Introduction

Perovskites-based structures including transition metal oxides have attracted much attention in the past decades, with the discovery of high-Tc superconductivity and colossal magnetoresistance1. More generally, these compounds exhibit various electronic orders, going from the canonical anti-ferromagnetic (AF) Mott insulator when the onsite repulsion is maximum because of strong electronic interactions, to Fermi liquid-like metals when carrier doping is such that screening prevents the system from localization. Depending on the cations and the doping level involved, charge, spin and orbital orders can appear in the ground state together with metallic and even superconducting phases. Transitions between these states can be tuned by temperature, magnetic or electric fields2. All these compounds can be seen as stacks of oxide layers where the charge neutrality is conserved in the unit cell (u.c.), but not necessarily in each layer. Therefore, the translation symmetry is locally broken at the interface and charge imbalance can develop. Like in band-gap engineering with semiconductors, it is possible to create artificial interface materials by growing thin layers of a transition metal oxide on top of another one. Recently, the observation of two-dimensional (2D) superconductivity3 and magnetic correlations4 at the interface between the two band insulators LaAlO3 and SrTiO3 has drawn a lot of attention. Another particularly interesting candidate is the homo-metallic structure LaTiO3/SrTiO3 that uses TiO2 plans as a building block5,6. Titanium is in the 3d0 state in the SrTiO3 layer which is a band insulator of 3.2 eV bandgap, whereas it is 3d1 in the LaTiO3 one which is therefore an AF Mott insulator due to strong correlations7. Providing the interface layer is TiO2, an extra electron is left in the structure every 2 u.c.8,9. As shown by photoemission10 and optical studies11, a two-dimensionnal electron gas (2-DEG) develops and extends a few u.c. beyond the interface.

Several theoretical approaches pointed out that an electronic reconstruction leads to an increase of the electronic density at the LaTiO3/SrTiO3 interface8,9,12,13. Okamoto and Millis8 proposed a phase diagram where different orbital and magnetic states occur as a function of the thickness of the LaTiO3 layer and the strength of the Mott-Hubbard parameter U/t (U is the Coulomb onsite repulsion energy, and t the hopping term between neighbour Ti sites). Fully polarized ferromagnetic metallic sub-bands are expected to form for thickness below 5 u.c. and U/t~8–10. However, Kancharla and Dagotto,14 taking into account both local and long-range Coulomb interactions, showed that strong AF fluctuations reminiscent of the magnetic order of the bulk compound persist in the metallic phase. As suggested by Larson15 and Okamoto16, lattice relaxation at the interface strongly modifies the band configuration, and may enhance the electronic correlations in the 2-DEG12. In this context, it is clear that the LaTiO3/SrTiO3 interface layer appears to be a unique system to study the physics of a 2-DEG influenced by strong electronic correlations. In this study, we show that this heterostructure undergoes a superconducting transition and we are able to characterize the 2-DEG as extending mostly towards the SrTiO3 side of the interface.

Results

Low-temperature transport measurements

We have grown epitaxial layers of LaTiO3 using pulsed laser deposition (PLD) on single crystal substrates of SrTiO3 cut along (100) and (110) crystallographic directions. The details of the growth conditions and X-ray characterizations are given in Methods section and in Supplementary Figures S1 and S2. In this study, we focus mainly on two LaTiO3/(100)SrTiO3 heterostructures whose thickness of 40 and 60 Å correspond to 10 and 15 u.c. respectively. The sheet resistance measured in a Van der Pauw geometry decreases with temperature, indicating a metallic behaviour of the interface (Fig. 1). At temperatures lower than 20 K the two samples exhibit an increase of resistance characteristic of weak localization in disordered 2D films. The heterostructures undergo a superconducting transition at Tconset≈310 mK for the 10 u.c. sample and at Tconset≈260 mK for the 15 u.c. sample (inset, Fig. 1). Thinner 5 u.c. (100) films and 20 to 100 u.c. thicker (100) films as well as (110) oriented films are not metallic at low temperature.

Figure 1: Sheet resistance of the hererostructures.
figure 1

Sheet resistance a of the 10 u.c. (black dots, left axis) and 15 u.c. (blue dots, right axis) LaTiO3/SrTiO3 samples in an intermediate range of temperature.The red lines correspond to quadratic fits of the form R(T)=AT2+R0. (Inset) Sheet resistance as a function of temperature showing the superconducting transitions at Tconset=310 mK for the 10 u.c. (black dots) and Tconset=260 mK for the 15 u.c. (blue dots), where Tconset is defined by a 10% drop of the resistance.

Full size image

In Figure 2, we show the current–voltage characteristics of the 10 u.c. sample measured at different temperatures. At low temperature, the I(V) curves show a clear critical current Ic of 5 μA corresponding to a critical current per unit width of 16.7 μA cm−1. For current much higher than Ic, the I(V) curves merge together on a linear Ohmic law with a resistance corresponding to the normal resistance. In the case of the 15 u.c. sample, the critical current per unit width is found to be 14 μA cm−1. Figure 3a shows the sheet resistance of the 10 u.c sample as a function of temperature measured for different values of a magnetic field applied perpendicularly to the sample. The magnetic field induces a transition from a superconducting state to a nonsuperconducting one. The dependence of the critical field as a function of temperature is linear close to Tc, which is consistent with the form taking into account a Landau–Ginsburg in-plane coherence length (Fig. 3b). The critical field extrapolated at T=0 is mT for the 10 u.c. sample and mT for the 15 u.c. sample (see Supplementary Fig. S3). At T=0, we found and . Measurements performed in a parallel magnetic field geometry give for the 10 u.c sample and for the 15 u.c sample. We thus extract the thickness of the 2D superconducting electron gas d15u.c.=12 nm and d15u.c.=13.5 nm. Note that this is an upper bound given the precision of the sample alignment in the parallel magnetic field. These values are close to the ones reported in LaAlO3/SrTiO3 heterostructures17,18. In disordered electronic system, weak localization produces a decrease of conductivity that can be experimentally revealed by varying the temperature. In the particular case of a 2D system, the conductivity takes the remarkable logarithmic dependence with temperature , where p depends on the process that limits the phase coherence; p=3 for electron–phonon scattering and p=1 for electron–electron scattering in the dirty limit19. Such logarithmic temperature dependence is observed on our samples (see Fig. 3c), thus confirming the 2D nature of the electron gas. The fit gives p=0.97±0.05 showing that the phase coherence is limited mainly by electron–electron scattering.

Figure 2: Current–voltage characteristics of the 10 u.c. sample.
figure 2

The critical current at low temperature is 5 μA corresponding to a critical current per unit width of 16.7 μA cm−1.

Full size image
Figure 3: Sheet resistance of the hererostructures under magnetic field.
figure 3

(a) Sheet resistance of the 10 u.c. sample as a function of temperature for different values of the perpendicular magnetic field. (b) Temperature dependence of the perpendicular critical field, defined as the magnetic field that suppresses 90% of the resistance drop. The red line indicates the linear dependence of the critical field with temperature close to Tc. (c) Conductivity of the 15 u.c. sample as a function of temperature for a perpendicular magnetic field corresponding to the critical field. The red line corresponds to the expression with p=0.97 indicating that phase coherence is limited by electron–electron scattering (p=1)19.

Full size image

Hall effect measurements.

To investigate the density and mobility of charge carriers, we performed Hall measurements at low temperature (Fig. 4). The experiment confirms that the sign of the hall coefficient is negative for both samples, indicating that electron-like charge carriers dominate the transport. The sheet carrier density was found to be 2 × 1014 cm2 for the 10 u.c. sample and 2.7 × 1013 cm2 for the 15 u.c. one. Taking the sheet resistance measured previously, we obtained a Hall mobility of 52 cm2 V−1s−1 for the 10 u.c. sample and 210 cm2 V−1s−1 for the 15 u.c. one. In an ideal picture, the interface between SrTiO3 and LaTiO3 can be observed as a Ti ions network, in the 4+ state in SrTiO3 and in the 3+ one in LaTiO3. Therefore, one electron is left every two cells on average at the interface, which corresponds to an areal density of approximately 3 × 1014 cm−2 (refs 8, 9). This is approximatively the electron density measured through Hall effect in the 10 u.c. sample (2 × 1014 cm−2), consistent with the value observed in LaTiO3/SrTiO3 superlattices6 by measuring the number of Ti3+ in the vicinity of the interface. Optical studies confirm that free carriers with densities of approximately 3 × 1014 cm-2 do exist in similar superlattices, with a typical mobility of 35 cm2/V−1s−1 and an effective mass m*≃2me11. The mobility that we measured on the 10 u.c. sample (52 ) is close to this value, which supports an effective mass close to 2me. The 15 u.c. has a lower sheet density of 2.7 × 1013 cm-2 and a Tconset of only 260 mK.

Figure 4: Hall effect.
figure 4

Hall voltage VH divided by the current I as a function of magnetic field for the 10 u. c. (black dots) and 15 u.c. (blue dots) LaTiO3/SrTiO3 samples, measured at 100 mK. Red solid lines correspond to linear fits. (Inset) Tc as a function of doping for SrTiO3 single crystals taken from reference27 (black open circles). The two red dots correspond to LaTiO3/SrTiO3 samples taking the thickness d10 u.c.=12 nm, d15 u.c.=13.5 nm. Here, we have defined Tc as the temperature at which the resistance reaches zero as this definition is more appropriate for comparison with the magnetic definition of the Tc used in reference27. Errors bars represent the uncertainty in the comparison of the two definitions of Tc.

Full size image

Discussion

It is known that LaTiO3 itself can be oxygen20,22,21 or Sr doped23, and thus becomes metallic. The key question is therefore: does superconductivity take place within a doped Mott insulator layer, namely oxygen- or Sr-doped LaTiO3, or within a 2-DEG formed at the LaTiO3/SrTiO3 interface, which extends mostly within the band insulator SrTiO3 (ref. 8)? And conversely, do the electronic correlations, which are known to be strong in the former case and moderate in the latter one16, have a role in that context? The recent works on LaTiO3/SrTiO3 superlattices6,11,24 clearly indicate that under proper growth conditions, the interface is abrupt, with no sizable Sr diffusion for deposition temperatures below 1,000 °C10 Optical spectroscopy confirms that the carrier properties in superlattices are different from the ones in La1−xSrxTiO3 compounds11. Moreover, the low-temperature transport properties are different in conducting doped LaTiO3 and doped SrTiO3. In both cases, electron–electron collisions dominate the scattering events according to the Fermi liquid picture. Figure 1 shows that, in an intermediate regime temperature, the temperature dependence of the resistance is well fitted by a quadratic law R(T)=AT2+R0 where the coefficient A depends on the Landau parameters, and therefore on the carrier density and the effective mass m*25. We obtained A=0.27Ω/□K2 for the 10 u.c. sample and A=0.11Ω/□K2 for the 15 u.c. sample. In Table 1, we summarize the different values of A found in the literature for doped LaTiO3 and doped SrTiO3 thin films and crystals and compare them to the values extracted from our experiment. The largest values of A reported in doped LaTiO3 are on the order of 1×10−9Ωcm/K2 before the system becomes insulating at low temperature (∂R/∂T<0 for T<100 K), whereas it is two orders of magnitude higher for doped SrTiO3. From the comparison, we see that the LaTiO3/SrTiO3 interface layer behave more like doped SrTiO3 than doped LaTiO3. These observations are consistent with an electronic reconstruction of the LaTiO3/SrTiO3 interface, leading to the formation of a few unit cells thick 2-DEG in the SrTiO3 that projects mostly towards the SrTiO3 substrate and makes it conducting layer8,15,16. As shown in Figure 4 inset, our data are consistent with the dependence of Tc with the carrier density of doped SrTiO3 reported in the literature27.

Table 1 Comparison of A parameters (given in Ωcm/K2) measured in the 10 u.c. and 15 u.c. LaTiO3/SrTiO3 samples (taking the thickness d10u.c.=12 nm, d15u.c.=13.5 nm), with data obtained from literature on thin films and crystals for similar doping.
Full size table

In summary, we have measured the electronic transport properties of LaTiO3/SrTiO3 heterostructures. The samples show a metallic behaviour and a superconducting transition is observed at low temperature. Our analysis shows that a 2-DEG is formed at the interface which is located mostly on the SrTiO3 substrate, in agreement with the electronic reconstruction scenario8. This discovery opens the possibility to study the interplay between superconductivity and different electronic orders predicted to take place with ultra-thin LaTiO3 films on SrTiO3. According to our results in terms of carrier density, mobility and gas thickness, it should be possible to modulate significantly the behaviour of the 2-DEG by adjusting the number of charge carriers with an electrostatic gate.

Methods

Growth of LaTiO3/SrTiO3 heterostructures

We have grown epitaxial layers of LaTiO3 using excimer laser-based PLD on commercially available (Crystal) single-crystal substrates of SrTiO3 cut along (100) and (110) crystallographic directions. Although the (100) subtrates were given a buffered HF treatment to expose TiO2-terminated surface, the (110) plane has Sr, Ti and oxygen ions on one plane and hence HF treatment is irrelevant in this case. The substrates were glued to the heater block of the PLD system and heated in oxygen pressure of 200 mtorr in the temperature range of 850–950 °C for 1 h to realize surface reconstruction. This process has been used routinely to grow epitaxial films and heterostructures of YBa2Cu3O6+x and hole-doped manganites. The source of LaTiO3 is a stoichiometric sintered target of 22 mm in diameter that was ablated in oxygen partial pressure of 1×10−4 torr with energy fluence of ~1 J cm−2 per pulse at a repetition rate of 3 Hz to realize a gowth rate of 0.12 Å s−1.

X-ray characterizations

The X-ray diffraction pattern of LaTiO3 films deposited on (100) substrate is shown in Supplementary Figure S1. After subtracting the contribution of the substrate, the lattice parameter of the film is found to be 3.956 Å (inset), in good agreement with previous studies28 and close to the one reported in bulk LaTiO3 (3.928 Å)29. The X-ray diffraction pattern (θ−2θ scan) around 32° of LaTiO3 films deposited on SrTiO3(110) is shown in panel a of Supplementary Figure S2. The (110) peak of the film is observed at 2θ=32.193°, close to the (110) peak of the substrate, which corresponds to a LaTiO3 lattice parameter of 3.928 Å. As shown in panel b of Supplementary Figure S2, the typical width of the rocking curve of the (110) peak is about 0.1° indicating a very good out-of-plane orientation of the layers.

Additional information

How to cite this article: Biscaras, J. et al. Two-dimensional superconductivity at a Mott insulator/band insulator interface LaTiO3/SrTiO3. Nat. Commun. 1:89 doi: 10.1038/ncomms1084 (2010).