Photonic integrated circuits (PICs) can enable high-speed data transmission, low power consumption, and a compact size, making them suitable for integration into edge devices. They are increasingly being used in signal processing for edge AI and sensor applications.
Silicon photonics is an expansive technology. It’s based on CMOS processing, heterogeneous integration, and advanced optical functionality. Silicon-on-insulator (SOI) is a key enabler. Some of the key elements of PICs include (Figure 1):
- Optical wave guides can be made with silicon or silicon nitride and enable efficient on-chip optical connectivity.
- Optical ring resonators serve as a fundamental building block and can be integrated with optical filters, modulators, multiplexers, and frequency comb generators. There are also more specialized designs, such as Fabry-Pérot resonators for lasers and interferometers, and Whispering Gallery Mode resonators, which are optimal for sensing and nonlinear optics functions.
- Modulators are used to manage photonics properties, such as phase, polarization, and intensity, to optimize PIC performance.
- Photodetectors provide connectivity between the optical and electrical sections of PICs and the outside.
- Optical coupling elements are used to combine, split, or redistribute optical signals. Grating couplers and edge couplers are the most common due to their unique advantages. Grating couplers are compatible with CMOS processing, while edge couplers require additional back-end processing steps, such as cleaving and polishing, but can provide connectivity to external lasers and electronics.
Figure 1. Examples of common PIC functional elements. (Image: Santec)
PICs in edge AI
PICs can support lower latencies for applications like LIDAR in autonomous vehicles and robots, enabling safer navigation. PICs consume significantly less energy than conventional ICs, an important consideration in energy-constrained edge devices.
Further energy savings can be realized since processing data on the edge device reduces the need for cloud connectivity and the associated energy requirements. Eliminating or minimizing the need for wireless connectivity also makes edge devices more resilient.
PICs in advanced sensors
PICs can be used in highly sensitive and compact sensors for edge applications. PICs, combined with edge AI, can implement real-time object and facial recognition, as well as other complex image processing functions, directly on the sensor chip.
PIC sensors are being used for environmental monitors that can detect pollutants and measure air or water quality. They are also used in food processing to measure parameters like ripeness and nutrient content.
PIC lab on a chip
PIC sensors can enable rapid diagnostics of medical and environmental conditions in the field, eliminating the need for remote laboratories. In many cases, they also have higher sensitivities and provide superior results. For some materials, attomolar concentrations, 10-18 moles per liter, or one quintillionth of a mole of a substance dissolved in one liter of solution, can be reliably detected with PIC sensors.
Integrated optical sensors for these applications are usually fabricated using silicon nitride (SiN). SiN waveguides have good sensitivity over the visible to near-infrared range and can be fabricated with a small bend radius. This enables a very long sensor to be “rolled up,” occupying very little space on the surface of a PIC. Longer sensors have higher sensitivity.
Some designs feature a coating on the sensor that alters its refractive index when it encounters target molecules. Changes in the refractive index are detected using photonic transducers and connected to the rest of the lab-on-a-chip using optical waveguides (Figure 2).
Figure 2. Lab-on-a-chip biosensors like this, based on PICs, can provide real-time data for medical diagnostics. (Image: Aventier)
Other photonics platforms
PICs are a versatile technology and can be made with materials beyond Si and SiN. Many PICs incorporate silica (SiO2) for functions like planar optical waveguides.
Figure 3. This tunable laser system utilizes hybrid integration, combining a low-loss SiN PIC with a high-performance active gain InP PIC. (Image: PhotonDelta)
Lithium Niobate (LiNbO3) can be used for making low-loss modulators. Its low optical index and broad transparency window make it well-suited for matching fiber inputs and outputs. Lithium Niobate on Insulator (LNOI) technology is under development for future PIC designs.
System cost and performance can be simultaneously optimized by mixing PIC technologies using heterogeneous or hybrid integration. For example, a tunable laser system has been developed for military systems that combines a low-loss SiN PIC with a high-performance active gain Indium Phosphide (InP) PIC (Figure 3).
Summary
PICs are being utilized in a growing range of edge applications, including data processing and sensor integration. They can reduce energy consumption and enhance processing capabilities, and can be implemented on single chips or through heterogeneous integration, which incorporates multiple semiconductors and sensor technologies.
References
Designing a Photonic Integrated Circuit: Best Practices for Simulation & Layout, SimuTech
Exploring Photonic Integrated Circuits and Optical ICs, Avantier
High-performance analog signal processing with photonic integrated circuits, Light Science and Applications
Introduction to Silicon Photonics, Santec
Photonic Computing Takes a Step Toward Fruition, APS Physics
Photonic Integrated Circuit (PIC), AyarLabs
What is a Photonic Integrated Circuit?, Ansys
What is a Photonic Integrated Circuit?, PhotonDelta
What is Integrated Photonics?, AIM Photonics
EEWorld Online related content
How are single photon sensors used in quantum computing?
Accelerating high-performance AI workloads with photonic chips
What’s the difference between a VCSEL and PCSEL?
How can in-package optical interconnects enhance chiplet generative AI performance?
How does UCIe on chiplets enable optical interconnects in data centers?
Filed Under: FAQ, Featured
