What problem does this paper attempt to address?

The problems that this paper attempts to solve are several key challenges faced when using Mach - Zehnder interferometer (MZI) - based photonic multipliers in photonic neural networks (PNNs). Specifically, these challenges include: 1. **Optical Loss**: Traditional MZIs have inevitable optical losses during transmission, which will affect the performance and scalability of PNNs. 2. **Crosstalk Noise**: Crosstalk noise between MZIs will further reduce the integrity of the signal, limiting the precision and scalability of PNNs. 3. **Fabrication Process Variations (FPVs)**: Minor changes in the manufacturing process will lead to unstable performance of MZIs, thereby affecting the overall performance of PNNs. 4. **Size Problem**: The size of traditional MZIs is relatively large, which is not conducive to integration and miniaturization, limiting the practicality and efficiency of PNNs. To address these issues, the paper proposes an optimized compact and low - loss MZI design based on phase - change materials (PCM). By optimizing the multimode interferometer (MMI) splitter and combiner and integrating a compact and lossless PCM phase modulator, this design significantly improves the performance of MZIs, thereby enhancing the scalability and accuracy of PNNs. Specific improvements include: - **Low Loss**: The designed MZI has a low loss of 0.2 dB. - **Low Crosstalk**: The crosstalk coefficient reaches - 38 dB. - **Compact Size**: The total length is only 52 μm. - **Robustness to Fabrication Process Variations**: It can maintain high performance even in the case of fabrication process variations. Through these optimizations, the paper shows that using the optimized PCM - MZI can significantly improve the inference accuracy of PNNs, especially more obvious in large - scale networks. Experimental results show that, compared with traditional MZIs, using the optimized PCM - MZI can reduce the loss of inference accuracy from 98% to 6%.

Compact and Low-Loss PCM-based Silicon Photonic MZIs for Photonic Neural Networks

Analysis of Optical Loss and Crosstalk Noise in MZI-based Coherent Photonic Neural Networks

Mastering Silicon Photonics Device Design for Scalable and Robust Optical Neural Networks

High-speed, low-loss silicon Mach-Zehnder modulators with doping optimization.

High-Baudrate Silicon Photonics Ring Resonator and Mach-Zehnder Modulators for Short-Reach Applications

An Implementation of Complex-Valued Matrix-Vector Multiplication Based on MZI Array for Photonic Neural Networks

Easily Scalable Photonic Tensor Core Based on Tunable Units with Single Internal Phase Shifters

Towards Phase-Error- and Loss-Tolerant Programmable MZI-Based Optical Processors for Optical Neural Networks

Design of a photonic unitary neural network based on MZI arrays

High speed silicon Mach-Zehnder modulator based on interleaved PN junctions.

Clements-Enhanced Complex-Valued Coherent Mesh with Balanced Detection Units for Photonic Neural Networks

Compact silicon photonic Mach-Zehnder modulators for high-energy physics

Ultra‐Compact and NonVolatile Nanophotonic Neural Networks

Low-loss Chip-Scale Programmable Silicon Photonic Processor

An integrated optical neural network chip based on Mach-Zehnder interferometers

High-speed programmable photonic circuits in a cryogenically compatible, visible-NIR 200 mm CMOS architecture

Characterization and Optimization of Integrated Silicon-Photonic Neural Networks under Fabrication-Process Variations

Ultracompact programmable silicon photonics using layers of low-loss phase-change material Sb$_2$Se$_3$ of increasing thickness

100 GHz Micrometer-Compact Broadband Monolithic ITO Mach-Zehnder Interferometer Modulator Enabling 3500 Times Higher Packing Density

Low-Loss Calibration-Free 2 × 2 Mach-Zehnder Switches With Varied-Width Multimode-Interference Couplers

Calibration-Free Silicon Photonic Non-Blocking 6 × 6 Mach-Zehnder Switch