Control of Oxygen Vacancy Ordering in Brownmillerite Thin Films via Ionic Liquid Gating

Oxygen defects and their atomic arrangements play a significant role in the physical properties of many transition metal oxides. The exemplary perovskite SrCoO3-δ (P-SCO) is metallic and ferromagnetic. However, its daughter phase, the brownmillerite SrCoO2.5 (BM-SCO), is insulating and an antiferromagnet. Moreover, BM-SCO exhibits oxygen vacancy channels (OVCs) that in thin films can be oriented either horizontally (H-SCO) or vertically (V-SCO) to the film’s surface. To date, the orientation of these OVCs has been manipulated by control of the thin film deposition parameters or by using a substrate-induced strain. Here, we present a method to electrically control the OVC ordering in thin layers via ionic liquid gating (ILG). We show that H-SCO (antiferromagnetic insulator, AFI) can be converted to P-SCO (ferromagnetic metal, FM) and subsequently to V-SCO (AFI) by the insertion and subtraction of oxygen throughout thick films via ILG. Moreover, these processes are independent of substrate-induced strain which favors formation of H-SCO in the as-deposited film. The electric-field control of the OVC channels is a path toward the creation of oxitronic devices.

† These authors contributed equally to this work. *e-mail: stuart.parkin@mpi-halle.mpg.de

Table of Contents:
Supporting Text Figure S1 to S9 Table S1 Supporting text

Structure refinements of H-SCO and V-SCO
Structural analyses of the V-SrCoO2.5 (V-SCO) and H-SrCoO2.5 (H-SCO) phases were carried out using a laboratory GaJet X-Ray source (=1.3414 Å ) and a six-circle x-ray diffractometer especially designed for probing (ultra-) thin epitaxial films and surfaces. The data collection was carried out under grazing incidence of the incoming beam (incidence angle µ=1 deg.) and by recording integrated intensities in transverse scans, i.e. by rotating the sample about its surface normal.
For H-SCO, 33 reflections were collected in total, which reduce to 26 by symmetry equivalence according to the 2mm plane group symmetry. For V-SCO we collected 22 reflections reducing to 17 by symmetry using the same point group symmetry. The average agreement of symmetry equivalent reflections is in the range of several percent. Subsequently, the squared structure factor amplitudes (|Fobs| 2 ) were derived by applying instrumental correction factors using standard crystallographic techniques (1). The structure analysis was carried out by least squares refinement of the |Fobs| 2 to the calculated ones (|Fcalc| 2 ) using the programs Shelx (2) and Prometheus (3) by allowing the atomic positions (x,y,z), site occupancy (θ) and atomic displacement parameters (ADP) to vary according to the space groups Pmmm (V-SCO) and Ima2 (H-SCO) (4). In addition, the refinement requires the consideration of the domain structure. There are two rotational structure domains for H-SCO that involve the overlap of reflections making an incoherent averaging of the reflection intensities reflected by each domain necessary. The refinement quality is measured by the unweighted residuum (Ru) and the Goodness of Fit parameter (5) which were found to be excellent in both cases. For V-SCO we derived Ru=0.03 (GOF=0.80), for H-SCO Ru=0.07 (GOF=1.13).
Figs. S8 and S9 show the schematic structural models for H-SCO and V-SCO. Here, blue, green and red balls represent the Co, Sr, and O ions, respectively. H-SCO can be considered as an approximate (√2 × √2 × 4) superstructure with respect to the bulk STO (001) substrate, while V-SCO corresponds to a (√2 × 1 × 1) superstructure (see also below). In Figure S8, we compare the film structure with the structural model published in Ref. [6]. This is achieved by superimposing the bulk model (balls in the foreground) on the structural model of the film (balls in the background). It is clearly evident that there exist only minor differences, especially near the Co atoms at z=1/4, but these lie within the experimental uncertainty. Figure S9 shows the model of the V-SCO film structure, which is characterized by the presence of two vacancy sites labelled (1) and (2)       according to the analysis of Ref. [6]. Only minor differences are observed. Numbers label interatomic distances in Å ngström units.     A channel with an area of 2×2 mm 2 was formed and then Ru (~5 nm)/ Au (~70 nm) electrodes were formed. The gate electrode with an area of 0.8×2 mm 2 was formed on the chiplet from the same Ru/Au bilayer and was placed adjacent to the device itself . The device was attached to the sample holder using double-sided tape. After placing the IL on the device surface, a Kapton film was attached to thin the IL. The contact wires were connected, then the ILG was performed during the XRD measurement. (b) A photo of the ionic liquid device for the in situ XRD measurement.