Data centres and AI systems waste up to 80% of their computing energy shuttling information between memory and processing units, a bottleneck that today’s electronics cannot fix. This project tackles that inefficiency by building a new kind of memory device that combines light and electricity to switch materials at the nanoscale. The researchers will use a technique called plasmon-enhanced light-matter interaction to watch ferroelectric switching in hafnium dioxide (HfO₂) in real time, without the electron-beam damage that blurs conventional microscopy. They aim to understand how oxygen vacancies, defects, and domain walls behave during switching, and to develop light-triggered ferroelectric switching for the first time. If successful, the work could guide the design of non-volatile memory and neuromorphic computing components that are faster, more reliable, and far more energy-efficient than current silicon-based architectures. This is fundamental materials science with a clear engineering payoff: better hardware for the data centres, autonomous vehicles, and AI systems that already underpin modern infrastructure.
View original technical description
We live in an information driven society, where we see proliferation of data centric technologies, e.g. self- driving vehicles, data centres, IoT and AI. Data complexity is explosively growing and data centers consume an increasing fraction of total world energy use. With the current von Neumann architectures up to ~80% of the computing energy is consumed in the data-transfer bottleneck between logic and memory on interconnects. To progress beyond this limit new types of device are needed. Neuromorphic systems, mimicking the brain nervous system, shine for proficiency in cognitive and data-intensive tasks, providing high computing efficiencies and low power consumption. Ferroelectric memories could offer the required technology for both non-volatile memory and neuromorphic computing. PlasmoFED links low-energy nanoscale device engineering and plasmon-enhanced light-matter interactions by implementing optically accessible memory devices to investigate ferroelectric switching materials in ambient conditions, in real-time and in-situ during device operation. We devise a non-destructive technique able to avoid electron-induced perturbation of the switching process present in traditional electron microscopy techniques. The industry-standard material HfO2 will be explored but with entirely new multifunctionality of ferroelectricity and ionic conductivity. PlasmoFED will focus on nanoscopic in-operando access to this hybrid switching process, tackling the current problems of stability in RRAMs. PlasmoFED will also address the current problems of scalability and reliability in FeRAMs aiming to understand the role of oxygen vacancies, defects and domain wall propagation in HfO2. The concept of light triggered ferroelectric switching will also be developed. PlasmoFED will provide critical knowledge for materials and device engineers to guide the creation of devices of unparalleled performance. The potential big win is new devices based on HfO2 for memory and AI applications.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know