A sphere-cut-splice crossover for the evolution of cluster structures.

Area of Science:

• Computational Chemistry
• Materials Science
• Evolutionary Computation

Background:

• Determining the lowest energy structures of atomic clusters is computationally challenging.
• Traditional genetic algorithms use plane-cut-splice crossover operators, which have limitations in global searching and local optimization speed.
• Existing methods struggle with efficiently exploring the complex energy landscapes of atomic clusters.

Purpose of the Study:

• To introduce a new sphere-cut-splice crossover operator for evolving atomic cluster structures.
• To enhance the performance of genetic algorithms in finding global minima for atomic clusters.
• To address the limitations of classical plane-cut-splice methods.

Main Methods:

• A novel crossover operator employing a spherical cutting and splicing strategy for atomic cluster structures.
• Implementation of the sphere-cut-splice crossover within a genetic algorithm framework.
• Testing and comparison against the traditional plane-cut-splice crossover using Lennard-Jones clusters (30 ≤ N ≤ 500).

Main Results:

• The sphere-cut-splice crossover significantly improves genetic algorithm performance compared to the plane-cut-splice method.
• Reduced average local minimizations (80.75%) and energy evaluations (83.86%) were observed with the sphere scheme.
• A mean speed-up ratio of 1.8207 was achieved, with larger speed-ups (2.3520) for clusters with 110 ≤ N ≤ 500.
• A modified strategy using optimized sphere sizes further improved global search capabilities.

Conclusions:

• The sphere-cut-splice crossover operator is a more effective method for evolving atomic cluster structures than plane-cut-splice operators.
• This approach offers substantial improvements in computational efficiency and global search capability, particularly for larger atomic clusters.
• The findings suggest a promising new direction for computational studies in atomic cluster structure prediction.