Ceramic ion conducting membranes and their technological

C. R. Acad. Sci. Paris, t. 1, Sdrie II c, p. 533-543, 1998
I~lectrochimie et photochimie/ Electrochemistry and photochemistry
(Chimie de 1'6tat solide et cristallochimie/Solid state chemistry and crystal chemistry)
Ceramic ion conducting membranes
and their technological applications
Brian STEELE
Centre for Ceramic Ion Conducting Membranes, Dept. of"Materials, Imperial College, London SW7 2BP, UK
E-mail: b.steeleez)ic.ac.uk
(Received 7 May 1998, accepted after revision 23 July 1998)
Abstract - Dense membranes of ceramic oxygen ion conductors are of great interest because of their ability to
transmit high oxygen fluxes with total selectivity. These membranes may be solid electrolytes exhibiting negligible electronic conductivity, or mixed ionic-electronic conductor~. Parameters controlling the ionic fluxes are
outlined, and the factors influencing material selection for a wtriety of applications are also summarised. These
applications include solid oxide fuel cells operating on natural gas, devices to generate oxygen from air, and reactors designed to produce hydrogen and syngas by the partial oxidation of methane. © AcadeMe des sciences/
Elsevier, Paris
ceramic membrane / ionic conductivity / oxygen fluxes/oxygeJa surface exchange
R6sum6 - Membranes c6ramiques & conduction ionique et leurs applications technologiques. Les
conducteurs d'ions oxyg~ne utilisant des membranes denses sont d un grand inter& en raison de leur aptitude ~i
emettre des flux dlevds d'oxyg&ne, tout en preservant une totale selectivite. Ces membranes peuvent etre des electrolytes solides de conductivite electronique ndgligeable, ou des conducteurs electronique-ionique mixtes. Les
parametres qui contr61ent ces flux ioniques sont m is en evidence et les facteurs qui influencent la selection des
materiaux resultants pour de multiples applications sont dgalement recapitules. Ces applications comprennent
les piles fi combustibles fi oxydes solides fonctionnant au gaz natt rel, les procedes permettant de produire de
l'oxygene fi partir de ['air, et les reacteurs conqus pour produire de l'k ydrog&ne et du syngaz ~tpartir de l'oxydation
partielle du methane. © Academic des sciences/Elsevier, Paris
membranes cdramiques / conductivitd ionique/flux d'oxygene/~change superficiel d'oxyg~ne
Version frangaise abr6gde
Certaines membranes cdramiques sont capables de supporter d'importants flux
(~1 A c m 2 ou 3,5 cm ~ 0 2 cm 2.min-l). Ces flux sont contr61& par certains param&tres,
concentration (C,) et la diffusion (D,) des lacunes d'ions oxyg&ne ( ~ ' ) , qui reprdsentent
l'amplitude de la conductivitd spdcifique ionique (0",) de celles-ci, ainsi que la cin&ique de
d'dchange interfaciale (k) de l'oxygSne ~t la surface :
1/202 + 1/~'+ 2e = O~
d'oxyg/me
tels que la
dgalement
la rdaction
(i)
Les ions oxyg&nes migrent d'un c6td ~ l'autre de la membrane sous l'influence d'un gradient de
potentiel dlectrochimique, et sont ensuite lib&& sous forme d'esp&es gazeuses par une reaction similaire ~ l'expression (i) se d&oulant en sens inverse. La charge dlectronique dolt &re pourvue d ' u n c6td
de la membrane et supprimde de l'autre. Ces flux dlectroniques peuvent &re fournis de mani~re
Invited paper.
B. SIeele
interne ~ l'aide d'un matdriau conducteur dlectronique-ionique mixte, ou peuvent &re apport& par
un conducteur dlectronique externe connectd aux faces opposdes de la membrane dlectrolyte c&amique. Notons qu'un important flux d'oxyg~ne fi travers une membrane d'Epaisseur L, n&essite de
lCaibles valeurs de rdsistivitd sp&ifique de surface (L/¢r) du mat&iau cEramique, ainsi qu'une faible
rdsistivitd interf:aciale, que l'on peut ddmontrer &re dqui~ alente ~tD*/~.k. I1 est clair que le changement
d'un flux d'oxyg~ne contr61d par une rdaction de surface ~i un transport interne survient lorsque les
deux termes prdcddents sont dgaux, c'est4l-dire pour L. = D %. D~s lors, ~ moins qu'une demarche ne
soit &ablie pour amdliorer le flux ~i travers I'interf:ace, il n'y a aucun avantage ~t fabriquer des membranes tr~s minces. Heureusement, les valeurs numdriques de D* et de k, ainsi que l'dpaisseur critique
L C, peuvent &re obtenues gffice ~l des mesures d'dchanges isotopiques 180/160. De plus, il est possible
d'obtenir des valeurs numdriques pour D,, et Q. avec des fluorites dopdes (A408) et des oxydes de perovskites (A2B206), car, pour ces structures, les anions ft. l'intdrieur du rEseau peuvent supporter
d'importantes concentrations de lacunes d'ions oxyg~ne. Quelques exemples peuvent etre citds, tels
que Zr0,85Y0,150 1,925 (YSZ) et Lao,6Sro,4CoO~_ ~. Par ailleurs, des recherches se poursuivent dgalement
sur les syst&mes pyrochlores (A2B2OT) et brown-millerites (A2B2Os), ainsi que sur des structures plus
complexes d'intercroissance interne.
Bien qu'initialement les membranes cdramiques aient ~tE &udides pour leur bonnes propri&& de
transport, d'autres facteurs deviennent cruciaux lorsque le matdriau est incorpord dans un syst~me rdel.
La membrane cdramique dolt non seulement &re thermochimiquement stable, mais dgalement thermom&aniquement robuste pour rdsister aux conditions opdratoires ; cela afin de s'assurer que des
composants fiables, ~i durde de vie prdvisible, puissent etre fabriqu& ~conomiquement dans l'optique
d'applications sp&ifiques. Ces applications comprenncnt les piles ~i combustibles ~t oxydes solides
(SOFC), les gdndrateurs d'oxyg~ne et les rdacteurs produisant du syngaz et de l'hydrog~ne par oxydation partielle du mdthane.
La technologie des SOFC peut &re ddpartag& en deux domaines principaux. Les syst~mes ~i haute
tempdrature (850-900 °C) pourront, dans l'avenir, atre couplds ~i une turbine ~l gaz pour fburnir des
rendements de l'ordre de 70 %. I1 est envisagd que la puissance de telles unit& soit, en fin de compte,
de l'ordre de 10-100 MW. L'unitd de demonstration de 100 kW de Westinghouse, qui a dtd mise en
service en mars 1998 aux Pays-Bas, constitue un jalon important pour le ddveloppement de cette technologie, que l'on pense pouvoir &re en mesure de co:~amercialiser lors de la prochaine ddcennie.
L'autre domaine concerne les piles ~i combustibles ~l oxydes solides fonctionnant ~l tempdrature intermddiaire (IT-SOFC). Le cofit des syst~mes SOFC ~l haute temperature depend principalement du
choix des matdriaux devant &re utilisds pour les plaques bipolaires et autres composants de ces syst~mes ayant un rapport avec la manipulation des gaz ~lce'.; temp&atures dlevEes. De faqon ~l utiliser des
matdriaux meilleurs march&, tels que l'acier inoxydable, il est souhaitable d'opdrer ~l une &helle de
temperature intermddiaire, de l'ordre de 500-700 °C. D autre dlectrolytes, tels que Ce0,gGd0,1OL95 et
La0,!)Sr0,1Ga0,sMg0,202,ss, sont en cours d'&ude. Pour des piles fbnctionnant 5 500 °C, la rdussite du
processus a ddj'a 6td rapportEe, et des entreprises, telles que Sulzer (SOFCO), ont 6tabli la fiabili[d
d'unit& de 1 k W ~i 650 °C. Cette technologie IT-SOFC devrait &re appliqude ~i petite &helle ~l des
unites C H P (1-1 O0 kW) fonctionnant au gaz naturel su r des sites r&identiels et commerciaux, ainsi
que pour des vdhicules dlectriques utilisant le mdthanol, et probablement l'essence.
La production d'oxyg6ne provenant de Fair peut &re rdalisde gr'ace -a une pompe dlectrique ~i oxygbne, qui prdsente de nombreuses caractdristiques identiq ues aux systbmes SO FC. Cependant, dans ce
cas, l'anode et la cathode doivent toutes les deux fonctionner avec une activitd d'oxygbne dlevde. Un
sdparateur, fonctionnant gr~ce ~i une diffdrence de pression, et comportant un conducteur mixte soumis 'h un gradient de pression partielle d'oxyg~ne, peut &re congu. Ce concept a l'avantage de supprimer les circuits dlectriques externes et la distribution de courant dlectrique interne.
Finalement, l'utilisation de membranes c&amiques pour la production de syngaz ~i partir du
mdthane suscite un intdr~t croissant. Cette rdaction s'effectue gdndralement fi grande &helle, par recyclage avec de la vapeur d'eau. N&nmoins, ce proces,,;us endothermique consommant beaucoup
d'dnergie et d'un prix de revient trSs 6levd, ne convient pas fi des sites isolds ou ~t petite dchelle. L'oxydation partielle du mdthane semb]e aujourd'hui plus attrayante, car l'oxygSne peut &re fourni par une
membrane cdramique conductrice ionique. Deux groupe:~ internationaux importants se sont d'ailleurs
crdds pour ddvelopper cette technologie. La contrainte principale concernant le mat&iau repose sur la
534
Ceramic ion conducting membranes
stabilitd de celui-ci sur un trSs large &entail d'activitd d'oxyg~ne, allant de l'air jusqu'~i un environnement fortement rdducteur. Le matdriau pdrovskite Lao,sSro,2Cro,eFeo,803 a dtd dtudid par BP America, et Argonne/Amoco ont rendu compte de donndes intdressantes sur le rapport
conversion/sdlectivitd, en utilisant un matdriau "a intercroissa~ ce interne, Sr4Fe4,sCo1,sO13. Une autre
faqon d'aborder le problSme consiste "a utiliser un matdriau composite comprenant deux phases,
comme par exemple les composites YSZ/Pd, dont une phase fournit la conductivitd ionique et l'autre,
la conductivitd dlectronique
Les membranes conductrices d'ions oxyg&ne ont un avenir prometteur devant elles, avec un dventail
large et varid d'applications. Bien que de nouvelles amdliorations soient ndcessaires concernant les performances, la fiabilitd chimique et mdcanique, la simplification du systbme et les cofits, le niveau des
connaissances actuelles permet d'envisager dans l'avenir le ddveloppement commercial de ces membranes conductrices d'ions oxygSne.
1. Introduction
The properties and applications of porous
inorganic membranes used in the chemical
industry are well documented [1, 2]. The
present survey, however, is concerned with the
science and technology of dense ceramic ion
conducting membranes (CICM) which are
emerging as an important class of materials
with many potential applications, particularly
in the ener D" sector. The: principle Factors generating this interest are energy conversion efficiency and reduced environmental pollutants.
Most activities, at present, are focussed on oxygen ion conducting membranes [3, 4], but
readers should also be aware of preliminary
investigations on protonic [5], and alkali ion
[6] conducting membranes.
The oxygen flux through a dense polycrystalline oxide membrane can be represented by the
schematic diagram represented in JTgureI.
Oxygen ions are injected into the membrane by
an interracial reaction of the type:
!/202 + Ve;" + 2e- u O)~
(1)
The oxygen ions migrate to the opposite side
of the membrane under the influence of an
electrochemical potential gradient, and are
then released as gaseous species by a similar
reaction to expression ill operating in the
reverse direction. Clearly electronic charge has
to be provided at the left hand side of the membrane and removed at the right hand side.
These electronic fluxes can be provided internally by using a mixed electronic and ionic conducting material (MIEC), or can be supplied by
j o 2- r
1
l:~E
~
RE
1
Figure 1. Schematic equivalent circuit fi)r oxygen flux
through dense ceramic oxide membrane.
Figure 1. Circuit dquivalent schdmatique pour le flux
d'oxyg&ne ~l travers une paroi c&amique dense d'oxyde.
an external electronic conductor connected to
the opposite faces of the CICM via electrodes if
the membrane is a predominately ionic
conductor (electrolyte). It should be immediately noted that high oxygen fluxes through
these membranes requires low values for Ro,
and also implies fast kinetics for the interracial
reactions represented by equation [1]. For this
introductory survey it will be assumed that
these interracial reactions can be represented by
a simila:" ohmic resistive term, R>:.
2. Materials design and selection
Ionic fluxes through CICM
materials
2.1.
Assuning simple ohmic behaviour the oxygen ion flux (Ji, m°l'cm 2"s 1) can be repre535
B.
Steele
sented by equation ~ (Appendix, box A) where
the area specific resistance (ASR) term (RxA)
has been replaced by L/(5 i, where Gi is the total
oxygen ion conductivity, and L (cm) is the
thickness of the ceramic membrane, it should
be noted that the actual ionic conductivity will
include contributions from both the intrinsic
grain (bulk) conductivity and grain boundary
conductivity, and that both need to be optiraised.
components in the system can be fabricated
from relatively cheap stainless steel alloys.
There is considerable interest, therefore, in synthesising novel oxide structures/compositions
which exhibit appropriate transport properties
at intermediate temperatures.
Design of alternative CICM materials
requires an understanding of the behaviour of
microscopic parameters, such as oxygen
vacancy diffusion coefficient (Dr), and concentration of oxygen vacancies (C,,) which are
included in expression y (Appendix, box A).
Useful values of D,, and Q can be provided by
doped fluorite (A4Os) and perovskite (A2B206)
oxides as the anion lattices in these structures
can accommodate significant concentrations of
oxygen ion vacancies. Examples include
Zrl.85Y0.15Ol.g25 (YSZ) (electrolyte), and
La0.6Sr0.4Co3_x (mixed conductor). However,
there are constraints on the concentrations of
mion vacancies that can be tolerated at intermediate temperatures due to ordering/clustering phenomena which effectively reduce the
concentration of mobile vacancies. Attention is
also being given to associated structures such as
Ehe pyrochlore (A2B207) and brown-millerite
(A2B205) systems which may be regarded as
having oxygen vacancies already contained
within the undoped stoichiometric composi-
For most technological applications it is
desirable that the oxygen fluxes through C I C M
components attain values around 1 A cm -, and
so with typical values of the electric potential in
the range 0.5-1.0 V it follows that the ASR
term should be as low as possible. Adopting a
typical target value of 0.15 ~ cm 2 enables the
membrane thickness at different temperatures
to be determined from figure 2 for a variety of
C I C M components. Fabrication procedures
limit the minimum thickness at which dense
pin-hole free ceramic membranes can be reliably manufactured to about 15 btm. Whilst
many of the conducting oxides in figure 2 have
sufficient ionic conductivity at elevated temperatures (> 800 °C) to allow the use of membranes thicker than 15 btm, it is desirable for
many applications to maintain the operating
temperature below 700 °C so that ancillary
Temperature (°C)
I
800 700
I
I
600
500
I
I
400
I
t
La Cr(Mg) 03
-------] ~
/
rR o = tJa = 0,15 ohmcm 2 ]
300
Stainless steel
Bi-polar
Cr-Fe (Y2Oa), Inconel - AI20~
"
Lao~Sro~CoosFeo20 a ~
-^ q, c'^ ~o c,
F
plate
material
1500pro
Self-supported
electrolytes
Mixe::lConduclors
150pm
5"
(o
09
-2
cn
O
_J
-3
1,5pm
-4
O.15pm
15pro
~'7 ~
08
~
~
- -'01G'=~
-'''Qa3(caNb'}%(H°)
I
I
I
I
I
1.0
1.2
1.4
IO00FF (K-1)
16
1.8
Supporled
electrolyles
~L
2.0
Figure 2. Specificionic conductivityvaluesfor selectedceramicoxide ion conducting membranesas a fhnction of reciprocal
temperature.
Figure 2. Choix de quelques valeursde conductivirdionique de c&amiquesconductricesen fonction de l'inversede la temp&ature.
536
Ceramic ion conducting membranes
tion. More complicated intergrowth structures,
such as the Ruddleston-Popper phases, e.g.
SrO(La0.rSr0.3MnO3) 2 ,are also being investigated, and further examples will be provided
later.
2.2.
Interfacial kinetics
As already indicated high fluxes (-1 A c m -2)
through C I C M materials must be accompanied
by rapid interfacial reactions. A value of 1 A
cm 2 translates into 1.56 x l0 is molecules O 2
cm-2.s -I (Appendix, box B). This flux requirement can be placed into perspective by considering how many 02 molecules per unit area (n)
collide with an oxide surface, assuming temperature and pressure valnes of 1000 K and
1 bar, respectively.
n = p/(2TcmkT) 1/2 = 2.2 x 1024cm-2.s -1
This implies that approximately 1 in 106 of
the molecules striking the surface has to be
adsorbed which requires high values for the
accommodation coefficient which is not normally a feature of high temperature gas/solid
reactions. Moreover, if we assume 10i4-1015
oxygen sites cm -2, then the site turnover rate is
about 10 ~s-i. Again this value represents a very
high rate for a heterogeneous reaction. It is not
surprising therefore that the surface exchange
reaction can often limit the oxygen flux
through C I C M materials.
Although the oxygen surface exchange process involves a series of individual reaction steps
including; adsorption, dissociation, charge
transfer, surface diffusion of intermediate surface species (eg: Oad, 02ad, Oa~t, 0 2-,,g,etc.), and
finally incorporation into a vacancy in the surface layer, the overall reaction can be represented by an expression similar to that already
given in equation (1). In subsequent discussions it is important to realise that the above
equation is not a charge-transfer reaction as it
involves the neutral combination of charged
species. The kinetics of' this surface exchange
process are governed by the oxygen surface
exchange coefficient (k.cm.s -1) which can be
determined by 180/t60 isotopic exchange
measurements [7]. The interracial resistive term
(R E in figure 1) can be transformed [4] into the
expression D*/Gik which incorporates values of
k. Taking into account these interracial reaction
resistive terms it follows that the oxygen flux is
now given by expression [3 in Box B. Clearly the
changeover from a situation in which the oxygen flux is controlled by the surface reaction to
one controlled by bulk diffusion is given by Ro
= R E for a single interface, i.e. L/G i = D*/c~ik,
from which a critical thickness (L~) can defined,
L~ = D*/k.
For a given value of k the oxygen flux will be
controlled by the surface exchange kinetics
until the critical membrane thickness is reached
(figure3), and there is no advantage to be
gained by fabricating very thin membranes
unless steps are taken to improve the flux
througa the interracial region. Relevant actions
could include increasing the effective surface
area, and improving k values by the deposition
of catalytically active surface species. A selection of relevant permeation data as a function
of reciprocal temperature is provided in
figure ~'.
2.3. O t h e r selection c r i t e r i a
It is obviously necessary to ensure that the
C I C M material is thermodynamically stable in
the oxTgen partial pressure gradient imposed
during operation. However it should also be
noted that even if a multicomponent nonstoichiometric phase is judged to be thermodynamically stable it can still decompose in an
oxygen partial pressure gradient due to kinetic
&mixing. if the mobilities of the slower moving cations are different then concentration
gradieF ts are developed in the oxide such that
Ie+I
le+O
'c
E
le-1
o
le-2
I~
le-3
E
o
o~-
o-~-~e-<-:.L-
- o _..._
.,g
le4
le5
o'1
~ le-6
0
-.JO~ La/CalColFe = 6:4:2:8
La/SrlColFe = 6:4:2:8
La/SrlCo = 5:5:10
~7~- La/SrlMn = 5:5:10
~x
"~\
\~\
le-7
le-8
le-4
I
le-3
I'-le-2
I
le-1
"---1
:'le+O
le+1
Membrane thickness I cm
Figure "i. Oxygen flux through oxide membranes at
800 °C calculated as a function of membrane thickness
using dala for D* and k [equation [3, box A].
Figure 3. Flux d'oxyg~ne~ttravers des membranesd'oxydes, calculds en fonction de leur dpaisseur, 'a 800 °C,
d'apr8s ks valeursde D* et k calculdesd'apr~s I'dquation[3
(box A).
537
B. S t e e l e
Oxygenpermeation
TempeqlRlJre.°C
~200 IlO0 1000 900
I
=.- 2
E
I
1
604]
700
r
800
l
1=1
-
~-
(el
>Ill - -
,,,-......
'
07
08
(a) StFeC-.,Oo.$03.l~,_ 6
(c) SrCOo.gFeo.2Oa4
(e) Lao 3S,o~;oOa.4
Ig) ~ sS¢osco034
-,<,
l
09
I
10
]
11
!
12
13
m0ur(~,xl
(b) (Bb~O~o2s-{Y203)o2s-Ag (35v~%}, 9011171
3. Applications
(cOLao.2Sro.sCoo.sFeo.zO~hl~
(f~ Lao.¢Sr0.,Ooo2Feos034
(h) YSZ-Pd (40voPI.), continuous PO phase
3.1. Fuel cell technology
Figure 4. Oxygen permeation data for selected membranes as a function of reciprocal temperature.
Figure 4. Donndes de pcrmdation de l'oxyg~ne pour quelques membranes choisies, en fonction de l'inverse de la
tempdrature.
the membrane interface exposed to the higher
oxygen partial pressure becomes enriched with
the faster moving cation species. The degradation of SrCoo.sFeo.203_ ~ membranes, for example, has been attributed to this phenomenon.
Performance degradation may also occur due to
the formation of oxyhydroxides, oxycarbonates, etc. as many of the constituent oxides
(e.g. La203, SrO) are very basic and may react
w i t h , H 2 0 or C O 2 present in the feed gases.
Structural failure due to thermal expansion
mismatch, stresses generated by temperature
gradients, crack growth due to stress corrosion
and other causes, are other factors that have to
be considered during the materials selection
and preliminary design stages. It is apparent
that many other properties in addition to the
transport behaviour have to be taken into
account to ensure that reliable components
with predictable lifetimes can be manufactured
economically for the proposed application.
2.4. Dual phase composite membranes
The comments in the preceding section indicate that it can be difficult to select a single
phase mixed conducting oxide that can satisfy
all the selection criteria. It is appropriate, therefore, to consider separating the ionic and electronic transport functions and to fabricate a
dual-phase composite membrane with two
538
interpenetrating percolation pathways for the
oxygen ions and electrons. Most investigators
in this area have used a combination of a
ceramic electrolyte with a noble metal (eg: Ag,
Pt, etc.), although preliminary measurements
have been conducted on dual oxide systems
such as: YSZ-LaMnO 3, and another obvious
candidate is C G O - L S C E Results for the
dual-phase system (Bi203)0.v5(Y203)0.25-Ag
(35 vol%) are included in the compilation of
permeation data reproduced in figure 4.
3.1.1. General aspects
Although the principles of fuel cells were first
demonstrated by Grove almost 100 years ago
[81 it has proved difficult to produce fuel cell
systems that can compete effectively with existing power generation systems based on combustion
processes.
The
electrochemical
oxidation of hydrogen rich fuels can provide
electrical generation system efficiencies in
excess of 60 % [9] with the formation of only
minor quantities of environmental pollutants
such as N O x due the absence of very high temperature combustion reactions. However unless
technological breakthroughs occur in the cost
of hydrogen and its storage then the concept of
a i~ydrogen economy will remain a futuristic
chimera. It is necessary, therefore, to assume
that for the foreseeable future, except for
selected niche markets, fuel cells will be supplied with a fossil fuel (e.g., natural gas) which
has to be converted to a hydrogen-rich fuel in
an appropriate reformer. This requirement
introduces problems for the polymeric, alkaline, and phosphoric acid fuel cells which typically operate in the temperature range
60-200 °C. For these systems, due to the endothermic steam reaction, the reformer cannot be
imegrated into the fuel cell stack, and energy
has to be supplied to an external reformer
wMch reduces the overall system efficiency. For
example, operating on pure hydrogen the electrical conversion efficiency of a polymer fuel
cell can approach 60 % whereas operating on
natural gas with an external fuel processor the
overall system efficiency drops to around 35 %
[10] which is not much better than a diesel
entwine operating under optimal load condi-
Ceramic ion conducting membranes
tions. This reduction in overall efficiency coupled with the associated increase in system
complexity makes it difficult at present for low
temperature fuel cell systems to be competitive
for power generation. Much effort at present is
concerned with developing alternative fuel
processing strategies (e.g. partial oxidation
routes) to try to reduce the efficiency losses
associated with external reformers.
In contrast the relatively high operating temperatures of the molten carbonate (MCFC) and
solid oxide (SOFC) fuel cell types allow the
reforming reaction to be integrated within the
fuel cell stack ensuring high electrical conversion efficiencies (50-55 %). The low power
density and complexity of the MCFC system
together with lifetime issues under real operating conditions are problems that still have to be
addressed for this type of fuel cell which
unlikely to be competitive at sizes less than
1 M W which is another barrier to entry into
the market. SOFC systems are much more versatile in that they are designed to operate over a
wide range of temperatures (500-1000°C)
with system sizes varying between 1 kW and
10 MW. Moreover, compared to other fuel cell
systems, the SOFC possesses other advantages
in that it is a two phase gas/solid system. This
overcomes many of the problems associated
with liquid electrolytes such as corrosion,
flooding, electrolyte distribution, and the
maintenance of stable triple-phase-boundary
(TPB) electrode/electrolyte regions. The high
efficiency of small SOFC units could make
them very useful for distributed systems operating on bio-fuel as the land area required per
kWh would be significantly reduced.
3.1.2. Solid oxide fuel cells
A useful introduction to SOFC technology
was published in 1995 by Minh and Takhashi
[11], and more recent developments suggest
that two approaches are being followed to commercialise SOFC systems. One strategy envisages relatively large SOFC systems (> 1 MW)
for central power generators combined with gas
turbines to provide overall system efficiencies
approaching 70 %. For comparison existing
combined cycle gas turbines (CCGT) units are
capable of generating electricity with full load
efficiencies in the 50-52 %. Exhaust gases from
the SOFC stack must enter the gas turbine at
temperatures in excess of 850 °C and so the
stack must be operated at relatively high tern-
peratures. The most advanced high temperature SOFC stack is the 100 kW Westinghouse
tubular system recently commissioned (Spring
1998) in the Netherlands. The tubular configuration has many advantages including the
absence of high temperature seals, and an effective design which allows strips of La(Ca)CrO 3
to be u~ed as interconnect material. The disadvantages include a relatively expensive manufacturing route, relatively low power densities
with each 1.5 m tube only producing 210 W,
and an assembly that is not very tolerant of
rapid temperature changes.
In at:erupt to overcome some of the disadvantages of the tubular configuration a number
of companies (e.g. Siemens, MHI) are attempting to develop planar SOFC stacks. Although
the individual cell components (figure 5),
(+) LSM / YSZ / Ni-YXZ (-),
are similar, the planar configuration introduces
sealing problems as well as involving difficult
decisions about the choice of material (bi-polar
plate) for connecting the individual cells
together. Components based on doped ceramic
LaCrO~ have not been satisfactory under real
operating conditions, and so special alloys (e.g.
Y20~ dispersion strengthened Cr0.95Fe0.5) have
been developed. Initial problems due to Cr
deposition at electrode/electrolyte interfaces
arising from the formation of gaseous species
such as CrO 3 and CrO(OH) have been overcome by the adoption of coating strategies.
However these special alloys are expensive, and
require costly fabrication technologies, and so
the commercial viability of these high temperature planar stacks involving relatively thin
(-100 btm) YSZ electrolyte plates is by no
means guaranteed.
Applications of SOFC stacks that do not
require integration with gas turbines can be
designed to operate at lower temperatures
(600-700°C) with stainless steel bi-polar
plates. Inspection of figure2 indicates that
thick fi]m (~0.15 t-tm) YSZ electrolytes satisfy a
target ASR value of0.15 ~ cm 2 at 700 °C, and
if the yttria is replaced by scandia the temperature could be reduced to around 670 °C which
is still adequate to sustain the methane steam
reforming reaction. However the CTE (10.5 ×
10-(' K l) of zirconia electrolytes is too low to
provide a good thermal expansion match with
ferritic stainless steel (CTE: ~12-13 × 10-6
K-l), and so rapid changes in temperature
could be difficult to tolerate. A more significant
539
B. Steele
end plate
anode:
Ni-ZrO 2 cermet
solid etectrotyte:
YzO~ stabilised ZrO z
cathode:
LeHnO 3, InzO 3- SnO z
bipolar separator prate:
oxide, Ni, Cr-alloy,
SiC, Si|N 4
anode
Figure 5. Planar SOFC configuration.
Figure 5. Configurationplanairedc SOFC (pile ~lcombustible"t membranesolide "abase d'oxyde).
problem is associated with the performance of
the cathode material (La0.TSr0.3MnO~+x),
which is invariably used for zirconia electrolytes. This manganite perovskite has low values
for oxygen diffusion and surface exchange at
intermediate temperatures which results in
excessive cathodic polarisation losses. It is not
possible to replace LSM by a more effective
mixed conducting cobaltite composition,
partly because of thermal expansion mismatch,
but principally because cobaltites react with
doped zirconia electrolytes to form resistive
phases such as La2Zr20 7 and SrZrOs. At
present, therefore, it does not appear likely that
YSZ electrolytes can be specified for high power
density stacks designed to operate in the temperature range 650-700 °C, although operation at somewhat higher temperatures
(-750 °C) appears feasible if appropriate cheap
metallic alloy bi-plates can be developed.
The electrolytes 8i2V0.gCu0.1Os.35, Ce0.9
Gd0.101.95 (CGO), La0.gSr0.1Ga0.sMg0.2Oe.ss,
have higher ionic conductivities than YSZ, and
thus appear to be more obvious contenders for
operation at intermediate temperature. However, the bismuth oxide electrolytes are not stable in the reducing environments established at
the anode, and La0.gSr01Ga0.sMg0.802.85 is not
stable in contact with the standard anode material Ni-YSZ, as well as being expensive.
540
Accordingly most investigations have involved
ceJia based electrolytes such as CGO. In contact with reducing fuel gases, a fraction of the
Ce 4+ ions can be reduced to Ce 3~. The associated partial electronic conductivity produces an
internal short circuit which reduces the overall
system efficiency. Whilst this electronic short
circuit is significant under open circuit (low
lo~Ld) conditions (> 0.9 V), its impact is small
under typical operating conditions (~0.7 V),
and SOFC stacks incorporating CGO electrolytes have already been successfully operated at
650 °C with system efficiencies around 45 %.
Another advantage of CGO is that there is
excellent thermal expansion match with ferritic
stainless steels used as bi-polar plates in these
intermediate temperature stacks. Moreover, the
cathode, La0.c,Sr0.4Co0.sFe0.203_,~ (LSCF) is also
stable in contact with CGO, and exhibits relatively small ASR values (0.1-0.3 ~ cm 2) in the
temperature range 650-700 °C. Investigations
are also underway to determine whether thin
layers of YSZ on the anode side of CGO electrolyte to reduce electronic conductivity are justified economically by improved stack
efficiency. The excellent cathode performance
of LSCF arise due the fact that it is a mixed conducting oxide (MIEC) which provides multiple
pathways for the oxygen ions to migrate to the
electrode/electrolyte interface. In fact the actual
Ceramic ion conducting membranes
electrochemical charge-transfer across this
interface is usually facile which has led Adler et
al. [12] to suggest a novel interpretation for the
relevant cathode behaviour. This paradigm has
been very successful in providing a quantitative
explanation of oxide cathode impedances in
terms of the associated transport properties (D*
and k values) which can be determined by independent measurements.
Examination of figure2, indicates that it
should be possible to use C G O at 500 °C provided appropriate cathodic performance can be
attained, and experiments are in progress with
alternative cathode materials to try to obtain
satisfactory power densities at 500 °C. Whilst
lower temperature SOFC operation is probably
not advantageous for methane steam reformers
it could be appropriate for methane partial oxidation reformers, and fbr use with methanol
fuel. The principle market for 500 °C operation would involve small compact stacks suitable for a variety of mobile applications
including caravans, yachts, and eventually electric vehicles. The ability of SOFC systems to
utilise CO as a fuel means that direct methanol
SOFC should be possible, and even for operation on LPG, gasoline, and diesel, the associated fuel processing unit can be relatively
simple compared to units required for the low
temperature polymer electrolyte fuel cell systems which require relatively pure H 2 (< 10
ppm CO) for their successful operation.
3.2. Oxygen generators
Two approaches are being investigated for
oxygen generation. One concept uses a voltage
driven system incorporating a ceramic electrolyte which functions as an electrochemical
pump transferring oxygen from air to a pure
oxygen anode compartment as indicated schematically below:
0 2, N2, (P02/), (-) 0 2-.2.- (+), 0 2, (P02/t)
The required applied voltage (E) is given by:
E= E o + IR + 1/1 + 1/2
where E 0 is the Nernst Voltage, i.e. E 0
= RT/4F ln(Po2///P02/), IR represents ohmic
losses, 1/1 is the activation polarisation losses at
both electrodes, and 1/2 represents concentration polarisation losses at both electrodes. One
advantage of the electrochemical production of
oxygen is that pure oxygen can be generated at
higher pressures than the air supplied to the
cathode. The performance of a multistack
YSZ-based generator has been described, but
CeO 2 cr Bi20 3 electrolyte based units operating at lower temperatures are expected to produce o>:ygen more cheaply. Applied voltages of
0.5 V can generate currents around 0.5 A cm -2.
This ccrresponds to 1.67 kW / kg 0 2 (Appendix, box B) which should be suitable for the
small scale generation of oxygen required for
medical and aerospace applications.
An alternative approach is to use a mixed
conductor and to provide the driving force for
oxygen transport by imposing a differential
oxygen partial gradient across the dense C I C M
membrane. The whole system can be operated
in excess of atmospheric pressure, but obviously
the mechanical integrity of such a system is of
major (oncern for thin membranes and appropriate 3orous supports have to be specified.
Although the use of mixed conductors for oxygen separation is conceptually simple many
scale-up problems remain to be solved economically, and it probable that small electrically
driven ~ystems will be commercialised first.
3.3. C~ramic p a r t i a l o x i d a t i o n reactors
Man), years ago it was recognised that useful
chemicals could be produced using oxygen ion
conducting membranes/electrolytes, and an
early investigation is provided by the oxidation
of NH3 to N O [13]. However, by the mid
1980's attention had focussed on other partial
oxidation reactions involving methane, and in
1995 very encouraging results were reported by
Balachandran et al. [14] for the production of
syngas IH 2 + CO) by partial oxidation of methane usi:lg oxygen fluxes through ceramic tubes
fabricated from mixed conducting oxides.
Excellent methane conversion levels and CO
and H , selectivity were obtained, and world
wide consortia co-ordinated by Air Products
and Praxair, respectively, have been formed to
exploit this technology. At present the production of syngas from methane is mostly accomplished by steam reforming which is a very
energy- and capital-intensive process as the reaction is highly endothermic. Dense oxygen ion
conducting membranes offer potential solutions
to several problems in methane conversion. The
associated reduction in plant size is also attractive to the oil industry as it should be possible to
convert surplus methane on oil rigs to syngas
and subsequently to liquid fuels such as meth541
B. Steele
anol, or hydrocarbons via Fischer-Tropsch reactions. There are vast reserves of natural gas in
remote gas fields, such as Alaska's North Slope,
and in off-shore locations, that are uneconomic
to exploit at present. Successful development of
the CICM partial oxidation route from natural
gas to liquid fuel could effectively boost the
world's oil reserves by an equivalent of at least
30 years consumption [15].
block
logical applications. Many opportunities exist
for further basic studies on the transport propcrties of existing and novel oxide materials in
association with other investigations seeking
further improvements in terms of performance,
chemical stability, mechanical reliability, system simplification , and cost. Nevertheless,
there is already sufficient knowledge established
to be confident of the future commercial success of CICM materials.
Appendix
Box A
Fe20~
block
#
= o/R
(~)
j = I/zFA=rl/zF.A.R
L
21¢L
2D*]
tT,. +-a-f;
P
DvflnP;'{2
CvdlnPo2
Figure 6. Crystal structure of Sr4Fe6 xCoxOl3.
Figure 6. Structure cristalline de la phase Sr4Fe6_xCo.O13.
(T)
j = 4LOlnPf, e
Box B
For syngas production the ceramic membrane has to operate under severe chemical
potential gradients with air on one side and
natural gas on the other. The initial work of
Balachandran using mixed conducting perovskite oxides in the system L a - S r - C o - F e - O
was unsuccessful due to decomposition of this
material when exposed to the reducing environment containing methane. An alternative
material based on Co doped Sr4Fe6OI3 has
been developed which exhibits a high partial
oxygen ion conductivity and kinetic stability, at
least, in methane. It appears that Co-doped
Sr4Fe6OI3 is an orthorhombic intergrowth
structure [16] incorporating perovskite layers
alternating with Fe202. 5 blocks (figure 6). The
successful development of this material has
stimulated R & D activities into alternative
materials including dual phase composite
mixed conducting materials.
4. C o n d u s i o n s
Clearly oxygen ion conducting membranes
have an exciting future in a variety of techno542
Fluxes
1 a c m -2 =2.6 pmol O2.cm-2.s -1
= 1.56 x 1018molecules 02 cm-2.s -I
= 8.32 × 10-8 kg.O2.s q
3.5 cm 3 0 2 cm-2.min -1
Fud Cell
Operating.parameters: 0.7 V at 0.3 Acm 2
(-0.2 Wcm-Q, efficiency: 50 %, fuel conversion 85 %.
Active cell area: 5000 cm 2 per kW, -1000 L
of H 2 required per kWh.
Oxygen generator
Operating parameter: 0.5 Acm -2 at 0.5 V
(0.;25 Wcm-2).
Active cell area: 571 cm 2 per L.min -1, i.e.
10 L (dm 3) requires 5710 cm .
Also 1.67 kWh required to generate 1 kg O 2,
i.e. 1670 kWh per tonne O 2.
At $0.06 per k w h , 02 costs ~$100 tonne q.
Ceramic ion conducting membranes
References
[1] Bhave R.R. (Ed.), Inorganic Membranes, Synthesis,
Characterisation, and Applications, Van Nostrand Reinhold, New York, 1991.
[2] Burggraaf A.J., Charpin J., Cot L. (Eds.), Inorganic
Membranes, Trans Tech Publications, Switzerland,
1991.
[3] Bouwmeester H.J.M., Burggraaf A.J., Dense ceramic
membranes for oxygen separation, in: Gellings P.J.,
Bouwmeester H.J.M. (Eds.), Solid State Electrochemistry, CRC Press, New York, 1997, pp. 481-553.
[4] Steele B.C.H., Dense ceramic ion conducting membranes, in: "Fuller H.L. et al. (Eds.), Oxygen Ion and
Mixed Conductors and Their Technological Applications, Kluwer Academic Publishers, Dordrecht, 1998
(to be published).
[5] Norby T., Larring Y., Curr. Opin. Solid State Mater. Sci.
2 (1997) 593-99.
[6] Yentekakis V., Moggridge, G.,Vayenas C.G., Lambert
R.M., J.Catal. 146 (1994) 293.
[7] Manning ES., Sirman J.D., Kilner J.A., Solid State Ionics 9~ (1997) 125-132.
[8] Grov: W.R., Phil. Mag. Ser. 3, 14 (1839) 127.
[9] Kordesch K., Sinrader G., Fuel Cells and Their Applications, VCH, Germany, 1996.
[10] Proc. 5~h Grove Fuel Cell Symposium, J. Power Sources
71 (]998) 1-372.
[l l] Mink N.Q., Takahashi T., Science and Technology of
Solid Oxide Fuel Cells, Elsevier, Amsterdam, 1995.
[12] Adler S.B., Lane J.A., Steele B.C.H, J. Electrochem. Soc.
143 ',1996) 3554 3564.
{13] Farr R.D., Vayenas C.G., J. Electrochem. Soc. 127
(198)) 1478.
[14] F;alachandran U., Dusek J.'I:, Mieville R.L., Poeppel
R.B., Kleefisch M.S., Pei S., Kobylinski T.P., Udovich
C.A. Bose A.C., Appl. Catal. A Gen. 133 (1995)
19 29.
[15] (]as io oil: a gusher for the millem~ium in: Business
Week 19 (May 1997) 57.
[16] Gugilla S., Manthiram A., J. Electrochem. Soc. 144
(1997 ) L120-122.
543