skip to content

Driscoll Research Group


Theme 1. Energy-Efficient Non-Volatile Memory

We are at the beginning of a Data Age. There is a wide range of rapidly growing data-centric technologies spanning Big Data, IoT, transport, medicine, electric vehicles, security, entertainment, and neuromorphic computing for AI. In most of these areas, memory energy dominates over compute energy and low-power non-volatile memory (NVM) is urgently needed. Also, neural network based deep learning embedded in edge computing devices could provide the solution to the high computing efficiencies.

Our three main projects in the field of low-power NVM are funded though Prof. Driscoll's Royal Academy of Engineering Research Chair, her ERC Advanced Grant, and an ECCS-EPSRC grant in collaboration with researchers from the USA. Here, we provide a general overview over our research in this field.

Conventional computer memory is either fast, but volatile (Randem Access Memory - RAM) or non-volatile, but slow (Hard Disk Drives (HDDs) or Magnetic Tape). The closest technology in between these two extremes is NAND Flash such as in Solid State Drives, which is non-volatile and achieves higher speed than HDDs through parallelisation of read and write operations. However, Flash requires much higher programming voltages than the other technologies, so that it is not suitable for low-power applications either.

In non-volatile memory such as HDDs and Magnetic Tape, information is stored in the magnetic polarisation of a storage material, whereas in RAM devices, information is stored in the form of charge on a dedicated capcitor (RRAM) or the charge on the gate terminals of two cross-coupled inverters (SRAM). Charge is prone to leaking out of such terminals, which makes the fast charge-based memores volatile. Our approach to designing new and energy-efficient computer memory relies on the nanoengineering of oxide thin films, where information can be stored in different materials properties other than charge. Such properties can be a stable electrical resistance, the direction of ferroelectric rather than ferromagnetic polarisation, or the arrangement of oxygen vacancies within an oxide. The video below provides more information.


Video 1. An introduction to Sustainable Memory Devices by Dr. Guliana di Martino, Dr. Chiara Ciccarelli,
and Prof. Judith Driscoll, narrated by Sunny Howard.


a) Resistive memory (and neuromorphic computing for AI): Scaleable, uniform and robust resistive switching.

Arguably, oxide memristors represent the most ideal NVM system in terms of simple composition, potential for lowest cost and highest density (Fig. 1). However, problems still remain in terms of scaling, uniformity and robustness. Memristors could offer the required, low power, technology for both non-volatile memory and neuromorphic computing. However, currently, there are problems of non-uniformity, difficulty to scale to small dimensions (20 nm and below), and robustness. These are all materials-related problems. We have demonstrated ways to overcome them in model systems (Fig. 2). We are currently working on translating these model systems to industry practical systems.

Figure 2. Oxide memristor films. (a) As-grown. (b) Schematic illustration of the strain mediated ME effect VAN films. (c) Magnetic hysteresis loops along the OOP direction and (d) magnetic hysteresis loops along the IP direction with and without applying in-situ voltages (plots courtesy of Chao Yun, Driscoll group).


b) Magnetoelectric memory: Magnetoeletricity above room temperature in simply- grown, self-assembled composite thin film systems.

Electric field control of magnetism (magnetoelectricity) could provide for ultra-high density non-volatile memory. No currents are passed during switching and so these systems have potential for ultra-low power. However, despite intensive research efforts, no practical materials systems have emerged. Interface-coupled, composite systems containing ferroelectric and ferri-/ferromagnetic elements have been the most promising, but they have many problems, e.g. substrate clamping, large unwanted currents (leakage), and they cannot be miniaturized to give high density recording. Through careful materials selection, design, and nanoengineering, we have demonstrated a high-performance room temperature magnetoelectric system (Fig. 3). A vertically aligned triple nanocomposite structure in which the strain coupling is independent of the substrate was used and a new, low leakage ferroelectric material was employed. Large converse magnetoelectric coefficients have been achieved of >10e-9 s/m.

Research Theme 3Figure 3. Large RT ME effect in composite film. (a) typical AFM image of sample surface. (b) Schematic illustration of the vertical strain mediated effect. (c) Magnetic hysteresis loops along the OOP direction with different voltages applied.


Video 2. Laboratory tour by Prof. Judith Driscoll and Dr. Giuliana Di Martino demonstrating their work in
the field of nanoplasmonics.

Example references:

Rui Wu, Ahmed Kursumovic, Xingyao Gao, Chao Yun, Mary E. Vickers, Haiyan Wang, Seungho Cho, and Judith L. MacManus-Driscoll. Design of a Vertical Composite Thin Film System with Ultralow Leakage To Yield Large Converse Magnetoelectric Effect. ACS Appl. Mater. Interfaces, 2018, 10 (21), pp 18237–18245 (DOI: 10.1021/acsami.8b03837).

Choi EM, Maity T, Kursumovic A, Lu P, Lee OJ, Bi Z, Park Y, Wu R, Gopalan V, Wang H, MacManus-Driscoll JL, Nanoengineering Giant Room Temperature Ferroelectricity into Orthorhombic SmMnO3 Films, Nature Communications. May 2020; 11, 2207.