Colossal Ionic Conductivity at Interfaces of Epitaxial ZrO2:Y2O3/SrTiO3 Heterostructures

See allHide authors and affiliations

Science  01 Aug 2008:
Vol. 321, Issue 5889, pp. 676-680
DOI: 10.1126/science.1156393


The search for electrolyte materials with high oxygen conductivities is a key step toward reducing the operation temperature of fuel cells, which is currently above 700°C. We report a high lateral ionic conductivity, showing up to eight orders of magnitude enhancement near room temperature, in yttria-stabilized zirconia (YSZ)/strontium titanate epitaxial heterostructures. The enhancement of the conductivity is observed, along with a YSZ layer thickness–independent conductance, showing that it is an interface process. We propose that the atomic reconstruction at the interface between highly dissimilar structures (such as fluorite and perovskite) provides both a large number of carriers and a high-mobility plane, yielding colossal values of the ionic conductivity.

Solid oxide fuel cells (SOFCs) have emerged as a promising nonpolluting technology for the short-to-medium–term substitution of fossil fuels (14). The conversion efficiency of chemical into electrical energy is limited by the transport of oxygen anions through an electrolyte material. So far, yttria-stabilized zirconia (Y2O3)x(ZrO2)1–x (YSZ) is the material mostly used in SOFCs because of its mechanical stability, chemical compatibility with electrodes, and high oxygen ionic conductivity. It is well known that doping ZrO2 with Y2O3 stabilizes the cubic fluorite structure of ZrO2 at room temperature and supplies the oxygen vacancies responsible for the ionic conduction, resulting in high values of the oxygen conductivity at high temperatures (57). A maximum value of 0.1 S/cm (where 1 S = 1 A/V) at 1000°C is observed for the 8 to 9 mole percent (mol %) yttria content (24). A severe drawback toward the final implementation of SOFCs is the relatively low room temperature ionic conductivity of this material, which imposes rather high operational temperatures around 800°C (14). The search for alternative electrolytes has not yet been successful in reaching the conductivity value of 0.01 S/cm desired for room temperature operation (14).

Only modest reductions in the operation temperature of SOFCs (500° to 700°C) can be anticipated with the recently proposed optimized electrolytes such as gadolinia-doped ceria and lanthanum gallates (811). On the other hand, the one to two orders of magnitude increase of the electrical conductivity reported (1214) in nano-crystalline samples as compared with single crystals outlines the importance of processing as an alternative route to increasing conductivity values toward the desired levels. Because modern thin film growth techniques allow a precise control of layer thickness and morphology, they provide a pathway for the production of solid electrolytes with optimized properties. Maier et al. found a substantial increase of the dc ionic conductivity of superlattices of CaF2 and BaF2 when the thickness of the individual layers was decreased down to 16 nm, assigned to a size effect due to the space charge regions being smaller than the layer thickness (15, 16). Kosacki et al. have reported enhanced conductivity in highly textured thin films of YSZ with thicknesses between 60 and 15 nm, reaching 0.6 S/cm at 800°C (17). Because reducing film thickness (and therefore increasing the fraction of material near the interface) produces such a noticeable conductivity enhancement, the interfaces themselves would seem to play a determining role in the outstanding conductivity properties observed.

To search for interface effects, we fabricated heterostructures where YSZ layers (with 8 mol % nominal yttria content) in the thickness range from 62 nm down to 1 nm were sandwiched between two 10-nm-thick layers of insulating SrTiO3 (STO). Also, superlattices were grown, alternating 10-nm-thick STO films with YSZ layers with thickness between 62 and 1 nm (18). Figure 1A displays a low-magnification (inset) and a high-resolution annular dark field (or Z-contrast) image of a [YSZ1nm/STO10nm]9 superlattice (with nine repeats), showing the excellent crystalline quality of the sample. The layers appear continuous and flat over long lateral distances (a few microns). The interfaces between the STO and the YSZ are seen to be atomically flat. From the high-magnification image it is possible to count the number of unit cells of STO and YSZ, nominally 25 of STO and 2 of YSZ. Most importantly, the YSZ is perfectly coherent with the STO, in agreement with x-ray diffraction (XRD) results (fig. S1), meaning that the ultrathin layer of YSZ grows rotated by 45° around the c axis and strains to match the STO lattice. Because the bulk lattice constants of STO and YSZ are 0.3905 (19) and 0.514 nm (20), respectively, the epitaxial growth of the YSZ on top of the STO ensures a large, expansive strain in the thin YSZ layers of 7% in the ab plane. Increasing the thickness of YSZ (for constant STO thickness) results in a loss of structural coherence, as reflected by a reduction of superlattice satellites in XRD. Electron microscopy observations confirm that the release of strain results in a granular morphology, although the growth remains textured.

Fig. 1.

(A) Z-contrast scanning transmission electron microscopy (STEM) image of the STO/YSZ interface of the [YSZ1nm/STO10nm]9 superlattice (with nine repeats), obtained in the VG Microscopes HB603U microscope. A yellow arrow marks the position of the YSZ layer. (Inset) Low-magnification image obtained in the VG Microscopes HB501UX column. In both cases a white arrow indicates the growth direction. (B) EEL spectra showing the O K edge obtained from the STO unit cell at the interface plane (red circles) and 4.5 nm into the STO layer (black squares). (Inset) Ti L2,3 edges for the same positions, same color code. All spectra are the result of averaging four individual spectra at these positions, with an acquisition time of 3 s each.

We plotted the lateral electrical conductivity (real part σ′) of the thinnest YSZ trilayer versus frequency in a double logarithmic plot (Fig. 2). The characteristic electrical response of ionic conductors (2123) is observed in the figure. The long range or σdc ionic conductivity of the material is obtained from the plateau found in σ′ versus frequency plots. In the presence of blocking effects due to grain boundaries or electrodes, a further decrease of σ′ (below bulk σdc values) may occur toward lower frequencies. For clarity, the value of σdc has been identified by using stars. The σdc value is found to be thermally activated, so when the temperature is reduced the conductivity curves shift downwards in Fig. 2. The inset in Fig. 2 displays Nyquist plots for the same sample. To determine the nature of the charge carriers, we measured the conductance of the samples by means of dc measurements. As can be observed in fig. S2, the dc conductance (open circles) is three to four orders of magnitude lower than the values obtained from ac measurements (solid squares) in the entire temperature measurement range. This result indicates that the electronic contribution to the ac measurements can be considered negligible, and thus, the measured ac transport is attributable to an ionic diffusion process.

Fig. 2.

Real part of the lateral electrical conductivity versus frequency of the trilayer with 1-nm-thick YSZ in a double log plot. Isotherms were measured in the range of 357 to 531 K. The solid line represents a NCL contribution (σ′ ∼ Aω, where A is a temperature-dependent proportionality factor and ω is the angular frequency), as explained in the text. Stars identify the value of σdc. The uncertainty of conductance measurements is 1 nS (10–2 S/cm in conductivity for the sample shown, see error bar). (Inset) Imaginary versus real part of the impedance (Nyquist) plots at 492, 511, and 531 K. Whereas the high-frequency contribution is a Debye-like process characterized by a conductivity exponent n = 0, the “grain boundary” term observed in the Nyquist plots shows a clear deviation from a Debye behavior, as reflected by the distorted impedance arcs.

In Fig. 3, the temperature dependence of the σdc of [STO10nm/YSZXnm/STO10nm] trilayers is shown together with data corresponding to a single crystal and the 700-nm thin film from (7). Whereas the “bulklike” samples (the thin film and the single crystal) show the well-known Arrhenius behavior with an activation energy of ∼1.1 eV, the trilayers show much larger conductivity values and smaller values of the activation energy. The thickest trilayer (62-nm YSZ) already shows an increase of about two orders of magnitude in the high-temperature dc conductivity, and the dc activation energy decreases to 0.72 eV. When decreasing the thickness of the YSZ layer to 30 nm, the dc conductivity increases another three orders of magnitude, and the activation energy decreases to 0.6 eV. The high values of the pre-exponential factor of ∼107 (ohm·cm)–1 are comparable to those found in other ion conductors (24) [see supporting online material (SOM) text]. If the thickness is further reduced all the way down to 1 nm (two unit cells of YSZ), the conductivity is observed to increase as the inverse of the YSZ layer thickness, but the conductance is essentially thickness-independent (bottom inset in Fig. 3). We can think of three parallel conduction paths due to the interfaces and the bulk YSZ and STO layers. The bulk conductivity of YSZ is 10–7 S/cm at 500 K, which would yield a conductance value of ∼10–14 S for 1-nm-thick layers. This value is much lower than the 10–6 S value measured with the ac technique. If we instead assume that the high conductance (G = 10–6 S) is due to electronic conduction in the STO, both ac and dc techniques would provide this same value, contrary to what is observed (fig. S2). Moreover, reported conductivity values in STO thin films (25) are also much lower than those necessary to explain the high conductance observed. Because bulk YSZ or STO contributions can be ruled out, an interface conduction mechanism is inferred.

Fig. 3.

Dependence of the logarithm of the long-range ionic conductivity of the trilayers STO/YSZ/STO versus inverse temperature. The thickness range of the YSZ layer is 1 to 62 nm. Also included are the data of a single crystal (sc) of YSZ and a thin film (tf) 700 nm thick [taken from (7)] with the same nominal composition. (Top inset) 400 K conductance of [YSZ1nm/STO10nm](ni/2) superlattices as a function of the number of interfaces, ni. (Bottom inset) Dependence of the conductance of [STO10nm/YSZXnm/STO10nm] trilayers at 500 K on YSZ layer thickness. Error bars are according to a 1 nS uncertainty of the conductance measurement.

To further test this scenario, we grew superlattices repeating the [YSZ1nm/STO10nm]growth unit. We found (top inset in Fig. 3) that conductance scales now with the number of interfaces up to a number of eight (four bilayer repetitions). There is a scaling breakdown in the figure, observed for a larger number of bilayer repetitions, most likely resulting from disorder building up in this highly strained structure. The experimental data indicates that the first STO/YSZ interface does not contribute to the large ionic conductivity observed in the samples, probably because the first STO layer is somehow different from the others as it is grown directly on the substrate. This scaling, together with the invariance of the conductance with the thickness of the YSZ, shows that the large conductivity values in these heterostructures originate truly at the interfaces between YSZ and STO. Our results indicate a superposition of two parallel contributions—one due to the bulk and one attributable to the interface—and the colossal ionic conductivity is observed as long as the interface conductance is larger than that of the bulk. The abrupt conductivity decrease when the thickness changes from 30 to 62 nm is most likely due to a degraded interface structure when the YSZ layers exceed the critical thickness.

The dc conductivity of the 1-nm YSZ layer shows a record value of 0.014 S/cm at 357 K, with an activation energy of 0.64 eV and an extrapolated value of 0.003 S/cm at 300 K. Thus, the threshold for the conductivity value that defines the feasibility for practical applications, 0.01 S/cm, is reached in these ultrathin films just slightly above room temperature. Previous enhancements of the conductivity in nanoscopic systems have been explained in terms of size effects (1316) and overlapping of space charge regions. However, the screening or Debye length in ionic conductors with high carrier concentrations, such as YSZ, is on the order of 0.1 nm (26). The conductivity values found here are in good agreement with a recent prediction of Kosacki et al. (17), who analyzed thicker YSZ thin films (15 nm) grown on MgO substrates, discussing the possible existence of an interface diffusion mechanism that would yield a dc conductivity ∼0.001 S/cm at room temperature, with an activation energy of 0.45 eV for a film thickness ∼1.6 nm. Consequently, and considering the good epitaxial quality of the heterostructures, we believe that strain and especially interfacial effects are at the origin of the enhanced conductivity.

To further investigate the role of the interfaces in the observed increase of ∼eight orders of magnitude in the ionic conductivity with respect to bulk YSZ values, we next present a detailed analysis of the YSZ/STO interfaces using atomic column resolution electron energy-loss spectroscopy (EELS). Figure 4A shows line traces corresponding to elemental concentrations of Ti and Sr obtained from EELS spectrum images of the [YSZ1nm/STO10nm]9 superlattice across several bilayers. The YSZ layers are the bright bands observed in the annular dark field (ADF) picture in the upper inset, which also shows in a green rectangle the area used for EELS analysis. The lower panel shows the normalized integrated intensity under the Sr M3 (dark yellow) and the TiL2,3 (red) absorption lines. The resulting two-dimensional (2D) images are shown in the inset. It can be observed that the Ti intensity is clearly higher than that of Sr in all of the interfaces, indicating that the STO termination layer is always a (TiO2) plane. YSZ grows epitaxially, rotating its cell 45° to accommodate half of the diagonal of the conventional unit cell [Math (aYSZ = 0.514 nm), where a is the lattice parameter] to the STO perovskite unit cell (aSTO = 0.390 nm) with a 7% in plane tensile strain on the YSZ. In this configuration, both structures are compatible because the FCC fluorite structure of YSZ keeps the positions of the atoms in the ab plane, the only difference being that the oxygen atoms of the fluorite are not in the z = 0 plane but are displaced to z = 1/4 along the c direction (Fig. 4B). The first YSZ plane in the (001) stacking sequence should be an Oplane at cYSZ/4, but these O sites are directly above the O atoms of the last TiO2 plane. Presumably, these sites are either vacant or the O atoms are displaced from their normal positions (as suggested by the shaded sites in Fig. 4B). Therefore, this interfacial O plane is likely to be highly disordered, even though the cation lattice remains coherent, which would thereby enable the enhancement in ionic conductivity.

Fig. 4.

(A) EELS chemical maps. The ADF image in the upper panel shows the area used for EELS mapping (spectrum image, marked with a green rectangle) in the [YSZ1nm/STO10nm]9 superlattice. The middle panel shows the averaged ADF signal acquired simultaneously with the EEL spectrum image, showing the STO (low-intensity regions) and YSZ (higher-intensity) layers. The lower panel shows the Ti (red) and Sr (dark yellow) EELS line traces across several consecutive interfaces. These line traces are averaged from the elemental 2D images shown in the insets, each framed with the same color code (red for Ti, dark yellow for Sr). Data was obtained in the VG Microscopes HB501UX. Because the STEM specimen was relatively thick (several tens of nanometers), the wide chemical interface profiles are most likely attributable to beam broadening. (B) Solid spheres model of the YSZ/STO interface showing: (1) The compatibility of the perovskite and fluorite (rotated) structures. (2) A side view of the interface between STO (at the bottom) and YSZ (on top) with realistic ionic radius. The shaded oxygen positions in the interface plane are presumed absent or displaced because of volume constraints, enabling the high ionic conductivity. (3) A 3D view of the interface, with the ionic radius reduced by half to better visualize the plane of oxygen vacancies introduced in the interface. The square symbol in the legend indicates the empty positions available for oxygen ions at the interface.

Further evidence for this interface structure comes from a close inspection of the fine structure of the O K edge (oxygen/potassium absorption edge) at the STO interface plane (Fig. 1B), which shows noticeable changes when compared to the O K edge from the middle of the STO layer. These changes are consistent with an enhanced density of O vacancies in this plane (27). However, we could not detect any change in the oxidation state of Ti at the interface (it is +4), which is in good agreement with the lack of electronic conduction observed (inset in Fig. 1B). Additionally, the fine structure of the O K edge within the YSZ ultrathin layers is completely different from that of bulk YSZ, as a consequence of the severe structural distortions in the oxygen octahedra attributable to the large strain accumulated (fig. S3 and SOM text). Thus, these results point to partial occupancy and high disorder in the interface oxygen plane, resulting in the introduction of a large number of interfacial oxygen vacancies and a substantial decrease in the activation energy for O migration. The STO side of the YSZ/STO interface may play a role in stabilizing disorder in the anionic sublattice, as has been recently reported for the LaAlO3/SrTiO3 interface (28, 29).

The analysis of the high-frequency dispersive ac conductivity above the dc plateau of Fig. 2 may provide some further insight into the colossal ionic conductivity. In ionic conductors, the dc plateau crosses over into a dispersive conductivity regime depending on frequency as a power law with a fractional exponent n (n < 1). This contribution is usually known as Jonscher's response and reflects the influence of ion-ion correlations on ion motion (2123, 30, 31). At even higher frequencies and lower temperatures, the power law dependence of the conductivity universally merges into an almost linear frequency-dependent term with n = 1, resulting in a regime with a nearly constant (dielectric) loss (NCL) (32, 33). NCL is weakly temperature dependent and has been ascribed to caged dynamics of mobile ions at short times (high frequencies) (34). Both power law regimes are usually distinct, and they are observed in particular in bulk or “thick” thin films of YSZ, where the n exponent has a value of ∼0.5 to 0.6 before merging into the NCL behavior (6, 7). In the case of our ultrathin YSZ heterostructures, either the Jonscher's response is absent (n= 0) or n is a very small value (Fig. 2), as shown by the dc plateau directly merging into a NCL (n = 1) term, which is plotted in the figure as a solid line to guide the eye. This is a very anomalous result indicating an uncorrelated ion diffusion process. The absence of the fractional power law regime has only been observed in systems with a carrier concentration low enough so that the carriers do not interact with each other. Such an explanation can be ruled out in our samples where the large conductivity values do not, by any means, indicate a reduced carrier density. We believe that the uncorrelated ion diffusion at the interface is another manifestation of conduction along the oxygen depleted/disordered interface plane. In fact, the absence of ion-ion correlations may produce a decrease in the activation energy of the conductivity (30, 31). The large in-plane expansive strain on the YSZ interface plane, together with the high concentration of vacant oxygen positions and probable positional disorder, surely contributes to the reduction in the activation energy and the resulting huge enhancement in ionic conductivity.

We have shown eight orders of magnitude enhancement of the ionic conductivity of YSZ in ultrathin films reaching values that enable practical application of the material in SOFCs slightly above room temperature. This result may have a special impact on single-chamber fuel cells where both electrodes are located on the same side of a thin electrolyte film deposited onto a substrate, and thus the ionic current flows in a lateral direction parallel to the substrate (35). The coherent interface between very dissimilar structures (in this case, fluorite and perovskite in YSZ/STO heterostructures) provides both a high carrier concentration and, simultaneously, a decreased activation energy, achieving a greatly enhanced mobility that accounts for the many orders of magnitude increase of the conductivity. The combination of epitaxial strain and suitable heterogeneous interfaces appears to be a key step in the design of artificial nanostructures with high ionic conductivity. This result is of major technological importance to achieve fast oxygen conduction at room temperature, and the outstanding electrical properties of the ultrathin YSZ/STO heterostructures may open the way to new and improved devices far beyond fuel cells.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article