GW Researchers Develop a GHz-Fast Electro-Optical Modulator

6 months ago by Biljana Ognenova

New types of optical neural networks and optical computing hardware might be closer than we think thanks to this super-fast micro-modulator.

Without signal modulators, we would be stuck with limited and expensive signal carrier systems that consume a lot of power to transmit information such as speech or data. 

Given the importance of convenient and affordable information transmitters we rely upon so much in the age of unrestrained and ubiquitous video streaming, the discovery made by researchers from the George Washington University is even more important, propelling the development of next-gen photonic reconfigurable devices and photonic integrated circuits (PIC) with a reduced size. 

 

High-Density Packaging with Indium Tin Oxide  

The research team led by Dr Volker Sorger, an associate professor of electrical and computer engineering at the GW University, has been working for over a decade to finally come up with a new type of electro-optical modulator by adding a thin layer of ITO (indium tin oxide) on a silicon photonic waveguide chip. 

ITO is a transparent conductive oxide, which, thanks to its refractive index, electrical conductivity, and optical transparency,  is widely used in LCDs, touchscreens, polymer-based electronics, smart windows, and thin-film photovoltaics. 

High-density packaging is essential for chip manufacturing. Applying ITO on a silicone-based chip solves the prevalent problem of packaging density for today’s optical modulators, which are typically between 1 millimetre and 1 centimetre in size. 

 

Red and black abstract illustration of a motherboard.
Red and black abstract illustration of a motherboard. 

 

Overcoming Signal Modulation Constraints of Silicon-Based Chips 

Silicon is popularly used as a passive base for PICs, but its light-matter interaction requires for building devices with a larger footprint. Solving the weak electro-optical effect with resonators is possible, but their inclusion restricts the operating range of the device and asks for increased energy consumption due to the integrated thermal elements.

This new compact device which is 1 micrometre in size accelerates signal modulation up to gigahertz speeds or 1 billion cycles per second, promising development of smaller, yet high-performing PICs.

Electrical data is encoded on the device by using tunable plasmonic ITO-based phase shifters that can operate at multiple light wavelengths in the industry-relevant IEEE C band spectrum, in frequencies ranging from 4 to 8 GHz, and travels through the optical modulator based on a Mach-Zehnder interferometer (MZI), creating efficient and fast optical applications.

 

Abstract simulation of fibre optics.
Abstract simulation of fibre optics. 

 

Optical Computing Hardware and Next-Gen Neural Networks

The most exciting potential application of this discovery might be in the area of electro-optic signal converters for industrial IT automation, which is facing the growing need to process information from sensors and convert it into industrial current signals

Fibre cables are already used for robotics control systems. Analogous signal converters could be developed in this area to enable greater torsion, flexing, strength, and bend radius for movable elements of industrial automated components.   

Another thought-provoking application includes new types of communication networks that can replicate complex cognitive processes of the human brain which conventional AI has failed to reproduce so far. 

 

Using Neural Networks to Accelerate AI

Today’s computer-based artificial neural networks require exhaustive computational resources and power consumption. Optical neural networks hold a lot of promise to match the human brain with their capabilities for handling complex computational tasks such as pattern recognition or risk management. 

Hybrid neural networks or more precisely optical hybrid neural networks that traditionally use optics for linear operations could be rearchitected to use micro-sized electro-optical modulators for processing the huge amounts of data produced from non-linear activation functions, necessary for learning and modelling complex data such as images and videos. 

Optics may still be behind electronics for data transmission, but the potential for accelerating AI algorithms remains a compelling research area for applied optics and may yet bring it forward.

Comments