Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.
Vera: I'm Vera.
Jocelyn: And I'm Jocelyn, here with Subrahmanyan.
Vera: We're looking at the paper: "Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars".
Jocelyn: Here's the gist of the paper. Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars is a novel method that employs an empirical distribution function (emPDF) to constrain the Milky Way mass profile using a large sample of halo stars from the Dark Energy Spectroscopic Instrument Milky Way Survey (DESI MWS).
Vera: The two tracer populations provide overall consistent measurements of the MW halo mass profile.
Jocelyn: The study employs the empirical distribution function (emPDF) method to constrain the Milky Way mass profile, which does not assume any functional form of the tracer distribution.
Vera: This method is a data-driven dynamical modeling approach that avoids parametric assumptions on the DF.
Title and authors: Vera: We’re diving into "Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars." The title itself tells you that they are using data-driven methods to figure out the mass profile of our Milky Way.
Jocelyn: And they use these specific tracer stars from the Dark Energy Spectroscopic Instrument Milky Way Survey, or DESI MWS. It’s a huge dataset that gives them a lot of information about stars in our local halo.
Subrahmanyan: The authors are Zhou, Han, Wang, Li, and several others. They are deep in this field of dynamical modeling and cosmology because they've spent years trying to infer the mass of our Milky Way using its stellar components.
Vera: What’s the real implication here for us? It’s about moving away from just assuming a simple mathematical shape for the distribution function and letting the data tell us what that shape is.
Jocelyn: They are using an empirical distribution function, or emPDF, which means they aren't forcing a specific functional form onto how these stars are spread in phase space. It’s more flexible than traditional modeling techniques.
Subrahmanyan: That flexibility is what allows them to test whether the data actually supports the dynamical equilibrium assumption that underpins many of these models, because they are fitting the model to the observed data.
Vera: So, instead of saying "the halo must be NFW," they are letting the stars dictate what we find about it through a statistical fitting process against their observations.
Jocelyn: It’s a very different way of approaching this problem than just using fixed theoretical predictions and seeing how close they are.
Subrahmanyan: And when you look at the results, they are trying to get constraints on the virial mass and concentration parameters for both blue horizontal branch stars and RR Lyrae variables. That’s what connects their stellar data back to the large-scale structure of the galaxy.
Vera: So, it’s about getting those numbers for M200c over solar masses and concentration that are directly informed by DESI DR2 data.
The paper's summary: Jocelyn: So, to summarize what they did in this paper, the authors are applying the emPDF method to constrain the Milky Way mass profile using two classes of tracers.
Vera: They use blue horizontal branch stars and RR Lyrae variables as their primary data sources for this dynamical modeling exercise.
Jocelyn: The goal is to derive best-constrained virial mass and concentration parameters from this combined data, which they find are consistent with each other.
Subrahmanyan: The summary says the two tracer populations provide overall consistent measurements of the Milky Way halo mass profile, which is a key finding because it shows that different types of stars can point toward the same underlying structure.
Vera: That consistency between the two tracer classes is important because it lends credibility to the final result, showing that they aren't just getting a lucky measurement from one population.
Jocelyn: They calibrate this method using mock observations from AuriDESI to make sure their approach is sound before applying it to real data.
Subrahmanyan: And the calibration process tests deviations from the steady-state assumption and the effects of observational errors, which sets up the foundation for how they handle real-world noise in their analysis.
Vera: When they test against mock samples, without observational errors, they find that the best constraints show relatively small deviations from what is considered true.
Jocelyn: But when observational error is included, that introduces a significant overall bias, which appears as a steeper slope and a higher mass at larger radii of the mass profile.
Subrahmanyan: So, the paper essentially shows that observational noise doesn't just add random scatter; it systematically shifts our inferred parameters in a predictable way toward larger values.
Vera: That systematic shift is something we have to be careful about when interpreting any new survey results.
The paper's improvements: Jocelyn: The authors propose several improvements to the method, focusing on making the emPDF more robust for real-world application.
Vera: One big one is incorporating a data-driven empirical distribution function into the AI systems to constrain dark matter halo mass profiles using tracer stars from DESI MWS data without assuming any functional form for the tracer distribution.
Jocelyn: This allows the AI systems to derive best-constrained virial mass and concentration of log M200c over solar masses equal twelve point five plus or minus zero point twenty-one minus zero point twelve plus zero point twenty-six minus zero point eighteen for both BHBs and RR Lyrae stars <ref:2611.05520#pg1>.
Subrahmanyan: That's a direct application of the method, which is powerful because it lets the AI learn from the full six dimensional phase space representation of these dynamical tracers.
Vera: They also suggest implementing a survey selection function correction within those AI systems to account for observational biases that depend on apparent magnitude, such as those described in section three point two of the paper <ref:2611.05520#pg1>.
Jocelyn: This allows the AI systems to apply the selection function S equals w to the empirical DF, ensuring that "An observed tracer sample follows the underlying DF modulated by the selection function" is correctly modeled when constructing likelihoods for mass inference.
Subrahmanyan: That’s a necessary step because it corrects for how observational biases affect our understanding of the distribution of stars we actually see, which is critical for accurate mass inference.
Vera: And they also develop a framework for quantifying systematic uncertainties in AI models by using mock observations, as described in section four point one.two of the paper <ref:2611.05520#pg1>.
Jocelyn: This allows the AI systems to estimate a "net systematic bias derived from our mock analysis" and incorporate it into the final parameter estimation, leading to results reported in the form of "x equals xtrue plus delta x plus sigma stat plus total uncertainty".
Subrahmanyan: That framework is essential because it moves beyond just reporting statistical uncertainty and gives us a more honest picture of the error budget, which is what we need when making real cosmological claims.
Conclusion: Vera: So, to wrap up this paper on "Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars," they show that combining their two tracer results with the independent satellite constraint leads to a very tight final measurement.
Jocelyn: The best-fit values are log M200c over solar masses equal to twelve point fourteen plus or minus zero point zero five minus zero point seventeen plus total uncertainty, and log c over solar mass equal to one point one six plus or minus zero point one <ref:2611.05520#pg1>.
Subrahmanyan: These final numbers represent a robust constraint on the Milky Way halo parameters, and they show how the constraints from different sources complement each other very effectively in parameter space.
Vera: The real implication is that this method gives us a way to use abundant DESI data to directly constrain the mass profile, rather than relying solely on simulations for these details.
Jocelyn: It shows that combining dynamical modeling with independent constraints from satellite dynamics is a powerful strategy for getting the tightest possible constraint on our local universe structure.
Subrahmanyan: To finish, I just want to say that this work provides a concrete, data-driven way to get those mass profile parameters using real stellar tracers and sophisticated AI tools.
Vera: It’s a method that is flexible enough to handle the complexities of observational noise while providing strong constraints on our local dark matter halo structure.
Jocelyn: We're ready to move on, but this paper on "Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars" gave us a great look at how to use empirical distribution functions to constrain the mass profile from real data.
Yanrui Zhou, Jiaxin Han, Wenting Wang, Zhaozhou Li, Alexander H. Riley
State Key Laboratory of Dark Matter Physics, School of Physics and Astronomy, Shanghai Jiao Tong University · School of Astronomy and Space Science, Nanjing University · Key Laboratory of Modern Astronomy and Astrophysics, Nanjing University · Centre for Astrophysics and Planetary Science, Racah Institute of Physics, The Hebrew University · Lund Observatory
astro-ph.GA
Submitted: 2026-11-03
Updated: 2026-11-03
Comments: 23 pages, 11 figures. This paper is a DESI Collaboration Key Project publication
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 85/100
The gist: Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars is a novel method that employs an empirical distribution function (emPDF) to constrain the…
Key concepts
- Empirical Distribution Function (emPDF)
- This is a data-driven technique used to model the distribution of tracer stars without assuming a fixed mathematical form. Instead of guessing how stars are distributed in space, the method directly uses the observed positions and velocities of halo stars to infer the underlying mass profile.
- Tracer Populations (BHB and RRL)
- The study uses two distinct groups of halo stars—Blue Horizontal Branch (BHB) stars and RR Lyrae (RRL) variables—as tracers. These different stellar populations provide complementary information about the Milky Way's gravitational potential, helping to build a more robust picture of the halo mass.
- Joint Constraint Modeling
- This involves combining the results from multiple independent sources, such as analyzing different tracer types (BHB and RRL) and incorporating constraints from satellite dynamics. Combining these diverse data sets significantly tightens the final measurements of the Milky Way's mass parameters.
Terminology
Summary
Constraining the Milky Way Mass profile with data driven distribution function using DESI DR2 halo stars is a novel method that employs an empirical distribution function (emPDF) to constrain the Milky Way mass profile using a large sample of halo stars from the Dark Energy Spectroscopic Instrument Milky Way Survey (DESI MWS). p. 1 Cherry Avenue, Tucson, AZ 85721, USA
43IRFU, CEA, Universit´e Paris-Saclay, F-91191 Gif-sur-Yvette, France
44Department of Physics and Astronomy, University of Waterloo, 200 University Ave W, Waterloo, ON N2L 3G1, Canada
45Perimeter Institute for Theoretical Physics, 31 Caroline St. North, Waterloo, ON N2L 2Y5, Canada
46Waterloo Centre for Astrophysics, University of Waterloo, 200 University Ave W, Waterloo, ON N2L 3G1, Canada
47Instituto de Astrof´ısica de Andaluc´ıa (CSIC), Glorieta de la Astronom´ıa, s/n, E-18008 Granada, Spain
48Departament de F´ısica, EEBE, Universitat Polit`ecnica de Catalunya, c/Eduard Maristany 10, 08930 Barcelona, Spain
49Universit´e Clermont-Auvergne, CNRS, LPCA, 63000 Clermont-Ferrand, France
50Department of Physics and Astronomy, Sejong University, 209 Neungdong-ro, Gwangjin-gu, Seoul 05006, Republic of Korea
51CIEMAT, Avenida Complutense 40, E-28040 Madrid, Spain
52Max Planck Institute for Extraterrestrial Physics, Gießenbachstraße 1, 85748 Garching, Germany
ABSTRACT
We employ a novel data-driven dynamical modeling method, the empirical distribution function
(emPDF), to constrain the Milky Way mass profile using a large sample of halo stars from the Dark
Energy Spectroscopic Instrument Milky Way Survey (DESI MWS). The uncertainties quoted outside and within brackets are statistical
and total errors including systematic errors calibrated from AuriDESI, respectively. The two tracer
populations provide overall consistent measurements of the MW halo mass profile.Where the paper says it
Where the paper says it
Methodology and Data Sources
The study employs the empirical distribution function (emPDF) method to constrain the Milky Way mass profile, which does not assume any functional form of the tracer distribution. This method is a data-driven dynamical modeling approach that avoids parametric assumptions on the DF. p. 3 It represents a data-driven dynamical
modeling method that avoids parametric assumptions
on the DF. A related example is the minimum-entropy
action-space approach of L. 2023) is observing a substantial sample of stars
in the MW outer stellar halo, which can serve as valuable dynamical tracers with line-of-sight velocity measurements. Among these, blue horizontal branch (BHB)
stars and RR Lyrae (RRL) variables 53 are particularly useful due to their nature as distance estimators
and their brightness, allowing them to be observed at
53 In addition to BHB and RRL, DESI also has a large sample of
giant stars observed in the outer stellar halo (e. In particular, our halo star samples are primarily composed of the BHB and RRL stars from the DR2
data. DESI MWS BHB sample
The BHB samples used in this study are constructed
following the same procedure as A. The total number of BHB stars observed by DESI DR1 is almost four times that of previous surveys beyond 50 kpc, with a non-negligible fraction detected beyond 100 kpc (A. Bystr¨om et al. Bystr¨om et al. (2024), but based on the DESI DR2 data. 2013, 2018,
2022). DESI MWS RR Lyrae sample
The DESI MWS RRL targets are selected from Gaia
Data Release 2 (DR2; G.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Calibration and Validation
The method is calibrated and validated using a number of realistic mock observations (AuriDESI). The calibration involves testing deviations from the steady-state assumption and the effects of observational errors using these mock observations. p. 1 The method does not assume
any functional form of the tracer distribution. Deviations from the steady-state assumption and
the effects of observational errors are further tested and calibrated using a number of realistic mock
observations (AuriDESI). This figure allows us to evaluate
and compare the results both across different halos observed from the same viewing direction and for the same
halo viewed from different directions. From these plots,
we find that without observational errors, the best constraints (green curves) show relatively smaller deviations
from the truth (black curves) in each panel. Moreover, any residual bias due to the emPDF estimator itself will also be accounted for when we calibrate
the result using the more realistic error-added mocks. In contrast to the results from error-free mocks, the
inclusion of observational error leads to a significant
overall bias, represented as a steeper slope and higher
mass at larger radii of the mass profile (blue curves with
hatched bands in Figure 4) than the truth (black).Where the paper says it
Where the paper says it
Where the paper says it
Constraints and Results
Applying the calibrated method to two classes of tracers, the researchers derive best-constrained virial mass and concentration for BHBs as log M200c/M⊙ = 12.05+0.21−0.12+0.26−0.18 and log c200c = 1. p. 1 Applying the calibrated method to two classes of tracers in DESI, we derive
a best-constrained virial mass and concentration of log M200c/M⊙ = 12. 18
and log c200c =
1.Where the paper says it
For RR Lyrae stars, they derive log M200c/M⊙ = 12.15+0.22−0.14+0.26−0.19 and log c200c = 1. p. 1 Applying the calibrated method to two classes of tracers in DESI, we derive
a best-constrained virial mass and concentration of log M200c/M⊙ = 12. 18
and log c200c =
1.Where the paper says it
Combining the halo star constraints with the independent constraint from satellite dynamics Z.-Z. Li et al. (2020a), they derive a robust and precise final measurement of the MW halo parameters of log M200c/M⊙ = 12. p. 1 Combining the halo
star constraints with the independent constraint from satellite dynamics Z. (2020a, simDF),
we derive a robust and precise final measurement of the MW halo parameters of log M200c/M⊙ =
12.Where the paper says it
Comparison with Other Methods
The joint constraint with satellites significantly tightens the constraints, as the satellite constraint has a more complementary orientation in parameter space. The resulting joint mass profile becomes more tightly constrained particularly at larger radii, where the satellites contribute most strongly. p. 9 Moreover, any residual bias due to the emPDF estimator itself will also be accounted for when we calibrate
the result using the more realistic error-added mocks. In contrast to the results from error-free mocks, the
inclusion of observational error leads to a significant
overall bias, represented as a steeper slope and higher
mass at larger radii of the mass profile (blue curves with
hatched bands in Figure 4) than the truth (black). Applying the calibrated method to two classes of tracers in DESI, we derive
a best-constrained virial mass and concentration of log M200c/M⊙ = 12. 18
and log c200c =
1.Where the paper says it
Where the paper says it
Conclusion
The study demonstrates the power of combining the self data-driven method emPDF, which can fully utilize 6D phase space information of dynamical tracers, with the abundant data of halo stars provided by DESI MWS. The calibrated estimator is then applied to the real DESI observations. p. 5 This
component describes the completeness fraction of spectroscopically observed DESI BHBs and RRLs relative
to the original photometric targets, which is a function
of the number of passes/exposures and Galactic coordinates, with weak dependence on apparent magnitudes
or stellar color (we will discuss this in Section 3. The observational probability is then applied to
the mock BHB and RRL samples by randomly discarding stars at each sky pointing, such that the mock completeness matches that of the real DESI observations. The two tracer
populations provide overall consistent measurements of the MW halo mass profile. Combining the halo
star constraints with the independent constraint from satellite dynamics Z. Deviations from the steady-state assumption and
the effects of observational errors are further tested and calibrated using a number of realistic mock
observations (AuriDESI). Applying the calibrated method to two classes of tracers in DESI, we derive
a best-constrained virial mass and concentration of log M200c/M⊙ = 12. Applying the calibrated method to two classes of tracers in DESI, we derive
a best-constrained virial mass and concentration of log M200c/M⊙ = 12. 18
and log c200c =
1.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Incorporate a data-driven empirical distribution function (emPDF) method into AI systems to constrain dark matter halo mass profiles using tracer stars from DESI MWS data without assuming any functional form for the tracer distribution. This enables AI systems to derive best-constrained virial mass and concentration of log M200c/M⊙ = 12.05+0.21−0.12+0.26−0.18 for BHBs and RR Lyrae stars, as derived in the abstract, by maximizing the posterior probability of a trial potential against the observed data likelihood under a steady-state assumption and spherical symmetry.
-
Implement a survey selection function correction within AI systems to account for observational biases that depend on apparent magnitude, such as those described in Section 3.2 of the paper. This allows AI systems to apply the selection function S(w) to the empirical DF, ensuring that
An observed tracer sample follows the underlying DF modulated by the selection function
is correctly modeled when constructing likelihoods for mass inference. -
Develop a framework for quantifying systematic uncertainties in AI models by using mock observations, as described in Section 4.1.2 of the paper. This allows AI systems to estimate a
net systematic bias derived from our mock analysis
and incorporate it into the final parameter estimation, leading to results reported in the form ofx = xtrue + ∆x + σstat(±σtot)
rather than just statistical uncertainty. -
Integrate joint constraint modeling by combining posterior constraints from multiple tracer populations and independent methods, such as combining BHB and RRL analyses with satellite dynamics constraints. This allows AI systems to derive a
joint posterior that provides one of the tightest yet well-calibrated constraints to date on the MW halo mass,
using a bivariate Gaussian approximation for the joint posterior of log M200c and log c200c. -
Utilize an ensemble approach, comparing results from different dynamical modeling methods like emPDF and oPDF to validate inference robustness. This allows AI systems to confirm that
the inferred mass profiles derived from oPDF are generally consistent with those from emPDF,
supporting the stability of the constraints on the MW mass profile against the choice of dynamical estimator.
Abstract
We employ a novel data-driven dynamical modeling method, the empirical distribution function (emPDF), to constrain the Milky Way mass profile using a large sample of halo stars from the Dark Energy Spectroscopic Instrument Milky Way Survey (DESI MWS). The method does not assume any functional form of the tracer distribution. Deviations from the steady-state assumption and the effects of observational errors are further tested and calibrated using a number of realistic mock observations (AuriDESI). Applying the calibrated method to two classes of tracers in DESI, we derive a best-constrained virial mass and concentration of M 200c/M=12.05-0.12+0.21(-0.18+0.26) and c 200c=1.20-0.16+0.11(-0.24+0.20) for BHBs. For RR Lyrae stars we derive M 200c/M=12.15-0.14+0.22(-0.19+0.26) and c 200c=1.16-0.12+0.09(-0.18+0.16). The uncertainties quoted outside and within brackets are statistical and total errors including systematic errors calibrated from AuriDESI, respectively. The two tracer populations provide overall consistent measurements of the MW halo mass profile. Combining the halo star constraints with the independent constraint from satellite dynamics, we derive a robust and precise final measurement of the MW halo parameters of M 200c/M=12.12-0.05+0.05(-0.04+0.06) and c 200c=1.16-0.03+0.05(-0.07+0.05), with correlation coefficients of ρ=-0.90(-0.54).
Sources
- Exploring the interaction between the MW and LMC with a large sample of blue horizontal branch stars from the DESI survey
- The DESI Experiment Part I: Science,Targeting, and Survey Design
- The DESI Experiment Part II: Instrument Design
- DESI 2024 III: Baryon Acoustic Oscillations from Galaxies and Quasars
- DESI 2024 IV: Baryon Acoustic Oscillations from the Lyman Alpha Forest
- DESI 2024 V: Full-Shape Galaxy Clustering from Galaxies and Quasars
- DESI 2024 II: Sample Definitions, Characteristics, and Two-point Clustering Statistics
- DESI 2024 VI: Cosmological Constraints from the Measurements of Baryon Acoustic Oscillations
- DESI 2024 VII: Cosmological Constraints from the Full-Shape Modeling of Clustering Measurements
- DESI DR2 Results II: Measurements of Baryon Acoustic Oscillations and Cosmological Constraints
- The DESI Experiment, a whitepaper for Snowmass 2013
- SpecDis: Value added distance catalogue for 4 million stars from DESI Year-1 data
- emPDF: Inferring the Milky Way mass with data-driven distribution function in phase space
- The DESI Y1 RR Lyrae catalog I: Empirical modeling of the cyclic variation of spectroscopic properties and a chemodynamical analysis of the outer halo
- The DESI Y1 RR Lyrae catalog II: The metallicity dependency of pulsational properties, the shape of the RR Lyrae instability strip, and metal rich RR Lyrae
Related papers
- The Pristine survey XXVIII. Journey to the Galactic outskirts: Mapping the outer halo red giant stars down to the very metal-poor end
- Origins of Extreme Emission-Line Ratios in z > 3 Galaxies: Insights from the Lumen Model
- The Kinematically Hot, Extremely Metal-Poor C-19 Stellar Stream in DESI DR2
- UV slopes of Starforming Galaxies in Strong Lensing fields at the Epoch of Reionization with JWST
- Evolution of Binaries Under Stochastic Perturbations
- The hydrodynamical response of cold circumgalactic clouds to quasar radiation