Professional Documents
Culture Documents
Semester – II
Paper Code – 20MAT22C4
PARTIAL DIFFERENTIAL
EQUATIONS
ISBN :
Price :
Publisher: Maharshi Dayanand University Press
Publication Year : 2021
Paper Code : 20MAT22C4
Partial Differential Equations
M. Marks = 100
Term End Examination = 80
Assignment = 20
Time = 3 Hours
Course Outcomes
Students would be able to:
CO1 Establish a fundamental familiarity with partial differential equations and their applications.
CO2 Distinguish between linear and nonlinear partial differential equations.
CO3 Solve boundary value problems related to Laplace, heat and wave equations by various methods.
CO4 Use Green's function method to solve partial differential equations.
CO5 Find complete integrals of Non-linear first order partial differential equations.
Section-I
Method of separation of variables to solve Boundary Value Problems (B.V.P.) associated with one
dimensional Heat equation. Steady state temperature in a rectangular plate, Circular disc, Semi-infinite
plate. The Heat equation in semi-infinite and infinite regions. Solution of three dimensional Laplace
equations, Heat Equations, Wave Equations in Cartesian, cylindrical and spherical coordinates. Method
of separation of variables to solve B.V.P. associated with motion of a vibrating string. Solution of Wave
equation for semi-infinite and infinite strings. (Relevant topics from the book by O’Neil)
Section-II
Partial differential equations: Examples of PDE classification. Transport equation – Initial value problem.
Non-homogeneous equations. Laplace equation – Fundamental solution, Mean value formula, Properties
of harmonic functions, Green function.
Section-III
Heat Equation – Fundamental solution, Mean value formula, Properties of solutions, Energy methods.
Wave Equation – Solution by spherical means, Non-homogeneous equations, Energy methods.
Section-IV
Non-linear first order PDE – Complete integrals, Envelopes, Characteristics, Hamilton Jacobi equations
(Calculus of variations, Hamilton ODE, Legendre transform, Hopf-Lax formula, Weak solutions,
Uniqueness).
Books Recommended:
I.N. Sneddon, Elements of Partial Differential Equations, McGraw Hill, New York.
Peter V. O’Neil, Advanced Engineering Mathematics, ITP.
L.C. Evans, Partial Differential Equations: Second Edition (Graduate Studies in Mathematics) 2nd
Edition, American Mathematical Society, 2010.
H.F. Weinberger, A First Course in Partial Differential Equations, John Wiley & Sons, 1965. M.D.
Raisinghania, Advanced Differential equations, S. Chand & Co.
Contents
Chapter Section Title of Chapter Page No.
0 0 Notations 1-9
(xii)
B x, r y R n x y r =closed ball in R with centre x and radius r>0
n
u x X f x 0
(iv) The sign function is defined by
1 if x 0
sgn x 0 if x 0
1 if x 0
u max u, 0
u min u, 0
u u u
u u u
(v) If u : U R m
(vi) The symbol fdS denotes the integral of f over n 1 dimensional surface in Rn
(vii) The symbol fdl denotes the integral of f over the curve C in R n
C
(viii) The symbol fdx denotes the volume integral of S over V R n and x V is an arbitrary point.
V
1
(ix) Averages:
B x ,r
fdy
n r n
B x ,r
fdy
u x u y
Lip u sup
x , yU x y
x y
u x u x hei u x
(i) lt
xi h 0 h
(a) A vector of the for 1, 2 ,..., n where each i is a non-negative integer is called
a multi- index of order 1 2 ... n
(b) For given multi-index ,define
u x
D u x
x 1 ...x n
1 n
(c) If i is a non-negative integer
D i u x D u x , i
The set of all partial derivatives of order i.
1
2
2
(d) D k u x D u
k
n
(iii) u u xi xi
i 1
=Laplacian of u
=trace of Hessian Matrix.
(iv) Let x, y R i.e.x x1 , x2 ,..., xn , y y1 , y2 ,..., yn
n
Then we write
Dxu ux1 ,..., uxn
D u u
y y1 ,..., u yn
The subscript x or y denotes the variable w.r.t. differentiation is being taken
(d) Function Spaces
k}
(e) C U u : U R u is inf initly differentiable
1
p
u dx ,1 p
p
u L U
p
U
The function u : U R is Lebsegue measurable over L if u L U
u L U
ess sup u
U
(v) Du Lp U
Du Lp U
2
Similarly, D u D 2u
Lp U Lp U
sufficiently close to x0 .
f x
f o g as x x0 ,provided lt 0
x x0 g x
Notations 5
2 Inequalities
(i) Convex Function
A function f : Rn R is said to convex function if
f ( x (1 ) y ) f ( x) (1 ) f ( y )
3 Calculus
(a) Boundaries
(i) The boundary U is C k if for each point x 0 U there exists r>0 and a C k function
: Rn1 R such that U B x0 , r x B x 0 , r x x ,..., x
n 1 n 1
Also, U is analytic if is analytic.
(ii) If U is C 1 , then along U , the outward unit normal at any point x0 U is denoted by
v x 0 v1 ,..., vn .
v
(b) Gauss-Green Theorem
Let U be a bounded open subset of R n and U is C 1 . u : U Rn and also u C U then
1
u xi dx uv dS
i
i 1, 2,..., n
U U
6 Partial Differential Equations
(c) Integration by parts formula
Let u, v C U then
1
uv dx uv dS
i
xi
U U
u vdx uvxi dx uv dS
i
Or xi
U U U
u vdx uvxi dx uv dS
i
Or xi
U U U
Proof: udx ux dx
i xi
U
u
dS
U
Hence proved.
v
(ii) Du.Dvdx uvdx udS
U U U
v
uvdx u dS (integrating by parts)
U u
v u
(iii) uv vu dx u v dS
U U
v
Proof: uvdx Du.Dvdx udS
U U U
u
Similarly, vudx Du.Dvdx udS
U U U
is a smooth and n-1 dimensional surface in R n .Suppose also f : R R is smooth and summable.
n
Then
f Du dx
u r
fdS dr
Rn
Cor. Taking u x x x0
fdx fdS dr
0 B x0 , r
Rn
for each point x0 R or we can say
n
d
fdx fdS
dr B x0 ,r B x ,r
0
1
c exp 2 if x 1
x : x 1
0 if x 1
The constant c is chosen so that dx 1
Rn
Def. We define
1 x
x : for every 0 .
n
Properties:
(i) The functions are C since x are C .
1 x
(ii) dx dx
Rn
n
Rn
=1
8 Partial Differential Equations
(g) Mollification of a function
If f : U R is locally integrable
We define the mollification of f
f : f in U
x y f y dy y f x y dy (by definition)
U B 0,
Properties:
(i) f C U
1
p
f dx if 1 p
p
f Lp U : U
ess sup f if p
U
Let B x, r be a ball with centre x and radius r and B 0,1 be an arbitrary point of B x, r and z be an
arbitrary point of B 0,1 then relation between y and z is y=x+rz.
CHAPTER - 1
HEAT, WAVE AND LAPLACE EQUATIONS
Structure
1.1 Introduction
1.2 Method of separation of variables to solve B.V.P. associated with one-dimensional Heat equation
1.3 Steady state temperature in a rectangular plate, Circular disc and semi-infinite plate
1.4 Solution of Heat equation in semi-infinite and infinite regions
1.5 Solution of three dimensional Laplace, Heat and Wave equations in Cartesian, Cylindrical and
Spherical coordinates.
1.6 Method of separation of variables to solve B.V.P. associated with motion of a vibrating string
1.7 Solution of wave equation for semi-infinite and infinite strings
1.1 Introduction
In this section, the temperature distribution is studied in several cases. For finding the temperature
distribution we require to solve the Heat equation with different Boundary Value Problem (B.V.P.),
whereas to find the steady state temperature distribution we require to attempt a solution of Laplace
equation and to obtain motion of vibrating string we find a solution of Wave equation.
1.1.1 Objective
The objective of these content is to provide some important results to the reader like:
(i) Temperature distribution in a bar with ends at zero temperature, insulated ends, radiating ends
and ends at different temperature.
(ii) Steady state Temperature distribution in a finite, semi-infinite and infinite plate
(iii) Heat conduction in semi-infinite and infinite bar
(iv) Solution of Heat, Laplace and Wave equation in various cases
1.2. Method of Separation of Variables to solve B.V.P. associated with One Dimensional Heat
Equation
A parabolic equation of the type
2u 1 u
---(1)
x 2 k t
k being a dissasivity (constant) and u x, t being temperature at a point x, t of a solid at time t is known
as Heat Equation in one dimension.
10 Partial Differential Equations
We now proceed to discuss the method of separation of variables to solve B.V.P., with boundary
conditions:
u 0, t 0 and u l , t 0 ... 2
and
u
u x, 0 f x and = x ... 3
t t 0
Suppose the solution of (1) is
u x, t X x T t ... 4
where X(x) is a function of x only and T(t) is a function of t only.
Therefore, we have
u dX
= T t
x dx
2u d 2 X
T t ... 5
x 2 dx 2
and
u dT
X x
t dt
Inserting (5) into (1), we obtain
d2X 1 dT
T t 2 X x
dx k dt
Dividing both sides by u(x,t)=X(x)T(t),we have
1 d 2 X 1 dT
... 6
X dx 2 kT dt
Now, L.H.S. of (6) is independent of t and R.H.S. is independent of x, either side of (6) can be equated to
2
some constant of separation. If constant of separation is p , then
1 d2X 1 dT
2
p2 and p2
X dx kT dt
2
d X
or p2 X 0 ... 7
dx 2
dT
and p 2 kT 0 ... 8
dt
These equations have the solutions
X x c1e px c2e px and T t Ae kp t ... 9
2
d2X
2
p2 X 0 ... 12
dx
dT
kp 2T 0 ... 13
dt
Solution of (12) is
X 0 c1 0 and
X l c2 sin pl 0
pl n for n 0, n being an integer.
n
p
l
For n0, we have infinite many solutions
12 Partial Differential Equations
n x
X n x an sin ; n =1,2,. .. ...15
l
n
2
dT
k T 0
dt l
dT kn 2 2
or nT 0, where n 2
dt l
Its general solution is
T t cnent ...16
giving
n x
u x, 0 bn sin f x ... 19
n 1 l
u n x nt
and nbn sin .e
t t 0 n 1 l t 0
n x
nbn sin x ... 20
n 1 l
From (19) and (20), the constant bn can be determined easily and thus, (18) represents the solution of
Heat equation.
1.2.1 Ends of the Bar Kept at Temperature Zero
Suppose we want the temperature distribution u(x,t) in a thin, homogeneous bar of length L, given that
the initial temperature in the bar at time zero in the section at perpendicular to the x-axis is specified by
u(x,0)=f(x). The ends of the bar are maintained at temperature zero for all time. The boundary value
problem modeling this temperature distribution is
Heat, Wave and Laplace Equations 13
u 2u
a 2 2 0 x L, t 0 ... 1
t x
u 0, t u L, t 0 t 0 ... 2
u x, 0 f x 0 x L ... 3
Put u x, t X x T t ... 4
The R.H.S. of this equation is a function of t only and L.H.S. a function of x only and these variables are
independent. We could, e.g. choose any t, we like, thereby fixing the right side of the equation at a constant
value. The left side would then have to equal this constant for all x. We therefore, conclude that X '' is
X
T'
constant. But then must equal the same constant, which we will designate (The negative sign is
a 2T
a convention; we would eventually get the same solution if we used ). The constant is called the
separation constant.
Thus, we have
X '' T '
2
X aT
giving us two ordinary differential equations
X " X 0
T ' a 2T 0
Now consider the boundary conditions. First
u 0, t X 0 T t 0
X 0 0 or T t 0
If T t 0 for all t, then the temperature in the bar is always zero. This is indeed the solution if f(x)=0.
Otherwise, we must assume that T(t) is non-zero for some t and conclude that
X (0) 0
u L, t X L T t 0
Similarly,
X L 0
14 Partial Differential Equations
We now have the following problems for X and T
X '' X 0
X 0 X L 0
and T'+ a 2T 0
We will solve for X x first because we have the most information about X. The problem is a regular
Strum-Liouville Problem on [0,L]. A value for for which the problem has a non-trivial solution is called
an eigen value of this problem. For such a , any non-trivial solution for X is called an eigen function.
Case 1: =0
Thus, there is only the trivial solution for this case. We conclude that 0 is not an eigen value of problem.
Case 2: 0
X x cekx de kx
Now, X 0 c d 0 c d
Therefore, X x ce kx ce kx c e kx e kx
Next, X L c e kL e kL 0
kL
Here, e e 0, because kL>0, so c=0. Therefore, there are no nontrivial solutions of the problems if
kL
is X x c cos kx d sin kx
Now, X 0 c 0 , so X x d sin kx .
Therefore, X L d sin kL 0
n 2 2
This is a eigen function of the given problem corresponding to the eigen value k 2
L2
n 2 2
Now, return to the problem for T with , the differential equation is
L2
n 2 2 a 2T
T ' 0
L2
with general solution
n2 2 a 2t
Tn t ane L2
This function satisfies the heat equation and the boundary conditions u 0, t u L, t 0 on t 0 To
satisfy the initial condition for a given n, however, we need
n x
un x, 0 cn sin f x
L
And this is possible only if f x is a constant multiple of this sine function. Usually, to satisfy the initial
condition we must attempt a superposition of all the un ' s :
n 2 2 a 2t
n x
u x, t cn sin e
L2
n 1 L
The initial condition now requires that
n x
u x, 0 f x cn sin
n 1 L
16 Partial Differential Equations
Which we recognize as the Fourier sine expansion of f x on [0.L]. Therefore, choose the cn ' s as the
Fourier sine coefficients of f x on [0,L]:
n x
L
2
cn f sin d
L0 L
With certain conditions on f x this Fourier sine series converges to f x for 0 x L and the formal
solution of the boundary value problem is
n a t
2 n n x L2
2 2 2
L
u x, t f sin d sin e
L n1 0 L L
Example: As a special example, suppose the bar is kept at constant temperature A, except at its ends,
which are kept at temperature zero. Then,
f x A 0 x L
and
n x
L
2 2A
cn A sin dx 1 cos n
L0 L n
=
2A
n
1 1
n
The solution in this case is
n2 2 a 2t
2A n x
u x, t 1 1 sin
n
L2
e
n 1 n L
2n 1 x
2 n 1 2 a 2t
2
4A 1
2
= sin e
L
n 1 2n 1 L
We got the last summation from the preceding line by noticing that 1 1 0 if n is even, so all the
n
terms in the series vanish for n even and we need only retain the terms with n odd. This is done by replacing
n with 2n 1 , there by summing over only the odd positive integers.
Problems: Solve the following boundary value problem:
u 2 u
2
1. a 0 x L, t 0
t x 2
u 0, t u L, t 0 t 0
u x, 0 x L x 0 x L
Heat, Wave and Laplace Equations 17
u 2u
2. 4 2 0 x L, t 0
t x
u 0, t u L, t 0 t 0
u x, 0 x 2 L x 0 x L
u 2u
3. 3 2 0 x L, t 0
t x
u 0, t u L, t 0 t 0
2 x
u x, 0 L 1 cos 0 x L
L
1.2.2 Temperature in a Bar with Insulated Ends
Consider heat conduction in a bar with insulated ends, hence no energy loss across the ends. If the initial
temperature is given by f x , then the temperature function is modeled by the B.V.P.
u 2u
a2 2 0 x L, t 0
t x
u u
0, t L, t 0 t 0
x x
u x, 0 f x 0 x L
We will solve for u x, t , leaving out some details, which are the same as in the preceding problem. Set
u x, t X x T t
X " X 0 ...1
X ' 0 X ' L 0 ... 2
X x cx d
X x constant 0
Case 2: 0
X x cekx de kx
Now,
X ' 0 kc kd 0 c d k 0
X x c e kx e kx
Next,
X ' L ck e kL e kL 0
This is zero only if c=0. But this forces X x 0 , so choosing negative eigen value.
Case 3: 0
Set k 2 , with k 0 .Then,
X " k 2 X 0
with general solution
X x c cos kx d sin kx
Now, X ' 0 dk 0
n 2 2
k2 ; for n=1,2,…
L2
Corresponding to such an eigen value, the eigen function is
n x
X n x cn cos , for n=1,2,…
L
We can combine case 1 and case 3, by writing the eigen values as
n 2 2
for n=0,1,2,...
L2
and eigen functions as
n x
X n x cn cos
L
This is a constant functions, corresponding to 0 , when n=0.
The equation for T is
n 2 2 a 2T
T ' 0
L2
When n=0, this has solutions
T0 t constant =d0
If n=1,2,…, then
n2 2 a2T
Tn t dne L2
Now let
u0 x, t constant =a 0
n 2 2 a 2t
n x
and un x, t an cos e
L2
, where an cn d n
L
Each of these functions satisfies the heat equation and boundary conditions. To satisfy the initial condition,
we must usually attempt a superposition of these functions:
n 2 2 a 2t
n x
u x, t un x, t a0 an cos e
L2
n 0 n 1 L
20 Partial Differential Equations
n x L2
L L
1
u x, t f d f cos d cos e
L0 L n1 0 L L
Example: Suppose the left half of the bar is initially at temperature A and the right half at temperature
zero. Then,
L
A , 0<x<
f x 2
0 L
, xL
2
L
2
1 A
a0 =
L0 Ad
2
L
n 2 A n
2
2
and a n =
L A cos
0
L
d
n
sin
2
The solution for this temperature distribution is
n 2 2 a 2t
A 2A 1 n n x
u x, t
2 n 1 n
sin
2
cos
L
e
L2
n
is zero if n is even and equals 1 if n=2k+1. We may omit all terms of this series in
k 1
Since sin
2
which the summation index is even, and sum over only the odd positive integers. This is done by replacing
n with 2n-1 in the function being summed. Then,
A 2 A 1 2n 1 x
n 1 2 n 1 2 a 2t
2
u x, t
2
cos e
L
2 n1 2n 1 L
Heat, Wave and Laplace Equations 21
Problems:
Solve the following B.V.P.’s:
u 2u
1. 2 0 x , t 0
t x
u u
0, t , t 0 t 0
x x
u x, 0 sin x 0 x
u 2u
2. 4 2 0 x 2 , t 0
t x
u u
0, t 2 , t 0 t 0
x x
u x, 0 x 2 x 0 x 2
3. A thin homogeneous bar of length L has insulated ends initial temperature B, a positive constant. Find
the temperature distribution in the bar.
4. A thin homogeneous bar of length L has initial temperature equal to a constant B and the right end
(x=L) is insulated, while the left end is kept at a zero temperature. Find the temperature distribution in the
bar.
5. A thin homogeneous bar of thermal diffusivity 9 and length 2 cm and insulated has its left end
maintained at temperature zero, while the right end is perfectly insulated. The bar has an initial temperature
given by f x x2 for 0<x<2. Determine the temperature distribution in the bar. What is lim u x, t ?
t
u 2u
a2 2 0 x L, t 0
t x
u
u 0, t 0, L, t Au L, t t 0
x
u x, 0 f x 0 x L
The boundary condition at L assumes that heat energy radiates from this end at a rate proportional to the
temperature at that end of the bar, A is a positive constant called the transfer co-efficient.
X " X 0
T ' a 2T 0
22 Partial Differential Equations
Since,
u 0, t X 0 T t 0, then
X 0 0
as T t 0 , implies that u x, t 0 which is possible only if f x 0 . The condition at the right end of
the bar implies that
X ' L T t AX L T t
X ' L AX L 0
X " X 0
X 0 X ' L AX L 0
From the strum-Liouville theorem, we can be confident that this problem has infinitely many eigen values
1 , 2 ,..., each of which is associated with a non-trivial solution, or eigen functions, X n x . We would
like, however, to know these solutions, so we will consider cases:
Cases 1: 0 ,
X x cx d
Since, X 0 0 d , then
X x cx
But then
X ' x c AX L AcL
Then,
c 1 AL 0
But 1 AL 0 , so c=0 and we get only the trivial solution from this case. This means that 0 is not an eigen
value of this problem.
Case 2: 0 , write k 2 , with k 0 .Then,
X " k 2 X 0 , so
X x cekx de kx
Heat, Wave and Laplace Equations 23
Now, X 0 c d 0 d c.
X x c e kx e kx 2c sinh kx
X x c cos kx d sin kx
k
or tan kL
A
Let z kL . Then, this equation is
z
tan z
AL
z z2
Since k , then n n2
L L
is an eigen value of this problem for each positive integer n which is shown in Figure below,
Figure: The eigen values of the problem for a bar with radiating ends with corresponding eigen function
24 Partial Differential Equations
z x
X n x an sin n
L
The equation for T is
a 2 zn2T
T ' 2 0
L
a 2 zn2t
So Tn t dne L2
n 1 L
zm x z n x
L
sin
0
sin
L L
dx 0
This is like the orthogonality relationship used to derive co-efficient of Fourier series and can be exploited
in the same way to find the
zn x
L
f x sin L
dx
cn 0
zn x
L
sin
2
dx
0 L
With this choice of co-efficient, the solution is
Heat, Wave and Laplace Equations 25
L z
f sin n d a 2 z 2t
L zn x L2 n
u x, t 0
L
sin e
n 1 2 zn
L
sin d
0 L
Problems:
1. A thin, homogeneous bar of thermal diffusivity 4 and length 6 cm with insulated sides, has its end
maintained at temperature zero. Its right end is radiating (with transfer co-efficient 1 ) into the
2
surrounding medium, which has temperature zero. The bar has an initial temperature given by
f x x 6 x . Approximate the temperature distribution u x, t by finding the fourth partial
sum of the series representation for u x, t .
u 2u
a 2 2 (0 x L , t 0)
t x
u (0, t ) T1 , u ( L, t ) T2 (t 0)
u ( x,0) f ( x) (0 x L)
Put u( x, t ) X ( x)T (t ) into the heat equation to obtain,
X " X 0
T ' a 2T 0
Unlike the preceding example, there is nothing in this partial differential equation that prevents separation
of the variables. The difficulty encountered here is with the boundary conditions which are non-
homogeneous ( u(0, t ) and u( L, t ) may be non-zero). To see the effect of this consider,
u (0, t ) X (0)T (t ) T1
If T1 0 , we could conclude that X (0) 0 . But if T1 0 , this equation forces us to conclude that
T1
T (t ) constant .This is a condition, we cannot except to satisfy. The boundary condition at L
X (0)
possess the same problem.
We attempt to eliminate the problem by perturbing the function. Set
u( x, t ) U ( x, t ) ( x)
26 Partial Differential Equations
We want to choose ( x) to obtain a problem, we can solve.
Substitute u( x, t ) into the partial differential equation to get
u 2 u
2
a a 2 "( x)
t x 2
We obtain the heat equation for U if "( x) 0 . Integrating twice, ( x) must have the form
( x) Cx D ...(1)
Now, consider the boundary conditions, first
u (0, t ) T1 U (0, t ) (0)
This condition becomes U (0, t ) 0 if we choose ( x) so that
(0) T1 ...(2)
The condition
u ( L, t ) T2 U ( L, t ) ( L)
becomes U ( L, t ) 0 if
( L) T2 ...(3)
Now, use (2) and (3) to solve for C and D in (1),
(0) D T1
1
and ( L) CL T1 T2 C (T2 T1 )
L
Thus, choose
1
( x) (T2 T1 ) x T1
L
with this choice, the boundary value problem for U ( x, t ) is
U 2U
a2 2
t x
U (0, t ) U ( L, t ) 0
1
U ( x, 0) u ( x, 0) ( x) f ( x) (T2 T1 ) x T1
L
We have solved this problem earlier, with the solution
n a t
2 n n x L2
2 2 2
L
1
U ( x, t ) f ( ) (T2 T1 ) x T sin d
sin e
L n1 0 L L L
Once, we know this function, then
1
u( x, t ) U ( x, t ) (T2 T1 ) x T1
L
Heat, Wave and Laplace Equations 27
1.3 Steady–State Temperature in Plates
The two-dimensional Heat equation is
u 2u 2u
a 2 2 2 a 2 2u
t x y
u
The steady-state case occurs when we set 0 . In this event, the Heat equation is Laplace’s equation
t
2u 0
A Dirichlet problem consists of Laplace’s equation, to be solved for (x,y) in a region R of the plane,
together with prescribed values the solution is to assumes on the boundary of R, which is usually a
piecewise smooth curve. If we think of R as a flat plate, then we are finding the steady-state temperature
distribution throughout a plate, given the temperature at all timers on its boundary.
1.3.1 Steady-State Temperature in a Rectangular Plate
Consider a flat rectangular plate occupying the region R in the xy-plane by 0 x a, 0 y b. Suppose
the right side is kept at constant temperature T, while the other sides are kept at temperature zero. The
boundary value problem for the steady-state temperature distribution is:
2u 0 (0 x a, 0 y b)
u ( x, 0) u ( x, b) 0 (0 x a)
u (0, y ) 0 (0 y b)
u (a, y ) T (0 y b)
Put u( x, y) X ( x)Y ( y) into Laplace’s equation to get
X "Y Y " X 0
X" Y"
X Y
Since the left side depends only on x and the right side only on y, and these variables are independent,
both sides must equal the same constant.
X" Y"
(say)
X Y
Now, use the boundary condition:
u ( x, 0) X ( x)Y (0) 0 Y (0) 0
u ( x, b) X ( x)Y (b) 0 Y (b) 0
and u(0, y) X (0)Y ( y) 0 X (0) 0
X " X 0
X (0) 0
and, Y must satisfy
Y " Y 0
Y (0) Y (b) 0
This problem for Y ( y ) was solved in the article (Ends of the bar kept at temperature zero) with X ( x) in
place of Y ( y ) and L in place of b.
n 2 2
n 2
b
with corresponding eigen functions
n y
Yn ( y ) bn sin for n=1,2,...
b
The problems for X is now
n 2 2
X " X 0
b2
X (0) 0
n x n x
n x
X n ( x) c e b e b 2cn sinh
b
n x n y
un ( x, y ) an sinh sin ; where a n 2bn cn
b b
For each n and any choice of the constant an this function satisfies Laplace’s equation and the zero
boundary conditions on three sides of the plate. For the non-zero boundary condition, we must use a
superposition
n y n a
u (a, y ) T an sin sinh
n 1 b b
Heat, Wave and Laplace Equations 29
This is a Fourier sine expansion of T on [a,b]. Therefore, choose the entire co-efficient
n y
sin as the Fourier sine co-efficient:
b
n a 2 n y
b
an sinh T sin dy
b b0 b
2T b
1 1
n
,
b n
We now have
2T 1 1 1n
an
n n a
sinh
b
The solution is
As we have done before, observe that 1 1 equals 0 if n is even, and equals 2 if n is odd. We can
n
therefore omit the even indices in this summation, writing the solution as:
4T
1 (2n 1) x (2n 1) y
u ( x, y )
n 1 (2n 1) a
sinh
b
sin
b
(2n 1) sinh
b
Problems: 1. Solve for the steady-state temperature distribution in a flat plate covering the region
0 x a, 0 y b, if the temperature on the vertical sides and the bottom side are kept at zero while the
temperature on the top side is a constant K.
2.Solve for the steady-state temperature distribution is a flat plate covering the region 0 x a, 0 y b,
if the temperature on the left side is a constant T1 and that on right side a constant T2 , while the top and
bottom sides are kept at temperature zero.
[Hint: Consider two separate problems. In the first, the temperature on the left side is T1 and the other
sides are kept at temperature zero. In the second, the temperature on the right side is T2 , while the other
sides are kept at zero. The sum of solutions of these problems is the solution of the original problem.]
30 Partial Differential Equations
Remark: It is possible to treat the case where the four sides are kept at different temperature (not
necessarily constant), by considering four plates, in each of which the temperature is non-zero on only
one side of the plate. The sum of the solutions of these four problems is the solution for the original plate.
1.3.2 Steady-State Temperature in a Circular Disc
Consider a thin disk of radius R, placed in the plane so that its centre is the origin. We will find the steady-
state temperature distribution u (r , ) as a function of polar co-ordinates. The Laplace’s equation in polar
co-ordinates is
2u 1 u 1 2u
0
r 2 r r r 2 2
for 0 r R and
Assume that the temperature is known on the boundary of the disk:
u R, f for -
In order to determine a unique solution for u, we will specify two additional conditions, First we seek a
bounded solution. This is certainly a physically reasonable condition. Second we assume periodically
conditions:
u u
u r , u r , and (r , ) (r, )
These conditions account for the fact that (r , ) and (r , ) are polar co-ordinates of the same point.
Attempt a solution
u(r, ) F (r )G( )
Similarly,
u u
(r , ) F (r )G '( ) (r , ) F (r )G '( )
G '( ) G '( )
G "( ) G( ) 0
G ( ) G( ) ... 2
G '( ) G '( )
This is a periodic Strum-Liouville problem and first we solve it by considering different cases:
Case 1: 0
In this case, the equation reduces to
G "( ) 0
Case 2: 0
Let n2
Then, the differential equation (2) is
32 Partial Differential Equations
G "( ) n 2G ( ) 0
Now,
G ( ) G ( ) ce n de n ce n de n
G ( ) c e n e n c d 0 c d
Also,
G '( ) G '( ) cn(e n e n ) cn(e n e n )
2cn 0 c 0
G ( ) 0
G "( ) k 2G ( ) 0
G0 ( ) c0
Gn ( ) cn cos(n ) d n sin(n )
Now, let n2 to get (1) as
rF "(r ) rF '(r ) n 2 F (r ) 0
This is a second order Euler differential equation with general solution
Fn (r ) an r n bn r n , for n=1,2,3...
and F0 (r ) a0 constant , for n=0
The requirement that the solution must be bounded forces to choose each bn 0 because
r n as r 0 (centre of the disk).
Combining cases, we can write
Fn (r ) an r n for n=0,1,2...
For n=0,1,2…, we now have functions of the form
un (r, ) Fn (r )Gn ( ) an r n cn cos(n ) dn sin(n )
1 1
An R n
f ( ) cos n d An
Rn f ( ) cos n d
1 1
and Bn R f ( ) sin n d Bn f ( ) sin n d
n
Rn
34 Partial Differential Equations
Example: As a specific example, suppose the disk has radius 3 and that f ( ) 2 . A routine
integration gives
A0 2 , An 0 for n=1,2,3...
2
and Bn ( 1) n 1
n.3n
The solution for this condition is
n
2 r
u (r , ) 2 (1) n 1 sin(n )
n 1 n 3
for 0 r 3 and .
Problems:
1. Find the steady-state temperature for a thin disk
3. Find the steady-state temperature in the flat wedge-shaped plate occupying the region
0 r k , 0 (in polar co-ordinates). The sides 0 and are kept at temperature zero and
the ark r k for 0 is kept at temperature T.
[Hint: The BVP for this situation is
2u 1 u 1 2u
0
r 2 r r r 2 2
u (r , 0) u (r , ) 0 (0 r k )
u (k , ) T (0 )
1.3.3 Steady-State Temperature Distribution in a Semi-infinite Strip
Find the steady-state temperature distribution in a semi-infinite strip x 0, 0 y 1, pictured in figure.
The temperature on the top side and bottom side are kept at zero, while the left side is kept at temperature
T.
The boundary value problem modelling this problem is:
Heat, Wave and Laplace Equations 35
2u 2u
0 0 y 1, x 0
x 2 y 2
u ( x, 0) 0 u ( x,1) (x 0)
u(0,y)=T (0 y 1)
n n2 2
The problem for X ( x) is now
X " n2 2 X 0
X n ( x) bnen x cnen x
X n ( x) cne n x
36 Partial Differential Equations
Thus, solution for each n is
which is Fourier sine expansion of T on [0,1]. Therefore, choose the entire co-efficient of sin(n y) as the
Fourier sine co-efficient:
1
d n 2 T sin(n y )dy
0
[As in above article]
2T
1 (1) n
n
Problem:
1. Find a steady-state temperature distribution in the semi-infinite region 0 x a, y 0 if the temperature
on the bottom and left sides are at zero and the temperature on the right side is kept at constant T.
2. Find the steady-state temperature distribution in the semi-infinite region 0 x 4, y 0 if the
temperature on the vertical sides are kept at constant T and temperature on the bottom side is kept at
zero.
[Hint: Assume two semi-infinite regions, first with left end at temperature T and right end and bottom
at temperature zero, second with right end at temperature zero and left end and bottom at temperature
zero. Sum of these two solutions is the solution of the original problem.]
3. Use your intuition to guess the steady-state temperature in a thin rod of length L if the ends are perfectly
insulated and the initial temperature is f ( x) for 0 x L.
2u 2u
0 (0 x L) (t 0)
t 2 x 2
u u
(0, t ) 0 ( L, t ) (t 0)
x x
u ( x, 0) f ( x) (0 x L)
Heat, Wave and Laplace Equations 37
1.3.4 Steady-State Temperature in a Semi-infinite Plate
The B.V.P. is
2u 2u
0 (0 y b, x 0)
x 2 y 2
u ( x, 0) 0 u ( x, b) ( x 0)
u (0, y ) T (0 y b)
n y
Yn an sin for n 1, 2,3...
b
n 2 2
with the eigen value n 2 .
b
Now the problem for X is
n2 2
X '' X 0
b2
The general solution is
n x n x
X n ( x) bne cne b b
For a bounded solution in the given domain, we have to assume bn 0 . Now the solution becomes
n x
X n ( x) cne b
. Thus the solution for each n by using the superposition is
38 Partial Differential Equations
n x
n y
u( x, y) dne b
sin where d n ancn
n 1 b
Now using the condition u(0, T ) T , we have
n y
u (0, y ) T d n sin which is a Fourier sine expansion of T on [0,1]. The
n 1 b
coefficient
n y
b
d n 2 T sin dy
0 b
2T
1 1
n
n
n x
2T n y b
u( x, y) 1 1 sin
n
b
e
n 1 n
1.3.5 Steady-State Temperature in an Infinite Plate
Suppose we want the steady-state temperature distribution in a thin, flat plate extending over the right
quarter plane x 0, y 0 . Assume that the temperature on the vertical side x 0 is kept at zero, while
the bottom side y 0 is kept at a temperature f ( x) .
2u 2u
0 ( x 0, y 0)
x 2 y 2
u (0, y ) 0 ( y 0)
u ( x, 0) f ( x) (x 0)
2
u ( x, y ) f ( ) sin(k )d sin(kx)e ky dk
0
0
Example: Assume that in the above problem
4 , 0 x2
f ( x)
0 , x2
Then,
2
ck
f ( ) sin(k )d
0
2
0
4sin( k ) d
8
1 cos(2k )
k
Thus,
8 1 cos 2k
0
ky
u ( x, y ) sin(kx)e dk
k
u 2u
a2 2 ( x 0, t 0)
t x
u ( x,0) f ( x) ( x 0)
u (0, t ) 0 (t 0)
As usual, we seek a solution, which is bounded.
Set,
u ( x, t ) X ( x)T (t ) to get
X " X 0 (x 0)
T ' a 2T 0 (t 0)
Now as in previous examples, we get
40 Partial Differential Equations
u ( x, t ) d k sin(kx)e a k t , where d k ak bk
2 2
For this choice the d k ' s are the Fourier sine integral co-efficient of f ( x) ; so
2
dk
f ( ) sin(k )d
0
With this choice of the co-efficient, the function defined by (1) is a solution of the problem.
Example: Suppose
x , 0 x
f ( x)
0 , x
2 2 sin(k )
Then, d k ( ) sin(k )d 1
0
k k
The solution is
2 sin(k )
u ( x, t ) 1 sin(kx)e
k 2 2t
dk
0
k k
There are no boundary conditions, so we impose the physically realistic condition that solutions should
be bounded. As usual, we seek a solution, which is bounded.
Set,
Heat, Wave and Laplace Equations 41
u ( x, t ) X ( x)T (t ) to get
X " X 0 ( x )
T ' a 2T 0 (t 0)
Now as in previous examples, we get
that satisfy the Heat equation and are bounded on the real line over all k>0. Now, using the superposition
u ( x, t ) (ak cos(kx) bk sin(kx))e a k t dk
2 2
...(1)
0
For this choice the ak ' s and bk ' s are the Fourier sine integral co-efficient of f ( x) ; so
1
ak
f ( ) cos(k )d
and
1
ak
f ( )sin(k )d
With this choice of the co-efficient, the function defined by (1) is a solution of the problem.
1.5 Solution of Heat, Laplace and Wave Equations
1.5.1 Solution of Three-Dimensional Heat Equations in Cartesian co-ordinates
It is a partial differential equation of the form:
u 2u 2u 2u
h2 2 2 2 ...(1)
t x y z
To find its solution by the method of separation of variables, suppose that the solution of (1) is
u( x, y, z, t ) X ( x)Y ( y)Z ( z)T (t ) ...(2)
1 d 2 X 1 d 2Y 1 d 2 Z 1 dT
2
2
2
2 ...(3)
X dx Y dy Y dz h T dt
42 Partial Differential Equations
Choosing the constant of separation such that
1 d2X 2
2 1 d Y
2
2 1 d Z 1 dT
2
p1 , 2
p2 , 2
p32 and 2 p 2 ,where p 2 p12 p22 +p32
X dx Y dy Z dz h T dt
Thus, we have the following three equations
d2X
2
p12 X 0
dx
d 2Y
2
p22Y 0
dy
d 2Z
p32 Z 0
dz 2
dT
p 2 h 2T 0
dt
with the solutions
X ( x) A cos p1 x B sin p1 x
Y ( y ) C cos p2 y D sin p2 y
Y ( y ) E cos p3 z F sin p3 z
T (t ) Ge p h t Ge ( p1 p2 p3 ) h t
2 2 2 2 2 2
p1 , p2 , p3 1
u 2u 2u
h2 2 2
t x y
The solution is
u ( x, y, t ) ( A cos p1 x B sin p1 x)(C cos p2 y D sin p2 y) Ee ( p1 p2 ) h t
2 2 2
p1 , p2 1
2u 1 u 1 2u 2u 1 u
...(1)
r 2 r r r 2 2 z 2 h2 t
To solve it by the method by separation of variables, we have
u(r , , z, t ) R(r )( )Z ( z)T (t ) ...(2)
Heat, Wave and Laplace Equations 43
giving
u dR 2u d 2 R
Z ( z )T (t ) , Z ( z )T (t )
r dr r 2 dr 2
2u d 2 2u d 2Z
R(r ) 2 Z ( z )T (t ) , R (r ) 2 T (t )
2 d z 2 dz
u dT
R(r ) Z ( z )
t dt
Substituting all these values in equation (1), we get
1 d 2 R 1 dR 1 d 2 d 2 Z 1 dT
2 2 2 2
R dr r dr r d 2
dz h T dt
1 d 2 d 2
and 2 2 0
2
...(5)
d 2
d
so that
1 d 2 R 1 dR 2
2 2
2 2
R dr r dr r
d 2 R 1 dR 2 2
2 R 0 where 2 2 - 2 ...(6)
dr 2 r dr r
with solution of (3) as
T (t ) ae h t
2 2
solution of (4) as
Z z b cos z c sin z
solution of (5) as
e cos f sin
The equation (6) is Modified Bessel’s Equation and the solution is
R(r ) AJ r BJ r for fractional
and
and
n2 p
2 x 1 n1 n p 2
Yn ( x) log J n ( x)
2 p 0 p x
Thus, the solution of Heat equation is
u r , , z, t ae
, ,
h 2 2t
b cos c sin e cos z f sin z AJ ( r ) BJ r
for fractional ,
and u r , , a, t ae h t
b cos c sin e cos z f sin z AJ ( r ) BY r
2 2
, ,
for integral .
Corollary: In 2-dimesnion, the cylindrical heat Equation is
2u 1 u 1 2u 1 u
2 2
r 2
r r r 2
h t
and the solution of Heat equation is
,
,
T (t ) ae
2 2
ht
( ) beim
and
1 d d m2 r 2 d 2 R 2r dR
sin 2 r n(n 1) (say)
2 2
sin d d sin R dr
2
R dr
giving
d 2R
2 r 2 n(n 1) R 0
dR
r2 2
2r ...(4)
dr dr
1 d d m2
sin n(n 1) 2 0
d
...(5)
sin d sin
d
Here (4), being homogeneous, if we put r es and D , reduces to
ds
D( D 1) 2 D n(n 1) R 0
or ( D n)( D (n 1)) R 0
R Aens Be ( n1) s
Ar n Br n 1
Putting cos in (5), so that
d d d d 1 d d
sin
d d d d sin d d
we have
d 2 d m2
1 n(n 1) 0
d d 1 2
d 2 d m2
or (1 2 ) 2 n ( n 1) 0
d2 d 1 2
Hence, summing overall n and trying superposition, the general solution of (1) may be expressed as
46 Partial Differential Equations
Bn
u (r , , , t ) (A r )(cos )e i e h t is required solution.
2 2
n
n
n 1
n , , m r
2V 2V 2V
0 ...(1)
x 2 y 2 z 2
To solve it by the method of separation of variables, we have
V ( x, y, z) X ( x)Y ( y)Z ( z ) ...(1)
2V d 2 X 2V d 2Y 2V d 2Z
giving 2 YZ, X 2 Z and XY 2
x dx 2 y 2 dy z 2 dz
so that (1) gives
1 d2X d2X
p 1
2
p12 X 0 ...(2)
X dx 2 dx 2
1 d 2Y d 2Y
2
p2
2
2
p2 2Y 0 ...(3)
Y dy dy
1 d 2Z d 2Z
p 2
p2Z 0 where p 2 p12 p2 2 ...(4)
Z dz 2 dz 2
The solutions of these equations are
X ( x) A cos p1 x B sin p1 x
Y ( y ) C cos p2 y D sin p2 y
Z ( z ) Ee pz Fe pz
The combined solution of (1) is
2V 2V
0 ...(1)
x2 y 2
To solve it by the method of separation of variables, we have
V ( x, y) X ( x)Y ( y) ...(2)
Heat, Wave and Laplace Equations 47
2V d 2 X 2V d 2Y
giving 2 Y and X 2
x 2 dx y 2 dy
so that (1) gives
1 d2X 1 d 2Y
2
2
p2
X dx Y dy
Now,
d2X
p2 X 0 ...(3)
dx 2
d 2Y
and p 2Y 0 ...(4)
dy 2
2V 1 V 1 2V 2V
0 ...(1)
r 2 r r r 2 2 z 2
Assuming that V (r, , z) R(r )( )Z ( z) , then (1) yields
1 d 2 R 1 dR 1 d 2 1 d 2 Z
0 ...(2)
R dr 2 rR dr r 2 d 2 Z dz 2
Since the variables are separated, we can take
1 2Z 1 d 2
2
and 2
z z 2 d 2
2Z d 2
2 2 z 0 and 2 0
z d 2
yielding the general solutions as
48 Partial Differential Equations
d 2 R 1 dR 2 2
2 R 0
dr 2 r dr r
for fractional .
for integral .
Corollary: 1. Taking constant A and B , the general solution can be written as
But Y (r ) as r 0 , therefore if it is finite along the line r 0 , then B 0 , hence the solution is
2V 1 V 1 2V
0 ...(1)
r 2 r r r 2 2
To solve it by the method of separation of variables, we take
giving
Heat, Wave and Laplace Equations 49
V dR 2V d 2 R
( ) ; 2 ( ) and
r dr r 2 dr
2V d 2
R(r ) 2
2 d
Substituting all these in the equation (1), we get
1 2 d 2R dR 1 d 2
r r n 2 (say)
R dr 2 dr d 2
so that we have
1 2 d 2R dR
r 2
r n (say)
2
R dr dr
d 2R dR 2
r2 2
r n R 0 ...(3)
dr dr
which is homogeneous and hence on putting
D( D 1) D n 2 r 0
D2 n2 r 0
1 d 2 2
n ...(4)
d 2
It has the solution
( ) C cos n D sin n
Vn r , Ar n Br n C cos n D sin n ...(5)
d 2R dR
r2 2
r 0 ...(6)
dr dr
d 2
and 0 ...(7)
d 2
Having the solution of (6) and (7) as
R(r ) c1 c2 log r
d1 d 2
r 2 d 2 R 2r dR 1 d d 1 d 2
R dr 2 R dr sin d sin 2
d d 2
1 d
2
d
2
2 2 0
d 2
d 2
giving
d 2R dR
r2 2
2r n(n 1) R 0 ...(2)
dr dr
1 d d 2
sin n(n 1) 2 0
d
...(3)
sin d sin
Heat, Wave and Laplace Equations 51
D( D 1) 2 D n(n 1) R 0
or ( D n)( D (n 1)) R 0
R Ae ns Be ( n 1) s
Ar n Br n 1
Putting cos in (4), so that
d d d d 1 d d
sin
d d d d sin d d
we have
d 2 d 2
1 n(n 1) 0
d d 1 2
d 2 d 2
or (1 ) 2 2
2
n(n 1) 0
d d 1 2
2u 2 u 2u 2u
2
c 2
t 2 x y 2 z 2
2u 2u 2u 1 2u
...(1)
x 2 y 2 z 2 c 2 t 2
u
0 at x 0, x a
x
u
0 at y 0, y a
y
u
0 at z 0, z a
z
and u( x, y, z, t ) 0 at t 0
To solve the problem, we shall use the method of separation of variables and assume that
u( x, y, z, t ) X ( x)Y ( y)Z ( z)T (t )
Now proceed as in previous examples to get
n1 n n ct
un1n2n3 ( x, y, z, t ) n1n2n3 cos cos 2 cos 3 cos n12 n22 n32
a a a a
Therefore, using the superposition, the general solution is
n1 n n ct 2
u ( x, y, z, t )
n1 , n2 , n3 1
n n n cos
1 2 3
a
x cos 2 y cos 3 z cos
a a a
n1 n22 n32
2u 2u 1 2u
...(1)
x 2 y 2 c 2 t 2
And the solution is given by
n1 n ct 2
u ( x, y, t )
n1 , n2 1
n1n2 cos
a
x cos 2 y cos
a a
n1 n22
2u 1 u 1 2u 2u 1 2u
...(1)
r 2 r r r 2 2 z 2 c 2 t 2
Let the solution is u(r, , z, t ) R( x)( )Z ( z)T (t ) ...(2)
1 d 2Z d 2Z
2
s2 2 s2Z 0 ...(5)
Z dz dz
Heat, Wave and Laplace Equations 53
The equation (1) becomes
d 2u 1 du q 2
2
2 s2 p2
dr r dr r
d u 1 du 2 q 2
2
( 2 ) 2 R 0 ...(6)
dr r dr r
where 2 s 2 p 2 . Equation (6) is the modified Bessel’s equation of order q has a solution
R(r ) A3 J q ( r ) B3 J q ( r ) for fractional q
Now
u (r , , z, t ) CJ r A cos( pct ) B sin( pct ) A cos(q ) B sin(q ) A cos(sz) B sin(sz)
p ,q , s
q 1 1 2 2 3 3
2u 2 u 1 2u cot u 1 2u 1 2u
r 2 r r r 2 2 r 2 r 2 sin 2 2 c 2 t 2
Assuming that the solution of (1) is
u(r , , , t ) R(r )( )( )T (t )
1 1
1 2 n
u (r , , , t ) A e im
A e ipct
CP m
(cos ) DP m
(cos ) Er 2
J ( pr ) Fr 2
J ( pr ) 1.6
n
n
1 1
n
p ,q , s 2 2
Method of separation of variables to solve B.V.P. associated with motion of a vibrating string
1.6.1 Solution of the problem of vibrating string with zero initial velocity and with initial
displacement
Let us consider an elastic string of length L , fastened at its ends on the x-axis and assume that it vibrates
in the xy plane. Initially the string is released from the rest and we want to find out the expression for
displacement function y ( x, t ). The B.V.P. modeling the motion of string is
54 Partial Differential Equations
2 y 2 y
2
a 0 x L, t 0 ... 1
t 2 x 2
y 0, t y L, t 0 t 0 ... 2
y x, 0 f x , ( 0 x L) ... 3
y
( x, 0) 0, 0 x L ... 4
t
Here, it is assumed that f x is the initial displacement of the string before release and initial velocity is
zero. We find the solution of equation (1) by separation of variables. For this, we set y x, t X x T t
and using this, we get
X '' T ''
XT '' a 2 X '' T or 2
X aT
Since the left side of this equation is a function of x only and right hand side is a function of t only where
x and t are independent, both must equal to some constant. Let the constant of separation is . The
above equation has become
From here we conclude that X (0) 0 . This assumes that T (t ) is non-zero for some t . Otherwise T 0 is
zero for all time and we get the trivial solution, i.e., string would not move and it is possible only when
f ( x) 0 means string is not displaced.
n x n 2 2
X n ( x) An sin with eigen values 2 for n 1, 2,3...,
L L
Now the problem of T (t ) is
n 2 2
T '' T 0
L2
With the condition y( x, 0) 0 X ( x)T '(0) 0 T '(0) 0. Otherwise the solution becomes trivial.
t
Thus the general solution is
Heat, Wave and Laplace Equations 55
n at n at
Tn (t ) Cn cos Dn sin
L L
By applying the condition T '(0) 0 , we get Dn 0 . Hence for fixed n , the solution for
n at
Tn (t ) Cn cos for n 1, 2,3...
L
Now, for a fixed n , the solution of equation (1) is
n x n at
yn ( x, t ) Bn sin cos where Bn AnCn
L L
Using the superposition, we obtain
n x n at
y ( x, t ) yn ( x, t ) Bn sin cos
n 1 n 1 L L
Now, using the condition (3), we have
n x
y ( x, 0) f ( x) Bn sin
n 1 L
2 L n
Which is Fourier sine series and the value of constant coefficient Bn is
L 0
f ( )sin
L
d
Thus, we have
2 L n n x n at
y( x, t ) yn ( x, t ) f ( )sin d sin cos
n 1 n 1 L
0
L L L
Corollary: In the above problem, if f ( x) is replaced by
L
x , 0 x
f ( x) 2
L x L
, xL
2
The coefficient
2 L2 n L n
Bn sin d L ( L ) sin d
L 0 L 2 L
4L n
2 2 sin
n 2
Thus the solution becomes
56 Partial Differential Equations
4L
1 n n x n at
y ( x, t )
2 n
n 1
2
sin
2
sin cos
L L
n
0 if n is even and sin 2k 1 1
k 1
Since sin if n is odd positive integer. The solution
2 2
is
1
k 1
4L
n n x n at
y ( x, t )
n 1 2n 1
2
sin sin
2 L
2 cos
L
1.6.2 Solution of the Problem of Vibrating String with Initial Velocity and Zero Initial Displacement
The B.V.P is
2 y 2 y
2
a 0 x L, t 0 ... 1
t 2 x 2
y 0, t y L, t 0 t 0 ... 2
y x, 0 0, ( 0 x L) ... 3
y
( x, 0) g ( x), 0 x L ... 4
t
Similarly to earlier article, the solution for X is
n x n 2 2
X n ( x) An sin with eigen values 2 for n 1, 2,3...,
L L
The solution for T is
n at n at
Tn (t ) Cn cos Dn sin
L L
And applying the condition (3), we have
y( x, 0) X ( x)T (0) 0 T (0) 0
n at
This implies Cn 0 and the solution for T is Tn (t ) Dn sin
n 1 L
Therefore, the general solution is
n x n at
y ( x, t ) yn ( x, t ) Bn sin sin where An Dn Bn … (5)
n 1 n 1 L L
Now, using condition (4), we have
Heat, Wave and Laplace Equations 57
y
( x, 0) g ( x)
t
y
n a n x
( x, 0) g ( x) Bn sin
t n 1 L L
2 L n
Bn
n a 0
f ( )sin d
L
2
n n x n at
L
n a
y( x, t ) f ( )sin d sin sin
n 1 0
L L L
2 y 2 y
2
a x 0, t 0 ... 1
t 2 x 2
y 0, t 0 t 0 ... 2
y x, 0 f ( x), ( x 0) ... 3
y
( x, 0) g ( x), x 0 ... 4
t
Here, this the problem of vibrating string with initial velocity and displacement. So we will separate the
problem in two parts: (i) zero initial velocity and with displacement (ii) with initial velocity and zero
displacement.
(i) Zero initial velocity
For this case g ( x) 0 . For a bounded solution, we firstly set y( x, t ) X ( x)T (t ) .
Case 1: If 0
Then X '' 0 X ( x) Ax
Which is unbounded solution on the given domain, unless A 0 . Thus, for this case we have a trivial
solution.
X ( x) Ae px Be px
X (0) 0 A B 0 A B
X ( x) A e px e px
Now,
Heat, Wave and Laplace Equations 59
which is unbounded solution for p 0 unless A 0 . Thus X ( x) 0 , again we get a trivial solution.
X ( x) A cos px B sin px
X (0) 0 C 0
Now,
X ( x) D sin px
Also it is given y ( x, 0) f ( x) E p sin px dp f ( x)
0
2
So Ep
f ( )sin p d
0
Which is unbounded solution on the given domain, unless A 0 . Thus, for this case we have a trivial
solution.
X ( x) Ae px Be px
X (0) 0 A B 0 A B
X ( x) A e px e px
Now,
Which is unbounded solution for p 0 unless A 0 . Thus X ( x) 0 , again we get a trivial solution.
T (0) 0 paC 0 C 0
Now,
T (t ) D sin( pat )
y ( x,0)
g ( x) paE p sin px dp g ( x)
t 0
Also it is given
2
Ep
pa f sin p d
0
2 1
2 y 2 y
2
a x , t 0 ... 1
t 2 x 2
y x, 0 f ( x), ( x ) ... 2
y ( x, 0)
g ( x), x ... 3
t
Similar to previous article, we will separate the problem in two parts: (i) zero initial velocity and with
displacement (ii) with initial velocity and zero displacement.
Case (i) Zero initial velocity
For this case g ( x) 0 . For a bounded solution, we firstly set y( x, t ) X ( x)T (t ) . Using this in equation,
we get
X '' T ''
X a 2T
In equation, the left side is a function of x only while right side is function of t . So each side must be
equal to some constant, let that separation of constant is . The equation becomes
X '' X 0
T '' a 2T 0
62 Partial Differential Equations
and the condition becomes
y
( x, 0) X ( x)T '(0) 0 T '(0) 0
t
Now we will discuss the cases for different values of .
Case 1: If 0
Then X '' 0 X ( x) Ax B
Which is unbounded solution on the given domain, unless A 0 . Thus solution is X ( x) B for the eigen
value.
X ( x) Ae px Be px
Since p 0 , the first term in right hand side Ae px is unbounded in the domain [0, ) and the second
term Be px is unbounded in the region (,0) Therefore, for a bounded solution, we have to assume that
A 0 and B 0 . Therefore X ( x) 0
X ( x) A cos px B sin px
X p ( x) Ap cos px Bp sin px
Is a solution for T . On the other hand, if p 2 , p 0 , then the equation T '' a 2 p 2T 0 has the solution
T t E cos pat F sin pat
Heat, Wave and Laplace Equations 63
T '(0) 0 paF 0 F 0
But
Tp (t ) E p cos pat
Also it is given y ( x, 0) f ( x) a p cos px bp sin px cos pat dp
1
ap
f cos p dp
Where
1
bp
f sin p dp
2
So E p
f ( ) sin p d
0
y ( x, 0)
Also it is given g ( x) pa a p cos px bp s in px sin pat dp g ( x) .
t
64 Partial Differential Equations
1
g cos p d
ap
ap
1
g sin p d
ap
bp
Problems:
1. Find the solution of B.V.P
2 y 2 y
2
a 0 x L, t 0
t 2 x 2
y 0, t y L, t 0 t 0
L
x ,0 x
y x, 0
2
Lx L
xL
,
2
y x
( x, 0) x cos 0 x L
t L
2. Find the solution of B.V.P.
2 y 2 y
2
a x 0, t 0
t 2 x 2
y (0, t ) 0 t 0
x(1 x) ,0 x 1
y x, 0
0 , x 1
y
( x, 0) 0 x 0
t
Some other problems
The Heat Equation in an Infinite Cylinder
Suppose we want the temperature distribution in a solid, infinitely long, homogeneous circular cylinder
of radius R. Let the z-axis be along the axis of the cylinder. In cylindrical co-ordinates the Heat equation
is:
u 2u 1 u 1 2u 2u
a 2 2u a 2 2
t r r r r 2 2 z 2
We assume that the temperature at any point in the cylinder depends only on the time t and the distance r
u u
from the z-axis, the axis of the cylinder. This means that 0 and the Heat equation is
z
Heat, Wave and Laplace Equations 65
u 2u 1 u
a2 2
t r r r
for t 0 . This means that the outer surface is kept at temperature zero.
Now as in previous articles (left as an exercise for readers) we obtain,
z r
J 0 n a2 zn2t R
z
u (r , t ) 2 2
2 R R2
R n 1 J1 ( zn )
e
0
f ( ) J 0 n d
R
Structure
2.1 Introduction
2.2 Transport Equation
2.3 Non-Homogeneous Equations
2.4 Laplace Equation and its Fundamental Solution.
2.5 Mean-Value Formula
2.6 Properties of Harmonic Functions
2.7 Green Function
2.1 Introduction
To model the physical problems, the partial differential equations (PDEs) are the common method. PDEs
can be used to describe a wide variety of phenomena such as sound, heat, diffusion, electrostatics,
electrodynamics, fluid dynamics, elasticity, gravitation and quantum mechanics, etc. In this chapter, we
will discuss about different types of the partial differential equations, their classifications and the classical
and weak solutions, etc.
Partial Differential Equation
A partial differential equation (PDE) is differential equation that contain an unknown function and its
partial derivate with respect to two or more variables i.e., let U be an open subset of R n . An expression of
the form
Example: The equation ut u x 0 is a partial differential equation, the unknown function is u and
independent variables are x and t.
a ( x) D u f ( x) … (2)
k
for a given function a k and f Here, it is clear that the coefficients of derivate are a function of x
only. The above equation is said to be homogeneous if f=0.
For example: ut u x 0 is a transport equation which is of first order, linear and homogeneous.
(b) Semi-linear Partial Differential Equation: A Partial Differential Equation (1) is said to semi-linear
PDE if it has the form
a ( x) D u a0 D 1u ,..., Du, u, x 0, … (3)
k
(c) Quasi-linear Partial Differential Equation: A Partial Differential Equation (1) is said to quasi PDE
if it has the form
Here, coefficient of highest order derivative are lower order derivative and function of x but not same
order derivatives.
For example: u u u u 0
x xx x t
(d) Nonlinear Partial Differential Equation: A Partial Differential Equation is non-linear in the highest
order derivatives.
For example: u 2u u u 0
xx x t
68 Partial Differential Equations
2.1.2 System of Partial Differential Equations
An expression of the form is said to be system of partial differential equations if it is represented by
is given and u u1, u 2 ,..., u m is the unknown function such that u : U R m
For example:
u divu 0 where u u1, u 2 , u3
Note: The classifications of system of partial differential equations are same as in case of a partial
differential equations.
2.1.3 Solution of PDE
An expression u which satisfies the given PDE (1) is called a solution of the Partial Differential Equation.
Well posed problem: A given problem in Partial Differential Equation is well posed (Hadaward) if it
satisfies
(i) existence
(ii) uniqueness
(iii) continuously depend on the data of given problem.
Classical Solution: If a solution of a given problem satisfies the above three conditions i.e., the solution
of kth order partial differential equation exists, is unique and is at least k times differentiable, then the
solution is called classical solution. Solutions of Wave equation, Lalpace, and Heat equation etc., are
classical solutions.
Weak Solution: If a solution of a given problem exists and is unique but does not satisfy the conditions
of differentiability, such solution is called weak solution.
For Example: The gas conservation equation
ut F ux 0
models a shock wave in particular situation. So solutions exists, is unique, but not continuous. Such
solution is known as weak solution.
Note: There are several physical phenomenon in which the problem has a unique solution, but does not
satisfy the condition of differentiability. In such cases, we cannot claim that we are not able to find the
solution rather such solutions are called weak solutions
2.2 Transport Equation
Laplace Equation and its solution 69
Homogeneous Transport Equation
The simplest partial differential equation out of four important equations is the Transport equation with
constant coefficient
ut b . Du 0 … (1)
function u=u(x, t ). Here x x1 ,..., xn R denotes a typical point in space and t 0 is the time variable.
n
ut b . Du 0 in R 0,
n
….(1)
ug on R n {t 0} ….(2)
where b R and g : R R is known and Du Dxu ux ,..., ux for the gradient of u with respect to
n n
1 n
Solution:
Let ( x, t ) be any point in the R n [0, ). To solve equation (1) , we observe the L.H.S. of equation (1)
carefully, we find that it denotes the dot product of ux1 ,..., uxn , u1 with b1 ,..., bn ,1 . So L.H.S. of equation
(1) tells that the derivative of u in the direction of b,1 is zero in R n1 dimensional space. So, the line
through ( x, t ) in the direction of (b,1) is
x( s) x sb
, s R … (3)
t (s) t s
z s u x sb, t s … (4)
u is constant on the line (4) through x, t with the direction b,1 R n1 .
Hence, if we know the value of u at any point on each such line, we know its value everywhere in
Rn 0, and it is given by equation (5).
Conversely, if g C1 , then u u x, t defined by (5) is indeed a solution of given initial value problem.
and
Du=Dg
Hence ut b.Du b.Dg b.Dg 0 for (x ,t) in Rn 0, and for t=0 u x, 0 g x on Rn t 0
Remark: If g is not C 1 , then there is obviously no C 1 solution of (1). But even in this case formula (5)
certainly provides a strong and in fact the only reasonable, candidate for a solution. We may thus
informally declare u x, t g x tb x R n , t 0 to be a weak solution of given initial value problem
even should g not be C 1 . This all makes sense even if g and thus u are discontinuous. Such a notion, that
a non-smooth or even discontinuous function may sometimes solve a PDE will come up again later when
we study nonlinear transport phenomenon.
2.3 Non-homogenous Problem
Theorem: Consider the non–homogeneous initial value problem of transport equation
ut b . Du f x, t in R n 0, ...(1)
ug on R n {t 0} ...(2)
where b R n , g : R n R, f : R n [0, ) R is known and Du Dxu ux1 ,..., uxn for the gradient of
u with respect to the spatial variables x . Solve the equation for u=u(x,t) with initial condition (2).
Solution: Fix a point x, t R n1 , as discussed before, the equation of line passing through x, t in the
direction of b,1 is given by z s u x sb, t s , where s is the parameter.
Differentiating this w. r. t. s
Integrating w. r. t. s from –t to 0
Laplace Equation and its solution 71
0 0
z s ds
t
f x sb, t s ds
t
0
z 0 z t f x sb, t s ds
t
t
u x, t u x bt , 0 f x b s t , s ds ( replacing by s )
0
t
u x, t u x bt , 0 f x b s t , s ds
0
t
u x, t g x bt f x b x t , s ds xR n
, t 0
0
It is the required solution of initial value problem for non-homogeneous transport equation.
2.4 Laplace’s Equation and its Fundamental Solution
We get the Laplace’s equation in several physical phenomenon such as irrotational flow of incompressible
fluid, diffusion problem etc. Let U Rn be a open set, x R and the unknown is u : U R , u u x
then, the Laplace’s equation is defined as
u 0 … (1)
and Poisson’s equation
u f
u x v r , …(2)
72 Partial Differential Equations
1
where r x x1 ... xn and v is to be selected (if possible) so that u 0 holds.
2 2 2
We note that
r 1 2
x1 ... xn2 2 2 xi i
1 x
xi 2 r
x 0
for i=1,2,…,n.
Thus, we have
xi
u xi v ' r ,
r
x2 x2
i v 'r i
1
and u x x v " r r 3
i i r2 r
for i=1,…,n.
So
n n xi 2 n 1 xi 2 n 1
u uxi xi v ''(r ) v '(r ) 2 v " r v 'r
i 1 i 1 r i 1 r r r
a log r b n 2
v r a
n2 b n 3
r
Laplace Equation and its solution 73
where a and b are constants.
Therefore, if r >0, the solution of Laplace’s equation is
a log x b n 2
u x a
x n2
b n 3
Without loss of generality, we take b=0. To find b, we normalize the solution i.e.
u x dx 1
Rn
So, the solution is
1
2 log x n 2
u x … (3)
1
n 3
n n 2 n x
n2
1
2 log x n 2
x … (4)
1
n 3
n n 2 n x
n2
2. Shifting the origin to a new point y, the Laplace’s equation remains unchanged. So x y is harmonic
for x y . If f : R n R is harmonic, then x y f y is harmonic for each y R n and x y .
x y f y dy
n
is harmonic.
R
Since u x x y f y dy
x
Rn
u ( x) ( x y) f ( y)dy … (5)
Rn
1
2 n log( x y ) f ( y )dy n 2
R
u ( x) … (6)
1 f ( y)
dy n 3
n n 2 n Rn x y n 2
For simplicity, we assume that the function f used in Poission’s equation is twice continuously
differentiable. Now, we show that u( x) defined by equation (5) satisfies
(i) u C 2 ( R n )
(ii) u f in R n .
Consequently, the function in (6) provided us with a formula for a solution of Poission’s equation.
Proof of (i):
Define u as,
u x x y f y dy
Rn
By change of variable x y z
u ( x) ( x) f ( x z)dz ( x) f ( x y)dy
Rn Rn
By definition . ux i .is
u ( x hei ) u ( x) f ( x hei ) f ( x)
( y ) dy (*)
h R n h
where h 0 is a real number ei R n , ei (0, 0,..., 0,1, 0,..., 0) with 1 in the ith position.
u ( x) f ( x y )
( y) dy (**)
xi Rn xi
for i 1, 2,3,..., n
Similarly,
2u ( x) 2 f ( x y )
( y) dy (***)
xi x j Rn xi x j
for i, j 1, 2,..., n
As the expression on the right hand side of equation (***) is continuous in the variable x, we see that
u C 2 (Rn )
u x y f x y dy
x
Rn
Then,
u x y f x y dy
x y x f x y dy
B 0, R B 0,
n
I J …(7)
where
I y x f x y dy … (8)
B 0,
J y x f x y dy … (9)
R B 0,
n
Now,
I y x f x y dy
B 0,
y dy
B0,
c D2 f
L R n
c 2 log n 2
c
2
n 3
76 Partial Differential Equations
Now,
J y x f x y dy
R B 0,
n
y y f x y dy , x y
R n B 0, x y
Integrating by parts
f x y
J D y Dy f x y dy y ds y
Rn B 0, B 0,
J K L
We take,
f x y
L y
B 0,
ds y
Df
L R n y ds y
B 0,
c log n 2
L … (10)
c n 3
Now K D y Dy f x y dy
R n B 0,
Integrating by parts
y
K y f x y dy f x y ds y
R n B 0, B 0,
y
f x y ds y (since is harmonic)
B 0,
Laplace Equation and its solution 77
D y , ,...,
y1 y2 yn
1 1
Also
yi yi n(n 2) (n) y n 2
(n 2) 1 y 1 yi
n(n 2) (n) y n 1
yi n (n) y n 1
y
1 y
y 0
n n y n
y y
and on B 0,
y
So,
y
on B 0,
1
.D y
n n n 1
Now, we have
1
f x y ds y
n n n 1 B0,
K
K f y ds y f y as 0 … (11)
B x ,
u x f x
This completes the proof. Thus u ( x) given by (5) is the solution of Poission’s equation.
r
(ii) fdx fds dr
0 B ( x0 , r )
B ( x0 , r )
d
(iii) fdx fds
dr B ( x0 ,r ) B ( x ,r )
0
78 Partial Differential Equations
2.5 Mean-value Theorem
Theorem: Mean-value formulas for Laplace’s equation
If u is a harmonic function. Then
u x uds udy … (1)
B x , r B x ,r
OR
If u is harmonic function, prove that u equals to both the average of u over the sphere B x, r and the
average of u over the entire ball B x, r provided B x, r U .
Set r : u y ds y … (2)
B x , r
Shifting the integral to unit ball, if z is an arbitrary point of unit ball then
r : u x rz ds z
B x , r
Then
'r Du x rz .zds z
B 0,1
u
'r ds y
B x , r
1
u y dy
n (n)r n 1 B (x ,r )
r
u y dy = 0 u 0 on B( x, r )
n Bx ,r
u x n r n
1
u x udy
n r n B x ,r
B x ,r
udy … (5)
Hence proved.
Converse of Mean- value Theorem
Theorem: If u C 2 U satisfies the mean value formula
u x uds
B x , r
Proof: Suppose that u is not harmonic, so u 0 . Therefore there exists a ball B x, r U such that
u 0 within B x, r .
Theorem: Let u C 2 U C U is harmonic within U .
u x0 max u ,
U
then u is constant within U .
Assertion (i) is the maximum principle for Laplace’s equation and (ii) is the strong maximum principle.
Proof: (ii) Suppose there exist a point x0 U such that
0
u x max u M
U
… (1)
Then for 0 r dist x , U , the mean value property implies
0
M
u x
0
B ( x0, r )
udy
1
( n) r n
B ( x0 , r )
udy
M
( n) r n
B ( x0 , r )
dy
M
Equality holds only if u M within B x0 , r . So we have, u y M for all y B x0 , r . To show that
this result holds for the set U .
Consider the set
X x U u x M
We prove that X is both open and closed.
X is closed since if x is the limit point of set X, then a sequence xn in X such that xn x
Laplace Equation and its solution 81
Since u is continuous so u xn u x .
So u x M
x X
X is closed.
u x udy
B x ,r
So x B x, r X .
Hence X is open.
But U is connected. The only set which is both open and closed in U is itself U . So U X .
Hence u x M x U . So u is constant in U .
Since U is harmonic, so by mean value theorem, there exists a ball B x0 , r U such that
u x0 uds y
B x0 , r
1
M u y ds y
n n r n 1 B x0 , r
u y
Hence x0 U .
u 0 in U
u g on U
where g 0 .
u g on U
Proof: Let u and u be two solutions of given boundary value problem, then
u f in U
u g on U
and
u f in U
u g on U
Let w u u
w 0 in U
w 0 on U
w is harmonic in U and w attains maximum value on boundary which is zero. If U is connected
then w is constant. So w 0 in U
Hence u u in U .
2.6.2 Regularity
In this property, we prove that if u C 2 is harmonic, then necessarily u C . Thus harmonic functions
are automatically infinitely differentiable.
Theorem: If u C U satisfies the mean value property for each ball B x, r U , then
u C U
Set u u in U … (1)
u x * u
1 x y
u y dy … (2)
n
U
x y he
u x he
i
1
n
U
i u y dy
… (3)
Now using (2) and (3), we have
x y hei x y
i
u x he u x 1
n
u y dy
h U h
u
Since C R n . So exists.
xi
Similarly D u exists for each multi-index .
So u C U .
We now show that u u on U .
Let x U then
u x x y u y dy
U
1 x y
u y dy
n
B x,
84 Partial Differential Equations
r
1
u y ds dr
n 0 Bx ,r
(using the cor. of coarea formula)
1 r
n n r u x dr
n 1
(by Mean value formula)
n
0
n n u x r n 1
n 0 r dr
u x y
n dy
B 0,
u x y dy (by definition)
B 0,
u x
So u u in U
and so u C U for each 0 .
Note: The above property holds for each 0 . It may happen u may not be smooth or even continuous
upto U .
2.6.3 Local Estimate for Harmonic Functions
Theorem: Suppose u is harmonic in U . Then
C
(i) D u x k u L1 B x , r
0
rn k 0 … (1)
2 nk
n 1 k
(ii) C0
1
, Ck k 1,... … (2)
n n
Proof: We prove this by induction.
For k 0, 0 .
To show u x 0 r n1 n u L1 B x, r
By mean value theorem
u x u y dy for each ball B x0 , r U
0
B x ,r
0
Laplace Equation and its solution 85
1
u x0 u y dy
n r n B x0 , r
1
u x0 u 1 … (3)
n r n L B x0 ,r
C0
D0u x0 u L1 B x ,r
rn 0
2n 1 n
where C1
n
Consider
2 2
uxi u
x12 i
x ... ux
xn2 i
u 0
xi
uxi x0 uxi dx
r
B x0 ,
2
1
r
n u xi dx
n r
B x0 ,
2
2
1
r
n uvi ds (By Gauss- Green Theorem)
n r
B x0 ,
2
2
86 Partial Differential Equations
2n
=
r
r
uvi ds
B x0 ,
2
2n
u r
L B x0 , … (4)
r 2
r r
If x B x0 , then B x, B x0 , r U .
2 2
By equation (3)
2n
u x u L1 B x, r
n rn 2
2n
n rn
u L1 B x , r
0
Hence
1 2 n
u L B x , r
0 2 n r
u L1 B x , r
0
… (5)
2n 1.n
u
x
i
x
0
n rn 1
u L1 B x , r
0
C
D u x 1 u L1 B x , r
0
rn 1 0
Hence result is true for k=1.
Assume that result is true for each multiindex of order less than or equal to k-1 for all balls in U . Fix
B ( x0 , r ) U and be multiindex with k
D u x0 D u
xi
kn
D u r … (6)
r L B x0 ,
k
r
If x B x0 , then
k
k 1
B x, r B x0 , r U
k
k 1
By assumption, in the ball B x, r
k
k 1
2n 1 n k 1
D u x0
n k 1
u k 1
L1 B x , … (7)
k 1
r
n
k
r
k
From (6) and (7)
k 1
kn 2 n k 1
n 1
D u x0
n k 1
u L1 B x ,r
r k 1 0
n r
k
2 nk
n 1 k
u L1 B x ,r
n r nk 0
Since,
n
1 k
1 for all k 2
2 2 k 1
Du x0 u xi x0 u xi dx
r
B x0 ,
2
2n
=
n r n
r
uvds ( By Guass Green’s theorem)
B x0 ,
2
2n
u r
L B x0 ,
r 2
r r
If x B x0 , then B x, B x0 , r
2 2
n
1 2
u x
n r
u L1 B x ,r
0
Hence
n
2n 2 1
uxi x0 u 1
r r n L B x0 ,r
2n 1 n
u 1
r n 1 n L B x0 ,r
n2n1
u L Rn 0 as r 0
r
Hence Du 0 .
So u is constant.
Theorem: Representation Formula
of the form
u x x y f y dy c xR n
Rn
x is bounded.
Laplace Equation and its solution 89
Let u be a solution of equation (1) which is represented as
u x y f y dy
Rn
and it is bounded.
Since f C 2 R n and x is bounded for n 3 . Let u be another bounded solution of equation (1)
Define w u u
w 0
and w is bounded ( difference of two bounded functions)
By Liouville’s theorem
w constant
or u u c
u uc
This is the required result.
x y f y dy
R2
2.6.5 Analytically
Theorem: If u is harmonic in U then u is analytic in U .
Proof: Suppose that x0 be any point in U . Firstly, we show that u can be represented by a convergent
power series in some neighbourhood of x0 .
Let 1 dist x0 , U
4
1
Then M u 1 … (1)
n r n L B x0 ,2r
for each x B x0 , r , B x, r B x0 , 2r U
By estimates of derivatives
ck
D u x0 u L1 B x ,r
r nk 0
90 Partial Differential Equations
2 nk
n 1 k
2 nk
n 1 k
So D u x0 u L1 B x ,r
L B x0 , r n r nk 0
2n 1 n
M … (2)
r
By Sterling formula
1
k
k 2 1
lim
2
k 0 k !e k
Hence,
ce !
… (3)
where ! n !
D u x0
x x0
!
The sum taken over all multiindices.
Laplace Equation and its solution 91
We claim that this power series converges, provided
r
x x0
2n 2 n3e
To verify this, let us compute for each N
The remainder term is
D u x0 t x x0 x x0
RN x
N !
For some 0 t 1 , t depending on x.
N
2n1 n2e r
N
RN x cM n2 3
r 2 n e
N
1
cM
N 2n
cM
0 as N 0
2N
Series is converges.
So u x is analytic in neighbourhood of x0 .
So u is analytic in U .
Choose x, y V , x y r . Then
1
u x udz udz
B x ,2 r
n 2n r n B y ,r
1 1
= udz u y
2n B y ,r
2n
2n u x u y … (2)
2n u y u x … (3)
Since V is connected, V is compact, so V can be covered by a chain of finite number of balls Bi i 1
cu y u x
1
So, u y u x cu y x, y V
c
Then, we have
u y y x y x u y dy
V
u y
u y y x y x
V
ds y
where denoting the outer unit normal vector on V . Also x y 0 for x y .
Then
y x u y dy
V
y x u y
u y y x ds y … (3)
U B x ,
Now
u y
y x ds y Du L B x ,
y x ds y
B x ,
1
c n n n1 o( ) 0 as 0 … (4)
n2
Also
y x y
u y
B x ,
ds y u y x
ds y
B 0,
Now
1 y
D y ,y0
n n y n
y y y
= u y x ds y u y x
1
ds y
y B 0, B 0,
n n n1
94 Partial Differential Equations
1
u y ds y
n n n 1 Bx ,
= u y ds y u x as 0 … (5)
B x ,
y x u
y x ydy u y y x ds y u x
U U
y x
Thus
u y x
u x y x u y ds y y x u y dy …(6)
U
U
This identity is valid for any point x U and for any function u C 2 U and it gives the solution of
problem defined by equation (1) and (2) provided that u y , u are known on the boundary U and the
value of u in U . But u is unknown to us along the boundary. Therefore, we have to eliminate u
to find the solution. For it, we define a correction term formula y (for fixed x) given by the
x
solution of
x 0 in U
x y x on U …. (7)
x u
u y u y dy U x
x x
u y ds
U
Then we have
x x u
x u y dy U dx
u y … (8)
U
y x x y
u x y x x
y u y dy u y dy … (9)
U U
G x, y y x x y x, y U , x y … (10)
G x, y
where Dy G x, y . y is the outer normal derivative of G with respect to the variable y. Also
we observe that equation (11) is independent of u .
Hence the boundary value problem given by equation (1) and (2) can be solved in term of Green’s function
and solution is given by equation (11) is known as Representation formula for Green’s
Function.
Note: Fix x U . Then regarding G as a function of y, we may symbolically write
G x in U
G = 0 on U
Write
v z G x, z , w z G y, z z U
Then
v z 0 z x , w z 0 z y
and w v 0 on U .
w v v w
v w ds z w v ds y
B y ,
… (1)
B x ,
w
vds Dw B x , ds
B x ,
c n1 0 as 0 … (2)
v
lim wds lim x z w z ds w x
0
B x , 0
B x ,
Now we have
v w
lim
0 w v ds z w x
B x ,
Similarly,
w v
lim
0 v w ds z v y
B y ,
w x v y
G x, y G y, x
Hence proved.
Definition: If x x1 ,..., xn1 , xn R , its reflection in the plane R is the point
n n
x x1 ,..., xn1 , xn .
n
Definition: Green’s function for the half space R is
G x, y y x y x x, y R
n
, x y
u g on Rn
with the help of Green’s function.
Laplace Equation and its solution 97
Solution: Let x, y R , x y .
n
By definition, G x, y y x y
x
x y y x … (1)
Clearly 0 in R
x n
Now
1
y x , n3
n n 2 n y x
n2
y1 x1
y x
y1 n n y x
n
y1 x1
2
2 1
n n y x n y x
n2
y12 n
-----------------------------------------------------------
-----------------------------------------------------------
------------------------------------------------------------
2 1
yn xn
2
yn n n y x
2 n
Adding y x 0 on Rn y x y x
So y x y x
G x, y y x y x is well defined.
G G
x, y DG.ˆ x, y
yn
G
y x y x
yn yn yn
yn xn yn xn
n
n n y x n n y x
n
2 xn
onR , y x
n
yx
n n x y
n
g y
u x
2 xn
n n Rn x y n
ds y xR n
K x, y
2 xn 1
n n x y n
x R , y R
n
n
n
is Poisson’s kernel for R .
Definition: If x R 0 , the point is called the point dual to x with respect to B 0,1
n x
x 2
x
Definition: Green’s function for the unit ball is G x, y y x x y x
x, y B 0,1 , x y .
Example: Solve the boundary value problem
u 0 in B 0,1
u g on B 0,1
G x, y y x y
We choose x y x y x
Laplace Equation and its solution 99
is harmonic for y x .
Or x y x is harmonic for y x
So, 0 in B 0,1
x
On B 0,1 :
x x y x
But
2 2
x1 xn
y x x y1 2 ... yn n
2 2 2
x
x x
2 1 2xy
2
x y 2 2
x x
x 2 1 2 xy y 1
x y 2 xy
2 2
x y
2
So x x y x y x .
So
G x, y y x x y x is well defined.
Now on B 0,1
G G
. , being the unit normal.
y
100 Partial Differential Equations
G y G
yi y 1
y y yi
yi x xi
2
G xi yi
yi n n x y n
n n x y
n
yi x yi
2
n n x y
n
G
1 x 2
n n x y
n
Therefore we have
1 x
2
u x g y n n x y n
ds y
B 0,1
u f in U
u g on U
where U is open, bounded and U is C 1 .
Proof: Let u be another solution of given problem.
Let w u u then w 0 in U
w 0 on U
Consider
wwdx w w
U U
xi xi
dx
Integrating by parts
Dw dx 0
2
U
Laplace Equation and its solution 101
Dw 0 in U
2
Dw 0 in U
w constant in U
But w 0 in U
Hence w 0 in U
u u
Hence uniqueness of solution.
Dirichlet’s Principle: Let us demonstrate that a solution of the boundary value problem for Poisson’s
equation can be characterized as the minimize of an appropriate functional.
Thus, we define the energy functional
1
I w Dw wfdx
2
U
2
I u min I w … (1)
wA
u g on U … (2)
0 u f u w dx
U
u u w dx f u w dx
U U
Integrating by parts
0 Du.D u w dx u w Du.vds f u w dx
U U U
Du.Du fu dx 0 Du.Dw fw dx
U U
102 Partial Differential Equations
U
2
Du fu dx Du.Dw fw dx
U
U
1
2
2 1
2
2
Du fu dx Du Dw fw dx
2
(By Cauchy-Schwartz’s inequality)
1 1
2 Du fu dx Dw fw dx
2 2
So
2
I u I w
Since u A , So
I u min I w
wA
i ' 0 for 0
1
i Du Dv u v
2
f dx
U
2
U
1
2
2
2
Du 2 Dv DuDv u v f dx
i ' 0 Du.Dv vf dx
U
Integration by parts
So u f in U .
ut u f x, t … (1)
ut u 0 … (2)
ut divF … (3)
where V is arbitrary.
Theorem: (Fundamental Solution)
Find the fundamental solution of homogeneous Heat equation
ut u 0 in U [0, ) ...(1)
where U R n is open.
104 Partial Differential Equations
Proof: It can be seen from the equation (1) that first order derivate involves w.r.t. to t and second order
derivate w.r.t. the space variables x1 , x2 ,..., xn . Consequently, if u solves the equation (1), so does
u( x, 2t ) for R .
So, we seek a solution of equation (1) of the form
1 x
u x, t v …. (2)
t t
for x R , t 0. Here, , are constants to be determined and the function v : R n R must be find.
n
Now, we simplify the equation (4) by putting 1 , so that the transformed equation involves the
2
variable y only and the equation is
v y y.Dv v 0 … (5)
v y w r where r y ….(6)
where w : R R .
From equation (5) and (6), we have
y yi
v y w '( y ) w '( y ) ( y r)
i
y r
yi r 1 yi 2
and v y y w ''(r ) w '(r ) 3
i i
r yi r r
Heat Equations 105
n 1
v y w " ' r
r
Using value of v( y ) in equation (4), we get
r n 1
w " w ' w 0 … (7)
2 r
r w ' n
r w ' ' 2 0
n 1
… (8)
rnw
r n1w ' a , where a is a constant
2
Assuming lim w, w ' 0 , we conclude a 0 , so
r
1
w ' rw … (9)
2
Integrating again, we have some constant b
r2
w be 4
…. (10)
where b is the constant of integration.
Combining (2) and (10) and our choices for , , we conclude that
x
2
b
n
e solves the Heat equation (1)
t 2 4t
u x, t dx 1
Rn
2
x
b
t
n
2
e n
4t
dx 1
R
b
2 t
n
n
1
2
t
106 Partial Differential Equations
1
b
4
n
2
1
2
x
e 4t
; x Rn , t 0
x, t 4 t 2
n
0 , x R , t 0
n
(ii) Sometimes, we write ( x, t ) x , t to emphasise that the fundamental solution is radial in the
variable r.
Theorem: Solution of Initial value problem
Solve the initial value (Cauchy) problem
ut u 0 in R n (0, ) ...(1)
ug on R n t 0 ...(2)
be the fundamental solution of the equation (1). From earlier article, we note that ( x, t ) ( x, t ) solves
the Heat equation away from the singularity (0,0) and thus so does ( x, t ) ( x y, t ) for each fixed y Rn
. Consequently, the convolution
2
x y
1
u x, t e 4t
g y dy
4 t
n
2
Rn
x y g y dy
n
… (4)
R
(ii) ut x, t u x, t 0 xR n
, t 0
So u C R n 0, .
(ii) ut x y, t g y dy
t
Rn
u x y, t g y dy
Rn
y x0 , y R n .
Then if x x 0
2
u x, t g x 0 x y, t g y g x dy
0
Rn
x y, t g y g x 0 dy
B x0 ,
x y, t g y g x 0 dy
Rn B x0 ,
IJ … (5)
Now I x y, t dy
Rn
J 2 g L x y, t dy
R B x ,
n 0
108 Partial Differential Equations
2
x y
c
n e 4t
dy
2
t R B x ,
n 0
2
y x0
c dy
n e 16t
2
t R n B x0 ,
r2
c
e
n
16t
r n 1dr 0 as t 0
2
t
lim u ( x, t ) g ( x 0 )
( x , y ) ( x0 ,0)
xR n ,t 0
Thus, we have proved that equation u(x,t) given by equation (4) is the solution of the initial value
problem.
Theorem: Non-homogeneous Heat Equation
Solve the initial value problem
ut u f in R n (0, )
u0 on R n t 0
associated with the non-homogeneous Heat equation, where f C12 R n 0, and f has compact
support.
Proof:
Define u as
2
t x y
1
u x, t e
4 t s
f y, s dyds xR n
, t 0 …(1)
4 t s
n
2
0 Rn
t
x y, t s f y, s dyds … (2)
0 Rn
Then
(i) u C12 R n 0,
Heat Equations 109
(ii) ut x, t u x, t f x, t xR n
, t 0
Proof: (i) Since has a singularity at (0,0) we cannot differentiate under the integral sign. Substituting
the variable x y 0, t s 0 and again converting to original variable.
t
ut x, t y, s f x y, t s dyds
0 Rn
2u 2
t
x , t y , s f x y, t s dyds i, j 1,..., n
xi x j 0 R n xi x j
Thus, ut , Dx2u C 2 R n 0, .
(ii) Now
t
ut x, t u x, t y, s x f x y, t s dyds y, t f x y, 0 dy
0 Rn t Rn
t
y , s f x y, t s dyds
y
Rn s
y , s y f x y, t s dyds y, t f x y, 0 dy
0 Rn s Rn
I J K … (3)
Now
J ft L
D2 f
L
y, s dyds c
0 Rn
Also, we have
t
I y ( y, s ) f ( x y, t s )dyds y, f x y , t dy
R n s R n
y, t f x y, 0 dy
Rn
y, f x y, t dy K … (4)
Rn
110 Partial Differential Equations
f x, y xR n
, t 0
t
(iii) u x, t y, s f x y, t s dyds
0 Rn
t
u
L R
n f
L R y, s dyds
n
0 Rn
t
f ds
0
f t
Taking limit as t 0
lim u x, t 0 for each x Rn .
t 0
T U T U T
Heat Equations 111
include to top U t T . The parabolic boundary T comprises the bottom and vertical sides of
U 0, T , but not the top.
1
E x, t; r y, s R n1 s t and x y, t s n
r
Heat Ball
3.2.1 Theorem: Mean-Value Property for the Heat Equation
Prove that
x y
2
1
u x, t n u y, s t s 2
dyds … (1)
4r E x ,t :r
for each Heat ball E x, t; r UT . It is assumed that u C1 UT solve the homogeneous Heat equation
2
ut u 0 in Rn 0, … (2)
Proof: The formula (1) is known as mean-value formula. We find that the right hand side of (1) involves
only u y, s for times s t . It is reasonable, as the value u x, t should not depend upon future times. We
may assume upon translating the space and time coordinates that
x 0, t 0 … (3)
So we can write Heat ball as
E r E 0,0; r … (4)
112 Partial Differential Equations
and set
1 y
r u y, s 2 dyds
r Er s
2
u ry, r s s
y
2
2
dyds (by shifting the variable) … (5)
E 1
1
y2 y 2
n1 yi u yi 2 2us dyds (Again shifting to original ball)
r Er s s
A B … (6)
We introduce the useful function
2
n y
log 4 s n log r … (7)
2 4s
Then
0, on E r …(8)
Since,
y, s r n on E r …(9)
be definition of Heat ball.
Now, we use (7) to write
n
1
B 4us yi yi dyds
r n 1 Er i 1
n
1
n 1 4nus 4 usyi yi dyds … (10)
r Er i 1
1 n y 2 n
n1 4nus 4 u yi yi dyds
r Er 2s 4s 2
i 1
Heat Equations 113
1 2n n
r n 1 4nus u y yi dyds A
s i 1 i
Er
This implies
'r A B
1 2n n
r n1 4nu u yi yi dyds
E r
s i 1
n
1 2n
n 1 4nu yi yi u y yi dyds = 0
i 1 r Er
s i
Therefore, is constant.
1 y
2
Thus r lim t u 0, 0 lim n 2
dyds 4u 0, 0 …(11)
t 0 t 0 t s
Et
2 2
1 y y
tn
E t
s2
dyds
E 1
s2
dyds 4 …(12)
x y
2
1
u x, t n u y, s t s 2
dyds … (14)
4r E x ,t ; r
Hence proved.
3.3 Properties of Solution
3.3.1 Theorem: Strong Maximum Principle for the Heat Equation
(ii) Furthermore, if U is connected and there exists a point x0 , t0 UT such that
u x0 , t0 max u
UT
Then u is constant in U t0 .
u x0 , t0 M max u
UT
E x0 , t0 ; r UT
M u x0 , t0
x0 y
2
1
n u y, s dyds M … (1)
t0 s
2
4r E x0 ,t0 ;r
Since
x0 y
2
1
1 n
4r t
E x0 ,t0 ;r s
2
dyds
0
Form equation (1), it is clear that equality holds only if u is identically equal to M within E x0 , t0 ; r .
Consequently
u y, s M for all y, s E x0 , t0 ; r
Draw any line segment L in U T connecting x0 , t0 with some other point y0 , s0 UT , with s0 t0 .
Consider
r0 min s s0 u x, t M for all po int s x, y L, s t t0
Since u is continuous, the minimum is attained. Assume r0 s0 .Then
u z0 , r0 M
Now fix any point x U and any time 0 t t0 . There exists points x0 , x1 ,..., xm , x such that the line
segments in R n connecting xi 1 to xi lie in U for i 1,..., m . (This follows since the set of points in U
which can be so connected to x0 by a polygonal path is nonempty, open and relatively closed in U ).
Select times t0 t1 ... tm t . Then the line segments in R n1 connecting xi 1 , ti 1 to xi , ti i 1,..., m
Remark: 1. From a physical perspective, the maximum principle states that the temperature at any point
x inside the road at any time (0 t T ) is less than the maximum of the initial distribution or the
maximum of temperature prescribed at the ends during the time interval [0,t].
2. The strong maximum principle implies that if U is connected and u C12 U T C U T satisfies
ut u 0 in UT
u0 on U 0, T
ug on U t 0
Let g C T , f C UT . Then there exists at most one solution u C12 U T C U T of the
initial/boundary-value problem
ut u f in U T
… (1)
u g on T
ut u f in U T
… (2)
u g on T
and
u t u f in U T
… (3)
u g on T
wt w ut ut (u u) 0
116 Partial Differential Equations
w 0 on T
apply previous theorem to w u u to get the result.
3.3.3 Regularity
Theorem: Smoothness
u C UT
Proof: We write
C x, t; r y, s x y r, t r 2
s t
To denote the closed circular cylinder of radius r , height r 2 , and top centre point x, t . Fix
3 r
C ' C x0 , t0 ; r , C " C x0 , t0 ; ,
4 2
Then
v 0 on R t 0
n
… (1)
and
Now, set
t
v x, t x y, t s f y, s dyds
0 Rn
We know that
Heat Equations 117
vt v f in R n 0, t0
… (2)
v 0 on R t 0
n
If u satisfies only the hypotheses of the theorem, we derive (4) with u u replacing u , being the
u x, t K x, t , y, s u y, s dyds x, t C ''
C
where
Since 1 on C ' .
1
We see u is C within C '' C x0 , t0 ; r
2
Theorem: Local Estimate for Solutions of the Heat Equation
There exists for each pair of integers k, l=0,1,…, a constant Ck ,l such that
Ck ,l
max Dxk Dtl u u L1 C x ,t ;r
r
C x ,t ; r k 2l n 2
2
for all cylinder C x, t ; r 2 C x, t ; r U T and all solutions u of the Heat equation in U T .
118 Partial Differential Equations
Proof: Fix some point in U T . Upon shifting the coordinates, we may as well assume the point is (0,0).
1 1
Suppose first that the cylinder C 1 C 0,0;1 lies in U T . Let C C 0, 0;
2 2
Then
1
u x, t K x, t , y, s u y, s dyds x, t C 2
C 1
Ckl u L1 C1
v x, t u rx, r 2t
According to (1)
1
Dxk Dtl v x, t Ckl v L1 C 1 x, t C 2
But Dxk Dtl v x, t r 2l k Dxk Dtl u rx, r 2t
1
and v L1 C 1 n 2 u L1 C r
r
Therefore,
C
max Dxk Dtl u kl
u L1 C r
2
C r r 2l k n 2
Note: If u solves the Heat equation within U T , then for each fixed time 0 t T , the mapping
x u x, t is analytic. However the mapping t u x, t is not in general analytic.
Heat Equations 119
3.4 Energy Methods
(a) Uniqueness
Theorem: There exists at most one solution u C12 U T of
ut u f in U T
… (1)
u g on T
Proof: If u be another solution, w u u solves
Set
e t w2 x, t dx 0 t T
U
Then
e t 2 wwt dx
U
2 wwdx
U
2 Dw dx 0
2
and so
e t e 0 0 0 t T
Consequently w u u in U T .
For this, suppose u and u are both smooth solutions of the Heat equation in U T , with the same boundary
conditions on U .
ut u 0 in UT
… (1)
u g on U 0, T
ut u 0 in UT
… (2)
u g on U 0, T
for some function g.
Theorem: Suppose u , u C 2 U T solve (1) and (2). If u x, t u x, t x U then
u u within U T .
Proof: Write w u u and set
e t w2 x, t dx 0 t T
U
Then
e t 2 Dw dx … (3)
2
Also
120 Partial Differential Equations
e t 4 Dw.Dwt dx
U
4 wwt dx … (4)
U
4 w dx
2
Since w=0 on U ,
Dw dx wwdx
2
U U
1 1
2 2 2
w dx w dx
2
U U
From (3) and (4)
2
e t 4 Dw2 dx
2
U
w2 dx 4 w dx
2
U U
e t e t
Hence
e t e t e t
2
0 t T … (5)
Now if e t 0 for all 0 t T , we are done. Otherwise there exists an interval t1 , t2 0, T with
e t 0 for t1 t t2 , e t2 0 … (6)
Write
f t log e t t1 t t2 … (7)
Then
e t e t
2
f t 0
e t e t 2
If 0 1,t1 t t2 then
f 1 t1 t 1 f t1 f t
Also
e 1 t1 t e t1 e t ,
1
and so
0 e 1 t1 t2 e t1
1
e t2
0 1
This inequality implies e t 0 for all times t1 t t2 , a contradiction.
CHAPTER - 4
WAVE EQUATIONS
Structure
4.1 Wave Equation – Solution by spherical means
4.2 Non-homogeneous equations
4.3 Energy methods for Wave Equation
4.5 Wave Equation
The homogeneous Wave equation is
utt u 0 … (1)
utt u f … (2)
Remarks: 1. The Wave equation is a simplified model equation for a vibrating string (n=1). For n=2, it
is membrane and it becomes an elastic solid for n=3. u(x,t) represents the displacement in some direction
of the point x at time t 0 for different values of n.
2. From physical perspective, it is obvious that we need initial condition on the displacement and velocity
at time t=0.
Solution of Wave equation by spherical means (for n=1)
Theorem: Derive the solution of the initial value problem for one-dimensional Wave equation
utt u xx 0 in R 0, … (1)
u g , ut h on R t 0 … (2)
u utt u xx 0 … (3)
t x t x
122 Partial Differential Equations
Set
v x, t u x, t … (4)
t x
Then, equation (4) becomes
vt x, t vx x, t 0 x R, t 0 … (5)
Equation (5) becomes the transport equation with constant coefficient (b=1).
Let v x,0 a x … (6)
We know that the fundamental solution of the initial-value problem consisting of transport equation (5)
and condition (6) is
v x, t a x t , x R, t 0 … (7)
Combining equation (4) and (7), we obtain
ut x, t ux x, t a x t in R 0, … (8)
Also
u x,0 g x in R … (9)
By virtue of initial condition (2), Equations (8) and (9) constitute the non-homogeneous transport
problem. Hence its solution is
t
u x, t g x t a x s t 1 s ds
0
x t
g x t
1
2 xt
a y dy … (10) x t 2s y
ut x,0 ux 0,0
h x g ' x , x R … (11)
Substituting (11) into (10)
x t
1
u x, t g x t h y g ' y dy
2 xt
x t
1 1
g x t g x t h y dy …(12)
2 2 x t
for x R, t 0 .
This is the d’ Alembert’s formula.
Wave Equations 123
Application of d’ Alembert’s Formula
utt u xx in R 0,
u g , ut h on R t 0 ...(1)
u0 on x 0 0,
where g, h are given, with
g 0 0, h 0 0 . … (2)
Solution: Firstly, we convert the given problems on the half-line into the problem on whole of R We do
so by extending the functions u, g , h to all of R by odd reflection method as below we set.
u x, t for x 0, t 0
u x, t … (3)
u x, t for x 0, t 0
g x for x 0
g x …(4)
g x for x 0
h x for x 0
h x …(5)
h x for x 0
Now, problem (1) becomes
utt uxx in R 0,
…(6)
u g , ut h on R t 0
Hence, d’ Alembert’s formula for one-dimensional problem (6) implies
x t
1 1
u x, t g x t g x t h y dy …(7)
2 2 x t
Recalling the definition of u, g , h in equations (3)-(5), we can transform equation (7) to read for
x 0, t 0
1 1
x t
g x t h y dy
g x t
2 2 x t if x t 0
u x, t …(8)
1 g x t g t x 1
x t
if 0 x t
2 h y dy
2 x t
utt u 0 in R3 0, …(1)
ug on R t 0
3
…(2)
ut h on R3 t 0 …(3)
As we know
U x; r , t u y, t ds y …(4)
B x , r
G x; r g y ds y …(5)
B x , r
H x; r h y ds y …(6)
B x , r
G rG, H rH …(8)
U tt U rr 0 in R 0,
U G on R t 0
…(9)
U t H on R t 0
U 0 on
r 0 0,
We note that the transformation in (7) and (8) convert the three-dimensional Wave equation into the
one-dimensional Wave equation.
From equation (7)
Utt rUtt
2
r U rr U r , Laplacian for n=3
r
Wave Equations 125
rU rr 2U r
U rU r r
Ur r
U rr … (10)
U x; r , t
u x, t lim
r 0 r
G t r G t r 1 t r
2r t r
lim H y dy
r 0
2r
G ' t H t …(13)
u x, t
t g y ds y
t h y ds y …(14)
t
B x ,t
B x ,t
But
g y ds y g x tz ds z … (15)
B x ,t B 0,1
Hence
g y ds y Dg x tz .zds z
t
B x ,t B 0,1
yx
Dg y . ds y … (16)
B x ,t t
for x R , t 0 .
3
The formula (17) is called KIRCHHOFF’S formula for the solution of the initial value problem (1)-(3)
in 3D.
4.6 Non-Homogeneous Problem
Now we investigate the initial-value problem for the non-homogeneous Wave equation
utt u f in R n 0,
… (1)
u 0, ut 0 on R t 0
n
Motivated by Duhamel’s principle, which says that one can think of the inhomogeneous problem as a set
of homogeneous problems each starting afresh at a different time slice t = t0. By linearity, one can add up
(integrate) the resulting solutions through time t0 and obtain the solution for the inhomogeneous problem.
Assume that u u x, t; s to be the solution of
… (2)
u ., s 0, ut ., s f ., s on R t s
n
and set
t
u x, t u x, t ; s ds xR n
, t 0 …(3)
0
utt u f in R n 0,
… (1)
u 0, ut 0 on R t 0
n
t
u x, t u x, t ; s ds xR n
, t 0 … (2)
0
Then
(i) u C 2 R n 0,
Wave Equations 127
(ii) utt u f in R 0,
n
Hence u C 2 R n 0, .
t
utt x, t ut x, t ; t utt x, t ; s ds
0
t
f x, t utt x, t ; s ds
0
Furthermore,
t t
u x, t u x, t ; s ds utt x, t ; s ds
0 0
Thus,
utt x, t u x, t f x, t x Rn , t 0
(iii) And clearly u x,0 ut x,0 0 for x Rn . Therefore equation (2) is the solution of equation (1).
Examples: Let us work out explicitly how to solve (1) for n=1. In this case, d’ Alembert’s formula gives
x t s
1
u x, t ; s f y, s dy
2 x t s
t x t s
1
u x, t f y, s dyds
2 0 x t s
t xs
i.e.
1
u x, t f y, t s dyds x R, t 0 … (5)
2 0 xs
u x, t; s t s f y, s dS
B x ,t s
So that
t
u x, t t s f y, s dS ds
B x ,t s
0
1
t
f y, s
4
0 B x ,t s
ts
dSds
1
t
f y, t r
4
0 B x , r
r
dSdr
Therefore,
1 f y, t y x
u x, t
4 yx
dy x R , t 0
3
B x ,t
utt u f in UT
u g on T … (1)
u h onU t 0
t
wtt w 0 in UT
w 0 on T
w 0 onU t 0
t
Set “energy”
Wave Equations 129
e t
1 2
wt x, t Dw x, t dx 0 t T
2
2U
e t wt wtt Dw.Dwt dx
U
wt wtt w dx 0
U
There is no boundary term since w=0, and hence wt 0 , on U 0, T . Thus for all
Cone of dependence
For this, suppose u C 2 solves
utt u 0 in Rn 0,
C x, t 0 t t , x x
0 0
t0 t .
CHAPTER - 5
NONLINEAR PARTIAL DIFFERENTIAL EQUATIONS
Structure
5.1 Non-linear First Order PDE – Complete integrals
5.2 Envelopes
5.3 Characteristics
5.1 Definition: Let U is an open sunset of Rn , x x1 ,..., xn R and let u : U R n R . A general form
n
F Du, u, x 0 , … (1)
for p R , z R, x U .
n
Here, “p” is the name of the variable for which we substitute the gradient Du and “z” is the variable for
which we substitute u ( x) . We also assume hereafter that F is smooth, and set
D p F Fp , Fp ,..., Fp
1 2 n
Dz F Fz .
Dx Fx , Fx ,..., Fx
1 2 n
Nonlinear Partial Differential Equations 131
Remark: The PDE F Du, u, x 0 is usually accompanied by a boundary condition of the form u g
on U . Such a problem is usually called a boundary value problem. Here our main concern is to search
solution for the non-linear PDE
Complete Integral: Consider the non-linear first order PDE
F Du, u, x 0 … (1)
Suppose first that A Rn is an open set. Assume for each parameter a a1 ,..., an A , we have a C 2
solution
u u x; a … (2)
of the PDE (1) and
ua1 u x1a1 ... uxn a1
ua u x1a2 ... u xn a2
2
Da u, Dxau 2
... ... ... ...
… (3)
uan u x1an ... u xn an
(ii) rank D u , D 2 u n
a xa x U , a A
Note: Condition (ii) ensures u x; a ”depends on all the n independent parameters a1 ,..., an ”.
u x ; a, b a.x b … (5)
Example 2: The Clairaut’s equation is the PDE
x.Du f Du u … (6)
where f : R R is given.
n
A complete integral is
u x ; a a x f a x U … (7)
for a Rn .
132 Partial Differential Equations
Example 3: The Hamilton-Jacobi Equation
ut H Du 0 … (8)
u x, t ; a, b a x tH a b … (9)
where t 0 .
Remark: For simplicity, in most of what follows, we restrict to n 2 . We call the two variables x, y .
Thus, we reduce to the case
F u x , u y , u , x, y 0 … (7)
In this case, the solution u u x, y is a surface in R 3 . The normal direction to the surface at each point is
given by the vector u x , u y , 1 .
5.2 Envelope
Definition: Let u u x ; a be a C 1 function of x and U and A are open subsets of Rn. Consider the vector
equation
Dau x; a 0 x U , a A … (1)
Suppose that we can solve (1) for the parameter a as a C 1 function of x ,
a x … (2)
Thus
Dau x; x 0 x U …(3)
We can call
v x : u x; x x U …(4)
is the envelope of the function u .; a aA
Remarks: We can build new solution of nonlinear first order PDE by forming envelope and such types
of solutions are called singular integral of the given PDE.
Theorem: Construction of new solutions
Suppose for each a A as above that u u .; a solves the partial differential equation
F Du, u, x 0 …(5)
Assume further that the envelope v , defined (3) and (4) above, exists and is a C 1 function. Then v solves (5)
as well.
Nonlinear Partial Differential Equations 133
Proof: We have v x u x; x
m
vxi x uxi x; x ua j x, x xji x
j 1
uxi x; x
for i 1,..., n .
F Dv x , v x , x F Du x; x , u x; x , x 0
Note: The geometric idea is that for each x U , the graph of v is tangent to the graph of u .; a for
a x . Thus Dv Dxu .; a at x , for a x .
u 2 1 Du
2
1 … (6)
x a 1
1
u x, a 1 x a
2 2
We find that
Dau
x a 0
1 x a
1
2 2
provided a x x .
p ( s ) Du x . , z ( s ) u ( x(.))
Then p (.) solves the ODE.
.
p D F p ( s ), z ( s ), x( s ) D F p ( s ), z ( s ), x( s ) p( s ) (3)
x z
.
and z(s) solves the ODE z ( s ) Dp F p( s ), z ( s ), x(s ) . p (s ) for those s such that x( s ) U
F Du, u, x 0 in U … (1)
We suppose that F and g are smooth functions. Now we derive the method of characteristics which solves
(1) and (2) by converting PDE into appropriates system of ODE. Initially, we would like to calculate u(x)
by finding some curve lying within U, connecting x with a point x0 and along which we can calculate
u. Since equation (2) says u g on . So we know the value of u at one end x0 and we hope then to able
to find the value of u all along the curve, and also at the particular point x.
Let us suppose the curve is described parametrically by the function
z s u x s … (3)
Set
p s Du x s … (4)
pi s uxi x s i 1,..., n . … (5)
So z . gives the values of u along the curve and p . records the values of the gradient Du .
n
pi s uxi x j x s x j s …(6)
j 1
Nonlinear Partial Differential Equations 135
d
where
ds
We can also differentiate the PDE (1) with respect to x
F F F
p Du, u, x u x s x s z Du, u, x u
n
j
Du, u, x 0 … (7)
xi
x j xi xi
j 1 j
We set
. j
F …(8)
x ( s) ( p( s), z ( s), x( s)) ( j 1, 2,..., n)
p j
Assuming (8) holds, we evaluate (7) at x x s and using equations (3) and (4), we have the identity
F F F
p p s , z s , x s u x s z p s , z s , x s p s x p s , z s , x s 0 Put
n
i
xi x j this
j 1 j i
the second equality holding by (5) and (8). We summarize by rewriting equation (8)-(10) in vector notation
p ( s) Dx F p ( s), z ( s), x ( s) Dx F p ( s), z ( s), x ( s) . p ( s)
z ( s) Dp F p ( s), z ( s), x ( s) . p ( s) ...(11)
x ( s) D p F p ( s), z ( s), x ( s)
This system of 2n+1 first order ODE comprises the characteristic equation of the nonlinear first order
PDE (1).
The functions p . p1 . ,..., p n . , z . , x . x1 . ,..., x n . are called the characteristics.
Remark: The characteristics ODE are truly remarkable in that they form a closed system of equations for
x . , z . u x . and p . Du x . , whenever u is a smooth solution of the general nonlinear PDE(1).
Now we discuss some special cases for which the structure of characteristics equations is especially
simple.
136 Partial Differential Equations
(a) Article
Let us consider the PDE of the form F Du, u, x 0 to be linear and homogeneous and thus has the form
F p, z, x b x . p c x z
x s Dp F b x
b x s (From last expression)
c x s z s
Thus
x s b x s
…(2)
z s c x s z s
comprise the characteristics equations for the linear first order PDE(1).
Example 5: Solve two dimensional system
x1u x2 x2u x1 u in U
…(3)
ug on
x2 , x1 . u x1 , u x2 u 0
We get,
b x s x2 , x1 ,
c x s 1
Now
b x s b1 x , b2 x
Nonlinear Partial Differential Equations 137
x2 , x1
b1 x x2 , b2 x x1
X s b X s
and z s c X s z s
Therefore z s z s
X s x2 s , x1 s
x1 s , x2 s x2 s , x1 s
x1 s x2 s and x2 s x1 s …(4)
Now x1 s x2 s x1 s
x1 s x1 s 0
Auxiliary equation is D2 1 0
D i
c3 0
138 Partial Differential Equations
Taking s 0 in (10)
x2 0 c2
c2 0
x1 s , x2 s x2 0 at s 0
Put s 0 in (11)
x1 0 c1
Let x x1 0 c1
0
x1 s x 0 cos s
x2 s x 0 sin s
Also we have
z s z s
dz
z s
ds
Integrating w.r.t.s
log z s log z 0
z
log 0 s
z
z z 0e s
z 0 z0
Therefore z s z 0 es
Also u g on
u x s ,0 g x1 s …(13)
We know that u x s z s
Nonlinear Partial Differential Equations 139
So u x1 s , x2 s z s
u x1 0 ,0 z 0 z 0 …(14)
z s g x0 es
Thus we have
x1 s c1 cos s x0 cos s
and z s g x0 es
x1 , x2 x1 s , x2 s x 0 cos s, x 0 sin s
x1 x 0 cos s and x2 x 0 sin s
Consider,
x12 x22 x 0
We have
x2
tan s
x1
x
s tan 1 2
x1
Thus
u x s z s g x0 es
x
e
arctan 2
u x s g x x
2
1
2
2
x1
F p, z, x b x, z . p c x, z
Now Dp F b x, z
X s Dp F b X s , z s
and z s Dp F . p
b X s , z s . p s
c X s , z s
Consequently
X s b X s , z s
..(2)
z s c X s , z s
are the characteristic equations for the quasilinear first order PDE (1).
Example 6: Consider a boundary-value problem for a semilinear PDE
ux1 ux2 u 2 in U
…(3)
u g on
b 1,1 and c z 2
Consequently
x1 s x 0 s, x 2 s s
z0 g x0
z s 1 sz 0
1 sg x 0
Nonlinear Partial Differential Equations 141
i.e. x x1 x2 , s x2 .
0
Then
g x0
u x1 , x2 u x s , x s z s
1 2
1 sg x 0
g x1 x2
,1 x2 g x1 x2 0
1 x2 g x1 x2
ux1 ux2 u in U
..(1)
u x2 on
2
p1 p1 , p 2 p 2
z 2p p
1 2
x1 p 2 , x 2 p1
We integrate these equations and we find
x1 s p20 e s 1 , x 2 s x 0 p10 e s 1
z s z 0 p10 p20 e2 s 1
p1 s p10e s , p 2 s p20e s
Since u x2 on , p20 u x2 0, x 0 2 x 0 .
2
x0
Therefore, the PDE ux1 ux2 u itself implies p10 p20 z 0 x 0 , and so p1
2 0
.
2
Thus we have,
142 Partial Differential Equations
1 x0 s
x s 2 x 0
e s
1 , x 2
s
2
e 1
z s x0 e2 s
2
x0
p1 s e s , p 2 s 2 x 0e s
2
x0 s
Fix a point x1 , x2 U . Choose s and x0 so that ( x1 , x2 ) ( x1 ( s), x 2 ( s)) (2 x 0 (e s 1), (e 1))
2
and so
u x1 , x2 u x1 s , x 2 s z s x 0 e2 s
2
x 4 x2
2
1
16
Exercise:
1. Find the characteristics of the following equations:
(a) x1ux1 x2ux2 2u, u ( x1 ,1) g ( x1 )
(b) ut b.Du f in R (0, ), b R , f f ( x, t )
n n
p s Dx H p s , x s
z s Dp H p s , x s . p s H p s , x s
x s D H p s , x s
p
ut H Du 0 in R 0,
n
ug on R n t 0
ut H Du, x 0
Nonlinear Partial Differential Equations 143
are Hamilton’s ODE
x D p H p ( s ), x( s )
p D H p ( s ), x( s )
x
which arise in the classical calculus of variations and in mechanics.
5.4.1 Derivation of Hamilton’s ODE from a Variational Principle (Calculus of Variation)
Article: Suppose that L : Rn Rn R is a given smooth function, which is called Lagrangian.
We write
L L q, x L q1 ,..., qn , x1 ,..., xn
and
D L L ...L
q q q
1 n
D
x L L ...L
x x
1 n
Where q, x R
n
For any two fix points x, y R and a time t 0 and we introduce the action functional
n
t
I w . L w s , w s ds … (2)
0
where the functions w . w1 . , w2 . ,..., wn . belonging to the admissible class
A w . C 2 0, t ; R n w 0 y, w t x
Thus, a C 2 curve w . belongs to A if it starts at the point y at time 0 and reaches the point x at time t.
According to the calculus of variations, we shall find a parametric curve x . A such that
i.e., we are seeking a function x . which minimizes the functional I . among all admissible candidates
w . A .
144 Partial Differential Equations
5.4.2 Theorem: Euler-Lagrange Equations
Prove that any minimizer x . A of I solves the system of Euler-Lagrange equations
(4)
d
Dq L x s , x s Dx L x s 0 s t
ds
v 0 v t 0 … (5)
and v v1 ,..., v n
For c R , we define
Then, w . belongs to the admissible class A and x . being the minimizer of the action functional and so
I x . I w .
i c I x . cv .
i ' 0 0 … (7)
Now we integrate (8) by parts in the first term inside the integral and using (5), we have
Nonlinear Partial Differential Equations 145
d
t
n
0 Lqi x , x Lxi x , x v i ds
i 1 0 ds
This identity is valid for all smooth functions v v1 ,..., v n satisfying (5) and so
d
ds
Lqi x , x Lxi x , x 0
for 0 s t , i 1,..., n
Remark: We see that any minimizer x . A of I . solves the Euler-Lagrange system of ODE. It is also
possible that a curve x . A may solve the Euler-Lagrange equations without necessarily being a
minimizer, in this case x . is a critical point of I . . So, we can conclude that every minimizer is a critical
point but a critical point need not be a minimizer.
5.4.3 Hamilton’s ODE:
Suppose C 2 function x . is a critical point of the action functional and solves the Euler-Lagrange equations.
Set
(1) p s Dq L x s , x s 0 s t
where p . is called the generalized momentum corresponding to the position x . and velocity x . .
p Dq L q, x
H p, x p.q p, x L q p, x , x p, x R
n
2m
The Hamiltonian is thus the total energy and the Lagrangian is the difference between the kinetic and
potential energy.
146 Partial Differential Equations
5.4.4 Theorem: Derivative of Hamilton’s ODE
x s Dp H p s , x s
(3) (0 s t )
p s Dx H p s , x s
x s q p s , x s
H n
q k L q k L
k
p , x p
p , x p, x q, x
q , x
xi k 1 xi qk xi xi
L
q, x (using (2))
xi
H n
q k L q k
and k
p , x q i
p , x p
p , x p, x
q , x
xi k 1 pi qk pi
Thus
H
pi
p s , x s qi p s , x s xi s
H L L
and
xi
p s , x s
xi
q p s , x s , x s x s , x s
xi
d L
ds qi
x s , x s
pi s
Hence
n
H i H i
H p s , x s
d
p x
ds i 1 pi xi
Nonlinear Partial Differential Equations 147
n
H H H H
0
i 1 pi xi xi pi
qRn
And
H p
lim
p p
(ii)Furthermore
LH* … (5)
Proof: For each fixed q , the function p p.q L q is linear, and the mapping
Indeed, if 0 1, p. pˆ R ,
n
148 Partial Differential Equations
H p 1 pˆ sup p 1 pˆ .q L q
sup p.q L q 1 sup pˆ .q L q
q
H p 1 H pˆ
H p sup p.q L q
qRn
p p
p L q
p p
p max L
B 0,
H p
Therefore, lim inf for all 0
p p
H p L q p.q p , q R n
and
L q sup p.q H p H * q
pRn
H * q sup p.q sup p.r L r
pR n
sup infn p. q r L r … (6)
pRn rR
since q L q is convex.
L r L q s. r q r R
n
H * q infn s. q r L r L q
rR
Nonlinear Partial Differential Equations 149
5.6 Hopf-Lax Formula
Consider the initial-value problem for the Hamilton-Jacobi equation
ut H Du 0 in R n 0,
… (1)
u g on R n t 0
We know that the calculus of variations problem with Lagrangian leads to Hamilton’s ODE for the
associated Hamilton H. Hence these ODE are also the characteristic equations of the Hamilton-Jacobi
PDE, we infer there is probably a direct connection between this PDE and the calculus of variations.
t
u x, t inf L w s ds g y w 0 y , w t x … (2)
0
is
x y
u x, t min tL g y … (3)
t
where, the infimum is taken over all C1 functions. The expression on the right hand side of (3) called
Hopf-Lax formula.
w s y
s
x y 0 s t
t
x y
t
u x, t L w s ds g y tL g y
0 t
and therefore
x y
u x, t infn tL g y
yR
t
1 t 1t
L w s ds L w s ds (by Jensen’s inequality)
t 0 t0
x y
t
tL g y L w s ds g y
t 0
and consequently
x y
infn tL g y u x, t
yR
t
Hence
x y
u x, t infn tL g y
yR
t
Lemma 1: (A functional identity)
For each x Rn and 0 s t , we have
x y
u x, t minn t s L u y, s … (1)
yR
ts
In other words, to compute u ., t , we can calculate u at time s and then use u ., s as the initial condition
on the remaining time interval s, t .
yz
u y, s sL g z … (2)
s
x z s x y s y z
Now since L is convex and 1 , we have
t t t s t s
xz s x y s yz
L 1 L L
t t ts t s
Thus
xz x y yz
u x, t tL g z t s L sL g z
t ts s
x y
t s L u y, s
ts
By (2). This inequality is true for each y R . Therefore, since y u y, s is continuous, we have
n
x y
u x, t minn t s L u y, s …. (3)
yR
ts
Nonlinear Partial Differential Equations 151
Now choose w such that
xw
u x, t tL g w …(4)
t
s s
and set y : x 1 w . Then x y x w y w .
t t t s t s
Consequently
x y
t s L u y, s
ts
xw yw
t s L sL g w
t s
xw
tL g w u x, t
t
By (4). Hence
x y
minn t s L u y , s u x, t …(5)
yR
ts
Lemma 2: (Lipschitz continuity)
x y
tL g y u x, t ….(6)
t
Then
xˆ z x y
u xˆ, t u x, t inf tL g z tL g y
z
t t
g xˆ x y g y Lip g xˆ x
Hence
u xˆ, t u x, t Lip g xˆ x
u xˆ , t u x, t Lip g x xˆ … (7)
152 Partial Differential Equations
Now select x Rn , t>0. Choosing y x in (*), we discover
u x, t tL 0 g x …(8)
Furthermore,
x y
u x, t minn tL g y
yR
t
x y
g x minn Lip g x y tL
yR
t
x y
g x t max Lip g z L z z
zR
n
t
g x t max H
B 0, Lip g
u x, t g x Ct
For
C:= max L 0 , max H … (9)
B 0, Lip g
Finally select x R ,0 tˆ t . Then Lip u ., t Lip g by (7) above. Consequently Lemma 1 and
n
u x, t u x, tˆ C t tˆ
x y
u x, t minn tL g y
yR
t
ut x, t H Du x, t 0 .
Nonlinear Partial Differential Equations 153
x hq y
u x hq, t h minn hL u y, t
yR
h
hL q u x, t .
Hence
u x hq, t h u x, t
L q .
h
Let h 0 , to compute
q.Du x, t ut x, t L q .
xz xz
ut x, t .Du x, t L
t t
0
This inequality and (10) complete the proof.
154 Partial Differential Equations
Lemma 3: (Semiconcavity)
g x z 2g x g x z C z
2
(11)
u x z , t 2u x, t u x z , t C z
2
for all x, z R , t 0 .
n
Remark: We say g is semiconcave provided (11) holds. It is easy to check (11) is valid if g is C 2 and
sup D2 g . Note that g is semiconcave if and only if the mapping x C 2
g x x is concave for some
R n
2
constant C.
x y
Proof: Choose y R so that u x, t tL g y . Then putting y z and y z in the Hopf-Lax
n
t
formulas for u x z, t and u x z, t , we find
u x z, t 2u x, t u x z, t
x y x y
tL g y z 2 tL g y
t t
x y
tL g y z
t
g y z 2g y g y z
C z ,
2
by (11)
We now prove that even if g is not semi-concave, the uniform convexity of H forces u to become semi-
concave for times t>0: it is a kind of mild regularizing effect for the Hopf-Lax solution of the initial- value
problem.
Nonlinear Partial Differential Equations 155
Lemma 4: (Semi-concavity Again)
Suppose that H is uniformly convex (with constant ) and u is defined by the Hopf-Lax formula. Then
1 2
u x z, t 2u x, t u x z, t z
t
for all x, z R , t 0 .
n
p p 1 1
H 1 2 H p1 H p2 p1 p2
2
(13)
2 2 2 8
Next we claim that for the Lagrangian L, we have estimate
1 1 q q 1
L q1 L q2 L 1 2 q1 q2
2
(14)
2 2 2 8
x y
Now choose y so that u x, t tL g y . Then using the same value of y in the Hopf-Lax formulas
t
for u x z, t and u x z, t , we calculate
u x z, t 2u x, t u x z, t
x z y x y
tL g y 2 tL g y
t t
x z y
tL g y
t
1 x z y 1 x z y x y
2t L L L
2 t 2 t t
2
1 2z 1 2
2t z ,
8 t t
The next-to-last inequality following from (14).
x, t Rn 0, . Then
ut x, t H Du x, t 0
156 Partial Differential Equations
x hq y
u x hq, t h minn hL u y, t
yR
h
hL q u x, t
Hence
u x hq, t h u x, t
L q
h
Let h 0 , to compute
and therefore
ut x, t H Du x, t ut x, t max q.Du x, t L q 0
qR
n
xz
u x, t tL g z
t
Fix h>0 and set
s s
s t h, y x 1 z
t t
xz yz
Then
t s
and
xz yz
u x, t u y, s tL g z sL g z
t s
xz
t s L
t
h h
u x, t u 1 x z , t h
t t L x z
h t
Let h 0 to compute
Nonlinear Partial Differential Equations 157
xz xz
.Du x, t ut x, t L
t t
Consequently
ut x, t H Du x, t ut x, t max q.Du x, t L q
qR n
xz xz
ut x, t .Du x, t L
t t
0
Hence ut x, t H Du x, t 0
Definition: We say that a Lipschitz Continuous function u : R 0, R is a weak solution of the
n
initial-value problem
ut H Du 0 in R n 0,
(15)
u g on R n t 0
provided
(b) ut x, t H Du x, t 0 for a.e. x, t R 0,
n
1
(c) u x z , t zu x, t u x z , t c 1 z
2
t
H is convex and
H p and g : R R is Lipschitz continuous. Then there
n
Assume H is C 2 and satisfies
lim
p p
exists at most one weak solution of the initial-value problem (15).
Proof: Suppose that u and u are two weak solutions of (15) and write w : u u .
Observe now at any point y, s where both u and u are differentiable and solve our PDE, we have
wt y, s ut y, s ut y, s
158 Partial Differential Equations
H Du y, s H Du y , s
1
H rDu y, s 1 r Du y, s dr
d
0
dr
1
DH rDu y, s 1 r Du y, s dr. Du y, s Du y , s
0
: b y, s .Dw y, s
Consequently
and
1
D 2u , D 2u C 1 I
s
Write
1
b y, s : DH rDu y, s 1 r Du y, s dr … (20)
0
Hence
0 k ,l 1
1
C 1 …(22)
s
For some constant C, in view of (17) and (19). Here we note that H convex implies D2 H 0 .
R : max DH p p max Lip u …(23)
C : x, t 0 t t , x x 0 0 R t0 t
Next write
e t v x, t dx
B x0 , R t0 t
e t vt dx R vdS
B x0 , R t0 t B x0 , R t0 t
R
B x0 , R t0 t
vdS by (21)
v b .v R dS
B x0 , R t0 t
divb v b b .Dvdx
B x0 , R t0 t
1
C 1 e t b b .Dvdx
t B x0 , R t0 t
160 Partial Differential Equations
by (22). The last term on the right hand side goes to zero as 0 , for a.e. t0 0 , according to (17), (18)
and the Dominated Convergence Theorem.
Thus
1
(24) e t C 1 e t for a.e. 0 t t0
t
z Lip u Lip u
v w u u 0 at t
Hence
u u Lip u Lip u
on B x0 , R t0 r
This inequality is valid for all 0 , and so u u in B x0 , R t0 r . Therefore, in particular,
u x0 , t0 u x0 , t0 .