Transition-state (TS) searches are central to computational studies of chemical reactions, yet advanced methods often require substantial effort to integrate into routine workflows. Consequently, users tend to rely on familiar software and established input formats. Here, we present dmf-g16, a Gaussian-specific front end to the Direct MaxFlux (DMF) reaction-path optimization method implemented in PyDMF. dmf-g16 enables DMF-based TS searches with minimal workflow changes: users simply replace the Gaussian executable with dmf-g16, while native QST2/QST3 input files remain unchanged. For QST inputs, DMF performs explicit path optimization using Gaussian as an external energy calculator, followed by TS refinement in Gaussian from the highest-energy path point. Benchmarks on 121 reactions show a substantial improvement in reliability over Gaussian QST2, increasing the success rate from 31.4% to 93.4%. Although path optimization adds computational cost, wall-clock time is typically only a few times that of QST2 and can be reduced through parallel energy evaluation.
Koda et al. (Mon,) studied this question.