Interpreting Cometary Water Activity Using the AI-Driven Thermophysical Model ThermoONet

Water ice is the primary volatile component of most comets. Its various forms and abundances across different comets provide critical clues about the formation and evolution of the Solar System, as well as planetary habitability. As comets approach the Sun, rising temperatures trigger the sublimation of surface and subsurface water ice. Within the water snowline, the resulting water vapor dominates the gas coma and acts as the primary driver of dust ejection, forming the dust coma and tail. Studying the spatial and temporal variations in cometary water activity enables us to infer key physical properties, such as nucleus size, porosity, composition, refractory-to-ice ratio, and erosion history.

A fundamental metric of water activity is the global water production rate, which quantifies the total outgassing of water vapor from the nucleus. Accurately interpreting this rate requires high-fidelity thermophysical modelling that accounts for the specific radiative environment surrounding the nucleus. Traditionally, such modelling is conducted by solving one-dimensional heat conduction equations numerically. Although numerous tools have been developed and successfully applied to data from telescopic and space-based observations, conventional numerical approaches are computationally intensive and often struggle with high-resolution shape models or large parameter spaces.

In this study, we present the application of our AI-based thermophysical model, ThermoONet, to the interpretation of cometary water production rates. ThermoONet is a general-purpose neural network capable of modelling a wide range of cometary nuclei with diverse physical characteristics [1]. It achieves comparable accuracy to traditional numerical methods while reducing computation time by five orders of magnitude. We demonstrate the utility of ThermoONet by analysing water production curves of dozens of comets observed by SOHO/SWAN [2]. Through curve fitting, we retrieve key properties such as nucleus sizes, offering new insights into cometary formation and evolutionary processes.

1. Zhao, S., Shi, X. and Lei, H., 2025. ThermoONet: Deep learning-based small-body thermophysical network: Applications to modeling the water activity of comets. Astronomy & Astrophysics, in press.

2. Combi, M.R., Mäkinen, T.T., Bertaux, J.L., Quémerais, E. and Ferron, S., 2019. A survey of water production in 61 comets from SOHO/SWAN observations of hydrogen Lyman-alpha: Twenty-one years 1996–2016. Icarus, 317, pp.610-620.