Axon-mimicking materials show promise for highly efficient computing

A team of researchers from Texas A&M University, Sandia National Lab-Livermore, and Stanford University are taking lessons from the brain to design materials for efficient computing. The newly discovered materials are the first of their kind – mimicking the behavior of an axon by spontaneously propagating an electrical signal. These findings could be important for the future of computing and artificial intelligence.

This study has been published Nature.

Any electrical signal propagating through a metal conductor loses amplitude due to the natural resistance of the metal. Modern computer processing (CPU) and graphics processing units can have up to 30 miles of copper wire that moves electrical signals within the chip. These losses add up quickly, requiring amplifiers to maintain pulse integrity. These design constraints affect the performance of current interconnect-dense chips.

To combat this limitation, researchers drew inspiration from molds. Axons are the part of a nerve cell or neuron in vertebrates that can conduct electrical impulses from the nerve cell body.

“Often, we want to send a data signal from one place to another, to a distant place,” said lead author Dr. Tim Brown, a postdoctoral scholar at Sandia National Lab and a former doctoral student in materials science and engineering. Texas A&M.

“For example, we may need to send an electrical pulse from the edge of a CPU chip to transistors near its center. Even for the best conducting metals, resistance at room temperature continues to scatter the transmitted signals, so we typically cut the transmission line. To amplify the signal, this requires energy, time, and space differently. It does: Some signals in the brain are transmitted across a distance of centimeters, but through axons, which are made of highly resistive organic material, without disrupting and amplifying the signals.”

An associate professor in the Department of Materials Science and Engineering at Texas A&M, Dr. According to Patrick Shamberger, axons are the communication highway. They communicate signals from one neuron to a neighboring neuron. Although neurons are responsible for processing signals, axons are like fiber optic cables that move signals from one neuron to its neighbor.

As with the axon model, the substances discovered in this study are in a primary state that allows them to spontaneously amplify a voltage pulse as it passes through the axon. The researchers took advantage of the electronic phase transition in lanthanum cobalt oxide, which makes it highly conductive when heated. This characteristic interacts with the small amount of heat generated when a signal passes through the material, resulting in a positive feedback loop.

The result is a set of exotic behaviors not found in ordinary passive electrical components—resistors, capacitors, inductors—including amplification of small disturbances, negative impedances, and unusually large phase shifts in AC signals.

According to Shamberger, these materials are unique because they exist in a semi-stable “Goldilocks state”. Electrical pulses do not decay or exhibit thermal runaway and break down. Conversely, if kept under constant current conditions, the material will naturally oscillate. The researchers determined that this behavior could be used to create spiking behavior and amplify the signal traveling on the transmission line.

“We mainly use internal instabilities in the material, which continuously strengthen an electronic pulse along the transmission line. This behavior was theoretically predicted by our co-author Dr. Stan Williams, and this is the first confirmation of its existence.”

These findings could be important in the future of computing, which is increasing demand for energy use. Data centers are expected to consume 8% of America’s energy by 2030, and artificial intelligence could dramatically increase that demand. In the long run, this is a step toward understanding dynamic materials and using biological inspiration to develop more efficient computing.