Photonic Matrix-Vector Multiplication With Low-Insertion-Loss and Non-Volatile Ge2Sb2Te3S2 Intensity Modulators
- 4 June 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Lightwave Technology
- Vol. 42 (12) , 4347-4354
Abstract
We propose a low-insertion-loss (IL) and non-volatile optical intensity modulator based on Ge 2 Sb 2 Te 3 S 2 (GSTS), a recently proposed Se-free wide-gap phase change material. Owing to the negligibly small optical absorption of amorphous GSTS at a wavelength of 1550 nm, a non-volatile intensity modulator with a low IL of less than 0.08 dB and a high extinction ratio (ER) of 19.6 dB is realized using GSTS with a device length of 8 μm, overcoming the trade-off relationship between IL and ER in conventional Ge 2 Sb 2 Te 5 (GST). Using visible laser light irradiation, we demonstrate multilevel modulation up to 24 levels by taking advantage of the small optical absorption of amorphous GSTS and the moderately large optical absorption of crystalline GSTS compared with GST. With the proposed GSTS intensity modulators, we demonstrate 4×4 matrix-vector multiplication using a recently proposed photonic tensor core, which shows a high tolerance to hardware error accumulation. On the basis of the experimental results, we numerically confirm that a high recognition accuracy of 97.33% can be obtained using a convolutional neural network with emulated optical convolutional operations.Keywords
Funding Information
- JST-CREST (JPMJCR2004)
- JST-Mirai Program (JPMJMI20A1)
- JSPS KAKENHI (JP20H02198, JP21J20272, JP22K14298, JP23H00172)
- Advanced Research Infrastructure for Materials and Nanotechnology in Japan
- Ministry of Education, Culture, Sports, Science and Technology (JPMXP1222UT1028)
This publication has 33 references indexed in Scilit:
- Si Microring Resonator Crossbar Array for On-Chip Inference and Training of the Optical Neural NetworkACS Photonics, 2022
- Photonic matrix multiplication lights up photonic accelerator and beyondLight: Science & Applications, 2022
- Optical coherent dot-product chip for sophisticated deep learning regressionLight: Science & Applications, 2021
- Non‐Volatile Reconfigurable Integrated Photonics Enabled by Broadband Low‐Loss Phase Change MaterialAdvanced Optical Materials, 2021
- An optical neural chip for implementing complex-valued neural networkNature Communications, 2021
- A New Family of Ultralow Loss Reversible Phase‐Change Materials for Photonic Integrated Circuits: Sb2S3 and Sb2Se3Advanced Functional Materials, 2020
- Nonvolatile Electrically Reconfigurable Integrated Photonic Switch Enabled by a Silicon PIN Diode HeaterAdvanced Materials, 2020
- Broadband transparent optical phase change materials for high-performance nonvolatile photonicsNature Communications, 2019
- Miniature Multilevel Optical Memristive Switch Using Phase Change MaterialACS Photonics, 2019
- Deep learning with coherent nanophotonic circuitsNature Photonics, 2017