Dark Energy Survey Year 6 Results: Cell-based Coadds and Metadetection Weak Lensing Shape Catalogue

TL;DR

The paper introduces the DES Year 6 Metadetection weak-lensing shape catalogue, released over 4422 deg$^2$ with 151,922,791 galaxies detected in riz bands and an effective number density of $n_{ m eff}=8.22$ gal/arcmin$^2$ and shape noise $σ_e=0.29$. It presents a two-stage pipeline combining cell-based coadds and the Metadetection shear estimator to mitigate blending- and detection-related biases, and validates the catalogue with extensive data- and image-simulation-based tests. The analysis finds no detectable multiplicative bias at the $m=(3.4\pm6.1)\times10^{-3}$ level (3$\sigma$), while PSF modeling residuals are present but subdominant; the work demonstrates methodologies essential for upcoming Stage-IV surveys and provides a public release of the DES Y6 Metadetection catalogue. The results mark a significant advancement in weak-lensing systematics control, paving the way for robust cosmology with LSST-era data and offering a framework for cross-checks with alternative shear methods such as BFD.

Abstract

We present the Metadetection weak lensing galaxy shape catalogue from the six-year Dark Energy Survey (DES Y6) imaging data. This dataset is the final release from DES, spanning 4422 deg$^2$ of the southern sky. We describe how the catalogue was constructed, including the two new major processing steps, cell-based image coaddition and shear measurements with Metadetection. The DES Y6 Metadetection weak lensing shape catalogue consists of 151,922,791 galaxies detected over riz bands, with an effective number density of $n_{\rm eff}$ =8.22 galaxies per arcmin$^2$ and shape noise of $σ_e$ = 0.29. We carry out a suite of validation tests on the catalogue, including testing for PSF leakage, testing for the impact of PSF modeling errors, and testing the correlation of the shear measurements with galaxy, PSF, and survey properties. In addition to demonstrating that our catalogue is robust for weak lensing science, we use the DES Y6 image simulation suite (Mau, Becker et al. 2025) to estimate the overall multiplicative shear bias of our shear measurement pipeline. We find no detectable multiplicative bias at the roughly half-percent level, with m = (3.4 $\pm$ 6.1) x $10^{-3}$, at 3$σ$ uncertainty. This is the first time both cell-based coaddition and Metadetection algorithms are applied to observational data, paving the way to the Stage-IV weak lensing surveys.

Figures (22)

• Figure 1: The cell-based coadd and DES coadd tile geometry for a corner of the DES coadd tile. The black dashed line shows the outer boundary of the 10k$\times$10k DES coadd tile. The inner black solid line shows the unique region of the coadd tile, which is defined on the sphere in right ascension and declination. Adjacent coadd tiles overlap the same area, but share adjacent unique region boundaries. The red solid lines show the unique regions for the cell-based coadds. In the interior of the coadd tile, these regions are 100$\times$100 pixels. The red dashed line (shown only for two of the cell-based coadds along the diagonal) shows the outer boundary of each cell-based coadd. Adjacent cell-based coadds with their outer boundaries overlap by 50 pixels. At the edge of the coadd tile, the unique region for the cell-based coadd is extended all the way to the outer boundary of the coadd tile.
• Figure 2: Left: Example false-color image of nine coadd cells (with borders indicated as white dotted lines) made from $gri$ bands. The inner 100 $\times$ 100 pixels are cut out from the whole coadd cell of 200 $\times$ 200 pixels by removing the 50 pixels buffer region around the cell. Middle: The mask image over the same region of the sky, where the grey region is excluded from the detection and further analysis. Right: The mfrac image over the same region of the sky, averaged over $riz$-bands. For each detection object, the average masked fraction is computed from the mfrac coadd image, and any objects with mfrac > 0.1 are excluded from our analysis. The interpolation and masked fraction differs for the bottom left cell relative to the two other cells on the bottom row due to the fact that 90-degree rotations are not applied to the input pixel masks near bright stars. See Sec. \ref{['sec:coadd']} for more details.
• Figure 3: Example image of our combined pixel-level and object-level mask and its associated coadd image near the local galaxy NGC0253. Several features are clearly apparent, including bright star masks (e.g., the mask hole right below the center of the large galaxy) and missing cell-based coadds due to bleeds from bright stars (e.g., rectangular missing regions sticking out of the bright star mask hole).
• Figure 4: Log density of objects excluded by the star-galaxy separation criteria, showing object S/N vs size ratio ($T/T_{\rm PSF}$; left) space and color space ($g-r$ vs $r-i$; right). All other selection criteria have been applied. Objects are chosen from 100 randomly selected patches out of 200 over the footprint. Well-measured stars populate the plot near $T/T_{\rm PSF}=0$, but not exactly on $T/T_{\rm PSF}=0$, due to how the shape and size priors are set for the Gaussian fit.
• Figure 5: Various statistics (number count, a root-mean-square of measured shape, shear response, shear weight) as a function of object S/N and size ratio (galaxy size/PSF size). Objects are binned into a grid of signal-to-noise and size ratio using 20 logarithmic bins with a limit of 10<S/N<1000 and 0.5<galaxy size/PSF size<5.0. The objects whose S/N is larger than 1000 are allocated in the last bin, and the same goes for the objects whose size ratio is larger than 5.0. The measured shapes for each sheared image are averaged to compute the shear response in each bin. This grid of shear response is smoothed by a Gaussian kernel of $\sigma$=2.0 to lower the noise in each bin. The shear weight is then computed from this smoothed response grid using Eqn. \ref{['eqn:shearweight']}.