On nonlocal quasilinear equations and their local limits (Q501636): Difference between revisions

The authors introduce a new class of quasilinear nonlocal operators of the form \[ L[u,Du]=(L_1+L_2)[u,Du], \] where \[ L_1[u,Du](x)=\int_{\mathbb{R}^P} \big(u(x+j_1(Du,z))-u(x)-j_1(Du,z)\cdot Du(x)\big)d\mu_1(z), \] \[ L_2[u,Du](x)=\int_{\mathbb{R}^P} \big(u(x+j_2(Du,z))-u(x)\big)d\mu_1(z), \] \(\mu_i\) are non-negative Lévy measures and \(j_i\) are suitable measurable functions, describing the strength and direction of the diffusion. The operators are allowed to degenerate (\(j_1 = 0\) or \(j_2 = 0\) in some set) and have arbitrary gradient growth, so the \(\infty\)-Laplacian, the \(p\)-Laplacian and strongly degenerate operators are included as particular cases. The main results obtained regard comparison, uniqueness, and existence results for viscosity solutions of linear and fully nonlinear equations of the type \[ F(u,Du,L[u,Du])=f(x)\quad \text{in}\;\mathbb{R}^N \] with \(F\) (degenerate) elliptic and strictly increasing in \(u.\) The authors identify the conditions under which the nonlocal operators converge to local quasilinear operators, and show that the solutions of the corresponding nonlocal equations converge to the solutions of the local limit equations. Finally, a (formal) stochastic representation formula is given for the solutions.