Research ArticleMATERIALS SCIENCE

Erasable and recreatable two-dimensional electron gas at the heterointerface of SrTiO3 and a water-dissolvable overlayer

See allHide authors and affiliations

Science Advances  16 Aug 2019:
Vol. 5, no. 8, eaaw7286
DOI: 10.1126/sciadv.aaw7286

Abstract

While benefiting greatly from electronics, our society also faces a major problem of electronic waste, which has already caused environmental pollution and adverse human health effects. Therefore, recyclability becomes a must-have feature in future electronics. Here, we demonstrate an erasable and recreatable two-dimensional electron gas (2DEG), which can be easily created and patterned by depositing a water-dissolvable overlayer of amorphous Sr3Al2O6 (a-SAO) on SrTiO3 (STO) at room temperature. The 2DEG can be repeatedly erased or recreated by depositing the a-SAO or dissolving in water, respectively. Photoluminescence results show that the 2DEG arises from the a-SAO–induced oxygen vacancy. Furthermore, by gradually depleting the 2DEG, a transition of nonlinear to linear Hall effect is observed, demonstrating an unexpected interfacial band structure. The convenience and repeatability in the creation of the water-dissolvable 2DEG with rich physics could potentially contribute to the exploration of next generation electronics, such as environment-friendly or water-soluble electronics.

INTRODUCTION

With the substantial and increasing demands of consumer electronics in today’s society, electronic waste has become one of the fastest growing segments of environmental pollution (13). Therefore, the next generation of electronics with environment-friendly features, such as water dissolvability and recyclability, is in high demand. One of the possible approaches to realizing this prospect is the application of removable nontoxic materials in electronic devices, where functionalities can be created and removed by introducing and removing the material at will. However, thus far, a facile process of creation and removal of the material in electronics is still challenging to achieve.

Sr3Al2O6 (SAO), which is easily dissolvable in water (4), allows a possibility to realize water-soluble or recyclable electronic devices. Recently, two-dimensional electron gases (2DEGs) have been facilely created and patterned at oxide heterointerfaces via room temperature deposition of various types of aluminum (Al)–based amorphous overlayers on SrTiO3 (STO), such as LaAlO3 (LAO), Al2O3, and YAlO3 (5, 6). The interfacial 2DEGs originate from the oxygen vacancies, which were created by the Al-induced redox reactions (58). However, in all these heterointerfaces, the defects, e.g., oxygen vacancies, are permanently constrained inside the heterostructures with fixed overlayers. This consequentially limits its application in recyclable electronics (911), despite the fact that the facile room temperature fabrication process of this 2DEG system is convenient and promising for future adoption of oxides in electronics and thermoelectrics (5). Fortunately, as Al is also a main element of the SAO, the SAO meets the criteria for the occurrence of the Al-induced redox reactions (58). Therefore, it is potentially feasible to easily create a tunable 2DEG by controlling the concentration and distribution of the oxygen vacancy using the SAO, while the inherent solubility of the SAO offers a possibility to explore future water-soluble or recyclable electronics (1215).

In this work, we report an erasable and recreatable 2DEG at the heterointerface between a water-dissolvable amorphous Sr3Al2O6 (a-SAO) overlayer and single-crystalline STO. In particular, the patterning, deposition, and dissolution of a-SAO thin film on STO substrate are all done at room temperature, leading to convenience and advantages in the fabrication and recycling processes. Furthermore, on the basis of photoluminescence (PL) measurement, we show that the 2DEG is induced by the oxygen vacancies at the interface, which could be conveniently removed by exposing the a-SAO overlayer in a humid environment, such as water or moisture. Last, during the a-SAO and oxygen vacancy removal process, a nonlinear-to-linear Hall resistance transition occurs and the 2DEG eventually completes a metal-to-insulator transition (MIT), which is well explained by a two-band model supported by the first-principles calculations.

RESULTS

Time-dependent electrical transport characterization

Various types of a-SAO/STO heterointerfaces were prepared at room temperature as described in Materials and Methods. The facile room temperature deposition is very beneficial to potential electronic applications. Additional amorphous LAO (a-LAO)/STO heterointerfaces were also prepared as a control sample to study the transport behavior of the stable oxygen vacancy as a function of air exposure time (tair). It has been reported that the epitaxial SAO single-crystal film is water soluble (4). Here, we found that the a-SAO thin film can also be easily dissolved in the deionized (DI) water. Figure 1 (A to C) shows the temperature-dependent sheet resistance, RS(T), of the 4-nm a-LAO/STO, 4-nm a-SAO/STO, and 10-nm amorphous STO (a-STO)–capped 4-nm a-SAO/STO interfaces in an ambient environment with a relative humidity of 70% and different tair. Right after deposition, all interfaces exhibit a similar metallic behavior, including RS(T), carrier density (nS), and carrier mobility (μH), in the entire temperature range of 2.5 to 300 K (see fig. S2). While the RS(T) of the metallic a-LAO/STO interface remains unchanged as tair increases, the a-SAO/STO, however, is markedly affected by the exposure to ambient conditions, as illustrated in Fig. 1B. The metallic behavior (dRS/dT > 0) of the 4-nm a-SAO/STO interface is preserved up to tair = 48 hours, with marked changes in high-temperature RS and negligible changes in low-temperature RS. As tair increases further (greater than 48 hours), the 4-nm a-SAO/STO interface exhibits an abrupt transition to a semiconducting behavior and eventually becomes insulating (see Fig. 1B). Therefore, we show that an MIT can be triggered by leaving the sample in air for a number of hours to absorb moisture from air. Furthermore, the MIT at the a-SAO/STO interface can be greatly slowed by capping a 10-nm a-STO thin film (Fig. 1C). Figure 1 (D and E) summarizes the tair dependence of RS(2.5 K) and RS(300 K) of 4-nm a-LAO/STO, 4-nm a-SAO/STO, and 10-nm a-STO–capped 4-nm a-SAO/STO. A notable increase of RS(300 K) by more than one order of magnitude can be seen in 4-nm a-SAO/STO by exposing under ambient conditions. In contrast, both RS(2.5 K) of all heterointerfaces and the RS(300 K) of both the 4-nm a-LAO/STO and 10-nm a-STO–capped 4-nm a-SAO/STO show almost tair-independent behaviors (tair ≤ 48 hours), further confirming the correlation between the MIT and the removal of a-SAO.

Fig. 1 Temperature-dependent transport property with different tair.

A 2DEG is formed at (A) the a-LAO/STO (001), (B) a-SAO/STO (001) interfaces, and (C) a-STO–capped a-SAO/STO (001). The 2DEG at the a-LAO/STO interface is much more stable than that at the a-SAO/STO interface under ambient conditions of room temperature and 70% of relative humidity. RS as a function of tair at 300 and 2.5 K for (D) 4-nm a-LAO/STO, 4-nm a-SAO/STO, and (E)10-nm a-STO–capped 4-nm a-SAO/STO, respectively. Inset: The schematic view of oxygen-vacancy–induced 2DEG exists at the (A) a-LAO/STO, (B) a-SAO/STO, and (C) a-STO–capped a-SAO/STO interfaces.

Water-dissolvable and recyclable electronic devices

Because of the water solvability, recyclable electronic devices become possible with erasable and recreatable 2DEG enabled by adding or removing the a-SAO overlayer. A Hall bar device was built on the a-SAO/STO heterointerface using photolithography, and the a-SAO overlayer was then repeatedly removed and deposited. Figure 2B shows optical images of the Hall bar device with and without the a-SAO overlayer. The Hall bar device of the heterointerface experiences three states in each repetition, as schematically shown in Fig. 2C, namely, (i) just after growth (t = 0 min and 0 min < t < 30 min), (ii) one drop of DI water added onto the device to dissolve a-SAO (t = 30 min), and (iii) the a-SAO completely dissolved (30 min < t < 60 or 0 min). When the 20-nm a-SAO deposited on STO (t = 0 min), a 2DEG is created with a sheet conductance at 300 K, σS(300 K), of 10−4 ohm−1. At t = 30 min, a drop of DI water is added on the device. Note that the dissolution of SAO in water is a fast process. Therefore, a marked change of conductivity occurs almost instantaneously by adding one drop of DI water on top of the a-SAO/STO heterointerface, while the electrical transport varies gradually by exposing the a-SAO/STO heterointerface in the air (see Fig. 1D). This immediately removes the 2DEG with at least two orders of magnitude decrease in σS(300 K) down to the measurement limit. At 60 min (denoted as t = 0 min), the above process is repeated on the same device, and similar removal and recreation of 2DEG are demonstrated. Several repetitions of this process with a period of 1 hour are followed, which verifies the stability and repeatability of the electronic device (see Fig. 2C). In addition, during the practical implementation into a possible water-soluble electronic device, the key is to isolate the SAO and 2DEG from water or moisture. Fortunately, SAO is insoluble in organic solvent. Hence, the easiest method to the stability issues is to encapsulate the 2DEG with materials only soluble to a specific organic solvent. For example, polymethyl methacrylate is a versatile polymeric material and highly soluble in a solvent like acetone not water. Hence, with the help of one or several of these packaging materials, a water-soluble electronic device may be developed. Therefore, we demonstrate that the highly reproducible and reversible control of the 2DEG at the a-SAO/STO heterointerface by water dissolution provides a possibility to construct next generation electronic devices with environment-friendly features.

Fig. 2 Water-dissolvable and recyclable electronic devices.

A 20-nm a-SAO/STO (001) heterointerface with a Hall bar pattern (A) before and (B) after the a-SAO overlayer dissolving into the DI water. The a-AlN in (A) denotes amorphous aluminium nitride. (C) Time dependence of sheet conductance, σS, at 300 K in ambient conditions for 20-nm a-SAO/STO (001). The vertical dash lines indicate that a drop of water was added on the Hall bar device, and the horizontal dash line indicates the measurement limit. The right panels schematically illustrated the three states of the electronic device: (top) metallic just after growth (at each t = 0 min), (middle) a fast MIT with one drop of DI water added on the device (at each t = 30 min), and (bottom) becoming insulating after the a-SAO fully dissolved into the DI water (at each 30 min < t < 60 or 0 min).

PL and the schematic mechanism of oxygen vacancies

Figure 3A shows the room temperature PL spectra of an as-received STO substrate, 4-nm a-LAO/STO, 4-nm a-SAO/STO, and 10-nm a-STO–capped a-SAO/STO under different treatments, including exposing to air and dissolving in water. In all the a-LAO/STO, a-SAO/STO, and a-STO–capped a-SAO/STO samples, a broad and intense PL emission centered at 430 nm (2.9 eV) was observed. This 2.9-eV PL peak corresponds to oxygen vacancies in STO (8, 16). The a-SAO/STO, a-STO–capped a-SAO/STO, and a-LAO/STO all exhibit high PL intensity, indicating a high concentration of oxygen vacancies at the interfaces. Since all these samples show similar metallic behavior, we propose that the emergent conductivity in the a-SAO/STO interface right after the deposition also originates from oxygen vacancies at the surface of STO (see Fig. 3B). This is similar to that at the a-LAO/STO interface (5, 6, 8) and is further supported by the correlation between the interfacial conductivity and concentration of oxygen vacancy. The PL spectrum of the as-received insulating STO substrate was also measured as a baseline. We found that the baseline is a broad and weak PL peak at 415 nm (3.0 eV), due to the defects in the bulk of the as-received STO substrate created during the crystal growth (8, 16). Because the defect density is very small, the as-received STO substrate is insulating. Therefore, using the PL spectrum of the as-received STO substrate as the baseline, any additional oxygen vacancies at the interface would produce a PL peak with a higher intensity than the baseline. As shown in Fig. 3A, after exposure to ambient conditions for 1 hour, the PL intensity of the a-SAO/STO interface notably decreases, demonstrating the effective removal of oxygen vacancies. This also eliminates the possibility of oxygen vacancies moving deeper into STO bulk in-gap state with lower energy (Fig. 3C) (1719). This process at the a-SAO/STO interfaces can also be substantially accelerated by dissolving the a-SAO/STO sample in DI water (see fig. S3). Eventually, by dissolving in DI water, both a-SAO/STO and a-STO–capped a-SAO/STO heterointerfaces eventually become insulating with the final PL intensity the same as the PL baseline of the as-received STO substrate. To summarize, as the a-SAO overlayer dissolves, the oxygen vacancy–related PL intensity is quenched, and the conductivity of the a-SAO/STO heterointerfaces vanishes. Therefore, we further conclude that the amount of oxygen vacancies at the a-SAO/STO heterointerface can be effectively controlled by removing the a-SAO overlayer and eventually determines the conductivity of the a-SAO/STO heterointerface.

Fig. 3 Room temperature PL and the schematic mechanism of oxygen vacancies formed and filled near STO surface.

(A) PL intensity of an as-received STO substrate, 4-nm a-LAO/STO, 4-nm a-SAO/STO, and 10-nm a-STO–capped 4-nm a-SAO/STO just after deposition, after 1 hour of air exposure and after dissolving in water. The illustration of the oxygen vacancies formed and filled near the surface of STO equilibrium: (B) between a-SAO film and its surface and (C) between the ambient conditions and its surface. The VO″ represents oxygen vacancy occupying the in-gap state of STO, and the H2O in the black squre box represents a water molecule under ambient conditions absorbed at the STO surface.

There are three mechanisms that have been intensively discussed to explain the origin of 2DEG created at the heterointerface of amorphous/crystalline layer and STO, namely, (i) electronic reconstruction to resolve the polar catastrophe (8), (ii) thermal interdiffusion at the interface (20), and (iii) creation of oxygen vacancies in the STO substrate due to redox reaction (5, 6, 8). First, the macroscopic a-SAO grown on top of STO could be ordered on a local length scale (21). However, the a-SAO may not be a polar oxide because it does not have well-ordered crystalline (LaO)+ or (AlO2) sublayers as the LAO. In addition, we performed an annealing experiment to exam the possibility of any polar catastrophe–related mechanism (8, 21). We have prepared a crystalline SAO/STO (001) heterointerface at 760°C under oxygen partial pressure of 1 × 10−6 torr and then in situ annealed at 600°C for 1 hour in 1 bar of oxygen partial pressure to remove any oxygen vacancy. If the main origin is indeed the polar catastrophe, then the annealed crystalline SAO/STO heterointerface shall remain conducting (8). However, our annealed sample turns insulating. Therefore, the mechanism of polar catastrophe is experimentally ruled out. Second, an Al-doped STO (SrTi1−xAlxO3−δ) could be formed if the thermal interdiffusion of Ti and Al happens at the a-SAO/STO heterointerface. However, it is reported that SrTi1−xAlxO3−δ has a bandgap of 3.5 eV, which is 0.3 eV larger than that of undoped STO (22). Hence, the second possible mechanism is also ruled out. Last, we propose that the most probable mechanism for the origin of 2DEG is the oxygen vacancies resulting from the Al-induced redox reactions (5, 6, 8). After adding the a-SAO layer, the interface becomes conducting because of the creation of oxygen vacancies. On the contrary, the interface becomes insulating because of the removal of oxygen vacancies by dissolving the a-SAO. This is consistent with the PL results in our study (see Fig. 3A).

Notably, this evolution of oxygen vacancy is in contrast to other previously reported STO-based amorphous 2DEG systems and our reference a-LAO/STO samples, where the RS(T) is stable (5). Normally, the oxygen vacancies are believed to remain in the STO after the removal of the a-SAO, because bulk oxygen vacancies are thermally stable at room temperature under ambient conditions. To be specific, oxygen atoms are expected to enter (or escape from) bulk STO only when both kinetic energy (typically >600°C) and oxidative (or reductive) conditions are simultaneously fulfilled (5, 23, 24). Hence, the results of tair-dependent conductivity and PL demonstrate an unexpected feature of the oxygen vacancies at the surface of STO, namely, that the oxygen vacancies at the surface can be removed when exposed to the humid environment instead of being permanent, like the bulk defects.

The oxygen vacancies are removed by the oxygen atoms either from the DI water or ambient O2. To identify the origin, we performed control experiments. First, we find that the RS(300 K) of the 4-nm a-SAO/STO heterointerface is almost constant after a storage of 12 hours in both 1-atm pure O2 and 1-atm pure N2 at room temperature. Second, while the heterointerface is sealed in 1-atm N2 environment, an abrupt MIT occurs only when a drop of DI water was added onto the surface. Hence, we can conclude that the oxygen atoms, which filling the oxygen vacancies, are from H2O instead of the O2 from air, as depicted in Fig. 3C. Previous studies have shown that the oxygen vacancies at the surface of many complex oxides, such as STO (25, 26), TiO2 (27, 28), and BaTiO3 (25, 29), are immediately filled at room temperature when exposed to water. The presence of hydroxide ions on the oxide surfaces is revealed by both scanning tunneling microscopy (27) and x-ray photoelectron spectroscopy (25, 26, 28, 29), which indicates that water dissociation at the oxygen vacancy into two hydroxide ions is an energetically favorable state (27, 29). Hence, the most possible chemical reaction to fill the oxygen vacancy is H2O+VO+OO×=2(OH)O. The equations are expressed in Kröger-Vink notation, whereVO, OO×, and (OH)O denote oxygen vacancy, a lattice oxygen, and hydroxide ions occupying oxygen sites at the STO surface layer, respectively. In contrast, the conductance of both the reduced STO single-crystal substrate and oxygen-deficient thick STO film is stable, which is ascribed to the itinerant electrons buried deeply below the STO surface (3032).

Two-band model

Unexpected complexity in the types, density, and mobility of the carriers can also be created at the a-SAO/STO heterointerface, depict of the mild room temperature deposition condition of a-SAO. Figure 4 shows the tair-dependent Hall resistance data with extracted nS and μS at T = 2.5 K. As the tair increases, Rxy(B) evolves from nonlinear (Fig. 4A) to almost linear behavior (Fig. 4B). In general, there are a number of possible origins of the nonlinear Rxy, including (i) anomalous Hall effect in ferromagnetic materials (33), (ii) magnetic-field–induced phase transition (34), (iii) spatially separated parallel conducting channels (3537), and (iv) different orbital occupancies of multiple types of itinerant carriers (14, 38, 39). In our case, the first two mechanisms can be ruled out as the system showed neither signature of ferromagnetic order nor magnetic-field–induced phase transition. In the case of spatially separated parallel conducting channels, the different types of charge carriers are generally spread in two different regions, i.e., the overlayer and the interface (3537). However, in our a-SAO/STO samples, the only conducting region is the surface, and the multiple types of carriers are not expected to be spread widely. We, therefore, conclude that the different orbital occupancies of itinerant carriers are the only reasonable origin for the observed nonlinear Hall effect. This is also in good agreement with the previous studies on the (001)-oriented STO-based 2DEG systems, where the itinerant electrons occupy the states of Ti 3dxy and 3dxz,yz energy bands (14, 38, 39). The nonlinear Hall effect in both cases can be well fitted by a two-band model. In the two-band conduction model, there are two types of carriers coexisting at the interface, and thus, the Rxy and Rxx can be written asRxy=RHB=(μ12n1+μ22n2)+(μ1μ2B)2(n1+n2)e[(μ1n1+μ2n2)2+(μ1μ2B)2(n1+n2)2]B(1)andRxx(0)=1e(μ1n1+μ2n2)(2)where RH is the Hall coefficient, and n1, n2 and μ1, μ2 are the densities and mobilities of the two types of carriers, respectively (33, 35, 36, 38, 40). As shown in Fig. 4A, the experimental data (hollow symbols) of the nonlinear Rxy can be fitted by Eq. 1 (solid lines) with two types of electrons. Figure 4 (C and D) summarizes the extracted two types of carrier density and carrier mobility as a function of tair from 0 to 48 hours. We first focus our discussion on tair from 0 to 18 hours. Figure 4 (C and D) shows that the minority electrons have a relatively constant density n1 of 3.75 × 1012 cm−2 with mobility μ1 slightly increasing from 2100 to 3500 cm2 V−1 s−1, whereas the majority of electrons have density n2 decreasing from 6.12 × 1013 to 1.56 × 1013 cm−2 with mobility μ2 notably increasing from 288 to 1261 cm2 V−1 s−1. In the process, as the a-SAO is gradually dissolved, the enhancement in the carrier mobility shall be attributed to the reduction of defect scattering (13) and electron-electron scattering (41, 42). We note here the apparent appearance of a crossover point at tair = 24 hours. For tair longer than 24 hours, the electron mobility μ1 and μ2 becomes much closer. Providing μ1 = μ2, Eq. 1 can be simplified to Rxy=1e(n1+n2)B, consistent with the almost linear B dependence of Rxy (Fig. 4B). Overall, n1 changes slightly, while n2 decreases more than one order of magnitude, accompanied by an increase in mobility for both carriers with μ2 (2242 cm2 V−1 s−1) approaching μ1 (3000 cm2 V−1 s−1) at the crossover point of tair = 24 hours.

Fig. 4 Two-band model fitting of Rxy(B) and tair-dependent carrier density and mobility.

Hall resistance of a-SAO/STO as a function of magnetic field at T = 2.5 K for various tair (A) nonlinear and (B) almost linear Hall effect. The ring data points are the measured data, and the solid lines are fitted curves using the two-band model. The tair dependence of carrier (C) densities and (D) mobilities at T = 2.5 K. Two types of electrons are denoted by n1, n2, μ1, and μ2.

First-principles calculations

We propose that the evolution of multiple types of carriers is due to the change of the density of oxygen vacancies formed at the a-SAO/STO heterointerface. To clarify this, first-principles calculations on a 2 × 2 × 9 supercell of cubic STO (space group Pm3¯m, no. 221) with different number of oxygen vacancies were carried out to investigate the band structure. Figure 5 (A to D) shows the perfect supercell and that with one, two, and three oxygen vacancies, respectively, where, for clarity, only partial atoms of the supercell along the vertical direction are displayed. The corresponding band structures are plotted in Fig. 5 (E to H), along the high-symmetry point: M = (0.5, 0.5, 0), Γ = (0, 0, 0), and X = (0.5, 0, 0). It is apparent that, at the Γ point, there are two types of conduction bands near the Fermi energy with different symmetries are occupied. It turns out that the parabolic bands are of Ti 3dxy character, and the asymmetry bands are of Ti 3dxz,yz character. As we can see, the more oxygen vacancies at the heterointerface, the more Ti 3dxz,yz orbitals are occupied. Moreover, the effective mass of electrons occupying the Ti 3dxy orbital is lighter than that of electrons occupying the Ti 3dxz,yz orbital, as the effective masses associated to different bands are proportional to the inverse of the second derivative along Γ-X and Γ-M directions in the k-space (43, 44). Therefore, on the basis of the orbital occupancies of the STO-based 2DEG systems (14, 37, 38, 4549), we conclude that the observed two types of electrons occupy different 3d orbitals: The first type of carriers with an almost constant carrier density (n1) and higher mobility (μ1) occupies the Ti 3dxy subband, and the second type of carriers with a higher carrier density (n2) and lower mobility (μ2) occupies the Ti 3dxz,yz subbands.

Fig. 5 First-principles calculations of electronic band structures of a 2 × 2 × 9 supercell of cubic STO with oxygen vacancies.

Side view of the supercell structure with (A) zero, (B) one, (C) two, and (D) three vacancies, where, for clarity, only four layers along the vertical direction are shown. Corresponding band structures with (E) zero, (F) one, (G) two, and (H) three vacancies are also plotted, respectively.

DISCUSSION

In summary, we present an erasable and recreatable 2DEG at the interface between water-dissolvable a-SAO and STO. This interfacial 2DEG is induced by oxygen vacancies at room temperature, which is verified by the PL measurement. Taking a-LAO/STO heterointerface with a stable 2DEG as a control sample, we speculate that the erasable 2DEG is due to the water solubility of a-SAO. During the dissolving process, the oxygen vacancies, which were created by a-SAO and are responsible for the occurrence of 2DEG, are refilled rapidly once exposed directly to the ambient conditions or water. Moreover, the removal rate of the oxygen vacancies is accompanied by the evolution of two types of carriers, demonstrating a tantalizing band structure in the easily created and patterned 2DEG. The facile process to erase and recreate 2DEG in this study is of possible help to the future novel electronics with environment-friendly features.

MATERIALS AND METHODS

Sample preparation

The a-LAO, a-SAO, and a-STO films were deposited under oxygen partial pressure of 1 × 10−6 torr by pulsed laser deposition using a KrF excimer laser (λ = 248 nm) with a repetition rate of 1 Hz and a laser fluence of 1.3 J/cm2. For the STO substrates, a singly TiO2 termination atomically flat surface was achieved by chemical etching and annealed at 950°C (50, 51). Commercial LAO, STO single crystals, and sintered SAO ceramic were used as targets. Films deposited at room temperature showed an amorphous structure, which was confirmed by in situ reflection high-energy electron diffraction (RHEED). The growth rate was calibrated with the in situ RHEED oscillations obtained during the growth of single-crystalline SAO-STO samples (see fig. S1). All the amorphous films showed flat surfaces with a root mean square roughness less than 0.5 nm, as confirmed by atomic force microscopy (AFM).

Electronic transport and PL measurement

The sheet resistance and Hall effect were carried in a Quantum Design physical property measurement system with temperature range of 300 down to 2.5 K and magnetic fields up to 9 T. To determine the exact carrier density, a Hall bar pattern on STO substrate was prepared by a standard photolithography process. The PL spectra were carried at room temperature by a He-Cd laser with an excimer wavelength of 325 nm. A broad characteristic PL peak of oxygen vacancies centered at 430 nm (2.9 eV) was observed for different oxygen-deficient samples.

First-principles calculations

First-principles calculations were carried out using the Vienna Ab Initio Simulation Package (52) with the projector-augmented wave potentials (53). The kinetic energy cutoff is fixed to 520 eV, and the Monkhorst-Pack k-point mesh (54) is taken as 6 × 6 × 1. The coordinates and the cell shape were fully relaxed until the forces acting on the atoms are all smaller than 0.05 eV/Å.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/8/eaaw7286/DC1

Section S1. RHEED oscillations of 4 unit cells (uc) of SAO/STO (001)

Section S2. Electrical transport properties of 4-nm a-SAO/STO and 4-nm a-LAO/STO

Section S3. Microscopy images of 20-nm a-SAO/STO (001) before and after dissolving into the DI water

Section S4. Microscopy images of 20-nm a-SAO/STO sample with different tair

Section S5. Electrical transport properties of 4-nm a-SAO/STO with different tair

Section S6. tair dependence of RS for 50-nm a-SAO/STO (001) and 10-nm a-STO–capped 4-nm a-SAO/STO

Section S7. Nonlinear to linear Hall effect of 4-nm a-SAO/STO heterointerface with different tair

Section S8. Magnetoresistance of 4-nm a-SAO/STO with different tair

Section S9. Growth oxygen partial pressure dependence of RS

Section S10. AFM images of reusable STO surface topography

Section S11. Discussion of the possible chemical reactions to dissolve a-SAO and remove oxygen vacancies at STO surface by water

Fig. S1. RHEED oscillations of 4 uc of single-crystalline SAO grown on STO (001).

Fig. S2. Electrical transport properties of 4-nm a-SAO/STO and 4-nm a-LAO/STO.

Fig. S3. Microscopy images of 20-nm a-SAO/STO (001) with Hall bar pattern.

Fig. S4. Microscopy images of the surface of 20-nm a-SAO/STO sample with different tair.

Fig. S5. Electrical transport properties of 4-nm a-SAO/STO with different tair.

Fig. S6. tair dependence of RS.

Fig. S7. Nonlinear to linear Hall effect of 4-nm a-SAO/STO heterointerface.

Fig. S8. Magnetoresistance of 4-nm a-SAO/STO at 2.5 K with different tair.

Fig. S9. Growth oxygen partial pressure dependence of RS at 300 K.

Fig. S10. AFM images of reusable STO surface topography.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We thank C. Mun Yin and K. Onggadinata for their fruitful discussion. Funding: X.R.W. acknowledges supports from the Nanyang Assistant Professorship grant from Nanyang Technological University and Academic Research Fund Tier 1 (RG108/17 and RG177/18) and Tier 3 (MOE2018-T3-1-002) from Singapore Ministry of Education. K. Hu acknowledges financial support from National Natural Science Foundation of China (grant no. 11647108) and the China Scholarship Council (grant no. 201808440038). Z.H., S.Z., and A.A. acknowledge the support from the NUS Academic Research Fund (AcRF Tier 1 grant nos. R-144-000-391-114 and R-144-000-364-112) and the Singapore National Research Foundation (NRF) under the Competitive Research Programs (CRP grant no. NRF-CRP15-2015-01). D.Q. and J.Y. acknowledge the support of the Australian Research Council (grant no. FT160100207 and FT160100205, respectively). Author contributions: X.R.W., W.L., and K.Ha. conceived and designed the experiments. K.Ha., Z.H., and S.Z. prepared the samples and performed the electrical transport property measurements. K.Hua., C.Y., and J.Ya. performed the room temperature PL measurement. K. Hu and X.L. carried out the first-principles calculations. All authors contributed to the discussion and interpretation of the results. K.Ha., W.L., and X.R.W. wrote the manuscript with help from all other authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article