Application Of BHMC Algorithm In The Stable Structure Optimation Of Alloy Clusters

Metallic clusters have great potentials in many fields such as physics and chemistry due to their unique property.Especially,the alloy clusters have received extensive attention because they can achieve bi-functional properties by the cooperative effect of different metals.For alloy clusters,the stable structure is extremely important for their multi-functional characteristics and applications.In theory,the structural optimization of alloy clusters based on potential functions is complicated,because the number of local minimum on the potential energy surface of the alloy cluster,will increase exponentially with the increase of atomic number.Therefore,it is critical to apply an effective algorithm for structural optimization of alloy clusters.So far,the traditional method is difficult to obtain satisfied solution because of their low efficiency,it is urgent to develop a global optimization algorithm based on evolutionary algorithm for structural optimization of alloy clusters.In this paper,an improved basin-hopping Monte Carlo(BHMC)method is developed to investigate the stable structures of Fe-Pt alloys.Also,a parallel strategy based on Spark's big data frame is considered to search the structures of Fe-Pt alloys with different Fe/Pt compositions and cluster size.A many-body Gupta potential is applied for describing the atomic interaction in alloy cluster system.The improved BHMC algorithm combines the optimized structure of single metal to improve the initial solution,meanwhile,a genetic local optimization operator is introduced to optimize the isomers of alloys,which can improve the global optimization ability of BHMC algorithm.Furthermore,the improved BHMC algorithm forms a suitable parallel model for structural optimization of Fe-Pt alloys by the Big data parallel framework and successfully realizes the parallelization of the algorithm.In this paper,the stable structures of the FenPt24-n(n = 0-24)bimetallic cluster have obtained by using improved BHMC algorithm,the effectiveness and stability of the improved algorithm are verified by experimental comparison.Also,the Spark-based parallel BHMC algorithm searches the same stable structures for the FenPt24-n(n = 0-24)alloys.Because the parallel algorithm only parallel processing the existing BHMC algorithm,it doesn't change the optimization strategy of the algorithm itself,so it can't find the stable structures of Fe-Pt alloys with more atomic number.In this case,the Spark-based parallel BHMC algorithm is mainly focused on comparing the efficiency of serial and parallel algorithms.The experimental results show that the ratio of optimal speed for the parallel algorithm is 6.44 times than those with the same number of steps in the existing experimental environment,and the parallel algorithm is better than the serial algorithm with the same run time.