Nonlinear stochastic partial differential equations with singular diffusivity and gradient Stratonovich noise (Q738908)

The paper deals with existence and uniqueness of a variational solution in terms of stochastic variational inequalities (SVI) to stochastic nonlinear diffusion equations with a highly singular diffusivity term and multiplicative Stratonovich gradient-type noise. Precisely, the authors study the problem \[ \begin{cases} dX_t\in \text{div} \big[\text{sgn\,}(\nabla X_t)\big]dt+\displaystyle\sum_{i=1}^N \langle b_i,\nabla X_t\rangle\circ d\beta^i_t & \text{in}\;(0,T)\times \mathcal{O},\\ X_0=x & \text{in}\;\mathcal{O},\\ \dfrac{\partial X_t}{\partial\nu}=0 & \text{in}\;(0,T)\times \partial\mathcal{O}, \end{cases} \] where \(\mathcal{O}\) is a convex, bounded and smooth domain in \(\mathbb{R}^d\) with \(d\geq 2,\) and \(\nu\) stands for the outward normal to \(\partial\mathcal{O}.\) The functions \(b_i: \overline{\mathcal{O}}\to\mathbb{R}^d\) are ``coefficient fields'' and \(\beta=(\beta^1,\ldots,\beta^N)\) denotes an \(N\)-dimensional Brownian motion on a filtered (normal) probability space \(\left(\Omega,\mathcal{F}, \{\mathcal{F}_t\}_{t\geq0},\mathbb{P}\right).\) The initial datum is chosen as \(x\in L^2(\Omega,\mathcal{F}_0,\mathbb{P},L^2(\mathcal{O})),\) while the multi-valued graph \(\xi\mapsto\text{sgn\,}(\xi),\) acting from \(\mathbb{R}^d\) into \(2^{\mathbb{R}^d}\) is defined by \[ \text{sgn\,}(\xi)= \begin{cases} \xi/|\xi| & \text{if}\;\xi\neq0,\\ \big\{\zeta\in\mathbb{R}^d:\;|\zeta|\leq 1\big\} & \text{if}\;\xi=0 \end{cases} \] for all \(\xi\in\mathbb{R}^d.\) A similar problem with a drift term given by a singular \(p\)-Laplace type operator \(\text{div} \big[|\nabla X_t|^{p-2}\nabla X_t\big]dt\) is studied as well. The authors derive a commutator relation for the unbounded noise coefficients in terms of a geometric Killing vector condition. Moreover, precise analysis is done on the connection between the nonlinear zero Neumann boundary conditions and the coefficient fields of the noise. This solves an open problem posed in [\textit{V. Barbu} et al., Stochastic Processes Appl. 123, No. 3, 934--951 (2013; Zbl 1267.60076)] and [\textit{V. Barbu} and \textit{M. Röckner}, J. Eur. Math. Soc. (JEMS) 17, No. 7, 1789--1815 (2015; Zbl 1327.60122)].