Dirichlet heat kernel estimates for fractional Laplacian with gradient perturbation (Q693715)

The gradient perturbation of the fractional Laplacian is the operator \[ {\mathcal L}^b=-(-\Delta)^{\alpha/2}+b\cdot \nabla, \] where \(b:\mathbb R^d\to \mathbb R^d\). Suppose that \(d\geq 2\), \(\alpha\in (1,2)\), and each component of \(b\) belongs to a certain Kato class. The main results of the paper are sharp two-sided estimates for the heat kernel of \({\mathcal L}^b\) in a bounded \(C^{1,1}\)-open set with zero exterior condition. The short-time estimates coincide with the ones obtained earlier by the same authors in the case of the fractional Laplacian showing that such estimates are stable under gradient perturbation. In addition, sharp Green function estimates and the boundary Harnack principle with the explicit decay rate are shown. The paper is rather technical, but the authors did a good job in the introductory section by explaining the earlier work, the difficulties in the current context, and four main ingredients of the proof, namely (i) the Lévy system for the process \(X^b\) corresponding to \({\mathcal L}^b\); (ii) the approximate stable-scaling of \(X^b\) in bounded \(C^{1,1}\) open set for small time; (iii) the sharp two-sided estimates of \(X^b\) and its dual process in \(C^{1,1}\) open sets with small diameter; (iv) the intrinsic ultracontractivity of (nonsymmetric) \(X^b\). The authors also point out the work of \textit{K. Bogdan} and \textit{T. Jakubowski} [Potential Anal. 36, No. 3, 455--481 (2012; Zbl 1238.35018)] where some of the results of the current paper were independently obtained.