The massive proliferation of connected devices and digital services is generating amazing amounts of new data, already known as the ‘oil’ of 21st century. How to quickly convert these data into useful information and store them efficiently is the current concern. Emerging Non-Volatile Memory (NVM) devices beyond flash memories are receiving substantial attention to cover the current data storage capacity gap, because they combine the non-volatility with a low latency, a high integrated density and low power consumption. In the last few years, NVM devices based on the electrically switchable resistance phenomenon have generated significant interest both in the industry and in the scientific community. Among ReRAM, metal oxide thin films are one of the most promising family of materials due to their compatibility with silicon mainstream fabrication technology, their simple metal-insulator-metal (MIM) structure and the good results of resistive switching (RS) they have already shown. A huge variety of binary metal oxides showing hysteretic RS, such as NiO, TiO2 or HfO2, have been extensively studied. Recently, particular attention has been drawn to the development of new ReRAM multilevel (ML) memories to significantly enhance the storage density, i.e. multiple distinguished states in the metal oxide can be used to store data, and enable new computing paradigms such as the neuromorphic computation. Although binary metal oxides present desirable working performance, their filamentary and stochastic nature (i.e. formation and breakdown of conductive filament paths of oxygen vacancies in the material) leads to a very strong variability, which severely limits their integration in a large matrix and disable their multilevel programming.
Perovskite oxides appear as strong candidates for the development of novel multilevel ReRAM switching devices, since they tolerate large variations in the oxygen content while maintaining their crystal structure. Currently, the ML storage properties of a number of perovskite materials and heterostructures has been successfully demonstrated. Despite the significant advances in perovskite-based ReRAM devices, the underlying RS mechanisms remain unclear and thus constitute a significant barrier to their widespread application . How to tune at best a perovskite resistive switch response or how to precisely control the size/resistance of the nanoscale defects causing the resistance change is still a critical challenge.
One of the major issues that this technology faces is to improve the reliability and stability of the MIM structure. The synthesis of perovskites demands high temperatures (typically above 500°C), which are incompatible with the thermal stability of metals. Metals suffer from quick thermal degradation, known as dewetting, in which agglomeration starts. The morphology of the metal evolves to a discontinuous and rough layer and finally breaks down into isolated particles, losing its functionality (breakdown of in-plane percolation) and leading to serious problems of device reproducibility and reliability. Up to know, the most adapted alternative is the use of an additional conductive oxide. However, this extra oxide film complicates enormously the study of the RS mechanisms, as it can contribute or modify the RS process through the mobility of O2- (or VO..). It is noteworthy to mention that TiN, which is also commonly used as conductive material, is not compatible with the synthesis of perovskites because it easily oxidizes from 450°C . Therefore, the development of new electrode-perovskite-electrode (EPE) structures which are thermally and chemically more stable, and which show an effective blocking to oxygen, is of great interest to control the RS of the functional layer and to be able to deeply investigate the RS mechanisms in perovskites.
The overall aim of the project is to move from fundamental studies of new EPE heterostructures based on perovskite thin films to their direct application as ML ReRAM devices, by:
1) Understanding of the nanoscale mechanisms governing the RS process.
2) Control, tuning and reproducibility of the RS behaviour.
3) Optimization of the RS response.
4) Demonstrating the effectiveness and application of the optimized EPE structures by designing and fabricating a reliable perovskite-based ML ReRAM device as a proof-of-concept.
