deepCool: Fast and Accurate Estimation of Cooling Rates in Irradiated Gas with Artificial Neural Networks

Accurate models of radiative cooling are a fundamental ingredient of modern cosmological simulations. Without such models, accreted baryonic gas cannot efficiently dissipate its energy and collapse to the centres of haloes to form stars. It is well established that local variations in the spectral energy distribution of the radiation field can drastically alter the cooling rate. In this talk I will introduce deepCool, a set of neural networks for calculating the total cooling, total heating and metal-line only cooling rates of irradiated astrophysical gases. deepCool performs to within a ~5% error of the photoionisation code CLOUDY, with extremely high execution speed and low memory cost. I will discuss the use of deepCool as a subroutine of the radiation hydrodynamics code RAMSES-RT, and suggest suitable methods for its deployment in large-scale cosmological simulations. Finally, I will discuss the Aspen simulations, in which we invoke similar methods to predict line emission from high-redshift galaxies. Such simulations are ideal for aiding the interpretation of observations from ALMA and JWST.