Dwarf galaxy halo masses from spectroscopic and photometric lensing in DESI and DES

TL;DR

This work implements two parallel weak-lensing paths to measure dwarf-galaxy halo masses: spectroscopic lenses from DESI and photometrically calibrated DES lenses. By fitting NFW profiles to the inner lensing signal, it yields precise halo masses and SHMR constraints across six stellar-mass bins, with cross-validation between spectroscopic and photometric samples confirming robustness of the one-halo measurements. The photometric approach, leveraging a large SOM-calibrated DESI training set, dramatically increases the dwarf-galaxy census and achieves a combined $S/N\approx38$ for $\log_{10}(M_*/M_\odot)<9.25$, enabling SHMR constraints down to $\log_{10}(M_*/M_\odot)\sim8$. The results establish a robust methodology for dwarf lensing and chart a course for transformative dark matter and galaxy-formation constraints with upcoming surveys like LSST, Euclid, and Roman.

Abstract

We present the most precise and lowest-mass weak lensing measurements of dwarf galaxies to date, enabled by spectroscopic lenses from the Dark Energy Spectroscopic Instrument (DESI) and photometric lenses from the Dark Energy Survey (DES) calibrated with DESI redshifts. Using DESI spectroscopy from the first data release, we construct clean samples of galaxies with median stellar masses $\log_{10}(M_*/M_{\odot})=8.3-10.1$ and measure their weak lensing signals with sources from DES, KiDS, and SDSS, achieving detections with $S/N$ up to 14 for dwarf galaxies ($\log_{10}(M_*/M_{\odot})<$9.25) -- opening up a new regime for lensing measurements of low-mass systems. Leveraging DES photometry calibrated with DESI, we extend to a photometric dwarf sample of over 700,000 galaxies, enabling robust lensing detections of dwarf galaxies with combined $S/N=38$ and a significant measurement down to $\log_{10}(M_*/M_{\odot})=8.0$. We show that the one-halo regime (scales $\lesssim 0.15h^{-1}\rm Mpc$) is insensitive to various systematic and sample selection effects, providing robust halo mass estimates, while the signal in the two-halo regime depends on galaxy color and environment. These results demonstrate that DESI already enables precise dwarf lensing measurements, and that calibrated photometric samples extend this capability. Together, they pave the way for novel constraints on dwarf galaxy formation and dark matter physics with upcoming surveys like the Vera C. Rubin Observatory's LSST.

Figures (15)

• Figure 1: Schematic summarizing the data used for the two analyses in this work. We measure galaxy--galaxy lensing using two samples of foreground lenses, each divided into six stellar mass subsamples, three of which target dwarf galaxies with $\log_{10}(M_*/M_{\odot})$$<$ 9.25, indicated as teal, green, and yellow circles in the figure, and with postage stamps of randomly-selected galaxies from each. The two samples are: (1) A spectroscopic sample from DESI (upper panel) with precise stellar masses for each galaxy (with a systematic error of $\sigma(M_*)=0.3$, Section \ref{['sec:desi-data']}) allowing for well-defined bin edges (right). The stacked spectrum from the lowest-mass subsample is shown. In the spectroscopic sample, we have 30,465 dwarf galaxies overlapping the DES footprint with a corresponding density of $\sim$50 galaxies/deg$^2$. This DESI sample is limited to the area that overlaps background lensing data, using sources from DES, KiDS, and SDSS (see Section \ref{['sec:sources']}). (2) A DES photometric sample, which we select and characterize using DESI (Section \ref{['sec:des-lens-sample']}) to achieve a $\sim\times$3 higher number density for dwarf galaxies and a substantially larger dwarf sample of 701,123 galaxies, albeit with wider stellar mass distributions ($\sigma(M_*)=0.45$). This number represents the spread in stellar mass for the SOM cells that contribute to the dwarf galaxy bins (Section \ref{['sec:sample-selection']}). For this photometric analysis, we use DES background sources.
• Figure 2: Stellar mass distributions for SAGA's background galaxy sample, as used in Thornton2024 (teal shaded region), and a substantially larger DESI sample used in this work. We use the DESI galaxies both as a spectroscopic lens sample (solid red lines, full sample) and with selection criteria imposed (Equation \ref{['eqn:selection']}, shaded red region) to choose a preferentially low-mass sample, which we use to calibrate a larger photometric sample of galaxies from DES. We divide the DESI-calibrated DES galaxies into six stellar mass bins (shown in Figure \ref{['fig:mass_bins']}), and we indicate here their median stellar masses at the top.
• Figure 3: Left: Weak lensing halo mass profiles for six stellar mass bins of DESI-fiducial lenses. We detect a signal in all bins: for the dwarf galaxies ($\log_{10}(M_*/M_{\odot})$ of $<$8.55, 8.55$-$8.9, 8.9$-$9.25), we find a $S/N$ of 5, 8, and 10, respectively, and a combined $S/N$ of 14. The curves represent the best-fit NFW profiles, with the shading showing the one-sigma uncertainty. The solid portion includes the six data points used in the fit. We also note the number of DESI lenses per bin. Table \ref{['tab:modeling']} includes the resulting halo mass constraints along with the median stellar masses and full profile signal-to-noise ratios. Right: Redshift distributions for the six stellar mass bins.
• Figure 4: Left: Distributions of mean stellar mass and redshift for DESI calibration sample galaxies within the SOM that is used for calibrating DES lenses. The blue squares in the redshift panel highlight SOM cells dropped because of $\gtrsim0.05$ dex impacts on the mean stellar masses from the $z>0.5$ galaxies (see Section \ref{['sec:stellar-mass-appendix']}). This cut removes $<3$% of the DES galaxies. Although the SOM is not directly trained to separate galaxies of different stellar masses, there is significant ($>2$ dex) variation of mean stellar mass across the cells. Right: $\rm \sigma_{log_{10}(M*/M_{\odot})}$ vs. $\langle$$\log_{10}(M_*/M_{\odot})$$\rangle$ for SOM cells in this work (pink) and in Thornton2024 (teal). The mean standard deviation has dropped from 0.58 to 0.39, as indicated by the horizontal lines. The thick lines mark the median standard deviation in bins that include 50 SOM cells. Compared to Thornton2024, we can extract more stellar mass bins from the SOM because of the increased number of cells and the decreased $\rm \sigma_{log_{10}(M*/M_{\odot})}$.
• Figure 5: Halo mass profiles for six stellar mass bins of DES photometric lens samples and corresponding NFW fits. We include the signal for an additional mass bin, pushing to even lower masses at $\log_{10}(M_*/M_{\odot})$$<8$. The stellar mass cutoffs in the legend refer to the mean stellar mass of the SOM cells that enter the bin, defined using the DESI calibrators within that cell. We also note the number of DES lenses per bin. The corresponding calibrated stellar mass distributions are shown in the right panel. The points and error bars in the stellar mass distribution panel represent the $16^{\rm th}$, $50^{\rm th}$, and $84^{\rm th}$ percentiles. Despite the overlap between the stellar mass distributions, the amplitude of the excess surface density measurements is generally increasing with stellar mass.