Reflected BSDEs and the obstacle problem for semilinear PDEs in divergence form (Q655320)

The author considers the Cauchy problem for semilinear parabolic equations in divergence form with obstacle. For this, he refines first former results [Probab. Theory Relat. Fields 125, No. 3, 393--407 (2003; Zbl 1027.60071)] on the stochastic representation of solutions of the Cauchy problem in divergence form without obstacle, and he discusses elements of the parabolic potential theory for the divergence operator \(L_t\) as well as the one-to-one correspondence between soft Radon measures and time-inhomogeneous additive functionals of the Markov family associated with \(L_t\). This allows the author to introduce the solution of the the Cauchy problem with obstacle as a pair \((u,\mu)\), where \(\mu\) is a positive Radon-measure satisfying a minimality condition. In general, when the obstacle function is irregular, a weak solution of the obstacle problem in the variational sense is not unique, but there is a minimal solution \(u\) if the obstacle is given as a lower barrier. The fact to consider the solution as a couple \((u,\mu)\) allows the author to get the existence, but also the uniqueness of the solution, where the first component \(u\) turns out to coincide with the weak minimal solution in the variational sense. The proof of the existence is done with the help of the stochastic interpretation of the solution through a reflected backward stochastic differential equation (BSDE). The last section of his paper the author devoted to a linear Cauchy problem with measure data, and he shows that, if the measure data is a soft Radon measure and satisfies some integrability assumption, then the unique renormalised solution can be interpreted stochastically through some simple BSDE.