Abstract: The recent identification of Quipu — a 428 Mpc filament of 68 X-ray-luminous galaxy clusters and currently the largest known cosmic structure — relied on a friends-of-friends (FoF) analysis of the CLASSIX cluster catalogue at a fixed redshift-dependent linking length [1]. Whether such supercluster results hold under a different clustering algorithm, or with a sparser input catalogue, has not been systematically tested. We compared six clustering methods (FoF at fixed and redshift-scaled linking length, DBSCAN, HDBSCAN, single-linkage hierarchical clustering, and a Gaussian Mixture Model) applied to 317 X-ray clusters in the MCXC redshift slice z = 0.03–0.07 [2], scoring each against the 185 published superstructure members [1] by binary F1. Robustness was measured by random subsampling, with a second slice (z = 0.04–0.08) testing slice-dependence. At full sample, the fixed-scale single-linkage methods achieved the highest F1 (0.74) but collapsed to 0.00 by 60% subsampling, whereas adaptive-density HDBSCAN held F1 ≥ 0.31 down to 20%. The results indicate a measurable trade-off between full-density accuracy and robustness under input sparsity, within the catalogue and hyperparameters reported here.
Keywords: galaxy clusters, superstructures, clustering algorithms, friends-of-friends, HDBSCAN, X-ray clusters, large-scale structure, computational physics
Cite this paper
Aaryan Kukar. (2026) How robustly can clustering algorithms recover cosmic superstructures from X-ray galaxy cluster catalogues? A method comparison under input sparsity. International Journal of Applied Physics, 11, 209-224

Copyright © 2026 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0


