Spark-based improved Basin-Hopping Monte Carlo algorithm for structural optimization of alloy clusters

CoPt alloy clusters are promising candidates for catalysts in fuel-cell applications because of their enhanced catalytic performances compared with pure Pt clusters. In this article, an improved Basin-Hopping Monte Carlo (BHMC) algorithm is proposed to optimize the stable structures of CoPt clusters with different sizes and compositions, and the Spark parallel framework is employed to accelerate the structural optimizations. The results show that the improved BHMC algorithm has higher probability to find the lowest-energy structure and more rapid convergence than the traditional BHMC one. Through the comparison of different threads during Spark parallel computing, it is found that the acceleration ratio increases with the thread number while decreases for large cluster due to the limitation of the resource allocation. By investigating the optimized structures of CoPt clusters, it is revealed that for small CoPt clusters, the most stable structures exhibit a wide variety at different Co/Pt ratios; for large CoPt clusters, however, their structures are independent of the Co/Pt ratios and close to the results of their monometallic counterparts. Besides, for all the CoPt clusters, Pt atoms tend to distribute near the surface and Co atoms are generally located in the interior.