Researchers at the University of Würzburg (Cluster of Excellence ctd.qmat) have shown that a single nanoscale device built on an oxide interface can act as a transistor, a memristor or a memcapacitor, depending on how it is wired. The work was published in Nature Communications (April 2026) and highlighted this week.
The material. Lanthanum aluminate (LaAlO3) and strontium titanate (SrTiO3) are both insulators, but at their interface they form a quasi-two-dimensional electron gas. By patterning this layer, the team creates microscopic conductive channels.
What makes it neuromorphic. Electrical current can displace oxygen atoms in the structure and change the channel's resistance. The device can therefore be "trained" by electrical stimulation, and its state depends on its history — the same principle as synaptic plasticity. As a memcapacitor, its ability to store charge likewise depends on its previous electrical state.
Why it matters. Conventional computers shuttle data between separate memory and processor, which is energy-hungry. One material platform that can switch, remember and adapt could simplify hardware neural networks, with possible long-term uses in self-learning wearable sensors. The study demonstrates the underlying device functions, not a complete AI system, and energy and reliability benefits remain to be established.
Paper: S. Pradhan et al., "Oxide interface-based polymorphic electronic devices for neuromorphic computing", Nature Communications (2026), DOI 10.1038/s41467-026-71642-2
Coverage: SciTechDaily / University of Würzburg
No comments:
Post a Comment