Expanding plausible biosignature gas candidates for detection in habitable exoplanet atmospheres

Abstract

Exoplanet science has become a central spotlight of astrophysics in the last decade, due in part to the discovery of more than 4400 exoplanets, of which 60 are potentially habitable. One rapidly growing branch of exoplanet research is atmosphere simulations for the search for signs of life with upcoming telescopes via remote detection of biosignature gases in exoplanet atmospheres. Existing work, however, has thus far only assessed ~ 20 molecules as potential biosignature gases. Recent work has proposed to study tens of thousands of small molecules as possible biosignature gases and constructed the All Small Molecules (ASM) Database. Assessing each of these molecules individually is not feasible, so we need to develop an efficacious approach to sort through these molecules to nd the best candidate biosignature gases. In this thesis, I provide the first practical solution to filter molecules in the ASM Database by introducing a triage framework.

Using the triage approach, I can quickly retrieve a compact cluster of plausible biosignature gas candidates that match the query spectrum to speed up the assessment. The same process allows us to construct mock prototype spectra for many unassessed gases to retrieve plausible biosignature gases in response to a remote observationthe triage approach proceeds by first clustering molecular IR spectra to identify highly detectable molecular subgroups. Next, the approach lters highly detectable clusters by the ability of the molecules contained therein to accumulate in an exoplanet atmosphere using proxy estimates for solubility, volatility, and photochemistry. Finally, the triage prioritizes the filtered molecules with notable biological functions or abundant production by life on Earth for further assessment.

Using this process substantially and eciently narrows the molecules in the ASM Database to a typically small candidate subset bounded only by the spectral resolution that the James-Webb Space Telescope (JWST) supports. Additional information or improvements to spectral resolution will undoubtedly further narrow the candidates, possibly to a unique association. A result that emerges is that I can distinguish among dierent groups of hydrocarbons even at a spectral resolution of [delta]v = 100 cm⁻¹ (R = 100 at 1 [mu]m and R = 10 at 10 [mu]m). Hydrocarbon molecules with the conjugated-diene and iso-ene structure are of particular interest because their smallest member, isoprene (C₅H₈), is produced in the highest amounts on Earth (equivalent to methane) by a broad spectrum of life; from bacteria to plants and mammals.

I further discover that isoprene has no abiotic false-positives, and it can be a potential biosignature gas for exoplanets with anoxic atmospheres transiting M dwarf stars. However, JWST observation simulations suggest that isoprene is indistinguishable from other conjugated-diene and iso-ene molecules. Identifying the exact molecule will require [delta]v = 10 cm⁻¹ (R = 1000 at 1 [mu]m and R = 100 at 10 [mu]m). Furthermore, the triage approach identifies the carbonyl group (molecules with C=O functional group) as highly detectable and prioritizes carbonyls for further assessment because all life produces carbonyl, and their chemistry is a fundamental pillar in Earth's biochemistry. The triage approach, however, correctly rules out carbonyls as possible biosignature gases due to their high solubility in water, low volatility, and rapid photochemical destruction.

In the study of the carbonyl group, I discover that carbonyl destruction may lead to a signicant accumulation of carbon monoxide (CO) in reducing H₂-dominated atmospheres. Although this offers a counter-argument to popular wisdom that carbon monoxide is an \anti-biosignature" gas, CO also has abiotic pathways that temper such an inference. For the first time, it is feasible for us to use a largely automated process to narrow down thousands of potential biosignature gas candidates. The projected savings over a purely human endeavor that the Machine Learning-enabled triage framework offers is an exciting new interdisciplinary outcome. I posit that the triage approach's efficacy is necessary for proposing new biosignature gas candidates, thus opening a new avenue, in conjunction with AI, for biosignature gas research.

With further improvements in resolution, even modest ones, biosignature detectability sharpens by sifting candidates far more finely, and we may even excitedly anticipate, uniquely. Keywords: Biosignature Gases, Molecular IR Spectroscopy, Triage, Machine Learning, Hierarchical Clustering, Exoplanet Atmosphere, Transmission Spectroscopy