He Xin, Lu Tian, Rong Chunying, Liu Wenjian, Ayers Paul W, Liu Shubin
Institute of Frontier Chemistry, School of Chemistry and Chemical Engineering, Shandong University, Qingdao, Shandong 266237, China.
Beijing Kein Research Center for Natural Sciences, Beijing 100022, China.
J Chem Theory Comput. 2024 Jul 23;20(14):6049-6061. doi: 10.1021/acs.jctc.4c00697. Epub 2024 Jul 12.
The Hohenberg-Kohn theorem of density functional theory (DFT) stipulates that energy is a universal functional of electron density in the ground state, so energy can be thought of having encoded essential information for the density. Based on this, we recently proposed to quantify energetic information within the framework of information-theoretic approach (ITA) of DFT (. , 157, 101103). In this study, we systematically apply energetic information to a variety of chemical phenomena to validate the use of energetic information as quantitative measures of physicochemical properties. To that end, we employed six ITA quantities such as Shannon entropy and Fisher information for five energetic densities, yielding twenty-six viable energetic information quantities. Then, they are applied to correlate with physicochemical properties of molecular systems, including chemical bonding, conformational stability, intermolecular interactions, acidity, aromaticity, cooperativity, electrophilicity, nucleophilicity, and reactivity. Our results show that different quantities of energetic information often behave differently for different properties but a few of them, such as Shannon entropy of the total kinetic energy density and information gain of the Pauli energy density, stand out and strongly correlate with several properties across different categories of molecular systems. These results suggest that they can be employed as quantitative measures of physicochemical properties. This work not only enriches the body of our knowledge about the relationship between energy and information, but also provides scores of newly introduced explicit density functionals to quantify physicochemical properties, which can serve as robust features for building machine learning models in future studies.
密度泛函理论（DFT）中的 Hohenberg-Kohn 定理规定，能量是基态电子密度的一个通用泛函，因此可以认为能量编码了密度的基本信息。基于此，我们最近提议在 DFT 的信息论方法（ITA）框架内量化能量信息（……，157，101103）。在本研究中，我们系统地将能量信息应用于各种化学现象，以验证能量信息作为物理化学性质定量度量的用途。为此，我们针对五种能量密度使用了诸如香农熵和费舍尔信息等六种 ITA 量，得到了二十六个可行的能量信息量。然后，将它们应用于与分子系统的物理化学性质相关联，包括化学键、构象稳定性、分子间相互作用、酸度、芳香性、协同性、亲电性、亲核性和反应性。我们的结果表明，不同的能量信息量对于不同的性质往往表现不同，但其中一些，例如总动能密度的香农熵和泡利能量密度的信息增益，表现突出，并与不同类别的分子系统中的几种性质强烈相关。这些结果表明它们可以用作物理化学性质的定量度量。这项工作不仅丰富了我们关于能量与信息之间关系的知识体系，还提供了大量新引入的显式密度泛函来量化物理化学性质，这些泛函可作为未来研究中构建机器学习模型的强大特征。
