11:30 AM - 11:45 AM
[PAE18-10] Telluric Correction with Forward Modeling for Exoplanet Atmosphere Characterization using High-resolution Spectroscopy
Keywords:High-resolution spectroscopy, Exoplanet atmosphere, Atmosphere composition, Infrared spectroscopy, Earth atmosphere, Telluric absorption
Ground-based high-resolution spectroscopy is a powerful tool for studying exoplanetary atmospheres, enabling species detection and detailed characterization via cross-correlation technique. However, telluric absorption significantly affects infrared wavelengths, which contain many important spectral features of atmospheric molecules, making precise correction methods necessary. Principal Component Analysis (PCA) is widely used for this purpose, particularly for short-period planets, as their large Doppler shifts distinguish them from telluric and stellar lines. Meanwhile, PCA struggles for long-period planet observations, where Doppler shifts are less pronounced, making these contaminations more challenging to remove.
Forward modeling, an alternative approach that does not rely on the planet’s Doppler shifts, has demonstrated success in removing telluric lines in optical wavelengths but remains less explored in the infrared due to stronger and more clustered lines that lead to reduced correction accuracy than PCA. This is shown in our work where we first assess its performance, prior to key problem identification, using CARMENES high-resolution infrared data of a hot Saturn, HD 149026 b, comparing it to the PCA. We use ExoJAX and ADAM optimizer for respectively cross-section computation and fitting purposes. After correction, forward modeling yields a tentative H2O signal at 3.3σ, whereas PCA achieves 4.8σ.
The possible error sources for this discrepancy, which is not uncommon in the literature, are not well understood yet. To evaluate this and improve the correction method, we identify the effects of possible sources of errors using synthetic atmospheric observations in the infrared considering known uncertainties. Specifically, we examine the impact of variable H2O profiles, biased instrumental profiles (IPs), and overlapping telluric and stellar lines.
We find that inaccurate adopted atmospheric H2O abundance profile could lead to large telluric residuals after correction that may hinder the exoplanet signal, and letting the abundances at each atmospheric layer fit as free parameters could significantly reduce the effect. We also examine the impact of inaccuracies and imprecise IP assumptions for both its shape and variability with wavelength, finding that they can weaken the recovered exoplanet signal or, in extreme cases, completely obscure it. Our results suggest that deriving a more accurate and precise IP, for example from laser frequency comb spectra, might be necessary to improve correction accuracy. Lastly, we find that overlapping telluric and stellar lines could also decrease correction performance to some extent, which forces us to exclude some regions, especially those with dense stellar lines, reducing the number of available telluric lines to fit. This then emphasizes the need for explicit stellar modeling.
Our investigation here paves the way for identifying areas to improve the forward modeling technique, which would be our future work, enhancing the reliability of telluric correction and, consequently, exoplanet atmospheric detection.
Forward modeling, an alternative approach that does not rely on the planet’s Doppler shifts, has demonstrated success in removing telluric lines in optical wavelengths but remains less explored in the infrared due to stronger and more clustered lines that lead to reduced correction accuracy than PCA. This is shown in our work where we first assess its performance, prior to key problem identification, using CARMENES high-resolution infrared data of a hot Saturn, HD 149026 b, comparing it to the PCA. We use ExoJAX and ADAM optimizer for respectively cross-section computation and fitting purposes. After correction, forward modeling yields a tentative H2O signal at 3.3σ, whereas PCA achieves 4.8σ.
The possible error sources for this discrepancy, which is not uncommon in the literature, are not well understood yet. To evaluate this and improve the correction method, we identify the effects of possible sources of errors using synthetic atmospheric observations in the infrared considering known uncertainties. Specifically, we examine the impact of variable H2O profiles, biased instrumental profiles (IPs), and overlapping telluric and stellar lines.
We find that inaccurate adopted atmospheric H2O abundance profile could lead to large telluric residuals after correction that may hinder the exoplanet signal, and letting the abundances at each atmospheric layer fit as free parameters could significantly reduce the effect. We also examine the impact of inaccuracies and imprecise IP assumptions for both its shape and variability with wavelength, finding that they can weaken the recovered exoplanet signal or, in extreme cases, completely obscure it. Our results suggest that deriving a more accurate and precise IP, for example from laser frequency comb spectra, might be necessary to improve correction accuracy. Lastly, we find that overlapping telluric and stellar lines could also decrease correction performance to some extent, which forces us to exclude some regions, especially those with dense stellar lines, reducing the number of available telluric lines to fit. This then emphasizes the need for explicit stellar modeling.
Our investigation here paves the way for identifying areas to improve the forward modeling technique, which would be our future work, enhancing the reliability of telluric correction and, consequently, exoplanet atmospheric detection.