Researchers have used biological conductive filaments to create neuromorphic memory transistors (memristors) that work effectively at very low power.
One of the main obstacles to the development of neuromorphic computing is the high power consumption of modern computers operating at a voltage of 1 V, while in the brain, 80 mV is sufficient for transmitting signals between neurons. After 10 years of experimentation, a team at the University of Massachusetts Amherst has developed a memristor that performs ultra-low energy consumption.
For creating electronic device scientists used protein nanowires from the bacteria Geobacter. Unlike silicon analogs, they are more stable in water and biological fluids, and their production is cheaper. and does not require harmful chemicals or high energy processes.
In experiments with a switch that transmits a pulsed charge through a tiny metal filament in a memristor, it was found that protein nanowires simplify the reduction of metals by changing the reactivity of their ions and the properties of electron transport..
In addition, since the on-off pulses cause changes in the metal filaments, the tiny device new branches and junctions are formed, which are 100 times smaller than the diameter of a human hair, which creates new connections, reminiscent of the learning process in the brain.
This allows the nanowire and memristor synapse to be modulated in conductance, or plasticity, so that it can emulate biological components for computation, the researchers said. Compared to a conventional computer, such a device has a learning ability that is not based on software..
Earlier, we also reported on the development of technology for efficient absorption of energy from metal by cleaving its chemical bonds.
text: Ilya Bauer, photo: UMass Amherst