{"id":"arxiv/2603.29072","url":"https://arxiv.org/abs/2603.29072","tags":["cond-mat.stat-mech","cs.LG","math.AT"],"score":0,"title":"How much of persistent homology is topology? A quantitative decomposition for spin model phase transitions","domain":"cond-mat.stat-mech","source":"arxiv","authors":["Matthew Loftus"],"reviews":0,"subject":"cond-mat.stat-mech","summary":"Point-cloud persistent homology (PH) -- computing alpha or Rips complexes on spin-position point clouds -- has been widely applied to detect phase transitions in classical spin models since Donato et al. (2016), with subsequent studies attributing the detection to the topological content of the persistence diagram. We ask a simple question that has not been posed: what fraction of the PH signal is genuinely topological? We introduce f_topo, a quantitative decomposition that separates the density-driven and topological contributions to any PH statistic by comparing real spin configurations…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-03-30","reactions":{}}