Stanford University researchers developed an optical device that enables the fine-tuning and changing of frequencies of individual photons in a stream of light, to virtually any mixture of colors. The new photonic architecture has potential applications and large implications for fields ranging from digital communications and artificial intelligence to quantum computing.
The structure consists of a low-loss wire for light, which acts as a sort of highway, with the photons being analogous to cars. The photons then enter a series of rings, similar to off-ramps in a highway cloverleaf. Each ring has a modulator that transforms the frequency of the passing photons. The device can be designed with as many rings as desired, and engineers are able to finely control the modulators to dial in the desired frequency transformation.
“This powerful new tool puts a degree of control in the engineer’s hands not previously possible,” said Shanhui Fan, senior author on the paper describing the work and a professor of electrical engineering at Stanford.
The color of a photon is determined by the frequency at which the photon resonates, which in turn is a factor of its wavelength. A red photon, for example, has a relatively low frequency and a wavelength around 650 nm. Blue has a much higher frequency with a wavelength of about 450 nm.
A simple transformation with this technology may entail shifting a photon from a wavelength of 500 to 510 nm, for example. As the human eye would register it, this would be a change from cyan to green.
The system is capable of much more complex and sophisticated transformations; to explain, Fan offered an example of an incoming light stream composed of 20% photons in the 500-nm range and 80% at 510 nm. With this new device, an engineer could fine-tune that ratio to 73% at 500 nm and 27% at 510 nm, all while preserving the total number of photons. The ratio could be 37% and 63%, for that matter. In quantum applications, where a single photon may have multiple colors, the device could change the ratio of those colors, too.
“We say this device allows for ‘arbitrary transformation,’ but that does not mean ‘random,’” said Siddharth Buddhiraju, first author on the paper and a graduate student in Fan’s lab during the research; he now works at Facebook Reality Labs. “Instead we mean that we can achieve any linear transformation that the engineer requires. There is a great amount of engineering control here.”
Among the applications the researchers envision are optical neural networks for AI, which perform neural computations with light rather than electrons. Existing methods for creating optical neural networks don’t actually change the frequencies of the photons, but simply reroute photons of a single frequency. Performing such neural computations through frequency manipulation could lead to much more compact devices, the researchers said.