Fractional Navier-Stokes Equations with Caputo Derivative Driven by Hermite Noise

Abstract: We study time-fractional stochastic Navier-Stokes equations on a bounded domain of $\R<sup>2$ (the restriction to dimension two is essential for the bilinear estimates via Sobolev embeddings) driven by a Hermite process $Z_H<sup>k$ of order $k\ge1$ and Hurst parameter $H\in(1/2,1)$. This class of noises generalizes fractional Brownian motion ($k=1$) and the Rosenblatt process ($k=2$). We construct the Wiener integral with respect to $Z_H<sup>k$ and establish sharp $L<sup>p$ estimates via hypercontractivity, explicitly capturing the dependence on $k$. Using a refined Hilbert-Schmidt estimate for the Mittag-Leffler operator, we prove that the stochastic convolution belongs to $\dot{H}<sup>ν$ under the condition $\al(1-ν)+2H&gt;2$. A fixed-point argument in a weighted space yields the existence, uniqueness, and Hölder regularity of mild solutions. We also prove a non-central limit theorem linking the solution to discrete approximations.

• The paper presents a novel formulation of 2D time-fractional Navier-Stokes equations with a Caputo derivative and non-Gaussian Hermite noise, generalizing traditional Gaussian models.
• It employs advanced techniques such as hypercontractivity and Hilbert-Schmidt estimates to rigorously establish local existence, uniqueness, and Hölder regularity of solutions.
• The work derives a non-central limit theorem linking discrete Hermite polynomial approximations to continuous Hermite-driven stochastic dynamics, highlighting its practical implications.

Fractional Navier-Stokes Equations with Caputo Derivative Driven by Hermite Noise: A Technical Synthesis

Introduction and Motivation

The paper addresses time-fractional stochastic Navier-Stokes equations (NSEs) in two dimensions, enhanced with a Caputo fractional derivative in time and subjected to Hermite-process-driven noise of arbitrary order $k\ge 1$. This construction generalizes previous treatments restricted to Gaussian noise or fractional Brownian motion, encompassing a broader class of non-Gaussian, long-range dependent drivers. The analysis focuses on bounded domains and adheres strictly to additive noise, given that the non-Gaussian structure of Hermite processes, especially for $k\ge 2$, presents substantial analytic obstacles for the multiplicative case.

The motivation arises from the physical relevance of memory and anomalous diffusion effects in complex and viscoelastic fluids, as well as the need for more realistic stochastic perturbations beyond standard Gaussian frameworks. The Caputo derivative permits physically interpretable initial conditions, and the generality of Hermite noise allows modeling noise effects that manifest non-central limit behaviors routinely observed in statistical hydrodynamics.

Model Formulation

The evolution equation considered is

$$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$

where $^C D_t^\alpha$ is the Caputo derivative, $A$ is the Stokes operator (arising from the Dirichlet-Laplacian restricted to divergence-free fields), $B(u)$ is the bilinear NSE nonlinearity, $f$ is a Lipschitz continuous deterministic force, and the stochastic term is driven by the formal time derivative of a Hermite process $Z_H^k$ with Hurst parameter $H \in (1/2,1)$. The system is cast in the mild (variation-of-constants) form, using Mittag-Leffler operators $E_\alpha$, $k\ge 2$0, and a stochastic convolution defined by

$k\ge 2$1

with all terms understood in the proper infinite-dimensional and weighted Sobolev settings.

Technical Contributions and Analytical Framework

1. Construction and Analysis of the Hermite Wiener Integral

The stochastic integral with respect to $k\ge 2$2 is formulated by direct extension from elementary functions via Wiener-Itô chaos expansions. For deterministic integrands $k\ge 2$3 in the Hermite covariance Hilbert space $k\ge 2$4, the resulting integral $k\ge 2$5 resides in the $k\ge 2$6-th Wiener chaos and satisfies a sharp hypercontractivity inequality: $k\ge 2$7 for every $k\ge 2$8. This succinctly exposes the polynomial moment growth inherent in non-Gaussian Hermite chaos (notably, for $k\ge 2$9 one recovers Gaussian $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$0 bounds).

2. Regularity of the Stochastic Convolution

Employing precise Hilbert-Schmidt estimates for the Mittag-Leffler operator $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$1 (obtained via spectral calculus and Weyl asymptotics), the regularity of $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$2 is established in the weighted Sobolev scale. The key requirement $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$3 (where $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$4 controls the spatial regularity) guarantees that, for admissible parameters and any $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$5,

$$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$6

where $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$7 scales as $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$8. Time increment regularity is shown by decomposing increments and leveraging the smoothing and scaling properties of $$^C D_t^\alpha u = -A u + B(u) + f(u) + \dot{Z}_H^k(t), \quad u(0) = u_0,$$9. The stochastic convolution thus admits explicit Hölder exponents determined by the interplay between $^C D_t^\alpha$0, $^C D_t^\alpha$1, and $^C D_t^\alpha$2.

3. Local Existence, Uniqueness, and Hölder Regularity of Solutions

Existence and uniqueness are proven via a contraction mapping principle in weighted $^C D_t^\alpha$3 spaces, counteracting the singularity of $^C D_t^\alpha$4 at $^C D_t^\alpha$5 with carefully chosen temporal weights. Bilinear estimates for $^C D_t^\alpha$6, Lipschitz control on $^C D_t^\alpha$7, and the $^C D_t^\alpha$8-hypercontractivity for the noise collectively yield local well-posedness for suitably regular initial data and parameter values.

Hölder regularity in time is established for the mild solution $^C D_t^\alpha$9, with the minimal Hölder exponent $A$0 given by

$A$1

reflecting the maximal regularity propagated by the initial condition, the smoothing action of the fractional evolution, and the Hermite noise, respectively.

4. Non-central Limit Theorems and Discrete Approximations

A non-central limit theorem is formulated connecting the continuum Hermite process $A$2 to the limit of discrete sums of Hermite polynomials of long-range dependent Gaussian sequences. Let $A$3 be such a sequence; then

$A$4

converges in finite-dimensional distribution to $A$5, and the associated solutions $A$6 to the approximated NSE converge in law to the continuous solution $A$7 within $A$8. This bridges the gap between discrete models (e.g., numerical approximations or physical systems with microscopic memory) and SPDEs with Hermite-driven forcing.

Implications and Perspectives

The systematic analysis of the stochastic Navier-Stokes system with time-fractional Caputo dynamics and Hermite noise advances the theoretical understanding of how non-Gaussian, memory-enhanced stochasticity propagates in active fluids. The explicit dependence of moment and regularity estimates on $A$9 and $B(u)$0 opens up new avenues for statistical inference and parameter identification. The non-central limit theorem crystallizes the universality of Hermite processes as macroscopic limits of nonlinear functionals of long-memory microscopic dynamics.

From a technical viewpoint, the deployment of hypercontractivity to handle Hermite chaos is robust and essential, given the unavailability of standard Gaussian calculus for higher $B(u)$1. The restriction to additive noise is necessary under current technology, as stochastic integration with random (mild) integrands in the Hermite setting remains an open challenge—future extensions will depend heavily on further developments in Malliavin calculus for non-Gaussian chaos.

The results also raise critical directions for future work: global regularity (even in $B(u)$2) is unresolved, the complete theory for multiplicative Hermite noise is lacking, and statistical estimation for physical parameters based on discrete data stands as an important open problem. Extensions to three dimensions confront the usual criticality barriers in fluid dynamics.

Conclusion

The paper establishes a full local existence, uniqueness, and regularity theory for two-dimensional time-fractional stochastic Navier-Stokes equations with additive Hermite noise and Caputo time derivatives. Moment bounds and path regularity reflect the higher-order, non-Gaussian structure of Hermite processes. The work not only generalizes Gaussian/fractional Brownian perturbations but also aligns the mathematical theory with observed non-Gaussian scaling phenomena in turbulent fluids with memory effects. The analytical program outlined here provides a foundation for subsequent advances in both stochastic PDEs and the theory of non-Gaussian, long-memory driven systems.

Reference: "Fractional Navier-Stokes Equations with Caputo Derivative Driven by Hermite Noise" (2604.10602)