Programmable integrated photonics is one of the most promising hardware acceleration schemes for deep learning. The programing method based on phase change materials makes the device non-volatile but reduces the resolution, which may result in a loss of performance. Here we propose an on-chip optical neural network implemented with a cascaded array of 1:1 couplers and waveguides with a digital non-volatile phase shifter. The Sb2Se3-assisted phase shifters can achieve a phase shift from 0 to 2π in steps of π/16 through 5 independent non-volatile reconfigurable units. We also proposed a Digital-Aware-Training regime to train this digital model. The network achieves 100% and 83.7% accuracy in the recognition of 4 Latin letters and iris project, respectively. Compared with conventional reconfigurable schemes, this computing platform has the characteristics of non-volatile, low power consumption, and high modulation robustness.
Ultra-compact optical switch is a crucial module for densely integrated on-chip network. In recent years phase-change materials (PCMs) enable design of ultra-compact reconfigurable devices due to their drastic refractive index change during phase transition. In this work we inverse design a photonic-crystal-like (PhC-like) subwavelength optical switch assisted by novel ultralow loss PCM: Sb2Se3. At amorphous state the refractive index of Sb2Se3 is similar to that of silicon, so the device work as a MMI and light outputs at output port 2. At crystalline state the refractive index of Sb2Se3 increases to 4.05 and the subwavelength structure guides light to output port 1. The proposed switch has a compact footprint of 2.75 × 16 μm2. At the center wavelength of 1550nm, the simulated insertion losses and crosstalks are 0.36 dB, -44.36 dB, 0.49 dB, and -40.37 dB under amorphous and crystalline states respectively.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.