Abstract A domain-decomposed method to simultaneously couple the classical Molecular Dynamics (MD) and Direct Simulation Monte Carlo (DSMC) methods is proposed. This approach utilises the MPI-based general coupling library, the Multiscale Universal Interface. The method provides a direct coupling strategy and utilises two OpenFOAM based solvers, mdFoam + and dsmcFoam + , enabling scenarios where both solvers assume one discrete particle is equal to one molecule or atom. The ultimate goal of this work is to enable complex multi-scale simulations involving micro, meso and macroscopic elements, as found with problems like evaporation. Results are presented to show the fundamental capabilities of the method in terms of mass and kinetic energy conservation between simulation regions handled by the different solvers. We demonstrate the capability of the method by deploying onto a large supercomputing resource, with attention paid to the scalability for a canonical NVT ensemble (a constant number of atoms N, constant volume V and constant temperature T) of Argon atoms. The results show that the method performs as expected in terms of mass conservation and the solution is also shown to scale reasonably on a supercomputing resource, within the known performance limits of the coupled codes. The wider future of this work is also considered, with focus placed on the next steps to expand the capabilities of the methodology to allow for indirect coupling (where the coarse-graining capability of the DSMC method is used), as well as how this will then fit into a larger coupled framework to allow a complete micro-meso-macro approach to be tackled.

中文翻译：

---

使用通用和高性能代码耦合库耦合分子动力学和直接仿真 Monte Carlo

摘要 提出了一种同时耦合经典分子动力学（MD）和直接模拟蒙特卡罗（DSMC）方法的域分解方法。这种方法利用了基于 MPI 的通用耦合库，即多尺度通用接口。该方法提供了一种直接耦合策略，并利用了两个基于 OpenFOAM 的求解器 mdFoam + 和 dsmcFoam +，从而实现了两个求解器都假设一个离散粒子等于一个分子或原子的场景。这项工作的最终目标是实现复杂的多尺度模拟，涉及微观、中观和宏观元素，如蒸发等问题。结果显示了该方法在不同求解器处理的模拟区域之间的质量和动能守恒方面的基本能力。我们通过部署到大型超级计算资源上来证明该方法的能力，并关注氩原子的规范 NVT 系综（恒定原子数 N、恒定体积 V 和恒定温度 T）的可扩展性。结果表明，该方法在质量守恒方面按预期执行，并且该解决方案还显示出在耦合代码的已知性能限制内在超级计算资源上合理扩展。还考虑了这项工作的更广阔的未来，重点放在扩展方法的能力以允许间接耦合（使用 DSMC 方法的粗粒度能力的地方）的后续步骤，以及这将如何然后融入一个更大的耦合框架，以允许处理完整的微观-中观-宏观方法。