Data for: "Prediction of Grain Boundaries from Bulk Structures with Score-based Denoising Models"

This folder contains data and figures included in the paper 'Prediction of Grain Boundaries from Bulk Structures with Score-based Denoising Models'. Folder named 'Figures': 1. including a file 'combined_dataframe_all.csv' which contains information about parameters of grain boundaries, corresponding tralations, level of added noise during denoising, grain boundary energy, and filterd entropy. 'ini_opt' means the structures obtained via direct optimization through MACE and 'fin_opt' means the structures obtained via optimization after denoising; 2. 'dH_steps' includes figures showing the change of information of entropy of atoms during the denoising processes; 3. 'Comparison' includes the figures which compare the denoised grain boundary structures with experimental and simulation results from previous study with permission of reuse included in the folder 'permission_reuse'. The comparison between the grain boundary structures of the same crystal structures with different elements is also included in this folder; 4. 'denoising_trajectory' includes the figures which describe the denoising trajectory of each tested grain boundary structure; 5. 'energy_difference' included figures comparing the grain boundary energy out of direct optimization and optimization after denoising; 6. 'adding_noise' includes the grain boundary energy with different levels of noise added; 7. 'small-medium' compares the energy obtained via MACE ''small'' model and ''medium'' model for graphene and FCC Cu. Other folders contain the data, code, structures, and trajectories of each tested grain boundary structures with certain parameters and they are stored in the same style. No matter which composite it is in the folder, the naming rules of the files and folders are in similar or repeating styles: The folder with name ended with '_Comparison' includes the grain boundary structures and corresponding denoising trajectories (.extxyz files) which is used for comparison with structures from previous literatures. The file ended with '.pt' is the trained model for denoising; Folders ended with numbers such as '_0.00', '0.10'... inform the level of noise added during denoising processes. Each folder inside these folders informs the grain boundary angles/orientations such as 3_111 means a Σ3{111} grain boundary; For graphene, the folder named, for example, 5.086_5.5 means 5.086°grain boundary which similar to 5.5° grain boundary mentioned in previous study but can have a periodic boundary condition under coincidence site lattice (CSL) framework; Files named 'denoiser.ipynb' are codes used for denoising the original structures and 'deop.ipynb' are codes include both denoising and optimizing processes via MACE. 'dH.ipynb' codes are for counting the information entropy values of the structures; Folder 'vacuum' contains two subfolders 'ini' and 'fin'. 'ini' contains denoised and optimized structures and trajectories towards the original built structures and 'fin' includes optimized structures and trajectories toward denoised grain boundary structures. '.xyz' files include structures before/after denoising and optimizing and '.extxyz' files include trajectories for denoising and optimizing. The number included in the file name, for example '-6.O_ini.xyz', indicates the translation along the grain boundary direction; Files named with 'ini.xyz' are structures built before any denoising or optimizing process; Files named with 'final_step.xyz' or 'denoised.xyz' or 'last_step.xyz' are grain boundary structures denoised; Files named with 'ini_optimised' or 'fin_optimised' means trajectories or structures that are optimized via MACE. 'ini' for direct optimization and 'fin' for optimization after denoising; Files named with 'denoising_adding_noise' are for trajectories or structures obtained via denoising process while adding extra level of noise; Files named with 'afteradding_noise' are for trajectories or structures obtained via denoising without adding noise. Folder 'dH' inside 'ini' and 'fin' are structures marked with information entropy and it can be visualized in OVITO by the modification named 'color coding'; If files in 'dH' is named with 'filtered', it means those files include information entropy values with filter (for details about the filter please read the paper) applied. (ps: Files named 'gb_generator.py', 'csl_generator.py', and 'io_file' are used for generating CSL with the package 'GB code'.) All structure folders follow this naming rule.