Department of Nuclear Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, South Korea.
J Chem Phys. 2018 Apr 14;148(14):144501. doi: 10.1063/1.5017198.
The performance of exchange-correlation functionals in density-functional theory (DFT) calculations for liquid metal has not been sufficiently examined. In the present study, benchmark tests of Perdew-Burke-Ernzerhof (PBE), Armiento-Mattsson 2005 (AM05), PBE re-parameterized for solids, and local density approximation (LDA) functionals are conducted for liquid sodium. The pair correlation function, equilibrium atomic volume, bulk modulus, and relative enthalpy are evaluated at 600 K and 1000 K. Compared with the available experimental data, the errors range from -11.2% to 0.0% for the atomic volume, from -5.2% to 22.0% for the bulk modulus, and from -3.5% to 2.5% for the relative enthalpy depending on the DFT functional. The generalized gradient approximation functionals are superior to the LDA functional, and the PBE and AM05 functionals exhibit the best performance. In addition, we assess whether the error tendency in liquid simulations is comparable to that in solid simulations, which would suggest that the atomic volume and relative enthalpy performances are comparable between solid and liquid states but that the bulk modulus performance is not. These benchmark test results indicate that the results of liquid simulations are significantly dependent on the exchange-correlation functional and that the DFT functional performance in solid simulations can be used to roughly estimate the performance in liquid simulations.
在密度泛函理论(DFT)计算中,交换关联泛函在液态金属中的性能尚未得到充分检验。在本研究中,对 Perdew-Burke-Ernzerhof(PBE)、Armiento-Mattsson 2005(AM05)、重新参数化的 PBE 用于固体以及局域密度近似(LDA)泛函进行了液态钠的基准测试。在 600 K 和 1000 K 下评估了对关联函数、平衡原子体积、体弹模量和相对焓。与可用的实验数据相比,原子体积的误差范围为-11.2%至 0.0%,体弹模量的误差范围为-5.2%至 22.0%,相对焓的误差范围为-3.5%至 2.5%,具体取决于 DFT 泛函。广义梯度近似泛函优于 LDA 泛函,而 PBE 和 AM05 泛函表现出最佳性能。此外,我们评估了液体模拟中的误差趋势是否与固体模拟中的误差趋势相当,这表明在固体和液体状态下,原子体积和相对焓的性能相当,但体弹模量的性能则不然。这些基准测试结果表明,液体模拟的结果高度依赖于交换关联泛函,并且固体模拟中的 DFT 泛函性能可以大致估计液体模拟中的性能。