With tiny hardware and antennas, a Virginia Commonwealth University professor hopes to usher in a new era of electronics design – one that could power the future of artificial intelligence, medical implants and more.
Supriyo Bandyopadhyay, Ph.D., professor of electrical and computer engineering in VCU’s College of Engineering, has built a portfolio of technologies, including hardware “matrix multipliers” that drive AI algorithms and minuscule antennas that use nanomagnets.
In the past year, his efforts have earned grants from the Virginia Innovation Partnership Corp. and VCU’s own Commercialization Fund. His matrix multiplication and antenna technologies are also supported by more than $450,000 in grants from the National Science Foundation.
“Supriyo’s research centers on cutting-edge concepts in electronics design,” said Brent Fagg, senior licensing manager at VCU TechTransfer and Ventures, which has submitted patents for Bandyopadhyay’s IP and provided funding for his work. “He is not merely redesigning the way things have been done but completely changing the way electronics components are created and used.”
Faster, more efficient and safer AI
In AI applications like Siri, Alexa, self-driving cars or text creator ChatGPT, the primary mathematical function to make it all work is called matrix multiplication. It’s a complex, intensive process typically done through software. “But software is sluggish, takes a ton of energy, it’s expensive, and it’s vulnerable to cyberattacks,” Bandyopadhyay said.
Instead, he has built matrix multipliers from tiny hardware — hundreds of times smaller than the width of a human hair — that allow for greater energy efficiency, compactness and resilience against cyberattacks.
“Trillions of these pieces will be required to power AI engines in the next decade,” he said.
Existing hardware matrix multipliers use signal-amplifying transistors and are volatile — meaning once power is cut, data is lost, and they rely on a network connection to pull from the cloud. Bandyopadhyay’s multipliers, meanwhile, are nonvolatile – they use magnets that don’t lose data when powered off. “A nonvolatile matrix multiplier is very attractive” for companies that make these devices, he said.
Beyond faster and more efficient processing, the increased safety of hardware multipliers is a standout benefit, Bandyopadhyay said, as AI becomes more prevalent in vehicles and computing systems that require critical security. The nonvolatile nature would also allow users — say in deep space, or deep mines — to access information without connecting to a network.
Riding the wave with better antennas
In applications requiring electromagnetic waves to pass through soil, ice, water, walls and other objects, the frequency must be kept low. (High-frequency waves attenuate rapidly with distance when moving through these elements.). Hence antennas that radiate at low frequencies are preferred for these applications, unlike in NextG networks where higher frequencies are preferred. Lower frequency electromagnetic waves have longer wavelengths. Unfortunately, in order to radiate effectively, a conventional antenna must be comparable in size to the wavelength. So an antenna radiating efficiently and sending out a strong signal at 1 megahertz must be about 300 meters long because the electromagnetic wavelength at 1 megahertz is 300 meters.
This is “an inherent shortcoming” to pushing out low-frequency signals, Bandyopadhyay said. “Imagine taking an antenna that large down a mine shaft or putting it in a submarine.”
Many applications can’t use a large antenna but still need a strong signal at a low frequency — for example, a medically implanted device.
“A medically implanted antenna cannot radiate at gigahertz frequencies since it would harm human tissue,” he said. “We want to overcome this limit and make a tiny antenna radiate just as efficiently as a large antenna. That requires a completely new method of making these devices.”
Bandyopadhyay has created unconventional antennas that use nanomagnets built on a piezoelectric substrate. When alternating voltage is applied, an acoustic wave is generated in the piezoelectric material which causes the magnets to emit electromagnetic waves because of a quantum mechanical coupling effect. His research introduces coupling electric charge oscillations (called plasmons) into the antennas, modifying the coupling and improving the antenna performance.
In collaboration with VCU Engineering colleague and Senior Associate Dean for Strategic Initiatives and Enrollment Management Erdem Topsakal, Ph.D., Anjan Barman from the S.N. Bose National Centre for Basic Sciences in India and several graduate students, Bandyopadhyay has demonstrated unconventional antennas that “break the theoretical limit on the radiation efficiencies” of conventional antennas and exceed them by more than 100,000 times, he said.
“These antennas have the potential to open up many new embedded applications,” he said, “such as medically implanted devices that communicate with external monitors, ultrasmall stealthy listening devices, personal communicators and wearable electronics — all while consuming minuscule amounts of energy.”
Barriers to entry
As Bandyopadhyay awaits hopeful patent protections, he is seeking more funding and qualified engineering students. He also is working to build relationships with industry, chiefly device manufacturers that would have use for his technology.
But even though he is confident in the technology, it’s still not an easy sell.
Bandyopadhyay recalls meeting with a defense contractor years ago to pitch an infrared photodetector that he believed was superior to the company’s existing technology. Afterward, talking to the CEO, he remembers the “illuminating” advice the leader gave him.
“He said, ‘You may have a better product than mine, but I couldn’t care less. Because my product is selling,’” Bandyopadhyay said, with the CEO then adding: “‘And the only way I pay any attention to you is if you have your own company and try to put me out of business.’”
“That’s the inertia you’re fighting against to get your better products out into the world,” Bandyopadhyay said, “but we are trying to shift that mindset.”