16
pages

- global solution
- sobolev space
- schrodinger equation
- linear wave
- navier stokes equations
- solution can
- defocusing semi-linear

Voir plus
Voir moins

Vous aimerez aussi

On

(

global

main

solutions

defo

to

w

a

=

defo

is

cusing

using

semi-linear

the

w

,

a

@

v

;

e

where

equation

1

Isab

H

elle

Gallagher

this

C.N.R.S.

[8]),

U.M.R.

since

8628,

In

D

consider

epartemen

t

j

de

1

Math

v

ect

ematiques

the

Univ

w

ersit

global

of

e

2

P

L

aris

es

Sud,

Stric

91405

orth

Orsa

b

y

[10

Cedex,

]).

F

duction

rance

W

Isabelle.Gallagher@math.u-

equation

psud

t

.fr

F

in

abrice

3

Planc

t

hon

=0

C.N.R.S.

;

U.M.R.

;

7598,

is

Lab

This

oratoire

with

d'Analyse

the

Num

and,

y

erique

lo

Univ

osedness

ersit

extends

ell-p

e

conserv

P

Hamiltonian

aris

2

6,

1

75252

k

P

.

aris

h

Cedex

k

05,

(it

F

estimates,

rance

is

fab@ann.jussieu.fr

that

Abstract

osedness

W

kno

e

sixties

pro

through

v

7

e

1

that

tro

the

and

3D

theorem

cubic

e

defo

the

cusing

semi-linear

2

w

a

+

v

3

e

0

equation

R

is

R

globally

(

w

@

ell-p

)

osed

t

for

=

data

0

in

the

)

Sob

(1.1)

olev

space

real

_

alued.

H

equation

s

sub-critical

where

resp

s

to

>

H

3

norm,

4

since

nonlinearit

This

is

result

cusing,

w

cal

as

ell-p

obtained

in

in

1

[11

to

]

w

follo

osedness

wing

the

Bour-

ation

gain's

the

metho

kr

d

k

([3

L

]).

+

W

2

e

presen

4

t

4

here

While

a

approac

dieren

go

t

bac

and

to

somewhat

80's

simpler

requires

argumen

hartz

t,

see

inspired

it

b

w

y

noting

previous

global

w

ell-p

ork

had

on

een

the

wn

Na

the

vier-Stok

(e.g.

es

]),

equations

1

([44

regularization

1

and

6

compactness

H

metho

step,

ds.

L

The

data

problem

e

of

(

lo

6

cal

is

w

u

ell-p

ys

osedness

where

for

splitting

lo

v

w

than

regularit

forces

y

particular

data

\

w

t

as

3

answ

of

ered

=

later,

time

and

the

denitiv

e

a

results

ma

w

2

ere

data

obtained

2

in

are

[14

1

]

that

in

.

a

regularit

more

0

general

_

framew

)

ork.

enables

Equation

1

(1.1)

ask

turns

smallest

out

3.

to

p

b

u

e

t

lo

of

cally

it

w

e

ell-p

W

osed

_

for

0

initial

_

data

v

in

a

_

e

H

j

1

J

2

equation

getting

_

(

H

4

1

all

2

v

and

0

ill-p

is

osed

)

b

assumption

elo

2

w

F

s

that

=

ed,

1

v

2

1

(whic

k

h

+

mak

(

es

1

sense

:

from

v

a

x

scaling

6

p

smo

oin

theorem,

t

2

of

h

view,

through

as

v

b

e

oth

the

u

equation

3

and

u

the

u

_

2

H

solution

1

=

2

alw

norm

deriv

are

regular

in

itself.

v

split

arian

0

t

s

under

=

the

v

same

0

rescaling).

1

A

norm

natural

2

question

2

is

norm.

then

ac

whether

b

these

frequency

lo

j

cal

with

solutions

2.

can

e

b

with

e

0

extended

global

globally

C

in

H

time,

\

at

(

least

),

for

of

some

and

range

size

1

k

2

_

whic

s

order

<

2

1.

.

In

usual

[3]

the

Bourgain

(

in

)

tro

0

duced

a

a

e

general

y

framew

is

ork

that

for

e

obtaining

(as

results

_

of

+

this

k

t

t

yp

1

e,

6

and

3

applied

6

it

2

to

+

the

)

2D

This

cubic

to

Sc

L

hr

L

odinger

(note

equation.

+

F

2

or

than

equation

w

(1.1),

in

Bourgain's

h

metho

,

d

is

yields

othness

global

ermits

w

x

ell-p

hartz

osedness

o

in

a

_

problem,

H

e

s

ed

for

2

s

u

>

+

3

v

4

2

,

(

as

)

pro

=

v

;

ed

:

in

e

[11].

the

W

at

e

t

in

0;

tend

is

to

a

giv

one

e

ativ

a

less

dieren

than

t

solution

pro

1.

of

e

of

the

this

result,

2

follo

H

wing

:

a

0

strategy

u

in

+

tro

0

duced

u

in

2

the

H

con

with

text

large

of

and

the

0

Na

_

vier-Stok

1

es

with

equations

small

in

One

[4]

y

(see

hiev

also

this

[7]

y

for

in

a

at

more

recen

t

J

approac

large

h).

.

When

W

compared,

solv

the

the

t

(1.1)

w

small

o

v

metho

,

ds

a

app

solution

ear

in

to

t

b

_

e

1

somewhat

)

dual

L

of

t

eac

L

h

x

other.

and

Let

norms

us

v

start

small

with

ha

a

e

simple

less

pro

2

of

v

whic

k

h

H

will

2

yield

h

global

of

w

2

ell-p

1

osedness

s

for

J

s

Remark

>

the

3

smallness

4

on

+

data

1

2

12

1

=

s

5

J

6

"

.

Since

or

the

later

equation

w

is

remark

globally

an

w

additional

ell-p

y

osed

preserv

for

so

large

in

data

w

in

ha

_

e

H

v

1

2

and

H

small

2

data

1

in

)

_

v

H

C

1

(

2

H

,

2

one

1

ma

)

y

L

b

t

e

L

tempted

x

to

follo

(

w

2

a

1

general

s

principle

J

of

(1.2)

nonlinear

will

in

us

terp

estimate

olation

in

and

3

claim

(

the

6

equation

)

to

that

b

2

e

1

globally

=

w

3

ell-p

smaller

osed

the

in

othness

b

e

et

for

w

the

een.

whic

T

is

o

4

mak

while

e

3

sense

the

of

smo

this

whic

heuristic,

p

w

an

e

6

pro

estimate

ceed

Stric

in

inequalities).

sev

T

eral

reco

steps.

er

What

solution

follo

our

ws

w

is

solv

an

a

informal

erturb

pro

equation,

of.

@

Note

t

that

from

+

no

3

w

3

on,

2

w

+

e

v

will

u

nev

0

er

u;

men

t

tion

j

the

=0

regularit

(

y

0

of

u

the

)

time

(1.3)

deriv

ativH

This

k

equation

ha

turns

the

out

2

to

s

b

L

e

L

lo

w

cally

b

w

space,

ell-p

2

osed

k

in

ds

_

t<T

H

0

1

older

,

1

on

trol

a

J

time

(1

in

trivial

terv

the

al

u

dep

u

end-

+

ing

j

only

2

on

T

k

1

u

4

0

v

k

v

_

the

H

J

1

s

.

)

This

large

is

as

easily

.

seen

4

through

ys

a

This

con

if

traction

ha

argumen

estimate

t

+

in

2

C

L

t

_

(

k

_

u

H

6

1

@

).

t

It

j

will

v

b

T

e

k

enough

t

to

u

pro

H

v

Z

e

6

the

Z

nonlinear

x

terms

em

to

the

b

T

e

+

L

J

1

3

t

(

(

h

L

H

2

s

x

osing

).

4

Using

2

Sob

Note

olev

k

em-

y

b

this

edding,

y

u

elo

3

really

2

argumen

C

ts

t

:

(

rene

L

tegrals

2

u

x

H

)

@

and

2

hence

1

is

k

lo

cally

k

L

1

1

t

t

L

.

2

On

k

the

(1.6)

other

t

hand,

2

v

u

2

6

L

j

3

t

t

T

(

um

L

t

6

with

x

=

)

k

for

_

v

k

0

k

in

+

_

4

H

giv

1

.

2

H

+

0

1

2

6

ds

(recall

2

Step

0

2),

L

whic

using

h

Sob

yields

eddings

v

tegrals

2

Recalling

u

obtain

2

2

L

s

1

T

t

2

(

2

L

+

2

T

x

2

)

3

using

(1.7)

H

es

older.

v

The

for

remaining

as

term

5

u

c

2

accordingly

v

needs

is

2

con

s

trolled

s

b

:

y

w

the

e

other

0

t

4

w

2

o,

)

and

alw

denoting

ossible

b

rescaling

y

explained

jk

in

part

kj

only

the

of

norm

at

in

w

the

optimal

con

3

traction

e

space,

e

w

estimates

e

time

obtain

energy

jk

k

u

k

kj

_

T

1

.

k

k

t

u

k

0

L

k

+

_

2

H

u

1

4

+

4

k

k

u

0

1

2

k

H

L

+

2

@

+

u

2

2

2

2

j

1

(

k

1

0

2

4

+

4

1

+

6

Z

s

0

)

u

T

v

1

t

3

j

jk

+

u

Z

kj

0

T

v

+

u@

T

u

jk

ds:

u

aking

kj

suprem

3

o

T

er

:

<

(1.4)

,

Th

H

us,

def

the

sup

linear

(

term

u

on

2

the

H

righ

+

t

@

can

u

b

2

e

2

absorb

k

ed

k

on

L

the

)

left,

es

as

T

so

H

on

+

as

T

T

T

.

k

2

k

6

L

j

x

(

+

1

3

2

T

+

T

1

k

6

k

s

6

)

ds;

and

H

w

and

e

olev

obtain

b

the

for

desired

in

result,

on

with

left.

T

(1.2),

.

e

inf

H

.

2

2

6

(1

j

)

(

H

1

T

2

3

+

2

1

(

6

3

s

)

)

H

;

2

1

T

k

3

u

J

0

2

k

s

2

;

_

whic

H

giv

1

con

!

o

:

er

(1.5)

T

4.

arbitrarily

T

T

o

long

extend

>

lo

6

cal

y

solutions

ho

to

J

global

,

ones,

one

w

T

e

3

then

2

need

(1

to

+

obtain

3

an

)

a

1

priori

(1.8)

b

that

ound

e

on

v

the

estimated

energy

u

of

k

a

L

solution

b

u

2

.

J

This

s

will

:

b

is

e

a

accomplished

p

through

b

the

a

energy

in

inequalit

as

y

b

,

w

pro

(2.1).

vided

last

one

is

can

the

con

non

trol

part

the

the

p

t,

erturbativ

least

e

one

terms

an

b

the

y

result

the

>

energy

4

of

w

u

will

.

v

Indeed,

to

w

the

e

on

ha

space

v

in

e

3

thew

arising

with

from

;

the

later,

energy

only

inequalit

Before

y

spaces

,

denition

to

explicit

lo

(1

w

s

er

L

the

already

2

theorem

3

nonlinear

s

2

factor

did

in

w

(1.7)

w

to

H

1

C

2

)(6

s

eral

.

of

In

of

fact,

one

ignoring

in

all

e

irrelev

metho

an

of

t

Since

epsilons,

the

(1.8)

frequency

will

in

b

whic

ecome

er

T

e.g.

k

0

u

H

0

)

k

(

2

0

_

)

H

s

1

the

k

order.

v

b

0

ense

k

olev

2

eed

_

[11].

H

the

1

inhomogeneous

2

y

.

(or

1

on

:

as

(1.9)

equations

T

equation.

o

e

pic

and

k

w

T

ma

as

in

large

0

as

,

w

pro

e

giv

w

a

an

frequency

t,

e

w

out

e

osition,

ha

in

v

v

e

1

to

pla

v

y

J

with

k

the

resp

)

ectiv

k

e

_

size

L

of

(1

u

3)

0

:

and

ceeding

v

of,

0

are

and

restriction

this

data

giv

disp

es

the

immediately

w

the

cal

s

and

>

nite

3

propagation,

4

tioned

restriction:

ely

tak

simply

e

Sob

k

y

v

terpart,

0

One

k

state

_

other

H

and

1

commen

2

asp

w

M

applying

1

to

k

e

hr

k

Pro

_

theorem

H

with

s

u

where

0

M

e

is

with

large.

spaces,

Then

simply

k

u

its

0

part

k

its

_

v

H

w

1

the

in

M

duction.

1

us

osition,

2

k

is

whic

k

cut.

_

w

H

oin

s

one

,

abstract

where

giv

s

y

=

olation.

will

+

k

1

_

2

2

(1

)

k

).

and

This

k

forces

1

2

H

<

all

0.

4

This

t

barrier

at

(

u

=

k

1

H

2

\

b

4

et

t

w

s

een

s

the

4

conserv

3

ation

(1.10)

la

pro

w

with

(here,

pro

_

sev

H

remarks

1

in

)

The

and

to

the

4

scaling

can

(here,

e

_

osed

H

at

1

exp

2

of

)

orking

app

lo

ears

Sob

to

spaces

b

use

e

the

ro

sp

oted

of

in

as

b

men

oth

in

Bourgain's

Alternativ

metho

,

d

could

and

replace

ours.

homogeneous

Indeed,

olev

in

b

all

their

the

coun

recen

as

t

[14].

progress

ma

for

also

other

the

(tec

for

hnically

nonlinearities

more

Klein-Gordon),

complicated)

w

equations

will

lik

t

e

that

KdV

ect

([6])

as

or

ell

NLS

on

([5])the

the

same

d

kind

other

of

lik

restrictions

the

app

Sc

ear,

odinger

through

2

for

of

KdV

the

it

W

solv

start

es

the

completely

of

the

0

w

v

ell-p

.

osedness

w

question

are

as

orking

the

L

equation

based

is

one

ill-p

y

osed

cut

b

data

elo

0

w

to

half-w

lo

a

frequency

y

u

to

and

the

high

scaling

part

([12

0

]).

as

Let

e

us

in

no

informal

w

of

state

the

the

tro

main

This

theorem.

es

Theorem

an

1

decomp

Let

with

(

parameter

0

J

;

h

the

1

at

)

h

2

e

(

W

_

ho

H

ev

s

p

\

t

L

that

4

needs

;

an

_

decomp

H

as

s

en

1

b

)

real

with

terp

s

Here

>

e

3

ha

4

e

.

u

Then

k

there

H

exists

a

J

unique

s

global

k

in

0

time

_

solution

s

to

k

(1.1).

0

Moreo

_

v

er,

2

w

(

e

s

ha

k

v

0

e

_

k

s

for

k

_

s:

H

slo

Let

sum

us

t

mak

Theorem

e

x;

a

Definition

remark

f

concerning

,

the

(

L

.

4

0

norms.

v

F

temp

or

e

an

b

y

f

>

L

0,

then

w

~

e

p

ha

exp

v

then

e

3

k

v

def

0

q

(

whic

)

1

)

(

k

x

L

2

4

f

=

p

x;

3

inhomogeneous

4

L

k

k

0

0

q

k

(

L

)

4

2

and

<

k

to

s

0

only

(

ards

the

)

s

k

b

_

.

H

the

s

only

=

Let

p;

s

i.e.

3

q

2

,

k

q

,

0

t

k

j

_

denote

H

e

s

k

:

)

(2.1)

(

It

and,

follo

ws

the

that

Cauc

as

x;

so

L

on

2

as

t

s

(

>

k

3

p

4

2

,

p

the

k

L

~

4

0

norm

p

of

2

If

0

p

can

b

b

homogeneous

e

_

made

(

arbitrarily

if

small

the

compared

m

to

f

the

to

_

as

H

distribution,

s

norm

2

b

y

L

rescaling

to

the

Z

data.

,

In

v

particular

hartz

since

w

k

in

u

([9,

0

p;

k

(

L

q

4

admissible

.

h

k

+

1

0

>

k

similarly

L

p;

4

.

w

x

e

(

conclude

b

that

o

the

at

quan

j

tit

,

y

=

k

Then

u

i!

0

x

k

p

4

q

L

2

4

k

can

)

b

;

e

u

con

)

trolled

F

b

)

y

to

k

with

u

y

0

u

k

)

2

t

_

x

H

j

1

p

,

(

and

k

w

0

e

~

assume

)

this

(

to

k

b

(

e

)

the

(

case

2

for

k

the

t

rest

~

of

(

the

0

pro

where

of;

the

that

t

will

5

b

;

e

).

useful

2

to

s

estimate

3

the

,

Hamiltonian

f

of

elongs

u

the

in

Beso

Section

space

2.3.

B

Finally

p;q

,

R

w

)

e

and

recall

if

some

partial

denitions

P

and

m

prop

j

erties

con

of

erges

the

w

w

f

a

a

v

ered

e

and

equation

sequence

whic

j

h

=

will

j

b

k

e

j

of

k

use

p

later.

elongs

First

`

w

(

e

)

de