🧠 This artificial device behaves like a neuron

Researchers at McGill University have developed a light-sensitive nanoscale structure that mimics a neuron in information processing. The neuron-like behavior originates from the materials themselves. As a result, energy consumption is lower compared to using similar devices combined with circuits or software.

Instead of collecting data and processing it elsewhere, the device detects and interprets light in the same location, much like an eye processes visual information.


According to the researchers, this innovation could improve the efficiency of visual technologies, such as artificial retinas and smart optical sensors. It could also revolutionize the manufacturing of artificial neural networks (ANNs), which are one of the foundations of machine learning.

"By using specific materials and nanostructures, we have developed the first device capable of faithfully reproducing the neuronal dynamics observed in a biological context," explains Songrui Zhao, lead author and associate professor of electrical and computer engineering.

A multi-layered light-sensitive device The researchers fabricated the device by stacking layers of atoms using a technique called molecular beam epitaxy. They then exposed it to light of different colors, intensities, and durations, and measured changes in the electrical signals within the material.

By analyzing the signals over a given period, they found that the device could combine incoming signals, briefly store information, and trigger a response once a certain threshold was reached. This mechanism resembles how a neuron processes information and shows that such behavior can arise directly from the physical properties of the material rather than from software or a complex circuit.

"By carefully designing the layers, we created a device whose response to light is tunable, which provides the basis for replicating the behavior of a single neuron," says Professor Zhao. "We succeeded in modulating the electrical current flow to achieve the desired behavior."

Building neural networks from scratch According to the researchers, since artificial neural networks (ANNs) are made up of a large number of interconnected neurons, this device could offer a new approach to system design.

"A single artificial neuron is like a cell that can be used as a building block to construct networks from scratch," adds Professor Zhao. "It is a somewhat crazy idea: creating something akin to a biological system from an inorganic material."

Such an approach could lead to more efficient information processing and applications in advanced computing, for example.

The researcher notes that future studies will focus on extending the device's light sensitivity range and optimizing its performance, as well as on applications such as data encryption, where processing information directly at the sensor level could enhance security.