A method to derive self-consistent NLTE astrophysical parameters for 4 million high-resolution 4MOST stellar spectra in half a day with invertible neural networks

Abstract

Modern spectroscopic surveys obtain spectra for millions of stars. However, classical spectroscopic methods can often be computationally expensive, rendering them impractical for the analysis of large datasets. We introduce a novel simulation-based deep-learning approach for the efficient analysis of high-resolution stellar spectra to be obtained with the upcoming high-resolution 4MOST spectrograph. We used a suite of synthetic non-local thermodynamic equilibrium (NLTE) spectra generated with Turbospectrum to mimic 4MOST observations and trained a conditional invertible neural network (cINN) for the purpose of predicting self-consistently stellar surface parameters and chemical abundances. The cINN is a neural network architecture that estimates full posterior distributions for the target stellar properties, providing an intrinsic uncertainty estimate. We evaluated the predictive performance of the trained cINN model on both synthetic data and observed spectra of stars. We found that our new cINN trained on NLTE synthetic spectra is capable of recovering stellar parameters with average errors ($\sigma$) of $33$ K for $T_{\mathrm{eff}}$, $0.16$ dex for $\log (g)$, and $0.12$ dex for [Fe/H], $0.1$ dex for [Ca/Fe], $0.11$ for [Mg/Fe], and $0.51$ dex for [Li/Fe], respectively, at a signal to noise ratio of 250 per Angstrom. From the analysis of the observed spectra of Gaia-ESO / 4MOST / PLATO benchmark stars, we verified that our NLTE estimates for stellar parameters and abundances are consistent with results obtained with the independent code TSFitPy. We conclude that the NLTE cINN is robust and can, theoretically, evaluate 4 million high-resolution 4MOST spectra in less than a day, using GPU acceleration.

25 pages, 14 figures, 7 tables, accepted for publication in Astronomy & Astrophysics

References in corpus (64)

• The Gaia mission
• A grid of MARCS model atmospheres for late-type stars I. Methods and general properties
• The Apache Point Observatory Galactic Evolution Experiment (APOGEE)
• Normalizing Flows: An Introduction and Review of Current Methods
• The GALAH Survey: Scientific Motivation
• The chemical make-up of the Sun: A 2020 vision
• The Cannon: A data-driven approach to stellar label determination
• Non-LTE calculations for neutral Na in late-type stars using improved atomic data
• NLTE line formation of Fe for late-type stars. I. Standard stars with 1D and <3D> model atmospheres
• NLTE line formation of Fe in late-type stars II: 1D spectroscopic stellar parameters
• Gaia FGK Benchmark Stars - Metallicity
• Gaia FGK Benchmark Stars: Effective temperatures and surface gravities
• The Payne: self-consistent ab initio fitting of stellar spectra
• Observational constraints on the origin of the elements. IV: The standard composition of the Sun
• NLTE abundances of Mn in a sample of metal-poor stars
• The wide-field, multiplexed, spectroscopic facility WEAVE: Survey design, overview, and simulated implementation
• Chromium: NLTE abundances in metal-poor stars and nucleosynthesis in the Galaxy
• The Gaia-ESO Public Spectroscopic Survey: Implementation, data products, open cluster survey, science, and legacy
• Atomic data for the Gaia-ESO Survey
• The Sixth Data Release of the Radial Velocity Experiment (RAVE) -- II: Stellar Atmospheric Parameters, Chemical Abundances and Distances
• Carbon, oxygen, and iron abundances in disk and halo stars. Implications of 3D non-LTE spectral line formation
• NLTE analysis of Co I/Co II lines in spectra of cool stars with new laboratory hyperfine splitting constants
• Abundances in the Local Region II: F, G and K Dwarfs and Subgiants
• Observational constraints on the origin of the elements. I. 3D NLTE formation of Mn lines in late-type stars
• The Gaia FGK Benchmark Stars - High resolution spectral library
• Ionization balance of Ti in the photospheres of the Sun and four late-type stars
• The Gaia-ESO Public Spectroscopic Survey: Motivation, implementation, GIRAFFE data processing, analysis, and final data products
• Non-LTE line formation for heavy elements in four very metal-poor stars
• Red Supergiant Stars as Cosmic Abundance Probes: II. NLTE Effects in J-band Silicon Lines
• Kinetic equilibrium of iron in the atmospheres of cool stars III. The ionization equilibrium of selected reference stars
• Non-local thermodynamic equilibrium stellar spectroscopy with 1D and <3D> models - I. Methods and application to magnesium abundances in standard stars
• NLTE analysis of Sr lines in spectra of late-type stars with new R-matrix atomic data
• The GALAH Survey: Non-LTE departure coefficients for large spectroscopic surveys
• Solar oxygen abundance
• The formation of the Milky Way halo and its dwarf satellites; a NLTE-1D abundance analysis. II. Early chemical enrichment
• Lithium abundance patterns of late-F stars: an in-depth analysis of the lithium desert
• APOGEE Kinematics I: Overview of the Kinematics of the Galactic Bulge as Mapped by APOGEE
• Assessment of [Fe/H] determinations for FGK stars in spectroscopic surveys
• Observational constraints on the origin of the elements II. 3D non-LTE formation of Ba ii lines in the solar atmosphere
• NLTE Chemical abundances in Galactic open and globular clusters
• Non-LTE Radiative Transfer with Turbospectrum
• The Gaia-ESO survey: 3D NLTE abundances in the open cluster NGC 2420 suggest atomic diffusion and turbulent mixing at the origin of chemical abundance variations
• PySME -- Spectroscopy Made Easier
• ERGO-ML I: Inferring the assembly histories of IllustrisTNG galaxies from integral observable properties via invertible neural networks
• Inference of cosmic-ray source properties by conditional invertible neural networks
• Observational constraints on the origin of the elements. VII. NLTE analysis of Y II lines in spectra of cool stars and implications for Y as a Galactic chemical clock
• A new method for the inversion of atmospheric parameters of A/Am stars
• High-resolution Spectroscopic Study of Dwarf Stars in the Northern Sky: Lithium, Carbon, and Oxygen Abundances
• Stellar Parameter Determination from Photometry using Invertible Neural Networks
• The RAdial Velocity Experiment (RAVE): Parameterisation of RAVE spectra based on convolutional neural networks
• Korg: a modern 1D LTE spectral synthesis package
• The stellar content of the Hamburg/ESO survey. III. Field horizontal-branch stars in the Galaxy
• An empirical recipe for inelastic hydrogen-atom collisions in non-LTE calculations
• 3D NLTE modelling of Y and Eu. Centre-to-limb variation and solar abundances
• Emission-line diagnostics of HII regions using conditional Invertible Neural Networks
• HST Low Resolution Stellar Library
• Noise-Net: Determining physical properties of HII regions reflecting observational uncertainties
• Inferring stellar parameters and their uncertainties from high-resolution spectroscopy using invertible neural networks
• Spectral classification of young stars using conditional invertible neural networks I. Introducing and validating the method
• A deep-learning approach to the 3D reconstruction of dust density and temperature in star-forming regions
• Disentangling stellar atmospheric parameters in astronomical spectra using Generative Adversarial Neural Networks
• M3DIS -- A grid of 3D radiation-hydrodynamics stellar atmosphere models for stellar surveys. II. Carbon-enhanced metal-poor stars
• Gaia FGK Benchmark Stars: spectral library, metallicities and abundances of $\alpha$ and Fe-peak elements of the third version
• Spectral classification of young stars using conditional invertible neural networks II. Application to Trumpler 14 in Carina