The performance of minima hopping and evolutionary algorithms for cluster structure prediction

Area of Science:

• Computational Chemistry
• Materials Science
• Chemical Physics

Background:

• Global optimization is crucial for predicting atomic cluster structures.
• Comparing different optimization algorithms is essential for advancing computational methods.
• Existing algorithms face challenges with complex energy landscapes.

Purpose of the Study:

• To compare the efficacy of evolutionary algorithms and minima hopping for global optimization in cluster structure prediction.
• To introduce and evaluate a novel average offspring recombination operator for evolutionary algorithms.
• To enhance the minima hopping method with a softening approach and improved feedback.

Main Methods:

• Implementation of an evolutionary algorithm with a new average offspring recombination operator.
• Enhancement of the minima hopping algorithm using a softening method and a stronger feedback mechanism.
• Testing both algorithms on atomic clusters with Lennard-Jones potentials, and silicon and gold clusters using force fields.

Main Results:

• Improved minima hopping demonstrated suitability for all tested homoatomic cluster problems.
• Evolutionary algorithms were more efficient for systems with compact, symmetric ground states (e.g., LJ(150)).
• Evolutionary algorithms failed for systems with complex energy landscapes and asymmetric ground states (e.g., LJ(75), silicon clusters >30 atoms).

Conclusions:

• Minima hopping is a robust method for global optimization in homoatomic cluster structure prediction.
• Evolutionary algorithms require further refinement to handle complex and asymmetric energy landscapes.
• The study provides insights into the strengths and weaknesses of both optimization approaches, guiding future algorithm development.