Neural-Network Chemical Emulator for First-Star Formation: Robust Iterative Predictions over a Wide Density Range

Abstract

We present a neural-network emulator for the thermal and chemical evolution in Population III star formation. The emulator accurately reproduces the thermochemical evolution over a wide density range spanning 21 orders of magnitude (10${}^{-3}$-10${}^{18}$ cm${}^{-3}$), tracking six primordial species: H, H${}_2$, e${}^{-}$, H${}^{+}$, H${}^{-}$, and H${}_2^{+}$. To handle the broad dynamic range, we partition the density range into five subregions and train separate deep operator networks (DeepONets) in each region. When applied to randomly sampled thermochemical states, the emulator achieves relative errors below 10% in over 90% of cases for both temperature and chemical abundances (except for the rare species H${}_2^{+}$). The emulator is roughly ten times faster on a CPU and more than 1000 times faster for batched predictions on a GPU, compared with conventional numerical integration. Furthermore, to ensure robust predictions under many iterations, we introduce a novel timescale-based update method, where a short-timestep update of each variable is computed by rescaling the predicted change over a longer timestep equal to its characteristic variation timescale. In one-zone collapse calculations, the results from the timescale-based method agree well with traditional numerical integration even with many iterations at a timestep as short as 10${}^{-4}$ of the free-fall time. This proof-of-concept study suggests the potential for neural network-based chemical emulators to accelerate hydrodynamic simulations of star formation.

19 pages, 7 figures, Accepted for publication in ApJ

References in corpus (30)

• DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators
• DGM: A deep learning algorithm for solving partial differential equations
• The Formation of the First Star in the Universe
• The Formation of the First Stars. I. The Primordial Star Forming Cloud
• A KInetic Database for Astrochemistry (KIDA)
• Protostellar Feedback Halts the Growth of the First Stars in the Universe
• Protostar Formation in the Early Universe
• The First Stars: Mass Growth Under Protostellar Feedback
• Star formation at very low metallicity. I: Chemistry and cooling at low densities
• The critical radiation intensity for direct collapse black hole formation: dependence on the radiation spectral shape
• The Birth of a Massive First-Star Binary
• Cholla : A New Massively-Parallel Hydrodynamics Code For Astrophysical Simulation
• The mass of first stars
• GAMER-2: a GPU-accelerated adaptive mesh refinement code -- accuracy, performance, and scalability
• Formation of Population III Stars in Fossil HII Regions: Significance of HD
• Rate of three-body recombination of hydrogen molecules during primordial star formation
• Gravitational instability in protostellar disks at low metallicities
• Reducing the complexity of chemical networks via interpretable autoencoders
• Formation of Massive and Wide First-star Binaries in Radiation Hydrodynamics Simulations
• Chemulator: Fast, accurate thermochemistry for dynamical models through emulation
• Quokka: A code for two-moment AMR radiation hydrodynamics on GPUs
• Incorporating Astrochemistry into Molecular Line Modelling via Emulation
• Upper stellar mass limit by radiative feedback at low-metallicities: metallicity and accretion rate dependence
• Magnetohydrodynamic effect on first star formation: prestellar core collapse and protostar formation
• Gravitational stability and fragmentation condition for discs around accreting supermassive stars
• Ionization degree and magnetic diffusivity in the primordial star-forming clouds
• Non-ideal magnetohydrodynamic simulations of the first star formation: the effect of ambipolar diffusion
• Neural networks: solving the chemistry of the interstellar medium
• Neural network-based emulation of interstellar medium models
• Bridging Machine Learning and Cosmological Simulations: Using Neural Operators to emulate Chemical Evolution

Cited by in corpus (1)

• Systematic selection of surrogate models for nonequilibrium chemistry