Explain the data–simulation discrepancy for SN+flat and SN+BPZ photo-z cosmology results

Determine the underlying causes of the significant discrepancies between data and simulation in the inferred w-shifts (Δw ≡ w_photo-z − w_spec-z) obtained when using the SN+flat redshift prior and the SN+BPZ Gaussian prior photometric redshift methods on the DES-SN5YR photometrically classified Type Ia supernova sample with host spectroscopic redshifts, including investigation of the noted differences in intrinsic scatter (σ_int), redshift scatter and outlier rates, and the low-redshift (z < 0.3) Hubble residual behavior.

Background

The analysis evaluates several photometric redshift approaches for Type Ia supernova cosmology using the DES-SN5YR photometrically classified sample with host spectroscopic redshifts as a reference. While multiple photo-z variants (e.g., SN+SOMPZ and SN+DNF) yield Δw values consistent with simulations, two cases—SN+flat prior and SN+BPZ Gaussian prior—show significant disagreement between the data-derived Δw and the simulation predictions.

The authors note higher σ_int for these cases and differences in redshift scatter/outlier rates, and that the largest differences are in the Hubble residuals at z < 0.3, but they cannot isolate a specific cause, indicating an unresolved issue in modeling or methodology for these photo-z configurations.

References

However, our data results for SN+flat prior and SN+BPZ prior are significantly discrepant with our simulation results. We are unable to attribute this to any particular cause but note that i) σ_int is higher for these two cases across data and simulations than the others and ii) the redshift scatter and outlier rate are higher in the simulations than the data.

— Evaluating Cosmological Biases using Photometric Redshifts for Type Ia Supernova Cosmology with the Dark Energy Survey Supernova Program  (2407.16744 - Chen et al., 2024) in Section 5.4 (Cosmology Results from Data)