The Dark Energy Survey (DES) Year 6 combined analysis of galaxy clustering and weak lensing (3×2pt) uncovered an unresolved internal inconsistency associated with lens tomographic redshift bin 2 of the MagLim++ galaxy sample: the posterior of the bin's leading redshift-uncertainty mode pushed strongly against its calibration prior, and posterior-predictive internal-consistency tests failed whenever the bin was included. After an extensive but inconclusive program of diagnostics, DES conservatively excluded lens bin 2 from its fiducial cosmology results and concluded that an unknown systematic effect, likely specific to that bin's redshift range (0.40<z<0.550.40 < z < 0.55), is at play. Subsequent companion papers have excluded magnification bias, the redshift-mode parametrization, clustering-redshift calibration components, and covariance modeling as causes, and systematic investigations of the lens sample have found no evidence connecting the anomaly to imaging artifacts or foreground contamination, but its origin — residual photometric-redshift calibration error, a stellar-contamination or selection artifact, an interplay of photometric systematics with the photo-zz-dependent sample selection, or a modeling deficiency — remains unidentified as of August 2026. The problem is to determine and validate the root cause such that bin 2 can be restored to the analysis with validated internal consistency.

Background

The Dark Energy Survey (DES) mapped ~5000 deg2^2 of the southern sky over six years (Y6) to constrain cosmology through weak gravitational lensing (the tiny coherent distortion of background galaxy shapes by foreground mass) and galaxy clustering. Its flagship "3×2pt" analysis combines three two-point correlation functions: cosmic shear (correlations of the shapes of 140 million source galaxies), angular clustering w(θ)w(\theta) of 9 million lens galaxies, and galaxy-galaxy lensing (the cross-correlation of lens positions with source shapes). Galaxies are divided into tomographic redshift bins, and the lens sample, MagLim++, is selected by magnitude and photometric-redshift (photo-zz) cuts; bin 2 spans 0.40<z<0.550.40 < z < 0.55 (mean z≈0.44z \approx 0.44), a range where DES griz filter transitions create strong color-redshift degeneracies.

Because the analysis is only as good as its redshift calibration, the true redshift distribution n(z)n(z) of each bin is estimated with SOMPZ — a Self-Organizing Map framework that combines deep-field multi-band photometry, wide-field data, and an image-simulation (Balrog) catalog — supplemented by WZ, clustering-redshift measurements that use cross-correlations with spectroscopic galaxies to constrain n(z)n(z). Residual n(z)n(z) uncertainty is marginalized over by mode projection: an ensemble of calibration realizations is decomposed into a few orthogonal modes Ui,β(z)U^{i,\beta}(z) per bin, whose amplitudes uβu^\beta become nuisance parameters with priors set by the calibration spread. The reported DES Y6 constraints in flat Λ\LambdaCDM are S8=σ8(Ωm/0.3)0.5=0.789±0.012S_8 = \sigma_8(\Omega_{\rm m}/0.3)^{0.5} = 0.789 \pm 0.012 and Ωm=0.333−0.028+0.023\Omega_{\rm m} = 0.333^{+0.023}_{-0.028}, about 1.8–2.6σ\sigma from combined primary CMB datasets (arXiv:2601.14559).

Before unblinding, DES runs posterior-predictive checks: a statistic ΔPPD\Delta_{\rm PPD} (a probability-of-data-type metric from the posterior predictive distribution) quantifies both goodness of fit and internal consistency between probe subsets, with values below 0.01 flagging failure. At unblinding, the first attempt failed multiple tests: the posterior of the leading lens-bin-2 redshift mode, ul2,1u_{\rm l}^{2,1}, peaked against the upper edge of its prior (the data preferring an n(z)n(z) shifted to higher redshift with its low-zz tail near z≈0.25z \approx 0.25 suppressed), and bin 2 contributed disproportionately to the internal tension (3×2pt ΔPPD=0.006\Delta_{\rm PPD} = 0.006). Diagnostics — widening WZ and mode priors, dropping WZ constraints entirely, replacing the mode basis with a "shift-and-stretch" parametrization, covariance variants, and relaxing the lens-bin-6 magnification prior — were all inconclusive; wide-prior runs remained discrepant at more than 5σ\sigma. DES therefore excluded all data-vector elements involving lens bin 2 from the fiducial analysis and concluded there is "an unknown systematic effect that is likely associated with the lens bin 2", noting that this redshift range has long posed challenges from DES filter transitions, color-redshift degeneracies, and photo-zz-dependent sample selection, and that it "will continue to be a substantial challenge for future photometric galaxy lensing surveys". Including bin 2 shifts cosmological parameters by only 0.4σ\sigma and would add ≲10%\lesssim 10\% constraining power, so the published numbers are not at stake; what is at stake is the diagnosis — restoring trust in one sixth of the lensing-galaxy sample and in the calibration methodology inherited by Stage-IV surveys (Rubin Observatory LSST, Euclid, Roman).

Problem Statement

Identify the root cause of the internal inconsistency that the DES Year 6 3×2pt analysis found in lens redshift bin 2 (the MagLim++ sample bin covering 0.40<z<0.550.40 < z < 0.55), whose leading redshift-uncertainty mode posterior ul2,1u_{\rm l}^{2,1} pushed against its calibration prior and caused posterior-predictive internal-consistency tests (ΔPPD\Delta_{\rm PPD}-type) to fail when the bin was included. The root cause is a specific physical or analysis-level cause — e.g., a residual photometric-redshift calibration error in the reconstructed n(z)n(z), a stellar-contamination or star-galaxy-separation artifact, an interplay of photometric systematics with the photo-zz-dependent MagLim++ selection, or a galaxy-bias/intrinsic-alignment modeling deficiency — supported by quantitative evidence showing that correcting for it restores internal consistency with bin 2 included: the bin's redshift-mode posteriors consistent with their calibration priors, and the same unblinding-stage posterior-predictive tests passing — the ΔPPD\Delta_{\rm PPD} goodness-of-fit checks and the pairwise internal-consistency checks between probe subsets defined in Appendix A of arXiv:2601.14559, excluding the combinations identified there as prior-volume dominated. Magnification bias, the clustering-redshift (WZ) components, the modal n(z)n(z) parametrization, and covariance modeling have already been excluded by the source and its companion papers and are therefore not the root cause.

References

  1. T. M. C. Abbott, M. Adamow, M. Aguena, et al. (DES Collaboration), 'Dark Energy Survey Year 6 Results: Cosmological Constraints from Galaxy Clustering and Weak Lensing', arXiv:2601.14559 (2026), DOI 10.48550/arXiv.2601.14559, https://arxiv.org/abs/2601.14559
  2. N. Weaverdyck, M. Rodríguez-Monroy, J. Elvin-Poole, I. Sevilla-Noarbe, A. Porredon, S. Avila, et al., 'Dark Energy Survey Year 6 Results: MagLim++ Lens Sample Selection and Measurements of Galaxy Clustering', arXiv:2601.14484 (2026), DOI 10.48550/arXiv.2601.14484, https://arxiv.org/abs/2601.14484
  3. G. Giannini et al., 'Dark Energy Survey Year 6 Results: Redshift Calibration of the MagLim++ Lens Sample', arXiv:2509.07964 (2025), DOI 10.48550/arXiv.2509.07964, https://arxiv.org/abs/2509.07964
  4. E. Legnani, J. Elvin-Poole, D. Anbajagane, D. Sanchez Cid, A. Ferté, et al., 'Dark Energy Survey Year 6 Results: Magnification modeling and its impact on galaxy clustering and galaxy-galaxy lensing cosmology', arXiv:2601.14833 (2026), DOI 10.48550/arXiv.2601.14833, https://arxiv.org/abs/2601.14833
  5. G. Giannini, G. Camacho-Ciurana, A. Whyley, J. Prat, J. Blazek, et al., 'Dark Energy Survey Year 6 Results: Galaxy-galaxy lensing', arXiv:2601.15175 (2026), DOI 10.48550/arXiv.2601.15175, https://arxiv.org/abs/2601.15175
  6. W. d'Assignies, G. M. Bernstein, B. Yin, G. Giannini, et al., 'Dark Energy Survey Year 6 Results: Clustering-redshifts and importance sampling of Self-Organised-Maps n(z)n(z) realizations for 3×23\times2pt samples', arXiv:2510.23565 (2025), DOI 10.48550/arXiv.2510.23565, https://arxiv.org/abs/2510.23565
  7. B. Yin, A. Amon, A. Campos, M. A. Troxel, et al., 'Dark Energy Survey Year 6 Results: Redshift Calibration of the Weak Lensing Source Galaxies', arXiv:2510.23566 (2025), DOI 10.48550/arXiv.2510.23566, https://arxiv.org/abs/2510.23566
  8. N. Weaverdyck, D. Schlegel, A. Raichoor, I. Sevilla-Noarbe, 'Taming Additive Systematics via Redshift-Bin-Optimized Star-Galaxy Separation', arXiv:2605.04220 (2026), DOI 10.48550/arXiv.2605.04220, https://arxiv.org/abs/2605.04220
  9. I. Pantos, L. Perivolaropoulos, 'Status of the S8S_8 Tension: A 2026 Review of Probe Discrepancies', arXiv:2602.12238 (2026), DOI 10.48550/arXiv.2602.12238, https://arxiv.org/abs/2602.12238
  10. DES Collaboration, T. M. C. Abbott, M. Aguena, A. Alarcon, et al., 'Dark Energy Survey Year 6 Results: Cosmological Constraints from Cosmic Shear', arXiv:2602.10065 (2026), DOI 10.48550/arXiv.2602.10065, https://arxiv.org/abs/2602.10065