Journal of Membrane Science 288 (2007) 175–187

Air separation using a catalytically modified mixed

conducting ceramic hollow fibre membrane module
Alan Thursfield, Ian S. Metcalfe ∗
School of Chemical Engineering and Advanced Materials, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, United Kingdom
Received 7 September 2006; received in revised form 8 November 2006; accepted 10 November 2006
Available online 21 November 2006

Abstract
A membrane reactor module consisting of four gas-tight hollow fibre membranes made of the mixed ionic-electronic conducting perovskite
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ has been tested for oxygen permeation in the temperature range of 650–1000 ◦ C with air on one side and helium on the
other. The module was operated under gradientless gas-phase conditions and in the absence of mass transfer limitations at lower temperatures. Also
at lower temperature the rate-determining step for oxygen permeation was found to be oxygen reduction with the oxygen permeation performance
of the module being significantly enhanced by deposition of a porous platinum catalyst film onto the air-side of the membranes. A stable but mass
transfer limited flux of 0.56 ␮mol cm−2 s−1 (0.73 ml STP cm−2 min−1 ) was achieved after a period of 4 days of operation at 1000 ◦ C and continued
at this rate for a further 4 days. A non-mass transfer limited flux of ca 0.14 ␮mol cm−2 s−1 was obtained at 800 ◦ C. The membrane module was
operated for 500 h before the onset of significant leakage. Post operation qualitative analysis indicated localised changes in the stoichiometry of
the membranes with no discernable trend, quantitative analysis on the other hand revealed that the perovskite had retained its starting composition.
Post-operation analysis of the membranes revealed the presence of sulfur in the form of sulfates of strontium, cobalt and calcium and a change
of surface microstructure for both the air- and lumen-side membrane surfaces. Possible sources of this sulfur contamination are: the glass-ceramic
sealing material employed, the helium sweep gas and atmospheric sulfur dioxide. This contamination may contribute to the decrease in the long-
term oxygen permeation rates. The metal cation stoichiometry of the perovskite showed changes in localised areas of the membranes following
oxygen permeation studies.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Perovskite; La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ ; Hollow fibre; Porous platinum catalyst film; Oxygen permeation

1. Introduction appropriate gas atmosphere compositions. Thus the material

The rich magnetic, optical, redox and catalytic properties of
metal oxides can be utilised for many applications [1]. Amongst
the most studied family of metal oxides is the highly versa-
tile cubic perovskite phase [2–4]. These ABO3 solid oxides
are formed from combining lanthanide or group II element
cations (A-site cation) and typically, thought not exclusively,
first row transition metal cations (B-site cation) into the cubic
close packed structure. They possess rich defect chemistry
created by doping of suitable cations into the structure and evi-
denced by the creation of oxygen vacancies within the oxygen
sublattice, VO•• , (represented in Kröger–Vink notation) and elec-
tronic n- and p-type conductivity at high temperature under
∗ Corresponding author. Tel.: +44 191 222 5279; fax: +44 191 222 5292.
E-mail address: I.Metcalfe@newcastle.ac.uk (I.S. Metcalfe).
is imparted with mixed ionic-electronic conducting (MIEC)
properties.
When formed into dense membranes and subjected to high
temperatures and oxygen chemical potential differences, these
membranes can transport pure oxygen through the combination
of mobile oxygen vacancies and electronic defects while simul-
taneously excluding the transport of other gas-phase species.
Interest has grown in applying MIEC membranes to industrial
processes such as air separation, oxidative coupling of methane
and synthesis gas production in catalytic membrane reactors
(CMR) as they offer product separation and oxygen supply in
one unit [5]. Commercial advantage can also be gained if freely
available atmospheric oxygen is used. Their inherent electronic
conductivity removes the necessity of external circuitry to pump
the oxygen from one side to the other, simplifying the reactor
and also removing electrode-membrane contact resistances, an
inherent problem with pure ionic conductors.

0376-7388/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2006.11.015
176 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
The mechanism of oxygen transport through MIECs involves
oxygen disassociation and reduction on the membrane surface
followed by incorporation into the perovskite lattice as oxygen
anions, OxO (in Kröger–Vink notation), that can migrate through
the crystal lattice via oxygen vacancy defect sites. This pro-
cess is activated and can be hindered by slow surface kinetics
at the membrane surface and long solid-state diffusional path
length i.e., large membrane thicknesses. The impact of these
two factors can be reduced by employing membranes of high
surface area and small wall thickness. This can be achieved with
membranes in the form of hollow fibres. The hollow fibres pro-
vide a high membrane surface area per unit volume of reactor
compared to the conventional disc and tubular forms typically
used. The hollow fibres have diameters typically ca 1–2 mm
with wall thicknesses in the range 0.2–0.3 mm; disk and tube
membranes typically have thicknesses of ca 1–2 mm. Interest in
hollow fibres is growing, and a number of groups have begun
work in fabricating and characterising the structure and oxygen
permeation performance of a number of perovskite based hol-
low fibre membranes [6–17]. One of the most studied perovskite
membrane materials is La0.6 Sr0.4 Co0.2 Fe0.8 O33 − δ (abbreviated
to LSCF6428) because of its stability. Stability of the crystal
structure is crucial if commercial exploitation is to be realised
as degradation over time caused by kinetic demixing of the con-
stituent cations under a thermodynamic gradient may lead to
membrane failure [18–20]. Catalytic modification of the mem-
brane surface can increase the rate of surface processes and
therefore the oxygen permeation performance when such pro-
cesses are limiting. This approach has been tried, and proved
successful, using photolithography to deposit regularly spaced
platinum disks onto a disk membrane of SrCo0.5 FeOx MIEC
[21]. The oxygen permeation achieved was up to double that of
an unmodified test sample of the same material in the tempera-
ture range 600–800 ◦ C.
In our previous work a hollow fibre membrane reactor of the
shell and tube type (with the tube replaced with a bundle of
hollow fibres) was tested for oxygen permeation and methane
oxidation using catalytically unmodified membranes [22]. The
reactor was operated under conditions under which there was
an axial oxygen chemical potential gradient on both sides of the
membrane giving rise to a gradient in oxygen permeation rate.
In contrast to this, a gradientless reactor would allow oxygen
permeation rates to be accurately determined as a function of
operating conditions.
In this contribution we report on the performance of a catalyt-
ically unmodified and catalytically modified MIEC LSCF6428
gradient-less hollow fibre reactor module with respect to oxygen
permeation over the temperature range of 650–1000 ◦ C and also
present information on post-operation characterisation.
2. Experimental
2.1. Characterisation
Commercial LSCF6428 (Praxair) was used to produce the
dense hollow fibre membranes by the phase inversion tech-
nique. This method and the membrane gas-tightness testing
apparatus are described elsewhere [13,14]. Crystallinity and
phase purity was ascertained by powder X-ray diffraction (XRD)
using a PANalytical X’Pert Pro diffractometer fitted with a
X’Celerator and employing Cu K␣ radiation, λ = 1.54180, or
Cu K␣1, λ = 1.54059. A nominal step size of 0.033 degrees two-
theta with a step time of 100 s in continuous mode was routinely
used. Diffractograms were recorded on ground membrane sam-
ples and whole hollow fibre samples. A Röntec Quantax 1.2
FEI 30XL ESEM FEG system was used to perform scanning
electron microscopy (SEM) for microstructure characterisation
and for standard-less qualitative elemental analysis by energy
dispersive X-ray analysis (EDS). Elemental mapping was per-
formed after EDS to ascertain possible segregation of elements.
Quantitative analysis of the bulk composition of the used hol-
low fibre membranes was performed by Inductively Coupled
Plasma Optical Emission Spectroscopy (ICP-OES) using a Uni-
cam 701 ICP-OES. Calibration for the elements of interest over
the range 0–100 ppm was done using certified single element
standards. The sample was introduced at a rate of 1 ml min−1
and the plasma was operated at a power of 1 kW with a coolant
flow of 13 L min−1 and an auxiliary flow of 0.5 L min−1 .
2.2. Catalytic surface modification
A module was made comprising of four hollow fibres whose
outer air-side surfaces were coated with a layer of platinum. A
platinum resinate (M603B supplied by Metalor, UK) containing
14 wt% platinum was crudely applied onto the outer surface of
the central 5 cm of each hollow fibre using a small fine brush. The
membranes were then fired using a programmed temperature
ramp of 10 ◦ C min−1 to 400 ◦ C, held for 1 h, then 10 ◦ C min−1 to
900 ◦ C, held for 1 h, and then 10 ◦ C min−1 to 25 ◦ C to remove the
organic resin and sinter the platinum metal onto the membrane
surface. The average surface loading of the catalytically modi-
fied hollow fibres was ca 3 mg cm−2 . Hollow fibre membranes
coated with platinum will be referred to hereafter as Pt-HF and
uncoated as HF. A diagram of the hollow fibre reactor module
is given in Fig. 1 with an expanded schematic of a single hollow
fibre indicating the central 5 cm exposed to the atmosphere.
2.3. Oxygen permeation experiments
The membrane performance apparatus employed consisted
of gas supply and gas flow control, the hollow fibre reactor
module with furnace and online gas chromatography for anal-
ysis of the gas stream. All experiments were performed at a
total pressure of one atmosphere on both sides of the mem-
brane. The gas supply to the lumen-side of the reactor module
was controlled by an Aalborg mass flow controller calibrated
for 100% helium and ranged from 0 to 200 ml min−1 . All gas
flows are quoted at standard temperature and pressure (STP).
The upstream and downstream gas flow rates through the lumen
were independently monitored using a digital flow meter (Var-
ian). The apparatus has been described elsewhere [22] and will
not be repeated here, the only difference being the reactor mod-
ules used in this work have no shell. A heating or cooling rate of
1 ◦ C min−1 was used when required throughout this investiga-
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
Fig. 1. Schematic of the shell-less hollow fibre membrane and magnified diagram of the platinum coated hollow fibre membrane area.
tion for the membrane modules. The temperature profile of this
furnace under operating conditions has been reported previously
[22]. In order to report fluxes and avoid the complication of oxy-
gen flux as a function of position along the membrane, a coating
of glass ceramic sealant; 70% inorganic filler, composed mostly
of barium sulfate, and also kaolin and silica flour, and the remain-
ing 30% is a binder, composed mostly of sodium and potassium
silicates was applied to the outside of the hollow fibres. This
application was designed to pacify the oxygen permeation con-
tribution from the membrane outside of the central linear heated
zone of 5 cm. The superficial active membrane surface area SA
thus provided by the module can be approximated by
2π(ro − ri )
SA = nl
ln(ro /ri )
where ro and ri are the outer and inner radius of a typical
hollow fibre, n is the number of hollow fibres and l is the
active working length. The radii ro and ri were typically 0.07
and 0.05 cm, respectively, although this is an approximation
because of irregular wall thicknesses. Assuming that only the
central linear heating zone of the furnace is used (l = 5 cm) gives
an active membrane area of ca 7.6 cm2 provided by the four
hollow fibres. This central membrane area was then coated with
a porous film of platinum for experiments with the catalytically
modified membrane.
One HF and one Pt-HF module were used for all of the exper-
iments described in this work. Throughout this work the side of
the membrane which is exposed to the constant oxygen partial
pressure of the atmosphere is referred to as the air-side; the par-
tial pressure of oxygen on this side of the membrane being pO2  .
The other side of the membrane, the lumen-side, is referred to
as the sweep-side. About 99.999% helium (BOC, UK) (possi-
ble contaminants are N2 at 5 ppm, O2 at 2 ppm, H2 at 1 ppm,
CO2 at 0.5 ppm, H2 O at 1 ppm and C1–C5 hydrocarbons at a
total of 0.5 ppm) was used as the inlet sweep gas and therefore
the inlet oxygen partial pressure at the lumen-side (denoted by
pO2  in ) was zero. The outlet oxygen partial pressure from the
lumen-side is denoted by pO2  out .
Permeation is determined by measuring oxygen evolution
through the hollow fibres on the lumen-side (as opposed to con-
177
sumption on the air-side) and is adjusted so as to not include
any contribution from leakage. The oxygen flux, jO2 , was deter-
mined by the percentage molar content of oxygen appearing in
the helium sweep gas, [O2 ], measured by gas chromatography
and using Eq. (2):
1 1 1
jO2 = [O2 ]F (2)
Vm 60 SA
where F is the flow rate of the helium sweep gas in ml (STP)
min−1 , Vm is the molar volume of the sweep gas (1 mol is
equivalent to 22400 ml at STP), the 1/60 factor converts the
rate from minutes to seconds and the final term SA is 7.6 cm2
to give the flux (area specific, based on the simple projected
(1) area of the active section of the hollow fibre membrane).
Oxygen permeation data were recorded 15 min after chang-
ing the gas flow conditions and 30 min after each temperature
change.
2.4. Membrane history
It can be important to know the history of membranes during a
series of experiments as past treatment can influence behaviour.
The treatment history of the LSCF6428 membranes in the HF
and Pt-HF modules is as described in the sequential series of
experiments shown in Figs. 4–6. The information is summarised
in Table 1. Treatment 1 represents the beginning of the exper-
iments with the modules being heated to the first operating
temperature under air and treatment 50 indicates the cessation
of the study on the Pt-HF as the module is cooled down to room
temperature under air. Also included are the heating and cooling
conditions between experiments and the duration of these steps.
The total operation time of the HF module was ca 43 h and for
the Pt-HF module ca 500 h.
3. Results and discussion
3.1. Characterisation
Prior to use each LSCF6428 hollow fibre was tested for gas-
tightness and samples of the hollow fibres were ground for
178 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187

Table 1
Membrane treatment summary collated from Figs. 4–6
Treatment Relevant figure Experimental conditions

Heating/hold (◦ C) Cooling/hold (◦ C) Helium sweep rate(s) (ml min−1 )c Duration/hoursd

1 4 and 5a 25–650 0e 10.9

2 4 and 5 650 10–187 3.2
3 4 and 5 650–700 10 1.3
4 4 and 5 700 10–180 3.2
5 4 and 5 700–750 10 1.3
6 4 and 5 750 10–173 3.2
7 4 and 5 750–800 10 1.3
8 4 and 5 800 10–170 2.8
9 4 and 5 800–850 10 1.3
10 4 and 5 850 9–123 2.4
11 4 and 5 850–900 9 1.3
12 4 and 5 900 10–137 2.4
13 4 and 5 900–950 10 1.3
14 4 and 5 950 9–134 2.4
15 4 and 5 950–1000 9 1.3
16 4 and 5 1000 9–182 3.2
17 5b 1000–950 9 1.3
18 5 950 13–135 2.4
19 5 950–900 13 1.3
20 5 900 10–138 2.4
21 5 900–850 10 1.3
22 5 850 9–125 2.4
23 5 850–800 9 1.3
24 5 800 11–170 2.8
25 5 800–750 11 1.3
26 5 750 11–173 3.2
27 5 750–700 11 1.3
28 5 700 10–172 3.2
29 5 700–650 10 1.3
30 5 650 8–171 3.2
31 5 650–700 8 1.3
32 5 700 11–172 3.2
33 5 700–750 11 1.3
34 5 750 11–168 3.2
35 5 750–800 11 1.3
36 5 800 11–164 2.8
37 5 800–850 11 1.3
38 5 850 11–125 2.4
39 5 850–900 11 1.3
40 5 900 10–136 2.4
41 5 900–950 10 1.3
42 5 950 9–137 2.4
43 5 950–1000 9 1.3
44 5 1000 10–161 2.4
45 6b 1000 10 24
46 6 1000–750 10 4.6
47 6 750 100 192
48 6 750–1000 10 4.6
49 6 1000 100 192
50 1000–25 0e 16.3
a For Pt-HF and HF modules.
b Pt-HF module only.
c Flow rate(s) during experiments and temperature changes.
d Includes wait times for temperature stabilisation, sweep gas flow stabilisation and analysis time.
e In air.

powder XRD, which revealed that the membranes employed in are provided in Fig. 2. EDS analysis of these surfaces did not
this work had retained the original perovskite structure (powder reveal the presence of foreign elements and elemental mapping
pattern not shown). Images of the inner and outer surfaces and (not shown) did not indicate segregation of any of the constituent
cross-sectional microstructure of the as-prepared hollow fibres perovskite metals.
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187 179

Fig. 2. Representative SEM images of (a) the outer surface of the hollow fibre
membrane, (b) the lumen-side surface and (c) fracture surface.

The grain size is no larger than 5 ␮m, the surface (Fig. 2a)
is clean and shows almost no pores or cavities. The lumen-side
(Fig. 2b) reveals the presence of cavities and an uneven surface
due to nodules of grains. A fracture surface image (Fig. 2c)
indicated a degree of pitting but no pores. We conclude that the
membranes were adequate for use in this investigation.
In Fig. 3 typical micrograph images of the outer surface of a
hollow fibre coated with platinum following sintering are shown.
The crudely applied coating is not continuous as there are areas
of bare membrane where the LSCF6428 crystal grains can be
seen (Fig. 3a), the coating itself has a dispersion of island parti- Fig. 3. Representative SEM images of the outer surface of a hollow fibre mem-
brane with a discontinuous porous coating of platinum catalyst showing (a) zones
cle sizes (Fig. 3b). The platinum islands have a porous surface
of bare membrane and zones of platinum coverage, (b) islands of platinum with
structure, some of which exhibit pits revealing the membrane a range of sizes, (c) fine porosity of a typical platinum island with cavity and (d)
surface (Fig. 3c). The film can be seen in profile (Fig. 3d) and the platinum layer in profile sintered onto the outer membrane wall.
has a fairly uniform thickness of ca 0.2 ␮m.
180 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
3.2. Oxygen permeation studies
To ensure a meaningful oxygen flux is determined we must
ensure that there are no significant axial oxygen partial pressure
gradients on the lumen-side. This can be confirmed by mon-
itoring the oxygen partial pressure difference between outlet
and inlet on the lumen-side and if this difference is typically
greater than 10% of the partial pressure driving force at the
inlet (i.e. 0.208 atm) then the lumen-side flow rate should be
increased. Here we choose a 10% difference in the driving force
as being significant as we do not at this stage know the exact
form of the dependence of flux on oxygen partial pressure but
this figure would be reasonable in most cases with the excep-
tion of highly non-linear behaviour. However, the ultimate test
that the difference in driving force is insignificant is that the
flux no longer depends on sweep gas flow rate as this flow rate
is increased. Furthermore, if external mass transfer resistances
Fig. 4. Comparison of the oxygen permeation performance of Pt-HF module represented by the () symbols with an HF module represented by the (夽) symbols
under the same conditions of temperature and helium sweep rate.
are unimportant, the overall mass transfer coefficient for oxygen
should be independent of the lumen-side flow rate. Indeed both
these conditions of gradientless operation and absence of mass
transfer resistances are satisfied when the flux is independent of
the lumen-side flow rate.
In Fig. 4 the oxygen permeation fluxes of the Pt-HF and HF
modules are compared under similar conditions. The hollow
fibre membranes in both modules were subject to treatments
1–16 inclusive as indicated in Table 1. At lower temperatures
there is a large range of sweep gas flow rates in which the flux
is independent of sweep gas flow rate. This indicates that the
membrane is operating under conditions where mass transfer is
unimportant and where the reactor is effectively gradientless.
At higher temperatures there is no region observed where flux
is independent of the sweep gas flow rate. Furthermore, at lower
temperatures the benefit of the catalyst is apparent, as up to
ca four times the oxygen permeation for a given set of condi-
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187 181

tions is obtained. This indicates that the reduction of oxygen on
the airside must be the rate-determining step in the absence of
a catalyst and that mass transfer cannot be controlling on the
air-side when no catalyst is present. The highest oxygen flux
obtained was 0.75 ␮mol cm−2 s−1 , equivalent to an electrical
current density of 300 mA cm−2 .
Following the first heating cycle presented in Fig. 4
(650–1000 ◦ C in 50 ◦ C steps, treatments 1–16 inclusive) a
cooling cycle (950–650 ◦ C in 50 ◦ C steps, treatments 17–30
inclusive) and a second heating cycle (700–1000 ◦ C in 50 ◦ C
steps, treatments 31–44 inclusive) were carried out to examine
the effect of rapid heating and cooling on the oxygen perme-
ation behaviour. Each cycle took ca 24 h to complete. The data
are presented in Fig. 5. The first heating cycle exhibits lower
fluxes to some degree at all temperatures except 1000 ◦ C. On
cooling from 1000 to 650 ◦ C the oxygen flux is higher at all
temperatures compared to the first heating cycle, although the
affect is less pronounced in the 850–1000 ◦ C range compared
to the 650–800 ◦ C range. The second heating cycle shows that
there is little difference to the oxygen fluxes obtained in the
preceding cooling cycle, although they are a little lower until
850 ◦ C is reached. On the first heating cycle from 650 to 1000 ◦ C
the oxygen vacancy concentration is being established but has
not reached its steady state under these conditions. Upon cool-
ing back to 650 ◦ C we observe higher oxygen fluxes than those
obtained during the first heating cycle due to the established
higher vacancy concentration, this new concentration is closer
to the steady state value for these operating conditions and upon
heating for the final time the increase in oxygen permeation is
therefore modest. The situation may be complicated by pos-
sible structural changes of the perovskite to phases such as
brownmillerite at low temperatures. This phase contains ordered
oxygen vacancies that present a barrier to solid state oxygen
diffusion.

Fig. 5. Oxygen permeation performance of a shell-less Pt-HF module over the temperature range 650–1000 ◦ C; () symbols represent the first heating cycle
incorporated here from Fig. 4, (䊋) symbols represent the cooling cycle and (䊉) symbols represent the second heating cycle.
182 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
Fig. 6. Temporal oxygen permeation of the Pt-HF module at 750 and 1000 ◦ C.
The (䊉) symbols represent the oxygen permeation at a constant helium sweep
of 100 ml min−1 at 750 ◦ C. In the top graph section the () symbols represent
the oxygen permeation at a constant helium sweep of 100 ml min−1 at 1000 ◦ C.
3.3. Long-term oxygen permeation studies on the Pt-HF
module
The oxygen permeation trend over longer periods of time than
those employed in the preceding data was investigated firstly at
750 ◦ C and then at 1000 ◦ C, the data are presented in Fig. 6. At
1000 ◦ C the module is operating under mass transfer limitations.
Prior to the investigation conducted at 750 ◦ C, the module was
held for 24 h at 1000 ◦ C (treatment 45) following on from the
last heating cycle of the experiment described in Fig. 5. The long
term study at 1000 ◦ C followed on from the 8 day period where
the module had been held for 8 days at 750 ◦ C.
At 750 ◦ C the oxygen flux is initially ca 0.06 ␮mol
cm−2 s−1 at a helium sweep rate of 100 ml min−1 . This value
is close to the oxygen permeation rate obtained during the
first heating cycle described above. Over time however the
flux decreases to ca 0.02 ␮mol cm−2 s−1 . A similar trend was
reported by Kruidhof et al. [23], who used a La0.6 Sr0.4 CoO3 − δ
membrane. They observed a large decrease in the oxygen per-
meation at 750 ◦ C over ca 400 h. This was explained in terms
of changes in the oxygen stoichiometry and ordering of oxy-
gen vacancies. At 1000 ◦ C the decrease is much less dramatic,
the permeation rate falling from ca 0.60 ␮mol cm−2 s−1 to ca
0.56 ␮mol cm−2 s−1 over the 8 day period. As well as oxygen
stoichiometry changes within the membranes, another factor is
the possible removal, deactivation or detachment of the plat-
inum catalyst layer over time. The removal or deactivation of
this layer would be expected to have a large impact on the oxy-
gen permeation at 750 ◦ C where surface processes dominate. In
contrast, at 1000 ◦ C, surface processes are less important and
the effect would be more modest. The decrease in the oxygen
flux with time has also been reported with tubular membranes of
LSCF6428 operated at 850 ◦ C [24,25] and more recently with
hollow fibre membranes of BaCox Fey Zrz O33 − δ also operated at
850 ◦ C [26]. The reason for this decrease was reported to be con-
tamination of the membrane with sulfur from the synthetic air
gas supply, in the form of sulfur dioxide forming both strontium
sulfate and cobalt sulfate [24,25]. This possibility is discussed
further in the post-operation studies section. Another exam-
ple of contamination affecting oxygen permeation has recently
been provided by Viitanen et al. [27] they observed a decrease
in the oxygen permeation rate of a disc LSCF6428 membrane
over time in a dual chamber cell. They found evidence for the
deposition of a contaminating silicon oxide layer formed under
operating conditions from siloxane-containing grease in the gas
valves apparatus. We cannot eliminate the possibility that a sim-
ilar contamination process has occurred in this study because of
the presence of significant quantities of glass-ceramic cement
in direct contact with the membranes which when subjected to
high temperatures may take part in solid-state reactions. This
possibility is discussed later.
During these long term experiments the atmospheric oxy-
gen leak into the helium sweep, possibly through membrane
degradation and/or aging of the glass-ceramic sealant, gradu-
ally increased at 1000 ◦ C from an initial value of 0.01 ␮mol s−1
(ca 0.02 ml min−1 ) to 0.06 ␮mol s−1 (ca 0.08 ml min−1 ) in the
100 ml min−1 of helium sweep gas, the experiment was termi-
nated at this point as it was deemed that an oxygen leak rate of
around 10% of the solid state oxygen permeate was too high.
3.4. Comparison with other geometries and materials
In order to put the performance of the LSCF6428 hollow fibre
membranes into context, we now compare the performance of
the HF and Pt-HF modules used in this work with that of previous
work using LSCF6428 in both disk and hollow fibre form and
where other perovskite materials have been used. The data are
summarised in Table 2.
Two studies have used LSCF6428 membranes with similar
thicknesses of 1.5 and 1.53 mm, respectively, [28,29]. Oxygen
permeation rates with air on one side and an inert sweep on the
other at 900 ◦ C were reported to be 0.1 and 0.08 ␮mol cm−2 s−1 ,
respectively. Under the same conditions, the hollow fibres used
in this work show a flux of 0.2 ␮mol cm−2 s−1 . The thickness
of the hollow fibres used here is 0.25 mm and, as such, if solid
state diffusion were rate determining we would expect to see a
flux some six times greater than the fluxes for the thicker mem-
branes. The fact that this is not the case may suggest (one should
be careful about drawing any conclusions from comparisons of
fluxes from different studies) some element of surface rate con-
trol. We note that there are differences between studies in which
hollow fibre LSCF6428 membranes have been used and that a
higher oxygen flux has been obtained with unmodified hollow
fibre membranes at a lower temperature (900 ◦ C) [14] than the
maximum used (1000 ◦ C) in this contribution. The differences
between studies may lie in the sample treatment history and the
use of a shell. A shell can protect the outer surface of the mem-
brane from contaminants that may be present in the atmosphere
or that may evolve from the furnace tube and deposit onto the
membrane and degrade performance over time.
The LSCF6428 membrane material is only one of an ever
increasing number of perovskites being investigated for oxygen
permeation and methane oxidation reactions. In recent years
Ba0.5 Sr0.5 Co0.8 Fe0.2 O33 − δ (BSCF5582) [8] and its zirconia
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187 183

Table 2
Comparison of oxygen permeation performance of hollow fibre membranes of LSCF6428 with other MIEC perovskite materials
Hollow fibre membrane material Thickness (mm) Temperature (◦ C) Flux jO2 a Atmospheres used Ref.

ml cm−2 min−1 ␮mol cm−2 s−1 Low pO2  /high pO2 

La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ 0.2 700–900 0.1–0.8 0.07–0.55 Ar/air [13]
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ 0.3 800–900 0.1–1.2 0.07–0.9 Ar/air [14]
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ 0.2 543–958 0.09–0.42 0.07–0.3 He/air [22]
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ 0.2 858 0.26–0.8 0.2–0.6 0.022 atm CH4 /He/air [22]
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ (HF module) 0.2 650–1000 0.007–0.65b 0.005–0.5b He/laboratory air This work
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ (Pt-HF module) 0.2 650–1000 0.02–1.0b 0.015–0.75b He/laboratory air This work
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ (Pt-HF module) 0.2 1000 0.73b,c 0.56b,c He/laboratory air This work
La0.6 Sr0.4 Co0.2 Fe0.8 O3 − δ (Pt-HF module) 0.2 750 0.03d 0.02d He/laboratory air This work
BaZrx Coy Fez O3 − δ e 0.18 800–900 5–7.6 2.98–5.65 0.04–0.06 O2 atm/He/air 26
Ba0.5 Sr0.5 Co0.8 Fe0.2 O3 − δ 0.2 700–950 0.69–5 0.51–3.72 He/air (no shell) 31
a STP assumed if not stated.
b Fluxes include mass and non-mass transfer limited values.
c Steady value after 4 days of operation.
d Value after 8 days of operation.
e Stoichiometry not given.

containing analogue BaZrx Coy Fez O33 − δ [26] (the stoichiom- peaks can be observed in the 2θ range 23–52◦ . These low inten-
etry of which has not been provided for proprietary reasons) sity peaks become noticeably stronger relative to the perovskite
have been developed and reported to possess impressive long phase when analysis of the intact membrane air-side surface is
term stability during oxygen permeation and hydrocarbon oxida- performed (Fig. 7c). We assign these peaks to strontium sulfate
tion studies. These materials have provided the highest reported as an excellent match with PDF 05–0593 was made. Additional
fluxes in the open literature to date, ca 5 and 7.6 times higher information from EDS described in the following discussion
than the maximum flux obtained from the Pt-HF membranes helped to confirm the presence of this phase. Platinum could
used here. not be detected in the Pt-HF membrane samples by XRD as its
reflections are coincident with that of the perovskite phase. The
3.5. Post operation characterisation intact membrane surface on the lumen-side on the other hand
shows only the perovskite phase (Fig. 7d).
After operation, each module was cooled down at 1 ◦ C min−1 The external membrane surface (air exposed side), membrane
to room temperature, removed from the furnace and visually wall cross section (the bulk) and the lumen-side membrane sur-
inspected. The HF module had a number of dark blue patches at face (helium sweep gas exposed side) microstructures of the
the hollow fibre-ceramic cement boundary. The Pt-HF module membranes were investigated by SEM and qualitative elemental
was inspected and it was observed that: (a) the metallic lustre analysis performed by EDS with elemental mapping.
of the platinum coating on the central 5 cm zone of the mem-
branes was no longer present, (b) approximately 2–3 cm either 3.5.1. HF module
side of this zone the pacifying cement coating was blackened, The air-side surface of the HF module was found to be pit-
and (c) there were patches of dark blue colouration with a shiny ted and covered in plates, an image of a representative example
smooth appearance in areas where the cement was blackened is provided in Fig. 8a. The pits are ca 50 ␮m in diameter and
and in areas were the cement retained its original colour. We are decorated with plates. EDS with elemental mapping (not
conclude that over the course of the oxygen permeation experi- shown) revealed the presence of lanthanum, strontium, cobalt,
ments the ceramic-glass water based cement and the hollow fibre iron, calcium and sulfur on the air-side membrane surface, rep-
had reacted and that cobalt had leached out of the LSCF6428 and resentative data are provided in Table 3. Calcium was found
deposited onto the cement surface. The affect was more severe
for the Pt-HF module possibly as it had been used for a longer Table 3
period of time at high temperature. Samples of the central 5 cm Representative post operation EDS data of the surfaces of the LSCF6428 mem-
hollow fibre membranes from both modules were removed for branes from the HF module
SEM, EDS, ICP and powder XRD analysis. Location Approximate atom (%)
Fig. 7a and b show the XRD powder patterns for the HF and
Pt-HF samples, respectively. These diffraction patterns are the La Sr Co Fe Ca S
bulk response and not surface specific, no preferred orientation Shell-side pit interior 0.8 16.9 8.9 1.4 26.6 45.3
was assumed for the crystallites. The powder diffraction patterns Shell-side pit edge 16.8 11.6 10.6 29.8 16.2 15.1
are very similar. The main perovskite structure has remained Shell-side away from pit 29.7 13.9 10.3 39.1 0 7.0
Bulk 33.3 22.5 8.9 35.2 0 ∼0.1
intact but with splitting of the main reflections indicating a Lumen-side 34.5 4.9 12.9 47.5 0 0.2
lowering of symmetry. In addition, a number of less intense
184 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
Fig. 7. Post operation powder XRD patterns of the (a) HF and (b) Pt-HF mem-
branes. Inset is the expanded region indicating the presence of strontium sulfate.
Also shown is a representative (c) XRD pattern of the intact air-side membrane
surface of the Pt-HF module strongly indicating the presence of strontium sulfate
and (d) XRD pattern of the intact lumen-side membrane surface of the Pt-HF
module revealing the absence of sulfate containing phases.
Fig. 8. Representative post operation SEM images of the HF module: (a) the
outer surface of the hollow fibre membrane showing a typical pit, (b) the outer
surface of the hollow fibre membrane away from pitted areas, and (c) the lumen-
side.
to be concentrated only at the pits along with particularly high
concentrations of sulfur. The calcium may have its origins in the
municipal water used in the hollow fibre forming process and
also from the furnace alumina tube insert. The chemical analysis
data sheet of the furnace tube indicated the presence of 0.14%
calcium oxide and 2.88% potassium oxide, however, potassium
was not observed in the EDS spectra of the membrane. To a
lesser degree, cobalt and strontium were also detected at these
pits, probably also in the form of sulfates while lanthanum and
iron were almost completely absent. The localisation of cobalt
and calcium sulfate at these pits meant that these sulfates were
not revealed in the powder XRD diffractogram. Away from the
pits (Fig. 8b) the membrane surface has broken up into small
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187 185

crystallites to give a rough surface, EDS and elemental map-

ping indicated an even covering of sulfur and the constituent
LSCF6428 perovskite metals in a range of stoichiometries, but
no calcium. The air-side sulfur away from the pits is probably
in the form of strontium sulfate as indicated by XRD.
The lumen-side surface was found to be decorated with crys-
tals and small, shallow cavities (Fig. 8c). The observed ridges
are a product of the hollow fibre forming process and not as
a result of microstructural change due to the imposed oxygen
permeation or other experimental conditions.
EDS revealed the presence of sulfur on the lumen-side and
surprisingly, within the membrane bulk at low levels. The form
which this bulk sulfur takes is not clear as the sulfide anion is
too large to take up positions at oxygen sites within the per-
ovskite structure. The presence of sulfur within the membrane
bulk has recently been reported in a similar study to this using
BaCox Fey Zrz O3 − δ hollow fibre membranes [26]. A tentative
explanation is that sulfates have migrated through fissures, too
small to be observed by SEM, at the membrane surface and have
entered and diffused through the bulk, though there is no direct
evidence for this. In addition we cannot rule out the possibility
of contamination during fracturing of the sample for analysis.

3.5.2. Pt-HF module

The air-side surface of the Pt-HF module was found to be
more complicated. The surface at the central 5 cm zone appeared
to be decorated with crystals of a range of sizes (Fig. 9a) but did
not show pits as was observed for the HF membranes, possi-
bly due to the platinum layer blocking their development. The
surface is markedly different to the preoperational air-side sur-
face (Fig. 3a–c) where clearly defined regions of platinum were
observed. Before operation there are Pt islands with clear surface
porosity. Post-operation, there is no clear porosity, possibly due
to sintering during operation. The removal of the metallic lustre
from the Pt-HF membranes did not result in a loss of the observed
enhancement of the oxygen permeation. EDS analysis (sum-
marised in Table 4) of the air-side membrane surface indicated
the presence of mostly sulfur (up to ca 60 atom %) and small
amounts of localised platinum (ca 3–7 atom %) along with the
constituent perovskite metals of the membrane in a wide range
of molar ratios. However, quantitative ICP of a bulk membrane
sample indicated that the nominal LSCF6428 stoichiometry was
maintained. Close to the pacifying cement-(Pt-HF) interface an
interesting structure can be observed resembling a continuum

Table 4
Representative post operation EDS data of the surfaces of the LSCF6428 mem-
branes from the Pt-HF module
Location Approximate atom (%)
La Sr Co Fe S Pt Al Si Ba

Shell-side (e.g. 1) 2.2 28.2 0.8 1.4 60.3 7.1 0 0 0

Shell-side (e.g. 2) 3.2 31.5 1.3 4.7 56.0 3.3 0 0 0
Shell-sidea 5.6 33.8 4.0 6.9 43.2 0 2.9 2.1 0.8 Fig. 9. Post operation representative SEM images of the Pt-HF module: (a) the
Bulk 22.1 25.6 12.9 33.0 0 0 6.4 0 0 outer surface of the hollow fibre membrane at the center, (b) the outer surface of
Lumen-side 27.3 18.9 8.5 35.7 0 0 9.6 0 0 the hollow fibre near the cement, (c) magnified image of (b) and (d) lumen-side
surface showing uneven surface with large raised protrusions and cavities.
a See Fig. 9b.
186 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
of connected depressions (Fig. 9b). It appears on first inspection
that either a matrix has been deposited on the membrane surface,
or, that grain growth followed by preferential smooth erosion of
the grains to leave a prominent grain boundary has occurred. A
higher magnification reveals that this surface structure contains
fine cracks (Fig. 9c). Such a structure was not observed with the
HF module, possibly due to its shorter period of operation at
high temperature. EDS analysis in this area again gave a wide
range of stoichiometries of the perovskite constituent metals but
with consistently higher proportions of strontium and sulfur in
an approximate 1:1 ratio indicating the presence of strontium
sulfate. Also present at low concentrations were the component
elements of the heavy grade cement.
The lumen-side, at the central 5 cm zone and within the
areas where the membrane was covered by the cement, revealed
the presence of large protrusions, cavities and deposits of iso-
lated crystal grains (Fig. 9d). The internal microstructure had
deformed in comparison to the pre-operation state (Fig. 2) prob-
ably indicating pending failure. EDS indicated that lanthanum,
strontium, cobalt and iron were in their nominal pre-operation
stoichiometric ratios but also that aluminium was present at sim-
ilar atom % concentrations as cobalt. No sulfur was detected on
the lumen-side surface.
EDS analysis of the Pt-HF membrane bulk indicated a mix of
the perovskite metal stoichiometries with no discernable pattern
and significant amounts of evenly distributed aluminium and the
absence of sulfur. The aluminium probably originates from the
kaolin component of the cement. At 1000 ◦ C over a period of 8
days it is possible that the Al3+ cation could diffuse through the
perovskite by solid-state reaction. EDS of the HF module did
not indicate the presence of aluminium within the instrument’s
detection limits, which may be attributed to its shorter time of
operation at 1000 ◦ C. We note also the absence of barium in both
sets of membrane samples within the detection limits of EDS.
The membrane cross-sectional microstructure of the HF and
Pt-HF membranes had not changed in comparison to the pre-
operation state and so micrographs are not shown.
3.5.3. Sulfur contamination and kinetic demixing
The presence of sulfur has four possible sources: (a) from
trace amounts of polyethersulfone from the phase inversion
method used to form the hollow fibres into a dope, (b) from
the cement, (c) from trace amounts of sulfur dioxide in the air
(in urban areas at the 1 ppb level) and (d) from the helium sweep
gas supply. The first possible source seems unlikely as sinter-
ing of the green was performed at 1200 ◦ C, sufficient to burn
off all of the organics used in the forming process. It may be
possible however that during this burning-off period the sulfur
reacts with strontium in the presence of air to form the stron-
tium sulfate phase but XRD and EDS did not show the presence
of sulfur or sulfur containing phases on either the lumen- or
air-side prior to use. The cement contains significant quantities
of sulfur and it is most likely that the observed sulfur is from
this source. A number of previous post-operation oxygen per-
meation studies have reported the presence of sulfur containing
phases on perovskite MIEC membranes with one source being
atmospheric sulfur dioxide which would certainly contribute to
sulfate formation but we cannot separate out its contribution
from that from the cement. The helium sweep gas supply has
also been cited in other work as a source of sulfur contamination
in the form of sulfur dioxide [24,25] with different helium gas
cylinders possibly containing different levels of sulfur contam-
ination. This may be the reason why sulfur was detected on the
lumen-side surface of the HF module but not on the lumen-side
of the Pt-HF module following the oxygen permeation studies.
Against this argument is the lack of sulfur contaminants in the
supplier’s specification of purity. The formation of the strontium
sulfate phase on the membrane air-side surface over time may
play a role in the observed decrease of the oxygen flux observed
at 750 and 1000 ◦ C (Fig. 6).
In addition, other phases such as cobalt sulfate, cobalt oxide,
strontium oxide and lanthanum oxide have been found, specif-
ically on LSCF6428 membranes [24,25]. The authors also
presented EDS data that indicated cobalt and iron had segre-
gated to the air-side of the membrane [25] in accordance with
the theory of kinetic demixing [18–20]. These additional phases,
and cobalt and iron enrichment at the air-side of the membrane,
could not be found in this post-operation study within the detec-
tion limits of the instrument used, only strontium sulfate was
found to be present with a high degree of certainty. The EDS
elemental analysis did not reveal any particular trend in the
perovskite cation stoichiometry as a function of location in the
membrane bulk indicating that on the time scale of the exper-
iment, detectable segregation of cobalt or iron at the air-side
region of the membrane had not occurred. The localised differ-
ences in the perovskite cation stoichiometry may indicate the
onset of demixing after ca 21 days (ca 500 h) of operation, but
the module was not used for a long enough period at high tem-
perature to conclusively show evidence, for example, of cobalt
and iron segregation at the air-side of the membrane. A study by
Lein et al. [30] has shown that after around 30 days of operation
at 1150 ◦ C kinetic demixing of La–Sr–Co–Fe oxide membranes
can occur.
4. Conclusions
Catalytically modified and unmodified gas-tight hollow fibre
membranes of LSCF6428 have been used for air separation in
the absence of axial oxygen partial pressure gradients. A stable
mass transfer limited flux of 0.56 ␮mol cm−2 s−1 was obtained
after 4 days of operation at 1000 ◦ C, the flux remained at this
level for a further 4 days before the study was terminated. A
non-mass transfer limited flux of ca 0.14 ␮mol cm−2 s−1 was
obtained at 800 ◦ C. The deposition of platinum catalyst onto
the surface of LSCF6428 MIEC hollow fibre membranes can
increase the rate of oxygen reduction at the air-side mem-
brane surface and provides an overall increase in the oxygen
permeation rate. This indicates that under these conditions oxy-
gen reduction on the air-side is rate-determining. The module
comprising the catalytically modified hollow fibres membranes
maintained sealing integrity for ca 500 h of operation involving
many changes of temperature and oxygen chemical potential dif-
ference across the membranes before the onset of a significant
leak.
 A. Thursfield, I.S. Metcalfe / Journal of Membrane Science 288 (2007) 175–187
The formation of strontium sulfate occurred on the air-side
membrane surface, the origins of which may lie in the presence
of sulfur dioxide in the atmosphere, impurities in the helium
sweep gas or the use of a barium sulfate-containing high tem-
perature glass-ceramic sealant. The formation did not have a
significant affect on the oxygen permeation rate on the time scale
of the experiments at 1000 ◦ C but at 750 ◦ C may play a role in
the observed drop over a period of 8 days. After prolonged use
at high temperature, observable amounts of cobalt leached out
of the LSCF6428 membrane material but not enough to cause a
bulk change in the perovskite stoichiometry as indicated by ICP.
The LSCF6428 perovskite reacted with the ceramic cement
but this did not lead to catastrophic failure of the membrane.
The presence of aluminium throughout the Pt-HF membranes
from a solid state reaction is of concern and indicates that an
alternative method of pacifying the membranes must be found
as with time LSCF6428 can react with the glass-ceramic sealant
employed. EDS data indicates localised changes in stoichiom-
etry of LSCF6428 under typical oxygen permeation conditions
employed in this work.
Acknowledgements
This work was funded by the EPSRC, grant number
GR/S12197. The authors wish to thank Dr. Kang Li and Dr. Tan
Xiaoyao, Imperial College London, for supplying the hollow
fibre membranes used in this work.
References
[1] J.L.G. Fierro (Ed.), Metal Oxides: Chemistry and Applications, Taylor and
Francis, Boca Raton, 2006.
[2] L.G. Tejuca, J.L.G. Fierro (Eds.), Properties and Applications of
Perovskite-Type Oxides, Marcel Dekker Inc, 1992.
[3] Catal. Today, 26 (1990) 133–231 all references therein.
[4] P.J. Gellings, in: H.J.M. Bouwmeester (Ed.), The Handbook of Solid State
Electrochemistry, Taylor and Francis, London, 1997.
[5] A. Thursfield, I.S. Metcalfe, The use of dense mixed ionic and electronic
conducting membranes for chemical production, J. Mater. Chem. 14 (2004)
2475.
[6] T. Schiestel, M. Kilgus, S. Peter, K.J. Caspary, H. Wang, J. Caro, Hollow
fibre perovskite membranes for oxygen separation, J. Membr. Sci. 258
(2005) 1.
[7] C. Tablet, G. Grubert, H. Wang, T. Schiestel, M. Schoeder, B. Langanke,
J. Caro, Oxygen permeation study of perovskite hollow fiber membranes,
Catal. Today 104 (2005) 126.
[8] S. Lui, G.R. Gavalas, Oxygen selective ceramic hollow fibre membranes,
J. Membr. Sci. 246 (2005) 103.
[9] J. Luyten, A. Buekenhoudt, W. Adriansens, J. Cooymans, H. Weyten, F.
Servaes, Preparation of LaSrCoFeO3 − x membranes, Solid State Ionics 135
(2000) 637.
187
[10] X. Tan, S. Liu, K. Li, Preparation and characterization of inorganic hollow
fibre membranes, J. Membr. Sci. 188 (2001) 87.
[11] S. Lui, X. Tan, K. Li, R. Hughes, Preparation and characterization of
SrCe0.95 Yb0.05 O2.975 hollow fibre membranes, J. Membr. Sci. 193 (2001)
249.
[12] X. Tan, K. Li, Modelling of air separation in a LSCF hollow-fibre membrane
module, AIChE J. 48 (2002) 1469.
[13] X. Tan, Y. Lui, K. Li, Preparation of LSCF ceramic hollow fiber membranes
for oxygen production by a phase-inversion/sintering technique, Ind. Eng.
Chem. Res. 44 (2005) 61.
[14] X. Tan, Y. Lui, K. Li, Mixed conducting ceramic hollow fiber membranes
for air separation, AIChE J. 51 (2005) 1991.
[15] X. Tan, K. Li, Oxidative coupling of methane in a perovskite hollow-fiber
membrane reactor, Ind. Eng. Chem. Res. 44 (2006) 142.
[16] K. Li, X. Tan, Y. Lui, Single-step fabrication of ceramic hollow fibers for
oxygen permeation, J. Membr. Sci. 272 (2006) 1.
[17] Y. Lui, X. Tan, K. Li, SrCe0.95 Yb0.05 O3 − α hollow fiber membrane and its
property in proton conduction, AIChE J. 52 (2006) 1577.
[18] M. Martin, Materials in thermodynamic potential gradients, J. Chem. Ther-
modyn. 35 (2003) 1291.
[19] M. Martin, Electrotransport and demixing in oxides, Solid State Ionics
136–137 (2000) 331.
[20] J.-O. Hong, O. Teller, M. Martin, H.-I. Yoo, Demixing of mixed oxide
(A, B)O in an oxygen potential gradient: numerical solution of the
time evolution of the demixing process, Solid State Ionics 123 (1999)
75.
[21] S.M. Murphy, D.A. Slade, K.J. Nordheden, S.M. Stagg-Williams, Increas-
ing oxygen flux through a dense oxygen permeable membrane by
photolithographic patterning of platinum, J. Membr. Sci. 277 (2006)
94.
[22] A. Thursfield, I.S. Metcalfe, Methane oxidation in a mixed ionic-electronic
conducting ceramic hollow fibre reactor module, J. Solid State Elec-
trochem. 10 (2006) 604.
[23] H. Kruidhof, H.J.M. Bouwmeester, R.H.E.v. Doon, A.J. Burggraaf, Influ-
ence of order-disorder transitions on oxygen permeability through selected
nonstoichiometric perovskite-type oxides, Solid State Ionics 63–65 (2002)
816.
[24] X. Nanping, L. Shiguang, J. Wanqin, S. Jun, Y.S. Lin, Experimental
and modeling study on tubular dense membranes for oxygen permeation,
AIChE J. 45 (1999) 2519.
[25] S. Li, W. Jin, P. Huang, N. Xu, J. Shi, Y.S. Lin, Tubular lanthanum cobaltite
perovskite type membrane for oxygen permeation, J. Membr. Sci. 166
(2000) 51.
[26] H. Wang, T. Schiestel, C. Tablet, M. Scroeder, J. Caro, Mixed oxygen ion
and electron hollow fiber membranes for oxygen separation, Solid State
Ionics 177 (2006) 2255.
[27] M.M. Viitanen, R.G.v. Welzenis, H.H. Brongersma, F.P.F.F. van Berkel,
Silica poisoning of oxygen membranes, Solid State Ionics 150 (2002)
223.
[28] S. Li, H. Qi, N. Xu, J. Shi, Tubular dense perovskite type membranes.
Preparation, sealing and oxygen permeation properties, Ind. Eng. Chem.
Res. 38 (1999) 5028.
[29] S. Diethelm, J. Van herle, Oxygen transport through dense La0.6 Sr0.4 Fe0.8
Co0.2 O3 − δ perovskite-type permeation membranes, J. Eur. Ceram. Soc. 24
(2004) 1319.
[30] H.L. Lein, K. Wiik, T. Grand, Kinetic demixing and decomposition of
oxygen permeable membranes, Solid State Ionics 177 (2006) 1587.