Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Solid State Ionics 189 (2011) 82–90

Contents lists available at ScienceDirect

Solid State Ionics


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s s i

Electrical conductivity of Ni–YSZ composites: Degradation due to Ni particle growth


M.H. Pihlatie a,b,⁎, A. Kaiser a, M. Mogensen a, M. Chen a
a
Risoe National Laboratory for Sustainable Energy, Technical University of Denmark, Fuel Cells and Solid State Chemistry Division, Roskilde, Denmark
b
VTT Technical Research Centre of Finland, POB 1000, FI-02044, Finland

a r t i c l e i n f o a b s t r a c t

Article history: The short-term changes in the electrical conductivity of Ni–YSZ composites (cermets) suitable for use in Solid
Received 3 November 2010 Oxide Fuel Cells (SOFC) were measured by an in-situ 4-point DC technique. The isothermal reduction was
Received in revised form 31 January 2011 carried out in dry, humidified or wet hydrogen at temperatures from 600 to 1000 °C. While the cermets
Accepted 4 February 2011
reduced at 600 °C showed a stable conductivity of 1000–1200 S/cm, rapid initial conductivity loss was
Available online 22 March 2011
observed at elevated temperatures. At 1000 °C the conductivity degraded nearly instantaneously to about
Keywords:
800 S/cm, and continued to decline fast to about 400 S/cm. At 850 °C, the presence of steam did have an
SOFC accelerating effect on the conductivity loss. Scanning Electron Microscopy of cermets reduced in different
Ni–YSZ conditions showed increasing particle size and loss of metal-to-metal percolation in the samples reduced at
Ni particle growth higher temperatures. The short-term changes in conductivity were modelled using two different semi-
Ostwald ripening empirical approaches. Thermodynamic calculations were carried out to assess the vaporisation of Ni in the
Electrical conductivity conditions tested. The rate and mechanisms of conductivity degradation due to Ni particle growth are
Degradation discussed in light of the measurements, modelling and literature data.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction particle nanocatalysis as discussed by Sehested et al. [3,4]. Holzer et al.


have recently published related work on Ni grain growth in Ni–CGO
The anode supported Solid Oxide Fuel Cells (SOFC) widely use Ni– composites [5]. Particle growth is more pronounced the smaller the
YSZ ceramic–metal composites (cermets) as anode support and anode initial grains are (or actually, the larger the initial surface area) and
materials. A sufficient electrical conductivity, in the order of 100 S/cm, the higher the temperature. Sehested et al. showed, both experimen-
has to be sustained throughout the operation life of the cells to ensure tally and through density functional theory (DFT) calculations, that
efficient anode current collection. The electrical conductivity of a Ni– humidity boosts Ni growth through increased surface diffusivity of the
YSZ cermet is known to degrade with time under operation at high Ni2–OH complex formed under humid conditions. The formation of
temperature due to the growth of Ni grains and loss of their the Ni2–OH complex is favoured by high steam to hydrogen ratio [3,4].
connectivity. The atomic mobility of the micron-sized Ni embedded Particle growth in the Ni network may result in performance
in the porous YSZ backbone is dependent on operation conditions and degradation through loss of electrical conductivity of the anode
is enhanced with increasing temperature. The mobility in Ni grains support (current collection). Also the electrochemical performance of
during operation may take place though the bulk material, grain the anode can be affected when the active three-phase boundary is
boundaries, through surface migration, or in the gas phase through reduced, resulting in an increase in the anode polarisation resistance.
evaporation and condensation. The combined effect of these could be Particle growth of Ni has been shown to contribute to the mechanisms
called sintering, which tends to lower the total surface energy of the behind redox instability, as first expressed by Klemensø [6] and
system. The morphological changes in Ni affect, however, only one further elaborated by Pihlatie [7], who showed that the redox
phase in the composite and do not lead to densification of the cermet expansion increases with increasing degree of Ni growth.
as a whole. Therefore, the term Ni particle growth will be used — The ceramic–metal Ni–YSZ interface is formed upon reduction of
irrespective of the Ni mass transport mechanism. In this process the the as-sintered composite. The energetics of the interface can be
largest Ni grains grow at the expense of the smallest grains. Fine nickel characterised by work of adhesion, equal to the work needed to
grains are known to grow in size in both microcomposites of Ni–YSZ separate two bonded surfaces. This is [8,9]
as shown in the works of Simwonis et al. and Jiang [1,2], and fine
Wadh = σm + σZrO2 −γm;ZrO2 : ð1Þ

⁎ Corresponding author at: VTT Technical Research Centre of Finland, POB 1000, FI-
02044, Finland. Tel.: + 358 400430395; fax: + 358 20 7227048. where σm and σZrO2 stand for cleaved metal and ceramic surface Gibbs
E-mail addresses: mikko.pihlatie@vtt.fi, mikko.pihlatie@iki.fi (M.H. Pihlatie). free energies and γm,ZrO2 is the Ni–zirconia interface energy, all per

0167-2738/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.ssi.2011.02.001
M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90 83

surface area when the equilibration of the ceramic and metal surfaces wetting. The surface free energies σ and interface energy γ are related
takes place. to the contact angle through Young equation
Nickel in Ni–YSZ cermets applied in SOFC's typically percolates
 
when the volume fraction of Ni exceeds about 35–40%; electrical S LS
cosθ = σZrO2 −γm;ZrO2 = σm ; ð5Þ
conductivity arises from the connectivity of the randomly packed
solid phases, each with its size distribution — this has been modelled
by Sunde [10] and related work. Ni particle growth in porous cermets where L refers to liquid and S solid phase. Furthermore, the work of
change the ability of the Ni network to percolate and a degradation of adhesion Wadh and the contact angle have been correlated through the
conductivity for composites originally above the percolation thresh- Young–Dupre equation
old can be observed; the growth can take place through different
L
mechanisms: evaporation–condensation, surface diffusion, grain Wadh = σm ð1 + cosθÞ; ð6Þ
boundary or volume (bulk) diffusion. Sintering of an aggregate of
particles has generally been described to depend on time t and so that the contact angle is determined by the adhesion energy
particle radius r according to [11,12] between the solid and the liquid phases, and the surface energy of the
liquid phase. Most metals in contact with refractory ceramics show
x 1 = n −p = n
= ð A⋅t Þ ⋅r ; ð2Þ relatively large contact angles in the non-wetting regime, for example
r contact angles of liquid Ni with different oxides. Tsoga et al. [15] and
where x is the contact radius at the neck between two spherical Mantzouris et al. [16] report decreasing wetting angles when the
particles during sintering, A is a constant and the exponents n and stabilised zirconia substrate was doped with TiO2. Furthermore, the
p vary depending on which sintering mechanism is dominating. actual interface energy will also depend on layers of components or
Evaporation–condensation (coarsening) corresponds to n = 3 and impurities segregated to the surfaces and the interface [17]. The
p = 2, surface diffusion (coarsening) to n = 5–7 and p = 4; for interfacial properties can also be evaluated using computational
sintering (densification) rate-limited by volume diffusion the methods such as DFT. For example, DFT calculations on the Ni–YSZ
values are n = 5 and p = 3, and by grain boundary diffusion n = 6 interface have shown that the ideal Ni(001)–c-ZrO2(001) interface is
and p = 4 [11]. The geometric dimensions of the particles affect energetically more stable than bonds on other interfaces [18]. For
their sintering rate as was shown in the original paper of Herring nanocomposites of Ni–YSZ, improved mechanical stability and
[13]; in a geometrically identical cluster of particles, an increasing hardness have been reported due to good interfacial adhesion. This
particle size will decrease the rate of sintering as shown by adhesion is weakened when the particle size increases, in fact in
Herring's scaling law microcomposites the enforcing effect from the metal cannot be
observed. It has been proposed that the good degree of epitaxy
 p
t ðr2 Þ r2 between Ni and tetragonal zirconia polycrystals (TZP) in nanocom-
= ð3Þ posites is due to very good lattice matching between ZrO2 and nickel
t ðr1 Þ r1
(in the preferred crystal orientation), and the evaporation–conden-
In Eq. (3), the time t to reach a particular degree of sintering sation grain growth mechanism operating in metallic nanoparticles
depends on both the particle size (r) and the sintering mechanism (p). during sintering [19]. According to this hypothesis, a change in growth
A ceramic–metal composite produced by wet ball milling, solid state mechanism takes place when going to larger grain sizes, and also, the
sintering and subsequent high temperature reduction contains existence of less favourable interface orientations decreases the work
particles with a size distribution typically ranging from a few tens of of separation. For Ni(001)–c-ZrO2(001) interface, work of separation
nanometres up to a few micrometres. The particle growth is most at 5000–5700 mJ/m2 has been calculated, whereas the Ni(011)–c-
pronounced in the finest fraction of the microstructure, which tends ZrO2(011) interface shows only 900 mJ/m2 [18]. For comparison, the
to disappear. The larger particles grow at the cost of the smaller ones; work of separation in a strong ceramic–metal interface of Al2O3/Nb
this thermodynamically-driven spontaneous process is generally has been calculated to be 9800 mJ/m2 [19].
known as Ostwald ripening. The actual mass transport in the system Electrical conductivity is a good in-situ measure to monitor and
may take place either through surface diffusion or via the gas phase. A quantify the microstructural changes in the cermet: Ni particle
possible growth mechanism via the gas phase may originate from the growth and the loss of percolation. The conductivity degradation (or
fact that the equilibrium vapour pressure over a spherically curved particle growth) is initially rapid and slows down to finally reach a
surface with a radius R, pR, is greater than that over a flat surface, p∞, plateau characteristic of the cermet as well as the operating
by the ratio [13] conditions. This type of behaviour is shown in the works of Klemensø
[20], Vassen et al. [21] and Holzer et al. [5], and follows also when
 
pR 2Ωo γ applying Eq. (2). The extent to which particle growth degrades the
= exp ; ð4Þ
p∞ kTR initial conductivity depends on the temperature and humidity of
operation, but also on the microstructure and composition of the
where Ωo is the atomic or molecular volume, γ the surface tension cermet. Electrical conductivity degradation data in the literature on
and k the Bolzmann constant. Naturally, this requires a sufficient realistic SOFC cermets is scarce. The present paper deals with the early-
partial pressure gradient to be sustained that allows the process to stage conductivity degradation — up to about 200 h of operation — of
proceed at a measurable rate. When a small particle is in the Ni–YSZ composites due to particle growth of Ni. The changes in the
vicinity of a larger particle, a diffusive flow of matter can develop electrical properties of the cermet were determined depending on
towards the larger particle because of the concentration gradient time, temperature and the humidity using an in-situ DC conductivity
caused by the higher equilibrium vapour pressure above the smaller measurement. The measured conductivities were fitted by two
particle. different approaches to describe the degradation behaviour. Thermo-
The stability of the metal–ceramic interface affects the long-term dynamic calculations using FactSage® were carried out to examine the
stability of the composite. Wetting of liquid metal droplets on ceramic evaporation and gaseous partial pressures of Ni species under different
substrates has been investigated by measuring the contact angle conditions. Further, the changes in the composite microstructure were
between the substrate and the droplet [14,15]. Depending on the examined by exposing a typical composite to a reducing atmosphere
relative surface energies, the two materials wet each other if the for different durations and at different temperatures, and examining
contact angle is less than 90°, else the interface is said to be non- the microstructures in a scanning electron microscope.
84 M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90

2. Experimental experiments comprised four tests, all at 850 °C, with the following
details:
Microcomposites of Ni and YSZ were used in the experiments. The
A. Dry 40% H2 diluted in Ar, resulting in p(Ar) = 0.59, p(H2) = 0.4,
samples tested were similar to those reported in [22,23]. For the
p(H2O) ≈ 0.003, log(p(O2))≈−23.6 and p(H2O)/p(H2) ≈ 0.008.
measurement, rectangular plates were cut from the green tape to
B. Dry 40% H2 diluted in He, resulting in p(He) = 0.59, p(H2) = 0.4,
approximately 6 × 37 mm2 size and 0.5–0.8 mm thickness. A hole of
p(H2O) ≈ 0.002, log(p(O2))≈−24.9 and p(H2O)/p(H2) ≈ 0.005.
3 mm diameter was made in both ends of each sample before
C. Wet H2 diluted in Ar with a small flow of O2, resulting in p(Ar) =
sintering at 1300–1400 °C.
0.70, p(H2) = 0.15, p(H2O) = 0.15, log(p(O2))≈−15.6, log(p
The measurement arrangement was a 4-point DC conductivity
(OH))≈−8.1 and p(H2O)/p(H2)≈1.
setup similar to that used in [20]. The sample was resting on a piece
D. Wet H2 diluted in He with a small flow of O2, resulting in p(He) =
of YSZ electrolyte plate placed on top of a rectangular porous YSZ
0.70, p(H2) = 0.15, p(H2O) = 0.15, log(p(O2))≈−15.4, log(p
piece, as sketched in Fig. 1 [7]. Pt current leads (ø0.3 mm) came
(OH))≈−8.1 and p(H2O)/p(H2) ≈ 1.
through the holes at the ends of the sample, were wrapped around
the sample a few times, after which the good contact was ensured The equilibrium gas compositions were calculated using the HSC
using Pt paste. The voltage probes were Ni wires placed 10 mm Chemistry® 6.1 software. The aim of the third set of experiments was
apart and pushed vertically down to the top surface of the sample to investigate the effect of humidity, gas compositions and the mass
by a weight of 130 g. The resistance was measured using a Keithley transport mechanism. If the mass transport occurs via the gas phase,
580 micro-ohmmeter (Keithley Instruments Inc.), which is a highly the light vs. heavy carrier gas might show differences in the
specialised micro-ohmmeter with a resolution down to 10 μΩ in the degradation behaviour of the composites. Each of the tests A–D
measurement range of relevance (around 200 mΩ). The experi- contain one redox cycle, that is, after the 1st reduction (of
ments were carried out in a vertical tube furnace with maximum approximately 217 h) the samples are re-oxidised and subsequently
temperature 1050 °C and the possibility to run with pure hydrogen re-reduced for another reduction period.
as well as wet reducing gas. Scanning Electron Microscopy (SEM) was carried out for several
The effect of different reduction temperatures and humidity samples after testing. The samples were mounted in epoxy, polished
conditions for the short-term (up to 200 h) conductivity changes and carbon coated. A field emission gun Zeiss Supra 35 SEM equipped
was measured. The test sequence was always to heat up the as- with both a lateral and Inlens secondary electron detector was used
sintered sample in air to the isothermal test temperature, flush the for the investigations. Low accelerating voltage (0.7–1 kV) was used
test chamber with nitrogen, and after that switch to the test for the electrons in order to obtain the charge contrast in the Inlens
atmosphere; the time line in the figures presenting the results has image, as described in [24]. The lateral secondary detector was used in
always the zero-point when the gas was (for the first time) parallel to obtain images of the uncharged microstructure. The
changed to reducing, i.e. when the (initial) reduction commences. samples for microscopy were exposed in a tube furnace to a
Different combinations of pure hydrogen and a diluting gas (N2, humidified (bubbling of the inlet gas through water at RT) mixture
Ar, He) were used. Humidification was done either by bubbling the of H2 (9%) and N2 (91%) to examine the effect of exposure time and
inlet gas through a water bottle at room temperature (referred to temperature on the microstructure. The exposure times were 5 and
as ‘humid’ conditions having 2–3% steam), or by a small controlled 96 h, and the different exposure temperatures were 643, 884 and
flow of oxygen into the furnace so that the humidity was 1026 °C.
produced inside the tube (referred to as ‘wet’ conditions). The
partial pressure of oxygen was logged with a p(O2) sensor 3. Results
adjacent to the samples. Measurement data were sampled with
4–15 min intervals. 3.1. Conductivity measurements
Three different sets of experiments were carried out. In the first
one, four experiments were carried out in “dry” hydrogen at Results from electrical conductivity measurements as a function of
isothermal test temperatures of 600, 750, 850 and 1000 °C, with time since the commencement of the reduction (the instant of change
two identical samples in each (the experimental value was taken as of the gas from inert to reducing) in pure dry hydrogen of the NiO–YSZ
the average of the two). The measured log(p(O2)/atm) at the composite are presented in Fig. 2 for the isothermal temperatures of
mentioned test temperatures were − 32.0, − 28.0, − 24.5 and 600, 750, 850 and 1000 °C, respectively. The measured levels of p(O2)
−21.8, respectively, and humidity less than 0.1% in all cases. Secondly, were given in experimental. At 600 °C, the initial conductivity is
to study the effect of humidity, two experiments were run at
temperatures 600 and 850 °C by controlled flows of H2 and O2. The
resulting conditions were log(p(O2))≈−23.0, log(p(OH))≈−12.1 at
600 °C, and log(p(O2))≈−16.5, log(p(OH))≈−8.3 at 850 °C. At both
temperatures, p(H2) and p(H2O) were about 80% and 20%, respec-
tively, and the resulting p(H2O)/p(H2) = 0.25. The third set of

Fig. 2. In-situ electrical conductivity based on 4-point DC measurement of Ni–YSZ


composite since the change of the gas to reducing dry hydrogen at isothermal
Fig. 1. Experimental setup for direct current conductivity measurements. temperatures from 600 to 1000 °C.
M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90 85

Fig. 3. In-situ electrical conductivity based on 4-point DC measurement of Ni–YSZ


composite since the change of the gas to reducing wet hydrogen at isothermal 600 and Fig. 5. Electrical conductivity of Ni–YSZ composite since the change of the gas to
850 °C. reducing in a dry mixture of 40 ml/min H2–60 ml/min He at 850 °C, together with the
measured oxygen partial pressure (experimental case B). Note that a redox cycle by
insertion of air is implemented at 197 h and the sample is re-reduced at 203 h.
slightly below 1200 S/cm and within the tested period the conduc-
tivity remains well above 1000 S/cm. By 50 h at 750 °C the Four additional tests were carried out to closer investigate the effect
conductivity has degraded to 900 S/cm. An isothermal hold at of humidity and the carrier gas. The results from experimental cases A to
850 °C leads to much sharper initial degradation so that at 160 h the D are shown in Figs. 4 to 7, respectively. Cases C and D suffer slightly from
measured value was only slightly above 600 S/cm (not shown). At noise in the signal due to the humidity/gas composition, but the average
1000 °C the initial degradation is remarkably fast so that the signal can be visually deduced. It can be observed that conductivity
maximum value measured was about 760 S/cm and already after degradation is slightly slower in dry H2 diluted with Ar (case A)
27 h the measured value was about 420 S/cm. It is noteworthy that compared to the case where the dilution is by He (case B). In wet
nearly all of the difference in the conductivity of the four measure- conditions the difference between Ar (case C) and He (case D) as the
ments in Fig. 2 is generated within the first 20 h after the start of the dilution gas becomes negligible. Secondly, complete re-oxidation in air
reduction; after that the rate of degradation in the present data set and subsequent re-reduction in cases A–D does not significantly change
does not vary significantly between the different temperatures. the early-stage electrical conductivity; the conductivity after the redox
The results from measurements made in wet conditions are shown cycle was initially slightly higher and degraded marginally faster in dry
in Fig. 3 for isothermal operation at 600 °C and 850 °C. The initial conditions (cases A and B), whereas in wet conditions (cases C and D) the
reduction at 600 °C in wet conditions is slower than in dry conditions slightly higher conductivity was retained until the end of the experiment.
(Fig. 2), furthermore, the initial conductivity is marginally above
1200 S/cm and after 160 h the measurement is still above 1100 S/cm, 3.2. Microscopy
that is, more than that observed in dry conditions after a shorter
period. In Fig. 3, the observed improvement of the conductivity within Fig. 8 graph A shows the secondary electron (SE) image of a Ni–YSZ
the first 10 h from the start of the reduction at 600 °C is due to slower composite exposed for 5 h to humidified 9% H2 diluted in N2 at 643 °C.
kinetics or reduction in wet conditions. The substantial steam partial An image obtained using an Inlens detector of the same area is shown
pressure slows down the removal of the steam generated upon in graph B. In the SE image, the lighter gray is metallic nickel, the
reduction of the NiO into Ni. Therefore it takes longer to achieve the darker gray is YSZ and black areas are pores. In the Inlens image
maximum conductivity. At 850 °C, the degradation of conductivity is (Fig. 8B), the brightest gray/white phase is percolating nickel and the
somewhat faster in wet conditions than in dry atmosphere. gray ‘islands’ appearing to hover over the plane of the image are non-

Fig. 4. Electrical conductivity of Ni–YSZ composite since the change of the gas to Fig. 6. Electrical conductivity of Ni–YSZ composite since the change of the gas to a
reducing in a dry mixture of 40 ml/min H2–60 ml/min Ar at 850 °C, together with the mixture of p(Ar) = 0.70, p(H2) = 0.15, p(H2O) = 0.15 at 850 °C, together with the
measured oxygen partial pressure (experimental case A). Note that a redox cycle by measured oxygen partial pressure (experimental case C). Note that a redox cycle by
insertion of air is implemented at 213 h and the sample is re-reduced at 221 h. insertion of air was implemented at 193 h and the sample was re-reduced at 200 h.
86 M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90

was not completely reduced after 5 h at 643 °C — this is plausible


remembering the times needed for full reduction at the lower
temperatures [23]. Images after an isothermal anneal at 643 °C of
96 h are shown in Fig. 9A (SE) and B (Inlens).
When the isothermal reduction was carried out for 5 h at 884 °C
and humidified 9% H2, the resulting microstructure is as shown in
Fig. 10A (SE) and B (Inlens). Further extending the reduction time to
96 h leads to microstructures presented in Fig. 11A (SE) and B
(Inlens). Still elevating the isothermal temperature to 1026 °C, Fig. 12
shows the microstructure after 5 h (A: SE and B: Inlens), and Fig. 13
after 96 h of exposure. With the magnification of the images in Figs. 8–
13 remaining the same, it is easy to observe the temperature
dependent increase in the characteristic Ni particle size and also the
loss of Ni percolation.

Fig. 7. Electrical conductivity of Ni–YSZ composite since the change of the gas to a
mixture of p(He) = 0.70, p(H2) = 0.15, p(H2O) = 0.15 at 850 °C, together with the 4. Analysis and discussion
measured oxygen partial pressure (experimental case D). Note that a redox cycle by
insertion of air was implemented at 230 h and the sample was re-reduced at 238 h. Conductivity measurements of the Ni–YSZ cermets showed that
significant short-term conductivity loss in dry reducing gas occurred
percolating nickel. Further, the Inlens shows some contrast between when reduction temperature reached about 850 °C and above and
the YSZ (darkest gray scale) and porosity (intermediate gray was very fast at 1000 °C. If the reducing gas was wet, this process was
appearing to be in the plane of the main image). By comparison of further amplified, but only above a certain temperature in the
microstructural features in Fig. 8 A and B the different phases can be neighbourhood of 800 °C. The examined cermet showed stable
identified. The darker area that can be perceived in the middle of conductivities of about 1100–1200 S/cm when reduced at 600 °C
especially larger Ni grains is with high probability NiO; the composite both in wet and dry atmosphere, whereas, for example, operation at

Fig. 8. SEM of a Ni–YSZ composite exposed for 5 h to humidified 9% H2 diluted in N2 at Fig. 9. SEM of a Ni–YSZ composite exposed for 96 h to humidified 9% H2 diluted in N2 at
643 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity. B: An 643 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity. B: An
Inlens detector image where light gray is percolating Ni, darker gray ‘islands’ appearing Inlens detector image where light gray is percolating Ni, darker gray ‘islands’ appearing
to hover over the plane of the image are non-percolating Ni. The darkest gray phase is to hover over the plane of the image are non-percolating Ni. The darkest gray phase is
YSZ and the medium gray areas in the plane of the image are porosity. YSZ and the medium gray areas in the plane of the image are porosity.
M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90 87

Fig. 10. SEM of a Ni–YSZ composite exposed for 5 h to humidified 9% H2 diluted in N2 at Fig. 11. SEM of a Ni–YSZ composite exposed for 96 h to humidified 9% H2 diluted in N2
884 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity. B: An at 884 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity.
Inlens detector image where light gray is percolating Ni, darker gray ‘islands’ appearing B: An Inlens detector image where light gray is percolating Ni, darker gray ‘islands’
to hover over the plane of the image are non-percolating Ni. The darkest gray phase is appearing to hover over the plane of the image are non-percolating Ni. The darkest gray
YSZ and the medium gray areas in the plane of the image are porosity. phase is YSZ and the medium gray areas in the plane of the image are porosity.

1000 °C dry or 850 °C rapidly degraded the conductivity to about 400– In Eq. (7), σo is understood as the maximum obtainable
500 S/cm. At the highest temperature tested (1000 °C) the maximum conductivity of the investigated cermet reduced in conditions under
value recorded was 760 S/cm, which means that about 500 S/cm which no conductivity loss due to Ni particle growth takes place;
(40%) of the maximum obtainable conductivity of the cermet based on the measurements the value used was σo = 1260 S/cm.
disappeared instantaneously and could not be measured even with Secondly, a linear combination of two exponential functions was used
continuous data logging with 3 min sampling steps. If the initial as shown in Eq. (8), where τ1 and τ2 are the time constants of the two
reference value for σo (or particle size) change was taken as the first functions. In both Eqs. (7) and (8) t is given in seconds.
measured value (here 760 S/cm at 1000 °C) instead of the maximum
obtainable (here 1260 S/cm), the ultra-rapid initial processes could
σ = σ0 −A1 ð1− expð−t = τ1 ÞÞ−A2 ð1− expð−t = τ2 ÞÞ ð8Þ
easily explain scatter in literature data also discussed by Holzer [5].
The observed good conductivity of the samples reduced at low
temperatures is in contrast to the results reported by Grahl-Madsen In Eq. (8), Ai reflect the fraction of the maximum obtainable (total)
et al. [25], who needed a high temperature or 1000 °C to reach the conductivity that disappears due to degradation through process i
maximum obtainable conductivity of the cermet. The probable reason with time constant τi. The measured conductivity data was analysed
for this difference is the coarse NiO mean particle size (d50) used by using both Eqs. (7) and (8) and the fitted parameters are shown in
Grahl-Madsen et al. (5.4–7.6 μm) compared to the present composites Table 1 from iterative fitting to Eq. (7) and in Table 2 from fitting to
with the d50 typically below 2 μm. The finer NiO phase can simply Eq.(8) using Matlab. The value of R2 gives a measure of the goodness
sinter and reach a better percolation at a lower temperature, while of the fit, with the value of 1 indicating a perfect fit to data. Generally,
reduction at higher temperatures removes the finest Ni fraction. The it can be noted that there were at least two different degradation
resulting reduction of the conductivity is attributed to particle growth processes with their respective time constants so that the quality of
and loss of percolation in the Ni phase after the initial phase. fits using Eq. (7) was less than using Eq. (8); a model using Eq. (7)
The measured conductivity data was modelled using two different typically overestimates the conductivity during early-in-life and
approaches. First, the expression from [11], Eq. (2), was rewritten as underestimates later when the conductivity plateau is approached.
The value of p is in the vicinity of 2 in all cases, whereas n is 4–5 at
 1 −p  lower temperatures and 6 at 850 °C or higher. It is questionable
σ = σ0 × 1−ð A × t Þn × r n ð7Þ
whether a physical meaning can be attached to the obtained values of
88 M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90

Fig. 12. SEM of a Ni–YSZ composite exposed for 5 h to humidified 9% H2 diluted in N2 at Fig. 13. SEM of a Ni–YSZ composite exposed for 96 h to humidified 9% H2 diluted in N2
1026 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity. at 1026 °C. A: SE image where lighter gray is Ni, darker gray YSZ and black is porosity.
B: An Inlens detector image where light gray is percolating Ni, darker gray ‘islands’ B: An Inlens detector image where light gray is percolating Ni, darker gray ‘islands’
appearing to hover over the plane of the image are non-percolating Ni. The darkest gray appearing to hover over the plane of the image are non-percolating Ni. The darkest gray
phase is YSZ and the medium gray areas in the plane of the image are porosity. phase is YSZ and the medium grey areas in the plane of the image are porosity.

n and p. A much better data fit was reached using Eq. (8) with two partial pressure remains at 600 °C at about 1E−12 atm or below even
different processes, one with a short and another with longer time at the highest p(H2O)/p(H2). The partial pressures increase nearly
constant. This confirms that there indeed are several growth exponentially when the temperature is elevated. For example at
processes overlapping. This appears plausible when considering the 850 °C, humidification of hydrogen by p(H2O)=1–4% gives a p(Ni-
initial size distribution in Ni particles. The finest Ni particle fraction is total) of about 8E−13–1E−12 atm, whereas the p(Ni-total) reaches
highly surface active and undergoes a rapid, if not nearly instanta- about 1E−9 at the highest p(H2O)/p(H2). For full assessment on
neous, morphological change if the circumstances are propitious for evaporative mass transport, a more sophisticated modelling approach
particle growth as also discussed by Holzer et al. [5]. It has also to be including microstructural effects should be employed. It cannot be
remembered that besides the Ni particle size distribution (PSD), there excluded, however, that evaporative mass transport does occur at
are several factors affecting the electrical conductivity of the cermet: high temperatures and especially in the presence of steam (high
total Ni content, the PSD of the YSZ phase, as well as total porosity and p(H2O)/p(H2)). This statement gets some support from the reported
pore size distribution. All these factors play a role on what changes the experiments, where the change in conductivity with time was found
aggregate of Ni particles with a PSD will undergo. In other words, the to be marginally faster with a lighter carrier gas despite the fact the
exact time behaviour will be cermet specific. A general observation the p(O2) was slightly lower; the lighter molecules of the carrier gas
from the measurements is that while the rapid conductivity loss was could facilitate gaseous mass transport of Ni species. Measured
completed within a few hours, the slower process takes at least tens of degradation in dry gas was faster when hydrogen was diluted with He
hours before the conductivity starts to stabilise. This time behaviour is (40% H2, 60% He, Fig. 5) compared to the case where H2 was diluted
in general agreement with results by Hauch et al. on time evolution of with Ar (40% H2, 60% He, Fig. 4). When the similar test was carried out
the polarisation resistance in Ni–YSZ cermets [26]. in wet gas (H2, steam, He or Ar), the cermet in the test where H2 was
It is also possible that the actual physical mechanisms change diluted with He (Fig. 7) degraded slightly faster than in the test where
between surface diffusion and evaporation–condensation. To assist in H2 was diluted with Ar (Fig. 6). The difference duel to the diluting gas
assessing the conditions for evaporation, thermodynamic calculations was smaller in wet conditions than in dry conditions.
on the total partial pressure of Ni species over solid Ni/NiO were The observed changes in conductivity are qualitatively explained
carried out in the temperature range 600–1000 °C using FactSage® in microscopy by increasing particle size in the Ni phase and loss of
Thermochemical Software and Databases [27]. Nine different steam to percolation in the cermet. For complete analysis of the microsctuc-
hydrogen ratios were calculated, ranging from very low values up to tures and direct comparison with the measured conductivity or
p(H2O)/p(H2)=8.9. Fig. 14 shows the results obtained. The total Ni literature data, microscopy combined with quantitative image
M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90 89

Table 1
Parameters obtained from fitting the measured conductivity data to Eq. (7) for either the first 33 h of the experiment, or the entire data series (the length in h given in the table).

Experimental case The first 33 h The entire measured time series

A n p R2 h A n p R2
a
600 °C dry 4.6E−23 4 2.2 0.39 29
750 °C dry 5.4E−22 5.1 2.1 0.75 54 3.4E−22 5.4 2.1 0.81
850 °C dry 1.7E−21 6 2.1 0.97 160 5.8E−21 6 2 0.97
1000 °C drya 5.4E−20 6 2.1 0.71 30
600 °C wet 3.6E−24 4 2.1 −0.2b 165 5.9E−24 4 2.1 0.61
850 °C wet 9.7E−21 6 2.1 0.89 90 6.3E−21 6 2.1 0.84
850 °C 40%He60%H2 (Fig. 5) 4.2E−21 6 2 0.99 197 3.1E−21 6 2 0.95
850 °C 40%Ar60%H2 (Fig. 4) 3.1E−21 6 2 0.87 214 1.7E−21 6 2 0.91
850 °C wet He (Fig. 7) 1.0E−21 6 2.3 0.73 230 6.3E−21 6 2.1 0.27b
850 °C wet Ar (Fig. 6) 4.3E−21 6 2.2 0.63 191 6.6E−21 6 2.1 0.39b
a
The duration of the experimental time series was 29 h at 600 °C and 30 h at 1000 °C.
b
The macroscopic fit is equal in quality to the other fits. The R2 value is taken down by a higher noise level in the signal observed under wet conditions.

Table 2
Parameters obtained from fitting the measured conductivity data to Eq. (8) for either the first 33 h of the experiment, or the entire data series (the length in h given in the table).

Experimental case The first 33 h The entire measured time series

A1 (S/cm) τ2 (s) A2 (S/cm) τ2 (s) R2 (h) A1 (S/cm) τ2 (s) A2 (S/cm) τ2 (s) R2

600 °C drya 82.3 0.01 119.0 9.47 0.44 29


750 °C dry 207.2 0.76 151.9 10.26 0.85 54 232.2 0.92 158.9 18.52 0.87
850 °C dry 317.9 0.08 218.7 12.02 0.97 160 381.8 0.27 261.1 43.25 0.97
1000 °C drya 621.6 0.19 207.1 4.57 0.97 30
600 °C wet 0.66 0.11 95.4 14.34 0.78 165 34.6 1e−12 103.7 51.97 0.87
850 °C wet 456.7 0.16 217.1 8.73 0.98 90 507.1 0.27 217.8 20.21 0.97
850 °C 40%He60%H2 (Fig. 5) 289.0 0.19 212.7 13.87 0.99 197 287.2 5e−15 285.8 26.46 0.96
850 °C 40%Ar60%H2 (Fig. 4) 321.0 0.21 131.5 13.63 0.98 214 359.8 4e−4 184.1 61.01 0.98
850 °C wet He (Fig. 7) 585.5 0.74 169.8 11.75 0.83 230 652.1 0.11 185.4 56.47 −1.2b
850 °C wet Ar (Fig. 6) 610.7 0.77 156.7 16.78 0.56 191 631.2 0.10 175.8 32.99 −0.4b
a
The duration of the experimental time series was 29 h at 600 °C and 30 h at 1000 °C.
b
The macroscopic fit is equal in quality to the other fits. The R2 value is taken down by a higher noise level in the signal observed under wet conditions.

analysis should be linked to samples where the conductivity was The changes in the electric conductivity were related in micros-
actually measured. copy analysis to particle growth and loss of percolation in the Ni
phase. No definite conclusions can be made on the details of the
5. Conclusions particle growth mechanisms; marginally faster early degradation was
observed when the H2 was diluted with He compared to the case
It was shown that the initial electrical conductivity and the short- where dilution was by Ar. It is possible that part of the observed
term conductivity loss of Ni–YSZ cermets depend closely on the particle growth and percolation loss originates from gas phase mass
operating conditions (operating temperature and steam content) and transport.
that the loss processes can be nearly instantaneous. Ni growth can be
reduced by lower the operation temperature.
The examined cermet type showed stable conductivities of about
1100–1200 S/cm when reduced at 600 °C both in wet and dry
atmosphere; the performance was slightly better and more stable in
wet conditions. The maximum obtainable conductivity of the
investigated cermet was estimated to be about 1260 S/cm.
Short-term changes in conductivity in dry reducing gas took place
when the reduction temperature was elevated. When the tempera-
ture reached about 850 °C and above, the changes were rapid and
substantial. If the reducing gas was wet, this process was further
amplified, but only above a certain temperature in the neighbourhood
of 800 °C. Operation at 1000 °C in dry conditions or at 850 °C and wet
conditions degraded the conductivity nearly instantaneously to about
750–800 S/cm and to about 400–500 S/cm within 30 h from the initial
reduction.
The measured conductivity data could be satisfactorily described
by a linear combination of two exponential functions, one with a short
and the other with a longer time constant. This implies that at least
two overlapping processes or mechanisms contribute to conductivity Fig. 14. Total partial pressure of Ni gas species over solid Ni/NiO depending on
change, possibly depending on the initial particle size distribution of temperature and the ratio p(H2O)/p(H2). The calculations were carried out using using
Ni and cermet microstructure. FactSage® Thermochemical Software and Databases [27].
90 M.H. Pihlatie et al. / Solid State Ionics 189 (2011) 82–90

Acknowledgements [9] M.C. Muñoz, S. Gallero, J.I. Beltrán, J. Cerdá, Surf. Sci. Rep. 61 (2006) 303–344.
[10] S. Sunde, J. Electrochem. Soc. 143 (3) (1996) 1123–1132.
[11] Y.-M. Chiang, D.P. Birnie, W.D. Kingery, Physical Ceramics; Principles for Ceramic
M. Pihlatie acknowledges Nikolaos Bonanos of Risoe DTU for Science and Engineering, John Wiley & Sons Inc, 1997.
support on the conductivity measurements and Karl Thydén of Risoe [12] W.D. Kingery, H.K. Bowen, D.R. Ulhmann, Introduction to ceramics, John Wiley &
Sons, New York, USA, 1976.
DTU for support in both the experimental work and microscopy. M. [13] C. Herring, J. Appl. Phys. 21 (April 1950) 301–303.
Pihlatie was financially supported during his Risoe employment by [14] P. Wynblatt, Acta Mater. 48 (2000) 4439–4447.
the Marie Curie Intra-European Fellowship, contract number MEIF- [15] A. Tsoga, P. Nikolopoulos, A. Naoumidis, Ionics 2 (1996) 427–434.
[16] X. Mantzouris, N. Zouvelou, D. Skarmoutsos, P. Nikolopoulos, F. Tiezt, J. Mater. Sci.
CT-2005-023882, as part of the European Commission's 6th frame- 40 (2005) 2471–2475.
work programme, as well as by Energinet.dk under the project PSO [17] M. Mogensen, K.V. Hansen, Impact of impurities and interface reaction on
2007-1-7124 SOFC R&D. Other authors were supported by the said electrochemical activity, in: W. Vielstich, H.A. Gasteiger, H. Yokokawa (Eds.),
Handbook of Fuel Cells — Fundamentals, Technology and Applications, Advances
Energinet.dk project.
in Electrocatalysis, Materials, Diagnostics and Durability, Volume 5, John Wiley &
Sons, Ltd, 2009.
References [18] J.I. Beltran, S. Gallego, J. Cerda, J.S. Moya, M.C. Muñoz, Phys. Rev. B 68 (2003)
075401.
[1] D. Simwonis, F. Tietz, D. Stöver, Solid State Ionics 132 (2000) 241–251. [19] J.S. Moya, S. Lopez-Esteban, C. Pecharromán, Prog. Mater. Sci. 52 (2007)
[2] S.P. Jiang, J. Mater. Sci. 38 (2003) 3775–3782. 1017–1090.
[3] J. Sehested, J.A.P. Gelten, I.N. Remediakis, H. Bengaard, J.K. Nørskov, J. Catal. 223 [20] T. Klemensø, M. Mogensen, J. Am. Ceram. Soc. 90 (2007) 3582–3588.
(2004) 432–443. [21] R. Vassen, D. Simwonis, D. Stöver, J. Mater. Sci. 36 (2001) 147–151.
[4] J. Sehested, J.A.P. Gelten, S. Helveg, Appl. Catal., A 309 (2006) 237–246. [22] M. Pihlatie, A. Kaiser, P.H. Larsen, M. Mogensen, J. Electrochem. Soc. 156 (3)
[5] L. Holzer, B. Münch, B. Iwanschitz, M. Cantoni, Th. Hocker, Th. Graule, J. Power (2009) B322–B329.
Sources (2010), doi:10.1016/j.jpowsour.2010.08.006. [23] M. Pihlatie, A. Kaiser, M. Mogensen, Solid State Ionics 180 (2009) 1100–1112.
[6] T. Klemensø, C. Chung, P.H. Larsen, M. Mogensen, J. Electrochem. Soc. 152 (11) [24] K. Thydén, Y.L. Liu, J.B. Bilde-Sørensen, Solid State Ionics 178 (2008) 1984–1989.
(2005) A2186–A2192. [25] L. Grahl-Madsen, P.H. Larsen, N. Bonanos, J. Engell, S. Linderoth, J. Mater. Sci. 41
[7] M. Pihlatie, Stability of Ni–YSZ Composites for Solid Oxide Fuel Cells During (2006) 1097–1107.
Reduction and Re-oxidation, Doctoral Dissertation, VTT Publications 740, VTT [26] A. Hauch, M. Mogensen, A. Hagen, Solid State Ionics (2010), doi:10.1016/j.
Technical Research Centre of Finland, Espoo, 978-951-38-7400-1, 20108 http:// ssi.2010.01.004.
www.vtt.fi/inf/pdf/publications/2010/P740.pdf. [27] C.W. Bale, P. Chartrand, S.A. Decterov, G. Eriksson, K. Hack, R. Ben Mahfoud, J.
[8] M.W. Finnis, J. Phys. Condens. Matter 8 (1996) 5811–5836. Melançon, A.D. Pelton, S. Petersen, Calphad J. 62 (2002) 189–228.

You might also like