ALT This work presents a novel first-principles approach for simulating electrostatic forces in polar materials, a crucial component often lacking in machine-learning interatomic potentials. Long-range electrostatic interactions play a fundamental role and, although machine learning force fields have rapidly improved, their description of the atomic environment is local, thus hampering their applicability to real-world problems. Few methods have been proposed to address this issue, and all of them require extensive retraining of the force field. The approach here overcomes these limitations: it is derived from first principles and eliminates the need for new ab-initio simulations on the training set. Furthermore, it can be easily implemented alongside existing and available force fields. The method is released as an open-source package that can be readily employed in simulations. It is highly versatile and user friendly.

ALT EDITORS' SUGGESTION Nonlocal machine-learned exchange functional for molecules and solids Here, the authors use machine learning and exact physical constraints to design a nonlocal exchange functional for both molecular and solid-state systems. The model is computationally efficient, achieves hybrid-DFT accuracy on molecular benchmarks, and improves the accuracy of band gap predictions over semilocal DFT. To demonstrate the efficiency and accuracy of the model, the authors compute charged point defect transition levels in silicon in good agreement with experiment, a task previously only possible with more expensive hybrid DFT calculations. Kyle Bystrom and Boris Kozinsky Phys. Rev. B 110, 075130 (2024)

ALT TOC image from the article