基于动力学蒙特卡洛模拟的数据集
收藏资源简介:
该数据集基于动力学蒙特卡洛模拟生成,用于研究二维多粒子系统的表面步骤波动及其相关的时间依赖性粗糙度。数据集包含了300条独立的KMC运行轨迹,每条轨迹包含1000个时间步长的状态快照,这些快照记录了原子边缘扩散过程中系统的演化。该数据集的创建旨在通过生成对抗网络(GANs)学习随机动力学,从而替代传统的模型并捕捉热波动。数据集的内容来源于简单的立方材料{100}表面的二维晶格KMC模拟,初始状态为一个条形区域,随着时间的推移,由于热波动,该区域逐渐变得粗糙。该数据集可用于加速模拟过程,并提高预测能力,同时考虑到对称性和守恒定律,无需详细了解潜在的势能景观即可训练模型。
This dataset is generated via kinetic Monte Carlo (KMC) simulations to investigate surface step fluctuations and their associated time-dependent roughness in two-dimensional multi-particle systems. It contains 300 independent KMC run trajectories, each with 1000 time-step state snapshots that record the system's evolution during atomic step edge diffusion processes. This dataset was developed to learn stochastic dynamics using Generative Adversarial Networks (GANs), replacing traditional models while capturing thermal fluctuations. The dataset originates from two-dimensional lattice KMC simulations on the {100} surface of a simple cubic material, starting with an initial stripe-shaped region that gradually becomes rough over time due to thermal fluctuations. It can be used to accelerate simulation workflows and enhance prediction capabilities, enabling model training without detailed knowledge of the underlying potential energy landscape while accounting for symmetry and conservation laws.




