Energy-Efficient Nonlinear Optical Micro-Device Arrays

Patent Status

Patent Pending

Brief Description

Optical neural networks (ONNs) are a promising computational alternative for deep learning due to their inherent massive parallelism for linear operations. However, the development of energy-efficient and highly parallel optical nonlinearities, a critical component in ONNs, remains an outstanding challenge.

To address this situation, researchers at UC Berkeley and Berkeley National Lab developed a nonlinear optical microdevice array (NOMA) compatible with incoherent illumination by integrating the liquid crystal cell with silicon photodiodes at the single-pixel level. The researchers fabricated NOMA with over half a million pixels, each functioning as an optical analog of the rectified linear unit at ultralow switching energy down to 100 femtojoules/pixel. The team demonstrated an optical multilayer neural network.

This work holds promise for large-scale and low-power deep ONNs, computer vision, and real-time optical image processing.

Suggested uses

• Large-scale and low-power deep optical neural networks: The array's design is ideal for building expansive and energy-efficient neural networks that use light for computation.

• Computer vision: The NOMA can be applied in systems that require fast and efficient processing of visual data.

• Real-time optical image processing: The device's high-speed, parallel processing capabilities make it suitable for applications needing instantaneous image manipulation and analysis.

Advantages

• Ultralow energy consumption: Each pixel operates at a remarkably low switching energy of 100 femtojoules, making the device highly energy-efficient.

• High parallelism and scalability: The array contains over half a million pixels, enabling massive parallel operations, which is crucial for deep learning and other computationally intensive tasks.

• Integration with incoherent light: The device's compatibility with incoherent illumination expands its usability and simplifies system design.

Related Materials