The grain and grain boundary impedance of sol-gel prepared thin layers of yttria stabilized zirconia (YSZ).

Separating grain and grain boundary impedance contributions of ion conducting thin films is a highly non-trivial task. Recently, it could be shown that long, thin, closely spaced, and interdigitally arranged electrodes enabled such a separation on pulsed laser deposited yttria stabilized zirconia (YSZ) thin films. In this contribution, the same approach was used to investigate YSZ layers prepared by the sol-gel route on sapphire substrates. Grain and grain boundary properties were quantified for layers between 28 and 168 nm thickness. Only for the thinnest of the investigated layers, a deviation from macroscopic bulk properties was found, which could be correlated to interfacial strain in the epitaxial layer. A dependence of the preferential orientation on the film thickness was found.

Introduction

During the last years, thin film ion conductors have stirred a lot of interest [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28]. This is partly due to their use as model systems to study material properties only observable in nanometer scaled systems, for which the term “nano-ionics” has been coined. However, also their possible use in several applications such as batteries, fuel cells, and sensors is of high importance. A major focus of research was laid on thin films of yttria stabilized zirconia (YSZ), a well known fast oxygen ion conductor. Regarding the ionic conductivity in YSZ thin films, different results have been published, some claiming increased ionic conductivity [8], [24], [29], [30], while others find the opposite effect [2], [23], [31]. In recent studies, the effect of interfacial strain on ion conduction has been studied in detail [6], [7], [15], [16] and also first theoretical treatments about this topic were published [25], [32]. These findings might very well help to explain previous partly contradicting studies. However, until recently the role of ion conduction in grain boundaries of the mostly polycrystalline thin layers was not clarified. This was particularly due to the fact that the unavoidable stray capacitance of the substrate inhibits straightforward separation of grain and grain boundary contributions by impedance spectroscopy. In Ref. [26] it could be shown that long, thin, closely spaced and interdigitally arranged electrodes may enable quantification of bulk and grain boundary properties also for very thin films. These studies on pulsed laser deposited (PLD) YSZ films also revealed that bulk and grain boundary properties stay unchanged down to very thin layers of about 20 nm [5], [26]. Moreover, it was found that the same electrode geometry allows investigation of conductivity anisotropies of YSZ layers deposited on silicon wafers covered by a thin silica buffer layer Ref. [33]. In the present contribution, we employ the same technique to YSZ layers prepared using the sol–gel route. Grain and grain boundary conductivity are evaluated in the temperature range from 200 to 500 °C and are compared for different layer thicknesses.

Experimental

Preparation of the YSZ layers

The YSZ sol was prepared according to the alkoxide route described in Ref. [34]. The final composition of the sol corresponded to 7.8% YSZ. Acetic acid (100% p.a.), Y(III) nitrate hexahydrate (99+%), and 2-propanol (99.5+%) were obtained from Merck, Germany. Zr(IV) butoxide solution (80 wt.% in 1-butanol) was purchased from Aldrich, USA.

Spin-coating was carried out on single crystalline 1 × 1 cm2 sapphire substrates (0001) (Crystec, Germany) using an SCC-200 spin coater (KLM, Germany) with rotational speeds between 1800 and 12,000 rpm. In one instance (layer E, see Table 1), the sol was diluted 1:1 with a mixture of 10 vol.% HNO3 conc. in 2-propanol. 80 μl of the YSZ sol was pipetted onto the spinning substrates and the rotational speed was maintained for 30 s. The sol-layers were dried at room temperature, then at 100 °C for 2 min each, and afterwards calcined at 500 °C for 10 min. Finally, the layers were sintered for 12 h at 1200 °C with heating and cooling ramps of 10 °C per minute. The preparation parameters of the samples used for this study are given in Table 1.

Table 1

Physical and electrical parameters of the YSZ layers under investigation, as well as the activation energies of macroscopic reference samples for comparison. All thin films were annealed at 1200 °C for 12 h. G.B. means grain boundary, Ea activation energy.

Sample name	Rotational speed[rpm]	Film thickness[nm]	Grain size[nm]	Ea total[eV]	Ea bulk[eV]	Ea G.B.[eV]	G.B. thickness[nm]
Layer A	1800	168 ± 7	–	1.03 ± 0.03	0.92 ± 0.04	–	–
Layer B	3600	100 ± 7	310	1.09 ± 0.01	1.01 ± 0.01	1.11 ± 0.01	2.2 ± 0.1
Layer C	7200	95 ± 4	260	1.09 ± 0.01	0.99 ± 0.00	1.11 ± 0.01	2.4 ± 0.1
Layer D	12,000	63 ± 3	260	1.10 ± 0.01	0.96 ± 0.02	1.13 ± 0.01	2.1 ± 0.1
Layer E	12,000a	28 ± 1	440	1.15 ± 0.01	1.20 ± 0.02	1.14 ± 0.02	3.5 ± 0.6
YSZ polycrystal	–	–	–	–	1.05	1.17	–
YSZ single crystal	–	–	–	–	1.11	–	–

Sol for the preparation of layer E was diluted 1:1 with 10 vol.% HNO3 conc. in 2-propanol.

Structural characterization

X-ray diffraction measurements were performed on an X'Pert PRO system (PANalytical, Germany). An X'Celerator detector (PANalytical, Germany) with a Ni K filter was used in Bragg-Brentano geometry with Cu Kα radiation. Final structural refinements were carried out with TOPAS 4.2 (Bruker, Germany).

Top view images of the layers were made by atomic force microscopy (AFM) using a NanoScope V multimode AFM (Bruker AXS (formerly Veeco), USA). From these images grain sizes were calculated using the EPQ program by IMTRONIC, Germany.

Film thicknesses were determined by sputtering a crater down to the YSZ/sapphire interface, controlled by time of flight secondary ion mass spectrometry (TOF-SIMS, Iontof, Germany). The resulting sputter crater depth was measured by digital holographic microscopy (DHM, Lyncée tec, Switzerland). For one sample (YSZ-5 FT, see Table 1), the thickness was determined by AFM profiling of a pore reaching to the substrate.

Electrical impedance spectroscopy

Impedance spectra were recorded in ambient air using an Alpha-A high-performance frequency analyzer (Novocontrol, Germany) in the frequency range of 1 MHz to 1 Hz with an amplitude of 1 V. Owing to the large number of grain boundaries in the films, this voltage was still in the linear regime but allowed improved data quality. Au/Cr interdigital electrodes of 10 μm (layer A, C, and E) or 5 μm (layer B and D) stripe width and 25 μm stripe distance were employed. Detailed information about the electrode preparation and the impedance measurements including the methode to separate grain and grain boundary contributions is found in Ref. [26]. 9.5% YSZ single crystals (Crystec, Germany) and 8% polycrystals (Tosoh, Japan) were also analyzed as macroscopic reference samples.

Experimental results and discussion

Microstructure of the YSZ films

Film thickness

Five YSZ layers were prepared with different rotational speeds during spin coating, hereafter denoted as A, B, C, D, and E, with A being the one prepared using the slowest speed. The rotational speeds during spin coating and the film thickness of each layer are given in Table 1.

As expected, there is an inverse relationship between rotational speed and YSZ film thickness. Layer E (28 nm), obtained after 50% dilution of the sol, was half as thick as layer D (63 nm) for the same rotational speed.

AFM measurements

The grain sizes, also given in Table 1, were evaluated using the linear intercept method i.e. counting the grain boundaries along straight lines in the AFM images in Fig. 1a) to d) and subsequent computer based statistical evaluation. Two different grain shapes can be distinguished in the AFM images in Fig. 1. Layer B (Fig. 1a)) exhibits dome-shaped and rather symmetric grains of about 300 nm width. In Fig. 1b) and c), showing the thinner layers C and D, also a second grain morphology of asymmetric and flat grains emerges (marked by arrows), hence the grain size given in Table 1 is an average over both morphologies. The thinnest layer E (Fig. 1d) 28 nm thin) consists only of very large and extremely flat grains. Also some pores can be seen which reach down to the substrate (arrow in Fig. 1d)). Unfortunately the AFM measurements of layer A were not suitable for grain size evaluation due to surface contaminations most likely from unremoved electrode material. AFM images of layers B to E were taken before electrode preparation.

Fig. 1

AFM images of the surfaces of a) layer B, b) layer C, c) layer D, and d) layer E. The arrows in b) and c) mark examples of different grain morphology. The arrow in d) marks a pore.

XRD measurements

The XRD patterns in Fig. 2 show magnifications of the YSZ (200) (Fig. 2a) and (111) (Fig. 2b) peaks recorded for all layers. Since the (200) plane and (100) plane are in parallel, we only refer to the latter in the following discussion. Rietveld analysis of the XRD patterns showed an increasing texture parallel to the (100) plane with decreasing film thickness as can be seen in Fig. 2a). The thinnest layer E is strongly oriented in the [100] direction with no other visible reflexes. On the other hand, there is a dominating texture in [111] direction for the thicker layers, see Fig. 2b). Only the thickest layer A is little textured, the grains being randomly orientated like in a sample prepared from powder. The experimental half-width of the high angle reflexes in the XRD pattern for the thinnest layer E could only be explained by including strain. However, for a reliable numeric evaluation more sophisticated XRD measurements are required. Strain values of a very similar system published elsewhere are listed below [35].

Fig. 2

XRD patterns of the YSZ a) (200) and b) (111) reflections of layers A to E.

These results suggest the existence of a strained region at the YSZ|Al2O3 interface of all layers consisting of large epitaxially grown grains, which is in excellent agreement with the data shown in Ref. [36]. This two dimensional large island growth is favored by the virtually defect free surface of the sapphire substrate. In contrast, surfaces with a higher defect density tend to form smaller grains with a (111) orientation [35]. A study modeling this behavior can be found in Ref. [37]. On top of this interface region, a part of the layer is largely textured in the (111) direction and exhibits smaller grains. All material above this second textured region (layer A) is randomly oriented.

Electrical properties of the YSZ films

As shown in Ref. [26], the special electrode configuration used here allows separation of bulk and grain boundary properties via impedance spectroscopy. In Fig. 3 an example of an impedance spectrum is shown, together with a fit result using the equivalent circuit in the inset of Fig. 3b). An Arrhenius diagram of the grain and the grain boundary conductivities is given in Fig. 4b), in Fig. 4a) the total conductivity is plotted for comparison. The grain boundary values are normalized to the electric grain boundary width d obtained from the grain boundary capacitance C:

Fig. 3

Impedance spectrum of layer D at 350 °C in a) Nyquist and b) Modulus plots. The inset in b) depicts the equivalent circuit to fit the data. RBulk and RGB are the resistances for bulk and grain boundaries in YSZ. CPEGB and CPEStray are the constant phase elements used to model the grain boundary and stray capacitance, see Ref. [26].

Fig. 4

Arrhenius plots of a) the total conductivies and b) the grain and grain boundary conductivities measured on layers A to E. The red and green lines are macroscopic bulk measurements plotted for comparison.

In Eq. (1) L denotes the electrode length, h the film thickness, D the spacing between the electrodes and d the lateral grain size. The dielectric constant of the grain boundaries ε is approximated with the value found for bulk YSZ. A detailed discussion can be found in Ref. [26]. The activation energies of these measurements are given in Table 1, along with the electrically relevant width of the grain boundary.

The total conductivities plotted in Fig. 4a) are all, with the exception of layer A, very similar and about one order of magnitude smaller than the bulk conductivity of a macroscopic polycrystal. At lower temperatures, layer A exhibits significantly higher conductivities and a lower activation energy than the other layers, which is most likely related to additional proton conductivity, cf. Refs. [38], [39], [40], [41]. The fact that this feature is solely observed for layer A can be caused by macroscopically visible cracks, which offer an additional adsorption area for ambient water and could lead to proton conducting paths [38]. All other layers exhibited no cracks visible under an optical microscope or in the AFM images.

The bulk conductivities of the layers, shown in Fig. 4b), correlate very well to each other and also to the macroscopic reference samples (except layer A at lower temperatures, see above). The activation energies for ion conduction in the grain bulk of layers A to D (ca. 1.0 eV, for A only the highest temperatures are considered) are in good agreement with the value of our macroscopic polycrystalline material of 1.05 eV or to 1.08 eV as given in Ref. [5].

For layer E a higher bulk activation energy and a lower bulk conductivity for lower temperatures are evaluated which we attribute to the strain found in the XRD measurements. Particularly the higher activation energy cannot be caused by the certain porosity visible in Fig. 1d). The influence of strain on oxygen ion transport in YSZ thin films has recently been dealt with in a series of experimental studies [6], [7], [15], [42]. It was found that a tensile strain in YSZ increases the ionic conductivity in an interface region and decreases the activation energy of ion transport, while a compressive strain has the opposite effect. In Ref. [35] large heterogeneous strains of 6.8% and − 7.5% have been evaluated for the b and c direction of large epitaxial (100) grains on a (0001) sapphire substrate, which is the case for layer E, therefore we expect similar values. However, the beneficial effect of a tensile strain has been calculated to have a maximum at about 4% [25], hence the strain in layer E is expected to have only impeding influence on ion transport. The absolute value of the activation energy for grain bulk transport of about 1.2 eV is in good agreement with the values reported in Ref. [16], where it increases from 1.1 eV for small strains to almost 1.3 eV. However, one has to keep in mind that the latter values include both grain bulk and grain boundary conduction.

The grain boundary conductivities and the respective activation energies of all layers do not differ within the error margin. The absolute conductivities are about two orders of magnitude smaller than the bulk values, which is generally explained by depletion of oxygen vacancies in the space charge regions adjacent to the grain boundaries [43]. The activation energy of oxygen transport through the grain boundaries of about 1.10 eV compares well to 1.15 eV in Ref. [5] (thicker layers) or 1.14 eV in Ref. [26] (thin PLD layers). Hence, all these results are in good agreement with previous studies on thin films and macroscopic crystals, see Refs. [5], [26], [27], [43].

The electrically relevant thickness of the grain boundaries, given in Table 1, was calculated from the grain boundary capacitance and the grain size from AFM measurements, see Ref. [26]. It does not include effects due to a possible variation of the grain-size perpendicular to the surface caused by the transition from (100) to (111) texture. Layers B to D all have a very similar grain boundary thickness of about 2.2 nm for grain sizes of 260 nm (layer C and D) and 310 nm (layer B), which is about half the size usually obtained on macroscopic samples and also on thicker layers [5]. For layer E with a grain size of 440 nm, a larger grain boundary thickness of 3.5 nm is calculated. For comparable grain sizes, values of 5.4 nm – measured on several 100 nm thick layers [5] – and 5 nm [44] or 5.4 nm [45] – measured on macroscopic polycrystals – can be found. However, in contrast to our data on layers B to E all these results were obtained on samples showing no crystallographic texture. In a previous study, we reported a grain boundary thickness of merely 1.0 nm for YSZ layers thinner than 100 nm with a grain size of about 20 nm, which were grown by pulsed laser deposition and were strongly textured in the (111) direction [26]. Hence, the low grain boundary thickness especially of layers B to D might be due to the fact that these layers are also textured to a high degree in the (111) direction. However, in order to verify the assumption that the width of the electrically relevant grain boundary region depends on the orientation between two adjacent grains, further investigations are required.

Conclusions

Polycrystalline YSZ layers of thicknesses between about 30 and 170 nm were prepared on single crystalline sapphire substrates using the sol–gel route. AFM investigations revealed two different grain morphologies, which could be linked to two different textures found in subsequent XRD measurements: Large and flat grains textured in the (100) plane and small dome-shaped grains textured in the (111) plane. The (100) texture becomes more pronounced close to the layer/substrate interface, with the (111) grains growing on top of them. Thicker layers become increasingly untextured close to the surface. For the thinnest of the investigated layers XRD also revealed a strain, which can be correlated with an increased activation energy of oxygen ion transport in the grain bulk. Apart from that, the conductivities of grain bulk and grain boundaries and the corresponding activation energies of all layers hardly differ from each other and from macroscopic reference samples. These results also further demonstrate the advantages of an optimized electrode geometry in separating grain bulk and grain boundary impedance even in very thin layers.