Aerosols, which are nano- to microscale particles suspended in the air, have a significant influence on climate, weather, health, and ecology. The size and composition of aerosol particles determine their interactions with atmospheric compounds. Among these, Sodium Chloride (NaCl) is the most abundant aerosol particle. Surface-sensitive Ambient Pressure X-ray Photoelectron Spectroscopy (APXPS) has revealed that NaCl aerosol particles undergo structural transformations depending on atmospheric humidity levels . However, deciphering the surface atomic arrangement responsible for the interaction of NaCl aerosols with the atmosphere under humid conditions remains a challenge.

To address this challenge, we investigate the surface structure and interactions of aerosols under humid conditions using density functional theory (DFT) and machine learning (ML). We employ Bayesian Optimization (BO) alongside DFT to model the adsorption geometry of atmospheric water on the NaCl aerosol surface. Specifically, we utilize the Bayesian Optimization Structure Search (BOSS) code to sample various configurations of atmospheric water on aerosol surfaces, enabling the learning of adsorption energy landscapes. Once the adsorption geometry is optimized, we apply the Δ self-consistent field (ΔSCF)  approach to compute the core electron binding energy of the Na 1s electron in NaCl aerosol particles. By comparing the Na 1s binding energy before and after atmospheric water adsorption, we can interpret changes in XPS spectra under humid conditions.

This study leverages ML-driven DFT to reveal atomic-scale interactions between aerosols and atmospheric water, providing insights into APXPS data. Additionally, it serves as a framework for exploring interactions between aerosols and more complex atmospheric compounds.