Diagnosing satellite-profile systematics in simplified DESI LRG HOD models using public Abacus full-box mocks

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract LRG HOD analyses infer satellite fraction and velocity bias using high-fidelity mocks, but simplified public-box analyses can reveal which observables drive these parameters. Using the public AbacusSummit base simulation at z = 0.8, a compact halo/subhalo-like candidate catalogue, and a DESI DR1 LRG clustering vector over 0.6 < z < 0.8, we fit exact-number-density SHAM-like and minimal-HOD samples to wp + ξ0 + ξ2. The baseline minimal-HOD grid favours a low satellite fraction, fsat = 0.04, with αs = 1.1, while observable splits show that this preference is driven mainly by small-scale wp, not by the full vector uniformly. A phenomenological radial-dilution stress test improves the diagnostic score from $\chi ^2_{\rm heur}=282.6$ to 222.3, but direct radial diagnostics show that q = 4.0 moves 0.723 of satellites beyond an estimated R200m, so it is not a physical profile model. An occupation-control test varying the satellite host-mass power shifts the selected host masses but leaves the median satellite radius nearly unchanged, and an NFW remapping gives only modest improvement. The low-fsat preference should therefore not be interpreted as a physical low satellite fraction without matched occupation freedom, profile freedom, selection and covariance. This is not a replacement for full DESI AbacusHOD inference, but a diagnostic study of satellite-profile systematics.