• Aucun résultat trouvé

w l. Formulation of the problem The Ritt theorem in several variables

N/A
N/A
Protected

Academic year: 2022

Partager "w l. Formulation of the problem The Ritt theorem in several variables"

Copied!
14
0
0

Texte intégral

(1)

The Ritt theorem in several variables

C. A. BEREI~STEII# a n d M. A. DOSTAL

University of Maryland and Stevens Institute of Technology*)

w l. Formulation of the problem

T h e m o t i v a t i o n for t h e p r o b l e m t r e a t e d in this n o t e arises in t h e t h e o r y of con- v o l u t i o n e q u a t i o n s . L e t U be a locally c o n v e x space o f f u n c t i o n s or d i s t r i b u t i o n s . I t is a s s u m e d t h a t b y m e a n s of a t r a n s f o r m a t i o n ~ o f F o u r i e r typeS), t h e s p a c e

U is i s o m o r p h i c to a s u b a l g e b r a s of t h e a l g e b r a ,vg = ~-~(C n) of all e n t i r e f u n c t i o n s in {3% U s i n g t h e i n v e r s e t r a n s f o r m a t i o n ~ 1 one can t r a n s f e r t h e o p e r a t i o n o f m u l t i p l i c a t i o n f r o m ~7 to U. T h e r e s u l t i n g ring m u l t i p l i c a t i o n in U is called t h e c o n v o l u t i o n a n d is d e n o t e d r * ~0 (r ~v C U). More generally, let G E ~A b e such t h a t t h e m u l t i p l i c a t i o n b y t h e f u n c t i o n G is a c o n t i n u o u s e n d o m o r p h i s m c}l'la o f U. I f %~ is t h e c o n t i n u o u s e n d o m o r p h i s m o f U c o r r e s p o n d i n g to c ~ a u n d e r t h e i s o m o r p h i s m ~, t h e n ~(fG is called a c o n v o l u t o r of t h e s p a c e U. S o m e t i m e s - - b u t n o t a l w a y s - - t h e r e exists a d i s t r i b u t i o n S (or a n o t h e r " g e n e r a l i z e d f u n c t i o n " ) s u c h t h a t t h e F o u r i e r t r a n s f o r m of S is t h e f u n c t i o n G, a n d for e a c h r C U, ~r162 c a n be i n t e r p r e t e d as t h e c o n v o l u t i o n S * r in s o m e g e n e r a l i z e d sense.

G i v e n f C U, consider t h e e q u a t i o n

S * u = f (1.1)

w h i c h is equivalent to the equation S(~)~;(~) = f(~). It is natural to ask for necessary a n d sufficient conditions for the solvability of (1.1) in U. A n obvious necessary condition is that

f/S

be entire. In order to conclude that ~ S is in U one usually

*) The authors wish to thank Instituto de Matematica Pura e Aplicada, l~io de Janeiro, for its support and hospitality.

1) cj can be either the classical Fourier-Laplace transformation or the Fourier-Borel trans- formation or another similar transformation depending upon the nature of the space U. The inverse transformation will be denoted by cj-1; instead of cj(r we shall write r Similarly,

stands for cj(U), etc.

(2)

268 C . A . B E I ~ E N S T E I N A N D ]VL A . D O S T A L

has to use additional properties of S and U, and this is often done b y combining a theorem of Paley-Wiener t y p e 2) with the following intermediary step: There is another subMgebra ~ of cA containing S and f and such that

(S) whenever F, G C ~ are such that F/G E ~ , then F/G E ~?~.

Obviously, if ~3 = (/ has property (S), then the above necessary condition for the solvability of (1.1) is also sufficient. However, this is rarely the case, and one usually has to consider a different subMgebra '%)6. In the sequel, subalgebras

~/~ with the property (r will be called stable. Denoting b y F~N the quotient field of ~?d, stability of ~N means F ~ N ~4 = 9:0.

As an illustration of the previous general argument, consider the following two examples:

Example 1. Let U = c~A ' be the dual of the space ~>4 endowed with the compact-open topology. B y means of the Fourier-Borel transform (eft w 2), U is isomorphic to the algebra U = E x p = the set of all entire functions of exponential type in {Y. The stability of E x p is a classical result of LindelSf [18] (eft also below).

Hence a necessary and sufficient condition for the solvability of equation (1.1) in the space of analytic functionals in 13" is that 0~S C :~q. Non-triviM examples of the use of the stability of Exp in the s t u d y of convolution equations can be found in H6rmander [13] and MMgrange [19, 20].

Example 2. Let U =- cE' be the space of distributions with compact support in R ~. Consider equation (1.1) with S , f C ~'. Then, although the function f/S is of exponential t y p e (provided it is entire), in generM j~S ~ c~,, unless S is in- vertible (cf. [12]). Hence c~, is not stable, and it is the non-stability of c?~, which causes serious difficulties in the s t u d y of such convolution equations. In order to overcome these difficulties, one often has to resort to intricate methods (cf. [13, 20]).

I t therefore seems interesting to examine the stability of those subalgebras of d which arise in the study of convolution equations. Below we list some of them:

~P: polynomials in (In;

E: finite exponential sums, i.e. entire functions of the form

2) For example, the Paley-Wiener-Schwartz theorem for the case of distributions, or the PSlya-Ehrenpreis-Martineau theorem (cf. w 2) when analytic functionMs are involved, etc.

(3)

T H E I~ITT T H E 0 1 ~ E ~ 117 S E V E I ~ A L V A R I A B L E S 269

H(~) = ~ hje<o~ "~>, (~ e C") (1.2)

j = l

where hj are complex numbers, and 0~ C (Y, also called frequencies of H, are given points in lY 3);

E2: exponential polynomials in C ~, i.e. functions of the form (1.2) with hj E 7 ; F,7: entire functions of the form HIP for some H C E2, P C ~?;

~ o : Fourier transforms of distributions q ) q ~g' such that, for some constants t > _ 0 , r > 0 , ~ > 0 and A real (all depending on q)),

^

m a x ]~b(~')l >_ c e x p [Aco(~) @ hi~l(~)], (V ~ C C"), (1.3) where A(~; r) denotes the p o l y d i s k of center ~ and radius r; ~ = ~ § i~; h[~ 1 is the supporting function of the s u p p o r t of q~; a n d ~o(~) = In (2 -t- 1~]). ~) (For the properties of such classes, see [3, 4, 5]);

~ : Fourier transforms of Schwartz distributions with compact support; or, more generally,

(r the same for Beurling distributions (cf3));

Exp: entire functions in C ~ of exponential type.

The stability of ~ is simple (ef. e.g. [9]). T h a t E is stable for n = 1 constitutes the so-called R i t t t h e o r e m [23]. F o r n > i this was p r o v e d b y Avanissian and Martineau [1]. E 2 is clearly non-stable. Indeed, for n = 1,

sin z

- - e E~ \ E~ (z e C). (1.4)

Z

I t can be easily established t h a t c~ 0) is stable [5]. ~ ' ' and more generally c~, 6 0 ) ,

are not stable (ef. E x a m p l e 2 above). The stability of E x p was p r o v e d b y E. Lindel6f [18] for n = 1. The extension to a r b i t r a r y n >_ 1 is due to Ehrenpreis [10] and Malgrange [19].

I t remains to be found, first how " u n s t a b l e " E 2 really is a), and secondly,

3) I n w h a t follows it will a l w a y s be a s s u m e d t h a t t h e frequencies Oj are pairwise d i s t i n c t a n d t h e coefficients hj are all non-zero. Besides, < , > denotes t h e bilinear p r o d u c t in On: <0, ~> = 01~1 + 9 . . + O n ~ .

4) A c t u a l l y , for c0 one m a y t a k e a n y c o n t i n u o u s s u b a d d i t i v e f u n c t i o n in R" satisfying c e r t a i n g r o w t h conditions. T h e n , i n s t e a d of ~ ' , one has to t a k e t h e space cd,~o of B e u r l i n g d i s t r i b u t i o n s (of. [2]).

~) i.e., to describe t h e s t r u c t u r e of e n t i r e f u n c t i o n s of t h e f o r m Y/G where F , G E E 2 (or, e q u i v a l e n t l y , F , G C E 9 .

(4)

270 C . A . B E I ~ E I ~ S T E I N A I ~ D M . A . D O S T A L

w h e t h e r E 2 is s t a b l e or not. O u r m a i n o b j e c t i v e in this n o t e is to s h o w t h a t b o t h questions h a v e a c o m m o n answer:

MAIN T~EOREM. The subalgebra E2 is stable. Hence the most general form of an entire function of the form FIG, F, G E E 2, (cf.5)) is

F H

G P (H ~ E2, P c ~). (1.5)

G i v e n a n e x p o n e n t i a l p o l y n o m i a l H , t h e g r e a t e s t c o m m o n d i v i s o r o f its coeffi- cients,

dH = (hi, . . . , hs), (1.6)

will be called the content of H. F u r t h e r m o r e , l e t f = HIP be a n y e l e m e n t o f E2"

T h e n we shall s a y t h a t HIP is a reduced form o f f, p r o v i d e d (P: dH) = 1. I f t h i s is so a n d f = H*/P*, for s o m e H* C E 2 a n d P* C ~ , t h e n it is e a s y to see t h a t PIP* a n d dHIdH.. H e n c e t h e following l e m m a holds:

LnMMA. Every function in E2 has a unique reduced form 6). Furthermore, let H, F E E~ be such that Q = H / F E ~ . Then Q[dH.

T h e n t h e m a i n t h e o r e m c a n be r e f o r m u l a t e d as follows:

T~EOI~EM 1. Let F, G E E 2 be such that FIG C ~i. Then there exist unique 6) H E E 2 and P E ~ such that F / G = H / P and (P, dH)= 1.

As a n a p p l i c a t i o n o f T h e o r e m 1 one obtains:

T ~ E O ~ E ~ 2. Let F, G E E 2 and F/G E d . Let HIP FIG. Then PIdG.

COROLLARY 1. Let F, G be exponential polynomials in C" such that FIG is entire and dc = 1. Then FIG is also an exponential polynomial.

COROLLARY 2. E is stable (n >_ 1).

As w a s m e n t i o n e d a b o v e , Corollary 2 for n - - 1 is t h e R i t t t h e o r e m [23]. O t h e r proofs o f R i t t ' s t h e o r e m were g i v e n b y H . Selberg [24], 1 ). D. L a x [16] a n d A. Shields [25]. Shields p r o v e s t h a t , for n ~- 1, t h e h y p o t h e s e s F E E 2, G E E a n d FIG E d i m p l y FIG E E,~. H e also m e n t i o n s t h a t , a c c o r d i n g to a n u n p u b l i s h e d r e s u l t o f

6) Unique up to a constant multiple of H and P.

be the reduced form of

(5)

THE I ~ I T T TttEORE)/s IN SEVEI~A~, VAll, IABLES 271 W. D. B o u w s m a , Corollary 1 holds w h e n n = 1. F i n a l l y , Corollary 2 is due t o V. A v a n i s s i a n a n d A. M a r t i n e a u (unpublished [1]). 7)

T h e m a i n t h e o r e m is e s t a b l i s h e d in Section 2. T h e o r e m 2, as well as a n o t h e r a p p l i c a t i o n o f T h e o r e m 1 are discussed in t h e concluding Section 3.

T h e o r e m s 1 a n d 2 were a n n o u n c e d in our n o t e [7].

F o r applications o f e x p o n e n t i a l p o l y n o m i a l s in one v a r i a b l e see [15].

w 2. Proof of the main theorem

F o r ~ E C ", ~ d e n o t e s t h e c o m p l e x c o n j u g a t e o f ~, i.e. ~ = ( ~ 1 , - . . , ~n).

W h e n ~ is considered as a p o i n t in R 2'~, t h e coordinates o f ~ are (Re gl, I m ~1 . . . . , R e ~n, I m ~ ) . T h e E u c l i d e a n i n n e r p r o d u c t in R m will be d e n o t e d < , >~. W e recall f r o m w 1 t h a t t h e bilinear p r o d u c t in C n, d e n o t e d s i m p l y b y < , >, is

< z , ~ > = Z l ~ l + . . . + z n ~ , z , ~ C ~.

I f F is an e x p o n e n t i a l p o l y n o m i a l w i t h frequencies a ~ , . . , %, i.e.

P

F(~) ~- Z Pj(~) e<"j'r (2.1)

j = l

w h e r e Pj are polynomials, we will d e n o t e b y [F] t h e c o n v e x hull o f t h e p o i n t s 51 . . . . , @ in R 2". L e t h[f ] be t h e s u p p o r t i n g f u n c t i o n of t h e set [F], i.e. for each ~ E C " = It 2"

h [ F ] ( ~ ) = m a x < x , C>2n = m a x <(~j, ~>2n ~ - m a x ] ~ e < a j , ~ > . ( 2 . 2 )

xE [F] 1 < j < p l < j <_p

W e shall n e e d a few simple facts a b o u t a n a l y t i c functionals, i.e. e l e m e n t s o f d ' (cf. [14, 17, 21, 26]). T h e F o u r i e r - B o r e l t r a n s f o r m /~(~) = #(e <'' :>) establishes a n i s o m o r p h i s m of t h e spaces ~4' a n d E x p . G i v e n # E ~ ' , t h e indicator p , of

# is d e f i n e d b y

- - log I~(t~)l p~(~) = lim

t-+co t

a n d its u p p e r - s e m i c o n t i n u o u s r e g u l a r i z a t i o n 15, 15,(~) = lim p,(~'), is p l u r i s u b h a r m o n i c .

7) Professor H. S. Shapiro has kindly informed us that several years ago he proved Corollary 2 by means of an inductive argument. His proof has not been published. Added in proof: In the meantime K. Kitagawa [27] published a proof of Corollary 1. His proof is rather sketchy.

A complete proof of Corollary 1 was recently announced by V. Avanissian and 1~. Gay in [28, 29].

(6)

272 C. A . t 3 E I % E N S T E I I ~ A:ND ~I. A . D O S T A L

A carrier of a n a n a l y t i c f u n c t i o n a l # is a c o m p a c t subset K of 6 ~ such t h a t for e v e r y n e i g h b o r h o o d U o f K t h e r e is a c o n s t a n t C such t h a t

1~(r _< c sup Ir l,

z E U

for all r E Q/. A c o m p a c t c o n v e x set K is called a convex support o f # if K is a m i n i m a l c o m p a c t c o n v e x carrier o f #, i.e. K is a carrier o f # such t h a t i f L is a n o t h e r carrier o f # a n d L c _ K , t h e n ch. L = K , w h e r e ch. L d e n o t e s t h e c o n v e x hull of L.

I n t h e sequel t h e P d l y a - E h r e n p r e i s - M a r t i n e a u t h e o r e m will be u s e d in t h e following f o r m u l a t i o n (cf. [14], Th. 5.2, Cor. 5.3):

T~nORE~ I. Let # E ~ ' , then

~,(~) ~- inf{hK(~): K carries #}.

Hence # has a unique convex support if and only i f ~5, is a convex function.

(2.3)

I n [4, 6] we established t h e following lower e s t i m a t e for e x p o n e n t i a l polynomials:

T~nORE~ I I . Let P E E2, i.e. P is an exponential polynomial. Then, for each

> O, there exists a constant C ~-- C(s, P) such that if f is an analytic function in the polydisk zl(~, s) = A,

A = {~' E C": m a x I~j - - $/] ~ e}, (2.4) J

then

If(~)[ chip](0 ~ C m a x [f(z)P(z)]. (2.5)

z E A

To e v e r y e x p o n e n t i a l p o l y n o m i a l t h e r e corresponds, via t h e F o u r i e r - B o r e l t r a n s - form, a u n i q u e #p E c9/' such t h a t /~p(~) = P(~). H e n c e we h a v e

COROLLARY 1. l~or every P E E2, [P] is the unique convex support of #p and

~.~(~) --p.~(~) ~ h~(~). (2.6)

(Indeed, it suffices to set

f

~ 1 in T h e o r e m I I a n d a p p l y T h e o r e m I).

L e t A be a c o m p a c t c o n v e x subset o f R m. F o r a n a r b i t r a r y 0 E R m, set

A ~ I f A ~ consists o f one p o i n t only, 0 is called a

regular direction o f A: T h e set o f all r e g u l a r directions will be d e n o t e d reg A.

T h e set o f all e x t r e m a l points will be d e n o t e d e x t A.

LnM~A 1 (cf. [8]). Let A and B be compact convex sets in R m. Then for each O,

(7)

T H E R I T T T H E O R E M I N S E V E R A L V A R I A B L E S 273

Moreover, every z in

A ~ gl e x t A = e x t (A~ (2.7)

(A -~ B ) ~ = A ~ ~- B ~, (2.8)

e x t (A § B) c__ e x t A -~ e x t B. (2.9) e x t (A + B) has a unique decomposition z = z 1 ~- z 2, z 1 E e x t A, z 2 E e x t B. Although the inclusion in (2.9) cannot be replaced by equality, one has for every 0 C reg A [J reg B,

(ext (A @ B)) ~ = (ext A ) ~ @ (ext B ) ~ (2.10)

( R e l a t i o n s (2.7)--(2.9) a r e o b v i o u s . T h e u n i q u e n e s s o f t h e d e c o m p o s i t i o n z = z 1 -}- z 2 w a s p r o v e d in [11]. E q u a t i o n (2.10) follows f r o m (2.7) a n d (2.8)).

F i n a l l y , a s i m p l e l e m m a on pieeewise linear f u n c t i o n s in R = will be n e c e s s a r y . G i v e n x 0C R '~ a n d 9 > 0, set

Bin(x0; ~) = {x e R~: Hx - -

x011

= m a x Ix i - Xo, i[ < ~}.

l < i < m

W e shall w r i t e B~(~) for B~(0; 9). L e t ~s be t h e class o f all c o n t i n u o u s f u n c t i o n s on B,,(1) w i t h t h e following p r o p e r t y : for e a c h r E ~ t h e r e e x i s t -N distinct v e c t o r s O j e R m ( j = 1 , . . . , - N ) s u c h t h a t for e a c h x e B m ( 1 ) , r 1 6 2

( x , 0i} ~ for s o m e j . G i v e n a f u n c t i o n f on a n o p e n c o n v e x s e t G _c R ", f will b e called piecewise linear on G if f o r e a c h x 0 E G t h e r e e x i s t a 0 > 0 a n d a n a f f i n e m a p p i n g Z of 1t m (i.e. Z(z) = A z @ x o for s o m e n o n - s i n g u l a r m • m a t r i x A) s u c h t h a t Z(0) = x0, z(B~(1)) --~ Bm(xo; O) c__G, a n d if r d e n o t e s t h e r e s t r i c t i o n o f f ~ t o B~(1), t h e n r is in c~.

LEMMA 2. Let f be a piecewise linear function defined on an open convex set G c t l m. T h e n f is convex i f (and only if) f is subharmonic.

Proof. I n v i e w o f t h e local c h a r a c t e r of c o n v e x i t y t h e l e m m a will follow if we show t h a t a n y s u b h a r m o n i c f u n c t i o n r r E ~ , is c o n v e x in B,,(~) for s o m e

~o _< t. W e c a n a s s u m e r = 0. Leg N ( r be t h e n u m b e r N c o r r e s p o n d i n g to b y definition. O u r claim is t r i v i a l w h e n e i t h e r N ( r = 1 or m = 1. A s s u m e t h a t it h a s b e e n p r o v e d for all i n t e g e r s 1, 2 . . . . , N - - 1 a n d a r b i t r a r y m (-N > 2).

F i x r to be a n y f u n c t i o n in c22 for w h i c h N ( r L e t 01 . . . . ,0N be t h e c o r r e s p o n d i n g v e c t o r s . Set V = { x E R m : ( x , 0 i - - 0 y ) = 0 , V i , j, i % j } a n d d = d i m V. Since N > 2 a n d 0~ are d i s t i n c t v e c t o r s , d < m, i.e. 0 < d < m - - 1.

Consider f i r s t t h e case d = 0, i.e. V = {0}. I f x 0 EBm(1) is a r b i t r a r y , x0 ~ 0, let -N0 be t h e t o t a l n u m b e r o f 0y'S for w h i c h r = (x0, 0y},,. T h e n _N o < 57, b e c a u s e d = 0. B y c o n t i n u i t y , in s o m e Bin(z0; d)c__Bm(1), one needs o n l y N 0 linear f u n c t i o n s to d e f i n e r B y t h e i n d u c t i o n h y p o t h e s i s r is c o n v e x in s o m e B,,(x0; do), 0 < do _< 8. H e n c e , it r e m a i n s to be s h o w n t h a t r is c o n v e x a t t h e

(8)

2 7 4 C~. A . B E R E ~ T S T E I k ' f f A N D M . A . I ) O S T A L

origin. L e t x~, x~ be a n y two distinct points in B~(1) such t h a t x 1 = 5X 2 for some cr ~ 0 . One has to show

r 1~- ( 1 - - ~)x2) < 2 r Jr ( 1 - - 2)r 0 < ~ < 1. (2.11) F o r m--~ 1, this is trivial because s u b h a r m o n i c i t y coincides w i t h convexity.

I f m > 1, there will be a vector y E Bin(1 ) linearly i n d e p e n d e n t of x 1 such t h a t x s - ~ y E B ~ ( 1 ) (i---- 1,21. Then for a n y k = 1 , 2 , . . . t h e s e g m e n t w i t h end- points X~, k = x~ d- k-~y does n o t contain t h e origin a n d b y t h e local c o n v e x i t y of r in B i n ( l ) ~ {0}, t h e i n e q u a l i t y (2.11) is satisfied for Xl, k, X2, k i n s t e a d of x 1, x 2. L e t t i n g k---> ~ , (2.11) follows b y continuity. Finally, for t h e case d > 1, one can assume t h a t

V = {x E Rm: xs = O for i > d}.

F o r x E R ~, set x = ( X s + l , . . . , x z ) a n d r r I t suffices to prove t h e c o n v e x i t y of r in B m s(0; ~) for some ~ < 1. However, since

dim {~ C Rm-S: <~, O~ -- Oj> m d = 0 V j, i i :/: j } = O, we are in t h e preceding case.

COROLLARY 2. Let f be a plurisubharmonic function in (~ which is piecewise linear in R 2'~- G ~. Then f is convex in R 2".

Proof of Theorem 1. L e t

be such t h a t

P

F(~) = ~ Pj(~)e<"J ' ~>, (P :e ~ ) (2.12)

j 1 q

G(~) ---- ~ Oj($le<bJ '~>, (Qj e ~1 (2.13 /

/ = 1

K = ~ c ~;d. F (2.14)

T h e n K C Exp. L e t ~o, #, #0 C d ' be such t h a t K = ~0, G = ~, F = ~o. Obviously, p~o _ < p , ~ p~o. Since P~0 =/5,0 = h[F], P , = /5, = big ] (cf. (2.6)), we o b t a i n from Theorems I, I I t h a t for e v e r y s > 0, there are constants C1, C 2 depending o n l y on s , F a n d G such t h a t for every ~ C ( F ,

eP."(~)]K(~)l _< C 1 m a x [F(z)l < C2ep.-0 (r162 (2.15)

z e ~ (~; ~)

This shows t h a t p~o < p."o --p.", hence b y (2.6)

P~0 --~ /5~0 = hM -- h[c]" (2.16)

(9)

THE RITT THEOREM IN SEVERAL VARIABLES 275 B y T h e o r e m 2, p:0 is a c o n v e x function. Since p~0 is also p o s i t i v e l y h o m o g e n e o u s o f o r d e r t, t h e r e exists a c o m p a c t c o n v e x set [K] _c R 2" such t h a t p:0 :-- h[g], hence b y (2.16),

IF] = [G] + [K]. (2.17)

( B y T h e o r e m I of this section, t h e set [K] is o b v i o u s l y t h e u n i q u e c o n v e x s u p p o r t of t h e f u n c t i o n a l ~0).

L e t ~/ be t h e f a m i l y of all linear varieties in R :~ of dimension 2n - - 1, each o f which contains a t least t w o d i f f e r e n t p o i n t s of t h e f o r m

P q

i = l j = l

where all t h e coefficients l~, mj are integers. T h e n t h e set 9~ = {0 C S 2~ ~: 0 /- A, for some A E CY} has m e a s u r e zero in S 2n-~, t h e u n i t sphere in R 2=. (Indeed, f i x a r b i t r a r y Zx, 9 9 zT (T _> 2) of t h e f o r m (2.18); t h e n t h e n o r m a l v e c t o r s to all A E ~ such t h a t z~ C A (i = 1 , . . . , T), d e f i n e a n algebraic s u b v a r i e t y of S 2~ 1 o f dimension < 2n - - 2.)

Obviously, one can assume t h a t t h e c o m p a c t sets IF], [G], [K] lie in R ~ , t h e positive o r t h a n t in R 2~. T h e set ~ l being o f m e a s u r e zero in S 2 n - l , o n e can f i n d v E (S 2~-~ ~ ~}?) [3 R+ ~. Since ~ ~ ~ / , v is a regular d i r e c t i o n for b o t h [F]

a n d [G]. H e n c e h~F~(v)----<c7j,~>2~ for e x a c t l y one ~j.. l ~ e n u m b e r i n g t h e a~'s, if necessary, a n d using t h e f a c t t h a t ~ q <]7, one can assume t h a t

h[F](V) = <a,, ~>2, > <a2, u>:, > . . . > <dr, r>z= > 0 (2.19) Similarly,

h[q(v) --~ <b~, v>:~ > . . . > <b~, v>2~ > 0. (2.20) Set k ~ - - - - a x - - b r B y L e m m a 1, k ~ E x t [ K ] , a n d k~ is t h e o n l y p o i n t o f [K]

for which h[Kl@) = <$~, v>2=.

Set

d ~ - < b 1 - - b2, ~)>2n, g A : ~__. {X e R2n: <X, ~'>2n > <~1, ~)>2n - - d}, H - : R 2n ~ H :b.

L e t r be such t h a t 5 ~ H + for i = 1 , . . . , r a n d 5 ~ r for i - - - - r - b l . . . . , p . Using t h e n o t a t i o n (2.12)--(2.14), set

[

fl(~) = F ( ; ) - p1(~)~<ol, ~>, g1($) G($) - - Ql(~)e <b'' ~>, K I ( g ) QI(g)K(r - - p i ( $ ) e <k,, ~>, F I ( $ ) Qx($)fl(g) - Pl(~)g~(g) e<k1" ~>

T h e n fl, El are e x p o n e n t i a l polynomials, K 1 is entire, a n d

(2.21)

F 1 = K1G. (2.22)

(10)

276 C. A . B E I ~ E N S T E I N A N D ~r A. D O S T A L

N e x t we claim t h a t

~ @ 5 2 C H - ( j = 2 . . . . ,q). (2.23)

Indeed, b y (2.19) a n d (2.20), <kl -5 bj, v}2~ = <51, u}2, -5 ( - - 51 -5 bj, v}2n <_

(5~, ~'>2~ + ( - b~ .5 b2, ~'>~ = (~1, ~>2,~ - - d.

Since 52 . . . . ,5~ E H +, it follows f r o m (2.23) t h a t none of the t e r m s w i t h fre- quencies aj (2 ~ j ~ r) can be cancelled in F 1 b y a t e r m coming from Pl(~)gl(~)e <k''~>. Moreover, it also shows t h a t a 1 c a n n o t be a f r e q u e n c y of F 1.

Hence, if x C [ F 1 ] , x @52 , a n d

h[F~](r) = (52, v}2, > (x, @2,. (2.24)

r _f _/

Thus, the frequencies of F1 are a2, . . . , a , a;+ 1 .. . . , %~, where { a t + l , . . . , a p l } i 8

a subset of (a~+l . . . . , @ , 5 1 - - b l @ t ) 2 , . . . , 5 1 - - E ) l " s b ~ } ~ H - and [F1]c__[F].

Indeed, [F~] E eh. ([f~] O { ~ "5 bj}j>_2) c [F] O ([K] "5 [G]) = [F].

N e x t we proceed w i t h ~1, G, K 1 in t h e same fashion as above w i t h F , G, K . H e n c e there is a ~1 E ~v4' such t h a t ;1 = / ( 1 a n d ~1 has a unique convex support [K1], a n d [F~] = [K~] "5 [G]. I n particular, b y (2.24) a n d (2.20),

(a~, ~)~o = hi~,](~ ) .5 hiG](~) = hi~,~(~) + (b,, ~}~.

}tenee h[K~](r) = (/c2, f}2,~ for a unique /c 2 C [K~]. On t h e other h a n d , b y L e m m a 1, /c 2 = a 2 - b 1. Set

I f~(~) = F~($) -- Q~(~)P~($)e <"~' :>

g~(~) = g~(~) (2.2~)

K~(~) = K~(~) - - P~(~)e <~-'' :>

[ F~($) = f~(~) -- P~($)g~(~)e <~o' ~>

Then f~ , F~ ~ E2, K~ ~ :~i a n d

Fe = K~G, (2.26)

a: is n o t a f r e q u e n c y of F~, b u t each a~, i = 3 , . . . , r is. The r e m a i n i n g frequencies a ~ + l , . . . , %~ form a subset of

{at+l, . .., %, a 1 -- 51 @ 5 2 .. . . , Ct 1 -- b~ ~- bq, a~ -- b I -~ b 2 .. . . . a 2 -- b~ "5 b~}.

Hence {5"+1, . .., ~ } ~ H - . Moreover, [Fe] ~ [F~], because

IF:] _~ ch ([fi] U {~: +

b~}j~:)~

[F1] U ([K1] -5 [G]) = [F~].

Continuing in t h e same fashion, one f i n a l l y constructs F~ ~ E}, a n d K~ ~ such t h a t (i) F~ = K~G, (ii) [F~] ~ _ H - [ 7 [F]. Since ~ e oH, t h e frequencies a} ~) of F~ can be n u m b e r e d so t h a t 5~ ~) is the o n l y point in [F~] for which (5~ ~), ~ ) ~ = h~,.~(~) a n d (~), ~)~ > (~), ~)~o > . . . . ~ 5 (~)~, ~)~o > 0. Set H I + =

{ x ~ R : ~ : ( x , r } : ~ > (c7~ ~ ) , v } ~ - d } , H i = R 2 ~ H +, a n d let r ~ > 1 be such

(11)

T H E : R I T T T I t E O I ~ E 1 V s I N S E V E R A L V A R I A B L E S 277 t h a t 5! ~) C H + for i = 1 , . . . , r~ a n d 5~ ~) C H~- for j > r~. I t is n o w clear t h a t we can r e p e a t t h e same p r o c e d u r e indefinitely. I f a t some p o i n t we o b t a i n

~Y = F~+~,~ .... +~N = 0, t h e t h e o r e m follows. H o w e v e r , this m u s t a c t u a l l y h a p p e n w h e n N is s u f f i c i e n t l y large. For, let N be so large t h a t H ~ [3 R+ ~ = O, h e n c e [ ~ ] c _ [ F J 1 3 H ~ = 0 a n d ~ 0 .

w 3. Applications

B y T h e o r e m 1, if F a n d G are e x p o n e n t i a l p o l y n o m i a l s such t h a t t h e q u o t i e n t K = FIG is entire, we can w r i t e K in t h e r e d u c e d form, K = H / P , w h i c h is u n i q u e l y d e t e r m i n e d (cf. w 1 and6)). N o w t h e question arises w h e n P ~ 1. T h e n e x t t h e o r e m gives a simple sufficient condition.

THEOREM 2. Let F, G e E 2 be such that FIG E ~ . Let H I P be the reduced form of .FIG. Then P divides d~. I n particular, P ~ 1 whenever de = 1.

Proof. Set

P

~(~) = ~ aj(=)e<~ '~

j - - 1 q

H(~) = ~ bj(~)e<~J ,~>.

j - - 1

F i r s t we shall p r o v e t h e following special case b y i n d u c t i o n on p.

(3.1)

(A) (i) P is irreducible (ii) de = 1. Then P ~ 1.

I f p = 1, t h e n b y (ii), al(z ) is a c o n s t a n t , a 1 ~ 0. H e n c e H / P is t h e r e d u c e d f o r m o f t h e e x p o n e n t i a l p o l y n o m i a l (1/al)e -< .... >F(z). I n view o f t h e u n i q u e n e s s o f t h e r e d u c e d form, P m u s t be c o n s t a n t .

Suppose n o w t h a t (A) holds w h e n e v e r G has a t m o s t p - 1 frequencies, p ~ 1 . T h e r e are t w o possible cases: e i t h e r P]bj for all j = 1 , . . . , q or P ~ b / for some j. I n t h e f i r s t case, P ~ 1 b y d e f i n i t i o n o f r e d u c e d form. H e n c e it suffices t o consider t h e second case when, a f t e r r e a r r a n g i n g t h e flj's if necessary, t h e r e is a q0 ~ 1 such t h a t P § bj, j = l . . . . , qo a n d bj = b ' P , bj* e ~ , for j = q0 + 1 , . . . , q. W e claim t h a t it suffices to consider t h e case q0 = q. I n d e e d , if qo < q, set

q0

F*(~) = F(~) - G(z) ~ b*(~)e<~i , % H*(z) = ~ bj(~)e<~J ' z>

J ~> qo j - - I

T h e n F*/G is e n t i r e a n d H * / P is its r e d u c e d form. T h e r e f o r e we shall a s s u m e

p + bj (V j). (3.2)

(12)

2 7 8 C . A . B E I t E N S T E I N A N D l'r A . ] : ) O S T A L

I t will be shown t h a t (3.2) leads to contradiction if P r constant, a n d this will prove (A). I t follows from w 2 t h a t [PF] = [H] + [G], a n d

[ P F ] = ch {s, + ~: i = 1 , . . . , p, j = 1 , . . . , q}.

L e t y be a f i x e d e x t r e m e point of the polyhedron [PF]. B y L e m m a 1, y = c~0 @ flL for e x a c t l y one i o a n d J0. R e n u m b e r i n g the frequencies one can assume t h a t i o = p , i.e.

@ 4- fiJo J= ~ + f i j ( i < p , V j ) . (3.3) Consider all jo's for which (3.3) holds. R e n u m b e r i n g t h e b/s one can assume t h a t there is some J, 1 < J < _ q , such t h a t (3.3) holds for all J0>--J, b u t does n o t hold for j0 < J . H e n c e each of the frequencies ~p + fij, j >_ J, appears in t h e p r o d u c t HG = P F e x a c t l y o~ce. B y the l e m m a in w l, this means ~hat Plaebj for j > J , t h u s b y (3.2) a n d (i),

Set

Then

a e ~-- ~tpP for some gp E ~ . (3.4)

G*(z) = a ( z ) _ aAz)e<~p , z >

0(~) = r

F ( z ) = F ( z ) - - ~,(z)e<~ '~>H(z), [I(z) = H(z)dc.(z).

(3.5)

t t / P is t h e reduced form of F / d . (3.6) Indeed, b y (3.5), _F/G = [I/P, a n d since (C/H, P) = 1, (da, P) = 1 means b y (i) t h a t P § da., where d~. = (al . . . . , % 1). However this follows from (ii) a n d (3.4). Since G has p -- 1 terms, the i n d u c t i o n hypothesis shows t h a t P is constant, which contradicts (3.2).

(B) N e x t assume t h a t P is irreducible a n d d G arbitrary. Assume t h a t P ~ d G.

I n particular P ~ constant. W r i t i n g G = dGG1, one can a p p l y (A) to FIG 1 = HdG/P. Hence P is a constant, a contradiction.

(C) Finally, let P a n d dG be arbitrary. I f P = P1 " " " Pr is t h e factorization of P into irreducible factors, t h e n t h e t h e o r e m follows b y applying (B) to each of the equations

( F - ~ P , ) l a = H I P j ( j - - 1 . . . . . ~).

i c j

A n o t h e r application of T h e o r e m 1 is t h e following s t a t e m e n t (Theorem 3), which gives a simple necessary condition for a quotient of two exponential poly- nomials to be entire.

(13)

T H E t ~ I T T T H E 0 ~ E M I N S E V E R A L V A R I A B L E S 279 Given a r b i t r a r y f i n i t e sets B = {/~1 . . . . , fiq}, C = { y ~ , . . . , y,} of points in R ~ we shall s a y t h a t t h e fii's are rational affine combinations of t h e yCs, if for some j0, ko a n d all j

fiJ -- fiJ. = i Wjk(yk -- yko), Wjk e Q. (3.7) k--1

I t is clear t h a t if (3.7) holds for some j0, k0, it holds w i t h suitable r a t i o n a l s wj,~ for a n y o t h e r pair j0, ko. T h e n e x t s t a t e m e n t is an e a s y consequence o f T h e o r e m 1.

THEOREM 3. Let F, G be exponential polynomials such that FIG is entire. Then the frequencies of G are rational affine combinations of the frequencies of F.

T h e p r o o f follows along similar lines as t h e p r o o f o f t h e t h e o r e m in Section 1 of [22].

References 1. AVA~ISSlA~, V., Oral communication, 1970.

2. BERE~STEI~, C. A. & DOSTAL, M. A., Analytically uniform spaces and their applications to convolution equations, Lecture Notes in Math., Vol. 256, Springer-Verlag (1972).

3. --))-- Sur une classe de functions enti~res, C . R . Acad. Sci. Paris, 274 (1972), 1149--1152.

4. --)7-- Some remarks on convolution equations, A n n . Inst. Fourier (Grenoble), 23 (1973), 55--74.

5. --~-- On convolution equations I, " L ' A n a l y s e harmonique dans le domaine complexe", Lecture Notes in Math., Vol. 336, Springer-Verlag (1973), 79--94.

6.--~)-- On convolution equations l I , to appear in "Froceedings of the Conference on Analysis, Rio de Janeiro 1972", (]:Iermana et Cie, Publ.).

7. --)7-- A lower estimate for exponential sums, Bull. Amer. Math. Soc. 80 (1974), 687--691;

cf. also Prelim. I~ep. 73T-B237 in Notices A . M . S . (August, 1973) p. A-492.

8. I)OSTAL, M. A., An analogue of a theorem of Vladimir Bernstein and its applications to

singular supports, P~vc. ;London Math. Soc. 19 (1969), 553--576.

9. EttREI~:PI~EIS, L., Mean-periodic functions I, Amer. J. Math. 77 (1955), 293--328.

10. --))-- Fourier analysis in several complex variables, Wiley-Interscience, New York, 1970.

11. I~UJIWARA, M . , U b e r den MittelkSrper zweier konvexen KSrper, Sci. Bep. Res. Inst. T6hoku Univ. 5 (1916), 275--283.

12. I-ISRMANDER, L., On the range of convolution operators, A n n . of Math. 76 (1962), 148-- 170.

13. --))-- Convolution equations in convex domains, Invent. Math. 4 (1968), 306--317.

14. I~ISELMAN, C. O., On entire functions of exponential type and indicators of analytic func- tionals, Acta Math. 117 (1967), 1--35.

15. LAIRD, P. G., Some properties of mean periodic functions, J. Austral. Math. Soc. 14 (1972), 424-- 432.

16. LAx, P. D., The quotient of exponential polynomials, Duke Math. J. 15 (1948), 967--970.

17. LELO~G, P., Fonctionnelles analytiques et fonctions enti~res (n variables), Les Presses de l ' U n i v , de Montrdal, 1968.

(14)

280 C. A. : B E R E N S T E I N A N D M. A. D O S T A L

18. LINDEL6F, E., Sur les fonctions enti~res d'ordre entier, Ann. Sci. Ecole Norm. Sup. 22 (1905), 369--395.

19. MALGRANGE, B., Existence et approximation des solutions des 6quations aux d6riv6es partielles et des dquations de convolution, Ann. Inst. Fourier (Grenoble), 6 (1956), 271--335.

20. --))-- Sur les 6quations de convolution, Rend. Sere. Mat. Univ. e Politec. Torino 19 (1959/

60), 19--27.

21. MARTINEAU, A., Sur les fonctionnelles analytiques, J. Analyse Math. 11 (1963), 1--164.

22. 1%I~, J. Fo, A factorization theory for functions 2J aieai ~, Trans. Amer. Math. Soc. 29 (1927), 584 596.

23. --~)-- On the zeros of exponential po]ynomials, Trans. Amer. Math. Soe. 31 (1929), 680-- 686.

24. S E L B E E G , I - L , ~ b e r einige transzendente Gleichungen, Avh. Norske Vid. Akad. Oslo, 1/10 (1931), 1--8.

25. S~IELDS, A., On quotients of exponential polynomials, Comm. Pure Appl. Math. 16 (1963), 2 7 - - 3 ] .

26. T ~ v E s , F., Linear partial differential equations with constant coefficients, Gordon & Breach, 1966.

Added in proof (ef. footnote 7):

27. •ITAGAWA, K., Sur les polyn6mes exponentiels, J. Math. Kyoto Univ. 13 (1973), 489-- 496.

28. AVANISSIA~, V. & GAY, ]~., Sur une transformation des fonctionnclles analytiques portables par des eonvexes compacts de C ~, et la convolution d'I~adamard, C. •. Acad. Sei., Paris, 279 (1974), 133--136.

2 9 . - - ~ - - Sur les fonetions enti~res arithm6tiques de type exponentiel et le quotient d'exponentielle-polyn6mes de plusieures variables (ibidem).

17eceived February d, 1974 C. A. Berenstein

University of Maryland I ) e p a r t m e n t of Mathematics

College Park, Maryland 20742, U.S.A.

M. A. Dostal

Stevens I n s t i t u t e of Technology D e p a r t m e n t of Mathemaflics

Hoboken, New Jersey 07030, U.S.A.

Références

Documents relatifs

to estimate RUL of the durational modeling because HMM use exponential time durations, so the new HSMM method is derived by adding a temporal component [1], [2]. Thus,

Then, in Section 5, for s = 1, we will prove that if the Taylor subdivision operator is convergent as a stationary vector subdivision scheme, then H A is convergent in the sense

Cluster representatives and health volunteers in the Region have been assisting in implementing priority health programmes at the community level, while maintaining strong

domain in JR..2, and showed that if the set of local maximum points of the boundary datum consists of N connected components, then the interior critical.. points

3.- Give the opposite 1. In a supermarket, the fresh fruit is usually near the checkout. The fruit is in plastic bags. The smell near the bread makes people feel thirsty. There

Однако ряд особенностей алгоритмов, видных лишь при сопоставлении разных методов решения одной и той же задачи, стал виден только по мере

If ›f is equal to 1, the reinitialization is random. If )f is large, the initial points are distributed based on the initial and convergence points of past searches, which induces

For example, many recent works deal with F I -modules (functors from the category of finite sets with injec- tions to abelian groups, see [1]); finitely generated F I -modules