Lithium-based molten salts have attracted significant attention due to their applications in energy storage, advanced fission reactors and fusion devices. Lithium fluorides and particularly 66.6%LiF-33.3¾F2 (Flibe) are of considerable interest in nuclear systems, as they show an excellent combination of desirable heat-transfer and neutron-absorption characteristics. For nuclear salts, the range of possible local structures, compositions, and thermodynamic conditions presents significant challenges in atomistic modeling. In this work, we demonstrate that atom-centered neural network interatomic potentials (NNIP) provide a fast and accurate method for performing molecular dynamics of molten salts. For LiF, these potentials are able to accurately model dimer interactions, crystalline solids under deformation, semi-crystalline LiF near the melting point and liquid LiF at high temperatures. For Flibe, NNIPs accurately predicts the structures and dynamics at normal operating conditions, high temperature-pressure conditions, and in the crystalline solid phase. Furthermore, we show that NNIP-based molecular dynamics of molten salts are scalable to reach long timescales (e.g., nanosecond) and large system sizes (e.g., 105 atoms), while maintaining ab initio accuracy and providing more than three orders of magnitude of computational speedup for calculating structure and transport properties.

中文翻译：

---

用稳健的神经网络原子间势建模LiF和FLiBe熔盐

锂基熔盐由于其在储能，先进裂变反应堆和聚变设备中的应用而备受关注。氟化锂，特别是66.6％LiF-33.3¾F2（氟化锂）在核系统中引起了广泛关注，因为它们显示出理想的传热和中子吸收特性的极佳组合。对于核盐，可能的局部结构，组成和热力学条件的范围在原子建模中提出了重大挑战。在这项工作中，我们证明了以原子为中心的神经网络原子间电势（NNIP）为执行熔盐的分子动力学提供了一种快速而准确的方法。对于LiF，这些电势能够准确地建模二聚体相互作用，变形下的结晶固体，接近熔点的半结晶LiF和高温下的液态LiF。对于Flibe，NNIP可以准确预测正常工作条件，高温压力条件以及结晶固相中的结构和动力学。此外，我们表明基于NNIP的熔融盐分子动力学可扩展以达到较长的时间尺度（例如，纳秒）和较大的系统大小（例如，105个原子），同时保持从头算的准确性并提供超过三个数量级的计算量加快了计算结构和运输性能的速度。