NEW EFFICIENT ALGORITHM TO GENERATE DOPED CRYSTAL STRUCTURE MODEL USED FOR AB INITIO SIMULATIONS
Abstract
Atomic site substitutions (doping, solid solutions etc.) are the common strategy to tune the properties of compounds. In order to predict the tuned property by simulation, one has to provide a crystal structural model with the substitutions. We develop an algorithmic implementation to generate crystal structure models with atomic substitutions used for ab initio materials simulations. Vast number of cases how to assign the sites to be replaced in an atomic crystal structure should be sorted by the space group theory into the clusters each of which accommodates the equivalent symmetry of the substitutions. The number, however, easily gets to explode amounting to billion or trillion, which cannot be handled by the conventional sorting tools. To overcome the combinatorial explosion beyond the capacity of the conventional tools, a different strategy to prevent from the classification over the whole samples should be used. In modern theory of combinatorics, various knowledge has been developed to prevent such a global surveying. In this study, we adopt an approach called the canonical augmentation. However, the core of this method, "m-function", has not been provided as a concrete form, but just as a recursive definition. We have implemented it as a concrete formulation for our purpose, and overwhelming speedup and memory saving have been demonstrated. A package implementation of our method is being released as SHRY , a Python implementation of the method. In addition to the expected memory savings, the overwhelming speed of the method is being verified. If the language implementation is replaced with C or other compilar-type languages, the speed is expected to be further overwhelmingly faster. As a tool for generating representative structure models for solid solutions, it will eliminate the inhibiting factors in the past for the materials tuning via solid solutions, bringing a revolutionary leap forward.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 16, 2024
- Source ID
- FA23862314013
Entities
People
- Ryo Maezono
Organizations
- Air Force Office of Scientific Research
- Japan Advanced Institute of Science and Technology
- United States Air Force