Physics-Guided Neural Modeling for Low-Dimensional Thermoelectric Module

A generic framework for combining the thermoelectric (TE) transport in low dimensional materials with neural networks is presented to advance the predictions of thermoelectric module. By training the integrand of transport integrals and implementing into circuit simulators, this approach directly predicts the TE performance under various electronic density of states (DOS) for low dimensional TE modules. This approach enables efficient modeling and produces physically meaningful results for thermoelectric properties.