Connecting Crystal Structures by Symmetry via Subgroup Matching
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 28 HE credits
Student thesis
Abstract [en]
Analyzing a reconstructive phase transition in crystalline materials requires the analysis of paths in which certain symmetries are preserved, in order to help find the most probable path the material takes in a real world scenario. Such a path can then be used, for example, for thermal dynamic integration, where small changes in the overall physics of a material are beneficial when calculating the change of a state variable as a material changes. In this project, a program for finding and generating symmetry preserving paths between crystalline materials was created. It provides an alternative to existing programs such as COMSUBS and p2ptrans. The program is similar in structure to COMSUBS, but uses a more sophisticated collision detection algorithm and is written in python, providing a modern and easily expandable program. Validating the functions of the program by comparing the program’s output to the expected output, found that 99.35% of generated subgroup structures could be validated and >99.8% of generated paths could be validated as symmetry preserving. The program can therefore be used for this purpose.
The program was used to generate paths between materials in the Materials Project database. The paths were then searched in order to find pairs of materials connected by a very short path. Approximately 5% of analyzed pairs of materials had an average path distance of less than 1 Å per atom and a unit cell volume difference of less than 20%. This relatively high percentage could pose a problem in AI modelling or other use cases where large datasets are used. Limited data diversity in the training data worsens the predictability in AI models made from it. In cases where specific materials are analyzed, this similarity can be a cause of confusion and uncertainty about which materials are the most stable and best to use. Further analysis using the program should be performed on other databases, especially those created using AI models, in order to determine the usefulness of the models.
Place, publisher, year, edition, pages
2024. , p. 46
Keywords [en]
crystalline, materials, material, path, search, subgroup, structure, space group, symmetry, preserving, simulation, program
National Category
Condensed Matter Physics
Identifiers
URN: urn:nbn:se:liu:diva-207867ISRN: LITH-IFM-A-EX--24/4376--SEOAI: oai:DiVA.org:liu-207867DiVA, id: diva2:1903238
Presentation
2024-08-30, Laplace N323, Fysikhuset, Olaus Magnus väg, 583 30, Linköping, 10:15 (English)
Supervisors
Examiners
2024-10-222024-10-032024-10-22Bibliographically approved