Skip to main content

Fast, accurate chemistry for hydrodynamical models

Chemistry is important to hydrodynamical models as it effectively sets the gas temperature by determining the abundances of important coolants. However, full chemical networks such as those used in UCLCHEM are typically too computationally expensive to include in a hydrodynamical model. Similarly, solving the chemistry in a time dependent way can also be a huge computational burden and equilibrium is often assumed instead.

Chemulator is a neural network based emulator which can calculate the time dependent temperature and chemical abundances of a gas. This significantly faster than solving the chemical network directly and can be embedded in hydrodynamical models to obtain accurate temperatures without the computational cost of a full chemical and radiative trasfer solver.

Authors, Contributors

Main author: Jonathan Holdship

Contributors: S. Viti, T. J. Haworth and J. D. Ilee

You are free to download, modify, and use the code for your work. The development of the code has been funded by STFC grants ST/H001794/1 and ST/J001511/1.

Fast, accurate chemistry for hydrodynamical models

A collection of statistical emulators for the UCLCHEM and RADEX astronomical codes. The UCLCHEM emulators can be used to obtain chemical abundances for various molecules under varying physical conditions. The RADEX emulator can be used to estimate the strength of molecular lines. Both emulators are built using neural networks. For more detail on the emulation procedure and its applications, please refer to the associated paper.

Authors, Contributors

Main author: Damien de Mijolla

Contributors: Serena Viti, Jonathan Holdship, Ioanna Manolopoulou, Jeremy Yates

The software is released, as is, under an MIT license.