Chemical interactions between aluminosilicate base sealants and the components on the anode side of solid oxide fuel cells

Article abstract of DOI:10.1149/1.1467945

The chemical interactions between several aluminosilicate glasses and components at the anode side of SOFC have been investigated. Severe reactions occur if the modifier ions are Ba and Ca. MgO base sealants have been investigated in detail. The formation of some detrimental phases are seen. Cordierite forms with many reaction mixtures of sealants with ZrO₂ stabilized with 8 mol % Y₂O₃ (as electrolyte) or nickel (as anode). On the other hand, with oxide dispersion strengthened Cr5Fe1Y₂O₃ and steel, another detrimental phase, cristobalite, is favored. There appears to be a competitive formation of these two detrimental phases in some cases. In some of the interactions none of the detrimental phases appears. The diffusion behavior of the cations across the interfaces has been investigated. Among all the cations, chromium diffuses the maximum. The diffusion coefficient of chromium for the diffusion couples of different sealants with ODS alloy have been determined.

Full text of DOI:10.1149/1.1467945

Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
A607
0013-4651/2002/149͑5͒/A607/8/$7.00 © The Electrochemical Society, Inc.
Chemical Interactions Between Aluminosilicate Base Sealants
and the Components on the Anode Side of Solid Oxide
Fuel Cells
N. Lahl, D. Bahadur,a K. Singh, L. Singheiser, and K. Hilpertz
¨
Institute for Materials and Processing in Energy Systems, Forschungszentrum Julich GmbH,
¨
D-52425 Julich, Germany
The chemical interactions between several aluminosilicate glasses and components at the anode side of SOFC have been inves-
tigated. Severe reactions occur if the modifier ions are Ba and Ca. MgO base sealants have been investigated in detail. The
formation of some detrimental phases are seen. Cordierite forms with many reaction mixtures of sealants with ZrO2 stabilized with
8 mol % Y2O3 ͑as electrolyte͒ or nickel ͑as anode͒. On the other hand, with oxide dispersion strengthened Cr5Fe1Y2O3 and steel,
another detrimental phase, cristobalite, is favored. There appears to be a competitive formation of these two detrimental phases in
some cases. In some of the interactions none of the detrimental phases appears. The diffusion behavior of the cations across the
interfaces has been investigated. Among all the cations, chromium diffuses the maximum. The diffusion coefficient of chromium
for the diffusion couples of different sealants with ODS alloy have been determined.
© 2002 The Electrochemical Society. ͓DOI: 10.1149/1.1467945͔ All rights reserved.
Manuscript submitted July 23, 2001; revised manuscript received November 26, 2001. Available electronically March 29, 2002.
The development of a suitable sealant is still an important de-
mand for planar solid oxide fuel cell ͑SOFC͒ systems consisting of
cell stacks because of the stringent requirements such as gas tight-
ness, insulating properties, matching thermal expansion coefficients,
good soldering behavior, and the chemical stability in both the oxi-
dizing and reducing atmospheres.1-3 The sealant, being in contact
with all other components of planar SOFCs, is prone to degradation
due to chemical interaction at the interfaces of these individual com-
ponents and subsequent structural modifications. Literature on this
aspect of interaction is rather limited.4-6 Several glass and glass
ceramics have been studied as candidate materials for sealants. We
have recently investigated many aluminosilicate glasses as potential
sealant in the system AO-SiO2-Al2O3-B2O3-N ͑A ϭ Ba, Ca, and
Mg; N ϭ nucleating agents͒.7 By detailed differential thermal
analysis ͑DTA͒, X-ray diffraction ͑XRD͒, thermal expansion, and
electron microscopy work, we have determined the important char-
acteristics of these glasses and their crystallization kinetics.7,8
The chemical interactions of selected sealants with other materi-
als of SOFC components at the anode side were investigated here.
These materials are (i) ZrO2 stabilized with 8 mol % Y2O3 ͑8YSZ͒
for the anode and the electrolyte, (ii) Ni for the anode, and (iii)
oxide dispersion strengthened ͑ODS͒ Cr5Fe1Y2O3 alloy as well as
the steel Fe18Cr1Al ͑DIN 1.4742͒ for the interconnect.
LaCrO3-based perovskites used as ceramic interconnect are not con-
sidered in this study. There is a general trend to use metallic inter-
connects in planar SOFC systems. The sealants were selected in
order to elucidate the influence of varying alkaline earth metal ions
as well as varying content of alumina and of TiO2 .
The rationale behind the choice of the specific aluminosilicate
systems investigated and the specific issues addressed are described
in the following. The different alkaline earth metals used, Ba, Ca,
and Mg, act as modifiers and have widely different chemical prop-
erties such as field strength, ionic radius, and electronegativity.
These properties have strong influence on the thermal expansion
coefficient as well as the reactivity. Our objective was to investigate
this influence. TiO2 is generally used as a nucleating agent in alu-
minosilicate glasses. TiO2 is known to induce phase separation in
glasses and hence influences significantly crystallization kinetics
and phase formation at the interfaces. We have intentionally varied
the TiO2 content to see its influence on the above-mentioned behav-
ior. The investigation of the kinetics and the products of the reac-
tions between the different sealant compositions and the SOFC com-
ponents is the task of this paper. Many of these reaction products
could be detrimental for SOFCs. Hence, it is necessary to look for
the conditions which inhibit or avoid the formation of detrimental
phases. Moreover, interdiffusion has to be investigated. The most
mobile species have to be identified and their diffusion coefficients
determined. Measures to reduce diffusion are of interest.
Experimental
The glass compositions considered here together with their label
are given in Table I. The details of the synthesis of glass and their
characterization are given elsewhere.7 Grain sizes were not deter-
mined for each sealant composition. The typical size ranged be-
tween 1 and 20 ␮m. The chemical interactions between the sealants
and the components of SOFC were investigated by annealing their
powder mixtures under oxidizing and reducing conditions. Oxidiz-
ing conditions means annealing under ambient air. Reducing condi-
tions were obtained by using initially a gas mixture composed of Ar
with 4% H2 , which was humidified by passing over a water bath at
84°C and subsequently through a condenser at 56°C, leading to a
H2 /H2O ratio of 4. The grain size of the SOFC components in the
powder mixtures amounted to Ͻ1 ␮m for 8YSZ, Ni, Cr5Fe1Y2O3 ,
and Fe18Cr1Al. Diffusion couples were generally made by sand-
wiching a polished glass plate in between a polished interconnect
alloy plate and a polished 8YSZ plate. In some cases the glass plate
in the sandwich was replaced by a glass paste made of glass powder
and an organic binder. The diffusion couple results by annealing the
sandwich at temperatures above the glass transition temperature.
Phase analysis was conducted by XRD ͑Philips 1050, Eind-
hoven, Netherland; Cu K␣ for all samples except sample MAST12
for which Co K␣ radiation is used͒. The chemical composition of
the phases were determined by scanning electron microscopy with
energy-dispersive X-ray analysis, SEM/EDX ͑CAMSCAN/Tracor
Northern, Middleton, WI͒ and transmission electron microscopy
with EDX, TEM/EDX ͑Philips model CM 200 with GIF, energy
filter, Eindhoven, Netherland͒. Using SEM/EDX, the diffusion pro-
file of different cations was determined across the interfaces.
Results and Discussion
Table II shows the details of the phases identified by XRD, SEM/
EDX, and TEM/EDX in the powder mixtures of the sealants with
different components of SOFCs after annealing them at 1000°C for
1000 h under oxidizing and in part reducing conditions. The mate-
rials chosen were 8YSZ, Ni, ODS, and steel. The ASiO3 ͑A
ϭ Ba,Ca,Mg͒ is a common product in most of the powder mixtures.
Besides this, other phases are formed. Some of these phases are
a Permanent address: Department of Metallurgical Engineering and Materials Sci-
ence, Indian Institute of Technology Bombay, Mumbai-400076, India.
z E-mail: k.hilpert@fz-juelich.de
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A608
Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
perature. The nucleating agent TiO2 was added to explore whether
the growth of these undesirable phases can be suppressed by chang-
ing the crystallization kinetics.
Table I. Nominal composition of the glasses „in mol %… and their
label.
Label
BaO
45
CaO
45
MgO
SiO2
Al2O3
B2O3
TiO2
Influence of alkaline earth metals on chemical inter-
action.—Figure 1 shows the typical SEM of interfaces of ͑a͒ BAS
ϩ ODS, ͑b͒ CAS ϩ ODS, and ͑c͒ MAS ϩ ODS heated at 1000°C
under three different conditions. The first one is heated under reduc-
ing conditions for 1000 h, the second one under oxidizing conditions
for 2 h, and the third one under oxidizing conditions for 1000 h.
Though the experimental conditions are different, the results ob-
tained for the interaction are generally valid for different sealants
containing BaO, CaO, and MgO. The BaO and CaO base glasses
show vigorous crystallization and reaction at the interface, whereas
the MgO base glass exhibits very sober crystallization, and reaction.
Similar results were obtained by experiments with powder mixtures.
The crystallization kinetics of these glasses has already been studied
by us earlier.7 Moreover, it has been shown that the CaO base glass
in contact with 8YSZ gives rise to the formation of m-ZrO2 , which
is detrimental to SOFC as mentioned previously. This is due to the
diffusion of yttrium from 8YSZ into the glass. The detailed discus-
sion of this interaction is given in an earlier paper from this
laboratory.9 We have recently observed that the activation energy of
crystallization increases significantly as we change the alkaline earth
ion from Ba to Ca to Mg. This has been explained in terms of
increasing field strengths of the alkaline earth ions.7 It has been
observed that the diffusion couple of steel with the sealant BAS
shows melting at the interface with detrimental corrosion attack of
BAS
CAS
45
45
5
5
5
5
MAS
45
45
5
5
MAS10
MAST10
MAST5
MAST12
40
38
43
35.6
45
45
45
43.1
10
10
5
5
5
5
3.6
2
2
4.9
12.8
known to be detrimental to SOFCs. The phase m-ZrO2 is formed
during the interaction of glass CAS with 8YSZ, which exhibits very
high specific volume compared to stabilized zirconia.9 A further det-
rimental phase is cordierite (Mg2Al4Si5O18) formed in several in-
teraction studies of MgO base glass with other components of
SOFCs, in agreement with the phase diagram of the
MgO-Al2O3-SiO2 system. The thermal expansion coefficient ͑TEC͒
of this phase is very low (2 ϫ 10Ϫ6/C) compared to that of other
components of the fuel cell which is detrimental to SOFCs. The
formation of another detrimental phase, cristobalite (SiO2), occurs
in many cases of interaction of the sealants with interconnect mate-
rials ͑ODS or steel͒. Cristobalite is known to undergo a structural
transformation at 200°C involving a large volume change, which
causes microcrack formation during cooling from the sealing tem-
Table II. Phases identified in powder mixtures of different sealants with SOFC components after their annealing at 1000°C over up to 1000 h
in air or in reducing conditions „see Experimental section….a
Ni
YSZ
ODS
Steel
Sealant
Air
Red
Air
Red
Air
Red
Air
BaSiO3
NiO
BaAl2Si2O8
BaSiO3
Ni
Ba2Al2O5
BaAl2Si2O8
CaSiO3
Ni
BaSiO3
BaAl2Si2O8
BaAl2Si2O8
Ba2Zr2Si3O12
BaSi2O5
BaCrO4
Cr2O3
BaSiO3
Ba2SiO4
Cr2O3
BaAl2Si2O8
CaSiO3
Cr2O3
BaCrO4
Ba3Fe32O51
Ba2Si3O8
BAS
CaSiO3
NiO
CaSiO3
CaSiO3
Ca2SiO4
m-ZrO2
Mg2SiO4
ZrSiO4
CaSiO3
Cr2O3
CaSiO3
Ca3Cr2(SiO4)3
(Fe0.6Cr0.4)2O3
CAS
CaAl2Si2O8
Ca3Zr(Si2O9)O2
Mg2SiO4
CaAl2Si2O8
Mg2SiO4
Ca3Cr2(SiO4)3
(Mg,Fe)2SiO4
Spineld
Ca3Cr2(SiO4)3
Cr2O3
Mg2SiO4
NiO
MAS
Ni
Spineld
ZrSiO4
(Fe0.6Cr0.4)2O3
c
e
e
e
SiO2
SiO2
SiO2
b
b
Spineld
Cr2O3
Mg2SiO4
NiO
Mg2SiO4
Ni
ZrSiO4
Mg2SiO4
ZrSiO4
Mg2SiO4
(Mg,Fe)2SiO4
Spineld
SiO2
MAS10
MAST10
MAST5
(Fe0.6Cr0.4)2O3
(Fe0.6Cr0.4)2O3
(Fe0.6Cr0.4)2O3
e
e
SiO2
NiO
Ni
Mg2SiO4
Spineld
Cr2O3
ZrSiO4
Mg2SiO4
ZrSiO4
Mg2TiO4
Spineld
e
e
SiO2
SiO2
Mg2SiO4
NiO
Mg2SiO4
8YSZ
Mg2SiO4
Mg2SiO4
Mg2TiO4
Mg2SiO4
Spineld
Spineld
ZrSiO4
ZrO2 ͑cub͒
b
b
e
ZrO2͑cub͒
SiO2
Cr2O3
b
b
e
SiO2
b
MAST12
Mg2SiO4
NiO
MgAl2O4
Mg2SiO4
ZrSiO4
(Mg,Fe)2SiO4
Spineld
ZrO2͑cub͒
SiO2 ,e MgCr2O4
b
b
a The phases MgSiO3 and Mg2Al4Si5O18 are present in the powder mixtures of MgO based sealants with other SOFC components unless otherwise
specified.
b No Mg2Al4Si5O18
.
c No MgSiO3 , no Mg2Al4Si5O18
d Means (Mg,Fe)(Cr,Al)2O4 .
e Represents crystobalite.
.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
A609
Figure 1. SEM micrographs after annealing at 1000°C of the following
interfaces: ͑a, top left͒ BAS ϩ ODS under reducing condition for 1000 h,
͑b, top right͒ CAS ϩ ODS under ambient air for 2 h, and ͑c, left͒
MASϩ ODS under ambient air for 1000 h.
Figure 2. ͑a͒ XRD pattern of powder mixtures of MAS and MAST5 with 8YSZ heated at 1000°C for 500 h in air. The XRD pattern of sealant MAST5 is shown
for comparison. ͑b͒ XRD pattern of powder mixture of MAST12 with 8YSZ under similar conditions as ͑a͒, with Co K␣ radiation.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A610
Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
Figure 3. Variation of Tg and Tc ͑determined by DTA͒ with different mol %
TiO2 as nucleating agent in MgO-Al2O3-SiO2-B2O3 glass composition.
Figure 5. XRD pattern of powder mixtures of ͑a͒ MAS ϩ ODS, ͑b͒
MAS10 ϩ ODS ͑100 h͒, and ͑c͒ MAS10 ϩ ODS ͑500 h͒ heated at 1000°C.
the alloy after annealing at 1000°C for 1000 h under reducing con-
ditions. FeO is possibly present under these conditions, forming
mixtures with low eutectic temperatures. Severe interactions were
also observed at the CAS/8YSZ interface leading to the diffusion of
Ca into the 8YSZ.
In view of the low activation energy of crystallization of BaO
and CaO base glasses and the extensive reaction of these glasses
with other components of SOFCs, we have further carried out de-
tailed investigations on MgO base glasses, which are now reported
and discussed.
and possibly Zr ions from 8YSZ diffuse to favor formation of
ZrSiO4 phase in one of the separated regions. The kinetics of this
phase formation is probably much faster than the incubation period
for the cordierite phase. Phases such as MgSiO3 and Mg2SiO4 get
favorably crystallized in the other phase-separated region. MgSiO3
is the second major phase of the reaction products of 8YSZ and the
sealants MAS or MAST5 which contain 0 or 2% TiO2 , respectively.
The Mg2SiO4 phase is present in minor amount. If the TiO2 content
is increased to 4.9% ͑MAST12͒, the second major product phase is
Mg2SiO4 . It seems that the content of TiO2 has a strong influence
on the nucleation and crystallization behavior of the glass. Both the
nucleation and crystallization temperatures decrease if the TiO2 con-
tent is raised ͑see Fig. 3͒. This could change the phase separation
phenomena and crystallization kinetics of the glass, resulting in an
increase of nuclei of Mg2SiO4 as more TiO2 is used as nucleating
agent. According to a report of Ray et al.,10 the increase in nuclei in
a glass causes a decrease of Tg and Tc values in DTA. As seen in
Fig. 3, the decrease in Tg and Tc with increase in TiO2 content may
refer to the increase of the nuclei of Mg2SiO4 which become maxi-
mum when the TiO2 content is 4.9% in sealant MAST12.
Figure 4 shows the XRD pattern of powder mixtures of ͑a͒
MAS ϩ steel, ͑b͒ MAST5 ϩ steel, ͑c͒ MAS10 ϩ steel, and ͑d͒
MAST10 ϩ steel heated at 1000°C for 1000 h in air between the 2␪
values of 10 and 40°. The diffraction line at about 10.5°, which
corresponds to the cordierite phase, and the line at about 21.9°,
which may originate from both the cordierite and the cristobalite
phase, is considered in the following. The vertical line drawn is a
guide for observing minor shift in line position, which may be due
to overlapping line positions of cordierite and cristobalite. The in-
tensity ratio of the two lines of cordierite ͑one at ϳ10.5° and the
other at ϳ21.9°͒ can be taken as a guideline to unravel qualitatively
the relative volume fractions of the cristobalite and the cordierite
phases. From this figure, it is clear that the cordierite phase is not
formed in the powder mixture of MAS ϩ steel, but the cristobalite
phase can be identified. In contrast to this, the cordierite and the
cristobalite phases are formed if the glass MAS was given the same
heat-treatment as the mixture. It appears that chromium from the
steel acts as an inhibitor for the growth of cordierite in these powder
mixtures. The amount of cordierite phase is negligible in the
MAST5 ϩ steel mixture. However, for the powder mixtures of seal-
ant MAS10 or MAST10 with steel, the amount of cordierite phase is
substantial and possibly the volume fraction of the cristobalite phase
is small. There seems to be a competition between the formation of
these two phases. Cordierite phase seems to be the favored phase if
the alumina content is 10%, as can be expected on the basis of the
phase diagram of the MgO-Al2O3-SiO2 system. It is possible that
Chemical interactions of MgO base glasses.—Figure 2a shows
the XRD pattern of powder mixtures of MAS and MAST5 with
8YSZ heated at 1000°C for 500 h in air. XRD data of the sealant
MAST5 ͑heat-treated under similar conditions͒ is given for compari-
son. Figure 2b shows the XRD pattern of powder mixtures of
MAST12 with 8YSZ under similar conditions as described previ-
ously. It is interesting to note that the glass MAS by the interaction
with 8YSZ gives rise to the formation of cordierite. However, the
cordierite phase is not formed by this interaction if the glass MAST5
with 2% of TiO2 is used. In contrast to this, the cordierite phase is
formed in the glass MAST5 after similar heat-treatment as the pow-
der mixture. The cordierite phase is again suppressed in the interac-
tion of 8YSZ with the sealant MAST12 containing 4.9% TiO2 as
nucleating agent. It is known that TiO2 induces phase separation,
Figure 4. XRD pattern of powder mixtures of ͑a͒ MAS ϩ steel, ͑b͒
MAST5 ϩ steel, ͑c͒ MAS10 ϩ steel, and ͑d͒ MAST10 ϩ steel heated at
1000°C for 1000 h.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
A611
Figure 7. XRD pattern of powder mixtures of ͑a͒ MAS10 ϩ Ni, ͑b͒ MAS
ϩ Ni, and ͑c͒ MAS after annealing under reducing conditions at 1000°C for
500 h.
Figure 6. Element mapping analysis of powder mixtures of MAS10
ϩ ODS heated at 1000°C for 500 h by TEM/EDX.
Figure 8. SEM micrographs of interface of MAST5 ϩ ODS ͑a, top left͒
without special treatment of the alloy surface before joining, ͑b, top right͒
after oxidizing the alloy surface before joining, and ͑c, left͒ diffusion profile
of chromium in ͑a͒ and ͑b͒. Both samples were given heat-treatment at
1000°C for 50 h.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A612
Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
initially the cristobalite phase is favored. However, with longer du-
ration, the cordierite gets nucleated. This conjuncture gets strength-
ened when we discuss the results of the powder mixtures of the
sealants MAS and MAS10 with ODS.
As shown in Table II, one of the reaction products of ODS with
most MgO base sealants is a spinel phase whose approximate com-
position is estimated as ͑Mg,Fe͒(Cr,Al)2O4 . Besides the spinel
phase, the cristobalite phase is also present in many reaction prod-
ucts. In addition, the cordierite phase is present if the Al2O3 content
of the sealant is 10%. Again, chromium seems to play the role of an
inhibitor for the cordierite phase.
Figure 5 shows the XRD patterns of powder mixtures of ͑a͒
MAS ϩ ODS, ͑b͒ MAS10 ϩ ODS ͑500 h͒, and ͑c͒ MAS10
ϩ ODS ͑100 h͒ heated at 1000°C. The role of Al2O3 content and the
competition between cristobalite and cordierite phase formation can
also be seen here. If the MAS10 ϩ ODS mixture is heated for 100
h, the cordierite phase is not seen, but the presence of cristobalite is
obvious. However, if heated for 500 h, the cordierite phase forms
whereas the cristobalite phase is not evident. Similarly, in the pow-
der mixture of MAS ϩ ODS heated at 1000°C for 500 h, the cordi-
erite phase is unable to nucleate although the presence of cristobalite
phase is obvious. Kata et al.11 and Imanaka et al.12 have shown that
cordierite as well as many other aluminum base ceramics can sup-
press the formation of cristobalite in borosilicate glass and the com-
mon factor for this suppression is the Al3ϩ ion. It has also been
demonstrated that cristobalite prevention depends on the amount of
alumina present in the composite. Jean and Gupta13,14 have dis-
cussed the suppression of cristobalite in the presence of a small
amount of alumina in borosilicate as well as high-silica glasses. The
suppression has been attributed to the strong affinity between Al3ϩ
and alkali ions forming a Kϩ- and Al3ϩ-rich reaction layer more
rapidly than the nucleation of cristobalite.13 However, this is only
possible if the alumina content is above a threshold concentration. It
may also be possible for the present set of samples that the content
of alumina plays an important role and the competitive formation
reaction between cristobalite and cordierite depends upon the con-
tent of Al3ϩ as well as temperature and time.
Figure 6 shows the approximate phase distribution of powder
mixtures of MAS10 ϩ ODS heated at 1000°C for 500 h on the basis
of element mapping by TEM/EDX. The detailed analysis ͑not in-
cluded here͒ shows an affinity between Mg and Cr. The affinity
between Mg and Cr ions has also been evidenced by SEM/EDX in
the diffusion couples of different sealants with ODS leading to the
formation of a spinel phase as discussed in the previous section.
Silicon is often found in the reaction zone in which chromium and
magnesium are diffused if the interface is annealed for a longer
duration. However, silica does not form any phase with chromia.
This is also known from the phase diagram of the binary and ternary
systems. The phase diagram of Cr2O3-SiO2 further suggests the
formation of cristobalite.
The XRD pattern of powder mixtures of ͑a͒ MAS ϩ Ni and ͑b͒
MAS10 ϩ Ni heated in reducing atmosphere at 1000°C for 500 h
are shown in Fig. 7. Figure 7c shows the XRD pattern of sealant
MAS heated under similar conditions as the powder mixtures for
comparison. Forsterite (Mg2SiO4) formation is shown by all the
XRD patterns in Fig. 7. The cordierite phase is formed in pure MAS
as well as in the mixture of MAS10 with Ni. It is interesting to note
that in the mixture of Ni with MAS, Ni seems to favor the formation
of forsterite (Mg2SiO4) phase, which presumably suppresses the
formation of cordierite in the powder mixture MAS ϩ Ni. As
shown in Fig. 7, the cordierite phase is not suppressed if the sealant
MAS is given the same heat-treatment as the mixtures. However, the
amount of forsterite seems to be small in this material and the
amount of MgSiO3 ͑protoenstatite͒ is the highest as compared to
that of other product phases formed. This phase is completely absent
in the powder mixture of sealant MAS with nickel. It seems that the
diffusion of Ni in the glass lattice is helpful in nucleating the for-
sterite (Mg2SiO4) phase which in turn suppresses the formation of
Figure 9. Typical diffusion profiles of different cations at the interface of ͑a,
left͒ MAS ϩ ODS and ͑b, right͒ MAST5 ϩ ODS annealed at 900°C for 380
and 12 h, respectively.
cordierite when the alumina content is low ͑5%͒. It has been dem-
onstrated that nickel promotes the forsterite formation in the
MgO-CaO-Al2O3-SiO2 glass system.15 Nickel in the ionic state can
substitute for the Mg2ϩ ion in the octahedral site of the forsterite
structure. Comparing the XRD pattern of MAS and MAS ϩ Ni
powder mixtures, it is pertinent that Ni induces the formation of
forsterite phase in the powder mixture of MAS ϩ Ni rapidly enough
so as to suppress the formation of cordierite. This is not the case for
the pure sealant MAS in which the favored phase is MgSiO3 ͑pro-
toenstatite͒.
Interdiffusion at the sealant/interconnect interface.—Figure 8
shows the SEM micrographs of the interface of MAST5 ϩ ODS ͑a͒
without special treatment of the alloy surface before joining, ͑b͒
after oxidizing the alloy surface before joining, and ͑c͒ the diffusion
profile of chromium in ͑a͒ and ͑b͒. The two samples were heat-
treated at 1000°C for 50 h. The interface made after oxidation of the
alloy shows a better adhesion, and the reaction at the interface is
significantly less as compared to the interface made without preoxi-
dation. The advantage of preoxidation has also been described ear-
lier in the literature.16-18 Oxygen plays an important role in promot-
ing strong bonds between metal and oxide ceramics. Oxygen as a
surface active agent lowers the surface tension of the metal, enhanc-
ing the spreading of the liquid, which leads to better adhesion. It has
been shown by Jiang and Silcox19 that the formation of a graded
chromium oxide layer at the interface yields good adhesion. Jiang
and Silcox19 have carried out a detailed investigation by electron
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
A613
Table III. Diffusion coefficient data for chromium in different sealants determined from the diffusion across the interface of selected sealants
with the ODS alloy under different conditions of annealing.
Diffusion coefficient (cm2 sϪ1
)
Sealant
Kind of glass
Heat-treatment
900°C/380 h
780°C/3 h ϩ 900°C/500 h
780°C/3 h ϩ 900°C/500 h
780°C/3 h ϩ 900°C/500 h
900°C/12 h
900°C/50 h
900°C/100 h
1000°C/24 h
1000°C/50 h
MAS
5% Al2O3
10% Al2O3
2.0ϫ10Ϫ14
1.56ϫ10Ϫ14
1.56ϫ10Ϫ14
3.1ϫ10Ϫ13
1.56ϫ10Ϫ14
3.5ϫ10Ϫ14
2.4ϫ10Ϫ14
1.1ϫ10Ϫ12
1.2ϫ10Ϫ12
MAS10
MAST10
MAST5
MAST5
MAST5
MAST5
MAST5
MAST5
10% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
5% Al2O3 ,2% TiO2
energy loss spectroscopy ͑EELS͒ at the interfaces of chromium/
alkaline earth boroaluminosilicate glasses. An important observation
of their EELS studies is that the initial oxidized layer consists of
Cr2ϩ ions, which possess high oxygen affinity, and is therefore con-
sidered to be chemically very active. If such a preoxidized layer is
joined with the magnesium aluminosilicate glass, there is a strong
tendency of the Cr2ϩ ion to go to higher valence states than 2ϩ. The
magnesium ion being a modifier ion has relatively high mobility.
Driven by the difference in the heats of oxide formation, ion ex-
change between Cr2ϩ and Mg2ϩ could occur, giving rise to metallic
magnesium and higher oxidation states of chromium. The close af-
finity between magnesium and chromium ions has already been
demonstrated by our SEM/EDX and TEM/EDX studies ͑cf. previous
section͒. Therefore, it is expected that magnesium will segregate at
the interface, having two effects: (i) the magnesium-rich layer will
form a barrier layer for the diffusion of chromium and (ii) the
SiO 4ϩ tetrahedral network will be severely distorted, thereby fur-
2. If the time of heat-treatment at a fixed temperature for a par-
ticular sample is varied, it was observed that the diffusion coefficient
decreases with increasing time for the heat-treatment ͑cf. results for
sealant MAST5͒. This can be explained by the crystallization of the
glass which restricts the movement of ions.
3. If the temperature of heat-treatment is varied but the duration
was kept fixed ͑cf. results of sealant MAST5 at 900°C over 50 h and
at 1000°C over 50 h͒, the diffusion coefficient is high for high
temperatures as expected.
4. For the interface of MAST5 ϩ ODS, with a preoxidized sur-
face of ODS, the diffusion coefficient amounts to
8
ϫ 10Ϫ14 cm2 sϪ1 ͑not shown in Table III͒, which is much less com-
pared to the value without preoxidation, which amounts to 1.2
ϫ 10Ϫ12 cm2 sϪ1. Both the samples were given the same heat-
treatment ͑1000°C for 50 h͒. This aspect is also reflected by Fig. 8
as discussed previously. Thus, the diffusion of chromium can be
reduced substantially by preoxidizing the surface of ODS before
making the interface.
͓
͔
4
ther reducing the open structure and increasing the viscosity of the
sealant inhibiting diffusion of chromium.
Conclusion
Figure 9 shows typical diffusion profiles of different cations at
the interface of ͑a͒ MAS ϩ ODS, and ͑b͒ MAST5 ϩ ODS an-
nealed at 900°C for 380 and 12 h, respectively. The two diffusion
profiles exhibit essentially similar features. The main diffusing ions
are chromium from ODS as well as Mg from the sealant. These
elements are present in significant amounts within the reaction layer.
This is in agreement with the formation of spinel phase in the reac-
tion layer containing magnesium and chromium ions, subsequently
suppressing the formation of cordierite. However, cristobalite, an-
other detrimental phase, is formed instead ͑see Table II͒.
From the diffusion profile of chromium across the interface of
ODS with different sealants, the diffusion coefficient was estimated.
The results are presented as relative atomic fraction vs. distance into
the reaction layer from the ODS to the glass side. The concentration
of Cr decreases from the ODS surface into the diffusion layer with
respect to time. The concentration of Mg decreases from sealant to
the ODS side. From the diffusion profile of the diffusing cations the
diffusion coefficient is evaluated by using the following relation
The chemical interactions between different aluminosilicate base
glass sealants and components on the anode side of SOFCs have
been investigated. The sealants with BaO and CaO as network modi-
fiers are not found suitable due to extensive crystallization and re-
action at the interfaces. MgO base glasses, which have been studied
in more detail with varying alumina content or with TiO2 as nucle-
ating agent, offer better possibilities. The formation of detrimental
phases such as cordierite and cristobalite was observed in different
powder mixtures and diffusion couples. We have made an attempt to
get insight into the conditions under which these phases are formed.
The implication of these investigations is that the formation of these
phases can be avoided by suitably choosing the composition and the
nucleating agent. The interdiffusion investigations reveal that the
diffusion of chromium is maximum across the interfaces of sealant/
interconnect. A rough estimate suggests that chromium may diffuse
up to 0.9 mm in a time period of 50,000 h if operated at 1000°C,
which is considered the lifetime of SOFCs.
The Institute for Materials and Processing in Energy Systems,
Forschungszentrum Julich GmbH, assisted in meeting the publication costs
of this article.
c x,t͒
x
͑
¨
ϭ 1 Ϫ erf
͓1͔
ͫ
ͬ
c0
2ͱDt
References
where c(x,t) is the concentration of ions at the distance, x, from the
surface of Cr5Fe1Y2O3 for a given diffusion time period, t. The
term c0 is the initial concentration of ions and D the diffusion co-
efficient. The term erf stands for error function and its values vary
from 0 to 1 for varying values of x2/4Dt between 0 and 3. The
diffusion coefficient of Cr in the sealants of the diffusion couples are
summarized in Table III.
1. K. L. Ley, M. Krumpelt, R. Kumar, J. H. Meisser, and I. Bloom, J. Mater. Res., 11,
1489 ͑1996͒.
2. S. V. Phillips, A. K. Dutta, and L. Lakin, in Proceedings of the 2nd International
Symposium on Solid Oxide Fuel Cells, F. Grosz, P. Zegers, S. C. Singhal, and O.
Yamamoto, Editors p. 737, Commission of the European Communities, Report
E4R 13456EN, Athens, Greece ͑1991͒.
3. Y. M. Sung, J. Mater. Sci., 31, 5421 ͑1996͒.
4. Y. Sakaki, M. Hattori, and Y. Esaki, in Solid Oxide Fuel Cells V, U. Stimming, S.
C. Singhal, H. Tagawa, and W. Lehnert, Editors, PV 97-40, The Electrochemical
Proceedings Series, p. 652, Pennington, NJ ͑1997͒.
Some of the noteworthy features are
1. Diffusion couples of different sealants given the same heat-
treatment yield essentially similar values for the diffusion coeffi-
cients.
¨
5. C. Gunther, G. Hofer, and W Kleinlein, in Solid Oxide Fuel Cells V, U. Stimming,
S. C. Singhal, H. Tagawa, and W. Lehnert, Editors, PV 97-40, p. 746, The Elec-
trochemical Society Proceedings Series, Pennington, NJ ͑1997͒.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A614
Journal of The Electrochemical Society, 149 ͑5͒ A607-A614 ͑2002͒
6. T. Horita, N. Sakai, and T. Kawada, Denki Kagaku, 61, 760 ͑1993͒.
7. N. Lahl, K. Singh, L. Singheiser, K. Hilpert, and D. Bahadur, J. Mater. Sci., 38,
3089 ͑2000͒.
8. N. Lahl, L. Singheiser, K. Hilpert, K. Singh, and D. Bahadur, in Solid Oxide Fuel
Cells VI, S. C. Singhal and M. Dokiya, Editors, PV 99-19, p. 1057, The Electro-
chemical Society Proceedings Series, Pennington, NJ ͑1999͒.
9. L. Kinderman, D. Das, D. Bahadur, R. Nikel, and K. Hilpert, J. Am. Ceram. Soc.,
80, 909 ͑1997͒.
12. Y. Imanaka, A. Shigenori, N. Kamehara, and K. Niwa, J. Am. Ceram. Soc., 78,
1265 ͑1995͒.
13. J. H. Jean and T. K. Gupta, J. Mater. Res., 8, 2393 ͑1993͒.
14. J. H. Jean and T. K. Gupta, Phys. Chem. Glasses, 36, 176 ͑1995͒.
15. L. Barbieri, A. M. Ferrari, C. Leonelli, T. Manfredini, G. C. Pellacani, S. Bruni, and
F. Cariatti, Phys. Chem. Glasses, 36, 176 ͑1995͒.
16. A. Zanchetta, P. Lefort, and E. Gabbay, J. Eur. Ceram. Soc., 15, 233 ͑1995͒.
17. A. A. Zanchetta, P. Lortholary, and P. Lefort, J. Alloys Compd., 228, 86 ͑1995͒.
18. C. Beraud, M. Courbiere, C. Esnouf, D. Juve, and D. Treheux, J. Mater. Sci., 24,
4545 ͑1989͒.
10. C. R. Ray, W. Huang, and D. E. Day, J. Am. Ceram. Soc., 74, 160 ͑1991͒.
11. K. Kata, Y. Shimada, and H. Takamizawa, IEEE Trans. Compon., Hybrids, Manuf.
Technol., CHMT-13, 448 ͑1990͒.
19. N. Jiang and J. Silcox, J. Appl. Phys., 87, 3768 ͑2000͒.
Downloaded on 2014-12-04 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).