PINION: physics-informed neural network for accelerating radiative transfer simulations for cosmic reionization

With the advent of the Square Kilometre Array Observatory (SKAO), scientists will be able to directly observe the Epoch of Reionization by mapping the distribution of neutral hydrogen at different redshifts. While physically motivated results can be simulated with radiative transfer codes, these simulations are computationally expensive and cannot readily produce the required scale and resolution simultaneously. Here we introduce the Physics-Informed neural Network for reIONization (PINION), which can accurately and swiftly predict the complete 4D hydrogen fraction evolution from the smoothed gas and mass density fields from pre-computed N-body simulation. We trained PINION on the C-2-Ray simulation outputs and a physics constraint on the reionization chemistry equation is enforced. With only five redshift snapshots, PINION can accurately predict the entire reionization history between z = 6 and 12. We evaluate the accuracy of our predictions by analyzing the dimensionless power spectra and morphology statistics estimations against C-2-Ray results. We show that while the network's predictions are in very good agreement with simulation to redshift z > 7, the network's accuracy suffers for z < 7. We motivate how PINION performance could be improved using additional inputs and potentially generalized to large-scale simulations.

Automated summary

With the advent of SKAO, the Epoch of Reionization can be observed directly by mapping the distribution of neutral hydrogen. We introduce PINION, a Physics-Informed neural Network that accurately and swiftly predicts the 4D hydrogen fraction evolution using pre-computed N-body simulation. PINION is trained on C-2-Ray simulation outputs and can accurately predict the reionization history between z=6 and 12.