Stoichiometry-driven switching between surface reconstructions on SrTiO$_3$(001)

Controlling the surface structure on the atomic scale is a major difficulty for most transition metal oxides; this is especially true for the ternary perovskites. The influence of surface stoichiometry on the atomic structure of the SrTiO$_3$(001) surface was examined with scanning tunneling microscopy, low-energy electron diffraction, low-energy He$^+$ ion scattering (LEIS), and X-ray photoelectron spectroscopy (XPS). Vapor deposition of 0.8 monolayer (ML) strontium and 0.3 ML titanium, with subsequent annealing to 850 {\deg}C in 4 $\times$ 10$^{-6}$ mbar O$_2$, reversibly switches the surface between c(4 $\times$ 2) and (2 $\times$ 2) reconstructions, respectively. The combination of LEIS and XPS shows a different stoichiometry that is confined to the top layer. Geometric models for these reconstructions need to take into account these different surface compositions.


Introduction
Strontium titanate (SrTiO 3 , STO) belongs to the class of cubic perovskite oxides and it is well known for its useful bulk and surface properties, e.g., a high dielectric constant at low temperatures [1], photocatalytic water-splitting [2], lattice matching for growth of high-T c superconductors [3], and the formation of two-dimensional electron gases at its surfaces [4,5], and at interfaces with other perovskites [6]. For most applications, a control of the surface structure at the atomic scale is of central importance.
The most important surface, STO(001), exhibits a large variety of reconstructions. Typical preparation procedures consist of sputtering with Ar + ions of different energies (~1 keV) and fluences, and annealing to high temperatures (~1000 °C) in various oxygen pressures (atmospheric to <10 -10 mbar O 2 ). Often the outcome also depends on the sample preparation history.

Experimental Methods
Nb-doped (0.5 wt %) SrTiO 3 (001) single crystals were purchased from MaTecK Company, Germany. After ultrasonic cleaning in acetone, the samples were introduced into a two-chamber UHV system. One chamber is equipped with evaporators (Sr, Ti), a sputter gun, a home-built quartz crystal microbalance (QCM) and a leak valve for admitting O 2 into the chamber. Its base pressure was below 10 -9 mbar. The second chamber, with a base pressure below 10 -10 mbar, was used for analysis with LEED (SpectaLEED, Omicron), STM (Aarhus 150, SPECS), XPS (non-monochromatized dualanode), and a scanning ion gun for LEIS. XPS spectra were acquired using Mg Kα radiation. For LEIS, 1 keV He + ions were used. Backscattered ions (scattering angle ϑ=137°) and photoelectrons (emission normal to sample surface) were detected with a SPECS Phoibos 100 hemispherical analyzer with 5-channel detection. The XPS peaks were fitted after subtracting a Shirley background and the ISS peak intensity was summed up over the peak area after subtraction of a linear background.
After loading the samples into the system, they were sputtered with 1 keV Ar + ions (~8×10 13 Ar + /cm 2 s) for typically 10 minutes and annealed at ~850 °C in 10 -6 mbar O 2 for 40 minutes. Titanium was evaporated from an electron beam evaporator (Omicron) and strontium was evaporated using a low-temperature effusion cell (CreaTec). The flux was monitored by a QCM. The temperature was measured with an optical pyrometer using an emissivity of 1. While this measurement method is not very accurate for oxides, the structures presented here are stable within a wide range of temperatures (700-900 °C), thus ambiguities in temperature measurement should not be a major a problem.
All samples were treated with the following sequence. In order to get a clean, well-defined surface structure, samples introduced to the UHV system were sputtered (at least once) and then annealed. A c(4×2) structure (determined by LEED or STM) was observed after this procedure. Sr was evaporated onto this surface and the sample was annealed after deposition. Sr was deposited until a change of surface reconstruction to the (2×2) structure was observed

Results
A two-domain c(4×2) reconstructed surface was obtained through multiple sputtering and annealing cycles; no other superstructures were detected by STM and LEED. Figure 1a) shows a large-scale STM image of this surface.
The inset shows the associated LEED pattern. This structure was identified as the two-domain c(4×2) reconstruction using the software LEEDpat30 [24]. The terraces are ~30 nm wide and the step-height is equivalent to one unit cell of STO (3.9 Å). An atomically resolved image of the two-domain c(4×2) structure and its corresponding fast Fourier transformation is shown in Figure 1    after subtracting a linear background) spectra.

Discussion
Reconstructions of STO(001) surfaces have been proposed to be formed by ordered oxygen vacancies (on vacuum-annealed samples) [11,27], Sr adatoms [16,19] or a double-layer TiO 2 structure [28][29][30]. For the reconstructions investigated here, mainly two structure proposals are found in the literature. Based on STM measurements and first-principles calculations, Kubo and Nozoye suggested a model consisting of ordered Sr adatoms [16].
Supported by a combined STM and density functional theory (DFT) study, Marks and coworkers suggested a double-layer TiO 2 structure forming both, the c(4×2) [29] and the (2×2) structure [30]. A striking feature of the two surfaces is the different appearance of the step edges. As already discussed in references [12,17,25] [34]). By varying the chemical potential of Ti or Sr (deposition of Ti or Sr), the surface will undergo a transition to a different geometric structure and therefore lower its free energy.

Conclusion
In conclusion, the present study shows that, on the STO (001)  inherently similar. These models incorporate the same building block with an equal density for the two structures. This is in contrast to the results presented here. Assuming the structure for the c(4×2) reconstruction proposed in ref. [29] is correct, at least a new model for (2×2) has to be found.