(38e) Recent Advances in Automated Transition State Theory Calculations: A Case Study Using Butanol

With the effects of global climate change becoming more apparent, researchers have been pushing for a shift from conventional fossil fuels to greener alternatives. However, these fuels need to be adequately studied a priori to make reasonable experimental choices. In these studies, detailed reaction mechanisms need to be constructed to explain the complex combustion phenomena. Because of the sheer number of thermodynamic and kinetic parameters, they are often estimated rather than determined experimentally. Ab initio calculations and transition state theory (TST) can determine thermodynamic and kinetic parameters more accurately than commonly used estimates. However, given the number of reactions and species present in even simple combustion systems, these calculations need to be automated.

Many researchers have been working on tools to perform these automated TST calculations, but for this work we aim to improve AutoTST, the automated transition state theory code first developed by Bhoorasingh et al. [1]. The first generation of AutoTST would start by creating a transition state geometry guess by using a functional group tree to decide on key distances in the reaction center and using RDKit [2], an open source cheminformatics toolkit, to generate the starting geometry. AutoTST would then employ Gaussian [3] to perform a series of partial optimizations to arrive at an optimized transition state complex. This complex is then validated through automated intrinsic reaction coordinate (IRC) calculations and validated geometries are then fed into Arkane [4] to obtain kinetic parameters. This work has improved AutoTST by: rewriting AutoTST as an independent, reusable, module; including the effects of 1-D hindered rotors; adding a systematic conformer analysis; and bypassing the expensive IRC calculations by using vibrational analysis to validate transition states.

To assess the impact of these improvements, we performed automated TST calculations on 1117 reactions and 298 species present in a model for the combustion of butanol [6] and compared these reaction rates and thermodynamic expressions against values calculated by the first generation of AutoTST, values estimated by RMG, and values published in the model. From here, we identified reactions and species where our calculations deviated the most from comparisons, perform benchmark calculations to generate high fidelity kinetics and thermodynamic parameters for these reactions and species, and compare and investigate the sources of these discrepancies. This framework will enable researchers the ability to generate large amounts of high fidelity kinetic and thermodynamic training data on-the-fly for machine-learning based estimation methods.

References:

[1] Bhoorasingh, P. L.; Slakman, B. L.; Khanshan, F. S.; Cain, J. Y.; West, R. H. Automated Transition State Theory Calculations for High-Throughput Kinetics. J. of Phys. Chem. A. 2017, 121, 6896-6904.

[2] RDKit: Open-source cheminformatics; http://www.rdkit.org

[3] Gaussian 16, Revision B.01, Frisch M. J. et al., Gaussian, Inc., Wallingford CT, 2016.

[4] Allen, J. W.; Green, W. H.; Arkane: http://reactionmechanismgenerator.github.io/RMG-Py/users/arkane/credits.html

[5] Gao, C. W.; Allen, J. W.; Green, W. H.; West, R. H. Reaction Mechanism Generator: Automatic Construction of Chemical Kinetic Mechanisms. Comput. Phys. Commun. 2016, 203, 212–225.

[6] Sarathy S. M. et al. A comprehensive chemical kinetic combustion model for the four butanol isomers. Comb. Flame. 2016, 159, 2028-2055.