Simulating warm dense matter that undergoes a wide range of temperatures and densities is challenging. Predictive theoretical models, such as quantum-mechanics-based first-principles molecular dynamics (FPMD), require a huge amount of computational resources. Herein, we propose a deep learning based scheme, called electron temperature dependent deep potential molecular dynamics (TDDPMD), for efficiently simulating warm dense matter with the accuracy of FPMD. The TDDPMD simulation is several orders of magnitudes faster than FPMD, and, unlike FPMD, its efficiency is not affected by the electron temperature. We apply the TDDPMD scheme to beryllium (Be) in a wide range of temperatures (0.4 to 2500 eV) and densities (3.50 to 8.25 g/cm$^3$). Our results demonstrate that the TDDPMD method not only accurately reproduces the structural properties of Be along the principal Hugoniot curve at the FPMD level, but also yields even more reliable diffusion coefficients than typical FPMD simulations due to its ability to simulate larger systems with longer time.

中文翻译：

---

通过电子温度相关的深位势分子动力学模拟暖致密物质

模拟经历各种温度和密度的温暖致密物质具有挑战性。预测性理论模型，例如基于量子力学的第一性原理分子动力学 (FPMD)，需要大量的计算资源。在此，我们提出了一种基于深度学习的方案，称为电子温度相关的深位势分子动力学（TDDPMD），用于以 FPMD 的精度有效模拟温暖的致密物质。TDDPMD 模拟比 FPMD 快几个数量级，并且与 FPMD 不同，其效率不受电子温度的影响。我们将 TDDPMD 方案应用于各种温度（0.4 至 2500 eV）和密度（3.50 至 8.25 g/cm$^3$）的铍 (Be)。