SOLID
STATE
El S1WIER
Solid State lonics
Oxygen
1QNlCS
89 (1996) X1-92
transport through La, _,Sr,FeO, _ 6 membranes
II. Permeation in air/CO, CO, gradients
J.E. ten Elshof*,
H.J.M.
Bouwmeeskr,
H. Verweij
Oxygen permeation irieasuremcnts were pcrfcmned 011 dense LA, _,Sr,FeO,~-,
(s = 0. I-0.4) membranes in large oxygen
activity gradients. i.e. air/CO, CO,. in the tcmpcrature range I L73-1.323 K, yielding fluxes up to 25 mmol n-’ s- ’ at 1273
K. The oxygen semipermeability
was linearly proportional with the CO parrial pressure and the strontium content. Ir was
concluded that the fluxes are limited by the surl‘ace exchange kinetics. Two simple models arc proposcd for the oxygen
cxchangc reaction in the presence of CO. In both tnodcls oxygen vacancies at the phase boundary play a dcfinilc role in the
exchange process. XPS and SEM analysis of the perovskite/CO,
CO, interfaces indicated segregation of strontium. being
present at the surface mostly in the limu OF SI-CO; and/or SrO. The enhanced oxygen hxcs in air/He gradients as they
were observed after exposure of the membrane sutiace IO CO are probably (indirectly) related to the level of strontium
it is suggested thal the scgregetion process is
segregation. Based on SRM analysis and surface profile measurements.
accompanied by an enlargement of the specific surface area, which may promote the spcrd of the overall pcrmcation process
under exchange-controlled
conditions.
1. Introduction
The defect chemistry of La, _~$r_J%O,. s (x = O0.6) is well understood as a function of St- content,
oxygen partial pressure and tcmpcrature 11.,21.Propertics such as the diffusion coet’ficicnt, electron and
electron hole mobilitics are also known 1%51. In
part I of this paper (61 it was found that the oxygensemipermeable
properties of La, _ YSr,FcO,_,Y menlbranes (<Y= 0. I-0.4) in small partial pressure gradients, i.e. air/He; can be described by a diffusion“Carrcsponclingauthor.
controlled
mechanism.
In the present paper the
oxygen permeation process through the same membranes is coupled to a simple chemical reaction, i.c.
the oxidation of CO. The presence of CO, CO,
mixtures results in large oxygen activity gradients
across the membrane. Preliminary results showed a
linear correlation bctwcen the oxygen flux and the
CO partial pressure, indicating a surface exchangecontrolled
flux
[7) under these circumstances.
Voorhoevc 181 divided oxidation catalysis by perovskite-type oxides AM03=6 into two types, suprafacial and intrafacial, depending on the predominant
mechanism.
CO oxidation
at high tcmpcrature
0167-2738/9h/SlS.00 Copyripht G 1996 Elsekier Science B.V. All rights reserved
PII SO 167.2738i96)00255-X
proceeds primarily through the partial removal of
oxygen from the perovskitc lattice and subscyucnt
oxidation of the oxide bulk, and is an example of
jntrafacial catalysis. Thermodynamics,
M-O bond
strengths and oxygen diff~~sivity arc the detcrmjning
factors for this mechanism. An example of suprafacial catalysis is the oxidation of CO at relatively low
temperatures: i.c. 400~650 K. It takes place by the
reaction between adsorbed species and is therefore
determined by the electronic configuration of localized sites on the surface. Most studies of CO
oxidation on perovskitc-type
oxides were stimulated
by the search for possible inexpensive substitutes of
noble metals in automotive
exhaust catalysts and
were therefore performed in the range of 400-800 K
at low concentrations
and large space velocities 1%
I IJ. Several relationships
between the solid state
chemistry and the catalytic activity of perovskitcs
have been proposed. The catalytic
activjtics
of
several LaMO i I ,j perovskitc-type
oxides (M = Cr.
Mn, Fe, Co, Ni) at 500 K could bc rclatcd to the
crystal-field stabilization
energy ]9] or Fermi-level
energy ]Sj, revealing the suprafacial nature of the
catalytic process under these conditions.
Nitadori ct al. 1121 observed that the catalytic
activities of perovskites
LnMO? with Ln and M
b&g a trivalent rare-earth and a first-row transition
metal cation, respectively,
are determined
by the
nature of the transition metal cation. This implies
that the activity may also be affcctcd by the oxidation state of the transition metal. For the sake of
experiment,
its valency can be altered by (partial)
substitution
of’ trjvalent Ln by di- or tetravalent
cations. The effect of partial substitution of La‘ ‘. by
Ca*+ , Sr2+, Ra’+ Ce”T and Th4 has been studied.
Doshi et al. 1131 found a correlation between the
p-type electrical conductivity
and the rate of CO
oxidation on LIc ,-~SrlM,_,,M~,O,..
6 (M = Co, Cr,
Fe, Mn; M’ = Nb, Tij. However, it should be noted
that even in unsubstituted LaMO,, and M”’ cations
co-exist
next to M” ’ in general, their relative
concentrations
being dependent on the oxygen partial
pressure and temperature.
Rao et al. 1141 noticed a decrease in the catalytic
activity of LaCoO,_,
when the catalyst was annealed in pure oxygen. It was suggested that this
pretreatment lowers the oxygen deficiency and that
the activity is reduced as a result. Nakamura et al.
[15,16] determined the rates of reduction by CO and
re-oxidation by 0, of samples La, _,Sr.rCoO, _+ (,x =
O-0.6) as a function of the oxygen non-stoichiometry
6. The steady-state catalytic activities were estimated
from the intcrscction of the oxidation and reduction
curves,
and were found to agree well with the
obsetved trend in activity. cspccially for large values
of x. The catalytic process was explained in terms of
a redox-like
(intrafacial)
mechanism.
The rcdox
cycle mvolves a combined process of surface reaction and bulk diffusion of oxy’gen. To account for
the discrepancies at small x, a suprafacial mechanism
earlier proposed by Tascon et al. [ l7J for LaCoO,
was suggested. The relationship
between catalytic
acti\;ity on one hand and oxygen vacancy conccntration and electrical conducti\;ity on the other has
been observed also on the perovskite-related
spincltype oxide La, ISrXCu04_,5 1IX].
In this paper the oxygen permeation properties of
La, _,Sr,FeO,.
6 membranes
(X = (X1-0.4)
are
studied when they are placed in an air/CO, CO?
gradient. An attempt is made to relate the catalytic
properties of La, _lSr,FeO.,_ a to its oxygen vacancy
concentration.
2. Theory
In small oxygen partial pressure gradients j,:,> /IP~~,
al relatively
high oxygen partial pressures (pxZ,
properties
of
p;,> IO -1 bar? the permeation
La, .~Sr,FeO,, 8 membranes can be modelled by a
diffusion mechanism [6]. The permeation flux can be
expressed in terms of the Wagner equation:
Here I, is the membrane thickness (m); qio,, and cr,,
arc the ionic and electronic conductivities
(S m-‘)
respectively;
IT:, and 1,:. are the oxygen partial
pressures at opiosite
sides of the membrane.
At
temperatures
of 1000- 1400 K and oxygen partial
pressures p. >l(j -’ bar. p-type electronic conduc-
tivity is predominant
in La, _,Sr,FcO,
,s and the
electronic transference number remains very close to
unity. The electronic conductivity
decreases with
decreasing oxygen partial pressure: until a minimum
is reached, where a transition from p-type to n-type
conductivity
occurs. The region of minimum electronic conductivity,
typically
lo-‘IO’” bar 02,
coincides with a plateau region in the oxygen nonstoichiomctry,
where S----n/2, as shown in Fig. I.
Fig. 2 illustrates the calculated oxygen permeation
fluxes. The partially limiting effect of the electronic
conductivity
on the oxygen permeation rate can be
seen, especially for x=0.4, as a slight flattening of
= 10 -“’ bar. Details of the
the curve at about /TV_)_
calculations have been ‘described elsewhere [ci].
3 ,._
I,
.-
--.__.__.___
_._..----
_.-.
X-O.1
..__ .------
-il
4
I/--
2.9
h
x-o.25
_.___._._.._-
-/
I
a
_..._.
+
------1,
2.8 1I/-.--- __..._...___.-x-0.4
I
_
i/
2.7 :- ___,_.____
_. __--_
-18
-15
-12
-9
-6
I
-_I
-3
0
&I (pbz/bar)
Fig. 1. Calculated oxygen contacts of La, ,Sr,FcCJ i__cifor x=
0.1 -0.4 as a function of log ()c,T at 1273 K.
I()()
.i __-_
_.____.-.-.-
-.--
2.2. SurJucr
reactinn
Almost no details are known about the mechanism
of oxygen exchange on perovskite-type
oxides. In
the absence of a chemical reaction occurring at one
of the gas/solid interfaces, the overall process for the
incorporation and release of oxygen can be written as
$ o2 + v;; $0;
+ 211.
using ~KCiger-Vink notation [I 91. Based on “O/ “0
exchange experiments, Boukamp ct al. [ 20) proposed
a two-step mechanism for the exchange on fluoritetype Bi, .5Er0.50J in O,-containing
atmospheres, in
which dissociative
adsorption plus charge transfer
appears to be rate-limiting
at high 1~“~.
O2 + 2e’ # 20;,, T
(3)
20,,, + 2e’ + 2V;, 4 20:) .
(4)
Based on an observed interrelationship
between the
diffusion coefficients L), (cm’ s -‘) and the surface
exchange coellicicnts k (cm s -‘) or several oxides of
the fluorite and perovskitc
types from ‘“O/ I60
exchange experiments,
Kilner [21] suggested the
possible involvcmcnt of surface oxygen vacancies in
the exchange of oxygen. Since D, =(IIv/3)*6
for
perovskites, where the vacancy diffusion coefficient
13, (cm2 s- ‘) is approximately constant, the correlation bctwcen I), and k suggests a correlation bctween li and 6 as well.
In the present study the effect of CO, CO, on
ex.changc
reaction
of perovskites
the oxygen
La ,_-x SrJeO,_,
is investigated. There is evidence
that the mechanism
of the surface reaction
is
changed drastically
by their presence, c.g. from
isotopic exchange on YSZ 1221. The overall reaction
for the oxidation of CO can be written as
CO+0;,+2h’+COz+VV;;.
-18
-15
-12
-9
-6
-3
0
log @0&M
Fig. 2. Calcula~cd oxygen tluxcs through I .OO mm mcmhrnncs at
1273 K as a fllnction ol’ the pcrmcatc side oxygen partial yrcssure,
when oxyga~ is supplied ~1 the feed side. Model pxametcrs
are
descrihcd elsewhere (61.
(2)
(5)
Yasuda and Hikita 12.31 found a linear colrehtion
between the surface exchange constant and the nonin
stoichiometry
of La 1 XCa,CrO,_., [x=0.1-0.3)
CO, CO? almosphercs from electrical conductivity
relaxation experiments.
This indicates that oxygen
vacancies also play an essential role under these
conditions.
3. Experimental
The preparation
of the La,_.,Sr,FcO,
A mcmbranes (x-= 0.1-0.4)
and experimental
procedure
have been described in part I of this paper 161. Tho
density of the mcmbrancs was measured by means of
an Archimedes method. Some samples of ],a,, XSr,,,2.
Fw,_,
were sputtered with a 50 nm thick porous
layer of Pt.
The experimental set-up is schematically depicted
in Fig. 3. Unless specifically 4latcd, air was supplied
to the feed side of the membrane.
Composition
analysis of the permeate stream was pelformcd by a
Varian 3300 gas chromatograph coupled to a LDCI
Milton Roy CL-l.0 integrator. 02, N, and CO were
separated by a molecular
sieve 13X. CO, was
separated by a Porapak-N column. Measurements
were performed only on Icak-fret membranes,
in
which case no nitrogen and oxygen were detected in
the permeate stream. The oxygen Huxes were calculated from the mass balance of oxygen-containing
species:
where F is the tlow rate at the pcrmcate side (m’ s ’
(STP)): and A the geometric surkice area at the
CO-side of the membrane (m2). ci” and cj,“’ arc the
inlet and outlet concentrations
of species i. During
the measurements,
F was kept at values between
7-25 ml/min (STP). Tn order to avoid deposition of
,
L+
:
i
I-
.“,_ ._) Vent
‘*-
_- --
-
i
_
carbon, the CO/CO,
ratio was always kept above
unity. Assumin& idcal gas mixing bchaviour,
the
oxygen partial pressure al the Ji~cJllbl-ane/gas
interface was calculated
from the outlet partial pressures
of CO and CO,:
.
(7?
The values ol’ the equilibrium
constant K were
obtained from a compilation of thermodynamic
data
given in Ref. [24].
-3.2. Bulk ana1wi.r
XRD spectra were taken on a Philips PW I’710
with Ni-filtered CuKa radiation. LaB, was added as
internal standard to powders oblained Irom crushed
sintcred membranes.
The samples were measured
with u 20 scan from 20” to 130” with steps of 0.02”
and the intensity was collected during 5-10 s. The
CuKa, (h= 1.5406 A) component was used in the
rciincmcnt procedure. The bulk concentrations of La,
Sr and I% were determined by X-ray Fluorcsccncc
Spectroscopy (XRF) using a Philips PW 14X0/ IO
X-ray spectrometer.
The membrane surface morphology was examined
by High Resolution Scanning Electron Microscopy
(HR-SEM) using a Hitachi S-800 Field Emission
Microscope operating at IS kV, which was coupled
to a Kevex Delta Range Energy Dispersive X-ray
analysis (EDX) system for surface clement analysis.
The surface profiles of the samples were measured
using a Sloan DEKTAK 3030. The X-ray Photoelectron Spectroscopy (XPS) esperiments
were carried
out in a K.ratos XSAM 800 system controlled by a
PDP I 1 microcomputer. For the excitation, an AIKa
radiation source (225 W j was used. The spectrometer was calibrated
and its linearity chcckcd by
measuring the Ag 3,,j,l Cu 2p3,2 and Au 4f7,> peaks
on clean sputtered samples. The position of the C 1s
peak, relative to its normal position at 284.8 eV, was
usccl to correct the measured binding energies (BE)
for electrostatic charging of the samples. The perovskite membrane surfaces were analyzed by measur-
in& the La 3d, Fc 2p and Sr 3d spectra of clean
spultcred samples. Analysis was performed using a
DS X00 software package. 100% Gaussians
were
used for ~hc synthesis of the Sr 3d peak. Internal
consistency was maintained by constraining the spinorbit coupling ABE(Sr 3d:,,,-Sr 3d,,]) to I .8 cV and
the relative intensities and areas of the 3d,,, and
3d R,2 components lo a ratio of 3:2.
High Resolution
Transmission
Electron Microscopy (HR-TEM) and Selecled Arca Hlcctron Diffraction (SAD)
was performed
with a Philips
CM30 TWIN/(S)‘I’EM opcraling al 300 kcV, with an
aperture of 0.1-0.2 pm (point resolution 2.3 A),
f&cd with a double tilt goniomcter
stage (245”).
Measurements
were performed on single-crystalline
grains at the perovskitc/CO
interface at room temperature. The membranes were locally thinned with a
Gatan dimple grinder, followed by ion milling.
and discussion
4. Results
No second phases were detected in any sample.
For fc=O. I-0.3, all peaks could be indexed on the
basis of an orthorhombic
unit cell, whereas x=0.4
showed a rhombohedral symmetry. The relined unit
cell parameters are given in Table 1. The results are
in good agreement with those obtained by others 151.
The bulk compositions
as determined by XRI; arc
found to agree well with their corresponding nominal
compositions.
For the calculation
of the cationic
stoichiometrics listed in Table 2. it was assumed that
the oxygen non-stoichiometries
ol the samples were
ncpligible. The membrane dcnsitics were determined
from an Archimedes
type of method. The values,
relative to theoretical, arc also lisled in Table 2.
Tahlc 1
L!nit cell daw from XRD :II rowi
I
Cryslal
0. I
0.2
0.3
0.4
OrthorhodGc
OrthorhondGc
Orthorholnhic
Khombohedral”
“Hexagonal
syslew
sexing.
tcnq>eruture
for r-0.
0.1
0.2
0.3
0.4
0.99
0.99
0.10
0.20
0.29
0.40
0.90
0.80
0.70
0.60
96 97
95-06
94 95
91-92
I .oo
1SMI
Fig. 4 shows the oxygen flux through I+,.,%,,,, *
at 1273 K versus the oxygen partial presFe%,
sum. In the oxygen partial prcssurc range in which
the measurements were performedl i.e. p. = lo-’ ’ IO-” bar the non-stoichiometry
rS~0.05. ‘rhc model
calculation is based on diffusion-limited
permeation
and would thus indicate an upper limit for the
oxygen flux. Though the observed fluxes exceed this
limit, the trends of theoretical
and experimental
behaviour differ substantially. Similar measurements
[or x=0.2 and x =0.3 showed no fluxes exceeding
the theoretically predicted bulk diffusion limit, but
the observed trends wern: similar to those seen for
La,,_,Sr,,FeO,
h‘. This suggests that application of
the bulk dicfusion model, used earlier to fit experimental data of oxygen permeation at rclativcly
high oxygen pressures 161, is not \;alid when CO,
CO, gas mixtures arc preseni at the oxygen-lean side
or the membranes.
It is not clear why the diffusion limit is exceeded
for x- 0.1. Possibly, difrusion paths parallel to bulk
diffusion contribute signilicantly to the overall di ffu-
12
-g
9-i
r
-
-
I
*
\
A
.--
_
.-__.
I
I-O.4
(’ (A)
n (A)
h (A,
5.5403( I )
5.525-I(6)
55210(Hj
5.5278(X)
5.5510(Z) 7.835 l(2)
5.55 I3(9)
7.8267(2)
5.547X( I )
7.8054(3)
13.436X( 1)
\
0 +. _.__
-16
-15
_
_
-12
.-
-9
log (pop4
.J?b_
-6
J
-3
0
sion process at low oxygen partial prcssurcs, c.g.
grain boundary diffusion. Furthermore, the \;acancy
diffusion coeRicicnts D, used in the model calculations have only been determined at relatively high
PO,? i.e. 10 --’- 1 bar. Although Dv is generally
considered
to vary little wit11 5. Yasuda and
Hishinuma reported changes of’ D, of up to an order
01 magnitude at large 6 in L,a,_,Sr,,.lCro_,,Ni~~,~~~~~ ,?
and La,,_,iSr,l,jiCrO,
fi 1251. indicating that extrapolation lo low oxygen partial pressures may not be
allowed.
The oxygen flux is found to increase linearly will1
the CO partial pressure, as illustrated in Fig. 5 for
x==O.2. In view of Eq. (5), this clearly indicates the
surface reaction or CO with lattice oxygen to be the
rate-limiting
step in the overall permeation process.
Further evidence for a surface exchange-controlled
llux is found from the observation thal a sample
spulercd
with Pt shows a signiticantly
increased
flux. Jn comparison
with the bare membrane, the
slope (d.loLl d/pto) is larger by a factor of 1.8t0.2
[7]. The absence 01‘ mass transfer limitations was
\Terified by variation of the total flow rate between
7-25 ml/min (STP) at constant ~~c-oa~ the outlet of
the reactor. No significant change of the permeation
was observed.
Variation of the membrane thickness belwcen 0.5
and 2.0 mm did not show any influence on the
observed flux. An influence was n&her round from
changing the oxygen pressure at the higher partial
pressure side. In the latter experiments, pco was kept
constant at a value of 0.12 bar, while the oxygen
partial pressure at the feed side was varied bctwecn
15 T .__
__.-.
_
7
Y; 10:
5
EE 5’
2
7
1
-c’
-.._.-
/
m
/
.-;
0
25 ,~x=o.l
o iI-‘---’
0
i
A
/
A without Pt layer
0 with 50 nm Pt layer I
---_ _-.
-------. 7
0.1
0.3
0.2
P,~ (bar)
Fig. 5. Oxygen flux through
I .O mm [hick L.;+,,Sr,,.,lW :__.,
membrunes ilt 1273 K i*sa function ol’ C:O partial pressure.
--
--
/7
--- - 7___
0.1
-1
0.2
0.3
pco (bar)
Fig. 6. Oxygen Ruxcs through 1A _ .Sr,l%O i_-o mncmbrancs
I273 K as ii runction of C:O p:Aitl pressure.
-
_L.. ._.
0
0.02 and 1 bar. Thus. the surface rexlion
at the
lower partial pressure side is f~11ly limiting the
oxygen Hux.
The effect or %-doping on permeation at 1273 K
is given in Fig. 6. The oxygen flux clearly increases
with increasing
concentration
of Sr. Due to instrumental
limitaiions,
the oxygen fluxes through
La,,Sr,,.,FeO,_
6 were too large to be measured
quantitatively, but the observed trends were comparable to those on other compositions. The slopes k,,,,
(d./oZ/ dpc,,) of the plots of Fig. 6 were determined
by least-squares fitting. Fig. 7 shows a linear relationship between k,,,, and x. By considering that the
oxygen partial pressures in these experiments were in
all cases in the range 10 “’ - IO- ‘” bar, in which
region
non-stoichiometry
of
the
OXygeJl
La , .,Sr,FeO, ,? is virtually constant, i.e. S=x/2:
we think that oxygen vacancies play a role in the
125
_.
__ _._
-.
-
1
at
apparent exchange mechanism. It is unlikely that the
proportionality
is directly related to the strontium
content, since catalytic studies on perovskites AMO,
showed that the activity was not related to the nature
of the A-site cation ]_26,27)).
To account for the observed bchaviour we may
consider an Elcy-Rideal
mechanism [28] in which a
surface oxygen anion can only be attacked by CO
when it is residing next IO a surface oxygen vacancy.
When S << 3 -- 8, the concentration of such vacancyanion pairs will be proportional
to the vacancy
concentration,
leading to a proportionality
between
the flux and (5. Since Ihe oxygen vacancy concentrations do not cxcecd 5% of the total oxygen
contents in the expcrimcnts presented here, [Vi; OF,] = [V,]. Thus
CO + Vi; - 0; -_) CO, + Vi-i - Vi_; + 2e’ .
(8)
A second explanation is that of a Langmuir-Hinshelwood type of mechanism in which CO adsorbs on a
surface oxygen vacancy. before it rcacls
with a
neighbouring
oxygen ion:
co + v, +
1%.
yv;;. .
*CO], ,
. . COlj + 0:) -CO,
(9)
+ 2Vi_i + 2e’.
The activation energy for oxygen permeation
was
determined
for x-O.3
in the tcmpcralure
range
1173-1323 K from an cxpcrimenl in which pco=
0.08 bar and [>co, =0.32 bar were measured at the
outlet. The Arrhenius ~101 is given in Fig. 8. For the
sake of comparison, the Arrhenius plot for the same
composition
(thickness
0.92 mm) in an air/He
gradient is also shown. 1.t is clear that under the
conditions covered by the expcrimcnt. the oxygen
flux is strongly enhanced by the presence of CO: but
the activation energy has ticcreased. The calculated
activation energy was found to be 3 12 19 kJ/mol.
Using Eel. (1 I.), WC may interpret this activation
energy as consisting of two contributions, namely the
acGvation energy for oxygen vacancy formation and
the activation energy [or k,,:
Since the CO/CO,
ratio was kept constant, the
oxygen partial pressure changes considerably
over
the temperature
interval ( 10-‘2-10
Ii bar). The
oxygen non-stoichiometry
remains almost constant.
\;arying Tram about 0.152 to 0.156 between 1173 and
&-----A-_
--A_.__
--
_~
air/CO, CO, gradient
(11)
[Vi.;] = 6 by definition. It is assumed that the value of
k, does not vary with the strontium content. It
follows that
(12)
From Fig. 7, a value of 1.26+0.16
bar -’ is found for k,,.
(13)
(10)
The first step is assumed to bc rate-limiting.
This
model is in agreement with a suggestion made by
Yasuda and Hikita for La,,,7Ca,,,3Cr0,_.,
1231. The
authors speculated that gaseous species involved in
the surface reaction (CO and CO,) may form a
temporary bond with vacant lattice sites present at
the surface. I1 should be noted that due 10 the high
temperatures, adsorption such as is assumed in this
model is physically less likely.
For both mechanisms described ahove, the kinetic
expression Tar the oxygen flux will be
.Jo2 = + k,lV,lpcx, .
The positive intercepts in Fig. 6 arc due lo the
occurrence of a finite flux in the absence of CO 161.
so that the overall expression for J,, can be written
2
as
mol m-’
s-’
air/He gradient
B
I
-8.0 7
0.75
20&l
2 kJ/mol
0.6
0.65
1ow-r (l/K)
Fig. X. Apparent
activation
energies liar oxygen pcrrneation
tlm@
Lu,,,Sr,, $0.
A in air/h
and air/CO.
CO;, Hc gntdiznts for I .O mm membranes.
1323 K. Thus, within expcrimcntal
ent activation energy almost ret&
error, the apparto that of k,:,.
In the oxidation models proposed above. it is
assumed that the surface composition is the same as
that of the bulk. However, segregation of earlh alkali
cations has been observed on several cobalt-based
perovskites 129-3 1 I. Furthermore, imposing an oxygen partial pressure gradient across the membrane
appears to further increase the level of segregation
L-311. Since
the reaction
model
assumes
8=
i [Sr,‘,,]: an increased strontium concentration
in the
near-surface layer may affect the local concentration
of oxygen lracancies. SEMI EDX and XPS analysis
were performed to investigate segregation and surface morphology.
Several samples were quickly cooled down from
reaction conditions to room temperature by quenching the quartz reactor in ice water. EDX analysis
of the pcrovskite/COICOz
interfaces
did not
show
il
significant
deviation
from
the bulk
composition. SEM analysis showed a second phase
present at the surface. next to the perovskite
phase. The morphology
of this phase is very
similar
to that
of SrCO,
as found
in a
previous
study on La,,.,Sr,,.,*Co,~_,Fe,,,0,_,
and
l_a,,.,Ba,,,,Co,,,Fc~~,~O~ _$ membranes after exposure
to air/methane
gradients [301. However, much highcr levels of Sr and Ba segregation were observed in
the latter case. The difference is most likely related
to the large diferencc in chemical stability bctwecn
cobalt- and iron-based pcrovskitcs 1321.
Semi-quantitative
analysis of the level of Sr
segregation was pet-formed by XPS. Previous work
on related pcrovskiles La, ,Sr,CoO,_,
(x-0.4, 0.6)
showed that the Sr 3d spectrum can be deconvolutcd
into a low (131.X cV) and a high (133.8-134.2
eV)
HE contribution [33 I. The low BE has an anomalously low value and was attributed to strontium in the
bulk of the perovskitc. The high BE was considered
to be due to the presence of minor second phases at
the surracc, SUCII as SrCO, with a reported BE of
133.1-133.5
cV [34:35], and SrO with a BE of
135. I - I 35.6 cV 134,361. The strontium concentration
at La,. ,Sr,FcO,
B membrane surfaces is shown in
Fig. 9(a). Both fresh and used samples are enriched
0
$
0.2
0.1
a.
0.3
0.4
IWW+La)lb,lk
:-- - - . used
0.6 . fresh samples
samples
!
P
2
with Sr. The spectrum was deconvoluted
into two
components with 3d,,, maxima at 131.4-131.8 eV
and 133.2-133.9 cV. These peaks were attributed to
strontium present in the form of La,SrrFeO, _n‘and
minor second phases. respectively.
Their relative
contributions to the total Sr 3d spectrum are given in
Table 3. In each case these contributions
are about
50% for the used samples, with the exception of the
membrane that had only been exposed to He. The
relative contributions of the high BE to the spectra of
fresh samples are much smaller (12- 19%). By using
only the low BE contribution of the Sr 3d spectrum
Tahie 3
1)cconvolution of XPS-Sr 3d spectrum into high ;wl low binding
energy (BRj conwibul.ions l’or used and f&h membranes
x
0. I
0.2
0.3
0.4
0.4”
Fresh
Used
high BL (‘%)
46
1x
54
47
31
10~ NJ(%)
54
52
4h
53
(5’)
“Sanlplc exposed only to air/Ile
high HF.( %)
low BE(X j
17
12
83
xx
19
81
gradient.
(denoted as Sr,,,) the Sr enrichment in the perovskitc
lattice near the surface can bc estimated, as given in
Fig. 9(b). Within experimental error, the fresh samples show little or no enrichmcnl,
whereas a slight
cnrichmcnt is found for the used samples. The La/ Fe
ratios were also determined, and were found to be
close to the bulk ratios for both the fresh and used
membranes.
In part I. of this paper [6] it was reported that
oxygen permeation through La, _ 1Sr, FeO., _.8 membranes in air/He gradients can be enhanced by
treatment of the membranes lower partial pressure
side in a CO containing atmosphere. XPS analysis of
La,,,Sr,,,,FeO,
,j surfaces that had been exposed for
48 h to an air/He and an air/O.4 bar CO gradient:
respectively.
indicate a higher level of segregation
for the latter sample. The [Sr/(SriLa)]
ratios at the
surface are 0.61 and 0.73, respectively. The differencc in strontium enrichment
in the near-sin&c
perovskile layers bctwecn the Iwo samples is somewhat smal Icr, i.c. [Sr,;,, /( Srlnr) + La 1 equals 0.52 and
0.59. respectively.
Thus it appears that strontium
segregates to the surrace where it forms SrCO,
and/or SrO, and that this process is accelerated by
the presence of COI CO, mixtures. It is very possible
that 111~level of segregation is correlated with the
enhanced rate of oxygen exchange after treatment.
For instance, it is conjectured that a larger spccilic
arca on which oxygen exchange can take place is
directly or indirectly related to the level of segregation. The initial increase of Ihc oxygen fluxes with
time, as described in part .I [(,I. may be indicative 01
the time-scale
on which the segregation
process
occurs.
Surface morphologies
or a fresh and a used
membrane (quenched rrom an air/CO gradient) of
La,,,Sr,,,FeO,_,
are shown in Fig. 10 and Fig. I I.
The membranes have a rclalive density of 96% and
were taken from the same batch of material. The
surface areas shown are indicative [or both types of
surfaces. It appears that the surface roughness of the
quenched membrane
is larger, which indicates a
larger specific surface area. SurfLice profile measurements also showed a small but reproducible increase
01’ the surface roughness for the used membranes. In
comparison w-ith fresh samplesl the surface profile
lengths of used membranes were larger by a factor of
1.4-1.6. These resulls are in agreement with studies
of Deng et al. [ 37,381, who showed theoretically that
an enlargement of the specific surface area may have
a very pronounced effect on the pcrmcation fluxes
through mixed conductors under conditions of surfact-exchange
control.
Miura ct al. 1391 showed that the oxygen pcrmeation
rate
through
dense
Lilms
of
La,,_,Sr,,_,Co,,,,Fe,,,~O~ _ci prepared by slip casting,
w-as increased significantly
when the surFace was
treated with a solution containing 2 M HNO.;. The
authors conclude that the acid treatment resulted in
the removal of a surface layer containing impurities
like SrO. Their XPS-Sr 3d spectrum of a freshly
prepared surface showed maxima at 132 eV and 134
eV, with a shoulder at I35 cV. After the acid
treatment, tbc low energy peak. assigned by the
authors to SrO or SrCO:,. had disappeared completcly. while the intensity of the shoulder at 135 CV had
increased. However, in contrast to their findings.
90
J.E. ten Elshqf
et al. I Solid State Ionicx
XPS measurements
in a study on perovskitc-related
superconductors
Bi,Sr,CaCu,O,
and Bi,Sr,CuO,
1401 indicate that the low energy peak is due to
strontium from the pcrovskite lattice. XPS experiments on SrCO, powder performed in this laboratory
showed the Sr 3d5,* maximum at 133.2 cV. Furthcrmore. Table 3 shows that treatment of the membrane
results in a strongly increased contribution
of the
high BE peak. Jt therefore appears that the acid
treatment [39 1 decomposes the perovskite surface
completely, forming SrO or SrCQ,, and is accompanied by an increased oxygen eschange rate. This
result would be in agreement with the observations
made here and in a methane coupling study on
La,,,,Sr,,,,Co,,,,Fe~~,~O~ membranes
1.311, where an
enhanced oxygen flux was reported after reduction of
the membrane surface by CH,.
X9 (1996)
XI -92
Previous investigations
on the relationship
between crystal structure and oxygen content of the
La, _,SrxFeO 3_6 system [5] showed phase transitions
to a cubic structure for x=0.3 and x=0.4 when the
non-stoichiometry
increased to a value of x/2. When
the non-stoichiometry
of these phases was small
(S-O),
rhombohedral
symmetry
was found in
and
a
mixture
of
rhombohedral
La,.,Sr,.,FeO,-8,
and orthorhombic in La,,,Sr,~,FeO,_,
[5]. To verify
the assumption of a strongly depleted oxygen concentration at the air/CO interface, SAED experiments were performed
on single crystals at the
interfaces
of three La,.,Sr,,,FeO,_,
membranes,
either fresh or quenched from air/CO or air/He,
respectively. In contrast to the rhombohedral symmetry found by XRD, only orthorhombic symmetry was
found in tht fresh sample, while both cubic (a=
4.00+0.15
A) and orthorhombic
symmetries were
observed in both used samples. A SAED pattern of a
cubic crystal of La,,Sr,.,FeO,
8 is shown in Fig.
12. TVleasurements on La,,,7Sr,,,,Fc0,
n‘ samples
gave similar results, with a unit ccl1 parameter a=
3.069.1
I A for the cubic grains.
Although the occurrence of cubic symmetry may
indicate a high level of non-stoichiomctry,
care
should be taken to attach too great a value to these
results, since the mutual occurrence of more than one
type of crystal symmetry, even in equilibrated samples, has been reported by Dann ct al. IS]. Furthermore, the presence of orthorhombic symmetry may
indicate that part of the interface has been re-oxidized during the quench. Finally,
cubic crystal
symmetry does not necessarily prove that the oxygen
content is strongly depicted, but may also bc an
intrinsic high temperature crystal structure, irrespcctivc of the level of non-stoichiomctry.
5. Conclusions
The oxygen semipermeability
of dense mixcdconducting
membranes
with nominal composition
in air/CO, CO, graLa l ,Sr,FeO,_,s (x=0.1-0.3)
dients appears to be limited by the rate of CO
meiry LObe present in several grains, while no cubic
symmetry was observed in fresh mcmbrancs.
Acknowledgments
The support of the Commission of the European
Communities
in the li-amework of the Joule programmc, sub-programmc
Energy from Fossil Resources, Hydrocarbons,
is gratelYully acknowledged.
MA. Smithas, E.G. Keim and A.H.J. van den Berg
(Centre for Materials Research) are thanked for
performing the SEM, EDX, SAED and XPS measurements. N.M.L.N.P. Closset (MESA Institute) is
gratefully thanked for performing the surface profile
measurements.
References
Ill J. Mi7osaki,
oxidation at the gas/sol id interface. The activation
energy found on La,_,Sri,.3Fe0,_,3
was 31+ 19 kl/
mol. which is much lower than in air/He gradients.
The oxygen fluxes are linearly proportional to the
CO partial pressure and can be explained by simple
models for the oxidation of CO. Essentially, these
models yield the same kinetic expression for the flux
and both are based on the assumption that the nonstoichiometry
parameter is approximately
qua1
w
1/2x at the perovskite/CO,
CO, interface. Although
segregation of strontium was found at the surface;
the cnrichmenl of Sr in the near-surl‘ace perovskite
layer was found to be small. SrCO, and possibly
SrO was found on the membrane surface. but no
signs of complete decomposition
of the perovskite
surface were observed. SEM analysis and profile
mcasuremcnts of the membrane surfaces indicated an
enlargement of the specific surl‘ace area alicr exposure to CO. which may explain the enhanced fluxes
that were observed.
SAED on sin&-crystalline
grains al the interface
of La, _.VSrxFeO,_, (x-=0.3, 0.4) showed cubic sym-
M. Yoshihiro, S. Yamauchi and K. Fucki. J.
Solid Slalc Chcm. 58 (19X.5) 257.
PI J.J. Mizusaki, M. Yoshihiro. S. Yamauchi and K. Fucki. J.
Solid Smtc C.:hcm. 67 ( 1987) 1.
131T. Ishigaki, S. Yarnauchi, K. Kishio. J. Mizuwki and K.
Fueki, .I. Solid Slate Chcm. 73 (19881 179.
1111. MiLusaki, T. S~tsamo~o. W.R. Camlou and H.K. Howcn, J.
Am. Ccram. SW. 66 (1983) 247.
r51 S.E. Dnnn, D.B. Cbric, M.T. Weller, _M.F. Thomas, A.D.
Al-R;~wwas.
J. Solid Stare CI~rrn. IO!, (199-l) 134.
161J.E. ren Elshof, H.J.M. Boowmcestcr and II. Verweij. Solid
State lonics 81 (1995) 97.
171 B.A. van Hasscl. J.R. tw Elshol’ imd H.J.M. Hoowmeestcr,
Appl. CMal. A: General I I9 (1994) 279.
[8l R.J.H. Voorhneve, J.P. Rcmeika and L.E. Trimhle. Ann. N.Y.
hcad. Sci. 3 (1976) .3.
r91 J.M.D. Tasc6n, S. Mendioroz and LG. ‘I‘ejuca. Z. Phys.
Chem. 124 (1981) 109.
11011.. Wachowski. Z. Phys. Chcm. 269 (1988) 743.
111I C. KremeniC. J.M.L. Nieto, J.M.D. Tnsc:cin and L.J. Tejuca,
J. Chem. Sot., Faraday Trans. 1. XI ( 1985) 939.
1121‘I’. Nitudori. T. lchiki and M. Misono, Bull. Oem. SW. Jpn.
61 (1988) 621.
[I31 K. Tbhi. C:.H. Alcock, N.Gunasekamn and J.J. Carhcrry. J.
Catnl. 140 ( 1993) 5.57.
il41 D.Y. Rae and D.K. C.~hakrabnrt~. Ind. J. Chcm. 23.4 ( 1984j
375.
1151 T. Nakamura, M. Misono and Y. Yoncdn, Hull. Chcm. Sot.
Jp. 55 ( 1982) 394.
[IhI T. Nakamura, M. Misono and Y. Yonedn. .I. C&II. 83 ( 1983)
151.
[I71 J.IM.T>. ‘l;~sCcin, J.I..G. Fierro
Chcm. 124 (19X1 ) 249.
and L.G. T+ca.
Z. I’hy~.
[ 191 PA. Kriigcr, The Chemistry of lmpcrfec~ Crystals, (NorthHolland. Amsterdam.
I 004).
1201 B.A. Houkamp, H.J.M. 13ouwmecster antI A.J. Burggraf. in:
Proc. 2nd Inl. Synp. on Ionic and Mixed Conductin@
Oxidca. cds. T.4. Rnmanar;ly~ullan. W.I.. M;orrel and H.L.
Tullcr, 1904. p. 141.
[21 1 J.A. Kilner. in: Proc. 2nd Int. Symp. on lunic and Mixed
Conducting Oxides, eds. T.A. ~~lmRn;lr;lyan;ll~, W.I.. Worrel
arid 11.1,. Tullcr. 1994, 1,. 174.
1221 E.Kh. Kurumchin and M.V I’erfilizv. Solid Slate lonics. 42
(IOOO) 129.
1231 1. Yasoda and T. Hikita, J. Elecuwhem.
SW. 141 (1994)
1208.
(241 I. Harin and 0. Knackc. Tbcnnochemical
Properties ot
lnorfanic
Substances
(Springer.
13erlin. IHcidelbcrf, New
York, 197?1).
12.51 I. Y;lsucla and M. Ilishinuma. Solid State lonics, X0 ( lO95j
141.
1261 T. Nitnclori. T. lchiki and M. Misono, Bull. <‘hem. Sot. Jpn.
61 (1988) 621.
1271 P.IE. Marli and A. 13aikcr, Catal. LLU. 26 (1994) 7 I-X4.
(281 Catalysis:
an hitegruled
Approach
10 tlomogencous.
Hctero~cneous
and Industrial (‘alalysis. Studies in Surface
Science and Catalysis. Vol. 79. &. J.A. Moulijn, PW.N.M.
\-an I.cruwcn and K.A.
van
Santcn
(Elsevicr. .L\nislerduin.
1993 j.
1291 K. Tabata. I. Ma~sumoto and S. Kohiki. J. Mat. Sci. 22
(I 987) X82.
I.301 J.E. ten lilshof, B.A. van Hasacl and 11..l.M. Houwmeester,
Catal. Today 3-s ( 1995) x97.
1311 J.E. tw Elshof. 1i.J.M. Bouwmcc$ter and H. Vcrweij, Appl.
C%al. X: General I.?0 ~IW5) I%.
1321 I’. \;akanwra, G. Petzow and l..J. Guucklcr, Mnt. Res. Rull.
I4 IlO79) 649.
1331 A.E:. Rocclu~t. P. Chulktx, J.F. Dobson. P.C. He+. S. blyhra
and J.G. Thompwn. Physica C I60 (19X9)252.
1341 J.F. Moulclcr, W.1:. Stickle. 1%. Sohol. K.D. Bomhen.
1landhook of X-ray Photozlcclron
Spcctroscop)-.
A Kcfercnce T3ooli 0L‘ Stantl;wd Spectra for Identification and Intcrprctalion of Xl.5 Data. ed. J Chaslain (Pcrkin-Elmer Corporation. Physical Electronics
Division. Eden Prairie, LSA,
1092).
[3Sl A.U. Christie. J. Let: 1. Sutherland and J.M. Walls: .Appl.
Surf. Sci. 15 (19X3) 224.
I.351tl. van 1Iwerc.n and J.A. Verhoevcn. .I. Elecn~n Spccwosc.
Relnt. I’h~~xm. ‘I (IWO) 265.
I.371 H. Deng. M. Zhou ancl I3. Axles.
Solid Slate lonia 74
(1994) 75.
1381 H. Dcnp, M. Zhou and H. Abcles, Solid Sblc lonics X0
(1995) 213.
[?A I N. Mioru. Y. Okamob
J. Talnnki, I<. Morinaga ;md N.
Ya~nazoe.Solid %arc lonics 79 (1995) 195.
[401 A.1:. Hocquct. .J.F. Dobson. t?C. Healy; S. Myhra ancl J.G.
Thoq’son.
Phys. SW. Sol. B I52 (1980) 5 IO.