ESI The Erwin Schrödinger International

Institute for Mathematical Physics

Boltzmanngasse 9
A-1090 Wien, Austria

On Some Equivalence Problems

for Differential Equations

A. V. Bocharov
V. V. Sokolov
S. I. Svinolupov

Vienna, Preprint ESI 54 (1993) October 7, 1993

Supported by Federal Ministry of Science and Research, Austria

 ON SOME EQUIVALENCE PROBLEMS
FOR DIFFERENTIAL EQUATIONS

A. V. Bocharov,♠ V. V. Sokolov,♥ and S. I. Svinolupov♦

International Erwin Schrödinger Institute for Mathematical

Physics, Pasteurgasse 4/7, A-1090 Wien, Austria.

August 22, 1993

Table of contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. Point Transform Linearizability of the Second Order Ordinary Differential Equa-
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. Contact linearizability of equations of the third order . . . . . . . . . . 5
4. Further Equivalence Problems . . . . . . . . . . . . . . . . . . . . 7
5. A few words concerning the techniques used . . . . . . . . . . . . . . 11

1. Introduction
From an intuitive viewpoint nonlinear differential equations that are integrable
by the modern (nonclassical) methods are differential equations reducible to linear
equations by transformations of some specific type. Besides if one analyzes a list
of known integrable equations one may easily find out that there are quite few
genuinely different integrable equations. Many integrable equations are reduced to
different (simpler) forms by appropriate changes of variables.
Therefore, the following problem seems to be rather important:
Find out whether a specific equation can be reduced by some trans-
formation to a certain model equation.
Certainly, to make a plausible problem formulation out of this we must specify
a class of transformations to be used.

1991 Mathematics Subject Classification. 58G37.

Key words and phrases. contact transformation, equivalence problem, linearizability, Painlevé
equations.

♠ Wolfram Research Inc.

♥ International Erwin Schrödinger Institute for Mathematical Physics
♦ Mathematical Institute, Ufa Center of Russian Academy

Typeset by AMS-TEX

Typeset by AMS-TEX
2 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

A classical choice would be to take point transformations or contact transforma-

tions. With this choice the corresponding problem is called the Cartan’s equiva-
lence problem.
The principal result of the present paper is a set of formulae that have been
made as effective as possible so that in the simplest cases a decision whether or not
a certain equation is equivalent to a linear may be made algorithmically.
We may think of these formulae and of their prospective generalizations as a
possible theoretical foundation for building a computer expert system on nonlinear
differential equations.
Since available Computer Algebraic Systems (such as Mathematica, Axiom, Re-
duce, Maple...) facilitate straightforward algebraic and differential operations with
symbolic expressions but do not provide much help with the inverse and more so-
phisticated operations, it is desirable to reduce each algorithm in the equivalence
theory to routine checking of appropriate differential identities.
We believe that it is possible to reach this high level of efficiency in many rea-
sonable classes of equivalence problems.

2. Point transform linearizability of the

second order ordinary differential equation
Let us show using the undergoing classical example ([1], [2]) how effective the
answer may be in a simple case.
Problem 1. Find necessary and sufficient conditions for an ordinary differential
equation of the form
d2 y dy
= F (x, y, ) (1)
dx2 dx
to be reduced by a transformation of the form
x̃ = φ(x, y), ỹ = ψ(x, y) (2)
to the equation
d2 ỹ
= 0. (3)
dx̃2
Solution.
(1) The equation (1) must be of the form
d2 y ¡ dy ¢3 ¡ dy ¢2 dy
= a(x, y) + 3b(x, y) + 3c(x, y) + d(x, y); (4)
dx2 dx dx dx
(2) Both of the following differential polynomials depending on the coefficients
a, b, c, d of the expression (4) must be identically zero:
K = αy + γx + aβ + 2bα + cγ, (5)
M = αx + βy − bβ − 2cα − dγ
where
α = bx − cy + ad − bc,
β = dy − cx + 2c2 − 2bd, (6)
2
γ = by − ax + 2b − 2ac
 On Some Equivalence Problems for Differential Equations 3

Assuming the condition for the (5) holds (and thus the equation is proven to be
linearizable), the next question would be how easy it is to find the functions φ and
ψ giving the actual linearizing transformation. E.g. is it easier than to solve the
original equation directly?
In a sense, it is, since it turns out that the required φ and ψ satisfy an overde-
termined system of linear differential equations.
Here is the precise formulation of this fact.
Proposition 1. Transformation (2) linearizes the equation (1) if φ and ψ are an
arbitrary pair of functionally independent solutions of the system
Xxx = 2pXx + cXx − dXy ,
Xxy = qXx + pXy + bXx − cXy , (7)
Xyy = 2qXy + aXx − bXy .
where p = −Sx /S, q = −Sy /S and where S stands for an arbitrary nonzero
solution of the system
Sxx − cSx + dSy − βS = 0,
Syy − aSx + bSy − γS = 0, (8)
Sxy − bSx + cSy + αS = 0.
Remark 1.
Provided conditions (5) hold, system (7), (8) is compatible in the following sense.
Define S, Sx , Sy , X, Xx , Xy arbitrarily at a generic point (x0 , y0 ) as initial con-
ditions for the system. Then there is a unique solution (S, X) of the system (7),
(8) with these initial conditions. In terms of the original unknowns φ and ψ it is to
be observed that the required linearizing transformation is defined by 8 parameters
(e.g. by values of φ, φx , φy , ψ, ψx , ψy , p, q at a generic point (x0 , y0 )).
This degree of arbitrariness is to be expected. Indeed, if a certain transformation
(2) linearizes equation (1), then its composition with any point symmetry of the
equation (3) does the same. It is well known however (see [3]) that symmetries of
the latter constitute the 8-parameter Lie group isomorphic to the SL(3). ¥
Remark 2.
It is easy to derive from system (7), (8) some relevant ordinary differential equa-
tions in each of the variables x and y. Consider, for example, system (8). If a = 0
then the second equation is ordinary. Assume that a 6= 0. Then it follows from this
equation that
1 b γ
Sx = Syy + Sy − S.
a a a
Substituting this expression for Sx to the third equation, we find that S satisfies
the following third-order liner ordinary differential equation:
ay ¡ bay − aby ¢ ¡ γy a − a y γ ¢
Syyy − Syy − + b2 − ac + γ Sy − − bγ − aα S = 0
a a a
In a similar way we get an ordinary differential equation in x. ¥
It is true that to solve the resulting equations is sometimes more difficult than to
integrate the original equation. Here is however a restricted version of linearization
problem which can really be solved in quadratures:
Restrict ourselves to transformations of the form
x̃ = x, ỹ = ψ(x, y). (9)
4 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

Proposition 2.
(1) Equation (4) is reduced to a linear equation of the form

yxx = 3f (x)yx + g(x)y + h(x) (10)

by a transformation of the form (9) if and only if

a = 0, 2bx − cy = 0, My = 0, (11)

where
M = 2dy − 6bd − 3cx + 9/2c2 .

(2) If (11) is valid, then equation (4) is reduced by the transformation (9), where
ψ(x, y) is an arbitrary solution of the equation

ψyy + 3bψy = 0

depending on y effectively 1 , to equation (10) with

2 ψyx 1 3 9
f= + c, g= M + fx − f 2 , h = ψxx − 3f ψx + dψy − gu. (12)
3 ψy 2 2 4

Remark 3.
It is easy to check that, due to (11) the functions f, g, h do not depend on y.
Thus the task of linearizing w.r.t. to the transformation group (9) is solved in
quadratures. ¥
Remark 4.
It is known [3] that using a point transformation of the form

x̃ = a(x), ỹ = b(x)y + c(x)

any equation of the form (10) may be brought to (3). However, in general, it can
not be done in quadratures [4]. ¥
Note that the reduction of (4) to (10) while investigating a specific nonlinear
equation means a considerable progress, because for linear equations the issue of
integrability in quadratures is very well understood and perfectly algorithmized.
Remark 5.
With the help of an original code written in muMATH we have performed a
complete testing of the equations of the form (4) from the Kamke reference book [5].
It turned out that more than one third of them are linearized using the algorithm
described in the Proposition 2. ¥

1 It is obvious that such a solution is obtainable in quadratures

 On Some Equivalence Problems for Differential Equations 5

3. Contact linearizability of equations of the third order

One of the results of the present paper is a list of conditions for reducibility of
a third-order ordinary differential equation to the equation y 000 = 0 by a contact
transformation. Though the problem of contact linearizability for third order ODE
have been considered in several papers (see, for example, [6]) we could not find
explicit formulas solving this problem at the algorithmic level.
Let us first recall some basic facts about contact transformations.
The most popular example of a proper contact transformation is the Legendre
transformation
x̃ = y1 , ỹ = y − xy1 ,
dy
where y1 = etc.
dx
Using the chain rule for differentiation, it is not difficult to find out how deriva-
tives are affected by this transformation. In particular,
1
ỹ1 = −x, ỹ2 = − .
y2

A generic contact transformation in the case of one independent variable is a trans-

formation of the form:

x̃ = φ(x, y, y1 ), ỹ = ψ(x, y, y1 ), (13)

where φ and ψ are any functions, satisfying the contactness condition

∂φ ¡ ∂ψ ∂ψ ¢ ∂ψ ¡ ∂φ ∂φ ¢
y1 + = y1 + . (14)
∂y1 ∂y ∂x ∂y1 ∂y ∂x

The contactness condition means exactly that ỹ1 does not depend on y2 :

∂ψ ∂ψ ∂ψ
y2 + y1 +
∂y1 ∂y ∂x
ỹ1 = = χ(x, y, y1 ).
∂φ ∂φ ∂φ
y2 + y1 +
∂y1 ∂y ∂x

Of course, the functions ψ, φ and χ must be functionally independent.

In case when φ and ψ do not depend on y1 , condition (14) holds automati-
cally. This means that the point transformation (2) is a special case of a contact
transformation.
In virtue of results by Bäcklund [7], (13) is the most general form of invertible
local transformation for the case of one dependent and one independent variable.
Theorem 1. An equation

d3 y dy d2 y
3
= F (x, y, , ) (15)
dx dx dx2
is reduced by a contact transformation (13) to

d3 ỹ
=0 (16)
dx̃3
6 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

if and only if
(1) it has the form

d3 y dy ¡ d2 y ¢3 dy ¡ d2 y ¢2 dy d2 y dy
3
= a(x, y, ) 2
+3b(x, y, ) 2
+3c(x, y, ) 2 +d(x, y, ); (17)
dx dx dx dx dx dx dx dx

(2) the coefficients a, b, c and d of equation (17) satisfy the following differential
identities:

L = 0, N = 0, 3H − K = 0, 3F − M = 0, (18)
h = 0, G = 0, f = 0, λ = 0, µ = 0,

where

K = ∂1 (α) + ∂(γ) + aβ + 2bα + cγ,

M = ∂(α) + ∂1 (β) − bβ − 2cα − dγ,
L = ∂(β) − 2∂y (d) − 2cβ − 2dα,
N = ∂1 (γ) − 2∂y (a) + 2bγ + 2aα,
F = ∂(α) + 2∂y (c) + bβ − dγ,
H = ∂1 (α) + 2∂y (b) − cγ + aβ,
Q = ∂(K) + ∂1 (M ) + 3/2α2 − 3/2βγ,
G = −3∂y (α) − ∂1 (M ) + ∂(K) + 2bM + 2cK,
f = 3∂y (β) − 2∂(M ) + 2cM + 2dK,
h = 3∂y (γ) + 2∂1 (K) + 2aM + 2bK,
λ = 2∂(Q) − 2∂y (M ) + 2αM − 2βK,
µ = 2∂1 (Q) + 2∂y (K) + 2αK − 2γM,
α = ∂(b) − ∂1 (c) + ad − bc,
β = ∂1 (d) − ∂(c) − 2bd + 2c2 ,
γ = ∂1 (b) − ∂(a) − 2ac + 2b2 .

∂ ∂ ∂ ∂
Here ∂, ∂y and ∂1 stand for the ∂x + y1 ∂y , ∂y and ∂y1 respectively.

Example 1.
Let us figure out when an equation of the from

d3 y d2 y dy
3
= f (x) 2
+ g(x) + h(x)y + s(x); (19)
dx dx dx

is transformable to equation (16).

Substituting the right hand side of the equation into the (18) we find out that
all of them, except L = 0, hold trivially. The condition L = 0 is then equivalent to
the relation
1 1 1 2 1
h = − f 00 + f f 0 − f g − f 3 + g 0 .
6 3 3 27 2
¥
 On Some Equivalence Problems for Differential Equations 7

Example 2.
Consider an equation of the form

d3 y ¡ d2 y ¢2 ± dy
= c (20)
dx3 dx2 dx
with a constant parameter c in the right-hand side. The only nontrivial linearization
condition for equation (20) is the relation N = 0. In this particular case it is
equivalent to the equation
2c3 − 9c2 + 9c = 0.
It follows that the nonlinear equation (20) is transformable into equation (16) if
and only if c = 3/2 or c = 3.

4. Further Equivalence Problems

Here we would like to discuss some unsolved or partially solved Cartan equiva-
lence problems.
4.1. Ordinary differential equations.
We consider the following problem to be important:
Problem 2. Find necessary and sufficient condition for equation (1) to be trans-
formable by a point transformation (2) to one of the 6 Painlevé transcendental
equations 2 .
The importance of this problem is in particular due to the fact that the nowadays
fashionable Painlevé integrability test ([8], [9]) has one basic drawback: it is not
invariant with respect to variable changes of the form (2).
In the paper [10] the problem of reducibility of the equation (4) to the first or
second Painlevé transcendental equations has been considered.
Recall that a standard form of these equations is:

d2 y
= y 2 + x, (21)
dx2
d2 y
= y 3 + xy + a (22)
dx2
The reducing transformations were selected within the smaller group of trans-
formations of the form
x̃ = φ(x), ỹ = ψ(x, y) (23)
The equivalence problem has been solved in [10] for the equation (21), and a
solution for the equation (22) has been suggested that appears to be incorrect for
various reasons. In particular the authors of [10] make a wrong assertion that
equations of the from (22) with different values of a are equivalent to each other.
Below we generalize the results of [10] concerning equation (21) giving criteria
for reducing equation (4) to equation (21) by a general point transformation of the
form (2).
2 It is easy to see that an equation, transformable to one of the Painlevé’s, must have the form
(4).
8 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

Lemma. Any equation of the form (4) with M = 0 is reduced to an equation with
K = 0 by the substitution x̃ = y, ỹ = x. Here K and M are differential expressions
defined by (5).
Proposition 2. If two equations of the form (4), both with K = 0, are related by
a transformation of the form (2) then this transformation must actually have the
form (23).
This proposition is very important in the sequel, because the first Painlevé equa-
tion (21) has the property K = 0. Therefore if the source equation (4) has either
K = 0 or M = 0, we easily fall into the context of the [10].
Theorem 2.
(1) Assume that for an equation (4) both K 6= 0 and M = 6 0. Then the equation
is reduced to the first Painlevé (21) if and only if the following identities
and inequations hold:

k = 0, m = 0, F x = Hy , R = 0, S1 6= 0, S3 6= 0, S4 6= 0,

where
1 4
k= KKx + KMy − M Ky − aM 2 − 2bKM − cM 2 ,
3 3
1 4
m = M My + M Kx − KMx + bM 2 + 2cKM + dK 2 ,
3 3
Ky M
F = +a + b,
K K
Mx K
H= −d − c,
M M
2 ¡ 1 ¢
S1 = 2
Hx − H 2 − cH + dF + 5β ,
5M 5
6
S2 = (S1 )x + HS1 ,
5
S2 ± 36 ± ±
S3 = 4( )x M + HS2 M 2 + 8d M 3 − S12 ,
M 5
± 24 ±
S4 = 2(S3 )x M + S3 H M.
5
R = (S4 )x + 3HS4 .

Here the expression for β defined in the same way as in the Theorem 1.
(2) The required transformation is given by the explicit formulae

−4/5 −2/5
x̃ = S3 S4 , ỹ = S1 S4

Remark 6.
Note that the this result is much more effective then in the case of the Problem
1. The required transformation is built in a straightforward manner using the
coefficients of the original equation (4).
The reason for such high efficiency is that the first Painlevé equation (21) does
not have point symmetries and therefore the reducing transformation is unique (see
 On Some Equivalence Problems for Differential Equations 9

Remark 1). The situation with the rest five Painlevé transcendental equations in
this respect is the same. ¥
Along with the Painlevé equations, integrable from the classical viewpoint are
those equations of the form (1) that admit a 2-dimensional Lie group of point
symmetries [11, p. 200]. Consequently, the following problem is of major practical
interest
Problem 3. Find necessary and sufficient conditions for a given equation of the
from (1) to be point-equivalent to one of the model equations, admitting a 2-dimensional
Lie group of point symmetries.
Remark 7.
A list of model equations, admitting 2-dimensional symmetry groups can be
found, for example, in [12]. The Problem 3 restricted to transformations of the
form (23) was considered in [13]. The results of the latter work would be of much
practical use should they be completed by providing recipes for building the required
equivalence transformations in quadratures. ¥
4.2. Partial differential equations.
From the viewpoint of possible computer implementations, it would be important
to obtain criteria of linearizability of simpler partial differential equations.
Let us point out the following generalization of the Problem 1 (that may come
unexpected for the reader).
Problem 4. Find criteria of reducibility of the equation

F (x, y, u, ux , uy , uxx , uxy , uyy ) = 0 (24)

to the linear ordinary differential equation

ũx̃x̃ = 0

it by a point transformation

x̃ = φ(x, y, u), ỹ = ψ(x, y, u), ũ = χ(x, y, u) (25)

Remark 8.
We have come across a number of research papers dealing with formal properties
of specific equations from the (24) class in the context when they are actually
reducible to (3) by transformations (25) (in a nontrivial way, the circumstance
being concealed from the authors). ¥
Let us list some of the simplest results relevant to the partial differential equa-
tions.
Theorem 3.
(1) Equation (24) is reduced to the ordinary differential equation ũx̃x̃ = 0 by a
point transformation

x̃ = x, ỹ = y, ũ = χ(x, y, u)

if and only if it has the form

uxy = A(x, y, u)ux uy + B(x, y, u)ux + C(x, y, u)uy + D(x, y, u)

10 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

where the coefficients A, B, C, D are related by the following identities:

Cu = Ax , B u = Ay , Bx = Cy , Du − Bx − AD + BC = 0.
(2) The function χ, defining the transformation, is available in quadratures as
a solution of the following compatible system:
(ln(χu ))x = −C,
(ln(χu ))y = −B,
(ln(χu ))u = −A,
χxy = −χu D.

Theorem 4. For the equation

ut = A(t, x, u)uxx + F (t, x, u, ux ) (26)
to reduce via a contact transformation to the equation
ũt̃ = ũxx
˜ (27)
it is necessary and sufficient that
(1) equation (26) is of the form
ut = A(t, x, u)uxx + B(t, x, u)u2x + C(t, x, u)ux + D(t, x, u);
(2) the following identities hold:
Au = 0, (28)
Cu /A − 2(B/A)x = 0, (29)
2
(2Cx − 4Du − 2CAx /A + (C − 4BD)/A)u − 4(B/A)t = 0, (30)
(K3 )x = 0,
where
3
K3 = 2A1/2 (K2 )x + Att A−1 − A2t A−2
2
1 3
K2 = A1/2 (CA−1 )t + A1/2 ( Axx − A2x A−1 − At A−1 − 2K1 )x ,
2 8
1 1 1
K1 = Cx − Du − CAx A−1 + (C 2 − 4BD)A−1 .
2 2 4
Remark 9.
In fact, it turns out that the required reducing contact transformation is neces-
sarily a point transformation of the following special form:
t = χ(t), x = φ(t, x), u = ψ(t, x, u).
Equations for the required functions χ, φ and ψ may be found in the paper [14].
¥
On the whole, situation here is much similar to the one described in section 1.
That is the problem of reducing the source equation to (27) cannot generally be
solved in quadratures. However, an important step for reducing it to a more general
linear equation
vt = α(t, x)vxx + β(t, x)vx + γ(t, x)v + δ(t, x)
is done in quadratures provided the conditions (28)-(30) are satisfied.
 On Some Equivalence Problems for Differential Equations 11

5. A few words concerning the techniques used

There are several formally different ways to solve equivalence problems for dif-
ferential equations.
The most popularized are approaches related to the theory of invariants as well
is to the G-structure theory (cf.[15], [16]).
Our approach however is based on a more or less straightforward step-by-step
(often a computer-aided) study of the overdetermined partial differential system
defining the required transformation. In the process of bringing this overdetermined
system to ”passive from” (cf.[17], [18] ), we get a number of compatibility conditions.
These conditions are differential identities for the coefficients of the source equation.
Let us, for example, present an overdetermined defining system relevant to the
problem 1.
It is easily seen that a transformation of the form (2) would imply the following
transformation of the second derivative:

d2 ỹ d2 y ¡ dy ¢−3 ¡ dy dy ¢¡ dy ¢−2
2
=J 2 φy + φx + ψyy ( )2 + 2ψyx + ψxx φy + φx −
dx̃ dx dx dx dx dx
¡ dy ¢¡ dy dy ¢¡ dy ¢−3
ψy + ψx φyy ( )2 + 2φyx + φxx φy + φx ,
dx dx dx dx
where
J = ψ y φx − ψx φy (32)
is the Jacobian of the transformation (2).
Substituting the above expressions into equation (3), we get the equation of the
form (4) the coefficients of which are defined by the formulae:

a = J −1 (ψy φyy − φy ψyy ),

1 1 2 2
b = J −1 ( ψx φyy − φx ψyy + ψy φyx − φy ψyx )
3 3 3 3
1 1 2 2
c = J −1 ( ψy φxx − φy ψxx + ψx φxy − φx ψxy (33)
3 3 3 3
−1
d = J (ψx φxx − φx ψxx ).

To solve the Problem 1, we should consider the four equations (33) as an overde-
termined system of nonlinear partial differential equations for the unknown func-
tions φ and ψ. The system turns out to be equivalent to the set of relations (5),
(7) and (8). Note that the S function of the system (8) is nothing else J −1/3 .
Acknowledgements
The authors wish to thank V.Driuma, N. Kamran, I.Krasil’shchik, Yu.Romanov-
ski, and V.Yumagouzhin for a lot of useful discussions and collaboration. One of
the authors (VVS) is thankful to A.Vinogradov and P.Michor for hospitality while
staying at ESI, where part of this work was done.

References
1. M.A.Tresse, Determination des invariants: ponctuels de l’equation differentielle du second
ordre y 00 = w(x, y, y 0 ), Hirzel, Leipzig, 1896.
12 A.V.Bocharov, V.V.Sokolov and S.I.Svinolupov

2. E.Cartan, Sur les varietes a connexion projective, Bull. Soc. Math. de France 52 (1924),
205-241.
3. Ovsiannikov L.V., Group Analysis of Differential Equations, Moscow: Nauka, 1978. (Rus-
sian)
4. I.Kaplansky, Introduction to Differential Algebra, ”Inostrannaia Literatura”, 1959. (Rus-
sian)
5. E. Kamke, Differentialgleichungen, Lösungsmethoden und Lösungen, Band 1, Gewöhnliche
Differentialgleichungen, Chelsea Publishing Company, New York, 1959.
6. Shiing-shen Chern, Sur la geometrie d’une equation differentielle du troisieme ordre, Comp-
tes Rendus de l’Academie des Sciences a Paris (1937), 1227.
7. A.V.Bäcklund, Uber Flächentransfromationen, Math. Ann. 9 (1876), 297-320.
8. P.Painlevé, Memoire sur les equations differentielles d’ont l’integrale generale est uniforme,
Bull. Soc. Math. de France 28 (1900), 201.
9. J.Weiss, M.Tabor, and G.Carnevale, The Painlevé Property for Partial Differential Equa-
tions, J. Math. Phys. 24 (1983), 522-526.
10. N.Kamran, K.G.Lamb, and W.F.Shadwick., The Local Equivalence Problem for d2 y/dx2 =
F (x, y, dy/dx) and the Painlevé transcendents, J. Differential Geometry 22 (1985), 139-150.
11. Peter J. Olver, Applications of Lie groups to differential equations, Springer-Verlag, 1986.
12. F.M. Mahomed, Lie Algebras Associated with Scalar Second-Order Ordinary Differential
Equations, J.Math.Phys. 12 (1989), 2770-2777.
13. L.Hsu and N.Kamran, Classification of Ordinary Differential Equations, Proceedings of the
London Mathematical Society 58, Part 2 (1989), 387-416.
14. V.V.Sokolov, S.I.Svinolupov, and T.Wolf, On Linearizable Evolution Equations of Second
Order, Phys. Lett. A 163 (1992), 415-418.
15. R.Liouville, Sur les Invariantes de Certaines Equations, J. de l’Ecole Polytechnique 59
(1889), 7-88.
16. C.Grisson, G.Thompson, and G.Wilkens, J. Differential Equations 77 (1989), 1-15.
17. J.Thomas, Differential Systems, AMS Colloquium Pub., N.Y. 21 (1937).
18. A.V.Bocharov, M.L.Bronstein, Efficiently Implementing Two Methods of the Geometrical
Theory of Differential Equations..., Acta Applicandae Mathematicae 16 (1989), 143-166.

Correspondence to: S.I.Svinolupov, Mathematical Institute of Ufa Center of

Russian Academy of Sciences, Chernyshevsky str. 112, 450000, Ufa, Russia
E-mail address: [email protected]