Stellar Structure Modeling using a Parallel Genetic Algorithm for Objective Global Optimization

Abstract

Genetic algorithms are a class of heuristic search techniques that apply basic evolutionary operators in a computational setting. We have designed a fully parallel and distributed hardware/software implementation of the generalized optimization subroutine PIKAIA, which utilizes a genetic algorithm to provide an objective determination of the globally optimal parameters for a given model against an observational data set. We have used this modeling tool in the context of white dwarf asteroseismology, i.e., the art and science of extracting physical and structural information about these stars from observations of their oscillation frequencies. The efficient, parallel exploration of parameter-space made possible by genetic-algorithm-based numerical optimization led us to a number of interesting physical results: (1) resolution of a hitherto puzzling discrepancy between stellar evolution models and prior asteroseismic inferences of the surface helium layer mass for a DBV white dwarf; (2) precise determination of the central oxygen mass fraction in a white dwarf star; and (3) a preliminary estimate of the astrophysically important but experimentally uncertain rate for the ^12C(alpha,gamma)^16O nuclear reaction. These successes suggest that a broad class of computationally-intensive modeling applications could also benefit from this approach.