Using a neural network approach to accelerate disequilibrium chemistry calculations in exoplanet atmospheres
Abstract
In this era of exoplanet characterization with JWST, the need for a fast implementation of classical forward models to understand the chemical and physical processes in exoplanet atmospheres is more important than ever. Notably, the timedependent ordinary differential equations to be solved by chemical kinetics codes are very timeconsuming to compute. In this study, we focus on the implementation of neural networks to replace mathematical frameworks in onedimensional chemical kinetics codes. Using the gravity gradient, temperaturepressure profiles, initial mixing ratios, and stellar flux of a sample of hotJupiter's atmospheres as free parameters, the neural network is built to predict the mixing ratio outputs in steady state. The architecture of the network is composed of individual autoencoders for each input variable to reduce the input dimensionality, which is then used as the input training data for an LSTMlike neural network. Results show that the autoencoders for the mixing ratios, stellar spectra, and pressure gradients are exceedingly successful in encoding and decoding the data. Our results show that in 90 per cent of the cases, the fully trained model is able to predict the evolved mixing ratios of the species in the hotJupiter atmosphere simulations. The fully trained model is ~10^{3} times faster than the simulations done with the forward, chemical kinetics model while making accurate predictions.
 Publication:

Monthly Notices of the Royal Astronomical Society
 Pub Date:
 September 2023
 DOI:
 10.1093/mnras/stad1763
 arXiv:
 arXiv:2306.07074
 Bibcode:
 2023MNRAS.524..643H
 Keywords:

 exoplanets;
 planets and satellites: atmospheres;
 planets and satellites: gaseous planets;
 Astrophysics  Earth and Planetary Astrophysics;
 Computer Science  Machine Learning
 EPrint:
 13 pages, 9 figures, accepted for publication at MNRAS