Deep learning of deformation-dependent conductance in thin films: Nanobubbles in graphene

Motivated by the ever-improving performance of deep learning techniques, we design a mixed input convolutional neural network approach to predict transport properties in deformed nanoscale materials using a height map of deformations, as can be obtained from scanning probe measurements, as input. We employ our approach to study electrical transport in a graphene nanoribbon deformed by a number of randomly positioned nanobubbles. Our network is able to make conductance predictions valid to an average error of 4.3%. We find that such low average errors are achieved by a redundant input of energy values, yielding predictions that are 30%-40% more accurate than conventional architectures. We demonstrate that the same method can learn to predict the valley-resolved conductance, with success specifically in identifying the energy at which intervalley scattering becomes prominent. We demonstrate the robustness of the approach by testing the pretrained network on samples with deformations differing in number and shape from the training data. We furthermore employ a graph theoretical analysis of the structure and outputs of the network and conclude that a tight-binding Hamiltonian can be effectively encoded in the first layer of the network, which is supported by numerical findings. Our approach contributes a theoretical understanding and a refined methodology to the application of deep learning for the determination of transport properties based on real-space disorder information.

Automated summary

Motivated by the improvement of deep learning techniques, we design a mixed input convolutional neural network to predict transport properties in deformed nanoscale materials using a height map of deformations as input. Our network achieves higher accuracy in conductance predictions by using redundant input of energy values, and it successfully predicts valley-resolved conductance.