Non-physical species in pressure-dependent networks: by the switch of an atom Sharon Haba, Nelly Mitnik, Mark E. Fuller, Alon Grinberg Dana Department of Chemical Engineering, Technion – Israel Institute of Technology, Israel Large-scale application of connectivity-based reaction templates, as used in automatically generated chemical kinetic models, 1 may result in non-physical species either in the model itself or in pressure-dependent networks used to compute rate coefficients. This problem, which has not received enough attention, results in ill-formulated networks and may introduce considerable errors into computed pressure-dependent rate coefficients, eventually propagating into the predicted observables. We begin by consider two structural isomers, ⋅ OONNH and ⋅ NNOOH , and show using several established electronic structure methods that the former corresponds to an energetic well while the latter is non-physical (Figure 1) at various optimization methods (DFT and CCSD(T)-F12 2 ). We use multi-dimensional scans of internal coordinates (bond distances and dihedral angles) to explain this difference. Next, we explore a pressure-dependent network on the N 2 O 2 H potential energy surface (PES). We show that an automatically generated PES may contain non-physical species and requires additional treatment. We identify all non-physical species in the network, correct the network by properly connecting physical isomers, e.g., using a neural network based method, 3 and show the effect of this modification on the computed well-skipping pressure- dependent reaction rate coefficients. Finally, we give recommendations on how to generalize and automate the approach to treat such cases when considering large numbers of PES networks.
Figure 1: ( A ) ⋅ OONNH and ( B ) ⋅ NNOOH optimized at the ωB97XD/Def2TZVP level of theory 4,5 starting from reasonable energetic well conformers embedded using force fields. 6 While structure A was found to represent an energetic well, structure B was found to break into smaller fragments when optimized. References 1. M. Liu, A. Grinberg Dana, M.S. Johnson, M.J. Goldman, A. Jocher, A.M. Payne, C.A. Grambow, K. Han, N.W. Yee, E.J. Mazeau, K. Blondal, R.H. West, C.F. Goldsmith, W.H. Green, Reaction Mechanism Generator v3.0: Advances in Automatic Mechanism Generation, J. Chem. Inf. Model. 2021, 61(6), 2686-2696, https://doi.org/10.1021/acs.jcim.0c01480 2. T.B. Adler, G. Knizia, H.-J. Werner, A simple and efficient CCSD(T)-F12 approximation, J. Chem. Phys. 2007, 127, 221106, https://doi.org/10.1063/1.2817618 3. L. Pattanaik, J.B. Ingraham, C.A. Grambow, W.H. Green, Generating transition states of isomerization reactions with deep learning, Phys. Chem. Chem. Phys. 2020, 22, 23618-23626, https://doi.org/10.1039/D0CP04670A 4. J.-D. Chai, M. Head-Gordon, Long-range corrected hybrid density functionals with damped atom–atom dispersion corrections, Phys. Chem. Chem. Phys. 2008, 10, 6615-6620, https://doi.org/10.1039/B810189B 5. F. Weigend, R. Ahlrichs, Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy, Phys. Chem. Chem. Phys. 2005, 7, 3297-3305, https://doi.org/10.1039/B508541A 6. A. Grinberg Dana, D. Ranasinghe, H. Wu, C. Grambow, X. Dong, M. Johnson, M. Goldman, M. Liu, W.H. Green, "ARC - Automated Rate Calculator", version 1.1.0, https://github.com/ReactionMechanismGenerator/ARC, DOI: 10.5281/ zenodo.3356849
P09
© The Author(s), 2022
Made with FlippingBook Learn more on our blog