American researchers have developed a new type of memristor with diffusive dynamics which could open a new avenue of neuromorphic computing hardware.
The idea behind neuromorphic computing is that the computer operates more like a human brain. To do so, it needs the functionality of a biological synapse and this is where the memristor comes in.
The team from the Electrical and Computer Engineering Department at the University of Massachusetts Amherst say memristors have become a leading candidate to enable neuromorphic computing by reproducing the functions in biological synapses and neurons in a neural network system, while providing advantages in energy and size.
Unfortunately most of the synaptic demonstrations with memristors to date have not implemented diffusive dynamics unless sub-threshold complementary metal–oxide–semiconductor (CMOS) computers equipped with large capacitors are used to simulate it.
They demonstrated a bio-inspired solution to the diffusive dynamics that is fundamentally different from the CMOS approach while sharing great similarities with synapses. This leads to a significant reduction in footprint, complexity, and energy-consumption.
“Specifically, we developed a diffusive-type memristor where diffusion of atoms offers a similar dynamics and the needed time-scales as its bio-counterpart, leading to a more faithful emulation of actual synapses, i.e., a true synaptic emulator,” they said.
Their approach offers not only a timing mechanism close to that of actual synapses but also other important features observed in synapses, such as the dynamical balance of Ca2+ concentration, depletion effects of mobile species, and interacted but separated physical entities for different functions and others.
"These enable us to more closely capture synaptic functions that could not be demonstrated before,” they said.
The working mechanism of the proposed novel memristor was confirmed with a combination of in-situ transmission electron microscopy and nanoparticle dynamics simulations. Used together with a regular drift-type memristor, the operating characteristics of the devices were verified experimentally by demonstrating some important synaptic functions, including both short-term and long-term plasticity.
[An illustration of the concept behind the new memristor. Photo: University of Massachusetts Amherst]