Abstract Using an automated optimization scheme, we generate a gallium projector augmented wave (PAW) data set with the 3d electrons treated as core electrons, in contrast to typical treatment of the 3d electrons as valence electrons. Treating the 3d electrons as core electrons reduces calculation sizes, enabling faster defect studies of GaN. We show that the results of defect calculations in GaN are not adversely affected by freezing the 3d electrons to create the new PAW. For all three point defects tested (gallium vacancy, gallium interstitial in the tetrahedral site, and gallium interstitial in the octohedral site), the defect energy levels do not deviate by more than 0.25 eV from results with 3d electrons treated as valence. We show speedups in the defect calculations using the new PAW from 1.3 to 2.9, with increasing speedup for larger supercell size. These speedups would translate to similar speedups in Born-Oppenheimer molecular dynamics studies of GaN defects. In addition, we share some insight into the development of the optimized PAW including the use of the arctangent of the logderivatives—a superior accuracy metric for automated optimization of logderivatives/scattering properties.

中文翻译：

---

使用平面波 DFT 更快地计算 GaN 中的缺陷的低价镓 PAW

摘要使用自动优化方案，我们生成了一个镓投影增强波 (PAW) 数据集，其中 3d 电子被视为核心电子，这与将 3d 电子作为价电子的典型处理形成对比。将 3d 电子视为核心电子可减少计算量，从而加快对 GaN 的缺陷研究。我们表明，冻结 3d 电子以创建新的 PAW 不会对 GaN 中的缺陷计算结果产生不利影响。对于测试的所有三个点缺陷（镓空位、四面体位置的镓间隙和八面体位置的镓间隙），缺陷能级与将 3d 电子视为价态的结果偏差不超过 0.25 eV。我们使用新的 PAW 从 1.3 到 2.9 显示了缺陷计算的加速，随着更大的超胞尺寸的加速。这些加速将转化为 GaN 缺陷的 Born-Oppenheimer 分子动力学研究中的类似加速。此外，我们分享了对优化 PAW 开发的一些见解，包括使用对数导数的反正切——一种用于自动优化对数导数/散射特性的卓越准确度度量。