Diffusion of Nd and Mo in lanthanum tungsten oxide

Solid State Ionics 274 (2015) 128–133

Contents lists available at ScienceDirect

Solid State Ionics journal homepage: www.elsevier.com/locate/ssi

Diffusion of Nd and Mo in lanthanum tungsten oxide Einar Vøllestad a, Markus Teusner b, Roger A. De Souza b, Reidar Haugsrud a,⁎ a b

University of Oslo, Department of Chemistry, Centre for Materials Science and Nanotechnology, FERMiO, Gaustadalleen 21, 0349 Oslo, Norway Institute of Physical Chemistry, RWTH Aachen University, Landoltweg 2, 52056 Aachen, Germany

a r t i c l e

i n f o

Article history: Received 18 November 2014 Received in revised form 9 March 2015 Accepted 9 March 2015 Available online xxxx Keywords: Lanthanum tungstate Cation diffusion SIMS Degradation Tracer diffusion

a b s t r a c t Cation diffusion in functional oxides exposed to electrochemical gradients may lead to kinetic demixing or decomposition and, consequently, determine the life-time of the functional component. Here we present chemical diffusion coefficients of Nd and Mo in the mixed proton–electron conductor lanthanum tungsten oxide, La28 − xW4 + xO54 + 3x/2 (LWO), measured at 1000 to 1200 °C in both oxidizing and reducing atmospheres. The bulk diffusivities of Nd and Mo were similar at all temperatures investigated and did not change significantly from oxidizing to reducing conditions. On these bases it is suggested that bulk diffusion of both Nd and Mo occurs via the La2 site on which both cations reside. Based on the low activation energy for bulk transport (~ 200 kJ∙mol−1) at temperatures below 1200 °C it is proposed that the cation defect concentrations are, in effect, frozen in. Preferential diffusion of Nd along the grain boundaries was rationalized based on space charge effects and depletion of W6 + and Mo6+ near the positively charged grain boundary core. Potential implications of kinetic demixing or decomposition of LWO membranes are also evaluated based on the present results. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Proton conducting oxides are interesting materials for the active component in diverse hydrogen based technologies, such as fuel cells, electrolyser cells, and membrane reactors for hydrogen separation and purification [1–3]. In operation, proton conducting membranes are exposed to elevated temperatures and gradients in hydrogen activity, which in most cases results in a driving force for cation migration. Although it is slow, transport of cations often determines the lifetime of functional oxides, as it can result in kinetic demixing or decomposition, and morphological instabilities on the membrane surfaces [4–10]. Investigations of the diffusion properties of the cations are therefore required to determine the durability of proton conducting oxides serving in electrochemical potential gradients. Several contributions are available on cation diffusion in oxide ion conducting oxides [11–17], but for proton conductors such studies are still scarce. Among proton conductors other than those based on perovskites, lanthanum tungsten oxide with the formula La28 − xW4+ xO54+ 3x / 2 has as of late been the subject of many studies [18–24]. Lanthanum tungsten oxide (LWO) is essentially a pure proton conductor up to 600 °C under wet conditions [19,25], and displays significant electronic conductivity above 750 °C under oxidizing and reducing conditions. The mixed proton- and electron conduction yield an appreciable hydrogen ⁎ Corresponding author. E-mail address: [email protected] (R. Haugsrud).

http://dx.doi.org/10.1016/j.ssi.2015.03.011 0167-2738/© 2015 Elsevier B.V. All rights reserved.

flux at these high temperatures [26–28]. Partial substitution of W6 + with Mo6+ results in an increased electronic conductivity without significantly affecting the proton conductivity [29], and Mo substituted LWO was recently shown to exhibit an unprecedented hydrogen permeation rate of 0.05 mL min−1 cm−2 at temperatures down to 700 °C [30]. Thus, LWO based materials may find potential application both as a mixed proton–electron conducting membrane for hydrogen separation and – in the regime with essentially pure proton conduction – as an electrolyte in fuel- and electrolyser cells. Investigation of cation diffusion in LWO is accordingly called for to evaluate its long term stability towards kinetic demixing and decomposition in such applications. A recent study on cation interdiffusion in LWO revealed similar bulk diffusion coefficients for La and W between 1200 and 1350 °C [31], indicating that both cations migrate by the same diffusion mechanism. Enhanced diffusion along the grain boundaries was only observed for La diffusion, but the possibility of W diffusion along grain boundaries could not be discounted due to experimental limitations inherent to the interdiffusion technique. In this study we examine cation transport in LWO at lower temperatures, from 1000 to 1200 °C, and determine diffusion profiles by Timeof-Flight Secondary Ion Mass Spectrometry (ToF-SIMS). Specifically we examine diffusion from a thin-film, Nd27W3.5Mo1.5O55.5 (NWMO) surface coating into LWO. Since the ionic radii and nominal valences of Nd and Mo are similar to La and W, respectively, this is regarded as a chemical tracer experiment. It also allows for the simultaneous determination of the diffusivity of both cations. The lower temperature and the

E. Vøllestad et al. / Solid State Ionics 274 (2015) 128–133

use of ToF-SIMS principally yield a better starting point for determining grain boundary transport. Such data is essential to firmly evaluate long term stability of LWO towards cation diffusion induced degradation under conditions relevant for applications. 2. Experimental 2.1. Sample preparation Powders of La27W5O55.5 and Nd27W3.5Mo1.5O55.5, synthesized by spray pyrolysis (CerPoTech, Norway), were used to manufacture dense polycrystalline substrates and the dense target for thin film deposition, respectively. The powders were uniaxially cold-pressed to pellets and sintered at 1550 °C for 3 h, yielding relative densities above 97%. The dense pellets were ground with SiC paper down to 4000 grit to obtain a uniform surface. The structure and phase purity was confirmed with XRD analysis. For the LWO substrates, subsequent polishing with diamond dispersions down to particle size of 0.1 μm gave a surface roughness of 20–40 nm, as determined from interference light microscopy (Wyko NT1100 Interferometer, Veeco, Inc. Plainview, NY). Thin films of NWMO were deposited on the LWO substrates with a Pulsed Laser Deposition (PLD) system from SURFACE (Germany). A KrF eximer laser with energy 300 mJ and a frequency of 7 Hz was utilized. The films were deposited in 0.05 mbar of O2 with a substrate temperature of 600 °C and target-substrate distance of 6.5 cm. Film thicknesses between 60 and 100 nm were obtained. The films crystallized in the fluorite-type F43m structure as previously reported for PLD deposited LWO films [32]. 2.2. Diffusion annealing Before the polished substrates were coated with the NWMO film, the substrates were pre-annealed under the same condition as the subsequent diffusion anneal with duration tpre − anneal ≥ 2tdiffusion − anneal. The specimens were sectioned in two; one piece was kept as a reference, while the other was used in the diffusion annealing. The annealing was done in a Probostat™ (NorECs, Norway) in wet (~2.8% H2O) air or wet 5% H2 (balanced by Ar) at temperatures ranging from 1000 °C to 1200 °C for diffusion times varying from 5 to 30 h. 2.3. SIMS analysis Secondary Ion Mass Spectrometry (SIMS) is a powerful tool to determine the distribution of isotopes and elements in solids [33]. In this study, depth profiling was utilized to monitor the intensity profiles of

3. Results 3.1. Evaluation of diffusion profiles Fig. 1 presents typical intensity profiles obtained by using a Cs+ sputtering beam. The intensities are normalized to the 18O− matrix signal to correct for minor machine instabilities during the measurement. Based on the regular shape of the intensity profiles it is assumed that the measured intensities are proportional to the concentrations of the corresponding cations. Fig. 1a presents the intensity profiles of a “zero-time” specimen which has only been subjected to the preannealing procedure. The NWMO film is clearly visible at shallow penetration depths (up to ~100 nm), followed by a sharp interface with the LWO substrate. In Fig. 1b, typical intensity profiles after diffusion annealing are presented. Significant in-diffusion of Nd and Mo into the LWO substrate is evident at penetration depths larger than 100 nm. Analysis of the intensity profiles requires closer consideration of the boundary conditions for the diffusion process. The tracer film and LWO substrate comprise a diffusion couple where interdiffusion of the cations is expected. Interdiffusion of La/Nd and W/Mo may also result in a directed flux of cation vacancies and change in the chemical composition in the substrate. The derived coefficients should accordingly be considered as chemical interdiffusion coefficients. A proper description of composition-dependent chemical diffusion coefficients requires a formal Boltzmann-Matano analysis of all four mobile cations on three different lattice sites with varying valence states. Our previous study of interdiffusion in LWO, using Nd and Mo as tracers for La and W, revealed that there is essentially no compositional dependence on the cation interdiffusion coefficients [31]. The diffusion profiles will therefore be analysed under the assumption that the diffusion coefficients are constant within the diffusion zones.

WO-

10

10-1 10-2

NdO10-3

MoO0

200

400

600

Depth (nm)

800

1000

101

-

LaO 0

Normalized Intensity (Ix / IO )

-

10-4

the chemical tracer cations (Nd and Mo) after diffusion annealing. Ei+ ther O+ 2 or Cs beams, both with energy 2 keV, were used for sputtering and raster scanned over an area of 250 μm × 250 μm. Positive secondary ions were monitored during depth profiling when sputtering with O+ 2 , whereas negative ions were monitored with a Cs+ beam. The secondary ions for analysis were generated by a 15 keV Ga+ ion beam, operated in bunch mode with a cycle time of 50 μs and rastered over an area of 80 μm × 80 μm with a resolution of 128 × 128 points [33]. Considering that the average grain size of the LWO substrates was approximately 10 μm, the analysed area should consist of several grains and grain boundaries. The sputtering time scale was converted to a depth scale by determining, post-analysis, the sputter crater depth by means of interference light microscopy.

b)

101

-

Normalized Intensity (Ix / IO )

a)

129

100

Nd diffusion : Dbulk = 5 x 10-16 cm2 s-1

10-1 10-2

Mo diffusion : Dbulk = 2 x 10-15 cm2 s-1

10-3 0

500

1000

1500

Depth [nm]

Fig. 1. Intensity profiles for a zero-time (a) and a diffusion annealed specimen (b). The red lines in (b) indicate fitted curves using Eqs. (1) and (2) for Nd and Mo diffusion, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

E. Vøllestad et al. / Solid State Ionics 274 (2015) 128–133

2    x : IðxÞ ¼ K exp − pffiffiffiffiffiffiffiffiffi 2 Dt

ð1Þ

Here, I is the relative intensity of the analysed element, K is a constant, x is the depth, t is the diffusion time and D is the effective diffusion coefficient. The red lines in Fig. 1b represent the best fit of Eq. (1) to the intensity profiles. For large depths there are distinct differences between the intensity profiles for the two cations; the Mo intensity approaches the background concentration rapidly, whereas a long diffusion tail is observed for the Nd profiles. The latter is attributed to enhanced grain boundary diffusion and in accordance with Harrison type-B kinetics. The analytical treatment of grain boundary diffusion is traditionally performed by means of Whipple-Le Claire analysis, in which the logarithm of the intensity is plotted against x6/5. In this work however, we use a solution derived by Chung and Wuensch [34] which is specifically derived for SIMS depth profiles: " 3=2

s  ω  Dgb ¼ Dbulk  t

1=2

10

A

  # ∂ lnI B : − 6=5 ∂η

ð2Þ

Here, Dgb and Dbulk are the grain boundary and bulk diffusion coefficients, respectively. ω is half the grain boundary width and s denotes the segregation factor, which can only be assumed to be unity for selfdiffusion. The term − ∂∂ ηln6=5I is the slope in a plot of ln I versus the dimenpffiffiffiffiffiffiffiffiffiffiffiffiffi sionless parameter η6/5 where η ¼ x= Dbulk t . The parameters A and B are obtained from tabulated values calculated by Chung and Wuensch

0 T = 1150 °C -1 Nd

-2

DNd = 3 x 10-13 cm2 s-1 gb

-

Fig. 1 shows that the amount of tracer ions in the thin film at the LWO/NWMO interface is reduced after the diffusion annealing compared to the zero-time specimen. Beyond the interface there is a decrease in intensity of the tracer ions typical of a radiotracer experiment. These are characteristics of diffusion from an instantaneous source, with the corresponding analytic solution to the diffusion equation:

Ln (Ix / IO )

130

-3 -4 -5 -6

Mo 0

10

20

η

30

40

6/5

Fig. 2. Logarithmic intensity profiles of Nd and Mo (background subtracted) versus η6/5. The slope in the region 6 b η b 10 (red line) is used to extract grain boundary diffusion coefficients from the Nd profiles. The bulk diffusion fits are indicated with the green lines. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 presents an Arrhenius plot of the grain boundary diffusion coefficients for Nd diffusion in air. The grain boundary diffusivities are three to four orders of magnitude larger than the bulk diffusivities, and exhibit an activation energy of (250 ± 90) kJ·mol−1. There is appreciable scatter in the grain boundary diffusivities relative to a linear Arrhenius behaviour, especially under reducing conditions. The error bars included for the grain boundary diffusivity comprise both the errors in determination of the slope − ∂∂ ηln6=5I and those associated with the bulk diffusion coefficients (cf. Eq. (3)). The surface roughness was significantly higher for the samples annealed under reducing conditions, which yields increased uncertainties for the determination of the grain boundary diffusion coefficients.

for different ranges of the slope − ∂∂ ηln6=5I [34]. Eq. (2) is only valid for 6 b η b 10, and when the dimensionless parameter β, β≡

ω  Dgb 3=2

Dbulk  t1=2

;

4. Discussion ð3Þ

is between 1 and 105. Fig. 2 presents representative profiles of logarithmic normalized intensities versus η6/5. There is a linear region for deep penetration depths in the Nd profiles, whereas the shapes of the Mo profiles do not display the same characteristic feature. Consequently, grain boundary diffusion coefficients are only determined for Nd. 3.2. Diffusion coefficients Fig. 3 presents logarithmic bulk diffusion coefficients for Nd and Mo diffusion in LWO annealed in air (a) and 5% H2 (b), both wet (pH2 O = 0.028 atm), as a function of reciprocal temperature. The error bars represent the maximum expected error, which includes errors from the sputter crater depth determination and diffusion occurred during ramps up and down in temperature. The Mo diffusivities are approximately one order of magnitude higher than the Nd diffusivities under oxidizing conditions. Both Nd and Mo diffusion display Arrhenius type behaviour with activation energies of approximately 200 and 250 kJ·mol−1, respectively. Under reducing conditions, the diffusivities of both species are essentially identical at all temperatures investigated, with activation energies of ~250 kJ·mol−1. Comparing Fig. 3a and b, we also note that the bulk diffusivities are essentially unaffected by the large change inpO2 going from 0.21 (air) to ~10−15 atm (5% H2).

4.1. Bulk diffusion The results reported herein reveal similar activation energies for bulk diffusion of Nd and Mo in LWO, with diffusion coefficients essentially independent of pO2 . Under reducing conditions the bulk diffusivities are essentially identical, whereas the Mo diffusion coefficients are approximately one order of magnitude higher than the Nd diffusivities under oxidizing conditions. The differences are, however, within the experimental error. In our previous work we suggested that the two cations migrate by the same mechanism [31]. Since the actual values for the Nd and Mo diffusion coefficient and their activation energies are comparable, the present data set supports this interpretation. The significant difference in valence and size of these cations would be expected to result in more distinct differences with separate migration trajectories. The present results are, therefore, discussed based on the assumption of a common diffusion mechanism. To promote the understanding of the diffusion mechanisms in LWO, a schematic of the defective fluorite-type structure of LWO [21,35] is presented in Fig. 5. LWO has three different cation sites, denoted La1, La2 and W. Combined experimental and theoretical calculations have revealed that 1/5 of the tungsten atoms in LWO reside on La2 sites [23,36]. Since the La2 sites are nearest neighbours both with W and La and form a continuous network for transport (cf. Fig. 5), we postulate that the rate determining step for both La and W diffusion is jumps into, or out of, vacant La2 sites.

E. Vøllestad et al. / Solid State Ionics 274 (2015) 128–133

a) -12

Ea = (240 ± 70) kJ mol

-1

H2

-13 -1

Mo

Ea = (250

Temperature (°C) 1100

±

90) kJ mol

-1

5% H2

Nd

-14

2

2

-14

1200

-12

Air

-1

log ( Dbulk / cm s )

1000

log ( Dbulk / cm s )

Mo

-13

b)

Temperature (°C) 1100

1200

131

-15 -16 -17

Nd

Ea = (200 ± 70) kJ mol Nd

-18

0.70

-1

0.75 3

-1

0.80

-15

Mo

-16 -17 -18

0.68

0.72 3

-1

-1

0.76

-1

10 T (K )

10 T (K )

Fig. 3. Arrhenius plots of Nd (▲) and Mo (□) bulk diffusion coefficients measured in air (a) and 5% H2 (b).

If we compare directly the present data set of bulk diffusivities with the corresponding interdiffusion coefficients from [31] (Fig. 6), there is good agreement between the values at overlapping temperatures. Interestingly, the temperature dependence increases across the experimental window (from 1000 to 1350 °C) covering both investigations. Assuming Arrhenius behaviour, the activation energy increases accordingly, from approximately 200 below 1200 °C to above 400 kJ·mol−1 at the higher temperatures. One would expect more than a single Arrhenius slope since grain boundary diffusion – with significantly lower activation energy than bulk – generally starts to dominate at lower temperatures. However, since these experimental approaches ideally yield decoupled bulk and grain boundary diffusion coefficients, a change from grain boundary to bulk transport with increasing temperature cannot explain the deviation from a straight-line Arrhenius behaviour in Fig. 6. The activation energy for cation diffusion comprises the enthalpy of formation and migration of cation vacancies. In the low temperature regime (b1200 °C), equilibration of the cation defects is slow compared to the equilibrium concentrations established during sintering at 1500 °C where the cations are considerably more mobile. Thus, one may speculate whether the concentration of cation vacancies is, in effect, frozen in below 1200 °C, and that the activation energy thereby represents only the migration enthalpy at these low temperatures. This behaviour has previously been reported for tracer diffusion in the perovskite

Temperature (°C) -8

1200

1100

La0.9Sr0.1Ga0.9Mg0.1O2.9, where cation defects were regarded as frozen in at temperatures 300° lower than the sintering temperature [11]. Following this logic, the migration enthalpy for cation diffusion is in the order of 200–250 kJmol−1, and the formation enthalpy for the defect by which the cations migrate is also in the order of 200–250 kJmol−1. In comparison, estimated formation and migration enthalpies for cation diffusion in the well-studied oxygen deficient fluorite yttria-stabilized zirconia (YSZ) are in the order of 100–150 and 400–500 kJmol−1, respectively [17,37]. We note that the formation enthalpy in YSZ is comparable, but slightly lower than what is estimated for LWO, whereas the migration enthalpy in YSZ is larger. The path for cation migration in the fluorite structure passes a saddle point between two nearestneighbouring oxygen atoms, and it has been proposed that the large migration enthalpy is mostly due to the large charge density the cations must surpass to make a successful jump [37]. In LWO, the La2 sites are 7-coordinated in a distorted cubic environment as a result of the defective oxygen sublattice. One may therefore speculate whether the relatively low migration enthalpy in LWO can be related to the departure from local cubic symmetry around the La2 sites, effectively reducing the charge density through which the cation migrate. 4.2. Grain boundary diffusion in LWO Analysis of the diffusion profiles in Fig. 4 shows that there is an enhanced diffusion of Nd along the grain boundaries in LWO, with activation energy in the order of 250 kJ·mol−1. Conversely, there were no indications of such enhanced grain boundary diffusion of W. These

1000

g.b.

-1

2

-1

Log (Dg.b. / cm s )

Ea = (250 ± 90) kJ mol -10

5% H2

-12

Air

-14 0.70

0.75 3

-1

0.80

-1

10 T (K ) Fig. 4. Arrhenius plot of the grain boundary Nd diffusion coefficients measured in air (open symbols) and 5% H2 (closed symbols), both wet.

Fig. 5. Illustration of the three different cation sites in the LWO structure, taken from [35]. The La2 site depicted here is edge sharing with other La2 sites in the b-direction, forming a continuous network for transport.

132

E. Vøllestad et al. / Solid State Ionics 274 (2015) 128–133

region [41,43]. Similar findings were recently reported for proton conducting oxides with resistive grain boundaries [44]. As of yet, there are no reports on the composition or structure of grain boundaries in LWO. Still, we may speculate on the effect of space charge layers on cation diffusion based on the results presented herein and the studies on similar fluorite-type oxides. A positively charged grain boundary core in LWO (i.e. accumulation of oxygen vacancies within the core) is likely to depress the concentration of W and Mo atoms on La2 sites with an effective positive charge. As a result, one would expect a lower diffusivity of W or Mo along the grain boundaries, in accordance with our observations. Moreover, the effectively negative cation vacancies will react to the electric potential induced by the depleted oxygen vacancies close to the grain boundary core, resulting in an accumulation of mobile cation vacancies that can facilitate faster diffusion of La and Nd along the grain boundaries. 4.3. Implications on the kinetic stability of LWO

Deff ¼ ð1−r ÞDbulk þ rDgb :

ð4Þ

Kinetic demixing or decomposition is rate limited by the faster moving species. As enhanced diffusion along the grain boundaries is only observed for La, one would expect that La is the faster diffusing cation species in LWO. Accordingly, accumulation of La on the high pO2 -side

0

120

100

0

0 140

-8

0 160 0

Temperature (°C) 180

findings consequently support the observations in our previous work [31]. We also note that the activation energy for grain boundary diffusion of Nd is almost identical to that of bulk transport. Similar behaviour has previously been reported for cation diffusion in zirconia and lanthanum nickel oxide [13,38,39]. Two potential causes for the similar activation energies between bulk and grain boundary transport have been suggested; i) a change in oxidation state of the diffusing cation or ii) the grain boundaries are in equilibrium with the surrounding atmosphere while the defect concentrations within the bulk remain frozen in. As grain boundary diffusion was only observed for La/Nd, both of which are resistive towards changes in oxidation state, the former explanation is considered unlikely in this case. The latter (ii) would suggest that the activation energy for grain boundary diffusion – in the order of 200–300 kJmol−1 – comprise a combination of both formation and migration enthalpy. Considering the preceding discussion on the formation and migration enthalpy for cation defects in the bulk, this would imply that the migration enthalpy does not contribute significantly to the activation energy if the formation enthalpy is identical to that in the bulk, or vice versa. To fully comprehend the processes governing grain boundary transport and its activation energy the present data set is too limited, and more studies using different diffusants and methodologies are needed. The origin of the exclusive transport of Nd along the grain boundaries is further discussed in light of the defect chemistry of LWO and space charge theory for grain boundaries in oxides. There are several reports on the grain boundary structure and composition of fluorite-type oxide ion conductors such as zirconia and ceria [40–42]. The lattice mismatch between two grains leads to reduced formation energy of vacancies at the grain boundary. In oxygen-deficient oxides, this leads to oxygen vacancies being accumulated at the boundary, resulting in an effectively positively charged grain boundary core. This further leads to a space charge region in the vicinity of the core (the space charge region) where the concentration of positively charged defects is depressed. More detailed studies on oxygen deficient fluorites have revealed indications of accumulation of acceptor dopants within the space charge

eff

[9]. The present results reveal that the bulk diffusion coefficients for Nd and Mo diffusion are similar in magnitude, and that only Nd is mobile along the grain boundaries. In order to evaluate the long term kinetic stability of LWO membranes, the system can be considered as within Harrison's type-A kinetic regime. Accordingly, the effective Nd diffusion coefficient has contributions from both the bulk and grain boundary diffusivities:

-10

log (Dbulk / cm2s-1)

Fig. 6. Comparison of bulk diffusion coefficients in LWO obtained by interdiffusion and chemical tracer diffusion. The lines are meant as guides-to-the-eye to illustrate the change in temperature dependency.

Fig. 7 compares the bulk cation diffusion coefficients of LWO to those of related systems [11,17,38,45–50]. Clearly, the values for LWO are significantly lower than for many state-of-the-art oxygen transport membranes, in particular the perovskite LaMnO3 and the Ruddlesden– Popper phase La2NiO4. Compared to other fluorites such as zirconia and ceria, however, the bulk and grain boundary diffusion coefficients in LWO are notably higher [40,51]. Martin has shown that kinetic demixing or decomposition can occur   DW is smaller than 1/2 if the ratio of the effective cation diffusivities Deff La

1. Mo: LWO - this study 2. Nd: LWO - this study 3. La/W: LWO (interdiff) [31] 4. Mn: LaMnO3 [44]

4

5. Pr: LaMnO3 [46]

5

-12 -14

14

15

6. Pr: La2NiO4 [38]

3

16

6

7. Co: La2NiO4 [38]

7

8. Fe: LaFeO3 [46]

8

1

-16

2

9 10

-18 13

12

11

11. Fe: LSGM [11] 12: Zr: BaTiO3 [45] 13: Sr: BaTiO3 [45]

-20 -22

9. Cr: La0.9Sr0.1FeO3 [49] 10. Cr: La0.95Sr0.05CrO3 [47]

0,5

0,6

0,7 3

-1

0,8

0,9

14. Y: YSZ [17] 15. Sc: YSC [17] 16. Zr: YSZ [17]

-1

10 T (K ) Fig. 7. Reported bulk diffusion coefficients for selected fluorite and fluorite-related systems obtained using chemical tracer diffusion measurements.

E. Vøllestad et al. / Solid State Ionics 274 (2015) 128–133

of a LWO membrane exposed to a pO2 -gradient facilitated by grain boundary diffusion may result in kinetic demixing or decomposition. For an operating temperature of 900 °C with bulk diffusion coefficients of 1 × 10−16 cm2·s−1 and grain boundary diffusivities 104 times higher, the effective La diffusion coefficient can be calculated from Eq. (4) as DLa eff = 4 × 10− 16 cm2 ⋅ s− 1. Here, the grain boundary width is assumed 1 nm and the average grain size 10 μm. The time to reach steady state demixing will most likely be determined by the slowest moving cation — in our case W diffusion with an es−16 cm2 ⋅ s−1. timated effective diffusion coefficient of DW eff = 1 × 10 Given these estimations, the thickness of a LWO membrane can be reduced to ~1 μm before a lifetime of 40,000 h is challenged by cation diffusion related degradation. On these bases we conclude that degradation due to kinetic demixing or decomposition is not likely to limit the durability of LWO-based components for hydrogen technologies. 5. Conclusions

[5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23]

The bulk chemical tracer diffusion coefficients of Nd and Mo in La28 − xW4+ xO54+ 3x / 2 (LWO) were similar in magnitude at all temperatures investigated (1000–1200 °C), with an activation energy of 200 ± 80 kJ·mol−1. The bulk diffusivities did not change significantly from oxidizing to reducing conditions. As Nd and Mo are expected to reside on La and W sites, respectively, it is suggested that bulk diffusion occurs via vacancies on the La2-sublattice where both cations are present. The low activation energies for bulk diffusion are attributed to frozen-in defect concentrations below 1200 °C, and thus only reflect the migration enthalpy. Enhanced diffusivity along the grain boundaries, 103–104 times faster than bulk diffusion, was only observed for Nd migration, with an activation energy of ~250 kJ·mol−1. The lack of Mo diffusion along the grain boundaries is attributed to space charge effects and depletion of the hexavalent Mo6+ at La2-sites near the positively charged grain boundary core. The bulk diffusivities reported for LWO are lower than some stateof-the-art oxygen transport membranes, but higher than other oxygen deficient fluorites such as YSZ. Kinetic demixing or decomposition for a LWO membrane exposed to an electrochemical potential gradient will be determined by the faster grain boundary diffusivity. Acknowledgements This publication has been produced with support from the BIGCCS Centre, performed under the Norwegian research programme Centres for Environment-friendly Energy Research (FME). The authors acknowledge the following partners for their contributions: Aker Solutions, ConocoPhilips, Det Norske Veritas, Gassco, Hydro, Shell, Statkraft, Statoil, TOTAL, GDF SUEZ and the Research Council of Norway (193816/S60). References [1] T. Norby, Nature 410 (6831) (2001) 877–878. [2] T. Norby, R. Haugsrud, Dense Ceramic Membranes, in Membranes for Energy Conversion, Wiley-VCH Verlag GmbH & Co KgaA, 2008. 169. [3] H. Iwahara, Y. Asakura, K. Katahira, M. Tanaka, Solid State Ionics 168 (3–4) (2004) 299–310. [4] H. Schmalzried, W. Laqua, Oxid. Met. 15 (3) (1981) 339–353.

[24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39]

[40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51]

133

M. Martin, H. Schmalzried, Ber. Bunsenges. Phys. Chem. 89 (2) (1985) 124–130. H. Schmalzried, J. Chem. Soc. Faraday Trans. 86 (8) (1990) 1273–1280. M. Martin, Mater. Sci. Rep. 7 (1) (1991) 1–86. M. Martin, in: Pennington (Ed.)The Electrochemical Society Proceedings Series, 1999, pp. 308–316. M. Martin, Solid State Ionics 136–137 (2000) 331–337. M. Martin, J. Chem. Thermodyn. 35 (8) (2003) 1291–1308. O. Schulz, M. Martin, C. Argirusis, G. Borchardt, Phys. Chem. Chem. Phys. 5 (11) (2003) 2308–2313. M. Palcut, R. Knibbe, K. Wiik, T. Grande, Solid State Ionics 202 (1) (2011) 6–13. N. Cebasek, Thesis at Department of Chemistry, University of Oslo, 2012. N. Čebašek, T. Norby, Z. Li, J. Electrochem. Soc. 159 (8) (2012) F461–F467. J.B. Smith, T. Norby, A. Fossdal, J. Am. Ceram. Soc. 89 (2) (2006) 582–586. S.P. Harvey, R.A. De Souza, M. Martin, Energy Environ. Sci. 5 (2) (2012) 5803. M. Kilo, M.A. Taylor, C. Argirusis, G. Borchardt, B. Lesage, S. Weber, S. Scherrer, H. Scherrer, M. Schroeder, M. Martin, J. Appl. Phys. 94 (12) (2003) 7547–7552. T. Shimura, S. Fujimoto, H. Iwahara, Solid State Ionics 143 (1) (2001) 117–123. R. Haugsrud, Solid State Ionics 178 (7–10) (2007) 555–560. R. Haugsrud, C. Kjølseth, J. Phys. Chem. Solids 69 (7) (2008) 1758–1765. A. Magraso, C. Frontera, D. Marrero-Lopez, P. Nunez, Dalton Trans. 46 (2009) 10273–10283. J.J. Roa, A. Magrasó, M. Morales, P. Núñez, M. Segarra, Ceram. Int. 37 (5) (2011) 1593–1599. A. Magraso, J.M. Polfus, C. Frontera, J. Canales-Vazquez, L.-E. Kalland, C.H. Hervoches, S. Erdal, R. Hancke, M.S. Islam, T. Norby, R. Haugsrud, J. Mater. Chem. 22 (5) (2012) 1762–1764. S. Erdal, L.-E. Kalland, R. Hancke, J. Polfus, R. Haugsrud, T. Norby, A. Magrasó, Int. J. Hydrog. Energy 37 (9) (2012) 8051–8055. A. Magrasó, J. Power Sources 240 (2013) 583–588. S. Erdal, Hydrogen in Oxides, in Department of Chemistry, University of Oslo, 2011. S. Escolástico, C. Solís, J.M. Serra, Solid State Ionics 216 (2012) 31–35. S. Escolástico, C. Solís, T. Scherb, G. Schumacher, J.M. Serra, J. Membr. Sci. 444 (2013) 276–284. M. Amsif, A. Magrasó, D. Marrero-López, J.C. Ruiz-Morales, J. Canales-Vázquez, P. Núñez, Chem. Mater. 24 (20) (2012) 3868–3877. S. Escolastico, J. Seeger, S. Roitsch, M. Ivanova, W.A. Meulenberg, J.M. Serra, ChemSusChem 6 (8) (2013) 1523–1532. E. Vøllestad, T. Norby, R. Haugsrud, Solid State Ionics 244 (2013) 57–62. E. Vøllestad, A. Gorzkowska-Sobas, R. Haugsrud, Thin Solid Films 520 (21) (2012) 6531–6534. R.A. De Souza, J. Zehnpfenning, M. Martin, J. Maier, Solid State Ionics 176 (15–16) (2005) 1465–1471. Y.C. Chung, B.J. Wuensch, J. Appl. Phys. 79 (11) (1996) 8323–8329. L.-E. Kalland, A. Magrasó, A. Mancini, C. Tealdi, L. Malavasi, Chem. Mater. 25 (11) (2013) 2378–2384. A. Magraso, C.H. Hervoches, I. Ahmed, S. Hull, J. Nordstrom, A.W.B. Skilbred, R. Haugsrud, J. Mater. Chem. A 1 (11) (2013) 3774–3782. M. Kilo, M.A. Taylor, C. Argirusis, G. Borchardt, R.A. Jackson, O. Schulz, M. Martin, M. Weller, Solid State Ionics 175 (1–4) (2004) 823–827. N. Čebašek, R. Haugsrud, Z. Li, T. Norby, J. Am. Ceram. Soc. 96 (2) (2013) 598–605. D. Waller, J.D. Sirman, J.A. Kilner, in: U. Stimming, et al., (Eds.), Proceedings of the Fifth International Symposium on Solid Oxide Fuel Cells, vol. 97, Electrochemical Society Inc., Pennington, 1997, pp. 1140–1149. K. Kowalski, A. Bernasik, A. Sadowski, J. Eur. Ceram. Soc. 20 (7) (2000) 951–958. Y. Lei, Y. Ito, N.D. Browning, T.J. Mazanec, J. Am. Ceram. Soc. 85 (9) (2002) 2359–2363. R.A. De Souza, Z.A. Munir, S. Kim, M. Martin, Solid State Ionics 196 (1) (2011) 1–8. C. Tian, S.-W. Chan, Solid State Ionics 134 (1–2) (2000) 89–102. C. Kjølseth, H. Fjeld, Ø. Prytz, P.I. Dahl, C. Estournès, R. Haugsrud, T. Norby, Solid State Ionics 181 (5–7) (2010) 268–275. S. Miyoshi, M. Martin, Phys. Chem. Chem. Phys. 11 (17) (2009) 3063–3070. S. Koerfer, R.A. De Souza, H.-I. Yoo, M. Martin, Solid State Sci. 10 (6) (2008) 725–734. M. Palcut, J.S. Christensen, K. Wiik, T. Grande, Phys. Chem. Chem. Phys. 10 (43) (2008) 6544–6552. J.B. Smith, T. Norby, Solid State Ionics 177 (7–8) (2006) 639–646. N. Sakai, K. Yamaji, T. Horita, H. Negishi, H. Yokokawa, Solid State Ionics 135 (1–4) (2000) 469–474. I. Wærnhus, N. Sakai, H. Yokokawa, T. Grande, M.-A. Einarsrud, K. Wiik, Solid State Ionics 178 (13–14) (2007) 907–914. P.-L. Chen, I.W. Chen, J. Am. Ceram. Soc. 79 (7) (1996) 1793–1800.