30-09-2021 | | By Sam Brown
Recently, researchers from IBM and Skoltech have announced the development of a hyper-efficient optical switch that can operate faster than current transistors while using less energy. What is optical computation, what did the researchers develop, and could optical parts help replace electronics switches in the future?
Optical computers are devices whose computational logic is powered using light (unlike traditional computers that use electricity). Optical computers can combine electrical and photonic components together, and such devices can also use light to control electrical flow or vice versa. Still, a true optical switch would control a light source with another light source.
Optical computers would exhibit multiple advantages compared to traditional computers due to the properties of light. The first major advantage is that light-based components typically have a higher SNR than electrical components. The second major advantage is that photons are able to operate at far higher frequencies than current flow, meaning that photonic components can theoretically operate at high frequencies and therefore produce faster processors.
However, one common misconception of photonic systems is that they are faster because light travels faster in a vacuum than electrical flow in conductors. Light is often guided through waveguides made of plastic or glass (such as fibre optic cables), which results in light travelling slower than an electrical current flowing through copper cables. Despite this slower speed, optic fibre cables can handle more data because of the lower SNR and the ability to send a wide range of different frequencies of photons down the same waveguide without interference.
Recently, researchers from Skoltech and IBM announced the development of a functioning optical switch that is extremely energy efficient and can switch from a control signal consisting of individual photons. The high speed of the switch (with a switching speed of 1 trillion operations per second) puts the new device over 1000 times faster than current transistors used in high-end CPUs. Furthermore, the new switch can operate at room temperature, removing the need for complex cooling mechanisms.
The switch itself has an amplification factor of 23,000, meaning that a single photon can result in the production of 23,000 photons, and this makes the switch ideal for high-frequency operation and amplification of low-light sources.
The construction of the switch itself consists of a 35nm thin organic semiconductor polymer surrounded by a highly reflective structure. This structure forms a microcavity, and the purpose of this microcavity is to trap incoming light for as long as possible. The result from photons trapped in the microcavity is photons interacting with excitations in the semiconductor, creating quasiparticles called excitation-polaritons.
Powering the switch utilises a primary bright laser that pumps the microcavity to produce thousands of identical quasiparticles, resulting in a Bose-Einstein condensate representing a 1 or 0. Controlling the state of the switch is done using a second beam that hits the microcavity before the main power beam, and this beam (which can consist of a single photon) changes the state of the quasiparticles before the main beam amplifies the signal.
The researchers noted that while single photons can control the switch, tens of photons provides a better SNR, which is often more important than increased energy efficiency. Furthermore, the researchers used clever techniques to help reduce the power consumed, such as utilising vibrations in the semiconductor to help pump the laser and carefully choosing a wavelength of light that best worked with their switch.
The device produced by the researchers is far from a practical device that could be incorporated into modern electronics (even the researchers noted this). However, the fact that it can operate at such high speeds while consuming little energy all while at room temperature is a significant step towards optical devices.
However, traditional electronic components have many advantages that optical devices simply do not have. For example, electric current can generate magnetic fields, making electricity ideal for controlling actuators and other mechanical systems. Secondly, electricity can be converted into a wide variety of energy forms, including light, heat, and sound. This is highly practical for peripheral devices and interacting with computers (such as monitors and keyboards).
If optical components find their way into commercial devices, they will most likely be in the form of co-processors that work with electrical circuits. For example, a math co-processor could be engineered using optical parts and be designed to handle specific mathematical operations that would normally be difficult to compute on standard silicon.
The researchers have demonstrated an amazing feat of engineering and could pave the way to practical optical components. If engineers can find a way to mount such devices onto semiconductors for use on a PCB, the optical technology race will start.