Dec 18 2018
An innovative electronic device created at the University of Michigan has the ability to directly model the behaviors of a synapse—a connection between two neurons.
A schematic of the molybdenum disulfide layers with lithium ions between them. On the right, the simplified inset shows how the molybdenum disulfide changes its atom arrangements in the presence and absence of the lithium atoms, between a metal (1T’ phase) and semiconductor (2H phase), respectively. (Image credit: Xiaojian Zhu, Nanoelectronics Group, University of Michigan)
For the first time, it has been possible to explore the way neurons compete for or share resources, using hardware without the need for complex circuits.
“Neuroscientists have argued that competition and cooperation behaviors among synapses are very important. Our new memristive devices allow us to implement a faithful model of these behaviors in a solid-state system,” stated Wei Lu, U-M professor of electrical and computer engineering and senior author of the study published in Nature Materials.
Memristors are electrical resistors that have memory—sophisticated electronic devices that regulate current on the basis of the history of the voltages applied to them. They have the ability to store and process data at the same time, rendering them quite efficient compared to conventional systems. They could enable innovative platforms that process an enormous number of signals in parallel and have the ability to perform advanced machine learning.
The memristor is an excellent model for a synapse. It simulates the manner in which the connections between neurons weaken or strengthen as signals pass through them. However, the variations in conductance essentially arise from variations in the shape of the channels of conductive material included in the memristor. These channels, as well as the memristor’s potential to conduct electricity, could not be precisely regulated in earlier devices.
Currently, the U-M researchers have created a memristor in which they have better control of the conducting pathways. They produced an innovative material from the semiconductor molybdenum disulfide—a “two-dimensional” material that can be stripped into layers just a few atoms thick. Lithium ions were injected by Lu and his colleagues into the gaps between molybdenum disulfide layers.
They discovered that when adequate lithium ions are present, the molybdenum sulfide modifies its lattice structure, thus allowing electrons to easily run through the film as if it were a metal. However, in regions with very few lithium ions, the molybdenum sulfide revives its original lattice structure and turns into a semiconductor, and it is hard for the electrical signals to get through.
The lithium ions can be slid with an electric field to easily rearrange them within the layer. This modifies the size of the areas that conduct electricity bit by bit, thus regulating the conductance of the device.
Because we change the ‘bulk’ properties of the film, the conductance change is much more gradual and much more controllable.
Wei Lu, Professor of Electrical and Computer Engineering, University of Michigan.
Apart from making the devices behave in a better way, the layered structure allowed Lu and his colleagues to link a number of memristors together through shared lithium ions—forming a type of connection that is also found in brains. The dendrite of a single neuron, or its signal-receiving end, might have various synapses linking it to the signaling arms of other neurons. Lu compared the availability of lithium ions to that of a protein that allows the growth of the synapses.
In case proteins, known as plasticity-related proteins, are released by the growth of one synapse, other adjacent synapses can also grow—this is cooperation. Neuroscientists have debated that cooperation between synapses assists in rapidly forming vivid memories that last for many years and develop associative memories, such as a scent that reminds one of his/her grandmother’s house, for instance. In case the protein is scarce, one synapse will grow at the expense of the other—and this competition pares down the brains’ connections and prevents them from exploding with signals.
Lu and his colleagues were able to demonstrate that these phenomena directly use their memristor devices. In the competition framework, lithium ions got drained away from one side of the device. There was an increase in the conductance of the side with the lithium ions, mimicking the growth, and the conductance of the device with very less lithium was limited.
In a cooperation framework, they developed a memristor network with four devices with the ability to exchange lithium ions, and subsequently siphoned a portion of the lithium ions from one device out to the others. In this scenario, besides the increase in the conductance of the lithium donor, there was an increase in the conductance of the other three devices too, despite the fact that their signals were not so strong.
At present, Lu and his colleagues are developing networks of memristors such as these to investigate their capability for neuromorphic computing, which emulates the circuitry of the brain.
The study was supported in part by the National Science Foundation. It was carried out in collaboration with the group of Xiaogan Liang, U-M professor of mechanical engineering.