DeepH-HONPAS dataset

DeepH-HONPAS dataset consists of the raw DFT data for the training of bilayer graphene and bilayer MoS2 structures. Also, the DFT results including overlap matrices of twisted structures are provided. This dataset is used with DeepH-HONPAS code.

To prepare the dataset, please unzip the dataset and set the right path by specifying "raw_dir" in the preprocess.ini file. After preprocess, the binary .HSX files will be translated into .h5 files for neural network training. Dataset files include:

Bilayer_graphene_PBE.zip: a set of bilayer graphene structures, calculated with PBE functional

Bilayer_graphene_HSE06.zip: a set of bilayer graphene structures, calculated with HSE06 functional

Bmos2_PBE.zip: a set of bilayer MoS2 structures, calculated with PBE functional

Bmos2_HSE06.zip: a set of bilayer MoS2 structures, calculated with HSE06 functional

To inference the large structures, please unzip the files that contain the twisted structure result and set the right path by specifying "OLP_dir" in the inference.ini file. After the inference step, the predicted Hamiltonians should be in the "work_dir". Further comparison of band structures are available with the .bands files. The overlap files include:

PBE results: TBG-7.34-DZ-PBE.zip, TBG-4.41-DZ-PBE.zip, TBG-2.88-DZ-PBE.zip, TBG-1.05-DZ-PBE.zip, TBmos2-7.34-DZ-PBE.zip, TBmos2-4.41-DZ-PBE.zip, TBmos2-2.88-DZ-PBE.zip

HSE06 results: TBG-7.34-DZ-HSE06.zip, TBG-4.41-DZ-HSE06.zip, TBG-2.88-DZ-HSE06.zip, TBmos2-7.34-DZ-HSE06.zip