Quantization-aware Photonic Homodyne computing for Accelerated Artificial Intelligence and Scientific Simulation

Modern problems in high-performance computing, ranging from training and inferencing deep learning models in computer vision and language models to simulating complex physical systems with nonlinearly-coupled equations, require exponential growth of computational resources. Photonic analog systems are emerging with solutions of intrinsic parallelism, high bandwidth, and low propagation loss. However, their application has been hindered by the low analog accuracy due to the electro-optic distortion, material nonlinearities, and signal-to-noise ratios. Here we overcome this barrier with a quantization-aware digital-photonic mixed-precision framework across chiplets for accelerated AI processing and physical simulation. Using Lithium Niobate photonics with channel equalization techniques, we demonstrate linear multiplication (9-bit amplitude-phase decoupling) in homodyne optical logics with 6-bit precision at the clock rate of 128 giga-symbol-per-second (128 GS/s), enabling AI processing with 6 ns latency. Codesign hardware-algorithms, including iterative solvers, sparse-dense quantization, and bit-sliced matrix multiplication, explore photonic amplitude and phase coherence for complex-valued, physics-inspired computation. In electromagnetic problems, our approach yields 12-bit solutions for partial differential equations (PDEs) in scattering problems that would conventionally require up to 32-bit and often even 64-bit precision. These results preserve digital-level fidelity while leveraging the high-speed low-energy photonic hardware, establishing a pathway toward general-purpose optical acceleration for generative artificial intelligence, real-time robotics, and accurate simulation for climate challenges and biological discoveries.

💡 Research Summary

The paper introduces a mixed‑precision photonic‑digital computing platform that overcomes the long‑standing accuracy barrier of analog optical processors. Using thin‑film lithium‑niobate (TFLN) electro‑optic modulators, the authors achieve independent amplitude and phase control, enabling 9‑bit amplitude‑phase decoupling and 6‑bit complex multiplication at an unprecedented 128 GS/s symbol rate. A homodyne detection scheme with a 50:50 coupler and balanced photodiodes simultaneously extracts the real and imaginary components of the product, while low‑speed integrated photodiodes perform time‑integrated readout with sub‑6 ns latency.

To address high‑speed distortions, a frequency‑domain channel equalization is applied: the system’s transfer function H(f) is measured via homodyne detection, and a pre‑emphasis filter 1/H(f) is applied to the input data. This reduces the standard deviation of encoding errors from ~6 % to ~1.6 %, effectively delivering ~7‑bit quantization accuracy even at 40 GS/s.

On the algorithmic side, the authors co‑design quantization‑aware training and bit‑sliced matrix multiplication for complex‑valued neural networks. In a two‑layer complex‑valued ONN for MNIST classification, channel equalization improves inference accuracy from 90.1 % to 93.4 % (digital baseline 94.8 %). A single‑layer real‑valued network runs at 128 GS/s, completing a 784‑by‑10 matrix‑vector multiply in 6.125 ns with a 4 % mean error.

For scientific simulation, the platform solves complex‑valued partial differential equations governing electromagnetic scattering. By encoding the field and material parameters as complex vectors, the system delivers 12‑bit solution fidelity within 30 µs, matching the precision of conventional 32‑ to 64‑bit digital solvers.

Energy efficiency is highlighted: the TFLN modulators operate at 2 V with sub‑10 mW power, supporting >100 GS/s data throughput. Integrated ADCs with ≥10‑bit resolution keep the digital‑optical interface low‑power. Overall, the work demonstrates that a quantization‑aware, mixed‑precision photonic architecture can provide high‑speed, low‑energy, and high‑precision computation for both AI inference and physics‑inspired simulations, opening a pathway toward general‑purpose optical accelerators for generative AI, real‑time robotics, climate modeling, and biomedical discovery.