Improved Configurational Sampling Protocol for Large Atmospheric Molecular Clusters

Abstract

The nucleation process leading to the formation of new atmospheric particles plays a crucial role in aerosol research. Quantum chemical (QC) calculations can be used to model the early stages of aerosol formation, where atmospheric vapor molecules interact and form stable molecular clusters. However, QC calculations heavily depend on the chosen computational method, and when dealing with large systems, striking a balance between accuracy and computational cost becomes essential. We benchmarked the binding energies and structures and found the B97-3c method to be a good compromise between the accuracy and computational cost for studying large cluster systems. Further, we carefully assessed configurational sampling procedures for targeting large atmospheric molecular clusters containing up to 30 molecules (approximately 2 nm in diameter) and proposed a funneling approach with highly improved accuracy. We find that several parallel ABCluster explorations lead to better guesses for the cluster global energy minimum structures than one long exploration. This methodology allows us to bridge computational studies of molecular clusters, which typically reach only around 1 nm, with experimental studies that often measure particles larger than 2 nm. By employing this workflow, we searched for low-energy configurations of large sulfuric acid–ammonia and sulfuric acid–dimethylamine clusters. We find that the binding free energies of clusters containing dimethylamine are unequivocally more stable than those of the ammonia-containing clusters. Our improved configurational sampling protocol can in the future be applied to study the growth and dynamics of large clusters of arbitrary compositions.

Assessment of Anharmonicities in Clusters: Developing and Validating a Minimum-Information Partition Function

“…Corrections that affect both the estimated formation free energies and the composition of the critical cluster (as demonstrated in Figure ) are therefore crucial. And as the most recent quantum chemical calculations extend to even larger and more complex cluster sizesdemonstrated by recent studies, such as those by Engsvang et al − examining clusters up to 30 or 60 moleculesthe question of the level of local anharmonicity becomes increasingly pertinent.…”

Reparameterization of GFN1-xTB for atmospheric molecular clusters: applications to multi-acid–multi-base systems

“…This highlights the need for including dynamics-based sampling procedures for atmospheric clusters even though the system might seem fairly rigid. It can also be envisioned that the improvement workow will be quite important when studying much larger (SA) 1-20 (base) 1-20 clusters as recently done by Engsvang et al 42,62 and Wu et al 38 For large clusters, the global minimum is tricky to locate, and adding dynamics-based congurational sampling might aid in the process.…”

“…PM stands for parameterization method indicating the model version. Of these methods, GFN1-xTB has shown to have the highest correlation with electronic binding energies at a higher level of theories 37,38 and have been shown to have a higher correlation with DFT trajectories for molecular dynamics than GFN2-xTB. 39 The reason GFN2-xTB performs worse than GFN1-xTB for atmospheric molecular clusters (oen involving sulfuric acid) is that there is a decrease in the number of d-functions for sulfur in the basis set for the newer GFN2-xTB model.…”

Cluster-to-particle transition in atmospheric nanoclusters

“…Here we study (SA) n (base) n clusters, with n = 1 − 15 and base = AM, MA, DMA and TMA. The SA-AM and SA-DMA systems were previously explored by Wu et al (2023) and additional sampling was carried out with CREST in this work.…”

“…However, sampling large clusters with n ≥ 5 presents an enormous challenge. To thoroughly explore the configurational space of the clusters we apply our recently identified configurational sampling workflow, which has been optimized towards the sampling of large cluster structures (Wu et al, 2023):…”