25.07.2013 Views

Stokes versus resurgence multipliers for ordinary linear differential ...

Stokes versus resurgence multipliers for ordinary linear differential ...

Stokes versus resurgence multipliers for ordinary linear differential ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Stokes</strong> <strong>versus</strong> <strong>resurgence</strong> <strong>multipliers</strong> <strong>for</strong> <strong>ordinary</strong> <strong>linear</strong><br />

di¤erential systems with a single arbitrary level<br />

Pascal REMY<br />

October 2, 2008<br />

Abstract<br />

We provide a proof of the summability-<strong>resurgence</strong> of solutions of any <strong>ordinary</strong> <strong>linear</strong><br />

di¤erential system with a single arbitrary level r following Ecalle’s method by regular perturbation<br />

and majorant series. We deduce from it a precise description of singularities in the<br />

Borel plane with an application to the exact correspondence between <strong>resurgence</strong> <strong>multipliers</strong><br />

(or connection constants) in the Borel plane and <strong>Stokes</strong> <strong>multipliers</strong> in the Laplace plane.<br />

0 Introduction<br />

The main two “e¤ective”approaches to the meromorphic classi…cation of <strong>ordinary</strong> <strong>linear</strong> di¤erential<br />

systems at 0 are, on one hand, the theory of summation (r-summation or multi-summation) — a<br />

full set of meromorphic invariants is then given by the <strong>Stokes</strong> matrices of the system— and, on the<br />

other hand, the theory of <strong>resurgence</strong>, a full set of invariants being then given by alien derivatives.<br />

The <strong>Stokes</strong> matrices here considered are the transition matrices between asymptotic solutions on<br />

each side of the singular summation directions (anti-<strong>Stokes</strong> directions) in the space C of the initial<br />

variable x, also called the Laplace plane. The alien derivatives are de…ned as an average of various<br />

analytic continuations of the Borel trans<strong>for</strong>m of solutions around their various singularities in the<br />

so-called Borel plane. In view to their numerical calculation one has to determine the connection<br />

constants (or, <strong>resurgence</strong> <strong>multipliers</strong>) linking analytic continuations of the Borel trans<strong>for</strong>m<br />

of solutions to a given fundamental solution at the various singularities. Numerically, it appears<br />

easier and of lower cost to calculate the <strong>resurgence</strong> <strong>multipliers</strong> associated with one speci…c analytic<br />

continuation than to calculate an average of several analytic continuations. However, while <strong>Stokes</strong><br />

matrices and alien derivatives are connected via a matrix-log <strong>for</strong>mula, to come back to <strong>Stokes</strong><br />

matrices from <strong>resurgence</strong> <strong>multipliers</strong> one needs to make explicit the connection between the two.<br />

In this paper, we are given an <strong>ordinary</strong> <strong>linear</strong> di¤erential system (in short, a di¤erential system<br />

or a system) of dimension n 2 with analytic coe¢ cients at 0 in C and rank r 1<br />

r+1 dY<br />

(1) x<br />

dx<br />

= A(x)Y ; A(x) 2 Mn(Cfxg)<br />

1


together with a <strong>for</strong>mal fundamental solution<br />

eY (x) = e F (x)x L e Q(1=x)<br />

that we assume to be prepared as follows:<br />

eF (x) = In + X<br />

Fmx m = In + O(x) 2 Mn(C[[x]]) satis…es<br />

m 1<br />

where In is the identity matrix of size n.<br />

L =<br />

JM<br />

j=1<br />

jInj<br />

eF (0) = In<br />

+ Jnj where Inj is the identity matrix of size nj,<br />

Jnj =<br />

2<br />

0<br />

6<br />

6.<br />

6<br />

4.<br />

1<br />

. .. . ..<br />

.. .<br />

3<br />

0<br />

7<br />

. 7<br />

15<br />

0 0<br />

is an irreducible Jordan block of size nj (Jnj = 0 if nj = 1) and the eigenvalue j satis…es<br />

Q 1<br />

x =<br />

JM<br />

j=1<br />

qj<br />

1<br />

x<br />

rank of the system)<br />

Re( j) 2 [0; 1[<br />

Inj is a diagonal matrix with polynomial entries in 1<br />

x<br />

qj<br />

1<br />

x<br />

= aj;r<br />

x r<br />

aj;r 1<br />

xr 1<br />

aj;1<br />

x<br />

1<br />

2<br />

x C<br />

h<br />

1<br />

i<br />

x<br />

of degree r (the<br />

or qj 0. It is well known that such conditions can be ful…lled by means of algebraic gauge<br />

trans<strong>for</strong>mations (see [BJL79] <strong>for</strong> example).<br />

In addition, without loss of generality, we assume that<br />

(2) 1 = 0 and q1 0;<br />

conditions that can always be ful…lled by means of a change of variable Y = x 1 e q1(1=x) Z.<br />

Finally, the qj’s are not supposed to be distinct; however, the assumption of a single level<br />

equal to r is equivalent to the conditions<br />

(3)<br />

qj 6 0 () aj;r 6= 0 (the exact degree is r)<br />

qj q` () aj;r = a`;r:<br />

2


When r = 1, M. Loday-Richaud and myself have made explicite in [L-RR] <strong>for</strong>mulæ linking<br />

the <strong>resurgence</strong> <strong>multipliers</strong> of the Borel trans<strong>for</strong>m b F ( ) to the <strong>Stokes</strong> <strong>multipliers</strong> (those entries of<br />

<strong>Stokes</strong> matrices that are not, a priori, trivial). Such <strong>for</strong>mulæ have been ful…lled from the precise<br />

description of singularities in the Borel plane that we have derived from a proof of the summability<strong>resurgence</strong><br />

of solutions following Ecalle’s method by perturbation and majorant series.<br />

The aim of this paper is to generalize these results to System (1) with arbitrary rank r 2. We<br />

treat this case by per<strong>for</strong>ming a reduction of the rank to one. Doing so we are not given a system<br />

of pure level r = 1 because, in general, there appears other levels smaller than one. Even so, the<br />

<strong>for</strong>mal series solutions of the reduced system keep being 1-summable, the singularities appearing<br />

in the Borel plane are no more singular regular but generically irregular and the analysis gets much<br />

more complicated. Formulæ thus obtained connect the <strong>resurgence</strong> <strong>multipliers</strong> of Borel trans<strong>for</strong>ms<br />

of solutions of the reduced system and the <strong>Stokes</strong> <strong>multipliers</strong> of the initial system (1). Actually,<br />

splitting e F (x) into J column-blocks accordingly the Jordan structure of L, we can restrict<br />

ourselves, without loss of generality, to the correspondence between the <strong>resurgence</strong> and <strong>Stokes</strong><br />

<strong>multipliers</strong> associated with the …rst column-block of e F (x) that we denote by e f(x). Indeed, any of<br />

the J column-blocks of e F (x) can be positioned at the …rst place by means of a permutation P on<br />

the columns of e Y (x) 1 .<br />

The organization of this paper is as follows. In Section 1, we collect in<strong>for</strong>mations on the<br />

Borel trans<strong>for</strong>mation and summablity-<strong>resurgence</strong> needed in the sequel. In Section 2, we apply<br />

rank reduction to System (1). In Section 3, we provide a proof of the summability-<strong>resurgence</strong> of<br />

<strong>for</strong>mal solutions of the reduced system following Ecalle’s method by perturbation and majorant<br />

series. Section 4 is devoted to the analysis of singularities in the Borel plane. In Section 5, we state<br />

<strong>for</strong>mulæ linking the <strong>Stokes</strong> <strong>multipliers</strong> to the <strong>resurgence</strong> <strong>multipliers</strong>. These ones are illustrated in<br />

Section 6 with three examples.<br />

1 Preliminaries<br />

1.1 Formal Borel trans<strong>for</strong>mation<br />

One calls <strong>for</strong>mal Borel trans<strong>for</strong>mation the <strong>linear</strong> operator<br />

eB : e'(t) = X<br />

'mt m 2 tC[[t]] 7 ! b'( ) = X<br />

m 1<br />

1 The new <strong>for</strong>mal fundamental solution is then<br />

m 1<br />

'm<br />

eY (x)P = e F (x)P x P 1 LP e P 1 Q(1=x)P<br />

3<br />

m 1<br />

(m 1)!<br />

2 C[[ ]]


etween the C-vector spaces tC[[t]] and C[[ ]].<br />

Recall that this operator extends, by setting e B(1) = the Dirac distribution at the origin, into<br />

2 d<br />

an isomorphism from the di¤erential C-algebra C[[t]]; +; ; t to the di¤erential C-algebra<br />

dt<br />

C C[[ ]]; +; ; that changes <strong>ordinary</strong> product e' e into the convolution product<br />

b' b Z<br />

( ) =<br />

0<br />

b'( ) b ( )d<br />

2 d<br />

(the integral is applied separately to each term of the series) and changes derivation t<br />

dt into<br />

multiplication by . It also changes multiplication by 1<br />

d<br />

into derivation allowing thus to extend<br />

t d<br />

the isomorphism e B from the meromorphic series C[[t]][t 1 ] to C[ (k) ; k 2 N] C[[ ]].<br />

When the series e' is 1-Gevrey 2 (we denote e'(t) 2 C[[t]]1), the <strong>for</strong>mal Borel trans<strong>for</strong>m b' is a<br />

convergent series and its sum (also denoted b') is called the minor of e'. In the special case when<br />

e' is convergent (e'(t) 2 Cftg), the minor b' is an entire function that grows at most exponentially<br />

at in…nity.<br />

1.2 Summable-resurgent series<br />

A summable-resurgent series is a 1-Gevrey series e'(t) whose the analytic continuation of the minor<br />

b'( ) outside its disc of convergence lives on a Riemann surface ^<br />

Cn (b') and grows exponentially<br />

at in…nity (cf. De…nition 1.1 below). In the <strong>linear</strong> single level case we can restrict ourselves to<br />

the case when the set of singularities of the analytic continuation is (b') := f0; !1; :::; !mg where<br />

!1; :::; !m 2 C . For more general de…nitions, we refer to [E85] or [Sau06].<br />

Given > 0, smaller than the minimal distance between elements of (b'), we call -sectorial<br />

region a domain of the universal covering ^<br />

Cn (b') of Cn (b') lying at a distance at least from<br />

(b') and composed of the three following parts:<br />

an open sector of bounded width at in…nity;<br />

a neighborhood of 0 in the principal sheet, say, an open disc centered at 0;<br />

a tubular neighborhood of a piecewise-C 1 path connecting 0 to after a …nite number of<br />

turns around all or part of points of (b').<br />

2 The coe¢ cients 'm satisfy: there exists C; A > 0 such that j'mj CA m m! <strong>for</strong> all m 0.<br />

4


a -sectorial region<br />

De…nition 1.1 A 1-Gevrey series e'(t) 2 C[[t]]1 is said to be summable-resurgent with singular<br />

support (b') if it satis…es the two following conditions:<br />

i. The minor b'( ) can be analytically continued to any -sectorial region of the above type.<br />

One says then that b' is an endless continuable analytic function with singular support (b').<br />

ii. On all -sectorial region of the above type, b' grows at most exponentially.<br />

The type of exponential growth depends, in general, on the region and is not bounded when<br />

the width of goes to in…nity.<br />

De…nition 1.2<br />

1. The 1-Gevrey series e'(t) 2 C[[t]]1 is said to be resurgent with singular support (b') if it<br />

satis…es Condition (i) of De…nition 1.1.<br />

2. It is said to be summable if it satis…es Conditions (i) and (ii) in restriction to sectors of<br />

C itself (sectors in the usual meaning) issuing from 0 and avoiding points of (b') at some<br />

distance > 0.<br />

3. It is said to be summable in a direction if it satis…es Conditions (i) and (ii) in restriction<br />

to some sector of C bisecting and avoiding points of (b') at some distance > 0.<br />

Note that a summable-resurgent series is there<strong>for</strong>e both resurgent and summable but the converse<br />

is false.<br />

The set of summable series in a direction is denoted by Cftg1; . Recall that the sum s (e') of<br />

e' 2 Cftg1; is given by<br />

Z 1ei s (e')(t) := b'( )e<br />

=t<br />

d<br />

0<br />

5


2 d<br />

and the operator s is an injective morphism from the di¤erential C-algebra Cftg1; ; +; ; t<br />

dt<br />

to the di¤erential C-algebra of holomorphic functions on sectors with vertex 0, bisecting direction<br />

and larger than a half-plane ([MR89]).<br />

De…nitions 1.1 and 1.2 extend naturally to the set<br />

n o<br />

e 1<br />

(t) = p(t ) + e'(t) ; p(t) 2 C[t] ; e'(t) 2 C[[t]]1 with (b') …nite<br />

of 1-Gevrey meromorphic series with …nite singular support. Indeed, by Borel trans<strong>for</strong>mation,<br />

b ( ) = X<br />

k=0<br />

pk (k) + b'( ) ; 2 N ; pk 2 C,<br />

and the singular support of b is ( b ) = (b').<br />

In particular, if e' is summable-resurgent (resp. resurgent, summable) with singular support<br />

(b') and if m is meromorphic at 0, then the product m e' keeps being summable-resurgent (resp.<br />

resurgent, summable) with singular support (b').<br />

1.3 Extension of <strong>for</strong>mal Borel trans<strong>for</strong>mation and summation operator<br />

To hold the case of systems with single arbitrary level in full generality, it is necessary to extend<br />

the operators e B and s to <strong>for</strong>mal expansions of the <strong>for</strong>m<br />

e'(t) = X<br />

m'm(t)<br />

m 0<br />

where, <strong>for</strong> all m 0, m 2 C and 'm belongs to the algebra<br />

(4) < t ; log t ; e P (t 1=r ) ; 2 C and P (t) 2 tC[t] ; deg(P ) r 1 ><br />

We recall in this section some points of the general theory of resurgent functions that will be<br />

used in this paper. For more details, we refer to [E85] or [Sau06].<br />

Denote by<br />

C1 the Riemann surface of the logarithm, i.e., the universal covering of C of base-point 1;<br />

C1f g the set of germs of (multi<strong>for</strong>m) analytic function at 0 2 C1, i.e., the set of germs of<br />

analytic function in a domain of the <strong>for</strong>m<br />

ae i<br />

with a continuous function h : R !]0; +1[.<br />

; 0 < a < h( ) ; 2 R C1<br />

6


The set Cf g of germs of analytic function at 0 2 C identi…es itself to the subset of C1f g <strong>for</strong>med<br />

by germs of regular analytic function at 0 2 C1. The quotient<br />

C := C1f g=Cf g<br />

is called set of microfunctions at 0 and the canonical mapping from C1f g to C is denoted by<br />

“can”. The variation of 2 C is the unique ' 2 C1f g de…ned by<br />

'( ) := var( )( ) = ( ) ( e 2i )<br />

<strong>for</strong> any arbitrary representative of in C1f g, i.e., such that can( ) = .<br />

1.3.1 Resurgence constant<br />

Let '(t) 2 C1ftg be with a subexponential growth at 0, i.e., such that<br />

(5) lim jtj log (j'(t)j) = 0 uni<strong>for</strong>mly on all sectors 1 < arg(t) < 2.<br />

jtj!0<br />

The Borel trans<strong>for</strong>m<br />

B (')( ) := 1<br />

2i<br />

Z<br />

'(t)e<br />

(the path is given Figure 1) of '(t) in any direction de…nes a germ of analytic function at<br />

0 2 C1 that can be analytically continued on C1 with an exponential growth at in…nity uni<strong>for</strong>mly<br />

on all sectors 0 1 < arg( ) < 0 2. The analytic continuation b' of any B (') is called minor of '.<br />

Note that the <strong>for</strong>mal Borel trans<strong>for</strong>mation e B of Section 1.1 is obtained by applying B separately<br />

to each term of the series.<br />

The set of '(t)’s satisfying Condition (5) is called set of <strong>resurgence</strong> constants 3 and is denoted<br />

by Rc. Note that all element of the algebra (4)is a <strong>resurgence</strong> constant.<br />

Figure 1<br />

3 The terminology “constant” is justi…ed by the fact that e' having no singularity, except 0, in C1, all its alien<br />

derivatives are zero.<br />

7<br />

=t dt<br />

t 2


Given ' 2 Rc, its minor e' can be non-integrable at 0. For example, '(t) = t 1=2 has <strong>for</strong> minor<br />

b'( ) = 2 p 3=2 . Consequently, the convolution product and the Laplace trans<strong>for</strong>mation L<br />

of Section 1.1 can be no more used in general. To circumvent this di¢ culty, one proceeds as follows:<br />

<strong>for</strong> any arbitrary u near 0 2 C1, one de…nes<br />

called major of ';<br />

one denotes by<br />

u( ) := 1<br />

2i<br />

the class of majors of ' modulo Cf g;<br />

Z 1=u<br />

0<br />

'(t)e<br />

=t dt<br />

t 2<br />

['] := can( u) 2 C<br />

one extends the Borel trans<strong>for</strong>m B (') of any ' 2 Rc into<br />

Note that the minor b' is the variation of ['].<br />

B (') := ([']; b')<br />

2 C1f g<br />

Doing that, the operator B is an isomorphism from the di¤erential C-algebra 2 d<br />

Rc; +; ; t<br />

dt<br />

to the di¤erential C-algebra C1; +; ~; of microfunctions at 0 the variation of which can be<br />

analytically continued on C1 with an exponential growth at in…nity uni<strong>for</strong>mly on all sectors<br />

0<br />

1 < arg( ) < 0 2 d<br />

2. It changes derivation t<br />

product ' into the convolution product<br />

dt<br />

['] ~ [ ] := can( u )<br />

de…ned like the class modulo Cf g of the truncated convolution product<br />

( u )( ) :=<br />

Z u<br />

u<br />

into multiplication by and changes <strong>ordinary</strong><br />

( ) ( )d 2 C1f g<br />

(u arbitrarily near 0 2 C1 such that 2]0; u[ and arg( u) = arg ) of any arbitrary<br />

representative and of ['] and [ ] in C1f g respectively.<br />

The reciprocal of B is the Laplace trans<strong>for</strong>mation operator L de…ned <strong>for</strong> all 2 C1 by<br />

Z<br />

(6) L ( )(t) :=<br />

0 ;"<br />

( )e<br />

Z 1ei =t<br />

d +<br />

"e i<br />

b'( )e<br />

where ( ) 2 C1f g is any representative of , b'( ) 2 C1f g is the variation of and 0 ;" is the<br />

circle with center 0 and radius " > 0 covered once in counterclockwise from "e i( 2 ) .<br />

8<br />

=t d


For simplicity, one can compare the class ['] to a speci…c major that one denotes by maj(').<br />

One assumes besides that this major is holomorphic on C1 and grows exponentially at in…nity<br />

uni<strong>for</strong>mly on all sectors 0 1 < arg( ) < 0 2 (such a major always exists, cf. [E85]). Doing that, <strong>for</strong><br />

all '; 2 Rc,<br />

(7)<br />

(8)<br />

2 d'<br />

maj t<br />

dt<br />

= maj(')<br />

maj(' ) = maj(') ~ maj( )<br />

and the Laplace trans<strong>for</strong>m L ([']) becomes<br />

Z<br />

L (['])(t) = L (maj('))(t) =<br />

maj(')( )e<br />

where is the Hankel loop around 0 like shown in Figure 2. Moreover,<br />

L (maj(')) = ' and L (maj(') ~ maj( )) = '<br />

Figure 2<br />

Complete this section by some classical <strong>for</strong>mulae:<br />

function ' minor b' major maj(')<br />

t m with m 2 N<br />

t with 2 CnZ<br />

m 1<br />

(m 1)!<br />

1<br />

( )<br />

t m with m 2 N 0 (or (m) )<br />

9<br />

1<br />

( )(1 e 2i )<br />

m 1<br />

(m 1)!<br />

=t d<br />

log<br />

2i<br />

= 1<br />

2i ei (1 ) 1<br />

( 1) m m!<br />

2i m+1


More generally, <strong>for</strong> all p 2 N,<br />

8<br />

maj(t log p ><<br />

t)( ) =<br />

>:<br />

pX<br />

k=0<br />

p+1 X<br />

k=0<br />

k<br />

k<br />

1 log k<br />

1 log k<br />

1.3.2 Extension of <strong>for</strong>mal Borel trans<strong>for</strong>mation<br />

Given a <strong>for</strong>mal expansion<br />

(9) e' = X<br />

m 0<br />

if 2 CnN<br />

if 2 N<br />

m'm<br />

where k; k 2 C<br />

where m 2 C and 'm 2 Rc <strong>for</strong> all m 0, one <strong>for</strong>mally de…nes the minor b' and a major of e'<br />

by setting<br />

b' := X<br />

m b'm and := X<br />

mmaj('m)<br />

Denote by<br />

m 0<br />

m 0<br />

H the set of <strong>for</strong>mal expansions (9) satisfying ( ) 2 C1f g;<br />

H ;exp the set of e' 2 H, the minors b' of which can be analytically continued on unlimited<br />

sectors bisecting direction with an exponential growth of order one at in…nity.<br />

The <strong>for</strong>mal Borel trans<strong>for</strong>mation e B of Section 1.1 extends to H by setting<br />

eB(e') = ([e']; b')<br />

where [e'] = can( ) and b' = var([e']). Note that, thus extended, e B can be seen like a <strong>linear</strong><br />

operator from the di¤erential C-algebra 2 d<br />

H; +; ; t<br />

dt<br />

to the di¤erential C-algebra C; +; ~; .<br />

2 d<br />

In particular, H ;exp; +; ; t<br />

dt<br />

is a di¤erential C-algebra. Note also that one has the inclusions<br />

Rc H ;exp H<br />

As previously, one compares the class [e'] to a speci…c major that one denotes by maj(e').<br />

In the case when e' 2 H ;exp, one de…nes the Laplace trans<strong>for</strong>m L ([e']) by Identity (6); it is<br />

an analytic function on a sector with vertex 0, bisecting direction and opening larger than a<br />

half-plane. The Laplace trans<strong>for</strong>m L thus de…ned is compatible with the convolution product ~:<br />

L (maj(e') ~ maj( e )) = L (maj(e'))L (maj( e ))<br />

10


Note that if one assumes that maj(e') can be analytically continued on unlimited sectors bisecting<br />

directions and 2 with an exponential growth of order one at in…nity (one can always choose<br />

it like this, cf. [E85]), one has<br />

Z<br />

L ([e'])(t) = L (maj(e'))(t) = maj(e')( )e<br />

=t<br />

d<br />

Complete this section by a proposition that will be used in the study of singularities and that<br />

describes the major maj(e') in the special case when e' reads e (t)t log p t with e (t) 2 C[[t]]1, 2 C<br />

and p 2 N.<br />

Proposition 1.3 Given e (t) 2 C[[t]]1 a 1-Gevrey series, 2 C and p 2 N,<br />

1. e (t)t log p t 2 H;<br />

2. one can choose <strong>for</strong> major<br />

8<br />

maj( e (t)t log p ><<br />

t)( ) =<br />

>:<br />

pX<br />

k=0<br />

p+1 X<br />

k=1<br />

qX<br />

k( )<br />

`=0 k=0<br />

k( ) log k<br />

pX<br />

1 log k<br />

`;k +` 1 log k +<br />

p+1 X<br />

k=1<br />

k( ) log k<br />

where k( ) is analytic at 0 and `;k is constant <strong>for</strong> all ` and k.<br />

if 2 CnZ<br />

if 2 N<br />

if = q 2 N<br />

Moreover, if e is summable-resurgent with singular support ( b ), then the analytic functions k<br />

are endless continuable with singular support ( b ) and grow exponentially on all -sectorial regions<br />

.<br />

1.3.3 Application to the summation<br />

Let<br />

e'(t) = X<br />

m 0<br />

mt m 2 C[[t]]1<br />

be a 1-Gevrey series. The minor b' and the major maj(e') being de…ned by<br />

8<br />

b'( ) =<br />

><<br />

0 +<br />

>:<br />

X<br />

m 1<br />

m<br />

m<br />

(m<br />

1<br />

1)!<br />

2 Cf g<br />

maj(e')( ) = 0<br />

2i +<br />

X m<br />

m<br />

(m<br />

!<br />

1 log<br />

1)! 2i<br />

2 C1f g<br />

m 1<br />

11


it results from Section 1.3.2 the following elementary lemma:<br />

Lemma 1.4<br />

1. e'(t) 2 Cftg1; () e'(t) 2 H ;exp.<br />

2. If e'(t) 2 Cftg1; , then the sum s (e') of e' in the direction<br />

Z<br />

is given by<br />

s (e')(t) = maj(e')( )e<br />

=t<br />

d<br />

We deduce from it the two propositions below that will be used to state the connections between<br />

<strong>Stokes</strong> and <strong>resurgence</strong> <strong>multipliers</strong>:<br />

Proposition 1.5 Let e'(t) 2 Cftg1; be a summable series in the direction and 2 Rc.<br />

1. e' 2 H ;exp.<br />

2. L (maj(e' )) = s (e') .<br />

Proof. The …rst point is straight<strong>for</strong>ward from Lemma 1.4, point 1, and the structure of algebra<br />

of H ;exp. As <strong>for</strong> the second point, it results from identities<br />

and Lemma 1.4, point 2.<br />

L (maj(e' )) = L (maj(e') ~ maj( )) = L (maj(e'))L (maj( ))<br />

Proposition 1.6 Let 2 C be and e'(t) 2 C[[t]]1 a 1-Gevrey series.<br />

If e'(t)t 2 H ;exp <strong>for</strong> any direction , then e' is summable in the direction and<br />

L (maj(e'(t)t )) = s (e')t<br />

Proof. Since t 2 Rc H ;exp, the structure of algebra of H ;exp implies e' 2 H ;exp. The series<br />

e' is there<strong>for</strong>e summable in the direction (Lemma 1.4) and we conclude by Proposition 1.5.<br />

2 Rank reduction<br />

Denote by<br />

A(x) = X<br />

Amx m<br />

m 0<br />

and, <strong>for</strong> k = 0; :::; r 1, A k (x) = X<br />

Fix, once and <strong>for</strong> all, a r-th root x = t 1=r of t and denote by<br />

:= exp 2i<br />

r<br />

12<br />

m 0<br />

Ak+mrx m


The r-reduced system of System (1) reads ([L-R01], [Lu72], [Tu63])<br />

2 dY<br />

(10) rt<br />

dt<br />

= A(t)Y<br />

where A(t) is the blocked matrix of size rn rn de…ned by<br />

2<br />

6<br />

A(t) = 6<br />

4<br />

A 0 (t) tA r 1 (t) tA 1 (t)<br />

A 1 (t) A 0 (t)<br />

.<br />

. ..<br />

.. .<br />

. ..<br />

. ..<br />

.<br />

.. . 0 A (t) r tA 1 (t)<br />

Ar 1 (t) A1 (t) A0 (t)<br />

To e Y (x) corresponds the <strong>for</strong>mal fundamental solution<br />

>:<br />

eY (t) = e F (t) e Y 0(t)<br />

of System (10) de…ned by<br />

8 2<br />

eF<br />

6<br />

eF ><<br />

(t) = 6<br />

4<br />

0 (t) t e F r 1 (t) t e F 1 eF<br />

(t)<br />

1 (t)<br />

.<br />

.<br />

Fe 0 (t)<br />

. ..<br />

. ..<br />

. ..<br />

. ..<br />

. ..<br />

e0 F (t)<br />

.<br />

.<br />

tFe r 1 (t)<br />

eF r 1 3<br />

7<br />

5<br />

(t) Fe 1 (t) Fe 0 (t)<br />

and<br />

8<br />

><<br />

>:<br />

2<br />

6<br />

eY<br />

6<br />

0(t) = 6<br />

4<br />

t<br />

Qk (t) := Q<br />

t<br />

eF k (t) = X<br />

m 0<br />

Fk+mrt m ; k = 0; :::; r 1<br />

.<br />

.<br />

3<br />

7<br />

5<br />

Mr<br />

1<br />

ktIn<br />

t L<br />

r e Q0(t) t L<br />

r L e Q1(t) t L<br />

r (r 1)L e Qr 1(t)<br />

L In<br />

r e Q0(t) t<br />

.<br />

L (r 1)In<br />

r e Q0(t) t<br />

k=0<br />

L In<br />

L In<br />

L In Q1(t) (r 1)(L In) Qr r e t r e 1(t)<br />

.<br />

. ..<br />

L (r 1)In<br />

r L In e Q1(t) t<br />

1<br />

k t 1=r ; k = 0; :::; r 1<br />

.<br />

L (r 1)In<br />

r (r 1)(L (r 1)In) e Qr 1(t)<br />

Note that e F (0) is not equal to Irn in general, but is always invertible since it is the lower triangular<br />

blocked matrix<br />

2<br />

In<br />

6<br />

eF<br />

6 F1<br />

(0) = 6<br />

4 .<br />

0<br />

. ..<br />

. ..<br />

. ..<br />

. ..<br />

3<br />

0<br />

7<br />

. 7<br />

0 5<br />

Fr 1 F1 In<br />

13<br />

3<br />

7<br />

5


A normal <strong>for</strong>m of System (10) reads<br />

2 dY<br />

rt<br />

dt<br />

= A0(t)Y<br />

where A0(t) is the blocked matrix of size rn rn de…ned by<br />

2<br />

rar + tL<br />

6<br />

6(r<br />

1)ar 1<br />

6<br />

A0(t) = 6 .<br />

6<br />

4 .<br />

ta1<br />

rar + tL<br />

. ..<br />

tIn<br />

.. .<br />

. ..<br />

.. .<br />

. ..<br />

rar + tL (r 2)tIn<br />

(r 1)tar<br />

.<br />

.<br />

ta1<br />

1<br />

with, <strong>for</strong> all k = 1; :::; r,<br />

a1 (r 1)ar 1 rar + tL (r 1)tIn<br />

ak =<br />

JM<br />

j=1<br />

aj;kInj<br />

The <strong>for</strong>mal series e F (t) is there<strong>for</strong>e uniquely determined by the transition system<br />

2 dF<br />

(11) rt<br />

dt<br />

= A(t)F F A0(t)<br />

jointly with the initial condition F (0) = e F (0) ([BJL79]).<br />

In the sequel, we are only interested in the column-block e f <strong>for</strong>med by the …rst n1 columns of<br />

eF . Recall that the aim of this paper is to make explicit connections between the <strong>Stokes</strong> <strong>multipliers</strong><br />

of the initial System (1) associated with e f and the <strong>resurgence</strong> <strong>multipliers</strong> of the <strong>for</strong>mal Borel<br />

trans<strong>for</strong>m b f of e f.<br />

3 Summability-<strong>resurgence</strong> of e f<br />

The main purpose of this section is to prove the following result:<br />

Theorem 3.1 The <strong>for</strong>mal series e f is summable-resurgent with singular support the points aj;r,<br />

j = 1; :::; J (leading terms of polynomials qj’s).<br />

Theorem 3.1 is proved below following Ecalle’s method by regular perturbation and majorant<br />

series quoted in [E85]. Such a proof having already been detailed in the case of systems with single<br />

level r = 1, we only give the main points and we refer to [L-RR] <strong>for</strong> more details.<br />

14<br />

3<br />

7<br />

5


3.1 Prepared system<br />

In order to simplify the proof of Theorem 3.1, we prepare System (10) by means of a suitably<br />

polynomial gauge trans<strong>for</strong>mation. Recall that such a gauge trans<strong>for</strong>mation does not a¤ect the<br />

conclusion of Theorem 3.1.<br />

Let<br />

P (x) = In + P1x + ::: + P2r 1x 2r 1 2 Mn(C[x])<br />

be satisfying P (x) e F (x) = In + O(x 2r ). Note that the gauge trans<strong>for</strong>mation Y 7 ! P (x)Y allows<br />

to write the matrix A(x) in the <strong>for</strong>m<br />

A(x) =<br />

rX<br />

k=1<br />

kakx r k + x r L + B(x) where B(x) 2 Mn(x 2r Cfxg)<br />

More precisely, splitting the matrix B(x) = Bj;` (x) into blocks accordingly the Jordan structure<br />

of L, the assumption of single level (Conditions (3)) asserts that<br />

(12) B j;` 8<br />

< O(x<br />

(x) =<br />

:<br />

2r ) if aj;r 6= a`;r<br />

O(x3r ) if aj;r = a`;r<br />

By rank reduction, the corresponding gauge trans<strong>for</strong>mation Y 7 ! P (t)Y with<br />

2<br />

6<br />

(13) P (t) := 6<br />

4<br />

allows to assume that<br />

(14)<br />

In + tPr tPr 1 + t 2 P2r 1 tP1 + t 2 Pr+1<br />

P1 + tPr+1<br />

and to write A(t) in the <strong>for</strong>m<br />

.<br />

.<br />

In + tPr<br />

. ..<br />

. ..<br />

. ..<br />

. ..<br />

.. . In + tPr tPr 1 + t 2 P2r 1<br />

Pr 1 + tP2r 1 P1 + tPr+1 In + tPr<br />

e F (t) = Irn + O(t 2 )<br />

(15) A(t) = A0(t) + B(t) where B(t) 2 Mrn(t 2 Cftg)<br />

More precisely, splitting successively the matrix B(t) into r2 blocks B (u;v) (t) of size n n, then<br />

each B (u;v) h<br />

(t) = B (u;v)j;` i<br />

(t) accordingly to the Jordan structure of L, Normalizations (12) imply<br />

(16) B (u;v)j;` 8<br />

O(t<br />

><<br />

(t) =<br />

2 ) if 1 v u r and aj;r 6= a`;r<br />

O(t4 ) if 1 u < v r and aj;r = a`;r<br />

>:<br />

O(t 3 ) if not<br />

15<br />

.<br />

.<br />

3<br />

7<br />

5


Under Assumptions (14) and (15), e f is uniquely determined by the …rst n1 columns of System<br />

(11) jointly with the initial condition e f(0) = Irn;n1 (the …rst n1 columns of the identity matrix Irn<br />

of size rn). Hence, the system<br />

2 df<br />

(17) rt<br />

dt A0(t)f + tfJn1 = Bf<br />

since 1 = 0 and a1;k = 0 <strong>for</strong> all k 2 f1; :::; rg (Conditions (2)).<br />

On the other hand, we adopt from now on the following notations: given a matrix M of size<br />

rn p with any p 2 N , we split successively M into r row-blocks M u; accordingly the block<br />

structure of A0, then each M u; into J row-blocks M u;j; accordingly the Jordan structure of L.<br />

The notation M u;j;`; refers to the `-th row in M u;j; ; the notation M ;q and M u;j;q to the q-th<br />

column in M and M u;j; respectively. M u;j;`;q denotes the entry of M at row ` in multi-row-blocks<br />

(u; j) and column q.<br />

3.2 Proof of Theorem 3.1<br />

Following Ecalle, we consider, instead of System (17), the regularly perturbed system<br />

2 df<br />

(18) rt<br />

dt A0(t)f + tfJn1 = Bf<br />

ful…lled by substituting B <strong>for</strong> B. An identi…cation of equal powers of shows that this system<br />

admits a unique <strong>for</strong>mal series solution of the <strong>for</strong>m<br />

ef(t; ) = X<br />

ef m(t) m<br />

m 0<br />

satisfying e f 0(t) = Irn;n1 and e f m(t) 2 Mrn;n1(t2C[[t]]) <strong>for</strong> all m 1. The proof proceeds by<br />

induction on m and is similar to the one detailed in [L-RR, Section 2]. In particular, <strong>for</strong> …xed<br />

j 2 f1; :::; Jg and m 1, System (18) shows that the system satis…ed by the e f u;j;`;q<br />

m ’s reads, <strong>for</strong><br />

u = 1; :::; r, ` = 1; :::; nj and q = 1; :::; n1,<br />

(19) rt 2 d<br />

dt<br />

aj;r e f u;j;`;q<br />

m t( j u + 1) e f u;j;`;q<br />

m<br />

Xu<br />

1<br />

(r + k u)aj;r+k u e f k;j;`;q<br />

m<br />

k=1<br />

16<br />

t<br />

rX<br />

k=u+1<br />

(k u)aj;k u e f k;j;`;q<br />

= te f u;j;`+1;q<br />

m te f<br />

m<br />

u;j;`;q 1<br />

m<br />

+ B u;j;`; e f ;q<br />

m 1


It turns out then, due to Normalizations (16), that <strong>for</strong> all u, j and m 1,<br />

ef u;j;<br />

2m 1(t) = O(t 3m 1 ) and<br />

8<br />

< O(t<br />

u;j;<br />

fe 2m (t) =<br />

:<br />

3m )<br />

O(t<br />

if aj;r = 0<br />

3m+1 ) if aj;r 6= 0<br />

Thus the series e f(t; ) can be rewritten as a series in t with polynomial coe¢ cients in . Consequently,<br />

by unicity, the <strong>for</strong>mal series e f(t) corresponds to e f(t; 1). Moreover, <strong>for</strong> all and in<br />

particular <strong>for</strong> = 1, the series e f(t; ) admits a <strong>for</strong>mal Borel trans<strong>for</strong>m with respect to t de…ned<br />

by<br />

'( ; ) = Irn;n1 + X<br />

'm( ) m<br />

where 'm( ) is the <strong>for</strong>mal Borel trans<strong>for</strong>m of e f m(t) <strong>for</strong> all m 1. System (19) implies a similar<br />

system on the entries of 'm: <strong>for</strong> …xed j 2 f1; :::; Jg and m 1, the 'u;j;`;q u = 1; :::; r, ` = 1; :::; nj and q = 1; :::; n1,<br />

m ’s satisfy <strong>for</strong> all<br />

(20) r( aj;r) d'u;j;`;q m<br />

d<br />

X<br />

with ' 0 = Irn;n1. More precisely,<br />

( j u + 1 r)' u;j;`;q<br />

m<br />

u 1<br />

d'<br />

(r + k u)aj;r+k u<br />

k=1<br />

k;j;`;q<br />

m<br />

d<br />

m 1<br />

rX<br />

(k u)aj;k u'<br />

k=u+1<br />

k;j;`;q<br />

m<br />

u;j;`;q 1<br />

m 'm = ' u;j;`+1;q<br />

+ d b B u;j;`;<br />

d<br />

' ;q<br />

m 1<br />

1. when aj;r = 0, the normalization of e Y (x) and the assumption of single level (Conditions (2)<br />

et (3)) imply aj;k = 0 <strong>for</strong> all k 2 f1; :::; r 1g and System (20) reads<br />

(21) r d'u;j;`;q m<br />

d<br />

( j u + 1 r)' u;j;`;q<br />

m<br />

u;j;`;q 1<br />

m 'm = ' u;j;`+1;q<br />

2. when aj;r 6= 0, System (20) is much more involved since the polynomial<br />

qj<br />

1<br />

x<br />

:= aj;r 1<br />

x r 1 :::<br />

aj;1<br />

x<br />

is no more necessarily equal to 0. Actually, System (20) reads<br />

dZm;j;q<br />

(22) Rj<br />

d<br />

where we denote by:<br />

= SjZm;j;q + Tm;j;q Zm;j;q 1<br />

17<br />

+ d b B u;j;`;<br />

d<br />

' ;q<br />

m 1


(a) Zm;j;q the rnj-dimensional column-vector de…ned by<br />

2 3<br />

Zm;j;q :=<br />

6<br />

4<br />

'1;j;q m<br />

.<br />

' r;j;q<br />

m<br />

7<br />

5 and Zm;j;0 := 0<br />

(b) Tm;j;q the rnj-dimensional column-vector de…ned by<br />

2<br />

d<br />

6<br />

Tm;j;q := 6<br />

4<br />

b B 1;j;;<br />

d<br />

' ;q<br />

.<br />

d<br />

m 1<br />

b B r;j;;<br />

3<br />

7<br />

5<br />

and<br />

2<br />

d<br />

6<br />

T1;j;q := 6<br />

4<br />

b B 1;j;;q<br />

d.<br />

d b B r;j;;q<br />

3<br />

7<br />

5<br />

d<br />

' ;q<br />

m 1<br />

(c) Rj and Sj the rnj-dimensional square matrices de…ned by<br />

2<br />

6<br />

Rj := 6<br />

4<br />

r(<br />

(r<br />

aj;r)Inj<br />

1)aj;r 1Inj<br />

.<br />

.<br />

r(<br />

0Inj<br />

aj;r)Inj<br />

.. .<br />

. ..<br />

.. .<br />

. .. r(<br />

.. .<br />

aj;r)Inj<br />

0Inj<br />

.<br />

.<br />

0Inj<br />

and<br />

Sj :=<br />

2<br />

6<br />

4<br />

aj;1Inj (r 1)aj;r 1Inj r( aj;r)Inj<br />

Lj rInj aj;1Inj (r 1)aj;r 1Inj<br />

0Inj Lj (r + 1)Inj<br />

.<br />

.<br />

..<br />

.<br />

. ..<br />

. ..<br />

. ..<br />

.. . Lj (2r 2)Inj aj;1Inj<br />

0Inj 0Inj Lj (2r 1)Inj<br />

Note that Rj and Sj are independent of m and q.<br />

Fix now > 0 and a -sectorial region associated with ( b f) = faj;r ; j = 1; :::; Jg as above<br />

(Section 1.2). We must prove that<br />

the series b f( ) = '( ; 1) = Irn;n1 + X<br />

'm( ) is convergent at 0 and can be analytically<br />

continued to ;<br />

m 1<br />

bf( ) grows at most exponentially on at in…nity.<br />

These two properties are ful…lled by applying a technique of majorant series similar to the one<br />

detailed in [L-RR, Section 2]. Consequently, we only provide below a sketch of their proof.<br />

Due to the de…nition of the two following properties hold:<br />

18<br />

d<br />

.<br />

.<br />

3<br />

7<br />

5<br />

3<br />

7<br />

5


(25)<br />

1. There exists a constant a > 0 such that <strong>for</strong> all 2<br />

(23) jarg( )j a<br />

2. There exists a constant K > 0 such that <strong>for</strong> all 2 , there is a piecewise-C 1 -path<br />

parametrized by arc length from 0 to such that the arc length s of all 2 satis…es<br />

(24) j j s K j j<br />

(j j denotes the modulus of the projection of on C).<br />

Let us, <strong>for</strong> u = 1; :::; r and j = 1; :::; J, the perturbed system<br />

8<br />

1<br />

CK<br />

1 Re<br />

j<br />

r<br />

u 1<br />

r<br />

gu;j; = Jnjgu;j; + gu;j; Jn1<br />

><<br />

>:<br />

j<br />

r<br />

= t<br />

u 1<br />

r<br />

rX<br />

k=u+1<br />

1 t g u;j;<br />

+ 1<br />

CK Iu;j;<br />

rn;n1<br />

Xu<br />

1<br />

jaj;r+k uj gk;j; k=1<br />

2I u;j;<br />

rn;n1 Jn1 + jBju;j;<br />

t<br />

g if aj;r = 0<br />

jaj;k uj g k;j; + Jnj tgu;j; + tg u;j; Jn1 + jBj u;j; g if aj;r 6= 0<br />

where the unknown g is, like b f, a rn n1-dimensional matrix splitted in multi-row-blocks g u;j; ,<br />

C is the positive constant<br />

C := max<br />

1 j J exp (2a jIm( j)j)<br />

and jBj denotes the series B in which the coe¢ cients are replaced by their module.<br />

System (25), like System (18), admits a unique <strong>for</strong>mal series solution<br />

eg(t; ) = X<br />

eg m(t) m<br />

m 0<br />

such that eg 0(t) = Irn;n1 and eg m(t) 2 Mrn;n1(t 2 C[[t]]) <strong>for</strong> all m 1. Moreover, this series satis…es<br />

the three following properties:<br />

<strong>for</strong> all u = 1; :::; r, j = 1; :::; J and m 1,<br />

eg u;j;<br />

2m 1(t) = O(t 3m 1 ) and eg u;j;<br />

8<br />

<<br />

2m (t) =<br />

:<br />

19<br />

O(t 3m ) if aj;r = 0<br />

O(t 3m+1 ) if aj;r 6= 0;


<strong>for</strong> all m 1, eg m(t) has non-negative coe¢ cients;<br />

<strong>for</strong> all , eg(t; ) determines an analytic function in t at 0.<br />

Hence, the <strong>for</strong>mal series eg(t) = eg(t; 1) is convergent at 0 as a series in eg m(t) as well as a series in<br />

t. Consequently, its <strong>for</strong>mal Borel trans<strong>for</strong>m<br />

( ) = Irn;n1 + X<br />

( m( ) 2 Mrn;n1( R + [[ ]]) is the <strong>for</strong>mal Borel trans<strong>for</strong>m of eg m(t) <strong>for</strong> all m 1) is an entire<br />

function with exponential growth at in…nity.<br />

Theorem 3.1 results then from Lemma 3.2 below.<br />

m 1<br />

m( )<br />

Lemma 3.2 The series (K ) is a majorant series <strong>for</strong> b f( ) = Irn;n1 + X<br />

'm( ).<br />

Proof. Prove the following inequalities: <strong>for</strong> all m 1, 2 , u = 1; :::; r, j = 1; :::; J and<br />

q = 1; :::; n1,<br />

(26) ' u;j;q<br />

m ( )<br />

(27)<br />

><<br />

>:<br />

u;j;q<br />

m (s )<br />

u;j;q<br />

m (K j j)<br />

The second inequality is straigth<strong>for</strong>ward from Condition (24) and the fact that has nonnegative<br />

coe¢ cients. As <strong>for</strong> the …rst inequality, it results from System (27) below satis…ed by the<br />

entries<br />

(u;j;`);q<br />

m :<br />

8<br />

1<br />

CK<br />

1 Re<br />

j<br />

r<br />

u 1<br />

r<br />

u;j;`;q<br />

m = u;j;`+1;q<br />

m<br />

d u;j;`;q<br />

m<br />

d<br />

j<br />

r<br />

Xu<br />

1<br />

k=1<br />

u 1<br />

r<br />

jaj;r+k uj<br />

1<br />

u;j;`;q<br />

m<br />

d k;j;`;q<br />

m<br />

d<br />

=<br />

+ u;j;`;q 1<br />

+ u;j;`+1;q<br />

m +<br />

m<br />

rX<br />

k=u+1<br />

20<br />

jaj;k uj<br />

u;j;`;q 1<br />

m<br />

+ dd jBj u;j;`;<br />

d<br />

k;j;`;q<br />

m<br />

+ dd jBj u;j;`;<br />

d<br />

m 1<br />

u;j;q<br />

m<br />

;q<br />

m 1 if aj;r = 0<br />

;q<br />

m 1 if aj;r 6= 0


In the case when aj;r = 0, the solution of Equation (21) given<br />

(28) ' u;j;`;q<br />

m ( ) =<br />

1<br />

r<br />

j<br />

r<br />

u 1<br />

r<br />

Z<br />

1<br />

0<br />

d b B u;j;`;<br />

d<br />

' ;q<br />

m 1( ) + ' u;j;`+1;q<br />

m ( ) '<br />

u;j;`;q 1<br />

m<br />

( )<br />

!<br />

j u 1<br />

+ r r d ;<br />

Conditions (23) and (24) allow, by majorizing (28), to prove the …rst inequality of (26).<br />

To conclude in the case when aj;r 6= 0, we rewrite System (27) in a similar <strong>for</strong>m to System (22)<br />

and we apply Grönwall Lemma to ' u;j;`;q<br />

m ( (s)). This achieves the proof of Lemma 3.2.<br />

4 Singularities of the Borel trans<strong>for</strong>m b f<br />

The aim of this section is to describe the precise possible <strong>for</strong>m of the singularities aj;r 6= 0. Contrary<br />

to the systems with single level 1 (see [L-RR, Section 3]), these singularities are no more singular<br />

regular but generically irregular. However, the kind of “irregularity” of each singularity can be<br />

directly read on the initial <strong>for</strong>mal fundamental solution e Y (x) since it depends only on polynomials<br />

qj.<br />

Following Ramis, we call front of ! the set<br />

F r(!) := fqj ; aj;r = !g<br />

Under our assumption of single level (Conditions (3)), the front of each singularity ! of b f is a<br />

singleton:<br />

! 1<br />

1<br />

!;r 1<br />

F r(!) = + q! with q! :=<br />

xr x<br />

x xr 1 !;1<br />

:::<br />

x<br />

In the case when q! 0 (we refer to such a singularity ! as a singularity with monomial front),<br />

the singularity ! is regular. More precisely, we …nd a simple-rami…ed structure (cf. De…nition<br />

4.1 below) similar to the one already met in the case of systems with single level 1 (see Section<br />

4.1, Theorem 4.2). On the other hand, when q! 6 0, the singularity ! becomes irregular and the<br />

polynomial q! determines completely this irregularity (see Section 4.2, Theorem 4.14).<br />

In order to simplify the notations, we use, from now on, the translation operator ! in !:<br />

! b f( ) = b f( + !)<br />

21


4.1 Singularities with monomial front<br />

Let ! be such a singularity. Theorem 4.2 below shows that ! admits a simple-rami…ed structure<br />

similar to the one already met in [L-RR]. Note that this result is not surprising since, the monomial<br />

!=x r in the initial system (1) providing by rank reduction the monomial !=t, System (10) is<br />

“morally”a system with single level 1 <strong>for</strong> !.<br />

Be<strong>for</strong>e to <strong>for</strong>mulate Theorem 4.2, recall the de…nition of the simple rami…ed singularities.<br />

De…nition 4.1 (simple-rami…ed singularity)<br />

A germ ' is said to be simple-rami…ed at ! 2 C if, modulo germs of analytic function at 0, !'<br />

reads<br />

where<br />

!'( ) =<br />

!;0 X<br />

p=0<br />

!;p log p +<br />

!;0+1<br />

X<br />

h!;p( ) log p + X !; X<br />

(H!; ;p( ) log p )<br />

p=1<br />

2 ! p=0<br />

i. the <strong>resurgence</strong> coe¢ cients !;p, on the one hand, and h!;p( ) and H!; ;p( ), on the other<br />

hand, are constant and analytic at 0 respectively;<br />

ii. the set of rami…cation exponents ! is a …nite subset of CnZ in which values are distinct<br />

modulo Z;<br />

iii. the logarithmic orders ( !;0; ( !; ) 2 !) are non-negative integers.<br />

Theorem 4.2 (description of singularities with monomial front)<br />

Let ! be any of aj;r 6= 0 with monomial front and q 2 f1; :::; n1g. The q-th column b f ;q<br />

simple-rami…ed singularity at !. Modulo germs of analytic function at 0, ! b f ;q<br />

reads<br />

! b f ;q<br />

N!;0+q<br />

X<br />

( ) =<br />

p=0<br />

;q<br />

N!;0+q+1<br />

X<br />

!;p p<br />

log +<br />

p=1<br />

h ;q<br />

!;p( ) log p + X<br />

2 !<br />

N!; +q<br />

X<br />

p=0<br />

H ;q<br />

!; ;p ( ) logp 1<br />

1<br />

of b f has a<br />

where ;q<br />

!;p 2 Mrn;1(C) and h ;q<br />

!;p( ); H ;q<br />

!; ;p ( ) 2 Mrn;1(Cf g). The set of rami…cation exponents !<br />

22


and the logarithmic orders (N!;0 + q; (N!; + q) 2 !) are given by<br />

8<br />

! =<br />

j<br />

r<br />

u 1<br />

; j 2 f1; :::; Jg and u 2 f1; :::; rg such that<br />

r<br />

><<br />

>:<br />

N!;0 = max nj 2 ; j 2 f1; :::; Jg such that qj<br />

N!; = max nj 2 ; j 2 f1; :::; Jg such that qj<br />

qj<br />

1<br />

x<br />

1<br />

x<br />

1<br />

x<br />

= !<br />

x r and ( j 6= 0 or u 6= 1)<br />

= !<br />

x r and j = 0<br />

!<br />

= and<br />

xr there exists u 2 f1; :::; rg such that<br />

j<br />

r<br />

u 1<br />

r =<br />

Moreover, the <strong>resurgence</strong> coe¢ cients of b f ;q<br />

at ! are germs of endless continuable analytic function<br />

with singular support the points aj;r ! that grow exponentially at in…nity.<br />

Note that the <strong>resurgence</strong> coe¢ cients of b f ;q<br />

at ! depend on the path chosen <strong>for</strong> the analytic<br />

continuation of b f.<br />

Proof. To prove Theorem 4.2, we follow Ecalle’s method by regular perturbation quoted in [E85]<br />

(see also [CNP93] and [Sau06]). Resuming the notations of Section 3, the <strong>for</strong>mal series e f reads<br />

X<br />

(29)<br />

f(t) e 1<br />

= P (t) ef m(t)<br />

where the matrix P is de…ned by (13) and the e f m’s by (19). Thus, after a <strong>for</strong>mal Borel trans<strong>for</strong>mation,<br />

Identity (29) becomes<br />

X<br />

(30) f( b 1<br />

) = P (0) + E 'm( )<br />

where E is an entire function (since P 1 (t) is analytic at 0) and the 'm’s are de…ned by (20).<br />

Ecalle’s method consists in showing that, <strong>for</strong> …xed q 2 f1; :::; n1g, the ' ;q<br />

m ’s have the same structure<br />

at the singularity !, then this structure is transmitted to b f ;q<br />

. The arguments used in this proof<br />

are similar to those detailed in [L-RR, Section 3] <strong>for</strong> systems with single level 1. Consequently, we<br />

only give the main points.<br />

In order to study the behavior of the ' m’s near !, we need to explicitly know their entries<br />

23<br />

m 0<br />

m 0


with u = 1; :::; r, j = 1; :::; J, ` = 1; :::; nj and q = 1; :::; n1.<br />

Recall that, in the case when aj;r = 0, the 'u;j;`;q m ’s are given by<br />

' u;j;`;q<br />

m<br />

(31) ' u;j;`;q<br />

m ( ) =<br />

1<br />

r<br />

j<br />

r<br />

u 1<br />

r<br />

Z<br />

1<br />

0<br />

d b B u;j;`;<br />

d<br />

' ;q<br />

m 1( ) + ' u;j;`+1;q<br />

m ( ) '<br />

u;j;`;q 1<br />

m<br />

In the case when aj;r 6= 0, the 'u;j;`;q m ’s are determined by System (22). More precisely,<br />

( )<br />

!<br />

j u 1<br />

+ r r d<br />

when aj;r = !, the polynomial qj 2 F r(!) is the monomial !=x r (! is a singularity with<br />

monomial front). Consequently, aj;k = 0 <strong>for</strong> all k 2 f1; :::; r 1g and System (22) splits into<br />

the di¤erential equations<br />

r( aj;r) d'u;j;`;q m<br />

d<br />

( j u + 1 r)' u;j;`;q<br />

m<br />

u;j;`;q 1<br />

m 'm = ' u;j;`+1;q<br />

+ d b B u;j;`;<br />

d<br />

' ;q<br />

m 1<br />

There<strong>for</strong>e, since ' m(0) = 0, it results that, <strong>for</strong> all u = 1; :::; r, ` = 1; :::; nj and q = 1; :::; n1,<br />

(32) ' u;j;`;q<br />

m<br />

( ) = 1<br />

j<br />

( !) r<br />

r<br />

Z<br />

0<br />

d b B u;j;`;<br />

d<br />

u 1<br />

r<br />

1<br />

' ;q<br />

m 1( ) + ' u;j;`+1;q<br />

m ( ) '<br />

u;j;`;q 1<br />

m<br />

( )<br />

!<br />

( !)<br />

j u 1<br />

+ r r d<br />

when aj;r 6= !, the calculation of the 'u;j;`;q m ’s is much more involved since System (22) has<br />

generally an irregular singularity at aj;r. Denote then by Fj( ) a fundamental solution at 0<br />

of the homogeneous system of (22). Note that Fj( ) is independant of q and m, analytic at 0<br />

(since aj;r 6= 0) and can be analytically continued on the universal covering of Cnfaj;rg. Thus,<br />

solving System (22) by the method of variation of constants, it results that the column-vector<br />

2 3<br />

Zm;j;q =<br />

6<br />

4<br />

'1;j;q m<br />

.<br />

' r;j;q<br />

m<br />

reads, <strong>for</strong> all u = 1; :::; r and q = 1; :::; n1,<br />

Z<br />

(33) Zm;j;q( ) = Fj( )<br />

0<br />

7<br />

5 ; Zm;j;q(0) = 0<br />

F 1<br />

j ( )R 1<br />

j ( ) (Tm 1;j;q( ) zm;j;q 1( )) d<br />

Note that Fj( ) and R 1<br />

j ( ) are analytic at ! since aj;r 6= !.<br />

24


We are left to prove that, <strong>for</strong> …xed q 2 f1; :::; n1g,<br />

i. the ' ;q<br />

m ’s, m 1, have the same structure at !: simple-rami…ed with the same set of<br />

rami…cation exponents ! and the same logarithmic orders (N!;0 + q; (N!; + q) 2 !);<br />

ii. this structure is successively transmitted to the series P ;q<br />

m 1 ' m ( ) and b f ;q<br />

.<br />

The …rst point is proved by recursion on m 1 applying Proposition 4.3 below to Identities (31),<br />

(32) and (33). For the second point, we use the uni<strong>for</strong>m convergence of Series P<br />

m 1 ' m( ) stated<br />

in Section 3.2, then we apply Proposition 4.3 again to Identity (30).<br />

As <strong>for</strong> the properties of the <strong>resurgence</strong> coe¢ cients, they result from the summability-<strong>resurgence</strong> of<br />

ef. For more details, we refer to [L-RR, Section 3].<br />

Proposition 4.3 (stability of simple-rami…ed singularities)<br />

Let e be an entire function with valuation v at the origin.<br />

Let ' be a germ of endless continuable analytic function with a …nite singular support .<br />

Assume that ' has a simple-rami…ed singularity at ! 2 with rami…cation exponents ! and<br />

logarithmic orders ( !;0; ( !; ) 2 !).<br />

Then, the convolution product e ' admits at ! a singularity with the same type: simple-rami…ed,<br />

same rami…cation exponents ! and same logarithmic orders ( !;0; ( !; ) 2 !): modulo germs of<br />

analytic function at 0,<br />

!(e ')( ) =<br />

!;0+1<br />

X<br />

p=1<br />

k!;p( ) log p + X !; X<br />

(K!; ;p( ) log p )<br />

2 ! p=0<br />

Note that no terms logp<br />

appears. Moreover, the valuation at 0 of the <strong>resurgence</strong> coe¢ cient k!;p<br />

is equal to v and the valuation of the <strong>resurgence</strong> coe¢ cient K!; ;p is v + 1.<br />

When the matrix L is diagonal, the structure of ! is more simpler, but, contrary to systems<br />

with single level 1, there always exists non-integer powers of , included when the matrix L is<br />

trivial:<br />

Corollary 4.4 Assume that the matrix L is diagonal 4 : L = diag(0; 2; :::; n).<br />

Let ! be any of aj;r 6= 0 with monomial front. Then, modulo germs of analytic function at the<br />

origin, ! b f reads<br />

! b f( ) = ! + h!( ) log + X<br />

1<br />

H!; ( )<br />

4 To simplify the notations in this case, we do not write the index q of columns.<br />

25<br />

2 !<br />

1


where ! 2 Mrn;1(C) and h!( ); H!; ( ) 2 Mrn;1(Cf g). The set of rami…cation exponents ! is<br />

given by<br />

! =<br />

j<br />

r<br />

u 1<br />

r<br />

; j 2 f1; :::; ng and u 2 f1; :::; rg such that<br />

4.2 Singularities with non-monomial front<br />

qj<br />

1<br />

x<br />

= !<br />

x r and ( j 6= 0 or u 6= 1)<br />

Fix now a singularity ! of b f with non-monomial front. The previous method does not provide a<br />

description of ! with enough accuracy. Indeed, System (22) has now an irregular singularity at !<br />

and the calculations become more involved. Actually, we can only assert that the singularity ! is<br />

no more singular regular but irregular.<br />

Be<strong>for</strong>e to start the study of this singularity, look at a simple example.<br />

4.2.1 Example<br />

Consider the system<br />

3 dY<br />

(34) x<br />

dx<br />

= 0 0<br />

x 4 2 + x Y<br />

with single level 2 and its <strong>for</strong>mal fundamental solution e Y (x) = e F (x)e Q(1=x) de…ned by<br />

eF (x) = I2 + O(x 4 ) and Q 1<br />

x<br />

= diag 0;<br />

The …rst column b f of the <strong>for</strong>mal Borel trans<strong>for</strong>m of e F has only one singularity: ( b f) = f1g, and<br />

it is with non-monomial front:<br />

F r(1) =<br />

1<br />

x2 1<br />

x<br />

The behavior of b f at 1 is provided by explicitly calculating the analytic continuation of b f near 1.<br />

The 2-reduced system of System (34) is given by<br />

2 dY<br />

2t<br />

dt =<br />

2<br />

0<br />

6<br />

6t<br />

4<br />

0 0 0<br />

2 0<br />

2<br />

0<br />

0<br />

t<br />

t<br />

0<br />

0 1 t2 3<br />

7<br />

5<br />

2 t<br />

Y<br />

The <strong>for</strong>mal series e f is there<strong>for</strong>e uniquely determined by the system<br />

2 df<br />

2t<br />

dt =<br />

2<br />

0<br />

6<br />

6t<br />

4<br />

0 0 0<br />

2 0<br />

2<br />

0<br />

0<br />

t<br />

t<br />

0<br />

0 1 t2 3<br />

7<br />

5<br />

2 t<br />

f<br />

26<br />

1<br />

x 2<br />

1<br />

x


jointly the initial condition e 2 3<br />

1<br />

6<br />

f(0) = I4;1 = 60<br />

7<br />

405<br />

0<br />

(cf. System (17)). Consequently, e f reads<br />

2<br />

1<br />

6<br />

ef(t)<br />

6ef<br />

= 6<br />

4<br />

1;2<br />

(t)<br />

0<br />

ef 2;2<br />

3<br />

7<br />

5<br />

(t)<br />

(since L = 0, we omit the index q of columns) where the <strong>for</strong>mal series e f 1;2<br />

and e f 2;2<br />

ef<br />

satisfy<br />

1;2<br />

(t) = e f 2;2<br />

(t) = O(t2 ) and the di¤erential system<br />

8<br />

><<br />

>:<br />

2t 2 de f 1;2<br />

dt<br />

2t 2 de f 2;2<br />

dt<br />

2 e f 1;2<br />

(2 t) e f 2;2<br />

te f 2;2<br />

= t2 ef 1;2<br />

By <strong>for</strong>mal Borel trans<strong>for</strong>mation, it results that b f reads<br />

2 3<br />

1<br />

6<br />

bf( ) = 60<br />

7<br />

405<br />

0<br />

+<br />

2<br />

0<br />

6<br />

6bf<br />

6<br />

4<br />

1;2<br />

( )<br />

0<br />

bf 2;2<br />

3<br />

7<br />

5<br />

( )<br />

where b f 1;2<br />

and b f 2;2<br />

satisfy b f 1;2<br />

( ) = b f 2;2<br />

( ) = O( ) and<br />

8<br />

>< 2( 1) b f 1;2<br />

1 b f 2;2<br />

=<br />

>:<br />

2( 1) b f 2;2<br />

+ 1 b f 2;2<br />

In particular, b f 1;2<br />

is the unique analytic solution at 0 of the di¤erential equation<br />

(35)<br />

8<br />

><<br />

2 dy<br />

4( 1) + (6<br />

d<br />

>:<br />

y(0) = 0<br />

7) y = 3 2<br />

Choose a determination of log such that log is a real number when > 0 and solve (35): <strong>for</strong><br />

all 2 [0; 1[,<br />

bf 1;2<br />

( ) = (1 ) 3 1<br />

2 e 4(1<br />

Z<br />

)<br />

3<br />

p<br />

1<br />

2<br />

e<br />

1<br />

4(1 ) d<br />

27<br />

0<br />

bf 1;2<br />

= 0<br />

= 0


where the path (0; ) is the segment [0; ]. More precisely, in terms of special functions, b f 1;2<br />

reads<br />

bf 1;2<br />

( ) = 3<br />

0<br />

@<br />

2<br />

2<br />

1<br />

where<br />

+ 1F1<br />

1 3 ; 2 2 ;<br />

1<br />

4(1 )<br />

(1 ) 2 e<br />

1F1<br />

1<br />

2<br />

1<br />

4(1 )<br />

1<br />

A<br />

+ e 1<br />

4 + 3<br />

2 1F1<br />

+1X<br />

3<br />

; ; =<br />

2<br />

n=0<br />

n<br />

n!(2n + 1)<br />

is the con‡uent hypergeometric function with parameters 1<br />

2<br />

1F1<br />

1 3<br />

;<br />

2 2 ;<br />

and 3<br />

2 .<br />

1<br />

4<br />

(1 ) 3<br />

2 e<br />

There<strong>for</strong>e, b f admits an irregular singularity at 1. Concerning the irregular parts e<br />

1 3 ; 2 2 ;<br />

1<br />

4(1 ) + 2<br />

1<br />

4(1 ) and<br />

1<br />

4(1 ) , we will show in Section 4.2.3, Proposition 4.15, that they result from the<br />

rami…ed exponentials e<br />

t 1=2<br />

of e Y ful…lled from the exponential e 1=x during the rank reduction.<br />

With this simple example, we see clearly that the structure of singularities with non-monomial<br />

front is generically more involved since irregular parts can occur. In particular, a study of System<br />

(22) similar to the one of Section 4.1 can not provide such a structure with enough accuracy. To<br />

circumvent this di¢ culty, we use an alternative approach based on microfunctions (or majors).<br />

4.2.2 Equations of exponential type and microsolutions<br />

Recall that the set C of microfunctions at 0 is the quotient<br />

C := C1f g=Cf g<br />

The canonical mapping from C1f g in C is denoted by “can”and “var”is the variation. For all<br />

2 C, we denote by C the translated of C in .<br />

Denote by @t := d<br />

dt the derivative with respect to t. Let p = an(t)@ n t + ::: + a1(t)@t + a0(t) be an<br />

equation of exponential type (the coe¢ cients aj are polynomials and the levels of p are not greater<br />

than 1). Its Borel trans<strong>for</strong>m bp = bm( )@ m + ::: + b1( )@ + b0( ) is also an equation of exponential<br />

type and the degree of bm( ) is equal to n (the order of p).<br />

Let 1; :::; s be the singular points of bp, i.e., the zeros of bm, and m1; :::; ms their multiplicity<br />

order. The space of microsolutions of bp in j, i.e., the space of 2 C j satisfying bp = 0 in C j , is<br />

mj-dimensional ([Ka83]).<br />

Let be now a direction such that the half-lines d j; issuying from j in the direction are<br />

distinct. If is a microsolution of bp in j, then var( ) = b' is a complete solution of bpby = 0<br />

28


and can be analytically continued with an exponential growth of order one at in…nity along d j; .<br />

Thus, the Laplace trans<strong>for</strong>m L j; ( ) (= the translated in j of (6)) of is well de…ned and it is<br />

a solution of py = 0.<br />

Consequently, the solutions of py = 0 can be obtained in this way:<br />

Theorem 4.5 ([E85], [Mal91]) For …xed j 2 f1; :::; sg, denote by ( j;k)1 k mj a basis of microsolutions<br />

of bp in j.<br />

Then, the set L j; ( j;k)’s, j = 1; :::; s and k = 1; :::; mj, is a basis of solutions of py = 0.<br />

The Laplace trans<strong>for</strong>mation de…nes an isomorphism between the set of microsolutions of bp and<br />

the set of solutions of p. We use partially this result as follows:<br />

Corollary 4.6 Let (e'1; :::; e'n) be a basis of <strong>for</strong>mal solutions of p:<br />

e'j = e je j=t with j 2 C and e j 2 t j e P(t 1=k j )C[[t]][log t]<br />

with j 2 C and Pj(t) belonging to tC[t] with degree not greater than kj 1.<br />

For all j 2 f1; :::; ng, choose a major maj( e j) 5 of e j and de…ne b j by<br />

b j := j maj( e j)<br />

1. b j is a major of e'j and it de…nes a multi<strong>for</strong>m analytic function in j.<br />

2. A basis of microsolutions of bp is given by (can( b 1); :::; can( b n)).<br />

4.2.3 Description of the singularity !<br />

In order to apply Corollary 4.6, we reduce System (10) into an equation of exponential type using<br />

Cyclic Vector Theorem ([D70]) and Birkho¤ algebraization Theorem ([Si90]).<br />

Let Y = MZ be, M(t) 2 GLrn(Cftg[t 1 ]), a meromorphic gauge trans<strong>for</strong>mation that trans<strong>for</strong>ms<br />

System (10) into the companion system of a <strong>linear</strong> di¤erential equation (E) of order rn with<br />

polynomial coe¢ cients. A <strong>for</strong>mal fundamental solution of (E) is<br />

eZ(t) = e G(t) e Y 0(t) where e G(t) = M 1 (t) e F (t)<br />

Thus, the levels of (E) are not greater than 1 and, consequently, (E) is an equation of exponential<br />

type. Note besides that, M(t) being a meromorphic at 0, e G keeps being summable-resurgent and<br />

the singularities of the q-th column of its minor b G are those of the q-th column of b F .<br />

5 maj( e j) is a function of …nite determination.<br />

29


Denote now by e G1;1 the …rst n1 columns of e G. The study of the analytic continuation of b G1;1<br />

near ! allows us to describe the structure of the singularity ! as singularity of b f. Indeed, the<br />

identity e f = M e G1;1 implies on the Borel trans<strong>for</strong>ms the identity<br />

(36) b f = c M b G1;1<br />

and c M reads<br />

cM =<br />

vX<br />

k=0<br />

mk (k) + be<br />

where mk 2 Mrn(C) <strong>for</strong> all k = 0; :::; v and be is an entire function.<br />

To study the analytic continuation of b G1;1 near !, we proceed as follows:<br />

i. we derive from the <strong>for</strong>mal fundamental solution e Z(t) a basis of <strong>for</strong>mal solutions of (E);<br />

ii. applying Corollary 4.6 to this <strong>for</strong>mal basis, we build a basis of microsolutions at ! of the<br />

Borel trans<strong>for</strong>m equation ( b E) of (E);<br />

iii. we interpret the analytic continuation of b G1;1 near ! from these microsolutions.<br />

>:<br />

i. A basis a <strong>for</strong>mal solutions of (E)<br />

Recall that e Y 0(t) is the blocked matrix of size rn rn de…ned by<br />

8 2<br />

6<br />

eY<br />

6<br />

><<br />

0(t) = 6<br />

4<br />

t L<br />

r eQ0(t) t L<br />

r LeQ1(t) t L<br />

L In<br />

t r e<br />

(r 1)L Qr r e 1(t)<br />

Q0(t)<br />

.<br />

L In<br />

t r L In Q1(t) e<br />

.<br />

. ..<br />

L In<br />

(r 1)(L In) Qr t r e 1(t)<br />

.<br />

L<br />

t<br />

(r 1)In<br />

r eQ0(t) 3<br />

7<br />

5<br />

L (r 1)In<br />

t r L In Q1(t) e<br />

L (r 1)In<br />

t r (r 1)(L (r 1)In) Qr e 1(t)<br />

Qk (t) := Q<br />

where = e 2i =r .<br />

Split successively e h<br />

G = eG1<br />

structure of e Y 0, then each e Gu =<br />

1<br />

k t 1=r ; k = 0; :::; r 1<br />

h eGu;1<br />

i<br />

eGr into r column-blocks of size rn n accordingly the block<br />

i<br />

into J column-blocks accordingly the Jordan<br />

eGu;J<br />

structure of L. Denote also by eg u;j;q the …rst entry of the q-th column of e Gu;j. Note that, <strong>for</strong><br />

…xed j 2 f1; :::; Jg, the eg u;j;q’s, u = 1; :::; r and q = 1; :::; nj, are summable-resurgent with singular<br />

support the points ak;r aj;r, k = 1; :::; J.<br />

A basis of <strong>for</strong>mal solutions of (E) is de…ned by<br />

(ez1; :::; ezr)<br />

30


where, <strong>for</strong> u = 1; :::; r,<br />

with, <strong>for</strong> all j 2 f1; :::; Jg and q 2 f1; :::; njg,<br />

8<br />

qX<br />

ezu;j;q(t) =<br />

><<br />

`=1<br />

>:<br />

and<br />

ezu = (ezu;1;1; :::; ezu;1;n1; ezu;2;1; :::; ezu;J;1; :::; ezu;J;nJ )<br />

ehu;j;`(t)t j=r (u 1) j logq ` ( u 1t1=r )<br />

e<br />

(q `)!<br />

qj( u 1t 1=r )<br />

ehu;j;`(t) = eg 1;j;`(t) + ( u 1 t 1=r ) 1 eg 2;j;`(t) + ::: + ( u 1 t 1=r ) (r 1) eg r;j;`(t)<br />

ii. A basis of microsolutions of ( b E)<br />

The singularities of ( b E) are located at the points = aj;r, j = 1; :::; J. Since (E) is an equation<br />

of exponential type, Corollary 4.6 applies: with each <strong>for</strong>mal solution ezu;j;q is associated a<br />

microsolution of ( b E) at aj;r. It is de…ned by can( b u;j;q) where<br />

b u;j;q = maj(ezu;j;q) = aj;r maj<br />

`=1<br />

qX<br />

`=1<br />

ehu;j;`(t)t j=r (u 1) j logq ` ( u 1t1=r )<br />

e<br />

(q `)!<br />

qj( u 1t 1=r !<br />

)<br />

Lemma 4.7 For all u 2 f1; :::; rg, j 2 f1; :::; Jg and q 2 f1; :::; njg,<br />

b<br />

u;j;q = aj;r maj<br />

qX<br />

ehu;j;`(t)t j=r (u 1) j logq ` ( u 1t1=r !<br />

)<br />

~ maj<br />

(q `)!<br />

e qj( u 1t 1=r )<br />

!<br />

where<br />

i. the major<br />

Nu;j;q := maj<br />

qX<br />

`=1<br />

ehu;j;`(t)t j=r (u 1) j logq ` ( u 1t1=r !<br />

)<br />

(q `)!<br />

satis…es the two following properties called Properties (Paj;r ):<br />

(a) Nu;j;q reads<br />

Nu;j;q( ) = X<br />

where 2 C, p 2 N and ' ;p( ) 2 Cf g;<br />

…nite<br />

' ;p( ) log p<br />

(b) the ' ;p’s are endless continuable with singular support the points ak;r aj;r, k = 1; :::; J,<br />

and grow exponentially at in…nity;<br />

31


ii. the major maj e qj( u 1 t 1=r ) is a major of e qj( u 1 t 1=r ) that is holomorphic on C1 and<br />

grows exponentially at in…nity.<br />

Proof. Point (i) is straight<strong>for</strong>ward from Proposition 1.3. Indeed, <strong>for</strong> all ` 2 f1; :::; qg, e hu;j;` reads<br />

ehu;j;`(t) = eg 1;j;`(t) + ( u 1 t 1=r ) 1 eg 2;j;`(t) + ::: + ( u 1 t 1=r ) (r 1) eg r;j;`(t)<br />

and the eg v;j;`’s, v = 1; :::; r, are summable-resurgent with singular support the points ak;r aj;r,<br />

k = 1; :::; J.<br />

As <strong>for</strong> Point (ii), it results from Section 1.3.1 and the fact that the exponential e qj( u 1t 1=r ) is a<br />

<strong>resurgence</strong> constant (the degree in t of qj<br />

u 1 1=r t r 1<br />

is equal to r < 1 at the most).<br />

d<br />

Note that Properties (Paj;r ) are stable by derivation<br />

d<br />

log and , 2 C.<br />

A basis of microsolutions at any singularity of ( b E) is given by<br />

can( b u;j;q) ; 1 u r and 1 q nj<br />

and convolution by polynomials in<br />

j;aj;r=<br />

Consequently, Lemma 4.7 above allows to describe the structure of all microsolutions of ( b E) at :<br />

Lemma 4.8 Let be a microsolution of ( b E) at .<br />

Choose a representative b of in e O , i.e., such that can( b ) = .<br />

Then, the function b translated by reads, modulo germs of analytic function at 0,<br />

b =<br />

Xr<br />

1<br />

k=0<br />

k ~ maj e q!( k t 1=r )<br />

where the k’s satisfy Properties (P ) <strong>for</strong> all k = 0; :::; r 1.<br />

iii. Analytic continuation of b G1;1 near !<br />

The study of microsolutions of ( b E) at ! allows us to describe the behavior of b G1;1 near !:<br />

Proposition 4.9 Modulo germs of analytic function at 0, ! b G1;1 reads<br />

(37) ! b G1;1 =<br />

Xr<br />

1<br />

k=0<br />

k ~ maj e q!( k t 1=r )<br />

where the k’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::r 1.<br />

32


To prove this result, we proceed by recursion on columns b G1;1;q, q = 1; :::; n1, of b G1;1 starting<br />

with q = 1. The calculations being similar <strong>for</strong> all columns, we only detail the case q = 1.<br />

The …rst column e G1;1;1 of e G1;1 reads<br />

2<br />

eg 1;1;1<br />

3<br />

6<br />

eG1;1;1 = 6 .<br />

4<br />

drneg 1;1;1<br />

dtrn 7<br />

5<br />

since the …rst column of the <strong>for</strong>mal fundamental solution e Z(t) of (E) is<br />

2 3<br />

with<br />

6<br />

4<br />

ez1;1;1<br />

.<br />

d rn ez1;1;1<br />

dt rn<br />

ez1;1;1(t) = e h1;1;1(t) = eg 1;1;1(t) + t 1=r eg 2;1;1(t) + ::: + t (r 1)=r eg r;1;1(t)<br />

Identity (37) in restriction to e G1;1;1 is proved in the two following lemmas.<br />

Lemma 4.10 Modulo germs of analytic function at 0, the …rst entry !bg 1;1;1 of ! b G1;1;1 reads<br />

(38) !bg 1;1;1 =<br />

Xr<br />

1<br />

k=0<br />

7<br />

5<br />

k;1 ~ maj e q!( k t 1=r )<br />

where the k;1’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::r 1.<br />

Proof. Recall that, <strong>for</strong> all u 2 f1; :::; rg, the major maj(ezu;1;1) de…nes a microsolution of ( b E) at<br />

0. Its variation<br />

bhu;1;1 = bg u;1;1 + u 1 \<br />

t 1=r eg2;1;1 + ::: + (u 1)(r 1) \<br />

t (r 1)=r egr;1;1<br />

is there<strong>for</strong>e a complete solution of ( b E) and the analytic continuation of b hu;1;1 near ! de…nes a<br />

microsolution of ( b E) at !. Consequently, Lemma 4.8 implies a system of the following <strong>for</strong>m:<br />

modulo germs of analytic function at 0,<br />

8<br />

><<br />

>:<br />

.<br />

!bg 1;1;1+ !<br />

\<br />

t 1=r eg 2;1;1 + :::+ ! \<br />

t (r 1)=r eg r;1;1 =<br />

Xr<br />

1<br />

k=0<br />

!bg 1;1;1 + ! \<br />

t 1=r eg 2;1;1 + ::: + r 1 ! \<br />

t (r 1)=r eg r;1;1 =<br />

!bg 1;1;1 + r 1 ! \<br />

t 1=r eg 2;1;1 + ::: +<br />

(r 1)2<br />

!<br />

33<br />

k;1 ~ maj e q!( k t 1=r )<br />

Xr<br />

1<br />

k=0<br />

\<br />

t (r 1)=r eg r;1;1 =<br />

k;2 ~ maj e q!( k t 1=r )<br />

Xr<br />

1<br />

k=0<br />

k;r 1 ~ maj e q!( k t 1=r )


where the k;u’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::r 1 and u = 1; :::; r. The solution of this<br />

system achieves the proof.<br />

Lemma 4.11 Denote by b G `<br />

1;1;1 the `-th entry of b G1;1;1, ` = 2; :::; rn.<br />

Identity (38) keeps holding <strong>for</strong> all `: modulo germs of analytic function at 0,<br />

(39) ! b G `<br />

1;1;1 =<br />

Xr<br />

1<br />

k=0<br />

k;` ~ maj e q!( k t 1=r )<br />

where the k;`’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::; r 1.<br />

Proof. Recall that, <strong>for</strong> ` = 2; :::; rn,<br />

(40)<br />

b G `<br />

1;1;1 = e B d` 1 eg 1;1;1<br />

dt ` 1 = d`<br />

d `<br />

Prove Identity (39) <strong>for</strong> ` = 2. Identity (40) implies that<br />

` 1 d` 2bg 1;1;1<br />

d ` 2<br />

! b G 2 d<br />

1;1;1 = !<br />

2bg 1;1;1<br />

d 2 + ! d<br />

!<br />

2bg 1;1;1<br />

d 2 + 2 dbg 1;1;1<br />

!<br />

d<br />

Thus, applying Lemma 4.10 and the properties of derivative of the convolution product ~, ! b G 2<br />

1;1;1<br />

reads, modulo germs of analytic function at 0,<br />

(41) ! b G 2<br />

1;1;1 =<br />

Xr<br />

1<br />

k=0<br />

d 2<br />

k;1<br />

d 2 ~ maj eq!( k t 1=r ) + !<br />

Xr<br />

1<br />

k=0<br />

d 2<br />

k;1<br />

d 2 ~ maj eq!( kt 1=r )<br />

Xr<br />

1<br />

+ 2<br />

k=0<br />

d k;1<br />

d ~ maj eq!( k t 1=r )<br />

where the d k;1<br />

’s and<br />

d<br />

d2 k;1<br />

d 2 ’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::; r 1. On the other hand,<br />

multiplication by is a ~-derivation. Hence, <strong>for</strong> k = 0; :::; r 1,<br />

with<br />

d 2<br />

d 2<br />

k;1<br />

d 2 ~ maj eq!( k t 1=r ) =<br />

d 2<br />

k;1<br />

d 2 ~ maj eq!( kt 1=r ) = d2<br />

= d2<br />

= d2<br />

k;1<br />

d 2 ~maj eq!( kt 1=r ) d<br />

+ 2<br />

k;1<br />

d 2 ~ maj eq!( kt 1=r )<br />

d<br />

d<br />

d<br />

k;1<br />

d<br />

~ maj t2<br />

2 dt eq!( kt 1=r ) (cf. (7))<br />

k;1<br />

2 ~ maj<br />

k;1<br />

2 ~ maj<br />

34<br />

k<br />

r t1 1=r q 0 !<br />

k<br />

r t1 1=r q 0 !<br />

1<br />

k t 1=r<br />

e q!( k t 1=r )<br />

1<br />

k t 1=r ~ maj e q!( k t 1=r )<br />

(cf. (8))


There<strong>for</strong>e,<br />

d 2<br />

k;1<br />

d 2 ~ maj eq!( kt 1=r ) = k ~ maj e q!( kt 1=r )<br />

where k satis…es Properties (P!). Indeed, the major<br />

maj<br />

k<br />

r t1 1=r q 0 !<br />

is a polynomial in log and with 2 CnN (cf. Section 1.3.1). Identity (41) achieves the proof<br />

<strong>for</strong> ` = 2. In the case when ` 3, the calculations are similar and left to the reader.<br />

1<br />

k t 1=r<br />

Conclusion. Description of the singularity ! as singularity of b f<br />

Applying an elementary adequate generalization of Proposition 4.3, we deduce from Proposition<br />

4.9 a …rst description of singularities with non-monomial front of b f:<br />

Proposition 4.12 Let ! be any of aj;r 6= 0 with non-monomial front.<br />

Modulo germs of analytic function at 0, ! b f reads<br />

! b f( ) =<br />

Xr<br />

1<br />

k=0<br />

!;k ~ maj e q!( k t 1=r ) ( )<br />

where the !;k’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::; r 1.<br />

The following elementary result allows us to precise this description (see Theorem 4.14 below).<br />

Lemma 4.13 Denote by<br />

and de…ne the support of q! by<br />

Denote also<br />

Then,<br />

1. the % polynomials<br />

are distinct two by two;<br />

q!<br />

1<br />

x<br />

:=<br />

!;r 1<br />

x r 1 :::<br />

a!;1<br />

x<br />

supp(q!) := fk 2 f1; :::; r 1g ; !;k 6= 0g<br />

:= pgcd(j 2 supp(q!); r) and % := r 2 f2; :::; rg<br />

q!<br />

1<br />

k t 1=r ; k 2 f0; :::; % 1g<br />

35


2. <strong>for</strong> …xed k 2 f0; :::; % 1g, only the polynomials<br />

are equal to q!<br />

1<br />

k t 1=r .<br />

q!<br />

1<br />

k+`% t 1=r ; ` 2 f0; :::; 1g<br />

Theorem 4.14 Let ! be any of aj;r 6= 0 with non-monomial front.<br />

Modulo germs of analytic function at 0, ! b f reads<br />

! b f( ) =<br />

% 1 X<br />

k=0<br />

'!;k ~ maj e q!( k t 1=r ) ( )<br />

where the '!;k’s satisfy Properties (P!) <strong>for</strong> all k = 0; :::; % 1.<br />

Theorem 4.14 is su¢ cient to make explicit connections between <strong>Stokes</strong> and <strong>resurgence</strong> <strong>multipliers</strong>.<br />

However, we will improve this result in the next section (see Corollary 5.10) precising the<br />

possible <strong>for</strong>m of the '!;k’s at 0.<br />

The majors maj e q!( k t 1=r ) having an irregular singularity at 0, this one is transmitted to<br />

! b f and we …nd the result on singularities with non-monomial front announced at the beginning of<br />

Section 4. A more precise study of this irregularity requires to know explicitly these majors. Their<br />

e¤ective calculation can be per<strong>for</strong>med by means of Theorem 4.5. Indeed, it is easy to determine<br />

an equation of exponential type satis…ed by the exponential e q!( k t 1=r ) . Doing that, we prove, in<br />

the case when q! is a monomial, that the majors maj e q!( k t 1=r ) express themselves in terms of<br />

the hypergeometric functions<br />

pFq(b1; b2; :::; bp; c1; c2; :::; cq; t) = X (b1)m(b2)m:::(bp)m t<br />

(c1)m(c2)m:::(cq)m<br />

m<br />

m!<br />

or<br />

0Fq( ; c1; c2; :::; cq; t) = X<br />

m 0<br />

m 0<br />

1<br />

(c1)m(c2)m:::(cq)m<br />

t m<br />

m!<br />

; 1 p q<br />

where b1; b2; :::; bp 2 C, c1; c2; :::; cq 2 Cn( N) and, <strong>for</strong> all 2 C, ( )m is the Pochhammer symbol:<br />

( )0 = 1 and ( )m = ( + 1):::( + m 1) <strong>for</strong> all m 1<br />

Proposition 4.15 ([Re07]) Let a 2 C be, r 2 and k 2 f1; :::; r 1g such that k<br />

r<br />

36<br />

is irreducible.


at 1=2<br />

1. When r = 2, the major maj e<br />

at 1=2<br />

maj e<br />

with suitable 1; 2 2 C.<br />

( ) = 1<br />

at 1=3<br />

2. When r = 3, the majors maj e<br />

at 1=3<br />

maj e<br />

at 2=3<br />

maj e<br />

( ) = 1<br />

( ) = 1<br />

0F1<br />

1F1<br />

with suitable 1; 2; 3; 1; 2; 3 2 C.<br />

at k=r<br />

3. When r 4, the major maj e<br />

at k=r<br />

maj e<br />

( ) =<br />

Xr<br />

1<br />

j=1<br />

reads<br />

4 +<br />

0<br />

@ 2<br />

2<br />

3=2 e a 2<br />

; 2<br />

3 ; a3 27<br />

0F1<br />

+ 4=3 2<br />

5<br />

6<br />

2 4a3 ; ; 3 27 2 1F1<br />

+ 5=3 2<br />

at 2=3<br />

and maj e<br />

reads<br />

7<br />

6<br />

; 4<br />

3 ; a3 27<br />

1F1<br />

2<br />

a<br />

read<br />

5=3 + 3<br />

1 3 a2 ; ; 2 2 4<br />

0<br />

@<br />

4 4a3 ; ; 3 27 2 2F2<br />

+ 7=3 3<br />

2<br />

9 3a3<br />

2<br />

1<br />

2<br />

; 1; 1<br />

2<br />

k 1Fr 2 b1;j; b2;j; :::; bk 1;j; c1;j; c2;j; :::; cr 2;j;<br />

j<br />

1+jk=r<br />

e a2<br />

4<br />

1<br />

A<br />

1F2 1; 4 5 ; 3 3 ; a3 27<br />

2 4a3 ; ; 3 27 2<br />

k<br />

kFr 1 b1; b2; :::; bk; c1; c2; :::; cr 1; k<br />

+ r<br />

with suitable 1; 2; :::; r 2 C and suitable parameters b1;j; :::; bk 1;j; b1; :::; bk 2 C, c1;j; :::; cr 2;j;<br />

c1; :::; cr 1 2 Cn( N) and k;j; k 2 C.<br />

In the case when q! is no more a monomial, the majors maj e q!( k t 1=r ) do not seem to<br />

read in terms of special functions. For example, the exponential e t 2=3 +t 1=3<br />

di¤erential equation<br />

(216t 5<br />

81t 6 )y 000 + (1080t 4<br />

can maj e t 2=3 +t 1=3<br />

216 3 by 000 + (64 144 + 1728 2<br />

324t 5 )y 00<br />

(144t 2<br />

k;j<br />

k<br />

being solution of the<br />

930t 3 + 180t 4 )y 0 + (64 64t 21t 2 )y = 0;<br />

is a microsolution at 0 of the Borel trans<strong>for</strong>m equation<br />

81 3 )by 00<br />

(352 3090 + 324 2 )by 0 + (909 180 )by = 0<br />

All solutions of this equation are holomorphic on C1 with an irregular singularity at 0. Thus, a<br />

basis (by1; by2; by3) of solutions is also a basis of microsolutions at 0 and the major maj e t 2=3 +t 1=3<br />

reads<br />

maj e t 2=3 +t 1=3<br />

= 1by1 + 2by2 + 3by3<br />

with suitable 1; 2; 3 2 C. The solutions by1, by2 and by3 being not “classical”functions, it is the<br />

same <strong>for</strong> maj et 2=3 +t 1=3<br />

.<br />

37<br />

2<br />

1<br />

A


5 <strong>Stokes</strong> <strong>multipliers</strong> and structure of singularities<br />

5.1 <strong>Stokes</strong> matrices and rank reduction<br />

We use the terminology left-right in C with the following meaning: a point moves to the left when<br />

its argument moves counterclockwise and to the right when it moves clockwise. We choose all<br />

arguments in ] 2 ; 0]. This choice corresponds to the usual choice [0; 2 [ at in…nity. We choose<br />

also the determination of log x such that log x is a real number when x > 0.<br />

The anti-<strong>Stokes</strong> directions of System (1) associated with the …rst block of columns e f of e F<br />

are the directions of maximal decay of the exponentials e qj(1=x) with qj 6= 0. To each non-zero<br />

polynomial qj correspond the r anti-<strong>Stokes</strong> directions<br />

k := arg(aj;r)<br />

r<br />

2k<br />

r<br />

2<br />

2(k + 1)<br />

r<br />

;<br />

2k<br />

r<br />

; k = 0; :::; r 1<br />

regularly distribued around the origin x = 0.<br />

Fix such a collection of anti-<strong>Stokes</strong> directions and consider, <strong>for</strong> " > 0 small enough, the natural<br />

realizations<br />

Y k "(x) = F k "(x)x L e Q(1=x)<br />

of e Y (x) in the directions k ", where F k "(x) denotes the uniquely determined r-sum of e F (x) at<br />

k ". The <strong>Stokes</strong> matrix In + C associated with k e Y in the direction k is the unipotent matrix<br />

characterized by the relation<br />

Y = Y + (In + C )<br />

k k<br />

k<br />

where the sums Y and Y + to the right and to the left of k respectively are the analytic continu-<br />

k<br />

k<br />

ation of Y k " and Y k+" respectively to the sector with vertex 0, bisecting direction k and opening<br />

. Note that such analytic continuations exist without any ambiguity when " > 0 is small enough.<br />

r<br />

For all k = 0; :::; r 1, denote by c the …rst n1 columns of C and split c into J row-blocks<br />

k k k<br />

c j;<br />

k<br />

accordingly the Jordan structure of L. The entries of cj;<br />

k are zero as soon as eqj(1=x) is not ‡at<br />

in the direction k. We refer to such entries as trivial entries of c k . The other entries of c k are<br />

called the <strong>Stokes</strong> <strong>multipliers</strong> of b Y associated with b f in the direction k.<br />

The rank reduction allows to collect the r directions k into an only direction<br />

= r 0 2] 2 ; 0]<br />

that is an anti-<strong>Stokes</strong> direction of System (10) associated with e f. Using similar notations to those<br />

seen above, the <strong>Stokes</strong> matrix Irn + C associated with e Y in the direction is the unipotent<br />

matrix characterized by the relation (see [L-R01, Proposition 4.2])<br />

Mr<br />

1<br />

Y = Y +(Irn + C ) where C = C k<br />

38<br />

k=0


which reads also<br />

(42) s ( e F ) s +( e F ) = s +( e F )Y 0C Y 1<br />

0<br />

where s ( e F ) and s +( e F ) are the sums of e F to the right and to the left of respectively. Recall<br />

(cf. Section 1.2) that<br />

s ( e Z 1ei F )(t) := bF ( )e<br />

=t<br />

d<br />

0<br />

The aim of Section 5 is to make explicit the <strong>Stokes</strong> <strong>multipliers</strong> associated with the …rst block of<br />

columns e f of e F (initial System (1)) in terms of <strong>resurgence</strong> <strong>multipliers</strong> (the resurgent coe¢ cients that<br />

are pertinent) associated with b f (r-reduced System (10)) and, reciprocally, to make explicit the<br />

structure of singularities of b f from the <strong>for</strong>mal fundamental solution e Y and the <strong>Stokes</strong> <strong>multipliers</strong><br />

associated with e f. These <strong>for</strong>mulae allow us, as in the case of systems with single level 1 (see<br />

[L-RR]), to state the existence of a <strong>linear</strong> isomorphism between the full set of <strong>Stokes</strong> <strong>multipliers</strong> of<br />

the initial System (1) and the full set of the <strong>resurgence</strong> <strong>multipliers</strong> of the r-reduced System (10) .<br />

5.2 Main result<br />

Denote by<br />

d the half-line d := [0; 1e i [ issuing from 0 toward 2 C ;<br />

( b f) = faj;r 6= 0 ; arg(aj;r) = g the set of singularities of b f on the half-line d ;<br />

J! = fj 2 f1; :::; Jg ; aj;r = !g <strong>for</strong> all singularity ! 2 ( b f);<br />

0p q the null matrix of size p q;<br />

c j;q<br />

k<br />

the q-th column of cj;<br />

k ;<br />

= e 2i =r and its conjugate in C.<br />

The <strong>resurgence</strong> coe¢ cients of b f at any singularity ! 2 ( b f) depend on the path chosen <strong>for</strong><br />

the analytic continuation of b f. We always choose the analytic continuation of b f to the right of all<br />

singularities met along d between 0 and ! like shown in Figure 3.<br />

Figure 3<br />

39


Since the set ( b f) is …nite and e f summable-resurgent, we can apply Ecalle’s method detailed<br />

in [L-RR, Section 4] to Identity (42). We obtain thus, <strong>for</strong> all q 2 f1; :::; n1g, the identities<br />

X<br />

(43)<br />

e !=t<br />

Z<br />

! b f ;q<br />

=t 1 X<br />

( )e d = s +(<br />

r<br />

e F )M ;q<br />

! (t)e !=t<br />

!2 ( b f)<br />

+<br />

!2 ( b f)<br />

where + is the Hankel loop around 0 like shown in Figure 2 and M ;q<br />

! is the rn-dimensional<br />

column-vector de…ned, <strong>for</strong> u = 1; :::; r and j = 1; :::; J, by<br />

8<br />

0nj 1 if j =2 J!<br />

(44) M u;j;q<br />

! (t) =<br />

><<br />

>:<br />

Xr<br />

1<br />

k=0<br />

( kt1=r ) Lj (u 1)Inj c j;q j;q 1<br />

c k k<br />

log( k t 1=r )+<br />

::: + ( 1)q 1 log q 1 ( kt1=r )<br />

c<br />

(q 1)!<br />

j;1<br />

e k q!( kt 1=r ) if j 2 J!<br />

The global identity (43) splits into identities term by term as proved below:<br />

Theorem 5.1 For all singularity ! 2 ( b f), the Laplace trans<strong>for</strong>m<br />

L +( ! b f ;q<br />

Z<br />

)(t) := ! b f ;q<br />

=t<br />

( )e d<br />

of ! b f ;q<br />

is connected to the <strong>Stokes</strong> <strong>multipliers</strong> associated with e f by the relation<br />

(45) L +( ! b f ;q<br />

)(t) = 1<br />

r s +( e F )M ;q<br />

! (t)<br />

The vectors M ;q<br />

! (t) are those de…ned above.<br />

Proof. The proof is straight<strong>for</strong>ward from Lemmas 5.2 and 5.3 proved just below.<br />

Indeed, due to the description of singularities given in Theorems 4.2 and 4.14, we deduce from<br />

Lemma 5.2 an explicit <strong>for</strong>mula <strong>for</strong> the Laplace integrals L +( ! b f ;q<br />

). Precisely,<br />

when ! 2 ( b f) is with monomial front (cf. Theorem 4.2), L +( ! b f ;q<br />

) reads<br />

(46) L +( ! b f ;q<br />

N!;0+q<br />

X<br />

)(t) =<br />

p=0<br />

+<br />

X<br />

;q<br />

!;p log p N!;0+q<br />

t +<br />

p=1<br />

s +( e k ;q<br />

!;p) log p t<br />

+ X<br />

2 !<br />

N!; +q<br />

X<br />

where ;q<br />

!;p 2 Mrn;1(C), e k ;q<br />

!;p 2 Mrn;1(tCftg 1; +) and e K ;q<br />

!; ;p 2 Mrn;1(Cftg 1; +);<br />

40<br />

p=0<br />

s +( e K ;q<br />

!; ;p ) logp t t


when ! 2 ( b f) is with non-monomial front (cf. Theorem 4.14), L +( ! b f ;q<br />

) reads<br />

L +( ! b f ;q<br />

% 1 X<br />

Z<br />

)(t) =<br />

=t<br />

d<br />

!<br />

e q!( kt 1=r )<br />

k=0<br />

k=0<br />

then<br />

(47) L +( ! b f ;q<br />

0<br />

% 1 X<br />

B<br />

X<br />

)(t) = @<br />

where 2 C, s 2 N and<br />

;s<br />

…nite<br />

+<br />

' ;q<br />

!;k ( )e<br />

s +( e ;q<br />

e ;q<br />

!;k; ;s 2 Mrn;1(Cftg 1; +).<br />

1<br />

C<br />

!;k; ;s )t logs tA<br />

e q!( kt 1=r )<br />

On the other hand, it results from (44) that the matrix product 1<br />

r s +( e F )M ;q<br />

! (t) reads similarly<br />

to (46) when ! is with monomial front and similarly to (47) when ! is with non-monomial front.<br />

Indeed,<br />

when ! 2 ( b f) is with monomial front, condition q! 0 implies that all exponentials<br />

e q!( k t 1=r ) are equal to 1;<br />

when ! 2 ( b f) is with non-monomial front, Lemma 4.13 implies the decomposition<br />

(48) M ;q<br />

% 1 X<br />

! (t) = M ;q<br />

!;k (t)eq!( kt 1=r )<br />

k=0<br />

where, <strong>for</strong> all k 2 f0; :::; % 1g, M ;q<br />

u = 1; :::; r and j = 1; :::; J, by<br />

!;k is the rn-dimensional column-vector de…ned, <strong>for</strong><br />

8<br />

0nj 1 if j =2 J!<br />

M u;j;q<br />

!;k (t) =<br />

><<br />

>:<br />

X1<br />

`=0<br />

( k+`% t1=r ) Lj (u 1)Inj c j;q j;q 1<br />

c k+`%<br />

k+`% log( k+`% t 1=r )+<br />

::: + ( 1)q 1 log q 1 ( k+`% t1=r )<br />

c<br />

(q 1)!<br />

j;1<br />

k+`%<br />

if j 2 J!<br />

There<strong>for</strong>e, we deduce from Lemma 5.3 that one can identify the exponentials e !=t on both sides<br />

of identity (43).<br />

Lemma 5.2 Let e' be a 1-Gevrey series and b' its minor:<br />

b'( ) = X<br />

m 0<br />

'm m 2 Cf g<br />

41


Given 2 C and ` 2 N, we denote by<br />

e' ;`(t) := 2i X<br />

For all p 2 N, the function b'( )<br />

e (t) :=<br />

pX<br />

k=0<br />

m 0<br />

d `<br />

dz `<br />

1 log p<br />

i z e<br />

(1 z) jz=m+<br />

is a major of<br />

Moreover, if e' is summable in the direction + , then<br />

'mt m 2 C[[t]]1<br />

p<br />

k e' ;p k(t)t log k t 2 t C[[t]]1[log t]<br />

1. e' ;` is summable in the direction + <strong>for</strong> all 2 C and ` 2 N;<br />

2. <strong>for</strong> all p 2 N,<br />

Proof. Recall the identity<br />

Z<br />

(49)<br />

Z<br />

+<br />

b'( )<br />

+<br />

1 log p e<br />

1 e<br />

=t d =<br />

=t d = 2i e i<br />

and more generally, <strong>for</strong> all p 2 N, the identity<br />

(50)<br />

Z<br />

+<br />

1 log p e<br />

=t d = 2i<br />

pX<br />

k=0<br />

p<br />

k<br />

pX<br />

k=0<br />

p<br />

k s +(e' ;p k)(t)t log k t<br />

t ; 2 C<br />

(1 )<br />

p d k<br />

dzp k<br />

i z e<br />

(1 z) jz=<br />

obtained by derivating p times with respect to . The function<br />

'(t) := 2i<br />

pX<br />

k=0<br />

p<br />

k<br />

p d k<br />

dzp k<br />

being a <strong>resurgence</strong> constant, Section 1.3.1 implies that<br />

i z e<br />

(1 z) jz=<br />

1 log p<br />

Consider the case p = 0 and prove the identity<br />

Z<br />

1 =t<br />

(51)<br />

b'( ) e d = s +(e' ;0)(t)t<br />

+<br />

t log k t ; 2 C<br />

t log k t<br />

is a major of '(t).<br />

The case when 2 Z being clear by the residue theorem (e' ;0 is polynomial since 1<br />

( m)<br />

= 0 <strong>for</strong><br />

all m 2 N), we assume 2 CnZ. From (49), it results that b'( ) 1 is a major of e' ;0(t)t . Since<br />

the variation of b'( ) 1 is<br />

var b'( )<br />

1 ( ) = 1 e 2i<br />

42<br />

b'( )<br />

1


the summability of e' implies that e' ;0(t)t 2 H + ;exp. Thus, Proposition 1.6 applies given the<br />

summability of e' ;0 in the direction + and Identity (51).<br />

A recursion allows to extend Identity (51) to any integer p<br />

a major of<br />

1. Indeed, b'( ) 1 p<br />

log being<br />

pX<br />

e (t) =<br />

p<br />

k e' ;p k(t)t log k t<br />

k=0<br />

(cf. (50)), the summability of e' implies once more that e 2 H + ;exp. On the other hand, the<br />

e' ;p k’s being summable in the direction + <strong>for</strong> all k = 1; :::; p, Proposition 1.5 shows that the<br />

e' ;p k(t)t log k t’s belong also to H + ;exp <strong>for</strong> k = 1; :::; p. There<strong>for</strong>e e' ;p(t)t 2 H + ;exp and Proposition<br />

1.6 applies again.<br />

Lemma 5.3 Let m 2 be an integer and '1, ..., 'm the functions de…ned on a sector with<br />

vertex 0, bisecting direction and larger than a half-plane by<br />

pj X<br />

'j(t) :=<br />

`=1<br />

s +( e hj;`)t ` log r` t e p`(t 1=r )<br />

where e hj;` 2 Cftg 1; +[t 1 ], ( `; r`) 2 C=Z N distinct and p`(t) 2 tC[t] of degree not greater than<br />

r 1 and distinct two by two.<br />

Assume that the 'j’s satisfy on an identity of the <strong>for</strong>m<br />

'1 + '2e !2=t + ::: + 'me !m=t = 0<br />

where arg(!j) = <strong>for</strong> all j = 2; :::; m and j!2j < j!3j < ::: < j!mj.<br />

Then, 'j 0 <strong>for</strong> all j 2 f1; :::; mg.<br />

Proof. The proof is similar to the one of [L-RR] and is left to the reader.<br />

Theorem 5.1 allows to state several results on the connections between the <strong>Stokes</strong> <strong>multipliers</strong><br />

and the structure of singularities in the Borel plane. Although these results were similar <strong>for</strong> all<br />

kinds of singularities, we prefer to distinguisch the case of singularities with monomial front and<br />

the case of singularities with non-monomial front.<br />

5.3 Connections between <strong>Stokes</strong> <strong>multipliers</strong> and structure of singularities<br />

with monomial front<br />

A …rst consequence of Theorem 5.1 is to make explicit the analytic continuation of e f to the right<br />

of its singularities with monomial front according to the <strong>for</strong>mal fundamental solution e Y and the<br />

<strong>Stokes</strong> <strong>multipliers</strong> associated with e f:<br />

43


Corollary 5.4 (structure of singularities with monomial front)<br />

For all ! 2 ( b f) with monomial front, ! b f ;q<br />

is a major of the matrix product 1<br />

r e F M ;q<br />

! .<br />

Reciprocally, knowing the structure of singularities with monomial front in the Borel plane, we<br />

can determine the <strong>Stokes</strong> <strong>multipliers</strong> associated with e f and non-zero monomials qj’s:<br />

Corollary 5.5 (<strong>Stokes</strong> <strong>versus</strong> <strong>resurgence</strong> <strong>multipliers</strong>)<br />

Let q 2 f1; :::; n1g be. Given ! 2 ( b f) a singularity of b f with monomial front (cf. Theorem 4.2),<br />

we consider b f ;q<br />

the analytic continuation of the q-th column of b f to the right of ! along d .<br />

For all j 2 J!, the matrices c j;q<br />

, k = 0; :::; r 1, are the unique solutions of the blocked system<br />

(52)<br />

where<br />

2<br />

6<br />

1. Cj;q := 6<br />

4<br />

8<br />

><<br />

>:<br />

2<br />

Inj<br />

6<br />

4 .<br />

Inj<br />

Inj<br />

Cj;q;1<br />

Cj;q;2<br />

.<br />

Cj;q;r<br />

3<br />

k<br />

Lj (r 1)Lj<br />

Lj In j (r 1)(Lj In j)<br />

.<br />

.. .<br />

Lj (r 1)In j (r 1)(Lj (r 1)In j)<br />

.<br />

3 2<br />

c<br />

7 6<br />

7 6<br />

7 6<br />

5 4<br />

j;q<br />

0<br />

c j;q<br />

1<br />

.<br />

c j;q<br />

r 1<br />

7<br />

5 is the rnj-dimensional column-vector de…ned by<br />

Cj;q;1 :=<br />

Cj;q;u :=<br />

N!;0+q<br />

X<br />

p=0<br />

N<br />

j<br />

!;<br />

r<br />

X<br />

p=0<br />

3<br />

7<br />

5 = rCj;q Xj;q<br />

p(0) 1;j;q<br />

!;p if j = 0<br />

+q<br />

u 1<br />

r<br />

p<br />

j<br />

r<br />

u 1<br />

r<br />

H u;j;q<br />

!; j<br />

r<br />

u 1<br />

r ;p(0) if j 6= 0 or u 6= 1<br />

where the complex numbers p( ) are de…ned, <strong>for</strong> p 2 N and 2 C, by<br />

p( ) := 2i dp<br />

dz p<br />

44<br />

i z e<br />

(1 z) jz=


2. Xj;q is the rnj-dimensional column-vector de…ned by<br />

2<br />

where, <strong>for</strong> …xed k 2 f0; :::; r 1g,<br />

8<br />

Xj;1;k = 0<br />

><<br />

>:<br />

and, <strong>for</strong> q 2,<br />

j;q 1<br />

Xj;q;k = (k log )c<br />

6<br />

Xj;q := 6 r 1<br />

4<br />

X<br />

k=0<br />

Xr<br />

1<br />

k=0<br />

Xr<br />

1<br />

k=0<br />

kLj Xj;q;k<br />

k(Lj In j) Xj;q;k<br />

.<br />

k(Lj (r 1)In j) Xj;q;k<br />

k + ::: + ( 1)` (k log ) `<br />

`!<br />

j;q `<br />

c<br />

Proof. System (52) results from the identi…cation of the terms in t j<br />

r<br />

3<br />

7<br />

5<br />

k + ::: + ( 1)q 1 (k log )<br />

(q 1)!<br />

j 2 J!, in the identity deduced from (45) by asymptotic expansion at the origin.<br />

The uniqueness is provided by the fact that the matrix<br />

2<br />

3<br />

is invertible 6 .<br />

Inj<br />

6<br />

4 .<br />

Inj<br />

Inj<br />

Lj (r 1)Lj<br />

Lj In j (r 1)(Lj In j)<br />

.<br />

.. .<br />

Lj (r 1)In j (r 1)(Lj (r 1)In j)<br />

.<br />

7<br />

5<br />

q 1<br />

c j;1<br />

k<br />

u 1<br />

r , u 2 f1; :::; rg and<br />

These two corollaries allow to state the existence of a <strong>linear</strong> isomorphism between the <strong>Stokes</strong><br />

and <strong>resurgence</strong> <strong>multipliers</strong> associated with singularities with monomial front:<br />

Proposition 5.6 (<strong>linear</strong> isomorphism)<br />

Given ! 2 ( b f) a singularity of b f with monomial front (cf. Theorem 4.2), we consider the<br />

analytic continuation of b f to the right of ! along d .<br />

For all j 2 J!, the Laplace trans<strong>for</strong>mation L + induces a <strong>linear</strong> isomorphism between<br />

6 Its inverse is<br />

1<br />

r<br />

2<br />

6<br />

4<br />

Inj Inj Inj<br />

Lj Lj In j Lj (r 1)In j<br />

.<br />

.<br />

(r 1)Lj (r 1)(Lj In j ) (r 1)(Lj (r 1)In j )<br />

45<br />

. ..<br />

.<br />

3<br />

7<br />

5


1. (case j = 0) the entries of c j;<br />

u 2 f2; :::; rg;<br />

2. (case j 6= 0) the entries of c j;<br />

k k=0;:::;r 1<br />

k k=0;:::;r 1<br />

and the entries of<br />

and the entries of Hu;j;<br />

!; j<br />

r<br />

1;j;<br />

!;0 and H u;j;<br />

u 1<br />

!; ;0(0)<br />

with<br />

r<br />

with u 2 f1; :::; rg.<br />

u 1 ;0(0) r<br />

Proof. Corollary 5.4 implies that, <strong>for</strong> all j 2 J! and q 2 f1; :::; n1g, the entries of 1;j;q<br />

!;p (resp.<br />

H u;j;q<br />

!; j are <strong>linear</strong> combinations in C of the entries of<br />

u 1 ;p(0)) r r<br />

1;j;s<br />

!;0 (resp. H u;j;s<br />

!; j<br />

s 2 f1; :::; qg. The result follows easily by Corollary 5.5.<br />

r<br />

De…nition 5.7 The entries of the matrices<br />

are called <strong>resurgence</strong> <strong>multipliers</strong> of b f at !.<br />

1;j;<br />

!;0 and H u;j;<br />

!; j<br />

r<br />

with<br />

u 1 ;0(0)) r<br />

u 1<br />

r ;0(0), j 2 J!, of Proposition 5.6<br />

By misuse of language, we also call <strong>resurgence</strong> <strong>multipliers</strong> the entries of the matrices 1;j;<br />

!;p and<br />

H u;j;<br />

!; j<br />

r<br />

u 1<br />

r ;p(0), j 2 J! and p 0, of Corollary 5.5. The entries of their q-th column de…ne the<br />

<strong>resurgence</strong> <strong>multipliers</strong> of b f ;q<br />

at !.<br />

The case when L is diagonal deserves to be precised since, contrary to the general case, it gives<br />

explicitly the <strong>linear</strong> isomorphism between <strong>Stokes</strong> and <strong>resurgence</strong> <strong>multipliers</strong>:<br />

Corollary 5.8 Assume that the matrix L is diagonal: L = diag(0; 2; :::; n).<br />

Given ! 2 ( b f) a singularity of b f with monomial front (cf. Corollary 4.4), we consider b f the<br />

analytic continuation of b f to the right of ! along d .<br />

For all j 2 J!, the <strong>Stokes</strong> <strong>multipliers</strong> c j<br />

1;j<br />

! and H u;j<br />

!; j<br />

r<br />

u 1<br />

r<br />

1. case j = 0:<br />

c j<br />

(0) by the following <strong>for</strong>mulae:<br />

k = 2i 1;j<br />

! + 2i<br />

1;j<br />

! = 1<br />

H u;j<br />

!;<br />

Xr<br />

1<br />

c<br />

2 ir<br />

j<br />

k=0<br />

u 1 (0) =<br />

r<br />

k<br />

rX<br />

u=2<br />

, k = 0; :::; r 1, are related to the <strong>resurgence</strong> <strong>multipliers</strong><br />

k<br />

i (u 1)<br />

k(u 1) e r<br />

(1 +<br />

i<br />

u 1 (1 + )e r<br />

(u<br />

r<br />

1)<br />

Xr<br />

1<br />

2 ir<br />

k=0<br />

u 1<br />

r )Hu;j<br />

!;<br />

46<br />

k(u 1) c j<br />

u 1 (0) <strong>for</strong> all k 2 f0; :::; r 1g<br />

r<br />

k<br />

<strong>for</strong> all u 2 f2; :::; rg


2. case j 6= 0:<br />

c j<br />

k<br />

H u;j<br />

!; j<br />

r<br />

= 2i<br />

rX<br />

u=1<br />

u (0) = 1<br />

r<br />

(1<br />

k( j u+1) e i ( j u+1)<br />

r<br />

(1<br />

2 ir<br />

j u+1<br />

r<br />

) Hu;j<br />

!; j<br />

r<br />

j u+1<br />

)e r i ( j u+1)<br />

r 1<br />

r X<br />

k=0<br />

u (0) 1<br />

<strong>for</strong> all k 2 f0; :::; r 1g<br />

r<br />

k( j u+1) c j<br />

k <strong>for</strong> all u 2 f1; :::; r 1g<br />

5.4 Connections between <strong>Stokes</strong> <strong>multipliers</strong> and structure of singularities<br />

with non-monomial front<br />

To each result previously enunciated in the case of singularities with monomial front corresponds<br />

an analogous result in the case of singularities with non-monomial front. Their proofs being similar<br />

to those that have already been made, we give only their terms.<br />

Just as in the case of singularities with monomial front, Theorem 5.1 provides the structure<br />

of singularities with non-monomial front in terms of the <strong>for</strong>mal fundamental solution e Y and the<br />

<strong>Stokes</strong> <strong>multipliers</strong> associated with e f:<br />

Corollary 5.9 (structure of singularities with non-monomial front)<br />

For all ! 2 ( b f) with non-monomial front, ! b f ;q<br />

is a major of the matrix product 1<br />

r e F M ;q<br />

! .<br />

In particular, Decomposition (48) of the matrix M ;q<br />

! allows, due to Proposition 1.3, to improve<br />

Theorem 4.14 given the precise possible <strong>for</strong>m of singularities with non-monomial front:<br />

Corollary 5.10 (description of singularities with non-monomial front)<br />

Let ! be any of aj;r 6= 0 with non-monomial front and q 2 f1; :::; n1g. The q-th column b f ;q<br />

translated by ! reads, modulo germs of analytic function at 0:<br />

! b f ;q<br />

% 1 X<br />

( ) =<br />

where, <strong>for</strong> all k 2 f0; :::; % 1g, ' ;q<br />

!;k<br />

k=0<br />

' ;q<br />

!;k ~ maj<br />

rami…ed singularity at 0.<br />

Precisely, modulo germs of analytic function at 0, ' ;q<br />

!;k reads<br />

' ;q<br />

N!;0+q<br />

X<br />

!;k ( ) =<br />

p=0<br />

;q<br />

N!;0+q+1<br />

X<br />

!;k;p p<br />

log +<br />

p=1<br />

e q!<br />

1<br />

k t 1=r<br />

!!<br />

( )<br />

1<br />

is a major of<br />

r e F M ;q<br />

;q<br />

!;k . In particular, ' !;k<br />

h ;q<br />

!;k;p ( ) logp + X<br />

47<br />

2 !<br />

N!; +q<br />

X<br />

p=0<br />

H ;q<br />

!;k; ;p ( ) logp 1<br />

of b f<br />

has a simple


;q<br />

where !;k;p 2 Mrn;1(C) and h ;q<br />

;q<br />

!;k;p ( ); H !;k; ;p ( ) 2 Mrn;1(Cf g). The set of rami…cation exponents<br />

! and the logarithmic orders (N!;0 + q; (N!; + q) 2 !) are given by<br />

8<br />

j u 1<br />

! =<br />

; j 2 f1; :::; Jg and u 2 f1; :::; rg such that<br />

r r<br />

><<br />

>:<br />

N!;0 = max (nj 2 ; j 2 f1; :::; Jg such that qj 2 F r(!) and j = 0)<br />

N!; = max (nj 2 ; j 2 f1; :::; Jg such that qj 2 F r(!) and<br />

there exists u 2 f1; :::; rg such that<br />

qj 2 F r(!) and ( j 6= 0 or u 6= 1)g<br />

j<br />

r<br />

u 1<br />

r =<br />

Moreover, the functions h ;q<br />

;q<br />

!;k;p and H !;k; ;p are germs of endless continuable analytic function with<br />

singular support the points aj;r ! that grow exponentially at in…nity.<br />

;q<br />

These functions and the constants !;k;p are so-called <strong>resurgence</strong> coe¢ cients of b f ;q<br />

at !.<br />

These two corollaries allow to state results similar to Corollary 5.5 and Proposition 5.6 concerning,<br />

on the one hand, the determination of <strong>Stokes</strong> <strong>multipliers</strong> from <strong>resurgence</strong> <strong>multipliers</strong> and, on<br />

the other hand, the existence of a <strong>linear</strong> isomorphism between <strong>Stokes</strong> and <strong>resurgence</strong> <strong>multipliers</strong>:<br />

Corollary 5.11 (<strong>Stokes</strong> <strong>versus</strong> <strong>resurgence</strong> <strong>multipliers</strong>)<br />

Let q 2 f1; :::; n1g be and k 2 f0; :::; % 1g. Given ! 2 ( b f) a singularity of b f with non-monomial<br />

front (cf. Corollary 5.10), we consider b f ;q<br />

the analytic continuation of the q-th column of b f to<br />

the right of ! along d :<br />

For all j 2 J!, the matrices c j;q<br />

, ` = 0; :::; 1, are the unique solutions of the blocked system<br />

2<br />

6<br />

4<br />

where<br />

k+`%<br />

kLj (k+%)Lj (k+( 1)%)Lj<br />

k(Lj In j ) (k+%)(Lj In j ) (k+( 1)%)(Lj In j )<br />

.<br />

.<br />

k(Lj ( 1)In j ) (k+%)(Lj ( 1)In j ) (k+( 1)%)(Lj ( 1)In j )<br />

1. the matrix<br />

2<br />

6<br />

4<br />

.. .<br />

.<br />

3 2<br />

7 6<br />

7 6<br />

7 6<br />

5 4<br />

kLj (k+%)Lj (k+( 1)%)Lj<br />

k(Lj In j ) (k+%)(Lj In j ) (k+( 1)%)(Lj In j )<br />

.<br />

.<br />

. ..<br />

3<br />

c j;q<br />

k<br />

c j;q<br />

k+%<br />

.<br />

c j;q<br />

7<br />

5<br />

k+( 1)%<br />

= rCk;j;q Xk;j;q<br />

k(Lj ( 1)In j ) (k+%)(Lj ( 1)In j ) (k+( 1)%)(Lj ( 1)In j )<br />

48<br />

.<br />

3<br />

7<br />

5


is invertible with inverse<br />

2<br />

2<br />

6<br />

2. Ck;j;q := 6<br />

4<br />

8<br />

><<br />

>:<br />

1<br />

Ck;j;q;1<br />

Ck;j;q;2<br />

.<br />

Ck;j;q;<br />

6<br />

4<br />

Ck;j;q;1 :=<br />

Ck;j;q;u :=<br />

3<br />

kLj k(Lj In j ) k(Lj ( 1)In j )<br />

(k+%)Lj (k+%)(Lj In j ) (k+%)(Lj ( 1)In j )<br />

.<br />

.<br />

(k+( 1)%)Lj (k+( 1)%)(Lj In j ) (k+( 1)%)(Lj ( 1)In j )<br />

.. .<br />

7<br />

5 is the nj-dimensional column-vector de…ned by<br />

N!;0+q<br />

X<br />

p=0<br />

N<br />

j<br />

!;<br />

r<br />

X<br />

p=0<br />

p(0) 1;j;q<br />

!;k;p if j = 0<br />

+q<br />

u 1<br />

r<br />

p<br />

j<br />

r<br />

u 1<br />

r<br />

H u;j;q<br />

!;k; j<br />

r<br />

.<br />

3<br />

7<br />

5<br />

u 1<br />

r ;p(0) if j 6= 0 or u 6= 1<br />

where the complex numbers p ( ) are de…ned, <strong>for</strong> p 2 N and 2 C, by<br />

p( ) := 2i dp<br />

dz p<br />

3. Xk;j;q is the nj-dimensional column vector de…ned by<br />

2<br />

6<br />

Xk;j;q := 6 1<br />

4<br />

X<br />

where, <strong>for</strong> …xed ` 2 f0; :::; 1g,<br />

8<br />

Xk;j;1;` = 0<br />

><<br />

>:<br />

`=0<br />

X1<br />

`=0<br />

and, <strong>for</strong> q 2,<br />

q 1 X<br />

Xk;j;q;` =<br />

p=1<br />

X1<br />

`=0<br />

i z e<br />

(1 z) jz=<br />

(k+`%)Lj Xk;j;q;`<br />

(k+`%)(Lj In j) Xk;j;q;`<br />

.<br />

(k+`%)(Lj ( 1)In j) Xk;j;q;`<br />

( 1) p ((k + `%) log ) p<br />

49<br />

p!<br />

3<br />

7<br />

5<br />

j;q p<br />

c k+`%


Proposition 5.12 (<strong>linear</strong> isomorphism)<br />

Given ! 2 ( b f) a singularity of b f with non-monomial front (cf. Corollary 5.10), we consider the<br />

analytic continuation of b f to the right of ! along d .<br />

For all j 2 J! and k 2 f0; :::; % 1g, the Laplace trans<strong>for</strong>mation L + induces a <strong>linear</strong> isomorphism<br />

between<br />

1. (case j = 0) the entries of c j;<br />

u 2 f2; :::; g;<br />

k+`%<br />

2. (case j 6= 0) the entries of c j;<br />

f1; :::; g.<br />

k+`%<br />

`=0;:::; 1<br />

`=0;:::; 1<br />

and the entries of<br />

and the entries of H u;j;<br />

1;j;<br />

!;k;0 and Hu;j;<br />

!;k;<br />

!;k; j<br />

r<br />

u 1 ;0(0)<br />

with<br />

r<br />

with u 2<br />

u 1 ;0(0) r<br />

Note that these results, and especially Corollary 5.11, are rather theoretical results. It does<br />

not seem that these ones allow the e¤ective calculation of the <strong>Stokes</strong> <strong>multipliers</strong> associated with<br />

ef and singularities with non-monomial front. Indeed, the study of the analytic continuation of b f<br />

near any of its singularities ! with non-monomial front does not allow to determine, in general,<br />

the <strong>resurgence</strong> <strong>multipliers</strong> at ! (see, e.g., Section 4.2.1). To circumvent this di¢ culty, we can<br />

proceed like in [L-R90] by …rst per<strong>for</strong>ming the singularity ! into a singularity with monomial front<br />

by means of a suitable algebraic change of the variable x in the initial System (1), then applying<br />

Corollary 5.6. See Section 6.3 below <strong>for</strong> an example.<br />

6 Examples<br />

We complete this paper by three simple examples where we can calculate the exact value of <strong>Stokes</strong><br />

<strong>multipliers</strong> from the study of the analytic continuation in the Borel plane. Of course, these examples<br />

are not general, but they suitably illustrate the use of previous results.<br />

6.1 Example 1<br />

In this …rst example, the matrix L is diagonal and b f admits two col<strong>linear</strong> singularities with<br />

monomial front.<br />

Consider the system<br />

3 dY<br />

(53) x<br />

dx =<br />

2<br />

6<br />

4<br />

0 0 0<br />

x 4 x 5 2 0<br />

x 4 + x 5 0 4 + x2<br />

2<br />

50<br />

3<br />

7<br />

5 Y


with single level 2 and its <strong>for</strong>mal fundamental solution e Y (x) = e F (x)xLeQ(1=x) de…ned by<br />

2<br />

1<br />

eF (x) = 4<br />

0<br />

1<br />

0<br />

3<br />

0<br />

05<br />

= I3 + O(x<br />

1<br />

4 ) ; L = diag 0; 0; 1<br />

2<br />

and Q 1<br />

x<br />

= diag 0;<br />

1<br />

;<br />

x2 The …rst column e f of e F has two anti-<strong>Stokes</strong> direction ( = 0 and = ) determined by<br />

the directions of maximal decay of e 1=x2 and e 2=x2.<br />

The <strong>Stokes</strong> matrices I3 + C0 and I3 + C<br />

associated with = 0 and = respectively are de…ned by<br />

2<br />

0 0<br />

3<br />

0<br />

2<br />

0 0<br />

3<br />

0<br />

C0 = 4 5 and C = 4 5<br />

c 2 0 0 0<br />

c 3 0 0 0<br />

c 2 0 0<br />

c 3 0 0<br />

Indeed, only the …rst column of e F is divergent. The <strong>Stokes</strong> <strong>multipliers</strong> c 2 0 and c 2 on the one hand<br />

and c 3 0 and c 3 on the other hand are calculated from the <strong>resurgence</strong> <strong>multipliers</strong> of b f at 1 and 2<br />

respectively by means of Corollary 5.8 (these two singularities are with monomial front and L is<br />

diagonal).<br />

The 2-reduced system of System (53) is de…ned by<br />

2<br />

2 dY<br />

2t<br />

dt =<br />

6<br />

4<br />

0 0 0 0 0 0<br />

t2 2 0 t3 0 0<br />

t2 0 4 + t<br />

2 t3 0 0 0 t<br />

0<br />

0<br />

0<br />

0<br />

t2 0 0 t2 2 t 0<br />

t2 0 0 t2 0 4 t<br />

2<br />

Consequently, the <strong>for</strong>mal series e f reads (cf. System (17))<br />

2<br />

1<br />

6<br />

6ef<br />

6<br />

ef = 6<br />

4<br />

1;2<br />

(t)<br />

ef 1;3<br />

(t)<br />

0<br />

ef 2;2<br />

(t)<br />

ef 2;3<br />

3<br />

7<br />

5<br />

(t)<br />

where the <strong>for</strong>mal series e f u;j<br />

, u = 1; 2 and j = 2; 3, satisfy e f u;j<br />

(t) = O(t 2 ) and the di¤erential<br />

51<br />

3<br />

7 Y<br />

7<br />

5<br />

2<br />

x 2


system<br />

8<br />

><<br />

>:<br />

2t 2 de f 1;2<br />

dt<br />

2t 2 de f 1;3<br />

dt<br />

2t 2 de f 2;2<br />

dt<br />

2t 2 de f 2;3<br />

dt<br />

2e f 1;2<br />

= t2 4 + t<br />

2 e f 1;3<br />

= t 2<br />

(2 t) e f 2;2<br />

= t2 4<br />

t<br />

2 e f 2;3<br />

= t 2<br />

By Borel trans<strong>for</strong>mation, the minor b f reads<br />

2<br />

2 3 0<br />

1 6<br />

6<br />

607<br />

6bf<br />

7 6<br />

6<br />

bf( ) = 607<br />

6<br />

7<br />

6<br />

607<br />

+ 6<br />

7 6<br />

405<br />

6<br />

4<br />

0<br />

1;2<br />

( )<br />

bf 1;3<br />

( )<br />

0<br />

bf 2;2<br />

( )<br />

bf 2;3<br />

3<br />

7<br />

5<br />

( )<br />

where b f u;j<br />

, u = 1; 2 and j = 2; 3, satisfy b f u;j<br />

(0) = 0 and the di¤erential system<br />

There<strong>for</strong>e,<br />

8<br />

><<br />

>:<br />

2( 1) b f 1;2<br />

=<br />

2( 2) db f 1;3<br />

d<br />

+ 3<br />

2 b f 1;3<br />

= 1<br />

2( 1) db f 2;2<br />

d + 3b f 2;2<br />

= 1<br />

2( 2) db f 2;3<br />

d<br />

+ 5<br />

2 b f 2;3<br />

= 1<br />

bf 1;2<br />

and b f 2;2<br />

have only one singularity, located at = 1, and their analytic continuations to<br />

the right of 1 along d0 = R + are given by<br />

bf 1;2<br />

( ) = 2( 1)<br />

and<br />

52<br />

b f 2;2<br />

( ) = i<br />

3 ( 1) 3=2 1<br />

3


f 1;3<br />

and b f 2;3<br />

have only one singularity, located at = 2, and their analytic continuations to<br />

the right of 2 along d0 are given by<br />

bf 1;3<br />

( ) = 27=4<br />

3 e 3i =4 ( 2) 3=4 + 2<br />

3<br />

We deduce thus from Corollary 5.8 that<br />

8<br />

><<br />

>:<br />

c 2 0 =<br />

c 3 0 = 2 3=4<br />

6.2 Example 2<br />

4 p<br />

3<br />

4<br />

3 ( 3<br />

4<br />

and<br />

i , c 2 = + 4p<br />

3<br />

16<br />

+<br />

) 5<br />

( 3<br />

4 ) i , c3 = 2 3=4 4<br />

3 ( 3<br />

4 )<br />

2;3<br />

fb ( ) = 29=4<br />

5 e 5i =4 ( 2) 5=4 + 2<br />

5<br />

In this second example, the matrix L has a 2-dimensional Jordan block and b f has only one<br />

singularity that is with monomial front.<br />

Consider the system<br />

3 dY<br />

(54) x<br />

dx =<br />

2<br />

4<br />

0 0<br />

3<br />

0<br />

5 Y<br />

x 4 + x 5 2 x 2<br />

x 4 + 2x 5 0 2<br />

with single level 2 and its <strong>for</strong>mal fundamental solution e Y (x) = e F (x)xLeQ(1=x) de…ned by<br />

2<br />

1<br />

eF (x) = 4<br />

0<br />

1<br />

0<br />

3<br />

0<br />

05<br />

= I3 + O(x<br />

1<br />

4 ) ;<br />

2<br />

0<br />

L = 40<br />

0<br />

0<br />

0<br />

0<br />

3<br />

0<br />

15<br />

0<br />

and Q 1<br />

x<br />

= diag 0;<br />

1<br />

;<br />

x2 The anti-<strong>Stokes</strong> directions of the …rst column e f of e F are the directions = 0 and = (the<br />

directions of maximal decay of e 1=x2)<br />

and the <strong>Stokes</strong> matrices in these directions are I3 + C0 and<br />

I3 + C respectively where<br />

2 3<br />

2 3<br />

C0 =<br />

The four <strong>Stokes</strong> <strong>multipliers</strong> c 2;1;1<br />

0<br />

0 0 0<br />

4c<br />

2;1;1<br />

0 0 05<br />

and C = 4<br />

c 2;2;1<br />

0 0 0<br />

i<br />

0 0 0<br />

5<br />

c 2;1;1 0 0<br />

c 2;2;1 0 0<br />

16<br />

5<br />

( 3<br />

4 )<br />

, c 2;2;1<br />

0 , c 2;1;1 and c 2;2;1 are calculated from the <strong>resurgence</strong> <strong>multipliers</strong><br />

of b f at 1 by means of Corollary 5.5 ( = 1 is a singularity with monomial front and L is not<br />

diagonal).<br />

53<br />

1<br />

x 2


The 2-reduced system of System (54) is de…ned by<br />

2<br />

2 dY<br />

2t<br />

dt =<br />

6<br />

4<br />

0 0 0 0 0 0<br />

t 2 2 t t 3 0 0<br />

t 2 0 2 2t 3 0 0<br />

0 0 0 t 0 0<br />

t 2 0 0 t 2 2 t t<br />

2t 2 0 0 t 2 0 2 t<br />

The <strong>for</strong>mal series e f reads there<strong>for</strong>e (cf. System (17))<br />

2<br />

1<br />

6<br />

6ef<br />

6<br />

ef(t) = 6<br />

4<br />

1;2;1;1<br />

(t)<br />

ef 1;2;2;1<br />

(t)<br />

0<br />

ef 2;2;1;1<br />

(t)<br />

ef 2;2;2;1<br />

3<br />

7<br />

5<br />

(t)<br />

where e f u;2;`;1<br />

, u = 1; 2 and ` = 1; 2, satisfy e f u;2;`;1<br />

(t) = O(t2 ) and the di¤erential system<br />

8<br />

2t2 de f 1;2;1;1<br />

dt<br />

2e f 1;2;2;1<br />

te f 1;2;2;1<br />

= t2 Thus, the minor b f reads<br />

><<br />

>:<br />

2t 2 de f 1;2;2;1<br />

dt<br />

2t 2 de f 2;2;1;1<br />

dt<br />

2t 2 de f 2;2;2;1<br />

dt<br />

bf( ) =<br />

2e f 1;2;2;1<br />

= t2 (2 t) e f 2;2;1;1<br />

(2 t) e f 2;2;2;1<br />

= 2t2 2<br />

2 3 0<br />

1 6<br />

6<br />

607<br />

6bf<br />

7 6<br />

6<br />

607<br />

6<br />

7<br />

6<br />

607<br />

+ 6<br />

7 6<br />

405<br />

6<br />

4<br />

0<br />

1;2;1;1<br />

( )<br />

bf 1;2;2;1<br />

( )<br />

0<br />

bf 2;2;1;1<br />

( )<br />

bf 2;2;2;1<br />

3<br />

7<br />

5<br />

( )<br />

54<br />

3<br />

7 Y<br />

7<br />

5<br />

te f 2;2;2;1<br />

= t2


where b f u;2;`;1<br />

, u = 1; 2 and ` = 1; 2, satisfy b f u;2;`;1<br />

(0) = 0 and the identities<br />

8<br />

><<br />

>:<br />

2( 1) b f 1;2;1;1<br />

= + 1 b f 1;2;2;1<br />

2( 1) b f 1;2;2;1<br />

=<br />

2( 1) db f 2;2;1;1<br />

d<br />

2( 1) db f 2;2;2;1<br />

d<br />

+ 3b f 2;2;1;1<br />

= 1 + b f 2;2;2;1<br />

+ 3b f 2;2;2;1<br />

= 2<br />

Solving these equations by the method of variation of constants, we deduce that the analytic<br />

continuations of the b f u;2;`;1<br />

’s to the right of 1 along d0 = R + read<br />

8<br />

><<br />

>:<br />

bf 1;2;1;1<br />

( ) =<br />

bf 1;2;2;1<br />

( ) = 2( 1)<br />

bf 2;2;1;1<br />

( ) =<br />

3 + i log( 1)<br />

+<br />

4( 1) 4( 1)<br />

3 5i<br />

( 1)<br />

9<br />

3=2 i<br />

3 ( 1) 3=2 log( 1) + 5<br />

9<br />

bf 2;2;2;1<br />

( ) = 2i<br />

3 ( 1) 3=2 + 2<br />

3<br />

Apply now Corollary 5.5: Identity (52) implies<br />

2<br />

1<br />

6<br />

60<br />

41<br />

0<br />

0<br />

1<br />

0<br />

1<br />

1<br />

0<br />

1<br />

0<br />

3 2<br />

i c<br />

1 7 6<br />

7 6<br />

i 5 4<br />

1<br />

2;1;1<br />

0<br />

c 2;2;1<br />

0<br />

c 2;1;1<br />

c 2;2;1<br />

3<br />

7<br />

5 =<br />

2<br />

3 + i<br />

6<br />

63<br />

6<br />

4<br />

4<br />

5i<br />

0<br />

9<br />

55<br />

0(0) + 1<br />

4 1(0)<br />

1<br />

2 0(0)<br />

1<br />

2<br />

2i<br />

3 0<br />

1<br />

2<br />

i<br />

3 1<br />

1<br />

2<br />

3<br />

7<br />

5


1<br />

where the complex numbers p(0) and p( ), p = 0; 1, are de…ned by<br />

2<br />

8<br />

><<br />

>:<br />

There<strong>for</strong>e,<br />

0(0) = 2i<br />

1(0) = 2i d<br />

dz<br />

0<br />

1<br />

8<br />

><<br />

>:<br />

1<br />

2<br />

1<br />

2<br />

c 2;1;1<br />

0<br />

c 2;2;1<br />

0<br />

c 2;1;1 =<br />

6.3 Example 3<br />

= 2i ei =2<br />

= 2i d<br />

dz<br />

= ip<br />

18<br />

i z e<br />

= 2<br />

(1 z) jz=0<br />

2 2i ( is Euler’s constant)<br />

( 3 = 4p<br />

) 2<br />

e<br />

i z<br />

(88 + 27p<br />

= ip<br />

3 (8 + 3p )<br />

8 p<br />

(1 z) jz= 1<br />

2<br />

24 9 p<br />

= p (4 8 + 8 log 2 + 4i )<br />

48 log 2)<br />

3 + 2 + ip<br />

18 (88 + 27p + 24 9 p + 48 log 2)<br />

c 2;2;1 = ip<br />

3 (8 3p )<br />

In this last example, we treat in detail the case of a singularity with non-monomial front. Let<br />

3 dY<br />

(55) x<br />

dx<br />

= 0 0<br />

x 4 2 + x Y<br />

be the system of Section 4.2.1. Recall that the <strong>for</strong>mal fundamental solution e Y is de…ned by<br />

eY (x) = e F (x)e Q(1=x) where<br />

eF (x) =<br />

1 0<br />

1 = I2 + O(x 4 ) and Q 1<br />

x<br />

= diag 0;<br />

The anti-<strong>Stokes</strong> directions of the …rst column e f of e F are once more = 0 and = (the<br />

directions of maximal decay of e 1=x2 1=x ) and the associated <strong>Stokes</strong> matrices are de…ned by I2+C0<br />

and I2 + C respectively where<br />

C0 =<br />

0 0<br />

c 2 0 0<br />

and C =<br />

56<br />

0 0<br />

c 2 0<br />

1<br />

x 2<br />

1<br />

x


The singularity = 1 of b f being with non-monomial front, the calculation of the <strong>Stokes</strong> <strong>multipliers</strong><br />

is much more involved than in the previous examples. Indeed, the study of the analytic continuation<br />

of b f near 1 (cf. Section 4.2.1) does not allow to calculate easily the <strong>resurgence</strong> <strong>multipliers</strong> and,<br />

there<strong>for</strong>e, the <strong>Stokes</strong> <strong>multipliers</strong>.<br />

To circumvent this di¢ culty, we per<strong>for</strong>m the singularity = 1 into a singularity with monomial<br />

front. Let us<br />

System (55) becomes<br />

3 dY<br />

(56) y<br />

dy =<br />

x = y<br />

1<br />

2<br />

4<br />

y<br />

2<br />

3<br />

0 0<br />

5 Y<br />

8y4 2<br />

(2 y) 3<br />

and the <strong>for</strong>mal fundamental solution e Y(y) = e F1(y)e Q1(1=y) associated with e Y (x) is de…ned by<br />

eF1(y) = e F (x(y))<br />

1 0<br />

0 e 1=4 =<br />

1 0<br />

0 e 1=4 + O(y 4 ) and Q1<br />

1<br />

y<br />

= diag 0;<br />

Systems (55) and (56) have the same <strong>Stokes</strong> matrices ([L-R90]). To normalize e F1 to e F1(0) = I2,<br />

we use the constant gauge trans<strong>for</strong>mation<br />

This one trans<strong>for</strong>ms System (56) into<br />

Z =<br />

3 dZ<br />

(57) y<br />

dy =<br />

2<br />

4<br />

1 0<br />

0 e 1=4 Y<br />

3<br />

0 0<br />

5 Z<br />

8e 1=4 y 4<br />

(2 y) 3 2<br />

and the <strong>for</strong>mal fundamental solution e Y(y) into the <strong>for</strong>mal fundamental solution e Z(y) = e F2(y)eQ1(1=y) where<br />

eF2(y) =<br />

1 0<br />

0 e 1=4 e F1(y) = I2 + O(y 4 )<br />

Systems (55) and (57) have once more the same <strong>Stokes</strong> matrices. The singularity = 1 being now<br />

with monomial front, we can apply Corollary 5.8 (the matrix L is null).<br />

The 2-reduced system of System (57) is de…ned by<br />

2 dY<br />

2t<br />

dt =<br />

2<br />

0<br />

6<br />

6T1(t)<br />

4 0<br />

0<br />

2<br />

0<br />

0<br />

tT2(t)<br />

t<br />

0<br />

0<br />

0<br />

3<br />

7<br />

5<br />

T2(t) 0 T1(t) 2 t<br />

Y<br />

57<br />

1<br />

y 2


where 8> <<br />

>:<br />

The …rst column e f 2 of e F 2 reads<br />

T1(t) = 16e 1=4 (4 + 3t)t2 64 48t + 12t2 = 16e 1=4X (2m 3)(m 1)<br />

t3 4<br />

m 2<br />

m tm T2(t) = 8e 1=4 (12 + t)t2 64 48t + 12t2 = 8e 1=4X (2m 1)(m 1)<br />

t3 4<br />

m 2<br />

m tm 2<br />

1<br />

6<br />

ef<br />

6ef<br />

2(t) = 6<br />

4<br />

1;2<br />

2 (t)<br />

0<br />

ef 2;2<br />

3<br />

7<br />

5<br />

2 (t)<br />

where the <strong>for</strong>mal series e f 1;2<br />

2 and e f 2;2<br />

2 satisfy e f 1;2<br />

2 (t) = e f 2;2<br />

2 (t) = O(t 2 ) and the di¤erential system<br />

8<br />

Thus, the minor b f 2 of e f 2 reads<br />

><<br />

>:<br />

2t2 de f 1;2<br />

2<br />

dt<br />

2t2 de f 2;2<br />

2<br />

dt<br />

bf 2( ) =<br />

2 e f 1;2<br />

2 = T1(t)<br />

(2 t) e f 2;2<br />

2 = T2(t)<br />

2 3<br />

1<br />

6<br />

60<br />

7<br />

405<br />

0<br />

+<br />

2<br />

0<br />

6<br />

6bf<br />

6<br />

4<br />

1;2<br />

2 ( )<br />

0<br />

bf 2;2<br />

3<br />

7<br />

5<br />

2 ( )<br />

where b f 1;2<br />

2 and b f 2;2<br />

2 satisfy b f 1;2<br />

2 (0) = b f 2;2<br />

2 (0) = 0 and the identities<br />

with 8> <<br />

>:<br />

8<br />

><<br />

>:<br />

2( 1) b f 1;2<br />

2 = b T1( )<br />

2( 1) db f 2;2<br />

2<br />

d + 3b f 2;2<br />

bT1( ) = 16e 1=4X 2m 1<br />

4m+1 (m 1)!<br />

m 1<br />

bT2( ) = 8e 1=4X 2m + 1<br />

4m+1 (m 1)!<br />

m 1<br />

58<br />

2 = d b T2<br />

d<br />

( )<br />

m ( + 2) ( 1)=4<br />

= e<br />

2<br />

m ( + 6) ( 1)=4<br />

= e<br />

4


There<strong>for</strong>e the analytic continuations of b f 1;2<br />

2 and b f 2;2<br />

2 to the right of 1 along d0 = R + are de…ned<br />

by 8> <<br />

where<br />

>:<br />

bf 1;2<br />

2 ( ) =<br />

bf 2;2<br />

2 ( ) = 1<br />

2<br />

= i<br />

4<br />

( + 2) 1)=4<br />

e(<br />

4( 1)<br />

Z<br />

3=2 ( 1)<br />

0<br />

e 1=4 + 3<br />

2 1F1<br />

We deduce thus from Corollary 5.8<br />

References<br />

8<br />

><<br />

>:<br />

c 2 0 = ip<br />

2<br />

c 2 = ip<br />

2<br />

( 1) 1=2 d e T2<br />

d ( )d = ( 1) 3=2 + Hol( 1)<br />

3 p<br />

1 3<br />

;<br />

2 2 ;<br />

1<br />

4<br />

2e 1=4 31F1<br />

3 p + 2e 1=4 + 31F1<br />

and Hol( ) 2 Cf g<br />

1 3<br />

;<br />

2 2 ;<br />

1 3<br />

;<br />

2 2 ;<br />

[BJL79] W. Balser, W. B. Jurkat, D. A. Lutz, A general theory of invariants <strong>for</strong> meromorphic<br />

di¤erential equations ; Part I, <strong>for</strong>mal invariants, Funkcialaj Ekvacioj, 22, 197 221, 1979<br />

[CNP93] B. Candelpergher, J.-C. Nosmas, F. Pham, Approche de la résurgence, Actualités Mathématiques,<br />

Hermann, Paris, 1993<br />

[D70] P. Deligne, Equations di¤érentielles à points singuliers réguliers, Lecture Notes in Mathematics,<br />

163, Springer-Verlag, 1970<br />

[E85] J. Ecalle, Les fonctions résurgentes, tome III : l’équation du pont et la classi…cation analytique<br />

des objets locaux, Publications Mathématiques d’Orsay, n 85 05, 1985<br />

[Ka83] M. Kashiwara, Systems of microdi¤erential equations, Progress in Mathematics, 34,<br />

Birkhäuser, Boston, 1983<br />

[L-R90] M. Loday-Richaud, Calcul des invariants de Birkho¤ des systèmes d’ordre deux, Funkcialaj<br />

Ekvacioj, 33, 161 225, 1990<br />

[L-R01] M. Loday-Richaud, Rank reduction, normal <strong>for</strong>ms and <strong>Stokes</strong> matrices, Expositiones<br />

Mathematicae, 19, 229 250, 2001<br />

[L-RR] M. Loday-Richaud, P. Remy, Resurgence and calculation of the <strong>Stokes</strong> <strong>multipliers</strong> with an<br />

error bound <strong>for</strong> <strong>linear</strong> ODE’s with a single level equal to one, in preparation<br />

59<br />

1<br />

4<br />

1<br />

4


[Lu72] D. A. Lutz, On the reduction of rank of <strong>linear</strong> di¤erential systems, Paci…c Journal of<br />

Mathematics, 42, n 1, 153 164, 1972<br />

[Mal91] B. Malgrange, Fourier trans<strong>for</strong>m and di¤erential equations, Recent Developments in<br />

Quantum Mechanics, A. Boutet de Monvel et al. (eds.), Kluwer Academic Publishers,<br />

33 48, 1991<br />

[MR89] J. Martinet, J.-P. Ramis, Théorie de Galois di¤érentielle et resommation, Computer Algebra<br />

and Di¤erential Equations, E. Tournier, éd., Academic Press, 117 214, 1989<br />

[Re07] P. Remy, Résurgence des systèmes di¤érentiels linéaires et calcul des matrices de <strong>Stokes</strong>,<br />

Thesis, Université d’Angers, September 19, 2007<br />

[Sau06] D. Sauzin, Resurgent functions and splitting problems, RIMS Koukyuroku, Kyoto, to appear<br />

[Si90] Y. Sibuya, Linear di¤erential equations in the complex domain : problems of analytic continuation,<br />

Translations of Mathematical Monographs, 82, American Mathematical Society,<br />

Providence, RI, 1990<br />

[Tu63] H. L. Turrittin, Reducing the rank of <strong>ordinary</strong> di¤erential equations, Duke Math. J., 30,<br />

271 274, 1963<br />

60

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!