Boundary comparison without global domination of linearized solutions
Establish linear boundary comparison for nonnegative solutions of the linearized fractional p-Laplacian without assuming a global domination condition of the form 0 ≤ F ≤ M_q U, including control of unbounded quotients and the exterior source term arising on the zero set of the background solution.
References
Two further questions concern the present operator. First, removing the global domination 0 ≤F ≤ MqU from linear boundary comparison requires control of unbounded quotients and the exterior source in (6.24). Second, a non-local counterpart of [26, Lemma 3.15] would give (9.9) C−1ε( ρ R)s+ε≤ u(Aρ(w))u(AR(w)) ≤ Cε( ρ R)s−ε, 0 < ρ < R. An iteration of flat comparisons might yield this sharper growth, with suitable control of the exterior data. A power bound of order s+ε is integrable at infinity in the natural (p−1)-tail when ε < s/(p−1). Neither question is settled or used in Theorem 1.1.
An NTA extension would still require the structural and local energy hypotheses of Lemma 7.11, together with normalization and maximum-lift bounds, uniformly over the ordered backgrounds. Under these inputs the quotient and deformation arguments need no approximating hyperplane. General NTA geometry alone has not been shown here to provide them.
A possible extension would impose quantitative conditions on the negative exterior data, for example smallness of Tail(u−, w, 4r)u(A) and Tail(v−, w, 4r)v(A) ,for solutions positive in the comparison patch. Whether such conditions suffice for a version of Theo- rem 1.1 requires additional boundary non-degeneracy and comparison arguments. No such extension is asserted here.