Abstract
Röckner and Zhang (Probab Theory Relat Fields 145, 211–267, 2009) proved the existence of a unique strong solution to a stochastic tamed 3D Navier–Stokes equation in the whole space and for the periodic boundary case using a result from Stroock and Varadhan (Multidimensional diffusion processes, Springer, Berlin, 1979). In the latter case, they also proved the existence of an invariant measure. In this paper, we improve their results (but for a slightly simplified system) using a self-contained approach. In particular, we generalise their result about an estimate on the \(L^4\)-norm of the solution from the torus to \({\mathbb {R}}^3\), see Lemma 5.1 and thus establish the existence of an invariant measure on \({\mathbb {R}}^3\) for a time-homogeneous damped tamed 3D Navier–Stokes equation, given by (6.1).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In the present paper we study the stochastic tamed Navier–Stokes equations (NSEs) on \({\mathbb {R}}^3\) which were introduced by Röckner and Zhang [30, 31]. We consider the following stochastic tamed NSEs with viscosity \(\nu \), \((t,x) \in [0,T] \times {{\mathbb {R}}^3}\)
subject to the incompressibility condition
and the initial condition
where p(t, x) and \({\widetilde{p}}_j(t,x)\) are unknown scalar functions.
We assume that the taming function \(g : {\mathbb {R}}_+ \rightarrow {\mathbb {R}}_+\) is smooth and satisfies, for some \(N \in {\mathbb {N}}\), the following condition
Finally we assume that \(\{W_t^j; t \ge 0, j = 1 ,2, \dots \}\) is a sequence of independent one-dimensional standard \({\mathbb {F}} = ({\mathcal {F}}_t)_{t \ge 0}\)-Brownian motions on a complete filtered probability space \((\Omega , {\mathcal {F}}, {\mathbb {F}}, {\mathbb {P}})\). The stochastic integral is understood as the Itô integral. The drift and diffusion coefficients are given as follows:
where \(\ell ^2\) denotes the standard Hilbert space consisting of all square summable sequences of real numbers endowed with standard norm \(\Vert \cdot \Vert _{\ell ^2}\). In the following f and \( \sigma \) are always assumed to be measurable with respect to all their variables.
In the case of classical deterministic Navier–Stokes equations on \({\mathbb {R}}^3\), if the initial data \(u_0 \in \mathrm {V}\) (see Sect. 2), then there exists only a local strong solution, see [35]. Cai and Jiu [11] studied the Navier–Stokes equations with damping on \({\mathbb {R}}^3\), where the damping was modeled by the term \(|u|^{\beta - 1} u\), \(\beta \ge 1\); the tamed term considered in the current paper corresponds to \(\beta = 3\). They proved the existence of a global weak solutionFootnote 1 for any \(\beta \ge 1\) and \(u_0 \in \mathrm {H}\), see Sect. 2 and they proved the existence of a global strong solutionFootnote 2 for \(\beta \ge 7/2\) and \(u_0 \in \mathrm {V}\cap L^{\beta +1}\). Moreover, they were able to show the uniqueness of strong solutions for \(7/2 \le \beta \le 5\). Later, Zhang et al. [40], by exploiting the Gagliardo-Nirenberg inequality, were able to lower down the parameter \(\beta \) to 3. Thus, establishing the existence of a global strong solution to Navier–Stokes equations with damping on \({\mathbb {R}}^3\) for \(\beta > 3\), \(u_0 \in \mathrm {V}\cap L^{\beta +1}\) and proving uniqueness whenever \(3 < \beta \le 5\). They also remarked that the critical value for \(\beta \) is \(\beta = 3\) [40, Remark 3.1]. But, Zhou [41, Theorem 2.1] was able to surpass this critical value of \(\beta \).
Moreover, for any \(\beta \ge 1\), he proved that the strong solution is unique in a larger class of weak solutions, see [41, Theorem 3.1]. The critical case of \(\beta = 3\) was studied by Röckner and Zhang [31, Theorem 1.1], where they proved the existence of a smooth unique global solution to the deterministic tamed 3D Navier–Stokes equations for very smooth initial data and deterministic forcing f. Moreover, they proved [31, Theorem 1.1] that this unique solution converges (in \(L^2(0,T; L^2({\mathcal {O}}))\)) to a bounded Leray-Hopf solution of 3D Navier–Stokes equations (if exists) on a bounded domain \({\mathcal {O}} \subset {\mathbb {R}}^3\). The non-explosion of the solution is due to the tamed term. Röckner and Zhang [32] also studied 3D tamed Navier–Stokes equations on a bounded domain \({\mathcal {O}} \subset {\mathbb {R}}^3\) with Dirichlet boundary conditions and proved the existence of a unique strong solution directly, based on the Galerkin approximation and on a kind of local monotonicity of the coefficients. Recently, You [39] proved the existence of a random attractor for the 3D damped (\(|u|^{\beta -2} u\)) Navier–Stokes equations with additive noise for \(4 < \beta \le 6\) with initial data \(u_0 \in \mathrm {V}\) on a bounded domain \({\mathcal {O}} \subset {\mathbb {R}}^3\) with smooth boundary.
Röckner and Zhang [30] proved the existence of a strong solution of the stochastic tamed NSEs (in probabilisitc sense) by invoking the Yamada-Watanabe theorem, thus proving the existence of a martingale solution to (1.1) (with more generalised noise term) in the absence of compact Sobolev embeddings and the pathwise uniqueness. They used the localization method to prove the tightness, a method introduced by Stroock and Varadhan [34]. In this paper, we present a self-contained proof of the same result. In order to prove the existence of a martingale solution, Röckner et al. used the Faedo-Galerkin approximation with the non-classical finite dimensional space \(H_n^1 = \text {span}\{e_i, i = 1 \cdots n\}\) where \({\mathcal {E}} = \{e_i\}_{i \in {\mathbb {N}}} \subset {\mathcal {V}}\) (see Sect. 2) is the orthonormal basis of \(H^1\). They also require that in the case of the periodic boundary conditions, \({\mathcal {E}}\) is an orthogonal basis of \(H^0\) which was essential in obtaining the \(L^4\)-estimate of the solution. We generalised this result to \({\mathbb {R}}^3\). Another reason for Röckner et al. to choose the periodic boundary conditions is the compactness of \(H^2 \hookrightarrow H^1\) embedding, which along with the \(L^4\)-estimate of the solution was crucial in establishing the existence of invariant measures. We don’t require this embedding and hence are able to obtain the existence of invariant measures for the damped tamed Navier–Stokes equations on \({\mathbb {R}}^3\).
In the present paper we prove the existence of a unique strong solution to the stochastic tamed 3D Navier–Stokes equation (1.1) under some natural assumptions \((\mathbf{H}1 )\)–\((\mathbf{H}2 )\) on the drift f and the diffusion \(\sigma \) (see Sect. 2). To prove the existence of strong solution we use the Yamada-Watanabe theorem [27, 38] which states that the existence of martingale solutions plus pathwise uniqueness implies the existence of a unique strong solution. In order to establish the existence of martingale solutions, instead of using the standard Faedo-Galerkin approximations we use a different approach motivated from [15, 20]. We study a truncated SPDE on an infinite dimensional space \(\mathrm {H}_n\), defined in the Sect. 4 and then use the tightness criterion, Jakubowski’s generalisation of the Skorohod’s theorem and the martingale representaion theorem to prove the existence of martingale solutions. The essential reason for us to incorporate this approximation scheme, is the non-commutativity of gradient operator \(\nabla \) with the standard Faedo-Galerkin projection operator \(P_n\) [4, Section 5]. The commutativity is essential for us to obtain a priori bounds. We also prove the existence of invariant measures, Theorem 6.1, for time homogeneous damped tamed Navier–Stokes equations 6.1 under the assumptions \((\mathbf{H}1 )^\prime - (\mathbf{H}3 )^\prime \) (see Sect. 6). We use the technique (Theorem 6.4) of Maslowski and Seidler [22], see also [5, 9], working with weak topologies to establish the existence of invariant measures. We show the two conditions of Theorem 6.4, boundedness in probability and sequentially weakly Feller property are satisfied for the semigroup \((T_t)_{t \ge 0}\), defined by (6.2). In contrast to the paper by Röckner and Zhang [30], a priori bound on \(L^4\)-norm of the solution plays an essential role in the existence of martingale solutions and not in the existence of invariant measures.
This paper is organised as follows. In Sect. 2, we recall some standard notations and results and set the assumptions on f and \(\sigma \). We also establish certain estimates on the tamed term which we use later in Sects. 4 and 5. We end the section by recalling a generalised version of the Gronwall Lemma for random variables from [14]. In Sect. 3, we establish the tightness criterion and state the Jakubowski’s generalisation of Skorohod theorem which we use along with a priori estimates obtained in the Sect. 5 to prove the existence of a martingale solution and path-wise uniqueness of the solution. In Sect. 4, we introduce our truncated SPDEs and describe the approximation scheme motivated by [15, 20], along with all the machinery required. Finally, in Sect. 6 we establish the existence of an invariant measure for damped tamed 3D Navier–Stokes equations (6.1).
2 Functional Setting
2.1 Notations and Basic Definitions
Let \((X, \Vert \cdot \Vert _X), (Y, \Vert \cdot \Vert _Y)\) be two real normed spaces. The space of all bounded linear operators from X to Y is denoted by \({\mathcal {L}}(X,Y)\). If \(Y = {\mathbb {R}}\), then \(X^\prime := {\mathcal {L}}(X, {\mathbb {R}})\) is called the dual space of X. The standard duality pairing is denoted by \({}_{X^\prime } \langle \cdot , \cdot \rangle _X\). If both spaces are separable Hilbert then by \({\mathcal {L}}_2(Y;X)\) we will denote the space of all Hilbert-Schmidt operators from Y to X endowed with the standard norm \(\Vert \cdot \Vert _{{\mathcal {L}}_2(Y;X)}\).
Assume that X, Y are Hilbert spaces with scalar products \(\langle \cdot , \cdot \rangle _X\) and \(\langle \cdot , \cdot \rangle _Y\) respectively. For a densely defined linear operator \(A : D(A) \rightarrow Y\), \(D(A) \subset X\), by \(A^*\) we denote the adjoint operator of A. In particular, \(D(A^*) \subset Y\), \(A^*: D(A^*) \rightarrow X\) and
Note that \(D(A^*) = Y\) if A is bounded.
Let \(C_0^\infty ({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) denote the set of all smooth functions from \({{\mathbb {R}}^3}\) to \({\mathbb {R}}^3\) with compact supports. For \(p \in [1, \infty ]\) the Lebesgue spaces of \({\mathbb {R}}^3\)-valued functions will be denoted by \(L^p({{\mathbb {R}}^3}; {\mathbb {R}}^3)\), and often by \(L^p\) whenever the context is understood. If \(p = 2\), then \(L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) is a Hilbert space with the scalar product given by
We define, see [36], the Bessel potential space \(H^{s,p}({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) for \(s \ge 0\) and \(p \in (1,\infty )\) as the space of all \({\mathbb {R}}^3\)-valued functions \(u \in L^p({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) such that \(\left( 1 - \Delta \right) ^{s/2} u \in L^p({{\mathbb {R}}^3}; {\mathbb {R}}^3)\), where \(\Delta \) is the Laplace operator in \({\mathbb {R}}^3\). In particular, \(H^1({{\mathbb {R}}^3}; {\mathbb {R}}^3):= H^{1,2}({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) is the space of all \(u \in L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) for which the weak derivatives \(D_iu \in L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\), \(i= 1, \dots , 3\). \(H^1({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) is a Hilbert space with the scalar product given by
where
Finally, for \(s \ge 0\), the space \(H^{s,2}=:H^s\) is also a Hilbert space endowed with the norm
where \({\hat{u}}\) denotes the Fourier transform of a tempered distribution u.
Let
where \(\Pi \) is the orthogonal projection from \(L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) to \(\mathrm {H}\). It is known, see e.g. [17], that \(\Pi =(1-D_iD_j\Delta ^{-1})_{i,j=1}^3\) is a pseudodifferential operator with matrix symbol \((\delta _{i,j}-D_iD_j\Delta ^{-1})_{i,j=1}^3\); \(\Pi \) is a bounded linear operator in \(H^{s,p}({\mathbb {R}}^3)\) by the Marcinkiewicz-Mihlin Theorem [21, 25], for all \(s \in {\mathbb {R}}\) and a \(p\in (1,\infty )\).
On \(\mathrm {H}\) we consider the scalar product and the norm inherited from \(L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) and denote them by \(\langle \cdot , \cdot \rangle _{\mathrm {H}}\) and \(|\cdot |_{\mathrm {H}}\) respectively, i.e.
On \(\mathrm {V}\) we consider the scalar product and norm inherited from \(H^1({{\mathbb {R}}^3}; {\mathbb {R}}^3)\), i.e.
where \(((\cdot , \cdot ))\) is defined in (2.1). \(\mathrm{D}(\mathrm {A})\) is a Hilbert space under the graph norm
where the inner product is given by
2.2 Some Operators
Let us consider the following tri-linear form
defined for suitable vector fields \(u, \mathrm {v}, w\) on \({\mathbb {R}}^3\). We will recall the fundamental properties of the form b which are valid in unbounded domains.
By the Sobolev embedding theorem and Hölder inequality, we obtain the following estimates
for some positive constant c. Thus the form b is continuous on \(\mathrm {V}\). Moreover, if we define a bilinear map B by \(B(u,w) := b(u,w, \cdot )\), then by inequality (2.5) we infer that \(B(u,w) \in \mathrm {V}^\prime \) for all \(u, w \in \mathrm {V}\) and that the following inequality holds
Moreover, the mapping \(B : \mathrm {V}\times \mathrm {V}\rightarrow \mathrm {V}^\prime \) is bilinear and continuous.
Let us, for any \(s > 0\), define the following standard scale of Hilbert spaces
If \( s > \frac{d}{2} + 1\) then by the Sobolev Embedding theorem,
Here \(C_b({\mathbb {R}}^d; {\mathbb {R}}^3)\) denotes the space of continuous and bounded \({\mathbb {R}}^3\)-valued functions defined on \({\mathbb {R}}^d\). If \(u, w \in \mathrm {V}\) and \(\mathrm {v}\in \mathrm {V}_s\) with \(s > \frac{3}{2}\) then
for some constant \(c > 0\). Thus b can be uniquely extended to the tri-linear form (denoted by the same letter)
and
At the same time, the operator B can be uniquely extended to a bounded linear operator
In particular, it satisfies the following estimate
We will also use the following notation, \(B(u) := B(u,u)\).
Let us assume that \(s > 1\). It is clear that \(\mathrm {V}_s\) is dense in \(\mathrm {V}\) and the embedding \(j_s : \mathrm {V}_s \hookrightarrow \mathrm {V}\) is continuous. Then there exists [7, Lemma C.1] a Hilbert space U such that \(U \subset \mathrm {V}_s\), U is dense in \(\mathrm {V}_s\) and
Therefore, the following embedding of the spaces hold
The following Gagliardo-Nirenberg interpolation inequality will be used frequently. Let \(q \in [1, \infty ]\) and \(m \in {\mathbb {N}}\). If
then there exists a constant \(C_{m,q}\) depending on m and q such that
Recall that \(\Pi \) is the orthogonal projection from \(L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\) to \(\mathrm {H}\). For any \(u \in \mathrm {H}\) and \(\mathrm {v}\in L^2({{\mathbb {R}}^3}; {\mathbb {R}}^3)\), we have
The Stokes operator \(\mathrm {A}:\mathrm{D}(\mathrm {A}) \rightarrow \mathrm {H},\) is given by
2.3 Assumptions
We now introduce the following assumptions on the coefficients f and \(\sigma \):
- (H1):
-
\(f:{{\mathbb {R}}^3}\times {\mathbb {R}}^3 \rightarrow {\mathbb {R}}^3\) is a continuous function such that there exists a constant \(C_{f} > 0\) and \(b_f \in L^1({\mathbb {R}}^3)\) such that for all \(x \in {{\mathbb {R}}^3}\), \(u \in {\mathbb {R}}^3\),
$$\begin{aligned} |f(x,u)|^2 \le C_{f}|u|^2 + b_f(x). \end{aligned}$$Moreover, for \(u_1, u_2 \in {\mathbb {R}}^3\)
$$\begin{aligned} |f(x,u_1) - f(x,u_2)| \le C_f |u_1 - u_2|, \qquad \forall \, x \in {\mathbb {R}}^3. \end{aligned}$$ - (H2):
-
A measurable function \(\sigma :[0,\infty )\times {{\mathbb {R}}^3}\rightarrow \ell ^2 \) of \(C^1\) class with respect to the x-variable and for any \(T > 0\) there exists a constant \(C_{\sigma ,T} > 0\) such that for all \(t \in [0,T], x \in {{\mathbb {R}}^3}\)
$$\begin{aligned} \Vert \partial _{x^j} \sigma (t,x)\Vert _{\ell ^2} \le C_{\sigma , T}, \quad j = 1,2,3 \end{aligned}$$and, for all \(t \in [0,\infty ), x \in {{\mathbb {R}}^3}\),
$$\begin{aligned} \Vert \sigma (t,x)\Vert ^2_{\ell ^2} \le \frac{1}{4}\,. \end{aligned}$$(2.11)
Below for the sake of simplicity the variable \(``x''\) in the coefficients will be dropped.
Define, for \(j \in {\mathbb {N}}\), a map \(G_j : [0,T] \times \mathrm {H}\rightarrow \mathrm{H}\) by
Then \(G : \mathrm {H}\rightarrow {\mathcal {L}}_2(\ell ^2; \mathrm {H})\) is given by
Let \(\{e_j\}_{j = 1}^\infty \) be the orthonormal basis of \(\ell ^2\) then
For simplicity we will assume that \(\nu = 1\). In particular, the function g defined by (1.4) will from now on be given by
Observe that the function g defined in this way satisfies
and
We are interested in proving the existence of solutions to (1.1)–(1.3). In particular, we want to prove the existence of a random divergence free vector field u and scalar pressure p satisfying (1.1) and (1.3). Thus we project equation (1.1) using the orthogonal projection operator \(\Pi \) on the space \(\mathrm {H}\) of the \(L^2\)-valued, divergence free vector fields. On projecting, we obtain the following abstract stochastic evolution equation:
![](http://media.springernature.com/lw452/springer-static/image/art%3A10.1007%2Fs00021-020-0480-z/MediaObjects/21_2020_480_Equ21_HTML.png)
where we assume that \(u_0\in \mathrm {V}\) and \(W(t) = (W^j(t))_{j=1}^\infty \) is a cylindrical Wiener process on \(\ell ^2\) and \(\{W^j(t), t \ge 0, j \in {\mathbb {N}}\}\) is an infinite sequence of independent standard Brownian motions. We will repeatedly use the following notation
We will need the following lemma later to obtain the a priori estimates.
Lemma 2.1
-
(i)
For any \(u \in \mathrm{D}(\mathrm {A})\)
$$\begin{aligned} |\langle B(u), u \rangle _{\mathrm {V}}| \le \frac{1}{2} |u|^2_{\mathrm{D}(\mathrm {A})} + \frac{1}{2} \big ||u|\cdot |\nabla u|\big |^2_{L^2}\,. \end{aligned}$$(2.18) -
(ii)
If \(u \in \mathrm {H}\), then
$$\begin{aligned} {\left\{ \begin{array}{ll} ((-g(|u|^2)u,u)) \le C_N |\nabla u|^2_{L^2} - 2 \big ||u|\cdot |\nabla u|\big |^2_{L^2},\\ \langle -g(|u|^2)u, u \rangle _{L^2} \le - \Vert u\Vert ^4_{L^4} + C_N|u|^2_\mathrm {H}, \end{array}\right. } \end{aligned}$$(2.19)where \(((\cdot , \cdot ))\) is defined in (2.1) and \(C_N > 0\) is a generic constant depending on N.
-
(iii)
If \(T > 0\) then \(\exists \)\(C_{\sigma , T} > 0\) such that for any \( t \in [0,T]\) and \(u \in \mathrm{D}(\mathrm {A})\),
$$\begin{aligned} \Vert G(t,u)\Vert ^2_{{\mathcal {L}}_2(\ell ^2;\mathrm {H})}&\le \frac{1}{4} |\nabla u(t)|^2_{L^2}\,, \end{aligned}$$(2.20)$$\begin{aligned} \Vert G(t,u)\Vert ^2_{{\mathcal {L}}_2(\ell ^2;\mathrm {V})}&\le \frac{1}{2} |\mathrm {A}\,u(t)|^2_{L^2} + C_{\sigma ,T}|\nabla u(t)|^2_{L^2}\,. \end{aligned}$$(2.21)
Proof
Let \(u \in \mathrm{D}(\mathrm {A})\). Since \(\langle B(u), u \rangle _\mathrm {H}= 0,\) using the Cauchy-Schwarz and the Young’s inequality we get
Let us introduce a function \(\phi :{\mathbb {R}}_+ \rightarrow {\mathbb {R}}\) such that \(g(r) = r - \phi (r)\) which can be written as
Also \(\phi ^\prime (r) = 1 - g^\prime (r)\), and there exists a constant \({{\widetilde{C}}}_N > 0\)Footnote 3 such that \(|\phi ^\prime (r)| \le {\widetilde{C}}_N\) for every \(r \ge 0 \). Moreover
Thus \(|\phi ^\prime (r) r|\) is bounded by some positive constant \(C_N\). Let \(u \in \mathrm {H}\), then using the definitions of g and \(((\cdot , \cdot ))\) we get
Thus, by the integration by parts formula we get
Using the above bound on \(|\phi ^\prime (r) r |\), we obtain
Since \(g(r) \ge 0\), \(|\phi (r)| \le r\) for all \(r \ge 0\). Thus using (2.23) in (2.22), we obtain
Now to prove the second inequality in (2.19), we take the similar approach. Let \(u \in \mathrm {H}\), then
By the definition of \(\phi \) there exists a constant \(C_N > 0\) such that \(|\phi (r)| \le C_N\) for all \(r > 0\), thus
This completes the proof of part (ii).
Now for (iii), by \((\mathbf{H}1 )\) and \((\mathbf{H}2 )\) we have
Secondly, noting that
and
by \((\mathbf{H}1 )\) and \((\mathbf{H}2 )\) we have
\(\square \)
On a purely heuristic level, by application of the Itô formula to the function \(\mathrm {H}\ni \xi \mapsto |\xi |^2_\mathrm {H}\in {\mathbb {R}}\) and a solution u to (1.1) and using Lemma 2.1, one obtains the following inequality
which could lead to a priori estimates that can be used further to prove the existence of the solution.
We will require the following version of the Gronwall Lemma [14, Lemma 3.9]\(:\)
Lemma 2.2
Let X, Y, I and \(\varphi \) be non-negative processes and Z be a non-negative integrable random variable. Assume that I is non-decreasing and there exist non-negative constants \(C,\alpha , \beta , \gamma , \eta \) with the following properties
and such that for some constant \({\widetilde{C}} > 0\) and all \( t \in [0,T]\),
If \(X \in L^\infty ([0,T] \times \Omega )\), then we have
3 Compactness
Let \(({\mathcal {O}}_R)_{R\in {\mathbb {N}}}\) be a sequence of bounded open subsets of \({{\mathbb {R}}^3}\) with regular boundaries \(\partial {\mathcal {O}}_R\) such that \({\mathcal {O}}_R \subset {\mathcal {O}}_{R+1}\). Let us consider the following functional spaces, which were already introduced in [7]:
\({\mathcal {C}}([0,T]; \mathrm {U}') :=\) the vector space of continuous functions \(u: [0,T] \rightarrow \mathrm {U}^\prime \), where \(\mathrm {U}^\prime \) is a vector space, with the topology \({\mathcal {T}}_1\) induced by the norm \(|u|_{{\mathcal {C}}([0,T]; \mathrm {U}^\prime )} := \sup _{t \in [0,T]}|u(t)|_{\mathrm {U}^\prime }\),
\(L^2_w(0,T; \mathrm {D}(\mathrm {A})) :=\) is the Hilbert space \(L^2(0,T; \mathrm {D}(\mathrm {A}))\) with the weak topology \({\mathcal {T}}_2\),
\(L^2(0,T; \mathrm {H}_{loc}) :=\) the space of measurable functions \(u : [0,T] \rightarrow \mathrm {H}\) such that for all \(R \in {\mathbb {N}}\)
with the topology \({\mathcal {T}}_3\) induced by the semi-norms \((q_{T,R})_{R \in {\mathbb {N}}}\).
The following lemma is inspired by the classical Dubinsky theorem [37, Theorem IV.4.1] (see also [19]) and the compactness result due to Mikulevicus and Rozovskii [26, Lemma 2.7].
Lemma 3.1
Let
and let \(\widetilde{{\mathcal {T}}}\)Footnote 4 be the supremum of the corresponding topologies. Then a set \({\mathcal {K}}\subset {\widetilde{{\mathcal {Z}}}}_T\) is \(\widetilde{{\mathcal {T}}}\)-relatively compact if the following two conditions hold:
-
(i)
\(\sup _{u \in {\mathcal {K}}} \int _0^T |u(s)|^2_{\mathrm {D}(\mathrm {A})}\,ds < \infty \,,\,\) i.e. \({\mathcal {K}}\) is bounded in \(L^2(0,T; \mathrm {D}(\mathrm {A}))\,,\)
-
(ii)
\(\lim _{\delta \rightarrow 0} \sup _{u \in {\mathcal {K}}} \sup _{\overset{s,t \in [0,T]}{|t-s|\le \delta }} {|u(t)-u(s)|}_{\mathrm {U}^{\prime }} =0\,.\)
The above lemma can be proved by modifying the proof of [7, Lemma 3.1], see also [37, Theorem IV.4.1].
Let \(\mathrm {V}_w\) denote the Hilbert space \(\mathrm {V}\) endowed with the weak topology.
\({\mathcal {C}}([0,T]; \mathrm {V}_w) :=\) the space of weakly continuous functions \(u: [0,T] \rightarrow \mathrm {V}\) endowed with the weakest topology \({\mathcal {T}}_4\) such that for all \(h \in \mathrm {V}\) the mappings
are continuous. In particular, \(u_n \rightarrow u\) in \({\mathcal {C}}([0,T]; \mathrm {V}_w)\) iff for all \(h \in \mathrm {V}:\)
Consider the ball
Let q be the metric compatible with the weak topology on \({\mathbb {B}}\). Let us recall the following subspace of the space \({\mathcal {C}}([0,T]; \mathrm {V}_w)\)
The space \({\mathcal {C}}([0,T]; {\mathbb {B}}_w)\) is metrizable with metric
Since by the Banach-Alaoglu theorem [33], the set \({\mathbb {B}}_w\) is compact, \(({\mathcal {C}}([0,T]; {\mathbb {B}}_w), \varrho )\) is a complete metric space.
The following lemma says that any sequence \((u_n)_{n \in {\mathbb {N}}} \subset L^\infty ([0,T]; {\mathbb {B}})\), convergent in \({\mathcal {C}}([0,T]; \mathrm {U}^\prime )\) is also convergent in the space \({\mathcal {C}}([0,T]; {\mathbb {B}}_w)\). The proof of the lemma is similar to the proof of [8, Lemma 2.1] (see also [6, Lemma 4.1]).
Lemma 3.2
Let \(r > 0\) and \(\left( u_n\right) _{n\in {\mathbb {N}}} \subset L^\infty (0,T; \mathrm {V})\) be a sequence of functions such that
-
(i)
\(\sup _{n \in {\mathbb {N}}} \sup _{s \in [0,T]} \Vert u_n(s)\Vert _\mathrm {V}\le r\,,\)
-
(ii)
\(u_n \rightarrow u\) in \({\mathcal {C}}([0,T]; \mathrm {U}^\prime )\,.\)
Then \(u,u_n \in {\mathcal {C}}([0,T]; {\mathbb {B}}_w)\) and \(u_n \rightarrow u\) in \({\mathcal {C}}([0,T]; {\mathbb {B}}_w)\) as \(n \rightarrow \infty \).
Proof
The lemma can be proved by following the steps of the proof of Lemma 4.1 [6] for our choice of functional spaces. \(\square \)
Let
and let \({\mathcal {T}}\) be the supremum of the corresponding topologies.
Now we formulate the compactness criterion analogous to the result due to Mikulevicus and Rozowskii [26], Brzeźniak and Motyl [7, Lemma 3.3] for the space \({\mathcal {Z}}_T\).
Lemma 3.3
Let \(\left( {\mathcal {Z}}_T, {\mathcal {T}}\right) \) be as defined in (3.6). Then a set \({\mathcal {K}}\subset {\mathcal {Z}}_T\) is \({\mathcal {T}}\)-relatively compact if the following three conditions hold
-
(a)
\(\sup _{u \in {\mathcal {K}}} \sup _{s \in [0,T]} \Vert u(s)\Vert _\mathrm {V}< \infty \,,\)
-
(b)
\(\sup _{u \in {\mathcal {K}}} \int _0^T |u(s)|^2_{\mathrm {D}(\mathrm {A})}\,ds < \infty \,\), i.e. \({\mathcal {K}}\) is bounded in \(L^2(0,T; \mathrm {D}(\mathrm {A}))\,,\)
-
(c)
\(\lim _{\delta \rightarrow 0} \sup _{u \in {\mathcal {K}}} \sup _{\underset{|t-s| \le \delta }{s,t \in [0,T]}}|u(t) - u(s)|_{\mathrm {H}} = 0\,.\)
Proof
Let us denote \(r = \sup _{u \in {\mathcal {K}}}\sup _{s \in [0,T]}\Vert u(s)\Vert _\mathrm {V}\), which in view of assumption (c), is \(<\infty \). Define the ball \({\mathbb {B}}\) of radius r by (3.3).
Let us notice that \({\mathcal {Z}}_T = {\widetilde{{\mathcal {Z}}}}_T \cap {\mathcal {C}}([0,T]; \mathrm {V}_w)\), where \({\widetilde{{\mathcal {Z}}}}_T\) is defined by (3.2). Let \({\mathcal {K}}\) be a subset of \({\mathcal {Z}}_T\). Because of the assumption (a) we may consider the metric space \({\mathcal {C}}([0,T]; {\mathbb {B}}_w) \subset {\mathcal {C}}([0,T]; \mathrm {V}_w)\) defined by (3.4) and (3.5). Because of the assumption (b), the restriction to \({\mathcal {K}}\) of the weak topology in \(L^2(0,T; \mathrm {D}(\mathrm {A}))\) is metrizable. Since the restrictions to \({\mathcal {K}}\) of the four topologies considered in \({\mathcal {Z}}_T\) are metrizable, compactness of a subset of \({\mathcal {Z}}_T\) is equivalent to its sequential compactness.
Let \((u_n)\) be a sequence in \({\mathcal {K}}\). By Lemma 3.1, the boundedness of the set \({\mathcal {K}}\) in \({L}^{2}(0,T; \mathrm {D}(\mathrm {A}))\) and assumption (c) imply that \({\mathcal {K}}\) is compact in \({\widetilde{{\mathcal {Z}}}}_T\). Hence in particular, there exists a subsequence, still denoted by \((u_n)\), convergent in \({\mathcal {C}}([0,T];\mathrm {U}^\prime )\). Therefore by Lemma 3.2 and assumption (a), \((u_n)\) is convergent in \({\mathcal {C}}([0,T]; {\mathbb {B}}_w)\). This completes the proof of the lemma. \(\square \)
3.1 Tightness
Let \(({\mathbb {S}}, \varrho )\) be a separable and complete metric space.
Definition 3.4
Let \(u \in {\mathcal {C}}([0,T]; {\mathbb {S}})\). The modulus of continuity of u on [0, T] is defined by
Let \((\Omega , {\mathcal {F}}, {\mathbb {P}})\) be a probability space with filtration \({\mathbb {F}}:= ({\mathcal {F}}_t)_{t \in [0,T]}\) satisfying the usual conditions, see [23], and let \((X_n)_{n \in {\mathbb {N}}}\) be a sequence of continuous \({\mathbb {F}}\)-adapted \({\mathbb {S}}\)-valued processes.
Definition 3.5
We say that the sequence \((X_n)_{n \in {\mathbb {N}}}\) of \({\mathcal {C}}([0,T]; {\mathbb {S}})\)-valued random variables satisfies condition \([{\mathbf {T}}]\) iff \(\forall \, \varepsilon>0, \forall \, \eta> 0,\, \exists \, \delta > 0:\)
Lemma 3.6
(See [8, Lemma 2.4]) Assume that \((X_n)_{n \in {\mathbb {N}}}\) satisfies condition \([{\mathbf {T}}]\). Let \({\mathbb {P}}_n\) be the law of \(X_n\) on \({\mathcal {C}}([0,T]; {\mathbb {S}}), n \in {\mathbb {N}}\). Then for every \(\varepsilon > 0\) there exists a subset \(A_\varepsilon \subset {\mathcal {C}}([0,T]; {\mathbb {S}})\) such that
and
Now we recall the Aldous condition which is connected with condition \([{\mathbf {T}}]\) (see [2, 24]). This condition allows to investigate the modulus of continuity for the sequence of stochastic processes by means of stopped processes.
Definition 3.7
A sequence \((X_n)_{n \in {\mathbb {N}}}\) satisfies condition \([{\mathbf {A}}]\) iff \(\forall \, \varepsilon > 0\), \(\forall \, \eta > 0\), \(\exists \, \delta > 0\) such that for every sequence \((\tau _n)_{n \in {\mathbb {N}}}\) of \({\mathbb {F}}\)-stopping times with \(\tau _n \le T\) one has
Lemma 3.8
(See [24, Theorem 3.2]) Conditions \([{\mathbf {A}}]\) and \([{\mathbf {T}}]\) are equivalent.
Using the compactness criterion from Lemma 3.3 and above results corresponding to Aldous condition we obtain the following corollary which we will use to prove the tightness of the laws defined by the truncated SPDE (4.25).
Corollary 3.9
(Tightness criterion) Let \((X_n)_{n \in {\mathbb {N}}}\) be a sequence of \({\mathbb {F}}\)-adapted continuous \(\mathrm{H}\)-valued processes such that
-
(a)
there exists a constant \(C_1 > 0\) such that
$$\begin{aligned} \sup _{n \in {\mathbb {N}}} {\mathbb {E}}\left[ \sup _{s \in [0,T]} \Vert X_n(s)\Vert ^2_{\mathrm{V}} \right] \le C_1, \end{aligned}$$ -
(b)
there exists a constant \(C_2 > 0\) such that
$$\begin{aligned} \sup _{n \in {\mathbb {N}}} {\mathbb {E}}\left[ \int _0^T |X_n(s)|^2_{\mathrm{D}(\mathrm {A})}\,ds \right] \le C_2, \end{aligned}$$ -
(c)
\((X_n)_{n \in {\mathbb {N}}}\) satisfies the Aldous condition \([{\mathbf {A}}]\) in \(\mathrm{H}\).
Let \({{\mathbb {P}}}_n\) be the law of \(X_n\) on \({\mathcal {Z}}_T\). Then for every \(\varepsilon > 0\) there exists a compact subset \(K_\varepsilon \) of \({\mathcal {Z}}_T\) such that
Proof
Let \(\varepsilon > 0\). By the Chebyshev inequality and (a), we infer that for any \(n \in {\mathbb {N}}\) and any \(r>0\)
Let \({R}_{1}\) be such that \(\frac{{C}_{1}}{{R}_{1}} \le \frac{\varepsilon }{3}\). Then
Let \({B}_{1}:= \left\{ X_n \in {\mathcal {Z}}_T :\, \, \sup _{s \in [0,T]}\Vert X_n(s) {\Vert }_{\mathrm {V}}^{2} \le {R}_{1} \right\} \).
By the Chebyshev inequality and (b), we infer that for any \(n \in {\mathbb {N}}\) and any \(r>0\)
Let \({R}_{2}\) be such that \(\frac{{C}_{2}}{{R}_{2}^{2}} \le \frac{\varepsilon }{3}\). Then
Let \({B}_{2} := \left\{ X_n \in {\mathcal {Z}}_T : \, \, \Vert X_n\Vert _{{L}^{2}(0,T; \mathrm {D}(\mathrm {A}))} \le {R}_{2} \right\} \).
By Lemmas 3.6 and 3.8 there exists a subset \({A}_{\frac{\varepsilon }{3}} \subset {\mathcal {C}}([0,T], \mathrm {H})\) such that \({{{\mathbb {P}} }}_{n} \bigl ( {A}_{\frac{\varepsilon }{3}}\bigr ) \ge 1 - \frac{\varepsilon }{3}\) and
It is sufficient to define \({K}_{\varepsilon } \) as the closure of the set \({B}_{1} \cap {B}_{2} \cap {A}_{\frac{\varepsilon }{3}}\) in \({\mathcal {Z}}_T\). By Lemma 3.3, \({K}_{\varepsilon }\) is compact in \({\mathcal {Z}}_T\). The proof is thus complete. \(\square \)
3.2 The Skorohod Theorem
We will use the following Jakubowski’s generalisation of the Skorohod theorem in the form given by Brzeźniak and Ondreját [10], see also [16].
Theorem 3.10
Let \({\mathcal {X}}\) be a topological space such that there exists a sequence \(\{f_m\}_{m \in {\mathbb {N}}}\) of continuous functions \(f_m : {\mathcal {X}} \rightarrow {\mathbb {R}}\) that separates points of \({\mathcal {X}}\). Let us denote by \({\mathcal {S}}\) the \(\sigma \)-algebra generated by the maps \(\{f_m\}\). Then
-
(a)
every compact subset of \({\mathcal {X}}\) is metrizable,
-
(b)
if \((\mu _m)_{m \in {\mathbb {N}}}\) is a tight sequence of probability measures on \(({\mathcal {X}}, {\mathcal {S}})\), then there exists a subsequence \((m_k)_{k \in {\mathbb {N}}}\), a probability space \((\Omega , {\mathcal {F}}, {\mathbb {P}})\) with \({\mathcal {X}}\)-valued Borel measurable variables \(\xi _k, \xi \) such that \(\mu _{m_k}\) is the law of \(\xi _k\) and \(\xi _k\) converges to \(\xi \) almost surely on \(\Omega \). Moreover, the law of \(\xi \) is a Radon measure.
Using Theorem 3.10, we obtain the following corollary which we will apply to construct a martingale solution of the stochastic tamed Navier–Stokes equations.
Corollary 3.11
Let \((\eta _n)_{n \in {\mathbb {N}}}\) be a sequence of \({\mathcal {Z}}_T\)-valued random variables such that their laws \(law (\eta _n)\) on \(({\mathcal {Z}}_T, {\mathcal {T}})\) form a tight sequence of probability measures. Then there exists a subsequence \((n_k)\), a probability space \(({\widetilde{\Omega }}, \widetilde{{\mathcal {F}}}, \widetilde{{\mathbb {P}}})\) and \({\mathcal {Z}}_T\)-valued random variables \({\widetilde{\eta }}\), \({\widetilde{\eta }}_k, k \in {\mathbb {N}}\) such that the variables \(\eta _k\) and \({\widetilde{\eta }}_k\) have the same laws on \({\mathcal {Z}}_T\) and \({\widetilde{\eta }}_k\) converges to \({\widetilde{\eta }}\) almost surely on \({\widetilde{\Omega }}\).
Proof
It is sufficient to prove that on each space appearing in the definition (3.6) of the space \({\mathcal {Z}}_T\), there exists a countable set of continuous real-valued functions separating points.
Since the spaces \({\mathcal {C}}([0,T]; \mathrm {U}^\prime )\) and \({L}^{2}(0,T; \mathrm {H}_{loc})\) are separable, metrizable and complete, this condition is satisfied, see [3], exposé 8.
For the space \({L}^{2}_{w}(0,T; \mathrm {D}(\mathrm {A}))\) it is sufficient to put
where \(\{ {\mathrm {v}}_{m}, m \in {\mathbb {N}}\} \) is a dense subset of \({L}^{2}(0,T; \mathrm {D}(\mathrm {A}))\).
Let us consider the space \({\mathcal {C}}([0,T];{\mathrm {V}}_{w})\). Let \(\{ {h}_{m}, \, m \in {\mathbb {N}}\} \) be any dense subset of \(\mathrm {V}\) and let \({{\mathbb {Q}}}_{T}\) be the set of rational numbers belonging to the interval [0, T]. Then the family \(\{ {f}_{m,t}, \, m \in {\mathbb {N}}, \, \, t \in {{\mathbb {Q}}}_{T} \} \) defined by
consists of continuous functions separating points in \({\mathcal {C}}([0,T];{\mathrm {V}}_{w})\). The statement of the corollary follows from Theorem 3.10, concluding the proof. \(\square \)
We end this section by giving the definitions of a martingale and strong solution to (2.17).
Definition 3.12
A stochastic basis \((\Omega , {\mathcal {F}}, {\mathbb {F}}, {\mathbb {P}})\) is a probability space equipped with the filtration \({\mathbb {F}} = \{{\mathcal {F}}_t\}_{t \ge 0}\) of its \(\sigma -\)field \({\mathcal {F}}\).
Definition 3.13
A martingale solution of (2.17) is a system
where \(\left( {\widehat{\Omega }}, \widehat{{\mathcal {F}}}, \widehat{{\mathbb {P}}}\right) \) is a probability space and \(\widehat{{\mathbb {F}}} = \left( \widehat{{\mathcal {F}}}_t\right) _{t\ge 0}\) is a filtration on it, such that
-
\({\widehat{W}}\) is an \(\ell ^2\)-valued cylindrical Wiener process on \(\left( {\widehat{\Omega }}, \widehat{{\mathcal {F}}}, \widehat{{\mathbb {F}}}, \widehat{{\mathbb {P}}}\right) \),
-
\({\widehat{u}}\) is \(\mathrm{D}(\mathrm{A})\)-valued progressively measurable process,\(\mathrm {V}\)-valued weakly continuous \(\widehat{{\mathbb {F}}}\)-adapted process such that
$$\begin{aligned}&{\widehat{u}}(\cdot , \omega ) \in {\mathcal {C}}([0,T]; \mathrm {V}_w) \cap L^2(0,T; \mathrm{D}(\mathrm {A})), \\&{\widehat{{\mathbb {E}}}}\left( \sup _{t\in [0,T]}\Vert {\widehat{u}}(t)\Vert ^2_{\mathrm {V}} + \int _0^T |{\widehat{u}}(t)|^2_{\mathrm{D}(\mathrm{A})}dt\right) < \infty \end{aligned}$$and
$$\begin{aligned} \begin{aligned}&\langle {\widehat{u}}(t), \mathrm {v}\rangle + \int _0^t \langle \mathrm {A}{\widehat{u}}(s), \mathrm {v}\rangle \,ds + \int _0^t \langle B({\widehat{u}}(s)), \mathrm {v}\rangle \,ds \\&\quad + \int _0^t \langle g(|{\widehat{u}}(s)|^2)\,{\widehat{u}}(s), \mathrm {v}\rangle \,ds = \langle u_0, \mathrm {v}\rangle + \int _0^t \langle f({\widehat{u}}(s)), \mathrm {v}\rangle \,ds \\&\quad + \left\langle \int _0^t G(s,{\widehat{u}}(s))\,dW(s) , \mathrm {v}\right\rangle . \end{aligned} \end{aligned}$$(3.9)for all \(t \in [0,T]\) and all \(\mathrm {v}\in {\mathcal {V}}\), \(\widehat{{\mathbb {P}}}\)-a.s.
Definition 3.14
We say that problem (2.17) has a strong solution if for every stochastic basis \((\Omega , {\mathcal {F}}, {\mathbb {F}}, {\mathbb {P}})\) and \(\ell ^2\)-valued cylindrical Wiener process W(t) on the given filtered probability space there exists a \(\mathrm{D}(\mathrm{A})\)-valued progressively measurable process,\(\mathrm {V}\)-valued continuous \({{\mathbb {F}}}\)-adapted process u such that
and satisfies (3.9) for all \(t \in [0,T]\) and all \(\mathrm {v}\in {\mathcal {V}}\), \({\mathbb {P}}\)-a.s.
Remark 3.15
The strong solution defined in Definition 3.14 is a probabilistically strong solution with the same regularity as a martingale solution. But, in Theorem 5.19 we show that the strong solution u is more regular, i.e. \(u \in {\mathcal {C}}([0,T]; \mathrm {V})\cap L^2(0,T; \mathrm {D}(\mathrm {A}))\).
4 Truncated SPDEs
The approximation scheme described in this section to define truncated SPDEs was first introduced by [15] and also later used by Manna et al. in [20].
In order to describe the approximation scheme, we will use the following notations and spaces.
where \(|\cdot |\) is the Euclidean norm on \({\mathbb {R}}^3\). We will use \({\mathcal {F}}(u)\) and \({\hat{u}}\) interchangeably to denote the Fourier transform of u. The inverse Fourier transform will be given by \({\mathcal {F}}^{-1}\).
We define \(\mathrm {H}_n\) as the subspace of \(\mathrm {H}\),
The norm on \(\mathrm {H}_n\) is inherited from \(\mathrm {H}\). For \(n \in {\mathbb {N}}\), let us define a map \(P_n\) by
Firstly, note that \(P_n : \mathrm {H}\rightarrow \mathrm {H}\) is a linear and bounded map. Moreover, \(\mathrm {Range}(P_n) \subset H_n\) and hence one can deduce that
In addition, \(P_n\) is an orthogonal projection onto \(\mathrm {H}_n\) i.e. \(\forall \, u \in \mathrm {H}\), \(u - P_n\,u \perp \mathrm {H}_n\). In other words
Indeed, for \(u \in \mathrm {H}\) and \(\mathrm {v}\in \mathrm {H}_n\), we have
Let us recall that \(\mathrm{D}(\mathrm {A}) := \mathrm {H}\cap H^{2,2}\) and the Stokes operator is given by
and \(\mathrm{D}(\mathrm {A})\) is a Hilbert space under the graph norm
Lemma 4.1
Let \(P_n\) be the orthogonal projection given by (4.1), then \(P_n : \mathrm {V}\rightarrow \mathrm {V}\) is uniformly bounded.
Proof
Let \(u \in \mathrm {V}\), then by the definition of \(P_n\) and \(\mathrm {V}\)
Thus we have shown that
and hence \(P_n\) is uniformly bounded in \(\mathrm {V}\). \(\square \)
Lemma 4.2
If \(u \in \mathrm{D}(\mathrm {A})\) then \(\Delta \, u \in \mathrm {H}\). In particular, if \(u \in \mathrm{D}(\mathrm {A})\) then \(\mathrm {A}\,u = - \Delta \,u\).
Proof
Since \(u \in \mathrm{D}(\mathrm {A})\), it is clear that \(\Delta \, u \in L^2\). Thus we are left to show that \(\mathrm{div}(\Delta \, u) = 0\) in the weak sense. Let \(\varphi \in C_0^\infty ({\mathbb {R}}^3)\), then using the definition of \(\mathrm{div}\) and \(\Delta \), we get
By definition \(\mathrm {A}\,u = - \Pi (\Delta \,u)\), but since \(\Delta \, u \in \mathrm {H}\), and \(\Pi : L^2 \rightarrow \mathrm {H}\) is an orthogonal projection, \(\Pi (\Delta \,u) = \Delta \,u\) and hence,
\(\square \)
Lemma 4.3
If \(n \in {\mathbb {N}}\), then \(H_n \subset \mathrm{D}(\mathrm {A})\) and
Proof
We start with proving the first statement. Let \(u \in \mathrm {H}_n\); by definition
Since \(u \in \mathrm {H}_n, \mathrm{supp}({\hat{u}}) \subset B_n\),
Thus we have proved that \(u \in \mathrm{D}(\mathrm {A})\) and hence \(\mathrm {H}_n \subset \mathrm{D}(\mathrm {A})\). Moreover, we showed that there exists a constant \(C_n > 0\), depending on n such that
Now in order to establish the equality (4.3), we just need to show that \(\mathrm {A}u \in \mathrm{H}_n\). Since \(u \in \mathrm{H}_n\), \(u \in \mathrm{D}(\mathrm {A})\). Hence from Lemma 4.1, \(\mathrm {A}\,u = - \Delta \,u\). We are left to show that \(\mathrm{supp}\,({\mathcal {F}}(\mathrm {A}u)) \subset B_n\). Using the definition of \(\mathrm {A}\,u\), we get following equalities
Thus
Hence \(\mathrm {A}u \in \mathrm{H}_n\). Since \(P_n : \mathrm {H}\rightarrow \mathrm {H}_n\) is an orthogonal projection, we infer that
\(\square \)
Lemma 4.4
If \(n \in {\mathbb {N}}\), then the map \(\mathrm {A}_n := \mathrm {A}\big |_{H_n} : H_n \rightarrow H_n,\) is linear and bounded.
Proof
In Lemma 4.2 we showed that \(\mathrm {A}_n\) is well defined and it’s straightforward to show it is linear. We are left to show that it is bounded. Let \(u \in \mathrm{H}_n\), then by Parseval equality and the definition of \(\mathrm {H}_n\)
Thus,
\(\square \)
Lemma 4.5
If \(n \in {\mathbb {N}}\), then the map
is well defined and Lipschitz on a ball \({\mathbb {B}}_R := \left\{ u \in \mathrm {H}_n : \Vert u\Vert _{\mathrm {H}_n} \le R \right\} \), \(R > 0\). Moreover
where \(B_n(u):= B_n(u,u)\) and \(((\cdot , \cdot ))\) is defined in (2.1).
Proof
We will show that \(\forall \, u,\mathrm {v}\in \mathrm {H}_n, B(u,\mathrm {v}) \in \mathrm {H}\). Since \(u, \mathrm {v}\in \mathrm{H}_n\), \(u, \mathrm {v}\in \mathrm{D}(\mathrm {A})\). Thus,
Since \(H^{s,2}({\mathbb {R}}^d) \hookrightarrow L^\infty ({\mathbb {R}}^d)\) for every \(s > \frac{d}{2}\), there exists a constant \(C > 0\) such that
In particular, it holds true for \(s = 2\). Thus, we have
Hence \(B(u, \mathrm {v}) \in \mathrm {H}\), which implies \(B_n(u,\mathrm {v}) \in \mathrm {H}_n\) and is well defined.
Let \(R > 0\) be fixed and \(u,\mathrm {v}\in {\mathbb {B}}_R\). Then, as before, using the embedding \(H^2 \hookrightarrow L^\infty \), we have
Since \(u, \mathrm {v}\in {\mathbb {B}}_R\), and using (4.4) and (4.14), we get
Since \(u \in \mathrm {H}_n\) and \(P_n\) is the orthogonal projection on \(\mathrm {H}\),
Also by using the definition of \(((\cdot , \cdot ))\) and the Cauchy-Schwarz inequality we get
\(\square \)
Lemma 4.6
The map
is well defined and Lipschitz on a ball \({\mathbb {B}}_R\), \(R > 0\). Moreover
Proof
Let \(u \in \mathrm {H}_n\), then by the definition of g (2.14), the estimate (2.15) and the embedding of \(H^1 \hookrightarrow L^6\), we have
Therefore \(g_n : \mathrm {H}_n \rightarrow \mathrm {H}_n\) is well defined. From above we can also infer that there exists a constant \(C_n > 0\) depending on n such that
Let \(R > 0\) be fixed and \(u, \mathrm {v}\in {\mathbb {B}}_R\). Then, using (2.16), we have
Since \(H^1 \hookrightarrow L^6\), we obtain
Since \(u , \mathrm {v}\in {\mathbb {B}}_R\), using (4.14), we get
Let \(u \in \mathrm {H}_n\), then using Lemmas 4.2 and 4.3, the definitions of \(g_n\) and \(((\cdot , \cdot ))\) we get
Also, note that
Hence the inequalities (4.12) can be established with the help of the above two relations and Lemma 2.1 (ii). This completes the proof of the lemma. \(\square \)
Lemma 4.7
Let f satisfy the assumption \((\mathbf{H1})\).Then the map
is well defined and Lipschitz.
Proof
Let \(u \in \mathrm {H}_n\), then by the assumption \((\mathbf{H1})\),
Therefore \(f_n : \, \mathrm {H}_n \rightarrow \mathrm {H}_n\) is well defined. Let \(u, \mathrm {v}\in \mathrm {H}_n\), then
\(\square \)
Lemma 4.8
Let \(\sigma \) satisfy the assumption \((\mathbf{H2})\). Then the map
is well defined and Lipschitz.
Proof
Let \(u \in \mathrm {H}_n\), then
Using (4.14), we infer
Thus \(G_n : \, \mathrm {H}_n \rightarrow {\mathcal {L}}_2(\ell ^2; \mathrm {H}_n)\) is well defined. Let \(u, \mathrm {v}\in \mathrm {H}_n\), then
Using (4.14), we infer
\(\square \)
Proposition 4.9
\(L^2, H^1\) and \(\mathrm{D}(\mathrm {A})\) norms on \(\mathrm {H}_n\) are equivalent (with constants depending on n).
Proof
Let \(u \in \mathrm {H}_n\), then using the Parseval’s identity
Thus if \(u \in \mathrm {H}_n\) then \(L^2\) and \(\mathrm {H}_n\) have equal norms. The equivalence of \(H^1\) and \(\mathrm {H}_n\) norms is established from (4.14). Using (4.5) and (4.21) we can establish equivalence of \(\mathrm{D}(\mathrm {A})\) and \(\mathrm {H}_n\) norms. \(\square \)
As discussed earlier in the introduction instead of using the standard Galerkin approximation of SPDE on the finite dimensional space we will look at truncated SPDEs on infinite dimensional space \(\mathrm {H}_n\). For every \(n \in {\mathbb {N}}\), we will establish the existence of a unique global solution to the truncated SPDE and obtain a priori estimates in order to prove the tightness of measures on a suitable space.
In order to study the truncated SPDE on \(\mathrm {H}_n\) we project the SPDE (2.17) on \(\mathrm {H}_n\) using \(P_n\). The projected SPDE on \(\mathrm {H}_n\) is given by
where \(u_n \in \mathrm {H}_n, u_0 \in \mathrm {V}\) and other operators \(\mathrm {A}_n, B_n, g_n, f_n\) and \(G_n\) are as defined in Lemmas 4.4 – 4.8.
Lemma 4.10
Let us define \(F :\, \mathrm {H}_n \rightarrow {\mathbb {R}}\) by
Then for every \(u \in \mathrm {H}_n\) there exists \(K_1 > 0\), independent of n, such that
Proof
From the definition of \(B_n, g_n\) and \(f_n\), we have
Using Lemma 4.8 and since \(u \in \mathrm {H}_n\), we get
where in the last inequality we used the Young’s inequality and hypothesis (H1). On rearranging, we get
for appropriately chosen \(K_1\). Thus, in particular
\(\square \)
We will use the following theorem from [1, Theorem 3.1] to prove Theorem 4.12. We have modified it in the way we will be using it.
Theorem 4.11
Let X be an abstract Hilbert space. Assume that \(\sigma \) and b satisfy the following conditions
-
(i)
For any \(R > 0\) there exists a constant \(C > 0\) such that
$$\begin{aligned} \Vert \sigma (u) - \sigma (\mathrm {v})\Vert _{{\mathcal {L}}_2(\ell ^2;X)} + \Vert b(u) - b(\mathrm {v})\Vert _X \le C \Vert u -\mathrm {v}\Vert ^2_X, \quad \Vert u\Vert _X,\Vert \mathrm {v}\Vert _X \le R\,. \end{aligned}$$ -
(ii)
There exists a constant \(K_1 > 0\) such that
$$\begin{aligned} \Vert \sigma (u)\Vert ^2_{{\mathcal {L}}_2(\ell ^2; X)} + 2 \langle u, b(u) \rangle _{L^2} \le K_1(1+\Vert u\Vert ^2_X), \quad \quad u \in X\,. \end{aligned}$$
Then for any X-valued \(\xi \), there exists a unique global solution \(u = \left( u(t)\right) _{t \ge 0}\) to
Theorem 4.12
Let the assumptions \((\mathbf{H1})\)–\((\mathbf{H2})\) hold. Then for every \(u_0 \in \mathrm {V}\) there exists a unique global solution \(u_n = \left( u_n(t)\right) _{t \ge 0}\) to
![](http://media.springernature.com/lw466/springer-static/image/art%3A10.1007%2Fs00021-020-0480-z/MediaObjects/21_2020_480_Equ66_HTML.png)
Proof
The proof is a direct application of Theorem 4.11. Using Lemmas 4.4 - 4.8, we can show that condition (i) of Theorem 4.11 is satisfied. In Lemma 4.10 we proved that condition (ii) is satisfied. Thus we have existence of the unique global solution \(u_n = \left( u_n(t) \right) _{t \ge 0}\) to (4.25). \(\square \)
5 Existence of Solution
5.1 A priori Estimates
In this subsection we will obtain certain a priori estimates for the solution \(u_n\) of (4.25). We will use these a priori estimates in Lemma 5.3 to prove the tightness of measures on the space \({\mathcal {Z}}_T\), defined in (3.6). We will also establish certain higher order estimates which will be required to prove the convergence of non-linear terms in later sections.
Let us fix \(T > 0\). For any \(R > 0\), define the stopping time
where \(u_n\) is the solution of (4.25). By the definition of a martingale solution one knows that for every \(n \ge 1\), \(\tau _R^n \nearrow \infty \) as \(R \nearrow \infty \).
Lemma 5.1
Let \(u_n\) be the solution of (4.25). For all \(\rho >0\) there exist positive constants \(C_1(\rho ), C_2(\rho )\) such that if \(\Vert u_0\Vert _\mathrm {V}\le \rho \), then
Moreover, for every \(\delta > 0\) there exists a constant \(C(\delta ) > 0\) such that if \(|u_0|_\mathrm {H}\le \delta \), then
Proof
Let \(u_n(t)\), \(t \ge 0\) be the solution of (4.25) then applying the Itô formula to \(\phi (\xi ) = |\xi |_{\mathrm{H}}^2\) and the process \(u_n(t)\), we get
Using Lemma 2.1, the Young’s inequality, assumption \((\mathbf{H}1 )\), boundedness of \(P_n\) in \(\mathrm {H}\), we get
Using the (4.12)\(_2\), we get
On rearranging we get
Now since
is a \({\mathbb {F}}\)-martingale, as by Lemma 2.1 and (5.1) we have the following inequalities
where to establish the last inequality we have used the equivalences of norm from Proposition 4.9. Thus, \({\mathbb {E}}[ \mu _n(t)] = 0\).
Hence applying Lemma 2.2 for the following three processes\(:\)
we obtain from (5.7), (2.26) is satisfied for \(\alpha = 1\) , \(Z = |u_0|^2_\mathrm {H}+ T|b_f|_{L^1}\) and \(\phi (r) = C_{f,N}\). Since \({\mathbb {E}}(I(t)) = 0\), (2.27) is satisfied and hence all inequalities for the parameters (see (2.25)) are trivially satisfied. Thus, if \(|u_0|_\mathrm {H}\le \delta \), we have
In particular,
Hence, using (5.8) and (5.9) we infer that
Since we are interested in the estimates involving \(\mathrm {V}\) norm of u. We apply the Itô formula to \(\phi (\xi ) = |\nabla \xi |_{L^2}^2\) and the process \(u_n(t)\), obtaining
where \(((\cdot , \cdot ))\) is as defined in (2.1).
Using Lemma 2.1, assumptions \((\mathbf{H}1 )\)–\((\mathbf{H}2 )\), boundedness of \(P_n\) in \(\mathrm {H}\), estimates (4.8), (4.12), the Cauchy-Schwarz and the Young’s inequality, we get
i.e.,
On rearranging, we have
Now since the process
is a \({\mathbb {F}}\)-martingale, as by Lemma 2.1 and (5.1) we have the following inequalities
where to establish the last inequality we have used the equivalences of norm from Proposition 4.9. Thus, \({\mathbb {E}}[ \mu _n(t)] = 0\).
Again as before, by applying Lemma 2.2 to processes
the inequalities (2.26) and (2.27) are satisfied. Thus, from (5.9) and (5.11), if \(\Vert u_0\Vert _\mathrm {V}\le \rho \), then
In particular,
From (5.14) and (5.13), we have the following estimate
Note that \(|u|^2_{\mathrm{D}(\mathrm {A})} := |u|^2_{L^2} + |\mathrm {A}\,u|^2_{L^2}\). Thus from (5.9) and (5.15) we can infer (5.3). On combining (5.9) and (5.14), we get
Using the Burkholder-Davis-Gundy inequality, the definition of \(\langle \cdot , \cdot \rangle _\mathrm {V}\) and the Young’s inequality, for every \( \varepsilon > 0\) we obtain
On combining (5.7) and (5.12), then using (5.3), (5.9), (5.14), (5.17) and Lemma 2.2, we can infer (5.2).
\(\square \)
In the next lemma we will use the estimates from Lemma 5.1 to establish higher order estimates.
Lemma 5.2
Let \(\tau _R^n\) be as defined in (5.1). For all \(\rho > 0\) and \(p \in [1,3]\) there exist positive constants \(C_1(p, \rho )\) , \(C_2(p, \rho )\) such that if \(\Vert u_0\Vert _\mathrm {V}\le \rho \), then
Proof
Let \(p \in [1, \infty )\). Then by using the Itô formula for \(\xi (t) = \Vert u_n(t)\Vert _{\mathrm{V}}^2\), \(\phi (x) = x^p,\) equations (5.5), (5.11) and the definition of \(\Vert \cdot \Vert _\mathrm {V}\), we obtain
Using Lemma 2.1, boundedness of \(P_n\) in \(\mathrm{V}\) and assumption \((\mathbf{H}1 )\), we can simplify (5.20)
On rearranging, we get
which on further simplification yields
As before we will show that
is a \({\mathbb {F}}\)-martingale. By Lemma 2.1 and (5.1) we have the following inequalities
where the finiteness of the integral follows from Proposition 4.9. Hence, \({\mathbb {E}}[ \mu _n(t)] = 0\).
Since \(|u_n(s)|_\mathrm {H}\le \Vert u_n(s)\Vert _\mathrm {V}\) and \(|\nabla u_n(s)|_{L^2} \le \Vert u_n(s)\Vert _\mathrm {V}\) on applying the modified version of the Gronwall Lemma (Lemma 2.2) for
and
we have
Using (5.22) in (5.20), we also obtain
Now we are left to show the estimate (5.18). Using the Burkholder- Davis- Gundy inequality, Lemma 2.1, we get
Using the definition of \(\mathrm {V}\)-norm, the Hölder inequality and the Young’s inequality, for \(\varepsilon > 0\) we obtain
Thus from (5.21) and using (5.23), (5.24) and Lemma 2.2, we have
Choosing \(\varepsilon \) small enough we get
\(\square \)
5.2 Tightness of Measures
For each \(n \in {\mathbb {N}}\), the solution \(u_n\) of the truncated equation (4.25) defines a measure \(law (u_n)\) on \(({\mathcal {Z}}_{T}, {\mathcal {T}})\), where \({\mathcal {Z}}_T\) was defined in (3.6) as
In this subsection we will prove that this sequence of measures defined on \({\mathcal {Z}}_{T}\) is tight.
Lemma 5.3
The set of measures \(\{law (u_n), n \in {\mathbb {N}}\}\) is tight on \(({\mathcal {Z}}_{T}, {\mathcal {T}})\).
Proof
We recall the definition of the stopping time, \(\tau _R^n\)
We will use Corollary 3.9 to prove the tightness of measures. According to estimates (5.2) and (5.3), conditions (a) and (b) are satisfied. Thus it is sufficient to prove that the sequence \((u_n)_{n \in {\mathbb {N}}}\) satisfies the Aldous condition \([\mathbf{A} ]\) in \(\mathrm {H}\). By (4.25), for \(t \in [0,T \wedge \tau _R^n]\) we have
Let \(s , t \in [0,T], s< t\) and \(\theta := t - s\). First we will establish estimates for each term of the above equality.
Ad.\(J_2^n\). Since \(\mathrm {A}: \mathrm {V}\rightarrow \mathrm {V}^\prime \), then by the Hölder inequality and (5.3), we have the following inequalities
Ad.\(J^n_3\). \(B: \mathrm {D}(\mathrm {A}) \times \mathrm {V}\rightarrow \mathrm {H}\) is bilinear and continuous and \(P_n : \mathrm {H}\rightarrow \mathrm {H}\) is bounded then by Lemma 4.5, the Cauchy-Schwarz inequality and (5.2) we have
Ad.\(J^n_4\). Since \(H^1 \hookrightarrow L^6\) then by the definition of g and estimate (5.18) (for \(p = 2\)), we have
Ad.\(J^n_5\). Using the assumption \(\mathbf {H1}\), (5.9) and the Cauchy-Schwarz inequality, we obtain the following inequalities
Ad.\(J^n_6\). Using the Itô isometry, Lemma 2.1 and (5.2), we obtain the following
Let us fix \(\kappa > 0\) and \(\varepsilon > 0\). By the Chebyshev inequality and estimates (5.26)–(5.29), we obtain
where \(i = 2, \dots , 5\). Let \(\delta _i = \dfrac{\kappa ^2}{c_i^2} \varepsilon ^2\). Then
By the Chebyshev inequality and (5.30), we have
Let \(\delta _6 = \dfrac{\kappa ^2}{c_6} \varepsilon \). Then
Since \([\mathbf{A} ]\) holds for each term \(J_i^n,~i= 1,2, \dots , 6\); we infer that it holds also for \((u_n)_{n \in {\mathbb {N}}}\). Thus, the proof of lemma can be concluded by invoking Corollary 3.9. \(\square \)
Now we will state the main theorem of this section.
Theorem 5.4
Let assumptions \((\mathbf {H1})\) and \((\mathbf {H2})\) be satisfied. Then there exists a martingale solution \(({\widehat{\Omega }}, \widehat{{\mathcal {F}}}, \widehat{{\mathbb {F}}}, \widehat{{\mathbb {P}}}, {\widehat{u}})\) of problem (2.17) such that
In the following subsection we will prove Theorem 5.4 in several steps.
5.3 Proof of Theorem 5.4
By Lemma 5.3 the set of measures \(\{law (u_n), n \in {\mathbb {N}}\}\) is tight on the space \(({\mathcal {Z}}_{T}, {\mathcal {T}})\) defined by (3.6). Hence by Corollary 3.11 there exists a subsequence \((n_k)_{k \in {\mathbb {N}}}\), a probability space \(({\widetilde{\Omega }}, \widetilde{{\mathcal {F}}}, \widetilde{{\mathbb {P}}})\) and, on this space, \({\mathcal {Z}}_{T}\)-valued random variables \({\widetilde{u}}, {\widetilde{u}}_{n_k}, k \ge 1\) such that
\({{\widetilde{u}}}_{{n}_{k}}\rightarrow {\widetilde{u}}\, \text{ in } \,{\mathcal {Z}}_{T}\)\(\widetilde{{\mathbb {P}}}\text{-a.s. }\) precisely means that
Let us denote the subsequence \(({\widetilde{u}}_{n_k})\) again by \(({{\widetilde{u}}_n})_{n \in {\mathbb {N}}}\).
By Theorem B.1, \({\mathcal {C}}([0,T]; \mathrm {H}_n)\) is a Borel subset of \({\mathcal {C}}([0,T];\mathrm{U}^\prime ) \cap L^2(0,T; \mathrm {H}_{loc})\). Since \(u_n \in {\mathcal {C}}([0,T]; \mathrm{H}_n)\), \({\mathbb {P}}\)-a.s., and \({{\widetilde{u}}_n}\), \(u_n\) have the same laws on \({\mathcal {Z}}_T\), thus
Since \({\mathcal {C}}([0,T]; \mathrm {V}) \cap {\mathcal {Z}}_T\) and \(L^2(0,T; \mathrm{D}(\mathrm {A})) \cap {\mathcal {Z}}_T\) are Borel subsets of \({\mathcal {Z}}_T\) (Theorem B.1) and \({{\widetilde{u}}_n}\) and \(u_n\) have the same laws on \({\mathcal {Z}}_T\); from (5.18) and (5.3), we have for \(p \in [1,3]\)
Also, \({\mathcal {C}}([0,T]; \mathrm {H}_n)\) is continuously embedded in \(L^4(0,T; L^4)\) and \({{\widetilde{u}}_n}\), \(u_n\) have same law \(\mu \) on \({\mathcal {C}}([0,T]; \mathrm {H}_n)\), therefore we have
Thus, by estimate (5.4) we infer
By inequality (5.35) we infer that the sequence \(({{\widetilde{u}}_n})\) contains a subsequence, still denoted by \(({{\widetilde{u}}_n})\) convergent weakly in \(L^2([0,T] \times {\widetilde{\Omega }}; \mathrm{D}(\mathrm {A}))\). Since by (5.32) \(\widetilde{{\mathbb {P}}}\)-a.s \({{\widetilde{u}}_n}\rightarrow {\widetilde{u}}\) in \({\mathcal {Z}}_{T}\), we conclude that \({\widetilde{u}} \in L^2([0,T] \times {\widetilde{\Omega }}; \mathrm{D}(\mathrm {A}))\), i.e.
Similarly by inequality (5.34) for \(p = 1\) we can choose a subsequence of \(({{\widetilde{u}}_n})\) convergent weak star in the space \(L^2({\widetilde{\Omega }}; L^\infty (0,T; \mathrm{V}))\) and, using (5.32), we infer that
For each \(n \ge 1\), let us consider a process \({\widetilde{M}}_n\) with trajectories in \({\mathcal {C}}([0,T]; \mathrm {H}_n)\) in particular in \({\mathcal {C}}([0,T];\mathrm{H})\) defined by
Lemma 5.5
\({\widetilde{M}}_n\) is a square integrable martingale with respect to the filtration \(\widetilde{{\mathbb {F}}}_n = (\widetilde{{\mathcal {F}}}_{n,t})\), where \(\widetilde{{\mathcal {F}}}_{n,t} = \sigma \{{\widetilde{u}}_n(s), s \le t\}\), with the quadratic variation
Proof
Indeed since \({{\widetilde{u}}_n}\) and \(u_n\) have the same laws, for all \(s, t \in [0,T]\), \(s \le t\), then for all bounded continuous functions h on \({\mathcal {C}}([0,s]; \mathrm {V}_w)\), and all \(\psi , \zeta \in \mathrm {V}_\gamma \)\((\gamma > \frac{d}{2})\), we have
and
\(\square \)
Lemma 5.6
Let us define a process \({\widetilde{M}}\) for \(t \in [0,T]\) by
Then \({\widetilde{M}}\) is a \(\mathrm {H}\)-valued continuous process.
Proof
Since \({\widetilde{u}} \in {\mathcal {C}}([0,T]; \mathrm {V})\) we just need to show that the remaining terms on the r.h.s. of (5.43) are \(\mathrm {H}\)-valued a.s. and well-defined.
Using the Cauchy-Schwarz inequality repeatedly and (5.37) we have the following inequalities
Since \(H^{k,p}({\mathbb {R}}^d) \hookrightarrow L^\infty ({\mathbb {R}}^d)\) for every \(k > d/p\), hence there exists a \(C > 0\) such that \(\Vert u\Vert _{L^\infty } \le C \,\Vert u\Vert _{H^{2,2}}\) for every \(u \in H^{2,2}({\mathbb {R}}^3)\). Thus by the Cauchy-Schwarz inequality, (5.37) and (5.38) we obtain the following estimate
We know that for \(d= 3\), \(H^{1,2} \hookrightarrow L^6\), thus using (2.15), (5.32) and (5.34), we get
Using the assumptions \((\mathbf{H}1 )\) and (5.38) we can show that
This concludes the proof of the lemma. \(\square \)
Lemma 5.7
Let \(\gamma > \frac{3}{2}\), \(u \in L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) and \((u_n)_{n \in {\mathbb {N}}}\) be a bounded sequence in \(L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) such that \(u_n \rightarrow u\) in \(L^2(0,T; \mathrm {H}_{loc})\). Then for all \(r,t \in [0,T]\) and all \(\psi \in \mathrm {V}_\gamma :\)
Here \(\langle \cdot , \cdot \rangle \) denotes the duality pairing between \(\mathrm {V}_\gamma \) and \(\mathrm {V}_{\gamma }^\prime \).
Proof
We will prove the lemma in two steps.
Step I
Let us fix \(\gamma > \frac{3}{2}\) and \(r,t \in [0,T]\). Assume first that \(\psi \in {\mathcal {V}}\). Then there exists a \(R > 0\) such that \(\text {supp}(\psi )\) is a compact subset of \({\mathcal {O}}_R\). There exists a constant \(C \ge 0\) such that
where we used (2.8) to establish the last inequality. We have
Thus using the estimate (5.45), the Hölder inequality, (2.16) and the Cauchy-Schwarz inequality, we obtain
Since \(u_n \rightarrow u\) in \(L^2(0,T; \mathrm {H}_{loc})\) we infer that (5.44) holds for every \(\psi \in {\mathcal {V}}\).
Step II
Let \(\psi \in \mathrm {V}_\gamma \) and \(\varepsilon > 0\). Then there exists a \(\psi _\varepsilon \in {\mathcal {V}}\) such that \(\Vert \psi _\varepsilon - \psi \Vert _{\mathrm {V}_\gamma } < \varepsilon .\) Hence, we get
Since \( {\mathcal {V}}\) is dense in \(\mathrm {V}_\gamma \), (5.45) holds for all \(\psi \in \mathrm {V}_\gamma \). In particular, there exists a constant \(C > 0\) such that
Using (5.46), (5.47) and the Cauchy-Schwarz inequality we have following inequalities
Hence by Step I and the assumptions on \(u, u_n\) there exists a \(M > 0\) such that
Since \(\varepsilon > 0\) is arbitrary we conclude the proof. \(\square \)
Corollary 5.8
Let \(\gamma > \frac{3}{2}\), \(u \in L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) and \((u_n)_{n \in {\mathbb {N}}}\) be a bounded sequence in \(L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) such that \(u_n \rightarrow u\) in \(L^2(0,T; \mathrm {H}_{loc})\). Then for all \(r,t \in [0,T]\) and all \(\psi \in \mathrm{V}_\gamma \)
Here \(\langle \cdot , \cdot \rangle \) denotes the dual pairing between \(\mathrm{V}_\gamma \) and \(\mathrm{V}_\gamma ^\prime \).
Proof
Let us fix \(\gamma > \frac{3}{2}\) and take \(r, t \in [0,T]\) and \(\psi \in \mathrm{V}_\gamma \). We have
We will consider each of these integrals individually. Using the estimate from (5.47), we have
Since the sequence \((u_n)_{n \in {\mathbb {N}}}\) is bounded in \(L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) and \(P_n \psi \rightarrow \psi \) in \(V_\gamma \), we infer that \(\lim _{n \rightarrow \infty } I_1(n) = 0\). By Lemma 5.7 we infer that
\(\square \)
Lemma 5.9
For all \(s, t \in [0,T]\) such that \(s \le t\) and \(\gamma > 3/2 :\)
-
(a)
\(\lim _{n \rightarrow \infty }\langle {\widetilde{u}}_n(t), P_n \psi \rangle = \langle {\widetilde{u}}(t), \psi \rangle ,~~ \widetilde{{\mathbb {P}}}\)-a.s., \( \psi \in \mathrm {V},\)
-
(b)
\(\lim _{n \rightarrow \infty } \int _s^t \langle \mathrm {A}{\widetilde{u}}_n(\sigma ), P_n \psi \rangle _\mathrm {H}\,ds = \int _s^t \langle \mathrm {A}{\widetilde{u}}(\sigma ), \psi \rangle _\mathrm {H}\,d \sigma ,~~ \widetilde{{\mathbb {P}}}\)-a.s., \(\psi \in \mathrm {H}\),
-
(c)
\(\lim _{n \rightarrow \infty } \int _s^t \langle B({\widetilde{u}}_n(\sigma )), P_n \psi \rangle \,d\sigma = \int _s^t\langle B({\widetilde{u}}(\sigma )), \psi \rangle \,d \sigma ,~~ \widetilde{{\mathbb {P}}}\)-a.s., \(\psi \in \mathrm {V}_\gamma \),
-
(d)
\(\lim _{n \rightarrow \infty } \int _s^t \langle g(|{\widetilde{u}}_n(\sigma )|^2) {\widetilde{u}}_n(\sigma ), P_n \psi \rangle \,d \sigma = \int _s^t \langle g(|{\widetilde{u}}(\sigma )|^2) {\widetilde{u}}(\sigma ), \psi \rangle \,d \sigma ,~ \widetilde{{\mathbb {P}}}\)-a.s., \(\psi \in \mathrm {V}_\gamma \),
-
(e)
\(\lim _{n \rightarrow \infty } \int _s^t \langle f({\widetilde{u}}_n(\sigma )), P_n \psi \rangle \,d \sigma = \int _s^t \langle f({\widetilde{u}}(\sigma )), \psi \rangle \,d \sigma ,~~ \widetilde{{\mathbb {P}}}\)-a.s., \(\psi \in \mathrm {V}_\gamma \),
where \(\langle \cdot , \cdot \rangle \) denotes the duality pairing between appropriate spaces.
Proof
Let us fix \(s, t \in [0,T]\), \(s \le t\) and \(\gamma > \frac{3}{2}\). By (5.32) we know that \({\mathbb {P}}\)-a.s.
Let \(\psi \in \mathrm {V}\). Since \({\widetilde{u}}_n \rightarrow {\widetilde{u}}\) in \({\mathcal {C}}([0,T]; \mathrm {V}_w)\)\(\widetilde{{\mathbb {P}}}\)-a.s., from (5.34) \({\widetilde{u}}_n\) is uniformly bounded in \({\mathcal {C}}([0,T]; \mathrm {V}_w)\) and \(P_n \psi \rightarrow \psi \) in \(\mathrm{V}\), thus
Hence we infer that assertion (a) holds.
Let \(\psi \in \mathrm {H}\). Since by (5.49) \({\widetilde{u}}_n \rightarrow {\widetilde{u}}\) in \(L^2_w(0,T; \mathrm{D}(\mathrm {A}) )\)\(\widetilde{{\mathbb {P}}}\)-a.s., from (5.35) \({{\widetilde{u}}_n}\) is uniformly bounded in \(L_w^2(0,T; \mathrm{D}(\mathrm {A}))\) and \(P_n \psi \rightarrow \psi \) in \(\mathrm{H}\). Thus, we have, \(\widetilde{{\mathbb {P}}}\)-a.s.,
Hence, we have shown that assertion (b) is true.
Assertion (c) follows directly for every \(\psi \in \mathrm {V}_\gamma \) from [7, Lemma B.1] and a modification of Corollary 5.8.
By (5.49) \({\widetilde{u}}_n \rightarrow {\widetilde{u}}\) in \(L^2(0,T; \mathrm {H}_{loc})\). From Lemma 5.1, (5.32) and (5.36) the sequence (\({\widetilde{u}}_n\)) is bounded in \(L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\) and \({\widetilde{u}} \in L^2(0,T; \mathrm {H}) \cap L^4(0,T; L^4)\). Thus, using Corollary 5.8 we infer that (d) holds for every \(\psi \in \mathrm {V}_\gamma \).
Now we are left to deal with (e). Let \(\psi \in \mathrm {V}_\gamma \),
Since \(\mathrm {V}_\gamma \hookrightarrow \mathrm {H}\), by the Cauchy-Schwarz inequality we have
Since \({\widetilde{u}}_n \rightarrow {\widetilde{u}}\) in \(L^2(0,T; \mathrm {H}_{loc})\) and \({\widetilde{u}}_n\) is a bounded sequence in \(L^2(0,T; \mathrm {H})\). \(I_1(n)\) can be shown to converge to zero as \(n \rightarrow \infty \) following the methodology of Lemma 5.7 and Corollary 5.8. Since \(P_n \psi \rightarrow \psi \) in \(\mathrm{V}_\gamma \), \(I_2(n) \rightarrow 0\) as \(n \rightarrow \infty \). This completes the proof of Lemma 5.9. \(\square \)
The proofs of Lemmas 5.10, 5.13 and 5.14 follow the similar methodology as that of Lemmas 5.6 - 5.8 [7] and Lemmas 5.9 - 5.11 [4].
Let h be the bounded continuous function on \({\mathcal {C}}([0,T]; \mathrm {V}_w)\).
Lemma 5.10
Let \(\gamma > \frac{3}{2}\). For all \(s, t \in [0,T]\), such that \(s \le t\) and all \(\psi \in \mathrm{V}_\gamma \)
Proof
Let us fix \(s, t \in [0,T], s \le t\) and \(\psi \in \mathrm{V}_\gamma \). By Eq. (5.39), we have
By Lemma 5.9, we infer that
In order to prove (5.50) we first observe that since \({{\widetilde{u}}_n}\rightarrow {\widetilde{u}}\) in \({\mathcal {Z}}_{T}\), in particular in \({\mathcal {C}}([0,T]; \mathrm {V}_w)\) and h is a bounded continuous function on \({\mathcal {C}}([0,T]; \mathrm {V}_w)\), we get
and
Let us define a sequence of \({\mathbb {R}}\)-valued random variables:
We will prove that the functions \(\{f_n\}_{n \in {\mathbb {N}}}\) are uniformly integrable in order to apply the Vitali’s convergence theorem. We claim that
Since, \(\mathrm {H}\hookrightarrow \mathrm {V}_\gamma ^\prime \) then by the Cauchy-Schwarz inequality, for each \(n \in {\mathbb {N}}\) we have
Since, \({\widetilde{M}}_n\) is a continuous martingale with quadratic variation defined in (5.40), by the Burkholder-Davis-Gundy inequality we obtain
Since, \(P_n :\mathrm {H}\rightarrow \mathrm {H}\) is a contraction and by Lemma 2.1, (5.18) for \(p = 1\), we have
Then by (5.55) and (5.57) we see that (5.54) holds. Since the sequence \(\{f_n\}_{n \in {\mathbb {N}}}\) is uniformly integrable and by (5.51) it is \(\widetilde{{\mathbb {P}}}\)-a.s. point-wise convergent, then application of the Vitali’s convergence theorem completes the proof of the Lemma. \(\square \)
Remark 5.11
Using Burkholder-Davis-Gundy inequality we have proved a stronger claim (5.56) than what we needed.
From Lemmas 5.5 and 5.10 we have the following corollary.
Corollary 5.12
For all \(s,t \in [0,T]\) such that \(s \le t :\)
Lemma 5.13
For all \(s,t \in [0,T]\) such that \(s \le t\) and all \(\psi , \zeta \in \mathrm{V}_\gamma \)
where \(\langle \cdot , \cdot \rangle \) denotes the appropriate duality pairing.
Proof
Let us fix \(s, t \in [0,T]\) such that \(s \le t\) and \(\psi , \zeta \in \mathrm {V}_\gamma \) and define \({\mathbb {R}}\)-valued random variables \(f_n\) and f for \(\omega \in {\widetilde{\Omega }}\) by
By Lemma 5.9 or more precisely by (5.51) and (5.52) we infer that \(\lim _{n \rightarrow \infty } f_n(\omega ) = f(\omega )\), for \(\widetilde{{\mathbb {P}}}\) almost all \(\omega \in {\widetilde{\Omega }}\).
We will prove that the functions \(\{f_n\}_{n \in {\mathbb {N}}}\) are uniformly integrable. We claim that for some \(r > 1\),
For each \(n \in {\mathbb {N}}\) as before we have
Since, \({\widetilde{M}}_n\) is a continuous martingale with quadratic variation defined in (5.40), by the Burkholder-Davis-Gundy inequality we obtain
Since, \(P_n :\mathrm {H}\rightarrow \mathrm {H}\) is a contraction and by Lemma 2.1 we have
Thus if \(r \in [1,3]\) then by (5.18), (5.53) and (5.59)–(5.61) we infer that (5.58) holds. Hence by application of the Vitali’s convergence theorem
\(\square \)
We will be using the following notations in following lemmata. \(\mathrm {V}^\prime ({\mathcal {O}}_R)\) is the dual space to \(\mathrm {V}({\mathcal {O}}_R)\), where
where
We recall that \(\mathrm {H}_{{\mathcal {O}}_R}\) is the space of restrictions to the subset \({{\mathcal {O}}_R}\) of elements of the space \(\mathrm {H}\) i.e.,
with the scalar product defined by
Lemma 5.14
The map \(G :\mathrm {H}_{{\mathcal {O}}_R} \rightarrow {\mathcal {L}}_2(\ell ^2; \mathrm {V}^\prime ({{\mathcal {O}}_R}))\) given by (2.13) is well defined and there exists some constant \(C_R > 0\) such that
Moreover, for every \(\psi \in {\mathcal {V}}\) the mapping \(\mathrm {H}\ni u \mapsto \langle G(u), \psi \rangle \in \ell ^2\) is continuous, if in the space \(\mathrm {H}\) we consider the Fréchet topology inherited from the space \(L^2_{loc}({\mathbb {R}}^3,{\mathbb {R}}^3)\).
Proof
Let \(\sigma =({\sigma }^1, ..., {\sigma }^{d}) : {\overline{{\mathcal {O}}}} \rightarrow {\mathbb {R}}^d\) and fix \(R > 0\). Let \(u \in {\mathcal {V}}({\mathcal {O}}_R)\). Then
Let \(\mathrm {v}\in {\mathcal {V}}({\mathcal {O}}_R)\). Since, \(\mathrm {v}\) on the boundary \(\partial {{\mathcal {O}}}_{R}\) is equal to zero, thus using the integration by parts formula, we obtain for \(\mathrm {v}\in {\mathcal {V}}({\mathcal {O}}_R)\)
Using the Hölder inequality, we obtain
Therefore, if we define a linear functional \({{\hat{B}}}_{R}\) by
we infer that it is bounded in the norm of the space \(\mathrm {V}({{\mathcal {O}}}_{R})\). Thus it can be uniquely extended to a linear bounded functional (denoted also by \({{\hat{B}}}_{R}\)) on \(\mathrm {V}({{\mathcal {O}}}_{R})\). Moreover, by estimate (5.65) we have the following inequality
or equivalently
Since by equality (2.13), \(G(u)(e_j) = \Pi \left[ (\sigma _j \cdot \nabla ) u\right] \), where \(\left\{ e_j\right\} _{j=1}^\infty \) is an orthonormal basis of \(\ell ^2\), we get by estimate (5.66)
Therefore, using the assumption \(\mathbf {(H2)}\), \(G(u) \in {\mathcal {L}}_2(\ell ^2,\mathrm {V}^{\prime }({{\mathcal {O}}}_{R}))\) and
By estimate (5.63) and the continuity of the embedding \({\mathcal {L}}_2(\ell ^2, \mathrm {V}^{\prime }({{\mathcal {O}}}_{R})) \hookrightarrow {\mathcal {L}}(\ell ^2, \mathrm {V}^{\prime }({{\mathcal {O}}}_{R}))\), we obtain
for some constant \(C(R)>0\). Thus, for any \(\psi \in \mathrm {V}({{\mathcal {O}}}_{R})\)
Now we identify \({}_{\mathrm {V}^\prime }\langle G(\cdot ), \psi \rangle _{\mathrm {V}}\) with the mapping \(\psi ^{**}G :\mathrm {H}\rightarrow (\ell ^2)^\prime \) defined by
Thus, from the inequality (5.67), we infer that
Therefore, if we fix \(\psi \in {\mathcal {V}}\) then, there exists \({R}_{0}>0\) such that \(\text{ supp }\, \psi \) is a compact subset of \({{\mathcal {O}}}_{{R}_{0}}\). Since G is linear, estimate (5.68) with \(R:={R}_{0}\) yields that the mapping
is continuous in the Fréchet topology inherited on the space \(\mathrm {H}\) from the space \({L}^{2}_{loc}({\mathbb {R}}^3 , {\mathbb {R}}^3 )\), concluding the proof of the lemma. \(\square \)
Lemma 5.15
(Convergence of quadratic variations) For any \(s, t \in [0, T]\) and \(\psi , \zeta \in \mathrm{V}_\gamma \), we have
where \(\langle \cdot , \cdot \rangle \) is the inner product in \(\ell ^2\).
Proof
Let us fix \(\psi , \zeta \in \mathrm {V}_\gamma \) and let us denote for \(\omega \in {\widetilde{\Omega }}\)
We will prove that the functions are uniformly integrable and convergent \(\widetilde{{\mathbb {P}}}\)-a.s. We start by proving that for some \(r > 1\),
Since, \({\mathcal {L}}_2(\ell ^2;H)\) is continuously embedded in \({\mathcal {L}}(\ell ^2; H)\), then by (2.20) there exists some \(c > 0\) such that
and thus
Using the Hölder inequality, we get
Thus
for some \({\widetilde{C}} > 0\). Hence by (5.34) for some \(r \in (1,3]\)
inferring (5.69).
Pointwise convergence Next, we have to prove the following pointwise convergence for a fix \(\omega \in {\widetilde{\Omega }}\), i.e. we will show that for a fix \(\omega \in {\widetilde{\Omega }}\)
Let us fix \(\omega \in {\widetilde{\Omega }}\) such that
-
(i)
\({{\widetilde{u}}_n}(\cdot ,\omega ) \rightarrow {\widetilde{u}}(\cdot ,\omega )\) in \({L}^{2}(0,T, {H}_{loc})\),
-
(ii)
\({\widetilde{u}}(\cdot ,\omega )\in {L}^{2}(0,T;H)\) and the sequence \(({{\widetilde{u}}_n}(\cdot ,\omega ){)}_{n\in {\mathbb {N}}}\) is bounded in \({\mathcal {C}}([0,T]; \mathrm {V})\).
Notice that, in order to prove (5.70) it is sufficient to prove that
We have
Let us consider the term \({I}_{1}(n)\). Since \(\psi \in \mathrm {V}_\gamma \), we have
By Lemma 2.1, the continuity of the embedding \({\mathcal {L}}_2 (\ell ^2, \mathrm {H}) \hookrightarrow {\mathcal {L}} (\ell ^2, \mathrm {H})\) and (ii), we infer that
for some constant \(K>0\). Thus
Let us move to the term \({I}_{2}(n)\) in (5.72). We will prove that for every \(\psi \in \mathrm {V}_\gamma \) the term \({I}_{2}(n)\) tends to zero as \(n \rightarrow \infty \). Assume first that \(\psi \in {\mathcal {V}} \). Then there exists \(R>0\) such that \(\text{ supp } \psi \) is a compact subset of \({{\mathcal {O}}}_{R}\). Since \({{\widetilde{u}}_n}(\cdot ,\omega ) \rightarrow {\widetilde{u}}(\cdot ,\omega )\) in \({L}^{2}(0,T;{H}_{loc})\), then in particular
where \({q}_{T,R}\) is the seminorm defined by (3.1). In other words, \({{\widetilde{u}}_n}(\cdot ,\omega ) \rightarrow {\widetilde{u}}(\cdot ,\omega )\) in \({L}^{2}(0,T;\)\({H}_{{{\mathcal {O}}}_{R}})\). Therefore, there exists a subsequence \(({{\widetilde{u}}}_{{n}_{k}}(\cdot ,\omega ){)}_{k} \) such that
Hence by Lemma 5.14 as \(k \rightarrow \infty \)
In conclusion, by the Vitali’s convergence theorem
Repeating the above reasoning for all subsequences, we infer that from every subsequence of the sequence \(\bigl ( G \bigl ( \sigma , {{\widetilde{u}}_n}(\sigma ,\omega ){\bigr ) }^*\psi {\bigr ) }_{n}\) we can choose the subsequence convergent in \({L}^{2}(s,t; \ell ^2)\) to the same limit. Thus, the whole sequence \(\bigl ( G \bigl ( \sigma , {{\widetilde{u}}_n}(\sigma ,\omega ){\bigr ) }^*\psi {\bigr ) }_{n}\) is convergent to \(G \bigl (\sigma , {\widetilde{u}}(\sigma ,\omega ){\bigr ) }^*\psi \) in \({L}^{2}(s,t; \ell ^2)\). At the same time
If \(\psi \in \mathrm {V}_\gamma \) then for every \(\varepsilon > 0\) we can find \({\psi }_{\varepsilon } \in {\mathcal {V}} \) such that \(\Vert \psi - {\psi }_{\varepsilon }\Vert _{V_\gamma } < \varepsilon \). By the continuity of embeddings \({\mathcal {L}}_2 (\ell ^2, \mathrm {H}) \hookrightarrow {\mathcal {L}} (\ell ^2, \mathrm {H}) \hookrightarrow {\mathcal {L}}(\ell ^2, \mathrm {V}_\gamma ^\prime )\), Lemma 2.1 and (ii), we obtain
for some positive constants c and C. Passing to the upper limit as \(n \rightarrow \infty \), we infer that
In conclusion, we proved that
which completes the proof of (5.71). Thus, by (5.69), convergence (5.70) and the Vitali’s convergence theorem, we conclude the proof of Lemma 5.15. \(\square \)
By Lemma 5.10 we can pass to the limit in (5.41). By Lemmas 5.13 and 5.15 we can pass to the limit in (5.42) as well. After passing to the limits we infer that for all \(\psi , \zeta \in \mathrm{V}_\gamma \) and all bounded continuous functions h on \({\mathcal {C}}([0,T]; \mathrm {V}_w)\):
and
where \(\langle \cdot , \cdot \rangle \) is the dual pairing between \(\mathrm {V}_\gamma ^\prime \) and \(\mathrm {V}_\gamma \).
From Lemmas 5.5, 5.13 and 5.15, we infer the following corollary.
Corollary 5.16
For \(t \in [0,T]\)
Continuation of the proof of Theorem 5.4. Now we apply the idea analogous to that used by Da Prato and Zabczyk, see [12, Section 8.3]. By Lemma 5.6, and Corollary 5.12, we infer that \({\widetilde{M}}(t), t \in [0, T]\) is an \(\mathrm {H}\)-valued continuous square integrable martingale with respect to the filtration \(\widetilde{{\mathbb {F}}} = (\widetilde{{\mathcal {F}}_t})\). Moreover, by Corollary 5.16 the quadratic variation of \({\widetilde{M}}\) is given by
Therefore by the martingale representation theorem, there exist
-
a stochastic basis \((\widetilde{{\widetilde{\Omega }}}, \widetilde{\widetilde{{\mathcal {F}}}}, \widetilde{\widetilde{{\mathbb {F}}}}, \widetilde{\widetilde{{\mathbb {P}}}})\),
-
a cylindrical Wiener process \(\widetilde{{\widetilde{W}}}(t)\),
-
and a progressively measurable process \({\widetilde{{\widetilde{u}}}}(t)\) such that for all \(t \in [0,T]\) and all \(\mathrm {v}\in {\mathcal {V}} \)
$$\begin{aligned}&\langle {\widetilde{{\widetilde{u}}}}(t) , \mathrm {v}\rangle - \langle \widetilde{{\widetilde{u}}}(0) , \mathrm {v}\rangle + \int _0^t \langle \mathrm {A}{\widetilde{{\widetilde{u}}}}(s), \mathrm {v}\rangle \,ds + \int _0^t \langle B({\widetilde{{\widetilde{u}}}}(s), {\widetilde{{\widetilde{u}}}}(s)), \mathrm {v}\rangle \, ds \\&\quad = \int _0^t \langle f({\widetilde{{\widetilde{u}}}}(s)) - g(|{\widetilde{{\widetilde{u}}}}(s)|^2) {\widetilde{{\widetilde{u}}}}(s), \mathrm {v}\rangle \,ds + \left\langle \int _0^t G(s, {\widetilde{{\widetilde{u}}}}(s))\,d\widetilde{{\widetilde{W}}}(s), \mathrm {v}\right\rangle . \end{aligned}$$
Thus, the conditions from Definition 3.13 hold with \(({\widehat{\Omega }}, \widehat{{\mathcal {F}}}, \widehat{{\mathbb {F}}}, \widehat{{\mathbb {P}}}) = (\widetilde{{\widetilde{\Omega }}}, \widetilde{\widetilde{{\mathcal {F}}}}, \widetilde{\widetilde{{\mathbb {F}}}}, \widetilde{\widetilde{{\mathbb {P}}}})\), \({\widehat{W}} = \widetilde{{\widetilde{W}}}\) and \({\widehat{u}} = \widetilde{{\widetilde{u}}}\). The proof of Theorem 5.4 is thus complete.
5.4 Uniqueness and Strong Solutions
In this subsection we will show that the solutions of (2.17) are pathwise unique and that the martingale solution of (2.17) is the strong solution. Let us recall the definition of pathwise unique solutions.
Definition 5.17
Let \((\Omega , {\mathcal {F}}, {\mathbb {F}}, {\mathbb {P}}, W, u^i), i = 1,2\) be the martingale solutions of (2.17) with \(u^i(0) = u_0, i= 1,2\). Then we say that the solutions are pathwise unique if \({\mathbb {P}}\)-a.s. for all \(t \in [0,T],\)\(u^1(t) = u^2(t)\).
Theorem 5.18
Assume that the assumptions \((\mathbf{H1})\) and \((\mathbf{H2})\) are satisfied. If \(u_1, u_2\) are two solutions of (2.17) defined on the same filtered probability space \(({\widehat{\Omega }}, \widehat{{\mathcal {F}}}, \widehat{{\mathbb {F}}}, \widehat{{\mathbb {P}}})\) then \(\widehat{{\mathbb {P}}}\)-a.s. for all \(t \in [0,T]\), \(u_1(t) = u_2(t)\).
This theorem has been proved in [30, Theorem 3.7].
Theorem 5.19
Assume that assumptions \((\mathbf {H1})\) and \((\mathbf {H2})\) are satisfied. Then there exists a path-wise unique strong solution \(u \in {\mathcal {C}}([0,T]; \mathrm {V}) \cap L^2(0,T; \mathrm{D}(\mathrm {A}))\) of (2.17) such that
Proof
Since by Theorem 5.4 there exists a martingale solution and by Theorem 5.18 it is pathwise unique, the existence of strong solution follows from [27, Theorem 2]. Moreover, it can be shown that \(u \in {\mathcal {C}}([0,T];\mathrm {V}) \cap L^2(0,T; \mathrm {D}(\mathrm {A}))\) by following the proof of [4, Lemma 3.6]. \(\square \)
6 Invariant Measures
In this section, we consider time homogeneous damped tamed NSEs, i.e. the coefficients \(f, \sigma \) are independent of t and furthermore \(f \in \mathrm {H}\) is not dependent on u. The time homogeneous damped tamed NSEs in abstract form are given by
![](http://media.springernature.com/lw408/springer-static/image/art%3A10.1007%2Fs00021-020-0480-z/MediaObjects/21_2020_480_Equ142_HTML.png)
where \(\mathrm {A}_\alpha = \alpha I - \nu \Delta \) for some \(\alpha \in {\mathbb {R}}\) and \(\nu > 0\) is the viscosity. The operator B and the cylindrical Wiener process \(W = \left( W_j\right) _{j=1}^\infty \) on \(\ell ^2\) is same as defined in Sect. 2 and \(G_j\) are as defined in (2.12).
Let \( {\mathcal {B}}_b(\mathrm {V})\) denote the set of all bounded and Borel measurable functions on \(\mathrm {V}\). For any \(\varphi \in {\mathcal {B}}_b(\mathrm {V})\), \(t \ge 0\), we define a function \(T_t \varphi :\mathrm {V}\rightarrow {\mathbb {R}}\) by
It follows from Theorem 6.2 and Ondrejat [28] (see also [5]) that \(T_t \varphi \in {\mathcal {B}}_b(\mathrm {V})\) and \(\left\{ T_t\right\} _{t \ge 0}\) is a semigroup on \({\mathcal {B}}_b(\mathrm {V})\). Also since this unique solution to (6.1) has a.e. path in \({\mathcal {C}}([0,T]; \mathrm {V})\), it is also a Markov semigroup (see [28, Theorem 27]). Moreover, \(\left\{ T_t \right\} _{t \ge 0}\) is a Feller semigroup, i.e. \(T_t\) maps \(C_b(\mathrm {V})\) into itself.
Next we state the main result of this section, regarding the existence of an invariant measure.
Theorem 6.1
Let for every \(\alpha > 0\), the assumptions \((\mathbf {H1})^\prime - (\mathbf {H3})^\prime \) be satisfied. Then there exists an invariant measure \(\mu \in {\mathcal {P}}(\mathrm {V})\) of the semigroup \((T_t)_{t \ge 0}\) defined by (6.2), such that for any \(t \ge 0\) and \(\varphi \in \mathrm{SC}_{b}(\mathrm {V}_w)\)
If \(T_t\) is sequentially weakly Feller Markov semigroup then for every \(\varphi \in \mathrm{SC}_b(\mathrm {V}_w)\), \(T_t \varphi \in \mathrm{SC}_b(\mathrm {V}_w) \subset B_b(\mathrm {V})\) (see [5, 22] for the definitions and inclusion of the spaces); therefore the integral on the l.h.s. in Theorem 6.1 makes sense.
Next we list the assumptions that we make on the coefficients f and \(\sigma \) along with a coercivity type assumption, see [29].
- (H1)\(^\prime \):
-
The function \(f:{\mathbb {R}}^3 \rightarrow {\mathbb {R}}^3\) is time independent and \(\mathrm {H}\)-valued.
- (H2)\(^\prime \):
-
A measurable function \(\sigma :{\mathbb {R}}^3 \rightarrow \ell ^2 \) of \(C^1\) class with respect to the x-variable and for all \(x \in {\mathbb {R}}^3\) there exists a constant \(C_\sigma > 0\) such that
$$\begin{aligned} \Vert \partial _{x^j} \sigma (x)\Vert _{\ell ^2} \le C_{\sigma }, \quad j = 1,2,3 \end{aligned}$$and, for all \(x \in {\mathbb {R}}^3\),
$$\begin{aligned} \Vert \sigma (x)\Vert ^2_{\ell ^2} \le \frac{1}{4}\,. \end{aligned}$$ - (H3)\(^\prime \):
-
there exists a \(\delta > 0\) such that
$$\begin{aligned} 2 \nu |\nabla u |^2_{L^2} - \Vert G(u)\Vert ^2_{{\mathcal {L}}_2(\ell ^2; \mathrm {H})} \ge 2\delta |\nabla u|^2_{L^2}, \quad u \in \mathrm {V}\,. \end{aligned}$$
The following theorem regarding the existence of a pathwise unique strong solution to the time homogeneous damped tamed NSEs (6.1) can be proved by modifying the proofs of Theorem 5.4 and Theorem 5.18 to incorporate the extra linear term \(\alpha \,u\).
Theorem 6.2
Assume that assumptions \((\mathbf {H1})^\prime \) and \((\mathbf {H2})^\prime \) are satisfied. Then for every \(u_0 \in \mathrm {V}\), there exists a path-wise unique strong solution u of (6.1) for every \(T > 0\) such that \(u \in {\mathcal {C}}([0,T]; \mathrm {V}) \cap L^2(0,T; \mathrm {D}(\mathrm {A}))\), \({\mathbb {P}}\)-a.s.
For fixed initial value \(u_0 = \mathrm {v}\in \mathrm {V}\) we denote the (path-wise) unique solution of (6.1), whose existence is proved in Theorem 6.2 by \(u(t;\mathrm {v})\).
Definition 6.3
We say that a family \(\{T_t\}_{t \ge 0}\) is sequentially weakly Feller iff
For a metric space \({\mathbb {U}}\), we use \({\mathcal {P}}({\mathbb {U}})\) to denote the family of all Borel probability measures on \({\mathbb {U}}\). We will use the following theorem from Maslowski-Seidler [22] to prove the existence of invariant measures.
Theorem 6.4
Assume that
-
(i)
the semigroup \(\{T_t\}_{t \ge 0}\), defined by (6.2) is sequentially weakly Feller in \(\mathrm {V}\),
-
(ii)
for any \(\varepsilon > 0\) there exists \(R > 0\) such that
$$\begin{aligned} \sup _{ T \ge 1} \frac{1}{T} \int _0^T {\mathbb {P}}(\{\Vert u(t;u_0)\Vert _\mathrm {V}> R\})\,dt < \varepsilon \,. \end{aligned}$$
Then there exists at least one invariant measure for (6.1).
6.1 Boundedness in Probability
Lemma 6.5
Let \(u_0 \in \mathrm {V}\). Then, under the assumptions of Theorem 6.1, for every \(\varepsilon > 0\), there exists \(R > 0\) such that
Proof
Using the Itô formula for the function \(\phi (\xi ) = |\xi |^2_\mathrm {H}\) and the process u(t), we have
Now we deal with each term on the r.h.s. of (6.4) one by one. Firstly let us notice that we have
Using the assumptions on f, for any \(\beta > 0\) we obtain the following estimate
Since u is the solution of (6.1) and satisfies the estimates (4.24) (courtsey, Theorem 6.2), we can show that the process
is a martingale. Thus, taking expectation in (6.4) and using the estimates (6.5) – (6.8), we infer
On rearranging, we get
Now using the assumption \((\mathbf{H}3 )^\prime \) in (6.9), we obtain
Choosing \(\beta = \frac{\alpha }{2}\) yields
Therefore for \(\gamma = \min {\{\frac{\alpha }{2}, \delta \}} > 0\),
Thus, for any \(T > 0\), we infer that
Using the Chebyshev inequality and inequality (6.10), we infer that for every \(T \ge 0\)
Now for sufficiently large \(R > 0\) depending on \(\varepsilon , |u_0|_\mathrm {H}\) and \(|f|_{\mathrm {H}}\), the assertion follows. \(\square \)
6.2 Sequentially Weak Feller Property
We are left to verify the assumption (i) of Theorem 6.4, i.e. the Markov semigroup \(\{T_t\}_{t \ge 0}\) is sequentially weakly Feller in \(\mathrm {V}\). In other words we want to show that for any \(t > 0\) and any bounded and weakly continuous \(\varphi : \mathrm {V}\rightarrow {\mathbb {R}}\), if \(\xi _n \rightarrow \xi \) weakly in \(\mathrm {V}\), then
The second named author in his PhD thesis proved that the martingale solutions of stochastic constrained Navier–Stokes equations continuously depend on the initial data [13, Theorem 5.7.7]. We have a similar result for time homogeneous damped tamed NSEs, which can be proved analogously, see also [9, Theorem 4.11].
Theorem 6.6
Assume that \((u_{0,n})_{n=1}^\infty \) is a \(\mathrm {V}\)-valued sequence that is convergent weakly to \(u_0 \in \mathrm {V}\). Let
be a martingale solution of problem (6.1) on \([0, \infty )\) with the initial data \({u}_{0,n}\). Then for every \(T > 0\) there exist
-
a subsequence \(({n}_{k}{)}_{k}\),
-
a stochastic basis \(\bigl ( {\widetilde{\Omega }}, {\widetilde{{\mathcal {F}}}}, {\widetilde{{\mathbb {F}}}} , \widetilde{{\mathbb {P}}}\bigr ) \),
-
an \(\ell ^2\)-valued cylindrical Wiener process \({\widetilde{W}}(t) = \left( {\widetilde{W}}^j(t)\right) _{j=1}^\infty \),
-
and \(\widetilde{{\mathbb {F}}}\)-progressively measurable processes \({\widetilde{u}}\), \(\big ({{\widetilde{u}}}_{{n}_{k}}\big )_{k \ge 1} \) (defined on this basis) with laws supported in \( {\mathcal {Z}}_{T}\) such that
$$\begin{aligned} {{\widetilde{u}}}_{{n}_{k}} \text{ has } \text{ the } \text{ same } \text{ law } \text{ as } {u}_{{n}_{k}} \text{ on } {\mathcal {Z}}_{T} \text{ and } {{\widetilde{u}}}_{{n}_{k}}\rightarrow {\widetilde{u}} \text{ in } {\mathcal {Z}}_{T}, \quad \widetilde{{\mathbb {P}}}\text{-a.s. } \end{aligned}$$(6.12)and the system
$$\begin{aligned} \bigl ( {\widetilde{\Omega }}, {\widetilde{{\mathcal {F}}}}, {\widetilde{{\mathbb {F}}}}, \widetilde{{\mathbb {P}}}, {\widetilde{W}},{\widetilde{u}}\bigr ) \end{aligned}$$is a martingale solution to problem (6.1) on the interval [0, T] with the initial data \(u_0\). In particular, for all \(t \in [0,T]\) and all \(\mathrm {v} \in {\mathcal {V}} \)
$$\begin{aligned}&{\langle {\widetilde{u}}(t) , \mathrm {v} \rangle }_{} + \int _{0}^{t} {\langle \mathrm {A}_\alpha {\widetilde{u}}(s) , \mathrm {v} \rangle }_{} \, ds + \int _{0}^{t} {\langle B({\widetilde{u}}(s)) , \mathrm {v} \rangle }_{} \, ds \nonumber \\&\qquad + \int _{0}^{t} {\langle g(|{\widetilde{u}}(s)|^2) {\widetilde{u}}(s) , \mathrm {v} \rangle }_{} \, ds \nonumber \\&\quad = {\langle {\widetilde{u}}(0) , \mathrm {v} \rangle }_{\mathrm {V}} + \int _{0}^{t} {\langle f , \mathrm {v} \rangle }_{} \, ds + {\Bigl < \int _{0}^{t} \sum _{j=1}^\infty G_j(s,{\widetilde{u}}(s))\,dW^j(s) , \mathrm {v} \Bigr >}_{}, \quad {\widehat{{\mathbb {P}}}}\text{-a.s. } \end{aligned}$$Moreover, the process \({\widetilde{u}}\) satisfies the following inequality
$$\begin{aligned} \widetilde{{\mathbb {E}}}\,\left[ \,\sup _{ s\in [0,T] } {\Vert {\widetilde{u}}(s)\Vert _\mathrm {V}}^{2}\, + \int _{0}^{T} |{\widetilde{u}}(s)|^2_{\mathrm{D}(\mathrm {A})}\, ds \,\right] < \infty \,. \end{aligned}$$(6.13)
We will need the uniqueness in law of solutions of (6.1), which is defined next.
Definition 6.7
Let \((\Omega ^i, {\mathcal {F}}^i, {\mathbb {F}}^i, {\mathbb {P}}^i, W^i, u^i), i = 1,2\) be martingale solutions of (6.1) with \(u^i(0) = u_0, i=1,2\). Then we say that the solutions are unique in law if
where \(law_{{\mathbb {P}}^i}(u^i), i = 1,2\) are by definition probability measures on \({\mathcal {C}}([0, \infty ); \mathrm {V}_w) \cap L^2([0, \infty ); \mathrm {D}(\mathrm {A}))\).
Lemma 6.8
Assume that assumptions \((\mathbf {H1})^\prime - (\mathbf {H3})^\prime \) are satisfied. Then the martingale solution of (6.1) is unique in law.
The proof of the above lemma is the direct application of Theorems 2 and 11 of [27] once we have proved the existence of a pathwise unique martingale solution of (6.1); which follows from Theorem 6.2.
Lemma 6.9
The semigroup \(\{T_t\}_{t \ge 0}\) is sequentially weakly Feller in \(\mathrm {V}\).
Proof
Let us choose and fix \( 0 < t \le T, \xi \in \mathrm {V}\) and \(\varphi :\mathrm {V}\rightarrow {\mathbb {R}}\) be a bounded weakly continuous function. Need to show that \(T_t \varphi \) is sequentially weakly Feller in \(\mathrm {V}\). For this aim let us choose a \(\mathrm {V}\)-valued sequence \((\xi _n)\) weakly convergent to \(\xi \) in \(\mathrm {V}\). Since the function \(T_t \varphi :\mathrm {V}\rightarrow {\mathbb {R}}\) is bounded, we only need to prove (6.11).
Let \(u_n(\cdot ) = u(\cdot ; \xi _n)\) be a strong solution of (6.1) on [0, T] with the initial data \(\xi _n\) and let \(u(\cdot ) = u(\cdot ; \xi )\) be a strong solution of (6.1) with the initial data \(\xi \). We assume these processes are defined on the stochastic basis \((\Omega , {\mathcal {F}}, {\mathbb {F}}, {\mathbb {P}})\). By Theorem 6.6 there exist
-
a subsequence \((n_k)_k\),
-
a stochastic basis \(({\widetilde{\Omega }}, \widetilde{{\mathcal {F}}}, \widetilde{{\mathbb {F}}}, \widetilde{{\mathbb {P}}})\), where \(\widetilde{{\mathbb {F}}} = \{\widetilde{{\mathcal {F}}}_s\}_{s \in [0,T]}\),
-
an \(\ell ^2\)-valued cylindrical Wiener process \({\widetilde{W}}(t)\) on \(({\widetilde{\Omega }}, \widetilde{{\mathcal {F}}}, \widetilde{{\mathbb {F}}}, \widetilde{{\mathbb {P}}})\),
-
and progressively measurable processes \({\widetilde{u}}(s), ({\widetilde{u}}_{n_k}(s))_{k \ge 1}\), \(s \in [0,T]\) (defined on this basis) with laws supported in \({\mathcal {Z}}_T\), where
$$\begin{aligned} {\mathcal {Z}}_T = {\mathcal {C}}([0,T]; \mathrm {U}^\prime ) \cap L^2_w(0,T; \mathrm {D}(\mathrm {A})) \cap L^2(0,T; \mathrm {H}_{loc}) \cap {\mathcal {C}}([0,T]; \mathrm {V}_w), \end{aligned}$$such that
$$\begin{aligned} {\widetilde{u}}_{n_k} \text{ has } \text{ the } \text{ same } \text{ law } \text{ as } u_{n_k} \text{ on } {\mathcal {Z}}_T \text{ and } {\widetilde{u}}_{n_k} \rightarrow {\widetilde{u}} \text{ in } {\mathcal {Z}}_T, \; \widetilde{{\mathbb {P}}}\text{-a.s. } \end{aligned}$$(6.14)and the system
$$\begin{aligned} ({\widetilde{\Omega }}, \widetilde{{\mathcal {F}}}, \widetilde{{\mathbb {F}}}, \widetilde{{\mathbb {P}}}, {\widetilde{W}}, {\widetilde{u}}) \end{aligned}$$(6.15)is a martingale solution to (6.1) on the interval [0, T] with the initial data \(\xi \).
In particular, by (6.14), \(\widetilde{{\mathbb {P}}}\)-almost surely
Since the function \(\varphi : \mathrm {V}\rightarrow {\mathbb {R}}\) is sequentially weakly continuous, we infer that \(\widetilde{{\mathbb {P}}}\)-a.s.,
Since the function \(\varphi \) is also bounded, by the Lebesgue dominated convergence theorem we infer that
From the equality of laws of \({\widetilde{u}}_{n_k}\) and \(u_{n_k}, k \in {\mathbb {N}}\), on the space \({\mathcal {Z}}_T\) we infer that \({\widetilde{u}}_{n_k}\) and \(u_{n_k}\) have same laws on \(\mathrm {V}_w\) and so
On the other hand, the r.h.s. of (6.17) is equal by (6.2), to \(T_t\varphi (\xi _{n_k})\).
Since u, by assumption, is a martingale solution of (6.1) with the initial data \(\xi \) and from the above, \({\widetilde{u}}\) is also a solution of (6.1) with the initial data \(\xi \). Thus, by Lemma 6.8, we infer that
Hence
As before, the r.h.s. of (6.18) is equal by (6.2), to \(T_t\varphi (\xi )\). Thus, by (6.16), (6.17) and (6.18), we infer
Using the subsequence argument, we infer that the whole sequence \((T_t \varphi (\xi _n))_{n \in {\mathbb {N}}}\) is convergent and
\(\square \)
Proof of Theorem 6.1
The existence of an invariant measure is established by using Theorem 6.4, Lemmas 6.5 and 6.9. Hence the proof of Theorem 6.1 is complete. \(\square \)
Remark 6.10
It has been suggested to the authors by the referee, that there should exist an invariant measure for the original non-damped, tamed 3-D Navier–Stokes equations (1.1-1.3). We agree with the premises of this suggestion for which we are grateful, yet we have decided to postpone a detailed study of this issue till another publication. Note that the existence of a unique invariant measure for the non-damped tamed 3-D Navier–Stokes equations was established by Röckner and Zhang [30] in a bounded domain (3-D torus).
Notes
i.e. the equation is satisfied in the weak sense and the solution u belongs to the space \( L^\infty (0,T; \mathrm {H})\cap L^2(0,T; \mathrm {V}) \cap L^{\beta +1}(0,T; L^{\beta + 1})\).
A pair of function (u, p) is a strong solution iff it is a weak solution and \(u \in L^\infty (0,T; \mathrm {V}) \cap L^2(0,T; H^2) \cap L^\infty (0,T; L^{\beta +1})\).
One can show that the constant \({{\widetilde{C}}}_N\) is independent of N. Moreover, \(|\phi ^\prime (r)| \le 1\) and thus \(|\phi ^\prime (r)r| \le (N+1)\) for every \(r > 0\).
\(\widetilde{{\mathcal {T}}}\) is the supremum of topologies \({\mathcal {T}}_1, {\mathcal {T}}_2, {\mathcal {T}}_3\), i.e. it is the coarsest topology on \({\widetilde{{\mathcal {Z}}}}_T\) that is finer than each of \({\mathcal {T}}_1, {\mathcal {T}}_2\) and \({\mathcal {T}}_3\).
References
Albeverio, S., Brzeźniak, Z., Wu, J.: 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, 309–322 (2010)
Aldous, D.: Stopping times and tightness. Ann. Probab. 6(2), 335–340 (1978)
Badrikian, A.: Séminaire sur les fonctions aléatoires linéaires et les mesures cylindriques. Lecture Notes in Mathematics, vol. 139. Springer, Berlin (1970)
Brzeźniak, Z., Dhariwal, G.: Stochastic Constrained Navier–Stokes Equations on \({\mathbb{T}}^2\). arxiv:1701.01385v2 (Submitted 2019)
Brzeźniak, Z., Ferrario, B.: Stationary solutions for stochastic damped Navier–Stokes equations in \({\mathbb{R}}^d\). Indiana Univ. Math. J. 68(1), 105–138 (2019)
Brzeźniak, Z., Hornung, F., Weis, L.: Martingale solutions for the stochastic nonlinear Schrödinger equation in the energy space. Probab. Theory Relat. Fields 174, 1273–1338 (2019)
Brzeźniak, Z., Motyl, E.: Existence of a martingale solution of the stochastic Navier–Stokes equations in unbounded 2D and 3D-domains. J. Differ. Equ. 254(4), 1627–1685 (2013)
Brzeźniak, Z., Motyl, E.: The existence of martingale solutions to the stochastic Boussinesq equations. Glob. Stoch. Anal. 1(2), 175–216 (2014)
Brzeźniak, Z., Motyl, E., Ondreját, M.: Invariant measure for the stochastic Navier–Stokes equations in unbounded 2D domains. Ann. Probab. 45(5), 3145–3201 (2017)
Brzeźniak, Z., Ondreját, M.: Stochastic geometric wave equations with values in compact Riemanninan homogeneous spaces. Ann. Probab. 41(3B), 1938–1977 (2013)
Cai, Z., Jiu, Q.: Weak and strong solutions for the incompressible Navier–Stokes equations with damping. J. Math. Anal. Appl. 343, 799–809 (2008)
Da Prato, G., Zabczyk, J.: Stochastic Equations in Infinite dimensions. Cambridge University Press, Cambridge (1992)
Dhariwal, G.: A study of constrained Navier–Stokes equations and related problems, PhD Thesis, University of York (2017)
Duan, J., Millet, A.: Large deviations for the Boussinesq equations under random influences. Stoch. Process. Appl. 119(6), 2052–2081 (2009)
Fefferman, C.L., McCormick, D.S., Robinson, J.C., Rodrigo, J.L.: Higher order commutator estimates and local existence for the non-resistive MHD equations and related models. J. Funct. Anal. 267, 1035–1056 (2014)
Jakubowski, A.: The almost sure Skorokhod representation for subsequences in nonmetric spaces, Teor. Veroyatn. Primen. 42(1), 209–216 (1997); translation in Theory Probab. Appl. 42(1), 167–174 (1998)
Kato, T., Ponce, G.: Well-posedness of the Euler and Navier–Stokes equations in the Lebesgue spaces \(L^p_s({\mathbb{R}}^2)\). Rev. Mat. Iberoam. 2(1–2), 73–88 (1986)
Kuratowski, K.: Topologie, Vol. I (French) 3’eme ed. Monografie Matematyczne, Tom XX, Polskie Towarzystwo Matematyczne, Warszawa (1952)
Lions, J.L.: Quelques méthodes de résolution des problèmes aux limites non linéaires. Dunod, Paris (1969)
Manna, U., Panda, A.A.: Higher order regularity and blow-up criterion for semi-dissipative and ideal Boussinesq equations. J. Math. Phys. 60(4), 22 (2019)
Marcinkiewicz, J.: Sur les multiplicateurs des séries de Fourier. Studia Math. 8, 78–91 (1939)
Maslowski, B., Seidler, J.: On sequentially weakly Feller solutions to SPDE’s. Rend. Lincei Mat. Appl. 10, 69–78 (1999)
Métivier, M.: Semimartingales. Gruyter, Berlin (1982)
Métivier, M.: Stochastic Partial Differential Equations in Infinite Dimensional Spaces. Scuola Normale Superiore, Pisa (1988)
Mihlin, S.G.: On the multipliers of Fourier integrals. Dokl. Akad. Nauk SSSR (N.S.) 109, 701–703 (1956)
Mikulevicius, R., Rozovskii, B.L.: Global \(L_2-\)solutions of stochastic Navier–Stokes equations. Ann. Probab. 33(1), 137–176 (2005)
Ondreját, M.: Uniqueness for stochastic evolution equations in Banach spaces. Diss. Math. 426, 1–63 (2004)
Ondreját, M.: Brownian representations of cylindrical local martingales, martingale problems and strong Markov property of weak solutions of SPDEs in Banach spaces. Czechoslov. Math. J. 55(130), 1003–1039 (2005)
Pardoux, E.: Stochastic partial differential equations and filtering of diffusion processes. Stochastics 3, 127–167 (1979)
Röckner, M., Zhang, X.: Stochastic tamed 3D Navier–Stokes equation: existence, uniqueness and ergodicity. Probab. Theory Relat. Fields 145, 211–267 (2009)
Röckner, M., Zhang, X.: Tamed 3D Navier–Stokes equation: existence, uniqueness and regularity. Infin. Dimens. Anal. Quantum Probab. Relat. Top. 12(4), 525–549 (2009)
Röckner, M., Zhang, T.: Stochastic 3D tamed Navier–Stokes equations: existence, uniqueness and small time large deviations principles. J. Differ. Equ. 252, 716–744 (2012)
Rudin, W.: Functional Analysis, 2nd edn. McGraw-Hill, New York (1991)
Stroock, D.W., Varadhan, S.R.S.: Multidimensional Diffusion Processes. Springer, Berlin (1979)
Temam, R.: Navier–Stokes Equations: Theory and Numerical Analysis. North-Holland Publishing Company, Amsterdam (1979)
Triebel, H.: Theory of Function Spaces, Monographs in Mathematics, vol. 78. Birkhäuser, Basel (1983)
Vishik, M.J., Fursikov, A.V.: Mathematical Problems of Statistical Hydrodynamics. Kluwer, Dordrecht (1988)
Yamada, T., Watanabe, S.: Watanabe On the uniqueness of solutions of stochastic differential equations. J. Math. Kyoto Univ. 11(1), 155–167 (1971)
You, B.: The existence of a random attractor for the three dimensional damped Navier–Stokes equations with additive noise. Stoch. Anal. Appl. 35(4), 691–700 (2017). https://doi.org/10.1080/07362994.2017.1311794
Zhang, Z., Wu, X., Lu, M.: On the uniqueness of strong solution to the incompressible Navier–Stokes equations with damping. J. Math. Anal. Appl. 377, 414–419 (2011)
Zhou, Y.: Regularity and uniqueness for the 3D incompressible Navier–Stokes equations with damping. Appl. Math. Lett. 25, 1822–1825 (2012)
Acknowledgements
Open access funding provided by TU Wien (TUW). The authors would like to thank the referee for the constructive and helpful feedback to improve the presentation of the paper.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Communicated by F. Flandoli
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The research of Gaurav Dhariwal was supported by the Department of Mathematics, University of York and partially supported by the Austrian Science Fund (FWF) Grants P30000, W1245, and F65. The research of Zdzisław Brzeźniak has been partially supported by the Leverhulme project Grant No. RPG-2012-514.
Appendices
Appendix A: Convergence of \(P_n\)
Lemma A.1
Let \(\gamma > \frac{3}{2}\) and \(P_n :\mathrm {H}\rightarrow \mathrm {H}_n\) be the orthogonal projection as given by (4.1) (for more details see Sect. 4). Then as \(n \rightarrow \infty \)
-
(i)
\(P_n \psi \rightarrow \psi \) in \(\mathrm {H}\) for \(\psi \in \mathrm {H}\),
-
(ii)
\(P_n \psi \rightarrow \psi \) in \(\mathrm {V}\) for \(\psi \in \mathrm {V}\),
-
(iii)
\(P_n \psi \rightarrow \psi \) in \(\mathrm {V}_\gamma \) for \(\psi \in \mathrm {V}_\gamma \,.\)
Proof
Let \(\psi \in \mathrm {H}\), then by (4.1) and Parseval’s equality we have
Now since \(\psi \in \mathrm {H}\) using Lebesgue dominated convergence theorem it can be shown that
which infers (i).
Let \(\psi \in \mathrm {V}\), then by (4.1) and the definition of \(\mathrm {V}\)-norm we get
Again using the Lebesgue dominated convergence theorem for \(\psi \in \mathrm {V}\), we get
thus proving (ii).
Let \(\psi \in \mathrm {V}_\gamma \), then by (4.1) and the definition of \(\mathrm {V}_\gamma \)-norm we get
Similarly as before it can be shown that for \(\psi \in \mathrm {V}_\gamma \),
which concludes the proof. \(\square \)
Appendix B: Kuratowski Theorem
The main objective of this appendix is to prove the following theorem (see [9, Lemma 4.2]).
Theorem B.1
Let \(T > 0\) and
Then the sets \({\mathcal {C}}([0,T];\mathrm {V}) \cap {\mathcal {Z}}_T\), \({\mathcal {C}}([0,T]; \mathrm {H}_n) \cap {\mathcal {Z}}_T\) and \(L^2(0,T; \mathrm{D}(\mathrm {A})) \cap {\mathcal {Z}}_T\) are Borel subsets of \({\mathcal {Z}}_T\).
The proof of the above theorem heavily relies on the following Kuratowski theorem [18].
Theorem B.2
Assume that \(X_1, X_2\) are the Polish spaces with their Borel \(\sigma \)-fields denoted respectively by \({\mathcal {B}}(X_1), {\mathcal {B}}(X_2)\). If \(\varphi :X_1 \rightarrow X_2\) is an injective Borel measurable map then for any \(E_1 \in {\mathcal {B}}(X_1)\), \(E_2 := \varphi (E_1) \in {\mathcal {B}}(X_2)\).
We will also need following abstract results to prove Theorem B.1
Lemma B.1
Let \(X_1, X_2\) and Z be topological spaces such that \(X_1\) is a Borel subset of \(X_2\). Then \(X_1 \cap Z\) is a Borel subset of \(X_2 \cap Z\), where \(X_2 \cap Z\) is a topological space too, with the topology given by
Proof
Since the Borel \(\sigma \)-filed on \(X_2 \cap Z\) is the smallest \(\sigma \)-field generated by \(\tau (X_2 \cap Z)\), i.e. \({\mathcal {B}}(X_2 \cap Z) = \sigma ( \tau (X_2 \cap Z))\), in order to prove the lemma it is enough to show that \(\forall \, Y \in {\mathcal {B}}(X_1)\)
Firstly, we show that (B.2) holds for all \(Y \in \tau (X_1)\). Since \(X_1 \in {\mathcal {B}}(X_2)\), \(X_1 \subset X_2\) and has trace topology from \(X_2\), i.e \(\forall \, Y \in \tau (X_1)\) there exists a \(C \in \tau (X_2)\) such that
As \(X_1 \in {\mathcal {B}}(X_2)\) there exists a countable collection \(\left\{ K_i\right\} _{i \in {\mathbb {N}}}\) of open subsets of \(X_2\) such that
Therefore,
Since \(C \in \tau (X_2)\), for every \(i \in {\mathbb {N}}\), \(C \cap K_i\) is open in \(X_2\) and there exists a collection \(\left\{ B_j\right\} _{j\in {\mathbb {N}}} \in \tau (X_2)\) such that
Thus
and for every \(j \in {\mathbb {N}}\), \(B_j \cap Z \in {\mathcal {B}}(X_2 \cap Z)\). Since \({\mathcal {B}}(X_2 \cap Z)\) is a \(\sigma \)-field, the countable union also belongs to \({\mathcal {B}}(X_2 \cap Z)\), proving (B.2) for every \(Y \in \tau (X_1)\).
Secondly, we implement the method of good sets to prove (B.2) for a larger class of subsets of \(X_1\). Let
Claim\(:\)\({\mathcal {G}}\) is a \(\sigma \)-field.
-
(i)
\(X_1 \in {\mathcal {G}}\) since \(X_1 \subset X_1\) and \(X_1 \in \tau (X_1)\) by the definition of topology.
-
(ii)
Let \(A \in {\mathcal {G}}\). We want to show that \(A^c := X_1 \setminus A \in {\mathcal {G}}\), i.e. \(A^c \subset X_1\) and \(A^c \cap Z \in {\mathcal {B}}(X_2 \cap Z)\). Since \(A \in {\mathcal {G}}\), \(A \subset X_1\) and \(A \cap Z \in {\mathcal {B}}(X_2 \cap Z)\). Clearly \(A^c = X_1\setminus A \subset X_1\).
Since \(A \cap Z \in {\mathcal {B}}(X_2 \cap Z)\), then by the definition of \(\sigma \)-field
$$\begin{aligned} {}^c \left( A \cap Z\right) := \left( X_2 \cap Z\right) \setminus (A \cap Z) \in {\mathcal {B}}(X_2 \cap Z)\,. \end{aligned}$$We have the following set relations
$$\begin{aligned} {}^c(A \cap Z)&= {}^c A \cup {}^c Z = \left[ (X_2 \cap Z)\setminus A \right] \cup \left[ (X_2 \cap Z) \setminus Z \right] \\&= \left[ (X_2 \setminus A) \cap Z \right] \cup \emptyset = (X_2 \setminus A ) \cap Z \\&= \left[ A^c \cup \left( X_2 \setminus X_1 \right) \right] \cap Z \\&= \left( A^c \cap Z \right) \cup \left[ \left( X_2 \setminus X_1 \right) \cap Z\right] \\&= \left( A^c \cap Z \right) \cup {}^cX_1\;. \end{aligned}$$Now in the above identity \({}^c\left( A \cap Z \right) \), \({}^cX_1\) belongs to \({\mathcal {B}}(X_2 \cap Z)\) and hence \(A^c \cap Z \in {\mathcal {B}}(X_2 \cap Z)\), inferring \(A^c \in {\mathcal {G}}\).
-
(iii)
Let \(\{A_i\}_{i \in {\mathbb {N}}} \in {\mathcal {G}}\). Then \(A_i \subset X_1\) for every \(i \in {\mathbb {N}}\) hence
$$\begin{aligned} \bigcup _{i \in {\mathbb {N}}} A_i \subset X_1\,. \end{aligned}$$Also, the following holds
$$\begin{aligned} \left( \bigcup _{i \in {\mathbb {N}}} A_i \right) \cap Z = \bigcup _{i \in {\mathbb {N}}} (A_i \cap Z)\,. \end{aligned}$$Since \(A_i \in {\mathcal {G}}\), \(A_i \cap Z \in {\mathcal {B}}(X_2 \cap Z)\) and as \({\mathcal {B}}(X_2 \cap Z)\) is a \(\sigma \)-field
$$\begin{aligned} \bigcup _{i \in {\mathbb {N}}} \left( A_i \cap Z \right) \in {\mathcal {B}}(X_2 \cap Z)\,. \end{aligned}$$
From \((i) - (iii)\) we can infer that \({\mathcal {G}}\) is a \(\sigma \)-field. We have already shown that \(\tau (X_1) \subset {\mathcal {G}}\) thus
Therefore, we have shown that for every \(Y \in {\mathcal {B}}(X_1)\), \(Y \cap Z \in {\mathcal {B}}(X_2 \cap Z)\). \(\square \)
Lemma B.2
Let \(X_1, X_2, Y\) be topological spaces such that \(X_1 \subset X_2\), \(X_1\) has trace topology from \(X_2\) and \(X_1 \cap Y = X_2 \cap Y\) then
Proof
The topologies of \(X_1 \cap Y\) and \(X_2 \cap Y\) denoted by \(\tau (X_1 \cap Y)\) and \(\tau (X_2 \cap Y)\) respectively are given by
Since \(X_1\) has a trace topology from \(X_2\), for every \(A \in \tau (X_1)\) there exists a \(C \in \tau (X_2)\) such that \(A = C \cap X_1\). Thus
Thus all we are left to show is \(C \cap X_1 \cap B = C \cap B\) for every \(C \in \tau (X_2)\) and \(B \in \tau (Y).\) Since \(X_1 \cap Y = X_2 \cap Y\), we have the following set relations
\(\square \)
We will need the following space\(:\)
\(L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) is complete under the family of semi-norms
In particular, it is a Frechét space with the metric
Remark B.1
\(L^2(0,T; \mathrm {H}_{loc}) \subset L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) and we can define a topology on \(L^2(0,T; \mathrm {H}_{loc})\) by restricting the metric d to \(L^2(0,T; \mathrm {H}_{loc})\). Hence \(L^2(0,T; \mathrm {H}_{loc})\) is a topological space with the trace topology from \(L^2_{loc}([0,T] \times {\mathbb {R}}^3)\).
Let us define a new topological space\(:\)
Note that \(\widetilde{{\mathcal {Z}}}_T\) and \({\mathcal {Z}}_T\) are same as a set. Because \(L^2_{loc}([0,T] \times {\mathbb {R}}^3) \cap L^2_w(0,T; \mathrm{D}(\mathrm {A}))\) and \(L^2(0,T; \mathrm {H}_{loc}) \cap L^2_w(0,T; \mathrm{D}(\mathrm {A}))\) are same as a set. \(L^2(0,T; \mathrm {H}_{loc}) \subset L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) and the only extra elements in \(L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) are the ones which are locally square integrable but have non-zero divergence. But the intersection of \(L^2_{loc}([0,T]\times {\mathbb {R}}^3)\) with \(L^2_w(0,T; \mathrm{D}(\mathrm {A}))\) eliminates those elements as the divergence free condition is imposed by the second set.
By Remark B.1 and Lemma B.2\(\widetilde{{\mathcal {Z}}}_T\) and \({\mathcal {Z}}_T\) have the same topologies. Thus we will prove Theorem B.1 for \(\widetilde{{\mathcal {Z}}}_T\) instead of \({\mathcal {Z}}_T\).
Proof of Theorem B.1
First of all \({\mathcal {C}}([0,T]; \mathrm {V}) \subset {\mathcal {C}}([0,T]; \mathrm{U}^\prime ) \cap L^2_{loc}([0,T] \times {\mathbb {R}}^3)\). Secondly, \({\mathcal {C}}([0,T]; \mathrm {V})\) and \({\mathcal {C}}([0,T]; \mathrm{U}^\prime ) \cap L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) are Polish spaces. And finally, since \(\mathrm {V}\) is continuously embedded in \(\mathrm{U}^\prime \), the map
is continuous and hence Borel. Thus, by application of the Kuratowski theorem (see Theorem B.2), \({\mathcal {C}}([0,T]; \mathrm {V})\) is a Borel subset of \({\mathcal {C}}([0,T]; \mathrm{U}^\prime ) \cap L^2_{loc}([0,T] \times {\mathbb {R}}^3)\). Therefore, by Lemma B.1\({\mathcal {C}}([0,T]; \mathrm {V}) \cap \widetilde{{\mathcal {Z}}}_T\) is a Borel subset of \({\mathcal {C}}([0,T]; \mathrm{U}^\prime ) \cap L^2_{loc}([0,T] \times {\mathbb {R}}^3) \cap \widetilde{{\mathcal {Z}}}_T\) which is equal to \(\widetilde{{\mathcal {Z}}}_T\). We can show in the same way in the case of \({\mathcal {C}}([0,T];\mathrm {H}_n) \cap {\mathcal {Z}}_T\).
Similarly we can show that \(L^2(0,T; \mathrm{D}(\mathrm {A})) \cap \widetilde{{\mathcal {Z}}}_T\) is a Borel subset of \(\widetilde{{\mathcal {Z}}}_T\). \(L^2(0,T; \mathrm{D}(\mathrm {A})) \hookrightarrow L^2_{loc}([0,T] \times {\mathbb {R}}^3)\) and both are Polish spaces thus by application of the Kuratowski theorem [Theorem B.2], \(L^2(0,T; \mathrm{D}(\mathrm {A}))\) is a Borel subset of \(L^2_{loc}([0,T]\times {\mathbb {R}}^3)\). Finally, we can conclude the proof of the theorem by Lemma B.1.
\(\square \)
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Brzeźniak, Z., Dhariwal, G. Stochastic Tamed Navier–Stokes Equations on \({\mathbb {R}}^3\): The Existence and the Uniqueness of Solutions and the Existence of an Invariant Measure. J. Math. Fluid Mech. 22, 23 (2020). https://doi.org/10.1007/s00021-020-0480-z
Accepted:
Published:
DOI: https://doi.org/10.1007/s00021-020-0480-z