RT2QMD: GPU-Accelerated nucleus-nucleus fragmentation event generator
收藏资源简介:
We developed a GPU-accelerated nucleus-nucleus fragmentation event generator. The Quantum Molecular Dynamics (QMD) model was implemented on GPU architecture. A corresponding evaporation model was also integrated to handle de-excitation. The developed models, RT2QMD, can handle nuclear collisions where the projectiles including oxygen isotopes and lighter nuclei, covering most of the situations in carbon-ion radiotherapy. The intended energy range of this package is 100 MeV/u to 500 MeV/u. This package was compared against the corresponding Geant4 models and experimental data. The RT2QMD showed good agreement with Geant4 and experimental data for neutron double-differential cross-section in the 290 MeV/u 12C(12C,xn), 12C(16O,xn) reactions, and the 230 MeV/u Cu(4He,xn) reaction. The fragment production cross-section from 12C - 12C reactions showed relatively large differences compared to Geant4 and experimental data, due to the simplified evaporation model. The RT2QMD ran on an NVIDIA RTX 4090 GPU, while the Geant4 models ran on an Intel Xeon Gold 6342 node using all 48 available threads. The computing speeds of RT2QMD were about 30 times faster than those of Geant4 for all reactions. This package is part of the GPU-based Monte Carlo code, RT2, to handle dose calculation in heavy-ion therapy.
本研究开发了一款基于图形处理器(Graphics Processing Unit, GPU)加速的核-核碎裂事件生成器。量子分子动力学(Quantum Molecular Dynamics, QMD)模型已基于GPU架构完成实现,同时集成了对应的蒸发模型以处理核退激发过程。这套开发完成的模型套件命名为RT2QMD,可处理包含氧同位素及轻核在内的弹核引发的核碰撞过程,覆盖碳离子放射治疗中的绝大多数应用场景。该工具包的适用能量范围为100 MeV/u至500 MeV/u。 本工具包已与对应的Geant4模型及实验数据开展对比验证:在290 MeV/u的碳-12与碳-12(12C(12C,xn))、碳-12与氧-16(12C(16O,xn))反应,以及230 MeV/u的铜与氦-4(Cu(4He,xn))反应中,RT2QMD在中子双微分截面的预测结果与Geant4模型及实验数据吻合良好;但受限于简化的蒸发模型,碳-12与碳-12反应的碎裂产物产生截面的预测结果与Geant4模型及实验数据存在相对较大的偏差。 RT2QMD模型运行于NVIDIA RTX 4090 GPU平台,而Geant4模型则运行于搭载Intel Xeon Gold 6342处理器的节点,并利用了全部48个可用线程。在所有测试反应中,RT2QMD的计算速度约为Geant4模型的30倍。本工具包是基于GPU的蒙特卡洛代码RT2的组成部分,用于重离子治疗中的剂量计算。



