Russian scientists have discovered a new method for enhancing the crystal structure prediction algorithms, thus rendering the discovery of novel compounds many times faster. The study results have been reported in Computer Physics Communications.
Considering the ubiquitous need for innovative technologies, chemists are required to continuously find novel higher performance materials that have better stability, weight, strength, and other characteristics. The advancements in materials science that are being desired by the modern world are almost limitless. The quest for novel materials is an arduous task and, if carried out experimentally, consumes plenty of time and cost because this usually involves testing a large number of compounds at varying conditions. While computers can help overcome this problem, they need smart algorithms: or else, sorting through potential choices can continue for an innumerable number of years until a good compound is identified.
In 2005, things changed when Artem R. Oganov, who is currently a Professor of Skoltech and Moscow Institute of Physics and Technology (MIPT), created USPEX—the evolutionary crystal structure prediction algorithm—which is perhaps the most effective algorithm in the field and employed by a countless number of researchers across the world.
This algorithm only has to know the type of atoms the crystal is made of, and it subsequently produces a small number of arbitrary structures whose stability is evaluated based on the energy of interaction that takes place between the atoms. Following this, an evolutionary mechanism begins, in which chemists created a natural selection, mutations, and crossover of the structures and their “descendants” until they identify compounds that are specifically stable.
In their latest study, researchers from MIPT, Skoltech, and Samara State Technical University, headed by Artem R. Oganov, enhanced the first step of SPEX, which produces initial structures. Demonstrating that purely arbitrary generation is not highly effective, chemists once again inspired by nature created an arbitrary structure generator on the basis of the database of the topological types of crystal structures, combining evolutionary methods devised by Oganov and topological methods devised by Professor Vladislav Blatov from Samara State Technical University.
Armed with the fact that almost all of the 200,000 inorganic crystal structures known till date are part of the 3,000 topological types, one can rapidly produce a wide range of structures comparable to the sought-for structure. In addition, the tests demonstrated that due to the novel generator, the evolutionary search handles the prediction tasks three times more quickly than its earlier version.
The 3,000 topological types are the result of abstraction applied to real structures. Going the other way round, you can generate nearly all the known structures and an infinite number of unknown but reasonable structures from these 3,000 types. This is an excellent starting point for an evolutionary mechanism. Right from the start you most likely sample an area close to the optimal solution. You either get the optimal solution right in the beginning, or get somewhere near it and then get it by evolutionary improvement.
Pavel Bushlanov, Researcher and Study First Author, Skoltech