Observational constraints on the origin of the elements. X. Combining NLTE and machine learning for chemical diagnostics of 4 million stars in the 4MIDABLE-HR survey

TL;DR

This paper addresses the challenge of deriving accurate stellar parameters and chemical abundances for millions of FGK-type stars by combining non-local thermodynamic equilibrium (NLTE) physics with machine learning. It introduces a four-layer NLTE Payne artificial neural network trained on a grid of $404{,}793$ NLTE spectra over 16 elements, enabling fast, self-consistent fitting of $T_{ m eff}$, $ m olinebreak log g$, [Fe/H], $ olinebreak \\xi_t$, and 17 elemental abundances with typical biases $ ext{<0.13}$ dex and dispersions $ ext{<0.16}$ dex when tested on degraded spectra. The authors validate the approach against 121 Gaia-ESO benchmark stars using TSFitPy and compare the derived abundances to the OMEGA+ Galactic chemical evolution (GCE) models, demonstrating that multi-element abundances can constrain the Galaxy’s formation history. They also integrate the NLTE Payne into the 4MIDABLE-HR survey pipeline (SAPP) for automated, scalable analysis of over $4$ million stars, and release the models and code publicly. Together, these results show that NLTE-informed ML spectroscopy can deliver precise chemical fingerprints at scale, enabling new inferences about Galactic evolution and element production sites.

Abstract

We present 4MOST-HR resolution Non-Local Thermal Equilibrium (NLTE) Payne artificial neural network (ANN), trained on 404,793 new FGK spectra with 16 elements computed in NLTE. This network will be part of the Stellar Abundances and atmospheric Parameters Pipeline (SAPP), which will analyse 4 million stars during the five year long 4MOST consortium 4: MIlky way Disc And BuLgE High-Resolution (4MIDABLE-HR) survey. A fitting algorithm using this ANN is also presented that is able to fully-automatically and self-consistently derive both stellar parameters and elemental abundances. The ANN is validated by fitting 121 observed spectra of low-mass FGKM type stars, including main-sequence dwarf, subgiant and giant stars down to [Fe/H] $\approx -3.4$ degraded to 4MOST-HR resolution, and comparing the derived abundances with the output of the classical radiative transfer code TSFitPy. We are able to recover all 18 elemental abundances with a bias <0.13 and spread <0.16 dex, although the typical values are <0.09 dex for most elements. These abundances are compared to the OMEGA+ Galactic Chemical Evolution model, showcasing for the first time, the expected performance and results obtained from high-resolution spectra of the quality expected to be obtained with 4MOST. The expected Galactic trends are recovered, and we highlight the potential of using many chemical elements to constrain the formation history of the Galaxy.

Figures (8)

• Figure 1: Distribution in a 2D histogram of synthetic spectra used to train the Payne. Top left panel shows the Kiel diagram, while the rest are distributions as a function of [Fe/H]. All abundances are chosen uniformly random in metallicity space, except for A(O)$< 8.87$ and A(C)$< 8.7$. There are less low metallicity giant model atmospheres, resulting in slightly less spectra at low metallicities. There are also no public MARCS models above the black line in the $T_{\rm eff}$-$\log g$ space, resulting in a lack of computed spectra in that regime.
• Figure 2: Kiel diagram of the fitted stellar sample with [Fe/H] in colour with PARSEC evolutionary tracks Bressan2012 in colour.
• Figure 3: Payne fit (red lines) to the HD 140283 and HD 84937 UVES spectra (black dots), degraded to $R \approx 20000$ resolution, in all three 4MOST-HR windows. The subplots are zoom-ins to different regions of the spectra.
• Figure 4: Comparison between the derived abundance from Payne and TSFitPy plotted in absolute A(X) units. The green line is a one-to-one comparison. Each subpanel shows the average difference (bias) and standard deviation (std) when comparing the abundances from the two sources.
• Figure 5: Estimated systematic error for stellar parameters and abundances. The error was estimated by taking spread of the difference between the derived parameter from Payne and literature (for $T_{\rm eff}$ and $\log g$ for benchmark stars from Heiter2015b) or TSFitPy.