Stochastic 2-D Navier-Stokes Equation with Artificial Compressibility
Utpal Manna, Jose-Luis Menaldi, and Sivaguru S. Sritharan

TL;DR
This paper investigates the stochastic 2-D Navier-Stokes equation with artificial compressibility, establishing existence, uniqueness, and the transition to incompressible flow using a local monotonicity approach.
Contribution
It provides the first rigorous proof of existence, uniqueness, and the incompressible limit for stochastic Navier-Stokes equations with artificial compressibility.
Findings
Proved existence and uniqueness of strong solutions.
Established the limit to incompressible flow.
Utilized local monotonicity of operators for analysis.
Abstract
In this paper we study the stochastic Navier-Stokes equation with artificial compressibility. The main results of this work are the existence and uniqueness theorem for strong solutions and the limit to incompressible flow. These results are obtained by utilizing a local monotonicity property of the sum of the Stokes operator and the nonlinearity.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Stochastic -D Navier-Stokes Equation with Artificial
Compressibility
Utpal Manna
Utpal Manna: Department of Mathematics, University of Wyoming, Laramie, WY 82071, USA
[email protected] http://uwyo.edu/utpal/ ,
J.L. Menaldi
J.L. Menaldi: Department of Mathematics, Wayne State University, Detroit, MI 48202, USA
[email protected] http://www.math.wayne.edu/ menaldi/ and
S.S. Sritharan
S.S. Sritharan: Department of Mathematics, University of Wyoming, Laramie, WY 82071, USA
[email protected] http://math.uwyo.edu/Prof_home/Sri.html
Abstract.
In this paper we study the stochastic Navier-Stokes equation with artificial compressibility. The main results of this work are the existence and uniqueness theorem for strong solutions and the limit to incompressible flow. These results are obtained by utilizing a local monotonicity property of the sum of the Stokes operator and the nonlinearity.
Key words and phrases:
Stochastic Navier-Stokes equations, Artificial compressibility, Local monotonicity
2000 Mathematics Subject Classification:
Primary 76D05; Secondary 35Q30, 60H15, 76D03, 76D06
- This research is supported by Army Research Office, Probability and Statistics Program, grant number DODARMY1736
1. Introduction
The stochastic Navier-Stokes equation is a well accepted model for atmospheric, aero and ocean dynamics. Chandrasekhar [5] and Novikov [14] first studied the Navier-Stokes equation with external random forces. After that several approaches have been poposed, from the classic paper by Bensoussan and Temam [4] to some more recent results, e.g., by Bensoussan [3], by Flandoli and Gatarek [8] and by Menaldi and Sritharan [13].
This paper is concerned with the existence and uniqueness of strong solutions for the Stochastic 2-D Navier-Stokes equation with artificial compressibility in bounded domains. The concept of artificial compressibility was first introduced by Chorin [6, 7] and Temam [17, 18], in order to overcome the computational difficulties connected with the incompressibility constraint. Using the classical Sobolev compactness embedding and exploiting the classical Lions [12] method of fractional derivatives, Temam in his papers [17, 18] and in his book [19](chapter 3) proved the existence, uniqueness and convergence of the deterministic Navier-Stokes equation with artificial compressibility in bounded domains.
In the rest of this section we formulate the abstract Navier-Stokes problem with artificial compressibility. We describe some standard well known results including the local monotonicity property of the Navier-Stokes operator. In Section 2 we establish certain new a priori estimates involving exponential weight for stochastic Navier-Stokes equation with artificial compressibility. These estimates play a fundamental role in the proof of existence and uniqueness of strong solutions proved in the second half of Section 2. The monotonicity argument used here is the generalization of the classical Minty-Browder method for dealing with local monotonicity. This method was first used by Menaldi and Sritharan [13] and for multiplicative noise by Sritharan and Sundar [16]. For similar ideas see also Barbu and Sritharan [1, 2]. In the last part of Section 2 we discuss the convergence of the corresponding perturbed problem. Here the use of local monotonicity avoids the classical method based on compactness and thus the results apply to unbounded domains and hence the existence and the uniqueness as well as convergence to incompressible flow are new even in the deterministic case.
2. Abstract Mathematical Framework and Local Monotonicity
Let be a bounded domain (for the sake of simplicity) with smooth boundary, the velocity and the pressure fields. The Navier-Stokes problem (with Newtonian constitutive relationship and artificial compressible medium) can be written as follows
[TABLE]
with the initial conditions
[TABLE]
where is a vanishing parameter, belong to is the kinematic viscosity, denotes pressure and is a scalar-valued function and the (force) field is in A solution should belongs to the space with and its dual space. The second equation in (2.1) is an artificial state equation of a slightly compressible medium and the extra term is a stabilization term to handle the nonlinearity. The standard spaces used here are as follows:
with the norm
[TABLE]
and with the norm
[TABLE]
Using the Gelfand triple we may consider or as a linear map from or into the dual of respectively. The inner product in the or is denoted by and the induced duality by Thus, for any and in and in we have
[TABLE]
[TABLE]
and
[TABLE]
It is clear that is a linear continuous operator from into Next, an integration by parts and Hölder inequality yields
[TABLE]
and in the right-hand-side we can use -norms to estimate the product
Lemma 2.1**.**
For any real-valued smooth functions and with compact support in the following hold:
[TABLE]
Proof.
The results stated above are classical and well known [10]. ∎
As in Temam [19], chapter 3, the non-linear term is a trilinear continuous form on
[TABLE]
where
[TABLE]
We have the following lemmas.
Lemma 2.2**.**
Let and be in the spaces and respectively. Then the following estimate holds:
[TABLE]
Proof.
We observe that
[TABLE]
Using the Hölder inequality,
[TABLE]
In Frobenius norm divergence can be estimated by gradient. Hence
[TABLE]
Using the equation (2.11) in Lemma 2.1 we get
[TABLE]
∎
Lemma 2.3**.**
Let and be in the space . Then the following estimate holds:
[TABLE]
Proof.
For any given and in , we have from (2.12)
[TABLE]
and
[TABLE]
Then using (2.16) we obtain
[TABLE]
Now
[TABLE]
Hence using (2.16) and (2.17) we have
[TABLE]
With the help of (2.16) and (2.17) last three terms of the right hand side vanish.Thus
[TABLE]
[TABLE]
which by Lemma 2.2 gives the estimate
[TABLE]
where and denotes the norm in and respectively.
Now using the fact that for any two real numbers , and any , with ,
[TABLE]
we obtain the estimate (2.15). ∎
Notice that is continuously included in and is a (nonlinear) continuous mapping from into its dual Hence the nonlinear operator can be considered as a map from into the space . Then combination of previous lemmas yield the following local monotonicity property.
Lemma 2.4**.**
For a given , let us denote by the (closed) -ball in
[TABLE]
then the nonlinear operator is monotone in the convex ball i.e.,
[TABLE]
* and *
Proof.
First, it is clear that
[TABLE]
and the equation(2.15) yields
[TABLE]
Summing these equations up we get the desired result (2.22). ∎
3. Stochastic -D Navier-Stokes Equation with
Artificial Compressibility
Let us consider the Navier-Stokes equation subject to a random (Gaussian) term i.e., the forcing field has a mean value still denoted by and a noise denoted by We can write (to simplify notation we use time-invariant forces) and the noise process as a series where and are regarded as -valued functions in and respectively. The stochastic noise process represented by is normal distributed in with a trace-class co-variance operator denoted by and given by
[TABLE]
We interpret the stochastic Navier-Stokes equations as an Itô stochastic equations in variational form
[TABLE]
in with the initial condition
[TABLE]
for any in the space and any in
A finite-dimensional (Galerkin) approximation of the stochastic Navier-Stokes equation can be defined as follows. Let be a complete orthonormal system (i.e., a basis) in the Hilbert space belonging to the space (and ). Denote by the -dimensional subspace of and of all linear combinations of the first elements Also denote by the image of
Consider the following stochastic ODE in
[TABLE]
in with the initial condition
[TABLE]
for any in the space and in The coefficients involved are locally Lipschitz and we need some a priori estimate to show global existence of a solution as an adapted process in the space
Proposition 3.1** (energy estimate).**
Under the above mathematical setting let
[TABLE]
Let be an adapted process in which solves the stochastic ODE (3.4). Then we have the energy equality
[TABLE]
which yields the following estimate for any
[TABLE]
for any Moreover, if we suppose
[TABLE]
then we also have
[TABLE]
for some constant depending only on and
Proof.
From (3.4) we notice that,
[TABLE]
It is clear that
[TABLE]
and the equation (2.8) yields
[TABLE]
Hence
[TABLE]
Using the equation we get from (2.6)
[TABLE]
Combining all the above results one can get from (3)
[TABLE]
Rearranging the terms we get the desired energy equality (LABEL:e2.7).
Next, we calculate the stochastic differential of the process
[TABLE]
to get
[TABLE]
Using the energy equality (LABEL:e2.7) we have
[TABLE]
Now using the inequality
[TABLE]
on we have
[TABLE]
Then (3.14) yields
[TABLE]
Rearranging the terms we have
[TABLE]
Next we integrate in to get
[TABLE]
Finally taking mathematical expectation and keeping in mind the expectation of a stochastic integral is zero, we have the desired result (3.8).
Similarly, consider
[TABLE]
and use Itô calculus. Here we check that its stochastic differential satisfies
[TABLE]
Using the energy equality (LABEL:e2.7) we get
[TABLE]
Simplification and rearrangement of the terms in the above equation yields
[TABLE]
Now in the first and the second terms on the right hand side of the above equation we apply the following elementary inequality
[TABLE]
where
Then choosing
[TABLE]
we get
[TABLE]
where the constant depends only on and .
Similarly with proper choices of and , we can prove that
[TABLE]
Then (3) yields
[TABLE]
Integrating the stochastic differential (3),then taking the sup norm in and finally taking the mathematical expectation we have
[TABLE]
By means of martingale inequality, we deduce
[TABLE]
where the constant depends only on and
Using (3.18) in (3) we get the desired estimate (3.10). ∎
Now we deal with the existence and uniqueness of the SPDE and its finite-dimensional approximation.
Proposition 3.2** (uniqueness).**
Let be a solution of the stochastic Navier-Stokes equation (SPDE) (LABEL:e2.2) with the regularity
[TABLE]
and let the data and satisfy the condition
[TABLE]
If in is another solution of the stochastic Navier-Stokes equation (LABEL:e2.2), then
[TABLE]
with probability for any and
Proof.
Indeed if and are two solutions then solves the deterministic equation
[TABLE]
with Notice that actually and are better processes, they belong to the space
Next, setting
[TABLE]
we have
[TABLE]
Using (2.6) we get
[TABLE]
Since , Lemma 2.3 and (3) yield
[TABLE]
Hence
[TABLE]
Hence, integrating in , we deduce (3.21), with probability . ∎
Each solution in the space of the stochastic Navier-Stokes equation actually belongs to a better space, namely the space in 2-D, Thus in 2-D, the uniqueness holds in the space
If a given adapted process in satisfies
[TABLE]
for any function in and some functions in and in then we can find a version of (which is still denoted by ) in satisfying the energy equality
[TABLE]
see e.g. Gyongy and Krylov [9], Pardoux [15].
Definition 3.3**.**
() A strong solution is defined on a given probability space (, , , ) as a valued function which satisfies the stochastic Navier-Stokes equation (LABEL:e2.2) in the weak sense and also the energy inequality
[TABLE]
where the constant depends only on and
Proposition 3.4** (2-D existence).**
Let and be such that
[TABLE]
for some Then there is adapted processes and with the regularity
[TABLE]
satisfying the stochastic Navier-Stokes equation (LABEL:e2.2) and the a priori bound (3.10) for every
Proof.
Denoting
[TABLE]
we have
[TABLE]
Then using the a priori estimate (3.10), it follows from the Banach-Alaoglu theorem that along a subsequence, the Galerkin approximations have the following limits:
[TABLE]
where has the Itô differential
[TABLE]
and the energy equality holds, i.e.,
[TABLE]
Now, for any adapted process in we define
[TABLE]
as an adapted, continuous (and bounded in ) real-valued process in . Then from the energy equality
[TABLE]
Integrating between and taking the mathematical expectation we have
[TABLE]
Considering the fact that the initial conditions and converge to and respectively in , and the lower-semi-continuity of the -norm, we deduce
[TABLE]
Next, by monotonicity on -balls, i.e. by Lemma 2.4, we have
[TABLE]
Rearranging the terms we find
[TABLE]
Taking limit in , we get
[TABLE]
Now we take with and is an adapted process in
.
Then we have
[TABLE]
Dividing by on both sides of the inequality above, and letting go to [math], one obtains
[TABLE]
Since is arbitrary, we conclude that . Thus the existence of a strong solution of the stochastic Navier-Stokes equation (LABEL:e2.2) has been proved. ∎
4. Convergence as
We will now study the asymptotic limit of . Let us consider the family of perturbed systems(depending on the positive parameter ).
[TABLE]
with the initial conditions
[TABLE]
This is a method to overcome the computational difficulties connected with the constraint ””. The equations (4.1)-(4.2) are easier to approximate than the original stochastic Navier-Stokes equation as the constraint ”” has been replaced by the evolution equation (4.2).
Here we will show how the solutions of the perturbed problems converge to the solutions of the incompressible stochastic Navier-Stokes equation as . The idea of the proof is similar to the deterministic case presented in Temam [19].
Let
[TABLE]
Then we can write the above mentioned perturbed systems as Itô Stochastic equations in variational form
[TABLE]
in with the initial conditions
[TABLE]
for any in the space and any in
Proposition 4.1**.**
As , the solutions of the equations (LABEL:e2.31)-(4.5) converge to the solution of the incompressible stochastic Navier-Stokes equation.
Proof.
First we should point out that the solutions of the equations (LABEL:e2.31)-(4.5) satisfy the monotonicity property in Lemma 2.4, the energy equality (LABEL:e2.7) and the a priori estimates (3.8) and (3.10). By virtue of these a priori estimates and using the Banach-Alaoglu theorem, along a subsequence the approximations have the following limits:
[TABLE]
Let us denote
[TABLE]
and
[TABLE]
Let be a scalar function on with . Multiplying the equation (LABEL:e2.31) by , integrating in and taking mathematical expectation, we obtain
[TABLE]
Now passing to the limit in (4.7) we have in the sense of distribution,
[TABLE]
Hence in the same sense
[TABLE]
Then passing to the limit in the following equation
[TABLE]
we get
[TABLE]
which implies that .
Hence
Now it is clear that
[TABLE]
Now using the same Minty-Browder monotonicity argument used in Proposition 3.4 we can show that in limit
[TABLE]
Using the results (4.10) and (4.11), in limit we have from equation (4)
[TABLE]
This proves that is a solution of the incompressible stochastic Navier-Stokes equation. ∎
Remark 4.2*.*
If one considers the multiplicative noise of the type considered in the hypotheses in Sritharan and Sundar [16], then under these conditions same a priori estimates (3.8)-(3.10) hold. Thus weakly in , where denote the space of linear operators such that is a Hilbert-Schmidt operator from to and the norm on the space is defined by and is a trace class operator. Hence with the help of Minty-Browder monotonicity arguments the existence and uniqueness proofs go through and we can also establish the limit to incompressible flow.
Acknowledgment. We thank the reviewer for the helpful comments.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] Barbu, V. and Sritharan, S.S.: m 𝑚 m -accretive quantization of the vorticity equation, Semigroups of operators: theory and applications (Newport Beach, CA, 1998) , 296–303, Progr. Nonlinear Differential Equations Appl. , 42 , Birkh user, Basel, 2000.
- 2[2] Barbu, V. and Sritharan, S.S.: Flow invariance preserving feedback controllers for the Navier-Stokes equation, J. Math. Anal. Appl. , 255 (2001), no. 1, 281–307.
- 3[3] Bensoussan, A.: Stochastic Navier-Stokes equations, Acta. Appl.Math. , 38 (1995), 267–304.
- 4[4] Bensoussan, A. and Temam, R.: Equations stochastique du type Navier-Stokes, J. Funct. Anal. , 13 (1973), 195–222.
- 5[5] Chandrasekhar, S.: Stochastic, Statistical and Hydromagnetic Problems in Physics and Astronomy , Selected Papers, Vol 3, University of Chicago Press, Chicago, 1989.
- 6[6] Chorin, A.J.: Numerical solutions of the Navier-Stokes equations, Math. Comp. , 22 (1968), 745–762.
- 7[7] Chorin, A.J.: On the convergence of discrete approximations to the Navier-Stokes equations, Math. Comp. , 23 (1969), 341-353.
- 8[8] Flandoli, F. and Gatarek, D.: Martingale and stationary solutions for the stochastic Navier-Stokes equation, Probab. Theory Rel. Fields , 102 (1995), 367–391.
