Neuromorphic computing is a relatively new discipline in which electronic circuits mimic neurological systems in the body. Recently, it has come to the forefront of electronics to achieve sophisticated artificial intelligence. To accomplish this, researchers aim to build a neuromorphic memristor, a type of electrical component that can function like real brain synapses.
A team at the University of Massachusetts Amherst has found a way to create a neuromorphic memristor. By using biological, but conductive filaments, the team created a tool that works efficiently on low power. One of the biggest hurdles to neuromorphic computing is that most computers operate at over one volt of electricity, whereas the brain works at around 80 millivolts – over 10 times lower than computers. However, this newly developed memristor can work at biological voltages of 40-100 millivolts, just like the human brain.
The voltage experiments were done in the lab run by Jun Yao, an electrical and computer engineering researcher. Yao said: “People probably didn’t even dare to hope that we could create a device that is as power-efficient as the biological counterparts in the brain. It’s a conceptual breakthrough and we think it’s going to cause a lot of exploration in electronics that work in the biological voltage regime”.
First author, Tianda Fu, used protein nanowires developed at the University of Massachusetts Amherst by microbiologist and co-author Derek Lovely. Protein nanowires are electrically conductive structures, produced by certain bacteria. The ones used in this report were from a type of bacterium called Geobacter.
Geobacter’s electrically conductive nanowires hold many advantages over the currently used silicon ones. Making silicon nanowires requires a large amount of energy and involves toxic chemicals. In contrast, biological nanowires are produced by shearing them off the bacteria. They are also more stable in water or bodily fluids, an important feature for biomedical applications.
To test the nanowires, Yao used a pulsing on-off pattern of positive-negative charge through a metal thread in a memristor. The protein nanowires affect the metal thread’s properties, causing them to change. This is because bacterial nanowires are living, and chemically alter metals to get their energy similarly to how we breathe oxygen.
As the device is subject to the on-off pulse, new branching and connections develop within the device. This creates an effect similar to learning; new connections are being made like in a human brain. In short, the device has a learning capability that is not based on software.
Future research in this field will be focused on fully exploring the “chemistry, biology, and electronics” of protein nanowires in memristors. An exciting prospect of this field is the possibility to develop devices that directly communicate with human neurons. An example given by the researchers is a less invasive and more sustainable heart monitor that interfaces directly with the body.