MagLim++ Galaxy Sample for DES Y6 Cosmology

• The paper introduces MagLim++ as a refined DES Y6 lens sample, extending the Y3 MagLim approach with improved magnitude cuts and redshift calibration.
• It employs catalog-level cleaning, joint masking, and a SOMPZ+WZ redshift framework to enhance systematics control and boost cosmological parameter constraints.
• MagLim++ supports DES Y6 analyses by providing a high-density, tomographically binned lens sample crucial for galaxy clustering and galaxy-galaxy lensing studies.

MagLim++ is the Dark Energy Survey (DES) Year 6 magnitude-limited lens-galaxy sample used as the fiducial position-tracer catalog for galaxy clustering, galaxy-galaxy lensing, and the DES Y6 $3\times2$pt cosmology analysis. It is the Year 6 successor to the DES Year 3 MagLim sample: it preserves the redshift-dependent $i$-band magnitude selection originally optimized for joint clustering and lensing information, while adding catalog-level cleaning, a unified joint footprint, and a new redshift-calibration framework intended to improve purity and systematics control at Stage-III statistical precision (Weaverdyck et al., 20 Jan 2026).

1. Historical origin and design logic

The immediate precursor of MagLim++ is the DES Y3 MagLim lens sample, introduced as an alternative to the red-sequence-based redMaGiC sample for combined galaxy clustering and galaxy-galaxy lensing. The Y3 optimization study defined MagLim through the simple cut

$i < 4\,z_{\rm phot}+18,$

with an additional bright-end cut

$i>17.5,$

using DES Y3 GOLD objects with SOF magnitudes and DNF photometric redshifts. The purpose was to maximize $2\times2$pt constraining power by balancing number density against photometric-redshift precision. Relative to redMaGiC, the optimized Y3 MagLim selection had $\sim 30\%$ wider redshift distributions and $\sim 3.5$ times more galaxies, and forecasted gains included a $40\%$ increase in the figure of merit for parameter pairs involving $\Omega_m$, $w$, and $i$0 in $i$1CDM (Porredon et al., 2020).

MagLim++ retains that same broad philosophy rather than replacing it. The Y6 galaxy-galaxy lensing and clustering papers describe it as a refined version of Y3 MagLim: still a magnitude-limited tracer population with roughly uniform lensing/clustering signal-to-noise with redshift, but augmented with additional contamination-removal cuts, improved masking, and improved redshift calibration. In that sense, the “++” suffix denotes a cleaned and revalidated continuation of the MagLim program rather than a fundamentally different class of lens selection (Giannini et al., 21 Jan 2026).

2. Selection, footprint, and tomographic structure

MagLim++ is drawn from DES Y6 Gold. Its defining magnitude selection is

$i$2

where $i$3 is the DNF photo-$i$4 estimate. The magnification paper further specifies that objects satisfy FLAGS_GOLD = 0, EXT_MASH = 4, and the DES Y6 joint mask over $i$5. The clustering paper describes the broad base selection as

$i$6

together with

$i$7

These descriptions are compatible: the former gives the core operational definition, while the latter records the broader pre-tomographic catalog cuts used in the clustering analysis (Legnani et al., 21 Jan 2026).

A common source of confusion is the role of $i$8. In the Y6 masking analysis this value appears as a depth requirement on the footprint,

$i$9

at the $i < 4\,z_{\rm phot}+18,$0 level using BDF_DEPTH photometry, not as a replacement for the object-level redshift-dependent MagLim++ selection. The depth cut defines where the sample can be supported reliably; the object selection itself remains redshift-dependent (Rodríguez-Monroy et al., 9 Sep 2025).

After the joint mask and quality selections, the usable footprint is $i < 4\,z_{\rm phot}+18,$1, and the MagLim++ sample contains $i < 4\,z_{\rm phot}+18,$2 galaxies, with surface density $i < 4\,z_{\rm phot}+18,$3. It is split into six tomographic bins with DNF $i < 4\,z_{\rm phot}+18,$4 edges $i < 4\,z_{\rm phot}+18,$5. The clustering paper reports corresponding mean redshifts $i < 4\,z_{\rm phot}+18,$6 (Weaverdyck et al., 20 Jan 2026).

Bin	$i < 4\,z_{\rm phot}+18,$7 range	$i < 4\,z_{\rm phot}+18,$8
1	$i < 4\,z_{\rm phot}+18,$9	1,852,541
2	$i>17.5,$0	1,335,298
3	$i>17.5,$1	1,413,743
4	$i>17.5,$2	1,783,837
5	$i>17.5,$3	1,391,524
6	$i>17.5,$4	1,409,284

A notable feature of the Y6 implementation is that lens bin 2 is excluded from the fiducial cosmological analysis. The Y6 papers state that this bin showed internal tension in the $i>17.5,$5pt analysis and a problematic posterior for its leading redshift nuisance mode, with no clear cause identified (Giannini et al., 21 Jan 2026).

3. Redshift calibration and representation of $i>17.5,$6

For MagLim++, the lens redshift distribution is calibrated with a SOMPZ $i>17.5,$7 WZ framework. SOMPZ uses a self-organizing-map transfer from deep multi-band photometry to the wide DES footprint, mediated by Balrog synthetic source injections; WZ adds clustering-redshift information through cross-correlations with spectroscopic reference samples, and the final combination is performed through importance sampling (Giannini et al., 9 Sep 2025).

The Y6 implementation introduces several upgrades relative to DES Y3. The SOM uses a noise-weighted metric in flux space, the Balrog catalog spans the full survey footprint and is about $i>17.5,$8 times larger than in Y3, and the uncertainty propagation includes sample variance, shot noise, photometric zero-point uncertainty, and redshift-sample uncertainty. The calibration generates $i>17.5,$9 redshift realizations per tomographic bin and achieves typical uncertainties on the mean redshift at the $2\times2$0–$2\times2$1 level, corresponding to a $2\times2$2–$2\times2$3 average reduction relative to DES Y3 (Giannini et al., 9 Sep 2025).

For cosmological inference, the MagLim++ lens $2\times2$4 is not treated as fixed. In the Y6 galaxy-galaxy lensing analysis it is represented by a mode-projection model,

$2\times2$5

with three modes per lens bin. The redshift-calibration paper arrives at the same effective compression for the SOMPZ+WZ case: the high-dimensional realization ensemble is reduced to three orthogonal modes per bin for use in inference, with only minor degradation in cosmological constraints after marginalization (Giannini et al., 21 Jan 2026).

This mode-based treatment marks a methodological shift from the DES Y3 practice of parameterizing lens-redshift uncertainty primarily through shifts and stretches. A plausible implication is that MagLim++ was designed not only as a cleaner lens catalog, but also as a lens catalog whose residual redshift uncertainty could be propagated in a more flexible basis.

4. Role in DES Y6 clustering, galaxy-galaxy lensing, and $2\times2$6pt inference

MagLim++ is the Y6 lens catalog for galaxy clustering, galaxy-galaxy lensing, and the full $2\times2$7pt DES cosmology program. In this role it provides the foreground overdensity field $2\times2$8, while source shapes come from Metadetection. On the retained scales of the fiducial analyses, the lens sample is modeled with linear, scale-independent galaxy bias, and the corresponding $2\times2$9 enters both clustering and galaxy-shear kernels (Giannini et al., 21 Jan 2026).

The clustering measurement is the angular two-point correlation function $\sim 30\%$0, measured with the Landy–Szalay estimator in 30 logarithmic bins from $\sim 30\%$1 to $\sim 30\%$2 arcmin, with only $\sim 30\%$3 entering cosmological inference. The Y6 clustering paper reports total signal-to-noise $\sim 30\%$4, reduced to $\sim 30\%$5 when restricted to linear scales. From clustering alone, the analysis finds

$\sim 30\%$6

and

$\sim 30\%$7

The same paper emphasizes that $\sim 30\%$8 alone does not separate galaxy bias from $\sim 30\%$9, so the $\sim 3.5$0 combinations are the direct clustering observables (Weaverdyck et al., 20 Jan 2026).

The Y6 galaxy-galaxy lensing measurement uses MagLim++ lenses and a Metadetection source sample of $\sim 3.5$1 million galaxies in four source bins. The full GGL signal reaches $\sim 3.5$2, a $\sim 3.5$3 improvement over DES Y3. After applying the cosmological scale cuts and excluding lens bin 2, the retained signal-to-noise becomes $\sim 3.5$4 for the linear-bias model and $\sim 3.5$5 for the nonlinear-bias model. Including point-mass marginalization reduces the effective values further to $\sim 3.5$6 and $\sim 3.5$7, respectively. The retained physical cuts are $\sim 3.5$8 for the linear-bias model and $\sim 3.5$9 for the nonlinear-bias model, chosen so that biases in $40\%$0 remain below $40\%$1 in validation studies (Giannini et al., 21 Jan 2026).

MagLim++ also supports shear-ratio measurements on small angular scales. Although those measurements are not used in the main cosmological analysis, they provide an internal consistency check on the source and lens redshift distributions and on shear calibration (Giannini et al., 21 Jan 2026).

5. Systematics control, masking, and magnification

The defining methodological feature of MagLim++ is aggressive systematics control at the catalog, footprint, and likelihood levels. Two catalog-level cuts are especially central in Y6. The first is a redshift-bin-optimized star-galaxy separation using unWISE near-infrared colors $40\%$2, applied in addition to morphology-based classification; the clustering paper states that this cut removes $40\%$3 of objects. The second is a self-organizing-map cut in $40\%$4 color space that removes compact color regions associated with poor photo-$40\%$5 behavior and strong QSO contamination; this removes $40\%$6 of objects. Together the SOM and star-galaxy cuts remove about $40\%$7 of the initial MagLim selection (Weaverdyck et al., 20 Jan 2026).

These Y6 choices were preceded by a Y3 study showing that residual stellar contamination in the MagLim lens sample varied strongly with tomographic bin, $40\%$8, with $40\%$9 overall contamination. That work argued for redshift-bin-optimized NIR separation precisely because morphology-only cuts leave additive contamination that is difficult to remove later by map-level weighting (Weaverdyck et al., 5 May 2026).

At the footprint level, DES Y6 adopted a unified joint mask built around MagLim++. The area evolves from a seed footprint of $\Omega_m$0 to $\Omega_m$1 after the baseline mask and to $\Omega_m$2 after the systematics mask, yielding a final joint footprint of $\Omega_m$3. The masking strategy combines outlier cuts on survey-property templates, hard cuts on cirrus indicators, and a leverage mask in the 19-dimensional template space. The masking paper argues that aggressive masking makes the Iterative Systematics Decontamination method more reliable, particularly when upgraded from linear to cubic fits to capture non-linear contamination (Rodríguez-Monroy et al., 9 Sep 2025).

Magnification is another first-order systematic for MagLim++ because the sample is defined by a magnitude-dependent selection rather than a red-sequence cut. The Y6 magnification analysis models the observed overdensity as

$\Omega_m$4

with Balrog simulations used to estimate the selection response. The fiducial magnification coefficients for the six bins are reported as

$\Omega_m$5

Neglecting magnification in simulated Y6 analyses produces shifts of $\Omega_m$6 in $\Omega_m$7 and $\Omega_m$8 in $\Omega_m$9; with cosmic shear included, the shifts remain $w$0 in $w$1 and $w$2 in $w$3. The same study reports that freeing the magnification bias in lens bin 2 leads to unphysical negative values, reinforcing the decision to exclude that bin, while also arguing that the underlying problem is likely not magnification alone but some other unmodeled systematic for which bin 2 acts as a proxy (Legnani et al., 21 Jan 2026).

Operationally, the Y6 GGL pipeline also uses random-point subtraction, boost-factor corrections for lens-source clustering, and large-scale-structure weights. The random catalogs are 50 times larger than the corresponding lens catalogs (Giannini et al., 21 Jan 2026).

6. Scientific lineage and broader use

Although MagLim++ is a Y6 construct, it belongs to a broader MagLim program that became central to DES cosmology in Y3. In the original Y3 $w$4pt analysis using the MagLim lens sample, the reported flat $w$5CDM constraints were

$w$6

and in flat $w$7CDM

$w$8

That analysis also found that extending to smaller scales with a non-linear galaxy-bias model improved constraining power by $w$9 in the $i$00–$i$01 plane and $i$02 in the $i$03–$i$04 plane (Porredon et al., 2021).

The same sample lineage supported several specialized analyses. A harmonic-space DES Y3 $i$05pt treatment using MagLim reported, for its fiducial four-bin extended-cut case,

$i$06

consistent with the configuration-space DES Y3 result (Faga et al., 2024). A small-scale galaxy-galaxy lensing HOD analysis found typical MagLim host halo masses $i$07–$i$08, satellite fractions $i$09–$i$10, and linear galaxy bias rising from $i$11 at low redshift to $i$12 at high redshift (Zacharegkas et al., 2021). In a Planck-lensing $i$13 analysis, the DES Y3 MagLim sample served as the photometric large-scale-structure tracer, yielding four-bin measurements

$i$14

consistent with $i$15CDM (Li et al., 6 Jan 2025).

The Y3 experience also established the limits of the sample. In the DES Y3 $i$16 SPT/Planck CMB-lensing analysis, the baseline four-bin MagLim sample was robust, but the two highest-redshift MagLim bins were retained only for diagnostics because clustering and galaxy-galaxy lensing preferred inconsistent galaxy-bias values there (Abbott et al., 2022). That history is directly relevant to MagLim++: the Y6 sample inherits the statistical advantages of a dense magnitude-limited lens catalog, but its architecture is explicitly shaped by lessons about contamination, redshift tails, magnification, and problematic tomographic bins that emerged in the Y3 program.

In this sense, MagLim++ is best understood as DES’s mature magnitude-limited lens sample: a high-density photometric tracer optimized not only for raw signal-to-noise, but for the more restrictive requirement that the full clustering-plus-lensing likelihood remain calibratable, maskable, and robust under the systematics budgets demanded by precision multi-probe cosmology.