, , ,

We consider the following integro-differential stochastic equation with infinite delay

\begin{matrix} X (t, θ) = \int_{- \infty}^{t} b (t - s) [Δ X (s, θ) + \frac{\partial W_{Γ}}{\partial s} (s, θ)] d s, t \geq 0, θ \in T^{d}, \end{matrix}

(1.1)

where

b \in L^{1} (R_{+})

Δ

is the Laplace operator and

T^{d}

is the

d

-dimensional torus.

In ( 1.1 ),

W_{Γ}

is a spatially homogeneous Wiener process with the space covariance

Γ

and taking values in the space of tempered distributions

S^{'} (T^{d}) .

Such equation arises, in the deterministic case, in the study of heat flow in materials of fading memory type (see [3] , [14] ).

In this paper we address the following question: under what conditions on the covariance

Γ

the process

X

takes values in a Sobolev space

H^{α} (T^{d})

, particularly in

L^{2} (T^{d})

We study a particular case of weak solutions under the basis of an explicit representation of the solution to ( 1.1 ) (cf. Definition 3.4 ).

Observe that equation ( 1.1 ) can be viewed (see [16] ) as the limiting equation for the stochastic Volterra equation

\begin{matrix} X (t, θ) = \int_{0}^{t} b (t - s) [Δ X (s, θ) + \frac{\partial W_{Γ}}{\partial s} (s, θ)] d s, t \geq 0, θ \in T^{d} . \end{matrix}

(1.2)

b

is sufficiently regular, we get, by differentiating ( 1.2 ) with respect to

t

\begin{matrix} \frac{\partial X}{\partial t} (t, θ) = b (0) [Δ X (t, θ) + \frac{\partial W_{Γ}}{\partial t} (t, θ)] + \int_{0}^{t} b^{'} (t - s) [Δ X (s, θ) + \frac{\partial W_{Γ}}{\partial s} (s, θ)] d s, \end{matrix}

(1.3)

where

t \geq 0

and

θ \in T^{d} .

Taking in ( 1.3 ),

b (t) \equiv 1

we obtain

\begin{matrix} {\begin{matrix} \frac{\partial X}{\partial t} (t, θ) & = & Δ X (t, θ) + \frac{\partial W_{Γ}}{\partial t} (t, θ), t > 0, θ \in T^{d}, \\ X (0, θ) & = & 0, θ \in T^{d} . \end{matrix} \end{matrix}

(1.4)

Similarly, taking

b (t) \equiv t

and differentiating ( 1.2 ) twice with respect to

t

we obtain

\begin{matrix} {\begin{matrix} \frac{\partial^{2} X}{\partial t^{2}} (t, θ) & = & Δ X (t, θ) + \frac{\partial W_{Γ}}{\partial t} (t, θ), t > 0, θ \in T^{d}, \\ X (0, θ) & = & 0, θ \in T^{d}, \\ \frac{\partial X}{\partial t} (0, θ) & = & 0, θ \in T^{d} . \end{matrix} \end{matrix}

(1.5)

It has been shown in [8,Theorem5.1] (see also [7,Theorem1] ) that equations ( 1.4 ) and ( 1.5 ) on the

d

-dimensional torus

T^{d}

have an

H^{α + 1} (T^{d})

-valued solutions if and only if the Fourier coefficients

(γ_{n})

of the space covariance

Γ

of the process

W_{Γ}

satisfy

\begin{matrix} \sum_{n \in Z^{d}} γ_{n} (1 + | n |^{2})^{α} < \infty . \end{matrix}

(1.6)

Observe that for both, stochastic heat ( 1.4 ) and wave ( 1.5 ) equations, the conditions are exactly the same, despite of the different nature of the equations. On the other hand, the obtained characterization form a natural framework in which nonlinear heat and wave equations can be studied.

In the present paper, we will prove that condition ( 1.6 ) even characterizes

H^{α + 1} (T^{d})

-valued solutions for the stochastic Volterra equation ( 1.1 ), provided certain conditions on the kernel

b

are satisfied. This is a strong contrast with the deterministic case, where regularity of ( 1.1 ) is dependent on the kernel

b

We use, instead of resolvent families, a direct approach to the equation ( 1.1 ) finding an explicit expression for the solution in terms of the kernel

b

. This approach reduces the considered problem to questions in harmonic analysis and lead us with a complete answer.

2 Preliminaries

Let

(Ω, ℱ, (ℱ_{t})_{t \geq 0}, P)

be a complete filtered probability space. By

T^{d}

we denote the

d

-dimensional torus which can be identified with the product

(- π, π]^{d}

. Let

D (T^{d})

and

D^{'} (T^{d})

denote, respectively, the space of test functions on

T^{d}

and the space of distributions. By

〈 ξ, φ 〉

we denote the value of a distribution

ξ

on a test function. We assume that

W_{Γ}

is a

D^{'} (T^{d})

-valued spatially homogeneous Wiener process (abbr. s.h. Wiener process) with covariance

Γ

which is a positive definite distribution. For more information about s.h. Wiener process we refer to [9] .

Any arbitrary s.h. Wiener process

W_{Γ}

is uniquely determined by its covariance

Γ

according to the formula

\begin{matrix} E 〈 W_{Γ} (t, θ), φ 〉 〈 W_{Γ} (τ, θ), ψ 〉 = m i n (t, τ) 〈 Γ, φ ◆ ψ_{(s)} 〉, \end{matrix}

(2.1)

where

φ, ψ \in D (T^{d})

and

ψ_{(s)} (η) = ψ (- η)

, for

η \in T^{d}

. Because

W_{Γ}

is spatially homogeneous process, the distribution

Γ = Γ (θ - η)

for

θ, η \in T^{d}

The space covariance

Γ

, like distribution in

D^{'} (T^{d})

, may be uniquely expanded (see, e.g. [18] or [5] ) into its Fourier series (with parameter

w = 1

because the period is

2 π

)

\begin{matrix} Γ (θ) = \sum_{n \in Z^{d}} e^{i (n, θ)} γ_{n}, θ \in T^{d}, \end{matrix}

(2.2)

convergent in

D^{'} (T^{d})

. In ( 2.2 ),

(n, θ) = \sum_{i = 1}^{d} n_{i} θ_{i}

and

Z^{d}

denotes the product of integers.

The coefficients

γ_{n}

, in the Fourier series ( 2.2 ), satisfy:

(1) $γ_{n} = γ_{- n}$ for $n \in Z^{d}$ ,

(2) the sequence

(γ_{n})

is slowly increasing, that is

\begin{matrix} \sum_{n \in Z^{d}} \frac{γ_{n}}{1 + | n |^{r}} < + \infty, for  some r > 0 . \end{matrix}

(2.3)

Let us introduce, by induction, the following set of indexes. Denote

Z_{s}^{1} : = N

, the set of natural numbers and define

Z_{s}^{d + 1} : = (Z_{s}^{1} \times Z^{d}) \cup {(0, n) : n \in Z_{s}^{d}}

. Let us notice that

Z^{d} = Z_{s}^{d} \cup (- Z_{s}^{d}) \cup {0}

. For instance, for

d = 2

Z_{s}^{2} = N \times Z \cup {(0, n) : n \in Z}

Now, the s.h. Wiener process

W_{Γ}

corresponding to the covariance

Γ

given by ( 2.2 ) may be represented in the form

\begin{matrix} W_{Γ} (t, θ) = \sqrt{γ_{0}} β_{0} (t) + \sum_{n \in Z_{s}^{d}} \sqrt{2 γ_{n}} [cos (n, θ) β_{n}^{1} (t) & + & sin (n, θ) β_{n}^{2} (t)], \end{matrix}

(2.4)

\begin{matrix} t \geq 0 and θ \in T^{d} . \end{matrix}

In ( 2.4 )

β_{0}, β_{n}^{1}, β_{n}^{2}, n \in Z_{s}^{d}

, are independent real Brownian motions and

γ_{0}, γ_{n}

are coefficients of the series ( 2.2 ). The series ( 2.4 ) is convergent in the sense of

D^{'} (T^{d})

Because any periodic distribution with positive period is a tempered distribution (see, e.g [5] ), we may restrict our considerations to the space

S^{'} (T^{d})

of tempered distributions. By

S (T^{d})

we denote the space of infinitely differentiable rapidly decreasing functions on the torus

T^{d}

Let us denote by

H^{α} = H^{α} (T^{d})

α \in R

, the real Sobolev space of order

α

on the torus

T^{d}

. The norms in such spaces may be expressed in terms of the Fourier coefficients (see [1] )

| | ξ | |_{H^{α}} = {(\sum_{n \in Z^{d}} (1 + | n |^{2})^{α} | ξ_{n} |^{2})}^{\frac{1}{2}} = {(| ξ_{0} |^{2} + 2 \sum_{n \in Z_{s}^{d}} (1 + | n |^{2})^{α} ((ξ_{n}^{1})^{2} + (ξ_{n}^{2})^{2}))}^{\frac{1}{2}},

where

ξ_{n} = ξ_{n}^{1} + i ξ_{n}^{2}, ξ_{=} {\bar{ξ}}_{- n}, n \in Z^{d}

There is another possibility to define the Sobolev spaces (see, e.g. [17] ). We say that a distribution

ξ \in S^{'} (T^{d})

belongs to

H^{α}, α \in R

, if its Fourier transform

\hat{ξ}

is a measurable function and

\int_{T^{d}} (1 + | λ |^{2})^{α} | \hat{ξ} (| λ |) |^{2} d λ < + \infty .

3 Main results

b \in L_{l o c}^{1} (R_{+})

and

μ \in C

, we shall denote by

r (t, μ)

the unique solution in

L_{l o c}^{1} (R_{+})

to the linear Volterra equation

\begin{matrix} r (t, μ) = b (t) + μ \int_{0}^{t} b (t - s) r (s, μ) d s, t \geq 0 . \end{matrix}

(3.1)

In many cases the function

r (t, μ)

may be found explicitly. For instance:

\begin{matrix} b (t) & \equiv 1, & r (t, μ) = e^{μ t} \end{matrix}

\begin{matrix} b (t) & = t, & r (t, μ) = \frac{sinh \sqrt{μ} t}{\sqrt{μ}} μ \neq 0 \end{matrix}

\begin{matrix} b (t) & = e^{- t}, & r (t, μ) = e^{(1 + μ) t} \end{matrix}

\begin{matrix} b (t) & = t e^{- t}, & r (t, μ) = e^{- t} \frac{sinh \sqrt{μ} t}{\sqrt{μ}} . \end{matrix}

Remark 3.1. Let us denote by

\hat{f} (k), k \in Z

, the

k

-th Fourier coefficient of

f

\hat{f} (k) = \frac{1}{2 π} \int_{0}^{2 π} e^{i k t} f (t) d t .

Given

b \in L^{1} (R_{+})

, we find that, for

F (t) : = \int_{- \infty}^{t} b (t - s) f (s) d s

, we have

\begin{matrix} \hat{F} (k) = \tilde{b} (i k) \hat{f} (k), k \in Z, \end{matrix}

(3.2)

where

\tilde{b} (λ) : = \int_{0}^{\infty} e^{- λ t} b (t) d t

denotes the Laplace transform of

b

We state the following hypothesis:

Definition 3.2. We say that a kernel

b \in L^{1} (R_{+})

is admissible for equation ( 1.1 ) if

{lim}_{| n | \to \infty} | n |^{2} \int_{0}^{\infty} [r (s, - | n |^{2})]^{2} d s = : C_{b}

exists.

Example 3.3. i) In case

b (t) = e^{- t}

we obtain

| n |^{2} \int_{0}^{\infty} [r (s, - | n |^{2})]^{2} d s = \frac{- | n |^{2}}{2 (1 - | n |^{2})}

and hence

C_{b} = \frac{1}{2} .

ii) In case

b (t) = t e^{- t}

we obtain

| n |^{2} \int_{0}^{\infty} [r (s, - | n |^{2})]^{2} d s = \frac{| n |^{2}}{4 + 4 | n |^{2}}

and hence

C_{b} = \frac{1}{4} .

Denote by

W_{n} (t, θ) : = cos (n, θ) β_{n}^{1} (t) + sin (n, θ) β_{n}^{2} (t), n \in Z_{s}^{d}

, that is the

n

-th element in the expansion ( 2.4 ).

Definition 3.4. By a solution

X (t, θ)

to the stochastic Volterra equation ( 1.1 ) we will understand the process of the form

\begin{matrix} X (t, θ) = \sqrt{γ_{0}} β_{0} (t) + \sum_{n \in Z_{s}^{d}} \sqrt{2 γ_{n}} \int_{- \infty}^{t} r (t - s, - | n |^{2}) d W_{n} (s, θ), \end{matrix}

(3.3)

where the function

r

is as above,

t \geq 0

and

θ \in T^{d}

The process

X

given by ( 3.3 ) is a particular form of the weak solution to the equation ( 1.1 ) (cf. [9] ) and takes values in the space

S^{'} (T^{d})

The following is our main result.

Theorem 3.5. Assume

b \in L^{1} (R_{+})

is admissible for ( 1.1 ). Then, the equation ( 1.1 ) has an

H^{α + 1} (T^{d})

-valued solution if and only if the Fourier coefficients

(γ_{n})

of the covariance

Γ

satisfy

\begin{matrix} \sum_{n \in Z^{d}} γ_{n} (1 + | n |^{2})^{α} < \infty . \end{matrix}

(3.4)

Remark 3.6. Particularly, the eq. ( 1.1 ) has

L^{2} (T^{d})

-valued solution if and only if

\sum_{n \in Z^{d}} \frac{γ_{n}}{1 + | n |^{2}} < \infty .

Proof. We shall use the representation ( 2.4 ) for the Wiener process

W_{Γ} (t, θ)

with respect to the basis:

1, cos (n, θ), sin (n, θ)

, where

n \in Z_{s}^{d}

and

θ \in T^{d}

. Equation ( 1.1 ) may be solved coordinatewise as follows.

Assume that

\begin{matrix} X (t, θ) = \sum_{n \in Z_{s}^{d}} [cos (n, θ) X_{n}^{1} (t) + sin (n, θ) X_{n}^{2} (t)] + X_{0} (t) . \end{matrix}

(3.5)

Introducing ( 3.5 ) into ( 1.1 ), we obtain

\begin{matrix} cos (n, θ) X_{n}^{1} (t) + sin (n, θ) X_{n}^{2} (t) & = & - | n |^{2} \int_{- \infty}^{t} b (t - s) [cos (n, θ) X_{n}^{1} (s) + sin (n, θ) X_{n}^{2} (s)] d s \end{matrix}

\begin{matrix} + \sqrt{2 γ_{n}} \int_{- \infty}^{t} b (t - s) [cos (n, θ) β_{n}^{1} (s) + sin (n, θ) β_{n}^{2} (s)] d s, \end{matrix}

or, equivalently

\begin{matrix} [cos (n, θ), sin (n, θ)] [\begin{matrix} X_{n}^{1} (t) \\ X_{n}^{2} (t) \end{matrix}] & = & - | n |^{2} \int_{- \infty}^{t} b (t - s) [cos (n, θ), sin (n, θ)] [\begin{matrix} X_{n}^{1} (s) \\ X_{n}^{2} (s) \end{matrix}] d s \end{matrix}

\begin{matrix} + & \sqrt{2 γ_{n}} \int_{- \infty}^{t} b (t - s) [cos (n, θ), sin (n, θ)] [\begin{matrix} d β_{n}^{1} (s) \\ d β_{n}^{2} (s) \end{matrix}] . \end{matrix}

Denoting

X_{n} (t) : = [\begin{matrix} X_{n}^{1} (t) \\ X_{n}^{2} (t) \end{matrix}] and β_{n} (t) : = [\begin{matrix} β_{n}^{1} (t) \\ β_{n}^{2} (t) \end{matrix}]

we arrive at the equation

\begin{matrix} X_{n} (t) = - | n |^{2} \int_{- \infty}^{t} b (t - s) X_{n} (s) d s + \sqrt{2 γ_{n}} \int_{- \infty}^{t} b (t - s) d β_{n} (s) . \end{matrix}

(3.6)

Taking Fourier transform in

t

, and making use of ( 3.2 ), we get the following solution to the equation ( 3.6 ):

X_{n} (t) = \int_{- \infty}^{t} r (t - s, - | n |^{2}) \sqrt{2 γ_{n}} d β_{n} (s) = \int_{0}^{\infty} r (s, - | n |^{2}) \sqrt{2 γ_{n}} d β_{n} (t - s) .

Hence, we deduce the following explicit formula for the solution to the equation ( 1.1 ):

\begin{matrix} \begin{matrix} X (t, θ) = \sqrt{γ_{0}} β_{0} (t) & + & \sum_{n \in Z_{s}^{d}} \sqrt{2 γ_{n}} [cos (n, θ) \int_{0}^{\infty} r (s, - | n |^{2}) d β_{n}^{1} (t - s) \\ + & sin (n, θ) \int_{0}^{\infty} r (s, - | n |^{2}) d β_{n}^{2} (t - s)] . \end{matrix} \end{matrix}

(3.7)

Since the series defining the process

X

converges in

S^{'} (T^{d}), P

-almost surely, it follows from the definition of the space

H^{α}

that

X (t) \in H^{α + 1}, P

-almost surely if and only if

\begin{matrix} \sum_{n \in Z^{d}} (1 + | n |^{2})^{α + 1} γ_{n} [{(\int_{- \infty}^{t} r (t - s, - | n |^{2}) d β_{n}^{1} (s))}^{2} + {(\int_{- \infty}^{t} r (t - s, - | n |^{2}) d β_{n}^{2} (s))}^{2}] < \infty . \end{matrix}

(3.8)

Because the stochastic integrals in ( 3.8 ) are independent Gaussian random variables, we obtain that ( 3.8 ) holds

P

-almost surely if and only if

\begin{matrix} \sum_{n \in Z^{d}} (1 + | n |^{2})^{α + 1} γ_{n} E [(\int_{- \infty}^{t} r (t - s, - | n |^{2}) d β_{n}^{1} (s))^{2} + (\int_{- \infty}^{t} r (t - s, - | n |^{2}) d β_{n}^{2} (s))^{2}] < \infty . \end{matrix}

(3.9)

Or equivalently, using properties of stochastic integrals, if and only if

\begin{matrix} \sum_{n \in Z^{d}} (1 + | n |^{2})^{α + 1} γ_{n} \int_{0}^{\infty} [r (s, - | n |^{2})]^{2} d s < \infty . \end{matrix}

(3.10)

Since

b

is admissible for the equation ( 1.1 ), we conclude that ( 3.10 ) holds if and only if

\sum_{n \in Z^{d}} (1 + | n |^{2})^{α + 1} \frac{γ_{n}}{| n |^{2}} < \infty,

and the proof is complete.

Concerning uniqueness, we have the following result.

Proposition 3.7. Assume

b \in L^{1} (R_{+})

is admissible for ( 1.1 ) and the following conditions hold:

(i)

\sum_{n \in Z^{d}} γ_{n} (1 + | n |^{2})^{α} < \infty,

(ii)

{\frac{1}{\hat{b} (i k)}}_{k \in Z} \subset C \ {- | n |^{2} : n \in Z^{d}}

Then, ( 1.1 ) has a unique

H^{α + 1} (T^{d})

-valued solution.

Proof. Let

X (t, θ)

be solution of

X (t, θ) = \int_{- \infty}^{t} b (t - s) Δ X (s, θ) d s .

Taking Fourier transform in

θ

and denoting by

X_{n} (t)

the

n

-th Fourier coefficient of

X (t)

, we obtain

X_{n} (t) = - | n |^{2} \int_{- \infty}^{t} b (t - s) X_{n} (s) d s

for all

n \in Z^{d} .

Taking now Fourier transform in

t

, we get

(1 + | n |^{2} \hat{b} (i k)) {\tilde{X}}_{n} (k) = 0

for all

k \in Z

and all

n \in Z^{d} .

According to (ii) we obtain that

{\tilde{X}}_{n} (k) = 0

for all

k \in Z

and all

n \in Z^{d} .

Hence, the assertion follows by uniqueness of the Fourier transform.

The following corollaries are an immediate consequence of Theorem 3.5 . The arguments are the same as in [7] . We give here the proof for the sake of completeness.

Corollary 3.8. Suppose

b \in L^{1} (R_{+})

is admissible for ( 1.1 ) and assume

Γ \in L^{2} (T^{d}) .

Then the integro-differential stochastic equation ( 1.1 ) has a solution with values in

L^{2} (T^{d})

for

d = 1, 2, 3 .

Proof. We have to check equation ( 3.4 ) with

α = - 1

. Note, that if

Γ \in L^{2} (T^{d})

then

\hat{Γ} = (γ_{n}) \in l^{2} (Z^{d}) .

Consequently

\sum_{n \in Z^{d}} \frac{γ_{n}}{1 + | n |^{2}} \leq (\sum_{n \in Z^{d}} {γ_{n}}^{2})^{1 / 2} (\sum_{n \in Z^{d}} \frac{1}{(1 + | n |^{2})^{2}})^{1 / 2} .

But

\sum_{n \in Z^{d}} {γ_{n}}^{2} < \infty

and

\begin{matrix} \sum_{n \in Z^{d}} \frac{1}{(1 + | n |^{2})^{q}} < \infty if and only if 2 q > d . \end{matrix}

(3.11)

Hence, the result follows.

Corollary 3.9. Suppose

b \in L^{1} (R_{+})

is admissible for ( 1.1 ) and assume

\hat{Γ} \in l^{p} (Z^{d})

for

1 < p \leq 2 .

Then the integro-differential stochastic equation ( 1.1 ) has a solution with values in

L^{2} (T^{d})

for all

d < \frac{2 p}{p - 1} .

Proof. Note that

\sum_{n \in Z^{d}} \frac{γ_{n}}{1 + | n |^{2}} \leq (\sum_{n \in Z^{d}} {γ_{n}}^{p})^{1 / p} (\sum_{n \in Z^{d}} \frac{1}{(1 + | n |^{2})^{q}})^{1 / q},

where

\frac{1}{p} + \frac{1}{q} = 1 .

Hence the result follows from ( 3.11 ) with

q = \frac{p}{p - 1} .

For

α = - 1

, the condition ( 3.4 ) can be written as follows.

Theorem 3.10. Let

b

be admissible for the equation ( 1.1 ). Assume that the covariance

Γ

is not only a positive definite distribution but is also a non-negative measure. Then the equation ( 1.1 ) has

L^{2} (T^{d})

-valued solution if and only if

\begin{matrix} (Γ, G_{d}) < + \infty, \end{matrix}

(3.12)

where

\begin{matrix} G_{d} (x) = \sum_{n \in Z^{d}} \int_{0}^{+ \infty} \frac{1}{\sqrt{(4 π t)^{d}}} e^{- t} e^{- \frac{| x + 2 π n |^{2}}{t}} d t, x \in T^{d} . \end{matrix}

(3.13)

The proof of Theorem 3.10 is the same that for Theorem 2, part 2) in [7] , so we omit it. For more details concerning the function

G_{d}

we refer to [7] and [10] .

Additionally, from properties of function

G_{d}

defined by ( 3.13 ) and the condition ( 3.12 ) we obtain the following result (see [8,Theorem4] ).

Corollary 3.11. Assume that

Γ

is a non-negative measure. Then equation ( 1.1 ) has function valued solutions:

i) for all

Γ

d = 1

; ii) for exactly those

Γ

for which

\int_{| θ | \leq 1} ln | θ | Γ (d θ) < + \infty

d = 2

; iii) for exactly those

Γ

for which

\int_{| θ | \leq 1} \frac{1}{| θ |^{d - 2}} Γ (d θ) < + \infty

d \geq 3

Let us see that the formula ( 3.7 ) of the solution provides the next regularity result. In order to prove hölderianity of

X

with respect to

t

, we need assumptions very similar to those in [2] .

Hypothesis (H) Assume that there exist

δ \in (0, 1)

and

C_{δ} > 0

such that, for all

τ \in (- \infty, t)

we have:

(i)

\int_{s}^{t} r^{2} (t - τ, - | n |^{2}) d τ \leq C_{δ} | n |^{2 (δ - 1)} | t - s |^{δ}

; (ii)

\int_{- \infty}^{s} [r (t - τ, - | n |^{2}) - r (s - τ, - | n |^{2})]^{2} d τ \leq C_{δ} | n |^{2 (δ - 1)} | t - s |^{δ}

Proposition 3.12. Assume that

\sum_{n \in Z^{d}} \frac{γ_{n}}{1 + | n |^{2}} < + \infty

. Under Hypothesis (H), the trajectories of the solution

X

to the equation ( 1.1 ) are almost surely

η

-Hölder continuous with respect to

t

, for every

η \in (0, δ / 2)

Proof. From the expansion ( 3.7 ) and properties of stochastic integral, we have

\begin{matrix} E & | | X (t, θ) - X (s, θ) | |_{L^{2}}^{2} = E | | \sqrt{γ_{0}} (β_{0} (t) - β_{0} (s)) \end{matrix}

\begin{matrix} + \sum_{n \in Z_{s}^{d}} \sqrt{2 γ_{n}} [cos (n, θ) (\int_{- \infty}^{t} r (t - τ, - | n |^{2}) d β_{n}^{1} (τ) - \int_{- \infty}^{s} r (s - τ, - | n |^{2}) d β_{n}^{1} (τ)) \end{matrix}

\begin{matrix} {+ sin (n, θ) (\int_{- \infty}^{t} r (t - τ, - | n |^{2}) d β_{n}^{2} (τ) - \int_{- \infty}^{s} r (s - τ, - | n |^{2}) d β_{n}^{2} (τ))] | |}_{L^{2}}^{2} \end{matrix}

\begin{matrix} = & (2 π)^{d} (γ_{0} | t - s | \end{matrix}

\begin{matrix} + \sum_{n \in Z_{s}^{d}} 2 γ_{n} [\int_{- \infty}^{s} [r (t - τ, - | n |^{2}) - r (s - τ, - | n |^{2})]^{2} d τ + \int_{s}^{t} r^{2} (t - τ, - | n |^{2}) d τ]) . \end{matrix}

According to assumptions (i) and (ii) of the Hypothesis (H), we have

E | | X (t, θ) - X (s, θ) | |_{L^{2}}^{2} \leq C_{δ} \sum_{n \in Z_{s}^{d}} 2 γ_{n} | n |^{2 (δ - 1)} | t - s |^{δ} .

Because

X

is a Gaussian process, then for any

m \in N

, there exists a constant

C_{m} > 0

that

E | | X (t, θ) - X (s, θ) | |_{L^{2}}^{2 m} \leq C_{m} {[C_{δ} \sum_{n \in Z_{s}^{d}} 2 γ_{n} | n |^{2 (δ - 1)}]}^{m} | t - s |^{m δ} .

Taking

m

such that

m δ > 1

and using the Kolmogorov test, we see that the solution

X (t, θ)

η

-Hölder continuous, with respect to

t

, for

η = δ / 2 - 1 / (2 m)

Example 3.13. Let us consider the particular case

b (t) = e^{- t}, t \geq 0

Then, by previous considerations,

r (t, - | n |^{2}) = e^{(1 - | n |^{2}) t}

. One can check that in this case the Hypothesis (H) is fulfilled.

References

R. Adams, Sobolev Spaces, Academic Press, New York, 1975.
Ph. Clément, G. Da Prato, Some results on stochastic convolutions arising in Volterra equations perturbed by noise, Rend. Math. Acc. Lincei. s. 9, 7, (1996) 147–153.
Ph. Clément, G. Da Prato, Existence and regularity results for an integral equation with infinite delay in a Banach space, Integral Equations and Operator Theory, 11, (1988), 480-500.
R. Dalang, N. Frangos, The stochastic wave equation in two spatial dimensions, The Annals of Probability 26 (1998), 187-212.
C. Gasquet, P. Witomski, Fourier Analysis and Applications, Springer-Verlag, New York Berlin, 1999.
A. Karczewska, Regularity and continuity of solutions to stochastic evolution equations, G. Da Prato and L. Tubaro (eds.), Stochastic partial differential equations and applications, Marcel Dekker, Lect. Notes Pure Appl. Math. 227, 309-323 (2002).
A. Karczewska, J. Zabczyk, A note on stochastic wave equations, G. Lumer and L. Weis (eds.), Evolution equations and their applications in physical and life sciences, Marcel Dekker, Lect. Notes Pure Appl. Math. 215, 501-511 (2001).
A. Karczewska, J. Zabczyk, Stochastic PDE's with function-valued solutions. Ph. Clément (ed.) et al., Infinite dimensional stochastic analysis. Royal Netherlands Academy of Arts and Sciences. Verh. Afd. Natuurkd., 1. Reeks, K. Ned. Akad. Wet. 52, 197-216 (2000).
A. Karczewska, J. Zabczyk, Regularity of solutions to stochastic Volterra equations, Rend. Math. Acc. Lincei. s. 9, 11, (2001) 141–154.
N.S. Landkof, Foundations of modern potential theory, Springer-Verlag, Berlin, 1975.
O. Lévêque, Hyperbolic stochastic partial differential equations driven by boundary noises, Ph.D. Thesis, Ecole Polytechnique Fédérale de Lausanne, 2001.
R.C. Mac Camy, Stability theorems for a class of functional differential equations, SIAM J. Appl. Math., 30 (1976), 557-576.
C. Mueller, Long time existence for the wave equation with a noise term, The Annals of Probability 25 (1) (1997), 133-151.
J.W. Nunziato, On heat conduction in materials with memory, Quart. Appl. Math. 29, (1971), 187-204.
G. Da Prato, A. Lunardi, Periodic solutions for linear integrodiferential equations with infinite delay in Banach spaces. Differential Equations in Banach spaces, Springer-Verlag, Lecture Notes in Mathematics 1223, 49-60 (1985).
J. Prüss. Evolutionary Integral Equations and Applications. Monographs Math., vol. 87, Birkhäuser Verlag, Basel, 1993.
M. Reed, B. Simon, Methods of modern mathematical physics, Vol. II, Academic Press, New York, 1975.
L. Schwartz, Méthodes mathématiques pour les sciences physiques, Hermann, Paris,1965.

Department of Mathematics, University of Zielona Gora, ul. Szafrana 4a, 65-246 Zielona Gora, Poland. E-mail address : A.Karczewska@im.uz.zgora.pl Universidad de Santiago de Chile, Departamento de Matematica, Facultad de Ciencias, Casilla 307-Correo 2, Santiago, Chile. E-mail address : clizama@lauca.usach.cl

Anna Karczewska

Carlos Lizama