Solid State Sciences 9 (2007) 32e42 www.elsevier.com/locate/ssscie
Redox behavior and transport properties of La0.52xCexSr0.5þxFeO3d and La0.52ySr0.5þ2yFe1yNbyO3d perovskites V.V. Kharton a,b,*, J.C. Waerenborgh c, A.V. Kovalevsky a, G.C. Mather d, A.P. Viskup b, M.V. Patrakeev e, P. Gaczyn´ski c, A.A. Yaremchenko a, V.V. Samakhval b a
b
Department of Ceramics and Glass Engineering, CICECO, University of Aveiro, 3810-193 Aveiro, Portugal Institute of Physicochemical Problems, Belarus State University, 14 Leningradskaya Street, 220050 Minsk, Belarus c Chemistry Department, ITN/CFMC-UL, Estrada Nacional 10, P-2686-953 Sacave´m, Portugal d Institute of Ceramics and Glass, CSIC, Cantoblanco, 28049 Madrid, Spain e Institute of Solid State Chemistry, Ural Division of RAS, 91 Pervomaiskaya Street, Ekaterinburg 620219, Russia Received 11 April 2006; received in revised form 20 October 2006; accepted 9 November 2006 Available online 17 January 2007
Abstract The effects of doping the mixed-conducting (La,Sr)FeO3d system with Ce and Nb have been examined for the solid-solution series, La0.52xCexSr0.5þxFeO3d (x ¼ 0e0.20) and La0.52ySr0.5þ2yFe1yNbyO3d (y ¼ 0.05e0.10). Mo¨ssbauer spectroscopy at 4.1 and 297 K showed that Ce4þ and Nb5þ incorporation suppresses delocalization of p-type electronic charge carriers, whilst oxygen nonstoichiometry of the Cecontaining materials increases. Similar behavior was observed for La0.3Sr0.7Fe0.90Nb0.10O3d at 923e1223 K by coulometric titration and thermogravimetry. High-temperature transport properties were studied with Faradaic efficiency (FE), oxygen-permeation, thermopower and totalconductivity measurements in the oxygen partial pressure range 105e0.5 atm. The hole conductivity is lower for the Ce- and Nb-containing perovskites, primarily as a result of the lower Fe4þ concentration. Both dopants decrease oxide-ion conductivity but the effect of Nb-doping on ionic transport is moderate and ion-transference numbers are higher with respect to the Nb-free parent phase, 2.2 103 for La0.3Sr0.7Fe0.9Nb0.1O3d cf. 1.3 103 for La0.5Sr0.5FeO3d at 1223 K and atmospheric oxygen pressure. The average thermal expansion coefficients calculated from dilatometric data decrease on doping, varying in the range (19.0e21.2) 106 K1 at 780e1080 K. Ó 2006 Elsevier Masson SAS. All rights reserved. Keywords: Perovskite; Lanthanumestrontiumeferrite; Mixed conductor; Ionic conductivity; Mo¨ssbauer spectroscopy; Hole transport
1. Introduction Synthesis gas (syngas, CO þ H2) is the feedstock for FischereTropsch and methanol synthesis whose production is likely to increase significantly in the coming years. This is partly because the efficient conversion of syngas to a liquid fuel could extend petrol reserves for decades [1e3]. In addition, syngas is currently the cheapest source of hydrogen * Corresponding author. Department of Ceramics and Glass Engineering, CICECO, University of Aveiro, 3810-193 Aveiro, Portugal. Tel.: þ351 234 370263; fax: þ351 234 425300. E-mail address:
[email protected] (V.V. Kharton). 1293-2558/$ - see front matter Ó 2006 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.solidstatesciences.2006.11.008
and likely to be of major importance in the first steps towards a hydrogen economy. Currently, commercial production involves the endothermic reforming of methane, a costly, energy-intensive process which yields a H2/CO ratio higher than the optimum required for useful conversion reactions. Syngas with an optimized ratio of components may be formed more efficiently from methane through partial oxidation, employing a dense ceramic membrane reactor with mixed oxide-ion and electronic conduction. One of the most promising groups of materials for this application is the lanthanume strontiumeferrite system with the perovskite structure, (La,Sr)FeO3d [4e11]. The electronic and ionic conductivities of (La,Sr)FeO3d is maximized for equivalent La and Sr
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
contents (x ¼ 0.5) [10,11]; the higher vacancy concentrations formed with greater Sr content result in substantial vacancy ordering and deterioration of transport properties. Although the system exhibits high oxygen permeability, thermodynamic and dimensional stability is insufficient under the large oxygen chemical-potential gradients required for membrane operation. Doping strategies have been employed to improve stability, principally through substitution of Fe with higher-valence, more redox-stable cations, such as Cr and Ti [4,9,11e13]. However, substitution with either of these species brings about a decrease in oxygen nonstoichiometry, ionic conductivity and oxygen permeability. If the potential of dense ceramic membrane reactors is to be realized, further efforts are required in order to optimize both stability and electrochemical performance of the most-promising mixed-conducting systems. In this report, Ce- and Nb-doping of the (La,Sr)FeO3d family is examined. Previous studies have indicated that the solubilities of these dopants in the related SrFeO3d system in air are relatively high, up to 20 at.% for Ce on the A site of the (ABO3) perovskite [14] and up to 50 at.% for Nb on the B site [15]. In the case of Ce-doped phases, site-size requirements would suggest that the A site is occupied by Ce3þ. However, the Ce3þ state is rather unusual in oxidizing conditions [16] and its presence in the perovskite structure may depend to a large degree on the overall composition; for example, Ce does not occupy the A site (either as Ce3þ or Ce4þ) in the strontium-deficient Sr1xCeO3d system [17]. The oxygen nonstoichiometry in (Sr,Ce)FeO3d was reported to be essentially independent of cerium content, 0.20 0.02 at room temperature in air [14]. This may suggest that the incorporation of either Ce3þ or Ce4þ is chargecompensated by a lower iron oxidation state. At the same time, the estimations of d [14] were made from the amount of desorbed oxygen, assuming that all Fe and Ce cations in argon atmosphere are in the þ3 and þ4 valence states, respectively. This assumption requires careful validation, especially considering that Fe4þ is present in SrFeO3d down to relatively low oxygen pressures [18]. In the case of SrFe(Nb)O3d, Nbdoping leads to important changes in the local environment of the Fe cations, although no cation ordering was found [15]. Here, we focus on the La0.52xCexSr0.5þxFeO3d and La0.52y Sr0.5þ2yFe1yNbyO3d solid-solution systems, which were chosen with a view to maintaining a constant iron oxidation state; hence, substitution with either Ce or Nb involves an increase in the Sr content, assuming fixed oxidation states of þ4 and þ5 for the Ce and Nb cations, respectively. Oxygen stoichiometry and iron oxidation states are, respectively, determined by coulometric titration and Mo¨ssbauer spectroscopy. Ionic and electronic transport is analyzed employing Faradaic efficiency, oxygen-permeation, thermopower and total-conductivity measurements. 2. Experimental Samples in the La0.52xCexSr0.5þxFeO3d and La0.52y Sr0.5þ2yFe1yNbyO3d systems were synthesised by solid-state reaction from high-purity La2O3, Ce(NO3)3$6H2O, SrCO3,
33
FeC2O4$2H2O and Nb2O5; La2O3 and Nb2O5 were previously annealed in air at 1473 and 1073 K, respectively, to remove adsorbates. After thermal decomposition of nitrate and carbonate components, reaction was conducted in air in the temperature range 1373e1523 K for 30e50 h with multiple intermediate grindings. Ceramics with 91e98% of theoretical density were obtained on pressing to 300e400 MPa and sintering in air at 1553e1773 K for 4e6 h. The samples were finally annealed in air at 1273 K for 3e4 h and slowly cooled in order to achieve equilibrium with air at low temperatures. Characterization of the materials was carried out employing X-ray diffraction (XRD), scanning electron microscopy coupled with energy-dispersive spectroscopy (SEM/EDS), inductivelycoupled plasma (ICP) spectroscopic analysis, testing of gastightness, dilatometry, thermogravimetric analysis (TGA) and coulometric titration (CT). Determination of electrochemical properties included measurements of steady-state oxygenpermeation fluxes and Faradaic efficiency (FE) to determine ion-transference numbers under zero oxygen chemical-potential gradient at p(O2) ¼ 0.21 atm. Total conductivity and Seebeck coefficient were also measured at 973e1223 K in the oxygen partial pressure range 105e0.5 atm. Detailed descriptions of the experimental procedures and equipment can be found in Refs. [8e12,18,19] and references therein. Powdered samples for XRD, TGA, CT and Mo¨ssbauer spectroscopy were obtained by grinding dense ceramics, firstly equilibrated in atmospheric air as described above. Changes in oxygen content with respect to a reference point were studied by the coulometric titration (CT) method combined with TGA, as described elsewhere [18e20]. The CT measurements were performed using double electrochemical cells [20] in an isothermal regime between 923 and 1223 K. After each isothermal redox cycle, the sample was re-oxidized and tested for reproducibility (Fig. 1A). For La0.5Sr0.5FeO3d, the total oxygen content in the reference point (1223 K, air) was determined by TGA using a Setaram SetSys 16/18 instrument. The standard TGA procedure included heating up to 1223 K in a flow of dry CO2-purified air, temperature cycling in the range 1023e 1223 K with a step of 50 K and equilibration at each temperature during 2e4 h, flushing the apparatus with argon, and then reduction in a flow of dry 10% H2e90% N2 mixture at 1373 K. The reduction process was carried out until the sample weight became time-independent, indicating complete reduction of the ferrite into iron metal, SrO and La2O3. The p(O2)eTed diagram, calculated from the CT data using the reference point obtained by TGA at 1223 K in air, was further validated by comparison with the total-conductivity data in reducing atmospheres. According to the well-known theoretical models for the defect chemistry of oxygen-deficient ferrites, confirmed experimentally (e.g., see Refs. [6,9,10,18,19,21] and references therein), the minimum of the conductivity vs. oxygen partial pressure curves corresponds to the state where the average oxidation state of iron ions is equal to 3þ; this state is often called the electrone hole equilibrium point. Indeed, the total oxygen content in La0.5Sr0.5FeO3d, observed experimentally under these conditions, was equal to 2.750 atoms per formula unit; the error (0.002) corresponds to the level of experimental uncertainty
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
34 2.96
A
Cycle 1 Cycle 2 2.94 973 K 2.92 1023 K 2.90
1073 K 1123 K
2.88 1173 K 1223 K
2.86 -0.7
-0.6
-0.5
-0.4
-0.3
-0.2
B
may give a significant error, the TGA data showed relatively good reproducibility (Fig. 2). Mo¨ssbauer spectra were collected at room temperature and at 4.1 K in transmission mode using a conventional constantacceleration spectrometer and a 25 mCi 57Co source in a Rh matrix. The velocity scale was calibrated using a-Fe foil. The absorbers were obtained by pressing the powdered samples (5 mg of natural Fe/cm2) into Perspex holders. Isomer shifts (IS, Table 2) are given relative to metallic a-Fe at room temperature. Low-temperature spectra were collected using a Janis bath cryostat (model SVT-400), with the sample immersed in liquid He. The spectra were fitted to Lorentzian lines using a non-linear least-squares method [25]. The relative areas and widths of both peaks in a quadrupole doublet were kept equal during refinement. The distributions of magnetic splittings were fitted according to the histogram method [26]. 3. Results and discussion
2.96 CT 2.94
2.92
3.1. Phase analysis and thermal expansion
TGA
973 K 1023 K
2.90 1073 K 2.88
1123 K
1173 K
2.86 -1.2
-1.0
-0.8
-0.6
-0.4
-0.2
log p(O2) (atm) Fig. 1. Reproducibility of the total oxygen content of La0.3Sr0.7Fe0.90Nb0.10O3d in the course of redox cycling of coulometric titration cell (A), and the oxygen content measured by TGA and calculated from the coulometric titration data using the reference point at 1223 K (B).
Single-phase perovskite was formed for all samples, with the exception of La0.1Ce0.2Sr0.7FeO3d equilibrated at low temperature in air; however, fast cooling of this composition in air or quenching from elevated temperatures (1200e1300 K) resulted in monophasic material. The impurity phases in the equilibrated material were identified as CeO2- and SrFe12O19-based solid solutions (Fig. 3). Higher concentrations of niobium cations in La1xSrxFeO3d (x ¼ 0.5e0.7) resulted in the segregation of LaNbO4 as a second phase. The solubility of Nb and Ce in (La,Sr)FeO3d is, hence, substantially lower than that found for SrFeO3-based systems [14,15]. Unit-cell parameters, listed in Table 1, appear to be governed by opposing factors. Substituting La3þ with the 0.0
heating continuous regime, 5 K/min
-0.2
)
-0.4
m/m0 (
of the CT and TGA techniques. Consequently, the electrone hole equilibrium point at 1223 K was used as the reference state for the CT data array in the case of La0.3Sr0.7Fe0.90Nb0.10O3d. For the latter composition, the complete reduction procedure could not be applied due to oxygen nonstoichiometry of binary niobium oxide phases, which are expected to form in strongly reducing atmospheres [22e24]. In moderately reducing environments corresponding to the electronehole equilibrium points [9e11,19], all niobium cations are expected to be pentavalent [22,24]. The absolute value of the oxygen concentration in the Nb-containing samples was thus identified and used as reference data for the arrays of the coulometric titration data. The changes in total oxygen content, obtained by TGA and calculated from the CT data using the reference point at 1223 K, showed excellent agreement throughout the studied temperature range (Fig. 1B), thus confirming that the CT cells were absent of leaks. We note that, although stagnated oxygen nonstoichiometry variations in La0.3Sr0.7Fe0.90Nb0.10O3d on heating and cooling
-0.6
-0.8
cooling
-1.0
steps after each 50 K, 1-2 K/min -1.2
600
800
1000
1200
T, K Fig. 2. Reproducibility of the relative weight changes of La0.3Sr0.7Fe0.90Nb0.10 O3d powder in air.
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
larger Sr2þ cation tends to increase volume, whereas the replacement of Fe3þ by the small Nb5þ leads to a contraction in unit cell. On doping with Ce, the cell volume also increases. The rhombohedral distortion of the parent phase decreases on doping with either Ce or Nb. In addition, grain size decreases on doping, as revealed by the microstructures shown in Fig. 4. The dilatometric curves of La0.52xCexSr0.5þxFeO3d and La0.52ySr0.5þ2yFe1yNbyO3d ceramics were similar to other ferrite-based perovskites [8,9,11]; the average thermal expansion coefficients (TECs) are summarized in Table 1. A transition to a regime with higher expansion is observed above 730e950 K, resulting from chemically-induced expansion of the lattice on losing oxygen with associated reduction of iron. The average TEC values for the Ce-doped samples are significantly lower compared to La0.5Sr0.5FeO3d, suggesting lower oxygen losses on heating due to a higher level of oxygen deficiency at room temperature, in agreement with the Mo¨ssbauer-spectroscopy data.
La0.4Sr0.6Fe0.95Nb0.05O3-
Intensity (a.u.)
La0.3Sr0.7Fe0.90Nb0.10O3-
La0.3Ce0.1Sr0.6FeO3-
CeO2
SrFe12O19
3.2. Mo¨ssbauer spectra and oxygen stoichiometry of the La0.52ySr0.5þ2yFe1yNbyO3d series
La0.1Ce0.2Sr0.7FeO326
20
30
40
50
2
28
60
30
32
70
34
35
36
The Mo¨ssbauer spectrum of La0.4Sr0.6Fe0.95Nb0.05O3d at 4.1 K (Fig. 5) is similar to that of La0.5Sr0.5FeO3d, analyzed elsewhere [27,28]. However, if only two distributions of magnetic hyperfine fields (Bhf) are considered in the refinement, as in the case of La0.5Sr0.5FeO3d, the distribution with the lower hISi and hBhfi is bimodal and shows a strong increase of IS with increasing Bhf. A significantly better fit is obtained if this distribution is split in two. The parameters of the two resultant distributions (Table 2) are typical of Fe5þ and localized
80
°
Fig. 3. XRD patterns of the title series equilibrated at low temperatures in air.
Table 1 Properties of La0.52ySr0.5þ2yFe1yNbyO3d and La0.52xCexSr0.5þxFeO3d ceramics Material
Phase composition
Unit-cell parameters a, nm
a,
La0.5Sr0.5FeO3d
PðR3CÞ
0.54927
60.27
La0.4Sr0.6Fe0.95Nb0.05O3d
PðR3CÞ
0.54917
60.21
La0.3Sr0.7Fe0.90Nb0.10O3d La0.3Ce0.1Sr0.6FeO3d
PðR3CÞ PðR3CÞ
0.55054 0.54984
60.13 60.06
La0.1Ce0.2Sr0.7FeO3d
PðPm3mÞ þ I
0.38949
e
Material
La0.5Sr0.5FeO3d La0.4Sr0.6Fe0.95Nb0.05O3d La0.3Sr0.7Fe0.90Nb0.10O3d La0.3Ce0.1Sr0.6FeO3d La0.1Ce0.2Sr0.7FeO3d
to
Average TECs
a 106 ; K1
T, K 350e950 950e1310 340e780 780e1080
12.4 23.7 13.1 21.2
340e700 700e1080 340e730 730e1080
12.3 20.7 12.8 19.0
Total conductivity, S cm1
1173 K
1223 K 4
6.8 10
1.3 103
1.4 103 3.6 104
2.2 103 6.4 104
3.0 104
5.5 104
Activation energy for ionic conduction
1223 K
1123 K
1023 K
370 K
T, K
Ea, kJ/mol
166 90.5 55.4 41.5 e
243 134 72.6 56.5 33.1
335 192 103 74.0 40.7
239 179 e 22.1 4.80
1123e1223 e 1073e1223 1073e1223 1023e1223
109 13 e 86 9 106 8 100 14
All data are given for atmospheric oxygen pressure. P corresponds to the cubic or rhombohedrally distorted perovskite phase (space groups are given in parentheses). I indicates minor phase impurities based on CeO2 and SrFe12O19. a is the linear thermal expansion coefficient (TEC) averaged in the given temperature range. to is the oxygen ion-transference number.
36
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
Fig. 4. SEM micrographs of La0.52ySr0.5þ2yFe1yNbyO3d ceramics: y ¼ 0 (A), 0.05 (B), and 0.10 (C).
Fe4þ, with predominant j d4i character, at 4.1 K [28e32]. The presence of the latter state is associated with a lower number of (Fe eg e O 2p e Fe eg) electron transfer pathways due to Nb5þ incorporation; this leads to a more localized jd4i character of the Fe4þ electronic structure, reflected in the values of IS. Since the niobium concentration is low, the localized Fe4þ state is only observed for approximately 10% of the total iron content. The majority of Fe4þ cations, about 40% of the
Fig. 5. Mo¨ssbauer spectra of La0.5Sr0.5FeO3d (A), La0.4Sr0.6Fe0.95Nb0.05O3d (B), La0.3Sr0.7Fe0.90Nb0.10O3d (C), La0.3Ce0.1Sr0.6FeO3d (D) and La0.1Ce0.2Sr0.7FeO3d (E), taken at 4.1 K. The calculated function plotted on the experimental points is the sum of two or three distributions of magnetic hyperfine fields (Table 2). In the case of La0.1Ce0.2Sr0.7FeO3d, an additional sextet due to an impurity phase is also present.
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
37
Table 2 Parametersa obtained from the Mo¨ssbauer spectra of ferrite-based materials Composition La0.5Sr0.5FeO3d
T, K
Valence state of iron ions
IS,
(mm/s)
QS, 3 (mm/s)
, T
10
3þ 5þ 3þ 4þ Itinerant
0.42 0.02 0.29 0.10
0.03 0.03 0.12 0.09
3þ 4þ Localized 5þ 3þ 4þ Localized 4þ Itinerant
0.42 0.09 0.03 0.32 0.03 0.09
3þ 4þ Localized 5þ 3þ 4þ Localized 4þ Itinerant
297 La0.4Sr0.6Fe0.95Nb0.05O3d
4.1
297
La0.3Sr0.7Fe0.9Nb0.1O3d
4.1
297
La0.3Ce0.1Sr0.6FeO3d
4.1 297
La0.1Ce0.2Sr0.7FeO3d
4.1
297
G, mm/s
I%
Average charge of iron cations
50.7 28.2 e e
þ3.5(0)
0.36 0.36
75 25 50 50
0.00 0.20 0.02 0.11 0.26 0.14
49.5 30.6 26.5 e e e
e e e 0.53 0.48 0.40
68 10 22 48 10 42
þ3.5(4)
0.45 0.10 0.01 0.33 0.00 0.12
0.01 0.12 0.08 0.42 0.27 0.33
49.6 31.3 26.1 e e e
e e e 0.44 0.34 0.43
65 16 19 46 13 41
þ3.5(4)
3þ 4þ Localized 3þ 4þ Localized
0.44 0.05 0.30 0.04
0.00 0.03 0.31 0.16
50.9 26.5 e e
e e 0.57 0.39
75 25 76 24
þ3.2(5)
3þb 4þ 3þ In phase impurities 3þb 4þ 3þ In phase impurities
0.43 0.08 0.44 0.25 0.03 0.35 0.31
0.04 0.01 0.47 0.52 0.33 0.40 0.16
50.0 26 52.1 e e 41.8 50.9
e e e 0.65 0.40 0.28 0.48
70 21 9 67 18 7 8
þ3.2(3)
þ3.5(0)
þ3.5(2)
þ3.5(4)
þ3.2(4)
þ3.2(1)
a IS, G and QS are the isomer shift relative to metallic a-Fe at 297 K, line-widths and the quadrupole splitting of a doublet in the paramagnetic state, respectively. For the magnetically split spectra fitted with a distribution of sextets, the average isomer shift , the quadrupole shift 3 and the average magnetic hyperfine field are given. I is the relative area. Estimated errors are 0.2 T for , <2% for I, and 0.02 mm/s for the other parameters. Average charge of iron cations is given for the perovskite phases. b The signal includes a partial contribution from impurity phase(s).
iron content, are in an itinerant state at room temperature, with predominant j d5L1j character (L1 stands for a hole in the anion valence band). This state leads to charge disproportionation (Fe4þ / Fe3þ þ Fe5) on cooling down to 4.1 K, usually at the same temperature when magnetic ordering occurs [28,30,31,33]. A microscopic phase separation seems to occur, therefore, as in the case of cobalt- and aluminium-substituted ferrites [28,30]; the high fraction of Fe5þ contributions at 4.1 K indicates that the largest domains remain in the charge-ordered (CO) state. The same contributions are present in the La0.3Sr0.7Fe0.90Nb0.10O3d spectrum collected at 4.1 K. At 297 K, the spectra of La0.52ySr0.5þ2yFe1yNbyO3d are characterized by a significant overlapping of contributions (Fig. 6). Nonetheless, the fit of these spectra is improved when based on the low-temperature data and model in comparison to that of a simpler two-doublet refinement. The broadening of the room-temperature absorption peak on increasing Nb content may be analyzed by doublets with increasingly greater quadrupole splittings (QS), attributable to a higher disorder in the lattice. In agreement with the low-temperature data, at 297 K the three contributions correspond to Fe3þ, localized Fe4þ with dominant jd4i character, and itinerant Fe4þ. A slight increase in IS values for all contributions to both the 297 and
4.1 K spectra is observed in La0.3Sr0.7Fe0.90Nb0.10O3d relative to La0.5Sr0.5FeO3d (Table 2). This suggests a decrease in covalent character of the FeeO bonds due to their disruption by Nb5þ and the inductive effect of competing NbeO bond, namely strong polarization of the anion electronic charge by the high positive charge of Nb5þ cations. For all La0.52ySr0.5þ2yFe1yNbyO3d (x ¼ 0e0.10) perovskites, the iron oxidation states calculated from the Mo¨ssbauer spectra (Table 2) suggest that there is virtually no oxygen deficiency, within the limits of experimental uncertainty. Hence, the incorporation of Nb5þ, charge-compensated by Sr2þ doping, increases the degree of hole localization without significantly affecting the average oxidation state of iron. This behavior indicates that the lower solubility of Nb5þ in (La,Sr)FeO3d in comparison to SrFeO3d [15] may result from low tolerance of the former ferrite lattice with respect to extra oxygen intercalation, at least at low temperatures. A different situation is observed, however, in the hightemperature range, when the role of hole localization becomes greater due to thermal expansion and oxygen loss from the perovskite lattice. At temperatures above 923 K, the oxygen nonstoichiometry of La0.3Sr0.7Fe0.90Nb0.10O3d is higher than that of the parent compound, La0.5Sr0.5FeO3d, by 0.03e0.05
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
38
La0.3Sr0.7Fe0.9Nb0.1O3La0.5Sr0.5FeO3-
10%
2.95
A
2.90
10%
923 K
B
973 K
923 K 1073 K 973 K
10%
Relative Transmission
2.85
1023 K 2.80
C
1073 K 1123 K 1173 K 1223 K -5
-4
-3
-2
-1
0
log p(O2) (atm)
5%
Fig. 7. Comparison of the p(O2)eTed diagrams of La0.5Sr0.5FeO3d and La0.3Sr0.7Fe0.90Nb0.10O3d, determined by CT and TGA. Solid lines are for visual guidance only.
5%
D
E -8
-4
0
+4
+8
Velocity ( mm/s ) Fig. 6. Room-temperature Mo¨ssbauer spectra of La0.5Sr0.5FeO3d (A), La0.4Sr0.6 Fe0.95Nb0.05O3d (B), La0.3Sr0.7Fe0.90Nb0.10O3d (C), La0.3Ce0.1Sr0.6FeO3d (D) and La0.1Ce0.2Sr0.7FeO3d (E). The calculated function plotted on the experimental points is the sum of either two or three doublets (Table 2). For La0.1Ce0.2 Sr0.7FeO3d, sextets due to impurity phases were also considered.
oxygen atoms per formula unit (Fig. 7). A similar tendency was earlier observed for (La,Sr)(Fe,Al)O3d [28], which was associated with less-covalent character of the FeeOeFe bonds. One additional factor responsible for such behavior
is, most likely, the higher Sr2þ content which promotes local oxygen-vacancy ordering, as occurs in the (La,Sr)FeO3d system [10]; the redox equilibria is thereby shifted towards greater oxygen nonstoichiometry and a lower Fe oxidation state. These hypotheses are supported by the data on ionic and hole transport, discussed below. In particular, the decrease in total conductivity, predominantly p-type electronic, is much greater than would be expected from the substitution of 5e 10 at% Nb5þ for Fe cations, assuming that Nb had no influence on the iron oxidation states (Table 1). 3.3. Mo¨ssbauer spectra of the La0.52xCexSr0.5þx FeO3d series The spectra of La0.3Ce0.1Sr0.6FeO3d (Figs. 5 and 6) are similar to those of the Nb-containing perovskites. Analyses of the low-temperature spectrum show, however, that there are no contributions with IS or Bhf typical of Fe5þ (Table 2), suggesting that Fe4þ disproportionation does not occur down to 4.1 K. This indicates that no itinerant Fe4þ can be expected at 297 K. Indeed, a two-doublet refinement consistent with the 4.1 K analyses, leads to a good fit of the 297 K spectrum. No improvement of the refinement is achieved when the fitting of a third doublet, with IS in the range typical of itinerant Fe4þ, is performed.
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
3.4. Ionic transport The p(O2) dependencies of oxygen-permeation flux, j, and specific oxygen permeability, J(O2), for membranes of La0.3Ce0.1Sr0.6FeO3d of different thicknesses are shown in Fig. 8. The specific permeability is related to oxygen flux density by [34]:
p2 JðO2 Þ ¼ jd ln p1
1 ð1Þ
where d is membrane thickness, and p2 and p1 are feed-side and permeate-side p(O2), respectively. As this term is proportional to j d, its value is independent of thickness when surface-exchange limitations are insignificant but increases with thickness when exchange kinetics are slow [8,9]. For La0.1Ce0.2Sr0.7FeO3d, the specific permeability is essentially
A
La0.3Ce0.1Sr0.6FeO3-
log j mol s
cm
-7.4
p2 = 0.21 atm
-7.6
-7.8
-8.0
-8.2 d=1.00 mm 1223 K
-8.4
d=1.43 mm
1173 K
1223 K
B
-8.6
mol s
cm
1173 K
-8.8
log J
The content of tetravalent iron in La0.3Ce0.1Sr0.6FeO3d is low, in the range 0.25e0.26 per formula unit, which precludes the formation of any domains with itinerant Fe4þ. The phase impurities present in La0.1Ce0.2Sr0.7FeO3d give rise to the magnetic sextets observed in the 297 K Mo¨ssbauer spectrum (Fig. 6), accounting for w15% of the total Fe content in the sample. This iron is in the 3þ oxidation state and is magnetically ordered at room temperature, as expected for SrFe12O19 hexaferrite identified by XRD (Fig. 3). The impurities are more difficult to analyze in the 4.1 K spectrum because their contributions are of low intensity and overlap those from the likewise magnetically-ordered perovskite. The only readily perceived contribution from impurities arises from the highest Bhf component which gives rise to sharp peaks that overlap the distribution of magnetic splittings from Fe3þ in the perovskite phase. Consequently, the exact stoichiometry of the perovskite solid solution is not determinable. However, according to the estimated relative areas both at 4.1 and 297 K, the Fe4þ fraction within the main phase should be lower than 0.25 per formula unit, consistent with the absence of charge disproportionation at low temperatures. The results clearly show that Ce-containing compositions exhibit a substantially higher Fe3þ fraction and a lower degree of hole delocalization than the Nb-containing analogues. This suggests that the great majority, if not all, the Ce cations are in the tetravalent state. At the same time, the increase in oxygen nonstoichiometry on doping is higher than expected for these valence states. In contrast to the Nb-containing perovskites, La0.52xCexSr0.5þxFeO3d samples are oxygen-deficient even at low temperatures. The conductivity data (Table 1) are consistent with the Mo¨ssbauer-spectroscopy results on heating to 1223 K. Such behavior cannot be explained by only considering simple relationships between the valence of dopant and host cations. One possible hypothesis relates to the incorporation of Ce4þ on the B sublattice, compensated stoichiometrically by A-site deficiency: the charge of A-site vacancies should be compensated via the creation of oxygen vacancies. This hypothesis could explain the relatively low solubility of the cerium cations in the (La,Sr)FeO3d lattice with respect to SrFeO3d [14].
39
-9.0 0.2
0.4
0.6
0.8
1.0
1.2
log p2/p1 Fig. 8. Dependence of the oxygen-permeation fluxes (A) and specific oxygen permeability (B) of La0.3Ce0.1Sr0.6FeO3d ceramic membranes on the oxygen partial pressure gradient. The feed-side oxygen pressure ( p2) is 0.21 atm.
independent of membrane thickness, indicating that oxygen transport is mainly limited by the bulk ambipolar conductivity rather than surface-exchange kinetics. Oxide-ion-transference numbers (to) in air, determined by the FE method and calculated from the permeation data [8], are listed in Table 1. The low to values suggest that the limiting process in permeation is the bulk ionic conductivity; similar behavior was observed for all the Ce- and Nb-containing materials in the measured temperature range (1073e1223 K). The oxide-ion conductivities of the title materials were calculated from the to and total-conductivity data, and are shown in Fig. 9. The oxide-ionic conductivity decreases on addition of Ce and Nb dopants, whereas the activation energy remains essentially unchanged, analogous to the effect on ionic transport of Ti doping in the (La,Sr)FeO3d system [11]. On analyzing the data in Fig. 9, one should note that, in the case of undoped La0.5Sr0.5FeO3d, the permeation is limited by both ionic conductivity and surface-exchange kinetics [11]. However, the limiting effect of the surface processes diminishes with increasing temperature; hence, to provide a working comparison with the Ce- and Nb-substituted phases, only the oxide-ionic conductivity of La0.5Sr0.5FeO3d at high temperature (1123e 1223 K) is presented.
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
40
La0.5Sr0.5FeO3-0.4
1223 K
La0.3Sr0.7Fe0.9Nb0.1O3-
2.0
1123 K
La0.3Ce0.1Sr0.6FeO3-
1023 K
S cm
1.6 -1.2
1/6
log
log
(S/cm)
1073 K
La0.1Ce0.2Sr0.7FeO3-
-0.8
1173 K
-1.6 1.2
-2.0
1/4
p(O2) = 0.21 atm 8.0
8.5
9.0
9.5
10.0
104/T, K-1 Fig. 9. Temperature dependence of the oxygen ionic conductivity of the ferritebased materials at atmospheric oxygen pressure.
The similar activation energies for ionic conduction (Table 1) suggest that the composition of the title materials has essentially no effect on the ion-transport mechanism. The decrease in ionic conductivity with Ce and Nb additions and greater Sr content is likely, therefore, to result from a lower concentration of mobile species (ions and/or vacancies). The high-valence Ce4þ and Nb5þ cations may inhibit migration of neighboring oxygens due to stronger, attractive Coulombic forces. Also, the more extensive oxygen nonstoichiometry of the Sr-rich compositions is likely to result in considerable vacancy ordering and associated defect-interaction terms, as discussed above. In the case of Cedoped compositions, an additional consideration is that Ce4þ may occupy the B sites, with the associated formation of Asite vacancies. In this instance, oxide-ion transport may be blocked due to local lattice distortions arising from the unoccupied A sites. The deleterious effect on oxide-ionic transport of doping with Nb is relatively minor. A possible reason for this is that local distortions in the vicinity of the Nb5þ cations may trap the neighboring oxygen anions, but prevent partial ordering in the oxygen sublattice [19]. 3.5. Electronehole conduction The total conductivity of La0.3Sr0.7Fe0.90Nb0.10O3d as a function of p(O2) in the oxidizing regime, p(O2) > 105 atm, is shown in Fig. 10. The p(O2) dependence of conductivity may be described by a p(O2)þ1/4 power law, indicative of a predominantly p-type electronic conduction mechanism. All the studied materials show similar behavior, typical of (La,Sr)FeO3d-based materials [9e11,19]. Table 1 lists selected total-conductivity values for a series of temperatures which indicate that, also in common with the parent La0.5Sr0.5FeO3d phase, the conductivity of the title materials becomes pseudo-
La0.3Sr0.7Fe0.9Nb0.1O30.8 -5
-4
-3
-2
-1
0
log p(O2) (atm) Fig. 10. Oxygen partial pressure dependence of the total conductivity of La0.3Sr0.7Fe0.90Nb0.10O3d.
metallic at high temperature. This transition is attributable to a decrease in concentration of p-type charge carriers (Fe4þ) which occurs on heating as a result of oxygen loss (Fig. 7). Supporting evidence for the nature of the conduction mechanism in oxidizing atmospheres is provided from Seebeckcoefficient (S ) data, presented in Fig. 11 for La0.1Ce0.2Sr0.7 FeO3d as a function of p(O2). The positive S sign, increasing with decreasing p(O2), is characteristic of p-type hole transport. Hence, the effect that doping with Ce and Nb has on the redox behavior of LSF does not extend to modifying the electronic conduction mechanism. The total conductivity and hole mobility as a function of the fraction of Fe4þ cations are shown in Fig. 12. The hole mobility (mp) was evaluated from [24,35]: sp ¼ epmp Nfu =Vuc
ð2Þ
where Vuc is the unit-cell volume calculated from the XRD data, Nfu is the number of formula units per unit cell, and p is the concentration of p-type charge carriers per formula unit. This last quantity was assumed to be equivalent to the concentration of Fe4þ; possible contributions from Ce3þ/4þ and Nb4þ/5þ were neglected, as the stability of these redox couples under oxidizing conditions is usually very low (e.g. [16,24] and references therein). Moreover, these couples should give rise to n-type electronic conduction, but the mobility of electrons is anticipated to be negligible. Instead, Ce and Nb cations in low oxidation states may be expected to act as hole traps. The temperature-dependence of hole mobility at a fixed hole concentration (Fig. 12) further validates that the conduction
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42 2.1
A
La0.1Ce0.2Sr0.7FeO3-
1.4
41
A
1.8
S cm
S cm
1.2 1.5
1223 K
1/6
1123 K
1.2
1/4
0.8
1173 K
log
log
1.0
1073 K 1023 K 0.9
0.6 240
1223 K 1173 K
B
B
1123 K
200
0.10
(cm2 V-1 s-1)
1073 K 1023 K
-1
V K )
La0.3Sr0.7Fe0.9Nb0.1O3-
0.12
0.08
p
S
160
120
80
0.06
-R/6F
-4
-3
-2
-1
0
log p(O2) (atm) Fig. 11. Oxygen partial pressure dependence of the total conductivity (A) and Seebeck coefficient (B) of La0.1Ce0.2Sr0.7FeO3d ceramics.
mechanism involves p-type transport, i.e. thermally-activated polaron hops. Accordingly, the mobility values under moderately oxidizing atmospheres are below the accepted threshold of w0.1 cm2 V1 s1, which is considered to separate smallpolaron and broadband conduction. The decrease in the hole mobility, which is observed on incorporation of either Ce or Nb in the La0.5Sr0.5FeO3d parent phase, is likely to be associated with several factors. These include the increase in oxygen-vacancy concentration and, in the case of Nb-doping, decrease in the total concentration of Fe cations, both of which break up, to some extent, the network of conducting FeeOeFe pathways. Additionally, local distortions near Nb have been observed by extended X-ray absorption fine structure (EXAFS) data in the related Sr(Fe,Nb)O3d system, and may promote localization of holes [15]. Similar effects are expected on Ce4þ incorporation in the B sites, which may also disrupt the FeeOeFe bonding.
4. Conclusions The effects of Ce- and Nb-doping of the candidate membrane-reactor material, (La,Sr)FeO3d, have been appraised with the objective of improving thermodynamic and
0.04
0.05
0.10
0.15
0.20
0.25
0.30
0.35
[Fe4+] / [Fe]total Fig. 12. Total conductivity and hole mobility of La0.3Sr0.7Fe0.90Nb0.10O3d as a function of the [Fe4þ]/[Fe] fraction.
dimensional stability under the large chemical-potential and thermal gradients encountered during membrane operation. Mo¨ssbauer spectroscopy, coulometric titration and totalconductivity studies of the selected La0.52xCexSr0.5þxFeO3d and La0.52ySr0.5þ2yFe1yNbyO3d systems indicate that doping with Ce and, to a lesser extent Nb, promotes localization of electron holes and increases oxygen nonstoichiometry. The latter effect may be partly associated with Ce4þ incorporation on the B sublattice of the perovskite and local vacancy ordering arising from the concomitant increase in Sr content, in agreement with ionic-conductivity data. Electronic conduction in oxidizing conditions involves ptype electronic transport as indicated by the p(O2) dependencies of total conductivity and Seebeck coefficient, the positive sign of thermopower and relatively low, temperature-activated hole mobility estimated using the oxygen nonstoichiometry data. The incorporation of Ce or Nb does not alter the conduction mechanism but lowers electronic conductivity, primarily due to a decrease in the concentrations of charge carriers and FeeOeFe bonds. Oxygen-permeation through the Nb- and Ce-containing membranes is limited by the bulk ambipolar conductivity rather than surface-exchange kinetics. Oxide-ion conductivity decreases with increasing niobium and cerium content, whilst
42
V.V. Kharton et al. / Solid State Sciences 9 (2007) 32e42
oxide-ion-transference numbers are higher in the Nb-doped system in comparison to the parent (La,Sr)FeO3d material. Although doping with Ce considerably lowers ionic and electronic conductivities, surface modification of ceramic membranes with Ce may still be of interest for improving catalytic properties. Acknowledgements This work was supported by the NATO Science for Peace program (project 978002), the FCT, Portugal (projects POCI/ CTM/58570/2004, SFRH/BPD/15003/2004 and SFRH/BPD/ 17649/2004), and the Russian Foundation for Basic Research (Grant 06-03-33099-a). We would like to thank A. Shaula, I. Marozau, D. Logvinovich and Y. Pivak for experimental assistance and helpful discussions. References [1] J. Kilner, S. Benson, J. Lane, D. Waller, Chem. Ind. (17) (1997) 907. [2] A.P.E. York, T. Xiao, M.L.H. Green, Top. Catal. 22 (2003) 345. [3] D.J. Wilhelm, D.R. Simbeck, A.D. Karp, R.L. Dickenson, Fuel Process. Technol. 71 (2001) 139. [4] T.J. Mazanec, in: H.U. Anderson, A.C. Khandkar, M. Liu (Eds.), Ceramic Membranes I, The Electrochemical Society, Pennigton, NJ, 1997, PV95-24, p. 16. [5] M. Schwartz, J.H. White, A.F. Sammells, U.S. Patent 6,214,757, 2001. [6] J.E. ten Elshof, H.J.M. Bouwmeester, H. Verweij, Solid State Ionics 81 (1995) 97. [7] J.E. ten Elshof, H. Verweij, Solid State Ionics 89 (1996) 81. [8] V.V. Kharton, A.L. Shaulo, A.P. Viskup, M. Avdeev, A.A. Yaremchenko, M.V. Patrakeev, A.I. Kurbakov, E.N. Naumovich, F.M.B. Marques, Solid State Ionics 150 (2002) 229. [9] V.V. Kharton, A.A. Yaremchenko, A.L. Shaula, A.P. Viskup, F.M.B. Marques, J.R. Frade, E.N. Naumovich, J.R. Casanova, I.P. Marozau, Defect Diffus. Forum 226e228 (2004) 141. [10] M.V. Patrakeev, J.A. Bahteeva, E.B. Mitberg, I.A. Leonidov, V.L. Kozhevnikov, K.R. Poeppelmeier, J. Solid State Chem. 172 (2003) 219. [11] E.V. Tsipis, M.V. Patrakeev, V.V. Kharton, A.A. Yaremchenko, G.C. Mather, A.L. Shaula, I.A. Leonidov, V.L. Kozhevnikov, J.R. Frade, Solid State Sci. 7 (2005) 355.
[12] V.V. Kharton, A.V. Kovalevsky, E.V. Tsipis, A.P. Viskup, E.N. Naumovich, J.R. Jurado, J.R. Frade, J. Solid State Electrochem. 7 (2002) 30. [13] Q. Ming, J. Huang, Y.L. Yang, M. Nersesyan, A.J. Jacobson, J.T. Richardson, D. Luss, Combust. Sci. Technol. 138 (1998) 279. [14] N.E. Trofimenko, H. Ullmann, J. Paulsen, R. Muller, Solid State Ionics 99 (1997) 201. [15] M.J. Akhtar, Z.N. Akhtar, J.P. Dragun, C.R.A. Catlow, Solid State Ionics 104 (1997) 147. [16] R.J. Panlener, R.N. Blumenthal, Thermodynamic study of nonstoichiometric cerium dioxide, U.S.A.E.C., 1972, Report no. COO-1441-18, p. 1. [17] G.C. Mather, J.R. Jurado, Bol. Soc. Esp. Cera´m. Vidrio 42 (2003) 311. [18] M.V. Patrakeev, I.A. Leonidov, V.L. Kozhevnikov, V.V. Kharton, Solid State Sci. 6 (2004) 907. [19] M.V. Patrakeev, E.B. Mitberg, A.A. Lakhtin, I.A. Leonidov, V.L. Kozhevnikov, V.V. Kharton, M. Avdeev, F.M.B. Marques, J. Solid State Chem. 167 (2002) 203. [20] M.V. Patrakeev, E.B. Mitberg, A.A. Lakhtin, I.A. Leonidov, Ionics 4 (1998) 191. [21] J. Mizusaki, Solid State Ionics 52 (1992) 79. [22] B.C.H. Steele, Ph.D. thesis, London University, 1965. [23] C.B. Alcock, S. Zador, B.C.H. Steele, Proc. Brit. Ceram. Soc. (8) (1967) 231. [24] P. Kofstad, Nonstoichiometry, Diffusion and Electrical Conductivity in Binary Metal Oxides, Wiley-Interscience, New York, 1972. [25] J.C. Waerenborgh, M.O. Figueiredo, J.M.P. Cabral, L.C.J. Pereira, Phys. Chem. Miner. 21 (1994) 460. [26] J. Hesse, A. Rubartsch, J. Phys. E 7 (1974) 526. [27] S.E. Dann, D.B. Currie, M.T. Weller, M.F. Thomas, A.D. Al-Rawwas, J. Solid State Chem. 109 (1994) 134e144. [28] V.V. Kharton, J.C. Waerenborgh, A.P. Viskup, S.O. Yakovlev, M.V. Patrakeev, P. Gaczyn´ski, I.P. Marozau, A.A. Yaremchenko, A.L. Shaula, V.V. Samakhval, J. Solid State Chem. 179 (2006) 1273. [29] P. Adler, S. Eriksson, Z. Anorg. Allg. Chem. 626 (2000) 118. [30] P. Adler, S. Ghosh, Solid State Sci. 5 (2003) 445. [31] J.C. Waerenborgh, D.P. Rojas, A.L. Shaula, G.C. Mather, M.V. Patrakeev, V.V. Kharton, J.R. Frade, Mater. Lett. 59 (2005) 1644. [32] P. Adler, J. Mater. Chem. 9 (1999) 471. [33] J.B. Yang, X.D. Zhou, Z. Chu, W.M. Hikal, Q. Cai, J.C. Ho, D.C. Kundac¸iya, W.B. Yelon, W.J. James, H.U. Anderson, H.H. Hamdeh, S.K. Malik, J. Phys.: Condens. Matter 15 (2003) 5093. [34] H.-H. Mo¨bius, in: Extended Abstracts of 37th Meeting of the International Society of Electrochemistry, vol. 1, Vilnius, Lithuania, 1986, p. 136. [35] J.B. Goodenough, Solid State Electrochemistry, in: P. Bruce (Ed.), Cambridge University Press, Cambridge, 1995, p. 43.
