Mean-field backward stochastic differential equations and related partial differential equations (Q734629)

Let \((\Omega,{\mathfrak I},P)\) be complete probability space, \(F=\{{\mathfrak I}_s: 0\leq s\leq T\}\) denote the natural filtration generated by \(d\)-dimensional standard Brownian motion \(\{B_t\}_{t\geq 0}\). Let \((\widetilde\Omega,\widetilde{{\mathfrak I}},\widetilde P)= (\Omega\times \Omega,{\mathfrak I}\otimes{\mathfrak I},P\otimes p)\) be the (non-completed) product of \((\Omega,{\mathfrak I}, P)\) with itself. For any \(\theta\in L^1(\widetilde\Omega,\widetilde{{\mathfrak I}},\widetilde P)\) the variable \(\theta(\cdot,\omega)\in L^1(\Omega,{\mathfrak I},P)\,P(d\omega)\)-a.s. and \[ E'[\theta(\cdot,\omega)]= \int_\Omega\theta(\omega', \omega)\,P(d\omega'). \] In the present paper, two equations are considered: \(\bullet\) mean-field backward stochastic differential equation (MFBSDE) \[ dX^{t,\zeta}_s= E'[b(s,(X^{t,\zeta}_s)', X^{t,\zeta}_s)]\,ds+ E'[\sigma(s, (X^{t,\zeta}_s)', X^{t, \zeta}_s)]\, dB_s,\quad s\in[t, T],\;X^{t,\zeta}_t= \zeta.\tag{1} \] and \(\bullet\) backward stochastic differential equation (BSDE) \[ \begin{gathered} - dYt^{t, \zeta}_s= E'[f(s,(X^{0,x_0}_s)', X^{t,\zeta}_s,(Y^{0,x_0}_s)', Y^{t,\zeta}_s, Z^{t,\zeta}_s)]\,ds- Z^{t,\zeta}_s dB_s,\;s\in[t, T],\\ Y^{t,\zeta}_T= E'[\Phi((X^{0,x_0}_T)', X^{t, \zeta}_T)].\end{gathered}\tag{2} \] The authors prove, that under the assumptions of Lipschitz type and the linear growth condition on coefficients \(b\), \(\sigma\), \(f\), \(\Phi\) the equation (1) has a unique strong solution \(X^{t,\zeta}_s\) and there exists unique solution \((Y^{t,\zeta}_s, Z^{t,\zeta}_s)\) to equation (2). They show too that, if the coefficients are deterministic then the function \[ u(t, x)= Y^{t,x}_s|_{s=t} \] is viscosity solution of partial differential equation (PDE) \[ {\partial\over\partial t}u(t,x)+ Au(t,x)+ E[f(t, X^{0,x_0}_t, x,u(t,X^{0,x_0}_t), u(t,x), Du(t,x), E[\sigma(t, X^{0,x_0}_t, x)])]= 0, \] \[ (t,x)\in [0,T)\times \mathbb{R}^n,\quad u(T, x)= E[\Phi(X^{0,x_0}_t, x)],\qquad x\in\mathbb{R}^n, \] with \[ Au(t,x)= \tfrac12\text{tr}(E[\sigma(t, X^{0,x_0}_t,x)]^T D^2u(t, X))+ Du(t, x) E[b(t, X^{0,x_0}_t, x)]. \]