Atomistic simulations of tensile test are often employed in studies of cleavage or decohesion of crystals and interfaces. However, the particular numerical implementation differ from model to model. Each individual model is characterized not only by a particular computational procedure but also by different loading conditions and the way the crystal breaks. Therefore, calculated values of strength can substantially differ and the particular choice of the model affects the predictions made. This paper is intended to illustrate differences in computational models and their results. Particular attention is paid to differences in computed values of fracture stress and its dependence on the size of computational supercell. It is shown that the fracture stress computed in models considering uniform stress distribution does not depend on the supercell size. On the other hand, the fracture stress decreases with increasing supercell size in models with localized strain.

中文翻译：

---

在计算拉伸试验中，超级单元尺寸和应变局部化的影响

拉伸测试的原子模拟通常用于研究晶体和界面的分裂或解聚。但是，特定的数值实现因模型而异。每个单独的模型不仅具有特定的计算过程，而且具有不同的加载条件和晶体破裂的方式。因此，强度的计算值可能会大不相同，并且模型的特定选择会影响所做的预测。本文旨在说明计算模型及其结果的差异。特别要注意断裂应力的计算值的差异及其对计算超级单元大小的依赖性。结果表明，在考虑均匀应力分布的模型中计算的断裂应力与超级单元尺寸无关。另一方面，在局部应变模型中，断裂应力随着超级单元尺寸的增加而减小。