Engineering the stoichiometry of a TiO₂-rich SrTiO₃(001) surface

The preparation of SrTiO₃(001) surfaces with atomic precision^1,2 has laid the foundation for the atomic-scale engineering of oxide heterostructures.^3–6 Owing to lattice, orbital, charge, and spin degrees of freedom, such systems exhibit fascinating phenomena including superconductivity,^7,8 monolayer magnetism,⁹ and tunable topological textures.^10,11

The stabilization of such phenomena depends crucially on the structural quality of the grown film or heterostructure. Prerequisite to achieving high-quality heterostructures is having a well-defined growth substrate. The ideal substrate for heteroepitaxy is atomically flat, has a minor lattice mismatch to the deposited material, does not react chemically with the deposited material, and has a singular chemical composition. SrTiO₃(001) substrates suitable for heterostructure growth are typically prepared by etching with an acidic solution and subsequent annealing at high temperatures.^1,2 The resulting surface has long been assumed to be a truncation of the bulk crystal on the TiO₂ planes. This assumption is based on three observations: the acid etchant used in the preparation process selectively etches SrO, ion spectroscopy shows an excess of Ti on the surface, and the surface steps have a height of one full unit cell.¹ 

However, later studies found that this simply truncated surface is not thermodynamically stable at high temperatures. Above $∼600 °C$⁠, the surface reconstructs in one (or several) of the possible superstructures.^12–24 All these superstructures comprise a TiO₂ adlayer on top of the final bulk TiO₂ layer, forming a TiO₂ double-layer (DL) surface. The coverage of this adlayer may differ, depending on the reconstruction.^21,25 Notably, the adlayer fully covers the surface for all reconstructions with a periodicity of twice the bulk unit cell.

As the annealing of SrTiO₃(001) substrates and the growth of most films on them takes place at such temperatures, the DL is likely to be present on SrTiO₃(001) surfaces used for film growth and, thus, affects the initial stage of film growth. This effect would be strongest on the properties of interfaces and films with a thickness of only a few unit cells. Possibly affected phenomena include those involving critical thicknesses,^26,27 charge transfer,^28–31 and orbital coupling across interfaces.⁵ Despite this wide variety of possibly affected systems, only a few recent articles report studies on the effects of the TiO₂ DL on epitaxial growth.^32–36 Similar effects of surface reconstructions are well-known for semiconductor heterostructures. For these systems, controlling the reconstruction during growth is paramount for achieving high-quality surfaces.^37–39 

In this Letter, we report a straightforward experiment designed to study the initial stage of epitaxial growth on SrTiO₃(001). To gain insight into the formation of the first SrTiO₃ unit cell during growth, we deposit less than one monolayer of SrO by pulsed laser deposition (PLD) and study the resulting surface morphology. We use typical experimental conditions for the layer-by-layer growth of complex-oxide films. After depositing SrO, we characterize the samples in situ by reflection high-energy electron diffraction (RHEED) and ex situ by atomic force microscopy (AFM).

The results presented in the following focus on several series of samples prepared under the same experimental conditions. To rule out any effect of these conditions, we have performed a variety of control experiments, which are discussed in the supplementary material. These include experiments using substrates prepared via the commonly used buffered-hydrofluoric acid (BHF) method,² for which we arrive at the same conclusion. We note that deposition at lower oxygen pressure yielded a different surface morphology, presumably caused by an incomplete oxidation of the SrO plume.⁴⁰ 

For our “standard” process, we used $5×5×0.3 mm3$ SrTiO₃(001) substrates provided by Shinkosha Corp. After being degreased in acetone, isopropanol, and DI water, the substrates were loaded into a PLD chamber. The substrates were then thermally prepared under an oxygen process pressure of $7.5×10−2 hPa$ by annealing with a newly developed CO₂-laser heating system at 1300 °C for 200 s.⁴¹ We measured the temperature on the back of the substrate T_sub using an optical pyrometer that was calibrated with respect to the melting point of a single-crystalline Al₂O₃ substrate.⁴² Unless stated otherwise, the ramp rate of T_sub was $2 K/s$⁠.

For the SrO deposition, T_sub was reduced to 853 °C after thermal preparation. The deposition was done by firing a small number of pulses (pls) from an excimer laser (Coherent, $λ=248 nm$⁠) at a sintered SrO target. The laser fluence was 2 Jcm⁻², the spot size was 2.68 mm², and the repetition frequency was 1 Hz. After deposition, the samples were kept at the deposition temperature for about 5 min to record RHEED images. Next, the heating laser was switched off abruptly to cool the samples to room temperature with an initial rate of several hundred °C/s. Control studies showed that the delay between deposition and cooling did not affect the morphology nor the reconstruction of the surface.

Figure 1(a) shows an AFM image of a SrTiO₃ surface on which 6 pls of SrO have been deposited. We observe a well-defined step-and-terrace structure with small islands on top of the terraces. These islands avoid the right-hand side of each terrace and have diameters between 20 and 60 nm. The height profiles in Fig. 1(c) show that the height of each island is 3.9 Å, the same as the height difference between terraces. There are no islands with a different height, as shown by the high-contrast image in Fig. 1(b) and the histogram in Fig. 1(d).

The island height is twice the expected height for single-layer SrO islands on an unreconstructed surface. This finding suggests that SrO interacts with a reconstructed SrTiO₃ surface to form islands with a height of one full SrTiO₃ unit cell.

To investigate the formation process of these islands, we prepared a series of samples with various amounts of deposited SrO. Figure 2 shows the surface morphology of such a series of samples. The sample labeled “0 pls” served as a control. It underwent the deposition process described above, but without SrO deposition. The terraces on its surface are about 200 nm wide and are separated by one-unit-cell-high steps.

The SrO deposition process modifies this surface in three stages. The first stage can be observed in Figs. 2(b) and 2(c), where the deposited SrO forms spatially separated islands with a height of one unit cell. Island diameters increase with SrO deposition with a roughly constant nucleation distance, i.e., the average distance between the centers of adjacent islands, of about 60 nm. On each terrace, we observe no islands in a region of approximately the same width as the nucleation distance. This observation suggests that islands forming at positions less than the nucleation distance away from a step edge are incorporated into the then advancing step. AFM phase contrast measurements, discussed in the supplementary material, support this scenario.⁴³ 

The island diameter does not increase further from 9 to 12 pls. Instead, the density of islands increases, the nucleation distance decreases, and islands start to merge. This second stage culminates in the surface morphology shown in Fig. 2(f), where we see that the islands have broken up, forming networks of narrow ridges and trenches. Whereas the height of these ridges is again 3.9 Å, their width is only about 10 nm. The island-free region at the edge of each terrace has the same width. These observations imply that the increased SrO coverage reduces the nucleation distance to such an extent that it becomes energetically beneficial for the surface to reorganize itself into the observed ridge-and-terrace structure.

In the third stage, the trenches are filled by the additionally deposited SrO. A comparison of Figs. 2(f) and 2(g) shows that the nucleation distance no longer changes, but the edges of the ridges have become softer.

Continued deposition of SrO results in the growth of a SrO film on the modified SrTiO₃ surface. Interestingly, Fig. 2(h) shows that the surface is then strongly disordered, with the original step-and-terrace only barely visible as diagonal lines. This disorder could be induced during growth, for example, due to the structural mismatch between SrO and SrTiO₃(001), or ex situ by humidity-induced degradation of SrO.

This peculiar evolution of the surface morphology as a function of the amount of deposited SrO is also reflected in the RHEED patterns obtained in situ at room temperature after deposition. Figure 3(b) shows the RHEED pattern obtained for the control sample. We observe slightly elongated specular and (10) diffraction spots. Vertical streaks highlighted by the yellow arrows are visible in the gaps between those spots. These half-order streaks imply the presence of a surface reconstruction with a periodicity of twice the bulk lattice constant.^13,44 In the following, we use the term $(2×n)$ to refer to all possible such reconstructions collectively.

Our measurements reveal that the deposition of SrO has two main effects on the RHEED patterns. First, the higher-order streaks fade with increasing number of laser pulses, such that the $(2×n)$ reconstruction disappears after 12 pls. Second, up to 12 pls, the main spots elongate in the propagation direction of the beam. Between 12 and 18 pls, the spots shorten again. Finally, the deposition of 40 pls of SrO results in a pattern with significant diffuse scattering. This behavior indicates the growth of a dominantly amorphous film and, therefore, supports the scenario outlined above that the SrO film is already strongly disordered during growth. Simultaneously, we observe faintly reemerging half-order streaks that are consistent with a $(2×n)$ reconstruction of the SrO surface.

For the 40-pls sample, we recorded the intensity of the specular RHEED spot as a function of SrO deposition time, see Fig. 3(a). We observe a small increase in intensity during the first two pulses, followed by two full oscillations. The signal fades after these two oscillations, in agreement with the amorphous growth implied by the diffuse image in Fig. 3(g). We note that epitaxial growth of SrO on SrTiO₃(001) is possible under growth conditions different to the ones we used here.^45,46 In those studies, the initial oscillations were irregular in a similar manner as shown in Fig. 3(a).

To understand this irregularity, we compare the RHEED results to those obtained by AFM. Starting from 0 pls, the surface is atomically flat and is fully reconstructed into one (or several) of the possible $(2×n)$ configurations. Depositing one or two pulses of SrO transfers spectral weight to the specular spot. We attribute this transfer to the mitigation of one of the surface reconstructions, thus enhancing long-range order. Between 2 and 12 pls, the signal is dominated by surface roughening and smoothing through island formation. After one cycle, the surface morphology transitions to a structure with narrow ridges and trenches. Next, the surface undergoes another cycle of roughening and smoothing until a SrO film starts to grow. In the current experimental conditions, this film is dominantly amorphous, and no further oscillations are observed.

To explain our experimental results, we propose a simple model for the initial phase of SrO growth on SrTiO₃(001). This model is illustrated in Fig. 4 and contains two key elements. First, the preparation process—thermal or chemical—of SrTiO₃(001) always results in a reconstructed surface with a TiO₂ DL. The observed $(2×n)$ symmetry of the reconstruction implies that the TiO₂ adlayer fully covers the surface.^21,25 Second, the deposition of SrO modifies this surface by simultaneously intercalating between the TiO₂ DL and growing on top of it. This intercalation is mediated by a continuous layer rearrangement, as is well established for $Srn+1TinO3n+1$ Ruddlesden–Popper compounds grown by molecular beam epitaxy (MBE).^47,48

If SrO intercalated only in the TiO₂ DL, the islands would have a height of 2 Å. Hence, SrO must also grow on top of the final TiO₂ layer to form a stable surface. This result reveals that the unreconstructed TiO₂ surface is not at all stable at high temperatures, in agreement with earlier results.¹⁵ 

The proposed model agrees well with the literature. It is clearly consistent with all direct observations of a TiO₂-rich surface reconstruction on the surface of SrTiO₃(001).^12–24 It is also consistent with the original observations by Kawasaki et al. listed in the introduction, which led to the assumption that the surface is truncated on the TiO₂ planes.¹ Previous reports on the deposition of SrO on SrTiO₃(001) show similar RHEED patterns^46,49 and/or oscillation signals.^45,46 Importantly, our model explains why the initial SrO DL is always missing during the growth of $Srn+1TinO3n+1$ Ruddlesden–Popper compounds.⁴⁷ 

The continuous layer rearrangement driving the SrO intercalation probably requires T_sub to exceed a minimum value. In the original reports on this subject,^47,48 substrate temperatures of $≥750$ °C were used. The deposition series we have performed to study the influence of T_sub by varying it show that our model is valid down to at least 700 °C. Below this temperature, the islands became too small to be observed by AFM, which would also have been the case if they had been 2 Å in height. The deposition of SrO on SrTiO₃(001) with $Tsub<700 °C$ has been studied by different groups using scanning tunneling microscopy (STM), the results of which show a more disordered surface with varying island heights.^46,50 Others used such low deposition temperatures in combination with pulsed-laser interval deposition to avoid the growth of multi-level islands.^49,51 We argue that T_sub was indeed insufficient in those studies to drive the full layer rearrangement.

Our work provides evidence that surface reconstructions influence the epitaxial growth of oxide films. This understanding is supported by recent studies of SrTiO₃(001) surfaces.^32–36 We expect that the influence of the surface reconstructions is greatest on ultrathin films and interfaces. For example, the TiO₂ DL was experimentally found to be key for stabilizing high-temperature superconductivity in the FeSe/SrTiO₃ system.⁵² To perovskite films, excess TiO₂ can be added in three ways: (i) by migrating to the surface during film growth;^33,35 (ii) by forming an atomically sharp ATiO₃ interface layer (A being the A-site cation of the grown film); and (iii) by intermixing with the first BO₂ layers of the grown film.²⁹ All these processes would strongly influence the physical properties of ultrathin films, for example, SrRuO₃⁹ or SrIrO₃.⁵³ However, we expect the strongest effect to be exerted on charge-transfer systems such as the LaAlO₃/SrTiO₃ interface. The physical properties of such systems depend crucially on the formal charge of each atomic layer,^28–30 which is strongly altered in any of these three scenarios.

It is, thus, relevant to explore ways to mitigate excess TiO₂. Furthermore, it seems worthwhile to tune this excess to study its effects on the properties of films and interfaces. We suggest that continuous layer rearrangement^47,48 is the crucial process to control in this exploration. The Ti and O partial pressures as well as the substrate temperature are the key parameters in that process. In this respect, we note that $(2×n)$ surface reconstructions have been observed on SrTiO₃(001) for a wide range of oxygen pressure, for example, UHV,^{13–15,17,19,22} $10−7–10−6$⁠,^23,24 $10−3$⁠,²⁰ and $>10−2 hPa$ (this work and Refs. 16 and 18). Finally, it is important to determine whether related similar reconstructions occur on different oxide surfaces and whether they affect the growth in a similar manner. Our previous RHEED studies show that thermally prepared NdGaO₃(001), DyScO₃(110), and TbScO₃(110) also have $(2×n)$ surface reconstructions.⁴² 

Compared to the field of semiconductor heterostructures,^37–39 surface reconstructions in oxide heterostructures are only scarcely investigated in terms of their effect on epitaxial growth.^35,36,44 Such investigation will make it possible to grow multilayers of even better quality, thus providing the basis for observing new emergent phenomena in this fascinating material class.

See the supplementary material for the results obtained on chemically prepared surfaces, a discussion of the phase contrast observed in AFM, RHEED images obtained along the [110] direction, and the effects on the surface morphology of changing the preparation and deposition temperatures and pressures.

We thank Hans Boschker, Bruce A. Davidson, and Gertjan Koster for valuable discussions; Gennady Logvenov for a careful reading of the manuscript; Lilli-Marie Pavka for textual editing; and Hans Boschker, Ingo Hagel, Prosper Ngabonziza, Sarah Parks, Sabine Seiffert, and Wolfgang Winter for technical support.