Scholarly article on topic 'Changes in Physical and Mechanical Properties of SOFC Ni–YSZ Composites Caused by Redox Cycling'

Changes in Physical and Mechanical Properties of SOFC Ni–YSZ Composites Caused by Redox Cycling Academic research paper on "Earth and related environmental sciences"

0
0
Share paper
Keywords
{""}

Academic research paper on topic "Changes in Physical and Mechanical Properties of SOFC Ni–YSZ Composites Caused by Redox Cycling"

Journal of The Electrochemical Society, 155 (5) B467-B472 (2008) B467

0013-4651/2008/155(5)/B467/6/$23.00 © The Electrochemical Society

Changes in Physical and Mechanical Properties of SOFC Ni-YSZ Composites Caused by Redox Cycling

D. Sarantaridis,z R. J. Chater, and A. Atkinson*

Department of Materials, Imperial College London, London SW7 2AZ, United Kingdom

The investigations of sequential cyclic reduction and oxidation (redox) of solid oxide fuel cell (SOFC) Ni-yttria-stabilized zirconia (YSZ) anode substrates presented here are aimed at monitoring and understanding changes in critical physical and mechanical properties as a function of degree of oxidation at 800°C. The commonly observed expansion on oxidation, and its irreversibility on subsequent reduction, was confirmed but was found to be greatly dependent on oxidation conditions, giving strains in the range 0.55-0.8%. Nevertheless, redox cycling was accompanied by reversible changes in elastic modulus and no significant damage was detected in the YSZ network. Microstructure observations of oxidized Ni particles revealed the formation of closed porosity of a similar magnitude to that observed in oxidized composites (ca. 5%), suggesting that this is a major cause of the expansion on oxidation of the composite.

© 2008 The Electrochemical Society. [DOI: 10.1149/1.2883731] All rights reserved.

Manuscript submitted October 30, 2007; revised manuscript received January 7, 2008. Available electronically March 11, 2008.

Anode-supported solid oxide fuel cells (SOFCs) using Ni-based substrates (cermet composites) are susceptible to damage during fault conditions (e.g., interruption of the fuel supply or too high a fuel utilization at high overvoltage) as a result of dimensional changes of the cermet caused by the oxidation of Ni to NiO and cycles of reduction and oxidation (redox). Although there is considerable variability in results reported by different researchers, some trends are consistently observed in redox experiments with Ni-yttria-stabilized zirconia (YSZ) cermets. Upon initial reduction of the NiO-YSZ precursor there is negligible dimensional change, whereas the first complete oxidation (all Ni converted to NiO) is accompanied by a linear expansion strain of approximately 1%. Successive redox cycles result in an accumulating expansion strain, revealing a nonreversible redox process for the cermet. This oxidation strain is recognized to be the critical parameter for driving damage (electrolyte cracking) in anode-supported cells (ASCs), which can take place for an oxidation strain as low as 0.1%/

Nevertheless, the details of how the cermet expands with advancing oxidation and what controls this mechanism are not established. The relationship between degree of oxidation and composite expansion is of great interest for estimating how much oxidation the cell can accommodate before failure. Regarding the actual expansion mechanism, there is published work that suggests expansion is due to the agglomeration of metallic nickel2,3 and the formation of porosity upon oxidation,4-6 based mainly on qualitative scanning electron microscopy (SEM) observations. We present in this paper controlled oxidation studies of Ni particles and Ni cermet substrates in which changes in dimensions, porosity, and elastic modulus were measured as a function of degree of oxidation.

While redox damage of the ASCs by electrolyte cracking is the expected failure mode, there is some controversy concerning the effect of redox on the integrity of the anode substrate itself. It has been reported in the literature that redox can cause microcracking or even catastrophic failure of the composite (see Ref. 1 and references therein). In the present work we assess damage to the composite from elastic modulus changes in combination with microstructural observations of the YSZ network.

Experimental

Samples.— The Ni-based composite specimens were provided by Forschungszentrum Juelich in the form of NiO/YSZ composite plates with composition 56 wt % NiO/44 wt % 8YSZ. The plates were pretreated under the same conditions as would be used for cosintering an electrolyte layer so that they were in the same state as they would be in a typical ASC. (The presence of the electrolyte and

* Electrochemical Society Active Member. z E-mail: dimitrios.sarantaridis@imperial.ac.uk

anode in the sintering of a production half-cell is not likely to affect the sintering of the substrate, because the substrate has much greater thickness than the functional layers. After fabrication of a production half-cell the electrolyte would put the substrate under tension upon cooling down after sintering. Again, this is a small effect on the substrate, because of its greater thickness, and at operating temperature the substrate's residual stress is estimated to be only ~ 2 MPa.7) Rectangular bars of ca. 40 X 4 X 1.5 mm for the interrupted oxidation experiments or 25 X 4 X 1.5 mm for the dilatom-etry measurements were cut from the plates. Spherical Ni particles (Alpha Aesar, 5-15 micrometers, 99% Ni) were used for the studies of individual particles. Samples of the porous YSZ network were prepared from the Ni-YSZ composites by completely dissolving the nickel using concentrated (16 molar) HNO3.

Mechanical/physical properties characterization.— The characterization of the redox behavior of the composites involved monitoring the linear strain, elastic modulus E, porosity, and microstuc-ture. Vernier callipers (10 ^m resolution) and dilatometry (Netzsch 402E dilatometer, 0.01 ^m resolution) were used for the strain measurements. The elastic modulus was measured by the impulse excitation of vibration technique8 using the GrindoSonic MK5 instrument. Out-of-plane excitation was performed on samples supported by foam pads, while the signal was recorded with a microphone. The porosity measurements were performed with the aid of a Sartorius (CP 124 S) balance, based on the Archimedes principle and vacuum impregnation of water.9 Thus, values for the bulk density (which corresponds to the mass of unit volume of the porous composite), apparent solid density (i.e., the density of the solid phase including closed pores), and volume fraction of open porosity could be obtained. In the complete reduced (Ni-YSZ) and oxidized states (NiO-YSZ), values for the closed porosity were also deduced using theoretical true solid densities for the known composition of the cermet (reduced 7.10 g/cm3, oxidized 6.31 g/cm3).

SEM (instrument: Jeol JSM 5610 LV) and focused ion beam (FIB, instrument: FEI FIB200) techniques were used for observing the YSZ network and sectioning/imaging oxidized Ni (NiO) particles. The YSZ samples were coated with gold before being examined by SEM.

Redox procedures.— Reduction of the composites was carried out in a tube furnace with 10% H2/N2 flowing at 50 mL/min and the following temperature program: heated up to 650°C at 3 K/min, held for 6 h, heated up to 900°C at 3 K/min, held for 10 h, and cooled down to room temperature (RT) at 3 K/min. A reduction experiment was also performed in situ in the dilatometer by heating up in N2 to 900°C (10 K/min) and then introducing 5% H2/N2 (50 mL/min) for 3 h at 900°C.

Interrupted oxidation of the cermets was performed at 800°C (in a box furnace with ambient air) in steps. A step consisted of insert-

1 1 1 1 1 1 1 1 1 i 1 i 1 1 1 1 1 i 1 i

• oxidation strain o

o re-oxidation strain

■O o

■ ° * *

i.i.i.i.i I.I. i . i . 1 . 1

0 10 20 30 40 50 60 70 80 90 100 degree of oxidation (%)

Figure 1. Expansion strain of anode substrates as a function of degree of oxidation. Error bars indicate the uncertainty introduced by the dimensional measurements.

ing the sample directly from RT into the furnace at 800°C, allowing it to oxidize for a few seconds or minutes, and then taking it out and letting it cool freely in air to RT. The degree of oxidation was determined after each step by the change in weight, and dimensional changes were measured using Vernier callipers. Oxidation in the dilatometer was performed in flowing 50 mL/min air at 800°C after the sample was heated up in N2 at 10 K/min. The nickel particles were oxidized in the box furnace at 800°C for 2 h (full oxidation) or a few seconds or minutes (partial oxidation).

Results and Discussion

Redox of anode substrates.— On initial reduction, the bar specimens showed no detectable length changes when measured with the callipers and only 0.017% contraction in the dilatometer, in agreement with most other published work on similar composites.1 Figure 1 shows the relation between degree of oxidation and expansion strain of the anode substrates (interrupted oxidation). After the first complete oxidation the substrate had expanded to a total linear strain of 0.55%, showing an almost linear relation between degree of oxidation and strain in the range of 20-90% oxidation. However, in the first oxidation steps (up to 20%) a slight contraction was recorded. After the first oxidation and reduction cycle, reoxidation of the composite showed a similar trend to the first oxidation, except that there was a constant extra strain (ca. 0.1%), resulting from the irrevers-ibility of the first cycle. As a result, the total strain after the second

Table I. Linear expansion strain of composites after total oxidation for different methods of oxidation.

Interrupted Direct Dilatometry

oxidation oxidation oxidation

(ambient air) (ambient air) (air flow)

Oxidation strain (%) 0.55 ± 0.03 0.65 ± 0.03 0.80 ± 0.01

complete oxidation was higher (0.63%) than after the first. (The slight vertical offset to the left of the reoxidation points in Fig. 1, 3, and 5 was probably caused by some loss of material during handling of the sample between the two redox cycles.) This accumulation of oxidation strain with redox cycling is in accordance with published studies. Klemens0 et al.2 have suggested that it is the result of the sintering and rearrangement of the metal Ni particles in the YSZ network. The sample again showed a small initial contraction during the second oxidation. This kind of contraction was not observed, however, in the dilatometer oxidation (Fig. 2).

Moreover, the different oxidation conditions produced different values of final expansion strain (Table I). The interrupted and direct oxidation both took place in ambient air (no flow), implying that interruption of the oxidation process affects the expansion behavior of the cermet. It seems that interruption of the early steps of oxidation causes the initial shrinkage and a final lower expansion on full oxidation (0.55%) when compared to direct oxidation (0.65%). The oxidation in the dilatometer showed a distinctly higher expansion strain on full oxidation (0.8%). The key feature in this experiment is that there was a flow of air for the oxidation in the dilatometer. Ettler et al.10 have drawn attention to the effects that incident oxidizing flows have on the kinetics and dimensional behavior of Ni/YSZ cermets. Also, the fact that only in the dilatometry case did the strain measurement take place at high temperature suggests that RT measurements may not be as accurate and may also be compounded by the effects of additional internal stresses caused by the different thermal expansion coefficients of the constituent materials in the composites. Nevertheless, in all cases an irreversible strain was observed after a complete oxidation/reduction cycle.

Young's modulus exhibited a single linear dependence on degree of oxidation (Fig. 3) for both first oxidation and reoxidation. As nickel oxidizes and the solid volume increases, the porosity decreases and the elastic modulus of the composite increases. Its reversibility and repeatability implies that no significant damage occurs within the composite structure. No macroscopic cracking of any samples was observed, but some slight warping of the samples was evident after the intermediate steps of interrupted oxidation. After full oxidation, however, the samples appeared to be flat. Dissolution of the Ni from a piece (ca. 7 X 4 X 1.5 mm) taken from the composite that had gone through two oxidation/reduction cycles re-

Figure 3. Elastic modulus E of anode substrates as a function of degree of oxidation.

Figure 5. Bulk density, apparent solid density, and open porosity of anode substrates as a function of degree of oxidation.

vealed a coherent YSZ network structure (Fig. 4) that retained a rectangular bulk shape. Some minor localized cracking might have occurred, but if it did it was not clearly detectable by SEM and had no major effect on the mechanical integrity of the YSZ network. Although this is consistent with the observed elastic modulus behavior, it is not what one would expect when taking into account the cermet's oxidation strain. It is hard to explain how a brittle ceramic network can accommodate an elongation of 0.6% without failing. It is possible that some degree of plastic deformation (creep) took place in the ceramic structure upon the increase in solid volume when the nickel was oxidized.

Monitoring of the evolution of porosity reveals that with increasing oxidation open porosity decreases, and the overall bulk density increases in an approximately linear fashion (Fig. 5). However, the apparent solid density was observed to decrease to 5.73 g/cm3 on complete oxidation, significantly lower than the 6.11 g/cm3 value measured for the as-received NiO/YSZ composite. Hence, closed porosity forms on oxidation, and in our case an increase of ca. 5% in closed pore volume was estimated to have occurred (Table II). It is apparent from the values in Table II that a redistribution of open and closed porosity takes place after the first reduction/oxidation cycle, but the relative porosity values remain practically unchanged after the second cycle.

There are two ways in which closed porosity can be formed. One is due to the redistribution and partial sintering of nickel grains in

Figure 4. Fracture surface of the YSZ network after two redox cycles and dissolution of nickel.

the reducing environment. The second is through the intrinsic properties of the nickel oxidation and reduction mechanisms at high temperatures. In order to clarify this, we conducted studies of the microstructural changes occurring during oxidation of nickel particles.

Oxidation ofNi particles.— It has been established that Ni oxidation at 500°C < T < 1000°C is controlled by outward diffusion of Ni++ through the oxide scale,11 a process that should result in the formation of voids at the metal/oxide interface or in the bulk of the metal if the metal/oxide interface is prevented from moving inward. Although oxidation of nickel has been widely studied, it is not clear how porosity develops during oxidation, especially for the case of the micrometer-size Ni particles/grains that are present in the cermets of interest. In experiments oxidizing cylindrical nickel wires, Hales and Hill12 showed that internal porosity developed in the nickel due to the inability of the circular metal/oxide interface to migrate inward during oxidation. For a spherical geometry the problem of the oxide/metal interface recession and how contact can be maintained at that interface is a similar issue. If the interface cannot recess inward then an outward Ni++ diffusion mechanism would be expected to create hollow-shell oxide particles with internal void formation equal to the volume of initially present Ni. It has been found that when interface recession is prevented in bulk specimens (e.g., by impurities that pin the interface), inward oxygen permeation through fissures in the oxide occurs and two-layered oxide scales form, with the boundary between the layers coinciding with the original metal surface.13 Recent studies show that high-purity nickel sheets develop a simple oxide layer (outward diffusion of Ni++)14 and that the orientation of the Ni grains may also affect the oxide morphology.15.

Figure 6 shows SEM pictures of the almost-spherical Ni particles and their appearance after full oxidation to NiO particles. The surface of the oxidized particles has changed significantly to show a more textured outer surface. Generally, this surface appearance is taken to signify porosity extending below the surface, what is commonly described as "spongelike" texture.4'5 When the same samples were imaged using the secondary electrons generated by the scanned primary ion beam in the FIB microscope, a similar texture was revealed, but less pronounced (Fig. 7a). However, when sputtered ions were used to form the image in the FIB microscope, Fig. 7b, the surface texture of the fully oxidized NiO particles appeared much smoother. This surface morphology is more aptly described as dimpled rather than porous. Sputtered ions originate from the top surface layers only, that is, they are of subnanometer origin and have an energy range that is at least an order of magnitude greater than

Table II. Properties of the composites after different successive stages of reduction and oxidation. (E, elastic modulus; pb, bulk density; p. apparent solid density; Pa, open porosity; and Ps closed porosity.)

Strain E pb pas Pa Ps

(±0.03,%) (±1.3, GPa) (±0.05, g/m3) (±0.05, g/cm3) (±0.3,%) (±0.3,%)

As received — 79 4.65 6.11 23.9 2.5

Reduced 0.00 32 4.02 6.90 41.8 1.6

Oxidized 0.55 74 4.57 5.73 20.3 7.3

Rereduced 0.14 31 4.04 6.87 41.2 1.9

Reoxidized 0.63 72 4.55 5.72 20.4 7.4

secondary electrons generated in the SEM. Both of these fundamental aspects mean that the ion image shows the surface topography with less enhancement of edges and thin features.

To investigate the subsurface porosity of the oxidized particles, we have exploited the ability to control the FIB current over 4 orders of magnitude. Localized scanning of the FIB at high currents in the nanoamp range sputters the Ni and NiO at a sufficient rate to form a cross section in situ through a particle. Subsequently, the section is imaged with the same ion beam on tilting the particle through 45°, but at the much lower ion current of approximately 50 pA. Images of such cross sections are shown in Fig. 8. There are structure changes both in terms of grain morphology and porosity as a result of oxidation. From a completely dense Ni particle, a NiO particle containing some porosity is formed on complete oxidation. However, a hollow core is not formed.

In order to monitor the porosity changes as oxidation advances, Ni particles were exposed to oxidizing conditions for time periods ranging from 30 s to 5 min, and then sectioned and observed in the FIB microscope (Fig. 9). With the exception of Fig. 9b, it can be observed that as oxidation proceeds, grain growth and formation of porosity take place. The particle of Fig. 9b is significantly smaller

20ku X5,eee 5» is iv sei

than the rest of the batch, and this probably resulted in a higher degree of oxidation than intended. After 5 min of oxidation, and using secondary ion imaging, the oxide scale around the particle is clearly visible (Fig. 9f). Pores are present in both the scale and the metal. A comparison of the fully (Fig. 8b) and partially (Fig. 9) oxidized particles shows that partial oxidation can be characterized by higher porosity formation than full oxidation, implying a complex oxidation mechanism. While outward Ni diffusion may dominate at the early oxidation stages, it is apparent that inward oxygen transport is also involved, given that the oxide/metal interface cannot recess in our spherical geometry.

The above observations show that the oxidation process of small particles does lead to some small amount of closed porosity, but not

Figure 6. SEM of (a) as received Ni and (b) fully oxidized Ni (NiO) particles. The secondary electron images were recorded using a beam energy of 20 keV.

Figure 7. FIB microscope images of fully oxidized Ni particles using (a) secondary electrons and (b) positively charged secondary ions.

Figure 8. FIB cross-sectional secondary electron images of (a) as received Ni and (b) fully oxidized Ni (NiO) particles.

nearly as large an amount as observed on oxidation of wires. Thus, the "swelling" on oxidation is much smaller than would be expected if hollow oxide particles had been formed. Werber16 has suggested that the presence of porosity in the initial Ni particles has an adverse effect on the oxidation swelling. However, this cannot account for the present observations, as the Ni particles were originally fully dense. Nevertheless, even the small amount of porosity observed in the FIB images can easily account for the ca. 5% increase in closed porosity measured for the oxidation of the cermet. Although this closed porosity formed by the oxidation mechanism can account for the swelling of the cermet, additional contributions from sintering of metallic Ni cannot be excluded, and it is likely that a combination of the two, nickel sintering and intrinsic oxidation properties, is involved.

Conclusions

Investigations on Ni-YSZ composites used in state-of-the-art SOFCs were performed involving monitoring changes in critical physical (dimensions and porosity) and mechanical (Young's modulus) properties upon redox cycling at 800°C. A correlation between degree of oxidation and changes in these properties was established using interrupted oxidation.

Initial reduction of the NiO-YSZ composite resulted in negligible shrinkage, whereas subsequent interrupted oxidation exhibited a total linear expansion of ca. 0.55%, following a close-to-linear relation with degree of oxidation. However, the oxidation expansion was shown to depend on the oxidation procedure (interruption of oxidation, measurement temperature, or air flow) giving values as high as 0.8%. It was established that the oxidation strain is not reversible and a residual strain of ca. 0.1% remains after the first oxidation/reduction cycle and adds on to the second. However, the oxidation expansion did not have a detectable effect on the integrity of the composite. The changes in elastic modulus of the composite

Figure 9. FIB cross-sectional secondary electron images of Ni particles oxidized at 800°C for (a) 30, (b) 60, (c) 90, (d) 180, and (e) 300 s. Image (f) is the same particle as in (e) but obtained using the secondary ion signal.

were reversible, and a coherent YSZ network structure was revealed after two redox cycles. Finally, measurement of porosity changes showed that closed porosity forms on oxidation, and this was estimated to be of the order of 5% of the volume of the composite. Sectioning of partially oxidized spherical Ni particles using FIB techniques revealed formation of closed pores, which can account for the increase in closed porosity and the expansion of the composite on oxidation.

It is clear that expansion of the Ni-YSZ composite on oxidation, which is the critical parameter for cell failure, is closely related to microstructural changes of the Ni particles during oxidation. It is shown in this paper that formation of closed porosity in the NiO particles due to intrinsic features of the oxidation mechanism makes a key contribution to the oxidation expansion of Ni cermet anodes and anode supports.

Acknowledgments

This work was carried out as part of the European Commission's Real-SOFC Project, contract no. SES6-CT-2003-502612. The authors thank R. Sweeney for assistance with the dilatometry measurements and colleagues at Forschungszentrum Juelich for provision of samples.

References

1. D. Sarantaridis and A. Atkinson, Fuel Cells, 7, 246 (2007).

2. T. Klemens0, C. Chung, P. H. Larsen, and M. Mogensen, J. Electrochem. Soc.,

152, A2186 (2005).

3. T. Klemens0, C. C. Appel, and M. Mogensen, Electrochem. Solid-State Lett., 9, A403 (2006).

4. J. Malzbender, E. Wessel, and R. W. Steinbrech, Solid State Ionics, 176, 2201 (2005).

5. D. Waldbillig, A. Wood, and D. G. Ivey, Solid State Ionics, 176, 847 (2005).

6. D. Waldbillig, A. Wood, and D. G. Ivey, J. Power Sources, 145, 206 (2005). 11. A. Atkinson, Rev. Mod. Phys., 57, 437 (1985).

7. W. Fischer, J. Malzbender, G. Blass, and R. W. Steinbrech, J. Power Sources, 150, 12. R. Hales and A. C. Hill, Corros. Sci., 12, 843 (1972).

73 (2005). 13. A. Atkinson and D. W. Smart, J. Electrochem. Soc., 135, 2886 (1988).

8. International Standard ISO 12680-1:2005. 14. A. M. Huntz, M. Andrieux, and R. Molins, Mater. Sci. Eng., A, 415, 21 (2006).

9. European Standard EN 1389:2003. 15. R. Peraldi, Mater. High. Temp., 20, 649 (2003).

10. M. Ettler, G. Blaß, and N. H. Menzler, Fuel Cells, 7, 349 (2007). 16. T. Werber, Solid State Ionics, 42, 205 (1990).