Concrete Resolution of Differential Problems using Tannakian
Categories
Jacques-Arthur Weil
Département de Mathématiques, Université de Limoges
Algorithms Seminar
April 19, 1999
[summary by Frédéric Chyzak]
A properly typeset version of this document is available in
postscript and in
pdf.
If some fonts do not look right on your screen, this might be
fixed by configuring your browser (see the documentation here).
Abstract
Given a linear ODE with polynomial coefficients, one easily finds
local information about its solutions. To obtain global information
of algebraic nature (operator factorization, explicit finite form,
algebraic relations between solutions), one classically reduces the
problem to determining rational or exponential solutions of auxiliary
linear ODE's. The latter are often uneasy to compute in practice, and
we show by a few examples how to advantageously substitute
differential systems that are simpler to construct, solve or study.
1 Solving Linear Differential Equations
The main question when studying a linear differential operator L is
how to ``solve'' for its solutions. ``Solving'', however, covers
several meanings. Throughout this text, L denotes a differential
operator acting on a function y in the variable x
by L(y)=any(n)+...+a0y for polynomials ai in x with
coefficients in a field C. This field is Q, Q_ or C in practice.
The simplest way to solve is the determination of local information,
like a basis of formal solutions in the neighbourhood of 0. The
general form of a formal solution is the formal series
y=x |
|
(p0(ln x)+p1(ln x)x1/r+...+pi(ln x)xi/r+···) |
for polynomials pi with uniformly bounded degrees. Here, r is
a positive integer, the ramification, and p0 is assumed to be
non-zero so as to ensure that the highest possible power has been
incorporated into the generalized exponent aÎ C[x1/r].
The power xa is nothing but expòa/x dx, the formal
solution of y'=(a/x)y. This approach by generalized exponents
is due to Van Hoeij [12] and unifies regular and irregular
singular expansions. A similar treatment was developed in the case of
systems by Barkatou [1] and
Pflügel [5].
Of course, the most generally understood acceptance of ``solving''
relates to resolution in closed form. By simultaneously considering
the bases of formal solutions in the neighbourhood of all possible
singularities of the operator L, namely, the zeroes of its leading
coefficient an(x), several algorithms are available to search for
solutions in various classes of closed form, like polynomial
solutions yÎ C[x], rational solutions yÎ C(x), exponential
solutions y for which y'/yÎ C(x), or liouvillian solutions y
for which y'/y is algebraic over C(x).
See [4, 13] and the references there.
Note that each solution s in the above classes supplies a
first-order right-hand factor of the operator L,
namely ¶-s'/s where ¶ denotes the derivation
operator with respect to x. A more general problem is that of the
factorization of operators from the ring C(x)[¶] of linear
differential operators with rational function coefficients, and the
search for higher-order right-hand factors. This relates to
differential Galois theory. More specifically, polynomial, rational,
and exponential solutions correspond to factorization in this ring,
whereas liouvillian solutions correspond to the more complex problem
of absolute factorization [13], i.e., factorization of an
operator LÎ C(x)[¶] with factors in K[¶] for an
algebraic closure K of C(x). In any case, factorization relates
to solving since any solution of any right-hand factor is a solution
of the original operator. Furthermore, specialized algorithms exist
for linear differential equations of small orders.
Right-hand factors of an operator are a first type of auxiliary
operators or lower order that simplify solving. More generally,
another form of ``solving'' the operator L is by looking for its
solutions that can be viewed as powers, products, or wronskians of an
auxiliary operator, or system of operators, of lower order. This is
the main discussion of the next sections. Applications include the
classification of solutions, connexion problems, number theory (by
looking for differential equations of minimal order), and the search
for first integrals of non-linear differential equations.
2 Lower Order Equations and Symmetric Power Solutions
As an example, consider the third-order equation y'''-4ry'-2r'y=0
(rÎ C(x)). It admits a basis of solutions of the
form (z1=y12,z2=y22,z3=y1y2), where both y1 and y2
are solutions of the same second-order equation y''=ry. To obtain
such special solutions of a higher-order operator L, the crucial
relation to be used is z1z2=z32. Indeed, considering the formal
solution z~i=xaiSi corresponding to the
expansion of each actual function zi, we obtain that the formal
expansion of the product z1z2 is the product of formal expansions
z~1z~2=xa1+a2S1S2.
Identifying those generalized exponents for L that can be a sum of
two terms therefore supplies a set of candidate exponents for the
auxiliary operator and the zi. Note that the original third-order
equation has been replaced by a ``simpler system'' consisting of a
second-order equation and a quadratic relation.
3 Liouvillian Solutions
To solve an operator L for its liouvillian solutions, one looks for
the possible irreducible polynomials P of the
form Xm-bm-1Xm-1-...-b0 such that P(u)=0
implies L(expò u dx)=0 [10]. Given the order n of
the operator, differential Galois theory shows that only finitely many
degrees are possible for the polynomial P. There exists an
algorithm to compute the list of the possible numbers m: for n=2,
the list is 1, 2, 4, 6, and 12; for n=3, it is 1, 3, 6, 9, 21,
and 36 [6, 7, 9]; for n=4 and higher, a formula
is known for the maximum number of the list.
By construction, the roots ui of P are logarithmic
derivatives yi'/yi of a solution of L,
and bm-1=åiui=åiyi'/yi is the logarithmic derivative
of the product Õiyi. A necessary and sufficient condition for
the existence of a polynomial P of degree m above, which describes
the liouvillian solutions of L is that there exists a polynomial of
degree m in solutions of L whose logarithmic derivative is
rational, and which is the product of linear factors. More
specifically, for a solution basis (z1,...,zm) of L the
product Õiyi is searched for under the
form Õi(ci,1z1+...+ci,mzm).
The search for liouvillian solutions therefore reduces to the search
for exponential solutions. To this end, the present work allows to
avoid computing the equation for the symmetric power, which is too
large, but prefers a more compact representation.
4 Factorization and Alternate Power Solutions
As another typical example, let us consider the search for a
right-hand factor H=¶2-b1¶-b0 of order 2 of the
operator L=¶4-a2¶2-a1¶-a0 of order 4. For
any solution basis (y1,y2) of H, the operator H is given by
the determinantal representation
H(y)= |
½ ½ |
|
½ ½ |
|
½ ½ ½ ½ |
y |
y1 |
y2 |
y' |
y1' |
y2' |
y'' |
y1'' |
y2'' |
|
½ ½ ½ ½ |
=y''- |
|
y'+ |
|
y
where
wi,j= |
½ ½ |
|
½ ½ |
. |
To obtain a factor of order 2, we now search for an exponential
solution and show that it can be interpreted as a
determinant w0,1. Let A be the companion matrix
A= |
æ ç ç ç è |
0 |
1 |
0 |
0 |
0 |
0 |
1 |
0 |
0 |
0 |
0 |
1 |
a0 |
a1 |
a2 |
0 |
|
ö ÷ ÷ ÷ ø |
,
and let
Y= |
æ ç ç ç è |
|
ö ÷ ÷ ÷ ø |
,
so that
Y'=AY. |
Let us introduce the
vector Z=(w0,1,w0,2,w0,3,w1,2,w1,3,w2,3)T.
In view of their definition, the wi,j satisfy differential
relations like w0,1'=w0,2,
w0,3'=w1,3+a0w0,0+a1w0,1+a2w0,2,
and so on. From them, we find a matrix
L2(A)= |
æ ç ç ç ç ç ç è |
0 |
1 |
0 |
0 |
0 |
0 |
0 |
0 |
1 |
1 |
0 |
0 |
a1 |
a2 |
0 |
0 |
1 |
0 |
0 |
0 |
0 |
0 |
1 |
0 |
-a0 |
0 |
a2 |
0 |
0 |
1 |
0 |
-a0 |
0 |
-a1 |
0 |
0 |
|
ö ÷ ÷ ÷ ÷ ÷ ÷ ø |
such that Z'=L2(A)Z. |
Again, we then only look for exponential solutions Z of the
matrix L2(A), which is easy to construct and contains more
information than the usual single auxiliary equation used for
factorization. Finally, one has to check that the solution Z is a
determinant. For this, a necessary and sufficient condition is the
Plücker relation, which here simply reduces
to w0,1w2,3-w0,2w1,3+w0,3w1,2=0.
To rephrase the method in a more formal way, introduce V, the
solution space of L. The search for Z is indeed a search for
objects in the 2-exterior power L2(V), i.e., the vector space
of linear combination of formal 2-exterior products vÙ w,
(v,w)Î V2, which satisfy the rule wÙ v=-vÙ w. Pure
exterior product uÙ v are interpreted as determinants. The
search for Z is therefore equivalent to the search for a pure
exterior product w0,1ÎL2(V) such that the
1-dimensional vector space Cw0,1 is stable under the action
of the differential Galois group of L.
Here the search for a second-order right-hand factor of a fourth-order
equation has been reduced to solving a ``simpler'' system of six
first-order equations.
5 Module and Dual Module Associated with an Operator
As an important tool for the study of a linear differential
operator L, one classically associates a canonical module in the
following way. For L in the algebra k[¶] of linear
differential operators with coefficients in a field k, one considers
the quotient M=k[¶]/k[¶]L of k[¶] by its left
ideal k[¶]L. The left module M can be viewed as the
module k[¶]y generated by a generic solution y of the
operator L. Linear constructs on and between solution spaces of
operators, like (direct or usual) sums, (symmetric or exterior or
usual commutative) products, (indefinite) integration, and so on,
correspond to constructs on and between the corresponding
k[¶]-modules.
A variant module is obtained by endowing the dual k-vector
space M* with a k[¶]-module structure. Let r be the
order of L, then M is of dimension r and its
dual M*=Homk(M,k) is isomorphic to kr. Now
let A be the companion matrix of L and (b1,...,br) be the
canonical basis of M*. The latter is turned into a
k[¶]-module by defining an operator Ñ on M* by
the action
(Ñ b1,...,Ñ br)T=-AT(b1,...,br)T
and letting ¶ act by Ñ. Thus, Ñ(am)=aÑ
m+a'm when aÎ k and mÎ M. From this Leibniz rule applied to
the product y1b1+...+yrbr=(y1,...,yr)(b1,...,br)T, we
derive the equality
Y'=AY for Y=(y1,...,yr)T
whenever Ñ(y1b1+...+yrbr)=0. Note that this
k[¶]-module structure on M* usually does not make it
the dual k[¶]-module Homk[¶](M,k),
for the operator L usually has no solution in k.
The modules M* allow for a better description of the
calculations suggested in the previous sections through a link between
the solution space V=Sol(L) and the
k[¶]-module M*. This link is obtained by introducing
the map f from V to M* defined
by f(y)=yb1+...+y(r-1)br. Calculations with elements of
the C-vector space Sol(L) have their counterparts
in the k[¶]-module M*. For example, one recovers the
determinants of the previous sections from the following identity for
exterior products in the module L2M*
f(y1)Ùf(y2)= |
|
wi-1,j-1 biÙ bj
with
wi,j= |
½ ½ |
|
½ ½ |
. |
Again, constructs at the level of solution spaces translate into
constructs at the level of the corresponding k[¶]-modules.
6 Tannakian Definition of the Differential Galois Group
This section is based on my (Chyzak's) study and tentatively reflects
what was not presented by the speaker for lack of time. It aims at
defining differential Galois groups by the Tannakian viewpoint, as an
alternative to Kolchin's more traditional and elementary definition by
differential extension fields. Interestingly, some properties are
easier to derive by the Tannakian viewpoint, for instance that it is a
linear algebraic group (i.e., a subgroup of GLn(C)
and an algebraic variety). Another consequence is the possibility to
rephrase algorithms in such a way that differential Galois theory, in
the sense of Kolchin, is only used as a classification tool to prove
the correction of the algorithms, while calculations take place at the
level of modules in a more efficient way. This presentation is based
on a discussion with the speaker, on conference proceedings by Ramis
and Martinet [8, Part 2, Chapter 1], and on unpublished
notes by Churchill [2, 3]. More direct
references may be works by Bertrand, Deligne, and Katz. The Tannakian
construction has a natural counterpart in difference Galois theory
[11, Section 1.4].
For comparison sake, Kolchin's definition of the differential Galois
group of a linear differential operator LÎ k[¶] is as
follows. Let C be the subfield of constants of k, n be the
order of L, and consider the Picard-Vessiot extensions k' of k
associated with L, i.e., the differential field extensions of k
that contain an n-dimensional C-vector space of solutions of L
and do not enlarge the constant field C. Then the differential
Galois group of L is defined as the group G of differential field
automorphisms (i.e., field automorphisms that respect the differential
structure) of any Picard-Vessiot extension k' that
additionally respect the action of k on k'. This mimics the
classical Galois theory for a polynomial PÎ k[X], where one
introduces the group of field automorphisms of a suitable
extension k' of k which contains all solutions of P and restrict
to the identity on k. While the (algebraic) Galois group of a
polynomial is a subgroup of a permutation group Sn, the
differential Galois group of an operator is a subgroup of the linear
group GLn(C) for the common field of constants C
of k and k'.
For its part, instead of a single extension k' of k, the Tannakian
presentation simultaneously considers a whole collection of
k[¶]-modules, and introduces the differential Galois group as
a group of internal transformations on this collection. Crucially,
each transformation has to transform all the modules in a way
compatible with the linear maps between the modules. Moreover, each
module M is associated with a solution set that can be viewed as the
kernel of the derivation on M, and the above-mentioned
transformations have to be compatible with taking solutions.
At the heart of the Tannakian construction are k-vector spaces V
that are closed under the action of an operator Ñ which extends
the action of the derivation on k by the Leibniz rule:
Ñ(af)=aÑ(f)+a'f, when aÎ k and fÎ V.
This makes V a k[¶]-module with ¶ acting
by Ñ.
From now on, we restrict to k[¶]-modules that are
finite-dimensional k-vector spaces. Fundamental examples are the
modules M=k[¶]/k[¶]L discussed in the previous
section. We also restrict to k=C(z). An element hÎkerÑ
is called a horizontal vector. As has been explained when discussing
dual modules M*, horizontal vectors in M* correspond to
solutions yÎ kr of the equation D y=0
where D=d/dz-A for (Ñ b1,...,Ñ
br)T=-AT(b1,...,br)T once a basis (b1,...,br)
of M* has been chosen. Rather than enlarging the space M
where we have a solution for L, as is the case in the traditional
differential Galois theory, we now enlarge the coefficient field
of M* so as to ensure the existence of a solution to D
and thus of horizontal vectors for Ñ. To this end, consider a
non-singular point aÎC of the operator L, and introduce the
field Ma of germs of meromorphic functions at a, which
is isomorphic to the field of convergent Laurent
series C{z-a}[(z-a)-1]. By Cauchy's theorem, the C-vector
space kerD, where D is now viewed as acting
on ( Ma)r, is of dimension r and supplies with
horizontal vectors of Ñ in MaÄC(z)M.
Note that kerD and kerÑ usually have no more structure
than that of C-vector spaces.
As an example, let us consider the C(z)[¶]-module generated
by the Bessel function of the first kind J0(z). It is a
two-dimensional C(z)-vector space with basis (J0(z),J1(z)),
and J0'=-J1. With the above notation,
æ è |
|
ö ø |
= |
æ è |
|
ö ø |
= |
æ è |
|
ö ø |
æ è |
|
ö ø |
,
so that
A= |
æ è |
|
ö ø |
. |
As a result of a simple computation, h=f1J0+f2J1 is a
horizontal vector of Ñ if and only if
æ è |
|
ö ø |
ÎkerD
=C z |
æ è |
|
ö ø |
ÅC z |
æ è |
|
ö ø |
, |
where Yn(z) are the Bessel functions of the second kind, and
where Jn and Yn now denote germs of the corresponding
functions (their local expansions at a¹0). This simplifies
to hÎC z(Y0J1-J0Y1), whence h is a constant by the
Wronskian relation Y0J1-J0Y1=2/p z.
To the module M above and any non-singular point aÎC, we have
just associated the C-vector space of horizontal vectors
of Ñ in the form of local expansions at a.
Denote ha(M) this vector space. We now proceed to associate
horizontal vectors to more involved module constructions.
Denote {M} the smallest class of k[¶]-modules
containing M and closed under finite direct sums and products,
finite symmetric and exterior products, dualization, and taking the
module of homomorphisms between two modules, and submodules. One can
extend Ñ from M to any VÎ{M} in a canonical way; in
particular:
Ñ| |
|
=
|
æ è |
|
ö ø |
,
Ñ| |
|
=Ñ|VÄ1|W+1|VÄÑ|W.
|
This class becomes a category for the usual k[¶]-module
morphisms. The map ha extends to {M} by
ha(V)=ker(Ñ| MaÄC(z)V). In
particular, ha(VÅ W)=ha(V)Åha(W)
and ha(VÄ W)=ha(V)Äha(W). The crucial fact
is that ha is compatible with the maps that are natural between
modules on the one hand and horizontal vector spaces on the other
hand. Specifically, any k[¶]-module homomorphism h between
two modules V and W induces a C-linear homomorphism ha(h)
between ha(V) and ha(W). This makes ha a functor,
in the sense that the diagram
To relate horizontal vectors at two non-singular points a and b,
consider the maps s that associate with any
k[¶]-module V a C-linear
map s(V):ha(V)®hb(V), subject to the
constraint that
Such a map s (from {M} to the linear morphisms in the
category of C-vector space) is called a morphism from (the
functor) ha to (the functor) hb. The collection of such
morphisms when a and b vary is a semigroup for composition. The
corresponding notion of isomorphisms (of functors) is obtained when
each of the linear maps s(V) is invertible. Two cases are of
interest: when a¹ b, one of those isomorphisms is provided by
analytic continuation along a path from a to b; when a=b, the
isomorphisms from ha into itself form a group (the group of
automorphisms of the functor ha). This group is the
differential Galois group of L, following the Deligne-Katz
definition.
References
- [1]
-
Barkatou (M. A.). --
An algorithm to compute the exponential part of a formal fundamental
matrix solution of a linear differential system. Applicable Algebra in
Engineering, Communication and Computing, vol. 8, n°1,
1997, pp. 1--23.
- [2]
-
Churchill (R. C.). --
Connections on modules. --
February 1997. Unpublished Notes for the Kolchin
Seminar in Differential Algebra.
- [3]
-
Churchill (R. C.). --
A comparison of the Kolchin and Deligne-Katz definitions of a
differential Galois group. --
February 1998. Unpublished Notes for the Kolchin
Seminar in Differential Algebra.
- [4]
-
Pflügel (Eckart). --
ISOLDE, a package for computing invariants of systems of ordinary
linear differential equations. In Salvy (Bruno) (editor), Algorithms
Seminar, 1997-1998, INRIA Research Report, pp. 79--82. --
1998.
- [5]
-
Pflügel (Eckhard). --
An algorithm for computing exponential solutions of first order
linear differential equations. In Küchlin (W.) (editor), ISSAC'97
(July 21--23, 1997. Maui, Hawaii, USA). pp. 164--171. --
ACM Press, 1997.
- [6]
-
Singer (Michael F.). --
Liouvillian solutions of nth order homogeneous linear differential
equations. American Journal of Mathematics, vol. 103, n°4,
1981, pp. 661--682.
- [7]
-
Singer (Michael F.) and Ulmer (Felix). --
Liouvillian and algebraic solutions of second and third order linear
differential equations. Journal of Symbolic Computation, vol. 16,
n°1, 1993, pp. 37--73.
- [8]
-
Tournier (E.) (editor). --
Computer Algebra and Differential Equations. --
Academic Press, Computational Mathematics and Applications,
1989.
- [9]
-
Ulmer (Felix). --
On Liouvillian solutions of linear differential equations. Applicable Algebra in Engineering, Communication and Computing, vol. 2,
n°3, 1992, pp. 171--193.
- [10]
-
Ulmer (Felix). --
Linear differential equations and liouvillian solutions. In Salvy
(Bruno) (editor), Algorithms Seminar, 1993--1994, INRIA Research
Report, pp. 41--44. --
1994.
- [11]
-
van der Put (Marius) and Singer (Michael F.). --
Galois Theory of Difference Equations. --
Springer, 1997, Lecture Notes in Mathematics.
- [12]
-
Van Hoeij (Mark). --
Formal solutions and factorization of differential operators with
power series coefficients. Journal of Symbolic Computation, vol. 24,
n°1, 1997, pp. 1--30.
- [13]
-
Weil (Jacques-Arthur). --
Absolute factorization of differential operators. In Salvy (Bruno)
(editor), Algorithms Seminar, 1996--1997, INRIA Research Report,
pp. 33--36. --
1997.
This document was translated from LATEX by
HEVEA.