Cation non-stoichiometry in Fe:SrTiO3 thin films and its effect on the electrical conductivity

The interplay of structure, composition and electrical conductivity was investigated for Fe-doped SrTiO3 thin films prepared by pulsed laser deposition. Structural information was obtained by reciprocal space mapping while solution-based inductively-coupled plasma optical emission spectroscopy and positron annihilation lifetime spectroscopy were employed to reveal the cation composition and the predominant point defects of the thin films, respectively. A severe cation non-stoichiometry with Sr vacancies was found in films deposited from stoichiometric targets. The across plane electrical conductivity of such epitaxial films was studied in the temperature range of 250–720 °C by impedance spectroscopy. This revealed a pseudo-intrinsic electronic conductivity despite the substantial Fe acceptor doping, i.e. conductivities being several orders of magnitude lower than expected. Variation of PLD deposition parameters causes some changes of the cation stoichiometry, but the films still have conductivities much lower than expected. Targets with significant Sr excess (in the range of several percent) were employed to improve the cation stoichiometry in the films. The use of 7% Sr-excess targets resulted in near-stoichiometric films with conductivities close to the stoichiometric bulk counterpart. The measurements show that a fine-tuning of the film stoichiometry is required in order to obtain acceptor doped SrTiO3 thin films with bulk-like properties. One can conclude that, although reciprocal space maps give a first hint whether or not cation non-stoichiometry is present, conductivity measurements are more appropriate for assessing SrTiO3 film quality in terms of cation stoichiometry.

The dissolving agent was prepared by mixing deionised water obtained by Barnstead™ Easypure ™ II (18.2 M cm -1 ), the concentrated acids and 1000 mg kg -1 Eu single element standard (Certipure®, Merck, Germany). The final concentration of Eu was adjusted to 1 mg kg -1 in the diluted acid mixture and was used as an internal standard to correct for possible signal drifts. After 30 min of dissolving time, the obtained sample liquid was transferred into a new polypropylene sample tube to remove the remaining substrate and to stop a possible dissolution process of the substrate. This whole process was conducted under ambient conditions at room temperature and the derived sample solutions were measured without any further dilution.
For signal quantification, single element standards (for details see Table 2) were mixed with the diluted acid mixture (3 vol% HNO 3 , 0.3 vol% HF) already containing the internal standard to performe an external calibration. Standard solutions with varying concentration levels from 0.2 to 12.6 mg kg -1 for the main components Sr and Ti and 0.002 to 0.126 mg kg -1 for the Fe dopant were prepared. With the obtained signal intensities, regression lines were derived to calculate the analyte concentration of the unknwon samples.
Samples and standards were analyzed with an iCAP 6500 ICP-OES spectrometer (ThermoFisher Scientific, Bremen, Germany) equipped with a MiraMist nebulizer and a cyclonic spray chamber (Glass Expansion, Port Melbourne, Australia). Sample-uptake was achieved with the peristaltic pump of the instrument (25 rpm, 0.64 mm ID pump tubing). Background-corrected emission signals were recorded in the radial viewing mode and processed using Qtegra software (Thermo Scientific, USA). Six replicates with an integration time of 10 s each were measured for samples as well as standard solutions. The optimized ICP-OES parameters and the monitored emission lines are summarized in Table 1. For each element several intense but non interfered emission lines were measured.
Observed signal intensities were normalized using the signal response for the internal standard (Eu), and finally converted into concentration units by means of the external calibration. By using the molar masses of each element, the mole fractions of the cations present in the investigated samples were calculated. Obtained Eu signals were constant over each measurement session (less than 5% relative standard deviation for the whole measurement period, indicating the absence of temporal trends), and no significant difference in Eu-response between samples and calibration standards was observed. given supporting the validity of this approach. The first argument refers to the electronic resistance R eon . This resistance is essentially the dc resistance and the resulting conductivity σ eon is plotted in The pseudo-intrinsic behaviour of SrTiO 3 thin films was already reported in Ref. [2] for films prepared from 0.4 % Fe doped stoichiometric targets. Interestingly, the rather pronounced high frequency shoulder found in our study on 2 % Fe doped films was not present for the 0.4 % Fe doped films discussed in Ref. [2,3]. In order to exclude the relevance of artefacts and to further understand the different shape of the spectra, we also prepared 0.4 % Fe doped films here (1.1 J/cm 2 , 5 Hz, investigated with La 0.6 Sr 0.4 CoO 3-δ microelectrodes prepared as described in Ref. [4].) All results shown in the earlier studies were excellently reproduced for these 0.  the electronic intrinsic conductivity [5] as well as with literature data [2], showing great agreement to both.
The more or less distorted arc with a high frequency intercept found here and in Ref. [2] for 0.4 % Fe doping can be fitted by a serial circuit of a R followed by one or two R-CPE (CPE = constant phase element). The serial R is a contact resistance and is not further considered. For main arcs with little distortion, one R-CPE element is sufficient and the corresponding capacitance then corresponds very well to the expected bulk capacitance of the thin film. However, not surprisingly, also a fit is possible with the transmission line model suggested in the main paper and representing the mixed conducting character of the SrTiO 3 films. Based on this model the small high frequency arc for 2 % Fe doped films indicates a total conductivity which is much higher than the pseudointrinsic electronic conductivity due to parallel ion conduction. The absence of the high frequency arc for 0.4 % Fe, on the other hand, thus suggests that either the ionic conductivity is very low and does not strongly enhance the total conductivity, or the corresponding chemical capacitance in the transition line model becomes too small to allow a separation of the arcs, or both.
We think that the appearance of some distortion in SI 6 Discussion of the capacitances found for the thin films deposited from

Sr overstoichiometric targets
As mentioned in the main paper, the high frequency arc in the impedance spectra found for thin films deposited from Sr overstoichiometric targets is attributed to the SrTiO 3 bulk due to the match with expected geometrical capacitances. The capacitances of the two other R-CPE circuits fit pretty well to space charge layers; for bulk SrTiO 3 with 2 % Fe, a relative permittivity of 150 and a space charge potential of 600 mV, for example, ca. 6 nm result for one space charge [6]. It is thus reasonable to assume that the corresponding capacitances are due to the space charge layers at the two electrodes, with the Nb:SrTiO 3 /Fe:SrTiO 3 junction probably having the smaller capacitance (due to two similarly thick space charge zones in both SrTiO 3 parts). Hence, the large low frequency arc would be the space charge between SrTiO 3 film and Pt. From the measured peak frequencies, we can estimate the space charge potential and a rather realistic value of ca. 700 mV is found [6]. Accordingly, the dc resistance does no longer reflect the electronic conductivity.

SI 7 Characterization of samples using optical microscopy after electrical measurements
From previous studies, we have atomic force microscopy (AFM) images and transmission electron microscopy (TEM) images available [2]. Using optical microscopy, we checked each sample after electrical characterization to make sure no changes in surface structure or crack formation happened during impedance measurement. Fig. SI 4 shows microscopy images of a typical sample after impedance measurements. Damages to the microelectrode after removing the Pt needle are indicated with arrows. Furthermore, we also show microelectrodes before and after measurements for thin films deposited from targets with 0 %, 5 %, 7 % and 11 % Sr excess in     The impedance spectra were measured at 327 °C. Fig. SI 10 shows the quality of the impedance measurements for the thin films deposited from bulk-like targets.  CPE-T2 5.6*10 -9 ± 1*10 -11 3.2 *10 -10 5.7 6*10 -9 -3*10 -9 4.8*10 -10 -1.4*10 -10 4 -7