Inverse design of organic light-emitting diode structure based on deep neural network

Abstract

In this work, we present a neural network that receives the layer thicknesses and refractive indices as the inputs and predicts the LEE response of a given OLED structure. The network is trained with a dataset containing 260,000 structure parameter-LEE pairs which were obtained from the in-house Chance-Prock-Silbey (CPS) model. The neural network was able to complete the spectrum predictions with 4×106 faster rate than the rigorous electromagnetic simulation based on CPS formulation depending on the degree of parallelization with RMSE of 1.86×10-3. Two different routes were taken for the inverse design of LEE responses. One was the inverse neural network built by joining additional layers to the pre-trained forward neural network and the other was the neural network assisted genetic algorithm (NNGA) which uses the forward network as the platform for LEE prediction. NNGA showed superior performances in the inverse design of both existing and non-existing LEEs whereas the inverse neural network was unmatched in terms of the computation speed. The NNGA also successfully solves two representative OLED optimization problems: maximizing LEE and minimizing angle-dependent white color variation. Considering that both problems possess large design spaces that are difficult to be addressed by conventional numerical approaches employing rigorous electromagnetic simulations, the neural network-based methodology presented in this work provides a promising platform for tackling computation-heavy optimization tasks with one-time computation cost. Furthermore, different aspects of the neural network including its one-to-many mapping behaviour, extrapolability, and comparison between the inverse design methods were investigated extensively.