Neuromorphic 'memtransistor' resembles a human brain
In this Internet of Things (IoT) era, massive amounts of data are produced, collected and transmitted through devices in real time — but the separation of memory and data processing units adversely affects the smooth functioning of optoelectronic devices. South Korean scientists have now designed a predictable optoelectronic device — a multifunctioning ‘memtransistor’ — to address these limitations.
The current computing systems which have separate memory and processing devices cause excess energy consumption and slow down data transmission. Even state-of-the-art 2D memtransistors — devices that can collect, store and process information — exhibit suboptimal electronic properties, such as unusually high operating voltages.
To overcome these limitations, scientists at Dongguk University designed a predictable multifunctioning memtransistor. As described in the journal Small Methods, they fabricated a highly efficient optoelectronic and memory device using two-dimensional (2D) materials — nanomaterials that are merely one or two atoms thick — by stacking 2D tellurium flakes on a thin rhenium disulfide flake, followed by the deposition of an aluminium oxide layer.
As explained by senior author Professor Hyunsik Im, the team developed an “electrically and optically tuneable p–n junction memtransistor fabricated with an Al2O3 encapsulated 2D Te/ReS2 van der Waals (vdW) heterostructure. This combines the favourable optical and electrical properties of p-type 2D Te and n-type ReS2 semiconductors with a stable Al2O3 charge trapping layer.”
In this optoelectronic memory device, multiple resistance states can be tuned by applying different voltages, or light powers. The transition between the high or low resistance states is controlled by carriers trapped in the Al2O3 layer under high electric fields. This causes an additional gate bias that tunes the Schottky barrier height at the ReS2/source electrode interface, while preserving p–n junction behaviours during the switching process, giving the device the added benefit of being electrically conductive, while being able to store memory efficiently.
Remarkably, the novel device is material-independent and scalable. Moreover, it allows the integration of additional electronic circuits for neuromorphic computing — a set of processes that attempt to mimic the brain’s architecture and data processing capabilities.
“The development of these highly efficient memtransistor-based synaptic devices can decrease circuit complexity and minimise power consumption for neuromorphic computing and visual information processing,” Prof Im said. “Mimicking synaptic activities in the human brain could become a much more manageable task in the near future.”
Please follow us and share on Twitter and Facebook. You can also subscribe for FREE to our weekly newsletter and bimonthly magazine.
Fully coupled annealing processor for enhanced problem solving
Researchers have designed a scalable, fully-coupled annealing processor with 4096 spins, and...
STMicroelectronics breaks 20 nm barrier for next-gen microcontrollers
STMicroelectronics has launched an advanced process based on 18 nm Fully Depleted Silicon On...
Chip opens door to AI computing at light speed
A team of engineers have developed a silicon-photonics chip that uses light waves, rather than...
