Local-Softening Stochastic Surface Walking for Fast\nExploration of Corrugated Potential Energy Surfaces

Global potential energy surface (PES)\nexploration provides a unique\nroute to predict the thermodynamic and kinetic properties of unknown\nmaterials, but the task is highly challenging for systems with tight\ncovalent bonds. Here, we develop the local-softening stochastic surface\nwalking (LS-SSW) method for scanning corrugated PESs. LS-SSW transforms\nthe vibrational mode space of a system by adding pairwise penalty\npotentials with a self-adaption mechanism, which helps to delocalize\nand soften the strong local modes. This allows the stochastic surface\nwalking (SSW) method to capture more efficiently the correct local\natomic movement toward nearby minima and simultaneously reduce the\nbarrier height of reactions. As a result, the local trapping time\nin searching for the corrugated PES is greatly reduced. LS-SSW can\nbe applied generally to the reaction pathway sampling and the global\nPES exploration of both clusters and crystals, the high efficiency\nof which is demonstrated in searching the reaction pathways between\nC<sub>4</sub>H<sub>6</sub> isomers, finding the global minimum of\ncarbon clusters up to 360 atoms, and constructing the global PES of\nFe<sub>7</sub>C<sub>3</sub> material.