Machine Learning-Driven Insights into Cloud CondensationNuclei Formation in Gaseous Exoplanet Atmospheres

Clouds on exoplanets are hypothesized to explain the absence of expected molecular or atomic absorption features in optical and near-infrared spectra. These observations are obtained using space-based telescopes such as CHEOPS, JWST, and the future PLATO mission, as well as ground-based telescopes like the VLT. These clouds form through the condensation of thermally stable materials onto cloud condensation nuclei (CCN) via gas-surface reactions, playing a crucial role in shaping the observed atmospheric properties. In rocky exoplanets, CCN are supplied by processes such as sandstorms, combustion, and volcanic eruptions. However, gaseous exoplanets lack direct sources of CCN. Instead, CCN form through a bottom-up nucleation process, where small molecules like TiO₂ undergo a series of chemical reactions to form larger molecular clusters [(TiO₂)_N], which grow until they reach a size sufficient to undergo a phase transition from gas to solid, ultimately forming CCN. Previous studies have explored nucleation using various theories, including Classical Nucleation Theory, Modified Classical Nucleation Theory, Non-Classical Nucleation theory, and Kinetic Nucleation Networks. All these approaches require thermochemical data for the nucleating species. While experimental studies have provided insights, limitations in replicating substellar atmospheric conditions, such as extreme temperatures and pressures, hinder their applicability. Quantum mechanical methods have been employed to address these challenges by optimizing cluster geometries and calculating thermochemical properties. However, these computationally expensive methods can take weeks to months for big clusters.

This project utilizes machine learning (ML) models to predict the geometric and thermochemical properties of large molecular clusters. The initial objective involves developing a comprehensive data catalog by integrating in-house molecular data with information from the literature. The dataset is utilized to train the ML models, which are then employed to predict the structural and thermochemical properties of larger molecular clusters. The ultimate goal is to identify clusters capable of undergoing phase transitions from the gas phase to the solid phase, serving as cloud condensation nuclei (CCN) essential for cloud formation in gaseous exoplanets. This work aligns with the scientific objectives of PLATO Work Packages 116700 and 116800. Additionally, it complements the goals of JWST Proposal 6045 (Cycle 3), titled “Detecting Ongoing Gas-to-Solid Nucleation on the Ultra-Hot Planet WASP-76 b”, which aims to observe a single transit of WASP-76 b using MIRI/LRS.