Abstract
We investigate the existence of a weak martingale solution for a two-dimensional critical viscoelastic flow of the Oldroyd type driven by pure jump Lévy noise. Due to the viscoelastic nature, noise in the equation modeling stress tensor is considered in the Marcus canonical form. Owing to the lack of dissipation and taking into account of the structure of the non-linear terms, the proof requires higher order estimates.
1. Introduction
1.1. Description of the deterministic model
It is well-known that the full classical Oldroyd type models for viscoelastic fluids (see Oldroyd [Citation1]) in can be written as (1.1a) (1.1a) (1.1b) (1.1b) (1.1c) (1.1c) (1.1d) (1.1d) where is the velocity vector field which is assumed to be divergence free, is the non-Newtonian part of the stress tensor (i.e., is a (2, 2) symmetric matrix), p is the pressure of the fluid, which is an unknown scalar, i.e., part of the solution. The parameters ν (the viscosity of the fluid), a (the reciprocal of the relaxation time), ν1 and ν2 (determined by the dynamical viscosity of the fluid, the retardation time and a) are assumed to be non-negative. is called the deformation tensor and is the symmetric part of the velocity gradient is a quadratic form in and for Oldroyd fluids one usually chooses where is a constant and is the vorticity tensor, and is the skew-symmetric part of velocity gradient. We impose an idealized condition (or far-field condition) on the fluid, i.e.,
The case κ = 0 in Equation(1.1)(1.1a) (1.1a) is a classical issue and analysis of the model under this situation is quiet challenging. For more review of such problems, we refer to [Citation2, Citation3] and the references cited therein. Authors in [Citation4] have studied the two dimensional model when there is a lack of dissipation in the velocity equation only, i.e., when and when has a special structure (that is, in the corotational case). As an improvization of [Citation4], the authors in [Citation5] have exploited the method of transferring dissipation and have obtained the existence of a unique global solution in three dimensional space provided the initial data is small in the H Sobolev-norm with It is worth noting that even when the coupling is critical with respect to the smoothing effect provided by the partial parabolic regularization (see, for instance, [Citation4] for details). When the diffusion is present in both the equations i.e., the problem is subcritical in dimension two, and global well-posedness of strong solutions can be obtained in [Citation6] due to the smoothing effect provided by the fully parabolic system. One can clearly observe that Equation(1.1a)(1.1a) (1.1a) is similar to Navier-Stokes equations (NSE) if We have fairly good understanding of two dimensional NSE in both deterministic and stochastic set-ups. Although in three dimensional NSE settings, there are still open questions on global solvability and only partial results are available in this direction. In this regard, we restrict ourselves to the two dimensional problem only.
Let us now briefly describe the connection between some deterministic systems with critical coupling which are close to our model. In particular, a critical coupling for the Boussinesq system has been studied in Hmidi et al. [Citation7, Citation8], Manna and Panda [Citation9], for the MHD system in Caflisch et al. [Citation10], for the liquid crystal model in Lin et al. [Citation11], to name a few. Many other interesting results are devoted to the Oldroyd-B model and related models and are available in [Citation12–16] and the references therein.
Stochastic analysis of the above critical coupled systems influenced by random forcing are very limited and quite recent (e.g. see Yamazaki [Citation17] for Boussinesq system with zero dissipation, Manna et al. [Citation18] for the non-resistive MHD system, Brzeźniak et al. [Citation19] for the liquid crystal model). Existence and asymptotic behavior of a linear visoelastic fluid equation perturbed by additive or multiplicative Wiener processes has been studied, in Barbu et al. [Citation20] and Razafimandimby [Citation21].
In this context, it is worthy to mention that some recent works have been devoted to understand the fluid behaviors in the micro-macro regimes. In order to build a micro-macro model, one needs to go down to the microscopic scale and make use of the kinetic theory to obtain a mathematical model for the evolution of the microstructures of the fluid (e.g. configurations of the polymer chains in the case of polymeric fluid). For a quick survey and more details, we refer to our earlier paper [Citation2, Citation3] and references therein.
1.2. The stochastic model
In the present work, we are interested in the mathematical analysis of a stochastic version of (1.1a)–(1.1d). Our work is motivated by the importance of external perturbation on the dynamics of the velocity field for fluids with memory. The viscoelastic property demands that the material must return to its original shape after any deforming external force has been removed (i.e., it will show an elastic response) even though it may take time to do so. Hence the equation modeling stress tensor (i.e. the equation for ) is invariant under coordinate transformation. Therefore the question of how to incorporate a suitable perturbation modeling the stress tensor without destroying its invariance property is a delicate one.
Hence for a complete knowledge of the effect of fluctuating forcing field on the behavior of the viscoelastic fluids, one needs to take into account the dynamics of and To initiate this kind of investigation, we propose a mathematical study of the following system of equations which basically describes an approximation of the system governing the viscoelastic fluids under the influence of fluctuating external forces.
1.2.1. Physical motivation for the choice of noise
In continuum mechanics, it is well-known that in a solid material, the elastic component of the stress can be attributed to the deformation of the bonds between the tiniest particles of the material. As a subject of interest, in a fluid, the elastic stress can be assigned to the change in the mean spacing of the particles. This phenomena in the fluid affects the particles’ interaction or collision rate and hence the transfer of momentum across the fluid takes place. Therefore, it is related to the microscopic thermal random component of the particles’ motion, which motivates us to establish more realistic models by incorporating stochastic processes. The randomness leads to a variate of new phenomena and may have highly non-trivial impact on the behavior of the solution. As per the existing literature, the viscous component of the stress arises from the macroscopic mean velocity of the particles. It can be attributed to friction or particle diffusion between adjacent parcels of the medium that have different mean velocities, which inspires us to incorporate noise (both continuous and jump) for both the equations (see [Citation22]) for further details.
1.3. Model problem
In this work we consider the following two dimensional stochastic viscoelastic equation ( and ) with pure jump noise (in Itô-Lévy sense for the velocity equation and Marcus canonical sense for the stress tensor equation) (1.2a) (1.2a) (1.2b) (1.2b) (1.2c) (1.2c) Each is a bounded function in is a -valued pure jump Lévy process with the intensity measure λ2 such that where i.e., where represents time homogeneous compensated Poisson random measure. Here represents time homogeneous compensated Poisson random measure on a certain measurable space Precise definition of will be stated later. The tensor product ⊗ denotes usual matrix multiplication. Properties of F have been stated in Assumption 2.3.
1.3.1. Why Marcus type noise?
The choice of the noise modeling the stress tensor for the Stratonovich type noise (for continuous noise) and Lévy noise (for jump noise) in the Marcus canonical form are due to its special property of invariance under diffeomorphism. We refer to our earlier paper [Citation2] where the EquationEquation (1.1a)(1.1a) (1.1a) for velocity field is perturbed by a very general noise of multiplicative Lévy type, whereas EquationEquation (1.1b)(1.1b) (1.1b) for the stress tensor is perturbed by a Wiener noise in the Stratonovich form. However, interested readers may look into the recent manuscript [Citation23] of the first author and his collaborator, where for the first time a certain kind of SPDE driven by Lévy noise in the Marcus canonical form has been studied. Such invariance property is known in the literature for Stratonovich integrals (see e.g. Brzeźniak and Elworthy [Citation24, Theorem 3.3 and Proposition 4.4]). However, one needs an analogue of the Stratonovich integral in the case of stochastic integral with respect to compensated Poisson random measure. The work of Marcus [Citation25], developed later by Applebaum and Kunita, (see e.g. Section 6.10 of [Citation26] and [Citation27]; see also [Citation28]) provides a framework to resolve this technical issue. In other words, one needs a result of the following type (which is framed in SDE set-up with notations as above). Authors of the current paper have studied the critically coupled system in [Citation2] (when ) under the influence of a Wiener process in Stratonovich form and in [Citation3] in presence of Marcus noise (when ) in
1.3.2. Novelty of our work
We construct a martingale solution of the stochastic critical Oldroyd-B system Equation(1.2)(1.2a) (1.2a) with zero dissipation (in the velocity equation) and a convection diffusion (in the equation for ) in dimension two. The proof is based on a variant of the Faedo-Galerkin method. Furthermore, we use tightness criteria (see Appendix A.2 of the current manuscript and the references cited therein) in certain spaces of cádlág functions and Jakubowski’s generalization of the Skorokhod-Theorem to nonmetric spaces; we refer to [Citation29, Citation30], Proposition 9 in [Citation31], Proposition 3.7 in Step IV of the current manuscript.
Since the original system Equation(3.1)(3.1a) (3.1a) lacks dissipation, it behaves more like an incompressible Euler equation. Hence mere -estimate on the velocity is not enough to talk about well-posedness of such problems and one needs to consider more regular initial data to obtain an enstrophy estimate. Besides, the rigorous analysis involved in this paper owing to the critical coupling of the random system in presence of pure jump noise is quiet technical and delicate compared to the deterministic ones, see [Citation4, Citation5].
Proposition 1.1.
Let L(t) be a -valued pure jump Lévy process with the intensity measure λ such that i.e., where represents time homogeneous compensated Poisson random measure. Let be complete C-vector fields. Define such that . Consider the following “Marcus” stochastic differential equation (1.3) (1.3) which is defined in the integral form as follows where solves Suppose, Y, an -valued process, is a solution to Equation(1.3)(1.3) (1.3) . Let be a C diffeomorphism. Define for each , the “Push-forward” of the vector fields and by as Let be such that
Then Y is a solution to Equation(1.3)(1.3) (1.3) if and only if is a solution to
Proof of the above result can be borrowed from [Citation32, Theorem B.1], where Brzeźniak and the first named author of this article have obtained the result for the pure jump case.
Let us briefly describe the content of the paper. In Section 2, we present detailed description of the functional spaces, the Marcus map and assumptions of this paper. In Section 3, we discuss statement of the main result and its proof.
Appendix A is split into two parts. The very first subsection has been presented by series of technical lemmas which are useful during the course of analysis made throughout the paper. The second subsection has been devoted to certain useful compactness and tightness criterion for cádlag functions and predictable processes.
2. Functional settings, assumptions and the Marcus map
We denote as the space of all valued compactly supported functions in We define and be the closure of in and respectively. For let Vs and be the closure of and in and respectively. It is clear from the definitions that and
Notation 2.1.
Throughout this paper, we denote as the dual of X where X is a Hilbert space and C as a generic constant.
We note the following:
H is a Hilbert space with scalar product and V is a Hilbert space with scalar product
is a Hilbert space with scalar product and is a Hilbert space with scalar product
2.1. Trilinear operators b and
We define the trilinear operators and by with and Let us define bilinear maps and by and Then we can show that and are linear and continuous.
2.2. Some additional operators and notations
If then by the Sobolev embedding theorem where denotes the space of continuous and bounded valued functions defined on We have the following conclusions: (2.1) (2.1) The following embeddings are continuous and dense. By Lemma C.1 in Appendix C of [Citation23], there exists a Hilbert space such that and (2.2) (2.2) Combining Equation(2.1)(2.1) (2.1) and Equation(2.2)(2.2) (2.2) we have Let us consider the adjoint of the following maps denoted by respectively and are defined as: Since all the embeddings are dense so are one-one. Let us define the map Note that i is compact and range of i is dense in H. We consider the adjoint of i defined as which is one-one. In the similar manner, one can have and Let us define by In the similar manner, we can define and corresponding to maps and respectively. We note that and Note that for each using the continuity of one can conclude existence of an element such that Similarly we can define by
Remark 1.
In view of [Citation23], one can observe that both are self adjoint and are compact.
Lemma 2.2.
For the following results hold:
For proof see [Citation23].
Let be a measurable space and let be the set of all valued measures on where On the set we consider the field defined as the smallest field for all the map is measurable.
Let be a measurable space. Let N1 on over a filtered probability space be the time homogeneous Poisson random measure (for definition, see e.g. [Citation33]), which is known to be a measurable function The formula defines a measure on called an intensity measure of N1. Moreover, for all and all such that the -valued process is an integrable martingale on The random measure on is called the compensator of N1 and is called the compensated time homogeneous Poisson random measure.
Similarly, one can define the time homogeneous Poisson random measure N2.
Assumption 2.3.
Below we state the assumptions made throughout this paper.
(A.1) is a filtered probability space, where is the filtration, satisfying the usual conditions, i.e.,
is complete on
for each contains all -null sets,
the filtration is right-continuous.
(A.2) N1 is a time homogeneous Poisson random measure on a measurable space over the above probability space with intensity measure λ1 (and compensator ).
(A.3) is a -valued, -adapted Lévy process of pure jump type defined on the above probability space with drift 0 and the corresponding time homogenous Poisson random measure N2.
(A.4) The intensity measure λ2 of N2 (with compensator ) is such that and where
(A.5) N1 and N2 are mutually independent.
(A.6) for each
(A.7) is a measurable function such that there exists a positive constant L such that (2.3) (2.3) and for each there exists a positive constant Cp such that (2.4) (2.4)
(A.8) is a measurable function such that there exists a positive constant Cp, depending upon such that
Assumption (A.8) is required for the gradient and curl estimates in Lemma 3.4 and Corollary 3.5.
2.3. The Marcus map
For general discussion about this topic, we refer to the Appendix A. For we consider the bounded linear map It is easy to observe that is also a continuous linear map. We define a generalized Marcus mapping such that for each fixed the function is the continuously differentiable solution of the ordinary differential equation Therefore, we can write
Notation.
Let us fix t = 1 and denote
EquationEquation (1.2)(1.2a) (1.2a) with notation is defined in the integral form as following: (2.5) (2.5) where is a -measurable random variable.
For we denote With these above notations, Equation(2.5)(2.5) (2.5) can be written as: Now, define a linear operator Then
Thus, if we denote then y satisfies
Hence
We then have the following useful result. We postpone the proof to Appendix A.
Lemma 2.4.
Let be defined by If then
3. Main result and key ideas of the proof
We now consider the following stochastic critical Oldroyd-B system (with zero dissipation in the velocity equation and a convection diffusion equation for ) in dimension two (3.1a) (3.1a) (3.1b) (3.1b) (3.1c) (3.1c) We first introduce the following functional spaces endowed with the respective topologies: with the topology induced by the Skorokhod metric with the topology generated by the seminorms where BR is a closed ball of radius R; with the weakest topology such that for all the mappings are continuous. In particular, in iff for all Similarly, one can define the spaces and
Now let us consider the spaces defined by (3.2) (3.2) and let be the supremumFootnote1 of the corresponding topologies in the spaces Let and be the product topology on
We now recall the definition of a weak martingale solution.
Definition 3.1.
A weak martingale solution of Equation(3.1)(3.1a) (3.1a) is a system where
is a filtered probability space with a filtration
is a time homogeneous Poisson random measure on over with the intensity measure λ1 and is a time homogeneous Poisson random measure on with the intensity measure λ2,
and
is an -progressively measurable process with -a.e. paths such that for all and all and satisfies the inequality where denotes the expectation with respect to
is an -progressively measurable process with -a.e. paths such that for all and all and satisfies the inequality
We now present our main result.
Theorem 3.1.
For the sake of brevity, we take Let Assumption 2.3 hold and let and . Then there exists a weak martingale solution to the system Equation(3.1)(3.1a) (3.1a) satisfying
3.1. Main ideas of the proof
Due to lack of dissipation in the velocity equation Equation(3.1)(3.1a) (3.1a) , we first incorporate an artificial dissipative term. In other words, for every ν0 being fixed, we consider the following approximating Oldroyd-B system: (3.3a) (3.3a) (3.3b) (3.3b) (3.3c) (3.3c) We now consider an Faedo-Galerkin approximation of the above system and let the Galerkin solution be Our strategy is the following: we prove that the system is a weak martingale solution to Equation(3.1)(3.1a) (3.1a) , which is obtained as a limit (as or ) of a random variable on the new probability space and this random variable has the same laws as the random variable on the original probability space The proof is split into a few steps.
Firstly, our aim is to have uniform energy bounds for the Galerkin solution which are also valid bounds for the Galerkin solution of the original system Equation(3.1)(3.1a) (3.1a) . Therefore the bounds need to be free from the choice of ν (and n). Toward this, we obtain a basic energy estimate for the Galerkin solution which reads, for where C depends on the parameter of the problem but is independent of n and Since the original system (3.1) lacks dissipation, it behaves more like an incompressible Euler equation. Hence mere -estimate on the velocity is not enough to talk about well-posedness of such problems and one needs to consider more regular initial data to obtain an enstrophy estimate. Toward this goal our aim is to get the -estimate on the finite dimensional projection on the vorticity, i.e. for Drawing motivation from [Citation4], we define the singular integral operator R, that operates on matrices in the following way: Now applying and R to (3.3), we get for Define where for being fixed.
Using the commutator property and a crucial simplification due to the identity we see that satisfies the following equation for Next, exploiting already obtained good bounds on and properties of R, we find an uniform energy bound in for which further yields Further, with the help of an associated elliptic problem, we prove As a next step, due to the above uniform bounds (which are independent of ν), tightness of the sequence of measures is tight on the space (see Equation(3.20)(3.20) (3.20) ). Next by choosing and employing Skorokhod-Jakubowski theorem, we have a new probability space and random variables on this new probability space, such that -a.s. As a last step, we show that on the new probability space the limiting sequence i.e. the system is a weak martingale solution of the original problem Equation(3.1)(3.1a) (3.1a) .
3.2. Proof of Theorem 3.1
We present the proof of Theorem 3.1 below.
Proof.
We split the proof in few steps.
Step I: (Faedo-Galerkin Approximation and Energy Estimate). By Remark 1, there exist orthonormal bases of respectively consist of eigenvectors of operators Let be the corresponding eigenvalues of and respectively. Therefore we have and Let us denote Let be defined as by For and let us consider the following maps:
Let us denote and
Let be a flow on assosciated to the vector field i.e., For we define Consider the following mappings where is defined by where of class such that Since and the mappings and are well defined. Moreover, and are globally Lipschitz continuous.
We now consider the following -valued modified approximated system: (3.4a) (3.4a) (3.4b) (3.4b) Since all relevant maps are globally Lipschitz (see Lemma A.1–A.3 in Appendix A), we have the following standard result, see e.g. [Citation34] for a reference.
Lemma 3.2.
For each there exists a unique global, -progressively measurable, -valued càdlàg process satisfying the modified Galerkin approximated system Equation(3.4)(3.4a) (3.4a) .
It follows easily, see for instance [Citation34], that for each the approximated system Equation(3.4)(3.4a) (3.4a) has a unique local maximal solution. In two dimensional setting, by a combination of the proof of [Citation34, Theorem 3.1] with Lemma 3.3 below, we infer that the approximated system Equation(3.4)(3.4a) (3.4a) has a unique global solution.
Step II: (Estimates for the velocity field and stress tensor)
Lemma 3.3.
Let Let and Then for every T > 0, there exists depending on (but is independent of n and ν), such that (3.5) (3.5)
Proof.
Estimate Equation(3.5)(3.5) (3.5) can be proved by applying Itô’s Lemma to the function (for the process ) and (for the process ) to the system Equation(3.4)(3.4a) (3.4a) . □
Step III: (Estimate of the vorticity)
Lemma 3.4.
Let and Then for every T > 0 there exists a constant depending on (but is independent of n and ν), such that (3.6) (3.6) where and is the finite dimensional approximation of
Proof.
Let us recall that R denote the singular integral operator that operates on matrices in the following way: We note that R is a degree 0 singular integral operator having the following properties (see [Citation4]):
R is a bounded linear transformation in
For divergence free we have (3.7) (3.7)
Now applying and R to Equation(3.4)(3.4a) (3.4a) , we get for Define
Let Then using properties of R, we see that satisfies (3.8) (3.8) Exploiting the properties of R and Young’s inequality, we now estimate the following two terms which lead to (3.9) (3.9) (3.10) (3.10) For fixed, let us define a stopping time by Therefore applying Itô’s formula to the function for the EquationEquation (3.8)(3.8) (3.8) , using Equation(3.9)(3.9) (3.9) , Equation(3.10)(3.10) (3.10) then integrating in and then taking supremum and expectation, we get (3.11) (3.11) where (3.12a) (3.12a) (3.12b) (3.12b) (3.12c) (3.12c) (3.12d) (3.12d) Using Burkholder-Davis-Gundy inequality, (A.8) in Assumption 2.3, and Young’s inequality, we get (3.13) (3.13) In order to estimate M2 term, we apply the boundedness property of R (see Equation(3.7)(3.7) (3.7) ) and part 1 of Lemma 2.4, and obtain (3.14) (3.14) Thus using Equation(3.14)(3.14) (3.14) , (A.4) in Assumption 2.3 and bounds on we have, (3.15) (3.15) Likewise, using (A.8) in Assumption 2.3, Lemma 3.3 and as we have (3.16) (3.16) and employing Lemma 2.4, (A.4) in Assumption 2.3, boundedness property of R and bounds on we achieve (3.17) (3.17) Finally we compute the third term on the left hand side of the inequality Equation(3.11)(3.11) (3.11) to obtain, (3.18) (3.18) which is finite because of Lemma 3.3 and Equation(3.5)(3.5) (3.5) .
Now substituting Equation(3.13)(3.13) (3.13) , Equation(3.15)(3.15) (3.15) –Equation(3.18)(3.18) (3.18) in Equation(3.11)(3.11) (3.11) , we have Using Gronwall’s inequality and noting that as we finally achieve Finally, using the definition of we have Note that, as for a fixed ν0, we have by the definition of that Hence the right hand side of the above inequality can be further made less than some constant, which is independent of n and ν. □
Corollary 3.5.
Under the above mathematical settings, there exists a constant C > 0, which depends on the parameters of the problem but is independent of n and ν, such that
Proof.
Following Lemma 3.2 of [Citation35], let us consider the following elliptic problem: (3.19) (3.19) where
Multiplying Equation(3.19)(3.19) (3.19) by and using integration by parts on we have This directly implies Then taking expectation we have by Equation(3.6)(3.6) (3.6) This completes the proof. □
Step IV: (Tightness)
Let the spaces defined by (3.20) (3.20) and let be the supremum of the corresponding topologies in the spaces Let and be the supremum of and We note that the solutions obtained from Galerkin approximation induces sequence of measures on the space Then we have the following result:
Lemma 3.6.
(Tightness of law) The sequence of measures is tight on the space
Proof.
We observe from Step II that there exists a constant such that for all (3.21a) (3.21a) (3.21b) (3.21b) Therefore from Equation(3.21)(3.21a) (3.21a) we conclude that first two conditions of Theorem A.9 for are satisfied. Now it is sufficient to prove that and satisfy Aldous condition in the space and respectively. We start with the former sequence. We will use Lemma A.7. Let be a sequence of stopping times such that
Let be a sequence of stopping times such that Then for we get We will show that each of satisfies condition Equation(A.27)(A.27) (A.27) in Lemma A.7. Let us fix being independent of time clearly satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7. Now exploiting the fact given by the embedding is continuous, then by Hölder’s inequality and Equation(3.21)(3.21a) (3.21a) , can be estimated as Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using is continuous and the continuous embedding and Equation(3.21)(3.21a) (3.21a) , we estimate as Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using Equation(3.21)(3.21a) (3.21a) and the embedding we have Therefore, satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using Equation(3.21)(3.21a) (3.21a) , the continuous embedding Itô-Lévy isometry and Assumption 2.3, we have Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 2 and ζ = 1.
Let us consider the equation for for being independent of time, Equation(A.27)(A.27) (A.27) in Lemma A.7 is automatically satisfied for any Now exploiting the fact given by the continuous embedding Hölder’s inequality and Equation(3.21)(3.21a) (3.21a) , can be estimated as Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using and the embedding we have Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using the embedding and Equation(3.21)(3.21a) (3.21a) we have for Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and
Using Itô-Lévy isometry, embedding Lemma A.3 and Equation(3.21)(3.21a) (3.21a) we have Thus satisfies Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 2 and ζ = 1.
Since is continuous, using Schwarz’s inequality, from Lemma A.3 and Equation(3.21)(3.21a) (3.21a) we get, Thus satisfies the condition Equation(A.27)(A.27) (A.27) in Lemma A.7 with α = 1 and This proves that the sequences and satisfy the Aldous condition in the space and respectively. Thus the proof is complete. □
Step V: (Use of Skorokhod Representation Theorem)
In separable metric spaces, one can apply Prokhorov Theorem (see [Citation36], Theorem II.6.7) and classical Skorokhod Representation Theorem (see [Citation37], Theorem 6.7) to obtain a.s. convergence from tightness. However, for our purpose, we need a modified version of Skorokhod Representation Theorem due to Jakubowski [Citation29, Citation30], which can be used for non-metrizable topological space. The version used in this paper is stated below (see also [Citation31], Proposition 9).
Proposition 3.7.
Let be a complete separable metric space and a topological space such that there is a sequence of continuous functions that separates points of . Denote and equip with the topology induced by the canonical projections . Let be a probability space and be a tight sequence of random variables in , where is the algebra generated by . Assume that there is a random variable η in such that . Then there are a subsequence and random variables in for on a common probability space with
for
in almost surely for
almost surely.
Note that, by we denote the set of all -valued Borel measures ξ on with for all for some sequence of Borel sets with and for all It is well known that is a complete separable metric space (see [Citation31]). Similarly, is a complete separable metric spaces.
In the line of notations used in Proposition 3.7, we denote Clearly, is a complete separable metric space. Since any infinite dimensional normed space with respect to weak topology in that space is not metrizable, both and are not metrizable, and so is or
Lemma 3.6 ensures that the sequence of measures is tight on the space Let for Then the set of measures is tight on
We choose By the Skorokhod-Jakubowski theorem (see Proposition 3.7 and also Lemma A.10), there exists a subsequence a filtered probability space with a filtration and on this space, -valued random variables such that
for all
in with probability 1 on as
for all
We will denote these sequences again by and Using the definiton of we have -a.s. (3.22a) (3.22a) (3.22b) (3.22b) It is easy to verify that the spaces are not a Polish spaces. The space is a Polish space. Then by Kuratowski theorem, is a Borel subset of Hence is a Borel subset of which is equal to (see Lemma 4.2 in [Citation38]). Thus we have the following proposition.
Proposition 3.8.
The set is a Borel subset of and the corresponding embedding transforms Borel sets into Borel subsets. A similar result is true for
The above result leads us to the following conclusion:
Corollary 3.9.
and take values in Hn and respectively. The laws of and are equal on and the laws of and are equal on
Thus one may conclude that the limiting processes and satisfy the following estimates for for some constant Cr (depending on r).
Step VI: (Convergence)
Let us define the following functionals for and for all (3.23a) (3.23a) (3.23b) (3.23b) (3.23c) (3.23c) (3.23d) (3.23d) Next, we claim that (3.24a) (3.24a) (3.24b) (3.24b) (3.24c) (3.24c) (3.24d) (3.24d) which is a direct consequence of the following Proposition.
Proposition 3.10.
If are defined by Equation(3.23)(3.23a) (3.23a) , then we have the following convergences: (3.25) (3.25) (3.26) (3.26)
We postpone the proof of the Proposition in Appendix A.1. Hence from Equation(3.24)(3.24a) (3.24a) , we infer that for Leb-almost all all and and -almost all we infer that satisfies Finally, since is dense in V and is dense in we infer that the above two equalities hold for all and all and respectively. This concludes that the system is a weak martingale solution of Equation(3.1)(3.1a) (3.1a) . Thus the proof of Theorem 3.1 is complete. □
Acknowledgements
The first author would like to thank Professor Zdzisław Brzeźniak for many useful discussions on Marcus canonical SDEs. The second author would like to thank Professor Erika Hausenblas for useful discussions and partial support from the FWF-Project P32295.
Disclosure statement
Both the authors declare that they have no conflict of interest.
Additional information
Funding
Notes
1 Supremum topology is the topology generated by the collection i.e., the topology generated from the collection as a subbasis. Let be the supremum of the corresponding topologies in the spaces This supremum topology is different from the subspace topology.
References
- Oldroyd, J.G. (1958). Non-Newtonian effects in steady motion of some idealized elastico-viscous liquids. Proc. Roy. Soc. Lond. Ser. A. 245:278–297.
- Manna, U., Mukherjee, D. (2017). Strong solutions of stochastic models for viscoelastic flows of Oldroyd type. Nonlinear Anal. 165:198–242. DOI: https://doi.org/10.1016/j.na.2017.10.003.
- Manna, U., Mukherjee, D. (2021). Weak solutions and invariant measures of stochastic Oldroyd-B type model driven by jump noise. J. Differ. Equ. 272:760–818. DOI: https://doi.org/10.1016/j.jde.2020.10.009.
- Elgindi, T. M., Rousset, F. (2015). Global regularity for some Oldroyd-B type models. Commun. Pur. Appl. Math. 68(11):2005–2021. DOI: https://doi.org/10.1002/cpa.21563.
- Elgindi, T. M., Liu, J. (2015). Global wellposedness to the generalized Oldroyd type models in R3. J. Differ. Equ. 259(5):1958–1966.
- Constantin, P., Kliegl, M. (2012). Note on global regularity for two-dimensional Oldroyd-B fluids with diffusive stress. Arch. Rational Mech. Anal. 206(3):725–740. DOI: https://doi.org/10.1007/s00205-012-0537-0.
- Hmidi, T., Keraani, S., Rousset, F. (2010). Global well-posedness for a Boussinesq-Navier-Stokes system with critical dissipation. J. Differ. Equ. 249(9):2147–2174. DOI: https://doi.org/10.1016/j.jde.2010.07.008.
- Hmidi, T., Rousset, F. (2011). Global well-posedness for the Euler-Boussinesq system with axisymmetric data. J. Funct. Anal. 260(3):745–796. DOI: https://doi.org/10.1016/j.jfa.2010.10.012.
- Manna, U., Panda, A. A. (2019). Higher order regularity and blow-up criterion for semi-dissipative and ideal Boussinesq equations. J. Math. Phys. 60(4):22.
- Caflisch, R. E., Klapper, I., Steele, G. (1997). Remarks on singularities, dimension and energy dissipation for ideal hydrodynamics and MHD. Commun. Math. Phys. 184(2):443–455. DOI: https://doi.org/10.1007/s002200050067.
- Lin, F.-H., Liu, C., Zhang, P. (2005). On hydrodynamics of viscoelastic fluids. Commun. Pure Appl. Math. 58(11):1437–1471. DOI: https://doi.org/10.1002/cpa.20074.
- Fang, D., Hieber, M., Zi, R. (2013). Global existence results for Oldroyd-B fluids in exterior domains: The case of non-small coupling parameters. Math. Ann. 357(2):687–709. DOI: https://doi.org/10.1007/s00208-013-0914-5.
- Hu, D., Lelievre, T. (2007). New entropy estimates for Oldroyd-B and related models. Commun. Math. Sci. 5(4):909–916. DOI: https://doi.org/10.4310/CMS.2007.v5.n4.a9.
- Lei, Z. (2006). Global existence of classical solutions for some Oldroyd-B model via the incompressible limit. Chin. Ann. Math. Ser. B. 27(5):565–580. DOI: https://doi.org/10.1007/s11401-005-0041-z.
- Lei, Z. (2010). On 2D viscoelasticity with small strain. Arch. Rational Mech. Anal. 198(1):13–37. DOI: https://doi.org/10.1007/s00205-010-0346-2.
- Ye, Z., Xu, X. (2016). Global regularity for the 2D Oldroyd-B model in the corotational case. Math. Meth. Appl. Sci. 39(13):3866–3879. DOI: https://doi.org/10.1002/mma.3834.
- Yamazaki, K. (2016). Global martingale solution for the stochastic Boussinesq system with zero dissipation. Stoch. Anal. Appl. 34(3):404–426. DOI: https://doi.org/10.1080/07362994.2016.1148615.
- Manna, U., Mohan, M. T., Sritharan, S. S. (2017). Stochastic non-resistive Magnetohydrodynamic system with Lévy noise. Random Oper. Stoch. Equ. 25(3):155–194.
- Brzeźniak, Z., Hausenblas, E., Razafimandimby, P. (2014). Stochastic nonparabolic dissipative systems modeling the flow of liquid crystals: Strong solution. Math. Anal. Incompress. Flow. 1875:41–72.
- Barbu, V., Bonaccorsi, S., Tubaro, L. (2014). Existence and asymptotic behavior for hereditary stochastic evolution equations. Appl. Math. Optim. 69(2):273–314. DOI: https://doi.org/10.1007/s00245-013-9224-2.
- Razafimandimby, P. (2010). On stochastic models describing the motions of randomly forced linear viscoelastic fluids. J. Inequal. Appl. 2010(1):932053. DOI: https://doi.org/10.1155/2010/932053.
- Constantin, P. (1997). Some Mathematical Aspects of Incompressible Fluid Mechanics. Partial Differential Equations and Their Applications (Toronto, on, 1995). CRM Proc. Lecture Notes, Vol. 12. Providence, RI: Amer. Math. Soc., pp. 51–69.
- Brzeźniak, Z., Motyl, E. (2013). Existence of a martingale solution of the stochastic Navier-Stokes equations in unbounded 2D and 3D domains. J. Differ. Equ. 254(4):1627–1685. DOI: https://doi.org/10.1016/j.jde.2012.10.009.
- Brzeźniak, Z., Elworthy, K. D. (2000). Stochastic differential equations on Banach manifolds. Methods Funct. Anal. Topol. 6(2):497–558.
- Marcus, S. L. (1981). Modelling and approximations of stochastic differential equations driven by semimartingales. Stochastics. 4(3):223–245. DOI: https://doi.org/10.1080/17442508108833165.
- Applebaum, D. (2009). Lévy Processes and Stochastic Calculus, 2nd ed. Cambridge: Cambridge University Press.
- Kunita, H. (2004). Stochastic Differential Equations Based on Lévy Processes and Stochastic Flows of Diffeomorphisms. Real and Stochastic Analysis. Boston, MA: Trends Math., pp. 305–373.
- Chechkin, A., Pavlyukevich, I. (2014). Marcus versus Stratonovich for systems with jump noise. J. Phys. A. Math. Theor. 47(34):342001. DOI: https://doi.org/10.1088/1751-8113/47/34/342001.
- Jakubowski, A. (1986). On the Skorokhod topology. Annales de l’IHP Probabilités et Statistiques. 22:263–285.
- Jakubowski, A. (1998). The almost sure Skorohod representation for subsequences in nonmetric spaces. Teor.Veroyatnost. i Primenen. 42(1):209–216. Translation in Theory Probab. Appl. (1997) 42(1), 167–174.
- Brzeźniak, Z., Hornung, F., Manna, U. (2020). Weak martingale solutions for the stochastic nonlinear Schrödinger equation driven by pure jump noise. Stoch. PDE. Anal. Comp. 8 (1):1–53. DOI: https://doi.org/10.1007/s40072-019-00141-x.
- Brzeźniak, Z., Manna, U. (2019). Weak solutions of a stochastic Landau-Lifshitz-Gilbert equation driven by pure jump noise. Commun. Math. Phys. 371 (3):1071–1129. DOI: https://doi.org/10.1007/s00220-019-03359-x.
- Motyl, E. (2013). Stochastic Navier–Stokes equations driven by Lévy noise in unbounded 3D domains. Potential Anal. 38(3):863–912.
- Albeverio, S., Brzeźniak, Z., Wu, J.L. (2010). Existence of global solutions and invariant measures for stochastic differential equations driven by Poisson type noise with non-Lipschitz coefficients. J. Math. Anal. Appl. 371(1):309–322. DOI: https://doi.org/10.1016/j.jmaa.2010.05.039.
- Bessaih, H. (1999). Martingale solutions for stochastic Euler equations. Stoch. Anal. Appl. 17(5):713–725. DOI: https://doi.org/10.1080/07362999908809631.
- Parthasarathy, K. R. (1967). Probability Measures on Metric Spaces. New York: Academic Press.
- Billingsley, P. (1999). Convergence of Probability Measures. Wiley Series in Probability and Statistics, 2nd ed. New York: John Wiley & Sons, Inc.
- Brzeźniak, Z., Motyl, E., Ondrejàt, M. (2017). Invariant measure for the stochastic Navier-Stokes equations in unbounded 2D domains. Ann. Probab. 45(5):3145–3201. DOI: https://doi.org/10.1214/16-AOP1133.
- Métivier, M. (1988). Stochastic Partial Differential Equations in Infinite-Dimensional Spaces, with a preface by G. Da Prato. Pisa: Scuola Normale Superiore di Pisa.
- Joffe, A., Métivier, M. (1986). Weak convergence of sequences of semimartingales with applications to multitype branching processes. Adv. Appl. Probab. 18(1):20–65. DOI: https://doi.org/10.2307/1427238.
- Motyl, E. (2014). Stochastic hydrodynamic-type evolution equations driven by Lévy noise in 3D unbounded domains–Abstract framework and applications. Stoch. Process. Appl. 124(6):2052–2097. DOI: https://doi.org/10.1016/j.spa.2014.01.009.
- Mikulevicius, R., Rozovskii, B. L. (2005). Global L2-solutions of stochastic Navier-Stokes equations. Ann. Probab. 33(1):137–176.
Appendix A.
Technical lemmas
We now state some useful results obtained by the first author and his collaborator [Citation32].
Lemma A.1.
There exists such that for any and
There exists such that for any and
There exists such that for any and
There exists such that for any and
Lemma A.2.
There exists a constant such that for any
Lemma A.3.
There exists a constant such that for any (A.1) (A.1)
Below we provide Proof of Lemma 2.4.
Proof of Lemma 2.4.
Proof. Since for all
Note that Hence This proves first part of the Lemma.
We begin with the observation that (A.2) (A.2) Hence as (A.3) (A.3) Similarly, one obtains (A.4) (A.4) Therefore using the estimates Equation(A.3)(A.3) (A.3) and Equation(A.4)(A.4) (A.4) in Equation(A.2)(A.2) (A.2) we have
A.1. Proof of Proposition 3.10
We now see using Fubini’s Theorem that (A.5) (A.5) Lemma A.4 (the following Lemma) ensures that each term on the right hand side of Equation(3.23a)(3.23a) (3.23a) converges to the right hand side of corresponding term in Equation(3.23c)(3.23c) (3.23c) in which further assures that right hand side of Equation(A.5)(A.5) (A.5) goes to zero as This verifies Equation(3.25)(3.25) (3.25) . Lemma A.5 (which is proven below) ensures that each term on the right hand side of Equation(3.23b)(3.23b) (3.23b) converges to the right hand side of corresponding term in Equation(3.23d)(3.23d) (3.23d) in which further verifies Equation(3.26)(3.26) (3.26) . This completes the proof.
Lemma A.4.
For all
Proof.
Let be fixed.
Owing to Equation(3.22)(3.22a) (3.22a) , in particular for almost all Also using and by employing Vitali Theorem we have
Equation Equation(3.22)(3.22a) (3.22a) ensures that which further with Vitali Theorem infer
Owing to Equation(3.22)(3.22a) (3.22a) , we have for any (A.6) (A.6) Let Let be fixed. We choose and note that Hence for using Equation(A.6)(A.6) (A.6) and Hölder’s inequality, we achieve for all and (A.7) (A.7) for some positive constant cr (depending on r). Therefore using Equation(A.7)(A.7) (A.7) and by Vitali Theorem, we conclude for all (A.8) (A.8) Hence Equation(A.7)(A.7) (A.7) , Equation(A.8)(A.8) (A.8) and the Dominated Convergence Theorem proves (iii).
Let We have, Now, using Hölder’s inequality and in we have For every since in Vs and we conclude that for all and all , (A.9) (A.9) Using Hölder’s inequality we obtain for all and (A.10) (A.10) Considering Equation(A.9)(A.9) (A.9) and Equation(A.10) (A.10) (A.10) and by the Vitali Theorem we achieve for all (A.11) (A.11) Hence, in view of Equation(A.10)(A.10) (A.10) , Equation(A.11)(A.11) (A.11) and the Dominated Convergence Theorem, we infer that (A.12) (A.12) Hence we have (iv).
As so Now using Equation(3.22)(3.22a) (3.22a) we have (A.13) (A.13) Let Now using Equation(3.22)(3.22a) (3.22a) we have (A.14) (A.14) for some positive constant Hence by Vitali Theorem, using Equation(A.13)(A.13) (A.13) and Equation(A.14)(A.14) (A.14) we have (v).
Let us consider Using Assumption 2.3 we have (A.15) (A.15) Furthermore, using Assumption 2.3, for every and we have the following inequality (A.16) (A.16) for some constant Since the restriction of Pn to the space H is the -projection onto hence by Equation(A.15)(A.15) (A.15) , Equation(A.16)(A.16) (A.16) and by the Vitali Theorem we infer that for all (A.17) (A.17) Since Equation(A.17)(A.17) (A.17) holds for all Moreover, and Equation(A.17)(A.17) (A.17) and the Dominated Convergence Theorem assures assertion (vi).
Using Itô-Lévy isometry, and the fact that for and Equation(A.17)(A.17) (A.17) and then exploiting Dominated Convergence Theorem we ensure assertion (vii).
Lemma A.5.
For all
Proof.
Proofs of (i)–(iv) are similar to the proofs of Lemma A.4 (i)–(iv), and hence omitted.
(v) Let Now using integration by parts and (3.22) we note that (A.18) (A.18) Let Now using Equation(3.22)(3.22a) (3.22a) we have for some positive constant Cr depending on r. Hence by Vitali Theorem, using Equation(A.18)(A.18) (A.18) we ensure assertion (v).
(vi) Using Lipschitz property of G and Equation(3.22)(3.22a) (3.22a) , we obtain for all for all -a.s. (A.19) (A.19) Moreover, for every every and every (A.20) (A.20) Since the restriction of to the space is the -projection onto then by Equation(A.19)(A.19) (A.19) , Equation(A.20)(A.20) (A.20) and by Vitali’s Theorem, we obtain (A.21) (A.21) Since Equation(A.21)(A.21) (A.21) holds for all
As for all From Equation(A.21)(A.21) (A.21) and the Itô isometry we have, (A.22) (A.22) Moreover, from Equation(A.20)(A.20) (A.20) and the Itô isometry, with we obtain, (A.23) (A.23) Finally, from Equation(A.22)(A.22) (A.22) , Equation(A.23)(A.23) (A.23) and using Dominated Convergence Theorem we obtain (vi).
(vii) Since, in exploiting the Lipschitz property of b, we obtain, (A.24) (A.24) Now using Lemma A.1, for we get, which further with Equation(A.24)(A.24) (A.24) , and Vitali’s theorem we get, (A.25) (A.25) Finally from Equation(A.25)(A.25) (A.25) and Dominated convergence theorem we get
A.2. Compactness and tightness criteria
Let be a complete separable metric space with a metric ρ. Let us denote by the set of all -valued functions defined on which are right continuous and have left limits (càdlàg functions) for every The space is endowed with the Skorokhod J-topology. For more details see Métivier’s [Citation39, Chapter II] and Billingsley’s [Citation37, Chapter 3].
A.2.1. The Aldous condition
Let be a complete, separable metric space. Let be a probability space with usual hypotheses.
Definition A.1.
Let and let be given. A modulus of continuity is defined by (A.26) (A.26) where is the set of all increasing sequences with the following property
Definition A.2.
Let be a sequence of -valued random variables. The sequence of laws of these processes is Tight if and only if
[T]
Definition A.3.
A sequence of -valued random variables satisfies the Aldous condtion if and only if
[A] such that for every sequence of -stopping times with one has
Lemma A.6.
Condition [A] implies condition [T].
Proof.
See Theorem 2.2.2 of [Citation40]. □
Lemma A.7.
Let be a separable Banach space and let be a sequence of E-valued random variables. Assume that for every sequence of -stopping times with and for every and the following condition holds (A.27) (A.27) for some and some constant Then the sequence satisfies the Aldous condtion in the space E.
Proof.
See [Citation33] for the proof. □
Theorem A.8.
(Compactness criteria for ) Let and be the supremum of the corresponding topologies. Let and be the supremum of the corresponding topologies. Define and as the supremum of and Then is relatively compact if the following three conditions are satisfied:
and all and
i.e. is bounded in
Proof of Theorem A.8 can be found in Lemma 3.3 in Brzeźniak and Motyl [Citation23], Lemma 4.1 in Motyl [Citation41], Theorem 2 of Motyl [Citation33], Lemma 2.7 in Mikulevicius and Rozovskii [Citation42].
Theorem A.9.
Let be a sequence of càdlàg -adapted -valued processes such that
there exists a positive constant C1 such that
there exists a positive constant C2 such that
(respectively ) satisfies the Aldous condition in (respectively ).
Let be the law of on . Then, for every , there exist compact subset of such that for i = 1, 2 and the sequence of measures is said to be tight on
For a relevant proof see Corollary 1, Motyl [Citation33].
Lemma A.10.
Let be a sequence of adapted -valued processes satisfying the Aldous condition in and Then, there are a subsequence and random variables and for on a common probability space with and for and -almost surely in for