AI 中文总结
该研究提出一种基于物理的数据驱动方法，通过多极展开与多元切比雪夫多项式近似，实现带电物体各向异性成对相互作用的高效建模，在芳香族分子上验证了其适用性，可用于带电材料的模拟。
AI 中文摘要
我们提出了一种基于物理的 data-driven 方法，用于建模存在长程静电作用时的各向异性成对相互作用。该方法将总相互作用分为两部分：长程静电相互作用采用截断至偶极级的 multipole expansion（多极展开）近似；短程残余相互作用采用多元 Chebyshev 多项式近似，该多项式拟合自有限数量构型的测量值。我们在一系列芳香族分子（苯、苯甲腈和苯氧负离子）上评估了该方法，发现其在短程相互作用采用适中截断距离时可产生满意结果。该方法可应用于带电合成材料与生物材料的复杂相互作用建模及动态模拟。
英文摘要
We formulate a physics-informed data-driven method for modeling anisotropic pairwise interactions in the presence of long-ranged electrostatics. The method separates the total interaction into a long-ranged electrostatic interaction that is approximated using a multipole expansion truncated at the dipole level and a short-ranged residual interaction that is approximated using multivariate Chebyshev polynomials fit to measurements from a limited number of configurations. We assess the approach on a sequence of aromatic molecules (benzene, benzonitrile, and phenoxide), finding that it produces satisfactory results using a modest cutoff distance for the short-ranged interaction. This method has applications for modeling complex interactions for, and conducting dynamic simulations of, synthetic and biological materials with charge.
