Project G
The needle in a haystack: hunting for exocomets
Julia Bodensteiner, Amelia Bayo & Giulia Roccetti
Even though the first exocomets (a.k.a "falling evaporating bodies", FEBs) were observationally discovered before the first exoplanets orbiting sun-like stars were found (by Ferlet in 1987), their discovery rates have not followed parallel. The presence of FEBs was first inferred from transient components in metallic lines in the famous Beta Pictoris system. In short, by taking many high-quality spectra of the star Beta Pictoris, one could literally see that the "fingerprints" of calcium in the spectra (so-called spectral lines), changed with time: "the star" was visible as one stable contribution, but also some random contributions that appeared and disappeared in time scales of hours and days. These "capricious" additional pockets of gas between us and the star were (and still are) interpreted as the gaseous tails of comets.
However, since this first exciting discovery, targeted individual star monitoring campaigns have reported only three other bonafide FEBs (comets!) hosting stars. More recently, larger survey-like initiatives have been conducted taking advantage of spectroscopic archival data originally obtained in the context of exoplanet searches. Despite these larger efforts, the census remains in the few tens and is severely affected by biases in the samples and data treatment.
Some of these biases can be "corrected" for by analysing more homogeneous data sets, and by using more sophisticated techniques that incorporate deep learning, for instance.
And we are in luck: just by the end of 2023, a huge effort was released by Barbieri to curate the archive of HARPS, one of the most stable "planets hunter" instruments through the last decades. The catalogue provides a reliable compilation of 289843 observations of 6488 unique astronomical objects covering very relevant metallic lines like Ca II H and K (where exocomets are "easier to spot").
This project consists in analysing (temporal) changes in the Ca II H and K lines for subsets of these ~6500 objects in a progressively automatic way with the ultimate goal of developing a fully functional anomaly detection flow (among other techniques testing autoencoders and variational recurrent autoencoders, similar to the one described in Sanchez-Saez et al. 2022 and in Bayo et al. 2024 in prep.). With this, we will hopefully be able to detect the spectroscopic signatures of new, so far undetected exocomets around other stars in our solar neighbourhood.