UCSD physicists receive $12.6 million from DOE for next-generation computing

UCSD physicists receive $12.6 million from DOE for next-generation computing

September 29, 2022 — The first generation of computers used vacuum tubes. The second, transistors and the third, integrated circuits. Each new generation has allowed computers to be faster, smaller and more energy efficient. Now, as the world expands beyond the limits of integrated circuits, what does fourth generation computing look like?

Q–MEEN-C research seeks to mimic the emerging complexity that makes the brain an efficient computer.

The answer may lie in quantum materials capable of achieving neuromorphic, or cerebral, computing capabilities with low power consumption. Since 2018, Quantum Materials for Energy Efficient Neuromorphic Computing (Q-MEEN-C) – led by the University of California, San Diego – has been at the forefront of this research. Now, through a highly competitive process, the Department of Energy (DOE) has announced renewed funding of $12.6 million for the center.

“This round of additional funding demonstrates the Department of Energy’s confidence in the work being done by Q-MEEN-C,” said UC San Diego Chancellor Pradeep K. Khosla. “The center embodies many of our guiding principles of collaboration and cutting-edge research. This achievement not only has a positive effect on the researchers, but also on the Department of Physics and on the entire university.

Q-MEEN-C is a DOE Energy Frontiers Research Center (EFRC), one of more than 40 established to help address the world’s most pressing energy technology challenges. The center, led by UC San Diego, is a collaborative effort that brings together researchers from around the world. They each bring unique expertise to an exciting scientific challenge: to create a brain-like computer with dramatically reduced power requirements.

“With current technology, to make a device that mimics the brain, the local energy requirements are so great that it’s not practical,” said Q-MEEN-C director and professor emeritus of physics Ivan K. Schuler. “During the semiconductor revolution, materials science helped developers identify silicon and germanium as ideal materials. It’s the same now, where we see quantum materials as the key to increasing computing power while reducing local energy consumption.

Quantum materials are a class of new materials that display more complex quantum mechanical behavior than silicon, and whose range of properties are ideally suited for more efficient and transformative neuromorphic computing.

When Q-MEEN-C was created, researchers sought to determine if quantum materials were even viable as an energy-efficient material for neuromorphic computing. Over the past four years, they have successfully demonstrated that quantum materials have great potential due to their unusual electronic and magnetic properties.

A summary of Q-MEEN-C’s key metrics since 2018, including members and publications.

“This is just the beginning,” said the center’s co-director and physics professor Alex Frañó. “Now that we have found viable materials, we are laying the groundwork for future research. The human brain is a network of neurons, synapses and dendrites – you can’t have a brain-like computer without a brain-like network. We can take these quantum materials and combine them with other materials to see how they react with each other as a step towards creating neuromorphic computer networks.

Frañó says the center takes a holistic approach to the problem – from a single electron in an atom to the complexity of a computer chip: “You have to understand the physics at every step.” This is called “emergence” – where the whole is more than the sum of its parts, even if it is not explicitly known how all the parts work together.

One of the main goals of developing neuromorphic computing is pattern recognition, classification, and learning – things the brain does remarkably well with minimal energy input. A human could see an image of the Golden Gate Bridge and an image of the Statue of Liberty and instantly differentiate between the two landmarks. A computer would have to individually analyze the billions of pixels in multiple images to come to the same conclusion. Add fog or rain or a different angle and it gets even more complicated.

Although we see this capability to some extent when photo software recognizes similar faces in a feed as the same person, there is a limit depending on how detailed the images are, how long it takes to categorize them and the amount of energy it requires. battery of your phone.

“It’s not just a software issue,” Schuller said. “You won’t achieve energy efficiency with software improvements. There must also be a new type of material.

One of the stated goals of EFRCs is to train future energy scientists, and DOE funding supports Q-MEEN-C students and postdoctoral researchers. “We are not only creating the next generation of knowledge, but also the next generation of researchers,” said Frañó. “One day, our students will lead their own research groups in neuromorphic computing.”

“We are driven by continuous wonder. Maybe neuromorphic computing won’t play out the way we envision it today, but it will play out one way or another. It could take decades and it’s probably beyond what we’re able to predict right now,” Schuller said. “But this next generation – they will see something wonderful.”

Funding provided by DOE #DE-SC0019273. A complete list of principal investigators and participating institutions is available on the Q-MEEN-C website.

Credit: Michelle Franklin, UCSD

#UCSD #physicists #receive #million #DOE #nextgeneration #computing

Leave a Comment

Your email address will not be published. Required fields are marked *