Journal of Membrane Science, 66 (1992) 211-226 211

Elsevier Science Publishers B.V., Amsterdam

Oxygen semipermeable solid oxide membrane

composites prepared by electrochemical vapor deposition

Y.S. Lin, K.J. de Vries, H.W. Brinkman and A.J. Burggraaf

Laboratory of Inorganic Chemistry, Materials Science and Catalysis, Department of Chemical Technology, University of
Twente, P.O. Box 217, 7500 AE Ens&de (Netherlands)
(Received June 10,199l; accepted in revised form September 30,199l)

Abstract

Ceramic membrane composites consisting of a coarse porous a-alumina or two-layer porous alumina
membrane support and an oxygen semipermeable gas tight thin (0.2-5 pm) yttria stabilized zirconia
(YSZ) film are prepared by the electrochemical vapor deposition (EVD) method. The minimum gas-
tight thickness of the YSZ films depends strongly on the average pore size of the support on which the
films are deposited by the EVD process. The oxygen permeation fluxes through such gas tight YSZ
membrane composites, measured in situ on the EVD apparatus, are in the range of 3 x lo-’ to 6 X lo-’
mol/cm’-set with an oxygen partial pressures of Pb, (high) x 3 x 10 m-2atm and P& (low) x 10 -’ atm,
much larger than the literature data for thicker YSZ pellets. During the oxygen permeation experiments
the rate-limiting step is found to be the bulk electrochemical transport in the grown YSZ films with a
thickness smaller than 10 pm.

Keywords: yttria stabilized zirconia; solid oxide membrane composites; oxygen permeation; electrochem-
ical vapor deposition; ceramic membranes

Introduction conia, YSZ) or fast mixed conductors (e.g. some

perovskite type materials) because they are
Ceramic membrane composites with an im- highly selective for oxygen transport [ 11. The
pervious but oxygen permeable thin dense elec- major challenge in preparation of these oxygen
trolyte film can be used in solid oxide fuel cells, permeable membranes is the fabrication of a
catalytic membrane reactors, gas sensors and very thin gas-tight film of these inorganic crys-
oxygen membrane separators, and therefore are talline materials supported by a porous support
of great importance in chemical, environmen- (which provides mechanical strength). An ox-
tal and energy conversion industries. The most ygen semipermeable solid oxide membrane film
interesting materials which can be used to make should be as thin as possible in order to maxi-
the oxygen semipermeable membrane films are mize the oxygen permeation flux since the elec-
fast ionic conductors (e.g. yttria stabilized zir- trochemical transport resistance is, in many
cases, proportional to the film thickness. Very
Correspondence to: Y.S. Lin, Department of Chemical En-
thin films can be fabricated without difficulty
gineering, 629 Rhodes Hall (ML 171) , University of Cin- from glassy or polymeric materials but not as
cinnati, Cincinnati, OH 45221-0171, USA.

0376-7388/92/$05.00 0 1992 Elsevier Science Publishers B.V. All rights reserved.

212 Y.S. LIN ET AL.

easily from the inorganic crystalline materials the resistance in the electrodes. For oxygen
of solid oxides. semipermeable membranes, a further decrease
Recent developments in the electrochemical of the electrolyte film thickness ( > 1 pm) with
vapor deposition (EVD ) technique, originally a sufficiently large surface oxygen exchange rate
developed by Westinghouse Electric Co. [ 21 for would be interesting in order to obtain a max-
fabrication of gas-tight solid electrolyte films imum oxygen permeation flux. Fabrication of
for fuel cell applications, showed that it is pos- such very thin gas-tight electrolyte films sup-
sible to use this technique for the manufacture ported on porous substrates has not been found
of oxygen semipermeable membrane compos- to be reported before.
ites with an ultrathin solid oxide electrolyte In order to better understand the EVD pro-
film. In the EVD process, a porous substrate cess, some theoretical and experimental work
(as the support in the membrane composite) has been carried out to investigate the kinetics
separates a mixture of two to three metal chlo- of the EVD film growing process [8-111. Car-
ride vapors and an oxygen source reactant olan and Michaels [8] and some other inves-
(water vapor or oxygen). Initially, in the so- tigators [ 9,101 reported theoretical analyses,
called chemical vapor deposition (CVD ) stage with some experimental results, of the growth
of the EVD process, the reactants from both of YSZ film on porous substrates by the EVD
sides of the substrate inter-diffuse into the sub- method. Only electrochemical transport in the
strate pores and react to form a solid solution bulk metal oxide film was considered. Follow-
of oxides deposited on the internal pore surface ing almost the identical procedure as employed
of the substrate. This solid deposit can even- for deriving the thickness-time relation for the
tually plug the substrate pore entrance in the Wagner scaling process [ 5,6], they found that
surface of the substrate exposed to the chloride the EVD kinetics can be described by the par-
vapors under certain conditions [3,4], result- abolic law (the film thickness is proportional
ing in a physical barrier which prevents direct to the square root of the deposition time) with
contact of the reactants on both sides of sub- a parabolic rate constant determined only by
strate. Since the deposit is an oxygen permea- the free-electronic conductivity [8,9] or to-
ble solid oxide, an oxide film can then grow on gether with the electronic-hole conductivity
the surface of the substrate in the chloride side [lo]. Observing some other experimental evi-
by the Wagner scaling process [5,6]. The dences on the EVD film growth data that can
unique feature of the EVD technique is its abil- not be explained by the parabolic law, Lin et al.
ity to make a gas-tight electrolyte film on a po- [ 111 recently reported a more systematic the-
rous support since the residual pores or pin- oretical and experimental study on the EVD ki-
holes, if there are any, can be plugged by the netics. They found that the EVD process in-
deposited oxides due to direct contact between volves four oxygen transport steps: (1)
the reactants in the residual pore openings. diffusion of oxygen reactant in the substrate
Previous studies on the EVD process focused pores (pore diffusion step); (2) reduction
on the growth of YSZ, films with a thickness charge transfer reaction at the oxygen/EVD
ranging from 5 to 50 pm on porous substrates film interface (reduction surface reaction step);
[2,7-lo]. For solid oxide fuel cells the thick- (3) electrochemical transport of oxygen ions
ness of the electrolyte in this range is accepta- and electrons in the EVD film (bulk diffusion
ble as long as the reduction of the ohmic loss step); and (4) oxidation charge transfer reac-
(due to the ionic conducting resistance) in the tion at the interface of the EVD film/metal
electrolyte is at a negligible level compared to chloride vapor (oxidation surface reaction
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 213

step). Lin et al. [ll] showed that either the rate-control made by Dou et al. [16]may not
pore diffusion step or the bulk diffusion step apply to the YSZ films prepared by EVD be-
can be rate limiting in the EVD film growing cause of the possible effects of different crys-
process, depending on the deposition condi- tallite size, much smaller film thickness ( -C10
tions such as substrate pore dimension, depo- pm) and different PO2 range. Therefore, inves-
sition temperature and pressure. tigation of the oxygen permeation through the
Oxygen permeation through the membrane YSZ film prepared by EVD is of great impor-
composites prepared by the EVD method tance in understanding the oxygen permeation
should, in principle, follow a similar mecha- mechanism through very thin electrolyte films.
nism as involved in the EVD film growing pro- In current literature, however, there are no ex-
cess: pore diffusion, reduction surface reaction, perimental studies reported on the oxygen per-
bulk diffusion and oxidation surface reaction. meation through stabilized zirconia films with
Among the limited number of studies on the a thickness smaller than 500 pm, due to una-
oxygen permeation through YSZ, most re- vailability of such thin gas-tight electrolyte
searchers [ 12-151 considered only the bulk dif- films supported on porous substrate and diffi-
fusion in the electrolyte film and found that the culties associated with sealing the thin electro-
oxygen permeation flux is reversily propor- lyte films on a permeation cell at high temper-
tional to the electrolyte thickness. They also atures for the permeation measurements.
found experimentally that the dependency of Furthermore, the previous oxygen permeation
the oxygen permeation flux on the oxygen par- work was all performed on stabilized zirconia
tial pressure is P6:j4 at lower PO, and P&/,”at thin pellets prepared by the conventional ce-
higher Po2, which is consistent with the theo- ramic powder-sintering process. No oxygen
retical prediction assuming the bulk diffusion permeation data on the electrolyte films pre-
as being the rate-limiting step. Dou et al. [ 161 pared by the EVD method is available in the
investigated the oxygen permeation in calcia- open literature. The main objectives of this pa-
stabilized zirconia films and considered both the per are to report the EVD preparation of ce-
reduction surface reaction step and the bulk ramic membrane composites consisting of a
diffusion step. They predicted, based on the ox- thin YSZ film with a thickness in the microm-
ygen permeation data on thicker ( > 500 pm) eter or nanometer scale and the oxygen per-
calcia-stabilized zirconia pellets, that a transi- meation through such gas-tight ceramic mem-
tion from the bulk diffusion rate-control to the brane composites.
surface reaction rate-control occurs when the
thickness of a calcia-stabilized zirconia sample Experimental
gets smaller than 270 pm at 1 atm of oxygen
partial pressure. The EVD experiments were performed in a
In the previous study [ 111, Lin et al. found home-built EVD apparatus consisting of a cen-
that the YSZ film growth rate (with a YSZ film tral reactor, water vapor delivery, metal chlo-
thickness of ca. 1 to 5 pm) during the EVD pro- ride vapor delivery and vacuum systems. The
cess is equivalent to an oxygen permeation flux central reactor was made of three dense alu-
which is several orders of magnitude larger than mina tubes of different size. A porous substrate
the value predicted using the surface reaction was cemented onto the end of the smaller tube,
rate coefficients reported by Dou et al. [ 161. separating the reactor into two chambers, as
This suggests that the prediction of the range depicted in Fig. l(A) which schematically
of the film thicknesses for the surface reaction shows the smallest and medium alumina tubes
214 Y.S. LIN ET AL.

A H20+A1r 2 mbar
umina on the coarse porous alumina substrates
I ‘1 h using the sol-gel method [ 191. The top-layer of
the membrane substrates had an average pore
diameter of 10 nm and a thickness of ca. 3 pm
and was thermally stable at 1000°C. Pore size
distributions of the coarse porous substrates
B H20+Aw 150 “-bar
and the top-layer of the membrane substrate
,,,..
are shown in Fig. 2. In the EVD experiments, a
!
Ar 150 mbar
‘\
‘\ 1 membrane substrate disk was cemented onto
the smallest tube in the central reactor with its
4
ZrCl4 YC13 I
I top-layer facing the chloride-chamber.
<lOOk <350°c 1odoE The deposition experiments were performed
in the EVD reactor with a total pressure of 2
Fig. 1. Schematic representation of the central EVD reac- mbar at both the chloride- and oxygen-cham-
tor during the film growing experiments (A) and the oxy- bers through which the gas-flows containing the
gen permeation experiments (B ) .
chloride vapors and water vapor were con-
stantly passed, as shown in Fig. 1 (A). During
of the central reactor. The ZrCl, and YCl, sub- deposition, the sublimation beds and substrate
limation beds were placed inside the medium were kept at the given temperatures. Typical
tube. The temperatures of the sublimation beds experimental conditions for growing YSZ film
and substrate were controlled by a six-zone are listed in Table 1. After a given period of time
heating element surrounding the largest dense of deposition (typically 20-100 min), the EVD
alumina tube which was coaxially placed out- film growing process was terminated by shut-
side of the medium tube (not shown in the fig- ting off the oxygen containing flow. Gas tight-
ure). In the EVD experiments, water vapor
ness of a substrate after deposition was checked
carried by air or other oxygen source reactant
by Ar permeation through the substrate at the
was introduced into the smallest tube (referred
to as the oxygen-chamber) and the chloride va-
pors were generated in the medium tube (re- 0.5 , LO.04

ferred to as the chloride-chamber) by passing

Ar through the sublimation beds. The volume
of the oxygen-chamber (375 ml) was about 20
times smaller than that of the chloride-cham-
-0.3
0.4 1

-
coarse
S”bStrate
Porous

10.03

0.03

e:

ber (8030 ml). The details on the EVD appa- i?iI

0.2 -
ratus are given elsewhere [ 17,181.
The following two types of disk-shaped ce- 0.1 -

ramic substrates (12 mm in diameter and ca. 2

mm in thickness) were used in the EVD exper- 1 1000
P&E RADdO~nm)
iments: (A) coarse porous a-alumina and (B )
two-layer alumina membrane composites. The Fig. 2. Normalized pore size distributions of the coarse po-
coarse porous substrates had an average pore rous substrate (by mercury porosimetry ) and the top-layer
of the membrane substrate (by nitrogen adsorption poro-
diameter of 0.2 pm and porosity of 50%. The
simetry) [P(r)dr has a physical meaning of the volume of
membrane substrates were prepared by coating the pores with a pore radius from F to r + dr to the total
a thin film of thermal stability improved y-al- volume of the all pores].
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 215

TABLE 1

Typical experimental conditions

EVD film growing experiments

Deposition temperature 1ooo”c
ZrCl, sublimation temperature 150°C
YCl, sublimation temperature 613°C
YClJZrCl, vapor (mole) ratio 115
Total pressure 2mbar (1.97X10-3atm)
ZrCl, vapor concentration 3.5 X 10 -g mol/cm3
Water vapor concentration 2.0X lo-’ mol/cm3
Water vapor/air (volume) ratio 2/3-1.1
PO, in chloride chamber < lo-l4 atm
PO, in oxygen chamber (with air/steam) 6X10p4atm
Oxygen permeation experiments
Total pressure in chloride chamber 150 mbar (0.148 atm)
Total pressure in the oxygen chamber 150 mbar (0.148 atm)
Permeation temperature 1ooo”c
PO,in chloride chamber - 10 -’ atm”
PO, in oxygen chamber <3X10-‘atm

“Estimated from the purity of Ar gas filled in the chloride chamber.

deposition temperature. This was done by in the sealing region between the edge of the
measuring pressure increase in the oxygen- substrate disk and the internal surface of the
chamber initially filled with Ar at 100 mbar and end part of the smallest tube were also filled by
isolated by closing all the inlet and outlet valves, the deposit during the deposition step. There-
with a constant pressure of 200 mbar in the fore the substrate was gas tight in the reactor
chloride-chamber filled with Ar. The Ar after deposition. This way of in-situ oxygen
permeability for the substrate (after deposi- permeation measurement avoided the high-
tion) of more than 200 times smaller than that temperature sealing problem and other diffi-
for the same substrate before deposition was culties associated with measuring the oxygen
used to roughly indicate the gas tightness. For permeation through these membrane compos-
some substrates after deposition, gas tightness ites using a special permeation apparatus.
was more accurately checked by measuring the In the in-situ oxygen permeation measure-
permeability in a home-built permeation mea- ments, the temperature of a substrate after de-
surement unit at room temperature. The thick- position was kept at a temperature similar to
ness and phase structure of grown zirconia/ytt- that for deposition (950-1050°C). The subli-
ria films were examined by SEM (Jeol JSM-35 mation beds were cooled down to 100’ C or lower
CF) and XRD (Philips PW 1710, Ni-filtered for ZrCl, and 350” C or lower for YCl, to mini-
Cu-Ka). mize the vapor pressure of the chloride. This
The oxygen permeation measurements on the was done to reduce the possible growth of the
ceramic membrane composites consisting of a YSZ film during oxygen permeation measure-
thin YSZ film grown by the EVD method were ments. This precaution may not have been
performed in the EVD apparatus after the EVD completely sufficient. A permeation experi-
film growing process. Since the EVD is a self- ment was started by flushing the oxygen-cham-
pore-plugging process, even pores or pinholes ber with an oxygen containing gas and the chlo-
216 Y.S. LIN ET AL.

ride-chamber with Ar for at least five minutes. sure change data of air or air/vapor from which
The chloride-chamber was filled with high pu- the slope of the blank run was subtracted.
rity Ar and the oxygen-chamber with air or air/
steam, both to a total pressure of 150 mbar Results and discussion
(0.148 atm). Both the reactor chambers were
isolated by closing all the inlet and outlet valves YSZ film growth
and the total pressure change in the oxygen- Gas-tight YSZ films of 2-8 ,um in thickness
chamber was monitored (by a personal com- were grown on the coarse porous substrates. It
puter ) , as shown in Fig. 1B. Before each per- is difficult, however, to grow gas-tight YSZ films
meation measurement, a blank run was per- with a thickness smaller than 1 pm on the coarse
formed with both reactor chambers filled porous substrates due, probably, to the larger
initially with Ar at 150 mbar to check possible average pore size of the coarse porous sub-
leakage through the closed inlet and outlet strates. The gas-tightness was proven by zero
valves. Typical experimental conditions for the He flux with a 4 bar pressure drop across the
oxygen permeation measurements are also EVD prepared membrane composite measured
given in Table 1. Figure 3 gives an example of at room temperature with the special permea-
the pressure change as function of time in the tion apparatus. The YSZ films grown on the
oxygen-chamber filled with Ar, air and air/ coarse porous substrates remain gas-tight after
steam. The slopes of the pressure change data three thermal cyclings between room temper-
are used to calculate the permeation flux (or ature and 1000’ C with a heating or cooling rate
leakage flux). There is a slight leakage through of lSO”C/hr.
the closed valves connected to the vacuum As reported in the previous study [ 111, the
pump, as indicated by the small slope of the rate-limiting step for the EVD film growth pro-
pressure change data for Ar shown in Fig. 3.
The fact that the pressure drop in the oxygen- TABLE 2

chamber when filled with air or air/steam is Effects of reactant partial pressures on YSZ film growth
much more substantial than when filled with rate*
Ar indicates a considerable amount of oxygen
flux permeating through the EVD prepared Effects of PO,
Oxygen source PO, Growth rate
membrane composite. The oxygen permeation
reactant (atm) (pm/hr)
flux was calculated from the slopes of the pres-
Air/Hz0 (1: 1) lo-“ - 1.5
OJH,O (1:l) 1O-3 - 1.5
Hz/H20 (2:3) lo-l4 1.2-1.6
Dried air or O2 lo-“ < 0.5

Effects of Pzch
T ZrCk( “C) PZrch Growth rate
(atm) (,am/hr)

s w
dud
AT
Air
00000 Air/Steam
140 6.4x 1O-5 2.0
150 1.3x 10-4 1.5
160 2.6x 1O-4 1.7
180 9.2x 10-4 1.5

Fig. 3. Pressure change in the oxygen-chamber as a func- “Values of all other parameters are the same as listed in
tion of time during oxygen permeation measurements. Table 1 for EVD film growing.
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 217

Fig. 4. SEM photographs of YSZ films grown at 700°C (a) and 1000°C (b) on the coarse porous substrates (Left: back-
scattered electron image; Right: secondary electron image).

cess on the coarse porous substrates is the dif- An average YSZ film growth rate of 1.5 pm/hr
fusion of the oxygen source reactant in the sub- at 1000oC was found on the coarse porous sub-
strate pores under the deposition conditions strates, consistent with the oxygen diffusion
(especially at low total pressure, e.g. 2 mbar). flux rate in the substrate pores. The YSZ film
218 Y.S. LIN ET AL.

growth rates on the coarse porous substrates TABLE 3

under the deposition conditions of different ox-

Effects of deposition temperature on the YSZ film growth
ygen partial pressure in the oxygen-chamber rate”
and different (zirconium) chloride vapor pres-
sure in the chloride-chamber are summarized Deposition Estimatedfilm Z~(lll)/Zor(Ol2)~
temp. (“C) thickness
in Table 2 to examine the effects of the reactant
(Pm)
concentrations on the YSZ film growth rates.
The results show that both the oxygen partial 700 1 0.2
pressure, in the studied range of lo-l4 to 10m3 900 1.2 0.3
850 1.2 0.8
atm, and the chloride vapor pressure have neg-
900 1.3 3.8
ligible effects on the film growth rates. Since 950 1.3 11
the rate-limiting step is diffusion of the oxygen 1000 3.0 22
source reactant (oxygen or HzO) in the pores, “Total deposition time 90-95 min; pore-closure time 30-45
the film growth rate is determined by the par- min.
tial pressure of the oxygen source reactant (all bRatio of XRD intensity for the reflection from cubic YSZ
ca. 5 x 10W4atm), and the substrate pore di- (111) to that from a-alumina (substrate) (012).
mension. The results given in Table 2 suggest
that the pore diffusion step is the rate-limiting phase YSZ to that of the (012) reflection of a!-
step under the current experimental conditions. alumina substrate for the six samples after de-
Deposition of YSZ on the coarse pore sub- position at different temperatures.
strates was also performed at different deposi- From the SEM photograph shown in Fig.
tion temperatures (700-1000°C) to study the 4 (a), it seems that a dense YSZ film could be
effects of temperature on the film growth rate. grown at a temperature as low as 700” C. How-
SEM photographs of the cross-sections of two ever, at 700°C it is expected that the bulk dif-
substrates after deposition at ?OO”C and fusion in the YSZ film should be the rate-lim-
1000°C are compared in Fig. 4 where in both iting step because the bulk diffusion resistance
cases a YSZ film can be seen on the top of each in the YSZ film at 700’ C is three to four orders
substrate. The film thickness estimated from of magnitude larger than that at 1000” C due to
the SEM photographs for several substrates the large activation energy for electron conduc-
(including the two shown in Fig. 4) after de- tion in YSZ [ 14,151. Therefore the EVD film
position at different temperatures are given in growth rate should be much smaller than 1.5
Table 3. Because YSZ was deposited on/in the pm/hr at 700°C. By a more careful examina-
rather coarse ceramic substrates, the estimated tion of the SEM photograph of the substrate
values of the film thickness from the SEM pho- after deposition at 700’ C (see Fig. 4 (a) ), it ap-
tographs should not be considered as very ac- pears that the dense-film-like YSZ crystallites
curate ones especially for the samples after de- are deposited, during the initial CVD stage, in
position at the lower temperatures. To provide (not on) the top part of the pores of the sub-
further information on the deposition results, strate. This can be further substantiated by the
XRD analysis was also performed on these XRD data discussed below.
samples after deposition. Figure 5 shows the The XRD patterns given in Fig. 5 for the
XRD patterns for the substrate after deposi- substrate after deposition at 700°C reveal very
tion at 700’ C. Table 3 also lists values of the strong reflections from a-alumina but rather
ratio R of the XRD intensity of the reflection weak reflections from cubic phase YSZ. The
from the (111) plane of the deposited cubic values of the XRD intensity ratio R [ = Ic ( 111) /
OXYGEN SEMIPERMEABLESOLID OXIDE MEMBRANE COMPOSITES 219

(1004) # 9024
yta 1000 ‘C

50 46 42
TWO THETA (3

Fig. 5. XRD pattern of YSZ grown at 700°C on the coarse porous substrate (a-alumina) (indexed with “c” for cubic phase
YSZ and “(Y”for cu-alumina).

la! (012 ) ] listed in Table 3 for this substrate and substrates. Because of the much smaller pore
the other two substrates after deposition at 800 size of the top-layer, the pore closure time was
and 850” C are all very small. Obviously, a typically ca. 5-10 min for deposition on the
smaller value of this XRD intensity ratio indi- membrane substrates, shorter than that on the
cates a thinner YSZ film or less coverage of the coarse pore substrates (30-45 min). With an
deposited YSZ on the surface of a substrate. additional 5-15 min deposition, ultrathin dense
Therefore, the rather small values of R for the YSZ films were grown on the surface of the
three substrates after deposition at the lower membrane substrate top-layers. Figure 6 is a
temperatures suggest that there is a very thin SEM photograph showing the cross-section of
or even no dense YSZ film on the surface of one of such membrane substrates after depo-
these substrates. The YSZ seen from the SEM sition. As in the CVD stage some of the YSZ
photographs is mainly composed of YSZ crys- may be deposited inside the substrate top-layer
tallites deposited in the substrate pores very which consists of ultrafine y-alumina crystal-
near the surface during the initial CVD stage. lites; the exact thickness of the dense YSZ film
When deposition temperature is higher, the formed in the EVD stage on the top-layer is very
XRD intensity ratio R increases with deposi- difficult, if not impossible, to be estimated from
tion temperature, showing an increase in the the SEM photograph. Figure 7 gives an XRD
coverage or thickness of the deposited films on pattern for one of the membrane substrates
the surface of the coarse pore substrate. This after deposition for 20 min and shows the ex-
indicates that the growth of a somewhat thicker istence of a cubic phase YSZ solid solution on
dense YSZ film on the substrate surface by the the membrane substrate surface. The XRD re-
EVD process is possible only at the elevated flections for a-alumina, although already cov-
temperatures because the EVD process relies ered by a thin y-alumina top-layer, are still very
on the electronic and ionic conductivities which strong. This indicates that the grown YSZ film
are strongly temperature dependent. on the top layer of the substrate must be rather
Following similar experimental conditions as thin.
listed in Table 1, gas-tight YSZ films were also As discussed before, under the EVD experi-
deposited on the top-layers of the membrane mental conditions listed in Table 1, the film
220 Y.S. LIN ET AL.

7 -Al,@

a-Al,O,

Fig. 6. SEM photograph of YSZ grown on the top-layer of the membrane substrate.

(113)

solo)

52 58 54 50 46 42 38 34 30 26 2
TWO THETA (‘)

Fig. ‘7.XRD pattern of YSZ grown on the membrane substrate (the same samples as of Fig. 6) (indexed with “c” for cubic
phase YSZ and “CT for a-alumina).

growth rate is determined by the diffusion of tion time after pore-closure (ca. 10 min), a
the oxygen source reactant in the substrate po- thickness of ca. 0.2 pm can be estimated for the
res. Since the diffusion resistance through the dense YSZ film deposited on the particular
membrane substrate consisting of a very thin membrane substrate discussed here. An accu-
alumina top-layer and the coarse pore sub- rate experimental method is under develop-
strate is ca. 20% larger than that through the ment for the measurement of the dense film
coarse pore substrate alone, the EVD film thickness on such a fine pore substrate. It is
growth rate on the membrane substrates should important to note that the membrane substrate
be ca. 1.2 ,um/hr, 20% percent smaller than that deposited with such a thin Y SZ film on the sur-
on the coarse pore substrate. Therefore from face of its top-layer is impervious to Ar, He and
the film growth rate (1.2 pm/hr) and deposi- Nz, as proven by the zero flux of these gases
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 221

through the membrane composites (subject to meation but also to some other phenomena such
a total pressure drop of 4 bar across the as adsorption of water vapor on the internal
membrane ) . surface of the oxygen-chamber. In most cases,
It is found that the minimum thickness of a it was found that the pressure drop for air/
YSZ film which is gas-tight on a porous sub- steam mixture became constant about 30 min
strate strongly depends on the average pore size after the introduction of the permeating gas in
and pore size distribution of the substrate sur- the oxygen chamber. Therefore the pressure
face on which the film grows. As a rule of thumb, change data at the steady-state were used, as
for a substrate with a reasonable uniform pore shown in Fig. 9, for calculating the oxygen per-
size distribution and porosity of ca. 0.5, the meation flux with air/steam mixture as the
minimum YSZ film thickness to be gas-tight permeating gas. For air the steady-state of the
should be roughly ca. 10 times the average pore pressure change rate was reached as soon as the
diameter of the substrates. Therefore for the permeating gas was introduced. To give an ex-
coarse pore substrates used in this study (pore ample, Fig. 10 shows the pressure changes as a
diameter=0.16 pm), a thickness of the YSZ function of time in the oxygen-chamber with
film larger than 2 pm can ensure the gas-tight- air as the permeating gas at different
ness of the grown film. For the deposition of temperatures.
YSZ films on the top-layers of the membrane
substrates (the top-layers have a pore diameter
of 10 nm, or slit pore width of 6 nm with a more
uniform pore size distribution), the minimum
gas-tight thickness can be as small as 60-100
nm. Therefore ultrathin gas-tight YSZ films can
be grown only on these membrane substrates.

Oxygen permeation
A negative pressure change in the oxygen- a
h -1.5 ~ , ,
0 10 30 40
chamber was found for all the in-situ oxygen Tim? (min)
permeation experimental runs after deduction
of leakage. For the mixture of air/steam as the Fig. 8. Pressure drop in the oxygen-chamber for air/steam
in the initial period of permeation experimental runs.
permeating gas, the pressure drop in the oxy-
gen-chamber is more rapid in the initial period
and becomes constant ca. lo-20 min after the
introduction of the permeating gas in the oxy-
gen-chamber, as shown in Fig. 3. If the initial
slopes of the pressure drop data are used to cal-
culate the permeation flux, the air/steam mix-
ture gives a much larger permeation flux than
the air. However, it is found that the initial
slopes of the pressure change data for air/steam
mixture, as shown in Fig. 8, decrease with in-
creasing temperature. This indicates that the
pressure drop for the air/steam mixture in the Fig. 9. Pressure change for air/steam in the oxygen-cham-
initial period is due not only to the oxygen per- ber at different temperatures.
 YSLINETAL.

set for air and 11.6~ lo-’ to 38.8~ lo-’ mol/

cm2-set for air/steam mixture. The leakage flux
in most cases is ca. lo%-20% of the total flux
measured. Therefore the accuracy of the oxy-
gen permeation flux measured should be within
? 20%. Assuming the pore diffusion being the
=m 95OT
0 rate-limiting step, the oxygen flux through the
anaon
L&La
1000°C
1050°C Air
coarse porous substrate under the conditions of
oxygen permeation measurements (P& = 0.03
0

atm and Po, < 10m5 atm, T= 1000°C where

Pb, and P& are the oxygen partial pressure in
Fig. 10. Pressure change for air in the oxygen-chamber at
different temperatures. the oxygen-chamber and the chloride-chamber
respectively) is ca. 110 X lo-’ mol/cm2-set, five
TABLE 4 to ten times larger than the permeation flux
measured on the composites as listed in Table
Oxygen permeation flux through YSZ films grown on the
4. Therefore, the pore diffusion is not the rate
coarse porous substrates
limiting step under the conditions of the oxy-
Sample Permeation Thickness of Permeation flux gen permeation experiments.
# temp (“C) YSZ (pm) ( 10-smol/cm2- The measured oxygen permeation fluxes for
set )
these YSZ composites listed in Table 4 are much
Air Air/steam larger than the literature data for the thicker
YSZ pellets at somewhat different experimen-
9015 950 5 3.54 5.65 tal conditions [ 12-161. Some of the published
1000 5 8.76 11.6
1050 5 15.6 14.9
data of free-electron and hole conductivities
9026 1000 4 12.4 19.1 calculated from the oxygen permeation data of
9063 1000 5 14.6 38.8 relatively thicker YSZ pellets [ 14,151 are given
9022 1000 3.5 10.7 - in Table 5. The oxygen permeation flux data
9032 1000 6 9.4 -
published for the thicker YSZ pellets can be ex-
trapolated to the present experimental condi-
The oxygen permeation fluxes are calculated tions using these electronic conductivity data
from the slopes of the pressure change data as with the following equation which is derived
shown in Figs. 9 and 10 by: based on the assumption of bulk electrochem-
ical transport being the rate-limiting step [ 111:
Joz = (V,IR7’,S) ( -@ldt) (1)
Jo2 = (RT/LF’){o;[ (P&)“4- (P&)1’4]
where V, is the volume of the oxygen-chamber
(375 cm3); T, is the average temperature in the +6:[ (P&-l’“- UC,)-““I} (2)
oxygen-chamber (363K); R the gas constant
and S the permeating area of the membrane where erg and ox are the electronic-hole and
composite (0.5 cm”). The measured oxygen free-electron conductivities; F the Faraday
permeation fluxes on the YSZ film supported constant; T the temperature; L the thickness of
on the coarse porous substrates are summa- YSZ; Pb, and P& are again the oxygen partial
rized in Table 4. The determined oxygen per- pressures in the higher pressure side (in the ox-
meation flux at 1000°C for different samples ygen-chamber) and lower pressure side (in the
ranges from 8.7 x lo-’ to 14.6 x lo-’ mol/cm2- chloride-chamber). The oxygen permeation
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 223

TABLE 5

Previous oxygen permeation experiments on YSZ and extrapolated permeation fluxes

Investigators Park & Blumenthal [ 151 Kleitz et al. [ 141

Electrolyte 8 m/o YSZ 9 m/o YSZ
Disk thickness (mm) 1.5 3-3.6
Temperature ( ‘C) 800-1050 1170-1550
@!, (ohm-1-cm-‘-atm-*/4)
-general formula I,3 x I07e-3.s8==~/KT 5 5X 105~-3.72ev/KT

-at 1000°C 5.8x 1O-9 1.0x 10-a

a; (ohm-1-cm-‘-atm”4)
-general formula 3 3x ~02~-‘.67ev,KT 1 4><1,,e-‘.52ev,KT
-at 1000°C 5.7x 10-5 1.6x 1O-5
Extrapolated permeation
flux (mol/cm’-set)* 4.8x lo-’ 1.4x 10-s

“Extrapolated to the present oxygen permeation conditions ( T= 1000", L = 5 pm, Pb, = 0.03 atm, P& = 10 -’ atm ) .

fluxes measured in the present study (listed in experiments and the in-situ oxygen permeation
Table 4) agree reasonably well with the extrap- experiments, as seen in Table 1. It is known that
olated data given in Table 5. If the surface re- the oxygen permeation flux through an YSZ
action rate coefficients measured by Dou et al. composite is proportional to PO, with the pore
[ 161 for calcia-stabilized zirconia is used to diffusion as the rate-limiting step and roughly
predict the oxygen permeation flux under the to (P& )I’4 with the bulk electrochemical
present conditions, a value of ca. 1.3~ lo-” transport as the rate-limiting step [ 111. There-
mol/cm’-set is obtained assuming the surface fore the difference in the oxygen partial pres-
reaction being the rate-limiting step. This is sures between these two types of different ex-
three orders of magnitude smaller than the ex- periments may result in a different rate-limiting
perimentally measured data presented in Table step for each type of experiment. The calcu-
4. It should be pointed out that the results of lated oxygen permeation fluxes, assuming the
Dou et al. were obtained for the thicker zirconia pore diffusion or bulk diffusion being the rate-
pellets under much higher oxygen partial pres- limiting step and using the method proposed by
sures compared to the present experiments. Lin et al. [ 111, are presented in Table 6 for these
Therefore the surface reaction may not be the two types of experiments. The step which gives
rate-limiting step for the very thin YSZ films the smaller predicted oxygen flux should be the
prepared by EVD under the current experi- rate-limiting step. In the oxygen permeation
mental conditions. experiments, the rate-limiting step is the bulk
The average YSZ film growth rate during electrochemical transport in the grown YSZ
EVD film growing experiments is ca. 1.5 pm/ film. During the EVD film growing experi-
hr at 1000 oC, which is equivalent to an oxygen ments the pore diffusion becomes the rate-lim-
permeation flux of ca. 1 X lo-’ mol/cm2-sec. iting step because of the lower Po, in both re-
This value is 5 to 10 times smaller than the per- actor chambers. The experimentally measured
meation fluxes reported in Table 4. This differ- oxygen fluxes in these two kinds of the experi-
ence is mainly due to the different oxygen par- ments agree with the oxygen flux predicted for
tial pressure in the oxygen-chamber and the the EVD film growing experiments (assuming
chloride-chamber between the EVD film growth the pore diffusion being the rate-limiting step)
224 Y.S. LIN ET AL.

TABLE 6
used for the oxygen permeation measurements
Comparison of oxygen permeation flux through YSZ film for YSZ deposited on the coarse porous sub-
during EVD film growing and oxygen permeation strates. The results are summarized in Table 7.
measurements As mentioned before, the exact thickness of the
EVD film Oxygen
deposited dense YSZ films on the membrane
growing Permeation substrates was difficult to measure. Further-
more, slight pre-deposition or post-deposition
PO, in oxygen-chamber 6 x~O--~ 3 x10-2
before or after introduction or termination of
(atm)
PO, in chloride- < lo-‘4 < 10-5 oxygen source reactant during the EVD film
chamber (atm ) growing process may also make it difficult to
Pore diffusion 2 x10-9 1 x10-7 accurately predict the exact film thickness in
controlled J$
the current experiments. Nevertheless, it is ac-
Bulk diffusion > 6 x10-’ N 1.5x10-s
controlled J& ceptable that the YSZ films grown on the mem-
Experimental N 1.0x10-9 N 1 x10-s brane substrates should be thinner than those
measured J& grown on the coarse porous substrates because
“Unit: (mol/cm’-set) of the shorter deposition time. Due to the
smaller thickness of YSZ films, the oxygen per-
TABLE 7 meation fluxes through the YSZ films on the
membrane substrates, as listed in Table 7, are
The oxygen permeation fluxes with air as permeating gas
through YSZ films grown on membrane substrates generally larger than those through YSZ on the
(1OOO~C) coarse porous substrates presented in Table 4.
These results further show that under the ex-
Sample Deposition Estimated Permeation flux
perimental conditions for oxygen permeation
# time YSZ thickness ( 10e9 mol/
(min) (pm) cm’-see) measurements the surface reaction obviously
does not play a dominant role. The bulk elec-
9037 10 0.2 20.9 trochemical transport in such thin YSZ films
9038 30 0.5 55.3
9040 30 0.5 55.3
is the rate-limiting step. This suggests that a
9048 30 0.5 31.5 further decrease in the YSZ film thickness,
9052 30 0.5 17.1 which, of course, should remain gas-tight, may
further raise the oxygen permeation flux.
and for the oxygen permeation (assuming bulk It can be also found in Table 4 that the oxy-
diffusion being the rate-limiting step). The gen permeation fluxes for air/steam mixture,
EVD film growth data and the oxygen permea- in most cases, are larger than those for air. Fur-
tion data are in general consistent to each other. thermore, the activation energy calculated from
This further supports that for such YSZ films the data of Run # 9015 using the Arrhenius plot
prepared by the EVD method the oxygen per- for air/steam mixture (130 kJ/mol) is about
meation flux is indeed at least larger than two third of that for air (198 kJ/mol). This
1 x lo-’ mol/cm’-set under the experimental suggests that addition of water vapor to the air
conditions investigated. may enhance the oxygen permeation flux and
The oxygen permeation fluxes through the reduce the activation energy for oxygen per-
YSZ films grown on the membrane substrates meation. It is also found that the YSZ film
were also measured in-situ using air as the per- growth rate with dried air or pure oxygen is
meating gas under the same conditions as those much slower than that with humid air, air/
OXYGEN SEMIPERMEABLE SOLID OXIDE MEMBRANE COMPOSITES 225

steam mixture or oxygen/steam as the oxygen avoided possible high temperature sealing
source reactant, as listed in Table 2. Water may problems and sample transferring inconveni-
play a similar role in enhancing the oxygen flux ence if performed in a specially designed per-
in the both cases. Although some researchers meation apparatus. The oxygen permeation
[ 201 have noticed that water vapor may in- fluxes through such YSZ membrane compos-
crease both the surface reaction rate and bulk itesareintherangeof3x10-gto6x10-8mol/
diffusion of the oxygen transport in solid oxide cm’-set, which is much larger than the litera-
electrolytes, the exact mechanism of the oxy- ture data for thicker YSZ pellets. The litera-
gen permeation with water vapor through the ture data of the oxygen permeation extrapo-
YSZ is an interesting topic for further research. lated to the present experimental conditions
It should be pointed out that the EVD appara- assuming bulk diffusion being the rate-limiting
tus used in this study was not originally de- step, however, agree well with the oxygen per-
signed for the oxygen permeation experiments. meation data reported in this study. The oxy-
The apparatus limitation, therefore, made it gen permeation fluxes through the thinner YSZ
difficult to investigate more accurately and films supported on the membrane substrates are
comprehensively the oxygen permeation phe- larger than those through the thicker YSZ films
nomena for such EVD prepared membrane grown on the coarse porous substrates. These
composites with thin YSZ films. Nevertheless, results suggest that during the oxygen permea-
the work reported here, to our best knowledge, tion experiments the rate-limiting step is the
represents the first effort on investigating the bulk electrochemical transport in the grown
oxygen permeation through solid electrolyte YSZ films. Furthermore, addition of water va-
films with a thickness thinner than 10 ,um. por in air enhances the oxygen permeation flux
and reduces the activation energy for oxygen
Conclusions permeation.

Ceramic membranes consisting of a gas-tight Acknowledgement

thin YSZ film with a thickness ranging from
0.2 to 5 pm on top of a porous alumina ceramic This investigation was supported by the
support can be prepared by the EVD method. Dutch Ministry of Economical Affairs (IOP
The minimum gas-tight thickness of the YSZ Technical Ceramics No. 87 A045). Mr. J. Mei-
films depends strongly on the average pore size jerink and Ir. G. Polhaar are acknowledged for
of the substrates on which the films are depos- the assistance in running the experiments.
ited. Gas-tight YSZ films thinner than 1 pm can
be grown on the two-layer membrane sub- List of symbols
strates consisting of a thin y-alumina top layer
with improved thermal stability. The rate-lim- F Faraday constant
iting step during the EVD film growing process Joz oxygen permeation flux ( mol/cm 2-set )
under the present film growing experimental L thickness of substrate disk or layer (mm
conditions and the specific supports used is the orpm)
diffusion of the oxygen source reactant in the P pressure in the oxygen-chamber (atm)
substrate pores. P& oxygen partial pressure at the lower pres-
Oxygen permeation through the supported sure side (atm)
YSZ films grown by the EVD method was in- Pb, oxygen partial pressure at the higher pres-
situ investigated on the EVD apparatus which sure side (atm)
226 Y.S. LIN ET AL.

gas constant 8 M.F. Carolan and J.M. Michaels, Growth rates and
mechanism of electrochemical vapor deposited yttria
permeation area of the substrate disk stabilized zirconia films, Solid State Ionics, 37 (1990)
(cm2) 189.
time (set or min ) 9 U.N. Pal and S.C. Singhal, Electrochemical vapor de-
position of yttria stabilized zirconia films, J. Electro-
average temperature of the oxygen-cham-
them. Sot., 137 (1990) 2937.
ber (“CorK) 10 N.J. Kiwiet and J. Schoonman, Electrochemical va-
permeation temperature ( ‘C or K) por deposition: theory and experiment, Proc. 25th Int.
volume of the oxygen-chamber (cm3) Sot. Energy Conversion Eng. Conf., P.A. Nelson et al.
(Eds.), AIChE, New York, NY, 1990, p. 240.
free-electron conductivity in the EVD
11 Y.S. Lin, L.G.J. de Haart, K.J. de Vries and A.J.
YSZ film at one atm of oxygen partial Burggraaf, A kinetic study of the electrochemical va-
pressure (ohm-1-cm-‘-atm-‘/4) por deposition of solid oxide electrolyte films on po-
hole conductivity in the EVD YSZ film at rous substrates,J. Electrochem. Sot., 137 (1990) 3960.
12 S.F. Palguev, V.K. Gilderman and A.D. Neujmin, Ox-
one atm of oxygen partial pressure
ygen permeability of stabilized zirconia solid electro-
(ohm-‘-cm-1-atm1/4) lyte, J. Electrochem. Sot., 122 (1975) 747.
13 J. Fouletier, P. Fabry and M. Kleitz, Electrochemical
semipermeability and the electrode microsystems in
References solid oxide electrolyte cells, J. Electrochem. Sot., 123
(1976) 204.
14 M. Kleitz, E. Fernandez, J. Fouletier and P. Farby,
A.J. Burggraaf, H.J.M. Bouwmeester, B.A. Boukamp Determination of electronic conductivities and ionic
and V. Zaspalis, Synthesis, microstructure and prop- domains of ZrO,-Y,O, by semipermeability measure-
erties of porous and dense ceramic membranes, in: J. ments, Adv. Ceram., 3 (1981) 349.
Nowotny (Ed.), Science of Ceramic Interfaces, Else- 15 J.H. Park and R.N. Blumenthal, Electronic transport
vier, Amsterdam, 1991, in press. in 8 molar percent ZrOz-YzO,, J. Electrochem. Sot.,
A.O. Isenberg, Growth of refractory oxide by electro- 136 (1989) 2867.
chemical vapor deposition (EVD) at elevated tem- 16 S. Dou, C.R. Masson and P.D. Paccy, Mechanism of
perature, in: J.D.E. MC Intyre, S. Srinavasan and F.G. oxygen permeation through lime-stabilized zirconia,
Will (Eds.), ECS Symp. Electrode Materials, Pro- J. Electrochem. Sot., 132 (1985) 1843.
cesses for Energy Conversion and Storage, Electro- 17 Y.S. Lin, L.G.J. de Haart, K.J. de Vries and A.J.
them Sot. Inc., Princeton, NJ, 77-6 (1977) 572. Burggraaf, Modification of ceramic membranes by
CVD and EVD for gas separation, catalysis and SOFC
Y.S. Lin, P. Fransen, K.J. de Vries and A.J. Burg-
application, in: G. de With, R.A. Terpstra and R. Met-
graaf, Experimental study on CVD modification of ce-
selaar (Eds. ) , Euro-Ceramics, Vol. 3-Engineering
ramic membranes, Proc. 11th Int. CVD Conf., K.E.
Ceramics, Proc. 1st Eur. Ceramic Sot. Conf., Maas-
Spear and G.W. Cullen (Eds.), The Electrochem. Sot.,
tricht, June 18-3, 1989, Elsevier, Amsterdam, 1989,
Pennington, NY, 1990, p. 539.
pp. 590-95.
Y.S. Lin and A.J. Burggraaf, Modelling and analysis
18 L.G.J. de Haart, Y.S. Lin, K.J. de Vries and A.J.
of CVD processes in porous media for ceramic com-
Burggraaf, Modified-CVD of nanoscale structure in
posite preparation, Chem. Eng. Sci., 46 (1991) 3067. and EVD of thin layers on porous ceramic mem-
P. Kofstad, High-Temperature Oxidation of Metals, branes, J. Europ. Ceram. Sot., 8 (1991) 59.
Wiley, New York, NY, 1966, Chap. 5, pp. 112-146. 19 Y.S. Lin and A.J. Burggraaf, Preparation and char-
L. Heyne, Electrochemistry of mixed ionic-electronic acterization of high-temperature thermally stable al-
conductors, in: S. Geller (Ed.), Solid Electrolytes, umina composite membranes, J. Am. Ceram. Sot., 74
Springer-Verlag, Berlin, 1977, p. 169. (1991) 219.
G. Dietrich and W. Schafer, Advances in the devel- 20 V.S. Stotz and C. Wagner, Die Lijslichkeit von Was-
opment of thin film cells for high temperature elec- serdampfund Wasserstoff in festen Oxiden, Ber. Bun-
trolysis, Int. J. Hydrogen Energy, 9 (1984) 747. senges., 70 (1966) 780.