Automatized Parametrization of SCC-DFTB Repulsive Potentials: Application to Hydrocarbons

In this work, we derive and test a new automatized strategy to construct repulsive potentials for the self-consistent charge density functional tight-binding (SCC-DFTB) method. This approach allows one to explore the parameter space in a systematic fashion in order to find optimal solutions. We find that due to the limited flexibility of the SCC-DFTB electronic part, not all properties can be optimized simultaneously. For example, the optimization of heats of formation is in conflict with the optimization of vibrational frequencies. Therefore, a special parametrization for vibrational frequencies is derived. It is shown that the performance of SCC-DFTB can be significantly improved using a more elaborate fitting strategy. A new fit for C and H is presented, which results in an average error of 2.6 kcal/mol for heats of formations for a large set of hydrocarbons, indicating that the performance of SCC-DFTB can be systematically improved also for other elements.

本研究推导并测试了一种全新的自动化策略，用于为自洽电荷密度泛函紧束缚（self-consistent charge density functional tight-binding, SCC-DFTB）方法构建排斥势。该策略可通过系统化的方式探索参数空间以获取最优解。研究发现，由于SCC-DFTB电子部分的灵活性有限，无法同时优化所有目标性质：例如，生成焓的优化与振动频率的优化存在冲突。为此，我们针对振动频率推导了专用参数化方案。研究表明，采用更为精细的拟合策略可显著提升SCC-DFTB的计算性能。本文针对碳（C）与氢（H）元素提出了全新的拟合方案，该方案在大型烃类数据集上的生成焓平均误差仅为2.6 kcal/mol，这说明SCC-DFTB的性能亦可通过系统化优化的方式在其他元素体系中得到提升。