The heterogeneous reaction of N2O5 with atmospheric aerosols plays a critical role in regulating the atmospheric NOx and in the reduction of tropospheric ozone. Due to the diverse chemical composition of tropospheric aerosol particles, the experimental reactive uptake coefficient of N2O5 shows a large variation. In order to develop a theory to systematically explain the variations in the observed reactive uptake coefficients, we need a detailed understanding of the elementary reaction mechanisms involved. In this work, we have investigated the reactions of N2O5 in bulk water and at the air-water interface to model the N2O5 reactivity in aerosols. Since bond breaking and formation is involved, the standard classical force filed is not suitable to model such reactive events. More computationally demanding ab initio molecular dynamics(AIMD) can be used but at the expense of computational cost. One way to overcome the cost of AIMD without compromising the accuracy is to leverage the recently developed machine learning techniques to train a prototype reactive model using high-level DFT data as the benchmark. In this talk, I will present the development of a machine-learned reactive force field to investigate the dissociation of N2O5 in bulk water and at the air-water interface. I will also show how the energetic and kinetics of N2O5 dissociation changes at the air-water interface compared to the bulk water. I will also discuss how the detailed understanding of the elementary reactive events can be used to assess the validity of the proposed uptake mechanism of N2O5 e.g. the resistor model.

Understanding the reactive uptake of N2O5 in the atmospheric aerosol using machine learning and ab initio molecular dynamics

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.