Isocapacitary constants associated with $p$-Laplacian on graphs
Bobo Hua, Lili Wang
TL;DR
This work develops a capacity-based framework for the $p$-Laplacian on graphs by introducing discrete isocapacitary constants that quantify how subsets separate in the graph. A discrete coarea formula links $p$-energy to level-set capacities, enabling two-sided, order-matched bounds for the first Dirichlet, Neumann, and Steklov eigenvalues in terms of $\alpha_p^D$, $\overline{\alpha}_p^N$, and $\overline{\alpha}_p^S$, respectively. The results hold for finite and infinite graphs (via exhaustion) and yield a unified geometric characterization that improves upon classical Cheeger-type bounds, with a constructive Steklov theory through graph sequences. An explicit path-graph example demonstrates the sharpness of the bounds, illustrating the effectiveness of the capacity approach in discrete nonlinear spectral problems.
Abstract
In this paper, we introduce isocapacitary constants for the $p$-Laplacian on graphs and apply them to derive estimates for the first eigenvalues of the Dirichlet $p$-Laplacian, the Neumann $p$-Laplacian, and the $p$-Steklov problem.
