🔭 astrophysics

DELVE Milky Way Satellite Galaxy Census I: Satellite Population and Survey Selection Function in DES, DELVE, and Pan-STARRS

This paper presents a comprehensive census of Milky Way satellite galaxies by quantifying detection efficiencies across the DES, DELVE, and Pan-STARRS surveys, ultimately inferring a total population of approximately 265 satellites and characterizing their luminosity function, spatial distribution, and size-luminosity relation.

Original authors: C. Y. Tan, A. Drlica-Wagner, A. B. Pace, W. Cerny, E. O. Nadler, A. Doliva-Dolinsky, D. Anbajagane, T. S. Li, J. D. Simon, A. K. Vivas, A. R. Walker, M. Adamów, K. Bechtol, J. L. Carlin, Q. O. Casey

Published 2026-03-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: C. Y. Tan, A. Drlica-Wagner, A. B. Pace, W. Cerny, E. O. Nadler, A. Doliva-Dolinsky, D. Anbajagane, T. S. Li, J. D. Simon, A. K. Vivas, A. R. Walker, M. Adamów, K. Bechtol, J. L. Carlin, Q. O. Casey, C. Chang, A. Chaturvedi, T. -Y. Cheng, A. Chiti, Y. Choi, D. Crnojević, P. S. Ferguson, R. A. Gruendl, A. P. Ji, G. Limberg, G. E. Medina, B. Mutlu-Pakdil, N. E. D. Noël, K. Overdeck, V. M. Placco, A. H. Riley, D. J. Sand, J. Sharp, N. F. Sherman, G. S. Stringfellow, R. H. Wechsler, M. Aguena, S. Allam, O. Alves, D. Bacon, D. Brooks, D. L. Burke, R. Camilleri, J. A. Carballo-Bello, A. Carnero Rosell, J. Carretero, L. N. da Costa, M. E. da Silva Pereira, T. M. Davis, J. De Vicente, S. Desai, S. Everett, B. Flaugher, J. Frieman, J. García-Bellido, D. Gruen, G. Gutierrez, K. Herner, S. R. Hinton, D. L. Hollowood, D. J. James, K. Kuehn, O. Lahav, S. Lee, J. L. Marshall, C. E. Martínez-Vázquez, P. Massana, J. Mena-Fernández, R. Miquel, J. Muir, J. Myles, R. L. C. Ogando, A. A. Plazas Malagón, A. Porredon, E. Sanchez, D. Sanchez Cid, I. Sevilla-Noarbe, M. Smith, E. Suchyta, M. E. C. Swanson, C. To, E. J. Tollerud, D. L. Tucker, V. Vikram, N. Weaverdyck, M. Yamamoto, A. Zenteno

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Milky Way, our home galaxy, as a massive, glowing city in the dark. For a long time, astronomers thought this city was surrounded by thousands of tiny, dim "neighborhoods" called satellite galaxies. These are small, faint islands of stars orbiting our galaxy, held together by invisible "dark matter" glue.

However, there was a problem. Theories predicted thousands of these neighborhoods, but we could only see a few dozen. It was like looking at a city skyline and seeing only the skyscrapers, missing all the tiny houses and sheds in the suburbs. This was known as the "Missing Satellites Problem."

The question was: Are the theories wrong, or are we just bad at finding the tiny, dim houses?

This paper, written by a massive team of astronomers (the DELVE and DES collaborations), says: "We are just bad at finding them."

Here is a breakdown of what they did, using simple analogies:

1. The Great Cosmic Sweep (The Data)

To find these hidden neighborhoods, the team didn't just look through one telescope. They combined data from three massive sky surveys:

DES (Dark Energy Survey): Like a high-powered flashlight scanning the southern sky.
DELVE: A specialized survey looking for faint objects in the southern sky.
Pan-STARRS: A survey covering the northern sky.

Together, these surveys acted like a giant, high-resolution net cast over about 91% of the sky away from the bright, messy center of our galaxy. They were looking for "overdensities"—clumps of stars that didn't belong to the background noise.

2. The Detective Work (The Search)

Finding these galaxies is hard because they are incredibly faint and look a lot like random stars or dust. To solve this, the team used two different "detective algorithms":

The "Ugali" Detective: This one is very strict. It uses complex math to ask, "Is this clump of stars statistically likely to be a real galaxy, or just a random accident?"
The "Simple" Detective: This one is faster but a bit noisier. It just looks for the densest spots in the star field.

The Trick: They only counted a galaxy as "found" if both detectives agreed. This is like having two independent witnesses confirm a sighting. If one says "I saw a ghost" and the other says "I saw nothing," you don't report a ghost. This rule ensured their list was pure—no fake alarms.

3. The "Fishing Net" Test (Selection Function)

Here is the most clever part of the paper. The team knew their "net" had holes. If a galaxy was too small, too far away, or too dim, their net would miss it. They couldn't just count what they found and say, "That's all there is."

So, they ran a simulation:

They took their real sky data and injected 100,000 fake, computer-generated galaxies into it.
They ran their detective algorithms on this fake data.
They checked: "Did our net catch the fake galaxy? If not, why? Was it too dim? Too far?"

This created a "Catchability Map." It told them exactly how good their net was at catching different types of galaxies. For example, they learned: "We catch 90% of galaxies this size, but only 10% of galaxies that size if they are that far away."

4. The Big Reveal (The Results)

Using their "Catchability Map," they corrected their count.

What they found: They successfully identified 49 known satellite galaxies in their data (out of 62 known ones). They missed a few because they were too faint or in a "blind spot" (like near the bright Magellanic Clouds).
What they predict: By mathematically filling in the holes where their net failed, they estimated the total number of satellite galaxies around the Milky Way.

The Verdict: There are likely 265 satellite galaxies orbiting the Milky Way (with a margin of error of roughly +80 or -50).

This number fits perfectly with the "Missing Satellites" theory. The galaxies weren't missing; they were just hiding in the dark, waiting for a better net.

5. A Little Clumping (Anisotropy)

The team also noticed something interesting about where these galaxies are.

They expected the satellites to be spread out evenly around the Milky Way, like sprinkles on a donut.
Instead, they found a slight clump of satellites near the Large Magellanic Cloud (LMC), a large neighboring galaxy.
The Analogy: It's like finding that most of the stray cats in a city are clustered around a specific food truck. The LMC is the food truck; it brought its own "satellite" cats with it when it arrived at the Milky Way.

Why Does This Matter?

This paper is a huge step forward because it proves that our understanding of the universe (the "Lambda-CDM" model) is correct. The "missing" galaxies were just too hard to see.

For Dark Matter: It confirms that dark matter clumps together in the way we think it does.
For Galaxy Formation: It helps us understand how the first tiny galaxies formed after the Big Bang.
For the Future: It sets the stage for the Vera C. Rubin Observatory, a new telescope that will be even better at finding these tiny, dim galaxies. The Rubin telescope will likely find hundreds more, turning our "neighborhood" census into a full census of the local universe.

In short: The team built a better net, tested it rigorously, and proved that the Milky Way is indeed surrounded by a bustling, crowded family of hundreds of tiny, dark-matter-rich galaxies. We just needed better glasses to see them.

1. Problem Statement

The standard $\Lambda$ CDM cosmological model predicts that the Milky Way (MW) should be surrounded by thousands of dark matter subhalos, many of which host luminous satellite galaxies. However, the observed number of MW satellites is significantly lower than predicted, a discrepancy historically known as the "Missing Satellites Problem."

Recent studies suggest this tension is largely resolved by observational incompleteness: current surveys miss the faintest, smallest, and most distant satellites. To rigorously test $\Lambda$ CDM predictions and constrain dark matter physics (e.g., particle mass, interaction cross-sections) and galaxy formation processes (e.g., reionization), a complete, uniform, and pure census of the MW satellite population is required. Previous efforts were limited by smaller survey footprints, shallower depths, or inconsistent selection criteria.

2. Methodology

2.1 Data Sources and Footprint

The authors constructed a unified stellar catalog using three wide-field optical surveys, covering $\sim 27,700$ deg $^2$ (approx. 91% of the high-Galactic-latitude sky, $|b| \ge 15^\circ$ ):

DES Year 6 (Y6): Dark Energy Survey, southern sky, $g \sim 24.7$ (10 $\sigma$ depth).
DELVE DR3: DECam Local Volume Exploration, southern sky (complementary to DES), $g \sim 24.2$ .
Pan-STARRS1 (PS1) DR1: Northern sky coverage, $g \sim 22.5$ .

Geometric Masking: Regions with high interstellar extinction ( $E(B-V) > 0.2$ ), high stellar density (near the Galactic plane), and areas around known massive systems (LMC, SMC, globular clusters) were masked to reduce false positives.

2.2 Detection Algorithms

Two independent algorithms were employed to identify satellite candidates as overdensities of resolved stars:

ugali: A likelihood-based approach comparing a background-only model against a model including a satellite galaxy (modeled as a Plummer profile with a specific isochrone). It outputs a Test Statistic (TS).
simple: A stellar-overdensity search that filters stars using isochrone cuts and identifies local density peaks, calculating Poisson significance (SIG).

Detection Criteria: To ensure a pure sample (minimizing false positives), strict thresholds were applied:

DES/DELVE: Candidates must be detected by both algorithms.
- ugali: $\sqrt{TS_{gr}} > 5.0$ (DES) or $> 8.0$ (DELVE) AND $\sqrt{TS_{gi}} > 5.0$ (DES) or $> 7.5$ (DELVE).
- simple: $SIG_{gr} > 5.5$ (DES) or $> 6.5$ (DELVE).
PS1: Adopted thresholds from Drlica-Wagner et al. (2020).
Physical Constraints: Only systems with half-light radii $r_{1/2} > 15$ pc and heliocentric distances $16 < D < 400$ kpc were included to avoid ambiguous compact systems and distant Local Group galaxies.

2.3 Selection Function and Efficiency Modeling

To infer the total population from the observed sample, the authors quantified the detection efficiency ( $P_{det}$ ):

Simulations: $10^5$ mock MW satellites with varying properties (stellar mass, distance, size, metallicity, age) were injected into the survey catalogs.
Recovery: The same detection pipeline was run on the simulated data to determine recovery rates.
Modeling: Two methods were used to parameterize efficiency:
1. Analytic Contours: 50% detectability contours in the $r_{1/2}$ vs. $M_V$ plane as a function of distance.
2. Machine Learning: An XGBoost classifier trained on simulation data to predict detection probability based on $M_V$ , $r_{1/2}$ , distance, and local stellar density.

2.4 Population Inference

An empirical probabilistic model was fitted to the recovered census data, accounting for the selection function. The model assumes:

Luminosity Function: Power law ( $PLF \propto 10^{-\beta M_V}$ ).
Radial Distribution: Spherically symmetric cored-NFW profile.
Size-Luminosity Relation: Linear relation between $\log(r_{1/2})$ and $M_V$ with Gaussian scatter.
Bayesian Inference: Markov Chain Monte Carlo (MCMC) was used to derive posterior distributions for model parameters and the total satellite count.

3. Key Contributions

Largest Uniform Census: This work presents the largest uniformly selected sample of MW dwarf satellites to date, recovering 49 out of 62 known satellites within the survey footprint. If bright satellites ( $M_V < -12.5$ ) outside the footprint are assumed detected, the total count rises to 52.
Rigorous Selection Function: The paper provides the most comprehensive characterization of detection efficiency for MW satellites across a combined footprint of DES, DELVE, and PS1, including public release of the XGBoost classifiers and HEALPix masks.
Pure Sample Definition: By imposing high detection thresholds, the authors established a "pure" sample free of spurious detections, essential for unbiased statistical inference.
Comparative Analysis: The study compares MW satellite properties directly with M31 (Andromeda) using consistent methodologies, allowing for robust host-halo mass scaling tests.

4. Key Results

Total Population Estimate: The empirical model predicts a total of $265^{+79}_{-47}$ satellite galaxies around the Milky Way within the range $-20 \le M_V \le 0$ , $15 \le r_{1/2} \le 3000$ pc, and $10 \le D_{GC} \le 300$ kpc.
Luminosity Function: The inferred luminosity function is steeper than previous estimates from SDSS, predicting a larger number of faint satellites ( $M_V \gtrsim -6$ ).
Size-Luminosity Relation: The data supports a relation where fainter galaxies are generally more compact. The model shows good agreement with observations, though outliers like Antlia II and Crater II (ultra-diffuse systems) exist.
Radial Distribution: The radial distribution of satellites is consistent with a cored-NFW profile, with no significant deviation from isotropy after correcting for selection effects, except for a specific anomaly.
Anisotropy: A mild anisotropy ( $\sim 2.3\sigma$ ) was detected in the declination distribution, concentrated near $\text{Dec} \approx -60^\circ$ . This is attributed to the Large Magellanic Cloud (LMC) satellite system, which brings a cluster of its own satellites (e.g., Carina II, III, Horologium I, Hydrus I, Phoenix II, Pictor II, Reticulum II) into the MW halo. Removing these LMC-associated satellites reduces the anisotropy significance to $< 1.5\sigma$ .
M31 vs. MW: Comparing the MW to M31, the M31-to-MW satellite count ratio is $3.2 \pm 1.2$ , consistent with the mass ratio of their host halos, supporting the linear scaling of subhalo numbers with host mass.

5. Significance

Dark Matter Constraints: By providing a precise, completeness-corrected census, this work offers a critical benchmark for testing alternative dark matter models (e.g., Warm Dark Matter, Self-Interacting Dark Matter) which predict different suppression rates of low-mass halos.
Reionization Physics: The faint end of the luminosity function is sensitive to the epoch of reionization. The results provide tighter constraints on the minimum halo mass required to form stars.
Future Surveys: The derived selection functions and efficiency models serve as a baseline for interpreting data from upcoming surveys like the Vera C. Rubin Observatory (LSST), which is expected to discover hundreds of additional ultra-faint satellites.
Methodological Standard: The paper sets a new standard for satellite searches by combining multiple surveys, using dual-algorithm verification, and releasing robust, public selection functions, facilitating direct comparison between observations and cosmological simulations.