High-throughput DFT calculations are key to screening existing/novel materials, sampling potential energy surfaces, and generating quantum mechanical data for machine learning. By including a fraction of exact exchange (EXX), hybrid functionals reduce the self-interaction error in semi-local DFT and furnish a more accurate description of the underlying electronic structure, albeit at a high computational cost that often prohibits such high-throughput applications. To address this challenge, we have constructed SeA (SeA=SCDM+exx+ACE), a robust, accurate, and efficient framework for high-throughput condensed-phase hybrid DFT in the PWSCF module of Quantum ESPRESSO (QE) by combining: (1) the non-iterative selected columns of the density matrix (SCDM) orbital localization scheme, (2) a black-box and linear-scaling EXX algorithm (exx), and (3) adaptively compressed exchange (ACE). Across a diverse set of non-equilibrium (H$_2$O)$_{64}$ configurations (with densities spanning 0.4 g/cm$^3$$-$1.7 g/cm$^3$), SeA yields a one$-$two order-of-magnitude speedup (~8X$-$26X) in the overall time-to-solution compared to PWSCF(ACE) in QE (~78X$-$247X speedup compared to the conventional EXX implementation) and yields energies, ionic forces, and other properties with high fidelity. As a proof-of-principle high-throughput application, we trained a deep neural network (DNN) potential for ambient liquid water at the hybrid DFT level using SeA via an actively learned data set with ~8,700 (H$_2$O)$_{64}$ configurations. Using an out-of-sample set of (H$_2$O)$_{512}$ configurations (at non-ambient conditions), we confirmed the accuracy of this SeA-trained potential and showcased the capabilities of SeA by computing the ground-truth ionic forces in this challenging system containing > 1,500 atoms.

翻译：高通量DFT计算是筛选现有/新型材料、采样势能面以及为机器学习生成量子力学数据的关键。通过包含部分精确交换（EXX），杂化泛函减少了半局域DFT中的自相互作用误差，并提供了更准确的电子结构描述，但其计算成本高昂，常阻碍此类高通量应用。为应对这一挑战，我们构建了SeA（SeA=SCDM+exx+ACE），这是一个在Quantum ESPRESSO（QE）的PWSCF模块中实现高通量凝聚相杂化DFT的鲁棒、准确且高效的框架，通过结合：（1）非迭代的密度矩阵选定列（SCDM）轨道局域化方案，（2）黑箱式线性标度EXX算法（exx），以及（3）自适应压缩交换（ACE）。在一组多样化的非平衡（H₂O）₆₄构型（密度范围从0.4 g/cm³到1.7 g/cm³）中，与QE中的PWSCF(ACE)相比，SeA在整体求解时间上实现了1-2个数量级的加速（约8倍至26倍），相较于传统EXX实现加速约78倍至247倍，并保持了能量、离子力及其他性质的高保真度。作为原理验证的高通量应用，我们利用SeA通过主动学习的数据集（包含约8,700个（H₂O）₆₄构型）在杂化DFT水平上训练了环境液态水的深度神经网络（DNN）势。使用（H₂O）₅₁₂构型（在非环境条件下）的样本外测试集，我们验证了该SeA训练势的准确性，并通过计算这一包含超过1,500个原子的挑战性系统中的真实离子力展示了SeA的能力。