Supplementary data for "An Atlas of Chirality-Dependent Electronic Structures of MoS2 Nanotubes from Deep Learning"

Training datasets for fine tune CHGNet MLIP and DeepH-E3 model for systematically predicting electronic structures of Mo2 nanotubes. Processed_Data_Train_DeepHE3.zip contains the training data for DeepH-E3 model. Dataset_for_finetune_CHGNet.zip is a json file contains structures, energies, forces, stress that used to fine tune CHGNet v0.3.0, which contains MoS2 monolayer and nanotubes (zigzag, chiral and armchair).Production_prediction_info_upload_figshare.csv contains all results info of nanotubes in the production prediction. Units for D0 and D are in Angstrom, Strain_MoMo, Strain_SS_Inner and Strain_SS_Outer are in %, Theta_degree is in degree, deephe3_gap is in EV, e_effmass and h_effmass are in m0, Delta_K_lumo_homo are in 1/Angstrom. All the data in this csv table are kept to six decimal places.Data for reproducing paper of: An Atlas of Chirality-Dependent Electronic Structures of MoS2 Nanotubes from Deep LearningThe trained models of CHGNet and DeepH-E3 for MoS2 from these datasets are on github: https://github.com/mmdl-westlake/ML4MoS2NT_ElectronicStructures.