金团簇结构的人工蜂群算法研究
The study of structures of gold clusters by Artifical Bee Colony algorithm

采用具有群体智能的蜂群优化算法(ABC)结合半经验Gupta原子间相互作用势对金团簇Aun(n=2~100)的稳定结构、结合能(Eb)、平均结合能(Eab)、平均间距(ra)、对称性(PG)、一阶差分(？E1)和二阶差分(？E2)进行了研究。结果表明: 一、ABC算法在中小尺度团簇上具有良好的全局寻优能力，其与遗传算法(GA)、动态格点搜索(DLS)等算法结果吻合非常好，当团簇尺度增大时，与其它算法一样，标准ABC算法的全局最优搜索能力也逐渐减弱；二、ABC算法中三参数取值对最优解获取有直接影响，当团簇尺度n小于60时，中等参数值即能获得最优解，当n大于60时，为平衡计算时间和精度，较大的最大迭代次数(gmax)、中等或偏小的可行解尺度(SN)和最大侦查蜂数目(glimit)取值是一种较好的选择；三、通过对ABC和文献结果比较，几个金团簇的新低能结构被提出和确认，另外，本文首次对Au91~Au99进行了系统描述。
TThe structures, binding energies, average binding energies, averaged distances, symmetries, the first and second differences of energies about gold clusters Aun(n=2~100) have been investigated by the artificial bee colony algorithm (ABC) in conjugation with the semi-empirical Gupta inter-atomic potential. The calculated results show that: first, ABC algorithm has an excellent capability on looking for the global minimum value to the small or medium size clusters, the results from ABC are in good agreement with those from genetic algorithm (GA) and dynamical lattice search (DLS); With the increase of the size of cluster, the capability of looking for the global minimum value for the standard ABC becomes weaken gradually, similar to others. Second, three parameters proposed in the ABC algorithm have a direct effect to the optimized results, if the size of cluster n is less than 60，it is easy to find the optimum solution, in opposite, there is a good choice if you give a value with large iteration (gmax), a small or medium trial solutions (SN) and scout bees (glimit). Third, by compared to the results of the ABC and references, several structures with lower energy have been proposed and confirmed. In addition, the Au91~Au99 have been described systematically in the first time.