中文摘要：

除中子外,聚变核心同时释放大量高能伽马,其能谱可反映聚变过程的关键物理参数,并为过程诊断提供重要信息.由于聚变伽马的时间与能量特性,需要设计高探测效率及能量分辨率的伽马谱仪.根据高能伽马谱仪的概念设计（gamma-to-electron magnetic spectrometer）,针对该系统中伽马-电子转换靶、电子偏转汇聚、电子探测等关键环节进行优化设计以提高系统探测效率及能量分辨率.其中采用Monte-Carlo程序Geant4模拟研究了伽马-电子转换靶中康普顿散射与多次库仑散射对由转换靶出射电子的能谱与角分布的影响.开发并行遗传算法对复杂几何偏转磁场参数进行优化,得到低强度（小于100 Gauss）复杂边界偏转磁场.根据系统优化设计结果,采用Geant4模拟了该系统对不同能量伽马的响应.此外,还可模拟该系统对特征聚变伽马能谱的测量,结果显示,该系统可在聚变中子产额分别为2.5×10^15及1.2×10^16条件下,对10—20 MeV高能伽马能谱测量实现能量分辨分别满足0.5 MeV（小于5%）及0.25 MeV（小于2.5%）,说明该系统可用于聚变过程伽马能谱的诊断.

英文摘要：

Apart from neutrons, the fusion core produces gamma rays during fusion reaction. The spectrum of gamma ray can provide very important information for fusion diagnosis. However, due to the gamma energy and yield in one fusion pulse being both lower, the gamma spectrometer used should have high detection efficiency and energy resolution. The concept of a Gamma-to-electron magnetic spectrometer （GEMS） provides the idea to build up such a spectrometer to meet this requirement. Based on this concept design, four important parts of this facility are investigated. The first part is the gamma-electron converter. The main physics processes include Compton scattering of gamma ray with converter material generating electron, the electron multiple Coulomb scattering （MCS） inside the converter and the electron attenuation. Affected by the thickness of convector, these processes give a complex influence on the detection efficiency and angular-energy distribution of the electrons which are emitted from the downstream face of the convector. The Monte Carlo code Geant4 is employed to investigate theeffects of Compton scattering, MCS and converter thick on the angular-energy distribution. The second one is the collimation. The collimation is used to select the forward direction, the performances of cutoff angle of the collimator on the detection efficiency and resolutions, the correlation between electron transportation direction and energy, are also studied using Geant4 code. The third part is the dipole magnetic field. There are several geometric and magnetic parameters, therefore, a multi-thread parallelized genetic algorithm is developed to obtain the best result. Both the irregular geometric shape and dipole magnetic field strength are optimized to achieve the best energy resolution and detection efficiency. The obtained magnetic field has an intensity of less than 100 Gauss, and its performance on gathering elections is also verified by Geant4 code. The last one is the location of electron detectors. The study shows