Rapid Preparation and SEM Microstructural Characterization of Nickel-Yttria-Stabilized Zirconia Cermets

J. Am. Ceram. Soc., 91 [10] 3405–3407 (2008) DOI: 10.1111/j.1551-2916.2008.02538.x r 2008 The American Ceramic Society

Journal Rapid Preparation and SEM Microstructural Characterization of Nickel–Yttria-Stabilized Zirconia Cermets Christian Monachon,w Aı¨ cha Hessler-Wyser, and Antonin Faes Interdisciplinary Centre for Electron Microscopy (CIME), Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland

Antonin Faes and Jan Van Herle Laboratory for Industrial Energy Systems (LENI), Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland

Enrico Tagliaferri HT ceramix SA, CH-1400 Yverdon-les-Bains, Switzerland

precluded the use of one-step SEM quantitative characterization techniques until now. Optical techniques also present some problems because the nickel BSC for light is so much higher than the two other phases that they cannot be distinguished. To avoid this problem, Simwonis et al.2 proposed a technique based on selective Fe2O3 sputtering on the nickel phase, which leads to a decrease of this phase’s BSC of light and thus a contrast enhancement for the two low-reflecting phases. Lee et al.3,4 proposed two other techniques. The first is based on physical separation of the phases—optical microscopy to take advantage of the nickel’s brightness compared with the other phases, SEM to obtain porosity, and then a final SEM analysis after complete etching of the nickel to reveal the zirconia structure. The second technique by Lee et al. is based on an optical micrograph only but requires a strong contrast enhancement via computational methods to reduce uncertainties concerning phases. The techniques presented by both Lee and Simwonis are accurate, but time consuming. Indeed, of these techniques requires impregnation by an epoxy resin before polishing to avoid filling the present porosity (which constitutes about 24 vol%) with the grinding residues. Although the impregnation by epoxy resin allows a wider range of techniques to be used with a good accuracy, it has two main drawbacks. Firstly, it implies significant skill to correctly impregnate a porous material, especially when its pore size approaches the submicrometric range. The second disadvantage of this technique is that embedding a small sample in a big epoxy matrix significantly increases the duration of the polishing. This paper describes a technique consisting of direct polishing without impregnation, followed by high-voltage BSE observation to avoid difficulties posed by the grinding residues.

Sample preparation and scanning electron microscopy imaging methods were developed to characterize solid-oxide fuel cell anodes made out of a nickel–yttria-stabilized zirconia porous cermet. The sample preparation uses only tripod polishing and does not require any resin impregnation. To overcome surface damages induced by polishing, as well as to increase contrast between phases, high acceleration voltages were used. The method gave results of a small systematic increase of the measured nickel proportion due to imaging conditions. A quick image analyzing method is also presented. I. Introduction

S

OLID-oxide fuel cells (SOFC) are seen as a promising future energy conversion technology. Their high working temperature (about 8001C for zirconia-based SOFC) is compensated by high current densities for small volumes and the possibility to recover high-temperature heat for cogeneration applications. This communication presents a quick sample preparation and microstructure characterization technique developed for nickel– yttria-stabilized zirconia (Ni–YSZ) cermet supports, the material commonly used as an anode in SOFCs. The Ni–YSZ cermet microstructure requires specific characterization tools, because (i) its efficiency as an anode relies to a great extent on the density of triple-phase boundaries (the three phases being metallic nickel, YSZ, and reducing gas), which can be estimated by imaging methods and (ii) depending on the fuel flow field and stack design, anode edges may suffer constant oxidation–reduction cycles of the nickel phase, which eventually leads to crack formation and propagation and then destruction of the cell. The technique presented here uses a tripod-polishing method, followed by scanning electron microscopy (SEM) with backscattered electron (BSE) imaging and a simple mathematical tool based on stereology1 to describe the main quantities of the microstructure. The main issue in the characterization of the Ni–YSZ structure by BSE SEM is that the backscattering coefficients (BSC) of Ni and YSZ are very close, which leads to little contrast and

II. Experimental Procedure Samples were prepared by tapecasting a slurry containing 55 wt% nickel oxide, 22.5 wt% (3 mol% yttria-doped zirconia), and 22.5 wt% (8 mol% yttria-doped zirconia). Samples were broken in a random direction, and then the fractured section was mechanically polished by a tripod method. The tripod (a small sample holder with three legs whose lengths are adjusted with micrometric screws) was used in order to produce a planar surface from a fractured sample by polishing on a mylar diamond lapping film, particle sizes 15, 6, 3, and 1 mm. The samples were then carbon coated as the zirconia phase is not conductive. SEM BSE images were taken on a Philips

M. Cinibulk—contributing editor

Manuscript No. 24351. Received February 25, 2008; approved May 8, 2008. w Author to whom correspondence should be addressed. e-mail: christian.monachon@ epfl.ch

3405

3406

Vol. 91, No. 10

Communications of the American Ceramic Society

Table I. Values Obtained for Volume Fractions of Each Phase Method

LIM method Theoretical value

Vol% Ni

Vol% YSZ

Vol% porosity

3273 29.4

4673 46.3

2273 24.3

LIM, lineal interception method; YSZ, yttria-stabilized zirconia.

Fig. 1. Image treatment and phase separation, applied to the nickel phase. Top left: raw microscope image with corresponding gray-level histogram; top right: image as treated by GIMP software with corresponding gray-level histogram; bottom: image after threshold filtering, showing the nickel phase in black.

(FEI Corporation Hillsboro, Oregon 97124) XL-30 SFEG SEM at 22 kV with a high beam intensity (large objective aperture, medium spot size). When insufficient for quantitative analysis, image contrasts were enhanced manually with the GIMP software5 (brightening tool). Information was extracted from the obtained images by the ImageJ software6 (thresholding methods, measurements). Figure 1 shows an example of image treatment. For a homogeneous microstructure, the minimum number of grains required to obtain reasonable statistics is about 400. For this work, 10 pictures containing about 100 grains each were analyzed. Every treated picture was compared by superimposing with its original to ensure that the treatment did not introduce errors. Typical dimensions of the microstructure were derived from calculated volume fraction and grain counting along random curves of known dimensions (lineal interception method, LIM method7), using formula, (1) where L3 is the typical length needed, L the total length of the random line, Vi the volume fraction of the considered phase, and N the number of grains of that phase crossed by the random line. The range of L3 is deduced from that of Vi, N being an integer and the influence of L on the range of L3 being negligible compared with that of Vi. In order to link the typical length L3 to the grain size, a multiplying factor is included accounting for the fact that a polished surface does not necessarily intersect the particles at their maximum diameter. For the equivalent spherical diameter of a grain this factor is 1.5.7,8

L3 ¼

LVi N

(1)

III. Results (1) Imaging Method Using high-acceleration voltage to obtain BSE images on our samples had two objectives: (i) the surface of the polished samples was contaminated by polishing residues (diamond, fine particles from the anode itself), through which high-voltage electrons penetrate and (ii) absolute BSC difference between the phases is increased (while relative differences remains unchanged), giving better contrast. The main drawback of this method is lower resolution because of higher mean free paths, and thus higher penetration of electrons in the material (about 1 mm, as calculated by the CASINO software9). This reduces the precision of the results, because it increases the apparent volume fraction of nickel (and YSZ to a smaller extent) to the detriment of the pore phase.

(2) Typical Length Measurements Table I shows the results obtained for volume fraction measurements, compared with the theoretical values calculated from weight fractions of the composite used for synthesis, including a 2.5% final porosity after sintering of the sample. The YSZ volume fraction was derived from the measurements made for the porosity and nickel fractions. The slight differences between calculated and theoretical values (mainly the nickel phase overestimation) can be explained by the high interaction volume of the electron beam used (high-acceleration voltage). A reasonable accuracy in the volume fraction measurements allowed us to give typical phase length values in our samples. The lengths were deduced from Eq. (1). This kind of measurement is only applicable for isomorphic microstructures, which is the case for our samples as several samples were broken in different macroscopic directions before polishing and gave similar results. The phase length results are given in Table II. This allows a quantitative approach for the cermet phases, which is otherwise difficult to obtain without significant computational treatment. (3) Specific Applications of the Method The presented method revealed two interesting facts: the nickel grain coarsening after fuel cell operation and the effect of different sintering powders on the microstructure of the anode. Nickel grain coarsening is relevant in the field of SOFCs because of its direct effect on the degradation of the anode electrochemical performance. The presented microstructural investigation method was applied to a 600-h stack conditionstested sample. It was compared with a sample of the same tape casting, as reduced for 112 h at 8001C in an N2 atmosphere containing 15 vol% H2. Table III shows the results obtained. The mean distance between nickel grains, l, is given by l¼

Lð1
Vi Þ N

(2)

according to the notation in Eq. (1). The measured values suffer high uncertainty due to the presence of very small grains (which coarsen rapidly due to their high curvature) but an increase of the size of the nickel grain as well as of the widening of the distance between them can be seen. This result is consistent with2 and may lead to further work. The same sample preparation method has been applied to different cermet configurations, leading to results directly related to the redox tolerance of the anode. The mass proportions were the same for all samples. In one case, the nickel oxide powder had a different grain size distribution and in the second case, 50% of 8YSZ zirconia (mean grain size 1 mm, measured by centrifugation) was replaced by 10YSZ zirconia with a mean grain size of 12 mm. Both compositions are of interest because of their increased resistance to redox cycling. Table IV presents the

Table II. Typical Lengths L3 Measured in Our Samples Phase

Ni YSZ Porosity YSZ, yttria-stabilized zirconia.

L3 ½mm

0.5270.11 0.4870.10 0.4170.19

October 2008

Table III. Typical Lengths L3 and l Measured in the 600 h—Aged and the As-Reduced Samples L3 ½mm

l (mm)

0.770.1 0.970.2

1.370.1 1.970.4

Testing time (h)

0 600

3407

Communications of the American Ceramic Society

results obtained. It shows that changing the grain size of a phase has an impact on the other phases and changes the properties of the anode. Especially, increase in dimension of the nickel grains seems to be the source of the improvement in cell redox stability. Again, the presented technique provided a quick and powerful tool for quantitative comparison of different microstructures.

the zirconia phase, a high acceleration voltage (22 kV) is used. The results obtained fromimages gained with this observation technique indicate a slight overestimation of the nickel phase volume fraction to the detriment of the pore fraction. This 2% overestimation is constant between samples and seems to be due to the higher BSC of the nickel phase, coupled to a thickness effect due to higher voltages. However, a comparatively short time is required (1 h for sample preparation, 1 h for imaging, 30 min for image treatment including measurements) for both preparation and analysis because of their simplicity, which is the main interest of this technique.

Acknowledgments The author is grateful to Prof. H. Hofmann (EPFL–Laboratory of Powder Technology) for advise on stereologic methods.

IV. Conclusion Current methods for characterizing nickel–zirconia cermets used for SOFC applications involve several time-consuming steps. A new, faster method is presented, which avoids using epoxy infiltration of the specimen, reduces the polishing time considerably, and allows for a single-step SEM observation of the sample to provide information on all three phases present. This method uses direct polishing of as-fractured samples without epoxy impregnation. The polishing is performed following the same procedure as for a TEM preparation method on a tripod to ensure a good planarity of the surface. However, polishing residues remain on the surface, making the observation more difficult. To overcome this surface contamination as well as to be able to distinguish the metallic nickel phase from

Table IV. Measurement of the Impact of Different Powders on the Typical Lengths of NiO in the Cermet’s Microstructure Composition

Standard composition NiO with different grain size With 12 mm zirconia added

L3 ½mm

l (mm)

0.7270.07 1.0570.10 0.9770.16

0.6970.07 0.9770.15 1.2470.2

References 1

J. Ohser and F. Mu¨cklich, Statistical Analysis of Microstructures in Materials Science. John Wiley & Sons Ltd, Chichester, 2000. 2 D. Simwonis, F. Tietz, and D. Sto¨ver, ‘‘Nickel Coarsening in Annealed Ni/8YSZ Anode Substrates for Solid Oxide Fuel Cells,’’ Solid State Ionics, 132, 241–51 (2000). 3 J.-H. Lee, H. Moon, H.-W. Lee, J. Kim, J.-D. Kim, and K.-H. Yoon, ‘‘Quantitative Analysis of Microstructure and its Related Electrical Property of SOFC Anode, Ni-YSZ Cermet,’’ Solid State Ionics, 148, 15–26 (2002). 4 K.-R. Lee, S. H. Choi, J. Kim, H.-W. Lee, and J.-H. Lee, ‘‘Viable Image Analysing Method to Characterize the Microstructure of the Ni/YSZ Cermet Anode of SOFC,’’ J. Power Sources, 140, 226–34 (2005). 5 M. I. Mendelson, ‘‘Average Grain Size in Polycrystalline Ceramics,’’ J. Am. Ceram. Soc., 52 [8] 443–6 (1969). 6 J. C. Wurst and J. A. Nelson, ‘‘Lineal Intercept Technique for Measuring Grain Size in Two-Phase Polycrystalline Ceramics,’’ J. Am. Ceram. Soc., 55 [2] 109–109 (1972). 7 D. Waldbillig, A. Wood, and D. G. Ivey, ‘‘Electrochemical and Microstructural Characterization of the Redox Tolerance of Solid Oxide Fuel Cell Anodes,’’ J. Power Sources, 145, 206–15 (2005). 8 Y. L. Liu, S. Primdahl, and M. Mogensen, ‘‘Effects of Impurities on Microstructure in Ni/YSZ-YSZ Half-Cells for SOFC,’’ Solid State Ionics, 161, 1–10 (2003). 9 Gnu Image Manipulation Program, http://www.gimp.org/ 10 ImageJ, http://rsb.info.nih.gov/ij/ 11 D. Drouin, A. R. Couture, D. Joly, X. Tastet, V. Aimez, and R. Gauvin, ‘‘CASINO V2.42—A Fast and Easy-to-Use Modeling Tool for Scanning Electron Microscopy and Microanalysis Users,’’ Scanning, 29, 92–101 (2007). &