Cobalt-doped Ca₁₂Al₁₄O₃₃ mayenite oxide ion conductors: phases, defects, and electrical properties

Abstract

Mayenite Ca₁₂Al₁₄O₃₃, as a good oxygen ion conductor with conductivity slightly lower than stabilized ZrO₂, has been investigated through doping strategy over the last few decades, but with little success in further improving its oxide ionic conductivity. Here, cobalt-doped Ca₁₂Al_14-xCo_xO_33+δ (0 ≤ x ≤ 1.6) materials were prepared by traditional solid-state reaction method, and then studied by complementary techniques, including X-ray diffraction (XRD), scanning electron microscope coupled with energy dispersion spectrum (EDS) analysis, X-ray photoelectron spectroscopy, and static lattice atomistic simulations. The results showed that these doped materials had much lower Co contents in the crystal structure than their nominal compositions, which was consistent with the high calculated defect formation energy (~ 6.25 eV). The minor divalent Co ions in the crystal structure would reduce the amount of mobile oxide ions and accordingly slightly decreased the bulk conductivities, while most of the Co ions existed in the form of Co₂O₃ and segregated along grain boundaries in the ceramic samples, which could apparently increase the grain boundary conductions of Ca₁₂Al₁₄O₃₃.

Introduction

Solid oxide fuel cells (SOFCs), as a clean energy conversion technology with high efficiency and fuel flexibility, are gaining increased attention over the last few decades [1,2,3]. Oxide ion–conducting electrolyte plays a crucial role in the working temperature range of SOFCs [4,5,6]. The yttrium-stabilized ZrO₂ (YSZ), as a traditional and most widely used electrolyte in the industry, can be used only for temperatures higher than 750 °C [7,8,9]. This high working temperature raises undesired reactions between the electrolyte and electrode materials as well as issues concerning thermal stresses. Thus, it is important to develop new oxide ion conductors with high ionic conductivity for SOFCs subjected to intermediate temperatures (500–750 °C) [9,10,11].

The oxide ion conductivity of mayenite Ca₁₂Al₁₄O₃₃ was first identified in 1988 by M. Lacerda et al. [12]. The Ca₁₂Al₁₄O₃₃ material possesses slightly lower oxide ion conduction than that of YSZ, which therefore makes Ca₁₂Al₁₄O₃₃ a competitive candidate for SOFC electrolytes. The parent Ca₁₂Al₁₄O₃₃ has a cubic crystal structure belonging to the I$\overline{4}$3d space group with a lattice constant of ~ 11.99 Å. For each unit cell, there are two Ca₁₂Al₁₄O₃₃ molecules. Twelve cages build up the positively charged framework, with two free O ions randomly occupying two different cages. The unit cell can therefore be represented by a chemical formula [Ca₂₄Al₂₈O₆₄]⁴⁺•2O²⁻. Each cage that composed of framework Ca, Al, and O atoms has a width of ~ 6 Å and surrounded by 8 other cages with ~ 3.7 Å wide windows shared, as shown in Fig. 1. The highly mobile free oxygen ions lead already for the parent Ca₁₂Al₁₄O₃₃ material to an oxide ion conductor. The oxygen migration in un-doped Ca₁₂Al₁₄O₃₃ material was examined by neutron diffraction [13], density functional theory (DFT), and molecular dynamics (MD) calculations [14, 15] as well as oxygen-isotope-exchange experiments [16, 17]. Irvine et al. made the first attempt to enhance the high-temperature ionic conduction of Ca₁₂Al₁₄O₃₃ material by zinc doping or zinc and phosphorus co-doping on the Al sites [18]. Then, Ebbinghaus et al. investigated the iron-doped mayenite [19]. Both of these cases resulted in slight decreases in ionic conduction for the doped samples. Studies of copper [20] or iridium [21] incorporation into mayenite were also reported, without effects of these dopants on the oxide ion conduction being examined. More recently, we reported the synthesis, defects, and electrical properties of Ga-doped Ca₁₂Al₁₄O₃₃ material [22]. Ga-doping on the Al site was shown to improve the bulk conductivity of Ca₁₂Al₁₄O₃₃.

Fig. 1

a Two exemplar connecting cages; b the window shared by two adjacent cages in Ca₁₂Al₁₄O₃₃. The green, cyan, yellow, and red balls represent free oxygen, framework Al, Ca, and O atoms, respectively; the value “3.7 Å” denotes the width of the window, which is exactly the longest distance between Ca and O atoms in this window

In this study, the Co ions, which have larger ionic sizes than the Al ions in the same coordination environment, were selected to substitute for Al in order to increase the lattice parameters and enlarge the width of the window of two connecting neighboring cages, which will accordingly benefit the migration of oxide ions. The effects of Co doping on the phase, structure, and electrical properties of Ca₁₂Al₁₄O₃₃ were then investigated. The results show that the raw material Co₂O₃ mainly existed as a secondary phase in the product and segregated along the grain boundaries of Ca₁₂Al₁₄O₃₃, which could apparently increase the grain boundary conduction, especially at temperatures below 700 °C. Only a small part of Co₂O₃ was reduced at high temperature into divalent Co²⁺ ions. These divalent Co²⁺ ions can successfully substitute the 4-coordinated Al ions in Ca₁₂Al₁₄O₃₃. However, this acceptor-doping would lead to a slight decrease in the bulk conductivity of Ca₁₂Al₁₄O₃₃, which may due to the decrease of free oxygen that caused by the acceptor-doping. The high defect formation energy for Co²⁺ ions substituting for 4-coordinated tetrahedral Al ions derived from the static lattice simulation agreed well with its narrow solid solution range.

Materials and methods

The samples of Ca₁₂Al_14-xCo_xO_33+δ were prepared by a traditional solid-state reaction method using CaCO₃ (Alfa Aesar, > 99.8% purity), Al₂O₃ (Alfa Aesar, > 99.997% purity), and Co₂O₃ (Alfa Aesar, > 99.99% purity) as starting raw materials. All raw materials were weighed stoichiometrically. The evenly mixed and well-ground raw materials were first fired at 1000 °C for 12 h to drive off CO₂. The precalcined powders, after regrinding, were then uniaxially pressed into pellets and sintered at 1250 °C for 24 h in the air to make the reaction fully completed and yield the target products. Archimedes principle was applied to estimate the densities of these prepared ceramics.

The XRD characterizations were performed on a Panalytical X’pert Pro X-ray diffractometer with Cu Kα radiation. For Rietveld refinement applications, the XRD data were collected over a 2θ range of 5–120°. Topas-Academic software [23] was used for these refinements. X-ray photoelectron spectroscopy (XPS) was recorded on ESCALAB 250Xi, Thermo Scientific (America), and a vacuum of 10⁻¹⁰ mbar was applied. The microstructure and EDS analyses were performed on a Hitachi (Tokyo, Japan) S4800 scanning electron microscope (SEM). Solartron 1260 frequency response analyzer was used to perform the AC impedance spectroscopy (IS) measurements over a 10⁷–10⁻¹ Hz frequency range. Prior to performing the IS measurements, the platinum paste was coated on the opposite faces of the pellets and subsequently fired at 750 °C for 1 h to remove organic components and form electrodes.

The energies of Co²⁺ ions substituting for Al³⁺ ions were studied by atomistic static lattice simulations, using the General Utility Lattice Program (GULP) [24, 25] based on the interatomic potential approach. Buckingham potential function [26] was used to model the interactions between ions, using the shell model [27] to describe the electronic polarizability for structural modeling. The interatomic potential parameters used for the atomistic simulations are listed in Table 1.

Table 1 Buckingham interatomic potential and shell model parameters used for the atomistic simulations

Results and discussion

XRD data

Figure 2a shows the XRD patterns of Ca₁₂Al_14-xCo_xO_33+δ (0 ≤ x ≤ 1.6). We can clearly see that no peaks from a secondary phase are observed for compositions x ≤ 1.2, and all the reflection peaks can be indexed with the mayenite structure phase Ca₁₂Al₁₄O₃₃ (PDF#70-2144). For x = 1.6, however, reflections from Co₂O₃ (PDF#02-0770) impurity appear. The refined cell parameters for different compositions were plotted in Fig. 2b which shows that the lattice parameters of the doped materials are close to each other and only slightly higher than that of the un-doped parent material, indicating deviations between the exactly Co contents in the crystal structure and the nominal compositions. According to Vegard’s law [28], unit cell parameters should vary linearly with composition for a continuous substitutional solid solution. Here, as the effective ionic radius of Al³⁺ (in a four-coordinated environment) is ~ 0.39 Å, smaller than that of Co²⁺/Co³⁺ (> 0.5 Å) [29], a linear lattice expansion obeying Vegard’s law is expected for Ca₁₂Al_14-xCo_xO_33+δ with increasing values of x, if all the Co atoms enter into the crystal structure and occupy the Al atoms’ sites. Rietveld structure refinements were then carried out on all these doped samples. For these Rietveld refinements, a parent cubic Ca₁₂Al₁₄O₃₃ structure (space group I$\overline{4}$3d) was used as the starting structure model. This model contains two Ca sites (24d), two Al sites (16c and 12a) in two different kinds of AlO₄ tetrahedron (3-linked and 4-linked, respectively, the numbers 3 and 4 denoting the bridging oxygen number of an AlO₄ tetrahedron corner-shared with other AlO₄ tetrahedra), and three oxygen sites (one 16c and two 48e). Results of refinements show that the Co contents in the crystal structure of all these doped samples are similar and that the Co atoms have a preferred distribution on the 3-linked Al2 site. The typical Rietveld fitted plot and the final refined structural parameters (including the Wyckoff site, fractional coordinate (x, y, z), occupancy, and isotropic thermal factor B_iso) of the Ca₁₂Al_12.8Co_1.2O_33+δ sample which has the highest nominal Co content (without secondary phases observed in the XRD patterns) are given in Fig. 3 and Table 2, respectively. The refined cell parameter a = 11.9959(1) Å, weighted profile R factor R_wp ≈ 5.16%, Bragg-intensity R factor R_B ≈ 3.78%. Such low R factors indicate the good reliability of the refinement. The refined occupancy of Co on the 3-linked Al2 site is ~ 0.022(1), corresponding to the chemical formula Ca₁₂Al_13.824(4)Co_0.176(4)O_33+δ. Therefore, the Co content in the crystal structure is believed to be much lower than that expected in the nominal formula. In fact, attempt to fix the Co content in the crystal structure as high as designed would lead to a much worse fitting result with R_wp > 18%. In addition, when we fixed all the Co atoms on the 4-linked Al site, the refinement also had a much worse result (R_wp ≈ 11%).

Fig. 2

a XRD patterns and b refined cell parameters with errors included of Ca₁₂Al_14-xCo_xO_33+δ (x = 0, 0.4, 0.8, 1.2, and 1.6)

Fig. 3

Rietveld fitting for XRD data of Ca₁₂Al_12.8Co_1.2O_33+δ

Table 2 Final refined structural parameters of Ca₁₂Al_12.8Co_1.2O_33+δ

SEM analysis

As stated above, the refined Co content in the crystal structure is much lower than expected. Therefore, it is essential to figure out the exact location of the remaining Co atoms. For this purpose, backscattered electron (BSE) SEM measurements combined with EDS elemental analysis were carried out. Normally, the contrast of BSE images is related only to the atomic number of elements on the surface of the sample and is independent of morphological contrasts. When the elemental distribution of the sample’s surface is not uniform, areas containing atoms with large mean atomic numbers show higher brightness than an area with atoms of low mean atomic numbers. Thus, different phases can be distinguished based on the brightness contrasts of the BSE images [30]. In the present study, the BSE images revealed secondary phases along the grain boundaries for all the doped materials. As the brightness contrast is clearer in the sample with higher cobalt content, the backscattered electron (BSE) images recorded on the ceramic sample of x = 1.2 were selected as examples and demonstrated in Fig. 4a. As can be clearly seen, the contrast between the grain and grain boundary makes it easy to conclude that the substance along the grain boundary can be ascribed to a secondary phase. The subsequent EDS elemental distribution map analyses (Fig. 4b–d) display clear Co element distribution areas (Fig. 4d), which in fact is exactly the secondary phase areas. Element concentration analysis on a selected dot also shows a much higher Co element concentration than that of the Al and Ca element, as shown in Fig. 4e, f. These results combining with the XRD pattern of the sample Ca₁₂Al_12.4Co_1.6O_33+δ displayed above, the secondary phase is most likely the starting material Co₂O₃. For the XRD patterns of Ca₁₂Al_14-xCo_xO_33+δ with x ≤ 1.2, the absence of peaks from Co₂O₃ is probably because the Co₂O₃ contents in these samples are below the detection limits of the X-ray diffractometer.

Fig. 4

SEM image (a) and EDS element distribution maps of Ca, Al, and Co, respectively (b–d) for the untreated Ca₁₂Al_12.8Co_1.2O_33+δ surface; the zoom-in picture of the area in the red rectangle in picture a (e); the element concentrations of the red dot in picture e (f), and the peak which is u-labeled corresponds to the Au element that sprayed on the surface of the ceramic pellet prior to measurements

XPS analysis

In addition, XPS measurements on Co-doped samples were used to shed light on the valence state of the Co atoms. Figure 5 displays the spectrum of the sample Ca₁₂Al_12.8Co_1.2O_33+δ. As is well documented, in the XPS spectra of Co(II) high-spin compounds, such as CoO, there are intense shake-up satellite peaks at ~ 787.0 and 804.0 eV; whereas for the XPS spectra of low-spin Co(III) compounds, the satellite peaks are weak or missing [31]. Thus, the Co 2p XPS spectrum of the well-known mixed valence oxide Co₃O₄ is symptomatic of weak shake-up satellite structures due to the minor Co(II) component [32]. In this work, besides the two main peaks around 780.81 and 796.61 eV, two apparent satellite peaks at ~ 785.41 and ~ 802.02 eV, with intermediate intensities, can be observed in the spectrum. These satellite peaks indicate the mixed valence nature of the Co element in the Ca₁₂Al_12.8Co_1.2O_33+δ sample. Since the Co atoms distributed along the grain boundaries are identified to exist in the form of Co₂O₃ in the trivalent state, minor divalent Co²⁺ ions therefore are assumed to have entered into the crystal structure of Ca₁₂Al₁₄O₃₃ and occupy the Al sites, as the divalent high-spin Co²⁺ ions, rather than the low-spin Co³⁺ ions, are usually more readily to occupy tetrahedral sites in the crystal. This is well documented and validated by experiments and DFT calculations [33]. A representative case is that of the crystal structure of mixed valence Co₃O₄ (space group Fd$\overline{3}$m, a ≈ 8.1 Å) where the Co²⁺ and Co³⁺ ions show highly ordered distribution in 4-coordinated tetrahedral sites and 6-coordinated octahedral sites, respectively.

Fig. 5

The Co 2p spectra of the nominal Ca₁₂Al_12.8Co_1.2O_33+δ sample

Static lattice simulation

As previously mentioned, the refined cell parameters of Ca₁₂Al_14-xCo_xO_33+δ indicated a very narrow solid solution, and the Co ions in the mayenite structure are mainly divalent. This narrow solid solution usually relates to high defect formation energies for substitution. For the purpose of verifying, defect formation energies were then calculated using static lattice simulation technology based on combination of dopant and vacancy defect energies and lattice energies of the binary oxides, and can be simply determined from the following defect equation:

$$ 2 CoO+2\;A{l_{Al}}^{\times}\to 2C{o}_{Al}\hbox{'}+A{l}_2{O}_3+{V}_{\mathrm{O}}^{\bullet \bullet } $$

(1)

i.e.,

$$ \varDelta {H}_{formation}=E\left({V_O}^{\bullet \bullet}\right)+2E\left(C{o}_{Al}^{\prime}\right)+E\left(A{l}_2{O}_3\right)-2\ E(CoO) $$

(2)

where the E(X) denotes the calculated total energy or point defect energy of the species of interest after geometric optimization.

The starting point was to reproduce the experimental crystal structure of Ca₁₂Al₁₄O₃₃, CoO, and Al₂O₃. The mayenite structure exhibited by Ca₁₂Al₁₄O₃₃ material is cubic (space group I$\overline{4}$3d) as mentioned before, while CoO and Al₂O₃ adapt cubic and rhombohedral structure, respectively. The interatomic potentials used for these calculations are shown in Table 1. The results shown in Table 3 demonstrate that the deviations between the calculated and experimental unit cell edges and volumes are below 2%.

Table 3 Calculated and experimental structural parameters of Ca₁₂Al₁₄O₃₃ (space group I$\overline{4}$3d), CoO (space group Fm$\overline{3}$_m), and Al₂O₃ (space group R$\overline{3}$cH)

The final calculated formation energy for Co²⁺ ions to substitute for 3-linked tetrahedral Al2 sites, derived from Eq. (2), was ~ 6.25 eV. For comparison, the possibility of Co²⁺ ions replacing 4-linked tetrahedral Al1 sites was also investigated, which resulted in the formation energy of ~ 7.53 eV, apparently higher than that in the 3-linked case. Generally, the defect formation energy below 3.0 eV implies a relatively wide solid solution. Otherwise, it probably would have resulted in a narrow solid solution. For example, our previous work reported that for LaMGa₃O₇ (M = Ba, Sr, Ca), the highest defect formation energy of ~ 2.74 eV occurred for La substituting for Ba with about 35% of the Ba ions being replaced by La ions [34]. Recently, it has been reported that for K/Na-doped SrSiO₃ materials, the lowest defect formation energy when K/Na substitutes for Sr is 4.13 eV, and that the K/Na-doped SrSiO₃ turned out to have an extremely low range of solid solutions [35]. In the present work, the calculated defect formation energies for Co²⁺ ions replacing Al atoms explain well the narrow solid solution in Ca₁₂Al_14-xCo_xO_33+δ and verify that the Co²⁺ ions show strong preference in the 3-linked tetrahedral Al2 sites. For comparison, the defect formation energies for Co³⁺ ions substitution for Al ions were also calculated in a similar way, and the results turned out to be ~ 7.36 eV and ~ 8.95 eV for Co³⁺ ions occupying the 3-linked tetrahedral Al2 site and 4-linked tetrahedral Al2 site, respectively, apparently higher than that in the case of Co²⁺ ions.

Electrical property

The electrical properties of the Ca₁₂Al_14-xCo_xO_33+δ ceramic samples were investigated by alternating current (AC) impedance spectroscopy (IS). For AC impedance spectroscopy measurements on a ceramic sample, the different relaxation times for grains (bulk) and grain boundaries make it accessible to separate their corresponding resistance, resulting in separate semicircles in the complex impedance spectrum. Therefore, one can calculate the bulk conductivity (σ_b), grain boundary conductivity (σ_b), and the total conductivity (σ_t) once the bulk resistance (R_b), grain boundary resistance (R_g), and total resistance (R_t = R_b + R_g) were determined. The densities of ceramic samples for IS measurements calculated based on the Archimedes’ principle are all higher than 91% of the theoretical predictions. The densities of these samples did not show regular variation as a function of Co content. The AC impedance measurements for all the samples were carried out within the temperature range of 300–950 °C.

Figure 6 displays several selected typical impedance spectra from parent Ca₁₂Al₁₄O₃₃ sample at various temperatures. At temperature 350 °C (Fig. 6a), the impedance complex plane plot is dominated by the semicircle arc related to grain boundary resistance (R_gb) response, and similar impedance spectra of Ca₁₂Al₁₄O₃₃ was also reported by J.T.S. Irvine et al. [18]. The inset clearly shows a small semicircle corresponding to bulk response (R_b) at high frequency. No arc related to electrode response was detected at this temperature. When the temperature reached 550 °C (Fig. 6b), besides the arcs from bulk and grain boundary responses, a small tail of Warburg-type electrode response (inclined spike which usually related with a diffusion process of ions) in the low-frequency region (< 10 Hz) with an associated large capacitance > 10⁻⁷ F/cm was observed, corresponding to the ionic conduction [36]. At temperature 650 °C (Fig. 6c), the arc from bulk response disappeared, which can be explained by the equation 2π f_maxRC = 1 where the f_max is the frequency at Z″_max, and R and C are resistance and associated capacitance of bulk, respectively [36]. Thus, when the resistance of bulk decreased with increasing temperature, the f_max would increase and out with respect to the frequency scale of the instrumentation (such as maximum 1 × 10⁷ Hz here). When temperature is ≥ 950 °C, the resistance of grain boundary is much smaller than that of bulk. As can be seen in Fig. 6d, the intercept on the Z′ axis at high frequency can be ascribed to the bulk resistance and is greater than 350 Ω cm, while the grain boundary resistance derived from the first semicircle on the spectra is only ~ 100 Ω cm.

Fig. 6

a–d Complex impedance plots for Ca₁₂Al₁₄O₃₃ at various temperatures. Selected frequencies (Hz) marked by the red square are labeled

Co-doping on Ca₁₂Al₁₄O₃₃ was revealed by complex impedance spectra to have a great effect on the electrical property, and can significantly increase the grain boundary conductivities. For illustration, Fig. 7 demonstrates the typical spectra of sample Ca₁₂Al_12.8Co_1.2O_33+δ recorded at various temperatures. At 350 °C, the spectrum (Fig. 7a) is composed of two severely overlapped and depressed arcs at high- and low-frequency regions related to bulk and grain boundary response, respectively, with a following small tail corresponding to electrode response arising from ionic conduction. The electrode response, however, was not observed at the same temperature for the parent material, indicating that the Co-doping can accelerate the migration of the free oxide ions in this material. Usually, a perfect semicircle with center on the real axis represents an ideal Debye type behavior corresponding to a single relaxation mechanism, and can be modeled through a parallel resistance-capacitance circuit. However, such perfect semicircular arcs are expected only in homogeneous materials. In ceramic systems, due to the variations in grain size, the semicircular arcs in the complex impedance plane corresponding to the grain and grain boundary effects would get inclined to the real axis with the center generally located below the real axis implying distribution of relaxation time [37, 38]. In such case, a simple parallel network model of bulk or grain boundary resistance-capacitance (R_b–C_b or R_gb–C_gb) is accepted as a poor approximation [39]. When this non-ideal frequency response is present, it is commonly accepted to employ distributed circuit elements in an equivalent circuit, and the most widely used is the constant phase element (CPE), which has a non-integer power dependence on the frequency [40]. Often a CPE is used in a model in place of a capacitor to compensate for non-homogeneity in the system. Therefore, in order to acquire the accurate resistances of bulk and grain boundary, here, equivalent circuit consisting of a serial combination of two circuits with three parallel R_i, C_i, and (CPE)_i elements (R, C, and CPE represent the device resistance, capacitance, and constant phase element, respectively; i is b for bulk or gb for grain boundary) were used to model these two responses, as illustrated in the inset in the top right corner of Fig. 7a. The bulk conductivity (σ_b) and grain boundary conductivity (σ_gb) are calculated from R_b and R_gb, respectively, while the total conductivity (σ_b) is calculated from the total resistance of R_b plus R_gb, i.e., R_b + R_gb. The obtained R_b (~ 1.1 × 10⁶ Ω cm) is higher than that of the un-doped Ca₁₂Al₁₄O₃₃ (3.21 × 10⁵ Ω cm) at the same temperature, but the R_gb (~ 9.0 × 10⁵ Ω cm) is much lower than that of the parent Ca₁₂Al₁₄O₃₃ (~ 4.75 × 10⁷ Ω cm). This indicates that the substitution of Co²⁺ for Al³⁺ in the crystal structure will decrease the bulk conductivity of Ca₁₂Al₁₄O₃₃, which can be most likely ascribed to the reduction of free oxygen content caused by this acceptor doping, and the segregation of Co₂O₃ along the grain boundary can significantly enhance the grain boundary conduction. When the temperature is ≥ 750 °C, the semicircle corresponding to the bulk impedance diminished. This is because the bulk impedance is out respect to the frequency scale of the instrumentation, as explained for Fig. 6c with a similar phenomenon. The grain boundary impedance dominated the spectra at this temperature, as can be appreciated from Fig. 7b.

Fig. 7

a, b Complex impedance plots of the sample Ca1₂Al_12.8Co_1.2O_33+δ. The experimental and calculated data are shown as black open hexagons and red solid line, respectively

Figure 8 shows the bulk, grain boundary, and total conductivities of Ca₁₂Al_14-xCo_xO_33+δ (x = 0, 0.4, 0.8, 1.2). It can be seen that the bulk conductivities of the doped samples are not improved as we have expected, but even slightly lower than that of the parent material, especially in the low-temperature regions (Fig. 8a). A similar behavior occurred with Zn-doped Ca₁₂Al_14-xZn_xO_33+δ materials as previously reported [18]. However, the bulk conductivities of these Co-doped materials at a given temperature are close to each other, which may be explained by the fact of close levels of Co contents in the crystal structure of those doped materials as stated before. In addition, all these doped materials have almost the same activation energy (~ 0.96 eV) which is slightly higher than that of the un-doped Ca₁₂Al₁₄O₃₃ (~ 0.79 eV). This activation energy increase is likely related to the local framework distortion arising from the replacement of aluminum by cobalt, and the similar case also occurred in the zinc- and/or phosphorus-doped Ca₁₂Al₁₄O₃₃ materials [18]. Although the bulk conduction was decreased with Co-doping, Fig. 8b shows apparently enhances in grain boundary conductivities at temperatures ≤ 650 °C. The improvements are limited at temperatures above 700 °C. This can be explained by the fact that at low temperatures, the oxide ions transport between grains is greatly blocked at grain boundaries in the un-doped material, making the grain boundary the main limiting factor for oxide ion migration. While for the Co-doped materials the existence of Co₂O₃ along the grain boundaries can greatly reduce the blocking resistance between the grains. At elevated temperatures, the resistance of grain boundaries becomes small and is no longer the main limiting factor for oxide ion migration even for the un-doped material; therefore, the effect of Co₂O₃ segregation along grain boundaries appears to lessen the increase of conduction at grain boundaries. Figure 8c shows the enhancements in total conductivities of series Ca₁₂Al_12.8Co_1.2O_33+δ samples.

Fig. 8

a Bulk, b grain boundary, and c total conductivities of Ca₁₂Al_14-xCo_xO_33+δ; dp O₂ dependence of total conductivity for the sample Ca₁₂Al_12.8Co_1.2O_33+δ

To verify whether there are electronic conductivities for those Co-doped materials, AC impedance measurements were performed on the Ca₁₂Al_12.8Co_1.2O_33+δ sample under different atmospheres with variable oxygen partial pressures between 400 and 850 °C. The results showed no apparent change in the total conductivity under variable atmospheres, as shown in Fig. 8d, indicating the pure ionic conduction nature in the Co-doped materials, and the electronic conduction, if any, was small.

Conclusion

In this work, a series of Co-doped Ca₁₂Al_14-xCo_xO_33+δ (0 ≤ x ≤ 1.6) materials were synthesized by a traditional solid-state method. Although the XRD did not detect a secondary phase for x ≤ 1.2, the Rietveld structure refinements together with SEM, EDS and XPS technologies revealed that only a small amount of Co element (Co²⁺) successfully substituted the 3-linked tetrahedral site Al atoms in the crystal structure of Ca₁₂Al₁₄O₃₃ and most of the Co atoms were distributed along the grain boundaries in the form of Co₂O₃. The high defect formation energy (~ 6.25 eV) calculated with a static lattice simulation technique agrees well with the narrow solid solution range for Co²⁺/Al³⁺ in the crystal structure. The substitution of Co²⁺ for Al³⁺ ions slightly reduced the bulk conductivities for the doped materials, but apparently increased the grain boundary conductions, and without electronic conduction introduced.