Essential m-dissipativity and hypocoercivity of Langevin dynamics with multiplicative noise (Q2115107)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Essential m-dissipativity and hypocoercivity of Langevin dynamics with multiplicative noise |
scientific article |
Statements
Essential m-dissipativity and hypocoercivity of Langevin dynamics with multiplicative noise (English)
0 references
15 March 2022
0 references
The authors study the properties of the Kolmogorov-type operator \[ L = \mathrm{Tr}(\Sigma D^{2}_{v} ) + b(v) \cdot D_{v} + v \cdot D_{x} - D_{x} \Phi(x) \cdot D_{v} , \] called the Langevin operator. This is associated with the Langevin dynamics \begin{align*} &dX(t) =V(t) dt ,\\ &dV(t)=b(V(t)) dt -\nabla \Phi(X(t)) dt + \sigma(V(t)) dW(t) , \end{align*} in \({\mathbb R}^{d}\). Here \(W(\cdot)\) is a \(d\)-dimensional Wiener process, \(\Phi\) is a suitable potential, \(\sigma\) is a variable diffusion matrix with weakly differentiable coefficients, \(b\) a vector field with components \[ b_{i}(v)=\sum_{j=1}^{d} \partial_{j} a_{ij}(v) -a_{ij}(v) v_j, \] for \(a_{ij} = \Sigma_{ij}\), \(\Sigma = \sigma \sigma^{T}\) and \(D_{v}, D_{x}\), \(D^{2}_{x}\) denote the gradients and the Hessian matrix. The operator \(L\) is studied in the weighted Hilbert space \(H=L^{2}({\mathbb R}^{2 d}, \mu)\), where \(\mu\) is a Gaussian measure on \(({\mathbb R}^{2 d}, \mathcal{B}({\mathbb R}^{2 d})\) weighted by an exponential factor \(e^{-\Phi(x) }\) corresponding to the action of the potential \(\Phi\). The authors prove that the Langevin operator \((L, C_{c}({\mathbb R}^{2d}))\) is closable on \(H\) and its closure \((L, D(L))\) generates a strongly continuous contraction semigroup \((T_t )_{t \ge 0}\) on \(H\). Moreover, the following contractivity result is obtained. For all \(\theta_1 \in (1,\infty)\) there exists a \(\theta_2 \in (0,\infty)\) such that \[ \| T_{t} g - (g,1)_{H} \|_{H} \le \theta_1 e^{-\theta_2 t} \| g -(g,1)_{H}\|_{H}, \quad \forall \, g \in H, \,\, t \ge 0. \] Such estimate provides important information on the asymptotic behavior of Langevin dynamics towards equilibrium and extends previous results by \textit{J. Dolbeault} et al. [C. R., Math., Acad. Sci. Paris 347, No. 9--10, 511--516 (2009; Zbl 1177.35054)] for constant diffusion matrices to the case of variable diffusion matrices with weak derivatives.
0 references
Langevin equation
0 references
multiplicative noise
0 references
hypocoercivity
0 references
dissipativity
0 references
self-adjointness
0 references
Fokker-Planck equation
0 references
0 references
0 references
0 references