Introduction

The transfer of electrons between a solid surface and adsorbed atomic or molecular species is fundamental in natural and synthetic processes, being at the heart of most catalytic reactions and many sensors. In special cases, metallic conduction can be induced at the surface of, for example, Si-terminated SiC1, or mixed-terminated ZnO2, in the presence of a hydrogen adlayer. Generally, only the surface atoms are significantly affected by adsorbates. However, remotely changing electronic states far from the adsorbed layer is possible, if these states are electrostatically coupled to the surface. Here we show that the surface adsorption of common solvents such as acetone, ethanol, and water can induce a large change in the conductivity at the buried interface between SrTiO3 substrates and LaAlO3 thin films3,4,5,6,7,8, and an adsorbate induced insulator to metal transition was observed when the thickness of LaAlO3, dLAO, is around 3 unit cells (uc). This phenomenon is observed only for polar solvents. Our result provides experimental evidence that adsorbates at the LaAlO3 surface induce accumulation of electrons at the LaAlO3/SrTiO3 interface, suggesting a general polarization-facilitated electronic transfer mechanism, which can be adapted for use in a range of oxide-based devices.

The intense interest in the LaAlO3/SrTiO3 interface3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22 has led to several recent experimental observations that suggest a close relationship between the interface and the LaAlO3 surface. These include the use of conducting atomic force microscopy probes to toggle a metal insulator transition10,11,12 through the writing of surface charge12,13, and the use of capping layers of SrTiO3 (ref. 14) or SrCuO2 (ref. 15) to alter the interface conductivity. The origin of these effects, and indeed the interface conductivity itself, remains in debate4,16,17,18,19,20,21,22. The transfer of electrons from the surface to the interface, to reconcile the polar discontinuity between the neutral {100} layers in SrTiO3 and the charged layers in {100} LaAlO3, is one model that naturally connects the electronic states of the interface and LaAlO3 surface4,18,19,20,21,22. However, despite these concepts and recent experimental progress, the effect of surface adsorbates has not been investigated. This issue is addressed in the present work. We have found that the exposure of LaAlO3/SrTiO3 samples to a polar solvent can increase the sheet carrier density, n2d, by more than 2×1013 cm−2, representing a change of the same order as the total charge density typical in this system5,6,7,8. Compared with the strong perturbations associated with other surface processes, that is, the extremely strong local electric field produced by the conducting atomic force microscopy probes10,11,12,13, or the structural variation by introducing capping layers grown at high temperatures14,15, the changes associated with room-temperature treatment using these solvents would naively be expected to be small. However, such processes result in a surprisingly large modulation of n2d, revealing a dramatic surface-interface coupling.

Results

Transport characterization

The fabrication of the conducting LaAlO3/SrTiO3 interfaces and the surface adsorption process (SAP) are described in the Methods. As shown in Figure 1a, labels 1 & 2, a SAP using acetone increased n2d from ∼1×1013 cm−2 to more than 3×1013 cm−2 for a dLAO=10 uc sample, over a wide temperature range (2 K≤T≤300 K). This remarkable modulation suggests that the acetone molecules have been adsorbed on the LaAlO3 surface and, in turn, change the electronic states at the buried interface. Heating the sample at T=380 K in a moderate vacuum (<103 Pa, with a helium background) for several hours produced negligible conduction change (not shown). However, the original sample state was recovered after heating at an elevated temperature of 653 K, evidenced by the decrease in n2d (Fig. 1a, label 3) and the subsequent increase after another SAP using water (Fig. 1a, label 4). These facts indicate that the surface adsorption is reversible but strong, not a physical adsorption driven by the relatively weak van der Waals force, and thus likely involves electron transfer. We note that as a function of n2d, the qualitative shape of the n2d(T) data at low temperatures is quite different. This is likely associated with the complex temperature and electric-field dependencies of SrTiO3, leading to a non-trivial variation of the confining potential as a function of n2d, as discussed elsewhere23.

Figure 1: Transport characterization.
figure 1

Temperature dependence of (a) Sheet carrier density, n2d, (b) Hall mobility, μH, and (c) Sheet resistance, RSheet, in different states of a dLAO=10 uc LaAlO3/SrTiO3 sample. The black closed circles represent the as-grown sample before experiencing any SAP. The red open circles represent the sample after an acetone SAP. The blue closed triangles represent the sample after heating at T=663 K in an O2 pressure of 1.33×10−2 Pa for 5 h. The pink open triangles represent the sample after a water SAP. The processing and measurement sequence follows the labelled order in (a). (d) Room-temperature conductivity, 1/Rsheet, of a dLAO=10 uc LaAlO3/SrTiO3 sample that was repeatedly processed by water SAP (red closed squares) and recovered by heating at 573 K in an oxygen flow for 3 h (blue closed circles). The black open circle represents the as-grown sample.

Full size image

Accompanying the n2d increase, a striking reduction in the Hall mobility, μH, from 3,000 cm2 V−1 s−1 to 600 cm2 V−1 s−1 (T=2 K) is observed after the acetone SAP (Fig. 1b). μH decreases quickly with increasing temperature, with a relatively small room temperature value of μH ∼6 cm2 V−1s−1 due to the dominance of phonon scattering. At this temperature, μH also has a very weak dependence on n2d, thus the conductivity 1/Rsheet is a reasonable and convenient index of n2d (Fig. 1a, c), and we will use it hereafter to characterize the effect of the SAP process on n2d for a variety of common solvents. To study the repeatability of the adsorption/desorption processes, we repeatedly performed water SAP on a LaAlO3/SrTiO3 sample followed by a heating step at 573 K in an oxygen flow, to reset the system (Fig. 1d). The sample switches reproducibly between a high-conductivity state after SAP and a low-conductivity state after heating.

Effect of SAP using different solvents

To explore the origin of the SAP-induced n2d, we have studied the effect of SAP using a variety of solvents with dLAO=10 uc samples. These solvents can be classified into three categories: non-polar, polar aprotic (no dissociable H+), and polar protic (dissociable H+). The results are summarized in Figure 2. Clearly, SAP using non-polar solvents produces no significant modulation in 1/Rsheet and by extension, n2d. SAP using a relatively weak polar solvent such as ethyl acetate produces a slight increase in n2d, whereas using more polar solvents produces a much larger increase, independent of whether the solvents are aprotic or protic. In Figure 2, we also show that Δ(1/RSheet) is well scaled with the molecular dipole moment divided by the molecular volume. These results suggest that the polar nature of solvents has a key role in determining the magnitude of the observed changes for the different solvents. In the context of an electrostatic mechanism, the scaling of the change with the dipole moment density would be reasonable, as the strength of the effect should be related to not only the molecular dipole moment, but also the areal density of the molecules on the surface.

Figure 2: Effect of SAP using different solvents.
figure 2

The difference in 1/Rsheet after and before SAP, Δ(1/Rsheet), reflects the change in n2d at room temperature. A clear increase in Δ(1/Rsheet) is observed after SAP using polar solvents, independent of their aprotic or protic character. Δ(1/Rsheet) is reasonably scaled with the molecular dipole moment (D) of each solvent divided by the corresponding molecular volume (Vm) (crosses). The line is a guide to the eye. In each category, the solvents are ordered from left to right by increasing polarizability. At least two samples were used for each solvent and all samples were dLAO=10 uc and in the as-grown state before SAP.

Full size image

Comparison between SAP and adsorption of solvent vapour

Further insights into the origin of the SAP-induced n2d can be made by comparing the effect of SAP with the effect of adsorption from a solvent vapour. It is well known that a thin layer of water coats all hydrophilic surfaces under ambient conditions24. Until now the adsorption of water on LaAlO3, particularly on the AlO2-terminated LaAlO3 surface, has been scarcely studied. A water-dipping study showed that, whereas La2O3 is highly hygroscopic, LaAlO3 only reacts weakly with water25. In contrast, the adsorption of water on Al2O3 is well understood. It is known that Al2O3 adsorbs 1 monolayer of water when the relative humidity (RH) is ∼35%, and more than 5 monolayers for RH >70% (refs 26,27). Thus, although these studies cannot be trivially extrapolated to the current heterostructures, we expect a thin layer of water with full coverage will coat the AlO2-terminated LaAlO3 surface when the LaAlO3/SrTiO3 sample is exposed to RH >70% air at room temperature. Such a surface layer has also been inferred from charge writing experiments28. Nonetheless, despite the existing water layer, we can still modulate the conductivity using the macroscopically large volumes of solvent in the SAP process.

Figure 3a shows that the conductivity of LaAlO3/SrTiO3 is still sensitive to water SAP using liquid even after exposing it to water-saturated atmosphere for several hours. A similar result was found when acetone was used (Fig. 3b). From these results, we can conclude that the thickness of adsorbate layer also has an important role in the SAP effect, as the SAP with liquid is expected to produce a thicker adsorbed solvent film compared with the vapour experiments.

Figure 3: Comparison between the effect of SAP and the effect of adsorption of solvent vapour.
figure 3

(a) Conductivity versus time plots for a sample exposed to (I) air of RH=30%, (II) air near the saturation point of water, and (III) followed by a water SAP. (b) Conductivity versus time plots for a sample exposed to (I) air of RH=30%, (II) air near the saturation point of acetone, and (III) followed by an acetone SAP. Note that surface adsorption of, for example, water, would have already taken place when the sample was kept in normal air. The exposure to the nearly saturated water (acetone) vapour for many hours should achieve a full and thicker coverage of adsorbates on the LaAlO3 surface. The samples are dLAO=10 uc, and the measurements were performed at ambient environment. The irregular perturbations in (a) were caused by fluctuations in the local illumination and room temperature.

Full size image

LaAlO3 thickness dependence

A key feature of the LaAlO3/SrTiO3 interface is that there exists a critical LaAlO3 thickness, dLAO ∼4 uc, below which the interface is insulating5. It is instructive to see the dependence of the effect of SAP on the LaAlO3 thickness. As shown in Figure 4a, for dLAO=2 uc sample and bare SrTiO3 substrate, no modulation in the conductivity is observed. For dLAO=3 uc, the as-grown sample is insulating and the acetone SAP drives the sample from an insulating state to a conducting state. In this case, the absolute modulation in the carrier density, as indexed by Δ(1/RSheet), is smaller than that in the thicker samples, but the sensitivity, defined as the ratio of the conductivity change with respect to that of the as-grown sample, is much larger. We speculate that, with suitable optimization, this large sensitivity may be exploited for sensor applications.

Figure 4: LaAlO3 thickness dependence.
figure 4

(a) The effect of SAP on samples of different LaAlO3 thicknesses. The blue closed circles represent 1/Rsheet of the as-grown samples. The red open circles represent 1/Rsheet of the samples after an acetone SAP. The difference after and before acetone SAP, Δ(1/Rsheet), is represented by black crosses. Sensitivity (green triangles) was defined as a ratio between Δ(1/Rsheet) and 1/Rsheet (as grown). The green arrow indicates the measurement limit. The dLAO=0 uc sample is a bare SrTiO3 substrate. The acetone SAP triggers a transition from an insulating state to a conducting state for a dLAO=3 uc sample. (b) The time dependence of 1/Rsheet after acetone SAP for dLAO=3, 5, and 10 uc samples. All measurements were performed at ambient environment. The small perturbations in the data, notably on the 3 uc curve, were caused by small fluctuations in the room temperature. Lines are guides for the eye.

Full size image

Figure 4b shows the time-dependent 1/Rsheet, immediately after an acetone SAP, of LaAlO3/SrTiO3 samples of different LaAlO3 thicknesses. As dLAO decreases, 1/Rsheet tends to relax more quickly, indicating solvent desorption. This observation suggests that the surface adsorption is directly related to the nature of the LaAlO3 surface, or its polar character. We note that a somewhat analogous observation has previously been found for the charge stability on the LaAlO3 surface in the same system12.

Discussion

Next, we discuss the possible mechanisms that may explain these data. We exclude electrostatic attraction because the polar adsorbates themselves are charge-neutral as a whole and will not change the electrostatic boundary condition of LaAlO3/SrTiO3 without the transfer of electrons29. Oxygen vacancies in the SrTiO3 (ref. 16), atomic interdiffusion17, and an electronic reconstruction4,18,19,20 have all been suggested to be the origin of the interfacial conductivity in the LaAlO3/SrTiO3 system. In the case of the SAP experiments, we exclude the first two of these as we assume that a SAP at room temperature does not cause significant atomic structure changes at the interface, and we focus on the electronic reconstruction mechanism.

As discussed in several papers19,20, and illustrated in Figure 5a, for an idealized LaAlO3 surface free from surface states, the valence band of LaAlO3, EVLAO is the electron source for the interface electrons. Here we assume no conductivity in the LaAlO3 film itself, and, for simplicity, do not consider other processes that do not affect the electrostatic state of the LaAlO3. An uncompensated potential, VUncom, across the LaAlO3 film of the order of the band gap of SrTiO3, EgSTO is needed to align the chemical potential over the whole structure. A simple electrostatic consideration shows that

, where σ0 is the charge density of one atomic layer, σInter is the total sheet electron density at the interface, e is the unit charge, and ɛLAO is the dielectric constant of LaAlO3. In reality, σInter is larger than n2d because of the trapping of some electrons at the interface19. Immediately, we find that

This simple relationship predicts the same thickness dependence of σInter as several more detailed theoretical calculations20,30, noting that VUncom is fixed for an ideal surface. In the work presented here, dLAO is fixed, and the SAP induced change in n2d is thus associated with a reduction in VUncom.

Figure 5: Sketches of band diagrams and electron transfer mechanisms.
figure 5

(a) Idealized surface. Electrons transfer from EVLAO of the LaAlO3 surface to the conduction band of SrTiO3, ECSTO, near the interface. eVUncom roughly equals to EgSTO. (b) Surface with aligned polar adsorbates. Electrons transfer from the surface adsorbates to ECSTO near the interface. The built-in potential across the adsorbates, Vad, effectively reduces VUncom across the LaAlO3 layer. In both (a,b) bulk stoichiometry and valences are shown for simplicity.

Full size image

We note that there are several possibilities of reduction of VUncom. In practice, all samples have surface states associated with inevitable defects, surface reconstructions, surface adsorbates and so on. If these surface states are within the bandgap of LaAlO3, they can replace EVLAO as more energetically favourable electron sources, where a relatively smaller VUncom is required to dope electrons into the SrTO3. As far as surface adsorbates are concerned, they may interact with surface atoms or oxygen defects, and cause complex surface states. Two recent theoretical papers have suggested that the dissociation of hydrogen21 and redox reactions22 on the LaAlO3 surface as possible electron sources. In both cases, VUncom is reduced compared with that of the idealized surface. The role of water on the LaAlO3 surface has also been recently stressed in theoretical calculations by S. Hellberg (personal communication). However, we caution that the observed accumulation of electrons by SAP cannot be fully explained by adsorbate-induced surface states alone, because the polar character of the adsorbates and their thickness were found to have a vital role in our experimental results.

With the above picture in mind, we can interpret our results as follows: the strong polarization in the LaAlO3 layer aligns to a degree the polar adsorbates nearest the surface, with a decaying trend as we move away into the adsorbate layer. As shown schematically in Figure 5b, this leads to an electrostatic potential across the adsorbate layer, Vad, quite similar to the built-in potential in the polar LaAlO3. As a result, electrons are sourced from the adsorbates instead of EVLAO of the surface LaAlO3 to the LaAlO3/SrTiO3 interface, and VUncom across the LaAlO3 is effectively decreased. σInter increases according to Equation (1), consistently explaining the observed SAP-induced accumulation of electrons at the LaAlO3/SrTiO3 interface. In this picture, the change in σInter is proportional to the value of Vad and the areal density of adsorbates. If we assume that the thickness of the adsorbate layer is the same for all solvents, and that all solvents are equally likely to transfer electrons it is simple to show that ΔσInter will be proportional to D/Vm, consistent with the experimentally observed scaling relation between Δ(1/RSheet) and D/Vm (Fig. 2).

Our present observation is quite distinct from the proposed water-cycle mechanism that accounts for the modulation of interface conductivity by biased conductive scanning probe, because we achieve conductivity changes without actively dissociating water molecules28. We note that in principle the biased scanning probe will also change the physical distribution of the adsorbates, (for example, water), in addition to chemical changes, which may also contribute to the modulation in the interface conductivity in the charge writing experiments10,11,12,13.

The observed close coupling between polar surface adsorbates and interface conductivity provides new insights into the origin of the electron gas at the LaAlO3/SrTiO3 interface, and demonstrates a new tuning parameter in controlling this state, which has the potential to be applied broadly in a wide range of novel oxide-based devices. These data also suggest that many of the conflicting (and often contradictory) studies of this popular system can be explained by this observation, that is, whether measurements are made in situ31, or after various surface exposures (for example, atmospheric water, or solvent cleaning)32,33,34, the sensitive electronic structure of the interface may be dramatically altered. For example, the predicted electric field across the LaAlO3 layer18,19,20 was not observed in X-ray photoemission studies32 that may be explained consistently because the surface adsorbates (in particular atmospheric water) induced by the ex-situ processing reduce VUncom, or equally, the electrical field across the LaAlO3 layer. In contrast, in the LaAlO3/SrTiO3 superlattices of multiple interfaces, the interfaces relatively far away from the surface may not be significantly affected by adsorbates34.

Methods

Sample fabrication

Using optical lithography and lift-off, a six-contact Hall bar (central bar length 50 μm, width 10 μm) with amorphous AlOx as a hard mask was patterned onto the atomically flat TiO2-terminated SrTiO3 (100) substrates. The LaAlO3 thin films were grown on the patterned substrates by pulsed laser deposition with the growth monitored by in-situ reflection high-energy electron diffraction. The conducting interfaces are confined within the Hall bar region. For the samples shown in this article, the growth was at 923 K in an O2 pressure of 1.33×10−3 Pa, after a pre-annealing at 1,223 K in an O2 pressure of 6.67×10−4 Pa for half an hour. After deposition, there followed a post-annealing step at 873 K in an O2 pressure of 4×104 Pa for one hour. During deposition, the laser repetition was 1 Hz and the laser energy density was 0.6 Jcm−2. The same phenomenology is found for samples grown in a variety of temperatures, oxygen pressures and laser conditions.

Electrical contact and measurement

The conducting interface was contacted by ultrasonic bonding with Al wire. The temperature-dependent sheet resistance, sheet carrier density and mobility were deduced from standard transport measurements using the patterned Hall bar. All transport measurements were carried out using a standard four-probe method. To test for the possibility of parallel conductivity through the adsorbate layer itself, contacts were made directly to the surface using silver epoxy. No significant surface conduction was detected both before and after SAP, clearly demonstrating that the conduction is at all times dominated by the LaAlO3/SrTiO3 interface.

Surface adsorption process

SAP was achieved by placing a drop of the liquid solvent on the sample surface at room temperature and blowing off all visible solvent within less than 10 s using dry nitrogen gas. Contacts were either left on the samples during the SAP, or reconnected afterwards; no significant difference in the conduction was observed by using either method.

Additional information

How to cite this article: Xie Y. et al. Control of electronic conduction at an oxide heterointerface using surface polar adsorbates. Nat. Commun. 2:494 doi: 10.1038/ncomms1501 (2011).