A Better Path from Reactant to Product

Molecules in chemical reactions undergo changes in their three-dimensional structure along the reaction coordinate connecting reactant and product. Molecular geometry interpolation aims to construct high-quality reaction coordinates consisting of a smoothly varying curve in configuration space that connects reactant and product through chemically reasonable structures. Because geometry interpolation does not rely on potential energy surface information, it can characterize the potential energy surface of a reaction inexpensively, or it can provide high-quality initial guess structures for transition state optimization using density functional theory or another quantum chemical level of theory.

The main result in our method is a new computational methods called TRICS (translation-rotation internal coordinates with splicing). It builds on a well-established internal coordinate interpolation approach, which is prone to failures due to changes in the internal coordinate system along the reaction pathway. TRICS addresses this problem by performing interpolation in multiple patches of configuration space that are each parameterized by a separate internal coordinate system, followed by an iterative procedure that “splices” these path segments together to yield a single smooth pathway. 

The performance of TRICS is compared against other recently published interpolation methods on large datasets of reactions. One area where TRICS stands out is the capability to predict barrierless reaction pathways, which often occur in photoreactions or radical reactions. Barrierless reactions do not have a transition state and can proceed in an ultrafast manner; therefore, identifying barrierless reactions is very important for accurate modeling of reaction kinetics. In all of our test cases, TRICS can generate smooth and barrierless reactions while other interpolation methods predicted pathways with spurious barriers. The results show that TRICS can be a very useful tool for characterizing reactive potential energy surfaces, especially in reactions where a single transition state cannot be found. 

For more information, please read the publication at doi.org/10.1063/5.0353998. 

The code is freely available at https://www.github.org/leeping/geomeTRIC.