Comparative study of Minnesota functionals performance on ferroelectric BaTiO3 and PbTiO3
Comparative study of Minnesota functionals performance on ferroelectric BaTiO3 and PbTiO3
复制标题
DOI：
10.1103/physrevmaterials.4.073802
复制
发表时间：
2020-07
影响因子：
3.4
通讯作者：
Maggie L. Kingsland;K. A. Lynch;S. Lisenkov;Xiao He;I. Ponomareva
中科院分区：
文献类型：
--
作者：
Maggie L. Kingsland;K. A. Lynch;S. Lisenkov;Xiao He;I. Ponomareva
基于密度泛函理论的模拟是铁电材料的计算研究和经典势能的参数化的主要工具。然而，预测往往强烈依赖于交换关联泛函。最流行的选择，LDA和GGA，往往会低估或高估一些结构、电气和能源属性。这些阻碍了将第一性原理模拟的范围扩展到有限温度和真实大小的高精度经典势的发展。在这项工作中，我们研究了最近发展起来的明尼苏达交换关联泛函在典型铁电材料${\mathm{Batio}}_{3}$和${\mathm{PbTiO}}_{3}$上的性能，并与一些流行的泛函进行了比较。我们发现，不同泛函对某些性质(四方性、位相能量差和极化)的预测之间存在很强的相关性。沿着相关线，我们发现了一系列泛函(包括来自明尼苏达套件的一些泛函)，它们的预测介于LDA和GGA之间，因此提供了改进。提出了一种相对于所选基准对功能性能进行相对排序的方法，并将其应用于确定${\mathm{Batio}}_{3}$和${\mathm{PbTio}}_{3}$的性能最好者。研究发现，这种性能与材料有关。因此，我们建议将这项工作中进行的性能评估用于在铁电材料上使用之前的“筛选”功能。速度快，计算成本低，这很可能导致改进的描述，特别是对于经典的势参数。明尼苏达州的功能套件被发现适合于这项任务。
Density-functional-theory-based simulations are the leading tool for the computational investigation of ferroelectrics and the parametrization of their classical potentials. However, the predictions often depend strongly on the exchange-correlation functional. The most popular choices, the LDA and GGA, tend to underestimate or overestimate some structural, electric, and energy properties. These impede the development of highly accurate classical potentials that extend the reach of first-principles simulations to finite temperatures and realistic sizes. In this work, we investigate the performance of recently developed Minnesota exchange-correlation functionals on the prototypical ferroelectrics ${\mathrm{BaTiO}}_{3}$ and ${\mathrm{PbTiO}}_{3}$ in comparison with some popular ones. We find that there exists a strong correlation between predictions for some properties (tetragonality, phase energy difference, and polarization) by different functionals. Along the correlation line, we find a range of functionals (including some from the Minnesota suite) whose predictions fall between those of the LDA and those of the GGA and, therefore, offer an improvement. A way to relatively rank functional performance with respect to chosen benchmarks is proposed and applied to identify the top performers for ${\mathrm{BaTiO}}_{3}$ and ${\mathrm{PbTiO}}_{3}$. The performance is found to be material dependent. Therefore, we propose that the performance assessment carried out in this work is employed to ``screen'' functionals prior to their use on ferroelectrics. Quick and computationally inexpensive, it is likely to lead to improved descriptions, especially for classical potential parametrization. The Minnesota functional suite is found to be suitable for this task.