SciDev.Net (Reino Unido)

Transistor that mimics neurons developed

Publicado em 12 maio 2017

Por José Tadeu Arantes, da Agência FAPESP

Speed read

  • Transistor can count, remember and perform simple arithmetic operations much like neurons
  • Innovation set to allow manufacturing equipment without additional memory unit
  • Tool is at early stage of development but shows capacity for complex operations

An international team of scientists has developed a transistor capable of mimicking some characteristics of neurons, such as counting, remembering and performing simple arithmetic operations.

The device still needs to undergo a long process of adjustment and refinement. However, it can lead to the development of a platform for the manufacture of super-compact calculators and other electronic equipment that need to have memory intrinsically linked to the transistor itself, or incorporate features at nanoscale without requiring an additional memory unit.

Standard computing equipment typically have separate processing and memory units that constantly exchange data. The prospect of storing data in the same physical unit can lead to devices — such as memsistors, memcapacitors and meminductors — with inherent memory and better performance due to the time and energy saved by avoiding the transfer of data

The transistor was developed through tests and modelling carried out jointly by researchers from the Federal University of São Carlos (UFSCar), Brazil, the University of Würzburg, Germany, and the University of South Carolina, United States. The research is supported by the Brazilian research foundation FAPESP.

In an article published in the journal Nano Letters, the authors explained their transistor consists of micrometric and nanoscale parts, and can process information in infinitely smaller scales of time, space and energy than do current transistors, such as the ones used in the average calculator.

"In this work, we demonstrate the ability of quantum dots-based transistors to perform complex operations directly in memory," said Victor Lopez Richard, professor in the physics department at UFSCar, and one of the project’s coordinators, in an interview for FAPESP’s newsletter.

“In addition, since quantum dots are sensitive to photons, we can say that the transistor is capable of perceiving light,” Richards added, explaining that this means the transistor can control its own electrical voltage. Having this capacity enables it to control the dynamics of quantum dots involved in the process and mimic how neurons respond to stimuli. 

However, he cautions that more research is needed before the transistor can be used as a technological tool. For example, it now only works at extremely low temperatures of approximately 4 Kelvin, which corresponds to the temperature of liquid helium.

"Our goal is to make it functional on other levels, including room temperature," Richards said.

Taken from a newsletter by FAPESP, a SciDev.Net donor, edited by SciDev.Net’s Latin America and the Caribbean desk