Abstract
We study the symmetric stochastic p-Stokes system, \(p \in (1,\infty )\), in a bounded domain. The results are two-fold: First, we show that in the context of analytically weak solutions, the stochastic pressure—related to non-divergence free stochastic forces—enjoys almost \(-1/2\) temporal derivatives on a Besov scale. Second, we verify that the velocity u of strong solutions obeys 1/2 temporal derivatives in an exponential Nikolskii space. Moreover, we prove that the non-linear symmetric gradient \(V(\mathbb {\epsilon } u) = (\kappa + \left| \mathbb {\epsilon } u\right| )^{(p-2)/2} \mathbb {\epsilon } u\), \(\kappa \ge 0\), which measures the ellipticity of the p-Stokes system, has 1/2 temporal derivatives in a Nikolskii space.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \({\mathcal {O}} \subset {\mathbb {R}}^n\) be a bounded domain, \(n \in {\mathbb {N}}\) and \(T > 0\) be finite. Given an initial datum \(u_0\) and a stochastic forcing \(G(\cdot ) \,\textrm{d}W\), we are seeking for a velocity field \(u: \Omega \times (0,T) \times {\mathcal {O}} \rightarrow {\mathbb {R}}^n\) and a pressure \(\pi : \Omega \times (0,T) \times {\mathcal {O}} \rightarrow {\mathbb {R}}\) that satisfy the relations
where \(S(\xi ):= \left( \kappa + \left| \xi \right| \right) ^{p-2} \xi \), \(\xi \in {\mathbb {R}}^{ n \times n}\), \(p \in (1,\infty )\), \(\kappa \ge 0\), \(\mathbb {\epsilon } u:= 2^{-1}\big ( \nabla u + (\nabla u)^T \big )\) and . The stochastic input \(G(\cdot ) \,\textrm{d}W\) is rendered via a cylindrical Wiener process W on an abstract separable Hilbert space U and a suitable noise coefficient G.
The system (1.1) is called stochastic symmetric p-Stokes system and describes the evolution of power-law fluids in the regime of laminar flow. Ultimately, one wants to substitute (1.1a) by
to include convective effects. This might lead to turbulent flow. Maybe most prominent is the case \(p=2\). Then the system reduces to the infamous stochastically forced Navier–Stokes equations. A derivation of this model and how meaningful choices of G might look like can be found in e.g. [50] (see also [12] and the references therein).
While the system (1.1) does not bother with the mathematical challenges of turbulence, it is still difficult to obtain improved regularity. There are several reasons:
-
(a)
Solutions to non-linear equations lack smoothness even for smooth data.
-
(b)
Limited regularity of the stochastic data \(G(\cdot ) \,\textrm{d}W\) leads to limited regularity of the solution.
-
(c)
Stochastic integration in general Banach spaces causes technical restrictions.
-
(d)
Delicate interplay between the divergence free constrain (1.1b) and the pressure.
In this article we focus on the temporal regularity for solutions to the system (1.1). Before we comment on the available literature on regularity and finally state the main contributions of this paper, we briefly discuss some results related to the existence of non-Newtonian fluids.
1.1 Existence
A mathematical theory for the existence of weak solutions to the deterministic counterpart of (1.1) enriched by the convective term was initiated by Ladyzhenskaya [39, 40] and Lions [43]. Motivated by the works, many authors contributed to the construction of solutions, e.g. [28, 46,47,48, 62]. Ultimately leading to the construction of a weak solution for \(p > 2n/(n+2)\). Below the compactness threshold \(W^{1,p}({\mathcal {O}}) \hookrightarrow \hookrightarrow L^2({\mathcal {O}})\) wild solutions may emerge, see e.g. [15].
The stochastic case is less known. An answer for solution independent noise coefficient was given in [57, 63]. The authors require the condition \(p > 9/5\) in the three dimensional setting. It was extended in [10] to cover more general noise coefficients and a unified condition \(p > (2n+2)/(n+2)\).
A construction for stochastic electro-rheological fluids, here p also depends on \(\Omega \times (0,T) \times {\mathcal {O}}\), is done in [13].
1.2 Regularity
Regularity is needed in order to control discretizations. In particular, regularity moderates the speed of convergence of numerical schemes, see e.g. [11]. Therefore, and out of mathematical curiosity, many authors have investigated the regularity of solutions to (1.1) and related models.
1.2.1 Deterministic
In general, there is a barrier that does not allow for arbitrary high regularity even though the data is smooth, cf. [37]. Still partial results can be obtained.
\(C^{1,\alpha }\)-regularity has been obtained in e.g. [2, 19, 20, 29]. Hölder regularity is sub-optimal from a numerical point of view. Instead, regularity of the expression \(V(\mathbb {\epsilon } u) = (\kappa + \left| \mathbb {\epsilon } u\right| )^{(p-2)/2} \mathbb {\epsilon } u\) is more suitable for the investigation of numerical algorithms as observed by Barrett and Liu in [4]. Indeed, the operator V measures the monotonicity of S in a natural way, i.e., for all \(A,B \in {\mathbb {R}}^{n\times n}\) it holds
A first result for non-singular shear thinning fluids, \(\kappa = 1\) and \(p \in (1,2)\), has been obtained in [51]. The author verifies local regularity \(V(\mathbb {\epsilon } u) \in W^{1,2}_{\textrm{loc}}({\mathcal {O}})\).
A similar result has been obtained for shear thickening fluids, \(p \ge 2\), in [9]. Additionally, local regularity for the non-linear tensor S and the pressure \(\pi \) was found in [38]. They show \(\nabla S(\mathbb {\epsilon } u), \nabla \pi \in L^{p'}_{\textrm{loc}}({\mathcal {O}})\).
In [21] regularity transfer of the data to the solution in terms of Campanato and weighted BMO (bounded mean oscillation) spaces is shown.
Spatial regularity results for the parabolic p-Stokes system has been obtained in [1]. The authors show that \(\nabla u \in L^\infty (0,T; L^2_{\textrm{loc}}({\mathcal {O}}))\) and \(\nabla V(\mathbb {\epsilon } u) \in L^2_{\textrm{loc}}((0,T)\times {\mathcal {O}})\). It was extended to the \(\phi \)-Stokes problem in [17].
As far as we know, improved temporal regularity for the parabolic symmetric p-Stokes system has not been analyzed in the literature. Only a few results that neglect the appearance of the pressure but keep the symmetric gradient are available. For example, the parabolic symmetric p-Laplace system has been studied by Frehse and Schwarzacher in [33] and by Burczak and Kaplický in [16]. In [33] it is shown that \(\partial _t u \in B^{1/2}_{2,\infty }(0,T; L^2({\mathcal {O}}))\). Here B denotes a Besov space (see Sect. 2.1 for more details).
1.2.2 Stochastic–Spatial Regularity
In general, the tools from deterministic theory are not available for stochastic equations. In particular, one cannot test the equation by a test function. Instead, one needs to perform an expansion for a suitable functional. This method is called Itô’s formula. While some test functions can be recovered in this way others cannot. Still, the deterministic regularity theory acts as a guiding example on what regularity can be expected.
It has been observed by Gess [35] that stochastic partial differential equations, arising as a stochastically perturbed gradient flow of an energy, can lead to improved regularity results. He proposes conditions that ensure the existence of a strong solution.
Fortunately, the Helmholtz-Leray projection \(\Pi _{\text {div}}\) (see (A. 6)) allows to reformulate (1.1a) and (1.1b) into
In other words, the Helmholtz-Leray projection factorizes the equations for the velocity field and the pressure. It enables to seek for the velocity independently of the pressure.
Now it is possible to interpret (A. 19a) as a stochastically perturbed gradient flow of the energy
induced by the potential \(\varphi _{\kappa }(t):= \int _0^t (\kappa + s)^{p-2} s \,\textrm{d}s\) on the space \(W^{1,p}_{0,\text {div}}({\mathcal {O}})\). Indeed, (A. 19a) can be written as
where \(D {\mathcal {J}}(u) = \Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u)\). Therefore, it is natural to look for regularity estimates that match the concept of strong solutions.
A similar approach is used by Breit and Gmeineder in [13]. They prove spatial regularity as in the deterministic case also for stochastically forced electro-rheological fluids and obtain \(\nabla u \in L^2(\Omega ; L^\infty (0,T; L^2({\mathcal {O}})))\) as well as \(\nabla V(\mathbb {\epsilon } u) \in L^2(\Omega \times (0,T) \times {\mathcal {O}})\). An analogous line of argumentation can be used to establish spatial regularity for the p-Stokes system.
1.2.3 Stochastic–Temporal Regularity
In sharp contrast to spatial regularity, where similar results as in the deterministic theory can be expected, is temporal regularity for stochastic partial differential equations. In general, temporal regularity is substantially worse compared to deterministic equations. This is due to the action of the random data \(G(\cdot ) \,\textrm{d}W\) on the solution u.
Even in the simplest case \(G\equiv 1\), we don’t expect u to exceed the regularity threshold induced by the Wiener process W. A precise regularity result was derived by Hytönen and Veraar in [36]. They show that \({\mathbb {P}}\)-a.s. \(W \in B^{1/2}_{\Phi _2,\infty }(0,T)\), where \(\Phi _2(t) = \exp (t^2)-1\), and \(W \not \in B^{1/2}_{p,q}(0,T)\) for any \(p \in [1,\infty ]\) and \(q \in [1,\infty )\). Therefore, a natural space for temporal regularity is the Nikolskii-space of functions with exponential second moments and 1/2 derivatives.
Only recently, precise regularity results for stochastic integrals in 2-smooth Banach spaces have been found by Ondreját and Veraar in [53]. Maybe most important, stochastic integration behaves as deterministic integration in the sense that improved regularity of the integrand leads to improved regularity of the stochastic integral itself. For example, if the integrand is bounded in time, then the stochastic integral is as regular as a Wiener process, i.e.,
Exactly this fact is used in [61] to derive temporal regularity for the stochastic p-Laplace system. In particular, we manage to recover the limiting temporal regularity \(u \in L^2(\Omega ; B^{1/2}_{\Phi _2,\infty }(0,T; L^2({\mathcal {O}})))\). In other words, the regularity of W fully transfers to regularity of u. Additionally, exploiting the V-coercivity, it is possible to derive non-linear gradient regularity \(V(\nabla u) \in L^2(\Omega ; B^{1/2}_{2,\infty }(0,T; L^2({\mathcal {O}})))\).
The stochastic p-Stokes system is even worse. A major obstacle in the derivation of higher regularity for the system (1.1) is the pressure. If \(G(\cdot ) \,\textrm{d}W\) is not divergence-free, then the pressure is very rough in time (see e.g. [18, 41]). This can be seen by formally decomposing the pressure into a deterministic and a stochastic component \(\pi = \pi _{\textrm{det}} + \pi _{\textrm{sto}}\), where
While the deterministic pressure can be treated by classical arguments, we observe that the stochastic pressure looks like white noise in time. This leads to severe difficulties in numerical discretizations as soon as we approximate the velocity field by discretely divergence-free fields rather than exactly divergence-free fields. More details can be found in [30, 31, 42] and the references therein.
1.3 Outline
In Sect. 2 we introduce the setup and present our main results.
In Sect. 3 we construct weak solutions and discuss regularity of pressure.
In Sect. 4 we prove the existence of strong solutions and verify temporal regularity of velocity and non-linear gradient.
More details on Besov norms, Helmholtz decomposition, Bogovskii operator, potentials, mapping properties of stochastic integrals, and gradient flows are presented in Appendix A.
2 Setup and Main Results
Let \({\mathcal {O}} \subset {\mathbb {R}}^n\), \(n \in {\mathbb {N}}\), be a bounded Lipschitz domain. For some given \(T>0\) we denote by \(I:= (0,T)\) the time interval and write \({\mathcal {O}}_T:= I \times {\mathcal {O}}\) for the time space cylinder. Moreover let \(\left( \Omega ,{\mathcal {F}}, ({\mathcal {F}}_t)_{t\in I}, {\mathbb {P}} \right) \) denote a stochastic basis, i.e., a probability space with a complete and right continuous filtration \(({\mathcal {F}}_t)_{t\in I}\). We write \(f \lesssim g\) for two non-negative quantities f and g if f is bounded by g up to a multiplicative constant. Accordingly we define \( > rsim \) and \(\eqsim \). We denote by c a generic constant which can change its value from line to line. Inner products and duality pairings are denoted by \(\left( \cdot , \cdot \right) \) and \(\langle \cdot , \cdot \rangle \), respectively.
2.1 Function Spaces
As usual, for \(1\le q < \infty \) we denote by \(L^q({\mathcal {O}})\) the Lebesgue space and \(W^{1,q}({\mathcal {O}})\) the Sobolev space. Moreover, \(W^{1,q}_0({\mathcal {O}})\) denotes the Sobolev spaces with zero boundary values. It is the closure of \(C^\infty _0({\mathcal {O}})\) (smooth functions with compact support) in the \(W^{1,q}({\mathcal {O}})\)-norm. We denote by \(W^{-1,q'}({\mathcal {O}})\) the dual of \(W^{1,q}_0({\mathcal {O}})\). The space of mean-value free Lebesgue functions is denoted by \(L^q_0({\mathcal {O}})\). The space of smooth, compactly supported and divergence-free vector fields is called \(C^\infty _{0,\text {div}}({\mathcal {O}})\) and its closure within the \(W^{1,p}\)-norm is abbreviated by \(W^{1,p}_{0,\text {div}}({\mathcal {O}})\). We do not distinguish in the notation between scalar-, vector- and matrix-valued functions.
For a Banach space \(\left( X, \left\| \cdot \right\| _X \right) \) let \(L^q(I;X)\) be the Bochner space of Bochner-measurable functions \(u: I \rightarrow X\) satisfying \(t \mapsto \left\| u(t)\right\| _X \in L^q(I)\). Moreover, \(C^0({\overline{I}};X)\) is the space of continuous functions with respect to the norm-topology. We also use \(C^{0,\alpha }({\overline{I}};X)\) for the space of \(\alpha \)-Hölder continuous functions. Given an Orlicz-function \(\Phi : [0,\infty ] \rightarrow [0,\infty ]\), i.e. a convex function satisfying \( \lim _{t \rightarrow 0} \Phi (t)/t = 0\) and \(\lim _{t \rightarrow \infty } \Phi (t)/t = \infty \) we define the Luxemburg-norm
The Orlicz space \(L^\Phi (I;X)\) is the space of all Bochner-measurable functions with finite Luxemburg-norm. For more details on Orlicz-spaces we refer to [25]. Given \(h \in I\) and \(u:I \rightarrow X\) we define the difference operator \(\tau _h: {\{{u: I \rightarrow X}\}} \rightarrow {\{{u: I\cap I - {\{{h}\}} \rightarrow X}\}} \) via \(\tau _h(u) (s):= u(s+h) - u(s)\). The Besov-Orlicz space \(B^\alpha _{\Phi ,r}(I;X)\) with differentiability \(\alpha \in (0,1)\), integrability \(\Phi \) and fine index \(r \in [1,\infty ]\) is defined as the space of Bochner-measurable functions with finite Besov-Orlicz norm \(\left\| \cdot \right\| _{B^\alpha _{\Phi ,r}(I;X)}\), where
The case \(r = \infty \) is commonly called Nikolskii-Orlicz space and abbreviated by \(N^{\alpha ,\Phi } = B^{\alpha }_{\Phi ,\infty }\). When \(\Phi (t) = t^p\) we write \(B^\alpha _{p,r}(I;X)\) and call it Besov space. If X is reflexive, we denote by \(B^{-\alpha }_{p',q'}(I;X') = \big ( B^{\alpha }_{p,q}(I;X) \big )'\).
Similarly, given a Banach space \(\left( Y, \left\| \cdot \right\| _Y \right) \), we define \(L^q(\Omega ;Y)\) as the Bochner space of Bochner-measurable functions \(u: \Omega \rightarrow Y\) satisfying \(\omega \mapsto \left\| u(\omega )\right\| _Y \in L^q(\Omega )\). The space \(L^q_{{\mathcal {F}}}(\Omega \times I;X)\) denotes the subspace of X-valued progressively measurable processes. We abbreviate the notation \(L^q_\omega L^q_t L^q_x:= L^q(\Omega ;L^q(I;L^q({\mathcal {O}}))) \) and \(L^{q-}:= \bigcap _{r< q} L^r\).
2.2 Structural Assumptions on the Stochastic Data
Let U be some separable Hilbert space and \({\{{u_j}\}}_{j\in {\mathbb {N}}}\) be a complete orthonormal system of U. We render the stochastic input via a cylindrical noise on the abstract space U.
Assumption 1
(Cylindrical Wiener process). We assume that W is an U-valued cylindrical Wiener process with respect to \(({\mathcal {F}}_t)\). Formally W can be represented as
where \({\{{\beta ^j}\}}_{j\in {\mathbb {N}}}\) are independent 1-dimensional standard Brownian motions.
We construct the noise coefficient G as a two parameter map. In its first parameter it acts as a Nemytskii operator whereas the dependence on the Wiener process is linear.
Definition 2
(Noise coefficient). Let \({\{{g_j}\}}_{j\in {\mathbb {N}}}: {\mathcal {O}} \times {\mathbb {R}}^N \rightarrow {\mathbb {R}}^N\). We define G by
G is uniquely determined by the sequence of functions \({\{{g_j}\}}\). In particular, regularity assumptions and summation properties need to be imposed on \({\{{g_j}\}}\) to obtain an operator G that is stable on specific function spaces.
The construction of the stochastic integral in the framework of Hilbert spaces can be done by Itôs isometry and requires the integrand to be an Hilbert-Schmidt operator, i.e., if \(G \in L^2_{{\mathcal {F}}}\big (\Omega \times I; L_2(U; L^2_x) \big )\), then
converges in \(L^2\big (\Omega ; C(I; L^2_x) \big )\). Moreover, we have the equivalence
2.3 Main Result I: Temporal Regularity of Pressure for Weak Solutions
We apply the general theory of monotone stochastic partial differential equations developed by Liu and Röckner [44], see also the book [45], to the p-Stokes system. More recently, the theory has been refined to cover more general equations, see e.g. [3, 55]. In this way, existence of weak solutions is almost immediate. The main novelty is a careful consideration of the regularity of the stochastic pressure.
Definition 3
Let \(u_0 \in L^2_\omega L^2_{\text {div}}\) be \({\mathcal {F}}_0\)-measurable. The tupel \((u,\pi )\) is called weak solution to (1.1) if
-
(a)
\(u \in L^2_\omega C_t^0 L^2_{\text {div}} \cap L^p_\omega L^p_t W^{1,p}_{0,x}\) is \(({\mathcal {F}}_t)\)-adapted,
-
(b)
for all \(t \in I\), \(\xi \in C^\infty _{0,\text {div}}\) and \({\mathbb {P}}\)-a.s. it holds
$$\begin{aligned} \int _{{\mathcal {O}}} (u(t) - u_0) \cdot \xi \,\textrm{d}x+ \int _0^t \int _{{\mathcal {O}}} S( \mathbb {\epsilon } u) :\nabla \xi \,\textrm{d}x\,\textrm{d}s= \sum _{j\in {\mathbb {N}}} \int _0^t \int _{{\mathcal {O}}} g_j(\cdot ,u)\cdot \xi \,\textrm{d}x\,\textrm{d}\beta ^j(s), \end{aligned}$$(2.6) -
(c)
\(\pi = \pi _{\textrm{det}} + \pi _{\textrm{sto}}\) with \(\pi _{\textrm{det}} \in L^{p'}_\omega L^{p'}_t L^{p'}_{0,x}\), \(\pi _{\textrm{sto}} \in L^{2}_\omega B^{-1/2}_{\Phi _2,\infty } \big ( W^{1,2}_x \cap L^2_{0,x} \big )\)
-
(d)
and for all \(\xi _{t,x} \in C^\infty _0({\mathcal {O}}_T)\) and \({\mathbb {P}}\)-a.s.
$$\begin{aligned} \langle \pi , \text {div}\xi _{t,x} \rangle = \int _{{\mathcal {O}}_T} S(\mathbb {\epsilon } u): \nabla \Pi _{\text {div}}^\perp \xi _{t,x} \,\textrm{d}x \,\textrm{d}t +\int _{{\mathcal {O}}_T} {\mathcal {I}}_W\big (G(u)\big ) \partial _t \Pi _{\text {div}}^\perp \xi _{t,x} \,\textrm{d}x \,\textrm{d}t. \end{aligned}$$(2.7)
Here \(\Pi _{\text {div}}^\perp \) is the complement of the Helmholtz–Leray projection and is defined in A. 6.
The general theory of monotone stochastic partial differential equations requires further assumptions on the noise coefficient.
Assumption 4
We assume that \({\{{g_j}\}}_{j\in {\mathbb {N}}} \in C^0({\mathcal {O}} \times {\mathbb {R}}^n; {\mathbb {R}}^n)\) with
-
(a)
(sublinear growth) for all \(v \in L^2_{\text {div}}\)
$$\begin{aligned} \left\| G(v)\right\| _{L_2(U;L^2_x)}^2 = \sum _{j\in {\mathbb {N}}} \left\| g_j\big (\cdot ,v(\cdot )\big )\right\| _{L^2_x}^2 \le c_{\text {growth}}(1+\left\| v\right\| _{L^2_x}^2), \end{aligned}$$(2.8) -
(b)
(Lipschitz continuity) for all \(v_1, \, v_2 \in L^2_{\text {div}}\) it holds
$$\begin{aligned} \left\| G(v_1) - G(v_2)\right\| _{L_2(U;L^2_x)}^2 \le c_{\text {lip}} \left\| u_1 - u_2\right\| _{L^2_x}^2. \end{aligned}$$(2.9)
We are ready to formulate our first main result.
Theorem 5
Let \(p \in (1,\infty )\) and \({\mathcal {O}}\) be a bounded \(C^2\)-domain. Additionally, let \(q \in [1,\infty )\), \(u_0 \in L^{q}_\omega L^2_{\text {div}}\) be \({\mathcal {F}}_0\)-measurable and Assumption 4 be satisfied. Then there exists a unique weak solution \((u,\pi )\) to (1.1) such that
where \({\mathcal {J}}\) is given by (1.4).
Moreover, it holds \(\pi = \pi _{\textrm{det}} + \pi _{\textrm{sto}} \) with
for any \(s, r \in (1,\infty )\) and \(\varepsilon >0\).
2.4 Main result II: Temporal Regularity of Velocity Field for Strong Solutions
We exploit the gradient flow structure (1.5) on the space of solenoidal vector fields in order to use the existence theory of strong solutions initiated by Gess [35]. In doing so, we provide a sufficient condition for the existence of strong solutions to the p-Stokes system.
Definition 6
Let \(u_0 \in L^2_\omega L^2_{\text {div}}\) be \({\mathcal {F}}_0\)-measurable. The tupel \((u,\pi )\) is called strong solution if it is a weak solution and additionally satisfies
-
(a)
\(\Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u) \in L^2_\omega L^2_t L^2_{\text {div}}\),
-
(b)
for all \(t \in I\) and \({\mathbb {P}}-a.s.\) it holds
$$\begin{aligned} u(t) - u_0 - \int _0^t \Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u) \,\textrm{d}s= \sum _{j\in {\mathbb {N}}} \int _0^t \Pi _{\text {div}} g_j(\cdot ,u) \,\textrm{d}\beta ^j(s) \end{aligned}$$(2.11)as an equation in \(L^2_{\text {div}}\).
In general, strong solutions only upgrade the regularity of the divergence-free component of the equation. In particular, the pressure does not enjoy higher regularity for strong solutions compared to weak solutions. Additional assumptions need to be imposed on the noise coefficient to ensure existence of strong solutions.
Assumption 7
We assume that \({\{{g_j}\}}_{j\in {\mathbb {N}}} \in C^1({\mathcal {O}} \times {\mathbb {R}}^n; {\mathbb {R}}^n)\) such that for all \(v \in W^{1,p}_{0,x} \cap L^2_{\text {div}}\) it holds
-
(a)
(stability)
$$\begin{aligned} \left\| \mathbb {\epsilon } \Pi _{\text {div}} G(v) \right\| _{L_2(U;L^p_x)} \lesssim 1+\left\| \mathbb {\epsilon } v\right\| _{L^p_x}, \end{aligned}$$(2.12) -
(b)
(boundary condition) and \(\Pi _{\text {div}} G(v) \in W^{1,p}_{0,x}\).
Clearly, \(G(v) = v\) satisfies Assumption 7. Generally, the stability can be verified using the Sobolev stability of \(\Pi _{\text {div}}\), cf. Theorem 32, and a suitable assumption on the data \({\{{g_j}\}}\), see Example 8. But it is non-trivial to verify the boundary condition; the Helmoltz–Leray projection is a non-local operator. So even if G(v) vanishes on the boundary, it is unclear if \(\Pi _{\text {div}} G(v)\) also vanishes.
Example 8
Let \(h \in C^\infty _{0,x}\). Define \(g_1(x,v):= \sin (x\cdot v) h(x)\) and \(g_j = 0\) for \(j \ge 2\). Set \(G(v)(x):= \sum _{j\in {\mathbb {N}}} g_j(x,v(x)) u_j\) for \(v \in W^{1,p}_{0,x} \cap L^2_{\text {div}}\), \(\{u_j\}_{j \in {\mathbb {N}}}\) an ONB of U.
Notice that the i-th derivative of the k-th component is given by
This implies \(G(v) = 0\) and \(\nabla G(v) = 0\) on \(\partial {\mathcal {O}}\). Now, using the Sobolev stability of \(\Pi _{\text {div}}\), Theorem 32,
Moreover, due to Poincaré’s inequality, boundedness of \({\mathcal {O}}\) and Korn’s inequality,
This allows us to establish (2.12).
Theorem 9
Let \(p \ge 2\), \(q \in [1,\infty )\) and \( {\mathcal {J}}(u_0) \in L^{q}_\omega \) be \({\mathcal {F}}_0\)-measurable. Moreover, let G satisfy Assumptions 4 and 7. Then there exists a unique strong solution \((u,\pi )\) to (1.1). Moreover, it holds
We extend the techniques developed in [61] to obtain improved temporal regularity for the velocity field and its non-linear symmetric gradient.
Theorem 10
Let \(p \in (1,\infty )\), Assumption 4 be satisfied and \((u,\pi )\) be a strong solution to (1.1).
-
(a)
(Polynomial integrability in time) Let \(q \in [1,\infty )\). Then \(u \in L^{2}_\omega B^{1/2}_{q,\infty } L^2_x\) and
$$\begin{aligned} \left\| u\right\| _{L^{2}_\omega B^{1/2}_{q,\infty } L^2_x} \lesssim \left\| u\right\| _{L^{2}_\omega L^\infty _t L^2_x} + \left\| \Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u)\right\| _{L^2_\omega L^2_t L^2_x} + 1. \end{aligned}$$(2.14) -
(b)
(Exponential integrability in time) Let \(N_2(t):= t^2 \ln (t+1)\), \(\Phi _2(t):= e^{t^2} - 1\) and \(u \in L^{N_2}_\omega L^\infty _t L^2_x\). Then \(u \in L^{2}_\omega B^{1/2}_{\Phi _2,\infty } L^2_x\) and
$$\begin{aligned} \left\| u\right\| _{L^{2}_\omega B^{1/2}_{\Phi _2,\infty } L^2_x} \lesssim \left\| u\right\| _{L^{N_2}_\omega L^\infty _t L^2_x} + \left\| \Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u)\right\| _{L^2_\omega L^2_t L^2_x} + 1. \end{aligned}$$(2.15)
Remark 11
Theorem 10 works for the full range \(p \in (1,\infty )\). But the existence of strong solutions is still open for \(p \in (1,2)\). Existence of strong solutions for deterministic power-law fluids including convection has been established in [6].
Exponential integrability requires slightly stronger moment estimates. In a similar way to the a priori bound (2.10a) for weak solutions, one can establish an estimate of the form
In other words, if the initial condition has \(N_2\) moments, so does the solution at later times and the assumption of part (b) is satisfied.
Theorem 12
Let \(p \ge 2\), Assumption 7 be satisfied and \((u,\pi )\) be a strong solution to (1.1). Additionally, assume \({\mathcal {J}}(u) \in L^1_\omega L^\infty _t\). Then \(V(\mathbb {\epsilon } u) \in L^2_\omega B^{1/2}_{2,\infty } L^2_x\) and
3 Temporal Regularity of Stochastic Pressure for Weak Solutions
The aim of this section is the discuss weak solutions to (1.1). In particular, we split the pressure into two terms; one is related to the diffusion operator \(S(\mathbb {\epsilon } u)\), while the other corresponds to stochastic data.
3.1 Existence of Weak Solutions
The general theory of monotone SPDEs of Liu and Röckner [44] covers the construction of the velocity variable. A similar construction for weak solutions to power-law fluids has been done in [10]. For the sake of completeness, we state the result and comment on the main ingredients.
Theorem 13
Let \(p \in (1,\infty )\), \(q \in [1,\infty )\), \({\mathcal {O}}\) be open, bounded, Assumption 4 be satisfied and \(u_0 \in L^q_\omega L^2_{\text {div}}\) be \({\mathcal {F}}_0\)-measurable.
Then there exists \(u \in L^q_\omega C^0_t L^2_{\text {div}} \cap L^{p q/2}_\omega L^{p q/2}_t W^{1,p}_{0,x} \) adapted to \(({\mathcal {F}}_t)\) such that u satisfies (2.6) and
The abstract conditions postulated by Liu and Röckner, that guarantee the existence of a weak solution, can be verified along the lines of [45, Example 4.1.9]. In fact, [45, Example 4.1.9] only deals with the case \(p \ge 2\). However, the general case \(p \in (1,\infty )\) can be recovered, if we adjust the underlying Gelfand triple to
The key tools for the a priori bound (3.1) are Itô’s formula for \(F(u) = \left\| u\right\| _{L^2_x}^q\) together with a Gronwall argument. While the case \(q=2\) is standard, the general case needs some minor changes. Since this is not the main focus of the article, we only refer to [60, Theorem 4.3.1] for more details.
Remark 14
The integrability in probability of the solution u is purely determined by the integrability in probability of the initial condition \(u_0\), i.e., for all \(q \in [1,\infty )\)
The limit case \(q = \infty \) needs to be excluded since the moment transfer is hindered by the Wiener process W. The most one can hope for is \(u \in L^{\Phi _2}_\omega L^2_x\) since \(W \in L_\omega ^{\Phi _2} \backslash L^\infty _\omega \). More details about integrability of Brownian motions in Banach spaces can be found in [36].
3.2 Reconstruction of the Pressure
Thanks to the Helmholtz-Leray projection \(\Pi _{\text {div}}\) it is possible to target the construction of the velocity u and the pressure \(\pi \) individually. The pressure corresponds to the residual error when testing (2.6) by smooth gradients rather than smooth divergence free fields, i.e., we use (2.7) as a definition of the pressure \(\pi \). Additionally, we artificially decompose \(\pi = \pi _{\textrm{det}} + \pi _{\textrm{sto}} \) into the components
and discuss them separately.
First, we have a look at the deterministic component.
Lemma 15
Let \({\mathcal {O}}\) be a bounded \(C^2\) domain and \( u \in W^{1,p}_{0,x}\). Then there exists \(\pi _{\textrm{det}} \in L^{p'}_{0,x}\) such that for all \(\xi \in C^\infty _{0,x}\)
and
Proof
We interpret (3.2a) distributionally, i.e., for \(\xi \in C^\infty _{0,x}\)
Without loss of generality we assume \(\langle \pi _{\textrm{det}}\rangle _{{\mathcal {O}}} = 0\).
Since we want to identify \(\pi _{\textrm{det}}\) as a proper function, we substitute \(\xi = {\mathcal {B}} \xi _0\), where \({\mathcal {B}}\) denotes the Bogovskii operator (see Theorem 34) and \(\xi _0 \in C^\infty _{0,x}\) with \(\langle \xi _0\rangle _{{\mathcal {O}}} = 0\), to obtain
Thus, using the symmetry of \(\Pi _{\text {div}}^\perp \) and integration by parts, (3.6) is equivalent to
where \({\mathcal {B}}^*\) is the adjoint operator of \({\mathcal {B}}\).
It remains to verify the boundedness of
We instead address the boundedness of the adjoint operator and use a duality argument.
The right hand side of (3.6) is estimated by Hölder’s inequality
We further estimate, due to the Sobolev stability of the Helmholtz-Leray projection (Theorem 32) and the Bogovskii operator (Theorem 34 with \(s = 0\), \(q = p\)),
Therefore, the linear operator \(A: L^p_{0,x} \rightarrow L^p_{0,x}\) is bounded.
The claim (3.3) follows by our construction. The inequality (3.4) is verified by a density argument
\(\square \)
Second, we investigate the stochastic pressure. The main difficulty is the limited time regularity. Therefore, we initially take a look at the time integrated pressure.
Lemma 16
Let \({\mathcal {O}}\) be a bounded domain with locally Lipschitz boundary and \({\mathcal {I}}_W\big (G(u) \big ) \in L^2_\omega B^{1/2}_{\Phi _2,\infty } L^2_x\). Then there exists \({\mathcal {K}}_{\textrm{sto}} \in L^2_\omega B^{1/2}_{\Phi _2,\infty } \big ( W^{1,2}_{0,x} \cap L^2_{0,x}\big ) \) such that for all \(t \in I\) and \(\xi \in C^\infty _{0,x}\) it holds
Moreover,
Proof
The proof proceeds similar to the one of Lemma 15.
First, note (3.8) is equivalent to
where \({\mathcal {B}}^*\) is the adjoint of the Bogovskii operator. In particular, \({\mathcal {K}}_{\textrm{sto}}\) is mean-value free.
Due to Theorem 34, one can check that \( {\mathcal {B}}^*: L^2_{0,x} \rightarrow W^{1,2}_{0,x} \cap L^2_{0,x}\) is bounded. Indeed, let \(\xi , \zeta \in C^\infty _{0,x}\) such that \(\langle \xi \rangle _{{\mathcal {O}}} = \langle \zeta \rangle _{{\mathcal {O}}} = 0\), then by Hölder’s inequality and Theorem 34 with \(s = -1\) and \(q = 2\),
Thus, using the density of smooth mean-value free functions within \(W^{1,2}_{0,x} \cap L^2_{0,x}\),
Finally, (3.11) together with \(L^2_x\)-stability of the Helmholtz-Leray projection establish
An application of the \(L^2_\omega B^{1/2}_{\Phi _2,\infty }\)-norm verifies (3.9). \(\square \)
Now, we have identified the regularity class for the integrated pressure. The next step is the transfer of the results to the pressure. For a smooth test function \(\xi \in C^\infty _0\big ( [0,T); C^\infty _{0,x} \big )\) we define the distribution
To shorten the notation we abbreviate \(X = W^{1,2}_{0,x} \cap L^2_{0,x}\).
Lemma 17
Let \(\alpha \in (0,1)\) and \(p,q \in (1,\infty )\). Moreover, assume that \({\mathcal {K}}_{\textrm{sto}} \in L^2_\omega B^{\alpha }_{p,q} X\). Then \(\pi _{\textrm{sto}} \in L^2_\omega B^{\alpha -1}_{p,q} X\) and
Proof
Let \(\xi \in C^\infty _0\big ( I; C^\infty _{0,x} \big )\) be spatially mean-value free. Due to (3.12) and duality
Note that \(\xi \) trivially extends by zero to the full space in a smooth way. Now, we can use the equivalent spectral characterization of Besov norms, cf. Corollary 27,
Since X is reflexive and \(p, q < \infty \) we can identify the dual using Lemma 28
An application of Lemma 29 yields
Going back to the Besov norm in terms of integrability of differences, cf. Theorem 26, and using that \(\xi \) is extended by zero
This concludes
Overall, revisiting (3.14) and using the density of smooth mean-value free functions in X together with (3.16),
Ultimately, due to the equivalent norm on the dual space, cf. Lemma 28,
Taking the square and expectation verifies (3.13). \(\square \)
Remark 18
So far the stochastic pressure was always estimated on suboptimal spaces. For example in [18, 41] the authors infer \(\pi _{\textrm{sto}} \in L^1_\omega W^{-1,\infty }_t L^2_{0,x}\). They neglect all additional information on the temporal differentiability of \({\mathcal {K}}_{\textrm{sto}}\) and only use the boundedness.
Whether a corresponding result of Lemma 17 remains valid on the scale of exponentially integrable Besov spaces boils done to the understanding of the interaction between derivatives and norms in the spirit of (3.15), i.e.,
In order to establish (3.15), we heavily relied on the spectral representation of Besov norms. However, the equivalence of the spectral representation and the definition in terms of integrated differences fails for \(p \in \{ 1, \infty \}\). Particularly, \(\Phi _2(t) = e^{t^2}-1\) is to close to \(\infty \) and whence \(\Phi _2^*\) is to close to 1.
Still, in regard of the regularity result of the integrated pressure, cf. Lemma 16, it is natural to define the set
Corollary 19
In the setting of Lemma 16 we have \(\pi _{\textrm{sto}} \in {\mathbb {B}}^{-1/2}_{\Phi _2,\infty }(0,T; X)\) \({\mathbb {P}}\)-a.s.
Proof
We only need to verify that the distribution \(\pi _{\textrm{sto}}\) defined by (3.12) is well-defined. The claim that \(\pi _{\textrm{sto}}\) belongs to \( {\mathbb {B}}^{-1/2}_{\Phi _2,\infty }(0,T; X)\) follows by the definition (3.18) and the assumption on \({\mathcal {K}}_{\textrm{sto}}\).
Let \(\xi \in C^\infty _0\big ((0,T); X' \big )\). Due to duality and Hölder’s inequality
Next, we take the square and expectation
Thus, we have shown that \(\pi _{\textrm{sto}}\) is well-defined. \(\square \)
3.3 Proof of Theorem 5
The previous results lead to a simple proof of Theorem 5.
Proof of Theorem 5
Due to the factorization (1.3) it is possible to construct the velocity field u independently of the pressure \(\pi \). Afterwards the pressure is reconstructed.
The existence of the velocity field and verification of the a priori bound (2.10a) is done in Theorem 13.
Section 3.2 deals with the pressure reconstruction; Lemma 15 verifies the regularity of the deterministic pressure; the stochastic pressure is handled in Lemma 16 and 17. In particular, Lemma 16 ensures that \({\mathcal {K}}_{\textrm{sto}} \in L^2_\omega B^{1/2}_{\Phi _2,\infty } X \hookrightarrow L^2_\omega B^{\alpha }_{s,r} X\) for any \(\alpha \in (0,1/2)\) and \(s,r \in (1,\infty )\). Therefore Lemma 17 implies (2.10c). \(\square \)
4 Temporal Regularity of Strong Solutions
Within this section we discuss the existence of strong solutions to (1.1). The main difference between weak and strong solutions lies in the fact that the latter can be interpreted in a point-wise manner, i.e., there is no need to interpret (1.1a) in a distributional sense.
4.1 Existence of Strong Solutions
Strong solutions for a related model have already been constructed in [13] by means of improved spatial regularity. They formally test (1.1) by the Laplacian. We on the other hand rely on the gradient flow structure (A. 19a), which – at least formally – corresponds to a test with \(-\Pi _{\text {div}}\text {div}S(\mathbb {\epsilon } u)\).
An application of Itô’s formula to \({\mathcal {J}}(u)^{q}\), \(q\in [1,\infty )\), provides
where \(D{\mathcal {J}}(u)\) and \(D^2{\mathcal {J}}(u)\) are given by (A. 23) and (A. 24), respectively. Most importantly, the deterministic integral in the first term has a sign, i.e.
The Itô corrections need to be interpreted as
and, recalling \(\varphi _{\kappa }(t) = \int _0^t (\kappa + s)^{p-2} s \,\textrm{d}s\),
A priori bounds for the energy can be established whenever the correction terms can be bounded with respect to the energy \({\mathcal {J}}(u)\).
The above computation have been made rigorous in the abstract theory developed by Gess in [35].
Proof of Theorem 9
We will only address the case \(q = 1\). More details on the general case can be found in [60, Theorem 4.3.1]. Our aim is to apply the result [35, Theorem 1.4], i.e., we need to check the assumptions (A1\()-(\)A6).
First of all, we relate our framework to the one used in [35]. We choose \(H = L^2_{\text {div}}\), \(S = W^{1,p}_{0,x}\) and \(V = H \cap S\). Additionally, we let \(\varphi (v) = {\tilde{\varphi }}(v) = {\mathcal {J}}(v)\) and \(B_t(v) \equiv \Pi _{\text {div}} G(v)\). Keep in mind that we use \(\varphi \) as the potential that defines the energy whereas Gess uses \(\varphi \) to denote the energy itself.
Ad (A1): First we show that \({\mathcal {J}}: W^{1,p}_{0,x} \rightarrow {\mathbb {R}}\) is continuous. Indeed, let \(u,v \in W^{1,p}_{0,x}\). Then, due to the fundamental theorem,
where \(w_\theta = v + \theta (u-v)\). Note that \(\left| \mathbb {\epsilon } w_\theta \right| \le \max \{ \left| \mathbb {\epsilon } u\right| , \left| \mathbb {\epsilon } v\right| \}\) for all \(\theta \in [0,1]\). This and Hölder’s inequality imply
Finally, we arrive at
The continuity is verified.
Moreover, using that \(\varphi '(t) = t^{p-1}\) is homogeneous of degree \(p-1\) and a substitution,
Since \(p \ge 2\) we find \(\varphi _{\kappa }(t)\) is increasing in \(\kappa \) for fixed t and whence \({\mathcal {J}}(2u) \le 2^p {\mathcal {J}}(u)\).
Ad (A2): Recall
Applying a weighted Young’s inequality one finds
In other words \(\varphi _{\kappa }\) behaves like the pth-power shifted along \(\kappa \). It follows using the monotonicity of powers and (4.1)
Let \({\{{w_k}\}}_{k \in {\mathbb {N}}} \subset W^{1,p}_{0,x}\). Now, using (A. 24), Hölder’s and Young’s inequalities and (4.2)
Ad (A3): Lemma 42 shows that \({\mathcal {J}}\) is strongly convex.
Ad (A4): Recall (A. 23). Therefore,
Ad (A5): Follows by the assumption on the noise coefficient, cf. (2.9).
Ad (A6): The major ingredient that enables strong solutions is the regularity of the noise. Recall (2.3) and (A2) with \(w_k = \Pi _{\text {div}}G(v)u_k = \Pi _{\text {div}} g_k(\cdot , v )\). It remains to estimate
An application of (2.12) and (4.2) show
Overall, we have verified (A1\()-(\)A6) and therefore we may apply [35, Theorem 1.4] which provides us with the a priori bound
A careful consideration of the proof shows that one can also obtain an estimate where the supremum in time is taken before the expectation. One additional term, due to the stochastic integral, needs to be considered, i.e.,
Due to the Burkholder-Davis-Gundy inequality
Invoking Hölder’s inequality
The Assumption 7 together with (4.2) show
Finally, Young’s inequality verifies
This allows for a Gronwall type argument that establishes the improved a priori estimate
\(\square \)
Remark 20
On a first view we could ask, if one could expand the q-energy (applying Itô’s formula to \(\left\| \epsilon u\right\| _{L^q_x}^q\)) of the symmetric p-Stokes system. However, additional difficulties arise to to the presence of the symmetric gradient. In particular, one needs to understand the interaction between projected symmetric p- and q-Laplacian. If one could establish a sign
then it is possible to obtain improved gradient regularity.
Regularity questions about \(\text {div}S(\mathbb {\epsilon } u)\) – and ultimately on \(\nabla S(\mathbb {\epsilon } u)\) – are out of reach, since
is only a distribution. The derivation of higher regularity for the pressure is a non-trivial task.
Fortunately, the strong formulation of equation (1.1) is sufficient to derive improved temporal regularity. We already established a similar result for the p-Laplace system in [61].
4.2 Exponential Besov-regularity
In general, it is delicate to derive temporal regularity on the limited threshold of order 1/2 for stochastic partial differential equations. A key example was derived by Hytönen and Veraar in [36]. They show that \({\mathbb {P}}\)-almost surely Wiener processes belong to the Besov space \(B^{1/2}_{\Phi _2,\infty }\), \(\Phi _2(t) = e^{t^2}-1\), and do not belong to the Besov space \(B^{1/2}_{p,q}\) for \(p,q < \infty \).
We show that strong solutions inherit similar temporal regularity properties from the driving Wiener process.
Proof of Theorem 10
We will only present part (b) of Theorem 10 in detail and comment on the necessary changes for part (a).
We start with the estimate for the Besov-Orlicz semi-norm
Using the strong formulation (2.11)
Invoking \(L^{\infty } \hookrightarrow L^{\Phi _2}\) and Hölder’s inequality
The second term is bounded by Theorem 35 (e), (A. 14) and (2.8)
where \(N_2(t) = t^2 \ln (t+1)\). This establishes the semi-norm estimate.
The full norm follows by the embedding \(L^\infty \hookrightarrow L^{\Phi _2}\)
We have verified (2.15).
Part (a) follows similarly to part (b), but we use the embedding \(L^\infty _t \hookrightarrow L^q_t\) and Theorem 35 (b) together with an extrapolation result, cf. [53, Remark 3.4]. \(\square \)
Remark 21
Theorem 10 immediately implies Hölder regularity up to almost 1/2 by an application of a Sobolev embedding (cf. [58, Sect. 2.3.2]), i.e., for any \(\alpha \in (0,1/2)\)
4.3 Nikolskii-regularity of Non-linear Gradient
Before we can turn our attention to investigate the regularity properties of the non-linear gradient, we need to understand how the stochastic integral operator interacts with linear projections and gradients.
Lemma 22
Let \(p \ge 2\) and Assumption 7 be satisfied. Then there exists a constant \(C > 0\) such that for all progressively measurable \(u \in L^{p}_\omega L^\infty _t \big ( L^2_{\text {div}} \cap W^{1,p}_{0,x}\big )\) it holds
Proof
Due to the linearity of \(\mathbb {\epsilon } \Pi _{\text {div}}\) and \({\mathcal {I}}\) we find
Therefore, using the time regularity of stochastic integrals, cf. Theorem 35 (b),
Since \(L^p_x\) is of 2-smooth, we can use (A. 14). This and the growth assumption (2.12) establish
\(\square \)
Now we are ready to prove the result on temporal regularity of the non-linear gradient.
Proof of Theorem 12
The V-coercivity, cf. Lemma 38, allows to rewrite
It follows, using the symmetry and integration by parts,
Applying the strong formulation (2.11)
The first term is estimated by Hölder’s and Young’s inequalities
The stochastic term needs a more refined analysis. Due to the symmetry
Using Lemma 38 and Young’s inequality
The first term can be absorbed. The remaining term is estimated by Hölder’s and Young’s inequalities
We are in position to estimate the non-linear expression \(V(\mathbb {\epsilon } u)\). Integrate (4.4) in time and rescale by \(h^{-1}\). Moreover, take the supremum over h and expectation. Finally, apply (4.5) and (4.6) to find
Fubini’s theorem and the a priori estimate (2.13) imply
The time regularity of the stochastic integral, cf. Lemma 22, allows to conclude
Choosing \(\delta > 0\) sufficiently small and (4.2) lead to
Lastly, using the equivalence \(\varphi '(t) t \eqsim \varphi (t)\),
All together, we proved
\(\square \)
Remark 23
Strong solutions to the symmetric stochastic p-Stokes system (1.1) are comparable in terms of temporal regularity to the stochastic p-Laplace system, cf. [61].
In particular, if one could prove increased integrability, e.g. \( \Pi _{\text {div}}\text {div}S(\mathbb {\epsilon } u) \in L^2_{\omega } L^{q}_t L^2_x\) for some \(q > 2\), then it would also be possible to show \(V(\mathbb {\epsilon } u) \in L^2_\omega B^{1/2}_{q,\infty } L^2_x\).
The temporal regularity enables a similar error analysis for time discretizations of (1.1) as done in e.g. [14, 26] for the p-Laplace system.
Data availability
On behalf of all authors, the corresponding author states that the manuscript has no associated data.
References
Acerbi, E., Mingione, G.: Regularity results for parabolic systems related to a class of non-Newtonian fluids. Ann. Inst. H. Poincaré Anal. Non Linéaire 21(1), 25–60 (2004)
Acerbi, E., Mingione, G.: Regularity results for stationary electro-rheological fluids. Arch. Ration. Mech. Anal. 164(3), 213–259 (2002)
Agresti, A., Veraar, M.: The critical variational setting for stochastic evolution equations. arXiv e-prints, arXiv:2206.00230 (2022) [math.PR]
Barrett, J.W., Liu, W.B.: Finite element approximation of the p-Laplacian. Math. Comp. 61(204), 523–537 (1993)
Belenki, L., Diening, L., Kreuzer, C.: Optimality of an adaptive finite element method for the p-Laplacian equation. IMA J. Numer. Anal. 32(2), 484–510 (2012). https://doi.org/10.1093/imanum/drr016
Berselli, L.C., Diening, L., Ružiǩa, M.: Existence of strong solutions for incompressible fluids with shear dependent viscosities. J. Math. Fluid Mech. 12(1), 101–132 (2010). https://doi.org/10.1007/s00021-008-0277-y
Bhatia, H., Norgard, G., Pascucci, V., Bremer, P.-T.: The Helmholtz-Hodge decomposition. a survey. IEEE Trans. Visual Comput. Graphics 19(8), 1386–1404 (2012)
Bogovskii, M.E.: Solutions of some problems of vector analysis, associated with the operators div and grad. Theory of cubature formulas and the application of functional analysis to problems of mathematical physics. Vol. 1980. Trudy Sem. S. L. Soboleva, No. 1. Akad. Nauk SSSR Sibirsk. Otdel., Inst. Mat., Novosibirsk, pp. 5-40, 149 (1980)
Breit, D.: Analysis of generalized Navier–Stokes equations for stationary shear thickening flows. Nonlinear Anal. 75(14), 5549–5560 (2012). https://doi.org/10.1016/j.na.2012.05.003
Breit, D.: Existence theory for stochastic power law fluids. J. Math. Fluid Mech. 17(2), 295–326 (2015). https://doi.org/10.1007/s00021-015-0203-z
Breit, D., Diening, L., Storn, J., Wichmann, J.: The parabolic p-Laplacian with fractional differentiability. IMA J. Numer. Anal. 41(3), 2110–2138 (2021). https://doi.org/10.1093/imanum/draa081
Breit, D., Feireisl, E., Hofmanova, M.: Compressible fluids driven by stochastic forcing: the relative energy inequality and applications. Comm. Math. Phys. 350(2), 443–473 (2017). https://doi.org/10.1007/s00220-017-2833-x
Breit, D., Gmeineder, F.: Electro-rheological fluids under random influences: martingale and strong solutions. Stoch. Partial Differ. Equ. Anal. Comput. 7(4), 699–745 (2019). https://doi.org/10.1007/s40072-019-00138-6
Breit, D., Hofmanova, M., Loisel, S.: Space-time approximation of stochastic p-Laplace-type systems. SIAM J. Numer. Anal. 59(4), 2218–2236 (2021). https://doi.org/10.1137/20M1334310
Buckmaster, T., Vicol, V.: Convex integration and phenomenologies in turbulence. EMS Surv. Math. Sci. 6(1–2), 173–263 (2019). https://doi.org/10.4171/emss/34
Burczak, J., Kaplicky, P.: Evolutionary, symmetric p-Laplacian. Interior regularity of time derivatives and its consequences. Commun. Pure Appl. Anal. 15(6), 2401–2445 (2016). https://doi.org/10.3934/cpaa.2016042
Burczak, J., Kaplicky, P.: Interior regularity of space derivatives to an evolutionary, symmetric \(\varphi \)-Laplacian. Monatsh. Math. 183(1), 71–101 (2017). https://doi.org/10.1007/s00605-016-1005-y
Carelli, E., Prohl, A.: Rates of convergence for discretizations of the stochastic incompressible Navier-Stokes equations. SIAM J. Numer. Anal. 50(5), 2467–2496 (2012). https://doi.org/10.1137/110845008
Coscia, A., Mingione, G.: Holder continuity of the gradient of p(x)-harmonic mappings. C. R. Acad. Sci. Paris Ser. I Math. 328(4), 363–368 (1999). https://doi.org/10.1016/S0764-4442(99)80226-2
Diening, L., Ettwein, F., Ružička, M.:“\(C^{1,\alpha }\)-regularity for electrorheological fluids in two dimensions”. NoDEA Nonlinear Differential Equations Appl. 14(1-2) (2007), pp. 207-217. https://doi.org/10.1007/s00030-007-5026-z
Diening, L., Kaplicky, P., Schwarzacher, S.: Campanato estimates for the generalized Stokes system. Ann. Mat. Pura Appl. (4) 193(6), 1779–1794 (2014). https://doi.org/10.1007/s10231-013-0355-5
Diening, L., Ružička, M.: Interpolation operators in Orlicz–Sobolev spaces. Numer. Math. 107(1), 107–129 (2007). https://doi.org/10.1007/s00211-007-0079-9
Diening, L., Ettwein, F.: Fractional estimates for non-differentiable elliptic systems with general growth. Forum Math. 20(3), 523–556 (2008). https://doi.org/10.1515/FORUM.2008.027
Diening, L., Fornasier, M., Tomasi, R., Wank, M.: A Relaxed Kaanov iteration for the p-poisson problem. Numerische Mathematik (2020)
Diening, L., Harjulehto, P., Hasto, P., Ružička, M.: Lebesgue and Sobolev spaces with variable exponents, vol. 2017. Lecture Notes in Mathematics. Springer, Heidelberg, 2011, pp. x+509. https://doi.org/10.1007/978-3-642-18363-8
Diening, L., Hofmanova, M., Wichmann, J.: An averaged space.time discretization of the stochastic p-Laplace system. Numerische Mathematik (2022). https://doi.org/10.1007/s00211-022-01343-7
Diening, L., Kreuzer, C.: Linear convergence of an adaptive finite element method for the p-Laplacian equation. SIAM J. Numer. Anal. 46(2), 614–638 (2008). https://doi.org/10.1137/070681508
Diening, L., Ružička, M., Wolf, J. (2010) Existence of weak solutions for unsteady motions of generalized Newtonian fluids. Ann. Sc. Norm. Super. Pisa Cl. Sci. 9(1), 1–46
Fan, X.: Global \(C^{1,\alpha }\) regularity for variable exponent elliptic equations in divergence form. J. Differ. Equ. 235(2), 397–417 (2007). https://doi.org/10.1016/j.jde.2007.01.008
Feng, X., Prohl, A., Vo, L.: Optimally convergent mixed finite element methods for the stochastic Stokes equations. IMA J. Numer. Anal. 41(3), 2280–2310 (2021). https://doi.org/10.1093/imanum/drab006
Feng, X., Qiu, H.: Analysis of fully discrete mixed finite element methods for time dependent stochastic Stokes equations with multiplicative noise. J. Sci. Comput. 88(2), 25–31 (2021). https://doi.org/10.1007/s10915-021-01546-4
Foias, C., Manley, O., Rosa, R., Temam, R.: Navier-Stokes equations and turbulence. Vol. 83. Encyclopedia of Mathematics and its Applications. Cambridge University Press, Cambridge, 2001, pp. xiv+347. https://doi.org/10.1017/CBO9780511546754
Frehse, J., Schwarzacher, S.: On regularity of the time derivative for degenerate parabolic systems. SIAM J. Math. Anal. 47(5), 3917–3943 (2015). https://doi.org/10.1137/141000725
Geisert, M., Heck, H., Hieber, M.: On the equation div u = g and Bogovskii’s operator in Sobolev spaces of negative order. Partial differential equations and functional analysis. Vol. 168. Oper. Theory Adv. Appl. Birkhauser, Basel, 2006, pp. 113–121. https://doi.org/10.1007/3-7643-7601-5_7
Gess, B.: Strong solutions for stochastic partial differential equations of gradient type. J. Funct. Anal. 263(8), 2355–2383 (2012). https://doi.org/10.1016/j.jfa.2012.07.001
Hytonen, T.P., Veraar, M. C.: On Besov regularity of Brownian motions in infinite dimensions. Probab. Math. Statist. 28(1), 143–162 (2008)
Iwaniec, T., Manfredi, J.J.: Regularity of p-harmonic functions on the plane. Rev. Mat. Iberoamericana 5(1–2), 1–19 (1989). https://doi.org/10.4171/RMI/82
Jin, B.J., Kang, K., Lee, J.: On the regularity of weak solutions to the stationary motion of the degenerate power-law fluids. Ann. Univ. Ferrara Sez. VII Sci. Mat. 60(1), 169–188 (2014). https://doi.org/10.1007/s11565-013-0196-6
Ladyženskaja, O.A.: New equations for the description of the motions of viscous incompressible fluids, and global solvability for their boundary value problems. Trudy Mat. Inst. Steklov. 102, 85–104 (1967)
Ladyzhenskaya, O.A.: The mathematical theory of viscous incompressible flow. Mathematics and its Applications, Vol. 2. Second English edition, revised and enlarged, Translated from the Russian by Richard A. Silverman and John Chu. Gordon and Breach Science Publishers, New York-London-Paris, 1969, pp. xviii+224
Langa, J.A., Real, J., Simon, J.: Existence and regularity of the pressure for the stochastic Navier–Stokes equations. Appl. Math. Optim. 48(3), 195–210 (2003). https://doi.org/10.1007/s00245-003-0773-7
Le, K.-N., Wichmann, J.: A class of space-time discretizations for the stochastic p-Stokes system. arXiv e-prints (2023). arXiv: 2307.13253 [math.NA]
Lions, J.-L.: Quelques methodes de resolution des problemes aux limites non lineaires. Dunod, Paris; Gauthier-Villars, Paris, pp. xx+554 (1969)
Liu, W., Rockner, M.: SPDE in Hilbert space with locally monotone coefficients. J. Funct. Anal. 259(11), 2902–2922 (2010). https://doi.org/10.1016/j.jfa.2010.05.012
Liu, W., Rockner, M.: Stochastic partial differential equations: an introduction. Universitext. Springer, Cham, 2015, pp. vi+266. https://doi.org/10.1007/978-3-319-22354-4
Malek, J., Nečas, J., Rokyta, M., Ružička, M.: Weak and measure-valued solutions to evolutionary PDEs. Vol. 13. Applied Mathematics and Mathematical Computation. Chapman & Hall, London, 1996, pp. xii+317. https://doi.org/10.1007/978-1-4899-6824-1
Malek, J., Nečas, J., Ružička, M.: On weak solutions to a class of non-Newtonian incompressible fluids in bounded three-dimensional domains: the case p 2. Adv. Differential Equations 6(3), pp. 257–302 (2001)
Malek, J., Nečas, J., Ružička, M.: On the non-Newtonian incompressible fluids. Math. Models Methods Appl. Sci. 3(1), 35–63 (1993). https://doi.org/10.1142/S0218202593000047
Mikulevicius, R.: On the Cauchy problem for stochastic Stokes equations. SIAM J. Math. Anal. 34(1), 121–141 (2002). https://doi.org/10.1137/S0036141001390312
Mikulevicius, R., Rozovskii, B.L.: Stochastic Navier–Stokes equations for turbulent flows. SIAM J. Math. Anal. 35(5), 1250–1310 (2004). https://doi.org/10.1137/S0036141002409167
Naumann, J.: On the differentiability of weak solutions of a degenerate system of PDEs in fluid mechanics. Ann. Mat. Pura Appl. 151(4), 225–238 (1988). https://doi.org/10.1007/BF01762796
van Neerven, J.: \(\gamma \)-radonifying operators-a survey. The AMSI-ANU Workshop on Spectral Theory and Harmonic Analysis. Vol. 44. Proc. Centre Math. Appl. Austral. Nat. Univ. Austral. Nat. Univ., Canberra, pp. 1–61 (2010)
Ondrejat, M., Veraar, M.: On temporal regularity of stochastic convolutions in 2-smooth Banach spaces. Ann. Inst. Henri Poincare Probab. Stat. 56(3), 1792–1808 (2020). https://doi.org/10.1214/19-AIHP1017
Pisier, G.: Martingales in Banach spaces. Vol. 155. Cambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, pp. xxviii+561 (2016)
Rockner, M., Shang, S., Zhang, T.: Well-posedness of stochastic partial differential equations with fully local monotone coefficients. arXiv e-prints, arXiv:2206.01107. (2022) [math.PR]
Temam, R.: Navier–Stokes equations. Revised. Vol. 2. Studies in Mathematics and its Applications. Theory and numerical analysis, With an appendix by F. Thomasset. North-Holland Publishing Co., Amsterdam-New York, pp. x+519 (1979)
Terasawa, Y., Yoshida, N.: Stochastic power law fluids: existence and uniqueness of weak solutions. Ann. Appl. Probab. 21(5), 1827–1859 (2011). https://doi.org/10.1214/10-AAP741
Triebel, H.: Theory of function spaces. Vol. 78. Monographs in Mathematics. Birkhauser Verlag, Basel, p. 284. (1983) https://doi.org/10.1007/978-3-0346-0416-1
Triebel, H.: Theory of function spaces. II. Vol. 84. Monographs in Mathematics. Birkhauser Verlag, Basel, pp. viii+370 (1992). https://doi.org/10.1007/978-3-0346-0419-2
Wichmann, J.: Regularity and Numerics for stochastic p-Laplace and symmetric p-Stokes systems. (2022). https://doi.org/10.4119/UNIBI/2967438
Wichmann, J.: On temporal regularity of strong solutions to stochastic p-Laplace systems. SIAM J. Math. Anal. 55(4), 3713–3730 (2023). https://doi.org/10.1137/21m1460491
Wolf, J.: Existence of weak solutions to the equations of non-stationary motion of non-Newtonian fluids with shear rate dependent viscosity. J. Math. Fluid Mech. 9(1), 104–138 (2007). https://doi.org/10.1007/s00021-006-0219-5
Yoshida, N.: Stochastic shear thickening fluids: strong convergence of the Galerkin approximation and the energy equality. Ann. Appl. Probab. 22(3), 1215–1242 (2012). https://doi.org/10.1214/11-AAP794
Funding
Open Access funding enabled and organized by CAUL and its Member Institutions. The research was partially supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - SFB 1283/2 2021 - 317210226, and by the Australian Government through the Australian Research Council’s Discovery Projects funding scheme (Grant no. DP220100937).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is 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.
Appendix A. Miscellaneous
Appendix A. Miscellaneous
1.1 A.1. Fourier Representation of Besov Norms
There are many different characterizations of Besov spaces. A nice overview can be found e.g. in the books of Triebel [58, 59]. Some results derived from a specific characterization are not obvious for others. In this appendix we introduce the Fourier representation of Besov spaces and state some properties.
Definition 24
A family of functions \({\{{\varphi _j}\}}_{j\in {\mathbb {N}}_0}\) is called smooth dyadic partition of unity, if the following conditions are satisfied:
-
(a)
\(\varphi _j \in C^\infty ({\mathbb {R}})\),
-
(b)
$$\begin{aligned} {\left\{ \begin{array}{ll} \textrm{supp}\, \varphi _0 \subset \{t \in {\mathbb {R}}: \,\left| t\right| \le 2 \}, \\ \textrm{supp}\, \varphi _j \subset \{t \in {\mathbb {R}}: \,2^{j-1} \le \left| t\right| \le 2^{j+1} \}, \quad j \in {\mathbb {N}}, \end{array}\right. } \end{aligned}$$
-
(c)
for any \(k \in {\mathbb {N}}_0\)
$$\begin{aligned} \sup _{t\in {\mathbb {R}}, j\in {\mathbb {N}}_0} 2^{jk} \left| \partial ^k \varphi _j(t)\right| < \infty , \end{aligned}$$ -
(d)
and for any \(t \in {\mathbb {R}}\)
$$\begin{aligned} \sum _{j\in {\mathbb {N}}_0} \varphi _j(t) = 1. \end{aligned}$$
Now we are in a good shape to define the Fourier representation of Besov spaces on the full space \({\mathbb {R}}\) with values in a Banach space X.
Definition 25
Let \(\alpha \in {\mathbb {R}}\), \(p \in [1,\infty ]\) and \(q\in [1,\infty ]\). We define
where
Here \({\mathcal {F}}\) and \({\mathscr {S}}'({\mathbb {R}};X)\) denote Fourier transform and X-valued tempered distributions, respectively.
In contrast to the definition (2.1), where integrability of scaled increments is measured, (A. 2) measures integrability of frequency localized and scaled values. Note that (A. 2) is defined explicitly even for \(\alpha <0\). A generalization of the explicit definition (2.1) for negative \(\alpha \) is not clear. Nevertheless, for \(\alpha > 0\) one can verify equivalence of (2.1) and (A. 2) as done in [59, Sect. 1.5.1].
Theorem 26
Let \(\alpha > 0\), \(p \in (1,\infty )\) and \(q \in [1,\infty ]\). Then \( B^{\alpha }_{p,q}({\mathbb {R}};X) = {\tilde{B}}^{\alpha }_{p,q}({\mathbb {R}};X)\) and for all \(f \in B^{\alpha }_{p,q}({\mathbb {R}};X)\) it holds
Corollary 27
Let \(\alpha > 0\), \(p \in (1,\infty )\) and \(q \in [1,\infty ]\). Then \( \big ( B^{\alpha }_{p,q}({\mathbb {R}};X) \big )' = \big ( {\tilde{B}}^{\alpha }_{p,q}({\mathbb {R}};X)\big )'\) with equivalent norms.
The Fourier approach allows to characterize the dual spaces, cf. [58, Sect. 2.11.2].
Lemma 28
Let X be reflexive, \(\alpha \in {\mathbb {R}}\), \(p\in [1,\infty )\) and \(q \in [1,\infty )\). Then \(\big ({\tilde{B}}^{\alpha }_{p, q}({\mathbb {R}};X) \big )' = {\tilde{B}}^{-\alpha }_{p',q'}({\mathbb {R}};X')\).
Additionally, we can control the interaction of derivatives and the differentiability parameter in the Besov-norm.
Lemma 29
Let \(\alpha \in {\mathbb {R}}\), \(p \in [1,\infty ]\) and \(q \in [1,\infty ]\). Additionally, assume \(u \in {\tilde{B}}^{\alpha +1}_{p,q}({\mathbb {R}};X)\). Then \(\partial _t u \in {\tilde{B}}^{\alpha }_{p,q}({\mathbb {R}};X)\) and
Proof
Notice by [58, Theorem in Sect. 2.3.8] we have the equivalence
The claim follows. \(\square \)
1.2 A.2. Helmholtz-decomposition and Bogovskii Operator
The Helmholtz decomposition is a powerful tool in the analysis of fluids. It allows for a complete decoupling of the governing equations of the velocity field and the pressure. For a recent survey we refer to [7]. More details can be found e.g. in [32, Chap. 2 Sect. 3]. We abbreviate \(W^{k,q}_x:= W^{k,q}({\mathcal {O}})\).
The general idea is to decompose a vector field into a divergence-free vector field and a gradient of a potential. Let \(v \in L^2_x\) and \(G_v\) be a solution to
Note that \(G_v\) is defined up to a constant. We will prescribe the constant by the condition \(\langle G_v\rangle _{{\mathcal {O}}} = 0.\)
Now, we define the Helmholtz-Leray projection \(\Pi _{\text {div}}\) and its orthogonal complement \(\Pi _{\text {div}}^\perp \) by
In order to understand the range of the operator \(\Pi _{\text {div}}\) and \(\Pi _{\text {div}}^\perp \) we introduce the spaces
The next lemma can be found e.g. in [56, Theorem 1.4].
Lemma 30
(Helmholtz decomposition). \(L^2_x\) splits into the direct sum of the closed subspaces \(L^2_{\text {div}}\) and \((L^2_{\text {div}})^\perp \), i.e.,
Moreover, \(\Pi _{\text {div}}\) and \(\Pi _{\text {div}}^\perp \) are the \(L^2\)-orthogonal projections onto \(L^2_{\text {div}}\) and \((L^2_{\text {div}})^\perp \), respectively.
Remark 31
Since \(\Pi _{\text {div}}\) and \(\Pi _{\text {div}}^\perp \) are \(L^2\)-orthogonal projections, an equivalent definition to (A. 6) is given by
Stability of the Helmholtz-decomposition is particularly important, when it comes to the reconstruction of the pressure. Clearly, as an \(L^2\)-projection it trivially holds
Higher order stability is closely linked to regularity of solutions to the Laplace equation. Indeed, if \(\text {div}v \in L^2_x\), then (A. 5) is equivalent to the Neumann-Laplacian
i.e., \(G_v = \Delta ^{-1} \text {div}v \) is a weak solution to (A. 10).
Therefore, it is not surprising that improved stability properties of the Laplacian transfer to stability of the Helmholtz-Leray projection, cf. [56, Remark 1.6].
Theorem 32
. Let \({\mathcal {O}}\) be a \(C^{k+1}\)-domain and \(q\in (1,\infty )\). Then
Theorem 32 can be proven using the fundamental solution of the Neumann-problem (A. 10) and Calderon-Zygmund estimates. For example the result for Bessel-potential spaces is presented in [49, Lemma 3.6 & Lemma 3.8].
Corollary 33
. Let \({\mathcal {O}}\) be a bounded \(C^2\)-domain and \(q \in (1,\infty )\). Then
-
(a)
\(\nabla \Pi _{\text {div}}: W^{1,q}_{0,\text {div}} \rightarrow L^q_x\) is bounded,
-
(b)
\(\Pi _{\text {div}} \text {div}: L^{q'}_x \rightarrow \big ( W^{1,q}_{0,\text {div}} \big )'\) is bounded,
-
(c)
for all \(u \in W^{1,q}_{0,\text {div}}\) and \(S \in L^{q'}_x\)
$$\begin{aligned} \langle -\Pi _{\text {div}} \text {div}S, u \rangle =\int _{{\mathcal {O}}} S :\nabla \Pi _{\text {div}} u \,\textrm{d}x. \end{aligned}$$(A. 12)
Proof
The claims immediately follow by Theorem 32 and the density of smooth functions. \(\square \)
The right inverse of the divergence is nowadays called Bogovskii operator. It traces back to his work [8] and is a useful tool for the regularity analysis of the pressure.
Theorem 34
(Bogovskii’s operator [34, Theorem 2.5]) Let \(q \in (1,\infty ) \) and \({\mathcal {O}} \subset {\mathbb {R}}^n\) be a bounded domain with a locally Lipschitz boundary. Then there exists \({\mathcal {B}}: C^\infty _{0,x} \rightarrow C^\infty _{0,x}\) such that
Moreover, \({\mathcal {B}}\) extends continuously to a bounded operator from \(W^{s,q}_{0,x}\) to \(W^{s+1,q}_{0,x}\) provided \(s > -2 + 1/q\).
1.3 A.3. Stability of Stochastic Integration
The construction of the stochastic integral in general Banach spaces is delicate and one needs to look for a suitable generalization of the Itô isometry. It turns out that \(\gamma \)-radonifying operators are natural in the context of stochastic integration. The following result was obtained by Ondreját and Veraar and contains optimal stability of the stochastic integral driven by a cylindrical Wiener process for \(\gamma \)-radonifying operators with values in a separable 2-smooth Banach space. For more details on \(\gamma \)-radonifying operators we refer to the survey [52].
Theorem 35
(Stability of stochastic integrals [53, Theorem 3.2]). Let E be a separable 2-smooth Banach space. Additionally, let \(q \in [1,\infty )\), \(p \in (2,\infty ]\), \(\alpha = \tfrac{1}{2} - \tfrac{1}{p}\), \(\Phi _2(t):= e^{t^2}-1\), \(N_q(t):= t^q \ln ^{q/2}(t+1)\) and \(G \in L^0_{\mathcal {F}}\big (\Omega ;L^p(I;\gamma (U;E))\big )\). Then
-
(a)
\({\mathcal {I}}_W(G) \in B^{\alpha }_{\Phi _2,\infty }(I;E)\) a.s.,
-
(b)
\(\left( {\mathbb {E}} \left[ \left\| {\mathcal {I}}_W(G)\right\| _{B^{\alpha }_{q,\infty }(I;E)}^{2q} \right] \right) ^\frac{1}{2q}\le C \sqrt{q} \left\| G\right\| _{L^{2q}(\Omega ;L^{p}(I;\gamma (U;E)))}\),
-
(c)
\(\left( {\mathbb {E}} \left[ \left\| {\mathcal {I}}_W(G)\right\| _{B^{\alpha }_{\Phi _2,\infty }(I;E)}^{q} \right] \right) ^\frac{1}{q}\le C \sqrt{q} \left\| G\right\| _{L^{\infty }(\Omega ;L^{p}(I;\gamma (U;E)))}\),
-
(d)
\(\left\| {\mathcal {I}}_W(G)\right\| _{L^{\Phi _2}(\Omega ;B^{\alpha }_{\Phi _2,\infty }(I;E)} \le C \left\| G\right\| _{L^\infty (\Omega ;L^p(I;\gamma (U;E)))}\),
-
(e)
\(\left\| {\mathcal {I}}_W(G)\right\| _{L^q(\Omega ;B^{\alpha }_{\Phi _2,\infty }(I;E))} \le C \sqrt{q} \left\| G\right\| _{L^{N_q}(\Omega ;L^p(I;\gamma (U;E)))}\).
We are particularly interested in the cases \(E = W^{1,p}_{0,x}\) for \(p \ge 2\). This space is 2-smooth, cf. [54]. Naturally, if E is 2-smooth, then Hilbert-Schmidt operators are \(\gamma \)-radonifying, i.e., \(L_2(U;E) \subset \gamma (U;E)\). Indeed,
where \(C_2\) is the martingale type 2 constant of E and \({\{{\gamma _j}\}}_{j\in {\mathbb {N}}} \sim {\mathcal {N}}(0,1)\) are independent and identically distributed.
1.4 A.4. Monotonicity and Potentials
A continuous, convex and strictly increasing function \(\phi \,:\, [0,\infty ) \rightarrow [0,\infty )\) satisfying
is called an N-function.
We say that \(\phi \) satisfies the \(\Delta _2\)–condition, if there exists \(c > 0\) such that for all \(t \ge 0\) holds \(\phi (2t) \le c\, \phi (t)\). By \(\Delta _2(\phi )\) we denote the smallest such constant. Since \(\phi (t) \le \phi (2t)\) the \(\Delta _2\)-condition is equivalent to \(\phi (2t) \eqsim \phi (t)\) uniformly in t. Note that if \(\Delta _2(\phi ) < \infty \) then \(\phi (t) \eqsim \phi (c\,t)\) uniformly in \(t\ge 0\) for any fixed \(c>0\). For a family \(\phi _\lambda \) of N-functions we define \(\Delta _2({\{{\phi _\lambda }\}}):= \sup _\lambda \Delta _2(\phi _\lambda )\).
By \(\phi ^*\) we denote the conjugate N-function of \(\phi \), which is given by \(\phi ^*(t) = \sup _{s \ge 0} (st - \phi (s))\). Then \(\phi ^{**} = \phi \).
More details can be found e.g. in [5, 22,23,24,25].
Definition 36
Let \(\phi \) be an N-function. We say that \(\phi \) is uniformly convex, if \(\phi \) is \(C^1\) on \([0,\infty )\) and \(C^2\) on \((0,\infty )\) and assume that
uniformly in \(t > 0\). The constants hidden in \(\eqsim \) are called the characteristics of \(\phi \).
Note that (A. 15) is stronger than \(\Delta _2(\phi ,\phi ^*)<\infty \). In fact, the \(\Delta _2\)-constants can be estimated in terms of the characteristics of \(\phi \).
Associated to an uniformly convex N-function \(\phi \) we define the tensors
Sometimes it is convenient to change the growth of an N-function near 0. One possibility is to introduce the shifted N-function \(\phi _\kappa \) for \(\kappa \ge 0\) by
We define \(S_\kappa \) and \(V_\kappa \) analogously to (A. 16). However, we neglect in the notation the dependence of \(S_\kappa \) and \(V_\kappa \) on \(\kappa \), since the characteristic of \(\varphi _{\kappa }\) is uniform in \(\kappa \ge 0\).
Lemma 37
([23, Lemma 27]). Let \(\phi \) be an N-function with \(\Delta _2(\phi , \phi ^*) < \infty \). Then the family \({\{{\varphi _\kappa }\}}_{\kappa \ge 0}\) satisfies \(\sup _{\kappa \ge 0} \Delta _2\big ( \varphi _\kappa , (\varphi _\kappa )^* \big ) < \infty \).
Lemma 38
(Equivalence lemma [23, Lemma 3]). We have
uniformly in \( P, Q \in \mathbb {R}^{n \times n}\). Moreover, uniformly in \( Q \in \mathbb {R}^{n \times n}\),
The constants depend only on the characteristics of \(\phi \).
Lemma 39
([23, Lemma 32]). Let \(\varphi \) be an uniformly convex N-function. Then for all \(\delta >0\) there exists \(C_\delta > 0\) such that for all \(t,\,u \ge 0\)
The following inequality follows immediately from Lemma 38 and Lemma 39.
Lemma 40
(Young type inequality). Let \(\phi \) be an uniformly convex N-function. Then for each \(\delta >0\) there exists \(C_\delta \ge 1\) (only depending on \(\delta \) and the characteristics of \(\phi \)) such that
for all \(P, Q, R\in {\mathbb {R}}^{N\times n}\).
Lemma 41
(Change of Shift [24, Lemma 42]). Let \(\phi \) be an uniformly convex N-function. Then for each \(\delta >0\) there exists \(C_\delta \ge 1\) (only depending on \(\delta \) and the characteristics of \(\phi \)) such that
for all \( P, Q\in \mathbb {R}^{n \times n}\) and \(t\ge 0\).
1.5 A.5. Gradient flow
The Helmholtz-Leray projection \(\Pi _{\text {div}}\) allows to reformulate (1.1a) and (1.1b) into
Now it is possible to interpret (A. 19a) as a stochastically perturbed gradient flow of the energy
induced by the potential \(\varphi _{\kappa }(t):= \int _0^t (\kappa + s)^{p-2} s \,\textrm{d}s\) on the space \(W^{1,p}_{0,\text {div}}\).
It has been observed by Diening and Kreuzer [27] that the energy gap of \(u \in W^{1,p}_x\) towards a minimizer \( u^* \in \textrm{arg min}_{v \in W^{1,p}_x} {\mathcal {J}}(v)\) is proportional to the distance measured in the V-quasi norm, i.e.,
In fact, the result can be generalized to energy gaps between arbitrary functions \(u,v \in W^{1,p}_x\).
Lemma 42
Let \(u,v \in W^{1,p}_x\). Then
Proof
Due to a Taylor expansion it holds
where the first and second order Gateaux derivatives are given by
and
respectively.
Following the proof of [27, Lemma 16] we find
This establishes the claim. \(\square \)
Clearly, if we insert a minimizer \(v = v^*\) in (A. 22), then we recover (A. 21). However, the estimate (A. 22) ensures that \({\mathcal {J}}\) is strongly convex.
In order to identify (A. 19a) as a perturbed gradient flow, we need to find a pointwise representation of the gradient \(D {\mathcal {J}}(u)\).
Proposition 43
Let \({\mathcal {O}}\) be a bounded \(C^2\)-domain and \(u \in W^{1,p}_{0,x}\). Then for all \(\xi \in W^{1,p}_{0,\text {div}}\)
If additionally \(\xi \in C^\infty _{0,\text {div}}\) and \(\Pi _{\text {div}} \text {div}S(\mathbb {\epsilon } u) \in L^1_{\textrm{loc}}\), then
Proof
Let \(\xi \in W^{1,p}_{0,\text {div}}\). Due to (A. 23), (A. 16) and the symmetry of \(S(\mathbb {\epsilon } u)\)
Next, we use that \(\Pi _{\text {div}}\) is a projection, \(\xi \in {\mathcal {R}}(\Pi _{\text {div}})\) and (A. 12),
This establishes (A. 26).
The second part follows by Riesz’s representation theorem. \(\square \)
Remark 44
We want to emphasize that in general it is not possible to decouple the projection and the divergence, i.e., \(\Pi _{\text {div}} \text {div}\) should be understood as one operator (defined by the duality relation (A. 12)) rather than the composition of two.
However, for smooth functions it is exactly the composition of the divergence and the Helmholtz-Leray projection.
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
Wichmann, J. Temporal Regularity of Symmetric Stochastic p-Stokes Systems. J. Math. Fluid Mech. 26, 20 (2024). https://doi.org/10.1007/s00021-024-00852-9
Accepted:
Published:
DOI: https://doi.org/10.1007/s00021-024-00852-9
Keywords
- SPDEs
- Non-linear Laplace-type systems
- Strong solutions
- Regularity
- Stochastic p-Stokes system
- power-law fluids
- Generalized fluids