Suppr超能文献

通过将轮廓变形方法与解析延拓相结合来加速芯能级计算

Accelerating Core-Level Calculations by Combining the Contour Deformation Approach with the Analytic Continuation of .

作者信息

Panadés-Barrueta Ramón L, Golze Dorothea

机构信息

Faculty of Chemistry and Food Chemistry, Technische Universität Dresden, Dresden 01062, Germany.

出版信息

J Chem Theory Comput. 2023 Aug 22;19(16):5450-5464. doi: 10.1021/acs.jctc.3c00555. Epub 2023 Aug 11.

Abstract

In recent years, the method has emerged as a reliable tool for computing core-level binding energies. The contour deformation (CD) technique has been established as an efficient, scalable, and numerically stable approach to compute the self-energy for deep core excitations. However, core-level calculations with CD face the challenge of higher scaling with respect to system size compared to the conventional quartic scaling in valence-state algorithms. In this work, we present the CD-WAC method [CD with analytic continuation (AC)], which reduces the scaling of CD applied to the inner shells from O(N) to O(N) by employing an AC of the screened Coulomb interaction . Our proposed method retains the numerical accuracy of CD for the computationally challenging deep core case, yielding mean absolute errors <5 meV for well-established benchmark sets, such as CORE65, for single-shot calculations. More extensive testing for different flavors proves the reliability of the method. We have confirmed the theoretical scaling by performing scaling experiments on large acene chains and amorphous carbon clusters, achieving speedups of up to 10× for structures of only 116 atoms. This improvement in computational efficiency paves the way for more accurate and efficient core-level calculations on larger and more complex systems.

摘要

近年来,该方法已成为计算芯能级结合能的可靠工具。轮廓变形(CD)技术已被确立为一种计算深芯激发自能的高效、可扩展且数值稳定的方法。然而,与价态算法中传统的四次方缩放相比,使用CD进行芯能级计算面临着随系统大小更高缩放比例的挑战。在这项工作中,我们提出了CD-WAC方法[带解析延拓(AC)的CD],通过对屏蔽库仑相互作用进行解析延拓,将应用于内壳层的CD缩放比例从O(N)降低到O(N)。我们提出的方法在计算具有挑战性的深芯情况时保持了CD的数值精度,对于诸如CORE65等成熟的基准集,单次计算的平均绝对误差<5 meV。对不同情况进行的更广泛测试证明了该方法的可靠性。我们通过对大的并苯链和非晶碳簇进行缩放实验,证实了理论缩放比例,对于仅116个原子的结构实现了高达10倍的加速。计算效率的这种提高为在更大、更复杂的系统上进行更准确、高效的芯能级计算铺平了道路。

https://cdn.ncbi.nlm.nih.gov/pmc/blobs/324a/10448726/5997f37ce68b/ct3c00555_0002.jpg

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍。

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索

文档翻译

学术文献翻译模型,支持多种主流文档格式。

立即体验