A synaptic transistor that learns while it computes

Material scientists at the Harvard School of Engineering and Applied Sciences (SEAS) have now created a new type of transistor that mimics the behaviour of a synapse. The  device simultaneously modulates the flow of information in a circuit and physically adapts to changing signals. We do have a lot of undiscovered as well as unusual properties in modern materials. The synaptic transistor could mark the beginning of a new kind of artificial intelligence: one embedded not in smart algorithms but in the very architecture of a computer. 

"There's extraordinary interest in building energy-efficient electronics these days," says principal investigator Shriram Ramanathan, associate professor of materials science at Harvard SEAS. "Historically, peole have been focused on speed, but with speed comes the penalty of power dissipation. With electronics becoming more and more powerful and ubiquitous, you could have a huge impact by cutting down the amount of energy they consume." 

The human mind, for all its phenomenal computing power, runs on roughly 20 watts of energy, so it offers a natural model for engineers. "The transistor we have demonstrated is really an analog to the synapse in our brains," says co-lead author Jian Shi, a postdoctoral felllow at SEAS. "each time a neuron initiates an action and another neuron reacts, the synapse between them increases the strength of its connection. And the faster neurons spike each time, the stronger the synaptic connection. Essentially, it memorizes the action between the neurons." 

In principle, a system integrating millions of tiny synaptic transistors and neuron terminals could take parallel computing into a new era of ultra-efficient high performance. 

So how does a synaptic transistor works? The synaptic transistor has a structure which is almost similar to that of a field effect transistor, where a bit of ionic liquid takes the place of the gate insulating layer between the gate electrode and the conducting channel, and that channel is composed of samarium nickelate (SmNiO3 or SNO) rather than field effect transistor's doped silicon. 

A synaptic transistor has an immediate response and a learning response. The immediate response is basically same as that of a field effect transistor- the amount of current that passes between the source and drain contacts varies with the amount of voltage applied to the gate electrode. The learning response is that the conductivity of the SNO layer varies in response to the STDP history of the synaptic transistor, essentially by shutting oxygen ions between the SNO and the ionic liquid. 

The electrical analog of strengthening a synapse is to increase the conductivity of the SNO, which essentially increases the gain of the synaptic transistor. Similarly, weakening a synapse is analogous to decreasing the electrical conductivity of the SNO. thereby lowering the gain. The artificial synapses have the flexibility to learn "more or less" how to perform a task, and then to learn how to improve its earlier performance etc. While the physical structure of Harvard's synaptic transistor has the potential t learn from history, in itself it contains no way to bias the transistor so as to properly control SNO's memory effect. This function is carried out by an external supervisory circuit that converts the time delay between input and output into a voltage applied to the ionic liquid that either drives ions into the SNO or removes them. In response, the synaptic transistors become self-optimizing within a circuit being subjected to learning experiences. 

The new transistor is inherently energy efficient. The nickelate belongs to an unusual class of materials, called corelated electron systems, that can undergo an insulator-metal transition. At a certain temperature when exposed to an external field the conductance of the material suddenly changes. "We exploit the extreme sensitivity of this material," says Ramanathan. "A very small excitation allows you to get a large signal, so the input energy required to drive this switching is potentially very small. That could translate into a large boost for energy efficiency." The beauty of this type of a device is that the 'learning behaviour is more or less temperature insensitive, that's a big advantage," says Ramanathan. 

The research was supported by the National Science Foundation (NSF), the Army Research Office's Multidisciplinary University Research Initiative, and the Air Force Office of Scientific Research. The team has also benefited from the facilities at the Harvard Centre for Nanoscale Systems, a memeber of the NSF-supported National Nanotechnology Infrastructure Network.