Using Independent Component Analysis to Detect Exoplanet Reflection Spectrum from Composite Spectra of Exoplanetary Binary Systems

The analysis of the wavelength-dependent albedo of exoplanets represents a direct way to provide insight into their atmospheric composition and to constrain theoretical planetary atmosphere modeling. Wavelength-dependent albedo can be inferred from the exoplanet's reflected light of the host star, but this is not a trivial task. In fact, the planetary signal may be several orders of magnitude lower (10(-4) or below) than the flux of the host star, thus making its extraction very challenging. Successful detection of the planetary signature of 51 Peg b has been recently obtained by using cross-correlation function or autocorrelation function techniques. In this paper we present an alternative method based on the use of Independent Component Analysis (ICA). In comparison to the above-mentioned techniques, the main advantages of ICA are that the extraction is "blind" i.e., it does not require any a priori knowledge of the underlying signals, and that our method allows us not only to detect the planet signal but also to estimate its wavelength dependence. To show and quantify the effectiveness of our method we successfully applied it to both simulated data and real data of an eclipsing binary star system. Eventually, when applied to real 51 Peg + 51 Peg b data, our method extracts the signal of 51 Peg but we could not soundly detect the reflected spectrum of 51 Peg b mainly due to the insufficient signal-to-noise ratio of the input composite spectra. Nevertheless, our results show that with "ad hoc" scheduled observations an ICA approach will be, in perspective, a very valid tool for studying exoplanetary atmospheres.