Expanding density-correlation machine learning representations for anisotropic coarse-grained particles

This record contains three datasets and the scripts used to generate figures in "Expanding density-correlation machine learning representations for anisotropic coarse-grained particles." This paper explores the theory and implementation of machine-learning descriptors for ellipsoidal bodies, extending the popular "Smooth Overlap of Atomic Positions" (SOAP) formalism. These case studies serve to demonstrate the different use cases of this technology. The three datasets are:
- Generated configurations of nematic and smectic liquid crystal systems, with a range of orientational order, characterized by the nematic order parameter
- Dimers of (1, 1.5, 2) ellipsoids at different interaction cutoffs and rotations, with computed Gay-Berne type energies
- Crystalline configurations of planar benzene molecules, with energetics computed using QuantumEspresso v7.046 using Perdew–Burke–Ernzerhof (PBE) pseudopotentials and cutoff parameters reported by Prandini et al., Grimme D3-dispersion correction, and a 3 × 3 Monkhorst–Pack k-point grid.

本记录包含三个数据集及其生成“扩展各向异性粗粒粒子的密度-相关性机器学习表示”一文中图表的脚本。该论文探讨了椭圆体机器学习描述符的理论与实现，扩展了广受欢迎的“原子位置平滑重叠”(SOAP)形式主义。这些案例研究旨在展示该技术的不同应用场景。三个数据集分别为：
- 生成的各向异性液晶系统的液晶和向列相配置，其取向有序度范围广泛，以向列相有序参数为特征
- 在不同相互作用截止值和旋转下的(1, 1.5, 2)椭圆体二聚体，并计算了Gay-Berne类型的能量
- 使用QuantumEspresso v7.046计算平面苯分子的晶体结构，能量计算采用Perdew–Burke–Ernzerhof (PBE)伪势和由Prandini等人报告的截止参数，Grimme D3分散校正以及3 × 3 Monkhorst–Pack k点网格。