不同的高斯过程方法在系外行星大气透射光谱数据处理中的比较研究

A Comparative Study of Different Gaussian Process Methods in Exoplanet Atmospheric Transmission Spectroscopy Data Processing

凌星法是研究系外行星大气的重要方法之一.在地面上对系外行星大气进行观测时,传统凌星法通常在目标星之外同时观测一颗或多颗参考星,利用较差光谱法并结合高斯过程回归,以同步监测并线性地消除凌星光变曲线内由于地球大气和仪器效应产生的系统噪声.尝试了一种新方法,即不使用较差光谱法,而是将参考星的光变曲线作为高斯过程的相关量来探索光变曲线中的系统噪声,该方法允许在目标星光变曲线和同时记录的时间序列间进行更灵活的非线性映射,以此详尽地探索凌星信号中存在的系统噪声.更进一步,还测试了没有参考星的情况,以模拟视场中无参考星的行星系统的凌星观测.对此进行了试验和研究,将此方法应用于一次热木星WASP-69b的凌星光谱数据中.研究发现,对于WASP-69b的这次凌星数据,新方法在使用参考星时获得的白光凌星深度精度优于传统方法,获得的分光光变曲线残差普遍小于传统方法得到的残差,而在没有参考星情况下,利用已知的凌星中心时间仍可以得出可接受的结果.更多针对凌星数据的测试,可进一步验证新方法的优势和适用性.

The transit method is one of the important ways to study the atmosphere of exoplanets.When observing the atmosphere of exoplanets on the ground,the traditional transit method usually observes one or more comparison stars nearby the target star at the same time,and a series of differential spectra can be obtained.The system noise caused by the earth’s atmosphere and instrument effect in the transit light curve will be modelled and removed using Gaussian process regression.Here we try a new method,that is,instead of using the differential spectroscopy,the light curves of the comparison stars are used as the relevant quantity of the Gaussian process to explore the system noise in the light curve,which allows for more flexible nonlinear mapping between the light curve of the target star and the simultaneously recorded time series,and thus can be used to explore the system noise existing in the transit signal in details.Furthermore,we also tested the case of no comparison stars.We applied the new method to one transit spectral data of the hot Jupiter WASP-69b.We find that for this WASP-69b transit data,the new method obtains better white-light transit depth accuracies than the traditional method when a reference star is used,the residuals of the spectroscopic light curves are generally smaller than those obtained by the traditional method,and acceptable results can still be obtained without a reference star and with a known transit center time.More tests on transiting data will further validate the advantages and applicability of the new method.