(Nanowerk News) Photonic chips have revolutionized data-heavy technologies. Alone or in conjunction with traditional electronic circuits, these laser-powered devices transmit and process information at the speed of light, making them a promising solution for data-hungry applications of artificial intelligence.
In addition to their unmatched speed, photonic circuits use far less energy than electronic circuits. Electrons move relatively slowly through the hardware, colliding with other particles and generating heat, while photons flow without losing energy, producing no heat at all. Unencumbered by the energy losses inherent in electronics, integrated photonics is poised to play a major role in continuous computing.
Photonics and electronics are separate fields of science and employ different architectural structures. Both, however, rely on lithography to determine the elements of their sequence and connect them in sequence. While photonic chips don’t use transistors that fill the increasingly shrinking and increasingly layered grooves of electronic chips, their complex lithographic patterns guide laser light through coherent circuits to form photonic networks that can perform computational algorithms.
But now, for the first time, researchers at the University of Pennsylvania School of Engineering and Applied Science have created a photonic device that provides programmable on-chip information processing without lithography, offering photonic speed coupled with superior accuracy and flexibility for artificial intelligence. applications (AI).
Achieving unparalleled light control, the device comprises spatially distributed optical gain and loss. The laser emits light directly on the semiconductor wafer, without the need for a defined lithographic path.
Liang Feng, Professor in the Departments of Materials Science and Engineering (MSE) and Electrical Systems and Engineering (ESE), along with Ph.D. student Tianwei Wu (MSE) and postdoctoral fellows Zihe Gao and Marco Menarini (ESE), introduce the microchip in a recent study published in Nature Photonics (“Reconfigurable, lithography-free integrated photonic processor”).
Silicon-based electronics systems have changed the computing landscape. But they have clear limitations: they are slow at processing signals, they work through data serially and not in parallel, and they can only be miniaturized to a certain extent. Photonics is one of the most promising alternatives because it can overcome all these drawbacks.
“But photonic chips aimed at machine learning applications face the constraints of a complicated fabrication process in which lithographic patterns are fixed, limited in reprogramming capabilities, prone to errors or damage, and expensive,” said Feng. “By eliminating the need for lithography, we are creating a new paradigm. Our chip overcomes those bottlenecks and offers better accuracy and best reconfigurability given the removal of all kinds of constraints from predefined features.”
Without lithography, the chip becomes a customizable data processing powerhouse. Since the pattern is not predefined and engraved, the device is intrinsically free from defects. Perhaps more impressively, the lack of lithography makes the microchip impressively reprogrammable, able to adjust its laser printed pattern for optimal performance, whether tasks are simple (multiple inputs, small data set) or complex (multiple inputs, large data set) tasks.
In other words, the complexity or minimalism of the device is a kind of living thing, adaptable in ways that scratched microchips cannot.
“What we have here is something very simple,” said Wu. “We can build and use it very quickly. We can integrate it easily with classic electronics. And we can reprogram it, changing laser patterns on the fly to achieve real-time reconfigurable computing for on-chip training of AI networks.”
A simple piece of semiconductor, this device is incredibly simple. It is this manipulation of the material properties of the plates that is the key to the research team’s breakthrough in projecting lasers into dynamically programmable patterns to reconfigure the computational functions of photonic information processors.
This ultimate reconfiguration capability is critical for real-time machine learning and AI.
“The interesting part,” says Menarini, “is how we control the light. Conventional photonic chips are technologies based on passive materials, meaning the materials scatter light, reflecting it back and forth. Our material is active. The pumping beam modifies the material in such a way that when the signal beam arrives, it can release energy and increase the amplitude of the signal.”
“This active nature is the key to this science, and the necessary solution to achieve our lithography-free technology,” added Gao. “We can use it to reroute optical signals and program optical information processing on the chip.”
Feng compared the technology to an artistic tool, a pen for drawing on a blank page.
“What we have achieved is exactly the same: pumping light is our pen for drawing photonic computing networks (images) on a non-patterned sheet of semiconductor wafer (blank page).”
But unlike indelible lines of ink, these beams of light can be drawn and redrawn, their patterns tracing countless paths into the future.