Equivalence of weak and viscosity solutions in fractional non-homogeneous problems (Q832510)

The nonlocal operator \[ (-\Delta_p)^s \phi(x)= C(p,s,n)\text{\,pr.v\,} \int_{\mathbb{R}^n} \frac{|\phi(x)- \phi(y)|^{p-2}(\phi(x)- \phi(y))}{|x-y|^{n+sp}}\,\text{cl}^ny, \] \(ps<n\), is a generalization of the \(p\)-Laplace operator. The linear case \(p=2\) is the classical fractional Laplacian \((-\Delta)^s\). The equation \[ (-\Delta_p)^su= F(x,u,D^p_su) \] is studied in a bounded domain \(\Omega\) in \(\mathbb{R}^n\), and here \[ D^p_su(x)= \int_{\mathbb{R}^n} \frac{|u(x)- u(y)|^p}{|x-y|^{n+sp}}\,\text{cl}^n y. \] As usual, the ``boundary values'' are given in the certain domain \(\mathbb{R}^n\setminus\Omega\). Some natural growth and smoothness properties of \(F\) are assumed. The authors study the equivalence between bounded weak supersolutions and bounded viscosity supersolutions. The concept of viscosity supersolutions is rather useless without a comparison principle. The gradient \(D^p_su\) on the right-hand side is not well suited for this principle. Theorem 1.1 states that bounded weak supersolutions are viscosity supersolutions. The opposite implication in Theorem 1.3 requires the comparison principle. Unfortunately, its validity cannot be deduced from the equation. The paper also contains many useful estimates. It ends with a discussion about comparison principles.