Stellar Reddening Based Extinction Maps for Cosmological Applications

The Big Picture

Imagine trying to photograph a distant city through a foggy window. You can wipe down your side of the glass, but you can’t control the fog itself. If the fog isn’t evenly distributed, your photos will be systematically skewed in ways that are hard to correct later. Cosmologists face exactly this problem. The Milky Way is full of interstellar dust, and every observation of a distant galaxy or supernova must pass through this dusty “fog” before reaching our telescopes.

To do science, astronomers must subtract the dust’s effect. But that correction is only as good as the maps they use.

The best dust maps we have are subtly contaminated. They’re not just measuring dust; they’re also picking up faint signals from distant galaxies, the very objects cosmologists are trying to study. Use a contaminated dust map to “clean” your galaxy survey, and you’ll inadvertently imprint a fingerprint of that contamination onto your results.

For next-generation surveys like DESI and the Rubin Observatory’s LSST, which aim to measure the universe’s properties at the sub-percent level, even tiny biases like this can ruin the science.

Nayantara Mudur, Core Francisco Park, and Douglas Finkbeiner at Harvard have built two new dust extinction maps designed to break this vicious cycle. Their maps are less entangled with the distant galaxies they’re supposed to be blind to, and the difference is measurable.

Key Insight: The most widely used dust maps inadvertently contain signals from distant galaxies, creating a circular contamination problem. By rigorously filtering stellar catalogs to remove extragalactic sources, Mudur et al. produce cleaner maps that introduce fewer systematic biases into cosmological analyses.

How It Works

The researchers focused on stellar-reddening based maps, which work by measuring how much redder stars appear compared to their known type. Think of it like checking whether a white streetlight looks orange, then using that color shift to infer how much atmosphere you’re looking through. For each star, the reddening reveals how much dust lies between us and that star.

The problem: real stellar catalogs aren’t perfectly clean. Galaxies and quasars sometimes sneak in, masquerading as stars. When a map-maker mistakes a galaxy for a star, the reddening inferred at its position is wrong. And since galaxies cluster together, this contaminates the dust map with exactly the galaxy-clustering signal you want to avoid.

The team’s core strategy was strict, multi-pronged filtering:

• Gaia astrometry cuts: Gaia is the European Space Agency’s star-mapping satellite. Objects with parallax (the tiny apparent position shift as Earth orbits the Sun, essentially zero for distant galaxies) or proper motion (a star’s slow drift across the sky) inconsistent with being a distant star were removed.
• Morphology cuts: Extended sources, which tend to be galaxies, were rejected using Pan-STARRS shape measurements.
• Color-based QSO rejection: Quasars have distinctive optical and near-infrared colors; suspicious combinations triggered exclusion.
• External catalog cross-matching: Known quasar catalogs provided additional veto lists.

With the cleaned stellar sample in hand, photometry from Pan-STARRS1 (a wide-field optical sky survey spanning five wavelength bands from 490–964 nm) and 2MASS (the Two Micron All Sky Survey, covering three near-infrared bands), plus Gaia parallaxes, were fed into the Bayestar stellar inference framework. Bayestar models each star’s distance and reddening simultaneously, producing a per-star estimate of the probable range of reddening and distance values. Integrating these along each line of sight produces a 2D extinction map.

The result: two maps with angular resolutions of 6.1 arcminutes and 15 arcminutes full-width half-maximum. (For scale, the full Moon spans about 30 arcminutes, so 6.1 arcminutes is quite fine.) The maps cover the high galactic latitudes where cosmological surveys operate. The finer map matches the SFD (Schlegel-Finkbeiner-Davis 1998) map in resolution, but with substantially less contamination from the background universe.

To measure success, the team computed angular cross-correlations between their extinction maps and tracers of cosmic structure, specifically quasar samples that probe the universe’s web of galaxies. The question is simple: do the bright spots on the dust map tend to line up with the bright spots on the galaxy map? A well-behaved dust map should show zero such alignment. The new maps outperform most existing stellar-reddening and emission-based maps on this metric, showing significantly reduced spurious correlations across a range of angular scales.

Why It Matters

Two cosmological measurements are especially sensitive to dust map contamination.

The first is magnification bias and lensing cross-correlations. These measurements use the way foreground galaxies bend and amplify light from background objects to map how matter is distributed throughout the universe. Any spurious correlation between the dust map and galaxy positions distorts these signals directly.

The second application is measuring fNL, the primordial non-Gaussianity parameter, a number encoding information about cosmic inflation and the earliest moments of the universe’s existence. Detecting a nonzero fNL is one of the grand goals of modern cosmology and requires picking out an incredibly subtle pattern in how galaxies cluster across the largest scales. Dust contamination that correlates with galaxy positions is precisely the kind of systematic that can mimic or mask this signal.

As Gaia’s astrometric catalog matures and photometric surveys push deeper, stellar catalogs will only improve. More precise parallaxes mean cleaner star-galaxy separation, which means even less contaminated maps.

The two-map design (6.1’ and 15’) is itself deliberate. The coarser map sacrifices resolution for better noise properties in sparse regions, giving analysts a choice of tool depending on the science. The authors also note that combining their maps with neutral hydrogen (HI) maps, which trace dust through a completely different physical channel, could offer a valuable independent cross-check.

Bottom Line: By carefully cleaning stellar catalogs of galaxies and quasars before building dust maps, this work delivers extinction maps with measurably lower spurious correlations with cosmic structure. For upcoming surveys trying to squeeze precision cosmology from billions of galaxy observations, that kind of improvement matters.