EELS analysis of cation intermixing around LaAlO3/SrTiO3 interfaces

Abstract

The pioneering work by Ohtomo and Hwang1 reported the formation of an electron gas with a large charge carrier density at the interface between two band insulators, LaAlO 3 (LAO) film on SrTiO 3 (STO) substrate. However the mechanisms of charge transfer and transport in this system are still not clearly established. Epitaxial films with 3 and 5 u.c. thickness were grown by Pulse Laser Deposition. The oxygen partial pressure during deposition was fixed at 10 ‐4 Torr, and the temperature reached 750°C. The parameter misfit between the substrate (aSTO = 3.905 A) and the film (aLAO = 3.791 A using a pseudo cubic description) did not induce strain relaxations via interfacial misfit dislocations as no dislocations could be detected along the foil observed for both samples, as seen on Figure 1. The 5 u.c. sample exhibited a conductive interface while for the 3 u.c. film the resistance exceeded our instrumental limits (> 100 MΩ). Thus, consistently with previous studies, the critical thickness lies between 3 and 5u.c. These samples are good candidates to investigate structural and/or chemical differences between conductive and insulating samples. EELS profiles across the interface of the selected samples were used to deduce the contribution of Ti 3+ to the Ti‐L 2,3 absorption edges. A minimum valence of Ti 3.9+ (+/‐ 0.05) was found located in the first unit cell below the interface of both samples (Figure 2). This would lead to a maximum theoretical density of free charge carriers of 6.6x10 13 (+/‐ 3.28x10 13 ) cm ‐2 if we assume that all the carriers originate from Ti 4+ reduction. Experimental measurements of Hall coefficient on the 5 u.c. sample below 10 K revealed a 2D charge carrier density (≈ 3x10 14 cm ‐2 ) that was comparable to theoretical density (3.3x10 14 cm ‐2 ). However, 2D charge carrier density at room temperature (n > 1.2x10 15 cm ‐2 ), was much higher than the density calculated based on EELS valence measurements. This suggests that the conduction was not purely bidimensional. The hypothesis of a quasi 2D conduction zone restricted to the first layers above and below the interface, still underestimates the charge carrier density with respect to the Hall measurements This would confirm the 3D nature of the conducting layer. At the partial pressure of 10 ‐4 Torr used during the PLD growth, no signature of oxygen vacancies could be detected in the O‐K edge recorded in the substrate and around the interface by EELS, as observed on Figure 3. The interfacial O‐K EELS spectra reflect intermixing rather than oxygen vacancies. Although a low level of oxygen non‐stoichiometry is not excluded, it would be insufficient to explain the sheet resistance measured and the differences between the 3 and 5 u.c. samples. The full presentation will combine this analysis with elemental profiles and strain analysis obtained by Medium‐Energy Ion Scattering (MEIS), and additional electrical measurements to give a rather complete description of these films. Neither electronic reconstruction nor anionic vacancies alone can explain the carrier density observed. Intermixing is demonstrated in the two samples, excluding a donor doping scenario as single mechanism. The measured c/a ratio are larger than those predicted by epitaxial strains obtained from an elastic calculation taking intermixing into account. This indicates that compressive electrostatic forces developed around the interface, and extended deeper into the substrate in the 3 u.c. sample, reducing the confinement and diluting the interfacial charge carrier. A complex competition between donor doping, structural distortions and reconstruction, and ionic compensation is revealed.