China 
Africa 
Explore chapters and articles related to this topic
Alloys and Environmental Related Issues
Published in Giacomo Giorgi, Koichi Yamashita, Theoretical Modeling of Organohalide Perovskites for Photovoltaic Applications, 2017
Fedwa El Mellouhi, Fahhad H. Alharbi, Carlo Motta, Sergey Rashkeev, Stefano Sanvito, Sabre Kais
High-throughput computational materials design and discovery is an emerging area of materials science combining advanced crystal structure prediction, thermodynamic and electronic-structure calculation, as well as machine learning and database construction (Curtarolo et al. 2013). The first step in materials discovery or in the design of materials with specific properties consists in determining their structural properties. Structural information is critical to the understanding and predicting materials properties and their functionalities. This information can be obtained by using computer simulation methods to map the high-dimensional potential free energy surface and find the global minimum among the large number of local minima (Dixon and Szego 1975; Gavezzotti 1994). It is well known, based on heuristic estimates that the number of local minima grows exponentially with increasing system size (Stillinger 1999). Minimization methods can be mainly classified into two groups, deterministic and stochastic. Deterministic methods, variations on Newton’s method, have the strength of being extremely fast, but have the weakness of often being trapped in local minima. Conversely, a stochastic method is far less likely to be trapped in local minima, but it can be shown that no stochastic method has the potential of converging to the global minimum in a finite number of steps. In principle, there is a need of a full visit of all local minima on the potential energy surface. To find a global minimum in a complex potential energy surface, it is not feasible to use deterministic methods and there is a need for adapting stochastic search strategies (Stillinger 1999).
Improvement of look ahead based on quadratic approximation for crystal structure prediction
Published in Science and Technology of Advanced Materials: Methods, 2022
Tomoki Yamashita, Hirotaka Sekine
Crystal structure prediction (CSP) becomes a key to the advancement of new materials design, with the development of computation capabilities. It has become possible to predict the stable crystal structure for given chemical composition using searching algorithms in conjunction with first-principles calculations. However, CSP is basically a quite difficult global optimization problem due to the exponential increase of the number of potential energy minima with respect to the system size. A great deal of effort has been put into overcoming this problem. So far, several searching algorithms in CSP have been successfully developed such as random search (RS) [1–4], simulated annealing [5,6], minima hopping [7,8], evolutionary algorithm (EA) [9–12], and particle swarm optimization (PSO) [13,14].
CrySPY: a crystal structure prediction tool accelerated by machine learning
Published in Science and Technology of Advanced Materials: Methods, 2021
Tomoki Yamashita, Shinichi Kanehira, Nobuya Sato, Hiori Kino, Kei Terayama, Hikaru Sawahata, Takumi Sato, Futoshi Utsuno, Koji Tsuda, Takashi Miyake, Tamio Oguchi
Research on data-driven materials development based on computer simulations has been actively conducted for the last decades. With the development of computational capabilities, it has become possible to obtain a large amount of reliable data from first-principles calculations at high speed. However, first-principles calculations cannot be directly applied to the design of new materials for new compositions and unknown structures, because such quantum-mechanical approaches require their crystal structures as input. In recent years, crystal structure prediction (CSP) methods have made it possible to discover new materials in conjunction with first-principles calculations. The history of new material discovery by CSP is well summarized in the review paper by the pioneers in this field [1]. So far, a great deal of effort has been devoted to developing the searching algorithms, such as random search (RS) [2–5], simulated annealing [6,7], minima hopping [8,9], evolutionary algorithm (EA) [10–13], and particle swarm optimization (PSO) [14,15]. In particular, EA and PSO are widely used and efficient algorithms as implemented in USPEX [11–13] and CALYPSO [14,15], respectively.