The stochastic reflection problem on an infinite dimensional convex set and BV functions in a Gelfand triple (Q439883): Difference between revisions

The authors consider the Dirichlet form \[ \mathcal{E} ^\rho(u,v)= \frac12 \int_H \langle Du,Dv \rangle \rho(z) \mu(dz), \] where \(\mu\) is a Gaussian measure on a Hilbert space \(H\) and \(\rho\) is a bounded variation (BV) function. The measure \(\mu\) has a covariance operator of the form \(\frac 12 A^{-1}\), where \(A:D(A) \subset H \rightarrow H\) is a linear self-adjoint operator on \(H\) such that \(\langle Ax,x \rangle \geq \delta |x|^2\) for some \(\delta>0\) and \(A^{-1}\) is of trace-type. The notion of BV function introduced in this paper is based on a Gelfand triple \((H_1,H, H_1^*)\), where \(H_1\) is a Hilbert space continuously and densely embedded in \(H\), and extends the definition of BV functions given by \textit{L. Ambrosio, G. Da Prato} and \textit{D. Pallara} [Atti Accad. Naz. Lincei, Cl. Sci. Fis. Mat. Nat., IX. Ser., Rend. Lincei, Mat. Appl. 21, No. 4, 405--414 (2010; Zbl 1206.26014)]. A key result is the characterization of BV functions by an integration by parts formula for \(\rho \cdot \mu\). Using the notion of BV function, the authors establish a Skorohod-type representation for the associated distorted Ornstein-Uhlenbeck process, in the case \(\rho=1_\Gamma\), where \(\Gamma\) is a convex set. If \(\Gamma\) is a regular convex set, then \(1_\Gamma \) is a BV function, and, as a consequence, one obtains the existence and uniqueness of a strong solution for the stochastic differential equation in the Hilbert space \(H\) \[ dX(t)+ (AX(t)+ N_{\Gamma}(X(t)) dt \ni dW(t), \] with initial condition \(X(0)=x\), where \(N_\Gamma(x)\) is the normal cone to \(\Gamma\) at \(x\) and \(W(t)\) is a cylindrical Wiener process in \(H\). These results are extended to the non symmetric case. The last section of the paper is devoted to extend these results to the case \(\Gamma= K_\alpha\), where \(K_\alpha =\{ f\in L^2(0,1)| f\geq -\alpha\}\) for some \(\alpha \geq 0\).