Professional Documents
Culture Documents
N.Kh. Ibraglmov
CONTENTS
Preface 90
Chapter I. Definitions and elementary applications 92
§1.1. One-parameter transformation groups 93
§1.2. Prolongation formulae 96
§1.3. Groups admissible by differential equations 98
§1.4. Integration and reduction of order using one-parameter groups 99
1.4.1. Integrating factor 99
1.4.2. Method of canonical variables 100
1.4.3. Invariant differentiation 101
§1.5. Defining equations 102
§1.6. Lie algebras 104
§1.7. Contact transformations 105
Chapter II. Integration of second-order equations admitting a two-dimensional 107
algebra
§2.1. Consecutive reduction of order 108
2.1.1. An instructive example 108
2.1.2. Solvable Lie algebras 109
§2.2. The method of canonical variables 111
2.2.1. Changes of variables and basis in an algebra 111
2.2.2. Canonical form of two-dimensional algebras 112
2.2.3. An integration algorithm 113
2.2.4. An example of implementation of the algorithm 114
Chapter III. Group-theoretical classification of second-order equations 115
§3.1. Equations admitting a three-dimensional algebra 115
3.1.1. Classification in the complex domain 116
3.1.2. Classification over the reals. Isomorphism and similarity 117
§3.2. The general classification result 120
§3.3. Two remarkable classes of equations 121
3.3.1. The equation 121
y" + F3(x,y)(y')3 + F2(x,y)(yf + Fl(x,y)y' + F(x,y) = 0.
Linearizability criteria
3.3.2. Equations y" + a(x)y' + b(x)y = c(x)y~3 122
90 N.Kh. Ibragimov
Preface
"The extreme importance of the work of Lie for the general development
of geometry cannot be overestimated: I am sure that in the near future it will
be even more so", wrote F. Klein in his reaction [1] to S. Lie's results on
group-theoretical foundations of geometry, which Lie submitted as his entry
for the first N.I. Lobachevskii prize. Without any doubt, this is also a fair
valuation of Lie's work in the area of differential equations.
Marius Sophus Lie was born on 17 December 1842 in the village
Nordfjordeid in Norway. He was the sixth (youngest) child in the family of
Johann Hermann Lie, a Lutheran pastor. Starting in 1857 he studied in
Christiania (now Oslo), first in a grammar school and then (1859 — 1865) in
the university. Among other events that determined the choice of his scientific
direction, of importance were: self-study in 1868 of the geometric works
of Chasles, Poncelet, and Plucker, a journey to Germany and France
(1869 — 1870) and a meeting there with Klein (which grew into a close
friendship and a long term partnership), Chasles, Jordan, and Darboux.
During 1872 — 1886 Lie worked in the university of Christiania, and from
1886 to 1898 in the university of Leipzig. Lie died on 18 February 1899
in Christiania. The life, development of ideas and work of the great
Norwegian mathematician are discussed in detail in the excellent book of
E.M. Polishchuk, "Sophus Lie" (Nauka, Leningrad 1983). This book reflects
in sufficient detail all the main aspects of Lie's scientific work, and enables the
reader to get close to Lie the man.
One of the more remarkable achievements of Lie was the discovery that the
majority of known methods of integration of ordinary differential equations,
which seemed up to that time artificial and internally disconnected, could be
derived in a unified manner using the theory of groups (see the book [2]).
Group analysis of differential equations 91
C H AP T ER I
Fig. 1. The hull of the Ricatti equation^ .y' + >>2—2/x2 = 0 is_a surface invariant under the
group of non- homogeneous stretchings χ = xe", y = ye~", y' = y'e~ 2a.
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWV
93
Fig. 2. The hull of the equation u' + u2- u- 2 = 0 obtained from the Ricatti equation
y' + y2—2/ x2 = 0 by the change of variables t = In x, u = xy.
If the conditions (1.3) hold n ot for all values of the parameter (in some
a and b, we call G a
fixed interval), but only for sufficiently small values ofzyxwvutsrqponmlkjihgfedcbaZYX
local one- parameter group. In group analysis of differential equations one deals
precisely with such local groups, which for brevity are called simply groups.
Transformations (1.1) are called point transformations (unlike, for example,
contact transformations, in which the transformed values x, y" depend also on
the derivative y', see §1.7), an d the group G is called a group of point
transformations. F rom (1.2), (1.3) we see that the inverse transformation to
(1.1) is obtained by changing the sign of the group parameter a:
(1.6)
i$. = η(φ , Ψ). Ψ α= = ο =y-
da
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXW
95
The tangent vector field (1.5) is sometimes also written as a first- order
differential operator
(1.7)
Theorems 1.1 —1.4 quoted below are well known. We omit the proofs of
the first three, while the proof of the fourth one is included as it uses the
method of passing to canonical variables we shall use in what follows.
Theorem 1.1. A function F(x, y) is an invariant if and only if it satisfies the
partial differential equation
(1.8) X F ^ H x , y ) - M ) ^
dxl ' Έ α~
if and only if
(1.12)
Theorem 1.3. Lei iAe system of equations (1.10) ac/mii α grow/» G, enrf suppose
that the tangent vector ξ(χ) of the group does not vanish on the surface Μ
defined by the equations (1.10). Then the system (1.9) can be equivalently
rewritten in the form
(1.10') Φ ,,(Α(.τ), . . . , / , - , ( *) ) «0 , fc- 1 s,
wAere 7i(x) / n _ i(x) w α basis of invariants {the collection of all functionally
independent invariants) of the group G. Equations (1.10) and (1.10') are
equivalent in the sense that they define the same surface M.
Theorem 1.4. Every one- parameter group of transformations (1.1) reduces to the
group of translations t — t + a, u = u with symbol X = 9/9/ by a suitable
change of variables
ί= ί( ζ , y), u= u( x, y).
(1.13) * = X( , , £ + *< «) - !- .
Therefore canonical variables are found from the equations
(1.14) X(O = 1, *( u ) - 0 .
dy DP Px + y'Pv + y"Pv.
(1.16) :
~dx D<f>
These are called the first and second prolongations of the infinitesimal
operator (1.7). The term prolongation formulae is frequently used to denote the
expressions for the additional coordinates:
group they admit. These tables have been constructed using Theorems 1.3
and 1.5.
(1.21) µ = - Έ 7Γ3
2 2
For this equation to be invariant, we must have b/ a — b = I/ a . Hence
b = I/ a. Therefore the equation admits a one- parameter group of stretchings
(which can be written in the form χ = xe", y = ye~°) with symbol
The resulting equation does not contain the independent variable. Therefore
u' = p(u), and the
its order can be lowered by the well- known substitutionzyxwvutsrqponmlkjihgfedcbaZYXWVUT
problem reduces to the integration of the Riccati equation
of the first- order equation (1.26), the solution of the original second- order
equation can be reduced to quadratures. Indeed, substituting known
expressions u(x, y) and o(x, y, y') into (1.27) leads to a first- order differential
equation, which admits the group G in view of the invariance of u and o, and
which can therefore be integrated in quadratures.
Let us use this second method to lower the order of the equation (1.24).
Writing down the first prolongation of the operator (1.25)
du y y* y
2
whence y "jy — do/ du + o . On substituting this expression into (1.24) we
obtain the first- order equation (1.26) in the form of the following Riccati
equation:
102 N.Kh. Ibragtmov
.28) r\ Fv = 0,
where ζι and ζ 2 are computed from the prolongation formulae (1.17') and
(1.18'). Equation (1.28) is called the defining equation for the group admitted
by the ordinary differential equation (1.19).
In what follows we shall consider differential equations written in the
explicit form
(1.29) = / ( * , y, y')·
In this case, after substituting the values of ζ 1 ( ζ 2 from (1.17'), (1.18') with y"
given by the right- hand side of (1.29), the defining equation (1.28) assumes the
form
(1.30)
H ere/ (x, y, y') is a known function (we are dealing with a given differential
equation (1.29)), while the coordinates ξ and η of the required admissible
operator (1.7) are unknown functions of x, y. Since the left- hand side of
(1.30) contains in addition to x, y also the quantity y' considered as an
independent variable, the defining equation decomposes into several independent
equations, thus becoming an overdetermined system of differential equations
for ξ, η . Solving this system, we find all the operators admitted by the
differential equation (1.29) under consideration.
Example. Let us find the operators
dy
admitted by the second order- equation
— y'%
%y
The left- hand side of this equation is a third- degree polynomial in the
variable y'. Therefore the defining equation decomposes into the following
four equations, obtained by setting the coefficients of the various powers of y'
equal to zero:
(1.32) (y'r- . | M y = 0,
(1.33) G,')·: ^ - 4
(1.34) „': 2^- ^ + (- 1- ^- 3^= 0,
(1.35) (y')·: ^ I + ± n j c + ( ^ - _ 2^ —η ) ^ = ϋ.
η,= —
ξ2= χ , η2= - 2.
This means that (1.31) admits two linearly independent operators
and that the set of all admissible operators is a two- dimensional vector space
with basis (1.36).
The Lie algebra is denoted by the same letter L , and we take its dimension to
be the dimension of the vector space L .
A general statement about properties of defining equations for second- order
ordinary differential equations is the following result due to Lie [2].
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWV
105
1.6. The set of all solutions of a defining equation (1.30) for second-
TheoremzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
order equations (1.29) forms an algebra L of dimension r ^ 8. The maximal
dimension r = 8 is attained if and only if equation (1.29) either is linear or can
be linearized by a change of variables.
Lie [6] calls the function W defined in this theorem the characteristic
function of the group of contact transformations. Construction of a group of
contact transformations admissible for an ordinary differential equation thus
reduces to the determination of the characteristic function, the function
W = W(x, y, p) of three variables x, y and ρ = y', by substitution into the
defining equation of an operator of the form
where the coordinate ζ 2 is computed from the formula (1.18'), while for ζ 3 we
have the analogous third prolongation formula
y" :W
( I f") ·: W
The system of differential equations in W thus obtained is easy to solve;
the result is
2
> (xp - 2yp) - t- C.o ( Xy-
^ d e p e n d s on ten arbitrary constants Ct, so equation (1.40) admits a ten-
parameter group of contact transformations. The Lie algebra of this group is
obtained by substituting the function W in the formula (1.39); it has the
following basis:
d
χ _. χ— XJL Y—x*JL X — u— X —
l 2 Λ x A J 5
~~diT* ~ dy ' 3— d y ' *~ dy ' ~ dx '
«*„
Fig. 3. The Huygens principle in the construction of a wave front coincides with the
statement that contact transformations χ - x+pt/ y/ l+p2, y = y- t\ \ J\ +p2, ρ = ρ
with characteristic function W = —y/ l +p2 forms a one- parameter group.
D evelopment of this idea of Lie together with the above- mentioned relation
of contact and conformal transformations has become subsequently the basis
of the solution of the problem on the H uygens principle (in the sense of
H adamard) for equations with a non- trivial conformal group (see [11] or [10],
Ch. 2, as well as the survey [12]).
CHAPTER II
λ
(2.2) ' =
Their commutator is
(2.3) [Xt. X,)=X,.
Therefore the vector space with basis (2.2) is a two- dimensional Lie
algebra L 2.
We write down the first prolongation of the operator X\ \
ι dy dy
and find its invariants, u — χ, ο = y'—y/ x. By Theorem 1.5 we have a
second- order differential invariant
y Λ
du ~ χ ~ τ* ~~ y u *
from which we have y" = do/ du + v/ u. Expressing the left- hand side of (2.1)
in the variables u, υ, we obtain the following first- order equation (1.26):
(2.4) *. ()
N ext let us find out how the operator X2 acts in the («, o)- plane. To that
end, let us prolong it in y' and pass to the variables u, 0 according to (1.13):
L = y ^^-
w
JLr^Y
+ yJL = XX.(u)
r^Y = J- +X.(u)- Jl- .
,- dy dy .. t - v du ' s ' dv
We have obtained the sretching operator Υ = υδ/δΉ, which is admitted by
equation (2.4). Therefore the operator Xx has enabled us to lower the order of
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWV
109
equationzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(2.1) and to reduce it to the form (2.4), while the operator X2 ensures
the ability to further integrate the resulting first- order equation. Construction
of the solution of the original equation is done in reverse order: first we find
the solution of equation (2.4):
Above we started the reduction of order using the first of the operators (2.2).
This was a completely random choice. Let us work out next whether anything
changes if we start with the operator X2. Its differential invariants are
u=x. v=y'/ y, y y y y
Using these, equation (2.1) assumes the form
(2.5) ~ + ν"· + ν - = O.
du u
U nlike equation (2.4), this Riccati equation does not have an obviously
integrable form. Let us see if the so far unused operator X\ provides us with
the simplifying change of variables. To that end we find how it acts in the
(w, o)- plane. We have
F rom Theorem 2.1 and the corollary of Theorem 2.2 it becomes clear why
the basis of methods of integration of second- order equations is taken to be
two- dimensional algebras (or, which amounts to the same, two- parameter
groups). It is sufficiently obvious also that if we want to integrate an
η > 2 by the method of consecutive order reduction, the
equation of orderzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
equation must admit a solvable «- dimensional algebra. F or a detailed
discussion of this question see, for example, [19], Theorems 2.60, 2.61,
and 2.64. H ere we restrict ourselves to the two- dimensional case.
d d
χ — χ - ^~
1 4
dy dx ' ' dy
d
IV X - =— X
=4- / 00.
It is also solved by two quadratures:
2/ = <
114 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
The results of Theorems 2.3 and 2.4, as well as the corresponding invariant
equations, are given in Table 6.4, while the integration algorithm is presented
as Table 6.6.
2.2.4. An example of the implementation of the algorithm.
Let us apply the group algorithm to the equation
(2.11) iT = 4 - - •
' y2 xy
1st step. Determination of the admissible algebra L 2. Solving the defining
equation (1.30), we find that
, r)=(CiX+Ci/ 2)y.
Therefore equation (2.11) admits an algebra L 2 with basis
dx dy dx 2 dy
According to Table 6.6 we can advance directly to the third step.
3rd step. Determination of the type of the algebra L 2. We have
Therefore the algebra L 2 is of type III in Table 6.4; for the correspondence to
be complete, we must change the sign of the operator X2, and then the basis
(2.12).Y —χ *— [ xii χ = χ d V ^
2
dx dy ' Ox 2 dy
has exactly the structure of type III in Table 6.4.
4th step. Determination of the integrating change of variable. H ere X\ is the
operator of the group of projective transformations of Table 6.1. We shall
use the change of variables given there (interchanging t and u):
the difference between them and the corresponding operators of type III in
Table 6.4 (the factor 1/2 in X2) are of no importance. Excluding the solutions
of (2.11) of the form
(2.14) y=Cx,
we write equation (2.11) in the form
• ±1 + J- = 0.
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWV
115
and if Cx φ 0, then
(2.16) U
= - L - | 1 i n l d t —l l + C,.
5th step. Solution in the original variables. Substituting into (2.15), (2.16) the
values (2.13) of the variables t, u, we obtain
(2.17) y
(2.18) C\ y 4 C2x 4- χ In + 6 7 = r 0.
CH AP TER III
(3.1) X l = _
f=f(x- y, y')-
Now we need to substitute into the defining equation the obtained value of/
and the coordinates ξ = χ, η = y of the second of the operators (3.1). As a
result, we obtain the equation zfx+f = 0, where ζ = x—y. The solution of
this equation has the form
(3.2)zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
ir
x— y
with an arbitrary function g(y'). Finally, the substitution into (1.30) of the
function / of (3.2) and of the coordinates ξ = χ2, η = y2 of the operator X
leads to the following equation for the function g(y')'-
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZY
117
It is both interesting and useful to give more detailed instructions for the
construction of similarity transformations. The main theorems on similarity of
(1)
groups of algebras were obtained by Lie ([13], p. 256]) and Eisenhart
([29], [30]). We shall now consider one of these in detail. First we need to
introduce the necessary notation.
L r and L r be two r- dimensional algebras of operators of the form
Let zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
1
(1.11) in the «- dimensional space of variables χ = (x , ..., x"). As we have
said already, for algebras to be^imilar they have to be isomorphic. Therefore
let us take the algebras L r and L r to be isomorphic, write them down in the
variables χ and x, respectively, and choose bases
so that the structure constants of the two algebras in these bases are the same.
We shall be interested in the case when r > n, while the ranks of the matrices
II ξαΙΙ, ΙΙξαΙΙ are the same and equal to n. Renumbering the indices, if
necessary, we shall assume that this is the value of the rank of the square
matrices || %f, \ \ , ||ξ£ |[ (i, h = 1, ..., ri). Then the following conditions hold:
(3- 4) Ι Ζ - φ , ν . Ι»'- Φ*V (»- 1 "; / "- «+ 1 r) ,
where (pph, φ ρΗ are functions of χ and x, respectively, and where we sum over
h from 1 to n. Under the assumptions made, we have the following theorem.
Theorem 3.1. For algebras L r and L r with the same structure and the same
number of variables x', x' to be similar, it is necessary and sufficientthat the
functions <pph, <pph satisfy the system of equations
(3.5) ψ / ( i )=<(>/(• *) (Λ= 1 //; p= w+ l, . . . . r)
and that this system does not lead to any relations either amongthe variables x'
l
or amongthe variables x .
Sketch of the proof Similarity of the algebras under consideration means that
there is a non- degenerate change of variables x' - f\ x), such that
(1)
Eisenhart used the term "equivalent groups". We have used "similar groups" (and
"similar algebras") in view of what is said in the first paragraph of §3.1. Two different
words are used in the Russian text for "similar" and "equivalent". (Ed.)
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWVU
119
where || ξ*Α || is the square matrix inverse to || ξ Α ' ||, that is,
U sing the conditions (3.5), it can be shown that the system (3.6'), and thus
the system (3.6), is compatible (a detailed treatment of this non- trivial step in
the argument can be found in [30], §22). This means that there is a change of
variables that establishes the similarity of the algebras L r and L T .
Determination of the change of variables. The similarity transformation is
determined by solving the mixed system of equations comprising functional
equations (3.5) and differential equations (3.6') (which are integrable in view
of (3.5)) with respect to x'.
Example 1. Let L 3 and L 3 be the first and the second algebras of Table 6.8.
They are complex isomorphic: their structures are the same and are of
type IX in Table 6.7 if for L 3 we use the basis
Thus, let us consider three- dimensional algebras L 3 and £ 3 in the plane with
the same structure, having, respectively, the bases
Jt1==(l +
* ) * + «„ " χ, = , * _ „ *
ox dy - dy * dx
x + { i + t )
and
Hence we have
With these functions φ Λ and φ Α , the system (3.5) can be written in the form
£y _ y x- j- y __ 1
. 1 - fzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
1 — ί^ ζ * 1 — £g χ
and we have the change of variable
l £
,._ —V „- ,±±IL·
which establishes the similarity of the algebras L 3 and L3 in the complex
plane. In this example, in order to find the change of variable, it was enough
to solve equations (3.5).
Example 2. Let us consider the algebras L 3 and L 3 with bases
Xl
~'W i'~dy~ > χ
*- ^ χ- δχ"τι'Ί Ϊ' " ~ " ~dx~~ r!f"dy~
Λ 3
and
They have the same structure (see Table 6.8). Equations (3.4), written as in
the previous example, have the form
Hence
φ ' = — xy, <ft
With these functions φ Α and cpA, the equations (3.5) assume the form
x- =xy, 2x—x+y
while after eliminating χ they lead to the relation
By Theorem 3.1 the existence of this relation between χ and y shows that
these algebras L3 and £,3 are not similar (neither over the reals, nor over the
field of complex numbers), though they are isomorphic.
l 2>
dx " * dy '
dz . „„ 1 dF. . 2 dFt
(4) equation (3.7) admits a twodimensional Lie algebra with a basis X\,
such that
(3.10)
122 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
(3.13) yy ' ^
A0Aa 2 A» „ . A. . Α-
Α.,, Aa A
w-| 2- 3
ar* 3 x* " xx x2 x2
x
F rom the condition zxy = zyx we have A2(2—Ai) + 9AoA3 = 0. Similarly,
from the condition wxy = wyx we obtain 3A3(l + A\ )—A\ = 0. Writing
A3 = —a, A2 = —b, we have then
Therefore we have the following assertion (see [25] and the references
contained therein).
Equation (3.13) is linearizable if and only if it has the form
of Table 6.9 let us replace y" by a general second- order differential expression
with variable coefficients. Dividing by the coefficient of the highest- order
derivative, we obtain the equation
(3.16) y"+a(x)y'+b(x)y=c(x)y- \
Its particular case,
(3.17) y'+bW y^ky- *, fc= const,
has long been discussed in the literature [32] —[35] as an example of a non-
linear equation with exceptional properties. It turns out, first of all, that the
general solution of (3.17) can be expressed in terms of solutions of the
corresponding linear equation
(3.17') y"+b(x)y- Q.
N amely, the solution of equation (3.17) with arbitrary initial data
y(xo) = }Ό Φ 0, y'(x0) = yi has the form
CHAPTER IV
(4.1) - ^ = zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
F* (i, x\ · ·. , x"), < = 1 n,
(4.6) nm&tr.
§4.2. Examples
Example 1. F or a linear homogeneous equation
dx/ dt=A{t)x
(4.8) 1 l * X2 * *3 « *
X
dx ' dx ' dx '
which generate a Lie algebra L3. F ormula (4.6) means that m ^ 3, so to
represent the general solution of the Riccati equation with arbitrary
coeffcieints we need at least three particular solutions. I n reality it suffices to
know three solutions, as any four solutions of the Riccati equation are related
by the condition that their cross- ratio be constant. In §4.3 we shall return to
a discussion of th e properties of the Riccati equation from the group-
theoretical point of view.
£ .= (*, 0). | 2 = ( j/ , 0 ) , |, =
Therefore in this case the operators (4.5) are
Thus, the system (4.9) admits two representations of the general solution: as a
linear superpositon of three solutions (Example 5), and as a "non- linear
superposition" of two solutions if the equalities (4.10) are solved for x, y.
so the general solution of the Riccati equation (4.7) is given by the formula
c u
(4.11) j —- i i- rC2ut __ ut + Kut
128 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
Let zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
X\ , ..., Χ4 be four solutions of the Riccati equation corresponding to some
particular values K\ , ..., AT4 of the constant Κ in (4.11). Then we have the
equality
x
,/ A0, *ι — *Ϊ . i — x* A\ — K 2 Κ ι— Κ4
4 1
ί · ^) ——— · r
__ , — κ —κ ' κ — κ '
which proves that the cross- ratio of any four solutions of the Riccati equation
is constant.
C H AP TER V
Δ2 sin 0 - ι
~ sine ae """ *ίηΗ)Ί ψ'
A solution that is invariant with respect to spatial rotations (spherically
symmetric) has the form « = R(r). F or it Laplace's equation leads to
(r2R')' — 0, whence we have
R= - ^L + C2.
(5.1)zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
F or the group of stretchings in u with symbol X = «9/ 6M we have as
invariants any function of x, y, z. Therefore no solution u = φ (χ, y, z) can
be invariant with respect to this group, as u is n ot an invariant.
Let us find invariant solutions for the group of stretchings in the
independent variables and the function u:
(5.2) x—ax, y=ay, z—az, u=a"u.
A function u = φ (χ, y, z) that is invariant with respect to the stretchings (5.2)
is called homogeneous of degree n. Thus, let us consider solutions of Laplace's
equation that are homogeneous of degree n. In this case, keeping in mind the
expansion of an arbitrary solution in homogeneous solutions, we shall take η to
be an integer.
Writing the operator of the group of stretchings (5.2) in spherical
coordinates in the form
d d
ν ,
Λ = r ——f nu - r— ,
or ou
we determine from the equation XJ = 0 a basis of invariants λ = ur~ n, θ, φ .
The invariant equation λ = Y n(Q, φ ) (the index η means that for each choice
of η we have a different function Y) provides us with the following general
form of an invariant solution:
(5.3) u=rnY n(Q, φ ).
Then
ΔΗ = ~ I - ^ r (nrnnY,,) + Γ "Δ 2 7 „ 1 = r"- * [n (n + 1) Υ n + Δ 2 Υη ],
and Laplace's equation assumes the form
(5.4) A 2 y u + n ( n + l ) F »= 0 .
Therefore the coefficient Y n of the invariant solution (5.3) is an eigenfunction of
the Laplace —Beltrami operator Δ2 (with eigenvalue n(n+ 1)); it is called a
spherical function of order n.
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXW
131
(5.5) «( Γ, β , φ )= = ^ β + A) + J ^ n r
» +
_^L- jΚη(θ, φ),
where the double integral is taken over the domain bounded by the
characteristics χ = XQ, y = JO and the curve γ. The function υ is called the
Riemann function, while the boundary- value problem (5.7), (5.8) which defines
this function is called the characteristic Cauchy problem, or the Goursat
problem.
if ι
Clearly its solution (solutions of the G oursat problem are unique) is the
function ο = 1.
Example 2. The telegraph equation
(5.9)
is one of the simplest types of equations (after uxy = 0) to which Riemann's
method is applicable. The G oursat problem (5.7), (5.8) in this case has the
form
(5.10) ν (, 1, • 1.
Usually, textbooks offer the following "m ethod" for solving it: let us look for
a solution of the problem (5.10) in the form
(5.11) v^V(z), 2
= (x- ar 0 ) ( # - ». ) -
Then we obtain the ordinary differential equation
Thus the Riemann function for the telegraph equation (5.9) is expressed in
terms of a Bessel function in the form
(5.12)
Example 3. Riemann [51], §9 applied the method he suggested to the equation
(5.13) i;xy - ; ———- ν = 0, I = const.
{% ~i" y)
(5.13') r ' U *. = l, P | I - * = 1 ·
Riemann reduces the resulting characteristic Cauchy problem to an ordinary
differential equation (which defines a special G auss hypergeometric function)
by assuming that Ό is a function of the one variable
(bii) (χ—*0)(y- y0)
( 5 14 ) l
' - (ζ, + νΜχ + ν) '
In the last two examples we could reduce the construction of the Riemann
function to the solution of an ordinary differential equation by having guessed
correctly the form of the solution of the G oursat problem. It turns out that
the ability to perform such a reduction is related to the fact that the equation
admits a sufficiently large group, and to the invariance of the conditionson
the characterstics with respect to a one- parameter subgroup of that group. I t
is the invariance principle, and not a lucky guess, that gives us the correct
form of the solution. Let us consider the equations of Examples 2 and 3
from this point of view.
Example 2'. The telegraph equation (5.9) admits a three- parameter group (in
addition to stretchings of the variable u and the infinite group consisting of
addition to u of any solution of the equation; this group is common to all
linear equations) with generators
admissible for the G oursat problem (5.10). Let us insist first on the
invariance of the characteristics χ = x0, y = jo- This invariance criterion can
be written in the form
[X(x- xt) ]s= rts,^a+ ^a:o= 0, [X (y- y0
Hence it can be seen that γ Φ 0 (otherwise α = β = 0). Therefore we can
set γ = 1 to obtain α = — x0, β = y0. The resulting operator
(5.15) X^- xJ- JL- ^- ^- A
134 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
dx dy
an invariant solution has the form ο = ο(µ). This is the invariant solution
found by Riemann; the variable ζ he used is related to the invariant (5.14) by
the functional relation ζ = µ/(1 — µ) and therefore is also an invariant.
Let us now move on to general statements. The starting point here will be
Lie's result [52] on the group classification of homogeneous equations (5.6).
Most convenient for our aims is the invariant (with respect to equivalence
transformations of linear equations) formulation of this result due to
Ovsyannikov [54] (see also [8], §9.6). Let us observe that in the assertions
below the admissible group is considered modulo addition to the dependent
variable of an arbitrary solution cp(;t, y) of the linear equation being
considered, that is, by an admissible algebra we mean a factor- algebra modulo
the ideal generated by the operators X = <p(x, j)6/ 8«. So let us start with the
following result of group classification.
Theorem 5.1. The equation
(5.18) t |i i ] e t t , + ( i ( i : , y)ux+b(x, y)uv+c(x, y)u=0
admits a four- dimensional Lie algebra if and only if the quantities
(5.19) p=k/ h, q=(ln\ h\ )Jh
(if h = 0, h and k have to be interchanged) are constant; it admits at most a
two- dimensional algebra if at least one of these quantities is non- constant. Here
we denote by h and k the Laplace invariants of equation (5.18):
(5.20) h
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXW
135
Corollary. The adjoint equation (5.7) admits a four- dimensional algebra if and
only if so does the original equation (5.18).
Indeed, in the mutually adjoint equations L[u] = 0 and L*[u] = 0 the
Laplace invariants h, k and Λ», fc, are related by a permutation: h = fc»,
k = Α.. This follows immediately from (5.7), (5.20). Therefore the
corresponding invariants (5.19) p, q and />*, q* are either constant for both
equations, or non- constant.
Remark. A general equivalence transformation of equations (5.18) has the
form
(5.23) *= / ( *) , y=g{y), a=>.{x,y)u.
The Laplace invariants are in reality invariant only with respect to a linear
transformation of the dependent variable u, without a change of the
coordinates x, y. On the other hand, the quantities ρ and q defined by (5.19)
are invariants of the general equivalence transformation (5.23) and therefore
arise naturally in the statement of Theorem 5.1.
N ow we can prove the main result of this section.
Theorem 5.2 [23]. Let the equation (5.6) have constant invariants (5.19). Then
the Goursat problem (5.7), (5.8) admits a one- parameter group and the Riemann
function can be determined from a second- orderordinary differential equation.
Proof. By Theorem 5.1 and its corollary, it suffices to consider equations (5.6)
for which the adjoint equation has already been reduced by a suitable
equivalence transformation to either of the forms (5.21) or (5.22), depending
on whether the invariant q* of the conjugate equation is or is not zero.
Suppose that q* = 0 and the adjoint equation has been reduced to the
form (5.21):
(5.21*) L*[u\ ^vxv+xvx+pyvt+p^xyv=Q, p*= con st.
136zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
for the function ν(µ), while the data (5.2Γ) on the characteristics beome the
initial condition F (0) = 1. So the construction of the Riemann function
reduces to solving equation (5.24) with this initial condition.
Suppose that q* Φ 0 and the adjoint equation has been reduced to the
form (5.22):
/ ?*. q+ —c o n st .
- — u — {), I s= const ?t 0,
then, after a preliminary analysis of the defining equations, it turns out that
this operator must have the following special form:
(5.27)
The second useful observation has to do with these equations and allows us to
give them a clear geometric meaning.
Let us start by noting that up till now we have actively used the tangent
vector field of a one- parameter group in the form of a first- order linear
differential operator (1.7). If we now consider the close connection between
138 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
( fts°) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
+ {+ ) °
and the conditions on the characteristics
(5.31)
/ + y0 V / *o +
from which
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZYXWV
139
After this, by integrating the third equation of (5.27'), which has the form
ά(σ + (ξ + τ\ )1/ (χ + γ)) = 0, we obtain σ = Ct+C\ l(y- x). Therefore the
operator admissible by equation (5.30) is equal to
Thus, the G oursat problem (5.30), (5.31) admits the one- parameter group with
the symbol (5.32).
An invariant solution has the form
The two conditions (5.31) on the characteristics become one initial condition
V(0) = 1.
(5.35) 7 —Ι / Γ / or / =
Taking, in particular, / = δ , we obtain from (5.35) the following
transformation of the δ- function:
(5.36)
— «a. (ή,
we obtain the infinitesimal transformation (5.35):
(5.35') f«/ -
For the δ- function this has the form
(5.36')
Group analysis of differential equationszyxwvutsrqponmlkjihgfedcbaZ
141
Using the Leray form on the surface (5.37) we can define the Dirac
measure δ (Ρ ) by
(5.38) <fi(P), φ> = [ φω.
A·
that is, equation (5.42). F urthermore, if dP/ dx* Φ 0 (for some i it must be
true in view of the condition grad Ρ Φ 0), then from (5.42) by differentiation
with respect to χ' we obtain the equality
™6(Ρ) + Ρ6'(Ρ)Ά=>0,
ox ox
which on division by dP/ 5x' gives us (5.43). F urther differentiation leads to
the equations (5.44).
F rom (5.44) we can see that the first- order equation
(5.45) Pf(P)+ mf(P)= 0
has the general solution
(m t) m
(5.46) f*= Ci6 - (P)+ C2P- .
5.3.2. Laplace's equation.
It is well known that Laplace's equation
Διι= 0
To that end, let us determine first which operators in L leave invariant the
singular point χ = 0. It is easy to see that this property is shared by all
operators (5.47) apart from the translation operators Xt (i = 1,...,»). F or
rotation operators X^ and for the stretchings Z\ and Z 2 it is obvious, while
the fact that Y t leaves the origin fixed, Y txk \ x=o = 0, is clear from the
k 2
equality Y tx = 2 χ ' χ *- |*1 δ ' *.
Let us return now to the equation (5.48). F irst of all, it admits the
operators Xtj, since the Laplacian and the δ- function are invariant with
respect to rotations (see (5.36)). As far as the operators Z\ and Z 2 are
concerned, we can form from them an admissible linear combination
To find the value of the constant coefficient k, let us write down the
infinitesimal invariance criterion of equation (5.48). F or that we need to
prolong the operator (5.49) to second derivatives, and also to add a
transformation of the δ- function according to (5.36'), in which in this case we
have Α- (ξ') = η. As a result we obtain the operator
(5.49') J W ^ +
/ ^ + (
*_ 1)U iJ L +
In a similar manner we can verify that the operators 7; are also admitted.
However, we do not need them and will regard them as excess symmetry of
the fundamental solution.
In accordance with the invariance principle, we shall look for a
solution of equation (5.48) that is invariant with respect to the group
with generating symbols (5.50). The rotation group has two invariants:
r = γ ( Λ 1 ) 2 + . . . + (xn) 2 and u. In these invariants the stretching operator of
(5.50) is written in the form
common to all the operators (5.50). By Theorem 1.3 an invariant solution can
be written in the form / = const, that is,
(5.51) a^C r 8 - ".
C = - .
By substitution into (5.48) we find the value of the constantzyxwvutsrqponmlkjihgfedcbaZY
r—- ,
(2 - n)iln
where Ω π is the surface area of the unit sphere in «- dimensional space.
Therefore the fundamental solution
(5.53) Η / - Δΐϊ= δ ( ί, χ)
with η- dimensional Laplacian Δ in the variables x'. Obvious analogues of
(5.50) are the operators of rotations, G alilean transformations, and stretchings:
(5.54) ^ w J L - ^ J L ,
4r + Ά« » 4-
ι
dt βχ du
As in the case of Laplace's equation they are singled out from the full algebra
of symmetry of the heat equation by being admissible for the equation (5.53).
Now let us use the invariance principle. Invariants for Xy are t, r, u. In
the space of these invariants the operators Yj are written in the form
Clearly, they can be obtained from the (n+ l)- dimensional version of (5.50) by
making one of the variables imaginary. From this analogy it is completely
clear that for the rotations Χί} and for the Lorentz transformations 7, there
are two invariants: u and τ = t2 — r2, and that in these invariants the last
operator of (5.58) has the form
(5.59) A +
( i_ n ) u . i-
From this it is also clear that the invariant common to all the operators (5.58)
is J = ta^- "~ 1^ 2, so an invariant solution can be sought in the form
1
(5.60) a - C x"- "" .
However, substitution of (5.60) into (5.57) shows that there is no such
solution. The reason is very simple: equation (5.57) has no classical
solutions. N onetheless, we shall find an invariant solution by moving into the
space of distributions. To this end, let us look for an invariant solution of
the form
(5.61) α = / (τ).
As this equality is written in terms of the invariants u and τ of the operators
Xij and Yi, it remains to write down the invariance condition with respect to
the operator (5.59). This condition has the form
or
(5.62)
which is the same as (5.45). Therefore its general solution for m = 0 is given
by (5.41'):
(5.64) / (τΗ
while for m Φ 0 it is given by (5.46):
(5.65)
Substitution into (5.57) leads to 2C\ = n~ m, C2 = 0. Returning to the
original notation, we obtain a fundamental solution of the wave equation for
any odd number η of spatial variables:
(5.66) ΐ ί = θ ( ί 2 - ζ *) / 2 for w==l,
(5.67) for
The group method of constructing fundamental solutions presented above
was applied recently [62] to wave equations in Lorentz spaces with a non-
trivial conformal group.
CHAPTER VI
SUMMARY OF RESULTS
Table 1
Frequently encountered groups on the plane
Rotation
χ = χ cos a+y sin a, X = y- L - xJ- / = ar»+ y* t — arctan — ,
dx dy
y = y cos a — x sin a u — f x* + y*
Lorentz transformation
χ - χ cosh a+y sinh a, J = y* 1«
1
tt = _ — In
y
ID. + * ,
—zyxwvutsrqponmlkjihgfedcbaZYXWV
y = y cosh a + χ sinh a dx dy 2 y —x
Galilean transformation
χ = x + ay, y = y
Group analysis of differential equations 147
Table 1 (contd.)
Canonical variables
G roup Symbol Invariant
in \ vti;li X = δ/ θί
Homogeneous stretching X= x ° y ° ( = In x, u = —
+
χ = xe", y = ye" dx dy y y
Non- homogeneous stretching
X = xJL + kyA- t = In x, u = .fi
χ = xe", y = ye1" ox dy y y
Projective transformation
x
- y
X
l - a x ' ) 1- ax * = *·- £- + *?.*. I
x
χ y
dx dy y
Table 2
Some first- order equations with a known admissible operator
(1) d .
x= d .
ν ' = F(x),
(XIII) x= d
xy' = y [In y - f- F(χ)], ••
Table 3
Some second- order equations with a known admissible operator
(I ) y = F( y, y') , x- ^;
= F( x, y') ,
— F(kx- \ - li X = l— — kJL·
dx dy
y" = ( i + y'*)' V ^ + ifv' )· ~
X^y- L .
~dx "dy '
(I I I )
dx
(VI) y" + q'(v) rV· 9'(y) T \
?(v) J(V) I OX
( Χ") y" y
dy
9(y)
(XIII) yy "= y'*+y*f ' f z, ^ 1 - In !/
wr X = x»J- .
> dy
Table 4
Canonical form of second- order equations admitting a two- dimensional Lie algebra
Basis of Li in Equation
Type Structure ai La
canonical variables
d d
1 (Χι. * , ] = ο, Χι\ / Χιφ Ο λr m
^^mm γ __ y' = / (sO
Π [Χι, Χ,} = 0, . λf 0 γ 0
y" = / (*)
d » Q d
III [Χι. ^ι VzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGF
Χ* Φ ο χ » =
~dy"' dx dy X
_ d y _ d
IV (Χι. Xt)**Xl, Λ] γ Λ J ^= U y
« = / (x) y'
dy ' dy
Table 5
Computation of Lie algebras of operators admissible by a second- order equation
Admissible operator
Χ^ Χ + ζ , / , + ζ , - ^ - ,
Second prolongation of
ζτ = - 0 (η) - y'0 (ξ) = η , + (η ν - ξ,) / - »"ξρ,
the operator ζ» = D (ζ,) - y·/ ) (|) = η«χ + (2η«ν - ξβ ) • +
+ (%ν - 2ξ«ν) ** - / n»» + (η» ~ 2ξχ - 3»^ ) ν β
Table 6
Algorithm for integration of second- order equations
using a two- dimensional Lie algebra by the method of canonical variables
4 Using Theorem 2.4, reduce the basis of the algebra La Determination of the
to one of Table 4 by a suitable change of variables integrating change of
x, y. By this change of variables reduce the equation variables
under consideration to canonical form of
corresponding type from Table 4 and integrate it
(§2.2.3)
Table 7
Real non- isomorphic structures of three- dimensional Lie algebras
(classification by Bianchi [26])
I 0 0 0
II 0 0 - Xt
III 0 0 - Xi
IV 0 Xt - (Xt + Xt)
V 0 Xt — Xi
VI 0 kXt, k φ 0 Η 1 — Xi
VI I 0 Xl+ qXt Xt — QXI
VI I I Xi χ» - 1Xt
IX X* Χι Xt
Remark. In the complex domain there remain only seven non- isomorphic structures:
type VII assumes the form of VI with k = {q+i)l(q- i) if we change to the complex
basis ΑΙ = X2 + iX\ , X2 = X2—iXu X3 = X3/ (q- i), while type IX becomes the same
as type VII in the basis X\ γ iY Υ — rY Υ = Xi + iXs.
Group analysis of differential equations 151
Table 8
Real non- similar three- dimensional Lie algebras an d invariant equations
Type in
Ν Basis of L 3 Equation
Table 7
dx dy '
d d y =
1· IX
A — —χ ' i » ·/
dx dy
2
dx dy
dx ' dy
1
VI I I dx~' dx 2 dy'
3
*~~ ~dx dy
d d
X — X —
4··
V
*~ ~dy~
„ a Y
a
A, = _ _ , Ai = _ — ,
5· VI I dx dy
d x
Xt = (qx + v)- g
j + (w — )- gg-
Χι = _ _ , Χ2 = _ _ ,
οχ oy
6 VI
for dx dy
III d d
ν v
for Al= _ — , Aj= Χ - τ- ι
7 y" = Cx *"*1
dy Oy
dx dy
γ 8 γ θ
Αι —. , Αϊ _ ,
8 dx dy
V Χ — d
d
y' = 0 .
σχ σι/
θ θ d
9 Χ — Χ —χ Χ —
dy dy dy
152 Table 8 (contd.)
Type in
Ν Basis of L 3 Equation
Table 7
10
X, «*
ax dy
IV
v d - τ d
Λ] = , A2 = X ——
x
11 y' = C e
dx
II d v
d v
d
= - — , Λ2 = _ _ , A S = I _ _
σι d <ty d dy
= zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED
v
- r - , A2 = χ _ _ ,
Sy oy
13··
Remarks. * In the complex domain, case l_reduces to case 2 by ^hanging to variables
x, y using the relations χ — (xy—l)fcx+y), y — —i(xy+l)/ (x+y), while 5 reduces
to 6 with k - (q + i)l(q—i) bythe change of variables χ - (y- ix)jl, y - (y+ix)[2.
• • In cases 4 and 13 there are no invariant second- order equations, so third- order
equations are quoted.
Table 9
Lie's group- theoretical classification of second- order equations
d
Gi γ
Ai = ——
** = / (». ιΟ
d dx d
Υ Υ
dx dy
G·
y« = J_/ (y')
dx dy dx dy
χ —v
dx dy
ν & ν ι d * .. & ν md . d y* = Cy~*
Gs =_ , A-s = 2 *_ + y _ , A - *- _ + * , _
X l s = I
x i ~ J L , x 2 = ^_, x9 = x J L + ( X + y )° y'=Ce~V
fc- a
dx dy &x oy y'=Cy'*- 1, ΗφΟ, 4- .1.2
, Xi = - s—dj
Xl = ——dx > X* = Xdy
ιfX»= X - 3—dx - j— .
dx dy dy dx
Xt = y _ _ , Xt = y — , X, = x *_ _ + x y - _ , y" = 0
dx dy dx dy
* — ' £ + • "£
Group analysis of differential equations 153
Table 10
Construction of the algebra admitted by a linear hyperbolic equation
in two variables
Equation u
xy +
α (
*·
v) u
x +
b ( x>
"' y
u + c (x tt
· V) = °
Laplace invariants A = ax + 06 — c, k = by - f ab — c
References
[1] F . Klein, Review of Theorie der Transformationgruppen, Bd. I ll, at the award of the
first N .I. Lobachevskii Prize, 22 October 1897, Typolithography of the Imperial
University, Kazan 1989, pp. 10- 28.
[2] S. Lie, Vorlesungen uber Differentialgleichungen mit bekannten infinitesimalen
Transformationen (revised and edited by D r G . Scheffers), B.G. Teubner, Leipzig 1893.
[3] , Klassifikation und Integration von gewonlichen Differentialgleichungen
zwischen x, y, die eine G ruppe von Transformationen gestatten, Arch, for M ath. VIII
(1883), 187- 453 (parts I — III), IX (1884), 431- 448 (part IV); see [4], Vol. 5, articles
x, xi, xrv, xvi.
[4] — , G esammelte Abhandlungen, B.G . Teubner, Leipzig; H . Aschehoug & Co.,
Oslo, Vol. 1 (1934); Vol. 2 (Part 1) (1935); Vol. 2 (Part 2) (1937); Vol. 3 (1922);
Vol. 4 (1929); Vol. 5 (1924); Vol. 6 (1927).
[5] - , Zur Theorie des Integrabilitatsfactors, Christ. F orh., Aar 1874,
Christiania 1875, pp. 242- 254; see [4], Vol. 3, article XIII.
[6] , G eometrie der Beruhrungstransformationen (Presented by Sophus Lie and
G eorg Scheffers), B.G . Teubner, Leipzig 1896.
[7] , Die infinitesimalen Beruhrungstransformationen der Optik, Leipz. Berichte 1,
131- 133 (1896); see [4], Vol. 6, article XXIV.
[8] L.V. Ovsyannikov, Gruppovoi analiz differentsial'nykh uravnenii, N auka, Moscow 1978.
M R 80d:58072.
Translation: G roup analysis of differential equations, Academic Press, New York —
London 1982. M R 83m:58082.
[9] R.L. Anderson and N .H . Ibragimov, Lie —Backlund transformations in applications,
SIAM, Philadelphia, PA 1979. M R 80c:58048.
154 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
N.Kh. Ibragimov
[50] A.N. Tikhonov and A.A. Samarskii, Uravneniya matematicheskoi fiziki, 5th ed.,
Nauka, Moscow 1977.
Translation: Equations of mathematical physics, McMillan, New York 1963.
MR 29 # 2498.
[51] G.F.B. Riemann, Uber die Fortpflanzung ebener Luftwellen von endlicher
Schwingungsweite, Abhandl. Koningl. Ges. Wiss. Gottingen 8 (1860).
[52] S. Lie, Uber die Integration durch bestimme Integrate von einer Classe linearer
partieller Differentialgleichungen, Arch. Math. Naturvid. 6 (1882), 328-368 (see [4],
Vol. 3, article XXXV).
[53] E.T. Copson, On the Riemann - Green function, Arch. Rational Mech. Anal. 1
(1958), 324-348. MR 20 # 4088.
[54] L.V. Ovsyannikov, Group properties of the S.A. Chaplygin equations, Zh. Prikl.
Mekh. i Tekhn. Fiz. 1960, no. 3, 126-145.
[55] J. Leray, Hyperbolic differential equations, Inst. for Advanced Study, Princeton,
NJ 1953. MR 16-139.
Translation: Giperbolicheskie differentsial'nye uravneniya, Nauka, Moscow 1984.
[56] I.M. Gel'fand and G.E. Shilov, Obobshchennye funktsii i deistviya nadnimi, Fizmatgiz,
Moscow 1959.
Translation: Generalized functions, Vols. I, II, Academic Press, New York-
London 1964, 1968. MR 29 # 3869; 37 # 5693.
[57] Y. Choquet-Bruhat, C. DeWitt-Morette, and M. Dillard-Bleick, Analysis, manifolds,
and physics, Parts I, II, North-Holland Publ. Co., Amsterdam-New York 1982,
1989. MR84a:58002; 91e:58001.
[58] Yu.Yu. Berest, Construction of fundamental solutions for Huygens' equations as
invariant solutions, Dokl. Akad. Nauk SSSR 317 (1991), 786-789.
= Soviet Math. Dokl. 43 (1991), 496-499.