Large deviation principle for a class of SPDE with locally monotone coefficients (Q2197842): Difference between revisions

The authors investigate stochastic partial differential equations (SPDEs) of the following form \[ dX_t = A(t,X_t)\,dt + B(t,X_t)\,dW_t, \] where \(W_t\) is a cylindrical Wiener process on a separable Hilbert space \(U\) and \(A\), resp.\ \(B\), are coefficients on \([0,T]\times V\) taking values in \(V^*\), resp.\ the Hilbert-Schmidt operators from \(U\to H\). As usual, \(V\hookrightarrow H \simeq H^* \hookrightarrow V^*\) is a Gelfand triplet. Moreover, the coefficients are assumed to have the following four properties \begin{itemize} \item[\((A_1)\)] Hemicountinuity: \(\mathbb R\ni s\mapsto \mbox{}_{V^*}\langle A(t,v_1+sv_2),v \rangle_V\) is continuous \item[\((A_2)\)] Local monotonicity: \(2\mbox{}_{V^*}\langle A(t,v_1)-A(t,v_2),v_1-v_2 \rangle_V + \|B(t,v_1)-B(t,v_2)\|_{HS}^2 \leq -\delta\|v_1-v_2\|_V^\alpha + (K+\rho(v_2))\|v_1-v_2\|_H^2\) for a suitable fixed \(\alpha>0\) and a function \(\rho\) on \(V\) which is bounded by \(C+\|u\|_V^\alpha\cdot \|v\|_H^\beta\) for some \(\beta\geq 0\) \item[\((A_3)\)] (Linear) growth: \(v\mapsto B(t,v)\) is bounded by \(C+\|v\|_H\) and is Lipschitz in \(v\) and \[ \|A(t,v)\|_{V^*}^{\alpha/(\alpha-1)} \leq K(1+\|v\|_V^\alpha)(1+\|v\|_H^\beta) \] \item[\((A_4)\)] Time regularity: \(\|B(t_1,v)-B(t_2,v)\|_{HS}^2\leq L(1+\|v\|_V)|t_1-t_2|^\gamma\) \end{itemize} Under these assumptions, the authors show that the \(\varepsilon\)-dependent solutions of the SPDEs \[ dX_t^\varepsilon = A(t,X_t^\varepsilon)\,dt + \varepsilon B(t,X_t^\varepsilon)\,dW_t,\quad X_0^\varepsilon = x\in H \] satisfy a large deviation principle on \(C([0,T],H)\cap L^\alpha([0,T],V))\) with a good rate function of the form \[ I(z) = \inf\left\{\frac 12\int_0^T \|\phi_s\|_U^2\,ds \mid z=z^\phi\in C([0,T],H)\cap L^\alpha([0,T],V)\; \phi\in L^2([0,T],U)\right\}. \] The proof relies on the weak convergence method and Laplace's principle.