getdocb7fe. 230KB Jun 04 2011 12:04:53 AM

J
i
on
Electr

o

u

a
rn l

o

f
P

c

r

ob
abil
ity

Vol. 9 (2004), Paper no. 23, pages 674-709.
Journal URL
http://www.math.washington.edu/∼ejpecp/

A General Analytical Result for Non-linear SPDE’s and Applications

Laurent Denis
Universit´e du Maine, Facult´e des Sciences,
D´epartement de Math´ematiques,
Laboratoire de Statistique et Processus,
Avenue Olivier Messiaen,
72085 Le Mans CEDEX 9, France
e-mail: ldenis@univ-lemans.fr
and
L. Stoica
University of Bucharest, Faculty of Mathematics,

Str. Academiei 14, Bucharest, RO -70109, Romania
e-mail: lstoica@pompeiu.imar.ro
Abstract: Using analytical methods, we prove existence uniqueness and
estimates for s.p.d.e. of the type
dut + Aut dt + f (t, ut ) dt + Rg (t, ut ) dt = h (t, x, ut ) dBt ,
where A is a linear non-negative self-adjoint (unbounded) operator, f is
a nonlinear function which depends on u and its derivatives controlled by
√
Au, Rg corresponds to a nonlinearity involving u and its derivatives of
the same order as Au but of smaller magnitude, and the right term contains
a noise involving a d-dimensional Brownian motion multiplied by a nonlinear function. We give a neat condition concerning the magnitude of these
nonlinear perturbations. We also mention a few examples and, in the case
of a diffusion generator, we give a double stochastic interpretation.

674

MSC: 60H15; 60G46; 35R60

Keywords: Stochastic Partial Differential Equations, Dirichlet space, Bilaplacian, Doubly stochastic representation.
Submitted to EJP on December 18, 2003. Final version accepted on

August 30, 2004.

675

1

Introduction

The starting point of this work is the following stochastic partial differential
equation in divergence form:
dut −

N
X

¢
¡
1
∂i ai,j ∂j ut dt + fe(t, x, ut , √ ∇ut σ)dt
2

i,j=1

(1)

µ
¶
d
X
1
1
ij
e
+
∂i a gej (t, x, ut , √ ∇ut σ) dt =
hi (t, ut , √ ∇ut σ)dBti ,
2
2
i=1
i=1
N

X

where a = 21 σσ ∗ is a symmetric non-negative definite matrix valued function
defined from RN into RN ×N , B = (B 1 , ..., B d ) is a d-dimensional Brownian
motion and fe, ge, e
h are only assumed to be Lipschitz continuous. We prove
existence, uniqueness and estimate the solution (with initial condition) under very weak conditions on the entries of a. They may be discontinuous
and even degenerated.
Due to the analytical methods we use, we see that in fact one can solve
a s.p.d.e. in a much more general context, namely
dut + Aut dt + f (t, ut ) dt + Rg (t, ut ) dt =

d
X

hi (t, ut ) dBti ,

(2)

i=1

where A is a self-adjoint non-negative operator defined on a Hilbert space,
f, g, h are adapted random functions defined on [0, +∞[×F where F is the
domain of A1/2 and Rg corresponds to a term of the same degree as A, but
of lower magnitude in some sense. The functional approach makes things
clearer and it also has the advantage of including as particular cases other
examples, like for instance the bi-laplacian in RN or the Ornstein-Uhlenbeck
operator in infinite dimension.
The plan of the paper is as follows. In Section 2 we set notations, hypotheses and we announce the main results. In Section 3 we treat the relevant deterministic equations and in Section 4 we prove existence, uniqueness
and estimates in terms of the data of the solution of the equation (2). Section 5 is devoted to some examples. Finally, in the last section we give the
probabilistic interpretation to the solutions of the above equation (1) under
the supplementary condition of uniform ellipticity.

2
2.1

Notations, hypotheses and main results
Hilbert framework and hypotheses

Let (H, (·, ·)) be a separable Hilbert space whose norm is simply denoted

by k·k and (Pt ) a symmetric strongly continuous semigroup on H. Let
676

(A, D (A)) be the infinitesimal generator of the semigroup. It is a nonnegative, densely defined, self-adjoint operator. The operator A1/2 will play
a special role in what follows, so that we introduce the notation F =
°
°2
¡
¢
D A1/2 , E (u, v) = (A1/2 u, A1/2 v) and E (u) = °A1/2 u° for u, v ∈ F.
The usual norm on F is given by kuk2F = E (u) + kuk2 , and we recall
that F is a Hilbert space with respect to this norm. The space D (A)
is also considered as a Hilbert space endowed with the norm defined by
kuk2D(A) = kAuk2 + kuk2 .
The aim of this paper is to study the solutions of the nonlinear stochastic partial differential equation of parabolic type (2), where (Bt ) represents
an adapted d−dimensional Brownian motion defined on a standard filtered
probability space (Ω, F, Ft , P ) . The letter R denotes a bounded linear operator R : K → F ′ defined on some Banach space K, with values in the dual
F ′ of F, and it is assumed to map a dense subspace J of K into H. We shall
adopt the usual convention that F ֒→ H ֒→ F ′ , so that we may write this
condition in the form RJ ⊂ H. The letters f, g, h denote time dependent

random functions which depend on the unknown solution u in a nonlinear
manner. They are defined as follows:
f

: R+ × Ω × F → H ,

g : R+ × Ω × F → K,

h : R+ × Ω × F → H d ,
where H d = H × · · · × H denotes the cartesian product of d copies of H.
These functions are assumed to be predictable and to satisfy the following
Lipschitz conditions
kf (t, ω, u) − f (t, ω, v)k ≤ C ku − vkF ,

kg (t, ω, u) − g (t, ω, v)kK

(H1)

≤ C ku − vk + αE (u − v)1/2 ,

kh (t, ω, u) − h (t, ω, v)kd ≤ C ku − vk + βE (u − v)1/2 ,

for u, v ∈ F, t ≥ 0 and almost all ω ∈ Ω, where C, α, β are positive constants.
The space H d is equipped with the product norm:
q
∀x = (x1 , ..., xd ) ∈ H d , kxkd = kx1 k2 + ... + kxd k2 .

The constants α and β should satisfy the following condition:
2α kRk + β 2 < 2,

(H2)

where kRk denotes the norm of the linear operator R. This means that
the size of the second degree perturbation introduced by R and the first degree perturbation associated with the Brownian motion, should be relatively
small. Indeed, in the case of the equation (1), the term Rg (u) contains second order derivatives of the unknown function u (see Subsection 5.1). But,
677

if the constant α is small, then the linear term Au controls the nonlinear
term Rg (u) . E.g. if one assumes that h ≡ 0, then the equation (1) becomes
a deterministic parabolic equation and the condition (H1), with β = 0, becomes the usual condition which ensures that the problem is well posed. On

the other hand, the fact that the constant β should be small is well-known
in the context of doubly stochastic BSDE (see [15]). In fact, E. Pardoux
proved in [14] that a condition on β is necessary for (1) to be a well-posed
stochastic parabolic equation.
Moreover it is assumed that
f (·, ·, 0) ∈ L2loc (R+ × Ω; H) ,
g (·, ·, 0) ∈

(H3)

L2loc (R+

h (·, ·, 0) ∈ L2loc

³

× Ω; K) ,
´
R+ × Ω; H d ,

where L2loc (R+ × Ω; H) denotes the set of H−valued functions u such that
for all T ≥ 0, u ∈ L2 ([0,T ] × Ω; H) .
Throughout this paper, we will assume that the hypotheses (H1), (H2),
(H3) are satisfied.

2.2

Notations

If (a, b) is an open interval we will use the notation L2loc ((a, b) ; H) to denote
the space of all functions u : (a, b) → H such that
Z s
kut k2 dt < ∞
r

for any closed interval [r, s] ⊂ (a, b) . The space L2loc (R+ ; H) will consist of
all measurable maps u : R+ → H such that all the following quantities are
finite
Z
T

0

kut k2 dt < ∞ , T > 0.

Similar spaces will be considered with H replaced by F and D (A) . We
denote by Fe the subspace of L2loc (R+ ; F ) consisting of all maps u that are
H -continuous and endow it with the seminorms
!1/2
Ã
Z
kukT :=

sup kut k2 + 2

0≤t≤T

T

0

E (ut ) dt

, T > 0.

The space of test functions in our study will be D = Cc∞ (R+ ) ⊗ D (A) , the
algebraic tensor product of the space of infinite differentiable functions on
[0, ∞) vanishing outside a finite interval with the domain of the infinitesimal
operator A. Since Cc∞ (R+ ) is dense in L2loc (R+ ) and D (A) is dense in F and
in H, it follows that D is dense both in L2loc (R+ ; F ) and in L2loc (R+ ; H) .
678

We now introduce some spaces of processes defined on our given probability space. We denote by P (H) the space of all predictable processes
u : R+ × Ω → H for which the quantities
Z T
kut k2 dt , T > 0,
E
0

are all finite. Factorizing the space P (H) by its subspace consisting of processes for which the above quantities vanish we get a subspace of L2loc (R+ × Ω; H) .
Similarly, P (F ) denotes the space of all predictable processes u : R+ × Ω →
F for which the quantities
Z T³
´
E
kut k2 + 2E (ut ) dt , T > 0,
0

are finite. By factorization with respect to the subspace of nulls of the above
quantities, P (F ) becomes a subspace of L2loc (R+ × Ω; F ) . Finally we denote
e (F ) the sub-vector space of processes u ∈ P (F ) with H -continuous
by P
trajectories, that is, u (·, ω) ∈ Fe almost surely. This space will be endowed
with the seminorms
³
´1/2
E kuk2T
, T > 0.

e (F ) is the basic space in which we are going to look for solutions.
The space P
The conditions (H1),(H3) imposed to f, g, h imply that if u ∈ P (F ) ,
then f (·, u· ) , hi (·, u· ) ∈ P (H) for all i ∈ {1, ..., d} and g (·, u· ) is predictable
and belongs to the space L2loc (R+ × Ω; K) . In particular we may define the
solution of the stochastic partial differential equation as follows.
Definition 1 We say that u ∈ P (F ) is a weak solution of the equation
(2) with initial condition u0 = x ∈ L2 (Ω, F0 , P ; H) if the following relation
holds almost surely, for each ϕ ∈ D,
Z ∞
[(us , ∂s ϕ) − E (us , ϕs ) − (f (s, us ) , ϕs ) − (Rg (s, us ) , ϕs )F ′ ,F ]ds+
0

+

d Z
X
i=1

∞
0

(hi (s, us ) , ϕs ) dBsi + (x, ϕ0 ) = 0.

In fact, we will work with the notion of mild solution, which will be
well-defined in section 4 (see Proposition 6).
Definition 2 We call u ∈ P (F ) a mild solution of the equation (2) with
initial condition u0 = x, if the following equality is verified almost surely,
for each t ≥ 0,
Z t
Z t
ut = P t x −
Pt−s f (s, us ) ds −
Pt−s Rg (s, us ) ds
0

+

d Z
X
i=1

679

0

0

t

Pt−s hi (s, us ) dBsi .

2.3

Main results

The main results of this work are the following.
Theorem 2.1 The notions of mild and weak solutions coincide, moreover
there exists a unique solution to equation (2) with initial condition u(0) =
e (F ) and satisfies the following estimate
x ∈ L2 (Ω, F0 , P ; H). It belongs to P
!
Ã
Z
E

µ
Z
≤ cE kxk2 +

sup kut k2 +

0≤t≤T

T

0

³

T

0

E (ut ) dt

´ ¶
kh (t, 0)k2d + kf (t, 0)k2 + kg (t, 0)k2K dt .

See Theorems 8 and 9 of Section 4. All along the proof, we make a special
effort to get optimal conditions on the constants of Lipschitz continuity of
the nonlinear terms. Our best result is that condition (H2) suffices to ensure
the validity of the above theorem. Results of this kind for diffusion equations
and even more precise information have already been obtained by Krylov
[11], Mikulevicius and Rozovskii [13], and G¨ongy and Rovira [10]. However
their methods do not seem to cover such general situations as our paper
does (measurable coefficients, without strict ellipticity condition). One also
has to mention the early work of Chojnowska-Michalik [5], which also treats
SPDE ’s in the general framework of Hilbert spaces, as we do. Our paper
uses more or less the same type of arguments (e.g. contraction principle and
semigroup properties), but we deal with two spaces: H and F. Therefore
we produce more precise estimates so that to take into consideration the
Lipschitz continuity of the nonlinear terms with respect to both the norms
of H and F (as it is stated in (H1)). And another thing that our paper
brings is the neat condition (H2).
In Corollary 12 we give a probabilistic interpretation of
¡ ijthe
¢ solution in
the case of the concrete equation (1), where the coefficients a are assumed
to be measurable and uniformly elliptic. That is, we show that the solution
admits a doubly stochastic representation, in the sense that u(t, XT −t ) is
expressed as a sum of stochastic integrals with respect to B and X, where
X is the process generated by A. The doubly stochastic representation was
first proved by E. Pardoux and S. Peng in [15]. They proved it directly under
different conditions and then used it to investigate the spde itself. Our result
follows from an approximation result stated in Lemma 10, which shows that
the solution of the linear stochastic equation is pathwise approximated by
the solutions of certain deterministic linear equations. In fact these deterministic equations are obtained from the stochastic equation by replacing
the Brownian with its usual polygonal approximation. The lemma holds in
the general Hilbert space framework and may have other applications.

680

3

Preliminaries on the deterministic linear equation

In this section we introduce some notations and gather some facts concerning
the deterministic linear equation, facts that are proved by standard methods
from the theory of semigroups of operators (see [8]). For example it is wellknow that if x ∈ H, then for all t > 0, Pt x ∈ D(A), and the map t → Pt x is
H−continuous on [0, +∞[ and H-differentiable on ]0, +∞[ and its derivative
is ∂t Pt x = −APt x. (The semigroup is even analytic, but we will not use this
fact.) We also have that ∀u ∈ D(A), ∀v ∈ F, E(u, v) = (Au, v), which
is useful when working with the notion of weak solution. We use several
notions of weak solutions taken from [12].
The next lemma allows one to extend the space of test functions ϕ permitted in the weak relations related to the solutions of evolution equations
and their time derivatives.
Lemma 1 If w : R+ → F has compact support and is F -differentiable with
F -continuous derivative ∂t w, then there exists a sequence (w n ) ⊂ D such
that
lim sup kwtn − wt kF = 0 ,
n

t

lim sup k∂t wn − ∂t wkF = 0 .
n

t

A similar approximation holds for H valued functions.
Proof. Assume that T > 0 is such that wt = 0, for each t > T. As
Cc∞ (R+ ) is dense in Cc (R+ ) and D(A) is dense in F , it is clear that there
exists a sequence (xn ) in D which converges uniformly to ∂w in F on [0, T ].
We put
Z
∀t ≥ 0, wtn = w(0) +

t

0

xns ds,

which clearly is an element in D. One verifies that
Z t
n
kxns − ∂s ws kF ds ≤ T sup kxnt − ∂t wt kF .
∀t ∈ [0, T ], kwt − wt kF ≤
0

t∈[0,T ]

The proof is complete.✷
Definition 3 A function u ∈ L2loc ((a, b) ; H) is said to admit a weak derivative on the open interval (a, b) if there exists a function ∂t u ∈ L2loc ((a, b) ; H)
(that we shall call the weak derivative of u) which satisfies the relation
Z b
Z b
(∂t ut , ϕ) dt,
(ut , ∂t ϕ) dt = −
a

a

for all test functions ϕ ∈ D such that ϕt = 0 for t outside a compact interval
contained in (a, b) .
681

Definition 4 Given x ∈ H and w ∈ L2loc (R+ ; H) , an s -weak solution of
the linear equation
∂t u + Au = w ,
(3)
with the initial condition u (0) = x, is a function u ∈ L2loc ((0, ∞) ; D (A)) ∩
C (R+ , H) which admits a weak derivative (in H) on (0, ∞) , such that the
relation
∂t ut + Aut = wt
is satisfied for almost all t > 0, and such that u0 = x.
The s in this name stands for the word “separate” and refers to the fact
that in this definition the operator ∂t + A splits in the sum of two separate
parts: ∂t and A. This is in contrast to another notion of weak solution we
will introduce later on in the Definition 4 and to its stochastic version from
the Definition 1.
The next lemma is the basic uniqueness result for the solutions considered in this paper. In particular it ensures uniqueness of s -weak solutions
of the above equation.
Lemma 2 If u ∈ L2loc (R+ ; F ) satisfies the relation
Z ∞
[(ut , ∂ϕt ) − E (ut , ϕt )]dt = 0,
0

with any ϕ ∈ D, then u ≡ 0, as an element of L2loc (R+ ; F ) .
R∞
Proof. Take w ∈ D and set vt = t Ps−t ws ds. A direct verification shows
that
∂t v = Avt − wt .
Assume that T is such that wt = 0 for t ≥ T, and hence also vt = 0, for
t ≥ T. Then one has
Z ∞
Z ∞
(ut , ∂vt − Avt ) dt =
(ut , wt ) dt =
−
0
0
Z ∞
[(ut , ∂vt ) − E (ut , vt )]dt.
=
0

By the preceding lemma one gets a sequence (v n ) ⊂ D such that vn → v,
uniformly in F together with the derivatives. The last expression written
with vn ,
Z
∞

0

[(ut , ∂vtn ) − E (ut , vtn )]dt = 0,

vanishes according to the hypothesis. In the limit we have
Z ∞
(ut , wt ) dt = 0.
0

682

Since w is arbitrary, it follows that u ≡ 0.✷
The next proposition gives a formula for the solution of equation (3) and
the basic estimates in terms of the data.
Lemma 3 If x ∈ H and w ∈ L2loc (R+ ; H) are given, then
Z t
Pt−s ws ds + Pt x , t ∈ R+ ,
ut =
0

is an s -weak solution of the linear equation (3), with initial condition u 0 = x.
f and it satisfies the following relations, with any T > 0,
Moreover u ∈ F
Z T
Z T
1
1
2
(wt , ut ) dt + kxk2 ,
E (ut ) dt =
kuT k +
(i)
2
2
0
0
¶
µ
Z T
2
2
2
T
kwt k dt .
(ii)
kukT ≤ e
kxk +
0

Besides these, if x ∈ F, the solution satisfies the relation
Z T³
Z T
´
2
2
kwt k2 dt.
k∂t uk + kAut k dt = E (x) +
E (uT ) +
0

(iii)

0

Proof. Fix T > 0 and assume first
it is
R tthat w ∈ D and x ∈ D(A). Then, 3/2
clear that the maps t → Pt x, t → 0 Pt−s ws ds belong to C([0, T ], D(A ))
and are H-differentiable with continuous derivatives. From this, we deduce
that t → ut belongs to C([0, T ], D(A)), it is also H-differentiable with continuous derivative and we have
∀t > 0,

dut
= wt − Aut .
dt

(4)

In particular it is an s -weak solution of (3) and moreover one can verify
dut
that ∂t ut =
belongs to F .
dt
Next we are going to establish the relations from the statement in this
particular case. Integrating by parts, we have
Z T
Z T
2
2
(ut , ∂t ut ) dt
(∂t ut , ut ) dt = kuT k − kxk −
0

0

so

2

kuT k − kxk

2

= 2

Z

Z

T

(∂t ut , ut ) dt
0
T

(wt − Aut , ut ) dt
µZ T
¶
Z T
E(ut ) dt
= 2
(wt , ut ) dt −
= 2

0

0

683

0

which is relation (i). The relation (ii) follows from this one by using the
Schwartz inequality
Z T
Z T
[kwt k2 + kut k2 ] dt
(wt , ut ) dt ≤
2
0

0

and then applying Gronwall’s lemma.
In order to deduce the third relation one starts with the following equality
kwt k2 = k∂t uk2 + kAut k2 + 2 (∂t u, Aut ) ,
which is simply a consequence of (4). Then one integrates both sides of this
equality. Since 2 (∂t u, Aut ) = 2E (∂t u, ut ) = ∂t E (ut ) , one has
Z T
2 (∂t u, Aut ) dt = E (uT ) − E (x) ,
0

and so we get the relation (iii) .
In order to obtain the result in the general case one approximates x
and w with objects of the preceding kind. The
R t three relations of the statement pass to the limit. The function u′t = 0 Pt−s ws ds always belongs to
L2loc (R+ ; D (A)) . The function u′′t = Pt x only belongs to L2loc ((0, ∞) ; D (A))
in general. But this suffices to fulfil the requirements of the definition of an
s -weak solution. ✷
Remark 1 The s -weak solution of the equation (3) always satisfies the
relation
Z t
Z t
wr dr,
Aur dr +
ut − u s = −
s

s

for any 0 < s ≤ t. If u (0) = x belongs to F, then the equality (iii) of the
preceding proposition ensures that Aur is integrable up to 0. Therefore, in
this particular case, the relation also holds with s = 0 and one has, for each
t > 0,
Z
Z
t

t

ut = x −

wr dr.

Aur dr +

0

0

Now we are going to treat equations involving the operator R.
Definition 5 Let x ∈ H, w ′ ∈ L2loc (R+ ; H) and w ′′ ∈ L2loc (R+ ; K) be
given. A weak solution of the equation
(∂t + A) u = w ′ + Rw′′

(5)

with initial condition u (0) = x is a function u ∈ L2loc (R+ ; F ) such that the
next relation is fulfilled with any ϕ ∈ D,
Z
¡
¢ ¡
¢
(x, ϕ0 ) + [(ut , ∂t ϕ) − E (ut , ϕt ) + wt′ , ϕt + Rwt′′ , ϕt F ′ ,F ]dt = 0.
684

Clearly an s -weak solution of the equation (3) is also a weak solution of
(5) with w ′ = w and w′′ = 0. The uniqueness of the solution associated to
given data follows again by Lemma 2. The existence will also be obtained by
an explicit formula. ToR treat the term containing R we need to give a sense
t
to the formal integral 0 Pt−s Rws ds with w ∈ L2loc (R+ ; K) . This integral is
well defined if w ∈ Cc∞ (R+ ) ⊗ J, where J is the space associated to R in the
introduction. Denote by U : Cc∞ (R+ ) ⊗ J → Fe the operator defined by
(U w)t =

Z

0

t

Pt−s Rws ds , t ≥ 0.

The next lemma proves that
R t we can extend it by continuity. Later we
will use the formal expression 0 Pt−s Rws ds rather than U w.

Lemma 4 The operator U admits a uniquely determined continuous extension
U : L2loc (R+ ; K) → Fe.
If w ∈ L2loc (R+ ; K) , then u = U w is a weak solution of the equation
(∂t + A) U = w, u (0) = 0.
Moreover it satisfies the following relations,
Z T
Z T
1
2
(Rwt , ut )F ′ ,F dt,
E (ut ) dt =
kuT k +
2
0
0
¶
µZ T
2
2
2
T
kwt kK dt .
kukT ≤ 2e kRk

(i)

(ii)

0

Proof. One easily sees that if w ∈ Cc∞ (Rt ) ⊗ J, then Rw ∈ C ∞ (R+ ) ⊗ H.
By Lemma 3, we deduce that u = U w is an s -weak solution of (3) with data
Rw and initial condition x = 0. ¿From relation (i) of that lemma we get
1
kU wT k2 +
2

Z

T
0

E (U wt ) dt =

Z

T

(Rwt , U wt ) dt.

0

Since, in general, one has (x, y) ≤ kxkF ′ kykF , as soon as x ∈ H and y ∈ F,
the last integral is dominated by
Z T
³
´1/2
kwt kK kU wt k2 + E (U wt )
dt.
kRk
0

Using Schwarz’s inequality and then Gronwall’s lemma one deduces the relation (ii) from the statement. The result with general w ∈ L2loc (R+ ; K)
follows by approximation.✷
Combining the preceding two lemmas one gets the following one.
685

Lemma 5 The equation (5) admits a unique solution given by
Z t
Z t
′
Pt−s Rws′′ ds , t ∈ R+ .
Pt−s ws ds +
ut = P t x +
0

0

It belongs to Fe and satisfies the following relations
Z T³
Z T
´
¢
¡ ′ ¢ ¡
1
1
wt , ut + Rwt′′ , ut F ′ ,F dt + kxk2 ,
E (ut ) dt =
kuT k2 +
2
2
0
0
µ
¶
Z T
Z T
° ′ °2
° ′′ °2
2
2
2
T
°
°
°
°
kukT ≤ 2e
kxk +
wt dt + kRk
wt dt .
0

4

0

Stochastic parabolic equations

In what follows, we consider stochastic integrals of processes in (P (H)) d
with respect to our Brownian motion. They have to be seen as Hilbert-space
valued stochastic integrals. For this, we refer to Da Prato and Zabczyk
[7], who deal with the more general case when the Brownian motion is
also infinite dimensional. So, we know that this stochastic integral has
good properties such as continuity in time, predictability and that a suitable
version of Itˆ
o’s formula holds.
One basic ingredient in the treatment of stochastic parabolic equations
is the following result. Let us note that the assertion (v) of this proposition
was proved in [5] under more general conditions.
Proposition 6 Let x ∈ L2 (Ω, F0 , P ; H), w = (w1 , ···, wd ) ∈ (P (H))d , w′ ∈
P (H) and w ′′ ∈ P (K) . We set
d Z
X

∀t ≥ 0, ut = Pt x +

i=1

t

0

Pt−s wi,s dBsi +

Z

t
0

Pt−s ws′ ds +

Z

0

t

Pt−s Rws′′ ds.

e (F ) and for each ϕ ∈ D, it verifies the folThen u has a version in P
lowing weak sense relation almost surely,
Z ∞
¡
¢ ¡
¢
[(us , ∂s ϕ) − E (us , ϕs ) + ws′ , ϕs + Rws′′ , ϕs F ′ ,F ]ds
0

d Z
X

+

i=1

∞
0

(wi,s , ϕs ) dBsi + (x, ϕ0 ) = 0.

Moreover, for all t ≥ 0:
kut k2 + 2

Z

0

t

E (us ) ds = kxk2 + 2
686

d Z
X
i=1

0

t

(us , wi,s ) dBsi +

(i)

Z

+

0

so

t¡

¢
us , ws′ ds +

2

E kut k + 2E
+E

µZ
0

0

t¡

t

E (us ) ds

0

µZ

Z

t¡

us , Rws′′

¶

= E kxk + E

¢

0

T

ds +
F,F ′

2

us , Rws′′ F,F ′

And we have the estimate
µ
Z
E kuk2T ≤ cE kxk2 +

¢

³

¶

ds + E

µZ

µZ
t

0

Z

t

kws k2d ds,

0

0

t¡

us , ws′

kws k2d ds

¶

¢

(ii)
¶

ds +

.

(iii)

¶
° °2 ° °2 ´
kwt k2d + °wt′ ° + °wt′′ °K dt ,

(iv)

where c is a constant which only depends on
R t T.
Finally, if w ′′ = 0, the random variables 0 us ds, t > 0, take value in
D(A) and they satisfy the next relation almost surely
ut + A

Z

t

us ds = x +

0

d Z
X

t

0

i=1

wi,s dBsi +

Z

t
0

ws′ ds.

(v)

Proof. We use the sequence of operators P 1 to construct a smoothing
n
approximation
n
wi,s
= P 1 wi,s , ws′,n = P 1 ws′ , zsn = P 1 Rws′′ , xn = P 1 x.
n

n

n

n

¡
¢d
Then (wn ) is a sequence of elements in L2 ([0, t] × Ω, D(A)) ∩ (P (H))d
¡
¢d
which converges to w in L2 ([0, t] × Ω, H) , (w′,n ) is a sequence of elements
in L2 ([0, t] × Ω, D(A)) which converges to w ′ in L2 ([0, t] × Ω, H) , (z n ) is a
sequence in L2 ([0, t] × Ω, D(A)) which converges to Rw ′′ in L2 ([0, t] × Ω, F ′ ),
for all t > 0, and (xn ) converges to x in L2 (Ω; H). Define for all n ∈ N∗ ,
unt = Pt xn +

d Z
X
i=1

0

t

n
Pt−s wi,s
dBsi +

Z

0

t

Pt−s ws′,n ds +

Z

t

0

Pt−s zsn ds.

We decompose u and un as
ut = u1t + u2t + u3t ,
2,n
3,n
unt = u1,n
t + ut + ut ,

with
u1t

=

d Z
X
i=1

t
0

Pt−s wi,s dBsi ,

u2t

= Pt x +

Z

0

687

t

Pt−s ws′ ds,

u3t

=

Z

0

t

Pt−s Rws′′ ds,

and
u1,n
t

=

d Z
X

t

0

i=1

n
Pt−s wi,s
dBsi ,

u2,n
t

Z t
Z t
3,n
′,n
Pt−s zsn ds.
= Pt x + Pt−s ws ds, ut =
n

0

0

We first study the stochastic part. One easily sees that u1,n = P 1 u1 and
n
Rt
hence, it is not difficult to see that u1,n
and 0 u1,n
s ds belong to D (A) , for
t
all t ≥ 0. Then we may perform the following calculations
Z tX
Z t
Z t
d Z tZ t
d Z s
X
1,n
1,n
i
1,n
1,n
APs−r wi,r
dsdBri =
APs−r wi,r dBr ds =
Aus ds =
us ds =
A
=−

d Z t³
X
0

i=1

0

0 i=1

0

0

1,n
Pt−r wi,r

−

1,n
wi,r

un

´

i=1

dBri

=

−u1,n
t

+

d Z
X

t

0

i=1

r

0

1,n
wi,r
dBri ,

which proves that
has a version which is a H-valued semimartingale; as
a consequence, it is H-continuous. Then, by Itˆ
o’s formula, one has
Z t
Z t
°2
°
° 1,n °2
° 1,n °
1,n
°ws ° ds =
°ut ° = 2 (u1,n
s , dus ) +
d
0

−2

Z

0

t

1,n
(u1,n
s , Aus )ds + 2

0

d Z
X
i=1

t

0

1,n
i
(u1,n
s , wi,s )dBs +

Z

1,n
1,n 1,n
As for all s ≥ 0 ,(u1,n
s , Aus ) = E(us , us ), we get

0

t°

°
°ws1,n °2 ds.
d

Z t
Z t
d Z t³
°
°
´
X
° 1,n °2
¡ 1,n ¢
° 1,n °2
1,n
1,n
i
°ws ° , (6)
E us ds = 2
us , wi,s dBs +
°ut ° + 2
d
0

i=1

0

0

which implies

¯ Z
¯ d Z
°
¯
¯X t ¡
°
°
T °
¢
° 1,n °2
¯
° 1,n °2
n
i¯
w
+
u1,n
,
w
dB
sup °ut ° ≤ 2 sup ¯
°
¯
s
i,s
s
t ° dt.
¯
d
t≤T ¯
t≤T
0
0
i=1

The Burkh¨older-Davies-Gundy’s inequality ensures that there exists a
constant c1 such that
¯
¯ 
¯
¯ d Z
d Z T ³
¯X
¯
¯X t ³
´2 ¯1/2
´
¯
¯
1,n
1,n
i¯
¯¯
ds¯  ≤
u1,n
u1,n
E sup ¯
s , wi,s
s , wi,s dBs ¯ ≤ c1 E
¯
¯
¯
¯
t≤T i=1 0
i=1 0
!
Ã
° ¯1/2
° ¯Z T °
°
° 1,n °2 ¯¯
° 1,n ° ¯¯
≤
≤ c1 E sup °ut ° × ¯
°wt ° dt¯
≤ εE
688

Ã

t≤T

°
°
° 1,n °2
sup °ut °
t≤T

!

d

0

c2
+ 1E
4ε

µZ

T

0

° ¶
°
° 1,n °2
°wt ° dt ,
d

for all ε > 0. So, taking ε small enough, we get
!
Ã
µZ T °
° ¶
°2
°
° 1,n °2
° 1,n °
≤ c2 E
E sup °ut °
°wt ° dt ,
t≤T

d

0

where c2 is another constant. Moreover, by
we get
µZ T
Z T
1
1,n
E(ut )dt ≤ E
E
2
0
0

taking the expectation in (6),

° ¶
°
° 1,n °2
°wt ° dt .
d

A similar calculation can be done for the difference u1,n − u1,m , so as to
deduce a suitable convergence of u1,n to u1 and establish the estimate
¶
µZ T
° 1 °2
2
°
°
kwt kd dt ,
(7)
E( u T ) ≤ cE
0

with a constant c which only depends on T . Now, one sees that unt = P 1 ut ,
n
so as previously we may perform the following calculation
unt +

Z

t

0

Auns ds = xn +

d Z
X

t

0

i=1

Z

n
wi,s
dBsi +

t

ws′,n ds +

0

Z

t
0

zsn ds ,

(8)

which shows that unt is a H-valued semimartingale. By applying Itˆ
o’s formula, one obtains
kunt k2
+

+2
Z

0

Z

t¡

t
0

E

(uns ) ds

uns , ws′,n

¢

n 2

= kx k + 2

ds +

Z

0

t

d Z t
X
¡
i=1

(uns , zsn ) ds

+

0

Z

0

¢ i
n
uns , wi,s
dBs +

t

kwsn k2d ds .

Thanks to Lemmas 3, 4 and the previous estimate, we know that uns →
us in the space L2 ([0, t] × Ω; F ), and hence one deduces that (uns , zsn ) →
(us , Rws′′ )F,F ′ in L1 ([0, t] × Ω; R). Therefore one easily sees that the above
relation passes to the limit and one gets the relation (ii) of our proposition.
The relation (iii) is obtained by taking the expectation.
In order to deduce the relation (i), one multiplies (8) with a test function
ϕ ∈ D and applies Itˆ
o’s formula over the interval [0, t] with t such that
ϕ(t) = 0,
Z t
¡
¢
[(uns , ∂s ϕ) − (Auns , ϕs ) + ws′,n , ϕs + (zsn , ϕs )F ′ ,F ]ds
0

+

d Z t
X
¡
i=1

0

¢
n
wi,s
, ϕs dBsi + (xn , ϕ0 ) = 0,
689

letting n tend to ∞, we get the relation (i).
To obtain the relation (iv) of the statement, we use (7) and Lemmas 3
and 4 which ensure that, almost surely
¶
µ
Z T
Z T
° ′′ °2
° °
° 2 °2
° ′ °2
°wt ° dt.
°u ° ≤ c kxk2 +
°wt ° dt , °u3 °2 ≤ c
T
K
T
0

0

This directly leads to the desired relation.
Assume now that w ′′ = 0. As for all n
unt + A

Z

0

t

uns ds = xn +

d Z
X
i=1

0

t

n
wi,s
dBsi +

Z

t
0

ws′,n ds,

we obtain the last assertion by letting n tends to infinity, using the fact that
A is a closed operator. ¤
If f, g and h do not depend on u then the equation (2) is a linear equation.
In this case the relation defining the notion of a mild solution simply becomes
a relation which defines the object that is a mild solution. Proposition 6
ensures that it is also a weak solution. In general we have the following
result.
Proposition 7 The notions of weak and mild solutions coincide. Any soe (F ) .
lution belongs to P

e (F ) and is a weak
Proof. The fact that any mild solution belongs to P
solution follows from Proposition 6.
Conversely, assume that u is a weak solution and define the process
vt = Pt x+

d Z
X
i=1

0

t

Z

Pt−s hi (s, us ) dBsi −

t

0

Pt−s f (s, us ) ds−

Z

t

Pt−s Rg (s, us ) ds.

0

We should prove that u = v. Comparing the value of the integral
Z ∞
[(us , ∂s ϕ) − E (us , ϕs )]ds,
0

obtained from the relation defining a weak solution, and the value of the
same integral with v in the place of u, given by the relation (i) of the
preceding proposition, we observe that the two are almost surely equal. So,
we deduce that
Z ∞
[(us − vs , ∂s ϕ) − E (us − vs , ϕs )]ds = 0,
0

almost surely, for each ϕ ∈ D. Since D contains a countable set which is
dense in it, we deduce that the relation holds with arbitrary ϕ ∈ D, outside
690

of a negligeable set in Ω. By Lemma 2 we deduce that u = v, almost surely,
concluding the proof. ✷
As the above propositions suggest, a first candidate for the space where
e (F ) . However the norms of this
the solutions are searched for would be P
space are too strong and would imply strong conditions on the Lipschitz
constants of the nonlinear terms to obtain a contraction inequality needed
to produce the solution. To avoid this, another space will be introduced so
that to obtain an existence result under relaxed conditions on the Lipschitz
constants (the constants α, β and kRk are particularly involved and the
condition (H2) is the optimal choice concerning them, with respect to the
b (F, T ) and
contraction argument we employ). The new space is denoted by P
consists of processes defined only on an interval [0, T ] of a certain length T >
0. More precisely, it is the space of all predictable processes u : [0, T ]×Ω → F
for which the norm
!1/2
Ã
Z
sup E kut k2 + 2E

kukE,T :=

0≤t≤T

T

E (ut ) dt

0

³
´
b (F, T ) , kk
is finite. By factorization with respect to the null sets, the pair P
E,T
becomes a Banach space.
The main existence result is the following theorem.
Theorem 8 There exists a unique solution to equation (2) with initial cone (F ) .
dition u (0) = x ∈ L2 (Ω, F0 , P ; H) and it belongs to P

Proof. The main point in the proof will be to show that the map Λ :
b (F, t) → P
b (F, t) , defined below, is a contraction with respect to the norm
P
k·kE,t for t small enough:
(Λu)t = Pt x −
+

Z

d Z
X
i=1

t

0

0

t

Pt−s f (s, us ) ds −

Z

t

Pt−s Rg (s, us ) ds
0

Pt−s hi (s, us ) dBsi .

b (F, t) ,
Proposition 6 gives, with u, v ∈ P
2

E kΛut − Λvt k + 2

−2

Z

0

Z

t

0

EE (Λus − Λvs ) ds =

t

E[(Λus − Λvs , f (s, us ) − f (s, vs ))+(Λus − Λvs , R (g (s, us ) − g (s, vs )))F,F ′ ]ds
+

Z

0

t

E kh (s, us ) − h (s, vs )k2d ds.
691

The integrands of the three terms in the right hand side are estimated, by
using the Lipschitz conditions (H1) and some elementary inequalities, as
follows
C2
kΛus − Λvs k2 + δ kus − vs k2F ,
δ

2 |(Λus − Λvs , f (s, us ) − f (s, vs ))| ≤

¯
¯
¯
¯
2 ¯(Λus − Λvs , R (g (s, us ) − g (s, vs )))F,F ′ ¯

≤ 2 kΛus − Λvs kF × kR (g (s, us ) − g (s, vs ))kF ′

≤ 2 kΛus − Λvs kF × kRk k(g (s, us ) − g (s, vs ))kK
³
´
≤ 2 kRk × kΛus − Λvs kF × C kus − vs k + αE (us − vs )1/2

1
≤ (δ + α kRk) kΛus − Λvs k2F + kRk2 C 2 kus − vs k2 + α kRk E (us − vs ) ,
δ
¶
µ
1
kus − vs k2 + β 2 (1 + δ) E (us − vs ) .
kh (s, us ) − h (s, vs )k2d ≤ C 2 1 +
δ

for any δ > 0 . Then one finds that
2

E kΛut − Λvt k + (2 − δ − α kRk) E

Z

0

t

E (Λus − Λvs ) ds ≤

¶Z t
C2
E kΛus − Λvs k2 ds
+ δ + α kRk
δ
0
µ
´¶ Z t
2 ³
C
2
2
+ δ+C +
E kus − vs k2 ds
kRk + 1
δ
0
Z t
¡
¢
E (us − vs ) ds.
+ δ + α kRk + β 2 (1 + δ) E

µ

0

Now, since by the hypothesis (H2) one has 2α kRk + β 2 < 2, we may choose
two positive constants θ 1 , θ2 , such that α kRk + β 2 < θ1 < θ2 < 2 − α kRk .
Then we take δ small enough to have δ + α kRk + β 2 (1 + δ) < θ 1 and
θ2 < 2 − δ − α kRk . Finally we conclude the preceding estimations obtaining
Z t
E (Λus − Λvs ) ds ≤
E kΛut − Λvt k2 + θ2 E
0

c3

Z

0

t

2

E kΛus − Λvs k ds + c4

Z

0

t

2

E kus − vs k ds + θ 1 E

Z

t
0

E (us − vs ) ds,

where c3 and c4 are constants which only depend on C, α, β and kRk . Gronwall’s lemma implies that, for any t,
Z t
E (Λus − Λvs ) ds ≤
E kΛut − Λvt k2 + θ2 E
0

692

ec3 t [c4

Z

0

t

E kus − vs k2 ds + θ 1 E

Z

t
0

E (us − vs ) ds].

The left hand side of this inequality is minorated by
µ
¶
Z t
θ2
2
E (Λus − Λvs ) ds
E kΛut − Λvt k + 2E
2
0
while the right hand side is dominated by
µ
¶
θ1
c3 t
e
c4 t +
ku − vk2E,t .
2
Thus we get the important inequality
(9)
kΛu − Λvk2E,t ≤ θ ku − vk2E,t ,
³
´
with θ := ec3 t c4 t + θ21 . Choosing t small enough to ensure that θ < 1 we
b (F, t) . The fixed point of this
obtain that Λ is a contraction on the space P
contraction u = Λu, is a solution of the mild equation on the interval [0, t].
Observing that mild solutions have a flow property, one extends the solution
on R+ . Finally, thanks to Proposition 6, it is clear that this solution belongs
e ).✷
to P(F
The next theorem shows that the solution satisfies estimates with respect
to the data similar to the estimate (iv) of Proposition 6.
Theorem 9 There exists a constant c which only depends on C, α, β, kRk
and T such that the following estimate holds for the solution of the equation
(2),
µ
Z T³
´ ¶
2
2
2
2
2
E kukT ≤ cE kxk +
kh (t, 0)kd + kf (t, 0)k + kg (t, 0)kK dt .
0

Proof. We shall use the inequality (9) established in the preceding proof.
We first remark that the length t and the constant θ have been chosen only
dependent on the constants C, α, β and kRk .
A solution satisfies the relation u = Λu and so we may write
√
kukE,t = kΛukE,t ≤ kΛu − Λ0kE,t + kΛ0kE,t ≤ θ kukE,t + kΛ0kE,t ,
which implies the estimate
kuk2E,t ≤

µ

1
√
1− θ

¶2

kΛ0k2E,t .

Since in general one has kvk2E,t ≤ E kvk2t , we may apply this and the estimate (iv) of Proposition 6 to dominate the last term of the inequality
obtaining
µ
¶2 µ
Z t³
´ ¶
1
√
kuk2E,t ≤ c
kh (s, 0)k2d + kf (s, 0)k2 + kg (s, 0)k2K ds .
E kxk2 +
1− θ
0
693

This estimate can be iterated over each interval of the form [nt, (n + 1) t] so
that we arrive at an estimate over an arbitrary interval [0, T ] ,
kuk2E,T

µ

2

≤ c5 E kxk +

Z

³

T

0

kh (s, 0)k2d

2

+ kf (s, 0)k +

kg (s, 0)k2K

´

¶

ds ,
(10)

with a new constant c5 , which depends on T.
Now we apply the estimate (iv) of Proposition 6 to u = Λu obtaining
E

kuk2T

µ
Z
2
≤ cE kxk +

T
0

³

kh (t, ut )k2d

2

+ kf (t, ut )k +

kg (t, ut )k2K

´

¶

dt .

Using the Lipschitz properties of f, g and h the last expression is dominated
by
µ
Z T³
´ ¶
2
2
2
2
kh (t, 0)kd + kf (t, 0)k + kg (t, 0)kK dt
c6 E kxk +
0

+c7 E

Z

RT ³

T

0

³

´
kut k2 + E (ut ) dt.

´
Finally, since we have E 0 kut k2 + E (ut ) dt ≤ (1 + T ) kuk2E,T , thanks to
inequality (10) we obtain the estimate asserted by the theorem.✷

5
5.1

Examples
The case of a diffusion generator in a finite-dimensional
state space

We can apply the above theorem to the standard finite dimensional diffusion
operator of (1):
dut −

N
X

¢
¡
1
∂i ai,j ∂j ut dt + fe(t, x, ut , √ ∇ut σ)dt
2
i,j=1

¶
µ
d
X
1
1
ij
e
hi (t, ut , √ ∇ut σ)dBti .
∂i a gej (t, x, ut , √ ∇ut σ) dt =
+
2
2
i=1
i=1
N
X

The second order operator associated with a is a semi-elliptic operator with
zero Dirichlet boundary conditions. To define it precisely we follow Section
3.1 of [9] . So the requirements of our framework are so weak that the
diffusion coefficients could be even degenerated and discontinuous. More
N
precisely,
let D be a domain
³
´ of the finite dimensional space R , and σ =
σ ij /i = 1, ..., N, j = 1, ..., k a measurable (N, k) -matrix field defined on
P
D. Then the square N -matrix field a = 21 σσ ∗ , aij = 21 kl=1 σ il σ jl , defined
694

on D is measurable, symmetric and non-negative definite.We only assume
that a is locally integrable with respect to the Lebesgue measure and that
it generates a Dirichlet space. To be more specific, let us define the bilinear
form
Z X
E (u, v) =
aij (x) ∂i u (x) ∂vj (x) dx, u, v ∈ Cc∞ (D) ,
D i,i

R
denote by (u, v) = D uv the usual scalar product in L2 (D) and set E1 (u, v) =
E (u, v)+(u, v) . What we assume is that (Cc∞ (D) , E1 ) is closable in the sense
of [9] (see Section 3.1 ). This means that the closure of Cc∞ (D) with respect
to the norm generated by E1 can be identified with a subspace of L2 (D) .
Still following [9], we know that this is the case if a is strictly elliptic or if
∂a
for all i, j ∈ {1, ..., N } , ∂xi,ji belongs to L2loc (D). This last condition shows
that we may consider the case where a is degenerated and even the totally
degenerated one, a ≡ 0, which corresponds to the case where we deal in fact
with an (infinite dimensional) ordinary stochastic differential equation.
The basic Hilbert space H is L2 (D) and F is the closure of Cc∞ (D)
1/2
with respect to E1 . The semigroup (Pt ) is the semigroup associated to this
Dirichlet space and it corresponds to zero boundary condition. The operator
A, its infinitesimal generator,
can be formally written as a divergence form
P
operator, i. e. A = − i,j ∂i (aij ∂j ).
Suppose that we have the predictable functions
fe

:

R+ × Ω × D × R × Rk → R,

ge = (e
g1 , ..., geN ) : R+ × Ω × D × R × Rk → RN ,
e
h = (e
h1 , ..., e
hd ) : R+ × Ω × D × R × Rk → Rd ,

that satisfy the Lipschitz conditions with respect
as follows,
¯
¡
¢¯
¯e
′ ′ ¯
e
¯f (t, ω, x, y, z) − f t, ω, x, y , z ¯ ≤
¯
¡
¢¯
¯ge (t, ω, x, y, z) − ge t, ω, x, y ′ , z ′ ¯
≤
a(x)
1
!
à d
X ¯¯
¡
¢¯¯2 2
′
′
hi (t, ω, x, y, z) − e
hi t, ω, x, y , z ¯
≤
¯e
i=1

to the last two variables
¯¢
¯ ¯
¡¯
C ¯y − y ′ ¯ + ¯z − z ′ ¯ ,
¯
¯
¯
¯
C ¯y − y ′ ¯ + α ¯ z − z ′ ¯ ,

¯
¯
¯
¯
C ¯y − y ′ ¯ + β ¯z − z ′ ¯ ,

where C, α, β are some constants, the seminorm |v|a(x) , for a vector v ∈ RN ,
P
ij
is expressed by |v|2a(x) =
i,j a (x) vi vj and the sign |·| stands for the
Euclidean norm in RN or Rk . Assume also that
fe(·, ·, ·, 0, 0) ∈ L2 ([0, T ] × Ω × D) , (∀) T > 0,

|e
g |a(·) (·, ·, ·, 0, 0) ∈ L2 ([0, T ] × Ω × D) , (∀) T > 0,
¯ ¯
¯e ¯
¯h¯ (·, ·, ·, 0, 0) ∈ L2 ([0, T ] × Ω × D) , (∀) T > 0.
695

(H ′ )

We denote by la2 the set of all classes of measurable functions u : D → RN
for which the seminorm

1/2
Z X
kuk 2 = 
aij (x) ui (x) uj (x) dx
La

D i,j

2
2
is finite. Then K
n is nothing but Loa , the quotient of la with respect to the
subvector space u ∈ la2 , kukL2a = 0 . The operator R is formally defined by

∀u = (u1 , ..., uN ) ∈ L2a , Ru =

X
i,j

¡
¢
∂i aij uj .

More precisely, as an element of the dual F ′ it is defined by
Z X
(Ru, v)F ′ ,F = −
aij (x)uj (x)∂i v(x) dx, ∀v ∈ Cc∞ (D) .
D i,j

For functions v ∈ F one has a well-defined generalized gradient so that
the
makes sense (∇uσ (x) is the vector in Rk of components
PN vector ∇uσ
i
i=1 (∂i u)σ j (x) , j = 1, ...k; see [2] for more details) and we may write
Z
1 XX i
σ l (x)σ jl (x)uj (x)∂i v(x) dx, ∀v ∈ F.
(Ru, v)F ′ ,F = −
2
D
i,j

l

¿From the inequality
¯
¯
¯
¯
¯(Ru, v)F ′ ,F ¯ ≤ kukL2a E 1/2 (v)

it follows that the operator norm of R satisfies kRk ≤ 1.
On the other hand, to the concrete functions f, g, h, one associates the
abstract operator-functions defined on F,
µ
¶
1
e
f : R+ × Ω × F → H, f (t, ω, u) (·) = f t, ω, ·, u (·) , √ ∇uσ (·) ,
2
µ
¶
1
g : R+ × Ω × F → K, g (t, ω, u) (·) = ge t, ω, ·, u (·) , √ ∇uσ (·) ,
2
µ
¶
1
d
e
h : R+ × Ω × F → H , h (t, ω, u) (·) = h t, ω, ·, u (·) , √ ∇uσ (·) .
2

The Lipschitz condition of g implies that for u, v ∈ F one has

kg (t, ω, u) − g (t, ω, v)kK
° µ
¶
µ
¶°
°
°
1
1
°
= °ge t, ω, ·, u (·) , √ ∇uσ (·) − ge t, ω, ·, v (·) , √ ∇vσ (·) °
° 2
2
2
La
696

°¯ µ
¶
µ
¶¯ °
°¯
¯ °
1
1
¯
°
= °¯ge t, ω, ·, u (·) , √ ∇uσ (·) − ge t, ω, ·, v (·) , √ ∇vσ (·) ¯¯ °
°
2
2
a L2
¯°
¯°
¯
°
°¯
¯°
¯°
¯ 1
°
°¯ 1
¯
¯
°
¯
°
°
≤ °C |u − v| + α ¯ √ (∇u − ∇v) σ ¯° ≤ C ku − vkL2 +α °¯ √ (∇u − ∇v) σ ¯¯°
° 2
2
2
L2
L
= C ku − vkL2 + αE 1/2 (u − v) .

This shows that the condition in (H1) is fulfilled. Similarly one can verify
the other conditions in (H1) concerning f and h. In the present case the
condition (H2) simply becomes 2α + β 2 < 2 and the condition (H3) immediately follows from the assumption (H ′ ). Thus we deduce from our main
theorem that under the condition 2α + β 2 < 2 the equation (1) has a unique
solution and it satisfies certain estimates. Under uniform ellipticity we will
prove a double stochastic formula in the last section.

5.2

An example with the bi-laplacian

We consider the following equation:
dut + ∆2 ut dt + fe(t, x, ut , ∇ut , D2 ut )dt
2

+∆e
g (t, x, ut , ∇ut , D ut )dt =
L2

¡

RN

¢

d
X
i=1

e
hi (t, x, ut , ∇ut , D2 ut )dBti ,

with initial condition u0 ∈
, where the state space is RN , ∇u =
(∂1 u, ...∂N u) is the gradient, D 2 u = (∂i ∂j u, i, j = 1, ...N ) is the Hessian of a
function u and ∆ is the Laplace operator. The functions fe, ge, e
h are defined
as follows
fe, ge : R+ × Ω × RN × R × RN × RN ×N → R,
e
h = (e
h1 , · · ·, e
hd ) : R+ × Ω × RN × R × RN × RN ×N → Rd .

They are assumed to be predictable, uniformly Lipschitz continuous with ree
e
spect
¡ to the last three
¢ variables, f (·, ·, ·, 0, 0, 0) , ge (·, ·, ·, 0, 0, 0) , hi (·, ·, ·, 0, 0, 0) ∈
2
N
L [0, T ] × Ω × R for any T > 0 , and i ∈ {1, · · ·, d} . More precisely this
continuity is expressed by
¯
¯ ¯
¯ ¯
¯¢
¡
¢¯¯
¡¯
¯e
¯f (t, ω, x, y, z, w) − fe t, ω, x, y ′ , z ′ , w′ ¯ ≤ C ¯y − y ′ ¯ + ¯z − z ′ ¯ + ¯w − w′ ¯ ,
¯ ¯
¯¢
¯
¯
¯
¡¯
¡
¢¯
¯ge (t, ω, x, y, z, w) − ge t, ω, x, y ′ , z ′ , w′ ¯ ≤ C ¯y − y ′ ¯ + ¯z − z ′ ¯ + α ¯w − w′ ¯ ,
v
u d ¯
¯2
uX ¯
¯ ¯
¯¢
¯
¯
¡¯
¯
t
hi (t, ω, x, y, z, w) − e
hi (t, ω, x, y, z, w)¯ ≤ C ¯y − y ′ ¯ + ¯z − z ′ ¯ + β ¯w − w′ ¯ ,
¯e
i=1

t ∈ R+ , ω ∈ Ω, x ∈ RN , y, y ′ ∈ R, z, z ′ ∈ RN , w, w′ ∈ RN ×N .
697

The constant C is arbitrary while α and β should satisfy the inequality
2αN 2 + β 2 N 4 < 2, which ensures the fulfillment of the technical condition
(H2). The bi-laplacian ∆2 is simply the square of the standard Laplacian
on RN . So one has
¡
¢
¡
¢
H = K = L2 RN , A = ∆2 , F = J = D (∆) = W 2,2 RN , R = ∆.

The “concrete” functions fe, ge, e
h of this example yield the following functions
acting on the functional spaces,
¡
¢
¡
¢
f (t, ω, u) = fe t, ω, ·, u, ∇u, D 2 u , g (t, ω, u) = ge t, ω, ·, u, ∇u, D 2 u ,
¡
¢
h (t, ω, u) = e
h t, ω, ·, u, ∇u, D 2 u ,

where f, g : R+ × F → H and h : R+ × F → H d . In order to check the
condition (H2) of the abstract framework , we use the following inequalities
¡
¢
k∂i ∂j uk ≤ k∆uk , u ∈ W 2,2 RN ,
p
¡
¢
−(u, ∆u) ≤ ε k∆uk + Cε kuk , u ∈ W 2,2 RN ,
k∇uk =

with ε arbitrary small and Cε depending on ε. So, one has

°¯
¯°
kg (t, ω, u) − g (t, ω, v)k ≤ C (ku − vk + k∇u − ∇vk) + α °¯D2 (u − v)¯°
≤ Cε ku − vk + (ε + αN 2 ) k∆(u − v)k

= Cε ku − vk + (ε + αN 2 )E(u − v)1/2 ,

and a similar relation for h. Moreover, it is clear that in this case, kRk ≤ 1.
The s.p.d.e. associated to the bi-laplacian models some physical phenomena (see [4] and the references therein for further results).

5.3

An example with an infinite dimensional state space

We use the notation of [3] and denote by (W, m) the Wiener space with
Wiener measure on it, H denotes the Cameron-Martin space, D is the Malliavin derivative, δ is its adjoint, the divergence. The s.p.d.e. we consider has
the form
1
g (t, ω, w, ut , Dut ) dt
dut − δDut dt + fe(t, ω, w, ut , Dut ) dt + δe
2
d
X
e
hi (t, ω, w, ut , Dut ) dBti (ω) .
=
i=1

The functions fe, ge, e
h are defined as follows
fe

:

R+ × Ω × W × R × H → R,

ge : R+ × Ω × W × R × H → H,
e
h = (e
h1 , · · ·, e
hd ) : R+ × Ω × W × R × H → Rd .
698

They are predictable, satisfy the integrability conditions
f (·, ·, ·, 0, 0) , hi (·, ·, ·, 0, 0) ∈ L2 ([0, t] × Ω × W ) , t > 0, i ∈ {1, · · ·, d}
g (·, ·, ·, 0, 0) ∈ L2 ([0, t] × Ω × W ; H) , t > 0,

and the Lipschitz condition in the form
¯
¯ ¢
¯ ¯
¡¯
¡
¢¯¯
¯e
¯f (t, ω, w, y, z) − fe t, ω, w, y ′ , z ′ ¯ ≤ C ¯y − y ′ ¯ + ¯z − z ′ ¯H ,
¯
¯
¯
¯
¯
¡
¢¯
¯ge (t, ω, w, y, z) − ge t, ω, w, y ′ , z ′ ¯
≤ C ¯y − y ′ ¯ + α ¯z − z ′ ¯H ,
H
v
u d ¯
¯2
uX ¯
¯
¯
¯
¯
¯
t
hi (t, ω, w, y, z) − e
hi (t, ω, w, y ′ , z ′ )¯ ≤ C ¯y − y ′ ¯ + β ¯z − z ′ ¯H ,
¯e
i=1

with an arbitrary constant C and α, β satisfying 2α + β 2 < 1.
In this case the state space is W and H = L2 (W, m) , A = − 21 δD
is the infinite
dimensional Ornstein-Uhlenbeck operator, F = D1,2 , and
R
E (u, v) = 21 hu, viH dm. The semigroup (Pt ) generated by A is the infinite
dimensional Ornstein-Uhlenbeck semigroup. The space K is L2 (W, m; H) ,
while R = δ. A direct verification shows that the conditions imposed on α
and β ensures applicability of the general result.

6

Doubly stochastic interpretation of solutions

The doubly stochastic interpretation of solutions of s.p.d.e. was first explored in [15] and turned out to be a powerful tool, giving a direct approach
to the study of s.p.d.e.’s. Further recent investigations are in [1] and [17].
In this section we only show that the doubly stochastic interpretation holds
in our case too.
We first construct an approximating sequence for the solution of the
following simple form of the main equation
dut + Aut dt =

d
X

wj,t dBtj , u0 = 0 .

j=1

We do this in the abstract setting. In fact, we already know an expression
for the solution and it was studied in Proposition 6. It will be approximated
by solutions of the following deterministic equation with suitable data w n ,
(∂t + A) v n = wn , v0n = 0 .
The next lemma may presumably have other applications. We are not
aware wether it is known.
699

Lemma 10 Let w = (w1 , ..., wd ) : R+ × Ω → (D (A))d , be such that w ∈
´d
³
e (F ) and
P
E sup kAwt k2d < ∞ , E sup kwt k4d < ∞ , (∀) T > 0.
t≤T

t≤T

Define the following process in the sense of Proposition 6,
ut =

d Z
X

0

j=1

and set
vtn

=

t

Z

t

0

Pt−s wj,s dBsj

Pt−s wsn ds,

where the integrand in the right hand side is defined by wtn =
for t ∈ [tni , tni+1 ) and tni = 2in , i ∈ N.
Then v n converges to u, in the following sense

³
´
j
n Bj
2
−
B
wj,ti
n
n
j=1
t
t

Pd

i+1

lim E sup kvtn − ut k2 = 0 , (∀) T > 0.
n

t≤T

As a consequence, there exists a subsequence (v nl )l∈N such that
lim sup kvtnl − ut k = 0 , a.s. (∀) T > 0.
l

t≤T

Proof. We are going to compare the approximands v n with the processes
un , defined by
unt

=

d Z
X
j=1

0

t

n
n
Pt−s w
ˆj,s
dBsj , w
ˆj,s
= wj,tni , s ∈ (tni , tni+1 ].

By Proposition 6 , we have
lim E sup kut − unt k2 = 0.

n→+∞

t≤T

We are first going to estimate unτ − vτn for some fixed dyadic number τ
which is written as k2nn for any large enough n, that we also considered as
fixed for the moment. One starts with the remark that we may write the
value of vτn in the form of a stoc