Chemulator: Fast, accurate thermochemistry for dynamical models through emulation

Context. Chemical modelling serves two purposes in dynamical models: accounting for the effect of microphysics on the dynamics and providing observable signatures. Ideally, the former must be done as part of the hydrodynamic simulation but this comes with a prohibitive computational cost that leads to many simplifications being used in practice.
Aims: We aim to produce a statistical emulator that replicates a full chemical model capable of solving the temperature and abundances of a gas through time. This emulator should suffer only a minor loss of accuracy when compared to a full chemical solver and would have a fraction of the computational cost allowing it to be included in a dynamical model.
Methods: The gas-grain chemical code UCLCHEM was updated to include heating and cooling processes, and a large dataset of model outputs from possible starting conditions was produced. A neural network was then trained to map directly from inputs to outputs.
Results: Chemulator replicates the outputs of UCLCHEM with an overall mean squared error (MSE) of 1.7 × 10⁻⁴ for a single time step of 1000 yr, and it is shown to be stable over 1000 iterations with an MSE of 3 × 10⁻³ on the log-scaled temperature after one timzze step and 6 × 10⁻³ after 1000 time steps. Chemulator was found to be approximately 50 000 times faster than the time-dependent model it emulates but can introduce a significant error to some models.