English

On the Boundary Behavior of Positive Solutions of Elliptic Differential Equations

Classical Analysis and ODEs 2014-04-30 v2 Analysis of PDEs

Abstract

Let uu be a positive harmonic function in the unit ball B1RnB_1 \subset \mathbb{R}^n and let μ\mu be the boundary measure of uu. Consider a point xB1x\in \partial B_1 and let n(x)n(x) denote the unit normal vector at xx. Let α\alpha be a number in (1,n1](-1,n-1] and A[0,+)A \in [0,+\infty) . We prove that u(x+n(x)t)tαAu(x+n(x)t)t^{\alpha} \to A as t+0t \to +0 if and only if μ(Br(x))rn1rαCαA\frac{\mu({B_r(x)})}{r^{n-1}} r^{\alpha} \to C_\alpha A as r+0r\to+0, where Cα=πn/2Γ(nα+12)Γ(α+12){C_\alpha= \frac{\pi^{n/2}}{\Gamma(\frac{n-\alpha+1}{2})\Gamma(\frac{\alpha+1}{2})}}. For α=0\alpha=0 it follows from the theorems by Rudin and Loomis which claim that a positive harmonic function has a limit along the normal iff the boundary measure has the derivative at the corresponding point of the boundary. For α=n1\alpha=n-1 it concerns about the point mass of μ\mu at xx and it follows from the Beurling minimal principle. For the general case of α(1,n1)\alpha \in (-1,n-1) we prove it with the help of the Wiener Tauberian theorem in a similar way to Rudin's approach. Unfortunately this approach works for a ball or a half-space only but not for a general kind of domain. In dimension 22 one can use conformal mappings and generalise the statement above to sufficiently smooth domains, in dimension n3n\geq 3 we showed that this generalisation is possible for α[0,n1]\alpha\in [0,n-1] due to harmonic measure estimates. The last method leads to an extension of the theorems by Loomis, Ramey and Ullrich on non-tangential limits of harmonic functions to positive solutions of elliptic differential equations with Holder continuous coefficients.

Keywords

Cite

@article{arxiv.1306.3571,
  title  = {On the Boundary Behavior of Positive Solutions of Elliptic Differential Equations},
  author = {A. A. Logunov},
  journal= {arXiv preprint arXiv:1306.3571},
  year   = {2014}
}