• 検索結果がありません。

NON-UNIQUE SOLVABILITY OF A CAUCHY PROBLEM FOR THE WAVE EQUATION IN QUASI-ANALYTIC ULTRADISTRIBUTION CATEGORY (Microlocal Analysis and Related Topics)

N/A
N/A
Protected

Academic year: 2021

シェア "NON-UNIQUE SOLVABILITY OF A CAUCHY PROBLEM FOR THE WAVE EQUATION IN QUASI-ANALYTIC ULTRADISTRIBUTION CATEGORY (Microlocal Analysis and Related Topics)"

Copied!
10
0
0

読み込み中.... (全文を見る)

全文

(1)

NON-UNIQUE SOLVABILITY OF A CAUCHY

PROBLEM FOR THE WAVE EQUATION IN

QUASI-ANALYTIC

ULTRADISTRIBUTION

CATEGORY

TAKASHI TAKIGUCHI (滝口孝志)

Department ofMathematics, National Defense Academy

(

防衛大学校数学教育室

)

Introduction

In this article, we prove non-uniqueness in an overdetermined Cauchy problem

(1) $\{$

$\frac{\partial^{2}u}{\partial t^{2}}-\Delta u=0$,

$\partial_{x}^{\alpha}u|_{x=x_{0}}=u_{\alpha}(t)$ for any $\alpha$,

where Ais the Laplacian on $\mathbb{R}^{n}$, $n\geq 2$

.

This is an inverse problem to reconstruct the wave from observation at

one

space point. This problemwasfirst introduced by L.Ehrenpreis [E], who proved uniqueness in this problem in distribution category, employing expansion by harmonic functions. As for uniqueness, FJohn [J] also proved it globally with respect to general real analytic time-like curves. For distribution solutions, another uniqueness result was proved by M.Nacinovich [N] in adifferent way. In 1993, S.Tanabe-T.Takiguchi [TT] proved that

(2) $\{\begin{array}{l}\frac{\partial^{2}u}{\partial t^{2}}-\Delta u=0\partial_{x}^{\alpha}u|_{x=x_{0}}=0\mathrm{f}\mathrm{o}\mathrm{r} \mathrm{a}\mathrm{n}\mathrm{y} \alpha\end{array}$

would imply that $u=0$ in aneighborhood of$x=x_{0}$ if$u$ is

a

non-quasi-anaIytic $(‘ \mathrm{N}\mathrm{Q}\mathrm{A}’$

for short) ultradistribution. In the

same

article, they introduced acounterexample by A.Kaneko which yields that uniqueness in this Cauchy problem does not hold for hyperfunctions.

For uniqueness in the Cauchy problem (1), the

case

where $u$ is aquasi-analytic $(‘ \mathrm{Q}\mathrm{A}’$

for short) ultradistribution is left open, which we study in this article

数理解析研究所講究録 1261 巻 2002 年 123-132

(2)

Ultradistributions

In this section,

we

review the definition of

ultradistributions.

Let $\Omega\subset \mathbb{R}^{n}$ be

an

open

subset and $M_{p},p=0,1$, $\cdots$, be

a

sequence ofpositive numbers.

Definition 1. $f\in \mathcal{E}(\Omega)=C^{\infty}(\Omega)$iscaUed

an

ultmdifferentiable fimction

of class $\{M_{p}\}$

(resp. $(M_{p})$) iffor any compact subset $K\subset\Omega$ there exist constants $h$ and $C$ (resp.

for any $K$ and for any $h>0$ there exists

some

$C$) such that

$\sup_{x\in K}|D^{\alpha}\varphi(x)|\leq Ch^{|\alpha|}M_{|\alpha|}$ for all $\alpha$

holds. Denote the set of the

ultradifferentiable

functions ofclass $\{M_{p}\}$ (resp. $(M_{p})$)

on

$\Omega$ by $\mathcal{E}^{\{M_{\mathrm{p}}\}}(\Omega)$ (resp. $\mathcal{E}^{(M_{\mathrm{p}})}(\Omega)$) and denote by

$D^{*}(\Omega)$ the set of$\mathrm{a}\mathrm{L}$ functions in

$\mathcal{E}^{*}(\Omega)$

with support compact in $\Omega$, where

$*=\{M_{p}\}$ or $(M_{p})$

.

For

a

compact subset $K\subset\Omega$ let

$D_{K}^{*}=\{\varphi\in D^{*}(\mathrm{R}^{n})$; suppf $\subset K\}$,

and

we

define

$D_{K}^{\{M_{\mathrm{p}}\},h}=$

{

$\varphi\in D_{K}^{\{M_{\mathrm{p}}\}}$ ; $\exists C$ such that

$\sup_{x\in K}|D^{\alpha}\varphi(x)|\leq Ch^{|\alpha|}M_{|\alpha|}$

}.

These spaces

are

endowed with natural structure oflocally

convex

spaces. For $\mathrm{N}\mathrm{Q}\mathrm{A}$ class,

we

impose the

following conditions

on

$M_{p}$

.

(M.0) (normalization)

$M_{0}=M_{1}=1$

.

(M.I) (logarithmic convexity)

$M_{p}^{2}\leq M_{p-1}M_{p+1}$, $p=1,2$,$\cdots$

.

(M.2) (stability under ultradifferential operators)

$\exists G$, $\exists H$ such that

$M_{p} \leq GH^{\mathrm{p}}\min_{0\leq q\leq p}M_{p}M_{q-p}$, p$=0,$1, $\cdots$

(3)

(M.3) (strong non-quasi-analyticity)

3$G$ such that $\sum_{q=p+1}^{\infty}\frac{M_{q-1}}{M_{q}}\leq Gp\frac{M_{p}}{M_{p+1}}$, $p=1,2$,

$\cdots$

(M.2) and (M.3)

are

often replaced by the following weaker conditions respectively; (M.2)’ (stabilityunder differential operators)

$\exists G$, $\exists H$ such that $M_{p+1}\leq GH^{p}M_{p}$, $p=0,1$,$\cdots$

(M.3)’

(non-quasi-analyticity)

$\sum_{p=1}^{\infty}\frac{M_{p-1}}{M_{p}}<\infty$

.

We note that if $\sigma>1$ then the Gevrey sequence

$M_{p}=(p!)^{\sigma}$

satisfies all the above conditions. For

more

details about NQA ultradifferentiable func-tions and NQA ultradistributions confer [Kol] and [Ko2].

In this article,

we

study QA ultradistributions. Let $N_{p}$, $p=0,1$,$\cdots$ , be asequence

of positive numbers. We impose the following conditions ((QA) and (NA)) instead of (M.3)

or

(M.3)’;

(QA) (quasi-analyticity)

$N_{p}\geq p!$, $p=0$,1, $\cdots$ , $\sum_{p=1}^{\infty}\frac{N_{p-1}}{N_{p}}=\infty$

.

Let $N_{p}$ be

asequence

ofpositive numbers

$\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\infty \mathrm{i}\mathrm{n}\mathrm{g}$ $(QA)$

.

If

$\lim\inf$ 0

$parrow\infty$

then $\mathcal{E}^{\{N_{\mathrm{p}}\}}$ is the class ofanalytic functions. We imposethe conditionthat $N_{p}$ does not

define the analytic class;

(NA) (non-analyticity)

$\lim_{parrow\infty}\sqrt{\frac{p!}{N_{p}}}=0$

.

(4)

If the sequence $N_{p}$ satisfies (M.I) and (QA), the sets $D^{(N_{\mathrm{p}})}$

and $D^{\{N_{p}\}}$

are

{0}

(cf. [C]$)$, however, we defifine the sheaves $D^{*}$

of $\mathrm{Q}\mathrm{A}$

ultradistributions

of class

$*$, where

$*=\{N_{p}\}$ or $(N_{p})$.

For asequence $M_{p}$ ofpositive numbers, we define its

associated$fi^{\mathrm{g}}nctions$

.

For $t>0$,

let

$\overline{M}(t):=\sup_{k}\frac{t^{k}}{M_{k}}$,

$M(t):= \sup_{k}\log\frac{t^{k}}{M_{k}}$,

At’(t) $:= \sup_{k}\frac{t^{k}k!}{M_{k}}$

.

Definition 2. $f\in D^{(M_{\mathrm{p}})’}$ (resp. $f\in D^{\{M_{\mathrm{p}}\}’}$) if

$f$ is expressed by the boundary value

of the holomorphic functions,

$f(x)=F_{1}(x+i\Gamma_{1}0)+\cdots+F_{m}(x+i\Gamma_{m}0)$,

where $i:=\sqrt{-1}$, $\Gamma_{j}$, $j=1$ ,$\cdots$ ,$m$

are

open

cones

in $\mathrm{R}\mathrm{n}$,

$F_{j}\in O(\{z\in \mathbb{C}^{n}$ ; $z\in$

$\mathbb{R}^{n}+i\Gamma_{j}’$, $|{\rm Im} z|<\exists\epsilon\})$, $j=1$,$\cdots$ ,$m$, for which, for any compact

set $K\subset \mathbb{R}^{n}$ there

exist constants $L$ and $C$ (resp. for any $L>0$ there exists $C$)

such that

$\sup_{x\in K}|F_{j}(x+iy)|\leq C\overline{M}(L/|y|)$

.

Note that, in $\mathrm{N}\mathrm{Q}\mathrm{A}$ case, this defifinition is equivalent

to the

one

by the duality (cf. [KO1]$)$

.

For

a function

defined

on

$\mathbb{R}^{n}$, its Fourier-Laplace transform is

$\hat{f}(\zeta):=\int_{\mathrm{R}^{n}}f(x)e^{-:x\cdot\zeta}dx$, $\zeta\in \mathbb{C}^{n}$

.

The Paley-Wiener theorem for $\dot{\mathrm{N}}\mathrm{Q}\mathrm{A}$

ultradistributions are

provedby H.Komatsu

(The-orem

1.1 in [Ko2]$)$. We extend this theorem for QA

ultradistributions

which

are

not

hyperfunctions. Note that the Paley-Wiener theorem for hyperfunctions

are

known (Theorem 8.1.1 in [Ka])

(5)

Proposition 3. (the Paley-Wiener theorem for ultradistributions) Assume that a se-quence$M_{p}$

of

positive numbers

saiisfies

(M.O), (M.1), $(M.2)$

and (NA). The following

conditions are equivalent.

i) $\hat{f}$ is the Fourier-Laplace

transform of

$f\in \mathcal{E}_{K}^{(M_{p})’}$ ( resp. $f\in \mathcal{E}_{K}^{\{M_{\mathrm{p}}\}’}$), where $\mathcal{E}_{K}^{*}$

is

the set

of

$ultradistr\dot{\mathrm{v}}butions$

of

the $class*whosesuppo\hslash s$ are contained in $K$

.

$\mathrm{i}\mathrm{i})$ There exist $L>0$ and $C>0$ (resp. $/or$ any $L>0$, there exists $C>0$) such that

$|\hat{f}(\xi)|\leq C\overline{M}(L|\xi|)$, $\xi\in \mathbb{R}^{n}$

and

for

any $\epsilon>0$ there exists $C_{\epsilon}$ such that

$|\hat{f}(\zeta)|\leq C_{\epsilon}\exp(H_{K}(\zeta)+\epsilon|\zeta|)$, $\zeta\in \mathbb{C}^{n}$

where

$H_{K}( \zeta):=\sup_{x\in K}{\rm Im} x\cdot\zeta$

is the support

function

of

$K$.

$\mathrm{i}\mathrm{i}\mathrm{i})$ There exist $L>0$ and $C>0$ (resp. $/or$ any $L>0$, there exists $C>0$) such that

$|\hat{f}(\zeta)|\leq C\overline{M}(L|\zeta|)e^{H_{K}(\zeta)}$, $\zeta\in \mathbb{C}^{n}$

.

The proof of this Proposition is obtained by modifying the proof ofTheorem 1.1 in [K02]. For this modification, we apply the Paley-Wiener theorem for hyperfunctions (Theorem 8.1.1 in [Ka]) and an estimate

$M(L| \zeta|)=\sup_{k}\frac{L^{k}|\zeta|^{k}}{M_{k}}\leq C\sup_{k}\frac{(\epsilon|\zeta|)^{k}}{k!}\leq Ce^{\epsilon|\zeta|}$.

Uniqueness ofafunction with analytic parameters In this section, we review the results ofthe following problem.

Problem 4. Let $f$ be a

function

defined

onRn. Assume that $f$ contains $x’$

as

analytic parameters at $x=0$, where $x=(x’, x’)\in \mathbb{R}^{n}$ and that the $restr\dot{\tau}ctions$ to $x=0$

of

$f$ and all its derivatives in $x’$ vanish;

$\partial_{x}^{\alpha},,f|_{\{x=0\}}=0$ for all $\alpha$

.

Under these conditions, judge whether $f=0$ in

some

neighborhood

of

$x=0$

.

The

answer

to this problem depends on the class where $f$ belongs and is closely

related to the uniqueness in (1), which we introduce in this section.

If $f$ is aNQA ultradistribution, the

answer

to Problem 4 is positive (cf. [B1], [TT]).

Applying this result, S.Tanabe-T.Takiguchi proved uniqueness in ($1\rangle$ in NQA

ultradis-tribution category

(6)

Theorem

5. (Theorem6.2 in [TT])

Assume

thatu is aNQA

ultradistribution

satisfying (2). Then u $=0$ in

some

neighborhood

of

{x

$=x_{0}\}$

.

The proofofthis theorem is too short and eaey to omit, which

we

introduce.

Pmof.

Since $\mathrm{a}\mathbb{I}$

conomak to $\{x=0\}$

are

non-characteristic

with respect to the

wave

operator, $u$contains$x$$\mathrm{a}\mathrm{e}$analytic

parametersat $x=0$

.

Thereforethe

answer

toProblem

4 proves the theorem. $\square$

It is also known that uniqueness is proved for NQA

ultradistributions even

if the

parameter $x’$ is weakened to QA

one

(cf. [B2]).

The

mswer

to Problem 4 is negative when $f$ is

a

hyperfunction. This

case

there is

a

famous counterexample by M.Sato (cf. Note 3.3 in [Ka]). J.Boman proved that

the

answer

to Problem 4is negative when $f$ is

a

QA

ultradistribution

by modifying

M.Sato’s counterexample (cf. [B3]).

The idea of J.Boman’s extension is the foUowing.

Assume

that $N_{p}$ satisfies (M.O),

(Af.1), (M.2)’, (QA) and (NA). Let

$E:=$

{

z

$\in \mathbb{C}$ ; $|z|<1$, Imz $\neq 0$

}.

Take such polynomiak$p_{k}(z)$ which approxinate $1/z$ unifomly in the wider

sense

in $E$

that

$|F( \tau, z)|\leq C_{r}M^{*}(\frac{r}{|{\rm Im} z|})$,

for $\forall r>0$, $\exists C_{r}$, where

$F( \tau, z):=\sum_{k=0}^{\infty}\frac{p_{k}(z)}{k!}\tau^{k}\in O((\mathbb{C}\backslash (-\infty, 0])\cross \mathbb{C})$

$F$ is a defining function of

a

QA

ultradistribution

$f$ of class $\{N_{p}\}$,

$f(\tau,x)=F(\tau,x+i0)-F(\tau,x-:0)$,

containing $\tau$

as

aholomorphic parameter. It is not

difficult to construct acounterex-ample in (Np) claae aPplying the inclusion relation between $\{N_{p}\}$ and $(N_{p})$ classes.

A.Kaneko proved that there exists ahyperfunction $u(t, x)\not\equiv \mathrm{O}$ in aneighborhood

of $\{x=0\}$ satisffing (2), applying Sato’s counterexample (cf. [TT]). We modify

A.Kaneko’s

idea and prove that $\mathrm{u}\mathrm{n}\cdot \mathrm{q}\mathrm{u}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{s}\mathrm{s}$ in (1) does not hold in the QA

ultradis-tribution category neither, in the proofof which,

we

utilize J.Boman’s counterexample

(7)

Ultradistribution solutions to partially hyperbolic partial differential equations

In this section,

we

study solvability of partially hyperbolic partial differential equa-tions in ultradistribution category. This solvability is one of the main tools to prove non-uniqueness in the Cauchy problem (1) in QA ultradistribution category.

We denote $x=(x_{1}, x’)=(x_{1}, x’, x’)\in \mathbb{R}^{n}$, where $x’=(x_{2}, \cdots, x_{k+1})$, $x’=$

$(x_{k+2}, \cdots, x_{n})$

.

Let $P(D)$ be

an

$m$-th order linear partial

differential

operator with

constant coefficients and $p_{m}(D)$ be its principal part. We

assume

that $\{x_{1}=0\}$ is

non-characteristic with respect to $P$

.

We consider the complexification $z=x+iy$ of

$x\in \mathbb{R}^{n}$ and apply similar notations for $x’$ and $x’$

.

We put

$\Omega_{A}:=\{x’\in \mathbb{R}^{k} ; |x’|<A\}$,

$U_{A}:=\{z’\in \mathbb{C}^{n-k-1} ; |z’|<A\}$,

$T_{A}:=\{x_{1}\in \mathbb{R} ; |x_{1}|<A\}$

.

Let $M_{p}$, $p=0,1$,$\cdots$ , be asequence of positive numbers satisfying (M.0), (M.I)

and (M.2)’. We denote by $D^{*}\prime \mathcal{O}(\Omega_{A}\cross U_{A})$ the space ofultradistributions of the class $*\mathrm{d}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{d}$ on $\mathbb{R}^{k}\cross \mathbb{C}^{n-k-1}$ containing $z’\in U_{A}$ as holomorphic parameters. For the

definition ofhyperfunctions and holomorphic parameters, confer [Ka]. In the

same

way,

we

define $D^{*}\prime O(T_{A}\cross\Omega_{A}\cross U_{A})$ on $\mathbb{R}\cross \mathbb{R}^{k}\cross \mathbb{C}^{n-k-1}$. We apply the

same

notions for

$\mathcal{E}^{*}\prime \mathcal{O}$

.

Our main purpose in this section is to prove the following theorem.

Theorem 6. Let $P$ be a partial

differential

operator

defined

above. Assume that the

sequence $M_{p}$

satisfies

(M.O), (M.I), (M.2) and (NA). Then the following conditions

are

equivalent

i) For any $A>0$, there exist such $0<a$,

$0<B<A$

that the initial value problem

$\{$

$P(D)u(x)=0$,

$\partial_{x_{1}}^{j}u|_{x_{1}=0}=u_{j}(x’, z’)$, $j=0,1$,$\cdots$ ,$m-1$,

where $uj\in \mathcal{E}^{*}\prime \mathcal{O}(\Omega_{A}\cross U_{A})$, allows an ultmdistribution solution $u(x_{1},x’, z’)$ with $\sup-$

port compact in $x’$

of

class $*=(M_{p})$ (resp. $\{M_{p}\}$ )

defined

on $T_{a}\cross\Omega_{B}\cross U_{B}$ which

contains $z’\in U_{B}$ as holomorphic parameters.

$\mathrm{i}\mathrm{i})$ There exist constants $\beta$, $\gamma$, $C$,$l$ (resp. there exist 73, $\gamma$ and

for

any 1there exists

$C$

$)$ such that

$|{\rm Im}\zeta_{1}|\leq\beta|{\rm Im}\zeta’|+\gamma|\zeta’|+M(l|\zeta’|)+C$,

for

$P(\zeta)=0$

.

(8)

Remark 7. i) The counterparts of Theorem 7for distributions and hyperfunctions

are

proved in [LK] in astronger form,

our

proof is amodification of their theory. In

E.G.Lee-A.Kaneko’s

theorems they do not

assume

that initial values and solutions

are

compactly supported in $x’$

.

For NQA ultradistributions, thisextension is possible

since

$D^{*}\prime O$ is

$\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{a}\mathbb{I}\mathrm{y}$ soft $\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{n}*\mathrm{d}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{s}\mathrm{N}\mathrm{Q}\mathrm{A}$class. For $\mathrm{Q}\mathrm{A}$

case,

we

have to prove partial

flabbiness

of$D^{*}\prime O$ for this extension.

\"u)

In the proof ofTheorem 6,

we

apply Proposition 3 to

estimate

the support with

respect to $x’$

.

\"ui)

What

we

claim in Theorem 6 is that

we

have

a

solution with holomorphic pa-rameter in

ultradistribution

category, especially in QA

ones.

Since the symbol of $P$ is

a

polynomial it is not the

case

that the term $M(l|\zeta’|)$ is valid, however,

our

theorem

holds for ageneral convolution operators. Therefore we state

our

theorem

as

Theorem

6.

The main theorem

In this section,

we

prove that uniqueness in (1) does not hold in QA

ultradistribution

category, to prove which, Theorem 6 and J.Boman’s counterexample play important

roles.

Theorem 8. Assume that the sequence $N_{p}$

satisfies

(M.O), (M.I), (M.2), (QA) and

(NA). There $exi_{\mathit{8}}ts$ such a $QA$ ultmd

$\dot{u}tr\dot{r}butionu(t,x)$

of

class $(N_{p})$

or

$\{N_{P}\}$ satisfying

(2) that $u(t,x)\not\equiv \mathrm{O}$ in any neighborhood

of

$x=x_{0}$

.

Proof.

For simplicity, let

us

assume

that $x_{0}=0$

.

Consider the Cauchy problem

(3) $\{\begin{array}{l}\frac{\partial^{2}u}{\partial t^{2}}-\Delta u=0u|_{x_{1}=0}=\varphi(x’,t),\frac{\partial u}{\partial x_{1}}|_{x_{1}=0}=\psi(x’,t)=0\end{array}$

where $x=(x_{1},x’)\in \mathrm{R}^{n}$ and $\varphi$ is J.Boman’s counterexample with holomorphic

param-eter $x’$

.

By the construction,

$\varphi$ is compactly supported in $t$

.

By virtue of Theorem 6,

the Cauchy problem (3) has alocal QA

ultradistribution

solution $u(t,x)$

near

$x_{1}=0$

.

By (1),

$\partial_{x,t}^{\alpha}u=\sum_{\beta}c_{\beta}\#_{x’,t}\partial_{x_{1}}u+\sum_{\gamma}c_{\gamma}\partial_{x,t}^{\gamma},u$,

(9)

where $\mathrm{C}\beta$, $c_{\gamma}=1$ or -1. We have

$\partial_{x,t}^{\alpha}u|_{x_{1}=0}=\sum_{\gamma}c_{\gamma}\partial_{x,t}^{\gamma},\varphi=,\sum_{\gamma,\gamma’}c_{\gamma’,\gamma}\prime\prime\partial_{t}^{\gamma’}\partial_{x}^{\gamma’},\varphi$.

Restricting both sides to $\{x’=0\}$ gives

us

$\partial_{x,t}^{\alpha}u|_{x=}0=\partial_{t}^{\gamma’}(\partial_{x}^{\gamma’},\varphi|_{x’=0})=0$,

because $\partial_{x}^{\gamma’},\varphi|_{x’=0}=0$

.

$\square$

Theorem 8completes the study ofuniqueness in the Cauchy problem (1).

Remark 9. Even inNQA ultradistributioncategory, uniquenessdoes not hold if initial values

are

restricted to finite order. More strongly,

we

construct acounterexample

In

distribution category. Let $m\in \mathrm{N}$

.

We have a local distribution solution $u(t, x)\not\equiv \mathrm{O}$ to

the Cauchy problem

$\{$

$\frac{\partial^{2}u}{\partial t^{2}}-\Delta u=0$,

$\partial_{x}^{\alpha}u|_{x=x\mathrm{o}}=0$ for $|\alpha|\leq m$

.

In fact, for simplicity, we

assume

that $x_{0}=0$. Consider the Cauchy problem

(4) $\{$

$\frac{\partial^{2}u}{\partial t^{2}}-\Delta u=0$,

$u|_{x_{1}=0}=(x_{2}\cdots x_{n})^{m+1}g(t)$, $\frac{\partial u}{\partial x_{1}}|_{x_{1}=0}=\psi(x’, t)=0$,

where $g(t)$ is adistribution of

one

variable. By Theorem 2 or 3 in [LK], (4) has $\mathrm{a}$

distribution solution $u(t, x)$

near

$x_{1}=0$. It is easy to show that $\partial_{x}^{\alpha}u|_{x=0}=0$ for

$|\alpha|\leq m$

.

In smoother classes where the counterpart of Theorem 6holds, the counterpart of Remark 9is proved. For example, $C^{\infty}$, ultradifferential and analytic classes

are

those

ones.

Note also that the argument in this article applies to ageneral linear partial

differential equation with analytic coefficients and areal analytic submanifold whose conormals

are

non-characteristic

(10)

REFERENCES

[B1] BomanJ., A local vanishing theoremfor distribuiions, C. R. Acad. Sci. Paris 315 (1992),

1231-1234.

[B2] –, Microlocal quasianalyticityfordistributions andultradistributions, Publ. RIMS, Kyoto Univ. 31 (1995), 10791992.

[B3] –., Uniqueness and non-uniquenessfor microanalytic continuation ofultradistributions, Contemporary Mathematics 259 (2000), 61-82.

[C] Carleman T., Fonction quasi analytique, Gauther-Villars et $\mathrm{C}\mathrm{i}\mathrm{e}$,

\’Editeurs,

Paris, 1926.

[E] Ehrenpreis L., A new Cauchy problem for partial differential equations (lecture at Univ. of

Tokyo ), Sugaku 26 (1974), 168 (notesby A.Kaneko), inJapanese.

[J] John F., On $l\dot{1}nearpan\dot{1}ald|.ffe\tau entalequat:ons$ with $analyt\dot{\iota}c$ coefficimts, Comm. Pure. Appl.

Math. 2 (1949), 209253.

[Ka] Kaneko A., Introduction to Hyperfunctions, Kluwer Academic Publishers,, Dordrecht, Boston, London, 1988.

[Kol] Komatsu H., $Ultmdi\epsilon tribut:ons,I,Structure$ theorems and characterization, J. $\mathrm{F}\mathrm{a}\mathrm{c}$. Sci. Univ.

Tokyo, Sect.lA 20 (1973), 25-105.

[Ko2] –, $Ultmd|.st_{l\dot{\mathrm{B}}}but:ons,II$, The kemel $theor\epsilon m$ and $ultmd|.st’\dot{\mathrm{r}}but\dot{\iota}onswth$ support in a

sub-manifold, Ibid24 (1977), 607-628.

[LK] $\mathrm{L}\mathrm{a}\mathrm{e}$ E.G.

and Kaneko A., Cond.Aon of$pah:al$ $hyperbol_{\dot{1}Cu}$. $y$ modulo a linear subvariety for

opemtors $w\cdot.th\omega nstant$ $coefficmts$, J. Fac. Sci. Univ.

$\mathrm{T}\mathrm{o}\mathrm{k}\mathrm{y}\mathrm{o},\mathrm{S}\mathrm{e}\mathrm{c}\mathrm{t}.\mathrm{l}\mathrm{A}$ $30$ (1989), 329346.

[N] Nacinovich M., Cauchy problem $fo\tau overdetem\dot{|}ned$ systems, Talk at the meeting “Analytic

SolutionsofPartial DifferentialEquations $n$

, Trento, 1992.

[TT] Tanabe S. and Takiguchi T., A local vanishing theorem for ultradistributions with analytic parameters, J. Fac. Sci. Univ. Tokyo Sec.1A 40 (1993), 607-621.

$\mathrm{D}_{\mathrm{E}\mathrm{P}\mathrm{A}\mathrm{R}\mathrm{T}\mathrm{M}\mathrm{B}\mathrm{N}\mathrm{T}}$ OF MATHEMATICS, NATIONAL

$\mathrm{D}_{\mathrm{B}\mathrm{F}\mathrm{B}\mathrm{N}\mathrm{S}\mathrm{B}}$ACADEUY, 1-10-20,

HASHIRIMIZU, $\mathrm{Y}\mathrm{o}\kappa 0-$

SUKA, KANAGAWA, 239-8686JAPAN $E-mal$ address: [email protected]

参照

関連したドキュメント

ABSTRACT: The decomposition method is applied to examples of hyperbolic, parabolic, and elliptic partlal differential equations without use of linearlzatlon techniques.. We

In the present paper, it is shown by an example that a unit disc counterpart of such finite set does not contain all possible T- and M-orders of solutions, with respect to

Solvability conditions for linear differential equations are usually formulated in terms of orthogonality of the right-hand side to solutions of the homogeneous adjoint

In this paper, we establish a Stroock-Varadhan support theorem for the global mild solution to a d (d ≤ 3)-dimensional stochastic Cahn-Hilliard partial differential equation driven by

We include applications to elliptic operators with Dirichlet, Neumann or Robin type boundary conditions on L p -spaces and on the space of continuous

Interesting results were obtained in Lie group invariance of generalized functions [8, 31, 46, 48], nonlinear hyperbolic equations with generalized function data [7, 39, 40, 42, 45,

Kilbas; Conditions of the existence of a classical solution of a Cauchy type problem for the diffusion equation with the Riemann-Liouville partial derivative, Differential Equations,

Here we do not consider the case where the discontinuity curve is the conic (DL), because first in [11, 13] it was proved that discontinuous piecewise linear differential