HETERO-INTERFACES FOR EXTREME ELECTRONIC ENVIRONMENTS
Alp Sehirlioglu (Case Western Reserve University)
The quasi-two-dimensional electron gas (Q-2D-EG) that forms at the interface between
two perovskite band insulators LaAlO3 (LAO) and SrTiO3 (STO) has stimulated extensive research
interest since its discovery in 2004. It has been shown that electrical properties of the interface
strongly depend on the film thickness. The insulating interface becomes conductive when the
film thickness exceeds a critical thickness (3 unit cell for LAO on STO). This conductive interface
is two-dimensional in characteristics. The physical origins of the Q-2D-EG formed at the
interface have been under intensive debate to date. Several mechanisms have been proposed,
such as the polar catastrophe at the polar LAO/non-polar STO interface, structural distortions at
the interface, oxygen vacancies introduced into the LAO/STO hetero-structure during the
growth of LAO, preferential cationic intermixing at the interface.
In 2006, Thiel et al. showed that the interface conductivity was tunable when the film
thickness was just below the critical thickness. A conductive atomic force microscopy (AFM) tip
was used to apply a positive voltage along the thickness of the film which created a conductive
region underneath the tip. This technique allowed writing of conducting wires in between
electrodes. Application of a negative voltage converted the interface underneath back into an
insulator. The tunability of the interface opened the path for a large number of applications;
the focus mostly has been on transistors. The exact origins of the tunability is also still under
intensive debate.
The potential advantages of oxide based hetero-interfaces for extreme environments are:
1- Capability to have high density non-volatile memory.
2- Incorporation of ultra-thin high-K dielectric (dielectric constant, K=25) film that
eliminates the need for a gate dielectric.
3- Insulating film and the substrate increases radiation hardening
4- Possibility of developing multi-state transistors
This technologically significant but still infant discovery holds great potential for nextgeneration electronics that can have both (i) higher information density and (ii) larger operation
domain.
Large number of parameters can affect the observed properties: (i) Substrate quality, (ii)
film composition, (iii) defects, (iv) strain development, (v) film thickness, (vi) electrode
materials, (vii) film surface conditions, (viii) in plane anisotropy. All these factors can affect
either the existence of the interface conductivity or just the magnitude of it. However, all have
different dependence on temperature. Therefore any extreme environment application
requires quantitative analysis of these parameters.
Fig. 1: φ-ω plots (left) for two STO substrates from two different vendors. Top crystal is of higher quality with
no observable domains. (Right) intensity- ω plots extracted from φ-ω plots at around φ =100o, showing the
domains clearly for the low quality crystal at the bottom.
Our investigations showed that the
quality of the substrates can be widely
different from company to company. Figure
1 shows φ-ω plots of two different STO
substrates of the same size from two
different companies. The data reveals that
the STO from company II has a large number
of domains. These domains can affect the
quality of the epitaxial film. The curvature of
the data is due to mis-cut angle and its value
can be calculated from XRD measurements.
The mis-cut angle leads to formation of
terraces on the surface of etched substrates,
the width of which increases with
decreasing mis-cut angle.
Etching is carried out to obtain a Titerminated surface that is required for the
formation of Q-2D-EG. The Q-2D-EG was
only observed when films were deposited
on Ti-terminated <001> oriented STO
crystals. The <001>-orientation provides AO
– BO2 stacking in perovskite phase (Fig. 2).
For example, a SrTiO3/LaAlO3 interface
normal to <001> produces a charge-
Fig. 2: a) At the AlO2/LaO/TiO2 interface half an electron is
added to the last Ti layer. This produces an interface dipole
that causes the electric field (E) to oscillate about 0 and the
potential (V) remains finite. The upper free surface is not
shown, but in this simple model the uppermost AlO2 layer
would be missing half an electron, which would bring the
electric field and potential back to zero at the upper
surface. The actual surface reconstruction is more
complicated b) The AlO2/SrO/TiO2 interface has half an
electron removed from the SrO plane in the form of oxygen
vacancies.
balanced layer of SrTiO3 – (SrO)0 or (TiO2)0 – intersecting with negatively charged (AlO2)1- or
positively charged (LaO)1+ layers, respectively. (this latter "LaO" layer should not be confused
with "LAO", the abbreviation for LaAlO3). The basic mechanism proposed on the origin of the Q2D-EG depends on the polarization discontinuity at the interface. This discontinuity results in
formation of a dipole between (AlO2)1-and (LaO)1+ layers; the electrostatic potential increases
with increasing thickness which would result in polar catastrophe. The potential is low when
the films are very thin (1-3 unit-cell thick) and the lattice polarization screening allows interface
to be insulating. However, the system is not sustainable for thicker films (>3 unit-cell). As the
electrostatic voltage increases, the valence band maximum (VBM) at the surface becomes equal
to the conduction band minimum (CBM) at the interface, resulting in charge flow from surface
to the interface. On a TiO2-terminated substrate surface AlO2/LaO/TiO2 hetero-interface forms;
the charge flow from the film surface starts filling up the conduction band Ti 3d states. The
polar catastrophe can be eliminated with addition of an electron per two unit cells by changing
the valence state of Ti ion from Ti4+to Ti3+. On average at the interface Ti4+ becomes Ti3.5+.
Longer etching times or lower pHs of the buffered Hydro-fluoric solutions (BHF) can
yield etching pits in the surface. Literature reveals a range of suggested pHs for surface
treatment, however, our results showed the best pH for etch pit free surfaces is 6.0 (Fig 3).
Pulse laser deposition is an effective technique of depositing multicomponent systems.
In order to obtain epitaxial films of multicomponent systems, ablation plume must consist of
primarily atomic or diatomic species that can be achieved through use of UV laser wavelengths
with pulse length is nanoseconds range. Even though for ceramics the composition of the film
generally is similar to the target, variations still can be observed. Our work indeed showed a
La/Al>1 in the film. The composition is also dependent on the angular relationship between the
plume and the substrate; La/Al ratio as high as 1.6 can be observed. In addition, the La/Al ratio
was found to be dependent on the target material. A commercial LaAlO3 target and a target
processed in our labs resulted in different La/Al ratios in the film, despite having the same
Fig. 3: (Left) AFM topography image of a 5x5 micron region of a Ti-terminated STO substrate prepared by the PI. At
pH =6 for the BHF solution a surface with no visible etch pits was achieved. (Middle-left) A higher magnification AFM
image for the same substrate showing the terraces that form due to the mis-cut angle. (Middle-right) Section
analysis of the same surface showing the step size is just below 4Å, indicating unit cell size steps which would be
expected from a Ti-terminated surface. (Right) AFM image of a Ti-terminated STO prepared by the PI with BHF
solution at pH = 5.7. Etch pits are visible on the surface compared with a etch-pit free surface.
phase and being deposited under the same processing conditions. The variations in La/Al ratio
can affect the electrical properties directly or indirectly through changes in structure and strain.
Even though their effect might not be catastrophic to the presence of Q-2D-EG it can affect the
quantitative analysis and modeling.
The interface properties are strongly influenced by the strain which exists in cube-oncube growth systems due to lattice mismatch. The epitaxially grown LAO film is strained in
LAO/STO heterostructure (≈3% lattice mismatch). The existence of strain field inevitably affects
measured electrical properties near the interface, from which the formation of Q-2D-EG at the
interface is evidenced. The presence of a critical thickness also indicates that strain in the
hetero-epitaxial LAO/STO system may have an effect on the formation of Q-2D-EG. It is also
worthwhile pointing out that interfacial strain in a perovskite heterostructure has been shown
to alter its physical properties. For example, thin film STO can exhibit strain induced
ferroelectricity due to constraints of the substrate. Tensile strain introduced by depositing the
STO on substrates with lattice mismatch was shown to eliminate the conductivity at the
interface, while increasing compressive stress did not destroy the conductivity but decreased
the mobility.
Reciprocal lattice mapping was conducted to measure the stress development,
interfacial strain and relaxation rate both in plate and out of plate. In addition, we investigated
possible unit cell tilting and determined that it does not occur in these systems. Our results
showed that these hetero-interfaces exhibited very low level of relaxation (<4%) with increasing
thickness up to 60 nm. Above 60 nm, the films started to peel off due to increased stress;
however the remaining parts of the film on the surface were still strained with little relaxation.
Strain development can be influenced by a wide range of factors such as compositional nonstoichiometry, lattice mismatch, thermal expansion mismatch and phase transformations.
(La1-x,Ndx)AlO3 targets with x=0, 0.1, 0.3 and 0.5 were prepared to investigate the strain
development at constant film thickness. The La/Nd ratio was verified in the films with ICP and
the correct perovskite phase and the changing lattice parameters where identified with XRD.
Quantitative XPS on the films showed that the correct La/Nd ratio was not obtained in the films
and especially for x=0.5 the film composition was widely varied from the target composition.
The presentation will show the effects of crystal quality, surface preparation, target
material processing and target material composition on the film compositions and the strain
development on the film. In addition, the relaxation processes will be presented as a function
of thickness for a given system. Electrical properties as a function of oxidation conditions will be
introduced as well as preliminary observations on thermal conductivity across the interface.
The last item is investigated in order to determine if there is any interaction between the
electrons traveling along the interface and the phonons traveling across the interface.