Dichotomy of the photo-induced 2-dimensional electron gas on SrTiO₃ surface terminations

Significance

Titanates are a group of oxide materials which are heavily studied due to their wide array of interesting physical properties. Here, we study the origin of one particular property, the 2-dimensional electron gas (2DEG), by growing SrTiO₃ and other titanate films with different surface terminations, using atomic layer-by-layer growth and probing their electronic structures. Surprisingly, we only observe the 2DEG on the A-site termination and not on the B-site termination. We use this dichotomy to help understand the origin of the 2DEG and suggest how this same methodology could be applied to other oxide systems.

Abstract

Oxide materials are important candidates for the next generation of electronics due to a wide array of desired properties, which they can exhibit alone or when combined with other materials. While SrTiO₃ (STO) is often considered a prototypical oxide, it, too, hosts a wide array of unusual properties, including a 2-dimensional electron gas (2DEG), which can form at the surface when exposed to ultraviolet (UV) light. Using layer-by-layer growth of high-quality STO films, we show that the 2DEG only forms with the SrO termination and not with the TiO₂ termination, contrary to expectation. This dichotomy of the observed angle-resolved photoemission spectroscopy (ARPES) spectra is similarly seen in BaTiO₃ (BTO), in which the 2DEG is only observed for BaO-terminated films. These results will allow for a deeper understanding and better control of the electronic structure of titanate films, substrates, and heterostructures.

There are few material systems which exhibit the wide range of relevant physical phenomena as ${\mathrm{S}\mathrm{r}\mathrm{T}\mathrm{i}\mathrm{O}}_3$ (STO). It is well known for having a high dielectric constant (1) and undergoing a superconducting transition at 0.3 K (2). When combined with other materials, either as a layered compound or as a thin film substrate, a much wider array of phenomena are observed. For example, when FeSe, an iron-based superconductor, is grown on STO, its superconducting $T_c$ increases from $8\hspace{0.17em}\text{K}$ to $\ge 60\hspace{0.17em}\text{K}$ (3, 4). In addition, at the interface of insulating STO and ${\mathrm{L}\mathrm{a}\mathrm{A}\mathrm{l}\mathrm{O}}_3$, a 2-dimensional electron gas (2DEG) emerges, along with superconductivity and ferromagnetism (5–8). A 2DEG can also be generated at the surface of STO alone by exposure to ultraviolet (UV) light, like that from a synchrotron lightsource (9–12). The UV light is believed to create oxygen vacancies through a double Auger process (12). The remaining electrons create a Ti³⁺/Ti⁴⁺ mixed-valence state and populate the Ti ${\mathrm{t}}_{2g}$ bands (12). Existence of subbands due to quantum confinement confirms that the excess electrons are trapped at the surface and contribute to the itinerant carrier density and the 2DEG (10). The quick refilling of the oxygen vacancies with low doses of oxygen suggests that the oxygen vacancies are also localized at the surface (12, 13).

The 2DEG on the surface of STO has been heavily studied by using angle-resolved photoemission spectroscopy (ARPES) (9–12, 14, 15). Measurements have been carried out on the (001), (110) (16), and (111) (17, 18) faces of STO by using either fractured single-crystal samples or commercially available substrate wafers. The fractured single-crystal samples likely have a mixed SrO/${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination (19), while the wafers are etched to create a ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination (20). Due in part to the observation of the 2DEG on as-received commercial STO substrate wafers, the 2DEG is generally associated with the ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ terminated surface. Under a wide variety of different preparation conditions, the 2DEG is consistently observed with very similar band structure (11).

More generally, perovskite oxides are typically divided into subunit cell layers that play different roles, like an active layer that drives the physical phenomena of interest sandwiched between passive doping layers. Cuprate superconductors are a classical example of this type of system and are divided into active superconducting ${\mathrm{C}\mathrm{u}\mathrm{O}}_2$ layers surrounded by passive charge reservoirs. In a recent experiment, Yan-Feng Lv et al. (21) observed dramatically different electronic structure on ${\mathrm{B}\mathrm{i}}_2{\mathrm{S}\mathrm{r}}_2{\mathrm{C}\mathrm{a}\mathrm{C}\mathrm{u}}_2{\mathrm{O}}_{8+\delta }$, a highly studied cuprate superconductor, as they exposed each of the different layers, using Ar⁺ ion sputtering. However, the best approach to probe the exotic properties of an active layer is still an open question. Molecular beam epitaxy (MBE) is the ideal tool to allow precise control of different terminations, without the surface damage caused by sputtering. A shuttered growth approach allows for exploration of a much wider variety of surfaces, since we are not restricted to natural crystal cleavage planes. While most research on surface 2DEG of STO assumes a ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination, further research into the pristine SrO termination could lead to new insights into the system.

In this work, we explore the differences in the 2DEG formation between homoepitaxial STO films with SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ terminations by combining synchrotron-based ARPES with in-situ MBE growth. We observe a clear 2DEG only on SrO-terminated STO films and not on films with a ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination. We explore different growth recipes and find that the accumulation of an extra SrO layer, either from a flux imbalance or a buffer layer at the substrate interface, can cause the STO to continuously rearrange during growth to promote the extra SrO layer to the surface, which leads to a 2DEG visible in ARPES.

Results

MBE Growth and ARPES Characterization.

We grew homoepitaxial STO using well-established methods (22, 23) on top of commercially available 0.05% Nb-doped STO (001) substrates. The substrates used in this study were etched and annealed at the vendor to create atomically flat, ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated surfaces with well-ordered steps and terraces. We used a shuttered approach in which Sr and Ti shutters were opened in turns to impinge on the heated substrate in a background of molecular oxygen. Specific details of the growth parameters can be found in Materials and Methods. We used reflection high-energy electron diffraction (RHEED) to monitor the quality and thickness of the film during growth. We aligned the RHEED in such a way that the intensity was at a maximum when the Sr shutter closed and a minimum when the Ti shutter closed. A similar alignment methodology is described in detail by H. Y. Sun et al. (24). We were able to observe large RHEED oscillations, as seen in (Fig. 1A), which are indicative of high-quality STO films. Ex situ X-ray photoelectron spectroscopy (XPS) and X-ray diffraction (XRD) measurements, which can be found in SI Appendix, further confirmed the high quality of the films.

Fig. 1.

Each row corresponds to different STO growth recipes: normal ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated STO, SrO-capped STO, SrO buffer followed by STO, and Ti-deficient STO, respectively. (A, D, G, and J) RHEED oscillations for films with different recipes. The background color represents which shutter was open: blue for Ti and green for Sr. The small arrow marks the intensity at the end of growth. (B, E, H, and K) ARPES spectra collected for each recipe once it was fully saturated. Each measurement was taken at a photon energy of 74 eV with the sample at 30 K. (C, F, I, and L) Schematic diagrams which represent the final structure after growth is completed.

The homoepitaxial, ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated STO films were then characterized by using ARPES, as shown in Fig. 1. Each film was exposed to UV light until the spectra intensity was saturated. The background intensity, particularly at higher binding energies, increased over time. However, the expected 2DEG did not develop near the Fermi energy, as observed in Fig. 1B. Using our shuttered growth approach, we could terminate the film with an SrO layer by ending with the Sr shutter open last (Fig. 1D). When SrO-terminated films were exposed to UV light, a very bright 2DEG quickly emerged, similar to what has been observed in other measurements on STO, as seen in Fig. 1E. This same phenomenon was observed as well in ${\mathrm{B}\mathrm{a}\mathrm{T}\mathrm{i}\mathrm{O}}_3$ (BTO), where the 2DEG was only visible on the BaO termination, as seen in SI Appendix, Fig. S1. Additionally, when we buried a single ${\mathrm{L}\mathrm{a}\mathrm{T}\mathrm{i}\mathrm{O}}_3$ (LTO) layer below the surface of a STO host lattice (25), we only observed the expected doped electronic structure when we terminated with SrO.

Previous studies indicated unexpected layer ordering and termination when a double SrO layer is deposited during growth (26, 27). To confirm our surface terminations, we performed ex situ XPS on our STO films, and the results are summarized in SI Appendix, Fig. S2. All of the films which showed a 2DEG also had similar Sr/Ti ratios. By using the stark contrast in 2DEG formation between the SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ terminations, we can further explore the growth mechanics and termination of various STO recipes. We started by growing a single-layer SrO buffer layer and then continued growth by using the nominal STO recipe of Sr and Ti deposition in turns. This resulted in a RHEED oscillation which was 180° “out-of-phase” compared with a typical STO growth: Sr shutter closed at an intensity minimum and Ti shutter closed at an intensity maximum (Fig. 1G). ARPES measurements revealed a 2DEG, which, along with ex situ XPS, confirmed SrO termination, despite the Ti shutter being open last, as seen in Fig. 1H.

The correlation between SrO termination and 2DEG formation is robust against disorder. If a double SrO layer is formed midgrowth due to off-stoichiometric rates, like has been reported (26), RHEED oscillations flip 180° “out of phase” from nominal STO growth (Fig. 1J). Provided that there were a few good oscillations at the end of growth, the 2DEG was always observed, despite the disorder buried a few unit cells beneath the surface, as observed in Fig. 1K.

Partial STO Layers.

Due to our shuttered approach, we can also explore the effect of partial STO layers on the formation of the 2DEG. A succession of partial STO layers were grown on the same film. In between each growth, the sample was irradiated with UV light while being measured with ARPES, with the results shown in Fig. 2. All of the partial-layer samples reacted when exposed to the UV light. For all of the partial layers, and the fully SrO-terminated layer, there was intensity which extended up to the Fermi energy. However, only the fully SrO-terminated layer had bright spectra with clear dispersion. These measurements indicated that the formation of the 2DEG is dependent upon a fully SrO-terminated layer, and not just excess SrO on the surface. When this experiment was repeated with completely separate samples grown for each partial-layer termination, it yielded similar results, as seen in SI Appendix, Fig. S3.

Fig. 2.

(A) Schematic RHEED oscillation to show where along the growth cycle each sample in the figure was stopped. (B–F) A series of ARPES spectra taken on the same sample, but stopped at 1/2 SrO layer, full SrO layer, 1/4 ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ layer, 3/4 ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ layer, and full ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ layer, respectively, during growth. Each sample was exposed to UV light and fully saturated before collecting the data. The color scale is the same for all spectra. Each measurement was taken at a photon energy of 74 eV with the sample at 30 K.

Doping Dependence.

We also explored how the electronic structure evolved with UV exposure with ARPES, and the results can be seen in Fig. 3. The SrO-terminated homoepitaxial STO films showed a weak signal from the very first measurement. This intensity quickly developed into a 2DEG, which progressively doped until a point of saturation (Fig. 3 A–E). On the bare substrate, the 2DEG evolved in a similar manner, as seen in Fig. 3 F–J. However, compared with the fully doped 2DEG on the SrO-terminated surface, the intensity of the 2DEG on the bare STO wafer was much weaker. The in-situ MBE-grown STO films with ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination showed a negligible response over the same range of exposure. If we instead removed these ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated films from vacuum and prepared them using the etching and annealing methods described in ref. 20, the response to UV light changes. The doping response of the treated ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated films was very similar to that of the bare substrate, as seen in Fig. 3 K–O.

Fig. 3.

Progressive UV-irradiation series for 3 different samples: SrO termination for A–E, STO substrate for F–J, and treated ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated MBE grown STO for K–O. F–O have the same color scale. A–E are too bright for this same color scale, so the intensity is reduced by a factor of 5. Each spectrum was taken by using a photon energy of 42 eV and with the sample at 30 K. Due to the rapid development of the 2DEG with UV exposure, each spectrum was collected in 21 s, with 3-min spacing between each panel shown.

Discussion

While a 2DEG can form at the surface of a treated STO wafer with ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination, our results indicate the unexpected importance of the pristine SrO termination. Most research has focused on the ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination, and there are only a few results which explore the role of SrO. There are theories and experiments that suggest that for the ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated STO, oxygen vacancies cluster near the surface, with the formation barrier being similar for the topmost SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ layers (13, 28). Ab initio calculations predict that while the ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$-terminated surface requires oxygen vacancies to create an accumulation layer for a 2DEG, it should form on a pristine SrO-terminated film without the need for UV irradiation (29). However, the surface morphology of the 2 terminations are significantly different, and disorder can affect the observed energies of in gap states created by oxygen vacancies (13, 30). Thus, modeling pristine surfaces may not capture the true energy landscape or how inhomogeneity in the different layers affects the oxygen-vacancy formation barrier.

The appearance of the 2DEG only on SrO-terminated films leads to 2 possible scenarios. Either the SrO termination layer is the source of the itinerant carriers, or its presence causes carriers created elsewhere in the film to accumulate at the surface. Recent experiments suggest that the scenario with itinerant charge coming solely from the topmost SrO layer is at least plausible (12, 15). Higher oxygen-vacancy formation rates have been observed in fractured STO crystals compared with annealed STO wafers, which hints that exposure of the SrO termination can increase the vacancy-formation rate (15).

While we only observed the 2DEG on the SrO-terminated films, both terminations showed similar evolution of the gap state when irradiated with UV light and thus likely have similar levels of oxygen vacancies, as seen in SI Appendix, Fig. S4. Despite the similar levels of oxygen vacancies in both the SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ terminations, it could still be possible that the vacancies only create itinerant carriers in the SrO layer, while the created carriers are localized for the ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ termination. The fact that the sharp contrast between the 2 terminations is still observed when charge is provided by a buried LTO layer, and not oxygen vacancies, indicates that the SrO termination is not the source of the itinerant carriers, and a different mechanism is responsible for the appearance of the 2DEG. Recent studies of RHEED patterns during growth have shown a variation of the inner potential between the 2 surface terminations (24). In addition, scanning tunneling spectroscopy studies revealed a 0.25-eV shift in the conduction band onset between the 2 terminations (13). Thus, there is evidence for variations in the surface potential, despite the traditional assumption that both SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ layers are electrically neutral. When combined with our results, it is evident that the SrO layer creates an accumulations layer, as shown by previous ab initio calculations (29). However, in order for the 2DEG to be observed, additional itinerant charges, like those from oxygen vacancies, are required. More work is needed to fully understand the subtle differences between the films and etched substrates traditionally used in MBE, as well as the differences between theory and experiment regarding the formation of the 2DEG. The effects of improved film quality during growth with ozone or using alternate methods such as hybrid MBE (31) still need to be explored.

The dichotomy observed between the SrO and ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ terminations provides interesting insights into lattice structures consisting of active and passive layers. Due to the Ti d-orbital character of the observed bands, we can assign ${\mathrm{T}\mathrm{i}\mathrm{O}}_2$ as the active layer in STO; however, we do not observe the oxygen vacancy-created 2DEG or 3D dispersion from a buried charge reservoir when probing the active layer directly. Naively, we would expect that directly probing the active layer would provide access to the underlying physics, but our results show that the reality is more nuanced. While the SrO layer certainly plays a unique role in STO, it is likely that other systems also require a passive layer termination to realize or protect the physics of the active layer.

Our results also suggest that the interface between STO and other materials should be further explored. In particular, our results could help to understand the origins of the 2DEG at the LAO/STO interface, in which termination is already known to play an important role. Understanding the interactions between charge reservoirs and active layers as materials are grown layer by layer could help us to control the relevant emergent properties. Our work helps open routes to explore the physics of interfaces and achieve a deeper understanding of titanates and other oxides materials in general.

Materials and Methods

MBE Growth.

All of the samples were grown on 0.05% Nb-doped STEP STO purchased from Shinkosha. The substrates were mounted to Inconel sample holders by using silver paste. Once loaded into the MBE system, the samples were degassed at 300 °C for 30 min. They were then exposed to $6\times 10^{-6}$ torr molecular oxygen partial pressure and heated to 550 °C for growth. The RHEED was aligned along the crystal [110] direction with the electron beam’s incident angle increased from zero until the first intensity minimum of the reflected (00) beam was observed. Two cells were used for growth: a differentially pumped source loaded with ultra-high-purity Sr (99.95%) and a high-temperature cell loaded with ultra-high-purity Ti (99.995%). The atomic flux and deposition rates were calibrated and set by using a quartz crystal monitor. A shuttered approach was used for growth, which for a typical STO recipe starts with the Sr shutter open first and ends with the Ti shutter open last. The films presented in this paper were grown to 50 unit cells thick; however, the same behavior is observed in much thinner films as well. Postgrowth, the samples were cooled down in the oxygen background and then transferred in situ to the ARPES chamber for investigating the electronic structure.

ARPES.

All ARPES measurements were conducted at beamline 5-2 of the Stanford Synchrotron Radiation Lightsource. The base pressure of the ARPES chamber was better than $4\times 10^{-11}$ torr. Measurements were taken by using linear polarization with a component of the electric field normal to the sample surface, in the second Brillouin zone, at photon energies between 32 and 84 eV, while the sample was at 30 K. The energy resolution for the measurements were between 10 and 20 meV.

Ex Situ Characterization.

XPS and XRD were done ex situ at the Stanford Nano Shared Facilities with a Phi Versaprobe and PANalytical X’Pert, respectively.