A Spectral Representation of a Weighted Random Vectorial Field: Potential Applications to Turbulence and the Problem of Anomalous Dissipation in the Inviscid Limit

Abstract

Let ${\mathfrak{G}}\subset\mathbb{R}^{3}$ with $vol(\mathfrak{G})\sim L^{3}$. Let ${\mathscr{T}}(x)$ be a Gaussian random field $\forall~x\in\mathfrak{G}$ with expectation $\mathbf{E}[{\mathscr{T}}(x)]=0$ and correlation $\mathbf{E}[{\mathscr{T}}(x)\otimes{\mathscr{T}}(y)]=K(x,y;\lambda)$, an isotropic and regulated kernel with correlation length $\lambda$. The field has a Karhunen-Loeve spectral representation ${\mathscr{T}}(x)=\sum_{I=1}^{\infty}\mathrm{Z}^{1/2}_{I}f_{I}(x)\otimes\mathscr{Z}_{I}$, with eigenvalues $\lbrace\mathrm{Z}_{I}\rbrace$, eigenfunctions $\lbrace f_{I}(x)\rbrace $ and Gaussian random variables $\mathscr{Z}_{I}$ with $\mathbf{E}[\mathscr{Z}_{I}]=0$ and $\mathbf{E}[\mathscr{Z}_{I}\otimes\mathscr{Z}_{J}]=\delta_{IJ}$. If $\mathfrak{G}$ contains incompressible fluid of viscosity $\nu$ with velocity $u_{a}(x,t)$ that evolves via the Navier-Stokes equations with a high 'Reynolds function' $\mathsf{RE}(x,t)=\tfrac{\|u_{a}(x,t)\|L}{\nu} $ then aspects of a turbulent flow with $\mathsf{RE}(x,t)\gg \mathsf{RE}_{*}$, a critical Reynolds number, might be represented by the 'weighted' random field $\mathscr{U}_{a}(x,t)= u_{a}(x,t)+\mathrm{A}u_{a}(x,t)\big(\mathsf{RE}(x,t)-\mathsf{RE}_{*}\big)^{\beta}\sum_{I=1}^{\infty} \mathrm{Z}^{1/2}_{I}f_{I}(x)\otimes\mathscr{Z}_{I}$ where random fluctuations and amplitude scale nonlinearly with $\mathsf{RE}(x,t)$, with mean $\mathbf{E}[{\mathscr{U}}_{a}(x,t)] =u_{a}(x,t)$. In the inviscid limit one can prove an anomalous dissipation-type law \begin{align} \lim_{\nu\rightarrow 0}\bigg(\lim_{u_{a}(x,t)\rightarrow {u}_{a}}\sup~\nu \int_{\mathfrak{G}}\int_{0}^{T}{\mathbf{E}}\bigg[\bigg|{\nabla}_{a}{\mathscr{U}}_{a}(x,s)\bigg|^{2}\bigg]d\mathcal{V}(x) ds\bigg)>0 \end{align} iff $\beta=\tfrac{1}{2}$ and $\sum_{I=1}^{\infty}\mathrm{Z}_{I}\int_{{\mathfrak{G}}}{\nabla}_{a}f_{I}(x){\nabla}^{a}f_{I}(x)d\mathcal{V}(x)>0$.