(38d) Autoneb: A Toolkit for Automating Reaction Network Calculation
AIChE Annual Meeting
2019
2019 AIChE Annual Meeting
Topical Conference: Applications of Data Science to Molecules and Materials
Reaction Path Analysis Using Advanced Data Science Methods
Sunday, November 10, 2019 - 4:30pm to 4:50pm
We present an automated toolkit for the first-principles studies of reaction pathways, especially in complicated reaction networks involving multiple conformers of the reaction species. Challenges remain in the automatic generation of conformers of an arbitrary molecule due to the high variability in bonding and charge state and the limited availability of suitable force field parameters needed in conventional conformer search methods, especially for charged species or radicals. We overcome this through automated enumeration of all possible combination of rotatable dihedrals, followed by automated pruning to generate featured conformers. Pruning of the conformers is achieved through a 3-step process: (1) elimination of conformers with atom clashes based on molecular fingerprint, (2) geometry optimization with density functional theory and elimination of energetically unfavorable conformers, (3) extraction of featured conformers via a clustering algorithm based on the geometries. For the surviving conformers, nudged elastic band (NEB) calculations are then automatically set up for each possible reactant-product conformer pair in the reaction network. A new interpolation scheme is implemented in AutoNEB for generating the initial structures of the âbeadsâ in NEB, in order to properly handle reactions involving complicated motion in the internal coordinates. Finally, automated post-processing aids the discovery of the most favorable route in the whole network, with visualization of the energy profile. This toolkit has been applied to the reaction network for gas-phase Pomeranz-Fritsch synthesis of isoquinoline and has provided an insightful explanation of the phenomena observed in mass spectroscopy.