Automated chemical resonance generation and structure filtration for kinetic modeling

Abstract

This work discusses efficient and automated methods for constructing a set of representative resonance structures for arbitrary chemical species, including radicals and biradicals, consisting of the elements H, C, O, N, and S. Determining the representative reactive structures of chemical species is crucial for identification of reactive sites and consequently applying the correct reaction templates to generate the set of important reactions during automated chemical kinetic model generation. We describe a fundamental set of resonance pathway types, accounting for simple resonating structures, as well as global approaches for polycyclic aromatic species. Automatically discovering potential localized structures along with filtration to identify the representative structures was shown to be robust and relatively fast. The algorithms discussed here were recently implemented in the Reaction Mechanism Generator (RMG) software. The final structures proposed by this method were found to be in reasonable agreement with quantum chemical computation results of localized structure contributions to the resonance hybrid.

Reaction Mechanism Generator v3.0: Advances in Automatic Mechanism Generation

“…To address the increase in computational requirements for handling additional resonance pathways and structures as well as to identify the representative localized structures, a heuristic-based filtration algorithm will identify representative resonance structures on-the-fly. This approach was shown to correspond well to quantum calculations …”

“…For surface chemistry, atom types representing generic surface sites have been added, along with quadruple bonds for carbon and silicon. The new set of 16 nitrogen and 26 sulfur atom types has been previously reported, and the remainder of the new atom types are summarized in Table .…”

“…Given that RMG uses localized representations of molecules (Lewis structures), it is important that the algorithm can generate and identify the structures which are most representative of the true behavior of a molecule in terms of reactivity. Thus, significant improvements have been made to resonance structure generation algorithms, in particular for aromatic species and heteroatoms. , For aromatic species, RMG can now generate Clar structures , in replacement of Kekulé structures, which are now considered unrepresentative by RMG and not used for reaction generation. For heteroatom molecules with lone pairs, more delocalization pathways are now recognized by RMG.…”

RMG Database for Chemical Property Prediction

“…In general, the existence of a published, experimentally-validated RMG mechanism involving a specific chemistry means some care has been taken inside the RMG database with regard to those chemistries and that at least the most relevant corrections for those reaction conditions are present. For example, Grinberg Dana et al 2018 demonstrated RMG’s capabilities on nitrogen compounds, Class et al 2019 demonstrated RMG’s capabilities on some sulfur chemistries, and Chu et al 2019 demonstrated RMG’s capabilities for small aromatic species. ,, However, the user should keep in mind that all large databases, including the RMG database, likely include some erroneous values  perfection is hard to achieve, and it is challenging to check tens of thousands of numbers.…”

New Data Structure for Computational Molecular Design with Atomic or Fragment Resolution

“…The resulting MDS of ethene is given in Table S3. Note that it is known that the third-row elements may not follow the octet rule and therefore the present algorithm may overlook some of the possible chemicals involving the third-row elements.…”