Skelton Jonathan M, Gunn David S D, Metz Sebastian, Parker Stephen C
Department of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom.
Department of Chemistry, University of Bath, Bath BA2 7AY, United Kingdom.
J Chem Theory Comput. 2020 Jun 9;16(6):3543-3557. doi: 10.1021/acs.jctc.9b01218. Epub 2020 May 20.
Accurately modeling the electronic structure of materials is a persistent challenge to high-throughput screening. A promising means of balancing accuracy against computational cost is non-self-consistent calculations with hybrid density-functional theory, where the electronic band energies are evaluated using a hybrid functional from orbitals obtained with a less demanding (semi)local functional. We have quantified the performance of this technique for predicting the physical properties of 16 tetrahedral semiconductors with bandgaps from 0.2 to 5.5 eV. Provided the base functional predicts a nonmetallic electronic structure, bandgaps within 5% of the PBE0 and HSE06 gaps can be obtained with an order of magnitude reduction in computing time. The positions of the valence and conduction band extrema and the Fermi level are well reproduced, enabling calculation of the band dispersion, density of states, and dielectric properties using Fermi's Golden Rule. While the error in the non-self-consistent total energies is ∼50 meV atom, the energy-volume curves are reproduced accurately enough to obtain the equilibrium volume and bulk modulus with minimal error. We also test the dielectric-dependent scPBE0 functional and obtain bandgaps and dielectric constants to within 2.5% of the self-consistent results, which amounts to a significant improvement over self-consistent PBE0 for a similar computational cost. We identify cases where the non-self-consistent approach is expected to perform poorly and demonstrate that partial self-consistency provides a practical and efficient workaround. Finally, we perform proof-of-concept calculations on CoO and NiO to demonstrate the applicability of the technique to strongly correlated open-shell transition-metal oxides.
准确地对材料的电子结构进行建模一直是高通量筛选面临的挑战。在计算成本与准确性之间取得平衡的一种有前景的方法是使用混合密度泛函理论进行非自洽计算,即在使用要求较低的(半)局域泛函获得的轨道基础上,利用混合泛函来评估电子能带能量。我们已经量化了该技术在预测16种带隙从0.2到5.5电子伏特的四面体半导体物理性质方面的性能。只要基础泛函预测出非金属电子结构,就能在计算时间减少一个数量级的情况下,获得与PBE0和HSE06带隙相差在5%以内的带隙。价带和导带极值的位置以及费米能级都能得到很好的再现,从而可以使用费米黄金定则来计算能带色散、态密度和介电性质。虽然非自洽总能量的误差约为每原子50毫电子伏特,但能量-体积曲线的再现精度足以在误差最小的情况下获得平衡体积和体模量。我们还测试了依赖于介电常数的scPBE0泛函,并获得了与自洽结果相差在2.5%以内的带隙和介电常数,这在相似的计算成本下相较于自洽的PBE0有了显著改进。我们确定了非自洽方法预期表现不佳的情况,并证明部分自洽提供了一种实用且高效的解决方法。最后,我们对CoO和NiO进行了概念验证计算,以证明该技术对强关联的开壳层过渡金属氧化物的适用性。