机器学习确定中低能区重离子碰撞参数分布研究  

Determination of the impact parameter distribution of low-intermediate energy heavy ion collisions using machine learning

作　　者：李理 张英逊[1,2] 崔莹 梁嘉杰 LI Li;ZHANG YingXun;CUI Ying;LIANG JiaJie(Department of Nuclear Physics,China Institute of Atomic Energy,Beijing 102413,China;Guangxi Key Laboratory of Nuclear Physics and Technology,Guangxi Normal University,Guilin 541004,China;Sino-French Institute of Nuclear Engineering and Technology,Sun Yat-sen University,Zhuhai 519082,China)

机构地区：[1]中国原子能科学研究院核物理研究所,北京102413 [2]广西师范大学广西核物理与核技术重点实验室,桂林541004 [3]中山大学中法核工程与技术学院,珠海519082

出　　处：《中国科学：物理学、力学、天文学》2022年第5期126-134,共9页Scientia Sinica Physica,Mechanica & Astronomica

基　　金：国家自然科学基金(编号:11875323,11705163,11790320,11790323,114222548,11961141003);国家重点基础研究发展计划(编号:2018YFA0404404);稳定基础科研支持资助项目(编号:WDJC-2019-13,BJ20002501);领创计划(编号:LC192209000701,LC202309000201)资助。

摘　　要：重离子碰撞的碰撞参数对动力学演化过程及产物分布都有重要影响.但是由于涨落效应的存在,实验观测量的某一值所对应的碰撞参数必然存在一定的分布.目前,对于实验观测量所对应的碰撞参数分布的确定主要有三种方法,即锐截断近似(Sharp Cut-Off Approximation)方法、神经网络方法和贝叶斯估算方法.本文以改进型量子分子动力学模型(ImQMD)为数据产生器,模拟产生了20000个事件.利用所产生的数据集,对上述三种方法给出的碰撞参数分布的划分进行了讨论和分析.最后,我们提出了一种不依赖于模型的重离子碰撞中心度确定方法,即非监督型机器学习方法——K-means聚类算法确定碰撞中心度.预期该方法能够对于中低能重离子碰撞中心度的确定提供另一种思路.The impact parameter is an important quantity for reasonably simulating heavy ion collisions(HIC)using the transport model,because many HIC observables obviously depend on it in the simulation.However,the impact parameter cannot be directly measured or determined from HIC observables,so reconstructing impact parameter distributions from HIC observables is highly desired.For intermediate energy HICs,impact parameters distribute in a wide range for the selected events due to the strong fluctuation mechanism.In this work,three methods for reconstructing impact parameter distributions are discussed,namely sharp cut-off approximation,Bayesian estimation,and neural network estimation of impact parameters.The dataset used for reconstructing the impact parameter distribution by machine learning methods are generated from an improved quantum molecular dynamics model(Im QMD).Finally,we propose to use an unsupervised machine learning method,namely the K-means algorithm,to sort the centrality of heavy ion collisions and its validity is also proved in theory.

关 键 词：碰撞参数分布 机器学习 神经网络 K-MEANS 

分 类 号：O571.6[理学—粒子物理与原子核物理] TP181[理学—物理]