Introduction: To answer the increasing need for data storage, several forms of memory called random access memories (RAM) have been developed. Oxide resistive RAM (OxRRAM), based on switching between a low and high conductive state, are considered as one of the most promising candidates for replacing FLASH technology in the next memory generation. Forming and breaking a nanometer-sized conductive area is commonly accepted as the physical phenomenon involved in the switching mechanism of OxRRAM [1]. Nevertheless, the nature of this filament is still highly debated because oxygen vacancies [2] on one side and and metallic migrations from the electrode on the other side [3] have been evidenced.By combining high spatial resolution and local chemical analyses coupling TEM with EELS and EDS, we propose to investigate and compare two approaches for confining and analyzing the filament of a state of the art OxRRAM device: (i) ex-situ polarization using a conductive atomic force microscopy (C-AFM) followed by FIB preparation and (ii) FIB preparation followed by in-situ polarization within the TEM. We will show the advantages and disadvantages of each approach, the principal challenge remaining the preparation of a TEM lamella which contain the nanometer-sized filament.
Ex-situ forming approach: This first approach consists to block the memory resistance in an operation state, then prepare a thin lamella by FIB before placing it in a TEM for analysis. Experiments are performed on small structures (< 100nm) with patterned top electrode to confine the filament in a sample whose size is suitable for TEM analysis. The resistance switching is realized with a C-AFM which allows both to localize the structure and to locally inject current in the memory (Fig. 1. a)). The I(V) curve of the forming operation (transition of a high resistance virgin state to a low resistance state) is shown in figure 1. a). A STEM image of the structure and the corresponding STEM-EELS map of the extracted Titanium signal are shown in figures 1. b) and 1. c), respectively. A Ti-rich region with a conical shape is clearly observed in the HfO2 layer (see the blue area inside the white dotted rectangle). This conical filament seems to connect the top electrode to the bottom electrode like previously reported [3]. We will discuss the limits of this method. How can we rigorously compare two states of the device in many lamellae with different thicknesses? Is the memory state stable over time or during the FIB preparation?
In-situ forming approach: In the second approach, the thin lamella is prepared from the memory. It is then loaded on a Nanofactory in-situ holder where a tip is used for contacting the device (Fig. 2 a)), applying a voltage and simultaneously monitoring the current. A protocol was developed to optimize sample preparation and electrical contacts with the probe and avoid mechanical stress and heating phenomena which can generate measurements artifacts or simply destroy the sample. Figure 2. b) presents the I(V) curve of the forming operation obtained within the TEM. Clearly, the forming step of the memory has been completed with 100µA compliance. No Titanium migration in the HfO2 layer has been observed by EDS (Fig. 3 a) and b)). EELS measurements before and after in-situ forming are on-going to investigate the fine structure of the Oxygen K edge in HfO2 and explain the change of resistance.
Conclusion: Two complementary polarization protocols have been developed in this work. Using C-AFM to electrically test memories is relatively easy to setup but does raise questions about external contamination and analysis of commercially available devices. With the in-situ approach, the associated experimental work is much heavier but allows complex electrical testing without any external contamination. Ti migration in the HfO2 layer was observed with the first approach, which suggests that the memory operates as a CBRAM (conductive bridge random access memory). A different behavior seems to occur with the in-situ approach. On-going studies of the Oxygen K edge fine structures will probably help to explain the observed behavior. The reproducibility of these first results will also be checked.
References
1. Waser, R., et al., Advanced Materials, 2009. 21(25-26): p. 2632-2663.
2. Calka, P., et al., Nanotechnology, 2013. 24(8).
3. Privitera, S., et al., Microelectronic Engineering, 2013. 109(0): p. 75-78.
Figures:

Figure 1. a) Forming curve measure with C-AFM (compliance current 400µA), b) STEM image of the memory biased with C-AFM and c) STEM-EELS map of extracted Titanium signal after forming. Ti migration area is seen inside the white dotted rectangle.

Figure 2. a) STEM image of the TEM lamella and the tip of the Nanofactory in-situ holder. The memory device is located inside the blue dotted rectangle and b) forming curve measured with the Nanofactory in-situ holder (compliance current 100µA).

Figure 3. EDS map after forming of a) Hafnium and b) Titanium. Maps have been done just under the area in contact with the tip. Similar maps have been observed along the memory after forming.
To cite this abstract:
Tristan Dewolf, Vincent Delaye, Nicolas Bernier, David Cooper, Nicolas Chevalier, Helen Grampeix, Christelle Charpin, Eric Jalaguier, Martin Kogelschatz, Sylvie Schamm-Chardon, Guillaume Audoit; Nano-characterization of switching mechanism in HfO2-based oxide resistive memories by TEM-EELS-EDS. The 16th European Microscopy Congress, Lyon, France. https://emc-proceedings.com/abstract/nano-characterization-of-switching-mechanism-in-hfo2-based-oxide-resistive-memories-by-tem-eels-eds/. Accessed: December 4, 2023« Back to The 16th European Microscopy Congress 2016
EMC Abstracts - https://emc-proceedings.com/abstract/nano-characterization-of-switching-mechanism-in-hfo2-based-oxide-resistive-memories-by-tem-eels-eds/