Lithium outdiffusion in LiTi2O4 thin films grown by pulsed laser deposition

We report surface chemical cation composition analysis of high quality superconducting LiTi2O4 thin films, grown epitaxially on MgAl2O4 (111) substrates by pulsed laser deposition. The superconducting transition temperature of the films was ~13.8 K. Surface chemical composition is crucial for the formation of a good metal/insulator interface for integrating LiTi2O4 into full-oxide spin-filtering devices in order to minimize the formation of structural defects and increase the spin polarisation efficiency. In consideration of this, we report a detailed angle resolved x-ray photoelectron spectroscopy analysis. Results show Li segregation at the surface of LiTi2O4 films. We attribute this process due to outdiffusion of Li toward the outermost LiTi2O4 layers. & 2016 Published by Elsevier B.V.


Introduction
In recent years, spin-filtering [1] has emerged as a promising way to polarise a charge current by means of a tunnel barrier with a ferromagnetic insulator (FI) and non-magnetic electrodes. Moreover, it has a theoretical capability to provide polarisation (P) up to ∼100%, rather than the limited 50% of conventional magnetic tunnel junctions with ferromagnetic electrodes [2]. The current polarisation occurs due to the difference in barrier heights for the two spin directions, created by the exchange splitting of the conduction band in the FI, at temperatures below the Curie temperature (T Curie ).
Spin-filter devices, working at low temperature, with high polarisation (P¼97%), have been studied extensively and are principally based on Eu-chalcogenide FIs with Curie temperature at the most 69 K [3][4][5] and GdN with a T Curie ¼30 K [6]. With the attempt to obtain spin currents by spin-filtering at temperatures closer to room temperature, interest has also focused on tunnel junctions based on magnetic transition metal oxides, with higher magnetic ordering temperatures, like BiMnO 3 [7], Sm 0.75 Sr 0.25 MnO 3 [8,9] and Pr 0.8 Ca 0.2 Mn 1À y Co y O 3 [10], but these devices showed spin polarisation only at cryogenic temperatures.
To achieve spin-filtering at a technologically useful temperature, one should consider spinel ferrites, such as CoFe 2 O 4 , MnFe 2 O 4 and NiFe 2 O 4 . These oxides, although they have a complex crystal structure and high magnetic-electronic properties sensitivity to structural and chemical defects, have T Curie well above room temperature (∼800 K) [11]. To date, a much lower than predicted polarisation (P o4%) has been demonstrated in CoFe 2 O 4 based tunnel junctions at 290 K [12,13]. The likely reason being the formation of anti-phase boundaries (APBs) in the barrier [14,15] which are detrimental for spin-filter efficiency as they locally change magnetic behavior and barrier height. Such defects are formed due to spinel oxides having a lattice parameter (a ¼0.8392-0.8511 nm) [11] almost double of the metallic layers (Au, Pt, LaNiO 3 , La 2/3 Sr 1/3 MnO 3 ) and substrates conventionally used in spin-filter devices. Achieving high spin-filter efficiency at room temperature is therefore primarily dependent on eliminating structural and chemical defects in ultra-thin ( o5 nm) epitaxial spinel ferrites films to be used in complex oxide barriers.
To reduce defects, the metal-superconductor spinel oxide LiTi 2 O 4 (a¼ 0.8405 nm) [16], closely-lattice matched to spinel ferrites, could be used as non-magnetic electrodes in all-spinel spinfilter devices, grown on spinel MgAl 2 O 4 (a ¼0.8080 nm) [17] substrates. As a result, an epitaxial system with near-perfect interfaces could be achieved, thus APBs and detrimental effects of interface anomalies on the polarisation of one or both the ferromagnetic layers should be minimized to greatly enhance the room temperature spin-filter efficiency.
Since the discovery of superconductivity by Johnston et al. [18] in 1973, LiTi 2 O 4 triggered high interest as it was the first oxide superconductor with a relatively high critical temperature (T c ¼11 K) and it was one of the very few ternary systems to exhibit an elevated critical temperature. LiTi 2 O 4 belongs to the Li 1 þ x Ti 1 À x O 4 ternary system, which also includes the electrical insulating Li 4/3 Ti 5/3 O 4 phase (x¼ ⅓), i.e. for > x 0 a metal to insulator composition-induced transition occurs. Even though sample reproducibility and lack of high quality LiTi 2 O 4 thin films are known issues [19][20][21] that have hampered the understanding of this spinel ternary system, there have been numerous studies on this unique spinel oxide. Recently, high quality epitaxial LiTi 2 O 4 thin films were successfully grown by PLD [22,23], thus opening the door for systematic experiments on LiTi 2 O 4 and for thin films effects investigation. Indeed, results from charge transport and tunneling spectra on LiTi 2 O 4 thin films have shown an anomalous magnetoresistance [24].
In this paper we report angle resolved x-ray photoelectron spectroscopy (AR-XPS) measurements on high quality epitaxial LiTi 2 O 4 thin films. The aim of this work is to investigate the surface chemical cation composition, as high quality surface layers are needed to integrate LiTi 2 O 4 in full-oxide spin-filtering devices.

Experimental details
A Li 4 Ti 5 O 12 ceramic target was prepared from a mixture of Li 2 CO 3 (Alfa-Aesar) and TiO 2 (Alfa-Aesar) by a standard solid-state reaction method [25]. The higher Li/Ti ratio (0.8) of the target was designed to compensate for the high loss of Li during ablation process [26]. LiTi 2 O 4 thin films were grown by PLD (KrF excimer laser, λ¼248 nm) on MgAl 2 O 4 (111) (CrysTec GmbH) substrates in vacuum (base pressure 10 À 6 Torr), with a laser fluence of 0.7 J/cm 2 and a repetition rate of 5 Hz. The substrate temperature was kept at 800°C during growth (unless stated otherwise) and cooled down to room temperature in vacuum after deposition.
Structural analysis was done using x-ray diffraction (XRD, PA-Nalytical high resolution x-ray diffractometer) with monochromatized CuKα 1 radiation (0.154 nm). Deposition rate was determined by measuring the thickness of ultra-thin films by x-ray reflectivity (XRR) analysis allowing then the controlled deposition of 200 nm-thick films. Due to the closely matched densities of MgAl 2 O 4 and LiTi 2 O 4 , XRR was performed on films grown on SrTiO 3 (110) (CrysTec GmbH) substrates. Four-point transport measurements were performed between 300 K and 4.2 K by direct Al-bonding to unpatterned films. Atomic force microscopy (AFM) was used to characterize the surface morphology of the films.
AR-XPS measurements (Monochromatized Al-Kα laboratory source (1486.6 eV), with a SPECS Phoibos 100 analyzer) were carried out to identify the surface cation composition of the thin films. Spectra were calibrated to the binding energy of C 1s at 284.8 eV and were fitted with Voigt functions for Ti 3 þ , Ti 4 þ and Li peaks to reduce the standard residual between the fit and the experimental data, using CasaXPS 2.3.15 software. The same fitting parameters were used for all spectra and a Shirley background was subtracted. For chemical quantification, we used the sensitivity factors 0.0568 and 0.473 for Li 1s and Ti 3s [27], respectively.

Results and discussion
To verify epitaxy and bulk phase purity of the LiTi 2 O 4 films, we measured out-of-plane XRD patterns (see Fig. 1a). Single phase epitaxial LiTi 2 O 4 film growth on MgAl 2 O 4 was confirmed since the spectra showed clear (111) and (222) Bragg's reflection peaks of the film and substrate, with the absence of additional peaks. The average out-of-plane lattice parameter obtained from LiTi 2 O 4 diffraction peaks position is almost identical to bulk value, a ¼ (0.84007 0.0002) nm. The high resolution XRD reciprocal space map (RSM), shown in Fig. 1b, of the substrate (531) peak and the film (531) peak indicates that the latter is fully relaxed in-plane.
XRR measurements show Kiessig fringes indicating smooth film surface and thus, a well controlled growth. A reflectivity scan of a 27 nm-thick LiTi 2 O 4 film is shown in Fig. 1b, together with the simulated fit. AFM imaging of the films surface (Fig. 1d) confirmed an RMS roughness of 0.4 nm.
As reported by Johnston et al. [19], electrical properties of the Li 1 þ x Ti 1 À x O 4 ternary system depend strongly on the stoichiometry (x.) Charge transport characterization on our samples confirmed the high quality thin film growth of the superconducting phase (x 0) Li-Ti-O spinel oxide. Fig. 2 shows the metallic behavior at room temperature - and a sharp normal to superconductor state transition (width o0.3 K) with a T c of 13.8 K (inset). The Fermi liquid behavior of the samples resistivity is confirmed by the variation of resistivity as T 2 at lower temperatures (blue line). The residual resistivity ρ 0 and residual resistivity ratio ρ ρ = RRR / 300 K 15 K of the films were 330 mΩ cm and 2.4, respectively, in accordance with the literature [23,24,28,29].
Both XRD and transport measurements strongly indicate that bulk properties of the LiTi 2 O 4 films are of very high quality. Film surface composition plays a critical role in tunnel devices, thus XPS measurements were carried out in order to determine the surface composition of the deposited LiTi 2 O 4 thin films. Interestingly, the two different spinel phases, LiTi 2 O 4 and Li 4/3 Ti 5/3 O 4 , not only have an evident stoichiometric Li to Ti ratio difference, but also the valence state of Ti differs: the average charge of the titanium ions increases from 3.5 þ (an equally mixed valence Ti 3 þ and Ti 4 þ ) for superconducting LiTi 2 O 4 to 4 þ for insulating Li 4/3 Ti 5/3 O 4 [30]. XPS is hence a very powerful probing method to differentiate between the two oxides and to get a clear understanding of the chemical composition of the film surface layers. Moreover, the probed depth is almost similar for Li 1 s and Ti 3 s as their binding energies are close to each other: 55 eV and 62 eV, respectively. In addition, any spurious effect arising from the analyzer transmission can be neglected so that the Li/Ti and Ti 3 þ /Ti 4 þ intensity ratios can be considered proportional to the ratio between the concentrations. Fig. 3 shows experimental XPS data as well as the fits of Li 1 s and Ti 3 s peaks of the LiTi 2 O 4 film at a bulk sensitive emission angle (θ ¼0°). The Ti 3s peak is clearly split in a Ti 3 þ ( $ 60 eV) and a Ti 4 þ ( $ 63 eV) component, indicating the presence of LiTi 2 O 4 . However, the spectral weight of Ti 3 þ is much less than Ti 4 þ (Ti 4 þ / Ti 3 þ ¼2.3) suggesting that Li 4/3 Ti 5/3 O 4 is present at the surface.
The formation of Li 4/3 Ti 5/3 O 4 is also confirmed by an increased Li/Ti ratio of 0.8 compared to 0.5 for LiTi 2 O 4 . To enhance clarity of the excess surface concentration of Li (green peak area, Fig. 3), the total spectral weight of the Li peak in order the have the correct Li to Ti ratio (0.5), taking into account the total Ti concentration, is depicted in Fig. 3 by the green-hatched area peak.
To quantify the LiTi 2 O 4 cation composition of the surface layers, AR-XPS measurements were performed. Varying the emission angle with respect to the surface normal, allows control of the probing depth, since the effective electron escape depth λ λ θ = cos eff . Fig. 4 shows the Ti 3s, Li 3s spectra and their corresponding total fitsnormalized to the Ti 3s area peaksof LiTi 2 O 4 film collected at various emission angles. By increasing θ (i.e. more surface sensitivity) the intensity of the Li 1s peak increases with respect to Ti, clearly indicating a segregation of Li at the LiTi 2 O 4 surface. Moreover, the Li 1s peak shape broadens towards higher binding energy, indicating that Li þ with a different chemical environment is present at the surface. An increase of the Ti 4 þ component with respect to the Ti 3 þ component with increasing θ was present as well.
We calculated the Li to Ti ratio for the different emission angles using the spectral weight of the Ti 3s peak and the spectral weight of the Li 1s peak. The clear increase of Li for larger θ is shown in Fig. 5a. Since the Li/Ti intensity ratio is larger than 0.5 and not constant with increasing emission angle, it indicates that the top surface layers are disproportionately Li-rich. Ti 4 þ to Ti 3 þ ratios were calculated from the respective components of the fitted Ti 3s envelopes (Fig. 5b). A clear increase of the Ti 4 þ /Ti 3 þ ratio was found at higher emission angles, indicating the presence of Li 4/ 3 Ti 5/3 O 4 at the top surface layers.
To resolve the composition of the surface layers, we modeled the observed Li/Ti and Ti 4 þ /Ti 3 þ ratios versus θ. In our model, we took into account that the intensity of the photoelectrons, created at depth z, has a damping factor λ ( − ) exp z/ eff , where the electron mean free path (λ) is approximately 2 nm [31]. A surface chemical composition of a mixture of the two different Li-Ti-O spinel compositions, LiTi 2 O 4 and Li 4/3 Ti 5/3 O 4 , and of an additional Li phase that mimics the excess of Li at the surface, was hypothesized. Here, the additional Li-phase, modeled by simply using Li þ , may come from LiCO 3 , LiOH or Li 2 O. Note that the presence of Li 2 O at the surface could be ruled out, as this phase is accompanied with an O 1 s peak at 528.6 eV [32] which was absent in our films (not shown here). With this simple model, we found that a good match (solid lines, Fig. 5) to the experimental data was found if the bulk of the film is LiTi 2 O 4 and a thin Li-enriched layer, $ 1.2 nm (1.5 u.c.), is present at the surface. More precisely, a mixture of Li þ and Li 4/3 Ti 5/3 O 4 would be present at the top surface layer and a LiTi 2 O 4 -Li 4/3 Ti 5/3 O 4 mixture towards the bulk. The surface composition is schematically depicted in the inset of Fig. 5b. The presence of a non-spinel Li 1 þ phase at the top surface layer would also clarify the peak broadening of the Li 1s at higher binding energy for increased emission angles.
The Li enrichment at the surface can be understood by considering the increased mobility of Li ions at elevated temperatures as a result of their low activation energy [33]. This leads to an outdiffusion of Li towards the surface that changes the chemical cation compositions of the top LiTi 2 O 4 layers. Similar Li-outdiffusion has been observed in e.g. LiNbO 3 crystals [34,35].
To increase the quality of the surface layers, we investigated whether the Li-outdiffusion may be reduced by lowering the growth temperature, even though previous studies have shown that lowtemperature growth is detrimental for the quality of LiTi 2 O 4 films [29]. By studying the top surface layers of a film grown at 600°C (with a T c of 10.7 K) we found that the Li-excess in the surface layers was significantly increased compared to the samples grown at 800°C. Fig. 6 shows the predominant Li 1s peak with respect to the Ti 3s peak for different collection angles. Here the Li to Ti ratio at θ¼0°was 4.270.2 and increased dramatically to 26.572.7 for θ¼75°. By modeling the data (Fig. 7) using the same procedure used for films grown at 800°C, we found that for low-temperature growth, the thickness of the Li-excess layer doubles to $ 2.5 nm. Moreover, Ti would be absent in the top $ 0.5 nm of the film. The surface composition is schematically depicted in the inset of Fig. 7b.
The increase in Li content at the surface for films grown at lower temperatures is counterintuitive as the outdiffusion may be expected to be reduced. However, previous studies have shown that Li can evaporate from the surface, where the evaporation rate increases with temperature [36,37]. Optimal surface layers are thus obtained when the competing processes of surface segregation and evaporation are balanced, i.e. the thickness of the Li-excess layer has to be kept minimal while keeping the outdiffusion low. Therefore, we have grown LiTi 2 O 4 films in a temperature range of 400-800°C. The found Li/Ti ratios are shown in inset of Fig. 7a. These Li/Ti ratios show that up to 700°C, the Li excess at the surface is extremely high, suggesting that the Li-evaporation is low. Increasing the growth temperature further, a clear drop in the Li/Ti ratio is present, indicating that Li evaporates from the surface. Taking also the bulk properties into account, our data suggest that a growth temperature of 800°C is optimal for both high quality LiTi 2 O 4 surfaces and bulk properties.
With the presence of Li outdiffusion from bulk to the surface and subsequent Li evaporation, Li-deficiency of the bulk LiTi 2 O 4 can be expected. However, since our films grown at 800°C show a very high T c , Li-deficiency in the bulk must be very low. Interestingly, a small Li deficiency could also explain our significantly higher T c (13.8 K) compared to bulk LiTi 2 O 4 (11.2 K), which is in agreement with the observations of Capponi et al. [38].
Even though aging effect [39] arising from surface exposure to   air contamination during the transfer from PLD to XPS may not be completely ruled out, AR-XPS measurements were also performed on a LiTi 2 O 4 film covered by ultra-thin (2 nm) capping layer of CoFe 2 O 4 deposited in-situ by PLD directly after the growth of LiTi 2 O 4 [40]. The corresponding Li 1s to Ti 3s intensity ratio is very similar to that shown in Fig. 5, suggesting that Li segregation might be an intrinsic phenomenon occurring during growth.

Conclusions
We have shown that the growth of metal oxide LiTi 2 O 4 thin films by PLD is affected by Li outdiffusion towards the surface. The thickness of the Li-rich layer was found to vary with growth temperature and it is minimized for samples grown at 800°C, as a result of a balance between Li outdiffusion from bulk and Li evaporation from the surface.
Further analyses are required to ascertain whether the change in chemical composition of the top surface layers affects the formation of APBs, thus hampering formation of a high quality interface with the insulator barrier when the LiTi 2 O 4 films are integrated in spin-filter devices. Work is underway to determine the effect of this layer on tunneling process when LiTi 2 O 4 is used as electrodes in tunnel junctions.