Ionic Liquid Gate-Induced Modifications of Step Edges at SrCoO2.5 Surfaces

Intense electric fields developed during gating at the interface between an ionic liquid and an oxide layer have been shown to lead to significant structural and electronic phase transitions in the entire oxide layer. An archetypical example is the reversible transformation between the brownmillerite SrCoO2.5 and the perovskite SrCoO3 engendered by ionic liquid gating. Here we show using in situ atomic force microscopy studies with photothermal excitation detection, that allows for high quality measurements in the viscous environment of the ionic liquid that the edges of atomically smooth terraces at the surface of SrCoO2.5 films are significantly modified by ionic liquid gating but that the terraces themselves remain smooth. The edges develop ridges that we show, using complementary X-ray photoemission spectroscopy studies, result from the adsorption of hydroxyl groups. Our findings exhibit a way of electrically controlled surface modifications in emergent ionitronic applications.

T he electrical control of the crystal or electronic phase of a material is a cornerstone of many of today's electronic devices. 1 Transistor devices that use ionic liquids (ILs) allow for the development of intense electric fields within an ultrathin electric double layer (EDL) that is formed at the interface between the IL and a thin film channel. 2−5 It has been shown that these fields can lead to ion migration into and out of the film that results in structural changes or phase transformations. 6−13 These changes have potential for applications in spintronic, 14,15 sensing, 16 bioinspired, 17 and neuromorphic 18 devices. Notably, the IL gating induced migration of ions can take place over very long distances up to microns, thus resulting in changes in physical properties of the entire bulk of a thin film. 19−21 However, the modification of the thin film surface itself has been little studied, 22,23 even though small changes in the state of the surface can strongly influence the electrochemical reactions that can take place in IL gating. 24,25 The existence of the viscous IL hinders the application of many surface sensitive characterization techniques. 26,27 There has been preliminary work investigating the ordering, dynamics, and interactions of ions in ILs at neutral and electrified solid/liquid interfaces using conventional atomic force microscopy (AFM). 28 −30 Here, using photothermal excitation 31,32 we employ in situ AFM to investigate the evolution of the surface of SrCoO x (SCO) thin films under IL gating. We find significant changes in the morphology of the edges of atomically smooth terraces at the surfaces of these films but the terraces themselves appear unchanged. With the help of X-ray photoemission spectroscopy (XPS), we show that subnanoscale ridge structures (RSs) that emerge at the terrace edges result from reaction with hydroxyl (OH − ) groups dissolved in the ionic liquid. Firstprinciples density functional calculations show that the terrace edges are more reactive than the terraces themselves, consistent with our experiments.

RESULTS AND DISCUSSION
IL Gate-Induced Terrace Edge Structure Modifications. In this study, 30 nm thick SrCoO 2.5 films were epitaxially grown by pulsed laser deposition on both (001) TiO 2 -terminated, undoped, and 0.5 wt % Nb-doped SrTiO 3 substrates. The films grow in a layer-by-layer growth mode, as shown by clear oscillations in reflection high-energy electron diffraction (RHEED) intensity (see Figure S1). The topographical evolution under IL gating was in situ mapped within an IL droplet which covered both the SrCoO 2.5 thin film and a lateral Au gate electrode. The AFM cantilever was immersed in the ionic liquid EMIM-TFSI (1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide) while the surface was scanned with a gate voltage (V G ) applied, as shown in Figure  1a. We refer to EMIM-TFSI as IL in the rest of the paper. In order to clearly identify the evolution of the surface structure without damaging it, an AC (tapping) mode was used. The in situ AFM imaging was captured using photothermal excitation, 32 which allows for high-resolution imaging comparable to that obtained by scanning without the IL (see Figure S2 for typical images that reveal the terraces in TiO 2terminated SrTiO 3 (100) in both air and IL).
Figure 1b−g illustrates typical high-resolution AFM topographical images scanned while V G was swept as follows: 0 → −1.8 → 0 → +1.8 → 0 V. Note that V G is applied for ∼5 min before each AFM scan at a fixed V G . Initially, the pristine SrCoO 2.5 surface shows clear (001) terraces (Figure 1b, V G = 0 V), consistent with a high quality epitaxial thin film (also see the RHEED results in Figure S1). The steps at the edges of the terraces are ∼4.0 Å high ( Figure S3), which is consistent with the pseudotetragonal structure of SrCoO 2.5 (a t = 3.905 Å and c t /4 = 3.936 Å). 33 By applying V G = −1.0 V for 5 min, the edges of each of the atomically flat terraces swell up, as shown in Figure 1c. The mean height of these ridged structures is h RS ∼ 1.3 Å (see Supporting Information for a definition of the mean height and method used to extract it). To quantify the change in h RS with V G , we summarize the evolution of h RS versus V G in Figure 1h. As V G is further decreased to −1.8 V, the RSs become more pronounced, as shown Figure 1d, and h RS is increased to ∼3.5 Å. Subsequently, when V G is switched to +0.5 V (Figure 1e), the RSs are maintained with an almost unchanged mean height of ∼3.5 Å, indicating that the RSs induced by IL gating are nonvolatile. Nevertheless, with further increases of V G to +1.0 and +1.8 V, h RS is reduced to ∼3.2 and ∼2.5 Å, respectively, as shown in Figure 1f,g. This confirms that the ridge structural changes are, at least in part, reversible. As distinct from the dramatic changes at the terrace edges, the flats portions of the terrace do not show any obvious response to the IL gating (the topographical profiles including RSs and terraces during a V G cycle are shown in Figure S3). The corresponding AFM phase images indicate that the chemical compositions of the terraces and edges undergo reversible and irreversible changes, respectively, during this gating cycle (see details in Supporting Information and Figures S4 and S5).
In the following, the conductivity and crystal structure of the SrCoO 2.5 thin films are investigated under IL gating. A thin film was fabricated into a transistor device in the form of Hallbars with lateral gate electrodes located in the vicinity of the channel. The channel is 400 μm long and 100 μm wide (ref 21). As V G is varied from 0 to −1.8 V, the channel resistance is reduced from 4.78 × 10 6 to 2.50 × 10 2 Ω (Figure 2a), and the initially insulating SrCoO 2.5 film becomes metallic. The resistance reverts back to 4.52 × 10 6 Ω when V G = +1.8 V is applied to the sample. By cycling V G according to 0 → −1.8 → Figure 2. IL gate-induced SCO phase transition. (a) Gate voltage-dependent resistance of a 30 nm thick SCO thin film at room temperature; the resistance was measured while cycling V G in the sequence: 0 → −1.8 → 0 → +1.8 → 0 V, as denoted by the blue arrows. Note that V G is incremented/decremented in steps of 0.015 V: the resistance is recorded at each step (this takes ∼2 s) after V G has been applied for 10 s (note that only every 10th data point is plotted). (b) X-ray diffraction patterns of an as-grown SrCoO 2.5 thin film (black) and the same film after IL gating at V G = −1.8 V (red); the Miller indices of the diffraction peaks are given in the figure (these are referenced to a pseudotetragonal unit cell 33 ). The X-ray diffraction (XRD) results that are shown in Figure 2b reveal that a significant structural change accompanies the IL gate-induced insulator−metal transition. The IL gating effect is nonvolatile so that the XRD could be carried out ex situ. The XRD of the pristine (as-grown) SrCoO 2.5 thin film clearly exhibits a characteristic quadrupled c-axis lattice constant expected for the brownmillerite phase which is the result of an ordered oxygen vacancy structure. Diffraction peaks corresponding to (002), (006), and (0010) of the brownmillerite phase SrCoO 2.5 can clearly be identified. A constant gate voltage V G = −1.8 V was then applied for 30 min through the IL to the device. The IL was then removed by washing the device in acetone followed by isopropyl alcohol (IPA). The (002), (006), and (0010) SrCoO 2.5 diffraction peaks disappear, accompanied by shifts of the (004) and (008) diffraction peaks to higher 2-theta values. These latter two peaks correspond to the (001) and (002) peaks of the perovskite structure SrCoO 3 , respectively. Thus, the XRD results show that IL gating at V G = −1.8 V induces a structural phase transition from SrCoO 2.5 to SrCoO 3 , which is consistent with the insulator to metal phase transition observed via the transport studies discussed above. Although the gating processes for XRD and transport measurements are different, a constant V G = −1.8 V is applied for 30 min for XRD while V G is gradually swept from 0 to −1.8 V for the transport measurements, both processes induce the full phase transition from SrCoO 2.5 to SrCoO 3 ( Figure S6). From the XRD peak positions, we find that the out-of-plane lattice parameter (within the pseudotetragonal structure) of the initial SrCoO 2.5 layer is reduced from c t /4 = 3.94 Å to c t = 3.80 Å at V G = −1.8 V. Scanning transmission electron microscopy (STEM) images and enlarged XRD patterns are given in Figure S7. However, such a shrinkage in the bulk lattice perpendicular to the surface of the film is not reflected in the AFM images and cannot account for the very large changes in the RSs at the terrace edges (also see Figure S8 for the STEM of ridge structure).
Interplay Between Hydroxyl and SrCoO x . The phase transition between SrCoO 2.5 and SrCoO 3 reflected in the XRD and transport data corresponds to the bulk of the thin film. Hence, to understand the mechanism for the surface evolution during IL gating, we turn to a surface sensitive technique, namely X-ray photoelectron spectroscopy (XPS), to characterize the chemical bonding at the surface averaged over many terraces and steps (the X-ray beam is 100 μm wide). XPS was carried out on the gated samples (V G = −1.5 V, the full phase transition from SrCoO 2.5 to SrCoO 3 is obtained as shown in Figure S6) immediately after the RSs emerge (monitored by in situ AFM) and after the IL was removed by ultrasonic cleaning of the sample in acetone followed by IPA. Figure 3a depicts the measured O-1s core-level spectrum and a corresponding multiple-peak fit to the data for the pristine SrCoO 2.5 sample. The fits to the XPS data use a CoO peak that is centered at 530.3 eV and a CoOH peak that is centered within the range 531.2 to 531.9 eV (these values are taken from refs 34 and 35). The main peak for the pristine SrCoO 2.5 is located at ∼530.4 eV, indicating that CoO bonding is dominant. In the sample that was gated at V G = −1.5 V, in addition to the main peak at ∼530.4 eV, an additional shoulder peak at the higher binding energy of ∼532.0 eV is found, which is evidence for Note that for all studies, extreme care is taken to eliminate any water dissolved in the IL by drying the IL for at least half of a day in vacuum (better than 10 −6 mbar) at 380 K. For the in situ AFM measurements, the IL and the sample are contained within a closed cell that is isolated from the surrounding air by a Viton O-ring seal, but some small amount of H 2 O will likely be absorbed by the IL during the experiment and when the IL is exposed briefly to the air during the application of the IL onto the sample before sealing the cell. Thus, there is always some small amount of residual water in the IL even after the baking procedure mentioned above (for example, ref  Figure S9). Figure 3g summarizes the mean height h RS and fwhm (full width at half-maximum) of the RSs induced by IL gating with distinct H 2 O doping levels (the mean fwhm definition and calculation methods are described in Supporting Information). After gating with V G = −1.5 V (20 min), the h RS in all three cases has a similar value of ∼4.0 Å. However, the mean fwhm of the RSs is increased from ∼20 to ∼94 nm when 0.1 vol % H 2 O is added to the IL. An even greater enhancement to ∼123 nm is found for the case of 1.0 vol % H 2 O. The simultaneous increase in both the mean fwhm of the RSs as well as R when using an IL with added H 2 O doping supports the hypothesis that the RSs are related to an interaction between OH − ions and the surface of SCO under negative V G . Moreover, R = 0.21 is found after a gating procedure in which V G is varied from 0 → −1.5 → 0 → +1.8 → 0 V, which implies that the resulting changes are only partially reversible ( Figure S10).
Electrical Property of the Ridge Structures. We now turn to investigate the electrical properties of the RS and terraces. To profile the conductivity of the RSs, we utilize conductive AFM (cAFM) mapping of the leakage current through a second set of similar SrCoO 2.5 films that were prepared, however, on conducting Nb-doped SrTiO 3 substrates. After the sample is gated at V G = −1.5 V for 20 min, the IL is removed by rinsing in acetone and IPA, and then cAFM is performed to map the leakage current through the terraces and the RSs. As shown in Figure 4a, one can clearly see the RSs at the terrace edges, which are almost the same as those measured in the IL (see Figure S11). Compared with homogeneous low leakage current of pristine, insulating SrCoO 2.5 ( Figure S12), the corresponding leakage current image (Figure 4b) illustrates that the RSs exhibit a lower ACS Nano www.acsnano.org Article leakage current (green) than that of the terraces (blue). This can more readily be seen in Figure 4c where line profiles of the topography and leakage current (white solid lines in Figure  4a,b, respectively) are compared. In addition, V G -dependent transport measurements were carried out on Hall bar devices that were patterned from samples prepared with several different thicknesses of SCO. In Figure 4d, the insulator-tometal-to-insulator transition of a 30 nm thick sample, as V G is varied according to 0 → −1.8 → 0 → +1.8 → 0 V, is almost fully reversible. The data shown here are the same data as those in Figure 2a but are plotted on a linear scale to magnify the resistance divergence after one V G cycle. The resistance of a thinner 15 nm thick sample after a similar gating cycle shows a much higher resistance as compared with the initial state. Data for 10 and 5 nm thick samples are also given in Figure 4d. It is clear that as the film thickness is decreased, the contribution of the insulating RSs becomes larger, which results in larger irreversible resistance changes. It is noteworthy that cAFM measurements can clearly distinguish the current variation (∼100 pA) between RS and terraces in the direction perpendicular to the film, suggesting that the increased irreversible behavior in transport measurements is mainly due to the more insulating RSs. H 2 O vol % dependent transport measurements were also performed ( Figure S13); these results show more pronounced irreversible behavior as the water level was increased.
Next we carry out first-principle density functional calculations to explore the role of the hydroxyl groups in forming the RSs. We established a slab model for calculating the binding energy (ΔE OH ) of hydroxyl adsorbed at different sites located at various distances from a terrace edge, respectively (Top 1−3 in Figure 5a−c). Here, SrCoO 3 is used in the calculations since hydroxyl is absorbed at negative gating voltages at which the phase transition from SrCoO 2.5 to SrCoO 3 has already taken place, as demonstrated in Figure 2. As SrCoO 2.5 thin films were grown in a layer-by-layer mode on the TiO 2 -terminated SrTiO 3 substrate (see RHEED oscillations in Figure S1), it is expected that the surface of the thin film should be CoO 2 -terminated. In the slab model used in the calculations, the terrace was constructed as an 8 × 1 (x × y) supercell with three SrCoO 3 layers in the z-direction, while the model for the terrace edge includes an additional layer in which in the x-direction is only 6 cells wide. The two lower SrCoO 3 layers were first fixed according to that of the calculated bulk structure and then the slab models were relaxed to obtain the ground state. Thereafter, neutral OH was added to the three positions Top 1, Top 2, and Top 3 corresponding to the three unique sites on the surface of the 6-cell wide terrace in the slab models (see Figure 5a−c). After relaxation of the final structures, the OH binding energies at these sites, ΔE OH , are obtained. The calculated ΔE OH (see Methods) shows that OH clearly prefers to be adsorbed at the terrace edge; the binding energy at the Top 1 site at the terrace edge is −3.32 eV while ΔE OH for the interior sites on the same terrace is significantly lower (−2.13 and −2.10 eV, respectively (see Figure 5a−c)). We also calculate E OH for OH − using a similar slab model by electron doping and find that the results show the same trend as for neutral OH (see Figure S14).
The local electronic density of states (DOS) for Co and OH, located at the edge (Top 1 model), are shown in Figure 5d (neutral OH) and Figure S14d (charged OH − ). The OH is clearly insulating, meanwhile the DOS near the Fermi energy (E F ) integrated over the 4 Co atoms underneath the hydroxyl (shown by the red dashed line in Figure 5a) is lower than that of the center 4 Co atoms away from the OH (shown by the blue dashed line in Figure 5a). Therefore, the quasi-onedimensional columnar structure formed from 4 Co atoms and the absorbed OH is less conducting, which is consistent with our experimental findings.

CONCLUSIONS
In summary, utilizing in situ AFM we find that IL gating of SrCoO x thin films results in significant modifications of the surface. While the morphology of the atomically flat terraces is not much changed, the edges of the terraces develop ridges that we find are due to the reaction of hydroxyl groups dissolved in the ionic liquid. These ridges are much less conducting than the terraces. Our findings imply the possibility of manipulating surfaces using ionic liquid gating when the surface has distinct morphological structures or creating nanoscale structures/interfaces with disparate conductivity, for example, with terrace edges whose spacing and orientations can be controlled by the miscut angle 36 or with hinges, formed, for example, from surface reconstructions that minimize the surface energy (e.g., MgO (111) surface 37−39 ). The observed terrace edge modification also shows the potential, for example, of nanostructured electrochemical interfaces in applications ranging from heterogeneous catalysis to biomedical and environmental sensing. 40,41 METHODS SrCoO 2.5 Thin Film Sample Preparation. SrCoO 2.5 thin films were grown on STO and Nb-STO (001) substrates at 750°C in an oxygen pressure of 5 × 10 −4 mbar, by pulsed laser deposition (PLD). 21 Before the deposition, all of the substrates (Crystec GmbH) were treated with buffered hydrofluoric acid [BOE 7:1 (HF:/NH 4 F = 12.5%:87.5%)] for 1 min at room temperature and then annealed at 950°C in oxygen atmosphere for 3 h to achieve TiO 2 -terminated atomic terraces. The thickness of the SrCoO 2.5 films for all of the in situ AFM, cAFM, STEM, and XPS measurements was ∼30 nm while SrCoO 2.5 samples with thicknesses of 5, 10, and 15 nm were used for the transport measurements. After deposition, the SrCoO 2.5 films were cooled down to room temperature in the same oxygen pressure as that used for the growth.
In Situ AFM Characterization of IL Gating-Induced Ridge Structures. The in situ AFM measurements were carried out in a Cypher (Asylum Research) atomic force microscope. An integrated blueDrive (Asylum Research) laser was used to actuate the oscillation of the cantilever (AC-55 TS, Olympus) in the IL. During AFM scanning, the cantilever, gate electrode plate (Au), and SCO sample were all covered by the IL (Figure 1a), and the gate voltage was applied by an external source meter (Keithley 2400). The ionic liquid 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM-TFSI), was used for all IL gating experiments. The IL and the devices were separately baked at 380 K in high vacuum (10 −7 mbar) for at least 12 h before the gating experiments were carried out. Except for the H 2 O-vol % dependent XPS measurements, dried IL was otherwise used for all measurements.
X-ray Photoelectron Spectroscopy. The X-ray photoelectron spectroscopy measurements were carried out in a Thermo Fischer Scientific K-Alpha fitted with MAGCIS and a dual mode argon ion source. Before the measurements, the gated samples were rinsed by IPA, and then immediately transferred to the XPS chamber. To remove the residual IL and other contaminants, the surface was gently treated by argon cluster ions (∼80 molecules of Argon) that do not penetrate the oxide surface and allow chemical state information to be preserved after such a cleaning process. 42 Al Kα radiation with the Xray source operated at 14 kV with a spot size of 100 μm was used.

ACS Nano
www.acsnano.org Article Conductive AFM. Before the cAFM experiments, the SCO sample grown on a Nb-doped SrTiO 3 substrate was gated (V G = −1.5 V for 20 min), and the development of the RSs was monitored by in situ AFM (see Figure S11). The Nb-doped SrTiO 3 substrate was sealed by paraffin to avoid current leakage. 21 After gating, the sample was rinsed by acetone and IPA to remove the residual IL and paraffin. In the cAFM measurements, a constant voltage of −5.0 V was applied to the sample holder (the tip is grounded) and the current flowing through the sample (perpendicular to the surface) is mapped. A 500 MΩ resistor was connected in series with the sample to prevent saturation of the amplifier during the measurements, and a low-noise trans-impedance amplifier (ORCA module, Asylum Research) was used that allowed for a current measurement as low as ∼1.5 pA. A conductive diamond probe (AD-2.8-AS, Adama Innovations) was used for cAFM, and the radius of the tip was ∼10 ± 5 nm.
Transport Measurements. The devices for transport measurements were prepared by photolithography and wet etching in the form of Hall-bars with lateral gate electrodes located in the vicinity of the channel. The channel was 400 μm long and 100 μm wide. Electrical contacts to the edges of the channel were formed from Au (60 nm)/ Cr (10 nm) that was deposited by thermal evaporation.
First-Principles Calculations. All calculations were performed by a density function theory (DFT) method as implemented in the Vienna ab initio Simulation Package (VASP). 43 The Perdew−Burke− Ernzerhof (PBE) pseudopotentials with the generalized gradient approximation (GGA) were used in the calculations. 44 All structures contain a vacuum space larger than 17 Å between successive slabs and were relaxed until the atomic forces on the atoms are less than 0.02 eV Å −1 . Spin polarization was included in all calculations. The Co atoms were set to be ferromagnetic with a Hubbard U set to be 2.5 Ev. 45 The binding energy was calculated from OH where E(slab + OH), E(slab), and E(OH) are the energies of the composite system, the clean slab, and the uncoordinated adsorbate, respectively.