Crystal structure prediction using ab initio evolutionary techniques: principles and applications (0911.3186v1)

Abstract: We have developed an efficient and reliable methodology for crystal structure prediction, merging ab initio total-energy calculations and a specifically devised evolutionary algorithm. This method allows one to predict the most stable crystal structure and a number of low-energy metastable structures for a given compound at any P-T conditions without requiring any experimental input. Extremely high success rate has been observed in a few tens of tests done so far, including ionic, covalent, metallic, and molecular structures with up to 40 atoms in the unit cell. We have been able to resolve some important problems in high-pressure crystallography and report a number of new high-pressure crystal structures. Physical reasons for the success of this methodology are discussed.

• The paper introduces a novel evolutionary algorithm that achieves near-100% success in predicting both stable and metastable crystal structures using ab initio calculations.
• The methodology efficiently explores the free energy landscape by integrating genetic operations such as heredity, mutation, and permutation.
• Applications include predicting high-pressure phases and novel material properties, thereby advancing autonomous material design and discovery.

Crystal Structure Prediction Using Evolutionary Algorithms

The paper "Crystal structure prediction using evolutionary algorithms: principles and applications" by Artem R. Oganov and Colin W. Glass presents a method that integrates ab initio total-energy calculations with a novel evolutionary algorithm to predict crystal structures. The methodology developed is noteworthy for its high success rate and applicability across diverse types of crystal structures, including ionic, covalent, metallic, and molecular systems without reliance on experimental data.

Methodological Overview

This paper introduces an advanced evolutionary algorithm designed to predict stable and low-energy metastable crystal structures at arbitrary pressure-temperature (P-T) conditions. This algorithm, encapsulated in the USPEX code, operates with minimal input, requiring only the atomic composition and P-T conditions, streamlining the process of crystal prediction substantially.

The cornerstone of this method is its reliance on evolutionary principles. It excels in global optimization of the energy landscape, where conventional methods like molecular dynamics and Monte Carlo simulations fall short due to the complexity and multitude of local minima in free energy surfaces. The algorithm begins with a diverse initial population of potential structures, which are locally optimized and iteratively refined using genetic operations such as heredity, mutation, and permutation. This causes an effective "zooming in" on low-energy regions of configuration space, thereby facilitating the identification of both stable and promising metastable structures.

Numerical Results and Claims

The authors report an impressive near-100% success rate in tests on various systems with up to 20 atoms per unit cell. The method has been validated through the prediction of several high-pressure phases that were previously unresolved or unknown. Notable structures predicted include stable phases like ε-oxygen and new phases of sulfur, as well as metastable phases of carbon and nitrogen.

This high accuracy is attributed to the algorithm's ability to explore the entire free energy landscape rather than merely its vicinity around known minima. Moreover, while current implementations efficiently handle structures with up to 30-40 atoms per cell, the paper hints at possible expansions for even larger systems.

Theoretical and Practical Implications

Theoretically, this research enhances our understanding of crystal structure prediction by illustrating the potency of evolutionary algorithms in navigating complex free energy landscapes. The findings underscore that even without experimental inputs, stable and metastable phases can be consistently predicted, which could significantly broaden our capacity to explore material properties under extreme conditions.

Practically, the evolution-based approach offers a versatile tool for material design, potentially expediting the discovery of novel materials with advantageous properties for technology and industry. The method's inherent scalability and parallelization demonstrate promise for computational efficiency in large-scale simulations.

Future Developments

The authors express ambitions to extend the method’s applicability to include variable compositions and explore potential stoichiometries, addressing challenges such as rationalizing stoichiometry in metallic alloys. Moreover, enhancements in treating disordered and aperiodic structures could broaden the scope of this methodology significantly.

In conclusion, the paper delineates a robust framework for crystal structure prediction, demonstrating the efficacy of evolutionary algorithms in this domain and setting the stage for future advancements in material science research. The capacity to predict structures from chemical composition alone is a substantial step towards autonomous material discovery and design.