Milky Way Mapper: Galactic Evolution Survey

• Milky Way Mapper is a large-scale stellar spectroscopy program that collects multi-epoch optical and near-IR spectra from over 5 million stars to derive atmospheric, kinematic, and chemical labels.
• The survey combines APOGEE and BOSS observations with automated pipelines like Zeta-Payne and ASPCAP to extract robust stellar parameters and elemental abundances.
• Its comprehensive dataset enables precise mapping of the Milky Way’s chemo-dynamical structure, open-cluster homogeneity, and evolution through advanced chemical cartography.

Milky Way Mapper (MWM) is the SDSS-V stellar-spectroscopy program designed to obtain multi-epoch optical and near-infrared spectra for more than 5 million stars across the entire sky, covering a large range in stellar mass, surface temperature, evolutionary stage, and age (Straumit et al., 2022). Within SDSS-V it serves as the survey’s stellar-spectroscopy engine, using APOGEE and BOSS observations to derive homogeneous stellar labels—atmospheric parameters, radial velocities, and chemical abundances—for studies of Galactic structure, chemical evolution, stellar evolution, and the formation history of the Milky Way (Mészáros et al., 9 Jun 2025).

1. Survey program and scientific scope

The defining feature of MWM is survey-scale stellar spectroscopy. The program aims to collect spectra for more than $5\times 10^6$ stars across the sky, with a major focus on low Galactic latitudes, and to use those data to study stellar masses, ages, evolution, rotation, multiplicity and binaries, Galactic structure traced by young stars, and chemical composition and kinematics (Straumit et al., 2022). About $10\%$ of the spectra are expected to be of hot stars of OBAF spectral types, a scientifically important subset for tracing young Galactic structure, stellar rotation and internal mixing, pulsations and asteroseismology, and binary evolution (Straumit et al., 2022).

MWM is observationally heterogeneous by design. In the APOGEE-based work used in DR19 and related analyses, the survey concentrates on FGKM stars cooler than $8000$ K for robust H-band atmospheric parameters and abundances, while specialized treatment is required for hotter stars (Mészáros et al., 9 Jun 2025). MWM observations in the open-cluster homogeneity study come from the APOGEE-style near-IR spectrographs on the Sloan Foundation Telescope at Apache Point Observatory and the duPont Telescope at Las Campanas Observatory, with abundances derived by ASPCAP, The Cannon, and The Payne (Sinha et al., 2024).

This combination of breadth, repeat spectroscopy, and homogeneous label production places MWM in the Galactic-archaeology regime: the program is not limited to source cataloguing, but is built to recover the chemo-dynamical structure of the Galactic disc from large stellar ensembles rather than from individual tracer populations.

2. Instrumentation and automated analysis

MWM uses two distinct spectroscopic regimes. APOGEE provides medium-resolution near-infrared spectroscopy at $R\sim 22{,}500$, while BOSS provides low-resolution optical spectroscopy at $R\sim 2{,}000$ (Straumit et al., 2022). For hot stars this distinction is decisive: APOGEE near-IR spectra of OBAF stars are dominated by hydrogen Brackett lines and contain few metal lines, whereas BOSS optical spectra are generally more informative for $T_{\rm eff}$, $\log g$, $[M/H]$, and radial velocity, unless extinction is high (Straumit et al., 2022).

For OBAF stars, MWM uses the fully automated Zeta-Payne pipeline, developed specifically because no established survey pipeline existed for that subset (Straumit et al., 2022). Zeta-Payne is a model-driven machine-learning method in which a neural network interpolates synthetic spectra and the stellar labels are obtained by minimizing a $\chi^2$ merit function. A central technical choice is to model the observed spectrum as a synthetic stellar spectrum multiplied by a residual response function expanded in Chebyshev polynomials, rather than relying on manual continuum normalization (Straumit et al., 2022). In simulated tests, convergence reliability is about $96\%$–$10\%$0 for both APOGEE and BOSS, but the average internal uncertainties are much smaller for BOSS; for example, $10\%$1 uncertainties are $10\%$2–$10\%$3 K for BOSS compared with $10\%$4–$10\%$5 K for APOGEE over the tested parameter space (Straumit et al., 2022).

For FGKM stars, the core analysis engine is ASPCAP, which performs pseudo-continuum normalization followed by a FERRE-based $10\%$6 fit to synthetic spectral grids (Mészáros et al., 9 Jun 2025). In DR19, ASPCAP delivers parameters for $10\%$7 stars, including all APOGEE-2 targets plus $10\%$8 newly observed stars from APO observed until 4 July 2023 (Mészáros et al., 9 Jun 2025). The release provides $10\%$9, $8000$0, radial velocities, and 24 abundance entries for 21 elements. The estimated precision is $8000$1–$8000$2 K in $8000$3 for giants and $8000$4–$8000$5 K for dwarfs, $8000$6–$8000$7 dex in $8000$8 for giants, and $8000$9–$R\sim 22{,}500$0 dex for multiple elements including metallicity, $R\sim 22{,}500$1, Mg, and Si (Mészáros et al., 9 Jun 2025). The element-quality assessment is explicit: $R\sim 22{,}500$2 are classified as excellent; Al, S, K, Cr, and Mn as good; Na, Ti, Co, Ce, and Nd as fair; and P, V, and Cu as poor (Mészáros et al., 9 Jun 2025).

3. Chemical cartography and low-dimensional abundance structure

A major MWM result is that Galactic disc abundance space is empirically low-dimensional. Using $R\sim 22{,}500$3 red giant stars from the Milky Way Mapper survey with $R\sim 22{,}500$4, one analysis models 16 elemental abundances

$R\sim 22{,}500$5

with a weighted non-negative matrix factorization, writing the shifted abundance matrix as $R\sim 22{,}500$6 with inverse-variance weights $R\sim 22{,}500$7 (Ness et al., 19 May 2026). The fiducial $R\sim 22{,}500$8 model reconstructs the measured abundances well: $R\sim 22{,}500$9 of stars have $R\sim 2{,}000$0 and $R\sim 2{,}000$1 have $R\sim 2{,}000$2 (Ness et al., 19 May 2026).

The four latent patterns are interpreted as shared nucleosynthetic channels associated with early core-collapse supernovae, late core-collapse supernovae, Type Ia supernovae, and asymptotic giant branch stars (Ness et al., 19 May 2026). The interpretation is deliberately cautious: these are shared statistical modes of variation rather than one-to-one direct yield vectors. Even so, the recovered fractions vary systematically with age, metallicity, position, and orbital properties. Older stars are dominated by the two SN II channels; younger stars show larger contributions from SN Ia and AGB enrichment; and the fractions change rapidly across the $R\sim 2{,}000$3-valley around $R\sim 2{,}000$4 (Ness et al., 19 May 2026). This supports a view in which the high-$R\sim 2{,}000$5 and low-$R\sim 2{,}000$6 sequences are different mixtures of the same underlying basis rather than chemically autonomous systems.

A second, larger analysis re-projects 16 abundances for $R\sim 2{,}000$7 red giant stars into the same four shared enrichment patterns and then groups stars by similarity in pattern fractions (Ness et al., 20 May 2026). The resulting enrichment pathways are continuous and chemo-dynamically ordered. Spatially, Channel 1 is strongest at small radii and large $R\sim 2{,}000$8, Channel 3 peaks near the mid-plane and intermediate radii, and Channel 4 strengthens toward the outer disc (Ness et al., 20 May 2026). Temporally, the paper identifies a transition in enrichment behaviour at approximately $R\sim 2{,}000$9 Gyr, interpreted as the onset of a more chemically mixed regime with increasing contributions from delayed sources (Ness et al., 20 May 2026). The broader implication is that MWM chemistry can be used as a compact generative coordinate system for disc evolution, linking abundance mixtures to radius, height, age, and orbit.

4. Open clusters, chemical tagging, and Galactic chemical-evolution models

MWM data are also used to test the chemical homogeneity of open clusters, a central assumption behind chemical tagging. A study combining SDSS-V Milky Way Mapper and APOGEE DR17 abundances measures intrinsic scatter in up to 20 abundances across 26 Milky Way open clusters (Sinha et al., 2024). At the $T_{\rm eff}$0 level it finds no measurable intrinsic abundance scatter in any element across any of the 26 clusters, with upper limits of $T_{\rm eff}$1 dex for most elements and about $T_{\rm eff}$2 dex for neutron-capture and weak-line species (Sinha et al., 2024). Giant stars in open clusters are found to be $T_{\rm eff}$3 dex more homogeneous than a matched sample of field stars (Sinha et al., 2024). This constrains intra-cloud pollution and turbulent mixing in progenitor molecular clouds, while also indicating that chemical tagging requires very high precision and usually additional information beyond light and iron-peak abundances.

At the Galactic scale, MWM DR19 abundances have been used to fit two-infall Galactic chemical-evolution models with $T_{\rm eff}$4 (Hegedűs et al., 31 May 2025). The modeling employs a golden sample of $T_{\rm eff}$5 stars, rounded in the abstract to $T_{\rm eff}$6, and analyzes 15 species across the Galactic disc (Hegedűs et al., 31 May 2025). The thin and thick discs are separated in the $T_{\rm eff}$7–$T_{\rm eff}$8 plane using

$T_{\rm eff}$9

The best-fit global model yields a primary formation timescale of $\log g$0 Gyr, a second infall ascending timescale of $\log g$1 Gyr, a relaxation timescale of $\log g$2 Gyr, and a second infall peak at $\log g$3 Gyr (Hegedűs et al., 31 May 2025). The inferred infall history is interpreted as evidence for a merger event about $\log g$4 Gyr ago and supports an inside-out formation scenario, with the locus of the low-Mg sequence shifting toward higher metallicity in the inner Galactic disc and the outer disc becoming more strongly dominated by the second infall component (Hegedűs et al., 31 May 2025).

5. Milky Way Mapper in the broader Galactic cartography ecosystem

The phrase “Milky Way Mapper” is not restricted to SDSS-V. Recent work uses closely related language for a wider set of Galactic cartography programs, each optimized for a different tracer population or ISM phase.

Survey or program	Primary tracer	Representative scope
Gaia (Hasan, 2021, Collaboration et al., 2022)	Astrometry, photometry, radial velocities	About $\log g$5 of the Galaxy’s stellar population, or roughly one billion sources, with microarcsecond astrometric precision; DR3 provides well over 33 million stars with full 6D phase space
OGLE Mira map (Iwanek et al., 2022)	Mira variable stars	65,981 Miras used to map the inner Galaxy; $\log g$6 kpc and $\log g$7
Augustus (Speagle et al., 4 Mar 2025)	Photometry + Gaia parallaxes	Distance, extinction, and stellar-parameter estimates for 170 million stars; 125 million high-quality posteriors with statistical distance uncertainties of $\log g$8 for well-constrained stellar types
APOGEE-based 3D dust map (Kh. et al., 2024)	Dust extinction from APOGEE stars	More than 44,000 stars; map extends 10 kpc along both $\log g$9 and $[M/H]$0 and 750 pc in $[M/H]$1, with a catalogue of large molecular clouds
WHAM Southern Sky Survey (Haffner et al., 2010)	Diffuse H$[M/H]$2 emission from the WIM	First all-sky, kinematic survey of diffuse H$[M/H]$3 emission; $[M/H]$4, one-degree beam, $[M/H]$5 spectral resolution
EBHIS (Winkel et al., 2010)	Neutral atomic hydrogen, H I	Unbiased, fully sampled H I survey of the whole northern hemisphere with $[M/H]$6 resolution and $[M/H]$7 channel spacing
MWISP (Yang et al., 9 Dec 2025, Su et al., 2019)	$[M/H]$8CO, $[M/H]$9CO, C$\chi^2$0O	DR1 covers $\chi^2$1 of the northern Galactic plane at $\chi^2$2 resolution and $\chi^2$3 velocity resolution
SDSS-V Local Volume Mapper (Kreckel et al., 2024)	Optical IFU mapping of ionized gas	Planned Milky Way footprint of about $\chi^2$4; Orion demonstration covers $\chi^2$5 with 195,000 spectra at $\chi^2$6 resolution

Within this ecosystem, MWM occupies the stellar-chemistry and radial-velocity layer. Gaia resolves the six-dimensional stellar skeleton of the disc and identifies non-axisymmetric structure such as the Local arm, the outer arm, and the bar (Collaboration et al., 2022). MWM adds precise abundance information and survey-scale stellar labels. The Mira-based map provides an independent three-dimensional reconstruction of the barred bulge and flaring disc (Iwanek et al., 2022), the APOGEE dust map recovers large-scale overdensities and cavities in the obscured plane (Kh. et al., 2024), Augustus extends high-latitude distance and extinction inference to 170 million stars (Speagle et al., 4 Mar 2025), and WHAM, EBHIS, MWISP, and LVM supply complementary maps of the ionized, atomic, and molecular ISM [(Haffner et al., 2010); (Winkel et al., 2010); (Yang et al., 9 Dec 2025); (Kreckel et al., 2024)].

6. Scientific significance, interpretation, and limitations

The central significance of MWM is that it converts stellar spectroscopy into a statistically interpretable map of Galactic evolution. Rather than using a single abundance tracer, recent MWM analyses exploit 16-element abundance vectors, Gaia-based orbital information, and large homogeneous samples to decode shared enrichment patterns and relate them to age, metallicity, radius, height above the plane, and orbital properties (Ness et al., 19 May 2026). This suggests a mode of Galactic inference in which the disc’s abundance structure is treated as generative and low-dimensional, while departures from that structure identify chemically and dynamically distinct populations, including accreted material and metal-poor disc stars (Ness et al., 19 May 2026).

Several limitations are equally well documented. For hot stars, APOGEE spectra are often less informative than BOSS spectra, and $\chi^2$7, $\chi^2$8, and $\chi^2$9 can be too uncertain for meaningful astrophysical interpretation in the near-IR unless extinction is severe (Straumit et al., 2022). In DR19, internally reported Astra abundance uncertainties are systematically smaller than empirical scatters from clusters and binaries, so they function as lower bounds rather than complete error estimates; cool-dwarf surface gravities are problematic below $96\%$0 K, and P, V, and Cu are too noisy for reliable scientific use (Mészáros et al., 9 Jun 2025). In the abundance-decoding framework, the recovered channels are explicitly not pure yield vectors, but shared empirical modes shaped by the input elements, uncertainties, and normalization choice (Ness et al., 19 May 2026). Related Galactic maps carry their own caveats: the 3D APOGEE dust map captures spiral-arm segments rather than a perfect global spiral pattern and is more reliable within about 4 kpc for some cavity identifications (Kh. et al., 2024), while the MWISP cloud catalog is explicitly described as interim and its completeness is biased toward nearby, brighter clouds (Yang et al., 9 Dec 2025).

Taken together, these constraints define Milky Way Mapper as a survey program and, more broadly, as a methodological regime. In the narrow sense, it is the SDSS-V program that produces large, homogeneous stellar spectroscopic catalogs. In the broader current literature, it is part of a coordinated multi-tracer effort to reconstruct the Milky Way in geometry, kinematics, chemistry, dust, and gas. The SDSS-V component is distinctive because it maps the Galaxy through stellar labels: not only where stars are, but how their atmospheric parameters, radial velocities, and multi-element abundances encode the disc’s assembly, enrichment, and mixing history.