Scholarlink

<p>The Interacting Quantum Atoms (IQA) method offers a route to calculate, directly from the two-particle-density-matrix, electron correlation energies for a quantum topological atom interacting with itself and with other atoms in a given system. Here, we corroborate the energetic transferability, at CCSD(T) level, of a variety of functional groups as they appear in both small organic molecules and water-containing complexes: methyl, amine, hydroxyl, water, carbonyl, carboxyl, amide, methylene and triple bonds. The dependence of these atomic correlation energies upon variation of a dihedral angle is very small (0.1&#xa0;kJ/mol), and that upon variation in bond length and angle reasonably small (units of kJ/mol for molecular dynamics-induced changes and tens of kJ/mol for extreme, manual changes). Overall, these findings give the green light for the machine learning of these physically-informed atomic properties. The beneficiary is the force field FFLUX, which will then be able to make predictions on dispersion effects, totally in line with the other types of IQA energy contributions that underpin it. Hence the familiar dispersion schemes that are typically bolted on become unnecessary.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On the transferability of pure electron correlation of quantum topologically defined molecular fragments at the CCSD(T) level

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月21日

查看全文
  • Mark A. Vincent,
  • Paul L. A. Popelier

摘要

The Interacting Quantum Atoms (IQA) method offers a route to calculate, directly from the two-particle-density-matrix, electron correlation energies for a quantum topological atom interacting with itself and with other atoms in a given system. Here, we corroborate the energetic transferability, at CCSD(T) level, of a variety of functional groups as they appear in both small organic molecules and water-containing complexes: methyl, amine, hydroxyl, water, carbonyl, carboxyl, amide, methylene and triple bonds. The dependence of these atomic correlation energies upon variation of a dihedral angle is very small (0.1 kJ/mol), and that upon variation in bond length and angle reasonably small (units of kJ/mol for molecular dynamics-induced changes and tens of kJ/mol for extreme, manual changes). Overall, these findings give the green light for the machine learning of these physically-informed atomic properties. The beneficiary is the force field FFLUX, which will then be able to make predictions on dispersion effects, totally in line with the other types of IQA energy contributions that underpin it. Hence the familiar dispersion schemes that are typically bolted on become unnecessary.