Search for: All records

Semiconductor moiré superlattices, characterized by their periodic spatial light emission, unveil a new paradigm of engineered photonic materials. Here, we show that ferroelectric moiré domains formed in a twisted hexagonal boron nitride (t-hBN) substrate can modulate light emission from an adjacent semiconductor MoSe₂monolayer. The electrostatic potential at the surface of the t-hBN substrate provides a simple way to confine excitons in the MoSe₂monolayer. The excitons confined within the domains and at the domain walls are spectrally separated because of a pronounced Stark shift. Moreover, the patterned light emission can be dynamically controlled by electrically gating the ferroelectric domains, introducing a functionality beyond other semiconductor moiré superlattices. Our findings chart an exciting pathway for integrating nanometer-scale moiré ferroelectric domains with various optically active functional layers, paving the way for advanced nanophotonics and metasurfaces. 

We demonstrate a new approach to confining excitons in a M oSe₂monolayer via the electrostatic potential from a twisted hBN substrate, which offers more flexibility in controlling exciton properties in a moiré superlattice. 

Recent advances in computer vision has led to a growth of interest in deploying visual analytics model on mobile devices. However, most mobile devices have limited computing power, which prohibits them from running large scale visual analytics neural networks. An emerging approach to solve this problem is to offload the computation of these neural networks to computing resources at an edge server. Efficient computation offloading requires optimizing the trade-off between multiple objectives including compressed data rate, analytics performance, and computation speed. In this work, we consider a “split computation” system to offload a part of the computation of the YOLO object detection model. We propose a learnable feature compression approach to compress the intermediate YOLO features with lightweight computation. We train the feature compression and decompression module together with the YOLO model to optimize the object detection accuracy under a rate constraint. Compared to baseline methods that apply either standard image compression or learned image compression at the mobile and perform image de-compression and YOLO at the edge, the proposed system achieves higher detection accuracy at the low to medium rate range. Furthermore, the proposed system requires substantially lower computation time on the mobile device with CPU only.