12.08.2013 Views

A Stochastic Model of Crystallization in an Emulsion - Laboratoire de ...

A Stochastic Model of Crystallization in an Emulsion - Laboratoire de ...

A Stochastic Model of Crystallization in an Emulsion - Laboratoire de ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

A <strong>Stochastic</strong> <strong>Mo<strong>de</strong>l</strong> <strong>of</strong> <strong>Crystallization</strong> <strong>in</strong> <strong>an</strong> <strong>Emulsion</strong><br />

G. Vallet - D. Trujillo<br />

<strong>Laboratoire</strong> <strong>de</strong> Mathématiques Appliquées, ERS-CNRS 2055<br />

Université <strong>de</strong> Pau et <strong>de</strong>s Pays <strong>de</strong> lAdour, 64000 Pau, Fr<strong>an</strong>ce<br />

guy.vallet@univ-pau.fr, david.trujillo@univ-pau.fr<br />

Abstract :<br />

The ma<strong>in</strong> feature <strong>of</strong> phase ch<strong>an</strong>g<strong>in</strong>g <strong>in</strong> dispersed medium is its r<strong>an</strong>dom behavior. The aim <strong>of</strong><br />

this work is to study a stochastic mo<strong>de</strong>l for the crystallization <strong>of</strong> emulsions dropplets. We adapt<br />

here some technics <strong>of</strong> stochastic partial differential equation to prove existence <strong>an</strong>d uniqueness <strong>of</strong> the<br />

solution. We then give some numerical simulations <strong>an</strong>d compare our results with the experimental<br />

observations.<br />

Keywords: <strong>Stochastic</strong> partial differential equations, non l<strong>in</strong>ear systems.<br />

AMS Classication: 34F05 - 35R60 - 60H15.<br />

1 Introduction<br />

The purpose <strong>of</strong> this paper is to use nonl<strong>in</strong>ear stochastic partial differential equations techniques to<br />

<strong>de</strong>scribe the phase ch<strong>an</strong>g<strong>in</strong>g <strong>in</strong> a dispersed medium (here <strong>an</strong> emulsion). A common way <strong>of</strong> <strong>de</strong>al<strong>in</strong>g<br />

with the microscopic properties <strong>of</strong> <strong>an</strong> emulsion has been to replace the thermodynamics unkowns<br />

by some k<strong>in</strong>d <strong>of</strong> averages (see Dumas et al [6] <strong>an</strong>d Vallet [16]).<br />

Exepted <strong>in</strong> some particular cases (see water emulsions presented at the end <strong>of</strong> the numerical<br />

simulations chapter), we obta<strong>in</strong> a reasonable mo<strong>de</strong>l by us<strong>in</strong>g a <strong>de</strong>term<strong>in</strong>istic one. Nevertheless, the<br />

study <strong>of</strong> the effects <strong>of</strong> small uctuations <strong>of</strong> the <strong>de</strong>term<strong>in</strong>istic mo<strong>de</strong>l by some stochastic perturbations<br />

is <strong>in</strong>terest<strong>in</strong>g :<br />

rstly, for physical mo<strong>de</strong>ll<strong>in</strong>g reasons (see for example Grecksch et al [8] or Hol<strong>de</strong>n et al [9] for the<br />

Strat<strong>an</strong>ovich <strong>in</strong>tegration),<br />

secondly, for mathematical reasons, because <strong>of</strong> the lack <strong>of</strong> compactness for that k<strong>in</strong>d <strong>of</strong> problem<br />

(see for example Bensouss<strong>an</strong> et al [3] , Lions et al [13] for the Strat<strong>an</strong>ovich <strong>in</strong>tegration or Hol<strong>de</strong>n<br />

et al[10] for stochastic conservation laws).<br />

Therefore, the theory <strong>of</strong> nonl<strong>in</strong>ear stochastic equations has been <strong>de</strong>velloped <strong>in</strong> m<strong>an</strong>y papers, us<strong>in</strong>g<br />

multi-valued functions, time discretization, Galerk<strong>in</strong> approximation, splitt<strong>in</strong>g methods, ..., ma<strong>in</strong>ly<br />

<strong>in</strong> one dimension space.<br />

For some stochastic calculus tools we refer to Friedm<strong>an</strong> [7], Grecksch et al [8], Hol<strong>de</strong>n et al [9]<br />

<strong>an</strong>d Oksendal et al[14];<br />

for stochastic partial differential equations we refer, moreover all <strong>of</strong> the above citations, to Bensouss<strong>an</strong><br />

[2], Breckner et al [4], Capińsky et al [5], Krylov [11], Kuo et al[12], without forgett<strong>in</strong>g<br />

Pardoux [15] <strong>an</strong>d Walsh [17].<br />

<strong>Mo<strong>de</strong>l</strong>l<strong>in</strong>g :<br />

One <strong>of</strong> the better known energy stor<strong>in</strong>g method is the latent heat one.<br />

It ma<strong>in</strong>ly consist <strong>in</strong> shift<strong>in</strong>g the temperature <strong>of</strong> a given material to a given value us<strong>in</strong>g a certa<strong>in</strong><br />

amount <strong>of</strong> energy, provi<strong>de</strong>d it is cheap enough. This br<strong>in</strong>gs about a phase ch<strong>an</strong>ge <strong>in</strong> the structures<br />

<strong>of</strong> the material. When the material reverts to its <strong>in</strong>itial temperature, the <strong>in</strong>verse phase ch<strong>an</strong>ge<br />

yields a part <strong>of</strong> the energy stored dur<strong>in</strong>g the rst phase ch<strong>an</strong>ge.<br />

Experimentation, on <strong>an</strong> <strong>in</strong>dustrial scale, has shown that, for efficiency reasons, the phase ch<strong>an</strong>g<strong>in</strong>g<br />

material (PCM) has to be scattered <strong>in</strong> droplets <strong>in</strong>to <strong>an</strong> other material, which does not ch<strong>an</strong>ge<br />

1


phase dur<strong>in</strong>g the experimentation. We thus consi<strong>de</strong>r <strong>an</strong> emulsion conta<strong>in</strong><strong>in</strong>g the phase-ch<strong>an</strong>ge<br />

material. However, repeat<strong>in</strong>g the melt<strong>in</strong>g-freez<strong>in</strong>g cycles, <strong>de</strong>stroys the dispersed structure <strong>of</strong> the<br />

emulsion. To avoid that, a surface-active agent (surfact<strong>an</strong>t) c<strong>an</strong> be used <strong>an</strong>d so the behavior <strong>of</strong><br />

the emulsion is closed to the behavior <strong>of</strong> solid matter. Therefore, the effect <strong>of</strong> free-convection c<strong>an</strong><br />

be ignored.<br />

The object <strong>of</strong> this paper is to talk about the crystallization <strong>of</strong> <strong>an</strong> emulsion.<br />

This process is based on two phenomena (cf. Dumas et al [6]) :<br />

1) the un<strong>de</strong>rcool<strong>in</strong>g, <strong>de</strong>ned as the difference between the melt<strong>in</strong>g temperature TF<br />

<strong>an</strong>d the<br />

crystallization temperature, <strong>in</strong>creases as the sample size <strong>of</strong> PCM <strong>de</strong>creases. For example, for water,<br />

3 <br />

3<br />

the un<strong>de</strong>rcool<strong>in</strong>g is 14 K for a few cm macrosamples <strong>an</strong>d is 38 K for a few m microsamples.<br />

Moreover, <strong>an</strong>y droplet which temperature is less th<strong>an</strong> TF<br />

may crystallize as soon as <strong>an</strong>y perturbation<br />

or shock occurs. Where from the second phenomenon :<br />

2) the ma<strong>in</strong> feature <strong>of</strong> emulsions crystallization <strong>in</strong> <strong>an</strong> <strong>in</strong>stable dispersed medium (here <strong>an</strong><br />

emulsion with temperature less th<strong>an</strong> TF<br />

) is its stochastic behavior. Samples <strong>of</strong> PCM which<br />

are apparently i<strong>de</strong>ntical, will not tr<strong>an</strong>sform at the same temperature dur<strong>in</strong>g the cool<strong>in</strong>g process.<br />

This leads to the notion <strong>of</strong> nucleation rate. The Nucleation Theorie gives us a function J,<br />

the<br />

<strong>de</strong>term<strong>in</strong>istic part <strong>of</strong> the crystallization speed, per unit <strong>of</strong> time. In the sequel, we consi<strong>de</strong>r the<br />

nucleation rate as a stochastic perturbation <strong>of</strong> this speed J.<br />

We assume that J is a positive lipschitz function <strong>of</strong> the temperature with, J(0) = 0 <strong>an</strong>d J( x)<br />

=<br />

0 , ∀x ∈ [ T , + ∞[<br />

.<br />

∞<br />

F<br />

One mays nd a presentation <strong>of</strong> the experimental context <strong>in</strong> the last section.<br />

Let us <strong>de</strong>note by u the temperature <strong>of</strong> the emulsion <strong>an</strong>d ϕ( t, x)<br />

the proportion <strong>of</strong> crystallized<br />

droplets, <strong>in</strong> the neighborhood <strong>of</strong> a po<strong>in</strong>t x <strong>an</strong>d a time t .<br />

Then, the mathematical mo<strong>de</strong>ll<strong>in</strong>g proposed by Dumas et al [6] is based on :<br />

the nonl<strong>in</strong>ear heat equation with a heat source term proportional to the speed <strong>of</strong> crystallization,<br />

i.e.<br />

∂u<br />

∂ϕ<br />

( u)<br />

= ,<br />

∂t<br />

∂t<br />

with some Fouriers type boundary conditions<br />

<strong>in</strong> <br />

(1)<br />

∂ u<br />

<br />

∂<br />

N<br />

( ) ∞<br />

= ( u u ) , on = ∂<br />

where u is a given temperature on the boundary ,<br />

the follow<strong>in</strong>g stochastic differential equation for the speed <strong>of</strong> crystallization<br />

∂ϕ<br />

∂t<br />

= (1 ϕ) J( u) + b( u, ϕ) <br />

∂w<br />

, <strong>in</strong><br />

∂t<br />

where 1 ϕ is the proportion <strong>of</strong> uncrystallized droplets, J( u) the nucleation rate, w the st<strong>an</strong>dard<br />

real Wiener process <strong>an</strong>d b the r<strong>an</strong>dom part <strong>of</strong> the nucleation phenomenon.<br />

A rigorous formulation <strong>of</strong> our system will be given below.<br />

2 Notations <strong>an</strong>d hypothesis<br />

From now on, we consi<strong>de</strong>r the follow<strong>in</strong>g notations :<br />

is a boun<strong>de</strong>d doma<strong>in</strong> <strong>of</strong> R with a smooth boundary <strong>an</strong>d <strong>an</strong> outward unit normal ,<br />

for T > 0 , Q is the cyl<strong>in</strong><strong>de</strong>r ]0 , T [ <strong>an</strong>d its lateral boundary ]0 , T [ ,<br />

2<br />

(2)<br />

(3)


( , Ϝ, P ) is a complete probability space, E the expectation <strong>an</strong>d ( Ϝt)<br />

t ∈[0<br />

,T ] a right cont<strong>in</strong>uous<br />

ltration such that Ϝ conta<strong>in</strong>s all Ϝ-null<br />

sets,<br />

<br />

<strong>de</strong>notes the real valued st<strong>an</strong>dard Ϝ -Wiener process,<br />

t ∈[0<br />

,T ]<br />

t<br />

H = L () , V = H () , H = L ( , H) = L ( , dP dx)<br />

,<br />

2 2 2<br />

V = L ( , V ) , H = L (0 , T, H) = L ( Q, dP dxdt)<br />

<strong>an</strong>d<br />

2 2<br />

V = L (]0 , T [ , V) = L (]0 , T [ <br />

, V ) ;<br />

|| . || <strong>an</strong>d || . || <strong>de</strong>note respectively the H <strong>an</strong>d V norms,<br />

V<br />

∞ ∞ ∞<br />

0 0<br />

∞<br />

∞<br />

0 0 0<br />

F F<br />

x<br />

. the function , <strong>de</strong>ne by ( x) = ( s) ds; which satises for all x <strong>an</strong>d y,<br />

x <br />

′<br />

. the function G, <strong>de</strong>ne by G( x) = ( s) ds,<br />

s>t s<br />

0<br />

2 1 2 2<br />

is a lipschitz cont<strong>in</strong>uous <strong>in</strong>creas<strong>in</strong>g function with (0) = 0 <strong>an</strong>d 0 < m M<br />

(therefore, this mo<strong>de</strong>l does not lead to a <strong>de</strong>generated parabolic equation),<br />

> 0 ; u ∈ L () , u 0 ; u ∈ L () <strong>de</strong>notes the cool<strong>in</strong>g temperature imposed on the<br />

boundary, with u positive, non <strong>in</strong>creas<strong>in</strong>g with respect to t <strong>an</strong>d regular,<br />

ϕ ∈ L () , 1 ϕ 0 ( ϕ = 0 <strong>in</strong> the experimentation),<br />

J is a lipschitz cont<strong>in</strong>uous, positive function with, J( x ) = 0,<br />

for <strong>an</strong>y x <strong>of</strong> R <strong>an</strong>d <strong>an</strong>y x <strong>of</strong><br />

[ T , ∞[<br />

, where T is the fusion temperature,<br />

b is a lipschitz cont<strong>in</strong>uous, positive function on R such that b( u, ϕ)<br />

= 0 if :<br />

. u 0<br />

(as the temperatures are <strong>in</strong> Kelv<strong>in</strong>, noth<strong>in</strong>g may happen for non positive temperatures),<br />

. u TF<br />

;<br />

(there is no crystallizations for a temperature above the fusion temperature),<br />

. ϕ 0;<br />

(some more crystallizations may occur only if some crystallizations already exist <strong>in</strong> the neighborhood,<br />

i.e. ϕ > 0),<br />

. ϕ 1;<br />

(there is noth<strong>in</strong>g else to crystallize if ϕ 1),<br />

<br />

<br />

( w( t))<br />

Moreover, for technical reasons, we need :<br />

3 Recalls on <strong>Stochastic</strong> Integration<br />

0<br />

( x)( y x) ( y) ( x) ( y)( y x ) ,<br />

. <strong>an</strong>d the function , <strong>de</strong>ne by ( x ) = max(0, m<strong>in</strong>( x, 1)) ,<br />

C is <strong>an</strong>y const<strong>an</strong>t that we do not need to precise.<br />

0<br />

We recall here some results concern<strong>in</strong>g the Itô <strong>in</strong>tegration, which c<strong>an</strong> be found <strong>in</strong> Grecksch et al<br />

[8].<br />

( , Ϝ, P, ( Ϝt) t0)<br />

is a st<strong>an</strong>dard ltered probability space i.e. ( , Ϝ,<br />

P ) is a probability space,<br />

( Ϝt) 0<br />

is <strong>an</strong> <strong>in</strong>creas<strong>in</strong>g family <strong>of</strong> <br />

algebras <strong>of</strong> Ϝ, Ϝ0 conta<strong>in</strong>s all Ϝ-null sets <strong>an</strong>d Ϝt<br />

=<br />

∩ Ϝ .<br />

3<br />

2<br />


( w( t)) t ∈[0<br />

,T ] is a cont<strong>in</strong>uous process with values <strong>in</strong> R , <strong>an</strong>d w(0)<br />

= 0.<br />

For every s < t, w( t) w( s)<br />

is a Gaussi<strong>an</strong> real valued r<strong>an</strong>dom variable with me<strong>an</strong> 0 <strong>an</strong>d<br />

vari<strong>an</strong>ce t s.<br />

Moreover, ∀s < t, ∀f<br />

Ϝ -measurable<br />

n<br />

L 2(0 ,T ; L 2(<br />

,H))<br />

E[( w( t) w( s)) f] = 0 <strong>an</strong>d E[( w( t) w( s)) ] = t s.<br />

About the Itô <strong>in</strong>tegration :<br />

<br />

2 2<br />

t<br />

∀f ∈ L (0 , T ; L ( , H)) , ( f( s) dw( s))<br />

is a cont<strong>in</strong>uous Ϝ -measurable process <strong>an</strong>d<br />

<br />

<br />

<br />

If f ⇀ f then<br />

<strong>an</strong>d<br />

s<br />

t0 t<br />

t t<br />

2<br />

2<br />

E[ f( s) dw( s)] = E[ f( s) ds ] .<br />

T<br />

T<br />

f ( s) dw( s) ⇀ f( s) dw( s)<br />

n<br />

L 2<br />

0 ( , ,P,H)<br />

0<br />

. .<br />

f ( s) dw( s) ⇀ f( s) dw( s ) .<br />

0<br />

L 2(<br />

, ,P, C([0<br />

,T ]))<br />

0<br />

The Fub<strong>in</strong>i Theorem :<br />

2 2<br />

Let f ∈ L (0 , T ; L ( , H)) <strong>an</strong>d h ∈ H with | f| h then<br />

t t <br />

f( s, ., x) dw( s) dx = f( s, ., x) dx dw( s) P a.s.<br />

The Itô formula :<br />

2 2<br />

Let f, g, M ∈ L (0 , T ; L ( , H)) such that ∀t ∈ [0 , T ] ,<br />

t t<br />

M( t) = M + g( s, ., . ) ds + f( s, ., . ) dw( s)<br />

then, for <strong>an</strong>y enough regular ϕ : R → R :<br />

n<br />

0<br />

0<br />

0 0 <br />

t<br />

ϕ( t, M( t)) = ϕ (0 , M ) + ∂ ϕ( s, M( s)) g( s, ., . ) ds+<br />

t t<br />

∂ ϕ( s, M( s)) g( s, ., . ) ds + ∂ ϕ( s, M( s)) f( s, ., . ) dw( s)+<br />

0<br />

2<br />

t<br />

1 2<br />

∂2, 2ϕ(<br />

s, M( s)) f( s, ., . ) ds.<br />

2 0<br />

<br />

So E ϕ( t, M( t)) dx = E ϕ (0 , M ) dx +<br />

0<br />

0<br />

t<br />

<br />

E ∂ ϕ( s, M( s)) g( s, ., . ) ds dx +<br />

0<br />

t<br />

<br />

E ∂ ϕ( s, M( s)) g( s, ., . ) ds dx +<br />

0<br />

t<br />

<br />

1<br />

2<br />

E ∂2, 2ϕ(<br />

s, M( s)) f( s, ., . ) ds dx .<br />

2<br />

0<br />

0<br />

0 0<br />

1<br />

2<br />

2<br />

0<br />

4<br />

0<br />

0<br />

1<br />

2<br />

2


The <strong>Stochastic</strong> energy equality :<br />

2 2 2 2<br />

Let f, M <strong>an</strong>d M <strong>in</strong> L (0 , T ; L ( , H )) , L (0 , T ; L ( , V )) <strong>an</strong>d<br />

2 2 ′<br />

L (0 , T ; L ( , V )) respectively, Ϝt<br />

-measurable, such that for t <strong>in</strong> [0 , T ] ,<br />

t t<br />

<br />

M( t) = M + M ( s) ds + f( s, ., . ) dw( s)<br />

then M c<strong>an</strong> be consi<strong>de</strong>r as a cont<strong>in</strong>uous adapted process <strong>an</strong>d, P a.s., for <strong>an</strong>y t,<br />

2<br />

|| M( t) || = || M ||<br />

t<br />

<br />

+ 2 < M ( s ) , M( s ) > ds+<br />

4 Denition <strong>of</strong> a Solution<br />

The couple ( u, ϕ)<br />

∈ V H is a solution <strong>of</strong> the system (1) - (2) <strong>an</strong>d (3) if<br />

n<br />

∞ ∞<br />

n<br />

V ′ ,V<br />

t t<br />

<br />

2<br />

2 M ( s) f( s) dx dw( s) + f ( s, ., . ) ds.<br />

∀v ∈ V, ∀t ∈ [0 , T ];<br />

t <br />

( u( t) u ) v dx +<br />

t <br />

∇( u) ∇v dx ds + <br />

∞<br />

( u u ) v d ds =<br />

<br />

5 Existence <strong>of</strong> a Solution<br />

5.1 Discretization <strong>of</strong> the system<br />

0<br />

<br />

t t<br />

ϕ( t) ϕ = (1 ϕ) J( u) ds + b( u, ϕ) dw( s ) .<br />

0<br />

u0 <strong>an</strong>d ϕ0 be<strong>in</strong>g given, let us <strong>in</strong>ductively construct the sequence ( u n, ϕn)<br />

n <strong>in</strong> the follow<strong>in</strong>g way :<br />

( u n+1 , ϕn+1) ∈ V H is the solution <strong>of</strong> the system : P a.s. <strong>an</strong>d <strong>an</strong>y v <strong>in</strong> V ;<br />

<br />

( u u ) v dx + t ( u ) v dx + t ( u u ) v d =<br />

∞<br />

n+1 n ∇ n+1 ∇ n+1 n+1<br />

<br />

<br />

<br />

0<br />

0 ϕ 1,<br />

<strong>an</strong>d<br />

( ϕ( t) ϕ ) v dx;<br />

In or<strong>de</strong>r to discretize the system (4) - (5), let us consi<strong>de</strong>r, for N 1 <strong>an</strong>d n ∈ { 0, 1, 2 , .., N}<br />

:<br />

<br />

n<br />

0 2<br />

0 0<br />

0 0<br />

0 0 <br />

<br />

0 0<br />

In or<strong>de</strong>r to prove the existence <strong>of</strong> a solution to our system, we propose to adapt the <strong>de</strong>monstration<br />

given <strong>in</strong> Breckner et al [4] or <strong>in</strong> Grecksch et al [8] for a stochastic partial differential equation. This<br />

method (the Rothe method) is based on a time discretization, implicit for the partial differential<br />

equation <strong>an</strong>d explicit for the stochastic differential equation (for the Itô <strong>in</strong>tegration).<br />

<br />

( ϕ ϕ ) v dx;<br />

n+1 n<br />

ϕ ϕ = tJ( u ) (1 ϕ ) + b( u , ϕ )( w w ) .<br />

n+1 n n n n n n+1 n<br />

0<br />

0<br />

t =<br />

T<br />

N<br />

= <br />

,<br />

t n T,<br />

w = w( n t ) ,<br />

u = u ( n t ) .<br />

5<br />

(4)<br />

(5)<br />

(6)<br />

(7)


Remark 1 For technical reasons (lost <strong>of</strong> the maximum pr<strong>in</strong>ciple for ( ϕn)<br />

), we need to use<br />

J( un) (1 ϕn) <strong>in</strong>stead <strong>of</strong> J( un) (1 ϕ n)<br />

. We then have to prove that, at the limit when N goes<br />

to <strong>in</strong>nity, the solution ϕ obta<strong>in</strong>ed satises 0 ϕ 1.<br />

Proposition 1 There exists a sequence ( u n, ϕn) n∈ , ( u n, ϕn)<br />

Ϝnt<br />

-measurable, with values <strong>in</strong><br />

V H,<br />

satisfy<strong>in</strong>g (6) <strong>an</strong>d (7).<br />

Pro<strong>of</strong>. The result leads from a classical xed po<strong>in</strong>t argument (see Grecksch et al [8] for example<br />

<strong>in</strong> the context <strong>of</strong> stochastic calculus). <br />

5.2 A priori estimates<br />

Lemma 2<br />

N<br />

2<br />

2<br />

∀n ∈ { 0, 1, .., N } , E|| ϕ || M so t<br />

E|| ϕ || M,<br />

N<br />

1<br />

Pro<strong>of</strong>. S<strong>in</strong>ce [( ϕ ) ( ϕ ) + ( ϕ ϕ ) ] = ( ϕ ϕ ) ϕ ,<br />

tJ( un) (1 ϕn) ϕn+1 + ϕn+1 b( u n, ϕn)( wn+1 w n)<br />

.<br />

<br />

As, ϕ b( u , ϕ ) is Ϝ -measurable, E ϕ b( u , ϕ )( w w ) =<br />

<br />

E ( ϕ ϕ ) b( u , ϕ )( w w ) + E { ϕ b( u , ϕ )( w w ) } =<br />

<br />

E ( ϕ ϕ ) b( u , ϕ )( w w ) <br />

1<br />

2<br />

E( ϕn+1 ϕn) + Ct. 4<br />

<br />

S<strong>in</strong>ce, E tJ( u ) (1 ϕ ) ϕ C t E( ϕ ) + 1 ,<br />

<br />

2 2<br />

2<br />

E( ϕ ) E( ϕ ) + E( ϕ ϕ ) C t E( ϕ ) + 1 .<br />

n n<br />

2<br />

E( ϕ ) + E( ϕ ϕ ) C t E( ϕ ) + C <br />

n<br />

2<br />

C t E( ϕ ) + C + C tE( ϕ ) .<br />

n<br />

2<br />

Whereas t 0,<br />

∃M > 0, ∀n ∈ { 0, 1, .., N 1}<br />

,<br />

n<br />

<strong>an</strong>d E|| ϕ ϕ || M.<br />

n=0<br />

1 2 2<br />

2<br />

[( ϕn+1) ( ϕn) + ( ϕn+1 ϕn)<br />

] =<br />

2<br />

∀ n, E|| ϕ || Ce Ce<br />

E|| ϕ ϕ || M t.<br />

n=0<br />

1 2 2<br />

2<br />

2 n+1 n n+1 n n+1 n n+1<br />

n n n n t n+1 n n n+1 n<br />

Whereafter,<br />

n+1 n n n n+1 n n n n n+1 n<br />

n+1 n n n n+1 n<br />

n n n+1 n+1<br />

2<br />

n+1 n n+1 n n+1<br />

2<br />

n+1<br />

2<br />

n+1<br />

2<br />

k=0<br />

k=0<br />

Gronwall lemma (cf. Ba<strong>in</strong>ov [1]) s<strong>in</strong>ce :<br />

k+1 k<br />

k=0<br />

2<br />

n+1 n<br />

k=0<br />

k n+1<br />

2<br />

k<br />

2 Cnt CT<br />

n+1<br />

2<br />

n+1 n<br />

6<br />

<br />

n<br />

k+1<br />

2


Pro<strong>of</strong>. Accord<strong>in</strong>g to equation (7 ),<br />

<strong>an</strong>d so,<br />

Lemma 4<br />

Thus, the lemma leads from the lemma 2. <br />

E|| ϕ ϕ || C t + Ct. <br />

N<br />

2<br />

2<br />

∀n ∈ { 0, 1, .., N } , E|| u || M so t<br />

E|| u || M,<br />

N<br />

1<br />

∞<br />

n+1 n+1 n+1 n+1 n n+1<br />

<br />

′<br />

n+1 n+1 n+1<br />

N<br />

2<br />

<strong>an</strong>d t<br />

E|| u || M.<br />

Pro<strong>of</strong>. Tak<strong>in</strong>g v = un+1<br />

as a test function <strong>in</strong> (6), one has :<br />

<br />

( u u ) u dx + t ∇( u ) ∇u<br />

dx+<br />

<br />

t ( u u ) u d ( ϕ ϕ ) u dx.<br />

<br />

S<strong>in</strong>ce, un is Fnt measurable, E ( ϕn n) n =<br />

+1 ϕ u +1 dx<br />

<br />

E ( ϕ ϕ ) ( u u ) dx + E ( ϕ ϕ ) u dx <br />

Lemma 5<br />

2 2 2<br />

n+1 n n n n n n+1 n<br />

|| ϕ ϕ || 2 t || J( u ) (1 ϕ ) || + 2 || b( u , ϕ )( w w ) || ,<br />

∃M<br />

> 0,<br />

n=0<br />

<br />

n n + n n n n n<br />

1 2<br />

2<br />

E|| ϕ +1 ϕ || E|| u +1 u || + t E J( u ) (1 ϕ ) u dx.<br />

4<br />

Moreover, as ∇( u ) = ( u ) ∇u<br />

,<br />

n n<br />

1<br />

n + k k<br />

k V<br />

2<br />

1<br />

+1 + <br />

2<br />

2<br />

2<br />

+1<br />

+1 2<br />

E|| u || E|| u u || m t E|| u || <br />

n n<br />

1<br />

∞<br />

+ k L2<br />

k k<br />

2<br />

<br />

0 + +<br />

2<br />

2<br />

+1 2 t<br />

2<br />

E|| u || E|| u || () E|| ϕ +1 ϕ || C.<br />

N n<br />

2<br />

V<br />

N<br />

1<br />

n=1<br />

n=0<br />

Pro<strong>of</strong>. S<strong>in</strong>ce is a lipschitz function, this result is obvious. <br />

n<br />

n=1<br />

n V<br />

n=0<br />

n+1 n n+1 n+1 n+1<br />

<br />

n+1 n n+1 n n+1 n n<br />

<br />

1 <br />

n n + n n n V<br />

2<br />

<br />

∞<br />

n n n<br />

1<br />

2 2<br />

2<br />

+1 +1 + +1<br />

4<br />

<br />

2<br />

( ) + + <br />

2<br />

+1 2<br />

E|| u || E|| u || E|| u u || m tE|| u || <br />

t<br />

E<br />

2<br />

u d E|| ϕ +1 ϕ || C t.<br />

Then, summ<strong>in</strong>g up from 0 to n,<br />

we get :<br />

∃M<br />

> 0,<br />

<br />

<br />

<br />

<br />

2 2<br />

n+1 n<br />

k=0<br />

k=0<br />

2<br />

n+1 n<br />

E|| u u || M,<br />

k=0<br />

k=0<br />

t E|| ( u ) || M <strong>an</strong>d E|| ( u ) ( u ) || M.<br />

7<br />

<br />

n<br />

n+1 n<br />

2<br />

2


Lemma 6<br />

Assum<strong>in</strong>g that u ∈ V ; ∃M > 0, ∀n ∈ { 0, 1,<br />

.., N } ,<br />

Pro<strong>of</strong>. Tak<strong>in</strong>g = ( n+1) ( n)<br />

as a test function <strong>in</strong> (6); for <strong>an</strong>y positive :<br />

<br />

<br />

2 2<br />

( n+1 n)( ( n+1 ) ( n)) + ||∇ n+1 || ||∇ n ||<br />

<br />

<br />

2<br />

||∇ n+1 n || n+1 n <br />

<br />

( ) ( )<br />

2<br />

+ <br />

v u u a<br />

t<br />

u u u u dx u u<br />

t<br />

[ ( u ) ( u )] + t<br />

2<br />

( u ) ( u ) d<br />

As t<br />

is small enough,<br />

2 1<br />

|| ′ || ∞<br />

2<br />

n<br />

2<br />

E|| G( u ) G( u ) || M<br />

k=0<br />

k+1 k<br />

n<br />

2<br />

<strong>an</strong>d t || ( u ) ( u ) || M.<br />

k=0<br />

k+1 k V<br />

∞<br />

n+1 L2 n n V<br />

<br />

2 <br />

2<br />

|| || () + || +1 ||<br />

2<br />

<br />

<br />

t<br />

a t<br />

u<br />

( u ) ( u ) +<br />

a<br />

2<br />

( ϕ ϕ )( ( u ) ( u )) dx.<br />

n+1 n n+1 n<br />

2<br />

n+1 n n+1 n n+1 n<br />

<strong>an</strong>d so : <br />

1<br />

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

+<br />

2 <br />

<br />

<br />

||∇ ( ) || ||∇ ( ) || + || ( ) ( ) || || ||<br />

2<br />

2<br />

<br />

2<br />

+ || n+1 n || V n+1 n n+1 n<br />

<br />

<br />

Moreover, one has :<br />

<br />

<br />

u u u u dx t u u d<br />

t<br />

t<br />

u u u u<br />

a<br />

a t<br />

( u ) ( u ) + ( ϕ ϕ )( ( u ) ( u )) dx.<br />

2<br />

E ( ϕ ϕ )( ( u ) ( u )) dx<br />

2 2<br />

2<br />

∞<br />

n+1 n n+1 n V n+1 L2<br />

2 u ()<br />

n+1 n n+1 n<br />

2<br />

2<br />

|| n+1 n|| + || n+1 n || <br />

2<br />

|| ||<br />

<br />

1<br />

b<br />

E ϕ ϕ E ( u ) ( u )<br />

2b<br />

b<br />

E ϕ<br />

2<br />

ϕ<br />

E ( u u )( ( u ) ( u )) dx .<br />

So, for b = <strong>an</strong>d a = , we get :<br />

′<br />

2 || || ∞<br />

n+1 n + n+1 n n+1 n<br />

2b<br />

<br />

<br />

1<br />

<br />

2 2<br />

( n+1 n)( ( n+1 ) ( n)) + ||∇ n+1 || ||∇ n ||<br />

4 <br />

<br />

2<br />

|| n+1 n || V n+1 n <br />

<br />

( ) ( )<br />

2<br />

+ <br />

t<br />

E u u u u dx E u E u<br />

t<br />

E ( u ) ( u ) + t E ( u ) ( u ) d<br />

4<br />

Hence, after summ<strong>in</strong>g up from 0 to n,<br />

we get<br />

k=0<br />

0<br />

<br />

t( ( u ) ( u )) ( u u )( ( u ) ( u ))<br />

2 2 ∞<br />

n+1 n n+1 L2<br />

2 + <br />

()<br />

C E|| ϕ ϕ || t E|| u || .<br />

n n<br />

1<br />

2<br />

2<br />

|| ( k+1) ( k)<br />

|| + || k+1 k || V<br />

4<br />

t<br />

E G u G u<br />

E ( u ) ( u ) +<br />

4<br />

k=0<br />

<br />

t<br />

2<br />

E||∇ ( un+1) || + t E ( un+1) d t E ( u0) d+<br />

2<br />

<br />

8


5.3 Existence<br />

N N<br />

N∈<br />

N∈<br />

t<br />

<br />

E u t E u C E ϕ ϕ .<br />

2<br />

||∇<br />

2<br />

0 || ||<br />

∞<br />

+1|| 2<br />

|| ||<br />

=0<br />

2 n<br />

n<br />

( ) + k L () +<br />

k+1 2<br />

k V<br />

k<br />

k=0<br />

So, lead<strong>in</strong>g from the above lemmas, one has<br />

n n∈<br />

n<br />

t<br />

2<br />

E|| ( uk+1) ( uk) || V C. <br />

2<br />

For <strong>an</strong>y sequences ( x ) <strong>of</strong> H (resp. <strong>of</strong> V),<br />

let us set :<br />

N<br />

N<br />

N<br />

x ( t) = x I<br />

( t) if t > 0 <strong>an</strong>d x (0) = x<br />

N <br />

N xk xk1<br />

x ( t)<br />

=<br />

( t ( k 1) t) + xk1 I[(<br />

k1)t,k t]<br />

( t ) .<br />

t<br />

( x ) <strong>an</strong>d ( x<br />

) are some sequences <strong>in</strong> H (resp. V)<br />

<strong>an</strong>d one has :<br />

<br />

|| x || t E|| x || ,<br />

N<br />

N 2 = <br />

k<br />

2<br />

H<br />

N<br />

k=1<br />

<br />

L∞<br />

( H)<br />

=<br />

k =1,..,N<br />

k<br />

2<br />

H<br />

N N <br />

N<br />

1<br />

2 t<br />

= 3<br />

k+1 k<br />

2<br />

H<br />

N <br />

∂t<br />

k=0<br />

N<br />

1<br />

2<br />

L 2(]0<br />

,T [ <br />

,V ′ ) = t<br />

E<br />

k=0<br />

xk+1 xk<br />

2 || V ′<br />

t<br />

One has the same k<strong>in</strong>d <strong>of</strong> results, replac<strong>in</strong>g H, H, H respectively by V, V,<br />

V.<br />

Lemma 7<br />

N<br />

N∈<br />

- ( ϕ ) is a boun<strong>de</strong>d sequence <strong>in</strong> L ( H)<br />

<strong>an</strong>d H as well,<br />

N<br />

N<br />

N∈<br />

- ( ϕ ϕ<br />

) converges toward 0 <strong>in</strong> H,<br />

N<br />

N<br />

N N<br />

N∈<br />

k ]( k1)t,k t]<br />

0<br />

- ( u ) ∈ is a boun<strong>de</strong>d sequence <strong>in</strong> H,<br />

L ( H) , L (0 , T ; L ( )) <strong>an</strong>d L ( ) as well,<br />

- ( u u<br />

) converges toward 0 <strong>in</strong> H.<br />

N N<br />

As ( u ) = ( u)<br />

,<br />

N ∞<br />

- ( ( u )) N∈<br />

is a boun<strong>de</strong>d sequence <strong>in</strong> L ( V)<br />

<strong>an</strong>d V as well,<br />

N<br />

N - ( ( u ) ( u)<br />

) converges toward 0 <strong>in</strong> H.<br />

N∈<br />

n<br />

1<br />

2<br />

2<br />

|| ( k+1) ( k) || + ||∇ n+1<br />

||<br />

4<br />

t<br />

E G u G u E ( u ) +<br />

4<br />

k=0<br />

k=1<br />

k=1<br />

k=0<br />

|| x || Max E|| x || ,<br />

|| x x || E|| x x || ,<br />

∂x<br />

|| || ||<br />

The a priori estimates lead to :<br />

∞<br />

∞ ∞<br />

5.3.1 Study <strong>of</strong> the stochastic differential equation<br />

2 2<br />

Accord<strong>in</strong>g to the lemma 7, it is possible to extract a subsequence <strong>of</strong> ( ϕ ) ∈ , still noted ( ϕ ) ∈ ,<br />

N<br />

∞<br />

such that ϕ converges weakly towards ϕ <strong>in</strong> H (<strong>an</strong>d L ( H) weak as well).<br />

Step 1 : energy equality<br />

9<br />

.<br />

N N<br />

N N<br />

(8)


Consi<strong>de</strong>r<strong>in</strong>g (7), one has, for <strong>an</strong>y <strong>in</strong>teger n,<br />

<strong>an</strong>d so,<br />

where ε( t) goes to 0 <strong>in</strong> H with t<br />

.<br />

Moreover, ( J( u ) (1 ϕ )) ∈ is a boun<strong>de</strong>d sequence <strong>in</strong> H,<br />

as well as ( b( u , ϕ )) ∈ ; so,<br />

there exists a subsequence <strong>of</strong> ( ϕ ) , still noted ( ϕ ) , <strong>an</strong>d, J <strong>an</strong>d B <strong>in</strong> H,<br />

such that :<br />

from H to H,<br />

one has<br />

n<br />

1 n<br />

1<br />

k k<br />

k=0<br />

k=0<br />

ϕ ϕ = t J( u ) (1 ϕ ) + b( u , ϕ )( w w )<br />

n<br />

0<br />

( n 1)t<br />

( n 1)t<br />

N N<br />

ϕ ( n t) ϕ (0) =<br />

N N<br />

J( u ) (1 ϕ ) ds +<br />

N N<br />

b( u , ϕ ) dw( s)+<br />

t J( u ) (1 ϕ ) + b( u , ϕ )( w w ) .<br />

Thus, for <strong>an</strong>y t <strong>in</strong> [0 , T ] <strong>an</strong>d n such that ( n 1)t < t nt, 0<br />

0 0 0 0 1 0<br />

t t<br />

N N<br />

ϕ ( t) ϕ (0) =<br />

N N<br />

J( u ) (1 ϕ ) ds +<br />

N N<br />

b( u , ϕ ) dw( s)+<br />

S<strong>in</strong>ce <strong>an</strong>d b are boun<strong>de</strong>d functions,<br />

0 0<br />

t J( u ) (1 ϕ ) + b( u , ϕ )( w w ) <br />

0 0 0 0 1 0<br />

t t<br />

n n<br />

J( u ) (1 ϕ ) ds +<br />

n n<br />

b( u , ϕ ) dw( s ) .<br />

( n1) t<br />

( n1)t t t<br />

N N<br />

ϕ ( t) ϕ (0) =<br />

N N<br />

J( u ) (1 ϕ ) ds +<br />

N N<br />

b( u , ϕ ) dw( s) + ε( t)<br />

0 0<br />

N N<br />

N<br />

N N<br />

N∈<br />

N∈<br />

N N<br />

J( u ) (1 ϕ ) ⇀ J ,<br />

N N<br />

b( u , ϕ ) ⇀ B.<br />

t t<br />

Remark<strong>in</strong>g that the applications u ↦→ u ds <strong>an</strong>d u ↦→ u dw( s)<br />

are cont<strong>in</strong>uous l<strong>in</strong>ear functions<br />

0 0<br />

t t t t<br />

N N<br />

J( u ) (1 ϕ ) ds +<br />

N N<br />

b( u , ϕ ) dw( s) ⇀ J ds + B dw( s ) .<br />

t t<br />

ϕ( t) ϕ(0) = J ds + B dw( s) <strong>in</strong> H;<br />

for <strong>an</strong>y t (s<strong>in</strong>ce the Itô <strong>in</strong>tegral implies that ϕ c<strong>an</strong> be chosen cont<strong>in</strong>uous).<br />

2at<br />

2<br />

Then, for <strong>an</strong>y a > 0 , by apply<strong>in</strong>g the Itô formula with f( t, x) = e || x||<br />

, we get :<br />

2at<br />

2<br />

e E|| ϕ( t) || E|| ϕ || =<br />

k k k+1 k<br />

0 0 0 0<br />

So, as N goes to <strong>in</strong>nity, we get,<br />

0 0<br />

10<br />

0 2<br />

0<br />

N N N<br />

(9)<br />

(10)


t<br />

t <br />

2as 2<br />

2as<br />

2<br />

2 a e E|| ϕ|| ds + e E 2 J ϕ dx + || B|| ds.<br />

Step 2 : discrete energy <strong>in</strong>equality<br />

Let us look for a similar formula to (11), <strong>in</strong> the discrete case.<br />

Let us multiply (7) by ϕ . Then,<br />

1 2 2<br />

2<br />

[( ϕn+1) ( ϕn) + ( ϕn+1 ϕn) ] = tJ( un) (1 ϕn) ϕn+1+<br />

2<br />

( ϕ ϕ ) b( u , ϕ )( w w ) + ϕ b( u , ϕ )( w w ) .<br />

2<br />

a( n+1)t 1 a( n+1)t 2 <strong>an</strong>t 2 2<br />

a( n+1)t 2<br />

[( e ϕn+1) ( e ϕn) + e ( ϕn+1 ϕn)<br />

] =<br />

2<br />

1 2 2 a( n+1)t 2<strong>an</strong>t 2<br />

a( n+1)t ( ϕn) ( e e ) + e tJ( un) (1 ϕn) ϕn+1+<br />

2<br />

2<br />

a( n+1)t e ( ϕ ϕ ) b( u , ϕ )( w w )+<br />

2<br />

a( n+1)t n n n n+1 n<br />

e ϕ b( u , ϕ )( w w ) .<br />

<br />

|| || || || || || <br />

<br />

<br />

<br />

<br />

|| ||<br />

<br />

<br />

<br />

1 2 a( n+1)t 2 2<strong>an</strong>t 2 2 a( n+1)t 2<br />

[ e E ϕn+1 e E ϕn + e E ϕn+1 ϕn<br />

]<br />

2<br />

2at 2<strong>an</strong>t 2 1 e<br />

2 a( n+1)t ae E ϕn<br />

+ e tE J( un) (1 ϕn) ϕn dx+<br />

2a<br />

<br />

2 a( n+1)t e tE J( u ) (1 ϕ )( ϕ ϕ ) dx+<br />

<br />

2 a( n+1)t 2<br />

2 e<br />

2<br />

E b( u n, ϕn) ( wn+1 wn) dx +<br />

E|| ϕn+1 ϕ n||<br />

.<br />

2<br />

2 a( n+1)t <br />

2<br />

n+1<br />

n+1 n n n n+1 n n n n n+1 n<br />

So, multiply<strong>in</strong>g aga<strong>in</strong> by e , we get :<br />

Thus,<br />

n+1 n n n n+1 n<br />

S<strong>in</strong>ce (7) implies that<br />

<br />

E J( u ) (1 ϕ )( ϕ ϕ ) dx = tE [ J( u ) (1 ϕ<br />

2<br />

)] dx,<br />

one has,<br />

e<br />

<br />

0<br />

<br />

2 a( n+1)t 2 2<strong>an</strong>t 2 2<strong>an</strong>t<br />

2<br />

e E|| ϕn+1|| e<br />

<br />

E|| ϕn|| 2a t e E|| ϕn||<br />

+<br />

2<br />

a( n+1)t 2e tE J( u ) (1 ϕ ) ϕ<br />

2<br />

dx + C( t)<br />

+<br />

<br />

2<br />

a( n+1)t t e E b( u , ϕ<br />

2<br />

) dx.<br />

Therefore, after summ<strong>in</strong>g up from 0 to N 1,<br />

we get,<br />

n n n+1 n<br />

n n n+1 n n n<br />

<br />

n n n<br />

n n<br />

N<br />

1<br />

2aNt<br />

N 2<br />

2akt<br />

2<br />

e E|| ϕ || E|| ϕ || 2a t e E|| ϕ || + C t+<br />

k=0<br />

N<br />

1<br />

<br />

2<br />

a( k+1)t 2 t e E J( u ) (1 ϕ ) ϕ dx+<br />

k=0<br />

k k k<br />

N<br />

1<br />

<br />

2<br />

a( k+1)t 2<br />

t e E b( u , ϕ ) dx,<br />

k=0<br />

<br />

<br />

0 2<br />

0<br />

<br />

11<br />

<br />

<br />

<br />

k k<br />

k<br />

(11)


<strong>an</strong>d,<br />

<br />

e E ϕ ( T ) E ϕ 2 a e E ϕ dt+<br />

2 2<br />

0<br />

<br />

2<br />

T<br />

aT<br />

||<br />

N<br />

|| || || <br />

0<br />

2at ||<br />

N 2<br />

||<br />

N<br />

1<br />

k=0<br />

2akt 2<br />

a( k+1)t T T <br />

2at N N N<br />

2at<br />

N N 2<br />

2 e E J( u ) (1 ϕ ) ϕ dx dt + e E b( u , ϕ ) dx dt.<br />

0<br />

0<br />

Step 3 : Lower semi-cont<strong>in</strong>uity<br />

T T<br />

S<strong>in</strong>ce applications u ↦→ u ds <strong>an</strong>d u ↦→ u dw( s)<br />

are l<strong>in</strong>ear cont<strong>in</strong>uous functions from H to<br />

H,<br />

0 0<br />

T T T T<br />

N N<br />

J( u ) (1 ϕ ) ds +<br />

N N<br />

b( u , ϕ ) dw( s) ⇀ J ds + B dw( s)<br />

H<br />

0 0 0 0<br />

<strong>an</strong>d th<strong>an</strong>ks to (10), ϕ ( T ) ⇀ ϕ( T ) <strong>in</strong> H.<br />

Let us set<br />

Remark that 0 <strong>an</strong>d that (11) implies,<br />

2aT N 2 2aT<br />

2<br />

T<br />

T <br />

+ E|| ϕ || 2 a<br />

2as 2<br />

e E|| ϕ|| ds +<br />

2as<br />

e E 2<br />

2<br />

J ϕ dx + || B|| ds =<br />

0 2<br />

0<br />

2aT 2 2aT<br />

N 2<br />

+ e E|| ϕ( T ) || = e lim E|| ϕ ( T ) || L<br />

where, we note L = E|| ϕ ||<br />

<br />

+ lim { 2 a e E|| ϕ || dt+<br />

0 2<br />

N∈<br />

2 N 2<br />

T<br />

0<br />

2at<br />

N 2<br />

T T <br />

2at N N N<br />

2at<br />

N N 2<br />

2 e E J( u ) (1 ϕ ) ϕ dx dt + e E b( u , ϕ ) dx dt } .<br />

S<strong>in</strong>ce the sequences<br />

0<br />

0<br />

T<br />

T <br />

2at<br />

N 2<br />

2at<br />

N N N<br />

( e E|| ϕ || dt ) , ( e E J( u ) (1 ϕ ) ϕ dx dt ) ,<br />

0<br />

0<br />

T<br />

T<br />

2at<br />

N N 2<br />

2at<br />

N 2<br />

( e E b( u , ϕ ) dx dt) <strong>an</strong>d e E|| ϕ ϕ || dt<br />

0<br />

N<br />

2C t ( e e ) + C t+<br />

Thus,<br />

<br />

N∈<br />

are boun<strong>de</strong>d <strong>in</strong> R,<br />

a subsequence c<strong>an</strong> be found, still <strong>in</strong><strong>de</strong>xed by N,<br />

such that all <strong>of</strong> the above<br />

sequences converge.<br />

consequently,<br />

T<br />

2at<br />

N 2<br />

L = E|| ϕ || 2 a lim e E|| ϕ || dt+<br />

0 2<br />

T <br />

2 lim<br />

2t<br />

e E<br />

N N N<br />

J( u ) (1 ϕ ) ϕ dx dt+<br />

0<br />

<br />

T <br />

lim<br />

2at<br />

e E<br />

N N 2<br />

b( u , ϕ ) dx dt.<br />

0<br />

E|| ϕ( T ) || lim E|| ϕ ( T ) || .<br />

= e lim E|| ϕ ( T ) || e E|| ϕ( T ) || .<br />

<br />

12<br />

0<br />

0<br />

<br />

0<br />

<br />

<br />

<br />

N∈<br />

(12)


Let us set,<br />

Remark that L<br />

<strong>an</strong>d that,<br />

T<br />

L = 2 a<br />

2at<br />

N 2<br />

e E|| ϕ ϕ || dt+<br />

T <br />

2<br />

2at<br />

e E<br />

N N N<br />

[ J( u ) (1 ϕ ) J( u) (1 ϕ)]( ϕ ϕ) dxdt<br />

T T<br />

+<br />

2at e E<br />

N N<br />

2<br />

( b( u , ϕ ) b( u, ϕ)) dxdt +<br />

2at<br />

N 2<br />

e E|| ϕ ϕ || dt.<br />

0<br />

<br />

0<br />

<br />

<br />

0<br />

T T<br />

(1 2 a + c( J, , b)) 2at N 2<br />

e E|| ϕ ϕ || dt + c( J, , b) 2at<br />

N 2<br />

e E|| u u || dt,<br />

0<br />

T T <br />

L = 2 a<br />

2at N 2<br />

e E|| ϕ || dt + 2<br />

2at<br />

e E<br />

N N N<br />

J( u ) (1 ϕ ) ϕ dxdt+<br />

0<br />

0<br />

T T<br />

<br />

2at N N 2<br />

2at<br />

N<br />

e E ( b( u , ϕ )) dxdt + 4 a e E[ ϕϕ dx] dt<br />

0<br />

<br />

T T T <br />

2at 2<br />

2at N 2<br />

2at<br />

2<br />

2 a e E|| ϕ|| dt + e E|| ϕ ϕ || dt + e E ( b( u, ϕ)) dx dt<br />

0<br />

0<br />

T T <br />

2<br />

2at e E<br />

N N<br />

b( u , ϕ ) b( u, ϕ) dxdt 2<br />

2at<br />

e E<br />

N N<br />

J( u ) (1 ϕ ) ϕ dxdt<br />

0<br />

<br />

T <br />

2 2at<br />

e E<br />

N<br />

J( u) (1 ϕ)( ϕ ϕ) dx dt.<br />

Thus, at the limit as N goes to <strong>in</strong>nity, one has :<br />

S<strong>in</strong>ce,<br />

N<br />

N<br />

0<br />

<br />

T<br />

lim L = L E|| ϕ || + 2 a<br />

2at<br />

2<br />

e E|| ϕ|| dt+<br />

0 2<br />

T T <br />

lim<br />

2at N 2<br />

e E|| ϕ ϕ || dt +<br />

2at<br />

e E<br />

2<br />

( b( u, ϕ)) dx dt<br />

0<br />

T T<br />

<br />

2at 2at<br />

2 e E B b( u, ϕ) dx dt 2 e E J ϕ dx dt.<br />

0<br />

<br />

T T <br />

2at 2<br />

2at<br />

e E ( b( u, ϕ)) dx dt 2 e E B b( u, ϕ) dx dt<br />

=<br />

0<br />

<br />

N<br />

N<br />

13<br />

0<br />

0<br />

0<br />

0<br />

0<br />

0<br />

<br />

0<br />

0<br />

<br />

<br />

<br />

0<br />

<br />

<br />

<br />

(13)


we get,<br />

T T <br />

2at 2<br />

2at<br />

2<br />

e E ( B b( u, ϕ)) dx dt e E B dx dt,<br />

T T <br />

2at 2<br />

2at<br />

2<br />

lim L = L E|| ϕ || + 2a<br />

e E|| ϕ|| dt e E B dx dt<br />

N<br />

0<br />

T T<br />

2<br />

2at e E J ϕ dx dt + lim<br />

2at<br />

N 2<br />

e E|| ϕ ϕ || dt+<br />

5.3.2 Study <strong>of</strong> the parabolic equation :<br />

Accord<strong>in</strong>g to the lemma 7, it is possible to extract a subsequence <strong>of</strong> ( u ) ∈ , still noted ( u ) ∈ ,<br />

N N ∞<br />

such that u (resp. ( u ) N∈<br />

) converges weakly towards u (resp. ) <strong>in</strong> V as well as <strong>in</strong> L ( V)<br />

weak- .<br />

1<br />

N N<br />

Let us consi<strong>de</strong>r v ∈ H ( Q ) , <strong>an</strong>d set v the step function <strong>an</strong>d v<br />

the piecewise aff<strong>in</strong>e cont<strong>in</strong>uous<br />

function from [0 , T ] to V, built from the sequence v = v( n t ) as <strong>in</strong>troduce <strong>in</strong> (8) .<br />

Step 1 : energy equality<br />

Accord<strong>in</strong>g to (6), for <strong>an</strong>y n, one has :<br />

0<br />

T <br />

2at<br />

2<br />

e E ( B b( u, ϕ)) dx dt.<br />

N<br />

<br />

T<br />

T<br />

2at 2<br />

2at<br />

N 2<br />

+ e E ( B b( u, ϕ)) dx dt + lim e E|| ϕ ϕ || dt.<br />

0<br />

<br />

<br />

<br />

n 1<br />

v +1 v<br />

unvn u0v0 dx t<br />

uk t<br />

<br />

∇ ∇ <br />

<br />

<br />

n 1<br />

n 1<br />

t ( u ) v dx + t ( u u ) v d =<br />

k=0<br />

<br />

<br />

<br />

<br />

0 2<br />

Follow<strong>in</strong>g (12), one nds that lim L<br />

Thus,<br />

0<br />

<br />

k+1 k+1<br />

<br />

k=0<br />

n<br />

<br />

1<br />

∞<br />

k+1 k+1 k+1<br />

<br />

v +1 v<br />

ϕnvn ϕ0v0 dx t<br />

ϕk t<br />

<br />

N N<br />

dx+<br />

<br />

<br />

<br />

<br />

<br />

∇ ∇ <br />

<br />

nt N<br />

N N N<br />

N<br />

u ( n t) v ( n t) u0v (0) dx u<br />

<br />

0 <br />

+<br />

nt n 1<br />

N N<br />

( ) + <br />

N <br />

( k+1 k)<br />

+<br />

∂v<br />

∂t<br />

dx ds<br />

u v dx ds t u u ∂v<br />

∂t dx<br />

0 <br />

nt <br />

<br />

N ∞,N<br />

N<br />

( u u ) v d ds =<br />

N N<br />

ϕ ( n t) v<br />

( n t) dx<br />

<br />

<br />

<br />

<br />

<br />

<br />

<br />

nt N<br />

n 1<br />

N<br />

N<br />

N<br />

<br />

ϕ v (0) dx ϕ + ( k+1 k)<br />

∂v<br />

∂v<br />

dx ds t ϕ ϕ<br />

∂t ∂t dx.<br />

0<br />

t [0 , T ] n ( n 1)t < t nt, <br />

t <br />

N N N<br />

u ( t) v ( t) u0v (0) dx <br />

N<br />

N <br />

u +<br />

∂v<br />

Therefore, for <strong>an</strong>y <strong>in</strong> <strong>an</strong>d for such that we get,<br />

∂t<br />

dx ds<br />

0<br />

<br />

n<br />

0<br />

k=0<br />

<br />

k=0<br />

<br />

0<br />

k=0<br />

<br />

0 <br />

<br />

0 <br />

14<br />

0 <br />

0<br />

0<br />

k k<br />

k k<br />

k=0<br />

<br />

<br />

dx.<br />

<br />

N N<br />

(14)


S<strong>in</strong>ce<br />

t t <br />

N N<br />

N ∞,N<br />

N<br />

∇( u ) ∇v dx ds + ( u u ) v d ds =<br />

0 0 <br />

<br />

<br />

t N<br />

N N N<br />

N <br />

ϕ ( t) v ( t) ϕ0v (0) dx ϕ<br />

<br />

0 <br />

<br />

+<br />

n N N<br />

[ ( ) ( )] <br />

n N N<br />

[ ( ) ( )] +<br />

∂v<br />

∂t<br />

dx ds<br />

u v t v n t dx ϕ v t v n t dx<br />

<br />

<br />

<br />

<br />

<br />

<br />

<br />

<br />

1<br />

+1 <br />

1<br />

+1 <br />

<br />

<br />

n t<br />

N<br />

n<br />

t<br />

n t<br />

n<br />

t<br />

+<br />

( ) <br />

<br />

( )<br />

N<br />

u<br />

n<br />

N<br />

k k<br />

n<br />

N<br />

k k<br />

∂v<br />

∂t<br />

∂v<br />

dx ds ϕ<br />

∂t<br />

dx ds<br />

t u u ∂v<br />

∂t<br />

dx t<br />

∂v<br />

ϕ ϕ<br />

∂t dx<br />

=0 <br />

k<br />

<br />

<br />

<br />

=0 <br />

k<br />

nt nt <br />

n N<br />

n ∞,n<br />

N<br />

∇( u ) ∇v dx ds ( u u ) v d ds.<br />

<br />

<br />

<br />

∞,n<br />

V<br />

√ √<br />

n<br />

|| || 2 || n|| 2<br />

n<br />

V<br />

V + || || V<br />

1<br />

<br />

<br />

+1 <br />

<br />

+1 <br />

=0<br />

<br />

n<br />

vk vk<br />

t E ( uk uk) t<br />

k<br />

k=0<br />

k=0<br />

0 0 <br />

<br />

0<br />

nt N<br />

n N N n ∂v<br />

u [ v ( t) v<br />

( n t)] dx = u<br />

( s) ds dx;<br />

∂t<br />

accord<strong>in</strong>g to (4), one has<br />

<br />

nt <br />

<br />

E<br />

<br />

nt <br />

n n<br />

∇( u ) ∇v dx ds E<br />

<br />

<br />

n ∞,n<br />

n<br />

( u u ) v d ds <br />

Moreover, us<strong>in</strong>g lemmas 4 <strong>an</strong>d 2, we get,<br />

<br />

<br />

<br />

t || v ||<br />

<br />

[2 E|| u || + || u<br />

<br />

|| ] <br />

t v t E u C C t v .<br />

0 <br />

0<br />

∞ ∞<br />

<br />

0<br />

<br />

<br />

<br />

dx<br />

<br />

<br />

<br />

<br />

1 1<br />

2<br />

+1 <br />

|| 2<br />

+1 || || || <br />

<br />

<br />

<br />

<br />

<br />

n<br />

n<br />

vk vk<br />

t<br />

E uk uk<br />

t<br />

<br />

C t|| (respectively with ); <strong>an</strong>d then, for <strong>an</strong>y ,<br />

|| <br />

<br />

t N<br />

N N N<br />

N <br />

( ) ( ) 0<br />

(0) <br />

<br />

0 <br />

+<br />

t t <br />

N N<br />

∇ ( ) ∇ +<br />

N ∞,N<br />

N<br />

( ) =<br />

∂v<br />

∂t<br />

ϕ t<br />

C t<br />

u t v t u v dx u ∂v<br />

∂t<br />

dx ds<br />

u v dx ds u u v d ds<br />

where ε( t) goes to 0 with t.<br />

t<br />

<br />

t<br />

n V<br />

t N<br />

N N N<br />

N <br />

ϕ ( t) v ( t) ϕ0v (0) dx ϕ + ( )<br />

∂v<br />

dx ds ε t<br />

∂t<br />

Let us consi<strong>de</strong>r, as proposed <strong>in</strong> Grecksch et al [8], r ∈ L (0 , T ) <strong>an</strong>d b ∈ L () . Then<br />

n V<br />

N<br />

T <br />

T <br />

E<br />

N N<br />

u ( t) v<br />

( t) r( t) b dx dt E<br />

N<br />

u v<br />

(0) r( t) b dx dt<br />

15<br />

t<br />

t<br />

t


A : V → V<br />

<strong>an</strong>d A ∈ V :<br />

∞<br />

∞<br />

T t <br />

N <br />

E u ∂v<br />

∂t<br />

′<br />

′<br />

0 0<br />

<br />

∞<br />

′<br />

V ′ ,V<br />

V ′ ,V<br />

∞<br />

0 0<br />

∞ ′ ∞ ∞<br />

∈<br />

= <br />

T t <br />

N N<br />

r( t) b dx ds dt + E ∇( u ) ∇v<br />

r( t) b dx ds dt+<br />

T t T <br />

E<br />

N ∞,N<br />

N<br />

( u u ) v r( t) b d ds dt = E<br />

N N<br />

ϕ ( t) v ( t) r( t) b dx dt<br />

0 0<br />

<br />

T <br />

T t N<br />

N<br />

N <br />

E ϕ v (0) r( t) b dx dt E ϕ ( )<br />

∂v<br />

r t b dx ds dt<br />

∂t<br />

0<br />

T<br />

+ E ε( t) r( t) b dt.<br />

T T <br />

E u( t) v( t) r( t) b dx dt E u v(0) r( t) b dx dt<br />

T t T t <br />

E u ( ) + ∇ ∇ ( ) +<br />

0 0 0 0 <br />

T t T <br />

∞<br />

( ) ( ) = ( ) ( ) ( ) <br />

∂v<br />

r t b dx ds dt E v r t b dx ds dt<br />

∂t<br />

E u u v r t b d ds dt E ϕ t v t r t b dx dt<br />

T T t <br />

E ϕ v(0) r( t) b dx dt E ϕ ( )<br />

∂v<br />

r t b dx ds dt.<br />

∂t<br />

the monotone operator :<br />

<br />

< Au, v > = ∇( u) ∇v<br />

dx + uv d,<br />

<br />

< A , v > = ∇ ∇v<br />

dx + uv d.<br />

t t<br />

∞<br />

u( t) u + A ds = F ds + ϕ( t) ϕ ,<br />

<br />

where F V is <strong>de</strong>ned by < F , v > u v d.<br />

N<br />

N<br />

0<br />

Thus, it follows that u ( T ) converges weakly <strong>in</strong> H towards some U <strong>an</strong>d<br />

T T<br />

∞<br />

U = u A ds + F ds + ϕ( T ) ϕ .<br />

∞<br />

0 0<br />

But, s<strong>in</strong>ce one c<strong>an</strong> choose u cont<strong>in</strong>uous, it ensures that u ( T ) ⇀ u( T ) <strong>in</strong> H.<br />

2at<br />

2<br />

Then, by apply<strong>in</strong>g the Itô formula to (15) with f( t, x) = e || x||<br />

<strong>an</strong>d s<strong>in</strong>ce<br />

N<br />

0<br />

0 0<br />

0 0<br />

Then, pass<strong>in</strong>g to the limit to <strong>in</strong>nity with N,<br />

one obta<strong>in</strong>s :<br />

Let us note :<br />

0 0 <br />

0 0 0 <br />

0 0<br />

0<br />

0<br />

t t<br />

ϕ( t) ϕ(0) = J ds + B dw( s ) ,<br />

<br />

<br />

N<br />

0<br />

0 0 <br />

0<br />

<br />

<br />

<br />

Then A is the weak limit <strong>in</strong> V <strong>of</strong> ( Au ) <strong>an</strong>d one has,<br />

0 0<br />

16<br />

0<br />

0<br />

(15)


2at<br />

2<br />

one has : e E|| u( t) || || u || =<br />

t t<br />

2 aE<br />

2as 2<br />

e || u( s) || ds 2 E<br />

2as ∞<br />

e < A , u > ds+<br />

t t t <br />

2 E<br />

2as e<br />

∞<br />

u u dds + 2 E<br />

2as e J u dxds + E<br />

2as<br />

e<br />

2<br />

B dxds.<br />

Step 2 : discrete energy <strong>in</strong>equality<br />

<br />

1 <br />

+<br />

2<br />

<br />

1<br />

Let us look for a similar formula <strong>in</strong> the discrete case. For this, remark that us<strong>in</strong>g the pro<strong>of</strong> <strong>of</strong><br />

lemma 4, we get<br />

2 2<br />

2<br />

|| un+1|| || un|| || un+1 un|| + t ∇( un+1 ) ∇un+1<br />

dx+<br />

2<br />

<br />

t u d t u u d =<br />

∞,n+1<br />

n+1<br />

<br />

t J( u ) (1 ϕ ) u dx + b( u , ϕ ) u dx ( w w ) .<br />

T<br />

2aT N 2<br />

at N<br />

( ) 0 2<br />

2<br />

2 2<br />

0<br />

N<br />

1<br />

2akt 2<br />

a( k+1)t 2as ∞ 2as<br />

<br />

2as 2 2aT<br />

2 || ||<br />

<br />

2aT<br />

N 2<br />

N<br />

N∈<br />

Therefore, us<strong>in</strong>g a similar <strong>de</strong>monstration as the one proposed for the sequence ( ϕ<br />

<br />

) , one has :<br />

e E|| u T || E|| u || a e E|| u || dt<br />

T T <br />

2<br />

2at e<br />

N N<br />

∇( u ) ∇u dx dt 2<br />

2at e<br />

N ∞,N<br />

N<br />

( u u ) u d dt+<br />

2C t ( e e ) + C t+<br />

T T <br />

2at N N N<br />

2at<br />

N N 2<br />

2 e E J( u ) (1 ϕ ) u dx dt + e E b( u , ϕ ) dx dt.<br />

0<br />

Step 2 : lower semi cont<strong>in</strong>uity<br />

Let us set<br />

2aT N 2 2aT<br />

2<br />

= e lim E|| u ( T ) || e E|| u( T ) || .<br />

Remark that 0 <strong>an</strong>d that<br />

0<br />

0<br />

<br />

0<br />

0 2<br />

0<br />

<br />

2<br />

n+1<br />

n n n+1 n n n+1 n+1 n<br />

<br />

0<br />

k=0<br />

0<br />

t t<br />

+ E|| u || 2 aE 2as 2<br />

e || u( s) || ds 2E<br />

2as ∞<br />

e < A , u > ds<br />

0 2<br />

0<br />

t<br />

t<br />

<br />

<br />

+2 E e u u d ds + 2 E e J u dx ds+<br />

0<br />

t<br />

<br />

E e B dx ds = + e E u( T ) =<br />

0<br />

e lim E|| u ( T ) || <br />

T T<br />

where, = E|| u || + 2 2at e E ∞ u u d dt + lim { 2a<br />

2at<br />

N 2<br />

e E|| u || dt<br />

0 2<br />

0<br />

T T <br />

2at N N<br />

2at<br />

N 2<br />

2 e E ∇( u ) ∇u dx dt 2 E e ( u ) d dt+<br />

0<br />

0<br />

0<br />

0<br />

0<br />

17<br />

<br />

0<br />

0<br />

<br />

<br />

<br />

<br />

<br />

(16)


T T <br />

2at N N N<br />

2at<br />

N N 2<br />

2 e E J( u ) (1 ϕ ) u dx dt + e E b( u , ϕ ) dx dt } .<br />

S<strong>in</strong>ce the sequences<br />

0<br />

0<br />

T<br />

T <br />

2at<br />

N 2<br />

2at<br />

N N<br />

( e E|| u || dt ) , ( e E ∇( u ) ∇u<br />

dx dt ) ,<br />

0<br />

N∈<br />

T T <br />

( E<br />

2at<br />

e<br />

N 2<br />

( u ) d dt ) , (<br />

2at<br />

e E<br />

N N N<br />

J( u ) (1 ϕ ) u dx dt)<br />

0<br />

<br />

N∈<br />

T<br />

2at<br />

N 2<br />

<strong>an</strong>d ( e E|| u u || dt)<br />

are boun<strong>de</strong>d <strong>in</strong> R,<br />

a subsequence c<strong>an</strong> be found, still <strong>in</strong><strong>de</strong>xed by N,<br />

such that the above sequences<br />

converge <strong>an</strong>d consequently,<br />

Let us set<br />

N∈<br />

T <br />

2at ∞<br />

= E|| u || + 2 e E u u d dt<br />

N∈<br />

T<br />

T <br />

2at N 2<br />

2at<br />

N N<br />

2a lim e E|| u || dt 2 lim e E ∇( u ) ∇u dx dt<br />

0<br />

T T <br />

2 limE<br />

2at e<br />

N 2<br />

( u ) ddt + 2 lim<br />

2at<br />

e E<br />

N N N<br />

J( u ) (1 ϕ ) u dxdt<br />

0<br />

T <br />

2at<br />

N N 2<br />

+ lim e E b( u , ϕ ) dx dt.<br />

T T<br />

= 2a 2at N 2<br />

e E|| u u || dt 2<br />

2at<br />

N N<br />

e < Au Au , u u > dt+<br />

N<br />

0<br />

T <br />

2<br />

2at<br />

e E<br />

N N N<br />

[ J( u ) (1 ϕ ) J( u) (1 ϕ)]( u u) dx dt+<br />

0<br />

T<br />

T<br />

2at N N<br />

2<br />

2at<br />

N 2<br />

e E ( b( u , ϕ ) b( u, ϕ)) dx dt + e E|| u u || dt.<br />

0<br />

T<br />

Remark that c( J, , b) 2at<br />

N 2<br />

e E|| ϕ ϕ || dt+<br />

Moreover, s<strong>in</strong>ce<br />

N<br />

<br />

<br />

<br />

0<br />

0<br />

0 2<br />

0<br />

0<br />

0<br />

0<br />

0<br />

<br />

T<br />

2at<br />

N 2<br />

(1 2 a + c( J, , b)) e E|| u u || dt.<br />

0<br />

T T<br />

= 2a 2at N 2<br />

e E|| u || dt 2<br />

2at<br />

N N<br />

e E < Au , u > dt+<br />

N<br />

0<br />

18<br />

0<br />

0<br />

0<br />

<br />

<br />

<br />

<br />

0<br />

<br />

<br />

N∈<br />

(17)


converges <strong>an</strong>d lim =<br />

S<strong>in</strong>ce,<br />

one has,<br />

T T <br />

2<br />

2at e E<br />

N N N<br />

J( u ) (1 ϕ ) u dxdt +<br />

2at<br />

e E<br />

N N 2<br />

( b( u , ϕ )) dxdt+<br />

0<br />

<br />

T T <br />

E|| u || + 2 e E u u ddt E|| u || 2 e E u u ddt+<br />

0 2<br />

0<br />

2 ∞<br />

0<br />

∞<br />

<br />

2<br />

at<br />

2at<br />

0<br />

<br />

T T<br />

2<br />

2at e E < Au, u > dt + 2<br />

2at<br />

N<br />

e E < Au, u > dt+<br />

0<br />

T T T<br />

2<br />

2at N<br />

e E < Au , u > dt + 4a 2at e E<br />

N<br />

uu dxdt 2a<br />

2at<br />

2<br />

e E|| u|| dt<br />

0<br />

0<br />

T T <br />

2at N 2<br />

2at<br />

2<br />

+ e E|| u u || dt + e E ( b( u, ϕ)) dx dt<br />

0<br />

0<br />

T T <br />

2<br />

2at e E<br />

N N<br />

b( u , ϕ ) b( u, ϕ) dxdt 2<br />

2at<br />

e E<br />

N N<br />

J( u ) (1 ϕ ) u dxdt<br />

0<br />

<br />

N N<br />

T <br />

2 2at<br />

e E<br />

N<br />

J( u) (1 ϕ)( u u) dx dt,<br />

0<br />

<br />

T T<br />

E|| u || 2<br />

2at e E<br />

∞<br />

u u d dt + 2<br />

2at<br />

N<br />

e E < Au , u > dt+<br />

0 2<br />

0<br />

<br />

T T T <br />

2 a<br />

2at 2<br />

e E|| u|| dt + lim<br />

2at N 2<br />

e E|| u u || dt +<br />

2at<br />

e E<br />

2<br />

( b( u, ϕ)) dxdt<br />

0<br />

0<br />

T T <br />

2 2at e E B b( u, ϕ) dx dt 2<br />

2at<br />

e E J u dx dt.<br />

0<br />

<br />

T T <br />

2at e E<br />

2<br />

( b( u, ϕ)) dx dt 2<br />

2at<br />

e E B b( u, ϕ) dx dt =<br />

0<br />

<br />

T T <br />

2at e E<br />

2<br />

( B b( u, ϕ)) dx dt <br />

2at<br />

e E<br />

2<br />

B dx dt,<br />

0<br />

<br />

T <br />

lim = E|| u || 2<br />

2at e E<br />

∞<br />

u u d dt+<br />

0 2<br />

T T <br />

2<br />

2at N<br />

e E < Au , u > dt <br />

2at<br />

e E<br />

2<br />

B dx dt<br />

0<br />

N<br />

19<br />

0<br />

0<br />

0<br />

0<br />

0<br />

0<br />

<br />

0<br />

0<br />

0<br />

<br />

0<br />

<br />

<br />

<br />

<br />

<br />

<br />

<br />

0


5.3.3 Conclusion<br />

So, accord<strong>in</strong>g to (16), lim <br />

T T<br />

2 2at e E J u dx dt + 2 a<br />

2at<br />

2<br />

e E|| u|| dt+<br />

0<br />

T T <br />

lim<br />

2at N 2<br />

e E|| u u || dt +<br />

2at<br />

e E<br />

2<br />

( B b( u, ϕ)) dx dt.<br />

0<br />

T<br />

T<br />

2at 2<br />

2at<br />

N 2<br />

+ e E ( B b( u, ϕ)) dx dt + lim e E|| u u || dt.<br />

0<br />

In or<strong>de</strong>r to conclu<strong>de</strong>, us<strong>in</strong>g (13) <strong>an</strong>d (17 ), we get : L<br />

<br />

⎡<br />

T T<br />

⎤<br />

(1 2 a + 2 c( J, , b)) ⎣ 2at N 2<br />

e E|| u u || dt +<br />

2at<br />

N 2<br />

e E|| ϕ ϕ || dt ⎦ .<br />

Then, for a 1 + c( J, , b ) , (14) <strong>an</strong>d (18) lead to :<br />

N N<br />

T <br />

2at<br />

2<br />

0 + + 2 e E ( B b( u, ϕ)) dx dt+<br />

T T<br />

lim<br />

2at N 2<br />

e E|| ϕ ϕ || dt + lim<br />

2at<br />

N 2<br />

e E|| u u || dt.<br />

0<br />

So ( u ) <strong>an</strong>d ( ϕ ) converge to u <strong>an</strong>d ϕ <strong>in</strong> H <strong>an</strong>d B = b( u, ϕ)<br />

.<br />

0<br />

As , <strong>an</strong>d J are Lipschitz functions, it is possible to i<strong>de</strong>ntify ( u ) , (1 ϕ) <strong>an</strong>d J( u)<br />

as the<br />

N N limits <strong>of</strong> the sequences ( ( u )) N , ( (1 ϕ )) N<br />

<strong>an</strong>d H respectively such that :<br />

N <strong>an</strong>d ( J( u )) N . Then, there exists u <strong>an</strong>d ϕ <strong>in</strong> V<br />

∀v ∈ V, ∀t ∈ [0 , T ];<br />

t t <br />

( u( t) u ) v dx + ∇( u) ∇v dx ds + <br />

∞<br />

( u u ) v d ds =<br />

<br />

0<br />

N<br />

<br />

In or<strong>de</strong>r to prove that ( u, ϕ) is a solution, we have to show that 0 ϕ 1 i.e. 0 ϕ( t)<br />

1.<br />

2<br />

So, consi<strong>de</strong>r<strong>in</strong>g a C ( R, R)<br />

function , the Itô formula gives :<br />

Remark 2 Th<strong>an</strong>ks to lemma (6), if u ∈ V then G( u) ∈ L ( , H ( Q))<br />

.<br />

<br />

0<br />

<br />

0<br />

0 0 <br />

<br />

t t<br />

ϕ( t) ϕ = (1 ϕ) J( u) ds + b( u, ϕ) dw( s ) .<br />

0<br />

t t<br />

′<br />

2<br />

[ ( ( ))] [ ( 0)]<br />

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

E ϕ t E ϕ E ϕ J u ϕ ds E b ( u, ϕ) ”( ϕ) ds.<br />

2<br />

<br />

0<br />

0<br />

0<br />

<br />

0 0<br />

0 0<br />

0<br />

0<br />

N N<br />

( ϕ( t) ϕ ) v dx<br />

Thus, if we suppose that ( x) = 0 for <strong>an</strong>y x <strong>in</strong> [0, 1] , E[ ( ϕ( t))] = 0 for <strong>an</strong>y t.<br />

Then, for <strong>an</strong>y t <strong>in</strong> [0 , T ] , 0 ϕ( t ) 1 , <strong>an</strong>d ( u, ϕ)<br />

is a solution.<br />

20<br />

+ <br />

0<br />

<br />

2 1<br />

(18)


6 Uniqueness<br />

Assume that ( u, ϕ) <strong>an</strong>d ( u, ϕ<br />

) <strong>in</strong> V H are two solutions. Then,<br />

<br />

t <br />

t <br />

( u u )( t) v dx + ∇[ ( u) ( u )] ∇v dx ds + ( u u ) v d ds =<br />

0 0 <br />

0 <br />

<br />

0 0<br />

0 <br />

0 <br />

0 <br />

0 <br />

0<br />

0 <br />

0 <br />

0 <br />

2 2<br />

V<br />

0 <br />

0 0 <br />

0<br />

<br />

0<br />

( ϕ ϕ )( t) v dx;<br />

t t<br />

ϕ( t) ϕ ( t) = [(1 ϕ) J( u) (1 ϕ ) J( u )] ds + [ b( u, ϕ) b( u, ϕ )] dw( s ) .<br />

S<strong>in</strong>ce 0 ϕ, ϕ 1,<br />

∀v ∈ V, ∀t ∈ [0 , T ];<br />

By apply<strong>in</strong>g the Itô formula (the stochastic energy equality Grecksch et al [8]), it follows that,<br />

<strong>an</strong>d<br />

Thus,<br />

t <br />

2<br />

E|| ( u u )( t) || + 2 E ∇[ ( u) ( u )] ∇[ u u ] dx ds+<br />

t <br />

t <br />

2<br />

2<br />

2 E ( u u ) d ds = E [ b( u, ϕ) b( u, ϕ )] dx ds+<br />

t <br />

2 E [(1 ϕ) J( u) (1 ϕ ) J( u )][ u u ] dx ds.<br />

t <br />

2<br />

2<br />

E|| ( ϕ ϕ )( t) || = E [ b( u, ϕ) b( u, ϕ )] dx ds+<br />

t <br />

2 E [(1 ϕ) J( u) (1 ϕ ) J( u )][ ϕ ϕ ] dx ds.<br />

t<br />

2<br />

E|| ( u u )( t) || + 2 E || ( u u )( s) || ds + E|| ( ϕ ϕ )( t)<br />

|| <br />

t <br />

t <br />

2<br />

2 E [ b( u, ϕ) b( u, ϕ )] dx ds + 2 E [ ϕ ϕ] J( u)[ u u ] dx ds+<br />

t <br />

2 E [ J( u ) J( u)] ϕ [ u u ] dx ds+<br />

t <br />

t <br />

2 E [ ϕ ϕ] J( u)[ ϕ ϕ ] dx ds + 2 E [ J( u ) J( u)] ϕ [ ϕ ϕ ] dx ds.<br />

t<br />

2<br />

E|| ( u u )( t) || + 2 E || ( u u )( s) || ds + E|| ( ϕ ϕ )( t)<br />

|| <br />

0<br />

2 2<br />

V<br />

t t<br />

C<br />

2<br />

E|| ( u u )( s) || ds + C<br />

2<br />

E|| ( ϕ ϕ )( s) || ds.<br />

Then, the Gronwall lemma leads to the uniqueness <strong>of</strong> the solution as well as to the cont<strong>in</strong>uity <strong>of</strong><br />

the solution with respect to the <strong>in</strong>itial conditions.<br />

21


7 Numerical Simulations<br />

Let us start with a presentation <strong>of</strong> the experimental context <strong>an</strong>d <strong>of</strong> the experimental observations<br />

(cf. Dumas et al [6]) :<br />

Several emulsions have been used. For example, Octa<strong>de</strong>c<strong>an</strong>e <strong>in</strong> water, glycerol <strong>an</strong>d TWEEN 80<br />

( P = 0, 50 ); water <strong>in</strong> motor oil ( P = 0, 25 <strong>an</strong>d P = 0, 50),<br />

where P <strong>in</strong>dicates the mass fraction. The<br />

experimental cell (see gure 1) is a vertical metallic tube closed by two isolated caps. This cyl<strong>in</strong><strong>de</strong>r<br />

conta<strong>in</strong>s a cage which supports 12 thermocouples. They are located regularly <strong>in</strong> a horizontal pl<strong>an</strong>e<br />

D, at different radii, for r = 0 for the thermocouple (1) to r = 27, 5 mm for the number (12) .<br />

Then, the thermocouple number (13) is located on the outer surface <strong>an</strong>d the number (14) <strong>in</strong> the<br />

<br />

bath, where the cell is immersed. This bath is cooled between 60 c <strong>an</strong>d 40 c at a const<strong>an</strong>t rate<br />

<br />

( 5 c/h to 30 c/h).<br />

Figure 1: Experimental cell<br />

On gure 2, we c<strong>an</strong> see the experimental temperature curves, <strong>of</strong> the different thermocouples,<br />

versus time dur<strong>in</strong>g the cool<strong>in</strong>g (for water, hexa<strong>de</strong>c<strong>an</strong>e <strong>an</strong>d octa<strong>de</strong>c<strong>an</strong>e emulsions). We observe,<br />

at the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the cool<strong>in</strong>g, before <strong>an</strong>y crystallization appears, a l<strong>in</strong>ear <strong>de</strong>creas<strong>in</strong>g <strong>of</strong> the<br />

temperature only due to the heat conduction <strong>in</strong>si<strong>de</strong> the emulsion (the temperature for r = 0 be<strong>in</strong>g<br />

always the highest). <strong>Crystallization</strong>s are <strong>de</strong>tected when the curves <strong>de</strong>viate from l<strong>in</strong>earity.<br />

22


As expected, the droplets near the <strong>in</strong>ner si<strong>de</strong> <strong>of</strong> the tube rst crystallize; but a part <strong>of</strong> the<br />

released energy heats up the axis region <strong>of</strong> the cyl<strong>in</strong><strong>de</strong>r, <strong>de</strong>lay<strong>in</strong>g the crystallizations <strong>of</strong> the droplets.<br />

Generally, the result is that the temperature near the axis region is practically const<strong>an</strong>t over a<br />

lapse <strong>of</strong> time ( <strong>de</strong>pends on P <strong>an</strong>d ).<br />

For the numerical simulations, we consi<strong>de</strong>r a l<strong>in</strong>ear heat equation <strong>an</strong>d a space nite difference<br />

method.<br />

u <strong>an</strong>d ϕ be<strong>in</strong>g given, we calculate the follow<strong>in</strong>g sequences ( u ) ( ϕ ) :<br />

0 0<br />

Figure 2: Experimental curves<br />

u +1 u a +1 ϕ +1 ϕ<br />

u =<br />

t 2 t<br />

n+1 n n+1 n<br />

a ∂u + u u + u ∞,n+1<br />

= ( u ) ,<br />

2 ∂<br />

2<br />

ϕ ϕ = tJ( u )(1 ϕ ) + coef.b( u , ϕ )( w w )<br />

n+1 n n n n n n+1 n<br />

12<br />

N (0, 1) U 6,<br />

i=1<br />

n n<br />

n n n n n n<br />

i<br />

a<br />

+ u<br />

,<br />

2<br />

where we use the follow<strong>in</strong>g estimation for the N (0, 1) Gaussi<strong>an</strong> law :<br />

U i be<strong>in</strong>g uniform laws on [0, 1] .<br />

One c<strong>an</strong> see <strong>in</strong> gures 3, 4 <strong>an</strong>d 5 some numerical curves. coef represents a multiplicative<br />

coefficient put <strong>in</strong> front <strong>of</strong> b <strong>in</strong> or<strong>de</strong>r to control the effect <strong>of</strong> the stochastic perturbation. If coef = 0,<br />

the mo<strong>de</strong>l is a <strong>de</strong>term<strong>in</strong>istic one (as if b = 0 cf.<br />

<strong>de</strong>pen<strong>de</strong>nce <strong>of</strong> with respect to coef.<br />

Vallet [16]) <strong>an</strong>d one c<strong>an</strong> see <strong>in</strong> gure 6 the<br />

23


Figure 3: Determ<strong>in</strong>istic case<br />

Figure 4: <strong>Stochastic</strong> case (small coef)<br />

24


Figure 5: <strong>Stochastic</strong> case (greater coef)<br />

Figure 6: Depend<strong>an</strong>ce <strong>of</strong> <br />

We see that the shapes <strong>of</strong> the curves are rather i<strong>de</strong>ntic to the experimental ones.<br />

On gures 7, 8 <strong>an</strong>d 9, we give, for each value <strong>of</strong> r, the function ϕ versus time. We c<strong>an</strong> see<br />

that the proportion <strong>of</strong> crystallized droplets ϕ is very small when the temperature is stabilized. ϕ<br />

<strong>in</strong>creases sharply only at the end <strong>of</strong> the temperature plateau <strong>an</strong>d the temperature <strong>de</strong>creases aga<strong>in</strong><br />

only just after the crystallizations <strong>of</strong> all the droplets.<br />

25


Figure 7: Determ<strong>in</strong>istic case<br />

Figure 8: <strong>Stochastic</strong> case (small coef)<br />

26


Figure 9: <strong>Stochastic</strong> case (greater coef)<br />

It is shown <strong>in</strong> Vallet [16] that <strong>in</strong> the <strong>de</strong>term<strong>in</strong>istic case, the mo<strong>de</strong>l gives non<strong>in</strong>creas<strong>in</strong>g temperature<br />

curves only. With coef = 0,<br />

the numerical curves give globally the same behavior. This<br />

behavior is generally the one observed <strong>in</strong> the experimentation. But, <strong>in</strong> one very particular case,<br />

the temperature may rise, <strong>in</strong>stead <strong>of</strong> the plateau (see gure 10). This ris<strong>in</strong>g occurs only with water<br />

emulsions <strong>in</strong> motor oil <strong>an</strong>d when P = 0, 5.<br />

Figure 10: Water emulsion with P = 0. 5<br />

It is known that a frozen water droplet takes a place more import<strong>an</strong>t th<strong>an</strong> a liquid one. Moreover,<br />

dur<strong>in</strong>g the crystallization, <strong>de</strong>ndrites appear. If a liquid droplet, with temperature un<strong>de</strong>r TF<br />

is touched, it crystallizes <strong>in</strong>st<strong>an</strong>t<strong>an</strong>eously. This crystallization excess <strong>in</strong>duce some more heat <strong>in</strong> the<br />

cyl<strong>in</strong><strong>de</strong>r, from which temperature rises.<br />

As, <strong>in</strong> our mo<strong>de</strong>l, wn+1 wn<br />

is with me<strong>an</strong> 0,<br />

that is, <strong>in</strong> me<strong>an</strong>, there are as m<strong>an</strong>y crystallizations<br />

as <strong>in</strong> the <strong>de</strong>term<strong>in</strong>istic case. In or<strong>de</strong>r to take <strong>in</strong>to account the above remark, we propose two<br />

numerical propositions :<br />

27


the rst one is to use | w w | <strong>in</strong>stead <strong>of</strong> w w <strong>in</strong> the comput<strong>in</strong>g (see gure 11),<br />

n+1 n n+1 n<br />

Figure 11: With | w w |<br />

n+1 n<br />

the second one is to consi<strong>de</strong>r that wn+1 wn<br />

has a positive me<strong>an</strong> (<strong>in</strong>stead <strong>of</strong> 0,<br />

see gure 12).<br />

That is, to <strong>in</strong>troduce a <strong>de</strong>rive vector to the Wiener process i.e. to consi<strong>de</strong>r the equation<br />

<strong>in</strong>stead <strong>of</strong><br />

∂ϕ<br />

∂t<br />

∂w<br />

= (1 ϕ) J( u) + b( u, ϕ)( cte + ) , <strong>in</strong> <br />

∂t<br />

∂ϕ<br />

∂t<br />

= (1 ϕ) J( u) + b( u, ϕ) <br />

∂w<br />

, <strong>in</strong> .<br />

∂t<br />

Figure 12: With positive me<strong>an</strong><br />

The shapes <strong>of</strong> the temperature curves are the same as the experimental one (cf. gure 10).<br />

But, we do not prove <strong>an</strong>y mathematical results for this stochastic context.<br />

An other way to take <strong>in</strong>to account this phenomenon is to modify the <strong>de</strong>term<strong>in</strong>istic mo<strong>de</strong>l by<br />

<strong>in</strong>troduc<strong>in</strong>g some convection terms, <strong>in</strong> the above equation, for example. To our knowledge, such a<br />

mo<strong>de</strong>l does not actually exist.<br />

28


References<br />

[1] D. Ba<strong>in</strong>ov, P. Simeonov, Integral Inequalities <strong>an</strong>d Applications,<br />

Mathematics <strong>an</strong>d its<br />

Applications, Vol. 57 , Kluwer Aca<strong>de</strong>mic Publishers (1992) .<br />

[2] A. Bensouss<strong>an</strong>, Some Existence results for <strong>Stochastic</strong> Partial Differential Equations,<br />

<strong>in</strong> <strong>Stochastic</strong><br />

Partial Differential Equations <strong>an</strong>s Applications, G. Da Prato <strong>an</strong>d L. Tubaro Eds,<br />

Pitm<strong>an</strong> Res. Notes Math. Ser., 268,<br />

(1992) , pp.37-53.<br />

[3] A. Bensouss<strong>an</strong>, R. Temam, Equations Stochastiques du Type Navier-Stokes,<br />

J. <strong>of</strong> Func. Anal.<br />

13,<br />

(1973) , pp.195-222.<br />

[4] H. Breckner, W. Grecksch, Approximation <strong>of</strong> solutions <strong>of</strong> stochastic evolution equations by<br />

Rothes method,<br />

report <strong>of</strong> the Institute <strong>of</strong> Optimization <strong>an</strong>d <strong>Stochastic</strong>, Halle University,<br />

n 13,<br />

(1997) .<br />

[5] M. Capiński, S. Peszat, Local Existence <strong>an</strong>d Uniqueness <strong>of</strong> strong solutions to 3-D stochastic<br />

Navier-Stokes Equations,<br />

NoDEA 4,<br />

(1997) , pp.185-200.<br />

[6] J.P. Dumas, M. Strub, M. Krichi, Etu<strong>de</strong> thermique <strong>de</strong>s ch<strong>an</strong>gements <strong>de</strong> phases d<strong>an</strong>s une<br />

émulsion, Rev. Gén. Therm. Fr., 341,<br />

Mai (1990) , p.267-273.<br />

[7] A. Friedm<strong>an</strong>, <strong>Stochastic</strong> Differential Equations <strong>an</strong>d Applications,<br />

Vol.1 & Vol.2, Aca<strong>de</strong>mic<br />

Press, New York, (1975) .<br />

[8] W. Grecksch, C. Tudor, <strong>Stochastic</strong> evolution equations,<br />

Aka<strong>de</strong>mie Verlag, Berl<strong>in</strong>, (1995) .<br />

[9] H. Hol<strong>de</strong>n, B. Øksendal, J. Uboe, T. Zh<strong>an</strong>g, <strong>Stochastic</strong> Partial Differential Equations :<br />

A <strong>Mo<strong>de</strong>l</strong><strong>in</strong>g, White Noise, Functional Approach,<br />

Birkhäuser, Berl<strong>in</strong>, (1996) .<br />

[10] H. Hol<strong>de</strong>n, N.H. Risebro, Conservation Laws with a r<strong>an</strong>dom Source,<br />

Applied Mathematics<br />

<strong>an</strong>d Optimization, Vol 36 - No 2 (1997) pp. 229-239.<br />

[11] N.V. Krylov, On Lp<br />

theory <strong>of</strong> stochastic partial differential equations <strong>in</strong> the whole space,<br />

SIAM<br />

<br />

J. Math. Anal., Vol. 27,<br />

n 2 , (1996) , pp.313-340.<br />

[12] H.H. Kuo, J. Xiong, <strong>Stochastic</strong> Differential Equations <strong>in</strong> White Noise Space, Inn. Dimens.<br />

Anal. Qu<strong>an</strong>tum Probab. Relat. Top. 1,<br />

No.4, (1998) , pp.611-632.<br />

[13] P.L. Lions, P.E. Soug<strong>an</strong>idis, Fully Nonl<strong>in</strong>ear Stochactic Partial Differential Equations,<br />

C. R.<br />

Acad. Sci. Paris, t. 326,<br />

Série 1, (1998) , pp.1085-1092.<br />

[14] B. Øksendal, <strong>Stochastic</strong> differential equations,<br />

Spr<strong>in</strong>ger-Verlag, Berl<strong>in</strong>, (1989) .<br />

[15] 1) E. Pardoux, Equations aux dérivées partielles stochastiques non l<strong>in</strong>éaires<br />

monotones, Etu<strong>de</strong> <strong>de</strong> solutions fortes <strong>de</strong> type Ito,<br />

Thèse, Université <strong>de</strong> Paris-Nord,<br />

Nov. (1975) .<br />

2) <strong>Stochastic</strong> partial differential equations <strong>an</strong>d lter<strong>in</strong>g <strong>of</strong> diffusion processes,<br />

<strong>Stochastic</strong>s, Vol.<br />

3,<br />

(1979) , pp.127-167.<br />

[16] 1) G. Vallet, Analyse mathématique <strong>de</strong>s tr<strong>an</strong>sferts thermiques d<strong>an</strong>s <strong>de</strong>s systèmes dispersés<br />

2<br />

<br />

subiss<strong>an</strong>t <strong>de</strong>s tr<strong>an</strong>sformations <strong>de</strong> phases, M AN, Vol. 27,<br />

n 7,<br />

(1993) , pp.895-923.<br />

2) G. Vallet, Mathematical mo<strong>de</strong>lization <strong>of</strong> the phase ch<strong>an</strong>ge heat tr<strong>an</strong>sfer <strong>in</strong> <strong>an</strong> emulsion,<br />

<strong>in</strong><br />

Progress <strong>in</strong> partial differential equations : the Metz surveys 2,<br />

M. Chipot (editor), Pitm<strong>an</strong><br />

<br />

Research Notes <strong>in</strong> Mathematics, n 296,<br />

(1993) , pp.55-66.<br />

[17] J.B. Walsh, An <strong>in</strong>troduction to stochastic partial differential equations,<br />

In Ecole dété <strong>de</strong><br />

Probabilité <strong>de</strong> Sa<strong>in</strong>t-Flour XIV, P.L. Hennequ<strong>in</strong> editor, Lecture Notes <strong>in</strong> Mathematics 1180,<br />

Spr<strong>in</strong>ger Verlag, Berl<strong>in</strong>, (1986) .<br />

29

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!