Novel components based on an oxide interface, developed by researchers at the ctd.qmat Cluster of Excellence in Würzburg, electronically replicate central functions of neural networks and open up new perspectives for energy-efficient hardware.
Image Credit: Jochen Thamm, think-design
Scientific Frontline: Extended "At a Glance" Summary: Neuromorphic Oxide-Interface Electronics
The Core Concept: A novel class of polymorphic electronic devices utilizes complex oxide materials to emulate the neural structure of the human brain, allowing hardware to process and store information simultaneously.
Key Distinction/Mechanism: Unlike traditional computing architecture that spatially separates processing and memory, this technology uses an ultrathin, conductive quasi-two-dimensional electron gas formed between two insulating oxides. Electrical currents displace oxygen atoms, altering electrical resistance and allowing the device to learn and adapt based on past activity, a process closely mimicking synaptic neuroplasticity.
Major Frameworks/Components:
- Lanthanum aluminate (\(\text{LaAlO}_3\)) and strontium titanate (\(\text{SrTiO}_3\)): The two insulating complex oxides that combine to create a highly conductive interface.
- Polymorphic nanoscale architecture: A single device that can function variably as a transistor (for current switching), a memristor (for resistance-based memory), and a memcapacitor (for electrical history-dependent capacitance).
- Quasi-two-dimensional electron gas: Microscopic electronic pathways that enable the precise, targeted control of charge carrier transport.
Branch of Science: Nanoelectronics, Solid-State Physics, Materials Science, and Neuromorphic engineering.
Future Application: The development of intelligent, self-learning biosensors for continuous, localized medical diagnostics (e.g., monitoring heart rate, blood pressure, and blood sugar), as well as foundational hardware for autonomous edge computing.
Why It Matters: By embedding memory and processing capabilities directly into single nanodevices, this brain-inspired computing platform promises to drastically reduce the massive energy consumption currently required to train artificial intelligence and machine learning models.
Researchers from the Würzburg branch of the Cluster of Excellence ctd.qmat have developed a new type of electronic device. These components open up new perspectives for energy-efficient hardware, particularly for applications in artificial intelligence.
Artificial intelligence currently requires vast amounts of energy, especially when training large models. While conventional computers typically separate processing and memory operations on a chip spatially, these functions are closely integrated in the human brain. Neurons (nerve cells) are connected via synapses and process signals directly at these junctions. When signals occur frequently or are particularly strong, these connections change. A central mechanism of learning is that the system stores previous activity and adapts its behavior. Because the connections in the brain change accordingly, this process is known as neuroplasticity.
Brain-inspired computing aims to mimic this principle of learning from experience; computer hardware should not only process information but also remember states and adapt. Researchers at the Universities of Würzburg and Dresden, working within the Cluster of Excellence ctd.qmat (Complexity, Topology, and Dynamics in Quantum Matter), have demonstrated that complex oxide materials are particularly well suited for this purpose. In a recent study, they present devices that electronically replicate essential properties of biological nervous systems.
Complex Oxides Combine Many Functions
The foundation of the new devices is an interface composed of two oxide materials produced at the Würzburg Chair for Experimental Physics IV: lanthanum aluminate (\(\text{LaAlO}_3\), or LAO) and strontium titanate (\(\text{SrTiO}_3\), or STO). Although both materials are electrically insulating on their own, an extremely thin conductive region known as a quasi-two-dimensional electron gas forms at their shared interface.
Through targeted microstructuring, researchers can create tiny "electron highways" where charge carrier traffic—that is, charge transport—can be precisely controlled. When current flows through the interface, oxygen atoms can be dislodged, altering the electrical resistance. Consequently, the conductivity of the structure can be deliberately adjusted. This process allows a device to be "trained," functioning similarly to a neural network that learns from external stimuli.
The complex oxides utilized in this process are among the most promising material platforms for novel electronics. "Complex oxides are a particularly exciting playground for us because they combine many electronic properties in a single material platform. That is exactly what makes them so interesting for a new generation of energy-efficient and adaptable computer hardware," explains Ralph Claessen, spokesperson for the Würzburg-Dresden Cluster of Excellence ctd.qmat at the University of Würzburg and coauthor of the study.
Elements for an Artificial Neural Network
The Würzburg team leveraged the versatility of the oxide platform to develop devices capable of mimicking the central functions of neurons and synapses. These include a transistor for switching current, a memristor acting as a resistance-based memory module, and a memcapacitor, whose capacitance depends on its electrical history. It is particularly noteworthy that a single nanoscale device—depending on its circuit configuration—can perform various tasks. It can operate as a transistor, memristor, or memcapacitor, functioning as a highly versatile electronic multitool.
"The exciting thing about our platform is that we can realize very different functions with the same material system. This brings us closer to hardware that not only computes but can also learn directly within the device and temporarily store information," says Soumen Pradhan, a postdoctoral researcher at the Chair of Applied Physics in Würzburg and the study's lead author.
Potential Applications for Smart Biosensors
Potential applications for this self-learning, brain-inspired computing platform include health monitoring and medical diagnostics.
In the long term, such technologies could be integrated into wearable sensor systems or bioelectronic applications. They could continuously monitor critical parameters—such as heart rate, blood pressure, body temperature, oxygen saturation, and blood glucose—and intelligently evaluate the data on-site: rapidly, energy-efficiently, and without the need for additional processing units.
Although the Würzburg physics team's recent findings stem from fundamental research, they highlight the immense potential of this new generation of hardware and were recently published in Nature Communications.
Additional information: The Cluster of Excellence ctd.qmat (Complexity, Topology, and Dynamics in Quantum Matter) at Julius-Maximilians-Universität Würzburg (JMU) and Technische Universität Dresden researches and develops novel quantum materials with tailor-made properties. Approximately 300 researchers from more than thirty countries are designing the foundations for future technologies at the intersection of physics, chemistry, and materials science. In 2026, the cluster entered the second funding period of the Excellence Strategy of the German federal and state governments, bringing an expanded focus on the dynamics of quantum processes.
Published in journal: Nature Communications
Title: Oxide interface-based polymorphic electronic devices for neuromorphic computing
Authors: Soumen Pradhan, Kirill Miller, Fabian Hartmann, Merit Spring, Judith Gabel, Berengar Leikert, Silke Kuhn, Martin Kamp, Victor Lopez-Richard, Michael Sing, Ralph Claessen, and Sven Höfling
Source/Credit: Julius-Maximilians-Universität Würzburg | Theresa Kunzelmann
Edited by: Scientific Frontline
Reference Number: ms070726_02