Speaker
Description
We are developing an open-source auto–differentiable spectral modeling code "ExoJAX" to perform atmospheric retrieval for high-resolution exoplanetary spectra by ab initio spectral fitting. This enables fully Bayesian inference of atmospheric parameters with the Hamiltonian Monte Carlo. Meanwhile, the parameter estimation of M dwarfs, which are important as host stars of various exoplanets, has mostly depended on pre-computed radiative-chemical equilibrium models. However, it has been reported in many literatures that the incompleteness of those models is a fundamental problem due to the spectral complexity caused by the cool atmosphere. We have retrieved the atmospheric parameters of the well-known M dwarf Barnard's star ($T_{\mathrm{eff}} \sim$ 3200 K) by applying the ExoJAX to its high-resolution near-infrared spectra ($R\sim$ 80000, $YJH$–band) independently of stellar models. We succeeded in reproducing well the molecular and atomic lines simultaneously at several wavelengths, and estimated the temperature-pressure profile and mixing ratios of each atomic and molecular species. Nevertheless, at other wavelengths, we found that the same way could not reproduce the absorption features at all, so we discuss the inaccuracy and missing absorption lines of the latest line list. The inspection of atmospheric retrieval of M dwarfs, for which is easier to obtain high-resolution spectra with high quality than in exoplanets, will highlight the path and caveats to accurate retrieval of exoplanets.
