• Aucun résultat trouvé

P(D) x") x" = O. Rn/W P(D) sx ~ ty P(D) P(D) On fundamental solutions supported by a convex cone

N/A
N/A
Protected

Academic year: 2022

Partager "P(D) x") x" = O. Rn/W P(D) sx ~ ty P(D) P(D) On fundamental solutions supported by a convex cone"

Copied!
40
0
0

Texte intégral

(1)

On fundamental solutions supported by a convex cone

ARNE ENQVIST

1. Introduction

L e t P(D) be a p a r t i a l differential o p e r a t o r in R n w i t h c o n s t a n t coefficients a n d /" a closed c o n v e x cone in R ' . T h u s we a s s u m e t h a t x, y E / ~ a n d s , t > O implies t h a t sx ~ ty E F. T h e p r o b l e m discussed here is to decide w h e n P(D) has a f u n d a m e n t a l solution w i t h s u p p o r t in F.

W h e n F is a p r o p e r cone, t h a t is, w h e n /~ contains n o s t r a i g h t line, this c o n . d i t i o n m e a n s precisely t h a t P(D) is h y p e r b o l i c w i t h r e s p e c t t o t h e s u p p o r t i n g planes of F which m e e t F o n l y a t t h e origin (Gs [4], see also H S r m a n d e r [5, T h e o r e m 5.6.2]). I n t h e o t h e r e x t r e m e case w h e r e /" is a h a l f space s u f f i c i e n t conditions were g i v e n long ago b y P e t r o w s k y (see G e l f a n d - - Shilov [2]), a n d a c o m p l e t e a n s w e r t o t h e q u e s t i o n was o b t a i n e d b y H S r m a n d e r [6]:

I n general t h e i n t e r s e c t i o n F gl ( , _ / ' ) -~ W is a linear subspace a n d x E F implies x ~- y E / 7 for e v e r y y E W. This shows t h a t _/1 is t h e inverse i m a g e in R n of t h e i m a g e V o f 2" in Rn/W u n d e r t h e q u o t i e n t m a p . I t is clear t h a t V is a p r o p e r cone. W e shall use t h e n o t a t i o n s n ' = d i m W , n" ---- n - - n' a n d c o o r d i n a t e s x ~ (x', x") such t h a t W is d e f i n e d b y x" = O. Also for n' > 0 a n d n" > 1 sufficient conditions for t h e existence o f a f u n d a m e n t a l solution o f P(D) w i t h s u p p o r t in F, analogous to t h o s e of P e t r o w s k y for n " ~ - 1, h a v e b e e n g i v e n b y Gindikin [3]. W e shall e x t e n d t h e s e in t h e d i r e c t i o n suggested b y t h e t e c h n i q u e used b y H S r m a n d e r [6]. H o w e v e r , w h e n n " > 1 t h e r e are poly- nomials such t h a t P ( ~ ' , D") is n o t h y p e r b o l i c for a n y ~'. This i n t r o d u c e s a new d i f f i c u l t y a n d in c o n s e q u e n c e o f this t h e r e s u l t is f a r f r o m complete.

I n Section 2 we i n v e s t i g a t e t h e general n e c e s s a r y conditions. T h e m e t h o d s u s e d are v e r y close to t h o s e of H 6 r m a n d e r [6]. I n t h e h y p e r b o l i c case t h e principal p a r t plays a v e r y i m p o r t a n t role. (See L. Svensson [9].) H e r e t h e principal p a r t does n o t give so m u c h i n f o r m a t i o n a b o u t = th e p o l y n o m i a l , a n d we h a v e n o t b e e n able t o f i n d a n y s u b s t i t u t e . H o w e v e r , in S e c t i o n 3 we s t u d y some s t a b i l i t y p r o p e r - ties o f t h e n e c e s s a r y conditions w h i c h allow us to c a r r y t h e m o v e r t o v a r i o u s poly- nomials r e l a t e d t o t h e b e h a v i o r of P s t i n f i n i t y .

(2)

2 A B N E E N Q V I S T

I n Section 4 we reduce t h e existence theorems to a priori estimates. To be able to p r o v e these estimates, we are forced to assume t h a t P(~', D") in some w a y can be w r i t t e n as a p r o d u c t of hyperbolic polynomials. I n Section 5 we give general mfficient conditions of t h a t t y p e a n d in Section 6 we investigate polynomials t h a t are p a r t i a l l y homogeneous.

W h e n deg P = 2, we solve our problem completely. (Section 7.) W e h a v e also tried to f i n d new m e t h o d s to prov e t h e existence of f u n d a m e n t a l solutions.

I n Section 8 we p r e s e n t a constructive m e t h o d in t h e case Of two variables.

The subject of this p~per was suggested to me b y I~rofessor L a r s t t 6 r m a n d e r , whose c o n s t a n t criticism a n d e n c o u r a g e m e n t h a v e been invaluable. I a m also v e r y grateful to h i m for m a n y valuable suggestions.

2. G e n e r a l n e c e s s a r y c o n d i t i o n s

L e t P(D) be a p a r t i a l differential operator in R" w i t h c o n s t a n t coefficients a n d let V be a proper, closed a n d convex cone in R"". Set / ~ = R~'• V, x : (x',x") 6 R " • R"" a n d Ix[ : maxl_<j_<,[xi]. F o r e v e r y real n u m b e r A we define t h e set V* as follows

V* = {v" e R~ <v", z"> ~ A, Ix"[ = 1, x" e V}.

THEORE~r 2.1. Assume that P(D) has a fundamental solution with support in 1".

Then the following condition is satisfied.

(2.1) There is a constant C such that the following is true: Let ~'----> ~"(~') be any analytic function such that P($', ~"(~')) = 0, I m ~"(~') 6 -- V* and [(~', ~"(~'))] < M for all ~' 6 Y2(~ 0, _R) : {~' 6 C ' ; I~' -- $~1 < R} where }~ s R" and R > O. Then m i n ( R , A ) ~ C l o g ( 2 + M ) .

W e need some l e m m a s for t h e proof.

LEMMA 2.2. I f P(D) has a fundamental solution with support in F, then for every compact neighbourhood K of zero there are constants C and # such that

(2.2) ]u(0){ ~

C ~

sup [D~P(D)u[, u E Cg(K).

[at -<~ -1'

Proof. I f P ( D ) E = 8, supp E C / ~ t h e n u(0) = E(P(D)u). Since t h e s u p p o r t of E is a subset of - - / ' , a n d /7 is regular in t h e sense of W h i t n e y , we conclude t h a t (2.2) holds for a r b i t r a r y K . (See Schwartz [8, p. 98].)

LEMMA 2.3. Let P(D) satisfy condition (2,2) and let g 6 C~(K) be equal to 1

(3)

O N F U N D A M E N T A L S O L U T I O N S S U P P O R T E D B Y A C O N V E X C O N E

i n a neighbourhood of zero. Set K ' = (--I") f'l supp (dz). T h e n there exist constants C and #, such that for all u e C~(R ~) with P ( D ) u = 0

(2.3) lu(0)t < C ~ sup ID~ul .

lal ~ # K'

Proof. We apply (2.2) to the function v = Zu, noting that P ( D ) v =

~ ~:o (P~(D)u)D~z/a!.

LE:~MA 2.4. Let ZN(t) ~- 1~ when ltl < 1/2N and zN(t) = 0 otherwise, t E R.

I f q N is the convolution of N factors ZN we have (i) supp ~N C (__ -2,1 1),.

>_0, f

~Ndt = 1,

(ii) dk~lv(t)/dt k is a measure with total mass < (2N) a when 0 < k < 57.

Proof. See the proof Of Lemma 2.2 in HSrmander [6].

In the following lemma we use the notation

~ ) N ( ~ , ) : j ~ - n " N ( ~ I l R ) . . " ~oN(~n, I R ) .

L~MMA 2.5. Let ~o E R ", and let F disc s = Y2(~0, R ). p Set

Then

(2.4)

be a function that is analytic in the poly-

ix'lkiu~(x')I

< ( 4 N / R ) k sup IF(C)l, 0 < k < N.

Proof. See the proof of L e m m a 2.3 in HSrmander [6].

Proof of Theorem 2.1. Consider

I t is clear t h a t P ( D ) u N = 0 and t h a t uP(0) = 1. Let K ' be the set in L e m m a 2.3.

We have to estimate u p and its derivatives in K'. B y hypothesis <x", I m $"(~')>

0 for x E K ' and $' E tg(~ 0,/~), and furthermore there is a constant 6 > 0, t

such that Ix' l > 6 or < x " , I m ~ " ( ~ ' ) > ~ A 6 for all ~ ' e z g ( ~ , / t ) . Set K 1 = { x e K ' ; I x ' l > 6} and g 2 = { x e K ' ; ( x ' , I m ~ " ( $ ' ) > ~ 6 A for all ~ ' e g ( ~ ' 0,R)}.

Using L e m m a 2.5 with k = iV we obtain

[uN(x) I ~ (4-1~/6R) N, X E K r

(4)

AI~NE E N Q V I S T

Differentiation of u ~ gives factors, which are coordinates of ~' or $"(~').

these are bounded by M we obtain

Similarly we obtain

Since sup

ID=u~l ~

C(1 + M)~(4N/aR) ~.

lal -< ~* Kt

sup [Dau~l ~ C(1 @M)*~e -aA.

lal-<~ K,

I f we use this in L e m m a 2.3 and observe t h a t u~r = 1 it follows t h a t (2.5) 1 __< C(1 + M)."((4N/bR) ~v + e-aA).

We choose the integer 32 so t h a t

(~/~)/(8e)

< N < (dR)/(4e) which is possible if R > (Se)/d. Then we obtain t h a t

1 < C(1 + M ) ' e -q~i~(R'~), where ~ 1 = min (1, d/(8e)), and this completes the proof.

The conditions in Theorem 2.1 are hard to apply since t h e y involve r a t h e r general polydiscs in P-l(0). However, when the function $' -~ $"(~') is algebraic of bounded degree and 1" is semialgebraic, it is possible to sharpen these conditions.

(A semiMgebraic set is a set t h a t can be defined by finitely m a n y real polynomial equations and inequalities.)

TI~EOR:EM 2.6. Let qo be an integer and assume that P(D) has a fundamental solution with support in the semialgebraic, closed and convex cone I ~. Then there are constants A o and R o such that i f the function C"" ~ ~' --~ ~"(() E C n" is analytic and algebraic of order ~ q 0 i n Q(~;, /?0) (t0 e R ' ) f and i f P(~', ~"(~')) = 0 in

! r

~(~o, Ro), then there is at least one point ~, C 0($o, -Ro) such that I m $"(~') ~ -- V~o.

We need some preliminaries before the proof. The function Theorem 2.6 shall satisfy the following conditions:

(2.6)

~' --> g"(~') in

~"(~') is analytic in D(~;,R). There are polynomials pj = p i ( $ ' , T), j = n ' q - 1 , : . . , n with ,degpj < q 0 , s u c h t h a t p j ( ~ ' , $ i ( $ ' ) ) = 0 in D(~;, R) and aj(()Aj(r 4 : 0 in D(~0,/~ ). Here aj and Aj denote respectively the coefficient of the t e r m of highest degree and the dis- criminant of pj as a polynomial in T.

Let U, be the set of all ($~,R,A) C R "+2 with If'0[ < t , R < t , such t h a t there exists a function $"(~'), satisfying (2.6) with [~"($')] < t, I m ~"(~') E -- V]

and P($', ~ " ( $ ' ) ) : - 0 for all $' in ~(~0, R), t

LEM~A 2.7. The function

(5)

O N F U N D A M E N T A L S O L U T I O N S S U P P O R T E D B Y A C O N V E X C O N E

f ( t ) : s u p { s ; ~ ( ~ o , R , A ) ~ U , , O < s < A , O < s < R } !

is algebraic for large t.

Proof. I f t h e condition (~, R, A) E U, can be built u p from equations a n d inequalities involving polynomials in real variables, t h e n t h e l e m m a follows f r o m t h e Tarski -- Seidenberg theorem. However, t h e condition (~, R, A) E U, can be expressed in t h e following way:

I~'ol < t, R < t. I n a d d i t i o n there are polynomials pj w i t h d e g p / _ < q0 a n d laj(~')Aj(~')l 2 4= 0 w h e n I~' -- $'01 < R, such t h a t for some ~j0 w i t h ]P~(~0, ~j0)I ~ = 0, j = n ' ~- 1 . . . n, we have I m 9 = I m ( v , , + l , . . . , ~ ) E - - V*, IP(~',~)I 2 ~ 0 a n d I~1 G t for all ~' w i t h

I$' -- ~'01 --< R, if ~i is t h e value for 0 = ~' of t h e unique continuous solution ~ of pi(O, ( r ) = 0 d e f i n e d on the line segment between ~ a n d ~', such t h a t a ~

Tjo

for 0 = ~0, !

B y L e m m a A.9 in H S r m a n d e r [6] this condition can be expressed in t h e required algebraic form.

Proof of Theorem 2.6. T h e o r e m 2.1 shows t h a t f(t) ~ C log (2 + t). Since f is increasing a n d algebraic for large t, lim~_~o~f(t) exists. L3t A o > lim~_~ f(t) a n d set R o = Ao/~, where y is t h e c o n s t a n t we get in L e m m a A2 in [6] for v = n' a n d M ~ deg (-[-[:,+l(ajAi)), whenever Pi are irreducible polynomials of degree q0. This proves t h e theorem, for a polydisc 9(~0, R) w i t h this radius con- !

tains one w i t h radius A o a n d real centre where a i a n d A i do n o t vanish.

COROLLARY 2.8. I f P ( D ) has a f u n d a m e n t a l solution with support in the closed, convex and semialgebraic cone I ~, then P satisfies the following condition:

(2.7) There are constants A o and R o such that i f ~'o E R ~', ~", ~" E C n~, ~" # 0 and f2(~' 0, R0) ~ ~' - ~ ~(~') E C is an analytic function satisfying the equation P(~', ~" ~ ~(~')~") ~ 0 when ~' C ~(~'o, Re), then there is at least one point ~' E ~2(~ o, Re) such that I m (~" ~- ~(~')~") ~ -- V*.

- - n X

R e m a r k 2 . 9 . A s s u m e t h a t P ( D ) E = 5, supp E ~ / " = {x C R ~ xn,+l > ~,+21 jI}.

I f ~(~,,+2 . . . ~,) E %' (R "+1) is t h e Fourier t r a n s f o r m of E w i t h respect to x ~ , + 2 , . . . , x~, we o b t a i n t h a t P ( D 1 , . . . , Dn,+v ~,,+2, 9 9 9 , $n)/~(~,,+2 . . . ~n)--

~ ( x l , . . . , x ~ , + l ) a n d /~ is a n a n a l y t i c function of ~ , + 2 , . . . , ~ w i t h values in ~'(R~'+~). N o w it follows from TrOves [10] (see also Appendix) t h a t the operator P ( D 1 , - . . , D~,+v ~ , , + 2 , - - . , $,) has c o n s t a n t strength. I t is n o t clear if this fol- lows from t h e conditions in Theorem 2.1.

Remark 2.10. I n t h e p r o o f of Theorem 2.1 (and Theorem 2.6) we h a v e o n l y

(6)

6 ARNE ]~NQWST

u s e d t h e e x i s t e n c e of a local f u n d a m e n t a l solution, i.e., a d i s t r i b u t i o n E s u c h t h a t P ( D ) E = ~ in a n e i g h b o u r h o o d of zero a n d s u p p E ~ 1".

W e f i n i s h t h i s s e c t i o n b y p r o v i n g a t h e o r e m w h i c h shows t h a t t h e r e is no re- s t r i c t i o n in a s s u m i n g t h a t F h a s i n t e r i o r p o i n t s . T h u s , we shall a s s u m e this l a t e r on.

THEOI~E~ 2.11. The operator P(D) has a fundamental solution with support in the hyperplane <x, N> --- 0 i f and only if P(~ ~ tN) ~ P(~) for all t E R, ~ E C ~.

Proof. W e c a n a s s u m e t h a t N - - ~ (0 . . . 0, 1) a n d P(D) = ~,~ a1(D')D..i L e t E be t h e f u n d a m e n t a l solution. L o c a l l y we c a n w r i t e E ~-- ~,o E2(x') | D~, ~(x,).

F r o m P ( D ) E = 6 we o b t a i n t h a t

ai(D,)Ek={(~o(X') for i ~ - O

+~=~ for i > O.

H e n c e , ao(D')E o = 8(x') a n d , if m > 1, am(D')E ~ = O, a,,(D')E~_ 1 ~- a,~_I(D')E, ~-

j - i - 1 ,

~- 0 . . . B y i n d u c t i o n it follows t h a t a a (D)E~_i = 0. T h u s 0 = (a,,(D'))~+lao(D')E o = (am(D'))~+l~(x'), w h i c h implies t h a t a~ = 0 for m > 1.

3. Stability of the necessary conditions

W e shall h e r e p r o v e a t h e o r e m w h i c h gives s o m e i n f o r m a t i o n a b o u t P a t i n f i n i t y .

THEOREm 3.1. Let P(D) satisfy (2.1) with respect to 1" ~ R ~" • V. Assume that ~j E R ~, s 1 ,t i e R, a i E C,, where sj and tj -+ ~ , j - + ~ and that there is a constant N such that 1~1] ~ t~, tj ~ s~, sj < t~ for all j. Furthermore, assume that

Qi( ) = + s/', + t/") j +

I f ~' E R ~" and Qo(~', ~") # 0 for some ~" E R ~", then Qo@, ~") ~ 0 for all ~" E C ~"

with I m ~" E - - i n t V*.

Proof. L e t ~' E R n' be s u c h t h a t deg~, Q0(~0,' ~") > 1 a n d a s s u m e t h a t ~" E {3n', I m $ " E - - i n t V* a n d Q0(~, $") - - 0.

L e t v E C a n d t a k e N " E R ~" s u c h t h a t d e g ~ Q 0 ( ~ , ~ " ~ ~:N") > 1 a n d let bi(~' ) b e t h e coefficient of t h e t e r m o f h i g h e s t degree w i t h r e s p e c t t o 9 of t h e p o l y n o m i a l Qy(~', ~" -~ zN"). N o w , consider Qj(~', ~" ~- zN") as a p o l y n o m i a l o f (~', ~) a n d w r i t e it as a p r o d u c t o f i r r e d u c i b l e f a c t o r s . Set d i : t h e p r o d u c t o f t h e d i s c r i m i n a n t s o f t h e s e f a c t o r s , c o n s i d e r e d as p o l y n o m i a l s of T. W e h a v e t h a t A 1 : Ay(~' ) : bi(~')di(~' ) ~ 0 is a p o l y n o m i a l of $' a n d if ~ : ~(~') is a c o n t i n -

(7)

O N F U N D A M E N T A L S O L U T I O N S S U P P O R T E D B Y A C O N V E X C O N E

uous solution of t h e e q u a t i o n Qj(~', ~" + ~N") = 0, t h e n z(~') is a n a l y t i c w h e n Aj(~') 4=- o.

L e t B c r be a ball w i t h centre ~ e I l"' a n d radius r > 0. Then b y L e m m a A2 in [6] there is a b a l l B o c B w i t h centre ~ ' e R " a n d radius y r > 0, such t h a t Ao(~' ) 4= 0 in B o. F u r t h e r m o r e , for e v e r y j there is a ball Bj C B 0 w i t h centre ~ E R ~' a n d radius y2r > 0 such t h a t Aj(~')4= 0 in 17i. This implies t h a t there is a ball /~ c B 0 w i t h real centre a n d radius y2r/2 = ylr > 0 a n d a subsequence Aik of A i such t h a t Aik(~' ) 4= 0 in /3 for all k. I n order n o t to complicate t h e n o t a t i o n s we assume t h a t this is t r u e for t h e whole sequence, i.e., Aj(~') ~ 0 in /~ for all j > 0 .

Now, consider t h e solutions of t h e e q u a t i o n Q0(~:, ~" -t- "oN") = 0, $' E/~.

These are a n a l y t i c in ~t a n d if the radius r > 0 above is small enough, t h e r e is a c o n s t a n t e > 0 such t h a t some of t h e solutions, s a y %, satisfies t h e condition I m (~" + Zo($')N ~) e -- V* for all ~' e B. I f ~0($') has t h e m u l t i p l i c i t y /, w h e n

$' E/~ a n d if U c 13 is a small n e i g h b o u r h o o d of zero t h e n Rouch6's t h e o r e m shows t h a t t h e e q u a t i o n Qj(~', ~ " + T N " ) = 0, ~ ' E / ~ has e x a c t l y /z solutions in z0(~') + U for large j . L e t Tj be one of these. T h e n Tj --> z0 u n i f o r m l y on /3 when j --> oo.

if j

a n d

This implies t h a t

I m (~" + zj(~')N") E -- V~2 for all is large. Thus, for large j

P(~ + sj~', ~: + tj(( +

~ / ( ) N " ) ) = o,

~'eB

for all ~ ' e / ~

I m (~: + tj(U + ,j(U)N")) e -- V*~/2 for all ~' 6 / ] .

ff

I t follows t h a t (2.1) is n o t valid since t h e n we would h a v e rain (tje/2, sj~lr) < C log (2 + [$j[ + Cl(sj + tj)) =

= 0 ( m i n (log tj, log sj)), j ---> 0%

which implies t h a t

rain (s, r) = O.

This completes t h e proof.

COROLLARY 3.2. Let P(D) have a fundamental solution with support in 1" =

= lt" X V and let p be the principal part of P. Then for every ~' e tl n" we have that p(~', D") is either hyperbolic with respect to V or zero.

(8)

A R I~TE :E]~Qu

4. Reduction of existence theorems to a priori estimates

I n this section we shall assume t h a t P satisfies t h e n e c e s s a r y conditions of T h e o r e m 2.1. L e t k E ~)/(R n) so t h a t for some c o n s t a n t s C a n d N

k(~ § V) --< k(O(1 + C l~I) ~.

(See p. 34 in [5].) I f 1 < p ~ oo we shall use t h e n o r m

( f

IluH~,~ = ( 2 ~ y " l k ( ~ ) a ( ~ ) l P d ~ , u e S ( R ~

As beibre, let /~ ~ R n" • V be a c o n v e x cone w i t h i n t e r i o r points. (Cf. T h e o r e m 2.11.) T h e n we are i n t e r e s t e d in t h e q u o t i e n t n o r m s d e f i n e d b y

l]ullpr.~k=inf{llvllp, k; v = u on / ' _ = - - / ' , v e ~ ( R " ) } , u e ~ ( R " ) .

Tm~o]cEM 4.1. Let P be a p o l y n o m i a l and a an element of c)s such that for every compact set K there is a constant C for which

U F

]l ]]1.1 <_ C ][P(n)u]l[-1/,, for all u C C : ( K ) .

Let 1 < pj .< ~ .a n d .kj e Oi(R"), . j = 1, 2, . T h e n for every f E B~,kj(R"~176 ~), j = 1, 2 . . . . , with s u p p f c F there is a solution u to P ( D ) u ~ f such that

E B l~ / ~ ' ~ . . . .

s u p p u C / ~ a n d u ~j,~ki~.,/, j--~ 1 , 2 ,

T h e p r o o f is r a t h e r long so we f i r s t p r o v e t h e following local version.

T ~ E o ~ E ~ 4.2. Let a e =X(R'), ~ c c R" and let P be a p o l y n o m i a l for which there is a constant C such that

U E

]] Ii~,~ ~ C I]P(D)u]]~-~/, for M1 u e C : ( - - / 2 ) .

T h e n there is a u C B| with s u p p u c I ~ such that P ( D ) u = ~ in ~ .

Proof. T h e e q u a t i o n P ( D ) u = (~ in Q m e a n s t h a t ~(P(D)v) -~ v(O) for all v e C ~ ( - - 9 ) . W e h a v e t h a t

Iv(0)] < HvHr-~ < C I]P(D)vH~-~/,, v e C : ( - - Y 2 ) . T h e n t h e linear f o r m

P ( D ) v - + v(O) on P ( D ) C ~ ( - - t 2 )

can be e x t e n d e d b y t h e H a h n - B a n a c h t h e o r e m to a linear f o r m g on C~(R~) such t h a t

I?~(W)] ~ CllWl]FI,-1/a

T h u s u is a d i s t r i b u t i o n w i t h s u p p o r t in /~, such t h a t P ( D ) u ~- 5 on .Q n a p u C Bo~,..

(9)

O ~ ~ U ~ D A M E ~ T A L S O L U T I O N S S U P P O R T E D B Y A C O N V E ~ C O N E

Using local f u n d a m e n t a l solutions of t h e t y p e i n t r o d u c e d in T h e o r e m 4.2 we can p r o v e t h e following a p p r o x i m a t i o n t h e o r e m .

T~EOREM 4.3. Let [22 C [22 be bounded open sets in R ~ and let P satisfy the conditions of Theorem 4.2 for every [2 c c R ~. Set _ F ~ int T' and denote by N 1 the set of solutions u C C+(12i) of the equation P ( D ) u = O, such that supp u c D 1 [3 F. Furthermore, let N i have the topology induced by C~ I f

# E ~g'(R"), F ~ 13 supp/~ ~ ~ , 1. ~ [3 s u p p P ( - - D ) # c ~ [22 ~ 1"~ Cl supp # c ~ [22, then the restriction to I-21 of the elements in N 2 f o r m a dense subset N': of N~.

Proof. I f we p r o v e t h a t e v e r y v C c~)'(~(~1) which is o r t h o g o n a l to N~ is also o r t h o g o n a l t o N 2, t h e s t a t e m e n t will follow f r o m t h e H a h n - B a n a c h t h e o r e m .

L e t [2a C D~ be o p e n b o u n d e d sets such t h a t [22 C C [2a a n d G [24 -t- [2a c 0 /~a. F u r t h e r m o r e , let E be a local f u n d a m e n t a l solution in [2a, i.e. P ( D ) E ~-- in [ 2 a , s u p p E C F . Set u = ~ 0 * E w h e r e ~0 E C~(1. [3 (0122) [3123). T h e n P ( D ) u ~--0 in s a n d s u p p u c F. T h u s

0 = v(~v 9 E ) = qp 9 E * ~(0) ~ (E 9 v)(~v), q~ E C F ( F N (C[22) Ct [23)"

L e t Z C C~~ be i in a n e i g h b o u r h o o d of ~ 2 a n d set # ~ Z ( ] ~ . v). T h e n F ~ gl supp # C ~ , F ~ Cl supp P ( - - D ) # C C ~2"

T h u s b y h y p o t h e s i s

/~~ I3 supp # C C [22.

H e n c e we can choose ~o E C~([21) such t h a t ~0 ~ 1 in a n e i g h b o u r h o o d of F ~ I f we set #2 = ~v# ~ yJ(/~*v) a n d v 1 ~ - P ( - - D ) # 2 we o b t a i n t h a t #1 E ~8'([21), v - - v 2 E ~'([21) a n d _P~ [3 supp (v - - v2) = O, which implies

( v - - v 2 ) ( u ) = 0 a n d v2(u ) = # 2 ( P ( D ) u ) ~-- 0 for all u E N 1. T h u s v(u)--~ 0.

LEMMA 4.4. There is a set [2 C R ~ such that i f D r ~ v[2 then the p a i r of sets D r and [2+: satisfies the hypotheses in Theorem 4.3 for all v > 1.

Proof. L e t p be t h e principal p a r t of P a n d let c o > 0. Set N 0 = (0, No),

,, . . ,, . r

N 2 ~ (Ni, coN1), . , N , = (N',,~oN~), w h e r e N o , . . , N ~ E i n t V * a n d p , 0 for j--~ 1, 2 . . . . , n. F u r t h e r m o r e , we choose t h e v e c t o r s Nj so t h a t e v e r y ~ E R n can be w r i t t e n as a linear c o m b i n a t i o n o f - - N o , N 1 , . . . , N , w i t h n o n - n e g a t i v e coefficients. (Observe t h a t this c o n d i t i o n is i n d e p e n d e n t o f co > 0.) Set

Q - - ~ { x E R " ; <x, N j > < ] , j : l , 2 , . . . , n, <x, N 0 > > - - I }

a n d D~ = v[2. W e shall p r o v e t h a t t h e s e sets will do if co > 0 is large enough.

(10)

l 0 A R N E ENQVIST

I t is clear t h a t s is a b o u n d e d n e i g h b o u r h o o d o f zero. W e n o w w a n t to s h o w t h a t

# e ~3'(R~), f ~ N s u p p # c D~+2, F ~ 13 s u p p

P(--D)/~ c c ~9~ ==>

f o 13 s u p p # C C ~ .

Set

Uj(s)

: {x 6 R ~ ; x" 6 i n t V, <x, N i > > s},

j

= 1, 2, . . . , n. T h e n t h e s t a t e m e n t a b o v e is a c o n s e q u e n c e of t h e following:

# 6 % " ( R ~ ) , # : 0 in

Uj(s), P ( - - D ) / ~ - O

in

Uj(t)=>#---- 0

in

Ui(t )

(j = 1 , 2 , . . . , n).

F r o m T h e o r e m 5.3.3 in H h r m a n d e r [5] we n o w get t h a t it is s u f f i c i e n t t o p r o v e t h a t e v e r y c h a r a c t e r i s t i c p l a n e t h a t i n t e r s e c t s

Ui(t )

also m e e t s Uj(s). A p l a n e t h a t does n o t m e e t

Ui(s )

h a s a n o r m a l t h a t lies in t h e d u a l cone U~ o f

Ui(O ).

H o w e v e r , U* is t h e c o n v e x hull of

{hNj ; h >

0} tJ F * , since t h i s is closed.

T h u s we h a v e t o p r o v e t h a t if

N = ( O , N " ) C F *

t h e n

p ( h N j - k N ) 4:0

for all h > 0. ( N o t e t h a t a p l a n e w i t h n o r m a l in F * is d e f i n e d b y a n e q u a t i o n in t h e x"

v a r i a b l e s , so it m e e t s

Uj(s)

if a n d o n l y if it m e e t s

Ui(t). )

H o w e v e r (4.1)

p ( h N j ~ - N ) = h m p ( N ~ , c o N j ' + h - ~ N ") =#0

f o r large o~ a n d h > 0.

To see t h i s we f i r s t o b s e r v e t h a t

h-iN" 6

V*, Nj' 6 V* a n d

p(N~, D")

is h y p e r - bolic w i t h r e s p e c t t o V b y R e m a r k 2.10 a n d T h e o r e m 3.1. I t follows f r o m T h e o - r e m 1.3 in [9] t h a t also

p(N~, --iD")

is h y p e r b o l i c w i t h r e s p e c t t o V, w h i c h b y T h e o r e m 5.5.4 in [5] implies (4.1).

Proof of Theorem

4.1. L e t ~O b e t h e sets d e f i n e d in L e m m a 4.4 a n d l e t

~% 6 C~~ be 1 in a n e i g h b o u r h o o d o f ~9~_ r T h e n it will be s u f f i c i e n t t o p r o v e t h a t t h e r e e x i s t % 6 [-]~o Blo~ pj, akj s u c h t h a t

II~,~(u~+, - u~)ll~j,o~ _< 2 -~, ~ _< ~, j _< ~,

P(D)u~+ x = f

in f 2 + 2 a n d s u p p u ~ + 1 c F.

I n f a c t , f o r s u c h u~ we o b t a i n t h a t u - - > u in

B~aki

as v - + oo, w h e r e

P(D)u ~ f

a n d s u p p u C / ' .

S e t u 1 =- E 9 (~3f), w h e r e E is a local f u n d a m e n t a l s o l u t i o n in a large neigh- b o u r h o o d of zero. T h e n

ui 6 rl~ Bpi, ak i, P(D)ul = f

in /2 2 a n d s u p p u 1 C /7.

W h e n u~ . . . . , u~ are chosen, we w a n t t o c o n s t r u c t u~+x. T h e n t h e r e is a dis- t r i b u t i o n

Vo 6 ['1~ Bpi,,k j

s u c h t h a t

P(D)v o ~ q~+af ~ f

in ~ + 2 , s u p p v 0 c F.

I f v = v 0 - - %, t h e n

P(D)v =- P(D)v o -- P(D)u~

= 0 in D~+I- Choose ~ 6

C~(F)

so t h a t s u p p to c --~62 1 a n d H~,(to * v - - v)[[vj, akj ~< 2-~-~ for # _< v, j < v. I t follows f r o m T h e o r e m 4.3 t h a t t h e r e is a C ~ f u n c t i o n w s u c h t h a t

P(D)w ---- 0

in f 2 + : , s u p p w c F a n d I I ~ , ( w - t o * v ) ] l e i , , k

i _ < 2

-~-1 f o r # < ~ , j < v . Set u~+ l : v o - w . T h e n

(11)

O~ F U N D A M E N T A L S O L U T I O N S S U P P O R T E D B ~ A C O ~ V E X C O ~ E

II~,(U.+l - u.)ll,j..kj -< 2-". /~ _< ~, j < ~, a n d

supp %+~ C F.

This proves t h e theorem.

P(D)%+~ = f in ~9+ 2

11

5. Sufficient conditions

I f t h e operator P(~', D") is hyperbolic w i t h respect to V for sufficiently m a n y

~' E r t h e n we can prove t h e existence of a f u n d a m e n t a l solution w i t h support in F = R =" • V. The following definition helps us to describe sets which suffice.

Definition 5.1. L e t 0 < ~ < 1 a n d let B C C " be a ball w i t h radius R a n d centre ~ . W e s a y t h a t a subset S of B is of type N~ if for an a r b i t r a r y , logarithmicMly plurisubh~rmonic f u n c t i o n g > 0 t h e following i n e q u a l i t y is t r u e

g((o) <-- (sup g)~-~(sup g)~

B S

(Cf. L e m m a 3.2 in [6].)

THEOREM 5.2. Let P be a polynomial in n = n' ~- n" variables and let N =

= (O,N#), where N " E V * = V*. Furthermore, let a($') ~ 0 be the coefficient of the term of highest degree of P($ + "vN) as a polynomial in T. Then the operator

l o c ~ a n "

P(D) has a fundamental solution E E B~o,a with s u p p E ~ / ' = x V if P satisfies the following condition:

(5.1) There are constants R > O, A and 8, 0 < (5 < 1, such that in every ball B with radius R and real centre ~0, t there is a subset S~o, of type N~ for which it is true that ~' E S~o,, I m p " E - - V~ ~ P ( $ ) ~ : 0 . I t is sufficient to prove t h e t h e o r e m ibr A = --1. (Cf. page 349 in [6].) F o r 1 < p < ~ , kE~)~(R ~') we set

Hul{pv-~ = inf{llvllp, k; v = u on V_ = - V , v eS(R~")}, u e 3 ( R ~ ' ) , where

\l/p

LEMMA 5.3. Let Q be a polynomial in n" variables with Q(~") # 0 when I m ~" E -- V*I and let ]c E ~?~(R~'). Then there is a constant C depending only on A, n" and deg Q such that

V _

Ilullp,~k < CllQ(D")ull~-k for all u E C~(Rn~).

(12)

12 A R N E E N Q V I S T

P r o @ Set

= o f )

Then s u p p E ~ V, Q(D")E = ~ and

,E(~,, _< C f *~(e")]/0(~") de",

where C only depends on A, n" and deg Q, t h a t is IIEIl~.g~ < C. I f g = Q(D")u on V_, then E . g = u on V_ so t h a t

Ilu]l~-~ _< liE * gig, ~ --< C Ilgllv, which shows t h a t

v_ < C "

IIullp, ~ _ IIQ(D )ullv.~ for all u ~ C~(R~").

LE~MA 5.4: I f h(~',x") is analytic in ~' and k C%~(R ~') then ]lh(~', ")llev,-~

is logarithmically plurisubharmonic.

Proof. Ilh($', ")llp~ is the norm of an analytic function with values in a Banaeh space, hence it is logarithmically plurisubharmonic.

LEMMA 5.5.

If

the polynomial P satisfies condition (5.1) with A = --1 and K c C R ~, then there is a constant C such that

sup II~(U, .)lIpv-~ < C sup [[P(~', D")~(U, ")llp~-~

R n t R n '

for all u e C : ( K ) , k e r Here ~(~', .) is the Fourier transform of u with respect to x'.

Proof. For ~'E St,, we have

ia(U)I II~(~', ")I[pv,-~ ~ C lIP(U, D")~(U, ")[[p,v-k < C G

where G = supRn, [IP(~ ', D")~(~', ")[Ip, k. The first estimate follows from L e m m a v_

5.3 and the second from the fact t h a t the function

~'-~ HP(r D")~(r .)[/,~-~

is logarithmically plurisubharmonic and of exponential type. The properties of the set S~0. now show t h a t

,, _ _ "'lr[V- "11--~/'~,~ <

la(~;)[ Ilu(~0, ")]LF,r~ < C ( s u p ra(~')l I1~(~', ,,~p,k, - _

]~'-$o'] _<R

_ < C ( s u p la(~')]

I1'~(~', ~,,~.k~

. ~ ' l V - ~ l - ~ G a , 0 < 6 < 1, I I n t

(13)

ON FUNDAMENTAL SOLUTIONS SUPPORTED B ~ A CO:XW-EX CONE 13 which implies t h a t

la(~')I I1~(~', .)11~-~ _ CO, ~' r R ' . H o w e v e r , b y L e m m a A 1 in [6] t h e r e is a 0 E A such t h a t

a(~') <_ 6' a(~' + zo) < C'

[a(~' + z0)I, z e C, [zl = 1.

T h u s

Since Ilu(~' +

zo,

.)11~ is s u b h a r m o n i e as a f u n c t i o n o f z we n o w o b t a i n t h e re- q u i r e d e s t i m a t e .

LEM~rA 5.6. Let P be a polynomial and b C ~ ( R " " ) . I f for every compact set K, there is a constant C 1 such that

(5.2) sup/1~(~', ")tip, v_ bk ~ C1 s u p I l P ( ~ t , D")~(~', 9 )II~,~, v_

i~.n" Rrt"

for all u e C~(K), k C ~)s then for every k C 7s

n)

there is a constant C such that (5.3)

Ilull~,

r_ ~ <_ c llP(D)ul]~- k for all u e C~o(g).

Proof. See t h e p r o o f o f T h e o r e m 3.10 in H S r m a n d e r [6].

Proof of Theorem 5.2. T h e t h e o r e m follows i m m e d i a t e l y f r o m L e m m a 5.5, L e m m a 5.6 a n d T h e o r e m 4.1.

6. Partially homogeneous operators L e t P be h o m o g e n e o u s in t h e ~" variables, t h a t is

P(~) = ~ a ~(~ )~ 9 ' "

lal ==

I f P(D) has a f u n d a m e n t a l solution w i t h s u p p o r t in A = {x ~ R n ; x n , + l >

c ~ , + 2 IxjI}, c > 0, t h e n b y R e m a r k 2.9 we k n o w t h a t t h e o p e r a t o r P ( D 1 .. . . . D~,+> ~,,+2, 9 9 9 , ~ ) has c o n s t a n t s t r e n g t h . This implies t h a t b(~') = a(~,0 ... 0)(~') is s t r o n g e r t h a n % for all ~, i.e. t h e r e is a c o n s t a n t C such t h a t

~ ( ~ ' ) < C b(~') for all ~ a n d all ~ ' C R ~'.

(6.U Set

= • = {Q ; Q(~") = limj.+~ (P(~j, ~")/b(~j)) for some sequence ~j E R "' such t h a t d ( ~ , b-l(0)) --~ m a s j - ~ m},

(14)

14 A R N E : E I ~ Q V I S T

w h e r e d(~j, b-a(O)) is t h e distance in C" f r o m ~ t o t h e zeros o f b. W e a s s u m e here t h a t b is n o t a c o n s t a n t , for this w o u l d i m p l y t h a t P is a p o l y n o m i a l in ~"

which is a trivial case. T h e set 2; is a c o m p a c t subset of t h e set o f all p o l y n o m i a l s o f degree m. F u r t h e r m o r e , we h a v e

THEORElV[ 6.1. I f the polynomial P(~) : ~1~[:0~ a~(~ )~ ' "~ satisfies (2.1) and ao~

is weaker than b for every a, then Q(D") is hyperbolic with respect to V for every Q in X~,.

Proof.

(Cf. the p r o o f of T h e o r e m 2.2 in H S r m a n d e r [7].) If P ( ~ ,

~")/b(~]) -+ Q(~"}

as j - + oo, t h e n w e c a n a s s u m e that e v e r y coordinate of ~i has fixed sign. I n p

order n o t to complicate the notations w e a s s u m e that all these coordinates are n o n - negative. It follows f r o m the Tarski-Seidenberg t h e o r e m that if Q ( ~ " ) = Z q ~ "~

t h e n

a ! ! . ,

(6.2) i n f { v I ;27 ] ~(V )/b(~7 ) -- q~L ~ <= 1/t, d(v', b-l(0)) ~ t, V~ ~ O , . . ~], ~ 0}

is an algebraic f u n c t i o n of t for large t. B y r e p e a t e d use of t h e same t h e o r e m we get t h a t t h e i n f i m u m o f n o n - n e g a t i v e ~2, w h e n t h e i n f i m u m in (6.2) is a t t a i n e d , is a n algebraic f u n c t i o n o f t for large t, a n d so on. F o r large t we h a v e t h e P u i s e u x series e x p a n s i o n

k~

v ' ( t ) =

0 9

--oo

a n d P(~'(t),~")/b(~'(t))-+Q(~") as t - + ~ . L e t s ~ O a n d consider P(#'(t) ~- t ~ ', ~")/b(~'(t)) = ~ . ( ~ a~)(#'(t))t~!zl~'~/fi!)~"~/b(~'(t)).

a t~

Since a s is w e a k e r t h a n b a n d d(v'(t), b-l(0)) ~ t it follows f r o m T h e o r e m 3.3.2 a n d L e m m a 4.1.1 in [5] t h a t a~)(#'(t))/b(~l'(t)) : O(t-I~l), t - + ~ , if fi 4: 0. T h u s , if 0 ~ s ~ 1 we h a v e

P(~l'(t) -~- t ~ ', ~")/b(~'(t)) :

= P(~'(t) + t ~ ', t~")/t"b(~'(t)) --+ Q@'), t -+ ~ . N o w t h e t h e o r e m follows f r o m T h e o r e m 3.1.

W e are going to s t u d y t w o special cases of partially h o m o g e n e o u s operators.

First w e will consider the case n" = 2 a n d later o n operators, s u c h that

Q(D")

is

strictly

hyperbolic for all Q c Xp. T h u s , let n" = 2 a n d P(~) = ~r~1=~ a~(~')~"%

First w e reformulate the necessary conditions.

T~EORE~ 6.2. Let P ( r Z a ( ~ )r where b : a(m,o ) is stronger than aa for all ~. Assume that P satisfies the necessary conditions of Theorem 2.6 with

(15)

ON ~ U N D A M E N T A L S O L U T I O N S S U P P O R T E D B Y k C O N V E X C O N E 15 respect to R " • W and let V = {x" ~ R ~ ; x++~ ~ Co ~ be a cone ~ W such that every Q ~ Z is hyperbolic with respect to V. T h e n for every c ~ c o there are constants R o and t such that i f P(~', ~(~'), 1 ) = 0 in a ball B with real centre ~' and radius R o, where ~ is analytic, then we have I m ~(~') = 0 for some ~' ~ B . Furthermore, l~(O')l ~ c i f d(O',b-~(O))~_t and O ' e B .

Proof. Assume t h a t d(~:, b-~(0)) - + ~ a n d P ( ~ , ~")/b(~:) --+ Q(~") E Z as v - + zc. I f P(~: + $', ~,(~'), 1) = 0 for ]~'] ~ R 0, t h e n we o b t a i n from Rouchd's t h e o r e m (cf. t h e p r o o f of T h e o r e m 3.1) t h a t there is a subsequence of %(~'), which converges to 2 as ~ - + 0o, u n i f o r m l y for I~'] ~ no, where Q(2, l) = 0, so t h a t

12[ ~ c 0. This proves the last s t a t e m e n t . W r i t e P as a p r o d u c t P(~) = b(~') I ~ (;,,,+~ -

~(;')~~

1

F r o m T h e o r e m 2.6 w e n o w obtain that there are constants A a n d R o s u c h that for e v e r y ~,,+2 E R there is a point ~' w i t h I~' -- ~'i ~ R o s u c h that I m ~,+2T(~') ~_ A . If w e let ~++2--> + oe w e see that there is a point ~' w i t h

I~'-- ~'I ~ R o s u c h that I m ~ ( ~ ' ) ~ 0.

THEOREM 6.3. Let P ( D ) satisfy the conclusions of Theorem 6.2 and set W ---- {x" C R 2", x,,+x ~_ c 4 ix,,+21}, where c is the constant in Theorem 6.2. T h e n the

nlor with support in A R " • W.

operator P ( D ) has a f u n d a m e n t a l solution E E - +, b

The p r o o f is similar to t h e p r o o f of Theorem 1.1 in H S r m a n d e r [6]. W r i t e P($', $,,+1,1) as a p r o d u c t of irreducible factors a n d let A be t h e p r o d u c t of their dis- criminants, w h e n t h e y are r e g a r d e d as polynomials of $++v Set R ~ ( b ( ~ ' ) A ( ~ ' ) ) M

where M is a large integer a n d set for v C C ~ ( R ~)

N~(v, ~') : m a x IR(C')bC)I [ l ~ (D++I -- Ti(C')D~'+2)v(C,^ ' ")l[pS,~

1

where t h e m a x i m u m is t a k e n over all labellings of t h e zeros ~i" (k E ~)s LnMMA 6.4. The f u n c t i o n N~(v, ~') is logarithmically plurisubharmonic.

Proof. See t h e p r o o f o f L e m m a 3.6 in H 6 r m a n d e r [6].

LEMM~_ 6.5. Let v C C~~ n) with v(x) = 0 for Ix' I ~ H . T h e n N ( V , ~') ~ e I-/lIm ~'1 SUp ~ 7 ( v , ~ ' ) , ~ --- O, 1 . . . m .

a n '

Proof. See t h e p r o o f of L e m m a 3.7 in H S r m a n d e r [6].

(16)

16 AI~NE ENQVIST

L~MMA 6.6. Let P(D) be as in Theorem 6.2. Then for every compact set K there is a constant C such that

s u p N , ( v , ~ ' ) < C s u p N ~ + l ( v , ~ ' ) ,

vEC~(K),

v = O, 1 , . . . , m - - 1.

i~.rd Rn ~

Proof. B y L e m m a A1 in H 6 r m a n d e r [6] we first choose a finite set A c R"' for polynomials of degree < d e g P . L e t B = R 0-~ 3. F o r e v e r y ~ ' C R n" a n d B > 0 we can f i n d a 0 E A such t h a t t h e distance from ~'~-zO to the zeros of R is a t least B w h e n [ z I = B . L e t

0 • = { ~ ' @ z O ; I z l = B , I m z ~ 0 }

a n d denote b y zg~, j = 0, l, 2, t h e points a t a distance < R o -~ j from O • I n w h a t follows we shall work in t h e sets /2+; t h e same a r g u m e n t s can be applied in t h e sets Y2j .

F r o m L e m m a 5.3 above a n d L e m m a 3.2 in H S r m a n d e r [6] we o b t a i n t h e fol- lowing estimates

sup IR(~')b(~')[ II ]z~ ( O - % 1 _ "vJ(~')Dn'+ 2)~)(~', ")Hp, w-k <~

D+ 1

(sup N v ( V ,

~t))X--d( SUp

[R(g')b(g')l ][ ~ (D,,+I -- ~j(~')Dn,+2)v(~ , ^ ' ")llp,~)~ ~ ~<

Im rv+ 1 ($') =0

v + l

__< (sup N(V, ~,))l-a( SUp ]R(~')b(~')] ]1 ~-~ (D,r -- 5(~')Dn'+2)v(g,~ ' ")[11,,~)~ a

Im vv+ 1 (~')=0

< (SUp Nv(V, ~'))l-~(sup N.+x(V , ~'))~ < C (sup N ( v , ~'))l-~(sup N~+l(V , ~'))~.

H e r e 0 < 6 < 1; t h e first i n e q u a l i t y follows from L e m m a 3.2 in H S r m a n d e r [6], t h e second is a consequence of L e m m a 5.3, a n d the f o u r t h follows f r o m L e m m a 6.5.

I f we n o w t a k e t h e m a x i m u m of t h e left side over all labellings of t h e zeros we o b t a i n t h a t

sup N~(v, ~') < C (sup ~V(v, ~'))~-~(sup N~+~(v, ~'))~.

DO+ R n' Rn t

The same e s t i m a t e holds for /2 o a n d since the f u n c t i o n N~(v, ~') is plurisub- harmonic, we can use the m a x i m u m principle for the f u n c t i o n N~(v, ~ ' + zO).

We o b t a i n t h a t

sup N~(v, ~') < C (sup/V~(v, ~'))~-~(sup ~+~(v, ~'))~,

Rr~P R n ' R n '

i.e.,

(17)

O N F U N D A ~ I ] ~ N T A L S O L U T I O N S S U P P O R T E D B Y 2~ C O N V E X C O N E 17 sup

N,(v, ~') < C

sup N +~(v, ~').

L ] ~ M ~ 6.7.

Let P(D) be as in Theorem

6.2.

Then for every compact set K there is a constant C such that

sup _R(~')g(~') I1~(~ ', .)H~.W~ ~ C sup ]7(~') IIP(~', D")~(~', .) H~,W~, if

v ~. C~(K).

R n ' B n '

Proof.

R e p e a t e d use of L e m m a 6.6 gives t h a t

sup IR(~')b(~')i Ii~(~', ")Jl~,~ = sup

No(V, ~') <

R n t R n '

< C sup

N,,(v, ~') = C

sup IR(~')I liP@,

D");;(~',

.)][p wZ <

l l n ' l t n '

_~ C sup _R (~') HP(~ ',

D")~(~',

.)[I,,WT~ = C G.

R n #

B y L e m m a 6.5 we h a v e

IR(~')b(C')I

II~(~', ")lip,w% _~

C G

for jim ~'l --~ const.

L e t A be t h e set we get f r o m L e m m a A1 in H S r m a n d e r [6] w h e n we a p p l y it to polynomials of degree _~ deg

(Rb).

F o r e v e r y ~' there is a 0 C A such t h a t

/~(~')g(~')

<_ C ~(~' -~ zO)g($' + zO) <_ C' ]R(~' ~- zO)b($' @ zO) l

for Izl = 1. This gives t h a t

~(~')b(~') II~(~' @

zO,

.)Hv,Wz < C G, lz[ = 1, which implies t h a t

where

~(~,)g(~,) ,A , jjv(~, ")11~,r < c G,

C is i n d e p e n d e n t of ~'.

Proof of Theorem

6.3. L e m m a 6.7 a n d L e m m a 5.6 show t h a t for e v e r y c o m p a c t set K a n d e v e r y k E ~ ( R n) there is a constant C such t h a t

[lullp,~gk ~ C IlPKn)ull~.~k, A_

u C C~(K),

so the t h e o r e m follows f r o m T h e o r e m 4.1.

We h a v e not been able to prove a n y analogues of T h e o r e m s 6.2 a n d 6.3 for a r b i t r a r y n", so we m u s t p u t some e x t r a conditions on t h e polynomials w h e n n" > 3. Therefore, we shall now s t u d y polynomials satisfying t h e following con- dition:

(18)

1 8 A R N E E N Q V I S T

(S) P(~)--~ ~.1~1-~ a~(~') ~'~ has real coefficients,

b--~ a(~.0 ... 0) is stronger t h a n a~ for all er a n d Q(D") is strictly hyperbolic w i t h respect to V for e v e r y Q E 2:.

L e t W be a convex cone such t h a t int W D V ~ {0} a n d set A --~ R"' • W. T h e n we can prove

THEOR]~M 6.8. I f P satisfies condition (S) then the operator P(D) has a fun- damental solution E E B 1~ oo, Y, with support in A.

F o r technical reasons we will consider Q o ( E ) - - P ( ~ ' , ~ " - - i N " ) where N " E i n t W*.

TH:EOR]~ 6.9. Let K C R ~ be a compact set and k E ~)<(R").

polynomial introduced above, then there is a constant C such that IlulI~,~ok ~-- C tiQo(D)uI[~,-k, u E C~(K).

I f Qo is the

Proof of Theorem 6.8. Set k(~) ~- 1/Q0(~ ). T h e n it follows f r o m T h e o r e m 6.9 a n d T h e o r e m 4.1 t h a t Qo(D) has a f u n d a m e n t a l solution E 0 E Bl~ w i t h s u p p o r t hi~ w i t h sup- in A. This implies t h a t P(D) has a f u n d a m e n t a l solution E E ~|

port in A. (Cf. page 349 in H S r m a n d e r [6].)

L~MMA 6.10. Assume that Q(D") is strictly hyperbolic with respect to V for every Q in X. Then there is a neighbourhood of Z, in the set of all real polynomials of degree m, such that all polynomials in this neighbourhood are strictly hyperbolic with respect to the cone W i f i n t W ~ V ~ { 0 } .

Proof. W e h a v e to prove t h a t there is a n e i g h b o u r h o o d of 27, in t h e set of all real homogeneous polynomials of degree m, such t h a t e v e r y p o l y n o m i a l Q in this n e i g h b o u r h o o d satisfies t h e following conditions:

Q(N") # 0 for all N " E W * ~ { 0 } .

The e q u a t i o n Q(E" + -oN") = 0, where N" E W*,

JN"] ~-- 1, ~" E R'", ]~"] = 1 a n d N" _1_ ~", has o n l y real simple zeros.

However, this is obvious for e v e r y p o l y n o m i a l in 27 satisfies these conditions a n d 27 is compact. (Cf. D e f i n i t i o n 5.5.1 in H S r m a n d e r [5].)

L e t P satisfy condition (S) a n d set Q~(~) = I(P(E', ~" - iN") + P(~)(~', ~" -- iN")) where N" E i n t W* a n d a - - ~ ( a ' , 0). The first step in t h e p r o o f of T h e o r e m 6.9 is t h e following lemma.

(19)

O ~ F U n D A M E n T A L S O L U T I O N S S U P P O R T E D B Y A C O ~ V E X C O ~ E 19 LEMMA 6.11. There is a constant C such that

9 I ~ _ #

I1~(~', )11~, ~0,~ ~< c ~ IIQ~<(8, D"),7(~',

.)ll,,~~

~=(~', o)

for all ~' ~ R ~', u~C~~ ~) and k~%(R~').

To prove this we need a l e m m a due to L. H S r m a n d e r .

LEMMA 6.12. Let s C C" be a convex set and let Q1 . . . Qm of degree < 2 5 , such that Qj(~) =4= o, j : 1 , . . . , m, when ~ ~ ~ . f u n c t i o n s gl . . . . , g,,, that are analytic in ~ , such that

and

1 = Z'Qj(~)gj(~), ~ C ~,

m

lm(~)l < C l ~ tQx(~)l, ~ex~.

1

Here C is a constant that depends only on m and 25.

be p o l y n o m i a l s T h e n there are

Proof. The v a r i a t i o n of t h e a r g u m e n t of Q1 is a t m o s t ~(N 1) w h e n ~ E 12, for i f $1, $ 2 e a 9 t h e n Qj(t~lJr- ( 1 - t ) ~ 2 ) = a I I ( t - & ) , where & ~[O, 1]. Since .(2 is s i m p l y connected we can choose a n a n a l y t i c b r a n c h of 01IN wj in ~ . W e o b t a i n t h a t t h e v a r i a t i o n of t h e a r g u m e n t of O1./N -t/# is < az(N ~ 1)/N < ~. T h u s t h e r e are constants aj C (3,. tajl ~- 1, such t h a t Iarg (ajQ)/N)] < u ( N ~ 1)/2N < ~ / 2 w h e n ~ E ~9. This implies t h a t there is a c o n s t a n t c > 0 such t h a t

c QIlN

I ; I < l~e (a~Qjn,) < tQSI, ~ e ~.

Set q : ~ "~ aj~gj ,~IlN . T h e n

(6.3) c Z IQjl 1IN < R e q ~_ lqi ~ X

IQjl 'IN

a n d

q=N (E

ajQ) N) = Qjhj.

1 1

T h u s

1 : ~ Qj(~)gj(~), ~ E D where gj(~) = hj(~)fl(~) -m~.

T h e f u n c t i o n hj is a s u m of t e r m s of t h e f o r m II(akQ~lN), where t h e p r o d u c t s consist of N m - N factors. Thus, according to (6.3)

lhji ~ C'xlq[ Nm-N,

m

Igj(~)l ~ C~lq(,~)l -~ ~ Cl(~, IQjl),

1

which shows t h a t

where t h e c o n s t a n t C depends o n l y on N a n d m.

(20)

20 ATONE ENQVIST

Proof of Lemma

6.11. We are going to a p p l y L e m m a 6.12 to t h e polynomials Qa"

Since

(a(fl(~')/b@)) -+ 0

if I~[ 4= 0 a n d

d(~',

b-~(O)) - + oo, we o b t a i n from L e m m a 6.10 t h a t there exists a c o n s t a n t t > 0 such t h a t Qa(~', ~") g= 0 if I m ~" E N" --

- - i n t

W*

a n d d @ , b-~(O)) >_ t. L e m m a 6.12 shows t h a t if d(~', b'~(O)) > t t h e n there are a n a l y t i c functions ga, ~', such t h a t

1 = SQa(~', ~")go,.~,($"), I m U' ~ - - W * , a n d

lga,~'(~")l <-- C/X

]Qa(~', r ~

C~/Oo(~', ~"),

I m ~" E - - W * . (Cf. t h e p r o o f of T h e o r e m 5.5.7 in [5].) Set

f"

" " "

T h e n we see i m m e d i a t e l y t h a t Ea, ~, e Boo, ~0, w i t h [lEa, ~,l],, ~o b o u n d e d b y a con- s t a n t t h a t does n o t d e p e n d on ~'. B y changing t h e i n t e g r a t i o n contour we o b t a i n t h a t supp Ea, ~, ~ W. Moreover,

~ ( ~ ' , x " ) =

~, Ea,~,*(Qa(~',D")~@,')),

u E C : ( R " ) .

a = ( a ' , o)

I f ha. ~, e S a n d ha, r =

Qa(~', D")a(~', x")

on W_ t h e n

h~,(x") = X(Ea, ~, .ha,~,)(x" )

= a(~', x") on W_. This implies t h a t

ll~(~', 9 )1[~,~0~ _< IIh/ll, ~;k ~ - < 211Ea,~, * ha, ~;111, ~ok <: _

C X h a,~'[I,,k"

'h

T h u s

a=(a', 0)

for all ~' E t l ~' w i t h

d@,

b-l(0)) > t. W h e n ~' C t l ~' is a r b i t r a r y , t h e n b y L e m - m a A 2 in H S r m a n d e r [6] there is a point 0' = 0'@) C R ~' w i t h [0'[ =<

t/7

a n d d(~' + 0', b-l(0)) > t. This implies t h a t

IJ~(~', 9 )1/2, ~o<~,+o,, .)~ -< 6' z ][Qa(~' + 0 , D")~(~', ~ - ' ")[[~, ~. ~ - Thus

~ a 2 03 u E C~~ ~) a n d k E ~2((Rn'), for all ~ ' E R '+,

N o w we w a n t to f i n d a b o u n d for t h e r i g h t - h a n d side of the estfmate in L e m m a 6.11. L e t t be t h e c o n s t a n t i n t r o d u c e d in t h e p r o o f of t h a t lemma. Thus we h a v e Qa(~', ~") :V 0 for all ~ if

d(~',

b-1(0)) ~ t a n d I m ~" C _N" -- i n t W*. F u r t h e r - more, let 0 ~ z E C ~ ( { x ' C R n ' ; Ix'I < 1}) a n d set Zo'(~')= Z ( ~ ' - - 0 ' ) . I f r

(21)

O N F U N D A M E N T A L S O L U T I O N S S U P P O R T E D B Y h C O N V E X C O N E 21 we define )/~,(~') b y setting Zo'($') = Xo,( I~e ~'). W h e n 0' 6. T = {~' 6. R~';

d ( ~ ' , b - l ( 0 ) ) > t + 2} and f i = (fi',0), Ifil = 1 we define

~(v) = ~(~, ~, o'; v) = ( ~ ) - " f xo,(~')Q~)(~)~(~)/Q~(~)d~, v 6.

C ~ ( R " ) .

J

LEMMA 6.13. Let E be defined as above. T h e n E 6. B 1~ and supp E C A = It ~' X W. Furthermore, i f q~ 6. C~~ ") then Ilq)EIl.,l is bounded uniformly with respect to 0'.

Proof. B y changing the integration contour in the immediately t h a t supp E C A.

Let K be a compact set and assume t h a t f l = (1,0 . . . . ,0).

then

~" variables, we obtain I f v 6. C ~ ( K )

f ~ '~

/~(v) = (2~)-"'

zo,(~ )v(~)/(~l- ~j(~~ =

rj

+

J

~j

where ~ o = ( $ e , . . . , ~ , ) and y j ~ ' R + i if I m v j < 0 and y j ~ R - - i if I m D > 0. Furthermore, ~j denotes the support of 0Zo,/0~ 1 between Y1 and R.

We get the estimate

I (v/1 _< f ,o f +

r1

f f

+ C . s u p I~(~1, ~~ d~ ~ < 6'1 I~(~)1 d~,

where C 1 is independent of 0', for fl~(~ + i~)l d~ _< e ~'~/l~(~)l d~.

We also need the equivalence between two norms t h a t we are going to use.

The following lemma, which is inspired b y Beurling [1], proves this.

LEMMA 6.14. Let K ' C R " b e a compact set and let ~v C C~(R"') be 1 on K ' . I f Xo, is the f u n c t i o n defined above then there is a constant C such that

supli~(~', 9 )lh,z _< c sup sup [l(v5 9 ~

(xo,u))(~,

~ ' ")lh,~

R n z l t n r o' E T

for all u e C ~ ( K ' X l t " ) and lce ~)~(R"").

(22)

22 ARNE ENQVlS~

Proof. L e t g be an element in the unit ball of the dual space of C~~ " ) e q u i p p e d w i t h the norm II " [ll.w~ a n d set v(x') = (u(x', "), g>. I f we p r o v e t h a t

(6.4) sup I~(~')t _<< C sup sup l(v~ 9 (Zr v 9 C~(K'),

R n~ R n~ O' ~ T

t h e s t a t e m e n t will follow as g varies.

A s s u m e t h a t (6.4) is false. Then there is a sequence v i 9 C~(K') such t h a t s u p Iv~(~')l = 1

R n ~

and

[(~ 9 (;Co.~j))(~')[ __< 1/j for all ~' 9 R": a n d 0' 9 T.

T a k e ~ 9 R " such t h a t 1~i(~)[ = 1 a n d set ~vi(~' ) = ~i(~' 3- ~ ) . T h e n we h a v e t h a t

I~/(~')[ < l~vi(O)[ = 1 for all ~ ' 9 R ~' a n d

] f ~(~' -- ~l')Z(~' 3- ~i -- O');vi(~')d~' <-- 1/j for all ~ ' 9 R '+ a n d 0 ' 9 T.

W e can n o w choose a s u b s e q u e n e e ~oik of wi such t h a t ~vik converge u n i f o r m l y t o h on e v e r y c o m p a c t set. Then h is analytic, ]h(~')] < [h(0)[ --~ 1 for all ~' 9 R ' . F u r t h e r m o r e , b y L e m m a A2 in H S r m a n d e r [6] we can choose a sequence 0~ 9 T such t h a t ~ ; - O; is b o u n d e d . Finally, choose a s u b s e q u e n e e of ~ ; k - O;k which converges t o --0' 0. T h e n

f w(~' ~')Z(~' Oo)hOl')d~l ' ~ 0 for all ~' 9 R"'

This implies t h a t ~ 9 ~Y-~(Zoo,h ) ~ O, so Xoo,h - - O, for ~-l(;G,h ) is analytic.

H o w e v e r , h is also analytic a n d ;~o0, ~ 0, so h---0, which contradicts t h a t lh(0)l = 1. This proves the lemma.

N o t e t h a t we h a v e only used t h a t T is defined b y some polynomial b 4= O, w i t h given degree. Thus, the constant C in the l e m m a depends only on the degree o f the polynomial defining T and not on the polynomial itself.

I f k 9 t h e n there are constants U a n d N such t h a t k(~ 3 - ~ ) (1 3-Cl~[)N/c(~) for all ~ , ~ 9 R n. F o r given C a n d N we shall here use the n o t a t i o n ~X(R n, C, N ) for all k 9 Oi(R ~) such t h a t k(~ 3- ~) < (1 3- C l~})Nk(~) for all ~, ~ / 9 R". W e shall n o w use L e m m a 6.13 a n d L e m m a 6.14 to p r o v e

LEMMA 6.15. Let K G R n be a compact set, C o, N C R+ and let Q~ be as before.

Then there is a constant C such that

(23)

O~N : F U I ~ D A M E ~ T A ~ L S O L I ) ' T I O ~ S S U P X : ' O t t T E D B Y A C O I N ~ V E X C O k N E 2 3

liQ~(D)ull~,-k <_ CllQo(D)ull~_k for all u C C~(K) and k E ~ ( R ' , C

O , N ) .

Proof.

L e t

u C C~(K)

a n d ~v E C ~ ( R ' ) , where ~ = 1 in a n e i g h b o u r h o o d of K.

F u r t h e r m o r e , let ~ E C~(R") be 1 in a n e i g h b o u r h o o d U of zero, such t h a t ( ( C U ) ~ K ) f l s u p p ~ v - ~ O. I f E = E ( ~ , fi, 0') is the distribution in L e m m a 6.13, Ifl] = 1, a n d S D g =

Q~(D)u

in A_, t h e n

~v((~vE) 9 g) = ~vZo,(D')Q~)(D)u

in A_.

Thus

II~y.o,(n')Q~)(D)ull~:-k < ]]~v((~E)

9 g)lll,~ _< Ilwlll, ~ llqEll~,x HgIi~, ~ ~< C [iglll,~, which implies t h a t

D' (~) D u

z_

livzo,( )Q~ ( ) ill,

~ <_ c

liQ~(D)ulI~,-~.

N o t e t h a t C is i n d e p e n d e n t of 0 ' ~ T. H o w e v e r , with Q ~ ) = Q~)(~', D"), sup I1(~ * (~o,Q~)~))(~', ")11~,~% _< c~ llv Zo,(n')Q~)(D)ull~,-~ <-- C2

llQ~(n)ull~-~.

I n view of L e m m a 6.14, this implies t h a t

sup [IQ~)(~ ', n")t~($',-)i11,~% < C IiQ=(n)ull~,%

1~. n~

i . e . ,

/ .

I1 (8) , *- ] ,

sup

Q~ (2, D")~(~',

.)Ill, k < C

IIQ~(~, D")R(~', ")l[~%d~'.

R n p J

B y using t h e technique in the p r o o f of T h e o r e m 3.10 in t t S r m a n d e r [6] we o b t a i n from this t h a t

H e n c e

[lQ~)(D)ullzl._k <_ C [IQ~(D)u[l~,-k.

IIQ~+~(D)ull~-k = 1

II(Qo(D)

-4-

2Q~)(D)

-- Qo@(D))uII1.A-k <---

<_ lIQo(n)uli~-k d-

IlQ~)(n)ulllA_k d- ilQ~')(D)UlilA,-k ~<

<~ C(llQ~(n)ull~-~ §

[IQ0(D)ull~-k).

This proves the lemma, for ~ = (~', 0) and fi = (fi', 0), lfil = 1 are arbitrary.

Proof of Theorem

6.9. B y L e m m a 6.11 we have

~Y' t

, , ")Ill, k,, < ~ z

fIQAD)ulr~-~,,

li~(~', ) i l l , ok" _< c liQ~(~, n")4(~',

~ -

(24)

2 4 AI~ l~7:E ENQVIST

w h e r e k , / ( ~ " ) = k(~', ~"). H o w e v e r , t h e r e are c o n s t a n t s C o a n d N such t h a t kr, E ~ ( ( R ~', C 0, N ) for all V'. T h u s L e m m a 6.15 implies t h a t for all V'

~ =

)11 , Q0 + _

By

using t h e t e c h n i q u e in t h e p r o o f of T h e o r e m 3.10 in [6] we o b t a i n f r o m this t h a t

Iju]l~:5, k < C IjQo(D)u[]~-k, u E C : ( K ) .

W e shall n o w i n v e s t i g a t e in detail t h e n e c e s s a r y conditions for p o l y n o m i a l s P s a t i s f y i n g t h e following condition:

( 6 . 5 ) - - a~(~ )~ , ' w h e r e b@) ~- a(m ' 0 .. . . . 0) (~') is real a n d

s t r o n g e r t h a n a~ for all ~. F u r t h e r m o r e , a s s u m e t h a t Q(~") is s t r i c t l y h y p e r b o l i c w i t h r e s p e c t to e v e r y v e c t o r in t h e i n t e r i o r o f t h e p r o p e r , closed a n d c o n v e x cone V* for all Q E 27, a n d t h a t N " --~ (1, 0, . . . , 0) E i n t V*.

N o t e t h a t we can m a k e b real b y m u l t i p l y i n g t h e p o l y n o m i a l b y t h e c o m p l e x c o n j u g a t e o f b. Observe also t h a t , since 2: is c o m p a c t , t h e r e is a smallest, p r o p e r , closed a n d c o n v e x cone V* D/V" s a t i s f y i n g t h e c o n d i t i o n in (6.5).

T~]~o~]~M 6.16. Let P be a polynomial satisfying (6.5). Then the following con- ditions are equivalent:

(i) P ( D ) is an evolution operator with respect to every half space containing R n' • (V ~ {O}) in its interior.

(ii) I m P(~) is dominated by P(~).

(iii) P satisfies the necessary conditions of Theorem 2.6 with respect to I " - ~ R n" • W for every cone W such that i n t W D V ~ ( 0 } .

W e are going t o p r o v e t h e t h e o r e m a f t e r some lemmas.

LENMA 6.17. Let a and b be polynomials such that a(~j)/b(~j)-+O, j--> oo, for all sequences ~j E R n such that d(~j, b-l(0)) --> ~ , j -+ oo. Then a is dominated

by b.

Proof. Set e(t) = sup {la(~)/b(~)l ; d(~, b=l(0)) ~ t > 0}. I t follows f r o m t h e a s s u m p t i o n t h a t e(t)-+ 0 as t--> oo. F r o m L e m m a A2 in [6] we o b t a i n t h a t t h e r e is a c o n s t a n t 7 > 0 such t h a t for e v e r y ~ E R n t h e r e i s a 0 E Rn,.TJOI < 1, such t h a t t h e d i s t a n c e f r o m ~ ~- tO t o t h e zeros o f ab is a t least t. N o w we h a v e t h a t

Références

Documents relatifs

Rédacteurs :Mohamed (le temps des pyramides), Besnik (Le Kosovo), Jean-Pierre (regard sur la dernière séance), Nelly (la dernière séance en émotions) Nelly et Christine (à chacun

Bochner's second theorem generalizes the classical Riemann removable singularity theorem.. Sharp results for this case are included in the results of

Sur la fonetion

Using rearrangements of matrix-valued sequences, we prove that with certain boundary conditions the solution of the one-dimensional SchrSdinger equation increases or

For stable spaces (~hat is, for spaces isomorphic to its square) one has the following stronger result.. isomorphic for every

In w 2 we give a defini- tion of microlocal approximability and also extend the definition of local approx- imability to pseudodifferential operators.. These

The class of differential operators of constant strength with variable coefficients (Definition 2.2 below) is closely related to operators with constant coeffcients..

B., The Cauchy problem with polynomial growth conditions for partial differential operators with constant