx uy − Yu ux uy ]ux
+ 2[Uy + (Uu − Yy )uy − Yu (uy )2
− (Ty + Tu uy )(uuxx + (ux )2 + uuyy + (uy )2 ) − Xy ux − Xu ux uy ]uy

+ Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3 − Yxx uy
− [Txx + 2Txu ux + Tuu (ux )2 + Tu uxx ](uuxx + (ux )2 + uuyy + (uy )2 )

− 2Tx utx − 2Tu ux utx − 2Yxu ux uy − Yuu (ux )2 uy + (Uu − 2Xx )uxx
+ Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3 − Xyy ux
− [Tyy + 2Tyu uy + Tuu (uy )2 + Tu uyy ](uuxx + (ux )2 + uuyy + (uy )2 )
− 2Ty uty − 2Tu uy uty − 2Xyu ux uy − Xuu (uy )2 ux + (Uu − 2Yy )uyy

− [Ut + (Uu − Tt − Xu ux − Yu uy )(uuxx + (ux )2 + uuyy + (uy )2 )

− Tu (uuxx + (ux )2 + uuyy + (uy )2 )2 − Xt ux − Yt uy ] = 0
(19.33)
function u. However we first observe that the coefficient of utx is −2uTx , the
coefficient of ux utx is −2uTu , the coefficient of uty is −2uTy , the coefficient
of uy uty is −2uTu and thus
Tx = 0, Ty = 0, Tu = 0, (19.34)
283
We thus obtain from (19.33) the new equation
(uxx + uyy )U
+ 2[Ux + (Uu − Xx )ux − Xu (ux )2 − Yx uy − Yu ux uy ]ux
+ 2[Uy + (Uu − Yy )uy − Yu (uy )2 − Xy ux − Xu ux uy ]uy

+ Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3
− Yxx uy − 2Yxu ux uy − Yuu (ux )2 uy + (Uu − 2Xx )uxx

− 3Xu ux uxx − Yu uy uxx − 2Yx uxy − 2Yu ux uxy (19.35)
+ Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3
− Xyy ux − 2Xyu ux uy − Xuu (uy )2 ux + (Uu − 2Yy )uyy

− [Ut + (Uu − Tt − Xu ux − Yu uy )(uuxx + (ux )2 + uuyy + (uy )2 )

− Xt ux − Yt uy ] = 0
The coefficients of ux uxx and uy uyy are −2uXu and −2uYu respectively. We
thus obtain two more equations
Xu = 0, Yu = 0 (19.36)
We then have the following dependence
T = T (t), X = X(t, x, y) and Y = Y (t, x, y) (19.37)
Therefore we get from (19.35) the equation
(uxx + uyy )U
+ 2[Ux + (Uu − Xx )ux − Yx uy ]ux + 2[Uy + (Uu − Yy )uy − Xy ux ]uy

+ Uxx + (2Uxu − Xxx )ux + Uuu (ux )2 − Yxx uy + (Uu − 2Xx )uxx
(19.38)
− 2Yx uxy + Uyy + (2Uyu − Yyy )uy + Uuu (uy )2

− Xyy ux + (Uu − 2Yy )uyy − 2Xy uxy u
− [Ut + (Uu − Tt )(uuxx + (ux )2 + uuyy + (uy )2 ) − Xt ux − Yt uy ] = 0
284
From the previous equation we see that the coefficients of ux uy and uxy are
−2(Yx + Xy ) and −2u(Yx + Xy ) respectively. We thus have the equation
Yx + Xy = 0 (19.39)
Let us continue with the rest of the coefficients:

Coefficient of uxx
U + u(Tt − 2Xx ) = 0 (19.40)
Coefficient of uyy
U + u(Tt − 2Yy ) = 0 (19.41)
Coefficient of (ux )2
Uu + uUuu + Tt − 2Xx = 0 (19.42)
Coefficient of (uy )2
Uu + uUuu + Tt − 2Yy = 0 (19.43)
Coefficient of ux
2Ux + u(2Uxu − Xxx − Xyy ) + Xt = 0 (19.44)
Coefficient of uy
2Uy + u(2Uyu − Yyy − Yxx ) + Yt = 0 (19.45)
Zeroth order coefficient
u(Uxx + Uyy ) − Ut = 0 (19.46)
We are now going to solve the system of equations (19.39)-(19.46), taking

into account (19.37).
Subtracting (19.40) from (19.41) or (19.42) from (19.43), we obtain
Xx = Yy (19.47)
Differentiation of (19.39) with respect to y and (19.47) with respect to x we

obtain
Yxy + Xyy = 0 and Xxx − Yxy = 0
285
respectively. Adding these two equations we derive the following equation
Xxx + Xyy = 0 (19.48)
Differentiation of (19.39) with respect to x and (19.47) with respect to y

yields
Yxx + Xyx = 0 and Xxy − Yyy = 0
respectively. Subtracting these two equations we derive the equation
Yxx + Yyy = 0 (19.49)
Because of (19.48) and (19.49), equations (19.44) and (19.45) take on the
form
2Ux + 2uUxu + Xt = 0 (19.50)
and
2Uy + 2uUyu + Yt = 0 (19.51)
respectively.
We also get from (19.40) the equation
U = u(2Xx − Tt )
and by double differentiation with respect to x and y we obtain the equations
Uxx = 2uXxxx and Uyy = 2uXxyy
respectively. Adding these two equations we have
Uxx + Uyy = 2u(Xxx + Xyy )x = 0 (19.52)
because of (19.48). Taking into account this equation, we get from (19.46)
that
Ut = 0 (19.53)
Differentiating (19.40) with respect to u we obtain
Uu + Tt − 2Xx = 0
286
and taking into account (19.42), we have the equation
Uuu = 0 (19.54)
Differentiating equation (19.40) with respect to x and then with respect to

u we obtain the equations
Ux = 2uXxx and Uxu = 2Xxx
respectively. Because of the above two equations, equation (19.50) yields
8uXxx + Xt = 0
Xxx = 0 and Xt = 0 (19.55)
and taking into account (19.48) that
Xyy = 0 (19.56)
From equation (19.41) differentiating with respect to y and then with respect
to u we obtain
Uy = 2uYyy and Uyu = 2Yyy
respectively. Because of these two equations, we obtain from (19.51) the

equation
8uYyy + Yt = 0
From the above equation we obtain the two equations
Yyy = 0 and Yt = 0 (19.57)
and because of (19.49) we also have
Yxx = 0 (19.58)
Since Xt = 0 and Ut = 0, differentiation of (19.40) with respect to t, yields
Ttt = 0 (19.59)
287
and thus
T = a1 t + a2 (19.60)
Since
Xxx = 0, Xyy = 0 and Xt = 0
we have X = X(x, y) and thus
X = a3 x + a4 y + a5 (19.61)
Since
Yxx = 0, Yyy = 0 and Yt = 0
we have Y = Y (x, y) and thus
Y = a7 x + a8 y + a6 (19.62)
Equation (19.39), because of (19.61) and (19.62) becomes a7 +a4 = 0 and thus
a7 = −a4 . Equation (19.47), because of (19.61) and (19.62), gives a3 = a8 .
Therefore, equation (19.62) takes on the form
Y = −a4 x + a3 y + a6 (19.63)
From equation (19.40), taking into account (19.61) and (19.60), we get
U = (2a3 − a1 )u (19.64)
The above equation is compatible to (19.54).

Conclusion. The operator Γ given by (19.3) is given by
∂ ∂ ∂ ∂
Γ=T +X +Y +U
∂t ∂x ∂y ∂u
∂ ∂ ∂
= (a1 t + a2 ) + (a3 x + a4 y + a5 ) + (−a4 x + a3 y + a6 )
∂t ∂x ∂y
∂
+ (2a3 − a1 )u (19.65)
∂u
∂ ∂ ∂ ∂ ∂ ∂
= a1 t − u + a2 + a3 x +y + 2u +
∂t ∂u ∂t ∂x ∂y ∂u
∂ ∂ ∂ ∂
+ a4 y −x + a5 + a6
∂x ∂y ∂x ∂y
288
Therefore the Lie Algebra of equation (19.27) is being spanned by the gen-
erators
∂ ∂ ∂ ∂ ∂ ∂
Γ(1) = t − u , Γ(2) = , Γ(3) = x +y + 2u
∂t ∂u ∂t ∂x ∂y ∂u
(19.66)
∂ ∂ ∂ ∂
Γ(4) =y − x , Γ(5) = , Γ(6) =
∂x ∂y ∂x ∂y
In order to find some solutions, we consider the Invariant Surface Condition
T ut + Xux + Y uy = U (19.67)
(i) The Invariant Surface Condition corresponding to Γ(1) takes the form
tut = −u (19.68)
The auxiliary system which corresponds to the above equation is

dt du
=− (19.69)
1 u
The above equation admits a general solution
F (x, y)
u= (19.70)
t
Using the above expression for u, equation (19.27) gives
(F Fx )x + (F Fy )y + F = 0 (19.71)
(ii) The Invariant Surface Condition corresponding to 2Γ(1) + Γ(3) takes the
form
2tut + xux + yuy = 0 (19.72)

dt dx dy du
= = = (19.73)
2t x y 0
The above system admits a general solution
y t
u=F , 2 (19.74)
x x
289
Introducing the similarity variables
y t
ξ= , η= (19.75)
x x2
and substituting u = F (ξ, η) into the original equation (19.27), we obtain
the equation
− Fηη + ξ 2 (Fξ )2 + 4ξηFξ Fη + 4η 2 (Fη )2 + ξ 2 F Fξξ + 4ξηF Fξη
(19.76)
+ 2ξF Fξ + 4η 2 F Fηη + 6ηF Fηη + Fξ + F Fξξ = 0
(iii) The Invariant Surface Condition corresponding to Γ(4) takes the form
−yux + xuy = 0 (19.77)
dx dy du
= = (19.78)
−y x 0
The above system admits a general solution
u = F (x2 + y 2 , t) (19.79)
Introducing the similarity variable ξ = x2 +y 2 and substitution of the function
u = F (ξ, t) into the original equation (19.27), we get the equation
Ft = 4ξ(Fξ )2 + 4ξF Fξξ + 4F Fξ (19.80)
Equations (19.71), (19.76) and (19.80) have to be reduced further using Lie
point symmetry analysis.
Using MAPLE. We are going to determine the symmetries of equation
(19.27) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
The system responds by U (x, y, t)
> N onlinEqn := diff(u, t) − diff(u ∗ diff(u, x), x) − diff(u ∗ diff(u, y), y) = 0;
290
∂ ∂ 2 ∂2
U (x, y, t) − U (x, y, t) − U (x, y, t) U (x, y, t)
∂t ∂x ∂x2
∂ 2 ∂2
− U (x, y, t) − U (x, y, t) U (x, y, t) =0
∂y ∂y 2
>DetSys:=DeterminingPDE(N onlinEqn); The program responds
∂ ∂
{ − ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = 0,
∂U ∂t
∂ ∂ ∂ ∂
− ξ1 (x, y, t, U ) = − ξ2 (x, y, t, U ), − ξ1 (x, y, t, U ) = − − ξ2 (x, y, t, U )
∂x ∂y ∂y ∂x
∂ ∂ ∂2
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂U ∂t ∂x2
∂2 ∂ ∂
2 − ξ2 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = 0
∂y ∂U ∂x
∂ ∂2 ∂2
ξ
− 3 (x, y, t, U ) = 0, − 3ξ (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂y ∂t2 ∂y∂x
∂ ∂
− η1 (x, t, U ) = U − − ξ3 (x, y, t, U ) + 2 − ξ2 (x, y, t, U ) }
∂t ∂y
{ − ξ1 (x, y, t, U ) = − C1x − − C3y + − C6,
− ξ2 (x, y, t, U ) = − C3x + − C1y + − C2, − ξ3 (x, y, t, U ) = − C3t + − C5
− η1 (x, y, t, U ) = U (2 − C1 − − C4)}
We could find the infinitesimals using the single command >Infinitesimals.

In fact, using
Inf2:=Infinitesimals(N onlinEqn), the program responds
[− ξ1 (x, y, t, U ) = 0,− ξ2 (x, y, t, U ) = 0,− ξ3 (x, y, t, U ) = 1, − η1 (x, y, t, U ) = 0],
[− ξ1 (x, y, t, U ) = y,− ξ2 (x, y, t, U ) = −x,− ξ3 (x, y, t, U ) = 0, − η1 (x, y, t, U ) = 0],
[− ξ1 (x, y, t, U ) = 0,− ξ2 (x, y, t, U ) = 0,− ξ3 (x, y, t, U ) = t, − η1 (x, y, t, U ) = −U ],
[− ξ1 (x, y, t, U ) = x,− ξ2 (x, y, t, U ) = y,− ξ3 (x, y, t, U ) = 0, − η1 (x, y, t, U ) = 2U ]
291
package called MathLie to equation (19.27). The MathLie is an AddOn
package to Mathematica (Reference [4]). We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the nonlinear equation using the command
In[2] := nonlinEqn = ∂t U [x, y, t] == ∂x (U [x, y, t] ∗ ∂x U [x, y, t])

+ ∂y (U [x, y, t] ∗ ∂y U [x, y, t]); nonlinEqn//LT F
Ut − Ux2 − Uy2 − U Ux,x − U Uy,y == 0
In[3] := pro2 = P rolongation[nonlinEqn, {U }, {x, y, t}]; pro2//LT F
Step 4. We now have to take into account the condition ∆ = 0, i.e. we sub-
stitute in the above expression Ut by (U ∗Ux )x +(U ∗Uy )y using the command
In[4] := prolong2 = pro2/. ∂t U [x, y, t]→∂x (U [x, y, t] ∗ ∂x U [x, y, t])

+ ∂y (U [x, y, t] ∗ ∂y U [x, y, t]); prolong2//LT F
Step 5. We collect together terms arranged with respect to Ux , Uy , Uxx , Uyy ,
Uxy , Uxt and Uyt using the command
In[5] := cprolong2 = Collect[prolong2, {∂t U [x, y, t], ∂x U [x, y, t], ∂y U [x, y, t],
∂xx U [x, y, t], ∂xx U [x, y, t], ∂yy U [x, y, t], ∂x,y U [x, y, t], ∂x,t U [x, y, t], ∂y,t U [x, t]}];
cprolong2//LT F
From the the output we ca obtained, we can easily read off the determining
equations. However we can find the determining equations with a single
command
In[6] := detnolinEqn = DeterminingEquations[nonlinEqn, {U }, {x, y, t},
{∂t U [x, y, t]}]; detnolinEqn//LT F
292
ξ1U == 0
ξ2U == 0
ξ3U == 0
ξ3y == 0
ξ3x == 0
− ξ2t − 2φ1y + U ξ2x,x + U ξ2y,y − 2U φ1y,U == 0
− ξ1t − 2φ1x + U ξ1x,x + U ξ1y,y − 2U φ1x,U == 0
φ1t − U φ1x,x − U φ1y,y == 0
ξ1y + ξ2x == 0
ξ1y + ξ2x == 0
− φ1 + 2U ξ1x − U φ1U == 0
2ξ1x − 2φ1U − U φ1U,U == 0
− φ1 + 2U ξ2y − U φ1U == 0
2ξ2y − 2φ1U − U φ1U,U == 0
− ξ3t + φ1U == 0
In[6] := infnonlinEqn = Infinitesimals[nonlinEqn, {U }, {x, y, t}];

infnonlinEqn//LT F
φ1 == −(k2 + 2 k4)U
ξ1 == −k3 − k4 x + 2 k6 y
ξ2 == −k5 − 2 k6 x − k4 y
ξ3 == k1 + k2 t
to determine some solutions of the heat equation. We consider the following
293
infinitesimals (for k2 = 2, k4 = −1 and all the rest k’s equal to zero)
xi[1][x, y, t, U ] = x;
xi[2][x, y, t, U ] = y;
xi[3][x, y, t, U ] = 2 ∗ t;
phi[1][x, y, t, U ] = 0;
In[7] := invar = xi[1] ∗ ∂x F [x, y, t, U ] + xi[2] ∗ ∂y F [x, y, t, U ]
+ xi[3] ∗ ∂t F [x, y, t, U ] + phi[1] ∗ ∂U F [x, y, t, U ] == 0
Out[7] := xFx + yFy + 2tFt == 0
Step 8. We solve the above equation using the DSolve[] command and then
reduce the equation into the form (19.76).
19.3. Example. The Laplace Equation. We consider the Laplace equa-
tion in the plane
uxx + uyy = 0 (19.81)
P r(2) Γ(∆)|∆=0 = 0 (19.82)
where
∆≡uxx + uyy (19.83)

∂ ∂
P r(2) Γ = U [xx] + U [yy] (19.84)
∂uxx ∂uyy
We have
P r(2) Γ(∆) = 0
∂ ∂ (19.85)
⇔ U [xx] + U [yy] ∆ = 0 ⇔ U [xx] + U [yy] = 0
∂uxx ∂uyy
294
The expressions of the infinitesimals U [xx] and U [yy] have been calculated
before and they are given by (19.23) and (19.24) respectively. However,
since the given equation is time-independent, we simply omit the partial
derivatives of the function u with respect to time t. We thus have
U [xx] = Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3
− Yxx uy − 2Yxu ux uy − Yuu (ux )2 uy + (Uu − 2Xx )uxx (19.86)
− Xyy ux − 2Xyu ux uy − Xuu (uy )2 ux + (Uu − 2Yy )uyy (19.87)
− 3Yu uy uyy − Xu ux uyy − 2Xy uxy − 2Xu uy uxy
Upon substitution of (19.86) and (19.87) into(19.85), we obtain the equation
Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3

(19.88)
− 3Yu uy uyy − Xu ux uyy − 2Xy uxy − 2Xu uy uxy = 0
We now have to take into account the condition ∆ = 0. We then substitute

uyy by −uxx into (19.88). We thus obtain the equation
Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3

(19.89)
− Xyy ux − 2Xyu ux uy − Xuu (uy )2 ux − (Uu − 2Yy )uxx
+ 3Yu uy uxx + Xu ux uxx − 2Xy uxy − 2Xu uy uxy = 0
We equate next the coefficients of the various partial derivatives of the func-
tion u to zero. We find
Coefficient of ux :
(2Uxu − Xxx ) − Xyy = 0 (19.90)
295
Coefficient of uy :
−Yxx + (2Uyu − Yyy ) = 0 (19.91)
Coefficient of uxy :
−2Yx − 2Xy = 0 (19.92)
Coefficient of ux uy :
−2Yxu − 2Xyu = 0 (19.93)
−3Xu + Xu = 0 (19.94)
Coefficient of uy uxx :
−Yu + 3Yu = 0 (19.95)
Coefficient of ux uxy :
−2Yu = 0 (19.96)
Coefficient of uy uxy :
−2Xu = 0 (19.97)
(Uu − 2Xx ) − (Uu − 2Yy ) = 0 (19.98)
Uuu − 2Xxu = 0 (19.99)
Coefficient of (uy )2 :
Uuu − 2Yyu = 0 (19.100)
−Xuu = 0 (19.101)
296
Coefficient of (uy )3 :
−Yuu = 0 (19.102)
Coefficient of (ux )2 uy :
−Yuu = 0 (19.103)
Coefficient of (uy )2 ux :
−Xuu = 0 (19.104)
Constant terms (zeroth-order derivative)
Uxx + Uyy = 0 (19.105)
We now have to solve the system of equation (19.90)-(19.105) in order to

determine the coefficients X, Y and U .
First of all, since Xu = 0 and Yu = 0, we conclude that X and Y are
independent of u, i.e.
X = X(x, y) and Y = Y (x, y) (19.106)
Therefore, we get either from (19.99) or (19.100) that Uuu = 0 and thus
U = A(x, y)u + B(x, y) (19.107)
Differentiating (19.92) and (19.98) first with respect to x and then with
respect to y, we obtain
Yxx + Xxy = 0, Yxy + Xyy = 0 (19.108)
and
−Xxx + Yxy = 0, −Xxy + Yyy = 0 (19.109)
respectively. Adding the first of (19.108) to the second of (19.109), and

then subtracting from the second of (19.108) the first of (19.109), we get the
equations
Yxx + Yyy = 0, Xxx + Xyy = 0 (19.110)
297
Because of the above equations, we obtain from (19.90) and (19.91) that
Uxu = 0 and Uyu = 0 (19.111)
Making use of (19.107), we obtain from the two equations of (19.111) that
Ax = 0 and Ay = 0 (19.112)
and thus
A(x, y) = α, α = constant (19.113)
In other words,
U = αu + B(x, y) (19.114)
Using the above expression of U , equation (19.105) becomes
Bxx + Byy = 0 (19.115)
Let us collect everything together. We have found so far that
X = X(x, y) and Y = Y (x, y), U = αu + B(x, y) (19.116)
where X and Y satisfy equations (19.98) and (19.92), i.e.
−Xx + Yy = 0, Xy + Yx = 0 (19.117)
and B(x, y) satisfies Laplace’s equation (19.115).

The generator of Lie point symmetries has thus the form
∂ ∂ ∂
Γ=X +Y +U
∂x ∂y ∂u
(19.118)
∂ ∂ ∂
= X(x, y) + Y (x, y) + αu + B(x, y)
∂x ∂y ∂u
In order to find a particular general solution of the equation uxx +uyy = 0, we
first consider two functions X and Y satisfying (19.117) and another function
B satisfying Laplace’s equation (19.115).
Case 1. We consider
X = x, Y =y and B = 0 (19.119)
298
The above choice satisfies the conditions (19.117). In this case the generator
of the symmetries is given by
∂ ∂ ∂
Γ=x +y + αu (19.120)
∂x ∂y ∂u
and the invariant surface condition reads
xux + yuy = αu (19.121)
y
u = xα F (19.122)
x
where F is an arbitrary function. Letting
y
ξ= (19.123)
x
Laplace’s equation (19.81) takes on the form
(1 + ξ 2 )F 00 (ξ) + 2(1 − α) ξ F 0 (ξ) − α(1 − α)F (ξ) = 0 (19.124)
We shall find some solutions of the above equation for some special values of
the parameter α.
1
For α = equation (19.124) becomes
2
1
(1 + ξ 2 )F 00 (ξ) + ξ F 0 (ξ) − F (ξ) = 0 (19.125)
4
with general solution given by
1 ξ
F (ξ) = C1 cosh arctanh p
2 1 + ξ2
1 (19.126)
ξ
+ C2 sinh arctanh p
2 1 + ξ2
Therefore the general solution of Laplace’s equation (19.81) is given by
(taking into account (19.122), (19.123) and (19.126))
1
1/2
y
u(x, y) = x C1 cosh arctanh p
2 x2 + y 2
1 (19.127)
y
+ C2 sinh arctanh p
2 x2 + y 2
299
For α = −1 equation (19.124) becomes
(1 + ξ 2 )F 00 (ξ) + 4 ξF 0 (ξ) + 2F (ξ) = 0 (19.128)

C1 + C2 ξ
F (ξ) = (19.129)
1 + ξ2
1
u(x, y) = (C1 x + C2 y) (19.130)
x2 + y2
(1 + ξ 2 )F 00 (ξ) + 6 ξF 0 (ξ) + 6F (ξ) = 0 (19.131)

ξ ξ2 − 1
F (ξ) = C1 + C 2 (19.132)
(ξ 2 + 1)2 (ξ 2 + 1)2
xy y 2 − x2
u(x, y) = C1 + C 2 (19.133)
(x2 + y 2 )2 (x2 + y 2 )2
(1 + ξ 2 )F 00 (ξ) + 8 ξF 0 (ξ) + 12F (ξ) = 0 (19.134)

3ξ 2 − 1 ξ 3 − 3ξ
F (ξ) = C1 + C 2 (19.135)
(ξ 2 + 1)3 (ξ 2 + 1)3
3xy 2 − x3 y 3 − 3x2 y
u(x, y) = C1 + C 2 (19.136)
(x2 + y 2 )3 (x2 + y 2 )3
300
(1 + ξ 2 )F 00 (ξ) + 10 ξF 0 (ξ) + 20F (ξ) = 0 (19.137)

ξ3 − ξ ξ 4 − 6ξ 2 + 1
F (ξ) = C1 + C 2 (19.138)
(ξ 2 + 1)4 (ξ 2 + 1)4
xy 3 − x3 y y 4 − 6x2 y 2 + x4
u(x, y) = C1 + C 2 (19.139)
(x2 + y 2 )4 (x2 + y 2 )4
(1 + ξ 2 )F 00 (ξ) + 12 ξF 0 (ξ) + 30F (ξ) = 0 (19.140)

5ξ 4 − 10ξ 2 + 1 ξ 5 − 10ξ 3 + 5ξ
F (ξ) = C1 + C 2 (19.141)
(ξ 2 + 1)5 (ξ 2 + 1)5
5xy 4 − 10x3 y 2 + x5 y 5 − 10x2 y 3 + 5x4 y
u(x, y) = C1 + C2 (19.142)
(x2 + y 2 )5 (x2 + y 2 )5
(1 + ξ 2 )F 00 (ξ) + 14 ξF 0 (ξ) + 42F (ξ) = 0 (19.143)

3ξ 5 − 10ξ 3 + 3ξ ξ 6 − 15ξ 4 + 15ξ 2 − 1
F (ξ) = C1 + C 2 (19.144)
(ξ 2 + 1)6 (ξ 2 + 1)6
3xy 5 − 10x3 y 3 + 3x5 y y 6 − 15x2 y 4 + 15x4 y 2 − x6
u(x, y) = C1 +C 2 (19.145)
(x2 + y 2 )6 (x2 + y 2 )6
301
Case 2. We consider
X = 1, Y = 1 and α = B = 0 (19.146)
The above choice satisfies the conditions (19.117). In this case we have
∂ ∂ ∂
Γ= + + αu (19.147)
∂x ∂y ∂u
The invariant surface condition reads
ux + uy = αu (19.148)
u = eαx G(ξ), ξ =y−x (19.149)
where G is an arbitrary function. Equation (19.81) then becomes
2G00 (ξ) − 2αG0 (ξ) + α2 G(ξ) = 0 (19.150)

n α α o
G(ξ) = eαξ/2 C1 cos ξ + C2 sin ξ (19.151)
2 2
(taking into account (19.149) and (19.151))
n α α o
u(x, y) = eα(y+x)/2 C1 cos (y − x) + C2 sin (y − x) (19.152)
2 2
Case 3. We consider
X = y, Y = −x and α = B = 0 (19.153)
The above choice satisfies the conditions (19.117). In this case we have
∂ ∂
Γ=y −x (19.154)
∂x ∂y
This choice of functions leads to rotational invariant solutions of Laplace’s
equation. The invariant surface condition reads
yux − xuy = 0 (19.155)
302
u = H(ξ), ξ = x2 + y 2 (19.156)
where H is an arbitrary function. Equation (19.81) then becomes
ξH 00 (ξ) + H 0 (ξ) = 0 (19.157)
H(ξ) = C1 + C2 ln(ξ) (19.158)

(taking into account (19.156) and (19.158))
u(x, y) = C1 + C2 ln(x2 + y 2 ) (19.159)
19.4. Example. The Wave Equation. We consider the wave equation
utt = uxx + uyy (19.160)
P r(2) Γ(∆)|∆=0 = 0 (19.161)
where
∆≡utt − uxx − uyy (19.162)

∂ ∂ ∂
P r(2) Γ = U [tt] + U [xx] + U [yy] (19.163)
∂utt ∂uxx ∂uyy
We have
∂ ∂ ∂
P r(2) Γ(∆) = U [tt] + U [xx] + U [yy] (utt − uxx − uyy )
∂utt ∂uxx ∂uyy (19.164)
[tt] [xx] [yy]
⇐⇒ U − U −U =0
303
Upon substitution of U [tt] , U [xx] and U [yy] in the previous equation, we get
Utt + (2Utu − Ttt )ut + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3 − Ytt uy
− [Xtt + 2Xtu ut + Xuu (ut )2 + Xu utt ]ux
− 2Xt uxt − 2Xu ut uxt − 2Ytu ut uy − Yuu (ut )2 uy + (Uu − 2Tt )utt
− 3Tu ut utt − Yu uy utt − 2Yt uyt − 2Yu ut uyt


(19.165)

− Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3


utt by uxx + uyy . Equation (19.165) then takes on the form
Utt + (2Utu − Ttt )ut + (Uuu − 2Ttu )(ut )2 − Tuu (ut )3 − Ytt uy
− [Xtt + 2Xtu ut + Xuu (ut )2 + Xu (uxx + uyy )]ux
− 2Xt uxt − 2Xu ut uxt − 2Ytu ut uy − Yuu (ut )2 uy
+ (Uu − 2Tt )(uxx + uyy ) − 3Tu ut (uxx + uyy )
− Yu uy (uxx + uyy ) − 2Yt uyt − 2Yu ut uyt



− Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3

304
the function u. However we do not proceed further and leave MAPLE to
perform the rest of the calculations.
Using MAPLE. We are going to determine the symmetries of the wave
equation (19.129) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
> W aveEqn := diff(u,t2) = diff(u,x2) + diff(u,y2);
∂2 ∂2 ∂2
U (x, y, t) = U (x, y, t) + U (x, y, t)
∂t2 ∂x2 ∂y 2
305
>DetSys:=DeterminingPDE(W aveEqn); The program responds
∂ ∂2 ∂2
{ − ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = − − ξ1 (x, y, t, U ) ,
∂U ∂t2 ∂y 2
∂2 ∂2 ∂3
− ξ 1 (x, y, t, U ) = − − ξ1 (x, y, t, U ) , 3 − ξ1 (x, y, t, U ) = 0,
∂x2 ∂y 2 ∂y
∂ ∂ ∂
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = − − ξ1 (x, y, t, U ) ,
∂U ∂x ∂y
∂ ∂ ∂2
− ξ2 (x, y, t, U ) = − ξ1 (x, y, t, U ), − ξ2 (x, y, t, U )
∂y ∂x ∂t2
∂2 ∂ ∂
= − ξ1 (x, y, t, U ), − ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U )
∂y∂x ∂U ∂t
∂ ∂ ∂
= − ξ1 (x, y, t, U ), − ξ3 (x, y, t, U ) = − ξ1 (x, y, t, U ),
∂x ∂x ∂t
∂ ∂ ∂2
− ξ3 (x, y, t, U ) = − ξ2 (x, y, t, U ), − η1 (x, y, t, U ) = 0,
∂y ∂t ∂U 2
∂2 ∂2 ∂2
− η 1 (x, y, t, U ) = − η 1 (x, y, t, U ) + − η1 (x, y, t, U ),
∂t2 ∂x2 ∂y 2
∂2 ∂3
− ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = 0,
∂y∂t ∂y 2 ∂x
∂2 1 ∂2 ∂2
− η1 (x, y, t, U ) = − − ξ1 (x, y, t, U ), − η1 (x, y, t, U )
∂t∂U 2 ∂x∂t ∂x∂U
1 ∂2 ∂2 1 ∂2
= ξ
− 1 (x, y, t, U ), η
− 1 (x, y, t, U ) = − − ξ1 (x, y, t, U )}
2 ∂y 2 ∂y∂U 2 ∂y∂x
>Inf1:=pdsolve(DetSys), the program’s output cannot be used as it is. That
is why we find the infinitesimals using the single command >Infinitesimals.
In fact, using
306
Inf2:=Infinitesimals(W aveEqn); we obtain eleven generators:

[− ξ1 (x, y, t, U ) = t,− ξ2 (x, y, t, U ) = 0,− ξ3 (x, y, t, U ) = x, − η1 (x, y, t, U ) = 0],
[− ξ1 (x, y, t, U ) = 0,− ξ2 (x, y, t, U ) = t,− ξ3 (x, y, t, U ) = y, − η1 (x, y, t, U ) = 0],
[− ξ1 (x, y, t, U ) = x,− ξ2 (x, y, t, U ) = y,− ξ3 (x, y, t, U ) = t, − η1 (x, y, t, U ) = 0],
[− ξ1 (x, y, t, U ) = 0,− ξ2 (x, y, t, U ) = 0,− ξ3 (x, y, t, U ) = 0, − η1 (x, y, t, U ) = U ],
[− ξ1 (x, y, t, U ) = y,− ξ2 (x, y, t, U ) = −x,− ξ3 (x, y, t, U ) = 0, − η1 (x, y, t, U ) = 0],
1 1 1
[− ξ1 (x, y, t, U ) = xy,− ξ2 (x, y, t, U ) = y 2 − x2 + t2 ,− ξ3 (x, y, t, U ) = yt,
2 2 2
1
− η1 (x, y, t, U ) = − yU ],
2
1 1 1
[− ξ1 (x, y, t, U ) = xt,− ξ2 (x, y, t, U ) = yt,− ξ3 (x, y, t, U ) = t2 + x2 + y 2 ,
2 2 2
1
− η1 (x, y, t, U ) = − tU ],
2
1 2 1 2 1 2
[− ξ1 (x, y, t, U ) = − x + y − t ,− ξ2 (x, y, t, U ) = −xy,
2 2 2
1
− ξ3 (x, y, t, U ) = −xt, − η1 (x, y, t, U ) = xU ]
2
Using the same procedure, we can find the symmetry generators of the wave
equation utt = uxx + uyy + uzz in three dimensions (see §19.6).
We shall now find a general solution of the wave equation considering the
generator (the fourth from the top)
∂ ∂
X (4) = t +x (19.167)
∂x ∂t
In this case the invariant surface condition takes on the form
tux + xut = 0 (19.168)
which admits the solution
u = U (y, ξ), ξ = −x2 + t2 (19.169)
307
Using the above substitution, the wave equation utt = uxx + uyy becomes
4ξUξξ + 4Uξ − Uyy = 0 (19.170)
The above equation can be solved by separation of variables. In fact consid-

ering that
U = F (ξ)G(y) (19.171)
G00 (y)
4 ξF 00 (ξ) + 4F 0 (ξ) − F (ξ) = (19.172)
G(y)
At this point we introduce the separation constant λ by
G00 (y) G00 (y)
= λ2 or = −λ2 , λ>0 (19.173)
G(y) G(y)
We then obtain
G(y) = C1 eλy + C2 e−λy or G(y) = C1 cos(λy) + C2 sin(λy) (19.174)
4 ξF 00 (ξ) + 4F 0 (ξ) − F (ξ) = ±λ2 (19.175)

p p
F (ξ) = K1 I0 ( ξ) + K2 K0 ( ξ)∓λ2 (19.176)
where I0 and K0 are the modified Bessel’s functions of zeroth order of the
first and second kind respectively. We thus have obtained the following two
solutions corresponding to the generator X (4) :
√ √
u = [K1 I0 ( −x2 + t2 ) + K2 K0 ( −x2 + t2 ) − λ2 ]
(19.177)
×[C1 eλy + C2 e−λy ]
and
√ √
u = [K1 I0 ( −x2 + t2 ) + K2 K0 ( −x2 + t2 ) + λ2 ]×
(19.178)
×[C1 cos(λy) + C2 sin(λy)]
308
where λ is any positive constant.
19.5. Example. The Laplace Equation in three dimensions.

We consider the 3-dimensional Laplace equation
uxx + uyy + uzz = 0, u = u(x, y, z) (19.179)
The generators of Lie algebra of symmetries for this equation are the following
∂ ∂ ∂ ∂
X (1) = , X (2) = , X (3) = , X (4) = u ,
∂x ∂y ∂z ∂u
∂ ∂ ∂ ∂ ∂
X (5) = x +y + z , X (6) = x −y ,
∂x ∂y ∂z ∂y ∂x
∂ ∂ ∂ ∂
X (7) = z − y , X (8) = x − z ,
∂y ∂z ∂z ∂x
(19.180)
∂ ∂ 1 ∂ 1 ∂
X (9) = xz + yz + (z 2 − x2 − y 2 ) − zu ,
∂x ∂y 2 ∂z 2 ∂u
1 ∂ ∂ ∂ 1 ∂
X (10) = (x2 − y 2 − z 2 ) + xy + xz − xu ,
2 ∂x ∂y ∂z 2 ∂u
∂ 1 ∂ ∂ 1 ∂
X (11) = xy + (y 2 − z 2 − x2 ) + yz − yu
∂x 2 ∂y ∂z 2 ∂u
Let us find some solutions of Laplace’s equation considering a linear com-
bination of some generators. For example considering X = X (5) + αX (4) ,
i.e.
∂ ∂ ∂ ∂
X=x +y +z + αu (19.181)
∂x ∂y ∂z ∂u
the invariant surface condition becomes
xux + yuy + zuz = αu (19.182)
The general sulution of the above equation is given by
y z
u = xα F (ξ, η), ξ = , η = (19.183)
x x
The above expression when substituted into the original equation (19.179)
yields
(1 + ξ 2 )Fξξ + (1 + η 2 )Fηη + 2ξηFξη
(19.184)
+ 2(1 − α)ξFξ + 2(1 − α)ηFη − α(1 − α)F = 0
309
Introducing a new similarity variable θ by
θ = κξ − λη (19.185)
equation (19.184) is being converted into
(κ2 + λ2 + θ2 )H 00 (θ) + 2(1 − α)θH 0 (θ) − α(1 − α)H(θ) = 0 (19.186)
For α = 1/2 the solution of equation (19.186) is given by

1 θ
H(θ) = C1 cosh arctanh √
2 κ2 + λ2 + θ2
1 (19.187)
θ
+ C2 sinh arctanh √
2 κ2 + λ2 + θ2
We thus obtain the solution of the 3−dimensional Laplace’s equation (using
(19.183), (19.185) and (19.187))

1/2
1 κy − λz
u(x, y, z) = x C1 cosh arctanh p
2 (κ2 + λ2 )x2 + (κy − λz)2
1 κy − λz
+ C2 sinh arctanh p
2 (κ2 + λ2 )x2 + (κy − λz)2
(19.188)
For α = −1 the solution of equation (19.186) is given by
C1 + C2 θ
H(θ) = (19.189)
κ2+ λ2 + θ 2
(19.183), (19.185) and (19.189))
C1 x + C2 (κy − λz)
u(x, y, z) = (19.190)
(κ2 + λ2 )x2 + (κy − λz)2
θ −κ2 − λ2 + θ2
H(θ) = C1 2 + C2 2 (19.191)
(κ + λ2 + θ2 )2 (κ + λ2 + θ2 )2
310
(19.183), (19.185) and (19.191))
x(κy − λz)
u(x, y, z) = C1
[(κ2+ λ2 )x2+ (κy − λz)2 ]2
(19.192)
−(κ2 + λ2 )x2 + (κy − λz)2
+ C2
[(κ2 + λ2 )x2 + (κy − λz)2 ]2
κ2 + λ2 − 3θ2 (−3κ2 − 3λ2 + θ2 )θ

H(θ) = C1 + C 2 (19.193)
(κ2 + λ2 + θ2 )3 (κ2 + λ2 + θ2 )3

(19.183), (19.185) and (19.193))
x[(κ2 + λ2 )x2 − 3(κy − λz)2 ]

u(x, y, z) = C1
[(κ2 + λ2 )x2 + (κy − λz)2 ]3
(19.194)
[−3(κ2 + λ2 )x2 + (κy − λz)2 ](κy − λz)
+ C2
[(κ2 + λ2 )x2 + (κy − λz)2 ]3
Note 1. Introducing a new similarity variable ρ by
ρ = ξ 2 + η2 (19.195)
equation (19.184) is being reduced to the equation
4ρ(1 + ρ)F 00 (ρ) + 2[2 + 2(3 − 2α)ρ]F 0 (ρ) − α(1 − α)F (ρ) = 0 (19.196)
The general solution of the above equation is expressed in terms of the

hypergeometric function 2 F1 (.)
α 1 − α 1 − 2α
F (ρ) = A· 2 F1 − , ; ;1 + ρ
2 2 2 (19.197)
2 + α 1 + α 3 + 2α
+ B· (1 + ρ)α+1/2 2 F1 , ; ;1 + ρ
2 2 2
where A, B are arbitrary constants and α 6= 1/2, −3/2.
Therefore the solution of the 3−dimensional Laplace’s equation is given by
311
α 1 − α 1 − 2α x2 + y 2 + z 2
α
u(x, y, z) = x A· 2 F1 − , ; ;
2 2 2 x2
x2 + y 2 + z 2 α+1/2
+ B· (19.198)
x2
2 + α 1 + α 3 + 2α x2 + y 2 + z 2
× 2 F1 , ; ;
2 2 2 x2
End of Note 1.
Let us further consider the symmetries associated to rotations along the
z−axis, i.e. the generator
∂ ∂
X (6) = x −y (19.199)
∂y ∂x
Solving the associated invariant surface condition, we obtain the function
u = U (η, ξ), ξ = x2 + y 2 , η=z (19.200)
Then the 3−dimensinal Laplace equation (19.179) becomes
4ξUξξ + 4Uξ + Uηη = 0 (19.201)
We shall develop two solution methods for the above equation:

First Method: We change the variables (ξ, η)−→(ρ, σ) by
ξ = λβ ρ, η = λγ σ (19.202)
Under the above change of variables, equation (19.201) takes on the form
4ρUρρ + 4Uρ + λβ−2γ Uσσ = 0 (19.203)
Invariance requires β − 2γ = 0, i.e. β = 2γ. We thus see that equation

(19.201) is invariant under the change of variables
ξ = λ2γ ρ, η = λγ σ (19.204)
Since
ξ = λ2γ ρ, η 2 = λ2γ σ 2 (19.205)
312
ξ
we may eliminate the scale parametr λ by taking the ratio . Therefore we
η2
can introduce a new similarity variable θ by
ξ
θ= (19.206)
η2
Equation (19.201) then yields
2θ(1 + θ)Uθθ + (2 + 3θ)Uθ = 0 (19.207)
The solution of the above equation is given by

√
U (θ) = C1 + C2 arctanh( 1 + θ) (19.208)
Therefore the solution of the 3−dimensional Laplace’s equation is given by

r
x2 + y 2
u(x, y, z) = C1 + C2 arctanh 1 + 1 + (19.209)
z2
Second Method: Equation (19.201) can be solved by factorization. Consid-
ering
U (η, ξ) = F (ξ)G(η) (19.210)
ξF 00 (ξ) + F 0 (ξ) 1 G00 (η)

+ =0 (19.211)
F (ξ) 4 G(η)
and thus
ξF 00 (ξ) + F 0 (ξ) 1 G00 (η)
=k and = −k (19.212)
F (ξ) 4 G(η)
where k is the separation constant. We thus have derived the equations
ξF 00 (ξ) + F 0 (ξ) − kF (ξ) = 0 (19.213)
and
G00 (η) + 4kG(η) = 0 (19.214)
313
For k = −λ2 (λ is a positive constant) the solution of the system of equations
(19.213) and (19.214) is given by
p p
F (ξ) = C1 J0 (2λ ξ) + C2 Y0 (2λ ξ) (19.215)
and
G(η) = D1 cosh(2λη) + D2 sinh(2λη) (19.216)
where J0 (x), Y0 (x) are the Bessel’s functions.

For k = λ2 (λ is a positive constant) the solution of the system of equations
(19.213) and (19.214) is given by
p p
F (ξ) = C1 I0 (2λ ξ) + C2 K0 (2λ ξ) (19.217)
and
G(η) = D1 cos(2λη) + D2 sin(2λη) (19.218)
where I0 (x), K0 (x) are the Bessel’s functions.

Therefore the solutions of the 3−dimensional Laplace’s equation (19.179) is
given by
n p p o
u(x, y, z) = C1 J0 (2λ x2 + y 2 ) + C2 Y0 (2λ x2 + y 2 )
(19.219)
× {D1 cosh(2λz) + D2 sinh(2λz)}
and
n p p o
u(x, y, z) = C1 I0 (2λ x2 + y 2 ) + C2 K0 (2λ x2 + y 2 )
(19.220)
× {D1 cos(2λz) + D2 sin(2λz)}
respectively.
Note 2. We may also determine the rotational invariant solutions of Laplace’s
equation by considering functions of the form
u = F (ξ), ξ = x2 + y 2 + z 2 (19.221)
Under the above choice, Laplace’s 3−dimensional equation takes on the form
2ξF 00 (ξ) + 3F 0 (ξ) = 0 (19.222)
314
F (ξ) = C1 + C2 ξ −1/2 (19.223)
Therefore the solution of Laplace’s equation is given by
u(x, y, z) = C1 + C2 (x2 + y 2 + z 2 )−1/2 (19.224)
19.6. Example. The Wave Equation in three dimensions.

We consider the 3-dimensional Wave equation
utt = uxx + uyy + uzz , u = u(x, y, z, t) (19.225)
The generators of Lie algebra of symmetries for this equation are given by
∂ ∂ ∂ ∂
X (1) = , X (2) = , X (3) = , X (4) = ,
∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂ ∂
X (5) = u , X (6) = t + z , X (7) = t +y ,
∂u ∂z ∂t ∂y ∂t
∂ ∂ ∂ ∂ ∂ ∂
X (8) = t + x , X (9) = x +y +z +t ,
∂x ∂t ∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂ ∂ ∂
X (10) = −z + y , X (11) = z − x , X (12) = y −x ,
∂y ∂z ∂x ∂z ∂x ∂y
1 ∂ ∂ ∂ ∂ ∂
X (13) = (x2 + t2 − y 2 − z 2 ) + xy + xz + xt − xu ,
2 ∂x ∂y ∂z ∂t ∂u
∂ ∂ ∂ 1 ∂ ∂
X (14) = xt + yt + zt + (t2 + x2 + y 2 + z 2 ) − tu ,
∂x ∂y ∂z 2 ∂t ∂u
∂ ∂ 1 ∂ ∂ ∂
X (15) = xz + yz + (t2 + z 2 − x2 − y 2 ) + tz − zu ,
∂x ∂y 2 ∂z ∂t ∂u
∂ 1 ∂ ∂ ∂ ∂
X (16) = xy + (t2 + y 2 − x2 − z 2 ) + xz + ty − yu
∂x 2 ∂y ∂z ∂t ∂u
(19.226)
Let us find some solutions of the Wave equation considering a linear com-
bination of some generators. For example considering X = X (9) + αX (5) ,
i.e.
∂ ∂ ∂ ∂ ∂
X=x +y +z + t + αu (19.227)
∂x ∂y ∂z ∂t ∂u
315
the invariant surface condition becomes
xux + yuy + zuz + tut = αu (19.228)
The general sulution of the above equation is given by

y z t
u = xα F (ξ, η, θ), ξ= , η= , θ= (19.229)
x x x
The above expression when substituted into the original equation (19.225)
yields
(1 + ξ 2 )Fξξ + (1 + η 2 )Fηη − (1 − θ2 )Fθθ + 2ξηFξη

+ 2ξθFξθ + 2ηθFηθ + 2(1 − α)ξFξ + 2(1 − α)ηFη (19.230)
+ 2(1 − α)θFθ − α(1 − α)F = 0
Introducing a new similarity variable ρ by
ρ=θ−ξ−η (19.231)
(1 + ρ2 )G00 (ρ) + 2(1 − α)ρG0 (ρ) − α(1 − α)G(ρ) = 0 (19.232)

1 ρ
G(ρ) = C1 cosh arctanh p
2 1 + ρ2
1 (19.233)
ρ
+ C2 sinh arctanh p
2 1 + ρ2
C1 + C2 ρ
G(ρ) = (19.234)
1 + ρ2
ρ −1 + ρ2
G(ρ) = C1 + C 2 (19.235)
(1 + ρ2 )2 (1 + ρ2 )2
316
1 − 3ρ2 −3ρ + ρ3
G(ρ) = C1 + C 2 (19.236)
(1 + ρ2 )3 (1 + ρ2 )3
Using (19.229), (19.231) and (19.233)-(19.236) one can find the solutions of
the original 3−dimensional Wave equation (19.225) expressed in terms of the
variables x, y, z and t. We thus obtain the following solutions:

1/2
1 x(t − y − z)
u(x, y, z, t) = x C1 cosh arctanh p
2 x2 + (t − y − z)2
1 x(t − y − z) (19.237)
+ C2 sinh arctanh p
2 x2 + (t − y − z)2
C1 x2 + C2 x(t − y − z)
u(x, y, z, t) = (19.238)
x2 + (t − y − z)2
x3 (t − y − z) [x2 − (t − y − z)2 ]x2

u(x, y, z, t) = C1 + C 2 (19.239)
[x2 + (t − y − z)2 ]2 [x2 + (t − y − z)2 ]2
x4 [x2 − 3(t − y − z)2 ]

u(x, y, z, t) = C1
[x2 + (t − y − z)2 ]3
(19.240)
x3 (t − y − z)[−3x2 + (t − y − z)2 ]
+ C2
[x2 + (t − y − z)2 ]3
Chapter V
Non Classical Symmetries
317
20 Non Classical Symmetries
20.1. Theory. We first state the following
Theorem 1. The Invariant Surface Condition
T ut + Xux = U (20.1)
is invariant under the infinitesimal transformations
t̂ = t + ·T (t, x, u) + O(2 ) (20.2)
x̂ = x + ·X(t, x, u) + O(2 ) (20.3)
û = u + ·U (t, x, u) + O(2 ) (20.4)
The proof is left to the reader.
It is evident that if we want to consider the invariance of the system
T ut + Xux = U
(20.5)
F (t, x, u, ut , ux , utx , utt , uxx , . . .) = 0
we have to impose the two conditions
P r(1) Γ(∆1 )|∆1 =0,∆2 =0 = 0 (20.6)
P r(n) Γ(∆2 )|∆1 =0,∆2 =0 = 0 (20.7)

where
∆1 ≡ T ut + Xux − U, ∆2 ≡ F (t, x, u, ut , ux , utx , utt , uxx , . . .) (20.8)
However the first condition (20.6) is automatically satisfied, thanks to
Theorem 1. That means that considering invariance of the equation
taking into account the Invariant Surface Condition, it suffices to consider
the condition
P r(n) Γ(∆)|∆=0 = 0 (20.9)
where ∆ ≡ F (t, x, u, ut , ux , utx , utt , uxx , . . .). On the other hand, we have
Theorem 2. We have the following relation
P r(n) (kΓ) = k P r(n) Γ (20.10)
318
if the Invariant Surface Condition
T ut + Xux = U (20.11)
is being satisfied.
From Theorem 2 we see that we can take k any function we wish. Therefore,
without loss of generality, we can take k = 1/T and rename X and U such
that X/T →X and U/T →U . We thus can take T = 1.
Conclusion. Let us collect everything together: Invariance of the system
T ut + Xux = U
(20.12)
can be obtained by
P r(n) Γ(∆)|∆=0 = 0, T ut + Xux = U (20.13)
where ∆ ≡ F (t, x, u, ut , ux , utx , utt , uxx , . . .), taking into account T = 1.

Equations (20.13) constitute the so-called
”Non Classical Symmetry Method”.
20.2. Non Classical Symmetry Analysis of the Heat Equation.

We consider the heat equation
ut = uxx (20.14)
and the condition P r(2) Γ(∆) = 0, ∆≡ut − uxx which is given by (13.99):
− Uxx + Txx ut − (2Uxu − Xxx )ux + 2Txu ut ux
− (Uuu − 2Xxu )(ux )2 + Tuu ut (ux )2 + Xuu (ux )3 (20.15)
+ Tu ut uxx + 3Xu ux uxx = 0
The above equation becomes for T = 1:
Ut + ut Uu − ux (Xt + ut Xu )
− Uxx − (2Uxu − Xxx )ux
(20.16)
− (Uuu − 2Xxu )(ux )2 + Xuu (ux )3
− (Uu − 2Xx )uxx + 3Xu ux uxx = 0
319
We now have to take into account the condition ∆ = 0, and the Invariant
Surface Condition for T = 1, i.e.
ut = uxx and ut = U − Xux (20.17)
Ut + (U − Xux )Uu − ux (Xt + (U − Xux )Xu )

− Uxx − (2Uxu − Xxx )ux
(20.18)
− (Uuu − 2Xxu )(ux )2 + Xuu (ux )3
− (Uu − 2Xx )(U − Xux ) + 3Xu ux (U − Xux ) = 0
The previous equation can be written as (arranged in powers of ux )
Ut − Uxx + 2U Xx
+ (2U Xu − Xt − 2Uxu + Xxx − 2XXx )ux (20.19)
+ (2Xxu − Uuu − 2XXu )(ux )2 + Xuu (ux )3 = 0
Equating to zero the various coefficients of the above expression we obtain

the following system
Ut − Uxx + 2U Xx = 0 (20.20)
2U Xu − Xt − 2Uxu + Xxx − 2XXx = 0 (20.21)

2Xxu − Uuu − 2XXu = 0 (20.22)
Xuu = 0 (20.23)
We are going to solve the above system of equations (20.20)-(20.23).
X = α(t, x)u + β(t, x) (20.24)
Because of the above expression, equation (20.22) gives
Uuu = −2α2 u + 2(αx − αβ) (20.25)
Integrating once with respect to u, we get from the above equation
Uu = −α2 u2 + 2(αx − αβ)u + γ(t, x) (20.26)
320
and by one more integration
1
U = − α2 u3 + (αx − αβ)u2 + γ(t, x)u + δ(t, x) (20.27)
3
Upon substituting the expressions found previously for X and U into equa-
tion (20.21), we obtain
2
− a3 u3 + (4ααx − 2α2 β)u2
3
+ (−αt − 3αxx + 2αx β + 2αβx + 2αγ)u (20.28)
− βt + βxx − 2ββx − 2γx + 2αδ = 0
Equating to zero all the coefficients of u in the above equation, we derive the
system
α3 = 0 (20.29)
4ααx − 2α2 β = 0 (20.30)

−αt − 3αxx + 2αx β + 2αβx + 2αγ = 0 (20.31)
−βt + βxx − 2ββx − 2γx + 2αδ = 0 (20.32)
Since α = 0 from (20.29), equations (20.30) and (20.31) are satisfied identi-
cally while we get from (20.32) the equation
βt − βxx + 2ββx + 2γx = 0 (20.33)
For α = 0, we have the following expressions for X and U :
X = β(t, x) and U = γ(t, x)u + δ(t, x) (20.34)
(γt − γxx + 2γβx )u + (δt − δxx + 2δβx ) = 0 (20.35)
We thus obtain from the above equation the system
γt − γxx + 2γβx = 0 (20.36)
δt − δxx + 2δβx = 0 (20.37)
321
Conclusion. The functions T , X and U are given by
T = 1, X = β(t, x) and U = γ(t, x)u + δ(t, x) (20.38)
where β(t, x), γ(t, x) and δ(t, x) satisfy the system of partial differential equa-
tions
βt − βxx + 2ββx + 2γx = 0 (20.39)
γt − γxx + 2γβx = 0 (20.40)

δt − δxx + 2δβx = 0 (20.41)
The solution of the system (20.39)-(20.41) is given by
ωt ν − ωνt νt ωx − νx ωt
β(t, x) = − , γ(t, x) = (20.42)
ωx ν − ωνx ωx ν − ωνx
and
δ(t, x) = β(t, x)·(u0 )x + (u0 )t − β(t, x)·u0 (20.43)
where u0 is any solution of the heat equation and the functions ω, ν satisfy
ωxx = ωt and νxx = νt respectively.
The system of equations (20.39), (20.40) and (20.41) was set up for the first
time by Bluman and Cole and was solved thirty years later by Elizabeth
Mansfield:
G.W. Bluman and J.D. Cole: ”The general similarity solution of the
heat equation”, J. Math. Mech. 18 (1969), 1025-1042
E.L. Mansfield: ”The Nonclassical Group Analysis of the Heat Equation”,
J. Math. Analysis and Applications 231 (1999), 526-542.
Chapter VI
Fundamental Solutions through
Lie Symmetries
322
21 Fundamental Solutions through
Lie Symmetries
21.1. The group action. Let the equation
F (x, t, u, ut , ux , · · · ) = 0 (21.1)
admits the symmetry generator
∂ ∂ ∂
Γ = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u) (21.2)
∂x ∂t ∂u
The group action generated by the vector field Γ can be obtained by solving
the first order system
d
X() = ξ(X(), T (), U ()), X(0) = x
d
d
T () = τ (X(), T (), U ()), T (0) = t (21.3)
d
d
U () = φ(X(), T (), U ()), U (0) = u
d
Example. The differential equation
ut = xuxx + aux , a>0 (21.4)
admits the symmetry generator
∂ ∂ ∂
Γ = 8xt + 4t2 − 4(x + at)u (21.5)
∂x ∂t ∂u
Therefore in this case we have the following values of the infinitesimals ξ, τ
and φ:
ξ = 8xt, τ = 4t2 , φ = −4(x + at)u (21.6)
The system (21.3) then takes the form
d
X() = 8 X()·T (), X(0) = x
d
d
T () = 4 T 2 (), T (0) = t (21.7)
d
d
U () = −4 (X() + a·T ()) U (), U (0) = u
d
323
The above system can be solved explicitly. Solving the second equation, we
find
t
T () = − (21.8)
4t − 1
Using the above value of T (), the first equation of system (21.7) becomes
d 8t
X() = − X(), X(0) = x (21.9)
d 4t − 1
x
X() = (21.10)
(4t − 1)2
The third equation then becomes
d x at
U () = −4 − U (), U (0) = u (21.11)
d (4t − 1)2 4t − 1
with solution given by
1 −4x
U () = ×exp u (21.12)
(4t − 1)a 4t − 1
where u is any solution of the original equation (21.4). Therefore
1 4x x t
U (x, t) = ×exp ×u , − (21.13)
(4t − 1)a 4t − 1 (4t − 1)2 4t − 1
The function U (x, t) is a solution of (21.4) if u is a solution too. Setting
λ = −4 and u = 1, we see that
1 −λx
Uλ (x, t) = ×exp (21.14)
(1 + λt)a 1 + λt
is also a solution of (21.4).
21.2. The Fundamental Solution. The solution of the Cauchy problem
F (t, x, u, ut , ux , · · · ) = 0, u(x, 0) = f (x) (21.15)
is given by
Z ∞
u(x, t) = f (y)K(t, x, y)dy, u(x, 0) = f (x) (21.16)
0
324
where K(t, x, y) is the kernel function called the fundamental solution of
the differential equation for the Cauchy problem.
In general, if
Z ∞
Uλ (x, t) = e−λy K(t, x, y)dy (21.17)
0
then K(t, x, y) can be obtained by the inverse Laplace transform of Uλ (x, t),
where Uλ (x, t) can be calculated using the procedure above.
We have the following formula for the inverse Laplace transform (for future
reference)
y µ − 1 p
L−1 λ−µ ek/λ = k Iµ−1 (2 ky), µ>0 (21.18)
k
where Iν is the modified Bessel function of the first kind of order ν.
Chapter VII
The Direct Method
22 The Direct Method of
Clarkson and Kruskal
The Direct Method of solving PDEs was introduced by Clarkson and
Kruskal:
P.A. Clarkson and M.D. Kruskal: ”New Similarity Reductions of the

Boussinesq Equation”. J. Math. Phys. 30 (1989) 2201-2213.
It is an interesting method which has been applied to a number of equa-

tions. However it cannot be used as a substitute to Lie symmetry analysis.
The main ingredients of the method will become evident while solving an
equation. In this Section we shall apply the Direct Method to the Burgers
equation:
ut + uux = uxx (22.1)
325
We suppose that there exists a solution of the form
u(x, t) = A(x, t) + B(x, t)U (z), z = z(x, t) (22.2)
where z = z(x, t) is the similarity variable.

Since
d
ut = At + Bt U + Bzt U 0 , U0 = U (22.3)
dz
d
ux = Ax + Bx U + Bzx U 0 , U0 = U (22.4)
dz
and
0 0 2 00 00d2
uxx = Axx + Bxx U + 2Bx zz U + Bzxx U + B(zx ) U , U = 2 U (22.5)
dz
the Burgers equation (22.1) becomes
At + Bt U + Bzt U 0 + (A + BU )(Ax + Bx U + Bzx U 0 )

− Axx + Bxx U + 2Bx zz U 0 + Bzxx U 0 + B(zx )2 U 00 = 0
B(zx )2 U 00 + [2zx Bx + Bzxx − B(zt + Azx )]U 0 − B 2 zx U U 0

(22.6)
− BBx U 2 + (Bxx − Bt − BAx − ABx )U + Axx − At − AAx = 0
We want to reduce the above equation into an ordinary differential equation

in U. For this, we introduce a set of functions Γn (z), n = 1, 2, . . ., 5 such that
(22.6) becomes
U 00 + Γ1 (z)U 0 + Γ2 (z)U U 0 + Γ3 (z)U 2 + Γ4 (z)U + Γ5 (z) = 0 (22.7)
Upon comparing (22.7) to (22.6), we obtain the system of equations
2zx Bx + Bzxx − B(zt + Azx ) = B(zx )2 ·Γ1 (z) (22.8)
−B 2 zx = B(zx )2 ·Γ2 (z) (22.9)

−BBx = B(zx )2 ·Γ3 (z) (22.10)
Bxx − Bt − BAx − ABx = B(zx )2 ·Γ4 (z) (22.11)
326
Axx − At − AAx = B(zx )2 ·Γ5 (z) (22.12)
In order to solve the system of equations (22.8)-(22.12), we apply the follow-
ing rules:
Rule 1. If A(x, t) has the form
A(x, t) = Ã(x, t) + B(x, t)Γ(z) (22.13)
we simply put Γ(z) = 0.

Rule 2. If B(x, t) is found to have the form
B(x, t) = B̃(x, t)Γ(z) (22.14)
we may put Γ(z) = 1.

Rule 3. If z(x, t) is determined from the implicit function
f (z) = z̃(x, t) (22.15)
where f (z) is an invertible function, then we may simply put f (z) = z.

Assume now that
zxx = 0 (22.16)
which means that z(x, t) is a linear function of x:
z = α(t)x + β(t) (22.17)
Equation (22.10) takes on the form (dividing by B)
Bx = −(zx )2 Γ3 (z) (22.18)
and for
Γ3 (z) = ∆03 (z) (22.19)
(the prime (0 ) denotes differentiation with respect to z) we get

∂
Bx = −(zx )2 Γ3 (z) = −(zx )2 ·∆03 (z) = −zx · ∆3 (z) (22.20)
∂x
and by integration
B(x, t) = −zx ·∆3 (z) (22.21)
327
taking into account that zx is independent of x, due to the assumed linearity
in z. From equation (22.21), according to Rule 2, we should have ∆3 (z) = 1
and then Γ3 (z) = 0 and
B(x, t) = −zx (22.22)
Since Bx (x, t) = −zxx = 0, we conclude that
B = B(t) (22.23)
From (22.9), we obtain on dividing by Bzx
B = −zx Γ2 (z) (22.24)
and according to Rule 2, Γ2 (z) = 1 and then zx = −B(t) and by integration
z = −x·B(t) + C(t) (22.25)
From (22.11), taking into account that Bx = 0, Bxx = 0 we have
−Bt − BAx = B(zx )2 ·Γ4 (z)

B 0 (t)
Ax = − − B 2 Γ4 (z) (22.26)
B(t)
According to Rule 1, we obtain from the above equation Γ4 (z) = 0 and thus
B 0 (t)
Ax = − (22.27)
B(t)
B 0 (t)
A(x, t) = − x + E(t) (22.28)
B(t)
Equation (22.12), using
B0
00 0
B B 2
Axx = 0, Ax = − , At = − − x + E 0 (t), zx = −B
B B B
328
takes on the form
00 0
B 0 (t) B0

B B 2 0
− x − E (t) − − x + E(t) −
B B B(t) B
2
= B·B Γ5 (z)

B 00 B 0 2 B0
x − 2x +E − E 0 = B 3 Γ5 (z) (22.29)
B B B
(prime denotes differentiation with respect to t). The above equation is linear
in x and since z is also linear in x (see equation (22.17)) we may choose
Γ5 (z) = −k 2 z with k constant and z given by (22.25). We thus obtain
B 00 B 0 2 B0
x − 2x +E − E 0 = −k 2 B 3 (−xB + C) (22.30)
B B B
Equating the coefficients of x and the constant terms, we get the system of
two equations
B 00 B 0 2
−2 = k2B 4 (22.31)
B B
B0
E − E 0 = −k 2 B 3 C (22.32)
B
Taking C = 0 for simplicity, the above system is equivalent to
BB 00 − 2(B 0 )2 − k 2 B 6 = 0, EB 0 − E 0 B = 0 (22.33)
Finally, equation (22.8) since (remember C = 0)
Bx = 0, zxx = 0, zt = −xB 0 , zx = −B (22.34)
with A given by (22.28), becomes

B0
0
−B(−xB ) − B − x + E (−B) = B·(−B)2 Γ1 (z) (22.35)
B
E = BΓ1 (z) (22.36)
329
The above equation is compatible to the second equation of the system
(22.33) if Γ1 (z) = m with m constant. Therefore
E = m·B(t) (22.37)
Since we have found all the Γ0 s, i.e.
Γ1 (z) = m, Γ2 (z) = 1, Γ3 (z) = 0, Γ4 (z) = 0, Γ5 (z) = −k 2 z (22.38)
equation (22.7) becomes (the prime (0 ) denotes differentiation with respect

to z)
U 00 + mU 0 + U U 0 − k 2 z = 0 (22.39)
Let us collect what we have found so far:

The solution of Burgers equation (22.1) is given by
u(x, t) = A(x, t) + B(t)U (z), z≡z(x, t) = −x·B(t) (22.40)
where
B 0 (t)
A(x, t) = − x + E(t), E(t) = m·B(t) (22.41)
B(t)
The function B(t) satisfies the first equation of the system (22.33)
BB 00 − 2(B 0 )2 − k 2 B 6 = 0 (22.42)
and U (z) is the solution of equation (22.39).

We first turn into (22.42). Under the substitution
1
B(t) = (22.43)
v(t)a
a·v −2a−1 v 00 + (a2 − a)v −2a−2 (v 0 )2 + k 2 ·v −6a = 0 (22.44)
The choice a = 1 eliminates the (v 0 )2 term and we thus obtain the equation
v 00 = −k 2 ·v −3 (22.45)
330
The above equation is an Ermakov type of equation admitting general so-
lution given by
K1 v 2 = −k 2 t2 + (K2 t − K1 )2 (22.46)
Therefore we get, using the above solution and (22.43) (for a = 1) that the
solutions of (22.42) are given by
22 −1/2
−k t + (K2 t − K1 )2
B(t) = ± (22.47)
K1
The solution of equation (22.45) was based on the following
Lemma. If w(x) is a nontrivial solution of the equation
y 00 + f (x)y = 0 (22.48)
then the general solution of the equation
y 00 + f (x)y = ay −3 (22.49)
is given by
Z
2

2 2 dx 2
K1 y = aw + w K2 + K1 (22.50)
w2
Let us now turn into equation (22.39). This equation can be written as
1 k2
(U 0 )0 + (mU )0 + (U 2 )0 − (z 2 )0 = 0 (22.51)
2 2
which can thus be integrated giving (taking the integration constant equal
to zero)
1 k2
U 0 + mU + U 2 − z 2 = 0 (22.52)
2 2
This is a Riccati differential equation which under the standard substitution
Q0 (z)
U =2 (22.53)
Q(z)
takes on the form
1
Q00 + mQ0 − k 2 z 2 Q = 0 (22.54)
4
331
The above is a hypergeometric equation admitting the general solution
m2 + 6k 3 1
−z(kz+2m)/4
Q(z) = ze C1 ·M , ; kz 2
8k 2 2
m2 + 6k 3 1 (22.55)
2
+ C2 ·U , ; kz
8k 2 2
where M and U are the known Kummer’s functions.

Using the above solution we find, using (22.53) that
N (k, m; z)
U (z) = (22.56)
D(k, m; z)
where
m2 + 6k 3 1
N (k, m; z) = C1 (2k 2 z 2 − 2kmz − 2k + m2 )·M , ; kz 2
8k 2 2
m2 − 2k 3 1
+ (6k − m2 )·M , ; kz 2
8k 2 2
m2 + 6k 3 1 (22.57)
2 2 2 2
+ C2 (2k z − 2kmz − 2k + m )·U , ; kz
8k 2 2
m2 − 2k 3 1
− 8k·U , ; kz 2
8k 2 2
and
m2 + 6k 3 1
2
D(k, m; z) = 2kz C1 ·M , ; kz
8k 2 2
m2 + 6k 3 1 (22.58)
2
+ C2 ·U , ; kz
8k 2 2
The function U (z) given in (22.56) should not be confused with Kummer’s
U −function appearing in (22.55), (22.57) and (22.58).
We thus have obtained the following Theorem
Theorem. The general solution of Burgers equation ut + uux = uxx is given
by
u(x, t) = A(x, t) + B(t)U (z) (22.59)
332
where
B 0 (t)
A(x, t) = − x + m·B(t) (22.60)
B(t)
The function B(t) is given by

22 −1/2
−k t + (K2 t − K1 )2
B(t) = ± (22.61)
K1
while U (z) is given by (22.56)-(22.58) with z = −xB(t).
Chapter VIII
Systems of Differential
Equations.
23 First Order Systems of
Lie Symmetry Analysis can also be applied to Systems of Differential Equa-
tions. In this Section we shall apply the method to systems of ODEs of the
first order.
We consider a system of two first order ODEs
ẋ = F (t, x, y), ẏ = G(t, x, y) (23.1)
Considering the operator

∂ ∂ ∂
Γ=T +X +Y (23.2)
∂t ∂x ∂y
and its first extension
∂ ∂
P r(1) Γ = Γ + X [t] + Y [t] (23.3)
∂ ẋ ∂ ẏ
333
the invariance condition of the system is being expressed by
P r(1) Γ(∆1 )|∆1 =0, ∆2 =0 = 0, P r(1) Γ(∆2 )|∆1 =0, ∆2 =0 = 0 (23.4)
where
∆1 = ẋ − F (t, x, y), ∆2 = ẏ − G(t, x, y) (23.5)
The extended infinitesimals are defined by
X [t] = Dt (X) − ẋDt (T ) (23.6)
Y [t] = Dt (Y ) − ẏDt (T ) (23.7)

where the total differential operator Dt is given by
∂ ∂ ∂
Dt = + ẋ + ẏ (23.8)
∂t ∂x ∂y
We have the following explicit expressions
X [t] = Xt + ẋXx + ẏXy − ẋ(Tt + ẋTx + ẏTy )

(23.9)
= Xt + (Xx − Tt )ẋ − Tx (ẋ)2 − Ty ẋẏ + Xy ẏ
Y [t] = Yt + ẋYx + ẏYy − ẏ(Tt + ẋTx + ẏTy )

(23.10)
= Yt + (Yy − Tt )ẏ − Ty (ẏ)2 − Tx ẋẏ + Yx ẋ
Example. Solve the system
ẋ = 2xy, ẏ = x2 + y 2 (23.11)
Solution. We set
∆1 = ẋ − 2xy and ∆2 = ẏ − x2 − y 2 (23.12)
We then have
P r(1) Γ(∆1 ) = 0
∂ ∂ ∂ [t] ∂ [t] ∂

⇔ T +X +Y +X +Y (ẋ − 2xy) = 0
∂t ∂x ∂y ∂ ẋ ∂ ẏ
⇔ X(−2y) + Y (−2x) + X [t] = 0
334
and substitution of X [t] by (23.9), we obtain
−2yX − 2xY + Xt + (Xx − Tt )ẋ − Tx (ẋ)2 − Ty ẋẏ + Xy ẏ = 0
In the above expression we substitute ẋ by 2xy and ẏ by x2 + y 2 , to account

for the conditions ∆1 = 0 and ∆2 = 0. We thus obtain
− 2yX − 2xY + Xt + (Xx − Tt )(2xy) − Tx (2xy)2

(23.13)
− Ty (2xy)(x2 + y 2 ) + Xy (x2 + y 2 ) = 0
We have similarly
P r(1) Γ(∆2 ) = 0
∂ ∂ ∂ ∂ ∂
⇔ T +X +Y + X [t] + Y [t] (ẏ − x2 − y 2 ) = 0
∂t ∂x ∂y ∂ ẋ ∂ ẏ
⇔ X(−2x) + Y (−2y) + Y [t] = 0
and substitution of Y [t] by (23.10), we obtain
−2xX − 2yY + Yt + (Yy − Tt )ẏ − Ty (ẏ)2 − Tx ẋẏ + Yx ẋ = 0
In the above expression we substitute ẋ by 2xy and ẏ by x2 + y 2 , to account

for the conditions ∆1 = 0 and ∆2 = 0. We thus obtain
− 2xX − 2yY + Yt + (Yy − Tt )(x2 + y 2 ) − Ty (x2 + y 2 )2

(23.14)
− Tx (2xy)(x2 + y 2 ) + Yx (2xy) = 0
We now have to solve the systems of equations (23.13) and (23.14). In order
to simplify calculations, we assume that
T = T (t), X = X(x), Y = Y (y) (23.15)
Under this assumption, (23.13) and (23.14) are being converted into
−yX − xY + xy(Xx − Tt ) = 0 (23.16)
and
−2xX − 2yY + (x2 + y 2 )(Yy − Tt ) = 0 (23.17)
335
respectively.
Taking the derivative of (23.16) with respect to t we obtain xy(−Ttt ) = 0
which means Ttt = 0 and thus
T (t) = at + b (23.18)
Because of the above expression, equations (23.16) and (23.17) take on the
form
−yX − xY + xy(Xx − a) = 0 (23.19)
and
−2xX − 2yY + (x2 + y 2 )(Yy − a) = 0 (23.20)
respectively. Differentiation of (23.19) twice with respect to y yields −yYyy =

0, i.e. Yyy = 0 and thus
Y = cy + δ (23.21)
Upon inserting Y given by (23.21) into (23.19) we get
−yX − x(cy + δ) + xy(Xx − a) = 0
(−X − cx + xXx − ax)y − xδ = 0
from which we get (since it holds for every x and y)
−X − cx + xXx − ax = 0 (23.22)
and
δ=0 (23.23)
We thus have that
Y =cδ (23.24)
Equation (23.20) then gives
−2xX − 2cy 2 + (x2 + y 2 )(c − a) = 0
336
which can also be written as (arranged in powers of x and y)
−(c + a)y 2 + (c − a)x2 − 2xX = 0
from which we obtain (since X is independent of y)
c + a = 0 and (c − a)x2 − 2xX = 0 (23.25)
We thus see that
c = −a (23.26)
From the second of (23.25) we obtain
X = −ax (23.27)
Conclusion: We have obtained
T = at + b, X = −ax, Y = −ay (23.28)
Considering for simplicity a = 1 and b = 0, the invariant surface conditions

have the form
trt − xrx − yry = 0, tut − xux − yuy = 0, tvt − xvx − yvy = 1 (23.29)
The solution of the above system is given by
r = R(tx, ty), u = U (tx, ty), v = ln(t) + V (tx, ty) (23.30)

y
r= , u = tx, v = − ln(x) (23.31)
x
i.e.
x = e−v , y = r e−v , t = u ev (23.32)
Under this change of variables, the original system takes on the form of two
decoupled equations
1 + 2ru 2r
ur = , vr = − (23.33)
1 − r2 1 − r2
337
The solution of the previous system is given by
−r + C1
u= , v = ln |r2 − 1| + C2 (23.34)
r2 − 1
y
− + C1 y 2
tx = x2 , − ln(x) = ln − 1 + C2 (23.35)
y x
−1
x
Solving with respect to x and y we get
α −α
x= , y= (23.36)
(αx + β)2 − 1 (αx + β)2 − 1
where we have set
C1 = −β and C2 = − ln(α) (23.37)
24 Second Order Systems of

We consider a system of two second order ODEs
ẍ = F (t, x, y, ẋ, ẏ), ÿ = G(t, x, y, ẋ, ẏ) (24.1)
Considering the operator

∂ ∂ ∂
Γ=T +X +Y (24.2)
∂t ∂x ∂y
and its second extension
∂ ∂ ∂ ∂
P r(2) Γ = Γ + X [t] + Y [t] + X [tt] + Y [tt] (24.3)
∂ ẋ ∂ ẏ ∂ ẍ ∂ ÿ
the invariance condition of the system is being expressed by
P r(2) Γ(∆1 )|∆1 =0, ∆2 =0 = 0, P r(2) Γ(∆2 )|∆1 =0, ∆2 =0 = 0 (24.4)
338
where
∆1 = ẍ − F (t, x, y, ẋ, ẏ), ∆2 = ÿ − G(t, x, y, ẋ, ẏ) (24.5)
The extended infinitesimals are defined by
X [tt] = Dt (X [t] ) − ẍDt (T ) (24.6)
Y [tt] = Dt (Y [t] ) − ÿDt (T ) (24.7)
where the total differential operator Dt is given by
∂ ∂ ∂ ∂ ∂
Dt = + ẋ + ẏ + ẍ + ÿ (24.8)
∂t ∂x ∂y ∂ ẋ ∂ ẏ
We can calculate the explicit expressions of the extended infinitesimals X [tt]
and Y [tt] . We have for the X [tt] :
X [tt] = Dt (X [t] ) − ẍDt (T )
∂ ∂ ∂ ∂ ∂
= + ẋ + ẏ + ẍ + ÿ Xt + ẋXx + ẏXy
∂t ∂x ∂y ∂ ẋ ∂ ẏ
∂ ∂ ∂
− ẋ(Tt + ẋTx + ẏTy ) − ẍ + ẋ + ẏ T
∂t ∂x ∂y
= Xtt + ẋXtx + ẏXty − ẋ(Ttt + ẋTtx + ẏTty ) (24.9)
+ ẋ(Xtx + ẋXxx + ẏXxy ) − (ẋ)2 (Ttx + ẋTxx + ẏTxy )
+ ẏ(Xty + ẋXxy + ẏXyy ) − ẋẏ(Tty + ẋTxy + ẏTyy )
+ ẍ(Xx − (Tt + ẋTx + ẏTy ) − ẋTx )
+ ÿ(Xy − ẋTy ) − ẍ(Tt + ẋTx + ẏTy )
and after collecting all terms together and rearranging, we obtain
X [tt] = Xtt + (2Xtx − Ttt )ẋ + (Xxx − 2Ttx )(ẋ)2 − Txx (ẋ)3
+ (Xx − 2Tt )ẍ − 3Tx ẋẍ − 2Ty ẍẏ + 2Xty ẏ
(24.10)
+ 2(Xxy − Tty )ẋẏ − 2Txy (ẋ)2 ẏ + Xy ÿ − Ty ẋÿ
+ Xyy (ẏ)2 − Tyy ẋ(ẏ)2
We find in a similar way that (simply interchanging X↔Y and x↔y)
Y [tt] = Ytt + (2Yty − Ttt )ẏ + (Yyy − 2Tty )(ẏ)2 − Tyy (ẏ)3
+ (Yy − 2Tt )ÿ − 3Ty ẏ ÿ − 2Tx ÿ ẋ + 2Ytx ẋ
(24.11)
+ 2(Yxy − Ttx )ẋẏ − 2Txy (ẏ)2 ẋ + Yx ẍ − Tx ẏẍ
+ Yxx (ẋ)2 − Txx (ẋ)2 ẏ
339
Notice that we have put together terms containing ẍ and ÿ. This arrangement
will simplify subsequent calculations when we have to take into account the
conditions ∆1 = 0 and ∆2 = 0.
ẍ = P (x, y), ÿ = Q(x, y) (24.12)
where Py 6=0 and Qx 6=0.

Solution. We set
∆1 = ẍ − P (x, y) and ∆2 = ÿ − Q(x, y) (24.13)
We then have
P r(2) Γ(∆1 ) = 0
∂ ∂ ∂ ∂ ∂
⇔ T +X +Y + X [t] + Y [t]
∂t ∂x ∂y ∂ ẋ ∂ ẏ
∂ ∂
+ X [tt] + Y [tt] (ẍ − P (x, y)) = 0
∂ ẍ ∂ ÿ
⇔X(−Px ) + Y (−Py ) + X [tt] = 0
After substituting X [tt] given in (24.10) into the above equation, we obtain
− XPx − Y Py + Xtt + (2Xtx − Ttt )ẋ + (Xxx − 2Ttx )(ẋ)2

− Txx (ẋ)3 + (Xx − 2Tt )ẍ − 3Tx ẋẍ − 2Ty ẍẏ + 2Xty ẏ
+ 2(Xxy − Tty )ẋẏ − 2Txy (ẋ)2 ẏ + Xy ÿ − Ty ẋÿ
+ Xyy (ẏ)2 − Tyy ẋ(ẏ)2 = 0
In the above equation we substitute ẍ by P (x, y) and ÿ by Q(x, y) to take

into account the conditions ∆1 = 0 and ∆2 = 0. We thus obtain the equation
− XPx − Y Py + Xtt + (2Xtx − Ttt )ẋ + (Xxx − 2Ttx )(ẋ)2

− Txx (ẋ)3 + (Xx − 2Tt )P − 3Tx P ẋ − 2Ty P ẏ + 2Xty ẏ
(24.14)
+ 2(Xxy − Tty )ẋẏ − 2Txy (ẋ)2 ẏ + Xy Q − Ty Qẋ
+ Xyy (ẏ)2 − Tyy ẋ(ẏ)2 = 0
340
We have similarly
P r(2) Γ(∆2 ) = 0
∂ ∂ ∂ ∂ ∂
⇔ T +X +Y + X [t] + Y [t]
∂t ∂x ∂y ∂ ẋ ∂ ẏ
∂ ∂
+ X [tt] + Y [tt] (ÿ − Q(x, y)) = 0
∂ ẍ ∂ ÿ
⇔X(−Qx ) + Y (−Qy ) + Y [tt] = 0
After substituting Y [tt] given in (24.11) into the above equation, we obtain
− XQx − Y Qy + Ytt + (2Yty − Ttt )ẏ + (Yyy − 2Tty )(ẏ)2

− Tyy (ẏ)3 + (Yy − 2Tt )ÿ − 3Ty ẏ ÿ − 2Tx ÿ ẋ + 2Ytx ẋ
+ 2(Yxy − Ttx )ẋẏ − 2Txy (ẏ)2 ẋ + Yx ẍ − Tx ẏẍ
+ Yxx (ẋ)2 − Txx (ẋ)2 ẏ = 0
In the above equation we substitute ẍ by P (x, y) and ÿ by Q(x, y) to take

into account the conditions ∆1 = 0 and ∆2 = 0. We thus obtain the equation
− XQx − Y Qy + Ytt + (2Yty − Ttt )ẏ + (Yyy − 2Tty )(ẏ)2

− Tyy (ẏ)3 + (Yy − 2Tt )Q − 3Ty Qẏ − 2Tx Qẋ + 2Ytx ẋ
(24.15)
+ 2(Yxy − Ttx )ẋẏ − 2Txy (ẏ)2 ẋ + Yx P − Tx P ẏ
+ Yxx (ẋ)2 − Txx (ẋ)2 ẏ = 0
We are going to solve the system of equations (24.14) and (24.15). We equate
to zero all the coefficients of the derivatives of x and y.
From equation (24.14) we obtain:
−XPx − Y Py + Xtt + (Xx − 2Tt )P + Xy Q = 0 (24.16)
The coefficient of ẋ is
(2Xtx − Ttt ) − 3Tx P − Ty Q = 0 (24.17)
The coefficient of ẏ is
−2Ty P + 2Xty = 0 (24.18)
341
The coefficient of ẋẏ is
2(Xxy − Tty ) = 0 (24.19)
The coefficient of (ẋ)2 is
Xxx − 2Ttx = 0 (24.20)
The coefficient of (ẏ)2 is
Xyy = 0 (24.21)
The coefficient of ẋ(ẏ)2 is
−Tyy = 0 (24.22)
The coefficient of (ẋ)2 ẏ is
−2Txy = 0 (24.23)
−Txx = 0 (24.24)
From equation (24.15) we obtain:

−XQx − Y Qy + Ytt + (Yy − 2Tt )Q + Yx P = 0 (24.25)
−2Tx Q + 2Ytx = 0 (24.26)
The coefficient of ẏ is
(2Yty − Ttt ) − 3Ty Q − Tx P = 0 (24.27)
The coefficient of ẋẏ is
2(Yxy − Ttx ) = 0 (24.28)
Yxx = 0 (24.29)
342
Yyy − 2Tty = 0 (24.30)
The coefficient of ẋ(ẏ)2 is
−2Txy = 0 (24.31)
The coefficient of (ẋ)2 ẏ is
−Txx = 0 (24.32)
−Tyy = 0 (24.33)
Let us now try to solve the system of equations (24.16)-(24.33). Despite the
fact that we don’t know the explicit expressions of P and Q, we can gain a
lot of information about the above system of equations.
First of all, differentiation of (24.18) and (24.26) with respect to y and x
respectively, we obtain the equations
−Tyy P − Ty Py + Xtyy = 0 and − Txx Q − Tx Qx + Ytxx = 0 (24.34)
Taking into account that Xyy = 0 from (24.21) and Yxx = 0 from (24.29)
and using also Tyy = 0, Txx = 0 (which appear in (24.22), (24.24) or (24.32),
(24.33)) we obtain from the above two equations (24.34) that
Ty Py = 0 and Tx Qx = 0
and thus Ty = 0 and Tx = 0 which means that
T = T (t) (24.35)
From (24.18) for Ty = 0 we obtain Xty = 0. From (24.20) for Ttx = 0 we get
Xxx = 0. From (24.21) we have Xyy = 0. Therefore since
Xty = 0, Xxx = 0 and Xyy = 0 (24.36)
we have that
X = α(t)x + βy + γ(t) (24.37)
343
From (24.26) for Tx = 0 we obtain Ytx = 0. From (24.29) we have Yxx = 0.
From (24.30) since Tty = 0 we get Yyy = 0. Therefore, since
Ytx = 0, Yxx = 0 end Yyy = 0 (24.38)
we get
Y = δx + (t)y + ζ(t) (24.39)
From equations (24.17) and (24.27), using the above found expressions for
T , X and Y we obtain
2α0 (t) − T 00 (t) = 0 and 20 (t) − T 00 (t) = 0
Integrating the above two equations we obtain

1 1
α(t) = T 0 (t) + a1 and (t) = T 0 (t) + a2 (24.40)
2 2
We thus obtain the following expressions of X and Y :
1
X= T 0 (t) + a1 x + βy + γ(t) (24.41)
2
1
Y = δx + T 0 (t) + a2 y + ζ(t) (24.42)
2
We are then left with the constant terms, i.e. equations (24.16) and (24.25)
respectively. These two equations can be integrated once we know the explicit
expressions of P (x, y) and Q(x, y).
x y
ẍ = , ÿ = (24.43)
(x2 + y 2 )2 (x2 + y 2 )2
Solution. In this case we recognize that
x y
P (x, y) = , Q(x, y) = (24.44)
(x2 + y 2 )2 (x2 + y 2 )2
We are now going to use equations (24.16) and (24.25) in order to evaluate
the explicit expressions of the infinitesimals T , X and Y .
Since
1 4x2 4xy
Px = − and Py = − (24.45)
(x2 + y 2 )2 (x2 + y 2 )3 (x2 + y 2 )3
344
4xy 1 4y 2
Qx = − and Qy = − (24.46)
(x2 + y 2 )3 (x2 + y 2 )2 (x2 + y 2 )3
equation (24.16) takes on the form (taking into account (24.41) and (24.42)
as well)

2 2 1 0
(3x − y ) T (t) + a1 x + βy + γ(t)
2
1
0
+ 4xy δx + T (t) + a2 y + ζ(t)
2
1 1
+ (x2 + y 2 )3 T 000 (t)x + γ 00 (t)
2 2
2 2 1 0
0
+ x(x + y ) T (t) + a1 − 2T (t)
2
+ βy(x2 + y 2 ) = 0
Upon arranging the above equation in powers of x and y and equating to

zero the coefficients, we obtain
a1 = 0, a2 = 0, δ + β = 0, γ(t) = 0, ζ(t) = 0, T 000 (t) = 0 (24.47)
We obtain the same relations using equation (24.25).

From T 000 (t) = 0 we get
T (t) = c3 t2 + 2c2 t + c1 (24.48)
and thus we have from (24.41) and (24.42)
X = (c3 t + c2 )x + c4 y, Y = −c4 x + (c3 t + c2 )y (24.49)
making the identifications β−→c4 and δ = −β−→ − c4 .

We thus obtain the following expression for the symmetry generator Γ:
∂ ∂ ∂
Γ=T +X +Y
∂t ∂x ∂y

2 ∂ ∂
= (c3 t + 2c2 t + c1 ) + (c3 t + c2 )x + c4 y
∂t ∂x

∂
+ − c4 x + (c3 t + c2 )y
∂y
345

∂ ∂ ∂ ∂ 2 ∂ ∂ ∂
Γ = c1 + c2 2t + x +y + c3 t + tx + ty
∂t ∂t ∂x ∂y ∂t ∂x ∂y
(24.50)
∂ ∂
+ c4 y −x
∂x ∂y
We thus see that the Lie algebra of symmetries of the original system is being
spanned by the generators
∂ ∂ ∂ ∂
Γ(1) = , Γ(2) = 2t + x +y
∂t ∂t ∂x ∂y
(24.51)
(3) ∂ ∂ ∂ ∂ ∂
Γ = t2 + tx + ty , Γ(4) = y −x
∂t ∂x ∂y ∂x ∂y
The invariant surface conditions corresponding to Γ(4) (the reader can easily
recognize that this is the rotation operator), are given by
yrx − xry = 0, yux − xuy = 0, yvx − xvy = 1 (24.52)
Solving the above system, we obtain
r = R(t, x2 + y 2 ), u = U (t, x2 + y 2 ),
y
v = arctan + V (t, x2 + y 2 )
x
We make the choice
p y
r = t, u = x2 + y 2 , v = arctan (24.53)
x
Under the above substitutions, the original system transforms into
u4 vr2 + 1 2ur vr
urr = , vrr = − (24.54)
u3 u
package called MathLie in solving second order system of equations. The
MathLie is an AddOn package to Mathematica (Reference [4]). We use the
following steps:
In[1]:=<<MathLie.m
346
Step 2. We enter the second order system of ODEs using the command

X[t] Y [t]
In[2] := sysEqns = ∂t,t X[t] == , ∂t,t Y [t] == ;
(X[t]2 + Y [t]2 )2 (X[t]2 + Y [t]2 )2
sysEqns//LT F
X
− + Xt,t == 0
(X + Y 2 )2
2
Y
− + Yt,t == 0
(X + Y 2 )2
2
Step 3. We can find the infinitesimals using a single command
In[3] := infEqns = Infinitesimals[sysEqns, {X, Y }, {t},

SubsitutionRules→{∂t,t X[t], ∂t,t Y [t]}];
infEqns//LT F
k3 X
φ1 == + k4 t X + k1 Y
2
k3 Y
φ2 == −k1 X + + k4 t Y
2
ξ1 == k2 + t(k3 + k4 t)
25 Systems of Partial Differential Equations

of the First Order
We consider a system of PDEs of the first order
F (t, x, u, v, ut , vt , ux , vx ) = 0 (25.1)
G(t, x, u, v, ut , vt , ux , vx ) = 0 (25.2)
347
Invariance of the system is being expressed by the conditions
P r(1) Γ(F )|F =0, G=0 = 0, P r(1) Γ(G)|F =0, G=0 = 0 (25.3)

∂ ∂ ∂ ∂
Γ=T +X +U +V (25.4)
∂t ∂x ∂u ∂v
and P r(1) Γ is the extended operator of the first order, defined by
∂ ∂ ∂ ∂
P r(1) Γ = Γ + U [t] + U [x] + V [t] + V [x] (25.5)
∂ut ∂ux ∂vt ∂vx

V [t] = Dt (V ) − vt Dt (T ) − vx Dt (X) (25.8)
V [x] = Dx (V ) − vt Dx (T ) − vx Dx (X) (25.9)
The total differential operators Dt and Dx are defined by
∂ ∂ ∂ ∂
Dt = + ut + utt + utx + ···+
(25.10)
∂ ∂ ∂
+ vt + vtt + vtx + ···
∂v ∂vt ∂vx
∂ ∂ ∂ ∂
Dx = + ux + utx + uxx + ···+
(25.11)
∂ ∂ ∂
+ vx + vtx + vxx + ···
∂v ∂vt ∂vx
After determining the infinitesimals T , X, U and V , we have to consider the
invariant surface conditions
T ut + Xux = U, T vt + Xvx = V (25.12)
348
in order to find the unknown functions u and v.
We shall calculate explicitly the coefficients U [t] , U [x] and V [t] , V [x] . We have
U [t] = Dt (U ) − ut Dt (T ) − ux Dt (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ut + vt U − ut + ut + vt T
∂t ∂u ∂v ∂t ∂u ∂v
∂ ∂ ∂ (25.13)
− ux + ut + vt X
∂t ∂u ∂v
= Ut + (Uu − Tt )ut − (ut )2 Tu + vt Uv − ut vt Tv
− ux Xt − ux ut Xu − ux vt Xv
U [x] = Dx (U ) − ut Dx (T ) − ux Dx (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ux + vx U − ut + ux + vx T
∂x ∂u ∂v ∂x ∂u ∂v
∂ ∂ ∂ (25.14)
− ux + ux + vx X
∂x ∂u ∂v
= Ux + (Uu − Xx )ux − (ux )2 Xu + vx Uv − ut Tx
− ux ut Tu − ut vx Tv − ux vx Xv
V [t] = Dt (V ) − vt Dt (T ) − vx Dt (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ut + vt V − vt + ut + vt T
∂t ∂u ∂v ∂t ∂u ∂v
∂ ∂ ∂ (25.15)
− vx + ut + vt X
∂t ∂u ∂v
= Vt + (Vv − Tt )vt − (vt )2 Tv + ut Vu − vx Xt
− vt ut Tu − vx ut Xu − vt vx Xv
V [x] = Dx (V ) − vt Dx (T ) − vx Dx (X)
∂ ∂ ∂ ∂ ∂ ∂
= + ux + vx V − vt + ux + vx T
∂x ∂u ∂v ∂x ∂u ∂v
∂ ∂ ∂ (25.16)
− vx + ux + vx X
∂x ∂u ∂v
= Vx + (Vv − Xx )vx − (vx )2 Xv + ux Vu − vt Tx
− ux vt Tu − vt vx Tv − vx ux Xu
ux
vx = u, vt = (25.17)
1 + u2
349
Solution. We denote by ∆1 and ∆2 the quantities
ux
∆1 = vx − u and ∆2 = vt − (25.18)
1 + u2
Invariance of the first equation is being expressed by (P r(1) Γ is given by
(25.5))
P r(1) Γ(∆1 ) = 0 ⇔ V [x] − U = 0 (25.19)
Substituting V [x] by the expression given in (25.16), the previous equation

becomes
Vx + (Vv − Xx )vx − (vx )2 Xv + ux Vu − vt Tx
(25.20)
− ux vt Tu − vt vx Tv − vx ux Xu − U = 0
In the above equation we substitute vx by u and ux by (1 + u2 )vt to account

for the conditions ∆1 = 0 and ∆2 = 0. We thus see that the condition
P r(1) Γ(∆1 )|∆1 =0, ∆2 =0 = 0 is equivalent to the equation
Vx + (Vv − Xx )u − u2 Xv + (1 + u2 )vt Vu − vt Tx
(25.21)
− (1 + u2 )(vt )2 Tu − uvt Tv − u(1 + u2 )vt Xu − U = 0

Vx + (Vv − Xx )u − u2 Xv − U
+ [(1 + u2 )Vu − Tx − uTv − u(1 + u2 )Xu ]vt (25.22)
− (1 + u2 )Tu (vt )2 = 0
Equating to zero the coefficients of the derivatives of the functions u and v,

we obtain the system
Vx + (Vv − Xx )u − u2 Xv − U = 0 (25.23)
(1 + u2 )Vu − Tx − uTv − u(1 + u2 )Xu = 0 (25.24)

−(1 + u2 )Tu = 0 (25.25)
Invariance of the second equation is being expressed by (P r(1) Γ is given by
(25.5))
1 2u
P r(1) Γ(∆2 ) = 0 ⇔ V [t] − U [x] · − U − ux = 0 (25.26)
1 + u2 (1 + u2 )2
350
Substituting the coefficients V [t] and U [x] given by (25.15) and (25.14) re-
spectively, the previous equation becomes
Vt + (Vv − Tt )vt − (vt )2 Tv + ut Vu − vx Xt
− vt ut Tu − vx ut Xu − vt vx Xv
1
− Ux + (Uu − Xx )ux − (ux )2 Xu + vx Uv − ut Tx
1 + u2 (25.27)
− ux ut Tu − ut vx Tv − ux vx Xv
2uU
+ ux = 0
(1 + u2 )2
In the above equation we substitute vx by u and ux by (1 + u2 )vt to account
for the conditions ∆1 = 0 and ∆2 = 0. We thus see that the condition
P r(1) Γ(∆2 )|∆1 =0, ∆2 =0 = 0 is equivalent to the equation
Vt + (Vv − Tt )vt − (vt )2 Tv + ut Vu − uXt

− vt ut Tu − uut Xu − uvt Xv
1
− Ux + (Uu − Xx )(1 + u2 )vt − (1 + u2 )2 (vt )2 Xu + uUv
1 + u2 (25.28)
2 2
− ut Tx − (1 + u )vt ut Tu − uut Tv − u(1 + u )vt Xv
2uU
+ (1 + u2 )vt = 0
(1 + u2 )2
1
Vt − uXt − (Ux + uUv )
1 + u2
2uU
+ (Vv − Tt ) − (Uu − Xx ) + vt
1 + u2 (25.29)
+ [−Tv + (1 + u2 )Xu ](vt )2

1
+ Vu − uXu + (Tx + uTv ) ut = 0
1 + u2
Equating to zero the coefficients of the various derivatives of the functions u
and v, we obtain the system of differential equations
1
Vt − uXt − (Ux + uUv ) = 0
1 + u2
351
2uU
(Vv − Tt ) − (Uu − Xx ) + =0
1 + u2
−Tv + (1 + u2 )Xu = 0
1
Vu − uXu + (Tx + uTv ) = 0
1 + u2
The above system can be written as (after multiplying by (1 + u2 ))
(1 + u2 )Vt − u(1 + u2 )Xt − Ux − uUv = 0 (25.30)
(1 + u2 )(Vv − Tt − Uu + Xx ) + 2uU = 0 (25.31)

(1 + u2 )Xu − Tv = 0 (25.32)
(1 + u2 )Vu − u(1 + u2 )Xu + Tx + uTv = 0 (25.33)
We are going to solve the system of equations (25.23)-(25.25) and (25.30)-
(25.33). Adding and subtracting (25.33) and (25.24) we obtain the two equa-
tions
Vu − uXu = 0 (25.34)
and
Tx + uTv = 0 (25.35)
From (25.25) we obtain that Tu = 0 and thus using (25.35) we get Tx = 0

and Tv = 0, which means that T depends on t only:
T = T (t) (25.36)
Because of that, we obtain from (25.32) that Xu = 0 and then from (25.34),
Vu = 0. At this stage we solve (25.23) with respect to U
U = Vx + (Vv − Xx )u − u2 Xv (25.37)
and then differentiate with respect to u, v and x, taking into account that
T , X and V are independent of u:
Uu = (Vv − Xx ) − 2uXv (25.38)
Uv = Vxv + (Vvv − Xxv )u − Xvv u2 (25.39)
352
Ux = Vxx + (Vvx − Xxx )u − Xxv u2 (25.40)
After substitution of (25.37)-(25.40) into (25.30) and (25.31), we obtain the
two equations
(Xvv − Xt )u3 + (2Xxv − Vvv + Vt )u2 + (Xxx − 2Xxv − Xt )u

(25.41)
+ (Vt − Vxx ) = 0
and
(2Vv − Tt )u2 + 2(Xv + Vx )u + 2Xx − Tt = 0 (25.42)
Equating the various coefficients of u to zero, we obtain from the two last
equations the following system
Xvv − Xt = 0 (25.43)
2Xxv − Vvv + Vt = 0 (25.44)

Xxx − 2Xxv − Xt = 0 (25.45)
Vt − Vxx = 0 (25.46)
and
2Vv − Tt = 0 (25.47)
Xv + Vx = 0 (25.48)
2Xx − Tt = 0 (25.49)
We differentiate next (25.47) and (25.48) with respect to v and x and we
obtain
Vxv = 0, Vvv = 0, Xxx = 0, Xxv = 0 (25.50)
We then derive from (25.44) and (25.45)
Vt = 0, Xt = 0 (25.51)
Differentiation with respect to t (25.49) gives Ttt = 0 from which we have
T = 2a1 t + a2 (25.52)
353
From (25.47) and (25.49), using this value of T , we get the equations
2Vv − 2a1 = 0 and 2Xx − 2a1 = 0 (25.53)
Integrating the above two equations with respect to v and x respectively, we
obtain
V = a1 v + α(x), X = a1 x + β(v) (25.54)
taking into account that V = V (x, v) and X = X(x, v). From (25.43) and
(25.46) we get (using the above expressions of X and V )
β 00 (v) = 0 and α00 (x) = 0 (25.55)
and thus
α(x) = a3 x + a4 and β(v) = a6 v + a5 (25.56)
Therefore
V = a1 v + a3 x + a4 and X = a1 x + a6 v + a5
The above expressions are consistent to (25.48) provided that a6 = −a3 . We
thus have
V = a1 v + a3 x + a4 and X = a1 x − a3 v + a5 (25.57)
Finally from (25.37) we have
U = (1 + u2 )a3 (25.58)
using the known expressions of V and X.
We are now in a position to write down the generator (25.4) of symmetries
in explicit form:
∂ ∂ ∂ ∂
Γ=T +X +U +V
∂t ∂x ∂u ∂v
∂ ∂ ∂
= (2a1 t + a2 ) + (a1 x − a3 v + a5 ) + (1 + u2 )a3
∂t ∂x ∂u
∂
+ (a1 v + a3 x + a4 ) (25.59)
∂v
∂ ∂ ∂ ∂
= a1 2t + x +v + a2
∂t ∂x ∂v ∂t
∂ ∂ ∂ ∂ ∂
+ a3 −v + (1 + u2 ) +x + a4 + a5
∂x ∂u ∂v ∂v ∂x
354
We thus have the following generators of the Lie Algebra of symmetries for
the system (25.17)
∂ ∂ ∂ ∂
Γ(1) = 2t +x + v , Γ(2) =
∂t ∂x ∂v ∂t (25.60)
(3) ∂ ∂ ∂ ∂ ∂
Γ = −v + (1 + u2 ) + x , Γ(4) = , Γ (5)
=
∂x ∂u ∂v ∂v ∂x
Solution 1. We shall find the general solution of the system corresponding
to the generator
∂ ∂ ∂
Γ(1) = 2t +x +v (25.61)
∂t ∂x ∂v
The invariant surface conditions corresponding to the above generator read
2tut + xux = 0, 2tvt + xvx = v (25.62)
with general solutions given by

x √ x
u=f √ and v = t·g √ (25.63)
t t
respectively. Introducing the similarity variable
x
ξ=√ (25.64)
t
we find the following expressions for the partial derivatives
1 x 1
vx = g 0 (ξ), vt = √ g(ξ) − g 0 (ξ), ux = √ f 0 (ξ) (25.65)
2 t 2t t
Therefore the system
ux
vx = u, vt = (25.66)
1 + u2
is being converted into
1
√ f 0 (ξ)
1 x t
g 0 (ξ) = f (ξ), √ g(ξ) − g 0 (ξ) = (25.67)
2 t 2t 1 + f 2 (ξ)
355
√
and then multiplying the second equation by 2 t and using the similarity
variable ξ defined by (25.64), we find the system
2f 0 (ξ)
g 0 (ξ) = f (ξ), g(ξ) − ξ·g 0 (ξ) = (25.68)
1 + f 2 (ξ)
Solution 2. We shall now find another general solution corresponding to
the linear combination of the generators Γ(2) − Γ(3) :
∂ ∂ ∂ ∂
Γ(2) − Γ(3) = +v − (1 + u2 ) −x (25.69)
∂t ∂x ∂u ∂v
The invariant surface conditions corresponding to the above generator read
ut + vux = −(1 + u2 ), vt + vvx = −x (25.70)
The general solution of the second equation is given by
v
arctan + t = F (x2 + v 2 ) (25.71)
x
However, since vx = u, the second equation of the original system becomes
vxx
vt = (25.72)
1 + vx2
Evaluating the partial derivatives of (25.71), after introducing the similarity
variable
η = x2 + v 2 (25.73)
we find
vt
+ 1 = 2vvt F 0 (η) (25.74)
v2
x 1+ 2
x
vx v
− 2
x x = 2(x + vv )F 0 (η) (25.75)
x
v2
1+ 2
x
vxx 2vx 2v v
x v 2vvx 2v 2
− 2 + 3 − 2 − 3
x x x − x x x2 x
v2 v 2 2 (25.76)
1+ 2 1+ 2
x x
= 4(x + vvx ) F (η) + 2(1 + (vx )2 + vvxx )F 0 (η)
2 00
356
Solving the system of the above three equations with respect to vt , vx and
vxx , we find (using any CAS like MAPLE or Mathematica)
η v + 2xη·F 0 (η)
vt = , vx = − (25.77)
−x + 2vη·F 0 (η) −x + 2vη·F 0 (η)
2η 2 [3F 0 (η) + 2η·F 00 (η) + 4η 2 (F 0 (η))3 ]
vxx = − (25.78)
(−x + 2vη·F 0 (η))3
Equation (25.72), because of (25.77) and (25.78) takes on the form (using
any CAS like MAPLE or Mathematica)
(x2 + v 2 ) + 4(x2 + v 2 )η 2 [F 0 (η)]2 + 6ηF 0 (η) + 4η 2 F 00 (η) + 8η 3 [F 0 (η)]3 = 0
which is equivalent to (since x2 + v 2 = η)
4ηF 00 (η) + 6F 0 (η) + 4η 2 {F 0 (η)}2 + 8η 2 {F 0 (η)}3 + 1 = 0 (25.79)
In view of the difficulty of the above equation, we try another method of

solution. In fact, upon substitution of vt given by the second equation of the
original system into the second of (25.70), we obtain
ux
+ vvx + x = 0 (25.80)
1 + u2
Dividing the first of (25.70) by 1 + u2 and substituting ux /(1 + u2 ) by vt , we
obtain the equation
ut
+ vvt + 1 = 0 (25.81)
1 + u2
Integrating (25.80) and (25.81), we obtain the equations
1 1
arctan(u)+ (v 2 +x2 )+f (t) = 0, arctan(u)+ v 2 +t+g(x) = 0 (25.82)
2 2
Since these two equations are identical, we should have that
1
arctan(u) + (v 2 + x2 ) + t = C (25.83)
2
We can set C = 0 without loss of generality. Substituting u = vx and using
(25.71) we obtain
1
arctan(vx ) + (v 2 + x2 ) + t = 0 (25.84)
2
357
1 η
F 0 (η) = − tan F + , η = v 2 + x2 (25.85)
2η 2
package called MathLie in solving first order system of partial differential
equations. The MathLie is an AddOn package to Mathematica (Reference
[4]).
We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the first order system of PDEs using the command

∂x U [x, t]
In[2] := sysEqns = ∂x V [x, t] == U [x, t], ∂t V [x, t] == ;
1 + U [x, t]2
sysEqns//LT F
− U + Vx == 0
Ux
− + Vt == 0
1 + U2
In[3] := infEqns = Infinitesimals[sysEqns, {U, V }, {x, t}, {},

SubsitutionRules→{∂x V [x, t], ∂x U [x, t]}]; infEqns//LT F
φ1 == −k4(1 + U 2 )
φ2 == k2 + k6 V − k4 x
ξ1 == k3 + k5 + k4 V + k6 x
ξ2 == k1 + 2 k6 t
358
26 Systems of Partial Differential Equations
of the Second Order
We consider a system of PDEs of the second order
F (x, y, u, v, ux , vx , uy , vy , uxx , ..., vyy ) = 0 (26.1)
G(x, y, u, v, ux , vx , uy , vy , uxx , ..., vyy ) = 0 (26.2)

Invariance of the system is being expressed by the conditions
P r(2) Γ(F )|F =0, G=0 = 0 andP r(2) Γ(G)|F =0, G=0 = 0 (26.3)

∂ ∂ ∂ ∂
Γ=X +Y +U +V (26.4)
∂x ∂y ∂u ∂v
and P r(1) Γ, P r(2) Γ are the extended operators of the first and second order,
defined by
∂ ∂ ∂ ∂
P r(1) Γ = Γ + U [x] + U [y] + V [x] + V [y] (26.5)
∂ux ∂uy ∂vx ∂vy
∂ ∂ ∂
P r(2) Γ = Γ(1) + U [xx] + U [xy] + U [yy]
(26.6)
[xx] ∂ ∂ ∂
+V + V [xy] + V [yy]
∂vxx ∂vxy ∂vyy
It is obvious that
X = X(x, y, u, v), Y = Y (x, y, u, v),
(26.7)
U = U (x, y, u, v), V = V (x, y, u, v)
U [x] = Dx (U ) − ux Dx (X) − uy Dx (Y ) (26.8)
U [y] = Dy (U ) − ux Dy (X) − uy Dy (Y ) (26.9)

V [x] = Dx (V ) − vx Dx (X) − vy Dx (Y ) (26.10)
V [y] = Dy (V ) − vx Dy (X) − vy Dy (Y ) (26.11)
359
and
U [xx] = Dx (U [x] ) − uxx Dx (X) − uxy Dx (Y ) (26.12)
U [xy] = Dx (U [y] ) − uxy Dx (X) − uyy Dx (Y ) (26.13)
U [yy] = Dy (U [y] ) − uxy Dy (X) − uyy Dy (Y ) (26.14)
[xx] [x]
V = Dx (V ) − vxx Dx (X) − vxy Dx (Y ) (26.15)
[xy] [y]
V = Dx (V ) − vxy Dx (X) − vyy Dx (Y ) (26.16)
V [yy] = Dy (V [y] ) − vxy Dy (X) − vyy Dy (Y ) (26.17)
The total differential operators Dx and Dy are defined by
∂ ∂ ∂ ∂
Dx = + ux + uxx + uxy + ···+
∂x ∂u ∂ux ∂uy
(26.18)
∂ ∂ ∂
+ vx + vxx + vxy + ···
∂v ∂vx ∂vy
∂ ∂ ∂ ∂
Dy = + uy + uxy + uyy + ···+
∂y ∂u ∂ux ∂uy
(26.19)
∂ ∂ ∂
+ vy + vxy + vyy + ···
∂v ∂vx ∂vy
We shall calculate explicitly some of the coefficients U [...] , V [...] .
We have for example
U [x] = Dx (U ) − ux Dx (X) − uy Dx (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + ux + vx U − ux + ux + vx X
∂x ∂u ∂v ∂x ∂u ∂v
∂ ∂ ∂ (26.20)
− uy + ux + vx Y
∂x ∂u ∂v
= Ux + (Uu − Xx )ux − (ux )2 Xu − ux vx Xv − ux uy Yu
− uy vx Yv − uy Yx − ux ut Tu − ut vx Tv − ut Tx + vx Uv
U [y] = Dy (U ) − ux Dy (X) − uy Dy (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + uy + vy U − ux + uy + vy X
∂y ∂u ∂v ∂y ∂u ∂v
∂ ∂ ∂ (26.21)
− uy + uy + vy Y
∂y ∂u ∂v
= Uy + (Uu − Yy )uy − (uy )2 Yu − ut Ty + (Uv − ut Tv )vy
− (uy Xu + vy Xv + Xy )ux − (vy Yv + ut Tu )uy
360
V [y] = Dy (V ) − vx Dy (X) − vy Dy (Y )
∂ ∂ ∂ ∂ ∂ ∂
= + uy + vy V − vx + uy + vy X
∂y ∂u ∂v ∂y ∂u ∂v
∂ ∂ ∂ (26.22)
− vy + uy + vy Y
∂y ∂u ∂v
= Vy + (Vv − Yy )vy − (vy )2 Yv + (Vu − vt Tu )uy
− (uy Xu + vy Xv + Xy )vx − (uy Yu + vt Tv )vy − vt Ty
and
U [xx] = Dx (U [x] ) − uxx Dx (X) − uxy Dx (Y )

∂ ∂ ∂ ∂ ∂ ∂ ∂ [x]
= + ux + uxx + uxy + vx + vxx + vxy U
∂x ∂u ∂ux ∂uy ∂v ∂vx ∂vy
∂ ∂ ∂ ∂ ∂ ∂
− uxx + ux + vx X − uxy + ux + vx Y
∂x ∂u ∂v ∂x ∂u ∂v
= Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 + (Uu − 2Xx )uxx
− Xuu (ux )3 − 2(ux Tu + vx Tv + Tx )uxt − (uxx Tu + vxx Tv + Txx )ut
− 2(ux vx Tuv + ux Txu + vx Txv )ut − [(ux )2 Tuu + (vx )2 Tvv ]ut
− (2vx Xv + 3ux Xu + uy Yu )uxx − 2(ux Yu + vx Yv + Yx )uxy
+ (Uv − uy Xv − ux Xv )vxx − (2vx Xuv + uy Yuu )(ux )2
+ (Uvv − ux Xvv − uy Yvv )(vx )2 + 2(Uuv − Xxv )ux vx
− 2(vx Yuv + Yxu )ux uy − (2vx Yxv + Yxx )uy + 2Uxv vx
(26.23)
361
U [yy] = Dy (U [y] ) − uxy Dy (X) − uyy Dy (Y )
∂ ∂ ∂ ∂ ∂ ∂ ∂ [y]
= + uy + uxy + uyy + vy + vxy + vyy U
∂y ∂u ∂ux ∂uy ∂v ∂vx ∂vy
∂ ∂ ∂ ∂ ∂ ∂
− uxy + uy + vy X − uyy + uy + vy Y
∂y ∂u ∂v ∂y ∂u ∂v
= Uyy + (2Uyu − Yyy )uy + (Uu − 2Yy )uyy + (Uuu − 2Yyu )(uy )2
− Yuu (uy )3 − 2(uy Tu + vy Tv + Ty )uyt − (uyy Tu + vyy Tv + Tyy )ut
− 2(uy vy Tuv + uy Tyu + vy Tyv )ut − [(uy )2 Tuu + (vy )2 Tvv ]ut
− (2vy Yv + 3uy Yu + ux Xu )uyy − 2(uy Xu + vy Xv + Xy )uxy
+ (Uv + ux Xv + uy Yv )vyy − (2vy Yuv + ux Xuu )(uy )2
+ (Uvv − ux Xvv − uy Yvv )(vy )2 + 2(Uuv − Yyv )uy vy
− 2(vy Xuv + Xyu )ux uy − (2vy Xyv + Xyy )ux + 2Uyv vy
(26.24)
Example. Solve the system of PDEs
ux + vy = 0, uux + vuy = uyy (26.25)
The above system appears in Boundary Layer Theory in the Mechanics of

Fluids.
Solution. We denote by ∆1 and ∆2 the quantities
∆1 = ux + vy and ∆2 = uux + vuy − uyy (26.26)
Invariance of the first equation is being expressed by
P r(2) Γ(∆1 ) = 0 ⇔ U [x] + V [y] = 0 (26.27)
where P r(2) Γ is given by (26.6). Substituting the coefficients U [x] and V [y]
given by (26.20) and (26.22) into the above equation, we obtain the equation
Ux + (Uu − Xx )ux − (ux )2 Xu − ux vx Xv − ux uy Yu

− uy vx Yv − uy Yx + vx Uv
(26.28)
+ Vy + (Vv − Yy )vy − (vy )2 Yv + Vu uy
− (uy Xu + vy Xv + Xy )vx − uy vy Yu = 0
In the above equation we have disregarded the time-dependent derivatives,

since the original system (26.25) is time-independent. In this equation we
362
substitute vy by −ux and uyy by uux + vuy , to account for the conditions
∆1 = 0 and ∆2 = 0.
We thus see that the condition P r(2) Γ(∆1 )|∆1 =0, ∆2 =0 = 0 is equivalent to the
equation
Ux + (Uu − Xx )ux − (ux )2 Xu − ux vx Xv − ux uy Yu

− uy vx Yv − uy Yx + vx Uv
(26.29)
+ Vy + (Vv − Yy )(−ux ) − (−ux )2 Yv + Vu uy
− (uy Xu + (−ux )Xv + Xy )vx − uy (−ux )Yu = 0
We now have to set to zero all the coefficients of the partial derivatives in
order to determine the quantities X, Y , U and V . Before that, the above
equation can be written as
Ux + Vy + (Uu − Xx )ux − (Vv − Yy )ux

− uy vx Yv + vx Uv − (uy Xu + Xy )vx (26.30)
− (ux )2 Xu − (ux )2 Yv + Vu uy − Yx uy = 0
We are now in a position to equate to zero the various coefficients of the

partial derivatives. We have:
Coefficient of ux
Uu − Xx − Vv + Yy = 0 (26.31)
Coefficient of uy
V u − Yx = 0 (26.32)
Coefficient of (ux )2
X u + Yv = 0 (26.33)
Coefficient of vx
U v − Xy = 0 (26.34)
Coefficient of uy vx
X u + Yv = 0 (26.35)
363
The constant term
Ux + Vy = 0 (26.36)
Invariance of the second equation is being expressed by
P r(2) Γ(∆2 ) = 0 ⇔ U ux + uU [x] + V uy + vU [y] − U [yy] = 0 (26.37)
where P r(2) Γ is given by (26.6).

Substituting the coefficients U [x] , U [y] and U [yy] given by (26.20), (26.21) and
(26.24) respectively, we obtain the equation

U ux + u Ux + (Uu − Xx )ux − (ux )2 Xu − ux vx Xv − ux uy Yu

− uy vx Yv − uy Yx + vx Uv + V uy

+ v Uy + (Uu − Yy )uy − (uy )2 Yu + Uv vy

− (uy Xu + vy Xv + Xy )ux − uy vy Yv
(26.38)
− Uyy + (2Uyu − Yyy )uy + (Uu − 2Yy )uyy + (Uuu − 2Yyu )(uy )2
− Yuu (uy )3
− (2vy Yv + 3uy Yu + ux Xu )uyy − 2(uy Xu + vy Xv + Xy )uxy
+ (Uvv − ux Xvv − uy Yvv )(vy )2 + 2(Uuv − Yyv )uy vy

− 2(vy Xuv + Xyu )ux uy − (2vy Xyv + Xyy )ux + 2Uyv vy = 0
In the above equation we have disregarded the time-dependent derivatives,

since the original system (26.25) is time-independent. In this equation we
also substitute vy by −ux and uyy by uux + vuy , to account for the conditions
∆1 = 0 and ∆2 = 0.
We thus see that the condition P r(2) Γ(∆2 )|∆1 =0, ∆2 =0 = 0 is equivalent to the
364
equation



+ v Uy + (Uu − Yy )uy − (uy )2 Yu − Uv ux

− (uy Xu − Xv ux + Xy )ux + ux uy Yv

− Uyy + (2Uyu − Yyy )uy + (Uu − 2Yy )(uux + vuy ) (26.39)
+ (Uuu − 2Yyu )(uy )2 − Yuu (uy )3

− (−2Yv ux + 3uy Yu + ux Xu )(uux + vuy )
− 2(uy Xu − Xv ux + Xy )uxy
+ (Uvv − ux Xvv − uy Yvv )(ux )2 − 2(Uuv − Yyv )ux uy

− 2(−Xuv ux + Xyu )ux uy − (−2Xyv ux + Xyy )ux − 2Uyv ux = 0
We now have to set to zero the coefficients of the partial derivatives of the
functions u and v. Before that, we see that some coefficients can be identified
more easily compared to some other ones. For example the coefficient of
(uy )3 is Yuu , the coefficient of uxy is 2Xy , the coefficient of ux uxy is −2Xv ,
the coefficient of uy uxy is 2Xu , the coefficient of vyy is −Uv , the coefficient
of ux vyy is −Xv , the coefficient of uy vyy is −Yv , the coefficient of (uy )2 is
2vYu − (Uuu − 2Yyu ), the coefficient of ux (uy )2 is Xuu , and the coefficient of
vy (uy )2 is 2Yuv . Therefore we have the following equations
Yuu = 0 (26.40)
Xy = 0, Xu = 0, Xv = 0 (26.41)
Uv = 0, Yv = 0 (26.42)
Yuv = 0, Xuu = 0, 2vYu − (Uuu − 2Yyu ) = 0 (26.43)
365
Because of the above four systems, equation (26.39) becomes



+ v Uy + (Uu − Yy )uy − Uv ux

− (uy Xu − Xv ux + Xy )ux + ux uy Yv (26.44)

− Uyy + (2Uyu − Yyy )uy + (Uu − 2Yy )(uux + vuy )
− (−2Yv ux + 3uy Yu + ux Xu )(uux ) − (−2Yv ux + ux Xu )(vuy )

+ (Uvv − ux Xvv − uy Yvv )(ux )2 − 2(Uuv − Yyv )ux uy

− 2(−Xuv ux + Xyu )ux uy − (−2Xyv ux + Xyy )ux − 2Uyv ux = 0
From the above equation we obtain:

Coefficient of (ux )3 is equal to Xvv , coefficient of ux vx is equal to −uXv ,
coefficient of vx is equal to uUv , coefficient of uy vx is equal to −uYv . We thus
obtain the system
Xv = 0, Uv = 0 and Yv = 0 (26.45)
We further equate to zero the coefficients of (ux )2 and ux and obtain the two
equations
vXv − 2uYv − Uvv − 2Xyv = 0 (26.46)
and
U − uXx − vUv − vXy + 2uYy + Xyy + 2Uyv = 0 (26.47)
respectively. We then equate to zero the terms which are independent of the
partial derivatives of the two functions. We obtain the equation
uUx + vUy − Uyy = 0 (26.48)
366
Equation (26.44), because of the previous four systems, gets simplified into
− u(ux uy Yu + uy Yx ) + V uy

+ v (Uu − Yy )uy − Xu ux uy + ux uy Yv

− (2Uyu − Yyy )uy + v(Uu − 2Yy )uy (26.49)
− (3uy Yu )(uux ) − (−2Yv ux + ux Xu )(vuy )

− 2(Uuv − Yyv )ux uy − 2(Xyu )ux uy = 0
Equating to zero the coefficient of uy , we obtain the equation

−uYx + V − 2Uyu + Yyy + vYy = 0 (26.50)
We finally equate to zero the coefficient of ux uy and obtain the equation
2uYu − vYv + 2Uuv − 2Yyv + 2Xyu = 0 (26.51)
We now have to solve the systems of equations (26.40)-(26.43), (26.45)-
(26.48) and (26.50)-(26.51).
Since Xu = 0 and Xy = 0 from (26.41) and Yv = Uv = 0 from (26.42), we
obtain from (26.51) that Yu = 0. Because of
Xu = 0, Xy = 0, Yu = 0 (26.52)
equation (26.47) takes the form
U − uXx + 2uYy = 0 (26.53)
From (26.34), (26.36) and (26.48) we obtain
Xxx = 0, Yxy = 0, Yyy = 0 (26.54)
Therefore
X = a1 x + a2 and Y = a3 y + a(x) (26.55)
where a(x) is an arbitrary function. Finally solving (26.53) with respect to
U and (26.50) with respect to V and substituting the known functions, we
find
U = uXx − 2uYy = a1 u − 2a3 u = (a1 − 2a3 )u (26.56)
367
and
V = uYx + 2Uyu − Yyy − vYy = a0 (x)u − a3 v (26.57)
respectively.
We are now in a position to obtain the explicit expression of the operator Γ,
defined in (26.4):
∂ ∂ ∂ ∂
Γ=X +Y +U +V
∂x ∂y ∂u ∂v
∂ ∂ ∂
= (a1 x + a2 ) + (a3 y + a(x)) + (a1 − 2a3 )u
∂x ∂y ∂u
∂
+ (a0 (x)u − a3 v) (26.58)
∂v
∂ ∂ ∂ ∂ ∂ ∂
= a1 x +u + a2 + a3 y − 2u −v
∂x ∂u ∂x ∂y ∂u ∂v
∂ ∂
+ a(x) + a0 (x)u
∂y ∂v
We thus have the following generators of the Lie Algebra of symmetries for
the system (26.25)
∂ ∂ ∂
Γ(1) = x + u , Γ(2) =
∂x ∂u ∂x
∂ ∂ ∂
Γ(3) =y − 2u −v , (26.59)
∂y ∂u ∂v
∂ ∂
Γ(4) = a(x) , Γ(5) = a0 (x)u
∂y ∂v
In order to find some general solutions of the system, we introduce a potential
w(x, y) such that
u = wy , v = −wx (26.60)
Then the first equation of (26.25) is being automatically satisfied, whereas
the second equation becomes
wy wxy − wx wyy = wyyy (26.61)
The invariant surface conditions
Xux + Y uy = U, Xvx + Y vy = V
368
taking into account
X = a1 x + a2 , Y = a3 y + a(x),
(26.62)
U = (a1 − 2a3 )u, V = a0 (x)u − a3 v
take the form
(a1 x + a2 )ux + (a3 y + a(x))uy = (a1 − 2a3 )u (26.63)
and
(a1 x + a2 )vx + (a3 y + a(x))vy = a0 (x)u − a3 v (26.64)
respectively.
Using relations (26.60), the invariant surface conditions (26.63) and (26.64)
become
(a1 x + a2 )wxy + (a3 y + a(x))wyy = (a1 − 2a3 )wy (26.65)
and
−(a1 x + a2 )wxx − (a3 y + a(x))wxy = a0 (x)wy + a3 wx (26.66)
respectively.
Integrating (26.65) with respect to y, we obtain (the integration constant is
taken equal to zero)
(a1 x + a2 )wx + (a3 y + a(x))wy = (a1 − 2a3 )w (26.67)
The above equation is to be solved in the special case a2 = a3 = 0.

In this case (26.67) becomes
a1 xwx + a(x)wy = a1 w (26.68)
The general solution of (26.68) is given by
w = x·F (ξ), ξ = y − J(x) (26.69)
where
Z
1 a(x)
J(x) = (26.70)
a1 x
369
Upon substitution of (26.69) into (21.61), we obtain
F 000 (ξ) − F (ξ)F 00 (ξ) − (F 0 (ξ))2 = 0 (26.71)
Equation (26.71) is a nonlinear ordinary differential equation which can be

simplified using Nucci’s reduction algorithm.
We introduce the variables
w1 = F (ξ), w2 = F 0 (ξ), w3 = F 00 (ξ) (26.72)
We then have
w10 = w2 , w20 = w3 , w30 − w1 w3 − (w2 )2 = 0 (26.73)
The above equations can also be written as

dw1 dw2 dw3
= w2 , = w3 , = w1 w3 + (w2 )2 (26.74)
dξ dξ dξ
Upon dividing the last two equations by the first one, we obtain the equations
dw2 w3 dw3 w1 w3 + (w2 )2
= and = (26.75)
dw1 w2 dw1 w2
Solving the first of the above equations with respect to w3 , i.e.
dw2
w3 = w2 (26.76)
dw1
and substitution into the second, we get the equation
d dw2 dw
2
w2 w2 = w1 w2 + (w2 )2
dw1 dw1 dw1
d2 w2 dw2 2 dw2
w2 2
+ − w1 − w2 = 0 (26.77)
dw1 dw1 dw1
We thus see that we have reduced the third order ODE (26.71) into a second
order ODE with unknown function w2 (w1 ).
Equation (26.77), written in moderate notation as
d2 y dy 2 dy
y 2
+ − x − y = 0, y = y(x) (26.78)
dx dx dx
370
will be solved using Lie point symmetry analysis.
Equation (26.78) admits the infinitesimals
{X = x, Y = 2y} (26.79)
We can go over to canonical coordinates (r, s) by solving the system

∂r ∂r
X+ Y =0
∂x ∂y
(26.80)
∂s ∂s
X+ Y =1
∂x ∂y
for X = x, Y = 2y. We find that the general solution of the above system is
given by
y y
r = F 2 , s = ln(x) + G 2 (26.81)
x x
We make the choice
y
r = 2 , s = ln(x) (26.82)
x
Inverting the above system, i.e. considering the substitution
x = es(r) , y = r e2s(r) (26.83)
equation (26.78) gets transformed into the equation

d2 s 2
ds 3 ds 2 ds
r − (6r − 3r) − (7r − 1) − =0 (26.84)
dr2 dr dr dr
The substitution
ds
U= (26.85)
dr
transforms equation (26.84) into the first order ODE
dU
r − (6r2 − 3r)U 3 − (7r − 1)U 2 − U = 0 (26.86)
dr
which is a first kind Abel’s equation. This equation under the substitution
r
U= (26.87)
Y (r)
371
gets transformed into the equation
dY
w + (7r − 1)w + 6r3 − 3r2 = 0 (26.88)
dr
The above equation under the substitution
Z
7
z = − (7r − 1)dr = − r2 + r = f (r) (26.89)
2
is being transformed into the equation
dY (z) 6r3 − 3r2
Y (z) − Y (z) = (26.90)
dz 7r − 1 r=f −1 (z)
Since
1 √
f −1 (z) = 1± 1 − 14z (26.91)
7
equation (26.90) splits into the two equations
√ 2 √
dY (z) 3 1 + 1 − 14z −5 + 2 1 − 14z
Y (z) − Y (z) = √ (26.92)
dz 343 1 − 14z
and
√ 2 √
dY (z) 3 −1 + 1 − 14z 5 + 2 1 − 14z
Y (z) − Y (z) = √ (26.93)
dz 343 1 − 14z
The above equations (26.92) and (26.93) are Abels’s quations of the second
kind and thus can be solved by Panayotounakos algorithm.
Integrating (26.85) and using (26.87), we find
Z Z
r
s = U (r)dr = dr (26.94)
Y (r)
where Y (r) has been determined from either (26.92) or (26.93) using the
Panayotounakos algorithm. Going back to the original variables, i.e. using
(26.82), we get from (26.94) that
Z
r
ln(x) = dr (26.95)
Y (r) r=y/x2
372
This is the general solution of equation (26.78).
Changing once more variables, i.e. x→w1 , y→w2 we see that the solution of
(26.77) is given by
Z
r
ln(w1 ) = dr (26.96)
Y (r) r=w2 /w2
1
We thus suppose that we have found w2 = w2 (w1 ), thanks to (26.96). Using

the first of (26.74), i.e.
dw1
= w2 (w1 ) (26.97)
dξ
we obtain w1 by solving the above differential equation and thus F (ξ):
w1 = F (ξ) (26.98)
After determining F (ξ), we can determine w from (26.69). The functions

u(x, y) and v(x, y) which satisfy the original system (26.25) are determine
from (26.60), i.e. by
u(x, y) = wy and v(x, y) = −wx (26.99)
Note. The Panayotounakos algorithm solves Abel’s equation of the sec-

ond kind as follows: The solution of the differential equation (Abel’s equation
of the second kind)
dy
yy 0 − y = f (x), y≡y(x), y0 = (26.100)
dx
is given by
1 1
y(x) = (x + 2C) r(x) + (26.101)
3 3
where C is a constant and r(x) are the roots of the cubic equation
r3 (t) + p r(t) + q = 0, t = ln |x + 2C| (26.102)
The quantities p and q are given by

1 1 a 3
p = b − a3 , q = c − ab + 2 (26.103)
3 3 3
373
a = −4, b = 3 + 4[g(t) + f (t)]e−t , c = −4[g(t) + 2f (t)]e−t (26.104)
and g(t) is a function defined by
[(t sin(t) + cos(t))ci(t) + cos2 (t)][4t ci(t) + cos(t)]e−t

g(t) = − 2f (t)
2[2t ci(t)]3
(26.105)
The function ci(t) is the known cosine integral function defined by
Zt
cosu − 1
ci(t) = γ + ln(t) + du (26.106)
u
0
where γ is the Euler-Mascheroni constant.

package called MathLie in solving second order system (26.65) of partial
differential equations. The MathLie is an AddOn package to Mathematica
(Reference [4]). We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the second order system of PDEs using the command
In[2] := sysEqns = {∂x U [x, y] + ∂y V [x, t] == 0,

U [x, y] ∗ ∂x U [x, y] + V [x, y] ∗ ∂y U [x, y] == ∂{y,2} U [x, y]};
sysEqns//LT F
Ux + Vy == 0
U Ux + V Uy − Uy,y == 0
In[3] := infEqns = Infinitesimals[sysEqns, {U, V }, {x, y}, {},

SubsitutionRules→{∂y V [x, y], ∂{y,2} U [x, y]}]; infEqns//LT F
374
φ1 == (k2 − 2 k3) U
φ2 == −k3 V + U F1x
ξ1 == k1 + k2 x
ξ2 == k3 y + F1
Example 2. Let us now consider another system of second order PDEs
ut = vxx , vt = uxx , u = u(x, t), v = v(x, t)
using MathLie. We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the second order system of PDEs using the command
In[2] := sysEqns = {∂t U [x, t] − ∂{x,2} V [x, t] == 0,
∂t V [x, t] − ∂{x,2} U [x, t] == 0}; sysEqns//LT F
Ut − Vx,x == 0
Vt − Ux,x == 0
In[3] := infEqns = Infinitesimals[sysEqns, {U, V }, {x, t}, {},
SubsitutionRules→{∂t U [x, t], ∂t V [x, t]}]; infEqns//LT F
− F2t + F1x,x == 0
− F1t + F2x,x == 0
ξ1 == k6 + k7 x − 2t(k3 + k5 x)
ξ2 == k1 + 2t(k7 − k5 t)
k4 x2
φ1 == (k4 + k5 t)U + V k2 + k3 x + + F1
2
k4 x2
φ2 == (k4 + k5 t)V + U k2 + k3 x + + F2
2
375
Step 4. We consider the following infinitesimals
In[4] := xinfi =
{xi[1][x, t, U, V ], xi[2][x, t, U, V ]}/.infEqns[1]/.
{k1→0, k2→0, k3→0, k4→1, k5→0, k6→0, k7→1}
with response
{x, 2t}
and
In[5] := uinfi =
{phi[1][x, t, U, V ], phi[2][x, t, U, V ]}/.infEqns[1]/.
{k1→0, k2→0, k3→0, k4→1, k5→0, k6→0, k7→1,
f ree[1][x, t]→0, f ree[2][x, t]→0}
with response
{U, V }
Step 5. Finally we use the LieReduction[] command
In[6] := redEqns =
LieReduction[sysEqns, {U, V }, {x, t}, {x, 2 ∗ t}, {U, V }];
LieT raditionalF orm[redEqns]/.zeta1→ξ//T ableF orm
Out[6] :=
t
== ξ
x2
U V
== F1 == F2
x x
F1ξ == 2ξ(F2ξ + 2ξF2ξ,ξ ) F2ξ == 2ξ(F1ξ + 2ξF1ξ,ξ )
The two equations at the bottom is a system of two coupled ODEs which
can easily be solved. Each one of F1 and F2 can be expressed in terms of the
error function.
376
Appendix. Normal Coordinates for a System of PDEs.
The invariance conditions for a system of two equations with two variables
x and t and two unknown functions u and v, u = u(x, t) and v = v(x, t)
E1 (x, t, u, v, ux , uxx , · · · , vx , vxx , · · · ) = 0
(26.107)
E1 (x, t, u, v, ux , uxx , · · · , vx , vxx , · · · ) = 0
leads to the determination of the symmetry generator
∂ ∂ ∂ ∂
Γ=X +T +U +V (26.108)
∂x ∂t ∂u ∂v
Let us now transform Γ by considering a transformation
(x, t, u, v)−→(r, s, w, p) (26.109)
with
r = r(x, t, u, v), s = s(x, t, u, v)
(26.110)
w = w(x, t, u, v), p = (x, t, u, v)
Using the chain rule of partial derivatives
∂ ∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
= + + +
∂x ∂x ∂r ∂x ∂s ∂x ∂w ∂x ∂p
∂ ∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
= + + +
∂t ∂t ∂r ∂t ∂s ∂t ∂w ∂t ∂p
∂ ∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
= + + +
∂u ∂u ∂r ∂u ∂s ∂u ∂w ∂u ∂p
∂ ∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
= + + +
∂v ∂v ∂r ∂v ∂s ∂v ∂w ∂v ∂p
we obtain
∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
Γ=X + + +
∂x ∂r ∂x ∂s ∂x ∂w ∂x ∂p
∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
+T + + +
∂t ∂r ∂t ∂s ∂t ∂w ∂t ∂p
∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
+U + + +
∂u ∂r ∂u ∂s ∂u ∂w ∂u ∂p
∂r ∂ ∂s ∂ ∂w ∂ ∂p ∂
+V + + +
∂v ∂r ∂v ∂s ∂v ∂w ∂v ∂p
377
∂
Γ = (Xrx + T rt + U ru + V rv )
∂r
∂
+ (Xsx + T st + U su + V sv )
∂s
(26.111)
∂
+ (Xwx + T wt + U wu + V wv )
∂w
∂
+ (Xpx + T pt + U pu + V pv )
∂p
We want the transformed system (26.107) to be independent of s under the

∂
transformation (26.109), i.e. we want Γ to be transformed into Γ = . We
∂s
thus obtain from (26.111) that
Xrx + T rt + U ru + V rv = 0
Xsx + T st + U su + V sv = 1
(26.112)
Xwx + T wt + U wu + V wv = 0
Xpx + T pt + U pu + V pv = 0
Solving the above system, we can determine the functions
r = r(ξ, ), s = s(η, )
(26.113)
w = w(θ, ), p = p(ρ, )
where ξ, η, are similarity variables, while the functions F (ξ), depend on the
similarity variables.
Under the above transformations, the system (26.107) gets transformed into
Ẽ1 (r, w, p, wr , · · · ) = 0
(26.114)
Ẽ2 (r, w, p, wr , · · · ) = 0
The above results can be generalized (along the same lines of reasoning) to
systems of any number of functions with any numbers of independent vari-
ables.
378
Chapter IX
The Optimal System
27 The Optimal System.
In this Section we shall develop methods of obtaining the Optimal System
of generators of the Lie Algebra of a given Differential Equation or a System
of Differential Equations (Ovsiannikov [1] and Olver [2]).
27.1 An Illustrative Example. The KdV Equation

As an Example we first consider the KdV equation (Olver [1], Ex. 2.44, Ex.
3.10, Ex. 3.11 and Ex. 3.12)
ut + uux + uxxx = 0 (27.1)
which admits the following symmetry generators
X1 = ∂ x , X2 = ∂t , X3 = t∂x + ∂u , X4 = x∂x + 3t∂t − 2u∂u (27.2)
We first consider the commutator Table [Xi , Xj ] = Xi Xj − Xj Xi , where
i, j = 1, 2, 3, 4.
X1 X2 X3 X4
X1 0 0 0 X1
X2 0 0 X1 3X2
X3 0 −X1 0 −2X3
X4 −X1 −3X2 2X3 0
The Commutator Table

We next form the table of the adjoint transformation matrix where each entry
(i, j) is being evaluated using the expansion
Ad(exp(Xi )Xj )
2 3 (27.3)
= Xj − [Xi , Xj ] + [Xi , [Xi , Xj ]] − [Xi , [Xi , [Xi , Xj ]]] + · · ·
1! 2! 3!
379
taking into account the commutation Table:
Ad X1 X2 X3 X4
X1 X1 X2 X3 X4 − X1
X2 X1 X2 X3 − X1 X4 − 3X2
X3 X1 X2 + X1 X3 X4 + 2X3
X4 e X1 e3 X2 e−2 X3 X4
The Adjoint Table
Let
X = a1 X 1 + a2 X 2 + a3 X 3 + a4 X 4 (27.4)
Case I. We assume that a4 6=0. We may take for simplicity a4 = 1. We thus

have
X = a1 X 1 + a2 X 2 + a3 X 3 + X 4 (27.5)
Let X 0 ≡Ad(exp(X3 )X) with X given by (27.5). We then obtain
X 0 ≡Ad(exp(X3 )X) = a1 Ad(exp(X3 )X1 ) + a2 Ad(exp(X3 )X2 )

+ a3 Ad(exp(X3 )X3 ) + Ad(exp(X3 )X4 )
(27.6)
= a1 X1 + a2 (X2 + X1 ) + a3 X3 + (X4 + 2X3 )
= (a1 + a2 )X1 + a2 X2 + (a3 + 2)X3 + X4
where in the third line we have used the adjoint transformation matrix. The
1
choice = − a3 eliminates X3 in (27.6). We then obtain from (27.6)
2
1
X 0 = a1 − a2 a3 X 1 + a2 X 2 + X 4 (27.7)
2
i.e.
X 0 = a01 X1 + a02 X2 + X4 (27.8)
380
1
where a01 = a1 − a2 a3 and a02 = a2 .
2
Let further X 00 ≡Ad(exp(X2 )X 0 ) with X 0 given by (27.8). We then obtain
X 00 ≡Ad(exp(X2 )X 0 ) = a01 Ad(exp(X2 )X1 ) + a02 Ad(exp(X2 )X2 )

+ Ad(exp(X2 )X4 ) = a01 X1 + a02 X2 + X4 − 3X2 (27.9)
0 0
= a1 X1 + (a2 − 3)X2 + X4
1
The choice = a02 eliminates X2 and thus
3
X 00 = a01 X1 + X4 (27.10)
Let finally X 000 ≡Ad(exp(X1 )X 00 ) with X 00 given by (27.10). We then obtain
X 000 ≡Ad(exp(X1 )X 00 ) = a01 Ad(exp(X1 )X1 ) + Ad(exp(X1 )X4 )

(27.11)
= a01 X1 + X4 − X1 = (a01 − )X1 + X4
The choice = a01 cancels X1 and thus
X 000 = X4 (27.12)
Case II. We assume that a4 = 0. In this case we consider
X = a1 X 1 + a2 X 2 + a3 X 3 (27.13)
Case II1. Let us first consider a3 6=0. We may take a3 = 1 and thus (27.13)
becomes
X = a1 X 1 + a2 X 2 + X 3 (27.14)
Let X 0 ≡Ad(exp(X2 )X) with X given by (27.14). We then obtain
X 0 ≡Ad(exp(X2 )X) = a1 Ad(exp(X2 )X1 ) + a2 Ad(exp(X2 )X2 )

+ Ad(exp(X2 )X3 ) = a1 X1 + a2 X2 + X3 − X1 (27.15)
= (a1 − )X1 + a2 X2 + X3
The choice = a1 eliminates X1 and thus
X 0 = a2 X 2 + X 3 (27.16)
381
Let further X 00 ≡Ad(exp(X4 )X 0 ) with X 0 given by (27.16). We then obtain
X 00 ≡Ad(exp(X4 )X 0 ) = a2 Ad(exp(X4 )X2 ) + Ad(exp(X4 )X3 )

(27.17)
= a2 e3 X2 + e−2 X3
which may also be written as, after rescaling
X 00 = a2 e5 X2 + X3 (27.18)
Depending on the sign of a2 , we can consider the coefficient of X2 as taking

one of the three values: +1, −1 or 0. Therefore
X 00 = X2 + X3 , X 00 = −X2 + X3 or X 00 = X3 (27.19)
Case II2. Consider the case a3 = a4 = 0, i.e.
X = a1 X 1 + a2 X 2 (27.20)
Let X 0 = Ad(exp(X3 )X) with X given by (27.20). We then obtain
X 0 = Ad(exp(X3 )X) = a1 Ad(exp(X3 )X1 ) + a2 Ad(exp(X3 )X2 )

(27.21)
= a1 X1 + a2 (X2 + X1 ) = (a1 + a2 )X1 + a2 X2
a1
The choice = − eliminates X1 and thus X 0 = a2 X2 . Taking a2 = 1 we
a2
get
X 0 = X2 (27.22)
Case II3. Consider the case a2 = a3 = a4 = 0, i.e. X = a1 X1 and for a1 6=0

we may take a1 = 1 and thus
X = X1 (27.23)
Taking into account (27.12), (27.19), (27.22) and (27.23) we get the following
optimal set of generators
X4 , X2 + X3 , −X2 + X3 X3 , X2 and X1 (27.24)
One can then determine invariant solutions corresponding to each one of

the above generators (27.24).
382
27.2 The Algorithmic Construction
Let G be a symmetry group and its corresponding Lie algebra G generated
by the vector fields (Refs. [3]-[16])
{X1 , X2 , · · · , Xn } (27.25)
Two elements X, Y ∈ G
n
X n
X
X= ai X i , Y = bi X i (27.26)
i=1 i=1
are equivalent, if there is a transformation g∈ G such that
Ad(g(Y )) = X (27.27)
The notation Ad(g) refers to the adjoint representation of g, i.e.
Ad(g(Y )) = g −1 Y g (27.28)
27.2.1 The Invariant. A real function Φ on the Lie algebra G is called an

invariant if
Φ(Ad(g(X))) = Φ(X) ∀X∈G and ∀g∈G (27.29)
If X and Y are equivalent (under the adjoint action) it is necessary that
Φ(X) = Φ(Y ) for any invariant Φ (27.30)
If
n
X
X= ai X i (27.31)
i=1
then the invariant Φ is a function of the coefficients a1 , a2 , · · · , an .

We first construct a table of the Commutation Relations (called Table I) with
enties (i, j) given by [Xi , Xj ] = Xi Xj − Xj Xi . Then taking any subgroup
g∈G with
n
X
g = eY , Y = bi X i (27.32)
i=1
383
to act on X, we obtain
Ad(exp(Y )X) = e−Y XeY
2
= X − [Y, X] + [Y, [Y, X]] − · · ·
2 (27.33)
= (a1 X1 + · · · + an Xn )
− [b1 X1 + · · · + bn Xn , a1 X1 + · · · + an Xn ] + O(2 )
According to the standard rules of the commutation relations, it is obvious
that
[b1 X1 + · · · + bn Xn , a1 X1 + · · · + an Xn ]
= b1 a2 [X1 , X2 ] + b1 a3 [X1 , X3 ] + · · · + b1 an [X1 , Xn ]
+ b2 a1 [X2 , X1 ] + b2 a3 [X2 , X3 ] + · · · + b2 an [X2 , Xn ] (27.34)
+ ···+
+ bn a1 [Xn , x1 ] + bn a2 [Xn , X2 ] + · · · + bn an−1 [Xn , Xn−1 ]
In other words, it follows from the above expression, that the commutator
[b1 X1 + · · · + bn Xn , a1 X1 + · · · + an Xn ] is an expression of the form
[b1 X1 + · · · + bn Xn , a1 X1 + · · · + an Xn ]
(27.35)
X
= aκ bλ [Xκ , Xλ ], κ, λ = 1, 2, · · · , n
κ6=λ
Taking into account the Commutator Table, the expression (27.35) can take
the form
[b1 X1 + · · · + bn Xn , a1 X1 + · · · + an Xn ]
(27.36)
= I1 X1 + I2 X2 + · · · + In Xn
where I1 , I2 , · · · , In are functions depending on ai , bi
Ik = Ik (a1 , · · · , an ; b1 , · · · , bn ) k = 1, 2, · · · , n (27.37)
Using (27.33) and (27.36), we obtain
Ad(exp(Y )X) = e−Y XeY

2
= X − [Y, X] + [Y, [Y, X]] − · · · (27.38)
2
= (a1 X1 + · · · + an Xn ) − (I1 X1 + · · · + In Xn ) + O(2 )
384
The quantities Ik , k = 1, 2, · · · , n can be written as
Ik = Ik1 (a1 , · · · , an )b1 + · · · + Ikn (a1 , · · · , an )bn (27.39)
The function Φ(a1 , a2 , · · · , an ) is an invariant, if and only if

∂Φ ∂Φ ∂Φ
I1 + I2 + · · · + In =0 (27.40)
∂a1 ∂a2 ∂an
Using the expressions (27.39), equation (27.40) takes on the form
∂Φ
(I11 b1 + I12 b2 + · · · + I1n bn )
∂a1
∂Φ
+ (I21 b1 + I22 b2 + · · · + I2n bn )
∂a2 (27.41)
+ ···+
∂Φ
+ (In1 b1 + In2 b2 + · · · + Inn bn ) =0
∂an

∂Φ ∂Φ ∂Φ
b1 I11 + I21 + · · · + In1
∂a1 ∂a2 ∂an

∂Φ ∂Φ ∂Φ
+ b2 I12 + I22 + · · · + In2
∂a1 ∂a2 ∂an (27.42)
+ ···+

∂Φ ∂Φ ∂Φ
+ bn I1n + I2n + · · · + Inn =0
∂a1 ∂a2 ∂an
Since the above equation should be true for any bk , k = 1, 2, · · · , n, we obtain

the following system of simultaneous PDEs
∂Φ ∂Φ ∂Φ
I11 + I21 + · · · + In1 =0
∂a1 ∂a2 ∂an
∂Φ ∂Φ ∂Φ
I12 + I22 + · · · + In2 =0
∂a1 ∂a2 ∂an (27.43)
···
∂Φ ∂Φ ∂Φ
I1n + I2n + · · · + Inn =0
∂a1 ∂a2 ∂an
385
Equations (27.43) determine the invariant Φ = Φ(a1 , a2 , · · · , an ).
27.2.2 The Adjoint Transformation Matrix. The entries (i, j) of the
Adjoint Transformation Table (called Table II) are obtained by expanding
Ad(exp(i Xi )Xj ) = e−i Xi Xj ei Xi (27.44)
taking into account the Commutator Table (Table I).
27.2.3 The General Transformation Matrix. Using the AdjointPTable,
we apply successively the adjoint action of X1 , X2 , · · · , Xn to X = ai Xi .
For example
Ad(exp(1 X1 )X) = Ad(exp(1 X1 )(a1 X1 + a2 X2 + · · · + an Xn ))
= a1 Ad(exp(1 X1 )X1 ) + a2 Ad(exp(1 X1 )X2 ) + · · · (27.45)
+ an Ad(exp(1 X1 )Xn )
Using the Adjoint Table, the above expression can be reduced to
Ad(exp(1 X1 )X) = (a1 , a2 , · · · , an )A1 (X1 , X2 , · · · , Xn )T (27.46)
Similarly we construct the matrices A2 , A3 , · · · , An as a result of separate
adjoint actions of X2 , X3 , · · · , Xn respectively. We then form the matrix A
which is the product of A1 , A2 , · · · , An :
A≡A(1 , 2 , · · · , n ) = A1 A2 · · ·An (27.47)
27.2.4 The Optimal System.To find an element of the optimal system,
we consider a representative element
n
X
Y = ãi Xi (27.48)
i=1
It is obvious that (ã1 , ã2 , · · · , ãn ) and (a1 , a2 , · · · , an ) are connected by

(ã1 , ã2 , · · · , ãn ) = (a1 , a2 , · · · , an )A (27.49)
where A is given in (27.47).
We come next to the crusial step of selecting the various ãi . First of all, we
have to find a general solution of the the system (27.43).
(a) If the general solution of the system depends on a single variable, say a1 ,
i.e.
Φ = F (a1 ) (27.50)
386
we may take
a1 = 1 or a1 = 0 (27.51)
For a1 = 1 we consider the representative element
Y = X1 (27.52)
We then replace
ã1 = 1 and ãk = 0 for k = 2, 3, · · · , n (27.53)
and solve the resulting system (27.49) with respect to 1 , · · · , n (in terms of
a1 , · · · , an ). If the system is consistent, then the choice Y = X1 is acceptable.
For a1 = 0, we consider the system of PDEs (27.43) and find the new invari-
ant of the resulting system. We then continue with the new values of the
coefficients, etc.
(b) If the general solution of the system (27.43) depends on two variables,
say a1 and a2 , i.e.
Φ = F (a1 , a2 ) (27.54)
we may consider
Φ = a1 a2 (27.55)
We consider the two cases: a1 a2 = 1 and a1 a2 = 0.

Let us first consider a1 a2 = 1. For a1 = 1 and a2 = 1, we consider
Y = X1 + X2 (27.56)
as a representative element. We then take ã1 = 1, ã2 = 1 and ãk = 0, for

k = 3, · · · , n and solve the system (27.49) with respect to 1 , · · · , n . If the
system is consistent, the choice Y = X1 + X2 is acceptable.
The choice a1 a2 = 0 leads to a number of choices like a1 = 1, a2 = 0 or
a1 = −1, a2 = 0 or a1 = 0, a2 = 0. In each of these cases we consider the
corresponding representative elements and examine the consistency of the
system (27.49).
There might be a number of even more complex expressions, which are
treated accordingly. The reader is advised to look carefully at the cited
387
literature.
27.2.5 Example. The KdV equation. Let us consider the KdV equation
ut + uux + uxxx = 0 (27.57)
which admits the following generators
X 1 = ∂x , X2 = ∂t , X3 = t∂x + ∂u , X4 = x∂x + 3t∂t − 2u∂u (27.58)
We first form the commutator table [Xi , Xj ] = Xi Xj − Xj Xi .
X1 X2 X3 X4
X1 0 0 0 X1
X2 0 0 X1 3X2
X3 0 −X1 0 −2X3
X4 −X1 −3X2 2X3 0
Using the above table with

4
X 4
X
X= ai X i , and Y = bi X i (27.59)
i=1 i=1
we find
Ad(exp(Y )X) = (a1 X1 + · · · + a4 X4 )
(27.60)
− (I1 X1 + I2 X2 + I3 X3 + I4 X4 ) + O(2 )
with
I1 = a4 b1 + a3 b2 − a2 b3 − a1 b4 , I2 = 3a4 b2 − 3a2 b4
(27.61)
I3 = −2a4 b2 + 2a3 b4 , I4 = 0
388
We then obtain the following system (along the lines of resoning in deriving
the system (27.43)) for the invariance Φ
∂Φ
a4 =0
∂a1
∂Φ ∂Φ
a3 + 3a4 =0
∂a1 ∂a2
(27.62)
∂Φ ∂Φ
a2 + 2a4 =0
∂a1 ∂a3
∂Φ ∂Φ ∂Φ
a1 + 3a2 − 2a3 =0
∂a1 ∂a2 ∂a3
Φ(a1 , a2 , a3 , a4 ) = F (a4 ) (27.63)
where F is an arbitrary function of a4 .
We next form the table of the adjoint transformation matrix
Ad X1 X2 X3 X4
X1 X1 X2 X3 X4 − X1
X2 X1 X2 X3 − X1 X4 − 3X2
X3 X1 X2 + X1 X3 X4 + 2X3
X4 e X1 e3 X2 e−2 X3 X4
Applying the adjoint action of X1 to

X = a1 X 1 + a2 X 2 + a3 X 3 + a4 X 4 (27.64)
we get
Ad(exp(1 X1 )(a1 X1 + a2 X2 + a3 X3 + a4 X4 ))
(27.65)
= (a1 − a4 1 )X1 + a2 X2 + a3 X3 + a4 X4
From the above equation we obtain the matrix
 
1 0 0 0
 0 1 0 0
A1 = 
 0 0 1 0
 (27.66)
−1 0 0 1
389
considering the corresponding equation (27.46).
Using similar procedure, we obtain the matrices A2 , A3 and A4
   
1 0 0 0 1 0 0 0
 0 1 0 0
 , A3 = 3 1 0 0 ,
 
A2 = 
−2 0 1 0  0 0 1 0
0 −32 0 1 0 0 23 1
  (27.67)
e4 0 0 0
 0 3e4 0 0
A4 = 
0 −2

0 e 4 0
0 0 0 1
respectively. Using the above matrices, we calculate the matrix A:

 
e4 0 0 0
 3 e4 e34 0 0
A = A1 A2 A3 A4 =  −2
 (27.68)
 −2 e 4
0 e 4
0
−(1 + 32 3 )e4 −32 e34 23 e−24 1
We have the following adjoint transformation equations
(ã1 , ã2 , ã3 , ã4 ) = (a1 , a2 , a3 , a4 )A (27.69)
We shall consider two cases: a4 = 1 and a4 = 0.

First Case. We consider the case with a4 = 1. We select a representative
X4 . Solving equations (27.70) for ã1 = 1, ã2 = 0, ã3 = 0, ã4 = 0, we get the
solution
1 1 1
1 = a1 − a2 a3 , 2 = a2 , 3 = − a3 (27.70)
3 3 2
which means that all the X4 + a1 X1 + a2 X2 + a3 X3 are equivalent to X4 .
Second Case. We consider the case with a4 = 0. Substituting a4 = 0 into
equations (27.63), leads to the new invariant
Φ(a1 , a2 , a3 ) = F (a22 a23 ) (27.71)
We take Φ = a22 a23 . We shall consider two cases: a22 a23 = 1 and a22 a23 = 0.
Let us first consider the case a22 a23 = 1. In this case we consider two repre-
sentative elements X2 + X3 and −X2 + X3 . For the representative element
390
X2 + X3 , equations (27.70) (for ã1 = 0, ã2 = 1, ã3 = 1, ã4 = 0) have the
solution
a1 1
2 = 0, 3 = − , 4 = − ln(a2 ) (27.72)
a2 3
For the representative element −X2 + X3 , equations (27.70) (for ã1 = 0, ã2 =
−1, ã3 = 1, ã4 = 0) have the solution
a1 1
2 = 0, 3 = − , 4 = − ln(−a2 ) (27.73)
a2 3
Let us consider next the case a22 a23 = 0.

(1) Let a3 6=0 and a2 = 0. We then consider two representative elements X3
and −X3 . Then a1 X1 +X3 is equivalent to X3 where the solution of equations
(27.70) (for ã1 = 0, ã2 = 0, ã3 = 1, ã4 = 0) is
a1 1
2 = , 4 = ln(a3 ) (27.74)
a3 2
while a1 X1 − X3 is equivalent to −X3 where the solution of equations (27.70)
(for ã1 = 0, ã2 = 0, ã3 = −1, ã4 = 0) is
a1 1
2 = , 4 = ln(−a3 ) (27.75)
a3 2
However X3 and −X3 are essentially equivalent.
(2) Let a3 = 0 and a2 6=0. For a2 6=0, we see that a1 X1 + a2 X2 is equivalent
to X2 and −X2 . For a3 = 0 and a2 = 0, a1 X1 is equivalent to X1 .
Collecting everything together, we find that the optimal system of Lie algebra
generators for the KdV equation is given by
X4 , X2 + X3 , −X2 + X3 , X3 , X2 , X1 (27.76)
which is exactly the same optimal system found previously in Section (27.1).
For 2-dimensional optimal systems, the reader can consult Ref. [5].
It is worth mentioning that there is also an algorithmic method of deriving
the invariants, once the commutation relations of the generators are given.
In this case one has to solve the so-called Jacobi system (Ref. [17], equation
(16)).
391
References
[1] L. V. Ovsiannikov: ”Group Analysis of Differential Equations”.
Elsevier 1982
[2] P. J. Olver: ”Applications of Lie Groups to Differential Equations”.
Springer 1986
[3] K-S. Chou, G-X. Li and C. Qu: ”A Note on Optimal Systems for
the Heat Equation”. J. Math. Anal. and Appl. 261 (2001) 741-751
[4] X. Hu, Y. Li and Y. Chen: ”A direct algorithm of one-dimensional
optimal system for the group invariant solutions”. arXiv: 1411.3797v1
[5] X. Hu, Y. Li and Y. Chen: ”The construction of two-dimensional
optimal system for the invariant solutions”. arXiv: 1411.3798v1
[6] N. Benoudina, Y. Zhang and C. M. Khalique: ”Lie symmetry
analysis, optimal system, new solitary wave solutions and conservation laws
of the Pavlov equation”. Oct. 2020 preprint
[7] S. Al-Omari, F. Zaman and H. Azad: ”Lie Symmetries, Optimal
System and Invariant Reductions to a Nonlinear Timoshenko System”.
Mathematics 5 (2017) 34
[8] M. Mu and C. Temuer: ”Lie Symmetries, 1-Dimensional Optimal Sys-
tem and Optimal Reductions of (1+2)-Dimensional Nonlinear Schrödinger
Equation”. J. App. Math. Phys. 2 (2014) 603-620
[9] W. Zhao, M. Munir, G. Murtaza and M. Athar: ”Lie symmetries
of Benjamin-Ono Equation”. Math. Bios. Eng. 18 (2021) 9496-9510
[10] A. Paliathanasis: ”Lie symmetry analysis and one-dimensional opti-
mal system for the generalized 2+1 Kadomtsev-Petviashvili equation”.
arXiv:2002.11585v1
[11] M. Nadjafikhah and F. Ahangari: ”Symmetry Analysis and Con-
servation Laws for the Hunter-Saxton Equation”.
Comm. Theor. Phys. 59 (2013) 335-348
[12] M. Nadjafikhah and F. Ahangari: ”Symmetry Reduction of Two-
Dimensional Damped Kuramoto-Sivashinsky Equation”.
Comm. Theor. Phys. 56 (2011) 211-217
[13] S. Kumar and S. K. Dhiman: ”Lie Symmetry Analysis, optimal
system, exact solutions and dynamics of solitons of a (3+1)-dimensional gen-
eralized BKP-Boussinesq Equation”. Pramana (2022) 96:31
[14] S. Kumar, W-X Ma and A. Kumar: ”Lie Symmetries, optimal
system and group-invariant solutions of the (3+1)-dimensional generalized
KP Equation”. Chinese J. Phys. 2020
392
[15] H. S. Tang, K. Y. Tong and B. H. Kee: ”Lie Symmetries, Optimal
System and Invariant Solutions of the Generalized Cox-Ingersoll-Ross Equa-
tion”. N. Wahi et al (Eds) ICMSS 2022, ACSR 98, pp. 103-113, 2023
[16] A. Tiwari, K. Sharma and R. Arora: ”Lie symmetry analysis,
optimal system and new exact solutions of a (3+1) dimensional nonlinear
evolution equation”. Nonlin. Eng. 10 (2021) 132-145
[17] J. N. Pecina-Cruz: ”An Algorithm to calculate the Invariants of any
Lie Algebra”. J. Math. Phys. 35 (1994) 3146-3162
Chapter X
Potential Symmetries
28 Potential Symmetries
Potential Symmetries is another type of symmetries for evolution type of
PDEs. Let us consider an evolution equation
ut = F (x, u, ux , uxx , · · · )
written in conserved form
ut = Φx
Introducing a potential function v by vx = u, equation ut = Φx is being
converted into the system
vx = u, vt = Φ
This is due to the fact that the compatibility condition vxt = vtx is equivalent
to the equation ut = Φx .
Let X be the generator of symmetries of the above system {vx = u, vt = Φ}
∂ ∂ ∂ ∂
X = ξ1 + ξ 2 + η1 + η2
∂x ∂t ∂u ∂v
Potential symmetries occur only if ξ , ξ 2 , η 1 have a non-trivial dependence
1
on v, i.e. if
∂ξ 1 2 ∂ξ 2 2 ∂η 1 2
+ + 6=0
∂v ∂v ∂v
393
For example, the Burgers equation ut + uux − uxx = 0 can be written as
1 2
ut = ux − u
2 x
Introducing the potential function v by vx = u, the above equation is con-

verted into the system
1
vx = u, vt = ux − u2
2
The procedure will become evident through an Example.
Example. We consider the Fokker-Planck equation
ut = uxx + xux + u (28.1)
First of all, we write this equation in conserved form
ut = (ux + xu)x (28.2)
We introduce a potential function v by vx = u. Equation (28.2) can thus be

reduced to the system
vx = u
(28.3)
vt = ux + xu
We determine next the Lie point symmetries of the above system. Let us
introduce the symmetry generator
∂ ∂ ∂ ∂
X=ξ +τ +η +φ (28.4)
∂x ∂t ∂u ∂v
Invariance of the system under Lie point transformations is being expressed
by the conditions
P r(1) X[∆1 ] = 0 and P r(1) X[∆2 ] =0 (28.5)

∆1 =0, ∆2 =0 ∆1 =0, ∆2 =0
where P r(1) X is the first prolongation of X, while ∆1 and ∆2 are given by
∆1 ≡ vx − u, ∆1 ≡ vt − ux − xu (28.6)
394
Applying the conditions (28.5), we find the following generators of Lie point
symmetries
∂ ∂ ∂ ∂
X 1 = −x e2t − e2t + (ux2 + 2xv + 2u)e2t + v(x2 + 1)e2t
∂x ∂t ∂u ∂v
∂ ∂ ∂
X2 = et − (ux + v)et − vxet
∂x ∂u ∂v
∂
X3 = e−t
∂x (28.7)
−2t ∂ −2t ∂ −2t ∂
X4 = −x e +e + ue
∂x ∂t ∂u
∂
X5 =
∂t
∂ ∂
X6 =u +v
∂u ∂v
and an additional infinite-dimensional symmetry due to the linearity of the
system. Only X 1 and X 2 exhibit potential symmetries due to the non-trivial
dependence on v:
∂ξ 2 ∂τ 2 ∂η 2
+ + 6= 0 (28.8)
∂v ∂v ∂v
Let us consider the generator X 1 and the corresponding invariant surface
conditions
ξux + τ ut − η = 0, ξvx + τ vt − φ = 0 (28.9)
written as
xux + ut + ux2 + 2vx + 2u = 0
(28.10)
xvx + vt + v(x2 + 1) = 0
F (− ln(x) + t) x2
v(x, t) = exp −
x 2
x2 G(− ln(x) + t) x2 (28.11)
u(x, t) = −F (− ln(x) + t)exp − + exp −
2 x2 2
Introducing the similarity variable z by
1 t
z = e− ln(x)+t = e (28.12)
x
395
the above solution of the system has the form
F (z) x2
v(x, t) = exp −
x 2
x2 (28.13)
G(z)
u(x, t) = − F (z) + 2 exp −
x 2
Upon substitution of u(x, t) given by the second expression of (28.13), into

the original Fokker-Planck equation (28.1), we obtain the equation
− {z 2 G00 (z) + 6zG0 (z) + 6G(z)}

(28.14)
+ x2 {z 2 F 00 (z) + 2zF 0 (z) − 2G(z)} = 0
z 2 G00 (z) + 6zG0 (z) + 6G(z) = 0

(28.15)
z 2 F 00 (z) + 2zF 0 (z) − 2G(z) = 0

1 A1 A2 B1
F (z) = + + B2 −
3 z3 z2 z (28.16)
A1 A2
G(z) = 3 + 2
z z
The above expressions, when substituted to the second of (28.13), and taking
1
into account that z = et , give
x

u = 3B1 xe−t − 3B2
x2 (28.17)
3 −3t 2 −2t
+ A1 (3x − x )e + 3A2 (1 − x )e exp −
2
After setting B1 = B2 = 0 and making the necessary simplifications, we get
x2
u = A1 (3x − x3 )e−t + 3A2 (1 − x2 ) exp −2t − (28.18)
2
Both solutions (28.17) and (28.18) satisfy the original Fokker-Planck equa-
tion.
396
References
[1] E. Pucci and G. Saccomandi: ”Potential symmetries and solutions
by reduction of partial differential equations”. J. Phys. A 26 (1993) 681-690
[2] M. L. Gandarias: ”New Potential Symmetries for some evolution equa-
tions”. Physica A 387 (2008) 2234-2242
[3] X. M. Wang, S. Bilige and Y. X. Bai: ”Potential Symmetries, Lie
Transformation Groups and Exact Solutions of KdV-Burgers Equation”.
Discontinuity, Nonlinearity, and Complexity 6 (2017) 1-9
Chapter XI
Generalized Symmetries
So far we have only considered Lie point symmetries. In this part we shall
consider Lie-Bäcklund Symmetries which are also called Generalized
Symmetries where the infinitesimals will depend on partial derivatives of
the unknown function(s).
29 Lie-Bäcklund Symmetries
We say that an evolution equation of the form
ut = F (x, u, ux , uxx , · · · ) (29.1)
admits a Lie-Bäcklund Symmetry if the generator X of the symmetry can

take the form
∂
X = Q(t, x, u, ux , uxx , · · · ) (29.2)
∂u
In order to apply the method, we have to consider a fixed order of symme-
try, say third order. In that case we take Q = Q(t, x, u, ux , uxx , uxxx ). For
example invariance of the potential form of Burgers equation ut = uxx + u2x
is expressed by the condition Dt Q = Dx2 Q + 2ux Dx Q.
The operators Dx and Dt will be taken as
∂ ∂ ∂ ∂ ∂
Dx = + ux + uxx + uxxx + uxxxx + ··· (29.3)
∂x ∂u ∂ux ∂uxx ∂uxxx
397
and
∂ ∂ ∂ ∂ ∂
Dt = + ut + utx + utxx + utxxx + ··· (29.4)
∂t ∂u ∂ux ∂uxx ∂uxxx
respectively. In the above operator expressions we have disregarded partial
∂ ∂
derivatives of the type , , etc. since we suppose that Q has the de-
∂ut ∂utx
pendence Q = Q(t, x, u, ux , uxx , · · · ). We have however included derivatives
of the type ut , utx , etc. in the coefficients of the operators (29.3) and (29.4),
since they can be expressed in terms of the partial derivatives ux , uxx , etc.
considering the evolution equation.
We shall also need the operator Dx2 . A simple calculation gives
∂2 ∂
Dx2 = (uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
2
∂ ∂2 ∂2 ∂2
+ uxxxx 2 + 2ux + 2uxx + 2uxxx
∂x∂uxxx ∂u∂uxxx ∂ux ∂uxxx ∂uxx ∂uxxx
2
∂ ∂
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
2
∂ ∂2 ∂2
+ uxxx 2 + 2ux + 2uxx
∂x∂uxx ∂u∂uxx ∂ux ∂uxx
2 2
2 ∂ ∂ ∂ ∂2
+ (uxx ) + uxx + uxx 2 + 2ux
∂u2x ∂u ∂x∂ux ∂u∂ux
2 2
∂ ∂ ∂ ∂2
+ (ux )2 2 + ux 2 + uxxxxx + 2
∂u ∂x∂u ∂uxxx ∂x
(29.5)
30 Lie-Bäcklund for the Burgers Equation in

Potential Form
Example 1. The Burgers Equation. Potential Form.
As our first Example we consider the potential form of Burgers equation
ut = uxx + u2x (30.1)
We shall consider third order symmetries. Invariance is being expressed by
the condition
Dt Q = Dx2 Q + 2ux Dx Q (30.2)
398
where
Q = Q(x, t, u, ux , uxx , uxxx ) (30.3)
Using the explicit form of these operators (Dx2 is given by (29.5), while Dx
and Dt by (29.3) and (29.4) respectively), the invariance condition of the
given equation takes on the form
∂ 2Q ∂Q
(uxxxx )2 + u xxxx
∂u2xxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
2 ∂ 2Q
∂ Q ∂Q ∂ 2Q
+ (uxx )2 2 + uxx + uxx 2 + 2ux
∂ux ∂u ∂x∂ux ∂u∂ux
∂ 2Q ∂ 2Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +
∂u ∂x∂u ∂uxxx ∂x2

∂Q ∂Q ∂Q ∂Q ∂Q
+ 2ux + ux + uxx + uxxx + uxxxx

∂Q ∂Q ∂Q ∂Q ∂Q
− + ut + utx + utxx + utxxx =0
(30.4)
We now have to express ut , utx , utxx and utxxx in terms of partial derivatives
with respect to x. We obtain by differentiation of the Burgers equation
ut = uxx + u2x with respect to x that
utx = uxxx + 2ux uxx (30.5)
utxx = uxxxx + 2(uxx )2 + 2ux uxxx (30.6)

utxxx = uxxxxx + 6uxx uxxx + 2ux uxxxx (30.7)
399
∂ 2Q ∂Q
(uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
2 2
∂ Q ∂Q ∂ Q ∂ 2Q
+ (uxx )2 2 + uxx + uxx 2 + 2ux
2 2
∂ Q ∂ Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +

∂Q ∂Q ∂Q ∂Q ∂Q
+ 2ux + ux + uxx + uxxx + uxxxx

∂Q ∂Q ∂Q
− + (uxx + u2x ) + (uxxx + 2ux uxx )
∂t ∂u ∂ux
∂Q
+ (uxxxx + 2(uxx )2 + 2ux uxxx )
∂uxx

∂Q
+ (uxxxxx + 6uxx uxxx + 2ux uxxxx ) =0
∂uxxx
(30.8)
400
Upon expanding, canceling opposite terms and rearranging, the previous
equation can be written in equivalent form
∂ 2Q
(uxxxx )2
∂u2xxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q
+ (uxxx )2 2
∂uxx
∂ 2Q ∂ 2Q ∂ 2Q
2 2 2
2∂ Q ∂ Q ∂ Q

+ (uxx ) + uxx 2 + 2ux
∂u2x ∂x∂ux ∂u∂ux
2 2 2
∂ Q ∂ Q ∂ Q
+ (ux )2 2 + ux 2 +
∂u ∂x∂u ∂x2
∂Q ∂Q
+ 2ux + (u2x )
∂x ∂u
∂Q 2 ∂Q ∂Q
− + 2(uxx ) + (6uxx uxxx ) =0
∂t ∂uxx ∂uxxx
(30.9)
We equate first the coefficients of (uxxxx )2 and uxxxx to zero:
∂ 2Q
=0 (30.10)
∂u2xxx
and
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2 + 2ux + 2uxx + 2uxxx = 0 (30.11)
We thus see from (30.10) that Q should be a linear function with respect to
uxxx . However the coefficient of uxxx should be a function of t only, since
considering the coefficients of uxxxx , uxxxx ux , uxxxx uxx and uxxxx uxxx appear-
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
ing in (30.9), i.e. , , and respectively,
we conclude that this expression should be independent of x, u, ux and uxx .
We thus have the following expression for Q:
Q = K(t)uxxx + B(x, t, u, ux , uxx ) (30.12)
401
We can now ignore the first two lines of equation (30.9). The rest of the
equation is
2 ∂ 2Q
(uxxx )
∂u2xx
∂ 2Q ∂ 2Q ∂ 2Q
2 2 2
2∂ Q ∂ Q ∂ Q

+ (uxx ) + uxx 2 + 2ux
∂u2x ∂x∂ux ∂u∂ux (30.13)
2 2 2
∂ Q ∂ Q ∂ Q
+ (ux )2 2 + ux 2 +
∂u ∂x∂u ∂x2
∂Q ∂Q
+ 2ux + (u2x )
∂x ∂u
∂Q 2 ∂Q ∂Q
− + 2(uxx ) + (6uxx uxxx ) =0
∂t ∂uxx ∂uxxx
We now consider equation (30.13) with Q given by (30.12).

From equation (30.13) we see that the coefficient of (uxxx )2 should be zero,
∂ 2Q ∂ 2B
= 0, i.e. = 0, which means that B is a linear function with
∂u2xx ∂u2xx
respect to uxx : B = b(x, t, u, ux )uxx + C(x, t, u, ux ). Equating on the other
∂ 2Q ∂b
hand the coefficient of uxxx ux to zero, i.e. = = 0, we see that b
∂u∂uxx ∂u
should be independent of u and thus
B = L(x, t, ux )uxx + C(x, t, u, ux ) (30.14)
which means that
Q = K(t)uxxx + L(x, t, ux )uxx + C(x, t, u, ux ) (30.15)
∂ 2Q ∂Q ∂L
The coefficient of uxxx uxx is equal to 2 −6 =2 − 6K(t),
∂ux ∂uxx ∂uxxx ∂ux
which should be zero. i.e.
∂L
− 3K(t) = 0 (30.16)
∂ux
We have next to equate to zero the coefficient of uxxx in equation (30.13)
taking into account the expression (30.15) for Q. This coefficient is equal to
402
∂ 2Q ∂K ∂L ∂K
2 − =2 − which should be equated to zero, and thus we
∂x∂uxx ∂t ∂x ∂t
have another differential equation for L:
∂L ∂K
2 − =0 (30.17)
∂x ∂t
The system of (30.16) and (30.17) admits a solution given by
1
L = 3K(t)ux + K 0 (t)x + F (t) (30.18)
2
Taking into account the above expression of L, the expression of Q given by
(30.15), takes on the form
1
Q = K(t)uxxx +3K(t)ux uxx + K 0 (t)x+F (t) uxx +C(x, t, u, ux ) (30.19)
2
We can now ignore the first two lines of equation (30.13) and also the last
term within the square brackets (which we have already taken into account).
We will then have the equation
2 ∂ 2Q
2∂ Q ∂ 2Q
(uxx ) + uxx 2 + 2ux
2 2 2
∂ Q ∂ Q ∂ Q
+ (ux )2 2 + ux 2 +
∂u ∂x∂u ∂x2 (30.20)
∂Q ∂Q
+ 2ux + (u2x )
∂x ∂u
∂Q ∂Q
− + 2(uxx )2 =0
∂t ∂uxx
with Q given by (30.19).
∂ 2Q ∂Q ∂ 2C
The coefficient of (uxx )2 in (30.20) is equal to − 2 = − 2L,
∂u2x ∂uxx ∂u2x
(with L given by (30.18)) which should be zero, i.e.
∂ 2C
= 6K(t)ux + K 0 (t)x + 2F (t) (30.21)
∂u2x
Upon integration twice with respect to ux , we obtain from the above equation
1
C = K(t)u3x + {K 0 (t)x + 2F (t)}u2x + G(x, t, u)ux + H(x, t, u) (30.22)
2
403
We thus have the following expression for Q, using (30.19) and (30.22):
1
0
Q = K(t)uxxx + 3K(t)ux uxx + K (t)x + F (t) uxx
2 (30.23)
3 1 0 2
+ K(t)ux + {K (t)x + 2F (t)}ux + G(x, t, u)ux + H(x, t, u)
2
From equation (30.20) we ignore terms we have taken into account and we
obtain the equation
∂ 2Q ∂ 2Q
uxx 2 + 2ux
∂x∂ux ∂u∂ux
2
2∂ Q
∂ 2Q ∂ 2Q (30.24)
+ (ux ) + ux 2 +
∂u2 ∂x∂u ∂x2
∂Q ∂Q ∂Q
+ 2ux + (u2x ) − =0
∂x ∂u ∂t
Taking into account (30.23) and ignoring the term uxxx which appears in the
∂Q
derivative (since we have already taken into account that term), equation
∂t
(30.24) becomes
∂ 2H ∂H ∂G ∂G 1
00 0
− + 2 u u
x xx + 2 − K (t)x − F (t) uxx
∂x2 ∂t ∂u ∂x 2
∂ 2H ∂ 2G ∂H ∂G
+ 2 + +2 − ux
∂x∂u ∂x2 ∂x ∂t (30.25)
∂ 2 H ∂H ∂ 2G ∂G 1 00
0 2
+ + +2 +2 − K (t)x − F (t) ux
∂u2 ∂u ∂x∂u ∂x 2
∂ 2 G ∂G
+ 2
+ u3x = 0
∂u ∂u
Equating to zero the coefficients of the partial derivatives of the function u,
we obtain from (30.25) the following system of PDEs
∂ 2 H ∂H
− =0 (30.26)
∂x2 ∂t
∂G
=0 (30.27)
∂u
∂G 1 00
2 − K (t)x − F 0 (t) = 0 (30.28)
∂x 2
404
∂ 2H ∂ 2G ∂H ∂G
2 + + 2 − =0 (30.29)
∂x∂u ∂x2 ∂x ∂t
∂ 2 H ∂H ∂ 2G ∂G 1 00
2
+ + 2 +2 − K (t)x − F 0 (t) = 0 (30.30)
∂u ∂u ∂x∂u ∂x 2
∂ 2 G ∂G
+ =0 (30.31)
∂u2 ∂u
From (30.27) we see that G does not depend on u. Therefore equation (30.31)
is satisfied automatically. Equation (30.30), taking into account (30.28) and
the fact that G is independent of u, takes on the form
∂ 2 H ∂H
+ =0 (30.32)
∂u2 ∂u
H = N (x, t)e−u + P (x, t) (30.33)
Integrating equation (30.28) with respect to x, we find the general expression

of the function G = G(x, t):
1 1
G(x, t) = K 00 (t)x2 + F 0 (t)x + M (t) (30.34)
8 2
∂P 1 1 1 1
= K 000 (t)x2 + F 00 (t)x − K 00 (t) + M 0 (t) (30.35)
∂x 16 4 8 2
Integration with respect to x, the above equation yields
1 000 1 1
P (x, t) = K (t)x3 + F 00 (t)x2 + {4M 0 (t) − K 00 (t)}x + R(t) (30.36)
48 8 8
Equation (30.26), because of (30.33), becomes
∂ 2 N −u ∂ 2 P ∂N −u ∂P
2
e + 2
= e + (30.37)
∂x ∂x ∂t ∂t
From the above equation we see that
∂ 2N ∂N
2
= (30.38)
∂x ∂t
405
i.e. the function N is any solution of the heat equation, while the function
P satisfies the similar equation
∂ 2P ∂P
2
= (30.39)
∂x ∂t
The above equation, because of (30.36), takes on the form
1 (iv) 1 1
K (t)x3 + {F 000 (t) − K 000 (t)}x2 + {4M 00 (t) − K 000 (t)}x
48 8 8 (30.40)
1
+ {4R0 (t) − F 00 (t)} = 0
4
Since the previous equation is true for every x, we obtain the following system
of ODEs
K (iv) (t) = 0, F 000 (t) − K 000 (t) = 0, 4M 00 (t) − K 000 (t) = 0,
(30.41)
4R0 (t) − F 00 (t) = 0
From equation K (iv) (t) = 0, we obtain
K(t) = a1 t3 + a2 t2 + a3 t + a4 (30.42)
Equation F 000 (t) − K 000 (t) = 0 because of (30.42), is equivalent to F 000 (t) = 6a1
and by integration
1
F (t) = a1 t3 + a5 t2 + a6 t + a7 (30.43)
2
3
Equation 4M 00 (t) − K 000 (t) = 0 is equivalent to M 00 (t) = a1 and by integra-
2
tion
3
M (t) = a1 t2 + a8 t + a9 (30.44)
4
1
Equation 4R0 (t) − F 00 (t) = 0 is equivalent to R0 (t) = (6a1 t + a5 ) and by
4
integration
3 1
R(t) = a1 t2 + a5 t + a10 (30.45)
4 4
406
Using the expressions for K(t), F (t) and M (t), we find that the function
G(x, t), using (30.34), is given by
1 1 3
G(x, t) = (3a1 t + a2 )x2 + (3a1 t2 + a5 t + a6 )x + a1 t2 + a8 t + a9 (30.46)
4 2 4
Using (30.36), we can determine the function P (x, t):
1 1 1
P (x, t) = a1 x3 + (6a1 t + a5 )x2 + (2a8 − a2 )x
8 8 4 (30.47)
3 2 1
+ a1 t + a5 t + a10
4 4
We are now ready to determine the characteristic function Q. This function
is given by (30.23). Using the expressions for the functions K(t), F (t), G(x, t)
and H(x, t, u) (given by (30.33) and (30.47)), we obtain the following expres-
sion for Q:
Q = (a1 t3 + a2 t2 + a3 t + a4 )uxxx + 3(a1 t3 + a2 t2 + a3 t + a4 )ux uxx

1 1 2
2 3
+ (3a1 t + 2a2 t + a3 )x + a1 t + a5 t + a6 t + a7 uxx
2 2
+ (a1 t3 + a2 t2 + a3 t + a4 )u3x
1 1
+ (3a1 t2 + 2a2 t + a3 )x + a1 t3 + a5 t2 + a6 t + a7 u2x
2 2
1 1
+ (3a1 t + a2 )x2 + (3a1 t2 + a5 t + a6 )x (30.48)
4 2
3
+ a1 t 2 + a8 t + a9 u x
4
1 1 1
+ a1 x3 + (6a1 t + a5 )x2 + (2a8 − a2 )x
8 8 4
3 2 1
+ a1 t + a5 t + a10 + N (x, t)e−u
4 4
407
where N (x, t) is any solution of the heat equation.

3
Q = a1 t3 (uxxx + 3ux uxx + u3x ) + t2 x(uxx + u2x )
2
3
2 3 2 3 2 1
3 3 2 3 2
+ tx + t x + t ux + x + tx + t
4 2 4 8 4 4
1
2 3 2 1 2
+ a2 t (uxxx + 3ux uxx + ux ) + tx(uxx + ux ) + t + x ux
2 4

1
+ a3 t(uxxx + 3ux uxx + u3x ) + x(uxx + u2x )
2 (30.49)

+ a4 uxxx + 3ux uxx + u3x

1 2 2 1 1
2 1
+ a5 t (uxx + ux ) + txux + x + t
2 2 8 4

2 1 2 1
+ a6 t(uxx + ux ) + xux + a7 uxx + ux + a8 tux + x
2 2
−u
+ a9 ux + a10 + N (x, t)e
From the above expression we see that the generators of the Lie-Bäcklund
symmetries are given by
3
Q[1] = t3 (uxxx + 3ux uxx + u3x ) + t2 x(uxx + u2x )
2 (30.50)
3 3 2 3 2 1 3 2 3 2
2 3
+ tx + t x + t ux + x + tx + t
4 2 4 8 4 4
1 1
Q[2] = t2 (uxxx + 3ux uxx + u3x ) + tx(uxx + u2x ) + t + x2 ux (30.51)
2 4
1
Q[3] = t(uxxx + 3ux uxx + u3x ) + x(uxx + u2x ) (30.52)
2
3
Q[4] = uxxx + 3ux uxx + ux (30.53)
1 2 2 1 1
2 1
Q[5] = t (uxx + ux ) + txux + x + t (30.54)
2 2 8 4
1
Q[6] = t(uxx + u2x ) + xux (30.55)
2
2
Q[7] = uxx + ux (30.56)
408
1
Q[8] = tux + x (30.57)
2
Q[9] = ux (30.58)
Q[10] = 1 (30.59)
and the infinite-dimensional subalgebra
Q[N ] = N (x, t)e−u (30.60)
where N (x, t) is any solution of the heat equation.
Let us now find some general solutions of the potential form of Burgers
equation using the generators of the Lie-Bäcklund symmetries.
Solution 1. We shall find the general solution of Burgers equation using
Q[6]. We thus first solve equation
1
t(uxx + u2x ) + xux = 0 (30.61)
2
with respect to u. The general solution of equation (30.61) is given by
√
x
u(x, t) = ln F (t) πt erf √ + G(t) (30.62)
2 t
where erf(x) is the known error function. Upon substitution of the above
expression into Burgers equation ut = uxx + u2x , we obtain the equation
√ d F (t) √ x d
t F (t) + √ πt erf √ + G(t) = 0 (30.63)
dt 2 t 2 t dt
From the above equation, we get the system of two ordinary differential
equations
√ d F (t) d
t F (t) + √ = 0, G(t) = 0 (30.64)
dt 2 t dt
C1
F (t) = √ , G(t) = C2 (30.65)
t
Therefore we obtain from (30.62), the general solution of Burgers equation
(30.1)
√
x
u(x, t) = ln C1 π erf √ + C2 (30.66)
2 t
409
Solution 2. We find next the general solution of Burgers equation using
Q[5]. We thus consider the equation
1 2 1 1 1
t (uxx + u2x ) + txux + x2 + t = 0 (30.67)
2 2 8 4
x2
u(x, t) = − + ln{F (t) − x G(t)} (30.68)
4t
Upon substitution of the above expression into Burgers equation (30.1), i.e.
ut = uxx + u2x , we obtain the equation

d d
F (t) + 2t F (t) − x 3G(t) + 2t G(t) = 0 (30.69)
dt dt
From the above equation we obtain the system of two ODEs

d d
F (t) + 2t F (t) = 0, 3G(t) + 2t G(t) = 0 (30.70)
dt dt
from which we get the solution
C1 C2
F (t) = √ , G(t) = √ (30.71)
t t t
We then obtain, using (30.68), the general solution of Burgers equation
x2 C
1 C2
u(x, t) = − + ln √ − x √ (30.72)
4t t t t
Solution 3. We find next the general solution of Burgers equation using
Q[3]. We thus consider the equation
1
t(uxxx + 3ux uxx + u3x ) + x(uxx + u2x ) = 0 (30.73)
2
The substitution w = ux converts (30.73) to the equation
1
t(wxx + 3wwx + w3 ) + x(wx + w2 ) = 0 (30.74)
2
410
Hx
w(x, t) = (30.75)
H
where
√ x x2
H≡H(x, t) = x πt F (t) erf √ + 2tF (t)exp − + x G(t) + 1 (30.76)
2 t 4t
We thus have
u(x, t) = ln(H(x, t)) (30.77)
Upon substitution of the above expression into Burgers equation ut = uxx +

u2x , we obtain the equation
√ x2

d x d
x πt F (t) + 2t F (t) erf √ + 2t F (t) + 2t F (t) exp −
dt 2 t dt 4t
d
+ 2xt G(t) = 0
dt
From the above equation we obtain the system of two ODEs
d d
F (t) + 2t F (t) = 0, G(t) = 0 (30.78)
dt dt
from which we get the solution
C1
F (t) = √ , G(t) = C2 (30.79)
t
We then obtain, using (30.77) and (30.76), the general solution of Burgers
equation
√ x 2 x2
u(x, t) = ln C1 πx erf √ + C1 √ exp − + C2 x + 1 (30.80)
2 t t 4t
31 Lie-Bäcklund for the Burgers Equation

Example 2. The Burgers Equation. We consider next the Burgers
equation
ut + uux − uxx = 0 (31.1)
411
Invariance is being expressed in the form
Dx2 Q − uDx Q − ux Q − Dt Q = 0 (31.2)
where
Q = Q(x, t, u, ux , uxx , uxxx ) (31.3)
We shall determine again the third order symmetries of the Burgers equa-
tion. Using the explicit form of the operators, as in previous Example, equa-
∂ 2Q ∂Q
(uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q
+ uxxxx 2 + 2ux + 2uxx
∂x∂uxxx ∂u∂uxxx ∂ux ∂uxxx
2
∂ Q
+ 2uxxx
∂uxx ∂uxxx
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
∂x∂uxx ∂u∂uxx ∂ux ∂uxx (31.4)
2 2
∂ Q ∂Q ∂ Q ∂ 2Q
+ (uxx )2 2 + uxx + uxx 2 + 2ux
2 2
∂ Q ∂ Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +

∂Q ∂Q ∂Q ∂Q ∂Q
−u + ux + uxx + uxxx + uxxxx
− ux Q

∂Q ∂Q ∂Q ∂Q ∂Q
with respect to x. Using the Burgers equation in the form
ut = uxx − uux (31.5)
and differentiation successively with respect to x, we obtain the following
identities
utx = uxxx − (ux )2 − uuxx (31.6)
412
utxx = uxxxx − 3ux uxx − uuxxx (31.7)
utxxx = uxxxxx − 3(uxx )2 − 4ux uxxx − uuxxxx (31.8)
Because of the above identities, equation (31.4) acquires the form
∂ 2Q ∂Q
(uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q
+ 2uxxx
∂uxx ∂uxxx
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
2 2
2∂ Q ∂Q ∂ Q ∂ 2Q
+ (uxx ) + uxx + uxx 2 + 2ux
∂u2x ∂u ∂x∂ux ∂u∂ux (31.9)
2 2
∂ Q ∂ Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +

∂Q ∂Q ∂Q ∂Q ∂Q
−u + ux + uxx + uxxx + uxxxx
− ux Q

∂Q ∂Q ∂Q
− + (uxx − uux ) + (uxxx − (ux )2 − uuxx )
∂t ∂u ∂ux
∂Q
+ (uxxxx − 3ux uxx − uuxxx )
∂uxx

2 ∂Q
+ (uxxxxx − 3(uxx ) − 4ux uxxx − uuxxxx ) =0
∂uxxx
413
Upon expanding, canceling opposite terms and rearranging, the above equa-
tion takes on the equivalent form
∂ 2Q
(uxxxx )2
∂u2xxx
∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q
+ 2uxxx
∂uxx ∂uxxx
2
∂ Q
+ (uxxx )2 2
∂uxx (31.10)
∂ 2Q ∂ 2Q ∂Q ∂ 2Q
+ uxxx 2 + 2ux + 4ux + 2uxx
∂x∂uxx ∂u∂uxx ∂uxxx ∂ux ∂uxx
∂ 2Q ∂Q 2
∂ Q 2
∂ Q ∂Q
+ (uxx )2 + 3 + u xx 2 + 2u x + 3ux
∂u2x ∂uxxx ∂x∂ux ∂u∂ux ∂uxx
∂ 2Q ∂Q 2
∂ Q 2
∂ Q
+ (ux )2 2
+ + ux 2 +
∂u ∂ux ∂x∂u ∂x2
∂Q ∂Q
−u − ux Q − =0
∂x ∂t
Since the above should be true for every u, etc. the coefficients of (uxxxx )2 and
uxxxx should vanish separately. We then obtain the following two equations
∂ 2Q
=0 (31.11)
∂u2xxx
and
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
+ ux + uxx + uxxx =0 (31.12)
From (31.11) we conclude that Q is a linear expression with respect to uxxx .
However the coefficient of uxxx should be a function of t only, since considering
the coefficients of uxxxx , uxxxx ux , uxxxx uxx and uxxxx uxxx appearing in (31.10),
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
i.e. , , and respectively, we see that
this coefficient should be independent of x, u, ux and uxx . We thus have the
following expression for Q:
414
Let us now ignore the first two lines of equation (31.10). We then have the
equation
∂ 2Q
(uxxx )2
∂u2xx
∂ 2Q ∂ 2Q ∂Q ∂ 2Q
+ uxxx 2 + 2ux + 4ux + 2uxx
∂x∂uxx ∂u∂uxx ∂uxxx ∂ux ∂uxx
∂ 2Q ∂Q ∂ 2Q ∂ 2
Q ∂Q (31.14)
+ (uxx )2 + 3 + u xx 2 + 2ux + 3ux
∂ 2Q ∂Q 2
∂ Q 2
∂ Q
+ (ux )2 2
+ + ux 2 +
∂u ∂ux ∂x∂u ∂x2
∂Q ∂Q
−u − ux Q − =0
∂x ∂t
with Q given by (31.13). From the first line of the above equation we see
∂ 2Q ∂ 2B
that since the coefficient of (uxxx )2 , should be zero, i.e. = = 0,
∂u2xx ∂u2xx
the function B has to be a linear function with respect to uxx . Therefore
B = b(x, t, u, ux )uxx + C(x, t, u, ux ) (31.15)
∂ 2Q ∂Q
The coefficients of uxxx ux , uxxx uxx and uxxx are 2 +4 − K(t) =
∂u∂uxx ∂uxxx
∂b ∂ 2Q ∂b ∂ 2Q ∂b
2 + 3K(t), 2 =2 and 2 − K 0 (t) = 2 − K 0 (t) re-
∂u ∂ux ∂uxx ∂ux ∂x∂uxx ∂x
spectively. We thus have the system of PDEs
∂b ∂b ∂b
2 + 3K(t) = 0, 2 = 0, 2 − K 0 (t) = 0 (31.16)
∂u ∂ux ∂x
The solution of the above system provides us with the general expression of
the function b:
1 3
b(x, t, u, ux ) = xK 0 (t) − uK(t) + L(t) (31.17)
2 2
We thus obtain, using (31.13) and (31.15) the following expression for Q:

1 0 3
Q = K(t)uxxx + xK (t) − uK(t) + L(t) uxx
2 2 (31.18)
+ C(x, t, u, ux )
415
We can now ignore the first two lines of (31.14). We thus obtain the equation
∂ 2Q ∂Q ∂ 2Q ∂ 2Q ∂Q
(uxx )2 +3 + u xx 2 + 2u x + 3ux
∂ 2Q ∂Q ∂ Q2 2
∂ Q
+ (ux )2 + + u x 2 + (31.19)
∂u2 ∂ux ∂x∂u ∂x2
∂Q ∂Q
−u − ux Q − =0
∂x ∂t
∂ 2Q ∂Q ∂ 2C
The coefficient of (uxx )2 is + 3 = + 3K(t), and setting this
∂u2x ∂uxxx ∂u2x
to zero, we have the equation
∂ 2C
+ 3 K(t) = 0 (31.20)
∂u2x
Integrating twice the above equation with respect to ux , we obtain

3
C(x, t, u, ux ) = − K(t)u2x + M (x, t, u)ux + N (x, t, u) (31.21)
2
Therefore we get the following expression for Q:

1 0 3
Q = K(t)uxxx + xK (t) − uK(t) + L(t) uxx
2 2 (31.22)
3
− K(t)u2x + M (x, t, u)ux + N (x, t, u)
2
It is now safe to disregard the first term of equation (31.19). We thus obtain
the equation
∂ 2Q ∂ 2Q ∂Q
uxx 2 + 2ux + 3ux
∂x∂ux ∂u∂ux ∂uxx
∂ 2Q ∂Q ∂ Q ∂ 2Q
2
+ (ux )2
+ + ux 2 + (31.23)
∂u2 ∂ux ∂x∂u ∂x2
∂Q ∂Q
−u − ux Q − =0
∂x ∂t
416
with Q given by (31.22). Upon expanding (31.23) and arranging with respect
to the powers of partial derivatives of the function u, we obtain
2 2
∂ 2M

∂ M 3 3 ∂ N 3 0
− K(t) (ux ) + +2 + K (t) (ux )2
∂u2 2 ∂u2 ∂x∂u 2
2
∂ 2N

∂ M ∂M ∂M
+ + −u −N − ux
∂x2 ∂x∂u ∂x ∂t

∂M 1 0 3
+2 + xK (t) − uK(t) + L(t) ux uxx (31.24)
∂u 2 2

∂M 1 00 0 0
+ 2 − xK (t) + uK (t) − L (t) uxx
∂x 2
∂ 2N ∂N ∂N
+ −u − =0
∂x2 ∂x ∂t
Equating to zero the coefficients of the partial derivatives in the above equa-
tion, we obtain the following system of differential equations
∂ 2M 3
2
− K(t) = 0 (31.25)
∂u 2
∂ 2N ∂ 2M 3 0
+ 2 + K (t) = 0 (31.26)
∂u2 ∂x∂u 2
∂ 2M ∂ 2N ∂M ∂M
+ − u − N − =0 (31.27)
∂x2 ∂x∂u ∂x ∂t
∂M 1 3
+ xK 0 (t) − uK(t) + L(t) = 0 (31.28)
∂u 2 2
∂M 1
2 − xK 00 (t) + uK 0 (t) − L0 (t) = 0 (31.29)
∂x 2
∂ 2N ∂N ∂N
− u − =0 (31.30)
∂x2 ∂x ∂t
We have now to integrate the above system of equations (31.25)-(31.30).
From equation (31.25) we obtain by a double integration with respect to u
3
M (x, t, u) = K(t)u2 + P (x, t)u + R(x, t) (31.31)
4
Using the above value for M , we get form (31.28) the following expression
for P :
1
P (x, t) = − xK 0 (t) − L(t) (31.32)
2
417
From (31.29), using the expressions for M and P , we get the equation
∂R 1 1
= xK 00 (t) + L0 (t)
∂x 4 2
from which, by a simple integration with respect to x, we get
1 1
R(x, t) = x2 K 00 (t) + xL0 (t) + F (t) (31.33)
8 2
Using the expressions for P and R given by (31.32) and (31.33), we get the
following expression for M :

3 2 1 0
M (x, t, u) = K(t)u − xK (t) + L(t) u
4 2 (31.34)
1 2 00 1 0
+ x K (t) + xL (t) + F (t)
8 2
From (31.26), using the above expression of M , we get
∂ 2N 1
2
+ K 0 (t) = 0
∂u 2
and under double integration with respect to u, we get the following expres-
sion of N :
1
N (x, t, u) = − K 0 (t)u2 + S(x, t)u + T (x, t) (31.35)
4
Using the above found expression of M and N given by (31.34) and (31.35)
respectively, we obtain from (31.27) the equation
1 1 1 1
− K 000 (t)x2 − L00 (t)x + −S + L0 (t) + K 00 (t)x u
8 2 2 4 (31.36)
∂S 1 00
+ + K (t) − F 0 (t) − T = 0
∂x 4
Finally, from equation (31.30), using the expression for N given by (31.35),
we get the equation
1 ∂S 2 ∂ 2 S ∂S ∂T ∂ 2T ∂T
K 00 (t) − u + 2
− − u + 2
− =0 (31.37)
4 ∂x ∂x ∂t ∂x ∂x ∂t
418
From the above equation we obtain the system, equating the zero all the
coefficients of u:
1 00 ∂S ∂ 2 S ∂S ∂T ∂ 2T ∂T
K (t) − = 0, − − = 0, − =0 (31.38)
4 ∂x ∂x2 ∂t ∂x ∂x 2 ∂t
From the above system we get
1
S(x, t) = K 00 (t)x + G(t)
4
1
T (x, t) = − K 000 (t)x2 − G0 (t)x + H(t) (31.39)
8
1 (iv) 1
K (t)x2 + G00 (t)x − K 000 (t) + H 0 (t) = 0
8 4
From the last equation we obtain
1 000
K (iv) (t) = 0, G00 (t) = 0 and K (t) + H 0 (t) = 0
4
and thus
3
K(t) = a1 t3 +a2 t2 +a3 t+a4 , G(t) = a5 t+a6 , H(t) = − a1 t+a7 (31.40)
2
Therefore, using the first two equations of (31.39), we have
1
S(x, t) = (3a1 t + a2 )x + a5 t + a6 (31.41)
2
and
3 3
T (x, t) = − a1 x2 − a5 x − a1 t + a7 (31.42)
4 2
1 1 1
− L00 (t)+G0 (t) x+ L0 (t)−G(t) u+ K 00 (t)−F 0 (t)−H(t) = 0 (31.43)
2 2 4
using the expressions for S(x, t) and T (x, t) from (31.41) and (31.42).
1 0 1 00
L (t) − G(t) = 0 and K (t) − F 0 (t) − H(t) = 0 (31.44)
2 4
419
and thus
3 1
L(t) = a5 t2 + 2a6 t + a8 , F (t) = a1 t2 + a2 − a7 t + a9 (31.45)
2 2
Let us now collect everything together. From (31.35) we get, using (31.40)-
(31.42), that
1
N (x, t, u) = − (3a1 t2 + 2a2 t + a3 )u2
4

1
+ (3a1 t + a2 )x + a5 t + a6 u (31.46)
2
3 3
− a1 x 2 − a5 x − a1 t + a7
4 2
From (31.34), we get using (31.40) and (31.45), the final expression for M :
3
M (x, t, u) = (a1 t3 + a2 t2 + a3 t + a4 )u2
4
1 2 2
− (3a1 t + 2a2 t + a3 )x + a5 t + 2a6 t + a8 u
2 (31.47)
1 3
+ (3a1 t + a2 )x2 + (a5 t + a6 )x + a1 t2
4 2
1
+ a2 − a7 t + a9
2
420
We are now ready to obtain the final expression for Q. Using (31.22) and
the expressions for K(t), L(t) and also M (x, t, u) and N (x, t, u) we have
Q = (a1 t3 + a2 t2 + a3 t + a4 )uxxx

1 3
+ (3a1 t2 + 2a2 t + a3 )x − (a1 t3 + a2 t2 + a3 t + a4 )u
2 2

+ a5 t2 + 2a6 t + a8 uxx
3
− (a1 t3 + a2 t2 + a3 t + a4 )u2x
2
3
+ (a1 t3 + a2 t2 + a3 t + a4 )u2
4
1
− (3a1 t2 + 2a2 t + a3 )x + a5 t2 + 2a6 t + a8 u
2
1 2 3 2 1
+ (3a1 t + a2 )x + (a5 t + a6 )x + a1 t + a2 − a7 t + a9 u x
4 2 2
1 2 2
1
− (3a1 t + 2a2 t + a3 )u + (3a1 t + a2 )x + a5 t + a6 u
4 2
3 3
− a1 x2 − a5 x − a1 t + a7
4 2
(31.48)
421

3 3
Q = a1 t3 uxxx + (t2 x − t3 u)uxx − t3 u2x
2 2
3
3 2 3 2 3 2 3 2 3 2 2 3 3 2 3
+ t u − t xu + tx + t ux − t u + txu − x − t
4 2 4 2 4 2 4 2
3 3 3 2 2 1 1
+ a2 t2 uxxx + tx − t2 u uxx − t2 u2x + t u − txu + x2 + t ux
2 2 4 4 2

1 1
− tu2 + xu
2 2
1
3 3 2 3 2 1 1 2
+ a3 tuxxx + x − tu uxx − tux + tu − xu ux − u
2 2 2 4 2 4

3 3 3
+ a4 uxxx − uuxx − u2x + u2 ux
2 2 4
2 2
+ a5 [t uxx + (−t u + tx)ux + tu − x]
+ a6 [2tuxx + (−2tu + x)ux + u]
+ a7 (−tux + 1) + a8 (uxx + uux ) + a9 ux
(31.49)
We thus have derived the generators of the Lie-Bäcklund algebra for the
Burgers equation
3 3
Q[1] = t3 uxxx + (t2 x − t3 u)uxx − t3 u2x
2 2
3 3 3 3
+ t3 u2 − t2 xu + tx2 + t2 ux (31.50)
4 2 4 2
3 2 2 3 3 2 3
− t u + txu − x − t
4 2 4 2
3 3
Q[2] = t2 uxxx + tx − t2 u uxx − t2 u2x
2 2 (31.51)
3 1 1 1 1
+ t2 u2 − txu + x2 + t ux − tu2 + xu
4 4 2 2 2
1 3 3
Q[3] = tuxxx + x − tu uxx − tu2x
2 2 2 (31.52)
3 1 1
+ tu2 − xu ux − u2
4 2 4
422
3 3 3
Q[4] = uxxx − uuxx − u2x + u2 ux (31.53)
2 2 4
2 2
Q[5] = t uxx + (−t u + tx)ux + tu − x (31.54)
Q[6] = 2tuxx + (−2tu + x)ux + u (31.55)
Q[7] = −tux + 1 (31.56)
Q[8] = uxx + uux (31.57)
Q[9] = ux (31.58)
32 Lie-Bäcklund Symmetries for the

Fokker-Planck equation
Example 3. As our third example we consider the Fokker-Planck equation
ut = uxx + xux + u, u = u(x, t) (32.1)
Invariance is being expressed by the condition
Dt Q = Dx2 Q + xDx Q + Q (32.2)
where
Q = Q(x, t, u, ux , uxx , uxxx ) (32.3)
We shall determine the third order symmetries of Fokker-Planck equation.

∂
That is why we consider only up to third order partial derivatives i.e. ,
∂ux
∂ ∂
and of the various operators. Using the explicit form of these
∂uxx ∂uxxx
423
operators, Fokker-Planck equation (32.1) takes on the form
∂ 2Q ∂Q
(uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
2 2
∂ Q ∂Q ∂ Q ∂ 2Q
+ (uxx )2 2 + uxx + uxx 2 + 2ux
2 2
∂ Q ∂ Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +

∂Q ∂Q ∂Q ∂Q ∂Q
+x + ux + uxx + uxxx + uxxxx +Q

∂Q ∂Q ∂Q ∂Q ∂Q
(32.4)
with respect to x. We obtain by differentiation of the Fokker-Planck equation
ut = uxx + xux + u with respect to x, that
utx = uxxx + 2ux + xuxx (32.5)
utxx = uxxxx + 3uxx + xuxxx (32.6)

utxxx = uxxxxx + 4uxxx + xuxxxx (32.7)
424
∂ 2Q ∂Q
(uxxxx )2 2
+ uxxxx
∂uxxx ∂uxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q ∂Q
+ (uxxx )2 2 + uxxx
∂uxx ∂ux
∂ 2Q ∂ 2Q ∂ 2Q
2 2
∂ Q ∂Q ∂ Q ∂ 2Q
+ (uxx )2 2 + uxx + uxx 2 + 2ux
2 2
∂ Q ∂ Q ∂Q ∂ 2Q
+ (ux )2 2 + ux 2 + uxxxxx +

∂Q ∂Q ∂Q ∂Q ∂Q
+x + ux + uxx + uxxx + uxxxx +Q

∂Q ∂Q ∂Q
− + (uxx + xux + u) + (uxxx + 2ux + xuxx )
∂t ∂u ∂ux
∂Q
+ (uxxxx + 3uxx + xuxxx )
∂uxx

∂Q
+ (uxxxxx + 4uxxx + xuxxxx ) =0
∂uxxx
(32.8)
425
Upon expanding, canceling opposite terms and rearranging, the previous
equation can be written in equivalent form
∂ 2Q
(uxxxx )2
∂u2xxx
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
∂ Q
+ (uxxx )2 2
∂uxx
∂ 2Q ∂ 2Q ∂ 2Q
2 2 2
∂ Q ∂ Q ∂ Q
+ (uxx )2 2 + uxx 2 + 2ux
∂ux ∂x∂ux ∂u∂ux
2 2 2
∂ Q ∂ Q ∂ Q ∂Q
+ (ux )2 2 + ux 2 + 2
+ x +Q
∂u ∂x∂u ∂x ∂x
∂Q ∂Q ∂Q ∂Q ∂Q
− +u + 2ux + 3uxx + 4uxxx =0
(32.9)
From the above equation we see that the coefficients of (uxxxx )2 and uxxxx
should be separately zero:
∂ 2Q
=0 (32.10)
∂u2xxx
and
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
+ ux + uxx + uxxx =0 (32.11)
From (32.10) we see that Q is a linear function with respect to uxxx . However
the coefficient of uxxx should be a function of t only, since considering the
coefficients of uxxxx , uxxxx ux , uxxxx uxx and uxxxx uxxx appearing in (32.9) i.e.
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
, , and respectively, we conclude that
this coefficient should be independent of x, u, ux and uxx . We thus have the
following expression for Q:
426
We can now ignore the first two lines of equation (32.9). The rest of the
equation is
∂ 2Q
(uxxx )2
∂u2xx
∂ 2Q ∂ 2Q ∂ 2Q
2 2 2
∂ Q ∂ Q ∂ Q
+ (uxx )2 2 + uxx 2 + 2ux (32.13)
∂ux ∂x∂ux ∂u∂ux
2 2 2
∂ Q ∂ Q ∂ Q ∂Q
+ (ux )2 2 + ux 2 + 2
+x +Q
∂u ∂x∂u ∂x ∂x
∂Q ∂Q ∂Q ∂Q ∂Q
− +u + 2ux + 3uxx + 4uxxx =0
We now consider equation (32.13) with Q given by (32.12).

From equation (32.13) we see that the coefficient of (uxxx )2 should be zero,
∂ 2Q ∂ 2B
= 0, i.e. = 0, which means that B is a linear function with respect
∂u2xx ∂u2xx
to uxx : B = b(x, t, u, ux )uxx + C(x, t, u, ux ). Equating on the other hand the
∂ 2Q ∂ 2B
coefficients of uxxx ux and uxxx uxx to zero, i.e. = = 0, and
∂u∂uxx ∂u∂uxx
∂ 2Q ∂ 2B
= = 0, we see that b should be independent of u and ux ,
∂ux ∂uxx ∂ux ∂uxx
i.e. b = L(x, t) and thus
B = L(x, t)uxx + C(x, t, u, ux ) (32.14)
which means that
Q = K(t)uxxx + L(x, t)uxx + C(x, t, u, ux ) (32.15)
We next have to equate to zero the coefficient of uxxx i.e. equation (32.13)
taking into account the expression (32.15).
∂L
This coefficient is given by 2 + K(t) − (K 0 (t) + 4K(t)) and thus we shall
∂x
have the differential equation
∂L
2 − K 0 (t) − 3K(t) = 0 (32.16)
∂x
427
We can now ignore the first two lines of equation (32.13). We will then have
the equation
2 ∂ 2Q
2∂Q ∂ 2Q
(uxx ) + uxx 2 + 2ux
2 2 2
∂ Q ∂ Q ∂ Q ∂Q
+ (ux )2 2 + ux 2 + 2
+x +Q (32.17)
∂u ∂x∂u ∂x ∂x
∂Q ∂Q ∂Q ∂Q ∂Q
− +u + 2ux + 3uxx + 4uxxx =0

∂ 2Q ∂ 2C
Equating the coefficient of (uxx )2 to zero, i.e. = = 0 we see that C
∂u2x ∂u2x
is linear with respect to ux , i.e. C = c(x, t, u)ux + D(x, t, u). We further see
∂ 2Q ∂c
that equating to zero the coefficient of uxx ux , i.e. = = 0, which
∂u∂ux ∂u
means that c does not depend on u. Therefore C(x, t, u, ux ) = M (x, t)ux +
D(x, t, u) and thus
Q = K(t)uxxx + L(x, t)uxx + M (x, t)ux + D(x, t, u) (32.18)
Let us now equate to zero the coefficient of uxx . This coefficient is equal to
2

∂M ∂ L ∂L ∂L
2 + 2 +x +L− + 3L . We thus have the differential equation
∂x ∂x ∂x ∂t
∂M ∂ 2L ∂L ∂L
2 + 2
+x − − 2L = 0 (32.19)
∂x ∂x ∂x ∂t
We can now ignore the first line of equation (32.17). We then obtain the
equation
2 ∂ 2Q ∂ 2Q
2∂ Q ∂Q
(ux ) 2
+ ux 2 + 2
+x +Q
∂u ∂x∂u ∂x ∂x (32.20)
∂Q ∂Q ∂Q ∂Q ∂Q
− +u + 2ux + 3uxx + 4uxxx =0

∂ 2Q ∂ 2D
Equating the coefficient of (ux )2 to zero, i.e. = = 0, we see that
∂u2 ∂u2
428
D is a linear function with respect to u. Therefore D = N (x, t)u + P (x, t)
and thus
Q = K(t)uxxx + L(x, t)uxx + M (x, t)ux + N (x, t)u + P (x, t) (32.21)
Equation (32.20) with Q given by (32.21), takes on the form (we ignore terms
proportional to uxxx and uxx , which have already been taken into account)
∂N ∂ 2 M ∂ 2N ∂ 2P
ux 2 + ux + u +
∂x ∂x2 ∂x2 ∂x2
∂M ∂N ∂P
+x ux + u+ + M ux + N u + P
∂x ∂x ∂x
∂M ∂N ∂P
− ux + u+ + N u + 2ux M = 0
∂x ∂t ∂t
The previous equation can be written as
∂N ∂ 2M ∂M ∂M
2 + +x − − M ux
∂x ∂x2 ∂x ∂t
∂ 2N ∂N ∂N
+ + x − u (32.22)
∂x2 ∂x ∂t
∂ 2P ∂P ∂P
+ +x − +P =0
∂x2 ∂x ∂t
following three equations
∂N ∂ 2M ∂M ∂M
2 + + x − −M =0 (32.23)
∂x ∂x2 ∂x ∂t
∂ 2N ∂N ∂N
2
+x − =0 (32.24)
∂x ∂x ∂t
and
∂ 2P ∂P ∂P
+ x − +P =0 (32.25)
∂x2 ∂x ∂t
We thus see that in order to determine the characteristic Q given by (32.21),
we have to solve the system of equations (32.16), (32.19) and (32.23)-(32.25)
in order to determine the unknown functions K(t), L(x, t), M (x, t), N (x, t)
and P (x, t).
429
33 Lie-Bäcklund for a nonlinear
wave equation
Example 4. A nonlinear wave equation. We consider the first order
nonlinear wave equation
ut = uux (33.1)
Invariance is being expressed by the equation
Dt Q = uDx Q + ux Q (33.2)
where we take Q the function
Q = Q(x, t, u, ux , uxx ) (33.3)
since we want to consider second order symmetries.

The operators Dx and Dt will be taken as
∂ ∂ ∂ ∂
Dx = + ux + uxx + uxxx (33.4)
∂x ∂u ∂ux ∂uxx
and
∂ ∂ ∂ ∂
Dt = + ut + utx + utxx (33.5)
∂t ∂u ∂ux ∂uxx
respectively.
In the above operator expressions we have disregarded partial derivatives of
∂ ∂
the type , , etc. since we suppose that Q = Q(t, x, u, ux , uxx ).
∂ut ∂utx
We have however included derivatives of the type ut , utx , utxx since they can
be expressed in terms of the partial derivatives ux , uxx , uxxx .
Therefore the invariance condition is being expressed as
∂ ∂ ∂ ∂
+ ut + utx + utxx Q
∂ (33.6)
∂ ∂ ∂
−u + ux + uxx + uxxx Q = ux Q
In the above expression we substitute
ut = uux , utx = (ux )2 + uuxx , utxx = 3ux uxx + uuxxx (33.7)
430
where the last two are differential consequences of the first equation. We
thus obtain
∂ ∂ ∂ ∂
+ uux + [(ux )2 + uuxx ] + [3ux uxx + uuxxx ] Q
∂ (33.8)
∂ ∂ ∂
−u + ux + uxx + uxxx Q = ux Q
Upon expanding, canceling opposite terms and rearranging, we get from the
above equation
∂Q ∂Q ∂Q ∂Q
−u + u2x + 3ux uxx = ux Q (33.9)
∂t ∂x ∂ux ∂uxx
Chapter XII
Recursion Operators
34 Recursion Operators
In our previous examples, considered in Chapter XI, we have considered third
order symmetries only. However calculations become rather involved once we
decide to consider higher order symmetries. That is why we have to consider
the so-called recursion operators. Using recursion operators we can find
higher order symmetries once we know the lower order ones.
For example, the KdV equation
ut = 6uux + uxxx
admits the generalized symmetries
G(1) = ux , G(2) = 6uux + uxxx ,

G(3) = 30u2 ux + 20ux uxx + 10uuxxx + uxxxxx ,
G(4) = 140u3 ux + 70u3x + 280uux u2x + 70u2 u3x + 70u2x u3x
+ 42ux u4x + 14ux u5x + u7x
The above form of KdV equation admits the recursion operator R given by
R = Dx2 + 4u + 2ux Dx−1
431
We can verify that
R(G(1) ) = R(ux ) = (Dx2 + 4u + 2ux Dx−1 )(ux ) = 6uux + uxxx = G(2)
R(G(2) ) = R(6uux + uxxx ) = (Dx2 + 4u + 2ux Dx−1 )(6uux + uxxx )

= 30u2 ux + 20ux u2x + 10uu3x + u5x = G(3)
In other words,
R(G(1) ) = G(2) , R(G(2) ) = G(3) , etc
We thus see that a recursion operator maps a generalized symmetry into

another generalized symmetry.
34.I. The Euler Method. Let us consider a linear PDE
L[u] = 0 (34.1)
where L is a linear differential operator and assume that this PDE admits a
Lie-Bäcklund symmetry generated by
∂
X = Q(x, u, ux , · · · ) (34.2)
∂u
Since (34.1) is linear, the symmetry condition takes the form
L[Q] = 0 when L[u] = 0 (34.3)
If however Q is linear in u, that is it can be written as
Q = R[u] (34.4)
for a linear operator R, then the determining equation becomes
LR[u] = 0 when L[u] = 0 (34.5)
In other words, if u solves L[u] = 0 then v = R[u] solves L[v] = 0. Therefore

if a linear differential equation which admits a Lie-Bäcklund symmetry with
Q linear in u, admits infinitely many Lie-Bäcklund symmetries.
Theorem 1. Let the linear differential equation L[u] = 0 admits a Lie-
Bäcklund symmetry
∂
X = Q(x, u, ux , uxx , · · · ) (34.6)
∂u
432
such that Q can be written as Q = R[u] for a linear differential operator R.
Then L[u] = 0 also admits the Lie-Bäcklund symmetries
∂
X = Rs Q(x, u, ux , · · · ) , s = 1, 2, 3, · · · (34.7)
∂u
In this case the linear operator R is called a recursion operator for the
equation L[u] = 0.
Let us consider next a general PDE
F (x, t, u, ux , uxx , · · · ) (34.8)
of order n≥2 admitting a Lie-Bäcklund symmetry generated by

∂
X = Q(x, u, ux , uxx , · · · ) (34.9)
∂u
The symmetry condition is actually
P r(n) F = 0
which is a linear PDE in Q. That is we have
L[u]Q = 0 (34.10)
for the linear operator

∂F ∂F ∂F ∂F 2
L[u] = + Dt + Dx + D + ··· (34.11)
∂u ∂ut ∂ux ∂uxx x
The linearized equation corresponding to (34.8) is by definition
L[u]v = 0 (34.12)
The above is also the necessary condition for
φ=u+v (34.13)
to solve equation (34.8). We thus see that every Lie-Bäcklund symmetry

v = Q admitted by equation (34.8) is a special solution of the linearized
equation.
Suppose next that (34.12) admits a nontrivial symmetry generated by
∂
X = (R[u]v) (34.14)
∂v
433
Then by Theorem 1, we see that
v = R[u]Q (34.15)
will also solve equation (34.12). We thus get the following

Theorem 2. If the PDE
F (x, u, ux , uxx , · · · ) = 0 (34.16)
admits the Lie-Bäcklund symmetry

∂
X = Q(x, u, ux , uxx , · · · ) (34.17)
∂u
and its associated linearized equation L[u]v = 0 admits a nontrivial symme-
try
∂
X = (R[u]v) (34.18)
∂v
then F (x, u, ux , uxx , · · · ) = 0 admits the infinite sequence of Lie-Bäcklund
symmetries
∂
X = ((R[u]v)s )Q , s = 1, 2, · · · (34.19)
∂u
Theorem 3. For a PDE F (x, u, ux , uxx , · · · ) = 0, the determining equation
for the recursion operators is
L[u]R[u]v = 0 (34.20)
for every u and v satisfying (34.8) and (34.12) respectively.

Theorem 4. The determining equation (34.20) for a recursion operator R[u]
is equivalent to the commutator condition
[L[u], R[u]]φ = 0 (34.21)
Theorem 5. For evolution equations of the form
ut = F (u, ux , uxx , · · · ) (34.22)
the commutator condition is equivalent to
[L̂[u], R[u]]φ = (Dt R[u])φ (34.23)
434
where
∂F ∂F ∂F 2
L̂[u] = + Dx + D + ··· (34.24)
∂u ∂ux ∂uxx x
Using recursion operators, we can determine if an equation is integrable with-
out knowing an explicit solution. We can also generate integrable hierarchies
by applying recursion operators to t−translation symmetries, i.e. R[u]ut .
Example. Consider the evolution equation
ut = u2 uxx (34.25)
This equation is a special case of (34.62) with a known recursion operator.

34.1. The Determining Equations. Let us consider evolution equations
of the form
ut = F (u, ux , uxx ) (34.26)
A recursion operator R[u] of this type of equations should satisfy the equation
[L̂[u], R[u]]φ = (Dt R[u])φ (34.27)
To determine R[u] we make an ansatz of the form
R[u] = QDx + H + I1 Dx−1 ◦J1 + I2 Dx−1 ◦J2 (34.28)
The quantities Q, H, I1 , I2 , J1 , J2 depend on u, ux , uxx , · · · . Considering the

above expression for R[u], we can expand both sides of (34.27) and find the
determining equations for the various quantities.
Before that, let us introduce the auxiliary notation
∂F ∂F ∂F
K= , E= , B= (34.29)
∂u ∂ux ∂uxx
435
The left hand side of (34.27) reads in expanded form
[L̂[u], R[u]]φ =
= (K + EDx + BDx2 )(QDx + H + I1 Dx−1 ◦J1 + I2 Dx−1 ◦J2 )φ
− (QDx + H + I1 Dx−1 ◦J1 + I2 Dx−1 ◦J2 )(K + EDx + BDx2 )φ

= φxx 2BDx Q − QDx B

+ φx EDx Q + BDx2 Q + 2BDx H − QDx E

+ φ EDx H + BDx2 H + 2BJ1 Dx I1 + BI1 Dx J1 + 2BJ2 Dx I1
(34.30)
+ BI2 Dx J2 − QDx K + I1 Dx (J1 B) + I2 Dx (J2 )

+ Dx−1 (J1 φ) KI1 + EDx I1 + BDx2 I1

+ Dx−1 (J2 φ) KI2 + EDx I2 + BDx2 I2

+ I1 Dx−1 Dx (J1 E)φ − J1 Kφ − Dx2 (J1 B)φ

−1 2
+ I2 Dx Dx (J2 E)φ − J2 Kφ − Dx (J2 B)φ
The right hand side of (34.27) gives upon expansion

(Dt R[u])φ = Dt (QDx + H + I1 Dx−1 ◦J1 + I2 Dx−1 ◦J2 ) φ
= φx Dt Q + φDt H + Dt I1 ·Dx−1 (J1 φ) + I1 Dx−1 (φDt J1 ) (34.31)
+ Dt I2 ·Dx−1 (J2 φ) + I2 Dx−1 (φDt J2 )
Comparing the two expanded members, i.e. (34.30) and (34.31), we obtain
the following set of determining equations
2BDx Q − QDx B = 0 (34.32)
EDx Q + BDx2 Q + 2BDx H − QDx E = Dt Q (34.33)

EDx H + BDx2 H + 2BJ1 Dx I1 + BI1 Dx J1 + 2BJ2 Dx I1
(34.34)
+ BI2 Dx J2 − QDx K + I1 Dx (J1 B) + I2 Dx (J2 ) = Dt H
KI1 + EDx I1 + BDx2 I1 = Dt I1 (34.35)
KI2 + EDx I2 + BDx2 I2 = Dt I2 (34.36)
436
Dx (J1 E) − J1 K − Dx2 (J1 B) = Dt J1 (34.37)
Dx (J2 E) − J2 K − Dx2 (J2 B) = Dt J2 (34.38)
34.2. Solution of the Determining Equations. We have a set of
seven determining equations (34.32)-(34.38) with six unknown quantities
Q, H, I1 , I2 , J1 , J2 . The determining equations do not have a general solu-
tion. Some of them however admit a general solution. On the other hand we
have to consider a special ansatz in solving some of these equations.
34.2.1. Solution of equation (34.32). Equation (34.32), taking into
account
F = F (u, ux , uxx )
∂F
B= (34.39)
∂uxx
∂ ∂ ∂ ∂ ∂
Dx = + ux + uxx + uxxx + uxxxx + ···
can be written as
∂Q ∂Q ∂Q ∂Q
2B ux + uxx + uxxx + uxxxx + ···
∂u ∂ux ∂uxx ∂uxxx
∂B (34.40)
∂B ∂B
− Q ux + uxx + uxxx =0
∂u ∂ux ∂uxx
Let us now equate to zero the coefficients of the partial derivatives of the
∂Q
function u. Equating to zero the coefficient of uxxxx , we get 2B =0
∂uxxx
∂Q
and thus = 0. Therefore
∂uxxx
Q = Q(u, ux , uxx ) (34.41)
∂Q ∂B
Equating to zero the coefficient of uxxx , we get 2B −Q = 0 and
∂uxx ∂uxx
dQ 1 dB
thus = and by integration
Q 2 B
Q = B 1/2 P 1/2 , P = P (u, ux ) (34.42)
437
Equating to zero the coefficient of uxx in equation (34.40), and taking into
∂ ∂B
account (34.42), we get 2B (BP )1/2 − Q = 0. This equation is equiv-
∂ux ∂ux
∂P ∂P
alent to B = 0, i.e. = 0 and thus
∂ux ∂ux
P = P (u) (34.43)
Equating to zero the coefficient of ux , and taking into account (34.42), we get
∂ ∂B ∂P
2B (BP )1/2 − (BP )1/2 = 0. This equation is equivalent to B = 0,
∂u ∂u ∂u
∂P
i.e. = 0 and thus
∂u
P = constant, P >0 (34.44)
Conclusion. Equation (34.32) admits the solution
∂F
Q = (BP )1/2 , B= , P >0 (34.45)
∂uxx
34.2.2. Solution of equations (34.35) and (34.36). These two equations
admit exact solutions too. We consider the following expressions of Ii , i = 1, 2
Ii = αi ux + βi ut + γi , i = 1, 2 (34.46)
where we have taken into account the most general expressions corresponding
to translational symmetries. Equations (34.35) and (34.36) can be written
in expanded form
K(αi ux + βi ut + γi ) + E(αi uxx + βi utx )
(34.47)
+ B(αi uxxx + βi utxx ) = αi uxx + βi utt
and taking into account that
∂F ∂F ∂F
K= , E= , B= (34.48)
∂u ∂ux ∂uxx
∂F ∂F ∂F
αi ux + uxx + uxxx
∂u ∂ux ∂uxx
∂F ∂F ∂F ∂F (34.49)
+ βi ut + utx + utxx + γi
∂u ∂ux ∂uxx ∂u
= αi uxt + βi utt
438
or, since uxt = Dx (F ) and utt = Dt (F )
∂F
αi Dx (F ) + βi Dt (F ) + γi = αi Dx (F ) + βi Dt (F ) (34.50)
∂u
∂F
The above equation is true if γi = 0, i.e. if γi = 0. Therefore
∂u
I1 = α1 ux + β1 ut , I2 = α2 ux + β2 ut (34.51)
We have seen that equations (34.32), (34.35) and (34.36) can be solved for any
form of the equation (34.22), i.e. for any form of the function F (u, ux , uxx ).
For the rest of the determining equations we shall consider, the function
F (u, ux , uxx ) belongs to one of the following categories :
I. The semilinear case
F = αuxx + β(u, ux )
II. The quasilinear case
F = α(u)uxx + β(u, ux )
III. The nonlinear case
F = α(u, ux )uxx + β(u, ux )
IV. The fully nonlinear case
F = F (u, ux , uxx )
Let us now solve the rest of the determining equations for some of the above
cases :
Case I. The semilinear case. In this case we take F as
F = α uxx + β(u, ux ) (34.52)
The quantities B and E are evaluated to be
∂F ∂F ∂β
B= = α, E= = (34.53)
∂uxx ∂ux ∂ux
Since B is a constant, we see from (34.45) that Q is a constant as well.
Therefore (34.33) takes on the form
2BDx H − QDx E = 0 (34.54)
439
The expanded form of the above equation is
∂H ∂H ∂H
2α ux + uxx + uxxx + ···
∂u ∂ux ∂uxx
2 2 (34.55)
∂ β ∂ β
− Q ux + uxx 2 = 0
∂u∂ux ∂ux
We equate to zero all the partial derivatives of the function u in order to

determine the quantity H. Equating to zero the coefficient of uxxx we find
∂H ∂H
2α = 0, i.e. = 0 and thus H = H(u, ux ). Equating to zero the
∂uxx ∂uxx
coefficient of uxx we get
∂H ∂ 2β
2α −Q 2 =0
∂ux ∂ux

∂ ∂β
2αH − Q =0 (34.56)
∂ux ∂ux
Similarly equating to zero the coefficient of ux we get
∂H ∂ 2β
2α −Q =0
∂u ∂u∂ux
∂ ∂β
2αH − Q =0 (34.57)
∂u ∂ux
From equations (34.56) and (34.57) we get that the quantity inside the big
∂β
parenthesis should be a constant, i.e. 2αH − Q = C1 from which we
∂ux
obtain that
Q ∂β
H= + C, H = H(u, ux ) (34.58)
2α ∂ux
We can similarly proceed further for the rest of the cases. We list below some
evolution equations and their corresponding recursion operators:
440
Equation 1.
β 0 (u) 2
ut = uxx + γux − u + β(u), β(u)6=0
β(u) x
(34.59)
β 0 (u)
R = Dx − ux
β(u)
Equation 2.
β 00 (u) 2
ut = uxx + β(u)ux + u , β 0 (u)6=0
β 0 (u) x
(34.60)
β 00 (u) 1 1
R = Dx + 0 ux + β(u) + ux Dx−1 β 0 (u)
β (u) 2 2
Equation 3.
β 00 (u) 2 β 0 (u)
ut = uxx + γ(u)ux + 0
ux + δ+ 0 , β 0 (u)6=0
β (u) β(u) β (u)
(34.61)
β 00 (u)
R = Dx + 0 ux
β (u)
Equation 4.
(a2 α(u) − a3 )
ut = α(u)uxx + ux
a1
α(u)α00 (u) 2

1
− α0 (u) + ux , a1 6=0
2 α0 (u)
p 1

1 0 α(u)α00 (u)
(34.62)
R = 2a1 α(u)Dx + 2a1 p α (u) + ux
α(u) 2 α0 (u)
p α0 (u)
+ 2a2 α(u) + (a3 ux + a1 ut )Dx−1 p3
α2 (u)
Equation 5.
uxx
ut = + α(u)ux
u2x
1 uxx (34.63)
R = Dx − 2 + a1 ux Dx−1
ux ux
441
Equation 6.
uxx
ut = + a3 uux + a4 ux
(a1 − a2 ux )2
1 a2 uxx a1 a3 u a1 a3 ux −1 (34.64)
R= Dx + 2
+ + Dx
a1 − a2 u x (a1 − a2 ux ) 2 2
Equation 7.
1
ut = + P (ux )
uxx
(34.65)
1
R= Dx
uxx
34.II. The Hereman Method. Let us now consider another approach of
determining recursion operators, using the KdV equation
ut = 6uux + uxxx (34.66)
This equation is invariant under the scaling
(t, x, u)−→(λ−3 t, λ−1 x, λ2 u) (34.67)
In fact, considering the transformation
û = λα u, x̂ = λβ x, t̂ = λγ t (34.68)
and taking into account

∂ û ∂u ∂ û ∂u ∂ 3 û 3
α−3β ∂ u
= λα−γ , = λα−β , = λ (34.69)
∂ t̂ ∂t ∂ x̂ ∂x ∂ x̂3 ∂x3
the KdV equation we considered above becomes
3
α−γ ∂u 2α−β ∂u α−3β ∂ u
λ = 6λ u +λ
∂t ∂x ∂x3
ut = 6λα+γ−β uux + λγ−3β uxxx
Invariance requires
α + γ − β = 0 and γ − 3β = 0
442
i.e. α = −2β, γ = 3β and for α = 2, we get
α = 2, β = −1 and γ = −3 (34.70)
We thus have
û = λ2 u, x̂ = λ−1 x, t̂ = λ−3 t (34.71)
In other words, the KdV equation is invariant under the transformation
(t, x, u)−→(λ−3 t, λ−1 x, λ2 u) (34.72)
Taking into account the above scaling, we assign the rank (or weight) of the
variables by
rank(u) = 2, rank(ux ) = 3, rank(unx ) = n + 2 (34.73)
We consider next the set L of powers of u of rank ≤ 7
L = {1, u, u2 , u3 } (34.74)
The following x−derivatives
∂ 3 ∂3 2
(u ) = 3u2 ux , (u ) = 6ux uxx + 2uuxxx
∂x ∂x3 (34.75)
∂5
(u) = u5x
∂x5
have rank 7. In other words, we have to specify in advance the highest rank
of symmetries we have to consider. From the above derivatives we obtain the
set of monomials of rank 7:
R = {u2 ux , ux u2x , uu2x , u5x } (34.76)
Therefore the generalized symmetry would be a linear combination of terms

of the above blocks of symmetry:
G = a1 u2 ux + a2 ux u2x + a3 uu2x + a4 u5x (34.77)
The constants ai , i = 1, 2, 3, 4 will be determined from the condition
Dt G = F 0 (u)[G] (34.78)
443
where F 0 (u)[G] is the Frechét derivative in the direction of G
m
0 ∂ X ∂F
F (u)[G] = F (u + G) = Dxi G (34.79)
∂ =0 i=0
∂uix
We obtain with F = 6uux + uxxx that

Dt G = F 0 (u)[G]⇐⇒
6(2a1 − 3a2 )u2x u2x + 6(a1 − 3a3 )uu22x + 6(a1 − 3a3 )uux u3x
(34.80)
+ 3(a2 − 10a4 )u23x + 3(a2 + a3 − 30a4 )u2x u4x
+ 3(a3 − 10a4 )ux u5x = 0
2a1 − 3a2 = 0, a1 − 3a3 = 0, a2 − 10a4 = 0,
(34.81)
a2 + a3 − 30a4 = 0, a3 − 10a4 = 0
from which we get
a1 a2 a3
= = = a4 (34.82)
30 20 10
The choice a4 = 1 leads us to the solution
a1 = 30, a2 = 20, a3 = 10, a4 = 1 (34.83)
We thus obtain the following expression for the generalized symmetry
G = 30u2 ux + 20ux u2x + 10uu2x + u5x (34.84)
Equation
ut = G (34.85)
with G given by (34.84), is a fifth order PDE, known as the Lax equation
in the completely integrable KdV hierarchy.
Let us now consider a mixing of x and t monomials of rank ≤2 , i.e.
L = {1, u, x, xu, t, tu, tu2 } (34.86)
We then have the following x−derivatives of the monomials contained in L
of rank 2:
Dx (xu) = u + xux , Dx (tu2 ) = 2tuux , Dx3 (tu = tu3x ) (34.87)
444
We thus have the following set of building blocks of symmetry
R = {u, xux , tuux , tu3x } (34.88)
The symmetry G we are looking for would be a linear combination of the

monomials in R
G = a1 u + a2 xux + a3 tuux + a4 tu3x (34.89)
The constants ai , i = 1, 2, 3, 4 will be determined from the condition
Dt G = F 0 (u)[G] (34.90)
We obtain with F = 6uux + uxxx that

Dt G = F 0 (u)[G]⇐⇒
(6a1 + 6a2 − a3 )uux + 3(a3 − 6a4 )tu22x + (3a2 − a4 )u3x (34.91)
+ 3(a3 − 6a4 )tux u3x = 0
6a1 + 6a2 − a3 = 0, a3 − 6a4 = 0, 3a2 − a4 = 0 (34.92)
from which we get

3a1 a3
= 3a2 = = a4 (34.93)
2 6
The choice a4 = 1 leads us to the solution
2 1
a1 = , a2 = , a3 = 6, a4 = 1 (34.94)
3 3
We thus obtain the following expression for the generalized symmetry
2 1
G = u + xux + 6tuux + tu3x (34.95)
3 3
We shall next calculate in detail the conserved density ρ for the KdV equation
ut + 6uux + uxxx = 0 (34.96)
of rank 6. The conserved density satisfies the equation
Dt ρ(x, t) + Dx J(x, t) = 0 (34.97)
445
where J(x, t) is the associated flux. The KdV equation is invariant under
the dilatation symmetry, proved above
(t, x, u)−→(λ−3 t, λ−1 x, λ2 u) (34.98)
We consider all operators of the form
DR−w(u)·i ui (x) (34.99)
with R = 6 and w(u) = 2, under the constraint

R
1≤i≤ , i.e. 1≤i≤3 (34.100)
w(u)
We then have the set of operators
i = 1−→Dx4 u = u4x
i = 2−→Dx2 u2 = 2u2x + 2uu2x (34.101)
i = 2−→Dx0 u3 = u3
From the above set we select a linear combination of terms which are not
total derivatives, i.e.
ρ = a1 u3 + a2 u2x (34.102)
The coefficients a1 and a2 are determined by the equation
Lu (Dt ρ) = 0 (34.103)
where Lu is Euler’s operator

∞
X ∂
Lu = (−1)k Dxk
∂uk
k=0 (34.104)
∂ ∂ ∂ ∂ ∂
= − Dx1 + Dx2 − Dx3 + Dx4 + ···
∂u ∂ux ∂uxx ∂u3x ∂u4x
and Dx , Dt are the total derivative operators with respect to x and t respec-
tively
∂ ∂ ∂
Dx = + ux + uxx + ··· (34.105)
∂x ∂u ∂ux
446
∂ ∂ ∂
Dt = + ut + utx + ··· (34.106)
∂t ∂u ∂ux
We have
∂ ∂ ∂
Dt ρ = + ut + utx (a1 u3 + a2 u2x )
∂t ∂u ∂ux (34.107)
2
= 3a1 ut u + 2a2 utx ux
and taking into account the substitutions (i.e. the differential consequences)
ut −→ − 6uux − u3x
(34.108)
utx −→ − 6u2x − 6uu2x − u4x
we obtain that
Dt ρ = −18a1 u3 ux − 3a1 u2 u3x − 12a2 u3x
(34.109)
− 12a2 uux u2x − 2a2 ux u4x
Therefore
Lu (Dt ρ) = 0⇐⇒
∂ ∂ ∂ ∂ ∂ (34.110)
− Dx1 + Dx2 − Dx3 + Dx4 (Dt ρ) = 0
∂u ∂ux ∂uxx ∂u3x ∂u4x
or more analytically
− 54a1 u2 ux − 6a1 uu2x − 12a2 ux u2x

− Dx (−18a1 u3 − 36a2 u2x − 12a2 uu2x − 2a2 u4x ) (34.111)
+ Dx2 (−12a2 uux ) − Dx3 (−3a1 u2 ) + Dx4 (−2a2 ux ) = 0
Taking into account the differential operator Dx we find
Dx (18a1 u3 + 36a2 u2x + 12a2 uu2x + 2a2 u4x )

(34.112)
= 54a1 u2 ux + 72a2 ux u2x + 12a2 (ux u2x + uu3x ) + 2a2 u5x
Dx2 (uux ) = 3ux u2x + uu3x (34.113)

Dx3 (u2 ) = 6ux u2x + 2uu3x (34.114)
and
Dx4 (ux ) = u5x (34.115)
447
Collecting everything together, after many simplifications, we find
Lu (Dt ρ) = 0⇐⇒18(a1 + 2a2 )ux u2x = 0 (34.116)
1
Therefore we should have a1 + 2a2 = 0 and for a1 = 1, we get a2 = − and
2
then
1
ρ = u3 − u2x (34.117)
2
The Kaup-Kupershmidt (KK) equation, which we consider next
25
ut = 5u2 ux + ux u2x + 5uu3x + u5x (34.118)
2
admits the dilatation symmetry
(t, x, u)−→(λ−5 t, λ−1 x, λ2 u) (34.119)
We then have the following values for the various weights
w(u) = 2, w(Dx ) = 1, w(Dt ) = 5 (34.120)
For the generalized symmetries of the KK equation we have the following

values of ranks
rank G(1) = 3, rank G(2) = 7, rank G(3) = 9,
(34.121)
rank G(4) = 13, rank G(5) = 15, rank G(6) = 19
Since
rank G(2) − rank G(1) 6=rank G(3) − rank G(2)
and
rank G(3) − rank G(1) = rank G(4) − rank G(2) = 6
we guess that rank R = 6 and s = 2, where there is the following relation

between the generalized symmetries and the recursion operator R
G(j+s) = RG(j) (34.122)
448
The recursion operator is given in general
R = R0 + R1 (34.123)
where R0 is a differential operator and R1 is an integral operator.

The differential part should be of rank 6 and thus taking all the possible
combinations of Dxi uj with rank Dxi uj = 6, we have
R0 = a1 Dx6 + a2 uDx4 + a3 ux Dx3 + a4 u2 Dx2 + a5 u2x Dx2

+ a6 uux Dx + a7 u3x Dx + a8 u3 I + a9 u2x I (34.124)
+ a10 uu2x I + a11 u4x I
Using the densities
ρ(1) = u and ρ(2) = 3u2x − 4u3 (34.125)
and the symmetries

25
G(1) = ux and G(2) = F (u) = 5u2 ux + ux u2x + 5uu3x + u5x (34.126)
2
we obtain
0 0
R1 = ã12 G(1) Dx−1 ρ(2) + ã21 G(2) Dx−1 ρ(1)
= ã12 ux Dx−1 (6ux Dx − 12u2 I) + ã21 G(2) Dx−1 I (34.127)
= a12 ux [Dx−1 (u2x I 2
+ 2u I) − ux I] + a13 G (2)
Dx−1 I
Substituting R = R0 + R1 into (F 0 (u) is the Frechét derivative)
∂R
+ R0 [F ] + R◦F 0 (u) − F 0 (u)◦R = 0 (34.128)
∂t
and equating to zero the partial derivatives of the function u, we obtain the
system
4a9 8a9 24a9 12a9
a1 = , a2 = , a3 = , a4 =
69 23 23 23
98a9 40a9 70a9 16a9
a5 = , a6 = , a7 = , a8 = (34.129)
69 23 69 69
82a9 26a9 2a9 4a9
a9 = a9 , a10 = , a11 = , a12 = , a13 =
69 69 69 69
449
The choice a9 = 69/4 gives us the solution
49
R = Dx6 + 6uDx4 + 18ux Dx3 + 9u2 Dx2 + u2x Dx2
2
35 3 69 2 41
+ 30uux Dx + u3x Dx + 4u I + ux I + uu2x I (34.130)
2 4 2
13 1 −1 2 (2) −1
+ u4x I + ux Dx (u2x + 2u )I + G Dx
2 2
Let us now turn to the Hirota-Satsuma system (a > 0)
ut = a(6uux + u3x ) − 2vvx

(34.131)
vt = −3uvx − v3x
This system models shallow water waves and admits infinity many densities
1
and generalized symmetries for a = . Let us list the first two of them
2
ρ(1) = u, ρ(2) = 3u2 − 2v 2 (34.132)
1 !
(1) ux (2) 3uux + u3x − 2vvx
G = , G = 2 (34.133)
vx −3uvx − v3x
We also have the following values for the weights: w(u) = w(v) = 2 and
w(Dt ) = 3. Using these values, we get
rank ρ(1) = 2, rank ρ(2) = 4 (34.134)
and

(1) 3 (2) 5
rank G = , rank G =
3 5
(34.135)
(3) 7 (4) 9
rank G = , rank G =
7 9
If we take rank Rij = 2 and s = 1 we get inconsistent results. We then take

s = 2 and thus rank Rij = 4. Therefore we get

(R0 )11 (R0 )12
R= + c41 G(1) Dx−1 ⊗Lu (ρ(2) )
(R0 )21 (R0 )22 (34.136)
(2) −1 (1)
+ c42 G Dx ⊗Lu (ρ )
450
where (R0 )ij is a linear combination of
{Dx4 , uDx2 , vDx2 , ux Dx , vx Dx , u2 , uv, v 2 , u2x , v2x } (34.137)
For example
(R0 )12 = c11 Dx4 + c12 uDx2 + c13 vDx3 + c14 ux Dx + c15 vx Dx
(34.138)
+ c16 u2 I + c17 uvI + c18 v 2 I + c19 u2x I + c20 v2x I
We also get

(1) ux
R1 = c41 G (1)
Dx−1 ⊗Lu (ρ(2) )
= c41 Dx−1 ⊗(6uI − 4vI)
vx
(34.139)
3ux Dx−1 uI −2ux Dx−1 vI

= c41
3vx Dx−1 uI −2vx Dx−1 vI
and

(2) F1 (u)
R1 = c42 G (2)
Dx−1 ⊗Lu (ρ(1) ) = c42 Dx−1 ⊗(I 0)
F2 (u)
(34.140)
F1 (u)Dx−1 0

= c42
F2 (u)Dx−1 0
(1) (2)
We then substitute R = R0 + R1 = R0 + R1 + R1 into the determining
equation
∂R
+ R0 [F ] + R◦F 0 (u) − F 0 (u)◦R = 0 (34.141)
∂t
to be able to determine the various coefficients cij . We thus obtain that

(R0 )11 (R0 )12
R= (34.142)
(R0 )21 (R0 )22
with
16 2
(R0 )11 = Dx4 + 8uDx2 + 12ux Dx + 16u2 I + 8u2x I − v I
3
+ 4ux Dx−1 uI + 12uux Dx−1 + 2u3x Dx−1 − 8vvx Dx−1
20 2 16 16 4 8
(R0 )12 = − vDx − vx Dx − uvI − v2x I − ux Dx−1 vI
3 3 3 3 3
451
(R0 )21 = −10vx Dx − 12v2x I + 4vx Dx−1 uI − 12uvx Dx−1 − 4v3x Dx−1
16 2 8
(R0 )22 = −4Dx4 − 16uDx2 − 8ux Dx − v I − vx Dx−1 vI
3 3
The expression (34.142) generates infinitely many generalized symmetries
which establishes that the Hirota-Satsuma system is completely integrable.
34.III. The Habibullin-Khakimova Method.
Let us consider the defining equation for the recursion operator R
d
R = [F ∗ , R] (34.143)
dt
where F ∗ is the linearization operator of an evolution equation
∂F
ut = F (u, u1 , u2 , · · · , uk ), 6=0 (34.144)
∂uk
defined by
∂F ∂F ∂F 2 ∂F k
F∗ = + Dx + Dx + · · · + D (34.145)
∂u ∂u1 ∂u2 ∂uk x
We define the operator L1 by the determinant of an (m + 1)×(m + 1) matrix
Dxm g (k1 ) Dxm−1 g (k1 ) ... g (k1 )

... ... ...
L1 U = ρ m (km ) m−1 (km ) (34.146)
Dx g Dx g ... g (km )
Dxm U Dxm−1 U ... U
where ρ is a function which depends on u, u1 , · · · and g (km ) are symmetries

which belong to one of the sets
S1 = {ux }, S2 = {ut }, S3 = {ux , ut }, S4 = {ux , ut , uτ } (34.147)
where uτ = g (3) is the closest symmetry of the evolution equation. It is also

obvious that
L1 g (km ) = 0 (34.148)
We suppose that the recursion operator R has the form
R = L−1
1 L2 (34.149)
452
The defining equation then becomes

d −1 ∗ −1
L L2 = F , L1 L2 (34.150)
dt 1

d d
L−1
1 L2 + L−1 L2 = F ∗ L−1 −1 ∗
1 L2 − L1 L2 F ⇔
dt dt 1 (34.151)
d
−2 d

L−1
1 L 2 − L1 L ∗ −1 −1
1 L2 = F L1 L2 − L1 L2 F
∗
dt dt
and multiplying by L−1
1 on the left, we obtain
d d
L2 − L−1
1 L1 L2 = L−1 ∗ −1 ∗
1 F L1 L2 − L2 F ⇔
dt dt
(34.152)
d d −1
L2 = L1 L1 + L1 F L1 L2 − L2 F ∗
−1 ∗ −1
dt dt
The last equation above can also be written as

d
L2 = AL2 − L2 F ∗ (34.153)
dt
where A is the operator
d
A= L1 L−1 −1 ∗ −1
1 + L1 F L1 (34.154)
dt
A similar equation is being satisfied by the operator L1 :
d
L1 = AL1 − L1 F ∗ (34.155)
dt
The order of the operator L2 is given by the formula
m2 = m + ord g (k1 +m) − ord g (k1 ) (34.156)
In general, a recursion operator R admits the representation

m
X
R = R0 + g (ki ) Dx−1 h(i) (34.157)
i=1
453
where R0 is a differential operator. We also have to mention that the knowl-
edge of the operators L1 and L2 allows us to write down the Lax pair
L2 Ψ = λL1 Ψ
d (34.158)
Ψ = F ∗Ψ
dt
for the evolution equation, where λ is the spectral parameter. Inversely,
knowing the Lax pair, we can determine the recursion operator.
For example for the KdV equation
ut = uu1 + u3 (34.159)
we consider that g (k1 ) = ux = u1 and thus for m = 1 we have
u2 u1
L1 U = ρ (34.160)
U1 U
and for ρ = −1 we get
L1 = u1 Dx − u2 (34.161)
The linearization operator F ∗ has the form
F ∗ = Dx3 + uDx + u1 (34.162)
The operator A should be a third order differential operator, i.e.
A = A(3) Dx3 + A(2) Dx2 + A(1) Dx + A(0) (34.163)
d
Using the equation L1 = AL1 − L1 F ∗ we can determine the coefficients
dt
A(j) , j = 3, 2, 1, 0 and then to determine A:
3u2 2 3u2 u4 u2 u3
A = Dx3 − Dx + u + 22 Dx + 3 u1 + − 2 (34.164)
u1 u1 u1 u1
The operator L2 should also be a third order differential operator, i.e.
L2 = b(3) Dx3 + b(2) Dx2 + b(1) Dx + b(0) (34.165)
454
d
Using the equation L2 = AL2 − L2 F ∗ we can determine the coefficients
dt
b(j) , j = 3, 2, 1, 0 and then to determine L2 :
2 2
L2 = u1 Dx3 − u2 Dx2 + uu1 Dx + u21 − uu2 (34.166)
3 3
Since m = 1, we see that m2 = 3 and local part of R, i.e. R0 is a second
order differential operator while we have only one non-local term. We thus
have the following representation
R = r(2) Dx2 + r(1) Dx + r(0) + u1 Dx−1 h (34.167)
Using the equation L2 = L1 R and the known expressions of L1 and L2 , we

determine r(2) , r(1) , r(0) and h. We thus obtain the following expression for
the recursion operator of the KdV equation ut = uu1 + u3
2 1
R = Dx2 + u + u1 Dx−1 (34.168)
3 3
References
[1] P. J. Olver: ”Evolution equations possessing infinitely many symme-
tries”. J. Math. Phys. 18(6) (1977) 1212-1215
[2] A. S. Fokas: ”Symmetries and Integrability”.
Stud. Appl. Math. 77 (1987) 253-299
[3] J. Häggblad: ”Symmetries and Recursion Operators of Nonlinear Dif-
ferential Equations”. Thesis, Lulea University of Technology, March 2006.
[4] Ü. Göktas: ”Algorithmic Computation of Symmetries and Recursion
Operators for Systems of Nonlinear Evolution and Differential-Difference
Equations”. Thesis, School of Mines, Colorado, May 1998.
[5] W. Hereman and Ü. Göktas: ”Integrability Tests for Nonlinear Evo-
lution Equations”. arXiv: solv-int/9904022v2
[6] D. E. Baldwin: ”Symbolic Algorithms and Software for the Painlevé
test and Recursion Operators for Nonlinear Partial Differential Equations”.
Thesis, School of Mines, Colorado, 2004.
[7] J. P. Wang: ”Symmetries and Conservation Laws of Evolution Equa-
tions”. Thesis, Vrije Universiteit of Amsterdam, 1998.
[8] J. P. Wang: ”A List of 1+1 Dimensional Integrable Equations and Their
Properties”. J. Nonlin. Math. Phys. 9 (2002) 213-233.
[9] I. T. Habitullin and A. R. Khakimova: ”On the Recursion Opera-
tors for Integrable Equations”. arXiv: 1805.01114v1
455
[10] M. Gürses, A. Karasu and V. V. Sokolov: ”On construction of
recursion operators from Lax representation ”.
J. Math. Phys. 40(12) (1999) 6473-6490
[11] A. B. Shabat (Ed): ”Encyclopedia of Integrable Systems”.
The Landau Institute, ver. 0032, 05.09.2007
[12] A. B. Shabat (Ed): ”Encyclopedia of Integrable Systems”.
The Landau Institute, ver. 0043, 31.12.2010
Chapter XIII
Conditional Symmetries
35 Theory of Conditional Symmetries
Conditional Symmetry is another type of symmetry which was introduced in
an attempt to solve nonlinear PDEs exhibiting poor Lie symmetry.
In order to present the main ingredients of the conditional symmetries method,
we consider first two differentiable functions K(t, u) and σ(t, x, u) which de-
pend on t, x, u, ux , uxx , · · · .
The function σ(t, x, u) is a conditional symmetry of the evolution equation
ut = K(t, u) (35.1)
if and only if
∂σ
= [K, σ] (35.2)
∂t
The commutator [K, σ] is given by
[K, σ] = K 0 σ − σ 0 K (35.3)
where σ 0 (u) is the Frechet derivative with respect to u, defined by
∂σ ∂σ ∂σ 2 ∂σ 3
σ 0 (u) = + ∂x + ∂x + ∂ + ··· (35.4)
∂u ∂ux ∂uxx ∂uxxx x
456
Equation (35.1) admits a Generalized Conditional Symmetry (briefly
denoted by GCS), if
∂σ
= [K, σ] + A(t, x, u, σ), A(t, x, u, 0) = 0 (35.5)
∂t
A(t, x, u, σ) is a differentiable function of t, x, u, ux , uxx , · · · and σ, σx , σxx , · · · .
From equation (35.5) we see that if σ does not depend explicitly on time t,
equation (35.1) admits a GCS σ if and only if
σ0K =0 (35.6)
σ=0
where, using (35.4),
∂σ ∂σ ∂K ∂σ ∂ 2 K ∂σ ∂ 3 K
σ 0 (u)K = K+ + + + ··· (35.7)
∂u ∂ux ∂x ∂uxx ∂x2 ∂uxxx ∂x3
36 Examples of Conditional Symmetries

Example 1. We consider the general equation
ut = g(u, ux )uxx + f (u, ux ) (36.1)
K = g(u, ux )uxx + f (u, ux ) (36.2)
We wish to find the general form of the expressions g(u, ux ) and g(u, ux ) so
that the given equation admits a GCS when
σ = uxxx + ux (36.3)
In this case we impose the condition
σ 0 (u)K =0 (36.4)
σ=0
The Frechet derivative σ 0 (u) in our case has the form

∂σ ∂σ ∂σ 2 ∂σ 3
σ 0 (u) = + ∂x + ∂x + ∂ (36.5)
∂u ∂ux ∂uxx ∂uxxx x
457
In the above expression we have truncated the Frechet derivative taking into
account the expression for σ. Since
∂σ ∂σ ∂σ ∂σ
= 0, = 1, = 0, =1 (36.6)
∂u ∂ux ∂uxx ∂uxxx
we obtain
∂K ∂ 3 K
σ 0 (u)K = + (36.7)
∂x ∂x3
Using any of the symbolic languages like MAPLE and Mathematica, we can
∂K ∂ 3 K
calculate + with K given by (36.2). The result should be arranged
∂x ∂x3
in powers of the second order partial derivative uxx . Before we write down
the complete expression of σ 0 (u)K, we note that the subscript 1 in f or g
denotes derivative with respect to u, while the subscript 2 denotes derivative
with respect to ux . For example
∂ 2f ∂ 3g
f12 = , g112 = (36.8)
∂u∂ux ∂ 2 u∂ux
Using this notation, we find
∂K ∂ 3 K
σ 0 (u)K = +
∂x ∂x3
= g222 (uxx ) + (f222 + 3g122 ux + 3g12 )(uxx )3
4
+ (3f122 ux + 3g112 u2x + g2 + 6g22 uxxx + 3g11 ux + 3f12 )(uxx )2

(36.9)
+ (g111 u3x + 3f11 ux + 3f112 u2x + f2 + g1 ux + 4g2 uxxxx
+ 9g12 ux uxxx + 4g1 uxxx + 3f22 uxxx )(uxx )
+ 3g2 (uxxx )2 + g(uxxx + uxxxxx ) + f1 (ux + uxxx ) + f2 uxxxx
+ 3g1 ux uxxxx + 3g11 uxxx (ux )2 + 3f12 ux uxxx + f111 u3x
We have now to take into account the condition σ = 0. We thus have to
make the substitution
uxxx = −ux (36.10)
and the differential consequences (which come by differentiation of the pre-

vious relation)
uxxxx = −uxx (36.11)
458
uxxxxx = −uxxx = ux (36.12)
We then obtain
∂K ∂ 3 K

0
σ (u)K = +
σ=0 ∂x ∂x3 σ=0
= g222 (uxx )4 + (f222 + 3g122 ux + 3g12 )(uxx )3
+ (3f122 ux + 3g112 u2x + g2 − 6g22 ux + 3g11 ux + 3f12 )(uxx )2
(36.13)
+ [g111 u3x + 3f11 ux + 3f112 u2x + f2 + g1 ux + 4g2 (−uxx )
+ 9g12 ux (−ux ) + 4g1 (−ux ) + 3f22 (−ux )](uxx )
+ 3g2 (ux )2 + f2 (−uxx )
+ 3g1 ux (−uxx ) + 3g11 (−ux )(ux )2 + 3f12 ux (−ux ) + f111 u3x
Canceling opposite terms and rearranging, the above expression takes on the
form
∂K ∂ 3 K

0
σ (u)K = +
σ=0 ∂x ∂x3 σ=0
= g222 (uxx )4 + (f222 + 3g122 ux + 3g12 )(uxx )3
(36.14)
+ 3(g112 u2x + f122 ux − 2g22 ux + g11 ux + f12 − g2 )(uxx )2
+ ux (g111 u2x + 3f112 ux − 9g12 ux + 3f11 − 6g1 − 3f22 )(uxx )
+ u2x [3(g2 − f12 ) + (f111 − 3g11 )ux ] = 0
Equating to zero the coefficients of uxx , we obtain the system of equations
g222 = 0, f222 + 3g122 ux + 3g12 = 0,

g112 u2x + f122 ux − 2g22 ux + g11 ux + f12 − g2 = 0,
(36.15)
g111 u2x + 3f112 ux − 9g12 ux + 3f11 − 6g1 − 3f22 = 0,
3(g2 − f12 ) + (f111 − 3g11 )ux = 0
Let us now briefly sketch the solution of the above system.

The first equation, g222 = 0, admits a general solution
g(u, ux ) = F (u)u2x + G(u)ux + H(u) (36.16)
where F (u), G(u) and H(u) are functions to be determined. Since
g2 = 2F (u)ux + G(u), g12 = 2F 0 (u)ux + G0 (u), g122 = 2F 0 (u)
459
the second equation of (36.15), f222 + 3g122 ux + 3g12 = 0, can be written as
f222 = −12F 0 (u)ux − 3G0 (u) (36.17)
and upon successive integrations with respect to ux , we get
f22 = −6F 0 (u)u2x − 3G0 (u)ux + L(u)

3
f2 = −2F 0 (u)u3x − G0 (u)u2x + L(u)ux + M (u)
2
and thus
1 1 1
f (u, ux ) = − F 0 (u)u4x − G0 (u)u3x + L(u)u2x + M (u)ux + N (u) (36.18)
2 2 2
Since (using the expressions (36.16) and (36.18) and differentiating properly)
g2 = 2F (u)ux + G(u), g22 = 2F (u), g11 = F 00 (u)u2x + G00 (u)ux + H 00 (u),

3
g112 = 2F 00 (u)ux + G00 (u), f12 = −2F 00 (u)u3x − G00 (u)u2x + L0 (u)ux + M 0 (u),
2
00 2 00 0
f122 = −6F (u)ux − 3G (u)ux + L (u)
the third equation g112 u2x +f122 ux −2g22 ux +g11 ux +f12 −g2 = 0 of the system
(36.15) takes on the form (upon arrangement in powers of ux )
5
− 5F 00 (u)u3x − G00 (u)u2x
2 (36.19)
+ {H 00 (u) − 6F (u) + 2L0 (u)}ux + M 0 (u) − G(u) = 0
Equating to zero all the coefficients of the powers of ux , we obtain from

(36.19) the system
F 00 (u) = 0, G00 (u) = 0, H 00 (u) − 6F (u) + 2L0 (u) = 0, M 0 (u) = G(u)
From the above system we get
F (u) = C1 u + C2 , G(u) = C3 u + C4 , M 0 (u) = C3 u + C4

(36.20)
H 00 (u) + 2L0 (u) = 6(C1 u + C2 )
Therefore
F (u) = C1 u + C2 (36.21)
460
G(u) = C3 u + C4 (36.22)
1
M (u) = C3 u2 + C4 u + C5 (36.23)
2
and
H 00 (u) + 2L0 (u) = 6(C1 u + C2 ) (36.24)
Let us now collect everything so far. We have the following expressions for
g(u, ux ) and f (u, ux ):
g(u, ux ) = (C1 u + C2 )u2x + (C3 u + C4 )ux + H(u) (36.25)

1 1 1
f (u, ux ) = − C1 u4x − C3 u3x + L(u)u2x
2 2 2 (36.26)
1
2
+ C3 u + C4 u + C5 ux + N (u)
2
The fourth equation g111 u2x + 3f112 ux − 9g12 ux + 3f11 − 6g1 − 3f22 = 0 of the
system (36.15), taking into account
g111 = H 000 (u), f112 = L00 (u)ux + C3 , g12 = 2C1 ux + C3

1
f11 = L00 (u)u2x + C3 ux + N 00 (u), g1 = C1 u2x + C3 ux + H 0 (u), (36.27)
2
f22 = −6C1 u2x − 3C3 ux + L(u)
becomes
9
H 000 (u) + L00 (u) − 6C1 u2x + 3{N 00 (u) − 2H 0 (u) − L(u)} = 0 (36.28)
2
Equating to zero the coefficients of ux , we obtain from the above equation
the system
9
H 000 (u) + L00 (u) − 6C1 = 0, N 00 (u) − 2H 0 (u) − L(u) = 0 (36.29)
2
Upon differentiation of (36.24) and combining the resulting equation with
the first of (36.29), we obtain L00 (u) = 0 and then
L(u) = C6 u + C7 (36.30)
461
Going back to (36.24) and taking into account the above expression of L(u),
we get H 00 (u) = 6C1 u + 6C2 − 2C6 from which we get
H(u) = C1 u3 + (3C2 − C6 )u2 + C8 u + C9 (36.31)
From the second of (36.29), taking into account the explicit expressions of
L(u) and H(u), we get the equation determining N (u):
N 00 (u) = 6C1 u2 + 3(4C2 − C6 )u + C7 + 2C8
Integrating the above equation we obtain the explicit value of the function
N (u):
1 1 1
N (u) = C1 u4 + (4C2 − C6 )u3 + (C7 + 2C8 )u2 + C10 u + C11 (36.32)
2 2 2
Let us now list the explicit values of the functions g(u, ux ) and f (u, ux ) using
(36.25) and (36.26) respectively:
g(u, ux ) = (C1 u + C2 )u2x + (C3 u + C4 )ux

(36.33)
+ C1 u3 + (3C2 − C6 )u2 + C8 u + C9
1 1 1
f (u, ux ) = − C1 u4x − C3 u3x + (C6 u + C7 )u2x
2 2 2
1 1 1
+ C3 u2 + C4 u + C5 ux + C1 u4 + (4C2 − C6 )u3 (36.34)
2 2 2
1
+ (C7 + 2C8 )u2 + C10 u + C11
2
The last equation 3(g2 − f12 ) + (f111 − 3g11 )ux = 0 of the system (36.15),
taking into account
g2 = 2(C1 u + C2 )ux + C3 u + C4 , f12 = C6 ux + C3 u + C4 ,

(36.35)
f111 = 12C1 u + 3(4C2 − C6 ), g11 = 6C1 u + 2(2C2 − C6 )
becomes an identity.
Therefore the system is consistent and the values of g(u, ux ) and f (u, ux ) are
given by (36.33) and (36.34) respectively.
Example 2. We consider next the equation
utt = uxxxx + uxx + u2x + u2 + au (36.36)
462
We shall prove that this equation admits the generalized symmetry given by
σ = uxxx + ux (36.37)
It suffices to prove that the condition
σ 0 (u)K =0 (36.38)
σ=0
is satisfied. Since
∂σ ∂σ ∂K ∂σ ∂ 2 K ∂σ ∂ 3 K
σ 0 (u)K = K+ + + (36.39)
∂u ∂ux ∂x ∂uxx ∂x2 ∂uxxx ∂x3
Example 3. We shall determine some general solutions of the nonlinear
diffusion equation
ut = (∆(u)ux )x (36.40)
assuming that it admits the generalized symmetry
σ = uxx + H(u)u2x + F (u)ux + G(u) (36.41)
In this case
K = (∆(u)ux )x = ∆0 (u)u2x + ∆(u)uxx (36.42)
We shall determine the coefficients H(u), F (u) and G(u) provided that the
condition
σ 0 (u)K =0 (36.43)
σ=0
is satisfied. We have the following expression for σ 0 (u)K

∂σ ∂σ ∂K ∂σ ∂ 2 K
σ 0 (u)K = K+ + (36.44)
∂u ∂ux ∂x ∂uxx ∂x2
Since (the prime denotes differentiation with respect to u)
∂σ ∂σ
= H 0 (u)u2x + F 0 (u)ux + G0 (u), = 2H(u)ux + F (u),
∂u ∂ux
(36.45)
∂σ
=1
∂uxx
463
and
∂K
= ∆00 (u)(ux )3 + 3∆0 (u)ux uxx + ∆(u)uxxx (36.46)
∂x
∂ 2K
= ∆000 (u)(ux )4 + 6∆00 (u)(ux )2 uxx
∂x2 (36.47)
+ 3∆0 (u)(uxx )2 + 4∆0 (u)ux uxxx + ∆(u)uxxxx
equation (36.44) becomes (arranged in powers of ux )
σ 0 (u)K =
= {H 0 (u)∆0 (u) + 2H(u)∆00 (u) + ∆000 (u)}u4x
+ {F (u)∆00 (u) + F 0 (u)∆0 (u)}u3x
+ {[6H(u)∆0 (u) + 6∆00 (u) + H 0 (u)∆(u)]uxx + G0 (u)∆0 (u)}u2x
+ {[3F (u)∆0 (u) + F 0 (u)∆(u)]uxx + 2[H(u)∆(u) + 2∆0 (u)]uxxx }ux
+ G0 (u)∆(u)uxx + ∆(u)uxxxx
(36.48)
We next have to take into account the condition σ = 0. We thus have to

substitute in (36.48)
uxx = −H(u)u2x − F (u)ux − G(u) (36.49)
and the differential consequences of the above relation
uxxx = −∆0 (u)u3x − 2H(u)ux uxx − F 0 (u)u2x

(36.50)
− F (u)uxx − G0 (u)ux
uxxxx = −H 00 (u)u4x − 5H 0 (u)u2x uxx − 2H(u)u2xx − 2H(u)ux uxxx

− F 00 (u)u3x − 3F 0 (u)ux uxx − F (u)uxxx − G00 (u)u2x (36.51)
− G0 (u)uxx
464
wherever they occur.
We then obtain the equation
σ 0 (u)K =
σ=0
= {∆000 − 4H∆00 + 5H 2 ∆0 − 3H 0 ∆0 − H 00 ∆ + 4H 0 H∆ − 2H 3 ∆}u4x
+ {4H 0 F ∆ + 2HF 0 ∆ + 9HF ∆0 − 4H 2 F ∆ − 5F ∆00
(36.52)
− F 00 ∆ − 3F 0 ∆0 }u3x
+ {−3G0 ∆0 + 8HG∆0 + 4F 2 ∆0 − 4H 2 G∆ + 4H 0 G∆ − 2HF 2 ∆
+ 2F F 0 ∆ − 6G∆00 − G00 ∆}u2x
+ {7F G∆0 + 2F 0 G∆ − 4HF G∆}ux + 3G2 ∆0 − 2HG2 ∆ = 0
Equating to zero all the coefficients of the different powers of ux , we obtain
the following system of ordinary differential equations
∆000 − 4H∆00 + 5H 2 ∆0 − 3H 0 ∆0 − H 00 ∆ + 4H 0 H∆ − 2H 3 ∆ = 0 (36.53)
4H 0 F ∆ + 2HF 0 ∆ + 9HF ∆0 − 4H 2 F ∆ − 5F ∆00
(36.54)
− F 00 ∆ − 3F 0 ∆0 = 0
− 6G∆00 − 3G0 ∆0 + 8HG∆0 + 4F 2 ∆0 − 4H 2 G∆ + 4H 0 G∆
(36.55)
− 2HF 2 ∆ + 2F F 0 ∆ − G00 ∆ = 0
7F G∆0 + 2F 0 G∆ − 4HF G∆ = 0 (36.56)
3G2 ∆0 − 2HG2 ∆ = 0 (36.57)
Example 4. We can check that equation
ut = uxx + (α + β(ln(u)) − γ 2 (ln(u))2 )u (36.58)
admits the generalized symmetry
σ = uxx − γux − u−1 u2x (36.59)
Since σ = 0 admits a general solution
u(x, t) = exp[F (t) + G(t)eγx ] (36.60)
using the above ansatz into the original equation, we obtain the system of
two decoupled differential equations
dF
= α + βF − γ 2 F 2 (36.61)
dt
465
dG
= (β + γ 2 − 2γ 2 F )G (36.62)
dt
The above system admits the solutions
(i) If k≡β 2 + 4αγ 2 > 0 then
√ k t 2
u(x, t) = C cos exp(γx + γ 2 t)
2 √ (36.63)
1 √ kt
+ 2 β − k tan
2γ 2
(ii) If k≡β 2 + 4αγ 2 < 0 then
√
−k t 2
u(x, t) = C cosh exp(γx + γ 2 t)
2 √ (36.64)
1 √ −k t
+ 2 β − −k tan
2γ 2
(iii) If k≡β 2 + 4αγ 2 = 0 then
exp(γx + γ 2 t) 1
u(x, t) = C 2
+ 2 (βt + 2) (36.65)
t 2γ t
References
[1] A. S. Fokas: ”A symmetry approach to exactly solvable evolution equa-
tions”, J. Math. Phys. 21 (1980) 1318-1325
[2] R. Z. Zhdanov: ”Conditional Lie-Bäcklund symmetry and reduction of
evolution equations”, J. Phys. A 28 (1995) 3841-3850
Chapter XIV
Contact Symmetries
37 Contact Symmetries.
Theory and an Example
Let a one-parameter transformation be given by
x̂ = f (x, y, p; ), ŷ = g(x, y, p; ), p̂ = h(x, y, p; ) (37.1)
466
with initial conditions
x = f (x, y, p; = 0), y = g(x, y, p; = 0), p = h(x, y, p; = 0) (37.2)
where p = y 0 . We consider next the infinitesimal version of the above trans-

formations
x̂ = x + ·ξ(x, y, p) + O(2 )
ŷ = y + ·η(x, y, p) + O(2 ) (37.3)
p̂ = p + ·ζ(x, y, p) + O(2 )
An infinitesimal contact transformation is defined if there exists a function
W = W (x, y, p), called characteristic function, such that
ξ = −Wp , η = W − pWp , ζ = Wx + pWy (37.4)
A second-order ODE
F (x, y, y 0 = p, y 00 ) = 0 (37.5)
is said to be invariant under a contact transformation, if the following con-

dition is being satisfied
∂F ∂F ∂F ∂F
ξ +η +ζ + ζ2 00 = 0 (37.6)
∂x ∂y ∂p ∂y
wherever F = 0. The function ζ2 is defined by
∂ζ ∂ζ 00 ∂ζ 00 ∂ξ
∂ξ 00 ∂ξ

ζ2 = +p +y −y +p +y (37.7)
∂x ∂y ∂p ∂x ∂y ∂p
Example. The Harmonic Oscillator. We consider the differential equa-
tion of the harmonic oscillator
F = y 00 + y = 0 (37.8)
Since
∂F ∂F ∂F ∂F
= 0, = 1, = 0, =1 (37.9)
∂x ∂y ∂p ∂y 00
equation (37.6) is equivalent to
η + ζ2 = 0 (37.10)
467
We calculate next ζ2 using the expression (37.7). Using (37.4), we find the
partial derivatives needed:
∂ζ ∂ζ ∂ζ
= Wxx + pWxy = Wxy + pWyy = Wxp + pWyp + Wy
∂x ∂y ∂p
(37.11)
∂ξ ∂ξ ∂ξ
= −Wpx = −Wpy = −Wpp
∂x ∂y ∂p
Therefore using the above expressions and (37.7), we obtain
ζ2 = Wxx + pWxy + p(Wxy + pWyy ) + y 00 (Wxp + pWyp + Wy )

− y 00 [−Wpx + p(−Wpy ) + y 00 (−Wpp )] (37.12)
2 00 00 2
= Wxx + 2pWxy + p Wyy + y (2Wxp + 2pWyp + Wy ) + (y ) Wpp
Using the above expression for ζ2 and the second expression of (37.4), equa-
Wxx +2pWxy +p2 Wyy −pWp +y 2 Wpp −2yWxp −2ypWpy −yWy +W = 0 (37.13)
where we have also made use of the substitution y 00 = −y.

Equation (37.13) has to be solved in order to determine the unknown char-
acteristic function W = W (x, y, p).
References
[1] F. Schwartz: ”Contact symmetries of the harmonic oscillator”,
J. Phys. A 16 (1983) L133-L135
Chapter XV
Conservation Laws
38 Variational Symmetries.
Let us first consider a one-dimensional Lagrangian
L(x, ẋ, t) (38.1)
468
where x is the position coordinate and t the time. The notation ẋ is used to
denote the derivative with respect to time:
dx
ẋ = (38.2)
dt
We also consider the one-parameter family of transformations
x̂ = x̂(x, t; ), x̂( = 0) = x

(38.3)
t̂ = t̂(x, t; ), t̂( = 0) = t
We introduce the notation

∂ x̂ ∂ t̂
X= , T = (38.4)
∂ =0 ∂ =0
A one-dimensional family of transformations is a variational symmetry if

dx̂ dt̂ dx d
L x̂, , t̂ = L x, , t + F (x, t; ) (38.5)
dt̂ dt dt dt
where F is some function.
A transformation satisfying (38.5) maps any solution of the Euler-Lagrange
equation into another solution. This follows from the fact that the Euler-
Lagrange equation determines the local extrema of the integral
Z t2
L(x, ẋ, t) dt (38.6)
t1
with fixed end-points (x1 , t1 ), (x2 , t2 ).

We differentiate (38.5) with respect to and then put = 0:

∂L ∂ x̂ ∂L ∂ t̂ ∂L ∂ dx̂ dt̂
· + · + ·
∂ x̂ ∂ =0 ∂ t̂ ∂ =0 ∂ ẋˆ ∂ dt̂ =0 dt =0 (38.7)
∂ dt̂ ∂ dF
+ L(x, ẋ, t) =
∂ dt =0 ∂ dt =0
We evaluate next the various quantities appearing in (38.7). We have in the
limit →0:
∂L ∂ x̂ ∂L
· = ·X (38.8)
∂ x̂ ∂ =0 ∂x
469
∂L ∂ t̂ ∂L
· = ·T (38.9)
∂ t̂ ∂ =0 ∂t

∂L ∂ dx̂ ∂L ∂ dx̂
· = · (38.10)
∂ ẋˆ ∂ dt̂ =0 ∂ ẋ ∂ dt̂ =0
and
∂ ∂

∂ dx̂
dt̂ dx̂ − dx̂ dt̂
= ∂ ∂
∂ dt̂ =0 (dt̂) 2
=0
∂ x̂ ∂ t̂ ∂ x̂ ∂ t̂
dt̂·d − dx̂·d d dx̂ d (38.11)
= ∂ ∂ = ∂ − · ∂
(dt̂)2 =0 d t̂ =0 dt̂ =0 dt̂ =0
dX dT
= − ẋ
dt dt
It is obvious that
dt̂
=1 (38.12)
dt =0
We have further
∂ dt̂ ∂ ∂ t̂ ∂ t̂ ∂x
= +
∂ dt =0 ∂ ∂t ∂x ∂t =0
∂ ∂ t̂ ∂ ∂ t̂
= + ẋ (38.13)
∂t ∂ =0 ∂ ∂x =0
d ∂ t̂ d
= + ẋ·0 = T
dt ∂ =0 dt
and similarly
∂ dF d ∂F d ∂F
= = G, G= (38.14)
∂ dt =0 dt ∂ =0 dt ∂ =0
Upon substitution of (38.8)-(38.14) into (38.7), we obtain the equation
∂L ∂L ∂L dX dT dT dG
X· + T· + · − ẋ +L = (38.15)
∂x ∂t ∂ ẋ dt dt dt dt
470
From the above equation we can determine the infinitesimals X and T .
Let us now recall the Euler-Lagrange equation of motion
d ∂L ∂L
= (38.16)
dt ∂ ẋ ∂x
Using the chain rule
dL ∂L ∂x ∂L ∂ ẋ ∂L
= + +
dt ∂x ∂t ∂ ẋ ∂t ∂t
∂L ∂L ∂L d ∂L ∂L ∂L
= ẋ + ẍ + = ẋ + ẍ + (38.17)
∂x ∂ ẋ ∂t dt ∂ ẋ ∂ ẋ ∂t
d ∂L ∂L
= ẋ +
dt ∂ ẋ ∂t
∂L
and solving with respect to , we obtain the equation
∂t
∂L d ∂L
= L− ẋ (38.18)
∂t dt ∂ ẋ
∂L d ∂L
In (38.17) we have substituted by because of the Euler-Lagrange
∂x dt ∂ ẋ
equation (38.16).
∂L
At this stage we substitute given by (38.18) into (38.15). We thus derive
∂t
the equation
∂L d ∂L ∂L dX dT dT dG
X· + T· L− ẋ + · − ẋ +L = (38.19)
∂x dt ∂ ẋ ∂ ẋ dt dt dt dt
∂L d ∂L
The above equation can be written as, after substituting by , thanks
∂x dt ∂ ẋ
to the Euler-Lagrange equation (38.16),
d ∂L d ∂L ∂L dX dT dT dG
X· +T· L− ẋ + · − ẋ +L = (38.20)
dt ∂ ẋ dt ∂ ẋ ∂ ẋ dt dt dt dt
It can also be expressed as

d ∂L ∂L dX dL dT
X· + + T +L
dt ∂ ẋ ∂ ẋ dt dt dt
(38.21)
∂L dT d ∂L
dG
− ẋ +T ẋ − =0
∂ ẋ dt dt ∂ ẋ dt
471
which is equivalent successively to

d ∂L d d ∂L dG
X + (L·T ) − T ẋ − =0
dt ∂ ẋ dt dt ∂ ẋ dt

d ∂L ∂L
X + L·T − T ẋ −G =0
dt ∂ ẋ ∂ ẋ

d ∂L ∂L
X + T L − ẋ −G =0 (38.22)
dt ∂ ẋ ∂ ẋ
From the last equation we conclude that the quantity
∂L ∂L
X + T L − ẋ −G (38.23)
∂ ẋ ∂ ẋ
is a constant of the motion.
Example. We consider the Lagrangian of a single particle moving in a
gravitational field
1
L = mẋ2 − mgx (38.24)
2
We shall determine the generators of the Lie algebra of symmetries and the
constants of the motion.
Solution. The generators of the Lie algebra of symmetries will be deter-
mined using equation (38.15). Going over to the more traditional notation,
i.e.
X→η and T →ξ (38.25)
we write equation (38.15) as

∂L ∂L ∂L dη dξ dξ dG
η· + ξ· + · − ẋ +L = (38.26)
∂x ∂t ∂ ẋ dt dt dt dt
For the Lagrangian given by (38.24), we calculate the partial derivatives
∂L ∂L ∂L
= −mg, = mẋ, =0 (38.27)
∂x ∂ ẋ ∂t
Using also the expression for the total derivative operator
d ∂ ∂
= + ẋ (38.28)
dt ∂t ∂x
472
we have
dη ∂η ∂η dξ ∂ξ ∂ξ dG ∂G ∂G
= + ẋ , = + ẋ , = + ẋ (38.29)
dt ∂t ∂x dt ∂t ∂x dt ∂t ∂x
Using the expressions (38.27), (38.29) and (38.24) into (38.26), we obtain the
equation

∂η ∂η ∂ξ ∂ξ
η(−mg) + mẋ + ẋ − ẋ + ẋ
∂t ∂x ∂t ∂x
1 ∂ξ (38.30)
2 ∂ξ ∂G ∂G
+ mẋ − mgx + ẋ = + ẋ
2 ∂t ∂x ∂t ∂x
We then equate to zero all the coefficients of ẋ in the above equation (since
the equation should be satisfied for every ẋ). We find that
∂η ∂ξ ∂G
m − mgx = (38.31)
∂t ∂x ∂x
The coefficient of ẋ2 is
∂η ∂ξ 1 ∂ξ
m −m + m =0 (38.32)
∂x ∂t 2 ∂t
The coefficient of ẋ3 is
∂ξ 1 ∂ξ
−m + m =0 (38.33)
∂x 2 ∂x
The constant term (the term which is independent of ẋ) is
∂ξ ∂G
−mgη − mgx = (38.34)
∂t ∂t
Equations (38.31)-(38.34) are the determining equations of the symmetries.
∂ξ
From equation (38.33) we get that = 0 and thus
∂x
ξ = ξ(t) (38.35)
∂η 1 ∂ξ
Equation (38.32), because of (38.35) takes on the form − = 0 and
∂x 2 ∂t
by integration
1
η = xξ 0 (t) + α(t) (38.36)
2
473
From (38.31), taking into account (38.35) and (38.36), we obtain the equation
1 ∂G
mxξ 00 (t) + mα0 (t) = (38.37)
2 ∂x
The equation for the constant term (38.34) becomes
3 ∂G
− mgxξ 0 (t) − mgα(t) = (38.38)
2 ∂t
∂ 2G ∂ 2G
Using Euler’s Theorem = , we obtain combining (38.37) and
∂t∂x ∂x∂t
(38.38), that
1 000 3
xξ (t) + α00 (t) = − gξ 0 (t) (38.39)
2 2
Since the above equation is satisfied for every x, we should have the system
3
ξ 000 (t) = 0 and α00 (t) = − gξ 0 (t) (38.40)
2
1 3
ξ(t) = a1 + a2 t + a3 t2 , α(t) = − a3 gt3 − a2 gt2 + a4 t + a5 (38.41)
2 4
Using the above expessions and equation (38.36) for η, we get the following
expresion
1 1 3
η(x, t) = x(a2 + 2a3 t) − a3 gt3 − a2 gt2 + a4 t + a5 (38.42)
2 2 4
Therefore the generator X of the symmetries of the Lagrangian L of a particle
moving freely in a gravitational field is given by
∂ ∂ ∂
X=ξ +η = (a1 + a2 t + a3 t2 )
∂t ∂x ∂t
(38.43)
1 1 3 3 2 ∂
+ x(a2 + 2a3 t) − a3 gt − a2 gt + a4 t + a5
2 2 4 ∂x

∂ ∂ 1 3 2 ∂
X = a1 + a2 t + x − gt
∂t ∂t 2 4 ∂x
(38.44)
∂ 1 ∂ ∂ ∂
+ a3 t2 + xt − gt3 + a4 t + a5
∂t 2 ∂x ∂x ∂x
474
Therefore the Lie algebra of symmetries of the Lagrangian is being spanned
by the five generators
∂ ∂ 1 3 2 ∂
X1 = , X2 = t + x− gt
∂t ∂t 2 4 ∂x (38.45)
2 ∂ 1 3 ∂ ∂ ∂

3 4 5
X =t + xt − gt , X =t , X =
∂t 2 ∂x ∂x ∂x
The constants G of the motion are determined by solving the system (38.37)
and (38.38), taking into account (38.41). The system is equivalent to
∂G 3 3
= − a2 mgt + a3 m x − gt2 + a4 m
∂x 2 2 (38.46)
∂G 3 3 2 1 3
= a2 mg − x + gt + a3 mg −3xt + gt − a4 mgt − a5 mg
∂t 2 4 2
Solving the above system, we obtain
3 1
G = − a2 mgxt + a3 m(x2 − 3gxt2 ) + a4 mx
2 2 (38.47)
1 1 1
+ a3 mg 2 t4 − a4 mgt2 − a5 mgt + C
8 2 2
Alternatively, the constant of the motion can be calculated equating (38.23)
to a constant. In that case we have to write G as
∂L ∂L
G=η + ξ L − ẋ (38.48)
∂ ẋ ∂ ẋ
and then use (38.41) and (38.42).
39 The Inverse Lagrangian Problem

Let us consider the equation
ẍ = F (x, ẋ, t) (39.1)
We want to find a function L(x, ẋ, t) such that the Euler-Lagrange equation
d ∂L ∂L
= (39.2)
dt ∂ ẋ ∂x
475
be equivalent to equation (39.1).
d
Taking into account that is a total derivative,
dt
d ∂ ∂ ∂
= + ẋ + ẍ (39.3)
dt ∂t ∂x ∂ ẋ
∂ ∂ ∂ ∂L ∂L
+ ẋ + ẍ =
∂t ∂x ∂ ẋ ∂ ẋ ∂x
The above equation, taking also into account that ẍ = F (x, ẋ, t), is equivalent
to
∂ ∂L ∂ ∂L ∂ 2L ∂L
+ ẋ + F (x, ẋ, t) 2 = (39.4)
∂t ∂ ẋ ∂x ∂ ẋ ∂ ẋ ∂x
The above is an equation which should be true for every x, ẋ and t.
Differentiation of (39.4) with respect to ẋ, we obtain
∂ ∂ 2L ∂ ∂L ∂ ∂ 2 L ∂F ∂ 2 L ∂ ∂ 2L ∂ ∂L
2
+ + ẋ 2
+ 2
+F 2
= (39.5)
∂t ∂ ẋ ∂x ∂ ẋ ∂x ∂ ẋ ∂ ẋ ∂ ẋ ∂ ẋ ∂ ẋ ∂ ẋ ∂x
∂ ∂L ∂ ∂L
After canceling the terms and and introducing the function M
∂x ∂ ẋ ∂ ẋ ∂x
by
∂ 2L
M= (39.6)
∂ ẋ2
we obtain from (39.5) the equation
∂ ∂ ∂ ∂F
M + ẋ M + F M = − M (39.7)
∂t ∂x ∂ ẋ ∂ ẋ
The above equation can be written in compact form as
dM ∂F
= −M (39.8)
dt ∂ ẋ
d
It is obvious that is a total derivative defined by (39.3) where ẍ has been
dt
substituted by F .
476
We thus have the following solution strategy:
1. Solve (39.8) for M .
2. Determine L from (39.6). That function will contain two arbitrary func-
tions of two variables x and t.
3. Find L up to a total derivative of a function with respect to time by
utilising equation (39.2).
Example. We consider the Emden-Fowler equation
2
ẍ + ẋ + x5 = 0 (39.9)
t
Determine the Lagrangian corresponding to this second order differential
equation.
Solution. Since
2
ẍ = − ẋ − x5 (39.10)
t
the function F is given by
2
F = − ẋ − x5 (39.11)
t
∂F 2
From the above function we evaluate the derivative = − . Therefore
∂ ẋ t
dM 2
= M (39.12)
dt t
A choice for the function M might be
M = t2 (39.13)
∂ 2L
= t2 (39.14)
∂ ẋ2
Integrating once the above equation, we obtain
∂L
= t2 ẋ + g(t, x) (39.15)
∂ ẋ
477
and one more integration gives
1
L = t2 ẋ2 + g(t, x)ẋ + h(t, x) (39.16)
2
Equation (39.4), taking into account (39.15) and (39.16), becomes
∂ 2 ∂ 2
t ẋ + g(t, x) + ẋ t ẋ + g(t, x)
∂t ∂t
2 ∂
+ − ẋ − x5 t2 ẋ + g(t, x) (39.17)
t ∂ ẋ
∂ 1 2 2
= t ẋ + g(t, x)ẋ + h(t, x)
∂x 2
The above equation is equivalent to the equation
∂g ∂g 2 ∂g ∂h
2tẋ + + ẋ − ẋ + x5 t2 = ẋ + (39.18)
∂t ∂x t ∂x ∂x
and this is also equivalent to
∂g ∂h
− x5 t2 = (39.19)
∂t ∂x
∂g ∂ x6 t2
= h+ (39.20)
∂t ∂x 6
We introduce a function Φ(x, t) by
∂Φ(x, t)
g= (39.21)
∂x
Therefore we obtain from (39.20) the equation
x6 t2 ∂Φ(x, t)
h+ = (39.22)
6 ∂t
We have thus expressed the unknown functions g(t, x) and h(t, x) in terms
of an auxiliary arbitrary function Φ(t, x):
∂Φ(x, t) 1 ∂Φ(x, t)
g(t, x) = , h = − x6 t2 + (39.23)
∂x 6 ∂t
478
The expression (39.16) of the Lagrangian, because of (39.23), gives
1 ∂Φ(x, t) 1 ∂Φ(x, t)
L = t2 ẋ2 + ẋ + h(t, x) − x6 t2 +
2 ∂x 6 ∂t
1 2 2 1 6 2 ∂Φ(x, t) ∂Φ(x, t)
L = t ẋ − x t + + ẋ (39.24)
2 6 ∂t ∂x
or equivalently
1 1 d
L = t2 ẋ2 − x6 t2 + Φ(x, t) (39.25)
2 6 dt
d
where is the total derivative. We thus see that the Lagrangian corre-
dt
sponding to the Emden-Fowler equation (39.9) is given by
1 1
L = t2 ẋ2 − x6 t2 (39.26)
2 6
modulo the total time derivative of an arbitrary function.
Let us now try to find the general solution of (39.9) using the symmetries
of the Lagrangian (39.26) and the expression (38.23) for the constant of
the motion. However we should find the infinitesimals of the variational
symmetry of the Lagrangian first. These are given by
ξ = 2at, η = −ax (39.27)
Using (38.23), we find that the constant of the motion is given by
3t3 ẋ2 + 3t2 xẋ + t3 x6 = k (39.28)
where k is an arbitrary constant. Solving this equation with respect to ẋ we
find
r
2 4
−t x± t4 x2 − t6 x6 + 4kt3
3
ẋ = (39.29)
2t3
We thus have been able to trasform the second order ODE (39.9) into a
first order one. However we can simplify more equation (39.29). We use the
substitution u = x2 t, i.e.
r
u(t)
x= (39.30)
t
479
Using this substitution, we can prove the following identities
1 du u
ẋ = − (39.31)
2tx dt t
and
4 2 4 6 6 3 3
4 3
t x − t x + 4kt = t 4k + u − u (39.32)
3 3
√ r
1 du u tx t 4
− = − 2 ± 2 4k + u − u3 (39.33)
2tx dt t 2t 2t 3
r
du 1 4
=± 4ku + u2 − u4 (39.34)
dt t 3
The above equation can be integrated by separation of variables.
Note. The Inverse Lagrangian Problem, i.e. the problem of determining
the Lagrangian (if it exists) corresponding to a given differential equation
(or a system of differential equations), was stated by Helmholtz who also
introduced the so-called Helmholtz Conditions, i.e. the conditions should
be satisfied for determining the Lagrangian given a differential equation. (see
Zwillinger below).
References
[1] G. F. T. del Castillo, C. A. Miron and R. I. B. Rojas: ”Variational
Symmetries of Lagrangians”. Rev. Mex. de Fisica E 59 (2013) 140-147
[2] D. Zwillinger: ”Handbook of Differential Equations”. Second Edition,
Academic Press 1992.
40 Hamilton’s Variational Principle

Let us suppose that there is a curve in the plane connecting two fixed points
x1 and x2 . This curve can be described by a function y(x). Nature picks
only one route for a system in going from x1 to x2 . We consider a functional
S expressed as an integral given by
Z x2
S= F (y(x), y 0 (x), x)dx (40.1)
x1
480
We want to find an equation satisfied by F which expresses the fact that the
evolution of the system follows the optimal path in going from x1 to x2 .
We introduce a parameter α and denote by y(α, x) all the neighboring func-
tions to the true y(x) trajectory. The function y(α, x) may be conveniently
represented by
y(α, x) = y(0, x) + α·η(x) (40.2)
where
η(x1 ) = η(x2 ) = 0 (40.3)
We then have a functional expressed by the parameter α :

Z x2
S(α) = F (y(α, x), y 0 (α, x), x)dx (40.4)
x1
where
∂
y 0 (α, x) = y(α, x) (40.5)
∂x
A necessary condition for S(α) to have an extremum is
∂S
=0 (40.6)
∂α α=0
The above condition means that S is independent of α in first order along

the path. This condition is also compatible to what was said above that the
evolution of a system follows the optimal path.
Performing the implied differentiation and assuming fixed end points, yields
Z x2
∂S ∂
= F (y(α, x), y 0 (α, x), x)dx
∂α ∂α x1
Z x2 (40.7)
∂F ∂y ∂F ∂y 0
= + 0 dx
x1 ∂y ∂α ∂y ∂α
We now have
∂y ∂y 0 ∂ 0 dη(x) dη(x)
= η(x), = y (x) + α = (40.8)
∂α ∂α ∂α dx dx
481
Because of the above relations, equation (40.7) becomes
Z x2
∂S ∂F ∂F dη(x)
= η(x) + 0 dx (40.9)
∂α x1 ∂y ∂y dx
Upon integration by parts for the second term of the above integrand,
Z x2 x2 Z x2
∂F dη(x) ∂F d ∂F
0 dx
dx = 0 η(x) − η(x) dx
x1 ∂y ∂y x1 x1 dx ∂y 0
Z x2 (40.10)
d ∂F
=− η(x) dx
x1 dx ∂y 0
where we have taken into account that

x2
∂F
η(x) = 0, since η(x2 ) = η(x1 ) = 0 (40.11)
∂y 0 x1
Therefore equation (40.9), because of (40.10), takes on the form

Z x2
∂S ∂F d ∂F
= η(x) − η(x) dx (40.12)
∂α x1 ∂y dx ∂y 0
Z x2
∂S ∂F d ∂F
= − η(x)dx (40.13)
∂α x1 ∂y dx ∂y 0
∂S
Using the condition = 0 and the fact that η(x) is an arbitrary func-
∂α α=0
tion, we obtain from equation (40.13) the equation
∂F d ∂F
− =0 (40.14)
∂y dx ∂y 0
This is a necessary condition for an extremum. Equation (40.14) is the

Euler-Lagrange equation in the plane.
Example. The brachistochrone problem. Consider a particle falling
from rest from a higher point to a lower point along a curve under the influ-
ence of gravity. The problem is to find the type of trajectory for which the
fall takes the shortest time.
482
Solution. Let the speed of the particle along the curve be v and an infinites-
imal arc length of the curve is ds where
r dy 2
p p
ds = dx2 + dy 2 = 1 + dx = 1 + (y 0 )2 dx (40.15)
dx
Conservation of energy implies
1
mgy = mv 2 (40.16)
2
i.e. potential energy transforms to kinetic energy. Solving with respect to v,
we obtain
p
v = 2gy (40.17)
ds
Since dt = , the time taken for a particle to travel from P1 to P2 is
v
Z P2 Z P2 s
ds 1 + (y 0 )2
t(P1 →P2 ) = = dx (40.18)
P1 v P1 2gy
We have to use the above equation as an extremum problem. The function
under variation is
s
1 + (y 0 )2
F (x, y, y 0 ) = (40.19)
2gy
We now have to use the Euler-Lagrange equations. We first remark that F
is independent of x. The total derivative of F with respect to x gives
dF ∂F ∂F ∂F
Dx F ≡ = + y0 + y 00 0 (40.20)
dx ∂x ∂y ∂y
∂F
Solving with respect to , we get
∂y
∂F dF ∂F ∂F
y0 = − − y 00 0 (40.21)
∂y dx ∂x ∂y
∂F d ∂F
Multiplying Euler-Lagrange − = 0 by y 0 gives
∂y dx ∂y 0
∂F d ∂F
y0 − y0 =0 (40.22)
∂y dx ∂y 0
483
∂F
Upon substitution of y 0 given by (40.21) into (40.22), we get
∂y
dF ∂F ∂F d ∂F
− − y 00 0 − y 0 =0 (40.23)
dx ∂x ∂y dx ∂y 0
∂F d ∂F
− + F − y0 0 = 0 (40.24)
∂x dx ∂y
∂F d ∂F
Since = 0, we obtain from (40.24) that F − y 0 0 = 0 and thus
∂x dx ∂y
∂F
F − y0 0 = C (40.25)
∂y
Since
s
1 + (y 0 )2 ∂F y0 1
F = , 0
= p ·√ (40.26)
2gy ∂y 1 + (y 0 )2 2gy
s
∂F 1 + (y 0 )2 y0 1
F − y 0 0 = C⇐⇒ − y0 p ·√ =C
∂y 2gy 0
1 + (y )2 2gy
1 + (y 0 )2 (y 0 )2

1
⇐⇒ p √ −√ =C
1 + (y 0 )2 2gy 2gy
1 1
⇐⇒ p ·√ =C
0
1 + (y ) 2 2gy
Squaring both sides of the last equation and rearranging, we obtain
1
1 + (y 0 )2 2gy = 2

(40.27)
C
and thus
dy 2 k 2 1
= − 1, k 2 = (40.28)
dx y 2gC 2
The solution of this last equation is given in parametric form by
1 1
x = k 2 (t − sint), y = k 2 (1 − cost) (40.29)
2 2
This curve is refer to literature as cycloid. Thus the particle will fall from
point P1 to point P2 in the shortest time, if it follows the path of a cycloid.
484
41 Some Auxiliary Operators and
their Identities
41.1. Euler-Lagrange Operator. The Euler-Lagrange operator is denoted
δ
by α and is defined by
δu
δ ∂ X ∂
α
= α
+ (−1)s Di1 Di2 · · ·Dis α (41.1)
δu ∂u s≥1
∂ui1 i2 ···is
where α labels the number of functions
α = 1, 2, 3, · · ·, m (41.2)
and the index i labels the number of independent variables (i = 1, 2, · · · , n).

The operators Di are defined as usual by
∂ α ∂ α ∂
Di = + u i + u ij + ··· (41.3)
∂xi ∂uα ∂uαj
Example. Suppose that α = 1, i.e. a single function, i = 3, i.e. 3-
independent variables (say x, y, z) and s = 2. In this case we have

δ ∂ ∂ ∂ ∂
= + (−1) Dx + Dy + Dz
δu ∂u ∂ux ∂uy ∂uz

∂ ∂ ∂
+ (−1)2 Dx2 + Dy2 + Dz2 (41.4)
∂uxx ∂uyy ∂uzz

∂ ∂ ∂
+ Dx Dy + Dx Dz + Dy Dz
∂uxy ∂uxz ∂uyz
Lemma 1. The following operator identity is true:

∂ ∂ ∂
Di , α ≡Di α − α Di = 0 (41.5)
∂u ∂u ∂u
Proof. This Lemma can be proved by direct computation. We have
∂ ∂
α ∂ α ∂
∂
Di α = + ui + uij + · · ·
∂u ∂xi ∂uα ∂uαj ∂uα
(41.6)
∂2 α ∂2 α ∂2
= + u i + uij + ···
∂xi ∂uα ∂(uα )2 ∂uαj ∂uα
485
and similarly
∂ ∂ ∂ α ∂ α ∂

Di = + ui α + uij α + · · ·
∂uα ∂uα ∂xi ∂u ∂uj
2 α
∂ ∂ui ∂ α ∂2
= + + u i
∂uα ∂xi ∂uα ∂uα ∂(uα )2
∂uα ∂ (41.7)
ij α ∂2
+ + u ij + ···
∂uα ∂uαj ∂uα ∂uαj
∂2 α ∂2 α ∂2
= + u i + uij + ···
∂xi ∂uα ∂(uα )2 ∂uαj ∂uα
where we have taken into account that

∂uαi ∂uαij
= = ··· = 0 (41.8)
∂uα ∂uα
∂ ∂
We thus see that Di − Di = 0, i.e.
∂uα ∂uα

∂
Di , α = 0 (41.9)
∂u
Lemma 2. The following operator equation holds true
δ
Di = 0 ∀α (41.10)
δuα
Theorem 1. A function f (x, u, u(1) , u(2) , · · ·, u(s) ) where
x = (x1 , x2 , · · ·, xn ), u = (u1 , u2 , · · ·, um ) (41.11)
is the divergence of a vector field
H = (h1 , h2 , · · ·, hn ) (41.12)
where
hi = hi (x, u, u(1) , u(2) , · · ·, u(s−1) ) (41.13)
i.e.
f = div(H)≡Di (hi ) (41.14)
486
if and only if
δf
=0 (41.15)
δuα
41.2. Lie-Bäcklund Operator. A Lie-Bäcklund operator X is defined by
∂ ∂ ∂ ∂
X = ξi i
+ η α α + ζiα α + ζiα1 i2 α + · · · (41.16)
∂x ∂u ∂ui ∂ui1 i2
where
ζiα = Di (wα ) + ξ j uαji , ζiα1 i2 = Di1 Di2 (wα ) + ξ j uαji1 i2 (41.17)
and similar expressions for the rest, where
wα = η α − ξ j uαj (41.18)
Note that in the Lie-Bäcklund operator the functions ξ and η also depend
on the derivatives of the differential variables, which is not the case when
dealing with point transformations.
The Lie-Bäcklund operator is the infinite order extension of the operator
∂ ∂
X = ξi i
+ ηα α (41.19)
∂x ∂u
The Lie-Bäcklund operator is also written as
∂ ∂ ∂
X = ξ i Di + w α α
+ Di (wα ) α + Di1 Di2 (wα ) α + · · · (41.20)
∂u ∂ui ∂ui1 i2
The Lie-Bäcklund operator truncates when acting on any differential func-

tion, i.e. it belongs to A. Indeed, let
∂ ∂
Xρ = ξρi i
+ ηρα α , ρ = 1, 2 (41.21)
∂x ∂u
be two Lie-Bäcklund operators (where the infinite order extension is always
understood). Their commutator is then given by
[X1 , X2 ] = X1 X2 − X2 X1
∂ ∂ (41.22)
= X1 (ξ2i ) − X2 (ξ1i ) + X 1 (η2
α
) − X 1 (η α
1 )
∂xi ∂uα
487
which is also a Lie-Bäcklund operator. This means that all Lie-Bäcklund
operators form an infinite-dimensional Lie algebra, which will be denoted by
the symbol LB .
41.3. The Operators N i .
Here we present two last operators needed for our analysis to find conserva-
tion laws.
Definition. Given a Lie-Bäcklund operator X, we define the operators N i
(i = 1, 2, · · ·, n) by the formal sum
∂ X δ
N i = ξ i + wα α
+ Di1 Di2 · · ·Dis (wα ) α (41.23)
∂ui s≥1
δui1 ···is
The fundamental identity is given in the following Theorem, which is

verified by direct calculation:
Theorem 2. The Euler-Lagrange, Lie-Bäcklund and the associated operator
N i are connected by the operator identity:
∂
X + Di (ξ i ) = wα + Di N i (41.24)
∂uα
42 Euler-Lagrange Equations
in Several Variables
We first introduce the necessary notation:
x = (x1 , x2 , · · ·, xn ) where n is the number of variables (42.1)
u = (u1 , u2 , · · ·, um ) where m is the number of functions (42.2)

We consider a differential function L∈A called Lagrangian, of first order
in an arbitrary volume V (V ⊂Rn ), spanned by x = (x1 , x2 , · · ·, xn ) with a
boundary denoted by ∂V .
We denote by S[u] a functional defined by
Z
S[u] = L(x, u, u(1) )dx (42.3)
V
Consider a small variation h(x) in u which leads to a Lagrangian
L(x, u + h, u(1) + h(1) ) (42.4)
488
Taylor expansion up to first order in h, gives
∂L α ∂L α
L(x, u + h, u(1) + h(1) ) = L(x, u, u(1) ) + h + α hi + O(h2 ) (42.5)
∂uα ∂ui
The variation δS[u] defined as
Z
δS[u] = {L(x, u + h, u(1) + h(1) ) − L(x, u, u(1) )}dx (42.6)
V
has thus the form

Z
∂L α ∂L α
δS[u] = h + h dx (42.7)
V ∂u
α ∂uαi i
Integrating the second term by parts, we find
Z Z Z
∂L α ∂L
α
∂L
h
α i
dx = D i α
h dx − D i α
hα dx (42.8)
V ∂u i V ∂u i V ∂u i
leading to
Z Z
∂L ∂L α ∂L
δS[u] = α
− Di α
h dx + D i α
hα dx (42.9)
V ∂u ∂u i V ∂u i
At this point we remind the Divergence Theorem (Gauss-Ostrogradsky)

Z Z
Di f (x, u, u(1) , · · ·) dx = f (x, u, u(1) , · · ·) pi dx (42.10)
V ∂V
where p is the unit vector normal to ∂V with components pi , i = 1, 2, · · ·, n.

Using the Divergence Theorem, equation (42.9) can be written as
Z Z
∂L ∂L α ∂L α i
δS[u] = α
− Di α h dx + α
h p dx (42.11)
V ∂u ∂ui ∂V ∂ui
Provided that the functions hα (x) vanish on the boundary ∂V , we get

Z
∂L ∂L α
δS[u] = α
− Di α h dx (42.12)
V ∂u ∂ui
A function u = u(x) is an extremum of the variational integral (42.3) if
δS[u] = 0. For an arbitrary volume V and infinitesimal increment h(x) that
vanish on ∂V , we must have
∂L ∂L
α
− Di α = 0 (42.13)
∂u ∂ui
489
The variational derivative
δL ∂L ∂L
α
≡ α − Di α (42.14)
δu ∂u ∂ui
is called the Euler-Lagrange equation.
43 Noether’s Theorem
Let G be a one-parameter group of transformations
x̂i = f i (x, u; ), i = 1, 2, · · ·, n (43.1)
ûα = φα (x, u; ), α = 1, 2, · · ·, m (43.2)

with generator
∂ ∂
X = ξ i (x, u) i
+ η α (x, u) α (43.3)
∂x ∂u
A variational integral (42.3)
Z
S[u] = L(x, u, u(1) )dx (43.4)
V
is said to be invariant under G if

Z Z
L(x̂, û, û(1) )dx̂ = L(x, u, u(1) )dx (43.5)
V̂ V
where V̂ ⊂Rn is the volume obtained from V by transformation (43.1)-(43.2).

Lemma 1. The variational integral (43.4) is invariant under the group G if
and only if
X(L) + L Di (ξ i ) = 0 (43.6)
where
∂L ∂L ∂L
X(L) = ξ i i
+ η α α + ζiα α (43.7)
∂x ∂u ∂ui
with
ζiα = Di (η α ) − uαj Di (ξ j ) (43.8)
490
Proof. Under a change of variables, the volume element transforms as
dV̂ = J dV (43.9)
where J is the Jacobian of the transformation.
We first have
Z Z
L(x̂, û, û(1) )dx̂ = L(x̂, û, û(1) )J dx (43.10)
V̂ V
where J is the Jacobian of the transformation x̂i = f i (x, u; ):

J = det|Dj (x̂i )| (43.11)
Since both integrals are now taken over the same arbitrary volume V , equa-
tion (43.5) is equivalent to
L(x̂, û, û(1) )J = L(x, u, u(1) ) (43.12)
Consider the infinitesimal transformation
x̂i = xi + ξ i , ûα = uα + η α , ûαi = uαi + ζiα (43.13)
Upon Taylor expansion of L(x̂, û, û(1) ) around = 0, yields
∂L ∂xi

L(x̂, û, û(1) ) = L(x, u, u(1) ) +
∂xi ∂ =0
(43.14)
∂L ∂uα ∂L ∂uαi

+ + + O(2 )
∂uα ∂ =0 ∂uαi ∂ =0
and since
∂xi ∂uα ∂uα
i α i
=ξ, =η , = ζiα (43.15)
∂ =0 ∂ =0 ∂ =0

L(x̂, û, û(1) ) = L(x, u, u(1) )

∂L i ∂L α ∂L α
+ i
ξ + α
η + ζ + O(2 )
α i
∂x ∂u ∂ui
∂L (43.16)
∂L α ∂L α
= L(x, u, u(1) ) + i
ξ + α η + α ζi + O(2 )
i
∂x ∂u ∂ui
2
= L(x, u, u(1) ) + ·X(L) + O( )
491
Taylor expansion of J around = 0 yields
∂J
J = J|=0 + + O(2 )
∂ =0
∂ x̂
(43.17)

= det|Dj (xi )| + det Dj + O(2 )
∂ =0
i 2
= 1 + ·Di (ξ ) + O( )
Thus the left hand side of equation (43.12), using (43.16) and (43.17), yields
L(x̂, û, û(1) ) J = {L(x, u, u(1) ) + ·X(L) + O(2 )}

×{1 + ·Di (ξ i ) + O(2 )}
(43.18)
= L(x, u, u(1) ) + ·X(L) + ·LDi (ξ i ) + O(2 )
= L(x, u, u(1) ) + {X(L) + LDi (ξ i )} + O(2 )
The above expression should be equal to the right hand side of (43.12), i.e.
L(x, u, u(1) ) + {X(L) + LDi (ξ i )} = L(x, u, u(1) )
from which we get
X(L) + L·Di (ξ i ) = 0 (43.19)
Definition. A vector T = (T 1 , T 2 , · · ·, T n ) is a conserved vector if
Di (T i ) = 0 (43.20)
Theorem 1. Let the variational integral

Z
S[u] = L(x, u, u(1) )dx (43.21)
V
be invariant under the group G with generator

∂ ∂
X = ξ i (x, u) i
+ η α (x, u) α (43.22)
∂x ∂u
Then the vector T i ∈A defined by
∂L ∂L
T i = Lξ i + wα α
= Lξ i + (η α − ξ j uαj ) α (43.23)
∂ui ∂ui
492
is a conserved vector of the Euler-Lagrange equation, i.e. Di (T i ) = 0 on the
solutions of Euler-Lagrange equations.
Proof. Since the variational integral is invariant, according to Lemma 1 of
this section, we shall have
X(L) + LDi (ξ i ) = 0 (43.24)
Using the following expression for X

∂ ∂
X = ξ i Di + w α α
+ Di (wα ) α (43.25)
∂u ∂ui
straightforward calculations yield
∂L α ∂L
X(L) + LDi (ξ i ) = ξ i Di (L) + wα + Di (w ) α + LDi (ξ i )
∂uα ∂ui
∂L ∂L ∂L
= ξ i Di (L) + LDi (ξ i ) + wα α + Di wα α − wα Di
∂u ∂ui ∂uαi
(43.26)
α ∂L α ∂L
∂L
i
= Di (Lξ ) + Di w +w − Di
∂uαi ∂uα ∂uαi
∂L δL
= Di Lξ i + wα α + wα α
∂ui δu
δL
Since = 0 and also X(L) + LDi (ξ i ) = 0, we obtain from (43.26) that
δuα
∂L
Di Lξ i + wα α = 0 (43.27)
∂ui
According to the definition of the conserved vector, i.e. Di (T i ) = 0, we derive

from (43.27) that
∂L ∂L
T i = Lξ i + wα α
= Lξ i + (η α − ξ j uαj ) α (43.28)
∂ui ∂ui
is a conserved vector.
We have seen that we can add any divergence type function F to the La-
grangian. The invariance condition (43.19), X(L) + LDi (ξ i ) = 0, is then
replaced by the divergence relation
X(L) + LDi (ξ i ) = Di (B i ) (43.29)
493
Theorem 2. Let
L(x, u, u(1) , · · ·, u(s) )∈A (43.30)
be a differential function of any order. It is also supposed that the Euler-

Lagrange equations are satisfied,
δL
= 0, α = 1, 2, · · ·, m (43.31)
δuα
Let X be a symmetry such that
X(L) + LDi (ξ i ) = Di (B i ), B i ∈A (43.32)
Then equations (43.31) admit a conservation law
Di (C i ) = 0 (43.33)
defined by
C i = N i (L) − B i , i = 1, 2, · · ·, n (43.34)
where the Noether operators N i are defined in (41.23).

Proof. This Theorem follows directly from the fundamental identity (41.24).
Indeed,
Di (C i ) = Di N i (L) − Di B i = 0 (43.35)
from which we get
Di N i (L) = Di B i = X(L) + LDi (ξ i ) (43.36)
δL
where we used the fact that α = 0.
δu
43.2. Many Symmetries. If a differential equation admits r symmetries
X1 , X2 , · · ·, Xr (43.37)
of the form
∂ ∂
Xµ = ξµi (x, u) i
+ ηµα (x, u) α , µ = 1, 2, 3, · · ·, r (43.38)
∂x ∂u
494
then (43.28) provides r conserved vectors T1 , T2 , · · ·, Tr with components
∂L
Tµi = Lξµi + (ηµα − ξµj uαj ) , µ = 1, 2, 3, · · ·, r (43.39)
∂uαi
Example 1. Special Relativity. The motion of a free particle in Special

Relativity is described by the Lagrangian
p
L = −mc2 1 − β 2 (43.40)
where
|v|2
β2 = , |v|2 = (v 1 )2 + (v 2 )2 + (v 3 )2 (43.41)
c2
The position vector x of the particle has coordinates
x = (x1 , x2 , x3 ) (43.42)
and the velocity v has coordinates
v = (v 1 , v 2 , v 3 ) (43.43)
such that
dx
v= (43.44)
dt
where t is the time and x is our differential variables (t is the only independent
variable).
The Lagrangian is invariant under the Lorentz group with generators
∂ ∂ ∂ i ∂
X0 = , Xi = , Xij = xj − x
∂t ∂xi ∂xi ∂xj (43.45)
∂ 1 ∂
X0i = t i + 2 xi
∂x c ∂t
Applying Noether’s theorem to each one of these generators with the given
Lagrangian, we wish to find the corresponding conservation laws.
The generators have the form
∂ ∂
X = ξ(t, x) + η α (t, x) α , α = 1, 2, 3 (43.46)
∂t ∂x
495
The conserved quantity which corresponds to each generator is given by
∂L
T = ξL + (η α − ξv α ) (43.47)
∂v α
where
∂L 1
α
= mγv α , γ=p (43.48)
∂v 1 − β2
∂
(i) Let X0 = . In this case we have (comparing this generator to (43.46))
∂t
ξ = 1, η1 = η2 = η3 = 0 (43.49)
Equation (43.47) yields

∂L
T = L − vα = −mc2 γ −1 − mγ|v|2 = −mc2 γ (43.50)
∂v α
Putting T = −E, we get
E = mc2 γ (43.51)
i.e. conservation of Energy.

∂
(ii) Let Xi = . In this case we have (comparing this generator to (43.46))
∂xi
ξ = 0, η α = δiα (43.52)
Equation (43.47) yields, denoting the conserved vector T = (T 1 , T 2 , T 3 ):

∂L ∂L ∂L
T1 = = mv 1 γ, T2 = = mv 2 γ, T3 = = mv 3 γ (43.53)
∂v 1 ∂v 2 ∂v 3
We thus see that the conserved vector is
T = mγv (43.54)
i.e. conservation of relativistic momentum.

∂ ∂
(iii) For the generator Xij = xj i − xi j we have a rotational symmetry.
∂x ∂x
In this case
ξ = 0, η α = xj δαi − xi δαj (43.55)
496
∂L ∂L
T12 = x2 1
− x1 2 = mγ(x2 v 1 − x1 v 2 ) (43.56)
∂v ∂v
∂L ∂L
T23 = x3 2 − x2 3 = mγ(x3 v 2 − x2 v 3 ) (43.57)
∂v ∂v
∂L ∂L
T31 = x1 3 − x3 1 = mγ(x1 v 3 − x3 v 1 ) (43.58)
∂v ∂v
The conserved quantity is T = (T23 , T31 , T12 ). Putting T = −L, we get the
conserved vector
L = mγ(x×v) (43.59)
i.e. conservation of angular momentum.
∂ 1 ∂
(iv) For the generator X0i = t i + 2 xi we have the conserved generator
∂x c ∂t
T = (T 1 , T 2 , T 3 ). In this case for X01 we have
1 1
ξ= x, η 1 = t, η2 = η3 = 0 (43.60)
c2
1 1 1 1 1 ∂L m 1 1 1 1
T01 = L 2 x + t − 2 x v 1
= − x + t − 2 x v mγv 1
c c ∂v γ c
m 1 1 1 1 1 2 m
1 1 2 γ 2 1 1 2
= − x + tmγv − 2 x mγ(v ) = −x + tv γ − 2 x (v ) (43.61)
γ c γ c
1 2

m (v )
= tv 1 γ 2 − x1 1 + 2 γ 2
γ c
At this point we recall the definition of β 2 considering the v 1 component only,
(v 1 )2 (v 1 )2 2
i.e. β = 2 and thus 1 + 2 γ = 1 + β 2 γ 2 . On the other hand, since
2
c c
2 1
γ = , we get 1 + β γ = γ 2 . Therefore
2 2
1 − β2
(v 1 )2 2
1+ γ = γ2 (43.62)
c2
Because of the above equation, we obtain from (43.61) the equation
m 1 2
T01 = tv γ − x1 γ 2 = mγ tv 1 − x1 (43.63)
γ
497
Considering the similar generators, we see that the quantity
Q = mγ(tv − x) (43.64)
is conserved, which is actually the relative center of mass.
Example 2. The Direct Method. In this example we shall apply the
condition (43.20), i.e. Di (T i ) = 0 to find the conservation laws corresponding
to the equation
uxt + u2 = 0, u = u(x, t) (43.65)
The above equation corresponds to the Maxwellian distribution. The
procedure of obtaining the conservation laws directly from Di (T i ) = 0 is
called the direct method.
Solution. The condition Di (T i ) = 0 can be written in expended form as
Dt T 1 + Dx T 2 |uxt +u2 =0 = 0 (43.66)
where
T 1 = T 1 (t, x, u, ut , ux )
(43.67)
T 2 = T 2 (t, x, u, ut , ux )
Since
∂ ∂ ∂
Dt = + ut + utx + ···
∂t ∂u ∂ux
(43.68)
∂ ∂ ∂
Dx = + ux + uxt + ···
∂x ∂u ∂ut
we get
Dt T 1 + Dx T 2 =
∂T 1 ∂T 1 ∂T 1 ∂T 2 ∂T 2 ∂T 2
= + ut + utx + + ux + uxt (43.69)
∂t ∂u ∂ux ∂x ∂u ∂ut
∂T 1 ∂T 2 ∂T 1
∂T 2 ∂T 1
∂T 2
= + + ut + ux + utx +
∂t ∂x ∂u ∂u ∂ux ∂ut
Upon substitution utx → − u2 to take account of the equation utx + u2 = 0,
we find that Dt T 1 + Dx T 2 |uxt +u2 =0 = 0 is equivalent to
∂T 1 ∂T 2 ∂T 1 ∂T 2 1 ∂T 2
2 ∂T
+ + ut + ux −u + =0 (43.70)
∂t ∂x ∂u ∂u ∂ux ∂ut
498
The choice
1 u2x 2 u2t
T = α(t, x, u) + β(t, x, u), T = γ(t, x, u) + δ(t, x, u) (43.71)
2 2
u2x u2 u2 u2
αt + βt + γx t + δx + ut αu x + βu + ux γu t + δu
2 2 2 2
2
− u (αux + γut ) = 0

1 1 1 1
αt u2x + γx u2t + αu ut u2x + γu ux u2t + (βu − γu2 )ut
2 2 2 2 (43.72)
2
+ (δu − αu )ux + (βt + δx ) = 0
Equating to zero the coefficients of the various partial derivatives of the

function u, we obtain that
The coefficient of u2x is
αt = 0 (43.73)
The coefficient of u2t is
γx = 0 (43.74)
The coefficient of ut u2x is
αu = 0 (43.75)
The coefficient of ux u2t is
γu = 0 (43.76)
βu − γu2 = 0 (43.77)
δu − αu2 = 0 (43.78)
499
βt + δx = 0 (43.79)
Since αt = αu = 0 and γx = γu = 0 we have that
α = α(x) and γ = γ(t) (43.80)
From (43.77) we get βu = γ(t)u2 and by integration with respect to u
1
β = γ(t)u3 + ζ(t, x) (43.81)
3
From (43.78) we get δu = α(x)u2 and by integration with respect to u
1
δ = α(x)u3 + η(t, x) (43.82)
3
Equation (43.79), because of (43.81) and (43.82), takes on the form
1
(γt + αx )u3 + (ζt + ηx ) = 0 (43.83)
3
γt + αx = 0 and ζt + ηx = 0 (43.84)
We then have γt + αx = 0 ⇒ γt = a1 and αx = −a1 . Integrating the two
previous equations with respect to t and x respectively, we have
γ(t) = a1 t + a2 and α(x) = −a1 x + a3 (43.85)
Let us now collect everything together. We have found
u2x 1
T 1 = (−a1 x + a3 ) + (a1 t + a2 )u3 + ζ(t, x)
2 3 (43.86)
2
u 1
T 2 = (a1 t + a2 ) t + (−a1 x + a3 )u3 + η(t, x)
2 3
where ζ(t, x) and η(t, x) satisfy ζt + ηx = 0. We set ζ = η = 0 since they
have a trivial contribution to T 1 , T 2 . Writing (43.86) as
1 1 1 1
T 1 = a1 − xu2x + tu3 + a2 u3 + a3 u2x
2 3 3 2 (43.87)
1 1 1 1
2 2 3 2 3
T = a1 tut − xu + a2 ut + a3 u
2 3 2 3
500
we obtain from the above expressions the following conserved vectors
1 1 1 1
T 1 = − xu2x + tu3 , T 2 = tu2t − xu3 (43.88)
2 3 2 3
1 1
T 1 = u3 , T 2 = u2t (43.89)
3 2
and
1 1
T 1 = u2x , T 2 = u3 (43.90)
2 3
Example 3. The Noether’s approach. In this example we shall de-
termine the conservation laws for equation uxt + u2 = 0, u = u(x, t) using
Noether’s approach.
Solution. The first step in applying Noether’s approach is to determine the
symmetry generator X
∂ ∂ ∂
X = ξ(t, x, u) + τ (t, x, u) + η(t, x, u) (43.91)
∂x ∂t ∂u
i.e. the infinitesimals ξ(t, x, u), τ (t, x, u) and η(t, x, u), using the equation
X(L) + LDi (ξ i ) = Di B i (43.92)
Here L is the Lagrangian corresponding to equation uxt + u2 = 0 given by

1 1
L = − ut ux + u3 (43.93)
2 3
It is obvious that equation uxt + u2 = 0 is a consequence of the variational
derivative
δL
=0 (43.94)
δu
with L given by (43.93). Since L contains up to first order derivatives, we
have to consider the first prolongation of X, i.e.
∂ ∂ ∂
P r(1) X = ξ(t, x, u) + τ (t, x, u) + η(t, x, u)
∂x ∂t ∂u (43.95)
∂ ∂
+ ζ [t] (t, x, u, ut , ux ) + ζ [x] (t, x, u, ut , ux )
∂ut ∂ux
501
The coefficients ζ [t] and ζ [x] are evaluated using the expressions
ζ [t] = Dt η − ut Dt τ − ux Dt ξ
(43.96)
ζ [x] = Dx η − ut Dx τ − ux Dx ξ
Since
∂ ∂ ∂ ∂
Dt = + ut + ··· , Dx = + ux + ··· (43.97)
∂t ∂u ∂x ∂u
we find
ζ [t] = Dt η − ut Dt τ − ux Dt ξ
∂η ∂η ∂τ ∂τ ∂ξ ∂ξ (43.98)
= + ut − ut + ut − ux + ut
∂t ∂u ∂t ∂u ∂t ∂u
and
ζ [x] = Dx η − ut Dx τ − ux Dx ξ
∂η ∂η ∂τ ∂τ ∂ξ ∂ξ (43.99)
= + ux − ut + ux − ux + ux
∂x ∂u ∂x ∂u ∂x ∂u
Equation (43.92), taking into account (43.93) and (43.95) becomes
1 1 1 1
η u2 + ζ [t] − ux + ζ [x] − ut + − ut ux + u3 (Dt τ + Dx ξ)
2 2 2 3 (43.100)
1 2
= Dt B + Dx B
The above equation can be written in equivalent form as (taking into account
the explicit expressions for ζ [t] and ζ [x] )

2 1 ∂η ∂η ∂τ ∂τ ∂ξ ∂ξ
η u − ux + ut − ut + ut − ux + ut
2 ∂t ∂u ∂t ∂u ∂t ∂u

1 ∂η ∂η ∂τ ∂τ ∂ξ ∂ξ
− ut + ux − ut + ux − ux + ux
2 ∂x ∂u ∂x ∂u ∂x ∂u
1 (43.101)
1 3 ∂τ ∂τ ∂ξ ∂ξ
+ − ut ux + u + ut + + ux
2 3 ∂t ∂u ∂x ∂u
∂B 1 ∂B 1 ∂B 2
∂B 2
= + ut + + ux
∂t ∂u ∂x ∂u
the function u.
502
1 ∂ξ 1 ∂τ 1 ∂τ
We find that the coefficients of ut u2x , ux u2t , u2t , u2x are , , and
2 ∂u 2 ∂u 2 ∂x
1 ∂ξ
respectively. We thus get ξu = ξt = 0 and τu = τx = 0. Therefore
2 ∂t
ξ = ξ(x) and τ = τ (t) (43.102)
The coefficients of ux and ut are given by
1 ∂η 1 3 ∂ξ ∂B 2 1 ∂η 1 3 ∂τ ∂B 1
− + u = and − + u =
2 ∂t 3 ∂u ∂u 2 ∂x 3 ∂u ∂u
respectively. These equations, because of (43.102) become
1 ∂η ∂B 2 1 ∂η ∂B 1
− = and − = (43.103)
2 ∂t ∂u 2 ∂x ∂u
The constant term of (43.101) is
1 ∂τ ∂ξ ∂B 1 ∂B 2
η u 2 + u3 + = +
3 ∂t ∂x ∂t ∂x
From the above equation, equating to zero the coefficients of u, we get
∂τ ∂ξ ∂B 1 ∂B 2
η = 0, + = 0, + =0 (43.104)
∂t ∂x ∂t ∂x
Because of the first equation of the above system, we get from (43.103) that
∂B 2 ∂B 1
= 0 and = 0. Therefore
∂u ∂u
B 1 = F (x, t) and B 2 = G(x, t) (43.105)
The functions F and G satisfy (see the last of (43.104)) the equation
Ft (x, t) + Gx (x, t) = 0 (43.106)

∂τ ∂ξ
From the second of (43.104), i.e. + = 0, taking also into account
∂t ∂x
∂τ ∂ξ
(43.102), we get = a1 and = −a1 and by integration with respect to
∂t ∂x
t and x respectively, we have
τ = a1 t + a2 , ξ = −a1 x + a3 (43.107)
503
Taking into account the above values of η, ξ and τ , we obtain from (43.98)
and (43.99) that
ζ [t] = −a1 ut and ζ [x] = a1 ux (43.108)
The conserved vectors are determined using the expression
δL X ∂L
T i = Bi − ξiL − W α α − Di1 ···is (W α ) α (43.109)
δui s≥1
∂ui1 ···is
where
W α = η α − ξ j uαj (43.110)
We thus get the following expressions for the two conserved vectors
δL
T 1 = B 1 − τ L − (η − τ ut − ξux )
δut
(43.111)
δL
T 2 = B 2 − ξL − (η − τ ut − ξux )
δux
Furthermore we have
δL ∂L 1 δL ∂L 1
= = − ux and = = − ut (43.112)
δut ∂ut 2 δux ∂ux 2
We may take F = G = 0, i.e.
B1 = B2 = 0 (43.113)
since they contribute trivially to T 1 and T 2 .
From the expressions (43.111), taking into account (43.112), (43.113) and the
expressions for η, ξ and τ we have found previously, we obtain the following
expressions for the conserved vectors
1 1
T 1 = −(a1 t + a2 ) − ut ux + u3
2 3

1 (43.114)
− − (a1 t + a2 )ut − (−a1 x + a3 )ux − ux
2
and
1 1
T 2 = −(−a1 x + a3 ) − ut ux + u3
2 3

1 (43.115)
− − (a1 t + a2 )ut − (−a1 x + a3 )ux − ut
2
504
After performing the relevant operations, we obtain the following two con-
served vectors
1 1
T 1 = − (a1 t + a2 )u3 − (−a1 x + a3 )(ux )2 (43.116)
3 2
and
1 1
T 2 = − (−a1 x + a3 )u3 − (a1 t + a2 )(ut )2 (43.117)
3 2
respectively. Since
1 1 1 1
T 1 = a1 − tu3 + xu2x + a2 − u3 + a3 − u2x
3 2 3 2 (43.118)
1 1 1 1
T 2 = a1 xu3 − tu2t + a2 − u2t + a3 − u3
3 2 2 3
we obtain from the above expressions the following conserved vectors
1 1 1 1
T 1 = − tu3 + xu2x , T 2 = xu3 − tu2t (43.119)
3 2 3 2
1 1
T 1 = − u3 , T 2 = − u2t (43.120)
3 2
and
1 1
T 1 = − u2x , T 2 = − u3 (43.121)
2 3
We have thus obtained the same set of conserved vectors as in the Direct
Method.
Example 4. The Variational Method. We shall determine the conserved
vectors corresponding to the equation uxt + u2 = 0 using the Variational
Method. The conserved vectors T 1 , T 2 are determined by the equation
Q(uxt + u2 ) = Di T i (43.122)
where Q is a quantity determined by the variational derivative

δ
[Q(uxt + u2 )] = 0 (43.123)
δu
505
δ
where is the variational derivative (the Euler operator with respect to u)
δu
δ ∂ ∂ ∂ ∂
= − Dx − Dt + Dx Dt + ··· (43.124)
δu ∂u ∂ux ∂ut ∂uxt
The quantity Q has the following functional dependence
Q = Q(x, t, u, ux , ut ) (43.125)
The variational equation (43.123) is equivalent to

∂ ∂ ∂ ∂
− Dx − Dt + Dx Dt + · · · [Q(uxt + u2 )] = 0
∂u ∂ux ∂ut ∂uxt
∂Q ∂
⇐⇒ (uxt + u2 ) + 2u·Q − Dx [Q(uxt + u2 )]
∂u ∂ux
∂ ∂
− Dt [Q(uxt + u2 )] + Dx Dt [Q(uxt + u2 )] = 0 (43.126)
∂ut ∂uxt

∂Q ∂Q
⇐⇒ (uxt + u2 ) + 2u·Q − Dx (uxt + u2 )
∂u ∂ux

∂Q 2
− Dt (uxt + u ) + Dx Dt (Q) = 0
∂ut
where
∂ ∂ ∂ ∂ ∂
Dx = + ux + uxt + uxx + uxxt + ··· (43.127)
∂x ∂u ∂ut ∂ux ∂uxt
and
∂ ∂ ∂ ∂ ∂
Dt = + ut + utt + utx + utxt + ··· (43.128)
∂t ∂u ∂ut ∂ux ∂uxt
We have further

∂Q ∂Q ∂Q
Dx (uxt + u ) = (uxt + u2 )Dx
2
+ Dx (uxt + u2 )
∂ux ∂ux ∂ux
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2 (43.129)
= (uxt + u ) + ux + uxt + uxx 2
∂ux ∂x ∂ux ∂u ∂ux ∂ut ∂ux
∂Q
+ (2u·ux + uxxt )
∂ux
506
and similarly

∂Q ∂Q ∂Q
Dt (uxt + u ) = (uxt + u2 )Dt
2
+ Dt (uxt + u2 )
∂ut ∂ut ∂ut
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
= (uxt + u2 ) + ut + utx + utt 2 (43.130)
∂ut ∂t ∂ut ∂u ∂ut ∂ux ∂ut
∂Q
+ (2u·ut + utxt )
∂ut
On the other hand, since
∂Q ∂Q ∂Q ∂Q
Dt (Q) = + ut + utt + utx (43.131)
we get
∂ ∂Q ∂Q ∂Q ∂Q
Dx Dt (Q) = + ut + utt + utx
∂x ∂t ∂u ∂ut ∂ux
∂ ∂Q
∂Q ∂Q ∂Q
+ ux + ut + utt + utx
∂u ∂t ∂u ∂ut ∂ux
∂ ∂Q
∂Q ∂Q ∂Q
+ uxx + ut + utt + utx
∂ux ∂t ∂u ∂ut ∂ux
∂ ∂Q ∂Q ∂Q ∂Q
+ uxt + ut + utt + utx
∂ut ∂t ∂u ∂ut ∂ux
2
∂ Q 2
∂ Q 2
∂ Q 2
∂ Q (43.132)
= + ut + utt + utx
∂t∂x ∂u∂x ∂ut ∂x ∂ux ∂x
∂ 2Q 2
∂ Q 2
∂ Q ∂ 2Q
∂t∂u ∂u∂u ∂ut ∂u ∂ux ∂u
∂ 2Q 2
∂ Q 2
∂ Q ∂ 2Q
+ uxx + ut + utt + utx 2
∂t∂ux ∂u∂ux ∂ut ∂ux ∂ux
∂ 2Q 2
∂ Q ∂ Q2 2
∂ Q
+ uxt + ut + utt 2 + utx
∂t∂ut ∂u∂ut ∂ut ∂ux ∂ut
507
The variational equation (43.126), because of (43.129), (43.130) and (43.132)
takes on the form
∂Q
(uxt + u2 ) + 2u·Q
∂u
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
− (uxt + u2 ) + ux + uxt + uxx 2
∂ux ∂x ∂u ∂u ∂ux ∂ut ∂ux
x
∂Q
+ (2u·ux + uxxt )
∂ux
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
2
− (uxt + u ) + ut + utx + utt 2
∂ut ∂t ∂ut ∂u ∂ut ∂ux ∂ut

∂Q
+ (2u·ut + utxt ) (43.133)
∂ut
∂ 2Q ∂ 2Q ∂ 2Q ∂ 2Q
+ + ut + utt + utx
∂t∂x ∂u∂x ∂ut ∂x ∂ux ∂x
∂ 2Q 2
∂ Q 2
∂ Q ∂ 2Q
∂t∂u ∂u∂u ∂ut ∂u ∂ux ∂u
∂ 2Q 2
∂ Q 2
∂ Q ∂ 2Q
+ uxx + ut + utt + utx 2
∂t∂ux ∂u∂ux ∂ut ∂ux ∂ux
∂ 2Q 2
∂ Q 2
∂ Q 2
∂ Q
+ uxt + ut + utt 2 + utx =0
∂t∂ut ∂u∂ut ∂ut ∂ux ∂ut
Equating to zero all the coefficients of the partial derivatives of the function
u and solving the determining equations, we get the following expression for
the quantity Q:
Q = (a1 t + a2 )ut + (−a1 x + a3 )ux (43.134)
The connection between the multiplier Q and the components (T 1 , T 2 ) of the

conserved vector is provided by equation (43.122), i.e.
Q(uxt + u2 ) = Dt T 1 + Dx T 2 (43.135)
The above expression, taking into account the explicit expression for the
multiplier Q and considering that T 1 , T 2 are of the first order, i.e. T 1 =
T 1 (x, t, u, ux , ut ) and T 2 = T 2 (x, t, u, ux , ut ), equation (43.135) can be written
508
in expanded form as
{(a1 t + a2 )ut + (−a1 x + a3 )ux }(uxt + u2 )

∂T 1 ∂T 1 ∂T 1 ∂T 1
= + ut + utt + utx (43.136)
∂T 2 ∂T 2 ∂T 2 ∂T 2
+ + ux + uxx + uxt
∂x ∂u ∂ux ∂ut
We now have to equate to zero all the coefficients of the partial derivatives
of the unknown function u. We find that
∂T 1
u2 (a1 t + a2 ) = (43.137)
∂u
∂T 2
u2 (−a1 x + a3 ) = (43.138)
∂u
The coefficient of utt is
∂T 1
=0 (43.139)
∂ut
∂T 2
=0 (43.140)
∂ux
The coefficient of uxt is
∂T 1 ∂T 2
(a1 t + a2 )ut + (−a1 x + a3 )ux = + (43.141)
∂ux ∂ut
∂T 1 ∂T 2
+ =0 (43.142)
∂t ∂x
From (43.139) and (43,140) we obtain that
T 1 = T 1 (x, t, u, ux ), T 2 = T 2 (x, t, u, ut ) (43.143)
509
Integrating (43.137) and (43.138) with respect to u and taking into account
(43.143), we obtain that
1
T 1 = (a1 t + a2 )u3 + F (x, t, ux ) (43.144)
3
and
1
T 2 = (−a1 x + a3 )u3 + G(x, t, ut ) (43.145)
3
The constant term equation (43.142) becomes, because of (43.144) and (43.145)
∂F ∂G
+ =0 (43.146)
∂t ∂x
From equation (43.141) taking again into account the expressions (43.144)
and (43.145), we get the equation
∂F ∂G
(a1 t + a2 )ut + (−a1 x + a3 )ux = + (43.147)
∂ux ∂ut
∂F ∂G
= (−a1 x + a3 )ux , = (a1 t + a2 )ut (43.148)
∂ux ∂ut
and upon integration
1 1
F = (−a1 x + a3 )u2x + f (x, t) G = (a1 t + a2 )u2t + g(x, t) (43.149)
2 2
∂f ∂g
+ =0 (43.150)
∂t ∂x
We may take f = g = 0 since these terms contribute trivially to the conserved
vectors. We thus have the following expressions for the conserved vectors
1 1
T 1 = (a1 t + a2 )u3 + (−a1 x + a3 )u2x
3 2 (43.151)
1 1
T 2 = (−a1 x + a3 )u3 + (a1 t + a2 )u2t
3 2
510
44 Adjoint Equations
We suppose that L is a linear differential operator of any order. The adjoint
operator of L is being denoted by L∗ and it is defined through the relation
v L[u] − u L∗ [v] = Di (P i ) ∀u, v∈A (44.1)
and P is a vector field
P = (P 1 , P 2 , · · ·, P n ) (44.2)
Equation
L∗ [v] = 0 (44.3)
is called the adjoint equation to L[u] = 0.
Consider a general second order operator L
L = aij (x)Di Dj + bi (x)Di + c(x), aij = aji , i, j = 1, 2, · · ·, n (44.4)
Theorem 1. The adjoint operator L∗ is uniquely determined and has the
form
L∗ [v] = Di Dj (aij v) − Di (bi v) + c v (44.5)
Proof. We form the quantity v L[u] and we use obvious identities
v L[u] = v aij (x)Di Dj u + v bi (x)Di u + v c(x) u
= Di (v aij (x)Dj u) − Di (v aij (x))Dj u (44.6)
i i
+ Di (v b (x)u) − uDi (v b (x)) + v c(x)u
We have further, using again obvious identities
− Di (v aij (x))Dj u = −Dj (uDi (vaij (x))) + uDi Dj (aij (x)v)
(44.7)
= −Di (uDj (vaij (x))) + uDi Dj (aij (x)v)
In the last step we interchanged the indices i, j and used the fact that aij is
symmetric. Hence
v L[u] = Di (v aij (x)Dj u) − Di (uDj (vaij (x)))
+ uDi Dj (aij (x)v) + Di (v bi (x)u) − uDi (v bi (x)) + v c(x)u

ij i
= u Di Dj (a (x)v) − Di (v b (x)) + v c(x) (44.8)

ij ij i
+ Di v a (x)Dj u − u Dj (v a (x)) + v b (x) u
511
From the above identity we obtain the following new identity

ij i
v L[u] − u Di Dj (a (x)v) − Di (v b (x)) + v c(x)
(44.9)
ij ij i
= Di v a (x)Dj u − u Dj (v a (x)) + v b (x) u
Comparing the above identity to (44.1), we get

L∗ = Di Dj (aij (x) v) − Di (v bi (x)) + c(x) v (44.10)
and
P i = v aij (x)Dj u − u Dj (v aij (x)) + v bi (x) u (44.11)
Note. The same result applies if u, v are m−dimensional vector functions,
where the coefficients are m×m matrices.
Definition. An operator is said to be self-adjoint if for any function u,
L[u] = L∗ [u] (44.12)
Theorem 2. The operator
L = aij (x)Di Dj + bi (x)Di + c(x) (44.13)
is self-adjoint if and only if
bi (x) = Dj (aij (x)) (44.14)
Proof. From equation
v L[u] − u L∗ [v] = Di (P i ) (44.15)
for u = v we shall have L[u] = L∗ [u] if and only if Di (P i ) = 0. For u = v we
have

i ij ij i
Di (P ) = Di u a (x)Dj u − u Dj (u a (x)) + u b (x) u

ij ij i
= Di u a (x)Dj u − Dj (u a (x)) + b (x) u
(44.16)
ij ij ij i
= Di u a (x)Dj u − a (x)Dj u − uDj (a (x)) + b (x) u

i ij
= Di u u b (x) − Dj (a (x))
512
The above has to be true for any u, therefore we should have
bi (x) − Dj (aij (x)) = 0 (44.17)
and this proves the Theorem.

Definition. We consider a system of s−order PDEs
Fα (x, u, u(1) , · · ·, u(s) ) = 0, α = 1, 2, · · ·, m (44.18)
The system of adjoint equations to (44.18) is defined by
δ(v β Fβ )
Fα∗ (x, u, v, · · ·, u(s) , v(s) )≡ = 0, α = 1, 2, · · ·, m (44.19)
δuα
where v = (v 1 , v 2 , · · ·, v m ) are new dependent variables, v = v(x).
Definition. A system (44.18) is said to be self-adjoint if the system ob-
tained from the adjoint equations (44.19) by the substitution v = u,
Fα∗ (x, u, u, · · ·, u(s) , u(s) ) = 0, α = 1, 2, · · ·, m (44.20)
is identical to the original system (44.18).

Example 1. Consider the heat equation
ut − c(x)uxx = 0 (44.21)
In this case
F = F (x, t, u, u(1) , u(2) ) = c(x)uxx − ut = 0 (44.22)
where x, t are the independent variables and u is the differential variable. By

definition we obtain the adjoint equation as
∗ ∗ δ
F = F (x, t, u, v, u(1) , u(2) , v(1) , v(2) ) = v{c(x)uxx − ut }
δu
∂ ∂ ∂ ∂ ∂
= − Dt − Dx + Dx2 + Dt2
∂u ∂ut ∂ux ∂uxx ∂utt
∂
(44.23)
+ Dx Dt v{c(x)uxx − ut }
∂uxt
∂ ∂
= −Dt + Dx2 v{c(x)uxx − ut }
∂ut ∂uxx
2
= −Dt (−v) + Dx (c(x) v) = vt + (c v)xx
513
Example 2. We consider next the KdV equation
ut = uux + uxxx (44.24)
F = F (t, x, u, u(1) , u(2) , u(3) ) = ut − uux − uxxx (44.25)
The adjoint equation is
F ∗ (t, x, u, v, u(1) , u(2) , u(3) , v(1) , v(2) , v(3) )

δ
= v{ut − uxxx − uux }
δu
∂ ∂ ∂ ∂ (44.26)

= − Dt − Dx − Dx3 v{ut − uxxx − uux }
∂u ∂ut ∂ux ∂uxxx
= −vux − Dt v + Dx (vu) + Dx3 v = −vux − vt + vx u + vux + vxxx
= −vt + vx u + vxxx
The adjoint equation to ut = uxxx + uux is vt = vx u + vxxx . If we replace v

by u in the adjoint equation, we obtain
F ∗ (t, x, u, v, u(1) , u(2) , u(3) , v(1) , v(2) , v(3) )

= −(ut − uxxx − uux ) (44.27)
= −F (t, x, u, u(1) , u(2) , u(3) )
We thus see that, according to the definition, the KdV equation is self-adjoint.
44.2. Constructing Lagrangians using the Adjoint Equation.
Theorem 3. Any system of s−order differential equations
Fα (x, u, u(1) , · · · , u(s) ) = 0, α = 1, 2, · · ·, m (44.28)
considered together with its adjoint

δ
Fα∗ (x, u, v, · · · , u(s) , v(s) ) = (v β Fβ ) = 0 (44.29)
δuα
has a Lagrangian. The simultaneous system (44.28) and (44.29) with 2m
dependent variables u = (u1 , · · · , um ) and v = (v 1 , · · · , v m ) is the system of
Euler-Lagrange equations with a Lagrangian L defined by
L = v β Fβ (44.30)
514
Proof. This is easy to verify, since
δL
= Fα (x, u, u(1) , · · · , u(s) ) = 0 (44.31)
δv α
and
δL
α
= Fα∗ (x, u, v, · · · , u(s) , v(s) ) = 0 (44.32)
δu
i.e. the Euler-Lagrange equation is satisfied.
We next state the following Lemma will be used for the construction of
Lagrangians:
Lemma. A function f (x, u, · · · , u(s) )∈A with several independent variables
x = (x1 , x2 , · · · , xn ) and several dependent variables u = (u1 , u2 , · · · , um ) is
a divergence of the vector field H = (h1 , h2 , · · · , hn ), hi ∈A, i.e.
f = divH≡Di (hi ) (44.33)
if and only if the following equations hold

δf
= 0, α = 1, 2, · · · , m (44.34)
δuα
Considering next linear second order partial differential equations of the form
L[u] = aij (x)uij + bi ui + c(x)u = 0 (44.35)
the Lagrangian (44.30) has the form
L = vL[u] = v[aij (x)uij + bi ui + c(x)u] (44.36)
We then have
δL ∂L
= = L[v] (44.37)
δv ∂v
and
δL ∂L ∂L ∂L
= Di Dj − D i +
δuα ∂uij ∂ui ∂u (44.38)
= Di Dj (a v) − Di (b v) + cv = L∗ [v]
ij i
515
Theorem 4. Let the linear operator L[u] be self-adjoint, i.e. L[u] = L∗ [u].
Then equation (44.35) is obtained from the Lagrangian
1
L = [c(x)u2 − aij (x)ui uj ] (44.39)
2
Proof. The Lagrangian (44.36) can be written in the form
L = v[aij (x)uij + bi ui + c(x)u]

(44.40)
= Dj (vaij ui ) − vui Dj (aij ) + vbi ui − aij ui vj + cuv
The first term of the second line is zero because of the above Lemma,
while the second and third terms annihilate each other because of the self-
adjointness condition bi (x) = Dj (aij ). Finally we set v = u, divide by two
and arrive at the Lagrangian (44.39).
Example 3. The heat equation considered along to its adjoint equation, has
a Lagrangian
L = v [ut − c(x)ux ] (44.41)
Proof. We calculate explicitly the derivatives (Euler-Lagrange equations)
δL
= ut − c(x)ux = 0 (44.42)
δv
δL δ
= {v [ut − c(x)ux ]}
δu δu
= F ∗ (t, x, u, v, u(1) , u(2) , v(1) , v(2) ) (44.43)
= vt + (cv)xx = 0
i.e. the Lagrangian (44.33) satisfies the Euler-Lagrange equation.
Example 4. The motion of a free particle in space is described by the
equation
∂ 2 xα
= 0, α = 1, 2, 3 (44.44)
∂t2
where
xα = xα (t) (44.45)
516
Here t is the independent variable. We shall determine the generators of the
symmetry of the free motion. Let X be the generator of the symmetries:
∂ ∂
X = ξ(t, x1 , x2 , x3 ) + η α (t, x1 , x2 , x3 ) α (44.46)
∂t ∂x
where summation over α is understood. The second prolongation is given by
∂ ∂ [α] ∂ [α] ∂
P r(2) X = ξ + η α α + ζ1 α
+ ζ2 (44.47)
∂t ∂x ∂ ẋ ∂ ẍα
where
[α]
ζ1 = Dt (η α ) − ẋα Dt (ξ) (44.48)
and
[α] [α]
ζ2 = Dt (ζ1 ) − ẍα Dt (ξ) (44.49)
Invariance of the equation is being expressed by the condition
∂ 2 xα
P r(2) X[∆] = 0 as long as ∆ = 0, ∆= (44.50)
∂t2
[α] [α]
Let us now calculate explicitly the coefficients ζ1 and ζ2 using the above
formulas. We have
[α]
ζ1 = Dt (η α ) − ẋα Dt (ξ)
∂η α ∂η α ∂ξ ∂ξ (44.51)
= + ẋβ β − ẋα + ẋγ γ
∂t ∂x ∂t ∂x
517
and
[α] [α]
ζ2 = Dt (ζ1 ) − ẍα Dt (ξ)
∂η α ∂η α ∂ξ ∂ξ
= Dt + Dt ẋβ β − Dt ẋα − Dt ẋα ẋγ γ
∂t ∂x ∂t ∂x
∂ξ ∂ξ
− ẍα + ẋγ γ
∂t ∂x
∂ ∂η α ∂η α ∂η α
β ∂ β
= + ẋ β
+ ẋ Dt β
+ β Dt ẋβ
∂t ∂x ∂t ∂x ∂x
∂ξ ∂ξ ∂ξ ∂ξ
α α α γ α γ
− ẋ Dt − Dt ẋ − ẋ ẋ Dt − γ Dt ẋ ẋ
∂t ∂t ∂xγ ∂x
∂ξ ∂ξ (44.52)
− ẍα + ẋγ γ
∂t ∂x
∂ 2ηα β ∂ 2 α
η β ∂

γ ∂
∂η α
= + ẋ + ẋ + ẋ
∂t2 ∂t∂xβ ∂t ∂xγ ∂xβ
α
∂η ∂ γ ∂

β α ∂

γ ∂
∂ξ
+ β + ẋ ẋ − ẋ + ẋ
∂x ∂t ∂xγ ∂t ∂xγ ∂t
∂ξ ∂

γ ∂

α α γ ∂

β ∂
∂ξ
− + ẋ ẋ − ẋ ẋ + ẋ
∂t ∂t ∂xγ ∂t ∂xβ ∂xγ
∂ξ ∂ ∂ ∂ξ ∂ξ
− γ + ẋβ β (ẋα ẋγ ) − ẍα + ẋβ β
∂x ∂t ∂x ∂t ∂x
After performing the necessary differentiations, we obtain the following ex-
[α]
pression for ζ2 :
[α] ∂ 2ηα 2 α
β ∂ η β
∂ 2ηα 2 α
γ ∂ η
ζ2 = + ẋ + ẋ + ẋ
∂t2 ∂t∂xβ ∂t∂xβ ∂xβ xγ
∂η α ∂ 2 xβ
2 β
γ ∂ x

α ∂ ξ
2
γ ∂ ξ
2
+ β + ẋ − ẋ + ẋ
∂x ∂t2 ∂xγ ∂t ∂t2 ∂xγ ∂t
2 α 2 α 2
∂ξ ∂ x

γ ∂ x α γ
∂ ξ β ∂ 2ξ (44.53)
− + ẋ − ẋ ẋ + ẋ
∂t ∂t2 ∂xγ ∂t ∂t∂xγ ∂xβ ∂xγ
∂ξ ∂ 2 xα γ ∂ 2 xγ α β ∂ξ
∂ 2 xα
γ ∂ 2 xγ α
− γ ẋ + ẋ − ẋ ẋ + β ẋ
∂x ∂t2 ∂t2 ∂xγ ∂xβ ∂t ∂x ∂t
∂ξ
β ∂ξ

α
− ẍ + ẋ
∂t ∂xβ
518
[α]
We see next that the invariance condition (44.42) is equivalent to ζ2 = 0,
∆=0
i.e.
[α]
P r(2) X[∆] = 0 ⇐⇒ ζ2 =0 (44.54)
∆=0 ∆=0
Taking into account the above condition and the expression (44.45), we ob-
tain the following invariance condition
∂ 2ηα 2 α
β ∂ η
2 α
β γ ∂ η
2
α∂ ξ
2
α β ∂ ξ
+ 2 ẋ + ẋ ẋ − ẋ − 2ẋ ẋ
∂t2 ∂t∂xβ ∂xβ xγ ∂t2 ∂t∂xβ (44.55)
2
∂ ξ
− ẋα ẋβ ẋγ β γ = 0
∂x x
Equating to zero the coefficients of ẋα , we obtain from the above equation
the following system of determining equations
∂ 2ηα
=0
∂t2
∂ ∂η α ∂ξ
2 β − δβα =0
∂t ∂x ∂t (44.56)
∂ ∂η α α ∂ξ

− 2δ β =0
∂xγ ∂xβ ∂t
∂ 2ξ
=0
∂xβ xγ
Example 5. We consider the equation
ux uxx + uyy = 0 (44.57)
This is a partial differential equation known from transonic gas dynamics.

Let
∂ ∂ ∂
X = ξ 1 (x, y, u) + ξ 2 (x, y, u) + η(x, y, u) (44.58)
∂x ∂y ∂u
The second extension (2nd-prolongation) has the form
∂ ∂ ∂ ∂ ∂
P r(2) X = ξ 1 + ξ2 +η + ζ1 + ζ2
∂x ∂y ∂u ∂ux ∂uy
(44.59)
∂ ∂ ∂
+ ζ11 + ζ12 + ζ22
519
where
ζ1 = Dx (η) − ux Dx (ξ 1 ) − uy Dx (ξ 2 ) (44.60)
We thus have the following generators

∂ ∂ ∂ ∂
X1 = , X2 = , X3 = , X4 = y ,
∂x ∂y ∂u ∂u
(44.61)
∂ ∂ ∂ ∂
X5 = x + 3u , X 6 = y − 2u
∂x ∂u ∂y ∂u
Considering the generator
∂ ∂ ∂
X5 + X6 = x +y +u (44.62)
∂x ∂y ∂u
the invariant surface condition becomes in this case
xux + yuy = u (44.63)

y
u = x·F (44.64)
x
y
ξ= (44.65)
x
the original equation takes on the form
ξ 3 F 0 (ξ) − ξ 2 F (ξ) = 1 (44.66)

1
F (ξ) = ξ − 3 + C (44.67)
3ξ
Going back to the original variables, the general solution of the equation is
given by
1 x3
u=y − 3 +C (44.68)
3y
520
Considering next the generator
∂ ∂ ∂
X 1 + X 2 + X 5 + X 6 = (x + 1) + (y + 1) +u (44.69)
∂x ∂y ∂u
the invariant surface condition becomes in this case
(x + 1)ux + (y + 1)uy = u (44.70)

y + 1
u = (x + 1)·F (44.71)
x+1
y+1
ξ= (44.72)
x+1
the original equation takes on the form
ξ 3 F 0 (ξ) − ξ 2 F (ξ) = 1 (44.73)

1
F (ξ) = ξ − 3 + C (44.74)
3ξ
Going back to the original variables, the general solution of the equation is
given by
1 (x + 1)3
u = (y + 1) − + C (44.75)
3 (y + 1)3
45 The Main Conservation Theorem

Consider a system of m differential equations of order s
Fα (x, u, u(1) , · · · , u(s) ) = 0, α = 1, 2, · · · , m (45.1)
with n independent variables x = (x1 , x2 , · · · , xn ) and m dependent variables

u = (u1 , u2 , · · · , um ). The adjoint system is
δ
Fα∗ (x, u, v, · · · , u(s) , v(s) ) = (v β Fβ ) = 0, α = 1, 2, · · · , m (45.2)
δuα
521
where v = (v 1 , v 2 , · · · , v m ) are new dependent variables, v = v(x).
Consider a point symmetry of the original system (45.1)
∂ ∂
X = ξi i
+ ηα α (45.3)
∂x ∂u
The condition for (45.1) to be invariant is
XFα = λβα Fβ (45.4)
The Lagrangian for the simultaneous system of equations (45.1) and (45.2)
is given by
L = v α Fα (45.5)
The extension of (45.3) to the adjoint variables is the generator
∂ ∂ ∂
Y = ξi i
+ η α α + η∗α α (45.6)
∂x ∂u ∂v
Let us now try to determine η∗α . First of all, the invariance condition
Y (L) + LDi (ξ i ) = 0 (45.7)
should be satisfied. Expanding the above condition (45.7), we find
Y (L) + LDi (ξ i ) = Y (v α Fα ) + v α Fα Di (ξ i )
= Y (v α )Fα + v α Y (Fα ) + v α Fα Di (ξ i )
= Y (v α )Fα + v α X(Fα ) + v α Fα Di (ξ i )
(45.8)
= η∗α Fα + v α λβα Fβ + v α Fα Di (ξ i )

α β α α i
= η∗ + v λβ + v Di (ξ ) Fα
Therefore invariance requires
η∗α + v β λαβ + v α Di (ξ i ) = 0 (45.9)
i.e.

η∗α β α α i
= − v λβ + v Di (ξ ) (45.10)
522
Thus if the original system admits the operator (45.3), the adjoint system
admits the operator
∂ ∂ ∂
Y = ξ i i + η α α − v β λαβ + v α Di (ξ i ) (45.11)
∂x ∂u ∂v α
We thus arrive at the following
Theorem 1. For a system described by (45.1) that admits a generator (45.3),
its adjoint system (45.2) inherits the symmetries of the original system, i.e.
the adjoint system admits the generator (45.6) extended according to (45.11).
Theorem 2. The Main Conservation Theorem. Every Lie point, Lie-
Bäcklund and non-local symmetry
∂ ∂
X = ξ i (x, u, u(1) , · · · ) i
+ η α (x, u, u(1) , · · · ) α (45.12)
∂x ∂u
of differential equations
Fα (x, u, u(1) , · · · , u(s) ) = 0, α = 1, 2, · · · , m (45.13)
provides a conservation law for the system of differential equations (45.13)
considered together with its adjoint system of equations.
Example 1. Angular Momentum Conservation. Consider the equation
of motion of a free particle of mass m in space
m ẍ = 0 (45.14)
where x = (x1 , x2 , x3 ) is the vector of spatial coordinates. The system of
differential equations we consider is given in our case by
Fα (t, xα , ẋα , ẍα ) = 0, α = 1, 2, 3 (45.15)
In the above equation t is the independent variable and xα are the dependent
variables. On the other hand, equations (45.15) are invariant under rotational
symmetry with generators
∂ ∂
Xij = xj i
− xi j (45.16)
∂x ∂x
We thus have the following three generators of rotational symmetry
∂ ∂
X12 = x2 1
− x1 2
∂x ∂x
∂ ∂
X13 = x3 1 − x1 3 (45.17)
∂x ∂x
∂ ∂
X23 = x3 2 − x2 3
∂x ∂x
523
We consider the generator X12 , i.e. rotation around x3 −axis. The corre-
sponding tangent vector field has components ξ = 0, η 1 = x2 , η 2 = −x1 and
η 3 = 0. The second prolongation corresponding to the generator X12 is given
by
∂ ∂ [α] ∂
P r(2) X12 = x2 1
− x1 2 + ζ2 (45.18)
∂x ∂x ∂ ẍα
where
[α]
ζ1 = Dt (η α ) − ẋα Dt (ξ)
[α] [α]
(45.19)
ζ2 = Dt (ζ1 ) − ẍα Dt (ξ)
[α] [α]
In calculating ζ1 and ζ2 we have to take into account that ξ = 0. We
obtain
[1]
ζ1 = Dt (η 1 ) = Dt (x2 ) = ẋ2
[2]
ζ1 = Dt (η 2 ) = −Dt (x1 ) = −ẋ1 (45.20)
[3]
ζ1 = 0
and
[1] [1]
ζ2 = Dt (ζ1 ) = Dt (ẋ2 ) = ẍ2
[2] [2]
ζ2 = Dt (ζ1 ) = −Dt (ẋ1 ) = −ẍ1 (45.21)
[3]
ζ2 = 0
We then have the following double extended operator

∂ ∂ ∂ ∂
P r(2) X12 = x2 1
− x1 2 + ẍ2 1 − ẍ1 2 (45.22)
∂x ∂x ∂ ẍ ∂ ẍ
The simultaneous system (45.14) considere together has the Lagrangian, ac-
cording to (45.5)
L = y β Fβ = my 1 ẍ1 + my 2 ẍ2 + my 3 ẍ3 (45.23)
The extension of X12 to the new adjoint variables y α is

∂ 1 ∂ α ∂
Y = x2 − x + η∗ (45.24)
∂x1 ∂x2 ∂y α
524
where η∗α is given by (45.10). Since ξ = 0, we get from (45.10) that
η∗α = −λαβ y β (45.25)
The quantities λαβ are obtained from the invariance condition (45.4). We
have in our case
P r(2) [X12 ]Fα = λβα Fβ (45.26)
The above equation can be written in expanded form (for α = 1, 2, 3 respec-

tively)
ẍ2 = λ11 ẍ1 + λ21 ẍ2 + λ31 ẍ3

− ẍ1 = λ12 ẍ1 + λ22 ẍ2 + λ32 ẍ3 (45.27)
0 = λ13 ẍ1 + λ23 ẍ2 + λ33 ẍ3
For these equations to be true, we should have λ21 = 1 and λ12 = −1 and all
other must be zero. Using (45.25), we get
η∗1 = −λ11 y 1 − λ12 y 2 − λ13 y 3 = y 2

η∗2 = −λ21 y 1 − λ22 y 2 − λ23 y 3 = −y 1 (45.28)
η∗3 = 0
In order to determine the conserved quantity associated to the extended

generator (45.24), we apply Noether’s theorem. The conserved quantity C is
given by
∂L ∂L ∂L
C = ξL + W α − Dt + Dt (η α ) α (45.29)
∂uαt α
∂utt ∂utt
where
uα = (x1 , x2 , x3 , y 1 , y 2 , y 3 )
uαt = (ẋ1 , ẋ2 , ẋ3 , ẏ 1 , ẏ 2 , ẏ 3 ) (45.30)
uαtt = (ẍ1 , ẍ2 , ẍ3 , ÿ 1 , ÿ 2 , ÿ 3 )
Since ξ = 0 and uα = (x2 , −x1 , 0, y 2 , −y 1 , 0) we get
∂L
W α = ηα and = 0 for every α = 1, 2, 3 (45.31)
∂uαt
525
Therefore the conserved quantity (45.29) takes on the form
C = −mx2 ẏ 1 + mx1 ẏ 2 + my 1 ẋ2 − my 2 ẋ1 (45.32)
Using the fact that system (45.14) is self-adjoint, we can replace the adjoint
variable y by x and substitute ẋ by v, (ẋ = v) where v is the velocity. We
then have from (45.32)
C = −mx2 v 1 + mx1 v 2 + mx1 v 2 − mx2 v 1
or
C = 2m(x1 v 2 − x2 v 1 ) (45.33)
We recognize that the above expression, after setting M = C/2, i.e.
M12 = m(x1 v 2 − x2 v 1 ) (45.34)
is the angular momentum.

Performing the same calculations for X13 and X23 , we arrive at the angular
momentum conservation
M = m(x×v) (45.35)
Example 2. The Maxwell Equations. We consider Maxwell’s equation

in vacuum
1 ∂E
= curlH, divE = 0
c ∂t (45.36)
1 ∂H
= −curlE, divH = 0
c ∂t
The above system contains six dependent variables, i.e. the components of
the electric field E = (E 1 , E 2 , E 3 ) and the magnetic field H = (H 1 , H 2 , H 3 )
and eight equations. Therefore the system is overdetermined. However the
number of equations in the Euler-Lagrange equations is equal to the number
of dependent variables. Consequently the system (45.36) cannot have a La-
grangian.
A variational problem in electrodynamics provides a Lagrangian for the wave
equation
1 ∂ 2A
4A − =0 (45.37)
c2 ∂t2
526
for the vector potential A of the electromagnetic field. This is not the case for
Maxwell’s equations. Let us now find a Lagrangian for the electromagnetic
field using Theorem 3 (§ 44.2). We note first that equations divE = 0 and
divH = 0 hold at any time provided that they are satisfied at initial time
t = 0. Hence they are merely initial conditions. We thus consider the
following determined system of differential equations (we set t0 = ct and
take t0 as new t):
∂H ∂E
curlE + = 0, curlH − =0 (45.38)
∂t ∂t
We introduce six new dependent variables, i.e. the components of the vectors
V = (V 1 , V 2 , V 3 ) and W = (W 1 , W 2 , W 3 ) (45.39)
and introduce the Lagrangian
∂H ∂E
L = V · curlE + + W · curlH − (45.40)
∂t ∂t
taking into account definition (44.30).
We can easily verify that Lagrangian (45.40) yields the system (45.38),
together with its adjoint, i.e.
δL ∂H δL ∂E
≡ curlE + = 0, ≡ curlH − =0 (45.41)
δV ∂t δW ∂t
and
δL ∂W δL ∂V
≡ curlV + = 0, ≡ curlW − =0 (45.42)
δE ∂t δH ∂t
If we set V = E and W = H, equations (45.42) coincode with (45.41).
Hence equations (45.38) are self-adjoint. Therefore we set V = E, W = H
in (45.40) divide by two and obtain the Lagrangian for the Maxwell equations
(45.38):

1 ∂H ∂E
L = E· curlE + + H· curlH − (45.43)
2 ∂t ∂t
The above Lagrangian can be written in expanded form (i.e. using Cartesian
coordinates) as
L = E 1 (Ey3 − Ez2 + Ht1 ) + E 2 (Ez1 − Ex3 + Ht2 ) + E 3 (Ex2 − Ey1 + Ht3 )
(45.44)
+ H 1 (Hy3 − Hz2 − Et1 ) + H 2 (Hz1 − Hx3 − Et2 ) + H 3 (Hx2 − Hy1 − Et3 )
527
Let us now consider the invariance of equations (45.38) with respect to the
group of transformations (3-dimensional rotation)
H 0 = H cosθ + E sinθ, E 0 = E cosθ − H sinθ (45.45)

with generator
3
∂ ∂ X ∂ i ∂

X=E −H ≡ Ei − H (45.46)
∂H ∂E i=1
∂H i ∂E i
and apply Noether’s Conservation Theorem. The first prolongation of this
generator is given by
∂ ∂ ∂ ∂ ∂ ∂
P r(1) X = E −H +Et −Ht +Ex −Hx +· · · (45.47)
∂H ∂E ∂Ht ∂Et ∂Hx ∂Ex
We then find with L given by (45.43), that

(1) 1
P r X(L) = − H·(curlE + Ht ) + E·(curlH − Et )
2
(45.48)
+ E·(−curlH + Et ) + H·(curlE + Ht ) = 0
Taking into account this condition and using formula (43.28), we can write
down the conservation law in the form
Dt (τ ) + divχ = 0 (45.49)
where
χ = (χ1 , χ2 , χ3 ), divχ = Dx (χ1 ) + Dy (χ2 ) + Dz (χ3 ) (45.50)
with τ given by (according to formula (43.28))
∂L ∂L 1 1
τ = E· − H· = [E·E − H·(−H)] = [E 2 + H 2 ] (45.51)
∂Ht ∂Et 2 2
Hence τ is the energy density. On the other hand, we can verify that χ is the
Poynting vector, χ = E×H by calculating the components of the conserved
vector (43.28). We thus have obtained the conservation of energy
E2 + H 2
Dt + div(E×H) = 0 (45.52)
2
The above formula appears in L. D. Landau and E. M. Lifshitz: ”The
Classical Theory of Fields”. Pergamon 1972.
528
46 The Double Reduction I
We define a non-local variable v by
T t = vx , T x = −vt (46.1)
We find similarity variables r, s, w such that
∂
X= (46.2)
∂s
In the similarity variables
T r = vs , T s = −vr (46.3)
so that the conservation law is rewritten as
Dr T r + Ds T s = 0 (46.4)
In order to write T r and T s in terms of the original variables, we note that
Dx = Dx (s)Ds + Dx (r)Dr , Dt = Dt (s)Ds + Dt (r)Dr (46.5)
and therefore
vx = Dx (s)vs + Dx (r)vr , vt = Dt (s)vs + Dt (r)vr (46.6)
The above equations, in view of (46.3), can be written as
T t = T r Dx (s) − T s Dx (r), −T x = T r Dt (s) − T s Dt (r) (46.7)
Solving the above system with respect to T s and T r we find
s T t Dt (s) + T x Dx (s)
T = (46.8)
Dt (r)Dx (s) − Dx (r)Dt (s)
and
T t Dt (r) + T x Dx (r)
Tr = (46.9)
The components T x , T t depend upon (x, t, u, u(1) , · · · , u(q−1) ) which means
that T r and T s depend on (s, r, w, wr , · · · , wr(q−1) ) for solutions invariant
under X. Therefore
Ds T s + Dr T r = 0 (46.10)
529
becomes
∂T s
+ Dr T r = 0 (46.11)
∂s
or
∂T s
Z
r
T = dr + f (s) (46.12)
∂s
Let us consider next the association of X with T . Since Xv = 0 = η v , it
follows that
Xvr = ζ r = Dr η v − Dr (ξ i )vi = 0, Xvs = 0 (46.13)
since ξ s = 1 and ξ r = 0. We thus see that in canonical coordinates
XT r = 0, XT s = 0 (46.14)
In other words, T r and T s are invariant under X. This means that

∂ r ∂ s
T = 0, T =0 (46.15)
∂s ∂s
The conservation law in canonical coordinates becomes
Dr T r = 0 (46.16)
We thus arrive at the following Theorem, which follows by integrating the

conservation law (46.16) :
Theorem (Sjöberg). A PDE F = 0 of order q with two independent vari-
ables, which admits a symmetry X that is associated to a conserved vector
T , is reduced to an ODE of order q − 1, namely
Tr = k (46.17)
where T r is given by (46.9) for solutions invariant under X.

Example 1. We consider the KdV equation
F = ut + uux + uxxx = 0 (46.18)
and the corresponding Lagrangian
L = v(ut + uux + uxxx ) (46.19)
530
The adjoint equation F ∗ can be evaluated taking into account the above
expression of L using the formula
F ∗ (t, x, u, v, u(1) , u(2) , u(3) , v(1) , v(2) , v(3) )

3
δL ∂L X δL
= = + (−1)m Di1 Di2 Di3
δu ∂u m=1 δui1 i2 i3
∂L ∂L ∂L ∂L
= − Dx − Dt − Dx3 (46.20)
∂u ∂ux ∂ut ∂uxxx
3
= vux − Dx (vu) − Dt (v) − Dx (v)
= vux − u(Dx v) − vDx u − vt − vxxx
= vux − vx u − vux − vt − vxxx
= −(vt + uvx + vxxx )
where we have used the following expressions for the operators Dx and Dt
∂ ∂ ∂ ∂
Dx = + ux + uxx + uxt + ··· (46.21)
∂x ∂u ∂ux ∂ut
∂ ∂ ∂ ∂
Dt = + ut + utx + utt + ··· (46.22)
∂t ∂u ∂ux ∂ut
Therefore the adjoint to KdV equation (46.13) is
vt + uvx + vxxx = 0 (46.23)
Replacing v by u in the adjoint equation, we have that
F ∗ (t, x, u, u, u(1) , u(2) , u(3) , u(1) , u(2) , u(3) ) = −(ut + uux + uxxx )
(46.24)
= −F (t, x, u, u(1) , u(2) , u(3) )
Therefore the KdV equation is self-adjoint.

Let us now consider one generator of Lie symmetries for the KdV equation
∂ ∂ ∂
X = −x − 3t + 2u (46.25)
∂x ∂t ∂u
i.e. in component form
ξ x = −x, ξ t = −3t, η = 2u (46.26)
531
The general expression for the conserved vectors T i for the system of equa-
tions (46.13) and (46.18),
ut + uux + uxxx = 0, vt + uvx + vxxx = 0
is given by
δ(vF ) X δ(vF )
T i = ξ i (vF ) + W + Di1 · · ·Dis (W ) (46.27)
δui s≥1
δuii1 ···is
The expression for the conserved vector T x is being obtained from the above
formula for i = x, i.e.
δ(vF ) X δ(vF )
T x = ξ x (vF ) + W + Di1 · · ·Dis (W ) (46.28)
δux s≥1
δuxi1 ···is
Let us now calculate separately the relevant terms in order to get the explicit
expression for T x . First of all, we have
W = η − ξ j uj = η − ξ x ux − ξ t ut
(46.29)
= 2u − (−x)ux − (−3t)ut = 2u + xux + 3tut
We also obtain
δ(vF ) ∂(vF ) ∂(vF ) ∂(vF ) ∂(vF )
= − Dt − Dx + Dt Dx
δux ∂ux ∂utx ∂uxx ∂uxxt
∂(vF ) (46.30)
+ Dx Dx = vu + Dx Dx (v) = vu + vxx
∂uxxx
while
X δ(vF ) δ(vF ) δ(vF )
Di1 · · ·Dis (W ) = Dt (W ) + Dx (W )
s≥1
δuxi1 ···is δuxt δuxx
(46.31)
δ(vF ) δ(vF ) δ(vF )
+ Dt Dx (W ) + Dt Dt (W ) + Dx Dx (W )
δuxtx δuxtt δuxxx
The only non-zero contributions to (46.28) come from the terms
δ(vF ) ∂(vF ) ∂(vF ) ∂(vF )
= − Dx − Dt
δuxx ∂uxx ∂uxxx ∂utxx (46.32)
= −Dx (v) = −vx
532
and
δ(vF ) ∂(vF )
= =v (46.33)
δuxxx ∂uxxx
On the other hand, we get
∂W ∂W ∂W ∂W
Dx (W ) = + ux + uxt + uxx
∂x ∂u ∂ut ∂ux (46.34)
= 3ux + 3tuxt + xuxx
and
∂W ∂W ∂W ∂W ∂W
Dx Dx (W ) = + ux + uxt + uxx + uxxt
∂x ∂u ∂ut ∂ux ∂uxt
∂W
(46.35)
+ uxxx (3ux + 3tuxt + xuxx )
∂uxx
= 4uxx + 3tuxxt + xuxxx
We thus obtain from (46.28) that
T x = (−x)v(ut + uux + uxxx ) + (2u + xux + 3tut )(vu + vxx )
+ (3ux + 3tuxt + xuxx )(−vx ) + (4uxx + 3tuxxt + xuxxx )v
and finally
T x = −xvut + 2u2 v + 3tvuut + 2uvxx + xux vxx + 3tut vxx
(46.36)
− 3vx ux − 3tvx uxt − xvx uxx + 4vuxx + 3tvuxxt
Using next the formula
δ(vF ) X δ(vF )
T t = ξ t (vF ) + W + Di1 · · ·Dis (W ) (46.37)
δut s≥1
δuti1 ···is
and performing the relevant calculations, we get the following expression for
the conserved vector T t
T t = v(−3tuxxx − 3tuux + xux + 2u) (46.38)
Since KdV is a self-adjoint equation, we simply substitute v by u in the above
expressions for the conserved vectors, we obtain the following expressions for
the conserved vectors of the KdV equation
T x = −xuut + 2u3 + 3tu2t + 6uuxx + 3tut uxx
− 3u2x − 3tux uxt + 3tuuxxt (46.39)
T t = −3tuuxxx − 3tu2 ux + xuux + 2u2
533
Using the identities
Dx (−tu3 ) = −3tu2 ux , −Dt (−tu3 ) = u3 + 3tu2 ut ,
1 1 1
Dx xu2 − u2 = xuux , −Dt xu2 = −xuut ,
2 2 2
3 2
Dx −3tuuxx + tux = −3tuuxxx , (46.40)
2
3
− Dt −3tuuxx + tu2x = 3uuxx + 3tut uxx
2
3
+ 3tuuxxt − u2x − 3tux uxt
2
the expressions (46.39) of the conserved vectors take on the form
1 3
T x = −Dt xu2 − Dt (−tu3 ) − Dt −3tuuxx + tu2x
2 2
3
+ u3 + 3uuxx − u2x (46.41)
2
3 1 3
T t = Dx −3tuuxx + tu2x + Dx (−tu3 ) + Dx xu2 + u2
2 2 2
Disregarding the total derivative terms, we get the following expressions for
the conserved vectors
3
T x = u3 + 3uuxx − u2x
2
t 3 2
T = u
2
and after an obvious rescaling,
2
T x = u3 + 2uuxx − u2x
3 (46.42)
T t = u2
The generators X of point symmetries associated to vector (46.37) can be

obtained by imposing the condition
X(T i ) + T i Dk (ξ k ) − T k Dk (ξ i ) = 0 (46.43)
After a simple calculation we get
ξ t = a, ξ x = b, η=0 (46.44)
534
Choosing a = 1, b = c, so that
∂ ∂
X= +c (46.45)
∂t ∂x
we can apply the double reduction method. Going over to normal coordi-
nates, i.e. solving the system
X(r) = 0, X(s) = 1, X(w) = 0 (46.46)
we obtain
r = x − ct, s = t, w=u (46.47)
We then obtain from

T t Dt (r) − T x Dx (r)
Tr =
that
2
T r = cw2 − 2wwrr + wr2 − w3 (46.48)
3
After setting T r = k (k a constant), we get a second order equation
d2 w dw 2 2 3
2w − + w − cw2 + k = 0 (46.49)
dr2 dr 3
The substitution
dw
p= , p = p(w) (46.50)
dr
taking into account the identity
d2 w dp
2
=p (46.51)
dr dw
converts equation (46.44) to the first order ODE
dp 1 1 c k −1
− p = − w2 + w − p (46.52)
dw 2w 3 2 2w
535
The above is a Bernoulli type differential equation which can be solved by
the substitution
z = p2 , z = z(w) (46.53)
Under the above substitution, equation (46.52) becomes a linear first order
ODE
dz 1 2 k
− z = − w2 + cw − (46.54)
dw w 3 w
1
z(w) = − w3 + cw2 + C1 w + k (46.55)
3
dw p
Therefore, since p = and p = ± z(w), the function w = w(r) is a
dr
solution of the differential equation
r
dw 1
= ± − w3 + cw2 + C1 w + k (46.56)
dr 3
The above equation can be solved by separation of variables and its solution
can be expressed in terms of the Weirstrass ℘−function.
Note. The third prolongation of X is
∂ ∂ ∂ ∂ ∂ ∂
P r(3) X = ξ x + ξt + η + η [t] + η [x] + η [xxx] (46.57)
∂x ∂t ∂u ∂ut ∂ux ∂uxxx
For ξ x = −x, ξ t = −3t and η = 2u we have
η [t] =, η [x] =, η [xxx] = (46.58)
Therefore
P r(3) X[F ] = 5(ut + uux + uxxx ) (46.59)
and thus λ = 5. Therefore
η ∗ = −[λ + Di (ξ i )]v = −[5 + Dx (−x) + Dt (−3t)]v = −v (46.60)
we thus see that

∂
Y =X −v (46.61)
∂v
536
is the symmetry generator of equation (46.18).
Example 2. We consider the heat equation ut = uxx . In this case we shall
consider two sets of generators:
Case I. We consider first the symmetry generators
∂ ∂
X1 = , X2 = (46.62)
∂x ∂t
which are associated with the conservation law
Dt (u) + Dx (−ux ) = 0 (46.63)
Let
X = X 1 + cX 2 (46.64)
Introducing canonical coordinates
s = x, r = cx − t (46.65)
the conservation law can be written as
Ds T s + Dr T r = 0 (46.66)
with
T r = u + cux = u + c2 ur (46.67)
Since T = (T r , T s ) is associated with X, according to the Theorem we have
T r = k1 =⇒ u + c2 ur = k1 (46.68)
The last equation can be integrated yielding the solution
r
u = k1 + k2 exp 2 (46.69)
c
Therefore
t − cx
u = k1 + k2 exp (46.70)
c2
is a solution of the heat equation ut = uxx invariant under X.
Case II. The heat equation admits also the generators
∂ ∂ ∂ ∂
X3 = u , X 4 = 2x + 4t − u (46.71)
∂u ∂x ∂t ∂u
537
There is no conservation law associated with X 3 . The conserved vector
f 0 u f 00 u f 0 u
x
T = 1/4 , 3/4 − 1/4 (46.72)
t t t
is associated with X 4 where
x
f y=√ (46.73)
t
satisfies the equation
1 1
f 00 − yf 0 + f = 0 (46.74)
2 4
The generators X 3 and X 4 commute. Thus if we use X 4 to reduce ut = uxx ,
X 3 will be a local symmetry of the reduced equation. Generator X 4 can be
written in terms of canonical coordinates
∂ x 1
X4 = with r = √ , s = ln(t), v = ut1/4 (46.75)
∂s t 4
The conservation law becomes
Ds T s + Dr T r = 0 (46.76)
where
T s = −f 0 v, T r = (2rf 0 − 4f 00 )v + 4f 0 vr (46.77)
and v = v(r) for solutions invariant under X 4 . We thus obtain, according to

the Theorem, after setting T r = k, i.e.
(2rf 0 − 4f 00 )v + 4f 0 vr = k (46.78)
In the new variables,

∂
X3 = v (46.79)
∂v
∂ vr
We transform X 3 to X 3 = with w = ln(v). Then wr = and equation
∂w v
(46.78) becomes
−4f 00 + 2rf 0 + 4wr f 0 = k (46.80)
538
r2
Z
k 0
w= dr + ln |f | − + ln(C) (46.81)
4f 0 4
with w = ln(v) and v = ut1/4 . Therefore
x2 Z k
−1/4 0
u=Ct f · exp − + dr (46.82)
4t 4f 0 √
r=x/ t
is a solution of the heat equation ut = uxx invariant under X 3 and X 4 .

Here f (r) satisfies the equation
1 1
f 00 − rf 0 + f = 0 (46.83)
2 4
whose solution can be expressed in terms of Bessel’s functions of the second
kind.
Example 3. The BBM equation
ut = utxx + uux (46.84)
admits the symmetry generators
∂ ∂
X1 = and X 2 = (46.85)
∂x ∂t
The conserved vectors associated to both the above generators are given by
1
1 2
T = u, −utx − u
2
1 1 2 1
2 2
T = u + ux , −uutx − (46.86)
2 2 3
1 1 4
3 3 2 2 2
T = u , ut − utx − u utx − u
3 4
1 2
We consider X = αX + X . The canonical coordinates of X are
s = t, r = αt − x, u = u(r) (46.87)
The r−components (T i )r = α(T i )t + (T i )x of the conserved vectors are
1 1
(T 1 )r = αu − utx − u2 = αu − αurr − u2
2 2
1 1
(T 2 )r = (αu2 − u3 ) + α3 u2r + αuurr (46.88)
3 2
1 1
(T 3 )r = αu3 − u4 + u2 − α2 u2rr + αu2 urr
3 4
539
We thus have three reduced equations
(T 1 )r = k1 , (T 2 )r = k2 , (T 3 )r = k3 (46.89)
The first reduced equation (T 1 )r = k1 is given in explicit form
1
αu − αurr − u2 = k1 (46.90)
2
dur 1 1
= αu − u2 − k1 du (46.91)
ur α 2
and by integration
1 1 3 k1
2
ur = C1 exp u − u − u (46.92)
2 6α α
A second integration with respect to r gives
Z 1 1 3 k1
exp u2 − u − u du = C1 r + C2 (46.93)
2 6α α
We thus see that
Z 1 1 3 k1
exp u2 − u − u du = C1 (αt − x) + C2 (46.94)
2 6α α
is a 4−parameter family of solutions of the BBM equation invariant under
the vector X.
The second reduced equation (T 2 )r = k2 is given in explicit form
1 1
(αu2 − u3 ) + α3 u2r + αuurr = k2 (46.95)
3 2
The third reduced equation (T 3 )r = k3 is given in explicit form
1 3 1 4
αu − u + u2 − α2 u2rr + αu2 urr = k3 (46.96)
3 4
Example 4. We shall apply the method of double reduction in a system of
PDEs. We consider the system
ρt + vρx + ρvx = 0
ρ(vt + vvx ) = 0 (46.97)
pt + vpx + λpvx = 0
540
Case I. For λ = 3, the system admits the symmetry generator
∂ ∂ ∂ ∂ ∂
X 1 = t2 + tx + (x − tv) − 3tp − tρ (46.98)
∂t ∂x ∂v ∂p ∂ρ
There are three conserved vectors associated to X 1
T 1 = (ρ, ρv)

T 2 = ρ(tv − x), ρv(tv − x) + tp (46.99)

3 2 2 2
T = ρ(tv − x) + t p, ρv(tv − x) + tp(3tv − 2x)
∂
The generator X 1 has a canonical form X 1 = with
∂s
dt dx dv dp dρ
2
= = = =
t tx x − tv 3tp −tρ
i
(46.100)
ds dr dw
= = = , i = 1, 2, 3
1 0 0
or
x 1 x
r= , s = − , w1 = x − v , w2 = pt3 , w3 = ρt (46.101)
t t t
The r−component of T 1 , T 2 , T 3 can be calculated by the use of the formula
(46.9):
w1 w3 w3 (w1 )2
(T 1 )r = , (T 2 )r = − − w2 ,
r r2 (46.102)
w3 (w1 )3 3w1 w2
(T 3 )r = +
r3 r
We thus see that the reduced system (when λ = 3) is
w1 w3 w3 (w1 )2
= k1 , + w2 = k2 ,
r r2 (46.103)
w3 (w1 )3 3w1 w2
+ = k3
r3 r
The above system has a solution
3 k2 r p
w1 = t± 9k2 r − 8k1 k3
4 k1 4k1
1 1p
w 2 = k2 ∓ 9k2 r − 8k1 k3 (46.104)
4 4
4k 2
w3 = √ 1
3k2 ± 9k2 r − 8k1 k3
541
Using the original variables, we get the following solution of the system
4k 2
ρ= r 1
x
3k2 t±t 9k2 − 8k1 k3
t
r
1 1 x (46.105)
p= 3
∓ 3 9k2 − 8k1 k3
4k2 t 4t t
r
3k2 1 x x
v=− ∓ 9k2 − 8k1 k3 +
4k1 t 4k1 t t t
Case II. We consider next the generator
∂ ∂ ∂
X2 = t −v + 2ρ (46.106)
∂t ∂v ∂ρ
which is associated with
1 1 1 λ
T4 = ρv 2 + p, ρv 3 + vp (46.107)
2 λ−1 2 λ−1
for λ arbitrary. Introducing canonical coordinates, we have
ρ
r = x, s = ln(t), w1 = p, w2 = vt, w3 = (46.108)
t2
Using equation (46.9), we get
1 λ
(T 4 )r = − ρtv 3 + tvp = k1 (46.109)
2 λ−1
Solving the previous equation with respect to p, we find
λ − 1 1 1
p= k1 + ρt v 3 (46.110)
λ 2 tv
The other two equations of the original system give
ρt + (vρ)x = 0 =⇒ 2w3 + w3 wr2 + w2 wr3 = 0 (46.111)
and
ρ(vt + vvx ) = 0 =⇒ w3 (w2 )(wr2 − 1) = 0 (46.112)
542
f (x) λ−1 1
If ρ = 0, v = , p = k1 ·
t λ f (x)
If ρ = 0, v = 0, p(x) is arbitrary.
kt2 x+c C1
If ρ = 3
,v= ,p=
(x + c) t x+c
References
[1] A. Sjöberg: ”Double Reduction of PDEs from the association of
symmetries with conservation laws with applications”.
Appl. Math. Comp. 184 (2007) 608-616.
[2] A. Sjöberg: ”On double reductions from symmetries and conservation
laws”. Nonlinear Analysis: Real World Applications 10 (2009) 3472-3477
[3] G. L. Caraffini and M. Galvani: ”Symmetries and exact solutions
via conservation laws for some partial differential equations of Mathematical
Physics”. Appl. Math. Comp. 219 (2012) 1474-1484
47 The Double Reduction II

In this section we generalize the double reduction method to functions of
more that two variables.
The theoretical part is an extension of the one-variable case and it is being
exhibited in the reference at the end of this section.
Example. We consider a nonlinear (2 + 1) wave equation
utt − (f (u)ux )x − (g(u)uy )y = 0 (47.1)
where f (u) and g(u) are arbitrary smooth functions. This equation admits
the conserved vector
T = (−ut , f (u)ux , g(u)uy ) (47.2)
and the following generators of Lie symmetries

∂ ∂ ∂
X1 = , X2 = , X3 = ,
∂t ∂x ∂y
(47.3)
∂ ∂ ∂
X4 = t + x +y
∂t ∂x ∂y
543
The associated symmetry X satisfies the equation
 t    t
T Dt ξ t Dx ξ t Dy ξ t T
X T x  − Dt ξ x x
Dx ξ Dy ξ x  x
T
Ty Dt ξ y y
Dx ξ Dy ξ y
Ty
 t (47.4)
T
t x y  x
+ (Dt ξ + Dx ξ + Dy ξ ) T =0
Ty
Considering the associated symmetry generators to be X 1 , X 2 and X 3 , the

linear combination
∂ ∂ ∂
X= + C1 + C2 (47.5)
∂t ∂x ∂y
∂
The above generator is being reduced to canonical form X = if
∂q
dt dx dy du dr ds dq dw
= = = = = = = (47.6)
1 C1 C2 0 0 0 1 0
i.e.
r = y − C2 t, s = x − C1 t, q = t, w(r, s) = u (47.7)
The conserved vector with respect to the new coordinates (r, s, q) satisfies
the equation
 r  t
T T
T s  = J(A−1 )T T x  (47.8)
q y
T T
where
 
Dt r Dt s Dt q
A−1 = Dx r Dx s Dx q  and J = det(A) (47.9)
Dy r Dy s Dy q
Then the reduced conserved form is
Dr T r + Ds T s = 0 (47.10)
544
where
T r = C22 wr + C1 C2 ws − g(w)wr
T s = C1 C2 wr + C12 ws − f (w)ws (47.11)
T q = −C2 wr − C1 ws
The reduced conserved form admits the symmetry

∂ ∂
X̃4 = r +s (47.12)
∂r ∂s
The reduced conserved form for the PDE by the associated symmetry satisfies
the formula
r r r
T Dr ξ r Ds ξ r T r s T
X − + (Dr ξ + Ds ξ ) = 0 (47.13)
Ts Dr ξ s Ds ξ s Ts Ts
It is clear that X̃4 is an associated symmetry and we thus can obtain a

∂ ∂
reduced conserved form by Y = r +s where the generator Y has the
∂r ∂s
∂
canonical form Y = when
∂m
dr ds dw dn dm dv
= = = = = (47.14)
r s 0 0 1 0
or
s
n= , m = ln(r), v(n) = w (47.15)
r
We can get the reduced conserved form using the formula
n r
T −1 T T
= J(A ) (47.16)
Tm Ts
where

−1 Dr n Dr m
A = and J = det(A) (47.17)
Ds n Ds m
Then the reduced conserved form is
Dn T n = 0 (47.18)
545
where
T n = vn [−C22 n2 + 2C1 C2 n + n2 g(v) + f (v) − C12 ]
(47.19)
T m = −vn [−C22 n + ng(v) + C1 C2 ]
The second step of double reduction is given by
vn [−C22 n2 + 2C1 C2 n + n2 g(v) + f (v) − C12 ] = C (47.20)
where C is a constant with
x − C1 t
n= and v = u (47.21)
y − C2 t
References
[1] A. H. Bokhari, A. Y. Dweik, F. D. Zaman, A. H. Kara and
F. M. Mahomed: ”Generalization of the double reduction theory”.
Nonlinear Analysis: Real World Applications 11 (2010) 3763-3769.
[2] S. C. Anco and G. W. Bluman: ”New conservation laws obtained
directly from symmetry action on a known conservation law”.
J. Math. Anal. Appl. 322 (2006) 233250.
Chapter XVI
Ricci Flow
The Ricci flow is an evolution equation introduced by Hamilton and used by
Grigori Perelman in proving the Poincare conjecture.
48 Two-Dimensional Ricci Flow

In this section we shall consider the two-dimensional Ricci flow. This equa-
tion has the form
∂
g(t) = −2Ric (48.1)
∂t
where g is the metric of a Riemannian manifold and Ric is Ricci’s tensor.
Using Cartesian coordinates (X, Y ) or complex variables
2x = Y + iX, 2y = Y − iX (48.2)
546
the metric tensor can be written in a conformally flat space as
1
ds2 = gij dxi dxj = 2eφ(x,y,t) dxdy = eφ(X,Y,t) (dX 2 + dY 2 ) (48.3)
2
Using equation (48.1), we see that function φ(X, Y, t) satisfies the equation
∂ φ
e = ∆φ (48.4)
∂t
where ∆ is the Laplacian. Introducing the field
u(x, y, t) = eφ (48.5)
ut = (ln(u))xy (48.6)
or the equivalent form
u2 ut + ux uy − uuxy = 0 (48.7)
We shall analyse equation (48.7) using Lie Symmetries.
Let us denote by ∆ the expression
∆ = u2 ut + ux uy − uuxy (48.8)
Introducing the second prolongation
∂ ∂ ∂ ∂
P r(2) Γ = T +X +Y +U
∂t ∂x ∂y ∂u
(48.9)
∂ ∂ ∂ ∂
+ U [t] + U [x] + U [y] + U [xy]
∂ut ∂ux ∂uy ∂uxy
invariance of the equation is being expressed by the condition
P r(2) Γ(∆) =0 (48.10)

∆=0
We now have
P r(2) Γ(∆) = 0⇐⇒

∂ ∂ ∂ ∂ ∂
T +X +Y +U + U [t]
∂t ∂x ∂y ∂u ∂u (48.11)
t
∂ ∂ ∂
+ U [x] + U [y] + U [xy] (u2 ut + ux uy − uuxy ) = 0
∂ux ∂uy ∂uxy
547
2uut U − uxy U + U [t] u2 + U [x] uy + U [y] ux + U [xy] (−u) = 0 (48.12)
Let us now list the explicit expressions of the coefficients U [t] , U [x] , U [y] and
U [xy] . We have that
U [t] = Ut + (Uu − Tt )ut − Xt ux − Yt uy
(48.13)
− Xu ut ux − Yu ut uy − Tu (ut )2
U [x] = Ux + (Uu − Xx )ux − Tx ut − Yx uy

(48.14)
− Yu ux uy − Tu ux ut − Xu (ux )2
U [y] = Uy + (Uu − Yy )uy − Ty ut − Xy ux
(48.15)
− Xu ux uy − Tu ut uy − Yu (uy )2
U [xy] = Uxy + (Uu − Xx − Yy )uxy − (2Xu ux + 2Yu uy + Tu ut )uxy
− Tu uy uxt − Ty uxt − Xu uy uxx − Xy uxx − Tu ux uyt − Tx uyt
− Yu ux uyy − Yx uyy − Xuu (ux )2 uy − Xxu ux uy − Xxy ux
(48.16)
− Xyu (ux )2 − Yuu ux (uy )2 − Yxu (uy )2 − Yxy uy − Yyu ux uy
− Tuu ut ux uy − Txu ut uy − Txy ut − Tyu ut ux + Uuu ux uy
+ Uxu uy + Uyu ux
548
Taking into account (48.12) and the above expressions, we obtain the equa-
tion

2uut U − uxy U + Ut + (Uu − Tt )ut − Xt ux − Yt uy

− Xu ut ux − Yu ut uy − Tu (ut ) u2
2

+ Ux + (Uu − Xx )ux − Tx ut − Yx uy

2
− Yu ux uy − Tu ux ut − Xu (ux ) uy

+ Uy + (Uu − Yy )uy − Ty ut − Xy ux
(48.17)
2
− Xu ux uy − Tu ut uy − Yu (uy ) ux

− Uxy + (Uu − Xx − Yy )uxy − (2Xu ux + 2Yu uy + Tu ut )uxy

− Yu ux uyy − Yx uyy − Xuu (ux )2 uy − Xxu ux uy − Xxy ux
− Xyu (ux )2 − Yuu ux (uy )2 − Yxu (uy )2 − Yxy uy − Yyu ux uy
− Tuu ut ux uy − Txu ut uy − Txy ut − Tyu ut ux

+ Uuu ux uy + Uxu uy + Uyu ux u = 0
We now have to take into account the condition ∆ = 0 substituting uxy by

uut + (ux uy )/u, i.e.
ux uy
uxy −→ uut + (48.18)
u
549
ux uy
2uut U − uut + U
u
+ Ut + (Uu − Tt )ut − Xt ux − Yt uy

− Xu ut ux − Yu ut uy − Tu (ut ) u2
2

+ Ux + (Uu − Xx )ux − Tx ut − Yx uy

2
− Yu ux uy − Tu ux ut − Xu (ux ) uy

+ Uy + (Uu − Yy )uy − Ty ut − Xy ux
(48.19)
− Xu ux uy − Tu ut uy − Yu (uy )2 ux
ux uy
− Uxy + (Uu − Xx − Yy ) uut +
u
ux uy
− (2Xu ux + 2Yu uy + Tu ut ) uut +
u
− Yu ux uyy − Yx uyy − Xuu (ux )2 uy − Xxu ux uy − Xxy ux − Xyu (ux )2
− Yuu ux (uy )2 − Yxu (uy )2 − Yxy uy − Yyu ux uy − Tuu ut ux uy

− Txu ut uy − Txy ut − Tyu ut ux + Uuu ux uy + Uxu uy + Uyu ux u = 0

the function u. First of all, we see that the coefficients of uy uxt and ux uyt is
equal to u·Tu , the coefficient of uxt is equal to u·Ty and the coefficient of uyt
is equal to u·Tx . We thus have
Tu = Tx = Ty = 0 (48.20)
and thus
T = α(t) (48.21)
We see further that the coefficient of uy uxx is equal to u·Xu , the coefficient
of uxx is equal to u·Xy , the coefficient of ux uyy is equal to u·Yu and the
550
coefficient of uyy is equal to u·Yx . Therefore
Xu = Xy = 0 and Yu = Yx = 0 (48.22)
and thus
X = β(t, x) and Y = γ(t, y) (48.23)

u u
x y
uut U − U
u
+ Ut + (Uu − Tt )ut − Xt ux − Yt uy u2

+ Ux + (Uu − Xx )ux uy
(48.24)
+ Uy + (Uu − Yy )uy ux
ux uy
− Uxy + (Uu − Xx − Yy ) uut +
u

+ Uuu ux uy + Uxu uy + Uyu ux u = 0
Equate now the coefficients of the various derivatives of the function u to

zero. We have
Coefficient of ux uy :
U
− + (Uu − Xx ) + (Uu − Yy ) − (Uu − Xx − Yy ) − u·Uuu = 0
u
The above equation can be simplified into
−U + u·Uu − u2 ·Uuu = 0 (48.25)
Coefficient of ux :
−u2 ·Xt + Uy − u·Uyu = 0 (48.26)
Coefficient of uy :
−u2 ·Yt + Ux − u·Uxu = 0 (48.27)
551
Coefficient of ut :
u·U + (Uu − Tt )u2 − (Uu − Xx − Yy )u2 = 0
The previous equation is equivalent to
U = (Tt − Xx − Yy )u (48.28)
Coefficient of zeroth order derivative
Uxy = u·Ut (48.29)
We are going to solve the system of equations (48.25)-(48.29).

From (48.28), taking into account (48.21) and (48.23) we find upon differ-
entiation that Ux = −Xxx u and Uxu = −Xxx . Therefore equation (48.27)
becomes Yt = 0 and then
Y = Y (y) (48.30)
Similarly combining (48.28) to (48.27), we obtain Xt = 0 and thus
X = X(x) (48.31)
From (48.28), taking into account (48.30) and (48.31) we obtain Uxy = 0 and
Ut = Ttt u and thus equation (48.29) takes on the form 0 = u2 Ttt from which
we get Ttt = 0. Therefore
T = a1 t + a2 (48.32)
Equation (48.25) becomes an identity taking into account the above found
relations.
We see further that the functions X = X(x) and Y = Y (y) remain undeter-
mined. We then can assign any values to these two functions. Considering
for example that (Tt −Xx −Yy ) is a constant and taking into account (48.30)-
(48.32) we conclude that X and Y should be linear, i.e.
X = a3 x + a4 , Y = a5 y + a6 (48.33)
Therefore we have from (48.28)
U = (a1 − a3 − a5 )u (48.34)
552
The generator of the symmetries of the 2-dimensional Ricci flow equation is
then given by
∂ ∂ ∂ ∂
Γ=T +X +Y +U
∂t ∂x ∂y ∂u
∂ ∂ ∂
= (a1 t + a2 ) + (a3 x + a4 ) + (a5 y + a6 ) (48.35)
∂t ∂x ∂y
∂
+ (a1 − a3 − a5 )u
∂u
The above vector operator can also be written as
∂ ∂ ∂ ∂ ∂
Γ = a1 t + u + a2 + a3 x −u
∂t ∂u ∂t ∂x ∂u
(48.36)
∂ ∂ ∂ ∂
+ a4 + a5 y −u + a6
∂x ∂y ∂u ∂y
We thus see that the Lie algebra of symmetries of the original equation is
being spanned by the following six generators
∂ ∂ ∂ ∂ ∂
Γ(1) = t + u , Γ(2) = , Γ(3) = x −u ,
∂t ∂u ∂t ∂x ∂u
(48.37)
∂ ∂ ∂ ∂
Γ(4) = , Γ(5) = y − u , Γ(6) =
∂x ∂y ∂u ∂y
We are now able to determine some general solutions of the original PDE
(48.7) using the previously determined symmetries.
We consider for example the combination Γ(4) + Γ(5) , i.e.
∂ ∂ ∂
Γ(4) + Γ(5) = +y −u (48.38)
∂x ∂y ∂u
Lagrange’s system of equations associated to (48.38), are given by
dx dy dt du
= = = (48.39)
1 y 0 −u
We find three independent solutions of the above system
y = C1 ex , t = C2 , yu = C3 (48.40)
We may thus consider two invariants
z = y e−x and u = f (z, w)y −1 (48.41)
553
We now have to convert the original equation to another PDE with unknown
function f (z, w). Since
∂f −1 ∂f ∂z ∂f ∂w −1
ut = y = + y = (fz ·0 + fw ·1)y −1 = fw y −1 (48.42)
∂t ∂z ∂t ∂w ∂t
∂f −1 ∂f ∂z ∂f ∂w −1
ux = y = + y = [fz ·(−ye−x ) + fw ·0]y −1
∂x ∂z ∂x ∂w ∂x (48.43)
= −e−x fz
∂ ∂ ∂ ∂f ∂ 2 f ∂z
uxy = ux = (−e−x fz ) = −e−x = −e−x 2
∂y ∂y ∂y ∂z ∂z ∂y
2
(48.44)
∂ f
= −e−x 2 e−x = −e−2x fzz
∂z
∂f −1 ∂f ∂z ∂f ∂w −1
uy = y − f ·y −2 = + y − f ·y −2
∂y ∂z ∂y ∂w ∂y (48.45)
= [fz ·e−x + fw ·0]y −1 − f ·y −2 = fz ·e−x y −1 − f ·y −2
Upon substitution of (48.42)-(48.45) into the original equation (48.7) we ob-
tain
u2 ut + ux uy − uuxy = (f 2 fw − z 2 fz2 + zf fz + z 2 f fzz )y −3 = 0 (48.46)
and thus
f 2 fw − z 2 fz2 + zf fz + z 2 f fzz = 0 (48.47)
We see that we have been able to reduce the original PDE into another
one with two independent variables. Since equation (48.47) is a nonlinear
PDE with two independent variables, we have to perform again a Lie point
symmetry analysis in order to reduce further this equation into an ordinary
differential equation.
Chapter XVII
Applications in Quantum
Mechanics
554
49 The Schrödinger Equation
Example. In this Example we consider the one-dimensional Schrödinger
equation
ut = uxx + V (x)u (49.1)
where V (x) is the potential.

We shall determine the condition the Potential V (x) should satisfy so that
Schrödinger equation be Lie-invariant.
We denote by ∆ the expression
∆≡ut − uxx − V (x)u (49.2)

We have for the second extension P r(2) Γ
P r(2) Γ(∆) = 0 ⇔
∂ ∂ ∂ ∂ ∂
⇔ T +X +U + U [t] + U [x]
(49.3)
[tt] ∂ [tx] ∂ [xx] ∂
+U +U +U ut − uxx − V (x)u = 0
⇔ −XV 0 (x)u − U V (x) + U [t] − U [xx] = 0
Substitution the known expressions of U [t] and U [xx] in the above expression
− X V 0 (x)u − U V (x)

+ Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )

(49.4)
2 2 3
+ (Uuu − 2Xxu )(ux ) − Tuu ut (ux ) − Xuu (ux )

555
In order to take into account the condition ∆ = 0, we substitute in the
previous equation ut by uxx + V (x)u. We then obtain the following equation
− X V 0 (x)u − U V (x)

+ Ut + (Uu − Tt − Xu ux )(uxx + V (x)u)

2
− Tu (uxx + V (x)u) − Xt ux

(49.5)
− Uxx − (Txx + 2Txu ux )(uxx + V (x)u) + (2Uxu − Xxx )ux
+ (Uuu − 2Xxu )(ux )2 − Tuu (uxx + V (x)u)(ux )2 − Xuu (ux )3


− Tu (uxx + V (x)u)uxx − 3Xu ux uxx = 0
We have to equate to zero all the coefficients of the partial derivatives of the
unknown function u.
First of all, we observe that the coefficients of utx and ux utx are −2Tx and
−2Tu respectively, which means that Tx = Tu = 0 and thus
T = T (t) (49.6)
On the other hand, equating to zero the coefficient of ux uxx we get the
equation Xu = 0, which means that X is independent of u and thus
X = X(x, t) (49.7)
Taking the two previous equations into account, equation (49.5) becomes
− X V 0 (x)u − U V (x)

+ Ut + (Uu − Tt )(uxx + V (x)u) − Xt ux
(49.8)

− Uxx + (2Uxu − Xxx )ux + Uuu (ux )2

+ (Uu − 2Xx )uxx = 0
556
From the above equation we get the following:
Coefficient of u2x :
Uuu = 0 (49.9)
Coefficient of ux :
−Xt − 2Uxu + Xxx = 0 (49.10)
2Xx − Tt = 0 (49.11)
Zeroth order derivative coefficient:
−X V 0 (x)u − U V (x) + Ut + (Uu − Tt )V (x)u − Uxx = 0 (49.12)
From (49.9) we see that U is a linear function with respect to u:
U = α(x, t)u + β(x, t) (49.13)
Using the above expression of U , equation (49.10) takes on the form
−Xt − 2αx + Xxx = 0 (49.14)
Equation (49.11) upon integration with respect to x gives

1
X = T 0 (t)x + γ(t) (49.15)
2
Therefore (49.14) because of the above expression of X, gives
1 1
αx = − T 00 (t)x − γ 0 (t) (49.16)
4 2
and integrating with respect to x, we get
1 1
α(x, t) = − T 00 (t)x2 − γ 0 (t)x + δ(t) (49.17)
8 2
Similarly, equation (49.12) because of (49.13) becomes
{−XV 0 − T 0 V + αt − αxx }u + (−βV + βt − βxx ) = 0 (49.18)
557
From the above equation, since it should be true for any u, we get the two
equations
−XV 0 − T 0 V + αt − αxx = 0 (49.19)
and
−βV + βt − βxx = 0 (49.20)
Equation (49.19), because of (49.17) and (49.15), gives

1 1
− T (t)x + γ(t) V 0 (x) − T 0 (t)V (x) − T 000 (t)x2
0
2 8 (49.21)
1 00 1
− γ (t)x + δ 0 (t) + T 00 (t) = 0
2 4
Since we do not know the explicit expression of the potential V (x), we can-
not safely equate to zero the various coefficients involved. We thus have to
consider a number of various choices for the potential.
50 A Nonlinear Schrödinger equation

Example. In this Example we consider the one-dimensional nonlinear Schrödinger
equation
iΨt = −Ψxx + V (x)Ψ + g(x)|Ψ|2 Ψ, Ψ = Ψ(x, t) (50.1)
where V (x) is an external potential and g(x) is a function which describes

the spatial modulation of nonlinearity.
This equation appears for example in Nonlinear Optics and the dynamics of
Einstein-Bose condensates.
Looking for localized stationary solutions of the form
Ψ(x, t) = u(x)·e−iλt (50.2)
we arrive at the following nonlinear eigenvalue problem
−uxx + V (x)u + g(x)u3 = λu (50.3)
with
lim u(x) = 0 (50.4)

|x|→∞
558
We wish to determine the possible pairs {V (x), g(x)} so as equation (50.3)
be invariant under Lie point symmetries.
Invariance of (50.3) is being expressed by the condition
P r(2) Γ(∆) =0 (50.5)

∆=0
where
∆≡ − uxx + V (x)u + g(x)u3 − λu (50.6)
We have for the second extension

P r(2) Γ(∆) = 0 ⇔
∂ ∂ [x] ∂ [xx] ∂

⇔ X +U +U +U −uxx + V (x)u
∂x ∂u ∂ux ∂uxx (50.7)

+ g(x)u3 − λu = 0
⇔ X{V 0 (x)u + g 0 (x)u3 } + U {V (x) + 3g(x)u2 − λ} − U [xx] = 0
Upon substitution of U [xx] into the above equation, we obtain the equation
X{V 0 (x)u + g 0 (x)u3 } + U {V (x) + 3g(x)u2 − λ}

(50.8)

We now have to take into account the condition ∆ = 0. We thus make the
substitution uxx = F where
F = F (x, u) = V (x)u + g(x)u3 − λu (50.9)
Equation (50.8) then takes on the form
X{V 0 (x)u + g 0 (x)u3 } + U {V (x) + 3g(x)u2 − λ}

(50.10)

+ (Uu − 2Xx )F − 3Xu F ux = 0
559
function u. We have that
Xuu = 0 (50.11)
Uuu − 2Xxu = 0 (50.12)
2Uxu − Xxx − 3F Xu = 0 (50.13)
The constant term is given by
X{V 0 (x)u + g 0 (x)u3 } + U {V (x) + 3g(x)u2 − λ}

(50.14)
− Uxx − (Uu − 2Xx )F = 0
We have to solve the system of equations (50.11)-(50.14). From equation

(50.11) upon integration with respect to u, we obtain
X = α(x)u + β(x) (50.15)
From (50.12), using the above expression for X, we get Uuu = 2α0 (x) and
upon integration with respect to u we get
U = α0 (x)u2 + γ(x)u + δ(x) (50.16)
From equation (50.13), using (50.15), (50.16) and the explicit expression for
F , we obtain the equation
−3g(x)α(x)u3 + 3[α00 (x) − (V (x) − λ)α]u + 2γ 0 (x) − β 00 (x) = 0 (50.17)
Equation to zero the coefficients of u we see that (since g(x)6=0)
α(x) = 0 and 2γ 0 (x) − β 00 (x) = 0 (50.18)
We thus have obtained so far
X = β(x), U = γ(x)u + δ(x), 2γ 0 (x) − β 00 (x) = 0 (50.19)
560
Finally, equation (50.14) yields
{2[γ(x) + β 0 (x)]g(x) + β(x)g 0 (x)}u3 + 3g(x)δ(x)u2

+ {β(x)V 0 (x) − γ 00 (x) + 2[V (x) − λ]β 0 (x)}u (50.20)
+ δ(x)V (x) + [V (x) − λ]δ(x) − δ 00 (x) = 0
From the above equation, equating to zero the coefficients of u, we get that
δ(x) = 0 and
2[γ(x) + β 0 (x)]g(x) + β(x)g 0 (x) = 0

(50.21)
β(x)V 0 (x) − γ 00 (x) + 2[V (x) − λ]β 0 (x) = 0
From the last of (50.18), upon integration with respect to x, we obtain

1
γ(x) = β 0 (x) + C (50.22)
2
and because of that, the first equation of (50.21) becomes
g 0 (x) β 0 (x) 2C
=− 3 + (50.23)
g(x) β(x) β(x)
Integration of the above equation gives
Z x
g(0) ds
g(x) = 3 ×exp −2C (50.24)
β (x) 0 β(s)
Let us now collect everything together:

1
X = β(x), U = γ(x)u, γ(x) = β 0 (x) + C,
2 (50.25)
1
β(x)V 0 (x) − β 000 (x) + 2 [V (x) − λ] β 0 (x) = 0
2
with g(x) given by (50.24).
Once the function β(x) is given, we can determine the form of the potential
V (x) from the second of (50.25).
Chapter XVIII
Applications in Epidemiology
561
51 The SIR Model
The most known model in Epidemiological research is the so-called SIR
model. This model consists of a nonlinear system of ordinary differential
equations which models the time evolution of three quantities:
S = S(t) denotes the susceptible individuals

I = I(t) denotes the infectious class (51.1)
R = R(t) denotes the recovery class
The system of equations is

dS SI
= −β
dt S+I +R
dI SI
=β − µI (51.2)
dt S+I +R
dR
= µI − αR
dt
where β, µ, α are positive parameters.
Let us now try to solve the above system. We solve the first equation with
respect to R
SI
R = −S − I − β (51.3)
dS
dt
We differentiate with respect to t the above equation,
dI dS dS d2 S
dR dS dI S + I − SI
=− − −β dt dt dt dt2 (51.4)
dS 2
dt dt dt
dt
and then we make the following substitutions

dR SI dI dS
→ µI − αR → µI − α − S − I − β , →− − µI (51.5)
dt dS dt dt
dt
562

SI
µI − α − S − I − β
dS
dt
dS dS dS d2 S
dS dS S − − µI + I − SI
=− − − − µI − β dt dt dt dt2
dS 2
dt dt
dt
The above equation can be written in equivalent form as
dS dS d2 S

SI
−µSI + (I − S) − SI
α S+I +β = −β dt dt dt2
dS dS 2
dt dt
dS 2
and multiplying by both members, we obtain
dt
dS 2 dS
α(S + I) + αβSI
dt dt
dS dS d2 S
= −β −µSI + (I − S) + βSI 2
dt dt dt
d2 S
Solving the above equation with respect to , we obtain the equation
dt2
d2 S dS 1 α + β α − β dS 2
= (α − µ) + + (51.6)
dt2 dt β S I dt
Case I. For β = α, the above equation takes on the form
d2 S dS 2 dS 2
= (α − µ) + (51.7)
dt2 dt S dt
written also as
SS 00 = (α − µ)S 0 + 2(S 0 )2 (51.8)
Once we solve the above equation, we have to determine next the other two
functions I(t) and R(t). In fact, introducing a new function V by
V =S+I (51.9)
563
and adding the first two equations of (51.2), we obtain
dV
= −µV + µS, I =V −S (51.10)
dt
We thus see that S can be determined using (51.8), we then determine V
solving the first of (51.10) and then I = V − S. The function R is then
determined by the last of (51.2). We thus have obtained the following
Theorem. The solution of the system (51.2) for β = α is being determined
by
d2 S dS 2 dS 2
= (α − µ) +
dt2 dt S dt
dV
= −µV + µS (51.11)
dt
dR
= µI − αR
dt
where
I =V −S (51.12)
Let us now turn into equation (51.7) (the first equation of (51.11)). We are
going to use three different methods in solving equation (51.7):
First Method. Dividing by S 2 this equation, we obtain the equation
S 00 S0 S 0 2
− (α − µ) − 2 =0 (51.13)
S S S
Introducing the substitution
S0
G= (51.14)
S
and using the identity
S 00
= G0 + G2 (51.15)
S
dG
= G2 + (α − µ)G (51.16)
dt
564
The above is a Riccati-type of equation which can be solved by using the
substitution
U0
G=− (51.17)
U
The above substitution converts (51.11) into the equation
U 00 − (α − µ)U 0 = 0 (51.18)
U = C1 + C2 e(α−µ)t (51.19)
S0 U 0
Upon combining (51.9) and (51.12), i.e. + = 0 and integration, we get
S U
A A
S= =
U C1 + C2 e(α−µ)t
B
S(t) = (51.20)
1 + A e(α−µ)t
The second equation of (51.11), because of the above expression of S(t), takes
on the form
dV µB
= −µV + (51.21)
dt 1 + A e(α−µ)t
µB
V (t) = Ce−µt − ×
A(α − 2µ)
α − 2µ 2α − 3µ (51.22)
1 −(α−µ)t −(α−µ)t
×2 F1 1, ; ;− e e
α−µ α−µ A
where we have used the formula
eµt
Z
dt
1 + A e(α−µ)t
α − 2µ 2α − 3µ (51.23)
1 1 −(α−µ)t −(α−2µ)t
=− ×2 F1 1, ; ;− e e
A(α − 2µ) α−µ α−µ A
565
The hypergeometric function 2 F1 (a, b; c; x) is defined by the series expansion
ab x a(a + 1)b(b + 1) x2
2 F1 (a, b; c; x) =1+ + + ··· (51.24)
c 1! c(c + 1) 2!
We then obtain that
I =V −S
B
= Ce−µt − (51.25)
1 + A e(α−µ)t
µB α − 2µ 2α − 3µ 1
− ×2 F1 1, ; ; − e−(α−µ)t e−(α−µ)t
A(α − 2µ) α−µ α−µ A
We finally have to evaluate the function R(t) using the third equation of
(51.11):
dR
= µI − αR (51.26)
dt
Z
−αt −αt
R = De + µe I(t)eαt dt (51.27)
where I(t) is given by (51.25).

Taking into account the explicit expression for I(t), we find
Z
J = I(t)eαt dt
(51.28)
µB
= CJ1 − J2 − BJ3
A(α − 2µ)
where
Z
1
J1 = e(α−µ)t dt = e(α−µ)t (51.29)
α−µ
Z α − 2µ 2α − 3µ 1 −(α−µ)t µt
J2 = F
2 1 1, ; ; − e e dt (51.30)
α−µ α−µ A
eαt
Z
J3 = dt
1 + A e(α−µ)t
(51.31)
1 µ µ
= 2 F1 1, 1 + ;2 + ; −Ae(α−µ)t eαt
α α−µ α−µ
566
The integral J2 can be considered, without loss of generality, as a definite
integral from 0 to t, since t > 0. In other words we have
Z t
α − 2µ 2α − 3µ 1
J2 = 2 F1 1, ; ; − e−(α−µ)s eµs ds (51.32)
0 α−µ α−µ A
Under the change of variable
es = x, i.e. s = ln(x) (51.33)
the integral J2 takes on the form

ln(x)
α − 2µ 2α − 3µ
Z
µ−1
1 −(α−µ)
J2 = x 2 F1 1, ; ;− x dx
1 α−µ α−µ A
Z ln(x) α − 2µ 2α − 3µ 1
= xµ−1 2 F1 1, ; ; − x−(α−µ) dx (51.34)
0 α−µ α−µ A
Z 1 α − 2µ 2α − 3µ 1
− xµ−1 2 F1 1, ; ; − x−(α−µ) dx
0 α−µ α−µ A
In evaluating the above integrals, we may use the formulas
 
Z t a1 , · · ·, ap ;
xa−1 (t − x)b−1 p Fq  cxk  dx
0 b1 , · · ·, bq ;
= B(a, b)ta+b−1 ×
 (51.35)
a+k−1

a a+1
 a1 , · · ·, ap , k , k , · · ·, k
; 
×p+k Fq+k  ctk
 

 a+b a+b+1 a+b+k−1 
b1 , · · ·, bq , , , , · · ·, ;
k k k
with a > 0, b > 0 and
 
Z 1 a2 , · · ·, ap ;
ta1 −1 (1 − t)b1 −a1 −1 p−1 Fq−1  zt  dt
0 b2 , · · ·, bq ;
  (51.36)
a1 , a2 , · · ·, ap ;
Γ(a1 )Γ(b1 − a1 )
= p Fq
 z 
Γ(b1 )
b1 , b2 , · · ·, bq ;
567
with p≤q + 1 and b1 > a1 > 0. These formulas have been taken from
E. D. Rainville: ”Special Functions”, Chelsea 1960.
Second Method. Since equation (51.7) does not contain the independent
variable t explicitly, we introduce the substitution
dS d2 S dP
P = , 2
=P (51.37)
dt dt dS
dP 2
− P =α−µ (51.38)
dS S
This is a linear first order differential equation with unknown function P =
P (S) with general solution given by
P = C·S 2 − (α − µ)S (51.39)
Going back, we can determine the original function S by solving the equation
dS
= C·S 2 − (α − µ)S (51.40)
dt
The above is a Riccati-type of equation which can be solved using the same
methodology as in the First Method.
Third Method. In this method, we shall use Lie point symmetries.
We introduce a more traditional notation y = y(t), in place of S = S(t). We
thus examine the Lie point symmetries of the equation
yy 00 − (α − µ)yy 0 − 2(y 0 )2 = 0, y = y(t) (51.41)
The second prolongation P r(2) Γ in our case is given by (T = T (t, y) and

Y = Y (t, y))
∂ ∂ ∂ ∂
P r(2) Γ = T +Y + Y [y] 0 + Y [yy] 00 (51.42)
∂t ∂y ∂y ∂y
Invariance under Lie point symmetries is being expressed by the condition
P r(2) Γ(∆) =0 (51.43)

∆=0
568
where ∆ is given by
∆ = yy 00 − (α − µ)yy 0 − 2(y 0 )2 (51.44)
We have further
P r(2) Γ(∆)
∂ ∂ ∂ ∂
= T +Y + Y [t] 0 + Y [tt] 00 yy 00 − (α − µ)yy 0 − 2(y 0 )2
∂t ∂y ∂y ∂y

= Y y 00 − (α − µ)y 0 − Y [t] (α − µ)y + 4y 0 + yY [tt]
We have the following expressions for the coefficients Y [t] and Y [tt] :
Y [t] = Yt + (Yy − Tt )y 0 − Ty (y 0 )2 (51.45)
and
Y [tt] = Ytt + (2Yty − Ttt )y 0 + (Yyy − 2Tty )(y 0 )2 − Tyy (y 0 )3
(51.46)
+ (Yy − 2Tt )y 00 − 3Ty y 0 y 00
respectively. We thus have that
P r(2) Γ(∆)

= Y y 00 − (α − µ)y 0

− (α − µ)y + 4y 0 Yt + (Yy − Tt )y 0 − Ty (y 0 )2
(51.47)
+ y Ytt + (2Yty − Ttt )y 0 + (Yyy − 2Tty )(y 0 )2 − Tyy (y 0 )3

00 0 00
+ (Yy − 2Tt )y − 3Ty y y
In the above equation we make the substitution
y 00 −→(α − µ)y 0 + 2y −1 (y 0 )2 (51.48)
569
to account for the condition ∆ = 0. We then obtain that the invariance
condition P r(2) Γ(∆) = 0 is equivalent to
∆=0

−1 0 2
Y 2y (y )

− (α − µ)y + 4y 0 Yt + (Yy − Tt )y 0 − Ty (y 0 )2

+ y Ytt + (2Yty − Ttt )y 0 + (Yyy − 2Tty )(y 0 )2 − Tyy (y 0 )3 (51.49)
+ (Yy − 2Tt )[(α − µ)y 0 + 2y −1 (y 0 )2 ]

0 0 −1 0 2
− 3Ty y [(α − µ)y + 2y (y ) ] = 0
We have next to equate to zero the coefficients of the derivatives of the

function y.
Equating the coefficient of (y 0 )3 to zero, we obtain the equation
yTyy + 2Ty = 0 (51.50)

B(t)
T (t, y) = A(t) + (51.51)
y
Equating the coefficient of (y 0 )2 we get the equation
2 2Y
Yyy − Yy + 2 = 2(α − µ)Ty + 2Tty (51.52)
y y
Equating to zero the coefficient of y 0 we obtain the equation
2 α−µ 1
Yty − Yt = Tt + Ttt (51.53)
y 2 2
Equating to zero the constant term (i.e. the term which does not contain
derivative of y) we get the equation
Ytt − (α − µ)Yt = 0 (51.54)
Y (t, y) = F (y) + G(y)e(α−µ)t (51.55)
570
Equation (51.52), taking into account the expressions of Y and T , takes on
the form
2{B 0 (t) + (α − µ)B(t)} + {y 2 F 00 (y) − 2yF 0 (y) + 2F (y)}
(51.56)
+ {y 2 G00 (y) − 2yG0 (y) + 2G(y)}e(α−µ)t = 0
Equation (51.53), taking into account the expressions for T and Y given by
(51.51) and (51.55) respectively, becomes
2 [ yG0 (y) − 2G(y)] e(α−µ)t
1 1 (51.57)
= y A0 (t) + A00 (t) + B 0 (t) + B 00 (t)
α−µ α−µ
The general solution of the system of equations (51.56) and (51.57) is given
by
F (y) = a2 y 2 + a1 y − a7 (α − µ)
G(y) = a5 y 2 + a6 y

1 (α−µ)t −(α−µ)t (51.58)
A(t) = − a6 e + a3 e + a4
α−µ
B(t) = a7 + a8 e−(α−µ)t
Therefore we get the following expressions

1 (α−µ)t −(α−µ)t
T (t, y) = − a6 e + a3 e + a4
α−µ
(51.59)
1
+ a7 + a8 e−(α−µ)t
y
and
Y = a2 y 2 + a1 y − a7 (α − µ) + (a5 y 2 + a6 y)e(α−µ)t (51.60)
The generator of the Lie symmetries is then given by
∂ ∂
Γ=T +Y
∂t ∂y

1 (α−µ)t −(α−µ)t ∂
= − a6 e + a3 e + a4
α−µ ∂t
(51.61)
1 ∂
+ a7 + a8 e−(α−µ)t
y ∂t

2 2 (α−µ)t ∂
+ a2 y + a1 y − a7 (α − µ) + (a5 y + a6 y)e
∂y
571
∂ ∂ e−(α−µ)t ∂ ∂
Γ = a1 y + a2 y 2 − a3 + a4
∂y ∂y α − µ ∂t ∂t

2 (α−µ)t ∂ (α−µ)t 1 ∂ ∂
+ a5 y e + a6 e − +y (51.62)
∂y α − µ ∂t ∂y
1 ∂ −(α−µ)t
∂ e ∂
+ a7 − (α − µ) + a8
y ∂t ∂y y ∂t
We thus see that the generators of the algebra of the Lie point symmetries
of the considered equation are the following
∂ ∂ e−(α−µ)t ∂ ∂
Γ(1) = y , Γ(2) = y 2 Γ(3) = −
, , Γ(4) =
∂y ∂y α − µ ∂t ∂t

∂ 1 ∂ ∂
Γ(5) = y 2 e(α−µ)t , Γ(6) = e(α−µ)t − +y (51.63)
∂y α − µ ∂t ∂y
1∂ ∂ e−(α−µ)t ∂
Γ(7) = − (α − µ) , Γ(8) =
y ∂t ∂y y ∂t
Case II. We analyse the Lie point symmetries of the system of (51.6) and
the last two equations appearing in (51.2). We introduce a more traditional
notation under the following changes
S→u, I→v, R→w (51.64)
which results in the following system of equations

1 α + β α − β 2
utt = (α − µ)ut + + ut
β u v
(51.65)
vt = −ut − µv
wt = µv − αw
The above system will be solved using Lie point symmetries. We define the
three quantities ∆1 , ∆2 and ∆3 by
1 α + β α − β 2
∆1 = utt − (α − µ)ut − + ut
β u v
(51.66)
∆2 = vt + ut + µv
∆3 = wt − µv + αw
572
Invariance of the system of equations (51.65) is being expressed by the con-
ditions
P r(2) Γ(∆1 ) =0
∆1 =0, ∆2 =0, ∆3 =0
P r(2) Γ(∆2 ) =0 (51.67)

∆1 =0, ∆2 =0, ∆3 =0
P r(2) Γ(∆3 ) =0
∆1 =0, ∆2 =0, ∆3 =0
where P r(2) Γ is the second prolongation defined by

∂ ∂ ∂ ∂
P r(2) Γ = T +U +V +W
∂t ∂u ∂v ∂w (51.68)
[t] ∂ [t] ∂ [t] ∂ ∂
+U +V +W + U [tt]
∂ut ∂vt ∂wt ∂utt
We have the following expressions for the coefficients U [t] , V [t] , W [t] and U [tt] .
U [t] = Ut + (Uu − Tt )ut − Tu u2t

(51.69)
+ Uv vt + Uw wt − Tv vt ut − Tw wt ut
V [t] = Vt + (Vv − Tt )vt − Tv vt2

(51.70)
+ Vu ut + Vw wt − Tu vt ut − Tw wt vt
W [t] = Wt + (Ww − Tt )wt − Tw wt2
(51.71)
+ Wu ut + Wv vt − Tu wt ut − Tv wt vt
and
U [tt] = Utt + (Uu − 2Tt − 3Tu ut − 2Tv vt − 2Tw wt )utt
+ (Uuu − 2Ttu )u2t − 2(Tuv vt + Tuw wt )u2t − Tuu u3t
+ (2Utu − Ttt )ut − (Tw wtt + Tv vtt )ut
(51.72)
− 2(Ttw wt + Ttv vt )ut − (Tvv vt2 + Tww wt2 )ut
+ 2(Uuv vt + Uuw wt − Tvw vt wt )ut + 2Uvw vt wt
+ 2Utv vt + 2Utw wt + Uvv vt2 + Uww wt2
573
We have further
P r(2) Γ(∆1 ) = 0
∂ ∂ ∂ ∂ ∂
⇐⇒ T + U +V +W + U [t]
∂t ∂u ∂v ∂w ∂ut
∂ ∂ ∂
+ V [t] + W [t] + U [tt] utt − (α − µ)ut
∂vt ∂wt ∂utt
1 α + β α − β 2 (51.73)
− + ut = 0
β u v
α+β 2 α−β
⇐⇒ U ut + V u2t
βu2 βv 2

[t] 2 α + β α − β
−U (α − µ) + + ut + U [tt] = 0
β u v
Using the expressions for U [t] and U [tt] we obtain from the above equation
1 α + β α−β 2
U+ V ut
β u2 v2

− Ut + (Uu − Tt )ut − Tu u2t + Uv vt

+ Uw wt − Tv vt ut − Tw wt ut ×

2 α + β α − β
× (α − µ) + + ut
β u v (51.74)
+ Utt + (Uu − 2Tt − 3Tu ut − 2Tv vt − 2Tw wt )utt
+ (Uuu − 2Ttu )u2t − 2(Tuv vt + Tuw wt )u2t − Tuu u3t
+ (2Utu − Ttt )ut − (Tw wtt + Tv vtt )ut
− 2(Ttw wt + Ttv vt )ut − (Tvv vt2 + Tww wt2 )ut
+ 2(Uuv vt + Uuw wt − Tvw vt wt )ut + 2Uvw vt wt
+ 2Utv vt + 2Utw wt + Uvv vt2 + Uww wt2 = 0
In the above equation we make the following substitutions

1 α + β α − β 2
utt −→(α − µ)ut + + ut
β u v
(51.75)
vt −→ − ut − µv
wt −→µv − αw
574
to account for the conditions ∆1 = 0, ∆2 = 0, ∆3 = 0. We thus obtain
1 α + β α−β 2
U + V ut
β u2 v2

− Ut + (Uu − Tt )ut − Tu u2t − (Uv − Tv ut )(ut + µv)

+ (Uw − Tw ut )(µv − αw) ×

2 α + β α − β
× (α − µ) + + ut
β u v
+ Utt + (Uu − 2Tt − 3Tu ut + 2Tv (ut + µv) − 2Tw (µv − αw))
1 α + β α − β 2 (51.76)
× (α − µ)ut + + ut
β u v
+ (Uuu − 2Ttu )u2t − 2(−Tuv (ut + µv) + Tuw (µv − αw))u2t
− Tuu u3t + (2Utu − Ttt )ut − (Tw wtt + Tv vtt )ut
− 2(Ttw (µv − αw) − Ttv (ut + µv))ut − (Tvv (ut + µv)2
+ Tww (µv − αw)2 )ut − 2(Uuv (ut + µv) + Uuw (µv − αw)
+ Tvw (ut + µv)(µv − αw))ut − 2Uvw (ut + µv)(µv − αw)
− 2Utv (ut + µv) + 2Utw (µv − αw) + Uvv (ut + µv)2
+ Uww (µv − αw)2 = 0
We have now to equate to zero the coefficients of the partial derivatives of

the functions u, v, w of the above equation. However we observe that the
coefficients of ut wtt and ut vtt are −Tw and −Tv respectively and since Tw = 0
and Tv = 0, we see that
T = T (t, u) (51.77)
575
1 α + β α−β 2
U + V ut
β u2 v2

− Ut + (Uu − Tt )ut − Tu u2t − Uv (ut + µv)
2 α + β α − β

+ Uw (µv − αw) (α − µ) + + ut
β u v
+ Utt + (Uu − 2Tt − 3Tu ut )
(51.78)
1 α + β α − β 2
× (α − µ)ut + + ut
β u v
+ (Uuu − 2Ttu )u2t − Tuu u3t + (2Utu − Ttt )ut
− 2(Uuv (ut + µv) + Uuw (µv − αw))ut
− 2Uvw (ut + µv)(µv − αw) − 2Utv (ut + µv)
+ 2Utw (µv − αw) + Uvv (ut + µv)2 + Uww (µv − αw)2 = 0
Equating further the coefficient of u3t to zero, we obtain the equation
1 α + β α − β
+ Tu + Tuu = 0 (51.79)
β u v
It is obvious that the solution of the previous equation will contain the vari-
able v which contradicts to (51.77). However, once we assume that α = β
then the solution of (51.79) is given by
1
T = F (t) + G(t) , α=β (51.80)
u
Assuming that α6=β then we should take Tu = 0 which means that
T = T (t), α6=β (51.81)
576
We shall consider the case α6=β first. In this case equation (51.78) becomes
1 α + β α−β 2
U + V ut
β u2 v2

− Ut + (Uu − Tt − Uv )ut − µvUv + (µv − αw)Uw ×

2 α + β α − β
× (α − µ) + + ut
β u v
+ Utt + (Uu − 2Tt )
(51.82)
1 α + β α − β 2
× (α − µ)ut + + ut
β u v
+ Uuu u2t + (2Utu − Ttt )ut
− 2(Uuv (ut + µv) + Uuw (µv − αw))ut
− 2Uvw (ut + µv)(µv − αw) − 2Utv (ut + µv)
+ 2Utw (µv − αw) + Uvv (ut + µv)2 + Uww (µv − αw)2 = 0
We consider next the P r(2) Γ(∆2 ). We have in this case
P r(2) Γ(∆2 ) = 0
∂ ∂ ∂ ∂ ∂
⇐⇒ T + U +V +W + U [t]
(51.83)
[t] ∂ [t] ∂ [tt] ∂
+V +W +U vt + ut + µv = 0
∂vt ∂wt ∂utt
⇐⇒µV + U [t] + V [t] = 0
Taking into account the explicit expressions of U [t] and V [t] , the previous
equation becomes
µV + Ut + (Uu − Tt )ut − Tu u2t

+ Uv vt + Uw wt − Tv vt ut − Tw wt ut
(51.84)
+ Vt + (Vv − Tt )vt − Tv vt2
+ Vu ut + Vw wt − Tu vt ut − Tw wt vt = 0
577
In the above equation we make the substitutions (51.75) to account for the
conditions ∆1 = 0, ∆2 = 0, ∆3 = 0. We thus obtain the equation
µV + Ut + (Uu − Tt )ut − Tu u2t
− (Uv − Tv ut )(ut + µv) + (Uw − Tw ut )(µv − αw)
(51.85)
+ Vt − (Vv − Tt )(ut + µv) − Tv (ut + µv)2 + Vu ut
+ Vw (µv − αw) + Tu (ut + µv)ut + Tw (µv − αw)(ut + µv) = 0
Taking into account that Tu = Tw = Tv = 0, the above equation gives
µV + Ut + (Uu − Tt )ut − Uv (ut + µv) + Uw (µv − αw)
(51.86)
+ Vt − (Vv − Tt )(ut + µv) + Vu ut + Vw (µv − αw) = 0
From the above equation, equating to zero the coefficient of ut and the con-
stant term (the terms which do not involve derivatives) we get the equations
(Uu + Vu ) − (Uv + Vv ) = 0 (51.87)
and
µV + Ut − µvUv + Uw (µv − αw)
(51.88)
+ Vt − µv(Vv − Tt ) + Vw (µv − αw) = 0
respectively.
Equation (Uu + Vu ) − (Uv + Vv ) = 0 admits the general solution
U + V = F (t, u + v, w) (51.89)
We consider finally the P r(2) Γ(∆3 ). We have in this case
P r(2) Γ(∆3 ) = 0
∂ ∂ ∂ ∂ ∂
⇐⇒ T + U +V +W + U [t]
(51.90)
∂ ∂ ∂
+ V [t] + W [t] + U [tt] wt − µv + αw = 0
∂vt ∂wt ∂utt
[t]
⇐⇒ − µV + αW + W = 0
Using the explicit expression for W [t] , the previous equation becomes
− µV + αW + Wt + (Ww − Tt )wt − Tw wt2
(51.91)
+ Wu ut + Wv vt − Tu wt ut − Tv wt vt = 0
578
In the above equation we make the substitutions (51.75) to account for the
conditions ∆1 = 0, ∆2 = 0, ∆3 = 0. We thus obtain the equation
− µV + αW + Wt + (Ww − Tt )(µv − αw) − Tw (µv − αw)2

+ Wu ut − Wv (ut + µv) − Tu (µv − αw)ut (51.92)
+ Tv (µv − αw)(ut + µv) = 0
Since Tu = Tv = Tw = 0, the above equation gives
− µV + αW + Wt + (Ww − Tt )(µv − αw)

(51.93)
+ Wu ut − Wv (ut + µv) = 0
From the above equation, equating to zero the coefficient of ut and the con-
stant term (the terms which do not involve derivatives) we get the equations
Wu − Wv = 0 (51.94)
and
−µV + αW + Wt + (Ww − Tt )(µv − αw) − µvWv = 0 (51.95)
respectively.
Equation Wu − Wv = 0 admits the general solution
W = G(t, u + v, w) (51.96)
The above system of determining equations admits the solution
T = a1 , U = a2 u, V = a2 v, W = a2 w + a3 e−αt (51.97)
References
[1] C. M. Nucci: ”Using Lie Symmetries in Epidemiology”.
Electr. J. Diff. Eqns. Conference 12 (2005) 87-101.
[2] V. Torrisi and C. M. Nucci: ”Applications of Lie Group Analysis to
a Mathematical Model which describes HIV transmission”,
in ”The Geometrical Study of Differential Equations”, J. A. Leslie and T. P.
Hobart (eds), AMS, Providence (2001), 31-40
579
52 HIV Epidemiology
In this section we consider the model
du1 u1 u2
= −βc − µu1
dt u1 + u2 + u3
du2 u1 u2
= βc − (µ + ν)u2 (52.1)
dt u1 + u2 + u3
du3
= νu2 − αu3
dt
connected to the HIV epidemiology.
We shall try to find a close-form solution to the above system. We solve the
first of (52.1) with respect to u3 :
u1 u2
u3 = −(u1 + u2 ) − βc (52.2)
du1
+ µu1
dt
and differentiate with respect to t:
du3 du
1 du2
=− +
dt dt dt
du du1 du1 d2 u du1
2 1
u1 + u2 + µu1 − u1 u2 + µ (52.3)
− βc dt dt dt dt2 dt
du 2
1
+ µu1
dt
du 2
1
Multiplying the above equation by + µu1 , we get
dt
du3 du1 2 du
1 du2 du1 2
+ µu1 = − + + µu1
dt dt dt dt dt
du d2 u (52.4)
2 du1 du1

1 du1
− βc u1 + u2 + µu1 − u1 u2 +µ
dt dt dt dt2 dt
In the above equation we make the following substitutions

du3 βcu1 u2
−→ νu2 − αu3 −→ νu2 + α (u1 + u2 ) + (52.5)
dt du1
+ µu1
dt
580
du1 du2
+ −→ − µu1 − (µ + ν)u2 (52.6)
dt dt
du2 du1
−→ − − µu1 − (µ + ν)u2 (52.7)
dt dt

βcu1 u2 du1 2
νu2 + α (u1 + u2 ) + + µu1
du1 dt
+ µu1
dt
du 2
1
= µu1 + (µ + ν)u2 + µu1
dt (52.8)
du du1 du1
1
− βc −u1 + µu1 + (µ + ν)u2 + u2 + µu1
dt dt dt
d2 u du1

1
− u1 u2 +µ
dt2 dt
du 2 du
1 1
(α − µ)(u1 + u2 ) + µu1 + αβc u1 u2 + µu1
dt dt
d2 u 1
du 2
1 du1
= βc u1 u2 2 + (u1 − u2 ) + 2µu21 (52.9)
dt dt dt

du1 2 3 2
+ (µ + ν)u1 u2 + µ u1 + µ(µ + ν)u1 u2
dt
In the above equation we divide by βc and then we solve with respect to
d2 u1
. We then obtain the equation
dt2
d2 u1

α − µ u1 + u2 u1 − u2 du1 2
= · −
dt2 βc u1 u2 u1 u2 dt

2µ(α − µ) u1 + u2 u1 du1
+ · − 2µ + (α − µ − ν) (52.10)
βc u2 u2 dt
µ2 (α − µ) u1 u1
+ (u1 + u2 )· − µ [µ(u1 + u2 ) + νu2 ] + αµu1
βc u2 u2
The obvious choice α − µ = βc, i.e. α = µ + βc reduces the equation into
d2 u1 2 du1 2 du1
2
= + (2µ + βc − ν) + µ(µ − ν + βc)u1 (52.11)
dt u1 dt dt
581
which is a second order nonlinear ordinary differential equation with respect
to u1 . We thus see that we have succeeded in decoupling the system of
equations (52.1). Equation (52.11) can be solved by three methods, as we
have done in previous section.
Adding the first two equations of system (52.1), we obtain the equation
d
(u1 + u2 ) = −(µ + ν)(u1 + u2 ) − νu2 (52.12)
dt
and introducing the new function U by
U = u1 + u2 (52.13)
we get the equation

dU
= −(µ + ν)U + νu1 (52.14)
dt
The above equation, when solved with respect to U , the function u2 can be
determined from u2 = U − u1 . The function u3 can then be determined
solving the third equation of (52.1) which is a linear first order ODE. We
thus have derived the following Theorem:
Theorem. The system of equations (52.1) is equivalent to the system
d2 u1 2 du1 2 du1
2
= + (2µ + βc − ν) + µ(µ − ν + βc)u1
dt u1 dt dt
dU (52.15)
= −(µ + ν)U + νu1 , U = u1 + u2
dt
du3
= −αu3 + ν(U − u1 )
dt
The above is a decoupled system of ODEs, which was derived under the
assumption α = µ + βc.
Case II. We analyse the Lie point symmetries of the system of (52.10) and
the last two equations appearing in (52.1). We introduce a more traditional
notation under the following changes
u1 →u, u2 →v, u3 →w (52.16)
582
which results in the following system of equations

α−µ u+v u−v
utt = · − (ut )2
βc uv uv

2µ(α − µ) u + v u
+ · − 2µ + (α − µ − ν) ut
βc v v
2 (52.17)
µ (α − µ) u u
+ (u + v)· − µ [µ(u + v) + νv] + αµu
βc v v
vt = −ut − µu − (µ + ν)v
wt = νv − αw
The above system will be solved using Lie point symmetries. We define the
three quantities ∆1 , ∆2 and ∆3 by

α−µ u+v u−v
∆1 = utt − · − (ut )2
βc uv uv

2µ(α − µ) u + v u
− · − 2µ + (α − µ − ν) ut
βc v v
(52.18)
µ2 (α − µ) u u
− (u + v)· + µ [µ(u + v) + νv] − αµu
βc v v
∆2 = vt + ut + µu + (µ + ν)v
∆3 = wt − νv + αw
Invariance of the system of equations (52.17) is being expressed by the con-
ditions
P r(2) Γ(∆1 ) =0
∆1 =0, ∆2 =0, ∆3 =0
P r(2) Γ(∆2 ) =0 (52.19)

∆1 =0, ∆2 =0, ∆3 =0
P r(2) Γ(∆3 ) =0
∆1 =0, ∆2 =0, ∆3 =0
where P r(2) Γ is the second prolongation defined by

∂ ∂ ∂ ∂
P r(2) Γ = T +U +V +W
∂t ∂u ∂v ∂w (52.20)
∂ ∂ ∂ ∂
+ U [t] + V [t] + W [t] + U [tt]
∂ut ∂vt ∂wt ∂utt
583
The expressions for the coefficients U [t] , V [t] , W [t] and U [tt] are known from
the previous section. We may then proceed along the same line of reasoning
as in the previous Section.
Chapter XIX
Some Special Equations
53 The Thomas Equation
The Thomas Equation. In this Example we shall find the Lie point sym-
metries of the Thomas equation
uxt + αut + βux + γut ux = 0, u = u(x, t) (53.1)
P r(2) Γ(∆)|∆=0 = 0 (53.2)
where
∆≡uxt + αut + βux + γut ux (53.3)

∂ ∂ ∂
P r(2) Γ = U [t] + U [x] + U [xt] (53.4)
∂ut ∂ux ∂uxx
We thus have
P r(2) Γ(∆) = 0
[t] ∂ [x] ∂ [xt] ∂

⇐⇒ U +U +U (∆) = 0 (53.5)
∂ut ∂ux ∂uxx
⇐⇒ (α + γux )U [t] + (β + γut )U [x] + U [xt] = 0
584
Using the known expressions for the coefficients U [t] , U [x] and U [xt] , the pre-
vious equation takes on the form

2
(α + γux ) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2
+ (β + γut ) Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut
+ Uxt + (Uxu − Txt )ut + (Utu − Xxt )ux (53.6)

− (Xu ux + Tu ut )uxt − Xt uxx − Tx utt
− Txu (ut )2 − Xtu (ux )2 − Xu ut uxx = 0
We introduce next the following substitution
uxt −→ − (αut + βux + γut ux ) (53.7)
to account for the condition ∆ = 0. We then obtain the equation

2
(α + γux ) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2
+ (β + γut ) Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut
+ Uxt + (Uxu − Txt )ut + (Utu − Xxt )ux

(53.8)
+ (Uuu − Ttu − Xxu )ux ut
− (Uu − Xx − Tt )(αut + βux + γut ux )
+ (Xu ux + Tu ut )(αut + βux + γut ux )
− Xt uxx − Tx utt − Xuu ut (ux )2 − Tuu ux (ut )2 − Tu ux utt
− Txu (ut )2 − Xtu (ux )2 − Xu ut uxx = 0
We now have to equate to zero the coefficients of the partial derivatives of the
function u. However instead of expanding the above equation, we can easily
identify some coefficients by a single glance. Equating to zero the coefficients
of uxx , utt , ux utt and ut uxx , we find
Xt = 0, Tx = 0, Tu = 0, Xu = 0 (53.9)
585
respectively. We thus have
T = T (t) and X = X(x) (53.10)

(α + γux ) Ut + (Uu − Tt )ut + (β + γut ) Ux + (Uu − Xx )ux
(53.11)
+ Uxt + Uxu ut + Utu ux + Uuu ux ut
− (Uu − Xx − Tt )(αut + βux + γut ux ) = 0
From the above equation we get the following coefficients, which are equated
to zero:
Coefficient of ux ut
γUu + Uuu = 0 (53.12)
Coefficient of ut
γUx + Uxu + αXx = 0 (53.13)
Coefficient of ux
γUt + Utu + βTt = 0 (53.14)
αUt + Uxt + βUx = 0 (53.15)
The general solution of (53.12) is given by
U = A(x, t) + B(x, t)e−γu (53.16)
Equations (53.13), (53.14) and (53.15), because of the above expression for
U , take the form
γAx + αXx = 0 (53.17)
γAt + βTt = 0 (53.18)

and
(αAt + Axt + βAx ) + (αBt + Bxt + βBx )e−γu = 0 (53.19)
586
respectively.
Given that X = X(x) and T = T (t), upon differentiation of (53.17) and
(53.18) with respect to t and x respectively, we obtain that
Axt = 0 (53.20)
On the other hand, equating to zero the expressions within the parentheses
in (53.19), we get
αAt + βAx = 0, αBt + Bxt + βBx = 0 (53.21)
where we have taken into account Axt = 0. The system of PDEs
αAt + βAx = 0, Axt = 0 (53.22)
admits the general solution

α
A(x, t) = a1 t − x + a2 (53.23)
β
Let us now determine X(x) and T (t). From (53.17), using (53.23), we find
γ γ
Xx = − Ax = a1
α β
from which by integration with respect to x we find
γ
X = a1 x + a3 (53.24)
β
From (53.18), using (53.23), we find
γ γ
Tt = − At = − a1
β β
from which by integration with respect to t we get
γ
T = −a1 t + a4 (53.25)
β
α
U = a1 t − x + a2 + B(x, t)e−γu (53.26)
β
587
where B(x, t) satisfies the second equation of (53.21).
We thus see that the generator of the Lie point symmetries is given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
γ ∂ γ ∂
= −a1 t + a4 + a1 x + a3 (53.27)
β ∂t β ∂x

α −γu ∂
+ a1 t − x + a2 + B(x, t)e
β ∂u

γ ∂ γ ∂ α ∂
Γ = a1 − t + x + t− x
β ∂t β ∂x β ∂u
∂ ∂ ∂ (53.28)
+ a2 + a3 + a4
∂u ∂x ∂t
∂
+ B(x, t)e−γu
∂u
We thus see that the generators of the Lie point symmetries of Thomas
equation are
∂ ∂ ∂
Γ1 = −γt + γx + (βt − αx)
∂t ∂x ∂u
∂
Γ2 =
∂u (53.29)
∂
Γ3 =
∂x
∂
Γ4 =
∂t
and the infinite subalgebra generated by
∂
Γ∞ = B(x, t)e−γu (53.30)
∂u
where B(x, t) is any solution of the equation αBt + Bxt + βBx = 0.
Let us now consider the invariant surface condition T ut +Xux = U associated
to the generator Γ1 ,
−γtut + γxux = βt − αx (53.31)
588
α β
u = − x − t + F (xt) (53.32)
γ γ
where F (xt) is an arbitrary function. Upon substitution of u given by (53.32)
into the original equation (53.1) we obtain the equation
αβ
ξF 00 (ξ) + γξ(F 0 (ξ))2 + F 0 (ξ) − =0 (53.33)
γ
where
ξ = xt (53.34)
is a similarity variable. The substitution
F 0 (ξ) = G(ξ) (53.35)
1 αβ 1
G0 (ξ) = −γG2 (ξ) − G(ξ) + (53.36)
ξ γ ξ
The above is a Riccati equation which under the substitution
1 Y 0 (ξ)
G= (53.37)
γ Y (ξ)
is being transformed into the second order equation
ξY 00 (ξ) + Y 0 (ξ) − αβY (ξ) = 0 (53.38)
This is a Bessel equation of zero-th order admitting a general solution
p p
Y = A·J0 (2 −αβξ) + B·Y0 (2 −αβξ) (53.39)
where J0 and Y0 are zero-th order Bessel functions of the corresponding ar-
gument. See any elementary treatise on Bessel functions like
F. Bowman: ”Introduction to Bessel Functions”, Dover 1958.
Combining (53.35) and (53.37) and integrating, we obtain the following ex-
pression for the function F (C1 = B/A)

1 p p
F = ln J0 (2 −αβξ) + C1 ·Y0 (2 −αβξ) + C2 (53.40)
γ
589
Using (53.32) and (53.40) and going back to original variables, we obtain the
general solution of Thomas equation
α β
u(x, t) = − x − t
γ γ
(53.41)
1 p p
+ ln J0 (2 −αβxt) + C1 ·Y0 (2 −αβxt) + C2
γ
54 The Fokker-Planck Equation

The Fokker-Planck Equation. In this Example we shall find the Lie point
symmetries of the Fokker-Planck equation
ut = uxx + (xu)x , u = u(x, t) (54.1)
P r(2) Γ(∆)|∆=0 = 0 (54.2)
where
∆≡ut − uxx − xux − u (54.3)

∂ ∂ ∂ ∂ ∂
P r(2) Γ = X +U + U [t] + U [x] + U [xx] (54.4)
∂x ∂u ∂ut ∂ux ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂ ∂ ∂ ∂ ∂
⇐⇒ X +U + U [t] + U [x] + U [xx] (∆) = 0 (54.5)
∂x ∂u ∂ut ∂ux ∂uxx
⇐⇒ − Xux − U + U [t] − xU [x] − U [xx] = 0
590
Using the known expressions for the coefficients U [t] , U [x] and U [xx] , the
previous equation becomes
− Xux − U
+ Ut + (Uu − Tt )ut − Xt ux − Tu (ut )2 − Xu ut ux

2
− x Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut



In the above equation we substitute ut by uxx + xux + u to account for the

condition ∆ = 0. We thus obtain the equation
− Xux − U
+ Ut + (Uu − Tt − Xu ux )(uxx + xux + u) − Xt ux
− Tu (uxx + xux + u)2

2
− x Ux + (Uu − Xx )ux − (Tx + Tu ux )(uxx + xux + u) − Xu (ux )

(54.7)
− Uxx + (2Uxu − Xxx )ux − (Txx + 2Txu ux )(uxx + xux + u)
+ (Uuu − 2Xxu )(ux )2 − Tuu (uxx + xux + u)(ux )2 − Xuu (ux )3


− Tu (uxx + xux + u)uxx − 3Xu ux uxx = 0
We have next to equate to zero the coefficients of the partial derivatives of the
function u. We can however simplify the procedure considerably if we equate
to zero the easily identified coefficients of utx , ux utx and ux uxx . Equating to
zero the above coefficients, we find Tx = 0, Tu = 0 and Xu = 0. We thus
have
T = T (t) and X = X(t, x) (54.8)
591
Because of the above expressions, equation (54.7) takes on the form
− Xux − U
+ Ut + (Uu − Tt )(uxx + xux + u) − Xt ux

− x Ux + (Uu − Xx )ux (54.9)

2
From the above equation we obtain the following

2Xx − Tt = 0 (54.10)
Uuu = 0 (54.11)
Coefficient of ux :
−X + x(Xx − Tt ) − Xt − 2Uxu + Xxx = 0 (54.12)
−U + Ut + u(Uu − Tt ) − xUx − Uxx = 0 (54.13)
From equation (54.10), upon integration with respect to x, we get
1
X = T 0 (t)x + γ(t) (54.14)
2
From equation (54.11) upon integration with respect to u, we get
U = α(x, t)u + β(x, t) (54.15)
1 1 1
αx = − T 0 (t)x − T 00 (t)x − [γ(t) + γ 0 (t)] (54.16)
2 4 2
and upon integration with respect to x we obtain the following expression
for α(x, t):
1 1 1
α(x, t) = − T 0 (t)x2 − T 00 (t)x2 − [γ(t) + γ 0 (t)]x + δ(t) (54.17)
4 8 2
592
Equation (54.13), because of (54.14) and (54.15), takes on the form
{αt − xαx − αxx − T 0 (t)}u − {β − βt + xβx + βxx } = 0 (54.18)
From the above equation, equating to zero the coefficients of u, we obtain

the two equations
αt − xαx − αxx − T 0 (t) = 0 (54.19)
and
β − βt + xβx + βxx = 0 (54.20)

1 1 1 1
− {T 000 (t)−4T 0 (t)}x2 − {γ 00 (t)−γ(t)}x+ T 00 (t)− T 0 (t)+δ 0 (t) = 0 (54.21)
8 2 4 2
Equating to zero the coefficients of x, we get the system of ordinary differ-
ential equations
T 000 (t) − 4T 0 (t) = 0 (54.22)
γ 00 (t) − γ(t) = 0 (54.23)

1 00 1
T (t) − T 0 (t) + δ 0 (t) = 0 (54.24)
4 2
The above system of equations (54.22)-(54.24) can be easily solved having
solution
T (t) = a1 + a2 e2t + a3 e−2t (54.25)
γ(t) = a4 et + a5 e−t (54.26)

δ(t) = a3 e−2t + a6 (54.27)
From (54.14), because of the known expressions of T (t) and γ(t), we get
X = (a2 e2t − a3 e−2t )x + a4 et + a5 e−t (54.28)
From (54.17), because of the known expressions of T (t), γ(t) and δ(t), we get
α(x, t) = −a2 e2t x2 − a4 et x + a3 e−2t + a6 (54.29)
593
Using the above expression of α(x, t), we obtain from (54.15)
U = (−a2 e2t x2 − a4 et x + a3 e−2t + a6 )u + β(x, t) (54.30)
where β(x, t) satisfies equation (54.20).

The generator of the Lie symmetries of the Fokker-Planck equation is then
given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂
= (a1 + a2 e2t + a3 e−2t )
∂t (54.31)
2t −2t t −t ∂
+ {(a2 e − a3 e )x + a4 e + a5 e }
∂x
∂
+ {(−a2 e2t x2 − a4 et x + a3 e−2t + a6 )u + β(x, t)}
∂u
∂ ∂ ∂ ∂
Γ = a1 + a2 e2t + e2t x − e2t x2 u
∂t ∂t ∂x ∂u
∂ ∂ ∂
+ a3 e−2t − e−2t x + e−2t u
∂t ∂x ∂u (54.32)
∂ ∂ ∂ ∂
+ a4 e t − et xu + a5 e−t + a6 u
∂x ∂u ∂x ∂u
∂
+ β(x, t)
∂u
Therefore the Lie algebra of symmetries of the Fokker-Planck equation are
spanned by the six generators
∂ ∂ ∂ ∂
X1 = , X 2 = e2t + e2t x − e2t x2 u
∂u ∂t ∂x ∂u
∂ ∂ ∂ ∂ ∂
X 3 = e−2t − e−2t x + e−2t u , X 4 = et − et xu (54.33)
∂t ∂x ∂u ∂x ∂u
∂ ∂
X 5 = e−t , X 6 = u
∂x ∂u
There is also a sub-algebra of infinite dimension with generator
∂
X ∞ = β(x, t) (54.34)
∂u
594
The invariant surface condition associated to the generator X 3 has the form
ut − xux = u (54.35)

1
u= F (t + ln(x)) (54.36)
x
Taking into account the above solution and considering the substitution
1
u= G(ξ), ξ = x et (54.37)
x
the original Fokker-Planck equation (54.1) takes on the form
ξ 2 G00 (ξ) − 2ξG0 (ξ) + 2G(ξ) = 0 (54.38)
This is a Legendre-type of equation with general solution
G(ξ) = Aξ 2 + Bξ (54.39)
We thus have obtained a general solution of the Fokker-Planck equation
u = (Axet + B)et (54.40)
55 The Fisher Equation

The Fisher Equation. In this Example we shall find the Lie point sym-
metries of the Fisher equation
ut = uxx + u(1 − u), u = u(x, t) (55.1)
P r(2) Γ(∆)|∆=0 = 0 (55.2)
where
∆≡ut − uxx − u(1 − u) (55.3)
595
∂ ∂ ∂
P r(2) Γ = U + U [t] + U [xx] (55.4)
∂u ∂ut ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂ ∂ ∂
⇐⇒ U + U [t] + U [xx] (∆) = 0 (55.5)
∂u ∂ut ∂uxx
⇐⇒ (−1 + 2u)U + U [t] − U [xx] = 0
Using the known expressions for U [t] and U [xx] , the above equation takes on
the form

2
(−1 + 2u)U + Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

(55.6)

In the above equation we substitute ut by uxx + u(1 − u) to account for the

condition ∆ = 0. We then obtain the equation
(−1 + 2u)U + Ut

+ (Uu − Tt − Xu ux )(uxx + u(1 − u)) − Tu (uxx + u(1 − u))2

− Xt ux

(55.7)
− Uxx + (2Uxu − Xxx )ux − (Txx + 2Txu ux )(uxx + u(1 − u))
+ (Uuu − 2Xxu )(ux )2 − Tuu (uxx + u(1 − u))(ux )2


− Tu (uxx + u(1 − u))uxx − 3Xu ux uxx = 0
596
We have next to equate to zero the coefficients of the partial derivatives of the
function u. We can however simplify the procedure considerably if we equate
to zero the easily identified coefficients of utx , ux utx and ux uxx . Equating to
zero the above coefficients, we find Tx = 0, Tu = 0 and Xu = 0. We thus
have
T = T (t) and X = X(t, x) (55.8)
Because of the above expressions, equation (55.7) takes on the form
(−1 + 2u)U + Ut

+ (Uu − Tt )(uxx + u(1 − u)) − Xt ux
(55.9)

2

2Xx − Tt = 0 (55.10)
Uuu = 0 (55.11)
−Xt − 2Uxu + Xxx = 0 (55.12)
The coefficient of zeroth order derivative is
(−1 + 2u)U + Ut + u(1 − u)(Uu − Tt ) = 0 (55.13)
From equation (55.10), upon integration with respect to x, we obtain

1
X(x, t) = T 0 (t)x + γ(t) (55.14)
2
From equation (55.11) upon integration with respect to u we obtain
U = α(x, t)u + β(x, t) (55.15)
597
1 1
αx = − T 00 (t)x − γ 0 (t) (55.16)
4 2
and upon integration with respect to x we have the following expression for
α(x, t):
1 1
α(x, t) = − T 00 (t)x2 − γ 0 (t)x + δ(t) (55.17)
8 2
{α + T 0 (t)}u2 + {2β + at − T 0 (t)}u + βt − β = 0 (55.18)
Equating the coefficients of u to zero, we obtain the following system of

equations
α + T 0 (t) = 0 (55.19)
2β + at − T 0 (t) = 0 (55.20)
and
βt − β = 0 (55.21)

1 1
− T 00 (t)x2 − γ 0 (t)x + δ(t) + T 0 (t) = 0 (55.22)
8 2
from which, equating to zero the coefficients of x, we obtain the system
T 00 (t) = 0, γ 0 (t) = 0, δ(t) = −T 0 (t) (55.23)
and thus
T (t) = a1 t + a2 , γ(t) = a3 , δ(t) = −a1 (55.24)
We then see that
α(x, t) = −a1 (55.25)
598
Because of the above expression of α, equations (55.20) and (55.21) are com-
patible each other if a1 = 0 and the value of β is zero. Therefore
α(x, t) = 0, β(x, t) = 0, T (t) = a2 , γ(t) = a3 , δ(t) = 0 (55.26)
and then
T (t) = a2 , X(x, t) = a3 and U = 0 (55.27)
The generator of Lie symmetries is given by

∂ ∂ ∂ ∂ ∂
Γ=T +X +U = a2 + a3 (55.28)
∂t ∂x ∂u ∂t ∂x
We thus see that the Lie algebra of symmetries is being spanned by the two
generators
∂ ∂
Γ1 = , Γ2 = (55.29)
∂t ∂x
∂ ∂
Γ1 + cΓ2 = +c (55.30)
∂t ∂x
the associated invariant surface condition ut + cux = 0 admits the solution
x
u=U t− (55.31)
c
Therefore considering the substitution
u = F (ξ), ξ = k(x − ωt) (55.32)
the Fisher equation (55.1), taking into account that
∂u dF ∂u dF ∂ 2u 2
2d F
= −kω , =k , = k
∂t dξ ∂x dξ ∂x2 dξ 2
takes on the form
k 2 F 00 (ξ) + k ωF 0 (ξ) + F − F 2 = 0 (55.33)
599
The above equation can be solved using the Tanh method. Considering the
expansion (ak , bk are constants)
n
X n
X
F = k
ak Y + bk Y −k , Y = tanh(ξ) (55.34)
k=0 k=1
with
dY 1
Y0 = = = 1 − tanh2 (ξ) = 1 − Y 2 (55.35)
dξ cosh2 (ξ)
and balancing the terms F 00 and F 2 we get n + 2 = 2n, i.e. n = 2. We thus

have the expansion
b1 b2
F = a0 + a1 Y + a2 Y 2 + + 2 (55.36)
Y Y
We first have to find the derivatives F 0 and F 00 where at each stage of differ-
entiation we have to substitute Y 0 by 1 − Y 2 . We thus get from (55.36) by
differentiation with respect to ξ
b1 2b2 dY
F 0 (ξ) = a1 + 2a2 Y − 2 − 3
Y Y dξ
(55.37)
b1 2b2
2
= a1 + 2a2 Y − 2 − 3 (1 − Y )
Y Y
Differentiation of (55.37) with respect to ξ yields
2b1 6b2 dY
F 00 (ξ) = 2a2 + 3 + 4 (1 − Y 2 )
Y Y dξ
b1 2b2 dY
− 2 a1 + 2a2 Y − 2 − 3 Y
Y Y dξ (55.38)
2b1 6b2
= 2a2 + 3 + 4 (1 − Y 2 )2
Y Y
b1 2b2
− 2 a1 + 2a2 Y − 2 − 3 Y (1 − Y 2 )
Y Y
We can now substitute (55.36), (55.37) and (55.38) in (55.33) and after col-
lecting the similar powers of Y and equating the corresponding powers of Y
to zero, we obtain a set of simultaneous equations from which we can deter-
mine the various coefficients appearing in the expansion (55.36). We derive
600
an overdetermined system of nine equations. Using five of them, we find the
following values
1 1 6
a0 = − 4k 2 − ω 2 , a1 = − kω, a2 = 6k 2
2 50 5 (55.39)
2 2
ω(100k − ω )
b1 = , b2 = 6k 2
250k
The values of k and ω are not independent each other. Consistency of the
rest four equations requires that
√ √ √ √
6 5 6 6 5 6
k= , ω= , k= , ω=− ,
24√ 6√ 24 √ 6 √ (55.40)
6 5 6 6 5 6
k=− , ω= , k=− , ω=−
24 6 24 6
or
√ √ √ √
6 5 6 6 5 6
k=i , ω=i , k=i , ω = −i ,
24√ 6√ 24 √ 6 √ (55.41)
6 5 6 6 5 6
k = −i , ω=i , k = −i , ω = −i
24 6 24 6
Therefore the solution of Fisher’s equation is given by
1 1 2 6
F = − 4k − ω + − kω tanh(ξ) + (6k 2 )tanh2 (ξ)
2
2 50 5
(55.42)
ω(100k 2 − ω 2 ) 1 6k 2
+ +
250k tanh(ξ) tanh2 (ξ)
or
1 1 2 6
u(x, t) = − 4k 2 − ω + − kω tanh(k(x − ωt))
2 50 5
ω(100k 2 − ω 2 ) (55.43)
+ (6k 2 )tanh2 (k(x − ωt)) + coth(k(x − ωt))
250k
+ 6k 2 coth2 (k(x − ωt))
where the values of k and ω are given by (55.40) or (55.41).
601
56 The Fitzhugh-Nagumo Equation
The Fitzhugh-Nagumo Equation. In this Example we shall find the Lie
point symmetries of the Fitzhugh-Nagumo equation
ut = uxx + u(1 − u)(u − a), u = u(x, t) (56.1)
P r(2) Γ(∆)|∆=0 = 0 (56.2)
where
∆≡ut − uxx − u(1 − u)(u − a) (56.3)

∂ ∂ ∂
P r(2) Γ = U + U [t] + U [xx] (56.4)
∂u ∂ut ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂ ∂ ∂
⇐⇒ U + U [t] + U [xx] (∆) = 0 (56.5)
∂u ∂ut ∂uxx
⇐⇒ [ 3u2 − 2(1 + a)u + a] U + U [t] − U [xx] = 0
Using the known expressions of the coefficients U [t] and U [xx] , the above
equation becomes
[ 3u2 − 2(1 + a)u + a] U


(56.6)

602
In the above equation we substitute ut by uxx + u(1 − u)(u − a) to account
for the condition ∆ = 0. We then obtain the equation
[ 3u2 − 2(1 + a)u + a] U

+ Ut + (Uu − Tt − Xu ux )[uxx + u(1 − u)(u − a)]
− Xt ux − Tu [uxx + u(1 − u)(u − a)]2

− Uxx + (2Uxu − Xxx )ux − (Txx + 2Txu ux )[uxx + u(1 − u)(u − a)]
(56.7)
2 2
+ (Uuu − 2Xxu )(ux ) − Tuu [uxx + u(1 − u)(u − a)](ux )

− Tu [uxx + u(1 − u)(u − a)]uxx − 3Xu ux uxx = 0

equate the easily identified coefficients first. We thus see that the coefficients
of ux uxx , utx and ux utx are 2Xu + 2Txu , 2Tx and 2Tu respectively. Therefore
we should have that Tx = Tu = 0 and Xu = 0 and thus
X = X(x, t) and T = T (t) (56.8)
[ 3u2 − 2(1 + a)u + a] U

+ Ut + (Uu − Tt )[uxx + u(1 − u)(u − a)] − Xt ux

− Uxx + (2Uxu − Xxx )ux (56.9)

2
+ Uuu (ux ) + (Uu − 2Xx )uxx = 0

The coefficient of (ux )2 is equal to
Uuu = 0 (56.10)
The coefficient of uxx is equal to
2Xx − Tt = 0 (56.11)
603
The coefficient of ux is equal to
Xxx − Xt − 2Uxu = 0 (56.12)
The constant term (terms which are independent of partial derivatives of the
function u) is equal to
[ 3u2 − 2(1 + a)u + a ] U + Ut − Uxx

(56.13)
+ (Uu − Tt )[u(1 − u)(u − a)] = 0
Integrating (56.10) with respect to u we obtain
U = α(x, t)u + β(x, t) (56.14)
Integration of (56.11) with respect to x gives

1
X = xT 0 (t) + γ(t) (56.15)
2
1 1
αx = − xT 00 (t) − γ(t) (56.16)
4 2
and by integration with respect to x gives
1 1
α(x, t) = − x2 T 00 (t) − xγ(t) + δ(t) (56.17)
8 2
We thus see that the function U becomes

1 2 00 1
U = − x T (t) − xγ(t) + δ(t) u + β(x, t) (56.18)
8 2
Finally, equation (56.13), because of (56.18), takes on the form
[ 3u2 − 2(1 + a)u + a ]

1 2 00 1
× − x T (t) − xγ(t) + δ(t) u + β(x, t)
8 2
(56.19)

1 2 000 1 0 1
+ − x T (t) − xγ (t) + δ (t) u + βt + T 00 (t)u − βxx
0
8 2 4

1 1
+ − x2 T 00 (t) − xγ(t) + δ(t) − T 0 (t) [u(1 − u)(u − a)] = 0
8 2
604
After expanding the above equation in terms of u, and equating the coeffi-
cients of the different powers of u to zero, we derive a set of four equations:
Coefficient of u3
1
− x2 T 00 (t) − xγ(t) + T 0 (t) + 2δ(t) = 0 (56.20)
4
Coefficient of u2
1 1
(a + 1)x2 T 00 (t) + (a + 1)xγ(t) − (a + 1)T 0 (t)
8 2 (56.21)
− (a + 1)δ(t) + 3β(x, t) = 0
Coefficient of u
1 1 1
− x2 T 000 (t) − xγ 0 (t) + T 00 (t) + aT 0 (t) + δ 0 (t)
8 2 4 (56.22)
− 2(a + 1)β(x, t) = 0
Coefficient of zeroth order of u
aβ(x, t) + βt − βxx = 0 (56.23)
The solution of the system of the determining equations yields
X = a1 , T = a2 , U =0 (56.24)
Therefore we obtain the generator of the Lie algebra of symmetries

∂ ∂
Γ = a1 + a2 (56.25)
∂x ∂t
We thus have the two generators
∂ ∂
Γ1 = , Γ2 = (56.26)
∂x ∂t
∂ ∂
Γ2 + cΓ1 = +c (56.27)
∂t ∂x
the associated invariant surface condition ut + cux = 0 admits the solution
x
u=U t− (56.28)
c
605
Therefore considering the substitution
u = F (ξ), ξ = k(x − ωt) (56.29)
the F-N equation (56.1), taking into account that
∂u dF ∂u dF ∂ 2u 2
2d F
= −kω , =k , = k
∂t dξ ∂x dξ ∂x2 dξ 2
takes on the form
k 2 F 00 (ξ) + kωF 0 (ξ) + F (1 − F )(F − a) = 0 (56.30)
The above equation can be solved using the Tanh method. Considering the
expansion (ak , bk are constants)
n
X n
X
F = k
ak Y + bk Y −k , Y = tanh(ξ) (56.31)
k=0 k=1
with
dY 1
Y0 = = 2
= 1 − tanh2 (ξ) = 1 − Y 2 (56.32)
dξ cosh (ξ)
and balancing the terms F 00 and F 3 we get n + 2 = 3n, i.e. n = 1. We thus

have the expansion
b1
F = a0 + a1 Y + (56.33)
Y
We first have to find the derivatives F 0 and F 00 where at each stage of differ-
entiation we have to substitute Y 0 by 1 − Y 2 . We thus get from (56.33) by
differentiation with respect to ξ
0
b1 dY b1
F (ξ) = a1 − 2 = a1 − 2 (1 − Y 2 ) (56.34)
Y dξ Y
Differentiation of (56.34) with respect to ξ yields
2b1 dY b1 dY
F 00 (ξ) = 3 (1 − Y 2 ) − 2 a1 − 2 Y
Y dξ Y dξ
(56.35)
2b1 b1
= 3 (1 − Y 2 )2 − 2 a1 − 2 Y (1 − Y 2 )
Y Y
606
We can now substitute (56.33), (56.34) and (56.35) into (56.30) and after
collecting the similar powers of Y and equating the corresponding powers
of Y to zero, we obtain a set of simultaneous equations from which we can
determine the various coefficients appearing in the expansion (56.33). We
derive an overdetermined system of seven equations. Using three of them,
we find the following values
√
1 2 1 √ √
a0 = − ω + a, a1 = ± 2 k, b1 = ± 2 k (56.36)
3 6 3
The values of k and ω are not independent √ each other.
√ Consistency of the
rest four equations for the choice a1 = 2 k, b1 = 2 k requires that
√ √ √ √
2 2 2 2
k= a, ω = − (a − 2) , k = − a, ω = − (a − 2) ,
√8 √2 8√ √2
2 2 2 2
k= a, ω = (−1 + 2a) , k = − a, ω = (−1 + 2a) ,
√8 2 √ 8 2 (56.37)
2 2
k= (a − 1), ω = − (1 + a) ,
8√ 2√
2 2
k=− (a − 1), ω = − (1 + a)
8 2
Therefore the solution of F-N equation (56.1) is given by
1 √2 1 √ √ 1
F = − ω + a + 2k tanh(ξ) + 2k (56.38)
3 6 3 tanh(ξ)
or
√
1 2 1 √
u(x, t) = − ω + a + 2k tanh(k(x − ωt)
3√ 6 3 (56.39)
+ 2k coth(k(x − ωt)
√ √
The above solution corresponds to the choice a1 = 2k, b1 = 2k where
k and ω are given by √ (56.37). Analogous
√ solutions are derived for othe
combinations of a1 = ± 2 k, b1 = ± 2 k.
607
57 A Nonlinear Diffusion Equation I
A Nonlinear Diffusion Equation I. In this Example we shall find the Lie
point symmetries of the nonlinear diffusion equation
uxx
ut = , u = u(x, t) (57.1)
(ux )2
P r(2) Γ(∆)|∆=0 = 0 (57.2)
where
∆≡ut (ux )2 − uxx (57.3)
∂ ∂ ∂
P r(2) Γ = U [t] + U [x] + U [xx] (57.4)
∂ut ∂ux ∂uxx
We thus have
P r(2) Γ(∆) = 0
[t] ∂ [x] ∂ [xx] ∂

⇐⇒ U +U +U (∆) = 0 (57.5)
∂ut ∂ux ∂uxx
⇐⇒ (ux )2 U [t] + 2ut ux U [x] − U [xx] = 0
Using the known expressions for the coefficients U [t] , U [x] and U [xx] , the above
equation takes on the form

2 2
(ux ) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2
+ 2ut ux Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut


608
In the above equation we substitute uxx by ut (ux )2 to account the condition
∆ = 0. Equation (57.6) then becomes

2 2
(ux ) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2

− 2Tx utx + (Uu − 2Xx )[ut (ux )2 ] − 2Tu ux utx

2
− (Tu ut + 3Xu ux )[ut (ux ) ] = 0

the function u. We thus derive the set of the determining equations from
which we can determine the coefficients X, T and U . They are given by
1
X = − (2t + u2 )a1 + a4 u + a5 x
8 (57.8)
a1 t
√
a1 u a6 a8
+ a6 a7 e e + √a1 u
e
1
T = a1 t2 + a2 t + a3 (57.9)
2
1
U = (a1 t + a2 )u − 2ta4 + a9 (57.10)
2
Therefore we get the following generators of the Lie algebra of symmetries
for the equation (57.1):
∂ ∂ ∂ ∂ 1 ∂
X1 = , X2 = x , X3 = , X4 = t + u
∂t ∂x ∂u ∂t 2 ∂u
∂ ∂ √ ∂ ea1 t ∂
X 5 = xu − 2t , X 6 = ea1 t e a1 u , X 7 = √a1 u (57.11)
∂x ∂u ∂x e ∂x
1 ∂ 1 ∂ 1 ∂
X 8 = − (xu2 + 2tx) + t2 + tu
8 ∂x 2 ∂t 2 ∂u
609
58 A Nonlinear Diffusion Equation and its
Potential form
A Nonlinear Diffusion Equation and its potential form.
In this Example we shall find the Lie point symmetries of the nonlinear
diffusion equation
u
x
ut = 2
u = u(x, t) (58.1)
u +1 x
and its potential form
uxx
ut = , u = u(x, t) (58.2)
u2x+1
Invariance of the first equation is being expressed by
P r(2) Γ(∆1 )|∆1 =0 = 0 (58.3)
where
∆1 ≡(u2 + 1)2 ut − (u2 + 1)uxx + 2uu2x (58.4)

∂ ∂ ∂ ∂
P r(2) Γ = U + U [t] + U [x] + U [xx] (58.5)
∂u ∂ut ∂ux ∂uxx
where we have omitted some terms which, when applied to ∆1 , give zero.
We thus have
P r(2) Γ(∆1 ) = 0
∂ ∂ ∂ ∂
⇐⇒ U + U [t] + U [x] + U [xx] (∆1 ) = 0
∂u ∂ut ∂ux ∂uxx (58.6)
⇐⇒{4u(u2 + 1)ut − 2uuxx + 2u2x }U + (u2 + 1)2 U [t]
+ 4uux U [x] − (u2 + 1)U [xx] = 0
610
Using the known expressions of the coefficients U [t] , U [x] and U [xx] , the pre-
vious equation becomes
{4u(u2 + 1)ut − 2uuxx + 2u2x }U

2 2 2
+ (u + 1) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2
+ 4uux Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut

2 (58.7)
− (u + 1) Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux


We shall now take into account the condition ∆1 = 0. We thus make the
substitution
uxx 2uu2x
ut −→ −
u2 + 1 (u2 + 1)2
611
in (58.7) and we obtain the equation
2uu2x
u
2 xx 2
4u(u + 1) 2 − − 2uuxx + 2ux U
u + 1 (u2 + 1)2
2uu2
u
xx
2 2
+ (u + 1) Ut + (Uu − Tt − Xu ux ) 2 − 2 x 2
u + 1 (u + 1)
2uu2x 2
u
xx
− Xt ux − Tu 2 −
u + 1 (u2 + 1)2
2uu2x
u
xx
+ 4uux Ux + (Uu − Xx )ux − (Tx − Tu ux ) 2 −
u + 1 (u2 + 1)2

2
− Xu (ux ) (58.8)
2uu2
u
xx
2
− (u + 1) Uxx − (Txx + 2Txu ux ) 2 − 2 x 2
u + 1 (u + 1)
+ (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2
u
xx 2uu2x
− Tuu 2 − (ux )2 − Xuu (ux )3
u + 1 (u2 + 1)2
2uu2x
u
xx
− Tu 2 − uxx − 3Xu ux uxx = 0
u + 1 (u2 + 1)2
After deriving the determining equations we find the following values of the
infinitesimal coefficients X, T and U
1
X = a1 x + a3 , T = a1 t + a2 , U =0 (58.9)
2
Invariance of the second equation is being expressed by
P r(2) Γ(∆2 )|∆2 =0 = 0 (58.10)
where
∆2 ≡ut (u2x + 1) − uxx (58.11)

∂ ∂ ∂
P r(2) Γ = U [t] + U [x] + U [xx] (58.12)
∂ut ∂ux ∂uxx
612
where we have omitted some terms which, when applied to ∆2 , give zero.
We thus have
P r(2) Γ(∆2 ) = 0
[t] ∂ [x] ∂ [xx] ∂

⇐⇒ U +U +U (∆2 ) = 0 (58.13)
∂ut ∂ux ∂uxx
⇐⇒ (u2x + 1)U [t] + 2ut ux U [x] − U [xx] = 0
Using the known expressions of the coefficients U [t] , U [x] and U [xx] , the pre-

2 2
(ux + 1) Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2


We shall now take into account the condition ∆2 = 0. We thus make the
substitution
uxx
ut −→
u2x + 1
613
in (58.14) and we obtain the equation

2 uxx
(ux + 1) Ut + (Uu − Tt − Xu ux ) 2 − Xt ux
ux + 1
u 2
xx
− Tu 2
ux + 1
u
xx
+2 2 ux Ux + (Uu − Xx )ux − Xu (ux )2
ux + 1
u
xx
− (Tx + Tu ux ) 2
ux + 1 (58.15)
u
xx
− Uxx − (Txx + 2Txu ux ) 2 + (2Uxu − Xxx )ux
ux + 1
u
xx
+ (Uuu − 2Xxu )(ux )2 − Tuu 2 (ux )2
ux + 1
u
xx
− Tu 2 uxx − 3Xu ux uxx = 0
ux + 1
After deriving the determining equations we find the following values of the
infinitesimal coefficients X, T and U
1 1
X = −a3 u + a1 x + a5 , T = a1 t + a2 , U = a1 u + a3 x + a4 (58.16)
2 2
59 A Nonlinear Diffusion Equation II

A Nonlinear Diffusion Equation II. In this Example we shall find the
Lie point symmetries of the nonlinear diffusion equation
ut = (K(u)ux )x , u = u(x, t) (59.1)
ut = K(u)uxx + K 0 (u)(ux )2 (59.2)
P r(2) Γ(∆)|∆=0 = 0 (59.3)
614
where
∆≡ut − K(u)uxx − K 0 (u)(ux )2 (59.4)

∂ ∂ ∂ ∂
P r(2) Γ = U + U [t] + U [x] + U [xx] (59.5)
∂u ∂ut ∂ux ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂
[t] ∂ [x] ∂ [xx] ∂

⇐⇒ U +U +U +U (∆) = 0
∂u ∂ut ∂ux ∂uxx (59.6)
⇐⇒ − U {K 0 (u)uxx + K 00 (u)(ux )2 } + U [t] − K(u)U [xx]
− 2K 0 (u)ux U [x] = 0
Using the known expressions of the coefficients U [t] , U [x] and U [xx] , the above
equation takes on the form
− U {K 0 (u)uxx + K 00 (u)(ux )2 }

− K(u) Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux

(59.7)


0 2
− 2K (u)ux Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut = 0
615
In the above equation, we substitute ut by K(u)uxx + K 0 (u)(ux )2 to account
for the condition ∆ = 0. In that case equation (59.7) becomes
− U {K 0 (u)uxx + K 00 (u)(ux )2 }
+ Ut + (Uu − Tt )[K(u)uxx + K 0 (u)(ux )2 ] − Xt ux
− Tu [K(u)uxx + K 0 (u)(ux )2 ]2 − Xu [K(u)uxx + K 0 (u)(ux )2 ]ux

− K(u) Uxx + (2Uxu − Xxx )ux
− (Txx + 2Txu ux )[K(u)uxx + K 0 (u)(ux )2 ]

+ (Uuu − 2Xxu )(ux )2 − Tuu [K(u)uxx + K 0 (u)(ux )2 ](ux )2
(59.8)

0 2
− Tu [K(u)uxx + K (u)(ux ) ]uxx − 3Xu ux uxx

0
− 2K (u)ux Ux + (Uu − Xx )ux − Xu (ux )2

0 2
− (Tx + Tu ux )[K(u)uxx + K (u)(ux ) ] = 0

equate to zero the easily identified coefficients. For example, the coefficients
of utx and ux utx are −2Tx and −2Tu respectively. We thus have Tx = 0 and
Tu = 0 and thus
T = T (t) (59.9)
− U {K 0 (u)uxx + K 00 (u)(ux )2 }
− Xu [K(u)uxx + K 0 (u)(ux )2 ]ux

− K(u) Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2
(59.10)

3
− Xuu (ux ) + (Uu − 2Xx )uxx − 3Xu ux uxx

0 2
− 2K (u)ux Ux + (Uu − Xx )ux − Xu (ux ) = 0
616
The coefficient of ux uxx is 2K(u)Xu and thus Xu = 0. Therefore
X = X(x, t) (59.11)
− U {K 0 (u)uxx + K 00 (u)(ux )2 }

− K(u) Uxx + (2Uxu − Xxx )ux
(59.12)
2
+ Uuu (ux ) + (Uu − 2Xx )uxx

0
− 2K (u)ux Ux + (Uu − Xx )ux = 0
−U K 0 (u) + K(u)(2Xx − Tt ) = 0 (59.13)
−U K 00 (u) − K(u)Uuu + K 0 (u)(2Xx − Uu − Tt ) = 0 (59.14)
The coefficient of ux is given by
−Xt − K(u)(2Uxu − Xxx ) − 2K 0 (u)Ux = 0 (59.15)
The terms which do not contain partial derivatives of the unknown function
u are given by
Ut − K(u)Uxx = 0 (59.16)
The solution of the system of the determining equations yields the following
values of the infinitesimal coefficients X, T and U :
1
X = a1 x + a3 , T = a1 t + a2 , U =0 (59.17)
2
617
60 The Axially Symmetric Wave Equation
The Axially Symmetric Wave Equation. In this Example we shall find
the Lie point symmetries of the axially symmetric wave equation
1
utt − uxx − ux = 0, u = u(x, t) (60.1)
x
P r(2) Γ(∆)|∆=0 = 0 (60.2)
where
1
∆≡utt − uxx − ux (60.3)
x
∂ ∂ ∂ ∂
P r(2) Γ = X + U [x] + U [tt] + U [xx] (60.4)
∂x ∂ux ∂utt ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂ ∂ ∂ ∂
⇐⇒ X + U [x] + U [tt] + U [xx] (∆) = 0 (60.5)
∂x ∂ux ∂utt ∂uxx
1 1
⇐⇒ X 2 ux − U [x] + U [tt] − U [xx] = 0
x x
618
Using the known expressions for the coefficients U [x] , U [tt] and U [xx] the pre-

1 1 2
X 2 ux − Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut
x x
(60.6)


1
In the above expression we substitute utt by uxx + ux to account for the
x
condition ∆ = 0. We thus obtain the equation

1 1 2
X 2 ux − Ux + (Uu − Xx )ux − Tx ut − Xu (ux ) − Tu ux ut
x x
1
− 2Xt utx + (Uu − 2Tt ) uxx + ux − 2Xu ut utx
x
1
− (Xu ux + 3Tu ut ) uxx + ux (60.7)
x



619
equate to zero the easily identified coefficients first. Equating to zero the
coefficients of ux utx and ut utx we find that Tu = 0 and Xu = 0. We thus have
T = T (x, t) and X = X(x, t) (60.8)
Taking into account the above, equation (60.7) takes on the form

1 1
X 2 ux − Ux + (Uu − Xx )ux − Tx ut
x x
+ Utt − Xtt ux + (2Utu − Ttt )ut + Uuu (ut )2
1
− 2Xt utx + (Uu − 2Tt ) uxx + ux
x (60.9)

− Uxx − Txx ut + (2Uxu − Xxx )ux

2
+ Uuu (ux ) − 2Tx utx + (Uu − 2Xx )uxx = 0
From the above equation we find that

−Xt + Tx = 0 (60.10)
The coefficients of (ut )2 and (ux )2 are
Uuu = 0 (60.11)
The coefficient of uxx is given by
−Tt + Xx = 0 (60.12)
1
Tx + 2Utu − Ttt + Txx = 0 (60.13)
x
1 1
2
X + (Xx − 2Tt ) + Xxx − Xtt − 2Uxu = 0 (60.14)
x x
The constant term (the terms which do not contain partial derivatives of the
function u) is
1
− Ux + Utt − Uxx = 0 (60.15)
x
620
Differentiating (60.10) and (60.12) with respect to t and x respectively and
adding the resulting equations, we obtain that
Xtt = Xxx (60.16)
Similarly upon differentiation of (60.10) and (60.12) with respect to x and t

respectively and addition of the resulting equations, we get
Txx = Ttt (60.17)
From (60.13), taking into account (60.17), we obtain

1
Tx + Utu = 0 (60.18)
x
From (60.14), because of (60.12) and (60.16), takes on the form
X − xTt − 2x2 Uxu = 0 (60.19)
From (60.11) we see that
U = α(x, t)u + β(x, t) (60.20)
Integration of (60.12) with respect to x gives
X = xTt + γ(t) (60.21)
Taking into account (60.20) and (60.21), equation (60.19) becomes
γ(t) − 2x2 αx = 0 (60.22)
from which we get γ(t) = 0 and αx = 0, i.e. α(x, t) = α(t). Therefore
U = α(t)u + β(x, t) and X = xTt (60.23)
Using the second of (60.23), equations (60.16) and (60.10) take on the form
Tttt = 0 and − xTtt + Tx = 0 (60.24)
respectively. Differentiating the second with respect to t and taking into

account the first, we get the equation Txt = 0 and thus
T = δ(x) + (t) (60.25)
621
Since Tttt = 0, we obtain that 000 (t) = 0 and thus
(t) = a2 t2 + a3 t + a4 (60.26)
The second equation of (60.24), taking into account (60.25) and (60.26), we
get δ 0 (x) = 2a2 x and thus
δ(x) = a2 x2 + a5 (60.27)
Therefore, using (60.25)-(60.27) we obtain the explicit expression of T :
T = a2 x2 + a2 t2 + a3 t + a4 (60.28)
where we have absorbed a5 into a4 . Using the second of (60.23) and (60.28),
we obtain
X = (2a2 t + a3 )x (60.29)
1
Equation (60.18), because of (60.10) becomes Utu = − Xt and taking into
x
account (60.23), we have α0 (t) = −Ttt = −2a2 and by integration
α(t) = −2a2 t + a5 (60.30)
We thus have
U = (−2a2 t + a5 )u + β(x, t) (60.31)
Let us now use equation (60.15), which has been left untouched so far. Using
(60.31), this equation becomes
1
− βx + βtt − βxx = 0 (60.32)
x
1 1
ξ = −x + t η = x + t (60.33)
2 2
equation (60.32) can be solved by separation of variables:
β = F (ξ)G(η) (60.34)
622
where F (ξ) and G(η) satisfy the coupled system of differential equations
d2 3 d
F (ξ) = F (ξ) (60.35)
dξ 2 F (ξ) dξ
and
d
2G(η) G(η)
d dη
G(η) = (60.36)
dη d
F (ξ) + 2(ξ − 2η) F (ξ)
dξ
Solving the above system of equations (60.35) and (60.36) we obtain the
value of the function β(x, t):
a6
β(x, t) = p p (60.37)
2a7 (x − t) − 2a8 · a8 + a7 (x + t)
Therefore the function U is given by

a6
U = (−2a2 t + a5 )u + p p (60.38)
2a7 (x − t) − 2a8 · a8 + a7 (x + t)
We thus have found that the values of the infinitesimal coefficients X, T and
U are given by (60.29), (60.28) and (60.38) respectively.
61 The Telegraph Equation

The Telegraph Equation. In this Example we shall find the Lie point
symmetries of the Telegraph equation
utt = uxx + u, u = u(x, t) (61.1)
P r(2) Γ(∆)|∆=0 = 0 (61.2)
where
∆≡utt − uxx − u (61.3)
623
∂ ∂ ∂
P r(2) Γ = U + U [tt] + U [xx] (61.4)
∂u ∂utt ∂uxx
We thus have
P r(2) Γ(∆) = 0
∂ ∂ ∂
⇐⇒ U + U [tt] + U [xx] (∆) = 0 (61.5)
∂u ∂utt ∂uxx
⇐⇒ − U + U [tt] − U [xx] = 0
Using the known values of the coefficients U [tt] and U [xx] , the previous equa-
tion takes on the form
− U + Utt − Xtt ux + (2Utu − Ttt )ut − 2Xtu ux ut



624
In the above equation we substitute utt by uxx +u to account for the condition
∆ = 0. We thus obtain the equation
− U + Utt − Xtt ux + (2Utu − Ttt )ut − 2Xtu ux ut
− 2Xt utx + (Uu − 2Tt )(uxx + u) − 2Xu ut utx
− (Xu ux + 3Tu ut )(uxx + u)



of the function u. However, just for the economy of the calculations, we can
equate to zero the coefficients of some easily identified coefficients, instead
of expanding the above equation into powers of the partial derivatives. For
example, equating to zero the coefficients of ux utx and ut utx , we get
Tu = 0 and Xu = 0 (61.8)
respectively. We thus have that
T = T (x, t) and X = X(x, t) (61.9)
− U + Utt − Xtt ux + (2Utu − Ttt )ut
+ Uuu (ut )2 − 2Xt utx + (Uu − 2Tt )(uxx + u)

− Uxx − Txx ut + (2Uxu − Xxx )ux (61.10)

2
+ Uuu (ux ) − 2Tx utx + (Uu − 2Xx )uxx = 0

The coefficient of (ut )2 and (ux )2 is
Uuu = 0 (61.11)
625
−Xtt − 2Uxu + Xxx = 0 (61.12)
2Utu − Ttt + Txx = 0 (61.13)
−Xt + Tx = 0 (61.14)
−Tt + Xx = 0 (61.15)
Finally the constant term (the terms which are independent of the partial
derivatives of the unknown function) is given by
−U + Utt + u(Uu − 2Tt ) − Uxx = 0 (61.16)
Differentiation of (61.14) with respect to x and (61.15) with respect to t and

adding the resulting equations, we get
Txx = Ttt (61.17)
Similarly differentiating (61.14) with respect to t and (61.15) with respect to

x and adding the resulting equations, we get
Xxx = Xtt (61.18)
Uxu = Utu = 0 (61.19)
The system of equations (61.11) and (61.19) admits the general solution
U = a1 u + F (x, t) (61.20)
Equation (61.16), because of the above expression of U , takes on the form
Ftt − Fxx − F + 2uTt = 0
626
From the above equation we obtain the two equations
Tt = 0 (61.21)
and
Ftt − Fxx − F = 0 (61.22)
Therefore
T = T (x) (61.23)
From (61.15) we get Xx = 0 and thus
X = X(t) (61.24)
Using (61.23) and (61.24), we obtain from (61.17) and (61.18) that Txx = 0
and Xtt = 0. Therefore
T = a2 x + a3 and X = a5 t + a4 (61.25)
However from (61.14) we have X 0 (t) = T 0 (x) and thus a5 = a2 . We thus

obtain from (61.25) that
T = a2 x + a3 and X = a2 t + a4 (61.26)
Let us now solve equation (61.22) to complete the picture. This equation
can be solved by separation of variables (factorization). Introducing the
similarity variables
1 1
ξ = −x + t, η = x+ t (61.27)
2 2
the function F gets factorized into
F (x, t) = G(ξ)H(η) (61.28)
where G(ξ) and H(η) satisfy the equations
d d 1
G(ξ) = µG(ξ) and H(η) = H(η) (61.29)
dξ dη 2µ
627
respectively, where µ is the separation constant. Solving the above equations,
we find
1
G(ξ) = C1 exp µ(−x + t) and H(η) = C2 exp (x + t) (61.30)
4µ
Therefore
1
F (x, t) = C1 C2 exp µ(−x + t) exp (x + t)
4µ
or
1 1
F (x, t) = a5 exp −µ + x+ µ+ t (61.31)
4µ 4µ
We thus have the following expression for the infinitesimal U , taking into
account (61.20) and (61.31)
1 1
U = a1 u + a5 exp −µ + x+ µ+ t (61.32)
4µ 4µ
The generator of the Lie symmetries is then given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂ ∂
= (a2 x + a3 ) + (a2 t + a4 ) + (61.33)
∂t ∂x
1 1 ∂
+ a1 u + a5 exp −µ + x+ µ+ t
4µ 4µ ∂u
∂ ∂ ∂ ∂ ∂
Γ = a1 u + a2 x + t + a3 + a4
∂u ∂t ∂x ∂t ∂x
(61.34)
1 1∂
+ a5 exp −µ + x+ µ+ t
4µ 4µ ∂u
We thus have the following generators of Lie algebra for the Telegraph equa-
tion
∂ ∂ ∂ ∂
X1 = u , X2 = x + t , X3 = ,
∂u ∂t ∂x ∂t
(61.35)
∂ 1 1∂
X4 = , X 5 = exp −µ + x+ µ+ t
∂x 4µ 4µ ∂u
628
62 The Laplace Equation
Case I. The Laplace Equation in Cylindrical Coordinates. We shall
determine the Lie symmetries of Laplace’s equation in cylindrical polar co-
ordinates. It is given by
∂ 2 ψ 1 ∂ψ ∂ 2 ψ
+ + 2 =0 (62.1)
∂ρ2 ρ ∂ρ ∂z
Invariance of the above equation under infinitesimal Lie point transforma-
tions is being expressed by the condition
P r(2) X[∆] =0 (62.2)

∆=0
where P r(2) X is the second prolongation of the vector field X defined by

∂ ∂ ∂ ∂ ∂ ∂
P r(2) X = R +Z +Ψ + Φ[ρ] + Φ[ρρ] + Φ[zz] (62.3)
∂ρ ∂z ∂ψ ∂ψρ ∂ψρρ ∂ψzz
and ∆ is the quantity defined by
∆ = ψρρ + ρ−1 ψρ + ψzz (62.4)
The infinitesimals Φ[ρ] , Φ[ρρ] and Φ[zz] are given by
Φ[ρ] = Ψρ + (Ψψ − Rρ )ψρ − Zρ ψz − Rψ (ψρ )2 − Zψ ψρ ψz (62.5)
Φ[ρρ] = Ψρρ + (Ψψ − 2Rρ )ψρρ + (2Ψρψ − Rρρ )ψρ − Zψ ψz ψρρ

− 3Rψ ψρ ψρρ + (Ψψψ − 2Rρψ )(ψρ )2 − Rψψ (ψρ )3 − Zρρ ψz (62.6)
− Zψψ ψz (ψρ )2 − 2Zρ ψzρ − 2Zψ ψρ ψzρ − 2Zρψ ψρ ψz
and
Φ[zz] = Ψzz + (Ψψ − 2Zz )ψzz + (2Ψzψ − Zzz )ψz − Rψ ψρ ψzz
− 3Zψ ψz ψzz + (Ψψψ − 2Zzψ )(ψz )2 − Zψψ (ψz )3 − Rzz ψρ (62.7)
− Rψψ ψρ (ψz )2 − 2Rz ψzρ − 2Rψ ψz ψzρ − 2Rzψ ψρ ψz
629
respectively. We thus have
1 1
P r(2) X[∆] = − 2
Rψρ + Φ[ρ] + Φ[ρρ] + Φ[zz]
ρ ρ

1 1 2
= − 2 Rψρ + Ψρ + (Ψψ − Rρ )ψρ − Zρ ψz − Rψ (ψρ ) − Zψ ψρ ψz
ρ ρ
+ Ψρρ + (Ψψ − 2Rρ )ψρρ + (2Ψρψ − Rρρ )ψρ − Zψ ψz ψρρ
− 3Rψ ψρ ψρρ + (Ψψψ − 2Rρψ )(ψρ )2 − Rψψ (ψρ )3 − Zρρ ψz (62.8)
− Zψψ ψz (ψρ )2 − 2Zρ ψzρ − 2Zψ ψρ ψzρ − 2Zρψ ψρ ψz
+ Ψzz + (Ψψ − 2Zz )ψzz + (2Ψzψ − Zzz )ψz − Rψ ψρ ψzz
− 3Zψ ψz ψzz + (Ψψψ − 2Zzψ )(ψz )2 − Zψψ (ψz )3 − Rzz ψρ
− Rψψ ψρ (ψz )2 − 2Rz ψzρ − 2Rψ ψz ψzρ − 2Rzψ ψρ ψz
In the above expression we substitute uzz by −(ρ−1 ψρ + ψρρ ) to account for

the condition ∆ = 0. We thus see that condition P r(2) X[∆] = 0 is
∆=0
equivalent to

1 1 2
− 2 Rψρ + Ψρ + (Ψψ − Rρ )ψρ − Zρ ψz − Rψ (ψρ ) − Zψ ψρ ψz
ρ ρ
+ Ψρρ + (Ψψ − 2Rρ )ψρρ + (2Ψρψ − Rρρ )ψρ − Zψ ψz ψρρ
− 3Rψ ψρ ψρρ + (Ψψψ − 2Rρψ )(ψρ )2 − Rψψ (ψρ )3 − Zρρ ψz
− Zψψ ψz (ψρ )2 − 2Zρ ψzρ − 2Zψ ψρ ψzρ − 2Zρψ ψρ ψz (62.9)
−1
+ Ψzz − (Ψψ − 2Zz )(ρ ψρ + ψρρ ) + (2Ψzψ − Zzz )ψz
+ (Rψ ψρ + 3Zψ ψz )(ρ−1 ψρ + ψρρ )
+ (Ψψψ − 2Zzψ )(ψz )2 − Zψψ (ψz )3 − Rzz ψρ
− Rψψ ψρ (ψz )2 − 2Rz ψzρ − 2Rψ ψz ψzρ − 2Rzψ ψρ ψz = 0
We now have to equate to zero the coefficients of the partial derivatives of the
function ψ. However, just for the economy of the calculations, we can equate
to zero some easily identified coefficients. For example, the coefficients of
ψρ ψzρ and ψz ψzρ are −2Zψ and −2Rψ respectively. We thus have
Zψ = 0 and Rψ = 0 (62.10)
630

1 1
− 2 Rψρ + Ψρ + (Ψψ − Rρ )ψρ − Zρ ψz
ρ ρ
+ Ψρρ + (Ψψ − 2Rρ )ψρρ + (2Ψρψ − Rρρ )ψρ
(62.11)
+ Ψψψ (ψρ )2 − Zρρ ψz − 2Zρ ψzρ
+ Ψzz − (Ψψ − 2Zz )(ρ−1 ψρ + ψρρ ) + (2Ψzψ − Zzz )ψz
+ Ψψψ (ψz )2 − Rzz ψρ − 2Rz ψzρ = 0
The terms (ψρ )2 and (ψz )2 have common coefficient Ψψψ . We thus have
Ψψψ = 0 (62.12)
The coefficient of ψρρ is −2Rρ + 2Zz . We then have
Rρ − Zz = 0 (62.13)
The coefficient of ψzρ is −2(Zρ + Rz ). We then have
Zρ + Rz = 0 (62.14)
The coefficient of ψρ is
1 1
− 2
R + (Ψψ − Rρ ) + (2Ψρψ − Rρρ )
ρ ρ
(62.15)
1
− (Ψψ − 2Zz ) − Rzz = 0
ρ
The coefficient of ψz is
1
− Zρ − Zρρ + 2Ψzψ − Zzz = 0 (62.16)
ρ
The constant term (the terms which do not contain any partial derivatives
of the function ψ) is
1
Ψρ + Ψρρ + Ψzz = 0 (62.17)
ρ
Differentiation of (62.13) with respect to z and (62.14) with respect to ρ and
subtracting the resulting equations we obtain the equation
Zρρ + Zzz = 0 (62.18)
631
Similarly differentiation of (62.13) with respect to ρ and (62.14) with respect
to z and adding the resulting equations we obtain the equation
Rρρ + Rzz = 0 (62.19)
From equation (62.12) we conclude that Ψ is a linear function with respect

to ψ:
Ψ = A(ρ, z)ψ + B(ρ, z) (62.20)
Equation (62.16), taking into account (62.18) and (62.20), becomes
Zρ = 2ρAz (62.21)
The above equation, since Zρ = −Rz according to (62.14), takes on the form
−Rz = 2ρAz which can also be written as (2ρA + R)z = 0 and by integration
we get
2ρA + R = F (ρ) (62.22)

1
Equation (62.15), taking into account (62.19) and (62.20), becomes − R+
ρ2
1
2Aρ + = 0, which can also be written, thanks to (62.13), in the form
ρ
R = 2ρ2 Aρ + ρRρ , which is equivalent to
1
R + 2A =0
ρ ρ
Integrating with respect to ρ the above equation we obtain

1
R + 2A = G(z)
ρ
and then
2ρA + R = ρG(z) (62.23)
Comparing (62.22) to (62.23) we have F (ρ) = ρG(z) and thus we see that
G(z) should be a constant, which we call a2 . Therefore F (ρ) = a1 ρ and then
2ρA + R = a2 ρ (62.24)
632
Solving the rest equations, we get the following expressions for R, Z and Ψ:
R = a2 ρ + a3 ρz
1
Z = a1 + a2 z + a3 (−ρ2 + z 2 ) (62.25)
2
1
Ψ = − a3 zψ + a4 ψ
2
The generator of the symmetries is thus given by
∂ ∂ ∂
X=R +Z +Ψ
∂ρ ∂z ∂ψ

∂ 1 2 2 ∂
= (a2 ρ + a3 ρz) + a1 + a2 z + a3 (−ρ + z ) (62.26)
∂ρ 2 ∂z

1 ∂
+ − a3 zψ + a4 ψ
2 ∂ψ
∂ ∂ ∂
X = a1 + a2 ρ + z
∂z ∂ρ ∂z
(62.27)
∂ 1 2 2 ∂ 1 ∂ ∂
+ a3 ρz + (−ρ + z ) − zψ + a4 ψ
∂ρ 2 ∂z 2 ∂ψ ∂ψ
Therefore the generators of the algebra of Lie symmetries of the original
equation (62.1) are given by
∂ ∂ ∂
X1 = , X2 = ρ + z ,
∂z ∂ρ ∂z
(62.28)
∂ 1 ∂ 1 ∂ ∂
X 3 = ρz + (−ρ2 + z 2 ) − zψ , 4
X =ψ
∂ρ 2 ∂z 2 ∂ψ ∂ψ
We shall determine next some general solutions of Laplace’s equation (62.1).
First Solution. We consider the solution associated to the generator X 1 .
ψ = F (ξ), ξ=ρ (62.29)
and thus the original equation is being converted into
ξF 00 (ξ) + F 0 (ξ) = 0 (62.30)
633
The general solution of the above equation is given by F (ξ) = A + B· ln(ξ)
and thus the solution of the original equation is
ψ(ρ, z) = A + B· ln(ρ) (62.31)
Second Solution. In this case we shall find the general solution associated
to the generator X 2 . Considering the auxiliary system
dρ dz dψ
= = (62.32)
ρ z 0
two invariants determined from the above system are
ρ
ξ= , ψ = F (ξ) (62.33)
z
The original equation thus becomes
1
(1 − ξ 2 )F 00 (ξ) + − 2ξ F 0 (ξ) = 0 (62.34)
ξ
1
F (ξ) = A + B·arctan p (62.35)
ξ2 − 1
Therefore the general solution of the original equation is given by
z
ψ(ρ, z) = A + B·arctan p (62.36)
ρ2 − z 2
Third Solution. In this case we shall find the general solution associated
to the linear combination of generators X 2 and a·X 4 , i.e. X 2 + a·X 4 :
∂ ∂ ∂
ρ +z + a·ψ (62.37)
∂ρ ∂z ∂ψ
Considering the auxiliary system
dρ dz dψ
= = (62.38)
ρ z aψ
two invariants determined from the above system are
ρ
ξ= , ψ = z 1/a F (ξ) (62.39)
z
634
The original equation thus becomes
2ξ 1 0 1 1
(1 + ξ 2 )F 00 (ξ) + 2ξ − + F (ξ) + 2 − F (ξ) = 0 (62.40)
a ξ a a
The general solution of the above equation is being expressed in terms of the
Hypergeometric function
1 a−1 a−2
2
F (ξ) = A· 2 F1 − , ; ; 1+ξ +
2a 2a 2a
a+2 (62.41)
a + 1 2a + 1 3a + 2
+ B·(1 + ξ 2 ) 2a 2 F1 , ; ; 1 + ξ2
2a 2a 2a
We thus find the general solution of the original equation is given by
1 a − 1 a − 2 z 2 + ρ2
ψ(ρ, z) = A· 2 F1 − , ; ; +
2a 2a 2a z2
(62.42)
z 2 + ρ2 a + 2 a + 1 2a + 1 3a + 2 z 2 + ρ2
+ B· 2a 2 F1 , ; ;
z2 2a 2a 2a z2
Case II. The 3-dimensional Laplace Equation in Cartesian
Coordinates. This equation has already been considered in §19.
Chapter XX
The Physics of Fluids
63 The Shallow Water Wave Equation
In this section we shall analyse the Lie point symmetries of the shallow water
wave equation
uxxxt + αux uxt + βut uxx − uxt − uxx = 0 (63.1)
where α and β are arbitrary constants.

Invariance of this equation is being expressed by the condition
P r(4) Γ(∆) =0 (63.2)

∆=0
635
where
∆ = uxxxt + αux uxt + βut uxx − uxt − uxx (63.3)
and P r(4) Γ is the fourth prolongation given by
∂ ∂ ∂ ∂ ∂ ∂
P r(4) Γ = T +X +U + U [t] + U [x] + U [xt]
∂t ∂x ∂u ∂ut ∂ux ∂uxt
(63.4)
∂ ∂
+ U [xx] + U [xxxt]
∂uxx ∂uxxxt
In the above expression for the fourth prolongation Γ(4) , we have omitted
terms which, when applied to ∆, give zero.
We thus have
P r(4) Γ(∆) =
∂ ∂ ∂ ∂ ∂ ∂
= T +X +U + U [t] + U [x] + U [xt]
∂t ∂x ∂u ∂ut ∂ux ∂uxt
∂ ∂
(63.5)
+ U [xx] + U [xxxt] uxxxt + αux uxt
∂uxx ∂u
xxxt
+ βut uxx − uxt − uxx
= βuxx U [t] + αuxt U [x] + (αux − 1)U [xt] + (βut − 1)U [xx] + U [xxxt]
We have the following expressions for the U [t] , U [x] , U [xt] , U [xx] and U [xxxt]
coefficients:
U [t] = Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu ) (63.6)
U [x] = Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu ) (63.7)
(63.8)
U [xt] = Uxt + (Uxu − Txt )ut + (Utu − Xxt )ux
− (Xu ux + Tu ut )uxt − Xt uxx − Tx utt (63.9)
− Txu (ut )2 − Xtu (ux )2 − Xu ut uxx
636
and
U [xxxt] = Uxxxt + (Uu − Tt − 3Xx )uxxxt − (4Xu ux + 2Tu ut )uxxxt
− (Xt + Xu ut )uxxxx + (3Uxxuu − Xxxxu − 3Txxtu )ut ux
+ (Uxxxu − Txxxt )ut + (3Uxxu − 3Txxt − Xxxx )uxt − Txxx utt
− Xtuuu (ux )4 + (Utuuu − 3Xxtuu )(ux )3 + 3(Uxtuu − Xxxtu )(ux )2
− Xuuuu ut (ux )4 + (Uuuuu − Ttuuu − Xxuuu )ut (ux )3
+ 3(Uxuuu − Txtuu − Xxxuu )ut (ux )2 − Txxxu (ut )2 − 3Txxuu (ut )2 ux
− 3Txuuu (ut )2 (ux )2 − Tuuuu (ut )2 (ux )3
− 3(Xxx + Txt − Uxu )uxxt − 6(Tu uxt + Xu uxx )uxxt − 6Xuu (ux )2 uxxt
− (9Xxu + 3Ttu − 3Uuu )ux uxxt − 6(Txu + Tuu ux )ut uxxt
+ (Utu − 3Xxt )uxxx + (Uuu − 3Xxu − Ttu )ut uxxx
+ (3Uxxtu − Xxxxt )ux + 3(Uxtu − Xxxt )uxx − 3Xtu (uxx )2
(63.10)
− 3Xuu ut (uxx )2 − 6Txu (uxt )2 − 6Tuu ux (uxt )2 − 3(Ttu + 3Xxu )uxx uxt
+ 3(Uuu − 2Tuu ut )uxt uxx − 12Xuu ux uxt uxx − 3(Txu + Tuu ux )utt uxx
− 3Txx uxtt − 3Tu uxx uxtt − 6Txu ux uxtt − 3Tuu (ux )2 uxtt
− 3(Tx + Tu ux )uxxtt + 6Uxuu ux uxt + 3(Uxuu − Txtu )ut uxx
+ 3(Uuuu − Ttuu − 3Xxuu )(ux )2 uxt − 3(Xxxu + Ttuu ux )ut uxx
+ 3(Utuu − 3Xxtu + Uuuu ut )ux uxx − Tuuu (ux )3 utt − 6Tuuu ut (ux )2 uxt
− 3Tuuu ux (ut )2 uxx − 6Txtu ux uxt − 3Txuu (ux )2 utt − 3Txuu (ut )2 uxx
− 12Txuu ut ux uxt − 3Txxu ux utt − 6Txxu ut uxt − Tuu (ut )2 uxxx
− 6Xxxu ux uxt − Tu utt uxxx − 4Xu uxt uxxx − 4Xtu ux uxxx
− 4Xuu ux ut uxxx − 6Xtuu (ux )2 uxx − 4Xuuu (ux )3 uxt
− 6Xuuu (ux )2 ut uxx − 9Xxuu ut ux uxx
637
Using the above values of the coefficients, the expression for the fourth pro-
longation becomes
P r(4) Γ(∆) =

= βuxx Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )

+ αuxt Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu )
+ (αux − 1)

× Uxt + (Uxu − Txt )ut + (Utu − Xxt )ux


2 2
− Txu (ut ) − Xtu (ux ) − Xu ut uxx
+ (βut − 1)

× Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux


+ Uxxxt + (Uu − Tt − 3Xx )uxxxt − (4Xu ux + 2Tu ut )uxxxt

638
− 6Xuuu (ux )2 ut uxx − 9Xxuu ut ux uxx
In the previous equation, we substitute uxxxt by −αux uxt − βut uxx + uxt + uxx
to account for the condition ∆ = 0. We thus obtain that the condition
639
P r(4) Γ(∆) = 0 is equivalent to the equation
∆=0

βuxx Ut + ut Uu − ut (Tt + ut Tu ) − ux (Xt + ut Xu )

+ αuxt Ux + ux Uu − ut (Tx + ux Tu ) − ux (Xx + ux Xu )
+ (αux − 1)×

× Uxt + (Uxu − Txt )ut + (Utu − Xxt )ux


2 2
− Txu (ut ) − Xtu (ux ) − Xu ut uxx
+ (βut − 1)×

× Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux


+ Uxxxt + (Uu − Tt − 3Xx )(−αux uxt − βut uxx + uxt + uxx )

− (4Xu ux + 2Tu ut )(−αux uxt − βut uxx + uxt + uxx )
640
− 6Xuuu (ux )2 ut uxx − 9Xxuu ut ux uxx = 0
We have next to equate to zero all the coefficients of the partial derivatives
of the function u. However, just for the economy of the calculations, we
consider first the easily identified coefficients of uxxtt , utt uxxx , uxt uxxx and
uxxxx . Equating to zero all these coefficients, we find
Tx = 0, Tu = 0, Xu = 0, Xt = 0 (63.11)
and thus
T = T (t), X = X(x) (63.12)
641
Therefore the equation P r(4) Γ(∆) = 0 becomes
∆=0

βuxx Ut + (Uu − Tt )ut

+ αuxt Ux + (Uu − Xx )ux
+ (αux − 1)×

× Uxt + Uxu ut + Utu ux

+ Uuu ux ut + (Uu − Xx − Tt )uxt
+ (βut − 1)× (63.13)

2
× Uxx + (2Uxu − Xxx )ux + Uuu (ux ) + (Uu − 2Xx )uxx

+ 3Uxxuu ut ux + Uxxxu ut + (3Uxxu − Xxxx )uxt
+ Utuuu (ux )3 + 3Uxtuu (ux )2 + Uuuuu ut (ux )3
+ 3Uxuuu ut (ux )2 − 3(Xxx − Uxu )uxxt + 3Uuu ux uxxt
+ Utu uxxx + Uuu ut uxxx + 3Uxxtu ux + 3Uxtu uxx
+ 3Uuu uxt uxx + 6Uxuu ux uxt + 3Uxuu ut uxx
+ 3Uuuu (ux )2 uxt + 3(Utuu + Uuuu ut )ux uxx = 0
Equating to zero either the coefficient of ux uxxt or ut uxxx , we get
Uuu = 0 (63.14)
642
Because of the above, equation (63.13) gets simplified to

βuxx Ut + (Uu − Tt )ut + αuxt Ux + (Uu − Xx )ux
+ (αux − 1)×

× Uxt + Uxu ut + Utu ux + (Uu − Xx − Tt )uxt
+ (βut − 1)× (63.15)

× Uxx + (2Uxu − Xxx )ux + (Uu − 2Xx )uxx

+ Uxxxu ut + (3Uxxu − Xxxx )uxt − 3(Xxx − Uxu )uxxt + Utu uxxx
+ 3Uxxtu ux + 3Uxtu uxx = 0
From the above equation we see that both coefficients of ut uxx and ux uxt ,
when equated to zero, give
Uu + Xx = 0 (63.16)
On the other hand, the coefficient of uxxx , when equated to zero, gives
Utu = 0 (63.17)
Let us now collect what we have found so far: Using Uuu = 0 and Utu = 0
we conclude that
U = A(x)u + B(x, t) (63.18)
We see further that Uu + Xx = 0 gives
A(x) + X 0 (x) = 0 (63.19)
since X = X(x).
Equating to zero the coefficients of uxxt , uxx and uxt we obtain
Xxx − Uxu = 0 (63.20)
βUt − Xx − Tt = 0 (63.21)
643
and
αUx + 3Uxxu − 2Xx − Xxxx = 0 (63.22)
respectively.
Let us now simplify further equation (63.15) taking into account the previ-
ously found relations:

(αux − 1)× Uxt + Uxu ut

(63.23)
+ (βut − 1)× Uxx + (2Uxu − Xxx )ux
+ Uxxxt + Uxxxu ut = 0
The coefficient of ux ut is
αUxu + β(2Uxu − Xxx ) = 0 (63.24)
The coefficient of ux is given by
αUxt − 2Uxu + Xxx = 0 (63.25)
The coefficient of ut is given by
−Uxu + βUxx + Uxxxu = 0 (63.26)
The term which is independent of the partial derivatives of the function u is

given by
−Uxt − Uxx + Uxxxt = 0 (63.27)
Equation (63.20), because of (63.18), takes on the form
X 00 (x) − A0 (x) = 0 (63.28)
The solution of the system of equations (63.19) and (63.28) yields
X(x) = a1 x + a2 , A(x) = −a1 (63.29)
Therefore we get from (63.18) that
U = −a1 u + B(x, t) (63.30)
644
Equations (63.21) and (63.22), because of the expressions of U and X, con-
stitute the system
βBt = a1 + T 0 (t), αBx = 2a1 (63.31)
The solution of the above system gives

a1 a1 1
B(x, t) = 2 x + t + T (t) + a3 (63.32)
α β β
We thus have
a1 a1 1
U = −a1 u + 2 x + t + T (t) + a3 (63.33)
α β β
Equations (63.24)-(63.27) become identities.
Therefore we have found the following values for the functions T (t), X(x)
and U (x, t, u):

2x t 1
T = T (t), X = a1 x + a2 , U = a1 − u + + + T (t) + a3 (63.34)
α β β
The generator of Lie point symmetries is thus given by
∂ ∂ ∂
Γ=T +X +U
∂t ∂x ∂u
∂ ∂
= T (t) + (a1 x + a2 ) (63.35)
∂t ∂x
2x t 1 ∂
+ a1 −u + + + T (t) + a3
α β β ∂u

∂ 2x t ∂ ∂ ∂ ∂ 1 ∂
Γ = a1 x + −u+ + +a2 +a3 +T (t) + (63.36)
∂x α β ∂u ∂x ∂u ∂t β ∂u
We thus have the following generators of the Lie algebra of symmetries of
the Shallow Water Wave Equation
(1) ∂ 2x t ∂ ∂
Γ =x + −u + + , Γ(2) = ,
∂x α β ∂u ∂x
(63.37)
∂ ∂ 1 ∂
Γ(3) = , Γ(4) = T (t) +
∂u ∂t β ∂u
645
We shall determine next some general solutions of the equation:
1
We consider the linear combination Γ(2) + Γ(4) with T (t) = 0 (the prime
F (t)
denoting derivative with respect to time t), i.e. we consider the generator
∂ 1 ∂ 1 ∂
X= + 0 + 0
(63.38)
∂x F (t) ∂t βF (t) ∂u
In order to find the invariants, we consider the auxiliary system
dx dt du
= F 0 (t) = βF 0 (t) (63.39)
1 1 1
dx dt
Equation = F 0 (t) admits the general solution x = F (t) + C1 and thus
1 1
we get the first invariant
z = x − F (t) (63.40)
dt du
Equation F 0 (t) = βF 0 (t) admits the general solution βu = t + C2 and
1 1
thus we get a second invariant w(z) given by
t
u= + w(z) (63.41)
β
Since (the prime on w denotes derivative with respect to z)
uxxxt = −F 0 (t)w(4) (z), ux = w0 (z), uxt = −F 0 (t)w00 (z),

1 (63.42)
ut = − F 0 (t)w0 (z), uxx = w00 (z)
β
the original equation (63.1) takes on the form
w(4) (z) + (α + β)w0 (z)w00 (z) − w00 (z) = 0 (63.43)
Upon integration of the above equation we obtain

1
w000 (z) + (α + β)(w0 (z))2 − w0 (z) = A (63.44)
2
where A is an arbitrary constant. The substitution
U = w0 (z) (63.45)
646
1
U 00 (z) + (α + β)U 2 (z) − U (z) = A (63.46)
2
Multiplying by U 0 (z) the above equation, we get
1
U 00 (z)U 0 (z) + (α + β)U 2 (z)U 0 (z) − U (z)U 0 (z) = AU 0 (z)
2
which by integration gives
1
(U 0 (z))2 + (α + β)U 3 (z) − U 2 (z) = AU (z) + B (63.47)
3
Introducing the transformation
1 12
U (z) = − Y (z) (63.48)
α+β α+β
(Y 0 (z))2 = 4Y 3 (z) − g2 Y (z) − g3 (63.49)
where
1
g2 = [1 + (α + β)A]
12 (63.50)
1 1
g3 = − − (α + β)[A + (α + β)B]
216 144
Equation (63.49) is the Weirstrass equation with general solution given by
Y (z) = ℘(z; g2 , g3 ) (63.51)

1 12
U (z) = − ℘(z; g2 , g3 ) (63.52)
α+β α+β
Combining (63.41), (63.45) and (63.52), we obtain the general solution of the
shallow water wave equation (63.1)
Z
t
u = + U (z)dz (63.53)
β z=x−F (t)
647
or explicitly
Z
t 1 12
u= + (x − F (t)) − ℘(z; g2 , g3 )dz (63.54)
β α+β α+β z=x−F (t)
where z = x − F (t) by (63.40).

References
P. A. Clarkson and E. L. Mansfield: ”Symmetry Reductions and Exact
Solutions of Shallow Water Wave Equations”. arXiv : solv-int/9409003v1.
64 The Euler Equations

Euler’s equations in Fluid Mechanics describe the motion of a fluid in terms
of the velocity u and its pressure p.
64.1 The 3-dimensional form of the Euler

equations
Euler’s equations in 3-dimensional form read
∂u
+ u·∇u = −∇p
∂t (64.1)
∇·u = 0
We denote by u = (u, v, w) the three components of the velocity vector. The

functions u, v and w depend on the three space variables x, y and z and the
time t. The above equations can be written in component form as
ut + uux + vuy + wuz + px = 0
vt + uvx + vvy + wvz + py = 0
(64.2)
wt + uwx + vwy + wwz + pz = 0
ux + vy + wz = 0
We denote by X the generator of Lie point symmetries where X is the vector

∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
(64.3)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂u ∂v ∂w ∂p
648
The functions {ξ 1 , · · ·, ξ 4 } and {φ1 , · · ·, φ4 } depend both on the space-time
variables x, y, z, t and the functions u, v, w and p.
The first prolongation of the vector X is denoted by P r(1) X and is given by
∂ ∂ ∂ ∂
P r(1) X = X + φx + φy + φz + φt
∂ux ∂uy ∂uz ∂ut
∂ ∂ ∂ ∂
+ ψx + ψy + ψz + ψt
∂vx ∂vy ∂vz ∂vt
(64.4)
∂ ∂ ∂ ∂
+ χx + χy + χz + χt
∂wx ∂wy ∂wz ∂wt
∂ ∂ ∂
+ πx + πy + πz
∂px ∂py ∂pz
We further denote by ∆(1) , ∆(2) , ∆(3) , ∆(4) the quantities
∆(1) = ut + uux + vuy + wuz + px

∆(2) = vt + uvx + vvy + wvz + py
(64.5)
∆(3) = wt + uwx + vwy + wwz + pz
∆(4) = ux + vy + wz
respectively.
The Lie symmetries of Euler’s equations can be revealed once we know the
explicit form of the functions {ξ 1 , · · ·, ξ 4 } and {φ1 , · · ·, φ4 }. The Lie symme-
tries are determined by the conditions
P r(1) X[∆(k) ] = 0, as long as ∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0 (64.6)
Applying the above conditions, we obtain
P r(1) X[∆(1) ] = φt + (φ1 )ux + (φ2 )uy + (φ3 )uz

(64.7)
+ uφx + vφy + wφz + π x
P r(1) X[∆(2) ] = ψ t + (φ1 )vx + (φ2 )vy + (φ3 )vz

(64.8)
+ uψ x + vψ y + wψ z + π y
P r(1) X[∆(3) ] = χt + (φ1 )wx + (φ2 )wy + (φ3 )wz
(64.9)
+ uχx + vχy + wχz + π z
P r(1) X[∆(4) ] = φx + ψ y + χz (64.10)
649
The above expressions should be transformed further, using the various ex-
pressions for the coefficients {φx , · · ·, π z } appearing in the first prolongation
formulas of X. These coefficients have been calculated using the formulas
(2.39), (2.43) or (2.44) of Olver (Reference [11]). The explicit expressions of
these coefficients are listed below.
φx = (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u
− px [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
(64.11)
− uy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− uz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− ut [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
φy = (φ1 )y + uy (φ1 )u + vy (φ1 )v + wy (φ1 )w − (uy )2 (ξ 2 )u

− py [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
(64.12)
− uy [vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y ]
− uz [uy (ξ 3 )v + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− ut [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
φz = (φ1 )z + uz (φ1 )u + vz (φ1 )v + wz (φ1 )w − (uz )2 (ξ 3 )u
− pz [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
(64.13)
− uy [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− uz [vz (ξ 3 )v + wz (ξ 3 )w + (ξ 3 )z ]
− ut [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
φt = (φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w − (ut )2 (ξ 4 )u
− pt [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [ut (ξ 1 )u + vt (ξ 1 )v + wt (ξ 1 )w + (ξ 1 )t ]
(64.14)
− uy [ut (ξ 2 )u + vt (ξ 2 )v + wt (ξ 2 )w + (ξ 2 )t ]
− uz [ut (ξ 3 )u + vt (ξ 3 )v + wt (ξ 3 )w + (ξ 3 )t ]
− ut [vt (ξ 4 )v + wt (ξ 4 )w + (ξ 4 )t ]
650
ψ x = (φ2 )x + ux (φ2 )u + vx (φ2 )v + wx (φ2 )w − (vx )2 (ξ 1 )v
− px [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [ux (ξ 1 )u + wx (ξ 1 )w + (ξ 1 )x ]
(64.15)
− vy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− vz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− vt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
ψ y = (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w − (vy )2 (ξ 2 )v
− py [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
(64.16)
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]
− vz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− vt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
ψ z = (φ2 )z + uz (φ2 )u + vz (φ2 )v + wz (φ2 )w − (vz )2 (ξ 3 )v
− pz [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
(64.17)
− vy [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− vz [uz (ξ 3 )u + wz (ξ 3 )w + (ξ 3 )z ]
− vt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
ψ t = (φ2 )t + ut (φ2 )u + vt (φ2 )v + wt (φ2 )w − (vt )2 (ξ 4 )v
− pt [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [ut (ξ 1 )u + vt (ξ 1 )v + wt (ξ 1 )w + (ξ 1 )t ]
(64.18)
− vy [ut (ξ 2 )u + vt (ξ 2 )v + wt (ξ 2 )w + (ξ 2 )t ]
− vz [ut (ξ 3 )u + vt (ξ 3 )v + wt (ξ 3 )w + (ξ 3 )t ]
− vt [ut (ξ 4 )u + wt (ξ 4 )w + (ξ 4 )t ]
χx = (φ3 )x + ux (φ3 )u + vx (φ3 )v + wx (φ3 )w − (wx )2 (ξ 1 )w
− px [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [ux (ξ 1 )u + vx (ξ 1 )v + (ξ 1 )x ]
(64.19)
− wy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− wz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− wt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
651
χy = (φ3 )y + uy (φ3 )u + vy (φ3 )v + wy (φ3 )w − (wy )2 (ξ 2 )w
− py [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
(64.20)
− wy [uy (ξ 2 )u + vy (ξ 2 )v + (ξ 2 )y ]
− wz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− wt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
χz = (φ3 )z + uz (φ3 )u + vz (φ3 )v + wz (φ3 )w − (wz )2 (ξ 3 )w
− pz [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
(64.21)
− wy [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− wz [uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]
− wt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
χt = (φ3 )t + ut (φ3 )u + vt (φ3 )v + wt (φ3 )w − (wt )2 (ξ 4 )w
− pt [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [ut (ξ 1 )u + vt (ξ 1 )v + wt (ξ 1 )w + (ξ 1 )t ]
(64.22)
− wy [ut (ξ 2 )u + vt (ξ 2 )v + wt (ξ 2 )w + (ξ 2 )t ]
− wz [ut (ξ 3 )u + vt (ξ 3 )v + wt (ξ 3 )w + (ξ 3 )t ]
− wt [ut (ξ 4 )u + vt (ξ 4 )v + (ξ 4 )t ]
π x = (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w − (px )2 (ξ 1 )p
− px [ux (ξ 1 )u + vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x − (φ4 )p ]
− px py (ξ 2 )p − px pz (ξ 3 )p − px pt (ξ 4 )p
(64.23)
− py [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− pz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− pt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
π y = (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w − (py )2 (ξ 2 )p
− px [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− py px (ξ 1 )p − py pz (ξ 3 )p − py pt (ξ 4 )p
(64.24)
− py [uy (ξ 2 )u + vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y − (φ4 )p ]
− pz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− pt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
652
π z = (φ4 )z + uz (φ4 )u + vz (φ4 )v + wz (φ4 )w − (pz )2 (ξ 3 )p
− px [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− pz px (ξ 1 )p − pz py (ξ 2 )p − pz pt (ξ 4 )p
(64.25)
− py [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− pz [uz (ξ 3 )u + vz (ξ 3 )v + wz (ξ 3 )w + (ξ 3 )z − (φ4 )p ]
− pt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
In previous formulas (64.7)-(64.10) we substitute the above listed expressions
for φx , · · ·, π z and we find successively, starting from equation (64.10).
P r(1) X[∆(4) ] =
= (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
+ (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w − (vy )2 (ξ 2 )v
− vx [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ] (64.26)
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]
+ (φ3 )z + uz (φ3 )u + vz (φ3 )v + wz (φ3 )w − (wz )2 (ξ 3 )w
− wz [uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]
653
In the above equation we make the substitutions
px −→ − (ut + uux + vuy + wuz )
py −→ − (vt + uvx + vvy + wvz )
(64.27)
pz −→ − (wt + uwx + vwy + wwz )
wz −→ − (ux + vy )
to account for the conditions ∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0.

We thus see that P r(1) X[∆(4) ] = 0 is equivalent to
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
(φ1 )x + ux (φ1 )u +vx (φ )v + wx (φ1 )w − (ux )2 (ξ 1 )u
1
+ (ut + uux + vuy + wuz )×

×[ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
+ (vt + uvx + vvy + wvz )×
×[vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
2 2 2
− vy [uy (ξ )u + wy (ξ )w + (ξ )y ]
+ (φ3 )z + uz (φ3 )u + vz (φ3 )v − (ux + vy )(φ3 )w − (ux + vy )2 (ξ 3 )w
+ (wt + uwx + vwy − (ux + vy )w)×
×[wx (ξ 1 )p + wy (ξ 2 )p − (ux + vy )(ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [uz (ξ 1 )u + vz (ξ 1 )v − (ux + vy )(ξ 1 )w + (ξ 1 )z ]
− wy [uz (ξ 2 )u + vz (ξ 2 )v − (ux + vy )(ξ 2 )w + (ξ 2 )z ]
+ (ux + vy )[uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]
− wt [uz (ξ 4 )u + vz (ξ 4 )v − (ux + vy )(ξ 4 )w + (ξ 4 )z ] = 0
the various functions. However, just for the economy of the calculations, we
654
equate to zero the easily identified coefficients of some partial derivatives. In
fact, we see that (ut )2 , (vt )2 and (wt )2 have common coefficient (ξ 4 )p . The
terms (uz )2 and (vz )2 have common coefficient w(ξ 3 )p . The terms ut wx and
vt wy have common coefficient −(ξ 4 )w . The coefficients of wt uz , uz vx and wt vz
are −(ξ 4 )u , −(ξ 3 )v and −(ξ 4 )v respectively. The coefficient of vz vx is equal
to w(ξ 1 )p + u(ξ 3 )p . Therefore equating to zero all these coefficients, we have
(ξ 4 )p = (ξ 4 )u = (ξ 4 )v = (ξ 4 )w = 0, (ξ 3 )p = (ξ 3 )v = 0, (ξ 1 )p = 0 (64.29)
We thus see that equation (64.28) gets simplified into
(φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u

+ (ut + uux + vuy + wuz )×[uy (ξ 2 )p − (φ1 )p ]
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
− uz [ux (ξ 3 )u + wx (ξ 3 )w + (ξ 3 )x ] − ut (ξ 4 )x
+ (vt + uvx + vvy + wvz )×[vy (ξ 2 )p − (φ2 )p ]
(64.30)
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]
− vz [uy (ξ 3 )u + wy (ξ 3 )w + (ξ 3 )y ] − vt (ξ 4 )y
+ (φ3 )z + uz (φ3 )u + vz (φ3 )v − (ux + vy )(φ3 )w
− (ux + vy )2 (ξ 3 )w
+ [wt + uwx + vwy − (ux + vy )w]×[wy (ξ 2 )p − (φ3 )p ]
+ (ux + vy )[uz (ξ 3 )u + (ξ 3 )z ] − wt (ξ 4 )z = 0
From the above equation we see that the coefficients of uz wx and vy uz are
−(ξ 3 )w and (ξ 3 )u respectively. We also find that the coefficients of uy vx ,
uy wx and vx wy are −(ξ 2 )v , −(ξ 2 )w and −(ξ 1 )w respectively. Furthermore
the coefficients of vx uy , ux vx , wt wy and wy uz are −(ξ 1 )u , −(ξ 1 )v , (ξ 2 )p and
−(ξ 2 )u respectively. Equating all these coefficients to zero, we have
(ξ 3 )w = (ξ 3 )u = 0, (ξ 2 )p = (ξ 2 )u = (ξ 2 )v = (ξ 2 )w = 0,
(64.31)
(ξ 1 )u = (ξ 1 )v = (ξ 1 )w = 0
655
We thus see that equation (64.30) takes on the form
(φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w

+ (ut + uux + vuy + wuz )×[−(φ1 )p ]
− ux (ξ 1 )x − uy (ξ 2 )x − uz (ξ 3 )x − ut (ξ 4 )x
+ (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w
+ (vt + uvx + vvy + wvz )×[−(φ2 )p ] (64.32)
− vx (ξ 1 )y − vy (ξ 2 )y − vz (ξ 3 )y − vt (ξ 4 )y
+ (wt + uwx + vwy − (ux + vy )w)×[−(φ3 )p ]
− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z − wt (ξ 4 )z = 0
Before we proceed further, let us put together (64.29) and (64.31):
(ξ 3 )p = (ξ 3 )u = (ξ 3 )v = (ξ 3 )w =0
(ξ 2 )p = (ξ 2 )u = (ξ 2 )v = (ξ 2 )w =0
(64.33)
(ξ 1 )p = (ξ 1 )u = (ξ 1 )v = (ξ 1 )w =0
(ξ 4 )p = (ξ 4 )u = (ξ 4 )v = (ξ 4 )w =0
From (64.32), equating to zero the coefficients, we find that

(φ1 )u − u(φ1 )p − (ξ 1 )x − (φ3 )w + w(φ3 )p + (ξ 3 )z = 0 (64.34)
The coefficient of uy is
−v(φ1 )p − (ξ 2 )x + (φ2 )u = 0 (64.35)
The coefficient of uz is
−w(φ1 )p − (ξ 3 )x + (φ3 )u = 0 (64.36)
−(φ1 )p − (ξ 4 )x = 0 (64.37)
The coefficient of vx is
(φ1 )v − u(φ2 )p − (ξ 1 )y = 0 (64.38)
656
The coefficient of vy is
(φ2 )v − v(φ2 )p − (ξ 2 )y − (φ3 )w + (ξ 3 )z = 0 (64.39)
The coefficient of vz is
−w(φ2 )p − (ξ 3 )y + (φ3 )v = 0 (64.40)
The coefficient of vt is
−(φ2 )p − (ξ 4 )y = 0 (64.41)
The coefficient of wx is
(φ1 )w − u(φ3 )p − (ξ 1 )z = 0 (64.42)
The coefficient of wy is
(φ2 )w − v(φ3 )p − (ξ 2 )z = 0 (64.43)
The coefficient of wt is
−(φ3 )p − (ξ 4 )z = 0 (64.44)
(φ1 )x + (φ2 )y + (φ3 )z = 0 (64.45)
657
P r(1) X[∆(1) ] =
= (φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w − (ut )2 (ξ 4 )u
− ut [vt (ξ 4 )v + wt (ξ 4 )w + (ξ 4 )t ]
+ (φ1 )ux + (φ2 )uy + (φ3 )uz

+ u (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]

4 4 4 4
− ut [ux (ξ )u + vx (ξ )v + wx (ξ )w + (ξ )x ]

+ v (φ1 )y + uy (φ1 )u + vy (φ1 )v + wy (φ1 )w − (uy )2 (ξ 2 )u
− py [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− uy [vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y ]

4 4 4 4
− ut [uy (ξ )u + vy (ξ )v + wy (ξ )w + (ξ )y ]

+ w (φ1 )z + uz (φ1 )u + vz (φ1 )v + wz (φ1 )w − (uz )2 (ξ 3 )u
− pz [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− ux [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− uy [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− uz [vz (ξ 3 )v + wz (ξ 3 )w + (ξ 3 )z ]

− ut [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
658
+ (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w − (px )2 (ξ 1 )p
− px [ux (ξ 1 )u + vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x − (φ4 )p ]
− px py (ξ 2 )p − px pz (ξ 3 )p − px pt (ξ 4 )p
− py [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− pz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− pt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]

(64.46)
wz −→ − (ux + vy )

We thus see that the condition P r(1) X[∆(1) ] = 0 is
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
equivalent to
(φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w − (ut )2 (ξ 4 )u

− ut [vt (ξ 4 )v + wt (ξ 4 )w + (ξ 4 )t ]
+ (φ1 )ux + (φ2 )uy + (φ3 )uz

+ u (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]

659

+ v (φ1 )y + uy (φ1 )u + vy (φ1 )v + wy (φ1 )w − (uy )2 (ξ 2 )u
− uy [vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y ]

4 4 4 4
− ut [uy (ξ )u + vy (ξ )v + wy (ξ )w + (ξ )y ]

+ w (φ1 )z + uz (φ1 )u + vz (φ1 )v − (ux + vy )(φ1 )w − (uz )2 (ξ 3 )u
+ (wt + uwx + vwy − w(ux + vy ))×
− ux [uz (ξ 1 )u + vz (ξ 1 )v − (ux + vy )(ξ 1 )w + (ξ 1 )z ]
− uy [uz (ξ 2 )u + vz (ξ 2 )v − (ux + vy )(ξ 2 )w + (ξ 2 )z ]
− uz [vz (ξ 3 )v − (ux + vy )(ξ 3 )w + (ξ 3 )z ]

4 4 4 4
− ut [uz (ξ )u + vz (ξ )v − (ux + vy )(ξ )w + (ξ )z ]
+ (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w

− (ut + uux + vuy + wuz )2 (ξ 1 )p
×[ux (ξ 1 )u + vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x − (φ4 )p ]
− (ut + uux + vuy + wuz )(vt + uvx + vvy + wvz )(ξ 2 )p
− (ut + uux + vuy + wuz )(wt + uwx + vwy − (ux + vy )w)(ξ 3 )p
+ (ut + uux + vuy + wuz )pt (ξ 4 )p
×[ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
+ (wt + uwx + vwy − (ux + vy )w)×
×[ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− pt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ] = 0
660
At this stage we substitute the values (64.33) into the above equation and
we get
(φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w + pt (φ1 )p
− ux (ξ 1 )t − uy (ξ 2 )t − uz (ξ 3 )t − ut (ξ 4 )t
+ (φ1 )ux + (φ2 )uy + (φ3 )uz

+ u (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w
+ (ut + uux + vuy + wuz )×[−(φ1 )p ]


+ v (φ1 )y + uy (φ1 )u + vy (φ1 )v + wy (φ1 )w
+ (vt + uvx + vvy + wvz )×[−(φ1 )p ]

1 2 3 4
− ux (ξ )y − uy (ξ )y − uz (ξ )y − ut (ξ )y

+ w (φ1 )z + uz (φ1 )u + vz (φ1 )v − (ux + vy )(φ1 )w
+ (wt + uwx + vwy − w(ux + vy ))×[−(φ1 )p ]

− ux (ξ 1 )z − uy (ξ 2 )z − uz (ξ 3 )z − ut (ξ 4 )z
+ (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w

+ (ut + uux + vuy + wuz )×[(ξ 1 )x − (φ4 )p ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )x ]
+ (wt + uwx + vwy − (ux + vy )w)×[(ξ 3 )x ]
− pt [(ξ 4 )x ] = 0
the various functions. We find that
The coefficient of pt is
(φ1 )p − (ξ 4 )x = 0 (64.47)
(φ1 )u − (ξ 4 )t − u(φ1 )p − u(ξ 4 )x − v(ξ 4 )y − w(ξ 4 )z + (ξ 1 )x − (φ4 )p = 0 (64.48)
(φ1 )v − v(φ1 )p + (ξ 2 )x = 0 (64.49)
661
(φ1 )w − w(φ1 )p + (ξ 3 )x = 0 (64.50)
− (ξ 1 )t + φ1 + u(φ1 )u − u2 (φ1 )p − u(ξ 1 )x − v(ξ 1 )y − w(φ1 )w
(64.51)
+ w2 (φ1 )p − w(ξ 1 )z + (φ4 )u + u[(ξ 1 )x − (φ4 )p ] − w(ξ 3 )x = 0
− (ξ 2 )t + φ2 − uv(φ1 )p − u(ξ 2 )x + v(φ1 )u − v(ξ 2 )y

(64.52)
− w(ξ 2 )z + v[(ξ 1 )x − (φ4 )p ] = 0
− (ξ 3 )t + φ3 − uw(φ1 )p − u(ξ 3 )x − v(ξ 3 )y + w(φ1 )u
(64.53)
− w(ξ 3 )z + w[(ξ 1 )x − (φ4 )p ] = 0
u(φ1 )v − uv(φ1 )p + (φ4 )v + u(ξ 2 )x = 0 (64.54)
v(φ1 )v − v 2 (φ1 )p − w(φ1 )w + w2 (φ1 )p + v(ξ 2 )x − w(ξ 3 )x = 0 (64.55)
−v(φ1 )p + (φ1 )v + (ξ 2 )x = 0 (64.56)
u(φ1 )w − uw(φ1 )p + (φ4 )w + u(ξ 3 )x = 0 (64.57)
(φ1 )w − w(φ1 )p + (ξ 3 )x = 0 (64.58)
Finally the constant term (terms which do not contain partial derivatives of
the unknown function) is
(φ1 )t + u(φ1 )x + v(φ1 )y + w(φ1 )z + (φ4 )x = 0 (64.59)
662
P r(1) X[∆(2) ] =
= (φ2 )t + ut (φ2 )u + vt (φ2 )v + wt (φ2 )w − (vt )2 (ξ 4 )v
− vt [ut (ξ 4 )u + wt (ξ 4 )w + (ξ 4 )t ]
+ (φ1 )vx + (φ2 )vy + (φ3 )vz

+ u (φ2 )x + ux (φ2 )u + vx (φ2 )v + wx (φ2 )w − (vx )2 (ξ 1 )v
− px [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [ux (ξ 1 )u + wx (ξ 1 )w + (ξ 1 )x ]

4 4 4 4
− vt [ux (ξ )u + vx (ξ )v + wx (ξ )w + (ξ )x ]

+ v (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w − (vy )2 (ξ 2 )v
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]


+ w (φ2 )z + uz (φ2 )u + vz (φ2 )v + wz (φ2 )w − (vz )2 (ξ 3 )v
− pz [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− vx [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− vy [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− vz [uz (ξ 3 )u + wz (ξ 3 )w + (ξ 3 )z ]

4 4 4 4
− vt [uz (ξ )u + vz (ξ )v + wz (ξ )w + (ξ )z ]
663
+ (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w − (py )2 (ξ 2 )p
− px [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− py px (ξ 1 )p − py pz (ξ 3 )p − py pt (ξ 4 )p
− py [uy (ξ 2 )u + vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y − (φ4 )p ]
− pz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− pt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]

(64.60)
wz −→ − (ux + vy )
to account for the conditions ∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0. We

thus see that the condition P r(1) X[∆(2) ] =0
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
is equivalent to
(φ2 )t + ut (φ2 )u + vt (φ2 )v + wt (φ2 )w − (vt )2 (ξ 4 )v

− vt [ut (ξ 4 )u + wt (ξ 4 )w + (ξ 4 )t ]
+ (φ1 )vx + (φ2 )vy + (φ3 )vz

+ u (φ2 )x + ux (φ2 )u + vx (φ2 )v + wx (φ2 )w − (vx )2 (ξ 1 )v
− vx [ux (ξ 1 )u + wx (ξ 1 )w + (ξ 1 )x ]

− vt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
664

+ v (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w − (vy )2 (ξ 2 )v
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]

4 4 4 4
− vt [uy (ξ )u + vy (ξ )v + wy (ξ )w + (ξ )y ]

+ w (φ2 )z + uz (φ2 )u + vz (φ2 )v − (ux + vy )(φ2 )w − (vz )2 (ξ 3 )v
+ (wt + uwx + vwy − (ux + vy )w)×
− vx [uz (ξ 1 )u + vz (ξ 1 )v − (ux + vy )(ξ 1 )w + (ξ 1 )z ]
− vy [uz (ξ 2 )u + vz (ξ 2 )v − (ux + vy )(ξ 2 )w + (ξ 2 )z ]
− vz [uz (ξ 3 )u − (ux + vy )(ξ 3 )w + (ξ 3 )z ]

4 4 4 4
− vt [uz (ξ )u + vz (ξ )v − (ux + vy )(ξ )w + (ξ )z ]
+ (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w − (vt + uvx + vvy + wvz )2 (ξ 2 )p

×[uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− (vt + uvx + vvy + wvz )(ut + uux + vuy + wuz )(ξ 1 )p
− (vt + uvx + vvy + wvz )(wt + uwx + vwy − (ux + vy )w)(ξ 3 )p
+ (vt + uvx + vvy + wvz )pt (ξ 4 )p
×[uy (ξ 2 )u + vy (ξ 2 )v + wy (ξ 2 )w + (ξ 2 )y − (φ4 )p ]
+ (wt + uwx + vwy − (ux + vy )w)×
×[uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− pt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ] = 0
665
In the above equation we substitute the values (64.33) and we get the equa-
tion
− vx (ξ 1 )t − vy (ξ 2 )t − vz (ξ 3 )t − vt (ξ 4 )t
+ (φ1 )vx + (φ2 )vy + (φ3 )vz

+ (ut + uux + vuy + wuz )×[−(φ2 )p ]

1 2 3 4
− vx (ξ )x − vy (ξ )x − vz (ξ )x − vt (ξ )x

+ (vt + uvx + vvy + wvz )×[−(φ2 )p ]



− vx (ξ 1 )z − vy (ξ 2 )z − vz (ξ 3 )z − vt (ξ 4 )z
+ (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w

+ (ut + uux + vuy + wuz )×[(ξ 1 )y ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )y − (φ4 )p ]
+ (wt + uwx + vwy − (ux + vy )w)×[(ξ 3 )y ]
− pt (ξ 4 )y = 0
(φ2 )p − (ξ 4 )y = 0 (64.61)
(φ2 )u − u(φ2 )p + (ξ 1 )y = 0 (64.62)
(φ2 )v − (ξ 4 )t − u(ξ 4 )x − v(φ2 )p − w(ξ 4 )z + (ξ 2 )y − (φ4 )p = 0 (64.63)
666
(φ2 )w − w(φ2 )p + (ξ 3 )y = 0 (64.64)
u(φ2 )u − u2 (φ2 )p − w(φ2 )w + w2 (φ2 )p + u(ξ 1 )y − w(ξ 3 )y = 0 (64.65)
−uv(φ2 )p + v(φ2 )u + v(ξ 1 )y + (φ4 )u = 0 (64.66)
−u(φ2 )p + (φ2 )u + (ξ 1 )y = 0 (64.67)
− (ξ 1 )t + φ1 + u(φ2 )v − u(ξ 1 )x − uv(φ2 )p − v(ξ 1 )y

(64.68)
− w(ξ 1 )z + u[(ξ 2 )y − (φ4 )p ] = 0
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ2 )v − v 2 (φ2 )p − v(ξ 2 )y − w(φ2 )w

(64.69)
+ w2 (φ2 )p − w(ξ 2 )z + (φ4 )v + v[(ξ 2 )y − (φ4 )p ] − w(ξ 3 )y = 0
− (ξ 3 )t + φ3 − u(ξ 3 )x − vw(φ2 )p − v(ξ 3 )y + w(φ2 )v − w(ξ 3 )z

(64.70)
+ w[(ξ 2 )y − (φ4 )p ] = 0
The terms wx and wy have common coefficient given by
(φ2 )w − w(φ2 )p + (ξ 3 )y = 0 (64.71)
The constant term (terms which do not contain partial derivatives of the
unknown function) is
(φ2 )t + u(φ2 )x + v(φ2 )y + w(φ2 )z + (φ4 )y = 0 (64.72)
667
P r(1) X[∆(3) ] =
= (φ3 )t + ut (φ3 )u + vt (φ3 )v + wt (φ3 )w − (wt )2 (ξ 4 )w
− pt [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wz [ut (ξ 3 )u + vt (ξ 3 )v + wt (ξ 3 )w + (ξ 3 )t ]
− wt [ut (ξ 4 )u + vt (ξ 4 )v + (ξ 4 )t ]
+ (φ1 )wx + (φ2 )wy + (φ3 )wz

+ u (φ3 )x + ux (φ3 )u + vx (φ3 )v + wx (φ3 )w − (wx )2 (ξ 1 )w
− px [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [ux (ξ 1 )u + vx (ξ 1 )v + (ξ 1 )x ]
− wz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]

4 4 4 4
− wt [ux (ξ )u + vx (ξ )v + wx (ξ )w + (ξ )x ]

+ v (φ3 )y + uy (φ3 )u + vy (φ3 )v + wy (φ3 )w − (wy )2 (ξ 2 )w
− py [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wy [uy (ξ 2 )u + vy (ξ 2 )v + (ξ 2 )y ]
− wz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]

− wt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]

+ w (φ3 )z + uz (φ3 )u + vz (φ3 )v + wz (φ3 )w − (wz )2 (ξ 3 )w
− wz [uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]

4 4 4 4
− wt [uz (ξ )u + vz (ξ )v + wz (ξ )w + (ξ )z ]
668
+ (φ4 )z + uz (φ4 )u + vz (φ4 )v + wz (φ4 )w − (pz )2 (ξ 3 )p
− px [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− pz px (ξ 1 )p − pz py (ξ 2 )p − pz pt (ξ 4 )p
− py [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− pz [uz (ξ 3 )u + vz (ξ 3 )v + wz (ξ 3 )w + (ξ 3 )z − (φ4 )p ]
− pt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]

(64.73)
wz −→ − (ux + vy )
to account for the conditions ∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0. We

thus see that the condition P r(1) X[∆(3) ] =0
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
is equivalent to
(φ3 )t + ut (φ3 )u + vt (φ3 )v + wt (φ3 )w − (wt )2 (ξ 4 )w

− pt [wx (ξ 1 )p + wy (ξ 2 )p − (ux + vy )(ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
+ (ux + vy )[ut (ξ 3 )u + vt (ξ 3 )v + wt (ξ 3 )w + (ξ 3 )t ]
− wt [ut (ξ 4 )u + vt (ξ 4 )v + (ξ 4 )t ]
+ (φ1 )wx + (φ2 )wy − (ux + vy )(φ3 )

+ u (φ3 )x + ux (φ3 )u + vx (φ3 )v + wx (φ3 )w − (wx )2 (ξ 1 )w
×[wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wx [ux (ξ 1 )u + vx (ξ 1 )v + (ξ 1 )x ]
+ (ux + vy )[ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]

− wt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
669

+ v (φ3 )y + uy (φ3 )u + vy (φ3 )v + wy (φ3 )w − (wy )2 (ξ 2 )w
×[wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− wy [uy (ξ 2 )u + vy (ξ 2 )v + (ξ 2 )y ]
+ (ux + vy )[uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]

4 4 4 4
− wt [uy (ξ )u + vy (ξ )v + wy (ξ )w + (ξ )y ]

+ w (φ3 )z + uz (φ3 )u + vz (φ3 )v − (ux + vy )(φ3 )w − (ux + vy )2 (ξ 3 )w
+ (wt + uwx + vwy − (ux + vy )w)×
+ (ux + vy )[uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]

4 4 4 4
− wt [uz (ξ )u + vz (ξ )v − (ux + vy )(ξ )w + (ξ )z ]

− (wt + uwx + vwy − (ux + vy )w)2 (ξ 3 )p
×[uz (ξ 1 )u + vz (ξ 1 )v − (ux + vy )(ξ 1 )w + (ξ 1 )z ]
− (wt + uwx + vwy − (ux + vy )w)(ut + uux + vuy + wuz )(ξ 1 )p
− (wt + uwx + vwy − (ux + vy )w)(vt + uvx + vvy + wvz )(ξ 2 )p
+ (wt + uwx + vwy − (ux + vy )w)pt (ξ 4 )p
×[uz (ξ 2 )u + vz (ξ 2 )v − (ux + vy )(ξ 2 )w + (ξ 2 )z ]
+ (wt + uwx + vwy − (ux + vy )w)×
×[uz (ξ 3 )u + vz (ξ 3 )v − (ux + vy )(ξ 3 )w + (ξ 3 )z − (φ4 )p ]
− pt [uz (ξ 4 )u + vz (ξ 4 )v − (ux + vy )(ξ 4 )w + (ξ 4 )z ] = 0
670
In the above equation we substitute the values (64.33) and we get the equa-
tion
− wx (ξ 1 )t − wy (ξ 2 )t + (ux + vy )(ξ 3 )t − wt (ξ 4 )t
+ (φ1 )wx + (φ2 )wy − (ux + vy )(φ3 )

+ (ut + uux + vuy + wuz )×[−(φ3 )p ]

1 2 3 4
− wx (ξ )x − wy (ξ )x + (ux + vy )(ξ )x − wt (ξ )x

+ (vt + uvx + vvy + wvz )×[−(φ3 )p ]

− wx (ξ 1 )y − wy (ξ 2 )y + (ux + vy )(ξ 3 )y − wt (ξ 4 )y


− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z − wt (ξ 4 )z

+ (ut + uux + vuy + wuz )×[(ξ 1 )z ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )z ]
+ (wt + uwx + vwy − (ux + vy )w)×[(ξ 3 )z − (φ4 )p ]
− pt (ξ 4 )z = 0
(φ3 )p − (ξ 4 )z = 0 (64.74)
(φ3 )u − u(φ3 )p + (ξ 1 )z = 0 (64.75)
(φ3 )v − v(φ3 )p + (ξ 2 )z = 0 (64.76)
671
(φ3 )w − (ξ 4 )t − w(φ3 )p + (ξ 3 )z − (φ4 )p = 0 (64.77)
(ξ 3 )t − φ3 + u(φ3 )u − u2 (φ3 )p + u(ξ 3 )x + v(ξ 3 )y − w(φ3 )w

(64.78)
+ w2 (φ3 )p + w(ξ 3 )z − (φ4 )w + u(ξ 1 )z − w[(ξ 3 )z − (φ4 )p ] = 0
The terms uy and uz have common coefficient
−u(φ3 )p + (φ3 )u + (ξ 1 )z = 0 (64.79)
(φ3 )v − v(φ3 )p + (ξ 2 )z = 0 (64.80)
(ξ 3 )t − φ3 + u(ξ 3 )x + v(φ3 )v − v 2 (φ3 )p + v(ξ 3 )y − w(φ3 )w

(64.81)
+ w2 (φ3 )p + w(ξ 3 )z − (φ4 )w + v(ξ 2 )z − w[(ξ 3 )z − (φ4 )p ] = 0
−v(φ3 )p + (φ3 )v + (ξ 2 )z = 0 (64.82)
− (ξ 1 )t + φ1 + u(φ3 )w − u(ξ 1 )x − v(ξ 1 )y − uw(φ3 )p

(64.83)
− w(ξ 1 )z + u[(ξ 3 )z − (φ4 )p ] = 0
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ3 )w − v(ξ 2 )y − wv(φ3 )p

(64.84)
− w(ξ 2 )z + v[(ξ 3 )z − (φ4 )p ] = 0
(φ3 )t + u(φ3 )x + v(φ3 )y + w(φ3 )z + (φ4 )z = 0 (64.85)
672
Let us now solve the system of the determining equations in order to find
the infinitesimals.
Solving the systems (64.37) and (64.47), (64.41) and (64.61), (64.44) and
(64.74) (in pairs), we find (φ1 )p = 0 and (ξ 4 )x = 0, (φ2 )p = 0 and (ξ 4 )y = 0,
(φ3 )p = 0 and (ξ 4 )z = 0 respectively. We thus have
(φ1 )p = (φ2 )p = (φ3 )p = 0 (64.86)
and
(ξ 4 )x = (ξ 4 )y = (ξ 4 )z = 0 (64.87)
It is now a good practice to put together all the determining equations taking
into account (64.33), (64.86) and (64.87).
From equation (64.66), written as v{(φ2 )u + (ξ 1 )y } + (φ4 )u = 0 we obtain
that (φ4 )u = 0 due to (64.62). From equation (64.54), written as u{(φ1 )v +
(ξ 2 )x } + (φ4 )v = 0 we obtain that (φ4 )v = 0 due to (64.49). From equation
(64.57), written as u{(φ1 )w + (ξ 3 )x } + (φ4 )w = 0 we obtain that (φ4 )w = 0
due to (64.50). We thus have
(φ4 )u = (φ4 )v = (φ4 )w = 0 (64.88)
It is obvious that in all the above determining equations, once used, we have
to take into account (64.33), (64.86), (64.87) and (64.88).
Let us now find the general form of the functions φ1 , φ2 and φ3 .
For the function φ1 we use equations (64.38), (64.42), (64.49) and (64.50).
Upon differentiation the above named equations with respect to u, v, w in
all the possible ways and taking into account (64.33), we get (φ1 )vv = 0,
(φ1 )ww = 0, (φ1 )uv = 0, (φ1 )uw = 0 and (φ1 )vw = 0. We thus see that φ1 is a
linear function with respect to v and w and thus admits the general form
φ1 = A1 (x, y, z, t, u)+B1 (x, y, z, t)v +C1 (x, y, z, t)w +D1 (x, y, z, t) (64.89)
all the possible ways and taking into account (64.33), we get (φ2 )uu = 0,
(φ2 )ww = 0, (φ2 )uv = 0, (φ2 )uw = 0 and (φ2 )vw = 0. We thus see that φ2 is a
linear function with respect to u and w and thus admits the general form
φ2 = A2 (x, y, z, t)u+B2 (x, y, z, t, v)+C2 (x, y, z, t)w +D2 (x, y, z, t) (64.90)
673
all the possible ways and taking into account (64.33), we get (φ3 )uu = 0,
(φ3 )vv = 0, (φ3 )uv = 0, (φ3 )uw = 0 and (φ3 )vw = 0. We thus see that φ3 is a
linear function with respect to u and v and thus admits the general form
φ3 = A3 (x, y, z, t)u+B3 (x, y, z, t, )v +C3 (x, y, z, t, w)+D3 (x, y, z, t) (64.91)
Using next equation (64.34) and the expressions for φ1 and φ3 , and differen-
tiation with respect to u, and taking into account (64.33), we get (A1 )uu = 0
which means that A1 is a linear function with respect to u, and thus we have
to rewrite the expression (64.89) as
φ1 = A1 (x, y, z, t)u + B1 (x, y, z, t)v + C1 (x, y, z, t)w + D1 (x, y, z, t) (64.92)
Using equation (64.63) and the expression (64.90) for φ2 and differentiation
with respect to v and taking into account equations (64.33) and (64.88), we
get (B2 )vv = 0 which means that B2 is a linear function of v and thus we
have to rewrite the expression (64.90) as
Using equation (64.77) and the expression (64.91) for φ3 and differentiation
with respect to w and taking into account equations (64.33) and (64.88), we
get (C2 )ww = 0 which means that C2 is a linear function of w and thus we
have to rewrite the expression (64.91) as
Equation (64.59), using the expression (64.92) for φ1 , takes on the form
u{(A1 )t + (D1 )x } + v{(B1 )t + (D1 )y } + w{(C1 )t + (D1 )z }

+ u2 (A1 )x + v 2 (B1 )y + w2 (C1 )z + uv{(B1 )x + (A1 )y } (64.95)
4
+ uw{(C1 )x + (A1 )z } + vw{(C1 )y + (B1 )z } + (D1 )t + (φ )x = 0
Equating the coefficients of u, v, w of the above equation to zero, we obtain

the following systems
(A1 )t + (D1 )x = 0, (B1 )t + (D1 )y = 0, (C1 )t + (D1 )z = 0 (64.96)
(A1 )x = 0, (B1 )y = 0, (C1 )z = 0, (D1 )t + (φ4 )x = 0 (64.97)
674
(B1 )x + (A1 )y = 0, (C1 )x + (A1 )z = 0, (C1 )y + (B1 )z = 0 (64.98)

4
+ uw{(C2 )x + (A2 )z } + vw{(C2 )y + (B2 )z } + (D2 )t + (φ )y = 0

(A2 )t + (D2 )x = 0, (B2 )t + (D2 )y = 0, (C2 )t + (D2 )z = 0 (64.100)
(A2 )x = 0, (B2 )y = 0, (C2 )z = 0, (D2 )t + (φ4 )y = 0 (64.101)

(B2 )x + (A2 )y = 0, (C2 )x + (A2 )z = 0, (C2 )y + (B2 )z = 0 (64.102)

4
+ uw{(C3 )x + (A3 )z } + vw{(C3 )y + (B3 )z } + (D3 )t + (φ )z = 0

(A3 )t + (D3 )x = 0, (B3 )t + (D3 )y = 0, (C3 )t + (D3 )z = 0 (64.104)
(A3 )x = 0, (B3 )y = 0, (C3 )z = 0, (D3 )t + (φ4 )z = 0 (64.105)

(B3 )x + (A3 )y = 0, (C3 )x + (A3 )z = 0, (C3 )y + (B3 )z = 0 (64.106)
From equation (64.45), using the expressions (64.92), (64.93) and (64.94) for
φ1 , φ2 and φ3 respectively, we obtain the equation
u{(A1 )x + (A2 )y + (A3 )z } + v{(B1 )x + (B2 )y + (B3 )z }

(64.107)
+ w{(C1 )x + (C2 )y + (C3 )z } + (D1 )x + (D2 )y + (D3 )z = 0
Equating to zero the coefficients of u, v, w and taking into account that

(A1 )x = 0, (B2 )y = 0 and (C3 )z = 0 we have found previously, we get
the systems
(A2 )y + (A3 )z = 0, (B1 )x + (B3 )z = 0, (C1 )x + (C2 )y = 0 (64.108)
675
and
(D1 )x + (D2 )y + (D3 )z = 0 (64.109)
We now take into account the above found relations (A1 )x = 0, (B1 )y =
0, (C1 )z = 0, (A2 )x = 0, (B2 )y = 0, (C2 )z = 0 and (A3 )x = 0, (B3 )y =
0, (C3 )z = 0 (see (64.97), (64.101) and (64.105) above) to rewrite the in-
finitesimals φ1 , φ2 and φ3 given by (64.92), (64.93) and (64.94) respectively,
into the form
φ1 = A1 (y, z, t)u + B1 (x, z, t)v + C1 (x, y, t)w + D1 (x, y, z, t) (64.110)
We come next to find the general expressions for the coefficients ξ 1 , ξ 2 , ξ 3 .
Equations (64.38) and (64.42), taking into account the expression (64.110)
for φ1 , become
B1 (x, z, t) = (ξ 1 )y and C1 (x, y, t) = (ξ 1 )z
respectively. Upon differentiation of the first with respect to y and the second
with respect to z, we obtain (ξ 1 )yy = 0 and (ξ 1 )zz = 0, which means that ξ 1
is a linear function with respect to y and z and thus admits the general form
ξ 1 = α1 (x, t) + β1 (t)y + γ1 (t)z (64.113)
since ξ 1 does not depend on other variables, due to (64.33).
for φ2 , become
A2 (y, z, t) = (ξ 2 )x and C2 (x, y, t) = (ξ 2 )z
respectively. Upon differentiation of the first with respect to x and the second
with respect to z, we obtain (ξ 2 )xx = 0 and (ξ 2 )zz = 0, which means that ξ 2
is a linear function with respect to x and z and thus admits the general form
ξ 2 = α2 (t)x + β2 (y, t) + γ2 (t)z (64.114)
for φ3 , become
A3 (y, z, t) = (ξ 3 )x and B3 (x, z, t) = (ξ 3 )y
676
respectively. Upon differentiation of the first with respect to x and the second
with respect to y, we obtain (ξ 3 )xx = 0 and (ξ 3 )yy = 0, which means that ξ 3
is a linear function with respect to x and y and thus admits the general form
ξ 3 = α3 (t)x + β3 (t)y + γ3 (z, t) (64.115)

Having found the general expressions for φ1 , φ2 , φ3 and ξ 1 , ξ 2 , ξ 3 it is time
to find more explicit expressions for the various coefficients, to ”shrink” the
dependence of these coefficients on the independent variables.
From equations (64.49) and (64.50), using (64.110) and (64.114), (64.115) we
find
B1 (x, z, t) + α2 (t) = 0, C1 (x, y, t) + α3 (t) = 0
respectively, from which we obtain that
B1 = B1 (t) and B1 (t) + α2 (t) = 0 (64.116)
and
C1 = C1 (t) and C1 (t) + α3 (t) = 0 (64.117)
find
A2 (y, z, t) + β1 (t) = 0, C2 (x, y, t) + β3 (t) = 0
A2 = A2 (t) and A2 (t) + β1 (t) = 0 (64.118)
and
C2 = C2 (t) and C2 (t) + β3 (t) = 0 (64.119)
find
A3 (y, z, t) + γ1 (t) = 0, B3 (x, z, t) + γ2 (t) = 0
677
A3 = A3 (t) and A3 (t) + γ1 (t) = 0 (64.120)
and
B3 = B3 (t) and B3 (t) + γ2 (t) = 0 (64.121)
Taking into account (64.116)-(64.121), we write φ1 , φ2 , φ3 as
φ1 = A1 (y, z, t)u + B1 (t)v + C1 (t)w + D1 (x, y, z, t) (64.122)
φ2 = A2 (t)u + B2 (x, z, t)v + C2 (t)w + D2 (x, y, z, t) (64.123)

φ3 = A3 (t)u + B3 (t)v + C3 (x, y, t)w + D3 (x, y, z, t) (64.124)
using (64.110)-(64.112) and ξ 1 , ξ 2 , ξ 3 as
ξ 1 = α1 (x, t) − A2 (t)y − A3 (t)z (64.125)
ξ 2 = −B1 (t)x + β2 (y, t) − B3 (t)z (64.126)

ξ 3 = −C1 (t)x − C2 (t)y + γ3 (z, t) (64.127)
using (64.113)-(64.115).
Before we proceed to the rest of the equations, we recognize that some of
them are identities and some other ones can be further simplified. For exam-
ple, we see that equation (64.55) is an identity due to (64.49) and (64.50).
Similarly equation (64.65) is an identity due to (64.62) and (64.64). On the
other hand, some equations are identical. For example equation (64.49) is
identical to (64.56), equation (64.50) is identical to (64.58), equation (64.75)
is identical to (64.79) and equation (64.76) is identical to both (64.80) and
(64.82).
The system (64.98) contains three equations. From the first one (B1 )x +
(A1 )y = 0, since B1 = B1 (t) we get (A1 )y = 0 and thus A1 = A1 (z, t). From
the second equation (C1 )x + (A1 )z = 0, since C1 = C1 (t) we get (A1 )z = 0
and thus A1 = A1 (t). The third equation (C1 )y + (B1 )z = 0 is an identity
since both C1 and B1 depend on t only.
The system (64.102) contains three equations. From the first one (B2 )x +
(A2 )y = 0, since A2 = A2 (t) we get (B2 )x = 0 and thus B2 = B2 (z, t). The
second equation (C2 )x +(A2 )z = 0 is an identity since both C2 and A2 depend
on t only. From the third equation (C2 )y + (B2 )z = 0, since C2 = C2 (t) we
678
get (B2 )z = 0 and thus B2 = B2 (t).
The system (64.106) contains three equations. The first one (B3 )x + (A3 )y =
0, is an identity since both B3 and A3 depend on t only. From the second
equation (C3 )x + (A3 )z = 0, since A3 = A3 (t) we get (C3 )x = 0 and thus
C3 = C3 (y, t). From the third equation (C3 )y + (B3 )z = 0, since B3 = B3 (t)
we get (C3 )y = 0 and thus C3 = C3 (t).
Equations (64.108) become identities, due to the t−dependence of the vari-
ous coefficients.
The system of equations (64.96) is a system of PDEs containing the function
D1 . From the first equation we obtain (D1 )x = −A01 (t) and by integration
with respect to x, D1 = −A01 (t)x + δ1 (y, z, t). Taking into account this ex-
pression of D1 , the second equation takes on the form (δ1 )y = −B10 (t) and
by integration with respect to y, we get δ1 = −B10 (t)y + δ2 (z, t). Therefore
D1 = −A01 (t)x − B10 (t)y + δ2 (z, t). The third equation, because of this expres-
sion of D1 , becomes (δ2 )z = −C10 (t) and by integration with respect to z we
get δ2 = −C10 (t)z + F10 (t) where F1 (t) is an arbitrary function of t. The prime
denotes derivative with respect to t. We thus have obtained the following
general expression of the function D1 :
D1 = −A01 (t)x − B10 (t)y − C10 (t)z + F10 (t) (64.128)
Solving the systems (64.100) and (64.104) along the same lines of reasoning
as we did for the system (64.96), we obtain the following expressions for D2
and D3 :
D2 = −A02 (t)x − B20 (t)y − C20 (t)z + F20 (t) (64.129)
D3 = −A03 (t)x − B30 (t)y − C30 (t)z + F30 (t) (64.130)

We are now in a position to write down the expressions for φ1 , φ2 , φ3 taking
into account (64.122)-(64.124), the fact that A1 , B2 and C3 depend on t only
(as we have established previously), and the expressions of D1 , D2 and D3
given by (64.128)-(64.130). We obtain the following expressions
φ1 = A1 (t)u + B1 (t)v + C1 (t)w

(64.131)
− A01 (t)x − B10 (t)y − C10 (t)z + F10 (t)
φ2 = A2 (t)u + B2 (t)v + C2 (t)w

(64.132)
− A02 (t)x − B20 (t)y − C20 (t)z + F20 (t)
679
φ3 = A3 (t)u + B3 (t)v + C3 (t)w
(64.133)
− A03 (t)x − B30 (t)y − C30 (t)z + F30 (t)
We should stress the fact that the various functions Ai (t), Bi (t) and Ci (t)
(i = 1, 2, 3) are not independent each other, as it will be established below.
It will also be proved that they are actually constants.
Let us continue next to the equation (64.78). This equation, because of
(64.75), (64.88) and after canceling opposite terms, is equivalent to
(ξ 3 )t − φ3 + u(ξ 3 )x + v(ξ 3 )y − w(φ3 )w + w(φ4 )p = 0 (64.134)
After substituting the expressions of ξ 3 and φ3 given by (64.127) and (64.133)

respectively, we obtain the equation
− u[A3 (t) + C1 (t)] − v[B3 (t) + C2 (t)] + w[−2C3 (t) + (φ4 )p ]

+ x[−C10 (t) + A03 (t)] + y[−C20 (t) + B30 (t)] (64.135)
+ [(γ3 )t + zC30 (t) − F30 (t)] = 0
following systems
A3 (t) + C1 (t) = 0, −C10 (t) + A03 (t) = 0 (64.136)
B3 (t) + C2 (t) = 0, −C20 (t) + B30 (t) = 0 (64.137)

−2C3 (t) + (φ4 )p = 0 (64.138)
(γ3 )t + zC30 (t) − F30 (t) = 0 (64.139)
The solution of (64.136) and (64.137) gives
A3 = −C1 = a1 and B3 = −C2 = a2 (64.140)
Integrating (64.139) with respect to t, we obtain
γ3 = −C3 (t)z + F3 (t) (64.141)
Integrating (64.138) with respect to p, we obtain
φ4 = 2C3 (t)p + F (x, y, z, t) (64.142)
We have next to determine the function F (x, y, z, t) introduced in the above

expression for φ4 . This can be done by solving the system of equations
680
(64.97), (64.101) and (64.105).
Equation (64.97), using (64.128) and (64.142) can be written as
Fx = A001 (t)x + B100 (t)y + C100 (t)z − F100 (t) (64.143)
and by integration with respect to x, we have

1
F = A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x + η1 (y, z, t) (64.144)
2
We thus have
1
φ4 = 2C3 (t)p+ A001 (t)x2 +B100 (t)xy +C100 (t)xz −F100 (t)x+η1 (y, z, t) (64.145)
2
Equation (64.101), using (64.145), gives
B100 (t)x + (η1 )y = A002 (t)x + B200 (t)y + C200 (t)z − F200 (t) (64.146)
from which we get
B100 (t) = A002 (t) (64.147)
and
(η1 )y = B200 (t)y + C200 (t)z − F200 (t) (64.148)
Integrating the above equation with respect to y we get

1
η1 = B200 (t)y 2 + C200 (t)yz − F200 (t)y + η2 (z, t) (64.149)
2
1
φ4 = 2C3 (t)p + A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x
2 (64.150)
1 00
+ B2 (t)y 2 + C200 (t)yz − F200 (t)y + η2 (z, t)
2
Equation (64.105), because of (64.150), yields
C100 (t)x + C200 (t)y + (η2 )z = A003 (t)x + B300 (t)y + C300 (t)z − F300 (t) (64.151)
681
from which we get
C100 (t) = A003 (t), C200 (t) = B300 (t) (64.152)
and
(η2 )z = C300 (t)z − F300 (t) (64.153)
Integrating the above equation with respect to z, we get

1
η2 = C300 (t)z 2 − F300 (t)z + F4 (t) (64.154)
2
We thus obtain the following expression for F (x, y, z, t)
1 1
F = A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x + B200 (t)y 2
2 2 (64.155)
00 00 1 00 2 00
+ C2 (t)yz − F2 (t)y + C3 (t)z − F3 (t)z + F4 (t)
2
and thus φ4 is given by
1
2 (64.156)
1 00 2 00 00 1 00 2 00
+ B2 (t)y + C2 (t)yz − F2 (t)y + C3 (t)z − F3 (t)z + F4 (t)
2 2
Equation (64.34), taking into account (64.131) and (64.133) for φ1 and φ3
respectively and (64.125) and (64.127) for ξ 1 and ξ 3 respectively, takes on
the form
A1 (t) − (α1 )x − C3 (t) + (γ3 )z = 0
and thus
(α1 )x = A1 (t) − 2C3 (t) (64.157)
where we have used (64.141) for γ3 . Integrating (64.157) with respect to x,

we obtain
α1 (x, t) = {A1 (t) − 2C3 (t)}x + F5 (t) (64.158)
682
Equation (64.39), using (64.132) and (64.133) for φ2 and φ3 respectively and
(64.126) and (64.127) for ξ 2 and ξ 3 respectively, takes on the form
B2 (t) − (β2 )y − C3 (t) + (γ3 )z = 0
and thus
(β2 )y = B2 (t) − 2C3 (t) (64.159)
where we have used (64.141) for γ3 . Integrating (64.159) with respect to y,
we get
β2 (y, t) = {B2 (t) − 2C3 (t)}y + F6 (t) (64.160)
Equation (64.48), taking into account (64.131) for φ1 , (64.125) for ξ 1 and
(64.156) for φ4 , becomes
A1 (t) − (ξ 4 )t + (α1 )x − 2C3 (t) = 0
and thus
(ξ 4 )t = 2{A1 (t) − 2C3 (t)} (64.161)
where we have taken into account (64.158) for α1 . Integrating (64.161) with
respect to t we obtain
Z
4
ξ = 2 {A1 (t) − 2C3 (t)}dt (64.162)
Let us write once more the expressions for ξ 1 , ξ 2 and ξ 3 taking into account
the expressions for α1 (x, t), β2 (y, t) and γ3 (z, t) given by (64.158), (64.160)
and (64.141) respectively. We have, using the previous expressions (64.125)-
(64.127):
ξ 1 = {A1 (t) − 2C3 (t)}x − A2 (t)y − A3 (t)z + F5 (t) (64.163)
ξ 2 = −B1 (t)x + {B2 (t) − 2C3 (t)}y − B3 (t)z + F6 (t) (64.164)
3
ξ = −C1 (t)x − C2 (t)y − C3 (t)z + F3 (t) (64.165)
From equation (64.83), upon substitution the known expressions for φi and
ξ i we get
{B1 (t) + A2 (t)}v + {C1 (t) + A3 (t)}w + 2{−A01 (t) + C30 (t)}x
(64.166)
+ {A02 (t) − B10 (t)}y + {A03 (t) − C10 (t)}z + F10 (t) − F50 (t) = 0
683
Equating to zero the coefficients of v, w, x, y, z we obtain the system of equa-
tions
B1 (t) + A2 (t) = 0, A02 (t) − B10 (t) = 0 (64.167)
C1 (t) + A3 (t) = 0, A03 (t) − C10 (t) = 0 (64.168)

−A01 (t) + C30 (t) = 0, F10 (t) − F50 (t) = 0 (64.169)
From (64.167) and (64.168) we get
B1 = −A2 = a3 and A3 = −C1 = a1 (64.170)
The second of the above relations is compatible to the first of (64.140). Equa-
tions (64.169) are new.
ξ i we get
{B1 (t) + A2 (t)}u + {C2 (t) + B3 (t)}w + {−A02 (t) + B10 (t)}x
(64.171)
+ 2{C30 (t) − B20 (t)}y + {B30 (t) − C20 (t)}z + F20 (t) − F60 (t) = 0
Equating to zero the coefficients of u, w, x, y, z we obtain the system of equa-

tions
B1 (t) + A2 (t) = 0, −A02 (t) + B10 (t) = 0 (64.172)
C2 (t) + B3 (t) = 0, B30 (t) − C20 (t) = 0 (64.173)

C30 (t) − B20 (t) = 0, F20 (t) − F60 (t) = 0 (64.174)
The first two systems are compatible to (64.140) and (64.170). It is only
equations (64.174) which are new.
ξ i we get
{A3 (t) + C1 (t)}u + {B3 (t) + C2 (t)}v + {C10 (t) − A03 (t)}x
(64.175)
+ {C20 (t) − B30 (t)}y = 0
Equating to zero the coefficients of u, v, x, y we obtain the two systems of

equations
A3 (t) + C1 (t) = 0, C10 (t) − A03 (t) = 0
684
and
B3 (t) + C2 (t) = 0, C20 (t) − B30 (t) = 0
The above two systems are compatible to (64.140).
Equation (64.69), after canceling opposite terms, taking into account that
(φ2 )p = 0, (φ4 )v = 0 and equation (64.64), is equivalent to
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ2 )v − w(ξ 2 )z − v(φ4 )p = 0 (64.176)
The above equation, after substitution of ξ 2 , φ2 and φ4 , takes on the form
{A2 (t) + B1 (t)}u + 2{B2 (t) − C3 (t)}v + {C2 (t) + B3 (t)}w
+ {−A02 (t) + B10 (t)}x + 2{−B20 (t) + C30 (t)}y (64.177)
+ {−C20 (t) + B30 (t)}z + F20 (t) − F60 (t) = 0
Before we proceed further, it is worth of mentioning that we have used a
number of equations in order to draw conclusions about the general form of
some infinitesimals. However, we can see next that all these equations are
compatible to the explicit expressions we have found about the coefficients
we have already determined. For example we see that equations (64.36) and
(64.42) are equivalent to C1 (t) + A3 (t) = 0 which is compatible to equation
A3 = −C1 = a1 . Similarly equations (64.40) and (64.43) are equivalent to
C2 (t) + B3 (t) = 0 which is compatible to B3 = −C2 = a2 . Furthermore,
equations (64.35) and (64.38) are equivalent to B1 (t) + A2 (t) = 0 which
is compatible to B1 = −A2 = a3 . On the other hand, we can easily see
that equations (64.49) and (64.56), equations (64.50) and (64.58), equations
(64.62) and (64.67), equations (64.64) and (64.71), equations (64.75) and
(64.79), and finally equations (64.76), (64.80) and (64.82) become identities
since they contain opposite terms which cancel out and they will not be used
any further.
At this stage, in order to simplify the calculations, we shall write down the
expressions for φ1 , · · · , φ4 and ξ 1 , · · · , ξ 4 using the values we have already
found:
A2 = −a3 , A3 = a1 , B1 = a3 , B3 = a2
(64.178)
C1 = −a1 , C2 = −a2
and
F60 = F20 , F50 = F10 (64.179)
685
Therefore, using the above values, we obtain from (64.131)-(64.133) and
(64.156) that
φ1 = A1 (t)u + a3 v − a1 w − A01 (t)x + F10 (t) (64.180)
φ2 = −a3 u + B2 (t)v − a2 w − B20 (t)y + F20 (t) (64.181)

φ3 = a1 u + a2 v + C3 (t)w − C30 (t)z + F30 (t) (64.182)
φ4 = 2C3 (t)p − F100 (t)x − F200 (t)y − F300 (t)z
1 1 1 (64.183)
+ A001 (t)x2 + B200 (t)y 2 + C300 (t)z 2 + F4 (t)
2 2 2
and from (64.163)-(64.165) and (64.162)
ξ 1 = {A1 (t) − 2C3 (t)}x + a3 y − a1 z + F1 (t) (64.184)
ξ 2 = −a3 x + {B2 (t) − 2C3 (t)}y − a2 z + F2 (t) (64.185)

3
ξ = a1 x + a2 y − C3 (t)z + F3 (t) (64.186)
Z
4
ξ = 2 {A1 (t) − 2C3 (t)}dt (64.187)
Let us now continue with the rest of the equations. Equations (64.68) and
(64.69) take on the form
{−A01 (t) + C30 (t)}x + {B2 (t) − C3 (t)}u = 0 (64.188)
and
{B2 (t) − C3 (t)}v − {B20 (t) − C30 (t)}y = 0 (64.189)
respectively. We thus get one more system
B2 (t) − C3 (t) = 0, −A01 (t) + C30 (t) = 0 (64.190)
{−A01 (t) + C30 (t)}x + {A1 (t) − C3 (t)}u = 0 (64.191)
and we thus obtain the equation
A1 (t) − C3 (t) = 0 (64.192)
686
Equations (64.52), (64.53) and (64.70) take on the form
2{A1 (t) − C3 (t)}v − B20 (t)y = 0 (64.193)
2{A1 (t) − C3 (t)}w − C30 (t)z = 0 (64.194)

and
2{B2 (t) − C3 (t)}w − C30 (t)z = 0 (64.195)
respectively. Equating to zero all the coefficients, we obtain from (64.192)-

(64.195) that
B2 = C3 = A1 = a4 (64.196)
The above equality is consistent to the equation (64.109). This can be proved
by taking into account equations (64.128)-(64.130).
We then obtain the following expressions for the infinitesimals, using equa-
tions (64.180)-(64.183) for φ1 , φ2 , φ3 and φ4
φ1 = a4 u + a3 v − a1 w + F10 (t) (64.197)
φ2 = −a3 u + a4 v − a2 w + F20 (t) (64.198)

φ3 = a1 u + a2 v + a4 w + F30 (t) (64.199)
φ4 = 2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t) (64.200)
and (64.184)-(64.187) for ξ 1 , ξ 2 , ξ 3 and ξ 4
ξ 1 = −a4 x + a3 y − a1 z + F1 (t) (64.201)
ξ 2 = −a3 x − a4 y − a2 z + F2 (t) (64.202)

ξ 3 = a1 x + a2 y − a4 z + F3 (t) (64.203)
ξ 4 = −2a4 t + a5 (64.204)
687
The generator X of Lie point symmetries is thus given by
∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂u ∂v ∂w ∂p
∂ ∂
= {−a4 x + a3 y − a1 z + F1 (t)} + {−a3 x − a4 y − a2 z + F2 (t)}
∂x ∂y
∂ ∂
+ {a1 x + a2 y − a4 z + F3 (t)} + {−2a4 t + a5 }
∂z ∂t
∂ ∂
+ {a4 u + a3 v − a1 w + F10 (t)} + {−a3 u + a4 v − a2 w + F20 (t)}
∂u ∂v
∂
+ {a1 u + a2 v + a4 w + F30 (t)}
∂w
∂
+ {2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)}
∂p
(64.205)

∂ ∂ ∂ ∂
X = a1 − z +x −w +u
∂x ∂z ∂u ∂w

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂
+ a2 − z +y −w +v + a3 y −x +v −u
∂y ∂z ∂v ∂w ∂x ∂y ∂u ∂v

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂
+ a4 − x −y −z − 2t + u +v +w + 2p
∂x ∂y ∂z ∂t ∂u ∂v ∂w ∂p

∂ ∂ ∂ ∂ ∂
+ a5 + F4 (t) + F1 (t) + F10 (t) − F100 (t)x
∂t ∂p ∂x ∂u ∂p

∂ 0 ∂ 00 ∂ ∂ 0 ∂ 00 ∂
+ F2 (t) + F2 (t) − F2 (t)y + F3 (t) + F3 (t) − F3 (t)z
∂y ∂v ∂p ∂z ∂w ∂p
(64.206)
We thus see that the generators of the algebra of Lie symmetries of Euler’s
equations are given by
∂ ∂ ∂ ∂
X 1 = −z +x −w +u (64.207)
∂x ∂z ∂u ∂w
688
∂ ∂ ∂ ∂
X 2 = −z +y −w +v (64.208)
∂y ∂z ∂v ∂w
∂ ∂ ∂ ∂
X3 = y −x +v −u (64.209)
∂x ∂y ∂u ∂v
∂ ∂ ∂ ∂
X 4 = −x −y −z − 2t
∂x ∂y ∂z ∂t
(64.210)
∂ ∂ ∂ ∂
+u +v +w + 2p
∂u ∂v ∂w ∂p
∂
X5 = (64.211)
∂t
∂
X 6 = F4 (t) (64.212)
∂p
∂ ∂ ∂
X 7 = F1 (t) + F10 (t) − F100 (t)x (64.213)
∂x ∂u ∂p
∂ ∂ ∂
X 8 = F2 (t) + F20 (t) − F200 (t)y (64.214)
∂y ∂v ∂p
∂ ∂ ∂
X 9 = F3 (t) + F30 (t) − F300 (t)z (64.215)
∂z ∂w ∂p
We shall now try to find some general solutions of Euler’s equations. We
consider the following cases:
CASE I. In this case we consider the linear combination of generators
Γ = X4 + X6 + X7 + X8 + X9 (64.216)

∂ ∂ ∂
Γ = (−x + F1 (t)) + (−y + F2 (t)) + (−z + F3 (t)) +
∂x ∂y ∂z
∂ ∂ ∂
+ (u + F10 (t)) + (v + F20 (t)) + (w + F30 (t)) + (64.217)
∂u ∂v ∂w
∂ ∂
+ {2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)} − 2t
∂p ∂t
Introducing the notation
x̂ = x − F1 (t), ŷ = y − F2 (t), ẑ = z − F3 (t) (64.218)
689
and solving the auxiliary system
dx dy dz
= = (64.219)
−x + F1 (t) −y + F2 (t) −z + F3 (t)
i.e. the system

dx̂ dŷ dẑ
− =− =− (64.220)
x̂ ŷ ẑ
we can introduce two invariants ξ and η defined by
ŷ ẑ
ξ= , η= (64.221)
x̂ x̂
Similarly, introducing the notation
û = u + F10 (t), v̂ = v + F20 (t), ŵ = w + F30 (t) (64.222)

du dv dw dx
0
= 0
= 0
= (64.223)
u + F1 (t) v + F2 (t) w + F3 (t) −x + F1 (t)
i.e. the system

dû dv̂ dŵ dx̂
= = =− (64.224)
û v̂ ŵ x̂
we can introduce the invariants U (ξ, η), V (ξ, η) and W (ξ, η) defined by
1 1 1
û = U (ξ, η), v̂ = V (ξ, η), ŵ = W (ξ, η) (64.225)
x̂ x̂ x̂
Finally solving the equation
dp dx
= (64.226)
2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t) −x + F1 (t)
i.e. the equation

dp̂ dx̂
= −2 (64.227)
p̂ x̂
690
we can introduce one more invariant P (ξ, η) defined by
1
p̂ = P (ξ, η) (64.228)
x̂2
where p̂ is given by
p̂ = 2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t) (64.229)
We are now able to write down the Euler equations in terms of the new
functions U (ξ, η), V (ξ, η), W (ξ, η) and P (ξ, η). They read
U 2 + ξU Uξ + ηU Uη − V Uξ − W Uη + 2P + ξPξ + ηPη = 0 (64.230)
U V + ξU Vξ + ηU Vη − V Vξ − W Vη − Pξ = 0 (64.231)
U W + ξU Wξ + ηU Wη − V Wξ − W Wη − Pη = 0 (64.232)
U + ξUξ + ηUη − Vξ − Wη = 0 (64.233)
We now have to solve the system of equations (64.230)-(64.233).
SOLUTION 1. Introducing the auxiliary variable χ by (Boisvert, Ames
and Srivastava [5])
χ=ξ−η (64.234)
and taking into account that Uξ = Uχ , Uη = −Uχ , etc., the above system
of PDEs can be reduced into the following system of ordinary differential
equations
U 2 + χU Uχ − V Uχ + W Uχ + 2P + χUχ = 0 (64.235)
U V + χU Vχ − V Vχ + W Vχ − Pχ = 0 (64.236)
U W + χU Wχ − V Wχ + W Wχ + Pχ = 0 (64.237)
U + χUχ − Vχ + Wχ = 0 (64.238)
The last of the above equations can be integrated with respect to χ and we
obtain the equation
V − W = χU (64.239)
691
where we have set the integration constant equal to zero.
Substitution of (64.239) into (64.235) results into the equation
U 2 + 2P + χPχ = 0 (64.240)
Substitution of (64.239) into (64.236) and (64.237) and subtracting the re-
sulting equations, we obtain the equation
−χU 2 + 2Pχ = 0 (64.241)
Upon elimination of Pχ between (64.240) and (64.241), we get the equation

1
P = − (2 + χ2 )U 2 (64.242)
4
Differentiation of the above equation with respect to χ and substitution the
resulting equation into (64.241), we obtain the following ordinary differential
equation:
2χU 2 + 2U Uχ + χ2 U Uχ = 0 (64.243)
Once we are able to solve the above equation, we could determine all the other
functions V , W and P . Let us now find the general solution of (64.243). The
general solution of this equation, written as
2χ
Uχ + U =0 (64.244)
2 + χ2
is given by
C
U= (64.245)
2 + χ2
where C is an arbitrary constant.
From (64.237), taking into account (64.239), we get
1
W = − Pχ (64.246)
U
Using the system of equations (64.241) and (64.246), we find
1
W = − χU (64.247)
2
692
Using next (64.239) and (64.247) we get the following expression for V :
1
V = χU (64.248)
2
We thus see that we have been able to express W , V and P in terms of the
function U , the last given by (64.245). Therefore using (64.247), (64.248)
and (64.242), we find the following expressions
χ χ C2 1
W = −C , V =C , P =− (64.249)
2(2 + χ2 ) 2(2 + χ2 ) 4 2 + χ2
The unkown functions can be expressed in terms of the similarity variables
ξ and η as
C C ξ−η
U= 2
, V = ,
2 + (ξ − η) 2 2 + (ξ − η)2
(64.250)
C ξ−η C2 1
W =− 2
, P = −
2 2 + (ξ − η) 4 2 + (ξ − η)2
Theorem 1. The system of equations (64.230)-(64.233) admits the general
solution given by (64.250).
Since (taking into account (64.221) and (64.218))
ŷ − ẑ (y − z) − (F2 (t) − F3 (t))
ξ−η = = (64.251)
x̂ x − F1 (t)
we can express the unkown functions given by (64.250), in terms of the
original variables:
[x − F1 (t)]2
U (ξ, η) = C
2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
C [x − F1 (t)] [(y − z) − (F2 (t) − F3 (t))]
V (ξ, η) =
2 2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
(64.252)
C [x − F1 (t)] [(y − z) − (F2 (t) − F3 (t))]
W (ξ, η) = −
2 2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
C2 [x − F1 (t)]2
P (ξ, η) = −
4 2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
Taking further into account (64.222) and (64.229), we get
u = û − F10 (t), v = v̂ − F20 (t), w = ŵ − F30 (t)
1 (64.253)
p = {p̂ + F100 (t)x + F200 (t)y + F300 (t)z − F4 (t)}
2
693
and then, considering (64.225) and (64.228) with x̂ = x − F1 (t), we have
1 1
u= U (ξ, η) − F10 (t) = U (ξ, η) − F10 (t) (64.254)
x̂ x − F1 (t)
1 1
v= V (ξ, η) − F20 (t) = V (ξ, η) − F20 (t) (64.255)
x̂ x − F1 (t)
1 1
w = W (ξ, η) − F30 (t) = W (ξ, η) − F30 (t) (64.256)
x̂ x − F1 (t)

1 1 00 00 00
p= P (ξ, η) + F1 (t)x + F2 (t)y + F3 (t)z − F4 (t)
2 x̂2
(64.257)
1 1 00 00 00
= P (ξ, η) + F1 (t)x + F2 (t)y + F3 (t)z − F4 (t)
2 [x − F1 (t)]2
Finally, substituting the expressions (64.252) into (64.254)-(64.257), we ob-
tain the following expressions for the original functions u, v, w and p:
x − F1 (t)
u=C 2 2
− F10 (t) (64.258)
2[x − F1 (t)] + [(y − z) − (F2 (t) − F3 (t))]
C (y − z) − (F2 (t) − F3 (t))
v= − F20 (t) (64.259)
2 2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
C (y − z) − (F2 (t) − F3 (t))
w=− − F30 (t) (64.260)
2 2[x − F1 (t)]2 + [(y − z) − (F2 (t) − F3 (t))]2
and
C2

1 1
p= −
2 4 2[x − F1 (t)] + [(y − z) − (F2 (t) − F3 (t))]2
2
(64.261)
00 00 00
+ F1 (t)x + F2 (t)y + F3 (t)z − F4 (t)
Theorem 2. The Euler equations (64.2) admit the general solution given
by (64.258)-(64.261). This solution contains three arbitrary time-dependent
functions F1 (t), F2 (t) and F3 (t) which can be determined from the initial
conditions.
CASE II. We shall find next a general solution to Euler’s equations corre-
sponding to the generator X 4 . We thus consider Lagrange’s auxiliary system
dx dy dz dt du dv dw dp
= = = = = = = (64.262)
−x −y −z −2t u v w 2p
694
dx dt √
Integrating the equation = we find x = C1 t and thus we introduce
−x −2t
the invariant ξ defined by
x
ξ=√ (64.263)
t
dx dt dx dt
Similarly integrating the equations = and = we find the
√ −y √ −2t −z −2t
general solutions y = C2 t and z = C3 t and therefore we introduce the
invariants η and θ defined by
y z
η=√ , θ=√ (64.264)
t t
respectively.
du dt dv dt dw dt
Integrating further the equations = , = , = , and
u −2t v −2t w −2t
dp dt √ √ √
= , we find u t = C4 , v t = C5 , w t = C6 and p t = C7 re-
2p −2t
spectively. We thus introduce the invariants U = U (ξ, η, θ), V = V (ξ, η, θ),
W = W (ξ, η, θ) and P = P (ξ, η, θ) defined by
U V W P
u= √ , v=√ , w=√ , p= (64.265)
t t t t
Using the above substitutions, Euler’s equations take on the form
1 1 ∂U 1 ∂U 1 ∂U
− U− ξ − η − θ
2 2 ∂ξ 2 ∂η 2 ∂θ
(64.266)
∂U ∂U ∂U ∂P
+U +V +W + =0
∂ξ ∂η ∂θ ∂ξ
1 1 ∂V 1 ∂V 1 ∂V
− V − ξ − η − θ
2 2 ∂ξ 2 ∂η 2 ∂θ
(64.267)
∂V ∂V ∂V ∂P
+U +V +W + =0
∂ξ ∂η ∂θ ∂η
1 1 ∂W 1 ∂W 1 ∂W
− W− ξ − η − θ
2 2 ∂ξ 2 ∂η 2 ∂θ
(64.268)
∂W ∂W ∂W ∂P
+U +V +W + =0
∂ξ ∂η ∂θ ∂θ
695
∂U ∂V ∂W
+ + =0 (64.269)
∂ξ ∂η ∂θ
The first observation is that the above system contains terms like
∂U ∂U ∂U ∂V ∂V ∂V
ξ +η +θ , ξ +η +θ and
∂ξ ∂η ∂θ ∂ξ ∂η ∂θ
∂W ∂W ∂W
ξ +η +θ
∂ξ ∂η ∂θ
which might mean that U, V, W would be homogeneous functions according
to Euler’s theorem. We thus consider the first order homogeneous functions
with respect to {ξ, η, θ} (modulo some constants)
U = a1 ξ + b1 η + c1 θ + k1
V = a2 ξ + b2 η + c2 θ + k2 (64.270)
W = a3 ξ + b3 η + c3 θ + k3
where ai , bi , ci , ki (i = 1, 2, 3) are parameters.
a1 + b 2 + c 3 = 0 (64.271)
Substituting c3 by −(a1 + b2 ) in (64.270), we have to consider the choice
U = a1 ξ + b1 η + c1 θ + k1
V = a2 ξ + b2 η + c2 θ + k2 (64.272)
W = a3 ξ + b3 η − (a1 + b2 )θ + k3
and then solve the system of equations (64.266)-(64.268).
Equation (64.266) yields (using (64.272))
∂P
F1 ≡ − (a1 − a21 − a2 b1 − a3 c1 )ξ − (b1 − a1 b1 − b1 b2 − b3 c1 )η
∂ξ
− [c1 − a1 c1 − b1 c2 + (a1 + b2 )c1 ]θ (64.273)

1
+ a1 − k1 + b1 k2 + c1 k3 = 0
2
∂P
F2 ≡ − (b2 − a2 b1 − b22 − b3 c2 )η − (a2 − a1 a2 − a2 b2 − a3 c2 )ξ
∂η
− (c2 − c1 a2 − c2 b2 + (a1 + b2 )c2 )θ (64.274)

1
+ b2 − k2 + k1 a2 + c2 k3 = 0
2
696
∂P
F3 ≡ − (b3 − b1 a3 + b3 a1 )η − (a3 − a2 b3 + a3 b2 )ξ
∂θ
+ [(a1 + b2 )2 + a3 c1 + a1 + b2 + b3 c2 ]θ (64.275)
1
+ k2 b3 − k3 + k1 a3 − k3 (a1 + b2 ) = 0
2
We thus have derived the system of equations (64.273)-(64.275) from which
we can determine the function P .
This system is integrable if the following Jacobi conditions (see for example
Forsyth [12], p. 437) are satisfied:
(F1 , F2 ) + {F1 , F2 } = 0
(F1 , F3 ) + {F1 , F3 } = 0 (64.276)
(F2 , F3 ) + {F2 , F3 } = 0
where
∂F ∂F ∂F1 ∂F2 ∂F1 ∂F2 ∂F1 ∂F2
1 2
(F1 , F2 ) = − + −
∂ξ ∂Pξ ∂Pξ ∂ξ ∂η ∂Pη ∂Pη ∂η
∂F ∂F (64.277)
1 2 ∂F1 ∂F2
+ −
∂θ ∂Pθ ∂Pθ ∂θ
∂F ∂F ∂F1 ∂F3 ∂F1 ∂F3 ∂F1 ∂F3
1 3
(F1 , F3 ) = − + −
∂F ∂F (64.278)
1 3 ∂F1 ∂F3
+ −
∂F ∂F ∂F2 ∂F3 ∂F2 ∂F3 ∂F2 ∂F3
2 3
(F2 , F3 ) = − + −
∂F ∂F (64.279)
2 3 ∂F2 ∂F3
+ −
and
∂F ∂F ∂F1 ∂F2
1 2
{F1 , F2 } = Pξ −
∂P ∂Pξ ∂Pξ ∂P
∂F ∂F (64.280)
1 2 ∂F1 ∂F2 ∂F ∂F
1 2 ∂F1 ∂F2
+ Pη − + Pθ −
∂P ∂Pη ∂Pη ∂P ∂P ∂Pθ ∂Pθ ∂P
697
∂F ∂F ∂F1 ∂F3
1 3
{F1 , F3 } = Pξ −
∂F ∂F (64.281)
1 3 ∂F1 ∂F3 ∂F ∂F
1 3 ∂F1 ∂F3
+ Pη − + Pθ −
∂F ∂F ∂F2 ∂F3
2 3
{F2 , F3 } = Pξ −
∂F ∂F (64.282)
2 3 ∂F2 ∂F3 ∂F ∂F
2 3 ∂F2 ∂F3
+ Pη − + Pθ −
It is obvious that in our case we have
{F1 , F2 } = {F1 , F3 } = {F2 , F3 } = 0 (64.283)
since F1 , F2 and F3 are all independent of P . Conditions (64.276) then yield

the system of equations
−a2 + a1 a2 + a2 b2 + a3 c2 + b1 − b1 a1 − b2 b1 − b3 c1 = 0 (64.284)
−a3 b2 − a3 + a2 b3 − c2 b1 + c1 b2 + c1 = 0 (64.285)
−b1 a3 + b3 + b3 a1 + c1 a2 − c2 a1 − c2 = 0 (64.286)
The system (64.284)-(64.286) admits many solutions, the simplest one given
by
b 1 = a2 , b3 = c2 , c1 = a3 (64.287)
We have now to redefine the ansatz (64.272) and modify the equations
(64.273)-(64.275) accordingly. We thus obtain
U = a1 ξ + a2 η + a3 θ + k1
V = a2 ξ + b2 η + c2 θ + k2 (64.288)
W = a3 ξ + c2 η − (a1 + b2 )θ + k3
and
∂P
= (a1 − a21 − a22 − a23 )ξ + (a2 − a2 b2 − c2 a3 − a2 a1 )η
∂ξ (64.289)
1
+ (a3 + a3 b2 − a2 c2 )θ + k1 − a1 k1 − a2 k2 − a3 k3
2
698
∂P
= (a2 − a2 b2 − c2 a3 − a2 a1 )ξ + (b2 − a22 − b22 − c22 )η
∂η (64.290)
1
+ (c2 + c2 a1 − a2 a3 )θ + k2 − a2 k1 − b2 k2 − c2 k3
2
∂P
= (a3 + a3 b2 − a2 c2 )ξ + (c2 + c2 a1 − a2 a3 )η
∂θ
− [(a1 + b2 )2 + a1 + b2 + a23 + c22 ]θ (64.291)
1
+ k3 − a3 k1 − c2 k2 + (a1 + b2 )k3
2
It can be checked that, taking into account (64.289)-(64.291), the integrabil-
ity conditions
∂ 2P ∂ 2P ∂ 2P ∂ 2P ∂ 2P ∂ 2P
= , = , = (64.292)
∂ξ∂η ∂η∂ξ ∂ξ∂θ ∂θ∂ξ ∂θ∂η ∂η∂θ
are satisfied.
The unknown function P is then determined by integrating the equation
dP = Pξ dξ + Pη dη + Pθ dθ (64.293)
where Pξ , Pη and Pθ are given by (64.289), (64.290) and (64.291) respectively.

dP = (a1 − a21 − a22 − a23 )ξ dξ + (b2 − a22 − b22 − c22 )η dη

− [(a1 + b2 )2 + a1 + b2 + a23 + c22 ]θ dθ
+ (a2 − a2 b2 − c2 a3 − a2 a1 )(ξdη + ηdξ)
+ (a3 + a3 b2 − a2 c2 )(ξdθ + θdξ)
+ (c2 + c2 a1 − a2 a3 )(ηdθ + θdη)
1 (64.294)
+ k1 − a1 k1 − a2 k2 − a3 k3 dξ
2
1
+ k2 − a2 k1 − b2 k2 − c2 k3 dη
2
1
+ k3 − a3 k1 − c2 k2 + (a1 + b2 )k3 dθ
2
699
Integrating the above equation we obtain the following expression for the
function P :
1 1
P = (a1 − a21 − a22 − a23 )ξ 2 + (b2 − a22 − b22 − c22 )η 2
2 2
1
− [(a1 + b2 ) + a1 + b2 + a23 + c22 ]θ2
2
2
+ (a2 − a2 b2 − c2 a3 − a2 a1 )ξη
+ (a3 + a3 b2 − a2 c2 )ξθ
+ (c2 + c2 a1 − a2 a3 )ηθ (64.295)
1
+ k1 − a1 k1 − a2 k2 − a3 k3 ξ
2
1
+ k2 − a2 k1 − b2 k2 − c2 k3 η
2
1
+ k3 − a3 k1 − c2 k2 + (a1 + b2 )k3 θ + k4
2
Theorem 3. The system of equations (64.266)-(64.269) admits the 10−parameter
solution given by (64.288) and (64.295).
We shall express now the solution of Euler’s equations in terms of the original
functions and the original variables. Since, according to (64.263)-(64.265),
x y z
ξ=√ , η=√ , θ=√ (64.296)
t t t
and
U V W P
u= √ , v=√ , w=√ , p= (64.297)
t t t t
the solutions (64.288) and (64.295) can be expressed in terms of the original
functions {u, v, w, p} and the original variables {x, y, z, t} as
u = (a1 x + a2 y + a3 z)t−1 + k1 t−1/2

v = (a2 x + b2 y + c2 z)t−1 + k2 t−1/2 (64.298)
w = [a3 x + c2 y − (a1 + b2 )z]t−1 + k3 t−1/2
700
and
1 1
p = (a1 − a21 − a22 − a23 )x2 t−2 + (b2 − a22 − b22 − c22 )y 2 t−2
2 2
1
− [(a1 + b2 )2 + a1 + b2 + a23 + c22 ]z 2 t−2
2
+ (a2 − a2 b2 − c2 a3 − a2 a1 )xyt−2
+ (a3 + a3 b2 − a2 c2 )xzt−2 + (c2 + c2 a1 − a2 a3 )yzt−2
1 (64.299)
+ k1 − a1 k1 − a2 k2 − a3 k3 xt−3/2
2
1
+ k2 − a2 k1 − b2 k2 − c2 k3 yt−3/2
2
1
+ k3 − a3 k1 − c2 k2 + (a1 + b2 )k3 zt−3/2
2
+ k4 t−1/2
respectively.
Note. The general solution of the system of equations (64.2) is given,
according to Boisvert, Ames and Srivastava (Ref. [5]) by
1 2C
1 C1 5 C
1 1 1
u= − x+ + y+ − + z t−1 + C2 t−1/2
6 3 6 3 3 3 3

5 C1 1 C1 1 2C1 1
v= − x+ + y+ + z t−1 − C2 t−1/2
6 3 6 3 3 3 3
1 1 (64.300)
w = (−x + y − z)t−1 + C2 t−1/2
3 3
1 2 2
p = − (x + y + z )t + (xy − xz + yz)t−2
2 2 −2
3 3
1 −3/2
+ C2 (x − y + z)t + C3 t−1
2
with a different parameterization compared to our own derivation.
A second solution of the system (64.284)-(64.286) is given by
H N M
b1 = , b3 = , c1 = (64.301)
G D D
701
where the quantities H, G, N, M and D are expressed in terms of the various
coefficients and are given explicitly by
H = 1 + a1 − a21 − a31 − 2a22 + 2a21 a22 + a42 − a1 a42 − a23 − a1 a23

+ b2 − 3a21 b2 − 2a31 b2 + 2a1 a22 b2 + 2a21 a22 b2 − a42 b2 − 2a23 b2
− 2a1 a23 b2 − b22 − 3a1 b22 − 3a21 b22 − a31 b22 + 2a22 b22 + 2a1 a22 b22
(64.302)
− a23 b22 − a1 a23 b22 − b32 − 2a1 b32 − a21 b32 + 3a2 a3 c2 + 3a1 a2 a3 c2
− a32 a3 c2 + 3a2 a3 b2 c2 + 3a1 a2 a3 b2 c2 − c22 − 2a1 c22 − a21 c22
− b2 c22 − 2a1 b2 c22 − a21 b2 c22
G = (a2 a3 − c2 − a1 c2 )(−a3 − a3 b2 + a2 c2 ) (64.303)

N = a2 a3 + a2 a3 b 2 − c 2 − a1 c 2 − b 2 c 2 − a1 b 2 c 2
H (64.304)
− (a3 + a3 b2 − a2 c2 )
G
M = −a3 − a1 a3 − a3 b2 − a1 a3 b2 + a2 c2 + a1 a2 c2
H (64.305)
+ (a2 a3 − c2 − a1 c2 )
G
and
D = −1 − a1 + a22 − b2 − a1 b2 (64.306)
respectively.
We have now to redefine the ansatz (64.272) and modify the equations
(64.273)-(64.275) accordingly. We thus obtain
H M
U = a1 ξ + η + θ + k1
G D
V = a2 ξ + b 2 η + c 2 θ + k 2 (64.307)
N
W = a3 ξ + η − (a1 + b2 )θ + k3
D
Working along the same lines of reasoning as before, one can determine the
function P with a different parameterization compared to that given in equa-
tion (64.295).
We shall next perform a symmetry analysis of the system (64.266)-(64.269)
and find some new solutions.
Lie Symmetry Analysis of the system (64.266)-(64.269).
702
Let Γ be the generator of Lie point symmetries for the system (64.266)-
(64.269), where Γ is the vector
∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ3
∂ξ ∂η ∂θ
(64.308)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂U ∂V ∂W ∂P
The functions {ξ 1 , ξ 2 , ξ 3 } and {φ1 , φ2 , φ3 , φ4 } depend on both the variables
{ξ, η, θ} and the functions {U, V, W, P }.
The first prolongation of the vector Γ is denoted by P r(1) Γ and is defined by
∂ ∂ ∂
P r(1) Γ = Γ + φξ + φη + φθ
∂Uξ ∂Uη ∂Uθ
∂ ∂ ∂
+ ψξ + ψη + ψθ
∂Vξ ∂Vη ∂Vθ
(64.309)
∂ ∂ ∂
+ χξ + χη + χθ
∂Wξ ∂Wη ∂Wθ
∂ ∂ ∂
+ πξ + πη + πθ
∂Pξ ∂Pη ∂Pθ
We furthe denote by ∆(1) , ∆(2) , ∆(3) and ∆(4) the quantities

1 1 1 1
∆(1) = − U − ξUξ − ηUη − θUθ
2 2 2 2 (64.310)
+ U Uξ + V Uη + W Uθ + Pξ
1 1 1 1
∆(2) = − V − ξVξ − ηVη − θVθ
2 2 2 2 (64.311)
+ U Vξ + V Vη + W Vθ + Pη
1 1 1 1
∆(3) = − W − ξWξ − ηWη − θWθ
2 2 2 2 (64.312)
+ U Wξ + V Wη + W Wθ + Pθ
and
∆(4) = Uξ + Vη + Wθ (64.313)
respectively.
The Lie symmetries of the system (64.266)-(64.269) can be determined once
703
we know the functions {ξ 1 , ξ 2 , ξ 3 } and {φ1 , φ2 , φ3 , φ4 }.
The Lie symmetries are determined by the conditions
P r(1) Γ[∆(k) ] = 0, k = 1, 2, 3, 4
as long as (64.314)
∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0
Using the definition of P r(1) Γ and the expressions ∆(1) , ∆(2) , ∆(3) , ∆(4) ,
we obtain
1 1 1 1
P r Γ[∆ ] = − Uξ ξ − Uη ξ − Uθ ξ + − + Uξ φ1
(1) (1) 1 2 3
2 2 2 2
1
2 3 ξ
+ Uη φ + Uθ φ + − ξ + U φ (64.315)
2
1 1
+ − η + V φ + − θ + W φθ + π ξ
η
2 2
1 1 1
P r(1) Γ[∆(2) ] = − Vξ ξ 1 − Vη ξ 2 − Vθ ξ 3 + Vξ φ1
2 2 2
1 1
+ − + Vη φ + Vθ φ + − ξ + U ψ ξ
2 3
(64.316)
2 2
1 1
+ − η + V ψ + − θ + W ψθ + πη
η
2 2
1 1 1
P r(1) Γ[∆(3) ] = − Wξ ξ 1 − Wη ξ 2 − Wθ ξ 3 + Wξ φ1
2 2 2
1 1
+ Wη φ2 + − + Wθ φ3 + − ξ + U χξ (64.317)
2 2
1 1
+ − η + V χη + − θ + W χθ + π θ
2 2
and
P r(1) Γ[∆(4) ] = φξ + ψ η + χθ (64.318)
Let us now list explicitly the various coefficients φξ , · · · , π θ which appear in
(64.315)-(64.318). We have
φξ = (φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U
− Pξ [Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ] (64.319)
− Uη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ]
− Uθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
704
φη = (φ1 )η + (φ1 )U Uη + (φ1 )V Vη + (φ1 )W Wη − (Uη )2 (ξ 2 )U
− Pη [Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ] (64.320)
− Uη [Vη (ξ 2 )V + Wη (ξ 2 )W + (ξ 2 )η ]
− Uθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
φθ = (φ1 )θ + (φ1 )U Uθ + (φ1 )V Vθ + (φ1 )W Wθ − (Uθ )2 (ξ 3 )U
− Pθ [Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ] (64.321)
− Uη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ]
− Uθ [Vθ (ξ 3 )V + Wθ (ξ 3 )W + (ξ 3 )θ ]
ψ ξ = (φ2 )ξ + (φ2 )U Uξ + (φ2 )V Vξ + (φ2 )W Wξ − (Vξ )2 (ξ 1 )V
− Pξ [Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uξ (ξ 1 )U + Wξ (ξ 1 )W + (ξ 1 )ξ ] (64.322)
− Vη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ]
− Vθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
ψ η = (φ2 )η + (φ2 )U Uη + (φ2 )V Vη + (φ2 )W Wη − (Vη )2 (ξ 2 )V
− Pη [Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ] (64.323)
− Vη [Vξ (ξ 1 )V + Uη (ξ 2 )U + Wη (ξ 2 )W + (ξ 2 )η ]
− Vθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
ψ θ = (φ2 )θ + (φ2 )U Uθ + (φ2 )V Vθ + (φ2 )W Wθ − (Vθ )2 (ξ 3 )V
− Pθ [Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ] (64.324)
− Vη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ]
− Vθ [Uθ (ξ 3 )U + Wθ (ξ 3 )W + (ξ 3 )θ ]
χξ = (φ3 )ξ + (φ3 )U Uξ + (φ3 )V Vξ + (φ3 )W Wξ − (Wξ )2 (ξ 1 )W
− Pξ [Wξ (ξ 1 )P + Wη (ξ 2 )P + Wθ (ξ 3 )P − (φ3 )P ]
− Wξ [Uξ (ξ 1 )U + Vξ (ξ 1 )V + (ξ 1 )ξ ] (64.325)
− Wη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ]
− Wθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
705
χη = (φ3 )η + (φ3 )U Uη + (φ3 )V Vη + (φ3 )W Wη − (Wη )2 (ξ 2 )W
− Pη [Wξ (ξ 1 )P + Wη (ξ 2 )P + Wθ (ξ 3 )P − (φ3 )P ]
− Wξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ] (64.326)
− Wη [Uξ (ξ 2 )U + Vη (ξ 2 )V + (ξ 2 )η ]
− Wθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
χθ = (φ3 )θ + (φ3 )U Uθ + (φ3 )V Vθ + (φ3 )W Wθ − (Wθ )2 (ξ 3 )W
− Pθ [Wξ (ξ 1 )P + Wη (ξ 2 )P + Wθ (ξ 3 )P − (φ3 )P ]
− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ] (64.327)
− Wη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ]
− Wθ [Wξ (ξ 1 )W + Uθ (ξ 3 )U + Vθ (ξ 3 )V + (ξ 3 )θ ]
π ξ = (φ4 )ξ + (φ4 )U Uξ + (φ4 )V Vξ + (φ4 )W Wξ
− Pξ [Uξ (ξ 1 )U + Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ − (φ4 )P ]
− Pη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ] (64.328)
− Pθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
− Pξ [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
π η = (φ4 )η + (φ4 )U Uη + (φ4 )V Vη + (φ4 )W Wη
− Pξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ]
− Pη [Uη (ξ 2 )U + Vη (ξ 2 )V + Wη (ξ 2 )W + (ξ 2 )η − (φ4 )P ] (64.329)
− Pθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
− Pη [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
π θ = (φ4 )θ + (φ4 )U Uθ + (φ4 )V Vθ + (φ4 )W Wθ
− Pξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ]
− Pη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ] (64.330)
− Pθ [Uθ (ξ 3 )U + Vθ (ξ 3 )V + Wθ (ξ 3 )W + (ξ 3 )θ − (φ4 )P ]
− Pθ [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
Let us consider the condition
P r(1) Γ[∆(4) ] =0 (64.331)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
first. Since
P r(1) Γ[∆(4) ] = φξ + ψ η + χθ
706
we obtain, taking into account (64.319), (64.323) and (64.327) that
P r(1) Γ[∆(4) ]
= (φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U
− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ]
+ (φ2 )η + (φ2 )U Uη + (φ2 )V Vη + (φ2 )W Wη − (Vη )2 (ξ 2 )V
(64.332)
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ]
+ (φ3 )θ + (φ3 )U Uθ + (φ3 )V Vθ + (φ3 )W Wθ − (Wθ )2 (ξ 3 )W
− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ]

1 1 1 1
Pξ −→ U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ (64.333)
2 2 2 2
1 1 1 1
Pη −→ V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ (64.334)
2 2 2 2
1 1 1 1
Pθ −→ W + ξWξ + ηWη + θWθ − U Wξ − V Wη − W Wθ (64.335)
2 2 2 2
and
Wθ −→ − (Uξ + Vη ) (64.336)
to account for the conditions ∆(1) = 0, ∆(2) = 0, ∆(3) = 0 and ∆(4) = 0.

Taking into account the above substitutions, the symmetry condition takes
707
on the form
(φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U

1 1 1 1
− U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ
2 2 2 2
×[Uξ (ξ )P + Uη (ξ )P + Uθ (ξ 3 )P − (φ1 )P ]
1 2
− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ]

1 1 1 1
− V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2
×[Vξ (ξ )P + Vη (ξ )P + Vθ (ξ 3 )P − (φ2 )P ]
1 2
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ]
(64.337)
+ (φ3 )θ + (φ3 )U Uθ + (φ3 )V Vθ − (φ3 )W (Uξ + Vη )
− (Uξ + Vη )2 (ξ 3 )W

1 1 1 1
− W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ
2 2 2 2

− V Wη + W (Uξ + Vη )
×[Wξ (ξ 1 )P + Wη (ξ 2 )P − (Uξ + Vη )(ξ 3 )P − (φ3 )P ]

− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ]
− Wη [Uθ (ξ 2 )U + Vθ (ξ 2 )V − (Uξ + Vη )(ξ 2 )W + (ξ 2 )θ ]
+ (Uξ + Vη ) [Wξ (ξ 1 )W + Uθ (ξ 3 )U + Vθ (ξ 3 )V + (ξ 3 )θ ] = 0
We equate the coefficients of the partial derivatives of the functions U, V, W
to zero. We thus have
The coefficient of Uξ Vξ is −(ξ 1 )V . Therefore
(ξ 1 )V = 0 (64.338)
The coefficients of Uξ Vθ and Uθ Vξ are (ξ 3 )V and −(ξ 3 )V respecively. Therefore
(ξ 3 )V = 0 (64.339)
708
The coefficients of Uθ Vη and Uη Vθ are (ξ 3 )U and −(ξ 3 )U respecively. Therefore
(ξ 3 )U = 0 (64.340)
The terms Uθ Wη and Uη Vη have common coefficient −(ξ 2 )U . Therefore
(ξ 2 )U = 0 (64.341)
The coefficient of Uη Wξ is −(ξ 2 )W . Therefore
(ξ 2 )W = 0 (64.342)
The coefficient of Wη Vξ is −(ξ 1 )W . Therefore
(ξ 1 )W = 0 (64.343)
1
The terms (Wξ )2 and (Vξ )2 have common coefficient U − ξ (ξ 1 )P .
2
Therefore
(ξ 1 )P = 0 (64.344)
1
The terms (Uη )2 and (Wη )2 have common coefficient V − η (ξ 2 )P .
2
Therefore
(ξ 2 )P = 0 (64.345)
1
The terms (Uθ )2 and (Vθ )2 have common coefficient W − θ (ξ 3 )P .
2
Therefore
(ξ 3 )P = 0 (64.346)
The coefficient of Uξ Vη is (2W − θ)(ξ 3 )P − 2(ξ 3 )W and since (ξ 3 )P = 0, we

have
(ξ 3 )W = 0 (64.347)
The coefficient of Uθ Wξ is −(ξ 1 )U − (ξ 3 )W and since (ξ 3 )W = 0, we have
(ξ 1 )U = 0 (64.348)
709
1 1
The coefficient of (Vη )2 is V − η (ξ 2 )P + W − θ (ξ 3 )P − (ξ 2 )V − (ξ 3 )W
2 2
and since we have already found (ξ 2 )P = (ξ 3 )P = (ξ 3 )W = 0, we get
(ξ 2 )V = 0 (64.349)
Let us collect all those preliminary results we have already found:
(ξ 1 )U = (ξ 2 )U = (ξ 3 )U = 0, (ξ 1 )V = (ξ 2 )V = (ξ 3 )V = 0,
(64.350)
(ξ 1 )W = (ξ 2 )W = (ξ 3 )W = 0, (ξ 1 )P = (ξ 2 )P = (ξ 3 )P = 0
We also find that

The coefficient of Uη Vξ is −(ξ 1 )U − (ξ 2 )V . The coefficient of (Uξ )2 is
1 1 1 3
U − ξ (ξ )P + W − θ (ξ )P − (ξ 1 )U − (ξ 3 )W . The coefficient of Vη Wξ
2 1 2 1
is − U − ξ (ξ )P − W − θ (ξ 1 )P + (ξ 1 )W . The coefficient of Uξ Wξ
3
2 2
1 3 1
is − U − ξ (ξ )P − W − θ (ξ 3 )P . The coefficient of Uξ Wη is
2 2
1 3 1 2
− V − η (ξ )P − W − θ (ξ )P + (ξ 3 )W . The coefficient of Uξ Uη is
2 2
1 1 1 2
V − η (ξ )P + U − ξ (ξ )P − (ξ 2 )U . The coefficient of Uξ Uθ is
2 2
1 1
W − θ (ξ )P + U − ξ (ξ 3 )P . The coefficient of Uη Uθ is
1
2 2
1 1
W − θ (ξ )P + V − η (ξ 3 )P .
2
2 2
All these coefficients vanish identically in view of (64.350). Therefore the
710
symmetry condition (64.337) takes the form
1
(φ1 )U − (ξ 1 )ξ − (φ3 )W + (ξ 3 )θ + W − θ (φ3 )P
2

1
− U − ξ (φ1 )P Uξ
2

1 1
1 2
+ (φ )V − (ξ )η − U − ξ (φ )P Vξ
2

1 1
1 3
+ (φ )W − (ξ )θ − U − ξ (φ )P Wξ
2

2 2
1 1
+ (φ )U − (ξ )ξ − V − η (φ )P Uη
2

3 3
1 1 (64.351)
+ (φ )U − (ξ )ξ − W − θ (φ )P Uθ
2
1
+ (φ2 )V − (ξ 2 )η − (φ3 )W + (ξ 3 )θ + W − θ (φ3 )P
2

1
− V − η (φ2 )P Vη
2

2 2
1 3
+ (φ )W − (ξ )θ − V − η (φ )P Wη
2

3 3
1 2
+ (φ )V − (ξ )η − W − θ (φ )P Vθ
2
1 1 1
+ U (φ1 )P + V (φ2 )P + W (φ3 )P + (φ1 )ξ + (φ2 )η + (φ3 )θ = 0
2 2 2
Equating the coefficients of the partial derivatives of the functions U, V, W
to zero, we obtain :
The coefficient of Uξ
1
(φ1 )U − (ξ 1 )ξ − (φ3 )W + (ξ 3 )θ + W − θ (φ3 )P
2 (64.352)
1 1
− U − ξ (φ )P = 0
2
The coefficient of Vξ
1
(φ1 )V − (ξ 1 )η − U − ξ (φ2 )P = 0 (64.353)
2
711
The coefficient of Wξ
1
(φ1 )W − (ξ 1 )θ − U − ξ (φ3 )P = 0 (64.354)
2
The coefficient of Uη
1
(φ2 )U − (ξ 2 )ξ − V − η (φ1 )P = 0 (64.355)
2
The coefficient of Uθ
1
(φ3 )U − (ξ 3 )ξ − W − θ (φ1 )P = 0 (64.356)
2
The coefficient of Vη
2 2 3 3 1 3

(φ )V − (ξ )η − (φ )W + (ξ )θ + W − θ (φ )P
2 (64.357)
1 2
− V − η (φ )P = 0
2
The coefficient of Wη
1
(φ2 )W − (ξ 2 )θ − V − η (φ3 )P = 0 (64.358)
2
The coefficient of Vθ
1
(φ3 )V − (ξ 3 )η − W − θ (φ2 )P = 0 (64.359)
2
The constant term
1 1 1
U (φ1 )P + V (φ2 )P + W (φ3 )P
2 2 2 (64.360)
+ (φ1 )ξ + (φ2 )η + (φ3 )θ = 0
Let us consider next the condition
P r(1) Γ[∆(1) ] =0 (64.361)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
712
We have (considering (64.315))
1 1 1 1
P r(1) Γ[∆(1) ] = − Uξ ξ 1 − Uη ξ 2 − Uθ ξ 3 + − + Uξ φ1
2 2 2 2
1
+ Uη φ2 + Uθ φ3 + − ξ + U φξ (64.362)
2
1 1
+ − η + V φη + − θ + W φθ + π ξ
2 2
We then take into account (64.319)-(64.321) for the φξ , φη , φθ and (64.328)
for π ξ respectively. After that, we make the substitutions (64.333)-(64.336)
Finally we take into account the values (64.350). The invariance condition
(64.361) then becomes
1 1 1 1
− Uξ ξ 1 − Uη ξ 2 − Uθ ξ 3 + − + Uξ φ1
2 2 2 2
2 3
+ Uη φ + Uθ φ
1
+ − ξ + U (φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ
2
1 1 1 1
1
+ (φ )P U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ
2 2 2 2
− Uξ (ξ 1 )ξ − Uη (ξ 2 )ξ − Uθ (ξ 3 )ξ
1
+ − η + V (φ1 )η + (φ1 )U Uη + (φ1 )V Vη + (φ1 )W Wη
2
1 1 1 1
+ (φ1 )P V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2
− Uξ (ξ 1 )η − Uη (ξ 2 )η − Uθ (ξ 3 )η
1
+ − θ + W (φ1 )θ + (φ1 )U Uθ + (φ1 )V Vθ − (φ1 )W (Uξ + Vη )
2
1 1 1 1
+ (φ1 )P W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ
2 2 2 2

1 2 3
− V Wη + W (Uξ + Vη ) − Uξ (ξ )θ − Uη (ξ )θ − Uθ (ξ )θ
713
+ (φ4 )ξ + (φ4 )U Uξ + (φ4 )V Vξ + (φ4 )W Wξ
1 1 1 1
1 4
− [(ξ )ξ − (φ )P ] U + ξUξ + ηUη + θUθ − U Uξ
2 2 2 2
− V Uη − W Uθ
1 1 1 1 (64.363)
− (ξ 2 )ξ V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2
1 1 1 1
− (ξ 3 )ξ W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ
2 2 2 2
− V Wη + W (Uξ + Vη ) = 0
714
Expanding the above equation and collecting terms with respect to the first
derivatives of the functions U, V, W we obtain the equation
1 1 1 1 1
− φ1 − U (ξ 1 )ξ − V (ξ 2 )ξ − W (ξ 3 )ξ + U (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V 2 + W 2 )(φ1 )P − (ξU + ηV + θW )(φ1 )P
2 4
1 1 1 1 1
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ1 )θ + (φ4 )ξ
2 2 2
1 1 1
1
1 4
1 1
+ − ξ + φ + U − ξ [(φ )U − (φ )P ] − V − η (ξ )ξ
2 2 2
1 1 1
− W − θ [(ξ )θ + (ξ 3 )ξ + (φ1 )W ] − (ξ 2 − θ2 )(φ1 )P
2 4
− (U 2 − W 2 )(φ1 )P + (ξU − θW )(φ1 )P + (φ4 )U Uξ

1 1 1
+ − ξ 2 + φ2 − U − ξ (ξ 2 )ξ − W − θ (ξ 2 )θ
2 2 2
1
− V − η [(ξ 1 )ξ − (ξ 2 )η + (φ1 )U − (φ4 )P ]
2
1 1 1 1 1
+ (ηU + ξV )(φ )P − ξη(φ )P − U V (φ )P Uη
2 4

1 1 1
+ − ξ 3 + φ3 − U − ξ (ξ 3 )ξ − V − η (ξ 3 )η
2 2 2
1
+ W − θ [(ξ 1 )ξ − (ξ 3 )θ + (φ1 )U − (φ4 )P ]
2
1 1 1 1 1
+ (ξW + θU )(φ )P − ξθ(φ )P − U W (φ )P Uθ
2 4

1 1 1
+ U − ξ [(ξ 2 )ξ + (φ1 )V ] + (ηU + ξV )(φ1 )P − ξη(φ1 )P
2 2 4

− U V (φ1 )P + (φ4 )V Vξ
715

1 1
+ V − η [(ξ 2 )ξ + (φ1 )V ] − W − θ [(ξ 3 )ξ + (φ1 )W ]
2 2

2 2 1 1 2 2 1 1
− (V − W )(φ )P − (η − θ )(φ )P + (ηV − θW )(φ )P Vη
4

1 1
+ W − θ [(ξ 2 )ξ + (φ1 )V ] + (θV + ηW )(φ1 )P
2 2

1
− ηθ(φ1 )P − V W (φ1 )P Vθ
4

1 1
+ V − η [(ξ 3 )ξ + (φ1 )W ] + (θV + ηW )(φ1 )P
2 2

1
− ηθ(φ1 )P − V W (φ1 )P Wη
4

1 1
+ U − ξ [(ξ 3 )ξ + (φ1 )W ] + (θU + ξW )(φ1 )P
2 2

1 1 1 4
− ξθ(φ )P − U W (φ )P + (φ )W Wξ = 0
4
Equating the coefficients of the derivatives to zero, we obtain the system of
equations :
The constant term
1 1 1 1 1
− φ1 − U (ξ 1 )ξ − V (ξ 2 )ξ − W (ξ 3 )ξ + U (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V + W )(φ )P − (ξU + ηV + θW )(φ1 )P
2 2 1
2 4 (64.364)
1 1 1 1 1 1
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ )θ
2 2 2
+ (φ4 )ξ = 0
1 1 1
1 1 4
1 1
− ξ + φ + U − ξ [(φ )U − (φ )P ] − V − η (ξ )ξ
2 2 2
1 1 1 2 (64.365)
− W − θ [(ξ )θ + (ξ )ξ + (φ )W ] − (ξ − θ )(φ1 )P
3 1 2
2 4
− (U 2 − W 2 )(φ1 )P + (ξU − θW )(φ1 )P + (φ4 )U = 0
716
1 2 2
1 2 1 2
− ξ + φ − U − ξ (ξ )ξ − W − θ (ξ )θ
2 2 2
1 1
− V − η [(ξ )ξ − (ξ )η + (φ )U − (φ4 )P ]
2 1
(64.366)
2
1 1
+ (ηU + ξV )(φ1 )P − ξη(φ1 )P − U V (φ1 )P = 0
2 4
1 3 3
1 3 1 3
− ξ + φ − U − ξ (ξ )ξ − V − η (ξ )η
2 2 2
1 1
+ W − θ [(ξ )ξ − (ξ )θ + (φ )U − (φ4 )P ]
3 1
(64.367)
2
1 1
+ (ξW + θU )(φ1 )P − ξθ(φ1 )P − U W (φ1 )P = 0
2 4
1 1 1
U − ξ [(ξ 2 )ξ + (φ1 )V ] + (ηU + ξV )(φ1 )P − ξη(φ1 )P
2 2 4 (64.368)
1 4
− U V (φ )P + (φ )V = 0
1 1
V − η [(ξ 2 )ξ + (φ1 )V ] − W − θ [(ξ 3 )ξ + (φ1 )W ]
2 2 (64.369)
1
− (V 2 − W 2 )(φ1 )P − (η 2 − θ2 )(φ1 )P + (ηV − θW )(φ1 )P
4
1 1
W − θ [(ξ 2 )ξ + (φ1 )V ] + (θV + ηW )(φ1 )P
2 2 (64.370)
1 1 1
− ηθ(φ )P − V W (φ )P = 0
4
1 1
V − η [(ξ 3 )ξ + (φ1 )W ] + (θV + ηW )(φ1 )P
2 2 (64.371)
1
− ηθ(φ1 )P − V W (φ1 )P = 0
4
717
1 1
U − ξ [(ξ 3 )ξ + (φ1 )W ] + (θU + ξW )(φ1 )P
2 2 (64.372)
1
− ξθ(φ1 )P − U W (φ1 )P + (φ4 )W = 0
4
P r(1) Γ[∆(2) ] =0 (64.373)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0

1 1 1
2 2 2
1 1
+ − + Vη φ + Vθ φ + − ξ + U ψ ξ
2 3
(64.374)
2 2
1 1
+ − η + V ψ + − θ + W ψθ + πη
η
2 2
We then take into account (64.322)-(64.324) for the ψ ξ , ψ η , ψ θ and (64.329)
for π η respectively. After that we make the substitutions (64.333)-(64.336)
718
1 1 1
− Vξ ξ 1 − Vη ξ 2 − Vθ ξ 3 + Vξ φ1
2 2 2
1
+ − + Vη φ + Vθ φ32
2
1
+ − ξ + U (φ2 )ξ + (φ2 )U Uξ + (φ2 )V Vξ + (φ2 )W Wξ
2
1 1 1 1
2
2 2 2 2
− Vξ (ξ 1 )ξ − Vη (ξ 2 )ξ − Vθ (ξ 3 )ξ
1
2
1 1 1 1
2 2 2 2
− Vξ (ξ 1 )η − Vη (ξ 2 )η − Vθ (ξ 3 )η
1
2
1 1 1 1
2 2 2 2
− V Wη + W (Uξ + Vη )

1 2 3
− Vξ (ξ )θ − Vη (ξ )θ − Vθ (ξ )θ
+ (φ4 )η + (φ4 )U Uη + (φ4 )V Vη + (φ4 )W Wη

1 1 1 1
1
− (ξ )η U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ
2 2 2 2
1 1 1 1
2 4
− [(ξ )η − (φ )P ] V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2
1 1 1 1
3
− (ξ )η W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ
2 2 2 2
− V Wη + W (Uξ + Vη )
(64.375)
719
1 1 1 1 1
− φ2 − U (ξ 1 )η − V (ξ 2 )η − W (ξ 3 )η + V (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V 2 + W 2 )(φ2 )P − (ξU + ηV + θW )(φ2 )P
2 4
1 2 1 2 1
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ2 )θ + (φ4 )η
2 2 2
1 1 1
+ − ξ 1 + φ1 − W − θ (ξ 1 )θ − V − η (ξ 1 )η
2 2 2
1 1
− U − ξ [(ξ )ξ − (ξ 2 )η − (φ2 )V + (φ4 )P ]
2
2 1 2 1 2
− U V (φ )P + (ξV + ηU )(φ )P − ξη(φ )P Vξ
2 2

1 1 1
+ − ξ 2 + φ2 − U − ξ (ξ 2 )ξ + V − η [(φ2 )V − (φ4 )P ]
2 2 2
1
− W − θ [(ξ 2 )θ + (ξ 3 )η + (φ2 )W ] − (V 2 − W 2 )(φ2 )P
2
2 1 2 2 2 4
+ (ηV − θW )(φ )P − (η − θ )(φ )P + (φ )V Vη
4

1 1 1
+ − ξ 3 + φ3 − U − ξ (ξ 3 )ξ − V − η (ξ 3 )η
2 2 2
1
+ W − θ [(ξ 2 )η − (ξ 3 )θ + (φ2 )V − (φ4 )P ]
2
2 1 2 1 2
− V W (φ )P + (ηW + θV )(φ )P − ηθ(φ )P Vθ
2 4

1 1
+ U − ξ [(ξ 1 )η + (φ2 )U ] − W − θ [(ξ 3 )η + (φ2 )W ]
2 2

2 2 2 1 2 2 2 2
− (U − W )(φ )P − (ξ − θ )(φ )P + (ξU − θW )(φ )P Uξ
4

1 1
+ V − η [(ξ 1 )η + (φ2 )U ] + (ξV + ηU )(φ2 )P
2 2

2 1 2 4
− U V (φ )P − ξη(φ )P + (φ )U Uη
4
720

1 1
+ W − θ [(ξ 1 )η + (φ2 )U ] + (ξW + θU )(φ2 )P
2 2

1
− ξθ(φ2 )P − U W (φ2 )P Uθ
4

1 3 1
+ U − ξ [(ξ )η + (φ2 )W ] + (ξW + θU )(φ2 )P
2 2
(64.376)
1 2 2
− ξθ(φ )P − U W (φ )P Wξ
4

1 1
+ V − η [(ξ 3 )η + (φ2 )W ] + (ηW + θV )(φ2 )P
2 2

1 2 2 4
− ηθ(φ )P − V W (φ )P + (φ )W Wη = 0
4
Equating to zero the coefficients of the derivatives of the functions U, V, W
we obtain the system of equations :
The constant term
1 1 1 1 1
− φ2 − U (ξ 1 )η − V (ξ 2 )η − W (ξ 3 )η + V (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V 2 + W 2 )(φ2 )P − (ξU + ηV + θW )(φ2 )P
2 4 (64.377)
1 2 1 2 1 2
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ )θ
2 2 2
4
+ (φ )η = 0
1 1 1
− ξ 1 + φ1 − W − θ (ξ 1 )θ − V − η (ξ 1 )η
2 2 2
1 1
− U − ξ [(ξ )ξ − (ξ )η − (φ )V + (φ4 )P ]
2 2
(64.378)
2
1 1
− U V (φ )P + (ξV + ηU )(φ2 )P − ξη(φ2 )P = 0
2
2 2
1 1 1
− ξ 2 + φ2 − U − ξ (ξ 2 )ξ + V − η [(φ2 )V − (φ4 )P ]
2 2 2
1 2
− W − θ [(ξ )θ + (ξ 3 )η + (φ2 )W ] − (V 2 − W 2 )(φ2 )P (64.379)
2
1
+ (ηV − θW )(φ2 )P − (η 2 − θ2 )(φ2 )P + (φ4 )V = 0
4
721
1 3 3
1 3 1 3
− ξ + φ − U − ξ (ξ )ξ − V − η (ξ )η
2 2 2
1 2
+ W − θ [(ξ )η − (ξ )θ + (φ )V − (φ4 )P ]
3 2
(64.380)
2
1 1
2
− V W (φ )P + (ηW + θV )(φ2 )P − ηθ(φ2 )P = 0
2 4
1 1
U − ξ [(ξ 1 )η + (φ2 )U ] − W − θ [(ξ 3 )η + (φ2 )W ]
2 2 (64.381)
2 2 2 1 2 2 2 2
− (U − W )(φ )P − (ξ − θ )(φ )P + (ξU − θW )(φ )P = 0
4
1 1
V − η [(ξ 1 )η + (φ2 )U ] + (ξV + ηU )(φ2 )P
2 2 (64.382)
1
− U V (φ2 )P − ξη(φ2 )P + (φ4 )U = 0
4
1 1
W − θ [(ξ 1 )η + (φ2 )U ] + (ξW + θU )(φ2 )P
2 2 (64.383)
1
− ξθ(φ2 )P − U W (φ2 )P = 0
4
1 3 1
U − ξ [(ξ )η + (φ2 )W ] + (ξW + θU )(φ2 )P
2 2 (64.384)
1 2 2
− ξθ(φ )P − U W (φ )P = 0
4
1 1
V − η [(ξ 3 )η + (φ2 )W ] + (ηW + θV )(φ2 )P
2 2 (64.385)
1
− ηθ(φ2 )P − V W (φ2 )P + (φ4 )W = 0
4
722
We finally consider the condition
P r(1) Γ[∆(3) ] =0 (64.386)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0

1 1 1
2 2 2
1 1
+ Wη φ2 + − + Wθ φ3 + − ξ + U χξ (64.387)
2 2
1 1
+ − η + V χη + − θ + W χθ + π θ
2 2
We then take into account (64.325)-(64.327) for the χξ , χη , χθ and (64.330)
for π θ respectively. After that, we make the substitutions (64.333)-(64.336)
723
1 1 1
− Wξ ξ 1 − Wη ξ 2 + (Uξ + Vη ) ξ 3 + Wξ φ1
2 2 2
1
+ Wη φ + − − (Uξ + Vη ) φ3
2
2
1
2
1 1 1 1
3
2 2 2 2
1 2 3
− Wξ (ξ )ξ − Wη (ξ )ξ + (Uξ + Vη ) (ξ )ξ
1
2
1 1 1 1
2 2 2 2
1 2 3
− Wξ (ξ )η − Wη (ξ )η + (Uξ + Vη ) (ξ )η
1 (64.388)
2
1 1 1 1
2 2 2 2
− V Wη + W (Uξ + Vη )

1 2 3
− Wξ (ξ )θ − Wη (ξ )θ + (Uξ + Vη ) (ξ )θ

1 1 1 1
1
− (ξ )θ U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ
2 2 2 2
1 1 1 1
2
− (ξ )θ V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2
1 1 1 1
3 4
− [(ξ )θ − (φ )P ] W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ
2 2 2 2
− V Wη + W (Uξ + Vη )
724
1 1 1 1 1
− φ3 − U (ξ 1 )θ − V (ξ 2 )θ − W (ξ 3 )θ + W (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V 2 + W 2 )(φ3 )P − (ξU + ηV + θW )(φ3 )P
2 4
1 3 1 3 1
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ3 )θ + (φ4 )θ
2 2 2
1 1 1
+ V − η [(ξ )θ + (φ3 )U ] + (ηU + ξV )(φ3 )P
2 2

1
− ξη(φ3 )P − U V (φ3 )P Uη
4

1 1
+ U − ξ [(ξ 2 )θ + (φ3 )V ] + (ηU + ξV )(φ3 )P
2 2

1
− ξη(φ3 )P − U V (φ3 )P Vξ
4

1 1
+ W − θ [(ξ 1 )θ + (φ3 )U ] + (θU + ξW )(φ3 )P
2 2

1 3 3 4
− ξθ(φ )P − U W (φ )P + (φ )U Uθ
4

1 2 1
+ W − θ [(ξ )θ + (φ3 )V ] + (θV + ηW )(φ3 )P
2 2

1 3 3 4
− ηθ(φ )P − V W (φ )P + (φ )V Vθ
4

1 1 1
+ ξ 3 − φ3 + U − ξ (ξ 3 )ξ − W − θ [(φ3 )W − (φ4 )P ]
2 2 2
1
+ V − η [(ξ 2 )θ + (ξ 3 )η + (φ3 )V ] − (V 2 − W 2 )(φ3 )P
2
1 2 2 3 3 4
− (η − θ )(φ )P + (ηV − θW )(φ )P − (φ )W Vη
4

1 1 1
+ ξ 3 − φ3 + U − ξ [(ξ 1 )θ + (ξ 3 )ξ + (φ3 )U ] + V − η (ξ 3 )η
2 2 2
1
− W − θ [(φ3 )W − (φ4 )P ] − (U 2 − W 2 )(φ3 )P
2
1 2 2 3 3 4
− (ξ − θ )(φ )P + (ξU − θW )(φ )P − (φ )W Uξ
4
725

1 1
+ − ξ 1 + φ1 − U − ξ [(ξ 1 )ξ − (ξ 3 )θ − (φ3 )W + (φ4 )P ]
2 2
1 1
− V − η (ξ 1 )η − W − θ (ξ 1 )θ − U W (φ3 )P
2 2
1 3 1 3
− ξθ(φ )P + (ξW + θU )(φ )P Wξ
4 2

1 2 1 3
+ − ξ + φ + V − η [(ξ )θ + (φ3 )W − (ξ 2 )η − (φ4 )P ]
2
2 2
1 2 1
− U − ξ (ξ )ξ − W − θ (ξ 2 )θ − V W (φ3 )P
2 2
1 3 1 3
− ηθ(φ )P + (ηW + θV )(φ )P Wη = 0
4 2
Equating to zero the coefficients of the derivatives of the functions U, V, W
we obtain the system of equations :
The constant term
1 1 1 1 1
− φ3 − U (ξ 1 )θ − V (ξ 2 )θ − W (ξ 3 )θ + W (φ4 )P
2 2 2 2 2
1 2 1
+ (U + V 2 + W 2 )(φ3 )P − (ξU + ηV + θW )(φ3 )P
2 4 (64.389)
1 3 1 3 1
+ U − ξ (φ )ξ + V − η (φ )η + W − θ (φ3 )θ
2 2 2
4
+ (φ )θ = 0
1 1
V − η [(ξ 1 )θ + (φ3 )U ] + (ηU + ξV )(φ3 )P
2 2 (64.390)
1
− ξη(φ3 )P − U V (φ3 )P = 0
4
1 1
U − ξ [(ξ 2 )θ + (φ3 )V ] + (ηU + ξV )(φ3 )P
2 2 (64.391)
1
− ξη(φ3 )P − U V (φ3 )P = 0
4
726
The coeficient of Uθ
1 1 1
W − θ [(ξ )θ + (φ3 )U ] + (θU + ξW )(φ3 )P
2 2 (64.392)
1
− ξθ(φ )P − U W (φ )P + (φ4 )U = 0
3 3
4
The coeficient of Vθ
1 2 1
W − θ [(ξ )θ + (φ3 )V ] + (θV + ηW )(φ3 )P
2 2 (64.393)
1
− ηθ(φ )P − V W (φ )P + (φ4 )V = 0
3 3
4
The coeficient of Vη
1 3 1 1
ξ − φ3 + U − ξ (ξ 3 )ξ − W − θ [(φ3 )W − (φ4 )P ]
2 2 2
1 2
+ V − η [(ξ )θ + (ξ 3 )η + (φ3 )V ] − (V 2 − W 2 )(φ3 )P (64.394)
2
1 2
− (η − θ2 )(φ3 )P + (ηV − θW )(φ3 )P − (φ4 )W = 0
4
Coefficient of Uξ
1 3 1 1
ξ − φ3 + U − ξ [(ξ 1 )θ + (ξ 3 )ξ + (φ3 )U ] + V − η (ξ 3 )η
2 2 2
1 3
− W − θ [(φ )W − (φ4 )P ] − (U 2 − W 2 )(φ3 )P (64.395)
2
1
− (ξ 2 − θ2 )(φ3 )P + (ξU − θW )(φ3 )P − (φ4 )W = 0
4
Coefficient of Wξ
1 1 1 1
− ξ + φ − U − ξ [(ξ )ξ − (ξ 3 )θ − (φ3 )W + (φ4 )P ]
1
2 2
1 1 1 1
− V − η (ξ )η − W − θ (ξ )θ − U W (φ3 )P (64.396)
2 2
1 1
− ξθ(φ )P + (ξW + θU )(φ3 )P = 0
3
4 2
727
Coefficient of Wη
1 1
− ξ 2 + φ2 + V − η [(ξ 3 )θ + (φ3 )W − (ξ 2 )η − (φ4 )P ]
2 2
1 2 1
− U − ξ (ξ )ξ − W − θ (ξ 2 )θ − V W (φ3 )P (64.397)
2 2
1 1
− ηθ(φ3 )P + (ηW + θV )(φ3 )P = 0
4 2
The solution of the system of the determining equations is given by
ξ 1 = a5 η + a6 θ + a1 , ξ 2 = −a5 ξ + a7 θ + a2 ,
(64.398)
ξ 3 = −a6 ξ − a7 η + a3
1 1
φ1 = a5 V + a6 W + a1 , φ2 = −a5 U + a7 W + a2 ,
2 2 (64.399)
3 1
φ = −a6 U − a7 V + a3
2
and
1 1 1 1
φ4 = a1 ξ + a2 η + a3 θ + a4 (64.400)
4 4 4 4
The generator of the symmetries is then given by (see (64.308))
∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ3
∂ξ ∂η ∂θ
(64.401)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂U ∂V ∂W ∂P
and taking into account (64.398)-(64.400),
∂ ∂
Γ = (a5 η + a6 θ + a1 ) + (−a5 ξ + a7 θ + a2 )
∂ξ ∂η
∂ 1 ∂
+ (−a6 ξ − a7 η + a3 ) + a5 V + a6 W + a1
∂θ 2 ∂U (64.402)
1 ∂ 1 ∂
+ −a5 U + a7 W + a2 + −a6 U − a7 V + a3
2 ∂V 2 ∂W
1 ∂
+ (a1 ξ + a2 η + a3 θ + a4 )
4 ∂P
We shall find some solutions of the system of equations (64.266)-(64.269)
based on the generator (64.402) when a5 = a6 = a7 = 0.
728
First of all, we shall determine the invariants of the system associated to the
symmetries of the equations. The invariants are determined by solving the
system of the invariant surface conditions
ξ 1 Uξ + ξ 2 Uη + ξ 3 Uθ = φ1
ξ 1 Vξ + ξ 2 Vη + ξ 3 Vθ = φ2
(64.403)
ξ 1 Wξ + ξ 2 Wη + ξ 3 Wθ = φ3
ξ 1 Pξ + ξ 2 Pη + ξ 3 Pθ = φ4
i.e. by solving the system (remember a5 = a6 = a7 = 0)

1
a1 Uξ + a2 Uη + a3 Uθ = a1
2
1
a1 Vξ + a2 Vη + a3 Vθ = a2
2 (64.404)
1
a1 Wξ + a2 Wη + a3 Wθ = a3
2
1
a1 Pξ + a2 Pη + a3 Pθ = (a1 ξ + a2 η + a3 θ + a4 )
4
The solution of the system of equations (64.404) is given by
1 a η − a ξ a θ − a ξ
1 2 1 3
U = ξ+F , (64.405)
2 a1 a1
1 a2 a η − a ξ a θ − a ξ
1 2 1 3
V = ξ+G , (64.406)
2 a1 a1 a1
1 a3 a1 η − a2 ξ a1 θ − a3 ξ
W = ξ+H , (64.407)
2 a1 a1 a1
a21 − a22 − a23 2 1 a2 a3 a4
P = ξ + η+ θ+ ξ
8a21 4 a1 a1 a1
a η − a ξ a θ − a ξ (64.408)
1 2 1 3
+K ,
a1 a1
where F, G, H and K are arbitrary functions.
a2 a3 a4
κ= , λ= , µ= (64.409)
a1 a1 a1
729
and
a1 η − a2 ξ
ζ≡ = η − κξ
a1
(64.410)
a1 θ − a3 ξ
χ≡ = θ − λξ
a1
equations (64.405)-(64.408) take now the form
1
U = ξ + F (ζ, χ) (64.411)
2
1
V = κξ + G(ζ, χ) (64.412)
2
1
W = λξ + H(ζ, χ) (64.413)
2
and
1 1
P = (1 − κ2 − λ2 )ξ 2 + (κη + λθ + µ)ξ + K(ζ, χ) (64.414)
8 4
respectively.
Our next task is to determine the functions F, G, H and K. Substituting the
expressions (64.411)-(64.414) to the system of equations (64.266)-(64.269)
we find a system of equations with unknown functions F, G, H and K. This
system consists of the equations
1
− (ζFζ + χFχ ) − (κFζ + λFχ )F + GFζ + HFχ
2 (64.415)
1
+ (κζ + λχ + µ) − (κKζ + λKχ ) = 0
4
1 1 1
− G + κF − (ζGζ + χGχ ) + GGζ + HGχ
2 2 2 (64.416)
− (κGζ + λGχ )F + Kζ = 0
1 1 1
− H + λF − (ζHζ + χHχ ) + GHζ + HHχ
2 2 2 (64.417)
− (κHζ + λHχ )F + Kχ = 0
and
1
− (κFζ + λFχ ) + Gζ + Hχ = 0 (64.418)
2
730
These equations have been written in such a form as to allow us to perform
some obvious manipulations leading to a solution.
In fact looking at the fourth equation, equation (64.418), we can make the
choice
κ=λ (64.419)
i.e. a2 = a3 and then introduce a new variable σ by
σ =ζ +χ (64.420)
Then every partial derivative of the fuctions F, G, H and K will be converted

into an ordinary derivative with respect to the new variable σ. In this respect,
1
Gσ + Hσ = 2κFσ − (64.421)
2
We are thus led to introduce another function Φ by
Φ=G+H (64.422)
because in this case equation (64.421) can be written as

1
Φσ = 2κFσ − (64.423)
2
which by integration yields
1
Φ = 2κF − σ + A (64.424)
2
Taking into account κ = λ, σ = ζ + χ and Φ = G + H, equations (64.415)-
(64.417) take the form
1 1 1
− σFσ − 2κF Fσ + ΦFσ + κσ + µ − 2κKσ = 0 (64.425)
2 4 4
1 1 1
− G + κF − σGσ + ΦGσ − 2κF Gσ + Kσ = 0 (64.426)
2 2 2
and
1 1 1
− H + κF − σHσ + ΦHσ − 2κF Hσ + Kσ = 0 (64.427)
2 2 2
731
respectively.
Our aim is to find an ODE which will contain only the unknown function Φ.
This ODE can be found by combining equations (64.425)-(64.427).
In fact substituting Φ by its expression given by (64.424) into (64.425) we
obtain the equation
A−σ 1 1µ
Kσ = Fσ + σ + (64.428)
2κ 8 8κ
The above expression can be written in terms of Φ using (64.423), i.e. using
1 1
Fσ = Φσ +
2κ 2
in the form
A−σ κ2 − 1 A + κµ
Kσ = Φσ + σ+ (64.429)
4κ2 8κ 2 8κ2
Adding (64.426) and (64.427) we obtain the equation
1 1 1 1
Kσ = σΦσ − ΦΦσ + κF Φσ + Φ − κF (64.430)
4 2 4 2
making use the definition of the function Φ.
The right hand side of the above equation can be expressed in terms of the
function Φ, using the expression (64.424):
1 1 1
Kσ = (σ − A)Φσ − σ + A (64.431)
2 8 4
We thus have derived two different expressions of Kσ given by (64.429) and
(64.431) respectively. Equating these two expressions, we obtain the first
order ordinary differential equation
2(2κ2 + 1)(σ − A)Φσ − (2κ2 − 1)σ + [(2κ2 − 1)A − κµ] = 0 (64.432)
2κ2 − 1 κµ
Φ(σ) = 2
σ+ ln(σ − A) + B (64.433)
2(2κ + 1) 2(2κ2 + 1)
732
where A, B are constants.
We are now going to express all other functions in terms of the Φ function.
In fact we obtain from (65.542) the function F :
1 1
F = Φ+ σ−A (64.434)
2κ 2
In order to determine the function K, we write equation (64.432) into a more
convenient form as
1 2κ2 − 1 (2κ2 − 1)A − κµ
(σ − A)Φσ = σ − (64.435)
2 4(2κ2 + 1) 4(2κ2 + 1)
Because of the above expression, equation (64.431) can be written as
2κ2 − 1 (2κ2 − 1)A − κµ 1 1

Kσ = 2
σ − 2
− σ+ A
4(2κ + 1) 4(2κ + 1) 8 4
or in a more simplified form
2κ2 − 3 2A + κµ
Kσ = σ + (64.436)
8(2κ2 + 1) 4(2κ2 + 1)
Integrating the above equation we obtain
2κ2 − 3 2A + κµ
K(σ) = 2
σ2 + σ+C (64.437)
16(2κ + 1) 4(2κ2 + 1)
Equations (64.426) and (64.427), substituting F and Kσ by (64.434) and
(64.431) respectively, take the new form
1 1
(σ − A)Gσ + G − Φ
2 4
(64.438)
2κ2 − 1 (2κ2 − 1)A − κµ
− σ+ =0
4(2κ2 + 1) 4(2κ2 + 1)
and
1 1
(σ − A)Hσ + H − Φ
2 4
(64.439)
2κ2 − 1 (2κ2 − 1)A − κµ
− σ+ =0
4(2κ2 + 1) 4(2κ2 + 1)
733
respectively. There is no need to try to solve the last two equations, since it
1
is a simple matter to verify that both of them are satisfied by Φ. Therefore
2
1
G=H= Φ (64.440)
2
Using equation (64.434) and the value of the function Φ given by (64.433),
we find the explicit value of the function F :
2κ2

1 κµ
F = σ+ ln(σ − A) + B − A (64.441)
2κ 2κ2 + 1 2(2κ2 + 1)
Using the expressions (64.411)-(64.414) and the values of the functions F, G, H

and K we have determined, we get the following values (remember λ = κ)
of the functions U, V, W and P :
2κ2

1 1 κµ
U = ξ+ σ+ ln(σ − A) + B − A (64.442)
2 2κ 2κ2 + 1 2(2κ2 + 1)
1 2κ2 − 1

1 κµ
V = κξ + σ+ ln(σ − A) + B (64.443)
2 2 2(2κ2 + 1) 2(2κ2 + 1)
1 2κ2 − 1

1 κµ
W = κξ + σ+ ln(σ − A) + B (64.444)
2 2 2(2κ2 + 1) 2(2κ2 + 1)
and
1 1
P = (1 − 2κ2 )ξ 2 + [κ(η + θ) + µ]ξ
8 4
(64.445)
2κ2 − 3 2A + κµ
+ 2
σ2 + σ+C
16(2κ + 1) 4(2κ2 + 1)
Let us now go back and express the solution in terms of the original variables.
We first recall that
x y z
ξ=√ , η=√ , θ=√ (64.446)
t t t
and
U V W P
u= √ , v=√ , w=√ , p= (64.447)
t t t t
734
On the other hand
a2 a4
κ= , µ= (64.448)
a1 a1
and (with λ = κ)
σ = ζ + χ = (η − κξ) + (θ − κξ) = η + θ − 2κξ
y+z a2 x
= √ −2 √
t a1 t
i.e.
a1 (y + z) − 2a2 x
σ= √ (64.449)
a1 t
Therefore the solutions of the original Euler equations are given by
x a2 a1 (y + z) − 2a2 x
u= + 2 2
·
2t 2a2 + a1 t

a1 a4 a1 (y + z) − 2a2 x
+ 2 2
ln √ −A (64.450)
4(2a2 + a1 ) a1 t

a1 (B − A) −1/2
+ t
2a2
a2 x 2a22 − a21 a1 (y + z) − 2a2 x
v= + 2 2
·
2a1 t 4a1 (2a2 + a1 ) t
(64.451)
a2 a4 a1 (y + z) − 2a2 x B −1/2
+ ln √ −A + t
4(2a22 + a21 ) a1 t 2
a2 x 2a22 − a21 a1 (y + z) − 2a2 x
w= + 2 2
·
2a1 t 4a1 (2a2 + a1 ) t
(64.452)
a2 a4 a1 (y + z) − 2a2 x B −1/2
+ ln √ −A + t
4(2a22 + a21 ) a1 t 2
and
a21 − 2a22 x2 1 a2 x(y + z) a4 x
p= · 2 + + 3/2
8a21 t 4a1 t2 t
2 2
2a2 − 3a1 [a1 (y + z) − 2a2 x]2
+ · (64.453)
16a21 (2a22 + a21 ) t2
2Aa21 + a2 a4 a1 (y + z) − 2a2 x
+ · + C t−1
4a1 (2a22 + a21 ) t3/2
where a1 , a2 , a4 are free parameters and A, B, C arbitrary constants.
735
64.2 Two-Dimensional Euler Equations
The 2−dimensional Euler equations have the form
ut + uux + vuy + px = 0
vt + uvx + vvy + py = 0 (64.454)
ux + vy = 0
The coefficients of the Lie symmetries are given by
ξ 1 = −a4 x + a3 y + F1 (t)
ξ 2 = −a3 x − a4 y + F2 (t) (64.455)
ξ 4 = −2a4 t + a5
and
φ1 = a4 u + a3 v + F10 (t)
φ2 = −a3 u + a4 v + F20 (t) (64.456)
φ4 = 2a4 p − F100 (t)x − F200 (t)y + F4 (t)
The above have been obtained from (64.201)-(64.204) and (64.197)-(64.200)

by ignoring ξ 3 and φ3 , the z−coordinate and the function w.
Renaming the various parameters and functions
a3 → a1 , a4 → a2 , a5 → a3 ,
(64.457)
ξ4 → ξ3, φ4 → φ3 , F4 → F3
expressions (64.455) and (64.456) take on the form
ξ 1 = −a2 x + a1 y + F1 (t)
ξ 2 = −a1 x − a2 y + F2 (t) (64.458)
ξ 3 = −2a2 t + a3
and
φ1 = a2 u + a1 v + F10 (t)
φ2 = −a1 u + a2 v + F20 (t) (64.459)
φ3 = 2a2 p − F100 (t)x − F200 (t)y + F3 (t)
736
respectively. Therefore the 2−dimensional Euler equations admit the sym-
metry generator
∂ ∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ 3 + φ1 + φ2 + φ3 (64.460)
∂x ∂y ∂t ∂u ∂v ∂p
∂ ∂
X = [−a2 x + a1 y + F1 (t)] + [−a1 x − a2 y + F2 (t)]
∂x ∂y
∂ ∂
+ (−2a2 t + a3 ) + [a2 u + a1 v + F10 (t)]
∂t ∂u (64.461)
0 ∂
+ [−a1 u + a2 v + F2 (t)]
∂v
∂
+ [2a2 p − F100 (t)x − F200 (t)y + F3 (t)]
∂p
Writing the previous expression as

∂ ∂ ∂ ∂
X = a1 y −x +v −u
∂x ∂y ∂u ∂v

∂ ∂ ∂ ∂ ∂ ∂
+ a2 − x −y − 2t + u +v + 2p
∂x ∂y ∂t ∂u ∂v ∂p
(64.462)
∂ ∂ ∂ 0 ∂ 00 ∂
+ a3 + F3 (t) + F1 (t) + F1 (t) − F1 (t)x
∂t ∂p ∂x ∂u ∂p

∂ ∂ ∂
+ F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p
we obtain the following set of generators for the 2−dimensional Euler equa-
tions
∂ ∂ ∂ ∂
X1 = y −x +v −u
∂x ∂y ∂u ∂v
∂ ∂ ∂ ∂ ∂ ∂
X 2 = −x −y − 2t + u +v + 2p
∂x ∂y ∂t ∂u ∂v ∂p
∂ ∂
X 3 = , X 4 = F3 (t) (64.463)
∂t ∂p
∂ ∂ ∂
X 5 = F1 (t) + F10 (t) − F100 (t)x
∂x ∂u ∂p
∂ ∂ ∂
X 6 = F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p
737
In solving the 2−dimensional Euler equations we shall need the invariant
surface conditions
ξ 1 ux + ξ 2 uy + ξ 3 ut = φ1
ξ 1 vx + ξ 2 vy + ξ 3 vt = φ2 (64.464)
ξ 1 px + ξ 2 py + ξ 3 pt = φ3
written in expanded form as
[−a2 x + a1 y + F1 (t)]ux + [−a1 x − a2 y + F2 (t)]uy
+ (−2a2 t + a3 )ut = a2 u + a1 v + F10 (t),
[−a2 x + a1 y + F1 (t)]vx + [−a1 x − a2 y + F2 (t)]vy
(64.465)
+ (−2a2 t + a3 )vt = −a1 u + a2 v + F20 (t),
[−a2 x + a1 y + F1 (t)]px + [−a1 x − a2 y + F2 (t)]py
+ (−2a2 t + a3 )pt = 2a2 p − F100 (t)x − F200 (t)y + F3 (t)
Solution I. We consider the case
a1 = a2 = 0, a3 = 1 (64.466)
In this case the system (64.465) takes on the form
F1 (t)ux + F2 (t)uy + ut = F10 (t)
F1 (t)vx + F2 (t)vy + vt = F20 (t) (64.467)
F1 (t)px + F2 (t)py + pt = F3 (t) − F100 (t)x − F200 (t)y
Let us consider the auxiliary system associated to the system (64.467):
dx dy dt du
= = = 0
F1 (t) F2 (t) 1 F1 (t)
(64.468)
dv dp
= 0 =
F2 (t) F3 (t) − F1 (t)x − F200 (t)y
00
The equation
dx dt
= (64.469)
F1 (t) 1
R
after integration yields F1 (t)dt = x + C1 and thus one invariant is given by
Z
ξ = x − F̃1 (t), F̃1 (t) = F1 (t)dt (64.470)
738
The equation
dy dt
= (64.471)
F2 (t) 1
R
after integration yields F2 (t)dt = y + C2 and thus another invariant is given
by
Z
η = y − F̃2 (t), F̃2 (t) = F2 (t)dt (64.472)
The equations
dt du dt dv
= 0 and = 0 (64.473)
1 F1 (t) 1 F2 (t)
after integration yield u = F1 (t) + C3 and v = F2 (t) + C4 respectively. We

thus obtain the invariant functions
U (ξ, η) = u − F1 (t) and V (ξ, η) = v − F2 (t) (64.474)
In integrating the equation

dt dp
= (64.475)
1 F3 (t) − F1 (t)x − F200 (t)y
00
we have to set x = ξ + F̃1 (t) and y = η + F̃2 (t) respectively. Therefore

integration yields
Z Z Z Z
F3 (t)dt − ξ F1 (t)dt − η F2 (t)dt − F100 (t)F̃1 (t)dt
00 00
Z (64.476)
00
− F2 (t)F̃2 (t)dt = p + C5
Let
Z
F̃3 (t) = F3 (t)dt (64.477)
Then since
Z Z
F100 (t)dt = F10 (t), F200 (t)dt = F20 (t) (64.478)
739
and (using integration by parts)
Z Z
F1 (t)F̃1 (t)dt = F1 (t)F̃1 (t) − F10 (t)F1 (t)dt
00 0
1
= F10 (t)F̃1 (t) − F12 (t)
Z Z2 (64.479)
1
= F20 (t)F̃2 (t) − F22 (t)
2

1 2
F̃3 (t) −ξF10 (t)− ηF20 (t)
− F10 (t)F̃1 (t) − F1 (t)
2
(64.480)
0 1 2
− F2 (t)F̃2 (t) − F2 (t) = p + C5
2
Based on the above expression, we are able to introduce another invariant
function P (ξ, η) by
p = P (ξ, η) − ξF10 (t) − ηF20 (t) + H(t) (64.481)
where

1 2 1 2
H(t) = F̃3 (t) − F10 (t)F̃1 (t) 0
− F1 (t) − F2 (t)F̃2 (t) − F2 (t) (64.482)
2 2
Using (64.470) and (64.472) (i.e. substituting ξ and η by x and y respec-
tively), equation (64.481) takes on the form
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t) (64.483)
where

1 2 2
K(t) = F̃3 (t) + F1 (t) + F2 (t) (64.484)
2
Considering the invariants ξ, η and U (ξ, η), V (ξ, η), P (ξ, η), i.e.
x = ξ + F̃1 (t), y = η + F̃2 (t), u = U (ξ, η) + F1 (t),

(64.485)
v = V (ξ, η) + F2 (t), p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t)
740
the 2−dimensional Euler equations (64.454) take the new form
U Uξ + V Uη + Pξ = 0
U Vξ + V Vη + Pη = 0 (64.486)
Uξ + Vη = 0
Symmetry analysis of equations (64.486). Let X be the generator of

symmetries of the equations (64.486):
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (64.487)
∂ξ ∂η ∂U ∂V ∂P
where ξ 1 , ξ 2 and φ1 , φ2 , φ3 are functions which depend on ξ, η, U, V, P . The

first prolongation is given by
∂ ∂ ∂ ∂
P r(1) X = X + U [ξ] + U [η] + V [ξ] + V [η]
∂Uξ ∂Uη ∂Vξ ∂Vη
(64.488)
∂ ∂
+ P [ξ] + P [η]
∂Pξ ∂Pη
Invariance of equations (64.486) is expressed by the conditions
P r(1) X[∆k ] = 0, k = 1, 2, 3 (64.489)

∆1 =0, ∆2 =0, ∆3 =0
where ∆1 , ∆2 , ∆3 are given by
∆1 = U Uξ + V Uη + Pξ
∆2 = U Vξ + V Vη + Pη (64.490)
∆3 = Uξ + Vη
A detailed symmetry analysis of the corresponding Navier-Stokes equations

has been performed in the next section. There is not any need to repeat
here the calculations, since equations (64.486) and (65.605) admit the same
symmetry generators.
The infinitesimal coefficents are given by (see (65.717) and (65.718))
ξ 1 = −a2 ξ + a3 η + a1 , ξ 2 = −a3 ξ − a2 η + a4 (64.491)
φ1 = a2 U + a3 V, φ2 = −a3 U + a2 V, φ3 = 2a2 P + a5 (64.492)
741
The generator
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (64.493)
∂ξ ∂η ∂U ∂V ∂P
of symmetries for the system (64.486) has now the form
∂ ∂
X = (−a2 ξ + a3 η + a1 ) + (−a3 ξ − a2 η + a4 )
∂ξ ∂η
(64.494)
∂ ∂ ∂
+ (a2 U + a3 V ) + (−a3 U + a2 V ) + (2a2 P + a5 )
∂U ∂V ∂P
Writing the above expression as
∂ ∂ ∂ ∂ ∂ ∂
X = a1 + a2 −ξ −η +U +V + 2P
∂ξ ∂ξ ∂η ∂U ∂V ∂P
∂ (64.495)
∂ ∂ ∂ ∂ ∂
+ a3 η −ξ +V −U + a4 + a5
∂ξ ∂η ∂U ∂V ∂η ∂P
one can read off the following generators
∂ ∂ ∂ ∂ ∂ ∂
X1 = , X 2 = −ξ −η +U +V + 2P
∂ξ ∂ξ ∂η ∂U ∂V ∂P
(64.496)
∂ ∂ ∂ ∂ ∂ ∂
X3 = η −ξ +V −U , X4 = , X5 =
∂ξ ∂η ∂U ∂V ∂η ∂P
Let us determine some solutions of the system (64.486) by considering the
ξ 1 Uξ + ξ 2 Uη = φ1
ξ 1 Vξ + ξ 2 Vη = φ2 (64.497)
ξ 1 Pξ + ξ 2 Pη = φ3
(−a2 ξ + a3 η + a1 )Uξ + (−a3 ξ − a2 η + a4 )Uη = a2 U + a3 V
(−a2 ξ + a3 η + a1 )Vξ + (−a3 ξ − a2 η + a4 )Vη = −a3 U + a2 V (64.498)
(−a2 ξ + a3 η + a1 )Pξ + (−a3 ξ − a2 η + a4 )Pη = 2a2 P + a5
I. For a2 = a3 = 0 the system (64.498) becomes
a1 Uξ + a4 Uη = 0
a1 Vξ + a4 Vη = 0 (64.499)
a1 P ξ + a4 P η = a5
742
The general solution of the above system is given by
a η − a ξ a η − a ξ
1 4 1 4
U =F , V =G ,
a1 a1
a η − a ξ (64.500)
a5 1 4
P = ξ+H
a1 a1
where F, G and H are arbitrary functions. At this point we introduce the
following notation
a4 a5 a1 η − a4 ξ
κ= , λ= , θ≡ = η − κξ (64.501)
a1 a1 a1
We will then have
U = F (θ), V = G(θ), P = λξ + H(θ) (64.502)
When the above expressions of the functions U, V and P are substituted in

the system (64.486), we shall derive the following system of equations
− κF F 0 + GF 0 + λ − κH 0 = 0 (64.503)
−κF G0 + GG0 + H 0 = 0 (64.504)

and
−κF 0 + G0 = 0 (64.505)
where the prime denotes differentiation with respect to θ.

Equation (64.505) can be solved with respect to G:
G = κF + C (64.506)
where C is an arbitrary constant. We substitute this value of G into each

one of the equations (64.503) and (64.504) and derive the equations
C 0 λ
H0 = F + (64.507)
κ κ
and
H 0 = −κCF 0 (64.508)
743
respectively. Equating the two different expressions of H 0 , we obtain the
equation
λ
CF 0 + =0 (64.509)
1 + κ2
λ
F (θ) = − θ+A (64.510)
C(1 + κ2 )
where A is an arbitrary constant.
Combining (64.507) and (64.508) (eliminating F 0 ) we obtain the equation
κλ
H0 = (64.511)
1 + κ2
from which, after integration with respect to θ, we obtain
κλ
H(θ) = θ+B (64.512)
1 + κ2
where B is an arbitrary constant. Substitution of (64.510) into (64.506) gives
the value of the function G:
κλ
G(θ) = − θ + κA + C (64.513)
C(1 + κ2 )
The solution of the 2−dimensional Euler equations (64.454) can be expressed
in terms of the original variables using (64.485), written down again for
convenience
x = ξ + F̃1 (t), y = η + F̃2 (t), u = U (ξ, η) + F1 (t),
(64.514)
where K(t) is given in (64.484).

First of all we have
a4
θ = η − κξ = [y − F̃2 (t)] − [x − F̃1 (t)]
a1
and after the necessary operations
(a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]

θ=− (64.515)
a1
744
We have further
u = U (ξ, η) + F1 (t) = F (θ) + F1 (t)
λ
=− θ + A + F1 (t)
C(1 + κ2 ) (64.516)
a1 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
= 2 2
+ F1 (t) + A
C(a1 + a4 ) a1
v = V (ξ, η) + F2 (t) = G(θ) + F2 (t)
κλ
=− θ + κA + C + F2 (t)
C(1 + κ2 )
a4 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)] (64.517)
= 2 2
C(a1 + a4 ) a1
a4
+ F2 (t) + A + C
a1
and
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t)
= λξ + H(θ) − F10 (t)x − F20 (t)y + K(t)
a5 κλ
= [x − F̃1 (t)] + 2
θ + B − F10 (t)x − F20 (t)y + K(t)
a1 1+κ
(64.518)
a5 a4 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
= [x − F̃1 (t)] − 2
a1 a1 + a24 a1

0 0 1 2 2
− F1 (t)x − F2 (t)y + F̃3 (t) + F1 (t) + F2 (t) + B
2
II. For a3 = a5 = 0, the system (64.498) becomes
(−a2 ξ + a1 )Uξ + (−a2 η + a4 )Uη = a2 U
(−a2 ξ + a1 )Vξ + (−a2 η + a4 )Vη = a2 V (64.519)
(−a2 ξ + a1 )Pξ + (−a2 η + a4 )Pη = 2a2 P
1 a η−a
2 4
U= F ,
a1 − a2 ξ a2 (a1 − a2 ξ)
1 a η−a
2 4
V = G , (64.520)
a1 − a2 ξ a2 (a1 − a2 ξ)
1 a η−a
2 4
P = H
(a1 − a2 ξ)2 a2 (a1 − a2 ξ)
745
where F, G and H are arbitrary functions. The above expressions of the
functions U, V and P are substituted into the system (64.486) and we obtain
the equations
a2 F (θF 0 + F ) + GF 0 + a2 θH 0 + 2a2 H = 0 (64.521)
a2 F (θG0 + G) + GG0 + H 0 = 0 (64.522)

and
1 0
θF 0 + F = − G (64.523)
a2
respectively. In the above equations all the functions F, G and H depend on
the variable θ, where we have set
a2 η − a4
θ= (64.524)
a2 (a1 − a2 ξ)
The prime denotes differentiation with respect to θ as well.
Solving (64.525) with respect to G, we get
G = −a2 θF + C (64.525)

Substitution of the function G given by (64.525) into the equations (64.521)
and (64.522), gives the two equations
CF 0 + a2 F 2 + a2 θH 0 + 2a2 H = 0 (64.526)
and
− a2 CθF 0 − a22 θF 2 + H 0 = 0 (64.527)
respectively. Equations (64.525) and (64.527) constitute a system of ordinary

differential equations with two unknown functions.
Eliminating F 0 between these two equations, we obtain the simple equation
(1 + a22 θ2 )H 0 + 2a22 θH = 0 (64.528)
The above equation, written as

d
[(1 + a22 θ2 )H] = 0 (64.529)
dθ
746
can be integrated immediatelly providing the solution
A
H(θ) = (64.530)
1 + a22 θ2

Substitution of H into (64.528), yields the equation
2Aa2
CF 0 + a2 F 2 + =0 (64.531)
(1 + a22 θ2 )2
The above equation, despite its complicated nature, can be solved using Lie
symmetry analysis. We find that the indinitesimal coefficients {ξ 1 , φ1 } are
given by
1 −2a2 θF + C
ξ1 = + θ2 and φ1 = (64.532)
a22 a2
The normal coordinates {r(θ, F ), s(θ, F )} are found by solving the system
1
2 ∂
−2a2 θF + C ∂
2
+ θ r(θ, F ) + r(θ, F ) = 0
a2 ∂θ a2 ∂F
(64.533)
1
2 ∂
−2a2 θF + C ∂
2
+ θ s(θ, F ) + s(θ, F ) = 1
a2 ∂θ a2 ∂F
r(θ, F ) = H1 (F + a22 θ2 F − Ca2 θ)

(64.534)
s(θ, F ) = a2 arctan(a2 θ) + H2 (F + a22 θ2 F − Ca2 θ)
where H1 , H2 are arbitrary functions of the corresponding arguments.

Let us take for simplicity
r(θ, F ) = F + a22 θ2 F − Ca2 θ and s(θ, F ) = a2 arctan(a2 θ) (64.535)
Inverting the above transformation, we get (where {r, s(r)} the new variables)
1
r + C tan s(r)
1 1
a
θ= tan s(r) , F (θ) = 12 (64.536)
a2 a2

1 + tan2 s(r)
a2
747
Under the above transformation, equation (64.531) takes on the form
d
a2 C + (C 2 + 2A + r2 ) s(r) = 0 (64.537)
dr
a2 C r
s(r) = − √ arctan √ +B (64.538)
C 2 + 2A C 2 + 2A
Comparing (64.538) and the second of (64.535), we get
a2 C r
a2 arctan(a2 θ) = − √ arctan √ +B (64.539)
C 2 + 2A C 2 + 2A
where (from the first of (64.535))
r = (1 + a22 θ2 )F − Ca2 θ (64.540)
We find from (64.539) that

√
r C 2 + 2A
arctan √ = [B − a2 arctan(a2 θ)] (64.541)
C 2 + 2A a2 C
and thus
r √C 2 + 2A
√ = tan [B − a2 arctan(a2 θ)] (64.542)
C 2 + 2A a2 C
The above equation, because of (64.540), becomes
(1 + a22 θ2 )F − Ca2 θ √C 2 + 2A
√ = tan [B − a2 arctan(a2 θ)] (64.543)
C 2 + 2A a2 C
Solving the above equation with respect to F , we obtain the following

expression for the solution of equation (64.531):
√ √C 2 + 2A
C 2 + 2A
F (θ) = tan [B − a 2 arctan(a 2 θ)]
1 + a22 θ2 a2 C
(64.544)
θ
+ Ca2
1 + a22 θ2
748
Using the above solution for F and (64.525), we determine the function G:
√ √C 2 + 2A
a2 C 2 + 2A θ
G(θ) = − tan [B − a2 arctan(a2 θ)]
1 + a22 θ2 a2 C
(64.545)
2 θ2
− Ca2 +C
1 + a22 θ2
We shall convert the above expressions for the various functions in terms of
the original variables. First of all we have
ξ = x − F̃1 (t), η = y − F̃2 (t) and

a2 η − a4 a2 [y − F̃2 (t)] − a4 (64.546)
θ≡ =
a2 (a1 − a2 ξ) a2 [a1 − a2 [x − F̃1 (t)]]
We also have
u = U (ξ, η) + F1 (t), v = V (ξ, η) + F2 (t),
(64.547)
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t)
where
1 a η−a
2 4
U (ξ, η) = F
a1 − a2 ξ a2 (a1 − a2 ξ)
1
= F (θ),
a1 − a2 [x − F̃1 (t)]
1 a η−a
2 4
V (ξ, η) = G
a1 − a2 ξ a2 (a1 − a2 ξ)
(64.548)
1
= G(θ),
a1 − a2 [x − F̃1 (t)]
1 a η−a
2 4
P (ξ, η) = 2
H
(a1 − a2 ξ) a2 (a1 − a2 ξ)
1
= H(θ)
(a1 − a2 [x − F̃1 (t)])2
and

1 2 2
K(t) = F̃3 (t) + F1 (t) + F2 (t) (64.549)
2
749
Solution II. We shall find the solution of the 2−dimensional Euler equa-
tions (64.454) considering the generaor X 2 (corresponding to the dilatation
subgroup) given in (64.463), written down here for convenience
∂ ∂ ∂ ∂ ∂ ∂
X 2 = −x −y − 2t + u +v + 2p (64.550)
∂x ∂y ∂t ∂u ∂v ∂p
The auxiliary system associated to the above generator is
dx dy dt du dv dp
= = = = = (64.551)
−x −y −2t u v 2p
Integrating
dx dt dy dt
= and =
−x −2t −y −2t
we obtain
x y
1/2
= C1 and = C2
t t1/2
respectively. We thus introduce the two invariants ξ and η by
ξ = xt−1/2 and η = yt−1/2 (64.552)
respectively. Similarly integrating

dt du dt dv dt dp
= , = , =
−2t u −2t v −2t 2p
we obtain
u v p
= C3 , = C4 , = C5
t−1/2 t−1/2 t−1
We thus introduce three more invariants U, V and P by
u = t−1/2 U (ξ, η), v = t−1/2 V (ξ, η), p = t−1 P (ξ, η) (64.553)
Substituting the above expressions into the Euler equations (64.454), we

derive the equations
1 1 ∂U 1 ∂U ∂U ∂U ∂P
− U− ξ − η +U +V + =0 (64.554)
2 2 ∂ξ 2 ∂η ∂ξ ∂η ∂ξ
750
1 1 ∂V 1 ∂V ∂V ∂V ∂P
− V − ξ − η +U +V + =0 (64.555)
2 2 ∂ξ 2 ∂η ∂ξ ∂η ∂η
and
∂U ∂V
+ =0 (64.556)
∂ξ ∂η
The coefficients {ξ 1 , ξ 2 } and {φ1 , φ2 , φ2 } of the generator
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (64.557)
∂ξ ∂η ∂U ∂V ∂P
of symmetries for the system (64.554)-(64.556) have the form
ξ 1 = a4 η + a1 , ξ 2 = −a4 ξ + a2 (64.558)
and
1 1
φ1 = a4 V + a1 , φ2 = −a4 U + a2 ,
2 2 (64.559)
1
φ3 = (a1 ξ + a2 η + a3 )
4
The above expressions is the 2−dimensional version of (64.398)-(64.400)
which we already know for the 3−dimensional case.
The invariant surface conditions are given by
ξ 1 Uξ + ξ 2 Uη = φ1
ξ 1 Vξ + ξ 2 Vη = φ2 (64.560)
ξ 1 Pξ + ξ 2 Pη = φ3
which can be written in expanded form as

1
(a4 η + a1 )Uξ + (−a4 ξ + a2 )Uη = a4 V + a1
2
1
(a4 η + a1 )Vξ + (−a4 ξ + a2 )Vη = −a4 U + a2 (64.561)
2
1
(a4 η + a1 )Pξ + (−a4 ξ + a2 )Pη = (a1 ξ + a2 η + a3 )
4
751
We shall consider the case a4 = 0. In this case the invariant surface conditions
take the form
1
a1 Uξ + a2 Uη = a1
2
1
a1 Vξ + a2 Vη = a2 (64.562)
2
1
a1 Pξ + a2 Pη = (a1 ξ + a2 η + a3 )
4
1 a η − a ξ
1 2
U (ξ, η) = ξ + F
2 a1
1 a2 a1 η − a2 ξ
V (ξ, η) = ξ+G (64.563)
2 a1 a1
1 a21 − a22 2 1 a2 a3 a η − a ξ
1 2
P (ξ, η) = 2
ξ + η+ ξ+H
2 a1 4 a1 a1 a1
Under the above substitutions, the reduced system of Euler’s equations
become
1 1 1
− θF 0 (θ) − κF (θ)F 0 (θ) + G(θ)F 0 (θ) + κθ + λ
2 4 4 (64.564)
0
− κH (θ) = 0
1 1 1
− θG(θ) − θG0 (θ) + κF (θ) − κF (θ)G0 (θ)
2 2 2 (64.565)
0 0
+ G(θ)G (θ) + H (θ) = 0
and
1
− κF 0 (θ) + G0 (θ) = 0 (64.566)
2
respectively, where we have introduced the notation
a2 a3 a1 η − a2 ξ
κ= , λ= , θ≡ = η − κξ (64.567)
a1 a1 a1
Integrating the third equation we derive the following expression which
expresses G in terms of F :
1
G = κF − θ + C (64.568)
2
752
where C is an arbitrary constant. Substitution of this expression for G into
the first two equations, equation (64.564) and equation (64.565), we obtain
C − θ 0 1λ
H0 = F + +θ (64.569)
κ 4 κ
and
3
H 0 = κ(C + θ)F 0 − θ + C (64.570)
4
respectively.
Equating the two expressions of H 0 , we obtain the equation
1
[ (1 + κ2 )θ + (−1 + κ2 )C ] F 0 − κθ + κC − λ = 0 (64.571)
4
ζ = (1 + κ2 )θ + (−1 + κ2 )C (64.572)
we find that the general solution of equation (64.571) is given by

κ λ 2Cκ3
F (θ) = θ + ln |ζ| − ln |ζ| + A (64.573)
1 + κ2 4(1 + κ2 ) (1 + κ2 )2
Eliminating θF 0 between (64.569) and (64.570), we get the equation
1 + κ2 0 κ2 − 3 λ C
H = 2CF 0 + θ+ + (64.574)
κ 4κ 4 κ
Integrating the above equation, we find
1 + κ2 κ2 − 3 2 λ C
H = 2CF + θ + + θ+B (64.575)
κ 8κ 4 κ
where B is an arbitrary constant.
65 The Navier-Stokes Equations

Navier-Stokes equations in Fluid Mechanics describe the motion of a fluid
in terms of the velocity u and its pressure p. These equations contain in
addition the viscosity of the fluid under consideration.
753
65.1 The 3-dimensional form of the Navier-Stokes
equations
The Navier-Stokes equations in 3-dimensional form read
∂u
+ u·∇u = −∇p + ν∆u
∂t (65.1)
∇·u = 0
where ν is the viscosity. We denote by u = (u, v, w) the three components of

the velocity vector. The functions u, v and w depend on three space variables
x, y and z and the time t. The above equations can be written in component
form as
ut + uux + vuy + wuz = −px + ν(uxx + uyy + uzz )
vt + uvx + vvy + wvz = −py + ν(vxx + vyy + vzz )
(65.2)
wt + uwx + vwy + wwz = −pz + ν(wxx + wyy + wzz )
ux + vy + wz = 0

∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
(65.3)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂u ∂v ∂w ∂p
The functions {ξ 1 , · · ·, ξ 4 } and {φ1 , · · ·, φ4 } depend both on the space-time

variables x, y, z, t and the functions u, v, w and p.
∂ ∂ ∂ ∂
P r(1) X = X + φx + φy + φz + φt
∂ ∂ ∂ ∂
(65.4)
∂ ∂ ∂ ∂
∂ ∂ ∂
+ πx + πy + πz
∂px ∂py ∂pz
754
The second prolongation of the vector X is denoted by P r(2) X and is given
by
P r(2) X = P r(1) X+
∂ ∂ ∂ ∂
+ φxx + φyy + φzz + φtt
∂uxx ∂uyy ∂uzz ∂utt
∂ ∂ ∂ ∂
+ ψ xx + ψ yy + ψ zz + ψ tt
∂vxx ∂vyy ∂vzz ∂vtt (65.5)
∂ ∂ ∂ ∂
+ χxx + χyy + χzz + χtt
∂wxx ∂wyy ∂wzz ∂wtt
∂ ∂ ∂
+ π xx + π yy + π zz
∂pxx ∂pyy ∂pzz
We further denote by ∆(1) , ∆(2) , ∆(3) , ∆(4) the quantities
∆(1) = ut + uux + vuy + wuz + px − ν(uxx + uyy + uzz )
∆(2) = vt + uvx + vvy + wvz + py − ν(vxx + vyy + vzz )
(65.6)
∆(3) = wt + uwx + vwy + wwz + pz − ν(wxx + wyy + wzz )
∆(4) = ux + vy + wz = 0
The Lie symmetries of the Navier-Stokes equations can be determined once
we know the explicit form of the functions {ξ 1 , · · ·, ξ 4 } and {φ1 , · · ·, φ4 }. The
Lie symmetries are determined by the conditions
P r(2) X[∆(k) ] = 0 as long as ∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0 (65.7)
Applying the above conditions, we obtain
P r(2) X[∆(1) ] = φt + (φ1 )ux + (φ2 )uy + (φ3 )uz
+ uφx + vφy + wφz + π x (65.8)
− νφxx − νφyy − νφzz
P r(2) X[∆(2) ] = ψ t + (φ1 )vx + (φ2 )vy + (φ3 )vz

+ uψ x + vψ y + wψ z + π y (65.9)
− νψ xx − νψ yy − νψ zz
P r(2) X[∆(3) ] = χt + (φ1 )wx + (φ2 )wy + (φ3 )wz
+ uχx + vχy + wχz + π z (65.10)
− νχxx − νχyy − νχzz
755
P r(2) X[∆(4) ] = φx + ψ y + χz (65.11)
The coefficients φx , · · · , π z appearing in P r(1) X have been evaluated before
in Euler’s equations (see (64.11)-(64.25)).
Let us now list the coefficients φxx , · · ·, χzz we shall use in evaluating the var-
ious P r(2) X[∆(k) ]. These coefficients have been evaluated using the formulas
(2.39), (2.43) or (2.44) of Olver (Reference [13]) and are listed below
756
φxx = −pxx [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− 2px [uxx (ξ 1 )p + uxy (ξ 2 )p + uxz (ξ 3 )p + uxt (ξ 4 )p
+ (ux )2 (ξ 1 )up + ux vx (ξ 1 )vp + ux wx (ξ 1 )wp + ux (ξ 1 )xp
+ ux uy (ξ 2 )up + uy vx (ξ 2 )vp + uy wx (ξ 2 )wp + uy (ξ 2 )xp
+ ux uz (ξ 3 )up + uz vx (ξ 3 )vp + uz wx (ξ 3 )wp + uz (ξ 3 )xp
+ ut ux (ξ 4 )up + ut vx (ξ 4 )vp + ut wx (ξ 4 )wp + ut (ξ 4 )xp
− ux (φ1 )up − vx (φ1 )vp − wx (φ1 )wp − (φ1 )xp ]
− (px )2 [ux (ξ 1 )pp + uy (ξ 2 )pp + uz (ξ 3 )pp + ut (ξ 4 )pp − (φ1 )pp ]
− uxx [3ux (ξ 1 )u + 2vx (ξ 1 )v + 2wx (ξ 1 )w + 2(ξ 1 )x
+ uy (ξ 2 )u + uz (ξ 3 )u + ut (ξ 4 )u − (φ1 )u ]
− 2uxy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− 2uxz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− 2uxt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
− vxx [ux (ξ 1 )v + uy (ξ 2 )v + uz (ξ 3 )v + ut (ξ 4 )v − (φ1 )v ]
(65.12)
− wxx [ux (ξ 1 )w + uy (ξ 2 )w + uz (ξ 3 )w + ut (ξ 4 )w − (φ1 )w ]
− (ux )2 [ux (ξ 1 )uu + 2vx (ξ 1 )uv + 2wx (ξ 1 )uw + 2(ξ 1 )ux
+ uy (ξ 2 )uu + uz (ξ 3 )uu + ut (ξ 4 )uu − (φ1 )uu ]
− (vx )2 [ux (ξ 1 )vv + uy (ξ 2 )vv + uz (ξ 3 )vv + ut (ξ 4 )vv − (φ1 )vv ]
− (wx )2 [ux (ξ 1 )ww + uy (ξ 2 )ww + uz (ξ 3 )ww + ut (ξ 4 )ww − (φ1 )ww ]
− ux [2vx (ξ 1 )xv + 2wx (ξ 1 )xw + (ξ 1 )xx ]
− uy [2ux (ξ 2 )xu + 2vx (ξ 2 )xv + 2wx (ξ 2 )xw + (ξ 2 )xx ]
− uz [2ux (ξ 3 )xu + 2vx (ξ 3 )xv + 2wx (ξ 3 )xw + (ξ 3 )xx ]
− ut [2ux (ξ 4 )xu + 2vx (ξ 4 )xv + 2wx (ξ 4 )xw + (ξ 4 )xx ]
− 2ux vx wx (ξ 1 )vw − 2ux uy vx (ξ 2 )uv − 2ux uy wx (ξ 2 )uv − 2uy vx wx (ξ 2 )vw
− 2ux uz vx (ξ 3 )uv − 2ux uz wx (ξ 3 )uv − 2uz vx wx (ξ 3 )vw − 2ut ux vx (ξ 4 )uv
− 2ut ux wx (ξ 4 )uv − 2ut vx wx (ξ 4 )vw
+ 2ux (φ1 )xu + 2vx (φ1 )xv + 2wx (φ1 )xw + 2ux vx (φ1 )uv + 2ux wx (φ1 )uw
+ 2vx wx (φ1 )vw + (φ1 )xx
757
φyy = −pyy [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− 2py [uyy (ξ 2 )p + uxy (ξ 1 )p + uyz (ξ 3 )p + uyt (ξ 4 )p
+ (uy )2 (ξ 2 )up + ux vy (ξ 1 )vp + ux wy (ξ 1 )wp + ux (ξ 1 )yp
+ uy uy (ξ 2 )vp + uy wy (ξ 2 )wp + uy uz (ξ 3 )up + uy (ξ 2 )yp
+ uz vy (ξ 3 )vp + uz wy (ξ 3 )wp + ux uy (ξ 1 )up + uz (ξ 3 )yp
+ ut uy (ξ 4 )up + ut vy (ξ 4 )vp + ut wy (ξ 4 )wp + ut (ξ 4 )yp
− uy (φ1 )up − vy (φ1 )vp − wy (φ1 )wp − (φ1 )yp ]
− (py )2 [ux (ξ 1 )pp + uy (ξ 2 )pp + uz (ξ 3 )pp + ut (ξ 4 )pp − (φ1 )pp ]
− uyy [3uy (ξ 2 )u + 2vy (ξ 2 )v + 2wy (ξ 2 )w + 2(ξ 2 )y
+ ux (ξ 1 )u + uz (ξ 3 )u + ut (ξ 4 )u − (φ1 )u ]
− 2uxy [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− 2uyz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− 2uyt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
− vyy [ux (ξ 1 )v + uy (ξ 2 )v + uz (ξ 3 )v + ut (ξ 4 )v − (φ1 )v ]
(65.13)
− wyy [ux (ξ 1 )w + uy (ξ 2 )w + uz (ξ 3 )w + ut (ξ 4 )w − (φ1 )w ]
− (uy )2 [ux (ξ 1 )uu + 2vy (ξ 2 )uv + 2wy (ξ 2 )uw + 2(ξ 2 )uy
− (vy )2 [ux (ξ 1 )vv + uy (ξ 2 )vv + uz (ξ 3 )vv + ut (ξ 4 )vv − (φ1 )vv ]
− (wy )2 [ux (ξ 1 )ww + uy (ξ 2 )ww + uz (ξ 3 )ww + ut (ξ 4 )ww − (φ1 )ww ]
− ux [2uy (ξ 1 )yu + 2vy (ξ 1 )yv + 2wy (ξ 1 )yw + (ξ 1 )yy ]
− uy [2vy (ξ 2 )yv + 2wy (ξ 2 )yw + (ξ 2 )yy ]
− uz [2uy (ξ 3 )yu + 2vy (ξ 3 )yv + 2wy (ξ 3 )yw + (ξ 3 )yy ]
− ut [2uy (ξ 4 )yu + 2vy (ξ 4 )yv + 2wy (ξ 4 )yw + (ξ 4 )yy ]
− 2ux uy vy (ξ 1 )uv − 2ux uy wy (ξ 1 )uw − 2ux vy wy (ξ 1 )vw − 2uy vy wy (ξ 2 )vw
− 2uy uz vy (ξ 3 )uv − 2uy uz wy (ξ 3 )uv − 2uz vy wy (ξ 3 )vw − 2ut uy vy (ξ 4 )uv
− 2ut uy wy (ξ 4 )uv − 2ut vy wy (ξ 4 )vw
+ 2uy (φ1 )yu + 2vy (φ1 )yv + 2wy (φ1 )yw + 2uy vy (φ1 )uv + 2uy wy (φ1 )uw
+ 2vy wy (φ1 )vw + (φ1 )yy
758
φzz = −pzz [ux (ξ 1 )p + uy (ξ 2 )p + uz (ξ 3 )p + ut (ξ 4 )p − (φ1 )p ]
− 2pz [uzz (ξ 3 )p + uxz (ξ 1 )p + uyz (ξ 2 )p + uzt (ξ 4 )p
+ (uz )2 (ξ 3 )up + ux vz (ξ 1 )vp + ux wz (ξ 1 )wp + ux (ξ 1 )zp
+ uy vz (ξ 2 )vp + uy wz (ξ 2 )wp + uy uz (ξ 2 )up + uy (ξ 2 )yp
+ uz vz (ξ 3 )vp + uz wz (ξ 3 )wp + ux uz (ξ 1 )up + uz (ξ 3 )zp
+ ut uz (ξ 4 )up + ut vz (ξ 4 )vp + ut wz (ξ 4 )wp + ut (ξ 4 )zp
− uz (φ1 )up − vz (φ1 )vp − wz (φ1 )wp − (φ1 )zp ]
− (pz )2 [ux (ξ 1 )pp + uy (ξ 2 )pp + uz (ξ 3 )pp + ut (ξ 4 )pp − (φ1 )pp ]
− uzz [3uz (ξ 3 )u + 2vz (ξ 3 )v + 2wz (ξ 3 )w + 2(ξ 3 )z
+ ux (ξ 1 )u + uy (ξ 2 )u + ut (ξ 4 )u − (φ1 )u ]
− 2uxz [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− 2uyz [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− 2uzt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
− vzz [ux (ξ 1 )v + uy (ξ 2 )v + uz (ξ 3 )v + ut (ξ 4 )v − (φ1 )v ]
(65.14)
− wzz [ux (ξ 1 )w + uy (ξ 2 )w + uz (ξ 3 )w + ut (ξ 4 )w − (φ1 )w ]
− (uz )2 [ux (ξ 1 )uu + 2vz (ξ 3 )uv + 2wz (ξ 3 )uw + 2(ξ 3 )uz
− (vz )2 [ux (ξ 1 )vv + uy (ξ 2 )vv + uz (ξ 3 )vv + ut (ξ 4 )vv − (φ1 )vv ]
− (wz )2 [ux (ξ 1 )ww + uy (ξ 2 )ww + uz (ξ 3 )ww + ut (ξ 4 )ww − (φ1 )ww ]
− ux [2uz (ξ 1 )zu + 2vz (ξ 1 )zv + 2wz (ξ 1 )zw + (ξ 1 )zz ]
− uy [2uz (ξ 2 )zu + 2vz (ξ 2 )zv + 2wz (ξ 2 )zw + (ξ 2 )zz ]
− uz [2vz (ξ 3 )zv + 2wz (ξ 3 )zw + (ξ 3 )zz ]
− ut [2uz (ξ 4 )zu + 2vz (ξ 4 )zv + 2wz (ξ 4 )zw + (ξ 4 )zz ]
− 2uy uz vz (ξ 2 )uv − 2uy uz wz (ξ 2 )uw − 2uy vz wz (ξ 2 )vw − 2uy vy wy (ξ 2 )vw
− 2uy uz vy (ξ 3 )uv − 2uy uz wy (ξ 3 )uv − 2uz vz wz (ξ 3 )vw − 2ut uz vz (ξ 4 )uv
− 2ut uz wz (ξ 4 )uv − 2ut vz wz (ξ 4 )vw
+ 2uz (φ1 )zu + 2vz (φ1 )zv + 2wz (φ1 )zw + 2uz vz (φ1 )uv + 2uz wz (φ1 )uw
+ 2vz wz (φ1 )vw + (φ1 )zz
759
ψ xx = −pxx [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− 2px [vxx (ξ 1 )p + vxy (ξ 2 )p + vxz (ξ 3 )p + vxt (ξ 4 )p
+ (vx )2 (ξ 1 )vp + ux vx (ξ 1 )up + ux vy (ξ 2 )up + vx (ξ 1 )xp
+ vx wx (ξ 1 )wp + vx vy (ξ 2 )vp + vy wx (ξ 2 )wp + vy (ξ 2 )xp
+ ux vz (ξ 3 )up + vx vz (ξ 3 )vp + vz wx (ξ 3 )wp + vz (ξ 3 )xp
+ vt ux (ξ 4 )up + vt vx (ξ 4 )vp + vt wx (ξ 4 )wp + vt (ξ 4 )xp
− (px )2 [vx (ξ 1 )pp + vy (ξ 2 )pp + vz (ξ 3 )pp + vt (ξ 4 )pp − (φ2 )pp ]
− vxx [3vx (ξ 1 )v + 2ux (ξ 1 )u + 2wx (ξ 1 )w + 2(ξ 1 )x
+ vy (ξ 2 )v + vz (ξ 3 )v + vt (ξ 4 )v − (φ2 )v ]
− 2vxy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− 2vxz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− 2vxt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
− uxx [vx (ξ 1 )u + vy (ξ 2 )u + vz (ξ 3 )u + vt (ξ 4 )u − (φ2 )u ]
(65.15)
− wxx [vx (ξ 1 )w + vy (ξ 2 )w + vz (ξ 3 )w + vt (ξ 4 )w − (φ2 )w ]
− (vx )2 [vx (ξ 1 )vv + 2ux (ξ 1 )uv + 2wx (ξ 1 )vw + 2(ξ 1 )xv
+ vy (ξ 2 )vv + vz (ξ 3 )vv + vt (ξ 4 )vv − (φ2 )vv ]
− (ux )2 [vx (ξ 1 )uu + vy (ξ 2 )uu + vz (ξ 3 )uu + vt (ξ 4 )uu − (φ2 )uu ]
− (wx )2 [vx (ξ 1 )ww + vy (ξ 2 )ww + vz (ξ 3 )ww + vt (ξ 4 )ww − (φ2 )ww ]
− vx [2ux (ξ 1 )xu + 2wx (ξ 1 )xw + (ξ 1 )xx ]
− vy [2ux (ξ 2 )xu + 2vx (ξ 2 )xv + 2wx (ξ 2 )xw + (ξ 2 )xx ]
− vz [2ux (ξ 3 )xu + 2vx (ξ 3 )xv + 2wx (ξ 3 )xw + (ξ 3 )xx ]
− vt [2ux (ξ 4 )xu + 2vx (ξ 4 )xv + 2wx (ξ 4 )xw + (ξ 4 )xx ]
− 2ux vx wx (ξ 1 )uw − 2ux vy vx (ξ 2 )uv − 2ux vy wx (ξ 2 )uv − 2vx vy wx (ξ 2 )vw
− 2ux vz vx (ξ 3 )uv − 2ux vz wx (ξ 3 )uw − 2vz vx wx (ξ 3 )vw − 2vt ux vx (ξ 4 )uv
− 2vt ux wx (ξ 4 )uv − 2vt vx wx (ξ 4 )vw
+ 2vx wx (φ2 )vw + (φ2 )xx
760
ψ yy = −pyy [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− 2py [vyy (ξ 2 )p + vxy (ξ 1 )p + uyz (ξ 3 )p + uyt (ξ 4 )p
+ (vy )2 (ξ 2 )vp + vx vy (ξ 1 )vp + vx wy (ξ 1 )wp + vx (ξ 1 )yp
+ uy vx (ξ 1 )up + uy vy (ξ 2 )up + vy wy (ξ 2 )wp + vy (ξ 2 )yp
+ uy vz (ξ 3 )up + vy vz (ξ 3 )vp + vz wy (ξ 3 )wp + vz (ξ 3 )yp
+ vt uy (ξ 4 )up + vt vy (ξ 4 )vp + vt wy (ξ 4 )wp + vt (ξ 4 )yp
− (py )2 [vx (ξ 1 )pp + vy (ξ 2 )pp + vz (ξ 3 )pp + vt (ξ 4 )pp − (φ2 )pp ]
− vyy [3vy (ξ 2 )v + 2uy (ξ 2 )u + 2wy (ξ 2 )w + 2(ξ 2 )y
+ vx (ξ 1 )v + vz (ξ 3 )v + vt (ξ 4 )v − (φ2 )v ]
− 2vxy [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− 2vyz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− 2vyt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
− uyy [vx (ξ 1 )u + vy (ξ 2 )u + vz (ξ 3 )u + vt (ξ 4 )u − (φ2 )u ]
(65.16)
− wyy [vx (ξ 1 )w + vy (ξ 2 )w + vz (ξ 3 )w + vt (ξ 4 )w − (φ2 )w ]
− (vy )2 [vx (ξ 1 )vv + 2uy (ξ 2 )uv + 2wy (ξ 2 )vw + 2(ξ 2 )yw
− (uy )2 [vx (ξ 1 )uu + vy (ξ 2 )uu + vz (ξ 3 )uu + vt (ξ 4 )uu − (φ2 )uu ]
− (wy )2 [vx (ξ 1 )ww + vy (ξ 2 )ww + vz (ξ 3 )ww + vt (ξ 4 )ww − (φ2 )ww ]
− vx [2uy (ξ 1 )yu + 2vy (ξ 1 )yv + 2wy (ξ 1 )yw + (ξ 1 )yy ]
− vy [2vy (ξ 2 )yu + 2wy (ξ 2 )yw + (ξ 2 )yy ]
− vz [2uy (ξ 3 )yu + 2vy (ξ 3 )yv + 2wy (ξ 3 )yw + (ξ 3 )yy ]
− vt [2uy (ξ 4 )yu + 2vy (ξ 4 )yv + 2wy (ξ 4 )yw + (ξ 4 )yy ]
− 2uy vx vy (ξ 1 )uv − 2uy vx wy (ξ 1 )uw − 2vx vy wy (ξ 1 )vw − 2uy vy wy (ξ 2 )vw
− 2uy vy vz (ξ 3 )uv − 2uy vz wy (ξ 3 )uv − 2vy vz wy (ξ 3 )vw − 2vt uy vy (ξ 4 )uv
− 2vt uy wy (ξ 4 )uv − 2vt vy wy (ξ 4 )vw
+ 2vy wy (φ2 )vw + (φ2 )yy
761
ψ zz = −pzz [vx (ξ 1 )p + vy (ξ 2 )p + vz (ξ 3 )p + vt (ξ 4 )p − (φ2 )p ]
− 2pz [vzz (ξ 3 )p + vxz (ξ 1 )p + vyz (ξ 2 )p + vzt (ξ 4 )p
+ (vz )2 (ξ 3 )vp + uz vx (ξ 1 )up + vx vz (ξ 1 )vp + vx (ξ 1 )zp
+ vx wz (ξ 1 )wp + uz vy (ξ 2 )up + vy vz (ξ 2 )vp + vy (ξ 2 )zp
+ vy wz (ξ 2 )wp + uz vz (ξ 3 )up + vz wz (ξ 3 )wp + vz (ξ 3 )zp
+ vt uz (ξ 4 )up + vt vz (ξ 4 )vp + vt wz (ξ 4 )wp + vt (ξ 4 )zp
− (pz )2 [vx (ξ 1 )pp + vy (ξ 2 )pp + vz (ξ 3 )pp + vt (ξ 4 )pp − (φ2 )pp ]
− vzz [3vz (ξ 3 )v + 2uz (ξ 3 )u + 2wz (ξ 3 )w + 2(ξ 3 )z
+ vx (ξ 1 )v + vy (ξ 2 )v + vt (ξ 4 )v − (φ2 )v ]
− 2vxz [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− 2vyz [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− 2vzt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
− uzz [vx (ξ 1 )u + vy (ξ 2 )u + vz (ξ 3 )u + vt (ξ 4 )u − (φ2 )u ]
(65.17)
− wzz [vx (ξ 1 )w + vy (ξ 2 )w + vz (ξ 3 )w + vt (ξ 4 )w − (φ2 )w ]
− (vz )2 [vx (ξ 1 )vv + 2uz (ξ 3 )uv + 2wz (ξ 3 )vw + 2(ξ 3 )zv
− (uz )2 [vx (ξ 1 )uu + vy (ξ 2 )uu + vz (ξ 3 )uu + vt (ξ 4 )uu − (φ2 )uu ]
− (wz )2 [vx (ξ 1 )ww + vy (ξ 2 )ww + vz (ξ 3 )ww + vt (ξ 4 )ww − (φ2 )ww ]
− vx [2uz (ξ 1 )zu + 2vz (ξ 1 )zv + 2wz (ξ 1 )zw + (ξ 1 )zz ]
− vy [2uz (ξ 2 )zu + 2vz (ξ 2 )zv + 2wz (ξ 2 )zw + (ξ 2 )zz ]
− vz [2uz (ξ 3 )zu + 2vz (ξ 3 )zw + (ξ 3 )zz ]
− vt [2uz (ξ 4 )zu + 2vz (ξ 4 )zv + 2wz (ξ 4 )zw + (ξ 4 )zz ]
− 2ux vx vz (ξ 1 )uv − 2uz vx wz (ξ 1 )uw − 2vx vz wz (ξ 1 )vw − 2uz vy vz (ξ 2 )uv
− 2uz vy wz (ξ 2 )uw − 2vy vz wz (ξ 2 )vw − 2uz vz wz (ξ 3 )uw − 2vt uz vz (ξ 4 )uv
− 2vt uz wz (ξ 4 )uw − 2vt vz wz (ξ 4 )vw
+ 2vz wz (φ2 )vw + (φ2 )zz
762
χxx = −pxx [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− 2px [wxx (ξ 1 )p + wxy (ξ 2 )p + wxz (ξ 3 )p + wxt (ξ 4 )p
+ (wx )2 (ξ 1 )wp + ux wx (ξ 1 )up + vx wx (ξ 1 )vp + wx (ξ 1 )xp
+ ux wy (ξ 2 )up + vx wy (ξ 2 )vp + wx wy (ξ 2 )wp + wy (ξ 2 )xp
+ ux wz (ξ 3 )up + vx wz (ξ 3 )vp + wz wx (ξ 3 )wp + wz (ξ 3 )xp
+ wt ux (ξ 4 )up + wt vx (ξ 4 )vp + wt wx (ξ 4 )wp + wt (ξ 4 )xp
− (px )2 [wx (ξ 1 )pp + wy (ξ 2 )pp + wz (ξ 3 )pp + wt (ξ 4 )pp − (φ3 )pp ]
− wxx [3wx (ξ 1 )w + 2vx (ξ 1 )v + 2ux (ξ 1 )u + 2(ξ 1 )x
+ wy (ξ 2 )w + wz (ξ 3 )w + wt (ξ 4 )w − (φ3 )w ]
− 2wxy [ux (ξ 2 )u + vx (ξ 2 )v + wx (ξ 2 )w + (ξ 2 )x ]
− 2wxz [ux (ξ 3 )u + vx (ξ 3 )v + wx (ξ 3 )w + (ξ 3 )x ]
− 2wxt [ux (ξ 4 )u + vx (ξ 4 )v + wx (ξ 4 )w + (ξ 4 )x ]
− uxx [wx (ξ 1 )u + wy (ξ 2 )u + wz (ξ 3 )u + wt (ξ 4 )u − (φ3 )u ]
− vxx [wx (ξ 1 )v + wy (ξ 2 )v + wz (ξ 3 )v + wt (ξ 4 )v − (φ3 )v ]
− (wx )2 [wx (ξ 1 )ww + 2ux (ξ 1 )uv + 2vx (ξ 1 )vw + 2(ξ 1 )xw
+ wy (ξ 2 )ww + wz (ξ 3 )ww + wt (ξ 4 )ww − (φ3 )ww ]
− (ux )2 [wx (ξ 1 )uu + wy (ξ 2 )uu + wz (ξ 3 )uu + wt (ξ 4 )uu − (φ3 )uu ]
− (vx )2 [wx (ξ 1 )vv + wy (ξ 2 )vv + wz (ξ 3 )vv + wt (ξ 4 )vv − (φ3 )vv ]
− wx [2ux (ξ 1 )xu + 2vx (ξ 1 )xv + (ξ 1 )xx ]
− wy [2ux (ξ 2 )xu + 2vx (ξ 2 )xv + 2wx (ξ 2 )xw + (ξ 2 )xx ]
− wz [2ux (ξ 3 )xu + 2vx (ξ 3 )xv + 2wx (ξ 3 )xw + (ξ 3 )xx ]
− wt [2ux (ξ 4 )xu + 2vx (ξ 4 )xv + 2wx (ξ 4 )xw + (ξ 4 )xx ]
− 2ux vx wy (ξ 2 )uv − 2ux wx wy (ξ 2 )uw − 2vx wx wy (ξ 2 )vw − 2ux vx wz (ξ 3 )uv
− 2ux wx wz (ξ 3 )uw − 2vx wx wz (ξ 3 )vw − 2wt ux wx (ξ 4 )uv − 2wt ux wx (ξ 4 )uw
− 2wt vx wx (ξ 4 )vw − 2ux vx wx (ξ 1 )uw
+ 2vx wx (φ3 )vw + (φ3 )xx
(65.18)
763
χyy = −pyy [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− 2py [wyy (ξ 2 )p + wxy (ξ 1 )p + wyz (ξ 3 )p + wyt (ξ 4 )p
+ (wy )2 (ξ 2 )wp + uy wx (ξ 1 )up + wx wy (ξ 1 )wp + wx (ξ 1 )yp
+ uy wy (ξ 2 )up + vy wy (ξ 2 )vp + uy wz (ξ 3 )up + wy (ξ 2 )yp
+ vy wz (ξ 3 )vp + wy wz (ξ 3 )wp + vy wx (ξ 1 )vp + wz (ξ 3 )yp
+ wt uy (ξ 4 )up + wt vy (ξ 4 )vp + wt wy (ξ 4 )wp + wt (ξ 4 )yp
− (py )2 [wx (ξ 1 )pp + wy (ξ 2 )pp + wz (ξ 3 )pp + wt (ξ 4 )pp − (φ3 )pp ]
− wyy [3wy (ξ 2 )w + 2vy (ξ 2 )v + 2uy (ξ 2 )u + 2(ξ 2 )y
+ wx (ξ 1 )w + wz (ξ 3 )w + wt (ξ 4 )w − (φ3 )w ]
− 2wxy [uy (ξ 1 )u + vy (ξ 1 )v + wy (ξ 1 )w + (ξ 1 )y ]
− 2wyz [uy (ξ 3 )u + vy (ξ 3 )v + wy (ξ 3 )w + (ξ 3 )y ]
− 2wyt [uy (ξ 4 )u + vy (ξ 4 )v + wy (ξ 4 )w + (ξ 4 )y ]
− uyy [wx (ξ 1 )u + wy (ξ 2 )u + wz (ξ 3 )u + wt (ξ 4 )u − (φ3 )u ]
− vyy [wx (ξ 1 )v + wy (ξ 2 )v + wz (ξ 3 )v + wt (ξ 4 )v − (φ3 )v ]
− (wy )2 [wx (ξ 1 )ww + 2uy (ξ 2 )uv + 2vy (ξ 2 )vw + 2(ξ 2 )yw
− (uy )2 [wx (ξ 1 )uu + wy (ξ 2 )uu + wz (ξ 3 )uu + wt (ξ 4 )uu − (φ3 )uu ]
− (vy )2 [wx (ξ 1 )vv + wy (ξ 2 )vv + wz (ξ 3 )vv + wt (ξ 4 )vv − (φ3 )vv ]
− wx [2uy (ξ 1 )yu + 2vy (ξ 1 )yv + 2wy (ξ 1 )yw + (ξ 1 )yy ]
− wy [2uy (ξ 2 )yu + 2vy (ξ 2 )yv + (ξ 2 )yy ]
− wz [2uy (ξ 3 )yu + 2vy (ξ 3 )yv + 2wy (ξ 3 )yw + (ξ 3 )yy ]
− wt [2uy (ξ 4 )yu + 2vy (ξ 4 )yv + 2wy (ξ 4 )yw + (ξ 4 )yy ]
− 2uy vy wx (ξ 1 )uv − 2uy wx wy (ξ 1 )uw − 2vy wx wy (ξ 1 )vw − 2uy vy wy (ξ 2 )uv
− 2uy vy wz (ξ 3 )uv − 2uy wy wz (ξ 3 )uw − 2vy wy wz (ξ 3 )vw − 2wt uy vy (ξ 4 )uv
− 2wt uy wy (ξ 4 )uv − 2wt vy wy (ξ 4 )vw
+ 2vy wy (φ3 )vw + (φ3 )yy
(65.19)
764
χzz = −pzz [wx (ξ 1 )p + wy (ξ 2 )p + wz (ξ 3 )p + wt (ξ 4 )p − (φ3 )p ]
− 2pz [wzz(ξ 3 )p + wxz (ξ 1 )p + wyz (ξ 2 )p + wzt (ξ 4 )p
+ (wz )2 (ξ 3 )wp + uz wx (ξ 1 )up + vz wx (ξ 1 )vp + wx (ξ 1 )zp
+ wx wz (ξ 1 )wp + uz wy (ξ 2 )up + vz wy (ξ 2 )vp + wy (ξ 2 )zp
+ wy wz (ξ 2 )wp + uz wz (ξ 3 )up + vz wz (ξ 3 )vp + wz (ξ 3 )zp
+ wt uz (ξ 4 )up + wt vz (ξ 4 )vp + wt wz (ξ 4 )wp + wt (ξ 4 )zp
− (pz )2 [wx (ξ 1 )pp + wy (ξ 2 )pp + wz (ξ 3 )pp + wt (ξ 4 )pp − (φ3 )pp ]
− wzz [3wz (ξ 3 )w + 2uz (ξ 3 )u + 2vz (ξ 3 )v + 2(ξ 3 )z
+ wx (ξ 1 )w + wy (ξ 2 )w + wt (ξ 4 )w − (φ3 )w ]
− 2wxz [uz (ξ 1 )u + vz (ξ 1 )v + wz (ξ 1 )w + (ξ 1 )z ]
− 2wyz [uz (ξ 2 )u + vz (ξ 2 )v + wz (ξ 2 )w + (ξ 2 )z ]
− 2wzt [uz (ξ 4 )u + vz (ξ 4 )v + wz (ξ 4 )w + (ξ 4 )z ]
− uzz [wx (ξ 1 )u + wy (ξ 2 )u + wz (ξ 3 )u + wt (ξ 4 )u − (φ3 )u ]
− vzz [wx (ξ 1 )v + wy (ξ 2 )v + wz (ξ 3 )v + wt (ξ 4 )v − (φ3 )v ]
− (wz )2 [wx (ξ 1 )ww + 2uz (ξ 3 )uw + 2vz (ξ 3 )vw + 2(ξ 3 )zw
− (uz )2 [wx (ξ 1 )uu + wy (ξ 2 )uu + wz (ξ 3 )uu + wt (ξ 4 )uu − (φ3 )uu ]
− (vz )2 [wx (ξ 1 )vv + wy (ξ 2 )vv + wz (ξ 3 )vv + wt (ξ 4 )vv − (φ3 )vv ]
− wx [2uz (ξ 1 )zu + 2vz (ξ 1 )zv + 2wz (ξ 1 )zw + (ξ 1 )zz ]
− wy [2uz (ξ 2 )zw + 2vz (ξ 2 )zv + 2wz (ξ 2 )zw + (ξ 2 )zz ]
− wz [2uz (ξ 3 )zu + 2vz (ξ 3 )zv + (ξ 3 )zz ]
− wt [2uz (ξ 4 )zu + 2vz (ξ 4 )zv + 2wz (ξ 4 )zw + (ξ 4 )zz ]
− 2uz vz wx (ξ 1 )uv − 2uz wx wz (ξ 1 )uw − 2vz wx wz (ξ 1 )vw − 2uz vz wy (ξ 2 )uw
− 2uz wx wz (ξ 2 )uw − 2vz wy wz (ξ 2 )vw − 2uz vz wz (ξ 3 )uv − 2wt uz vz (ξ 4 )uv
− 2wt uz wz (ξ 4 )uw − 2wt vz wz (ξ 4 )vw
+ 2vz wz (φ3 )vw + (φ3 )zz
(65.20)
765
Equation P r(2) X[∆(4) ] is equivalent to
P r(2) X[∆(4) ] =
= (φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w − (ux )2 (ξ 1 )u
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]
+ (φ3 )z + uz (φ3 )u + vz (φ3 )v + wz (φ3 )w − (wz )2 (ξ 3 )w
− wz [uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]
px −→ − (ut + uux + vuy + wuz ) + ν(uxx + uyy + uzz )

py −→ − (vt + uvx + vvy + wvz ) + ν(vxx + vyy + vzz )
(65.22)
pz −→ − (wt + uwx + vwy + wwz ) + ν(wxx + wyy + wzz )
wz −→ − (ux + vy )

We thus see that P r(2) X[∆(4) ] =0
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
is equivalent to
766
+ [ut + uux + vuy + wuz − ν(uxx + uyy + uzz )]×
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
+ [vt + uvx + vvy + wvz − ν(vxx + vyy + vzz )]×
2 2 2
− vy [uy (ξ )u + wy (ξ )w + (ξ )y ]
+ [wt + uwx + vwy − (ux + vy )w − ν(wxx + wyy + wzz )]×
+ (ux + vy )[uz (ξ 3 )u + vz (ξ 3 )v + (ξ 3 )z ]
− wt [uz (ξ 4 )u + vz (ξ 4 )v − (ux + vy )(ξ 4 )w + (ξ 4 )z ] = 0
equate to zero the easily identified coefficients of some partial derivatives. In
fact, we see that (ut )2 , (vt )2 and (wt )2 have common coefficient (ξ 4 )p . The
terms (uz )2 and (vz )2 have common coefficient w(ξ 3 )p . The terms ut wx and
vt wy have common coefficient −(ξ 4 )w . The coefficients of wt uz , uz vx and wt vz
are −(ξ 4 )u , −(ξ 3 )v and −(ξ 4 )v respectively. The coefficient of vz vx is equal
to w(ξ 1 )p + u(ξ 3 )p . Therefore equating all these coefficients to zero, we have
(ξ 4 )p = (ξ 4 )u = (ξ 4 )v = (ξ 4 )w = 0, (ξ 3 )p = (ξ 3 )v = 0, (ξ 1 )p = 0 (65.24)
767
We thus see that equation (65.23) gets simplified into

+ [(ut + uux + vuy + wuz ) − ν(uxx + uyy + uzz )]×[uy (ξ 2 )p − (φ1 )p ]
− ux [vx (ξ 1 )v + wx (ξ 1 )w + (ξ 1 )x ]
− uz [ux (ξ 3 )u + wx (ξ 3 )w + (ξ 3 )x ] − ut (ξ 4 )x
+ [(vt + uvx + vvy + wvz ) − ν(vxx + vyy + vzz )]×[vy (ξ 2 )p − (φ2 )p ]
− vy [uy (ξ 2 )u + wy (ξ 2 )w + (ξ 2 )y ]
− vz [uy (ξ 3 )u + wy (ξ 3 )w + (ξ 3 )y ] − vt (ξ 4 )y
+ [(wt + uwx + vwy − (ux + vy )w) − ν(wxx + wyy + wzz )]×
×[wy (ξ 2 )p − (φ3 )p ]
+ (ux + vy )[uz (ξ 3 )u + (ξ 3 )z ] − wt (ξ 4 )z = 0
(65.25)
From the above equation we see that the coefficients of uz wx and vy uz are
−(ξ 3 )w and (ξ 3 )u respectively. We also find that the coefficients of uy vx ,
uy wx and vx wy are −(ξ 2 )v , −(ξ 2 )w and −(ξ 1 )w respectively. Furthermore
the coefficients of vx uy , ux vx , wt wy and wy uz are −(ξ 1 )u , −(ξ 1 )v , (ξ 2 )p and
−(ξ 2 )u respectively. Equating all these coefficients to zero, we have
(ξ 3 )w = (ξ 3 )u = 0, (ξ 2 )p = (ξ 2 )u = (ξ 2 )v = (ξ 2 )w = 0,
(65.26)
(ξ 1 )u = (ξ 1 )v = (ξ 1 )w = 0
768
We thus see that equation (65.25) takes on the form
(φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w

+ [(ut + uux + vuy + wuz ) − ν(uxx + uyy + uzz )]×[−(φ1 )p ]
+ (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w
+ [(vt + uvx + vvy + wvz ) − ν(vxx + vyy + vzz )]×[−(φ2 )p ]
+ [(wt + uwx + vwy − (ux + vy )w) − ν(wxx + wyy + wzz )]×[−(φ3 )p ]
− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z − wt (ξ 4 )z = 0
(65.27)
Before we proceed further, let us put together (65.24) and (65.26):
(ξ 3 )p = (ξ 3 )u = (ξ 3 )v = (ξ 3 )w =0
(ξ 2 )p = (ξ 2 )u = (ξ 2 )v = (ξ 2 )w =0
(65.28)
(ξ 1 )p = (ξ 1 )u = (ξ 1 )v = (ξ 1 )w =0
(ξ 4 )p = (ξ 4 )u = (ξ 4 )v = (ξ 4 )w =0
From (65.27) we find that

The terms uxx , uyy and uzz have common coefficient ν(φ1 )p . The terms
vxx , vyy and vzz have common coefficient ν(φ2 )p . The terms wxx , wyy and
wzz have common coefficient ν(φ3 )p . We thus have
(φ1 )p = (φ2 )p = (φ3 )p = 0 (65.29)
Taking into account (65.29), and equating to zero the coefficients of ut , vt

and wt , we obtain from (65.27) that
(ξ 4 )x = (ξ 4 )y = (ξ 4 )z = 0 (65.30)
769
We thus see that equation (65.27) becomes
(φ1 )x + ux (φ1 )u + vx (φ1 )v + wx (φ1 )w

− ux (ξ 1 )x − uy (ξ 2 )x − uz (ξ 3 )x
+ (φ2 )y + uy (φ2 )u + vy (φ2 )v + wy (φ2 )w
(65.31)
− vx (ξ 1 )y − vy (ξ 2 )y − vz (ξ 3 )y
− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z = 0
(φ1 )u − (ξ 1 )x − (φ3 )w + (ξ 3 )z = 0 (65.32)
−(ξ 2 )x + (φ2 )u = 0 (65.33)
−(ξ 3 )x + (φ3 )u = 0 (65.34)
(φ1 )v − (ξ 1 )y = 0 (65.35)
(φ2 )v − (ξ 2 )y − (φ3 )w + (ξ 3 )z = 0 (65.36)
−(ξ 3 )y + (φ3 )v = 0 (65.37)
(φ1 )w − (ξ 1 )z = 0 (65.38)
(φ2 )w − (ξ 2 )z = 0 (65.39)
770
(φ1 )x + (φ2 )y + (φ3 )z = 0 (65.40)
Let us consider next equation P r(2) X[∆(1) ] = 0.

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
This equation is equivalent to
(φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w

+ (φ1 )ux + (φ2 )uy + (φ3 )uz


1 2 3 4
− ux (ξ )x − uy (ξ )x − uz (ξ )x − ut (ξ )x


− ux (ξ 1 )y − uy (ξ 2 )y − uz (ξ 3 )y − ut (ξ 4 )y


− ux (ξ 1 )z − uy (ξ 2 )z − uz (ξ 3 )z − ut (ξ 4 )z
+ (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w

+ [(ut + uux + vuy + wuz ) − ν(uxx + uyy + uzz )]×[(ξ 1 )x − (φ4 )p ]
+ [(vt + uvx + vvy + wvz ) − ν(vxx + vyy + vzz )]×[(ξ 2 )x ]
+ [(wt + uwx + vwy − (ux + vy )w) − ν(wxx + wyy + wzz )]×[(ξ 3 )x ]

− ν − uxx [2(ξ 1 )x − (φ1 )u ] − 2uxy (ξ 2 )x − 2uxz (ξ 3 )x
+ vxx (φ1 )v + wxx (φ1 )w + (ux )2 (φ1 )uu + (vx )2 (φ1 )vv + (wx )2 (φ1 )ww
− ux (ξ 1 )xx − uy (ξ 2 )xx − uz (ξ 3 )xx
+ 2ux (φ1 )xu + 2vx (φ1 )xv + 2wx (φ1 )xw + 2ux vx (φ1 )uv

1 1 1
+ 2ux wx (φ )uw + 2vx wx (φ )vw + (φ )xx
771

− ν − uyy [2(ξ 2 )y − (φ1 )u ] − 2uxy (ξ 1 )y − 2uyz (ξ 3 )y
+ vyy (φ1 )v + wyy (φ1 )w + (uy )2 (φ1 )uu + (vy )2 (φ1 )vv + (wy )2 (φ1 )ww
− ux (ξ 1 )yy − uy (ξ 2 )yy − uz (ξ 3 )yy
+ 2uy (φ1 )yu + 2vy (φ1 )yv + 2wy (φ1 )yw + 2uy vy (φ1 )uv

1 1 1
+ 2uy wy (φ )uw + 2vy wy (φ )vw + (φ )yy

− ν − uzz [2(ξ 3 )z − (φ1 )u ] − 2uxz (ξ 1 )z − 2uyz (ξ 2 )z
+ vzz (φ1 )v + wzz (φ1 )w + (uz )2 (φ1 )uu + (vz )2 (φ1 )vv
+ (ux + vy )2 (φ1 )ww − ux (ξ 1 )zz − uy (ξ 2 )zz − uz (ξ 3 )zz
+ 2uz (φ1 )zu + 2vz (φ1 )zv − 2(ux + vy )(φ1 )zw + 2uz vz (φ1 )uv

1 1 1
− 2uz (ux + vy )(φ )uw − 2vz (ux + vy )(φ )vw + (φ )zz = 0
In deriving the above equation, we substituted in equation (65.8) the ex-

pressions for the coefficients φt , φx , φy , φz , π x and φxx , φyy , φzz , we then made
the substitutions (65.22) and then we used the values (65.28), (65.29) and
(65.30).
The terms ux vx , uy vy and uz vz have common coefficient −2ν(φ1 )uv . The
terms ux wx , uy wy have common coefficient −2ν(φ1 )uw . The terms uz ux ,
uz vy have common coefficient 2ν(φ1 )uw . The terms vx wx , vy wy have com-
mon coefficient −2ν(φ1 )vw . The terms vz ux , vz vy have common coefficient
2ν(φ1 )vw . We thus have
(φ1 )uv = (φ1 )uw = (φ1 )vw = 0 (65.41)
The coefficients of (ux )2 , (uy )2 and (uz )2 are −ν(φ1 )uu − ν(φ1 )ww , −ν(φ1 )uu
and −ν(φ1 )uu respectively. The coefficients of (vx )2 , (vy )2 and (vz )2 are
−ν(φ1 )vv , −ν(φ1 )vv − ν(φ1 )ww and −ν(φ1 )vv respectively. The terms (wx )2
and (wy )2 have common coefficient −ν(φ1 )ww . We thus have
(φ1 )uu = (φ1 )vv = (φ1 )ww = 0 (65.42)
Equating the coefficients of uxy , uxz and uyz to zero we obtain the equations
(ξ 2 )x + (ξ 1 )y = 0 (65.43)
772
(ξ 3 )x + (ξ 1 )z = 0 (65.44)
and
(ξ 3 )y + (ξ 2 )z = 0 (65.45)
Equating the coefficients of uxx , uyy and uzz to zero, we obtain the equations
(ξ 1 )x − (φ1 )u + (φ4 )p = 0 (65.46)
2(ξ 2 )y − (ξ 1 )x − (φ1 )u + (φ4 )p = 0 (65.47)

and
2(ξ 3 )z − (ξ 1 )x − (φ1 )u + (φ4 )p = 0 (65.48)
The terms vxx , vyy and vzz have common coefficient −ν[(ξ 2 )x + (φ1 )v ], while
the terms wxx , wyy and wzz have common coefficient −ν[(ξ 3 )x + (φ1 )w ]. We
thus obtain two more equations
(ξ 2 )x + (φ1 )v = 0 (65.49)
and
(ξ 3 )x + (φ1 )w = 0 (65.50)
Because of the above equations, we are left with
(φ1 )t + ut (φ1 )u + vt (φ1 )v + wt (φ1 )w

+ (φ1 )ux + (φ2 )uy + (φ3 )uz


1 2 3
− ux (ξ )x − uy (ξ )x − uz (ξ )x


− ux (ξ 1 )y − uy (ξ 2 )y − uz (ξ 3 )y


− ux (ξ 1 )z − uy (ξ 2 )z − uz (ξ 3 )z
773
+ (φ4 )x + ux (φ4 )u + vx (φ4 )v + wx (φ4 )w
+ (ut + uux + vuy + wuz )×[(ξ 1 )x − (φ4 )p ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )x ]
+ [(wt + uwx + vwy − (ux + vy )w)]×[(ξ 3 )x ]

− ν − ux (ξ 1 )xx − uy (ξ 2 )xx − uz (ξ 3 )xx

1 1 1 1
+ 2ux (φ )xu + 2vx (φ )xv + 2wx (φ )xw + (φ )xx

− ν − ux (ξ 1 )yy − uy (ξ 2 )yy − uz (ξ 3 )yy

1 1 1 1
+ 2uy (φ )yu + 2vy (φ )yv + 2wy (φ )yw + (φ )yy

− ν − ux (ξ 1 )zz − uy (ξ 2 )zz − uz (ξ 3 )zz

1 1 1 1
+ 2uz (φ )zu + 2vz (φ )zv − 2(ux + vy )(φ )zw + (φ )zz = 0
From the above equation we find that

(φ1 )u − (ξ 4 )t + [(ξ 1 )x − (φ4 )p ] = 0 (65.51)
(φ1 )v + (ξ 2 )x = 0 (65.52)
(φ1 )w + (ξ 3 )x = 0 (65.53)
− (ξ 1 )t + φ1 + u(φ1 )u − u(ξ 1 )x − v(ξ 1 )y − w(φ1 )w − w(ξ 1 )z
+ (φ4 )u + u[(ξ 1 )x − (φ4 )p ] − w(ξ 3 )x + ν(ξ 1 )xx − 2ν(φ1 )xu (65.54)
+ ν(ξ 1 )yy + ν(ξ 1 )zz + 2ν(φ1 )zw = 0
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ1 )u − v(ξ 2 )y − w(ξ 2 )z
(65.55)
+ v[(ξ 1 )x − (φ4 )p ] + ν(ξ 2 )xx + ν(ξ 2 )yy − 2ν(φ1 )yu + ν(ξ 2 )zz = 0
774
− (ξ 3 )t + φ3 − u(ξ 3 )x − v(ξ 3 )y + w(φ1 )u − w(ξ 3 )z

(65.56)
+ w[(ξ 1 )x − (φ4 )p ] + ν(ξ 3 )xx + ν(ξ 3 )yy + ν(ξ 3 )zz − 2ν(φ1 )zu = 0
u(φ1 )v + (φ4 )v + u(ξ 2 )x − 2ν(φ1 )xv = 0 (65.57)
v(φ1 )v − w(φ1 )w + v(ξ 2 )x − w(ξ 3 )x − 2ν(φ1 )yv + 2ν(φ1 )zw = 0 (65.58)
w(φ1 )v + w(ξ 2 )x − 2ν(φ1 )zv = 0 (65.59)
u(φ1 )w + (φ4 )w + u(ξ 3 )x − 2ν(φ1 )xw = 0 (65.60)
v(φ1 )w + v(ξ 3 )x − 2ν(φ1 )yw = 0 (65.61)
The constant term (the terms which do not include partial derivatives of the
unknown function)
(φ1 )t + u(φ1 )x + v(φ1 )y + w(φ1 )z + (φ4 )x

(65.62)
− ν(φ1 )xx − ν(φ1 )yy − ν(φ1 )zz = 0
775
Equation P r(2) X[∆(2) ] = 0 is equivalent to the
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
equation
(φ2 )t + ut (φ2 )u + vt (φ2 )v + wt (φ2 )w

+ (φ1 )vx + (φ2 )vy + (φ3 )vz


− vx (ξ 1 )x − vy (ξ 2 )x − vz (ξ 3 )x


1 2 3
− vx (ξ )y − vy (ξ )y − vz (ξ )y


− vx (ξ 1 )z − vy (ξ 2 )z − vz (ξ 3 )z
+ (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w

+ [(ut + uux + vuy + wuz ) − ν(uxx + uyy + uzz )]×[(ξ 1 )y ]
+ [(vt + uvx + vvy + wvz ) − ν(vxx + vyy + vzz )]×[(ξ 2 )y − (φ4 )p ]
+ [(wt + uwx + vwy − (ux + vy )w) − ν(wxx + wyy + wzz )]×[(ξ 3 )y ]

− ν − vxx [2(ξ 1 )x − (φ2 )v ] − 2vxy (ξ 2 )x − 2vxz (ξ 3 )x + uxx (φ2 )u
+ wxx (φ2 )w + (vx )2 (φ2 )vv + (ux )2 (φ2 )uu + (wx )2 (φ2 )ww
− vx (ξ 1 )xx − vy (ξ 2 )xx − vz (ξ 3 )xx

2 2 2
776

− ν − vyy [2(ξ 2 )y − (φ2 )v ] − 2vxy (ξ 1 )y − 2vyz (ξ 3 )y + uyy (φ2 )u
+ wyy (φ2 )w + (vy )2 (φ2 )vv + (uy )2 (φ2 )uu + (wy )2 (φ2 )ww
− vx (ξ 1 )yy − vy (ξ 2 )yy − vz (ξ 3 )yy

2 2 2

− ν − vzz [2(ξ 3 )z − (φ2 )v ] − 2vxz (ξ 1 )z − 2vyz (ξ 2 )z + uzz (φ2 )u
+ wzz (φ2 )w + (vz )2 (φ2 )vv + (uz )2 (φ2 )uu + (ux + vy )2 (φ2 )ww
− vx (ξ 1 )zz − vy (ξ 2 )zz − vz (ξ 3 )zz

2 2 2
In deriving the above equation, we substituted in (65.9) the various coeffi-

cients φt , φx , · · · we have also made the substitutions (65.22) and we have
taken into account (65.28), (65.29) and (65.30).
equate to zero the easily identified coefficients of some partial derivatives.
In fact, we see that the coefficients of ux vx , ux wx and vx wx are −2ν(φ2 )uv ,
−2ν(φ2 )uw and −2ν(φ2 )vw respectively. Similarly the coefficients of uy vy ,
uy wy and vy wy are −2ν(φ2 )uv , −2ν(φ2 )uw and −2ν(φ2 )vw respectively. The
coefficients of uz vz , uz wz and vz wz are −2ν(φ2 )uv , −2ν(φ2 )uw and −2ν(φ2 )vw
respectively. Equating all these coefficients to zero we get
(φ2 )uv = (φ2 )uw = (φ2 )vw = 0 (65.63)
The terms (ux )2 , (uy )2 and (uz )2 have common coefficient −ν(φ2 )uu . The
terms (vx )2 , (vy )2 and (vz )2 have common coefficient −ν(φ2 )vv . The terms
(wx )2 , (wy )2 and (wz )2 have common coefficient −ν(φ2 )ww . Equating all these
coefficients to zero, we have
(φ2 )uu = (φ2 )vv = (φ2 )ww = 0 (65.64)
777
The coefficients of vxy , vxz and vyz are 2ν[(ξ 2 )x + (ξ 1 )y ], 2ν[(ξ 3 )x + (ξ 1 )z ] and
2ν[(ξ 3 )y + (ξ 2 )z ]respectively. Equating to zero all these coefficients, we have
(ξ 2 )x + (ξ 1 )y = 0, (ξ 3 )x + (ξ 1 )z = 0, (ξ 3 )y + (ξ 2 )z = 0 (65.65)
The terms uxx , uyy and uzz have common coefficient −ν[(ξ 1 )y + (φ2 )u ] and
thus
(ξ 1 )y + (φ2 )u = 0 (65.66)
The terms wxx , wyy and wzz have common coefficient −ν[(ξ 3 )y + (φ2 )w ] and
thus
(ξ 3 )y + (φ2 )w = 0 (65.67)
The coefficient of vxx is −ν[(ξ 2 )y − (φ4 )p ] + ν[2(ξ 1 )x − (φ2 )v ] and then we have
the equation
−[(ξ 2 )y − (φ4 )p ] + [2(ξ 1 )x − (φ2 )v ] = 0 (65.68)
The coefficient of vyy is −ν[(ξ 2 )y − (φ4 )p ] + ν[2(ξ 2 )y − (φ2 )v ] and thus we get
the equation
−[(ξ 2 )y − (φ4 )p ] + [2(ξ 2 )y − (φ2 )v ] = 0 (65.69)
The coefficient of vzz is −ν[(ξ 2 )y − (φ4 )p ] + ν[2(ξ 3 )z − (φ2 )v ] and thus we

obtain the equation
−[(ξ 2 )y − (φ4 )p ] + [2(ξ 3 )z − (φ2 )v ] = 0 (65.70)
778
We thus see that equation P r(2) X[∆(2) ] =0
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
gets simplified into
(φ2 )t + ut (φ2 )u + vt (φ2 )v + wt (φ2 )w

+ (φ1 )vx + (φ2 )vy + (φ3 )vz


− vx (ξ 1 )x − vy (ξ 2 )x − vz (ξ 3 )x


1 2 3
− vx (ξ )y − vy (ξ )y − vz (ξ )y


− vx (ξ 1 )z − vy (ξ 2 )z − vz (ξ 3 )z
+ (φ4 )y + uy (φ4 )u + vy (φ4 )v + wy (φ4 )w

+ (ut + uux + vuy + wuz )×[(ξ 1 )y ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )y − (φ4 )p ]
+ [(wt + uwx + vwy − (ux + vy )w)]×[(ξ 3 )y ]

− ν − vx (ξ 1 )xx − vy (ξ 2 )xx − vz (ξ 3 )xx

2 2 2 2

− ν − vx (ξ 1 )yy − vy (ξ 2 )yy − vz (ξ 3 )yy

2 2 2 2

− ν − vx (ξ 1 )zz − vy (ξ 2 )zz − vz (ξ 3 )zz

2 2 2 2
779
We may proceed now for the other coefficients using the above equation. We
have that
(φ2 )u + (ξ 1 )y = 0 (65.71)
(φ2 )v − (ξ 4 )t + [(ξ 2 )y − (φ4 )p ] = 0 (65.72)
(φ2 )w + (ξ 3 )y = 0 (65.73)
u(φ2 )u − w(φ2 )w + u(ξ 1 )y − w(ξ 3 )y − 2ν(φ2 )xu = 0 (65.74)
− (ξ 1 )t + φ1 + u(φ2 )v − u(ξ 1 )x − v(ξ 1 )y − w(ξ 1 )z

(65.75)
+ u[(ξ 2 )y − (φ4 )p ] + ν(ξ 1 )xx − 2ν(φ2 )xv + ν(ξ 1 )yy + ν(ξ 1 )zz = 0
u(φ2 )w + u(ξ 3 )y − 2ν(φ2 )xw = 0 (65.76)
v(φ2 )u + (φ4 )u + v(ξ 1 )y − 2ν(φ2 )yu = 0 (65.77)
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ2 )v − v(ξ 2 )y − w(φ2 )w − w(ξ 2 )z

+ (φ4 )v + v[(ξ 2 )y − (φ4 )p ] − w(ξ 3 )y + ν(ξ 2 )xx + ν(ξ 2 )yy (65.78)
− 2ν(φ2 )yv + ν(ξ 2 )zz + 2ν(φ2 )zw = 0
v(φ2 )w + (φ4 )w + v(ξ 3 )y − 2ν(φ2 )yw = 0 (65.79)
780
w(φ2 )u + w(ξ 1 )y − 2ν(φ2 )zu = 0 (65.80)
− (ξ 3 )t + φ3 − u(ξ 3 )x − v(ξ 3 )y + w(φ2 )v − w(ξ 3 )z

(65.81)
+ w[(ξ 2 )y − (φ4 )p ] + ν(ξ 3 )xx + ν(ξ 3 )yy + ν(ξ 3 )zz − 2ν(φ2 )zv = 0
(φ2 )t + u(φ2 )x + v(φ2 )y + w(φ2 )z + (φ4 )y

(65.82)
− ν(φ2 )xx − ν(φ2 )yy − ν(φ2 )zz = 0
Equation P r(2) X[∆(3) ] = 0 is equivalent to

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
(φ )t + ut (φ )u + vt (φ3 )v + wt (φ3 )w
3 3
+ (φ1 )wx + (φ2 )wy − (ux + vy )(φ3 )


− wx (ξ 1 )x − wy (ξ 2 )x + (ux + vy )(ξ 3 )x


− wx (ξ 1 )y − wy (ξ 2 )y + (ux + vy )(ξ 3 )y


− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z

+ [(ut + uux + vuy + wuz ) − ν(uxx + uyy + uzz )]×[(ξ 1 )z ]
+ [(vt + uvx + vvy + wvz ) − ν(vxx + vyy + vzz )]×[(ξ 2 )z ]
+ [(wt + uwx + vwy − (ux + vy )w) − ν(wxx + wyy + wzz )]×[(ξ 3 )z − (φ4 )p ]
781

− ν − wxx [2(ξ 1 )x − (φ3 )w ] − 2wxy (ξ 2 )x − 2wxz (ξ 3 )x + uxx (φ3 )u
+ vxx (φ3 )v + (wx )2 (φ3 )ww + (ux )2 (φ3 )uu + (vx )2 (φ3 )vv
− wx (ξ 1 )xx − wy (ξ 2 )xx + (ux + vy )(ξ 3 )xx

3 3 3

− ν − wyy [2(ξ 2 )y − (φ3 )w ] − 2wxy (ξ 1 )y − 2wyz (ξ 3 )y + uyy (φ3 )u
+ vyy (φ3 )v + (wy )2 (φ3 )ww + (uy )2 (φ3 )uu + (vy )2 (φ3 )vv
− wx (ξ 1 )yy − wy (ξ 2 )yy + (ux + vy )(ξ 3 )yy

3 3 3

− ν − wzz [2(ξ 3 )z − (φ3 )w ] − 2wxz (ξ 1 )z − 2wyz (ξ 2 )z + uzz (φ3 )u
+ vzz (φ3 )v + (ux + vy )2 (φ3 )ww + (uz )2 (φ3 )uu + (vz )2 (φ3 )vv
− wx (ξ 1 )zz − wy (ξ 2 )zz + (ux + vy )(ξ 3 )zz

3 3 3
In deriving the above equation, we substituted in (65.10) the various coef-

ficients φt , φx , · · · we have also made the substitutions (65.22) and we have
taken into account (65.28), (65.29) and (65.30).
equate to zero the easily identified coefficients of some partial derivatives.
In fact, we see that the coefficients of ux vx , ux wx and vx wx are −2ν(φ3 )uv ,
−2ν(φ3 )uw and −2ν(φ3 )vw respectively. Similarly the coefficients of uy vy ,
uy wy and vy wy are −2ν(φ3 )uv , −2ν(φ3 )uw and −2ν(φ3 )vw respectively. The
coefficients of uz vz , uz ux and uz vy are −2ν(φ3 )uv , 2ν(φ3 )uw and 2ν(φ3 )uw
respectively. The terms vz ux and vz vy have common coefficient 2ν(φ3 )vw .
Equating all these coefficients to zero we get
(φ3 )uv = (φ3 )uw = (φ3 )vw = 0 (65.83)
782
The term ux vy has coefficient −2ν(φ3 )ww . The term (ux )2 has coefficient
−ν(φ3 )uu − ν(φ3 )ww , while the terms (uy )2 and (uz )2 have common coeffi-
cient −ν(φ3 )uu . The terms (vx )2 , (vy )2 and (vz )2 have coefficients −ν(φ3 )vv ,
−ν(φ3 )vv − ν(φ3 )ww and −ν(φ3 )vv respectively. The terms (wx )2 and (wy )2
have common coefficient −ν(φ3 )ww . Equating all these coefficients to zero,
we have
(φ3 )uu = (φ3 )vv = (φ3 )ww = 0 (65.84)
The coefficients of wxy , wxz and wyz are 2ν[(ξ 2 )x + (ξ 1 )y ], 2ν[(ξ 3 )x + (ξ 1 )z ]

and 2ν[(ξ 3 )y + (ξ 2 )z ]respectively. Equating to zero all these coefficients, we
have
(ξ 2 )x + (ξ 1 )y = 0, (ξ 3 )x + (ξ 1 )z = 0, (ξ 3 )y + (ξ 2 )z = 0 (65.85)
The terms uxx , uyy and uzz have common coefficient −ν[(ξ 1 )z + (φ3 )u ] and
thus
(ξ 1 )z + (φ3 )u = 0 (65.86)
The terms vxx , vyy and vzz have common coefficient −ν[(ξ 2 )z + (φ3 )v ] and
thus
(ξ 2 )z + (φ3 )v = 0 (65.87)
The coefficient of wxx is −ν[(ξ 3 )z − (φ4 )p ] + ν[2(ξ 1 )x − (φ3 )w ] and then we

have the equation
−[(ξ 3 )z − (φ4 )p ] + [2(ξ 1 )x − (φ3 )w ] = 0 (65.88)
The coefficient of wyy is −ν[(ξ 3 )z − (φ4 )p ] + ν[2(ξ 2 )y − (φ3 )w ] and thus we get
the equation
−[(ξ 3 )z − (φ4 )p ] + [2(ξ 2 )y − (φ3 )w ] = 0 (65.89)
The coefficient of wzz is −ν[(ξ 3 )z − (φ4 )p ] + ν[2(ξ 3 )z − (φ3 )w ] and thus we

obtain the equation
−[(ξ 3 )z − (φ4 )p ] + [2(ξ 3 )z − (φ3 )w ] = 0 (65.90)
783
We thus see that equation P r(2) X[∆(3) ] =0
∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
gets simplified into
(φ3 )t + ut (φ3 )u + vt (φ3 )v + wt (φ3 )w

+ (φ1 )wx + (φ2 )wy − (ux + vy )(φ3 )


− wx (ξ 1 )x − wy (ξ 2 )x + (ux + vy )(ξ 3 )x


1 2 3
− wx (ξ )y − wy (ξ )y + (ux + vy )(ξ )y


− wx (ξ 1 )z − wy (ξ 2 )z + (ux + vy )(ξ 3 )z

+ (ut + uux + vuy + wuz )×[(ξ 1 )z ]
+ (vt + uvx + vvy + wvz )×[(ξ 2 )z ]
+ [wt + uwx + vwy − (ux + vy )w]×[(ξ 3 )z − (φ4 )p ]

− ν − wx (ξ 1 )xx − wy (ξ 2 )xx + (ux + vy )(ξ 3 )xx

3 3 3 3

− ν − wx (ξ 1 )yy − wy (ξ 2 )yy + (ux + vy )(ξ 3 )yy

3 3 3 3

− ν − wx (ξ 1 )zz − wy (ξ 2 )zz + (ux + vy )(ξ 3 )zz

3 3 3 3
784
Let us now proceed further by equating to zero the rest of the coefficients of
the above equation. We find that
(φ3 )u + (ξ 1 )z = 0 (65.91)
(φ3 )v + (ξ 2 )z = 0 (65.92)
(φ3 )w − (ξ 4 )t + [(ξ 3 )z − (φ4 )p ] = 0 (65.93)
(ξ 3 )t − φ3 + u(φ3 )u + u(ξ 3 )x + v(ξ 3 )y − w(φ3 )w + w(ξ 3 )z
− (φ4 )w + u(ξ 1 )z − w[(ξ 3 )z − (φ4 )p ] − ν(ξ 3 )xx − 2ν(φ3 )xu (65.94)
− ν(ξ 3 )yy − ν(ξ 3 )zz − 2ν(φ3 )zw = 0
u(φ3 )v + u(ξ 2 )z − 2ν(φ3 )xv = 0 (65.95)
− (ξ 1 )t + φ1 + u(φ3 )w − u(ξ 1 )x − v(ξ 1 )y − w(ξ 1 )z
(65.96)
+ u[(ξ 3 )z − (φ4 )p ] + ν(ξ 1 )xx − 2ν(φ3 )xw + ν(ξ 1 )yy + ν(ξ 1 )zz = 0
v(φ3 )u + v(ξ 1 )z − 2ν(φ3 )yu = 0 (65.97)
The coefficients of vy is
(ξ 3 )t − φ3 + u(ξ 3 )x + v(φ3 )v + v(ξ 3 )y − w(φ3 )w + w(ξ 3 )z

− (φ4 )w + v(ξ 2 )z − w[(ξ 3 )z − (φ4 )p ] − ν(ξ 3 )xx − ν(ξ 3 )yy (65.98)
− 2ν(φ3 )yv − ν(ξ 3 )zz + 2ν(φ3 )zw = 0
− (ξ 2 )t + φ2 − u(ξ 2 )x + v(φ3 )w − v(ξ 2 )y − w(ξ 2 )z

(65.99)
+ v[(ξ 3 )z − (φ4 )p ] + ν(ξ 2 )xx + ν(ξ 2 )yy − 2ν(φ3 )yw + ν(ξ 2 )zz = 0
785
w(φ3 )u + (φ4 )u + w(ξ 1 )z − 2ν(φ3 )zu = 0 (65.100)
w(φ3 )v + (φ4 )v + w(ξ 2 )z − 2ν(φ3 )zv = 0 (65.101)
(φ3 )t + u(φ3 )x + v(φ3 )y + w(φ3 )z + (φ4 )z − ν(φ3 )xx

(65.102)
− ν(φ3 )yy − ν(φ3 )zz = 0
We shall now start solving the determining equations of the Lie symmetries
of Navier-Stokes equations.
Using equations (65.41) and (65.42), and similarly (65.63) and (65.64) and
then (65.83) and (65.84) in pairs, we see that the infinitesimals φ1 , φ2 and φ3
are linear functions of the variables u, v, w and thus admit the representation
φ1 = A1 (x, y, z, t)u+B1 (x, y, z, t)v +C1 (x, y, z, t)w +D1 (x, y, z, t) (65.103)
and
respectively.
Equation (65.62), using the above expressions for φ1 , φ2 and φ3 , takes on the
form
{(A1 )t + (D1 )x − ν[(A1 )xx + (A1 )yy + (A1 )zz ]}u
+ {(B1 )t + (D1 )y − ν[(B1 )xx + (B1 )yy + (B1 )zz ]}v
+ {(C1 )t + (D1 )z − ν[(C1 )xx + (C1 )yy + (C1 )zz ]}w
(65.106)
+ (A1 )x u2 + (B1 )y v 2 + (C1 )z w2 + {(B1 )x + (A1 )y }uv
+ {(C1 )x + (A1 )z }uw + {(C1 )y + (B1 )z }vw
+ (D1 )t + (φ4 )x − ν[(D1 )xx + (D1 )yy + (D1 )zz ] = 0
786
Equating the coefficients of the various variables u, v, w to zero, we obtain
the systems of equations
(A1 )t + (D1 )x − ν[(A1 )xx + (A1 )yy + (A1 )zz ] = 0

(B1 )t + (D1 )y − ν[(B1 )xx + (B1 )yy + (B1 )zz ] = 0 (65.107)
(C1 )t + (D1 )z − ν[(C1 )xx + (C1 )yy + (C1 )zz ] = 0
(A1 )x = 0, (B1 )y = 0, (C1 )z = 0

(65.108)
(B1 )x + (A1 )y = 0, (C1 )x + (A1 )z = 0, (C1 )y + (B1 )z = 0
(D1 )t + (φ4 )x − ν[(D1 )xx + (D1 )yy + (D1 )zz ] = 0 (65.109)
Equation (65.82), using the above expressions for φ1 , φ2 and φ3 , given by
(65.103)-(65.105), takes on the form

(65.110)
+ {(C2 )x + (A2 )z }uw + {(C2 )y + (B2 )z }vw
+ (D2 )t + (φ4 )y − ν[(D2 )xx + (D2 )yy + (D2 )zz ] = 0

(A2 )t + (D2 )x − ν[(A2 )xx + (A2 )yy + (A2 )zz ] = 0

(B2 )t + (D2 )y − ν[(B2 )xx + (B2 )yy + (B2 )zz ] = 0 (65.111)
(C2 )t + (D2 )z − ν[(C2 )xx + (C2 )yy + (C2 )zz ] = 0
(A2 )x = 0, (B2 )y = 0, (C2 )z = 0

(65.112)
(B2 )x + (A2 )y = 0, (C2 )x + (A2 )z = 0, (C2 )y + (B2 )z = 0
(D2 )t + (φ4 )y − ν[(D2 )xx + (D2 )yy + (D2 )zz ] = 0 (65.113)
787
Equation (65.102), using the above expressions for φ1 , φ2 and φ3 , given by
(65.103)-(65.105), takes on the form

(65.114)
+ {(C3 )x + (A3 )z }uw + {(C3 )y + (B3 )z }vw
+ (D3 )t + (φ4 )z − ν[(D3 )xx + (D3 )yy + (D3 )zz ] = 0

(A3 )t + (D3 )x − ν[(A3 )xx + (A3 )yy + (A3 )zz ] = 0

(B3 )t + (D3 )y − ν[(B3 )xx + (B3 )yy + (B3 )zz ]} = 0 (65.115)
(C3 )t + (D3 )z − ν[(C3 )xx + (C3 )yy + (C3 )zz ] = 0
(A3 )x = 0, (B3 )y = 0, (C3 )z = 0

(65.116)
(B3 )x + (A3 )y = 0, (C3 )x + (A3 )z = 0, (C3 )y + (B3 )z = 0
(D3 )t + (φ4 )z − ν[(D3 )xx + (D3 )yy + (D3 )zz ] = 0 (65.117)
Since (A1 )x = 0, (B1 )y = 0, (C1 )z = 0, (A2 )x = 0, (B2 )y = 0, (C2 )z = 0 and
(A3 )x = 0, (B3 )y = 0, (C3 )z = 0, according to the systems (65.108), (65.112)
and (65.116) respectively, we can rewrite φ1 , φ2 , φ3 as

and
Equation (65.40), taking into account the above expressions for the infinites-
imals φ1 , φ2 , φ3 takes on the form
{(A2 )y + (A3 )z }u + {(B1 )x + (B3 )z }v + {(C1 )x + (C2 )y }w

(65.121)
+ (D1 )x + (D2 )y + (D3 )z = 0
788
We thus obtain from the above equation the following systems
(A2 )y + (A3 )z = 0, (B1 )x + (B3 )z = 0, (C1 )x + (C2 )y = 0 (65.122)
and
(D1 )x + (D2 )y + (D3 )z = 0 (65.123)
We come next to find the general expressions for ξ 1 , ξ 2 and ξ 3 . Equations

(65.35) and (65.38), taking into account the expression (65.118) for φ1 , we
get the two equations
B1 (x, z, t) = (ξ 1 )y and C1 (x, y, t) = (ξ 1 )z (65.124)
Upon differentiation of the first equation with respect to y and the second
with respect to z, we get (ξ 1 )yy = 0 and (ξ 1 )zz = 0 respectively, which means
that ξ 1 is a linear function with respect to y and z and thus admits the
representation
ξ 1 = α1 (x, t) + β1 (t)y + γ1 (t)z (65.125)
since ξ 1 does not depend on any other variables, due to (65.28).

for φ2 , we get the two equations
A2 (y, z, t) = (ξ 2 )x and C2 (x, y, t) = (ξ 2 )z (65.126)
Upon differentiation of the first equation with respect to x and the second
with respect to z, we get (ξ 2 )xx = 0 and (ξ 2 )zz = 0 respectively, which means
that ξ 2 is a linear function with respect to x and z and thus admits the
representation
ξ 2 = α2 (t)x + β2 (y, t) + γ2 (t)z (65.127)

for φ3 , we get the two equations
A3 (y, z, t) = (ξ 3 )x and B3 (x, z, t) = (ξ 3 )y (65.128)
Upon differentiation of the first equation with respect to x and the second
with respect to y, we get (ξ 3 )xx = 0 and (ξ 3 )yy = 0 respectively, which means
789
that ξ 3 is a linear function with respect to x and y and thus admits the
representation
ξ 3 = α3 (t)x + β3 (t)y + γ3 (z, t) (65.129)

Equations (65.49) and (65.52) are identical. The same is true for (65.50) and
(65.53). These equations take the form
α2 (t) + B1 (x, z, t) = 0, α3 (t) + C1 (x, y, t) = 0
using the expressions for ξ 2 , ξ 3 and φ1 . We thus obtain from the above two
equations that
B1 = B1 (t) and α2 (t) + B1 (t) = 0, i.e. α2 (t) = −B1 (t) (65.130)
as well as
C1 = C1 (t) and α3 (t) + C1 (t) = 0, i.e. α3 (t) = −C1 (t) (65.131)
β1 (t) + A2 (y, z, t) = 0, β3 (t) + C2 (x, y, t) = 0
equations that
A2 = A2 (t) and β1 (t) + A2 (t) = 0, i.e. β1 (t) = −A2 (t) (65.132)
as well as
C2 = C2 (t) and β3 (t) + C2 (t) = 0, i.e. β3 (t) = −C2 (t) (65.133)
γ1 (t) + A3 (y, z, t) = 0, γ2 (t) + B3 (x, z, t) = 0
equations that
A3 = A3 (t) and γ1 (t) + A3 (t) = 0, i.e. γ1 (t) = −A3 (t) (65.134)
790
as well as
B3 = B3 (t) and γ2 (t) + B3 (t) = 0, i.e. γ2 (t) = −B3 (t) (65.135)
We can then rewrite φ1 , φ2 , φ3 from (64.118)-(64.120), using the above found

expressions of the various coefficients, as
φ1 = A1 (y, z, t)u + B1 (t)v + C1 (t)w + D1 (x, y, z, t) (65.136)
φ2 = A2 (t)u + B2 (x, z, t)v + C2 (t)w + D2 (x, y, z, t) (65.137)

and
φ3 = A3 (t)u + B3 (t)v + C3 (x, y, t)w + D3 (x, y, z, t) (65.138)
We also have the following expressions for ξ 1 , ξ 2 and ξ 3 :
ξ 1 = α1 (x, t) − A2 (t)y − A3 (t)z (65.139)
ξ 2 = −B1 (t)x + β2 (y, t) − B3 (t)z (65.140)

and
ξ 3 = −C1 (t)x − C2 (t)y + γ3 (z, t) (65.141)
The second line of (65.108) contains three equations. From the first one,
(B1 )x +(A1 )y = 0, since B1 = B1 (t), we get (A1 )y = 0 and thus A1 = A1 (z, t).
From the second equation, (C1 )x + (A1 )z = 0, since C1 = C1 (t), we get
(A1 )z = 0 and thus A1 = A1 (t). The third equation, (C1 )y + (B1 )z = 0, is an
identity, since both C1 and B1 depend on t only.
The second line of (65.112) contains three equations. From the first one,
(B2 )x +(A2 )y = 0, since A2 = A2 (t), we get (B2 )x = 0 and thus B2 = B2 (z, t).
The second equation, (C2 )x + (A2 )z = 0, is an identity, since both C2 and
A2 depend on t only. From the third equation, (C2 )y + (B2 )z = 0, since
C2 = C2 (t), we get (B2 )z = 0 and thus B2 = B2 (t).
The second line of (65.116) contains three equations. The first one, (B3 )x +
(A3 )y = 0, is an identity, since both B3 and A3 depend on t only. From the
second equation, (C3 )x + (A3 )z = 0, since A3 = A3 (t), we get (C3 )x = 0 and
thus C3 = C3 (y, t). The third equation, (C3 )y + (B3 )z = 0, since B3 = B3 (t),
we get (C3 )y = 0 and thus C3 = C3 (t).
791
We thus see that since A1 = A1 (t), B2 = B2 (t) and C3 = C3 (t), expressions
(65.136)-(65.138) have the new form
φ1 = A1 (t)u + B1 (t)v + C1 (t)w + D1 (x, y, z, t) (65.142)

and
The system (65.107) gets simplified into
(A1 )t + (D1 )x = 0, (B1 )t + (D1 )y = 0, (C1 )t + (D1 )z = 0 (65.145)
and it is thus a system of PDEs containing the function D1 as the unknown

function. From the first equation, (D1 )x = −A01 (t) we get by integration
D1 = −A01 (t)x + δ1 (y, z, t). Taking into account this expression of D1 , the
second equation becomes (δ)1 = −B10 (t) and by integration with respect to
y, we get δ1 = −B10 (t)y +δ2 (z, t). Therefore D1 = −A01 (t)x−B10 (t)y +δ2 (z, t).
The third equation, because of this expression of D1 becomes (δ2 )z = −C10 (t)
and by integration with respect to z we have δ2 = −C10 (t)z + F10 (t) where
F1 (t) is an arbitrary function, the prime denoting derivative with respect to
time t. We thus have obtained the following expression for the function D1 :
D1 = −A01 (t)x − B10 (t)y − C10 (t)z + F10 (t) (65.146)
Solving the systems (65.111) and (65.115) along the same lines of reasoning
as we did for the system (65.107), we obtain the following expressions for D2
and D3 :
D2 = −A02 (t)x − B20 (t)y − C20 (t)z + F20 (t) (65.147)
D3 = −A03 (t)x − B30 (t)y − C30 (t)z + F30 (t) (65.148)

We now have to write down the new expressions for φ1 , φ2 and φ3 taking into
account the above found expressions for D1 , D2 and D3 .
φ1 = A1 (t)u + B1 (t)v + C1 (t)w

(65.149)
− A01 (t)x − B10 (t)y − C10 (t)z + F10 (t)
792
φ2 = A2 (t)u + B2 (t)v + C2 (t)w
(65.150)
− A02 (t)x − B20 (t)y − C20 (t)z + F20 (t)
and
φ3 = A3 (t)u + B3 (t)v + C3 (t)w
(65.151)
− A03 (t)x − B30 (t)y − C30 (t)z + F30 (t)
Equation (65.57), written as u{(φ1 )v + (ξ 2 )x } + (φ4 )v − 2ν(φ1 )xv = 0 and

using (64.49) (or, that is the same, (65.52)) and the fact that (φ1 )xv = 0,
gives (φ4 )v = 0. Equation (65.60), written as u{(φ1 )w + (ξ 3 )x } + (φ4 )w −
2ν(φ1 )xw = 0 and using (65.50) (or, that is the same, (65.53)) and the fact
that (φ1 )xw = 0, gives (φ4 )w = 0. Equation (65.77), written as v{(φ2 )u +
(ξ 1 )y } + (φ4 )u − 2ν(φ2 )yu = 0 and using (65.66) (or, that is the same, (65.71))
and the fact that (φ2 )yu = 0, gives (φ4 )u = 0. We thus have obtained that
(φ4 )u = (φ4 )v = (φ4 )w = 0 (65.152)
We turn next to equation (65.94). This equation, using equation (65.86)
(or, that is the same, (65.91)), the expressions for ξ 3 and φ3 and relations
(65.152), and after canceling opposite terms, takes on the form
− {A3 (t) + C1 (t)}u − {B3 (t) + C2 (t)}v + {(φ4 )p − 2C3 (t)}w
+ {A03 (t) − C10 (t)}x + {B30 (t) − C20 (t)}y (65.153)
0 0
+ (γ3 )t + C3 (t)z − F3 (t) − ν(γ3 )zz = 0
Equating the coefficients of the above equation to zero, we get the systems
A3 (t) + C1 (t) = 0, A03 (t) − C10 (t) = 0 (65.154)
B3 (t) + C2 (t) = 0, B30 (t) − C20 (t) = 0 (65.155)
(φ4 )p = 2C3 (t) (65.156)
(γ3 )t + C30 (t)z − F30 (t) − ν(γ3 )zz = 0 (65.157)
From the first two systems we obtain
A3 = −C1 = a1 and B3 = −C2 = a2 (65.158)
where a1 and a2 are constants. Integrating (65.156) with respect to p we
obtain
φ4 = 2C3 (t)p + F (x, y, z, t) (65.159)
793
Equation (65.32) using the known expressions of ξ 1 , ξ 3 and φ1 and also
(65.156), takes on the form
A1 (t) − (α1 )x − C3 (t) + (γ3 )z = 0 (65.160)
From the above equation upon differentiation with respect to z we obtain

(γ3 )zz = 0 and thus (65.157) gives
(γ3 )t + C30 (t)z − F30 (t) = 0
which upon integration with respect to t gives
γ3 = −C3 (t)z + F3 (t) (65.161)
(α1 )x = A1 (t) − 2C3 (t) (65.162)
Integrating with respect to x, we obtain from the above equation
α(x, t) = {A1 (t) − 2C3 (t)}x + F5 (t) (65.163)
B2 (t) − (β2 )y − C3 (t) + (γ3 )z = 0
which, because of (65.161) gives
(β2 )y = B2 (t) − 2C3 (t)
and thus by integration with respect to y we get
β2 = {B2 (t) − 2C3 (t)}y + F6 (t) (65.164)
A1 (t) − (ξ 4 )t + (α1 )x − (φ4 )p = 0
(ξ 4 )t = 2{A1 (t) − 2C3 (t)}
794
and by integration with respect to t we have
Z
4
ξ = 2 {A1 (t) − 2C3 (t)}dt (65.165)
Let us now write down the expressions for ξ 1 , ξ 3 and ξ 3 , taking into account
(65.139)-(65.141) and the expressions for α1 (x, t), β2 (y, t) and γ3 (z, t) given
by (65.163), (65.164) and (65.161) respectively:
ξ 1 = {A1 (t) − 2C3 (t)}x − A2 (t)y − A3 (t)z + F5 (t) (65.166)
ξ 2 = −B1 (t)x + {B2 (t) − 2C3 (t)}y − B3 (t)z + F6 (t) (65.167)

and
ξ 3 = −C1 (t)x − C2 (t)y − C3 (t)z + F3 (t) (65.168)
We shall next determine the function F (x, y, z, t) appearing in equation

(65.159) for φ4 . For this purpose we use the system of equations (65.109),
(65.113) and (65.117) taking into account the explicit expressions of D1 , D2
and D3 given by (65.146)-(65.148) respectively.
Fx = A001 (t)x + B100 (t)y + C100 (t)z − F100 (t) (65.169)
and by integration with respect to x, we have

1
F = A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x + η1 (y, z, t) (65.170)
2
We thus have
1
φ4 = 2C3 (t)p+ A001 (t)x2 +B100 (t)xy +C100 (t)xz −F100 (t)x+η1 (y, z, t) (65.171)
2
Equation (65.113), using (65.171), gives
B100 (t)x + (η1 )y = A002 (t)x + B200 (t)y + C200 (t)z − F200 (t) (65.172)
from which we get
B100 (t) = A002 (t) (65.173)
795
and
(η1 )y = B200 (t)y + C200 (t)z − F200 (t) (65.174)
Integrating the above equation with respect to y we get

1
η1 = B200 (t)y 2 + C200 (t)yz − F200 (t)y + η2 (z, t) (65.175)
2
1
2 (65.176)
1 00
+ B2 (t)y 2 + C200 (t)yz − F200 (t)y + η2 (z, t)
2
Equation (65.117), because of (65.176), gives
C100 (t)x + C200 (t)y + (η2 )z = A003 (t)x + B300 (t)y + C300 (t)z − F300 (t) (65.177)
from which we get
C100 (t) = A003 (t), C200 (t) = B300 (t) (65.178)
and
(η2 )z = C300 (t)z − F300 (t) (65.179)
Integrating the above equation with respect to z, we get

1
η2 = C300 (t)z 2 − F300 (t)z + F4 (t) (65.180)
2
We thus obtain the following expression for F (x, y, z, t)
1 1
F = A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x + B200 (t)y 2
2 2 (65.181)
1
+ C200 (t)yz − F200 (t)y + C300 (t)z 2 − F300 (t)z + F4 (t)
2
and thus φ4 is given by
1
φ4 = 2 C3 (t) p + A001 (t)x2 + B100 (t)xy + C100 (t)xz − F100 (t)x
2 (65.182)
1 00 2 00 00 1 00 2 00
+ B2 (t)y + C2 (t)yz − F2 (t)y + C3 (t)z − F3 (t)z + F4 (t)
2 2
796
We continue with some other equations. The next few equations will be
simplified taking into account (65.166)-(65.168) for ξ 1 , ξ 2 , ξ 3 , expressions
(65.149)-(65.152) for φ1 , φ2 , φ3 and also the values (65.158) for the coeffi-
cients A3 , C1 , B3 , C2 .
Equation (65.78) gets simplified into the equation
{B1 (t) + A2 (t)}u + {B10 (t) − A02 (t)}x − 2{B20 (t) + C30 (t)}y
(65.183)
+ 2{B2 (t) − C3 (t)} + F20 (t) − F60 (t) = 0
Equating to zero the coefficients of the above equation we get
B1 (t) + A2 (t) = 0, B10 (t) − A02 (t) = 0 (65.184)
B20 (t) + C30 (t) = 0, B2 (t) − C3 (t) = 0, F20 (t) − F60 (t) = 0 (65.185)
and we thus obtain
B1 = −A2 = a3 , B2 = C3 = a4 , F60 (t) = F20 (t) (65.186)
{C1 (t) + A3 (t)}u + {C2 (t) + B3 (t)}v = 0 (65.187)
which is compatible to (65.158).

− 2{A01 (t) − C30 (t)}x + {A02 (t) − B10 (t)}y + {A03 (t) − C10 (t)}z
(65.188)
+ {A2 (t) + B1 (t)}v + {A3 (t) + C1 (t)}w + F10 (t) − F50 (t) = 0
From the above equation by equating to zero the coefficients, we obtain the
systems
A01 (t) − C30 (t) = 0, F50 (t) = F10 (t) (65.189)
A02 (t) − B10 (t) = 0, A2 (t) + B1 (t) = 0 (65.190)

A03 (t) − C10 (t) = 0, A3 (t) + C1 (t) = 0 (65.191)
The last two systems are compatible to (64.186) and (64.158).
{B10 (t) − A02 (t)}x − 2{B20 (t) − C30 (t)}y + {−C20 (t) + B30 (t)}z
(65.192)
+ {B1 (t) + A2 (t)}u + {C2 (t) + B3 (t)}w + F20 (t) − F60 (t) = 0
797
Equating to zero the coefficients of the above equation we obtain the systems
B10 (t) − A02 (t) = 0, B1 (t) + A2 (t) = 0 (65.193)
−C20 (t) + B30 (t) = 0, C2 (t) + B3 (t) = 0 (65.194)
B20 (t) − C30 (t) = 0, F60 (t) = F20 (t) (65.195)
The above systems are compatible to (65.186) and (65.158).
2{A1 (t) − C3 (t)}u − 2{A01 (t) − C30 (t)}x + {A2 (t) + B1 (t)}v
+ {A02 (t) − B10 (t)}y + {A3 (t) + C1 (t)}w + {A03 − C10 (t)}z (65.196)
+ F10 (t) − F50 (t) = 0
Equating to zero the coefficients of the previous equation, we get the systems
A1 (t) − C3 (t) = 0, A01 (t) − C30 (t) = 0 (65.197)
A2 (t) + B1 (t) = 0, A02 (t) − B10 (t) = 0 (65.198)
A3 (t) + C1 (t) = 0, A03 (t) − C10 (t) =0 (65.199)
The last two systems are compatible to (65.158) and (65.186). From the
system (65.197) we get
A1 = C3 = a4 (65.200)
where we have also taken into account (65.186).
Let us now collect the values of all the coefficients we have already found,
using (65.158), (65.186) and (65.200):
A1 = a4 , A2 = −a3 , A3 = a1
B1 = a3 , B2 = a4 , B3 = a2
(65.201)
C1 = −a1 , C2 = −a2 , C3 = a3
F10 (t) = F50 (t), F20 (t) = F60 (t)
Using the above values of the coefficients, we are able to write down the
infinitesimals φ1 , · · ·, φ4
φ1 = a4 u + a3 v − a1 w + F10 (t)
φ2 = −a3 u + a4 v − a2 w + F20 (t)
(65.202)
φ3 = a1 u + a2 v + a4 w + F30 (t)
φ4 = 2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)
798
and ξ 1 , · · ·, ξ 4
ξ1 = −a4 x + a3 y − a1 z + F1 (t)
ξ2 = −a3 x − a4 y − a2 z + F2 (t)
(65.203)
ξ3 = a1 x + a2 y − a4 z + F3 (t)
ξ4 = −2a4 t + a5
Using the above values of the infinitesimals, we can prove that all the other
equations which have not been used one or the other way, are satisfied iden-
tically.
∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂u ∂v ∂w ∂p
∂ ∂
= {−a4 x + a3 y − a1 z + F1 (t)} + {−a3 x − a4 y − a2 z + F2 (t)}
∂x ∂y
∂ ∂
+ {a1 x + a2 y − a4 z + F3 (t)} + {−2a4 t + a5 }
∂z ∂t
∂ ∂
+ {a4 u + a3 v − a1 w + F10 (t)} + {−a3 u + a4 v − a2 w + F20 (t)}
∂u ∂v
0 ∂
+ {a1 u + a2 v + a4 w + F3 (t)}
∂w
∂
+ {2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)}
∂p
(65.204)
799

∂ ∂ ∂ ∂
X = a1 − z +x −w +u
∂x ∂z ∂u ∂w

∂ ∂ ∂ ∂
+ a2 − z +y −w +v
∂y ∂z ∂v ∂w

∂ ∂ ∂ ∂
+ a3 y −x +v −u
∂x ∂y ∂u ∂v

∂ ∂ ∂ ∂
+ a4 − x −y −z − 2t
∂x ∂y ∂z ∂t
(65.205)
∂ ∂ ∂ ∂
+u +v +w + 2p
∂u ∂v ∂w ∂p

∂ ∂ ∂ 0 ∂ 00 ∂
+ a5 + F4 (t) + F1 (t) + F1 (t) − F1 (t)x
∂t ∂p ∂x ∂u ∂p

∂ ∂ ∂
+ F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p

∂ 0 ∂ 00 ∂
+ F3 (t) + F3 (t) − F3 (t)z
∂z ∂w ∂p
We thus see that the generators of the algebra of Lie symmetries of Navier-
Stokes equations are given by
∂ ∂ ∂ ∂
X 1 = −z +x −w +u (65.206)
∂x ∂z ∂u ∂w
∂ ∂ ∂ ∂
X 2 = −z +y −w +v (65.207)
∂y ∂z ∂v ∂w
∂ ∂ ∂ ∂
X3 =y −x +v −u (65.208)
∂x ∂y ∂u ∂v
∂ ∂ ∂ ∂
X4 = −x −y −z − 2t
∂x ∂y ∂z ∂t
(65.209)
∂ ∂ ∂ ∂
+u +v +w + 2p
∂u ∂v ∂w ∂p
∂
X5 = (65.210)
∂t
∂
X6 = F4 (t) (65.211)
∂p
800
∂ ∂ ∂
X 7 = F1 (t) + F10 (t) − F100 (t)x (65.212)
∂x ∂u ∂p
∂ ∂ ∂
X 8 = F2 (t) + F20 (t) − F200 (t)y (65.213)
∂y ∂v ∂p
∂ ∂ ∂
X 9 = F3 (t) + F30 (t) − F300 (t)z (65.214)
∂z ∂w ∂p
The generators of Lie algebra of symmetries for the Navier-Stokes equations
are identical to those of Euler’s equations. Compare (65.206)-(65.214) to
(64.207)-(64.215).
We shall now try to find some general solutions of Navier-Stokes equations.
We consider the following cases:
CASE I. In this case we consider the linear combination of generators
Γ = X4 + X6 + X7 + X8 + X9 (65.215)
which can be written in full as

∂ ∂ ∂
Γ = (−x + F1 (t)) + (−y + F2 (t)) + (−z + F3 (t)) +
∂x ∂y ∂z
∂ ∂ ∂
+ (u + F10 (t)) + (v + F20 (t)) + (w + F30 (t)) + (65.216)
∂u ∂v ∂w
∂ ∂
+ {2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)} − 2t
∂p ∂t
x̂ = x − F1 (t), ŷ = y − F2 (t), ẑ = z − F3 (t) (65.217)

dx dy dz
= = (65.218)
−x + F1 (t) −y + F2 (t) −z + F3 (t)
i.e. the system
dx̂ dŷ dẑ
− =− =− (65.219)
x̂ ŷ ẑ
we can introduce two invariants ξ and η defined by
ŷ ẑ
ξ= , η= (65.220)
x̂ x̂
801
Similarly, introducing the notation
û = u + F10 (t), v̂ = v + F20 (t), ŵ = w + F30 (t) (65.221)

du dv dw dx
0
= 0
= 0
= (65.222)
u + F1 (t) v + F2 (t) w + F3 (t) −x + F1 (t)
i.e. the system
dû dv̂ dŵ dx̂
= = =− (65.223)
û v̂ ŵ x̂
we can introduce the invariants U (ξ, η), V (ξ, η) and W (ξ, η) defined by
1 1 1
û = U (ξ, η), v̂ = V (ξ, η), ŵ = W (ξ, η) (65.224)
x̂ x̂ x̂
Solving finally the equation
dp dx
= (65.225)
2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t) −x + F1 (t)
i.e. the equation
dp̂ dx̂
= −2 (65.226)
p̂ x̂
we can introduce one more invariant P (ξ, η) defined by
1
p̂ = P (ξ, η) (65.227)
x̂2
where p̂ is given by
p̂ = 2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t) (65.228)
We are now able to write down the Navier-Stokes equations in terms of the
new functions U (ξ, η), V (ξ, η), W (ξ, η) and P (ξ, η). They read
− U 2 − ξU Uξ − ηU Uη + V Uξ + W Uη − 2P − ξPξ − ηPη

− ν 2U + 4ξUξ + 4ηUη + Uξξ + Uηη + ξ 2 Uξξ (65.229)

+ η 2 Uηη + 2ξηUξη = 0
802
− U V − ξU Vξ − ηU Vη + V Vξ + W Vη + Pξ

− ν 2V + 4ξVξ + 4ηVη + Vξξ + Vηη + ξ 2 Vξξ (65.230)

+ η 2 Vηη + 2ξηVξη = 0
− U W − ξU Wξ − ηU Wη + V Wξ + W Wη + Pη

− ν 2W + 4ξWξ + 4ηWη + Wξξ + Wηη + ξ 2 Wξξ (65.231)

2
+ η Wηη + 2ξηWξη = 0
−U − ξUξ − ηUη + Vξ + Wη = 0 (65.232)

The above system admits as solutions homogeneous functions of {ξ, η} of
second order and beyond.
I. First Method of Solution. Introducing the auxiliary variable χ by
(Boisvert, Ames and Srivastava [5])
χ=ξ−η (65.233)
and taking into account that Uξ = Uχ , Uη = −Uχ , etc., the system (65.229)-
(65.232) of PDEs can be reduced into the following system of ordinary dif-
ferential equations
− U 2 − χU Uχ + V Uχ − W Uχ − 2P − χUχ
(65.234)
− ν 2U + 4χUχ + 2Uχχ + χ2 Uχχ = 0
− U V − χU Vχ + V Vχ − W Vχ + Pχ
(65.235)
− ν 2V + 4χVχ + 2Vχχ + χ2 Vχχ = 0
− U W − χU Wχ + V Wχ − W Wχ − Pχ
(65.236)
− ν 2W + 4χWχ + 2Wχχ + χ2 Wχχ = 0
−U − χUχ + Vχ − Wχ = 0 (65.237)
The last equation (65.237) can be integrated with respect to χ, yielding the
equation
V − W = χU − C1 (65.238)
803
Substitution of (65.238) into (65.234) results into the equation

2 2
−U − CUχ − 2P − χPχ − ν 2U + 4χUχ + 2Uχχ + χ Uχχ = 0 (65.239)
Substitution of (65.238) into (65.235) and (65.236) and subtracting the re-
sulting equations, we obtain the equation

−χU 2 + 2Pχ − ν 6χU + 4Uχ + 6χ2 Uχ + 2χUχχ + χ3 Uχχ = 0 (65.240)
Upon eliminating Pχ between (65.239) and (65.240) we get the equation

1 1
P = − U 2 − χ2 U 2 − CUχ − ν 2U + 3χ2 U + 6χUχ
2 2
(65.241)
3 2 1 4
+ 3χ Uχ + 2Uχχ + 2χ Uχχ + χ Uχχ
4
Differentiation of the above equation with respect to χ and substitution the
resulting equation into (65.240) we obtain a third order ordinary differential
equation with a single unknown function:

2
2χU + 2U Uχ + χ U Uχ + CUχχ + ν 12χU + 12Uχ + 18χ2 Uχ
2
1 (65.242)
+ 12χUχχ + 6χ3 Uχχ + 2Uχχχ + 2χ2 Uχχχ + χ4 Uχχχ = 0
4
Once we are able to solve the above equation, we could determine all the
other functions V , W and P . For example, function P can be determined
from (65.241), function W from (65.236) and V from (65.238). In fact, we
obtain from (65.236) using (65.238) that

U W + CWχ + ν 2W + 4χWχ + 2Wχχ + χ2 Wχχ = −Pχ (65.243)
where Pχ can be obtained from (65.240):

1 1
Pχ = χU 2 + ν 6χU + 4Uχ + 6χ2 Uχ + 2χUχχ + χ3 Uχχ (65.244)
2 2
From (65.238) we obtain that
V = W + χU − C (65.245)
804
where W can be determined by (65.243)-(65.244).
Equations (65.234)-(65.245) were derived along the lines of reasoning of the
paper by Boisvert, Ames and Srivastava (Ref. [5]).
One solution of (65.242) is given by
U =−6ν (65.246)
Pχ = 0 (65.247)
and thus equation (65.243) becomes

C
(2 + x2 )Wχχ + 4χ + Wχ − 4W = 0 (65.248)
ν
The above equation for C = 0 admits the general solution
( √ )
4 + 3χ2 3 2χ χ
W = C1 χ + C2 + arctan √ (65.249)
8(2 + χ2 ) 16 2
We finally obtain from (65.245) (for C = 0)

( √ )
4 + 3χ2 3 2χ χ
V = (C1 − 6ν)χ + C2 + arctan √ (65.250)
8(2 + χ2 ) 16 2
Equation (65.241) provides the value of the function P :
P = −12ν 2 (65.251)
Conclusion. The system of equations (65.234)-(65.237) is satisfied by
U = −6ν
( √ )
4 + 3χ2 3 2χ χ
V = (C1 − 6ν)χ + C2 + arctan √
8(2 + χ2 ) 16 2
( √ ) (65.252)
4 + 3χ2 3 2χ χ
W = C1 χ + C2 + arctan √
8(2 + χ2 ) 16 2
P = −12ν 2
805
Let us go back and express everything in terms of the original variables.
Using (65.217) and (65.220), we obtain
ŷ y − F2 (t) ẑ z − F3 (t)
ξ= = , η= =
x̂ x − F1 (t) x̂ x − F1 (t)
(65.253)
(y − z) − [F2 (t) − F3 (t)]
χ=ξ−η =
x − F1 (t)
On the other hand, using (65.221) and (65.224), we have
U (ξ, η) = x̂û = {x − F1 (t)}{u + F10 (t)}
V (ξ, η) = x̂v̂ = {x − F1 (t)}{v + F20 (t)}
W (ξ, η) = x̂ŵ = {x − F1 (t)}{w + F30 (t)}
i.e.
1
u= U (ξ, η) − F10 (t)
x − F1 (t)
1
v= V (ξ, η) − F20 (t) (65.254)
x − F1 (t)
1
w= W (ξ, η) − F30 (t)
x − F1 (t)
We finally get using (65.227) and (65.228) that
P (ξ, η) = x̂2 p̂ = {x − F1 (t)}2 {2p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)}
i.e.
1
p= P (ξ, η)
2{x − F1 (t)}2
(65.255)
1
+ {F100 (t)x + F200 (t)y + F300 (t)z − F4 (t)}
2
Conclusion. The solution of the system of equations (65.229)-(65.232) is
given by (65.254) and (64.255) where U, V, W are given by (65.252) and χ is
given in (65.253).
Based on this Conclusion, we arrive at the following
Theorem 1. The solution of the system of equations (65.229)-(65.232) is
given by
6ν
u=− − F10 (t) (65.256)
x − F1 (t)
806
( √ )
4 + 3χ2
χ
1 3 2χ
v= (C1 − 6ν)χ + C2 + arctan √
x − F1 (t) 8(2 + χ2 ) 16 2
− F20 (t)
(65.257)
( √ )
4 + 3χ2

1 3 2χ χ
w= C1 χ + C2 + arctan √
x − F1 (t) 8(2 + χ2 ) 16 2 (65.258)
− F30 (t)
where
(y − z) − [F2 (t) − F3 (t)]
χ= (65.259)
x − F1 (t)
and
6ν 2
p=−
{x − F1 (t)}2 (65.260)
1
+ {F100 (t)x + F200 (t)y + F300 (t)z − F4 (t)}
2
Note. No other solution of equation (65.242) is known so far. The claim
by Boisvert, Ames and Srivastava (Ref. [5]) that (65.242) is satisfied by
U = −12νχ−2 does not seem to be true.
CASE II. We shall now try to find a general solution of the Navier-Stokes
equations using the generator X 4 . We thus consider Lagrange’s auxiliary
system
dx dy dz dt du dv dw dp
= = = = = = = (65.261)
−x −y −z −2t u v w 2p
dx dt √
Integrating the equation = we find x = C1 t and thus we introduce
−x −2t
the invariant ξ defined by
x
ξ=√ (65.262)
t
dy dt dz dt
Similarly integrating the equations = and = we find the
√ −y √ −2t −z −2t
general solutions y = C2 t and z = C3 t and therefore we introduce the
807
invariants η and θ defined by
y z
η=√ , θ=√ (65.263)
t t
respectively.
du dt dv dt dw dt
Integrating further the equations = , = , = , and
u −2t v −2t w −2t
dp dt √ √ √
= , we find u t = C4 , v t = C5 , w t = C6 and p t = C7 re-
2p −2t
spectively. We thus introduce the invariants U = U (ξ, η, θ), V = V (ξ, η, θ),
W = W (ξ, η, θ) and P = P (ξ, η, θ) defined by
U V W P
u= √ , v=√ , w=√ , p= (65.264)
t t t t
Using the above substitutions, the Navier-Stokes equations take on the form
1 1 ∂U 1 ∂U 1 ∂U ∂U ∂U ∂U
− U− ξ − η − θ +U +V +W
2 2 ∂ξ 2 ∂η 2 ∂θ ∂ξ ∂η ∂θ
2 2 2 (65.265)
∂P ∂ U ∂ U ∂ U
+ −ν 2
+ 2
+ =0
∂ξ ∂ξ ∂η ∂θ2
1 1 ∂V 1 ∂V 1 ∂V ∂V ∂V ∂V
− V − ξ − η − θ +U +V +W
2 2 ∂ξ 2 ∂η 2 ∂θ ∂ξ ∂η ∂θ
2 2 2 (65.266)
∂P ∂ V ∂ V ∂ V
+ −ν 2
+ 2
+ =0
∂η ∂ξ ∂η ∂θ2
1 1 ∂W 1 ∂W 1 ∂W ∂W ∂W ∂W
− W− ξ − η − θ +U +V +W
2 2 ∂ξ 2 ∂η 2 ∂θ ∂ξ ∂η ∂θ
2 2 2 (65.267)
∂P ∂ W ∂ W ∂ W
+ −ν + + =0
∂θ ∂ξ 2 ∂η 2 ∂θ2
∂U ∂V ∂W
+ + =0 (65.268)
∂ξ ∂η ∂θ
For ν = 0 the above system reduces to the system (64.266)-(64.269) of the

Euler equations.
Before we proceed to finding solutions of the system (65.265)-(65.268), we
have to cite a solution found by Boisvert, Ames and Srivastava (Ref. [6])
808
which is expressed in terms of the error function and is given by

2 2 −1/2
U = a3 erf [ν(4a1 + 8a2 )] (2a2 ξ − a1 η − a1 θ)
(65.269)
1 2a2 a1 a2
+ − 2 2 2 ξ+ 2 (η + θ) + a4
2 a1 + 2a2 a1 + 2a22

a2 a3 2 2 −1/2
V = erf [ν(4a1 + 8a2 )] (2a2 ξ − a1 η − a1 θ)
a1
(65.270)
a1 a2 1 a22 a2 a4
+ 2 2
ξ+ − + 2 2
(η + θ) +
a1 + 2a2 4 a1 + 2a2 a1
W =V (65.271)
1 a2 1 a2
P = − 2 2 2 ξ2 + − 2 1 2 (η 2 + θ2 )
8 a1 + 2a2 16 4a1 + 8a2
(65.272)
a1 a2 1 a1
+ 2 (ξη + ξθ) + − ηθ + a5
a1 + 2a22 8 2a21 + 4a22
Using (65.262)-(65.264) we can recover the original variables and functions
from (65.269)-(65.272).
However our solution is quite different to that of Ref. [5], expressed in terms
of hypergeometric functions (confluent hypergeometric and generalized hy-
pergeometric functions). We start with a symmetry analysis of the system.
Lie Symmetry Analysis of the system (65.265)-(65.268).
Let Γ be the generator of Lie point symmetries for the system (65.265)-
(65.268), where Γ is the vector
∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ3
∂ξ ∂η ∂θ
(65.273)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂U ∂V ∂W ∂P
The functions {ξ 1 , ξ 2 , ξ 3 } and {φ1 , φ2 , φ3 , φ4 } depend on both the variables
{ξ, η, θ} and the functions {U, V, W, P }.
809
The first prolongation of the vector Γ is denoted by P r(1) Γ and is defined by
∂ ∂ ∂
P r(1) Γ = Γ + φξ + φη + φθ
∂Uξ ∂Uη ∂Uθ
∂ ∂ ∂
+ ψξ + ψη + ψθ
∂Vξ ∂Vη ∂Vθ
(65.274)
∂ ∂ ∂
+ χξ + χη + χθ
∂Wξ ∂Wη ∂Wθ
∂ ∂ ∂
+ πξ + πη + πθ
∂Pξ ∂Pη ∂Pθ
The second prolongation of the vector Γ is denoted by P r(2) Γ and is defined
by
P r(2) Γ = P r(1) Γ
∂ ∂ ∂
+ φξξ + φηη + φθθ
∂Uξξ ∂Uηη ∂Uθθ
∂ ∂ ∂ (65.275)
+ ψ ξξ + ψ ηη + ψ θθ
∂Vξξ ∂Vηη ∂Vθθ
∂ ∂ ∂
+ χξξ + χηη + χθθ
∂Wξξ ∂Wηη ∂Wθθ
∂
We have not included cross-terms like φξη etc., since mixed derivatives
∂Uξη
like Uξη do not appear in the original system. We have not also included terms
line π ξξ , π ηη and π θθ and similar cross-terms since second order derivatives
like Pξξ do not appear in the original system.
We furthe denote by ∆(1) , ∆(2) , ∆(3) and ∆(4) the quantities
1 1 1 1
∆(1) = − U − ξUξ − ηUη − θUθ
2 2 2 2 (65.276)
+ U Uξ + V Uη + W Uθ + Pξ − ν(Uξξ + Uηη + Uθθ )
1 1 1 1
∆(2) = − V − ξVξ − ηVη − θVθ
2 2 2 2 (65.277)
+ U Vξ + V Vη + W Vθ + Pη − ν(Vξξ + Vηη + Vθθ )
1 1 1 1
∆(3) = − W − ξWξ − ηWη − θWθ
2 2 2 2 (65.278)
+ U Wξ + V Wη + W Wθ + Pθ − ν(Wξξ + Wηη + Wθθ )
810
and
∆(4) = Uξ + Vη + Wθ (65.279)
respectively.
The Lie symmetries of the system (65.265)-(65.268) can be determined once
we know the functions {ξ 1 , ξ 2 , ξ 3 } and {φ1 , φ2 , φ3 , φ4 }.
P r(2) Γ[∆(k) ] = 0, k = 1, 2, 3, 4
as long as (65.280)
(1) (2) (3) (4)
∆ = 0, ∆ = 0, ∆ = 0, ∆ =0
Using the definition of P r(2) Γ and the expressions ∆(1) , ∆(2) , ∆(3) , ∆(4) ,
we obtain
1 1 1 1
2 2 2 2
1
+ Uη φ2 + Uθ φ3 + − ξ + U φξ
2 (65.281)
1 1
η θ
+ − η+V φ + − θ+W φ
2 2
+ π − νφ − νφ − νφθθ
ξ ξξ ηη
1 1 1
2 2 2
1 1
+ − + Vη φ2 + Vθ φ3 + − ξ + U ψ ξ
2 2 (65.282)
1 1
+ − η + V ψη + − θ + W ψθ
2 2
+ π − νψ − νψ − νψ θθ
η ξξ ηη
1 1 1
2 2 2
1 1
+ Wη φ2 + − + Wθ φ3 + − ξ + U χξ
2 2 (65.283)
1 1
+ − η + V χη + − θ + W χθ
2 2
+ π − νχ − νχ − νχθθ
θ ξξ ηη
811
and
P r(2) Γ[∆(4) ] = φξ + ψ η + χθ (65.284)
Let us now list explicitly the various coefficients φξ , · · · , χθθ which appear in
(65.281)-(65.284). We have
φξ = (φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U

− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ] (65.285)
φη = (φ1 )η + (φ1 )U Uη + (φ1 )V Vη + (φ1 )W Wη − (Uη )2 (ξ 2 )U

− Pη [Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ] (65.286)
− Uη [Vη (ξ 2 )V + Wη (ξ 2 )W + (ξ 2 )η ]
− Uθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
φθ = (φ1 )θ + (φ1 )U Uθ + (φ1 )V Vθ + (φ1 )W Wθ − (Uθ )2 (ξ 3 )U
− Pθ [Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ] (65.287)
− Uη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ]
− Uθ [Vθ (ξ 3 )V + Wθ (ξ 3 )W + (ξ 3 )θ ]
ψ ξ = (φ2 )ξ + (φ2 )U Uξ + (φ2 )V Vξ + (φ2 )W Wξ − (Vξ )2 (ξ 1 )V
− Pξ [Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uξ (ξ 1 )U + Wξ (ξ 1 )W + (ξ 1 )ξ ] (65.288)
− Vη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ]
− Vθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
ψ η = (φ2 )η + (φ2 )U Uη + (φ2 )V Vη + (φ2 )W Wη − (Vη )2 (ξ 2 )V
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ] (65.289)
812
ψ θ = (φ2 )θ + (φ2 )U Uθ + (φ2 )V Vθ + (φ2 )W Wθ − (Vθ )2 (ξ 3 )V
− Pθ [Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ] (65.290)
− Vη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ]
− Vθ [Uθ (ξ 3 )U + Wθ (ξ 3 )W + (ξ 3 )θ ]
χξ = (φ3 )ξ + (φ3 )U Uξ + (φ3 )V Vξ + (φ3 )W Wξ − (Wξ )2 (ξ 1 )W
− Pξ [Wξ (ξ 1 )P + Wη (ξ 2 )P + Wθ (ξ 3 )P − (φ3 )P ]
− Wξ [Uξ (ξ 1 )U + Vξ (ξ 1 )V + (ξ 1 )ξ ] (65.291)
− Wη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ]
− Wθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
χη = (φ3 )η + (φ3 )U Uη + (φ3 )V Vη + (φ3 )W Wη − (Wη )2 (ξ 2 )W
− Pη [Wξ (ξ 1 )P + Wη (ξ 2 )P + Wθ (ξ 3 )P − (φ3 )P ]
− Wξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ] (65.292)
− Wη [Uξ (ξ 2 )U + Vη (ξ 2 )V + (ξ 2 )η ]
− Wθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
χθ = (φ3 )θ + (φ3 )U Uθ + (φ3 )V Vθ + (φ3 )W Wθ − (Wθ )2 (ξ 3 )W
− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ] (65.293)
π ξ = (φ4 )ξ + (φ4 )U Uξ + (φ4 )V Vξ + (φ4 )W Wξ
− Pξ [Uξ (ξ 1 )U + Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ − (φ4 )P ]
− Pη [Uξ (ξ 2 )U + Vξ (ξ 2 )V + Wξ (ξ 2 )W + (ξ 2 )ξ ] (65.294)
− Pθ [Uξ (ξ 3 )U + Vξ (ξ 3 )V + Wξ (ξ 3 )W + (ξ 3 )ξ ]
− Pξ [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
π η = (φ4 )η + (φ4 )U Uη + (φ4 )V Vη + (φ4 )W Wη
− Pξ [Uη (ξ 1 )U + Vη (ξ 1 )V + Wη (ξ 1 )W + (ξ 1 )η ]
− Pη [Uη (ξ 2 )U + Vη (ξ 2 )V + Wη (ξ 2 )W + (ξ 2 )η − (φ4 )P ] (65.295)
− Pθ [Uη (ξ 3 )U + Vη (ξ 3 )V + Wη (ξ 3 )W + (ξ 3 )η ]
− Pη [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
813
π θ = (φ4 )θ + (φ4 )U Uθ + (φ4 )V Vθ + (φ4 )W Wθ
− Pξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + Wθ (ξ 1 )W + (ξ 1 )θ ]
− Pη [Uθ (ξ 2 )U + Vθ (ξ 2 )V + Wθ (ξ 2 )W + (ξ 2 )θ ] (65.296)
− Pθ [Uθ (ξ 3 )U + Vθ (ξ 3 )V + Wθ (ξ 3 )W + (ξ 3 )θ − (φ4 )P ]
− Pθ [Pξ (ξ 1 )P + Pη (ξ 2 )P + Pθ (ξ 3 )P ]
φξξ = −Pξξ [(ξ 1 )P Uξ + (ξ 2 )P Uη + (ξ 3 )P Uθ − (φ1 )P ]
− 2Uξη [(ξ 2 )P Pξ + (ξ 2 )U Uξ + (ξ 2 )V Vξ + (ξ 2 )W Wξ + (ξ 2 )ξ ]
− 2Uξθ [(ξ 3 )P Pξ + (ξ 3 )U Uξ + (ξ 3 )V Vξ + (ξ 3 )W Wξ + (ξ 3 )ξ ]
− Uξξ [2(ξ 1 )P Pξ + 3(ξ 1 )U Uξ + 2(ξ 1 )V Vξ + 2(ξ 1 )W Wξ
+ 2(ξ 1 )ξ + (ξ 2 )U Uη + (ξ 3 )U Uθ − (φ1 )U ]
− Vξξ [(ξ 1 )V Uξ + (ξ 2 )V Uη + (ξ 3 )V Uθ − (φ1 )V ]
− Wξξ [(ξ 1 )W Uξ + (ξ 2 )W Uη + (ξ 3 )W Uθ − (φ1 )W ]
− (Pξ )2 [(ξ 1 )P P Uξ + (ξ 2 )P P Uη + (ξ 3 )P P Uθ − (φ1 )P P ]
− (Uξ )2 [2(ξ 1 )U P Pξ + (ξ 1 )U U Uξ + 2(ξ 1 )U V Vξ + 2(ξ 1 )U W Wξ
+ 2(ξ 1 )ξU + 2(ξ 2 )U U Uη + (ξ 3 )U U Uθ − (φ1 )U U ]
− 2Pξ [(ξ 1 )V P Uξ Vξ + (ξ 1 )W P Uξ Wξ + (ξ 2 )U P Uξ Uη + (ξ 1 )ξP Uξ
+ (ξ 2 )V P Uη Vξ + (ξ 2 )W P Uη Wξ + (ξ 2 )ξP Uη + (ξ 3 )U P Uθ Uξ
(65.297)
+ (ξ 3 )V P Uθ Vξ + (ξ 3 )W P Uθ Wξ + (ξ 3 )ξP Uθ − (φ1 )U P Uξ
− (φ1 )V P Vξ − (φ1 )W P Wξ − (φ1 )ξP ]
− (Vξ )2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη + (ξ 3 )V V Uθ − (φ1 )V V ]
− (Wξ )2 [(ξ 1 )W W Uξ + (ξ 2 )W W Uη + (ξ 3 )W W Uθ − (φ1 )W W ]
− Uξ [2(ξ 1 )V W Vξ Wξ + 2(ξ 1 )ξV Vξ + 2(ξ 1 )ξW Wξ + (ξ 1 )ξξ
+ 2(ξ 2 )U V Uη Vξ + 2(ξ 2 )U W Uη Wξ + 2(ξ 2 )ξU Uη
+ 2(ξ 3 )U V Uθ Vξ + 2(ξ 3 )U W Uθ Wξ + 2(ξ 3 )ξU Uθ
− 2(φ1 )U V Vξ − 2(φ1 )U V Vξ − 2(φ1 )U W Wξ − 2(φ1 )ξU ]
− 2Vξ [(ξ 2 )V W Uη Wξ + (ξ 2 )ξV Uη + (ξ 3 )V W Uθ Wξ + (ξ 3 )ξV Uθ
− (φ1 )V W Wξ − (φ1 )ξV ]
− 2Wξ [(ξ 2 )ξW Uη + (ξ 3 )ξW Uθ − (φ1 )ξW ]
− (ξ 2 )ξξ Uη − (ξ 3 )ξξ Uθ + (φ1 )ξξ
814
φηη = −Pηη [(ξ 1 )P Uξ + (ξ 2 )P Uη + (ξ 3 )P Uθ − (φ1 )P ]
− Uηη [(ξ 1 )U Uξ + 2(ξ 2 )P Pη + 3(ξ 2 )U Uη + 2(ξ 2 )V Vη
+ 2(ξ 2 )W Wη + 2(ξ 2 )η − (ξ 3 )U Uθ − (φ1 )U ]
− 2Uηθ [(ξ 3 )P Pη + (ξ 3 )U Uη + (ξ 3 )V Vη + (ξ 3 )W Wη + (ξ 3 )η ]
− 2Uξη [(ξ 1 )P Pη + (ξ 1 )U Uη + (ξ 1 )V Vη + (ξ 1 )W Wη + (ξ 1 )η ]
− Vηη [(ξ 1 )V Uξ + (ξ 2 )V Uη + (ξ 3 )V Uθ − (φ1 )V ]
− Wηη [(ξ 1 )W Uξ + (ξ 2 )W Uη + (ξ 3 )W Uθ − (φ1 )W ]
− (Pη )2 [(ξ 1 )P P Uξ + (ξ 2 )P P Uη + (ξ 3 )P P Uθ − (φ1 )P P ]
− (Uη )2 [(ξ 1 )U U Uξ + 2(ξ 2 )U P Pη + 2(ξ 2 )U V Vη + 2(ξ 2 )U W Wη
+ 2(ξ 2 )ηU + (ξ 3 )U U Uθ + (ξ 2 )U U Uη − (φ1 )U U ]
− (Vη )2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη + (ξ 2 )V V Uθ − (φ1 )V V ]
− (Wη )2 [(ξ 1 )W W Uξ + (ξ 2 )W W Uη + (ξ 3 )W W Uθ − (φ1 )W W ] (65.298)
− 2Pη [(ξ 1 )U P Uξ Uη + (ξ 1 )V P Uξ Vη + (ξ 1 )W P Uξ Wη + (ξ 1 )ηP Uξ
+ (ξ 2 )V P Uη Vη + (ξ 2 )W P Uη Wη + (ξ 2 )ηP Uη + (ξ 3 )U P Uη Uθ
+ (ξ 3 )V P Uθ Vη + (ξ 3 )W P Uθ Wη + (ξ 3 )ηP Uθ − (φ1 )U P Uη
− (φ1 )V P Vη − (φ1 )W P Wη − (φ1 )ηP ]
− 2Uη [(ξ 1 )U V Uξ Vη + (ξ 1 )U W Uξ Wη + (ξ 1 )ηU Uξ + (ξ 2 )V W Wη Vη
+ (ξ 2 )ηV Vη + (ξ 2 )ηW Wη + (ξ 3 )U V Uθ Vη + (ξ 3 )U W Uθ Wη
+ (ξ 3 )ηU Uθ − (φ1 )U V Vη − (φ1 )U W Wη − (φ1 )ηU ]
− 2Uξ [(ξ 1 )V W Vη Wη + (ξ 1 )ηV Vη + (ξ 1 )ηW Wη ]
− 2Uθ [(ξ 3 )V W Vη Wη + (ξ 3 )ηV Vη + (ξ 3 )ηW Wη ]
− (ξ 1 )ηη Uξ − (ξ 2 )ηη Uη − (ξ 3 )ηη Uθ + (φ1 )ηη
+ 2 [(φ1 )V W Vη Wη + (φ1 )ηV Vη + (φ1 )ηW Wη ]
815
φθθ = −Pθθ [(ξ 1 )P Uξ + (ξ 2 )P Uη + (ξ 3 )P Uθ − (φ1 )P ]
− 2Uηθ [(ξ 2 )P Pθ + (ξ 2 )U Uθ + (ξ 2 )V Vθ + (ξ 2 )W Wθ + (ξ 2 )θ ]
− Uθθ [(ξ 1 )U Uξ + (ξ 2 )U Uη + 2(ξ 3 )P Pθ + 3(ξ 3 )U Uθ + 2(ξ 3 )V Vθ
+ 2(ξ 3 )W Wθ + 2(ξ 3 )θ − (φ1 )U ]
− 2Uξθ [(ξ 1 )P Pθ + (ξ 1 )U Uθ + (ξ 1 )V Vθ + (ξ 1 )W Wθ + (ξ 1 )θ ]
− Vθθ [(ξ 1 )V Uξ + (ξ 2 )V Uη + (ξ 3 )V Uθ − (φ1 )V ]
− Wθθ [(ξ 1 )W Uξ + (ξ 2 )W Uη + (ξ 3 )W Uθ − (φ1 )W ]
− (Pθ )2 [(ξ 1 )P P Uξ + (ξ 2 )P P Uη + (ξ 3 )P P Uθ − (φ1 )P P ]
− 2Pθ [(ξ 1 )U P Uξ Uθ + (ξ 1 )V P Uξ Vθ + (ξ 1 )W P Uξ Wθ + (ξ 1 )θP Uξ
+ (ξ 2 )U P Uη Uθ + (ξ 2 )V P Uη Vθ + (ξ 2 )W P Uη Wθ + (ξ 2 )θP Uη
+ (ξ 3 )V P Uθ Vθ + (ξ 3 )W P Uθ Wθ + (ξ 3 )θP Uθ − (φ1 )U P Uθ
− (φ1 )V P Vθ − (φ1 )W P Wθ − (φ1 )θP + (ξ 3 )U P (Uθ )2 ] (65.299)
− (Uθ )2 [(ξ 1 )U U Uξ + (ξ 2 )U U Uη + (ξ 3 )U U Uθ + 2(ξ 3 )U V Vθ
+ 2(ξ 3 )U W Wθ + 2(ξ 3 )θU − (φ1 )U U ]
− 2Uθ [(ξ 1 )U V Uξ Vθ + (ξ 1 )U W Uξ Wθ + (ξ 1 )θU Uξ + (ξ 2 )U V Uη Vθ
+ (ξ 2 )U W Uη Wθ + (ξ 2 )θU Uη + (ξ 3 )V W Vθ Wθ + (ξ 3 )θV Vθ
+ (ξ 3 )θW Wθ − (φ1 )U V Vθ − (φ1 )U W Wθ − (φ1 )θU ]
− (Vθ )2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη + (ξ 3 )V V Uθ − (φ1 )V V ]
− (Wθ )2 [(ξ 1 )W W Uξ + (ξ 2 )W W Uη + (ξ 3 )W W Uθ − (φ1 )W W ]
− 2Uξ [(ξ 1 )V W Vθ Wθ + (ξ 1 )θV Vθ + (ξ 1 )θW Wθ ]
− 2Uη [(ξ 2 )V W Vθ Wθ + (ξ 2 )θV Vθ + (ξ 2 )θW Wθ ]
− (ξ 1 )θθ Uξ − (ξ 2 )θθ Uη − (ξ 3 )θθ Uθ + (φ1 )θθ
+ 2 [(φ1 )V W Vθ Wθ + (φ1 )θV Vθ + (φ1 )θW Wθ ]
816
ψ ξξ = −Pξξ [(ξ 1 )P Vξ + (ξ 2 )P Vη + (ξ 3 )P Vθ − (φ2 )P ]
− Uξξ [(ξ 1 )U Vξ + (ξ 2 )U Vη + (ξ 3 )U Vθ − (φ2 )U ]
− 2Pξ [(ξ 1 )P Vξξ + (ξ 2 )P Vξη + (ξ 3 )P Vξθ + (ξ 1 )U P Uξ Vξ
+ (ξ 1 )W P Vξ Wξ + (ξ 1 )ξP Vξ + (ξ 2 )U P Uξ Vη + (ξ 2 )V P Vξ Vη
+ (ξ 2 )W P Vη Wξ + (ξ 2 )ξP Vη + (ξ 3 )U P Uξ Vθ + (ξ 3 )V P Vξ Vθ
+ (ξ 3 )W P Vθ Wξ + (ξ 3 )ξP Vθ − (φ2 )U P Uξ − (φ2 )V P Vξ
− (φ2 )W P Wξ − (φ2 )ξP ]
− 2Vξη [(ξ 2 )U Uξ + (ξ 2 )V Vξ + (ξ 2 )W Wξ + (ξ 2 )ξ ]
− 2Vξθ [(ξ 3 )U Uξ + (ξ 3 )V Vξ + (ξ 3 )W Wξ + (ξ 3 )ξ ]
− Vξξ [2(ξ 1 )U Uξ + 3(ξ 1 )V Vξ + 2(ξ 1 )W Wξ + 2(ξ 1 )ξ + (ξ 2 )V Vη
+ (ξ 3 )V Vθ − (φ2 )V ]
− Wξξ [(ξ 1 )W Vξ + (ξ 2 )W Vη + (ξ 3 )W Vθ − (φ2 )W ]
(65.300)
− (Pξ )2 [(ξ 1 )P P Vξ + (ξ 2 )P P Vη + (ξ 3 )P P Vθ − (φ2 )P P ]
− (Uξ )2 [(ξ 1 )U U Vξ + (ξ 2 )U U Vη + (ξ 3 )U U Vθ − (φ2 )U U ]
− (Vξ )2 [2(ξ 1 )U V Uξ + 2(ξ 1 )V P Pξ + (ξ 1 )V V Vξ + 2(ξ 1 )V W Wξ
+ 2(ξ 1 )ξV + (ξ 2 )V V Vη + (ξ 3 )V V Vθ − (φ2 )V V ]
− (Wξ )2 [(ξ 1 )W W Vξ + (ξ 2 )W W Vη + (ξ 3 )W W Vθ − (φ2 )W W ]
− 2Uξ [(ξ 1 )U W Vξ Wξ + (ξ 1 )ξU Vξ + (ξ 2 )U V Vξ Vη + (ξ 2 )U W Vη Wξ
+ (ξ 2 )ξU Vη + (ξ 3 )U V Vξ Vθ + (ξ 3 )U W Vθ Wξ + (ξ 3 )ξU Vθ
− (φ2 )U V Vξ − (φ2 )U W Wξ − (φ2 )ξU ]
− Vξ [2(ξ 1 )ξW + (ξ 1 )ξξ − 2(φ2 )V W Wξ − 2(φ2 )ξV ]
− Vη [2(ξ 2 )V W Vξ Wξ + 2(ξ 2 )ξV Vξ + (ξ 2 )ξW Wξ + (ξ 2 )ξξ ]
− Vθ [2(ξ 3 )V W Vξ Wξ + 2(ξ 3 )ξV Vξ + 2(ξ 3 )ξW Wξ + (ξ 2 )ξW Wξ
+ (ξ 3 )ξξ ] + 2(φ2 )ξW Wξ + (φ2 )ξξ
817
ψ ηη = −Pηη [(ξ 1 )P Vξ + (ξ 2 )P Vη + (ξ 3 )P Vθ − (φ2 )P ]
− Uηη [(ξ 1 )U Vξ + (ξ 2 )U Vη + (ξ 3 )U Vθ − (φ2 )U ]
− 2Pη [(ξ 1 )P Vξη + (ξ 2 )P Vηη + (ξ 3 )P Vηθ + (ξ 1 )U P Uη Vξ
+ (ξ 1 )V P Vη Vξ + (ξ 1 )W P Vξ Wη + (ξ 1 )ηP Vξ + (ξ 2 )U P Uη Vη
+ (ξ 2 )V P (Vη )2 + (ξ 2 )W P Vη Wη + (ξ 2 )ηP Vη + (ξ 3 )U P Uη Vθ
+ (ξ 3 )V P Vη Vθ + (ξ 3 )W P Vθ Wη + (ξ 3 )ηP Vθ − (φ2 )U P Uη
− (φ2 )V P Vη − (φ2 )W P Wη − (φ2 )ηP ]
− (Pη )2 [(ξ 1 )P P Vξ + (ξ 2 )P P Vη + (ξ 3 )P P Vθ − (φ2 )P P ]
− Vηη [(ξ 1 )V Vξ + 2(ξ 2 )U Uη + 3(ξ 2 )V Vη + 2(ξ 2 )W Wη + 2(ξ 2 )η
+ (ξ 3 )V Vθ − (φ2 )V ]
− 2Vηθ [(ξ 3 )U Uη + (ξ 3 )V Vη + (ξ 3 )W Wη + (ξ 3 )η ]
− 2Vξη [(ξ 1 )U Uη + (ξ 1 )V Vη + (ξ 1 )W Wη + (ξ 1 )η ]
(65.301)
− Wηη [(ξ 1 )W Vξ + (ξ 2 )W Vη + (ξ 3 )W Vθ − (φ2 )W ]
− (Uη )2 [(ξ 1 )U U Vξ + (ξ 2 )U U Vη + (ξ 3 )U U Vθ − (φ2 )U U ]
− (Vη )2 [(ξ 1 )V V Vξ + 2(ξ 2 )U V Uη + (ξ 2 )V V Vη + 2(ξ 2 )V W Wη
+ 2(ξ 2 )ηV + (ξ 3 )V V Vθ − (φ2 )V V ]
− (Wη )2 [(ξ 1 )W W Vξ + (ξ 2 )W W Vη + (ξ 3 )W W Vθ − (φ2 )W W ]
− 2Uη [(ξ 1 )U V Vξ Vη + (ξ 1 )U W Vξ Wη + (ξ 1 )ηU Vξ + (ξ 2 )U W Vη Wη
+ (ξ 2 )ηU Vη + (ξ 3 )U V Vη Vθ + (ξ 3 )U W Vθ Wη + (ξ 3 )ηU Vθ
− (φ2 )U V Vη − (φ2 )U W Wη − (φ2 )ηU ]
− 2Vη [(ξ 1 )V W Vξ Wη + (ξ 1 )ηV Vξ + (ξ 2 )ηW Wη − (φ2 )ηV
+ (ξ 3 )V W Vθ Wη + (ξ 3 )ηV Vθ − (φ2 )V W Wη ]
− 2Wη [(ξ 1 )ηW Vξ + (ξ 3 )ηW Vθ − (φ2 )ηW ]
− (ξ 1 )ηη Vξ − (ξ 2 )ηη Vη − (ξ 3 )ηη Vθ + (φ2 )ηη
818
ψ θθ = −Pθθ [(ξ 1 )P Vξ + (ξ 2 )P Vη + (ξ 3 )P Vθ − (φ2 )P ]
− (Pθ )2 [(ξ 1 )P P Vξ + (ξ 2 )P P Vη + (ξ 3 )P P Vθ − (φ2 )P P ]
− (Uθ )2 [(ξ 1 )U U Vξ + (ξ 2 )U U Vη + (ξ 3 )U U Vθ − (φ2 )U U ]
− (Vθ )2 [(ξ 1 )V V Vξ + (ξ 2 )V V Vη + 2(ξ 3 )U V Uθ + (ξ 3 )V V Vθ
+ 2(ξ 3 )V W Wθ + 2(ξ 3 )θV − (φ2 )V V ]
− (Wθ )2 [(ξ 1 )W W Vξ + (ξ 2 )W W Vη + (ξ 3 )W W Vθ − (φ2 )W W ]
− Uθθ [(ξ 1 )U Vξ + (ξ 2 )U Vη + (ξ 3 )U Vθ − (φ2 )U ]
− 2Pθ [(ξ 1 )P Vξθ + (ξ 2 )P Vηθ + (ξ 3 )P Vθθ + (ξ 1 )U P Vξ Uθ
+ (ξ 1 )V P Vξ Vθ + (ξ 1 )W P Vξ Wθ + (ξ 2 )U P Uθ Vη
+ (ξ 2 )V P Vη Vθ + (ξ 2 )W P Vη Wθ + (ξ 3 )U P Uθ Vθ
+ (ξ 3 )W P Vθ Wθ + (ξ 3 )V P (Vθ )2 + (ξ 1 )θP Vξ + (ξ 2 )θP Vη
+ (ξ 3 )θP Vθ − (φ2 )U P Uθ − (φ2 )V P Vθ − (φ2 )W P Wθ
− (φ2 )θP ]
(65.302)
− 2Vηθ [(ξ 2 )U Uθ + (ξ 2 )V Vθ + (ξ 2 )W Wθ + (ξ 2 )θ ]
− Vθθ [(ξ 1 )V Vξ + (ξ 2 )V Vη + 2(ξ 3 )U Uθ + 3(ξ 3 )V Vθ
+ 2(ξ 3 )W Wθ + 2(ξ 3 )θ − (φ2 )V ]
− 2Vξθ [(ξ 1 )U Uθ + (ξ 1 )V Vθ + (ξ 1 )W Wθ + (ξ 1 )θ ]
− Wθθ [(ξ 1 )W Vξ + (ξ 2 )W Vη + (ξ 3 )W Vθ − (φ2 )W ]
− 2Uθ [(ξ 1 )U V Vξ Vθ + (ξ 1 )U W Vξ Wθ + (ξ 1 )θU Vξ
+ (ξ 2 )U V Vη Vθ + (ξ 2 )U W Vη Wθ + (ξ 2 )θU Vη
+ (ξ 3 )U W Vθ Wθ − (φ2 )U V Vθ + (ξ 3 )θU Vθ
− (φ2 )U W Wθ − (φ2 )θU ]
− 2Vθ [(ξ 1 )V W Vξ Wθ + (ξ 1 )θV Vξ + (ξ 3 )θW Wθ − (φ2 )V W Wθ
+ (ξ 2 )V W Vη Wθ + (ξ 2 )θV Vη − (φ2 )θV ]
− Vξ [2(ξ 1 )θW Vξ + (ξ 1 )θθ ] − Vη [2(ξ 2 )θW Wθ + (ξ 2 )θθ ]
− (ξ 3 )θθ Vθ + 2(φ2 )θW Wθ + (φ2 )θθ
819
χξξ = −Pξξ [(ξ 1 )P Wξ + (ξ 2 )P Wη + (ξ 3 )P Wθ − (φ3 )P ]
− (Pξ )2 [(ξ 1 )P P Wξ + (ξ 2 )P P Wη + (ξ 3 )P P Wθ − (φ3 )P P ]
− (Uξ )2 [(ξ 1 )U U Wξ + (ξ 2 )U U Wη + (ξ 3 )U U Wθ − (φ3 )U U ]
− (Wξ )2 [2(ξ 1 )U W Uξ + 2(ξ 1 )V W Vξ + (ξ 1 )W W Wξ + 2(ξ 1 )ξW
+ (ξ 2 )W W Wη + (ξ 3 )W W Wθ − (φ3 )W W ]
− (Vξ )2 [(ξ 1 )V V Wξ + (ξ 2 )V V Wη + (ξ 3 )V V Wθ − (φ3 )V V ]
− 2Pξ [(ξ 1 )P Wξξ + (ξ 2 )P Wξη + (ξ 3 )P Wξθ + (ξ 1 )U P Uξ Wξ
+ (ξ 1 )V P Vξ Wξ + (ξ 1 )W P (Wξ )2 + (ξ 2 )U P Uξ Wη
+ (ξ 2 )V P Vξ Wη + (ξ 2 )W P Wξ Wη + (ξ 2 )ξP Wη
+ (ξ 3 )U P Uξ Wθ + (ξ 3 )V P Vξ Wθ + (ξ 3 )W P Wξ Wθ
+ (ξ 1 )ξP Wξ + (ξ 3 )ξP Wθ − (φ3 )U P Uξ − (φ3 )V P Vξ
− (φ3 )W P Wξ − (φ3 )ξP ]
− Uξξ [(ξ 1 )U Wξ + (ξ 2 )U Wη + (ξ 3 )U Wθ − (φ3 )U ]
(65.303)
− Vξξ [(ξ 1 )V Wξ + (ξ 2 )V Wη + (ξ 3 )V Wθ − (φ3 )V ]
− 2Wξη [(ξ 2 )U Uξ + (ξ 2 )V Vξ + (ξ 2 )W Wξ + (ξ 2 )ξ ]
− 2Wξθ [(ξ 3 )U Uξ + (ξ 3 )V Vξ + (ξ 3 )W Wξ + (ξ 3 )ξ ]
− Wξξ [2(ξ 1 )U Uξ + 2(ξ 1 )V Vξ + 3(ξ 1 )W Wξ + 2(ξ 1 )ξ
+ (ξ 2 )W Wη + (ξ 3 )W Wθ − (φ3 )W ]
− 2Uξ [(ξ 1 )U V Vξ Wξ + (ξ 1 )ξU Wξ + (ξ 2 )U V Vξ Wη
+ (ξ 2 )U W Wξ Wη + (ξ 2 )ξU Wη + (ξ 3 )U V Vξ Wθ
+ (ξ 3 )U W Wξ Wθ + (ξ 3 )ξU Wθ − (φ3 )U V Vξ
− (φ3 )U W Wξ − (φ3 )ξU ]
− 2Vξ [(ξ 1 )ξV Wξ + (ξ 2 )V W Wη Wξ + (ξ 2 )ξV Wη + (ξ 3 )V W Wξ Wθ
+ (ξ 3 )ξV Wθ − (φ3 )V W Wξ − (φ3 )ξV ]
− Wξ [(ξ 1 )ξξ − 2(φ3 )ξW ] − Wη [(ξ 2 )ξξ − 2(ξ 2 )ξW Wξ ]
− Wθ [(ξ 3 )ξξ − 2(ξ 3 )ξW Wξ ] + (φ3 )ξξ
820
χηη = −Pηη [(ξ 1 )P Wξ + (ξ 2 )P Wη + (ξ 3 )P Wθ − (φ3 )P ]
− Uηη [(ξ 1 )U Wξ + (ξ 2 )U Wη + (ξ 3 )U Wθ − (φ3 )U ]
− Vηη [(ξ 1 )V Wξ + (ξ 2 )V Wη + (ξ 3 )V Wθ − (φ3 )V ]
− Wηη [(ξ 1 )W Wξ + 2(ξ 2 )U Uη + 2(ξ 2 )V Vη + 3(ξ 2 )W Wη
+ 2(ξ 2 )η + 2(ξ 3 )W Wθ − (φ3 )W ]
− 2Wηθ [(ξ 3 )U Uη + (ξ 3 )V Vη + (ξ 3 )W Wη + (ξ 3 )η ]
− 2Wξη [(ξ 1 )U Uη + (ξ 1 )V Vη + (ξ 1 )W Wη + (ξ 1 )η ]
− 2Pη [(ξ 1 )P Wξη + (ξ 2 )P Wηη + (ξ 3 )P Wηθ + (ξ 1 )U P Uη Wξ
+ (ξ 1 )V P Vη Wξ + (ξ 1 )ηP Wξ + (ξ 2 )U P Uη Wη
+ (ξ 2 )V P Vη Wη + (ξ 2 )ηP Wη + (ξ 3 )U P Uη Wθ
+ (ξ 3 )V P Vη Wθ + (ξ 3 )W P Wη Wθ + (ξ 3 )ηP Wθ
+ (ξ 2 )W P (Wη )2 − (φ3 )U P Uη − (φ3 )V P Vη
− (φ3 )W P Wη − (φ3 )ηP ] (65.304)
2 1 2 3 3
− (Pη ) [(ξ )P P Wξ + (ξ )P P Wη + (ξ )P P Wθ − (φ )P P ]
− (Uη )2 [(ξ 1 )U U Wξ + (ξ 2 )U U Wη + (ξ 3 )U U Wθ − (φ3 )U U ]
− (Vη )2 [(ξ 1 )V V Wξ + (ξ 2 )V V Wη + (ξ 3 )V V Wθ − (φ3 )V V ]
− (Wη )2 [(ξ 1 )W W Wξ + 2(ξ 2 )U W Uη + 2(ξ 2 )V W Vη
+ (ξ 2 )W W Wη + 2(ξ 2 )ηW + (ξ 3 )W W Wθ − (φ3 )W W ]
− 2Uη [(ξ 1 )U V Vη Wξ + (ξ 1 )U W Wη Wξ + (ξ 1 )ηU Wξ + (ξ 2 )U V Vη Wη
+ (ξ 2 )ηU Wη + (ξ 3 )U V Vη Wθ + (ξ 3 )U W Wη Wθ + (ξ 3 )ηU Wθ
− (φ3 )U V Vη − (φ3 )U W Wη − (φ3 )ηU ]
− 2Vη [(ξ 1 )ηV Wξ + (ξ 2 )ηV Wη + (ξ 3 )V W Wη Wθ + (ξ 3 )ηV Wθ
− (φ3 )V W Wη − (φ3 )ηV ]
− Wη [(ξ 2 )ηη + 2(ξ 1 )ηW Wξ + 2(ξ 3 )ηW Wθ − 2(φ3 )ηW ]
− (ξ 1 )ηη Wξ − (ξ 3 )ηη Wθ + (φ3 )ηη
821
χθθ = −Pθθ [(ξ 1 )P Wξ + (ξ 2 )P Wη + (ξ 3 )P Wθ − (φ3 )P ]
− Uθθ [(ξ 1 )U Wξ + (ξ 2 )U Wη + (ξ 3 )U Wθ − (φ3 )U ]
− Vθθ [(ξ 1 )V Wξ + (ξ 2 )V Wη + (ξ 3 )V Wθ − (φ3 )V ]
− 2Pθ [(ξ 2 )P Wηθ + (ξ 3 )P Wθθ + (ξ 1 )P Wξθ + (ξ 2 )U P Uθ Wξ
+ (ξ 1 )V P Vθ Wξ + (ξ 1 )W P Wθ Wξ + (ξ 1 )θP Wξ + (ξ 2 )U P Uθ Wη
+ (ξ 2 )V P Vθ Wη + (ξ 2 )W P Wη Wθ + (ξ 2 )θP Wη + (ξ 3 )U P Uθ Wθ
+ (ξ 3 )V P Vθ Wθ + (ξ 3 )W P (Wθ )2 + (ξ 3 )θP Wθ − (φ3 )U P Uθ
− (φ3 )V P Vθ − (φ3 )W P Wθ − (φ3 )θP ]
− (Pθ )2 [(ξ 1 )P P Wξ + (ξ 2 )P P Wη + (ξ 3 )P P Wθ − (φ3 )P P ]
− (Uθ )2 [(ξ 1 )U U Wξ + (ξ 2 )U U Wη + (ξ 3 )U U Wθ − (φ3 )U U ]
− (Vθ )2 [(ξ 1 )V V Wξ + (ξ 2 )V V Wη + (ξ 3 )V V Wθ − (φ3 )V V ]
− (Wθ )2 [2(ξ 3 )V W Vθ + (ξ 1 )W W Wξ + (ξ 2 )W W Wη
(65.305)
+ 2(ξ 3 )θW + (ξ 3 )W W Wθ − (φ3 )W W ]
− 2Wηθ [(ξ 2 )U Uθ + (ξ 2 )V Vθ + (ξ 2 )W Wθ + (ξ 2 )θ ]
− Wθθ [(ξ 1 )W Wξ + (ξ 2 )W Wη + 2(ξ 3 )U Uθ + 2(ξ 3 )V Vθ
+ 3(ξ 3 )W Wθ + 2(ξ 3 )θ − (φ3 )W ]
− 2Wξθ [(ξ 1 )U Uθ + (ξ 1 )V Vθ + (ξ 1 )W Wθ + (ξ 1 )θ ]
− 2Uθ [(ξ 1 )U V Vθ Wξ + (ξ 1 )U W Wθ Wξ + (ξ 1 )θU Wξ + (ξ 2 )U V Vθ Wη
+ (ξ 2 )U W Wη Wθ + (ξ 2 )θU Wη + (ξ 3 )U V Vθ Wθ + (ξ 3 )U V (Wθ )2
+ (ξ 3 )θU Wθ − (φ3 )U V Vθ − (φ3 )U W Wθ − (φ3 )θU ]
− 2Vθ [(ξ 1 )V W Wξ Wθ + (ξ 2 )V W Wη Wθ + (ξ 1 )θV Wξ + (ξ 2 )θV Wη
+ (ξ 3 )θV Wθ − (φ3 )θV ]
− 2Wθ [(ξ 1 )θW Wξ + (ξ 2 )θW Wη − (φ3 )V W Vθ − (φ3 )θW ]
− (ξ 1 )θθ Wξ − (ξ 2 )θθ Wη − (ξ 3 )θθ Wθ + (φ3 )θθ
Let us consider the condition
P r(2) Γ[∆(4) ] =0 (65.306)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
first. Since
P r(2) Γ[∆(4) ] = φξ + ψ η + χθ
822
we obtain, taking into account (65.285), (65.289) and (65.293) that
P r(2) Γ[∆(4) ]
= (φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U
− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ]
(65.307)
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ]
+ (φ3 )θ + (φ3 )U Uθ + (φ3 )V Vθ + (φ3 )W Wθ − (Wθ )2 (ξ 3 )W
− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ]
1 1 1 1
Pξ −→ U + ξUξ + ηUη + θUθ − U Uξ − V Uη − W Uθ
2 2 2 2 (65.308)
+ ν(Uξξ + Uηη + Uθθ )
1 1 1 1
Pη −→ V + ξVξ + ηVη + θVθ − U Vξ − V Vη − W Vθ
2 2 2 2 (65.309)
+ ν(Vξξ + Vηη + Vθθ )
1 1 1 1
Pθ −→ W + ξWξ + ηWη + θWθ − U Wξ − V Wη − W Wθ
2 2 2 2 (65.310)
+ ν(Wξξ + Wηη + Wθθ )
and
Wθ −→ − (Uξ + Vη ) (65.311)
823
Taking into account the above substitutions, the symmetry condition takes
on the form
(φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ − (Uξ )2 (ξ 1 )U

1 1 1 1
2 2 2 2

+ ν(Uξξ + Uηη + Uθθ )
×[Uξ (ξ 1 )P + Uη (ξ 2 )P + Uθ (ξ 3 )P − (φ1 )P ]
− Uξ [Vξ (ξ 1 )V + Wξ (ξ 1 )W + (ξ 1 )ξ ]

1 1 1 1
2 2 2 2

+ ν(Vξξ + Vηη + Vθθ )
(65.312)
×[Vξ (ξ 1 )P + Vη (ξ 2 )P + Vθ (ξ 3 )P − (φ2 )P ]
− Vξ [Uη (ξ 1 )U + Wη (ξ 1 )W + (ξ 1 )η ]
− (Uξ + Vη )2 (ξ 3 )W

1 1 1 1
2 2 2 2

− V Wη + W (Uξ + Vη ) + ν(Wξξ + Wηη + Wθθ )
×[Wξ (ξ 1 )P + Wη (ξ 2 )P − (Uξ + Vη )(ξ 3 )P − (φ3 )P ]

− Wξ [Uθ (ξ 1 )U + Vθ (ξ 1 )V + (ξ 1 )θ ]
− Wη [Uθ (ξ 2 )U + Vθ (ξ 2 )V − (Uξ + Vη )(ξ 2 )W + (ξ 2 )θ ]
+ (Uξ + Vη ) [Wξ (ξ 1 )W + Uθ (ξ 3 )U + Vθ (ξ 3 )V + (ξ 3 )θ ] = 0
We equate the coefficients of the partial derivatives of the functions U, V, W
to zero. We thus have
824
The coefficient of Uξ Vξ is −(ξ 1 )V . Therefore
(ξ 1 )V = 0 (65.313)
The coefficients of Uξ Vθ and Uθ Vξ are (ξ 3 )V and −(ξ 3 )V respecively. Therefore
(ξ 3 )V = 0 (65.314)
The coefficients of Uθ Vη and Uη Vθ are (ξ 3 )U and −(ξ 3 )U respecively. Therefore
(ξ 3 )U = 0 (65.315)
The terms Uθ Wη and Uη Vη have common coefficient −(ξ 2 )U . Therefore
(ξ 2 )U = 0 (65.316)
The coefficient of Uη Wξ is −(ξ 2 )W . Therefore
(ξ 2 )W = 0 (65.317)
The coefficient of Wη Vξ is −(ξ 1 )W . Therefore
(ξ 1 )W = 0 (65.318)
1
The terms (Wξ )2 and (Vξ )2 have common coefficient U − ξ (ξ 1 )P .
2
Therefore
(ξ 1 )P = 0 (65.319)
1
The terms (Uη )2 and (Wη )2 have common coefficient V − η (ξ 2 )P .
2
Therefore
(ξ 2 )P = 0 (65.320)
1
The terms (Uθ )2 and (Vθ )2 have common coefficient W − θ (ξ 3 )P .
2
Therefore
(ξ 3 )P = 0 (65.321)
825
The coefficient of Uξ Vη is (2W − θ)(ξ 3 )P − 2(ξ 3 )W and since (ξ 3 )P = 0, we
have
(ξ 3 )W = 0 (65.322)
The coefficient of Uθ Wξ is −(ξ 1 )U − (ξ 3 )W and since (ξ 3 )W = 0, we have
(ξ 1 )U = 0 (65.323)
1 1
The coefficient of (Vη )2 is V − η (ξ 2 )P + W − θ (ξ 3 )P − (ξ 2 )V − (ξ 3 )W
2 2
and since we have already found (ξ 2 )P = (ξ 3 )P = (ξ 3 )W = 0, we get
(ξ 2 )V = 0 (65.324)
The terms Uξξ , Uηη and Uθθ have common coefficient −ν[(ξ 1 )P Uξ + (ξ 2 )P Uη +
(ξ 3 )P Uθ ]+ν(φ1 )P and since we have already found (ξ 1 )P = (ξ 2 )P = (ξ 3 )P = 0,
we obtain
(φ1 )P = 0 (65.325)
The terms Vξξ , Vηη and Vθθ have common coefficient −ν[(ξ 1 )P Vξ + (ξ 2 )P Vη +
(ξ 3 )P Vθ ]+ν(φ2 )P and since we have already found (ξ 1 )P = (ξ 2 )P = (ξ 3 )P = 0,
we obtain
(φ2 )P = 0 (65.326)
The terms Wξξ , Wηη and Wθθ have common coefficient −ν[(ξ 1 )P Wξ +(ξ 2 )P Wη +
(ξ 3 )P Wθ ] + ν(φ2 )P and since we have already found (ξ 1 )P = (ξ 2 )P = (ξ 3 )P =
0, we obtain
(φ3 )P = 0 (65.327)
Let us collect all those preliminary results we have already found:
(ξ 1 )U = (ξ 2 )U = (ξ 3 )U = 0, (ξ 1 )V = (ξ 2 )V = (ξ 3 )V = 0,
(ξ 1 )W = (ξ 2 )W = (ξ 3 )W = 0, (ξ 1 )P = (ξ 2 )P = (ξ 3 )P = 0, (65.328)
(φ1 )P = (φ2 )P = (φ3 )P = 0
826
Therefore the symmetry condition takes the form
(φ1 )ξ + (φ1 )W Wξ + (φ1 )V Vξ + (φ1 )U Uξ
− Uξ (ξ 1 )ξ − Uη (ξ 2 )ξ − Uθ (ξ 3 )ξ
+ (φ2 )η + (φ2 )U Uη + (φ2 )V Vη + (φ2 )W Wη
(65.329)
− Vξ (ξ 1 )η − Vη (ξ 2 )η − Vθ (ξ 3 )η
− Wξ (ξ 1 )θ − Wη (ξ 2 )θ + (Uξ + Vη ) (ξ 3 )θ = 0
Equating the coefficients of the partial derivatives of the functions U, V, W
to zero, we obtain :
(φ1 )U − (ξ 1 )ξ − (φ3 )W + (ξ 3 )θ = 0 (65.330)
(φ1 )V − (ξ 1 )η = 0 (65.331)
(φ1 )W − (ξ 1 )θ = 0 (65.332)
(φ2 )U − (ξ 2 )ξ = 0 (65.333)
(φ3 )U − (ξ 3 )ξ = 0 (65.334)
(φ2 )V − (ξ 2 )η − (φ3 )W + (ξ 3 )θ = 0 (65.335)
(φ2 )W − (ξ 2 )θ = 0 (65.336)
(φ3 )V − (ξ 3 )η = 0 (65.337)
827
The constant term
(φ1 )ξ + (φ2 )η + (φ3 )θ = 0 (65.338)
P r(2) Γ[∆(1) ] =0 (65.339)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
Since (considering (65.281))

1 1 1 1
2 2 2 2
1
+ Uη φ2 + Uθ φ3 + − ξ + U φξ
2 (65.340)
1 1
+ − η + V φη + − θ + W φθ
2 2
+ π − νφ − νφ − νφθθ
ξ ξξ ηη
and taking into account (65.285)-(65.287) for the φξ , φη , φθ respectively, (65.294)

for π ξ and (65.297)-(65.299) for φξξ , φηη , φθθ respectively. We then make the
substitutions (65.308)-(65.311) to account for the conditions ∆(1) = 0, ∆(2) =
0, ∆(3) = 0 and ∆(4) = 0. Finally we take into account the values (65.328).
1 1 1 1
− Uξ ξ 1 − Uη ξ 2 − Uθ ξ 3 + − + Uξ φ1
2 2 2 2
2 3
+ Uη φ + Uθ φ
1
2

1 2 3
− Uξ (ξ )ξ − Uη (ξ )ξ − Uθ (ξ )ξ
1
2

1 2 3
− Uξ (ξ )η − Uη (ξ )η − Uθ (ξ )η
1
+ − θ + W (φ1 )θ + (φ1 )U Uθ + (φ1 )V Vθ
2

1 1 2 3
− (φ )W (Uξ + Vη ) − Uξ (ξ )θ − Uη (ξ )θ − Uθ (ξ )θ
828
+ (φ4 )ξ + (φ4 )U Uξ + (φ4 )V Vξ + (φ4 )W Wξ

1 4 1 1 1 1
− [(ξ )ξ − (φ )P ] U + ξUξ + ηUη + θUθ
2 2 2 2

− U Uξ − V Uη − W Uθ + ν(Uξξ + Uηη + Uθθ )

2 1 1 1 1
− (ξ )ξ V + ξVξ + ηVη + θVθ − U Vξ
2 2 2 2

− V Vη − W Vθ + ν(Vξξ + Vηη + Vθθ )

3 1 1 1 1
− (ξ )ξ W + ξWξ + ηWη + θWθ − U Wξ
2 2 2 2

− V Wη + W (Uξ + Vη ) + ν(Wξξ + Wηη + Wθθ )

− ν − 2Uξη (ξ 2 )ξ − 2Uξθ (ξ 3 )ξ
− Uξξ [2(ξ 1 )ξ − (φ1 )U ] + Vξξ (φ1 )V + Wξξ (φ1 )W

+ (Uξ )2 (φ1 )U U + (Vξ )2 (φ1 )V V + (Wξ )2 (φ1 )W W
− Uξ [(ξ 1 )ξξ − 2(φ1 )U V Vξ − 2(φ1 )U W Wξ − 2(φ1 )ξU ]
+ 2Vξ [(φ1 )V W Wξ + (φ1 )ξV ]

1 2 3 1
+ 2Wξ (φ )ξW − (ξ )ξξ Uη − (ξ )ξξ Uθ + (φ )ξξ

− ν Uηη (φ1 )U − 2Uηθ (ξ 3 )η − 2Uξη (ξ 1 )η
+ Vηη (φ1 )V + Wηη (φ1 )W + (Uη )2 (φ1 )U U + (Vη )2 (φ1 )V V

+ (Wη )2 (φ1 )W W + 2Uη [(φ1 )U V Vη + (φ1 )U W Wη + (φ1 )ηU ]

1 1 1
+ 2 [(φ )V W Vη Wη + (φ )ηV Vη + (φ )ηW Wη ]
829

− ν − 2Uηθ (ξ 2 )θ − Uθθ [2(ξ 3 )θ − (φ1 )U ] − 2Uξθ (ξ 1 )θ
+ Vθθ (φ1 )V + Wθθ (φ1 )W + (Uθ )2 (φ1 )U U

+ 2Uθ [(φ1 )U V Vθ − (φ1 )U W (Uξ + Vη ) + (φ1 )θU ]
(65.341)
+ (Vθ )2 (φ1 )V V + (Uξ + Vη )2 (φ1 )W W
− (ξ 1 )θθ Uξ − (ξ 2 )θθ Uη − (ξ 3 )θθ Uθ + (φ1 )θθ

1 1 1
+ 2 [−(φ )V W Vθ (Uξ + Vη ) + (φ )θV Vθ − (φ )θW (Uξ + Vη )] = 0
The terms Uξ Vξ , Uη Vη and Uθ Vθ have common coefficient −2ν(φ1 )U V . The

terms Uθ Uξ and Uθ Vη have common coefficient 2ν(φ1 )U W . The terms Uη Wη
and Uξ Wξ have common coefficient −2ν(φ1 )U W . The terms Uξ Vθ and Vη Vθ
have common coefficient 2ν(φ1 )V W . The terms Vη Wη and Vξ Wξ have com-
mon coefficient −2ν(φ1 )V W . The coefficint of the term Uξ Vη is −2ν(φ1 )W W .
The terms (Uη )2 and (Uθ )2 have common coefficint −ν(φ1 )U U . The coeffi-
cient of the term (Uξ )2 is −ν(φ1 )U U − ν(φ1 )W W . The terms (Vξ )2 and (Vθ )2
have common coefficient −ν(φ1 )V V . The coefficient of the term (Vη )2 is
−ν(φ1 )V V − ν(φ1 )W W . The terms (Wη )2 and (Wξ )2 have common coefficint
−ν(φ1 )W W . Equating all these coefficients to zero, we obtain
(φ1 )U U = (φ1 )U V = (φ1 )U W = 0

(65.342)
(φ1 )V V = (φ1 )V W = (φ1 )W W = 0
The coefficient of the term Uηθ is 2ν[(ξ 2 )θ + (ξ 3 )η ]. Therefore we shall have
(ξ 2 )θ + (ξ 3 )η = 0 (65.343)
The coefficient of the term Uξη is 2ν[(ξ 2 )ξ + (ξ 1 )η ]. Therefore we shall have
(ξ 2 )ξ + (ξ 1 )η = 0 (65.344)
The coefficient of the term Uξθ is 2ν[(ξ 3 )ξ + (ξ 1 )θ ]. Therefore we shall have
(ξ 3 )ξ + (ξ 1 )θ = 0 (65.345)
The coefficient of the term Uξξ is ν[(ξ 1 )ξ + (φ4 )P − (φ1 )U ]. Therefore we shall
have
(ξ 1 )ξ + (φ4 )P − (φ1 )U = 0 (65.346)
830
The coefficient of the term Uηη is ν[−(ξ 1 )ξ +2(ξ 2 )η +(φ4 )P −(φ1 )U ]. Therefore
we shall have
−(ξ 1 )ξ + 2(ξ 2 )η + (φ4 )P − (φ1 )U = 0 (65.347)
The coefficient of the term Uθθ is ν[−(ξ 1 )ξ +2(ξ 3 )θ +(φ4 )P −(φ1 )U ]. Therefore
we shall have
−(ξ 1 )ξ + 2(ξ 3 )θ + (φ4 )P − (φ1 )U = 0 (65.348)
The terms Vξξ , Vηη and Vθθ have common coefficient −ν[(φ1 )V + (ξ 2 )ξ ].
Therefore we shall have
(φ1 )V + (ξ 2 )ξ = 0 (65.349)
The terms Wξξ , Wηη and Wθθ have common coefficient −ν[(φ1 )W + (ξ 3 )ξ ].
Therefore we shall have
(φ1 )W + (ξ 3 )ξ = 0 (65.350)
In view of (65.342)-(65.350), equation (65.341) becomes

1 1 1 1
− Uξ ξ 1 − Uη ξ 2 − Uθ ξ 3 + − + Uξ φ1
2 2 2 2
2 3
+ Uη φ + Uθ φ
1
2

1 2 3
− Uξ (ξ )ξ − Uη (ξ )ξ − Uθ (ξ )ξ
1
2

1 2 3
− Uξ (ξ )η − Uη (ξ )η − Uθ (ξ )η
1
+ − θ + W (φ1 )θ + (φ1 )U Uθ + (φ1 )V Vθ
2

1 1 2 3
− (φ )W (Uξ + Vη ) − Uξ (ξ )θ − Uη (ξ )θ − Uθ (ξ )θ
831
+ (φ4 )ξ + (φ4 )U Uξ + (φ4 )V Vξ + (φ4 )W Wξ

1 4 1 1 1 1
− [(ξ )ξ − (φ )P ] U + ξUξ + ηUη + θUθ
2 2 2 2

− U Uξ − V Uη − W Uθ

2 1 1 1 1
− (ξ )ξ V + ξVξ + ηVη + θVθ − U Vξ
2 2 2 2

− V Vη − W Vθ

3 1 1 1 1
− (ξ )ξ W + ξWξ + ηWη + θWθ − U Wξ
2 2 2 2

− V Wη + W (Uξ + Vη )

− ν − Uξ [(ξ 1 )ξξ − 2(φ1 )ξU ] + 2Vξ (φ1 )ξV

1 2 3 1
+ 2Wξ (φ )ξW − (ξ )ξξ Uη − (ξ )ξξ Uθ + (φ )ξξ

1 1 1
− ν 2Uη (φ )ηU + 2 [(φ )ηV Vη + (φ )ηW Wη ]

− ν 2Uθ (φ1 )θU − (ξ 1 )θθ Uξ − (ξ 2 )θθ Uη − (ξ 3 )θθ Uθ + (φ1 )θθ
(65.351)
1 1
+ 2 [(φ )θV Vθ − (φ )θW (Uξ + Vη )] = 0
At this stage we collect together the terms with respect to Uξ , Uη , Uθ , Vξ , Vη , Vθ

and Wξ , Wη and then equate to zero all these coefficints:
The coefficient of Uξ is
ν[(ξ 1 )ξξ + (ξ 1 )ηη + (ξ 1 )θθ ] + 2ν[(φ1 )θW − (φ1 )ξU ]

+ U [(φ1 )U − (φ4 )P ] − W [(ξ 1 )θ + (ξ 3 )ξ + (φ1 )W ]
1 (65.352)
+ [η(ξ 1 )η + θ(ξ 1 )θ + θ(ξ 3 )ξ − ξ(φ1 )U + θ(φ1 )W + ξ(φ4 )P ]
2
1
− ξ 1 + φ1 − (ξ 1 )η V + (φ4 )U = 0
2
832
The coefficient of Uη is
ν[(ξ 2 )ξξ + (ξ 2 )ηη + (ξ 2 )θθ ] − 2ν(φ1 )ηU − (ξ 2 )ξ U

+ V [(ξ 1 )ξ − (ξ 2 )η + (φ1 )U − (φ4 )P ] − (ξ 2 )θ W
1 (65.353)
+ [η(ξ 2 )η − η(ξ 1 )ξ + θ(ξ 2 )θ + ξ(ξ 2 )ξ − η(φ1 )U + η(φ4 )P ]
2
1
− ξ 2 + φ2 = 0
2
The coefficient of Uθ is
ν[(ξ 3 )ξξ + (ξ 3 )ηη + (ξ 3 )θθ ] − 2ν(φ1 )θU − (ξ 3 )η V
+ W [(ξ 1 )ξ − (ξ 3 )θ + (φ1 )U − (φ4 )P ] − (ξ 3 )ξ U
1 (65.354)
+ [η(ξ 3 )η − θ(ξ 1 )ξ + θ(ξ 3 )θ + ξ(ξ 3 )ξ − θ(φ1 )U + θ(φ4 )P ]
2
1
− ξ 3 + φ3 = 0
2
The coefficient of Vξ is
1
(φ4 )V − 2ν(φ1 )ξV + U − ξ [(φ1 )V + (ξ 2 )ξ ] = 0 (65.355)
2
The coefficient of Vη is
1 1

1 1
2ν[(φ )θW − (φ )ηV ] + V − η [(φ )V + (ξ 2 )ξ ]
2 (65.356)
1 1 3
− W − θ [(φ )W + (ξ )ξ ] = 0
2
The coefficient of Vθ is
1
− 2ν(φ1 )θV + W − θ [(φ1 )V + (ξ 2 )ξ ] = 0 (65.357)
2
The coefficient of Wξ is
1
(φ4 )W − 2ν(φ1 )ξW + U − ξ [(φ1 )W + (ξ 3 )ξ ] = 0 (65.358)
2
The coefficient of Wη is
1
− 2ν(φ1 )ηW + V − η [(φ1 )W + (ξ 3 )ξ ] = 0 (65.359)
2
833
1
− ν [(φ1 )ξξ + (φ1 )ηη + (φ1 )θθ ] + (φ4 )ξ − φ1
2
1 1 1
+ U (φ1 )ξ − (ξ 1 )ξ + (φ4 )P + V (φ1 )η − (ξ 2 )ξ (65.360)
2 2 2
1 1
+ W (φ1 )θ − (ξ 3 )ξ − [ξ(φ1 )ξ + η(φ1 )η + θ(φ1 )θ ] = 0
2 2
P r(2) Γ[∆(2) ] =0 (65.361)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
We have that (considering (65.282))

1 1 1
2 2 2
1 1
2 3
+ − + Vη φ + Vθ φ + − ξ + U ψ ξ
2 2 (65.362)
1 1
+ − η + V ψ + − θ + W ψθ
η
2 2
+ π η − νψ ξξ − νψ ηη − νψ θθ
Substituting ψ ξ , ψ η , ψ θ given by (65.288)-(65.290), π η given by (65.295)

and ψ ξξ , ψ ηη , ψ θθ given by (65.300)-(65.302) and also Pξ , Pη , Pθ , Wθ given
by (65.308)-(65.311) and finally taking into account (65.328), we obtain the
834
equation
1 1 1 1
− Vξ ξ − Vη ξ − Vθ ξ + Vξ φ + − + Vη φ2 + Vθ φ3
1 2 3 1
2 2 2 2
1
2

1 2 3
− Vξ (ξ )ξ − Vη (ξ )ξ − Vθ (ξ )ξ
1
2

1 2 3
− Vξ (ξ )η − Vη (ξ )η − Vθ (ξ )η
1
2

1 2 3
− Vξ (ξ )θ − Vη (ξ )θ − Vθ (ξ )θ

+ (φ4 )η + (φ4 )U Uη + (φ4 )V Vη + (φ4 )W Wη
1 1 1 1
2 2 2 2
1
+ ν(Uξξ + Uηη + Uθθ ) (ξ )η
1 1 1 1
2 2 2 2
+ ν(Vξξ + Vηη + Vθθ ) [(ξ 2 )η − (φ4 )P ]
1 1 1 1
− W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ − V Wη
2 2 2 2

3
+ W (Uξ + Vη ) + ν(Wξξ + Wηη + Wθθ ) (ξ )η
835

− ν Uξξ (φ2 )U − 2Vξη (ξ 2 )ξ − 2Vξθ (ξ 3 )ξ
− Vξξ [2(ξ 1 )ξ − (φ2 )V ] + Wξξ (φ2 )W + (Uξ )2 (φ2 )U U

+ (Vξ )2 (φ2 )V V + (Wξ )2 (φ2 )W W
+ 2Uξ [(φ2 )U V Vξ + (φ2 )U W Wξ + (φ2 )ξU ]
− Vξ [(ξ 1 )ξξ − 2(φ2 )V W Wξ − 2(φ2 )ξV ] − Vη (ξ 2 )ξξ

3 2 2
− Vθ (ξ )ξξ + 2(φ )ξW Wξ + (φ )ξξ

− ν Uηη (φ2 )U − Vηη [2(ξ 2 )η − (φ2 )V ]
− 2Vηθ (ξ 3 )η − 2Vξη (ξ 1 )η + Wηη (φ2 )W + (Uη )2 (φ2 )U U

+ (Vη )2 (φ2 )V V + (Wη )2 (φ2 )W W
(65.363)
+ 2Uη [(φ2 )U V Vη + (φ2 )U W Wη + (φ2 )ηU ]
+ 2Vη [(φ2 )ηV + (φ2 )V W Wη ] + 2Wη (φ2 )ηW

1 2 3 2
− (ξ )ηη Vξ − (ξ )ηη Vη − (ξ )ηη Vθ + (φ )ηη

− ν (Uθ )2 (φ2 )U U + (Vθ )2 (φ2 )V V + (Uξ + Vη )2 (φ2 )W W
+ Uθθ (φ2 )U − 2Vηθ (ξ 2 )θ − Vθθ [2(ξ 3 )θ − (φ2 )V ]

− 2Vξθ (ξ 1 )θ + Wθθ (φ2 )W
+ 2Uθ [(φ2 )U V Vθ − (φ2 )U W (Uξ + Vη ) + (φ2 )θU ]
+ 2Vθ [−(φ2 )V W (Uξ + Vη ) + (φ2 )θV ] − (ξ 1 )θθ Vξ − (ξ 2 )θθ Vη

3 2 2
− (ξ )θθ Vθ − 2(φ )θW (Uξ + Vη ) + (φ )θθ = 0
The above equation is the expanded version of the invariance condition

(65.361). Equate next to zero the coefficients of the partial derivatives of
the functions U, V and W :
The terms (Uθ )2 and (Uη )2 have common coefficient −ν(φ2 )U U while the co-
efficient of (Uξ )2 is −ν(φ2 )U U − ν(φ2 )W W . Therefore we obtain the equation
(φ2 )U U = (φ2 )W W = 0 (65.364)
836
The terms Uξ Vξ , Uη Vη and Uθ Vθ have common coefficient −ν(φ2 )U V . There-
fore we have the equation
(φ2 )U V = 0 (65.365)
The terms Uθ Uξ and Uθ Vη have common coefficient 2ν(φ2 )U W while the terms
Uη Wη and Uξ Wξ have common coefficient −2ν(φ2 )U W . Therefore we get the
equation
(φ2 )U W = 0 (65.366)
The terms (Vξ )2 and (Vθ )2 have common coefficient −ν(φ2 )V V while the co-
efficient of (Vη )2 is −ν(φ2 )V V − ν(φ2 )W W . Therefore we obtain the equation
(φ2 )V V = (φ2 )W W = 0 (65.367)
The terms Vθ Uξ and Vθ Vη have common coefficient 2ν(φ2 )V W while the terms
Vη Wη and Vξ Wξ have common coefficient −2ν(φ2 )V W . Therefore we get the
equation
(φ2 )V W = 0 (65.368)
The terms (Wξ )2 and (Wη )2 have common coefficient −ν(φ2 )W W while the
coefficient of Uξ Vη is −2ν(φ2 )W W . Therefore we obtain the equation
(φ2 )W W = 0 (65.369)
Collecting (65.364)-(65.369) we obtain
(φ2 )U U = (φ2 )U V = (φ2 )U W = 0

(65.370)
(φ2 )V V = (φ2 )V W = (φ2 )W W = 0
The terms Uξξ , Uηη and Uθθ have common coefficient −ν[(ξ 1 )η + (φ2 )U ]. We
thus have the equation
(ξ 1 )η + (φ2 )U = 0 (65.371)
The coefficient of Vηη is ν[(ξ 2 )η + (φ4 )P − (φ2 )V ]. We thus have the equation
(ξ 2 )η + (φ4 )P − (φ2 )V = 0 (65.372)
837
The coefficient of Vθθ is ν[2(ξ 3 )θ − (ξ 2 )η + (φ4 )P − (φ2 )V ]. We thus have the
equation
2(ξ 3 )θ − (ξ 2 )η + (φ4 )P − (φ2 )V = 0 (65.373)
The coefficient of Vξξ is ν[2(ξ 1 )ξ − (ξ 2 )η + (φ4 )P − (φ2 )V ]. We thus have the

equation
2(ξ 1 )ξ − (ξ 2 )η + (φ4 )P − (φ2 )V = 0 (65.374)
The terms Wξξ , Wηη and Wθθ have common coefficient −ν[(ξ 3 )η + (φ2 )W ].
We thus have the equation
(ξ 3 )η + (φ2 )W = 0 (65.375)
The coefficient of Vηθ is given by 2ν[(ξ 2 )θ + (ξ 3 )η ]. Therefore we get the

equation
(ξ 2 )θ + (ξ 3 )η = 0 (65.376)
The coefficient of Vξη is given by 2ν[(ξ 1 )η + (ξ 2 )ξ ]. Therefore we get the

equation
(ξ 1 )η + (ξ 2 )ξ = 0 (65.377)
The coefficient of Vξθ is given by 2ν[(ξ 1 )θ + (ξ 3 )ξ ]. Therefore we get the

equation
(ξ 1 )θ + (ξ 3 )ξ = 0 (65.378)
838
Taking into account (65.370)-(65.378), equation (65.363) gets simplified to
1 1 1 1
− Vξ ξ 1 − Vη ξ 2 − Vθ ξ 3 + Vξ φ1 + − + Vη φ2 + Vθ φ3
2 2 2 2
1
2

1 2 3
− Vξ (ξ )ξ − Vη (ξ )ξ − Vθ (ξ )ξ
1
2

1 2 3
− Vξ (ξ )η − Vη (ξ )η − Vθ (ξ )η
1
2

1 2 3
− Vξ (ξ )θ − Vη (ξ )θ − Vθ (ξ )θ

+ (φ4 )η + (φ4 )U Uη + (φ4 )V Vη + (φ4 )W Wη
1 1 1 1
− U + ξUξ + ηUη + θUθ − U Uξ − V Uη
2 2 2 2
1
− W Uθ (ξ )η
1 1 1 1
− V + ξVξ + ηVη + θVθ − U Vξ − V Vη
2 2 2 2
− W Vθ [(ξ 2 )η − (φ4 )P ]
1 1 1 1
2 2 2 2

3
− V Wη + W (Uξ + Vη ) (ξ )η
839

− ν 2Uξ (φ2 )ξU − Vξ [(ξ 1 )ξξ − 2(φ2 )ξV ] − Vη (ξ 2 )ξξ

3 2 2
− Vθ (ξ )ξξ + 2(φ )ξW Wξ + (φ )ξξ

− ν 2Uη (φ2 )ηU + 2Vη (φ2 )ηV + 2Wη (φ2 )ηW
(65.379)
1 2 3 2
− (ξ )ηη Vξ − (ξ )ηη Vη − (ξ )ηη Vθ + (φ )ηη

− ν 2Uθ (φ2 )θU + 2Vθ (φ2 )θV − (ξ 1 )θθ Vξ − (ξ 2 )θθ Vη

3 2 2
− (ξ )θθ Vθ − 2(φ )θW (Uξ + Vη ) + (φ )θθ = 0
Collecting the terms with respect to the first order partial derivatives, and
equating to zero the various coefficients, we obtain from the above equation
the following
1 1 1 3
U − ξ [(ξ )η + (φ )U ] − W − θ [(ξ )η + (φ2 )W ]
2
2 2 (65.380)
+ 2ν[(φ2 )θW − (φ2 )ξU ] = 0
1 1
V − η [(ξ )η + (φ2 )U ] + (φ4 )U − 2ν(φ2 )ηU = 0 (65.381)
2
1
W − θ [(ξ 1 )η + (φ2 )U ] − 2ν(φ2 )θU = 0 (65.382)
2
1 3
V − η [(ξ )η + (φ2 )W ] + (φ4 )W − 2ν(φ2 )ηW = 0 (65.383)
2
1 3
U− ξ [(ξ )η + (φ2 )W ] − 2ν(φ2 )ξW = 0 (65.384)
2
840
ν[(ξ 3 )ξξ + (ξ 3 )ηη + (ξ 3 )θθ ] − 2ν(φ2 )θV − (ξ 3 )η V

+ W [(ξ 2 )η − (ξ 3 )θ + (φ2 )V − (φ4 )P ] − (ξ 3 )ξ U
1 (65.385)
+ [η(ξ 3 )η − θ(ξ 2 )η + θ(ξ 3 )θ + ξ(ξ 3 )ξ − θ(φ2 )V + θ(φ4 )P ]
2
1
− ξ 3 + φ3 = 0
2
ν[(ξ 1 )ξξ + (ξ 1 )ηη + (ξ 1 )θθ ] − 2ν(φ2 )ξV − (ξ 1 )η V

+ U [(ξ 2 )η − (ξ 1 )ξ + (φ2 )V − (φ4 )P ] − (ξ 1 )θ W
1 (65.386)
+ [η(ξ 1 )η + θ(ξ 1 )θ + ξ(ξ 1 )ξ − ξ(ξ 2 )η − ξ(φ2 )V + ξ(φ4 )P ]
2
1
− ξ 1 + φ1 = 0
2
ν[(ξ 2 )ξξ + (ξ 2 )ηη + (ξ 2 )θθ ] + 2ν[(φ2 )θW − (φ2 )ηV ] − (ξ 2 )ξ U

− W [(ξ 2 )θ + (ξ 3 )η + (φ2 )W ] + V [(φ2 )V − (φ4 )P ] + (φ4 )V
1 (65.387)
+ [θ(ξ 2 )θ + ξ(ξ 2 )ξ + θ(ξ 3 )η − η(φ2 )V + θ(φ2 )W + η(φ4 )P ]
2
1
− ξ 2 + φ2 = 0
2
1
− ν [(φ2 )ξξ + (φ2 )ηη + (φ2 )θθ ] + (φ4 )η − φ2
2

2 1 1

2 1 2 1 4
+ U (φ )ξ − (ξ )η + V (φ )η − (ξ )η + (φ )P (65.388)
2 2 2
1 1
+ W (φ2 )θ − (ξ 3 )η − [ξ(φ2 )ξ + η(φ2 )η + θ(φ2 )θ ] = 0
2 2
P r(2) Γ[∆(3) ] =0 (65.389)

∆(1) =0, ∆(2) =0, ∆(3) =0, ∆(4) =0
841
We have that (considering (65.283))
1 1 1
2 2 2
1 1
+ Wη φ2 + − + Wθ φ3 + − ξ + U χξ
2 2 (65.390)
1 1
+ − η + V χη + − θ + W χθ
2 2
+ π − νχ − νχ − νχθθ
θ ξξ ηη
Substituting χξ , χη , χθ given by (65.291)-(65.293), π θ given by (65.296) and

χξξ , χηη , χθθ given by (65.303)-(65.305) and also Pξ , Pη , Pθ , Wθ given by
(65.308)-(65.311) and finally taking into account (65.328), we obtain the
842
equation
1 1 1
− Wξ ξ 1 − Wη ξ 2 − Wθ ξ 3 + Wξ φ1
2 2 2
1
+ Wη φ + − + Wθ φ3
2
2
1
2

1 2 3
− Wξ (ξ )ξ − Wη (ξ )ξ + (Uξ + Vη ) (ξ )ξ
1
2

1 2 3
− Wξ (ξ )η − Wη (ξ )η + (Uξ + Vη ) (ξ )η
1
2

1 2 3
− Wξ (ξ )θ − Wη (ξ )θ + (Uξ + Vη ) (ξ )θ

1 1 1 1
2 2 2 2
1
+ ν(Uξξ + Uηη + Uθθ ) (ξ )θ
1 1 1 1
2 2 2 2
+ ν(Vξξ + Vηη + Vθθ ) (ξ 2 )θ
1 1 1 1
− W + ξWξ + ηWη − θ(Uξ + Vη ) − U Wξ − V Wη
2 2 2 2

3 4
+ W (Uξ + Vη ) + ν(Wξξ + Wηη + Wθθ ) [(ξ )θ − (φ )P ]
843

− ν (Uξ )2 (φ3 )U U + (Wξ )2 (φ3 )W W + (Vξ )2 (φ3 )V V
+ Uξξ (φ3 )U + Vξξ (φ3 )V − 2Wξη (ξ 2 )ξ − 2Wξθ (ξ 3 )ξ

− Wξξ [2(ξ 1 )ξ − (φ3 )W ]
+ 2Uξ [(φ3 )U V Vξ + (φ3 )U W Wξ + (φ3 )ξU ]
+ 2Vξ [(φ3 )V W Wξ + (φ3 )ξV ] − Wη (ξ 2 )ξξ

1 3 3 3
− Wξ [(ξ )ξξ − 2(φ )ξW ] + (Uξ + Vη ) (ξ )ξξ + (φ )ξξ

− ν Uηη (φ3 )U + Vηη (φ3 )V − Wηη [2(ξ 2 )η − (φ3 )W ]
− 2Wηθ (ξ 3 )η − 2Wξη (ξ 1 )η + (Uη )2 (φ3 )U U

+ (Vη )2 (φ3 )V V + (Wη )2 (φ3 )W W (65.391)
+ 2Uη [(φ3 )U V Vη + (φ3 )U W Wη + (φ3 )ηU ]
+ 2Vη [(φ3 )V W Wη + (φ3 )ηV ] − Wη [(ξ 2 )ηη − 2(φ3 )ηW ]

1 3 3
− (ξ )ηη Wξ + (ξ )ηη (Uξ + Vη ) + (φ )ηη

− ν Uθθ (φ3 )U + Vθθ (φ3 )V + (Uθ )2 (φ3 )U U + (Vθ )2 (φ3 )V V
+ (Uξ + Vη )2 (φ3 )W W − 2Wηθ (ξ 2 )θ − Wθθ [2(ξ 3 )θ − (φ3 )W ]

− 2Wξθ (ξ 1 )θ + 2Uθ [(φ3 )U V Vθ − (φ3 )U W (Uξ + Vη ) + (φ3 )θU ]
+ 2Vθ (φ3 )θV − 2(Uξ + Vη ) [(φ3 )V W Vθ + (φ3 )θW ]

1 2 3 3
− (ξ )θθ Wξ − (ξ )θθ Wη + (ξ )θθ (Uξ + Vη ) + (φ )θθ = 0
The above equation is the expanded version of the invariance condition

(65.389). Equate next to zero the coefficients of the partial derivatives of
the functions U, V and W :
The terms (Uη )2 and (Uθ )2 have common coefficient −ν(φ3 )U U while the co-
efficient of (Uξ )2 is −ν(φ3 )U U − ν(φ3 )W W . Therefore we obtain the equations
(φ3 )U U = (φ3 )W W = 0 (65.392)
The terms Uξ Vξ , Uη Vη and Uθ Vθ have common coefficient −ν(φ3 )U V . There-

fore we have the equation
(φ3 )U V = 0 (65.393)
844
The terms Uθ Uξ and Uθ Vη have common coefficient 2ν(φ3 )U W while the terms
Uη Wη and Uξ Wξ have common coefficient −2ν(φ3 )U W . Therefore we get the
equation
(φ3 )U W = 0 (65.394)
The terms (Vξ )2 and (Vθ )2 have common coefficient −ν(φ3 )V V while the co-
efficient of (Vη )2 is −ν(φ3 )V V − ν(φ3 )W W . Therefore we obtain the equation
(φ3 )V V = (φ3 )W W = 0 (65.395)
The terms Vθ Uξ and Vθ Vη have common coefficient 2ν(φ3 )V W while the terms
Vη Wη and Vξ Wξ have common coefficient −2ν(φ3 )V W . Therefore we get the
equation
(φ3 )V W = 0 (65.396)
The terms (Wξ )2 and (Wη )2 have common coefficient −ν(φ3 )W W while the
coefficient of Uξ Vη is −2ν(φ3 )W W . Therefore we obtain the equation
(φ3 )W W = 0 (65.397)
Collecting (65.392)-(65.397) we obtain
(φ3 )U U = (φ3 )U V = (φ3 )U W = 0

(65.398)
(φ3 )V V = (φ3 )V W = (φ3 )W W = 0
The terms Uξξ , Uηη and Uθθ have common coefficient −ν[(ξ 1 )θ + (φ3 )U ]. We
(ξ 1 )θ + (φ3 )U = 0 (65.399)
The terms Vξξ , Vηη and Vθθ have common coefficient −ν[(ξ 2 )θ + (φ3 )V ]. We
(ξ 2 )θ + (φ3 )V = 0 (65.400)
The coefficient of Wηη is ν[2(ξ 2 )η − (ξ 3 )θ + (φ4 )P − (φ3 )W ]. We thus have the

equation
2(ξ 2 )η − (ξ 3 )θ + (φ4 )P − (φ3 )W = 0 (65.401)
845
The coefficient of Wθθ is ν[(ξ 3 )θ + (φ4 )P − (φ3 )W ]. We thus have the equation
(ξ 3 )θ + (φ4 )P − (φ3 )W = 0 (65.402)
The coefficient of Wξξ is ν[2(ξ 1 )ξ − (ξ 3 )θ + (φ4 )P − (φ3 )W ]. We thus have the

equation
2(ξ 1 )ξ − (ξ 3 )θ + (φ4 )P − (φ3 )W = 0 (65.403)
The coefficient of Wηθ is given by 2ν[(ξ 2 )θ + (ξ 3 )η ]. Therefore we get the

equation
(ξ 2 )θ + (ξ 3 )η = 0 (65.404)
The coefficient of Wξη is given by 2ν[(ξ 1 )η + (ξ 2 )ξ ]. Therefore we get the

equation
(ξ 1 )η + (ξ 2 )ξ = 0 (65.405)
The coefficient of Wξθ is given by 2ν[(ξ 1 )θ + (ξ 3 )ξ ]. Therefore we get the

equation
(ξ 1 )θ + (ξ 3 )ξ = 0 (65.406)
846
Taking into account (65.398)-(65.406), equation (65.391) gets simplified to
1 1 1
− Wξ ξ 1 − Wη ξ 2 − Wθ ξ 3 + Wξ φ1
2 2 2
1
+ Wη φ2 + − + Wθ φ3
2
1
2

1 2 3
− Wξ (ξ )ξ − Wη (ξ )ξ + (Uξ + Vη ) (ξ )ξ
1
2

1 2 3
− Wξ (ξ )η − Wη (ξ )η + (Uξ + Vη ) (ξ )η
1
2

1 2 3
− Wξ (ξ )θ − Wη (ξ )θ + (Uξ + Vη ) (ξ )θ

1 1 1 1
− U + ξUξ + ηUη + θUθ − U Uξ − V Uη
2 2 2 2
1
− W Uθ (ξ )θ
1 1 1 1
− V + ξVξ + ηVη + θVθ − U Vξ − V Vη
2 2 2 2
− W Vθ (ξ 2 )θ
1 1 1 1
2 2 2 2

3 4
− V Wη + W (Uξ + Vη ) [(ξ )θ − (φ )P ]
847

− ν 2Uξ (φ3 )ξU + 2Vξ (φ3 )ξV − Wη (ξ 2 )ξξ

3 3 3
+ 2Wξ (φ )ξW + (Uξ + Vη ) (ξ )ξξ + (φ )ξξ

− ν 2Uη [(φ3 )U V Vη + (φ3 )U W Wη + (φ3 )ηU ]
+ 2Vη (φ3 )ηV − Wη [(ξ 2 )ηη − 2(φ3 )ηW ]

(65.407)
1 3 3
− (ξ )ηη Wξ + (ξ )ηη (Uξ + Vη ) + (φ )ηη

− ν 2Uθ [(φ3 )U V Vθ − (φ3 )U W (Uξ + Vη ) + (φ3 )θU ]
+ 2Vθ (φ3 )θV − 2(Uξ + Vη ) (φ3 )θW

1 2 3 3
− (ξ )θθ Wξ − (ξ )θθ Wη + (ξ )θθ (Uξ + Vη ) + (φ )θθ = 0
Collecting the terms with respect to the first order partial derivatives, and
equating to zero the various coefficients, we obtain from the above equation
the following :
1 1
V − η [(ξ )θ + (φ3 )U ] − 2ν(φ3 )ηU = 0 (65.408)
2
1
W − θ [(ξ 2 )θ + (φ3 )V ] + (φ4 )V − 2ν(φ3 )θV = 0 (65.409)
2
1 1
W − θ [(ξ )θ + (φ3 )U ] + (φ4 )U − 2ν(φ3 )θU = 0 (65.410)
2
1
U − ξ [(ξ 2 )θ + (φ3 )V ] − 2ν(φ3 )ξV = 0 (65.411)
2
− ν[(ξ 3 )ξξ + (ξ 3 )ηη + (ξ 3 )θθ ] + 2ν[(φ3 )θW − (φ3 )ηV ]
1 1
+ V − η [(ξ 2 )θ + (ξ 3 )η + (φ3 )V ] + U − ξ (ξ 3 )ξ (65.412)
2 2
1 4 1
+ W − θ [(φ )P − (φ3 )W ] − (φ4 )W + ξ 3 − φ3 = 0
2 2
848
ν[(ξ 2 )ξξ + (ξ 2 )ηη + (ξ 2 )θθ ] − 2ν(φ3 )ηW
1 3
+ V − η [(ξ )θ − (ξ 2 )η + (φ3 )W − (φ4 )P ]
2
1 2 1 2 (65.413)
− U − ξ (ξ )ξ − W − θ (ξ )θ
2 2
1 2
− ξ + φ2 = 0
2
− ν[(ξ 3 )ξξ + (ξ 3 )ηη + (ξ 3 )θθ ] + 2ν[(φ3 )θW − (φ3 )ξU ]
1 1
+ U − ξ [(ξ 1 )θ + (ξ 3 )ξ + (φ3 )U ] + V − η (ξ 3 )η (65.414)
2 2
1 4 1
+ W − θ [(φ )P − (φ3 )W ] − (φ4 )W + ξ 3 − φ3 = 0
2 2
ν[(ξ 1 )ξξ + (ξ 1 )ηη + (ξ 1 )θθ ] − 2ν(φ3 )ξW
1
+ U − ξ [(ξ 3 )θ − (ξ 1 )ξ + (φ3 )W + (φ4 )P ] (65.415)
2
1 1 1
− V − η (ξ 1 )η − W − θ (ξ 1 )θ − ξ 1 + φ1 = 0
2 2 2
1
− ν [(φ3 )ξξ + (φ3 )ηη + (φ3 )θθ ] + (φ4 )θ − φ3
2
1 1
+ U (φ3 )ξ − (ξ 1 )θ + V (φ3 )η − (ξ 2 )θ
2 2 (65.416)
1 1
+ W (φ3 )θ − (ξ 3 )θ + (φ4 )P
2 2
1
− [ξ(φ3 )ξ + η(φ3 )η + θ(φ3 )θ ] = 0
2
We are now in a position to start solving the determining equations.
Considering the equations (65.342), (65.370) and (65.398), we see that the
functions φ1 , φ2 and φ3 are linear functions of U, V and W and thus admit
the representation
φ1 = X1 (ξ, η, θ)U + Y1 (ξ, η, θ)V + Z1 (ξ, η, θ)W + F1 (ξ, η, θ) (65.417)
849
and
respectively.
Using the expression (65.417) for φ1 , equation (65.360) yields

1 1
− ν[(X1 )ξξ + (X1 )ηη + (X1 )θθ ] − X1 + (F1 )ξ − (ξ 1 )ξ
2 2

1 1 1 1
+ (φ4 )P − ξ(X1 )ξ − η(X1 )η − θ(X1 )θ U
2 2 2 2

1 1
+ − ν[(Y1 )ξξ + (Y1 )ηη + (Y1 )θθ ] − Y1 + (F1 )η − (ξ 2 )ξ
2 2

1 1 1
− ξ(Y1 )ξ − η(Y1 )η − θ(Y1 )θ V
2 2 2
(65.420)
1 1 3
+ − ν[(Z1 )ξξ + (Z1 )ηη + (Z1 )θθ ] − Z1 + (F1 )θ − (ξ )ξ
2 2

1 1 1
− ξ(Z1 )ξ − η(Z1 )η − θ(Z1 )θ W
2 2 2
+ [(Y1 )ξ + (X1 )η ]U V + [(Z1 )ξ + (X1 )θ ]U W + [(Z1 )η + (Y1 )θ ]V W
+ (X1 )ξ U 2 + (Y1 )η V 2 + (Z1 )θ W 2 − ν[(F1 )ξξ + (F1 )ηη + (F1 )θθ ]
1 1 1 1
+ (φ4 )ξ − F1 − ξ(F1 )ξ − η(F1 )η − θ(F1 )θ = 0
2 2 2 2
Equating to zero all the coefficients of the functions U, V and W , we obtain
from the above equation the system of equations
1 1
− ν[(X1 )ξξ + (X1 )ηη + (X1 )θθ ] − X1 + (F1 )ξ − (ξ 1 )ξ
2 2 (65.421)
1 4 1 1 1
+ (φ )P − ξ(X1 )ξ − η(X1 )η − θ(X1 )θ = 0
2 2 2 2
1 1
− ν[(Y1 )ξξ + (Y1 )ηη + (Y1 )θθ ] − Y1 + (F1 )η − (ξ 2 )ξ
2 2 (65.422)
1 1 1
− ξ(Y1 )ξ − η(Y1 )η − θ(Y1 )θ = 0
2 2 2
850
1 1
− ν[(Z1 )ξξ + (Z1 )ηη + (Z1 )θθ ] − Z1 + (F1 )θ − (ξ 3 )ξ
2 2 (65.423)
1 1 1
− ξ(Z1 )ξ − η(Z1 )η − θ(Z1 )θ = 0
2 2 2
(Y1 )ξ + (X1 )η = 0, (Z1 )ξ + (X1 )θ = 0, (Z1 )η + (Y1 )θ = 0 (65.424)
(X1 )ξ = 0, (Y1 )η = 0, (Z1 )θ = 0 (65.425)
− ν[(F1 )ξξ + (F1 )ηη + (F1 )θθ ]
1 1 1 1 (65.426)
+ (φ4 )ξ − F1 − ξ(F1 )ξ − η(F1 )η − θ(F1 )θ = 0
2 2 2 2
The system (65.425) implies that X1 , Y1 and Z1 are independent of ξ, η and
θ respectively and thus the function φ1 admits the representation
φ1 = X1 (η, θ)U + Y1 (ξ, θ)V + Z1 (ξ, η)W + F1 (ξ, η, θ) (65.427)

1 1
− ν[(X2 )ξξ + (X2 )ηη + (X2 )θθ ] − X2 + (F2 )ξ − (ξ 1 )η
2 2

1 1 1
− ξ(X2 )ξ − η(X2 )η − θ(X2 )θ U
2 2 2

1 1
+ − ν[(Y2 )ξξ + (Y2 )ηη + (Y2 )θθ ] − Y2 + (F2 )η − (ξ 2 )η
2 2

1 1 1 1
+ (φ4 )P − ξ(Y2 )ξ − η(Y2 )η − θ(Y2 )θ V
2 2 2 2
(65.428)
1 1 3
+ − ν[(Z2 )ξξ + (Z2 )ηη + (Z2 )θθ ] − Z2 + (F2 )θ − (ξ )η
2 2

1 1 1
− ξ(Z2 )ξ − η(Z2 )η − θ(Z2 )θ W
2 2 2
1 1 1 1
+ (φ4 )η − F2 − ξ(F2 )ξ − η(F2 )η − θ(F2 )θ = 0
2 2 2 2
851
1 1
− ν[(X2 )ξξ + (X2 )ηη + (X2 )θθ ] − X2 + (F2 )ξ − (ξ 1 )η
2 2 (65.429)
1 1 1
− ξ(X2 )ξ − η(X2 )η − θ(X2 )θ = 0
2 2 2
1 1
− ν[(Y2 )ξξ + (Y2 )ηη + (Y2 )θθ ] − Y2 + (F2 )η − (ξ 2 )η
2 2 (65.430)
1 4 1 1 1
+ (φ )P − ξ(Y2 )ξ − η(Y2 )η − θ(Y2 )θ = 0
2 2 2 2
1 1
− ν[(Z2 )ξξ + (Z2 )ηη + (Z2 )θθ ] − Z2 + (F2 )θ − (ξ 3 )η
2 2 (65.431)
1 1 1
− ξ(Z2 )ξ − η(Z2 )η − θ(Z2 )θ = 0
2 2 2
(Y2 )ξ + (X2 )η = 0, (Z2 )ξ + (X2 )θ = 0, (Z2 )η + (Y2 )θ = 0 (65.432)
(X2 )ξ = 0, (Y2 )η = 0, (Z2 )θ = 0 (65.433)
− ν[(F2 )ξξ + (F2 )ηη + (F2 )θθ ]
1 1 1 1 (65.434)
+ (φ4 )η − F2 − ξ(F2 )ξ − η(F2 )η − θ(F2 )θ = 0
2 2 2 2
852

1 1
− ν[(X3 )ξξ + (X3 )ηη + (X3 )θθ ] − X3 + (F3 )ξ − (ξ 1 )θ
2 2

1 1 1
− ξ(X3 )ξ − η(X3 )η − θ(X3 )θ U
2 2 2

1 1
+ − ν[(Y3 )ξξ + (Y3 )ηη + (Y3 )θθ ] − Y3 + (F3 )η − (ξ 2 )θ
2 2

1 1 1
− ξ(Y3 )ξ − η(Y3 )η − θ(Y3 )θ V
2 2 2
(65.436)
1 1 3
+ − ν[(Z3 )ξξ + (Z3 )ηη + (Z3 )θθ ] − Z3 + (F3 )θ − (ξ )θ
2 2

1 1 1 1
+ (φ4 )P − ξ(Z3 )ξ − η(Z3 )η − θ(Z3 )θ W
2 2 2 2
1 1 1 1
+ (φ4 )θ − F3 − ξ(F3 )ξ − η(F3 )η − θ(F3 )θ = 0
2 2 2 2
1 1
− ν[(X3 )ξξ + (X3 )ηη + (X3 )θθ ] − X3 + (F3 )ξ − (ξ 1 )θ
2 2 (65.437)
1 1 1
− ξ(X3 )ξ − η(X3 )η − θ(X3 )θ = 0
2 2 2
1 1
− ν[(Y3 )ξξ + (Y3 )ηη + (Y3 )θθ ] − Y3 + (F3 )η − (ξ 2 )θ
2 2 (65.438)
1 1 1
− ξ(Y3 )ξ − η(Y3 )η − θ(Y3 )θ = 0
2 2 2
1 1
− ν[(Z3 )ξξ + (Z3 )ηη + (Z3 )θθ ] − Z3 + (F3 )θ − (ξ 3 )θ
2 2 (65.439)
1 4 1 1 1
+ (φ )P − ξ(Z3 )ξ − η(Z3 )η − θ(Z3 )θ = 0
2 2 2 2
(Y3 )ξ + (X3 )η = 0, (Z3 )ξ + (X3 )θ = 0, (Z3 )η + (Y3 )θ = 0 (65.440)
(X3 )ξ = 0, (Y3 )η = 0, (Z3 )θ = 0 (65.441)
853
− ν[(F3 )ξξ + (F3 )ηη + (F3 )θθ ]
1 1 1 1 (65.442)
+ (φ4 )θ − F3 − ξ(F3 )ξ − η(F3 )η − θ(F3 )θ = 0
2 2 2 2
Taking into account the new expressions for φ1 , φ2 and φ3 , equation (65.338)
becomes
[(X2 )η + (X3 )θ ]U + [(Y1 )ξ + (Y3 )θ ]V + [(Z1 )ξ + (Z2 )η ]W
(65.444)
+ (F1 )ξ + (F2 )η + (F3 )θ = 0
Equating to zero the coefficients of the functions U, V and W , we obtain the
system of equations
(X2 )η + (X3 )θ = 0, (Y1 )ξ + (Y3 )θ = 0, (Z1 )ξ + (Z2 )η = 0 (65.445)
(F1 )ξ + (F2 )η + (F3 )θ = 0 (65.446)
Special care must be taken in considering the functional dependence of X2 , X3 ,
Y1 , Y3 , Z1 , Z2 and F1 , F2 , F3 on the independent variables ξ, η, θ by looking
at the expressions for φ1 , φ2 and φ3 given by (65.427), (65.435) and (65.443)
respectively.
Let us next move on to determining the general representation of the func-
tions ξ 1 , ξ 2 and ξ 3 . Using the expression (65.427) for φ1 , we obtain from
equations (65.331) and (65.332) the equations
Y1 (ξ, θ) = (ξ 1 )η and Z1 (ξ, η) = (ξ 1 )θ (65.447)
respectively. Differentiation of the first equation with respect to η and the
second with respect to θ, we obtain the equations (ξ 1 )ηη = 0 and (ξ 1 )θθ = 0
respectively. This means that ξ 1 is a linear function of η and θ. On the other
hand, since ξ 1 does not depend on U, V, W and P due to (65.328), it will
admit the representation
ξ 1 = α1 (ξ) + β1 (ξ)η + γ1 (ξ)θ (65.448)
Using the expression (65.435) for φ2 , we obtain from equations (65.333) and
(65.336) the equations
X2 (η, θ) = (ξ 2 )ξ and Z2 (ξ, η) = (ξ 2 )θ (65.449)
854
respectively. Differentiation of the first equation with respect to ξ and the
second with respect to θ, we obtain the equations (ξ 2 )ξξ = 0 and (ξ 2 )θθ = 0
respectively. This means that ξ 2 is a linear function of ξ and θ. On the other
ξ 2 = α2 (η) + β2 (η)ξ + γ2 (η)θ (65.450)
Using the expression (65.443) for φ3 , we obtain from equations (65.334) and
(65.337) the equations
X3 (η, θ) = (ξ 3 )ξ and Y3 (ξ, θ) = (ξ 3 )η (65.451)
respectively. Differentiation of the first equation with respect to ξ and the

second with respect to η, we obtain the equations (ξ 3 )ξξ = 0 and (ξ 3 )ηη = 0
respectively. This means that ξ 3 is a linear function of ξ and η. On the other
ξ 3 = α3 (θ) + β3 (θ)ξ + γ3 (θ)η (65.452)
Using the expression (65.427) for φ1 and (65.450) for ξ 2 , we obtain from
(65.349) that Y1 (ξ, θ) + β2 (η) = 0. This equation implies that each function
should be a constant: Y1 (ξ, θ) = c1 and β2 (η) = −c1 .
(65.350) that Z1 (ξ, η) + β3 (θ) = 0. This equation implies that each function
should be a constant: Z1 (ξ, η) = c2 and β3 (θ) = −c2 .
(65.371) that X2 (η, θ) + β1 (ξ) = 0. This equation implies that each function
should be a constant: X2 (η, θ) = c3 and β1 (ξ) = −c3 .
(65.375) that Z2 (ξ, η) + γ3 (θ) = 0. This equation implies that each function
should be a constant: Z2 (ξ, η) = c4 and γ3 (θ) = −c4 .
(65.399) that X3 (η, θ) + γ1 (ξ) = 0. This equation implies that each function
should be a constant: X3 (η, θ) = c5 and γ1 (ξ) = −c5 .
(65.400) that Y3 (ξ, θ) + γ2 (η) = 0. This equation implies that each function
should be a constant: Y3 (ξ, θ) = c6 and γ2 (η) = −c6 .
855
Using the above found values of the various functions, we get the following
expressions for φ1 , φ2 , φ3 and ξ 1 , ξ 2 , ξ 3 :
φ1 = X1 (η, θ)U + c1 V + c2 W + F1 (ξ, η, θ)

φ2 = c3 U + Y2 (ξ, θ)V + c4 W + F2 (ξ, η, θ) (65.453)
φ3 = c5 U + c6 V + Z3 (ξ, η)W + F3 (ξ, η, θ)
and
ξ 1 = α1 (ξ) − c3 η − c5 θ
ξ 2 = α2 (η) − c1 ξ − c6 θ (65.454)
ξ 3 = α3 (θ) − c2 ξ − c4 η
Equations (65.343), (65.376) and (65.404) are identical and in view of (65.454),
they give −c6 − c4 = 0, i.e. c6 = −c4 . Equations (65.344), (65.377) and
(65.405) are identical and in view of (65.454), they give −c3 − c1 = 0, i.e.
c3 = −c1 . Equations (65.345), (65.378) and (65.406) are identical and in
view of (65.454), they give −c5 − c2 = 0, i.e. c5 = −c2 .
Therefore (65.453) and (65.454) take the new form
φ1 = X1 (η, θ)U + c1 V + c2 W + F1 (ξ, η, θ)

φ2 = −c1 U + Y2 (ξ, θ)V + c4 W + F2 (ξ, η, θ) (65.455)
φ3 = −c2 U − c4 V + Z3 (ξ, η)W + F3 (ξ, η, θ)
and
ξ 1 = α1 (ξ) + c1 η + c2 θ
ξ 2 = α2 (η) − c1 ξ + c4 θ (65.456)
ξ 3 = α3 (θ) − c2 ξ − c4 η
respectively.
Considering the system (65.424) and taking into account Y1 = c1 , the first
equation (Y1 )ξ +(X1 )η = 0 gives (X1 )η = 0 and thus X1 = X1 (θ). Taking into
account that Z1 = c2 , the second equation (Z1 )ξ + (X1 )θ = 0 gives X10 (θ) = 0
and thus X1 = k1 . The third equation (Z1 )η + (Y1 )θ = 0 of the system is an
identity in view of Z1 = c2 and Y1 = c1 .
Considering the system (65.432), and taking into account X2 = −c1 , the first
equation (Y2 )ξ + (X2 )η = 0 gives (Y2 )ξ = 0 and thus Y2 = Y2 (θ). The second
equation (Z2 )ξ + (X2 )θ = 0 is an identity, since Z2 = c4 and X2 = −c1 . The
856
third equation (Z2 )η + (Y2 )θ = 0 since Z2 = c4 , gives Y20 (θ) = 0 and thus
Y2 = k2 .
Considering the system (65.440), and taking into account Y3 = −c4 and
X3 = −c2 , the first equation (Y3 )ξ + (X3 )η = 0 becomes an identity. The
second equation (Z3 )ξ + (X3 )θ = 0, since X3 = −c2 , yields (Z3 )ξ = 0 and
thus Z3 = Z3 (η). The third equation (Z3 )η + (Y3 )θ = 0 in view of Y3 = −c4
gives (Z3 )η = 0, i.e. Z30 (η) = 0 and thus Z3 = k3 .
Taking into account
X1 = k1 , Y2 = k2 , Z3 = k3 (65.457)
we get from (65.455) the new expressions for φ1 , φ2 , φ3 :
φ1 = k1 U + c1 V + c2 W + F1 (ξ, η, θ)
φ2 = −c1 U + k2 V + c4 W + F2 (ξ, η, θ) (65.458)
φ3 = −c2 U − c4 V + k3 W + F3 (ξ, η, θ)
The system of equations (65.445) becomes an identity.

Taking into account (65.456) and (65.458) we obtain from equations (65.355),
(65.358), (65.381), (65.383), (65.409) and (65.410) that (φ4 )V = 0, (φ4 )W = 0,
(φ4 )U = 0, (φ4 )W = 0, (φ4 )V = 0 and (φ4 )U = 0, while equations (65.356),
(65.357), (65.359), (65.380), (65.382), (65.384), (65.408) and (65.411) are
identities. We thus have derived
(φ4 )U = (φ4 )V = (φ4 )W = 0 (65.459)
Equations (65.422) and (65.423) yield (F1 )η = 0 and (F1 )θ = 0 respectively.

Therefore
F1 = F1 (ξ) (65.460)
Equations (65.421) and (65.426) are equivalent to

1 1 1
(F1 )ξ − α10 (ξ) + (φ4 )P = k1 (65.461)
2 2 2
and
1
ν(F1 )ξξ + (ξF1 )ξ = (φ4 )ξ (65.462)
2
857
respectively. Integrating (65.462) we obtain
1
ν(F1 )ξ + ξF1 = φ4 + n1 (η, θ) (65.463)
2
Equations (65.429) and (65.431) yield (F2 )ξ = 0 and (F2 )θ = 0 respectively.
Therefore
F2 = F2 (η) (65.464)

1 1 1
(F2 )η − α20 (η) + (φ4 )P = k2 (65.465)
2 2 2
and
1
ν(F2 )ηη + (ηF2 )η = (φ4 )η (65.466)
2
1
ν(F2 )η + ηF2 = φ4 + n2 (ξ, θ) (65.467)
2
Equations (65.437) and (65.438) yield (F3 )ξ = 0 and (F3 )η = 0 respectively.
Therefore
F3 = F3 (θ) (65.468)

1 1 1
(F3 )θ − α30 (θ) + (φ4 )P = k3 (65.469)
2 2 2
and
1
ν(F3 )θθ + (θF3 )θ = (φ4 )θ (65.470)
2
1
ν(F3 )θ + θF3 = φ4 + n3 (ξ, η) (65.471)
2
858
Using any of the (65.463), (65.467) or (65.471) we see that φ4 is independent
of P , i.e.
(φ4 )P = 0 (65.472)
Equations (65.346), (65.347) and (65.348) are equivalent to
α10 (ξ) + (φ4 )P = k1 (65.473)
−α10 (ξ) + (φ4 )P = k1 − 2α20 (η) (65.474)

and
−α10 (ξ) + (φ4 )P = k1 − 2α30 (θ) (65.475)
respectively.
α20 (η) + (φ4 )P = k2 (65.476)
−α20 (η) + (φ4 )P = k2 − 2α30 (θ) (65.477)

and
−α20 (η) + (φ4 )P = k2 − 2α10 (ξ) (65.478)
respectively.
α30 (θ) + (φ4 )P = k3 (65.479)
−α30 (θ) + (φ4 )P = k3 − 2α20 (η) (65.480)

and
−α30 (θ) + (φ4 )P = k3 − 2α10 (ξ) (65.481)
respectively.
α10 (ξ) − α30 (θ) = k1 − k3 (65.482)
and
α20 (η) − α30 (θ) = k2 − k3 (65.483)
859
respectively.
Using (φ4 )P = 0, we obtain from (65.473), (65.476) and (65.479) that α10 (ξ) =
k1 , α20 (η) = k2 and α30 (θ) = k3 and by integration
α1 (ξ) = k1 ξ + q1 , α2 (η) = k2 η + q2 , α3 (θ) = k3 θ + q3 (65.484)
On the other hand, combining (65.461) and (65.469), (65.464) and (65.472),
(65.467) and (65.475), we obtain the equations F10 (ξ) = α10 (ξ), F20 (η) = α20 (η)
and F30 (θ) = α30 (θ) respectively and by integration F1 (ξ) = α1 (ξ) + m1 ,
F2 (η) = α2 (η) + m2 and F3 (θ) = α3 (θ) + m3 . Using further the expressions
(65.484) we arrive at
F1 (ξ) = k1 ξ + (q1 + m1 ), F2 (η) = k2 η + (q2 + m2 ),
(65.485)
F3 (θ) = k3 θ + (q3 + m3 )
Considering (65.456) and (65.458) and using (65.484) and (65.485), we arrive
at the new expressions for ξ 1 , ξ 2 , ξ 3 and φ1 , φ2 , φ3 :
ξ 1 = k1 ξ + c1 η + c2 θ + q1
ξ 2 = −c1 ξ + k2 η + c4 θ + q2 (65.486)
ξ 3 = −c2 ξ − c4 η + k3 θ + q3
and
φ1 = k1 U + c1 V + c2 W + k1 ξ + (q1 + m1 )
φ2 = −c1 U + k2 V + c4 W + k2 η + (q2 + m2 ) (65.487)
φ3 = −c2 U − c4 V + k3 W + k3 θ + (q3 + m3 )
Solving the system of equations (65.462), (65.466) and (65.470), using the
values (65.485), i.e.
1 1
(φ4 )ξ = k1 ξ + (q1 + m1 ), (φ4 )η = k2 η + (q2 + m2 )
2 2 (65.488)
1
(φ4 )θ = k3 θ + (q3 + m3 )
2
we determine the function φ4 :
1 1 1 1
φ4 = k1 ξ 2 + (q1 + m1 )ξ + k2 η 2 + (q2 + m2 )η
2 2 2 2 (65.489)
1 2 1
+ k3 θ + (q3 + m3 )θ + c
2 2
860
Using again the values (65.484) and taking into account (φ4 )P = 0, we obtain
from equations (65.474), (65.475), (65.477), (65.478), (65.480) and (65.481)
that k1 = k2 = k3 . Let
k = k1 = k2 = k3 (65.490)
Let us proceed further since a number of equations have been left with-
out considering their solutions. We see that the main task of determining
ξ 1 , ξ 2 , ξ 3 and φ1 , φ2 , φ3 , φ4 has been completed. However we have to check
the consistency of the remaining equations to the values of the coefficients
we have already found.
Equation (65.352) is equivalent to (taking into account (65.486), (65.487),
(65.489) and (65.490))
1
2kU + q1 + m1 = 0 (65.491)
2
Consistency requires
1
k = 0, m1 = − q1 (65.492)
2
1
2kV + q2 + m2 = 0 (65.493)
2
1
k = 0, m2 = − q2 (65.494)
2
1
2kW + q3 + m 3 = 0 (65.495)
2
1
k = 0, m3 = − q3 (65.496)
2
1
2kW + q3 + m 3 = 0 (65.497)
2
861
1
k = 0, m3 = − q3 (65.498)
2
1
2kU + q1 + m1 = 0 (65.499)
2
1
k = 0, m1 = − q1 (65.500)
2
1
2kV + q2 + m2 = 0 (65.501)
2
1
k = 0, m2 = − q2 (65.502)
2
1
2kW + q3 + m 3 = 0 (65.503)
2
1
k = 0, m3 = − q3 (65.504)
2
1
2kV + q2 + m2 = 0 (65.505)
2
1
k = 0, m2 = − q2 (65.506)
2
862
1
2kW + q3 + m 3 = 0 (65.507)
2
Consistency requires that
1
k = 0, m3 = − q3 (65.508)
2
1
2kU + q3 + m3 = 0 (65.509)
2
1
k = 0, m1 = − q1 (65.510)
2
We thus have established that
1 1 1
k = 0, m1 = − q1 , m2 = − q2 , m3 = − q3 (65.511)
2 2 2
Therefore we obtain from (65.486), (65.487) and (65.489) the following values
of the various infinitesimals
ξ 1 = c1 η + c2 θ + q 1
ξ 2 = −c1 ξ + c4 θ + q2 (65.512)
ξ 3 = −c2 ξ − c4 η + q3
1
φ1 = c1 V + c2 W + q1
2
1
φ2 = −c1 U + c4 W + k2 η + q2 (65.513)
2
1
φ3 = −c2 U − c4 V + q3
2
and
1 1 1
φ4 = q1 ξ + q2 η + q3 θ + c (65.514)
4 4 4
863
Letting for convenience
1
q 1 → a1 , q2 → a2 , q3 → a3 , c → a4
4 (65.515)
c 1 → a5 , c2 → a6 , c4 → a7
we obtain
ξ 1 = a5 η + a6 θ + a1 , ξ 2 = −a5 ξ + a7 θ + a2 ,
(65.516)
ξ 3 = −a6 ξ − a7 η + a3
1 1
φ1 = a5 V + a6 W + a1 , φ2 = −a5 U + a7 W + a2 ,
2 2 (65.517)
1
φ3 = −a6 U − a7 V + a3
2
and
1 1 1 1
φ4 = a1 ξ + a2 η + a3 θ + a4 (65.518)
4 4 4 4
The generator of the symmetries is then given by (see (65.273))
∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ3
∂ξ ∂η ∂θ
(65.519)
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂U ∂V ∂W ∂P
and taking into account (65.516)-(65.518),
∂ ∂
Γ = (a5 η + a6 θ + a1 ) + (−a5 ξ + a7 θ + a2 )
∂ξ ∂η
∂ 1 ∂
+ (−a6 ξ − a7 η + a3 ) + a5 V + a6 W + a1
∂θ 2 ∂U (65.520)
1 ∂ 1 ∂
+ −a5 U + a7 W + a2 + −a6 U − a7 V + a3
2 ∂V 2 ∂W
1 ∂
+ (a1 ξ + a2 η + a3 θ + a4 )
4 ∂P
We shall find some solutions of the system of equations (65.265)-(65.268)
based on the generator (65.520) when a5 = a6 = a7 = 0.
864
ξ 1 Uξ + ξ 2 Uη + ξ 3 Uθ = φ1
ξ 1 Vξ + ξ 2 Vη + ξ 3 Vθ = φ2
(65.521)
ξ 1 Wξ + ξ 2 Wη + ξ 3 Wθ = φ3
ξ 1 Pξ + ξ 2 Pη + ξ 3 Pθ = φ4
i.e. by solving the system (remember a5 = a6 = a7 = 0)

1
a1 Uξ + a2 Uη + a3 Uθ = a1
2
1
a1 Vξ + a2 Vη + a3 Vθ = a2
2 (65.522)
1
a1 Wξ + a2 Wη + a3 Wθ = a3
2
1
a1 Pξ + a2 Pη + a3 Pθ = (a1 ξ + a2 η + a3 θ + a4 )
4
The solution of the system of equations (65.522) is given by
1 a η − a ξ a θ − a ξ
1 2 1 3
U = ξ+F , (65.523)
2 a1 a1
1 a2 a η − a ξ a θ − a ξ
1 2 1 3
V = ξ+G , (65.524)
2 a1 a1 a1
1 a3 a1 η − a2 ξ a1 θ − a3 ξ
W = ξ+H , (65.525)
2 a1 a1 a1
a2 − a22 − a23 2 1 a2 a3 a4
P = 1 ξ + η + θ + ξ
8a21 4 a1 a1 a1
a η − a ξ a θ − a ξ (65.526)
1 2 1 3
+K ,
a1 a1
where F, G, H and K are arbitrary functions.
a2 a3 a4
κ= , λ= , µ= (65.527)
a1 a1 a1
865
and
a1 η − a2 ξ
ζ≡ = η − κξ
a1
(65.528)
a1 θ − a3 ξ
χ≡ = θ − λξ
a1
Equations (65.523)-(65.526) take now the form
1
U = ξ + F (ζ, χ) (65.529)
2
1
V = κξ + G(ζ, χ) (65.530)
2
1
W = λξ + H(ζ, χ) (65.531)
2
and
1 1
P = (1 − κ2 − λ2 )ξ 2 + (κη + λθ + µ)ξ + K(ζ, χ) (65.532)
8 4
respectively.
Our next task is to determine the functions F, G, H and K. Substituting the
expressions (65.529)-(65.532) to the system of equations (65.265)-(65.268)
we find a system of equations with unknown functions F, G, H and K. This
system consists of the equations
1
− (ζFζ + χFχ ) − (κFζ + λFχ )F + GFζ + HFχ
2
1
+ (κζ + λχ + µ) − (κKζ + λKχ ) (65.533)
4
2 2
− ν (1 + κ )Fζζ + 2κλFζχ + (1 + λ )Fχχ = 0
1 1 1
− G + κF − (ζGζ + χGχ ) + GGζ + HGχ
2 2 2
− (κGζ + λGχ )F + Kζ (65.534)

2 2
− ν (1 + κ )Gζζ + 2κλGζχ + (1 + λ )Gχχ = 0
1 1 1
− H + λF − (ζHζ + χHχ ) + GHζ + HHχ
2 2 2
− (κHζ + λHχ )F + Kχ (65.535)

− ν (1 + κ2 )Hζζ + 2κλHζχ + (1 + λ2 )Hχχ = 0
866
and
1
− (κFζ + λFχ ) + Gζ + Hχ = 0 (65.536)
2
These equations have been written in such a form as to allow us to perform
some obvious manipulations leading to a solution.
In fact looking at the fourth equation, equation (65.536), we can make the
choice
κ=λ (65.537)
i.e. a2 = a3 and then introduce a new variable σ by
σ =ζ +χ (65.538)
Then every partial derivative of the fuctions F, G, H and K will be converted

into an ordinary derivative with respect to the new variable σ. In this respect,
1
Gσ + Hσ = 2κFσ − (65.539)
2
We are thus led to introduce another function Φ by
Φ=G+H (65.540)
because in this case equation (65.539) can be written as

1
Φσ = 2κFσ − (65.541)
2
1
Φ = 2κF − σ + A (65.542)
2
Taking into account κ = λ, σ = ζ + χ and Φ = G + H, equations (65.533)-
(65.535) take the form
1 1 1
− σFσ − 2κF Fσ + ΦFσ + κσ + µ − 2κKσ
2 4 4 (65.543)
2
− 2ν(2κ + 1)Fσσ = 0
867
1 1 1
− G + κF − σGσ + ΦGσ − 2κF Gσ + Kσ
2 2 2 (65.544)
− 2ν(2κ2 + 1)Gσσ = 0
and
1 1 1
− H + κF − σHσ + ΦHσ − 2κF Hσ + Kσ
2 2 2 (65.545)
2
− 2ν(2κ + 1)Hσσ = 0
respectively.
Our aim is to find an ODE which will contain only the unknown function Φ.
This ODE can be found by combining equations (65.542)-(65.545).
In fact substituting Φ by its expression given by (65.542) into (65.543) we
obtain the equation
2ν(2κ2 + 1) A−σ 1 1µ
2Kσ = − Fσσ + Fσ + σ + (65.546)
κ κ 4 4κ
The above expression can be written in terms of Φ using (65.542), i.e. using
1 1 1
Fσσ = Φσσ and Fσ = Φσ +
2κ 2κ 2
in the form
ν(2κ2 + 1) A−σ
2Kσ = − 2
Φσσ + Φσ
κ 2κ2 (65.547)
κ2 − 1 A + κµ
+ 2
σ+
4κ 4κ2
1
2Kσ = 2ν(2κ2 + 1)Φσσ + σΦσ − ΦΦσ
2 (65.548)
1
+ 2κF Φσ + Φ − κF
2
making use the definition of the function Φ.
The right hand side of the above equation can be expressed in terms of the
function Φ, using the expression (65.542):
1 1
2Kσ = 2ν(2κ2 + 1)Φσσ + (σ − A)Φσ − σ + A (65.549)
4 2
868
We thus have derived two different expressions of Kσ given by (65.547) and
(65.549) respectively. Equating these two expressions, we obtain the ordinary
ν(2κ2 + 1)2 2κ2 + 1 2κ2 − 1

Φσσ + (σ − A)Φσ − σ
κ2 2κ2 4κ2 (65.550)
(2κ2 − 1)A − κµ
+ =0
4κ2
The general solution of the above equation can be expressed in terms of
hypergeometric functions (the conluent hypergeometric function and a gen-
eralized hypergeometric function)
κµ
Φ(σ) = C + Z(ν, κ; σ) − 2
(σ − A)2
8ν(1 + 2κ )
(65.551)
3 (σ − A)2
× 2 F2 1, 1; , 2;
2 4ν(1 + 2κ2 )
where Z(ν, κ; σ) is a function expressed in terms of the confluent hypergeo-

metric function given explicitly by
2κ2 − 1
Z(ν, κ; σ) = σ
2
2(1 + 2κ2 ) A2
+B √ exp (σ − A)
π 4ν(1 + 2κ2 )
1 3 (σ − A)2
× 1 F1 ; ; − (65.552)
2 2 4ν(1 + 2κ2 )
κµ 2
1 3 (σ − A)2
+ (σ − A) 1 F1 ; ; −
4ν(1 + 2κ2 ) 2 2 4ν(1 + 2κ2 )
1 3 (σ − A)2
× 1 F1 ; ;
2 2 4ν(1 + 2κ2 )
and A, B, C are constants.

We are now going to express all the other functions in terms of the Φ function.
In fact we obtain from (65.542) the function F :
1 1
F = Φ+ σ−A (65.553)
2κ 2
869
In order to determine the function K, we write equation (65.550) into a more
convenient form as
1
ν(2κ2 + 1)Φσσ + (σ − A)Φσ
2
2 (65.554)
2κ − 1 (2κ2 − 1)A − κµ
= σ−
4(2κ2 + 1) 4(2κ2 + 1)
Equation (65.549) can be written as
1 1 1
Kσ = ν(2κ2 + 1)Φσσ + (σ − A)Φσ − σ + A (65.555)
2 8 4
Taking into account the first member of (65.554), the expression for Kσ of
the above equation takes the form
2κ2 − 1 (2κ2 − 1)A − κµ 1 1
Kσ = 2
σ − 2
− σ+ A
4(2κ + 1) 4(2κ + 1) 8 4
2κ2 − 3 2A + κµ
Kσ = 2
σ+ (65.556)
8(2κ + 1) 4(2κ2 + 1)
2κ2 − 3 2A + κµ
K(σ) = 2
σ2 + σ+D (65.557)
16(2κ + 1) 4(2κ2 + 1)
where D is an arbitrary constant.
Equations (65.544) and (65.545), substituting F and Kσ by (65.553) and
(65.556) respectively, take the new form
1 1
2ν(2κ2 + 1)Gσσ + (σ − A)Gσ + G − Φ
2 4
2 2 (65.558)
2κ − 1 (2κ − 1)A − κµ
− 2
σ+ =0
4(2κ + 1) 4(2κ2 + 1)
and
1 1
2ν(2κ2 + 1)Hσσ + (σ − A)Hσ + H − Φ
2 4
(65.559)
2κ2 − 1 (2κ2 − 1)A − κµ
− σ+ =0
4(2κ2 + 1) 4(2κ2 + 1)
870
respectively. There is no need to try to solve the last two equations, since it
1
is a simple matter to verify that both of them are satisfied by Φ. Therefore
2
1
G=H= Φ (65.560)
2
Using the values of the functions F given by (65.553), G, H by (65.560) and
K by (65.557), equations (65.529)-(65.532) become (for λ = κ)
1 1 1
U = ξ+ Φ+ σ−A (65.561)
2 2κ 2
1 1
V = κξ + Φ (65.562)
2 2
1 1
W = κξ + Φ (65.563)
2 2
and
1 1
P = (1 − 2κ2 )ξ 2 + [κ(η + θ) + µ]ξ
8 4
(65.564)
2κ2 − 3 2 2A + κµ
+ σ + σ + D
16(2κ2 + 1) 4(2κ2 + 1)
where Φ is given by (65.551) and (65.552).
Let us now go back and express the solution in terms of the original variables.
We first recall that
x y z
ξ=√ , η=√ , θ=√ (65.565)
t t t
and
U V W P
u= √ , v=√ , w=√ , p= (65.566)
t t t t
On the other hand
a2 a4
κ= , µ= (65.567)
a1 a1
and (with λ = κ)
σ = ζ + χ = (η − κξ) + (θ − κξ) = η + θ − 2κξ
y+z a2 x
= √ −2 √
t a1 t
871
i.e.
a1 (y + z) − 2a2 x
σ= √ (65.568)
a1 t
Therefore the solutions of the original Navier-Stokes equations are given by

x a1 −1/2 a1 (y + z) − 2a2 x −1/2
u= + Φ(σ) t + −At (65.569)
2t 2a2 2a1 t
a2 x 1
v= + Φ(σ) t−1/2 (65.570)
2a1 t 2
a2 x 1
w= + Φ(σ) t−1/2 (65.571)
2a1 t 2
and
a21 − 2a22 x2 1 a2 x(y + z) a4 x
p= · + + 3/2
8a21 t2 4a1 t2 t
2 2
2a2 − 3a1 [a1 (y + z) − 2a2 x]2
+ · (65.572)
16a21 (2a22 + a21 ) t2
2
2Aa1 + a2 a4 a1 (y + z) − 2a2 x
+ 2 2
· 3/2
+ D t−1
4a1 (2a2 + a1 ) t
where a1 , a2 , a4 are free parameters. The function Φ(σ) which appears in
(65.569)-(65.571) is given by (65.551)-(65.552) where κ and µ are given in
(65.567) and σ in (65.568).
65.2 Two-Dimensional Navier-Stokes Equations

The 2−dimensional Navier-Stokes equations have the form
ut + uux + vuy + px − ν(uxx + uyy ) = 0
vt + uvx + vvy + py − ν(vxx + vyy ) = 0 (65.573)
ux + vy = 0
The coefficients of the Lie symmetries are given by
ξ 1 = −a4 x + a3 y + F1 (t)
ξ 2 = −a3 x − a4 y + F2 (t) (65.574)
ξ 4 = −2a4 t + a5
872
and
φ1 = a4 u + a3 v + F10 (t)
φ2 = −a3 u + a4 v + F20 (t) (65.575)
φ4 = 2a4 p − F100 (t)x − F200 (t)y + F4 (t)
The above have been obtained from (65.203) and (65.202) by ignoring ξ 3 and
φ3 , the z−coordinate and the function w.
Renaming the various parameters and functions
a3 → a1 , a4 → a2 , a5 → a3 ,
(65.576)
ξ4 → ξ3, φ4 → φ3 , F4 → F3
expressions (65.574) and (65.575) take on the form
ξ 1 = −a2 x + a1 y + F1 (t)
ξ 2 = −a1 x − a2 y + F2 (t) (65.577)
ξ 3 = −2a2 t + a3
and
φ1 = a2 u + a1 v + F10 (t)
φ2 = −a1 u + a2 v + F20 (t) (65.578)
φ3 = 2a2 p − F100 (t)x − F200 (t)y + F3 (t)
respectively. Therefore the 2−dimensional Navier-Stokes equations admit

the symmetry generator
∂ ∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ 3 + φ1 + φ2 + φ3 (65.579)
∂x ∂y ∂t ∂u ∂v ∂p
∂ ∂
X = [−a2 x + a1 y + F1 (t)] + [−a1 x − a2 y + F2 (t)]
∂x ∂y
∂ ∂
+ (−2a2 t + a3 ) + [a2 u + a1 v + F10 (t)]
∂t ∂u (65.580)
∂
+ [−a1 u + a2 v + F20 (t)]
∂v
∂
+ [2a2 p − F100 (t)x − F200 (t)y + F3 (t)]
∂p
873
Writing the previous expression as

∂ ∂ ∂ ∂
X = a1 y −x +v −u
∂x ∂y ∂u ∂v

∂ ∂ ∂ ∂ ∂ ∂
+ a2 − x −y − 2t + u +v + 2p
∂x ∂y ∂t ∂u ∂v ∂p
(65.581)
∂ ∂ ∂ 0 ∂ 00 ∂
+ a3 + F3 (t) + F1 (t) + F1 (t) − F1 (t)x
∂t ∂p ∂x ∂u ∂p

∂ ∂ ∂
+ F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p
we obtain the following set of generators for the 2−dimensional Navier-Stokes
equations
∂ ∂ ∂ ∂
X1 = y −x +v −u
∂x ∂y ∂u ∂v
∂ ∂ ∂ ∂ ∂ ∂
X2 = −x −y − 2t + u +v + 2p
∂x ∂y ∂t ∂u ∂v ∂p
∂ ∂
X3 = , X 4 = F3 (t) (65.582)
∂t ∂p
∂ ∂ ∂
X5 = F1 (t) + F10 (t) − F100 (t)x
∂x ∂u ∂p
∂ ∂ ∂
X6 = F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p
In solving the 2−dimensional Navier-Stokes equations we shall need the in-
variant surface conditions
ξ 1 ux + ξ 2 uy + ξ 3 ut = φ1
ξ 1 vx + ξ 2 vy + ξ 3 vt = φ2 (65.583)
1 2 3 3
ξ px + ξ py + ξ pt = φ
[−a2 x + a1 y + F1 (t)]ux + [−a1 x − a2 y + F2 (t)]uy
+ (−2a2 t + a3 )ut = a2 u + a1 v + F10 (t),
[−a2 x + a1 y + F1 (t)]vx + [−a1 x − a2 y + F2 (t)]vy
(65.584)
+ (−2a2 t + a3 )vt = −a1 u + a2 v + F20 (t),
[−a2 x + a1 y + F1 (t)]px + [−a1 x − a2 y + F2 (t)]py
+ (−2a2 t + a3 )pt = 2a2 p − F100 (t)x − F200 (t)y + F3 (t)
874
Solution I. We consider the case
a1 = a2 = 0, a3 = 1 (65.585)
In this case the system (65.584) takes on the form
F1 (t)ux + F2 (t)uy + ut = F10 (t)

F1 (t)vx + F2 (t)vy + vt = F20 (t) (65.586)
F1 (t)px + F2 (t)py + pt = F3 (t) − F100 (t)x − F200 (t)y
Let us consider the auxiliary system associated to the system (65.686):

dx dy dt du
= = = 0
F1 (t) F2 (t) 1 F1 (t)
(65.587)
dv dp
= 0 =
F2 (t) F3 (t) − F1 (t)x − F200 (t)y
00
The equation
dx dt
= (65.588)
F1 (t) 1
R
after integration yields F1 (t)dt = x + C1 and thus one invariant is given by
Z
ξ = x − F̃1 (t), F̃1 (t) = F1 (t)dt (65.589)
The equation
dy dt
= (65.590)
F2 (t) 1
R
after integration yields F2 (t)dt = y + C2 and thus another invariant is given
by
Z
η = y − F̃2 (t), F̃2 (t) = F2 (t)dt (65.591)
The equations
dt du dt dv
= 0 and = 0 (65.592)
1 F1 (t) 1 F2 (t)
875
after integration yield u = F1 (t) + C3 and v = F2 (t) + C4 respectively. We
thus obtain the invariant functions
U (ξ, η) = u − F1 (t) and V (ξ, η) = v − F2 (t) (65.593)
In integrating the equation
dt dp
= (65.594)
1 F3 (t) − F1 (t)x − F200 (t)y
00
we have to set x = ξ + F̃1 (t) and y = η + F̃2 (t) respectively. Therefore

integration yields
Z Z Z Z
F3 (t)dt − ξ F1 (t)dt − η F2 (t)dt − F100 (t)F̃1 (t)dt
00 00
Z (65.595)
00
− F2 (t)F̃2 (t)dt = p + C5
Let
Z
F̃3 (t) = F3 (t)dt (65.596)
Then since
Z Z
F100 (t)dt = F10 (t), F200 (t)dt = F20 (t) (65.597)
and (using integration by parts)

Z Z
00
0
1
= F10 (t)F̃1 (t) − F12 (t)
Z Z2 (65.598)
00 0
1
= F20 (t)F̃2 (t) − F22 (t)
2

0 0 0 1 2
F̃3 (t) − ξF1 (t) − ηF2 (t) − F1 (t)F̃1 (t) − F1 (t)
2
(65.599)
1
− F20 (t)F̃2 (t) − F22 (t) = p + C5
2
876
Based on the above expression, we are able to introduce another invariant
function P (ξ, η) by
p = P (ξ, η) − ξF10 (t) − ηF20 (t) + H(t) (65.600)
where

1 2 1 2
H(t) = F̃3 (t) − F10 (t)F̃1 (t) 0
− F1 (t) − F2 (t)F̃2 (t) − F2 (t) (65.601)
2 2
Using (65.589) and (65.591) (i.e. substituting ξ and η by x and y respec-
tively), equation (65.600) takes on the form
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t) (65.602)
where

1 2 2
K(t) = F̃3 (t) + F1 (t) + F2 (t) (65.603)
2
Considering the invariants ξ, η and U (ξ, η), V (ξ, η), P (ξ, η), i.e.
x = ξ + F̃1 (t), y = η + F̃2 (t), u = U (ξ, η) + F1 (t),
(65.604)
the 2−dimensional Navier-Stokes equations (65.573) take the new form
U Uξ + V Uη + Pξ − ν(Uξξ + Uηη ) = 0
U Vξ + V Vη + Pη − ν(Vξξ + Vηη ) = 0 (65.605)
Uξ + Vη = 0
Symmetry analysis of equations (65.605). Let X be the generator of
symmetries of the equations (65.605):
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (65.606)
∂ξ ∂η ∂U ∂V ∂P
where ξ 1 , ξ 2 and φ1 , φ2 , φ3 are functions which depend on ξ, η, U, V, P . The
first prolongation is given by
∂ ∂ ∂ ∂
P r(1) X = X + U [ξ] + U [η] + V [ξ] + V [η]
∂Uξ ∂Uη ∂Vξ ∂Vη
(65.607)
∂ ∂
+ P [ξ] + P [η]
∂Pξ ∂Pη
877
For the second prolongation we have the expression
∂ ∂
P r(2) X = P r(1) X + U [ξξ] + U [ηη]
∂Uξξ ∂Uηη
(65.608)
[ξξ] ∂ ∂
+V + V [ηη]
∂Vξξ ∂Vηη
Invariance of equations (65.605) is expressed by the conditions
P r(2) X[∆k ] = 0, k = 1, 2, 3 (65.609)

∆1 =0, ∆2 =0, ∆3 =0
where ∆1 , ∆2 , ∆3 are given by

∆1 = U Uξ + V Uη + Pξ − ν(Uξξ + Uηη )
∆2 = U Vξ + V Vη + Pη − ν(Vξξ + Vηη ) (65.610)
∆3 = Uξ + Vη
The various coefficents U [ξ] , · · · , V [ηη] have been evaluated explicitly and are
given by
U [ξ] = (φ1 )ξ + [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]Pξ − (ξ 1 )U Uξ2
+ [(φ1 )U − (ξ 1 )ξ ]Uξ + (φ1 )V Vξ − (ξ 2 )ξ Uη − (ξ 1 )V Uξ Vξ (65.611)
− (ξ 2 )U Uη Uξ − (ξ 2 )V Uη Vξ
U [η] = (φ1 )η + [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]Pη − (ξ 2 )U Uη2

+ [(φ1 )U − (ξ 2 )η ]Uη + (φ1 )V Vη − (ξ 1 )η Uξ − (ξ 1 )U Uξ Uη (65.612)
− (ξ 1 )V Uξ Vη − (ξ 2 )V Uη Vη
V [ξ] = (φ2 )ξ + [(φ2 )P − (ξ 1 )P Vξ − (ξ 2 )P Vη ]Pξ − (ξ 1 )V Vξ2
+ [(φ2 )V − (ξ 1 )ξ ]Vξ + (φ2 )U Uξ − (ξ 2 )ξ Vη − (ξ 1 )U Uξ Vξ (65.613)
− (ξ 2 )U Uξ Vη − (ξ 2 )V Vη Vξ
V [η] = (φ2 )η + [(φ2 )P − (ξ 1 )P Vξ − (ξ 2 )P Vη ]Pη − (ξ 2 )V Vη2
+ [(φ2 )V − (ξ 2 )η ]Vη + (φ2 )U Uη − (ξ 1 )η Vξ − (ξ 1 )U Vξ Uη (65.614)
− (ξ 1 )V Vξ Vη − (ξ 2 )U Uη Vη
P [ξ] = (φ3 )ξ − [(ξ 2 )ξ + (ξ 2 )U Uξ + (ξ 2 )V Vξ ]Pξ − (ξ 1 )P Pξ2
+ [(φ3 )P − (ξ 1 )ξ ]Pξ + (φ3 )V Vξ + (φ3 )U Uξ − (ξ 1 )U Uξ Pξ (65.615)
− (ξ 1 )V Vξ Pξ − (ξ 2 )P Pη Pξ
878
P [η] = (φ3 )η − [(ξ 1 )η + (ξ 1 )U Uη + (ξ 1 )V Vη ]Pη − (ξ 2 )P Pη2
+ [(φ3 )P − (ξ 2 )η ]Pη + (φ3 )U Uη + (φ3 )V Vη − (ξ 1 )P Pξ Pη (65.616)
− (ξ 2 )U Uη Pη − (ξ 2 )V Vη Pη
U [ξξ] = (φ1 )ξξ + Pξξ [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]

− 2Pξ (ξ 1 )P Uξξ + (ξ 2 )P Uξη + (ξ 1 )U P Uξ2 + (ξ 1 )V P Uξ Vξ
+ (ξ 1 )ξP Uξ + (ξ 2 )U P Uξ Uη + (ξ 2 )V P Vξ Uη − (φ1 )ξP

2 1 1
+ (ξ )ξP Uη − (φ )U P Uξ − (φ )V P Vξ
− Pξ2 [(ξ 1 )P P Uξ + (ξ 2 )P P Uη − (φ1 )P P ] + 2(φ1 )ξV Vξ

(65.617)
+ Uξ [2(φ1 )ξU − (ξ 1 )ξξ ] − 2Uξη [(ξ 2 )U Uξ + (ξ 2 )V Vξ + (ξ 2 )ξ ]
− Vξ2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη − (φ1 )V V ] − 2(ξ 2 )U V Uη Uξ Vξ
− Uξ2 [(ξ 1 )U U Uξ + 2(ξ 1 )U V Vξ + (ξ 2 )U U Uη + 2(ξ 1 )ξU
− (φ1 )U U ] − Uη [(ξ 2 )ξξ + 2(ξ 2 )ξU Uξ ]
+ Uξξ [(φ1 )U − (ξ 2 )U Uη − 2(ξ 1 )V Vξ − 3(ξ 1 )U Uξ − 2(ξ 1 )ξ ]
− 2Vξ [(ξ 1 )ξV Uξ + (ξ 2 )ξV Uη − (φ1 )U V Uξ ]
+ Vξξ [(φ1 )V − (ξ 1 )V Uξ − (ξ 2 )V Uη ]
U [ηη] = (φ1 )ηη + Pηη [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]

− 2Pη (ξ 1 )P Uξη + (ξ 2 )P Uηη + (ξ 2 )U P Uη2 + (ξ 1 )U P Uξ Uη
+ (ξ 2 )ηP Uη + (ξ 1 )V P Uξ Vη + (ξ 2 )V P Vη Uη − (φ1 )ηP

1 1 1
+ (ξ )ηP Uξ − (φ )U P Uη − (φ )V P Vη
− Pη2 [(ξ 1 )P P Uξ + (ξ 2 )P P Uη − (φ1 )P P ] + 2(φ1 )ηV Vη

(65.618)
+ Uη [2(φ1 )ηU − (ξ 2 )ηη ] − 2Uξη [(ξ 1 )U Uη + (ξ 1 )V Vη + (ξ 1 )η ]
− Vη2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη − (φ1 )V V ] − 2(ξ 1 )U V Uη Uξ Vη
− Uη2 [(ξ 1 )U U Uξ + 2(ξ 2 )U V Vη + (ξ 2 )U U Uη + 2(ξ 2 )ηU
− (φ1 )U U ] − Uξ [(ξ 1 )ηη + 2(ξ 1 )ηU Uη ]
+ Uηη [(φ1 )U − (ξ 1 )U Uξ − 2(ξ 2 )V Vη − 3(ξ 2 )U Uη − 2(ξ 2 )η ]
− 2Vη [(ξ 1 )ηV Uξ + (ξ 2 )ηV Uη − (φ1 )U V Uη ]
+ Vηη [(φ1 )V − (ξ 1 )V Uξ − (ξ 2 )V Uη ]
879
V [ξξ] = (φ2 )ξξ + Pξξ [(φ2 )P − (ξ 1 )P Vξ − (ξ 2 )P Vη ]

− 2Pξ (ξ 1 )P Vξξ + (ξ 2 )P Vξη + (ξ 1 )V P Vξ2 + (ξ 1 )U P Uξ Vξ
+ (ξ 1 )ξP Vξ + (ξ 2 )U P Uξ Vη + (ξ 2 )V P Vξ Vη − (φ2 )ξP

2 2 2
+ (ξ )ξP Vη − (φ )U P Uξ − (φ )V P Vξ
− Pξ2 [(ξ 1 )P P Vξ + (ξ 2 )P P Vη − (φ2 )P P ] + 2(φ2 )ξV Vξ (65.619)

2 1 2 2 2
+ 2Vξ Uξ [(φ )U V − (ξ )ξU ] − 2Vξη [(ξ )U Uξ + (ξ )V Vξ + (ξ )ξ ]
− Uξ2 [(ξ 1 )U U Vξ + (ξ 2 )U U Vη − (φ2 )U U ] − 2(ξ 2 )U V Vη Uξ Vξ
− Vξ2 [(ξ 1 )V V Vξ + 2(ξ 1 )U V Uξ + (ξ 2 )V V Vη + 2(ξ 1 )ξV
− (φ2 )V V ] − Vξ [(ξ 1 )ξξ + 2(ξ 2 )ξV Vη ] + 2(φ2 )ξU Uξ
+ Vξξ [(φ2 )V − (ξ 2 )V Vη − 2(ξ 1 )U Uξ − 3(ξ 1 )V Vξ − 2(ξ 1 )ξ ]
− Vη [2(ξ 2 )ξU Uξ + (ξ 2 )ξξ ] + Uξξ [(φ2 )U − (ξ 1 )U Vξ − (ξ 2 )U Vη ]
V [ηη] = (φ2 )ηη + Pηη [(φ2 )P − (ξ 1 )P Vξ − (ξ 2 )P Vη ]

− 2Pη (ξ 1 )P Vξη + (ξ 2 )P Vηη + (ξ 2 )V P Vη2 + (ξ 1 )U P Vξ Uη
+ (ξ 1 )ηP Vξ + (ξ 1 )V P Vξ Vη + (ξ 2 )U P Vη Uη − (φ2 )ηP

2 2 2
+ (ξ )ηP Vη − (φ )U P Uη − (φ )V P Vη
− Pη2 [(ξ 1 )P P Vξ + (ξ 2 )P P Vη − (φ2 )P P ] + 2(φ2 )ηU Uη

(65.620)
+ Vη [2(φ2 )ηV − (ξ 2 )ηη ] − 2Vξη [(ξ 1 )U Uη + (ξ 1 )V Vη + (ξ 1 )η ]
− Uη2 [(ξ 2 )U U Vη + (ξ 1 )U U Vξ − (φ2 )U U ] − 2(ξ 1 )U V Uη Vξ Vη
− Vη2 [(ξ 1 )V V Vξ + 2(ξ 2 )U V Uη + (ξ 2 )V V Vη + 2(ξ 2 )ηV
− (φ2 )V V ] − Vξ [(ξ 1 )ηη + 2(ξ 1 )ηV Vη ]
+ Vηη [(φ2 )V − (ξ 1 )V Vξ − 2(ξ 2 )U Uη − 3(ξ 2 )V Vη − 2(ξ 2 )η ]
− 2Uη [(ξ 1 )ηU Vξ + (ξ 2 )ηU Vη − (φ2 )U V Vη ]
+ Uηη [(φ2 )U − (ξ 1 )U Vξ − (ξ 2 )U Vη ]
Let us first consider the condition
P r(2) X[∆3 ] =0 (65.621)

∆1 =0, ∆2 =0, ∆3 =0
880
where ∆3 is given by
∆3 = Uξ + Vη (65.622)
Using (65.608) and the above expression, we have
P r(2) X[∆3 ] = U [ξ] + V [η] (65.623)
Substitution of U [ξ] and V [η] by their explicit expressions (65.611) and (65.614)
repsectively, we obtain
P r(2) X[∆3 ]

= (φ1 )ξ + [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]Pξ − (ξ 1 )U Uξ2
+ [(φ1 )U − (ξ 1 )ξ ]Uξ + (φ1 )V Vξ − (ξ 2 )ξ Uη − (ξ 1 )V Uξ Vξ

2 2
− (ξ )U Uη Uξ − (ξ )V Uη Vξ
(65.624)

+ (φ2 )η + [(φ2 )P − (ξ 1 )P Vξ − (ξ 2 )P Vη ]Pη − (ξ 2 )V Vη2
+ [(φ2 )V − (ξ 2 )η ]Vη + (φ2 )U Uη − (ξ 1 )η Vξ − (ξ 1 )U Vξ Uη

1 2
− (ξ )V Vξ Vη − (ξ )U Uη Vη
In the above expression we make the substitutions
Pξ −→ − U Uξ − V Uη + ν(Uξξ + Uηη ) (65.625)
Pη −→ − U Vξ − V Vη + ν(Vξξ + Vηη ) (65.626)

and
Vη −→ − Uξ (65.627)
881
to account for the conditions ∆1 = 0, ∆2 = 0 and ∆3 = 0 respectively.
We thus see that the condition (65.621) is equivalent to the equation
(φ1 )ξ + [(φ1 )P − (ξ 1 )P Uξ − (ξ 2 )P Uη ]
×[−U Uξ − V Uη + ν(Uξξ + Uηη )]
− (ξ )U Uξ + [(φ1 )U − (ξ 1 )ξ ]Uξ + (φ1 )V Vξ − (ξ 2 )ξ Uη
1 2
− (ξ 1 )V Uξ Vξ − (ξ 2 )U Uη Uξ − (ξ 2 )V Uη Vξ
(65.628)
+ (φ2 )η + [(φ2 )P − (ξ 1 )P Vξ + (ξ 2 )P Uξ ]
×[−U Vξ + V Uξ + ν(Vξξ + Vηη )]
− (ξ )V Uξ − [(φ2 )V − (ξ 2 )η ]Uξ + (φ2 )U Uη − (ξ 1 )η Vξ
2 2
− (ξ 1 )U Vξ Uη + (ξ 1 )V Vξ Uξ + (ξ 2 )U Uη Uξ = 0
Equate to zero all the coefficients of the partial derivatives of the functions
U and V . We first see that the coefficients of Uη2 and Vξ2 are V (ξ 2 )P and
U (ξ 1 )P respectively. We thus have
(ξ 1 )P = (ξ 2 )P = 0 (65.629)
The terms Uξξ and Uηη share the common coefficient ν(φ1 )P while ν(φ2 )P is
the common coefficient of the terms Vξξ and Vηη . Therefore
(φ1 )P = (φ2 )P = 0 (65.630)
Taking into account (65.629) and (65.630), equation (65.628) takes on the
form
(φ1 )ξ − (ξ 1 )U Uξ2 + [(φ1 )U − (ξ 1 )ξ ]Uξ + (φ1 )V Vξ
− (ξ 2 )ξ Uη − (ξ 1 )V Uξ Vξ − (ξ 2 )U Uη Uξ − (ξ 2 )V Uη Vξ
(65.631)
+ (φ2 )η − (ξ 2 )V Uξ2 − [(φ2 )V − (ξ 2 )η ]Uξ + (φ2 )U Uη
− (ξ 1 )η Vξ − (ξ 1 )U Vξ Uη + (ξ 1 )V Vξ Uξ + (ξ 2 )U Uη Uξ = 0
The terms Uξ2 and Uη Vξ have common coefficient −[(ξ 1 )U + (ξ 2 )V ]. Therefore

we have the equation
(ξ 1 )U + (ξ 2 )V = 0 (65.632)
Equating to zero the coefficients of Vξ , Uη and Uξ we obtain the equations
−(ξ 1 )η + (φ1 )V = 0 (65.633)
882
−(ξ 2 )ξ + (φ2 )U = 0 (65.634)
and
−(ξ 1 )ξ + (ξ 2 )η + (φ1 )U − (φ2 )V = 0 (65.635)
respectively. We finally equate to zero the constant term
(φ1 )ξ + (φ2 )η = 0 (65.636)
P r(2) X[∆1 ] =0 (65.637)

∆1 =0, ∆2 =0, ∆3 =0
∆1 = U Uξ + V Uη + Pξ − ν(Uξξ + Uηη ) (65.638)
P r(2) X[∆1 ] = φ1 Uξ + φ2 Uη + U U [ξ] + V U [η]

(65.639)
+ P [ξ] − νU [ξξ] − νU [ηη]
We substitute in the above expression the U [ξ] , U [η] , P [ξ] , U [ξξ] and U [ηη]
given by (65.611), (65.612), (65.615), (65.617) and (65.618) respectively. We
then make the substitutions (65.625)-(65.627) and take into account (65.629)-
(65.630).
883
We then see that the condition (65.637) is equivalent to the equation
φ1 Uξ + φ2 Uη

+ U (φ1 )ξ − (ξ 1 )U Uξ2 + [(φ1 )U − (ξ 1 )ξ ]Uξ + (φ1 )V Vξ

2 1 2 2
− (ξ )ξ Uη − (ξ )V Uξ Vξ − (ξ )U Uη Uξ − (ξ )V Uη Vξ

+ V (φ1 )η − (ξ 2 )U Uη2 + [(φ1 )U − (ξ 2 )η ]Uη − (φ1 )V Uξ

1 1 1 2
− (ξ )η Uξ − (ξ )U Uξ Uη + (ξ )V Uξ Uξ + (ξ )V Uη Uξ

+ (φ3 )ξ − [(ξ 2 )ξ + (ξ 2 )U Uξ + (ξ 2 )V Vξ − (φ3 )P + (ξ 1 )ξ
+ (ξ 1 )U Uξ + (ξ 1 )V Vξ ]×[−U Uξ − V Uη + ν(Uξξ + Uηη )]

3 3
+ (φ )V Vξ + (φ )U Uξ

− ν (φ1 )ξξ + 2(φ1 )ξV Vξ
+ Uξ [2(φ1 )ξU − (ξ 1 )ξξ ] − 2Uξη [(ξ 2 )U Uξ + (ξ 2 )V Vξ + (ξ 2 )ξ ]

− Vξ2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη − (φ1 )V V ] − 2(ξ 2 )U V Uη Uξ Vξ
− Uξ2 [(ξ 1 )U U Uξ + 2(ξ 1 )U V Vξ + (ξ 2 )U U Uη + 2(ξ 1 )ξU
− (φ1 )U U ] − Uη [(ξ 2 )ξξ + 2(ξ 2 )ξU Uξ ]
+ Uξξ [(φ1 )U − (ξ 2 )U Uη − 2(ξ 1 )V Vξ − 3(ξ 1 )U Uξ − 2(ξ 1 )ξ ]
− 2Vξ [(ξ 1 )ξV Uξ + (ξ 2 )ξV Uη − (φ1 )U V Uξ ]

1 1 2
+ Vξξ [(φ )V − (ξ )V Uξ − (ξ )V Uη ]
884

− ν (φ1 )ηη − 2(φ1 )ηV Uξ
+ Uη [2(φ1 )ηU − (ξ 2 )ηη ] − 2Uξη [(ξ 1 )U Uη − (ξ 1 )V Uξ + (ξ 1 )η ]

− Vη2 [(ξ 1 )V V Uξ + (ξ 2 )V V Uη − (φ1 )V V ] + 2(ξ 1 )U V Uη Uξ2
− Uη2 [(ξ 1 )U U Uξ − 2(ξ 2 )U V Uξ + (ξ 2 )U U Uη + 2(ξ 2 )ηU
(65.640)
− (φ1 )U U ] − Uξ [(ξ 1 )ηη + 2(ξ 1 )ηU Uη ]
+ Uηη [(φ1 )U − (ξ 1 )U Uξ + 2(ξ 2 )V Uξ − 3(ξ 2 )U Uη − 2(ξ 2 )η ]
− 2Vη [(ξ 1 )ηV Uξ + (ξ 2 )ηV Uη − (φ1 )U V Uη ]

1 1 2
+ Vηη [(φ )V − (ξ )V Uξ − (ξ )V Uη ] = 0
Equate to zero all the coefficients of the derivatives of the functions U and
V . The coefficients of Vξ Uηη and Vξ Uξξ are −ν(ξ 1 )V and ν(ξ 1 )V respec-
tively. The coefficients of Uη Uηη and Uη Uξξ are 3ν(ξ 2 )U and ν(ξ 2 )U respec-
tively. The coefficients of Uξ Uηη , Vξ Uξη , Uη Vηη , Vξ Vηη , Uη Vξξ and Vξ Vξξ are
−2ν(ξ 2 )V , 2ν(ξ 2 )V , ν(ξ 2 )V , −ν(ξ 2 )V , ν(ξ 2 )V and −ν(ξ 2 )V respectively. The
terms Uξ Uξξ and Uη Uξη share the common coefficient 2ν(ξ 1 )U . Equating all
these coefficients to zero, we obtain the equations
(ξ 1 )U = (ξ 2 )U = 0
(65.641)
(ξ 1 )V = (ξ 2 )V = 0
885
Taking into account (65.641), equation (65.640) becomes
φ1 Uξ + φ2 Uη

1 1 1 1 2
+ U (φ )ξ + [(φ )U − (ξ )ξ ]Uξ + (φ )V Vξ − (ξ )ξ Uη

1 1 2 1 1
+ V (φ )η + [(φ )U − (ξ )η ]Uη − (φ )V Uξ − (ξ )η Uξ

+ (φ3 )ξ + [−(ξ 2 )ξ + (φ3 )P − (ξ 1 )ξ ]

3 3
×[−U Uξ − V Uη + ν(Uξξ + Uηη )] + (φ )V Vξ + (φ )U Uξ

− ν (φ1 )ξξ + 2(φ1 )ξV Vξ
(65.642)
+ Uξ [2(φ1 )ξU − (ξ 1 )ξξ ] − 2Uξη (ξ 2 )ξ + Vξ2 (φ1 )V V
+ Uξ2 (φ1 )U U − Uη (ξ 2 )ξξ + Uξξ [(φ1 )U − 2(ξ 1 )ξ ]

1 1
+ 2Vξ Uξ (φ )U V + Vξξ (φ )V

− ν (φ1 )ηη − 2(φ1 )ηV Uξ
+ Uη [2(φ1 )ηU − (ξ 2 )ηη ] − 2Uξη (ξ 1 )η + Uξ2 (φ1 )V V

+ Uη2 (φ1 )U U − Uξ (ξ 1 )ηη + Uηη [(φ1 )U − 2(ξ 2 )η ]

1 1
− 2Uξ Uη (φ )U V + Vηη (φ )V = 0
Equating the coefficients of Uη2 , Vξ2 , Uξ Vξ and Uη Uξ to zero, we obtain
(φ1 )U U = (φ1 )V V = (φ1 )U V = 0 (65.643)
We next equate to zero the rest of the coefficients. The coefficient of Uξη is
2ν[(ξ 1 )η + (ξ 2 )ξ ]. We thus have the equation
(ξ 1 )η + (ξ 2 )ξ = 0 (65.644)
The coefficient of Uηη is ν[(φ3 )P − (φ1 )U + 2(ξ 1 )η − (ξ 1 )ξ ]. We thus have the

equation
(φ3 )P − (φ1 )U + 2(ξ 2 )η − (ξ 1 )ξ = 0 (65.645)
886
The coefficient of Uξξ is ν[(φ3 )P − (φ1 )U + (ξ 1 )ξ ]. We thus have the equation
(φ3 )P − (φ1 )U + (ξ 1 )ξ = 0 (65.646)
The terms Vηη and Vξξ have common coefficient −ν[(ξ 2 )ξ + (φ1 )V ]. Therefore
we have the equation
(ξ 2 )ξ + (φ1 )V = 0 (65.647)
Equating to zero the coefficient of Uη , we get the equation
φ2 + V [(ξ 1 )ξ − (ξ 2 )η ] − U (ξ 2 )ξ + V [(φ1 )U − (φ3 )P ]

(65.648)
+ ν[(ξ 2 )ηη + (ξ 2 )ξξ − 2(φ1 )ηU ] = 0
Equating to zero the coefficient of Uξ , we get the equation
φ1 − V [(ξ 2 )ξ + (ξ 1 )η + (φ1 )V ] + U [(φ1 )U − (φ3 )P ]

(65.649)
+ (φ3 )U + ν[(ξ 1 )ηη + (ξ 1 )ξξ ] + 2ν[(φ1 )ηV − (φ1 )ξU ] = 0
Equating to zero the coefficient of Vξ , we get the equation
U [(ξ 2 )ξ + (φ1 )V ] + (φ3 )V − 2ν(φ1 )ξV = 0 (65.650)
Equating to zero the constant term, we obtain the equation
V (φ1 )η + U (φ1 )ξ + (φ3 )ξ − ν[(φ1 )ξξ + (φ1 )ηη ] = 0 (65.651)
P r(2) X[∆2 ] =0 (65.652)

∆1 =0, ∆2 =0, ∆3 =0
∆2 = U Vξ + V Vη + Pη − ν(Vξξ + Vηη ) (65.653)
P r(2) X[∆2 ] = φ1 Vξ + φ2 Vη + U V [ξ] + V V [η]

(65.654)
+ P [η] − νV [ξξ] − νV [ηη]
887
We substitute in the above expression the V [ξ] , V [η] , P [η] , V [ξξ] and V [ηη]
given by (65.613), (65.614), (65.616), (65.619) and (65.620) respectively. We
then make the substitutions (65.625)-(65.627) and take into account (65.629)-
(65.630) and (65.641).
We then see that the condition (65.652) is equivalent to the equation
φ1 Vξ − φ2 Uξ

2 2 2 2
+ U (φ )ξ − [(φ )V + (φ )U Uξ − (ξ )ξ ]Uξ

2 2 2 2 1
+ V (φ )η − [(φ )V − (ξ )η ]Uξ + (φ )U Uη − (ξ )η Vξ

+ (φ3 )η − [(ξ 1 )η − (φ3 )P + (ξ 2 )η ]

3 3
×[−U Vξ + V Uξ + ν(Vξξ + Vηη )] + (φ )U Uη + (φ )V Vη

(65.655)
− ν (φ2 )ξξ + 2(φ2 )ξV Vξ + 2Vξ Uξ (φ2 )U V − 2Vξη (ξ 2 )ξ
+ Uξ2 (φ2 )U U + Vξ2 (φ2 )V V − Vξ (ξ 1 )ξξ + 2(φ2 )ξU Uξ

2 1 2 2
+ Vξξ [(φ )V − 2(ξ )ξ ] − Vη (ξ )ξξ + Uξξ (φ )U

− ν (φ2 )ηη + 2(φ2 )ηU Uη + Vη [2(φ2 )ηV − (ξ 2 )ηη ]
− 2Vξη (ξ 1 )η + Uη2 (φ2 )U U + Vη2 (φ2 )V V − Vξ (ξ 1 )ηη

2 2 2 2
+ Vηη [(φ )V − 2(ξ )η ] − 2Uη Uξ (φ )U V + Uηη (φ )U = 0
Equating the coefficients of Uη2 , Uξ Vξ , Uη Uξ and Vξ2 to zero, we obtain the

system of equations
(φ2 )U U = (φ2 )U V = (φ2 )V V = 0 (65.656)
We next equate to zero the rest of the coefficients. The coefficient of Vηη is
ν[(ξ 2 )η − (φ2 )V + (φ3 )P ]. We thus have the equation
(ξ 2 )η − (φ2 )V + (φ3 )P = 0 (65.657)
888
The coefficient of Vξξ is ν[2(ξ 1 )ξ − (ξ 2 )η − (φ2 )V + (φ3 )P ]. We thus have the
equation
2(ξ 1 )ξ − (ξ 2 )η − (φ2 )V + (φ3 )P = 0 (65.658)
The coefficient of Uξη is 2ν[(ξ 1 )η + (ξ 2 )ξ ]. We thus have the equation
(ξ 1 )η + (ξ 2 )ξ = 0 (65.659)
The terms Uξξ and Uηη have common coefficient −ν[(ξ 1 )η + (φ2 )U ]. We then
have the equation
(ξ 1 )η + (φ2 )U = 0 (65.660)
Equating the coefficient of Uη to zero, we get the equation
V [(ξ 1 )η + (φ2 )U ] + (φ3 )U − 2ν(φ2 )ηU = 0 (65.661)
Equating the coefficient of Uξ to zero, we get the equation
− φ2 + U [(ξ 1 )η + (ξ 2 )ξ + (φ2 )U ] + V [(φ3 )P − (φ2 )V ]

(65.662)
− (φ3 )V − ν[(ξ 2 )ηη + (ξ 2 )ξξ ] + 2ν[(φ2 )ηV − (φ2 )ξU ] = 0
Equationg to zero the coefficient of Vξ , we get the equation
φ1 − V (ξ 1 )η + U [(ξ 2 )η − (ξ 1 )ξ + (φ2 )V − (φ3 )P ]

(65.663)
+ ν[(ξ 1 )ξξ + (ξ 1 )ηη − (φ2 )ξV ] = 0
Equating the constant term to zero, we obtain the equation
V (φ2 )η + U (φ2 )ξ + (φ3 )η − ν[(φ2 )ξξ + (φ2 )ηη ] = 0 (65.664)
We shall solve the determining equations. Because of (65.643), the function

φ1 is a linear function of U and V :
φ1 = D1 (ξ, η)U + E1 (ξ, η)V + Z1 (ξ, η) (65.665)
Substituting the above expression into the equation (65.651) of the constant
term, we get the equation
(D1 )ξ U 2 + (E1 )η V 2 + [(D1 )η + (E1 )ξ ]U V
+ [(Z1 )ξ − ν(D1 )ξξ − ν(D1 )ηη ]U + [(Z1 )η − ν(E1 )ξξ − ν(E1 )ηη ]V (65.666)
+ (φ3 )ξ − ν(Z1 )ξξ − ν(Z1 )ηη = 0
889
Equating to zero all the coefficients of the functions U and V we arrive at
the following system of equations
(D1 )ξ = 0 and (E1 )η = 0 (65.667)
(D1 )η + (E1 )ξ = 0 (65.668)

(Z1 )ξ − ν(D1 )ξξ − ν(D1 )ηη = 0 (65.669)
(Z1 )η − ν(E1 )ξξ − ν(E1 )ηη = 0 (65.670)
and
(φ3 )ξ − ν(Z1 )ξξ − ν(Z1 )ηη = 0 (65.671)
Because of (65.667), we obtain
D1 = D1 (η) and E1 = E1 (ξ) (65.672)
Equation (65.668) then gives D10 (η) + E10 (ξ) = 0 and thus
D10 (η) = a1 and E10 (ξ) = −a1 (65.673)
and by integration
D 1 = a1 η + a2 and E1 = −a1 ξ + a3 (65.674)
Equations (65.669) and (65.679) then take the form (Z1 )ξ = 0 and (Z1 )η = 0
respectively and thus
Z 1 = a4 (65.675)
We thus obtain the following representation for the function φ1 using the
expression (65.665) and (65.674)-(65.675):
φ1 = (a1 η + a2 )U + (−a1 ξ + a3 )V + a4 (65.676)
Finally equation (65.671) yields
(φ3 )ξ = 0 (65.677)
Using equations (65.656) we conclude that the function φ2 is a linear function

of U and V and thus will admit the following representation
φ2 = D2 (ξ, η)U + E2 (ξ, η)V + Z2 (ξ, η) (65.678)
890
The equation (65.664) of the constant term then gives
(D2 )ξ U 2 + (E2 )η V 2 + [(D2 )η + (E2 )ξ ]U V

+ [(Z2 )ξ − ν(D2 )ξξ − ν(D2 )ηη ]U + [(Z2 )η − ν(E1 )ξξ − ν(E2 )ηη ]V (65.679)
+ (φ3 )η − ν(Z2 )ξξ − ν(Z2 )ηη = 0
Equating to zero all the coefficients of the functions U and V we arrive at

the following system of equations
(D2 )ξ = 0 and (E2 )η = 0 (65.680)
(D2 )η + (E2 )ξ = 0 (65.681)

(Z2 )ξ − ν(D2 )ξξ − ν(D2 )ηη = 0 (65.682)
(Z2 )η − ν(E2 )ξξ − ν(E2 )ηη = 0 (65.683)
and
(φ3 )η − ν(Z2 )ξξ − ν(Z2 )ηη = 0 (65.684)
Because of (65.680), we obtain
D2 = D2 (η) and E2 = E2 (ξ) (65.685)
Equation (65.681) then gives D20 (η) + E20 (ξ) = 0 and thus
D10 (η) = a5 and E10 (ξ) = −a5 (65.686)
and by integration
D 1 = a5 η + a6 and E2 = −a5 ξ + a7 (65.687)
Equations (65.682) and (65.683) then take the form (Z2 )ξ = 0 and (Z2 )η = 0
respectively and thus
Z 2 = a8 (65.688)
We thus obtain the following representation for the function φ2 using the
expression (65.678) and (65.687)-(65.688):
φ2 = (a5 η + a6 )U + (−a5 ξ + a7 )V + a8 (65.689)
891
Finally equation (65.684) yields
(φ3 )η = 0 (65.690)
Using the expression (65.689) for φ2 , we obtain from (65.660) the equation
(ξ 1 )η + a5 η + a6 = 0 (65.691)
and by integration with respect to η we obtain (remember the function ξ 1 is

independent of U, V and P )
1
ξ 1 = − a5 η 2 − a6 η + K1 (ξ) (65.692)
2
Using the expression (65.676) for φ1 , we obtain from (65.647) the equation
(ξ 2 )ξ + (−a1 ξ + a3 ) = 0 (65.693)
and by integration with respect to ξ we obtain (remember the function ξ 2 is

independent of U, V and P )
1
ξ 2 = a1 ξ 2 − a3 ξ + K2 (η) (65.694)
2
The functions K1 (ξ) and K2 (η) can be determined by considering the equa-
tion
(ξ 1 )ξ = (ξ 2 )η (65.695)
which can be obtained by subtracting (65.645)-(65.646). Using the above

equation and the expressions (65.692) and (65.694) for ξ 1 and ξ 2 respectively,
we obtain K10 (ξ) = K20 (η) and thus K10 (ξ) = k and K20 (η) = k. We thus obtain
by integration
K1 (ξ) = kξ + m1 and K2 (η) = kη + m2 (65.696)
Therefore ξ 1 and ξ 2 become

1
ξ 1 = − a5 η 2 − a6 η + kξ + m1 (65.697)
2
and
1
ξ 2 = a1 ξ 2 − a3 ξ + kη + m2 (65.698)
2
892
respectively. However there is still another relation between ξ 1 and ξ 2 which
is equation (65.659). Using this equation and the above two expressions for
ξ 1 and ξ 2 , we obtain the equation
a1 ξ − a5 η − a6 − a3 = 0 (65.699)
and thus
a1 = 0, a5 = 0 and a6 = −a3 (65.700)
Therefore the expressions for ξ 1 and ξ 2 given by (65.697) and (65.698) take
the new form
ξ 1 = a3 η + kξ + m1 (65.701)
and
ξ 2 = −a3 ξ + kη + m2 (65.702)
respectively. On the other hand, the expressions (65.676) and (65.689) for
φ1 and φ2 respectively take the new form
φ1 = a2 U + a3 V + a4 (65.703)
and
φ2 = −a3 U + a7 V + a8 (65.704)
Using (65.701)-(54.704), we obtain from (65.645) and (65.646)
(φ3 )P = a2 − k (65.705)
and from (65.650)
(φ3 )V = 0 (65.706)
From (65.648) using (65.701)-(65.705) we obtain (a7 + k)V + a8 = 0 and thus
a7 = −k and a8 = 0 (65.707)
From (65.649), using (65.701)-(65.705) we obtain
(φ3 )U + (a2 + k)U + a4 = 0 (65.708)
893
Using (65.701)-(65.705) we obtain from (65.657) and (65.658)
a7 = a2 (65.709)
Equation (65.663) gives the first of (65.707). Equation (65.662) gives (φ3 )V =
(a2 − 2a7 − k)V − a + 8 and using (65.707), we obtain
(φ3 )V = (a2 + k)V (65.710)
Comparing (65.706) and (65.710), we obtain
k = −a2 (65.711)
From (65.661) we obtain (φ3 )U = 0 and because of (65.708), we get k = −a2

(which we have already found) and
a4 = 0 (65.712)
Let us collect everything together. So far we have found
a4 = 0, a8 = 0, k = −a2 , a7 = a2 (65.713)
and
(φ3 )ξ = (φ3 )η = (φ3 )V = (φ3 )U = 0, (φ3 )P = 2a2 (65.714)
Therefore (65.701)-(65.704) become
ξ 1 = −a2 ξ + a3 η + m1 , ξ 2 = −a3 ξ − a2 η + m2 (65.715)
φ1 = a2 U + a3 V, φ2 = −a3 U + a2 V (65.716)
Taking into account (65.714), we get by integration of (φ3 )P = 2a2 that
φ3 = 2a2 P + a5 (65.717)
Upon renaming m1 → a1 and m2 → a4 , relations (65.715)-(65.717) take the

final form
ξ 1 = −a2 ξ + a3 η + a1 , ξ 2 = −a3 ξ − a2 η + a4 (65.718)
φ1 = a2 U + a3 V, φ2 = −a3 U + a2 V, φ3 = 2a2 P + a5 (65.719)
894
The generator
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (65.720)
∂ξ ∂η ∂U ∂V ∂P
of symmetries for the system (65.605) has now the form
∂ ∂
X = (−a2 ξ + a3 η + a1 ) + (−a3 ξ − a2 η + a4 )
∂ξ ∂η
(65.721)
∂ ∂ ∂
+ (a2 U + a3 V ) + (−a3 U + a2 V ) + (2a2 P + a5 )
∂U ∂V ∂P
Writing the above expression as
∂ ∂ ∂ ∂ ∂ ∂
X = a1 + a2 −ξ −η +U +V + 2P
∂ξ ∂ξ ∂η ∂U ∂V ∂P
∂ (65.722)
∂ ∂ ∂ ∂ ∂
+ a3 η −ξ +V −U + a4 + a5
∂ξ ∂η ∂U ∂V ∂η ∂P
one can read off the following generators
∂ ∂ ∂ ∂ ∂ ∂
X1 = , X 2 = −ξ −η +U +V + 2P
∂ξ ∂ξ ∂η ∂U ∂V ∂P
(65.723)
∂ ∂ ∂ ∂ ∂ ∂
X3 = η −ξ +V −U , X4 = , X5 =
∂ξ ∂η ∂U ∂V ∂η ∂P
Let us determine some solutions of the system (65.605) by considering the
ξ 1 Uξ + ξ 2 Uη = φ1
ξ 1 Vξ + ξ 2 Vη = φ2 (65.724)
ξ 1 Pξ + ξ 2 Pη = φ3
(−a2 ξ + a3 η + a1 )Uξ + (−a3 ξ − a2 η + a4 )Uη = a2 U + a3 V
(−a2 ξ + a3 η + a1 )Vξ + (−a3 ξ − a2 η + a4 )Vη = −a3 U + a2 V (65.725)
(−a2 ξ + a3 η + a1 )Pξ + (−a3 ξ − a2 η + a4 )Pη = 2a2 P + a5
I. For a2 = a3 = 0 the system (65.725) becomes
a1 Uξ + a4 Uη = 0
a1 Vξ + a4 Vη = 0 (65.726)
a1 P ξ + a4 P η = a5
895
a η − a ξ a η − a ξ
1 4 1 4
U =F , V =G ,
a1 a1
a η − a ξ (65.727)
a5 1 4
P = ξ+H
a1 a1
where F, G and H are arbitrary functions. At this point we introduce the
following notation
a4 a5 a1 η − a4 ξ
κ= , λ= , θ≡ = η − κξ (65.728)
a1 a1 a1
We will then have
U = F (θ), V = G(θ), P = λξ + H(θ) (65.729)
When the above expressions of the functions U, V and P are substituted in

the system (65.605), we shall derive the following system of equations
− κF F 0 + GF 0 + λ − κH 0 − νκ2 F 00 − νF 00 = 0 (65.730)
−κF G0 + GG0 + H 0 − νκ2 G00 − νG00 = 0 (65.731)

and
−κF 0 + G0 = 0 (65.732)
where the prime denotes differentiation with respect to θ.

Equation (65.732) can be solved with respect to G:
G = κF + C (65.733)
where C is an arbitrary constant. We substitute this value of G into each

one of the equations (65.730 and (65.731) and we derive the equations
ν(1 + κ2 ) 00 C 0 λ
H0 = − F + F + (65.734)
κ κ κ
and
H 0 = ν(1 + κ2 )F 00 − κCF 0 (65.735)
896
respectively. Equating the two different expressions of H 0 , we obtain the
equation
λ
ν(1 + κ2 )F 00 − CF 0 − =0 (65.736)
1 + κ2
ν(1 + κ2 ) Cθ λ
F (θ) = A exp 2
− θ+B (65.737)
C ν(1 + κ ) C(1 + κ2 )
where A, B are arbitrary constants.

Eliminating F 00 between (65.735) and (65.736), we obtain the equation
κλ
H0 = − (65.738)
1 + κ2
from which, after integration with respect to θ yields
κλ
H(θ) = − θ+D (65.739)
1 + κ2
where D is an arbitrary constant. Substitution of (65.737) into (65.733) gives
the value of the function G:
νκ(1 + κ2 ) Cθ κλ
G(θ) = A exp 2
− θ
C ν(1 + κ ) C(1 + κ2 ) (65.740)
+ κB + C
The solution of the 2−dimensional NS equations (65.573) can be expressed

in terms of the original variables using (65.604), written down again for
convenience
x = ξ + F̃1 (t), y = η + F̃2 (t), u = U (ξ, η) + F1 (t),
(65.741)
where K(t) is given in (65.603).

a4
θ = η − κξ = [y − F̃2 (t)] − [x − F̃1 (t)]
a1
897
and after the necessary operations
(a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]

θ=− (65.742)
a1
We have further
u = U (ξ, η) + F1 (t) = F (θ) + F1 (t)
ν(1 + κ2 ) Cθ λ
=A exp 2
− θ + B + F1 (t)
C ν(1 + κ ) C(1 + κ2 )
ν(a21 + a24 )
=A
Ca2 (65.743)
( 1 )
Ca21 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
×exp −
ν(a21 + a24 ) a1
a1 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
+ 2 2
+ F1 (t) + B
C(a1 + a4 ) a1
v = V (ξ, η) + F2 (t) = G(θ) + F2 (t)

νκ(1 + κ2 ) Cθ κλ
=A exp 2
− θ
C ν(1 + κ ) C(1 + κ2 )
+ κB + C + F2 (t)
νa4 (a21 + a24 )
=A
Ca3
( 1 ) (65.744)
Ca21 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
×exp −
ν(a21 + a24 ) a1
a4 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
+ 2 2
C(a1 + a4 ) a1
a4
+ F2 (t) + B + C
a1
898
and
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t)
= λξ + H(θ) − F10 (t)x − F20 (t)y + K(t)
a5 κλ
= [x − F̃1 (t)] − θ + D − F10 (t)x − F20 (t)y + K(t)
a1 1 + κ2
(65.745)
a5 a4 a5 (a4 x − a1 y) + [a1 F̃2 (t) − a4 F̃1 (t)]
= [x − F̃1 (t)] + 2
a1 a1 + a24 a1

0 0 1 2 2
− F1 (t)x − F2 (t)y + F̃3 (t) + F1 (t) + F2 (t) + D
2
II. For a3 = a5 = 0, the system (65.725) becomes
(−a2 ξ + a1 )Uξ + (−a2 η + a4 )Uη = a2 U

(−a2 ξ + a1 )Vξ + (−a2 η + a4 )Vη = a2 V (65.746)
(−a2 ξ + a1 )Pξ + (−a2 η + a4 )Pη = 2a2 P

1 a η−a
2 4
U= F ,
a1 − a2 ξ a2 (a1 − a2 ξ)
1 a η−a
2 4
V = G , (65.747)
a1 − a2 ξ a2 (a1 − a2 ξ)
1 a η−a
2 4
P = H
(a1 − a2 ξ)2 a2 (a1 − a2 ξ)
where F, G and H are arbitrary functions. The above expressions of the
functions U, V and P are substituted into the system (65.605) and we obtain
the equations
a2 F (θF 0 + F ) + GF 0 + a2 θH 0 + 2a2 H
(65.748)
− νa22 (θ2 F 00 + 4θF 0 + 2F ) − νF 00 = 0
a2 F (θG0 + G) + GG0 + H 0
(65.749)
− νa22 (θ2 G00 + 4θG0 + 2G) − νG00 = 0
and
1 0
θF 0 + F = − G (65.750)
a2
899
respectively. In the above equations all the functions F, G and H depend on
the variable θ, where we have set
a2 η − a4
θ= (65.751)
a2 (a1 − a2 ξ)
The prime denotes differentiation with respect to θ as well.
Solving (65.750) with respect to G, we get
G = −a2 θF + C (65.752)

Substitution of the function G given by (65.752) into the equations (65.748)
and (65.749), gives the two equations
− ν(a22 θ2 + 1)F 00 − 4νa22 θF 0 + CF 0 − 2νa22 F + a2 F 2

(65.753)
+ a2 θH 0 + 2a2 H = 0
and
νa2 θ(1 + a22 θ2 )F 00 + 2νa2 (1 + 3a22 θ2 )F 0 − a2 CθF 0 − a22 θF 2
(65.754)
+ 6νa32 θF − 2νa22 C + H 0 = 0
respectively. Equations (65.753) and (65.754) constitute a system of ordinary

differential equations with two unknown functions.
Eliminating F 00 between these two equations, i.e. solving (65.753) with
respect to ν(a22 θ2 + 1)F 00 ,
ν(a22 θ2 + 1)F 00 = −4νa22 θF 0 + CF 0 − 2νa22 F + a2 F 2

(65.755)
+ a2 θH 0 + 2a2 H
and substituting into (65.754), we obtain the simple equation
(1 + a22 θ2 )H 0 + 2a22 θH + 2νa2 (1 + a22 θ2 )F 0 + 4νa32 θF = 2νa22 C (65.756)
The above equation, written as

d d
[(1 + a22 θ2 )H] + 2νa2 [(1 + a22 θ2 )F ] = 2νa22 C (65.757)
dθ dθ
can be integrated immediatelly providing the solution
(1 + a22 θ2 )H + 2νa2 (1 + a22 θ2 )F = 2νa22 Cθ + A (65.758)
900
Solving the above equation with respect to H and substitution into (65.754),
νθ(1 + a22 θ2 )F 00 + 6νa22 θ2 F 0 − CθF 0 − a2 θF 2 + 6νa22 θF
1 − a22 θ2 θ (65.759)
+ 2νCa2 2 2 2
− 2Aa2 − 2νCa2 = 0
(1 + a2 θ ) (1 + a22 θ2 )2
The above equation, despite its complicated nature, can be reduced into an
Abel-type first order ODE using Lie symmetry analysis.
Using symmetry analysis, we find that the indinitesimal coefficients {ξ 1 , φ1 }
are given by
ξ 1 = 1 + a22 θ2 and φ1 = a2 (−2a2 θF + C) (65.760)
The normal coordinates {r(θ, F ), s(θ, F )} are found by solving the system
∂ ∂
(1 + a22 θ2 ) r(θ, F ) + a2 (−2a2 θF + C) r(θ, F ) = 0
∂θ ∂F (65.761)
∂ ∂
(1 + a22 θ2 ) s(θ, F ) + a2 (−2a2 θF + C) s(θ, F ) = 1
∂θ ∂F
r(θ, F ) = H1 (F + a22 θ2 F − Ca2 θ)
1 (65.762)
s(θ, F ) = arctan(a2 θ) + H2 (F + a22 θ2 F − Ca2 θ)
a2
where H1 , H2 are arbitrary functions of the corresponding arguments.
Let us take for simplicity
1
r(θ, F ) = F + a22 θ2 F − Ca2 θ and s(θ, F ) = arctan(a2 θ) (65.763)
a2
Inverting the above transformation, we get (where {r, s(r)} the new variables)
1 r + C tan(a2 s(r))
θ= tan(a2 s(r)), F (θ) = (65.764)
a2 1 + tan2 (a2 s(r))
d2 d 3
ν 2 s(r) + (a2 r2 − 4νa22 r + 2a2 A + a2 C 2 ) s(r)
dr dr (65.765)
d 2
+C s(r) = 0
dr
901
The substitution
d
Y (r) = s(r) (65.766)
dr
d
ν Y (r) + a2 (r2 − 4νa2 r + 2A + C 2 )Y 3 (r) + CY 2 (r) = 0 (65.767)
dr
which is an Abel-type of ODE, as promised. The above equation can be
solved only in some special cases. However it can be reduced into a Abel
equation of second type and then solved by the versatile Panayotounakos
algorithm.
The function s(r) can be calculated by integration of (65.766). Let
Z
s(r) = Y (r)dr ≡ M (r) (65.768)
The original function F can be determined through the equation (using

(65.763))
Z
1
arctan(a2 θ) = Y (r)dr ≡ M (r)|r=(1+a22 θ2 )F −Ca2 θ (65.769)
a2
For C = 0 equation (65.767) admits the solution
√
3ν
Y (r) = ± p (65.770)
2a2 r − 12a22 νr2 + 12a2 Ar
3
Integrating the above function we arrive at

√
K 3ν (65.771)
s(r) = ± F (m, z)
M
where we have defined
q
K = −3νa2 + 9a22 ν 2 − 6A
q
M = 6a2 K(18a22 ν 2 − A) + 18Aa22 ν
s
K (65.772)
m=
2(K + 3a2 ν)
r
r+K
z=
K
902
F (m, z) is the elliptic integral defined by (see any textbook on special func-
tions like Rainville [15], Whittaker and Watson [16])
Z z
dt
F (m, z) = p (65.773)
0 (1 − t )(1 − m2 t2 )
2
Then according to (65.769) (for C = 0 and taking into account the definition
of z in (65.772))
√ r
1 K 3ν (1 + a22 θ2 )F (θ) + K (65.774)
arctan(a2 θ) = ± F m,
a2 M K
We now have to solve the above equation with respect to F (θ). It is known
that if (Whittaker and Watson [16], §21.7, p.480)
Z z
dt
u= p = F (m, z) (65.775)
0 (1 − t )(1 − m2 t2 )
2
then
z = sn(u, m) (65.776)
where sn(u, m) is Jacobi’s sine function.

Since
r
(1 + a22 θ2 )F (θ) + K M
F m, =± √ arctan(a2 θ)
K Ka2 3ν
we have
r+K M
= sn2 m, √ arctan(a2 θ) (65.777)
K Ka2 3ν
and thus

2 M

r = −K 1 − sn m, √ arctan(a2 θ) (65.778)
Ka2 3ν
or
M
r = −K cn2 m, √ arctan(a2 θ) (65.779)
Ka2 3ν
903
and since r = (1 + a22 θ2 )F (θ), we obtain
M
(1 + a22 θ2 )F (θ) = −K cn2 m, √ arctan(a2 θ)
Ka2 3ν
From the above equation we obtain the function F (θ):
K 2
M
F (θ) = − cn m, √ arctan(a 2 θ) (65.780)
(1 + a22 θ2 ) Ka2 3ν
The other two functions G and H are determined from (65.752) and (65.758)
respectively.
For A = C = 0 equation (65.767) admits the solution
√
3ν
Y (r) = ± p (65.781)
2a2 r − 12νa22 r2 + D
3
where D is an arbitrary constant. Solving the equation

d
s(r) = Y (r)
dr
to recover s(r), we find for D = 0 and ignoring constants of integration
1 √r − 6a
2
s(r) = ± √ arctan √ (65.782)
a2 3ν 6a 2 ν
Comparing (65.782) and the second of (65.763), we get
1 √r − 6a
2
± √ arctan √ = arctan(a2 θ) (65.783)
3ν 6a2 ν
where (from the first of (65.763) for C = 0)
r = (1 + a22 θ2 )F (65.784)
Combining the above two equations we find (for A = C = 0)

√

6a2 2
F (θ) = 1 + ν·tan ± 3ν arctan(a2 θ) (65.785)
1 + a22 θ2
904
Using the above solution for F and (65.758) for A = C = 0 we find the value
of the function H:
12νa22 √

2
H(θ) = − 1 + ν·tan ± 3ν arctan(a2 θ) (65.786)
1 + a22 θ2
Finally equation (65.752) gives the function G:
6a22 θ √

2
G(θ) = − 1 + ν·tan ± 3ν arctan(a2 θ) (65.787)
1 + a22 θ2
We shall convert the above expressions for the various functions in terms of
the original variables. First of all we have
ξ = x − F̃1 (t), η = y − F̃2 (t) and
a2 η − a4 a2 [y − F̃2 (t)] − a4 (65.788)
θ≡ =
a2 (a1 − a2 ξ) a2 [a1 − a2 [x − F̃1 (t)]]
The original functions are recovered through the relations
u = U (ξ, η) + F1 (t), v = V (ξ, η) + F2 (t),
(65.789)
p = P (ξ, η) − F10 (t)x − F20 (t)y + K(t)
where
1 a η−a
2 4
U (ξ, η) = F
a1 − a2 ξ a2 (a1 − a2 ξ)
1
= F (θ),
a1 − a2 [x − F̃1 (t)]
1 a η−a
2 4
V (ξ, η) = G
a1 − a2 ξ a2 (a1 − a2 ξ)
(65.790)
1
= G(θ),
a1 − a2 [x − F̃1 (t)]
1 a η−a
2 4
P (ξ, η) = H
(a1 − a2 ξ)2 a2 (a1 − a2 ξ)
1
= H(θ)
(a1 − a2 [x − F̃1 (t)])2
and

1 2 2
K(t) = F̃3 (t) + F1 (t) + F2 (t) (65.791)
2
905
Let us make some comments about the general solution of equation (65.767).
The substitution (Appendix A)
1
w(r) = (65.792)
Y (r)
a2 2 C
ww0 = (r − 4νa2 r + 2A + C 2 ) + w (65.793)
ν ν
The second substitution
ν
r= z (65.794)
C
4a2 ν 2

0 a2 ν 2 2
w(z)w (z) − w(z) = z − z + 2A + C 2 (65.795)
C C C
i.e. to an equation of the form
w(z)w0 (z) − w(z) = f (z) (65.796)
The above is an Abel equation of the second type and thus solvable by the
Panayotounakos algorithm (Appendix B). The function F (θ) is recovered
using (65.769) while G and H are determined from (65.752) and (65.758)
respectively.
Solution II. We shall find the solution of the 2−dimensional NS equations

(65.573) considering the generaor X 2 (corresponding to the dilatation sub-
group) given in (65.582), written down here for convenience
∂ ∂ ∂ ∂ ∂ ∂
X 2 = −x −y − 2t + u +v + 2p (65.797)
∂x ∂y ∂t ∂u ∂v ∂p
The auxiliary system associated to the above generator is
dx dy dt du dv dp
= = = = = (65.798)
−x −y −2t u v 2p
906
Integrating
dx dt dy dt
= and =
−x −2t −y −2t
we obtain
x y
= C1 and = C2
t1/2 t1/2
respectively. We thus introduce the two invariants ξ and η by
ξ = xt−1/2 and η = yt−1/2 (65.799)
respectively. Similarly integrating

dt du dt dv dt dp
= , = , =
−2t u −2t v −2t 2p
we obtain
u v p
= C3 , = C4 , = C5
t−1/2 t−1/2 t−1
We thus introduce three more invariants U, V and P by
Substituting the above expressions into the Navier-Stokes equations (65.573)

we derive the equations
1 1 ∂U 1 ∂U ∂U ∂U ∂P
− U− ξ − η +U +V +
2 2 ∂ξ 2 ∂η ∂ξ ∂η ∂ξ
∂ 2U 2 (65.801)
∂ U
−ν + =0
∂ξ 2 ∂η 2
1 1 ∂V 1 ∂V ∂V ∂V ∂P
− V − ξ − η +U +V +
2 2 ∂ξ 2 ∂η ∂ξ ∂η ∂η
∂ 2V 2 (65.802)
∂ V
−ν 2
+ =0
∂ξ ∂η 2
and
∂U ∂V
+ =0 (65.803)
∂ξ ∂η
907
The coefficients {ξ 1 , ξ 2 } and {φ1 , φ2 , φ2 } of the generator
∂ ∂ ∂ ∂ ∂
X = ξ1 + ξ2 + φ1 + φ2 + φ3 (65.804)
∂ξ ∂η ∂U ∂V ∂P
of symmetries for the system (65.801)-(65.803) have the form
ξ 1 = a4 η + a1 , ξ 2 = −a4 ξ + a2 (65.805)
and
1 1
φ1 = a4 V + a1 , φ2 = −a4 U + a2 ,
2 2 (65.806)
1
φ3 = (a1 ξ + a2 η + a3 )
4
The above expressions is the 2−dimensional version of (65.516)-(65.518)
which we have derived explicitly for the 3−dimensional case.
The invariant surface conditions are given by
ξ 1 Uξ + ξ 2 Uη = φ1
ξ 1 Vξ + ξ 2 Vη = φ2 (65.807)
ξ 1 Pξ + ξ 2 Pη = φ3
which can be written in expanded form as

1
(a4 η + a1 )Uξ + (−a4 ξ + a2 )Uη = a4 V + a1
2
1
(a4 η + a1 )Vξ + (−a4 ξ + a2 )Vη = −a4 U + a2 (65.808)
2
1
(a4 η + a1 )Pξ + (−a4 ξ + a2 )Pη = (a1 ξ + a2 η + a3 )
4
We shall consider the case a4 = 0. In this case the invariant surface conditions
take the form
1
a1 Uξ + a2 Uη = a1
2
1
a1 Vξ + a2 Vη = a2 (65.809)
2
1
a1 Pξ + a2 Pη = (a1 ξ + a2 η + a3 )
4
908
1 a η − a ξ
1 2
U (ξ, η) = ξ + F
2 a1
1 a2 a η − a ξ
1 2
V (ξ, η) = ξ+G (65.810)
2 a1 a1
2 2
1 a1 − a2 2 1 a2
a3 a η − a ξ
1 2
P (ξ, η) = 2
ξ + η+ ξ+H
2 a1 4 a1 a1 a1
Under the above substitutions, the reduced system of the NS equations
(65.801)-(65.803) become
1 1 1
− θF 0 (θ) − κF (θ)F 0 (θ) + G(θ)F 0 (θ) + κθ + λ
2 4 4 (65.811)
− κH 0 (θ) − ν(1 + κ2 )F 00 (θ) = 0
1 1 1
− θG(θ) − θG0 (θ) + κF (θ) − κF (θ)G0 (θ)
2 2 2 (65.812)
+ G(θ)G0 (θ) + H 0 (θ) − ν(1 + κ2 )G00 (θ) = 0
and
1
− κF 0 (θ) + G0 (θ) = 0 (65.813)
2
respectively, where we have introduced the notation
a2 a3 a1 η − a2 ξ
κ= , λ= , θ≡ = η − κξ (65.814)
a1 a1 a1
Integrating the third equation, we derive the following expression which
expresses G in terms of F :
1
G(θ) = κF (θ) − θ + C (65.815)
2
where C is an arbitrary constant. Substitution of this expression for G into
the first two equations (65.811) and (65.812), we obtain
ν(1 + κ2 ) 00 C − θ 0 1 λ
H0 = − F + F + +θ (65.816)
κ κ 4 κ
and
3
H 0 = νκ(1 + κ2 )F 00 + κ(C + θ)F 0 − θ + C (65.817)
4
909
respectively. Eliminating the function F 00 between the two expressions of H 0 ,
we get the equation
2 0 0 1 2 1
(1 + κ )H = 2κCF + (κ − 3)θ + κλ + C (65.818)
4 4
The above equation expresses H in terms of F . On the other hand, equating
the two expressions of H 0 , we obtain the equation
ν(1 + κ2 )2 F 00 + [ (1 + κ2 )θ + (−1 + κ2 )C ] F 0 − κθ
1 (65.819)
+ κC − λ = 0
4
ζ = (1 + κ2 )θ + (−1 + κ2 )C (65.820)
we find that the general solution of equation (65.819) is given by
√ C 2 (1 − κ2 )2

κ 2 ×exp
F (θ) = B + + A 1 + κ
1 + κ2 2ν(1 + κ2 )3
1 3 ζ2
×ζ× 1 F1 ; ; −
2 2 2ν(1 + κ2 )3
8Cκ3 − λ(1 + κ2 )

− 2
×(ζ ) 2(1 + κ2 )3 (65.821)
8ν(1 + κ2 )5
1 3 ζ2 1 3 ζ2
× 1 F1 ; ; − 1 F1 ; ;
2 2 2ν(1 + κ2 )3 2 2 2ν(1 + κ2 )3
ζ2

3
− 2 F2 1, 1; , 2;
2 2ν(1 + κ2 )3
where A and B are arbitrary constants.
From (65.818) we find by integration with respect to θ the function H:
2κC κ2 − 3 2 κλ + 4C
H(θ) = F (θ) + θ + θ+D (65.822)
1 + κ2 8(1 + κ2 ) 4(1 + κ2 )
where D is an arbitrary constant.
The functions have now to be expressed in terms of the original variables.
a1 η − a2 ξ a1 yt−1/2 − a2 xt−1/2 a1 y − a2 x
θ= = = √ (65.823)
a1 a1 a1 t
910
and then
ζ = (1 + κ2 )θ + (−1 + κ2 )C
a2 + a2 a1 y − a2 x (65.824)
= 1 2 2· √ + (−1 + κ2 )C
a1 a1 t
We also obtain
1 a η − a ξ 1
1 2
U (ξ, η) = ξ + F = xt−1/2 + F (θ)
2 a1 2
1 a2 a η − a ξ 1 a
1 2 2
V (ξ, η) = ξ+G = xt−1/2 + G(θ)
2 a1 a1 2 a1
1 a2 −1/2 a2 1 a1 y − a2 x
= xt + F (θ) − √ +C
2 a1 a1 2 a1 t
1 a21 − a22 2 1 a2 a3 a η − a ξ
1 2
P (ξ, η) = 2
ξ + η+ ξ+H
2 a1 4 a1 a1 a1
2
1 a1 − a2 x 2 2
1 a2 −1/2 a3
(65.825)
= 2
+ yt + xt−1/2 + H(θ)
2 a1 t 4 a1 a1
1 a21 − a22 x2 1 a2 −1/2 a3 −1/2
= + yt + xt
2 a21 t 4 a1 a1
2Ca1 a2 a22 − 3a21 a1 y − a2 x 2
+ 2 F (θ) + √
a1 + a22 8(a21 + a22 ) a1 t
a2 a3 + 4Ca21 a1 y − a2 x
+ √ +D
4(a21 + a22 ) a1 t
The original functions are recovered through the relations
The function F (θ) is given in (65.821) where θ and ζ are given by (65.823)
and (65.824) respectively.
Appendix A. Transformations of Abel’s equations.
Abel’s equation of the first kind
dy
= A3 (x)y 3 + A2 (x)y 2 + A1 (x)y (65.827)
dx
under the substitution
H(x) Z
y(x) = , H(x) = exp A1 (x) dx (65.828)
w(x)
911
takes on the form
dw(x)
w(x) = B1 (x) w(x) + B0 (x) (65.829)
dx
where B1 (x), B0 (x) are functions expressed in terms of A3 (x), A2 (x) and
A0 (x). The transformation
Z
z = B1 (x) dx≡f (x) (65.830)
converts equation (6.3) into the equation
dw(z) B0 (x)
w(z) − w(z) = (65.831)
dz B1 (x) x=f −1 (z)
which is Abel’s equation of second kind and thus accessible by Panayotounakos

algorithm. The reader can consult Polyanin and Zaitsev (Reference [19]).
Appendix B. The Panayotounakos algorithm.
The Panayotounakos algorithm solves Abel’s equation of the second kind
as follows: The solution of the differential equation (Abel’s equation of the
second kind) (see References [17] and [18])
dy
yy 0 − y = f (x), y≡y(x), y0 = (65.832)
dx
is given by
1 1
y(x) = (x + 2C) r(x) + (65.833)
3 3
where C is a constant and r(x) are the roots of the cubic equation
r3 (t) + p r(t) + q = 0, t = ln |x + 2C| (65.834)
The quantities p and q are given by

1 1 a 3
p = b − a3 , q = c − ab + 2 (65.835)
3 3 3
a = −4, b = 3 + 4[g(t) + f (t)]e−t , c = −4[g(t) + 2f (t)]e−t (65.836)
912
and g(t) is a function defined by
[(t sin(t) + cos(t))ci(t) + cos2 (t)][4t ci(t) + cos(t)]e−t

g(t) = − 2f (t)
2[2t ci(t)]3
(65.837)
The function ci(t) is the known cosine integral function defined by

Zt
cosu − 1
ci(t) = γ + ln(t) + du (65.838)
u
0
where γ is the Euler-Mascheroni constant.
References
[1] V. V. Puhnachev: ”Group properties of the equations of Navier-Stokes
in the plane”. Journal of Appl. Mech. and Tech. Phys. 1 (1960) 83-90.
[2] V. O. Bytev: ”Invariant Solutions of the Navier-Stokes Equations”.
Z. Prikl. Mekh. Tekh. Fiziki 6 (1972) 56-64.
[3] L. V. Kapitanskii: ”Group-Theoretic Analysis of the Navier-Stokes
Equations in the Rotationally Symmetric Case and Some New Exact Solu-
tions”. Zapiski Nauch. Sem. Lening. Otdel. Mat. Inst. im. Steklova Akad.
Nauk SSSR Vol. 84 (1979) 89-107.
[4] S. P. Lloyd: ”The infinitesimal group of the Navier-Stokes equations”.
Acta Mechanica 38 (1981) 85-98.
[5] R. E. Boisvert: ”Group Analysis of the Navier-Stokes equations”.
PhD Thesis, Georgia Institute of Technology 1982
[6] R. E. Boisvert, W. F. Ames and U. N. Srivastava: ”Group
properties and new solutions of Navier-Stokes equations”.
J. Eng. Math. 17 (1983) 203-221.
[7] W. F. Ames and M. C. Nucci: ”Analysis of fluid equations by group
methods”. J. Eng. Math. 20 (1985) 181-187.
[8] T. A. Bright: ”New solutions to the Euler equations using Lie group
analysis and high order numerical techniques”.
PhD Thesis, Georgia Institute of Technology 1993
[9] T. A. Bright: ”Solutions to Euler Equations using Lie Group Analysis”.
Int. J. Nonl. Mech. 32 (1997) 303-316.
913
[10] W. I. Fushchich, W. M. Shtelsn and S. L. Slavutsky: ”Reduction
and exact Solutions of the Navier-Stokes Equations”.
J. Phys. A 24 (1991) 971-984.
[11] W. Fushchych and R. Popowych: ”Symmetry Reductions and
Exact Solutions of the Navier-Stokes Equations”.
Nonlin. Math. Phys. 1 (1994) 75-113.
[12] X. Hu, Z. Dong and Y. Chen: ”Symmetry Reductions and Exact
Solutions of the (2 + 1)−Dimensional Navier-Stokes Equations”.
Z. Naturforsch. 65a (2010) 504-510.
Springer 2000
[14] A. R. Forsyth: ”A Treatise on Differential Equations”. Dover 1996
[15] E. D. Rainville: ”Special Functions”. Chelsea 1960
[16] E. T. Whittaker and G. N. Watson: ”A Course of Modern
Analysis”. Cambridge University Press 1973
[17] D. Panayotounakos: ”Exact analytic solutions of unsolvable classes
of first and second order nonlinear ODEs (Part I: Abels equations)”.
Applied Math. Lett. 18 (2005) 155-162
[18] D. E. Panayotounakos, Th. Zarmpoutis and P. Sotiropoulos:
”The General Solutions of the Normal Abel’s Type Nonlinear ODE of the
Second Kind”. IAENG Int. J. Appl. Math. 43:3 (2013).
[19] A. D. Polyanin and V. F. Zaitsev: ”Handbook of Exact Solutions
for Ordinary Differential Equations”. Second Edition, CRC 2003.
Chapter XXI
General Relativity
66 The Einstein Equations for
Axially Symmetric Gravitational Fields
The line element for axially symmetric gravitational field in vacuum has the
form
ds2 = e2ψ dt2 − e−2ψ [e2γ (dρ2 + dz 2 ) + ρ2 dφ2 ] (66.1)
914
which is the Weyl-Lewis-Papapetrou metric. In the above expression for the
line element, the variables ρ, z, φ and t are Weyl’s canonical coordinates while
ψ(ρ, z) and γ(ρ, z) are unknown functions determined from Einstein’s field
equations.
Einstein’s equations in vacuum read
1
Rµν − gµν R = 0 (66.2)
2
where Rµν and R are the Ricci tensor and the scalar curvature respectively,
expressed in terms of the components of the metric tensor gµν (the gravita-
tional potentials).
Einstein’s equations for the metric (66.1) have the form
∂ 2 ψ 1 ∂ψ ∂ 2 ψ
Mψ≡ 2 + + 2 =0 (66.3)
∂ρ ρ ∂ρ ∂z
and
γρ = ρ [(ψρ )2 − (ψz )2 ], γz = 2ρψρ ψz (66.4)
We thus see that the main task is to determine the function ψ, since the
function γ can be obtained by solving the system (66.4). Equation (66.3)
has been solved using Lie symmetry methods in Section 62.
The generators of the algebra of Lie point symmetries of equation (66.3) are
given by (62.28), i.e.
∂ ∂ ∂
X1 = , X2 = ρ + z ,
∂z ∂ρ ∂z
(66.5)
∂ 1 ∂ 1 ∂ ∂
X 3 = ρz + (−ρ2 + z 2 ) − zψ , 4
X =ψ
∂ρ 2 ∂z 2 ∂ψ ∂ψ
First Solution. The solution of (66.3) associated to the generator X 1 is
given by
ψ(ρ, z) = A + B· ln(ρ) (66.6)
Second Solution. The general solution of (66.3) associated to the generator

X 2 is given by
z
ψ(ρ, z) = A + B·arctan p (66.7)
ρ2 − z 2
915
Third Solution. The general solution of (66.3) associated to the linear
combination of generators X 2 and X 4 , i.e.
∂ ∂ ∂
X 2 + a·X 4 = ρ +z +aψ (66.8)
∂ρ ∂z ∂ψ
is being expressed in terms of the hypergeometric equation
1 a − 1 a − 2 z 2 + ρ2
ψ(ρ, z) = A· 2 F1 − , ; ; +
2a 2a 2a z2
(66.9)
z 2 + ρ2 a + 2 a + 1 2a + 1 3a + 2 z 2 + ρ2
+ B· 2a · 2 F1 , ; ;
z2 2a 2a 2a z2
67 The Einstein-Maxwell Equations for

Axially Symmetric Gravitational Fields
The line element of a static axially symmetric space-time is given by Weyl’s
expression
ds2 = −e2φ dt2 + e−2φ [r2 dφ2 + e2µ (dr2 + dz 2 )] (67.1)
where φ and µ are functions which depend on r and z only.
The Einstein-Maxwell equations (in units where 8πG = 1, c = 1) for a given
metric are given by
Rµν = Tµν
1
Tµν = Fµρ Fνρ − gµν Fρσ F ρσ (67.2)
4
µν µν
F;ν = 0, F = Aµ,ν − Aν,µ
where Rµν is Ricci’s tensor, Tµν is the electromagnetic energy tensor and
Fµν is the electromagnetic field tensor. The semicolon denotes covariant
differentiation and the comma partial differentiation. We denote by
Aµ = (φ, A, 0, 0) (67.3)
the 4-potential. Using Weyl’s element, the Einstein-Maxwell equations for a
purely magnetostatic field read
1 e2φ
φrr + φr + φzz − 2 (A2r + A2z ) = 0
r 2r (67.4)
1
Arr − Ar + Azz + 2(Ar φr + Az φz ) = 0
r
916
and
e2φ 2
µr = r(φ2r − φ2z ) + (Ar − A2z )
2r (67.5)
e2φ
µz = 2rφr φz + Ar Az
r
We thus see that once we know φ and A we would be able to determine the
function µ by solving the PDEs (67.5). We are going to solve the system
(67.4) using Lie point symmetries. We denote by ∆1 and ∆2 the expressions
1 e2φ
∆1 ≡φrr + φr + φzz − 2 (A2r + A2z )
r 2r (67.6)
1
∆2 ≡Arr − Ar + Azz + 2(Ar φr + Az φz )
r
The second prolongation in our case is defined by
∂ ∂ ∂ ∂
P r(2) Γ = ξ 1 + ξ2 + φ1 + φ2
∂r ∂z ∂φ ∂A
∂ ∂ ∂ ∂
+ φ[r] + φ[rr] + φ[z] + φ[zz] (67.7)
∂φr ∂φrr ∂φz ∂φzz
∂ ∂ ∂ ∂
+ A[r] + A[rr] + A[z] + A[zz]
∂Ar ∂Arr ∂Az ∂Azz
where ξ 1 , ξ 2 , φ1 and φ2 are functions where each one of them depends on the
variables r, z, φ and A.
Invariance of the system of equations ∆1 = 0 and ∆2 = 0 is being expressed
by the conditions
P r(2) Γ(∆1 ) = 0 and P r(2) Γ(∆2 ) =0 (67.8)

∆1 =0 ∆2 =0
Let us now list the various infinitesimal coefficients. They are given by
φ[r] = (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − (φr )2 (ξ 1 )φ − φr Ar (ξ 1 )A

(67.9)
− φr φz (ξ 2 )φ − φz Ar (ξ 2 )A − φz (ξ 2 )r + Ar (φ1 )A
φ[z] = (φ1 )z + [(φ1 )φ − (ξ 2 )z ]φz − (φz )2 (ξ 2 )φ − φr Az (ξ 1 )A

(67.10)
− φr φz (ξ 1 )φ − φz Az (ξ 2 )A − φr (ξ 1 )z + Az (φ1 )A
917
A[r] = (φ2 )r + [(φ2 )A − (ξ 1 )r ]Ar − (Ar )2 (ξ 1 )A − φr Ar (ξ 1 )φ
(67.11)
− Ar Az (ξ 2 )A − φr Az (ξ 2 )φ − Az (ξ 2 )r + φr (φ2 )φ
A[z] = (φ2 )z + [(φ2 )A − (ξ 2 )z ]Az − (Az )2 (ξ 2 )A − Ar Az (ξ 1 )A
(67.12)
− Ar φz (ξ 1 )φ − φz Az (ξ 2 )φ − Ar (ξ 1 )z + φz (φ2 )φ
φ[rr] = (φ1 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }φrz
+ {(φ1 )φ − 2(ξ 1 )r − 3φr (ξ 1 )φ − 2Ar (ξ 1 )A − φz (ξ 2 )φ }φrr
+ {(φ1 )A − φr (ξ 1 )A − φz (ξ 2 )A }Arr − (φr )3 (ξ 1 )φφ
+ {(φ1 )φφ − 2(ξ 1 )rφ − 2Ar (ξ 1 )Aφ − φz (ξ 2 )φφ }(φr )2 (67.13)
+ {(φ1 )AA − φr (ξ 1 )AA − φz (ξ 2 )AA }(Ar )2 + 2Ar (φ1 )rA
+ 2{(φ1 )Aφ − (ξ 1 )rA }φr Ar − 2{φr (ξ 2 )Aφ + (ξ 2 )rA }Ar φz
+ {2(φ1 )rφ − (ξ 1 )rr }φr − {(ξ 2 )rr + 2φr (ξ 2 )rφ }φz
φ[zz] = (φ1 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }φrz
+ {(φ1 )φ − 2(ξ 2 )z − 3φz (ξ 2 )φ − 2Az (ξ 2 )A − φr (ξ 1 )φ }φzz
+ {(φ1 )A − φr (ξ 1 )A − φz (ξ 2 )A }Azz − (φz )3 (ξ 2 )φφ
+ {(φ1 )φφ − 2(ξ 2 )zφ − 2Az (ξ 2 )Aφ − φr (ξ 1 )φφ }(φz )2 (67.14)
+ {(φ1 )AA − φr (ξ 1 )AA − φz (ξ 2 )AA }(Az )2 + 2Az (φ1 )zA
+ 2{(φ1 )Aφ − (ξ 2 )zA }φz Az − 2{φz (ξ 1 )Aφ + (ξ 1 )zA }Az φr
+ {2(φ1 )zφ − (ξ 2 )zz }φz − {(ξ 1 )zz + 2φz (ξ 1 )zφ }φr
A[rr] = (φ2 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }Arz
+ {(φ2 )A − 2(ξ 1 )r − 3Ar (ξ 1 )A − 2φr (ξ 1 )φ − Az (ξ 2 )A }Arr
+ {(φ2 )φ − Ar (ξ 1 )φ − Az (ξ 2 )φ }φrr − (Ar )3 (ξ 1 )AA
+ {(φ2 )AA − 2(ξ 1 )rA − 2φr (ξ 1 )Aφ − Az (ξ 2 )AA }(Ar )2 (67.15)
+ {(φ2 )φφ − Ar (ξ 1 )φφ − Az (ξ 2 )φφ }(φr )2 + 2φr (φ2 )rφ
+ 2{(φ2 )Aφ − (ξ 1 )rφ }φr Ar − 2{φr (ξ 2 )Aφ + (ξ 2 )rA }Ar Az
+ {2(φ2 )rA − (ξ 1 )rr }Ar − {(ξ 2 )rr + 2φr (ξ 2 )rφ }Az
918
A[zz] = (φ2 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }Arz
+ {(φ2 )A − 2(ξ 2 )z − 3Az (ξ 2 )A − 2φz (ξ 2 )φ − Ar (ξ 1 )A }Azz
+ {(φ2 )φ − Ar (ξ 1 )φ − Az (ξ 2 )φ }φzz − (Az )3 (ξ 2 )AA
+ {(φ2 )AA − 2(ξ 2 )zA − 2φz (ξ 2 )Aφ − Ar (ξ 1 )AA }(Az )2 (67.16)
+ {(φ2 )φφ − Ar (ξ 1 )φφ − Az (ξ 2 )φφ }(φz )2 + 2φz (φ2 )zφ
+ 2{(φ2 )Aφ − (ξ 2 )zφ }φz Az − 2{φz (ξ 1 )Aφ + (ξ 1 )zA }Ar Az
+ {2(φ2 )zA − (ξ 2 )zz }Az − {(ξ 1 )zz + 2φz (ξ 1 )zφ }Ar
We get the following expression for P r(2) Γ(∆1 )
P r(2) Γ(∆1 ) =

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂r ∂z ∂φ ∂A
∂ ∂ ∂ ∂
+ φ[r] + φ[rr] + φ[z] + φ[zz]

[r] ∂ [rr] ∂ [z] ∂ [zz] ∂
+A +A +A +A (67.17)
1 e2φ 2 2

φrr + φr + φzz − 2 (Ar + Az )
r 2r
1 2φ e2φ
e 1
= ξ 1 − 2 φr + 3 (A2r + A2z ) + φ1 − 2 (A2r + A2z ) + φ[r]
r r r r
e2φ e2φ
+ φ[rr] + φ[zz] + − 2 Ar A[r] + − 2 Az A[z]
r r
The above expression can be written in expanded form (taking into account
the explicit form of the various infinitesimal coefficients)
919
(2)
1
1 e2φ 2 2

1
e2φ
2 2

P r Γ(∆1 ) = ξ − 2 φr + 3 (Ar + Az ) + φ − 2 (Ar + Az )
r r r
1
+ (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − (φr )2 (ξ 1 )φ − φr Ar (ξ 1 )A
r

2 2 2 1
− φr φz (ξ )φ − φz Ar (ξ )A − φz (ξ )r + Ar (φ )A

+ (φ1 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }φrz

+ {(φ1 )φφ − 2(ξ 1 )rφ − 2Ar (ξ 1 )Aφ − φz (ξ 2 )φφ }(φr )2

1 1 2 2
+ {2(φ )rφ − (ξ )rr }φr − {(ξ )rr + 2φr (ξ )rφ }φz

+ (φ1 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }φrz (67.18)
+ {(φ1 )φ − 2(ξ 2 )z − 3φz (ξ 2 )φ − 2Az (ξ 2 )A − φr (ξ 1 )φ }φzz
+ {(φ1 )φφ − 2(ξ 2 )zφ − 2Az (ξ 2 )Aφ − φr (ξ 1 )φφ }(φz )2

1 2 1 1
+ {2(φ )zφ − (ξ )zz }φz − {(ξ )zz + 2φz (ξ )zφ }φr
e2φ
+ − 2 Ar (φ2 )r + [(φ2 )A − (ξ 1 )r ]Ar − (Ar )2 (ξ 1 )A − φr Ar (ξ 1 )φ
r

2 2 2 2
− Ar Az (ξ )A − φr Az (ξ )φ − Az (ξ )r + φr (φ )φ
e2φ
+ − 2 Az (φ2 )z + [(φ2 )A − (ξ 2 )z ]Az − (Az )2 (ξ 2 )A − Ar Az (ξ 1 )A
r

1 2 1 2
− Ar φz (ξ )φ − φz Az (ξ )φ − Ar (ξ )z + φz (φ )φ
920
1 e2φ
In the above expression we substitute φzz by −φrr − φr + 2 (A2r + A2z ) to
r 2r
account for the condition ∆1 = 0.
We thus see that condition P r(2) Γ(∆1 ) = 0 is equivalent to
∆1 =0
1 1 e2φ 2 2

1
e2φ
2 2

ξ − 2 φr + 3 (Ar + Az ) + φ − 2 (Ar + Az )
r r r
1
+ (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − (φr )2 (ξ 1 )φ − φr Ar (ξ 1 )A
r

2 2 2 1

+ (φ1 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }φrz

+ {(φ1 )φφ − 2(ξ 1 )rφ − 2Ar (ξ 1 )Aφ − φz (ξ 2 )φφ }(φr )2

1 1 2 2
+ {2(φ )rφ − (ξ )rr }φr − {(ξ )rr + 2φr (ξ )rφ }φz

+ (φ1 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }φrz
+ {(φ1 )φ − 2(ξ 2 )z − 3φz (ξ 2 )φ − 2Az (ξ 2 )A − φr (ξ 1 )φ }×

1 e2φ
× −φrr − φr + 2 (A2r + A2z )
r 2r
+ {(φ1 )φφ − 2(ξ 2 )zφ − 2Az (ξ 2 )Aφ − φr (ξ 1 )φφ }(φz )2

1 2 1 1
+ {2(φ )zφ − (ξ )zz }φz − {(ξ )zz + 2φz (ξ )zφ }φr
921
e2φ
+ − 2 Ar (φ2 )r + [(φ2 )A − (ξ 1 )r ]Ar − (Ar )2 (ξ 1 )A − φr Ar (ξ 1 )φ
r

2 2 2 2
− Ar Az (ξ )A − φr Az (ξ )φ − Az (ξ )r + φr (φ )φ
e2φ
+ − 2 Az (φ2 )z + [(φ2 )A − (ξ 2 )z ]Az − (Az )2 (ξ 2 )A − Ar Az (ξ 1 )A
r

1 2 1 2
− Ar φz (ξ )φ − φz Az (ξ )φ − Ar (ξ )z + φz (φ )φ = 0
We now have to equate to zero the coefficients of the partial derivatives

of the unknown functions φ and A. However, just for the economy of the
calculations, we can equate first some easily identified coefficients and the
simplify further the equation for the condition P r(2) Γ(∆1 ) = 0. For
∆1 =0
e2φ 2
example we see that the coefficient of Ar φr Az is (ξ )φ , the coefficient of
r2
e2φ 1
Az φz Ar is (ξ )φ , the coefficient of φr Arr is −(ξ 1 )A and the coefficient of
r2
Az φrr is 2(ξ 2 )A . The terms Arr and Azz have common coefficient (φ1 )A . We
thus obtain the equations
(ξ 1 )φ = (ξ 1 )A = 0, (ξ 2 )φ = (ξ 2 )A = 0, (φ1 )A = 0 (67.19)
922
Therefore, because of (67.19), the condition P r(2) Γ(∆1 ) = 0 gets sim-
∆1 =0
plified into the equation
1 e2φ e2φ
ξ 1 − 2 φr + 3 (A2r + A2z ) + φ1 − 2 (A2r + A2z )
r r r
1
+ (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − φz (ξ 2 )r
r

+ (φ1 )rr − 2(ξ 2 )r φrz + {(φ1 )φ − 2(ξ 1 )r }φrr

1 2 1 1 2
+ (φ )φφ (φr ) + {2(φ )rφ − (ξ )rr }φr − (ξ )rr φz

+ (φ1 )zz − 2(ξ 1 )z φrz
1 e2φ
+ {(φ1 )φ − 2(ξ 2 )z }× −φrr − φr + 2 (A2r + A2z )
r 2r
+ {2(φ1 )zφ − (ξ 2 )zz }φz − (ξ 1 )zz φr
e2φ
2 2 1 2 2
+ − 2 Ar (φ )r + [(φ )A − (ξ )r ]Ar − Az (ξ )r + φr (φ )φ
r
e2φ
2 2 2 1 2
+ − 2 Az (φ )z + [(φ )A − (ξ )z ]Az − Ar (ξ )z + φz (φ )φ = 0
r
From the above equation, we that the terms Ar φr and Az φz have common
e2φ e2φ
coefficient − 2 (φ2 )φ while the terms Ar and Az have coefficients − 2 (φ2 )r
r r
e2φ 2
and − 2 (φ )z respectively. We then obtain the equations
r
(φ2 )r = (φ2 )z = (φ2 )φ = 0 (67.20)
e2φ 2
The terms φrz and Ar Az have coefficients −2{(ξ 2 )r + (ξ 1 )z } and {(ξ )r +
r2
(ξ 1 )z } respectively. We thus have the equation
(ξ 2 )r + (ξ 1 )z = 0 (67.21)
923
Equating to zero the coefficients of (Ar )2 and (Az )2 we obtain (after the
necessary simplifications) two more equations
1 1 1
ξ − φ1 + {(φ1 )φ − 2(ξ 2 )z } − {(φ2 )A − (ξ 1 )r } = 0 (67.22)
r 2
and
1 1 1
ξ − φ1 + {(φ1 )φ − 2(ξ 2 )z } − {(φ2 )A − (ξ 2 )z } = 0 (67.23)
r 2
respectively.
Taking into account the above equations, the condition P r(2) Γ(∆1 ) =0
∆1 =0
gets simplified even more into the equation
1 1
1 1 1 1 2
ξ − 2 φr + (φ )r + [(φ )φ − (ξ )r ]φr − φz (ξ )r
r r

+ (φ1 )rr + {(φ1 )φ − 2(ξ 1 )r }φrr

1 2 1 1 2
+ (φ )φφ (φr ) + {2(φ )rφ − (ξ )rr }φr − (ξ )rr φz
1
+ (φ1 )zz + {(φ1 )φ − 2(ξ 2 )z }× −φrr − φr
r

+ {2(φ1 )zφ − (ξ 2 )zz }φz − (ξ 1 )zz φr = 0
From the above equation we see that the coefficient of φrr is −2{(ξ 1 )r −(ξ 2 )z }
and then we get the equation
(ξ 1 )r − (ξ 2 )z = 0 (67.24)
It is obvious that the above equation can also be obtained by subtracting
equations (67.22) and (67.23).
Equating the coefficients of φr and φz to zero, after the necessary simplifica-
tions, we get the equations
1 1 1
− ξ + {2(ξ 2 )z − (ξ 1 )r } + {2(φ1 )rφ − (ξ 1 )rr } − (ξ 1 )zz = 0 (67.25)
r2 r
and
1
− (ξ 2 )r − (ξ 2 )rr + {2(φ1 )zφ − (ξ 2 )zz } = 0 (67.26)
r
924
Equating finally to zero the constant term, we obtain the equation
1
(φ1 )rr + (φ1 )r + (φ1 )zz = 0 (67.27)
r
We also obtain the following expression for P r(2) Γ(∆2 )
P r(2) Γ(∆2 ) =

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂r ∂z ∂φ ∂A
∂ ∂ ∂ ∂
+ φ[r] + φ[rr] + φ[z] + φ[zz]

[r] ∂ [rr] ∂ [z] ∂ [zz] ∂ (67.28)
+A +A +A +A
1
Arr − Ar + Azz + 2(Ar φr + Az φz )
r
1 1 1
= 2 ξ Ar + 2Ar φ[r] + 2Az φ[z] + − + 2φr A[r] + A[rr]
r r
[z] [zz]
+ 2φz A + A
the explicit form of the various infinitesimal coefficients)
925
1
P r(2) Γ(∆2 ) = 2 ξ 1 Ar
r
+ 2Ar (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − (φr )2 (ξ 1 )φ − φr Ar (ξ 1 )A

2 2 2 1

+ 2Az (φ1 )z + [(φ1 )φ − (ξ 2 )z ]φz − (φz )2 (ξ 2 )φ − φr Az (ξ 1 )A

1 2 1 1
− φr φz (ξ )φ − φz Az (ξ )A − φr (ξ )z + Az (φ )A
1
+ − + 2φr (φ2 )r + [(φ2 )A − (ξ 1 )r ]Ar − (Ar )2 (ξ 1 )A
r
− φr Ar (ξ 1 )φ − Ar Az (ξ 2 )A − φr Az (ξ 2 )φ

2 2
− Az (ξ )r + φr (φ )φ

+ (φ2 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }Arz

+ {(φ2 )AA − 2(ξ 1 )rA − 2φr (ξ 1 )Aφ − Az (ξ 2 )AA }(Ar )2

2 1 2 2
+ {2(φ )rA − (ξ )rr }Ar − {(ξ )rr + 2φr (ξ )rφ }Az

+ 2φz (φ2 )z + [(φ2 )A − (ξ 2 )z ]Az − (Az )2 (ξ 2 )A − Ar Az (ξ 1 )A

1 2 1 2
926

+ (φ2 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }Arz
+ {(φ2 )A − 2(ξ 2 )z − 3Az (ξ 2 )A − 2φz (ξ 2 )φ − Ar (ξ 1 )A }Azz

+ {(φ2 )AA − 2(ξ 2 )zA − 2φz (ξ 2 )Aφ − Ar (ξ 1 )AA }(Az )2

2 2 1 1
+ {2(φ )zA − (ξ )zz }Az − {(ξ )zz + 2φz (ξ )zφ }Ar
1
In the above expression we substitute Azz by −Arr + Ar − 2(Ar φr + Az φz )
r
to account for the condition ∆2 = 0.
927
We thus see that condition P r(2) Γ(∆2 ) = 0 is equivalent to
∆2 =0

1 1
ξ Ar + 2Ar (φ1 )r + [(φ1 )φ − (ξ 1 )r ]φr − (φr )2 (ξ 1 )φ
r2

1 2 2 2 1
− φr Ar (ξ )A − φr φz (ξ )φ − φz Ar (ξ )A − φz (ξ )r + Ar (φ )A

+ 2Az (φ1 )z + [(φ1 )φ − (ξ 2 )z ]φz − (φz )2 (ξ 2 )φ − φr Az (ξ 1 )A

1 2 1 1
− φr φz (ξ )φ − φz Az (ξ )A − φr (ξ )z + Az (φ )A
1
+ − + 2φr (φ2 )r + [(φ2 )A − (ξ 1 )r ]Ar − (Ar )2 (ξ 1 )A
r
− φr Ar (ξ 1 )φ − Ar Az (ξ 2 )A − φr Az (ξ 2 )φ

2 2
− Az (ξ )r + φr (φ )φ

+ (φ2 )rr − 2{(ξ 2 )r + φr (ξ 2 )φ + Ar (ξ 2 )A }Arz

+ {(φ2 )AA − 2(ξ 1 )rA − 2φr (ξ 1 )Aφ − Az (ξ 2 )AA }(Ar )2

2 1 2 2
+ {2(φ )rA − (ξ )rr }Ar − {(ξ )rr + 2φr (ξ )rφ }Az

+ 2φz (φ2 )z + [(φ2 )A − (ξ 2 )z ]Az − (Az )2 (ξ 2 )A − Ar Az (ξ 1 )A

1 2 1 2
928

+ (φ2 )zz − 2{(ξ 1 )z + φz (ξ 1 )φ + Az (ξ 1 )A }Arz
+ {(φ2 )A − 2(ξ 2 )z − 3Az (ξ 2 )A − 2φz (ξ 2 )φ − Ar (ξ 1 )A }×

1
× −Arr + Ar − 2(Ar φr + Az φz )
r
+ {(φ2 )AA − 2(ξ 2 )zA − 2φz (ξ 2 )Aφ − Ar (ξ 1 )AA }(Az )2

2 2 1 1
+ {2(φ )zA − (ξ )zz }Az − {(ξ )zz + 2φz (ξ )zφ }Ar = 0
In the above equation we make use of the relations
(ξ 1 )φ = (ξ 1 )A = 0, (ξ 2 )φ = (ξ 2 )A = 0, (φ1 )A = 0 (67.29)
and
(φ2 )r = (φ2 )z = (φ2 )φ = 0 (67.30)
we have found previously.

Therefore the condition P r(2) Γ(∆2 ) = 0 gets simplified into
∆2 =0

1 1 1 1 1 2
ξ Ar + 2Ar (φ )r + [(φ )φ − (ξ )r ]φr − φz (ξ )r
r2

1 1 2 1
+ 2Az (φ )z + [(φ )φ − (ξ )z ]φz − φr (ξ )z
1
2 1 2
+ − + 2φr [(φ )A − (ξ )r ]Ar − Az (ξ )r
r

+ − 2(ξ 2 )r Arz + {(φ2 )A − 2(ξ 1 )r }Arr

2 2 1 2
+ (φ )AA (Ar ) − (ξ )rr Ar − (ξ )rr Az

2 2 1
+ 2φz [(φ )A − (ξ )z ]Az − Ar (ξ )z
929

+ − 2(ξ 1 )z Arz + {(φ2 )A − 2(ξ 2 )z }×
1
× −Arr + Ar − 2(Ar φr + Az φz )
r
+ (φ2 )AA (Az )2 − (ξ 2 )zz Az − (ξ 1 )zz Ar = 0
From the above equation we find that the terms (Ar )2 and (Az )2 have com-
mon coefficient (φ2 )AA . We thus get the equation
(φ2 )AA = 0 (67.31)
The coefficient of Arr is −2{(ξ 1 )r − (ξ 2 )z } which when equated to zero gives

equation (67.24). The terms Arz , Az φr and Ar φz have common coefficient
−2{(ξ 2 )r + (ξ 1 )z } which, when equated to zero, gives equation (67.21).
We thus get a further simplification

1 1 1 1 1
ξ Ar + 2Ar (φ )r + [(φ )φ − (ξ )r ]φr
r2

1 1 2
+ 2Az (φ )z + [(φ )φ − (ξ )z ]φz
1
2 1 2
+ − [(φ )A − (ξ )r ]Ar − Az (ξ )r
r

2 1 1 2
+ 2[(φ )A − (ξ )r ]φr Ar + − (ξ )rr Ar − (ξ )rr Az
+ 2[(φ2 )A − (ξ 2 )z ]φz Az
1
+ {(φ2 )A − 2(ξ 2 )z }× Ar − 2(Ar φr + Az φz )
r

2 1
− (ξ )zz Az − (ξ )zz Ar = 0
From the above equation we see that the coefficient of Ar is

1 1 1
2
ξ + 2(φ1 )r + [(ξ 1 )r − 2(ξ 2 )z ] − (ξ 1 )rr − (ξ 1 )zz = 0 (67.32)
r r
The coefficient of Az is
1
2(φ1 )z + (ξ 2 )r − {(ξ 2 )rr + (ξ 2 )zz } = 0 (67.33)
r
930
The coefficient of Ar φr is, after the necessary simplifications
(φ1 )φ + 2{(ξ 2 )z − (ξ 1 )r } = 0
from which, taking into account (67.24), we get
(φ1 )φ = 0 (67.34)
The coefficient of Az φz after the necessary simplifications is (67.34).

We solve next the determining equations. First of all, using (67.24) and
(67.21) we can prove that
(ξ 1 )rr + (ξ 1 )zz = 0 (67.35)
and
(ξ 2 )rr + (ξ 2 )zz = 0 (67.36)
Secondly, using (67.29), (67.30) and (67.34) we have the following dependence
of the various unknown coefficients ξ 1 , ξ 2 , φ1 , φ2 on the various variables
ξ 1 = ξ 1 (r, z), ξ 2 = ξ 2 (r, z) (67.37)
and
φ1 = φ1 (r, z), φ2 = φ2 (A) (67.38)
From equation (67.31), taking into account that φ2 depends on A only, we

get that
φ2 = a1 A + a2 (67.39)
From equation (67.33), taking into account equation (67.36) and substituting
(ξ 2 )r by −(ξ 1 )z (taking into account (67.21)), we obtain the equation
2r(φ1 )z − (ξ 1 )z = 0
which can also be written as (2rφ1 − ξ 1 )z = 0 and thus
2rφ1 − ξ 1 = F (r) (67.40)
931
where F (r) is a function to be determined. Equation (67.22), taking into
account (67.34), (67.39) and (67.24), is equivalent to
1 1
ξ − φ1 = a1 (67.41)
r
Solving the system of equations (67.40) and (67.41), we find
F (r)
φ1 = a1 + (67.42)
r
and
ξ 1 = 2a1 r + F (r) (67.43)
From equation (67.21), taking into account (67.43), we find (ξ 2 )r = 0 and

thus
ξ 2 = G(z) (67.44)
where G(z) is a function to be determined. From equation (67.24), taking

into account (67.43) and (67.44), we obtain the equation
2a1 + F 0 (r) − G0 (z) = 0 (67.45)
From equation (67.25), taking into account (67.24), (67.34) and (67.35), is
equivalent to
1
− ξ 1 + (ξ 2 )z = 0
r
F (r)
−2a1 − + G0 (z) = 0 (67.46)
r
F (r)
F 0 (r) − =0 (67.47)
r
The above is a first order ODE with general solution
F (r) = a3 ·r (67.48)
932
We then have from (67.46) that G0 (z) = 2a1 + a3 and by integration
G(z) = (2a1 + a3 )z + a4 (67.49)
Equation (67.44), taking into account (67.49), gives the explicit expression
of ξ 2 :
ξ 2 = (2a1 + a3 )z + a4 (67.50)
From equation (67.43), taking into account (67.48), we obtain the explicit
expression of ξ 1 :
ξ 1 = (2a1 + a3 )r (67.51)
Finally from equation (67.42), taking into account (67.48), we obtain
φ1 = a1 + a3 (67.52)
All the rest equations, i.e. (67.23), (67.26), (67.27) and (67.32) are satisfied
by the above found values of φ1 , φ2 , ξ 1 and ξ 2 .
The generator Γ of Lie symmetries is given by
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + φ1 + φ2
∂r ∂z ∂φ ∂A
∂ ∂ ∂
= [(2a1 + a3 )r] + [(2a1 + a3 )z + a4 ] + (a1 + a3 ) (67.53)
∂r ∂z ∂φ
∂
+ (a1 A + a2 )
∂A

∂ ∂ ∂ ∂ ∂
Γ = a1 2r + 2z + +A + a2
∂r ∂z ∂φ ∂A ∂A
(67.54)
∂ ∂ ∂ ∂
+ a3 r + z + + a4
∂r ∂z ∂φ ∂z
Therefore the Lie algebra of the symmetries is being spanned by the four
generators
∂ ∂ ∂ ∂ ∂
Γ1 = 2r + 2z + +A , Γ2 =
∂r ∂z ∂φ ∂A ∂A
(67.55)
∂ ∂ ∂ ∂
Γ3 = r + z + , Γ4 =
∂r ∂z ∂φ ∂z
933
We shall try next to find some solutions of Einstein’s equations associated to
these symmetries.
Solution I. Considering the generator Γ3 , the invariant surface conditions
read
rφr + zφz = 1, rAr + zAz = 0 (67.56)

z
φ = ln(r) + F (ξ), A = G(ξ), ξ= (67.57)
r
Under the substitutions (67.57) the system of equations (67.4) take the form
1
(1 + ξ 2 )F 00 (ξ) + ξ F 0 (ξ) − e2F (ξ) (1 + ξ 2 )(G0 (ξ))2 = 0 (67.58)
2
and
(1 + ξ 2 )G00 (ξ) + ξ G0 (ξ) + 2(1 + ξ 2 )G0 (ξ)F 0 (ξ) = 0 (67.59)
respectively. We then have to solve the system of equations (67.58) and

(67.59). Dividing by (1 + ξ 2 )G0 (ξ) equation (67.59), we obtain the equation
G00 (ξ) ξ
+ + 2F 0 (ξ) = 0 (67.60)
G0 (ξ) 1 + ξ2
The substitution G0 (ξ) = H(ξ) converts (67.60) into the equation
H 0 (ξ) ξ
+ + 2F 0 (ξ) = 0 (67.61)
H(ξ) 1 + ξ2
1
ln(H(ξ)) + ln(1 + ξ 2 ) + 2F (ξ) = a1 (67.62)
2
p
H(ξ) 1 + ξ 2 = K e−2F (ξ) (67.63)
where K is a positive constant. Therefore

p
G0 (ξ) 1 + ξ 2 = K e−2F (ξ) (67.64)
934
Taking into account the above expression, equation (67.58) takes on the form
1
(1 + ξ 2 )F 00 (ξ) + ξ F 0 (ξ) − K 2 e−2F (ξ) = 0 (67.65)
2
The substitution
w = e−2F (ξ) (67.66)
ξ K2
ww00 − (w0 )2 + ww 0
+ w3 = 0 (67.67)
1 + ξ2 1 + ξ2
This is a rather complicated equation which can be solved using Lie symmetry
analysis. In fact considering the infinitesimals (after Lie symmetry analysis
of (67.67))
p
X = 1 + ξ2, Y = 0 (67.68)
and going to normal coordinates {r, s} (r = r(ξ, w) and s = s(ξ, w)) by

solving the system (do not confuse r to the spatial variable)
∂r ∂r ∂s ∂s
X +Y = 0, X +Y =1 (67.69)
∂ξ ∂w ∂ξ ∂w
we get the solution
r = M (w), s = arcsinh(ξ) + N (w) (67.70)
where M and N are arbitrary functions. We take for simplicity
r=w and s = arcsinh(ξ) (67.71)
Inverting the above transformation we obtain
w=r and ξ = sinh(s(r)) (67.72)
Using the transformation (67.72), equation (67.67) takes the form
d2 s ds 2 3 ds
3
r + − K r =0 (67.73)
dr2 dr dr
935
The substitution
ds
y= (67.74)
dr
transforms (67.73) into a Bernoulli-type of equation
ry 0 (r) + y(r) = K 2 r3 (y(r))3 (67.75)

1
y=± √ (67.76)
r −2K 2 r + K1
Combining (67.76) and (67.74), we determine the function s = s(r)
s
1
s = ±2K1 arctanh K1 −2K 2 r + 2 − K2 (67.77)
K1
Since r = w and s = arcsinh(ξ) by (67.71), equation (67.77) yields

s
2
1
arcsinh(ξ) = ±2K1 arctanh K1 −2K w + 2 − K2 (67.78)
K1
Inverting the above expression, we have

q arcsinh(ξ) + K
2
−2K 2 K12 w + 1 = tanh ± (67.79)
2K1
Solving with respect to w, we obtain the solution of (67.67):
arcsinh(ξ) + K
1 2 2
w(ξ) = 2 2
1 − tanh ± (67.80)
2K K1 2K1
1 2 arcsinh(ξ) + K2

w(ξ) = sech (67.81)
2K 2 K12 2K1
with no ± sign since sech(x) is an even function.
From (67.66) we obtain, using the function w(ξ) given by (67.81)

1 1 2 arcsinh(ξ) + K2

F (ξ) = − ln sech (67.82)
2 2K 2 K12 2K1
936
Equation (67.64), because of (67.66) yields
w(ξ)
G0 (ξ) = K (67.83)
1 + ξ2
Integrating the above expression (taking into account (67.81)), results in
1 arcsinh(ξ) + K
2
G(ξ) = tanh + K3 (67.84)
KK1 2K1
The functions F (ξ) and G(ξ) given by (67.82) and (67.84) constitute the
solution of the system (67.58)-(67.59).
Therefore the functions φ and A, taking into account (67.57) and (67.82),
(67.84) are given by

1 1 2 arcsinh(z/r) + K2

φ(r, z) = ln(r) − ln sech (67.85)
2 2K 2 K12 2K1
and
1 arcsinh(z/r) + K
2
A(r, z) = tanh + K3 (67.86)
KK1 2K1
respectively. Expressions (67.85) and (67.86) constitute the solution of the
system (67.4).
We next have to determine the function µ using the PDE system (67.5).
That will complete the solution of the Einstein-Maxwell system.
In fact, using (67.85) and (67.86), we get from (67.5) the relations

1
µr = 4K12 (r2 + z 2 ) + (−r2 + z 2 )
4K12 r(r2 + z 2 )
√ arcsinh(z/r) + K (67.87)
2 2 2
− 4K1 z r + z tanh
2K1
and
1 1 arcsinh(z/r) + K
2 1 z
µz = √ tanh − 2 2
(67.88)
K1 r + z
2 2 2K1 2K1 r + z 2
Integrating the above equation with respect to z we obtain
arcsinh(z/r) + K
2 1
µ = 2 ln cosh − ln(r2 + z 2 ) + F (r) (67.89)
2K1 4K12
937
Comparing the derivative of µ (given by (67.89)) with respect to r and com-
paring to the corresponding expression (67.87), we obtain the equation
1 + 4K12 1
F 0 (r) = (67.90)
4K12 r
and by integration
1 + 4K12
F (r) = ln(r) + K4 (67.91)
4K12
Therefore we obtain the following expression for the function µ:
arcsinh(z/r) + K
2 1
µ = 2 ln cosh − ln(r2 + z 2 )
2K1 4K12
(67.92)
1 + 4K12
+ ln(r) + K4
4K12
Solution II. Considering the generator Γ4 , both functions φ and A would
depend on a single variable r:
φ = F (r), A = G(r) (67.93)
The system (67.4) then becomes
1 e2F (r) 0
F 00 (r) + F 0 (r) − (G (r))2 = 0 (67.94)
r 2r2
1
G00 (r) − G0 (r) + 2F 0 (r)G0 (r) = 0 (67.95)
r
Dividing by G0 (r) the above equation takes the form
G00 (r) 1
− + 2F 0 (r) = 0 (67.96)
G0 (r) r
Introducing the function H = G0 (r) the above equation becomes
H 0 (r) 1
− + 2F 0 (r) = 0 (67.97)
H(r) r
and by integration ln(H(r)) − ln(r) + 2F (r) = a1 , i.e.
H(r) = Kre−2F (r)
938
Therefore
G0 (r) = Kre−2F (r) (67.98)
Because of the above expression, equation (67.94) takes on the form

1 1
F 00 (r) + F 0 (r) − K 2 e−2F (r) = 0 (67.99)
r 2
The substitution
w = e−2F (r) (67.100)
transforms equation (67.99) into the equation

1
ww00 − (w0 )2 + ww0 + K 2 w3 = 0, w = w(r) (67.101)
r
This equation can be solved using Lie symmetry analysis. In fact considering
the infinitesimals (after Lie symmetry analysis of (67.101))
X = r, Y = −2w (67.102)
and going to normal coordinates {r̂, ŝ} (r̂ = r̂(r, w) and ŝ = ŝ(r, w)) by
solving the system
∂r̂ ∂r̂ ∂ŝ ∂ŝ
X +Y = 0, X +Y =1 (67.103)
∂r ∂w ∂r ∂w
we get the solution
r̂ = M (r2 w), ŝ = ln(r) + N (r2 w) (67.104)
where M and N are arbitrary functions. We take for simplicity
r̂ = r2 w, and ŝ = ln(r) (67.105)
Inverting the above transformation we obtain
r = exp(ŝ(r)) and w = r exp(−2ŝ(r)) (67.106)
Using the transformation (67.106), equation (67.101) takes the form

d2 ŝ dŝ 2 3 dŝ
3
r̂ + − K r̂ =0 (67.107)
dr̂2 dr̂ dr̂
939
The substitution
dŝ
y= (67.108)
dr̂
transforms (67.107) into a Bernoulli-type of equation
r̂y 0 (r̂) + y(r̂) = K 2 r̂3 (y(r̂))3 (67.109)

1
y=± √ (67.110)
r̂ −2K 2 r̂ + K1
Combining (67.110) and (67.108), we determine the function ŝ = ŝ(r̂)
s
1
ŝ = ±2K1 arctanh K1 −2K 2 r̂ + 2 − K2 (67.111)
K1
Since r̂ = r2 w and ŝ = ln(r) by (67.105), equation (67.111) yields

s
2 2
1
ln(r) = ±2K1 arctanh K1 −2K r w + 2 − K2 (67.112)
K1
Inverting the above expression, we have

q ln(r) + K
2
−2K 2 K12 r2 w + 1 = tanh ± (67.113)
2K1
Solving with respect to w, we obtain the solution of (67.101):
ln(r) + K
1 2 2
w(r) = 2 2 2
1 − tanh ± (67.114)
2K K1 r 2K1
1 2 ln(r) + K2

w(r) = sech (67.115)
2K 2 K12 r2 2K1
with no ± sign since sech(x) is an even function.
From (67.100) we obtain, using the function w(r) given by (67.114)

1 1 2 ln(r) + K2

F (r) = − ln sech (67.116)
2 2K 2 K12 r2 2K1
940
Equation (67.98), because of (67.100) yields
G0 (r) = Krw(r) (67.117)
Integrating the above expression (taking into account (67.115)), results in

1 ln(r) + K
2
G(r) = tanh + K3 (67.118)
KK1 2K1
The functions F (r) and G(r) given by (67.116) and (67.118) constitute the
solution of the system of equations (67.94) and (67.95). Using (67.93), we
obtain the following expressions for the functions φ and A:

1 1 2 ln(r) + K2

φ = − ln sech (67.119)
2 2K 2 K12 r2 2K1
1 ln(r) + K
2
A= tanh + K3 (67.120)
KK1 2K1
The above expressions constitute the solution of the system (67.4). We next
have to determine the function µ which satisfies the PDE system (67.5).
In fact, using
e2φ 0
µr = 2r(φ0 (r))2 + (A (r))2 (67.121)
r
we find
2 2 ln(r) + K
2
µr = + tanh (67.122)
r K1 r 2K1
and by integration
ln(r) + K
2
µ = 2 ln(r) + 4 ln cosh + K4 (67.123)
2K1
Solution III. Considering the generator Γ1 , the invariant surface conditions
read
2rφr + 2zφz = 1, 2rAr + 2zAz = A (67.124)

1 √ z
φ= ln(r) + F (ξ), A= r G(ξ), ξ= (67.125)
2 r
941
In this case the system (67.4) becomes
(1 + ξ 2 )F 00 (ξ) + ξF 0 (ξ)
(67.126)

1 2F (ξ) 1 2 0 2 0 2
− e G (ξ) − ξG(ξ)G (ξ) + (1 + ξ )(G (ξ)) = 0
2 4
1
(1 + ξ 2 )G00 (ξ) + ξG0 (ξ) − ξF 0 (ξ)G(ξ) − G(ξ)
4 (67.127)
+ 2(1 + ξ 2 )F 0 (ξ)G0 (ξ) = 0
Solution IV. Considering the generator Γ1 − Γ3 , given explicity by
∂ ∂ ∂
Γ1 − Γ3 = r +z +A (67.128)
∂r ∂z ∂A
the invariant surface conditions read
rφr + zφz = 0, rAr + zAz = A (67.129)

z
φ = F (ξ), A = r G(ξ), ξ= (67.130)
r
In this case the system (67.4) becomes
(1 + ξ 2 )F 00 (ξ) + ξF 0 (ξ)
(67.131)

1 2F (ξ) 2 0 2 0 2
− e G (ξ) − 2ξG(ξ)G (ξ) + (1 + ξ )(G (ξ)) = 0
2
(1 + ξ 2 )G00 (ξ) + ξG0 (ξ) − 2ξF 0 (ξ)G(ξ) − G(ξ)

(67.132)
+ 2(1 + ξ 2 )F 0 (ξ)G0 (ξ) = 0
68 Einstein Equations
with Axial Symmetry and Rotation
The line element corresponding to an axially symmetric space-time which
also includes rotation, appears to have the form
ds2 = f (dt − w dφ)2 − ρ2 f −1 dφ2 − eµ (dρ2 + dz 2 ) (68.1)
942
The functions f, w and µ depend on two variables ρ and z only. Einstein’s
equations associated to the above line element read
1 f4
f fρρ + fρ + fzz − fρ2 − fz2 + 2 (wρ2 + wz2 ) = 0
ρ ρ
(68.2)
1
f wρρ − wρ + wzz + 2(fρ wρ + fz wz ) = 0
ρ
where the function µ satisfies the system of PDEs
fρ 1 ρ 2 2 1 f2 2
µρ = − + 2
(f ρ − f z ) − (wρ − wz2 )
f 2f 2 ρ
(68.3)
fz ρ f2
µz = − + 2 f ρ f z − w ρ w z
f f ρ
Therefore once we can determine f and w, we would be able to determine
the third function µ solving the system (68.3).
68.I Lie Symmetry Analysis of the Field Equations.
Let us now solve the system of equations (68.2) using Lie point symmetries.
We denote by ∆1 and ∆2 the expressions
1 f4
∆1 ≡f fρρ + fρ + fzz − fρ2 − fz2 + 2 (wρ2 + wz2 )
ρ ρ
(68.4)
1
∆2 ≡f wρρ − wρ + wzz + 2(fρ wρ + fz wz )
ρ
The second prolongation in this case is defined by
∂ ∂ ∂ ∂
P r(2) Γ = ξ 1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz] (68.5)
∂fρ ∂fρρ ∂fz ∂fzz
∂ ∂ ∂ ∂
+ w[ρ] + w[ρρ] + w[z] + w[zz]
∂wρ ∂wρρ ∂wz ∂wzz
where ξ 1 , ξ 2 , φ1 and φ2 are functions each one depending on the variables
ρ, z, f and w.
Invariance of the system of equations ∆1 = 0 and ∆2 = 0 is being expressed
by the conditions
P r(2) Γ(∆1 ) = 0 and P r(2) Γ(∆2 ) =0 (68.6)

∆1 =0, ∆2 =0 ∆1 =0, ∆2 =0
943
Let us now list the various infinitesimal coefficients. They are given by
f [ρ] = (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w
(68.7)
− fρ fz (ξ 2 )f − fz wρ (ξ 2 )w − fz (ξ 2 )ρ + wρ (φ1 )w
f [z] = (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − (ξ 2 )w fz wz

(68.8)
− fρ fz (ξ 1 )f − fρ wz (ξ 1 )w − fρ (ξ 1 )z + wz (φ1 )w
w[ρ] = (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
(68.9)
− wρ wz (ξ 2 )w − fρ wz (ξ 2 )f − wz (ξ 2 )ρ + fρ (φ2 )f
w[z] = (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f
(68.10)
− wρ wz (ξ 1 )w − fz wρ (ξ 1 )f − wρ (ξ 1 )z + fz (φ2 )f
f [ρρ] = (φ1 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }fzρ
+ {(φ1 )f − 2(ξ 1 )ρ − 3fρ (ξ 1 )f − 2wρ (ξ 1 )w − fz (ξ 2 )f }fρρ
+ {(φ1 )w − fρ (ξ 1 )w − fz (ξ 2 )w }wρρ − (fρ )3 (ξ 1 )f f
+ {(φ1 )f f − 2wρ (ξ 1 )f w − 2(ξ 1 )f ρ − fz (ξ 2 )f f }(fρ )2 (68.11)
+ {(φ1 )ww − fρ (ξ 1 )ww − fz (ξ 2 )ww }(wρ )2 + 2wρ (φ1 )wρ
+ 2{(φ1 )f w − (ξ 1 )wρ }fρ wρ + {2(φ1 )f ρ − (ξ 1 )ρρ }fρ
− 2{wρ (ξ 2 )f w + (ξ 2 )f ρ }fρ fz − {(ξ 2 )ρρ + 2wρ (ξ 2 )wρ }fz
f [zz] = (φ1 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }fzρ
+ {(φ1 )f − 2(ξ 2 )z − 3fz (ξ 2 )f − 2wz (ξ 2 )w − fρ (ξ 1 )f }fzz
+ {(φ1 )w − fρ (ξ 1 )w − fz (ξ 2 )w }wzz − (fz )3 (ξ 2 )f f
+ {(φ1 )f f − 2wz (ξ 2 )f w − 2(ξ 2 )f z − fρ (ξ 1 )f f }(fz )2 (68.12)
+ {(φ1 )ww − fρ (ξ 1 )ww − fz (ξ 2 )ww }(wz )2 + 2wz (φ1 )wz
+ 2{(φ1 )f w − (ξ 2 )wz }fz wz + {2(φ1 )f z − (ξ 2 )zz }fz
− 2{wz (ξ 1 )f w + (ξ 1 )f z }fρ fz − {(ξ 1 )zz + 2wz (ξ 1 )ww }fρ
w[ρρ] = (φ2 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }wzρ
+ {(φ2 )w − 2(ξ 1 )ρ − 3wρ (ξ 1 )w − 2fρ (ξ 1 )f − wz (ξ 2 )w }wρρ
+ {(φ2 )f − wρ (ξ 1 )f − wz (ξ 2 )f }fρρ − (wρ )3 (ξ 1 )ww
+ {(φ2 )ww − 2fρ (ξ 1 )f w − 2(ξ 1 )wρ − wz (ξ 2 )ww }(wρ )2 (68.13)
+ {(φ2 )f f − wρ (ξ 1 )f f − wz (ξ 2 )f f }(fρ )2 + 2fρ (φ2 )f ρ
+ 2{(φ2 )f w − (ξ 1 )f ρ }fρ wρ + {2(φ2 )wρ − (ξ 1 )ρρ }wρ
− 2{wρ (ξ 2 )f w + (ξ 2 )f ρ }fρ wz − {(ξ 2 )ρρ + 2wρ (ξ 2 )wρ }wz
944
w[zz] = (φ2 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }wzρ
+ {(φ2 )w − 2(ξ 2 )z − 3wz (ξ 2 )w − 2fz (ξ 2 )f − wρ (ξ 1 )w }wzz
+ {(φ2 )f − wρ (ξ 1 )f − wz (ξ 2 )f }fzz − (wz )3 (ξ 2 )ww
+ {(φ2 )ww − 2fz (ξ 2 )f w − 2(ξ 2 )wz − wρ (ξ 1 )ww }(wz )2 (68.14)
+ {(φ2 )f f − wρ (ξ 1 )f f − wz (ξ 2 )f f }(fz )2 + 2fz (φ2 )f z
+ 2{(φ2 )f w − (ξ 2 )f z }fz wz + {2(φ2 )wz − (ξ 2 )zz }wz
− 2{fz (ξ 1 )f w + (ξ 1 )zw }wρ wz − {(ξ 1 )zz + 2fz (ξ 1 )zf }wρ
The expression P r(2) Γ(∆2 ) yields
P r(2) Γ(∆2 )

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz]

[ρ] ∂ [ρρ] ∂ [z] ∂ [zz] ∂
+w +w +w +w (68.15)
1
f wρρ − wρ + wzz + 2(fρ wρ + fz wz )
ρ
f 1
= 2 ξ 1 wρ + φ1 wρρ − wρ + wzz + 2wρ f [ρ] + 2wz f [z] +
ρ ρ
f
+ − + 2fρ w[ρ] + f w[ρρ] + 2fz w[z] + f w[zz]
ρ
the explicit expressions of the infinitesimal coefficients)
(2) f 1 1
1
P r Γ(∆2 ) = 2 ξ wρ + φ wρρ − wρ + wzz +
ρ ρ

+ 2wρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w

2 2 2 1
− fρ fz (ξ )f − fz wρ (ξ )w − fz (ξ )ρ + wρ (φ )w

+ 2wz (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − fz wz (ξ 2 )w

1 1 1 1
− fρ fz (ξ )f − fρ wz (ξ )w − fρ (ξ )z + wz (φ )w
945
f
+ − + 2fρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2
− wρ wz (ξ )w − fρ wz (ξ )f − wz (ξ )ρ + fρ (φ )f

+ f (φ2 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }wzρ

+ {(φ2 )ww − 2fρ (ξ 1 )f w − 2(ξ 1 )wρ − wz (ξ 2 )ww }(wρ )2

2 2 2 2
− 2{wρ (ξ )f w + (ξ )f ρ }fρ wz − {(ξ )ρρ + 2wρ (ξ )wρ }wz

+ 2fz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2
− wρ wz (ξ )w − fz wρ (ξ )f − wρ (ξ )z + fz (φ )f

+ f (φ2 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }wzρ

+ {(φ2 )ww − 2fz (ξ 2 )f w − 2(ξ 2 )wz − wρ (ξ 1 )ww }(wz )2

1 1 1 1
− 2{fz (ξ )f w + (ξ )zw }wρ wz − {(ξ )zz + 2fz (ξ )zf }wρ
(68.16)
In the above expression we substitute fzz by −fρρ − ρ−1 fρ − f 3 ρ−2 (wρ2 + wz2 ) +
f −1 (fρ2 + fz2 ) to account for the condition ∆1 = 0 and wzz by −wρρ + ρ−1 wρ −
2f −1 (fρ wρ + fz wz ) to account for the condition ∆2 = 0. We thus see that
condition P r(2) Γ(∆2 ) = 0 is equivalent to
∆1 =0, ∆2 =0
946
f 1 2
2
ξ wρ − φ1 (fρ wρ + fz wz )+
ρ f

+ 2wρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w

2 2 2 1

+ 2wz (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − fz wz (ξ 2 )w

1 1 1 1
f
+ − + 2fρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2



2 2 2 2

+ 2fz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2
947

+ {(φ2 )w − 2(ξ 2 )z − 3wz (ξ 2 )w − 2fz (ξ 2 )f − wρ (ξ 1 )w }×

× −wρρ + ρ−1 wρ − 2f −1 (fρ wρ + fz wz )
+ {(φ2 )f − wρ (ξ 1 )f − wz (ξ 2 )f }

× −fρρ − ρ−1 fρ − f 3 ρ−2 (wρ2 + wz2 ) + f −1 (fρ2 + fz2 )
− (wz )3 (ξ 2 )ww

1 1 1 1
− 2{fz (ξ )f w + (ξ )zw }wρ wz − {(ξ )zz + 2fz (ξ )zf }wρ = 0
We shall equate to zero all the coefficients of the partial derivatives of the
unknown functions f and w in the above equation. However, just for the
economy of the calculations, we shall equate to zero the easily identified
coefficients first. For example, we find that the coefficients of wρ wρρ , wρ fρ fz
and wz fρ fz are 2f (ξ 1 )w , 2(ξ 2 )f and −2(ξ 1 )f respectively. We thus have
(ξ 1 )w = (ξ 1 )f = 0, (ξ 2 )f = 0 (68.17)
The coefficient of (wz )2 fz is 2(ξ 2 )w − 2f (ξ 2 )f w and since (ξ 2 )f = 0 we get
(ξ 2 )w = 0 (68.18)
Therefore we obtain the following simplified equation

f 1 2
2
ξ wρ − φ1 (fρ wρ + fz wz )+
ρ f

1 1 1 2 1
+ 2wρ (φ )ρ + [(φ )f − (ξ )ρ ]fρ − fz (ξ )ρ + wρ (φ )w

1 1 2 1 1
+ 2wz (φ )z + [(φ )f − (ξ )z ]fz − fρ (ξ )z + wz (φ )w
f
2 2 1 2 2
+ − + 2fρ (φ )ρ + {(φ )w − (ξ )ρ }wρ − wz (ξ )ρ + fρ (φ )f
ρ
948

+ f (φ2 )ρρ − 2(ξ 2 )ρ wzρ + {(φ2 )w − 2(ξ 1 )ρ }wρρ
+ (φ2 )f fρρ + (φ2 )ww (wρ )2 + (φ2 )f f (fρ )2 + 2fρ (φ2 )f ρ

2 2 1 2
+ 2(φ )f w fρ wρ + {2(φ )wρ − (ξ )ρρ }wρ − (ξ )ρρ wz

2 2 2 1 2
+ 2fz (φ )z + {(φ )w − (ξ )z }wz − wρ (ξ )z + fz (φ )f

+ f (φ2 )zz − 2(ξ 1 )z wzρ + {(φ2 )w − 2(ξ 2 )z }

−1 −1
× −wρρ + ρ wρ − 2f (fρ wρ + fz wz )

+ (φ2 )f × −fρρ − ρ−1 fρ − f 3 ρ−2 (wρ2 + wz2 ) + f −1 (fρ2 + fz2 )
+ (φ2 )ww (wz )2 + (φ2 )f f (fz )2 + 2fz (φ2 )f z

2 2 2 1
+ 2(φ )f w fz wz + {2(φ )wz − (ξ )zz }wz − (ξ )zz wρ = 0
From the above equation we obtain the following coefficients:

The terms fρ wz and fz wρ share the common coefficient −2[(ξ 1 )z +(ξ 2 )ρ ] while
the coefficient of wzρ is −2f [(ξ 1 )z + (ξ 2 )ρ ]. We thus have the equation
(ξ 1 )z + (ξ 2 )ρ = 0 (68.19)
The coefficients of fρ wρ and fz wz yield

1 1
− φ + (φ1 )f + f (φ2 )f w − 2{(ξ 1 )ρ − (ξ 2 )z } = 0 (68.20)
f
and
1 1
− φ + (φ1 )f + f (φ2 )f w = 0 (68.21)
f
respectively. Subtracting the above two equations we get the equation
(ξ 1 )ρ − (ξ 2 )z = 0 (68.22)
Because of the above equation, equation (68.20) is identical to (68.21).

From (68.19) and (68.22) we obtain easily the equations
(ξ 1 )ρρ + (ξ 1 )zz = 0 (68.23)
949
and
(ξ 2 )ρρ + (ξ 2 )zz = 0 (68.24)
The terms (fρ )2 and (fz )2 share a common coefficient which, when equated
to zero, yields
3(φ2 )f + f (φ2 )f f = 0 (68.25)
The terms (wρ )2 and (wz )2 share the common coefficient
f4 2
2(φ1 )w + f (φ2 )ww − (φ )f = 0 (68.26)
ρ2
The coefficient of fρ yields
f 2
(φ2 )ρ − (φ )f = 0 (68.27)
ρ
The coefficient of wρ yields
f 1 1 f 1 2 f 2
ξ + 2(φ )ρ + {(ξ )ρ − (ξ )z } − (ξ )z + 2f (φ2 )wρ
ρ2 ρ ρ
− {(ξ 1 )ρρ + (ξ 1 )zz } = 0
The previous equation, because of (68.22) and (68.23), becomes
f 1 f
2
ξ + 2(φ1 )ρ − (ξ 2 )z + 2f (φ2 )wρ = 0 (68.28)
ρ ρ
The coefficient of wz , when equated to zero, gives
f 2
2(φ1 )z + (ξ )ρ + 2f (φ2 )wz − f {(ξ 2 )ρρ + (ξ 2 )zz } = 0
ρ
and because of (68.24), we obtain the equation
f 2
2(φ1 )z + (ξ )ρ + 2f (φ2 )wz = 0 (68.29)
ρ
The coefficient of wρρ , taking into account (68.23), becomes identically zero.
Similarly the coefficient of fρρ is identically zero.
The coefficient of fz yields the equation
2(φ2 )z + 2f (φ2 )f z = 0 (68.30)
950
The constant term, when equated to zero yields the equation
1
− (φ2 )ρ + (φ2 )ρρ + (φ2 )zz = 0 (68.31)
ρ
Expression P r(2) Γ(∆1 ) yields
P r(2) Γ(∆1 )

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz]

[ρ] ∂ [ρρ] ∂ [z] ∂ [zz] ∂
+w +w +w +w
1
2 2 f4 2 2
(68.32)
f fρρ + fρ + fzz − fρ − fz + 2 (wρ + wz )
ρ ρ
f 4
2f 1
= ξ 1 − 2 fρ − 3 (wρ2 + wz2 ) + φ1 fρρ + fρ + fzz
ρ ρ ρ
3
4f f
+ 2 (wρ2 + wz2 ) + − 2fρ f [ρ]
ρ ρ
2f 4 2f 4
+ f f [ρρ] − 2fz f [z] + f f [zz] + 2 wρ w[ρ] + 2 wz w[z]
ρ ρ
951
(2) 1 f
2f 4 2 2

P r Γ(∆1 ) = ξ − 2 fρ − 3 (wρ + wz )
ρ ρ
3
1
1 4f 2 2

+ φ fρρ + fρ + fzz + 2 (wρ + wz )
ρ ρ
f
+ − 2fρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w
ρ

2 2 2 1

+ f (φ1 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }fzρ

+ {(φ1 )f f − 2wρ (ξ 1 )f w − 2(ξ 1 )f ρ − fz (ξ 2 )f f }(fρ )2

2 2 2 2
− 2{wρ (ξ )f w + (ξ )f ρ }fρ fz − {(ξ )ρρ + 2wρ (ξ )wρ }fz

− 2fz (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − (ξ 2 )w fz wz

1 1 1 1

+ f (φ1 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }fzρ

+ {(φ1 )f f − 2wz (ξ 2 )f w − 2(ξ 2 )f z − fρ (ξ 1 )f f }(fz )2

1 1 1 1
− 2{wz (ξ )f w + (ξ )f z }fρ fz − {(ξ )zz + 2wz (ξ )ww }fρ
952
2f 4

+ 2 wρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2
2f 4

+ 2 wz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f
ρ

1 1 1 2
In the above expression we substitute fzz by −fρρ − ρ−1 fρ − f 3 ρ−2 (wρ2 + wz2 ) +
f −1 (fρ2 + fz2 ) to account for the condition ∆1 = 0 and wzz by −wρρ + ρ−1 wρ −
2f −1 (fρ wρ + fz wz ) to account for the condition ∆2 = 0. We thus see that
condition P r(2) Γ(∆1 ) = 0 is equivalent to
∆1 =0, ∆2 =0
f 2f 4 2 3f 3 1 2
ξ 1 − 2 fρ − (w ρ + w 2
z )+ φ1 (w 2
ρ + w 2
z ) + (f + f 2
)
ρ ρ 3 ρ2 f ρ z
f
ρ

2 2 2 1



2 2 2 2


1 1 1 1
953

+ {(φ1 )f − 2(ξ 2 )z − 3fz (ξ 2 )f − 2wz (ξ 2 )w − fρ (ξ 1 )f }×

× −fρρ − ρ−1 fρ − f 3 ρ−2 (wρ2 + wz2 ) + f −1 (fρ2 + fz2 )
+ {(φ1 )w − fρ (ξ 1 )w − fz (ξ 2 )w }

× −wρρ + ρ−1 wρ − 2f −1 (fρ wρ + fz wz )
− (fz )3 (ξ 2 )f f

1 1 1 1
2f 4

+ 2 wρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2
2f 4

+ 2 wz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f
ρ

1 1 1 2
− wρ wz (ξ )w − fz wρ (ξ )f − wρ (ξ )z + fz (φ )f = 0
The above equation, taking into account that
(ξ 1 )f = (ξ 1 )w = 0, (ξ 2 )f = (ξ 2 )w = 0 (68.33)
954
we have found previously (see (68.17) and (68.19)), takes on the form
f 2f 4 3f 3 1 2
ξ 1 − 2 fρ − 3 (wρ2 + wz2 ) + φ1 (w 2
ρ + w 2
z ) + (f + f 2
)
ρ ρ ρ2 f ρ z
f
+ − 2fρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − fz (ξ 2 )ρ + wρ (φ1 )w
ρ

+ f (φ1 )ρρ − 2(ξ 2 )ρ fzρ + {(φ1 )f − 2(ξ 1 )ρ }fρρ + (φ1 )w wρρ
+ (φ1 )f f (fρ )2 + (φ1 )ww (wρ )2 + 2wρ (φ1 )wρ

1 1 1 2
+ 2(φ )f w fρ wρ + {2(φ )f ρ − (ξ )ρρ }fρ − (ξ )ρρ fz

1 1 2 1 1
− 2fz (φ )z + [(φ )f − (ξ )z ]fz − fρ (ξ )z + wz (φ )w

+ f (φ1 )zz − 2(ξ 1 )z fzρ
+ {(φ1 )f − 2(ξ 2 )z }×

−1 3 −2 2 2 −1 2 2
× −fρρ − ρ fρ − f ρ (wρ + wz ) + f (fρ + fz )

+ (φ1 )w × −wρρ + ρ−1 wρ − 2f −1 (fρ wρ + fz wz )
+ (φ1 )f f (fz )2 + (φ1 )ww (wz )2 + 2wz (φ1 )wz

1 1 2 1
+ 2(φ )f w fz wz + {2(φ )f z − (ξ )zz }fz − (ξ )zz fρ
2f 4

2 2 1 2 2
+ 2 wρ (φ )ρ + {(φ )w − (ξ )ρ }wρ − wz (ξ )ρ + fρ (φ )f
ρ
2f 4

2 2 2 1 2
+ 2 wz (φ )z + {(φ )w − (ξ )z }wz − wρ (ξ )z + fz (φ )f = 0
ρ
The coefficients of fρ fz and fzρ are 2{(ξ 2 )ρ + (ξ 1 )z } and −2{(ξ 2 )ρ + (ξ 1 )z }
respectively and thus, when equated to zero, yield equation (68.19) we have
found previously. Similarly the coefficient of wρ wz yields (68.19).
The coefficient of (fρ )2 is given by
1 1
φ − (φ1 )f + 2{(ξ 1 )ρ − (ξ 2 )z } + f (φ1 )f f = 0
f
955
and because of (68.22), we get
1 1
φ − (φ1 )f + f (φ1 )f f = 0 (68.34)
f
The coefficient of (fz )2 is also given by (68.34).
The coefficient of (wρ )2 is given by
2f 4 1 3f 3 1 1 f4
− ξ + φ + f (φ )ww + {2(φ2 )w − (φ1 )f }
ρ3 ρ2 ρ2
2f 4
− 2 {(ξ 1 )ρ − (ξ 2 )z } = 0
ρ
and because of (68.22), we get the equation
2f 4 1 3f 3 1 1 f4
− 3
ξ + 2
φ + f (φ )ww + 2
{2(φ2 )w − (φ1 )f } = 0 (68.35)
ρ ρ ρ
The coefficient of (wz )2 is also given by (68.36). The terms wρ fρ and fz wz
share the same coefficient
2f 4 2
−4(φ1 )w + 2(φ1 )wf + (φ )f = 0 (68.36)
ρ2
The coefficient of fρρ is −2f {(ξ 1 )ρ −(ξ 2 )z } which, because of (68.22), vanishes
identically. The coefficient of wρρ is identically zero.
The coefficient of wz yields the equation
2f 4 2
2(φ1 )wz + (φ )z = 0 (68.37)
ρ2
The coefficient of wρ gives
2f 1 1 2f 4 2
(φ )w + 2f (φ )wρ + 2 (φ )ρ = 0 (68.38)
ρ ρ
The coefficient of fz , when equated to zero, yields the equation
f
− (ξ 2 )ρ − 2(φ1 )z − f (ξ 2 )ρρ + f {2(φ1 )zf − (ξ 2 )zz } = 0
ρ
f
− (ξ 2 )ρ − 2(φ1 )z + 2f (φ1 )zf = 0 (68.39)
ρ
956
The coefficient of fρ , when equated to zero, yields the equation
f 1 1 1 f 2
− ξ − 2(φ ) ρ + 2f (φ )f ρ + (ξ )z
ρ2 ρ
f
− {(ξ 1 )ρ − (ξ 2 )z } − f {(ξ 1 )ρρ + (ξ 1 )zz } = 0
ρ
and because of (68.22) and (68.23), gets simplified into the equation
f 1 f
− 2
ξ − 2(φ1 )ρ + 2f (φ1 )f ρ + (ξ 2 )z = 0 (68.40)
ρ ρ
From the constant term we obtain
1 1
(φ )ρ + (φ1 )ρρ + (φ1 )zz = 0 (68.41)
ρ
We shall determine next the solution of the determining equations. First of
all, equation (68.34) is a PDE which contains one function only and admits
the general solution
φ1 = F (ρ, z, w)f + G(ρ, z, w)f · ln(f ) (68.42)
Equations (68.25), (68.27) and (68.30) is a system of PDEs containing φ2 as

the unknown function. The general solution of this system is given by
1
φ2 = H(w) + J(w) (68.43)
ρ2 f 2
However the function φ2 satisfies equation (68.31) as well. Substituting φ2
given by (68.43) into (68.31) we obtain that J(w) = 0. Therefore we have
the following general expression for φ2 :
φ2 = H(w) (68.44)
Substituting φ1 and φ2 given by (68.42) and (68.44) respectively into (68.21)

we get G(ρ, z, w) = 0. Therefore we have the following expression for φ1 :
φ1 = F (ρ, z, w)f (68.45)
From equation (68.36), taking into account (68.44) and (68.45) we get Fw = 0
and thus F = F (ρ, z). Therefore we have the following expression for φ1 :
φ1 = F (ρ, z)f (68.46)
957
Equations (68.37) and (68.38) are satisfied identically for φ1 and φ2 given by
(68.46) and (68.44) respectively.
(ξ 2 )ρ = 0 and taking into account (68.33), we see that
ξ 2 = ξ 2 (z) (68.47)
Adding equations (68.28) and (68.40) we get the equation (φ1 )f ρ +(φ2 )wρ = 0
from which, taking into account (68.46) and (68.44), we obtain Fρ = 0, i.e.
F = F (z) and thus
φ1 = F (z)f (68.48)
Therefore equations (68.28) and (68.40) are equivalent to the equation
f 1 f 2
− ξ + (ξ )z = 0
ρ2 ρ
i.e. to the equation
ξ 1 = ρ (ξ 2 )z (68.49)
H 00 (w) = 0, i.e. H(w) = a1 w + a2 and thus
φ2 = a1 w + a2 (68.50)
the equation
ξ 1 = ρ (a1 + F (z)) (68.51)
Taking into account (68.47) we obtain from (68.19) that (ξ 1 )z = 0, and

because of (68.33),
ξ 1 = ξ 1 (ρ) (68.52)
Since ξ 1 = ξ 1 (ρ) we conclude from (68.51) that F (z) should be a constant,

which we call a3 :
F (z) = a3 (68.53)
958
Therefore, taking into account (68.51), we get
ξ 1 = ρ (a1 + a3 ) (68.54)
Using (68.48) and (68.53), we see that
φ1 = a3 f (68.55)
Using (68.22) and (68.54), we get the equation
(ξ 2 )z = a1 + a3 (68.56)
and by integration, taking into account (68.47),
ξ 2 = (a1 + a3 )z + a4 (68.57)
The above expression of ξ 2 is compatible to equation (68.49).
We have thus found the following expressions for ξ 1 , ξ 2 and φ1 , φ2
ξ 1 = (a1 + a3 ) ρ, ξ 2 = (a1 + a3 )z + a4
(68.58)
φ1 = a3 f, φ2 = a1 w + a2
The generator Γ of Lie symmetries is thus given by
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂
= (a1 + a3 )ρ + [(a1 + a3 )z + a4 ] (68.59)
∂ρ ∂z
∂ ∂
+ a3 f + (a1 w + a2 )
∂f ∂w

∂ ∂ ∂ ∂
Γ = a1 ρ + z +w + a2
∂ρ ∂z ∂w ∂w
(68.60)
∂ ∂ ∂ ∂
+ a3 ρ + z +f + a4
∂ρ ∂z ∂f ∂z
Therefore the Lie algebra of symmetries of the system of differential equations
is being spanned by the four generators
∂ ∂ ∂ ∂
Γ1 = ρ +z +w , Γ2 = ,
∂ρ ∂z ∂w ∂w
(68.61)
∂ ∂ ∂ ∂
Γ3 = ρ + z + f , Γ4 =
∂ρ ∂z ∂f ∂z
959
68.II Simplified Version of the Field Equations. In the previous sub-
section we considered the system of equations (68.2). However this system
can be further simplified. We write again this system for ease of reference.
1 f4
f fρρ + fρ + fzz − fρ2 − fz2 + 2 (wρ2 + wz2 ) = 0
ρ ρ
(68.62)
1
f wρρ − wρ + wzz + 2(fρ wρ + fz wz ) = 0
ρ
We first see that the second equation of the above system can be written as
(ρ−1 f 2 wρ )ρ + (ρ−1 f 2 wz )z = 0 (68.63)
Therefore there exists a function u such that
uρ = −ρ−1 f 2 wz and uz = ρ−1 f 2 wρ (68.64)
The original system (68.62) then takes the form

1
f fρρ + fρ + fzz − fρ2 − fz2 + u2ρ + u2z = 0
ρ
(68.65)
1
f uρρ + uρ + uzz − 2(fρ uρ + fz uz ) = 0
ρ
The first equation of the above system comes from the substitutions
ρ ρ
wz = − uρ and wρ = uz (68.66)
f2 f2
into the first equation of (68.62), while the second is the compatibility con-
dition uρz = uzρ taking into account the two expressions of (68.64).
We shal change notation temporarily considering in place of the function u
the function w, thus to avoid a number of changes, and at the end of the
calculations we shall change notation again. In other words we shall consider
instead of (68.65) the system
1
f fρρ + fρ + fzz − fρ2 − fz2 + wρ2 + wz2 = 0
ρ
(68.67)
1
f wρρ + wρ + wzz − 2(fρ wρ + fz wz ) = 0
ρ
960
Let us now solve the system of equations (68.67) using Lie point symmetries.
We denote by ∆ ˜ 1 and ∆˜ 2 the expressions
˜ 1 ≡f fρρ + 1 fρ + fzz − fρ2 − fz2 + wρ2 + wz2

∆
ρ
(68.68)
˜
1
∆2 ≡f wρρ + wρ + wzz − 2(fρ wρ + fz wz ) = 0
ρ
The second prolongation in this case is defined by
∂ ∂ ∂ ∂
P r(2) Γ = ξ 1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz] (68.69)
∂ ∂ ∂ ∂
+ w[ρ] + w[ρρ] + w[z] + w[zz]
where ξ 1 , ξ 2 , φ1 and φ2 are functions each one depending on the variables

ρ, z, f and w.
Invariance of the system of equations ∆ ˜ 1 = 0 and ∆
˜ 2 = 0 is being expressed
by the conditions
˜ 1)
P r(2) Γ(∆ ˜ 2)
= 0 and P r(2) Γ(∆ =0 (68.70)
˜ 1 =0, ∆
∆ ˜ 2 =0 ˜ 1 =0, ∆
∆ ˜ 2 =0
The various infinitesimal coefficients are given by
f [ρ] = (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w

(68.71)
− fρ fz (ξ 2 )f − fz wρ (ξ 2 )w − fz (ξ 2 )ρ + wρ (φ1 )w
f [z] = (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − (ξ 2 )w fz wz

(68.72)
− fρ fz (ξ 1 )f − fρ wz (ξ 1 )w − fρ (ξ 1 )z + wz (φ1 )w
w[ρ] = (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
(68.73)
− wρ wz (ξ 2 )w − fρ wz (ξ 2 )f − wz (ξ 2 )ρ + fρ (φ2 )f
w[z] = (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f
(68.74)
− wρ wz (ξ 1 )w − fz wρ (ξ 1 )f − wρ (ξ 1 )z + fz (φ2 )f
961
f [ρρ] = (φ1 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }fzρ
+ {(φ1 )f f − 2wρ (ξ 1 )f w − 2(ξ 1 )f ρ − fz (ξ 2 )f f }(fρ )2 (68.75)
− 2{wρ (ξ 2 )f w + (ξ 2 )f ρ }fρ fz − {(ξ 2 )ρρ + 2wρ (ξ 2 )wρ }fz
f [zz] = (φ1 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }fzρ
+ {(φ1 )f f − 2wz (ξ 2 )f w − 2(ξ 2 )f z − fρ (ξ 1 )f f }(fz )2 (68.76)
− 2{wz (ξ 1 )f w + (ξ 1 )f z }fρ fz − {(ξ 1 )zz + 2wz (ξ 1 )ww }fρ
w[ρρ] = (φ2 )ρρ − 2{(ξ 2 )ρ + fρ (ξ 2 )f + wρ (ξ 2 )w }wzρ
+ {(φ2 )ww − 2fρ (ξ 1 )f w − 2(ξ 1 )wρ − wz (ξ 2 )ww }(wρ )2 (68.77)
− 2{wρ (ξ 2 )f w + (ξ 2 )f ρ }fρ wz − {(ξ 2 )ρρ + 2wρ (ξ 2 )wρ }wz
w[zz] = (φ2 )zz − 2{(ξ 1 )z + fz (ξ 1 )f + wz (ξ 1 )w }wzρ
+ {(φ2 )ww − 2fz (ξ 2 )f w − 2(ξ 2 )wz − wρ (ξ 1 )ww }(wz )2 (68.78)
− 2{fz (ξ 1 )f w + (ξ 1 )zw }wρ wz − {(ξ 1 )zz + 2fz (ξ 1 )zf }wρ
962
˜ 2 ) yields
The expression P r(2) Γ(∆
P r(2) Γ(∆˜ 2)

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz]

[ρ] ∂ [ρρ] ∂ [z] ∂ [zz] ∂
+w +w +w +w (68.79)
1
f wρρ + wρ + wzz − 2(fρ wρ + fz wz )
ρ
f 1 1
= − 2 ξ wρ + φ wρρ + wρ + wzz − 2wρ f [ρ] − 2wz f [z] +
1
ρ ρ
f
+ − 2fρ w[ρ] + f w[ρρ] − 2fz w[z] + f w[zz]
ρ
˜ 2 ) = − f ξ 1 wρ + φ1 wρρ + 1 wρ + wzz +

P r(2) Γ(∆
ρ2 ρ

− 2wρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w

2 2 2 1

− 2wz (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − fz wz (ξ 2 )w

1 1 1 1
963
f
+ − 2fρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2



2 2 2 2
(68.80)
− 2fz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2



1 1 1 1
− 2{fz (ξ )f w + (ξ )zw }wρ wz − {(ξ )zz + 2fz (ξ )zf }wρ
In the above expression we substitute fzz by −fρρ − ρ−1 fρ − f −1 (wρ2 + wz2 ) +

˜ 1 = 0 and wzz by −wρρ − ρ−1 wρ +
f −1 (fρ2 + fz2 ) to account for the condition ∆
2f −1 (fρ wρ + fz wz ) to account for the condition ∆ ˜ 2 = 0. We thus see that
˜ 2)
condition P r(2) Γ(∆ = 0 is equivalent to
˜ 1 =0, ∆
∆ ˜ 2 =0
964
f 1 2
− 2
ξ wρ + φ1 (fρ wρ + fz wz )+
ρ f

− 2wρ (φ1 )ρ + [(φ1 )f − (ξ 1 )ρ ]fρ − (ξ 1 )f (fρ )2 − fρ wρ (ξ 1 )w

2 2 2 1

− 2wz (φ1 )z + [(φ1 )f − (ξ 2 )z ]fz − (ξ 2 )f (fz )2 − fz wz (ξ 2 )w

1 1 1 1
f
+ − 2fρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f
ρ

2 2 2 2



2 2 2 2

− 2fz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2
965

+ {(φ2 )w − 2(ξ 2 )z − 3wz (ξ 2 )w − 2fz (ξ 2 )f − wρ (ξ 1 )w }×

× −wρρ − ρ−1 wρ + 2f −1 (fρ wρ + fz wz )
+ {(φ2 )f − wρ (ξ 1 )f − wz (ξ 2 )f }

× −fρρ − ρ−1 fρ − f −1 (wρ2 + wz2 ) + f −1 (fρ2 + fz2 )
− (wz )3 (ξ 2 )ww

1 1 1 1
− 2{fz (ξ )f w + (ξ )zw }wρ wz − {(ξ )zz + 2fz (ξ )zf }wρ = 0
We shall equate to zero all the coefficients of the partial derivatives of the
unknown functions f and w in the above equation. However, just for the
economy of the calculations, we shall equate to zero the easily identified
coefficients first. For example, we find that the coefficients of wρ wρρ , wρ fρ fz
and wz fρ fz are −2f (ξ 1 )w , 2(ξ 2 )f and 2(ξ 1 )f respectively. We thus have
(ξ 1 )w = (ξ 1 )f = 0, (ξ 2 )f = 0 (68.81)
The coefficient of (wz )2 fz is 4(ξ 2 )w − 2f (ξ 2 )f w and since (ξ 2 )f = 0 we get
(ξ 2 )w = 0 (68.82)
966
Therefore we obtain the following simplified equation
f 1 2
− 2
ξ wρ + φ1 (fρ wρ + fz wz )+
ρ f

1 1 1 2 1
− 2wρ (φ )ρ + [(φ )f − (ξ )ρ ]fρ − fz (ξ )ρ + wρ (φ )w

1 1 2 1 1
− 2wz (φ )z + [(φ )f − (ξ )z ]fz − fρ (ξ )z + wz (φ )w
f
2 2 1 2 2
+ − 2fρ (φ )ρ + {(φ )w − (ξ )ρ }wρ − wz (ξ )ρ + fρ (φ )f
ρ

+ f (φ2 )ρρ − 2(ξ 2 )ρ wzρ + {(φ2 )w − 2(ξ 1 )ρ }wρρ
+ (φ2 )f fρρ + (φ2 )ww (wρ )2 + (φ2 )f f (fρ )2 + 2(φ2 )f ρ fρ

2 2 1 2
+ 2(φ )f w fρ wρ + {2(φ )wρ − (ξ )ρρ }wρ − (ξ )ρρ wz

2 2 2 1 2
− 2fz (φ )z + {(φ )w − (ξ )z }wz − (ξ )z wρ + (φ )f fz

+ f (φ2 )zz − 2(ξ 1 )z wzρ
+ {(φ2 )w − 2(ξ 2 )z }×


+ (φ2 )f × −fρρ − ρ−1 fρ − f −1 (wρ2 + wz2 ) + f −1 (fρ2 + fz2 )
+ (φ2 )ww (wz )2 + (φ2 )f f (fz )2 + 2(φ2 )f z fz

2 2 2 1
+ 2(φ )f w fz wz + {2(φ )wz − (ξ )zz }wz − (ξ )zz wρ = 0
From the above equation we obtain the following coefficients:

The terms fρ wz and fz wρ share the common coefficient 2[(ξ 1 )z + (ξ 2 )ρ ] while
the coefficient of wzρ is −2f [(ξ 1 )z + (ξ 2 )ρ ]. We thus have the equation
(ξ 1 )z + (ξ 2 )ρ = 0 (68.83)
The coefficients of fρ wρ and fz wz yield
1 1
φ − 2{(φ1 )f + (φ2 )w } + 4(ξ 1 )ρ + 2f (φ2 )f w = 0 (68.84)
f
967
and
1 1
φ − 2{(φ1 )f + (φ2 )w } + 4(ξ 2 )z + 2f (φ2 )f w = 0 (68.85)
f
respectively. Subtracting the above two equations we get the equation
(ξ 1 )ρ − (ξ 2 )z = 0 (68.86)
Because of the above equation, equation (68.84) is identical to (68.85).

From (68.83) and (68.86) we obtain easily the equations
(ξ 1 )ρρ + (ξ 1 )zz = 0 (68.87)
and
(ξ 2 )ρρ + (ξ 2 )zz = 0 (68.88)
The terms (fρ )2 and (fz )2 share a common coefficient which, when equated
to zero, yields
−(φ2 )f + f (φ2 )f f = 0 (68.89)
The terms (wρ )2 and (wz )2 share the common coefficient
−2(φ1 )w − (φ2 )f + f (φ2 )ww = 0 (68.90)
The term fρ has coefficient which when equated to zero yields the equation
−2(φ2 )ρ + 2f (φ2 )f ρ = 0 (68.91)
The coefficient of wρ yields
f 1 f f
− 2
ξ − 2(φ1 )ρ − {(ξ 1 )ρ − (ξ 2 )z } + (ξ 2 )z + 2f (φ2 )wρ
ρ ρ ρ
1 1
− f {(ξ )ρρ + (ξ )zz } = 0
The previous equation, because of (68.86) and (68.87), becomes
f 1 1 f 2
− 2
ξ − 2(φ ) ρ + (ξ )z + 2f (φ2 )wρ = 0 (68.92)
ρ ρ
968
The coefficient of wz , when equated to zero, gives
f 2
− 2(φ1 )z − (ξ )ρ + 2f (φ2 )wz − f {(ξ 2 )ρρ + (ξ 2 )zz } = 0
ρ
and because of (68.88), we obtain the equation
f 2
− 2(φ1 )z − (ξ )ρ + 2f (φ2 )wz = 0 (68.93)
ρ
The coefficient of wρρ , taking into account (68.86), becomes identically zero.
Similarly the coefficient of fρρ is identically zero.
The coefficient of fz yields the equation
−2(φ2 )z + 2f (φ2 )f z = 0 (68.94)
The constant term, when equated to zero yields the equation

1 2
(φ )ρ + (φ2 )ρρ + (φ2 )zz = 0 (68.95)
ρ
˜ 1 ) yields
Expression P r(2) Γ(∆
P r(2) Γ(∆˜ 1)

∂ ∂ ∂ ∂
= ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂w
∂ ∂ ∂ ∂
+ f [ρ] + f [ρρ] + f [z] + f [zz]

[ρ] ∂ [ρρ] ∂ [z] ∂ [zz] ∂ (68.96)
+w +w +w +w
1
f fρρ + fρ + fzz − (fρ2 + fz2 ) + (wρ2 + wz2 )
ρ
f 1 f
1 1
= ξ − 2 fρ + φ fρρ + fρ + fzz + − 2fρ f [ρ]
ρ ρ ρ
+ f f [ρρ] − 2fz f [z] + f f [zz] + 2wρ w[ρ] + 2wz w[z]
969
˜ 1 ) = ξ 1 − f fρ + φ1 fρρ + 1 fρ + fzz

P r(2) Γ(∆
ρ2 ρ
f
ρ

2 2 2 1



2 2 2 2


1 1 1 1



1 1 1 1
970

+ 2wρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f

2 2 2 2

+ 2wz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2
In the above expression we substitute fzz by −fρρ − ρ−1 fρ − f −1 (wρ2 + wz2 ) +

˜ 1 = 0 and wzz by −wρρ − ρ−1 wρ +
f −1 (fρ2 + fz2 ) to account for the condition ∆
2f −1 (fρ wρ + fz wz ) to account for the condition ∆ ˜ 2 = 0. We thus see that
˜ 1)
condition P r(2) Γ(∆ = 0 is equivalent to
˜ 1 =0, ∆
∆ ˜ 2 =0
f 1 1
ξ 1 − 2 fρ + φ1 − (wρ2 + wz2 ) + (fρ2 + fz2 )
ρ f f
f
ρ

2 2 2 1



2 2 2 2


1 1 1 1
971

+ {(φ1 )f − 2(ξ 2 )z − 3fz (ξ 2 )f − 2wz (ξ 2 )w − fρ (ξ 1 )f }×

+ {(φ1 )w − fρ (ξ 1 )w − fz (ξ 2 )w }

− (fz )3 (ξ 2 )f f

1 1 1 1

+ 2wρ (φ2 )ρ + {(φ2 )w − (ξ 1 )ρ }wρ − (wρ )2 (ξ 1 )w − fρ wρ (ξ 1 )f

2 2 2 2

+ 2wz (φ2 )z + {(φ2 )w − (ξ 2 )z }wz − (wz )2 (ξ 2 )w − fz wz (ξ 2 )f

1 1 1 2
− wρ wz (ξ )w − fz wρ (ξ )f − wρ (ξ )z + fz (φ )f = 0
The above equation, taking into account that
(ξ 1 )f = (ξ 1 )w = 0, (ξ 2 )f = (ξ 2 )w = 0 (68.97)
972
we have found previously (see (68.81) and (68.82)), takes on the form
f 1 1
ξ 1 − 2 fρ + φ1 − (wρ2 + wz2 ) + (fρ2 + fz2 )
ρ f f
f
1 1 1 2 1
+ − 2fρ (φ )ρ + [(φ )f − (ξ )ρ ]fρ − fz (ξ )ρ + wρ (φ )w
ρ

+ f (φ1 )ρρ − 2(ξ 2 )ρ fzρ + {(φ1 )f − 2(ξ 1 )ρ }fρρ + (φ1 )w wρρ
+ (φ1 )f f (fρ )2 + (φ1 )ww (wρ )2 + 2wρ (φ1 )wρ

1 1 1 2
+ 2(φ )f w fρ wρ + {2(φ )f ρ − (ξ )ρρ }fρ − (ξ )ρρ fz

1 1 2 1 1
− 2fz (φ )z + [(φ )f − (ξ )z ]fz − fρ (ξ )z + wz (φ )w

+ f (φ1 )zz − 2(ξ 1 )z fzρ
+ {(φ1 )f − 2(ξ 2 )z }×


1 −1 −1
+ (φ )w × −wρρ − ρ wρ + 2f (fρ wρ + fz wz )
+ (φ1 )f f (fz )2 + (φ1 )ww (wz )2 + 2wz (φ1 )wz

1 1 2 1
+ 2(φ )f w fz wz + {2(φ )f z − (ξ )zz }fz − (ξ )zz fρ

2 2 1 2 2
+ 2wρ (φ )ρ + {(φ )w − (ξ )ρ }wρ − wz (ξ )ρ + fρ (φ )f

2 2 2 1 2
+ 2wz (φ )z + {(φ )w − (ξ )z }wz − wρ (ξ )z + fz (φ )f = 0
The coefficients of fρ fz and fzρ are 2{(ξ 2 )ρ + (ξ 1 )z } and −2{(ξ 2 )ρ + (ξ 1 )z }

respectively and thus, when equated to zero, yield equation (68.83) we have
found previously. Similarly the coefficient of wρ wz yields (68.83).
The coefficient of (fρ )2 is given by
1 1
φ − (φ1 )f + 2{(ξ 1 )ρ − (ξ 2 )z } + f (φ1 )f f = 0
f
973
and because of (68.86), we get
1 1
φ − (φ1 )f + f (φ1 )f f = 0 (68.98)
f
The coefficient of (fz )2 is also given by (68.98).

The coefficient of (wρ )2 is given by
1 1
− φ + f (φ1 )ww + 2(φ2 )w − (φ1 )f
f
− 2{(ξ 1 )ρ − (ξ 2 )z } = 0
and because of (68.86), we get the equation

1 1
− φ + f (φ1 )ww + 2(φ2 )w − (φ1 )f = 0 (68.99)
f
The coefficient of (wz )2 is also given by (68.99). The terms wρ fρ and fz wz

share the same coefficient
2f (φ1 )wf + 2(φ2 )f = 0 (68.100)
The coefficient of fρρ is −2f {(ξ 1 )ρ −(ξ 2 )z } which, because of (68.86), vanishes
identically. The coefficient of wρρ is identically zero.
The coefficient of wz yields the equation
2f (φ1 )wz + 2(φ2 )z = 0 (68.101)
The coefficient of wρ gives
2f (φ1 )wρ + 2(φ2 )ρ = 0 (68.102)
The coefficient of fz , when equated to zero, yields the equation

f
− (ξ 2 )ρ − 2(φ1 )z − f (ξ 2 )ρρ + f {2(φ1 )zf − (ξ 2 )zz } = 0
ρ
f
− (ξ 2 )ρ − 2(φ1 )z + 2f (φ1 )zf = 0 (68.103)
ρ
974
The coefficient of fρ , when equated to zero, yields the equation
f 1 f
− 2
ξ − 2(φ1 )ρ + 2f (φ1 )f ρ + (ξ 2 )z
ρ ρ
f
− {(ξ 1 )ρ − (ξ 2 )z } − f {(ξ 1 )ρρ + (ξ 1 )zz } = 0
ρ
and because of (68.86) and (68.87), gets simplified into the equation
f 1 1 1 f 2
− ξ − 2(φ ) ρ + 2f (φ )f ρ + (ξ )z = 0 (68.104)
ρ2 ρ
From the constant term we obtain
1 1
(φ )ρ + (φ1 )ρρ + (φ1 )zz = 0 (68.105)
ρ
We shall determine next the solution of the determining equations. First of
all, equation (68.98) is a PDE which contains one function only and admits
the general solution
φ1 = F (ρ, z, w)f + G(ρ, z, w)f · ln(f ) (68.106)
Equations (68.89), (68.91) and (68.94) is a system of PDEs containing φ2 as

the unknown function. The general solution of this system is given by
φ2 = H(w) + f 2 J(w) (68.107)
Equation (68.100), because of (68.106) and (68.107), takes the form
(Fw + Gw + 2J)f + Gw ln(f ) = 0
Gw = 0 and Fw + 2J = 0 (68.108)
On the other hand, equation (68.90) yields (thanks to (68.106) and (68.107))
Jf 2 + (H 00 (w) − 2Fw − 2Jw ) = 0 (68.109)
from which we get J = 0 and H 00 (w) − 2Fw − 2Jw = 0. Therefore the second
of (68.108) yields Fw = 0 and thus H 00 (w) = 0. Therefore
H(w) = a1 w + a2 (68.110)
975
Since J = 0, comparing (68.107) and (68.110) we get
φ2 = a1 w + a2 (68.111)
For Fw = 0 and Gw = 0, we obtain from (68.106) that
φ1 = F (ρ, z)f + G(ρ, z)f · ln(f ) (68.112)
Equations (68.101) and (68.102) become identities because of (68.111) and

(68.112).
From equation (68.92) we obtain using (68.111) and (68.112)
1 1 1 2
− ξ − 2F ρ + (ξ )z − 2Gρ ln(f ) = 0
ρ2 ρ
The above equation yields the system of equations
1 1 1
− 2
ξ − 2Fρ + (ξ 2 )z = 0 (68.113)
ρ ρ
and
Gρ = 0 (68.114)
From equation (68.93) we obtain using (68.111) and (68.112)

1
(ξ 2 )ρ + 2Fz + 2Gz ln(f ) = 0
ρ
The above equation yields the system of equations
1 2
(ξ )ρ + 2Fz = 0 (68.115)
ρ
and
Gz = 0 (68.116)
Equation (68.103) becomes because of (68.112)

1
− (ξ 2 )ρ + 2Gz = 0
ρ
976
and since Gz = 0 because of (68.116), we get that (ξ 2 )ρ = 0, i.e.
ξ 2 = ξ 2 (z) (68.117)
Equation (68.104) becomes because of (68.112)

1 1 1
− 2
ξ + 2Gρ + (ξ 2 )z = 0
ρ ρ
and since Gρ because of (68.114), we get the equation
1 1 1 2
− ξ + (ξ )z = 0 (68.118)
ρ2 ρ
For Gρ = Gz = 0, we have
G(ρ, z) = a4 (68.119)
From equation (68.115) since (ξ 2 )ρ = 0, we get Fz = 0, i.e.
F = F (ρ) (68.120)
We thus obtain from (68.112) the following expression of φ1
φ1 = F (ρ)f + a4 f · ln(f ) (68.121)
Comparing (68.113) and (68.118) we get Fρ = 0, i.e. F = a3 and thus
φ1 = a3 f + a4 f · ln(f ) (68.122)
Equation (68.118), since (ξ 2 )z = (ξ 1 )ρ = 0 because of (68.86), takes on the

form
1
(ξ 1 )ρ − ξ 1 = 0 (68.123)
ρ
ξ 1 = a5 ρ (68.124)
Using the above expression of ξ 1 in (68.86), we obtain the equation (ξ 2 )z = a5

and by integration with respect to z, we get
ξ 2 = a5 z + a6 (68.125)
977
We have thus found the following expressions for ξ 1 , ξ 2 and φ1 , φ2
ξ 1 = a5 ρ, ξ 2 = a5 z + a6
(68.126)
φ1 = a3 f + a4 f · ln(f ), φ2 = a1 w + a2
Changing notation from w to u, we see that the infinitasimals of the Lie

algebra of symmetries for the system (68.65) is
ξ 1 = a5 ρ, ξ 2 = a5 z + a6
(68.127)
φ1 = a3 f + a4 f · ln(f ), φ2 = a1 u + a2
The generator Γ of Lie symmetries for the system (68.65) is thus given by
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + φ1 + φ2
∂ρ ∂z ∂f ∂u
∂ ∂
= a5 ρ + (a5 z + a6 ) (68.128)
∂ρ ∂z
∂ ∂
+ {a3 f + a4 f · ln(f )} + (a1 u + a2 )
∂f ∂u
∂ ∂ ∂ ∂
Γ = a1 u + a2 + a3 f + a4 f ln(f )
∂u ∂u ∂f ∂f
(68.129)
∂ ∂ ∂
+ a5 ρ + z + a6
∂ρ ∂z ∂z
Therefore the Lie algebra of symmetries of the system of differential equations

(68.65) is being spanned by the four generators
∂ ∂ ∂ ∂
Γ1 = u , Γ2 = , Γ3 = f , Γ4 = f ln(f )
∂u ∂u ∂f ∂f
(68.130)
∂ ∂ ∂
Γ5 = ρ + z , Γ6 =
∂ρ ∂z ∂z
Solutions. Let us now try to find some general solutions of the system of
equations (68.65). The invariant surface conditions
ξ 1 fρ + ξ 2 fz = φ 1 , ξ 1 uρ + ξ 2 uz = φ2 (68.131)
978
take the form in view of (68.127),
a5 ρfρ + (a5 z + a6 )fz = a3 f + a4 f ln(f ),
(68.132)
a5 ρuρ + (a5 z + a6 )uz = a1 w + a2
We then have to find the invariants corresponding to the original system by
determining the general solution of the above system.
We consider the case a3 = a1 and a4 = 0. We then obtain in place of (68.132)
the system
a5 ρfρ + (a5 z + a6 )fz = a1 f
(68.133)
a5 ρuρ + (a5 z + a6 )uz = a1 u + a2
The above system can also be written as (dividing through by a5 )
ρfρ + (z + a)fz = kf
(68.134)
ρuρ + (z + a)uz = ku + m
where
a6 a1 a2
a= , k= , m= (68.135)
a5 a5 a5
Solving the system (68.134), we obtain the general solution
m z+a
f = ρk F (ξ), u=− + ρk G(ξ), where ξ = (68.136)
k ρ
We thus have determined the three invariants ξ and F, G of the system
(68.65). Substitution of f and u given by (68.136) into (68.65), we obtain
the system
(G0 (ξ))2 + {kG(ξ) − ξG0 (ξ)}2 − (F 0 (ξ))2 − {kF (ξ) − ξF 0 (ξ)}2
(68.137)
+ F (ξ){k 2 F (ξ) + (1 − 2k)ξF 0 (ξ) + (1 + ξ 2 )F 00 (ξ)} = 0
− 2F 0 (ξ)G0 (ξ) − 2{kG(ξ) − ξG0 (ξ)}{kF (ξ) − ξF 0 (ξ)}
(68.138)
+ F (ξ){k 2 G(ξ) + (1 − 2k)ξG0 (ξ) + (1 + ξ 2 )G00 (ξ)} = 0
We shall only consider the case k = 1. Other choices might also be possible.
However the choice k = 0 should be excluded since this case is being prohib-
ited by the appearance of the term −m/k in the expression for u.
For k = 1 equations (68.137) and (68.138) become
− G0 (ξ) [ξG(ξ) − (1 + ξ 2 )G0 (ξ)] + G(ξ) [G(ξ) − ξG0 (ξ)]
(68.139)
+ F 0 (ξ) [ξF (ξ) − (1 + ξ 2 )F 0 (ξ)] + (1 + ξ 2 )F (ξ)F 00 (ξ) = 0
979
and
2F 0 (ξ) [ξG(ξ) − (1 + ξ 2 )G0 (ξ)]
(68.140)
+ F (ξ) [G(ξ) − ξG0 (ξ) − (1 + ξ 2 )G0 (ξ)] = 0
respectively. Considering (see also Bhutani, Singh and Kalra [1])
ξG(ξ) − (1 + ξ 2 )G0 (ξ) = 0 (68.141)
we see that equation (68.140) is automatically satisfied. The general solution

of (68.141) is given by
p
G(ξ) = K 1 + ξ 2 (68.142)
Because of the above expression of G, equation (68.139) takes on the form
d2 F (ξ) dF (ξ) 2 ξF (ξ) dF (ξ) K2

F (ξ) − + + =0 (68.143)
dξ 2 dξ 1 + ξ 2 dξ 1 + ξ2
A Lie-symmetry analysis of (68.143) reveals that the above equation admits
the infinitesimals
p p
[X = 1 + ξ 2 , Y = 0], [X = 1 + ξ 2 arcsinh(ξ), Y = F ] (68.144)
Case I. Going over to normal coordinates (r, s) by solving the system

∂r ∂r
X +Y =0
∂ξ ∂F
(68.145)
∂s ∂s
X +Y =1
∂ξ ∂F
p
for X = 1 + ξ 2 , Y = 0, i.e.
p ∂r ∂r
1 + ξ2 + 0· =0
∂ξ ∂F
(68.146)
p ∂s ∂s
1 + ξ2 + 0· =1
∂ξ ∂F
we find that
r = H1 (F ) and s = arcsinh(ξ) + H2 (F ) (68.147)
980
where H1 and H2 are arbitrary functions. We take for simplicity
r=F and s = arcsinh(ξ) (68.148)
Let us now consider the inverse of the above transformation:
ξ = sinh(s(r)), F (ξ) = r (68.149)
d2 s ds ds 3
r 2+ − K2 =0 (68.150)
dr dr dr
The above equation can be converted into a Bernoulli differential equation
ds
under the substitution H = . We thus obtain the general solution of
dr
(68.150)
r
s(r) + B = A arctan √ (68.151)
A2 K 2 − r2
where A, B are constants. Inverting the above equation results in
r s(r) + B
√ = tan (68.152)
A2 K 2 − r 2 A
Going to the variables (ξ, F ) according to (68.148), the above expression
takes on the form
F (ξ) arcsinh(ξ) + B
p = tan (68.153)
A2 K 2 − F 2 (ξ) A
The above equation can be solved with respect to F (ξ), providing the
following expression
arcsinh(ξ) + B
F (ξ) = ±AK sin (68.154)
A
Case II. Going
p over to normal coordinates (r, s) by solving the system (68.145)
for X = 1 + ξ 2 arcsinh(ξ), Y = F , i.e.
p ∂r ∂r
1 + ξ 2 arcsinh(ξ) +F =0
∂ξ ∂F
(68.155)
p ∂s ∂s
1 + ξ 2 arcsinh(ξ) +F =1
∂ξ ∂F
981
we find that
F F
r = H1 and s = ln[arcsinh(ξ)] + H2 (68.156)
arcsin(ξ) arcsin(ξ)
where H1 and H2 are arbitrary functions. We take for simplicity
F
r= and s = ln[arcsinh(ξ)] (68.157)
arcsin(ξ)
Let us now consider the inverse of the above transformation:
ξ = sinh(es(r) ), F (ξ) = r·es(r) (68.158)
d2 s ds 2 ds
2 2
ds 3
r 2 +r + + (r − K ) =0 (68.159)
dr dr dr dr
The above equation can be converted into an Abel’s differential equation
ds
under the substitution H = . Remarkably enough, Abel’s equation admits
dr
a closed-form solution. The general solution of (68.159) is now given by
s(r)
res(r)
e + B = A arctan p (68.160)
A2 K 2 − (res(r) )2
where A, B are constants. Inverting the above equation results in
res(r) es(r) + B
p = tan (68.161)
A2 K 2 − (res(r) )2 A
Going to the variables (ξ, F ) according to (68.158), the above expression

takes on the form
F (ξ) arcsinh(ξ) + B
p = tan (68.162)
A2 K 2 − F 2 (ξ) A
The above equation can be solved with respect to F (ξ), providing the
following expression
arcsinh(ξ) + B
F (ξ) = ±AK sin (68.163)
A
982
We thus see that either choice of the infinitesimals (68.144) leads to the same
solution of (68.143).
Conclusion I. For k = 1 the solution of the system (68.137)-(68.138) is
given by (68.142) and (68.154):
p arcsinh(ξ) + B
G(ξ) = K 1 + ξ 2 , F (ξ) = ±AK sin (68.164)
A
Taking into account (68.136) we arrive at our
Conclusion II. For k = 1 we obtain
arcsinh(ξ) + B
f (ρ, z) = ±AKρ sin
A (68.165)
m p
u(ρ, z) = − + Kρ 1 + ξ 2
k
where
z+a
ξ= (68.166)
ρ
Using (68.66) and (68.165), we obtain (taking the plus sign of f )
z+a
1 1 arcsinh( ) + B
2 ρ
wz = − 2 p csec
A K ρ2 + (z + a)2 A
z+a (68.167)
arcsinh( ) + B
1 z+a 2

ρ
wρ = 2 p csec
A K ρ ρ2 + (z + a)2 A
There is only one function left, the function µ to be determined. This function
satisfies the system of equations (68.3). This system, using (68.167) and the
first of (68.165), takes on the form
1 A2 [ρ2 + (z + a)2 ] + (z + a)2 − ρ2
µρ = − ·
2A2 ρ[ρ2 + (z + a)2 ]
(68.168)
1 z+a
µz = 2 · 2
A ρ + (z + a)2
The above system admits the solution
1 ρ2 + (z + a)2 1
µ(ρ, z) = · ln − ln(ρ) (68.169)
2A2 ρ 2
983
References.
[1] N. Goyal and R. K. Gupta: ”On Symmetries and Exact Solutions
of Einstein Vacuum Equations for Axially Symmetric Gravitational Fields”.
Int. J. Math. Comp. Sci. 6 (2012) 838-841
[2] N. Goyal and R. K. Gupta: ”New Exact Solutions of the Einstein-
Maxwell Equations for Magnetostatic Fields”.
Chin. Phys. B 21 (2012) 090401
[3] O. P. Bhutani, K. Singh and D. K. Kalra: ”On certain classes of
exact solutions of Einstein equations for rotating fields in conventional and
non conventional form”. Int. J. Eng. Sci. 41 (2003) 769-786
Chapter XXII
Differential Geometry
69 The Minimal Surface Equation
The mean curvature K of a surface u = u(x, y) is given by the expression
(1 + u2y )uxx − 2ux uy uxy + (1 + u2x )uyy
K= (69.1)
2(1 + u2x + u2y )3/2
The equation K = 0 gives rise to a nonlinear PDE
(1 + u2y )uxx − 2ux uy uxy + (1 + u2x )uyy = 0 (69.2)
which is called the minimal surface equation.
P r(2) Γ(∆)|∆=0 = 0 (69.3)
where
∆≡(1 + u2y )uxx − 2ux uy uxy + (1 + u2x )uyy (69.4)
∂ ∂ ∂
P r(2) Γ = X +Y +U
∂x ∂y ∂u
(69.5)
[x] ∂ [y] ∂ [xy] ∂ [xx] ∂ [yy] ∂
+U +U +U +U +U
∂ux ∂uy ∂uxy ∂uxx ∂uyy
984
We have the following expressions for the various infinitesimals (details of
the calculations are given in Section 19)
U [x] = Ux + (Uu − Xx )ux − Xu (ux )2 − Yx uy − Yu ux uy (69.6)
U [y] = Uy + (Uu − Yy )uy − Yu (uy )2 − Xy ux − Xu ux uy (69.7)

U [xx] = Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3
− Xyy ux − 2Xyu ux uy − Xuu (uy )2 ux + (Uu − 2Yy )uyy (69.9)
U [xy] = Uxy + (Uyu − Xxy )ux + (Uxu − Yxy )uy
− Xy uxx − Yx uyy − Xuu (ux )2 uy − Yuu ux (uy )2 (69.10)
− Xyu (ux )2 − Yxu (uy )2 − 2Xu ux uxy − 2Yu uy uxy
− Xu uy uxx − Yu ux uyy
Therefore P r(2) Γ(∆) is equivalent to
P r(2) Γ(∆) = (2ux uyy − 2uy uxy )U [x] + (2uy uxx − 2ux uxy )U [y]
(69.11)
− 2ux uy U [xy] + (1 + u2y )U [xx] + (1 + u2x )U [yy]
985
and using the explicit expressions of the various infinitesimals, we obtain
P r(2) Γ(∆) =
= (2ux uyy − 2uy uxy )[Ux + (Uu − Xx )ux − Xu (ux )2 − Yx uy − Yu ux uy ]
+ (2uy uxx − 2ux uxy )[Uy + (Uu − Yy )uy − Yu (uy )2 − Xy ux − Xu ux uy ]

− 2ux uy Uxy + (Uyu − Xxy )ux + (Uxu − Yxy )uy

− Xy uxx − Yx uyy − Xuu (ux )2 uy − Yuu ux (uy )2
− Xyu (ux )2 − Yxu (uy )2 − 2Xu ux uxy − 2Yu uy uxy


+ (1 + uy ) Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3
2


+ (1 + ux ) Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3
2

(69.12)
(1 + u2y )uxx + (1 + u2x )uyy

In the above expression we substitute uxy by to
2ux uy
account for the condition ∆ = 0. We thus see that P r(2) Γ(∆)|∆=0 = 0 is
equivalent to
986
2ux uyy [Ux + (Uu − Xx )ux − Xu (ux )2 − Yx uy − Yu ux uy ]
(1 + u2y )uxx + (1 + u2x )uyy
− 2uy ×
2ux uy
×[Ux + (Uu − Xx )ux − Xu (ux )2 − Yx uy − Yu ux uy ]
+ 2uy uxx [Uy + (Uu − Yy )uy − Yu (uy )2 − Xy ux − Xu ux uy ]
(1 + u2y )uxx + (1 + u2x )uyy
− 2ux ×
2ux uy
×[Uy + (Uu − Yy )uy − Yu (uy )2 − Xy ux − Xu ux uy ]

− 2ux uy Uxy + (Uyu − Xxy )ux + (Uxu − Yxy )uy
(1 + u2y )uxx + (1 + u2x )uyy

+ (Uu − Xx − Yy )
2ux uy
+ (Uuu − Xxu − Yyu )ux uy
− Xy uxx − Yx uyy − Xuu (ux )2 uy − Yuu ux (uy )2
− Xyu (ux )2 − Yxu (uy )2
(1 + u2 )uxx + (1 + u2 )uyy
y x
(69.13)
− 2(Xu ux + Yu uy )
2ux uy


2
+ (1 + uy ) Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3

− 3Xu ux uxx − Yu uy uxx
(1 + u2 )uxx + (1 + u2 )uyy
y x
− 2(Yx + Yu ux )
2ux uy

+ (1 + u2x ) Uyy + (2Uyu − Yyy )uy + (Uuu − 2Yyu )(uy )2 − Yuu (uy )3

− 3Yu uy uyy − Xu ux uyy
(1 + u2 )uxx + (1 + u2 )uyy
y x
− 2(Xy + Xu uy ) =0
2ux uy
987
Upon expanding and arranging all the terms properly, we obtain
− (Xy + Yx )uxx − (Xy + Yx )uyy − (Xy + Yx )uxx u2y − (Xy + Yx )u2x uyy
+ (Uxx + Uyy )ux uy + (2Uxu − Xxx − Xyy )u2x uy + (2Uyu − Yxx − Yyy )u2y ux
+ (Uuu + Uyy − 2Xxu )u3x uy + (2Xxy − Yuu − Yyy )u3x u2y
+ (2Yxy − Xuu − Xxx )u2x u3y − (Yu + Uy )u3x uyy − (Yu + Uy )ux uyy
+ (Yu + Uy )ux uxx u2y − (Yu + Uy )ux uxx − (Ux + Xu )uxx uy
+ (Ux + Xu )uyy uy u2x − (Ux + Xu )uxx u3y − (Ux + Xu )uy uyy
+ 2(Xx − Uu )ux uy uyy + 2(Yy − Uu )ux uy uxx − 2(Xyu + Yxu + Uxy )u2x u2y
+ (Uxx + Uuu − 2Yyu )ux u3y − (Yxx + Yuu )ux u4y − (Xyy + Xuu )uy u4x = 0
(69.14)
of the function u. Equation (69.14) contains 24 terms in total. However
only 15 of them are different each other. We thus obtain the following set of
equations
Xy + Yx = 0 (69.15)
Uxx + Uyy = 0 (69.16)

2Uxu − Xxx − Xyy = 0 (69.17)
2Uyu − Yxx − Yyy = 0 (69.18)
Uuu + Uyy − 2Xxu = 0 (69.19)
2Xxy − Yuu − Yyy = 0 (69.20)
2Yxy − Xuu − Xxx = 0 (69.21)
Yu + Uy = 0 (69.22)
Ux + Xu = 0 (69.23)
Xx − Uu = 0 (69.24)
Yy − U u = 0 (69.25)
Xyu + Yxu + Uxy = 0 (69.26)
Uxx + Uuu − 2Yyu = 0 (69.27)
988
Yxx + Yuu = 0 (69.28)
Xyy + Xuu = 0 (69.29)
Let us now try to solve the determining equations (67.15)-(67.29).
Upon differentiation of equation (69.22) with respect to y and (69.25) with
respect to u and subtracting the resulting equations we obtain Uuu +Uyy = 0.
Because of this equation, we obtain from (69.19) that
Xxu = 0 (69.30)
Differentiation of (69.23) with respect to x, we get Uxx +Xxu = 0 and because

of (69.30) we get
Uxx = 0 (69.31)
Because of the previous equation, we obtain from (69.16) the equation
Uyy = 0 (69.32)
Differentiation of (69.22) with respect to y and (69.25) with respect to u and

subtracting the resulting equations, we get
Uuu = 0 (69.33)
The system of equations (69.31)-(69.33) admits the solution
U = [(a1 y + a2 )x + a3 y + a4 ]u + (a5 y + a6 )x + a7 y + a8 (69.34)
Differentiation of equation (69.23) upon with respect to u and (69.24) with

respect to x and adding the resulting equations, we obtain Xuu + Xxx = 0.
Because of this equation, we get from (69.21) the equation
Yxy = 0 (69.35)
From equation (69.27), because of (69.31) and (69.33) we obtain
Yyu = 0 (69.36)
Differentiation of (69.15) with respect to y gives Xyy + Yxy = 0 and because

of (69.35), we obtain
Xyy = 0 (69.37)
989
Because of the above equation, we get from (69.29) that
Xuu = 0 (69.38)
The system of equations (69.30), (69.37) and (69.38) admits a general solution
given by
X = F (x) + G(x)y + (a9 y + a10 )u (69.39)
Differentiation of (69.15) with respect to u gives Xyu + Yxu = 0 and using

(69.39) we obtain the equation
Yxu = −a9 (69.40)
The system of equations (69.35), (69.36) and (69.40) admits the general
solution
Y = H(x) + K(y) + L(u) − a9 xu (69.41)
Equations (69.34), (69.39) and (69.41) provide the general expressions for the
infinitesimals U, X and Y respectively. We shall now calculate the various
functions which appear in the above expressions.
Equation (69.29), using these expressions of X, Y and U takes on the form
a1 u + a5 = 0 from which we get that
a1 = 0 and a5 = 0 (69.42)
Equation (69.23) also gives (taking into account that a1 = a5 = 0) a2 u +

a9 y + a6 + a10 = 0 from which we obtain that
a2 = 0, a9 = 0 and a10 = −a6 (69.43)
Therefore we have the following expressions for X, Y and U :

X = F (x) + G(x)y − a6 u
Y = H(x) + K(y) + L(u) (69.44)
U = (a3 y + a4 )u + a6 x + a7 y + a8
Using the above expressions, equation (69.20) takes on the form L00 (u) +
K 00 (y) = 0 and then L00 (u) = 0 and K 00 (y) = 0 which means that
L(u) = b1 u + b2 and K(y) = c1 y + c2 (69.45)
990
Equation (69.21) gives F 00 (x) + G00 (x)y = 0 and then F 00 (x) = 0 and G00 (x) =
0 which means that
F (x) = B1 x + B2 and G(x) = C1 x + C2 (69.46)
Equation (69.28) becomes H 00 (x) + L00 (u) = 0 and thus H 00 (x) = 0 and
L00 (u) = 0 which means that
H(x) = A1 x + A2 (69.47)
since we have already taken into account L00 (u) = 0 by the first expression
of (69.45).
Equation (69.25) becomes K 0 (y) − a3 y − a4 = 0 which can be written as
c1 − a3 y − a4 = 0 from which we get
a3 = 0 and c1 = a4 (69.48)
Let us now write down again the expressions for X, Y and U in the light of
(69.45)-(69.48):
X = B1 x + B2 + (C1 x + C2 )y − a6 u
Y = A1 x + A2 + a4 y + c2 + b1 u + b2 (69.49)
U = a4 u + a6 x + a7 y + a8
Equation (69.22) takes the form b1 + a7 = 0 from which we get
b1 = −a7 (69.50)
Equation (69.15) transforms into C1 x + C2 + A1 = 0 from which we get
C1 = 0 and C2 = −A1 (69.51)
Equation (69.24) becomes B1 − a4 = 0 and thus
B1 = a4 (69.52)
Equations (69.16), (69.17), (69.18), (69.19), (69.27) and (69.29) are satisfied
automatically. Therefore the expressions (69.49) take the form
X = a4 x − A1 y − a6 u + B2
Y = A1 x + a4 y − a7 u + A2 + b2 + c2 (69.53)
U = a6 x + a7 y + a4 u + a8
991
Using the assignments
A1 −→a2 , B2 −→a3 , A2 + b2 + c2 −→a5 , a8 −→a1 (69.54)
we obtain the final expressions for X, Y and U :
X = a4 x − a2 y − a6 u + a3
Y = a2 x + a4 y − a7 u + a5 (69.55)
U = a6 x + a7 y + a4 u + a1
The generator of the Lie symmetries is thus given by

∂ ∂ ∂
Γ=X +Y +U
∂x ∂y ∂u
∂ ∂
= (a4 x − a2 y − a6 u + a3 ) + (a2 x + a4 y − a7 u + a5 ) (69.56)
∂x ∂y
∂
+ (a6 x + a7 y + a4 u + a1 )
∂u
∂ ∂ ∂ ∂ ∂ ∂
Γ = a1 + a2 −y +x + a3 + a4 x +y
∂u ∂x ∂y ∂x ∂x ∂y
(69.57)
∂ ∂ ∂ ∂ ∂
+ a5 + a6 −u +x + a7 −u +y
∂y ∂x ∂u ∂y ∂u
We thus see that the generators of the Lie algebra of symmetries for the
Minimal Surface PDE are
∂ ∂ ∂ ∂
X1 = , X 2 = −y + x , X3 =
∂u ∂x ∂y ∂x
∂ ∂ ∂ ∂ ∂
X4 = x + y , X5 = , X 6 = −u +x (69.58)
∂x ∂y ∂y ∂x ∂u
∂ ∂
X 7 = −u +y
∂y ∂u
There is a discrepancy between our calculations and those found by N. Bı̂lǎ
∂
(see Reference below). In fact in X 4 there is the additional term u which
∂u
does not appear in our own calculations. This is supported using MAPLE
or Mathematica software.
992
Let us now find a general solution of the Minimal Surface PDE using the
combination X = X 1 + X 2 , i.e.
∂ ∂ ∂
X = −y +x + (69.59)
∂x ∂y ∂u
The auxiliary Largange system corresponding to the above generator X has
the form
dx dy du
= = (69.60)
−y x 1
dx dy
The first equation, i.e. = , written as xdx + ydx = 0, admits the
−y x
general solution x2 + y 2 = C 2 , i.e.
p
x 2 + y 2 = C1 (69.61)
dy du dy du
The second equation = , written as p 2 = after integration
x 1 C1 − y 2 1
gives
y y
C2 + arctan p 2 = u or C2 = u − arctan (69.62)
C1 − y 2 x
Using (69.61) and the second of (69.62), we can introduce the two invariants,
named ξ and U by
p y
ξ = x2 + y 2 and U (ξ) = u − arctan (69.63)
x
The original PDE (69.2) can thus be transformed into the nonlinear ODE
ξ(ξ 2 + 1)U 00 (ξ) + (ξ 2 + 2)U 0 (ξ) + ξ 2 (U 0 (ξ))3 = 0 (69.64)
If U 0 (ξ) = 0, i.e. U (ξ) = a, then we obtain from (69.63) the solution

y
u(x, y) = arctan +a (69.65)
x
which is the equation of a helicoid.
If U 0 (ξ)6=0, the substitution
Y (ξ)
U 0 (ξ) = (69.66)
ξ
993
transforms (69.64) into
ξ(ξ 2 + 1)Y 0 (ξ) + Y (ξ) = −Y 3 (ξ) (69.67)
This is a Bernoulli differential equation with general solution

s
ξ2 + 1
Y (ξ) = , A>0 (69.68)
Aξ 2 − 1
Therefore after integration of

s
1 ξ2 + 1
U 0 (ξ) =
ξ Aξ 2 − 1
we obtain
1 2Aξ 2 + A − 1 + 2H
U (ξ) = √ ln √
2 A 2 A (69.69)
1 (A − 1)ξ 2 − 2
+ arctan +B
2 H
where
p
H= (ξ 2 + 1)(Aξ 2 − 1) (69.70)
We thus have the following solution of the Minimal Surface PDE (after sub-
stitution of ξ 2 by x2 + y 2 )
1 2A(x2 + y 2 ) + A − 1 + 2p(x2 + y 2 + 1)(A(x2 + y 2 ) − 1)

u(x, y) = √ ln √
2 A 2 A
1 (A − 1)(x2 + y 2 ) − 2 y
+ arctan p + arctan +B
2 (x2 + y 2 + 1)(A(x2 + y 2 ) − 1) x
(69.71)
The above solution corresponds to a helicoidal surface.

Using MAPLE. We are going to determine the symmetries of the Minimal
Surface Equation using MAPLE.
>restart;
994
>with(DEtools):
>with(PDEtools):
> u := U (x, y);
The system responds by U (x, y)
> mse := (1 + (diff(u, y))2 ) ∗ diff(u, x, x) − 2 ∗ diff(u, x) ∗ diff(u, y) ∗ diff(u, x, y)

+ (1 + (diff(u, x))2 ) ∗ diff(u, y, y) = 0;

∂ 2 ∂ 2 ∂ ∂ ∂ 2
1+ U (x, y) U (x, y) − 2 U (x, y) U (x, y) U (x, y)
∂y ∂x2 ∂x ∂y ∂y∂y
∂ 2 ∂ 2
+ 1+ U (x, y) U (x, y) = 0
∂x ∂y 2
>DetSys:=DeterminingPDE(mse); The program responds
∂ ∂ ∂ ∂
{ ξ
− 1 (x, y, U ) = ξ
− 2 (x, y, U ), ξ
− 1 (x, y, U ) = − ξ
− 2 (x, y, U ) ,
∂x ∂y ∂y ∂x
∂2 ∂2 ∂2
− 1ξ (x, y, U ) = 0, ξ
− 2 (x, y, U ) = 0, − ξ2 (x, y, U ) = 0,
∂U 2 ∂U 2 ∂x2
∂2 ∂ ∂
2 − ξ2 (x, y, U ) = 0, − η1 (x, t, U ) = − ξ2 (x, y, U ),
∂y ∂U ∂y
∂ ∂ ∂ ∂
η
− 1 (x, y, U ) = − ξ
− 1 (x, y, U ) , η
− 1 (x, y, U ) = − ξ
− 2 (x, y, U ) ,
∂x ∂U ∂y ∂U
∂2 ∂2 ∂2
− ξ2 (x, y, U ) = 0, − ξ2 (x, y, U ) = 0, − ξ2 (x, y, U ) = 0}
∂x∂U ∂y∂U ∂y∂x
>Inf1:=pdsolve(DetSys); the program responds
{ − ξ1 (x, y, U ) = − C2x − − C1y + − C5U + − C6, − ξ2 (x, y, U ) = − C1x

+ − C4U + − C2y + − C3, − η1 (x, y, U ) = − C2U − − C5x − − C4y + − C7

fact, using
995
Inf2:=Infinitesimals(mse); the program responds
[− ξ1 (x, y, U ) = 0,− ξ2 (x, y, U ) = 1, − η1 (x, y, U ) = 0],

[− ξ1 (x, y, U ) = 1,− ξ2 (x, y, U ) = 0, − η1 (x, y, U ) = 0],
[− ξ1 (x, y, U ) = 0,− ξ2 (x, y, U ) = 0, − η1 (x, y, U ) = 1],
[− ξ1 (x, y, U ) = −y,− ξ2 (x, y, U ) = x, − η1 (x, y, U ) = 0]
[− ξ1 (x, y, U ) = x,− ξ2 (x, y, U ) = y, − η1 (x, y, U ) = U ]
[− ξ1 (x, y, U ) = 0,− ξ2 (x, y, U ) = U, − η1 (x, y, U ) = −y]
[− ξ1 (x, y, U ) = U,− ξ2 (x, y, U ) = 0, − η1 (x, y, U ) = −x]
References.
[1] T. H. Colding and W. P. Minicozzi II: ”A Course in Minimal
Surfaces”. AMS 2011
[2] R. Osserman: ”A Survey of Minimal Surfaces”. Dover 2002
[3] U. Dierkes, S. Hildebrandt and F. Sauvigny: ”Minimal Surfaces”.
Springer 2010
[4] N. Bı̂lǎ: ”Lie Groups Applications to Minimal Surfaces PDE”.
Proceedings of the Workshop on Global Analysis, Differential Geometry
and Lie Algebras, Gr. Tsagas (Ed.), Balkan Society of Geometers, 1996.
Chapter XXIII
Financial Applications
70 The Black-Scholes Equation
The Black-Scholes equation was introduced by Black and Scholes
F. Black and M. Scholes: ”The Pricing of Options and Corporate Liabil-
ities”. Journal of Political Economy 81 (1973) 637-659,
in an attempt to pricing European Options.
The Black-Scholes equation has the form
1
ut + σ 2 x2 uxx + rxux − ru = 0, u = u(x, t) (70.1)
2
996
where σ is the volatility and r is the interest rate (both considered to be
constant). The derivation of this equation is based on Itô calculus using the
arbitrage principle.
71.1. Lie Symmetries. We shall determine the Lie symmetries of this
equation. Let ∆(x, t, u(2) ) be defined by the expression
1
∆(x, t, u(2) )≡ut + σ 2 x2 uxx + rxux − ru (70.2)
2
We introduce the vector field Γ (the generator of symmetries) by
∂ ∂ ∂
Γ = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u) (70.3)
∂x ∂t ∂u
The second prolongation in our case has the form
∂ ∂ ∂ ∂ ∂
P r(2) Γ = Γ + φx + φt + φxx + φxt + φtt (70.4)
∂ux ∂ut ∂uxx ∂uxt ∂utt
Invariance of the BS-equation under infinitesimal transformations is being
expressed by the condition
P r(2) Γ[∆(x, t, u(2) )] = 0 as long as ∆(x, t, u(2) ) = 0 (70.5)
The above condition can be written as

1
ξ(σ 2 xuxx + rux ) − rφ + φt + rxφx + σ 2 x2 φxx = 0 (70.6)
2
We have the following expressions for the coefficients φt , φx and φxx :
φt = φt − ξt ux + (φu − τt )ut − ξu ux ut − τu u2t (70.7)
φx = φx − τx ut + (φu − ξx )ux − τu ux ut − ξu u2x (70.8)

φxx = φxx + (2φxu − ξxx )ux − τxx ut + (φuu − 2ξxu )u2x
− 2τxu ux ut − ξuu u3x − τuu u2x ut + (φu − 2ξx )uxx (70.9)
− 2τx uxt − 3ξu ux uxx − τu ut uxx − 2τu ux uxt
The functions τ , ξ and φ have the following dependence on the corresponding
variables
τ = τ (t), ξ = ξ(x, t), φ = φ(x, t, u) (70.10)
997
Using the expressions (70.7)-(70.10), equation (70.6) takes on the form
ξ(σ 2 xuxx + rux ) − rφ + [φt − ξt ux + (φu − τt )ut ]

1
+ rx[φx + (φu − ξx )ux ] + σ 2 x2 [φxx + (2φxu − ξxx )ux + φuu u2x (70.11)
2
+ (φu − 2ξx )uxx ] = 0
1
We substitute in the above equation ut by − σ 2 x2 uxx − rxux + ru to account
2
for the condition ∆ = 0:

2 1 2 2
ξ(σ xuxx + rux ) − rφ + φt − ξt ux + (φu − τt ) − σ x uxx − rxux + ru
2
1
+ rx[φx + (φu − ξx )ux ] + σ 2 x2 [φxx + (2φxu − ξxx )ux + φuu u2x
2
+ (φu − 2ξx )uxx ] = 0
function u(x, t) in the above equation. We find
Coefficient of u2x :
1 2 2
σ x φuu = 0 (70.12)
2
Coefficient of ux :
1
−ξt − rx(φu − τt ) + rξ + rx(φu − ξx ) + σ 2 x2 (2φxu − ξxx ) = 0 (70.13)
2
1 1
− σ 2 x2 (φu − τt ) + σ 2 xξ + σ 2 x2 (φu − 2ξx ) = 0 (70.14)
2 2
Coefficient of zeroth-order derivative
1
φt + ru(φu − τt ) − rφ + rxφx + σ 2 x2 φxx = 0 (70.15)
2
Let us try now to solve the determining equations (70.12)-(70.15). From
equation (70.12) we obtain φuu = 0 and then
φ(x, t, u) = α(x, t)u + β(x, t) (70.16)
998
1 1
ξx − ξ = τt (70.17)
x 2
The above equation admits the solution
1
ξ(x, t) = (x· ln(x))τt + x·ρ(t) (70.18)
2
where ρ(t) is a function to be determined. From equation (70.13), taking
into account (70.16) and (70.18), we get the equation
σ 2 − 2r 1 1 ln(x) 1
σ 2 αx = τt + τtt + ρ0 (t) (70.19)
4 x 2 x x
Integrating the above equation with respect to x, we obtain
σ 2 − 2r 1
2
σ α(x, t) = τt ln(x) + τtt (ln(x))2 + ρ0 (t) ln(x) + θ(t) (70.20)
4 4
where θ(t) is function to be determined. From equation (70.15), using (70.16),

1 2 2
at − rτt + rxαx + σ x αxx u
2 (70.21)
1 2 2
+ βt − rβ + rxβx + σ x βxx = 0
2
Equating to zero the coefficients of u, we obtain the two equations
1
at − rτt + rxαx + σ 2 x2 αxx = 0 (70.22)
2
and
1
βt − rβ + rxβx + σ 2 x2 βxx = 0 (70.23)
2
The above equation expresses the fact that β(x, t) is any solution of the
original BS-equation. Equation (70.22) can be written as
1
(ln(x))2 τttt + 4 ln(x)ρ00 (t) + 4θ0 (t) = (σ 2 + 2r)2 τt − σ 2 τtt
2 (70.24)
2 0
+ 2(σ − 2r)ρ (t)
999
taking into account (70.20). The above equation is being satisfied if
τttt = 0, ρ00 (t) = 0 (70.25)
and
1
4θ0 (t) = (σ 2 + 2r)2 τt − σ 2 τtt + 2(σ 2 − 2r)ρ0 (t) (70.26)
2
From (70.25) we get
τ (t) = a1 + a2 t + a3 t2 (70.27)
and
ρ(t) = a4 + a5 t (70.28)
1 1 1
θ0 (t) = (σ 2 + 2r)2 (a2 + 2a3 t) − σ 2 a3 + (σ 2 − 2r)a5 (70.29)
8 2 2
and by integration
1 1 1
θ(t) = (σ 2 + 2r)2 (a2 t + a3 t2 ) − σ 2 a3 t + (σ 2 − 2r)a5 t + a6 (70.30)
8 2 2
Therefore equation (70.20) gives
1 1 1
α(x, t) = 2
(ln(x))2 a3 + 2 (σ 2 − 2r)(a2 + 2a3 t) + 2 (ln(x))a5
2σ 4σ σ (70.31)
1 2 2 2 1 2 1 1
+ 2 (σ + 2r) (a2 t + a3 t ) + 2 (σ − 2r)a5 t − a3 t + 2 a6
8σ 2σ 2 σ
We are now able to determine the function φ(x, t, u) using (70.16) and the
above expression for α(x, t):
1 1 1
φ(x, t, u) = 2
(ln(x))2 a3 + 2 (σ 2 − 2r)(a2 + 2a3 t) + 2 (ln(x))a5
2σ 4σ σ
1 2 2 2 1 2 1 1
+ 2 (σ + 2r) (a2 t + a3 t ) + 2 (σ − 2r)a5 t − a3 t + 2 a6 u + β(x, t)
8σ 2σ 2 σ
(70.32)
1000
From (70.18), using (70.27) and (70.28), we can find ξ(x, t):
1
ξ(x, t) = [x· ln(x)](a2 + 2a3 t) + x(a4 + a5 t) (70.33)
2
Using the expressions (70.33), (70.27) and (70.32) and we can find the explicit
expression for the vector Γ introduced in (70.3):

1 ∂
Γ= [x· ln(x)](a2 + 2a3 t) + x(a4 + a5 t)
2 ∂x
∂
+ (a1 + a2 t + a3 t2 )
∂t
1 2 1 2 1
+ (ln(x)) a 3 + (σ − 2r)(a 2 + 2a 3 t) + (ln(x))a5
2σ 2 4σ 2 σ2
1 1 1 1
+ 2 (σ 2 + 2r)2 (a2 t + a3 t2 ) + 2 (σ 2 − 2r)a5 t − a3 t + 2 a6 u
8σ 2σ 2 σ
∂
+ β(x, t)
∂u

∂ 1 ∂ ∂
Γ = a1 + a2 [x· ln(x)] +t
∂t 2 ∂x ∂t
1 ∂
2 2 2
+ 2 2(σ − 2r) ln(x) + (σ + 2r) t u
8σ ∂u

∂ ∂ 1
+ a3 [x· ln(x)]t + t2 + 2 (ln(x))2 + (σ 2 − 2r)(ln(x))t
∂x ∂t 2σ
∂ (70.34)
1 2 2 2 2
+ (σ + 2r) t − σ t u
4 ∂u
∂ ∂
∂ 1 1 2
+ a4 x + a5 xt + ln(x) + 2 (σ − 2r)t u
∂x ∂x σ2 2σ ∂u
1 ∂ ∂
+ a6 2 u + β(x, t)
σ ∂u ∂u
Therefore the Lie algebra of the infinitesimal transformations of the Black-
Scholes equation is being spanned by the six vector fields
∂
X1 = (70.35)
∂t
1001
1 ∂ ∂
X2 = [x· ln(x)] +t
2 ∂x ∂t (70.36)
1 2 ∂
+ 2 2(σ − 2r) ln(x) + (σ 2 + 2r)2 t u
8σ ∂u
∂ ∂ 1
X3 = [x· ln(x)]t + t2 + 2 (ln(x))2 + (σ 2 − 2r)(ln(x))t
∂x ∂t 2σ (70.37)
1 2 ∂
+ (σ + 2r)2 t2 − σ 2 t u
4 ∂u
∂
X4 = x (70.38)
∂x
∂ 1 1 ∂
X5 = xt + 2 ln(x) + (σ 2 − 2r)t u (70.39)
∂x σ 2 ∂u
1 ∂
X6 = 2 u (70.40)
σ ∂u
and the infinite dimensional subalgebra
∂
X∞ = β(x, t) (70.41)
∂u
where β(x, t) is an arbitrary solution of the original BS-equation.
71.2. Some General Solutions to the BS equation. We consider next
an equivalent, simplified form of the BS equation
ut = x2 uxx + kxux − ku (70.42)
Using a similar analysis as before, we find that the Lie algebra of symmetries
of the BS equation is being spanned by six infinitesimal generators
∂
X1 = (70.43)
∂t
1 ∂ ∂
X2 = (x· ln(x)) +t
2 ∂x ∂t (70.44)
1 1 ∂
− (k − 1) ln(x) + (k + 1)2 t u
4 4 ∂u
∂ ∂ 1
X3 = (x· ln(x)) t + t2 − (ln(x))2
∂x ∂t 4 (70.45)
1 1 1 ∂
+ (k − 1)(ln(x)) t + (k + 1)2 t2 + t u
2 4 2 ∂u
1002
∂
X4 = x (70.46)
∂x
∂ 1 ∂
X5 = xt − [ ln(x) + (k − 1)t ]u (70.47)
∂x 2 ∂u
∂
X6 = u (70.48)
∂u
and the infinite-dimensional subalgebra
∂
Xβ = β(x, t) (70.49)
∂u
where β(x, t) is an arbitrary solution of the original BS equation (71.42).
We shall determine now the general solution of the BS equation corresponding
to the generator X2 . We shall determine first the invariants corresponding
to X2 . We thus solve the system
dx dt du
= =
1
(x· ln(x)) t 1 (70.50)
2 − [ (k − 1) ln(x) + (k + 1)2 t ]
4
From equation
dx dt
1 =
2
(x· ln(x)) t
we get after integration
√ √
ln(lnx) = ln( t) + A, i.e. ln(x) = C1 · t
which means that one invariant is

ln(x)
y= √ (70.51)
t
From equation
dt du
=
t 1
− [ (k − 1) ln(x) + (k + 1)2 t ]
4
we get
(k − 1) ln(x) + (k + 1)2 t du
− dt =
4t u
1003
and using equation (71.51), we have the equation
√
(k − 1)y t + (k + 1)2 t du
− dt =
4t u
Integrating the above equation, we get
1 √
− [ 2(k − 1)y t + (k + 1)2 t ] = ln(u) + B
4
and thus
1
C2 u = exp − [ 2(k − 1) ln x + (k + 1)2 t ]
4
Therefore, another invariant is given by
1
v = u · exp [ 2(k − 1) ln x + (k + 1)2 t ] , v = v(y) (70.52)
4
We find next that
1 y 1
ut = − 2 vy + (k + 1)2 v exp − [ 2(k − 1) ln x + (k + 1)2 t ] (70.53)
4 t 4
1 1 1 1
ux = √ vy − (k − 1)v exp − [ 2(k − 1) ln x + (k + 1)2 t ] (70.54)
x t 2 4
and
1 1 1 vy 1 1
uxx = 2 vyy − √ vy − (k − 1) √ + (k − 1)v + (k − 1)2 v
x t t t 2 4
1 (70.55)
2
× exp − [ 2(k − 1) ln x + (k + 1) t ]
4
We thus see that equation ut = x2 uxx + kxux − ku gets transformed into
1 y
− 2 vy + (k + 1)2 v
4 t
1 1 1 vy 1 1
= x2 · 2 vyy − √ vy − (k − 1) √ + (k − 1)v + (k − 1)2 v
x t t t 2 4
1 1
1
+ kx· √ vy − (k − 1)v − k·v
x t 2
1004
1
vyy + y vy = 0 (70.56)
2
Using the substitution
1
v(y) = w(y)· exp − y 2 (70.57)
8
and also taking into account that
1 1
vy = − yw + wy · exp − y 2 (70.58)
4 8

1 1 1 1
vyy = − + y 2 w − ywy + wyy · exp − y 2 (70.59)
4 16 2 8
equation (71.56) gets transformed into

1 1 2 1 1 1
− + y w − ywy + wyy + y − yw + wy = 0
4 16 2 2 4
which is equivalent to Weber’s equation
1 1
wyy = + y2 w (70.60)
4 16
1 y 1 y
w(y) = A·U , √ + B·V ,√ (70.61)
2 2 2 2
and thus
1 y 1 y 1
v(y) = A·U , √ + B·V , √ × exp − y 2 (70.62)
2 2 2 2 8
where U and V are the parabolic cylindrical functions (M. Abramowitz
and S. Stegun, §19.1). Going back to the original variables, we obtain the
general solution of the BS squation
1 ln(x) 1 ln(x)
u(x, t) = A·U , √ + B·V , √
2 2t 2 2t
1 1 (70.63)
2 2
× exp − [ 2(k − 1) ln x + (k + 1) t ] − (ln(x))
4 8t
1005
71 Portfolio Optimization and
Stochastic Control
We introduce a solution procedure for the Hamilton-Jacobi-Bellman (HJB)
equation associated to a portfolio optimisation problem. The HJB equa-
tion is derived under the principle of Dynamic Programming. The solution
method relies on determining the underlying Lie point symmetries of the
equation. On finding the symmetries, we consider a linear combination of
the infinitesimal generators, and the two invariants that they determine. Us-
ing the transformations based on the two invariants, we convert the HJB
equation to an ordinary non-linear differential equation, which can in turn
be converted to an Abel differential equation. Abel’s differential equation
admits a closed form solution. We obtain twelve families of solutions.
I. Introduction. Stochastic Control (Bertsekas [6], [7], Evans [19], Fleming
and Soner [20], Ingersoll [28], Øksendal [42], Pham [46], Soner [49], Touzi
[52]) is an optimisation procedure used in many areas of financial applica-
tions, such as
[1]Portfolio Optimization (Benth and Karlsen [5], Korn and Korn [31], Korn
[32], Kraft [33])
[2]Option Pricing (Baron and Jensen [3])
[3]Consumption-Investment Strategies (Davis, Panas and Zariphopoulou [15],
Elliott and Kopp [18], Karatzas, Lehoczky, Sethi and Shreve [29], Karatzas,
Lehoczky and Shreve [30] and Zariphopoulou [53])
[4]Stochastic Interest Rates (Lewis [36])
[5]Bond Pricing (Kraft [33]) and Bond Portfolio Optimization (Puhle [47])
[6]Asset Allocation (Munk [41])
[7]Risk (Bielecki and Pliska [8], Dmitrasinovic-Vivodic, Lari-Lavassani, Li
and Ware [16])
[8]DSGE models (Papanicolaou [45])
[9]Pairs Trading (Mudchanatongsuk, Primbs and Wong [40])
[10]Pension Funds (Boulier, Trussant and Florens [13], Gerrard, Haberman
and Vigna [21])
[11]Insurance (Hipp, [24])
[12]Debt Crises (Stein, [50])
[13]Various other applications (Bucci et al. [14])
The basic equation used in the optimization process is the Hamilton-Jacobi-
Bellman (HJB) equation, derived under the principle of Dynamic Program-
1006
ming (Bellman [4]). Despite successful predictions of the HJB equation for
such models, there is no available general solution procedure. The solution
obtained so far is based on separation of variables. It is the purpose of this
paper to introduce a solution procedure (and to find some general solutions)
based on Lie Symmetries.
The Lie symmetry approach in solving differential equations was introduced
by S. Lie (Lie [37] and Lie [38]) and is best described in some recent ref-
erences, like Anderson and Ibragimov [1], Bluman and Kumei [9], Dresner
[17], Hereman [23], Ibragimov [25], Ibragimov [26], Olver [43], Ovsiannikov
[44] and Stephani [51]. This technique was used for the first time in Finance
by Ibragimov and Gazizov [27]. The same technique was also used by the
author in solving the Bensoussan-Crouhy-Galai equation, which models the
stochastic equity volatility (Antoniou [2]).
There is however a growing list of papers in applying Lie symmetry methods
to Finance. For a (non-complete) set of references, see Bordag [10], Bordag
and Chmakova [11], Bordag and Frey [12], Goard [22], Leach, O’Hara and
Sinkala [35] and Silberberg [48].
The paper is organized as follows: In section 2 we consider a simple model,
which leads to the Hamilton-Jacobi-Bellman equation. In section 3 the Lie
symmetry analysis to this equation is performed. In other words we find
the infinitesimal generators of the Lie group. In section 4 we find general
solutions to the equation, by first considering a linear combination of the in-
finitesimal generators and then by determining the invariants of the equation.
The invariants allow us to convert the equation into a nonlinear, second order
ordinary differential equation. Using another transformation, we convert this
nonlinear equation to an Abel equation possessing a closed-form solution. In
section 5 we present our conclusions and the twelve families of our solution.
II. A Portfolio Selection Model and the HJB Equation.
In this section we consider a selection portfolio model, optimized under the
principle of Dynamic Programming, leading to an HJB equation. The model
we consider is a Brownian model, in which the money market account and
stocks are modeled by
dS0 (t) = S0 (t)r(t)dt, S0 (0) = 1

(71.1)
dS(t) = S(t)[µ(t)dt + σdW (t)], S(0) > 0
The coefficients r, µ, and σ are progressively measurable processes being

uniformly bounded. Let also
1007
φ(t): be the number of stocks invested at time t.
φ0 (t)S0 (t): be the money invested in the money market account at time t.
X φ (t) = φ0 (t)S0 (t) + φ(t)S(t): be the wealth of the investor.
φ(t)S(t)
π(t) = : be the proportion invested in stock.
X(t)
1 − π(t): be the proportion invested in the money market account.
The following proposition holds true (see for example Kraft [33], [34] or
Øksendal [42]).
Proposition. The wealth dynamics of a self-financing trading strategy is
described by the so-called wealth equation
dX(t) = X(t)[(r + π(µ − r))dt + πσdW (t)]
(71.2)
X(0) = x0 , x0 ≥0
We come now to the portfolio optimization one investor has with initial
wealth X(0) = x0 > 0 and would like to have optimal final wealth X ∗ (T )
under an optimal strategy π ∗ . For a given initial wealth x0 > 0, the investor
maximizes the following utility functional
sup E[U (X π (T ))] (71.3)

π∈A
where
A = {π : π admissible and E[U (X π (T ))− ] < ∞} (71.4)
and X π (t) is given by (71.2):

π π
dX (t) = X (t) [r(t) + π(t)(µ(t) − r(t))]dt + π(t)σdW (t)
(71.5)
π
X (0) = x0
The function U (x), defined for positive real numbers, is called the utility
function, which is strictly concave, strictly increasing and continuously dif-
ferentiable, and satisfies the conditions
U 0 (0) = lim U 0 (x) = ∞, U 0 (∞) = lim U 0 (x) = 0 (71.6)

x↓0 x↑∞
Let us define the value function
G(t, x) = sup E t,x [U (X π (T ))] (71.7)

π∈A
1008
where
E t,x [U (X π (T ))] = E t,x [U (X π (T ))|X(t) = x], G(T, x) = U (x) (71.8)
Suppose now that there is an optimal strategy π ∗ . The corresponding HJB

equation, derived under the Principle of Dynamic Programming, is given by

1 2 2 2
Gt (t, x) + sup Gx (t, x)[r + (µ − r)π]x + σ π x Gxx (t, x) = 0 (71.9)
π 2
for any admissible strategy π and terminal condition G(T, x) = U (x).
We consider the case where r, µ and σ are all constants. Pointwise maxi-
mization over π yields the first order condition
Gx (t, x)(µ − r)x + σ 2 x2 Gxx (t, x)π(t) = 0 (71.10)
which implies that the optimal strategy is given by

µ − r Gx (t, x)
π∗ = − (71.11)
σ 2 xGxx (t, x)
Since the derivative of (71.10) with respect to π is σ 2 x2 Gxx (t, x), the can-
didate π ∗ is the global maximum if Gxx < 0. Substituting π ∗ into the HJB
equation (71.9), yields a PDE for G(t, x):
1 G2 (t, x)
Gt (t, x) + rxGx (t, x) − k 2 x =0 (71.12)
2 Gxx (t, x)
where
µ−r
k= (71.13)
σ
Equation (71.12) also appears in the classical book on ”Stochastic Differential
Equations” by Bernt Øksendal (Øksendal [42], page 247, equation (11.2.50)).
The traditional way of solving equation (71.12) is by separation of variables.
In fact, considering that (Merton [39], Øksendal [42])
1 γ
G(t, x) = x f (t)1−γ (71.14)
γ
equation (71.12) yields the first order differential equation
f 0 (t) = Kf (t) (71.15)
1009
where
γ 1
K=− r+ k2
1−γ 2(1 − γ)
The above differential equation (71.15) has the solution f (t) = e−K(T −t) and
satisfies the condition f (T ) = 1. Therefore the solution (71.14) written as
1 γ −K(T −t) 1−γ
G(t, x) = x e (71.16)
γ
1
is compatible with the condition G(T, x) = U (x)≡ xγ . From (71.16) and
γ
(71.11) we can determine the value function and the optimal strategy respec-
tively. For the type of solution we have obtained, we can check that Gxx < 0
if γ < 1. In Section IV we provide our own method of solution, based on Lie
Symmetry Analysis.
III. Lie Symmetry Analysis of the HJB Equation
We now consider the HJB equation (71.12) written in the equivalent form
1
ut uxx + rxux uxx − k 2 u2x = 0, u = u(x, t) (71.17)
2
where we have changed the notation of the function from G(t, x) to the more
traditional notation u(x, t).
We are going to perform a Lie symmetry analysis of this equation. By this
we mean that we shall determine the infinitesimal generators of the Lie group
of the underlying symmetries of the differential equation. These would then
be used to determine the invariants needed to convert the partial differential
equation into an ordinary differential equation.
Let us denote by V the vector field (the generator of the symmetries) defined
by
∂ ∂ ∂
V = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u) (71.18)
∂x ∂t ∂u
where ξ(x, t, u), τ (x, t, u) and φ(x, t, u) are smooth functions of their argu-
ments. The second prolongation of V is defined by
∂ ∂ ∂ ∂ ∂
P r(2) V = V + φx + φt + φxx + φxt + φtt (71.19)
1010
The symmetries are determined by the equation (Olver [43], Theorem 2.31)
P r(2) V [∆(x, t, u(2) )] = 0, as long as ∆(x, t, u(2) ) = 0 (71.20)
The quantity ∆(x, t, u(2) ) is defined by

1
∆(x, t, u(2) ) = ut uxx + rxux uxx − k 2 u2x (71.21)
2
We implement next the condition P r(2) V [∆(x, t, u(2) )] = 0. We have that
P r(2) V [∆(x, t, u(2) )] = 0

1
⇔P r(2) V [ut uxx + rxux uxx − k 2 u2x ] = 0
2
ξ(rux uxx ) + φx (rxuxx − k 2 ux ) + φt uxx + φxx (ut + rxux ) = 0 (71.22)
The coefficients φx , φt and φxx are calculated to be (Olver [43], Example

2.38)
φx = φx + (φu − ξx )ux − ξu u2x − τx ut − τu ux ut (71.23)
φt = φt + (φu − τt )ut − τu u2t − ξt ux − ξu ux ut (71.24)

The functions τ , ξ and φ have the following dependence
τ = τ (t), ξ = ξ(x, t), φ = φ(x, t, u) (71.26)
The above dependence can also be proved rigorously by fully expanding

(71.22) ans equating to zero all the coefficients of the partial derivatives
of the function u. Equation (71.22) thus becomes
ξ(rux uxx ) + (rxuxx − k 2 ux )[φx + (φu − ξx )ux ]

+ (φt − ξt ux )uxx + (φu − τt )uxx ut
(71.27)
+ [φxx + (2φxu − ξxx )ux + φuu u2x + (φu − 2ξx )uxx ]ut
+ rxux [φxx + (2φxu − ξxx )ux + φuu u2x + (φu − 2ξx )uxx ] = 0
1011
We now have to take into account the condition ∆(x, t, u(2) ) = 0. We there-
1 u2
fore substitute ut by −rxux + k 2 x into the previous equation (71.27):
2 uxx
ξ(rux uxx ) + (rxuxx − k 2 ux )[φx + (φu − ξx )ux ]

1 u2
+ (φt − ξt ux )uxx + (φu − τt )uxx −rxux + k 2 x
2 uxx
2
+ [φxx + (2φxu − ξxx )ux + φuu ux + (φu − 2ξx )uxx ]× (71.28)
1 2 u2x
× −rxux + k
2 uxx
+ rxux [φxx + (2φxu − ξxx )ux + φuu u2x + (φu − 2ξx )uxx ] = 0
We equate to zero all the coefficients of the partial derivatives of the function
u:
1 2
k φuu = 0 (71.29)
2
1 2
k (2φxu − ξxx ) = 0 (71.30)
2
1 2
k φxx = 0 (71.31)
2
−k 2 φx = 0 (71.32)
Coefficient of (ux )2 uxx :
1
− k 2 τ 0 (t) = 0 (71.33)
2
Coefficient of (uxx )2 :
rxφx + φt = 0 (71.34)
1012
Coefficient of ux (uxx )2 :
rξ − ξt + rx[τ 0 (t) − ξx ] = 0 (71.35)

Equations (71.29)-(71.35) are the determining equations of the Lie symme-
tries of equation (71.17). We are now going to solve the system of equations
(71.29)-(71.35).
From equation (71.29) we get that φuu = 0, which means that φ is a linear
function with respect to u:
φ(x, t, u) = α(x, t)·u + β(x, t) (71.36)
From equation (71.32) we get that φx = 0. Therefore, using (71.36), we have
αx ·u + βx = 0 (71.37)
from which we get αx = 0 and βx = 0. Therefore the functions α and β are

independent of x. On the other hand, since φx = 0, we get from (71.34) that
φt = 0. Therefore αt ·u + βt = 0, from which we have that αt = 0 and βt = 0,
which means that α and β are constants. We then have
φ(x, t, u) = a1 ·u + a2 (71.38)
Equation (71.31) becomes an identity. From equation (71.33) we get that

τ 0 (t) = 0 and thus τ (t) is a constant:
τ (t) = a3 (71.39)
We get from (71.30), because of (71.38), that ξxx = 0. Therefore ξ(x, t) is a

linear function with respect to x:
ξ(x, t) = f (t)·x + g(t) (71.40)
where f (t) and g(t) are functions to be determined. Using the above expres-
sion, equation (71.35) becomes
−f 0 (t)x + [r·g(t) − g 0 (t)] = 0 (71.41)
from which we have, since this equation should be true for any x, that
f 0 (t) = 0 and r·g(t) − g 0 (t)
1013
We thus have
f (t) = a4 and g(t) = a5 ert (71.42)
Therefore the function ξ(x, t) has the form
ξ(x, t) = a4 x + a5 ert (71.43)
Relations (71.38), (71.39) and (71.43) allow us to write down the generator
of the symmetries:
∂ ∂ ∂
V = (a4 x + a5 ert ) + a3 + (a1 u + a2 ) (71.44)
∂x ∂t ∂u
Therefore we have the following
Theorem 1. The Lie algebra of the infinitesimal transformations of the HJB
equation (71.17) is spanned by the five vector fields
∂ ∂ ∂ ∂ ∂
X1 = u , X2 = , X3 = , X4 = x , X5 = ert (71.45)
∂u ∂u ∂t ∂x ∂x
IV. General Solution to the HJB Equation.
In order to find a solution to the HJB equation (71.17), we have to deter-
mine the invariants of the equation. For this purpose we consider a linear
combination of the above found generators X1 , · · ·, X5 of the symmetries. We
consider the combination
X3 + λX4 + µX5 + νX1 + ρX2 (71.46)
In other words we consider the generator

∂ ∂ ∂
+ (λx + µert ) + (νu + ρ) (71.47)
∂t ∂x ∂u
where the parameters λ, µ, ν and ρ will be determined on the basis of find-
ing closed form solutions to the equation. The parameter µ appearing in
(71.46) and (71.47) not to be confused with µ appearing in the definition of
k, equation (71.13). Based on (71.47), we consider the auxiliary system
dt dx du
= rt
= (71.48)
1 λx + µe νu + ρ
1014
In order to solve the first equation
dt dx
= (71.49)
1 λx + µert
of the above system, we have to consider two cases: λ6=r and λ = r. Details
are given in the Appendix.
IV.1. Case 1. Solution of the HJB equation in the λ6=r case.
If λ6=r, equation (71.49) has the general solution
µ (r−λ)t
C1 = e−λt x − e (71.50)
r−λ
Therefore one invariant of the HJB equation (71.17) is given by
µ (r−λ)t
y = e−λt x − e (71.51)
r−λ
The second equation of the system (71.48)
dt du
= (71.52)
1 νu + ρ
admits the general solution
C2 + νt = ln(νu + ρ) (71.53)
Therefore another invariant is given by v = v(y) where

1 νt
u= e ·v(y) − ρ (71.54)
ν
Using (71.54) and (71.51), we find that the partial derivatives of the function
u transform as

1 µ (r−λ)t 0
ut = v(y) − λy + e v (y) eνt (71.55)
ν r−λ
1 (r−λ)t 0 1
ux = e v (y), uxx = e(r−2λ)t v 00 (y) (71.56)
ν ν
Using (71.55) and (71.56), the original HJB equation (71.17) is being trans-
formed into
1
νv(y)v 00 (y) + (r − λ)yv 0 (y)v 00 (y) − k 2 (v 0 (y))2 = 0 (71.57)
2
1015
which is a nonlinear, second order ordinary differential equation. Under the
transformation
v 0 (y)
w(y) = (71.58)
v(y)
0
1 2 2
[ν + (r − λ)yw(y)]w (y) + ν − k w (y) + (r − λ)yw3 (y) = 0 (71.59)
2
which is an Abel differential equation. Introducing the notation
1
γ = ν − k2 and δ = r − λ (δ6=0) (71.60)
2
equation (71.59) can be written as
γw2 + δyw3
w0 (y) = − (71.61)
ν + δyw
which in turn, under the substitution
1
w(y) = (71.62)
Z(y)
takes on the form
γZ(y) + δy
Z 0 (y) = (71.63)
νZ(y) + δy
Introducing the function V (y) by
Z(y) = y V (y) (71.64)
we transform equation (71.63) into the equation
γV (y) + δ
V + y V 0 (y) = (71.65)
νV (y) + δ
which is equivalent to the equation (after separation of variables)

νV + δ 1
2
V0 =− , V = V (y) (71.66)
νV + (δ − γ)V − δ y
1016
In solving (71.66), we distinguish the following cases:
1
Case 4.1.I. Let γ = 0, i.e. ν = k 2 (ν > 0). Equation (71.66) can be
2
written as
νV + δ 1
2
V0 =− , V = V (y) (71.67)
νV + δV − δ y
Denoting by D≡δ 2 + 4νδ the discriminant of the trinomial νV 2 + δV − δ, we
have the following subcases:
Subcase 4.1.Ia. For D > 0, i.e. δ(δ + 4ν) > 0⇔(δ > 0 or δ < −4ν), i.e.
(r − λ > 0 or r − λ < −4ν)⇔(λ < r or λ > r + 4ν), equation (71.67) admits
a solution given implicitly by
λη + δ νξ + δ
·ln|V − η| − ·ln|V − ξ| = −ln(y) + A (71.68)
ν(η − ξ) ν(η − ξ)
where ξ and η are the two real roots of the trinomial given by
√ √
−δ + D −δ − D
ξ= , η= (71.69)
2ν 2ν
respectively, and A is an arbitrary constant.
Subcase 4.1.Ib. For D < 0, i.e. δ(δ + 4ν) < 0⇔(−4ν < δ < 0), i.e.
(−4ν < r − λ < 0)⇔(r < λ < r + 4ν), equation (71.67) admits a solution
given implicitly by
1 δ V + ξ
·ln[(V + ξ)2 + η 2 ] + arctan = −ln(y) + A (71.70)
2 2νη η
where ξ and η are two parameters defined by
√
δ −D
ξ= and η = (71.71)
2ν 2ν
Subcase 4.1.Ic. For D = 0, i.e. δ = −4ν⇔λ = r + 4ν, equation (71.67)
admits a solution given implicitly by
2
+ ln|V − 2| = −ln(y) + A (71.72)
V −2
1
Case 4.1.II. Let γ6=0, i.e. ν6= k 2 . We shall distinguish two cases: γ = δ
2
1017
and γ6=δ.
Subcase 4.1.IIa. If γ = δ, equation (71.66) becomes
νV + δ 0 1
V = − (71.73)
νV 2 − δ y
In this case we have to consider two cases: ν6=δ and ν = δ. In the former
case we may set either δ = νω 2 or δ = −νω 2 , ω > 0. We thus have the
following:
Sub-subcase 4.1.IIa1. For δ = νω 2 we obtain from (71.73) the equation
V + ω2 0 1
2 2
V =− (71.74)
V −ω y
which gives upon integration
1 1 V − ω
ln(V 2 − ω 2 ) + ω· ln = − ln(y) + A (71.75)
2 2 V +ω
Sub-subcase 4.1.IIa2. For δ = −νω 2 we obtain from (71.73) the equation
V − ω2 0 1
V = − (71.76)
V 2 + ω2 y
which gives upon integration
1 2 2
V
ln(V + ω ) − ω· arctan = − ln(y) + A (71.77)
2 ω
Sub-subcase 4.1.IIa3. For δ = ν we obtain from (71.73) the equation
1 1
V0 =− (71.78)
V −1 y
and by integration
ln |V − 1| = − ln(y) + A (71.79)

Subcase 4.1.IIb. If γ6=δ, denoting by D≡(δ − γ)2 + 4νδ the discriminant
of the trinomial νV 2 + (δ − γ)V − δ, we have the following:
1018
Sub-subcase 4.1.IIb1. For D≡(δ − γ)2 + 4νδ > 0, equation (71.66) admits
λη + δ νξ + δ
· ln |V − η| − · ln |V − ξ| = − ln(y) + A (71.80)
ν(η − ξ) ν(η − ξ)
where ξ and η are the two real roots of the trinomial, given by
√ √
γ−δ+ D γ−δ− D
ξ= and η = (71.81)
2ν 2ν
Sub-subcase 4.1.IIb2. For D≡(δ − γ)2 + 4νδ < 0, equation (71.66) admits
1 2 2 δ − νξ V + ξ
ln[(V + ξ) + η ] + arctan = − ln(y) + A (71.82)
2 νη η
where ξ and η are two parameters defined by
√
δ−γ −D
ξ= and η = (71.83)
2ν 2ν
Sub-subcase 4.1.IIb3. For D≡(δ − γ)2 + 4νδ = 0, equation (71.66) admits
δ 1
ξ− + ln(V + ξ) = − ln(y) + A (71.84)
ν V +ξ
where ξ is the double root of the trinomial given by
γ−δ
ξ= (71.85)
2ν
and A is an arbitrary constant.
Going back, i.e. using (71.58), (71.62) and (71.64), we see that v = v(y) can
be determined by integrating the equation
v 0 (y) 1
= (71.86)
v(y) yV (y)
1019
where V (y) is given implicitly by any of the cases previously considered.
Therefore we find that
Z y ds
v(y) = B· exp (71.87)
y0 s V (s)

The function u(x, t), i.e. G(x, t) which enters the optimization strategy
(71.11), is given by (71.54) and the variable y which combines the two vari-
ables x and t is given by (71.51):
1
Z y ds
νt
G(x, t) = B·e · exp −ρ
ν y0 s V (s) (71.88)
−λt µ (r−λ)t
y =e x− e
r−λ
Case 2. Solution of the HJB equation in the λ = r case.
In this case the system (71.48) has the form
dt dx du
= rt
= (71.89)
1 rx + µe νu + ρ
The above system admits two independent solutions
C1 = x·e−rt − µt and C2 + νt = ln(νu + ρ) (71.90)
Therefore in this case we have the two invariants

1 νt
y = x·e−rt − µt and u = e ·v(y) − ρ (71.91)
ν
Using the above invariants, we find that the partial derivatives of the function
u transform as

1 0
ut = v(y) − [r(y + µt) − µ]v (y) eνt (71.92)
ν
1 (ν−r)t 0 1
ux = e v (y), uxx = e(ν−2r)t v 00 (y) (71.93)
ν ν
Using (71.92) and (71.93), equation (71.17) transforms into
1
νv(y)v 00 (y) + µv 0 (y)v 00 (y) − k 2 (v 0 (y))2 = 0 (71.94)
2
1020
which is a nonlinear, second order ordinary differential equation. Under the
transformation
v 0 (y)
w(y) = (71.95)
v(y)
(ν + µw)w0 = −γw2 − µw3 , w = w(y) (71.96)
which is and Abel differential equation. This equation may also be treated
as an equation with separable variables. In solving (71.96) we shall consider
two cases: µ = 0 and µ6=0.
Case 4.2.I. If µ = 0, we find that equation (71.96) is equivalent to
νw0 (y) = −γw2 (71.97)

ν
w(y) = (71.98)
γy + A
v 0 (y)
where A is an arbitrary constant. Integrating further w(y) = , we find
v(y)
that
v(y) = B(γy + A)ν/γ (71.99)
where B is another arbitrary constant.

Case 4.2.II. If µ6=0, equation (71.96) takes on the form
γw2 + µw3
w0 (y) = − (71.100)
ν + µw
Subcase 4.2.IIa. If γ6=0, under the substitution
γ
w(y) = (71.101)
γY (y) − µ
γ 2Y
Y 0 (y) = (γ6=ν) (71.102)
νγY + µ(γ − ν)
1021
and by integration we obtain
νγY + µ(γ − ν) ln(Y ) = γ 2 y + A (71.103)
Therefore v(y) is the solution to the equation
v 0 (y) γ
= (71.104)
v(y) γY (y) − µ
where Y (y) is given implicitly by (71.103). Therefore we find that

Z y γds
v(y) = B·exp (71.105)
y0 γY (s) − µ

1
Subcase 4.2.IIb. For γ = 0, (i.e. ν = k 2 ), equation (71.96) becomes
2
µw3
w0 (y) = − (71.106)
ν + µw
1
w= (71.107)
X(y)

µ
X 0 (y) = (71.108)
νX + µ
νX 2 (y) + 2µX(y) = 2µy + A (71.109)
where A is an arbitrary constant. Therefore v(y) is the solution to the equa-

v 0 (y) 1
tion = where X(y) is given implicitly by (71.109). Hence we find
v(y) X(y)
that
Z y ds
v(y) = B·exp (71.110)
y0 X(s)
1022
The variable y which connects x and t is given by the first of (71.91):
y = x·e−rt − µt (71.111)
Important Notice. The parameter k was defined in (71.13) and is being

expressed in termas of µ, r and σ. The parameter µ was also used in (71.46)
and (71.47). However despite of the same notation, these two parameters are
different in nature. We thus have to change notation and redefine k by
µ̃ − r
k= (71.112)
σ
with the obvious renaming µ→µ̃ in (71.13).
Before we write down the various solutions of the equation (71.17), i.e.
(71.12), we have to say that the solutions contain some of the parameters
λ, µ, ν and ρ introduced in (71.46) and (71.47) and the auxiliary parameters
1
γ = ν − k 2 and δ = r − λ. The original parameters of the theory are µ̃, r,
2
σ and k.
(a). In Case I (λ6=r) there correspond nine (9) solutions. In all solutions the
unknown function u(x, t), i.e. G(x, t), is given by
1 νt ρ
G(x, t) = e ·v(y) − (71.113)
ν ν
while the parameter y connecting the variables x and t is given by
−λt
µ rt
y=e x− e (71.114)
r−λ
The function v(y) is given by
Z y ds
v(y) = B·exp (71.115)
y0 sV (s)
while the function V (y) satisfies the equation

νV + δ 1
2
V 0 (y) = − (71.116)
νV + (δ − γ)V − δ y
Therefore the nine solutions come from the nine cases (corresponding to
the choice of the various parameters) which correspond to the solutions of
1023
equation (71.116). The solutions are given in Case 4.1 above.
(b). In Case II (λ = r) there correspond three (3) solutions. In all solutions
the unknown function u(x, t), i.e. G(x, t), is given by
1 νt ρ
G(x, t) = e ·v(y) − (71.117)
ν ν
while the parameter y connecting the variables x and t is given by
y = xe−rt − µt (71.118)
The function v(y) is given by
v(y) = B(γy + A)ν/γ , µ=0 (71.119)
or
Z y
γds
v(y) = B·exp , µ6=0, γ6=0 (71.120)
y0 γY (s) − µ
with Y (y) given implicitly by (71.103), or finally by
Z y ds
v(y) = B·exp , µ6=0, γ = 0 (71.121)
y0 X(s)
with X(y) given implicitly by (71.109).

Conclusions and Discussion. Collecting the results of the previous Sec-
tion, we arrive at the following
Theorem 2. The HJB equation (71.17) or (71.12)
1 G2 (t, x)
Gt (t, x) + rxGx (t, x) − k 2 x =0 (71.122)
2 Gxx (t, x)
admits the following twelve families of general solutions:
Solution I. The first solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.123)
µ
y = e−λt x − e(r−λ)t , µ6=0, λ6=r
r−λ
where V (y) is a function defined implicitly by
λη + δ νξ + δ
· ln |V − η| − · ln |V − ξ| = − ln(y) + A (71.124)
ν(η − ξ) ν(η − ξ)
1024
The parameters ξ and η are defined by
√ √
−δ + D −δ − D
ξ= and η = (71.125)
2ν 2ν
respectively, where δ = r − λ (δ6=0) and D≡δ 2 + 4νδ. The parameter ν is
1
such that ν = k 2 (ν > 0) and the parameter λ satisfies either one of the
2
inequalities λ < r or λ > r + 4ν (the last two conditions ensure that D > 0).
Solution II. The second solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.126)
−λt µ (r−λ)t
y =e x− e , µ6=0, λ6=r
r−λ
1 2 2 δ − νξ V + ξ
ln[(V + ξ) + η ] + arctan = − ln(y) + A (71.127)
2 νη η
√
δ −D
ξ= and η = (71.128)
2ν 2ν
respectively, where δ = r − λ (δ6=0) and D≡δ 2 + 4νδ. The parameter ν is
1
such that ν = k 2 (ν > 0) the parameter λ takes values within the range
2
r < λ < r + 4ν (the last condition ensures that D < 0).
Solution III. The third solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.129)
−λt µ (r−λ)t
y =e x− e , µ6=0, λ6=r
r−λ
2
+ ln |V − 2| = − ln(y) + A (71.130)
V −2
1025
The parameter δ is defined by δ = r − λ (δ6=0) and satidfies the equation
1
δ = −4ν, i.e. λ = r + 4ν. The parameter ν is such that ν = k 2 (ν > 0).
2
Solution IV. The fourth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.131)
−λt µ (r−λ)t
y =e x− e , µ6=0, γ6=0, λ6=r, γ = δ
r−λ
1 1 V − ω
ln[V 2 − ω 2 ] + ω ln = − ln(y) + A (71.132)
2 2 V +ω
1
with δ = νω 2 , δ = r − λ (δ6=0) and ν 6= k 2 .
2
Solution V. The fifth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.133)
µ (r−λ)t
y = e−λt x − e , µ6=0, γ6=0, λ6=r, γ = δ
r−λ
1 V
ln[V 2 + ω 2 ] − ω· arctan = − ln(y) + A (71.134)
2 ω
1
with δ = −νω 2 , δ = r − λ (δ6=0) and ν 6= k 2 .
2
Solution VI. The sixth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.135)
µ
y = e−λt x − e(r−λ)t , µ6=0, γ6=0, λ6=r, γ = δ
r−λ
ln |V − 1| = − ln(y) + A (71.136)
1026
1
The parameters satisfy δ = ν, δ = r − λ (δ 6= 0) and ν 6= k 2 .
2
Solution VII. The seventh solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.137)
µ
y = e−λt x − e(r−λ)t , µ6=0, λ6=r
r−λ
λη + δ νξ + δ
· ln |V − η| − · ln |V − ξ| = − ln(y) + A (71.138)
ν(η − ξ) ν(η − ξ)
√ √
γ−δ+ D γ−δ− D
ξ= and η = (71.139)
2ν 2ν
respectively, where δ = r−λ (δ6=0) and D≡(δ−γ)2 +4νδ > 0. The parameter
1
γ is defined by γ = ν − k 2 with γ 6= δ and γ6=0.
2
Solution VIII. The eighth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.140)
−λt µ (r−λ)t
y =e x− e , µ6=0, λ6=r
r−λ
1 δ − νξ V + ξ
ln[(V + ξ)2 + η 2 ] + arctan = − ln(y) + A (71.141)
2 νη η
√
δ−γ −D
ξ= and η = , (71.142)
2ν 2ν
respectively, where δ = r−λ (δ6=0) and D≡(δ−γ)2 +4νδ < 0. The parameter
1
γ is defined by γ = ν − k 2 with γ 6= δ and γ6=0.
2
Solution IX. The ninth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 sV (s) ν (71.143)
−λt µ (r−λ)t
y =e x− e , µ6=0, λ6=r
r−λ
1027
δ 1
ξ− + ln(V + ξ) = − ln(y) + A (71.144)
ν V +ξ
The parameters are given by
δ−γ
ξ= , D = (δ − γ)2 + 4νδ = 0 (71.145)
2ν
1
where δ = r − λ (δ6=0). The parameter γ is defined by γ = ν − k 2 with
2
γ 6= δ and γ6=0.
Solution X. The tenth solution is given by
B νt ρ
G(x, t) = e (γy + A)ν/γ −
ν ν (71.146)
−rt
y = xe
A, B are arbitrary constants and ν is a free parameter with ν6=0 and γ =
1
ν − k 2 with γ6=0.
2
Solution XI. The eleventh solution is given by
B νt Z y γ ds ρ
G(x, t) = e ·exp −
ν y0 γY (s) − µ ν (71.147)
−rt
y = xe − µt, µ6=0
where Y (y) is a function defined implicitly by
(νγ)Y + µ(γ − ν) ln(Y ) = γ 2 y + A, γ6=0 (71.148)
A, B are arbitrary constants and µ, ν are free parameters with µ6=0, ν6=0 and
1
γ = ν − k 2 with γ6=0.
2
Solution XII. The twelfth solution is given by
B νt Z y ds ρ
G(x, t) = e ·exp −
ν y0 X(s) ν (71.149)
−rt
y = xe − µt, µ6=0
where X(y) is a function defined implicitly by
νX 2 + 2µX = 2µy + A (71.150)
1028
1
provided that ν = k 2 , γ = 0. A, B are arbitrary constants and µ, ν are free
2
parameters with µ6=0, ν6=0.
Discussion. Only one solution has been found in the literature so far. It
is the solution given in (71.146). The equation (71.16) is a member of this
family of solutions. In fact, we can get from the solution (71.146) for A = 0
and ρ = 0 that
B 1−ν/γ
G(x, t) = (γx)ν/γ e−ν(T −t) (71.151)
ν
a solution similar in form to (71.16). It is obvious that a similar boundary
condition is being satisfied
B
G(x, T ) = U (x) = (γx)ν/γ
ν
The other solutions appear for the first time in this paper. It is obvious
that the parameters and the various constants need to be adjusted to match
the appropriate boundary conditions. It remains to be seen if the general
solutions found in this article are to have any practical interest in Finance
and Stochastic Control. In any case, the Hamilton-Jacobi-Bellman equation
associated to an optimization problem seems to have a rich mathematical
structure and thus deserves further investigation.
This research has also been published in ”Researchgate”.
Appendix. Solution of (71.49). Equation (71.49) can be written as
dx − (λx + µert )dt = 0
Using standard methods, we find that the above equation admits e−λt as
an integrating factor. Therefore we have to find the general solution of the
complete first order differential equation
e−λt dx − (λxe−λt + µe(r−λ)t )dt = 0 (71.152)
We suppose that (71.152) admits a solution of the form
U (x, t) = C (71.153)
Since
∂U ∂U
dU = dt + dx = 0 (71.154)
∂t ∂x
1029
on comparing this equation with (71.152) we obtain the following system of
∂U
= −(λxe−λt + µe(r−λ)t ) (71.155)
∂t
∂U
= e−λt (71.156)
∂x
Integrating (71.155) with respect to t, we need to distinguish two case: r6=λ
and r = λ.
Case 1. If r6=λ, we obtain from equation (71.155) by integration with respect
to t the solution
µ (r−λ)t
U = xe−λt − e + θ(x) (71.157)
r−λ
where θ(t) is a function to be determined. From (71.157) we obtain by
differentiation with respect to x that
∂U
= e−λt + θ0 (x) (71.158)
∂x
and then comparing (71.158) to (71.156), we find θ0 (x) = 0, i.e. θ(t) = C1 .
Therefore, using (71.157) and (71.153) we obtain the following solution of
(71.49) when r6=λ:
µ (r−λ)t
xe−λt − e =C (71.159)
r−λ
which is (71.50).
Case 2. If r = λ, we get from (71.155)
∂U
= −λxe−λt − µ (71.160)
∂t
which upon integration with respect to t gives us
U = xe−λt − µt + η(x) (71.161)
where η(x) is a function to be determined. From (71.161) we obtain
∂U
= e−λt + η 0 (x) (71.162)
∂x
1030
and then comparing (71.162) and (71.156), we get η 0 (x) = 0, i.e. η(t) = C.
Therefore comparing (71.161) and (71.153) we get the following solution of
(71.49) when r = λ:
xe−λt − µt = C (71.163)
which is the first of (71.90).
References
[1] R. L. Anderson and N. H. Ibragimov: ”Lie-Bäcklund Transforma-
tions in Applications ”. SIAM, Philadelphia 1979.
[2] S. M. Antoniou: ”A General Solution to the Stochastic Equity Volatil-
ity Equation”. SKEMSYS report (available from Researchgate)
[3] E. N. Baron and R. Jensen: ”A stochastic control approach to the
pricing of options”. Mathematics of Operations Research 15 (1990) 49-79
[4] R. Bellman: ”Dynamic Programming”.
Princeton University Press 1957.
[5] F. E. Benth and K. H. Karlsen: ”A Note on Merton’s portfolio se-
lection problem”. Stochastic Analysis and Applications 23 (2005) 687-704
[6] D. P. Bertsekas: ”Dynamic Programming and Optimal Control”. Athena
Scientific, Belmont Mass. 1995
[7] D. P. Bertsekas: ”Dynamic Programming and Stochastic Control”.
Academic Press, Inc. Orlando, FL 1976
[8] T. R. Bielecki and S. R. Pliska: ”Risk sensitive asset management
with transaction costs”. Finance and Stochastics 4 (2000) 1-33
[9] G. Bluman and S. Kumei: ”Symmetries and Differential Equations”.
Springer Verlag, 1989
[10] L. A. Bordag: ”Symmetry reductions of a nonlinear model of financial
derivatives”, arXiv:math.AP/0604207
[11] L. A. Bordag and A. Y. Chmakova: ”Explicit solutions for a non-
linear model of financial derivatives” Int. Journal of Theoretical and Applied
Finance 10 (2007) 1-21
[12] L. A. Bordag and R. Frey: ”Nonlinear option pricing models for
illiquid markets: scaling properties and explicit solutions”.
Preprint arXiv:0708.1568v1 [math.AP] 11 Aug 2007
[13] J-F Boulier, E Trussant and D. Florens: ”A dynamic model for
pension funds management”. Proceedings of the 5th AFIR Int. Colloquium
1995 (pp. 361-384)
[14] A. Bucci, C. Colapinto, M. Forster and D. La Torre: ”On Human
1031
capital and Economic Growth with Random Technology Shocks”. Preprint,
State University of Milan
[15] M. H. A. Davis, V. G. Panas and T. Zariphopoulou: ”European
Option Pricing with Transaction Costs”. SIAM Journal on Control and Op-
timization 31 (1993) 470-493
[16] G. Dmitrasinovic-Vivodic, A. Lari-Lavassani, X. Li and A.
Ware: ”Dynamic Portfolio Selection under Capital-at-Risk”. University of
Calgary preprint
[17] L. Dresner: ”Applications of Lies Theory of Ordinary and Partial Dif-
ferential Equations”. IOP Publishing, 1999
[18] R. E. Elliott and P. E. Kopp: ”Mathematics of Financial Markets”.
Springer 2005
[19] L. C. Evans: ”Partial Differential Equations”. Graduate Studies in
Mathematics, Vol. 19, AMS 1997
[20] W. H. Fleming and H. M. Soner: ”Controlled Markov Processes
and Viscosity Solutions”. Springer, 1993
[21] R. Gerrard, S. Haberman and E. Vigna: ”Optimal Investment
Choices Post Retirement in a Defined Contribution Pension Scheme”. City
University, Cass Business School preprint
[22] J. Goard: ”New Solutions to the Bond-Pricing Equation via Lie’s Clas-
sical Method”. Mathematical and Computer Modeling 32 (2000) 299-313
[23] W. Hereman: ”Review of Symbolic Software for the Computation of
Lie Symmetries of Differential Equations”. Euromath Bulletin 2, No.1 Fall
1993
[24] C. Hipp: ”Stochastic Control with Application in Insurance”. Univer-
sity of Karlsruhe report
ferential Equations”. Wiley, New York
[26] N. H. Ibragimov (ed.): ”CRC Handbook of Lie Group Analysis of
Differential Equations” Vol.1 1994, Vol.2 1995, Vol.3 1996 CRC Press, Boca
Raton, FL.
[27] N. H. Ibragimov and R. K. Gazizov: ”Lie Symmetry Analysis of
differential equations in finance”. Nonlinear Dynamics 17 (1998) 387-407
[28] J. E. Ingersoll, Jr: ”Theory of Financial Decision Making” Savage:
Rowman and Litterfield 1987
[29] I. Karatzas, J. P. Lehoczky, S. Sethi and S. E. Shreve: ”Explicit
solution of a general consumption/investment problem”. Math. Oper. Res.
11 (1986) 261-294
1032
[30] I. Karatzas, J. P. Lehoczky and S. E. Shreve: ”Existence and
uniqueness of multi-agent equilibrium in a stochastic, dynamic consump-
tion/ investment model”. Math. Oper. Res. 15 (1990) 80-128
[31] R. Korn and E. Korn: ”Option Pricing and Portfolio Optimization”.
AMS, 2001
[32] R. Korn: ”Optimal Portfolios: Stochastic Models for Optimal Invest-
ments and Risk Management in Continuous Time”. World Scientific, 1997
[33] H. Kraft: ”Optimal Portfolios with stochastic short rates and default-
able assets”. Springer, 2004
[34] H. Kraft: ”Financial Mathematics I: Stochastic Calculus, Option Pric-
ing, Portfolio Optimization”. Lecture Notes, University of Kaiserslautern,
August 2005
[35] P.G.L. Leach, J.G. OHara and W. Sinkala: ”Symmetry-based
solution of a model for a combination of a risky investment and a riskless
investment”. J. Math. Analysis and Applications 334 (2007) 368-381
[36] A. L. Lewis: ”Option Valuation under Stochastic Volatility”. Finance
Press, second printing 2005
[37] S. Lie: ”Gesammelte Werke”, Vols. 1-7. F. Engel and P. Hergard
(Eds.) Teubner, Leipzig 1899
[38] S. Lie: ”Vorlesungen über Differentialgleichungen mit bekannten in-
finitesimalen Transformationen”. Teubner, Leipzig 1912
[39] R. C. Merton: ”Continuous-time Finance”. Blackwell 1990
[40] S. Mudchanatongsuk, J. A. Primbs and W. Wong: ”Optimal
Pairs Trading”. Stanford University preprint
[41] C. Munk: ”Dynamic Asset Allocation”. Lecture Notes, University of
Southern Denmark, March 2008
[42] B. Øksendal: ”Stochastic Differential Equations”. Springer, 2005
Graduate Texts in Mathematics, vol.107, Springer Verlag, N.Y. 1993
[44] L. V. Ovsiannikov: ”Group Analysis of Differential Equations” Aca-
demic Press, New York 1982
[45] D. Papanicolaou: ”Investment-Specific Technological Change and
Asset Prices”. Kellogg School of Management, Northwestern University
preprint 2008.
[46] H. Pham: ”Continuous-time Stochastic Control and Optimization with
Financial Applications”. Springer 2009
[47] M. Puhle: ”Bond Portfolio Optimization”. Springer 2008
[48] G. Silberberg: ”Discrete Symmetries of the Black-Scholes Equation”.
1033
Proceedings of 10th Int.Conference in Modern Group Analysis.
[49] H. M. Soner: ”Stochastic Optimal Control in Finance”.
Lectures at Pisa, 2003
[50] J. L. Stein: ”Stochastic Optimal Control, International Finance and
Debt Crises”. Oxford University Press 2006
[51] H. Stephani: ”Differential Equations:Their Solutions Using Symme-
tries”. Cambridge University Press, 1989
[52] N. Touzi: ”Stochastic Optimal Control Problems, Viscosity Solutions,
and Application to Finance”. Lectures at Pisa, 2002
[53] T. Zariphopoulou: ”Optimal investment and consumption models
with non-linear stock dynamics”.
Math. Methods Oper. Res. 50 (1999) 271-296
72 A General Solution to the Stochastic

Equity Volatility Equation
(The BCG Equation)
We find a general solution to a non-linear PDE modeling the stochastic eq-
uity volatility as a function of the stock price and time. This equation was
introduced by Bensoussan, Crouhy and Galai (BCG). Our solution relies on
the exploitation of the underlying Lie symmetries of the given equation. We
also determine the classical Lie symmetries of a similar in nature non-linear
PDE which models the volatility of the stock price as a function of the total
asset value and time.
1. Introduction. The Black-Scholes option pricing model (Black and Sc-
holes [10]) was derived on the assumption of constant volatility. However
this assumption does not appear to be true in everyday market transactions,
giving rise to the so-called volatility smile (see for example Bollersley, Ray
and Kroner [12]). The volatility dynamics of the stochastic volatility models
(Fouque, Papanicolaou and Sircar [15], Heston [18], Lewis [23] and Anto-
niou [3]) or the stochastic volatility models combined with jump diffusion
processes (Bates [7] and Kou [22]) appears to be exogenous and cannot be
justified on economic grounds.
We should mention that the Heston model (Heston [18]) has been generalized
by S. Antoniou (Antoniou [3]) where in this broad class of models have been
1034
considered time-dependent coefficients, since in this case we can capture a
more detailed movement of the option price under stochastic volatility. It
is the Bensoussan-Crouhy-Galai (BCG) approach introduced some time ago
(Bensoussan, Crouhy and Galai [8] and [9]) that considers the stochastic na-
ture of volatility to be endogenous and comes from the impact of a change
in the value of the firm’s assets on the financial leverage.
Let us briefly sketch the BCG approach. Let V be the firm assets which
follow a log-normal process
dV = µV V dt + σ̄V dBt (72.1)
where µV and σ̄ are the asset’s return and volatility respectively. The volatil-
ity σ̄ is considered as being constant and Bt is a standard Brownian motion.
Let also S = S(V, t) be the equity of the firm. The equity is considered a
call option on the assets of the firm with a strike price K equal to the face
value of the firm liabilities and an expiration T equal to the average liability
maturity.
Therefore S follows the process
dS = µSdt + σ(S, t)S dBt (72.2)
and satisfies an equation similar to the Black-Scholes equation:

1
St + σ̄ 2 V 2 SV V + rV SV − rS = 0 (72.3)
2
Using Ito’s Lemma, we have
∂S ∂S 1 ∂ 2S
dS = dt + dV + dV ·dV (72.4)
∂t ∂V 2 ∂V 2
where dV ·dV = σ̄ 2 V 2 dt using the Ito-McKean rules. Therefore
∂S ∂S 1 ∂ 2S 2 2
dS = dt + (µV V dt + σ̄V dBt ) + σ̄ V dt
∂t ∂V 2 ∂V 2
∂S ∂S 1 ∂ 2S 2 2 ∂S
dS = + µV V + 2
σ̄ V dt + σ̄V dBt (72.5)
∂t ∂V 2 ∂V ∂V
1035
Comparing the coefficients of Brownian motion terms between (72.2) and
∂S
(72.5), we obtain the relation σ(S, t)S = σ̄V from which we get
∂V
∂S V
σ(S, t) = σ̄ · (72.6)
∂V S
The above equation is an implicit functional form for σ(S, t). Upon elimi-
nation of the asset term, Bensoussan, Crouhy and Galai have been able to
derive the following nonlinear partial differential equation
∂σ 1 2 2 ∂ 2 σ 2 ∂σ
+ σ S + (r + σ )S =0 (72.7)
∂t 2 ∂S 2 ∂S
which models the stochastic equity volatility as a function of the firm’s asset.
A brief sketch of the proof of this equation is given in Appendix A.
We also introduce the volatility of S as a function of V , denoted by σ̃(V, t)
and defined by
V
σ̃(V, t) = σ̄SV (V, t) (72.8)
S(V, t)
This function satisfies the nonlinear partial differential equation
∂ σ̃ 1 2 2 ∂ 2 σ̃ ∂ σ̃
+ σ̄ V 2
+ (r + σ̄σ̃)V =0 (72.9)
∂t 2 ∂V ∂V
A sketch of the proof of the above equation is given in Appendix B.
Introducing further the function θ̃(V, t) by
θ̃(V, t) = S(V, t)(σ̃(V, t) − σ̄) (72.10)
it is proved that θ̃(V, t) satisfies the equation
1 2 2 ∂ 2 θ̃ ∂ θ̃
θ̃t + σ̄ V 2
+ rV − rθ̃ = 0 (72.11)
2 ∂V ∂V
For a proof see Appendix C.
We also obtain from equation (72.10) that
θ̃(V, t)
σ̃(V, t) = + σ̄ (72.12)
S(V, t)
1036
which provides another solution strategy for determining the function σ̃(V, t),
solving the BS-type of equations (72.3) and (72.11), instead of solving directly
equation (72.9).
The core of the BCG model relies on PDE (72.7) (see equation (11) of [3])
that models the stochastic equity volatility as a function of the firm’s asset
and the time. This equation is a nonlinear PDE, where the unknown function
is the stochastic equity volatility function σ(S, t), S being the firm’s asset
and t the time variable.
We have found a general solution to the Bensoussan-Crouhy-Galai equation
(72.7), long believed to be as one not having an analytical solution. In solv-
ing this equation, we have found the Lie point symmetries of the equation.
We then determined two invariants, which allowed us to convert the equation
into an ordinary, nonlinear, second order DE. Under two successive transfor-
mations, we transformed this last DE into an Emden-Fowler equation, which
admits an analytical solution given by the error function erfi(x) of imaginary
argument. We have also determined the Lie point symmetries of the second
equation (72.9) for further investigation.
Despite the fact that the BCG approach is -to the best of our knowledge- one
of the most serious candidates in modeling the stochastic equity volatility,
it has not received proper attention. That might come from the non-linear
nature of the equations involved and the lack of analytical solutions. A fur-
ther reformulation of the BCG approach, using the closed-form solution of
the above-mentioned equations is under current study.
The paper is organized as follows:
In Section 2 we determine the Lie symmetries of differential equation (72.7).
In Section 3 we find a general solution of the BCG equation, taking into ac-
count the generators of the Lie symmetry. In appendix A we derive the BCG
equation (72.7). In Appendix B we derive equation (72.9). In Appendix C
we derive equation (72.11).
In Appendix D we prove an auxiliary Theorem that helps to convert an or-
dinary DE into an Emden-Fowler equation. In Appendix E we solve the
Emden-Fowler equation. In Appendix F we give a brief description on how
to find the inverse of the error function. In Appendix G we consider another
general solution of the equation, which is based on a generator not considered
in deriving the general solution in Section 3, using a paraphrasis of the so-
called tanh-method. In Appendix H we list the Lie symmetries of the second
equation, which can be used for the determination of a general, closed-form
solution.
1037
2. Lie Symmetries of the BCG Equation. Lie’s method of symmetry
is a well-established method initiated by S. Lie (see the original references
Lie [24] and [25]). There is however a more recent literature (Anderson and
Ibragimov [2], Bluman and Kumei [11], Dresner [14], Hereman [12], Ibrag-
imov [15], [16], Olver [26], Ovsiannikov [27], Stephani [28]). In principle,
this method can be used both for ODEs and PDEs (or systems of these
equations), linear or nonlinear, in many areas of science and engineering. In
finance, the first paper which applied Lie symmetry methods, was that by
N. H. Ibragimov and R. K. Gazizov (Ibragimov and Gazizov [21]). See also
the papers by S. Antoniou (Antoniou [4], [5] and [6] and references therein).
Theorem 1. Equation (72.7) can be transformed into the equation

∂u 1 2 ∂ 2 u 1 2 ∂u
+ u + r + u =0 (72.13)
∂t 2 ∂x2 2 ∂x
where σ(S, t) = u(x, t) and x = ln(S).
Proof. We consider the BCG equation (72.7). We make the substitution
x = ln(S) (72.14)
Since
∂ 2σ ∂ 2 u ∂u ∂σ ∂u
S2 = − and S = (72.15)
∂S 2 ∂x2 ∂x ∂S ∂x
the BCG equation (72.7) takes the form (72.13).
Theorem 2. The generators of the Lie Symmetries of equation (72.13) are
given by
∂ ∂ ∂ 1 ∂
X1 = , X2 = rt+t − u
∂t ∂x ∂t 2 ∂u (72.16)
∂ ∂ ∂
X3 = −e−(x−rt) + e−(x−rt) u , X4 =
∂x ∂u ∂x
Proof. Let ∆(x, t, u(2) ) be defined by
∂u 1 2 ∂ 2 u 1 2 ∂u
∆(x, t, u(2) ) = + u + r + u (72.17)
∂t 2 ∂x2 2 ∂x
Introduce the vector field X (the generator of the symmetries) by
∂ ∂ ∂
X = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u) (72.18)
∂x ∂t ∂u
1038
The second prolongation of the vector field is defined in our case by the
equation
∂ ∂ ∂ ∂ ∂
P r(2) X = X + φx + φt + φxx + φxt + φtt (72.19)
Invariance of the equation under the symmetry of the vector field X is imple-
mented by the equation P r(2) X[∆(x, t, u(2) )] = 0, as long as ∆(x, t, u(2) ) = 0.
We have
∂u 1 2 ∂ 2 u

(2) (2) (2) 1 2 ∂u
P r X[∆(x, t, u )]⇐⇒P r X + u + r+ u
∂t 2 ∂x2 2 ∂x (72.20)
1 1
⇐⇒φ(uuxx + uux ) + r + u2 φx + φt + u2 φxx
2 2
The coefficients φx , φt and φxx are calculated to be (Olver [26], Example
2.38)
φx = φx + (φu − ξx )ux − ξu u2x − τx ut − τu ux ut (72.21)
φt = φt + (φu − τt )ut − τu u2t − ξt ux − ξu ux ut (72.22)

The functions τ , ξ and φ have the following dependence
τ = τ (t), ξ = ξ(x, t), φ = φ(x, t, u) (72.24)
The above dependence can also be proved rigorously by fully expanding

(72.20) and equating to zero all the coefficients of the partial derivatives
of the function u. Expression (72.20) thus becomes
1
φ(uuxx + uux ) + r + u2 [φx + (φu − ξx )ux ]
2
+ φt + (φu − τt )ut − ξt ux (72.25)

1 2 2
+ u φxx + (2φxu − ξxx )ux + φuu ux + (φu − 2ξx )uxx
2
1039
1
In the above expression we substitute ut by − 12 u2 uxx − r+ u2 ux to account
2
for the condition ∆ = 0. We thus see that P r(2) X[∆(x, t, u(2) )] = 0 is
∆=0
equivalent to the equation
1 2
φ(uuxx + uux ) + r + u [φx + (φu − ξx )ux ]
2
1 1 2
2
+ φt − ξt ux + (φu − τt ) − u uxx − r + u ux (72.26)
2 2
1 2 2
+ u φxx + (2φxu − ξxx )ux + φuu ux + (φu − 2ξx )uxx = 0
2
All the coefficients of the partial derivatives of the function u have to be set
equal to zero:
Coefficient of u2x
1 2
u φuu = 0 (72.27)
2
Coefficient of uxx
1 1
uφ − u2 (φu − τt ) + u2 (φu − 2ξx ) = 0 (72.28)
2 2
Coefficient of ux
1 1
uφ + r + u2 (φu − ξx ) − ξt − r + u2 (φu − τt )
2 2 (72.29)
1 2
+ u (2φxu − ξxx ) = 0
2
Coefficient of zero-th order derivative of u
1 2 1
r + u φx + φt + u2 φxx = 0 (72.30)
2 2
We now have to solve the system of the determining equations (72.27)-
(72.30). From (72.27) we get φuu = 0, which means that φ is a linear function
with respect to u:
φ(x, t, u) = α(x, t) u + β(x, t) (72.31)
1040
From equation (72.28), after canceling opposite terms and dividing by u, we
obtain
1
φ + uτt − uξx = 0 (72.32)
2
In the above equation we substitute φ by the expression given by (72.31) and
1
α + τt − ξx u + β = 0 (72.33)
2
1
α + τt − ξx = 0 and β = 0 (72.34)
2
from equation (72.30), after substituting φ by (72.31) (taking into account
that β = 0), we obtain
1
(rαx + αt )u + (αx + αxx )u3 = 0
2
From the previous equation we obtain the system
rαx + αt = 0, αx + αxx = 0 (72.35)
Equation (72.29), after substitution of φ by (72.31) (with β = 0), can be

written as

1 1
[ r(τt − ξx ) − ξt ] + α + (τt − ξx ) + (2αx − ξxx ) u2 = 0
2 2
from the above equation we obtain the equations
r(τt − ξx ) − ξt = 0 (72.36)
and
1 1
α + (τt − ξx ) + (2αx − ξxx ) = 0 (72.37)
2 2
Since (from the first of (72.34))
1
ξx = α + τt , ξxx = αx (72.38)
2
1041
1
αx + α = − τ t (72.39)
2
The previous equations admits the general solution
1
α(x, t) = − τ 0 (t) + g(t)e−x (72.40)
2
where g(t) is a function to be determined. We can prove that the solution
(72.40) is compatible to the second of (72.35). After substituting (72.40) into
the first of (72.35), we get the equation
1
− τ 00 (t) + [g 0 (t) − rg(t)]e−x = 0 (72.41)
2
τ 00 (t) = 0 (72.42)
and
g 0 (t) − rg(t) = 0 (72.43)
Therefore
τ (t) = a1 + a2 t (72.44)
and
g(t) = a3 ert (72.45)
We thus obtain from (72.40)

1
α(x, t) = − a2 + a3 e−(x−rt) (72.46)
2
From (72.31), taking into account the above expression of α(x, t), we get
(with β = 0)
1
φ(x, t, u) = − a2 u + a3 e−(x−rt) u (72.47)
2
1042
From the first of (72.38), taking into account (72.46) and (72.44), we have
ξx = a3 e−(x−rt) and by integration with respect to x,
ξ = −a3 e−(x−rt) + f (t) (72.48)
where f (t) is a function to be determined. From (72.36), using the known

expressions of ξ and τ we can determine the function f (t). The differential
equation we obtain is f 0 (t) = ra2 which, when integrated, gives
f (t) = a2 rt + a4 (72.49)
We are now in a position to derive the function ξ(x, t) from (72.48):
ξ(x, t) = a2 rt − a3 e−(x−rt) + a4 (72.50)
The vector X, the generator of the symmetries, can then be expressed as

∂ ∂ ∂
X = ξ(x, t, u) + τ (x, t, u) + φ(x, t, u)
∂x ∂t ∂u
∂ ∂
= a2 rt − a3 e−(x−rt) + a4 + (a1 + a2 t) (72.51)
∂x ∂t
1 ∂
+ − a2 + a3 e−(x−rt) u
2 ∂u
The previous expression can be written as
∂ ∂ ∂ 1 ∂
X = a1 + a2 rt +t − u
∂t ∂x ∂t 2 ∂u (72.52)
−(x−rt) ∂ −(x−rt) ∂ ∂

+ a3 e +e u + a4
∂x ∂u ∂x
From the above expression we get the generators (of the symmetries of the
equation (72.13)) given by (72.16).
3. A General Solution of the BCG equation (72.7)
After determining the Lie symmetries of the BCG equation, we are now in a
position to find its general solution.
3.1 Some preliminary Theorems. We state and prove the following
Theorem 3. Equation (72.13), under the substitutions
r
r
u(x, t) = v(y), y = x − rt (72.53)
1 + rt
1043
takes the form
vvyy + vvy = 1 (72.54)
Proof. We consider a linear combination of the generators X1 , X2 and X4 :
Γ = λX1 + µX2 + X4 (72.55)
where λ and µ are constants to be determined. We thus consider the gener-

ator
∂ ∂ 1 ∂
Γ = (λ + µt) + (µrt + 1) − µu (72.56)
∂t ∂x 2 ∂u
Any other choice would depend on the boundary or/and initial conditions.
We can now determine the invariants corresponding to generator Γ given by
(72.56). For this purpose, we are going to solve the system
dt dx du
= = (72.57)
λ + µt µrt + 1 1
− µu
2
Two independent solutions of the above system are
1 − λr
x = rt + k ln(λ + µt) + C1 , k= (72.58)
µ
and
u = C2 (λ + µt)−1/2 (72.59)
We thus get the invariants

p
v(y) = λ + µt u, y = x − rt − k ln(λ + µt) (72.60)
Using the above substitutions, we obtain

1 kµ
ut = − µ(λ + µt)−3/2 v − r + (λ + µt)−1/2 vy
2 λ + µt
(72.61)
ux = (λ + µt)−1/2 vy
uxx = (λ + µt)−1/2 vyy
1044
1 1 1
− µv − kµvy + v 2 vyy + v 2 vy = 0 (72.62)
2 2 2
1
Under the choice k = 0, i.e. λ = , the previous equation becomes
r
vvyy + vvy = 1
where we have taken µ = 1 (any other choice might be possible).
Theorem 4. Under the transformation
y = ln(ξ) (72.63)
equation (72.54) is being transformed into the equation
1
ξ 2 vξξ + 2ξvξ = (72.64)
v
Proof. Introducing the transformation (72.63), since
vy = ξvξ and vyy = ξ 2 vξξ + ξvξ (72.65)
equation (72.54) takes the form (72.64).
Theorem 5. equation (72.64) admits a similarity transformation
ξ = η, v = η −1 w, w = w(η) (72.66)
and takes the final form
1
wηη = (72.67)
w
which is an Emden-Fowler type of equation. The proof is in Appendix D.
Theorem 6.
I. The general solution of equation (72.67) is given by
Z
dw
η=± 2
+B (72.68)
ln(w ) + 2A
where A and B are arbitrary constants.
II. The integral appearing in (72.68) can be expressed in terms of the erfi(x),
r r
π 2A + ln(w2 )
−A
η = ±e erfi +B (72.69)
2 2
1045
the error function of complex argument (Appendix F).
Proof.
I. Appendix D. p √
II. Under the substitution ln(w2 ) + 2A = 2 z, i.e.
2
w = e−A ez (72.70)
the integral in (72.68) takes the form

√ −A Z z2
η =± 2e e dz + B (72.71)
from which, taking into account (72.132) and readjusting the constants A
and B, we arrive at (72.69). The following Theorem provides the general
solution of the Emden-Fowler equation (72.67).
Theorem 7. The function w = w(η) which is a solution to (72.67), is given
by
r

−1 2
π −A
w(η) = exp [(erfi) (ζ)] − A , ζ = ± e (η − B) (72.72)
2
The function (erfi)−1 (x) is the inverse error function of complex argument.
Proof. From (72.68) we get
p
erfi( A + ln(w)) = ζ (72.73)
where ζ is defined in (72.72). Inverting (72.73) we find
A + ln(w) = (erfi)−1 (ζ)

p
(72.74)
Solving the previous equation with respect to w, we find w(η) given in (72.72).
3.2. A General Solution of the BCG Equation.
Let us now put everything together. We see that the final solution depends
on the variable η. Therefore we have to find the relation between η and the
original variables.
We have that η = ξ (Theorem 5), ξ = ey (Theorem 4) and y = x − rt
(Theorem 3) where x = ln(S). Therefore η = ey = ex−rt = eln(S)−rt and thus
η = S e−rt (72.75)
1046
Concerning the unknown
r function, we see that σ(S, t) = u(x, t) (Theorem
r 1
1), where u(x, t) = v(y) (Theorem 3) and v(y) = w(η) (Theorem
1 + rt η
4). The function w(η) is given by w(η) = exp [(erfi)−1 (ζ)]2 − A (Theorem
1
−1 2

7). Therefore v(y) = exp [(erfi) (ζ)] − A written as
S e−rt
ert
v(y) = exp [(erfi)−1 (ζ)]2 − A (72.76)
S
Theorem 8. The general solution of equation (72.1) is given by
r ert
r
−1 2
σ(S, t) = exp [(erfi) (ζ)] − A
1 + rt S
r (72.77)
π −A
ζ=± e (η − B), η = S e−rt
2
This is the main result ofrthis section.
π −A
Note. We could set K = e and thus ζ = ±K(η − B) where K is a
2
positive constant (K > 0).
3.3. Inverting the error function.
We now have to find an analytical expression for the inverse of the error func-
tion of imaginary argument. In Appendix F we prove the relation (72.138):
(erfi)−1 (ζ) = −i (erfi)−1 (iζ) (72.78)
where (erfi)−1 (x) is the inverse error function of imaginary argument, imple-
mented in many Symbolic Languages. Using (72.133) and (72.138) we prove
that (erfi)−1 (x) has the series expansion (72.140).
Appendix A. We shall prove equation (72.1). We first prove the following
Lemma. The function V (S, t) satisfies the equation
1
Vt + σ 2 S 2 VSS + rSVS − rV = 0 (72.79)
2
Proof. The function V (S, t) is the inverse of the function S(V, t). Differen-
tiation of the equation S≡S(V (S, t), t) with respect to S, we find
1 = SV VS (72.80)
1047
Differentiation of (72.80) with respect to S, we get
0 = SV VSS + SV V (VS )2 (72.81)
Therefore
SV VSS VSS
SV V = − = − (72.82)
(VS )2 (VS )3
We also have
∂S ∂V Vt
St = − · =− (72.83)
∂V ∂t VS
We thus see that equation (72.3), i.e.
1
St + σ̄ 2 V 2 SV V + rV SV − rS = 0
2
becomes
Vt 1 V
SS 1
− + σ̄ 2 V 2 − 3
+ rV − rS = 0 (72.84)
VS 2 (VS ) VS
On the other hand we obtain from (72.6) that
1 V
σ(S, t) = σ̄ (72.85)
VS S
from which there follows
σ̄ 2 V 2 = σ 2 S 2 (VS )2 (72.86)
Vt 1 2 2 2
V
SS 1
− + σ S (VS ) − + rV − rS = 0 (72.87)
VS 2 (VS )3 VS
which is equivalent to (72.79).

Proof of (72.7). We now proceed in proving equation (72.7). We first define
the auxiliary function
ψ(S, t) = ln(V (S, t)) (72.88)
1048
It then follows from equation (72.85) that
σ̄ 1
σ(S, t) = (72.89)
S ψS
From the definition of ψ(S, t) we obtain

Vt VS VSS VS 2 VSS
ψt = , ψS = , ψSS = − = − (ψS )2 (72.90)
V V V V V
Therefore we have from the previous three equations
Vt = V ψt , VS = V ψS , VSS = V [ψSS + (ψS )2 ] (72.91)
Substituting the relations (72.91) and (72.89) into (72.79), we obtain the
equation
1 2 ψSS
ψt + σ̄ + 1 + rSψS − r = 0 (72.92)
2 ψS2
Since (as follows from (72.89))

σ̄
ψS = (72.93)
σS
we obtain by differentiation with respect to S
σ̄(σ + SσS )
ψSS = − (72.94)
σ2S 2
Therefore we have, using the previous two expressions
ψSS σ + SσS
2
=− (72.95)
ψS σ̄
From equation (72.92), using (72.93) and (72.95), we get

1 rσ̄
ψt = − [σ̄ 2 − σ̄(σ + SσS )] − +r (72.96)
2 σ
Differentiation of the previous equation with respect to S, gives us
1 σS
ψtS = σ̄(2σS + SσSS ) + rσ̄ 2 (72.97)
2 σ
1049
Differentiation of (72.93) with respect to t we obtain
σ̄ σt
ψSt = − (72.98)
S σ2
Equating the two expressions (72.97) and (72.98) we obtain that
σ̄ σt 1 σS
− 2
= σ̄(2σS + SσSS ) + rσ̄ 2
Sσ 2 σ
The above is equivalent to
1
σt + S 2 σ 2 σSS + S(r + σ 2 )σS = 0 (72.99)
2
The above is equation (72.7).
Appendix B. We shall prove equation (72.9), i.e.
∂ σ̃ 1 2 2 ∂ 2 σ̃ ∂ σ̃
+ σ̄ V 2
+ (r + σ̄σ̃)V =0
∂t 2 ∂V ∂V
We define the function
φ(V, t) = ln(S(V, t)) (72.100)
From the definition of the function φ(V, t) we obtain

St SV SV V S 2 S
V VV
φt = , φV = , φV V = − = − (φV )2 (72.101)
S S S S S
Therefore we obtain from the above three relations that
St = Sφt , SV = SφV , SV V = S [ φV V + (φV )2 ] (72.102)
using the previous three relations, equation (72.3) takes on the form
1
φt + σ̄ 2 V 2 ( φV V + (φV )2 ) + rV φV − r = 0 (72.103)
2
We also get from (72.8), using the definition (72.100) that
σ̃(V, t) = σ̄V φV (72.104)
from the above relation we obtain

σ̃
φV = (72.105)
σ̄V
1050
and then
1
φV V = (V σ̃V − σ̃) (72.106)
σ̄V 2
Upon substitution of (72.105) and (72.106) into (72.103), we get the equation
1 r
φt = − [σ̄(V σ̃V − σ̃) + σ̃ 2 ] − σ̃ + r (72.107)
2 σ̄
Upon differentiation of the previous equation with respect to V we obtain
1 r
φtV = − (σ̄V σ̃V V + 2σ̃σ̃V ) − σ̃V (72.108)
2 σ̄
Differentiation of (72.105) with respect to t yields
σ̃t
φV t = (72.109)
σ̄V
Equating the two expressions (72.108) and (72.109) we get
σ̃t 1 r
= − (σ̄V σ̃V V + 2σ̃σ̃V ) − σ̃V (72.110)
σ̄V 2 σ̄
which is is equivalent to
∂ σ̃ 1 2 2 ∂ 2 σ̃ ∂ σ̃
+ σ̄ V + (r + σ̄σ̃)V =0
∂t 2 ∂V 2 ∂V
The above is equation (72.9).
Appendix C. We shall prove equation (72.11), i.e.
1 ∂ 2 θ̃ ∂ θ̃
θ̃t + σ̄ 2 V 2 2
+ rV − rθ̃ = 0
2 ∂V ∂V
From the defining equation (72.10) we obtain successively
θ̃t = St (σ̃ − σ̄) + S σ̃t

θ̃V = SV (σ̃ − σ̄) + S σ̃V (72.111)
θ̃V V = SV V (σ̃ − σ̄) + 2SV σ̃V + S σ̃V V
1051
Therefore
1 ∂ 2 θ̃ ∂ θ̃
θ̃t + σ̄ 2 V 2 + rV − rθ̃
2 ∂V 2 ∂V
= [St (σ̃ − σ̄) + S σ̃t ]
1
+ σ̄ 2 V 2 [SV V (σ̃ − σ̄) + 2SV σ̃V + S σ̃V V ]
2
(72.112)
+ rV [SV (σ̃ − σ̄) + S σ̃V ] − rS(σ̃ − σ̄)

1 2 2
= (σ̃ − σ̄) St + σ̄ V SV V + rV SV − rS
2

1 2 2
+ S σ̃t + σ̄ V σ̃V V + rV σ̃V + σ̄ 2 V 2 SV σ̃V
2
Since
1
St + σ̄ 2 V 2 SV V + rV SV − rS = 0, equation (72.3) (72.113)
2
and
1
σ̃t + σ̄ 2 V 2 σ̃V V + rV σ̃V = −σ̄σ̃V σ̃V , from equation (72.9) (72.114)
2
equation (72.112) gives
1 ∂ 2 θ̃ ∂ θ̃
θ̃t + σ̄ 2 V 2 + rV − rθ̃
2 ∂V 2 ∂V
(72.115)
= S(−σ̄σ̃V σ̃V ) + σ̄ 2 V 2 SV σ̃V
= σ̄V σ̃V (−S σ̃ + σ̄V SV ) = 0
where in the last stage we used equation (72.8).

Let us now prove the equation
1
θt + σ 2 S 2 θSS + rSθS − rθ = 0 (72.116)
2
This equation appears in BCG [4] without proof and is being used as an
approximation method in solving equation (72.7). Since θ(S, t) is defined by
θ(S, t) = S [ σ(S, t) − σ̄ ], we have
θt = S σt , θS = σ − σ̄ + SσS
(72.117)
θSS = 2σS + SσSS
1052
Therefore
1
θt + σ 2 S 2 θSS + rSθS − rθ
2
1
= S σt + σ 2 S 2 (2σS + SσSS ) + rS(σ − σ̄ + SσS ) − rS(σ − σ̄) (72.118)
2
1 2 2 2
= S σt + σ S σSS + S(r + σ )σS = 0
2
where in the last step we used equation (72.7).
Appendix D. Proof of Theorem 5. We try a transformation of the form
ξ = ηα, v = η β w, w = w(η) (72.119)
Under this transformation we find that the partial derivatives are transformed
into
2β β−1 1/α 2
2ξvξ = η ξ w + η β ξ 1/α wη (72.120)
α α
and

2 β(β − 1) β1
ξ vξξ = ξ+ − 1 η η β−2 ξ 1/α w
α α α
(72.121)
2β 1/α 1 1
β−1 1/α 1 β 2/α
+ ξ + − 1 η η ξ wη + 2 η ξ wηη
α2 α α α
We thus find that under the transformation (72.119), differential equation
(72.64) takes the form

β(β − 1) β1
ξ+ − 1 η η β−2 ξ 1/α w
α α α

2β 1/α 1 1 1
+ 2
ξ + − 1 η η β−1 ξ 1/α wη + 2 η β ξ 2/α wηη (72.122)
α α α α
2β β−1 1/α 2
+ η ξ w + η β ξ 1/α wη = η −β w−1
α α

β(β − 1) β1
ξ+ + 1 η η β−2 ξ 1/α w
α α α

2β 1/α 1 1 1
+ 2
ξ + + 1 η η β−1 ξ 1/α wη + 2 η β ξ 2/α wηη (72.123)
α α α α
2β β−1 1/α 2
+ η ξ w + η β ξ 1/α wη = η −β w−1
α α
1053
The coefficients of w and wη of the previous equation vanish simultane-
ously under the choice
α = 1 and β = −1 (72.124)
Therefore equation (72.64) under the transformation
ξ = η, v = η −1 w (72.125)
takes the form
wηη = w−1 (72.126)
This is an Emden-Fowler type of equation.

Appendix E. In this Appendix we shall solve the Emden-Fowler equation
(72.67). Consider the equation
d2 y 1
2
= (72.127)
dx y
dy
Multiplying this equation by 2 , we get the equation
dx
dy d2 y dy 1
2 = 2
dx dx2 dx y
d dy 2 dy 1
=2
dx dx dx y
dy 2 R dy
and integrating with respect to x we find = 2 from which we
dx y
dy 2
obtain = ln(y 2 ) + 2A where A is an integration constant. Therefore
dx
dy p
we have = ± ln(y 2 ) + 2A and by a further integration
dx
Z
dy
± p +B =x (72.128)
ln(y 2 ) + 2A
1054
where B is another integration constant.
Note. The general form of the Emden-Fowler equation
d2 y
= axm y n (72.129)
dx2
can be solved for a variety of choices for the exponents m and n using Lie
Symmetry Analysis.
Appendix F. The error function is denoted by erf(x) and it is defined by
the integral
Z z
2 2
erf(z) = √ e−x dx (72.130)
π 0
The error function of imaginary argument is denoted by erfi(x) and it is
defined by
erfi(x) = −i erf(ix) (72.131)
We can prove, using (72.130) and (72.131) that
Z z
2 2
erfi(z) = √ ex dx (72.132)
π 0
which means that erfi(x) is a real-valued function. The inverse of the error
function is a real-valued function given by
∞ √
−1
X ck π 2k+1
erf (x) = x
k=0
2k + 1 2
k−1
(72.133)
X cm ck−1−m
c0 = 1, ck =
m=0
(m + 1)(2m + 1)
and is being implemented in many of the well-known Symbolic Languages.

The first few terms of the previous expansion (72.133) are given by
√
−1 π π 7π 2 5 127π 3 7
erf (x) = x + x3 + x + x + ··· (72.134)
2 12 480 40320
In order to determine the function w(η) we have to invert the erfi(x) function.
Let us explain briefly on how to invert the erfi(x) function. Suppose we want
to invert the function
erfi(x) = z (72.135)
1055
Since, according to (72.131), erfi(x) = −i erf(ix), equation (72.135) can be
written as −i erf(ix) = z and thus erf(ix) = iz, from which we get ix =
erf−1 (iz) or
x = −i erf−1 (iz) (72.136)
Inverting directly (72.135), we have
x = erfi−1 (z) (72.137)
Comparing (72.136) and (72.137) we obtain the formula
erfi−1 (z) = −i erf−1 (iz) (72.138)
Using the series expansion (72.133) of the error function, we get from (72.136)
that
∞ √
X ck π 2k+1
x = −i iz (72.139)
k=0
2k + 1 2
Since −i·i2k+1 = i2k+4 = i2k = (−1)k , taking into account (72.139) and
(72.137), we get
∞ √
−1
X (−1)k ck π 2k+1
erfi (z) = z (72.140)
k=0
2k + 1 2
Details about the error function and its inverse can be found for example in
references [1], [13] and [29].
Appendix G. In trying to find a general solution of the BCG equation
(72.13), we have used a linear combination of the generators X1 , X2 and X4 .
It is obvious that the generator X3 gives rise to the so-called tanh-method
of traveling-wave solutions (see for example [13] and references therein). In
this case we introduce a new variable ξ by ξ = x − rt. Using the fact that
ut = −rU 0 (ξ), ux = U 0 (ξ), uxx = U 00 (ξ)

1 1
−rU 0 (ξ) + U 2 U 00 + r + U 2 U 0 = 0
2 2
1056
which is equivalent to U 00 + U 0 = 0 and admits the general solution
U = C1 + C2 e−ξ (72.141)
This solution however is not of any practical importance (see the beginning
of section 3 on p.66 of reference [8]).
Appendix H. Lie Symmetries of the BCG equation (72.9). The BCG
equation (72.9) under the substitution x = ln(V ) takes on the form
1 1
ut + σ̄ 2 uxx + r − σ̄ 2 + σ̄u ux = 0 (72.142)
2 2
where σ̃(V, t) = u(x, t) with x = ln(V ).
The generators of the Lie symmetries of equation (72.142) are given by
∂ ∂ ∂ ∂
X1 = , X2 = (x − ρt)− 2t + u
∂t ∂x ∂t ∂u
∂ ∂ ∂
X3 = −σ̄tx + σ̄t2 + (−σ̄tu + x − ρt) (72.143)
∂x ∂t ∂u
∂ ∂
X4 = , X5 =
∂u ∂x
1
where ρ = r − σ̄ 2 . For a proof see the paper in Researchgate.
2
The above generators can be used in finding a solution of the BCG equation
(72.2) along the lines of Section 3. However one can use (72.12) in determin-
ing σ̃(V, t).
References
[1] M. Abramowitz and I. A. Stegun: ”Handbook of Mathematical
Functions”. Dover, 1972
[2] R. L. Anderson and N. H. Ibragimov: ”Lie-Bcklund Transforma-
tions in Applications”. SIAM, Philadelphia 1979
[3] S. Antoniou: ”Stochastic Volatility models with closed-form solutions”.
SKEMSYS report (available from Researchgate).
[4] S. Antoniou: ”Portfolio Optimization and Stochastic Control. Closed-
form solutions to the HJB equation through Lie Symmetries”.
[5] S. Antoniou: ”Pricing Asian Options. The Geometric Average Case”.
1057
[6] S. Antoniou: ”The Arithmetic Average Case of Asian Options: Reduc-
ing the PDE to the Black-Scholes Equation”.
[7] D. Bates: ”Jumps and Stochastic Volatility: Exchange Rate Processes
Implicit in PHLX Deutsch Mark Options”.
Review of Financial Studies 9 (1996) 69-107
[8] A. Bensoussan, M. Crouhy and D. Galai: ”Stochastic equity volatil-
ity related to the leverage effect. I. Equity volatility behaviour”. Applied
Mathematical Finance 1 (1994) 63-85
[9] A. Bensoussan, M. Crouhy and D. Galai: ”Stochastic equity volatil-
ity related to the leverage effect II: valuation of European equity options and
warrants”. Applied Mathematical Finance 2 (1995) 43-59
[10] F. Black and M. Scholes: ”The pricing of options and corporate
liabilities”. Journal of Political Economy 81 (1973) 637-654
Springer Verlag, 1989.
[12] T. Bollerslev, Y. Chou Ray and K. F. Kroner: ”ARCH modeling
in finance:a review of the theory and empirical evidence”. Journal of Econo-
metrics 52 (1992) 5-59
[13] L. Carlitz: ”The inverse of the error function”.
Pacific J. Math. 13 (1963) 459-470
[14] L. Dresner: ”Applications of Lies Theory of Ordinary and Partial Dif-
ferential Equations”. IOP Publishing, 1999
[15] J-P. Fouque, G. Papanicolaou and K. R. Sircar: ”Derivatives in
Financial Markets with Stochastic Volatility”. CUP 2000.
Lie Symmetries of Differential Equations”.
Euromath Bulletin 2, No.1 Fall 1993
[17] W. Hereman and W. Malfliet: ”The Tanh method: A Tool to Solve
Nonlinear Partial Differential Equations with Symbolic Software”.
Institute of Mines report
[18] S. Heston: ”A closed-form Solution for Options with Stochastic
Volatility with Applications to Bond and Currency Options”.
Review of Financial Studies 6 (1993) 327-343
Differential Equations”. Vol.1 1994, Vol.2 1995, Vol.3 1996
1058
CRC Press, Boca Raton, FL.
[22] S. Kou: ”A jump diffusion model for option pricing”.
Management Science 48 (2002) 1086-1101
[23] A. Lewis: ”Option Valuation under Stochastic Volatility with Mathe-
matica Code”. Finance Press 2000.
[24] S. Lie: ”Gesammelte Werke”. Vols. 1-7. F. Engel and P. Hergard
(Eds.) Teubner, Leipzig 1899
[26] P. J. Olver: ”Applications of Lie Groups to Differential Equations”,
Graduate Texts in Mathematics, vol.107, Springer Verlag, N.Y. 1993
Academic Press, New York 1982
[28] H. Stephani: ”Differential Equations:Their Solutions Using Symme-
[29] Wikipedia: ”Error Function”. Available online.
Acknowledgements. I would like to thank Prof. A. Bensoussan for kindly

reading a preliminary draft of this investigation.
73 Asian Options.
The Geometric Average Case
We consider the valuation of path-dependent contingent claims where the
underlying asset follows a geometric average process. Considering the no-
arbitrage PDE of these claims, we first determine the underlying Lie point
symmetries. After determination of the invariants, we transform the PDE to
the Black-Scholes (BS) equation with time-dependent coefficients. We then
transform the BS equation into the heat equation which is solved using Pois-
son’s formula taking into account the payoff condition. We thus obtain a
closed-form solution for the pricing of Asian Options in the geometric aver-
age case. This procedure appears for the first time in the finance literature.
1. Introduction. A path-dependent option is an option whose payoff de-
pends on the past history of the underlying asset. In other words these
1059
options have payoffs that do not depend on the asset’s value at expiry. Two
very common examples of path-dependent options are Asian options and
lookback options.
The terminal payoff of an Asian option depends on the type of averaging
of the underlying asset price over the whole period of the option’s lifetime.
According to the way of taking average, we distinguish two classes of Asian
options: arithmetic and geometric options.
A lookback option is another type of path-dependent option, whose payoff
depends on the maximum or minimum of the asset price during the lifetime
of the option.
The valuation of path-dependent (Asian) European options is a difficult prob-
lem in mathematical finance. There are only some simple cases where the
price of path-dependent contingent claims can be obtained in closed-form
(Angus [3], Barucci et al. [8], Bergman [9], Geman and Yor [23], Kemma
and Vorst [33], Rubinstein [42], Wilmott [49], Zhang [51]). However in both
cases (geometric and arithmetic average) there are some numerical proce-
dures available (Barucci et al. [8], Haug [25]).
In the case of arithmetic average there is no closed-form solution. However
in one of our previous paper (Antoniou [6]) we have been able to convert the
PDE (73.12) into the BS equation using symmetry methods.
If the underlying asset price follows a lognormal stochastic process, its geo-
metric average has a lognormal probability density and in this case there is
a closed-form solution (Angus [3]).
In this paper we provide a rigorous justification of the transformation which
converts equation (73.14) into the BS equation, using symmetry methods.
On the other hand, we provide our own method of solution to the time-
dependent BS equation.
The paper is organized as follows: In Section 2 we consider the general
problem of pricing the path-dependent contingent claims. In Section 3 we
consider the Lie point symmetries of the partial differential equation (equa-
tion (73.19)) for the path-dependent geometric average options. Full details
of the calculation are provided in Appendix A. In Section 4, considering
one invariant of the PDE, we convert the equation into the time-dependent
Black-Scholes equation. In section 5, we transform the BS equation to the
heat equation. In section 6 we find a closed-form solution, using Poisson’s
formula and the payoff condition.
Because of the complexity of the calculations, we have included a great deal
of calculation details, supplemented by three Appendices.
1060
2. Path-Dependent Contingent Claims. Suppose that an option pays
off at expiration time T an amount that is a function of the path taken by
the asset between time zero and T. This path-dependent quantity can be
represented by an integral of some function of the asset over the time period
0≤τ ≤T
Z T
A(T ) = f (S, τ )dτ (73.1)
0
where f (S, t) is a suitable function. Therefore for every t, 0≤t≤T , we have

Z t
A(t) = f (S, τ )dτ (73.2)
0
or, in differential form,
dA = f (S, t)dt (73.3)
We have thus introduced a new state variable A, which will later appear in
our PDE. We are now going to derive the PDE of pricing the path-dependent
option. To value the contract, we consider the function V (S, A, t) and set up
a portfolio containing on of the path-dependent options and a number ∆ of
the underlying asset:
Π = V (S, A, t) − ∆·S (73.4)
The change in the value of this portfolio is given by, according to Itô’s for-
mula, by
∂V 1 ∂ 2V ∂V ∂V
dΠ = + σ 2 S 2 2 dt + dA + − ∆ dS (73.5)
∂t 2 ∂S ∂A ∂S
Choosing
∂V
∆= (73.6)
∂S
to hedge the risk and using (73.3), we find that
∂V 1 ∂ 2V ∂V
dΠ = + σ 2 S 2 2 + f (S, t) dt (73.7)
∂t 2 ∂S ∂A
1061
This change is risk-free and thus earns the risk-free rate of interest r,
dΠ = rΠdt (73.8)
leading to the pricing equation
∂V 1 ∂ 2V ∂V ∂V
+ σ 2 S 2 2 + f (S, t) + rS − rV = 0 (73.9)
∂t 2 ∂S ∂A ∂S
upon combining (73.7), (73.8) and (73.4), and where ∆ in (73.4) has been
substituted by (73.6) for the value of ∆. The previous PDE is solved by
imposing the condition
V (S, A, T ) = Ω(S, A) (73.10)
where T is the expiration time.

If A is considered to be an arithmetic average state variable, i.e.
Z t
A= S(τ )dτ (73.11)
0
then PDE (73.9) becomes
∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV = 0 (73.12)
∂t 2 ∂S ∂A ∂S
If A is considered to be a geometric average state variable, i.e.
Z t
A= ln(S(τ ))dτ (73.13)
0
∂V 1 ∂ 2V ∂V ∂V
+ σ 2 S 2 2 + ln(S) + rS − rV = 0 (73.14)
∂t 2 ∂S ∂A ∂S
We list below the payoff types we have to consider along with PDEs (73.12)
and (73.14):
For average strike call, we have the payoff V (S, A, T )
max(S − A, 0) (73.15)
1062
For average strike put, we have the payoff
max(A − S, 0) (73.16)
For average rate call, we have the payoff
max(A − E, 0) (73.17)
For average rate put, we have the payoff
max(E − A, 0) (73.18)
where E is the strike price.

This is an introduction to the path-dependent options. Further detaila can
be found in Wilmott’s treatise (Wilmott [49], Vol.2 and Vol.3) on Quantita-
tive Finance.
3. Lie Symmetries of the Partial Differential Equation. The tech-
nique of Lie Symmetries was introduced by S. Lie (Lie [36] and [37]) and
is best described in Anderson [2], Bluman and Anco [11], Bluman and Cole
[12], Bluman and Kumei [13], Campbell [17], Cantwell [18], Cohen [19], Dick-
son [20], Dresner [21], Emanuel [22], Hydon [27], Ibragimov [28] and [29],
Ince [31], Olver [38], Ovsiannikov [39], Page [40], Schwarz [43], Steeb [46],
Stephani [47] and Wulfman [50]. This technique was for the first time used
in partial differential equations of Finance by Ibragimov and Gazizov [30].
The same technique was also used by the author in solving the Bensoussan-
Crouhy-Galai equation (Antoniou [4]) , the HJB equation for a portfolio
optimization problem (Antoniou [5]) and the PDE reduction to the BS equa-
tion in the case of arithmetic average asian options (Antoniou [6]).
There is however a growing lists of papers in applying this method to Fi-
nance. For a non complete set of references, see Antoniou [7], Bordag [14],
Bordag and Chmakova [15], Bordag and Frey [16], Goard [24], Leach et al.
[34], Pooe et al. [41], Silberberg [42], Sinkala et al. [45].
Equation (73.12) of the arithmetic average case has been considered before
(Antoniou [6]) where it was reduced to the Black-Scholes equation using again
group-theoretic methods.
We now consider the PDE (73.14)
∂V 1 ∂ 2V ∂V ∂V
+ σ 2 S 2 2 + ln(S) + rS − rV = 0
∂t 2 ∂S ∂A ∂S
1063
We shall determine the Lie point symmetries of the previous equation. We
follow closely Olver [37]. Let ∆(S, A, t, V (2) ) be defined by
∂V 1 ∂ 2V ∂V ∂V
∆(S, A, t, V (2) )≡ + σ 2 S 2 2 + ln(S) + rS − rV (73.19)
∂t 2 ∂S ∂A ∂S
We introduce the vector field X (the generator of the symmetries) by
∂ ∂ ∂
X = ξ1 (S, A, t, V ) + ξ2 (S, A, t, V ) + ξ3 (S, A, t, V )
∂S ∂A ∂t (73.20)
∂
+ φ(S, A, t, V )
∂V
We calculate in Appendix A the coefficients ξ1 , ξ2 , ξ3 and φ and we find that
the Lie algebra of the infinitesimal transformations of the original equation
(73.14) is spanned by the eight vectors
∂
X1 = ,
∂t
1 ∂ ∂ 3 ∂
X2 = (S· ln(S)) +t + A
2 ∂S ∂t 2 ∂A
σ 2 − 2r (σ 2 − 2r)2

∂
+ ln(S) + t + rt V
4σ 2 8σ 2 ∂V
σ 2 − 2r

∂ ∂ 2 ∂
X3 = [(S· ln(S))t + 3SA] + 3tA +t + 2t ln(S)
∂S ∂A ∂t 4σ 2
(σ 2 − 2r)2 2 3(σ 2 − 2r)

2 2 2 ∂
+ 2 (ln(S)) + 2
t + rt − 2t + 2
A V (73.21)
σ 8σ 2σ ∂V
∂ ∂ ∂
X4 = , X5 = S +t
∂A ∂S ∂A
σ 2 − 2r

∂ 3 ∂ 2 2 ∂
X6 = 2tS +t + 2 (ln(S)) + 2
t V
∂S ∂A σ σ ∂V
2

2 ∂ 3 ∂ 6 3(σ − 2r) 2 6 ∂
X7 = 3t S +t + 2 t(ln(S)) + t − A V
∂S ∂A σ 2σ 2 σ2 ∂V
∂
X8 =V
∂V
and the generator of the infinite-dimensional sub-algebra
∂
Xβ = β(S, A, t) (73.22)
∂V
1064
where β(S, A, t) is an arbitrary solution of the original PDE (73.14).
4. Reduction of the equation. The invariant associated to the generator
X5 is
y = A − t· ln(S) (73.23)
This transformation, in a slightly different form, was used by Angus [3].

However in our case we derived this transformation rigorously using group-
theoretic methods. Considering the substitution
V (S, A, t) = u(y, t) (73.24)
we can transform (73.14) into the equation

1 1
ut + σ 2 − r tuy + σ 2 t2 uyy − ru = 0 (73.25)
2 2
y = ln(z) (73.26)
we transform equation (73.25) to the time-dependent BS-equation

1 2 1 22 1
ut + σ − r t + σ t zuz + σ 2 t2 z 2 uzz − ru = 0 (73.27)
2 2 2
Details are given in Appendix B.
5. Solution of the time-dependent BS-equation. The Black-Scholes
equation
∂V 1 ∂ 2V ∂V
+ σ 2 (t)S 2 2 + r(t)S − r(t)V = 0 (73.28)
∂t 2 ∂S ∂S
with time-dependent coefficients can, in complete analogy to the reasoning
deveped in Appendix C, be converted into the heat equation. The reader is
advised to look at Appendix C before start reading this section. Introducing
the transformation
V (t, S) = e−f (t) U (t, S) (73.29)
where
Z T
f (t) = r(s)ds (73.30)
t
1065
∂U 1 ∂ 2U ∂U
+ σ 2 (t)S 2 2 + r(t)S =0 (73.31)
∂t 2 ∂S ∂S
We consider next the transformation
x = ln(S) + α(t)
(73.32)
τ = β(t)
where α(t) and β(t) are functions to be determined next. Under this trans-
formation, since
∂U ∂U 0 ∂U 0
= β (t) + α (t) (73.33)
∂t ∂τ ∂x
and
∂U ∂U ∂ 2U ∂ 2U ∂U
S = , S2 2
= 2
− (73.34)
∂S ∂x ∂S ∂x ∂x
∂U ∂U 0 1 2 ∂ 2 U ∂U ∂U
β 0 (t) + α (t) + σ (t) − + r(t) =0 (73.35)
∂τ ∂x 2 ∂x2 ∂x ∂x
∂U 0 1 ∂ 2U 1 ∂U
β (t) + σ 2 (t) 2 + α0 (t) − σ 2 (t) + r(t) =0 (73.36)
∂τ 2 ∂x 2 ∂x
The choice
1 1
β 0 (t) = − σ 2 (t) and α0 (t) − σ 2 (t) + r(t) = 0 (73.37)
2 2
converts equation (73.31) into the heat equation
∂U ∂ 2U
= (73.38)
∂τ ∂x2
Therefore the required transformation, following from (73.37), reads
1 T 2
Z Z T
1
α(t) = σ (s)ds and β(t) = r(s) − σ 2 (s) ds (73.39)
2 t t 2
1066
where T is the strike time.
Conclusion. The Black-Scholes equation
∂V 1 ∂ 2V ∂V
+ σ 2 (t)S 2 2 + r(t)S − r(t)V = 0 (73.40)
∂t 2 ∂S ∂S
with time-dependent coefficients, under the transformation
Z T
−f (t)
V (t, S) = e U (t, S), f (t) = r(s)ds (73.41)
t
and
x = ln(S) + α(t), τ = β(t)
1 T 2
Z T
(73.42)
Z
1
α(t) = σ (s)ds, β(t) = r(s) − σ 2 (s) ds
2 t t 2
is converted into the heat equation
∂U ∂ 2U
= (73.43)
∂τ ∂x2
The above argument of solving the time-dependent Black-Scholes equation,
appears for the first time in literature. For another method see Wilmott [49],
Vol.2.
6. The Option Pricing Formula. We shall now derive the option pricing
formula for the Geometric Average case of Asian Options. We shall follow
the next steps:
Step 1. We consider equation (73.27). We shall transform this equation
into the Heat Equation, along the lines of the previous section. Under the
transformation
u(z, t) = e−r(T −t) U (z, t) (73.44)
since
ut = re−r(T −t) U + e−r(T −t) Ut
uz = e−r(T −t) Uz , uzz = e−r(T −t) Uzz
equation (73.27) transforms into the equation

1 2 1 22 1
Ut + σ − r t + σ t z Uz + σ 2 t2 z 2 Uzz = 0 (73.45)
2 2 2
1067
Under the new transformation (see equation (73.159) Appendix C)
x = ln(z) + α(t), τ = β(t) (73.46)
since
Ut = Uτ β 0 (t) + Ux α0 (t)
zUz = Ux , z 2 Uzz = Uxx − Ux

1 1
β 0 (t)Uτ + σ 2 t2 Uxx + σ 2 − r t + α0 (t) Ux = 0 (73.47)
2 2
The choice
1 1
β 0 (t) = − σ 2 t2 and α0 (t) = − σ 2 − r t (73.48)
2 2
Uτ = Uxx (73.49)
The transformations (73.48) give us by integration on the interval [t, T ]:
1 2 T 2
Z
1
β(t) = σ s ds = − σ 2 (T 3 − t3 ) (73.50)
2 t 6
and
1 Z T
2 11 2
α(t) = σ −r sds = − σ − r (T 2 − t2 ) (73.51)
2 t 2 2
Step 2. We shall derive the option pricing formula under the boundary
condition
A +
f (S, A) = exp −K (73.52)
T
From the formula (73.23) we obtain
A y t
= + ln(S) (73.53)
T T T
1068
and then
A y + t ln(S)
exp = exp (73.54)
T T
Since y = ln(z) and ln(z) = x − α(t), with α(t) given by (73.51), we obtain
x − α(t) + t ln(S) +
f (S, A) = exp −K (73.55)
T
Using Poisson’s formula (see for example Tikhonov and Samarskii [48] or
Lebedev [35]) we have
Z +∞
(x − w)2
1 −
U (x, τ ) = √ U (w, 0)e 4τ dw (73.56)
2 πτ −∞
where U (x, 0) can be derived from (73.55) for τ = 0, i.e. t = T :

x + T ln(S)
U (x, 0) = max exp − K, 0 (73.57)
T
Therefore
Z +∞ (x − w)2
1 w + T ln(S) −
U (x, τ ) = √ max exp − K, 0 e 4τ dw
2 πτ −∞ T
(73.58)

w−x
v= √
2 τ
i.e.
√ √
w = x + 2 τ v, dw = 2 τ dv (73.59)
1
Z +∞ x + 2√τ v + T ln(S)

2 √
U (x, τ ) = √ max exp − K, 0 e−v 2 τ dv
2 πτ −∞ T
(73.60)
1069
We now have that
x + 2√τ v + T ln(S)
exp −K >0 (73.61)
T
if
K
T ln −x
v> S
√ ≡ω (73.62)
2 τ
We thus get from (73.60) and (73.62) that
1
Z +∞ x + 2√τ v + T ln(S)
2
U (x, τ ) = √ exp − K e−v dv (73.63)
π ω T
or
1
Z +∞ x + 2√τ v + T ln(S)
2
U (x, τ ) = √ exp −v dv
π ω T
Z +∞ (73.64)
1 2
−√ K e−v dv
π ω
Using the identity

√
x + 2 τ v + T ln(S) τ 2 x τ
− v2 = − v − + + 2 + ln(S) (73.65)
T T T T
1 xT + τ Z +∞
τ 2

U (x, τ ) = √ Sexp exp − v − dv
π T2 ω T
Z +∞ (73.66)
1 2
−√ K e−v dv
π ω
The substitution
τ
s=v− (73.67)
T
converts equation (73.66) into the equation
1 xT + τ Z +∞ 2 1
Z +∞
2
−s
U (x, τ ) = √ S exp 2
e ds − √ K e−v dv (73.68)
π T ψ π ω
1070
where
K
T ln −x τ
ψ= S
√ − (73.69)
2 τ T
At this stage we consider the complementary error function erfc(x) (for
the special functions used in the text see for example Abramowitz and Stegun
[1] or Lebedev [35]) defined by
Z +∞
2 2
erfc(x) = √ e−ξ dξ (73.70)
π x
Therefore equation (73.68) yields
1 xT + τ 1
U (x, τ ) = S exp 2
erfc(ψ) − K erfc(ω) (73.71)
2 T 2
There is a known identity
1 √
erfc(x) = Φ(− 2x) (73.72)
2
where Φ(x) is the cumulative function of the normal distribution N (0, 1) with
mean 0 and variance 1, given by (see for example Abramowitz and Stegun
[1])
Z x
1 2
Φ(x) = √ e−u /2 du (73.73)
2π −∞
Equation (73.71) then gives
xT + τ √ √
U (x, τ ) = S exp Φ(− 2ψ) − KΦ(− 2ω) (73.74)
T2
Let us now express everything in terms of the original variables. Since x =
ln(z) + α(t) and ln(z) = y = A − t ln(S), we finally have
x = A − t ln(S) + α(t) (73.75)
where α(t) is given by (73.51). We also have that τ = β(t) (see equation
(73.46)) where β(t) is given by (73.59). Introducing the notation
√ √
d1 = − 2 ψ and d2 = − 2 ω (73.76)
1071
we find
K
A − t ln(S) − T ln + α(t) √2β(t)
d1 = p S + (73.77)
2β(τ ) T
and
K
A − t ln(S) − T ln + α(t)
d2 = p S (73.78)
2β(τ )
Therefore (73.74) becomes

A − t ln(S) + α(t) τ
U (x, τ ) = S exp + 2 Φ(d1 ) − KΦ(d2 ) (73.79)
T T
Collecting everything together, we obtain the following option pricing formula
for the Asian Options, for the geometric Average case:
A − t ln(S) + α(t)
−r(T −t) β(t)
V (S, A, t) = e S exp + 2 Φ(d1 ) − KΦ(d2 )
T T
(73.80)
Appendix A. We consider the partial differential equation (73.14)
∂V 1 ∂ 2V ∂V ∂V
+ σ 2 S 2 2 + ln(S) + rS − rV
∂t 2 ∂S ∂A ∂S
The Lie symmetries of the previous equation are to be determined next.
Step 1. Let ∆(S, A, t, V (2) ) be defined by
∂V 1 ∂ 2V ∂V ∂V
∆(S, A, t, V (2) )≡ + σ 2 S 2 2 + ln(S) + rS − rV (73.81)
∂t 2 ∂S ∂A ∂S
Step 2. Introduce the vector field X (the generator of the symmetries) by
∂ ∂ ∂
∂S ∂A ∂t (73.82)
∂
+ φ(S, A, t, V )
∂V
1072
Step 3. The second prolongation is defined in our case by the expression
∂ ∂ ∂ ∂
P r(2) X = X + φS + φA + φt + φSS (73.83)
∂VS ∂VA ∂Vt ∂VSS
Step 4. The Lie symmetries of the equation are determined by the condition
P r(2) X[∆(S, A, t, V (2) )] = 0 as long as ∆(S, A, t, V (2) ) = 0 (73.84)
Implementation of the equation P r(2) X[∆(S, A, t, V (2) )] = 0:

We have
P r(2) X[∆(S, A, t, V (2) )] = 0
∂V 1 ∂ 2V ∂V ∂V
⇐⇒P r(2) X[ + σ 2 S 2 2 + ln(S) + rS − rV ] = 0
∂t 2 ∂S ∂A ∂S
1 (73.85)
⇐⇒ξ1 [σ 2 SVSS + rVS + VA ] − rφ + rSφS + ln(S)φA + φt
S
1 2 2 SS
+ σ S φ =0
2
Step 5. Calculation of φS , φA , φt , φSS . We have the following expressions
φS = φS + (φV − ξ1S )VS − ξ1V VS2 − ξ3S Vt

(73.86)
− ξ2S VA − ξ3V VS Vt − ξ2V VS VA
φA = φA + (φV − ξ2A )VA − ξ2V VA2 − ξ3A Vt

(73.87)
− ξ1A VS − ξ1V VA VS − ξ3V Vt VA
φt = φt + (φV − ξ3t )Vt − ξ3V Vt2 − ξ1t VS
(73.88)
− ξ2t VA − ξ2V VA Vt − ξ1V Vt VS
φSS = φSS + (2φSV − ξ1SS )VS − ξ3SS Vt + (φV V − 2ξ1SV )VS2
− 2ξ3SV VS Vt − ξ1V V VS3 − ξ3V V VS2 Vt + (φV − 2ξ1S )VSS
− 2ξ2S VSA − 3ξ1V VS VSS − ξ2V VA VSS − 2ξ3V VS VSt (73.89)
− 2ξ2SV VA VS − 2ξ3S VSt − 2ξ2V VS VSA − ξ3V Vt VSS
− ξ2V V VS2 VA − ξ2SS VA
1073
Step 6. Therefore, using the previous expressions for φS , φA , φt and φSS ,
we obtain from the last equation of Step 4:
1
ξ1 [σ 2 SVSS + rVS + VA ] − rφ
S
+ rS φS + (φV − ξ1S )VS − ξ1V VS2 − ξ3S Vt


+ ln(S) φA + (φV − ξ2A )VA − ξ2V VA2 − ξ3A Vt


+ φt + (φV − ξ3t )Vt − ξ3V Vt2 − ξ1t VS (73.90)


1 2 2
+ σ S φSS + (2φSV − ξ1SS )VS − ξ3SS Vt + (φV V − 2ξ1SV )VS2
2
− 2ξ2S VSA − 3ξ1V VS VSS − ξ2V VA VSS − 2ξ3V VS VSt

2
− ξ2V V VS VA − ξ2SS VA = 0
1074
Step 7. We now have to take into account the condition ∆(S, A, t, V (2) ) = 0.
1
We substitute Vt by − σ 2 S 2 VSS − rSVS − ln(S)VA + rV into (73.90):
2
1
S
+ rS φS + (φV − ξ1S )VS − ξ1V VS2 − (ξ2S + ξ2V VS )VA
1
2 2
− (ξ3S + ξ3V VS ) − σ S VSS − rSVS − ln(S)VA + rV
2

+ ln(S) φA + (φV − ξ2A )VA − ξ2V VA2 − ξ1A VS − ξ1V VA VS
1
2 2
− (ξ3A + ξ3V VA ) − σ S VSS − rSVS − ln(S)VA + rV
2

+ φt + (φV − ξ3t )Vt − ξ1t VS − ξ2t VA
1 2
(73.91)
− ξ3V − σ 2 S 2 VSS − rSVS − ln(S)VA + rV
2
1
2 2
− (ξ2V VA + ξ1V VS ) − σ S VSS − rSVS − ln(S)VA + rV
2

1
+ σ 2 S 2 φSS + (2φSV − ξ1SS )VS + (φV V − 2ξ1SV )VS2 − ξ1V V VS3
2
− (ξ3SS + 2ξ3SV VS + ξ3V V VS2 + ξ3V VSS )×
1
× − σ 2 S 2 VSS − rSVS − ln(S)VA + rV
2
− 2ξ2SV VA VS − 2ξ3S VSt − 2ξ2V VS VSA + (φV − 2ξ1S )VSS

2
Step 8. Equate all the coefficients of the partial derivatives of the function V
to zero. We are then going to have a system of partial differential equations.
Before that, and just for the economy of the calculations, we observe that
we get some simple expressions looking at the coefficients of the mixed prtial
1
derivatives. In fact the coefficient of the derivative VSA is σ 2 S 2 (−2ξ2S ) and
2
1 2 2
that of VS VSA is σ S (−2ξ2V ) which means that ξ2S = 0 and ξ2V = 0.
2
1075
Therefore the coefficient ξ2 depends only on A and t:
ξ2 = ξ2 (A, t) (73.92)
1
The coefficient of the derivative VS VSt is σ 2 S 2 (−2ξ3V ) and that of VSt is
2
1 2 2
σ S (−2ξ3S ) which means that ξ3V = 0 and ξ3S = 0. Therefore the coeffi-
2
cient ξ3 depends only on A and t:
ξ3 = ξ2 (A, t) (73.93)
Because of (73.92) and (73.93), equation (73.91) becomes

1
S
2
+ rS φS + (φV − ξ1S )VS − ξ1V VS

+ ln(S) φA + (φV − ξ2A )VA − ξ1A VS − ξ1V VA VS

+ φt + (φV − ξ3t )Vt − ξ1t VS − ξ2t VA (73.94)
1
− (ξ1V VS ) − σ 2 S 2 VSS − rSVS − ln(S)VA + rV
2

1
+ σ 2 S 2 φSS + (2φSV − ξ1SS )VS + (φV V − 2ξ1SV )VS2
2

3
− ξ1V V VS − 3ξ1V VS VSS + (φV − 2ξ1S )VSS = 0
The coefficient of VS VSS is the sum of the terms

1 1
(−ξ1V ) − σ 2 S 2 and (−3ξ1V ) − σ 2 S 2
2 2
and therefore ξ1V = 0, which means that
ξ1 = ξ1 (S, A, t) (73.95)
1076
Therefore (73.94) gets simplified into

2 1
ξ1 [σ SVSS + rVS + VA ] − rφ + rS φS + (φV − ξ1S )VS
S

+ ln(S) φA + (φV − ξ2A )VA − ξ1A VS
(73.96)

1 2 2 2
+ σ S φSS + (2φSV − ξ1SS )VS + φV V VS + (φV − 2ξ1S )VSS = 0
2
From (73.96) we get the following coefficients (which are equated to zero):
Coefficient of VS2
1 2 2
σ S φV V = 0 (73.97)
2
Coefficient of VA
1
ξ1 + (ln(S))2 ξ3A + (ln(S))(ξ3t − ξ2A ) − ξ2t = 0 (73.98)
S
Coefficient of VSS
1 1
σ 2 Sξ1 + σ 2 S 2 (ξ3t − 2ξ1S ) + σ 2 S 2 (ln(S))ξ3A = 0 (73.99)
2 2
Coefficient of VS
rξ1 + rS(ξ3t − ξ1S ) − (ln(S))ξ1A + rS(ln(S))ξ3A
1 (73.100)
+ σ 2 S 2 (2φSV − ξ1SS ) − ξ1t = 0
2
Zero-th order coefficient
− rφ + rSφS + (ln(S))φA − r(ln(S))V ξ3A + φt
1 (73.101)
+ rV (φV − ξ3t ) + σ 2 S 2 φSS = 0
2
Step 9. We now have to solve the system of determining equations (73.97)-
(73.101). From (73.97) we get φV V = 0 and thus φ is a linear function with
respect to V :
φ = α(S, A, t)V + β(S, A, t) (73.102)
1077
From (73.99) we obtain
1 1 1
ξ1S − ξ1 = ξ3t + ln(S)ξ3A (73.103)
S 2 2
which is a linear differential equation with unknown function ξ1 . The solution
of this equation is given by
1 1
ξ1 = S ln(S)ξ3t + S(ln(S))2 ξ3A + S θ(A, t) (73.104)
2 4
where θ(A, t) is a function to be determined. From (73.98), using (73.104),
we get the equation
5 3
(ln(S))2 ξ3A + ξ3t − ξ2A (ln(S)) − ξ2t + θ(A, t) = 0 (73.105)
4 2
Since the previous equation should hold for any value of S, we get from this
equation the following three equations:
ξ3A = 0 (73.106)
3
ξ3t − ξ2A = 0 (73.107)
2
−ξ2t + θ(A, t) = 0 (73.108)
From (73.93) and (73.106) we get that
ξ3 = f (t) (73.109)
3
Using (73.107) and (73.109) we obtain ξ2A = f 0 (t) and thus
2
3
ξ2 = f 0 (t) A + g(t) (73.110)
2
From (73.108), using (73.110) we get
3
θ(A, t) = f 00 (t) A + g 0 (t) (73.111)
2
Because of (73.106), equation (73.104) gets simplified into
1
ξ1 = S ln(S)ξ3t + S θ(A, t) (73.112)
2
1078
Equation (73.100), because of (73.102) and (73.112), takes the form
1 ln(S)
αS = F (A, t) + G(A, t) (73.113)
S S
where
1 σ 2 − 2r 4
F (A, t) = ξ3t + θt (A, t) (73.114)
4 σ2 σ2
and
1 2 4
G(A, t) = ξ3tt + 2 θA (A, t) (73.115)
4 σ2 σ
Integrating (73.113) with respect to S, we obtain the following expression for
α(S, A, t):
1
α(S, A, t) = F (A, t) ln(S) + G(A, t)(ln(S))2 + ρ(A, t) (73.116)
2
where ρ(A, t) is a function to be determined. From (73.101), using (73.102),
we get the equation
− r(α·V + β) + rS(αS ·V + βS ) + (αA ·V + βA ) ln(S)

− rV ξ3A ln(S) + (αt ·V + βt ) + r(α − ξ3t )V (73.117)
1
+ σ 2 S 2 (αSS ·V + βSS ) = 0
2
Equating the coefficients of V to zero, we get from the previous equation the
following two equations
− rα + rSαS + (ln(S))αA − r(ln(S))ξ3A + αt

1 (73.118)
+ r(α − ξ3t ) + σ 2 S 2 αSS = 0
2
and
1
−rβ + rSβS + (ln(S))βA + βt + σ 2 S 2 βSS = 0 (73.119)
2
Equation (73.119) expresses the fact that the function β(S, A, t) introduced
in (73.102) is a solution of the original PDE.
1079
We consider now equation (73.118). Using the expression (73.116), equation
(73.118) becomes

1 2 1 2 2
− (σ − 2r)F (A, t) − rξ3t + ρt (A, t) + σ S G(A, t)
2 2

1 2
+ − (σ − 2r)G(A, t) + ρA (A, t) + Ft (A, t) ln(S) (73.120)
2

1 1
+ FA (A, t) + Gt (A, t) (ln(S))2 + GA (A, t)(ln(S))3 = 0
2 2
Since this equation should hold for any value of S, we obtain the following
four equations:
1 1
− (σ 2 − 2r)F (A, t) − rξ3t + ρt (A, t) + σ 2 S 2 G(A, t) = 0 (73.121)
2 2
1
− (σ 2 − 2r)G(A, t) + ρA (A, t) + Ft (A, t) = 0 (73.122)
2
1
FA (A, t) + Gt (A, t) = 0 (73.123)
2
GA (A, t) = 0 (73.124)
Before proceeding to the solution of the above system of equations, it is
instructive to find some equivalent expressions for the functions F (A, t) and
G(A, t). In fact, using (73.109) and (73.111), we find
1 σ 2 − 2r 0

4 3 000 00
F (A, t) = f (t) + 2 f (t) A + g (t) (73.125)
4 σ2 σ 2
and
2 00
G(A, t) = f (t) (73.126)
σ2
Equation (73.126) is compatible to (73.124). Equation (73.123), because of
(73.125) and (73.126) is equivalent to f 000 (t) = 0 which means that f (t) is a
second degree polynomial with respect to t:
f (t) = a1 + a2 t + a3 t2 (73.127)
1080
and using (73.109), we get
ξ3 = a1 + a2 t + a3 t2 (73.128)
Because of (73.127), the functions F (A, t) and G(A, t) become
σ 2 − 2r 1 00
F (A, t) = (a 2 + 2a 3 t) + g (t) (73.129)
4σ 2 σ2
and
4
G(A, t) = a3 (73.130)
σ2
respectively. Equation (73.122), because of (73.129) and (73.130), gives
3(σ 2 − 2r) 1
ρA (A, t) = 2
a3 − 2 g 000 (t) (73.131)
2σ σ
On the other hand, equation (73.121) gives
2
1 2 σ − 2r 1 00
ρt (A, t) = (σ − 2r) (a2 + 2a3 t) + 2 g (t)
2 4σ 2 σ (73.132)
+ r(a2 + 2a3 t) − 2a3
Equations (73.131) and (73.132) is a system of DEs with unknown functions

g(t) and ρ(A, t). The compatibility condition ρAt (A, t) = ρtA (A, t), taking
into account (73.131) and (73.132), is equivalent to g (iv)(t)=0 from which there
follows that g(t) is a third degree polynomial of t:
g(t) = a4 + a5 t + a6 t2 + a7 t3 (73.133)
Therefore (73.131) and (73.132) take the form

3
ρA (A, t) = 2
[(σ 2 − 2r)a3 − 4a7 ] (73.134)
2σ
and
2
1 2 σ − 2r 1
ρt (A, t) = (σ − 2r) (a2 + 2a3 t) + 2 (2a6 + 6a7 t)
2 4σ 2 σ (73.135)
+ r(a2 + 2a3 t) − 2a3
1081
respectively. Integrating them with respect to A and t respectively, we obtain
3
ρ(A, t) = 2
[(σ 2 − 2r)a3 − 4a7 ]·A + h(t) (73.136)
2σ
and
(σ 2 − 2r)2 2 σ 2 − 2r
ρ(A, t) = 2
(a 2 t + a 3 t ) + 2
(2a6 t + 3a7 t2 )
8σ 2σ (73.137)
+ r(a2 t + a3 t2 ) − 2a3 t + k(A)
where h(t) and k(A) are two functions to be determined. Comparing the two
previous expressions, we obtain the function ρ(A, t):
(σ 2 − 2r)2 2 σ 2 − 2r
ρ(A, t) = (a 2 t + a 3 t ) + (2a6 t + 3a7 t2 )
8σ 2 2σ 2
+ r(a2 t + a3 t2 ) − 2a3 t (73.138)
3
+ 2 [(σ 2 − 2r)a3 − 4a7 ]·A + a8
2σ
Using (73.133), function F (A, t) becomes (using equation (73.129))
σ 2 − 2r 2
F (A, t) = (a 2 + 2a 3 t) + (a6 + 3a7 t) (73.139)
4σ 2 σ2
From (73.110), using the expressions for f (t) and g(t) given by (73.127) and
(73.133) respectively, we get the following expression of ξ2 :
3
ξ2 (A, t) = (a2 + 2a3 t)A + a4 + a5 t + a6 t2 + a7 t3 (73.140)
2
From (73.111), using again the expressions for f (t) and g(t), we get the
following expression for θ(A, t):
θ(A, t) = 3a3 tA + a5 t + 2a6 t + 3a7 t2 (73.141)
From (73.112), using the expressions (73.128) and (73.141), we get the fol-
lowing expression for ξ1 (S, A, t):
1
ξ1 (S, A, t) = [S· ln(S)](a2 + 2a3 t) + S(3a3 A + a5 + 2a6 t + 3a7 t2 ) (73.142)
2
1082
From (73.116), using the expressions for F (A, t), G(A, t) and ρ(A, t) given
by (73.139), (73.130) and (73.138) respectively, we obtain the following ex-
pression for α(S, A, t):
2
σ − 2r 2
α(S, A, t) = (a2 + 2a3 t) + 2 (a6 + 3a7 t) ln(S)
4σ 2 σ
2 (σ − 2r)2
2
+ 2 a3 (ln(S))2 + (a2 t + a3 t2 )
σ 8σ 2 (73.143)
σ 2 − 2r
+ (2a6 t + 3a7 t2 ) + r(a2 t + a3 t2 ) − 2a3 t
2σ 2
3
+ 2 [(σ 2 − 2r)a3 − 4a7 ]·A + a8
2σ
Therefore we can now determine the function φ(S, A, t, V ) introduced in
(73.102):
2
σ − 2r 2
φ(S, A, t, V ) = (a2 + 2a3 t) + 2 (a6 + 3a7 t) ln(S)
4σ 2 σ
2 (σ − 2r)2
2
+ 2 a3 (ln(S))2 + (a2 t + a3 t2 )
σ 8σ 2 (73.144)
σ 2 − 2r 2 2
+ (2a6 t + 3a7 t ) + r(a2 t + a3 t ) − 2a3 t
2σ 2
3 2
+ 2 [(σ − 2r)a3 − 4a7 ]·A + a8 V + β(S, A, t)
2σ
1083
The vector X, the generator of the symmetries (Step 2), can thus be expressed
as
∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + φ
∂S ∂A ∂t ∂V
1 2 ∂
= [S· ln(S)](a2 + 2a3 t) + S(3a3 A + a5 + 2a6 t + 3a7 t )
2 ∂S

3 ∂
+ (a2 + 2a3 t)A + a4 + a5 t + a6 t2 + a7 t3
2 ∂A
∂
+ (a1 + a2 t + a3 t2 )
∂t
σ 2 − 2r (73.145)
2
+ (a2 + 2a3 t) + 2 (a6 + 3a7 t) ln(S)
4σ 2 σ
2 2 (σ − 2r)2
2
+ 2 a3 (ln(S)) + 2
(a2 t + a3 t2 )
σ 8σ
σ 2 − 2r
+ (2a6 t + 3a7 t2 ) + r(a2 t + a3 t2 ) − 2a3 t
2σ 2
3 2
∂
+ 2 [(σ − 2r)a3 − 4a7 ]·A + a8 V + β(S, A, t)
2σ ∂V
The previous expression can then be written as

∂ 1 ∂ ∂ 3 ∂
X = a1 + a2 [S· ln(S)] +t + A
∂t 2 ∂S ∂t 2 ∂A
σ 2 − 2r (σ 2 − 2r)2 ∂
+ ln(S) + t + rt V
4σ 2 8σ 2 ∂V

∂ ∂ ∂
+ a3 [tS· ln(S) + 3AS] + 3tA + t2
∂S ∂A ∂t
σ 2 − 2r 2 (σ 2 − 2r)2 2
2
+ t ln(S) + (ln(S)) + t
2σ 2 σ2 8σ 2 (73.146)
3(σ 2 − 2r) ∂

2
+ rt − 2t + A V
2σ 2 ∂V
∂ ∂ ∂ ∂ ∂
+ a4 + a5 S +t + a8 V + β(S, A, t)
∂A ∂S ∂A ∂V
2 ∂
∂V
∂ ∂ 2 σ − 2r
+ a6 2tS + t2 + ln(S) + t V
∂S ∂A σ2 σ2 ∂V
2

2 ∂ 3 ∂
6 3(σ − 2r) 2 6 ∂
+ a7 3t S +t + t ln(S) + t − 2A V
∂S ∂A σ2 2σ 2 σ ∂V
1084
Therefore the generators of the symmetries are given by (73.21) and (73.22).
Appendix B. Reduction of equation (73.14) to the time-dependent
BS-equation.
We consider the generator X5 given by
∂ ∂
X5 = S +t (73.147)
∂S ∂A
In order to find the invariant corresponding to this generator, we have to solve
dS dA
the equation = . The general solution of this equation is t ln(S) =
S t
A + C1 . Therefore the invariant associated to the generator X5 is
y = A − t· ln(S) (73.148)
Considering the substitution
V (S, A, t) = u(y, t) (73.149)
the partial derivatives transform as
Vt = ut − (ln(S))uy , SVS = −tuy , S 2 VSS = t2 uyy + tuy (73.150)
Using the previous substitutions, equation (73.14) is being transformed to

1 1
ut + σ 2 − r tuy + σ 2 t2 uyy − ru = 0 (73.151)
2 2
y = ln(z) (73.152)

1 2 1 22 1
ut + σ − r t + σ t zuz + σ 2 t2 z 2 uzz − ru = 0 (73.153)
2 2 2
which is the Black-Scholes equation (Black and Scholes [10], Wilmott [49])
with time-dependent coefficients.
Appendix C. The Lie algebra of the infinitesimal transformations of the
constant-coefficient Black-Scholes equation
∂F 1 ∂ 2F ∂F
+ σ 2 S 2 2 + rS − rF = 0 (73.154)
∂t 2 ∂S ∂S
1085
contains the two generators
∂ ∂
F and
∂F ∂t
of its underlying symmetries. We consider a linear combination of the above
generators
∂ ∂
F +λ (73.155)
∂F ∂t
and try to determine the corresponding invariant to the (73.155). Solving
dF
the equation = λdt we find as the general solution ln(F ) = λt + C, from
F
which we determine the invariant G:
F (t, S) = eλt G(t, S) (73.156)
We then substitute (73.156) into (73.154) and obtain the equation

∂G 1 2 2 ∂ 2 G ∂G
+ σ S 2
+ rS + (r − λ)G = 0
∂t 2 ∂S ∂S
The choice λ = r removes the last term of the above equation, leading to
∂G 1 2 2 ∂ 2 G ∂G
+ σ S 2
+ rS =0 (73.157)
∂t 2 ∂S ∂S
Therefore the transformation
F (t, S) = e−r(T −t) G(t, S) (73.158)
converts equation (73.154) into (73.157) where T is the strike time.

We can furthe transform (73.157) into the Heat Equation. The crucial point
∂
is that equation (73.157) contains S as generator of its infinitesimal Lie
∂S
Symmetries. Therefore we introduce the transformation
ξ = ln(S) + a(t) and τ = b(t) (73.159)
where a(t) and b(t) are functions to be determined next.Under this transfor-
mation, since
∂G ∂G 0 ∂G 0 ∂G ∂G ∂ 2G ∂ 2 G ∂G
= b (t) + a (t), S = , S2 = −
∂t ∂τ ∂ξ ∂S ∂ξ ∂S 2 ∂ξ 2 ∂ξ
1086
∂G ∂G 0 1 2 ∂ 2 G ∂G ∂G
b0 (t) + a (t) + σ − + r =0 (73.160)
∂τ ∂ξ 2 ∂ξ 2 ∂ξ ∂ξ
∂G 0 1 ∂ 2G 1 ∂G
b (t) + σ 2 2 + a0 (t) − σ 2 + r =0 (73.161)
∂τ 2 ∂ξ 2 ∂ξ
The choice
1 1
b0 (t) = − σ 2 and a0 (t) − σ 2 + r = 0 (73.162)
2 2
∂G ∂ 2G
= (73.163)
∂τ ∂ξ 2
Therefore, the required transformation, introduced in (73.162), reads

1 1 2
a(t) = σ 2 (T − t), b(t) = r − σ (T − t) (73.164)
2 2
where T is the strike time.
References
[1] M. Abramowitz and I. A. Stegun: ”Handbook of Mathematical
[2] R. L. Anderson and N. H. Ibragimov: ”Lie-Bcklund Transforma-
tions in Applications”. SIAM, Philadelphia 1979
[3] J. Angus: ”A note on pricing Asian Derivatives with continuous geo-
metric averaging”. Journal of Future Markets 19 (1999) 845-858
[4] S. M. Antoniou: ”A General Solution to the Bensoussan-Crouhy-Galai
PDE appearing in Finance”. SKEMSYS preprint
[5] S. M. Antoniou: ”Closed-Form Solutions to the Hamilton-Jacobi-
Bellman Equation”, SKEMSYS preprint
[6] S. M. Antoniou: ”The Arithmetic Average Case of Asian Options:Reducing
the PDE to the Black-Scholes Equation”. SKEMSYS preprint
[7] S. M. Antoniou: ”Lie Symmetry Analysis of the partial differential
1087
equations appearing in Finance”. SKEMSYS report
[8] E. Barucci, S. Polidoro and V. Vespri: ”Some results on partial
differential equations and Asian Options”. University of Pisa preprint
[9] Y. Z. Bergman: ”Pricing path contingent claims”.
Research in Finance 5 (1985) 229-241
[10] F. Black and M. Scholes: ”The pricing of options and corporate
liabilities”. Journal of Political Economy 81 (1973) 637-654
[11] G. W. Bluman and S. C. Anco: ”Symmetry and Integration Meth-
ods for Differential Equations”. Springer, 2002
[12] G.W. Bluman and J. D. Cole: ”Similarity Methods for Differential
Equations”. Springer-Verlag, 1974
derivatives”. arXiv:math.AP/0604207
linear model of financial derivatives”. International Journal of Theoretical
and Applied Finance 10 (2007) 1-21
illiquid markets: scaling properties and explicit solutions” arXiv:0708.1568v1
[math.AP] 11 Aug 2007
[17] J. E. Campbell: ”Introductory Treatise on Lies Theory of Finite Con-
tinuous Transformation Groups” Oxford at the Clarendon Press 1903
[18] B. J. Cantwell: ”Introduction to Symmetry Analysis”.
Cambridge University Press 2002
[19] A. Cohen: ”An Introduction to the Lie Theory of one-parameter
groups: With Applications to the solution of Differential Equations”.
D. C. Heath and Co. Boston 1911
[20] L.E. Dickson: ”Differential equations from the group standpoint”.
Ann. Math. 25 (1924) 287-378
[21] L. Dresner: ”Applications of Lie’s Theory of Ordinary and Partial
Differential Equations”. IOP Publishing, 1999
[22] G. Emanuel: ”Solution of Ordinary Differential Equations by Contin-
uous Groups”. Chapman and Hall 2001
[23] H. Geman and M. Yor: ”Bessel Processes, Asian Options and per-
petuities”. Mathematical Finance 3 (1993) 349-375
[24] J. Goard: ”New Solutions to the Bond-Pricing Equation via Lie’s Clas-
sical Method”, Mathematical and Computer Modelling 32 (2000) 299-313
1088
[25] E. G. Haug: ”Option Pricing Formulas”.
Second Edition, McGraw Hill 2007
Lie Symmetries of Differential Equations”.
Euromath Bulletin 2, No.1 Fall 1993
[27] P. E. Hydon: ”Symmetry Methods for Differential Equations: A Be-
ginners Guide”. Cambridge University Press 2000
Differential Equations” Vol.1 1994, Vol.2 1995, Vol.3 1996 CRC Press, Boca
Raton, FL.
[31] L. Ince: ”Ordinary Differential Equations”, Dover 1956
[32] L. Jiang and M. Dai: ”Convergence Analysis of Binomial Tree
Method for American-type Path-dependent Options”.
Preprint, Peking University
[33] A G Z Kemna and A C F Vorst: ”A pricing method for options”
Journal of Banking and Finance 14 (1990) 113-129
[34] P. G. L. Leach, J.G. OHara and W. Sinkala: ”Symmetry-based
investment”. Journal of Mathematical Analysis and Applications, 2007
[35] N. N. Lebedev: ”Special Functions and their Applications”.
Dover 1972
[36] S. Lie: ”Gesammelte Werke”, Vols. 1-7, F. Engel and P. Hergard (Eds).
Teubner, Leipzig 1899
Graduate Texts in Mathematics, vol.107. Springer Verlag, New York 1993
[40] J. M. Page: ”Ordinary Differential Equations with an Introduction to
Lie’s Theory of the Group of one parameter”.
McMillan and Co. New York 1897
[41] C. A. Pooe, F. M. Mahomed and C. Wafo Soh: ”Fundamental
Solutions for Zero-Coupon Bond Pricing Models”.
1089
Nonlinear Dynamics 36 (2004) 69-76
[42] M. Rubinstein: ”Exotic Options”.
Report, University of California at Berkeley, 1991
[43] F. Schwarz: ”Algorithmic Lie Theory for Solving Ordinary Differential
Equations”. Chapman and Hall/CRC, 2008
[44] G. Silberberg: ”Discrete Symmetries of the Black-Scholes Equation”.
Proceedings of 10th Intern. Conf. in Modern Group Analysis.
[45] W. Sinkala, P. G. L. Leach and J.G. OHara: ”Zero-coupon bond
prices in the Vasicek and CIR models: Their computation as group-invariant
solutions”. Mathematical Methods in the Applied Sciences, 2007
[46] Willi-Hans Steeb: ”Continuous Symmetries, Lie Algebras, Differen-
tial Equations and Computer Algebra”. World Scientific 2007
[47] H. Stephani: ”Differential Equations: Their Solutions Using Symme-
[48] A. N. Tikhonov and A. A. Samarskii: ”Equations of Mathematical
Physics”. Pergamon 1963
[49] P. Wilmott: ”Quantitative Finance”.
Three-Volume Set, Second Edition, Wiley 2006
[50] C. E. Wulfram: ”Dynamical Symmetry”. World Scientific 2011
[51] P. G. Zhang: ”Exotic Options”. 2nd Edition, World Scientific 1998
74 Asian Options.
The Arithmetic Average Case
We consider the path-dependent contingent claims where the underlying asset
follows an arithmetic average process. Considering the no-arbitrage PDE of
these claims, we first determine the underlying Lie Point Symmetries. After
determination of the invariants, we transform the PDE to the Black-Scholes
(BS) equation. We then transform the BS equation into the heat equation
and considering appropriate boundary condition, we find an explicit solution
of the option pricing equation. This procedure appears for the first time in
the finance literature.
1. Introduction. A path-dependent option is an option whose payoff de-
pends on the past history of the underlying asset. In other words these
options have payoffs that do not depend on the assets value at expiry. Two
1090
very common examples of path-dependent options are Asian options and
lookback options.
The terminal payoff of an Asian option depends on the type of averaging
of the underlying asset price over the whole period of the options lifetime.
According to the way of taking average, we distinguish two classes of Asian
options: arithmetic and geometric options.
A lookback option is another type of path-dependent option, whose payoff
depends on the maximum or minimum of the asset price during the lifetime
of the option. The valuation of path-dependent (Asian) European options
is a difficult problem in mathematical finance. There are only some simple
cases where the price of path-dependent contingent claims can be obtained
in closed-form (refs. [1]-[8]).
If the underlying asset price follows a lognormal stochastic process, then its
geometric average has a lognormal probability density and in this case there
is a closed-form solution (ref. [5]). This is not however the case by consider-
ing arithmetic average. In this case there is no closed-form solution. However
in both cases there are some numerical procedures available (ref. [8]).
It is the purpose of this article to provide a closed-form general solution for
the arithmetic path-dependent options.
The paper is organized as follows: In Section 2 we consider the general prob-
lem of pricing the path-dependent contingent claims. In Section 3 we con-
sider the Lie Point Symmetries of the Partial Differential Equation (equation
(3.1)) for the path-dependent arithmetic average options. Full details of the
calculation are provided in Appendix A. In Section 4, considering two invari-
ants of the PDE, we convert the equation into the Black-Scholes equation.
In section 5, we transform the BS equation to the heat equation. We then
find the general solution of our PDE (Theorem 3). In section 6 we find an
explicit solution to the option pricing equation under appropriate boundary
condition. Because of the complexity of the calculations, we have included a
great deal of calculation details, supplemented by many Appendices.
2. Path-Dependent Contingent Claims
Suppose that an option pays off at expiration time T an amount that is a
function of the path taken by the asset between time zero and T . This path-
dependent quantity can be represented by an integral of some function of the
asset over the time period 0≤τ ≤T
Z T
A(T ) = f (S, τ )dτ (74.1)
0
1091
where f (S, t) is a suitable function. Therefore for every t, 0≤t≤T , we have
Z t
A(t) = f (S, τ )dτ (74.2)
0
or, in differential form,
dA = f (S, t)dt (74.3)
We have thus introduced a new state variable A, which will later appear in
our PDE. We are now going to derive the PDE of pricing the path-dependent
option. To value the contract, we consider the function V (S, A, t) and set up
a portfolio containing one of the path-dependent options and a number ∆ of
the underlying asset:
Π = V (S, A, t) − ∆·S (74.4)
The change in the value of this portfolio is given by, according to Itô’s for-
mula, by
∂V 1 2 2 ∂ 2V ∂V ∂V
dΠ = + σ S dt + dA + − ∆ dS (74.5)
∂t 2 ∂S 2 ∂A ∂S
Choosing
∂V
∆= (74.6)
∂S
to hedge the risk and using (74.3), we find that
∂V 1 ∂ 2V ∂V
dΠ = + σ 2 S 2 2 + f (S, t) dt (74.7)
∂t 2 ∂S ∂A
This change is risk-free and thus earns the risk-free rate of interest r,
dΠ = rΠdt (74.8)
leading to the pricing equation
∂V 1 ∂ 2V ∂V ∂V
+ σ 2 S 2 2 + f (S, t) + rS − rV = 0 (74.9)
∂t 2 ∂S ∂A ∂S
1092
upon combining (74.7), (74.8) and (74.4), and where ∆ in (74.4) has been
substituted by (74.6) for the value of ∆. The previous PDE is solved by
imposing the condition
V (S, A, T ) = Ω(S, A) (74.10)
where T is the expiration time.

If A is considered to be an arithmetic average state variable, i.e.
Z t
A= S(τ )dτ (74.11)
0

∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV = 0 (74.12)
∂t 2 ∂S ∂A ∂S
If A is considered to be a geometric average state variable, i.e.
Z t
A= ln(S(τ ))dτ (74.13)
0

∂V 1 2 2 ∂ 2V ∂V ∂V
+ σ S 2
+ ln(S) + rS − rV = 0 (74.14)
∂t 2 ∂S ∂A ∂S
We list below the payoff types we have to consider along with PDEs (74.12)
and (74.14):
For average strike call, we have the payoff V (S, A, T )
max(S − A, 0) (74.15)
For average strike put, we have the payoff
max(A − S, 0) (74.16)
For average rate call, we have the payoff
max(A − E, 0) (74.17)
For average rate put, we have the payoff
max(E − A, 0) (74.18)
1093
where E is the strike price.
This is an introduction to the path-dependent options. Further details can
be found in Wilmott’s treatise (Wilmott [49], Vol.2 and Vol.3) on Quantita-
tive Finance.
3. Lie Symmetries of the Partial Differential Equation. The tech-
nique of Lie Symmetries was introduced by S. Lie (Lie [11] and [12]) and is
best described in refs [13]-[28]. This technique was for the first time used
in partial differential equations of Finance by Ibragimov and Gazizov [29].
The same technique was also used by the author in solving the Bensoussan-
Crouhy-Galai equation (Antoniou [30]) and the HJB equation for a portfolio
optimization problem (Antoniou [31]). There is however a growing list of
papers in applying this method to Finance. For a non complete set of refer-
ences, see [29]-[40].
We now consider the PDE (74.12)
∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV = 0
∂t 2 ∂S ∂A ∂S
We shall determine the Lie point symmetries of the previous equation. We
follow closely Olver [15]. Let ∆(S, A, t, V (2) ) be defined by
∂V 1 ∂ 2V ∂V ∂V
∆(S, A, t, V (2) )≡ + σ2S 2 2 + S + rS − rV (74.19)
∂t 2 ∂S ∂A ∂S
We introduce the vector field X (the generator of the symmetries) by
∂ ∂ ∂
∂S ∂A ∂t (74.20)
∂
+ φ(S, A, t, V )
∂V
We calculate in Appendix A the coefficients ξ1 , ξ2 , ξ3 and φ and we find that
the Lie algebra of the infinitesimal transformations of the original equation
(74.12) is spanned by the five vectors
∂ ∂ ∂ ∂
X1 = , X2 = , X3 = S +A
∂t ∂A ∂S ∂A (74.21)
∂ ∂ 2 ∂ ∂
X4 = 2AS + A2 + 2 (S − rA)V , X5 = V
∂S ∂A σ ∂V ∂V
and the generator of the infinite-dimensional sub-algebra
∂
Xβ = β(S, A, t) (74.22)
∂V
1094
where β(S, A, t) is an arbitrary solution of the original PDE (74.12).
4. Reduction of the equation. We shall now try to transform the pricing
differential equation into another by reducing the number of independent
variables. We shall retain the time variable and try to find a combination
of the other two variables S and A. The best candidate for this job is the
generator X4 . We state and prove the following Theorem:
Theorem 1. Any solution of the partial differential equation
∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV = 0 (74.23)
∂t 2 ∂S ∂A ∂S
can be expressed into the form
S
V (S, A, t) = An u(x, t)×exp b (74.24)
A
where the function u(x, t) satisfies the Black-Scholes equation
1
ut + σ 2 x2 uxx + rxux − ru = 0 (74.25)
2
with
S 2
x= , n = −rb, b= (74.26)
A2 σ2
Proof. Considering the generator X4 , we can determine two invariants of
the PDE. For this purpose we have to solve the following system
dS dA σ2 dV
= 2 = (74.27)
2AS A 2 (S − rA)V
dS dA S
The differential equation = 2 admits general solution 2 = C1 from
2AS A A
which we obtain our first invariant called x:
S
x= (74.28)
A2
The second differential equation can be written as (taking into account S =
2
r dV
xA ), b x − dA = where b is defined in (74.26). By integration of
A V
1095
this equation we find V = C2 A−rb ebxA . Therefore another invariant of the
PDE is the function u(x, t), realated to V (S, A, t) by
S
V (S, A, t) = An u(x, t)×exp b (74.29)
A
2r
with n = −rb = − . Upon substitution of V (S, A, t) given by (74.29) into
σ2
equation (74.23) we find that the function u(x, t) satisfies the BS-equation
1
ut + σ 2 x2 uxx + rxux − ru = 0 (74.30)
2
The proof is in Appendix B.
5. Transformation of the BS equation into the heat equation. Under
the transformation
z = ln(x) (74.31)
and taking into account that
xux = uz , x2 uxx = uzz − uz
the BS-equation (74.30) takes the form

1
ut + σ 2 uzz + ρuz − ru = 0 (74.32)
2
where
1
ρ = r − σ2 (74.33)
2
τ =T −t (74.34)
where T is the expiry time, equation (74.32) becomes

1
uτ = σ 2 uzz + ρuz − ru (74.35)
2
At this stage we use the following
Lemma. The equation
ut = a2 uxx + bux + cu (74.36)
1096
v = e−ct u(t, x = y − bt) (74.37)
can be converted into the heat equation
vt = a2 vyy (74.38)
The proof is given in Appendix C. Using this Lemma, equation (74.35) under
the substitution
u(τ, z) = e−rτ v(τ, y), y = z + ρτ (74.39)
is being transformed into the equation
1
vτ = σ 2 vyy (74.40)
2
Introducing further a new independent variable ζ by
σ
ζ=√ (74.41)
2
equation (74.40) is being transformed into the heat equation
wτ = wζζ (74.42)
where
w = w(τ, ζ) = v(τ, y) (74.43)
Theorem 2. The solution of the BS-equation (74.30) is given by
u(x, t) = e−rτ w(τ, ζ) (74.44)
where
√ √
2 2
ζ= y= (ln(x) + ρτ ), τ =T −t (74.45)
σ σ
and w(τ, ζ) satisfies the heat equation wτ = wζζ .
Using Theorems 1 and 2 we get the following
Theorem 3. The differential equation (74.23) has the general solution
S
V (S, A, t) = A−m w(τ, ζ)×exp b − rτ (74.46)
A
1097
where the function w(τ, ζ) satisfies the heat equation wτ = wζζ .
2
The variables ζ and τ = T − t are defined in (74.45), while m = rb, b = ,
σ2
ρ is given by (74.33) and x by (74.28).
Therefore equation (74.46) can be written in full notation as
1 2r/σ2 2 S
V (S, A, t) = w(τ, ζ)·exp − rτ (74.47)
A σ2 A
where
√
2 S 1 2
ζ= ln 2 + r − σ τ (74.48)
σ A 2
6. Solution of the pricing equation under the boundary condition.
We are going to derive the solution of the pricing equation under the condition
A +
V (S, A, T ) = exp −K (74.49)
T
r
S
A= (74.50)
x
Therefore
√
r 1 2r/σ2 x 2r/σ2
A 1 S x rγ S
= , = = , = xS (74.51)
T T x A S S A
where
1
γ= (74.52)
σ2
Combining (74.47) with (74.49) and taking into account (74.51), we obtain
( )
1 rS x rγ √
max exp − K, 0 = w(ζ, 0)·exp(2γ xS) (74.53)
T x S
Therefore
( )
1 rS x −rγ √
w(ζ, 0) = max exp − K, 0 ·exp(−2γ xS) (74.54)
T x S
1098
Using Poisson’s formula
Z +∞ (ζ − η)2
1
w(ζ, τ ) = √ w(η, 0)×exp − dη (74.55)
2 πτ −∞ 4τ
and taking into account (74.54), we obtain
Z +∞ ( s )
1 1 S η −rγ
w(ζ, τ ) = √ max exp − K, 0 ×
2 πτ −∞ T η S (74.56)
p (ζ − η)2
×exp(−2γ ηS)×exp − dη
4τ
η−ζ
ξ= √ (74.57)
2 τ
i.e.
√ √
η = ζ + 2 τ ξ, dη = 2 τ dξ (74.58)
Z +∞ ( s )
1 1 S
w(ζ, τ ) = √ max exp √ − K, 0 ×
2 πτ −∞ T ζ +2 τ ξ
(74.59)
ζ + 2√τ ξ −rγ q √ √
× ×exp(−2γ (ζ + 2 τ ξ)S)×exp(−ξ 2 )2 τ dξ
S
We have further
s
1 S
exp √ >K (74.60)
T ζ +2 τ ξ
if
S ζ
ξ< √ 2
− √ ≡ω (74.61)
2 τ (T · ln(K)) 2 τ
Therefore we obtain from (74.59) that
Z ω ( s )
1 1 S
w(ζ, τ ) = √ exp √ −K ×
π −∞ T ζ +2 τ ξ
(74.62)
ζ + 2√τ ξ −rγ q √
× ×exp −2γ (ζ + 2 τ ξ)S ×exp(−ξ 2 ) dξ
S
1099
Z ω s
1 1 S
w(ζ, τ ) = √ exp √ ×
π −∞ T ζ +2 τ ξ
ζ + 2√τ ξ −rγ q √
× ×exp −2γ (ζ + 2 τ ξ)S ×exp(−ξ 2 ) dξ (74.63)
ZSω √
K ζ + 2 τ ξ −rγ q √
−√ ×exp −2γ (ζ + 2 τ ξ)S ×
π −∞ S
×exp(−ξ 2 ) dξ
The substitution
√
ζ +2 τ ξ
p= (74.64)
S
i.e.
pS − ζ S
ξ= √ , dξ = √ dp (74.65)
2 τ 2 τ

θ
√
Z
1 1
−rγ

w(ζ, τ ) = √ exp √ ×p ×exp −2γS p ×
π −∞ T p
pS − ζ 2 S
×exp − √ √ dp
2 τ 2 τ
θ
(74.66)
√
Z
K
−√ p−rγ ×exp −2γS p ×
π −∞
pS − ζ 2 S
×exp − √ √ dp
2 τ 2 τ
where
√
ζ +2 τ ω 2ζ 1
θ≡ = − (74.67)
S S (T · ln(K))2
1100
Equation (74.66) can also be written as
S ζ 2 Z θ
w(ζ, τ ) = √ exp − p−rγ ×
2 πτ 4τ −∞
S2 ζS √ 1
2
×exp − p + √ p − 2γS p + √ dp
4τ 2 τ T p
2 Z θ
(74.68)
KS ζ
− √ exp − p−rγ ×
2 πτ 4τ −∞
S2 ζS √
2
×exp − p + √ p − 2γS p dp
4τ 2 τ
Collecting the results of this Section and Theorem 3, we arrive at
Theorem 4. The solution of the PDE (74.12) for the Asian arithmetic
options
∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV
∂t 2 ∂S ∂A ∂S
under the condition
A +
V (S, A, T ) = exp −K
T
is given by the formula (74.47)
1 2r/σ2 2 S
V (S, A, t) = w(τ, ζ)·exp 2 − rτ
A σ A
where
√
2 S 1 2
ζ= ln 2 + r − σ τ
σ A 2
with w(ζ, τ ) given by (74.68)
S ζ 2 Z θ
w(ζ, τ ) = √ exp − p−rγ ×
2 πτ 4τ −∞
S2 ζS √ 1
×exp − p2 + √ p − 2γS p + √ dp
4τ 2 τ T p
KS ζ2 θZ
− √ exp − p−rγ ×
2 πτ 4τ −∞
S2 ζS √
×exp − p2 + √ p − 2γS p dp
4τ 2 τ
1101
A more simplified expression of the option pricing formula for the Asian
Arithmetic Options is to be published elsewhere.
6. Additional Notes. One could use some general solutions of the heat
equation and then under some appropriate bounary conditions to find ex-
plicit expression(s) of the option pricing formula. We think two solutions are
appropriate to complete the picture. One of them is the solution expressed
in terms of parabolic cylindtical functions and the other expressed in terms
of the Airy functions. Both solutions are based on a Lie symmetry analysis
of the heat equation.
To start with, it is known that the Lie algebra of the infinitesimal symmetries
of the heat equation wτ = wζζ is produced by the six vector fields
∂ ∂ ∂ ∂ ∂
X1 = , X2 = , X3 = w , X4 = ζ + 2τ
∂ζ ∂τ ∂w ∂ζ ∂τ
(74.69)
∂ ∂ ∂ 2 ∂ 2 ∂
X5 = 2τ − wζ , X6 = 4τ ζ + 4τ − (ζ + 2τ )w
∂ζ ∂w ∂ζ ∂τ ∂w
and the infinite dimensional subalgebra
∂
Xα = α(ζ, τ ) (74.70)
∂w
where α(ζ, τ ) is an arbitrary solution of the heat equation (refs. [15], [37]).
6.1. The solution that is invariant with respect to the generator
X2 + X6 + aX3 = 4τ ζ∂ζ + (4τ 2 + 1)∂τ − (ζ 2 + 2τ − a)w∂w (74.71)
(a is a constant) is given by (refs. [15], [37])
( √ a −√2ζ
)
1 a 2ζ
w(ζ, τ ) = √4
K1 ·W , √ + K2 ·W , √
4τ 2 + 1 2 4τ 2 + 1 2 4τ 2 + 1
(74.72)
ζ 2τ a
×exp − 2 + arctan(2τ )
4τ + 1 2
where W (a, z) is the parabolic cylindrical function of imaginary argument
(see Appendix D). The solution given by Olver [15] (Example 3.17, page
208) contains some misprints. We have repeated the calculations (Antoniou
[37]) and we have found the solution given in (74.72).
6.2. The solution that is invariant with respect to the generator
X2 − X5 = ∂τ − 2τ ∂ζ + wζ∂w (74.73)
1102
is given by (refs. [15], [37])
2τ 3

2 2
w(ζ, τ ) = K1 ·Ai(ζ + τ ) + K2 ·Bi(ζ + τ ) exp ζτ + (74.74)
3
where Ai(x) and Bi(x) are the Airy functions (Appendix E).
Both solutions (74.72) and (74.74) we have considered, are not singular for
τ = 0, i.e. t = T . This is because we want to take into account the payoff
conditions (74.15)-(74.18) valid for t = T .
Appendix A. We consider the partial differential equation (74.12)
∂V 1 ∂ 2V ∂V ∂V
+ σ2S 2 2 + S + rS − rV
∂t 2 ∂S ∂A ∂S
The Lie symmetries of the previous equation are to be determined next.
Step 1. Let ∆(S, A, t, V (2) ) be defined by
∂V 1 ∂ 2V ∂V ∂V
∆(S, A, t, V (2) )≡+ σ2S 2 2 + S + rS − rV (74.75)
∂t 2 ∂S ∂A ∂S
Step 2. Introduce the vector field X (the generator of the symmetries) by
∂ ∂ ∂
∂S ∂A ∂t (74.76)
∂
+ φ(S, A, t, V )
∂V
Step 3. The second prolongation is defined in our case by the expression
∂ ∂ ∂ ∂
P r(2) X = X + φS + φA + φt + φSS (74.77)
∂VS ∂VA ∂Vt ∂VSS
Step 4. The Lie symmetries of the equation are determined by the condition
P r(2) X[∆(S, A, t, V (2) )] = 0 as long as ∆(S, A, t, V (2) ) = 0 (74.78)
Implementation of the equation P r(2) X[∆(S, A, t, V (2) )] = 0:
We have
P r(2) X[∆(S, A, t, V (2) )] = 0
∂V 1 ∂ 2V ∂V ∂V
⇐⇒P r(2) X[ + σ2S 2 2 + S + rS − rV ] = 0
∂t 2 ∂S ∂A ∂S
1 (74.79)
⇐⇒ξ1 [σ 2 SVSS + rVS + VA ] − rφ + rSφS + SφA + φt
S
1 2 2 SS
+ σ S φ =0
2
1103
Step 5. Calculation of φS , φA , φt , φSS . We have the following expressions
φS = φS + (φV − ξ1S )VS − ξ1V VS2 − ξ3S Vt

(74.80)
φA = φA + (φV − ξ2A )VA − ξ2V VA2 − ξ3A Vt

(74.81)
φt = φt + (φV − ξ3t )Vt − ξ3V Vt2 − ξ1t VS
(74.82)
φSS = φSS + (2φSV − ξ1SS )VS − ξ3SS Vt + (φV V − 2ξ1SV )VS2
− 2ξ2S VSA − 3ξ1V VS VSS − ξ2V VA VSS − 2ξ3V VS VSt (74.83)
− ξ2V V VS2 VA − ξ2SS VA
1104
Step 6. Therefore, using the previous expressions for φS , φA , φt and φSS ,
we obtain from the last equation of Step 4:
1
S
+ rS φS + (φV − ξ1S )VS − ξ1V VS2 − ξ3S Vt


+ S φA + (φV − ξ2A )VA − ξ2V VA2 − ξ3A Vt


+ φt + (φV − ξ3t )Vt − ξ3V Vt2 − ξ1t VS (74.84)


1 2 2
+ σ S φSS + (2φSV − ξ1SS )VS − ξ3SS Vt + (φV V − 2ξ1SV )VS2
2

2
1105
Step 7. We now have to take into account the condition ∆(S, A, t, V (2) ) = 0.
1
We substitute Vt by − σ 2 S 2 VSS − rSVS − SVA + rV into (74.84):
2
1
S
+ rS φS + (φV − ξ1S )VS − ξ1V VS2 − (ξ2S + ξ2V VS )VA
1
2 2
− (ξ3S + ξ3V VS ) − σ S VSS − rSVS − SVA + rV
2

+ S φA + (φV − ξ2A )VA − ξ2V VA2 − ξ1A VS − ξ1V VA VS
1
2 2
− (ξ3A + ξ3V VA ) − σ S VSS − rSVS − SVA + rV
2

1 2
2 2 (74.85)
− ξ3V − σ S VSS − rSVS − SVA + rV
2
1
2 2
− (ξ2V VA + ξ1V VS ) − σ S VSS − rSVS − SVA + rV
2

1
+ σ 2 S 2 φSS + (2φSV − ξ1SS )VS + (φV V − 2ξ1SV )VS2 − ξ1V V VS3
2
− (ξ3SS + 2ξ3SV VS + ξ3V V VS2 + ξ3V VSS )×
1
2 2
× − σ S VSS − rSVS − SVA + rV
2
− 2ξ2SV VA VS − 2ξ3S VSt − 2ξ2V VS VSA + (φV − 2ξ1S )VSS

2
Step 8. Equate all the coefficients of the partial derivatives of the function V
to zero. We are then going to derive the system of the determining partial dif-
ferential equations. Before that, and just for the economy of the calculations,
we observe that we get some simple expressions looking at the coefficients of
the mixed prtial derivatives. In fact the coefficient of the derivative VSA is
1 2 2 1
σ S (−2ξ2S ) and that of VS VSA is σ 2 S 2 (−2ξ2V ) which means that ξ2S = 0
2 2
1106
and ξ2V = 0. Therefore the coefficient ξ2 depends only on A and t:
ξ2 = ξ2 (A, t) (74.86)
1
The coefficient of the derivative VS VSt is σ 2 S 2 (−2ξ3V ) and that of VSt is
2
1 2 2
σ S (−2ξ3S ) which means that ξ3V = 0 and ξ3S = 0. Therefore the coeffi-
2
cient ξ3 depends only on A and t:
ξ3 = ξ2 (A, t) (74.87)
Because of (74.86) and (74.87), equation (74.85) becomes

1
S
2
+ rS φS + (φV − ξ1S )VS − ξ1V VS

+ S φA + (φV − ξ2A )VA − ξ1A VS − ξ1V VA VS

+ φt + (φV − ξ3t )Vt − ξ1t VS − ξ2t VA (74.88)
1
− (ξ1V VS ) − σ 2 S 2 VSS − rSVS − SVA + rV
2

1
+ σ 2 S 2 φSS + (2φSV − ξ1SS )VS + (φV V − 2ξ1SV )VS2
2

3
− ξ1V V VS − 3ξ1V VS VSS + (φV − 2ξ1S )VSS = 0
The coefficient of VS VSS is the sum of the terms

1 1
(−ξ1V ) − σ 2 S 2 and (−3ξ1V ) σ 2 S 2
2 2
and thus ξ1V = 0, which means that
ξ1 = ξ1 (S, A, t) (74.89)
1107
Therefore (74.88) becomes
1
S
+ rS φS + (φV − ξ1S )VS

+ S φA + (φV − ξ2A )VA − ξ1A VS
(74.90)

1 2 2
+ σ S φSS + (2φSV − ξ1SS )VS + φV V VS2
2

+ (φV − 2ξ1S )VSS = 0
From the above equation we get the following coefficients, which have to be
set to zero:
Coefficient of VS2 :
1 2 2
σ S φV V = 0 (74.91)
2
Coefficient of VA :
S 2 ξ3A + S(ξ3t − ξ2A ) + ξt − ξ2t = 0 (74.92)
Coefficient of VSS :
1 1
σ 2 Sξ1 + σ 2 S 2 (ξ3t − 2ξ1S ) + σ 2 S 3 ξ3A = 0 (74.93)
2 2
Coefficient of VS :
rξ1 + rS(ξ3t − ξ1S ) − Sξ1S + rS 2 ξ3A

1 (74.94)
+ σ 2 S 2 (2φSV − ξ1SS ) − ξ1t = 0
2
Zeroth-order coefficient
− rφ + rSφS + SφA − rSV ξ3A + φt + rV (φV − ξ3t )
1 (74.95)
+ σ 2 S 2 φSS = 0
2
1108
We now have to solve the system of the determining equations (74.91)-
(74.95). From (74.91) we get φV V = 0 and therefore φ is a linear function
with respect to V :
φ = α(S, A, t)V + β(S, A, t) (74.96)
From (74.93) we get the equivalent equation

1 1 1
ξ1S − ξ1 = ξ3t + Sξ3A (74.97)
S 2 2
The above is a linear first order differential equation with unknown function
ξ1 and admits a general solution given by
1 1
ξ1 = (S ln(S))ξ3t + S 2 ξ3A + Sθ(A, t) (74.98)
2 2
where θ(A, t) is a function to be determined.
From (74.92), using the expression for ξ1 given by (74.98), we obtain the
equation
3 1
ξ3A S 2 + ξ3t (S ln(S)) + [ξ3t − ξ2A + θ(A, t)]S − ξ2t = 0 (74.99)
2 2
Since the previous equation should be true for every S, each one of the
coefficients has to be zero. We thus obtain four equations
ξ3A = 0, ξ3t = 0 (74.100)
ξ3t − ξ2A + θ(A, t) = 0 (74.101)

ξ2t = 0 (74.102)
From (74.100), taking into account (74.87), we see that ξ3 is a constant
ξ3 = a1 (74.103)
From (74.102), taking into account (74.86), there follows that ξ2 depends on
A only:
ξ2 = f (A) (74.104)
We then have from (74.101) that
θ(A, t) = f 0 (A) (74.105)
1109
From (74.94), using the expressions (74.96) and (74.98) can take the form
σ 2 αS = f 00 (A) and then, after integration
σ 2 α(S, A, t) = f 00 (A)S + ρ(A, t) (74.106)
From (74.95), taking into account the expressions for φ and ξ3 given by
(74.96) and (74.103) respectively, we obtain the equation

1 2 2
− rβ + rSβS + SβA + βt + σ S βSS
2
(74.107)
1 2 2
+ αt + rSαS + SαA + σ S αSS V = 0
2
Equating the coefficients of V to zero, we obtain the next two equations

1
−rβ + rSβS + SβA + βt + σ 2 S 2 βSS = 0 (74.108)
2
and
1
αt + rSαS + SαA + σ 2 S 2 αSS = 0 (74.109)
2
Equation (74.108) expresses the fact that β(S, A, t) introduced in (74.96) is
any solution of the original PDE (74.12). Equation (74.109) can take the
form (taking into account (74.106))
ρt (A, t) + [ρA (A, t) + rf 00 (A)]S + f 000 (A)S 2 = 0 (74.110)
Since the previous equation should hold for any value of S, we get the fol-
lowing three equations
ρt (A, t) = 0 (74.111)
ρA (A, t) + rf 00 (A) = 0 (74.112)

f 000 (A) = 0 (74.113)
The above equation means that f is a second degree polynomial with respect
to A:
f (A) = a2 + a3 A + a4 A2 (74.114)
1110
From equation (74.111) we get that
ρ = g(A) (74.115)
But then, equation (74.112) gives g 0 (A) + 2ra4 = 0, and by integration
g(A) = −2ra4 A + a5 (74.116)
Therefore
ρ(A, t) = −2ra4 A + a5 (74.117)
Let us now collect everything together. From (74.98), taking into account
(74.103), (74.105) and (74.114) we obtain the following expression for ξ1 :
ξ1 = S(a3 + 2a4 A) (74.118)
ξ2 = a2 + a3 A + a4 A2 (74.119)
From (74.106), using (74.114) and (74.117), we obtain the following expres-
sion for α:
1
α(S, A, t) = 2 (2a4 S − 2ra4 A + a5 ) (74.120)
σ
We thus have the following expression for φ:
1
φ(S, A, t) = (2a4 S − 2ra4 A + a5 )V + β(S, A, t) (74.121)
σ2
We are now in a position to write down the expression for the vector X, the
generator of the symmetries, introduced in (74.76):
∂ ∂ ∂
X = S(a3 + 2a4 A) + (a2 + a3 A + a4 A2 ) + a1
∂S ∂A ∂t
(74.122)
1 ∂
+ 2 (2a4 S − 2ra4 A + a5 )V + β(S, A, t)
σ ∂V
∂ ∂ ∂ ∂ 1 ∂
X = a1 + a2 + a3 S +A + a5 2 V
∂t ∂A ∂S ∂A σ ∂V (74.123)
∂ ∂ 2 ∂ ∂
+ a4 2AS + A2 + 2 (S − rA)V + β(S, A, t)
∂S ∂A σ ∂V ∂V
1111
Therefore we conclude from (74.123) that the Lie algebra of the infinitesimal
transformations of the original equation (74.12) is spanned by the five vectors
given by (74.21) and the infinite dimensional Lie sub-algebra with generator
given by (74.22).
Appendix B. In this Appendix we transform the original PDE (74.12) into
the Black-Scholes equation. Using the transformation (74.29), we can express
the various partial derivatives of the function V (S, A, t) in terms of the partial
derivatives of the function u(x, t). We find that
S
n
Vt = ut A exp b
A
S
VS = (ux An−2 + buAn−1 )exp b
A (74.124)
S
n−4 n−3 2 n−2
VSS = (uxx A + 2bux A + b uA )exp b
A
S
VA = (−2ux SAn−3 + nuAn−1 − buSAn−2 )exp b
A
Substituting the above expressions into the origival equation (74.12) we ob-
tain the equation
1
An ut + σ 2 (S 2 An−4 )uxx
2
1 2 2 n−3 2 n−2 2 n−3
+ σ (2b)(S A ) + r(S A ) − 2(S A ) ux
2

1 (74.125)
+ σ 2 b2 (S 2 An−2 ) + rb(S 2 An−1 ) + n(S 2 An−1 ) − b(S 2 An−2 )
2

n
− rA u = 0
2
Using S = xA2 and b = 2 , we may simplify the coefficients of (74.125)
σ
equation. In fact we find
S 2 An−4 = x2 An
1 2
σ (2b)(S 2 An−3 ) + r(S 2 An−2 ) − 2(S 2 An−3 ) = rxAn
2 (74.126)
1 2 2 2 n−2
σ b (S A ) + rb(S 2 An−1 ) + n(S 2 An−1 ) − b(S 2 An−2 )
2
− rAn = −rAn
1112
Substituting the previous expressions in (74.125) we get the equation
1
An ut + σ 2 x2 uxx + rxux − ru = 0
2
1
ut + σ 2 x2 uxx + rxux − ru = 0 (74.127)
2
which is the well-known Black-Scholes equation.
Appendix C. In this appendix we shall transform the equation
ut = a2 uxx + bux + cu (74.128)
to the heat equation. The coefficients a, b and c are constants.

We have already proved in Section 13.3 (see (13.251)) that the substitution
v = e−ct ·u(t, x = y − bt) (74.129)
converts equation (74.128) into the heat equation vt = a2 vyy . We are not
going to repeat the calculations in this Section.
Appendix D. Weber’s Differential Equation. Abramowitz and Stegun
([41], §19.17) define as standard solutions of Weber’s equation
1
y 00 = a − x2 y (74.130)
4
the functions W (a, ±x) where
(cosh(πa))1/4 √
W (a, ±x) = √ G1 y1 ∓ 2 G3 y2 (74.131)
2 π
with
1 1 3 1
G1 = Γ + ia and G3 = Γ + ia (74.132)
4 2 4 2
and y1 , y2 are defined by the series expansions
x 2 2 1 x4 3 7 x6 4 15 x8
y1 = 1+a + a − + a − a + a −11a2 + +· · · (74.133)
2! 2 4! 2 6! 4 8!
1113
and
x3 2 3 x5 3 13 x7 4 9
2 63 x
y2 = x+a + a − + a − a + a −17a + +· · · (74.134)
3! 2 5! 2 7! 4 9!
xn
respectively. In the previous expressions, the nonzero coefficients an of
n!
are connected by
1
an+2 = a·an − n(n − 1)an−2 (74.135)
4
It is also worth of noting that W (a, ±x) can be expressed in terms of the
Confluent Hypergeometric functions ([41], §19.25)
r 3 1 1 r 2G 1 1 1
−3/4 G1 2 3
W (a, ±x) = 2 H − , a, x ± xH − , a, x2
G3 4 2 4 G1 4 2 4
where
H(m, n, x) = e−ix M (m + 1 − in; 2m + 2; 2ix) (74.136)
Note. Weber’s equation can be obtained for example by separation of vari-

ables for the Laplace equation in parabolic cylindrical coordinates (ref.[42],
Chapter 10).
Appendix E. The Airy equation. Airy’s differential equation
w00 − zw = 0, w = w(z) (74.137)
has solutions the linearly independent functions (ref.[41], §10.4)
Ai(z) and Bi(z) (74.138)
which are called Airy functions. These functions are defined by

√
Ai(z) = c1 f (z) − c2 g(z), Bi(z) = 3 [c1 f (z) + c2 g(z)] (74.139)
where
∞
X 1 z 3k
1 3 1·4 6 1·4·7 9
f (z) = 1 + z + z + z + ··· = 3k (74.140)
3! 6! 9! k=0
3 k (3k)!
1114
and
∞
X 2 z 3k+1
2 2·5 7 2·5·8 10
g(z) = z + z 3 + z + z +··· = 3k (74.141)
4! 7! 10! k=0
3 k (3k + 1)!
respectively. The constants c1 and c2 are given by

Bi(0) 1
c1 = Ai(0) = √ = 2/3 ,
3 3 Γ(2/3)
(74.142)
Bi0 (0) 1
c2 = −Ai0 (0) = √ = 1/3
3 3 Γ(1/3)
The Airy functions can also be expressed in terms of Bessel’s functions of
fractional order ([41], §10.4.14 and §10.4.18)
r r
1 z z
Ai(z) = K1/3 (ζ), Bi(z) = [I−1/3 (ζ) + I1/3 (ζ)] (74.143)
π 3 3
√
where ζ = 32 z z.
References
[1] M. Rubinstein: ”Exotic Options”. Preprint, University of California
at Berkeley, 1991
[2] P. G. Zhang: ”Exotic Options” Second Edition, World Scientific 1998
[3] P. Wilmott: ”Quantitative Finance”.
Three-Volume Set, Second Edition, Wiley 2006
[4] A G Z Kemna and A C F Vorst: ”A pricing method for options”.
Journal of Banking and Finance 14 (1990) 113-129
[5] J. Angus: ”A note on pricing Asian Derivatives with continuous geo-
metric averaging”. Journal of Future Markets 19 (1999) 845-858
[6] Y. Z. Bergman: ”Pricing path contingent claims”.
Research in Finance 5 (1985) 229-241
[7] H. Geman and M. Yor: ”Bessel Processes, Asian Options and perpe-
tuities”. Mathematical Finance 3 (1993) 349-375
[8] E. Barucci, S. Polidoro and V. Vespri: ”Some results on partial
differential equations and Asian Options”. University of Pisa preprint
[9] L. Jiang and M. Dai: ”Convergence Analysis of Binomial Tree Method
for American-type Path-dependent Options”. Preprint, Peking University
[10] E. G. Haug: ”Option Pricing Formulas”.
1115
Second Edition, McGraw Hill 2007
[11] S. Lie: ”Gesammelte Werke”, Vols. 1-7, F. Engel and P. Hergard (Eds),
[13] J. M. Page: ”Ordinary Differential Equations with an Introduction to
Lie’s Theory of the Group of one parameter”.
McMillan and Co. New York 1897
[14] A. Cohen: ”An Introduction to the Lie Theory of one-parameter
groups: With Applications to the solution of Differential Equations”.
D. C. Heath and Co. Boston 1911
Graduate Texts in Mathematics, vol.107. Springer, NY 1993
[16] H. Stephani: ”Differential Equations: Their Solutions Using Symme-
Differential Equations”. IOP Publishing, 1999
tions in Applications” SIAM, Philadelphia 1979
Differential Equations”. Vol.1 1994, Vol.2 1995, Vol.3 1996.
CRC Press, Boca Raton, FL.
Lie Symmetries of Differential Equations”. Euromath Bulletin 2, No.1, 1993
[24] P. E. Hydon: ”Symmetry Methods for Differential Equations: A Be-
ginners Guide”. CUP 2000
[26] B. J. Cantwell: ”Introduction to Symmetry Analysis”. CUP 2002
[27] Steeb Willi-hans: ”Continuous Symmetries, Lie Algebras, Differential
Equations and Computer Algebra”. World Scientific 2007
[28] F. Schwarz: ”Algorithmic Lie Theory for Solving Ordinary Differential
1116
Equations”. Chapman and Hall/CRC, 2008
[30] S. M. Antoniou: ”A General Solution to Stochastic Equity Volatility
Equation (the Bensoussan-Crouhy-Galai Equation)”, SKEMSYS preprint
[31] S. M. Antoniou: ”Portfolio Optimization using the Hamilton-Jacobi-
Bellman Equation”. SKEMSYS preprint
[32] J. Goard: ”New Solutions to the Bond-Pricing Equation via Lies Clas-
sical Method”. Mathematical and Computer Modelling 32 (2000) 299-313
derivatives”. arXiv:math.AP/0604207
linear model of financial derivatives”.
Int. J. Theor. Appl. Finance 10 (2007) 1-21
illiquid markets: scaling properties and explicit solutions”.
arXiv:0708.1568v1 [math.AP] 11 Aug 2007
[36] G. Silberberg: ”Discrete Symmetries of the Black-Scholes Equation”
Proceedings of 10th International Conference in Modern Group Analysis.
[37] S. M. Antoniou: ”Lie Symmetry Analysis of the partial differential
equations appearing in Finance”. SKEMSYS Report
[38] W. Sinkala, P. G. L. Leach and J.G. O’Hara: ”Zero-coupon bond
prices in the Vasicek and CIR models: Their computation as group-invariant
solutions”. Mathematical Methods in the Applied Sciences, 2007
[39] C. A. Pooe, F. M. Mahomed and C. Wafo Soh: ”Fundamental
Solutions for Zero-Coupon Bond Pricing Models”.
Nonl. Dyn. 36 (2004) 69-76
[40] P. G. L. Leach, J.G. O’Hara and W. Sinkala: ”Symmetry-based
investment”. J. Math. Anal. Appl. 2007
[41] M. Abramowitz and I.A. Stegun: ”Handbook of Mathematical
[42] N. N. Lebedev: ”Special Functions and their Applications”.
Dover 1972
1117
Chapter XXIV
Applications in Cosmology
75 The Kompaneets Equation
The Kompaneets Equation was introduced by Kompaneets and was later
used by Sunyaev and Zeldovich. The classical Kompaneets Equation has the
form
1
ut = 2 [x4 (ux + u2 + u)]x (75.1)
x
We also have the Generalized Kompaneets Equation which appears to have
the form
1
ut = 2 [x4 (ux + f (u))]x (75.2)
x
In expanded form, equations (75.1) and (75.2) have the form
ut = x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x (75.3)
and
ut = x2 uxx + [ f 0 (u)x2 + 4x ] ux + 4f (u)x (75.4)
respectively.
75.1. Classical Lie Symmetries.
Let us first try to solve equation (75.3). Introducing the notation
∆ = ut − x2 uxx − [ (2u + 1)x2 + 4x ] ux − 4(u2 + u)x (75.5)
invariance of this equation is being expressed by the condition
P r(2) Γ(∆) =0 (75.6)

∆=0
where
∂ ∂ ∂
P r(2) Γ = T +X +U
∂t ∂x ∂u (75.7)
∂ ∂ ∂
+ U [t] + U [x] + U [xx]
∂ut ∂ux ∂uxx
1118
We shall need the following expressions (see (12.14), (12.15) and (13.13))
U [t] = Ut + (Uu − Tt )ut − Xt ux − Tu (ut )2 − Xu ut ux (75.8)
U [x] = Ux + (Uu − Xx )ux − Tx ut − Xu (ux )2 − Tu ux ut (75.9)
(75.10)
We now have
∂ ∂ ∂
(2)
P r Γ(∆) = T + X +U
∂t ∂x ∂u (75.11)
[t] ∂ [x] ∂ [xx] ∂

+U +U +U (∆) = 0
∂ut ∂ux ∂uxx
which, in expanded form, because of (75.5), takes on the form
X 2xuxx + [2(2u + 1)x + 4]ux + 4(u2 + u)

+ U 2x2 ux + 4(2u + 1)x − U [t]

(75.12)
+ U [x] (2u + 1)x2 + 4x + x2 U [xx] = 0

Using (75.8)-(75.10), the above equation becomes

X 2xuxx + [2(2u + 1)x + 4]ux + 4(u2 + u)

+ U 2x2 ux + 4(2u + 1)x

2
− Ut + (Uu − Tt )ut − Xt ux − Tu (ut ) − Xu ut ux

2

+ (2u + 1)x + 4x Ux + (Uu − Xx )ux − Tx ut

2
− Xu (ux ) − Tu ux ut (75.13)

2
+ x Uxx − Txx ut + (2Uxu − Xxx )ux − 2Txu ut ux


1119
in (75.13) ut by x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x and we get the
equation
X 2xuxx + [2(2u + 1)x + 4]ux + 4(u2 + u)

+ U 2x2 ux + 4(2u + 1)x

− Ut + (Uu − Tt − Xu ux ) x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x

2 2 2 2
− Xt ux − Tu x uxx + [ (2u + 1)x + 4x ] ux + 4(u + u)x

+ (2u + 1)x + 4x Ux + (Uu − Xx )ux − Xu (ux )2
2

2 2 2
− (Tx + Tu ux ) x uxx + [ (2u + 1)x + 4x ] ux + 4(u + u)x

2
+ x Uxx − (Txx + 2Txu ux ) x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x

+ (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2

− Tuu (ux )2 x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x


− Tu uxx x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x

− 3Xu ux uxx = 0
(75.14)

the function u. However, instead of expanding the previous equation, we can
see that some coefficients can be traced immediately. That procedure may
simplify the whole calculation. For example the coefficient of utx is −2Tx
and the coefficient of ux utx is −2Tu . We thus have Tx = 0 and Tu = 0, which
means that
T = T (t) (75.15)
On the other hand, in the remaining part of the equation (after considering
that T = T (t)), the coefficient of ux uxx is −2x2 Xu and thus Xu = 0 which
1120
means that
X = X(t, x) (75.16)
Because of the previous two equations, equation (75.14) gets simplified to

the equation
X 2xuxx + [2(2u + 1)x + 4]ux + 4(u2 + u)

+ U 2x2 ux + 4(2u + 1)x

− Ut + (Uu − Tt ) x2 uxx + [ (2u + 1)x2 + 4x ] ux + 4(u2 + u)x

− Xt ux (75.17)

2

+ (2u + 1)x + 4x Ux + (Uu − Xx )ux

2 2
+ x Uxx + (2Uxu − Xxx )ux + Uuu (ux ) + (Uu − 2Xx )uxx = 0
From the above equation, we can now equate to zero all the coefficients of
the function u. We have
The coefficient of (ux )2 is equal to x2 Uuu and thus Uuu = 0 from which we
get
U = a(t, x)·u + b(t, x) (75.18)
The coefficient of uxx , when equated to zero, is equal to
2xX − (Uu − Tt )x2 + x2 (Uu − 2Xx ) = 0

1 1
Xx − X = Tt (75.19)
x 2
which is a linear first order differential equation with respect to x. This
equation admits a solution is given by
1
X(t, x) = Tt ln(x) + c(t) x (75.20)
2
1121
The coefficient of ux , when equated to zero, is equal to
X[2(2u + 1)x + 4] + 2x2 U + (Tt − Xx )[(2u + 1)x2 + 4x]

(75.21)
+ Xt + x2 (2Uxu − Xxx ) = 0
The coefficient of zero order derivative is given by
4(u2 + u)X + 4(2u + 1)xU − Ut − 4(Uu − Tt )(u2 + u)x

(75.22)
+ [(2u + 1)x2 + 4x]Ux + x2 Uxx = 0
We substitute the functions U and X given by (75.18) and (75.20) respec-

tively to both equations (75.21) and (75.22) and we obtain the equations
(arranged in powers of u)

x [1 + ln(x)]Tt + 2c(t) + 2a(t, x) u
1
+ [3 + x + x ln(x)]Tt + 2xc(t) (75.23)
2
+ 4xb(t, x) + 4xax + Ttt ln(x) + 2c0 (t) = 0
and

2x 2a(t, x) + xax + [2 + ln(x)]Tt + 2c(t) u2

+ 8x b(t, x) + 2x[2 + ln(x)] Tt + 4x c(t) − at + x2 axx
(75.24)
+ (x2 + 4x)ax + 2x2 bx u
+ x2 bxx + (x2 + 4x)bx + 4xb(t, x) − bt = 0
respectively.
Equating to zero the coefficients of u, we obtain from (75.23) the following
two equations
[1 + ln(x)]Tt + 2c(t) + 2a(t, x) = 0 (75.25)
and
[3 + x + x ln(x)]Tt + 2xc(t)
(75.26)
+ 4xb(t, x) + 4xax + Ttt ln(x) + 2c0 (t) = 0
1122
and from (75.24) the following three equations
2 a(t, x) + xax + [2 + ln(x)]Tt + 2c(t) = 0 (75.27)
8x b(t, x) + 2x[2 + ln(x)] Tt + 4x c(t)

(75.28)
− at + x2 axx + (x2 + 4x)ax + 2x2 bx = 0
x2 bxx + (x2 + 4x)bx + 4xb(t, x) − bt = 0 (75.29)
From equation (75.25) we can determine the function a(t, x):
1
a(t, x) = −c(t) − [1 + ln(x)]Tt (75.30)
2
Using the above value of a(t, x) into (75.26), we can determine the function
b(t, x). We find
1
b(t, x) = − [x(1 + ln(x)) + 1]Tt + 2[x c(t) + c0 (t)] + ln(x) Ttt (75.31)
4x
1
Using the above values of a(t, x) and b(t, x) into (75.27) we find Tt = 0 and
2
thus
T (t) = a1 (75.32)
From equation (75.28), using (75.30)-(75.32) we find c0 (t) = 0 and thus
c(t) = a2 (75.33)
Finally from (75.29), taking into account (75.30)-(75.33) we find c(t) = 0.

Therefore
a2 = 0 (75.34)
We thus see that the generator of the Lie symmetries is given by

∂
Γ= (75.35)
∂t
75.2. Non-Classical Kompaneets.
In determining the Non-Classical symmetries of the Kompaneets equation,
1123
we consider as usual the Invariant Surface Condition Xux + T ut − U = 0 for
T = 1, i.e.
ut = U − Xux (75.36)
We consider again equation (75.13) and we substitute ut by U − Xux and

thus uxx by
U − Xux (2u + 1)x + 4 4(u2 + u)

− ux − (75.37)
x2 x x
We then obtain the equation (for T = 1)
U − Xu 4(u2 + u)

x (2u + 1)x + 4
X 2x − ux −
x2 x x

+ [2(2u + 1)x + 4]ux + 4(u2 + u)
+ U 2x2 ux + 4(2u + 1)x

− Ut + (Uu − Xu ux )(U − Xux ) − Xt ux

2
+ (2u + 1)x + 4x Ux + (Uu − Xx )ux − Xu (ux ) 2 (75.38)

+ x Uxx + (2Uxu − Xxx )ux + (Uuu − 2Xxu )(ux )2 − Xuu (ux )3
2
4(u2 + u)

U − Xux (2u + 1)x + 4
+ (Uu − 2Xx ) − ux −
x2 x x
2

U − Xux (2u + 1)x + 4 4(u + u)
− 3Xu ux − ux − =0
x2 x x
1124
− x2 ·Xuu (ux )3

+ 2XXu + (3x + 6x u + 12x)Xu − 2x Xxu + x Uuu (ux )2
2 2 2 2

+ (−2U + 12xu2 + 12xu)Xu + (2x2 + 8x + 4x2 u + 2X)Xx
− (4x + x2 + 2x2 u)Uu + Xt + 2x2 Uxu − x2 Xxx

2X (75.39)
2
+ 2x U + X − − 2 ux
x

2U
+ X· − 2(u2 + u)X − (4xu2 + 4xu)Uu + 4(2u + 1)xU
x

2 2
+ (8xu + 8xu − 2U )Xx + x Uxx − Ut = 0
Equating to zero the coefficients of the partial derivatives of the function u,

we obtain from equation (75.39) the following system of equations
Xuu = 0 (75.40)
2XXu + (3x2 + 6x2 u + 12x)Xu − 2x2 Xxu + x2 Uuu = 0 (75.41)

(−2U + 12xu2 + 12xu)Xu + (2x2 + 8x + 4x2 u + 2X)Xx
− (4x + x2 + 2x2 u)Uu + Xt + 2x2 Uxu − x2 Xxx (75.42)
2X
2
+ 2x U + X − −2 =0
x
2U
X· − 2(u2 + u)X − (4xu2 + 4xu)Uu + 4(2u + 1)xU
x (75.43)
+ (8xu + 8xu2 − 2U )Xx + x2 Uxx − Ut = 0
X = α(x, t)u + β(x, t) (75.44)
Because of the above expression of the function X, equation (75.41) gives us
2{α(x, t)u + β(x, t)} α(x, t) + (3x2 + 6x2 u + 12x) α(x, t)

(75.45)
− 2x2 αx (x, t) + x2 Uuu = 0
1125
Integrating twice with respect to u the above equation, we can determine the
function U :

1 1
U = 2 δ(x, t) + γ(x, t)u − α(x, t) [α2 (x, t) + 3x2 ] u3
x 3
(75.46)
3 2
2 2 2
− α(x, t) β(x, t) + x + 6x u + x αx (x, t)u
2
where γ(x, t) and δ(x, t) are arbitrary functions to be determined later on.
Using the expressions (75.44) and (75.46) for the functions X and U respec-
tively, we find an equation which, when arranged in powers of u, takes the
form
2
α(α2 + 5x2 α + 6x4 )u3
3
+ {(−4αx + 6α)x4 + 36αx3 + (−4ααx + 2αβ + 4α2 )x2
+ 18α2 x + 2α2 β}u2
+ {(−6αx + 3αxx + 3α + 4βx )x4 + (36α − 24αx )x3 (75.47)
+ (2αβ − 2αx β + 70α + αt − 2αβx )x2 + 12αβx − 2αγ}u
+ {(2βx − βxx )x4 + 8βx x3 + (2δ + 2γx + βt − 2β + 2ββx − γ)x2
+ (−2β 2 − 8γ)x − 2αδ} = 0
We now have to equate to zero the coefficients of u in the above equation.

Doing so, we arrive at the following set of equations
α2 + 5x2 α + 6x4 = 0 (75.48)
(−4αx + 6α)x4 + 36αx3 + (−4ααx + 2αβ + 4α2 )x2

(75.49)
+ 18α2 x + 2α2 β = 0
(−6αx + 3αxx + 3α + 4βx )x4 + (36α − 24αx )x3
(75.50)
+ (2αβ − 2αx β + 70α + αt − 2αβx )x2 + 12αβx − 2αγ = 0
(2βx − βxx )x4 + 8βx x3 + (2δ + 2γx + βt − 2β + 2ββx − γ)x2
(75.51)
+ (−2β 2 − 8γ)x − 2αδ = 0
Equation (75.48) is a quadratic equation with respect to α which admits two
solutions:
α(x, t) = −3x2 or α(x, t) = −2x2 (75.52)
1126
Case I. If α(x, t) = −3x2 , we find from (75.49) the function β:
1
β(x, t) = − (3x + 1)x (75.53)
2
Using the values of α and β in (75.50) we determine the function γ:
7
γ(x, t) = (9x + 11)x2 (75.54)
6
We finally obtain from (75.51), using the values of the functions α, β and γ,
the function δ:
1
δ(x, t) = (36x2 + 154x + 163)x (75.55)
24
Case II. If α(x, t) = −2x2 , we find, using the set of equations (75.49)-(75.51)
that
1
β(x, t) = −(x − 4)x, γ(x, t) = (x2 + 32x + 44)x2
2 (75.56)
1
δ(x, t) = (x3 + 36x2 + 132x + 128)x
12
References.
[1] A. S. Kompaneets: ”The Establishment of Thermal Equilibrium
between Quanta and Electrons”. Sov. Phys. JETP 4 (1957) 730.
Chapter XXV
Additional Equations
76 The mKdV-KP Equation
We consider the mKdV-KP equation in the form
3
ut − ux + 6u2 ux + uxxx + uyy = 0, u = u(t, x, y) (76.1)
2 x
1127
Invariance of this equation under Lie point transformations is being expressed
by the condition
P r(4) X[∆]|∆=0 = 0 (76.2)
where P r(4) X is the fourth prolongation of the vector field X
∂ ∂ ∂ ∂
P r(4) X = X + φ[t] + φ[x] + φ[tx] + φ[xx]
∂ut ∂ux ∂ux ∂uxx
(76.3)
∂ ∂
+ φ[xxxx] + φ[yy]
∂uxxxx ∂uyy
with
∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 +φ (76.4)
∂t ∂x ∂y ∂u
and
3
∆ ≡ ut − ux + 6u2 ux + uxxx + uyy (76.5)
2 x
Implementation of the condition (76.2) leads to the determination of ξ 1 , ξ 2 , ξ 3

and φ:
ξ 1 = a1 − 3a5 t, ξ 2 = a2 + a4 y + a5 (3t − x),
(76.6)
ξ 3 = a3 − 2a4 t − 2a5 y, φ = a5 u
We thus see that the algebra of Lie symmetries of equation (76.1) is spanned
by the generators
∂ ∂ ∂ ∂ ∂
X1 = , X2 = , X3 =
, X4 = y − 2t ,
∂t ∂x ∂y ∂x ∂y
(76.7)
∂ ∂ ∂ ∂
X 5 = −3t + (3t − x) − 2y +u
∂t ∂x ∂y ∂u
The optimal system of generators is given by
X1
αX 1 + X 2
αX 1 + X 4 (76.8)
αX 1 + βX 2 + X 3
αX 1 + βX 2 + X 5
1128
I. Considering X 1 , we introduce the (independent of t) function u(t, x, y) =
F (x, y) which, when substituted to (76.1), yields the equation
2Fxxxx + 12F 2 Fxx + 24F (Fx )2 − 3Fxx + 2Fyy = 0 (76.9)
II. Considering αX 1 + X 2 , we introduce the similarity variables

αx − t
ξ= , η=y (76.10)
α
and the function
u(t, x, y) = F (ξ, η) (76.11)
which, when substituted to equation (76.1), yields the equation
2α(Fξξξξ + Fηη ) + 12αF 2 Fξξ − (2 + 3α)Fξξ + 24αF (Fξ )2 = 0 (76.12)
III. Considering αX 1 + X 4 , we introduce the similarity variables
αy + t2 3α2 x − 2t3 − 3αty

ξ= , η= (76.13)
α 3α2
and the function
u(t, x, y) = F (ξ, η) (76.14)
2α(Fηηηη + Fξξ ) + 12αF 2 Fηη − (2ξ + 3α)Fηη + 24αF (Fη )2 = 0 (76.15)
IV. Considering αX 1 + βX 2 + X 3 we introduce the similarity variables

αy − t αx − βt
ξ= , η= (76.16)
α α
and the function
u(t, x, y) = F (ξ, η) (76.17)
2αFηηηη +2αFξξ −2Fξη +12αF 2 Fηη −(2β +3α)Fηη +24αF (Fη )2 = 0 (76.18)
1129
V. Considering αX 1 + βX 2 + X 5 we introduce the similarity variables
y 2x + 3t − 3α − 2β
ξ= p , η= √ (76.19)
3
(3t − α)2 2 3 3t − α
and the function
F (ξ, η)
u(t, x, y) = √
3
(76.20)
3t − α
Fηηηη + Fξξ + 12F (Fη )2 − 2ξFξη + (6F 2 − η)Fηη − 2Fη = 0 (76.21)
A. Solution of equation (76.9). Let us first solve equation (76.9). This
∂ ∂
equation is invariant under the the symmetry generator + . It thus
∂x ∂y
admits the similarity variable ξ and the function G(ξ) given by
ξ = −x + y, F (x, y) = G(ξ) (76.22)
and then it gets transformed into the equation
2G(4) + (12G2 − 1)G00 + 24G(G0 )2 = 0 (76.23)
The above equation cannot be reduced further using Lie symmetries. In that
case we apply Nucci’s reduction algorithm. We introduce the notation
w1 = G(ξ), w2 = G0 (ξ), w3 = G00 (ξ), w4 = G000 (ξ) (76.24)
We thus see that
dw1
= w2
dξ
dw2
= w3 (76.25)
dξ
dw3
= w4
dξ
Equation (76.23) then, because of the notation we have introduced, trans-
forms into (since w40 = G(4) )
dw4 (1 − 12w12 )w3 − 24w1 w22
= (76.26)
dξ 2
1130
Our aim is to transform (76.26) into an equation which will contain w2 as
the dependent variable and w1 as the new independent variable.
Dividing the second and the third of (76.25) and also (76.26) by the first of
(76.25), we obtain the system
dw2 w3
=
dw1 w2
dw3 w4
= (76.27)
dw1 w2
dw4 (1 − 12w12 )w3 − 24w1 w22
=
dw1 2w2
We have (solving the first two of (76.27) with respect to w3 and w4 respec-
tively)

dw2 dw3 d dw2
w3 = w2 and w4 = w2 = w2 w2 (76.28)
dw1 dw1 dw1 dw1
Upon substitution of w3 and w4 given above into the last of (76.27), we obtain
the equation

d d dw2 2
dw
2
2 w2 w2 = (1 − 12w1 ) − 24w1 w2 (76.29)
dw1 dw1 dw1 dw1
We thus have been able to transform equation (76.23) into an equation with
w1 as independent variable with w2 as the unknown function. The previ-
ous equation can be written as (we see that the right hand side is a total
derivative)

d d dw2 1 dw2 d
w2 w2 = − (6w12 w2 ) (76.30)
dw1 dw1 dw1 2 dw1 dw1
and thus it can be integrated once, giving (we have set the integration con-
stant equal to zero)

d dw2 1
w2 w2 = w2 − 6w12 w2 (76.31)
dw1 dw1 2
Dividing by w2 the above equation, we obtain the equation

d dw2 1
w2 = − 6w12
dw1 dw1 2
1131
which when integrated yields
dw2 1
w2 = w1 − 2w13 + C1 (76.32)
dw1 2
d
(w2 ) = w1 − 4w13 + 2C1
dw1 2
and integrating once more,
1
w22 = w12 − w14 + 2C1 w1 + 2C2
2
Solving with respect to w2 we get
1
q
w2 = 2w12 − 4w14 + 8C1 w1 + 8C2 (76.33)
2
dw1
Since w2 = w10 = , we have to solve the equation
dξ
dw1 1
q
= 2w12 − 4w14 + 8C1 w1 + 8C2 (76.34)
dξ 2
The above equation can be solved by separation of variables and its solution
can be expressed in terms of Weierstrass ℘ function. This can be done using
the following Lemma. Based on this Lemma, the reader can work out the
details of converting (76.34) into the appropriate form and thus express the
solution in terms of Weierstrass function.
Lemma. Let
a0 x4 + 4a1 x3 + 6a2 x2 + 4a3 x + a4 = f (x) (76.35)
be any quartic polynomial which has no repeated factors. Let also

g2 = a0 a4 − 4a1 a3 + 3a22
(76.36)
g3 = a0 a2 a4 + 2a1 a2 a3 − a22 − a0 a23 − a21 a4
be its invariants. Let also z be defined by

Z x
dt
z= p (76.37)
x0 f (t)
1132
where x0 is any root of the equation f (x) = 0. By Taylor’s theorem (taking
into account that f (x0 ) = 0) we have
f (t) = 4A3 (t − x0 ) + 6A2 (t − x0 )2 + 4A1 (t − x0 )3 + A0 (t − x0 )4 (76.38)
where
A0 = a0 , A1 = A0 x0 + a1 , A2 = a0 x20 + 2a1 x0 + a2
(76.39)
A3 = a0 x30 + 3a1 x0 62 + 3a2 x0 + a3
Introducing the substitutions
(t − x0 )−1 = τ, (x − x0 )−1 = ξ (76.40)
the integral becomes

Z ∞
dτ
z= √ (76.41)
ξ 4A3 τ + 6A2 τ 2 + 4A1 τ + A0
3
To remove the second term of the cubic appearing under the square root of
the above integral, write
1 −1 1 −1
τ = σ − A2 A3 , ξ = s − A2 A3 (76.42)
2 2
and we get
Z ∞
dσ
z= p (76.43)
s 4σ − (3A2 − 4A1 A3 )σ − (2A1 A2 A3 − A32 − A0 A23 )
3 2
We can verify that
g2 = 3A22 − 4A1 A3 and g3 = 2A1 A2 A3 − A32 − A0 A23 (76.44)
where g2 and g2 are the invariants of the original quartic. Therefore

Z ∞
dσ
z= p (76.45)
s 4σ 3 − g2 σ − g3
and thus
s = ℘(z; g2 , g3 ) (76.46)
1133
Since
1 −1
x = x0 + A3 s − A2 (76.47)
2
we obtain
−1
1 0 1 00
x = x0 + f (x0 ) ℘(z; g2 , g3 ) − f (x0 ) (76.48)
4 24
This formula for x is to be regarded as the integral equivalent to the relation

Z x
dt
z= p (76.49)
x0 f (t)
If
Z x
dt
z= p (76.50)
a f (t)
where a is any constant, not necessarily a zero of f (x) and f (x) is a quartic
polynomial with no repeated factors, then

1 0 1 00 1
[f (a)] ℘ (z) + f (a) ℘(z) − f (a) + f (a)f 00 (a)
1/2 0
2 24 24
x = a+ 2 (76.51)
1 00 1
2 ℘(z) − f (a) − f (a)f (iv) (a)
24 48
Consider the equation

Z ∞
dt
z= p (76.52)
3
4t − g2 t − g3
ζ
determining z in terms of ζ. The path of integration can be any curve which

does not pass through a zero of the polynomial
4t3 − g2 t − g3 (76.53)
We obtain by differentiation
dζ 2
= 4ζ 3 − g2 ζ − g3 (76.54)
dz
1134
and so
ζ = ℘(z + α) (76.55)
The result that the equation

Z ∞
dt
z= p
ζ 4t3 − g2 t − g3
is equivalent to the equation
ζ = ℘(z) (76.56)
is sometimes written in the form

Z ∞
dt
z= p (76.57)
℘(z) 4t3 − g2 t − g3
The Weierstrass function ℘(z) satisfies the differential equation

d℘(z) 2
= 4 ℘3 (z) − g2 ℘(z) − g3 (76.58)
dz
References
[1] E.T. Whittaker and G.N. Watson: ”A Course of Modern Analysis”.
Fourth Edition, Cambridge Univ. Press, 1927
[2] W.S. Burnside and A.W. Panton: ”The Theory of Equations”.
Third Edition, Dublin Univ. Press 1892
(End of Lemma)
B. Solution of equation (76.12). Equation (76.12) admits two symme-

∂ ∂ ∂ ∂
try generators and . Considering the combination + , we may
∂ξ ∂η ∂ξ ∂η
introduce the invariants
θ = −ξ + η, F (ξ, η) = H(θ) (76.59)
2αH (iv) + (12αH 2 − α − 2)H 00 + 24αH(H 0 )2 = 0 (76.60)
1135
Integrating the above equation twice, we get
α+2
H 00 + 2H 3 − H + C2 θ + C1 = 0 (76.61)
2α
The above equation in the special case of C1 = C2 = 0, written as
α+2
H 00 + 2H 3 − H=0 (76.62)
2α
admits a solution expressed in terms of Jacobi’s sn() function.
Lemma. An equation of the form
u00 (x) + a·u3 (x) + b·u(x) = 0 (76.63)
u(x) = m C2 ·sn(m(kx + C1 ), nC2 ) (76.64)
where
s r p
2b a −a(a + 2b)
m= , k= b+ , n= (76.65)
a + 2b − a C22 2 a + 2b
C. Solution of equation (76.15). Equation (76.15) admits the symmetry

∂
generator . Introducing the invariants
∂η
F (ξ, η) = H(θ), θ=ξ (76.66)
H 00 (θ) = 0 (76.67)
H(θ) = C1 θ + C2 (76.68)
Going back to the original variables, we obtain the solution

αy + t2
u(t, x, y) = C1 + C2 (76.69)
α
1136
D. Solution of equation (76.18). Equation (76.18) admits the symmetry
∂ ∂ 1 ∂
generators Γ1 = and Γ2 = 2 − . Considering the combination
∂η ∂ξ α ∂η
1 2 ∂ 1 ∂
Γ +Γ =2 + 1− (76.70)
∂ξ α ∂η
the invariant surface condition gives
1
2Fξ + 1 − Fη = 0 (76.71)
α
We thus consider the similarity variable θ and the function H(θ) defined by
(1 − α)ξ + 2αη
θ= , F (ξ, η) = H(θ) (76.72)
2α
We then find that H(θ) satisfies the equation
1 + 5α2 + 4αβ − 24α2 H 2 00
H (iv) − H + 12(H 0 )2 H = 0 (76.73)
4α2
Using the identity
(H 3 )00 = 3H 2 H 00 + 6H(H 0 )2
and integrating twice equation (76.70), we get the equation

1 + 5α2 + 4αβ
H 00 + 2H 3 − H + C1 θ + C2 = 0 (76.74)
4α2
The above equation in the special case of C1 = C2 = 0, written as
1 + 5α2 + 4αβ
H 00 + 2H 3 − H=0 (76.75)
4α2
admits a solution expressed in terms of Jacobi’s sn() function.
E. Solution of equation (76.21). Since equation (76.21) admits no sym-
metry generator, we can consider translation in η , i.e.
F (ξ, η) = H(θ), θ=ξ (76.76)
and thus equation (76.21) takes on the form
H 00 (ξ) = 0 (76.77)
1137
We then get
H(θ) = C1 θ + C2 (76.78)
Going back to the original variables, we obtain the solution

y 1
u(t, x, y) = C1 + C2 √ (76.79)
3t − α 3
3t − α
Reference
M.S. Hashemi, S. Abbasbandy, M.S. Alhuthali and H.H. Alsulami:
”Conservation Laws and Symmetries of mKdV-KP Equation”.
Rom. J. Phys. 60 (2015) 904-917.
77 The Zabolotskaya-Khokhlov Equation

I. The Zabolotskaya-Khokhlov Equation. This equation has the form
uxt − u2x − uuxx − uyy = 0, u = u(x, y, t) (77.1)

(77.1) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
> ZaboKhokh := diff(u, x, t)−(diff(u, x))2 −u∗diff(u, x, x)−diff(u, y, y) = 0;

∂2 ∂ 2 ∂2
U (x, y, t) − U (x, y, t) − U (x, y, t) U (x, y, t)
∂x∂t ∂x ∂x2
2
∂
− 2 U (x, y, t) = 0
∂y
1138
Step 4. We find the determining equations using
>DetEqns:=DeterminingPDE(ZaboKhokh); The program responds
∂
{ − ξ1 (x, y, t, U ) = 0,
∂U
∂ ∂ ∂
− ξ1 (x, y, t, U ) = − − ξ3 (x, y, t, U ) + 2 − ξ2 (x, y, t, U ) ,
∂x ∂t ∂y
∂ 1∂ ∂
− ξ1 (x, y, t, U ) = − ξ2 (x, y, t, U ), − ξ2 (x, y, t, U ) = 0,
∂y 2 ∂t ∂U
∂ ∂2
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂x ∂y 2
∂ ∂ ∂
∂U ∂x ∂y
∂2 3 ∂2
ξ
− 3 (x, y, t, U ) = − ξ2 (x, y, t, U ),
∂t2 2 ∂y∂t
∂ ∂
− η1 (x, t, U ) = −2U − ξ3 (x, y, t, U ) + 2 − ξ2 (x, y, t, U ) U
∂t ∂y
∂
− − ξ1 (x, y, t, U ) }
∂t
Solving the above system, we can determine the infinitesimals. The above
system can be solved using the command >pdsolve. In fact using
>Inf1:=pdsolve(DetEqns); the program responds
1 d 1 d2
2
{ − ξ1 (x, y, t, U ) = − F 1(t) + 6 − C1 x + 2 − F 1(t) y
3 dt 6 dt
1 d
+ − F 3(t) y +− F 4(t),
2 dt
2 d
− ξ2 (x, y, t, U ) = − F 1(t) + − C1 y +− F 3(t),
3 dt
− ξ3 (x, y, t, U ) =− F 1(t),
2 d 1 d2
− η1 (x, y, t, U ) = − U − F 1(t) + 2U − C1 − − F 1(t)
3 dt 3 dt2
3 2
1 d

2 1 d

− − F 1(t) y − − F 3(t) y
6 dt3 2 dt2
d
− − F 4(t) }
dt
1139
We could find the infinitesimals using the single command
Inf2:=Infinitesimals(ZaboKhokh);
The generator Γ of point symmetries is then given by
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ 3 + η1
∂x ∂y ∂t ∂u
1 d2

1 d
2 1 d
= F 1(t) + 6C1 x + F 1(t) y + F 3(t) y
3 dt 6 dt2 2 dt

∂ 2d ∂ ∂
+ F 4(t) + F 1(t) + C1 y + F 3(t) + F 1(t)
∂x 3 dt ∂y ∂t
2 3

2 d 1 d
1 d

+ − u F 1(t) + 2uC1 − 2
F 1(t) − 3
F 1(t) y 2
3 dt 3 dt 6 dt
2

1 d
d ∂
− 2
F 3(t) y − F 4(t)
2 dt dt ∂u
II. The Zabolotskaya-Khokhlov Equation with a Dissipative Term.
This equation has the form
uxt + u2x + uuxx + λuxxx + µuyy = 0, u = u(x, y, t) (77.2)
(77.2) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
> ZaboKhokhDiss := diff(u, x, t) + (diff(u, x))2 + u ∗ diff(u, x, x)
+ λ ∗ diff(u, x, x, x) + µ ∗ diff(u, y, y) = 0;
∂2 ∂ 2 ∂2
U (x, y, t) + U (x, y, t) + U (x, y, t) U (x, y, t)
∂x∂t ∂x ∂x2
∂3 ∂2
+λ U (x, y, t) + µ U (x, y, t) = 0
∂x3 ∂y 2
1140
>DetEqns:=DeterminingPDE(ZaboKhokhDiss); The program responds
∂ ∂ 2 ∂
{ − ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = − ξ2 (x, y, t, U ),
∂U ∂x 3 ∂y
∂
∂ 1 ∂t − ξ2 (x, y, t, U ) ∂
− ξ1 (x, y, t, U ) = − , − ξ2 (x, y, t, U ) = 0,
∂y 2 µ ∂U
∂ ∂2
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂x ∂y 2
∂ ∂ 4 ∂
− ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = − ξ2 (x, y, t, U )
∂U ∂t 3 ∂y
∂ ∂ ∂2
ξ
− 3 (x, y, t, U ) = 0, ξ
− 3 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂x ∂y ∂y∂t
2 ∂ ∂
η
− 1 (x, t, U ) = − ξ
− 2 (x, y, t, U ) U+ − ξ1 (x, y, t, U )}
3 ∂y ∂t
d
2 1 dt − F 2(t) y
{ − ξ1 (x, y, t, U ) = x − C1 − +− F 3(t),
3 2 µ
− ξ2 (x, y, t, U ) = − C1y +− F 2(t),
4
− ξ3 (x, y, t, U ) = − C1t +− C2,
3
d2
2 1 − F 2(t) y d
= − U − C1 − dt2
− η1 (x, y, t, U ) + − F 3(t)}
3 2 µ dt
1141
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ 3 + η1
∂x ∂y ∂t ∂u
d
1 dt F 2(t) y

2 ∂
= C1x − + F 3(t)
3 2 µ ∂x

∂ 4 ∂
+ [C1 y + F 2(t)] + C1 t + C2
∂y 3 ∂t
d2
1 dt2 F 2(t) y

2 d ∂
+ − C1 u − + F 3(t)
3 2 µ dt ∂u
3
After multiplying all the coefficients ξ 1 , ξ 2 , ξ 3 , η 1 by − µ and renaming
4
1 3
α = − µ C1 , β = − µ C2 ,
2 4 (77.3)
3 3
Q(t) = F2 (t), R(t) = − F3 (t)
8 4
and making the substitutions ξ 1 →X, ξ 2 →Y, ξ 3 →T, η 1 →U , we get
X = αx + Q0 (t)y + µR(t)
3
Y = αy − 2µQ(t)
2 (77.4)
T = 2αt + β
U = −αu + Q00 (t)y + µR0 (t)
This rescaling and renaming was necessary in order to make contact with the
notation introduced by Tajiri (Ref. [1]).
The similarity solutions of the Z-K equation will be found by considering the
system
dx dy dt du
= = = (77.5)
X Y T U
We shall only consider the case α = β = 0. For the other two cases the
reader is advised to look at Ref. [1]. In this case, considering the system of
equations
dt dx dx dy dy du
= , = , = (77.6)
T X X Y Y U
1142
we obtain the similarity variables
Q00 (t)y 2 + 2R0 (t)y
u=− + H(ρ, t)
4µQ(t)
(77.7)
Q0 (t)y 2 + 2R(t)y
ρ = t, η = x +
4µQ(t)
Upon substituting the above to the equation (77.2), we obtain the equation
R2 (t) 2 Q0 (t) Q00 (t)
λHηηη + HHηη + Hηη + Hη + Hη + Hηρ − =0
4µQ2 (t) 2Q(t) 2Q(t)
Integrating the above equation with respect to η, we get the equation
R2 (t) Q0 (t) Q00 (t)
λHηη + HHη + Hη + H + Hρ − η = S(ρ) (77.8)
4µQ2 (t) 2Q(t) 2Q(t)
where S(ρ) is an arbitrary function of ρ.
Considering the transformation
η
τ 0 = ρ, ξ 0 = √
Q
(77.9)
1 Q0 R2
H = √ G(ρ, η) + √ η − √
Q 2 Q 4µQ Q
1
Q(τ 0 )Gτ 0 + Q0 (τ 0 )G + GGξ0 + λ Gξ0 ξ0 = 0 (77.10)
2
where we have imposed the boundary condition G(τ 0 , ξ 0 = ∞) = 0.
Furthermore, the choice Q(t) = 1, τ = −τ 0 /λ, ξ = −ξ 0 /λ converts (77.10)
into Burgers equation
Gτ + GGξ − Gξξ = 0 (77.11)
We have thus converted the Z-K equation into the Burgers equation under
the similarity transformations
t x + R(t)y/2µ
τ =− , ξ=−
λ λ
0 2 (77.12)
R (t)y + R (t)/2
u=− + G(τ, ξ)
2µ
References
[1] M. Tajiri: ”Similarity Reductions of the Zabolotskaya-Khokhlov Equa-
tion with a Dissipative Term”. Nonlin. Math. Phys. 2 (1995) 392-397
1143
78 The Kuramoto-Sivashinsky Equation
I. The Kuramoto-Sivashinsky Equation. This equation has the form
ut + αuux + βuxx + γuxxxx = 0 (78.1)

(78.1) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> KSEqn := diff(u, t) + α ∗ u ∗ diff(u, x) + β ∗ diff(u, x, x)

+ γ ∗ diff(u, x, x, x, x) = 0;
∂ ∂ ∂2
U (x, t) + α U (x, t) U (x, t) + β U (x, t)
∂t ∂x ∂x2
∂4
+γ U (x, t) =0
∂x4
>DetEqns:=DeterminingPDE(KSEqn); The program responds
∂ ∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂x
∂2 ∂
− ξ1 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂t2 ∂U
∂ ∂
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0
∂t ∂x
∂
− ξ1 (x, t, U )
∂t }
− η1 (x, t, U ) =
α
1144
− C1
{ − ξ1 (x, t, U ) =− C1 t +− C2, − ξ2 (x, t, U ) =− C3, − η1 (x, t, U ) = }
α
Inf2:=Infinitesimals(KSEqn);
∂ ∂ ∂
Γ = ξ1 + ξ 2 + η1
∂x ∂t ∂u
∂ ∂ C1 ∂
= (C1 t + C2) + C3 +
∂x ∂t α ∂u
∂ 1 ∂ ∂ ∂
= C1 t + + C2 + C3
∂x α ∂u ∂x ∂t
We thus have the following symmetry generators
∂ 1 ∂ ∂ ∂
X1 = t + , X2 = , X3 = (78.2)
∂x α ∂u ∂x ∂t
II. The Generalized Kuramoto-Sivashinsky Equation. This equation
has the form
ut + uux + αuxx + βuxxx + γuxxxx = 0 (78.3)

(78.3) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> KSEqnG := diff(u, t) + u ∗ diff(u, x) + α ∗ diff(u, x, x)

+ β ∗ diff(u, x, x, x) + γ ∗ diff(u, x, x, x, x) = 0;
1145
∂ ∂ ∂2
U (x, t) + U (x, t) U (x, t) + α U (x, t)
∂t ∂x ∂x2
∂3 ∂4
+β U (x, t) + γ U (x, t) = 0
∂x3 ∂x4
>DetEqnsG:=DeterminingPDE(KSEqnG); The program responds
∂ ∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂x
∂2 ∂
2 − ξ1 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂t ∂U
∂ ∂
− ξ2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0
∂t ∂x
∂
− η1 (x, t, U ) = − ξ1 (x, t, U )}
∂t
>Inf1G:=pdsolve(DetEqnsG); the program responds
{ − ξ1 (x, t, U ) =− C1 t +− C2, − ξ2 (x, t, U ) =− C3, − η1 (x, t, U ) =− C1}

Inf2G:=Infinitesimals(KSEqnG);
∂ ∂ ∂
Γ = ξ1 + ξ 2 + η1
∂x ∂t ∂u
∂ ∂ ∂
= (C1 t + C2) + C3 + C1
∂x ∂t ∂u
∂ ∂ ∂ ∂
= C1 t + + C2 + C3
∂x ∂u ∂x ∂t
Therefore the Lie algebra of the symmetry generators is spanned by the
generators
∂ ∂ ∂ ∂
X1 = t + , X2 = , X3 = (78.4)
∂x ∂u ∂x ∂t
1146
Considering the generator X 3 + νX 2 , i.e.
∂ ∂
X 3 + νX 2 = +ν (78.5)
∂t ∂x
we can introduce the similarity variables ξ = x − νt and u = F (ξ). Equation
γF (iv) (ξ) + βF 000 (ξ) + αF 00 (ξ) − νF 0 (ξ) + F (ξ)F 0 (ξ) = 0
The above equation can be integrated once giving

1
γF 000 (ξ) + βF 00 (ξ) + αF 0 (ξ) − νF (ξ) + F 2 (ξ) = K (78.6)
2
where K is an arbitrary constant.
We shall look for solutions of the form
p
X
F = an Y n (ξ) (78.7)
n=0
where the function Y (ξ) satisfies either the Bernoulli
Y 0 (ξ) = A Y (ξ) + B Y 2 (ξ) (78.8)
or the Riccati equation
Y 0 (ξ) = A Y (ξ) + B Y 2 (ξ) + C (78.9)
We have the following solutions for the Bernoulli

cosh[A(ξ + L)] + sinh[A(ξ + L)]
Y (ξ) = A (78.10)
1 − Bcosh[A(ξ + L)] − Bsinh[A(ξ + L)]
and Riccati equations

B Θ 1
Y (ξ) = − − tanh Θ(ξ + L)
2A 2A 2

1
sech 2 Θξ (78.11)
B Θ 1
Y (ξ) = − − tanh Θξ +
2A 2A 2 1 2A 1
L tanh 2 Θξ − sinh 2 Θξ
Θ
1147
respectively, where Θ2 = B 2 − 4AC and L is an arbitrary constant.
After balancing F 000 and F 2 in (78.5) we find that p = 3. Therefore we have
the expansion
F = a0 + a1 Y (ξ) + a2 Y 2 (ξ) + a3 Y 3 (ξ) (78.12)
Upon substitution of the above expansion in (78.5), make use of Bernoulli

equation and equate to zero the coefficients of the various powers of Y , we
can determine the various coefficients:
α = A2 γ, β = 4Aγ, a0 = ν − 6A3 γ, a1 = −120A2 Bγ,
(78.13)
a2 = −240AB 2 γ, a3 = −120B 3 γ
Therefore equation (78.5) admits a solution of the form (78.11) where Y (ξ) is
given by (78.9) and the various coefficients are given by (78.12). Finally we
substitute the expansion (78.11) into (78.5), make use of Riccati equation and
equate to zero the coefficients of the various powers of Y , we can determine
the various coefficients:
A = 1, B = 3, C = 1, α = 365γ, β = −36γ − 4γk,
a0 = −990γ + 60γk + ν, a1 = 60γ + 180γk, a2 = 60γk, (78.14)
a3 = −120γ
where k is any solution of the equation k 2 + 18k + 1 = 0.

Therefore equation (78.5) admits a solution of the form (78.11) where Y (ξ)
are given by (78.10) and the various coefficients are given in (78.13).
III. The Generalized 2-dimensional Kuramoto-Sivashinsky Equa-
tion. This equation has the form
ut − αux + 4uux + uxx + β(uxx + uyy )x + (uxxxx + uyyyy ) = 0 (78.15)

(78.15) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
1148
> KSEqn2D := diff(u, t) − α ∗ diff(u, x) + 4 ∗ u ∗ diff(u, x)
+ diff(u, x, x) + β ∗ diff(diff(u, x, x) + diff(u, y, y), x)
+ (diff(u, x, x, x, x) + diff(u, y, y, y, y)) = 0;

∂ ∂ ∂
U (x, y, t) − α U (x, y, t) + 4U (x, y, t) U (x, y, t)
∂t ∂x ∂x
∂2 ∂3 ∂3
+ 2 U (x, y, t) + β U (x, y, t) + U (x, y, t)
∂x ∂x3 ∂y 2 ∂x
∂4 ∂4
+ 4 U (x, y, t) + 4 U (x, y, t) = 0
∂x ∂y
>DetEqns2D:=DeterminingPDE(KSEqn2D); The program responds
∂ ∂ ∂
{ − ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = 0, − ξ1 (x, y, t, U ) = 0
∂U ∂x ∂y
∂2 ∂ ∂
2 − ξ1 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂t ∂U ∂t
∂ ∂ ∂
∂x ∂y ∂U
∂ ∂ ∂
∂t ∂x ∂y
1∂
− η1 (x, y, t, U ) = − ξ1 (x, y, t, U )}
4 ∂t
>Inf12D:=pdsolve(DetEqns2D); the program responds
{ − ξ1 (x, y, t, U ) =− C1 t +− C2, − ξ2 (x, y, t, U ) =− C3,

1
− ξ3 (x, y, t, U ) =− C4, − η1 (x, t, U ) = − C1}
4
Inf22D:=Infinitesimals(KSEqn2D);
1149
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ 3 + η1
∂x ∂y ∂t ∂u
∂ ∂ ∂ 1 ∂
= (C1 t + C2) + C3 + C4 + C1
∂x ∂y ∂t 4 ∂u
∂ 1 ∂ ∂ ∂ ∂
= C1 t + + C2 + C3 + C4
∂x 4 ∂u ∂x ∂y ∂t
Therefore the generators of the Lie algebra of symmetries for the generalized
2−dimensional Kuramoto-Sivashinsky equation are given by
∂ 1 ∂ ∂ ∂ ∂
Γ1 = t + , Γ2 = , Γ3 = , Γ4 = (78.16)
∂x 4 ∂u ∂x ∂y ∂t
References
[1] C. M. Khalique: ”Exact Solutions of the Generalized Kuramoto-Sivashinsky
Equation”. Caspian J. Math. Sci. 1(2) (2012) 109-116
[2] M. Nadjafikhah and F. Ahangari: ”Lie symmetry analysis of the
two-dimensional generalized Kuramoto-Sivashinsky equation”.
Math. Sci. 6:3 (2012)
79 The Kadomtsev-Petviashvili Equation

79.1. The (2+1)-dimensional Kadomtsev-Petviashvili equation.
The Kadomtsev-Petviashvili equation is the two-dimensional version of the
KdV equation and usually appears in one of the two equivalent forms
3 1 3
ut + uux + uxxx + uyy = 0 (79.1)
2 4 x 4
(ut − 6uux + uxxx )x + 3uyy = 0 (79.2)
I. We are going to determine the symmetries of equation (79.1) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
1150
3 1
> KP Eqn := diff(diff(u, t) + ∗ u ∗ diff(u, x) + ∗ diff(u, x, x, x), x)
2 4
3
+ ∗ diff(u, y, y) = 0;
4
∂2 3 ∂ 2 3 ∂2
U (x, y, t) + U (x, y, t) + U (x, y, t) U (x, y, t)
∂x∂t 2 ∂x 2 ∂x2
4 2
1 ∂ 3 ∂
+ U (x, y, t) + U (x, y, t) = 0
4 ∂x4 4 ∂y 2
>DetEqns:=DeterminingPDE(KP Eqn); The program responds
∂ ∂ 1 ∂
∂U ∂x 2 ∂y
∂ 2∂ ∂
− ξ1 (x, y, t, U ) = − − ξ2 (x, y, t, U ), − ξ2 (x, y, t, U ) = 0,
∂y 3 ∂t ∂U
∂ ∂2
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂x ∂y 2
∂ ∂ 3 ∂
− ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = − ξ2 (x, y, t, U ),
∂U ∂t 2 ∂y
∂ ∂
− ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = 0,
∂x ∂y
2∂ ∂
− η1 (x, y, t, U ) = − ξ1 (x, y, t, U ) − U − ξ2 (x, y, t, U ) }
3 ∂t ∂y
1151
1 d 1 d2
2 2 d
{ − ξ1 (x, y, t, U ) = − F 1(t) x − − F 1(t) y − − F 2(t) y
2 dt 3 dt2 3 dt
d
+− F 3(t), − ξ2 (x, y, t, U ) = − F 1(t) y +− F 2(t),
dt
3 1 d2
− ξ3 (x, y, t, U ) = − F 1(t) +− C1, − η1 (x, t, U ) = − F 1(t) x
2 3 dt2
2 d3
2 4 d2 2d d
− − F 1(t) y − − F 2(t) y + − F 3(t) − U − F 1(t) }
9 dt3 9 dt2 3 dt dt
Inf2:=Infinitesimals(KP Eqn);
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ 3 + η1
∂x ∂y ∂t ∂u
2

1 d 1 d

2 2 d ∂
= F 1(t) x − 2
F 1(t) y − F 2(t) y + F 3(t)
2 dt 3 dt 3 dt ∂x

d ∂ 3 ∂
+ F 1(t) y + F 2(t) + F 1(t) + C1
dt ∂y 2 ∂t
2 3 2
1 d 2 d

2 4 d
2d
+ 2
F 1(t) x − 3
F 1(t) y − 2
F 2(t) y+ F 3(t)
3 dt 9 dt 9 dt 3 dt
d ∂
−u F 1(t)
dt ∂u
II. We are going to determine the symmetries of equation (79.2) using MAPLE.
>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, y, t);
> KP Eqn := diff(diff(u, t) − 6 ∗ u ∗ diff(u, x) + diff(u, x, x, x), x)

+ 3 ∗ diff(u, y, y) = 0;
1152
∂2 ∂ 2 ∂2
U (x, y, t) − 6 U (x, y, t) − 6 U (x, y, t) U (x, y, t)
∂x∂t ∂x ∂x2
4 2
∂ ∂
+ 4 U (x, y, t) + 3 2 U (x, y, t) = 0
∂x ∂y
>DetEqns:=DeterminingPDE(KP Eqn); The program responds
∂ ∂ 1 ∂
∂U ∂x 2 ∂y
∂ 1∂ ∂
− ξ1 (x, y, t, U ) = − − ξ2 (x, y, t, U ), − ξ2 (x, y, t, U ) = 0,
∂y 6 ∂t ∂U
∂ ∂2
− ξ2 (x, y, t, U ) = 0, − ξ2 (x, y, t, U ) = 0,
∂x ∂y 2
∂ ∂ 3 ∂
− ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = − ξ2 (x, y, t, U ),
∂U ∂t 2 ∂y
∂ ∂
− ξ3 (x, y, t, U ) = 0, − ξ3 (x, y, t, U ) = 0,
∂x ∂y
1∂ ∂
η
− 1 (x, y, t, U ) = − ξ
− 1 (x, y, t, U ) − U ξ
− 2 (x, y, t, U ) }
6 ∂t ∂y
1 d 1 d2
2 1 d
{ − ξ1 (x, y, t, U ) = − F 1(t) x − − F 1(t) y − − F 2(t) y
2 dt 12 dt2 6 dt
d
+− F 3(t), − ξ2 (x, y, t, U ) = − F 1(t) y +− F 2(t),
dt
3 1 d2
ξ
− 3 (x, y, t, U ) = − F 1(t) + − C1, η
− 1 (x, t, U ) = − − F 1(t) x
2 12 dt2
1 d3
2 1 d2 1d d
+ − F 1(t) y + − F 2(t) y − − F 3(t) − U − F 1(t) }
72 dt3 36 dt2 6 dt dt
Inf2:=Infinitesimals(KP Eqn);
1153
∂ ∂ ∂ ∂
Γ = ξ1 + ξ2 + ξ 3 + η1
∂x ∂y ∂t ∂u
2

1 d 1 d
2 1 d ∂
= F 1(t) x − 2
F 1(t) y − F 2(t) y + F 3(t)
2 dt 12 dt 6 dt ∂x

d ∂ 3 ∂
+ F 1(t) y + F 2(t) + F 1(t) + C1
dt ∂y 2 ∂t
2 3
1 d2

1 d
1 d
2

+ − F 1(t) x + F 1(t) y + F 2(t) y
12 dt2 72 dt3 36 dt2
1d d ∂
− F 3(t) − u F 1(t)
6 dt dt ∂u
80 The Rosenau Equation

The Rosenau equation has the form
ut + ux + uux + uxxxt = 0, u = u(x, t) (80.1)
We are going to determine the symmetries of equation (80.1) using MAPLE.

>restart;
>with(DEtools):
>with(PDEtools):
> u := U (x, t);
> RosEqn := diff(u, t) + diff(u, x) + u ∗ diff(u, x) + diff(u, x$3, t) = 0;
The program responds with

∂ ∂ ∂
U (x, t) + U (x, t) + U (x, t) U (x, t)
∂t ∂x ∂x
∂4
+ 3 U (x, t) = 0
∂x ∂t
1154
>DetEqns:=DeterminingPDE(RosEqn); The program responds
∂ ∂ ∂
{ − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0, − ξ1 (x, t, U ) = 0,
∂U ∂t ∂x
∂ ∂ ∂2
ξ
− 2 (x, t, U ) = 0, ξ
− 2 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0,
∂U ∂x ∂t2
∂
η
− 1 (x, t, U ) = − ξ
− 2 (x, t, U ) (1 + U )
∂t
{ − ξ1 (x, t, U ) =− C3, − ξ2 (x, t, U ) =− C1 t +− C2,

− η1 (x, t, U ) = −− C1 −− C1 U }

Inf2:=Infinitesimals(RosEqn);
∂ ∂ ∂
Γ = ξ1 + ξ 2 + η1
∂x ∂t ∂u
∂ ∂ ∂
= C3 + (C1 t + C2) − C1 (1 + u)
∂x ∂t ∂u
∂ ∂ ∂ ∂
= C1 t − (1 + u) + C2 + C 3
∂t ∂u ∂t ∂x
We thus have the following symmetry generators
∂ ∂ ∂ ∂
X1 = t − (1 + u) , X2 = , X3 = (80.2)
∂t ∂u ∂t ∂x
Chapter XXVI
Fractional Differential
Equations
1155
81 Lie Symmetries of Fractional
81.1. Preliminaries. We first introduce some preliminary notions about
Fractional Calculus necessary for the symmetry analysis.
The Riemann-Liouville derivative for a function f (t), denoted by Dtµ f (t),
is defined by
 n
d
 n f (t),
 µ=n
 dt

µ
Dt f (t) = (81.1)
n
d


I n−µ f (t), 0 ≤ n − 1 < µ < n


dtn
where n is a positive integer and I ν f (t) is the Rieman-Liouville fractional
integral of order ν defined by
 1 R t (t − s)ν−1 f (s)ds, ν > 0

I ν f (t) = Γ(ν) 0 (81.2)

f (t), ν=0

where Γ(ν) is the Gamma function.

The Rieman-Liouville time-fractional partial derivative of order µ for a
function u(t, x) is defined by
 n
∂
 n u(t, x),
 µ=n
Dtµ u(t, x) = ∂t 1 n R
∂ t
 (t − s)n−µ−1 u(s, x)ds, 0 ≤ n − 1 < µ < n
Γ(n − µ) ∂tn 0

(81.3)
The Erdélyi-Kober fractional differential operator is defined by
δ, α
D{β 1 , β2 , ··· , βm }
g
n−1 m
Y X 1 ∂ δ+α, n−α (81.4)
= δ+j+ − ξk K{β1 , β2 , ··· , βm } g (ξ1 , ξ2 , · · · , ξm )
j=0 k=1
βk ∂ξk
where
(
[α] + 1, α∈N
/
n= (81.5)
α, α∈N
1156
δ, α
and the generalized Erdélyi-Kober integral operator K{β 1 , β2 , ··· , βm }
g
is defined by

δ, α
K{β 1 , β2 , ··· , βm }
g (ξ1 , ξ2 , · · · , ξm )
 1 R
∞ α−1 −(δ+α)
 Γ(α) 1 (ν − 1) ν

 g(ξ1 ν 1/β1 , ξ2 ν 1/β2 , · · · , ξm ν 1/βm )dν, α > 0
=


g(ξ1 , ξ2 , · · · , ξm ), α = 0

(81.6)
The fractional Leibnitz rule for the product f (t)g(t) is defined by
∞
α
X α
Dt (f (t)g(t)) = Dtα−n f (t)Dtn g(t) (81.7)
n
n=0
where
(−1)n−1 αΓ(n − α)

α
= (81.8)
n Γ(1 − α)Γ(n + 1)
81.2. The Condition of Invariance. A second-order, time-fractional
partial differential equation of the form
∂tα u = F (x, t, u, ux , uxx ) (81.9)
is invariant under Lie point symmetries if
P r(α,2) X(∆)|∆=0 = 0, ∆ ≡ ∂tα u − F (x, t, u, ux , uxx ) (81.10)
where the second prolongation P r(α,2) X is defined by
∂ ∂ ∂
P r(α,2) X = X + φ0t α
+ φ[x] + φ[xx] (81.11)
∂(∂t u) ∂ux ∂uxx
with
∂ ∂ ∂
X = ξ1 + ξ2 +φ (81.12)
∂t ∂x ∂u
Equation (81.10) can be written explicitly as

0 1 ∂F 2 ∂F ∂F [x] ∂F [xx] ∂F
φt − ξ −ξ −φ −φ −φ =0 (81.13)
∂t ∂x ∂u ∂ux ∂uxx ∆=0
1157
The various quantities appearing in (81.11) are defined by
φ0t = Dtα (φ) + ξ 2 Dtα (ux ) − Dtα (ξ 2 ux ) + Dtα (Dt (ξ 1 )u)
(81.14)
− Dtα+1 (ξ 1 u) + ξ 1 Dtα+1 (u)
φ[x] = Dx (φ) − ux Dx (ξ 2 ) − ut Dx (ξ 1 ) (81.15)

[xx] [x] 1 2
φ = Dx (φ ) − uxt Dx (ξ ) − uxx Dx (ξ ) (81.16)
Taking into account the Leibnitz rule, the coefficient φ0t
takes the form
α ∞
1 ∂ u α
X
0
φt = −αDt (ξ ) α − Dtn (ξ 2 )Dtα−n (ux ) + Dtα (φ)
∂t n
n=1
∞ (81.17)
X α
− Dtn+1 (ξ 1 )Dtα−n (u)
n+1
n=1
Using the chain rule, we have

m k
dm f (g(t)) X X k 1 r d
m k
k−r d f (g(t))
= (−g(t)) (g(t) ) (81.18)
dtm r k! dtm dtk
k=0 r=0
and after setting f (t) = 1, we obtain

∞
∂ αφ ∂ αu ∂ α φu X α ∂ n φu α−n
Dtα (φ) = α + φu α − u α + D (u) + µ (81.19)
∂t ∂t ∂t n ∂tn t
n=1
where
∞ X n X m Xk−1
X α n k 1 tn−α (−u)r
µ=
n m r k! Γ(n + 1 − α)
n=2 m=2 k=2 r=0 (81.20)
m k−r n−m+k
∂ (u ) ∂ φ
× m n−m
∂t ∂t ∂uk
Taking into account (81.19), we get the following final expression for φ0t
∂ αφ 1 ∂ u
α
∂ α φu
φ0t = + (φ u − αD t (ξ )) − u +µ
∂tα ∂tα ∂tα
∞
α ∂ α φu

X α
+ − n+1 1
Dt (ξ ) Dtα−n (u) (81.21)
n ∂tα n+1
n=1
∞
X α
− Dtn (ξ 2 )Dtα−n (ux )
n
n=1
1158
81.3. Example. The Burgers Equation. The shall determine the Lie
point symmetries of the time-fractional Burgers equation
∂tα u = uxx + uux (81.22)
where ∂tα u is the Riemann-Liouville derivative of order α:
∂tα u ≡ Dtα u, 0<α<1 (81.23)
Invariance of (81.22) results in the equation

0 [x] [xx]
φt − φux − uφ − φ = 0, ∆≡∂tα u − uxx − uux (81.24)
∆=0
Expanding (81.24) and equating the coefficients of the various partial deriva-
tives of the function u, we derive the system
ξu2 = ξt2 = ξu1 = ξx1 = φuu = 0

α n α
∂ (φ ) − Dtn+1 (ξ 1 ) = 0, n = 1, 2, · · ·
n t u n+1
2 (81.25)
ξxx − αξt1 − 2φxu + uξx2 − φ = 0
2ξx2 − αξt1 = 0
∂tα (φ) − u∂tα (φu ) − φxx − φx = 0

2a1
ξ1 = t, ξ 2 = a1 x + a2 , φ = −a1 u (81.26)
α
We thus have evaluated the generator X of the Lie point symmetries (see
(81.12))
2a1 ∂ ∂ ∂
X= t + (a1 x + a2 ) − a1 u (81.27)
α ∂t ∂x ∂u
Therefore the Lie algebra of the symmetries of Burgers equation is being
spanned by the generators
∂ ∂ 2t ∂ ∂
X1 = , X2 = x + −u (81.28)
∂x ∂x α ∂t ∂u
1159
Considering the generator X2 , the corresponding invariant surface condition
yields
dx αdt du
= =− (81.29)
x 2t u
From the above system, we obtain the invariants
u = t−α/2 f (ζ), ζ = xt−α/2 (81.30)
Under the above substitution, Burgers equation (81.22) takes on the form

1−3α/2, α
d2 f df
D2/α f (ζ) = 2 + f (81.31)
dζ dζ
where Dβδ, α is the Erdélyi-Kober fractional differential operator

Dβδ, α f (ζ)
n−1 (81.32)
Y 1 d δ+α, n−α
= δ+j− ζ Kβ f (ζ)
j=0
β dζ
with
(
[α] + 1, α∈N
/
n= (81.33)
α, α∈N
and Kβδ, α is the Erdélyi-Kober fractional integral operator

Kβδ, α f (ζ)
 1 R
∞ α−1 −(δ+α)
 Γ(α) 1 (ν − 1) ν

 f (ζν 1/β )dν, α > 0 (81.34)
=


f (ζ), α = 0

Operators (81.32) and (81.34) are special cases of the corresponding defini-
tions (81.4) and (81.6).
Equation (81.31) is an ordinary fractional differential equation which does
not admit any explicit solution. However the purpose of this section was
1160
to show that Fractional Calculus may be of some use in revealing the Lie
symmetries of Differential equations, although the reduced equations might
not admit any explicit solutions.
References
[1] I. Podlubny: ”Fractional Differential Equations”. AP, NY 1999
[2] D. Baleanu, K. Diethelm, E. Scalas and J. Trujillo: ”Fractional
Calculus Models and Numerical Methods”. World Scientific 2012
[3] R.K. Gazizov, A.A. Kasatkin and S.Yu. Lukashchuk: ”Symmetry
Properties of fractional diffusion equations”. Phys. Scr. T136 (2009) 014016
[4] R.K. Gazizov and A.A. Kasatkin: ”Construction of exact solutions
for fractional order differential equations by the invariant subspace method”.
Comp. Math. Appl. 66 (2013) 576-584
[5] E. Buckwar and Yu. Luchko: ”Invariance of a Partial Differential
Equation of Fractional Order and the Lie Group of Scaling Transformations”.
J. Math. Anal. Appl. 227 (1998) 81-97
[6] R. Sahadevan and T. Bakkyaraj: ”Invariant analysis of time frac-
tional generalized Burgers and Korteweg-de Vries equations”.
J. Math. Anal. Appl. 393 (2012) 341-347
[7] M.S. Hashemi and D. Baleanu: ”On the Time Fractional Generalized
Fisher Equation”. Comm. Theor. Phys. 65 (2016) 11-16
Chapter XXVII
Special Solution Methods
of Nonlinear PDEs
82 The Tanh-Coth Method
Let
E(U, U 0 , U 00 , · · · ) = 0 (82.1)
be a nonlinear ODE containing the unknown function U depending on a single

variable x and the derivatives U 0 , U 00 , · · · with respect to x. We suppose that
1161
the function U can be expanded in a series
n
X n
X
U= k
ak Y + bk Y −k (82.2)
k=0 k=1
where the expansion variable Y is the hyperbolic tangent
Y = tanh(x) (82.3)
and ak , bk are constant coefficients. The derivative of Y is given by

1
Y0 = = 1 − tanh2 (x) = 1 − Y 2 (82.4)
cosh2 (x)
Therefore after taking the derivative of U given by (82.2), we substitute Y 0

by the quantity 1 − Y 2 . We thus see that the highest expansion term of
the first derivative will contain Y n−1 ·Y 0 = Y n−1 ·(1 − Y 2 )∼Y n+1 , i.e. after
differentiation the highest exponent will be n + 1 for the first derivative, n + 2
for the second derivative, n + k for the k−th order derivative. The m−th
power of U will contain the n·m as the highest exponent.
The maximal value of n is determined by balancing the highest order deriva-
tive to the highest nonlinear term of the equation. That will then fix the
value of n. Upon substitution the expansion (82.2) into the original equa-
tion, taking into account Y 0 = 1 − Y 2 at each stage of differentiation, we
obtain a polynomial type of equation with respect to Y . Equating the coeffi-
cients of the different powers of Y to zero, we get an overdetermined system
of equations from which we can determine the coefficients ak and bk .
The reader might find interesting the following relations which connect the
derivatives
dU dU
= (1 − Y 2 )
dx dY
d2 U 2 dU
2
2 2d U
= −2Y (1 − Y ) + (1 − Y )
dx2 dY dY 2
d3 U 2 2 dU 2
2 2d U
= 2(1 − Y )(3Y − 1) − 6Y (1 − Y )
dx3 dY dY 2
3
dU
+ (1 − Y 2 )3 3
dY
1162
Example 1. Considering the equation
k 2 F 00 + kωF 0 + F − F 2 = 0 (82.5)
and balancing the second derivative F 00 to the power F 2 , i.e. n + 2 = 2n, we

get n = 2. We thus have to consider the expansion
b1 b2
F = a0 + a1 Y + a2 Y 2 + + 2 (82.6)
Y Y
Further details are given in Section 55.
Example 2. Considering the equation
k 2 F 00 + kωF 0 + F (1 − F )(F − a) = 0 (82.7)
and balancing the second derivative F 00 to the power F 3 , i.e. n + 2 = 3n, we

get n = 1. We thus have to consider the expansion
b1
F = a0 + a1 Y + (82.8)
Y
Further details are given in Section 56.
Example 3. Let us now consider the cubic modified Boussinesq equation
2
utt + uxxt + uxxxx − (u3 )xx = 0, u = u(x, t) (82.9)
9
ξ = k(x − ωt) (82.10)
we shall find the traveling wave solutions of the given equation using the
Tanh method. Using this change of variables and taking into account that
d2 U d3 U
utt = (−kω)2 , uxxt = k 2
(−kω) ,
dξ 2 dξ 3
(82.11)
d4 U d2
uxxxx = k 4 4 , (u3 )xx = k 2 2 (U 3 )
dξ dξ
d2 U d3 U 2 4 d4 U 2 d
2
(−kω)2 + k 2
(−kω) + k − k (U 3 ) = 0 (82.12)
dξ 2 dξ 3 9 dξ 4 dξ 2
1163
Integrating twice the above equation and taking the integration constants
equal to zero, we obtain the equation
2 2 d2 U dU
k 2
− kω + ω2U − U 3 = 0 (82.13)
9 dξ dξ
Balancing the second derivative to U 3 we find n + 2 = 3n and thus n = 1,

i.e.
U = a0 + a1 Y with Y 0 (ξ) = 1 − Y 2 (82.14)
We find
U 0 = a1 Y 0 = a1 (1 − Y 2 ) (82.15)
and
U 00 = −2a1 Y Y 0 = −2a1 Y (1 − Y 2 ) (82.16)

2 2
k {−2a1 Y (1 − Y 2 )} − kωa1 (1 − Y 2 )
9 (82.17)
+ ω 2 (a0 + a1 Y ) − (a0 + a1 Y )3 = 0
Upon expanding, the above equation becomes

4
a1 k − a1 Y 3 + a1 (kω − 3a0 a1 )Y 2
2 2
9 (82.18)
4 2
+ a1 ω − 3a0 − k Y + ω 2 a0 − kωa1 − a30 = 0
2 2
9
Equating to zero the coefficients of Y , we obtain the system (with a1 6=0)
4 2 4
k − a21 = 0, kω − 3a0 a1 = 0, ω 2 − 3a20 − k 2 = 0,
9 9 (82.19)
ω 2 a0 − kωa1 − a30 = 0
The above system admits the following solutions:

Solution I
1 1 3
a0 = ω, a1 = ω, k= ω (82.20)
2 2 4
1164
Solution II
1 1 3
a0 = ω, a1 = − ω, k=− ω (82.21)
2 2 4
Solution III
1 1 3
a0 = − ω, a1 = ω, k=− ω (82.22)
2 2 4
Solution IV
1 1 3
a0 = − ω, a1 = − ω, k= ω (82.23)
2 2 4
Considering the Solution I, since Y = tanh(ξ), we get from (82.14)
1 1 3
U = ω + ω·tanh(ξ), ξ = k(x − ωt) = ω(x − ωt) (82.24)
2 2 4
and thus
1 1 3
u(x, t) = ω + ω·tanh ω(x − ωt) (82.25)
2 2 4
We similarly find the solutions
1 1 3
u(x, t) = ω − ω·tanh − ω(x − ωt) (82.26)
2 2 4
1 1 3
u(x, t) = − ω + ω·tanh − ω(x − ωt) (82.27)
2 2 4
and
1 1 3
u(x, t) = − ω − ω·tanh ω(x − ωt) (82.28)
2 2 4
83 The Sech Method

Let
E(U, U 0 , U 00 , · · · ) = 0 (83.1)
1165
n
X n
X
U= k
ak Y + bk Y −k (83.2)
k=0 k=1
where the expansion variable Y is the hyperbolic secant
Y = sech(x) (83.3)
and ak , bk are constant coefficients. The derivative of Y is given by

p
Y 0 = −tanh(x)·sech(x) = − 1 − sech2 (x)·sech(x)
i.e.
√
Y0 =− 1−Y2Y
0
Therefore after taking
√ the derivative of U given by (83.2), we substitute Y
2
by the quantity − 1 − Y Y . We thus see that the highest expansion term
n−1 0 n−1
√
of the first derivative will contain Y ·Y = −Y · 1 − Y Y ∼ Y n+1 , i.e.
2
after differentiation the highest exponent will be n + 1 for the first derivative,
n + 2 for the second derivative, n + k for the k−th order derivative. The
m−th power of U will contain the n·m as the highest exponent.
The maximal value of n is determined by balancing the highest order deriva-
tive to the highest nonlinear term of the equation. That will then fix the
value of n. Upon substitution the expansion (83.2) into the original equa-
tion, taking into account Y 0 = 1 − Y 2 at each stage of differentiation, we
obtain a polynomial type of equation with respect to Y . Equating the coeffi-
cients of the different powers of Y to zero, we get an overdetermined system
of equations from which we can determine the coefficients ak and bk .
The reader might find interesting the following relations which connect the
derivatives
dU √ dU
= −Y 1 − Y 2
dx dY
d2 U 2 dU 2
2
2 d U
= Y (1 − 2Y ) + Y (1 − Y )
dx2 dY dY 2
1166
d3 U √ 2

2 2 dU 2 d U
= −Y 1 − Y (1 − 6Y ) + 3Y (1 − 2Y )
dx3 dY dY 2
3

dU
+ Y 2 (1 − Y 2 ) 3
dY
Example. Let us consider the cubic modified Boussinesq equation
utt − uxxxx − (u3 )xx = 0, u = u(x, t) (83.4)
u(x, t) = U (ξ), ξ = k(x − ωt) (83.5)
utt = (−kω)2 U 00 (ξ), uxxxx = k 4 U 0000 (ξ), (u3 )xx = k 2 (U 3 (ξ))00 (83.6)
(−kω)2 U 00 (ξ) − k 4 U 0000 (ξ) − k 2 (U 3 (ξ))00 = 0
k 2 U 0000 (ξ) − ω 2 U 00 (ξ) + (U 3 (ξ))00 = 0 (83.7)
k 2 U 000 (ξ) − ω 2 U 0 (ξ) + (U 3 (ξ))0 = 0 (83.8)
where we have set the integration constant equal to zero. A second integra-
tion yields
k 2 U 00 (ξ) − ω 2 U (ξ) + U 3 (ξ) = 0 (83.9)
where we have set again the integration constant equal to zero.

Balancing U 00 to U 3 , i.e. n + 2 = 3n, we get n = 1 and we thus consider the
expansion
√
U = a0 + a1 Y (ξ), Y 0 = − 1 − Y 2 Y (83.10)
The first derivative of U is given by

√
U 0 = a1 Y 0 = −a1 1 − Y 2 Y (83.11)
1167
We then calculate the second derivative
Y2 √
U 00 = a1 √ Y 0 − a1 1 − Y 2 Y 0
1−Y2
Y2 √ √ √ (83.12)
= a1 √ (− 1 − Y 2 Y ) − a1 1 − Y 2 (− 1 − Y 2 Y )
1−Y2
= a1 Y − 2a1 Y 3
Equation (83.9), because of (83.10) and (83.12) becomes
k 2 (a1 Y − 2a1 Y 3 ) − ω 2 (a0 + a1 Y ) + (a0 + a1 Y )3 = 0 (83.13)
Expanding and rearranging, the above equation yields
a1 (a21 − 2k 2 )Y 3 + 2a0 a21 Y 2 + a1 (k 2 − ω 2 + 3a20 )Y
(83.14)
+ a30 − a0 ω 2 = 0
Equating to zero the coefficients of Y , we obtain the system
a1 (a21 − 2k 2 ) = 0, 2a0 a21 = 0, a1 (k 2 − ω 2 + 3a20 ) = 0,
(83.15)
a30 − a0 ω 2 = 0
Since a1 6=0 (otherwise we get the trivial solution u = 0) the solution of the
above system is given by
√
a0 = 0, a1 = ± 2 k, ω = ±k (83.16)
We thus have the solution
√
U = ± 2 k·Y (ξ), ξ = k(x±kt) (83.17)
Since Y = sech(ξ), we obtain the following solutions of equation (83.4)
√ √
u(x, t) = ± 2 k·sech[k(x − kt)], u(x, t) = ± 2 k·sech[k(x + kt)] (83.18)
where k is a free parameter.
References
D. Baldwin, Ü. Göktas, W. Hereman, L. Hong, R. S. Martino,
J. C. Miller: ”Symbolic computation of exact solutions expressible in hy-
perbolic and elliptic functions for nonlinear PDEs”.
J. Symb. Comp. 37 (2004) 669-705
1168
84 The Auxiliary Equation Method I
84.1 The Bernoulli Equation Method
Let
E(U, U 0 , U 00 , · · · ) = 0 (84.1)

n
X n
X
U= k
ak Y (x) + bk Y −k (x) (84.2)
k=0 k=1
where the function Y (x) satisfies the Bernoulli equation
Y 0 (x) = A Y (x) + B Y 2 (x) (84.3)
while ak , bk and A, B are constants.

Balancing the highest order derivative to the highest order nonlinear term, we
determine the positive integer n, i.e. the highest order expansion term. After
each differentiation of (84.2) with respect to x, we substitute the derivative
Y 0 (x) by the expression given in (84.3). It is obvious that the highest of
the first derivative is n + 1, that of the second derivative n + 2, and the
k−th order derivative n + k. The m−th power will contain the m·n as the
highest exponent. Upon substitution of U and its derivatives U 0 , U 00 , · · · to
the original equation, we obtain a polynomial type of equation with respect
to Y . Equating to zero all the coefficients of Y , we obtain a system of
overdetermined equations. The solution of this system, provides the values
of ak , bk and A, B.
The solution of the Bernoulli equation is given by

cosh[A(x + m)] + sinh[A(x + m)]
Y (x) = A (84.4)
1 − Bcosh[A(x + m)] − Bsinh[A(x + m)]
where m is an arbitrary constant.
Example. We consider the equation
1
γF 000 (x) + βF 00 (x) + αF 0 (x) − νF (x) + F 2 (x) = k (84.5)
2
1169
where k is an arbitrary constant.
After balancing F 000 and F 2 , i.e. n + 3 = 2n, we find that n = 3. Therefore
we have the expansion
F = a0 + a1 Y (x) + a2 Y 2 (x) + a3 Y 3 (x) (84.6)
Upon substitution of the above expansion in (84.5), make use of Bernoulli

equation and equate to zero the coefficients of the various powers of Y , we
can determine the various coefficients:
α = A2 γ, β = 4Aγ, a0 = ν − 6A3 γ, a1 = −120A2 Bγ,
(84.7)
a2 = −240AB 2 γ, a3 = −120B 3 γ
Therefore equation (84.5) admits a solution of the form (84.6) where Y (x) is
given by (84.4) and the various coefficients are given by (84.7).
84.2 The Riccati Equation Method

In the Riccati equation method, we take again the expansion (84.3) with
derivative Y 0 given by
Y 0 (x) = A Y 2 (x) + B Y (x) + C (84.8)
The Riccati equation (84.8) admits the general solutions given by
A ∆ 1
Y (x) = − − tanh ∆(x + m)
2B 2B 2
1
A ∆ 1 sech ∆x (84.9)
Y (x) = − − tanh ∆x + 2
1 2A 1
2B 2B 2
m tanh ∆x − sinh ∆x
2 ∆ 2
respectively, where ∆2 = A2 − 4BC and m is an arbitrary constant.
After balancing F 000 and F 2 in (84.5) we find that n = 3. Therefore we have
the expansion
F = a0 + a1 Y (x) + a2 Y 2 (x) + a3 Y 3 (x) (84.10)
Finally we substitute the expansion (84.6) into (84.5), make use of Riccati
equation (84.8) and equate to zero the coefficients of the various powers of
1170
Y , we can determine the various coefficients:
A = 1, B = 3, C = 1, α = 365γ, β = −36γ − 4γk,
a0 = −990γ + 60γk + ν, a1 = 60γ + 180γk, a2 = 60γk, (84.11)
a3 = −120γ
where k is any solution of the equation k 2 + 18k + 1 = 0.

Therefore equation (84.5) admits a solution of the form (84.10) where Y (x)
are given by (84.9) and the various coefficients are given in (84.11).
85 The Exp-function Method

Let
E(U, U 0 , U 00 , · · · ) = 0 (85.1)
the function U can be expanded as
ac ·exp(cx) + · · · + a−d ·exp(−dx)
U (x) = (85.2)
bp ·exp(px) + · · · + b−q ·exp(−qx)
The relation between c and p can be established by balancing the highest
order linear term with the highest order nonlinear term. Similarly to deter-
mine the relation between d and q we balance the lowest order linear term
with the lowest order nonlinear term. We have the following expressions for
the derivatives of the function U containing the c and p terms
C1 ·exp[(p + c)x] + · · ·
U 0 (x) =
C2 ·exp(2px) + · · ·
C3 ·exp[(3p + c)x] + · · ·
U 00 (x) =
C4 ·exp(4px) + · · ·
C5 ·exp[(7p + c)x] + · · ·
U 000 (x) =
C6 ·exp(8px) + · · ·
etc., and also the power
Ck ·exp(kcx) + · · ·
U k (x) =
Ck0 ·exp(kpx) + · · ·
1171
We find similarly the derivatives containing the d and q terms
· · · + D1 ·exp[−(q + d)x]
U 0 (x) =
· · · + D2 ·exp(−2qx)
· · · + D3 ·exp[−(3q + d)x]
U 00 (x) =
· · · + D4 ·exp(−4qx)
· · · + D5 ·exp[−(7q + d)x]
U 000 (x) =
· · · + D6 ·exp(−8qx)
etc., and also the power
· · · + Dk ·exp(−kdx)
U k (x) =
· · · + Dk0 ·exp(−kqx)
Example. We consider the PDE
ut + 2αux − uxxt + βu2 ux = 0, u = u(x, t) (85.3)
u(x, t) = U (ξ), ξ = k(x − ωt) (85.4)
we shall find the traveling wave solutions using the exp-function method.
The given equation gets transformed into
ωk 2 U 000 (ξ) + βkU 2 (ξ)U 0 (ξ) + (2kα − ω)U 0 (ξ) = 0
Integrating once the above equation (and taking the integration constant
equal to zero), we get the equation
1
ωk 2 U 00 (ξ) + βkU 3 (ξ) + (2kα − ω)U (ξ) = 0 (85.5)
3
Balancing U 00 with U 3 for the c and p terms, i.e.
C1 ·exp[(3p + c)ξ] + · · ·
U 00 (ξ) =
C2 ·exp(4pξ) + · · ·
and
C3 ·exp(3cξ) + · · · C3 ·exp[(p + 3c)ξ] + · · ·
U 3 (ξ) = =
C4 ·exp(3pξ) + · · · C4 ·exp(4pξ) + · · ·
1172
we find 3p + c = p + 3c, i.e.
c=p (85.6)
Similarly balancing U 00 with U 3 for the d and q terms, i.e.
· · · + D1 ·exp[−(3q + d)ξ]
U 00 (ξ) =
· · · + D2 ·exp(−4qξ)
and
· · · + D3 ·exp(−3dξ) · · · + D3 ·exp[−(3d + q)ξ]
U 3 (ξ) = =
· · · + D4 ·exp(−3qξ) · · · + D4 ·exp(−4qξ)
we find −(3q + d) = −(3d + q), i.e.
q=d (85.7)
We can consider any values of the pairs (c, p) and (d, q) like
(I) c = p = 1 and d = q = 1. In this case
a1 eξ + a0 + a−1 e−ξ
U= (85.8)
b1 eξ + b0 + b−1 e−ξ
(II) c = p = 2 and d = q = 1. In this case
a2 e2ξ + a1 eξ + a0 + a−1 e−ξ
U= (85.9)
b2 e2ξ + b1 eξ + b0 + b−1 e−ξ
(III) c = p = 1 and d = q = 2. In this case
a1 eξ + a0 + a−1 e−ξ + a−2 e−2ξ
U= (85.10)
b1 eξ + b0 + b−1 e−ξ + b−2 e−2ξ
(IV) c = p = 2 and d = q = 2. In this case
a2 e2ξ + a1 eξ + a0 + a−1 e−ξ + a−2 e−2ξ
U= (85.11)
b2 e2ξ + b1 eξ + b0 + b−1 e−ξ + b−2 e−2ξ
Considering the first choice, upon substituting (85.8) into (85.5) we get an
equation of the form
1 1
3 (b1 e + b0 + b−1 e−ξ )3
ξ
(85.12)
ξ 2ξ 3ξ −ξ −2ξ −3ξ
×{A0 + A1 e + A2 e + A3 e + B1 e + B2 e + B3 e } = 0
1173
where A0 , A1 , A2 , A3 and B1 , B2 , B3 are expressions depending on a1 , a0 , a−1
and b1 , b0 , b−1 . Equating to zero the coefficients of exp(±nξ), n = 0, 1, 2, 3
we obtain the system of seven equations
A0 = 0, A1 = 0, A2 = 0, A3 = 0, B1 = 0, B2 = 0, B3 = 0 (85.13)
We obtain ten solutions in total. Considering one of them, i.e.
a1 b 0 a1 b−1
a0 = , a1 = a1 , a−1 = , b0 = b0 , b1 = b1 , b−1 = b−1 ,
b1 b1
(85.14)
6αb21 + βa21
ω=
3b21
we get from (85.8)
a1 b0 a1 b−1 −ξ
a1 e ξ + + e
b1 b1
6αb21 + βa21
U= , ξ = k x − t (85.15)
b1 eξ + b0 + b−1 e−ξ 3b21
Taking into account (85.4), we find that the solution of (85.3) is given by
a1 b1 e2ξ + a1 b0 eξ + a1 b−1 6αb21 + βa21
u(x, t) = , ξ = k x − t (85.16)
b21 e2ξ + b0 b1 eξ + b1 b−1 3b21
In a similar fashion, we obtain all the other solutions.
86 The (G0/G) Method

Let us suppose that we have a differential equation of the form
F (u, ux , uxx , · · · ) = 0 (86.1)
where u is the unknown function depending on the variable x. We expand
the function u in series
Xn G0 k X n G0 −k
u= ak + bk (86.2)
k=0
G k=1
G
and then balance the highest order derivative term to the most nonlinear
term. The function G satisfies the second order ordinary differential equation
G00 + λG0 + µG = 0 (86.3)
1174
where λ and µ are constants.
The first derivative of the expansion (86.2) will contain a combination of
terms of the form
G00 G0 2 G0 k−1 G00 G0 2 G0 −k−1
− and − (86.4)
G G G G G G
Upon substitution of G00 /G by −λ(G0 /G) − µ according to (86.3), the first
derivative of the expansion (86.2) will contain a combination of terms
G0 G0 2 G0 k−1 G0 G0 2 G0 −k−1
−λ −µ− and −λ −µ− (86.5)
G G G G G G
We thus see that the highest order terms of the first derivative of the expan-
sion (86.2), taking into account (86.3), will be
G0 k+1
(86.6)
G
Therefore the derivative of the expansion (86.2) taking into account (86.3)
will increase the highest order of the expansion terms by one.
Balancing the highest order derivative term to the most nonlinear term we
determine n. The substitution of the given expansion (for fixed n) into the
given equation (86.1) and equating to zero the various coefficients of (G0 /G),
we can determine the coefficients ak , bk .
Equation (86.3) admits the solutions
1√ 1√
G0 1 √ Acosh ∆ x + Bsinh ∆x λ
= ∆ 21 √ 21 √ − (86.7)
G 2 2

Asinh ∆ x + Bcosh ∆x
2 2
when ∆≡λ2 − 4µ > 0,
1√ 1√
G0 1 √ −Asin −∆ x + Bcos −∆ x λ
= −∆ 2
1√ 2
1√ − (86.8)
G 2 2

Acos −∆ x + Bsin −∆ x
2 2
when ∆≡λ2 − 4µ < 0, and
G0 B λ
= − (86.9)
G A + Bx 2
1175
when ∆≡λ2 − 4µ = 0.
Example 1. Consider the regularized long wave (RLW) equation
ut + ux + α(u2 )x − βuxxt = 0 (86.10)
which describes ion sound waves in a plasma.
We shall determine the traveling wave solutions of the above equation. Upon
introducing the similarity variable
ξ = k(x − ωt) (86.11)
ut = −(ωk)U 0 (ξ), ux = kU 0 (ξ), (u2 )x = 2kU (ξ)U 0 (ξ)
(86.12)
uxxt = −(ωk)k 2 U 000 (ξ)
βωk 2 U 000 (ξ) + 2αU (ξ)U 0 (ξ) + (1 − ω)U 0 (ξ) = 0 (86.13)
Integrating the above equation once and taking the integration constant equal
to zero, we get the equation
βωk 2 U 00 (ξ) + αU 2 (ξ) + (1 − ω)U (ξ) = 0 (86.14)
Balancing the second derivative U 00 (ξ) to the square U 2 (ξ) we obtain n + 2 =
2n, i.e. n = 2. We will then have the expansion
G0 (ξ) G0 (ξ) 2 G0 (ξ) −1
U (ξ) = a0 + a1 + a2 + b1
G(ξ) G(ξ) G(ξ)
G0 (ξ) −2 (86.15)
+ b2
G(ξ)
We differentiate the above expression with respect to ξ,
00 0
0 G (ξ) G (ξ) 2
U (ξ) = a1 −
G(ξ) G(ξ)
G0 (ξ) G00 (ξ) G0 (ξ) 2
+ 2a2 −
G(ξ) G(ξ) G(ξ)
G0 (ξ) −2 G00 (ξ) G0 (ξ) 2
− b1 −
G(ξ) G(ξ) G(ξ)
G0 (ξ) −3 G00 (ξ) G0 (ξ) 2
− 2b2 −
G(ξ) G(ξ) G(ξ)
1176
and then we substitute G00 (ξ)/G(ξ) by −λ(G0 (ξ)/G(ξ)) − µ. We then get the
expression
G0 (ξ) G0 (ξ) −2 G0 (ξ) −3
0
U (ξ) = a1 + 2a2 − b1 − 2b2
G(ξ) G(ξ) G(ξ)
G0 (ξ) G0 (ξ) 2 (86.16)
× −λ −µ−
G(ξ) G(ξ)

G0 (ξ) 3 G0 (ξ) 2
0
U (ξ) = −2a2 − (a1 + 2λa2 )
G(ξ) G(ξ)
G0 (ξ) G0 (ξ) −1
− (λa1 + 2µa2 ) + (b1 λ + 2b2 ) (86.17)
G(ξ) G(ξ)
G0 (ξ) −2 G0 (ξ) −3
+ (2b2 λ + b1 µ) + 2µ + b 1 − a1 µ
G(ξ) G(ξ)
We differentiate with respect to ξ the above expression and then we substitute

again G00 (ξ)/G(ξ) by −λ(G0 (ξ)/G(ξ)) − µ. We then get the expression
G0 (ξ) 4 G0 (ξ) 3
00
U (ξ) = 6a2 + (2a1 + 10λa2 )
G(ξ) G(ξ)
G0 (ξ) 2
+ (8µa2 + 4λ2 a2 + 3λa1 )
G(ξ)
G0 (ξ)
+ (2a1 µ + λ2 a1 + 6λµa2 )
G(ξ)
G0 (ξ) −1
(86.18)
+ (2b1 µ + λ2 b1 + 6λb2 )
G(ξ)
G0 (ξ) −2
+ (8b2 µ + 4λ2 + 3λµb1 )
G(ξ)
G0 (ξ) −3 G0 (ξ) −4
+ (2b1 µ2 + 10λµb2 ) + 6µ2 b2
G(ξ) G(ξ)
2
+ 2a2 µ + 2b2 + b1 λ + a1 λµ
Substitution of (86.15), (86.17) and (86.18) into (86.14), yields the equation
B1 B2 B3 B4
A4 X 4 + A3 X 3 + A2 X 2 + A1 X + + 2 + 3 + 4 + A0 = 0 (86.19)
X X X X
1177
G0 (ξ)
where X≡ and A4 , A3 , A2 , A1 , B1 , B2 , B3 , B4 , A0 are given explicitly by
G(ξ)
A4 = αa22 + 6βωk 2 a2 (86.20)
A3 = 2αa1 a2 + 2βωk 2 a1 + 10βωk 2 a2 λ (86.21)

A2 = 8βωk 2 a2 µ + αa21 + 2αa0 a2 + 4βωk 2 a2 λ2 + a2
(86.22)
− ωa2 + 3βωk 2 a1 λ
A1 = βωk 2 a1 λ2 − ωa1 + 2αa2 b1 + 6βωk 2 a2 λµ + 2αa0 a1
(86.23)
+ 2βωk 2 a1 µ + a1
B1 = βωk 2 b1 λ2 − ωb1 + 2αa0 b1 + 6βωk 2 b2 λ + 2αa1 b2
(86.24)
+ 2βωk 2 b1 µ + b1
B2 = 8βωk 2 b2 µ + αb21 + 2αa0 b2 + 4βωk 2 b2 λ2 + b2
(86.25)
− ωb2 + 3βωk 2 b1 λµ
B3 = 2βωk 2 b1 µ2 + 10βωk 2 b2 λµ + 2αb1 b2 (86.26)
B4 = 6βωk 2 b2 µ2 + αb22 (86.27)
A0 = 2αa2 b2 − ωa0 + a0 + βωk 2 a1 λµ + 2βωk 2 a2 µ2
(86.28)
+ 2βωk 2 b2 + αa20 + βωk 2 b1 λ + 2αa1 b1
Since equation (86.19) should hold for every X, we would have the following
system of equations
A4 = 0, A3 = 0, A2 = 0, A1 = 0, A0 = 0,
(86.29)
B1 = 0, B2 = 0, B3 = 0, B4 = 0
Solving the above system, we obtain the following three solution sets:
Solution Set I
−1 + ω 6βωk 2
(1) a0 = , a1 = 0, a2 = − , b1 = 0
α α
(86.30)
3 (1 − ω)2 1 −1 + ω
b2 = − · , µ = · , λ = 0, ω = ω
128 βωk 2 α 16 βωk 2
3 −1 + ω 6βωk 2
(2) a0 = · , a1 = 0, a2 = − , b1 = 0
4 α α
(86.31)
3 (1 − ω)2 1 −1 + ω
b2 = − · , µ = − · , λ = 0, ω = ω
128 βωk 2 α 16 βωk 2
1178
3 1 − ω + βωk 2 λ2 6βωk 2 λ
(3) a0 = − · , a1 = − ,
2 α α
6βωk 2 1 1 − ω + βωk 2 λ2 (86.32)
a2 = − , b1 = 0, b2 = 0, µ = · ,
α 4 βωk 2
λ = λ, ω = ω
1 −1 + ω + 3βωk 2 λ2 6βωk 2 λ
(4) a0 = − · , a1 = − ,
2 α α
6βωk 2 1 −1 + ω + βωk 2 λ2 (86.33)
a2 = − , b1 = 0, b2 = 0, µ = · ,
α 4 βωk 2
λ = λ, ω = ω
Solution Set II
6βk 2 µ
(1) a0 = − , a1 = 0, a2 = 0,
(4βk 2 µ + 1 − βk 2 λ2 )α
6βk 2 λµ 6βk 2 µ2
b1 = − , b2 = − , (86.34)
(4βk 2 µ + 1 − βk 2 λ2 )α (4βk 2 µ + 1 − βk 2 λ2 )α
1
ω=
4βk µ + 1 − βk 2 λ2
2
βk 2 (λ2 + 2µ)
(2) a0 = , a1 = 0, a2 = 0,
(4βk 2 µ − 1 − βk 2 λ2 )α
6βk 2 λµ 6βk 2 µ2
b1 = , b2 = , (86.35)
(4βk 2 µ − 1 − βk 2 λ2 )α (4βk 2 µ − 1 − βk 2 λ2 )α
1
ω=−
4βk µ − 1 − βk 2 λ2
2
6βk 2 µ 6βk 2 λ
(3) a0 = − , a1 = − ,
(4βk 2 µ + 1 − βk 2 λ2 )α (4βk 2 µ + 1 − βk 2 λ2 )α
6βk 2
a2 = − , b1 = 0, b2 = 0, (86.36)
(4βk 2 µ + 1 − βk 2 λ2 )α
1
ω=
4βk µ + 1 − βk 2 λ2
2
1179
βk 2 (λ2 + 2µ) 6βk 2 λ
(4) a0 = , a1 = ,
(4βk 2 µ − 1 − βk 2 λ2 )α (4βk 2 µ − 1 − βk 2 λ2 )α
6βk 2
a2 = , b1 = 0, b2 = 0 (86.37)
(4βk 2 µ − 1 − βk 2 λ2 )α
1
ω=−
4βk µ − 1 − βk 2 λ2
2
Solution Set III
4βk 2 µ
(1) a0 = − , a1 = 0,
(16βk 2 µ − 1)α
6βk 2 6βk 2 µ2
a2 = , b1 = 0, b2 = , (86.38)
(16βk 2 µ − 1)α (16βk 2 µ − 1)α
1
λ = 0, ω = − 2
16βk µ − 1
6βk 2 µ 6βk 2 λ
(2) a0 = − , a1 = − ,
(4βk 2 µ + 1 − βk 2 λ2 )α (4βk 2 µ + 1 − βk 2 λ2 )α
6βk 2
a2 = − , b1 = 0, b2 = 0, λ = λ, (86.39)
(4βk 2 µ + 1 − βk 2 λ2 )α
1
ω=
4βk µ + 1 − βk 2 λ2
2
βk 2 (λ2 + 2µ) 6βk 2 λ

(3) a0 = , a1 = ,
(4βk 2 µ − 1 − βk 2 λ2 )α (4βk 2 µ − 1 − βk 2 λ2 )α
6βk 2
a2 = , b1 = 0, b2 = 0, λ = λ (86.40)
(4βk 2 µ − 1 − βk 2 λ2 )α
1
ω=−
4βk µ − 1 − βk 2 λ2
2
Equations (86.30)-(86.40) provide the required set of coefficients which have

to be substituted in the expansion (86.15). On the other hand, the value
of the ratio (G0 /G) is found taking into account (86.7)-(86.9). After these
calculations, we are now in a position to find the traveling wave solutions of
the RLW equation (86.10).
1180
87 The Sine-Cosine Method
Let us suppose that we have a differential equation of the form
F (u, ux , uxx , · · · ) = 0 (87.1)
where u is the unknown function depending on a single variable x. The
Sine-Cosine is another method which makes use one of the substitutions
u = λ cosm (µx), λ, m, µ 6= 0 (87.2)
or
u = λ sinm (µx), λ, m, µ 6= 0 (87.3)
Example. We consider the KdV equation
ut + auux + uxxx = 0 (87.4)
u(x, t) = U (ξ), ξ = k(x − ωt) (87.5)
ut = (−kω)U 0 (ξ), ux = kU 0 (ξ), uxxx = k 3 U 000 (ξ) (87.6)
the KdV equation takes on the form
k 2 U 000 (ξ) + aU (ξ)U 0 (ξ) − ωU 0 (ξ) = 0
1
k 2 U 00 (ξ) + aU 2 (ξ) − ωU (ξ) = 0 (87.7)
2
where we have set the integration constant equal to zero. Considering the
cosine substitution
U = λ cosm (µξ), λ, m, µ 6= 0 (87.8)
− λ(ω + k 2 µ2 m)cosm (µξ) + λk 2 µ2 m(m − 1)sin2 (µξ)cosm−2 (µξ)
1 (87.9)
+ aλ2 cos2m (µξ) = 0
2
1181
In the above equation we have to express everything in terms of cos(µξ),
making use of the identity sin2 (µξ) = 1 − cos2 (µξ). We thus obtain the
equation
− (ω + k 2 µ2 m)cosm (µξ) − k 2 µ2 m(m − 1)cosm (µξ)

1 (87.10)
+ k 2 µ2 m(m − 1)cosm−2 (µξ) + aλ cos2m (µξ) = 0
2
Equating the exponents of the cosine in the last two terms of the above
equation we get m − 2 = 2m, i.e.
m = −2 (87.11)
We then equate to zero the coefficients of the similar powers of the cosine
function. We have
− (ω + k 2 µ2 m) − k 2 µ2 m(m − 1) = 0
1
k 2 µ2 m(m − 1) + aλ = 0
2
For m = −2 the above system becomes
1 1
k 2 µ2 = − ω, k 2 µ2 = − aλ (87.12)
4 12
from which we get
3ω 1√
λ= and µ = ± −ω (87.13)
a 2k
Therefore we obtain the solution
√
3ω 1 3ω 2 −ω
U (ξ) = √ or U (ξ) = sec ± ξ (87.14)
a −ω a 2k
cos2 ± ξ
2k
The solution of the original KdV equation (87.4) is then given by
√
3ω 2 −ω
u(x, t) = sec ± (x − ωt) (87.15)
a 2
1182
88 The Projective
Riccati Equation Method
We consider an evolution equation of the general form
ut = G(u, ux , uxx , ...) or utt = G(u, ux , uxx , ...) (88.1)
where u is a smooth function. We introduce a new variable ξ given by
ξ = k(x − ωt) (88.2)
where k and ω are constants. Changing variables and introducing a new

function U (ξ) by u(x, t) = U (k(x − ωt)) ≡ U (ξ), since
ut = (−ωk)U 0 (ξ), ux = kU 0 (ξ), uxx = k 2 U 00 (ξ), . . . (88.3)
equations (88.1) are being converted to ordinary differential equations of the

form
dU dU 2 d2 U
(−ωk) = G(u, k ,k , . . .) (88.4)
dξ dξ dξ 2
or
2 2 dU dU 2 d2 U
ω k = G(u, k ,k , . . .) (88.5)
dξ dξ dξ 2
respectively.
Equations (88.4) or (88.5) can be solved considering expansions of the form
(Abdou [1])
N
X
U = a0 + f k−1 [ak f (ξ) + bk g(ξ)] (88.6)
k=0
where a0 , ak and bk are constant coefficients to be determined.

Balancing the highest order derivative term to the higher nonlinear term of
the equations (88.4) or (88.5), we can determine the maximum number N of
the expansion terms.
The functions f (ξ) and g(ξ) satisfy the system
f 0 (ξ) = pf (ξ)g(ξ) (88.7)
1183
g 0 (ξ) = q + pg 2 (ξ) − rf (ξ) (88.8)
2 1h r2 + δ 2 i
g (ξ) = − q − 2rf (ξ) + f (ξ) (88.9)
p q
The system of equations (88.7) and (88.8) admit five families of solutions,
depending on the values of the parameters δ, p, q and r.
(I) If δ = λ2 − µ2 and pq < 0 then
q
f (ξ) = √ √ (88.10)
r + λ·sinh −pqξ + µ·cosh −pqξ
√ √ √
−pq λ·cosh −pqξ + µ·sinh −pqξ
g(ξ) = · √ √ (88.11)
p r + λ·sinh −pqξ + µ·cosh −pqξ
with
1h r2 + λ2 − µ2 2 i
g 2 (ξ) = − q − 2rf (ξ) + f (ξ) (88.12)
p q
(II) If δ = −λ2 − µ2 and pq > 0 then
q
f (ξ) = √ √ (88.13)
r + λ·sin pqξ + µ·cos pqξ
√ √ √
pq λ·cos pqξ + µ·sin pqξ
g(ξ) = · √ √ (88.14)
p r + λ·sin pqξ + µ·cos pqξ
with
1h r 2 − λ2 − µ2 2 i
g 2 (ξ) = − q − 2rf (ξ) + f (ξ) (88.15)
p q
(III) If q = 0, then
1
f (ξ) = pr (88.16)
ξ 2 + σξ + ζ
2
1 prξ + σ
g(ξ) = − · pr (88.17)
p ξ 2 + σξ + ζ
2
with
2r h σ 2 2rζ i
g 2 (ξ) = f (ξ) + 2 − f 2 (ξ) (88.18)
p p p
1184
where σ and ζ are free parameters.
(IV) If p = ±1 and δ = −r2 then
q 2
f (ξ) = + ψ(ξ) (88.19)
6r pr
12ψ 0 (ξ)
g(ξ) = (88.20)
q + 12ψ 0 (ξ)
where ψ(ξ) satisfies the Weirstrass equation
q2 pq 3
(ψ 0 (ξ))2 = 4ψ 3 (ξ) − ψ(ξ) − (88.21)
12 216
with solution ψ(ξ) = ℘(ξ).
The relation between f and g is given by
2r p
g 2 (ξ) = f (ξ) − (88.22)
p q
r2
(V) If p = ±1 and δ = − then
25
5q 5pq 2
f (ξ) = + (88.23)
6r 72ψ(ξ)
qψ 0 (ξ)
g(ξ) = − (88.24)
[pq + 12ψ(ξ)]ψ(ξ)
with ψ(ξ) = ℘(ξ) and
1 24r2 2
g 2 (ξ) = − q − 2rf (ξ) + f (ξ) (88.25)
p 25q
Example. We consider the equation (Antoniou [2], Appendix A, subsection

A.IV)
3
AA00 − (A0 )2 + 2m2 A2 − 8s2 A4 = 0, A = A(ξ) (88.26)
2
This is a generic equation which has appeared extensively in applying the
Riccati equation method with variable expansion coefficients (see next section
1185
§89 and the References at the end of that section).
We shall consider the expansion
A(ξ) = a0 + a1 f (ξ) + b1 g(ξ) (88.27)
Upon substituting (88.27) into (88.26) and equating to zero the coefficients
of f i (i = 0, 1, 2, 3, 4), g and f i g (i = 1, 2, 3), we find eight solutions, where
to every solution correspond two or three sub-families of solutions.
Family I. This family corresponds to the Solution I
m ρb1 m2
a0 = ± , a1 = , b1 = b1 , p = 8s2 b1 , q=−
4s2 2m 2s2 b1
where ρ is any root of the equation ρ2 = δ + r2 .

From (88.27) and Solution I, we obtain
m ρb1
A(ξ) = ± 2
+ f (ξ) + b1 g(ξ) (88.28)
4s 2m
m2
with p = 8s2 b1 , q = − with ρ satisfying the equation ρ2 = δ + r2 .
2s2 b1
Sub-family Ia. Since pq = −4m2 < 0, we consider the choice δ = λ2 − µ2 .
The functions f (ξ) and g(ξ) in (88.28) are given by (88.10) and (88.11)
respectively:
√
m2 1 m2 2 ξU 0 (ξ)
f (ξ) = − 2 · and g(ξ) = − 2 ·
2s b1 U (ξ) 2s b1 U (ξ)
where
p p
U (ξ) = r + λ·sinh(2|m| ξ) + µ·cosh(2|m| ξ) (88.29)
We thus obtain
√
m 1 ρm − ξ·U 0 (ξ)
A(ξ) = ± 2 − 2 · (88.30)
4s 4s U (ξ)
where ρ is any root of the equation ρ2 = λ2 − µ2 + r2 .

Sub-family Ib. If p = ±1 and δ = −r2 , we have ρ = 0 and then a1 = 0. In
m
this case we have A(ξ) = ± 2 + b1 g(ξ) where g(ξ) is given by (88.20).
4s
1186
1
(Ib.a) For p = 1, we have b1 = and q = −4m2 . Therefore g(ξ) =
8s2
12℘0 (ξ)
and thus
12℘0 (ξ) − 4m2
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.31)
4s2 2s2 12℘0 (ξ) − 4m2
where ℘(ξ) is the Weirstrass function, satisfying the equation
2 3 4m4 8m6
(℘(ξ)) = 4℘ (ξ) − ℘(ξ) + (88.32)
3 27
1
(Ib.b)For p = −1, we have b1 = − 2 and q = 4m2 . Therefore g(ξ) =
8s
12℘0 (ξ)
and thus
12℘0 (ξ) + 4m2
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.33)
4s2 2s2 12℘0 (ξ) + 4m2
r2 24r2
Sub-family Ic. If p = ±1 and δ = − , ρ satisfies the equation ρ2 =
25 25
while f (ξ) and g(ξ) are given by (88.23) and (88.24) respectively.
ρ 1 1
(Ic.a) For p = 1, we have a1 = · 2 , b1 = 2 and q = −4m2 . Therefore
2m 8s 8s
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − + and g(ξ) =
3r 9℘(ξ) [12℘(ξ) − 4m2 ]℘(ξ)
We thus have the following expression for A(ξ):
m 5ρm m2 1 m2 ℘0 (ξ)
A(ξ) = ± 2 + − + · (88.34)
4s 24s2 3℘(ξ) r 2s2 ℘(ξ)[12℘(ξ) − 4m2 ]
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.35)
3 27
ρ 1 1
(Ic.b) For p = −1, we have a1 = − · 2 , b1 = − 2 and q = 4m2 .
2m 8s 8s
Therefore
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − and g(ξ) = −
3r 9℘(ξ) [12℘(ξ) + 4m2 ]℘(ξ)
1187
We thus have the following expression for A(ξ):
m 5ρm m2 1 m2 ℘0 (ξ)
A(ξ) = ± + − + · (88.36)
4s2 24s2 3℘(ξ) r 2s2 ℘(ξ)[12℘(ξ) + 4m2 ]
Note. If δ = −λ2 − µ2 and pq > 0 leads to m2 < 0. The case q = 0 cannot

be considered either, since it leads to m = 0 and then the coefficient of f (ξ)
in (88.28) becomes infinite.
Family II. This family corresponds to the Solution II
m ρb1 m2
a0 = ± , a1 = , b1 = b1 , p = −8s2 b1 , q=−
4s2 2m 2s2 b1
where ρ is any root of the equation ρ2 = δ + r2 .

From (88.27) and Solution II, we obtain
m ρb1
A(ξ) = ± 2
+ f (ξ) + b1 g(ξ) (88.37)
4s 2m
m2
with p = −8s2 b1 , q=− and ρ satisfies the equation ρ2 = δ + r2 .
2s2 b1
Sub-family IIa. Since pq = 4m2 > 0, we first consider the case δ = −λ2 −µ2 .
The functions f (ξ) and g(ξ) are given by (88.13) and (88.14) respectively:
m2 1 2|m| Z(ξ)
f (ξ) = − · and g(ξ) = − ·
2s2 b1 V (ξ) 4s2 b1 V (ξ)
where
p p
Z(ξ) = λ·cos(2|m| ξ) + µ·sin(2|m| ξ) (88.38)
p p
V (ξ) = r + λ·sin(2|m| ξ) + µ·cos(2|m| ξ) (88.39)
We thus obtain
m 1 ρm + |m|Z(ξ)
A(ξ) = ± 2
− 2· (88.40)
4s 4s V (ξ)
where ρ is any root of the equation ρ2 = r2 − λ2 − µ2 .

Sub-family IIb. If p = ±1 and δ = −r2 then ρ = 0 and thus a1 = 0. In
m
this case we have A(ξ) = ± 2 + b1 g(ξ) where g(ξ) is given by (88.20).
4s
1188
1
(IIb.a) For p = 1, we have b1 = − and q = 4m2 . Therefore g(ξ) =
8s2
12℘0 (ξ)
and thus
12℘0 (ξ) + 4m2
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.41)
4s2 2s2 12℘0 (ξ) + 4m2
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) − (88.42)
3 27
1
(IIb.b) For p = −1, we have b1 = and q = −4m2 . Therefore g(ξ) =
8s2
12℘0 (ξ)
and thus
12℘0 (ξ) − 4m2
m 3 ℘0 (ξ)
A(ξ) = ± 2 + 2 · (88.43)
4s 2s 12℘0 (ξ) − 4m2
r2
Sub-family IIc. If p = ±1 and δ = − then ρ satisfies the equation
25
24r2
ρ2 = . The functions f (ξ) and g(ξ) are given by (88.23) and (88.24)
25
respectively.
1 ρ 1
(IIc.a) For p = 1, we have we have b1 = − 2 , a1 = − · 2 and
8s 2m 8s
q = 4m2 . We then have the following expressions for f (ξ) and g(ξ)
10m2 10m4 4m2 ℘0 (ξ)

f (ξ) = + and g(ξ) = −
3r 9℘(ξ) [4m2 + 12℘(ξ)]℘(ξ)
Therefore
m 5ρm h 1 m2 i m2 ℘0 (ξ)
A(ξ) = ± 2 − + + 2· (88.44)
4s 24s2 r 3℘(ξ) 2s ℘(ξ)[4m2 + 12℘(ξ)]
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) − (88.45)
3 27
1189
1 ρ 1
(IIc.b) For p = −1, we have we have b1 = 2
, a1 = · 2 and
8s 2m 8s
q = −4m2 . We then have the following expressions for f (ξ) and g(ξ)
10m2 10m4 4m2 ℘0 (ξ)

f (ξ) = − − and g(ξ) =
3r 9℘(ξ) [4m2 + 12℘(ξ)]℘(ξ)
Therefore
m 5ρm h 1 m2 i m2 ℘0 (ξ)
A(ξ) = ± − + + · (88.46)
4s2 24s2 r 3℘(ξ) 2s2 ℘(ξ)[4m2 + 12℘(ξ)]
Note. If δ = λ2 − µ2 and pq < 0 leads to m2 < 0. The case q = 0 cannot be

considered either, since it leads to m = 0 and then the coefficient of f (ξ) in
(88.37) becomes infinite.
Family III. This family corresponds to the Solution III
m 4m2 a21 − r2 b21 2 m2
a0 = ± , a 1 = a 1 , b 1 = b 1 , δ = , p = 8s b 1 , q = −
4s2 b21 2s2 b1
From (88.27) and Solution III, we obtain
m
A(ξ) = ± + a1 f (ξ) + b1 g(ξ) (88.47)
4s2
4m2 a21 − r2 b21 2 m2
where δ = , p = 8s b 1 , q = −
b21 2s2 b1
Sub-family IIIa. Since pq = −4m < 0 for δ = λ2 − µ2 , i.e. λ2 − µ2 =
2
4m2 a21 − r2 b21

the functions f (ξ) and g(ξ) are given by (88.10) and (88.11)
b21
respectively:
√ 0
m2 1 1 ξU (ξ)
f (ξ) = − 2
· , g(ξ) = 2 ·
2b1 s U (ξ) 4s U (ξ)
where
p p
Therefore
√
m 2m2 a1 1 1 ξU 0 (ξ)
A(ξ) = ± 2 − 2
· + 2· (88.49)
4s 4b1 s U (ξ) 4s U (ξ)
1190
where the following relation holds 4m2 a21 = (λ2 − µ2 + r2 )b21 .
4m2 a21 − r2 b21
Sub-family IIIb. If p = ±1 and δ = −r2 , i.e. −r2 = then
b21
a1 = 0 and g(ξ) is given by (88.20).
1 12℘0 (ξ)
(IIIb.a) If p = 1, then b1 = 2 , q = −4m2 and g(ξ) = .
8s 12℘0 (ξ) − 4m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.50)
4s2 8s2 12℘0 (ξ) − 4m2
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.51)
3 27
1 12℘0 (ξ)
(IIIb.b) If p = −1, then b1 = − , q = 4m2 and g(ξ) = .
8s2 12℘0 (ξ) + 4m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.52)
4s2 8s2 3℘0 (ξ) + m2
r2 r2 4m2 a21 − r2 b21

Sub-family IIIc. If p = ±1 and δ = − , i.e. − = then
25 25 b21
f (ξ) and g(ξ) are given by (88.23) and (88.24) respectively.
1
(IIIc.a) For p = 1, we have b1 = 2 and q = −4m2 and thus
8s
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − + and g(ξ) =
3r 9℘(ξ) [12℘(ξ) − 4m2 ]℘(ξ)
Therefore
m 10a1 m2 1 m2 m2 ℘0 (ξ)
A(ξ) = ± + − + + · (88.53)
4s2 3 r 3℘(ξ) 8s2 ℘(ξ)[3℘(ξ) − m2 ]
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.54)
3 27
1191
with 100m2 a21 = 24r2 b21 .
1
(IIIc.b) For p = −1, we have b1 = − and q = 4m2 and thus
8s2
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − and g(ξ) = −
3r 9℘(ξ) [12℘(ξ) + 4m2 ]℘(ξ)
Therefore
m 10a1 m2 1 m2 m2 ℘0 (ξ)
A(ξ) = ± + − + · (88.55)
4s2 3 r 3℘(ξ) 8s2 ℘(ξ)[3℘(ξ) + m2 ]
Note. If δ = −λ2 − µ2 and pq > 0 leads to m2 < 0. The choice q = 0 leads

to m = 0.
Family IV. This family corresponds to the Solution IV
m 4m2 a21 − r2 b21 2 m2

a0 = ± , a 1 = a 1 , b 1 = b 1 , δ = , p = −8s b 1 , q =
4s2 b21 2s2 b1
From (88.27) and Solution IV, we obtain

m
A(ξ) = ± + a1 f (ξ) + b1 g(ξ) (88.56)
4s2
4m2 a21 − r2 b21 2 m2
where δ = , p = −8s b 1 , q = .
b21 2s2 b1
Sub-family IVa. Since pq = −4m2 < 0 for δ = λ2 − µ2 , i.e. λ2 − µ2 =
4m2 a21 − r2 b21
then f (ξ) and g(ξ) are given by (88.10) and (88.11) respec-
b21
tively:
√ 0
m2 1 1 ξU (ξ)
f (ξ) = 2 · , g(ξ) = − 2 ·
2s b1 U (ξ) 4s U (ξ)
where
p p
Therefore
√
m m 2 a1 1 1 ξU 0 (ξ)
A(ξ) = ± 2 + 2 · − 2· (88.58)
4s 2s b1 U (ξ) 4s U (ξ)
1192
with 4m2 a21 = (λ2 − µ2 + r2 )b21 .
4m2 a21 − r2 b21
Sub-family IVb. If p = ±1 and δ = −r2 , i.e. −r2 = , we
b21
have a1 = 0.
1
(IVb.a) For p = 1, we have b1 = − q = −4m2 and thus from (C.27) we
8s2
12℘0 (ξ)
get g(ξ) = and then
12℘0 (ξ) − 4m2
m 3 ℘0 (ξ)
A(ξ) = ± 2 − 2 · 0 (88.59)
4s 8s 3℘ (ξ) − m2
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.60)
3 27
1
(IVb.b) For p = −1, we have b1 = , q = 4m2 and thus form (C.27) we
8s2
12℘0 (ξ)
get g(ξ) = and then
12℘0 (ξ) + 4m2
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.61)
4s2 8s2 3℘0 (ξ) + m2
r2 r2 4m2 a21 − r2 b21
Sub-family IVc. If p = ±1 and δ = − , i.e. − = then
25 25 b21
f (ξ) and g(ξ) are given by (88.23) and (88.24) respectively.
1
(IVc.a) For p = 1, we have b1 = − 2 , q = −4m2 and then
8s
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − + and g(ξ) =
3r 9℘(ξ) [12℘(ξ) − 4m2 ]℘(ξ)
Therefore
m 10a1 m2 1 m2 m2 ℘0 (ξ)
A(ξ) = ± + − + − · (88.62)
4s2 3 r 3℘(ξ) 8s2 ℘(ξ)[3℘(ξ) − m2 ]
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.63)
3 27
1193
3r2
with a21 = . This last equation comes from equating the two different
800m2 s4
expressions of δ and using the value of b1 .
1
(IVc.b) For p = −1, we have b1 = 2 , q = 4m2 and then
8s
10m2 10m4 4m2 ℘0 (ξ)
f (ξ) = − and g(ξ) = −
3r 9℘(ξ) [12℘(ξ) + 4m2 ]℘(ξ)
Therefore
m 10a1 m2 1 m2 m2 ℘0 (ξ)
A(ξ) = ± 2 + − − 2· (88.64)
4s 3 r 3℘(ξ) 8s ℘(ξ)[3℘(ξ) + m2 ]
Note. The case δ = −λ2 − µ2 and pq > 0 leads to m2 < 0. The choice q = 0
leads to m = 0.
Family V. This family corresponds to the Solution V
m 2 m2
a0 = ± , a1 = 0, b1 = b 1 , p = 4s b 1 , q = −
4s2 4s2 b1
From (88.27)and Solution V, we obtain
m
A(ξ) = ± + b1 g(ξ) (88.65)
4s2
2
where p = 4s2 b1 , q = − 4sm2 b1 .
Sub-family Va. Since pq = −4m2 < 0, the √ choice δ = λ2 − µ2 , the function
0
m 2 ξ·X ξ)
g(ξ) is given by (88.11): g(ξ) = 2 · where
4s b1 X(ξ)
p p
X(ξ) = r + λ·sinh(|m| ξ) + µ·cosh(|m| ξ) (88.66)
Therefore
√
m m ξX 0 (ξ)
A(ξ) = ± 2 + 2 · (88.67)
4s 2s X(ξ)
Sub-family Vb. If p = ±1 and δ = −r2 , then g(ξ) is given by (88.20).
1 12℘0 (ξ)
(Vb.a) For p = 1, we have b1 = 2 , q = −m2 and then g(ξ) = .
4s 12℘0 (ξ) − m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.68)
4s2 s2 12℘0 (ξ) − m2
1194
m4 m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.69)
12 216
1 2 12℘0 (ξ)
(Vb.b) For p = −1, we have b1 = − , q = m and then g(ξ) = .
4s2 12℘0 (ξ) + m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.70)
4s2 s2 12℘0 (ξ) + m2
r2
Sub-family Vc. If p = ±1 and δ = − then g(ξ) is given by (88.24).
25
1 m2 ℘0 (ξ)
(Vc.a) For p = 1, we get b1 = 2 , q = −m2 and then g(ξ) = .
4s ℘(ξ)[12℘0 (ξ) − m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± + · (88.71)
4s2 4s2 ℘(ξ)[12℘(ξ) − m2 ]
1 2 m2 ℘0 (ξ)
(Vc.b) For p = −1, we get b1 = − , q = m and then g(ξ) = .
4s2 ℘(ξ)[12℘(ξ) + m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± + · (88.72)
4s2 4s2 ℘(ξ)[12℘(ξ) + m2 ]
Note. The case δ = −λ2 − µ2 and pq > 0 leads to m2 < 0. The case q = 0
leads to m = 0.
Family VI. This family corresponds to Solution VI
m 2 m2
a0 = ± , a1 = 0, b1 = b 1 , p = 8s b 1 , q = −
4s2 2s2 b1
and r is any root of r2 + δ = 0.
From (88.27)and Solution VI, we obtain
m
A(ξ) = ± + b1 g(ξ) (88.73)
4s2
2
where p = 8s2 b1 , q = − 2sm2 b1 and r is any root of r2 + δ = 0.
Sub-family VIa. Since pq = −4m2 < 0, the choice δ = λ2 − µ2 , i.e.
1195
√
2 2 2 1 ξ·U 0 (ξ)
r + λ − µ = 0 and g(ξ) is given by (88.11): g(ξ) = 2 · where
4s b1 U (ξ)
p p
Therefore
√
m m ξU 0 (ξ)
A(ξ) = ± 2 + 2 · (88.75)
4s 4s U (ξ)
where r satisfies the equation r2 + λ2 − µ2 = 0.

Sub-family VIb. If p = ±1 and δ = −r2 (i.e. r is any real) then g(ξ) is
given by (88.20).
1 12℘0 (ξ)
(VIb.a) For p = 1, we get b1 = 2 , q = −4m2 and thus g(ξ) = .
8s 12℘0 (ξ) − 4m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.76)
4s2 8s2 3℘0 (ξ) − m2
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.77)
3 27
1 2 12℘0 (ξ)
(VIb.b) For p = −1, we get b1 = − 2 , q = 4m and thus g(ξ) = .
8s 12℘0 (ξ) + 4m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.78)
4s2 8s2 3℘0 (ξ) + m2
r2 2 r2
Sub-family VIc. If p = ±1 and δ = − , i.e. r − = 0 then g(ξ) is
25 25
given by (88.24).
1 4m2 ℘0 (ξ)
(VIc.a) For p = 1, we get b1 = 2 , q = −4m2 and then g(ξ) = .
8s ℘(ξ)[12℘0 (ξ) − 4m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± + · (88.79)
4s2 8s2 ℘(ξ)[3℘(ξ) − m2 ]
1196
1 2 4m2 ℘0 (ξ)
(VIc.b) For p = −1, we get b1 = − , q = 4m and then g(ξ) = − .
8s2 ℘(ξ)[12℘(ξ) + 4m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± + · (88.80)
4s2 8s2 ℘(ξ)[3℘(ξ) + m2 ]
leads to m = 0.
Family VII. This family corresponds to the Solution VII
m 2 m2
a0 = ± , a1 = 0, b1 = b 1 , p = −8s b 1 , q =
4s2 2s2 b1
and r is any root of r2 + δ = 0.
From (88.27)and Solution VII, we obtain
m
A(ξ) = ± + b1 g(ξ) (88.81)
4s2
m2
where p = −8s2 b1 , q = and r is any root of r2 + δ = 0.
4s2 b1
Sub-family VIIa. Since pq = −4m2 < 0, the choice δ = λ2 −√µ2 , i.e.
2 2 2 1 ξ·U 0 (ξ)
r + λ − µ = 0, then g(ξ) is given by (88.11): g(ξ) = − 2 ·
4s b1 U (ξ)
where
p p
U (ξ) = r + λ·sinh(2|m| ξ) + µ·cosh(2|m| ξ)
Therefore
√ 0
m 1 ξU (ξ)
A(ξ) = ± 2 − 2 · (88.82)
4s 4s U (ξ)
where r satisfies the equation r2 + λ2 − µ2 = 0.
Sub-family VIIb. If p = ±1 and δ = −r2 (i.e. r is any real) then g(ξ) is
given by (88.20).
1
(VIIb.a) For p = 1, we have b1 = − 2 , q = −4m2 and thus g(ξ) =
8s
12℘0 (ξ)
. Therefore
12℘0 (ξ) − 4m2
m 3 ℘0 (ξ)
A(ξ) = ± − · (88.83)
4s2 8s2 3℘0 (ξ) − m2
1197
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.84)
3 27
1 2 12℘0 (ξ)
(VIIb.b) For p = −1, we have b1 = , q = 4m and thus g(ξ) = .
8s2 12℘0 (ξ) + 4m2
Therefore
m 3 ℘0 (ξ)
A(ξ) = ± + · (88.85)
4s2 8s2 3℘0 (ξ) + m2
r2 r2
Sub-family VIIc. If p = ±1 and δ = − , i.e. r2 − = 0 then g(ξ) is
25 25
given by (88.24).
1 2 4m2 ℘0 (ξ)
(VIIc.a) For p = 1, we get b1 = − , q = −4m and then g(ξ) = .
8s2 ℘(ξ)[12℘(ξ) − 4m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± 2 − 2 · (88.86)
4s 8s ℘(ξ)[3℘(ξ) − m2 ]
1 2 4m2 ℘0 (ξ)
(VIIc.b) For p = −1, we get b1 = , q = 4m and then g(ξ) = − .
8s2 ℘(ξ)[12℘(ξ) + 4m2 ]
Therefore
m m2 ℘0 (ξ)
A(ξ) = ± − · (88.87)
4s2 8s2 ℘(ξ)[3℘(ξ) + m2 ]
leads to m = 0.
Family VIII. This family corresponds to the Solution VIII
ρp 4m2
a0 = 0, a1 = 2 , b1 = 0, p = p, q = −
8s m p
and ρ is any root of ρ2 = δ + r2 .

From (88.27) and Solution VIII, we obtain
ρp
A(ξ) = f (ξ) (88.88)
8s2 m
1198
4m2
where p = p, q = − and ρ is any root of the equation ρ2 = δ + r2 .
p
Sub-family VIIIa. Since pq = −4m2 < 0, the choice δ = λ2 − µ2 , i.e.
4m2 1
ρ2 = λ2 − µ2 + r2 , then f (ξ) is given by (88.10): f (ξ) = − · where
p U (ξ)
p p
Therefore
ρm 1
A(ξ) = − 2
· (88.90)
2s U (ξ)
where ρ satisfies the equation ρ2 = λ2 − µ2 + r2 .

r2 24 2
Sub-family VIIIb. If p = ±1 and δ = − , leads to ρ2 = r and then
25 25
f (ξ) is given by (88.23).
10m2 10m4
(VIIIb.a) For p = 1, we have q = −4m2 and thus f (ξ) = − +
3r 9℘(ξ)
4m4 8m6
(℘0 (ξ))2 = 4℘3 (ξ) − ℘(ξ) + (88.91)
3 27
Therefore
5ρm 1 m2
A(ξ) = − + (88.92)
12s2 r 3℘(ξ)
24r2
and ρ is any root of the equation ρ2 = .
25
10m2 10m4
(VIIIb.b) For p = −1, we have q = 4m2 and thus f (ξ) = − .
3r 9℘(ξ)
Therefore
5ρm 1 m2
A(ξ) = − − (88.93)
12s2 r 3℘(ξ)
leads to m = 0. The case δ = −r2 cannot be considered since in this case
ρ = 0 and then a1 = 0 (i.e. A(ξ) = 0).
References
1199
[1] M. A. Abdou: ”Exact Solutions for Nonlinear Evolution Equations via
Extended Projective Riccati Equation Expansion Method”.
Electr. J. Theor. Phys. 4 (2007) 17-30
[2] S. M. Antoniou: ”The Riccati Equation Method with Variable
Expansion Coefficients. I. Solving the Burgers Equation”.
SKEMSYS report available form ResearchGate
89 The Riccati Equation Method with

Variable Expansion Coefficients
This method was introduced by S. M. Antoniou (Antoniou [1]-[9]) and has
applied successfuly in solving a number of nonlinear equations.
The Method. We consider an evolution equation of the general form
ut = G(u, ux , uxx , ...) or utt = G(u, ux , uxx , ...) (89.1)
where u is a smooth function. We introduce a new variable ξ given by
ξ = k(x − ωt) (89.2)
where k and ω are constants. We change variables and introduce a new
function U (ξ) by u(x, t) = U (k(x − ωt)) ≡ U (ξ).
Since
ut = (−ωk)U 0 (ξ), ux = kU 0 (ξ), uxx = k 2 U 00 (ξ), . . . (89.3)
equations (89.1) become ordinary differential equations
dU dU 2 d2 U
(−ωk) = G(u, k ,k , . . .) (89.4)
dξ dξ dξ 2
and
dU dU 2 d2 U
ω2k2 = G(u, k ,k , . . .) (89.5)
dξ dξ dξ 2
Equations (89.4) or (89.5) will be solved considering expansions of the form
n
X
U (ξ) = ak Y k (89.6)
k=0
1200
or
n n
X X bk
U (ξ) = ak Y k + (89.7)
k=0 k=1
Yk
where all the expansion coefficients depend on the variable ξ,

ak ≡ ak (ξ), bk ≡ bk (ξ), for every k = 0, 1, 2, . . .n
contrary to the previously considered cases where the expansion coefficients
were considered as constants. The function Y (ξ) satisfies Riccati’s equation
Y 0 (ξ) = A + BY 2 (89.8)
where again the coefficients A and B depend on the variable ξ.
In solving equations (89.4) or (89.5), we consider the expansions (89.6) or
(89.7) and then we balance the nonlinear term with the highest derivative of
the function U (ξ) which determines n (the number of the expansion terms).
Equating similar powers of the function Y (ξ) we can determine the various
coefficients and thus find the solution of the equation considered.
A remarkable feature of the method is that we obtain quite new solutions and
apart of that, we get a proliferation of solutions, once we continue to apply
the method in all the intermediate equations, leading to the final solution of
the original equation.
Example. The Newell-Whitehead equation and its solutions.
The NW equation was introduced by Newell and Whitehead and belongs to
the general class of reaction-diffusion equations. This equation describes in
particular the Rayleigh-Benard convection. The NW equation has also been
considered and solved through a variety of methods before.
We consider the NW equation in the form (a > 0, b > 0)
ut = uxx + au − bu3 (89.9)

and try to find traveling wave solutions of this equation (Antoniou [3]).
We introduce a new variable ξ given by
ξ = x − ωt (89.10)
where ω is a non-zero constant, ω 6= 0. Changing variables and introducing
a new function U (ξ) by u(x, t) = U (x − ωt) ≡ U (ξ) since
ut = (−ω)U 0 (ξ), ux = U 0 (ξ), uxx = U 00 (ξ), . . . (89.11)
1201
equation (89.9) becomes an ordinary differential equation
U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0 (89.12)
We consider the extended Riccati equation method in solving equation

(89.12), since the Riccati equation method leads to trivial results.
We thus consider the expansion
n n
X
k
X bk
U (ξ) = ak Y +
k=0 k=1
Yk
and balance the second order derivative term with the third order of (3.4).
We then find that n = 1. The proof might go as follows: The first derivative
U 0 (ξ) contains the highest order term Y n−1 Y 0 and upon the substitution
Y 0 →A + BY 2 the highest order term becomes Y n+1 . The second derivative
term U 00 (ξ) contains the highest order term Y n Y 0 and upon the substitution
Y 0 →A + BY 2 the highest order term becomes Y n+2 . The nonlinear term
U 3 contains the highest order term U 3n . Therefore balancing the second
derivative term U 00 (ξ) with the nonlinear term U 3 (ξ) leads to the equation
n + 2 = 3n from which we obtain n = 1. We thus have
b1
U (ξ) = a0 + a1 Y + (89.13)
Y
where again all the coefficients a0 , a1 and b1 depend on ξ, and Y satisfies
Riccati’s equation Y 0 = A + BY 2 . From equation (89.13) we obtain (taking
into account Y 0 = A + BY 2 )
b01 b1 A
U 0 (ξ) = (a00 − b1 B + a1 A) + a1 BY 2 + − 2 (89.14)
Y Y
U 00 (ξ) = (a000 + 2a01 A + a1 A0 − b1 B 0 − 2b01 B)
+ (a001 + 2a1 AB)Y + (a1 B 0 + 2a01 B)Y 2 + (2a1 B 2 )Y 3
(89.15)
b00 + 2b1 AB 2b01 A + b1 A0 2b1 A2
+ 1 − +
Y Y2 Y3
1202
Therefore equation (89.12), under the substitution (89.13), (89.14) and (89.15),
becomes
(a000 + 2a01 A + a1 A0 − b1 B 0 − 2b01 B)
+ (a001 + 2a1 AB)Y + (a1 B 0 + 2a01 B)Y 2 + (2a1 B 2 )Y 3
b001 + 2b1 AB 2b01 A + b1 A0 2b1 A2
+ − +
Y Y2 Y3 (89.16)
b0 b1 A
+ ω (a00 − b1 B + a1 A) + a01 Y + a1 BY 2 + 1 − 2
Y Y
b1 b1 3
+ a a0 + a1 Y + − b a0 + a1 Y + =0
Y Y
Upon expanding and equating the coefficients of Y to zero, we obtain a system
of differential equations from which we can determine the various expansion
coefficients and the coefficients A and B of Riccati’s equation. We obtain
coefficient of Y 3 :
−ba31 + 2a1 B 2 = 0 (89.17)
coeffficient of Y 2 :
2a01 B − 3ba0 a21 + a1 B 0 + ωa1 B = 0 (89.18)
coefficient of Y :
a001 + aa1 + ωa01 − 3b(b1 a21 + a1 a20 ) + 2a1 AB = 0 (89.19)
coefficient of Y 0 :
a000 + aa0 + 2a01 A − b1 B 0 + a1 A0 − 2b01 B + ω(a00 + a1 A − b1 B)
(89.20)
− b(a30 + 6a0 a1 b1 ) = 0
coefficient of Y −1 :
b001 + ab1 + 2b1 AB + ωb01 − 3b(a20 b1 + a1 b21 ) = 0 (89.21)
−3ba0 b21 − ωAb1 − b1 A0 − 2b01 A = 0 (89.22)
2b1 A2 − bb31 = 0 (89.23)
1203
We now have to solve the system of equations (89.17)-(89.23) supplemented
by the Riccati equation Y 0 = A + BY 2 .
From equations (89.17) and (89.23), ignoring the trivial solutions, we obtain
r r
2 2
a1 = ± B and b1 = ± A (89.24)
b b
respectively. We thus consider
r the following
r four cases.
2 2
3.I. Case I. For a1 = B and b1 = A, we obtain
b b
from (89.22)
r
2 A0
6a0 = − ω+3 (89.25)
b A
from (89.18)
r
2 B0
6a0 = ω+3 (89.26)
b B
from (89.19)
B 00 B0
+ ω − 4AB + a − 3ba20 = 0 (89.27)
B B
from (89.21)
A00 A0
+ ω − 4AB + a − 3ba20 = 0 (89.28)
A A
from (89.20)
r
2
a000 + ωa00 + aa0 − 12a0 AB − ba30 + (AB 0 − A0 B) = 0 (89.29)
b
Equating the two different expressions of a0 , equations (89.25) and (89.26),
A0 B 0
we obtain the equation 3 + = −2ω, which by integration yields
A B
2ω
AB = K exp − ξ (89.30)
3
1204
where K is a positive constant, K > 0.
Using (89.26) and (89.30), we obtain from (89.27) the equation
B 00 3 B 0 2 2ω ω2
− − 4Kexp − ξ + a − =0 (89.31)
B 2 B 3 6
The previous equation, under the substitution
B0
F = (89.32)
B
transforms into the equation
1 2ω ω2
F 0 = F 2 + 4Kexp − ξ − a + (89.33)
2 3 6
which is a Riccati differential equation. Under the standard transformation
w0
F = −2 (89.34)
w
equation (89.33) takes on the form of a second order linear differential equa-
tion
2ω a ω 2
w00 + 2Kexp − ξ − + w=0 (89.35)
3 2 12
The substitution z = eµξ transforms (89.35) into
2ω
2
d w dw 1 − a ω2
z2 2 + z + 2 2Kz 3µ − + w=0 (89.36)
dz dz µ 2 12
2ω
ω −
The choice µ = − converts z 3µ into z 2 and thus equation (89.35) takes
3
on the form
2
2d w dw 18K 2 3 9a
z +z + z + − 2 w=0 (89.37)
dz 2 dz ω2 4 2ω
The substitution
18K 3 9a
ρ2 = and − ν2 = − 2 (89.38)
ω2 4 2ω
1205
d2 w dw
z2 2
+z + (ρ2 z 2 − ν 2 )w = 0 (89.39)
dz dz
which is Bessel’s differential equation with general solution
w = C1 Jν (ρz) + C2 Yν (ρz) (89.40)

3√2K ω
−ξ 3√2K − ω ξ
w(ξ) = C1 Jν e 3 + C 2 Yν e 3 (89.41)
ω ω
where
√
18a − 3ω 2
ν= (6a > ω 2 ) (89.42)
2ω
Bessel’s functions are known to be defined by
k
1 2
1 ν X∞ −4z Jν (z)cos(νπ) − J−ν (z)
Jν (z) = z and Yν =
2 k=0
Γ(ν + k + 1)k! sin(νπ)
respectively.
The constants C1 and C2 in (89.41) are chosen such that w(ξ) 6= 0, ∀ξ.
It is not trivial to ensure that, since the Bessel functions are oscillating at
infinity. However if the constant ω > 0, then we have not such a problem in
√ ω
3 2 − ξ
(89.41) w(ξ) 6= 0, ∀ξ because e 3 → 0 as ξ → ∞.
ω
B0 w0
Combining (89.32) and (89.34), we derive the equation + 2 = 0, from
B w
which by integration we obtain
C
B(ξ) = (89.43)
w2 (ξ)
with w given by (89.41). Using equations (89.30) and (89.43), we determine
the coefficient A:
K 2 2ω
A(ξ) = w (ξ)exp(− ξ) (89.44)
C 3
1206
We thus get
r r
2 1 2 2 2ω
a1 = C and b1 = C w (ξ)exp − ξ (89.45)
b w2 (ξ) b 3
0 0
and from (89.26), since BB + 2 ww = 0,
r
1 2 w0 (ξ)
a0 (ξ) = ω−6 (89.46)
6 b w(ξ)
Equation (89.29), using (89.26) and (89.30), can be expressed in terms of

B0
F = as
B
2ω
00 1 3
0 1 2 ω2 − ξ
F − F +ω F − F + a− F − 8KF e 3
2 2 6 (89.47)
2ω
8K − ξ ω
− ωe 3 + (18a − ω 2 ) = 0
3 54
The above equation should be compatible to (89.33) (i.e. the solution (89.41)
should satisfy (89.47) given that F is connected to w(ξ) through (89.34)).
According to the results of Appendix A, we have the constraints
1 9a
ν2 = , ω2 =
4 2
and
3√2K 3√2K 3√2K 3√2K
Jν × Yν+1 − Jν+1 × Yν =0
ω ω ω ω
We turn now to the determination of the function Y which satisfies Riccati’s
equation. Under the substitution
1 u0
Y =− (89.48)
Bu
Riccati’s equation Y 0 = A + BY 2 becomes
B0
u00 − u0 + ABu = 0 (89.49)
B
1207
The substitution
√
u= By (89.50)
001 h B 00 3 B 0 2 i 2ω
y + − y + Kexp − ξ y = 0 (89.51)
2 B 2 B 3
C B 00 3 B 0 2
taking also into account (89.30). Since B = 2 , we have − =
w (ξ) B 2 B
w00
−2 and using (89.40), we obtain, using the various formulas for the deriva-
w
tives and recurrence relations of Bessel’s functions, that
w00 2 2 2ω
−2 = ρ exp − ξ − ν 2 ω 2
w 9 3
h 2ω ων 2 i
y 00 + λ2 exp − ξ − y=0 (89.52)
3 3
where
1
λ2 = ρ2 ω 2 + K = 3K (89.53)
9
Equation (89.52) can be solved by converting it to Bessel’s equation, along
the lines of reasoning in transforming (89.35) into (89.39). We thus get the
following solution of equation (89.52):
3√3K ω 3√3K ω
y = C̃1 Jν e− 3 ξ + C̃2 Yν e− 3 ξ (89.54)
ω ω
The same index ν appears in both expressions of the Bessel’s function (89.54)
and (89.41), and is given by (89.42). We then obtain from (89.48), using
(89.50) and (89.43)
1 y 0 (ξ) w0 (ξ)
Y = − w2 (ξ) − (89.55)
C y(ξ) w(ξ)
1208
Therefore we arrive at the following
Solution I. The solution of equation (89.12)
U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0

b1
is given by (89.13), U (ξ) = a0 + a1 Y + where
Y
r r
1 2 w0 (ξ) 2 1
a0 (ξ) = ω−6 , a1 = C
6 b w(xi) b w2 (ξ)
r
K 2 2 2ω 1 2 y 0 (ξ) w0 (ξ)
b1 = w (ξ)exp − ξ and Y = − w (ξ) −
C b 3 C y(ξ) w(ξ)
with w(ξ) and y(ξ) being expressed in terms of Bessel’s functions by (89.41)
and (89.54) respectively. Therefore
2ω
r
1 2 0
y (ξ) exp − ξ
U (ξ) = ω−6 − 6K 0 3
6 b y(ξ) y (ξ) w0 (ξ)
−
y(ξ) w(ξ)
w0 (ξ) ω
In the above solution we have to take into account the condition = ,
w(ξ) 6
according to (89.99) of Appendix A. We thus have the following solution
expressed by Theorem I:
Theorem I. The traveling wave solutions of the Newell-Whitehead equation
ut = uxx + au − bu3 , derived under the substitution ξ = x − ωt, i.e. the
solutions of the equation U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0, are given by
r
1 2h y 0 (ξ) y 0 (ξ) ω −1 2ω i
U (ξ) = ω−6 − 6K − × exp − ξ (89.56)
6 b y(ξ) y(ξ) 6 3
The above solution is subject to the constraints

1 9a
ν2 = , ω2 = (89.57)
4 2
and
3√2K 3√2K 3√2K 3√2K
Jν × Yν+1 − Jν+1 × Yν = 0 (89.58)
ω ω ω ω
1209
r r
2 2
3.II. Case II. For a1 = B and b1 = − A, we obtain
b b
from (89.22)
r
2 A0
6a0 = ω+3 (89.59)
b A
from (89.18)
r
2 B0
6a0 = ω+3 (89.60)
b B
from (89.19)
B 00 B0
+ ω + 8AB − 3ba20 + a = 0 (89.61)
B B
from (89.21)
A00 A0
+ ω + 8AB − 3ba20 + a = 0 (89.62)
A A
from (89.20)
r r
2 2
a000 + ωa00 + aa0 + 12a0 AB − ba30 +3 (AB)0 + 2ω AB = 0 (89.63)
b b
Equating the two different expressions for a0 , equations (89.59) and (89.60),
A0 B0
we find that = and by integration
A B
B = −s2 A (89.64)
where s is real. Equation (89.62) then, using (89.59) and (89.64), becomes
3
AA00 − (A0 )2 + 2m2 A2 − 8s2 A4 = 0 (89.65)
2
where
ω2
2m2 = a − , m real (89.66)
6
1210
Equation (89.65) has been solved by a variety of methods in Paper I (Anto-
niou [1], Appendices A, B and C). In Appendix C of this paper we list all
the solution methods and the corresponding solutions we have already found
in Paper I.
A0
Equation (89.63) expressed in terms of G = gives, taking into account
A
(89.59)
1 1 ω2
G00 − G3 + ω G0 − G2 + a − G
2 2 6 (89.67)
ω
− 24s2 AA0 − 8ωs2 A2 + (18a − ω 2 ) = 0
54
All the solutions which satisfy (89.65) should be substituted in the above
equation. The resulting expressions will provide compatibility conditions
between the various parameters and constants. According to the results of
Appendix B, we find that (89.65) (expressed in terms of G, see (89.107)) is
compatible to (89.67) if
9a
ω2 = (89.68)
2
We now turn to the determination of the function Y which satisfies Riccati’s
equation. Under the substitution
1 v0
Y =− (89.69)
Bv
Riccati’s equation Y 0 = A + BY 2 becomes
B0
v 00 − v 0 + ABv = 0 (89.70)
B
A0
v 00 − v 0 − s2 A2 v = 0 (89.71)
A
v0
From the above equation we obtain = ±sA and then (see eqs (3.53)-(3.60)
v
of paper I, Antoniou [1])
1
Y (ξ) = ± (89.72)
s
1211
Collecting the results of this Section, we arrive at
Solution II. The solution of equation (89.12)
U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0

b1
Y
r r r
1 2 A0 2 2 2 1
a0 = ω+3 , a1 = −s A, b1 = − A and Y = ±
6 b A b b s
and A is any solution of the equation (89.65). We thus have the following
solution expressed by Theorem II:
Theorem II. The traveling wave solutions of the Newell-Whitehead equation
ut = uxx + au − bu3 , derived under the substitution ξ = x − ωt, i.e. the
r r
1 2 A0 (ξ) 2 1
U (ξ) = ω+3 − 2s A(ξ), Y = , (89.73)
6 b A(ξ) b s
r r
1 2 A0 (ξ) 2 1
U (ξ) = ω+3 + 2s A(ξ), Y = − , (89.74)
6 b A(ξ) b s
9a
with ω 2 = while the various functions A(ξ) which are solutions of (89.65)
2
are given in Appendix C. r r
2 2
3.III. Case III. For a1 = − B and b1 = A, we obtain
b b
from (89.22)
r
2 A0
6a0 = − ω+3 (89.75)
b A
from (89.18)
r
2 B0
6a0 = − ω+3 (89.76)
b B
from (89.19)
B 00 B0
+ ω + 8AB − 3ba20 + a = 0 (89.77)
B B
1212
from (89.21)
A00 A0
+ ω + 8AB − 3ba20 + a = 0 (89.78)
A A
from (89.20)
r r
2 2
a000 + ωa00 + aa0 + 12a0 AB − ba30 −3 (AB)0 − 2ω AB = 0 (89.79)
b b
Equating the two different expressions for a0 , equations (89.75) and (89.76),
1
we conclude that B = −s2 A, which is (89.64). We thus obtain again Y = ± .
s
Equation (89.78) gives, because of (89.75), equation (89.65).
Equation (89.79) also gives, taking into account (89.75) and (89.64)
1 1 ω2
G00 − G3 + ω G0 − G2 + a − G
2 2 6 (89.80)
2 0 2 2 ω 2
− 24s AA − 8ωs A − (18a − ω ) = 0
54
All the solutions which satisfy (89.78) (i.e.(89.65)) should be substituted in
the above equation. The resulting expressions will provide compatibility
conditions between the various parameters and constants. According to the
9a
results of Appendix B, we find ω 2 = . We thus obtain the following
2
Solution III.. The solution of equation (89.12)
U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0
b1
Y
r r r
1 2 A0 2 2 2 1
a0 = − ω+3 , a1 = s A, b1 = A and Y = ±
6 b A b b s
and A is any solution of the equation (89.65). We thus have the following
solution expressed by Theorem III:
Theorem III. The traveling wave solutions of the Newell-Whitehead equa-
tion ut = uxx + au − bu3 , derived under the substitution ξ = x − ωt, i.e. the
r r
1 2 A0 (ξ) 2 1
U (ξ) = − ω+3 + 2s A(ξ), Y = , (89.81)
6 b A(ξ) b s
1213
r r
1 2 A0 (ξ) 2 1
U (ξ) = − ω+3 − 2s A(ξ), Y = − , (89.82)
6 b A(ξ) b s
9a
with ω 2 = while the various functions A(ξ) which are solutions of (89.65)
2
are given in Appendix C. r r
2 2
3.IV. Case IV. For a1 = − B and b1 = − A, we obtain
b b
from (89.22)
r
2 A0
6a0 = ω+3 (89.83)
b A
from (89.18)
r
2 B0
6a0 = − ω+3 (89.84)
b B
from (89.19)
B 00 B0
+ ω − 4AB + a − 3ba20 = 0 (89.85)
B B
from (89.21)
A00 A0
+ ω − 4AB + a − 3ba20 = 0 (89.86)
A A
from (89.20)
r
2
a000 + ωa00 + aa0 − 12a0 AB − ba30 − (AB 0 − A0 B) = 0 (89.87)
b
Equating the two different expressions of a0 , equations (89.83) and (89.84),
A0 B0
we obtain the equation 3 + = −2ω, which by integration gives
A B
(89.30). We then obtain from (89.85), using (89.84) and (89.30), the equation
(89.31), which can be converted to equation (89.39) with solution given by
(89.41). We also obtain again (89.55) as the solution of Riccati’s equation.
B0
Equation (89.87) can be expressed in terms of F = as
B
1 1 ω2 2ω
F 00 − F 3 + ω F 0 − F 2 + a − F − 8KF e− 3 ξ
2 2 6
8K − 2ω ξ ω
− ωe 3 − (18a − ω 2 ) = 0
3 54
1214
The solution (89.41) is to be substituted in the above equation to get a com-
patibility condition between the parameters and the constants. We obtain
as in Case I, the same set of compatibility conditions. Therefore we arrive
at the following
Solution IV. The solution of equation (89.12)
U 00 (ξ) + ωU 0 (ξ) + aU (ξ) − bU 3 (ξ) = 0

is given by (89.13), U (ξ) = a0 + a1 Y + bY1 where
r r
1 2 w0 (ξ) 2 1
a0 (ξ) = − ω−6 , a1 = −C
6 b w(ξ) b w2 (ξ)
r y 0 (ξ) w0 (ξ)
K 2 2 2ω 1
b1 = − w (ξ)exp − ξ and Y = − w2 (ξ) −
C b 3 C y(ξ) w(ξ)
with w(ξ) and y(ξ) being expressed in terms of Bessel’s functions by (89.41)
and (89.54) respectively. Therefore
r
1 2h y 0 (ξ) y 0 (ξ) w0 (ξ) −1 2ω i
U (ξ) = −ω + 6 + 6K − × exp − ξ
6 b y(ξ) y(ξ) w(ξ) 3
0 (ξ) ω
In the above solution we have to take into account the condition ww(ξ) = ,
6
according to (89.99) of Appendix A. We thus have the following solution ex-
pressed by Theorem IV:
Theorem IV. The traveling wave solutions of the Newell-Whitehead equa-
tion ut = uxx + au − bu3 , derived under the substitution ξ = x − ωt, i.e. the
r
1 2h y 0 (ξ) y 0 (ξ) ω −1 2ω i
U (ξ) = −ω + 6 + 6K − × exp − ξ (89.88)
6 b y(ξ) y(ξ) 6 3
The above solution is subject to the constraints

1 9a
ν2 = , ω2 = (89.89)
4 2
and
3√2K 3√2K 3√2K 3√2K
Jν × Yν+1 − Jν+1 × Yν = 0 (89.90)
ω ω ω ω
1215
We list below the derivative of the function y(ξ) which appears in Solutions
I and IV. Using the general formulas for the derivatives of Bessel’s functions,
we obtain
√ h 3√3K ω 3√3K ω i
0 −ω −
ξ
y (ξ) = 3Ke 3 C̃1 Jν ξ
e 3 + C̃2 Yν e− 3 ξ
ω ω
1 h 3√3K ω 3√3K ω i (89.91)
−3ξ −3ξ
− ων C̃1 Jν e + C̃2 Yν e
3 ω ω
Appendix A. In this Appendix we consider the issue of expressing the
compatibility condition (89.47) in terms of the constants and the various
parameters. We first consider (89.47)
1 1 1 2ω
F 00 − F 3 + ω F 0 − F 2 + a − ω 2 F − 8Ke− 3 ξ F
2 2 6 (89.92)
8 2ω
− 3 ξ ω 2
− Kωe + (18a − ω ) = 0
3 54
We consider the above equation with (89.33):
1 2ω
ω2
F 0 − F 2 = 4Ke− 3 ξ − a − (89.93)
2 6
Substituting F 0 − 12 F 2 in (89.92) by the expression given in (89.93), we obtain
1 1 4 2ω 2ω
F 00 − F 3 + a − ω 2 F + Kωe− 3 ξ − 8Ke− 3 ξ F
2 6 3 (89.94)
2 2 2
+ ω ω −a =0
3 9
We multiply by F equation (89.93),
1 3 ω2
0 − 2ω ξ
− F = −F F + 4Ke 3 F+ −a F (89.95)
2 6
1
and we substitute − F 3 given by (89.95) into (89.94) and we obtain the
2
equation
4 2ω 2ω 2 2
F 00 − F F 0 + Kωe− 3 ξ − 4Ke− 3 ξ F + ω ω 2 − a = 0 (89.96)
3 3 9
Upon differentiation of (89.93) with respect to ξ, we find
8 2ω
F 00 − F F 0 = − Kωe− 3 ξ (89.97)
3
1216
ω 2ω 2 2 2
− 3 ξ
−4K +F e + ω ω −a =0 (89.98)
3 3 9
ω
The above equation should hold for every ξ. We thus have + F = 0 and
3
2 2
ω − a = 0, i.e.
9
ωw(ξ) − 6w0 (ξ) = 0 (89.99)
and
9a
ω2 = (89.100)
2
respectively, where w(ξ) is given by (89.41). Combining (89.100) with the
second of (89.38), we obtain
1
ν2 = (89.101)
4
Upon expanding (89.99) in a series around ξ = 0 (this is best facilitated using
any of the known Computer Algebra Systems) we find that (89.99) is true to
every order, provided that
3√2K 3√2K
C1 Jν + C 2 Yν =0 (89.102)
ω ω
and
3√2K 3√2K
C1 Jν+1 + C2 Yν+1 =0 (89.103)
ω ω
We thus have
√
√
3 2K 3 2K
C1 Yν
ω
Yν+1 ω
= − √ = − √ (89.104)
C2 Jν 3 ω2K Jν+1 3 ω2K
From the above relations we find that

3√2K 3√2K 3√2K 3√2K
Jν × Yν+1 − Jν+1 × Yν = 0 (89.105)
ω ω ω ω
1217
The above equation provides essentially a relation which determines the con-
stant K in terms of a (since ω is expressed through a by (A.9)). Considering
Case IV, we find the same compatibility conditions as in Case I.
Appendix B. In this Appendix we consider the issue of expressing the com-
patibility condition (89.67) in terms of the constants and the various param-
eters. We first consider (89.67)
1 1 ω2
G00 − G3 + ω G0 − G2 + a − G − 24s2 AA0 − 8ωs2 A2
2 2 6 (89.106)
ω 2
+ (18a − ω ) = 0
54
We similarly express (89.65) in terms of G:
0 1 2 2 2
ω2
G − G = 8s A − a − (89.107)
2 6
Upon substituting G0 − 21 G2 given by (89.107) into (89.106),we obtain
1 ω2 2ω 2ω 2
G00 − G3 + a − G − 24s2 AA0 + −a =0 (89.108)
2 6 3 9
From (89.107), multiplying by G, we find
1 ω2
− G3 + a − G = −GG0 + 8s2 A2 G (89.109)
2 6
Equation (89.108) then gives, because of (89.109),
2ω 2ω 2
G00 − GG0 + 8s2 A2 G − 24s2 AA0 + −a =0 (89.110)
3 9
Differentiating (89.107) with respect to ξ, we derive
G00 − GG0 = 16s2 AA0 (89.111)
Therefore, because of (89.111), equation (89.110) becomes
2 0 2 2 2 0 2ω 2ω 2
16s AA + 8s A G − 24s AA + −a =0
3 9
A0 2ω 2ω 2
which, because of G = , gives us the relation − a = 0 from
A 3 9
which we obtain
9a
ω2 = (89.112)
2
1218
Considering Case III, we find that (89.78) and (89.79) are compatible if
equation (89.112) holds true as well.
Appendix C. In this Appendix we list the solution methods and the corre-
sponding solutions of the equation (89.65)
3
A(ξ)A00 (ξ) − (A0 (ξ))2 + 2m2 A2 (ξ) − 8s2 A4 (ξ) = 0 (89.113)
2
C.I. First Method. We consider an expansion of the form
A(ξ) = a0 + a1 ϕ(ξ) (89.114)
where ϕ(ξ) satisfies Jacobi’s differential equation
d p
ϕ(ξ) = λ + µϕ2 + ρ2 ϕ4 (89.115)
dξ
where λ, µ and ρ are constant real parameters. Upon substituting (89.114)
into (89.113), taking into account (89.115) and equating to zero the coeffi-
cients of the different powers of ϕ, we can determine the coefficients a0 , a1
and the function ϕ(ξ). We find that the coefficients a0 and a1 are given by
the relations
ρ2 m2
a21 = , a20 = (89.116)
16s2 16s2
and the function ϕ(ξ) by
m C tanh(mξ) + 1
ϕ(ξ) = − · (89.117)
ρ C + tanh(mξ)
We thus have, using (89.114), the four values (89.116) and equation (89.117),
the following expressions for A(ξ)
ρ m m C tanh(mξ) + 1
A(ξ) = + − ·
4s 4s ρ C + tanh(mξ)
A(ξ) = − − ·
(89.118)
A(ξ) = − + − ·
A(ξ) = − − − ·
1219
C.II. Second Method.. We consider the (G0 /G)− expansion of the form
G0
A(ξ) = a0 + aa (89.119)
G
with a0 and a1 constants and G a function depending on ξ : G = G(ξ). Upon
substituting (89.119) into (89.113) and equating to zero the coefficients of the
1
different powers of G, and solving the resulting system, we get a21 = ,
16s2
2
m
a20 = 2 and
4s
a1
G = C1 + C2 ξ + C3 exp −4m2 ξ (89.120)
a0
Therefore

2 2 a1
G 0 D 1 − 4m a 1 exp −4m a0
ξ
=
G D + D ξ + a exp −4m ξ 2 a1
2 1 0 a0
with D1 and D2 constants. We thus have

2 a1

2
G0 h D1 − 4m a1 exp −4m ξ i
a0
A(ξ) = a0 + a1 = a0 + a1 a1
G D2 + D1 ξ + a0 exp −4m2 ξ
a0
The above expression gives us the following four expressions (corresponding
to the four combinations of signs for the coefficients a0 and a1 ):
m2
m 1 h D1 − exp(−2mξ) i
A(ξ) = + s
2s 4s D2 + D1 ξ + m exp(−2mξ)
2s
2
m
m 1 h D1 + s exp(2mξ) i
A(ξ) = −
2s 4s D2 + D1 ξ + m exp(2mξ)
2s (89.121)
m2
m 1 h D1 − s exp(2mξ) i
A(ξ) = − +
2s 4s D2 + D1 ξ − m exp(2mξ)
2s
2
m
m 1 h D1 + s exp(−2mξ) i
A(ξ) = − −
2s 4s D2 + D1 ξ − m exp(−2mξ)
2s
1220
C.III. Third Method. Using the transformation
1
A(ξ) = (89.122)
w2 (ξ)
equation (89.113) transforms into the Ermakov equation
w00 (ξ) − m2 w(ξ) = −4s2 w−3 (ξ) (89.123)
The solution of Ermakov’s equation is given by
(2mC2 ±C1 e2mξ )2 − 16m2 s2 −2mξ

w(ξ)2 = ×e
4m2 C1
We thus obtain, using the above expression and (89.122), that A(ξ) is given
by
4m2 C1 ·e2mξ
A(ξ) = (89.124)
(2mC2 ±C1 e2mξ )2 − 16m2 s2
The above expression can also be written in terms of tanh(mξ) as
4m2 C1 [1 − tanh2 (mξ)]

A(ξ) =
[(2mC2 ±C1 ) − (2mC2 ∓C1 )tanh(mξ)]2 − 16m2 s2 [1 − tanh(mξ)]2
C.IV. Fourth Method. This method is based on the Extended Projective
Riccati Equation method and has already applied for the equation (89.65) in
the previous Section.
References
Expansion Coefficients. I. Solving the Burgers Equation”.
Expansion Coefficients. II. Solving the KdV Equation”.
Expansion Coefficients. III. Solving the Newell-Whitehead Equation”.
Differ. Equ. Appl. Vol 7, No 1 (2015) 93-132. doi:10.7153/dea-07-07
Expansion Coefficients. IV. Solving the Fitzhugh-Nagumo Equation”.
1221
[5] S. M. Antoniou: ”Some General Solutions to the Duffing Equation”.
[6] S. M. Antoniou: ”Some General Solutions to the Painlevé-PII Equa-
tion”. SKEMSYS report available form ResearchGate
[7] S. M. Antoniou: ”Some General Solutions to the Painlevé-PII Equa-
tion. The Auxiliary Equation Method”.
[8] S. M. Antoniou: ”Some General Solutions to the Painlevé-PIV Equa-
tion”. SKEMSYS report available form ResearchGate
[9] S. M. Antoniou: ”Closed-form Solutions of a reduced form of the Gross-
Pitaevskii Equation”. SKEMSYS report available form ResearchGate.
90 The Abel Equation Method with

Variable Expansion Coefficients
This method was introduced by S. M. Antoniou (Antoniou [1]) to determining
solutions of the Van der Pol equation.
We consider a nonlinear ordinary differential equation
F (U, Ux0 , Uxx

00
, ...) = 0 (90.1)
where U is the unknown function and x is the independent variable. The

subscripts denote derivatives with respect to the independent variable x.
Equation (90.1) can be solved by considering expansions of the form
n
X
U (x) = ak Y k (90.2)
k=0
or
n n
X X
k bk
U (x) = ak Y + (90.3)
k=0 k=1
Yk
where all the expansion coefficients depend on the variable x,
ak ≡ak (x) and bk ≡bk (x) for every k = 0, 1, 2, 3, ...
1222
Expansion of the form (90.3) refers as the extended Abel equation method.
The function Y ≡Y (x) satisfies Abel’s equation
Y 0 = A + BY + CY 2 + DY 3 (90.4)
where again the coefficients A, B, C and D depend on the variable x.

In solving equation (90.1) we consider the expansions (90.2) or (90.3) and
then we balance the nonlinear term with the highest order derivative term of
the function U (x) which determines n (the number of the expansion terms).
Equating similar powers of the function Y (x) we can determine the various
coefficients and thus find the solution of the equation considered.
The Van der Pol equation. We consider Van der Pol’s differential equation
x00 + µ(x2 − λ)x0 + αx = 0 (90.5)
and apply the Abel equation method. We thus consider the expansion
n
X
x(t) = ak Y k
k=0
and balance the second order derivative term with the nonlinear term. We
then find that n = 1. The proof might go as follows:
the first derivative x0 (x) contains the highest order term Y 0 Y n−1 and upon
the substitution Y 0 →A + BY + CY 2 + DY 3 the highest order term becomes
Y n−1 Y 3 ≡Y n+2 . The second derivative x00 (t) will contain the highest order
term Y n+1 Y 0 and upon the substitution Y 0 →A + BY + CY 2 + DY 3 the
highest order term contains Y n+1 Y 3 ≡Y n+4 . The nonlinear term x2 (t)x0 (t)
contains the highest order term Y 2n Y n+2 ≡Y 3n+2 . Therefore balancing the
second derivative term x00 (t) with the nonlinear term x2 (t)x0 (t) leads to the
equation n + 4 = 3n + 2 from which we obtain n = 1. We thus have the
expansion
x(t) = a0 + a1 Y (90.6)
where again the coefficients a0 and a1 depend on t, and Y satisfies Abel’s

equation Y 0 = A+BY +CY 2 +DY 3 . From equation (90.6) we obtain (taking
into account Y 0 = A + BY + CY 2 + DY 3 )
x0 (t) = a00 + a01 Y + a1 (A + BY + CY 2 + DY 3 ) (90.7)
1223
x00 (t) = a000 + a001 Y + a01 (A + BY + CY 2 + DY 3 )+
+ a1 (A0 + B 0 Y + C 0 Y 2 + D0 Y 3 )+ (90.8)
+ a01 (B + 2CY + 3DY 2 )(A + BY + CY 2 + DY 3 )
Therefore equation (90.5), under the substitution (90.6), (90.7) and (90.8),
becomes
[a000 + a001 Y + a01 (A + BY + CY 2 + DY 3 )+
+ a1 (A0 + B 0 Y + C 0 Y 2 + D0 Y 3 )+
+ a01 (B + 2CY + 3DY 2 )(A + BY + CY 2 + DY 3 )] (90.9)
+ µ[(a0 + a1 Y )2 − λ][a00 + a01 Y + a1 (A + BY + CY 2 + DY 3 )]
+ α(a0 + a1 Y ) = 0
Upon expanding and equating the coefficients of Y to zero, we obtain a

system of differential equations from which we can determine the various
expansion coefficients a0 , a1 and the coefficients A, B, C, D of Abel’s equation.
We obtain:
Coefficient of Y 5 :
a1 D(µa21 + 3D) = 0 (90.10)
a1 (2µa0 a1 D + µa21 C + 5CD) = 0 (90.11)
a1 (4BD + 2C 2 + D0 ) + 2a01 D + µ(a20 − λ)a1 D+
(90.12)
+ 2µa0 a21 C + µa21 (a01 + a1 B) = 0
2a01 C + µ(a20 − λ)a1 C + 2µa0 a1 (a01 + a1 B)+
(90.13)
+ µa21 (a00 + a1 A) + a1 (3BC + C 0 + 3AD) = 0
Coefficient of Y :
2a01 B + a001 + a1 (B 0 + B 2 + 2AC) + µ(a20 − λ)(a01 + a1 B)
(90.14)
+ 2µa0 a1 (a00 + a1 A) + αa1 = 0
1224
a000 + αa0 + 2a01 A + µ(a20 − λ)(a00 + a1 A) + a1 (A0 + AB) = 0 (90.15)
We now have to solve the system of ordinary differential equations (90.10)-

(90.15), supplemented by Abel’s equation
Y 0 = A + BY + CY 2 + DY 3 (90.16)
From equation (90.10), ignoring the trivial solutions, i.e. supposing a1 D6=0,
we obtain
µa21 = −3D (90.17)
Using the above relation, we obtain from (90.11)
µa0 a1 = −C (90.18)
Because of (90.17) and (90.18), equation (90.12) takes on the form
a1 (BD − λµD + 2C 2 + D0 ) − Da01 − 7a0 CD = 0 (90.19)
We also get from (90.13) using again (90.17) and (90.18)
3Da00 + a0 C 2 + a1 (λµ − B)C − a1 C 0 = 0 (90.20)
a001 + [2B + µ(a20 − λ)]a01 + (B 2 + B 0 − λµB + α)a1

(90.21)
− 2Ca00 − BCa0 = 0
From (90.15), using (90.17), (90.18) and the relation

3D
a1 = a0 (90.22)
C
which comes dividing (90.17) by (90.18), we get the equation
D 0 D
a000 + µ(a20 − λ)a00 + 6 a0 + a00 A+
C C (90.23)
AD 3D 0 3ABD
+ α − AC − 3λµ + A + a0 = 0
C C C
1225
Using either (90.19) or (90.20) and (90.22), we obtain from either one of
them, the following relation
C2 C0
a00 = −λµ + B − + a0 (90.24)
3D C
From (90.17) and (90.18) we obtain a20 = −C 2 /(3µD) and then
C
a0 = ± √ (90.25)
−3µD
where we suppose that µD < 0.
From (90.25), with either choice of signs, we get
C0 1 D0
a00 = − a0 (90.26)
C 2D
Comparison of (90.24) and (90.26) results into the equation
C2 1 D0
λµ + =B+ (90.27)
3D 2D
Equation (90.23), because of a20 = −C 2 /(3µD) and (90.27), takes on the new
form
00
AD 1 D0 0
a0 + 6 −B− a
C 2D 0 (90.28)
D 0 AD 3D 0 3ABD
+ α+6 A − AC − 3λµ + A + a0 = 0
C C C C
Since there are two values of a0 according to (90.25), we obtain from (90.21)
the following two equations:
For
C
a0 = √
−3µD
we have
C2 0
a001 + 2B − λµ − a + (B 2 + B 0 − λµB + α)a1
3D 1
(90.29)
C2 C 0 D0
−√ B+2 − =0
−3µD C D
1226
For
C
a0 = − √
−3µD
we have
C2 0
a001 + 2B − λµ − a + (B 2 + B 0 − λµB + α)a1
3D 1
(90.30)
C2 C 0 D0
+√ B+2 − =0
−3µD C D
respectively.
Since
C 3D
a0 = ± √ and a1 = ± √
−3µD −3µD
according to (90.25) and (90.22), we distinguish two cases:

Case I. For
C 3D
a0 = √ and a1 = √
−3µD −3µD
we obtain from (90.28) and (90.29) the equations
D00 C 00 C 0 D0 D 0 2 AD C 0 D0
−2 +3 −2 + B−6 2 −
D C C D D C C D (90.31)
D 0 AD 3D 0 3ABD
−2 6 A − AC − 3λµ + A + − 2α = 0
C C C C
and
D00 0
C 0 D0 C 0 D0
+ 2B − 2B 2 − + 2λµ B + 2 −
D C D C D (90.32)
C 0 D0
−2 + 2(α − λµ) = 0
C D
respectively.
Case II. For
C 3D
a0 = − √ and a1 = − √
−3µD −3µD
1227
we obtain from (90.28) and (90.30) the same set of equations (90.31) and
(90.32) as above.
In the following we shall consider the case A = 0 in order to solve the system
of equations.
Conclusion. Van der Pol’s equation
x00 + µ(x2 − λ)x0 + αx = 0, x = x(t)
admits solutions of the form
x(t) = a0 + a1 Y
where a0 = a0 (t) and a1 = a1 (t) are time-dependent coefficients and Y = Y (t)

is an auxiliary function satisfying Abel’s equation of the first kind
Y 0 = BY + CY 2 + DY 3
where B, C and D are time-dependent coefficients: B = B(t), C = C(t) and

D = D(t) and satisfy the system of equations (90.27), (90.32) and (90.31) for
A = 0. The coefficients a0 = a0 (t) and a1 = a1 (t) are found to be determined
by
C 3D
a0 = ± √ and a1 = ± √
−3µD −3µD
respectively. We next have to solve the system of equations (90.27), (90.31)

and (90.32). For the rest of the calculations the reader should consult the
following Reference.
References
[1] S. M. Antoniou: ”Some General Solutions to the Van der Pol Equation.
II. The Abel Equation Method”.
SKEMSYS report available online form ResearchGate.
91 The Jacobi Elliptic Equation Method I

Let
F (x, u, u0 , u00 , · · · ) = 0 (91.1)
1228
be a nonlinear ODE where x is the independent variable and u = u(x) is the
unknown function. The Jacobi elliptic equation method consists in expanding
the unknown function u in a power series
n
X sn0 (x) k n
X sn0 (x) −k
u= ak + bk (91.2)
k=0
sn(x) k=0
sn(x)
n
X cn0 (x) k n
X cn0 (x) −k
u= ak + bk (91.3)
k=0
cn(x) k=0
cn(x)
n
X dn0 (x) k n
X dn0 (x) −k
u= ak + bk (91.4)
k=0
dn(x) k=0
dn(x)
where sn(x; m), cn(x; m) and dn(x; m) are Jacobi’s elliptic functions, called
sine, cosine and third kind elliptic functions respectively. These functions
obey the identities
sn2 (x) + cn2 (x) = 1, dn2 (x) + m2 sn2 (x) = 1 (91.5)
The derivatives are given by
sn0 (x) = cn(x) dn(x), cn0 (x) = −sn(x) dn(x),

(91.6)
dn0 (x) = −m2 sn(x) cn(x)
We also have the following useful relations
sn00 (x)
= 2m2 sn2 (x) − m2 − 1
sn(x)
cn00 (x)
= 2m2 sn2 (x) − 1 (91.7)
cn(x)
dn00 (x)
= 2m2 sn2 (x) − m2
dn(x)
We first determine n by balancing the higher order derivative to the highest

order nonlinear term and then we substitute one of the expansions to the orig-
inal equation. At each stage of differentiation we have to take into account
the relations (91.7). After that, we eliminate the denominators and thus we
and up with an equation which contains the products snk (x)cnm (x)dnn (x).
Equating to zero the coefficients of all those products, we obtain a system of
1229
equations from which we determine the various coefficients ak and bk .
Example 1. We wish to find the traveling wave solutions of the generalized
shallow water wave equation
αux uxt + βut uxx − uxt − uxx + uxxxt = 0 (91.8)
u = U (ξ), ξ = k(x − ωt) (91.9)
−ωk 2 U 000 (ξ) + ω(β − α)U (ξ)U 0 (ξ) + (ω − 1)U 0 (ξ) = 0 (91.10)
Integrating the previous equation and setting the integration constant equal
to zero, we obtain the equation
1
−ωk 2 U 00 (ξ) + ω(β − α)U 2 (ξ) + (ω − 1)U (ξ) = 0 (91.11)
2
Balancing U 00 to U 2 , i.e. n + 2 = 2n, we find n = 2. We consider the first
expansion (91.2) for n = 2:
sn0 (ξ) sn0 (ξ) 2
U = a0 + a1 + a2
sn(ξ) sn(ξ)
sn0 (ξ) −1 sn0 (ξ) −2 (91.12)
+ b1 + b2
sn(ξ) sn(ξ)
The first derivative of the above expression is given by

00
sn (ξ) sn0 (ξ) 2

0
U (ξ) = a1 −
sn(ξ) sn(ξ)
00
sn (ξ) sn0 (ξ) 2 sn0 (ξ)

+ 2a2 −
sn(ξ) sn(ξ) sn(ξ)
00 0
0 (91.13)
sn (ξ) sn (ξ) 2
sn (ξ) −2
− b1 −
00 0
0
sn (ξ) sn (ξ) 2
sn (ξ) −3
− 2b2 −
1230
In the above expression we substitute sn00 (ξ)/sn(ξ) by 2m2 sn2 (ξ) − m2 − 1
(according to the first of (91.7)) and we thus get
sn0 (ξ) 2
0 2 2 2
U (ξ) = 2m sn (ξ) − m − 1 −
sn(ξ)
sn0 (ξ) sn0 (ξ) −2 sn0 (ξ) −3 (91.14)
× a1 + 2a2 − b1 − 2b2
Differentiation of the above expression yields
sn0 (ξ) sn00 (ξ) sn0 (ξ) 2
00 2 0
U (ξ) = 4m sn(ξ)sn (ξ) − 2 −
sn0 (ξ) sn0 (ξ) −2 sn0 (ξ) −3
× a1 + 2a2 − b1 − 2b2
sn0 (ξ) 2
+ 2m2 sn2 (ξ) − m2 − 1 − (91.15)
sn(ξ)
sn0 (ξ) −3 sn0 (ξ) −4
× 2a2 + 2b1 + 6b2
sn(ξ) sn(ξ)
00 0

sn (ξ) sn (ξ) 2
× −
sn(ξ) sn(ξ)
In the above expression we substitute sn00 (ξ)/sn(ξ) by 2m2 sn2 (ξ) − m2 − 1.
We thus obtain
U 00 (ξ)
= 4m2 sn(ξ)sn0 (ξ)
sn0 (ξ) sn0 (ξ) −2 sn0 (ξ) −3
× a1 + 2a2 − b1 − 2b2
sn0 (ξ) sn0 (ξ) 2
−2 2m2 sn2 (ξ) − m2 − 1 −
sn(ξ) sn(ξ)
sn0 (ξ) sn0 (ξ) −2 sn0 (ξ) −3
× a1 + 2a2 − b1 − 2b2 (91.16)
sn0 (ξ) 2
+ 2m2 sn2 (ξ) − m2 − 1 −
sn(ξ)
sn0 (ξ) −3 sn0 (ξ) −4
× 2a2 + 2b1 + 6b2
sn(ξ) sn(ξ)
sn0 (ξ) 2
× 2m2 sn2 (ξ) − m2 − 1 −
sn(ξ)
1231
We finally substitute the derivative sn0 (ξ) by cn(ξ) dn(ξ) according to the
first of (91.6). The final expression for the second derivative U 00 (ξ) is thus
given by
U 00 (ξ)
= 4m2 sn(ξ) cn(ξ) dn(ξ)
cn(ξ) dn(ξ) cn(ξ) dn(ξ) −2
× a1 + 2a2 − b1
sn(ξ) sn(ξ)
cn(ξ) dn(ξ) −3
− 2b2
sn(ξ)
cn(ξ) dn(ξ) cn(ξ) dn(ξ) 2
2 2 2
−2 2m sn (ξ) − m − 1 −
sn(ξ) sn(ξ)
cn(ξ) dn(ξ) cn(ξ) dn(ξ) −2
× a1 + 2a2 − b1 (91.17)
sn(ξ) sn(ξ)
cn(ξ) dn(ξ) −3
− 2b2
sn(ξ)
cn(ξ) dn(ξ) 2
2 2 2
+ 2m sn (ξ) − m − 1 −
sn(ξ)
cn(ξ) dn(ξ) −3 cn(ξ) dn(ξ) −4
× 2a2 + 2b1 + 6b2
sn(ξ) sn(ξ)
cn(ξ) dn(ξ) 2
× 2m2 sn2 (ξ) − m2 − 1 −
sn(ξ)
We now have to substitute in (91.11) the second derivative U 00 (ξ) by the

above expression while for the function U (ξ) we have to use the expression
cn(ξ) dn(ξ) cn(ξ) dn(ξ) 2
U = a0 + a1 + a2
sn(ξ) sn(ξ)
cn(ξ) dn(ξ) −1 cn(ξ) dn(ξ) −2 (91.18)
+ b1 + b2
sn(ξ) sn(ξ)
which comes from (91.12) after substitution of sn0 (ξ) by cn(ξ) dn(ξ).
After these substitutions in (91.11) we have to multiply through by
cn(ξ) dn(ξ) 4
sn(ξ)
1232
to eliminate coefficients. We will then derive an equation which will contain
an expression where each term has the form snk (ξ)cnm (ξ)dnp (ξ). Equating
to zero the coefficients of each one of these products we get a system of
simultaneous equations from which we determine the coefficients. All these
calculations have to be performed by one of the leading Symbolic Languages,
like Macsyma, Maple, Mathematica, MATLAB or Axiom. The complete
set of solutions is given in the reference paper. We only provide one of the
solutions
p
1 − 48ωk 2 − ω + 2 768ω 2 k 4 + (ω − 1)2
a0 = −
ω(α − β)
a1 = a2 = b 1 = 0

2 2
p 2
2 4 2
64ωk − 28ωk − 768ω k + (ω − 1) − (ω − 1) (91.19)
3
b2 = ·
4ω 2 k 2 α−β
s p
−28ωk 2 − 768ω 2 k 4 + (ω − 1)2
m=±
4ωk 2
The above solution has to be substituted in (91.12).

References
[1] Y. Xiao, X. Xue and H. Zhang: ”A New Extended Jacobi Elliptic
Function Expansion Method and its Applications in the Generalized Shallow
Water Wave Equation”. J. Appl. Math. Article ID 896748 (2012).
92 The Jacobi Elliptic Equation Method II

Let
F (x, u, u0 , u00 , · · · ) = 0 (92.1)
be a nonlinear ODE where x is the independent variable and u = u(x) is

the unknown function. The Jacobi elliptic equation method (second version)
consists in expanding the unknown function u in a power series
n
X n
X
u= ak Φk + bk Φ−k (92.2)
k=0 k=0
1233
where Φ is a function which satisfies the differential equation
β γ
(Φ0 (x)) = ρ + αΦ2 (x) + Φ4 (x) + Φ6 (x) (92.3)
2 3
Example. We consider a system of the two component BKP hierarchy
ut − uxxx − uyyy − 6(ux v + uvx + uy w + uwy ) = 0
vy = ux (92.4)
w x = uy
where
u = u(x, y, t), v = v(x, y, t), w = w(x, y, t) (92.5)
u(x, y, t) = U (ξ), v(x, y, t) = V (ξ), w(x, y, t) = W (ξ)
(92.6)
ξ = kx + λy + νt + δ
the system (92.4) takes on the form
νU 0 − (k 3 + λ3 )U 000 − 6{k(U V )0 + λ(U W )0 } = 0
λV 0 − kU 0 = 0 (92.7)
kW 0 − λU 0 = 0
Balancing U 000 to U V 0 , i.e. n + 6 = n + (n + 2), we find n = 4. We thus
consider the expansions
U = a0 + a1 Φ(ξ) + a2 Φ2 (ξ) + a3 Φ3 (ξ) + a4 Φ4 (ξ) (92.8)
V = b0 + b1 Φ(ξ) + b2 Φ2 (ξ) + b3 Φ3 (ξ) + b4 Φ4 (ξ) (92.9)
W = c0 + c1 Φ(ξ) + c2 Φ2 (ξ) + c3 Φ3 (ξ) + c4 Φ4 (ξ) (92.10)
In the above expansions we have not considered the negative exponents ex-
pansions for simplicity.
Differentiation of (92.8)-(92.10) and taking into account (92.3), we convert
the system (92.7) into a system of equations which contains equations with
powers of the function Φ. Equating the coefficients to zero, we get a system
of equations from which we determine the various constant coefficients.
References
[1] M. K. Elboree: ”The Jacobi Elliptic Function Method and its Appli-
cation for Two Component BKP Hierarchy Equation”.
Comp. Math. Appl. 62 (2011) 4402-4414
1234
93 The Weierstrass Equation Method
Let
F (x, u, u0 , u00 , · · · ) = 0 (93.1)

unknown function. The Weirstrass equation method consists in expanding
the unknown function u in a power series
n
X n
X n
X n
X
u= ak Φk + bk Φ−k + ck Φk−1 Φ0 + dk Φ−k Φ0 (93.2)
k=0 k=0 k=0 k=0
where Φ is a function which satisfies the equation

p
Φ0 (x) = a + b Φ2 (x) + c Φ4 (x) (93.3)
The solutions of the above equation are expressed in terms of Weirstrass

function ℘(x; g2 , g3 ) which is a solution of the differential equation
(℘0 (x))2 = 4℘3 (x) − g2 ℘(x) − g3 (93.4)
where g2 , g3 are the so-called invariants of the equation. The second deriva-
tive is given by
1
℘00 (x) = 6℘3 (x) − g2 (93.5)
2
Equation (93.3) admits the following solutions, expressed in terms of Weirstrass
function:
Solution 1.
r
1 b
Φ= ℘(x; g2 , g3 ) − (93.6)
c 3
Solution 2.
s
3a
Φ=
3℘(x; g2 , g3 ) − b
(93.7)
4b2 − 12ac 4b(9ac − 2b2 )
g2 = , g3 =
3 27
1235
Solution 3.
s
12a ℘(x; g2 , g3 ) + 2a(2b + ∆1,2 )
Φ=
12℘(x; g2 , g3 ) + ∆1,2
√
−5b± 9b2 − 36ac
∆1,2 = , (93.8)
2
1
g2 = − (5b∆1,2 + 4b2 + 33abc),
12
1
g3 = − (−21b2 ∆1,2 − 20b3 + 63ac∆1,2 + 27abc)
216
Solution 4.
√
(6 ℘(x; g2 , g3 ) + b) a
Φ= (93.9)
3 ℘0 (x; g2 , g3 )
Solution 5.
3 ℘0 (x; g2 , g3 )
Φ= √
c 6℘(x; g2 , g3 ) + b (93.10)
1 1
g2 = (b2 + 12ac), g3 = (36abc − b3 )
12 216
Solution 6.
r
15b ℘(x; g2 , g3 )
Φ= −
2c 3℘(x; g2 , g3 ) + b (93.11)
2
5b 2b2 b3
a= , g2 = , g3 =
36c 9 54
Example 1. We consider the (2 + 1)−dimensional modified Kadomchev-
Petviashvili equation (mKP)
(ut + 3λu2 ux + uxxx )x + uyy = 0 (93.12)
u = U (ξ), ξ = αx + βy − ωt (93.13)
− ωαU 00 (ξ) + 3λ {2U (ξ)(αU 0 (ξ))2 + α2 U 2 (ξ)U 00 (ξ)}
+ α4 U 0000 (ξ) + β 2 U 00 (ξ) = 0
1236
− ωαU 00 + 3λα2 {2U (U 0 )2 + U 2 U 00 } + α4 U 0000 + β 2 U 00 = 0
or, what is equivalent,
− ωαU 00 + 3λα2 (U 2 U 0 )0 + α4 U 0000 + β 2 U 00 = 0
Integrating the above equation, we get (setting the integration constant equal
to zero)
−ωαU 0 + 3λα2 (U 2 U 0 ) + α4 U 000 + β 2 U 0 = 0
Integrating once again we obtain the equation
α4 U 00 + (β 2 − ωα)U + λα2 U 3 = 0 (93.14)

Balancing U 00 to U 3 , i.e. n + 2 = 3n we find n = 1. We thus consider the
expansion
b1 Φ0 (ξ)
U = a0 + a1 Φ(ξ) + + c1 Φ0 (ξ) + d1 (93.15)
Φ(ξ) Φ(ξ)
We find the first derivative U 0 (ξ) taking into account (93.3). We then differ-
entiate again to find the second derivative U 00 (ξ) taking into account (93.3).
We finally substitute the resulting expressions to (93.14) and substituting
the various coefficients of Φ(ξ) to zero, we get a system of equations from
which we determine a0 , a1 , b1 , c1 , d1 . The values of these coefficients have to
be substituted in (93.15) and (93.3).
Example 2. We consider the KdV equation
ut = 6uux + uxxx (93.16)
u = U (ξ), ξ = k(x − ωt) (93.17)
(−ωk)U 0 (ξ) = 6kU (ξ)U 0 (ξ) + k 3 U 000 (ξ)
1237
k 2 U 000 (ξ) + ωU 0 (ξ) + 6U (ξ)U 0 (ξ) = 0
k 2 U 00 (ξ) + ωU (ξ) + 3U 2 (ξ) = 0 (93.18)
Balancing U 00 to U 3 , i.e. n + 2 = 2n, we find n = 2. We thus consider the

expansion
b1 b2
U = a0 + a1 Φ(ξ) + a2 Φ2 (ξ) + + 2
Φ(ξ) Φ (ξ)
(93.19)
Φ0 (ξ) Φ0 (ξ)
+ c1 Φ0 (ξ) + c2 Φ(ξ)Φ0 (ξ) + d1 + d2 2
Φ(ξ) Φ (ξ)
We find the first derivative U 0 (ξ) taking into account (93.3). We then differ-
entiate again to find the second derivative U 00 (ξ) taking into account (93.3).
We finally substitute the resulting expressions to (93.18) and substituting
the various coefficients of Φ(ξ) to zero, we get a system of equations from
which we determine a0 , a1 , a2 , b1 , b2 , c1 , c2 , d1 , d2 . The values of these coeffi-
cients have to be substituted in (93.19) and (93.3).
References
[1] E. A. Saied, R. G. A. El-Rahman and M. I. Ghonamy: ”A general-
ized Weirstrass elliptic function expansion method for solving some nonlinear
partial differential equations”. Comp. Math. Appl. 58 (2009) 1725-1735
94 The Homotopy Method

Let Ω be a region in n−dimensional Euclidean space and Γ its boundary. Let
also a nonlinear differential equation expressed in operator form as
A(u) = Φ(x), x∈Ω (94.1)
supplied with a boundary condition

∂u
B u, = 0, x∈Γ (94.2)
∂n
Here u is the unknown function and x is the independent variable. Φ(x) is a
smooth function.
1238
We shall usually work in 2−dimensions where u will be the unknown function
of an evolution equation depending on the variables x and t.
The operator A can be considered of consisting of two parts, one linear
denoted by L and another nonlinear denoted by N . Therefore equation
(94.1) will be written as
L(u) + N (u) = Φ(x) (94.3)
We consider a homotopy
v(x, p) : Ω×[0, 1]−→R (94.4)
with
H(v, p) = (1 − p)[L(v) − L(u0 )] + p[A(v) − Φ(x)] = 0, p∈[0, 1] (94.5)
where v is an auxilliary variable and u0 is an initial guess of a solution of

equation (94.3), not as yet specified. The expression (94.5) can be written
in equivalent form as
L(v) − L(u0 ) + pL(u0 ) + p[N (v) − Φ(x)] = 0 (94.6)
The auxiliary function v is chosen as
v = v0 + pv1 + p2 v2 + p3 v3 + · · · (94.7)
such that
u = lim v = v0 + v1 + v2 + v3 + · · · (94.8)
p→1
Example. We consider the Fisher equation

∂u ∂ 2u
= + au(1 − u), u(x, 0) = φ(x) (94.9)
∂t ∂x2
The operators L and N are thus given by
∂v ∂ 2v
L(v) = , N (v) = − + av(1 − v) (94.10)
∂t ∂x2
∂v ∂u0 ∂ 2v ∂u0
2
− =p + av − av − (94.11)
∂t ∂t ∂x2 ∂t
1239
After plugging the expansion (94.7) into (94.11), we get the equation
∂v ∂v1 ∂v2 ∂v3 ∂u
0 0
+p + p2 + p3 + ··· −
∂t ∂t ∂t ∂t ∂t
∂ 2v ∂ 2
v ∂ 2
v ∂ 2
v
0 1 2 2 3 3
=p + p + p + p + · · ·
∂x2 ∂x2 ∂x2 ∂x2 (94.12)
2 2 2
+ pa{(v0 − v0 ) + p(v1 − 2v0 v1 ) + p (v2 − 2v0 v2 − v1 ) + · · · }
∂u0
−p
∂t
Equating the coefficients of similar powers of p, we obtain the following sys-
tem of equations
∂v0 ∂u0
p0 : − =0 (94.13)
∂t ∂t
∂v1 ∂ 2 v0 ∂u0
p1 : = 2
+ a(v0 − v02 ) − (94.14)
∂t ∂x ∂t
2
∂v2 ∂ v1
p2 : = + a(v1 − 2v0 v1 ) (94.15)
∂t ∂x2
∂v3 ∂ 2 v2
p3 : = + a(v2 − 2v0 v2 − v12 ) (94.16)
∂t ∂x2
and in general
k
k+1 ∂vk+1 ∂ 2 vk X
p : = + a vk − vk vk−j (94.17)
∂t ∂x2 j=0
At this stage we have to take into account the initial condition. Taking
u0 = u(x, 0) = φ(x) (94.18)
we obtain from (94.13) that
v0 = φ(x) (94.19)
∂v1
= φ00 (x) + a[φ(x) − φ2 (x)] (94.20)
∂t
1240
and by integration on [0, t] we obtain
v1 = {φ00 (x) + a[φ(x) − φ2 (x)]}t (94.21)
Equation (94.15) then takes on the form

∂v2
= {φ(iv) (x) + a[φ00 (x) − 2φ(x)φ00 (x) − 2(φ0 (x))2 ]}t
∂t (94.22)
+ a(1 − 2φ(x)){φ00 (x) + a[φ(x) − φ2 (x)]}t
and by integration on [0, t] we obtain

1
v2 = {φ(iv) (x) + a[φ00 (x) − 2φ(x)φ00 (x) − 2(φ0 (x))2 ]}t2
2 (94.23)
1
+ a(1 − 2φ(x)){φ00 (x) + a[φ(x) − φ2 (x)]}t2
2
Continuing in this way, we can determine all the successive functions vk
to the degree of approximation we like. On the other hand, the solution,
approximate or exact will depend on the initial guess φ(x).
The solution u then of the original equation will be given by (94.8):
u = v0 + v1 + v2 + v3 + · · · (94.24)
For the given equation we shall have

u = φ(x) + {φ00 (x) + a[φ(x) − φ2 (x)]}t
1
+ {φ(iv) (x) + a[φ00 (x) − 2φ(x)φ00 (x) − 2(φ0 (x))2 ]}t2
2 (94.25)
1
+ a(1 − 2φ(x)){φ00 (x) + a[φ(x) − φ2 (x)]}t2
2
+ ···
The reader can try any choice of the function φ(x) and then determine the
corresponding solution of Fisher’s equation.
95 The F-expansion Method

Let
E(x, u, u0 , u00 , · · · ) = 0 (95.1)
1241
be a nonlinear ODE where x is the independent variable and u = u(x) is
the unknown function. The F −expansion method consists in expanding the
unknown function u in a power series
n n
X X
k bk
u= ak F + (95.2)
k=0 k=1
Fk
where the function F satisfies the differential equation
(F 0 )2 = P F 4 + Q F 2 + R (95.3)
The expansion coefficients ak , bk and P, Q, R are all considered to be constant

quantities. Equation (95.3) admits 52 solutions and they are listed in Table 1
of Reference [3]. We then have to balance the highest order derivative to the
maximum power of the nonlinear term in order to determine the maximum
number n of the expansion terms in (95.2).
After determining n, we substitute the expansion (95.2) into the original
equation (95.1) and in each stage of differentiation we substitute F 0 by
p
F0 = P F4 + Q F2 + R (95.4)
according to (95.3). At the end, the original equation will be an expression

containing powers of F (x). Equating to zero all the coefficients of F we
obtain a system of equations from which we can determine all the constant
coefficients ak , bk and P, Q, R.
Note 1. Intead of considering the expansion (95.2) we can consider the
expansion
n
X n
X
u= ak F k + bk F k−1 F 0 (95.5)
k=0 k=0
This expansion usually refers as the extended F −expansion method.

Note 2. Intead of considering equation (95.3) we can consider the equation
(F 0 )2 = P F 2 + Q F 3 + R F 4 (95.6)
The solutions of the above equation are listed in Table 1 of Ref. [4].
Note 3. The solutions of the general equation
(y 0 )2 = A + By + Cy 2 + Dy 3 + Ey 4 (95.7)
1242
are listed in §11, Chapter 7 of H. T. Davis (Ref. [5]).
Example. We consider the reduced Ostrowski equation
u2 ut − ux uxt + uuxxt = 0, u = u(x, t) (95.8)
We shall determine the traveling wave solutions of this equation. Introducing

similarity variables ξ and U by
u(x, t) = U (ξ), ξ = x − ct (95.9)
−U 2 U 0 + U 0 U 00 − U U 000 = 0 (95.10)
Integrating once, we get the equation (taking the integration constant equal
to zero)
1
U U 00 − (U 0 )2 + U 3 = 0 (95.11)
3
Balancing U U 00 and U 3 , i.e. n + (n + 2) = 3n, we find n = 2 and we thus
have to consider the expansion
b1 b2
U = a0 + a1 F + a2 F 2 + + 2 (95.12)
F F
where F satisfies equation (95.3).
We calculate the first and second derivatives of the above expression (taking
into account the derivative of F 0 given by (95.4) in each stage of differenti-
ation) and then we substitute these expressions into (95.11). Collecting the
terms with respect to powers of F , we obtain a system of thirteen equations
which contain the coefficients a0 , a1 , a2 , b1 , b2 and P, Q, R. Solving the sys-
tem, we obtain the following set of non-trivial solutions
Solution I
a0 = 0, a1 = 0, a2 = 0, b1 = 0, b2 = b2
1 (95.13)
P = 0, Q = Q, R = − b2
6
Therefore the solution of (95.11) is given by
b2 1
U= , where (F 0 )2 = QF 2 − b2 (95.14)
F2 6
1243
Solution II
a0 = 0, a1 = 0, a2 = a2 , b1 = 0, b2 = 0
1 (95.15)
P = − a2 , Q = Q, R = 0
6
1
U = a2 F 2 , where (F 0 )2 = − a2 F 4 + QF 2 (95.16)
6
Solution III
a0 = a0 , a1 = 0, a2 = 0, b1 = 0, b2 = b2
1 a2 1 1 (95.17)
P = − 0 , Q = − a0 , R = − b 2
6 b2 3 6
b2 1 a20 4 1 1
U = a0 + , where (F 0 )2 = − F − a0 F 2 − b 2 (95.18)
F2 6 b2 3 6
Solution IV
a0 = a0 , a1 = 0, a2 = a2 , b1 = 0, b2 = 0
1 1 1 a2 (95.19)
P = − a2 , Q = − a0 , R = − 0
6 3 6 a2
1 1 1 a20
U = a0 + a2 F 2 , where (F 0 )2 = − a2 F 4 − a0 F 2 − (95.20)
6 3 6 a2
Solution V
1 a20
a0 = a0 , a1 = 0, a2 = , b1 = 0, b2 = b2
4 b2
(95.21)
1 a20 1 1
P =− , Q = − a0 , R = − b 2
24 b2 6 6
1 a20 2 b2
U = a0 + F + 2 , where
4 b2 F
2
(95.22)
1 a0 4 1 1
(F 0 )2 = − F − a0 F 2 − b 2
24 b2 6 6
1244
In all the above solutions (95.14), (95.16), (95.18), (95.20) and (95.22), we
have to adjust the coefficients to the solutions given in Table 1 of Ref.[3] or
to consider the solutions given in §11 of Chapter 7, Ref. [5]. This is left to
the reader. All the above calculations have been performed by MAPLE.
References
[1] M. A. Abdou: ”The Extended F-Expansion Method and its Applica-
tion for a class of Nonlinear Evolution Equations”.
Chaos, Solitons and Fractals 31 (2007) 95-104
[2] Y-M. Zhao: ”F-Expansion Method and Its Application for Finding New
Exact Solutions to the Kudryashov-Sinelshchikov Equation”.
J. Appl. Math. Vol 2013 Article ID 895760
[3] A. Filiz, M. Ekici and A. Sonmezoglou: ”F-Expansion Method and
New Exact Solutions of the Schrödinger-KdV Equation”.
The Scientific World Journal Vol 2014 Article ID 534063
[4] F. Kangalgil and F. Ayaz: ”New Exact Travelling Wave Solutions for
the Ostrovsky Equation”. Phys. Lett. A 372 (2008) 1831-1835
[5] H. T. Davis: ”Introduction to Nonlinear Differential and Integral Equa-
tions”. Dover 1962
96 The Adomian Decomposition Method

The method will become clear considering the example of a linear PDE.
We consider the heat equation
ut = uxx , 0 < x < a, t≥0 (96.1)
subject to the initial condition
u(x, 0) = φ(x) (96.2)
The original equation can be written as
∂ ∂2
Lt u = Lx u, where Lt = , Lx = (96.3)
∂t ∂x2
Inverting Lt and taking into account the initial condition, we get the formal
solution

u = u(x, 0) + L−1
t Lx u(x, t) (96.4)
1245
It is obvious that
Z t
−1
Lt () = ()dt (96.5)
0
and
Z xZ x
L−1
x () = ()dxdx (96.6)
0 0
At this point we formally expand u(x, t) into a series

∞
X
u(x, t) = uk (x, t) (96.7)
k=0
or in expanded form
u(x, t) = u0 + u1 + u2 + · · · , uk = uk (x, t) (96.8)

∞
X X ∞
uk = u(x, 0) + L−1
t L x uk (96.9)
k=0 k=0
or equivalently in expanded form (taking into account that u(x, 0) = φ(x))

u0 + u1 + u2 + · · · = φ(x) + L−1
t L (u
x 0 + u1 + u2 + · · · ) (96.10)
u0 = φ(x)
(96.11)
uk+1 = L−1
t L (u
x k ) , k = 0, 1, 2, · · ·
We thus have

k = 0 : u1 = L−1
t L (u
x 0 ) = L −1
t L x (φ(x))
Z t (96.12)
= L−1t φ00 (x) = φ00 (x)dt = φ00 (x) t
0
1246

k = 1 : u2 = L−1
t L (u
x 1 ) = L −1
t L x (φ00
(x)t)
Z t
t2 (96.13)
= L−1t φ 0000
(x)t = φ0000
(x)dt = φ 0000
(x)
0 2
and in general
tn
k = n : un = φ(2n) (x) (96.14)
n!
Therefore we have found the following formal solution of equation (96.1)
under the initial condition (96.2)
∞
X tn
u(x, t) = φ(x) + φ(2n) (x) (96.15)
n=1
n!
Considering the problem

ut = uxx , 0 < x < π, t≥0
u(x, 0) = sinx (96.16)
u(π, t) = 0, u(0, t) = 0
since u(x, 0) = φ(x) = sinx, we obtain using (96.12), (96.13), etc that

k = 0 : u1 = L−1
t L (u
x 0 ) = L −1
t L x (sinx)
Z t (96.17)
−1
= Lt −sinx = (−sinx)dt = −t sinx
0

k = 1 : u2 = L−1
t L (u
x 1 ) = L −1
t Lx (−t sinx)
Z t
t2 (96.18)
= L−1t t sinx = (sinx) tdt = sinx
0 2
We thus obtain
t2
u(x, t) = u0 + u1 + u2 + · · · = sinx − t sinx + sinx + · · ·
2 (96.19)
2
t
= sinx 1 − t + + · · ·
2
i.e. the exact solution
u(x, t) = e−t sinx (96.20)
1247
The boundary conditions u(π, t) = 0, u(0, t) = 0 are satisfied automatically.
Similarly the problem
ut = uxx , 0 < x < π, t≥0

u(x, 0) = x + cosx (96.21)
u(π, t) = π − e−t , u(0, t) = e−t
leads to the exact solution
u(x, t) = x + e−t cosx (96.22)
The extension of the Adomian decomposition method to nonlinear differ-

ential equations (ODEs or PDEs) is not a trivial task.
In that case we have to make use of the Adomian polynomials (Adomian
[1], [2], Wazwaz [3]).
We express the nonlinear terms F (u) in the form
∞
X
F (u) = An (u0 , u1 , · · · , un ) (96.23)
n=0
where the Adomian polynomials An are defined by

∞
1 dn
X
i
An = F( λ ui ) (96.24)
n! dλn n=0 λ=0
The first few of them are given by
A0 = F (u0 ) (96.25)
A1 = u1 F 0 (u0 ) (96.26)
1 2 00
A2 = u2 F 0 (u0 ) + u F (u0 ) (96.27)
2! 1
1 3 000
A3 = u3 F 0 (u0 ) + u1 u2 F 00 (u0 ) +u F (u0 ) (96.28)
3! 1
1 1
A4 = u4 F 0 (u0 ) + u22 + u1 u3 F 00 (u0 ) + u21 u2 F 000 (u0 )
2! 2! (96.29)
1 4 0000
+ u1 F (u0 )
4!
1248
For example for F = uux we find that the first four polynomials are given by
A0 = u0 (u0 )x (96.30)
A1 = (u0 )x u1 + u0 (u1 )x (96.31)

A2 = (u0 )x u2 + (u1 )x u1 + u0 (u2 )x (96.32)
A3 = (u0 )x u3 + (u1 )x u2 + (u2 )x u1 + (u3 )x u0 (96.33)
Example 1. The differential equation
ut + uux = 0, u(x, 0) = x (96.34)
can be written after separating the nonlinear part as
Lt u = −uux (96.35)
From the above equation, taking into account the initial condition, we obtain
u = x − L−1
t (uux ) (96.36)
The decompositions
∞
X ∞
X
u= un , uux = An (96.37)
n=0 n=0
convert (96.36) into the equation

∞
X X∞
un = x − L−1
t ( An ) (96.38)
n=0 n=0
From the above equation we get the recursive relations
u0 = x
(96.39)
uk+1 = −L−1
t (Ak ), k = 0, 1, 2, · · ·
The first four Adomian polynomials for uux have already been calculated in
(96.30)-(96.33). For u0 = x we find from (96.30) that
A0 = x (96.40)
1249
Therefore
Z t
u1 = −L−1
t (A0 ) = −Lt−1 (x) =− (x)dt = −xt (96.41)
0
From (96.31) for u0 = x and u1 = −xt we find
A1 = −2xt (96.42)
Therefore
Z t
u2 = −L−1
t (A1 ) = −L−1
t (−2xt) =− (−2xt)dt = xt2 (96.43)
0
From (96.32) for u0 = x, u1 = −xt and u2 = xt2 we find
A2 = 3xt2 (96.44)
Therefore
Z t
u3 = −L−1
t (A2 ) = −L−1 2
t (3xt ) =− (3xt2 )dt = −xt3 (96.45)
0
From (96.33) for u0 = x, u1 = −xt, u2 = xt2 and u3 = −xt3 we find
A3 = −4xt3 (96.46)
Therefore
Z t
u4 = −L−1
t (A3 ) = −L−1 3
t (−4xt ) =− (−4xt3 )dt = xt4 (96.47)
0
The solution of the equation is thus given by
u = u0 + u1 + u2 + u3 + u4 + · · ·
or, taking into account u0 = x and (96.41), (96.43), (96.45), (96.47), we

obtain
u = x − xt + xt2 − xt3 + xt4 + · · ·
x (96.48)
= x(1 − t + t2 − t3 + t4 + · · · ) = , |t| < 1
1+t
1250
We have thus derived an explicit solution of equation (96.34). The Adomian
decomposition is a remarkable method and can be used to solve more com-
plicated equations (ODEs and PDEs) and systems of differential equations
as well. The reader should take advantage of the many good books available,
like Wazwaz (Ref. [3]).
References
[1] G. Adomian: ”Solving Frontier Problems of Physics: The Decomposi-
tion Method”. Springer (1994).
[2] G. Adomian: ”A new approach to nonlinear partial differential equa-
tions”. J. Math. Anal. Appl. 102 (1984) 420-434.
[3] A-M. Wazwaz: ”Partial Differential Equations and Solitary Waves
Theory”. Springer 2009
97 The Simple Equation Method

Let
F (x, u, u0 , u00 , · · · ) = 0 (97.1)

unknown function. The simple equation method consists in expanding the
n
X Φ0 (x) k
u= ak (97.2)
k=0
Φ(x)
where Φ = Φ(x) is an unknown function which has to be determined.

The expansion (97.2) has to be substituted in (97.1). Before that, we have to
determine n, the maximum power of (Φ0 /Φ)k in the expansion (97.2). This
can be done by balancing the higher order derivative to the nonlinear term of
the highest power. Substituting (97.2) (after finding n) to (97.1) we obtain
an equation which contains both the coefficients ak and the ratios (Φ0 /Φ)k .
Multiplying this last equation by the highest power of Φ (i.e. eliminating the
denominators) we obtain an equation which has to be arranged in terms of
the powers Φk . Equating to zero the coefficients of all the powers of Φ, we
obtain a system of equations which contains both the unkown coefficients ak
and the differential equation(s) for Φ. We thus see that the simple equation
method determines both the coefficients and the unkown function Φ in the
1251
expansion (97.2).
An equation of the form (97.1) might be obtained from a nonlinear PDE
considering the traveling wave solutions.
Example. We consider the nonlinear PDE
ut − uxxt + 2αux + βu2 ux = 0 (97.3)
which is called the simplified Camassa-Holm equation. We are looking the
traveling wave solutions of this equation. Introducing the similarity variables
u = U (ξ), ξ = k(x − ωt) (97.4)
ut = (−kω)U 0 (ξ), ux = kU 0 (ξ), uxxt = k 2 (−kω)U 000 (ξ) (97.5)
ωk 2 U 000 (ξ) + (2α − ω)U 0 (ξ) + βU 2 (ξ)U 0 (ξ) = 0 (97.6)
1
ωk 2 U 00 + (2α − ω)U + βU 3 = 0 (97.7)
3
Balancing U 00 to U 3 , i.e. n + 2 = 3n we find n = 1 and thus we have to
consider the expansion
Φ0 (x)
U = a0 + a1 (97.8)
Φ(x)
Since
Φ00 (x) Φ0 (x) 2

0
U (ξ) = a1 − (97.9)
Φ(x) Φ(x)
000
00 Φ (x) Φ0 (x)Φ00 (x) Φ0 (x) 2
U (ξ) = a1 −3 +2 (97.10)
Φ(x) Φ2 (x) Φ(x)
equation (97.7) becomes because of (97.8) and (97.10)
000
2 Φ (x) Φ0 (x)Φ00 (x) Φ0 (x) 2
ωk a1 −3 +2
Φ(x) Φ2 (x) Φ(x)
Φ0 (x) 1 Φ0 (x) 3 (97.11)
+ (2α − ω) a0 + a1 + β a0 + a1 =0
Φ(x) 3 Φ(x)
1252
Upon expanding the above equation, multiplying through by Φ3 and rear-
ranging, we get the equation
1
βa30 + (2α − ω)a0 Φ3
3
+ [ωk 2 a1 Φ000 + a1 {−ω + βa20 + 2α}Φ0 ]Φ2
(97.12)
+ [−3ωk 2 a1 Φ00 Φ0 + βa0 a21 (Φ0 )2 ]Φ
0 3 1

3 2
+ (Φ ) βa + 2ωk a1 = 0
3 1
Equating to zero the coefficients of the different powers of Φ, we obtain the
1 3
βa + (2α − ω)a0 = 0 (97.13)
3 0
ωk 2 a1 Φ000 + a1 {−ω + βa20 + 2α}Φ0 = 0 (97.14)
−3ωk 2 a1 Φ00 Φ0 + βa0 a21 (Φ0 )2 = 0 (97.15)
1
(Φ0 )3 βa31 + 2ωk 2 a1 = 0 (97.16)
3
From equations (97.13) and (97.16) we obtain, considering that a0 6=0, a1 6=0
and Φ0 6=0 (the case a0 = 0 may be considered separately)
3(ω − 2α) 6ωk 2
a20 = , a21 = − (97.17)
β β
From the above two equations we determine a0 and a1 respectively. From
(97.15), upon dividing by a1 Φ0 , considered different than zero, we get the
−3ωk 2 Φ00 + βa0 a1 Φ0 = 0 (97.18)
The solution of this equaton is given by

βa a
0 1
Φ(ξ) = C1 + C2 · exp ξ (97.19)
3ωk 2
After plugging the above solution to (97.14), we obtain the consistency con-
dition
a20 a21 β 2
+ (2α + a20 β)ω − ω 2 = 0
9k 2
1253
which ω has to satisfy. However this condition becomes an identity, taking
into account (97.17).
We thus have found that the traveling wave solutions of (97.3) are given by
(97.8) where a0 and a1 are given by (97.17) and Φ by (97.19):
βa a
0 1
2
a0 a1 β exp ξ
U (ξ) = a0 + 3ωk 2 (97.20)
βa a
3ωk 2 0 1
C + exp ξ
3ωk 2
where C = C1 /C2 .
References
[1] N. Islam, Asaduzzaman and S. Ali: ”Exact Wave Solutions to
the simplified modified Camassa-Holm equation in Mathematical Physics”.
AIMS Mathematics 5 (2019) 26-41
[2] A. J. M. Jawad, M. D. Petkovic and A. Biswas: ”Modified simple
equation method for nonlinear evolution equations”.
Appl. Math. Comp. 217 (2010) 869-877
98 The Auxiliary Equation Method II

Let
F (x, u, u0 , u00 , · · · ) = 0 (98.1)
be a nonlinear ODE where u is the unknown function and x is the inde-
pendent variable. The simple equation method consists in expanding the
n
X
u= ak Φk (x) (98.2)
k=0
where Φ = Φ(x) is a function which is a solution of a differential equation.

Let in our case consider that Φ satisfies the equation
dΦ(x) 2
= A0 Φ2 (x) + A1 Φ3 (x) + A2 Φ4 (x) (98.3)
dx
The expansion (98.2) has to be substituted into the original equation taking
into account the derivative of Φ at each stage of differentiation, according
1254
to the differential equation (98.3). However, we have to determine n, the
maximum power of Φk in the expansion (98.2). This can be done by balanc-
ing the higher order derivative to the nonlinear term of the highest power.
Substituting (98.2) (after finding n) to (98.1) and taking into account the
derivative of Φ, we obtain an equation which contains both the coefficients
ak and the powers Φk . Equationg to zero the coefficients of the powers Φk
to zero, we obtain a system of equations from which we determine ak and
A0 , A1 , A2 .
The solution of (98.3) is given by
√A
2 0
−A0 A1 ·sech ± x
Φ= 2
√A 2 , A0 > 0 (98.4)
0
A21 − A0 A2 1 − tanh ± x
2
or
√
2A0 ·sech( A0 x)
Φ= p 2 √ , A0 > 0, A21 − 4A0 A2 > 0 (98.5)
A1 − 4A0 A2 − A1 sech( A0 x)
Example. We wish to find the traveling wave solutions of the generalized

reaction Duffing equation
utt + αuxx + βu + γu2 + δu3 = 0, u = u(x, t) (98.6)
u = U (ξ), ξ = k(x − ωt) (98.7)
utt = (−kω)2 U 00 (ξ), uxx = k 2 U 00 (ξ) (98.8)
k 2 (α + ω 2 )U 00 + βU + γU 2 + δU 3 = 0 (98.9)
Balancing U 00 to U 3 , i.e. n + 2 = 3n, we find n = 1 and we thus have to

consider the expansion
U = a0 + a1 Φ (98.10)
1255
We have to make two successive differentiations in order to find U 00 , taking
into account that we have to substitute
p
Φ0 = A0 Φ2 + A1 Φ3 + A2 Φ4 (98.11)
at each stage of differentiation. We have
p
U 0 = a1 Φ0 = a1 A0 Φ2 + A1 Φ3 + A2 Φ4 (98.12)
and
{2A0 Φ + 3A1 Φ2 + 4A2 Φ3 }Φ0
U 00 = a1 √
2 A0 Φ2 + A1 Φ3 + A2 Φ4
√
{2A0 Φ + 3A1 Φ2 + 4A2 Φ3 } A0 Φ2 + A1 Φ3 + A2 Φ4 (98.13)
= a1 √
2 A0 Φ2 + A1 Φ3 + A2 Φ4
1
= a1 (2A0 Φ + 3A1 Φ2 + 4A2 Φ3 )
2
Substitution of (98.13) and (98.10) into (98.9) yields
(2αk 2 a1 A2 + 2ω 2 k 2 a1 A2 + δa31 )Φ3
3 3
+ αk 2 a1 A1 + γa21 + 3δa0 a21 + ω 2 k 2 a1 A1 Φ2
2 2 (98.14)
+ (2γa0 a1 + αk a1 A0 + ω k a1 A0 + 3δa20 a1 + βa1 )Φ
2 2 2
+ βa0 + γa20 + δa30 = 0

Equating to zero all the coefficients of Φ, we obtain the following system of
equations
2αk 2 a1 A2 + 2ω 2 k 2 a1 A2 + δa31 = 0 (98.15)
3 2 3
αk a1 A1 + γa21 + 3δa0 a21 + ω 2 k 2 a1 A1 = 0 (98.16)
2 2
2γa0 a1 + αk a1 A0 + ω k a1 A0 + 3δa20 a1 + βa1 = 0
2 2 2
(98.17)
βa0 + γa20 + δa30 = 0 (98.18)
Solving the above system of equations (98.15)-(98.18) we get two families of
solutions
Family I.
3k 2 (α + ω 2 )
a0 = 0, a1 = − A1 (98.19)
2γ
1256
and
β 9δk 2
A0 = − 2 2
, A1 = A1 , A2 = − 2
(α + ω 2 )A21 (98.20)
k (α + ω ) 8γ
Family II. If µ is any solution of the quadratic equation
δµ2 + γµ + β = 0 (98.21)
then
a0 = µ, a1 = a1 (98.22)
and
γµ + 2β 2 γ + 3δµ
A0 = , A1 = − 2 a1 ,
k 2 (α
+ω )2 3 k (α + ω 2 )
(98.23)
1 δ
A2 = − 2 a2
2 k (α + ω 2 ) 1
Considering the above solutions, we first determine Φ given by (98.4)-(98.5)
and then the solution U given by (98.10).
References
[1] Sirendaoreji and S. Jiong: ”Auxiliary Equation Method and its
Applications to Nonlinear Evolution Equations”.
Int. J. Mod. Phys. 14 (2003) 1-11
99 The Variational Iteration Method

This method makes use of the Lagrange multiplier in a different setting
(Inokuti, Sekine and Mura [1] and references therein). Recently this method
was introduced by J. H. He (He [2], [3], [4]) to solving nonlinear differential
equations. Many examples are provided by A-M Wazwaz (Wazwaz [5]).
Let a DE be defined by
L(u) + N (u) = f (x) (99.1)
where L is a linear and N a nonlinear differential operator. Then the solution
of this equation can be determined by an iteration procedure defined through
the correction functional
Z x
un+1 (x) = un (x) + λ(ξ)[Lun (ξ) + N ũn (ξ) − f (ξ)]dξ (99.2)
0
1257
where λ is a Lagrange multiplier which can be determined optimally through
the variational procedure. The notation ũn means that it should be consid-
ered as a restricted variation, i.e.
δũn = 0 (99.3)
After determining λ, the solution can be found by

Z x
un+1 (x) = un (x) + λ(ξ)[Lun (ξ) + N un (ξ) − f (ξ)]dξ
0 (99.4)
n = 0, 1, 2, · · · where u0 is an initial guess
Example 1. The solution of the equation
u00 (x) − xu(x) + 1 = 0, u(−1) = 0, u(1) = 0 (99.5)
can be expressed in terms of the Airy functions. However considering the

variational iteration method we can find (a numerical) solution using the
iteration
Z x
un+1 (x) = un (x) + λ(ξ) [ u00n (ξ) − ξ ũn (ξ) + 1 ] dξ (99.6)
0
Using variation theory and taking into account δũn = 0, we have

Z x
δun+1 (x) = δun (x) + δ λ(ξ) [ u00n (ξ) − ξ ũn (ξ) + 1 ] dξ
Z 0x (99.7)
= δun (x) + λ(ξ) δu00n (ξ) dξ
0
At this point we use integration by parts. Since

Z x ξ=x Z x
λ(ξ) δu00n (ξ) 0
dξ = λ(ξ) δu (ξ) − λ0 (ξ) δu0n (ξ) dξ
0 ξ=0 0
and
Z x ξ=x Z x
0
λ (ξ) δu0n (ξ) 0
dξ = λ (ξ) δu(ξ) − λ00 (ξ) δun (ξ) dξ
0 ξ=0 0
1258
we have
Z x ξ=x ξ=x
λ(ξ) δu00n (ξ) 0
dξ = λ(ξ) δu (ξ) 0
− λ (ξ) δu(ξ)
0 ξ=0 ξ=0
Z x (99.8)
+ λ00 (ξ) δun (ξ) dξ
0
At the lower end-point the variation is considered to be zero, i.e. δun (0) = 0
and thus we get, combining (99.7) and (99.8)
Z x
δun+1 (x) = δun (x) + λ(ξ) δu00n (ξ) dξ
0
0 0
= δun (x) + λ(ξ) δu (ξ) − λ (ξ) δu(ξ) (99.9)
ξ=x ξ=x
Z x
0

0 0
δun+1 (x) = (1 − λ (ξ)) δun (ξ) + λ(ξ) δu (ξ)
ξ=x ξ=x
Z x (99.10)
0
[1 − λ0 (ξ)]ξ=x = 0, [λ(ξ)]ξ=x = 0, [λ00 (ξ)]ξ=x = 0 (99.11)
λ(ξ) = ξ − x (99.12)
Therefore the solution of the original equation (99.5) will be found by con-
sidering the iteration (using (99.12) and (99.6))
Z x
un+1 (x) = un (x) + (ξ − x) [ u00n (ξ) − ξ un (ξ) + 1 ] dξ (99.13)
0
Using an initial guess for u0 , we can determine (using any of the CAS) the
functions u1 , u2 , u3 , · · · at any order of approximation we like. The boundary
1259
conditions then have to be considered for the solution of the highest approx-
imation solution. The initial guess should contain two parameters (since we
have two boundary conditions).
Example 2. We shall now consider the PDE
ut = uxx + sinx, 0 < x < π, t>0

u(0, t) = e−t , u(π, t) = −e−t (99.14)
u(x, 0) = cosx
Considering variation with respect to t, we have

Z t ∂u (x, ξ) ∂ 2 ũ (x, ξ)
n n
un+1 (x, t) = un (x, t) + λ(ξ) − − sinx dξ (99.15)
0 ∂ξ ∂x2
Therefore, taking into account δũn = 0, we have
Z t ∂
δun+1 (x, t) = δun (x, t) + λ(ξ) δun (x, ξ) dξ
0 ∂ξ
Z t
= δun (x, t) + [λ(ξ)δun (x, t)]ξ=t − λ0 (ξ)δun (x, t)dξ
0
Z t
= [(1 + λ(ξ))δun (x, t)]ξ=t − λ0 (ξ)δun (x, t)dξ
0
From the above equation, we obtain according to the variational principle
[1 + λ(ξ)]ξ=t = 0 and [λ0 (ξ)]ξ=t = 0 (99.16)
The above system yields the value of the Lagrange multiplier:
λ(ξ) = −1 (99.17)
We thus have to consider the recursion sequence

Z t
∂un (x, ξ) ∂ 2 un (x, ξ)
un+1 (x, t) = un (x, t) − − − sinx dξ (99.18)
0 ∂ξ ∂x2
in order to find the solution of the equation (99.14).
For n = 0 and u0 (x, 0) = cosx, since
∂u0 (x, ξ) ∂ 2 u0 (x, ξ)

= 0 and = −cosx
∂ξ ∂x2
1260
equation (99.18) yields
Z t
u1 (x, t) = cosx − 0 + cosx − sinx dξ
0 (99.19)
= cosx − t cosx + t sinx
For n = 1 and u1 (x, t) given by (99.19), since
∂u1 (x, ξ)
= −cosx + sinx and
∂ξ
∂ 2 u1 (x, ξ)
= −cosx + ξ cosx − ξ sinx
∂x2
Z t
∂u1 (x, ξ) ∂ 2 u1 (x, ξ)
u2 (x, t) = u1 (x, t) − − − sinx dξ
0 ∂ξ ∂x2
= cosx − t cosx + t sinx
Z t
− (−cosx + sinx) − (−cosx + ξ cosx − ξ sinx)
0
(99.20)
− sinx dξ
Z t
= cosx − t cosx + t sinx − −ξ cosx + ξ sinx dξ
0
1 1
= cosx − t cosx + t sinx + t2 cosx − t2 sinx
2 2
We thus see that u2 (x, t) is given by
t2 t2
u2 (x, t) = cosx 1 − t + + sinx t − (99.21)
2! 2!
Continuing this way, we find that u3 (x, t) is given by
t2 t3 t2 t3
u3 (x, t) = cosx 1 − t + − + sinx t − + (99.22)
2! 3! 2! 3!
Therefore for any n we should have
t2 t3 (−1)n tn
un (x, t) = cosx 1 − t + − + · · · +
2! 3! n! (99.23)
t2 t3 (−1)n+1 tn
+ sinx t − + + · · · +
2! 3! n!
1261
and for n→∞,
u(x, t) = e−t cosx + (1 − e−t )sinx (99.24)
which is the solution of equation (99.14).

Example 3. In this example we consider the equation
utt = uxx + sinx, 0 < x < π, t > 0

u(0, t) = 0, u(π, t) = 0 (99.25)
u(x, 0) = sinx, ut (x, 0) = sinx
Considering variation with respect to t, we have

Z t ∂ 2 u (x, ξ) ∂ 2 ũ (x, ξ)
n n
un+1 (x, t) = un (x, t) + λ(ξ) − − sinx dξ
0 ∂ξ 2 ∂x2
(99.26)
Therefore, taking into account δũn = 0, we have

Z t ∂2
δun+1 (x, t) = δun (x, t) + λ(ξ) δu n (x, ξ) dξ
0 ∂ξ 2

∂
= δun (x, t) + λ(ξ) δun (x, ξ) − [λ0 (ξ)δun (x, ξ)]
∂ξ ξ=t
Z t
+ λ00 (ξ)δun (x, ξ)dξ
0

0 ∂
= [(1 − λ (ξ))δun (x, ξ)]ξ=t + λ(ξ) δun (x, ξ)
∂ξ ξ=t
Z t
+ λ00 (ξ)δun (x, ξ)dξ
0
From the above equation, we obtain according to the variational principle
[1 − λ0 (ξ)]ξ=t = 0 [λ(ξ)]ξ=t = 0 and [λ00 (ξ)]ξ=t = 0 (99.27)
The above system yields the value of the Lagrange multiplier:
λ(ξ) = ξ − t (99.28)
1262
We thus have to consider the recursive sequence
un+1 (x, t) = un (x, t)

Z t ∂ 2 u (x, ξ) ∂ 2 u (x, ξ)
n n
(99.29)
− (ξ − t) − − sinx dξ
0 ∂ξ 2 ∂x2
in order to find the solution of the equation (99.25).

For n = 0 and u0 (x, 0) = sinx + t sinx, since
∂ 2 u0 (x, ξ) ∂ 2 u0 (x, ξ)
= 0 and = −sinx − ξ sinx
∂ξ 2 ∂x2

Z t
u1 (x, t) = sinx + t sinx − (ξ − t) ξ sinx dξ
0 (99.30)
3
t
= sinx + t sinx − sinx
3!
For n = 1 and u1 (x, t) given by (99.30), since
∂ 2 u1 (x, ξ)
= −ξ sinx and
∂ξ 2
∂ 2 u1 (x, ξ) ξ3
= −sinx − ξ sinx + sinx
∂x2 3!
Z t ∂u (x, ξ) ∂ 2 u (x, ξ)
1 1
u2 (x, t) = u1 (x, t) − (ξ − t) − − sinx dξ
0 ∂ξ ∂x2
Z t
t3 ξ3
(99.31)
= sinx + t sinx − sinx − (ξ − t) − sinxdξ
3! 0 3!
t3 t5
= sinx + t sinx − sinx + sinx
3! 5!
We thus see that u2 (x, t) is given by
t3 t5
u2 (x, t) = sinx + sinx t − + (99.32)
3! 5!
1263
Continuing this way, we find that u3 (x, t) is given by
t3 t5 t7
u3 (x, t) = sinx + sinx t − + − (99.33)
3! 5! 7!
Therefore for any n we should have
t3 t5 t2n+1
un (x, t) = sinx + sinx t − + + · · · + (−1)n (99.34)
3! 5! (2n + 1)!
and for n→∞,
u(x, t) = sinx + sinx sint (99.35)
which is the solution of equation (99.25).

References
[1] M. Inokuti, H. Sekine and T. Mura: ”General Use of the Lagrange
Multiplier in Nonlinear Mathematical Physics”.
In ”Variational Methods in the Mechanics of Solids”, S. Nemat-Nasser (Ed.)
Pergamon Press 1980, pages 156-162.
[2] J. H. He: ”Variational iteration method - a kind of nonlinear analytical
technique: some examples”. Int. J. Nonlinear Mech, 34 (1999) 799-808.
[3] J. H. He: ”Variational iteration method for autonomous ordinary dif-
ferential systems”. Appl. Math. Comput. 114 (2000) 115-123.
[4] J. H. He: ”Some asymptotic methods for strongly nonlinearly equa-
tions”. Int J. Modern Math. 20 (10) (2006) 1141-1199.
[5] A-M. Wazwaz: ”Partial Differential Equations and Solitary Waves
100 Painlevé Analysis

A partial differential equation has the Painlevé property when the solu-
tions of the PDE are ”single-valued” about the movable singularity manifold.
If the singularity manifold is given by
φ(z1 , z2 , · · · , zn ) = 0 (100.1)
and
u = u(z1 , z2 , · · · , zn ) (100.2)
1264
is a solution of the PDE, then we assume that
∞
X
α
u=φ uk φk (100.3)
k=0
where
φ = φ(z1 , z2 , · · · , zn ), uk = uk (z1 , z2 , · · · , zn ) (100.4)
are analytic functions of (zk ) in a neighborhood of the manifold (100.1) and

α is an integer.
Substitution of (100.3) into the PDE determines the value(s) of α and defines
the recursion relations for {uk }. The recursion relations of {uk }, when written
as
(n + 1)(n − a1 )· · ·(n − ak )φm

x uk = F (u0 , u1 , · · · , un−1 , φ, z) (100.5)
determine the ”resonances” at the points −1, a1 , · · · , ak , since at these points

the recursion relations are not defined. The point n = −1 corresponds to the
arbitrary singularity manifold (φ=0).
If the resonance equations are compatible, then the PDE has the Painlevé
property, i.e. is conjectured to be integrable. This is the so-called WTC
approach (Weiss, Tabor and Carnivale [1]).
Finding the recursion relations between {uk } is a rather demanding task. An
algorithmic procedure has been developed in (Xie and Chen [4]) using the
Wu-Ritt algorithm. However there is a ”fast” way of identifying the reso-
nances. This will become clear at the Examples which follow.
A connection between the Painlevé property and the Bäcklund transforma-
tions and Lax pairs has been established (Weiss [2]). On the other hand there
is a connection between symmetry and the Painlevé property (Strampp [5]).
A similar to the PDEs Painlevé method has been developed for the ODEs as
well. For a classification of ODEs which pass the Painlevé test, the reader
has to consult (Levi, Sekera and Winternitz [6]).
Some additional References on this important subject appear at the end of
this Section.
Example 1. The KdV Equation. We consider the KdV equation given
by
ut + uux + µ uxxx = 0 (100.6)
1265
After substitution of
u = φα (x, t)u0 (x, t) (100.7)
into (100.6), the leading order singularity analysis yields α = −2. In fact we
derive the equation
φ2α u0 (u0 )x + φα [(u0 )t + µ(u0 )xxx ] + φ2α−1 αu20 φx

+ φα−1 [αu0 φt + 3αµφx (u0 )xx + 3αµφxx (u0 )x + αµφxxx u0 ]
+ φα−2 [3µα2 φx φxx u0 − 3αµ(φx )2 (u0 )x − 3αµφxx φx u0 (100.8)
+ 3µα2 (φx )2 (u0 )x ]
+ φα−3 [µα3 (φx )3 u0 − 3µα2 (φx )3 u0 + 2µα(φx )3 u0 ] = 0
From the aove equation we see that the most singular terms are φ2α−1 and
φα−3 . Equating these two terms, we get 2α − 1 = α − 3 and thus
α = −2 (100.9)
Equation (100.8) for α = −2 takes on the form
φ−4 u0 (u0 )x + φ−2 [(u0 )t + µ(u0 )xxx ]

+ φ−3 [−2φt u0 − 6µφx (u0 )xx − 6µφxx (u0 )x − 2µφxxx u0 ]
(100.10)
+ φ−4 [18µφx φxx u0 + 18µ(φx )2 (u0 )x ]
+ φ−5 [−2φx u20 − 24µ(φx )3 u0 ] = 0
Equating to zero the most singular term, which is the coefficient of φ−5 , i.e.
−2φx u20 − 24µ(φx )3 u0 = 0 we can determine u0 :
u0 = −12µ(φx )2 (100.11)
We thus see that leading order singularity analysis determines both the ex-
ponent α and the leading term u0 of the expansion (100.3).
We thus have the expansion
∞
X
−2
u=φ uk φ k (100.12)
k=0
1266
Inserting the expansion (100.12) into (100.6) leads to the recursion relation
k
X
(uk−3 )t + (k − 4)uk−2 φt + [uk−m ((um−1 )x + (m − 2)φx um )]
m=0
+ µ[(uk−3 )xxx + (k − 4)(3(uk−2 )xx φx + 3(uk−2 )x φxx + uk−2 φxxx ) (100.13)
+ (k − 3)(k − 4)(3(uk−1 )x φ2x + 3uk−1 φx φxx )
+ (k − 2)(k − 3)(k − 4)φ3x uk ] = 0
After solving with respect to uk , the above relation yields
µφ3x (k + 1)(k − 4)(k − 6)uk = F (uk−1 , · · · , u0 , φt , φx , φxx , · · · )
(100.14)
for k = 0, 1, 2, · · ·
It is obvious that the above equality is not true for k = −1, 4, 6. These values
correspond to the ”resonances” of the recursion relations.
Putting k = 0, 1, 2, · · · in (100.13) we obtain
k=0: u0 = −12µ(φx )2 (100.15)
k=1: u1 = 12µφxx (100.16)

k=2: φx φt + u2 φ2x − 3µφ2xx + 4µφx φxxx = 0 (100.17)
k=3: φxt + u2 φxx − u3 φ2x + µφxxxx = 0 (100.18)
∂
k=4: (φxt + u2 φxx − u3 φ2x + µφxxxx = 0) = 0 (100.19)
∂x
By (100.18) the compatibility condition (100.19) is satisfied identically. We
get similarly the compatibility condition for k = 6.
Since we get compatibility of the resonance equations, the KdV equation
passes the Painlevé test.
Setting the resonance functions u4 and u6 equal to zero,
u4 = u6 = 0 (100.20)
and requiring u3 = 0, we can establish that
uk = 0 for k≥3 (100.21)
provided that u2 satisfies the original KdV equation. We thus have
φx φt + u2 φ2x − 3µφ2xx + 4µφx φxxx = 0 (100.22)
1267
and
φxt + u2 φxx + µφxxxx = 0 (100.23)
using (100.17) and (100.18) respectively. Upon elimination of u2 between
(100.22) and (100.23), i.e. solving (100.23) with respect to u2 ,
µφxxxx + φxt
u2 = − (100.24)
φxx
and substitution into (100.22), we get the equation
µφ
xxxx + φxt

φx φt + − φ2x − 3µφ2xx + 4µφx φxxx = 0
φxx
φxx
The above equation, after multiplying through by − takes on the form
φ3x
φ3xx

φxt φt φxx φxxxx φxxx φxx
− +µ −4 +3 3 =0 (100.25)
φx φ2x φx φ2x φx
Since
φxt φt φxx ∂ φt
− = (100.26)
φx φ2x ∂x φx
and
φxxxx φxxx φxx φ3xx ∂ ∂ φxx 1 φxx 2
−4 +3 3 = − (100.27)
φx φ2x φx ∂x ∂x φx 2 φx
∂ φt ∂ φ 1 φ 2
xx xx
+µ − =0 (100.28)
∂x φx ∂x φx 2 φx
At this stage we introduce the notation
∂ φxx 1 φxx 2
{φ; x} = − (100.29)
∂x φx 2 φx
where {φ; x} designates the so-called ”Schwarzian derivative”.
Therefore equation (100.28) yields
∂ φt
+ µ{φ; x} = 0 (100.30)
∂x φx
1268
from which we obtain that
φt
+ µ{φ; x} = C (100.31)
φx
A remarkable feature of the Schwarzian derivative is that it is invariant under
the Möbius group acting on the dependent variable, i.e.

aφ + b
; x ≡{φ; x} (100.32)
cφ + d
On the other hand, taking into account
u = φ−2 (u0 + u1 φ + u2 φ2 ) (100.33)
with u0 and u1 given by (100.15) and (100.16) respectively, we get
φ2x φxx
u = −12µ + 12µ + u2 (100.34)
φ2 φ
∂2
u = 12µ ln(φ) + u2 (100.35)
∂x2
which is the Bäcklund transformation for the KdV equation.
Note 1. The resonances can also be evaluated using the substitution
u = u0 φα + k φr+α (100.36)
Specializing to our case for α = −2 and u0 = −12µ(φ2x ), i.e.
u = −12µ(φ2x ) φ−2 + k φr−2 (100.37)
and substitution of (100.37) into the original equation ut + uux + µ uxxx = 0,

we find

2r−5 2
(r − 2)φx φ k + µ[r3 − 9r2 + 14r + 24](φ3x )φr−5

+ µ[3r2 − 15r − 6]φx φxx φr−4 + [µ(r − 2)φxxx − 2φx + rφt ]φr−3 k
− 360µ2 (φ3x )φxx φ−4 + 24[µ(φ2x )φt + 7µ2 (φ2x )φxxx + 12µ2 φx (φxx )2 ]φ−3
− 24[µφx φxt + µ2 φx φxxxx + 3µ2 φxx φxxx ]φ−2 = 0
1269
From the above equation we see that the most singular term is φr−5 , which
means that we should have r3 −9r2 +14r+24 = 0, i.e. (r+1)(r−4)(r−6) = 0
from which we get
r = −1, 4, 6 (100.38)
Note 2. Equation (100.31) for C = 0 admits a number of solutions, for

example
φ = a + b tanh(kx − ωt + δ) if ω = −2µk 3 (100.39)
or
φ = a + b coth(kx − ωt + δ) if ω = −2µk 3 (100.40)
or even
φ = 1 + c exp(kx − ωt + δ) if ω = k 3 (100.41)
Truncating (100.35) at the constant level term, i.e. taking
∂2
u = 12µ ln(φ) (100.42)
∂x2
and taking
φ = 1 + exp(kx − k 3 t + δ) (100.43)
according to (100.41), we get from (100.42) the solution
exp(kx − k 3 t + δ)
u = 12µk 2 (100.44)
[exp(kx − k 3 t + δ) + 1]2
Lemma. The following identity holds true
eζ

1 2 ζ
= 1 − tanh (100.45)
(1 + eζ )2 4 2
Proof. We have
eζ 1 (1 + eζ )2 − (1 − eζ )2 1 − eζ 2

1
= = 1− (100.46)
(1 + eζ )2 4 (1 + eζ )2 4 1 + eζ
1270
On the other hand,
eζ − e−ζ e2ζ − 1
tanh(ζ) = = (100.47)
eζ + e−ζ e2ζ + 1
and thus
eζ − 1 ζ
= tanh (100.48)
eζ + 1 2
Using (100.46) and (100.48), we get (100.45).
Using the above Lemma, we can write equation (100.44) in the equivalent
form
3

2 kx − k t + δ

2
u = 3 µk 1 − tanh (100.49)
2
or
kx − k 3 t + δ
u = 3 µk 2 sech2 (100.50)
2
The above is the one-soliton solution of the KdV equation. Therefore Painlevé
analysis provides us both the Bäcklund transformation and the one-soliton
solution of the KdV equation.
Note 3. The functions u0 , u1 , · · · appearing in the perturbative expansion
may be calculated successively using the substitutions
u = u0 φα (100.51)
u = φα (u0 + u1 φ) (100.52)
u = φα (u0 + u1 φ + u2 φ2 ) (100.53)
etc. in the original equation. In other words there is no need to consider the
recursion equation.
The first substitution
u = φα (x, t)u0 (x, t)
has already been considered before (equation (100.7)) and we have found α =
−2 and u0 = −12µ(φx )2 (see equations (100.9) and (100.11) respectively).
1271
Let us now consider the substitution u = φα (u0 + u1 φ) with α = −2 and
u0 = −12µ(φx )2 , i.e.
u = φ−2 (−12µ(φx )2 + u1 φ) (100.54)
Under this substitution, the original KdV equation (100.6) yields
φ−1 ((u1 )t + µ(u1 )xxx )

+ φ−2 [−µφxxx u1 − 3µφx (u1 )xx − 24µφx φxt − 24µ2 φx φxxxx
+ u1 (u1 )x − 72µ2 φxx φxxx − u1 φt − 3µφxx (u1 )x ]
(100.55)
+ φ−3 [288µ2 φx (φxx )2 − 18µu1 φx φxx + 168µ2 (φx )2 φxxx
+ 24µφt (φx )2 − u21 φx − 6µ(φx )2 (u1 )x ]
+ φ−4 [30µ(u1 − 12µφxx )(φx )3 ] = 0
The most singular term of this equation is φ−4 . Equating to zero the coeffi-
cient of this term, yields
u1 = 12µφxx (100.56)
Let us next consider the substitution u = φα (u0 + u1 φ + u2 φ2 ) with α = −2

and u0 = −12µ(φx )2 , u1 = 12µφxx , i.e.
u = φ−2 (−12µ(φx )2 + 12µφxx φ + u2 φ2 ) (100.57)
Under this substitution, the original KdV equation (100.6) yields
(u2 )t + u2 (u2 )x + µ(u2 )xxx

+ φ−1 [12µu2 φxxx + 12µ2 φ5x + 12µφxxt + 12µφxx (u2 )x ]
+ φ−2 [−24µφx φxt − 12µ(φx )2 (u2 )x + 24µ2 φxx φxxt (100.58)
− 60µ2 φx φ4x − 36µu2 φx φxx − 12µφt φxx ]
+ φ−3 [24µ(φt φx + u2 (φx )2 + 4µφx φxxx − 3µ(φxx )2 )φx ] = 0
The most singular term of the above equation is φ−3 . Equating to zero this
term, we get the equation
φt φx + u2 (φx )2 + 4µφx φxxx − 3µ(φxx )2 = 0 (100.59)
The above is equation (100.17) derived by another method. We can continue

in this way and recover equations (100.18), (100.19) and even more.
1272
Note 4. The Painlevé method can also be applied to ordinary deifferential
equations. Let us consider the equation
u00 + au2 + bu + c = 0, u = u(x) (100.60)
The substitution
u = u0 (x)φα (x) (100.61)
leads to the equation
au20 φ2α + u000 φα + bu0 φα + c

(100.62)
+ φα−1 (2αu00 φ0 + αu0 φ00 ) + φα−2 (−αu0 φ00 + α2 u0 φ00 ) = 0
The most singular terms of the above equation are φ2α and φα−2 . Equating
these two terms, we get 2α = α − 2 and thus
α = −2 (100.63)
c + φ−2 (u000 + bu0 ) + φ−3 (−4u00 φ0 − 2u0 φ00 )

(100.64)
+ φ−4 (au20 + 6u0 (φ0 )2 ) = 0
The most singular term of the above equation is φ−4 . Equating the coefficient
of this term to zero, we get au20 + 6u0 (φ0 )2 = 0 from which we determine u0 :
6
u0 = − (φ0 )2 (100.65)
a
The substitution u = φ−2 (u0 + u1 φ), i.e.
6
u = φ−2 − (φ0 )2 + u1 φ (100.66)
a
into (100.60) leads to the equation
c + φ−1 (u001 + bu1 )

12 12 6b
+ φ−2 −2u01 φ0 − u1 φ00 − (φ00 )2 − φ0 φ000 + au21 − (φ0 )2 (100.67)
a a a
60
+ φ−3 −10(φ0 )2 u1 + (φ0 )2 φ00 = 0
a
1273
The most singular term of the above equation is φ−3 . Equating to zero this
60
term, i.e. −10(φ0 )2 u1 + (φ0 )2 φ00 = 0, we determine u1 :
a
6
u1 = φ00 (100.68)
a
The substitution u = φ−2 (u0 + u1 φ + u2 φ2 ), i.e.
6 6
u = φ−2 − (φ0 )2 + φ00 φ + u2 φ2 (100.69)
a a
into (100.60) leads to the equation
00 2 −1

00 6 (iv) 6b 00
u2 + bu2 + au2 + c + φ 12u2 φ + φ + φ
a a (100.70)
6b 18 24
−2 0 2 00 2 0 000 0 2
+φ − (φ ) + (φ ) − φ φ − 12u2 (φ ) = 0
a a a
Equating to zero the coefficients of the singular terms φ−1 and φ−2 , we obtain
the equations
6 6b
12u2 φ00 + φ(iv) + φ00 = 0 (100.71)
a a
and
6b 0 2 18 00 2 24 0 000
− (φ ) + (φ ) − φ φ − 12u2 (φ0 )2 = 0 (100.72)
a a a
respectively. Solving (100.71) with respect to u2 we find
1 φ(iv) b
u2 = − 00
− (100.73)
2a φ 2a
Upon substitution of u2 given by (100.73) into (100.72) yields the equation
00 2 φ(iv) 0 2
0 000
3(φ ) − 4φ φ + 00 (φ ) = 0
φ
φ00
Multiplying the above equation by we obtain the equation
(φ0 )3
φ(iv) φ00 3 φ00 φ000

+ 3 − 4 =0 (100.74)
φ0 φ0 (φ0 )2
1274
Using (100.27) and (100.29), we write (100.74) into the form
d
{φ; x} = 0 (100.75)
dx
where {φ; x} is the Schwarzian derivative. Therefore we get from (100.75)
that
{φ; x} = k (100.76)
On the other hand we have

u0 u1
u= 2 + + u2 (100.77)
φ φ
6 φ0 2 6 φ00 1 φ(iv) b
u=− + − 00
− (100.78)
a φ a φ 2a φ 2a
or even
6 d2 1 φ(iv) b
u= ln(φ) − − (100.79)
a dx2 2a φ00 2a
We thus see that after conducting Painlevé analysis, the solution of equation
(100.60) is given by the Bäcklund transformation (100.79) where φ is the
solution of the Schwarzian equation (100.76).
The solution of (100.76) may be taken as
φ = C1 + C2 tanh(C·x + δ)
φ = C1 + C2 coth(C·x + δ) (100.80)
if − 2c2 = k
or even
φ = C1 + C2 tan(C·x + δ)
φ = C1 + C2 cot(C·x + δ) (100.81)
if 2c2 = k
Alternatively, we may consider the truncated expression

6 d2
u= ln(φ) (100.82)
a dx2
1275
and then substitute into (100.60) in order to determine the solution of the
equation.
Note 5. We may consider equation (100.60) with variable coefficients, i.e.
u00 + a(x)u2 + b(x)u + c(x) = 0 (100.83)
Upon the substitution
u = u0 (x)φα (x) (100.84)
into (100.83), we find in complete analogy to Note 4, that the most singular
term occurs for α = −2 and that
6
u0 = − (φ0 )2 (100.85)
a(x)
after equating to zero the coefficient of the most singular term φ−4 . The
substitution u = φ−2 (u0 + u1 φ), i.e.
−2
6 0 2

u=φ − (φ ) + u1 φ (100.86)
a(x)
after equating to zero the coefficient of the most singular term φ−3 leads to
the determination of u1 :
6 00 12 a0 (x) 0
u1 = φ − φ (100.87)
a(x) 5 a2 (x)
The substitution u = φ−2 (u0 + u1 φ + u2 φ2 ), i.e.
6 00 12 a0 (x) 0

−2
6 0 2 2

u=φ − (φ ) + φ − φ φ + u 2 φ (100.88)
a(x) a(x) 5 a2 (x)
after equating to zero the coefficients of the singular terms φ−2 and φ−1 yields
the two equations
33 (a0 (x))2 9 a00 (x) b(x) 6 a0 (x) φ00

u2 = − + − +
25 a3 (x) 10 a2 (x) 2 a(x) 5 a2 (x) φ0
(100.89)
2 φ000 3 φ00 2
− +
a(x) φ0 2 a(x) φ0
1276
and
2 a0 (x) 0 108 (a0 (x))2 00 72 (a0 (x))2 0
12 φ00 − φ u2 + φ − φ
5 a(x) 5 a3 (x) 5 a4 (x)
12 a000 (x) 0 b(x) 00 6 (iv) 72 a0 (x) 000
− φ + 6 φ + φ − φ (100.90)
5 a2 (x) a(x) a(x) 5 a2 (x)
72 a00 (x)a0 (x) 0 12 b(x)a0 (x) 0 54 a00 (x) 00
+ φ − φ − φ =0
5 a3 (x) 5 a2 (x) 5 a2 (x)
respectively. Upon substituting the value of u2 given by (100.89) into (100.90)
provides an equation with unknown function φ.
Combining (100.88) and (100.89) we get
6 φ0 2 6 φ00 12 a0 (x) φ0
u=− + −
a(x) φ a(x) φ 5 a2 (x) φ
0 00
33 (a (x))2
9 a (x) b(x) 6 a0 (x) φ00
− + − + (100.91)
25 a3 (x) 10 a2 (x) 2 a(x) 5 a2 (x) φ0
2 φ000 3 φ00 2
− +
a(x) φ0 2 a(x) φ0
6 d2 12 a0 (x) φ0
u= ln(φ) −
a(x) dx2 5 a2 (x) φ
33 (a0 (x))2 9 a00 (x) b(x) 6 a0 (x) φ00
− + − + (100.92)
25 a3 (x) 10 a2 (x) 2 a(x) 5 a2 (x) φ0
2 φ000 3 φ00 2
− +
a(x) φ0 2 a(x) φ0
The above expression determines the solution of the original equation (100.83)
where φ is the solution of the equation which is being derived by elimination
of u2 between (100.89) and (100.90).
However considering the truncated expression at the ”constant” level, i.e.
ignoring u2
6 d2 12 a0 (x) φ0
u= ln(φ) − (100.93)
a(x) dx2 5 a2 (x) φ
and substitution into the original equation (100.83) we can determine φ and
then the solution u.
1277
Appendix I. In this Appendix we shall determine the general solutions of
the Schwarz equation
d Φ00 1 Φ00 2
− = C, Φ = Φ(x) (100.94)
dx Φ0 2 Φ0
where we make the convenient choice
C = ±2k 2 (100.95)
The substitution
Φ00
u= 0 (100.96)
Φ
converts the original equation into
1
u0 = u2 ±2k 2 (100.97)
2
The above is a Riccati type of differential equation which can be converted
into the second order differential equations
y 00 + k 2 y = 0 and y 00 − k 2 y = 0 (100.98)

y0
u = −2 (100.99)
y
Equations (100.98) admit the solutions
y = C1 sin(kx) + C2 cos(kx) (100.100)
and
y = C1 ekx + C2 e−kx (100.101)
respectively. We then obtain the corresponding solutions
C1 cos(kx) − C2 sin(kx)
u1 = −2k (100.102)
C1 sin(kx) + C2 cos(kx)
1278
and
C1 ekx − C2 e−kx
u2 = −2k (100.103)
C1 ekx + C2 e−kx
combining (100.99) with (100.100) and (100.101) respectively. Using (100.96),
we obtain the equations
Φ00 Φ00
= u1 and = u2 (100.104)
Φ0 Φ0
The above two equations admit the solutions
C4
Φ(x) = C3 + (100.105)
C1 tan(kx) + C2
and
C4
Φ(x) = C3 + 2kx
(100.106)
C1 e + C2
respectively. The above two solutions should be written in the form
b
Φ(x) = a + (100.107)
c tan(kx) + 1
and
b
Φ(x) = a + (100.108)
c e2kx +1
respectively.
Appendix II. We shall next sketch the general solution of Schwarz’s equa-
tion
Φt
+ µ{Φ; x} = C (100.109)
Φ
where
∂ Φxx 1 Φxx 2
{Φ; x} = − (100.110)
∂x Φx 2 Φx
is the Schwarzian derivative. The substitution
Φ = eF (t) Ψ(x) (100.111)
1279
F 0 (t)
+ µ{Ψ; x} = C (100.112)
F (t)
Introducing the separation constant λ by
F 0 (t)
=λ (100.113)
F (t)

C −λ
{Ψ; x} = (100.114)
µ
which is similar in form to (100.94) solved in Appendix I. On the other hand,
the general solution of (100.113) is given by
F (t) = C1 eλt (100.115)
The general solution is obtained using (100.111), (100.115) and the solutions
of (100.114) taking into account Appendix I.
References
[1] J. Weiss, M. Tabor and G. Carnevale: ”The Painlevé property for
partial differential equations”. J. Math. Phys. 24 (1983) 522-526
[2] J. Weiss: ”The Painlevé property for partial differential equations. II.
Bäcklund transformations, Lax pairs, and the Schwarzian derivative”.
J. Math. Phys. 24 (1983) 1405-1413
[3] J. Weiss: ”On classes of integrable systems and the Painlevé property”.
J. Math. Phys. 25 (1984) 13-24
[4] F. Xie and Y. Chen: ”An algorithmic method in Painlevé analysis of
PDE”. Comp. Phys. Comm. 154 (2003) 197-204
[5] W. Strampp: ”Symmetries and the Painlevé Property”.
Prog. Theor. Phys. 76 (1986) 802-809
[6] D. Levi, D. Sekera and P. Winternitz: ”Lie point symmetries and
ODEs passing the Painlevé test”. arXiv: 1712.09811v2
[7] R. Conte (Ed): ”The Painlevé Property One Century Later”.
Springer 1999
[8] A. S. Fokas, A. Its, A. Kapaev and V. Yu. Novokshenov (Eds):
1280
”Painlevé Transcendents. The Riemann-Hilbert Approach”. AMS 2006
[9] D. Levi and P. Winternitz (Eds): ”Painlevé Transcendents”.
Springer 1992
[10] R. Conte and M. Musette: ”The Painlevé Handbook”.
Springer 2008
[11] R. Conte: ”The Painlevé Approach to Nonlinear Ordinary Differential
Equations”. arXiv: solv-int/9710020v1
[12] M. Musette: ”Painlevé Analysis of Partial Differential Equations”.
arXiv: solv-int/9804003v1
[13] K. Iwasaki, H. Kimura, S. Shimomura and M. Yoshida: ”From
Gauss to Painlevé”. Vieweg, Braunschweig 1991
101 The Inverse Scattering Method.

The Lax Pair. The Darboux Method
101.1 The Inverse Scattering Method. This method was intro-
duced to find the soliton solutions of the KdV equation
ut − 6uux + uxxx = 0 (101.1)
There is a number of different formulations of this method which the reader
can find in the cited literature.
The main idea behind the method is the determination of the scattering data,
i.e. the asymptotic behavior of both the bound and the continuum states of
the potential u(x, t), which is associated to the Schrödinger equation
−ψxx (x, t) + u(x, t)ψ(x, t) = λ(t)ψ(x, t) (101.2)
where t is simply a parameter (not the time itself). The potential is supposed
to satisfy the KdV equation and decays to zero as x→±∞.
The scattering data are connected to the KdV solution via the Gefand-
Levitan-Marchenko equation (GLM equation)
Z ∞
K(x, y; t) + B(x + y; t) + K(x, s; t)B(s + y; t)ds = 0 (101.3)
x
where B(z; t) collects the scattering data, i.e.

N
1 +∞
X Z
2 −κn z
B(z; t) = cn e + b(k, t)eikz (101.4)
n=1
2π −∞
1281
The potential (i.e. the solution of the KdV equation) is calculated using the
equation
∂
u(x, t) = −2 K(x, x; t) (101.5)
∂x
where K(x, y; t) is the kernel of the GLM integral equation (101.3). For a
proof of (101.3)-(101.5), see any of the References at the end of this Section.
Let us now turn into the calculation details. First of all, we shall try to
determine how the eigenvalues λ(t) change with reepect to the t−parameter.
We consider the equation
2
∂
−λt ψ = − (u − λ) Ψ (101.6)
∂x2
where
Ψ = ψt + ψxxx − 3(u + λ)ψx (101.7)
Equation (101.6) comes upon combining the Schrödinger equation (101.2)
and the KdV equation (101.1). Details are presented in the Appendix of this
sub-Section.
We consider the bound states ψn first. We multiply (101.6) by ψn and then
use the Schrödinger equation (u − λn )ψn = (ψn )xx to simplify the result:
∂ 2Ψ
(−λn )t ψn2 = ψn − (u − λn )ψn Ψ
∂x2
∂ 2Ψ ∂ 2 ψn (101.8)
= ψn 2 − Ψ
∂x ∂x2
∂ ∂Ψ ∂ψn
= ψn −Ψ
∂x ∂x ∂x
Integrating the above equation with respect to x from −∞ to +∞, yields
Z +∞ ∂Ψ x=+∞
2 ∂ψn
− (λn )t ψn dx = ψn −Ψ (101.9)
−∞ ∂x ∂x x=−∞
The eigenfunctions ψn should vanish for x→±∞ and thus the right hand
side of (101.9) is equal to zero. Since the left hand integral is simply a
normalization constant, equation (101.9) holds true if (λn )t = 0, i.e.
dλn
=0 (101.10)
dt
1282
Therefore the eigenvalues of the bound states are invariant with respect to t:
λn (t) = λn (0) (101.11)
Equation (101.6) implies that

2
∂
− (u − λ) Ψ = 0
∂x2
i.e. function Ψ is an eigenfunction of Schrödinger’s equation with eigenvalue

λn :
∂ 2 Ψn
− uΨn = λn Ψn (101.12)
∂x2
Thus Ψn should have the form
Ψn (x, t) = An (t)ψn (x, t) + Bn (t)φn (x, t) (101.13)
i.e. a linear combination of two independent functions ψn and φn .

We further suppose that ψn and φn satisfy the conditions
lim ψn (x, t) ∝ e∓κn x

x→±∞
(101.14)
lim φn (x, t) ∝ e±κn x
x→±∞
where κn = −λ2n . Therefore we must set Bn (t) = 0 in order to avoid expo-

nentially unbounded behavior. Thus Ψn and ψn must be proportional:
Ψn = (ψn )t + (ψn )xxx − 3(u + λn )(ψn )x = An ψn (101.15)
The normalization constant Nn (t) for ψn (x, t) satisfies the usual requirement
Z +∞
2
Nn (t) ψn2 dx = 1 (101.16)
−∞
Differentiation of the above equation with respect to t and using (101.15),

yields
(Nn ψn )x + Nn (ψn )xxx − 3(u + λn )Nn (ψn )x = 0 (101.17)
1283
Using (101.14), we see that the general asymptotic form of Nn ψn in the limit
x→∞ is
lim [Nn (t)ψn (x, t)] = cn (t)e−κx (101.18)
x→∞
Substituting this asymptotic form into equation (101.17), we obtain the dif-
ferential equation for cn (t)
dcn
− 4κ3n cn = 0 (101.19)
dt
3
cn (t) = cn (0)e4κn t (101.20)
Therefore the proportionality constant cn (t) of the asymptotic decay of the
bound state wave function (introduced in (101.18)) is given by (101.20).
Let us now turn into the eigenfunctions and eigenvalues of the free states. It
is known from Quantum Mechanics that
(
a(k)e−ikx , x → −∞
ψ(x)−→ (101.21)
e−ikx + b(k)eikx , x → +∞
where a(k) is the transmission coefficient and b(k) is the reflection co-
efficient.
Going back to equation (101.12), we may represent Ψ as
Ψk (x, t) = Ak (t)ψk (x, t) + Bk (t)φk (x, t) (101.22)
where Ψk and ψk are both eigenfunctions of the Schrödinger equation with
the same eigenvalue. The Wronskian W of these functions must be a function
of t only, i.e.
∂
(ψk Ψk,x − Ψk ψk,x ) = 0 (101.23)
∂x
Evaluation of W in either one of the limits x → ±∞ yields
ψk Ψk,x − Ψk ψk,x = 0 (101.24)
The last equation implies that both Ψk and ψk are the same to within a
proportionality constant that may depend on t
Ψk (x, t) = Ak (t)ψk (x, t) (101.25)
1284
This proportionality constant Ak (t) may be evaluated by taking the limit
Ψ
k (bt − 4ikλk b)eikx + 4ikλk e−ikx
lim = (101.26)
x→+∞ ψk beikx + e−ikx
where for the right hand side we used the lower case of (101.21). Introducing
the notation
A = bt − 4ikλk b, B = 4ikλk , C = b, D = 1 (101.27)
Aeikx + Be−ikx
Ak =
Ceikx + De−ikx
Ak (Ceikx + De−ikx ) = Aeikx + Be−ikx

⇐⇒(A − Ak C)eikx + (B − Ak D)e−ikx = 0
From the above equation we obtain the system of two equations
A − Ak C = 0 and B − Ak D = 0 (101.28)
Equation B − Ak D = 0 can be written as 4ikλk − Ak = 0 and thus (with

λk = k 2 ),
Ak = 4ik 3 (101.29)
i.e. Ak is independent of t.
Equation B − Ak C = 0 can be written as (bt − 4ikλk b) − 4ik 3 ·b = 0, using
the value for Ak we have already found, and thus (with λk = k 2 )
bt − 8ik 3 b = 0 (101.30)

3t
b(k, t) = b(k, 0)e8ik (101.31)
The evolution of the transmission coefficient a(k) is obtained taking the limit
Ψ
k at e−ikx − 4λk (−ikae−ikx ) at + 4iak 3
lim = = (101.32)
x→−∞ ψk ae−ikx a
1285
Since the above limit must be equal to Ak = 4ik 3 , we get
at + 4iak 3
= 4ik 3
a
and thus at = 0, i.e.
a(k, t) = a(k, 0) (101.33)
Conclusion. The Scattering Data.
I. The eigenvalues λn (t) and the normalization constants cn (t) of the bound
states change according to
3
λn (t) = λn (0), cn (t) = cn (0)e4κn t (101.34)
II. For the free states, the evolution equations for the transmission a(k, t)
and reflection b(k, t) coefficients are
3t
a(k, t) = a(k, 0) and b(k, t) = b(k, 0)e8ik (101.35)
Since we know the scattering data, we may calculate the kernel K of the
GLM equation and then the potential.
We suppose now that the potential has one bound state with eigenvalue
λ = −κ2 and vanishing reflection coefficient b = 0. In this case B defined in
(101.4) has the form
3
B(z; t) = c2 (t)e−κz = c20 e8κ t e−κz (101.36)
where c0 = c(0). The GLM equation (101.3) then reads
3
K(x, y; t) + c20 e8κ t e−κ(x+y)
Z ∞
2 8κ3 t −κy
(101.37)
+ c0 e e K(x, s; t)e−κs ds = 0
x
At this stage we use the method of separated kernels considering the simple
choice
K(x, y; t) = f (x, t)e−κy (101.38)
where f (x, t) is a function to be determined. Upon combining (101.37) and
(101.38) we obtain after evaluating the integral and cancelling the common
factor e−κy
3 c20 8κ3 t
f (x, t) + c20 e8κ t e−κx + e f (x, t)e−2κx = 0 (101.39)
2κ
1286
The above is an algebraic equation which when solved with respect to f (x, t),
yields
3
c20 e8κ t e−κx
f (x, t) = − (101.40)
c20 8κ3 t −2κx
1+ e e
2κ
Therefore using (101.38) and (101.40) we get the following expression for the
kernel K(x, y; t) of the GLM equation
3
c2 e8κ t e−κ(x+y)
K(x, y; t) = − 0 2 (101.41)
c
1 + 0 e8κ3 t e−2κx
2κ
What we actually need is the expression K(x, x; t) for the kernel in order to
find the potential using (101.5). We thus obtain
3
c20 e8κ t e−2κx
K(x, x; t) = − (101.42)
c2
1 + 0 e8κ3 t e−2κx
2κ
Introducing a constant x0 by
c20 = 2κe2κx0 (101.43)
we can write (101.42) as
2
e−2κ(x−x0 −4κ t)
K(x, x; t) = −2κ (101.44)
1 + e−2κ(x−x0 −4κ2 t)
or more conveniently as
2
e−κ(x−x0 −4κ t)
K(x, x; t) = −2κ κ(x−x0 −4κ2 t) (101.45)
e + e−κ(x−x0 −4κ2 t)
The derivative of the above expression is evaluated to be
2
∂ e−2κ(x−x0 −4κ t)
K(x, x; t) = 4κ2 (101.46)
∂x (1 + e−2κ(x−x0 −4κ2 t) )2
Using the identity (100.45), we have
e−2F

1 1 1
−2F 2
= 1 − tanh (−F ) = sech2 (−F ) = sech2 (F ) (101.47)
2
(1 + e ) 4 4 4
1287
We thus find the following expression for the potential (combining (101.46),
(101.47) and (101.5) with F = κ(x − x0 − 4κ2 t))
u(x, t) = −κ2 sech2 [κ(x − x0 − 4κ2 t)] (101.48)
The above expression for the potential corresponds to the one-soliton solution
of the KdV equation with amplitude −κ2 and speed 4κ2 , moving to the right,
and centered initially at the point x0 .
Appendix. In this Appendix we shall prove equation (101.6). The reader
can also consult Vvedensky (Ref. [5], Chapter 10).
We first consider Schrödinger’s equation (101.2) written as
ψxx = (u − λ)ψ (101.49)
Differentiation of the above equation with respect to x results in
ψxxx = ux ψ + (u − λ)ψx (101.50)
Solving this equation with respect to ux ψ yields
ux ψ = ψxxx − (u − λ)ψx (101.51)
We shall derive (101.6) using the following steps:

Step 1. We differentiate equation (101.2) with respect to t
−ψxxt + ut ψ + uψt = λt ψ + λψt
ψxxt − (ut − λt )ψ − (u − λ)ψt = 0 (101.52)
Step 2. We substitute ut by 6uux − uxxx making use of the KdV equation

(101.1)
ψxxt − (6uux − uxxx − λt )ψ − (u − λ)ψt = 0
[ψxxt − (u − λ)ψt ] − (6uux − uxxx )ψ + λt ψ = 0

2
∂
− (u − λ) ψt − (6uux − uxxx )ψ + λt ψ = 0 (101.53)
∂x2
1288
Step 3. We use the identity
(ux ψ)xx = uxxx ψ + ux ψxx + 2uxx ψx (101.54)
and solve with respect to uxxx ψ
uxxx ψ = (ux ψ)xx − ux ψxx − 2uxx ψx (101.55)
We then substitute ψxx by (u − λ)ψ according to Schrödinger’s equation

(101.49)
uxxx ψ = (ux ψ)xx − ux (u − λ)ψ − 2uxx ψx (101.56)

2
∂
uxxx ψ = − (u − λ) (ux ψ) − 2uxx ψx (101.57)
∂x2
Step 4. We substitute the expression uxxx ψ given by (101.57) into (101.53)

and we get the equation
2 2
∂ ∂
− (u − λ) ψt + − (u − λ) (ux ψ)
∂x2 ∂x2
− 2uxx ψx − 6uux ψ + λt ψ = 0
The above equation is quivalent to

2
∂
− (u − λ) (ψt + ux ψ) − 2(uxx ψx + 3uux ψ) + λt ψ = 0 (101.58)
∂x2
Step 5. We simplify the expression uxx ψx +3uux ψ which appears in (101.58).

We consider the identity
(uψx )xx = uxx ψx + 2ux ψxx + uψxxx (101.59)
and then substitute ψxx and ψxxx by the right hand side of (101.49) and
(101.50) respectively:
(uψx )xx = uxx ψx + 2ux (u − λ)ψ + u[ux ψ + (u − λ)ψx ]

(101.60)
= (uxx ψx + 3uux ψ) − 2λux ψ + (u − λ)uψx
1289
Solving with respect to uxx ψx + 3uux ψ, we get
uxx ψx + 3uux ψ = (uψx )xx + 2λux ψ − (u − λ)uψx
(101.61)
= [(uψx )xx − (u − λ)uψx ] + 2λux ψ
In the above expression we substitute ux ψ by ψxxx − (u − λ)ψx according to

(101.51)
uxx ψx + 3uux ψ = [(uψx )xx − (u − λ)uψx ] + 2λ[ψxxx − (u − λ)ψx ] (101.62)

2
∂
uxx ψx + 3uux ψ = − (u − λ) (uψx )
∂x2
2
∂
+ 2λ − (u − λ) ψx (101.63)
∂x2
2
∂
= − (u − λ) (uψx + 2λψx )
∂x2
Step 6. We substitute the expression uxx ψx + 3uux ψ given by the right hand
side of (101.63) into the corresponding expression of (101.58) and we get the
equation
2
∂
− (u − λ) (ψt + ux ψ)
∂x2
2 (101.64)
∂
−2 − (u − λ) (uψx + 2λψx ) + λt ψ = 0
∂x2
2
∂
− (u − λ) (ψt + ux ψ − 2uψx − 4λψx ) + λt ψ = 0 (101.65)
∂x2
At this stage we introduce the function Ψ by
Ψ = ψt + ux ψ − 2uψx − 4λψx (101.66)
or upon substituting ux ψ by ψxxx − (u − λ)ψx according to (101.51), we get

the final form of Ψ
Ψ = ψt + ψxxx − 3(u + λ)ψx (101.67)
1290
2
∂
− (u − λ) Ψ + λt ψ = 0 (101.68)
∂x2
which is equation (101.6).

101.2 The Lax Pair. The Lax pair method, introduced by Peter Lax
(Lax [6]) is an atempt to identify integrable equations (through linearization).
The method works as follows:
Let L be a linear operator for which
LΨ = λΨ, Ψ = Ψ(x, t) (101.69)
Let also M be another operator for which
Ψt = M Ψ (101.70)
We differentiate (101.69) with respect to t and we obtain the equation
Lt Ψ + LΨt = λt Ψ + λΨt (101.71)
From the above equation we can derive the following, making use of (101.69)
and (101.70) (supposing M and λ commute)
Lt Ψ + LM Ψ = λt Ψ + λM Ψ
= λt Ψ + M (λΨ) (101.72)
= λt Ψ + M LΨ
Lt Ψ + LM Ψ − M LΨ = λt Ψ (101.73)
(Lt + LM − M L)Ψ = λt Ψ (101.74)
Provided that λt = 0, (which makes the problem isospectral) we obtain from

equation (101.74) that
Lt + LM − M L = 0 (101.75)
1291
or
Lt + [L, M ] = 0 (101.76)
after introducing the commutator [L, M ] of the operators L and M by
[L, M ] = LM − M L (101.77)
Equation (101.76) is known as Lax’s equation and the pair L, M is called
the Lax pair. Given a nonlinear PDE, once we can find the pair L, M such
that Lax’s equation is being satisfied, then this PDE can be solved by say
the Darboux transformation.
For a given PDE, it is not always obvious (and it is sometimes rather hard)
to find the Lax pair of operators.
Note. Lt denotes the derivative of operator L with respect to time and can
be computed using the obvious identity
Lt Ψ = (LΨ)t − LΨt
On the other hand, the commutator [L, M ] can be computed by
[L, M ]Ψ = L(M Ψ) − M (LΨ)
Example 1. Consider the operators
L = Dx2 + u (101.78)
and
M = −4Dx3 − 6uDx + (f (t) − 3ux )I (101.79)
where f (t) is an arbitrary function and I is the identity operator.
Calculating LM and M L we find
LM Ψ = L(M Ψ) = (Dx2 + u)[(−4Dx3 − 6uDx + (f (t) − 3ux )I)Ψ]
= −4Dx5 Ψ − 10uDx3 Ψ + (f (t) − 15ux )Dx2 Ψ
− (6u2 + 12uxx )Dx Ψ + (f (t)u − 3uux − 3u3x )Ψ
and
M LΨ = M (LΨ) = (−4Dx3 − 6uDx + (f (t) − 3ux )I)[(Dx2 + u)Ψ]
= −4Dx5 Ψ − 10uDx3 Ψ + (f (t) − 15ux )Dx2 Ψ
− (6u2 + 12uxx )Dx Ψ + (f (t)u − 9uux − 4u3x )Ψ
1292
respectively. Therefore
[L, M ]Ψ = (LM − M L)Ψ = (6uux + u3x )Ψ (101.80)
Since
Lt Ψ = ut Ψ (101.81)
combining (101.81) and (101.80), we obtain
Lt Ψ + [L, M ]Ψ = 0⇐⇒ut Ψ + (6uux + u3x )Ψ = 0 (101.82)
In other words Lax’s equation is equivalent to
ut + 6uux + u3x = 0 (101.83)
which is the KdV equation. Therefore the choice of L and M given by

(101.78) and (101.79) is the correct choice of the Lax pair for the KdV equa-
tion.
101.2a The Matrix Representation of Operators. Using the matrix
representation of operators for Lax pairs, things can be simplified consider-
ably. The matrix formalism for Lax pairs was introduced by Ablowitz, Kaup,
Newell and Segur (Ref. [7]). This is a formalism that allows a construction
that avoids the need to consider higher order Lax operators (the so-called
AKNS method).
To start with, let X and T be two matrices which satisfy the system of
Dx Ψ = XΨ (101.84)
and
Dt Ψ = T Ψ (101.85)
where Ψ is now a matrix. For example if L is of order 2, then Ψ is a column

ψ
Ψ= (101.86)
φ
while X and T are 2×2 matrices. If L is of order n, then Ψ is a column

consisting of n components while X and T are n×n matrices.
1293
Let us now consider the compatibility condition [Dx , Dt ]Ψ = 0 for the two
equations (101.84) and (101.85). We have, applying the Leibnitz rule and
repeated use of the two equations
Dx (Dt Ψ) − Dt (Dx Ψ)
= Dx (T Ψ) − Dt (XΨ)
= {(Dx T )Ψ + T (Dx Ψ)} − {(Dt X)Ψ + X(Dt Ψ)}
(101.87)
= {(Dx T )Ψ + T (XΨ)} − {(Dt X)Ψ + X(T Ψ)}
= (Dx T − Dt X)Ψ + (T X − XT )Ψ
= (Dx T − Dt X)Ψ + [T, X]Ψ = O
We thus get from the above equation
Dx T − Dt X + [T, X] = O (101.88)
or even
Tx − Xt + [T, X] = O (101.89)
The above is the so-called zero-curvature equation. It is the analog of

the Lax equation for the matrix representation of operators.
In the following we shall develop a method for explicit construction of the
Lax pairs and a method of discovering solutions of PDEs using the zero-
curvature equation and the Darboux transformation.
Example 2. Consider the operators

0 1
X= (101.90)
λ−u 0
and

f (t) + ux −(4λ + 2u)
T = (101.91)
−4λ2 + 2λu + 2u2 + uxx f (t) − ux
We have

0 1
Xt = (101.92)
−ut 0

uxx −2ux
Tx = (101.93)
2λux + 4uux + uxxx −uxx
1294
and

−uxx 2ux
[T, X] = T X − XT = (101.94)
(−2λ + 2u)ux uxx
Therefore the zero-curvature equation (101.89) becomes
Tx − Xt + [T, X]

uxx −2ux 0 1
= −
2λux + 4uux + uxxx −uxx −ut 0
(101.95)

−uxx 2ux
+
(−2λ + 2u)ux uxx

0 0
= =O
ut + 6uux + uxxx 0
From the above equation we get ut + 6uux + uxxx = 0, i.e. we get the KdV
equation. We thus see that the choice for X and T given by (101.90) and
(101.91) is the correct choice, for it leads to the KdV equation.
We shall now develop a method of determining the Lax pairs using the
scaling invariance of the given PDE (Larue [8], Hickman et al. [9]).
Example 3. We shall determine the Lax pair operators L and M for the
Lax fifth-order KdV equation
3 2 2
ut + α u ux + 2αux uxx + αuuxxx + u5x = 0 (101.96)
10
This equation is invariant under the scaling symmetry
(x, t, u)→(κ−1 x, κ−5 t, κ2 u) (101.97)
We thus assign the weights
w(x) = −1, w(∂x ) = 1

w(t) = −5, w(∂t ) = 5 (101.98)
w(u) = 2
We also have
w(uxxx ) = w(u) + 3w(∂x ) = 5
(101.99)
w(u5x ) = w(u) + 5w(∂x ) = 7
1295
Therefore L and M will be of order 2 and 5 respectively, according to the
assignments
w(u) = 2 −→ w(L) = 2
(101.100)
w(∂t ) = 5 −→ w(M ) = 5
We will thus have the following expressions for the L and M operators
L = Dx2 + a1 uI (101.101)
and
M = a2 Dx5 + a3 uDx3 + a4 ux Dx2 + (a5 u2 + a6 uxx )Dx
(101.102)
+ (a(t) + a7 uux + a8 uxxx )I
The various coefficients a1 , · · ·, a8 will be determined by the Lax equation

Lt + [L, M ] = 0. In fact, we have
L t = a1 u t I (101.103)
LM = (Dx2 + a1 u)(M ) = Dx2 (M ) + a1 u(M ) (101.104)

M L = {a2 Dx5 + a3 uDx3 + a4 ux Dx2 + (a5 u2 + a6 uxx )Dx
(101.105)
+ (a(t) + a7 uux + a8 uxxx )}(Dx2 + a1 u)
Explicit evaluation of LM and M L leads to
Lt + [L, M ] = Lt + LM − M L =
= a1 ut I + (2a3 − 5a1 a2 )ux Dx4 + (a3 + 2a4 − 10a1 a2 )uxx Dx3
+ (a4 + 2a6 − 10a1 a2 )u3x Dx2 − (3a1 a3 + a1 a4 + a7 )uux Dx2
+ (a1 a4 + 4a5 + a7 )uux Dx2 + (2a5 + 2a7 − 3a1 a3 )uuxx Dx (101.106)
+ 2(a5 + a7 − 2a1 a4 )u2x Dx + (a6 + 2a8 − 5a1 a2 )u4x Dx
− a1 a5 u2 ux I + (a7 − a1 a3 )uu3x I
+ (3a7 − a1 a4 + a1 a6 )ux uxx I + (a8 − a1 a2 )u5x I = 0
At this stage we have to simplify further the above expression using the
equation LΨ = λΨ, i.e.
(Dx2 + a1 u)Ψ = λΨ (101.107)
1296
Dx2 = (λ − a1 u)I (101.108)
and its differential consequences
Dx3 = −a1 ux I + (λ − a1 u)Dx

(101.109)
Dx4 = −a1 uxx I − 2a1 ux Dx + (λ − a1 u)2 I
3 2 2
Substituting (101.108)-(101.109) and also ut by − α u ux − 2αux uxx −
10
αuuxxx − u5x in (101.106), we obtain the equation
Lt + [L, M ] = Lt + LM − M L =
= 2[a1 (5a1 a2 − 2a3 − a4 ) + a5 + a7 ]u2x Dx
− λ(10a1 a2 − a3 − 2a4 )uxx Dx
+ 2[a1 (5a1 a2 − 2a3 − a4 ) + a5 + a7 ]uuxx Dx
− (5a1 a2 − a6 − 2a8 )u4x Dx
+ λ[a1 (10a1 a2 − 5λa2 − 7a3 ) + 2λa3 + 4a5 ]uux I (101.110)
3
− a1 α2 + 5a21 a2 − 5a1 a3 + 5a5 u2 ux I
10
− [a1 (2α − 15a1 a2 + 3a3 + 3a4 + a6 ) − 3a7 ]ux uxx I
− [a1 (α − 10a1 a2 + a3 + a4 + 2a6 ) − a7 ]ux u3x I
− λ(10a1 a2 − a4 − 2a6 )u3x I − [a1 (1 + a2 ) − a8 ]u5x I = 0
Equating to zero all the above coefficients, we obtain the following system of
equations (two of them are identical, that is why we have only nine equations
instead of ten)
a1 (5a1 a2 − 2a3 − a4 ) + a5 + a7 = 0
10a1 a2 − a3 − 2a4 = 0
5a1 a2 − a6 − 2a8 = 0
a1 (10a1 a2 − 5λa2 − 7a3 ) + 2λa3 + 4a5 = 0
3
a1 α2 + 5a21 a2 − 5a1 a3 + 5a5 = 0
10
a1 (2α − 15a1 a2 + 3a3 + 3a4 + a6 ) − 3a7 = 0
1297
a1 (α − 10a1 a2 + a3 + a4 + 2a6 ) − a7 = 0
10a1 a2 − a4 − 2a6 = 0
a1 (1 + a2 ) − a8 = 0
The solution of the above system yields
1
a1 = α, a2 = −16, a3 = −4α, a4 = −6α,
10 (101.111)
3 3 3
a5 = − α2 , a6 = −5α, a7 = − α2 , a8 = − α
10 10 2
where all the coefficients are independent of λ as it should be.
Therefore the Lax pairs L and M for the equation (101.96) are given by
1
L = L = Dx2 + αuI (101.112)
10
and
3
M = −16Dx5 − 4αuDx3 − 6αux Dx2 − α2 u2 + 5αuxx Dx
10 (101.113)
3 2 3
+ a(t) − α uux − αuxxx I
10 2
respectively.
We shall next consider the problems of conversion the {L, M } pair to {X, T }
pair and vice-versa.
Conversion I. {L, M }−→{X, T }. To be specific, it is supposed that we
know the Lax pair {L, M } for the KdV equation ut + 6uux + uxxx = 0 and
we wish to find the {X, T } pair for the same equation. In other words let us
suppose that we have
Lψ = λψ, ψt = M ψ (101.114)
where
L = Dx2 + uI (101.115)
and
M = −4Dx3 − 6uDx + (a(t) − 3ux )I (101.116)
1298
For the {X, T } pair, we are looking for a function

ψ
Ψ= (101.117)
φ
and 2×2 matrices X and T such that Ψx = XΨ and Ψt = T Ψ, i.e.

ψx A B ψ
= (101.118)
φx C D φ
and

ψt E F ψ
= (101.119)
φt G H φ
Let us set
ψx = φ (101.120)
and thus since φx = ψxx , we get
Lψ = λψ⇐⇒(Dx2 + u)ψ = λψ⇐⇒ψxx = (λ − u)ψ
(101.121)
⇐⇒φx = (λ − u)ψ
Equations (101.120) and (101.121) can be written in matrix form as

ψx 0 1 ψ
= (101.122)
φx λ−u 0 φ

0 1
X= (101.123)
λ−u 0
In order to find T , we have to express ψt and φt as a linear combination
of ψ and φ. Since, according to (101.121) φx = (λ − u)ψ, we obtain by
differentiation with respect to x (remember ψx = φ)
φxx = λψx − ux ψ − uψx = (λ − u)φ − ux ψ (101.124)
Equation ψt = M ψ can thus be written as
ψt = −4Dx3 ψ − 6uDx ψ + (a(t) − 3ux )ψ
= −4Dx2 φ − 6uφ + (a(t) − 3ux )ψ
(101.125)
= −4{(λ − u)φ − ux ψ} − 6uφ + (a(t) − 3ux )ψ
= −(4λ + 2u)φ + {a(t) + ux }ψ
1299
For φt we obtain, taking into account the previous expression for ψt
φt = (ψx )t = (ψt )x
= [−(4λ + 2u)φ + {a(t) + ux }ψ]x
= −(4λ + 2u)φx − 2ux φ + {a(t) + ux }ψx + uxx ψ (101.126)
= −(4λ + 2u)(λ − u)ψ − 2ux φ + {a(t) + ux }φ + uxx ψ
= (−4λ2 + 2λu + 2u2 + uxx )ψ + {a(t) − ux }φ
Equations (101.125) and (101.126) can be written in matrix form as

ψt a(t) + ux −(4λ + 2u) ψ
= 2 2 (101.127)
φt −4λ + 2λu + 2u + uxx a(t) − ux φ
Therefore the matrix T is given by

a(t) + ux −(4λ + 2u)
T = (101.128)
−4λ2 + 2λu + 2u2 + uxx a(t) − ux
Conversion II. {X, T }−→{L, M }. It is supposed next that we know the

Lax pair {X, T } for the KdV equation ut + 6uux + uxxx = 0 and we wish
to find the pair {L, M } for the same equation. Let us suppose that we have
first

ψx 0 1 ψ
= (101.129)
φx λ−u 0 φ
which means that
ψx = φ and φx = (λ − u)ψ
or
ψx = φ and ψxx = (λ − u)ψ
The above can be written as
φ = ψx and L̂ψ = 0, L̂ = Dx2 − (λ − u)I (101.130)
Defining L by
L = L̂ + λI = Dx2 + uI (101.131)
1300
to eliminate λ, we have obviously
Lψ = λψ (101.132)
which means that L is given by (101.131).

Considering next the equation

ψt a(t) + ux −(4λ + 2u) ψ
=
φt −4λ2 + 2λu + 2u2 + uxx a(t) − ux φ
we obtain
ψt = {a(t) + ux }ψ − (4λ + 2u)φ (101.133)
and
φt = (−4λ2 + 2λu + 2u2 + uxx )ψ + {a(t) − ux }φ (101.134)
Substituting φ by ψx in (101.133), we obtain
ψt = {a(t) + ux }ψ − (4λ + 2u)ψx (101.135)
We have next to eliminate λ from (101.135) and thus determine the M

operator. Differentiating λψ = Lψ with respect to x we get (λψ)x = (Lψ)x =
Dx (Dx2 + u)ψ or in expanded form
λψx = ψxxx + ux ψ + uψx (101.136)
We substitute λψx in (101.135) by the expression given in (101.136), and we

get
ψt = {a(t) + ux }ψ − 4(ψxxx + ux ψ + uψx ) − 2uψx
The above can be written as

ψt = −4ψxxx − 6uψx + {a(t) − 3ux }ψ
= (−4Dx3 − 6uDx + (a(t) − 3ux ))ψ
From the above equation we obtain the operator M , in accordance to the

equation ψt = M ψ :
M = −4Dx3 − 6uDx + (a(t) − 3ux )I (101.137)
1301
101.3 Lax Pair, Darboux Transformation and Exact Solutions
of Nonlinear PDEs.
Let Ψ be a solution of the Sturm-Liouville equation (S-L equation)
−Ψxx + uΨ = λΨ (101.138)
and
Ψ1 = Ψ1 (x, λ1 ) (101.139)
be another solution corresponding to the eigenvalue λ = λ1 . Let also σ1 be

defined by the logarithmic derivative of Ψ1 :
d Ψ1x
σ1 = ln(Ψ1 ) = (101.140)
dx Ψ1
The Darboux transformation (DT) of Ψ (Matveev and Salle [10])
Ψ −→ Ψ[1] (101.141)
is defined by
d Ψ1x W (Ψ1 , Ψ)
Ψ[1] = − σ1 Ψ = Ψ x − Ψ= (101.142)
dx Ψ1 Ψ1
where W (Ψ1 , Ψ) = Ψ1 Ψx − ΨΨ1x is the Wronskian of the pair of functions
Ψ1 and Ψ.
The importance of Darboux transformations lies to the following
Theorem 1. The function Ψ[1] satisfies the equation
−Ψxx [1] + u[1]Ψ[1] = λΨ[1] (101.143)
where
d2
u[1] = u − 2σ1x = u − 2 ln Ψ1 (101.144)
dx2
In other words, the S-L equation is covariant with respect to the action of
the DT:
Ψ −→ Ψ[1], u −→ u[1] (101.145)
1302
The above process can be continued from Ψ[1] to Ψ[2] by
d Ψ0 [1] d Ψ0
Ψ[2] = − 2 − 1 Ψ
dx Ψ2 [1] dx Ψ1
(101.146)
Ψ1x
Ψ2 [1] = Ψ2x − Ψ2
Ψ1
where Ψ2 [1] is the solution corresponding to λ = λ2 . The potential u[2] is
given in this case by
d2 d2
u[2] = u[1] − 2 ln Ψ2 [1] = u − 2 lnW (Ψ1 , Ψ2 ) (101.147)
dx2 dx2
There is the following important
Theorem 2. The function
W (Ψ1 , · · · , ΨN , Ψ)
Ψ[N ] = (101.148)
W (Ψ1 , · · · , ΨN )
satisfies the differential equation
−Ψxx [N ] + u[N ]Ψ[N ] = λΨ[N ] (101.149)
where
d2
u[N ] = u − 2 lnW (Ψ1 , · · · , ΨN ) (101.150)
dx2
The above Theorem is due to Crum and it is named after him.
Example 4. We shall determine the one-soliton solution of the KdV equa-
tion
ut + 6uux + uxxx = 0 (101.151)
using the Lax pairs
L = Dx2 + u, M = −4Dx3 − 6uDx − 3ux (101.152)
and the Darboux transformation.

We have in the case under consideration
Lψ = λψ =⇒ (Dx2 + u)ψ = λψ =⇒ ψxx + uψ = λψ
1303
and thus
ψxx = (λ − u)ψ (101.153)
On the other hand, we will also have (Theorem 1)
ψxx [1] = (λ − u[1])ψ[1] (101.154)
where
ψ10
ψ[1] = ψx − σψ, σ= (101.155)
ψ1
Combining the above two equations (101.154) and (101.155), we obtain the
equation
ψxx [1] = (λ − u[1])(ψx − σψ)
(101.156)
= (λ − u[1])ψx − σ(λ − u[1])ψ
Differentiating twice (101.155) we obtain the equation
ψxx [1] = ψxxx − 2σx ψx − σψxx − σxx ψ (101.157)
We have next to calculate ψxxx . In fact, differentiation of (101.153) with

respect to x, we get
ψxxx = (λ − u)ψx − ux ψ (101.158)
Substituting ψxxx and ψxx given by (101.158) and (101.153) respectively in

(101.157), yields the equation
ψxx [1] = {(λ − u)ψx − ux ψ} − 2σx ψx − σ(λ − u)ψ − σxx ψ

(101.159)
= (λ − u − 2σx )ψx − {ux + σxx + σ(λ − u)}ψ
We have thus obtained two expressions for ψxx [1], the one given by (101.156)
and the other given by (101.159). Equating the coefficients of ψx in these
two expressions we get λ − u[1] = λ − u − 2σx , i.e.
u[1] = u + 2σx (101.160)
We can take u = 0 for simplicity. Therefore for u = 0 the Lax pair becomes
L = Dx2 , M = −4Dx3 (101.161)
1304
and thus
( (
Lψ1 = λψ1 ψ1xx = λψ1
=⇒ (101.162)
ψ1t = M ψ1 ψ1t = −4ψ1xxx
With ψ1 = F (x, t), we have to solve the system {Fxx = λF, Ft = −4Fxxx }
Fxx = λF, Ft = −4λFx (101.163)
Equation Ft = −4λFx admits general solution given by

−x + 4λt
F = G(ξ), ξ= (101.164)
4λ
Equation Fxx = λF yields
G00 (ξ) − 16λ3 G(ξ) = 0 (101.165)
Taking
k2
λ= (101.166)
4
k 3 2
G00 (ξ) − G(ξ) = 0 (101.167)
2
k3 k3
G(ξ) = C1 ·exp ξ + C2 ·exp − ξ (101.168)
2 2
We thus have found that
k3 k3 1
ψ1 = C1 ·exp ξ + C2 ·exp − ξ , ξ=− x+t (101.169)
2 2 k2
Therefore the solution of the KdV equation is given by (101.60) for u = 0,
i.e. u[1] = 2σx , or removing the bracket, by
ψ1x
u = 2σx , σ= (101.170)
ψ1
1305
with ψ1 given by (101.169).
We obtain from (101.69) by differentiation with respect to x
1
k3 k 3
ψ1x = − k C1 ·exp ξ − C2 ·exp − ξ (101.171)
2 2 2
and thus
1
k3 k 3
− k C1 ·exp ξ − C2 ·exp − ξ
ψ1x 2 2 2
σ= = 3 3
ψ1 k k
C1 ·exp ξ + C2 ·exp − ξ
2 2
The above expression can also be written as (with C = C1 /C2 )
k3 k3
1 C·exp 2 ξ − exp − 2 ξ
σ=− k k3 k3 (101.172)
2
C·exp ξ + exp − ξ
2 2
Differentiation with respect to x yields (taking C = 1)
k2
σx = 1 1 2 (101.173)
2 2
exp k(−x + k t) + exp − k(−x + k t)
2 2

1 1
σx = k 2 sech2 k(x − k 2 t) (101.174)
4 2
Taking into account (101.170) and (101.174) we obtain the solution of the
KdV equation:
1 1
u = k 2 sech2 k(x − k 2 t) (101.175)
2 2
Example 5. We shall determine the solution of the Boussinesq equation
(Gu, Hu and Zhou [11], pages 86-87)
3utt + (6uux + uxxx )x = 0 (101.176)
1306
using the Lax pairs
3
L = Dx3 + uDx + v, M = Dx2 + u, 4vx = 3ut + 3uxx (101.177)
2
and the Darboux transformation.
We have in the case under consideration

3 3
Lψ = λψ =⇒ Dx + uDx + v ψ = λψ
2
3
=⇒ ψxxx + uψx + vψ = λψ
2
and thus
3
ψxxx = (λ − v)ψ − uψx (101.178)
2
On the other hand, we will also have (Theorem 1)
3
ψxxx [1] = (λ − v[1])ψ[1] − u[1]ψx [1] (101.179)
2
where
ψ10
ψ[1] = ψx − σψ, σ= (101.180)
ψ1
From the above expression we get by differentiation with respect to x
ψx [1] = ψxx − σx ψ − σψx (101.181)
Substituting the above two expressions (101.180) and (101.1181) into (101.179),
3
ψxxx [1] = (λ − v[1])(ψx − σψ) − u[1](ψxx − σx ψ − σψx )
2
3 3
= − u[1]ψxx + λ − v[1] + σu[1] ψx (101.182)
2
2
3
+ σx u[1] − σ(λ − v[1]) ψ
2
Differentiating twice (101.181) we obtain the equation
ψxxx [1] = ψxxxx − 3σxx ψx − 3σx ψxx − σψxxx − σxxx ψ (101.183)
1307
We have to calculate ψxxxx next. In fact, differentiation of (101.178) with
respect to x, we get

3 3
ψxxxx = − uψxx + λ − v − ux ψx − vx ψ (101.184)
2 2
Substituting ψxxxx and ψxxx given by (101.184) and (101.178) respectively in
(101.183), yields the equation
3 3 3
ψxxx [1] = − u − 3σx ψxx + λ − v − 3σxx − ux + σu ψx
2 2 2 (101.185)
+ {−vx − σxxx + σ(v − λ)}ψ
We have thus obtained two expressions for ψxxx [1], one given by (101.182)
and the other by (101.185). Equating the coefficients of ψxx in these two
3 3
expressions, we get − u[1] = − u − 3σx , i.e.
2 2
u[1] = u + 2σx (101.186)
We can take u = 0 for simplicity. Therefore for u = 0 the Lax pair becomes
L = Dx3 , M = Dx2 (101.187)
and thus
( (
Lψ1 = λψ1 ψ1xxx = λψ1
=⇒ (101.188)
ψ1t = M ψ1 ψ1t = ψ1xx
For ψ1 = F and λ = k 3 we have to solve the system
Fxxx = k 3 F, Ft = Fxx (101.189)
The secular equation for Fxxx = k 3 F is m3 − k 3 = 0 with roots given by
m1 = k,
m2 = ηk, m3 = θk
√ √
1 3 1 3 (101.190)
η=− − i, θ = − + i, η 2 = θ, θ2 = η
2 2 2 2
Therefore the general solution of Fxxx = k 3 F is given by
F = F1 (t)eηkx + F2 (t)eθkx + F3 (t)ekx (101.191)
1308
The above solution should satisfy equation Ft = Fxx , i.e.
{F10 (t) − k 2 η 2 F1 (t)}eηkx + {F20 (t) − k 2 θ2 F2 (t)}eθkx
(101.192)
+ {F30 (t) − k 2 F3 (t)}ekx = 0
F10 (t) − k 2 η 2 F1 (t) = 0, F20 (t) − k 2 θ2 F2 (t) = 0,
(101.193)
F30 (t) − k 2 F3 (t) = 0
2 θt 2 ηt 2t
F1 (t) = C1 ek , F2 (t) = C2 ek , F3 (t) = C3 ek (101.194)
where we have taken into account that η 2 = θ and θ2 = η.

We thus have the following solution of the system (101.189):
F = C1 exp{k(ηx + kθt)} + C2 exp{k(θx + kηt)}

(101.195)
+ C3 exp{k(x + kt)}
We have further
ψ1x Fx
σ= =
ψ1 F
and thus
C1 η exp(f1 ) + C2 θ exp(f2 ) + C3 exp(f3 )
σ=k (101.196)
C1 exp(f1 ) + C2 exp(f2 ) + C3 exp(f3 )
where
f1 = k(ηx + kθt), f2 = k(θx + kηt), f3 = k(x + kt) (101.197)
Taking for simplicity C1 = C2 = 1 and C3 = 0, the expression for σ given by

(101.196) gets simplified into
1

√ √3
σ = k − 1 + 3 tan k(−x + kt) (101.198)
2 2
Therefore the solution of the Boussinesq equation is given by
√
3 2 2 3
u = 2σx = − k sec k(−x + kt) (101.199)
2 2
1309
Note 1. The transition from (101.196) to (101.198) was implemented using
the commands ComplexExpand[ ] and then Simplify[ ] of ”Mathematica”.
Note 2. Taking C1 = C2 = C3 = 1 in (101.196) we get
√
ef + 3 sin(h) − cos(h)
σ=k (101.200)
ef + 2 cos(h)
with
√
3 3
f = k(x + kt) and h = k(−x + kt) (101.201)
2 2
The solution is then given by
√
1 − ef [cos(h) − 3 sin(h)]
2
u = 2σx = −6k (101.202)
[ef + 2 cos(h)]2
or in full form
N
u = −6k 2 (101.203)
D
where
3
N = 1 − exp k(x + kt)
2
√ √ √3 (101.204)
3
× cos k(−x + kt) − 3 sin k(−x + kt)
2 2
and
3 √3 2
D = exp k(x + kt) + 2 cos k(−x + kt) (101.205)
2 2
Note 3. Considering the Boussinesq equation (Matveev and Salle [10], §3.6)
1
utt − uxx + (u2 )xx + uxxxx = 0 (101.206)
3
we have to consider the Lax pair
3
L = Dx3 + (2u − 1)Dx + v, 4vx = 3uxx − 3ut
4 (101.207)
2
M = −Dx + uI
1310
References
[1] Z. S. Agranovich and V. A. Marchenko: ”The Inverse Problem in
Scattering Theory”. Gordon and Breach 1963
[2] M. J. Ablowitz and H. Segur: ”Solitons and the Inverse Scattering
Transform”. SIAM 1981
[3] V. E. Zakharov, S. V. Manakov, S. P. Novikov and L. P. Pitaevskii:
”Theory of Solitons. The Method of the Inverse Scattering Problem”.
Consultants Bureau, New York 1984
[4] M. A. Ablowitz and P. A. Clarkson: ”Solitons, Nonlinear Evolution
Equations and Inverse Scattering”. Cambridge University Press (1991)
[5] D. Vvedensky: ”Partial Differential Equations with Mathematica”.
Addison-Wesley (1994)
[6] P. D. Lax: ”Integrals of Nonlinear Equations of Evolution and Solitary
Waves”. Comm. Pure Appl. Math. 21 (1968) 467-490
[7] M. J. Ablowitz, D. J. Kaup, A. C. Newell and H. Segur. ”The
inverse scattering transform - Fourier analysis for nonlinear problems”.
Stud. Appl. Math. 53 (1974) 249-315
[8] J. Larue: ”Symbolic Verification of Operator and Matrix Lax Pairs for
some Completely Integrable Nonlinear Partial Differential Equations”.
Thesis, School of Mines, Colorado 2011
[9] M. Hickman, W. Hereman, J. Larue and U. Göktas: ”Scaling
Invariant Lax Pairs of Nonlinear Evolution Equations”. arXiv:1110.0586v1
[10] V. B. Matveev and M. A. Salle. ”Darboux Transformation and
Solitons”. Springer 1991
[11] G. Gu, H. Hu and Z. Zhou: ”Darboux Transformations in Integrable
Systems”. Springer 2005
102 The Hirota Direct Method

This method was introduced by R. Hirota (Hirota [1], [2]) in determining
multi-soliton solutions of the KdV equation
ut + 6uux + uxxx = 0 (102.1)
Hirota’s analysis starts with the introduction of the bilinear form

∂ ∂
Dx (f ·g) = − f (x)·g(x0 ) (102.2)
∂x ∂x0 x0 =x
1311
Dx (f ·g) = fx g − f gx (102.3)
The most general bilinear form is given by

∂ ∂ m ∂ ∂ n
Dxm Dtn (f ·g) = − 0 − 0 f (x, t)·g(x0 , t0 ) (102.4)
∂x ∂x ∂t ∂t x0 =x, t0 =t
The term bilinear comes from the fact that it acts on two functions.
Some useful identities are the following
∂ mf
Dxm (f ·1) = (102.5)
∂xm
Dxm (f ·g) = (−1)m Dxm (g·f ) (102.6)
Dxm (f ·f ) = 0, for m odd (102.7)
Dxm Dtn (ek1 x+ω1 t ·ek2 x+ω2 t )
(102.8)
= (k1 − k2 )m (ω1 − ω2 )n e(k1 +k2 )x+(ω1 +ω2 )t
Let P (Dx , Dt ) be a polynomial in Dx and Dt . Then it follows from (102.5)
and (102.8) that
P (k1 − k2 , ω1 − ω2 )
P (Dx , Dt )(ek1 x+ω1 t ·ek2 x+ω2 t ) =
P (k1 + k2 , ω1 − ω2 ) (102.9)

(k1 +k2 )x+(ω1 +ω2 )t
×P (Dx , Dt ) e ·1
The above is a very useful identity in implementing Hirota’s method.

We consider the KdV equation (102.1) in applying Hirota’s method. We
shall first transform KdV equation into the so-called Hirota bilinear form.
Let us consider a smooth function f (x, t). We introduce the transformation
∂2
u(x, t) = 2 ln(f (x, t)) (102.10)
∂x2
i.e.
f fxx − fx2
u(x, t) = 2 (102.11)
f2
1312
Since
2 4
ut = 2
(ft fxx + f fxxt − 2fx fxt ) − 3 (f fxx − fx2 )ft (102.12)
f f
2 4
ux = 2
(−fx fxx + f fxxx ) − 3 (f fxx − fx2 )fx (102.13)
f f
and
48 120 20
uxxx = 5
(fx )5 − 4 (fx )3 fxx + 3 {3fx (fxx )2 + 2(fx )2 fxxx }
f f f
(102.14)
10 2
− 2 (fx f4x + 2fxx fxxx ) + f5x
f f
the KdV equation (102.1) takes on the form

2
− 3 6fx (fxx )2 + 5f fx f4x + f ft fxx − 8(fx )2 f3x − 2ft (fx )2
f
(102.15)
2 2
− 2f fxx f3x + 2f fx fxt − f fxxt − f f5x = 0
The above equation can also be written in equivalent form as

∂ 2 2
f fxt − fx ft + f f4x − 4fx f3x + 3(fxx ) =0 (102.16)
∂x f 2
Integrating the above equation once and setting the integration constant
equal to zero, we obtain the equation
f fxt − fx ft + f f4x − 4fx f3x + 3(fxx )2 = 0 (102.17)
This is the bilinear form of the KdV equation. The above equation has to
be expressed in terms of Hirota’s D−operators. In fact we have, according
to the definition of Dx and Dt
Dt Dx (f ·f ) = 2(f fxt − fx ft ) (102.18)
and
Dx4 (f ·f ) = 2(f f4x − 4fx f3x + 3(fxx )2 ) (102.19)
1313
Therefore equation (102.17) can be expressed as
(Dt Dx + Dx4 )(f ·f ) = 0 (102.20)
or even as
P (D)(f ·f ) = 0, P (D) = Dt Dx + Dx4 (102.21)
Therefore if f is a solution of (102.20), then f can be used in determining

the function u(x, t) from (102.11), i.e. the solution of the KdV equation.
At this atage we consider the perturbative expansion of the function f
f = 1 + f1 + 2 f2 + 3 f3 + · · · (102.22)
We then calculate the product f ·f
f ·f = (1 + f1 + 2 f2 + 3 f3 + · · · )
·(1 + f1 + 2 f2 + 3 f3 + · · · )
(102.23)
= 1·1 + (f1 ·1 + 1·f1 ) + 2 (f2 ·1 + 1·f2 + f1 ·f1 )
+ 3 (f3 ·1 + f2 ·f1 + f1 ·f2 + 1·f3 ) + · · ·
If P (D) is any polynomial operator (like for example the one in (102.21))
with
P (D)(f ·f ) = 0 (102.24)
then since P (D) is supposed to be a linear operator,
P (D)(f ·f ) = P (D)(1·1) + P (D)(f1 ·1 + 1·f1 )

+ 2 P (D)(f2 ·1 + f1 ·f1 + 1·f2 ) (102.25)
3
+ P (D)(f3 ·1 + f2 ·f1 + f1 ·f2 + 1·f3 ) + · · · = 0
Since the previous equation holds for any , we should have
coefficient of 0 : P (D)(1·1) = 0 (102.26)
coefficient of 1 : P (D)(f1 ·1 + 1·f1 ) = 0 (102.27)

coefficient of 2 : P (D)(f2 ·1 + f1 ·f1 + 1·f2 ) = 0 (102.28)
coefficient of 3 : P (D)(f3 ·1 + f2 ·f1 + f1 ·f2 + 1·f3 ) = 0 (102.29)
1314
and in general
n
X
coefficient of n : P (D) fk ·fn−k = 0, f0 = 1 (102.30)
k=0
θi = ki x + ωi t, fi = eθi = eki x+ωi t (102.31)
we can determine the soliton solutions of PDEs in the following cases
I. one-soliton solution :
(102.32)
f (x, t) = 1 + f1 , f1 = eθ1
II. two-soliton solution :

(102.33)
f (x, t) = 1 + f1 + 2 f2 , f1 = eθ1 + eθ2
III. three-soliton solution :
(102.34)
f (x, t) = 1 + f1 + 2 f2 + 3 f3 , f1 = eθ1 + eθ2 + eθ3
and so on. At the end of the calculation we may set = 1 for simplicity.
Some useful identities might be the following
P (D)(f ·1) = P (∂)f and P (D)(1·f ) = P (−∂)f (102.35)
and since P (D) is always even, P (−∂) = P (∂) and thus
P (D)(f ·1 + 1·f ) = 2P (∂)f (102.36)
for any P (D) and any f . We also have
P (Dx )(eθ1 ·eθ2 ) = P (k1 − k2 )eθ1 +θ2 (102.37)
P (Dt )(eθ1 ·eθ2 ) = P (ω1 − ω2 )eθ1 +θ2 (102.38)

P (Dx , Dt )(eθ1 ·eθ2 ) = P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 (102.39)
Note. The reader will easily recognise the similarity between the equation
(102.10) and the WTC method of Painlevé analysis. The interested reader
can also find details to the relevant paper
J. D. Gibbon, P. Radmore, M. Tabor and D. Wood: ”The Painlevé Property
and Hirota’s Method”, Stud. Appl. Math. 72 (1985) 39-63.
1315
102.1 One-soliton solution of the KdV equation. For the KdV equation
we have derived the equation (102.21) and since
f = 1 + f1 , f1 = eθ1 (102.40)
we have
f ·f = 1·1 + (1·f1 + f1 ·1) + 2 f1 ·f1 (102.41)
and thus P (D)(f ·f ) = 0 is equivalent to
P (D)(1·1) + P (D)(1·f1 + f1 ·1) + 2 P (D)(f1 ·f1 ) = 0 (102.42)
From the above equation we obtain the following three equations
P (D)(1·1) = 0, P (D)(1·f1 + f1 ·1) = 0, P (D)(f1 ·f1 ) = 0 (102.43)
The first of the above equations is a trivial identity. The second equation
yields (considering (102.36))
P (D)(1·f1 + f1 ·1) = 0 ⇐⇒ 2P (∂)(1·f1 + f1 ·1) = 0

⇐⇒ 2(∂x ∂t + ∂x4 )eθ1 = 0 ⇐⇒ (k1 ω1 + k14 )eθ1 = 0 (102.44)
⇐⇒ k1 ω1 + k14 = 0
Solving with respect to ω1 we get
ω1 = −k13 (102.45)
which is the dispersion relation for ω.

The third equation of (102.43) yields (considering (102.39))
P (D)(f1 ·f1 ) = 0 ⇐⇒ P (k1 − k1 , ω1 − ω1 )e2θ1 = 0 (102.46)
which is a trivial identity.

3 2
Since ω1 = −k13 , we obtain f1 = ek1 x−k1 t = ek1 (x−k1 t) and thus (for = 1 and
k1 = k)
2 t)
f = 1 + f1 = 1 + ek(x−k (102.47)
Considering equation (102.11), since

2 t)
f fxx − (fx )2 = k 2 ek(x−k (102.48)
1316
we get
2
f fxx − fx2 k 2 ek(x−k t)
u(x, t) = 2 =2
f2 {1 + ek(x−k2 t) }2

1 1
u(x, t) = k 2 sech2 k(x − k 2 t) (102.49)
2 2
which is the famous one-soliton solution of the KdV equation.
102.2 Two-soliton solution of the KdV equation. For the two-soliton
solution of the KdV equation we have to consider
f = 1 + f1 + 2 f2 , f1 = eθ1 + eθ2 (102.50)

f ·f = 1·1 + (1·f1 + f1 ·1) + 2 (1·f2 + f1 ·f1 + f2 ·1)
(102.51)
+ 3 (f1 ·f2 + f2 ·f1 ) + 4 (f2 ·f2 )
P (D)(1·1) + P (D)(1·f1 + f1 ·1) + 2 P (D)(1·f2 + f1 ·f1 + f2 ·1)

(102.52)
+ 3 P (D)(f1 ·f2 + f2 ·f1 ) + 4 P (D)(f2 ·f2 ) = 0
From the above equation we obtain the following equations
P (D)(1·1) = 0, P (D)(1·f1 + f1 ·1) = 0,

P (D)(1·f2 + f1 ·f1 + f2 ·1) = 0, P (D)(f1 ·f2 + f2 ·f1 ) = 0, (102.53)
P (D)(f2 ·f2 ) = 0
Equation P (D)(1·1) = 0 is a trivial identity. Equation P (D)(1·f1 + f1 ·1) = 0

with f1 = eθ1 + eθ2 yields
2P (∂)f1 = 0 ⇐⇒ 2P (∂)(eθ1 + eθ2 ) = 0

⇐⇒P (∂)eθ1 + P (∂)eθ2 = 0
(102.54)
⇐⇒(∂x ∂t + ∂x4 )eθ1 + (∂x ∂t + ∂x4 )eθ2 = 0
⇐⇒(k1 ω1 + k14 )eθ1 + (k2 ω2 + k24 )eθ2 = 0
1317
From the above equation we obtain the two equations k1 ω1 + k14 = 0 and
k2 ω2 + k24 = 0 and then solving with respect to ω1 and ω2 we get
ω1 = −k13 and ω2 = −k23 (102.55)
respectively. These are the two dispersion relations for ω1 and ω2 .

Equation P (D)(1·f2 + f1 ·f1 + f2 ·1) = 0 is equivalent to
P (D)(1·f2 + f2 ·1) + P (D)(f1 ·f1 ) = 0 (102.56)
Since f1 = eθ1 + eθ2 , taking into account P (D)(eθi ·eθi ) = 0, we get
P (D)(f1 ·f1 ) = P (D)((eθ1 + eθ2 )·(eθ1 + eθ2 )) = 2P (D)(eθ1 ·eθ2 )

(102.57)
= 2P (k1 − k2 , ω1 − ω2 )eθ1 +θ2
Therefore equation (102.56) is equivalent to
2P (∂)f2 + 2P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 = 0 (102.58)
From the above equation we conclude that f2 should have the form
f2 = A(1, 2)eθ1 +θ2 (102.59)
Because of the above expression, equation (102.58) becomes
A(1, 2)P (∂)eθ1 +θ2 + P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 = 0 (102.60)
A(1, 2)P (k1 + k2 , ω1 + ω2 )eθ1 +θ2 + P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 = 0

(102.61)
⇐⇒ A(1, 2)P (k1 + k2 , ω1 + ω2 ) + P (k1 − k2 , ω1 − ω2 ) = 0
from which we get
P (k1 − k2 , ω1 − ω2 )
A(1, 2) = − (102.62)
P (k1 + k2 , ω1 + ω2 )
We thus see that f2 has the form
P (k1 − k2 , ω1 − ω2 ) θ1 +θ2
f2 = − e (102.63)
P (k1 + k2 , ω1 + ω2 )
1318
Since P (D) = Dx Dt + Dx4 , we have, taking into account ω1 = −k13 and
ω2 = −k23 , that
P (k1 − k2 , ω1 − ω2 ) (k1 − k2 )(ω1 − ω2 ) + (k1 − k2 )4
=
P (k1 + k2 , ω1 + ω2 ) (k1 + k2 )(ω1 + ω2 ) + (k1 + k2 )4
(k1 − k2 )(−k13 + k23 ) + (k1 − k2 )4 −3k1 k2 (k1 − k2 )3
= = (102.64)
(k1 + k2 )(−k13 − k23 ) + (k1 + k2 )4 3k1 k2 (k1 + k2 )3
(k1 − k2 )2
=−
(k1 + k2 )2
Therefore f2 becomes
(k1 − k2 )2 θ1 +θ2
f2 = e (102.65)
(k1 + k2 )2
and f is given by
(k1 − k2 )2 θ1 +θ2
f = 1 + eθ1 + eθ2 + e ,
(k1 + k2 )2 (102.66)
θ1 = k1 x − k13 t = k1 (x − k12 t), θ2 = k2 x − k23 t = k2 (x − k22 t)
i.e.
2 2 (k1 − k2 )2 (k1 +k2 )x−(k13 +k23 )t
f = 1 + ek1 (x−k1 t) + ek2 (x−k2 t) + e (102.67)
(k1 + k2 )2
The last two equations of (102.53) become identities.
The two-soliton solution is then given by
f fxx − (fx )2
u(x, t) = 2 (102.68)
f2
with f given in (102.67).
102.3 Three-soliton solution of the KdV equation. For the three-
soliton solution of the KdV equation we have to consider
f = 1 + f1 + 2 f2 + 3 f3 , f1 = eθ1 + eθ2 + eθ3 (102.69)
f ·f = 1·1 + (1·f1 + f1 ·1) + 2 (1·f2 + f1 ·f1 + f2 ·1)
+ 3 (1·f3 + f1 ·f2 + f2 ·f1 + f3 ·1)
(102.70)
+ 4 (f1 ·f3 + f2 ·f2 + f3 ·f1 ) + 5 (f2 ·f3 + f3 ·f2 )
+ 6 (f3 ·f3 )
1319
P (D)(1·1) + P (D)(1·f1 + f1 ·1) + 2 P (D)(1·f2 + f1 ·f1 + f2 ·1)
+ 3 P (D)(1·f3 + f1 ·f2 + f2 ·f1 + f3 ·1)
(102.71)
+ 4 P (D)(f1 ·f3 + f2 ·f2 + f3 ·f1 ) + 5 P (D)(f2 ·f3 + f3 ·f2 )
+ 6 P (D)(f3 ·f3 ) = 0
From the above equation we obtain the following equations
P (D)(1·1) = 0, P (D)(1·f1 + f1 ·1) = 0,
P (D)(1·f2 + f1 ·f1 + f2 ·1) = 0,
P (D)(1·f3 + f1 ·f2 + f2 ·f1 + f3 ·1) = 0, (102.72)
P (D)(f1 ·f3 + f2 ·f2 + f3 ·f1 ) = 0,
P (D)(f2 ·f3 + f3 ·f2 ) = 0, P (D)(f3 ·f3 ) = 0
We shall next determine the implications of each one of the above equations.
Equation P (D)(1·1) = 0 is a trivial identity. Equation P (D)(1·f1 + f1 ·1) = 0
for f1 = eθ1 + eθ2 + eθ3 takes the equivalent form
2P (∂)f1 = 0 ⇐⇒ P (∂)eθ1 + P (∂)eθ2 + P (∂)eθ3 = 0
⇐⇒(∂x ∂t + ∂x4 )eθ1 + (∂x ∂t + ∂x4 )eθ2 + (∂x ∂t + ∂x4 )eθ3 = 0 (102.73)
⇐⇒(k1 ω1 + k14 )eθ1 + (k2 ω2 + k24 )eθ2 + (k3 ω3 + k34 )eθ3 = 0
From the above equation we obtain the equations k1 ω1 +k14 = 0, k2 ω2 +k24 = 0
and k3 ω3 + k34 = 0. Solving with respect to ω1 , ω2 and ω3 , we get
ω1 = −k13 , ω2 = −k23 , ω3 = −k33 (102.74)
which are the dispersion relations for ω1 , ω2 and ω3 .
Equation P (D)(1·f2 + f1 ·f1 + f2 ·1) = 0 is equivalent to
P (D)(1·f2 + f2 ·1) + P (D)(f1 ·f1 ) = 0 (102.75)
Since f1 = eθ1 + eθ2 + eθ3 , taking into account P (D)(eθi ·eθi ) = 0, we get
P (D)(f1 ·f1 ) = P (D)((eθ1 + eθ2 + eθ3 )·(eθ1 + eθ2 + eθ3 ))
= 2P (D)(eθ1 ·eθ2 ) + 2P (D)(eθ2 ·eθ3 ) + 2P (D)(eθ3 ·eθ1 )
= 2P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 (102.76)
θ2 +θ3
+ 2P (k2 − k3 , ω2 − ω3 )e
+ 2P (k3 − k1 , ω3 − ω1 )eθ3 +θ1
1320
Therefore equation (102.75) is equivalent to
2P (∂)f2 + 2P (k1 − k2 , ω1 − ω2 )eθ1 +θ2

(102.77)
+ 2P (k2 − k3 , ω2 − ω3 )eθ2 +θ3 + 2P (k3 − k1 , ω3 − ω1 )eθ3 +θ1 = 0
From the above equation we see that f2 should have the form
f2 = A(1, 2)eθ1 +θ2 + A(2, 3)eθ2 +θ3 + A(3, 1)eθ3 +θ1 (102.78)
Because of the above expression, equation (102.77) can be written as
A(1, 2)P (∂)eθ1 +θ2 + A(2, 3)P (∂)eθ2 +θ3 + A(3, 1)P (∂)eθ3 +θ1
+ P (k1 − k2 , ω1 − ω2 )eθ1 +θ2 + P (k2 − k3 , ω2 − ω3 )eθ2 +θ3
+ P (k3 − k1 , ω3 − ω1 )eθ3 +θ1 = 0
{A(1, 2)P (k1 + k2 , ω1 + ω2 ) + P (k1 − k2 , ω1 − ω2 )}eθ1 +θ2

+ {A(2, 3)P (k2 + k3 , ω2 + ω3 ) + P (k2 − k3 , ω2 − ω3 )}eθ2 +θ3
+ {A(3, 1)P (k3 + k1 , ω3 + ω1 ) + P (k3 − k1 , ω3 − ω1 )}eθ3 +θ1 = 0
From the above equation we obtain the system of three equations
A(1, 2)P (k1 + k2 , ω1 + ω2 ) + P (k1 − k2 , ω1 − ω2 ) = 0
A(2, 3)P (k2 + k3 , ω2 + ω3 ) + P (k2 − k3 , ω2 − ω3 ) = 0
A(3, 1)P (k3 + k1 , ω3 + ω1 ) + P (k3 − k1 , ω3 − ω1 ) = 0
from which we obtain the values of the three coefficients:
P (k1 − k2 , ω1 − ω2 )
A(1, 2) = −
P (k1 + k2 , ω1 + ω2 )
P (k2 − k3 , ω2 − ω3 )
A(2, 3) = −
P (k2 + k3 , ω2 + ω3 )
P (k3 − k1 , ω3 − ω1 )
A(3, 1) = −
P (k3 + k1 , ω3 + ω1 )
The above three expressions can be written as
(k1 − k2 )2 (k2 − k3 )2 (k3 − k1 )2
A(1, 2) = , A(2, 3) = , A(3, 1) = (102.79)
(k1 + k2 )2 (k2 + k3 )2 (k3 + k1 )2
1321
using the dispersion relations (102.74), as we did in the two-soliton case.
Equation P (D)(1·f3 + f1 ·f2 + f2 ·f1 + f3 ·1) = 0 can be written as
P (D)(1·f3 + f3 ·1) + P (D)(f1 ·f2 + f2 ·f1 ) = 0

(102.80)
⇐⇒2P (∂)f3 + 2P (∂)(f1 ·f2 ) = 0
For f1 = eθ1 + eθ2 + eθ3 and f2 given by (102.78), we find

θ1 +θ2 +θ3
P (∂)(f1 ·f2 ) = e × A(1, 2)P (k3 − k1 − k2 )
(102.81)
+ A(1, 3)P (k2 − k1 − k3 ) + A(2, 3)P (k1 − k2 − k3 )
Using (102.80) and (102.81) we see that f3 should have the form
f3 = B(1, 2, 3)eθ1 +θ2 +θ3 (102.82)
Therefore we obtain from (102.80)

θ1 +θ2 +θ3 θ1 +θ2 +θ3
B(1, 2, 3)P (∂)(e )+e × A(1, 2)P (k3 − k1 − k2 )
(102.83)
+ A(1, 3)P (k2 − k1 − k3 ) + A(2, 3)P (k1 − k2 − k3 ) = 0

B(1, 2, 3)P (k1 + k2 + k3 , ω1 + ω2 + ω3 ) + A(1, 2)P (k3 − k1 − k2 )
(102.84)
+ A(1, 3)P (k2 − k1 − k3 ) + A(2, 3)P (k1 − k2 − k3 ) = 0
from which we obtain the value of B(1, 2, 3):

P
AP
B(1, 2, 3) = − (102.85)
P (k1 + k2 + k3 , ω1 + ω2 + ω3 )
where
X
AP = A(1, 2)P (k3 − k1 − k2 ) + A(1, 3)P (k2 − k1 − k3 )
(102.86)
+ A(2, 3)P (k1 − k2 − k3 )
1322
Equation P (D)(f1 ·f3 + f2 ·f2 + f3 ·f1 ) = 0 can be converted to the equivalent
form (B = B(1, 2, 3))

2θ1 +θ2 +θ3
e BP (k2 + k3 , ω2 + ω3 ) + A(1, 2)A(1, 3)P (k2 − k3 , ω2 − ω3 )

θ1 +2θ2 +θ3
+e BP (k1 + k3 , ω1 + ω3 ) + A(1, 2)A(2, 3)P (k1 − k3 , ω1 − ω3 )

θ1 +θ2 +2θ3
+e BP (k1 + k2 , ω1 + ω2 ) + A(1, 3)A(2, 3)P (k1 − k2 , ω1 − ω2 ) = 0
From the above equation we see that B(1, 2, 3) should have the form
B(1, 2, 3) = A(1, 2)A(2, 3)A(1, 3) (102.87)
The two last equations of (102.72) P (D)(f2 ·f3 + f3 ·f2 ) = 0, P (D)(f3 ·f3 ) = 0
are satisfied identically.
Let us collect everything together. We have for = 1 that
f = 1 + eθ1 + eθ2 + eθ3

+ A(1, 2)eθ1 +θ2 + A(2, 3)eθ2 +θ3 + A(3, 1)eθ3 +θ1 (102.88)
+ A(1, 2)A(2, 3)A(3, 1)eθ1 +θ2 +θ3
where A(i, j) are given in (102.79).

At this point we should mention that using (102.84) and (102.87) we obtain
the consistency condition
A(1, 2)A(2, 3)A(3, 1)P (k1 + k2 + k3 , ω1 + ω2 + ω3 )

+ A(1, 2)P (k3 − k1 − k2 ) + A(1, 3)P (k2 − k1 − k3 )
(102.89)

+ A(2, 3)P (k1 − k2 − k3 ) = 0
The three-soliton solution is then given using
f fxx − (fx )2
u(x, t) = 2 (102.90)
f2
with f given in (102.88).
There is a straightforward generalization of Hirota’s method to the N-soliton
1323
solutions. The reader should at this point consult the book by Hirota (Ref.
[2]) and the relevant references.
102.4 Some other equations solved by Hirota’s Method. We list
below some other equations which have been solved using Hirota’s method:
I. The Boussinesq Equation. The Boussinesq equation
utt − uxx − 3(u2 )xx − uxxxx = 0 (102.91)
∂2 f fxx − (fx )2
u(x, t) = 2 2 ln(f (x, t)) = 2 (102.92)
∂x f2
takes on the form

1
4
− 2f 3 f6x + 12f 2 fx f5x + (−36f fx2 + 6f 2 fxx − 2f 3 )f4x
f
+ 2f 3 fttxx − 4f 2 f3x
2
+ (48fx3 + 8f 2 fx )f3x − 4f 2 fx ftxx
− 4f 2 ft ftxx + 12f fxx
3
+ (6f 2 − 36fx2 )fxx
2
− 2f (12fx2 − 2ft2 + f ftt )fxx + 16f fx ft ftx − 12fx2 ft2

4 2 2 2
+ 12fx − 4f ftx + 4f fx ftt = 0

∂2 1 2

2 4
Dt − Dx − Dx (f ·f ) = 0
∂x2 f 2
Integrating twice and setting the integration constants equal to zero, we get
the equation

2 2 4
Dt − Dx − Dx (f ·f ) = 0 (102.93)
We are looking for the two-soliton solution of the Boussinesq equation. In

this case we consider the choice
f = 1 + f1 + 2 f2 , f1 = eθ1 + eθ2 (102.94)
and follow exactly the same steps as in the KdV equation. We thus arrive at
the following form of the function f (after setting = 1)
f = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2
P (k1 − k2 , ω1 − ω2 ) (102.95)
A(1, 2) = − , P (D) = Dt2 − Dx2 − Dx4
P (k1 + k2 , ω1 + ω2 )
1324
The dispersion relations are evaluated to have the form
q q
ω1 = ± 1 + k1 , ω2 = ± 1 + k22
2
(102.96)
Using the above values of ω1 and ω2 with the negative sign, we arrive at the
following explicit form of A(1, 2):
p p
1 + k12 1 + k22 − 1 − 2k12 − 2k22 + 3k1 k2
A(1, 2) = p p (102.97)
1 + k12 1 + k22 − 1 − 2k12 − 2k22 − 3k1 k2
The two-soliton solution of the Boussinesq equation has the form

N N1 + N2 + N3
u(x, t) = 2 2
=2 (102.98)
D D2
where
N1 = A(1, 2)[k12 eθ2 + k22 eθ1 + (k1 + k2 )2 ]eθ1 +θ2
N2 = [(k1 − k2 )2 eθ2 + k12 ]eθ1
(102.99)
N3 = k22 eθ2
D = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2
II. The Sawada-Kotera Equation. The Sawada-Kotera equation
ut + 45u2 ux + 15ux uxx + 15uuxxx + u5x = 0 (102.100)

∂w(x, t)
u= (102.101)
∂x
is being converted into its potential form
wxt + 45wx2 wxx + 15wxx wxxx + 15wx w4x + w6x = 0 (102.102)
We consider the transformation

∂ fx
u(x, t) = 2 ln(f (x, t)) = 2 (102.103)
∂x f
1325
takes on the form

1
− 2fx ft − 12fx f5x + 30fxx f4x + 2f ftx + 2f f6x
f2

2
− 20f3x = 0

1 6

D x D t + Dx (f ·f ) = 0
f2
i.e. equivalent to

6
Dx Dt + Dx (f ·f ) = 0 (102.104)
We are looking for the two-soliton solution of the Sawada-Kotera equation.

In this case we consider the choice
f = 1 + f1 + 2 f2 , f1 = eθ1 + eθ2 (102.105)
f = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2

P (k1 − k2 , ω1 − ω2 ) (102.106)
A(1, 2) = − , P (D) = Dx Dt + Dx6
P (k1 + k2 , ω1 + ω2 )
The dispersion relations are evaluated to be
ω1 = −k14 , ω2 = −k24 (102.107)
Using the above values of ω1 and ω2 , we arrive at the following explicit form
of A(1, 2):
(k12 − k1 k2 + k22 )(k1 − k2 )2

A(1, 2) = 2 (102.108)
(k1 + k1 k2 + k22 )(k1 + k2 )2
The two-soliton solution of the Sawada-Kotera equation has the form
N N1 + N2 + N3
u(x, t) = 2 2
=2 (102.109)
D D2
1326
where
N1 = A(1, 2)[k12 eθ2 + k22 eθ1 + (k1 + k2 )2 ]eθ1 +θ2
N2 = [(k1 − k2 )2 eθ2 + k12 ]eθ1
(102.110)
N3 = k22 eθ2
D = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2
III. The 2-dimensional Kadomtsev-Petviashvili Equation. The two-

dimensional Kadomtsev-Petviashvili equation
(ut + 6uux + uxxx )x ±3uyy = 0 (102.111)

∂2 f fxx − (fx )2
u(x, t) = 2 ln(f (x, t)) = 2 (102.112)
∂x2 f2
takes on the form
∂2 1

4 2
Dx Dt + Dx ±3Dy (f ·f ) = 0
∂x2 f 2
Integrating twice and setting the integration constants equal to zero, we get
the equation

4 2
Dx Dt + Dx ±3Dy (f ·f ) = 0 (102.113)
We are looking for the two-soliton solution of the two-dimensional Kadomtsev-

Petviashvili equation. In this case we consider the choice
f = 1 + f1 + 2 f2 , f1 = eθ1 + eθ2

(102.114)
θ1 = k1 x + `1 y + ω1 t, θ2 = k2 x + `2 y + ω2 t
f = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2

P (k1 − k2 , `1 − `2 , ω1 − ω2 )
A(1, 2) = − , (102.115)
P (k1 + k2 , `1 + `2 , ω1 + ω2 )
P (D) = Dx Dt + Dx4 ±3Dy2
1327
The dispersion relations are evaluated to be
k14 + 3`21 k24 + 3`22

ω1 = − , ω2 = − (102.116)
k1 k2
Using the above values of ω1 and ω2 , we arrive at the following explicit form
of A(1, 2):
(k2 `1 + k2 k12 − k1 k22 − k1 `2 )(−k2 `1 + k2 k12 − k1 k22 + k1 `2 )

A(1, 2) =
(−k2 `1 + k2 k12 + k1 k22 + k1 `2 )(k2 `1 + k2 k12 + k1 k22 − k1 `2 )
(102.117)
The two-soliton solution of the Boussinesq equation has the form

N N1 + N2 + N3
u(x, t) = 2 2
=2 (102.118)
D D2
where
N1 = A(1, 2)[k12 eθ2 + k22 eθ1 + (k1 + k2 )2 ]eθ1 +θ2
N2 = [(k1 − k2 )2 eθ2 + k12 ]eθ1
(102.119)
N3 = k22 eθ2
D = 1 + eθ1 + eθ2 + A(1, 2)eθ1 +θ2
References
[1] R. Hirota: ”Exact solution of the Korteweg-de Vries equation for
multiple collision of solitons”. Phys. Rev. Lett. 27 (1971) 1192-1194
[2] R. Hirota: ”Direct Method in Soliton Theory”.
Cambridge Univ. Press 2004
[3] Y. Matsuno: ”Bilinear Transformation Method”. Academic Press 1984
[4] C. Rogers and W. F. Shadwick: ”Bäcklund Transformations and
their Applications”. Academic Press 1982
[5] A. Wazwaz: ”Partial Differential Equations and Solitary Waves
[6] W. Hereman and W. Zhuang: ”Symbolic Computation of Solitons
via Hirota’s Bilinear Method”. Colorado School of Mines Report 1994
[7] G. W. Griffiths: ”Hirota Direct Method”. City Univ. Report 2012
1328
103 The Algebro-Geometric Method
This method makes use of Riemann’s Surfaces. The subject is rather ex-
tensive and cannot be covered in this section. The book by Belokolos et al
provides a self-contained exposition of the subject.
References
[1] E. D. Belokolos, A. I. Bobenko, V. Z. Enol’skii, A. R. Its and
V. B. Matveev: ”Algebro-Geometric Approach to Nonlinear Integrable
Equations”. Springer 1994.
A Appendix
In this Appendix we transform a second order ordinary differential equation
dy d2 y
F x, y, , 2 = 0 (A.1)
dx dx
into a second order ordinary differential equation
ds d2 s
Φ r, s, , 2 = 0 (A.2)
dr dr
under a change of variables (x, y) → (r, s).
We introduce the transformation
x = α(r, s), y = β(r, s), s = s(r) (A.3)
Since
∂α ∂α ∂β ∂β
dx = dr + ds and dy = dr + ds (A.4)
∂r ∂s ∂r ∂s
we have
∂β ∂β
dy dr + ds
= ∂r ∂s (A.5)
dx ∂α ∂α
dr + ds
∂r ∂s
and dividing by dr both members of the fraction in the right hand side of
the above relation, we find
∂β ∂β ds
dy + ·
= ∂r ∂s dr (A.6)
dx ∂α ∂α ds
+ ·
∂r ∂s dr
1329
The above equation provides the transformation rule for the first derivative.
We shall now find the transformation law for the second derivative
d2 y
(A.7)
dx2
Using the identity
d dy d2 y dx
= 2· (A.8)
dr dx dx dr
we obtain
d dy
2
dy
= dr dx (A.9)
dx2 dx
dr
We have further, using (A.6)
∂β ∂β ds
dy + ·
d = d ∂r ∂s dr =
dx ∂α ∂α ds
+ ·
∂r ∂s dr
∂α ∂α ds ∂β ∂β ds ∂β ∂β ds ∂α ∂α ds
+ · ·d + · − + · ·d + ·
= ∂r ∂s dr ∂r ∂s dr ∂r ∂s dr ∂r ∂s dr
∂α ∂α ds 2
+ ·
∂r ∂s dr
(A.10)
∂β ∂β ds ∂β ∂β ds ∂β ds
d + · =d +d · + ·d (A.11)
∂r ∂s dr ∂r ∂s dr ∂s dr
∂β ∂ 2 β ∂ 2β
d = 2 dr + ds (A.12)
∂r ∂r ∂r∂s
∂β ∂ 2β ∂ 2β
d = dr + 2 ds (A.13)
∂s ∂s∂r ∂s
ds dr·d2 s − ds·d2 r d2 s ds d2 r
d = = − dr (A.14)
dr dr2 dr2 dr dr2
Since r is the independent variable, we have
d2 r
=0
dr2
1330
Therefore we obtain from (A.14)
ds d2 s
d = dr (A.15)
dr dr2
Using (A.11), (A.12), (A.13) and (A.15), we find
∂β ∂β ds
d + · = B (2) (r, s)dr (A.16)
∂r ∂s dr
where
∂ 2β ∂ 2 β ds ∂ 2 β ds 2 ∂β d2 s
B (2) = + 2 + 2 + (A.17)
∂r2 ∂s∂r dr ∂s dr ∂s dr2
We find similarly
∂α ∂α ds
d + · = A(2) (r, s)dr (A.18)
∂r ∂s dr
where
∂ 2α ∂ 2 α ds ∂ 2 α ds 2 ∂α d2 s
A(2) = + 2 + 2 + (A.19)
∂r2 ∂s∂r dr ∂s dr ∂s dr2
We thus obtain from (A.10)
∂α ∂α ds ∂β ∂β ds
(2)
dy + · ·B (r, s)dr − + · ·A(2) (r, s)dr
d = ∂r ∂s dr ∂r
∂α ∂α ds 2
∂s dr (A.20)
dx
+ ·
∂r ∂s dr
Therefore we obtain from (A.9) using (A.20):
d dy ∂α ∂α ds
(2)
∂β ∂β ds
d2 y + · ·B (r, s) − + · ·A(2) (r, s)
= dr dx = ∂r ∂s dr ∂r ∂s dr
dx2 dx ∂α ∂α ds 3
+ ·
dr ∂r ∂s dr
(A.21)
Equation (A.21) provides the law of transformation for the second order
derivative under the transformation (A.3).
We introduce the notation
∂α ∂α ds ∂β ∂β ds
A(1) (r, s) = + · and B (1) (r, s) = + · (A.22)
∂r ∂s dr ∂r ∂s dr
1331
It is obvious that
d (1) d (1)
A (r, s) = A(2) (r, s) and B (r, s) = B (2) (r, s) (A.23)
dr dr
according to (A.18) and (A.16) respectively.
Therefore (A.21) can be written as
d dy
d2 y A(1) (r, s)·B (2) (r, s) − B (1) (r, s)·A(2) (r, s)
= dr dx = 3 (A.24)
dx2 dx
(1)
A (r, s)
dr
B Appendix
In this Appendix we transform a third order ordinary differential equation
dy d2 y d3 y
F x, y, , 2 , 3 = 0 (B.1)
dx dx dx
into a third order ordinary differential equation
ds d2 s d3 s
Φ r, s, , 2 , 3 = 0 (B.2)
dr dr dr
under a change of variables (x, y) → (r, s).
We introduce the transformation
x = α(r, s), y = β(r, s), s = s(r) (B.3)
We have already found the transformation laws for the first and second order
derivatives in Appendix A. For the third order derivative we use the formula
d d2 y
d3 y 2
= dr dx (B.4)
dx 3 dx
dr
In order to find the explicit expression of the third derivative in (B.4), we
first evaluate the differential of the expression (A.17), written down next for
convenience
∂ 2β ∂ 2 β ds ∂ 2 β ds 2 ∂β d2 s
B (2) (r, s) = 2 + 2 + 2 + (B.5)
∂r ∂r∂s dr ∂s dr ∂s dr2
1332
We have for the differentials of the various terms of (B.5):
∂ 2β
∂ 3β ∂ 3β
d dr=+ ds (B.6)
∂r2 ∂r3 ∂r2 ∂s
∂ 2 β ds ∂ 3β ∂ 3β ds ∂ 2 β d2 s
d 2 =2 dr + ds + 2 dr (B.7)
∂r∂s dr ∂r2 ∂s ∂r∂s2 dr ∂r∂s dr2
where we have used (A.15) for the differential d(ds/dr).
∂ 3β
∂ 2 β ds 2 ∂ 3 β ds 2 ∂ 2 β ds d2 s
d = dr + 3 ds +2 2 dr (B.8)
∂s2 dr ∂r∂s2 ∂s dr ∂s dr dr2
∂β d2 s ∂ 2 β ∂ 2 β d2 s ∂β d2 s
d = dr + ds + d (B.9)
∂s dr2 ∂r∂s ∂s2 dr2 ∂s dr2
Since
dr2 ·d3 s − d2 s·d(dr2 )
d2 s dr2 ·d3 s − d2 s·2dr·d2 r
d = =
dr2 dr4 dr4 (B.10)
3 2 2 3 2
ds d sd r ds dr
= 3 dr − 2 2 2 dr = 3 dr since =0
dr dr dr dr dr2
Therefore (B.9) becomes because of (B.10)
∂β d2 s ∂ 2β ∂ 2 β d2 s ∂β d3 s
d = dr + 2 ds + dr (B.11)
∂s dr2 ∂r∂s ∂s dr2 ∂s dr3
Taking into account (B.5) and using (B.6)-(B.8) and (B.11), we obtain
d (2) ∂ 3β ∂ 3 β ds
B (r, s) = 3 + 2 +
dr ∂r ∂r ∂s dr
∂ 3β ∂ 3 β ds ds ∂ 2 β d2 s
+2 + + 2
∂r2 ∂s ∂r∂s2 dr dr ∂r∂s dr2 (B.12)
∂ 3β 3
∂ β ds ds 2 ∂ 2 β ds d2 s
+ + + 2
∂r∂s2 ∂s3 dr dr ∂s2 dr dr2
∂ 2β 2 2
∂ β ds d s ∂β d s
3
+ + 2 +
∂r∂s ∂s dr dr2 ∂s dr3
d (2)
B (3) (r, s)≡ B (r, s) (B.13)
dr
1333
and collecting all terms together in (B.12), we find
∂ 3β ∂ 3 β ds ∂ 3 β ds 2
B (3) (r, s) = + 3 + 3
∂r3 ∂r2 ∂s dr ∂r∂s2 dr (B.14)
2 2 3
∂ β d s ∂ β ds 3 ∂ β ds d2 s ∂β d3 s
2
+3 + + 3 +
∂r∂s dr2 ∂s3 dr ∂s2 dr dr2 ∂s dr3
We find similarly, changing β→α, that
d (2)
A(3) (r, s)≡ A (r, s) (B.15)
dr
where
∂ 3α ∂ 3 α ds ∂ 3 α ds 2
A(3) (r, s) = + 3 + 3
∂r3 ∂r2 ∂s dr ∂r∂s2 dr (B.16)
2 2 3
∂ α d s ∂ α ds 3 ∂ α ds d2 s ∂α d3 s
2
+3 + + 3 +
∂r∂s dr2 ∂s3 dr ∂s2 dr dr2 ∂s dr3
Using the expression (A.21) we find by differentiation with respect to r :
d d2 y N (r, s)
= 4 (B.17)
dr dx2

A(1) (r, s)
where

N (r, s) = A(1) (r, s)B (3) (r, s) − B (1) (r, s)A(3) (r, s) A(1) (r, s)
(B.18)
− 3 A(1) (r, s)B (2) (r, s) − B (1) (r, s)A(2) (r, s) A(2) (r, s)
Therefore, using (B.4) and (B.17), we obtain
d d2 y
3
dy 2 N (r, s)
= dr dx = 5 (B.19)
dx 3 dx
A(1) (r, s)
dr
The above equation provides the transformation law for the third derivative
under a change of variables given by (B.3).
1334
C Appendix
In this Appendix we shall solve some 1st order ordinary differential equations
using MAPLE.
dy (y − x· ln(x))2
= + ln(x), y = y(x) (C.1)
dx x2
>restart;
>with(DEtools):
>with(PDEtools):
(y(x) − x ∗ ln(x))2
> ode := diff(y(x), x) = + ln(x); (C.2)
x2
(y(x) − x ln(x))2
diff(y(x), x) = + ln(x) (C.3)
x2
>symgen(ode);
[− ξ = x, − η = x + y] (C.4)
That means in our notation X = x and Y = x + y.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ (x + y) = 0 : (C.5)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ (x + y) = 1 : (C.6)
1335
> pdsolve({eq1, eq2}, {r(x, y), s(x, y)}); (C.7)

−y + x ln(x) −y + x ln(x)
r(x, y) =− F 1 − , s(x, y) = ln(x) +− F 2 −
x x
(C.8)

take
y − x ln(x)
r= s = ln(x) (C.9)
x
> itr := {x = exp(s(r)), y(x) = exp(s(r)) ∗ (r + s(r))}; (C.10)
The system responds
x = es(r) , y(x) = es(r) (r + es(r) )

(C.11)

> dchange(itr, ode, simplif y); (C.12)
The system responds with a rather complicated expression, which, when

simplified, takes the form
d
(r2 − r − 1) s(r) = 1 (C.13)
dr
d
Step 7. From the above equation, solving with respect to s(r) we obtain
ds
the equation
1
> eqn := diff(s(r), r) = ; (C.14)
r2 − r − 1
1336
The system responds
d 1
s(r) = 2 (C.15)
ds r −r−1
> dsolve(eqn); (C.16)

2√ 1 √
s(r) = − 5 arctanh (2r − 1) 5 +− C1 (C.17)
5 5
Step 8. Using the initial variables, we transform (C.17) into
√ √
2 5 5 y
ln(x) = − arctanh 2 − ln(x) − 1 + C (C.18)
5 5 x
Solving the previous equation with respect to y, we obtain
√ √5 √
5 5 x
y= tanh − ln(x) + C + x· ln(x) + (C.19)
2 2 2 2
which is the general solution of the original differential equation.
dy x + cos(ey + (x + 1)e−x )
= , y = y(x) (C.20)
dx ex+y
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x) = (x+cos(exp(y(x))+(x+1)∗exp(−x)))/exp(x+y(x));

(C.21)

d x + cos(ey(x) + (x + 1)e−x )
y(x) = (C.22)
dx ex+y(x)
1337
>symgen(ode,HINT=[f(x),x*g(y)]);
[− ξ = ex , − η = xe−y ] (C.23)
That means in our notation X = ex and Y = xe−y .
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ ex + diff(r(x, y), y) ∗ x ∗ e−y = 0 : (C.24)

> eq2 := diff(s(x, y), x) ∗ ex + diff(s(x, y), y) ∗ x ∗ e−y = 1 : (C.25)

−x−y −x−y y
{r(x, y) =− F 1 (xe +e + 1)e ,
(C.27)
s(x, y) = −e−x +− F 2 (xe−x−y + e−x−y + 1)ey }

take
r = (x + 1)e−x + ey s = −e−x (C.28)
> itr := {x = −ln(−s(r)), y(x) = ln(r +(1−ln(−s(r)))∗s(r))}; (C.29)
The system responds
{x = −ln(−s(r)), y(x) = ln(r + (1 − ln(−s(r)))s(r))} (C.30)
1338
d
cos(r) s(r) = 1 (C.32)
dr
Step 7. From the above equation, entering
1
> eqn := diff(s(r), r) = ; (C.33)
cos(r)
The system responds

d 1
s(r) = (C.34)
ds cos(r)
> dsolve(eqn); (C.35)
s(r) = ln(sec(r) + tan(r)) +− C1 (C.36)
Step 8. Using the initial variables, i.e. taking into account (C.28), we trans-
form (C.36) into
1 + sin[(x + 1)e−x + ey ]
−e−x = ln + C1 (C.37)
cos[(x + 1)e−x + ey ]
which is the general solution of the original differential equation.

d 2
− (y ) + y 2 − 2y + 5 = 0, y = y(x) (C.38)
dx
Solution. This is equation (E.55) which is to be solved by Lie symmetry
analysis. We use the following steps:
>restart;
>with(DEtools):
1339
>with(PDEtools):
> ode := −diff(y(x)2 , x) + y(x)2 − 2 ∗ y(x) + 5 = 0; (C.39)

d
−2y(x) y(x) + y(x)2 − 2y(x) + 5 = 0 (C.40)
dx
>symgen(ode);
[− ξ = 1, − η = 0] (C.41)
That means in our notation X = 1 and Y = 0.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ 0 = 0 : (C.42)

> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ 0 = 1 : (C.43)
{r(x, y) =− F 2(y), s(x, y) = x +− F 1(y)} (C.45)

take
r=y s=x+y (C.46)
> itr := {x = s(r) − r, y(x) = r}; (C.47)
1340
The system responds
{x = s(r) − r, y(x) = r} (C.48)


d
(r2 − 2r + 5) s(r) = r2 + 5 (C.50)
dr
> eqn := (r2 − 2 ∗ r + 5) ∗ diff(s(r), r) = r2 + 5; (C.51)
The system responds

d
(r2 − 2r + 5) s(r) = r2 + 5 (C.52)
dr
Solving the above equation using the command
>dsolve(eqn);
1 1
s(r) = r + ln(r2 − 2r + 5) + arctan r − +− C1 (C.53)
2 2
form (C.53) into
y − 1
x = ln(y 2 − 2y + 5) + arctan + C1 (C.54)
2
which is the general solution of the original differential equation (C.38).
dy
x2
x − y x·ln + 2 = 0, y = y(x) (C.55)
dx y
1341
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x ∗ diff(y(x), x) − y(x) ∗ (x ∗ ln(x2 /y(x)) + 2) = 0; (C.56)

d x2
x y(x) − y(x) x ln +2 =0 (C.57)
dx y(x)

>symgen(ode);
2y
[− ξ = 1, − η = ] (C.58)
x
2y
That means in our notation X = 1 and Y = .
x
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
2y
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ =0: (C.59)
x
2y
> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ =1: (C.60)
x

n y y o
r(x, y) =− F 1 2 , s(x, y) = −e−x +− F 2 2 (C.62)
x x
1342
take
y
r= s = −e−x (C.63)
x2
2
> itr := {x = −ln(−s(r)), y(x) = r ∗ ln(−s(r)) }; (C.64)
The system responds

2
x = −ln(−s(r)), y(x) = r ln(−s(r)) (C.65)
Step 6. We make a change of variables (from x, y(x) to r, s(r)) to the original


d 1
s(r) + r s(r) ln =0 (C.67)
dr r
1
> eqn := s(r) + r ∗ diff(s(r), r) ∗ ln = 0; (C.68)
r
The system responds
d 1
s(r) + r s(r) ln =0 (C.69)
dr r
>dsolve(eqn);
1
s(r) =− C1 ln (C.70)
r
1343
form (C.70) into
x2
−e−x = C1 ·ln (C.71)
y
which, when solved with respect to y becomes

y = x2 ·exp C·e−x (C.72)
This is the general solution of the original differential equation (C.55).

dy
x(y + 4) − y 2 − 2y − 2x = 0, y = y(x) (C.73)
dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x ∗ (y(x) + 4) ∗ diff(y(x), x) − y(x)2 − 2 ∗ y(x) − 2 ∗ x = 0; (C.74)

d
x (y(x) + 4) y(x) − y(x)2 − 2 y(x) − 2x = 0 (C.75)
dx
>symgen(ode);
[− ξ = x2 + 4x, − η = 4x + xy] (C.76)

That means in our notation X = x2 + 4x and Y = 4x + xy.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1344
> eq1 := diff(r(x, y), x)∗(x2 +4∗x)+diff(r(x, y), y)∗(4∗x+x∗y) = 0 : (C.77)

> eq2 := diff(s(x, y), x)∗(x2 +4∗x)+diff(s(x, y), y)∗(4∗x+x∗y) = 1 : (C.78)

y + 4 y + 4
1 x
r(x, y) =− F 1 , s(x, y) = ln +− F 2 (C.80)
x+4 4 x+4 x+4
take
y+4 1 x
r= , s = ln (C.81)
x+4 4 x+4
> itr := {x = 4 ∗ exp(4 ∗ s(r))/(1 − exp(4 ∗ s(r))),
(C.82)
y(x) = 4 ∗ r/(1 − exp(4 ∗ s(r))) − 4};
The system responds
4 e4s(r)

4r
x= , y(x) = −4 (C.83)
1 − e4s(r) 1 − e4s(r)

d d d
2
−r + 4 s(r) r − 6 s(r) r + 2 s(r) = 0 (C.85)
dr dr dr
> eqn := (4 ∗ r2 − 6 ∗ r + 2) ∗ diff(s(r), r) = r; (C.86)
1345
The system responds
d
(4r2 − 6r + 2) s(r) = r (C.87)
dr
>dsolve(eqn);
1 1
s(r) = ln(r − 1) − ln(2r − 1) +− C1 (C.88)
2 4
form (C.88) into
1 x 1 y + 4 1 y+4
ln = ln − 1 − ln 2 −1 +C
4 x+4 2 x+4 4 x+4
1 y + 4
1 x y+4
ln 2 − 1 = ln A −1
4 x+4 x+4 2 x+4
The above equation can also be simplified as

x(2y − x + 4) y−x
ln = 2 ln A·
(x + 4)2 x+4
x(2y − x + 4) = C·(y − x)2 (C.89)

d 2
(y ) + a·y 2 + b·y + c = 0, y = y(x) (C.90)
dx
>restart;
>with(DEtools):
1346
>with(PDEtools):
> ode := diff(y(x)2 , x) + a ∗ y(x)2 + b ∗ y(x) + c = 0; (C.91)

d
2y(x) y(x) + a y(x)2 + b y(x) + c = 0 (C.92)
dx
>symgen(ode);
[− ξ = 1, − η = 0] (C.93)
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y

{r(x, y) =− F 2(y), s(x, y) = x +− F 1(y)} (C.97)

take
r=y s=x+y (C.98)
> itr := {x = s(r) − r, y(x) = r}; (C.99)
1347
The system responds
{x = s(r) − r, y(x) = r} (C.100)


d
(ar2 + br + c) s(r) = ar2 + (b − 2)r + c (C.102)
dr
> eqn := (a ∗ r2 + b ∗ r + c) ∗ diff(s(r), r) = a ∗ r2 + (b − 2) ∗ r + c; (C.103)
the system responds

d
(ar2 + br + c) s(r) = ar2 + (b − 2)r + c (C.104)
dr
>dsolve(eqn);
2ar + b
2b arctan √
ln(ar2 + br + c) −4ca + b2
s(r) = r − − √ +− C1 (C.105)
a a −4ca + b2
form (C.105) into
1 2 2b 2ay + b
x = − ·ln(ay + by + c) − √ arctan √ + C (C.106)
a −4ca + b2 −4ca + b2
d 2 d
(y ) + a (y) + b·y 2 + c·y + e = 0, y = y(x) (C.107)
dx dx
1348
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x)2 , x)+a∗diff(y(x), x)+b∗y(x)2 +c∗y(x)+e = 0; (C.108)

d d
2y(x) y(x) + a y(x) + b y(x)2 + c y(x) + e = 0 (C.109)
dx dx
>symgen(ode);
[− ξ = 1, − η = 0] (C.110)
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y

{r(x, y) =− F 2(y), s(x, y) = x +− F 1(y)} (C.114)

take
r=y s=x+y (C.115)
1349
> itr := {x = s(r) − r, y(x) = r}; (C.116)
The system responds
{x = s(r) − r, y(x) = r} (C.117)
d
(br2 + cr + e) s(r) = br2 + (c − 2)r + e − a (C.119)
dr
> eqn := (b ∗ r2 + c ∗ r + e) ∗ diff(s(r), r) = b ∗ r2 + (c − 2) ∗ r + e − a; (C.120)
the system responds
d
2
(br + cr + e) s(r) = br2 + (c − 2)r + e − a (C.121)
dr
>dsolve(eqn);
2br + c
2 arctan √ a
ln(br2 + cr + e) 4eb − c2
s(r) = r − − √
b 4eb − c2
2br + c (C.122)
2 arctan √ c
4eb − c 2
+ √ +− C1
4eb − c2 b
Step 8. Using the initial variables, i.e. taking into account (C.115), we
transform (C.122) into
c
1 2 a −
b arctan √2by + c + C (C.123)

x = − ·ln(by 2 + cy + e) − √
b 4eb − c2 4eb − c2
1350
dy 2y 3 (x − y − xy)
= , y = y(x) (C.124)
dx x(x − y)2

>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x) = 2∗y(x)3 ∗(x−y(x)−x∗y(x))/(x∗(x−y(x))2 ); (C.125)
d 2y(x)3 (x − y(x) − xy(x))

y(x) = (C.126)
dx x(x − y(x))2

>symgen(ode);
[ − ξ = x2 , − η = y 2 ] (C.127)
We make the choice X = −x2 and Y = −y 2 using our notation .
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ (−x2 ) + diff(r(x, y), y) ∗ (−y 2 ) = 0 : (C.128)

> eq2 := diff(s(x, y), x) ∗ (−x2 ) + diff(s(x, y), y) ∗ (−y 2 ) = 1 : (C.129)
1351
y − x y − x
1
r(x, y) =− F 1 − , s(x, y) = +− F 2 − (C.131)
xy x xy
take
x−y 1
r= s= (C.132)
xy x
> itr := {x = 1/s(r), y(x) = 1/(r + s(r))}; (C.133)
The system responds

1 1
x= , y(x) = (C.134)
s(r) r + s(r)

d
(r2 − 2r + 2) s(r) = −r2 (C.136)
dr
> eqn := (r2 − 2 ∗ r + 2) ∗ diff(s(r), r) = −r2 ; (C.137)
the system responds

d
(r2 − 2r + 2) s(r) = −r2 (C.138)
dr
>dsolve(eqn);
s(r) = −r − ln(r2 − 2r + 2) +− C1 (C.139)
1352
1 x−y x − y 2 x−y
=− − ln +2 +2 +C (C.140)
x xy xy xy

x − y 2 x−y
1 = −y·ln +2 + 2 + C·y (C.141)
xy xy
dy y x2
= + , y = y(x) (C.142)
dx x x+y
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x) = y(x)/x + x2 /(x + y(x)); (C.143)
d y(x) x2
y(x) = + (C.144)
dx x x + y(x)
>symgen(ode);
x2 −2yx − y 2 + 2x3
[− ξ = 0, − η = ], [− ξ = 0, − η = − ] (C.145)
x+y x+y
x2
We make the choice X = 0 and Y = using our notation .
x+y
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1353
x2
> eq1 := diff(r(x, y), x) ∗ 0 + diff(r(x, y), y) ∗ =0: (C.146)
x+y
x2
> eq2 := diff(s(x, y), x) ∗ 0 + diff(s(x, y), y) ∗ =1: (C.147)
x+y
1 y2 y

r(x, y) =− F 1(x), s(x, y) = + +− F 2(x) (C.149)
2 x2 x

take
1 y2 y
r = x, s= + (C.150)
2 x2 x
p
> itr := {x = r, y(x) = −1 + 1 + 2 ∗ s(r) r}; (C.151)
The system responds

n p o
x = r, y(x) = −1 + 1 + 2 s(r) r (C.152)


d
s(r) = 1 (C.154)
dr
1354
> eqn := diff(s(r), r) = 1; (C.155)
the system responds

d
s(r) = 1 (C.156)
dr
>dsolve(eqn);
s(r) = r +− C1 (C.157)
1 y2 y
+ =x+C (C.158)
2 x2 x
y 2 + 2xy = 2x3 + A·x2 (C.159)

dy x3
= − xy, y = y(x) (C.160)
dx y
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x) = x3 /y(x) − x ∗ y(x) (C.161)

d x3
y(x) = − x y(x) (C.162)
dx y(x)
1355
>symgen(ode);
1 x2
[− ξ = 0, − η = − + y] (C.163)
y y
1 x2
We make the choice X = 0 and Y = − + y, using our notation .
y y
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1 x2
> eq1 := diff(r(x, y), x) ∗ 0 + diff(r(x, y), y) ∗ − + y = 0 : (C.164)
y y
1 x2
> eq2 := diff(s(x, y), x) ∗ 0 + diff(s(x, y), y) ∗ − + y = 1 : (C.165)
y y

1 2 2
r(x, y) =− F 1(x), s(x, y) = ln(1 − x + y ) +− F 2(x) (C.167)
2
take
1
r = x, s = ln(1 − x2 + y 2 ) (C.168)
2
p
> itr := {x = r, y(x) = −1 + r2 + exp(2 ∗ s(r))}; (C.169)
The system responds

n p o
x = r, y(x) = −1 + r2 + e2s(r) (C.170)
1356

d
s(r) = −r (C.172)
dr
> eqn := diff(s(r), r) = −r; (C.173)
the system responds

d
s(r) = −r (C.174)
dr
>dsolve(eqn);
1
s(r) = − r2 +− C1 (C.175)
2
1 1
ln(1 − x2 + y 2 ) = − x2 + C (C.176)
2 2
1 − x2 + y 2 = exp(−x2 + A) (C.177)

dy 3xy + 2y 3
=− 2 , y = y(x) (C.178)
dx x + 3xy 2
1357
>restart;
>with(DEtools):
>with(PDEtools):
3 ∗ x ∗ y(x) + 2 ∗ y(x)3
> ode := diff(y(x), x) = − (C.179)
x2 + 3 ∗ x ∗ y(x)2
d 3xy(x) + 2y(x)3
y(x) = − 2 (C.180)
dx x + 3xy(x)2

>symgen(ode,HINT=[f(x),g(y)]);
[− ξ = 2x, − η = y] (C.181)
We make the choice X = 2x and Y = y, using our notation .
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ 2 ∗ x + diff(r(x, y), y) ∗ y = 0 : (C.182)

> eq2 := diff(s(x, y), x) ∗ 2 ∗ x + diff(s(x, y), y) ∗ y = 1 : (C.183)

y y
1
r(x, y) =− F 1 √ , s(x, y) = ln(x) +− F 2 √ (C.185)
x 2 x
1358
take
y 1
r = √ , s = ln(x) (C.186)
x 2
> itr := {x = exp(2 ∗ s(r)), y(x) = r ∗ exp(s(r))}; (C.187)
The system responds
x = e2s(r) , y(x) = res(r)

(C.188)
d
7(r + r3 ) s(r) = −1 − 3r2 (C.190)
dr
> eqn := 7 ∗ (r + r3 ) ∗ diff(s(r), r) = −1 − 3 ∗ r2 ; (C.191)
the system responds
d
7(r + r3 ) s(r) = −1 − 3r2 (C.192)
dr
>dsolve(eqn);
1
s(r) = − ln(r(1 + r2 )) +− C1 (C.193)
7
1 1 y y 2
ln(x) = − ln √ 1 + +C (C.194)
2 7 x x
1359
7
y y 2 2
x √ 1+ =A
x x
or
6 2
y 2 2
x y 1+ = A⇐⇒x4 y 2 (x + y 2 )2 = A
x
and thus
x2 y(x + y 2 ) = B (C.195)
dy y − 2xy 2
= , y = y(x) (C.196)
dx x + 2x2 y
>restart;
>with(DEtools):
>with(PDEtools):
y(x) − 2 ∗ x ∗ y(x)2
> ode := diff(y(x), x) = (C.197)
x + 2 ∗ x2 ∗ y(x)
d y(x) − 2xy(x)2
y(x) = (C.198)
dx x + 2x2 y(x)
>symgen(ode,HINT=[f(x),g(y)]);
[− ξ = x, − η = −y] (C.199)
We make the choice X = x and Y = −y, using our notation .
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1360
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ (−y) = 0 : (C.200)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ (−y) = 1 : (C.201)
{r(x, y) =− F 1(xy), s(x, y) = ln(x) +− F 2(xy)} (C.203)

take
r = xy, s = ln(x) (C.204)
> itr := {x = exp(s(r)), y(x) = r ∗ exp(−s(r))}; (C.205)
The system responds
x = es(r) , y(x) = re−s(r)

(C.206)


d
2r s(r) = 1 + 2r (C.208)
dr
> eqn := 2 ∗ r ∗ diff(s(r), r) = 1 + 2 ∗ r; (C.209)
1361
the system responds
d
2r s(r) = 1 + 2r (C.210)
dr
>dsolve(eqn);
1
s(r) = r + ln(r) +− C1 (C.211)
2
1
ln(x) = xy + ln(xy) + C (C.212)
2
x
ln − 2xy = A (C.213)
y
dy 2y 2 − 2xy + ay
= 2 , y = y(x) (C.214)
dx y − x2 + ay
>restart;
>with(DEtools):
>with(PDEtools):
2 ∗ y(x)2 − 2 ∗ x ∗ y(x) + a ∗ y(x)

> ode := diff(y(x), x) = (C.215)
y(x)2 − x2 + a ∗ y(x)
d 2y(x)2 − 2xy(x) + ay(x)

y(x) = (C.216)
dx y(x)2 − x2 + ay(x)
1362
>symgen(ode,HINT=[y*f(x),g(y)]);
[− ξ = y, − η = y] (C.217)
We make the choice X = y and Y = y, using our notation .
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ y + diff(r(x, y), y) ∗ y = 0 : (C.218)

> eq2 := diff(s(x, y), x) ∗ y + diff(s(x, y), y) ∗ y = 1 : (C.219)
{r(x, y) =− F 1(−x + y), s(x, y) = ln(y) +− F 2(−x + y)} (C.221)

take
r = y − x, s = ln(y) (C.222)
> itr := {x = exp(s(r)) − r, y(x) = exp(s(r))}; (C.223)
The system responds
x = es(r) − r, y(x) = es(r)

(C.224)

1363
d
r2 s(r) = 2r + a (C.226)
dr
> eqn := r2 ∗ diff(s(r), r) = 2 ∗ r + a; (C.227)
the system responds

d
r2 s(r) = 2r + a (C.228)
dr
>dsolve(eqn);
a
s(r) = 2 ln(r) − +− C1 (C.229)
r
a
ln(y) = 2 ln | − x + y| − +C (C.230)
−x + y
y a
ln 2
= +C (C.231)
(x − y) x−y

We can express the above solution in parametric form. In fact, under the
substitution
x−y =t (C.232)
we obtain from (C.231) that

y a
ln 2 = + C (C.233)
t t
1364
and thus
y a
= A·exp (C.234)
t2 t
Using (C.232) and (C.234) we get
a a
x = t + A·t2 exp , y = A·t2 exp (C.235)
t t
This is a parametric solution of the differential equation (C.214). It is obvious
however that we could use any smooth function f (t) instead of t in (C.232).
In that case we would have obtained the following parametric solution
a a
x = f (t) + A·f (t)2 exp , y = A·f (t)2 exp (C.236)
f (t) f (t)

dy
y + 3xy + x3 = 0, y = y(x) (C.237)
dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x) ∗ diff(y(x), x) = −3 ∗ x ∗ y(x) − x3 ; (C.238)

d
y(x) y(x) = −3xy(x) − x3 (C.239)
dx
>symgen(ode);
x4
[− ξ = 0, − η = 2y + 3x2 + ] (C.240)
y
1365
x4
We make the choice X = 0 and Y = 2y + 3x2 + , using our notation .
y
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
x4
> eq1 := diff(r(x, y), x) ∗ 0 + diff(r(x, y), y) ∗ (2y + 3x2 + ) = 0 : (C.241)
y
x4
> eq2 := diff(s(x, y), x) ∗ 0 + diff(s(x, y), y) ∗ (2y + 3x2 + ) = 1 : (C.242)
y

2 1 2
r(x, y) =− F 1(x), s(x, y) = ln(y + x ) − ln(2y + x ) +− F 2(x) (C.244)
2
take
1 (y + x2 )2
r = x, s = ln (C.245)
2 2y + x2
> itr := {x = r, y(x) = −r2 + exp(2 ∗ s(r))

p (C.246)
±exp(s(r)) −r2 + exp(2 ∗ s(r))};
The system responds

n p o
x = r, y(x) = −r2 + e2s(r) ±es(r) −r2 + e2s(r) (C.247)

1366
d
s(r) = 0 (C.249)
dr
Step 7. From the above equation, we get
s(r) = C (C.250)
Step 8. Using the initial variables, i.e. taking into account (C.247), we get
√
y = −x2 + C 2 ±C −x2 + C 2 , C > 0 (C.251)

Example 15. Find the general solution of the differential equation
dy 2 dy
x − ay + b = 0, y = y(x) (C.252)
dx dx
Solution. The given equation admits the infinitesimal symmetries
X = 2x, Y = y (C.253)
Solving the normal coordinates system, we get the solution

y 1 y
r = F √ , s = ln(x) + G √
x 2 x

y 1
r=√ and s = ln(x) (C.254)
x 2
We thus consider the transformation
x = e2·s(r) , y(x) = r·es(r) (C.255)
under which equation (C.252) becomes

d 2 d
(2ar2 − r2 − 4b) s(r) + 2r(a − 1) s(r) + 1 = 0 (C.256)
dr dr
1367
From the above equation we get the two solutions
√
d (1 − a)r + a2 r2 − 4b
s(r) = ± (C.257)
dr (2a − 1)r2 − 4b
Integrating the previous equation we get

1−a (a − 1)r
s(r) = ± 2 arctanh √
2(2a − 1) a2 r2 − 4b
√
(C.258)
2 2a 2 2
+ ln[(2a − 1)r − 4b] + ln[ar + a r − 4b] + C
1−a
The solution of the original equation is being obtained from (C.258) taking
into account (C.254).
Example 16. Find the general solution of the differential equation
dy 2
xy 2 = ay 2 + bx = 0, y = y(x) (C.259)
dx
Solution. The given equation admits the infinitesimal symmetries
X = 2x, Y = y (C.260)
Solving the normal coordinates system, we get the solution

y 1 y
r = F √ , s = ln(x) + G √
x 2 x

y 1
r=√ and s = ln(x) (C.261)
x 2
We thus consider the transformation
x = e2·s(r) , y(x) = r·es(r) (C.262)
under which equation (C.259) becomes

d 2 d 2
r2 1 + r s(r) = 4(ar2 + b) s(r) (C.263)
dr dr
1368
From the above equation we get the two solutions
√
d (r2 + 2 ar2 + b)r
s(r) = ± (C.264)
dr −r4 + 4ar2 + 4b
Integrating the previous equation we get
1
r2 − 2a √ar2 + b
s(r) = ± 2a·arctanh − 2(n1 ) arctanh
4m 2m n1
√ar2 + b (C.265)
4 2
+ 2(n2 ) arctanh − m· ln[r − 4ar − 4b] + C
n2
where
√ p p
m= a2 + b, n1 = b + 2a(a − m), n2 = b + 2a(a + m) (C.266)
The solution of the original equation is being obtained from (C.265) taking
into account (C.261).
D Appendix
In this Appendix we shall find the symmetries of a 2nd order ordinary dif-
ferential equations using ”Mathematica”.
Example. Find the symmetries of the differential equation
dy 2
d2 y x −y
= dx , y = y(x) (D.1)
dx2 x3
Solution. We shall use the MathLie software. We use the following steps:
In[1]:=<<MathLie.m
Step 2. We enter the second order ODE using the command
1
In[2] := ode = ∂{x,2} y[x] ==3
∗ (x ∗ ∂x y[x] − y[x])2 ; ode//LT F
x
(−y + x yx )2
yx,x ==
x3
1369
Step 3. We find the determining equations with a single command

detode//LT F
− 4yξ1 + 2xφ1 + 2xy(ξ1 )x − 3y 2 (ξ1 )y − 2x2 (φ1 )x − x3 (ξ1 )x,x

+ 2x3 (φ1 )x,y == 0
3y 2 ξ1 − 2xyφ1 − 2xy 2 (ξ1 )x + 2x2 y(φ1 )x + xy 2 (φ1 )y
+ x4 (φ1 )x,x == 0
ξ1 + 4y(ξ1 )y − x(φ1 )y − 2x2 (ξ1 )x,y + x2 (φ1 )y,y == 0
(ξ1 )y + x(ξ1 )y,y == 0

k1 y
φ1 == k1 + k6 x + ey/x (−k7 + k8 x) + k2 y +
x
+ k3 xy − e−y/x (k4 y + x(k4 + k5 y))
ξ1 == k1 + k2 x + k3 x2 − e−y/x x(k4 + k5 x)
E Appendix
In this Appendix we shall solve some 2nd order ordinary differential equations
using MAPLE.
Example 1. Solve the equation
dy 2
d2 y x −y
= dx , y = y(x) (E.1)
dx2 x3
1370
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x$2) = (x ∗ diff(y(x), x) − y(x))2 /x3 ; (E.2)

d 2
d2 x y(x) − y(x)
y(x) = dx (E.3)
dx 2 x3
>symgen(ode);
[− ξ = 0, − η = x], [− ξ = x, − η = y], [− ξ = x2 , − η = xy] (E.4)
We make the choice X = x2 and Y = xy using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x2 + diff(r(x, y), y) ∗ x ∗ y = 0 : (E.5)

> eq2 := diff(s(x, y), x) ∗ x2 + diff(s(x, y), y) ∗ x ∗ y = 1 : (E.6)
> pdsolve({eq1, eq2}, {r(x, y), s(x, y)}); (E.7)

y y
1
r(x, y) =− F 1 , s(x, y) = − +− F 2 (E.8)
x x x
We take
y 1
r= s=− (E.9)
x x
1371
1 r
> itr := {x = − , y(x) = − }; (E.10)
s(r) s(r)
The system responds

1 r
x=− , y(x) = − (E.11)
s(r) s(r)
> dchange(itr, ode, simplif y); (E.12)
The system responds with an expression which, when simplified, gives

d2 d
s(r) = − s(r) (E.13)
dr2 ds
Step 7. Entering the commands
> eqn := diff(s(r), r$2) = −diff(s(r), r) : (E.14)
> dsolve(eqn); (E.15)

the system responds
s(r) =− C1 +− C2 e−r (E.16)
Step 8. Using the initial variables according to (E.9), we transform (E.16)

into
1 y
− = C1 + C2 exp − (E.17)
x x
y 1
exp − =A B+
x x
and inverting
y 1
− = ln(A) + ln B +
x x
1372
or

1
−y = ln(A) + ln B + ·x
x
The above equation can be cast to the form
Bx + 1
y = − ln(A) − ln ·x
x
x
y = C1 + ln ·x (E.18)
C2 x + 1
under the obvious identifications
C1 = −ln(A), B = C2
Expression (E.18) gives the general solution of the differential equation (E.1).
d2 y dy 2 dy 3
y 2− + 4y = 0, y = y(x) (E.19)
dx dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x)∗diff(y(x), x$2)−(diff(y(x), x))2 +4∗y(x)∗(diff(y(x), x))3 = 0;

(E.20)
d2 d 2 d 3
y(x) y(x) − y(x) + 4 y(x) y(x) =0 (E.21)
dx2 dx dx
>symgen(ode);
[− ξ = 1, − η = 0], [− ξ = −x + y 2 , − η = 0], [− ξ = 2x, − η = y] (E.22)
1373
We make the choice X = 1 and Y = 0 using our notation.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ 0 = 0 : (E.23)

> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ 0 = 1 : (E.24)
{r(x, y) =− F 2(y), s(x, y) = x +− F 1(y)} (E.26)
We take
r=y s=x+y (E.27)
> itr := {x = s(r) − r, y(x) = r}; (E.28)
The system responds
{x = s(r) − r, y(x) = r} (E.29)


d2 d
−r s(r) − s(r) + 1 + 4r = 0 (E.31)
dr2 dr
1374
> eqn := −r ∗ diff(s(r), r$2) − diff(s(r), r) + 1 + 4 ∗ r = 0 : (E.32)

the system responds
s(r) = r2 +− C1 ln(r) + r +− C2 (E.34)

into
x + y = y 2 + C1 ·ln(y) + y + C2 (E.35)
x = y 2 + C1 ·ln(y) + C2 (E.36)
Note. Equation (E.31) is an Euler-Legendre type of equation and can be
solved by the substitution r = ez . Under this substitution it can be con-
d2 s
verted into 2 = ez (1 + 4ez ) which can be solved easily.
dz
d2 y dy 2
2y − 3 − 4y 2 = 0, y = y(x) (E.37)
dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := 2 ∗ y(x) ∗ diff(y(x), x$2) − 3(diff(y(x), x))2 − 4 ∗ y(x)2 = 0; (E.38)

d2 d 2
2 y(x) y(x) − 3 y(x) − 4 y(x)2 = 0 (E.39)
dx2 dx
1375
>symgen(ode);
[− ξ = 1, − η = 0], [− ξ = 0, − η = y] (E.40)
We make the choice X = 0 and Y = y using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ 0 + diff(r(x, y), y) ∗ y = 0 : (E.41)

> eq2 := diff(s(x, y), x) ∗ 0 + diff(s(x, y), y) ∗ y = 1 : (E.42)
{r(x, y) =− F 1(x), s(x, y) = ln(y) +− F 2(x)} (E.44)
We take
r = x s = ln(y) + x (E.45)
> itr := {x = r, y(x) = exp(s(r) − r)}; (E.46)
The system responds
x = r, y(x) = es(r)−r

(E.47)

1376
d2 d 2 d
−2 s(r) + s(r) − 2 s(r) +5=0 (E.49)
dr2 dr dr
> eqn := −2∗diff(s(r), r$2)+diff(s(r), r)2 −2∗diff(s(r), r)+5 = 0 : (E.50)

the system responds
s(r) = r − ln(− C1 sin(r) −− C2 cos(r)) (E.52)

into
ln(y) + x = x − ln[C1 sin(x) + C2 cos(x)]
This equation can also be written as

1
y= (E.53)
C1 sin(x) + C2 cos(x)
Note 1. Using the substitution
d d2 dd dp dp ds dp
p= s(r), 2
s(r) = s(r) = = · =p (E.54)
dr dr dr dr dr ds dr ds
equation (E.49) takes on the form
dp
−2p + p2 − 2p + 5 = 0 (E.55)
ds
which is a first order differential equation. This last equation can also be
solved by symmetry analysis. See Example 3, Appendix C.
Note 2. The original equation (E.37), using the substitution
d d2 dd dp dp dy dp
p= y(x), y(x) = y(x) = = · = p (E.56)
dx dx2 dx dx dx dy dx dy
1377
takes on the form
d 2
y (p ) − 3p2 − 4y 2 = 0 (E.57)
dy
which under the further substitution u = p2 becomes a linear differential

equation
du 3
− u = 4y (E.58)
dy y
d2 y dy 2
y − + 2y 3 = 0, y = y(x) (E.59)
dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x) ∗ diff(y(x), x$2) − (diff(y(x), x))2 + 2 ∗ y(x)3 = 0; (E.60)

d2 d 2
y(x) y(x) − y(x) + 2 y(x)3 = 0 (E.61)
dx2 dx
>symgen(ode);
1
[− ξ = 1, − η = 0], [− ξ = − x, − η = y] (E.62)
2
We make the choice X = 1 and Y = 0 using our notation.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1378
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ 0 = 0 : (E.63)

> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ 0 = 1 : (E.64)
{r(x, y) =− F 2(y), s(x, y) = x +− F 1(y)} (E.66)
We take
r = y, s=x+y (E.67)
> itr := {x = s(r) − r, y(x) = r}; (E.68)
The system responds
{x = s(r) − r, y(x) = r} (E.69)


d2
3 d 3 d
3 2 d
3
r 2
s(r) − 2r s(r) + 6r s(r) + (1 − 6r ) s(r) = 1 − 2r3
dr dr dr dr
> eqn := r ∗ diff(s(r), r$2) − 2 ∗ r3 ∗ (diff(s(r), r))3

(E.71)
+ 6 ∗ r3 ∗ (diff(s(r), r))2 + (1 − 6 ∗ r3 ) ∗ (diff(s(r), r)) = 1 − 2 ∗ r3 :
1379
the system responds
s
1
s(r) = r + 2− C1 arctanh −4r + 2 −
C1 +− C2,
− C1
s (E.73)
1
s(r) = r − 2− C1 arctanh −4r + 2 −
C1 +− C2
− C1

into
s
1
x = ±2C1 arctanh C1 −4y + 2 + C2 (E.74)
C1
s
1 x−C
2
C1 −4 y + 2 = tanh ±
C1 2C1
or even
1 1

1 x − C 2
2
y= 2
− tanh ± (E.75)
4C1 4 C1 2C1
d2 y dy 2
y + a + b y 2 = 0, y = y(x) (E.76)
dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x) ∗ diff(y(x), x$2) + a ∗ (diff(y(x), x))2 + b ∗ y(x)2 = 0; (E.77)
1380
d2 d 2
y(x) y(x) + a y(x) + b y(x)2 = 0 (E.78)
dx2 dx
>symgen(ode);
[− ξ = 1, − η = 0], [− ξ = 0, − η = y] (E.79)

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y

{r(x, y) =− F 1(x), s(x, y) = ln(y) +− F 2(x)} (E.83)
We take
r = x, s = ln(y) + x (E.84)
> itr := {x = r, y(x) = exp(s(r) − r)}; (E.85)
1381
The system responds
x = r, y(x) = es(r)−r

(E.86)


d2 d 2 d
s(r) + (1 + a) s(r) − 2(1 + a) s(r) +1+a+b=0
dr2 dr dr
> eqn := diff(s(r), r$2) + (1 + a) ∗ (diff(s(r), r))2

(E.88)
− 2 ∗ (1 + a) ∗ (diff(s(r), r)) + 1 + a + b = 0 :

the system gives a solution which can be converted into the equivalent form
1 1 1a+1
s(r) = − −2(a + 1)r + A(r) − ln − B(r) (E.90)
2a+1 4 b
where
p
A(r) = 2r −b(a + 1), B(r) = (C1 ·eA(r) + C2 )2 (E.91)

into
1 1 1a+1
x + ln(y) = − −2(a + 1)x + A(x) − ln − B(x) (E.92)
2a+1 4 b

1 1 1a+1
y(x) = exp − −2(a + 1)x + A(x) − ln − B(x) − x
2a+1 4 b
(E.93)
1382
The case a = −1 needs to be examined separately.
d2 y dy
x + [2 − x·f (x)] = y·f (x), y = y(x) (E.94)
dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x∗diff(y(x), x$2)+(2−x∗f (x))∗(diff(y(x), x)) = y(x)∗f (x); (E.95)

d2 d
x y(x) + (2 − xf (x)) y(x) = y(x)f (x) (E.96)
dx2 dx
>symgen(ode);
1
[− ξ = 0, − η = y], [− ξ = 0, − η = ] (E.97)
x
1
We make the choice X = 0 and Y = using our notation.
x
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1
> eq1 := diff(r(x, y), x) ∗ 0 + diff(r(x, y), y) ∗ =0: (E.98)
x
1
> eq2 := diff(s(x, y), x) ∗ 0 + diff(s(x, y), y) ∗ = 1 : (E.99)
x
1383
{r(x, y) =− F 2(x), s(x, y) = yx +− F 1(x)} (E.101)
We take
r = x, s = yx (E.102)
s(r)
> itr := {x = r, y(x) = }; (E.103)
r
The system responds

s(r)
x = r, y(x) = (E.104)
r

d2 d
s(r) − f (r) s(r) =0
dr2 dr
> eqn := diff(s(r), r$2) − f (r) ∗ (diff(s(r), r)) = 0 : (E.106)

the system gives a solution which can be converted into the equivalent form
Z R
s(r) =− C1 + e f (r)dr dr − C2 (E.108)
1384
into
1 Z R
y= C1 + C2 e f (x)dx dx (E.109)
x
which is the general solution of the original differential equation (E.94).
d2 y dy dy 2
x(x + y) 2 + (x − y) +x − y = 0, y = y(x) (E.110)
dx dx dx
Solution.We use the following steps:
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x ∗ (x + y(x)) ∗ diff(y(x), x$2) + (x − y(x)) ∗ (diff(y(x), x))

+ x ∗ (diff(y(x), x))2 − y(x) = 0;
(E.111)
d2 d
x(x + y(x)) y(x) + (x − y(x)) y(x)
dx2 dx (E.112)
d 2
+x y(x) − y(x) = 0
dx
>symgen(ode);
1
[− ξ = −x, − η = x], [− ξ = x, − η = y], [− ξ = 0, − η = − y1(x + y)(2 +− y1)]
2
(E.113)
We make the choice X = x and Y = y using our notation.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1385
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ y = 0 : (E.114)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ y = 1 : (E.115)

n y y o
r(x, y) =− F 1 , s(x, y) = ln(x) +− F 2 (E.117)
x x
We take
y
r= , s = ln(x) (E.118)
x
> itr := {x = exp(s(r)), y(x) = r ∗ exp(s(r))}; (E.119)
The system responds
x = es(r) , y(x) = res(r)

(E.120)


d2 d 2 d
−(1 + r) s(r) + 2(1 + r) s(r) + s(r) = 0
dr2 dr dr
> eqn := −(1+r)∗diff(s(r), r$2)+2∗(1+r)∗(diff(s(r), r))2 +diff(s(r), r) = 0 :

(E.122)
1386
the system gives the solution
1
s(r) = − ln(−− C1r2 − 2− C1r − 2− C2) (E.124)
2
into
1 y 2 y
ln(x) = − ln C1 + 2C1 + C2 (E.125)
2 x x
y 2 y
2
x C1 + 2C1 + C2 = 1 (E.126)
x x
C1 y 2 + 2C1 xy + C2 x2 = 1 (E.127)
The above is the general solution of the original differential equation (E.110).
2 d2 y dy 2
x (x + y) 2 − x − y = 0, y = y(x) (E.128)
dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x2 ∗(x+y(x))∗diff(y(x), x$2)−(x∗diff(y(x), x)−y(x))2 = 0 (E.129)
d2 d 2
x2 (x + y(x)) y(x) − x y(x) − y(x) =0 (E.130)
dx2 dx
>symgen(ode);
[− ξ = −x, − η = x], [− ξ = x, − η = y], [− ξ = x2 , − η = yx] (E.131)
1387
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y


n y y o
r(x, y) =− F 1 , s(x, y) = ln(x) +− F 2 (E.135)
x x
We take
y
r= , s = ln(x) (E.136)
x
The system responds

(E.138)


d2 d 2 d
−(1 + r) 2
s(r) + (1 + r) s(r) − s(r) = 0
dr dr dr
1388
> eqn := −(1+r)∗diff(s(r), r$2)+(1+r)∗(diff(s(r), r))2 −diff(s(r), r) = 0 :

(E.140)

s(r) = −ln(−− C1ln(1 + r) −− C2) (E.142)

into
y
ln(x) = −ln C1 ·ln 1 + + C2 (E.143)
x
y
x C1 ·ln 1 + + C2 = 1 (E.144)
x
The above is the general solution of the original differential equation (E.128).
d2 y dy dy
(x + y) = 2 + 1 , y = y(x) (E.145)
dx2 dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := (x + y(x)) ∗ diff(y(x), x$2) = (2 ∗ diff(y(x), x) + 1) ∗ diff(y(x), x)

(E.146)

d2 d d
(x + y(x)) y(x) = 2 y(x) + 1 y(x) (E.147)
dx2 dx dx
1389
>symgen(ode);
[− ξ = 1, − η = −1], [− ξ = x, − η = y], [− ξ = x2 , − η = −x2 −2xy−2y 2 ] (E.148)
We make the choice X = 1 and Y = −1 using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ (−1) = 0 : (E.149)

> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ (−1) = 1 : (E.150)
{r(x, y) =− F 1(x + y), s(x, y) = x +− F 2(x + y)} (E.152)
We take
r = x + y, s = 2x + y (E.153)
> itr := {x = −r + s(r), y(x) = −s(r) + 2 ∗ r}; (E.154)
The system responds
{x = −r + s(r), y(x) = −s(r) + 2r} (E.155)

1390
d2 d 3 d 2 d
r s(r) + s(r) − 6 s(r) + 11 s(r) − 6 = 0
dr2 dr dr dr
> eqn := r ∗ diff(s(r), r$2) + (diff(s(r), r))3 − 6 ∗ (diff(s(r), r))2
(E.157)
+ 11 ∗ diff(s(r), r) − 6 = 0 :
p
2− C1 − 2r2 + −r2 (− C1 − r2 )
Z
s(r) = 2
dr +− C2,
− C1 − r
p (E.159)
−2− C1 + 2r2 + −r2 (− C1 − r2 )
Z
s(r) = − 2
dr +− C2
− C1 − r

into
p
2C1 − 2r2 + −r2 (C1 − r2 )
Z
2x + y = dr + C2 ,
C1 − r 2 r=x+y
p (E.160)
2C1 − 2r2 − −r2 (C1 − r2 )
Z
2x + y = dr + C2
C1 − r 2 r=x+y
The above equations constitute the general solutions of the original differen-
tial equation (E.145).
d2 d 3
4
x y(x) + x y(x) − y(x) = 0 (E.161)
dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := x4 ∗ diff(y(x), x$2) + (x ∗ diff(y(x), x) − y(x))3 = 0; (E.162)
1391
d2 d 3
x4 y(x) + x y(x) − y(x) =0 (E.163)
dx2 dx
>symgen(ode);
[− ξ = 0, − η = x], [− ξ = x, − η = y] (E.164)

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y


n y y o
r(x, y) =− F 1 , s(x, y) = ln(x) +− F 2 (E.168)
x x
We take
y
r = , s = ln(x) (E.169)
x
The system responds

(E.171)
1392

d2 d 2
− s(r) + s(r) + 1 = 0
dr2 dr
> eqn := −diff(s(r), r$2) + (diff(s(r), r))2 + 1 = 0 : (E.173)

s(r) = −ln(− C1sin(r) −− C2cos(r)) (E.175)

into
y y
ln(x) = −ln C1 ·sin + C2 ·cos (E.176)
x x
y y
x C1 ·sin + C2 ·cos =1 (E.177)
x x
d2
y(x) = ex ·y(x) (E.178)
dx2
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x$2) = exp(x) ∗ y(x); (E.179)
1393
d2
y(x) = ex y(x) (E.180)
dx2
>symgen(ode);
1

[− ξ = 0, − η = y], [− ξ = 0, − η = BesselI 0, 2e 2 x ],
1
(E.181)
x
[− ξ = 0, − η = BesselK 0, 2e 2 ]

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y

{r(x, y) =− F 1(x), s(x, y) = ln(y) +− F 2(x)} (E.185)
We take
r = x, s = ln(y) (E.186)
> itr := {x = r, y(x) = exp(s(r))}; (E.187)
1394
The system responds
x = r, y(x) = es(r)

(E.188)


d2 d 2
s(r) + s(r) + 1 = 0
dr2 dr
> eqn := diff(s(r), r$2) + (diff(s(r), r))2 = er : (E.190)

the system gives the solution expressed in terms of Bessel functions which,
using the usual notation, reads
1
s(r) = − r
2 1 1
C1 I0 2e 2 r + C2 K0 2e 2 r (E.192)
+ ln 1 1 1 1
I1 2e 2 r K0 2e 2 r + K1 2e 2 r I0 2e 2 r

into
1
ln(y) = − x
2 1 1
C1 I0 2e 2 x + C2 K0 2e 2 x (E.193)
+ ln 1 1 1 1
I1 2e 2 x K0 2e 2 x + K1 2e 2 x I0 2e 2 x

 1 1 
x x
2  C 1 I0 2e 2 + C 2 K 0 2e 2 
y= 1 1 1 1 (E.194)
x  I 2e 2 x K 2e 2 x + K 2e 2 x I 2e 2 x 
1 0 1 0
1395
Note. Equations of the form
d2 y
= f (x)·y, y = y(x) (E.195)
dx2
can be solved using the Green-Liouville transformation
G(ξ) dξ
x = x(ξ), y(x) = , ξ0 =
(ξ 0 )1/2 dx
Since
dy dG 1 0 −3/2 00
= (ξ 0 )1/2 − (ξ ) ξ G
dx dξ 2
and
d2 y 2
0 3/2 d G 3 1
2
= (ξ ) 2
+ (ξ 0 )−5/2 (ξ 00 )2 G − (ξ 0 )−3/2 ξ 000 G
dx dξ 4 2
equation (E.195) becomes
d2 G f (x) ξ 000 3 (ξ 00 )2
= G + − G
dξ 2 (ξ 0 )2 2(ξ 0 )3 4 (ξ 0 )4
d2 G f (x)
= + S G
dξ 2 (ξ 0 )2
where S is the Schwartzian derivative of the function ξ = ξ(x)
ξ 000 3 (ξ 00 )2
S= −
2(ξ 0 )3 4 (ξ 0 )4
d2 y
For example, the equation 2
= ex y under Green-Liouville transformation
dx
d2 G f (x)
takes on the form = + S G. Therefore the choice (ξ 0 )2 = ex ,
dξ 2 (ξ 0 )2
1 d2 G
i.e. ξ = 2ex/2 with S = − 2 , transforms the given equation into =
4ξ dξ 2
1
1 − 2 G, which admits the general solution
4ξ
p
G = ξ(C1 I0 (ξ) + C2 K0 (ξ))
1396
Using the above solution, we can go back to the original variables.
d2 d 2 d
y(x) 2 y(x) + y(x) + αx y(x) − 2α·y(x) = 0 (E.196)
dx dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x) ∗ diff(y(x), x$2) + (diff(y(x), x))2

(E.197)
+ α ∗ x ∗ diff(y(x), x) − 2 ∗ α ∗ y(x) = 0;

d2 d 2 d
y(x) y(x) + y(x) + αx y(x) − 2αy(x) = 0 (E.198)
dx2 dx dx
>symgen(ode);
[− ξ = x, − η = 2y] (E.199)
We make the choice X = x and Y = 2y using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ 2 ∗ y = 0 : (E.200)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ 2 ∗ y = 1 : (E.201)
1397
n y y o
r(x, y) =− F 1 2 , s(x, y) = ln(x) +− F 2 2 (E.203)
x x
We take
y
r = 2 , s = ln(x) (E.204)
x
> itr := {x = exp(s(r)), y(x) = r ∗ exp(2 ∗ s(r))}; (E.205)
The system responds
x = es(r) , y(x) = re2s(r)

(E.206)


d2
2 d
3 d 2 d
r s(r) − 6r s(r) − (7r + α) s(r) − s(r) = 0 (E.208)
dr2 dr dr dr
The substitution
d
s(r) = U (r) (E.209)
dr
converts equation (E.208) into
dU
r − 6r2 U 3 − (7r + α)U 2 − U = 0, U = U (r) (E.210)
dr
The above is an Abel equation of the first kind which under the substitution
r
U (r) = (E.211)
w(r)
takes on the form
dw
w· = −(7r + α)w − 6r3 , w = w(r) (E.212)
dr
1398
Since (integrating (E.209) and using (E.211))
Z Z
r
s(r) = U (r)dr = dr (E.213)
w(r)
going back to the original variables by (E.204), we find that the solution of
the original equation (E.196) is given implicitly by
Z
r
ln(x) = dr (E.214)
w(r) r=y/x2
In the above we supposed that we know the general solution of (E.212). In

fact we can transform (E.212) under the substitution
Z
7
z = − (7r + α)dr = − r2 + αr = f (r) (E.215)
2
into the second kind Abel equation
dw(z) 6r3
w(z) − w(z) = (E.216)
dz 7r + α r=f −1 (z)
√
or, since f −1 (z) = (α± α2 − 14z)/7,
√
dw(z) 6 (α± α2 − 14z)3
w(z) − w(z) = √ (E.217)
dz 343 2α± α2 − 14z
The above equation can be solved using the Panayotounakos algorithm.
d2 2
d d 2
2y(x) 2 y(x) + xy (x) y(x) − 6 y(x) = 0 (E.218)
dx dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := 2 ∗ y(x) ∗ diff(y(x), x$2) + x ∗ y(x)2 ∗ (diff(y(x), x))
(E.219)
− 6 ∗ (diff(y(x), x))2 = 0;
1399
d2 d d 2
2
2y(x) y(x) + xy (x) y(x) − 6 y(x) =0 (E.220)
dx2 dx dx
>symgen(ode);
1
[− ξ = − x, − η = y] (E.221)
2
1
We make the choice X = − x and Y = y using our notation.
2
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1
> eq1 := diff(r(x, y), x) ∗ − x + diff(r(x, y), y) ∗ y = 0 : (E.222)
2
1
> eq2 := diff(s(x, y), x) ∗ − x + diff(s(x, y), y) ∗ y = 1 : (E.223)
2

n o
r(x, y) =− F 1 yx2 , s(x, y) = −2ln(x) +− F 2 yx2 (E.225)
We take
r = x2 y, s = −2 ln(x) (E.226)

1
> itr := {x = exp(− ∗ s(r)), y(x) = r ∗ exp(s(r))}; (E.227)
2
1400
The system responds
n o
− 21 s(r) s(r)
x=e , y(x) = re (E.228)


d2 1 d 3 1 d 2 3 d
s(r) + r(r + 6) s(r) + (r + 14) s(r) + s(r) =0
dr2 4 dr 4 dr r dr
(E.230)
The substitution
d
s(r) = Y (r) (E.231)
dr
dY 1 1 3
+ r(r + 6)Y 3 + (r + 14)Y 2 + Y = 0, Y = Y (r) (E.232)
dr 4 4 r
1
Y (r) = (E.233)
r3 w(r)
takes on the form
dw r + 14 r+6
w· = 3
w+ , w = w(r) (E.234)
dr 4r 4r5
Z Z
dr
s(r) = Y (r)dr = 3
(E.235)
r w(r)
Z
dr
−2 ln(x) = 3
(E.236)
r w(r) r=x2 y
1401
Z
r + 14 1 7
z= 3
dr = − − 2 = f (r) (E.237)
4r 4r 4r
into
dw(z) r+6
w(z) − w(z) = 2 (E.238)
dz r (r + 14) r=f −1 (z)
where
√
−1 −1± 1 − 112z
f (z) = (E.239)
8z
The above equation (E.238) can be solved by Panayotounakos algorithm.
d2 d 2 1 d
y(x) 2 y(x) + y(x) + x y(x) = 0 (E.240)
dx dx 2 dx
>restart;
>with(DEtools):
>with(PDEtools):
1
> ode := y(x) ∗ diff(y(x), x$2) + (diff(y(x), x))2 + x ∗ (diff(y(x), x)) = 0
2
(E.241)
d2 d 2 1 d
y(x) y(x) + y(x) + x y(x) =0 (E.242)
dx2 dx 2 dx
>symgen(ode);
[− ξ = x, − η = 2y] (E.243)
1402
We make the choice X = x and Y = 2y using our notation.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ 2 ∗ y = 0 : (E.244)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ 2 ∗ y = 1 : (E.245)

n y y o
r(x, y) =− F 1 2 , s(x, y) = ln(x) +− F 2 2 (E.247)
x x
We take
y
r= , s = ln(x) (E.248)
x2
> itr := {x = exp(s(r)), y(x) = r ∗ exp(2 ∗ s(r))}; (E.249)
The system responds
x = es(r) , y(x) = re2s(r)

(E.250)


d2 d 3 14r + 1 d 2 1 d
s(r) −(6r+1) s(r) − s(r) − s(r) = 0 (E.252)
dr2 dr 2r dr r dr
1403
The substitution
d
s(r) = Y (r) (E.253)
dr
dY 14r + 1 2 1
= (6r + 1)Y 3 + Y + Y, Y = Y (r) (E.254)
dr 2r r
r
Y (r) = (E.255)
w(r)
takes on the form
dw 14r + 1
w· =− w − (6r + 1)r2 , w = w(r) (E.256)
dr 2
Z Z
r
s(r) = Y (r)dr = dr (E.257)
w(r)
Z
r
ln(x) = dr (E.258)
w(r) r=y/x2

Z
14r + 1 7 1
z=− dr = − r2 − r = f (r) (E.259)
2 2 2
into
dw(z) 2(6r + 1)
w(z) − w(z) = (E.260)
dz (14r + 1) r=f −1 (z)
where
√
−1 −1± 1 − 56z
f (z) = (E.261)
14
1404
The above equation (E.260) is a second type Abel equation of the form
yy 0 − y = f (x) and thus can be solved by Panayotounakos algorithm.
d2 d 2
2
d
(1 − y(x)) y(x) + 2 y(x) + 2x (y(x) − 2y(x) + 1) y(x)
dx2 dx dx
3 2
+ 3y(x) − 9y(x) + 9y(x) − 3 = 0
(E.262)

>restart;
>with(DEtools):
>with(PDEtools):
> ode := (1 − y(x)) ∗ diff(y(x), x$2) + 2 ∗ (diff(y(x), x))2

+ 2 ∗ x ∗ (y(x)2 − 2 ∗ y(x) + 1) ∗ (diff(y(x), x)) (E.263)
+ 3 ∗ y(x)3 − 9 ∗ y(x)2 + 9 ∗ y(x) − 3 = 0

d2 d 2 d
2
(1 − y(x)) y(x) + 2 y(x) + 2x (y(x) − 2y(x) + 1) y(x)
dx2 dx dx
3 2
+ 3y(x) − 9y(x) + 9y(x) − 3 = 0
(E.264)

>symgen(ode);
1
[− ξ = − x, − η = −1 + y] (E.265)
2
1
We make the choice X = − x and Y = −1 + y using our notation.
2
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1405
1
> eq1 := diff(r(x, y), x) ∗ − x + diff(r(x, y), y) ∗ (−1 + y) = 0 : (E.266)
2
1
> eq2 := diff(s(x, y), x) ∗ ∗ − x + diff(s(x, y), y) ∗ (−1 + y) = 1 : (E.267)
2

n o
r(x, y) =− F 1 (−1 + y)x2 , s(x, y) = −2 ln(x) +− F 2 (−1 + y)x2
(E.269)
We take
r = (−1 + y)x2 , s = −2 ln(x) (E.270)

1
> itr := {x = exp(− ∗ s(r)), y(x) = 1 + r ∗ exp(s(r))}; (E.271)
2
The system responds
n 1
o
x = e− 2 s(r) , y(x) = 1 + res(r) (E.272)


d2 r(r − 6) d 3 2r − 7 d 2 3 d
s(r) − s(r) − s(r) + s(r) = 0 (E.274)
dr2 4 dr 2 dr r dr
The substitution
d
s(r) = Y (r) (E.275)
dr
1406
dY r(r − 6) 3 2r − 7 2 3
= Y + Y − Y, Y = Y (r) (E.276)
dr 4 2 r
1
Y (r) = (E.277)
r3 w(r)
takes on the form

dw 2r − 7 (r − 6)
w· =− 3
w− , w = w(r) (E.278)
dr 2r 4r5
Z Z
dr
s(r) = Y (r)dr = 3
(E.279)
r w(r)
Z
dr
−2 ln(x) = 3
(E.280)
r w(r) r=(−1+y)x2

2r − 7
Z
7 1
z=− dr = − + = f (r) (E.281)
2r3 4r2 r
into
dw(z) r−6
w(z) − w(z) = 2 (E.282)
dz 2r (2r − 7) r=f −1 (z)
where
√
−1 1± 1 − 7z
f (z) = (E.283)
2z
1407
d2 d 2
2
d
y(x) y(x) − 3 y(x) − 2xy(x) y(x) − 3y(x)3 = 0 (E.284)
dx2 dx dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := y(x) ∗ diff(y(x), x$2) − 3 ∗ (diff(y(x), x))2

(E.285)
− 2 ∗ x ∗ y(x)2 ∗ (diff(y(x), x)) − 3 ∗ y(x)3 = 0

d2 d 2 d
2
y(x) y(x) − 3 y(x) − 2xy(x) y(x) − 3y(x)3 = 0 (E.286)
dx2 dx dx
>symgen(ode);
[− ξ = x, − η = −2y] (E.287)
We make the choice X = x and Y = −2y using our notation.

∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x + diff(r(x, y), y) ∗ (−2y) = 0 : (E.288)

> eq2 := diff(s(x, y), x) ∗ x + diff(s(x, y), y) ∗ (−2y) = 1 : (E.289)
1408
r(x, y) =− F 1(yx2 ), s(x, y) = ln(x) +− F 2(yx2 )

(E.291)
We take
r = yx2 , s = ln(x) (E.292)
> itr := {x = exp(s(r)), y(x) = r ∗ exp(−2 ∗ s(r))}; (E.293)
The system responds
x = es(r) , y(x) = re−2s(r)

(E.294)

ode using the command (this single command is a substitute of Appendix B)

d2 d 3 d 2 3 d
s(r) −r(r−6) s(r) +(2r−7) s(r) + s(r) = 0 (E.296)
dr2 dr dr r dr
The substitution
d
s(r) = Y (r) (E.297)
dr
dY 3
= r(r − 6)Y 3 − (2r − 7)Y 2 − Y, Y = Y (r) (E.298)
dr r
1
Y (r) = (E.299)
r3 w(r)
takes on the form
dw 2r − 7 r−6
w· = 3
w− 5 , w = w(r) (E.300)
dr r r
1409
Z Z
dr
s(r) = Y (r)dr = 3
(E.301)
r w(r)
Z
dr
ln(x) = 3
(E.302)
r w(r) r=yx2

2r − 7
Z
2 7
z= 3
dr = − + 2 = f (r) (E.303)
r r 2r
into
dw(z) r−6
w(z) − w(z) = − 2 (E.304)
dz r (2r − 7) r=f −1 (z)
where
√
−1 −2± 4 + 14z
f (z) = (E.305)
2z
F Appendix
In this Appendix we shall find the symmetries of a 3rd order ordinary differ-
ential equation using ”Mathematica”. The required software is MathLie.
Example. Find the symmetries of the Chazy equation
d3 y d2 y dy 2
= 2y − 3 (F.1)
dx3 dx2 dx
1410
In[1]:=<<MathLie.m
Step 2. We enter the third order ODE using the command
In[2] := ode = ∂{x,3} y[x] == 2 ∗ y[x] ∗ ∂{x,2} y[x] − 3 ∗ (∂x y[x])2 ;

ode//LT F
yx,x,x == −3yx2 + 2yyx,x

detode//LT F
(ξ1 )y == 0
(φ1 )y,y == 0
6(φ1 )x + 2y(ξ1 )x,x − 4y(φ1 )x,y − (ξ1 )x,x,x + 3(φ1 )x,x,y == 0
− 2y(φ1 )x,x + (φ1 )x,x,x == 0
− 2φ1 − 2y(ξ1 )x − 3(ξ1 )x,x + 3(φ1 )x,y == 0
(ξ1 )x + (φ1 )y == 0
φ1 == −6 k3 − (k2 + 2 k3 x)y
ξ1 == k1 + x(k2 + k3 x)
1411
G Appendix
In this Appendix we shall determine the symmetries and then solve a 3rd
order ordinary differential equation using ”MAPLE”.
Example. Solve the Chazy equation
d3 y d2 y dy 2
= 2y − 3 (G.1)
dx3 dx2 dx
>restart;
>with(DEtools):
>with(PDEtools):
> ode := diff(y(x), x$3) = 2∗y(x)∗(diff(y(x), x$2))−3∗(diff(y(x), x))2 (G.2)
d3 d2 d 2
y(x) = 2y(x) y(x) − 3 y(x) = 0 (G.3)
dx3 dx2 dx
>DetSys:=DeterminingPDE(ode);
∂ ∂3
− ξ1 (x, y) = 0, − ξ1 (x, y) = 0,
∂y ∂x3
∂2 ∂ (G.4)
− η1 (x, y) = −3 − ξ1 (x, y) − − ξ1 (x, y) y
∂x2 ∂x
The above system can be solved using the command pdsolve:
>pdsolve(DetSys):
1 2
− ξ1 (x, y) = − C1x + − C2x+ − C3,− η1 (x, y) = (− − C1x− − C2)y −3 − C1
2
(G.5)
1412
From (G.5) we can read off the coefficients
1
X = C 1 x2 + C 2 x + C 3 , Y = −C1 xy − C2 y − 3C1 (G.6)
2
and thus
∂ ∂
Γ=X +Y
∂x ∂y
1 ∂ ∂
= C 1 x2 + C 2 x + C 3 + (−C1 xy − C2 y − 3C1 ) (G.7)
2 ∂x ∂y
1 ∂ ∂ ∂ ∂ ∂
= C 1 x2 − (xy + 3) + C2 x −y + C3
2 ∂x ∂y ∂x ∂y ∂x
Therefore the generators of the Lie algebra are given by
1 ∂ ∂ ∂ ∂ ∂
Γ(1) = x2 − (xy + 3) , Γ(2) = x − y , Γ(3) = (G.8)
2 ∂x ∂y ∂x ∂y ∂x
The command Infinitesimals gives the infinitesimal coefficients X and Y :
>Infinitesimals(ode);
The system responds
[− ξ1 (x, y) = 1, − η1 (x, y) = 0], [− ξ1 (x, y) = x, − η1 (x, y) = −y],

1 (G.9)
[− ξ1 (x, y) = x2 , − η1 (x, y) = −xy − 3]
2
Let us now use the command ”symgen”:
>symgen(ode);
[− ξ = 1, − η = 0], [− ξ = −x, − η = y],

(G.10)
[− ξ = x2 , − η = −6 − 2xy]
We thus see that ”Infinitesimals” and ”symgen” give identical results.

Equations (G.8), (G.9) and (G.10) are the same.
Let us first make the choice X = 1 and Y = 0 using our notation.
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
1413
> eq1 := diff(r(x, y), x) ∗ 1 + diff(r(x, y), y) ∗ 0 = 0 : (G.11)

> eq2 := diff(s(x, y), x) ∗ 1 + diff(s(x, y), y) ∗ 0 = 1 : (G.12)
> pdsolve({eq1, eq2}, {r(x, y), s(x, y)}); (G.13)
{r(x, y) =− F 1(y), s(x, y) = x +− F 2(y)} (G.14)
We take
r = y, s=x+y (G.15)
> itr := {x = s(r) − r, y(x) = r}; (G.16)
The system responds
{x = s(r) − r, y(x) = r} (G.17)

> dchange(itr, ode, simplif y); (G.18)
The system responds with an expression which, when simplified, has the form
d3 d d3 d2 2 d 2 d2
s(r) − s(r) s(r) + 2 s(r) + 2r s(r) s(r)
dr3 dr dr3 dr2 dr dr2
d d2 d2 d 3 d 2
− 4r s(r) s(r) + 2r s(r) + 3 s(r) − 9 s(r)
dr dr2 dr2 dr dr
d
+9 s(r) − 3 = 0
dr
(G.19)
1414
The substitution
d
s(r) = U (r) (G.20)
dr
converts equation (G.19) into
d2 U dU 2 dU
(1 − U ) 2
+ 2 + 2r(U 2 − 2U + 1)
dr dr dr (G.21)
3 2
+ 3U − 9U + 9U − 3 = 0, U = U (r)
The above is a second order ODE which can be reduced further using Lie
point symmetries. This was done in Example 15, Appendix E. The infinites-
imal coefficients are given by
1
X = − x, Y = −1 + y (G.22)
2
Considering the system of determining the normal coordinates, we find that
r = (−1 + y)x2 , s = −2 ln(x) (G.23)
Inverting the above transformation, equation (G.21) transforms into

d2 d 3 d 2 d
3 2 2
4r s(r) −(r −6r ) s(r) −(4r −14r) s(r) +12 s(r) =0
dr2 dr dr dr
(G.24)
The substitution
d
s(r) = u(r) (G.25)
dr
transforms (G.24) into the equation
du
4r − (r3 − 6r2 )u3 − (4r2 − 14r)u2 + 12u = 0 (G.26)
dr
du r(r − 6) 3 2r − 7 2 3
= u + u − u (G.27)
dr 4 2 r
1415
1
u(r) = (G.28)
r3 w(r)
takes on the form

dw 2r − 7 r−6
w· =− 3
w− , w = w(r) (G.29)
dr 2r 4r5
Since (integrating (G.25) and using (G.28))
Z Z
dr
s(r) = u(r)dr = 3
(G.30)
r w(r)
going back to the original variables by (G.23), we find that the solution of
equation (E.21) is given implicitly by
Z
dr
−2 ln(x) = 3
(G.31)
r w(r) r=(−1+y)x2
In the above we supposed that we know the general solution of (G.29). In

fact we can transform (G.29) under the substitution
2r − 7
Z
7 1
z=− 3
dr = − 2 + = f (r) (G.32)
2r 4r r
into
dw(z) r−6
w(z) − w(z) = 2 (G.33)
dz 2r (2r − 7) r=f −1 (z)
where
√
−1 1± 1 − 7z
f (z) = (G.34)
2z
The above equation (G.33) is a second type Abel equation of the form yy 0 −
y = f (x) and thus can be solved by Panayotounakos algorithm.
We turn next to the choice
X = x2 , Y = −6 − 2xy (G.35)
1416
and we continue to Step 4 for this choice (Steps 1-3 are the same as before
in this Example):
∂r ∂r ∂s ∂s
·X(x, y) + ·Y (x, y) = 0, ·X(x, y) + ·Y (x, y) = 1
∂x ∂y ∂x ∂y
> eq1 := diff(r(x, y), x) ∗ x2 + diff(r(x, y), y) ∗ (−6 − 2xy) = 0 : (G.36)

> eq2 := diff(s(x, y), x) ∗ x2 + diff(s(x, y), y) ∗ (−6 − 2xy) = 1 : (G.37)
> pdsolve({eq1, eq2}, {r(x, y), s(x, y)}); (G.38)

1
r(x, y) =− F 1((6 + xy)x), s(x, y) = − +− F 2((6 + xy)x) (G.39)
x
We take
1
r = (6 + xy)x, s=− (G.40)
x
1
> itr := {x = − , y(x) = r ∗ s(r)2 + 6 ∗ s(r)}; (G.41)
s(r)
The system responds

1
> itr := {x = − , y(x) = r s(r)2 + 6 s(r)}; (G.42)
s(r)

> dchange(itr, ode, simplif y); (G.43)
1417
The system responds with an expression which, when simplified, has the form
d2 2 d d3 d 2 d2
3 s(r) − s(r) s(r) + 2r s(r) s(r)
dr2 dr dr3 dr dr2 (G.44)
d 3
+3 s(r) = 0
dr
The substitution
d
s(r) = Y (r) (G.45)
dr
converts equation (G.44) into
d2 Y dY 2 dY
Y 2 −3 − 2rY 2 − 3Y 3 = 0, Y = Y (r) (G.46)
dr dr dr
The above is a second order ODE which can be reduced further using Lie
point symmetries. This was done in Example 16, Appendix E. The infinites-
imal coefficients are given by
X = x, Y = −2y (G.47)
Considering the system of determining the normal coordinates, we find that
r = yx2 , s = ln(x) (G.48)
Inverting the above transformation, i.e.
x = es(r) , y = r·e−2s(r) (G.49)
equation (G.46) transforms into
d2 d 3 d 2 d
3 2 2
r s(r) − (r − 6r ) s(r) + (2r − 7r) s(r) + 3 s(r) = 0
dr2 dr dr dr
(G.50)
The substitution
d
s(r) = u(r) (G.51)
dr
transforms (G.50) into the equation
du
r − (r3 − 6r2 )u3 + (2r2 − 7r)u2 + 3u = 0 (G.52)
dr
1418
du 3
= r(r − 6)u3 − (2r − 7)u2 − u (G.53)
dr r
1
u(r) = (G.54)
r3 w(r)
takes on the form
dw 2r − 7 r−6
w· = 3
w− 5 , w = w(r) (G.55)
dr r r
Since (integrating (G.51) and using (G.54))
Z Z
dr
s(r) = u(r)dr = 3
(G.56)
r w(r)
going back to the original variables by (G.48), we find that the solution of
equation (G.46) is given implicitly by
Z
dr
ln(x) = (G.57)
r3 w(r) r=yx2
In the above we supposed that we know the general solution of (G.55). In
fact we can transform (G.55) under the substitution
2r − 7
Z
2 7
z= 3
dr = − + 2 = f (r) (G.58)
r r 2r
into
dw(z) r−6
w(z) − w(z) = − 2 (G.59)
dz r (2r − 7) r=f −1 (z)
where
√
−1 −2± 4 + 14z
f (z) = (G.60)
2z
The above equation (G.59) is a second type Abel equation of the form yy 0 −
y = f (x) and thus can be solved by Panayotounakos algorithm.
Note. Considering the generator Γ(2) , Chazy equation is being reduced into
a very complicated equation which does not admit any symmetry and thus
is not worth of trying further.
1419
H Appendix
In this Appendix we shall determine the symmetries of a 1st order partial
differential equation (1st order PDE) using ”MAPLE”.
Example. Find the symmetries of the first order partial differential equation
ut = uux + (ux )2 (H.1)
>restart;
>with(DEtools):
>with(PDEtools):
>u := U (x, t)
The system responds with U (x, t)
> pde := diff(u, t) = u ∗ (diff(u, x)) + (diff(u, x))2 (H.2)
∂ ∂ ∂ 2
U (x, t) = U (x, t) U (x, t) + U (x, t) = 0 (H.3)
∂t ∂t ∂x
>DetSys:=DeterminingPDE(pde);
∂
− ξ1 (x, t, U ) = 0,
∂U
∂ ∂
− ξ1 (x, t, U ) = U − η1 (x, t, U ) −− η1 (x, t, U ),
∂t ∂U
∂ ∂
− ξ1 (x, t, U ) = 0, − ξ2 (x, t, U ) = 0, (H.4)
∂x ∂U
∂ ∂ ∂
− ξ2 (x, t, U ) = − − η1 (x, t, U ) , − ξ2 (x, t, U ) = 0,
∂t ∂U ∂x
∂ ∂ ∂2
η
− 1 (x, t, U ) = 0, η
− 1 (x, t, U ) = 0, − η1 (x, t, U ) = 0
∂t ∂x ∂U 2
The above system can be solved using the command > pdsolve:
>pdsolve(DetSys):
1420
− ξ1 (x, t, U ) = − − C2 t + − C3, − ξ2 (x, t, U ) = −− C1 t + − C4

(H.5)
− η1 (x, t, U ) =− C1 U + − C2
The generator of symmetries is then given by

∂ ∂ ∂
Γ = ξ1 + ξ 2 + η1
∂x ∂t ∂u
∂ ∂ ∂
= (−C2 t + C3) + (−C1 t + C4) + (C1 u + C2) (H.6)
∂x ∂t ∂u
∂ ∂ ∂ ∂ ∂ ∂
= C1 −t + u + C2 −t + + C3 + C4
∂t ∂u ∂x ∂u ∂x ∂t
Therefore the Lie algebra of symmetries is being spanned by the vectors
∂ ∂ ∂ ∂
X 1 = −t +u , X 2 = −t +
∂t ∂u ∂x ∂u (H.7)
3 ∂ 4 ∂
X = , X =
∂x ∂t
I Appendix
In this Appendix we shall find the symmetries of a system of three PDEs
(taken from the flow in a polytropic gas) using ”MathLie”. The radial and
angular velocity fields will be denoted by V r and V θ respectively while we
shall denote by H the depth of the fluid above a flat bottom. All these
functions depend on the three variables r, θ and t:
V r(r, θ, t), V θ(r, θ, t), H(r, θ, t) (I.1)
Solution. We shall use the MathLie software. We use the following steps:
In[1]:=<<MathLie.m
We then enter each one function separately.
Step 2. We enter the function V r in ”Mathematica” format
In[2]:=V r = vr[r, θ, t];
Step 3. We enter the function V θ in ”Mathematica” format
In[3]:=V r = vr[r, θ, t];
1421
Step 4. We enter the function H in ”Mathematica” format
In[4]:=H = h[r, θ, t];
Step 5. We enter the system of PDEs using the command
V θ ∗ ∂θ V r
In[5] := eqnSys = {∂t V r + V r ∗ ∂r V r + + ∂r H,
r
V θ ∗ ∂θ V r ∂θ H
∂t V θ + V r ∗ ∂r V θ + + ,
r r
V θ ∗ ∂θ H ∂ (V r ∗ r) ∂ V
r θ θ
∂t H + V r ∗ ∂r H + +H ∗ + };
r r r
eqnSys//LT F
vθ (vr)θ
hr + vr (vr)t + (vr)t + == 0
r
hθ vθ (vθ)θ
+ vr (vθ)r + (vθ)t + == 0
r r
hθ vθ vr + r (vr)r (vθ)θ
ht + hr vr + +h + == 0
r r r
In[6] := detSys = DeterminingEquations[eqnSys, {vr, vθ, h},

{r, θ, t}, {∂t V r, ∂t V θ, ∂t H}, {}]; detSys//LT F
The program responds citing a system of thirty PDEs. They are listed on
pp. 233-234 of Ref [4]. This shows the advanced symbolic capabilities of the
”MathLie” program.
J Appendix
Magnetohydrodynamics.
Reduction of Equations
1422
Abstract. We reduce the equations of Magnetohydrodynamics (MHD)
into a system of Ordinary Differential Equations of a single variable, using
Lie symmetries.
Keywords: Magnetohydrodynamics (MHD), Lie Symmetries, Lie Symme-

try Reduction.
Attention: This Appendix contains the subsections K, L, M and N where
each equation is being equmerated separately according to the section it be-
longs.
K Introduction
Magnetohydrodynamics is the study of motion of electrically conducting
fluids in the presence of a magnetic fluid. It was Hannes Alfvén who estab-
lished MHD as a separate subject (Alfvén [1]). The reader can consult some
specialized references on the subject looking at Refs. [2]-[8]. A historical
account of MHD is best provided by S. Molokov, R. Moreau and K. K. Mof-
fat (Ref. [9]). Magnetohydrodynamics is an essential ingredient for plasma
physics and plasma astrophysics (Refs. [10]-[22]).
The applications of Lie symmetries to the MHD equations can reduce consid-
erably the labor of finding exact solutions. Some basic references are those
of Nucci [23], Fuchs [25], Picard [26], Popovych [24], Cheviakov [28] and
Wu [27]. In this paper we consider the MHD equations and determine their
Lie symmetries. We then reduce the equations into a system of seven non-
linear ODEs of a single variable.
L The MHD Equations

The equations of Magnetohydrodynamics is a combination of Navier-Stokes
and Maxwell equations. Let u = (u, v, w) be the three components of the
velocity of the fluid, p = (px , py , pz ) the components of the pressure and
H = (f, g, h) the magnetic field components in 3−dimensions. It is sup-
posed that the components of the velocity, the pressure and the magnetic
field vectors depend both on the spatial components (x, y, z) and the time t.
We also denote by ν the viscosity, by µ the permeability and by σc the elec-
trical conductivity of the fluid. Let us also define the coefficient of magnetic
1423
viscosity νm by
c2
νm = (L.1)
4πµσc
The MHD equations can be written in Cartesian coordinates as
ut + uux + vuy + wuz − ν(uxx + uyy + uzz ) + px

1 (L.2)
+ {g(gx − fy ) − h(fz − hx )} = 0
4πµ
vt + uvx + vvy + wvz − ν(vxx + vyy + vzz ) + py

1 (L.3)
+ {h(hy − gz ) − f (gx − fy )} = 0
4πµ
wt + uwx + vwy + wwz − ν(wxx + wyy + wzz ) + pz
1 (L.4)
+ {f (fz − hx ) − g(hy − gz )} = 0
4πµ
ux + vy + wz = 0 (L.5)
ft − (ug − vf )y + (wf − uh)z − νm (fxx + fyy + fzz ) = 0 (L.6)
gt − (vh − wg)z + (ug − vf )x − νm (gxx + gyy + gzz ) = 0 (L.7)
ht − (wf − uh)x + (vh − wg)y − νm (hxx + hyy + hzz ) = 0 (L.8)
fx + gy + hz = 0 (L.9)
We further denote by ∆(1) , ∆(2) , ∆(3) , ∆(4) , ∆(5) , ∆(6) , ∆(7) and ∆(8)
the quantities
∆(1) = ut + uux + vuy + wuz − ν(uxx + uyy + uzz ) + px

1 (L.10)
+ {g(gx − fy ) − h(fz − hx )}
4πµ
∆(2) = vt + uvx + vvy + wvz − ν(vxx + vyy + vzz ) + py

1 (L.11)
+ {h(hy − gz ) − f (gx − fy )}
4πµ
∆(3) = wt + uwx + vwy + wwz − ν(wxx + wyy + wzz ) + pz
1 (L.12)
+ {f (fz − hx ) − g(hy − gz )}
4πµ
1424
∆(4) = ux + vy + wz (L.13)
∆(5) = ft − (ug − vf )y + (wf − uh)z − νm (fxx + fyy + fzz ) (L.14)
∆(6) = gt − (vh − wg)z + (ug − vf )x − νm (gxx + gyy + gzz ) (L.15)
∆(7) = ht − (wf − uh)x + (vh − wg)y − νm (hxx + hyy + hzz ) (L.16)
and
∆(8) = fx + gy + hz (L.17)
respectively.
M Lie Symmetries of the MHD Equations

∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4 (M.1)
∂u ∂v ∂w ∂p
∂ ∂ ∂
+ ψ1 + ψ2 + ψ3
∂f ∂g ∂h
The functions {ξ 1 , · · ·, ξ 4 }, {φ1 , · · ·, φ4 } and {ψ 1 , ψ 2 , ψ 3 } depend both on the

space-time variables x, y, z, t and the functions u, v, w, f, g, h and p.
1425
∂ ∂ ∂ ∂
P r(1) X = X + φx + φy + φz + φt
∂ ∂ ∂ ∂
∂ ∂ ∂ ∂
∂ ∂ ∂
+ πx + πy + πz (M.2)
∂px ∂py ∂pz
∂ ∂ ∂ ∂
+ ft + fx + fy + fz
∂ft ∂fx ∂fy ∂fz
∂ ∂ ∂ ∂
+ gt + gx + gy + gz
∂gt ∂gx ∂gy ∂gz
∂ ∂ ∂ ∂
+ ht + hx + hy + hz
∂ht ∂hx ∂hy ∂hz
The second prolongation of the vector X is denoted by P r(2) X and is given
by
P r(2) X = P r(1) X+
∂ ∂ ∂ ∂
+ φxx + φyy + φzz + φtt
∂uxx ∂uyy ∂uzz ∂utt
∂ ∂ ∂ ∂
+ ψ xx + ψ yy + ψ zz + ψ tt
∂vxx ∂vyy ∂vzz ∂vtt
∂ ∂ ∂ ∂
+ χxx + χyy + χzz + χtt
∂wxx ∂wyy ∂wzz ∂wtt
∂ ∂ ∂ (M.3)
+ π xx + π yy + π zz
∂pxx ∂pyy ∂pzz
∂ ∂ ∂
+ f xx + f yy + f zz
∂fxx ∂fyy ∂fzz
∂ ∂ ∂
+ g xx + g yy + g zz
∂gxx ∂gyy ∂gzz
∂ ∂ ∂
+ hxx + hyy + hzz
∂hxx ∂hyy ∂hzz
The Lie symmetries of MHD equations can be revealed once we know the
explicit form of the functions {ξ 1 , · · ·, ξ 4 }, {φ1 , · · ·, φ4 } and {ψ 1 , ψ 2 , ψ 3 }.
1426
P r(2) X[∆(k) ] = 0, k = 1, 2, · · · , 8 as long as

∆(1) = 0, ∆(2) = 0, ∆(3) = 0, ∆(4) = 0, (M.4)
∆(5) = 0, ∆(6) = 0, ∆(7) = 0 and ∆(8) = 0
The first step is to calculate the various P r(2) X[∆(k) ], k = 1, 2, · · · , 8.

We find
P r(2) X[∆(1) ] = φt + (φ1 )ux + (φ2 )uy + (φ3 )uz
+ uφx + vφy + wφz + π x − νφxx − νφyy − νφzz
(M.5)
1
+ [ψ 2 (gx − fy ) − ψ 3 (fz − hx ) − gf y − hf z + gg x + hhx ]
4πµ
P r(2) X[∆(2) ] = ψ t + (φ1 )vx + (φ2 )vy + (φ3 )vz

+ uψ x + vψ y + wψ z + π y − νψ xx − νψ yy − νψ zz
(M.6)
1
+ [−ψ 1 (gx − fy ) + ψ 3 (hy − gz ) + f f y − f g x − hg z + hhy ]
4πµ
P r(2) X[∆(3) ] = χt + (φ1 )wx + (φ2 )wy + (φ3 )wz
+ uχx + vχy + wχz + π z − νχxx − νχyy − νχzz
(M.7)
1
+ [ψ 1 (fz − hx ) − ψ 2 (hy − gz ) + f f z + gg z − f hx − ghy ]
4πµ
P r(2) X[∆(4) ] = φx + ψ y + χz (M.8)
P r(2) X[∆(5) ] = −φ1 (gy + hz ) + φ2 fy + φ3 fz
+ ψ 1 (vy + wz ) − ψ 2 uy − ψ 3 uz − gφy − hφz + f ψ y + f χz + f t (M.9)
+ vf y + wf z − ug y − uhz − νm f xx − νm f yy − νm f zz
P r(2) X[∆(6) ] = φ1 gx − φ2 (fx + hz ) + φ3 gz
− ψ 1 vx + ψ 2 (wz + ux ) − ψ 3 vz + gφx − f ψ x − hψ z + gχz − vf x (M.10)
+ g t + ug x + wg z − vhz − νm g xx − νm g yy − νm g zz
P r(2) X[∆(7) ] = φ1 hx + φ2 hy − φ3 (fx + gy )
− ψ 1 wx − ψ 2 wy + ψ 3 (ux + vy ) + hφx + hψ y − f χx − gχy (M.11)
− wf x − wg y + ht + uhx + vhy − νm hxx − νm hyy − νm hzz
P r(2) X[∆(8) ] = f x + g y + hz (M.12)
1427
The coefficients {φx , · · · , hzz } appearing in P r(2) X[∆(k) ], k = 1, · · · , 8 can
be evaluated using the formulas (2.39), (2.43) or (2.44) of Olver (Ref. [30]).
After evaluating the various P r(2) X[∆(k) ], k = 1, · · · , 8 we have to make the
substitutions
px −→ − (ut + uux + vuy + wuz ) + ν(uxx + uyy + uzz )
1 (M.13)
− {g(gx − fy ) − h(fz − hx )}
4πµ
py −→ − (vt + uvx + vvy + wvz ) + ν(vxx + vyy + vzz )
1 (M.14)
− {h(hy − gz ) − f (gx − fy )}
4πµ
pz −→ − (wt + uwx + vwy + wwz ) + ν(wxx + wyy + wzz )
1 (M.15)
− {f (fz − hx ) − g(hy − gz )}
4πµ
wz −→ − (ux + vy ) (M.16)
ft −→ (ug − vf )y − (wf − uh)z + νm (fxx + fyy + fzz ) (M.17)
gt −→ (vh − wg)z − (ug − vf )x + νm (gxx + gyy + gzz ) (M.18)
ht −→ (wf − uh)x − (vh − wg)y + νm (hxx + hyy + hzz ) (M.19)
hz −→ − (fx + gy ) (M.20)
to account for the conditions ∆(1) = 0, · · · , ∆(8) = 0.
The reader might gain some experience on the details of calculations by
consulting similar papers, like Ref. [29].
Using the above procedure, we have been able to evaluate the infinitesimals
{φ1 , · · ·, φ4 }
φ1 = a4 u + a3 v − a1 w + F10 (t)
φ2 = −a3 u + a4 v − a2 w + F20 (t)
(M.21)
φ3 = a1 u + a2 v + a4 w + F30 (t)
φ4 = 2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)
the infinitesimals {ψ 1 , ψ 2 , ψ 3 }
ψ 1 = a4 f + a3 g − a1 h
ψ 2 = −a3 f + a4 g − a2 h (M.22)
ψ 3 = a1 f + a2 g + a4 h
1428
and the infinitesimals {ξ 1 , · · ·, ξ 4 }
ξ1 = −a4 x + a3 y − a1 z + F1 (t)
ξ2 = −a3 x − a4 y − a2 z + F2 (t)
(M.23)
ξ3 = a1 x + a2 y − a4 z + F3 (t)
ξ4 = −2a4 t + a5
The infinitesimals {φ1 , · · ·, φ4 } and {ξ 1 , · · ·, ξ 4 } are identical to the corre-

sponding infinitesimals of the Navier-Stokes equations.
The infinitesimals {ψ 1 , ψ 2 , ψ 3 } may be derived from {φ1 , φ2 , φ3 } after setting
F10 (t) = F20 (t) = F30 (t) = 0.
∂ ∂ ∂ ∂
X = ξ1 + ξ2 + ξ3 + ξ4
∂x ∂y ∂z ∂t
∂ ∂ ∂ ∂
+ φ1 + φ2 + φ3 + φ4
∂u ∂v ∂w ∂p
∂ ∂ ∂
+ ψ1 + ψ2 + ψ3
∂f ∂g ∂h
∂
= {−a4 x + a3 y − a1 z + F1 (t)}
∂x
∂
+ {−a3 x − a4 y − a2 z + F2 (t)}
∂y
∂ ∂
+ {a1 x + a2 y − a4 z + F3 (t)} + {−2a4 t + a5 }
∂z ∂t (M.24)
∂
+ {a4 u + a3 v − a1 w + F10 (t)}
∂u
∂
+ {−a3 u + a4 v − a2 w + F20 (t)}
∂v
∂
+ {a1 u + a2 v + a4 w + F30 (t)}
∂w
∂
+ {2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)}
∂p
∂ ∂
+ (a4 f + a3 g − a1 h) + (−a3 f + a4 g − a2 h)
∂f ∂g
∂
+ (a1 f + a2 g + a4 h)
∂h
1429

∂ ∂ ∂ ∂ ∂ ∂
X = a1 − z +x −w +u −h +f
∂x ∂z ∂u ∂w ∂f ∂h

∂ ∂ ∂ ∂ ∂ ∂
+ a2 − z +y −w +v −h +g
∂y ∂z ∂v ∂w ∂g ∂h

∂ ∂ ∂ ∂ ∂ ∂
+ a3 y −x +v −u +g −f
∂x ∂y ∂u ∂v ∂f ∂g

∂ ∂ ∂ ∂ ∂ ∂
+ a4 − x −y −z − 2t + u +v
∂x ∂y ∂z ∂t ∂u ∂v
(M.25)
∂ ∂ ∂ ∂ ∂
+w + 2p + f +g +h
∂w ∂p ∂f ∂g ∂h

∂ ∂ ∂ 0 ∂ 00 ∂
+ a5 + F4 (t) + F1 (t) + F1 (t) − F1 (t)x
∂t ∂p ∂x ∂u ∂p

∂ ∂ ∂
+ F2 (t) + F20 (t) − F200 (t)y
∂y ∂v ∂p

∂ 0 ∂ 00 ∂
+ F3 (t) + F3 (t) − F3 (t)z
∂z ∂w ∂p
We thus see that the generators of the algebra of Lie symmetries of MHD
equations are given by
∂ ∂ ∂ ∂ ∂ ∂
X 1 = −z +x −w +u −h +f (M.26)
∂x ∂z ∂u ∂w ∂f ∂h
∂ ∂ ∂ ∂ ∂ ∂
X 2 = −z +y −w +v −h +g (M.27)
∂y ∂z ∂v ∂w ∂g ∂h
∂ ∂ ∂ ∂ ∂ ∂
X3 =y −x +v −u +g −f (M.28)
∂x ∂y ∂u ∂v ∂f ∂g
∂ ∂ ∂ ∂ ∂ ∂
X4 = −x −y −z − 2t + u +v
∂x ∂y ∂z ∂t ∂u ∂v
(M.29)
∂ ∂ ∂ ∂ ∂
+w + 2p + f +g +h
∂w ∂p ∂f ∂g ∂h
∂
X5 = (M.30)
∂t
∂
X6 = F4 (t) (M.31)
∂p
1430
∂ ∂ ∂
X 7 = F1 (t) + F10 (t) − F100 (t)x (M.32)
∂x ∂u ∂p
∂ ∂ ∂
X 8 = F2 (t) + F20 (t) − F200 (t)y (M.33)
∂y ∂v ∂p
∂ ∂ ∂
X 9 = F3 (t) + F30 (t) − F300 (t)z (M.34)
∂z ∂w ∂p
We shall now reduce the MHD equations into a system of ODEs of a single
variable.
N Reduction of Equations
ξ 1 ux + ξ 2 uy + ξ 3 uz + ξ 4 ut = φ1
ξ 1 vx + ξ 2 vy + ξ 3 vz + ξ 4 vt = φ2
ξ 1 wx + ξ 2 wy + ξ 3 wz + ξ 4 wt = φ3
ξ 1 px + ξ 2 py + ξ 3 pz + ξ 4 pt = φ4 (N.1)
ξ 1 fx + ξ 2 fy + ξ 3 fz + ξ 4 ft = ψ 1
ξ 1 gx + ξ 2 gy + ξ 3 gz + ξ 4 gt = ψ 2
ξ 1 hx + ξ 2 hy + ξ 3 hz + ξ 4 ht = ψ 3
i.e. by solving the system (in expanded form)
(−a4 x + a3 y − a1 z + F1 (t))ux + (−a3 x − a4 y − a2 z + F2 (t))uy

+ (a1 x + a2 y − a4 z + F3 (t))uz + (−2a4 t + a5 )ut (N.2)
= a4 u + a3 v − a1 w + F10 (t)
(−a4 x + a3 y − a1 z + F1 (t))vx + (−a3 x − a4 y − a2 z + F2 (t))vy

+ (a1 x + a2 y − a4 z + F3 (t))vz + (−2a4 t + a5 )vt (N.3)
= −a3 u + a4 v − a2 w + F20 (t)
(−a4 x + a3 y − a1 z + F1 (t))wx + (−a3 x − a4 y − a2 z + F2 (t))wy
+ (a1 x + a2 y − a4 z + F3 (t))wz + (−2a4 t + a5 )wt (N.4)
= a1 u + a2 v + a4 w + F30 (t)
1431
(−a4 x + a3 y − a1 z + F1 (t))px + (−a3 x − a4 y − a2 z + F2 (t))py
+ (a1 x + a2 y − a4 z + F3 (t))pz + (−2a4 t + a5 )pt (N.5)
= 2a4 p − F100 (t)x − F200 (t)y − F300 (t)z + F4 (t)
(−a4 x + a3 y − a1 z + F1 (t))fx + (−a3 x − a4 y − a2 z + F2 (t))fy
+ (a1 x + a2 y − a4 z + F3 (t))fz + (−2a4 t + a5 )ft (N.6)
= a4 f + a3 g − a1 h
(−a4 x + a3 y − a1 z + F1 (t))gx + (−a3 x − a4 y − a2 z + F2 (t))gy
+ (a1 x + a2 y − a4 z + F3 (t))gz + (−2a4 t + a5 )gt (N.7)
= −a3 f + a4 g − a2 h
(−a4 x + a3 y − a1 z + F1 (t))hx + (−a3 x − a4 y − a2 z + F2 (t))hy
+ (a1 x + a2 y − a4 z + F3 (t))hz + (−2a4 t + a5 )ht (N.8)
= a1 f + a2 g + a4 h
Reduction I. This case corresponds to the choice
a1 = a2 = a3 = a4 = 0 and a5 = 1 (N.9)
i.e. to the linear combination of generators
Γ = X5 + X6 + X7 + X8 + X9 (N.10)
The above can be written in expanded form as

∂ ∂ ∂ ∂
Γ= + F1 (t) + F2 (t) + F3 (t) +
∂t ∂x ∂y ∂z
∂ ∂ ∂
+ F10 (t) + F20 (t) + F30 (t) + (N.11)
∂u ∂v ∂w
∂
+ {F4 (t) − F100 (t)x − F200 (t)y − F300 (t)z}
∂p
The auxiliary system associated to the symmetry generator (N.11) is given

by
dt dx dy dz du dv dw
= = = = 0 = 0 = 0
1 F1 (t) F2 (t) F3 (t) F1 (t) F2 (t) F3 (t)
(N.12)
dp
=
F4 (t) − F1 (t)x − F200 (t)y − F300 (t)z
00
1432
Z
x̂ = x − F̂1 (t), F̂1 (t) = F1 (t) dt
Z
ŷ = y − F̂2 (t), F̂2 (t) = F2 (t) dt (N.13)
Z
ẑ = z − F̂3 (t), F̂3 (t) = F3 (t) dt
we can introduce the invariants {U, V, W, P } and {F, G, H} defined by
u = U (x̂, ŷ, ẑ) + F1 (t)

v = V (x̂, ŷ, ẑ) + F2 (t) (N.14)
w = W (x̂, ŷ, ẑ) + F3 (t)
p = P (x̂, ŷ, ẑ) − xF10 (t) − yF20 (t) − zF30 (t) + K(t)
(N.15)
Z
1 2 2 2
K(t) = [F1 (t) + F2 (t) + F3 (t)] + F̂4 (t), F̂4 (t) = F4 (t) dt
2
and
f = F (x̂, ŷ, ẑ)
g = G(x̂, ŷ, ẑ) (N.16)
h = H(x̂, ŷ, ẑ)
respectively.
We are now able to write down the MHD equations in terms of the new
functions {U, V, W, P } and {F, G, H}. They read
U Ux̂ + V Uŷ + W Uẑ − ν(Ux̂x̂ + Uŷŷ + Uẑẑ ) + Px̂

1 (N.17)
+ [G(Gx̂ − Fŷ ) − H(Fẑ − Hx̂ )] = 0
4πµ
U Vx̂ + V Vŷ + W Vẑ − ν(Vx̂x̂ + Vŷŷ + Vẑẑ ) + Pŷ

1 (N.18)
+ [H(Hŷ − Gẑ ) − F (Gx̂ − Fŷ )] = 0
4πµ
U Wx̂ + V Wŷ + W Wẑ − ν(Wx̂x̂ + Wŷŷ + Wẑẑ ) + Pẑ
1 (N.19)
+ [F (Fẑ − Hx̂ ) − G(Hŷ − Gẑ )] = 0
4πµ
1433
Ux̂ + Vŷ + Wẑ = 0 (N.20)
and
(U G − V F )ŷ − (W F − U H)ẑ + νm (Fx̂x̂ + Fŷŷ + Fẑẑ ) = 0 (N.21)
(V H − W G)ẑ − (U G − V F )x̂ + νm (Gx̂x̂ + Gŷŷ + Gẑẑ ) = 0 (N.22)

(W F − U H)x̂ − (V H − W G)ŷ + νm (Hx̂x̂ + Hŷŷ + Hẑẑ ) = 0 (N.23)
Fx̂ + Gŷ + Hẑ = 0 (N.24)
Introducing the generator Γ̂ of Lie symmetries corresponding to the equations
(N.17)-(N.24)
∂ ∂ ∂
Γ̂ = ξˆ1 + ξˆ2 + ξˆ3
∂ x̂ ∂ ŷ ∂ ẑ
∂ ∂ ∂ ∂ (N.25)
+ φ̂1 + φ̂2 + φ̂3 + φ̂4
∂U ∂V ∂W ∂P
∂ ∂ ∂
+ ψ̂ 1 + ψ̂ 2 + ψ̂ 3
∂F ∂G ∂H
and after performing a Lie symmetry analysis, we find the following coeffi-
cients
ξˆ1 = −c4 x̂ + c3 ŷ − c1 ẑ
ξˆ2 = −c3 x̂ − c4 ŷ − c2 ẑ (N.26)
ξˆ3 = c1 x̂ + c2 ŷ − c4 ẑ
φ̂1 = c4 U + c3 V − c1 W
φ̂2 = −c3 U + c4 V − c2 W
(N.27)
φ̂3 = c1 U + c2 V + c4 W
φ̂4 = 2c4 P
and
ψ̂ 1 = c4 F + c3 G − c1 H
ψ̂ 2 = −c3 F + c4 G − c2 H (N.28)
ψ̂ 3 = c1 F + c2 G + c4 H
1434
Lagrange’s auxiliary system associated to the generator (of the dilatation
subgroup, corresponding to the choice c1 = c2 = c3 = 0, c4 = 1)
∂ ∂ ∂
Γ̂ = −x̂ − ŷ − ẑ
∂ x̂ ∂ ŷ ∂ ẑ
∂ ∂ ∂ ∂ (N.29)
+U +V +W + 2P
∂U ∂V ∂W ∂P
∂ ∂ ∂
+F +G +H
∂F ∂G ∂H
reads
dx̂ dŷ dẑ dU dV dW dP
− =− =− = = = =
x̂ ŷ ẑ U V W 2P
(N.30)
dF dG dH
= = =
F G H
The auxiliary system leads to the introduction of the following invariants
ŷ ẑ
ξ= , η= (N.31)
x̂ x̂
and
1 1 1 1
U= X(ξ, η), V = Y (ξ, η), W = Z(ξ, η), P = 2 R(ξ, η)
x̂ x̂ x̂ x̂ (N.32)
1 1 1
F = L(ξ, η), G = M (ξ, η), H = N (ξ, η)
x̂ x̂ x̂
Under the above substitutions (N.31)-(N.32), equations (N.17)-(N.24) take
on the form
− X 2 − X(ξXξ + ηXη ) + Y Xξ + ZXη
− ν[2X + 4(ξXξ + ηXη ) + (1 + ξ 2 )Xξξ + 2ξηXξη + (1 + η 2 )Xηη ]
− 2R − (ξRξ + ηRη ) (N.33)
1
− [M (M + ξMξ + ηMη + Lξ ) + N (N + ξNξ + ηNη + Lη )] = 0
4πµ
− XY − X(ξYξ + ηYη ) + Y Yξ + ZYη
− ν[2Y + 4(ξYξ + ηYη ) + (1 + ξ 2 )Yξξ + 2ξηYξη + (1 + η 2 )Yηη ]
+ Rξ (N.34)
1
+ [N (Nξ − Mη ) + L(M + ξMξ + ηMη + Lξ )] = 0
4πµ
1435
− XZ − X(ξZξ + ηZη ) + Y Zξ + ZZη
− ν[2Z + 4(ξZξ + ηZη ) + (1 + ξ 2 )Zξξ + 2ξηZξη + (1 + η 2 )Zηη ]
+ Rη (N.35)
1
+ [M (Mη − Nξ ) + L(N + ξNξ + ηNη + Lη )] = 0
4πµ
−X − (ξXξ + ηXη ) + Yξ + Zη = 0 (N.36)
(XM − Y L)ξ − (ZL − XN )η + νm [2L + 4(ξLξ + ηLη )
(N.37)
+ (1 + ξ 2 )Lξξ + 2ξηLξη + (1 + η 2 )Lηη ] = 0
(Y N − ZM )η + 2(XM − Y L) + ξ(XM − Y L)ξ
+ η(XM − Y L)η + νm [2M + 4(ξMξ + ηMη ) (N.38)
+ (1 + ξ 2 )Mξξ + 2ξηMξη + (1 + η 2 )Mηη ] = 0
(Y N − ZM )ξ + 2(ZL − XN ) + ξ(ZL − XN )ξ
+ η(ZL − XN )η − νm [2N + 4(ξNξ + ηNη ) (N.39)
+ (1 + ξ 2 )Nξξ + 2ξηNξη + (1 + η 2 )Nηη ] = 0
and
−L − (ξLξ + ηLη ) + Mξ + Nη = 0 (N.40)
respectively.
Introducing a new variable θ by (Nucci [23])
θ =ξ−η (N.41)
and taking into account that Xξ = Xθ , Xη = −Xθ , etc., we can convert
the system of equations (N.33)-(N.40) to a system of ordinary differential
equations with respect to the variable θ.
Let us first transform (N.36). This equation can be written as
−X − θXθ + (Y − Z)θ = 0
The above equation is equivalent to (Y − Z)θ = (θX)θ which by integration
yields (we take the inegration constant equal to zero)
Y − Z = θX (N.42)
Similarly equation (N.40) can be written as
−L − θLθ + (M − N )θ = 0
1436
which is equivalent to (M − N )θ = (θL)θ and by integration yields (we take
the inegration constant equal to zero)
M − N = θL (N.43)
Equation (N.37) can be written with respect to the variable θ as
(XM − Y L)θ + (ZL − XN )θ

+ νm [2L + 4θLθ + (2 + θ2 )Lθθ ] = 0
[X(M − N ) − L(Y − Z)]θ

(N.44)
+ νm [2L + 4θLθ + (2 + θ2 )Lθθ ] = 0
Since, according to (N.43) and (N.42) M − N = θL and Y − Z = θX, we get

X(M − N ) − L(Y − Z) = X(θL) − L(θX) = 0 and thus equation (N.44) is
equivalent to
2L + 4θLθ + (2 + θ2 )Lθθ = 0 (N.45)
The above is a second order ODE admitting a closed-form solution

k1 θ + k2
L= (N.46)
2 + θ2
Equation (N.38) expressed in terms of the variable θ reads
− (Y N − ZM )θ + 2(XM − Y L) + θ(XM − Y L)θ

+ νm [2M + 4θMθ + (2 + θ2 )Mθθ ] = 0
[−(Y N − ZM ) + θ(XM − Y L)]θ + (XM − Y L)

(N.47)
+ νm [2M + 4θMθ + (2 + θ2 )Mθθ ] = 0
Since
− (Y N − ZM ) + θ(XM − Y L)
= −(Y N − ZM ) + (Y − Z)M − Y (M − N ) = 0
1437
(using θX = Y − Z and θL = M − N ), equation (N.47) takes on the form
XM − Y L + νm [2M + 4θMθ + (2 + θ2 )Mθθ ] = 0 (N.48)
The above equation can be solved with respect to Y :

1
Y = XM + νm [2M + 4θMθ + (2 + θ2 )Mθθ ] (N.49)
L
Equation (N.39) expressed in terms of the variable θ, takes on the form
(Y N − ZM )θ + 2(ZL − XN ) + θ(ZL − XN )θ
− νm [2N + 4θNθ + (2 + θ2 )Nθθ ] = 0
[(Y N − ZM ) + θ(ZL − XN )]θ + (ZL − XN )
(N.50)
− νm [2N + 4θNθ + (2 + θ2 )Nθθ ] = 0
Since
(Y N − ZM ) + θ(ZL − XN )
= (Y N − ZM ) + Z(M − N ) − (Y − Z)N = 0
(using θL = M − N and θX = Y − Z), equation (N.50) takes on the form

ZL − XN
(N.51)
− νm [2N + 4θNθ + (2 + θ2 )Nθθ ] = 0
The above equation can be solved with respect to Z:
1
Z= XN + νm [2N + 4θNθ + (2 + θ2 )Nθθ ] (N.52)
L
Subtracting (N.49) and (N.52) we get

1 νm
Y − Z = X(M − N ) + [2M + 4θMθ + (2 + θ2 )Mθθ ]
L L
(N.53)
2
− [2N + 4θNθ + (2 + θ )Nθθ ]
Since Y − Z = θX and M − N = θL, equation (4.53) yields
2M + 4θMθ + (2 + θ2 )Mθθ = 2N + 4θNθ + (2 + θ2 )Nθθ (N.54)
1438
Therefore equation (N.49) can be written as
1
Y = XM + νm [2N + 4θNθ + (2 + θ2 )Nθθ ] (N.55)
L
and since M = N + θL, we obtain the following expression for Y :
1
Y = XN + νm [2N + 4θNθ + (2 + θ2 )Nθθ ] + θX (N.56)
L
Equations (N.34) and (N.35) expressed in terms of the variable θ, take on
the form
− XY − θXYθ + Y Yθ − ZYθ
− ν[2Y + 4θYθ + (2 + θ2 )Yθθ ] + Rθ
(N.57)
1
+ [N (Nθ + Mθ ) + L(M + θMθ + Lθ )] = 0
4πµ
and
− XZ − θXZθ + Y Zθ − ZZθ
− ν[2Z + 4θZθ + (2 + θ2 )Zθθ ] − Rθ
(N.58)
1
+ [−M (Mθ + Nθ ) + L(N + θNθ − Lθ )] = 0
4πµ
respectively. Subtracting equations (N.57) and (N.58) (forming the difference
(N.57)-(N.58)), we obtain the equation
− X(Y − Z) − θ(Yθ − Zθ )X + (Y − Z)Yθ − (Y − Z)Zθ
− ν[2(Y − Z) + 4θ(Yθ − Zθ ) + (2 + θ2 )(Yθθ − Zθθ )]

1
+ L(M − N ) + θL(Mθ − Nθ ) + 2LLθ (N.59)
4πµ

+ N (Nθ + Mθ ) + M (Nθ + Mθ ) + 2Rθ = 0
We have further
L(M − N ) + θL(Mθ − Nθ ) + 2LLθ
+ N (Nθ + Mθ ) + M (Nθ + Mθ )
= LθL + (M − N )(Mθ − Nθ ) + 2LLθ + N (Nθ + Mθ ) (N.60)
+ M (Nθ + Mθ )
= θL2 + 2(M Mθ + N Nθ + LLθ )
1439
Therefore equation (N.59) becomes
− X(Y − Z) − θ(Yθ − Zθ )X + (Y − Z)(Yθ − Zθ )

(N.61)
1 2
+ θL + 2(M Mθ + N Nθ + LLθ ) + 2Rθ = 0
4πµ
We have further, taking into account Y − Z = θX, that
− X(Y − Z) − θ(Yθ − Zθ )X + (Y − Z)(Yθ − Zθ )

= −X(θX) − θ(θX)θ X + θXθ(θX)θ (N.62)
− ν[2(θX) + 4θ(θX)θ + (2 + θ2 )(θX)θθ ]
= −θX 2 − ν[2θX + 4θ(θX)θ + (2 + θ2 )(θX)θθ ]
Therefore equation (N.61) gets simplified into
− θX 2 − ν[2θX + 4θ(θX)θ + (2 + θ2 )(θX)θθ ]

(N.63)

1 2 2 2 2
+ θL + (M + N + L )θ + 2Rθ = 0
4πµ
where we have made use of the identity
2(M Mθ + N Nθ + LLθ ) = (M 2 + N 2 + L2 )θ (N.64)
Equation (N.33) expressed in terms of the variable θ, takes on the form
− X 2 − θXXθ + Y Xθ − ZXθ
− ν[2X + 4θXθ + (2 + θ2 )Xθθ ] − 2R − θRθ
(N.65)
1
− [M (M + θMθ + Lθ ) + N (N + θNθ − Lθ )] = 0
4πµ
We have
− θXXθ + Y Xθ − ZXθ = −θXXθ + (Y − Z)Xθ
(N.66)
= −θXXθ + θXXθ = 0
1440
making use of Y − Z = θX, and
M (M + θMθ + Lθ ) + N (N + θNθ − Lθ )
= M 2 + θM Mθ + M Lθ + N 2 + θN Nθ − N Lθ
= M 2 + N 2 + θ(M Mθ + N Nθ ) + (M − N )Lθ (N.67)
= M 2 + N 2 + θ(M Mθ + N Nθ ) + θLLθ
1
= M 2 + N 2 + θ(M 2 + N 2 + L2 )θ
2
making use of M − N = θL and the identity (N.64).
Therefore equation (N.65) takes on the form
− X 2 − ν[2X + 4θXθ + (2 + θ2 )Xθθ ] − 2R − θRθ
(N.68)

1 2 2 1 2 2 2
− M + N + θ(M + N + L )θ = 0
4πµ 2
Let us next solve equation (N.63) with respect to Rθ :

1 2 1 2
Rθ = θX + ν θX + 2θ(θX)θ + (2 + θ )(θX)θθ
2 2
(N.69)
1 1 2 1 2 2 2
− θL + (M + N + L )θ
4πµ 2 2
We then substitute the above expression into (N.68) and solve the resulting
equation with respect to R:
1 1 1
R = − X 2 − (θX)2 − ν Ω(X)
2 4 2
1 1 1
(N.70)
2 2 2
+ (θL) − (M + N )
4πµ 4 2
where Ω(X) is the expression
Ω(X) = (2 + θ2 )X + 4θXθ + 2θ2 (θX)θ + (2 + θ2 )Xθθ
1
+ θ(2 + θ2 )(θX)θθ
2
written as
1
Ω(X) = θ(θX) + 2θ2 (θX)θ + θ(2 + θ2 )(θX)θθ
2 (N.71)
2
+ [(2 + θ )X]θθ
1441
making use of the identity
(2 + θ2 )Xθθ + 4θXθ + 2X = [(2 + θ2 )X]θθ
Differentiation of (N.70) with respect to θ yields

1 1
Rθ = −XXθ − (θX)(θX)θ − ν [Ω(X)]θ
2 2 (N.72)
1 1 1 2 2
+ (θL)(θL)θ − (M + N )θ
4πµ 2 2
We thus have obtained two different expressions of Rθ , equations (N.69) and
(N.72) respectively. Equating these two expressions, we obtain the equation

1 2 1 2
θX + ν θX + 2θ(θX)θ + (2 + θ )(θX)θθ
2 2

1 1 2 1 2 2 2
− θL + (M + N + L )θ
4πµ 2 2 (N.73)
1 1
= −XXθ − (θX)(θX)θ − ν [Ω(X)]θ
2 2
1 1 1 2 2
+ (θL)(θL)θ − (M + N )θ
4πµ 2 2

1 1 2 2
The term − (M + N )θ which appears to both terms of the above
4πµ 2
equation, cancels out. We thus obtain from (N.73) the equivalant equation
1 2 1
θX + XXθ + (θX)(θX)θ
2 2
1 2 1
+ ν θX + 2θ(θX)θ + (2 + θ )(θX)θθ + [Ω(X)]θ (N.74)
2 2

1
− θL2 + (L2 )θ + (θL)(θL)θ = 0
8πµ
The above is an ordinary differential equation of third order with respect to
the variable θ where X is the unknown function, since L has already been
found (it is given by equation (N.46)).
1 1 1
XXθ + (θX)(θX)θ = (X 2 )θ + [(θX)2 ]θ (N.75)
2 2 4
1442
and
1 1
θX + 2θ(θX)θ + (2 + θ2 )(θX)θθ + [Ω(X)]θ
2 2
1
(N.76)
= 2θ(θX) + (2 + θ2 )(θX)θ + Ω(X)
2 θ
equation (N.74) becomes

1 2 1 2 1
θX + (X )θ + [(θX)2 ]θ
2 2 4
1
+ ν[2θ(θX) + (2 + θ2 )(θX)θ + Ω(X)]θ (N.77)
2
1 2 2
− θL + (L )θ + (θL)(θL)θ = 0
8πµ
However, since L is given by (N.46), we find
k1 θ + k2
θL2 + (L2 )θ + (θL)(θL)θ = k1 (N.78)
2 + θ2
and thus (N.77) takes on the form (after multiplying by 2)
1
θX 2 + (X 2 )θ + [(θX)2 ]θ
2
+ ν[2θ(θX) + (2 + θ2 )(θX)θ + Ω(X)]θ (N.79)
k1 k1 θ + k2
− =0
4πµ 2 + θ2
We have further the identities
1 1
(X 2 )θ + [(θX)2 ]θ = [(2 + θ2 )X 2 ]θ (N.80)
2 2
and
2θ(θX) + (2 + θ2 )(θX)θ = [(2 + θ2 )(θX)]θ (N.81)
Because of the above two identities, equation (N.79) assumes the form
1
θX 2 + [(2 + θ2 )X 2 ]θ + ν[[(2 + θ2 )(θX)]θ + Ω(X)]θ
2
k1 k1 θ + k2 (N.82)
− =0
4πµ 2 + θ2
1443
Let us next try to determine the equation which will contain the unknown
function N . Combining (N.58) and (N.69) (i.e. substitution of Rθ given by
(N.69) into (N.58)), we obtain, after the relevant simplifications and cancel-
lations, the equation

1 2 1 2
θX + ν θX + 2θ(θX)θ + (2 + θ )(θX)θθ
2 2
+ XZ + ν[2Z + 4θZθ + (2 + θ2 )Zθθ ] (N.83)

1 1 2
− θL + LN = 0
4πµ 2
However the function Z is being expressed in terms of N through the equation
(N.52). Writing (N.52) as
X
Z= N + νm Φ(N ) (N.84)
L
with
Φ(N ) = [2N + 4θNθ + (2 + θ2 )Nθθ ]L−1 (N.85)
one can transform (N.83) into the equation

1 2 1 2
θX + ν θX + 2θ(θX)θ + (2 + θ )(θX)θθ
2 2
X
+X N + νm Φ(N )
L
X X
+ν 2 N + νm Φ(N ) + 4θ N + νm Φ(N ) (N.86)
L L θ
X
+ (2 + θ2 ) N + νm Φ(N )
L θθ

1 1 2
− θL + LN = 0
4πµ 2
The above is a fourth order ODE with unknown function N (since L is a
known function and X has been determined before from (N.82)).
Using the identity
1
θX + 2θ(θX)θ + (2 + θ2 )(θX)θθ
2 (N.87)
1
= [2θ(θX) + (2 + θ2 )(θX)θ ]θ
2
1444
and introducing the function Θ(N ) by
X
Θ(N ) = N + νm Φ(N ) (N.88)
L
equation (N.86) takes on the form
1 2 1
θX + ν[2θ(θX) + (2 + θ2 )(θX)θ ]θ + X· Θ(N )
2 2
+ ν 2 Θ(N ) + 4θ· Θ(N )θ + (2 + θ2 )· Θ(N )θθ (N.89)

1 1 2
− θL + LN = 0
4πµ 2
2θ(θX) + (2 + θ2 )(θX)θ = [(2 + θ2 )(θX)]θ (N.90)
and
2 Θ(N ) + 4θ· Θ(N )θ + (2 + θ2 )· Θ(N )θθ
= [2θ· Θ(N ) + (2 + θ2 )· Θ(N )θ ]θ (N.91)
= [(2 + θ2 )· Θ(N )]θθ
equation (N.89) assumes its final form

1 2 1
θX + ν [(2 + θ2 )(θX)]θθ + X· Θ(N )
2 2
1 1 2
(N.92)
2
+ ν [(2 + θ )· Θ(N )]θθ − θL + N L = 0
4πµ 2
Conclusion I. The equations of Magnetohydrodynamics (L.2)-(L.9) admit
the Lie symmetry generators {X 1 , · · · , X 9 } given by (M.26)-(M.34). The
linear combination
Γ = X5 + X6 + X7 + X8 + X9 (N.93)
converts the MHD equations (L.2)-(L.9) (after considering the invariants

(N.13)-(N.16)) to the system of equations (N.17)-(N.24).
A further reduction of (N.17)-(N.24) leads to the determination of the in-
finitesimals {ξˆ1 , ξˆ2 , ξˆ3 }, {φ̂1 , φ̂2 , φ̂3 , φ̂4 } and {ψ̂ 1 , ψ̂ 2 , ψ̂ 3 } given by (N.26)-
(N.28).
1445
Considering the dilatation subgroup, expressed by the generators (N.29),
after introducing the invariants (N.31)-(N.32), equations (N.17)-(N.24) re-
duce to (N.33)-(N.40). Introducing a new variable θ by θ = ξ − η, equations
(N.33)-(N.40) reduce to a system of ODEs depending on a single variable θ.
The functions {X, Y, Z, R} and {L, M, N } depending on θ, have to be
evaluated according to the following order:
(a) The function L is given by (N.46).
(b) The function X is determined by solving (N.82) where Ω(X) is the ex-
pression given by (N.71).
(c) The function N is determined by solving (N.92), where Θ(N ) and Φ(N )
are given by (N.88) and (N.85) respectively.
(d) The function Z is determined by (N.84) where Φ(N ) is given by (N.85).
(e) The function M is determined from (N.43), i.e. M = N + θL.
(f) The function Y is determined from (N.56).
(g) The function R is determined from (N.70).
Conclusion II. The original functions {u, v, w, p} and {f, g, h} of the MHD
equations (L.2)-(L.9) are determined by combining (N.14)-(N.16) with (N.31)-
(N.32). We thus have
1
u= X(θ) + F1 (t)
x̂
1
v = Y (θ) + F2 (t)
x̂ (N.94)
1
w = Z(θ) + F3 (t)
x̂
1
p = 2 R(θ) − xF10 (t) − yF20 (t) − zF30 (t)
x̂
and
1 1 1
f= L(θ), g= M (θ), h= N (θ) (N.95)
x̂ x̂ x̂
where
ŷ − ẑ
θ =ξ−η = (N.96)
x̂
with {x̂, ŷ, ẑ} given by (N.13). End of Appendix J.
1446
References
[1] H. Alfvén: ”Existence of electromagnetic-hydrodynamic waves”.
Nature 150 (1942) 405-406.
[2] J. P. Feidberg: ”Ideal Magnetohydrodynamics”. Plenum Press 1987
[3] P. H. Roberts: ”An Introduction to Magnetohydrodynamics”.
Longmans 1967.
[4] R. J. Hosking and R. L. Dewar: ”Fundamental Fluid Mechanics and
Magnetohydrodynamics”. Springer 2016.
[5] R. Moreau: ”Magnetohydrodynamics”. Springer 1990.
[6] D. Biskamp: ”Nonlinear Magnetohydrodynamics”.
Cambridge University Press 1997.
[7] H. P. Goedbloed and S. Poedts: ”Principles of Magnetohydrody-
namics. With Applications to Laboratory and Astrophysical Plasmas”.
Cambridge University Press 2004.
[8] H. P. Goedbloed, R. Keppens and S. Poedts: ”Advanced Mag-
netohydrodynamics. With Applications to Laboratory and Astrophysical
Plasmas”. Cambridge University Press 2010.
[9] S. Molokov, R. Moreau and H. K. Moffat: ”Magnetohydrodynam-
ics. Historical Evolution and Trends”. Springer 2007.
[10] K. Miyamoto: ”Plasma Physics and Controlled Nuclear Fusion”.
Springer 2005.
[11] C. Chinderi and M. Velli: ”Basics of Plasma Astrophysics”.
Springer 2015.
[12] B. V. Somov: ”Cosmic Plasma Physics”. Springer 2000.
[13] B. V. Somov: ”Plasma Astrophysics”. Part I. Springer 2006.
[14] B. V. Somov: ”Plasma Astrophysics”. Part II. Springer 2006.
[15] P. A. Sturrock: ”Plasma Physics”. Cambridge University Press 1994.
[16] T. Tajima: ”Computational Plasma Physics. With Applications to
Fusion and Astrophysics”. Taylor and Francis 2004.
[17] V. E. Fortov, I. T. Iakubov and A. G. Khrapak: ”Physics of
Strongly Coupled Plasma”. Clarendon Press, Oxford 2006.
[18] M. Goossens: ”An Introduction to Plasma Astrophysics and Magne-
tohydrodynamics”. Springer 2003.
[19] W. Baumjohann and R. A. Treumann: ”Basic Space Plasma
Physics”. Imperial College Press 1997.
1447
[20] L. F. Burlaga: ”Interplanetary Magnetohydrodynamics”.
Oxford University Press 1995.
[21] E. R. Priest: ”Solar Magnetohydrodynamics”. D. Reidel 2000.
[22] E. R. Priest and A. W. Hood (eds.): ”Advances in Solar System
Magnetohydrodynamics”. Cambridge University Press 1991.
[23] M. C. Nucci: ”Group Analysis for MHD Equations”.
Atti Sem. Mat. Fis. Univ. Modena XXXIII (1984) 21-34
[24] V. Popovych: ”On Lie Reduction of the MHD Equations to Ordinary
Differential Equations”. Sym. Nonlin. Math. Phys. Vol. 1 (1997) 227-231
[25] J. C. Fuchs: ”Symmetry groups and similarity solutions of MHD Equa-
tions”. J. Math. Phys. 32 (1991) 1703-1708
[26] P. Y. Picard: ”Some spherical solutions of ideal Magnetohydrody-
namic Equations”. J. Nonlin. Math. Phys. 14 (1991) 578-588
[27] J. Wu: ”Symmetry Analysis of the Nonlinear MHD Equations”.
Phys. Lett. A 185 (1994) 304-309
[28] A. Cheviakov: ”Bogoyavlenskij Symmetries of Isotropic and Anisotropic
MHD Equilibria as Lie Point Transformations”.
Proc. Inst. Math. NAS Ukraine 50 (2004) 69-76
[29] S. Antoniou: ”The Euler and Navier-Stokes Equations. Lie Symme-
tries and Exact Solutions”. Version 2. SKEMSYS report 2022
Springer 2000
1448
General References
[1] W. F. Ames: ”Nonlinear Partial Differential Equations in Engineering”.
Vol 1 and Vol 2, Academic Press, New York 1965 and 1972.
tions in Applications”. SIAM 1979
[3] D. J. Arrigo: ”Symmetry Analysis of Differential Equations”.
Wiley 2015
[4] G. Baumann: ”Symmetry Analysis of Differential Equations with
Mathematica R ”. Springer, TELOS R 2000. The ”MathLie” program.
[5] G. W. Bluman and S. Kumei: ”Symmetries and Differential
Equations”. Springer-Verlag 1989
[6] G. W. Bluman and S. C. Anco: ”Symmetry and Integration Methods
for Differential Equations”. Springer-Verlag 2002
[7] G. W. Bluman, A. F. Cheviakov and S.C. Anco: ”Applications of
Symmetry Methods to Partial Differential Equations”. Springer-Verlag 2010
[8] J. E. Campbell: ”Introductory Treatise on Lie’s Theory of Finite
Continuous Transformation Groups”. Oxford 1903
[9] B. J. Cantwell: ”Introduction to Symmetry Analysis”. CUP 2002
[10] R. Cherniha, M. Serov and O. Pliukhin: ”Nonlinear Reaction-
Diffusion-Convection Equations”. CRC 2018
[11] A. Cohen: ”An Introduction to the Lie Theory of One-Parameter
Groups”. D. C. Heath and Co. 1911
[12] L. E. Dickson:”Differential Equations from the Group Standpoint”.
Ann. Math. Second Series, Vol. 25, No. 4 (Jun., 1924), pp. 287-378
Differential Equations”. IOP Publishing 1999
[14] L. P. Eisenhart: ”Continuous Groups of Transformations”.
Dover 1961
[16] N. Euler and W. H. Steeb: ”Continuous Symmetries, Lie Algebras
and Differential Equations”. Mannheim 1992
[17] W. I. Fushchych, W. M. Shtelen and M. I. Serov: ”Symme-
try Analysis and Exact Solutions of Equations of Nonlinear Mathematical
Physics”. Kluwer, Dordrecht, The Netherlands 1993
1449
[18] A. G. Hansen: ”Similarity Analysis of Boundary Value Problems
in Engineering”. Prentice Hall 1964
[19] T. Hawkins: ”The Emergence of the Theory of Lie Groups: An Essay
in the History of Mathematics 1869-1926”. Springer-Verlag, Berlin 2000
[20] T. Hawkins: ”Jacobi and the birth of Lie’s Theory of Groups”.
Arch. Hist. Exact Sc. 24 (1991) 187-278
[21] J. M. Hill: ”Differential Equations and Group Methods for Scientists
and Engineers”. CRC Press 1992
[22] P. E. Hydon: ”Symmetry Methods for Differential Equations”.
CUP 1999
[23] N. H. Ibragimov: ”Elementary Lie Group Analysis and Ordinary
Differential Equations”. Wiley 2001
[24] N. H. Ibragimov: ”Transformation Groups Applied to Mathematical
Physics”. D. Reidel 1985
[25] N. H. Ibragimov: ”CRC Handbook of Lie Group Analysis of Differ-
ential Equations”. Volume 1, 1994. Volume 2, 1995. Volume 3 1996.
CRC Press.
[26] N. H. Ibragimov: ”A Practical Course in Differential Equations and
Mathematical Modelling”. World Scientific 2015
[27] B. Komrakov and V. Lychagin: ”Symmetries and Integrals”.
University of Oslo, Matematisk Institutt report, June 1993
[28] S. Lie: ”Theory of Transformation Groups”. Editor and Translator:
Joël Merker, Springer 2015
[29] S. Lie: ”Theorie der Transformationsgruppen”. Erster Abschnitt.
[30] S. Lie: ”Theorie der Transformationsgruppen”. Zweiter Abschnitt.
[31] S. Lie: ”Theorie der Transformationsgruppen”. Dritter und Letzter
Abschnitt. Teubner, Leipzig 1893
[32] S. Lie: ”Geometrie der Berührungstransformationen”.
[33] S. Lie: ”Vorlensungen über Differentialgleichungen mit bekannten
infinitesimalen Transformationen”. Teubner, Leipzig 1912
[34] E. L. Mansfield: ”A Practical Guide to the Invariant Calculus”.
CUP 2010
1450
[36] W. Miller: ”Symmetry and Separation of Variables”.
Addison-Wesley, 1977
[37] E. Noether: ”Invariante Variationsprobleme”. Nachr. d. König.
Gesellsch. d. Wiss. zu Göttingen, Math-phys. Klasse 1918 235-237 (English
Translation: E. Noether and M. A. Tavel ”Invariant Variation Problems”,
Transport Theory and Stat. Mech. 1 (1971) 183-207)
[38] M. C. Nucci: ”Interactive REDUCE Programs for Calculating Lie
Point, Nonclassical, Lie-Bäcklund and Approximate Symmetries of Differen-
tial Equations”. In ”CRC Handbook of Lie Group Analysis of Differential
Equations”, VOL. III, N. H. Ibragimov (Ed.) CRC Press, 1996
Second Edition, Springer-Verlag 1993
[40] P. J. Olver: ”Equivalence, Invariants and Symmetry”. CUP 1995
Academic Press, 1982
[42] A. D. Polyanin and V. F. Zaitsev: ”Handbook of Exact Solutions
for Ordinary Differential Equations”. Second Edition, CRC 2003
[43] C. Rogers and W. F. Ames: ”Nonlinear Boundary Value Problems
in Science and Engineering”. Academic Press, New York 1989
[44] I. Shingareva and C. Lizrraga-Celaya: ”Solving Nonlinear Partial
Differential Equations with Maple and Mathematica”. Springer 2011
[45] S. Steinberg and R. de Melo Marinho Jr: ”Practical Guide to the
Symbolic Computation of Symmetries of Differential Equations”.
arXiv:1409.8364
[46] H. Stephani: ”Differential Equations. Their Solution using
Symmetries”. CUP 1989
[47] A. R. Tresse: ”Sur les invariants differéntial de groupes continus de
transformations”. Acta Math. 18 (1894) 1-88, euclid.acta.1485881898
[48] E. Vessiot: ”Sur l’integration des systèmes différentiels des groupes
continus de transformations”.
Acta Math. 28 (1904) 307-349, euclid.acta.1485887115
1451

01-S - Antoniou - Lie Symmetries, Differential Equations and Applications

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

01-S - Antoniou - Lie Symmetries, Differential Equations and Applications

Uploaded by

Copyright:

Available Formats

Lie Symmetries,

Scientific Knowledge Engineering

Keywords: Lie Symmetries, Ordinary Differential Equations, Partial

Chapter I. General Considerations

Chapter II. Ordinary Differential Equations

Chapter III. Partial Differential Equations

Chapter IV. Higher Order Partial Differential

Chapter VI. Fundamental Solutions

Chapter VII. The Direct Method

Chapter VIII. Systems of Differential Equations

Chapter IX. The Optimal System

Chapter X. Potential Symmetries

Chapter XI. Generalized Symmetries

Chapter XII. Recursion Operators

Chapter XIII. Conditional Symmetries

Chapter XV. Conservation Laws

Chapter XVI. Ricci Flow

Chapter XVII. Applications in Quantum Mechanics

Chapter XVIII. Applications in Epidemiology

Chapter XIX. Some Special Equations

Chapter XXI. General Relativity

Chapter XXII. Differential Geometry

Chapter XXIII. Financial Applications

Chapter XXIV. Applications in Cosmology

Chapter XXV. Additional Equations

Chapter XXVI. Fractional Differential Equations

(a, b) → φ(a, b) (1.1)

2 One-parameter Lie group

(x1 , x2 , ..., xn ) → (x̂1 , x̂2 , ..., x̂n )

x→x̂, where x = (x1 , x2 , ..., xn ) and x̂ = (x̂1 , x̂2 , ..., x̂n )

x̂i = f i (x1 , x2 , ..., xn ; ), or x̂i = f i (x; ), i = 1, 2, ..., n

is called a one-parameter local Lie group if

f i (f (x, a), b) = f i (x, φ(a, b)), with f i (x; 0) = xi (2.1)

where φ(a, b) is a given composition law, for example φ(a, b) = a + b.

φ(a, φ(b, c)) = a + (b + c) = (a + b) + c = φ(φ(a, b), c) (2.6)

which means that associativity is being satisfied too.

3 Invariance of Differential Equations

where λ is a scale parameter. Since

x = λα x̂, y = λ−α ŷ (3.5)

x = e x̂, y = e− ŷ (3.6)

x = eλ x̂, y = eµ ŷ (3.7)

and require invariance of (3.1). We will then obtain λ + µ = 0, i.e. recover

x = eλ x̂, y = eµ ŷ (3.9)

The above equation is equivalent to

which can also be written as

x = e x̂, y = e−2 ŷ (3.12)

x = λα x̂, t = λβ t̂, u = λγ û (3.14)

where λ is a scale parameter. Since

x = λα x̂, t = λ2α t̂, u = λγ û (3.17)

x̂ = f (x, y; ), ŷ = g(x, y; ) (4.1)

f (x, y; 0) = x and g(x, y; 0) = y (4.2)

We expand x̂ and ŷ in a Taylor series around  = 0:

x̂ = x + X(x, y) + O(2 )

Expressions (4.6) are referred to as infinitesimal transformations while

x̂ = x + x2 + O(2 ), ŷ = y + xy + O(2 ) (4.8)

The Lie equations are given by

ln(ŷ) = − ln( + C1 ) + C =⇒ ln(ŷ) = − ln( + C1 ) + ln(C2 )

We can now find the explicit expressions for x̂ and ŷ:

x̂i = f i (x; ) (4.17)

f i (f (x, a); ) = f i (x; a + ), with f i (x; 0) = xi (4.18)

We can prove that

xi = f i (x̂; −) (4.19)

Upon expanding x̂i = f i (x; ) to first order in  we get

x̂i = f i (x1 , x2 , ..., xn ; ), or x̂i = f i (x; ), i = 1, 2, ..., n

x = e x̂, y = e− ŷ (3.6)

x = eλ x̂, y = eµ ŷ (3.7)

x = eλ x̂, y = eµ ŷ (3.9)

x = e x̂, y = e−2 ŷ (3.12)

x̂ = f (x, y; ), ŷ = g(x, y; ) (4.1)

We expand x̂ and ŷ in a Taylor series around = 0:

x̂ = x + X(x, y) + O(2 )

x̂ = x + x2 + O(2 ), ŷ = y + xy + O(2 ) (4.8)

ln(ŷ) = − ln( + C1 ) + C =⇒ ln(ŷ) = − ln( + C1 ) + ln(C2 )

x̂i = f i (x; ) (4.17)

f i (f (x, a); ) = f i (x; a + ), with f i (x; 0) = xi (4.18)

xi = f i (x̂; −) (4.19)

Upon expanding x̂i = f i (x; ) to first order in we get

x̂i = xi + ξ i (x) (4.20)

be the infinitesimal generator of Φ . The infinitesimal generator of the pro-

G1 : (x, t, u)−→(x + , t, u) (7.35)

x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

G2 : (x, t, u)−→(x, t + , u) (7.41)

G3 : (x, t, u)−→(a + x, t + , u) (7.47)

x̂ = C1 , t̂ = C2 ·e , 1 + û = C3 ·e− (7.50)

x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

x̂ = x, t̂ = t·e , û = (u + 1)·e− − 1 (7.52)

G4 : (x, t, u)−→(x, t·e , (u + 1)·e− − 1) (7.53)

x̂ = + C1 , t̂ = C2 ·e , 1 + û = C3 ·e− (7.56)

x̂ = + x, t̂ = t·e , û = (u + 1)·e− − 1 (7.58)

G5 : (x, t, u)−→( + x, t·e , (u + 1)·e− − 1) (7.59)

x̂ = C1 , 1 + t̂ = C2 ·e , 1 + û = C3 ·e− (7.62)

x̂|=0 = x =⇒ (C1 )|=0 = x =⇒ C1 = x

x̂ = x, t̂ = (t + 1)·e − 1, û = (u + 1)·e− − 1 (7.64)

G6 : (x, t, u)−→(x, (t + 1)·e − 1, (u + 1)·e− − 1) (7.65)