1. Introduction
The primary focus of this study lies in investigating nonlinear systems of differential equations:
where
is a bounded domain in
. This class of models captures the behavior of dynamical systems influenced by stochastic perturbations together with time-delay effects, operating within a confined spatial domain. Such formulations arise naturally in describing physical or biological processes, for instance, thermal conduction with hereditary memory, constrained population growth, or the transmission of signals in bounded media subject to random disturbances such as noise or defects. The boundary condition
on
corresponds to physically relevant situations: insulated boundaries in thermal physics, prescribed displacements in elasticity, or absorbing interfaces in ecological systems.
The associated problem in the complete space can be expressed as
where
and
. Here,
represents the delay interval. We define the elliptic operator
by
A generalized setting extends the framework from finite regions to the entire space, thereby addressing phenomena such as diffusion, wave propagation, and transport without spatial restrictions. In this broader context, one encounters models for fluid motion under stochastic influences, electromagnetic waves traversing random media, or financial processes evolving in unbounded domains. Stochastic forcing in these equations accounts for uncertainties including turbulent fluctuations, market variability, or randomly varying external excitations.
Nonlinear systems of differential equations, specifically of the forms (
1) and (
2), were originally introduced in prior studies [
1,
2,
3] as effective mathematical representations for processes that exhibit a strong reliance on their historical states. Memory effects of this type manifest across numerous scientific and engineering disciplines. In population dynamics and epidemiology, delays represent gestation or incubation intervals that modify reproductive or infection rates [
4]. Within viscoelastic materials, current stress–strain relations depend explicitly on prior deformation history [
5]. In control theory, feedback mechanisms are typically subject to signal delays that critically impact stability [
6]. In thermal science, generalized conduction models with memory kernels describe non-Fourier heat transfer [
7]. Likewise, in neuroscience, synaptic delays are decisive in shaping the dynamical patterns of neuronal assemblies [
8]. These diverse scenarios underscore how delay-driven nonlinear systems provide realistic representations of complex temporal behavior.
The study of nonlinear differential systems with memory and delay has a substantial history, both in bounded and unbounded domains. Foundational contributions include the delay-dependent stability criteria in [
9], the rigorous treatment of hereditary effects in nonlinear evolution equations developed in [
10,
11,
12], and the deployment of functional-analytic methods in [
13,
14]. Collectively, these works established core methodologies for analyzing systems in which the present state depends fundamentally on the past. Distinct from earlier studies, the present work broadens the scope to encompass a wider variety of nonlinearities within second-order systems, formulated both on bounded domains and on
. The approach integrates operator-theoretic techniques with iterative constructions to establish existence, uniqueness, and qualitative dynamics. This dual formulation and methodological synthesis provide a unified perspective on problems that were traditionally treated in narrower contexts.
The primary objective of this current study is to demonstrate the existence and uniqueness of invariant measures for systems (
1) and (
2) based on the Phragmén–Lindelöf method [
15]. Specifically, we will employ the Phragmén–Lindelöf theorem proposed by Li and Chen [
15], which comprises the following key steps: Our first focus is on establishing the existence of the solution for systems (
1) or (
2). To achieve this, we will explore a specific functional space where the corresponding transition semigroup exhibits Feller properties. This analysis will provide valuable insights into the long-term behavior and stability of the system. In addition to investigating the existence of the solution, we will also delve into the compactness of the semigroup
generated by
. Through this investigation, we will gain a deeper understanding of the stability and convergence properties of the semigroup, which are crucial for comprehending the overall behavior of the system. Furthermore, we will explore the boundedness in probability of the relevant equation under appropriate initial conditions. By examining the behavior of the solution within a certain probability range, we can ascertain its reliability and predictability under various circumstances.
Lately, this method has been utilized to ascertain the presence of an unchanging measure across various categories of partial differential equations (PDEs). Notable examples include fluid equations, nonlinear Schrödinger systems, quasilinear elliptic systems, and nonlocal partial differential equations. For instance, Alghanmi et al. [
16] explored the solvability of coupled systems of nonlinear implicit differential equations incorporating
-fractional derivatives subject to anti-periodic boundary conditions, thereby contributing to the refinement of fractional-order modeling. Hao et al. [
17] introduced a companion-based multi-level finite element framework for identifying multiple solutions of nonlinear differential equations, thus advancing computational methodologies within nonlinear analysis. Jiao et al. [
18] carried out a detailed investigation of solution structures for complex nonlinear partial differential difference equations in
, which deepened the understanding of discrete–continuous hybrid systems. Lan [
19] derived both existence and uniqueness theorems for nonlinear Cauchy-type problems governed by first-order fractional differential equations, offering a rigorous mathematical foundation for fractional dynamical systems. Moreover, Xu et al. [
20] established novel solution results for various classes of product-type nonlinear PDEs in
, thereby expanding analytic techniques applicable to higher-dimensional complex domains.
This study examines the existence of an invariant measure in the phase space
and focuses on the mild solutions of system (
2) in the phase space
, where
is a weighted space. Previous research has mainly investigated the systems (
1) and (
2) in
, which offers a simpler problem. However, in order to apply the Phragmén–Lindelöf method, it is necessary to work in
, which is the primary focus of our research. Additionally, we establish the uniqueness and existence of the solution. The existing literature has explored the conditions for the existence and uniqueness of the solution, as well as the Feller and Markov properties in the aforementioned spaces. However, our study expands upon this work by investigating the more complex problem in the phase space
and demonstrating the existence and uniqueness of the solution.
To enhance the clarity and coherence of this article, the content is structured as follows: In
Section 2 we introduce the notations. In
Section 3, we formulate the main results.
Section 4,
Section 5,
Section 6,
Section 7 and
Section 8 are dedicated to completing the proofs of them, namely Theorems 1, 2, and 4–6. Notably, in Theorem 5, we not only prove the existence of an invariant measure but also present an insightful example illustrating the application of Theorem 5 to nonlinear systems of integral differential equations.
Section 9 provides a conclusive summary.
2. Preliminaries
In what follows, we consider the domain
, which can either be a bounded domain with a boundary
that satisfies the Lyapunov condition or
is equal to
. The parameters
and
are introduced through the weight functions
These weight functions are introduced to regulate the asymptotic behavior of solutions and measures in the unbounded setting
, ensuring proper decay or integrability within the associated functional framework. More specifically:
The parameter determines the primary decay imposed on functions in weighted Sobolev or Banach spaces. Larger values of correspond to a steeper decay in the weight , leading to stricter integrability conditions for solutions and invariant measures.
The auxiliary exponent arises in defining dual or comparison weights, such as , and naturally appears in duality arguments or estimates involving test functions. Its role is to guarantee the finiteness of integrals where both and are present.
The inequality
highlighted in Theorem 4 plays a decisive role. Since
n represents the spatial dimension, this condition ensures that the imposed decay is sufficiently strong to balance the volume growth of
as
together with the weaker contribution from
. Consequently, it guarantees the tightness of measure families and compactness of embeddings in weighted spaces, both of which are indispensable for proving the existence of invariant measures.
Next, we present the following spaces such as
with
The coefficients
of
, as defined in (
3), exhibit Hölder continuity with
. These coefficients are also symmetric, bounded, and satisfy the ellipticity condition
Here,
is explicitly introduced as a fixed positive constant ensuring uniform ellipticity. It is part of the underlying hypothesis rather than a free parameter.
We enforce homogeneous Dirichlet boundary conditions on the boundary of
in the case that
is bounded. In this scenario,
If
, then one has that
Let
represent the fundamental solution for
. It can be deduced from references such as [
21], p. 311, that there are two positive constants
and
such that
holds for
and
.
Notably, in inequality (
5), the two positive constants
and
rely not only
, but on the constants
,
n,
, the Hölder constants, and maximum values of the coefficients of
. The estimates display various properties when the operator is represented in the divergence form
, as demonstrated in [
22], p. 202, denoted as
where
Here, the quantity
depends solely on the ellipticity constant
, the time horizon
, and the Hölder/supremum norms of the coefficients of
, but remains independent of the spatial variable
x. Since the same
arises for
and
, it follows that
holds.
Lemma 1. Given any positive value of , there exists a positive constant such thatholds true. Proof. Notice that (
4) yields that
for a given positive
. So
□
Put
and
, where
denotes the identity mapping.
It is a semigroup defined on the function space
, and its generator is denoted as
. By applying Lemma 1, it follows for all
and for any
,
holds. The aforementioned estimation facilitates the extension of the semigroup
into a linear mapping that operates from
to itself. Since
is densely packed within
, it can be inferred that
demonstrates strong continuity within
.
Let
,
, and
be equipped with an orthonormal basis
. Here, it is assumed that
and
. We introduce the operator
as an element of
. The operator
satisfies the conditions of being non-negative,
, and
. Let
be a complete probability space. We define
which denotes a
-Wiener process with values in
on the interval
. Here,
represents standard, one-dimensional, mutually independent Wiener processes. Additionally, we consider a normal filtration denoted by
that satisfies the following conditions:
Let
be denoted. According to [
23], [Lemma 6.2.2], it is concluded that
. In line with [
24], we introduce the multiplication operator denoted as
, which can be defined as follows: for a fixed
,
for any
. As both
and
hold, this operator is well defined, resulting in
being a Volterra–Fredholm operator. Moreover, the operator ∧ also satisfies the properties of being a Volterra–Fredholm operator, as stated in the following condition:
where
. So if
is a predictable process such that
then one has that
with
Moreover,
Let us make the assumptions that f and satisfy the following conditions:
- (i)
There exists a positive constant
such that
holds for any
.
- (ii)
Both functionals f and map elements from to .
We define a stochastic process
to be a mild solution of (
1) or (
2) if it satisfies the following equation:
subject to the given conditions:
Hence, the phase space of the problem can be represented as the Sobolev space . In this scenario, the variable belongs to if it can be expressed as , where is an element of and is an element of . Here, is defined as , with taking values in the interval .
3. Main Results
Theorem 1. Assume that f and τ fulfill conditions (i) and (ii), while is a measurable stochastic process, where . This process is independent of Φ
and satisfies the conditionsUnder these conditions, a unique mild solution of the system (
1)
(or (
2)
) exists on the interval . Furthermore, the following inequality holds for any : Theorem 2. For any such that , and any such that , we define the functions and as follows:Given the conditions specified in Theorem 1, there exists a constant such that the following inequality holds: The subsequent proposition demonstrates the continuity of the trajectories of the solution .
Proposition 1. Given that is a mild solution of either (
1)
or (
2)
, and taking into account the conditions stipulated in Theorem 1, we can deduce that exhibits probability continuity at associated with the norm . This can be expressed as follows: Proof. Notice that
In [
25], [Theorem 1.4.2], Pazy finds the density of
in
, and the first term can be shown to converge to zero. Likewise, the second term converges to zero as
x tends to zero, owing to the boundedness of the integrand. □
The Banach space
is defined to include bounded real Borel functions from
to
. The existence of the solution for all
is ensured through the arbitrary selection of
in Theorem 1, which implies the corresponding existence of
for
. By substituting
with
for
as the initial interval, we can establish the uniqueness and existence of solutions for
, which is denoted as
. Likewise,
represents a shift of the solution
, with the property that
, and for
,
.
Following the work of [
26], we introduce the family of shift operators as
Let denote the minimal -algebra that includes . It should be emphasized that the independence of from the -algebra , defined as the minimal sigma-algebra including for , is worth noting.
For any nonstochastic with and , is an -measurable stochastic function that takes values in , where for . By defining , we establish that y maps into itself. Theorem 1 yields the following result.
Proposition 2. Scientifically speaking, the family of the operators (3)
satisfieswhere and . Let
represent a
-algebra consisting of Borel subsets of
. Therefore,
naturally denotes the probability measure
defined on
as follows:
This measure
can be regarded as the transition function associated with the stochastic process
. Similar to the finite dimensional case discussed in [
24], p. 47, we can substantiate that it fulfills the criteria of a transition probability. In this manner, we obtain the following result.
Theorem 3. Considering the assumptions of Theorem 1, we can conclude that the process functions as a Markov process on . The transition function, denoted by , is given by (
13).
Proposition 3. For any , one has that Proof. If we denote
, then one has that
and
. On the other hand,
where
represents a
-Wiener process once more. In this way, the function
solves
The equation
satisfies the same conditions, where
and
. The only distinction is that
is a solution to (
14). But, due to the identical distribution of
and
, the distribution of
is equal to that of
and is therefore independent of
t. Consequently, the distribution of
fits into the distribution of
. This yields the desired result:
□
For
,
, and
, we define
Based on Proposition 3, we obtain the expression , which can be denoted as . Taking into account Proposition 1 and Theorem 2, the subsequent result is derived.
Proposition 4. Assuming conditions of Theorem 1 hold, we can conclude that possesses the Feller property and is stochastically continuous. Specifically, it satisfies the following property: If we introduce the function , then the central result of our study is presented in the subsequent theorem.
Theorem 4. If we suppose that the conditions of Theorem 1 are satisfied and (
11)
possesses a solution in that exhibits boundedness in probability for satisfyingwe can conclude the existence of an invariant measure μ on , denoted as . This measure satisfies the following equality: Remark 1. It is clear that condition (
15)
is equivalent to the following inequality: Theorem 5. Given the following assumptions:
and ;
The conditions stated in Theorem 1 are satisfied;
For a certain , is bounded above by , where ;
There exists a function such that , where ;
The functions satisfy , , and meet the conditions:
Consequently, we have the following result:which serves as a sufficient condition for probability boundedness. In this section, we finally consider the weight
. Therefore, we define the following spaces:
The semigroup (
7) now exhibits an exponential estimate given by:
where
represents the principal eigenvalue of
. We may conventionally extend the
-Wiener process
to
as
Here,
refers to a separate
-Wiener process that is unrelated to
.
Definition 1. A process , with values in , is considered to be a mild solution of the system (
1)
for if the following conditions hold: Theorem 6. Given a sufficiently small Lipschitz constant δ (see (
31)
for the precise requirement), the system (
1)
has a unique solution designated as , which is specified for . Furthermore, it satisfies the inequality stated asIn addition to this, the solution possesses an exponential attraction property. This implies the existence of positive quantities and ϑ for which the conditionholds true, for any chosen initial conditions and , and for any alternative solution with the initial conditions and . 4. Proofs of Theorem 1
We will prove Theorem 1 in this section.
Let
, where
, be the space of
-measurable processes for all
. It is equipped with the norm
, defined as
multiplied by the integral of
over the interval
. Now, we define the expression
for
as follows:
while setting
for
, with
. Using this definition, we can derive the following inequality:
By (
8), one has that
It follows from conditions (i) and (ii) for
f that
In order to estimate
, we make use of (
10) and [
27], [Lemma 4.1]. By applying the definition of the Sobolev–Schmidt norm as given in (
9), we obtain the following expression:
which is similar to the estimation in (
16). By combining these estimates, we obtain the following expression:
.
To demonstrate the contractive property of ∨, we consider arbitrary
. Let us proceed with the step-by-step calculations.
It follows from (
17) that
As a result, for sufficiently small
, inequalities (
18) and (
19) indicate that the mapping ∨ possesses a distinct fixed point within
, which corresponds to the solution sought in (
11). Moreover, if we consider the problem over intervals
with
, the continuity of the solution, almost surely, in the
norm guarantees both the existence and uniqueness of the solution over the interval
.
The demonstration of estimate (
12) is the final remaining task. By referencing Equation (
11), we can deduce that, for any
, the following holds:
We consider two distinct scenarios:
and
. For the case where
, the following inequality (
21) holds:
In the case where
, we have the inequality (
22) as follows:
Combining (
20)–(
22), one has that
By treating the final term as a distinct entity, it can be deduced that
Combining the aforementioned estimations, it follows that
thereby concluding the proof.
6. Proofs of Theorem 4
To establish the validity of Theorem 4, it is necessary to utilize several auxiliary lemmas.
Lemma 2. The operatoris a Sobolev–Schmidt operator for every fixed . Proof. By [
28], p. 91, an orthonormal basis
exists in
satisfying
. It can be easily verified that if
is an orthonormal basis in
, where
, then
is an orthonormal basis in
. Therefore, we have
However,
So
Hence, the proof is complete. □
Corollary 1. Based on the proof of Lemma 2, it can be demonstrated that constitutes a compact operator when mapping from to , where .
It follows from the method presented in [
22], p. 311, that we can express Equations (
23) and (
24) as follows:
Equation (
23) represents the solution
at a prescribed time
in terms of the initial condition, the inhomogeneous term
, and a perturbation involving
against the measure
. The operator
functions as an evolution operator (or fundamental solution) propagating both the initial data and the cumulative effects of forcing and perturbations. Equation (
24) generalizes this representation to the shifted time
, thereby yielding an iterative form of the variation-of-constants formula. In essence, it reveals how the solution develops beyond
, once again expressed through initial data transported by
and the nonlinear terms accumulated over time. These formulas constitute modifications of the classical variation-of-constants formula (Duhamel’s principle), adapted here to the abstract setting involving delay or measure-driven terms. Their significance lies in offering explicit representations of solutions in terms of initial states and forcing, which form the basis for compactness and fixed-point arguments.
It is possible to immediately apply the justifications in [
27], [Theorem 11.29], to Equation (
23).
Lemma 3. Under the condition that and , the operator can be regarded as a compact mapping. Specifically, it maps from to and is defined as Remark 2. The observed compactness in suggests a comparable compactness in .
Proof of Lemma 3. Define
We will utilize the Phragmén–Lindelöf theorem [
15] in its infinite dimensional version. In order to demonstrate this, we must establish the following:
- (i)
For any fixed value of in a range , the set is compact within ;
- (ii)
There exists a positive constant
such that if
such that if
and for all
with
, where one has that
To verify (i), assuming a fixed
in the range
and
, we introduce
It is obvious that
belongs to
. By employing Corollary 1, we can ascertain that
is a compact operator from
to
. Based on the approach outlined in [
23], p. 137,
converges strongly to
as
approaches 0. Consequently,
is compact, that is to say, (i) is satisfied.
To establish (ii), we fix
and
in such a manner that
, and
. Then, one has that
Through direct calculations, we obtain
Since
is compact, it is also strongly continuous for
. Therefore,
, where
. Moreover, the integrand in
is bounded by
. Thus, by utilizing the dominated convergence theorem, one has that
as
, thereby concluding the proof of the lemma. □
Consider any positive value of
and introduce
such that
with
,
and
. Based on Lemma 2, Corollary 1, and Lemma 3, it can be concluded that
is compact in
.
Lemma 4. According to Theorem 1’s conditions, there exists a positive constant c such that for arbitrary and with , the inequalityholds true, where and . Proof. The factorization formula can be represented as follows:
It follows from Theorem 1’s conditions (
25), (
26),and (
27) that
It follows from (
11) and Hausdorff–Young’s inequality that
Similarly,
Hence, if
,
, and
then it follows from the definition of
that
. Suppose that
. It follows from (
28) and (
29) that
We complete the proof. □
The remaining part of the proof for Theorem 4 can be derived in a similar manner as demonstrated in [
24], [Theorem 11.29].