Determine the cause of the DES Y6 lens bin 2 anomaly

Identify the root cause of the internal inconsistency observed in the Dark Energy Survey Year 6 three–two–point (3×2pt) analysis for the MagLim++ lens galaxy sample’s redshift bin 2, which manifested as posterior predictive distribution test failures and a posterior for one of its redshift-distribution modes pushing strongly against its prior.

Background

In the final blinding stage of the DES Year 6 3×2pt cosmology analysis, several posterior predictive distribution (PPD) checks failed, specifically involving the MagLim++ lens galaxy sample’s redshift bin 2. The posterior for one of the n(z) mode coefficients in this bin pushed strongly against its prior, indicating an internal inconsistency.

Despite targeted diagnostic tests—altering redshift-mode priors, testing alternative redshift parametrizations, checking covariance modeling, and relaxing magnification priors—the collaboration could not isolate a specific cause. As a conservative measure, all data involving lens bin 2 were excluded from the fiducial analysis, after which all unblinding tests passed.

The issue appears confined to lens bin 2 and likely involves an unknown systematic effect in the relevant redshift range; nevertheless, including bin 2 does not meaningfully alter the final cosmological constraints, underscoring the robustness of the 3×2pt combination. Resolving the cause remains important for future photometric lensing surveys and for confidently using small-scale and higher-redshift information.