机器学习时代经验能量函数的系统性改进
Systematic Improvement of Empirical Energy Functions in the Era of Machine Learning
AI总结：
本研究针对纯水、二氯甲烷及溶剂化离子体系，系统评估经验力场中静电、极化等分项替换的效果，量化各分项对力场性能的提升，为机器学习拟合的力场参数化提供参考。
摘要：
本文针对纯水、二氯甲烷（DCM）以及溶剂化的K⁺和Cl⁻离子，定量评估了经验力场中各分项的定向替换所产生的影响。在静电作用方面，评估了点电荷（PCs）、拟合分子静电势的基于机器学习（ML）的最小分布电荷（MDCM），以及基于库仑积分的静电方法。通过向现有力场（此处为CHARMM）添加由碎片分子轨道（FMO）推导的极化能，研究了显式包含二阶项的影响。研究表明，各向异性静电作用可降低水的均方根误差（RMSE）1.6 kcal/mol、DCM的RMSE 0.8 kcal/mol，以及溶剂化Cl⁻团簇的RMSE 0.4 kcal/mol。额外的极化项对DCM可忽略不计，但能显著改善纯水（误差降低1.1 kcal/mol）和Cl⁻团簇（误差降低0.4 kcal/mol）的误差，并且是描述溶剂化K⁺的关键，可使其RMSE降低2.3 kcal/mol。研究发现，12-6 Lennard-Jones函数形式与PC和MDCM静电方法结合时表现令人满意，但不适用于考虑静电穿透能的描述。通过将严格的两体方法与自洽参考数据进行比较，评估了多体贡献的重要性：DCM可用两体势很好地近似，而水以及溶剂化的K⁺和Cl⁻离子需要显式的多体校正。本工作系统量化了哪些分项能够提升现有力场的性能，以及应使用何种参考数据来以易处理的方式参数化这些分项，从而用于纯体系和多相体系的机器学习拟合。
英文摘要：
The impact of targeted replacement of individual terms in empirical force fields is quantitatively assessed for pure water, dichloromethane (DCM), and solvated K$^+$ and Cl$^-$ ions. For the electrostatics, point charges (PCs) and machine learning (ML)based minimally distributed charges (MDCM) fitted to the molecular electrostatic potential are evaluated together with electrostatics based on the Coulomb integral. The impact of explicitly including second-order terms is investigated by adding a fragment molecular orbital (FMO)-derived polarization energy to an existing force field, in this case CHARMM. It is demonstrated that anisotropic electrostatics reduce the RMSE for water (by 1.6 kcal/mol), DCM (by 0.8 kcal/mol) and for solvated Cl$^-$ clusters (by 0.4 kcal/mol). An additional polarization term can be neglected for DCM but notably improves errors in pure water (by 1.1 kcal/mol) and in Cl$^-$ clusters (by 0.4 kcal/mol) and is key to describing solvated K$^+$, reducing the RMSE by 2.3 kcal/mol. A 12-6 Lennard-Jones functional form is found to perform satisfactorily with PC and MDCM electrostatics, but is not appropriate for descriptions that account for the electrostatic penetration energy. The importance of many-body contributions is assessed by comparing a strictly 2-body approach with self-consistent reference data. DCM can be approximated well with a 2-body potential while water and solvated K$^+$ and Cl$^-$ ions require explicit many-body corrections. The present work systematically quantifies which terms improve the performance of an existing force field and what reference data to use for parametrizing these terms in a tractable fashion for ML fitting of pure and heterogeneous systems.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
