arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2410.01464cond-mat.mtrl-scics.LGphysics.comp-ph
收藏

用于加速晶体材料中原子输运模拟的 Flow Matching

Flow Matching for Accelerated Simulation of Atomic Transport in Crystalline Materials

Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA(材料科学与工程系,麻省理工学院);Energy Storage Research Alliance, Argonne National Laboratory, 9700 South Cass Avenue, Lemont, Illinois 60439, United States(能源存储研究联盟,阿贡国家实验室);Computational and Systems Biology Program, Massachusetts Institute of Technology, Cambridge, MA 02139, USA(计算与系统生物学计划,麻省理工学院)

Juno Nam, Sulin Liu, Gavin Winter, KyuJung Jun, Soojung Yang, Rafael Gómez-Bombarelli

AI总结:

本文提出 LiFlow 生成框架,利用 flow matching、原子位移传播与局部校正加速晶体 SSE 中锂扩散模拟,并在大规模候选材料基准上实现高精度泛化和最高 600,000 倍加速。

摘要:

原子输运支撑着储能和电子等技术中材料的性能,但其模拟在计算上仍然要求很高。特别是,对固态电解质(SSE)中的离子扩散进行建模,需要能够克服传统从头算分子动力学(AIMD)尺度限制的方法。我们提出 LiFlow,这是一个用于加速晶体材料 MD 模拟的生成式框架,将该任务表述为原子位移的条件生成。该模型采用 flow matching,并使用 Propagator 子模型生成原子位移、使用 Corrector 对非物理几何结构进行局部修正,同时结合基于 Maxwell-Boltzmann 分布的自适应先验,以考虑化学和热学条件。我们在一个数据集上对 LiFlow 进行基准测试,该数据集包含 4,186 个 SSE 候选材料在四个温度下的 25 ps 锂扩散轨迹。对于未见过的成分,该模型在锂均方位移(MSD)预测上获得 0.7–0.8 的稳定 Spearman 秩相关。此外,LiFlow 能够从较短的训练轨迹泛化到更大的超胞和更长的模拟,同时保持较高精度。与第一性原理方法相比,LiFlow 实现高达 600,000 倍的加速,使显著更大长度和时间尺度上的可扩展模拟成为可能。

英文摘要:

Atomic transport underpins the performance of materials in technologies such as energy storage and electronics, yet its simulation remains computationally demanding. In particular, modeling ionic diffusion in solid-state electrolytes (SSEs) requires methods that can overcome the scale limitations of traditional ab initio molecular dynamics (AIMD). We introduce LiFlow, a generative framework to accelerate MD simulations for crystalline materials that formulates the task as conditional generation of atomic displacements. The model uses flow matching, with a Propagator submodel to generate atomic displacements and a Corrector to locally correct unphysical geometries, and incorporates an adaptive prior based on the Maxwell-Boltzmann distribution to account for chemical and thermal conditions. We benchmark LiFlow on a dataset comprising 25-ps trajectories of lithium diffusion across 4,186 SSE candidates at four temperatures. The model obtains a consistent Spearman rank correlation of 0.7-0.8 for lithium mean squared displacement (MSD) predictions on unseen compositions. Furthermore, LiFlow generalizes from short training trajectories to larger supercells and longer simulations while maintaining high accuracy. With speed-ups of up to 600,000$\times$ compared to first-principles methods, LiFlow enables scalable simulations at significantly larger length and time scales.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑