近似密度泛函下固体极限中的分段线性偏差
Deviations from piecewise linearity in the solid-state limit with approximate density functionals
摘要:
在精确的密度泛函理论(DFT)中,基态总能量是整数电子点之间的一系列线性段,这种情况称为“分段线性”。偏离该条件表明对电子结构的预测能力差,特别是对电离能、基本能隙和电荷转移的预测能力差。在这篇文章中,我们重新审视了线性偏差(即,通过考虑两种不同的接近方式,在固态极限中的曲率):增加尺寸的大型有限系统和由具有周期性边界条件的越来越大的参考单元表示的晶体。我们表明,曲率接近消失值在这两个限制,即使是泛函产生差的电子结构的预测,因此不能被用作固体中的诊断或建设性的工具。我们发现,接近零曲率是不同的,在每一个的两个限制,由于存在一个补偿的背景电荷在周期性的情况下。基于这些研究结果,我们提出了一个新的标准的功能建设和评价,来自曲率的大小依赖性,以及一个实用的方法来评估这个标准。对于大的有限系统,我们进一步表明,曲率是由最高占据本征态的自相互作用。这些发现说明了各种固体,半导体纳米晶体和长链烷烃的计算研究。
英文摘要:
In exact density functional theory (DFT) the total ground-state energy is a series of linear segments between integer electron points, a condition known as "piecewise linearity". Deviation from this condition is indicative of poor predictive capabilities for electronic structure, in particular of ionization energies, fundamental gaps, and charge transfer. In this article, we take a new look at the deviation from linearity (i.e., curvature) in the solid-state limit by considering two different ways of approaching it: a large finite system of increasing size and a crystal represented by an increasingly large reference cell with periodic boundary conditions. We show that the curvature approaches vanishing values in both limits, even for functionals which yield poor predictions of electronic structure, and therefore can not be used as a diagnostic or constructive tool in solids. We find that the approach towards zero curvature is different in each of the two limits, owing to the presence of a compensating background charge in the periodic case. Based on these findings, we present a new criterion for functional construction and evaluation, derived from the size-dependence of the curvature, along with a practical method for evaluating this criterion. For large finite systems we further show that the curvature is dominated by the self-interaction of the highest occupied eigenstate. These findings are illustrated by computational studies of various solids, semiconductor nanocrystals, and long alkane chains.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。