ROLE OF ACTIVE ELECTRODE PURITY IN ECM MEMRISTIVE DEVICES

In last decades the request for progressively larger dataset elaboration has pushed integrated system technology towards its limit. The development of novel technologies and computational paradigms came out to be necessary for overcoming miniaturization issues and fulfilling future applications requests. In this scenario, redox based resistive switching random access memory (ReRAM), based on memristive devices, has proven to be a promising candidate. Memristive devices encode information in the form of internal resistance variation (called resistive switching), caused by the formation and destruction of highly conductive filaments through a thin oxide layer. The fundamental processes underlying switching effect are redox reactions and ionic motion, driven by an electric field. ReRAM shows suitable properties for digital implementations such as non-volatility, short write/read times, low power consumption, long endurance and retention. Moreover, the opportunity to program the resistance state in an analog way and the device capability to show spike timing dependent plasticity pave the way to neuromorphic applications. Alongside this favourable features, the full understanding of material-properties relation must still be deepened and the lack of complete design rules, together with the intrinsic high variability due to stochastic switching nature, prevents this technology from large scale employment at present. In this work the impact of copper electrode purity upon the electrical properties of an electrochemical metallization memory (ECM) is investigated. The specific cell structure under study is a metal-insulator-metal (MIM) stack composed by a platinum reference electrode, a Ta2O5 insulating layer and a structured copper active electrode. It is produced by sputtering deposition and photolithographic lift-off in a common bottom electrode fashion. The electrical behavior is characterized through I-V sweep measurements, cyclic voltammetry and pulse measurement for set time evaluation. It has been observed that, regarding set time, low purity devices exhibit a faster switching in the low voltage region, where the time limiting process is assumed to be electron transfer. Furthermore, quasi-static figures of merit, namely set voltage, reset voltage, low-resistance state, high resistance state and resistance ratio show systematic lower variabilities for high purity devices. For this reason, active metal purity engineering can be expected to have a fundamental role in the overcoming of ReRAM variability issue, permitting to approach the target of industrial production.