Surface-sensitive experimental methods like ambient pressure X-ray photoelectron spectroscopy (APXPS) are powerful methods for probing intricate evaporation-condensation processes on atmospheric aerosol nanoparticles. Subsequently, the resultant spectra need to be interpreted and analyzed to reveal the underlying structures. Electronic structure calculations have been used to decode the structures of adsorbed molecules on various surfaces by generating simulated XPS spectra and comparing them to experimental results. However, their extensive use is limited by the high computational cost required for conclusive and accurate results.  

In this study, we will use machine learning methods to efficiently correlate atmospheric molecular adsorption geometries on aerosol particles to their corresponding XPS spectra. Data generated from experimental APXPS and computational XPS will be used to form comprehensive datasets. We will engineer novel descriptors and evaluate them to improve spectra prediction. In describing the complex molecular adsorption geometries, the descriptors developed in this study will facilitate further data-driven studies in other fields where adsorption plays a key role. Furthermore, reliable XPS spectra prediction will accelerate atmospheric research by expediting interpretation and alleviating the need for resource-intensive experimental measurements.