Neural network representation and optimization of thermoelectric states of multiple interacting quantum dots

We perform quantum master equation calculations and machine learning to investigate the thermoelectric properties of multiple interacting quantum dots (MQD), including electrical conductance, Seebeck coefficient, thermal conductance and the figure of merit (ZT). We show that by learning from the data obtained from the QME, the thermoelectric states of the MQD can be represented well by a two-layer neural network. We also show that after training, the neural network was able to predict the thermoelectric properties of the MQD with much less computational cost compared to the QME approach. Based on the neural network, we further optimize the MQD to achieve a high ZT and power factor. This work presents a powerful route to study, represent, and optimize interacting quantum many-body systems.