colabfit/water_ice_JCP_2020

water ice JCP 2020数据集从单个参考的从头计算模拟出发，利用主动学习方法扩展到新的状态点，并描述了核的量子性质。该数据集在814个参考计算的基础上训练的最终模型，在各种条件下（包括常温常压下的液态水、不同的冰相以及空气-水界面）均取得了优秀的结果，并且考虑了核量子效应。数据集包含了8000个独特的分子配置，2088000个原子，元素包括氢和氧，性质包括能量、原子力和Cauchy应力。

The water ice JCP 2020 dataset starts from a single reference ab initio simulation, uses active learning to expand to new state points, and describes the quantum nature of the nuclei. The final model, trained on 814 reference calculations, yields excellent results under a range of conditions, including liquid water at ambient and elevated temperatures and pressures, different phases of ice, and the air-water interface, all incorporating nuclear quantum effects. The dataset contains 8000 unique molecular configurations, 2088000 atoms, elements include Hydrogen and Oxygen, and properties include energy, atomic forces, and Cauchy stress.