Global geometry optimization of clusters using a growth strategy optimized by a genetic algorithm

Bernd Hartke

DOI: https://doi.org/10.1016/0009-2614%2895%2900587-T

1995-05-24

Abstract:A new strategy for global geometry optimization of clusters is presented. Important features are a restriction of search space to favorable nearest-neighbor distance ranges, a suitable cluster growth representation with diminished correlations, and easy transferability of the results to larger clusters. The strengths and possible limitations of the method are demonstrated for Si10 using an empirical potential.

Chemical Physics

What problem does this paper attempt to address?

The problems that this paper attempts to solve are several key difficulties in cluster geometric optimization, especially the problem of finding the lowest - energy configuration for clusters composed of a given number of atoms or molecules. Specifically, the paper focuses on the following challenges:

1. **The number of local minima increases exponentially with the cluster size**: As the number of atoms included in the cluster increases, the number of possible local minima increases dramatically, which makes it more difficult to find the global minimum.
2. **Appropriate problem representation**: Selecting appropriate coordinates or parameters for optimization is an important issue, and different representation methods may lead to different optimization effects.
3. **Appropriate restriction of the search space**: How to effectively restrict the search space to reduce unnecessary consumption of computing resources while ensuring that the global optimal solution can be found.
4. **High - cost ab initio potential energy surface calculations**: Ab initio calculations of the potential energy surface require a large amount of computing resources, which is a major bottleneck in practical applications.

To address these challenges, the authors propose a new strategy - GOALS (Global Optimization by Assembly of Local Structure), which improves cluster geometric optimization in the following ways:

- **Transferring knowledge from smaller clusters to larger ones**: Using the known global optimal structure information of small clusters to construct the initial geometric configurations of larger clusters.
- **Avoiding simple coordinates for direct optimization of individual atoms**: This method usually results in highly correlated parameters, thus affecting the optimization effect.
- **Constructing methods to avoid generating obviously unreasonable structures**: Ensuring that the generated cluster structures are physically meaningful.

The paper demonstrates the effectiveness of the GOALS strategy through a specific implementation case - the optimization of small silicon clusters (such as Si\(_{10}\)) - and compares it with traditional optimization methods, proving that this strategy has a significant advantage in increasing the probability of finding the global minimum.