Direct observation of oxygen vacancy-driven structural and resistive phase transitions in La2/3Sr1/3MnO3

Resistive switching in transition metal oxides involves intricate physical and chemical behaviours with potential for non-volatile memory and memristive devices. Although oxygen vacancy migration is known to play a crucial role in resistive switching of oxides, an in-depth understanding of oxygen vacancy-driven effects requires direct imaging of atomic-scale dynamic processes and their real-time impact on resistance changes. Here we use in situ transmission electron microscopy to demonstrate reversible switching between three resistance states in epitaxial La2/3Sr1/3MnO3 films. Simultaneous high-resolution imaging and resistance probing indicate that the switching events are caused by the formation of uniform structural phases. Reversible horizontal migration of oxygen vacancies within the manganite film, driven by combined effects of Joule heating and bias voltage, predominantly triggers the structural and resistive transitions. Our findings open prospects for ionotronic devices based on dynamic control of physical properties in complex oxide nanostructures.


Supplementary Note 1. STEM-EELS analysis
The initial perovskite structure, brownmillerite lattice, and third phase of the LSMO film were characterized using STEM-EELS. The O K edge fine structure in EELS spectra provides information on excitations from O 1s electrons to 2p bands. The pre-peak of the LSMO O K edge correlates with the occupation of the Mn 3d band, providing information on the oxidation state of Mn ions. A drop in the pre-peak intensity and a shift towards higher energy losses indicate an increase in oxygen vacancy concentration [1][2][3]. Supplementary Figure 3a depicts STEM-EELS spectra for the different structural phases of LSMO. The pre-peak of the O K edge is reduced for the brownmillerite structure (both horizontal and vertical) in comparison to the EELS spectrum of the original perovskite lattice (yellow curve). This observation confirms a reduction of the Mn oxidation state via electro-thermal migration of oxygen vacancies towards the LSMO contact area during the application of positive voltage pulses. After the perovskite structure is re-established by negative voltage pulses (blue curve), the oxidation state of Mn ions is also restored. The pre-peak of the O K edge is further reduced for the perovskite structure 3 with enhanced out-of-plane lattice spacing (third phase, pink curve). The concentration of oxygen vacancies in the third phase is thus higher than that of the oxygen-deficient brownmillerite phase. The conclusions on oxygen vacancy concentration are further supported by a comparison of the Mn white line L3/L2 peak ratio (Supplementary Figure 3b). Since an increase of the L3/L2 peak ratio implies a lowering of the Mn oxidation state [1][2][3], the concentration of oxygen vacancies grows in the following order: perovskite lattice  brownmillerite structure  third phase. Migration of oxygen vacancies between the LSMO film and Nb-doped STO substrate was also inspected by STEM-EELS.

Supplementary Note 3. Stability of structural phases and resistance states
The stability of the different structural phases after resistive switching was monitored using in situ TEM.
Supplementary Figure 11 shows an example for the brownmillerite phase. In this experiment, the original perovskite phase is first transformed into the brownmillerite structure by a positive voltage pulse of Vp = +4.5 V. As a result, the resistance increases. The resistance and lattice structure are subsequently monitored at a small measurement voltage of Vm = 0.2 V. The resistance remains constant for 155 minutes (Supplementary Figure 11a), after which it increases due to an abrupt disconnect between the metal tip and the LSMO film (see Supplementary Figure 11d). The change in resistance is thus not caused by an instability of the brownmillerite phase but, rather, by an artefact in the in situ TEM measurement. The brownmillerite structure remains stable well beyond the duration of this experiment. In a separate test, a TEM specimen with brownmillerite LSMO film was stored for two months. STEM-HAADF imaging after two months did not reveal any relaxation into the original perovskite phase (Supplementary Figure   12). The brownmillerite structure and accompanying high resistance state that are induced by combined effects of applied electric field and Joule heating are thus stable over time. The stability of the third phase was also assessed. Supplementary Figure 13 summarizes an experiment wherein the perovskite LSMO film is first transformed into a coexisting brownmillerite and third phase. STEM imaging of the same area after 10 days (Supplementary Figure 13d) reveals that the boundary between the two structural phases has only meandered a bit in favor of the brownmillerite structure. This observation suggests very slow re-ordering of oxygen vacancies near the phase boundary.
Supplementary Figure 11. a, Resistance versus time for a LSMO film with a brownmillerite structure.
In this experiment, the brownmillerite phase is first induced by a positive voltage pulse of Vp = +4.5 V and the resistance is subsequently monitored at a measurment voltage of Vm = 0.2 V. b-d, STEM images of the metal tip/LSMO contact area during several stages of the experiment.