In Search of the Best Low-Cost Methods for Efficient Screening of Conformers

Locating the lowest energy conformer is crucial for the
accurate
computation of equilibrium properties of molecular systems. This paper
examines the performance of efficient low-cost methods in terms of
the alignment and relative energies of their energy minima against
the benchmark revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ potential energy
surface. The low-cost methods considered include GFN-FF, GFN2-xTB,
DFTB3, HF-3c, B97-3c, PBEh-3c, and r2SCAN-3c composite
methods against a diverse test set of 20 compounds including alkanes,
perfluoroalkyl molecules, peptides, open-shell radicals, and Zn(II)
complexes of varying sizes. The “3c” composite methods
are generally more accurate, but are at least 2–3 orders of
magnitude more expensive than tight-binding methods which have energy
minima that align well with the benchmark potential energy surface.
The findings of this paper were further exploited to introduce a simple
strategy involving Grimme’s CENSO energy-sorting algorithm
that resulted in up to an order of magnitude reduction in computational
time for locating the lowest energy conformer on the revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ
surface.

定位最低能量构象对于准确计算分子体系的平衡性质至关重要。本文探讨了在能量最小值对齐和相对能量方面，一系列高效且低成本方法的性能，这些方法与基准revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ势能面进行了对比。所考虑的低成本方法包括GFN-FF、GFN2-xTB、DFTB3、HF-3c、B97-3c、PBEh-3c以及r2SCAN-3c复合方法，它们针对包括烷烃、全氟烷基分子、肽、开壳自由基以及不同尺寸的Zn(II)配合物等20种多样的测试集进行了评估。通常，“3c”复合方法在准确性方面更胜一筹，但其成本至少比紧束缚方法高两个至三个数量级，而紧束缚方法的能量最小值与基准势能面吻合良好。本文的研究成果进一步被利用，以引入一种简单策略，涉及Grimme的CENSO能量排序算法，该算法在revDSD-PBEP86-D4/def2-TZVPP//MP2/cc-pVTZ表面定位最低能量构象时，实现了计算时间至少降低一个数量级的成果。