Interdisciplinary Center for Scientific Computing, Ruprecht-Karls University, Im Neuenheimer Feld 205, 69120 Heidelberg, Germany.
Fysikum, Stockholm University, Albanova, 10691 Stockholm, Sweden.
J Chem Theory Comput. 2021 Mar 9;17(3):1618-1637. doi: 10.1021/acs.jctc.0c01082. Epub 2021 Feb 5.
The performance of several standard and popular approaches for calculating X-ray absorption spectra at the carbon, nitrogen, and oxygen K-edges of 40 primarily organic molecules up to the size of guanine has been evaluated, focusing on the low-energy and intense 1s → π* transitions. Using results obtained with CVS-ADC(2)-x and fc-CVS-EOM-CCSD as benchmark references, we investigate the performance of CC2, ADC(2), ADC(3/2), and commonly adopted density functional theory (DFT)-based approaches. Here, focus is on rather than on of transition energies and intensities-in other words, we target relative energies and intensities and the spread thereof, rather than absolute values. The use of exchange-correlation functionals tailored for time-dependent DFT calculations of core excitations leads to error spreads similar to those seen for more standard functionals, despite yielding superior absolute energies. Long-range corrected functionals are shown to perform particularly well compared to our reference data, showing error spreads in energy and intensity of 0.2-0.3 eV and ∼10%, respectively, as compared to 0.3-0.6 eV and ∼20% for a typical pure hybrid. In comparing intensities, state mixing can complicate matters, and techniques to avoid this issue are discussed. Furthermore, the influence of basis sets in high-level calculations is investigated, showing that reasonably accurate results are obtained with the use of 6-311++G**. We name this benchmark suite as XABOOM (X-ray absorption benchmark of organic molecules) and provide molecular structures and ground-state self-consistent field energies and spectroscopic data. We believe that it provides a good assessment of electronic structure theory methods for calculating X-ray absorption spectra and will become useful for future developments in this field.
评估了几种标准和流行的方法在计算 40 种主要有机分子的碳、氮和氧 K 边 X 射线吸收光谱的性能,重点关注低能和强的 1s→π*跃迁。使用 CVS-ADC(2)-x 和 fc-CVS-EOM-CCSD 的结果作为基准参考,我们研究了 CC2、ADC(2)、ADC(3/2)和常用的基于密度泛函理论(DFT)的方法的性能。这里,重点是过渡能量和强度的相对值,换句话说,我们针对的是相对能量和强度及其分布,而不是绝对值。用于核心激发的时间相关 DFT 计算的交换相关泛函的使用导致误差分布与更标准的泛函相似,尽管绝对能量更好。与我们的参考数据相比,长程校正泛函表现特别好,在能量和强度方面的误差分布分别为 0.2-0.3 eV 和 10%左右,而典型的纯混合泛函的误差分布分别为 0.3-0.6 eV 和 20%左右。在比较强度时,态混合可能会使问题复杂化,并且讨论了避免此问题的技术。此外,还研究了高水平计算中基组的影响,结果表明,使用 6-311++G**可以获得相当准确的结果。我们将此基准套件命名为 XABOOM(有机分子的 X 射线吸收基准),并提供分子结构、基态自洽场能量和光谱数据。我们相信,它为计算 X 射线吸收光谱的电子结构理论方法提供了很好的评估,并将为该领域的未来发展提供有用的参考。