Clemson researchers develop biodegradable AI material

Clemson University researchers says a new advanced material created by them and researchers in the Czech Republic could play a key role in helping computers think more like brains and then biodegrade under certain conditions at the end of its lifespan.
The trans-Atlantic research team created the material, a polymer they call pB6CC, and used it to build memristors, which are a type of electronic component, according to a news release.
PB6CC is based on carbazole, a type of organic chemical building block that is used in some electronics. What makes pB6CC unique is that it combines properties favorable for advanced computing with the potential to biodegrade under specific environmental conditions.

Researchers built pB6CC into memristors because such devices could help move computing beyond today’s conventional chips and toward systems that mimic how the brain processes information, the news release stated.

Those systems, known as spiking neural networks, show promise for reducing the energy needed to power artificial intelligence systems.

The memristors act like synapses, changing their response based on the timing and history of signals. In spiking neural networks, this lets the system strengthen or weaken connections over time and process information in short bursts, much like the brain, using less energy than conventional chips.

Stephen Foulger, the Greg-Graniteville Endowed Chair and Professor at Clemson, said the project underscores the often overlooked but increasingly important role that advanced materials play in shaping the future of computing.
“Advanced materials are at the crest of a wave of next-generation AI technologies,” he said in the release. “There are a lot of startups in Silicon Valley that are looking at alternative designs of chips. I’m not talking about just faster and better. I’m talking about completely different methodologies to process information and to encode information, like through spiking neural networks.”

One of the problems with conventional chips is that they contribute to the growing problem of electronic waste and have limited recycling options, Foulger said.

The memristors created from pB6CC could help address that problem. They could biodegrade because pB6CC can be broken down by specific bacteria, Foulger said.
Memristors would also have advantages over conventional chips in manufacturing, Foulger said. While conventional chips are made in complex fabrication plants, memristors could be manufactured much less expensively using printing techniques, he said.

Kyle Brinkman, chair of the Department of Materials Science and Engineering at Clemson, said the research reflects the university’s leadership in advanced materials.

“This kind of work shows how Clemson is at the forefront of creating materials with real-world application,” he said in the release. “I congratulate Dr. Foulger and his team on this achievement and the impact it could have on future computing technologies.”
Researchers described their work in a paper published in April in the journal Applied Electronic Materials.

The title of the paper is, “Toward a Hardware Spiking Neural Network: Learning and Adaptation with an Environmental Sustainable Polymer Memristor.”
Foulger served as corresponding author.

Co-authors were Yuriy Bandera, Travis Wanless and Igor Luzinov of Clemson’s Department of Materials Science and Engineering; Olivia Cobb and Michael G. Sehorn of Clemson’s Department of Genetics and Biochemistry; Lubomir Kostal of the Institute of Physiology, Computational Neuroscience Group, Czech Academy of Sciences; Jiří Pfleger of the Otto Wichterle Centre of Polymer Materials and Technologies, Institute of Macromolecular Chemistry; and Jarmila Vilčáková of the Department of Physics and Materials Engineering, Faculty of Technology, Tomas Bata University.