A Truncation Test of the DESI Evolving Dark Energy Signal: Three Supernova Calibrations Compared

Abstract

The DESI DR2 collaboration reported ∼4σ evidence for evolving dark energy (w0 > −1, wa < 0) from baryon acoustic oscillations (BAO) combined with Type Ia supernova (SN Ia) distances. A persistent concern is whether low-redshift SN calibration systematics drive this signal. We introduce a redshift truncation scan—progressively excising low-z SNe and re-fitting CPL cosmology—and apply it to three SN datasets with independent absolute calibrations: Pantheon+ (SH0ES Cepheid distance ladder), DESY5 (internal BAO-anchored calibration), and DES-Dovekie (white dwarf SEDs + Gaia spectrophotometry). All chains use the full Planck 2018 TTTEEE_lite likelihood. We show that the signal is not dominated by photometric calibration. DES-Dovekie serves as a positive control: its CPL signal, already weakened by the recalibration, collapses under truncation—the signature of a calibration-driven signal. In contrast, Pantheon+ and DESY5 retain a ∼2σ deviation (Δχ2 ≈ 4) after all z ≤ 0.12 SNe are removed that does not weaken with further truncation. Quantitatively, in χ2 space the signal decomposes into photometric calibration (Δχ2 = 7.6, 2.8σ) and low-z non-calibration (Δχ2 = 7.4, 2.7σ), which together with a residual consistent with the CMB+BAO baseline (Δχ2 ≈ 2, 1.4σ) reproduce the DESY5 total (4.0σ) to within a few percent. This truncation-immune residual is either genuine dark energy evolution or a shared SN Ia standardization systematic beyond calibration.