Observations of the Sun as a star have been key to guiding models of stellar atmospheres and additionally provide useful insights on the effects of granulation and stellar activity on radial velocity measurements. Any high resolution solar spectra will contain absorption features from Earth’s atmosphere, and it can be useful to remove these telluric features when generating a solar atlas. Additionally, it can interesting to study the telluric spectra on their own.

This work consisted of creating a telluric-corrected solar atlas covering 0.5-1.0 µm derived from solar spectra taken with a Fourier Transform Spectrograph (FTS) at the Institut für Astrophysik, Göttingen. At first, we thought this endeavor would be a breeze. But then, we remembered that Earth’s rotation is quite small (<1 km/s). With such a small velocity, the stellar lines wouldn’t move much with respect to the telluric lines, keeping the same regions blocked. The dataset did however included observations over a large range in airmass, and so we could utilize this to distinguish tellurics from solar lines, but we would have to look at many spectra at the same time (see the figure below). At these resolutions and signal-to-noise ratios, telluric models have too many discrepancies to be useful without fine tuning the parameters, but forward modeling a full model of Earth’s atmosphere for several spectra at a time would be much too computationally intensive. And so, the plan of attack become obvious: we would simplify the atmosphere to a single layer. Each line would be fit with a single Lorentz profile, initiated with its center wavelength and width recorded in the HITRAN catalog. With this architecture, we could fit a bunch of spectra at once.

So, how did it work out? It turns out modeling Earth’s atmosphere as a single layer isn’t half bad. It was definitely better than using a full forward model with incorrect line parameters. We were able to divide out the telluric spectra from the data and then de-shift and average up the remaining solar spectra to create the atlas.

Some fun stuff came out of the telluric fits as well. Because we fit each line individually starting from the catalog value, we effectively solved for how much each absorption feature shifted. Comparisons between the best-fit telluric line shift and the HITRAN catalog value that summarizes how much we should expect it to shift showed excellent agreement. For HITRAN these values come from lab measurements from different sources and we could even see that some datasets were a bit better than others ;).

This work also identified a few absorption features with larger offsets relative to the catalogued line parameters and a few missing. These were corrected in HITRAN 2020 – you can read about that here!