Complex oxides for computing beyond von Neumann

Reducing transistor dimensions cannot sustain the growing demand for better technology. To reduce the power consumption while increasing the technological performance we can take inspiration from the brain – a naturally energy-efficient system and focus on integrating more intelligent components on chips. The brain’s efficiency largely stems from the co-location of memory and processing units, which we can emulate using smart materials and devices that perform the functionalities of neurons and synapses. Leading candidates for this are devices that can switch between multiple resistive states through an external stimulus. Anouk Goossens' thesis focuses on building such devices using complex oxides - a class of materials that is highly tunable due to a strong coupling between different degrees of freedom.  

Goossens explores resistive switching in interfacial memristors based on metal contacts on the unconventional semiconductor Nb-doped SrTiO3. These devices are tunable from both sides of the interface. By reducing the metal electrode area the resistance ratio is enhanced, which is an unconventional but desirable effect – downscaling tends to negatively impact device performance, but Goossens found the opposite to be the case. Altering the doping concentration allowed her to control a range of parameters including the stochasticity, memory window and switching speed.

Goossens also worked on spintronic devices using ferromagnetic SrRuO3 layers. Goossens: 'we show the ability to control the magnetic anisotropy and influence the magnetisation using current – two key parameters for device scalability. By integrating these layers into magnetic tunnel junctions we demonstrate relatively large changes in resistance and multiple non-volatile resistance states.'

Dissertation: https://hdl.handle.net/11370/ac34e663-5a09-427f-9c1c-9128f2e11ea8