A new microelectronics device can program and reprogram computer hardware as needed by using electrical pulses.
What if a computer could learn to rewire its circuits based on the information it receives?
A multi-institutional collaboration, involving the U.S. Department of Energy (DOE) Argonne National Laboratory, has created material that can be used to make computer chips that can do just that. This is achieved by using so-called “neuromorphic” circuitry and computer architecture to replicate brain functions. Purdue University professor Shriram Ramanathan led the team.
“The human brain can be transformed as a result of learning new things,” said Subramanian Sankaranarayanan, a co-author of the paper with a joint appointment at Argonne and the University of Illinois. Chicago. “We’re now creating a tool for machines to reconfigure their circuits in a brain-like way.”
With this capability, artificial intelligence -based computers can perform difficult jobs faster and more accurately while using less energy. One example is the analysis of complex medical images. Autonomous cars and space robots that can rewire their circuits depending on experience are a more futuristic example.
The key material of the new device consists of neodymium, nickel, and oxygen and is called perovskite nickelate (NdNiO3). The team lined this material with hydrogen and attached electrodes to it that allowed electrical pulses to be applied at different voltages.
“How much hydrogen is in nickelate, and where it is, changes the electronic properties,” Sankaranarayanan said. “And we can change its location and concentration with different electrical pulses.”
“This material has a multi-layered personality,” added Hua Zhou, a co-author of the paper and Argonne physicist. “It has two common functions of everyday electronics-turning on and blocking electricity as well as storing and discharging electricity. What is really new and unique is the addition of two functions that similar to the separate nature of synapses and neurons in the brain.A neuron is a nerve cell that connects to other nerve cells through synapses.Neurons begin to sense the outside world.
For its contribution, the Argonne team conducted a computational and experimental characterization of what would happen to the nickelate device under different voltages. To that end, they rely on the DOE Office of Science Argonne’s user facilities: the Advanced Photon Source, Argonne Leadership Computing Facility, and Center for Nanoscale Materials.
The experimental results show that the voltage change controls the movement of hydrogen ions within the nickel. A specific voltage concentrates hydrogen at the center of the nickel, producing neuron -like behavior. A different voltage shuttles hydrogen from the center, giving a synapse-like character. At different voltages, the resulting locations and hydrogen concentrations emit on-off currents on the computer chips.
“Our calculations revealing this mechanism at the atomic scale are extremely robust,” said Argonne scientist Sukriti Manna. The team relies on computational horsepower not only at the Argonne Leadership Computing Facility but at the National Energy Research Scientific Computing Center, a DOE Office of Science user facility at Lawrence Berkeley National Laboratory.
Confirmation of the mechanism came, in part, from the 33-ID-D beamline experiments at the Advanced Photon Source.
“Over the years we have had a very productive partnership with the Purdue group,” Zhou said. “Here, the team determined how the atoms inside the nickel are arranged under different voltages. Particularly important is to track the material’s response on the atomic scale to the action of hydrogen.”
With the team’s nickelate device, scientists will work to create a network of artificial neurons and synapses that can learn and change from experience. This network grows or shrinks as it is presented with new information and is thus able to work with extreme energy efficiency. And that energy efficiency translates into lower operating costs.
Brain -inspired microelectronics with team devices as a building block could have a bright future. This is especially so because the device can be made at room temperature by techniques consistent with the methods of the semiconductor industry.
The Argonne -related work was funded by the DOE Office of Basic Energy Sciences, as well as the Air Force Office of Scientific Research and the National Science Foundation.
Reference: “Reconfigurable perovskite nickelate electronics for artificial intelligence” by Hai-Tian Zhang, Tae Joon Park, ANM Nafiul Islam, Dat SJ Tran, Sukriti Manna, Qi Wang, Sandip Mondal, Haoming Yu, Suvo Banik, Shaobo Cheng, Hua Zhou , Sampath Gamage, Sayantan Mahapatra, Yimei Zhu, Yohannes Abate, Nan Jiang, Subramanian KRS Sankaranarayanan, Abhronil Sengupta, Christof Teuscher and Shriram Ramanathan, 3 February 2022, Science.