Volume 438, Number 3, August II 2005
|Page(s)||1021 - 1028|
|Published online||18 July 2005|
An artificial neural network approach to the solution of molecular chemical equilibrium
Istituto Nazionale di Astrofisica (INAF), Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, 50125 Florence, Italy e-mail: email@example.com
2 High Altitude Observatory, NCAR, 3450 Mitchell Ln, Boulder CO 80307-3000, USA
Accepted: 19 April 2005
A novel approach is presented for the solution of instantaneous chemical equilibrium problems. The chemical equilibrium can be considered, due to its intrinsically local character, as a mapping of the three-dimensional parameter space spanned by the temperature, hydrogen density and electron density into many one-dimensional spaces representing the number density of each species. We take advantage of the ability of artificial neural networks to approximate non-linear functions and construct neural networks for the fast and efficient solution of the chemical equilibrium problem in typical stellar atmosphere physical conditions. The neural network approach has the advantage of providing an analytic function, which can be rapidly evaluated. The networks are trained with a learning set (that covers the entire parameter space) until a relative error below 1% is reached. It has been verified that the networks are not overtrained by using an additional verification set. The networks are then applied to a snapshot of realistic three-dimensional convection simulations of the solar atmosphere showing good generalization properties.
Key words: molecular processes / astrochemistry / methods: numerical
© ESO, 2005
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.