论文部分内容阅读
介绍了自主编制的3维全电磁粒子模拟大规模并行程序NEPTUNE的基本情况。该程序具备对多种典型高功率微波源器件的3维模拟能力,可以在数百乃至上千个CPU上稳定运行。该程序使用时域有限差分(FDTD)方法更新计算电磁场,采用Buneman-Boris算法更新粒子运动状态,运用质点网格法(PIC)处理粒子与电磁场的耦合关系,最后利用Boris方法求解泊松方程对电场散度进行修正,以确保计算精度。该程序初步具备复杂几何结构建模能力,可以对典型高功率微波器件中常见的一些复杂结构,如任意边界形状的轴对称几何体、正交投影面几何体,慢波结构、耦合孔洞、金属线和曲面薄膜等进行几何建模。该程序将理想导体边界、外加波边界、粒子发射与吸收边界及完全匹配层边界等物理边界应用于几何边界上,实现了数值计算的封闭求解。最后以算例的形式,介绍了使用NEPTUNE程序对磁绝缘线振荡器、相对论返波管、虚阴极振荡器及相对论速调管等典型高功率微波源器件进行的模拟计算情况,验证了模拟计算结果的可靠性,同时给出了并行效率的分布情况。
The basic situation of NEPTUNE, a large-scale parallel program of 3-D all-electric particle simulation, is introduced. This program has the ability of 3-D simulation of many typical high-power microwave source devices and can run stably on hundreds or even thousands of CPUs. The program updates the electromagnetic field by FDTD method, updates the particle motion state by using Buneman-Boris algorithm, and uses the particle grid method (PIC) to deal with the coupling relationship between the particle and the electromagnetic field. Finally, the Boris method is used to solve the Poisson equation Electric field divergence correction, to ensure the accuracy of the calculation. The program initially has the capability of modeling complex geometries and can simulate some of the complex structures commonly found in typical high-power microwave devices such as axisymmetric geometry of arbitrary boundary shapes, orthogonal projection geometry, slow-wave structures, coupling holes, metal lines and Surface film for geometric modeling. The program applies the physical boundary such as ideal conductor boundary, external wave boundary, particle emission and absorption boundary and complete matching layer boundary to the geometric boundary, which achieves the closed solution of numerical calculation. Finally, an example is given to simulate the typical high power microwave source devices such as magnetically insulated line oscillator, relativistic return wave tube, virtual cathode oscillator and relativistic klystron using NEPTUNE program. The simulation calculation The reliability of the results, at the same time gives the distribution of parallel efficiency.