A team of researchers at the University of Würzburg, working within the ctd.qmat research cluster, has developed a miniature electronic component that simultaneously performs computations, stores data and adapts to incoming signals, infohub.kz reports.
In conventional computers, the processor and memory are physically separate, so the constant transfer of data between them requires enormous amounts of energy — especially when running artificial intelligence. The human brain, by contrast, processes information directly in synapses, the connections between neurons that reconfigure themselves under the influence of electrical impulses. The new technology reproduces this biological principle.
The device is built on two oxide materials — lanthanum aluminate and strontium titanate. Individually, neither conducts current, but at the interface where they meet an ultrathin conductive layer forms. By passing electricity through it, the scientists were able to shift oxygen atoms and thereby change the material's resistance. As a result, the component can literally be "trained": it remembers previous activity and alters its properties.
A single nanochip can handle three tasks at once, depending on how it is connected: it can work as an ordinary current switch (a transistor), store data through resistance (a memristor) or hold a charge based on past experience (a memcapacitor). In the long run, this universal foundation could enable energy-efficient neural networks and smart wearable health sensors that analyze body readings directly on the skin without sending data to servers.