Nature Photonics: Pintus, Dumont, Moody, and Bowers "A New Optical Memory Platform for Super-fast Calculations"
An international team that includes ECE's Paolo Pintus, Mario Dumont, Galan Moody, and John Bowers develops a faster, more durable optical platform for in-memory computation
From the COE News article "A New Optical Memory Platform for Super-fast Calculation"
Constant progress in reducing the size and increasing the performance of the circuits that power computers and smartphones have been the norm for several decades. But Moore’s Law is ending as physical limitations (i.e., the number of transistors that can physically fit on a chip and the increasing problem of heat as a result of packing them ever more densely) is slowing the rate of performance increases. Computing capacity is gradually plateauing, even as artificial intelligence, machine learning, and other data-intensive applications demand ever greater computational power. It has become essential, therefore, to develop novel technologies to address this challenge. A potential solution comes from photonics, which offers lower energy consumption and reduced latency than electronics.
One of the most promising approaches to harnessing the potential of optical computing is in-memory computing, which requires the use of photonic memories. Passing light signals through these memories makes it possible to perform operations nearly instantaneously. Until now, however, solutions proposed for creating such memories have faced challenges that include low switching speeds and limited programmability.
Now, an international team of researchers has developed a groundbreaking photonic platform to overcome those limitations. Paolo Pintus, who works as a project scientist with UC Santa Barbara professor of electrical and computer engineering (ECE) John Bowers and ECE associate professor Galan Moody and is an assistant professor at the University of Cagliari (Sardinia), coordinated the project with University of Pittsburgh assistant professor Nathan Youngblood, Institute of Science Tokyo professor Yuya Shoji, and Mario Dumont, who received his PhD in the Bowers lab last summer. Their findings were published in the October issue of the journal Nature Photonics.
The team utilized a magneto-optical material, cerium-substituted yttrium iron garnet (YIG), the optical properties of which dynamically change in response to external magnetic fields. By employing tiny magnets to store data and control the propagation of light within the material, the researchers pioneered a new class of magneto-optical memories. The innovative platform leverages light to perform calculations at significantly higher speeds and with much greater efficiency than can be achieved using traditional electronics.
This new type of memory has switching speeds one hundred times faster than those of state-of-the-art photonic integrated technology (PIC), they consume about one-tenth the power, and they can be reprogrammed multiple times to perform different tasks. While current state-of-the-art optical memories have a limited lifespan and can be written up to one thousand times, the team demonstrated that magneto-optical memories can be rewritten more than 2.3 billion times, equating to a potentially unlimited lifespan.
“These unique magneto-optical materials make it possible to use an external magnetic field to control the propagation of light through them,” Pontus said. “In this project, we use an electrical current to program micro-magnets and store data. The magnets control the propagation of light within the Ce:YIG material, allowing us to perform complex operations, such as matrix-vector multiplication, which lies at the core of any neural network.”
The authors believe that their findings could mark the beginning of a revolution in optical computing, paving the way for practical applications in the near future.
COE News – "A New Optical Memory Platform for Super-fast Calculation"
