A new method to measure the spectra of transiting exoplanet atmospheres using multi-object spectroscopy
Traditionally, ground-based spectrophotometric observations probing transiting exoplanet atmospheres have employed a linear map between comparison and target star light curves (e.g. via differential spectrophotometry) to correct for systematics contaminating the transit signal. As an alternative to this conventional method, we introduce a new Gaussian Processes (GP) regression-based method to analyse ground-based spectrophotometric data. Our new method allows for a generalised non-linear mapping between the target transit light curves and the time series used to detrend them. This represents an improvement compared to previous studies because the target and comparison star fluxes are affected by different telluric and instrumental systematics, which are complex and non-linear. We apply our method to six Gemini/GMOS transits of the warm (T$_{\rm eq}$ = 990 K) Neptune HAT-P-26b. We obtain on average $\sim$20 % better transit depth precision and residual scatter on the white light curve compared to the conventional method when using the comparison star light curve as a GP regressor and $\sim$20% worse when explicitly not using the comparison star. Ultimately, with only a cost of 30% precision on the transmission spectra, our method overcomes the necessity of using comparison stars in the instrument field of view, which has been one of the limiting factors for ground-based observations of the atmospheres of exoplanets transiting bright stars. We obtain a flat transmission spectrum for HAT-P-26b in the range of 490-900 nm that can be explained by the presence of a grey opacity cloud deck, and indications of transit timing variations, both of which are consistent with previous measurements.