Atomic structures, conformers and thermodynamic properties of 32k atmospheric molecules

Abstract

Low-volatile organic compounds (LVOCs) drive key atmospheric processes, such as new particle formation (NPF) and growth. Machine learning tools can accelerate studies of these phenomena, but extensive and versatile LVOC datasets relevant for the atmospheric research community are lacking. We present the GeckoQ dataset with atomic structures of 31,637 atmospherically relevant molecules resulting from the oxidation of α-pinene, toluene and decane. For each molecule, we performed comprehensive conformer sampling with the COSMOconf program and calculated thermodynamic properties with density functional theory (DFT) using the Conductor-like Screening Model (COSMO). Our dataset contains the geometries of the 7 Mio. conformers we found and their corresponding structural and thermodynamic properties, including saturation vapor pressures (pSat), chemical potentials and free energies. The pSat were compared to values calculated with the group contribution method SIMPOL. To validate the dataset, we explored the relationship between structural and thermodynamic properties, and then demonstrated a first machine-learning application with Gaussian process regression.

Computational Tools for Handling Molecular Clusters: Configurational Sampling, Storage, Analysis, and Machine Learning

“…NNs have been used to model large sulfuric acid–dimethylamine clusters and the NN potential ANI-2x has been benchmarked for small dimer clusters . KRR/GPR has been used to predict cluster binding energies, − , saturation vapor pressures of organic molecules, , and chemical potentials of organic molecules in atmospherically relevant solutions …”

“…48 NNs have been used to model large sulfuric acid−dimethylamine clusters 47 and the NN potential ANI-2x 91 has been benchmarked for small dimer clusters. 92 KRR/GPR has been used to predict cluster binding energies, [44][45][46]61 saturation vapor pressures of organic molecules, 41,42 and chemical potentials of organic molecules in atmospherically relevant solutions. 43 Our ML-oriented subpackage, JKML, offers an interface between the JKQC-constructed database files (e.g., those stored in ACDB 2.0) and two ML programs, quantum machine learning (QML 93 ) and SchNetPack.…”

“…Machine learning (ML) methodologies have proven to be highly advantageous due to their capacity to replace time-intensive QC calculations. Several recent studies have harnessed ML techniques to investigate molecular clusters. − To assist in the creation of ML models for molecular cluster studies, we introduce Jammy Key for machine learning (JKML). JKML streamlines the training of ML models and their various applications such as predictions of energy/forces and even the creation of ML-based calculators.…”

Data‐Driven Compound Identification in Atmospheric Mass Spectrometry

“…Figure shows a first visualization of this overlap. The figure presents a t‐stochastic neighborhood embedding (t‐SNE) analysis for three atmospheric molecular datasets (here referred to as Gecko, [ 165,166 ] Wang, [ 167 ] and Quinones [ 168,169 ] ) and two datasets of drug and metabolite compounds, representative of those in mass spectral databases (nablaDFT [ 170,171 ] and Massbank of North America [ 103 ] ). t‐SNE clustered the compounds according to the similarity of their (molecular) topological fingerprints.…”

“…In addition, community datasets, such as refs. [68, 166, 187, 188], could complement data infrastructures. They offer distinct advantages such as having been purposefully curated with design criteria like similarity and balance in mind.…”