Prototype code of the tight-binding hamiltonian construction neural network model. Note that it is not necessarily high-performance and well-optimized since it is a prototype code for a research work

tensorflow1.x （1.15） numpy （1.16.0）

Check the example_basic_method.py to construct a TB model for the InSe nanoribbon. (basic TBHCNN model in the manuscrippt)

Check the example_variation_one.py to optimize a given Wannier TB model for 2-D black phosphorus. (Variation1 in the manuscrippt)

Check the example_variation_two.py to construct a TB model in Slater-Koster form for 13-atom-wide armchair graphene nanoribbon (13-AGNR). (Variation2 in the manuscrippt)

[Note that the Variation2 code in model.py now is suitble specifically for the 13-AGNR system because it uses directly the fixed fitting formula for 13-AGNR]

These scripts are self-explanatory and can be easily understood with the manuscript. With the provided default parameter settings, they can be run directly to obtain the corresponding results.

in ./data/input file folder we provide the reference bands data for training the TBHCNN for InSe nanoribbon, Si of the diamond structure, and GaN of the wurtzite structure. Also, we provide the reference

bands data and unoptimized Wannier TB model as a template for training the Variation1 of TBHCNN for 2-D black phosphorus, and we provide the reference bands data and coordinate file (xyz file) for training the Variation2 of TBHCNN for 13-AGNR.

And in ./data/output file folder we store their corresponding result files for the above mentioned systems.

The Input files for the ab-initio and transport calculations within the manuscript.7z file provides the input files necessary to reproduce the results within the manuscrippt