The problem being addressed is to achieve efficient and robust materials for non-volatile memory and neuromorphic computing by materials innovation. The problem is very important both for new forms of computing with much greater power and speed and also for achieving operation at much low power.
The objectives are:
1. Translate high performance key model memristor system to industry platform film to meet industry performance metrics. The industry platform films will be comprised of sub-5 nm, columnar structured, ionic thin films (preferably of HfO2, but other systems if their performance is better) with conducting column boundaries.
2. Use entirely new hybrid/VAN composites to enable unprecedented scaling, uniformity and robustness.
Both 1 and 2 arre for non-volatile memory and neuromorphic omputing.
We explored group IV oxide systems for oxygen defect resistive switching, namely HfO2, CeO2, and ZrO2, combinations of these systems, and also doping in them. They are semiconductor industry compatible. By exploring the combination and by doping our aim was to a) induce partial engineered filaments that act as contacts to the interface layers, and b) control oxygen vacancy content to achieve interface switching control.
In terms of combining the group IV oxides, e.g. HfO2/ CeO2, we learned how to create guided filaments for controlled switching (Dou H et al. ASC Applied Electronic Materials: Nov. 2021; 3, 5278). This is very beneficial for eliminating the high voltage forming process that is typically required to initiate the switching process.
In terms of doping effects, created a very interesting doped amorphous HfO2 system which shows high performance, uniform, and reproducible multi-level resistive memory (Hellenbrand M et al, Science Advances, Jun. 2023, 9, eadg1946) (Fig. 1). The material was deposited at an industry-process-compatible temperature of 400 °C. The films are grown on TiN on Si with W top electrodes and thus have excellent potential for CMOS compatible devices.
The above work gave results which represent a breakthrough in terms of uniformity, a current bottleneck. This is because we created an interface switching device. The system has significant improvements over other emerging memory technologies for CMOS-compatible non-volatile embedded memory applications. It could replace embedded Flash memory in IoT applications, and be in edge computing and neuromorphic and analogue computing purposes.
We also did some basic studies on the influence of protons on resistive switching in SrTiO3 (STO). STO serves as useful test case system and it is of interest as a buffer forsilicon. We found protons from moisture in the atmosphere can play a role in resistive switching in STO (Kunwar S, Advanced Electronic Materials Oct. 2022, 2200816). However, we did not find that protons dominated switching with oxides grown on top of the STO, at least not under normal atmospheres.
We also explored ion channel systems and explored brain-like ions for switching We have 2 key papers in preparation demonstrating close mimicking (for the first time) of these brain-like ion motion and switching across channel structures.
We also focused on studies about whether we can achieve a memristor by exploiting a Schottky-to-Ohmic transition. With such a transition, the memristor is expected to exhibit a significantly larger on/off ratio compared to other structures, e.g. conventional ferroelectric tunnel junctions (FTJs). We made ionic HfO2 by doping in 5% Y to induce oxygen vacancies. We made model epitaxial films od very high quality, 4.5-nm-thick Y-doped HfO2 (YHO) film by pulsed laser deposition (PLD) on LSMO-buffered Nb-STO substrates. The ferroelectricity of the bare films was investigated by X-ray diffraction (XRD) (Fig. 2) and piezoresponse force microscopy (PFM) (Fig. 3). The films are ferroelectric. Current-voltage (J-V) characteristics of the YHO films were carried out in two-terminal device configuration (Fig. 4). We found a Schottky-to-Ohmic transition with a large on/off ratio of more than 500, much larger than the on/off ratio of conventional FTJs. The semiconducting nature of our YHO films, results in radically different current–voltage characteristics to standard tunnel junctions or memristors which are formed from fully depleted insulating ferro-electric barriers. Hence, we have a polarization-modulated transition from Schottky-barrier-controlled charge transport to Ohmic conduction (Muller M, et al. Applied Physics Letters, Aug. 2022; 121 09350).
We also explored other oxygen ion conducting and inorganic ferroelectric systems, i.e. new complex ferroelectric/multiferroic perovskites, and yttria stabilised zirconia. These help to formulate our overall understanding of combining ionic and ferroelectric effects. We did HAXPES studies on these systems and papers are in progress on the works.
Finally, we undertook basic studies of plasmon-enhanced spectroscopy on doped HfO2 films, Hf0.5Zr0.5O2 (HZO) films, to learn about mechanisms of resistive switching. This work with Dr. Giuliana DiMartino, an ERC STG grantee. The plasmonic field confines in a region where agglomeration of atomic defects can be triggered by applying an external voltage, allowed us track in-operando ion migration within the switching material. This allows switching in real-time and in-situ during device operation.
We achieved a breakthrough in the understanding of fundamental physical effects governing switching (Fig. 5). We viewed nanoscale direct tracking of oxygen vacancy migration in 5nm HZO during a pre wake-up stage and showed a structural phase change which influences the device wake-up and which leads to device fatigue. Those effects alter the remanent polarisation value of a device (i.e. increasing during wake-up and decreasing during fatigue), causing the expected device performance to vary over consecutive electronic switching cycles (Jan A et al. Advanced Functional Materials, March 2023; 33, 2214970).
