New features from transparent thin films of EuTiO3

ABSTRACT The almost multiferroic perovskite EuTiO3 (ETO) has been prepared as films on substrates of SrTiO3. For all prepared film thicknesses highly transparent insulating films with atomically flat surfaces and excellent orientation have been grown. They were characterized by X-ray diffraction, magnetic susceptibility and birefringence measurements and found to exhibit bulk properties, namely an antiferromagnetic transition at TN = 5.1 K and a structural transition at TS = 282 K. The latter could only be identified due to the high transparency of the samples since the optical band gap is of the order of 4.5 eV and larger than observed before for any bulk and thin film samples.

during vacuum annealing, and has been removed by post-annealing the samples. The major problem in thin film preparation relates, however, to the above-mentioned leakage currents which render it impossible to measure the polarization or permittivity directly. Also, there is no report, so far, on the unambiguous observation of the structural phase transition of ETO in thin films.

Experimental
For these reasons we have prepared thin films of ETO on substrates of SrTiO 3 . Epitaxial EuTiO 3 (ETO) single crystalline films were synthesized by using pulsed laser deposition (PLD) process from the ETO polycrystalline target. The excellent quality of the films is not only a result of the careful PLD ablation process but also due to the target quality. The ETO target used here differs from others since the ETO samples were synthesized by repeatedly heating mixtures of Ti 2 O 3 and Eu 2 O 3 at 1300 C with intimate grinding in between, which ensures optimal target properties. [8] The films were grown on SrTiO 3 [001] (STO) single crystal substrates provided by Crystec Company (Germany). For the PLD ablation process a KrF excimer laser with a wave length of 248 nm was used. The energy density on the target was »1.6 J/cm 2 and the frequency of the pulsed laser beam was 10 Hz. The deposition rate was »0.257 A /pulse and was calibrated by measuring the film thickness with a profilometer. The substrate was heated by a resistive heater and kept constant during the film growth at 600 C according to the radiation pyrometer reading. During the film deposition we maintained an oxygen flow with a flow rate 4.4 sccm, while the pressure in the deposition chamber was 1 £ 10 ¡5 mbar.
Films of various thicknesses (50, 100, 257, 500, and 1000 nm) were grown always on STO substrates, however, also with varying substrate thicknesses (1À0.1 mm) and were always highly transparent (Figure 1(a)) with resistivities larger than 100 MV. They were characterized by scanning electron microscopy (SEM) (Figure 1(b)) showing a smooth and homogenous surface, atomic force microscopy (AFM) (Figure 1(c)) evidencing a surface roughness of less than 0.25 nm and X-ray diffraction (Cu Ka 1 radiation) (Figure 1(d)). It is remarkable that the widths of reflections of our samples are very narrow and free from fringes most likely due to the rather large sample thickness. No features could be resolved which would be consistent with a c-axis shrinkage or expansion. Marginal differences between the lattice constants of ETO and STO appear upon blowing up the data and are consistent with the reported lattice parameters. From these data it is concluded that all films are cubic at room temperature and strain free.

Results
The low-temperature magnetic properties of thin films investigated so far differ significantly from bulk materials reporting ferromagnetism below 5 K in c-axis expanded samples.
[16À20] In order to characterize the low-temperature magnetic properties of our films, the temperature dependence of the magnetic moment m was measured using a SQUID magnetometer (Quantum Design MPMS-XL). The data have been corrected for the diamagnetic background stemming from the STO substrate by measuring the substrate separately and subtracting these data from the ones of substrate plus film. The magnetic susceptibility was measured in two modes, namely in plane and perpendicular to the film. In both orientations an anomaly at T N D 5.1 K was observed consistent with antiferromagnetic order and with previously obtained results,[16À20] however, slightly smaller than observed in our best ceramics with T N D 5.7 K. [8] For the in-plane measurements the anomaly at T N was less well resolved, as compared to the perpendicular mode configuration. Typical data corrected for the substrate in comparison to results obtained for bulk ceramic samples are shown in Figure 2.
We used spectroscopic ellipsometry to determine the optical band gap in ETO. Room temperature ellipsometric spectra in the spectral range from 0.6 to 6.5 eV were measured on a »500-nmthick ETO film at several angles of incidence (63 À75 ) with a J.A. Woollam Co. VASE variable angle spectroscopic ellipsometer. A numerical regression procedure [21,22] based on a three-layer optical model (consisting of 0.1-mm-thick STO substrate, ETO film and surface roughness layers) has been employed to extract the complex dielectric function of the ETO film. The thickness of the film and surface roughness layers (the letter was treated as a 50% bulk ETO film and 50% void mixture) were determined to be 482 § 2 nm and 5.1 § 0.4 nm, respectively. Figure   Another characteristic of ETO is its structural phase transition from cubic to tetragonal at T S D 282 K [8] which has been debated in the literature, where values of T S ranging from 130 to 310 K have been reported early after its discovery.
[25À28] By improving the sample quality T S has been limited to a narrow region around 282 K § 8 K.
[29À32] The detection of T S is similarly difficult, as in STO, [33] since at low temperature the c/a ratio (1.0014) slightly deviates from one. [34] Specific heat, thermal expansion, [8,30] resonant ultrasound (RUS) [31,32] and muon spin rotation (mSR) measurements [35,36] all reveal clear anomalies related to T S . Unfortunately these methods are hampered for our thin film samples due to the large background signal from the substrate. However, the high transparency of our samples admits to perform birefringence measurements which have already been successfully carried out on STO to identify its structural phase transition at T S D 105 K. [37,38] These measurements together with the large resistivity and spectroscopic ellipsometry imply a band gap which is substantially larger than that of STO.
Birefringence measurements were made on an ETO film with thickness of 1 mm oriented in [100] direction and deposited on a single STO crystal substrate. The film and substrate together were of 85 mm thickness. The sample was placed in a high-precision Linkam TMSG600 temperature stage    The inset shows the same, however, for bulk ceramic samples. [8] combined with the birefringence imaging system. [37] The temperature was controlled to within §0.1 K, and the measurements were made with cooling and heating rates of 0.5 K min ¡1 . Prior to performing the experiment, the sample was rejuvenated at a temperature of 470 K for half an hour. The Metripol Birefringence Imaging System (Oxford Cryosystems) consists of a polarizing microscope equipped with a computer-controlled plane-polarizer capable of being rotated to fixed angles a from a reference position, a circularly polarizing analyzer and a CCD camera. The intensity measured at any position is given by: where I 0 is the intensity of light that passes through the sample and represents its transmittance, F is the angle of an axis of the optical indicatrix of the specimen projected onto the image measured from a pre-determined direction, i.e. from the horizontal axis, and d is the phase difference between the polarized light components, and is given by: where λ is the wavelength of the light, d is the thickness of the thin film and Dn D ðn 1 À n 2 Þ is socalled plano-birefringence, i.e. the birefringence measured as seen in projection down the microscope axis. In all experiments the wavelength of 570 nm was used. Below the phase transition at T S D 282 K from the cubic Pm-3m to tetragonal I4/mcm, the ETO film is in phase possessing a single optic axis. We have verified that the ETO film was oriented such that its z-axis, i.e.
[001] direction of the tetragonal unit cell, was in plane of the surface of the substrate, and the main axis of the indicatrix was oriented 0 or 90 degree to the predetermined horizontal axis mentioned above. Hence, the measuring light beam passing through the ETO/STO sample was not parallel to the optic axis ([001] direction for the tetragonal symmetry), and the light was split into the ordinary and extraordinary ray. This set-up enabled us to observe the birefringence Dn. Since absolute values of Dn are determined by this method, the two orientations of the optic indicatrix were not an obstacle to obtain the temperature dependence of the birefringence. Because the film was of [100] orientation, the measuring light beam passing through the ETO/STO sample was not parallel to the optic axis (which is of [001] direction for the tetragonal symmetry), then the light was split into the ordinary and extraordinary ray, and thus the birefringence Dn could be observed.
By measuring several images with varying angle a, it is possible to fit for each pixel position the quantities I 0 , sin d and ' separately and then to plot images in false color representing these three values. In this way the birefringence can be detected with a very high sensitivity of the order of 10 ¡6 . One of the interesting features of the Metripol system is the possibility to separate the birefringence of the sample from the effective retardation, which is a sum of true sample birefringence and the one stemming from the background. This effective retardation can be calculated according to the formula [37]: where d m D sin À1 ðjsind m jÞ and ' m are the phase shift and orientation angle measured in the ETO/ STO sample, respectively, and d b D sin À1 ðjsind b jÞ and ' b are the phase shift and orientation angle measured as background values in the region outside the ETO/STO sample, respectively. Figure 4 shows the temperature changes of the birefringence of ETO upon cooling and heating. These data are representative for tens of regions chosen for calculations of birefringence. Sizes of these regions varied from 50 £ 50 mm 2 to 500 £ 500 mm 2 . Note that the birefringence of the STO substrate disappears above 150 K. [38] Up to 150 K the birefringence in STO is clearly temperature dependent and thus may produce local (non-homogenous) strains acting on the ETO film. For this reason we did not cool the sample far below 150 K. [38] In order to exclude any errors, birefringence experiments have been performed on the STO substrate only. These experiments on a 100-mm-thick STO single crystal confirmed that a non-zero birefringence, of the order of 2 £ 10 ¡6 , was observed in the temperature range from T S D 105 K to approximately 160 K, while above 160 K no birefringence could be revealed.
In the investigated ETO/STO sample we did observe domain structure which could have an effect on the Dn values. This implies that we cannot exclude a domain structure which might evolve perpendicularly to the sample surface, namely, along the direction in which the light passes through the sample. In this case the measured value of Dn is an 'effective' one.
As is shown in Figure 4, above 190 K the linear Dn(T) cooling and heating run in fact does not evidence a thermal hysteresis. Below 190 K Dn(T) is nonlinear and shows a thermal hysteresis which may reflect the influence of the STO substrate caused by strain.  Another evidence for the phase transition in ETO at 282 K is obtained from the temperature dependence of the discrepancy DF (after background subtraction) of the angle between the main axis of the optical indicatrix and the predetermined horizontal axis ( Figure 5, right). DF clearly changes around 282 K ( Figure 5, left), and before this transition temperature the precursor effect is distinctly observed. The temperature range of its existence has been calculated as DT D 1.1 Ã T S ¡ T S D 28.2 K. The origin of this precursor has been addressed [43] and is analogous to similar effects observed in other perovskites. [38] Between 190 and 282 K the data can be well described by conventional Landau theory, where Dn % (T S ¡ T) 2b with b D 0.5. This is analogous to STO, [37,38] but unlike STO the tetragonal domains are not observable due to domain wall pinning or freezing, consistent with RUS data. [31,32,44]

Conclusions
We conclude that our ETO films are of excellent quality because they are highly transparent with a band gap of 4.53 eV with AFM ordering below 5.1 K and T S D 282 K. The band gap of 4.53 eV ( Figure 3) is much larger than ever reported before.
[45À49] Electronic structure calculations for ETO arrive at a band gap of 0.77 eV [47,48] which increases rapidly when Ti is replaced by Zr or Hf where in both latter compounds the gap is distinctly larger than 2eV. This large variation of the band gap in this series has been attributed to the transition metal d energy levels of the Ti 3d, Zr 4d and Hf 5d states which constitute the conduction bands and move to higher energy values with increasing principal quantum number. This result leads to a variety of inconsistencies with experimental data, as the nearest and next nearest neighbor exchange constants derived for this series cannot be explained with the band gap values. Also, the huge jump of the gap energy from the Ti compound to the Zr one [50] (1.6 eV) is difficult to explain by the increase in the principal quantum number. [51] Our new determination of the band gap of 4.53 eV removes the inconsistencies for the exchange constants mentioned above.
To summarize, we have successfully grown transparent ETO films on STO substrates of outstanding quality. The characterization of the films has evidenced that independent of the film thickness the samples become antiferromagnetic at T N D 5.1 K. The transparency of all films together with spectroscopic ellipsometry data demonstrates that the band gap is much larger (4.53 eV) than ever reported. This is attributed to the absence of Ti 3C defect states which are present in previously investigated bulk and film samples. In addition, it allowed to detect the structural phase transition in thin films at T S D 282 K by means of birefringence measurement with similar features as those observed in STO single crystals.

Acknowledgement
It is a pleasure to acknowledge Viola Duppel for recording the SEM image and Pirmin Ganter for providing the AFM image. Micha» G orny passed away on January 18th, 2016. He was a high class specialist in the electronics and mechanics, and extremely talented experimentalist. It was a great pleasure to me to co-operate with him, and first of all to receive his help while modifying or constructing experimental setups. I was deeply impressed with his modest way of life and his readiness to help in laboratory work at any time he was asked for. The presence of his name in this paper is connected with his important contribution in modification of low temperature setup and with a very special tribute he deserves from my side (K.R.).

Disclosure statement
No potential conflict of interest was reported by the authors.