Machine learning and reaction dynamics: From spectroscopic constants of diatomic molecules to buffer gas chemistry

This thesis explores the spectroscopic properties and chemistry of diatomic molecules, which hold significant promise for applications in areas like quantum information and ultracold chemistry. Firstly, the Diatomic Molecular Spectroscopy Database, accessible through a dynamic website, has been implemented. This database predominantly consolidates spectroscopic information while enabling the computation and visualization of Franck-Condon factors, and is adaptable for user contributions. Based on this database, machine learning models have been built to effectively reveal relationships among spectroscopic constants, with input features based on constituent atoms' group and period. Similarly, a comprehensive dataset of contemporary experimental electric dipole moments has been created. Utilizing this dataset, it has been shown that a machine learning model can accurately predict dipole moments using spectroscopic constants. The availability of precise spectroscopic data allows for a rigorous assessment of advanced quantum chemistry methods. Specifically, we investigated the accuracy of coupled-cluster with single, double, and perturbative triple excitations [CCSD(T)] in predicting electric dipole moments when combined with different basis sets. Additionally, the hyperfine constants for the a³\Pi state of aluminum monofluoride (AlF) have been computed and compared to experimental values. Our study underscores the significance of a thorough evaluation encompassing both experimental and theoretical methodologies. AlF and calcium monofluoride (CaF), among other metal monofluorides, have emerged as highly promising options for experiments involving laser cooling and trapping of cold molecules. We have compared the efficiency of different fluorine-donor molecules producing AlF and CaF through metal atom ablation in a buffer gas cell. Additionally, we present an efficient machine learning method for fitting the potential energy surface of AlF-AlF system, trained on relevant configurations from molecular dynamics simulations at the CCSD(T) level.

Author dc:creator

• Xiangyue, Liu

• Machine learning
• Molecular dynamics
• Spectroscopic constants
• Diatomic molecules
• Buffer gas cell chemistry
• Laser cooling

Licence dc:rights.uri

• https://creativecommons.org/licenses/by-nc-sa/4.0/

Language dc:language

eng

Identifier URI

http://dx.doi.org/10.17169/refubium-41868

OAI record oai:identifier

https://refubium.fu-berlin.de/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Arefubium.fu-berlin.de%3Afub188%2F42143