MCMC方方法在星系形成半解析模模型中的应用

The MCMC Implementation in Semi-Analytical Model of Galaxy Formation

星系形成的半解析模型是理解星系形成中的重子物理过程的重要方法,但存在显著的缺点:物理参数太多,调控过程复杂。MCMC(Markov chain Monte Carlo)方法是现代统计计算中最重要的算法之一,通过MCMC方法可以得到星系形成半解析模型中众多物理参数的有效范围。简要介绍了半解析模型的主要物理过程和MCMC方法,综述了近年来MCMC方法在星系形成半解析模型中的应用和成果。这些结果表明MCMC方法对于限制半解析模型有很好的作用,对更好地理解星系形成中的物理过程起到促进作用。

Semi-analytical model （SAM） of galaxy formation is an important tool to understand the baryon physics processes in galaxy formation and evolution. It forms and evolves galaxies in the dark matter halos using dark matter halo merger trees and baryon physics processes. Because of the poor understanding of baryon physics processes in galaxy formation, there are many free parameters in SAM. From SAM, people can get many galaxy properties such as stellar mass function, luminosity function, correlation function and so on. The model is usually adjusted by eyes to get the appropriate prediction which could agree well with the observation. This makes SAM very difficult and unreliable in determining model parameters. MCMC （Markov chain Monte Carlo） method is an algorithm to run numerical simula- tion by random numbers which are generated by Markov chain. It is a Bayesian statistical technique for probing complex, highly degenerated probability distributions. It could sample a multidimensional space with a probability proportional to the likelihood that the model describes the observational constraints. The MCMC implementation in SAM could help to search the free parameter space automatically and get more convinced parameter estimations. In recent years, many SAM researchers have applied MCMC method into SAM and got some promising results. These studies show that the MCMC method is an excellent tool to make SAM much more powerful in understanding the galaxy baryon physics processes.