Abstract
Cosmological surveys must correct their observations for the reddening of extragalactic objects by Galactic dust. Existing dust maps, however, have been found to have spatial correlations with the large-scale structure of the Universe. Errors in extinction maps can propagate systematic biases into samples of dereddened extragalactic objects and into cosmological measurements such as correlation functions between foreground lenses and background objects and the primordial non-gaussianity parameter $f_{NL}$. Emission-based maps are contaminated by the cosmic infrared background, while maps inferred from stellar-reddenings suffer from imperfect removal of quasars and galaxies from stellar catalogs. Thus, stellar-reddening based maps using catalogs without extragalactic objects offer a promising path to making dust maps with minimal correlations with large-scale structure. We present two high-latitude integrated extinction maps based on stellar reddenings, with a point spread function of full-width half-maximum 6.1' and 15'. We employ a strict selection of catalog objects to filter out galaxies and quasars and measure the spatial correlation of our extinction maps with extragalactic structure. Our galactic extinction maps have reduced spatial correlation with large scale structure relative to most existing stellar-reddening based and emission-based extinction maps.
Introduction
To study the distant universe, astronomers must look through the "fog" of dust in our own Milky Way. This dust absorbs and scatters light, making extragalactic objects appear redder and dimmer than they actually are. To correct for this, we use dust maps. However, existing maps have a "contamination" problem. Emission-based maps (which look at the infrared glow of dust) often accidentally include the Cosmic Infrared Background—the collective light from the very distant galaxies cosmologists are trying to study. Extinction-based maps (which use the reddening of individual stars) often suffer from "intruders": distant quasars or galaxies that look like stars in imaging data.
If a dust map contains signals from distant galaxies, any cosmological analysis using that map will be systematically biased. For next-generation surveys like DESI and LSST, which aim for sub-percent precision, these small correlations between "local" dust and "distant" structure are a major roadblock.
Methodology
The authors built their maps using the Bayestar framework, which performs Bayesian inference on the distance and extinction of individual stars. The core innovation of this work is a rigorous secondary selection process to ensure the "stars" used are actually stars and are located behind the bulk of the Milky Way's dust.
SVM-based Filtering
They trained Support Vector Machines (SVMs) on spectroscopically confirmed objects to create complex color-cut boundaries. These filters use combinations of optical (Pan-STARRS1, SDSS) and mid-infrared (WISE) data to distinguish point-like quasars from true stars. The inclusion of WISE mid-IR data was critical. Quasars have much flatter spectra into the infrared compared to stars; the $z - W1$ color is a powerful "smoking gun" for identifying extragalactic interlopers.
Validation and Constraints
The authors used Gaia EDR3 parallaxes to enforce a distance cut ($d_{50} \sin(b) > 400$ pc), ensuring the stars are far enough above the Galactic plane to have passed through the integrated "sheet" of dust. To validate the maps, they used a cross-correlation technique to see if the dust patterns in their maps aligned with the known positions of galaxies at various redshifts ($z$). A "clean" map should show zero correlation with these distant structures.
Key Results
The primary achievement of this work is the dramatic reduction in extragalactic contamination. In the Southern Galactic Hemisphere test set, the aggressive filtering reduced quasar (QSO) contamination from 15% down to approximately 0.04%.
The new maps (available at 6.1’ and 15’ resolution) show an angular cross-correlation with large-scale structure that is broadly consistent with zero across all redshifts. This decoupling is essential for measuring parameters like primordial non-gaussianity ($f_{NL}$) without systematic bias. While the maps avoid the biases of emission-based maps (like the standard SFD map), they remain consistent with them in low-extinction regions where the CIB contamination is minimal.
Conclusion
By prioritizing sample purity over stellar density, the authors have created a map that is safer for precision cosmology. While these maps have higher statistical noise than previous efforts due to the strict selection, they represent a significant step forward in removing LSS leakage. Future surveys like the Vera C. Rubin Observatory (LSST) will observe stars several magnitudes fainter, allowing for maps that are both high-purity and low-noise, potentially reaching even higher angular resolutions.
Summary updated