Heuristic Optimization Of The Binary-morse Cluster

Cluster is made of many atoms.These atoms of the cluster stay in a stable status through the forces between them.The heuristic optimization of cluster is to find the best configuration of a cluster,which has the lowest energy.This is a typical non-deterministic polynomial-hard problem because the complexity of the problem grows exponentially as the cluster's size grows.The study of binary cluster is of great importance.On the one hand,there are plenty of physical and chemical properties that cannot be found in the single cluster exsiting in the binary cluster.These properties could be guidance to material synthesis,such as synthesis of medicine,alloy,and catalyst and so on.On the other hand,binary cluster offers the opportunity to get a new structure through the choice of atomic types,cluster size and composition.This paper focuses on binary Morse cluster.We use Morse potential to describe the interaction between atoms of the cluster.Morse potential is one of the pair-potential,which is very simple.Compared to single cluster,the search for the best configuration of of the binary cluster offers a considerable challenge.Because binary cluster can present complex potential energy surfaces compared to single cluster.Besides,there are plenty of isomers with the same geometric structure,but differing in the labeling of the atoms.So,the search for the energy-lowest cluster is more difficult.Algorithms used in this paper utilize L-BFGS as a basic optimization method.Algorithms used in this paper also include global perturbation,dynamic lattice search,surface operator,a two-phase turn-over method.We use these algorithms to search for the best configuration of clusters which consist of 38 and 75 atoms.In general,the first three belong to perturbation methods.The last one is a moving method.This paper optimized the surface operator and come up with a two-phase turn-over algorithm to search for the best configuration of the cluster whose size is 38 and 75.The new algorithm is proved more efficient in searching for the best configuration of a cluster.