Programable Multi-Level Graphene/PZT Memristor Based on Highly Conductive Neutral Domain Walls

Domain wall nanoelectronics is an attractive device concept for information processing, which relies on the 2D interfaces between domains of electrical polarization. Within a ferroelectric material – possessing spontaneous polarization switchable by an electric field – such domain walls can represent a functional entity with properties distinctly different form the parent phase. In particular, conductive domain walls in the normally insulating ferroelectric draw a lot of interest for memristor-type elements with potential applications in neuromorphic networks, but progress in the field is impeded by a relatively low conductivity of the individual domain walls – typically in the picoampere range. Until now, sufficient conductance for single channel devices was generally only reported for charged domain walls, which often require elaborate poling procedures and, in many cases, form unstable transient configurations. Here we demonstrate a programable multi-level memory device, which is based on highly conductive, yet nominally neutral and therefore nonvolatile single 180° domain walls in the archetypical ferroelectric system – tetragonal Pb(Zr,Ti)O₃ (PZT). The memristor structure consists of 60nm PZT epitaxially grown on a DyScO₃/SrRuO₃ substrate with a graphene top electrode. The high electrical conductivity combined with robust mechanical properties for ultra-thin layers of graphene offer unparalleled opportunities for through-electrode domain dynamics monitoring combined with analysis of memristive characteristics. Furthermore, the graphene top electrode enables us to use a variety of techniques for domain wall motion-control including electrical poling or local mechanical pressure applied through the graphene layer via the AFM-probe.<br/>Our nanometer-size memristive device containing a single conductive domain wall shows remarkable current levels of over 100nA for read pulses of 2V, and On/Off ratios of over 4 orders of magnitude. These domain walls – formed between adjacent c-domains with opposed polarity – could be written, reconfigured, and erased by the application of sufficiently high electric fields, which locally alter the polarization state of the ferroelectric material. Once written these channels show excellent positional stability during readout and robust retention within several days. We were able to demonstrate the injection of domain walls underneath our graphene electrode by applying short low-voltage (&lt;6V) pulses through the conductive AFM probe. The presence of the injected DWs switched our two-terminal device from a pristine high-resistive state into a low-resistive state with high read-out currents. The reached low-resistive state was stable under application of several tens of read-out pulses and could be completely reversed by a “reset”-pulse of opposing voltage bias. The device showed no degradation over the span of our measurement tests, which included hundreds of switching and reading pulses. The excellent quality of the PZT/graphene contact enables us to obtain high-resolution piezo-force microscopy images of the domain configurations for different resistivity states and therefore associate the low-R/high-R states with the total number of injected domain walls. Because of the enhanced stability of the mixed-domain configurations due to the graphene electrode, it was also possible to gradually change the domain geometry and modulate the resistivity level accordingly. Through the application of consecutive single pulses and pulse trains with positive and negative biases slightly above the coercive field, a minimum of tens of different resistivity levels could be addressed within the device. The achieved performance including the high current in the on-state originating from only a few single domain walls, strong retention, and addressable multi-level resistance states together with the low-voltage read and write pulses demonstrate the potential of PZT/graphene bilayers for future neuromorphic applications.