Structural Optimization Of Metal Nanoclusters Based On Swarm Intelligence Algorithms

Geometric structure of metal nanoclusters determines their physical and chemical properties such as the optical,electrical,magnetic,and catalytic performances to a large extent.Therefore,to optimizate and predict the stable structure of metal nanoclusters are necessary for understanding their unique properties.Also,it is a key aspect in study of metal nanoclusters.So far,because of the difficulty in accurate characterization of nanocluster structures in experiments,the theoretical calculations become the main strategy in the study of nanocluster structures.However,the traditional theoretical calculations such as first-principles calculations have a severve limitation in cluster size and only can handle small nanoclusters with specific configuration and high symmetries due their computational complexity and time-consuming.To develop new computational methods with high performance has great theoretical significance for structural optimization of metal nanoclusters.Theoretically,the structural optimization of metal nanoclusters is a typical global optimization problem and belongs to a NP-hard problem.It is very difficult to be accurately solved by the algirothms available.With the rapid development of computer technology,and the introduction of multi-population searching and parallel computation in intelligent optimization(IO)algorithms make it become increasingly mature.Therefore,to develop theoretical calculations based on swarm intelligence optimization(SIO)algorithm becomes necessary for researches on structures and properties of metal nanoclusters.Furthermore,the structures and properties of metal nanoclusters change significantly with increasing size of nanoclusters,displaying a strong small size effect.Therefore,in this paper we divide the metal nanoclusters into metal clusters(small nanosystems)and metal nanoparticles(large nanosystems)according to their sizes,and develop the corresponding SIO algorithms for these nanosystems according to their structural characteristics.The main researches of this thesis are given as follows.(1)A high-performance parallel multi-strategy differential evolution(PMDE)algorithm is proposed for structural optimization of transition metal clusters(N=5?100),due to the limitation of quantum-mechanical calculations and the deficiencies of heuristic algorithms such as slow convergence rate and easily trappinginto local optima in structural optimization of metal clusters.The algorithm introduces multi-population multi-strategies in mutation and elite pool method for maintaining the diversity of populations in evolution and improving the algorithm convergence.The effectiveness of PMDE algorithm are verified by comparing the results with those of the DFT calculations.Furthermore,the structures of Fe clusters and Cr clusters with the atomic number up to 100 are optimized.Two new structures for Fe83 and Fe84 with lower energies are found by using the PMDE algorithm.(2)To overcome the strong randomicity and slow convergence of traditional random methods and the drawbacks of IO algorithms with trapping into local minimum easily and time-consuming,an improved discrete particle swarm optimization(IDPSO)algorithm is developed for structural optimization of alloy nanoparticles.The methods of swap operator and swap sequence are introduced into the IDPSO algorithm,for solving the discrete combination optimization problem of alloy nanoparticles by changing the movement pattern of particles.Additionally,an adjustment probability is proposed for avoiding the algorithm trapping into local minimum prematurely.The stable structures of tetrahexahedral Pt-based nanoparticles with different size and compositions have been obtained by using the IDPSO algorithm.(3)A Basin hopping genetic algorithm(BHGA)combining with DFT calculations is proposed for structural optimization of adsorbed and supported alloy clusters,focusing on overcoming the faults of size limitation of DFT calculations and approximate solution of IO algorithms on structural optimization of alloy nanoclusters.The stable structures of alloy clusters have been obtained by using the BHGA-DFT method.Furthermore,the adsorption of CO molecular and the effect of TiO2 surface supported on Pd-Ir alloys are analyzed by calculating the adsorption energy and surface binding energy.The combination of SIO and DFT calculations significantly improves the accuracy for structural optimization of adsorbed alloy clusters.It overcomes not only the defects of IO algorithm with being unable to find exact solution,but also the limitation of first-principles calculations on cluster size.As are forementioned,from the perspective of intelligent computation,in this thesis I develop the optimization methods based on the SIO algorithms(PMDE algorithm,IDPSO algorithm,BHGA)to solve the structure optimization of transition metallic clusters,Pt-based alloy nanoparticles and adsorbed alloy clusters.The structural optimization of metal nanoclusters is an issue which needs to be addressed in the fields of chemistry and physics.Therefore,this thesis belongs to fundamental research of cross-subject.On one hand,the research extends and complements the global optimization.On the other hand,it provides the theoretical interpretation and guide for experimental researches and applications of nanoclusters.Also,it provides an important scientific basis for the design and application of new materials.