Artificial neural networks have shown effectiveness in the inverse design of nanophotonic structures; however, the numerical accuracy and algorithm efficiency are not analyzed adequately in previous reports. In this Letter, we demonstrate the convolutional neural network as an inverse design tool to achieve high numerical accuracy in plasmonic metasurfaces. A comparison of the convolutional neural networks and the fully connected neural networks show that convolutional neural networks have higher generalization capabilities. We share practical guidelines for optimizing the neural network and analyzed the hierarchy of accuracy in the multi-parameter inverse design of plasmonic metasurfaces. A high inverse design accuracy of ± 8 n m for the critical geometrical parameters is demonstrated.

The performance of AlGaN-based light-emitting diodes (LEDs) emitting at UVA-UVC regions can be severely compromised due to the polarization difference (∆P) between the last quantum barrier (LQB) and the electron blocking layer (EBL). In this work, the different situations of the bandgap difference (∆Eg) and ∆P of InAlN/AlGaN and AlGaN/AlGaN heterojunctions fully strained on GaN and AlN substrates are discussed. It shows that the InAlN/AlGaN heterojunctions could produce positive or negative sheet charges at the heterointerface under ∆Eg >0, which could not be realized by the conventional AlGaN/AlGaN heterojunctions. To demonstrate and utilize the feature, the polarization-modulated InAlN LQBs with 0.14-0.16 indium compositions of 320 nm UVB LEDs are designed and investigated. It is observed that the InAlN LQBs could replace the conventional AlGaN LQB to improve electron confinement and hole injection by affecting effective barrier heights. By modulating the LQB/EBL polarization using InAlN, the proposed UV LED has a 32% enhancement in internal quantum efficiency and lower efficiency droop (from 16.9% to 0.7%) compared with the conventional one without modulation. The operation voltage at the same current also significantly decreases. The improvement of optical output power and wall plug efficiency at 60 mA in proposed structures are near 90% and 100%, respectively. This study provides a novel and highly effective methodology for development of high efficiency UV LEDs.

No abstract