使用核碰撞仿真与机器学习研究原子核内中子密度分布与核子核子关联

从高统计量核碰撞数据中提取原子核结构的信息是高能与低能核物理交叉学科领域新兴的热门研究方向。这些研究包括从大量的核碰撞实验或仿真数据中提取原子核变形因子，中子皮厚度及阿尔法团簇等核结构信息。在核碰撞的计算机仿真程序中，原子核一般被建模为多个独立核子的集团，未考虑因两体，三体相互作用及张量力引起的两核子关联对碰撞初态涨落和末态粒子关联的影响。本项目计划在核结构中加入参数化的中子密度分布与核子核子关联，研究其对核碰撞末态粒子分布的影响。使用并改进机器学习中的点云神经网络，将末态粒子在动量空间的分布映射到初态中子密度分布与核子核子在坐标空间的关联，为从核碰撞实验数据中提取核结构信息提供指引。核子在原子核内的分布以及核子核子关联是低能与高能核物理共同关注的重要科学问题。核碰撞蒙特卡洛仿真与机器学习，同时提供了大数据生成与智能数据分析，是创新的交叉学科研究方法。

Extracting nuclear structure from high statistical heavy-ion collision data is a new research area in the interdisciplinary researches of high and low energy nuclear physics. Hot topics include extracting the nuclear shape deformation, neutron skin and $\alpha$-clusters, from a big amount of nuclear collision data from experiments or simulations. Nucleus are usually modeled as a cluster of independent nucleons in the computer Monte Carlo simulations of high-energy heavy-ion collisions, without taking into account the effect of nucleon-nucleon correlation caused by two-body, three-body interaction and tensor force to the initial state fluctuation and final state correlation. The current project plans to add parameterized neutron density distribution and nucleon-nucleon correlation in nuclear structure, to study their effects on the final state particle distribution, and to map the momentum distribution of final state particles to the neutron density and spatial correlation of nucleon pairs, using improved Point-Cloud neural network from machine learning tool box, to provide a guideline for extracting nuclear structure from nuclear-collision experiment. The distribution and correlation of nucleons in nucleus are the common interests of low energy and high energy nuclear physics. The nuclear-collision Monte Carlo simulations and machine learning, which provide both big data generation and smart data analysis, is novel interdisciplinary research method.