English

Unique-neighbor Expanders with Better Expansion for Polynomial-sized Sets

Combinatorics 2024-10-22 v1 Discrete Mathematics

Abstract

A (d1,d2)(d_1,d_2)-biregular bipartite graph G=(LR,E)G=(L\cup R,E) is called left-(m,δ)(m,\delta) unique-neighbor expander iff each subset SS of the left vertices with Sm|S|\leq m has at least δd1S\delta d_1|S| unique-neighbors, where unique-neighbors mean vertices with exactly one neighbor in SS. We can also define right/two-sided expanders similarly. In this paper, we give the following three strongly explicit constructions of unique-neighbor expanders with better unique-neighbor expansion for polynomial-sized sets, while sufficient expansion for linear-sized sets is also preserved: (1) Two-sided (n1/3ϵ,1ϵ)(n^{1/3-\epsilon},1-\epsilon) lossless expanders for arbitrary ϵ>0\epsilon>0 and aspect ratio. (2) Left-(Ω(n),1ϵ)(\Omega(n),1-\epsilon) lossless expanders with right-(n1/3ϵ,δ)(n^{1/3-\epsilon},\delta) expansion for some δ>0\delta>0. (3) Two-sided-(Ω(n),δ)(\Omega(n),\delta) unique-neighbor expanders with two-sided-(nΩ(1),1/2ϵ)(n^{\Omega(1)},1/2-\epsilon) expansion. The second construction exhibits the first explicit family of one-sided lossless expanders with unique-neighbor expansion for polynomial-sized sets from the other side and constant aspect ratio. The third construction gives two-sided unique-neighbor expanders with additional (1/2ϵ)(1/2-\epsilon) unique-neighbor expansion for two-sided polynomial-sized sets, which approaches the 1/21/2 requirement in Lin and Hsieh (arXiv:2203.03581). Our techniques involve tripartite product recently introduced by Hsieh et al (STOC 2024), combined with a generalized existence argument of biregular graph with optimal two-sided unique-neighbor expansion for almost all degrees. We also use a new reduction from large girth/bicycle-freeness to vertex expansion, which might be of independent interest.

Keywords

Cite

@article{arxiv.2410.07061,
  title  = {Unique-neighbor Expanders with Better Expansion for Polynomial-sized Sets},
  author = {Yeyuan Chen},
  journal= {arXiv preprint arXiv:2410.07061},
  year   = {2024}
}

Comments

To appear in SODA 2025

R2 v1 2026-06-28T19:14:44.179Z