Combining characteristics of a memristor with a transistor mimics the multiple synapses of neurons
Illustration: Northwestern University
Computers that operate more like the human brain than computers—a field sometimes referred to as neuromorphic computing—have promised a new era of powerful computing.
While this all seems promising, one of the big shortcomings in neuromorphic computing has been that it doesn’t mimic the brain in a very important way. In the brain, for every neuron there are a thousand synapses—the electrical signal sent between the neurons of the brain. This poses a problem because a transistor only has a single terminal, hardly an accommodating architecture for multiplying signals.
Now researchers at Northwestern University, led by Mark Hersam, have developed a new device that combines memristors—two-terminal non-volatile memory devices based on resistance switching—with transistors to create what Hersam and his colleagues have dubbed a “memtransistor” that performs both memory storage and information processing.
This most recent research builds on work that Hersam and his team conducted back in 2015 in which the researchers developed a three-terminal, gate-tunable memristor that operated like a kind of synapse.
While this work was recognized as mimicking the low-power computing of the human brain, critics didn’t really believe that it was acting like a neuron since it could only transmit a signal from one artificial neuron to another. This was far short of a human brain that is capable of making tens of thousands of such connections.
“Traditional memristors are two-terminal devices, whereas our memtransistors combine the non-volatility of a two-terminal memristor with the gate-tunability of a three-terminal transistor,” said Hersam to IEEE Spectrum. “Our device design accommodates additional terminals, which mimic the multiple synapses in neurons.”
Hersam believes that these unique attributes of these multi-terminal memtransistors are likely to present a range of new opportunities for non-volatile memory and neuromorphic computing.
Hersam and his team used molybdenum disulfide in their work back in 2015. However, in that instance they just used flakes of the material. In this most recent work, they used a continuous film of polycrystalline molybdenum disulfide that includes a large number of smaller flakes. This made it possible to scale up the device from just a single flake to a number of devices across an entire wafer.
Once they had fabricated memtransistors uniformly across an entire wafer, Hersam and his colleagues added contacts.
“Thus far, we have demonstrated seven terminals (six terminals in direct contact with the molybdenum disulfide channel and a seventh gate terminal), but additional terminals should be achievable using higher resolution lithography,” said Hersam.
The multi-terminal memtransistors have distinctive electrical characteristics, according to Hersam. For one, they have gate-tunability that allows dynamic adjustment of the electrical characteristics through the application of a gate potential. They also have large on/off switching ratios with high cycling endurance and long-term retention of states.
Perhaps the key feature is that the multiple terminals mimic the multiple synapses in neurons and enable gates to be tuned in such a way that the memtransistor is capable of more functions than would be possible to achieve using standard two-terminal memristors.
“For example, the conductance between a pair of two floating electrodes (pre-synaptic and post-synaptic neurons) can be varied by an order of magnitude by applying voltage pulses to the modulatory terminals,” explained Hersam.
Hersam believes that these memtransistors can serve as a foundational circuit element for neuromorphic computing. Of course, scaling up from dozens of these devices to the billions that are available today in conventional transistors must still be done. But Hersam does not see any fundamental barriers to doing this. In fact, Hersam and his team are already moving toward this aim.
He added: “We are now working on making smaller and faster memtransistors, which should possess lower operating voltages and more efficient neuromorphic computation. We are also exploring the integration of memtransistors into more complicated circuits that are suitable for non-volatile memory and advanced neuromorphic architectures.”