Tuneable multilevel capacitance switching in hafnium-based oxides

Abstract

Memcapacitors are devices exhibiting history-dependent capacitance in response to applied electrical stimuli and have attracted increasing attention as potential building blocks for adaptive, reconfigurable, and energy-efficient electronic systems. Unlike conventional tunable capacitors that rely on continuous biasing or mechanical actuation, memcapacitors offer non-volatile capacitance tuning, enabling circuits that retain their programmed functionality without static power consumption. However, practical implementations that are compact, CMOS-compatible, and capable of reliable multilevel operation remain limited. This thesis therefore investigates memcapacitive behaviour in metal–dielectric–metal device architectures. Hafnium oxide is chosen as the dielectric material due to it’s high dielectric constant, ultrathin scalability, and CMOS compatibility, while appropriate compositional modification also enabling ferroelectric behaviour.

These versatile properties allow hafnium oxide to be explored across two distinct physical regimes: defect-mediated switching in conventional hafnium oxide and polarisation-driven switching in ferroelectric hafnium zirconium oxide (HZO). This provides a unified framework to study the evolution of memcapacitive behaviour from binary to multilevel operation. In the first part of the thesis, memcapacitive behaviour is examined in conventional hafnium oxide memristor. Although primarily investigated for resistive switching, the geometry is inherently capacitive, offering an opportunity to study capacitance modulation. Electrical characterisation reveals binary capacitance switching with a capacitance ratio of 1:3 and a memory window of ∼3 pF in a 20×20 μm2 device, programmed using voltage pulses up to 5 V. Scaling the device area increases the absolute capacitance and enhances the capacitive ratio, confirming that the modulation scales with device geometry. Finally, an equivalent circuit model is developed to quantitatively capture the coupled resistive–capacitive behaviour. Building on this binary baseline, the thesis then investigates ferroelectric HZO capacitors, where capacitance is controlled by switchable polarisation rather than defect dynamics. These devices exhibit a substantially enlarged non-volatile capacitance window of approximately 24 pF and enable access to more than eight stable capacitance states, with endurance ∼10^6 cycles and retention beyond 10^5 s. The states are well separated, with switching behaviour analogous to potentiation–depression, and can be programmed using voltages within ±3 V for a 60 × 60 μm2 device size. Area-scaling analysis reveals a progressive reduction of the capacitive memory window with decreasing device size. At 20 × 20 μm2, the memory window becomes comparable to that observed in memristive devices, while voltage-dependent analogue switching remains intact. In this context, a single reconfigurable ferroelectric device can deliver capacitance densities of approximately 27 fF/μm2, which are competitive with current MIM technologies, while simultaneously providing non-volatile and programmable capacitance states. This combination offers a pathway to reduce overall capacitor footprint by replacing multiple fixed-value elements with a single tunable memcapacitor. Further studies of frequency-dependent be- haviour, dielectric loss, and DC leakage establish the operational limits and practical constraints governing reliable multistate performance. To verify circuit-level integration, the HZO devices were employed in a tunable high-pass filter and a relaxation RC oscillator. The filter cut-off frequency was tuned by 4.4 kHz, while the oscillator frequency was tuned by 6.5 kHz, directly translating non-volatile capacitance states into programmable frequency responses. In addition, complementary electrostatic force microscopy measurements confirm local charge and polarisation modulation, providing microscale evidence of stable multistate capacitive behaviour. Overall, this thesis presents a comprehensive progression of memcapacitive behaviour in hafnium-based devices, from binary defect-mediated switching to multilevel ferroelectric operation, supported by electrical, microscopic, and circuit-level analyses. The results establish hafnium-based ferroelectric capacitors as a scalable and versatile platform for non-volatile analogue capacitance tuning, with relevance to reconfigurable electronics, adaptive analogue systems, and emerging memory-centric computing architectures.