Amit Kumar Singh1 2 3 Serge Blonkowski1 Martin Kogelschatz2

1, STMicroelectronics, Crolles, , France
2, LTM-CNRS, Grenoble, , France
3, CEA-LETI, Grenoble, , France

The ongoing research in the field of non-volatile memories aims to overcome the limitation of flash memories in terms of scalability. Several emerging memory technologies, especially phase-change random access memories (PCRAM) and oxide based RAMs (Ox-RAMs) are already in production or explored as potential successors. A major interest of OxRAM is its ability to change the Set and Reset resistance values according to the applied electrical constraints, which is an attractive feature for neuromorphic applications.

In order to improve the understanding of OxRAM resistive switching and area scaling, electrical characterization at nanometric scale by Conductive-AFM (C-AFM) has been performed generally in atmospheric environment. In order to avoid the experimental artefacts from electro-chemical reactions due to humidity, we explore the resistive switching of OxRAMs at nanometric scale by C-AFM under vacuum. Moreover, we compare C-AFM measurements under vacuum to the ones made on standard devices in order to provide information on the switching mechanisms and scalability.

We report the observation of resistive switching cycles performed by Atomic Force Microscopy under 10-7 Torr vacuum environment by using two samples with different bottom electrodes, TiN/HfO2 and TiN/Ti/HfO2. A detailed experimental study concerning the effect of: the polarity, the nature of the C-AFM tip, the bottom electrode, the serial resistance and compliance current on the resistive switching current- voltage cycling characteristics is reported. In particular, the influence of the tip work function and wear out on current-voltage characteristics is analyzed. Those complete switching current-voltage cycling characteristics are compared to those measured on TiN/Ti/HfO2/TiN devices. The main Set and Reset characteristics are discussed. The large set voltage observed at nanometric scale with respect to the standard devices is analyzed for example. SEM EDX measurements have been carried out to evaluate the chemical composition of the C-AFM tip after I-V measurements on TiN/HfO2 and TiN/Ti/HfO2 samples. The experimental data are then compared to model based simulations.

It appears that the difference between C-AFM measurements and those obtained on devices cannot be explained only by top electrode area scaling. It is found that the nature of the tip and the quality of its contact with the substrate surface determine the differences of switching characteristics between all the above mentioned experiments. The model results indicate that those differences might be due to differences in filament morphology according to the nature of the top electrode or C-AFM tip.