You are on page 1of 21

PROGRESS REPORT

Membrane Reactors www.advmat.de

Microstructural and Interfacial Designs of Oxygen-


Permeable Membranes for Oxygen Separation
and Reaction–Separation Coupling
Xuefeng Zhu and Weishen Yang*

pure oxygen will rapidly increase. Cur-


Mixed ionic–electronic conducting oxygen-permeable membranes can rapidly rently, the commercialized technology for
separate oxygen from air with 100% selectivity and low energy consumption. the generation of pure oxygen is the cryo-
Combining reaction and separation in an oxygen-permeable membrane genic distillation of air. However, the cost
reactor significantly simplifies the technological scheme and reduces the of cryogenic distillation technology for
oxygen separation remains high. There-
process energy consumption. Recently, materials design and mechanism
fore, it is facing a great energy penalty, i.e.
investigations have provided insight into the microstructural and interfacial >10% loss in net output, for oxyfuel and
effects. The microstructures of the membrane surfaces and bulk are closely precombustion power plants using pure
related to the interfacial oxygen exchange kinetics and bulk diffusion kinetics. oxygen produced by cryogenic distillation
Therefore, the permeability and stability of oxygen-permeable membranes with factories.[1] Other than cryogenic distil-
a single-phase structure and a dual-phase structure can be adjusted through lation, pressure swing adsorption and
polymer membranes are the commercial-
their microstructural and interfacial designs. Here, recent advances in the
ized technologies for oxygen separation
development of oxygen permeation models that provide a deep understanding from air, but they are inefficient for the
of the microstructural and interfacial effects, and strategies to simultaneously production of oxygen with purity higher
improve the permeability and stability through microstructural and interfacial than 95%.
design are discussed in detail. Then, based on the developed high-performance Recently, a production method based
on mixed ionic–electronic conducting
membranes, highly effective membrane reactors for process intensification
(MIEC) membranes, which can produce
and new technology developments are highlighted. The new membrane high-purity oxygen with higher energy
reactors will trigger innovations in natural gas conversion, ammonia synthesis, efficiency and lower investment cost than
and hydrogen-related clean energy technologies. Future opportunities and the cryogenic distillation technology, has
challenges in the development of oxygen-permeable membranes for oxygen gained much attention from academic and
separation and reaction–separation coupling are also explored. industrial communities.[2] In contrast to
the physical permeation behavior occur-
ring on microporous membranes and
polymer membranes, this mechanism is
1. Introduction an electrochemical permeation of oxygen through MIEC mem-
branes. This behavior demonstrates that the permeation selec-
Oxygen is an important industrial feed gas that is utilized in tivity of MIEC membranes toward oxygen is 100%. Although
many industries, such as chemical industries, ferrous met- the MIEC membranes are dense ceramics, their oxygen perme-
allurgy, glass making, and environmental protection. If the ability is comparable to that of microporous membranes and
energy consumption for pure oxygen separation can be signifi- is much higher than that of polymer membranes. In the past
cantly reduced, then not only will the production costs of the 20 years, the United States Department of Energy and Air Prod-
abovementioned processes be decreased remarkably but also ucts & Chemicals, Inc. (APCI) have invested hundreds of mil-
less pollution and CO2 will be discharged into the environment. lions of dollars to commercialize MIEC ceramic membranes to
With the development of oxyfuel and precombustion technolo- produce pure oxygen.[3] This revolutionary new technology can
gies for power generation and CO2 capture, the demand for significantly reduce the production cost (>30%) compared to
that of the current cryogenic distillation technology.[3,4]
Macroscopically, the oxygen permeation process includes
Prof. X. Zhu, Prof. W. Yang
three steps, i.e., oxygen surface exchange on the feed side,
State Key Laboratory of Catalysis
Dalian Institute of Chemical Physics ambipolar transport of oxygen ions and electrons across the
Chinese Academy of Sciences membrane bulk (including the grains and grain boundaries
Dalian 116023, China (GBs)), and oxygen surface exchange on the permeation side,
E-mail: yangws@dicp.ac.cn as illustrated in Figure 1a. On the microscopic scale, the oxygen
The ORCID identification number(s) for the author(s) of this article exchange kinetics on the gas–solid interfaces and the diffusion
can be found under https://doi.org/10.1002/adma.201902547.
kinetics across the grain boundaries, as shown in Figure 1b,
DOI: 10.1002/adma.201902547 cannot be neglected. In addition to their use in the separation

Adv. Mater. 2019, 1902547 1902547  (1 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

of pure oxygen from air, MIEC membranes are applied as reac-


tors in which the reaction and separation processes can be Xuefeng Zhu is a professor of
coupled to simplify the process and save energy, as shown in Dalian Institute of Chemical
Figure 1c. In MIEC membrane reactors, catalytic reactions are Physics (DICP), Chinese
integrated with the oxygen separation process. For an oxidation Academy of Science (CAS),
reaction with a large negative free energy, the deep oxidation China. He received his Ph.D.
reactions limit the selectivity of the target product. MIEC mem- degree from DICP, CAS in
branes control the oxygen input into the reactor, which results 2007, and then worked in
in a low oxygen partial pressure in the membrane reactor, so Arizona State University
a high selectivity for the target product can be obtained by (USA) as a postdoctoral
avoiding the deep oxidation. For a reducing reaction with posi- researcher in 2008. His
tive free energy, such as water decomposition for the production research interests involve
of hydrogen, its conversion is limited by the thermodynamic mixed conducting ceramic
equilibrium. The MIEC membrane selectively removes oxygen membranes for gas separation, catalytic membrane
from the reactor, thus shifting the reaction equilibrium to the reactors for process intensification, solid oxide cells, and
product side. The requirements for MIEC membranes used for electrocatalysis for energy conversion and storage.
chemical reactions in membrane reactors depend on the reac-
tion conditions. The permeation flux, permeation stability, and Weishen Yang is a chair
structural stability under catalytic reaction conditions are the professor of Dalian Institute
most important performance indicators for the MIEC mem- of Chemical Physics (DICP),
branes. Therefore, many studies have focused on developing Chinese Academy of Sciences
MIEC membranes that can satisfy the operational conditions of (CAS), China. He received his
catalytic reactions. Ph.D. from CAS in 1990. As a
There are several excellent reviews and books that compre- visiting scholar, he worked at
hensively discuss the progress in the study of MIEC membranes Birmingham University (UK)
in recent years.[5,6] In these studies, the basic principles, mate- in 1989, and the University
rial development, membrane reactor properties, experimental of Southern California
methods, and theoretic models were well-introduced. Compo- (USA) in 2001. He works
sition change (or element-doping) is the most commonly used on the rational design and
strategy to develop new membrane materials. However, this molecular-level engineering of functional nanomaterials
strategy does not always effectively improve both the stability for applications in catalysis and separation (e.g., inorganic
and permeability of MIEC membranes. Recently, development membranes).
of the microstructural and interfacial design of MIEC mem-
branes was suggested as an important strategy to enhance the
membrane performance, but this approach has not been thor- permeability of membranes can be improved simultaneously.[8–10]
oughly discussed in a review article. With the help of theoretic In this work, we describe the recent efforts to build models to
model studies, the microstructural and interfacial effects can be aid in the understanding of the important effects of the surface
well described.[7] With the advanced strategies, the stability and and grain boundaries on the stability of the membranes, design

Figure 1.  a) Macroprocess of oxygen permeation. b) Micromechanism of oxygen exchange on the gas–solid interfaces, oxygen transport in the crystals
and across the grain boundaries of a MIEC membrane. c) The reaction–separation coupling in MIEC membrane reactors.

Adv. Mater. 2019, 1902547 1902547  (2 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

novel MIEC membrane materials for oxygen transport and cata- transport characteristics of the electrons (or holes) and oxygen
lytic reactions, and utilize the developed membranes to construct ions (or oxygen vacancies) in a given region are constant regard-
membrane reactors for new separation and reaction processes less of the location and chemical potential of oxygen. 2) The
with high efficiencies. The goal of this study is to understand diffusion of the gaseous oxygen is fast enough for the concen-
the permeation mechanism; design MIEC membrane materials tration polarization to be ignored. 3) All the steps of the oxygen
with improved performance by controlling the bulk lattice, grain permeation are under isothermal conditions, and the law of
boundaries and surface structure; and to develop new reaction– mass action is applicable to the oxygen exchange reactions on
separation couplings for process intensification. Recent develop- the gas–solid interfaces. A detailed deduction process is shown
ments indicate that the technology based on MIEC membranes in the literature.[7] The general transport equation, including
is promising for the production of pure oxygen and for process the limitation of the bulk ambipolar diffusion and interfacial
intensification via reaction–separation coupling. exchange, is expressed as

1 1 1 1
jO2 = ∆µOtot2 = 2 2 tot ∆µOtot2 (1)
2. Mechanism and Model of Oxygen Permeation 4 F r ′ + r + r ′′
2 2 b
4 F r

For separation processes that are dominated by the oxygen where r′, rb, and r″ are the permeation resistance through the
ionic–electronic (or oxygen vacancy-hole) ambipolar diffusion interface I, bulk region and interface II, respectively; rtot is the
across the MIEC membrane bulk, the Wagner equation can total permeation resistance across the membrane; F is the Far-
be used to determine the influence of the temperature, oxygen aday constant; and ∆µOtot2 is the oxygen chemical potential gra-
partial pressures and membrane thickness on the oxygen per- dient across the membrane. The interfacial resistances depend
meation flux. However, when the interfacial exchange reactions on the oxygen partial pressure, i.e.,
are fast enough or the membrane thickness is thick enough,
r ′ = r0′ (POI2 /P0 )
−0.5
the bulk ambipolar diffusion may be the rate-limiting step. In (2)
most cases, the interfacial oxygen exchange reactions on the two
sides are involved in controlling the permeation process. Several
r ′′ = r0′′ (POII2 /P0 )
−0.5
model equations have been developed to illustrate the mixed (3)
control of the permeation process by the oxygen exchange
reactions on the interfaces and the bulk ambipolar diffusion. where POI2 and POII2 are the oxygen partial pressures of the
Bouwmeester et al. introduced an important concept, the char- feed and permeation sides, respectively, P0, r′0, and r″0 are the
acteristic thickness, to modify the Wagner equation to fit the oxygen partial pressure at 1 atm and the permeation resist-
oxygen permeation process.[11] No clear relationship between the ance constants of the interfaces I and II, respectively, at oxygen
permeation flux and the oxygen partial pressure was provided partial pressure of 1 atm, and r′0, r″0, and rb are the kinetic
in the modified equation. Jacobson and coworkers analyzed the parameters that are obtained by fitting the experimental data to
net oxygen interfacial exchange flux on the surfaces of MIEC Equations (1)–(3). This permeation model has been success-
membranes and developed a general equation.[12] This equation fully used to fit the permeation data of perovskite-type MIEC
is too complicated to exhibit a clear dependency relationship membranes (Figure 2b) and dual-phase (DP) membranes.[15]
between the oxygen permeation flux and the experimental con- The kinetic parameters, i.e., the oxygen diffusion coefficient
ditions (i.e., temperature, membrane thickness, and oxygen (DO) and the interfacial exchange coefficients (k′ and k″), are
partial pressure). Lin et al. proposed a general equation by con- derived from the three permeation resistances
sidering the diffusion of charged species in the bulk region of
membranes and the oxygen exchange reactions on the gas–solid RT L
DO ≈ Dambi = (4)
interfaces.[13] Xu et al. simplified Lin’s model by introducing 4F 2c Ob2− r b
several assumptions to fit perovskite MIEC membranes.[14] This
simplified model is widely applied in permeation modeling,
RT 1
but it is only applicable to a few membranes with low oxygen k′ = (5)
4F 2c ′O2− r ′
vacancy concentrations, and it cannot give a unique solution to
fit the model parameters.[15] A proper model is important for
understanding the permeation mechanism of a MIEC mem- RT 1
k ′′ = (6)
brane. With the model equation, researchers can evaluate the 4F 2c ′′O2− r ′′
observed relationships between the oxygen permeation flux and
the experimental conditions, which can provide guidance for the where Dambi is the ambipolar diffusion coefficient; if the ionic
development of membranes and catalysts to accelerate the bulk conductivity is much lower than the electronic conductivity for
diffusion and the interfacial exchange reactions, respectively, the MIEC material, DO ≈ Dambi; R is the gas constant; T is the
and can aid in the understanding of the degradation mecha- temperature; L is the membrane thickness; c Ob2−, c O′ 2−, and c O′′2− are
nisms and approaches to improving the stability. the concentrations of oxygen ions in the membrane bulk, inter-
We built an oxygen permeation model by considering the face I, and interface II, respectively. The kinetic parameters of
oxygen ionic and electronic transport across the membrane MIEC materials are usually determined through tracer diffusion
bulk and the two gas–solid interfaces, as illustrated in Figure 2a. or electrical conduction relaxation techniques. The kinetic
To establish the model, three assumptions were made. 1) The parameters obtained via tracer diffusion are in an equilibrium

Adv. Mater. 2019, 1902547 1902547  (3 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 2.  a) Schematic graph of the oxygen permeation process for Zhu’s model. a) Reproduced with permission.[15] Copyright 2017, American Insti-
tute of Chemical Engineers, published by John Wiley and Sons. b) Comparison of the model data calculated by Zhu’s model with the experimental
data for Ba0.5Sr0.5Co0.8Fe0.2O3−δ (BSCF). The solid circles represent the experimental data and the colored surfaces represent the calculated results.
c) Permeation fluxes of the BCF membranes coated with various catalyst layers. c) Reproduced with permission.[8] Copyright 2016, American Institute
of Chemical Engineers, published by John Wiley and Sons. d) Changes in the permeation resistances with operation times for the BSCF membrane.
Reproduced with permission.[9] Copyright 2015, American Institute of Chemical Engineers, published by John Wiley and Sons.

state, and those obtained via the electrical conduction relaxa- The distribution of the three permeation resistances can dis-
tion are in a nonsteady state. MIEC membranes are used for tinctly reveal which step dominates the oxygen permeation pro-
oxygen permeation under steady state conditions; thus, the data cess. The rate-determining step may change with operational
obtained in equilibrium or a nonsteady state cannot reflect the temperature, oxygen partial pressure of both sides, and the thick-
real kinetics of the oxygen permeation under the steady state ness of the membranes. The resistance distribution map can be
conditions. However, the kinetic parameters obtained from used to easily determine how to improve the permeation flux and
modeling the permeation data to the equations were close to the to what extent it can be improved.[7–9] Another important applica-
values for the membrane in a real working steady state. tion of the permeation model is to reveal the degradation mecha-
The activation energy of each step can be obtained from the nism of membranes. The increases in the three permeation
model analysis, which aids in the evaluation of the kinetic pro- resistances may result in performance degradation. The perfor-
cesses of the bulk diffusion and interfacial exchange. For example, mance degradation of the BSCF membrane was regarded as the
the permeation flux of the 0.5 mm thick BaCe0.05Fe0.95O3−δ (BCF) result of the phase transformation at intermediate-to-low tem-
membrane was slightly improved (<6%) by applying porous perature (ILT) in the literature.[5,6,16] However, the model analysis
Sm0.5Sr0.5CoO3−δ (SSC) layers to both surfaces, and the apparent showed that the phase transformation in the membrane bulk
permeation activation energy was almost unchanged (Figure 2c). made a small contribution to the degradation, but the increase
A superficial but reasonable explanation to the above results is that in the interfacial exchange resistance predominated, as shown in
the oxygen permeation through the 0.5 mm thick BCF membrane Figure 2d.[9] A detailed discussion of the degradation mechanism
was controlled by the bulk diffusion. However, the model analysis of MIEC membranes is presented in a following section.
shows that the bulk diffusion resistance is only 30–40% of the total Each model applies to specific ranges, since models are devel-
resistance. The SSC layer decreased the interfacial exchange resist- oped based on appropriate assumptions. In Zhu’s model, the
ances but simultaneously increased the bulk diffusion resistance transport properties of the electrons/holes and oxygen ions/
due to the reaction between the SSC and BCF.[8] When the porous vacancies in a given zone were assumed to be constant, regardless
catalyst layer was changed from SSC to BCF, the permeation flux of their positions and the oxygen chemical potentials.[7] There-
of the 0.5 mm thick BCF membrane was improved by 30–45%. fore, Zhu’s model is applicable to MIEC materials whose oxygen

Adv. Mater. 2019, 1902547 1902547  (4 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

vacancy concentrations have a weak dependence on the oxygen a perovskite lattice. Therefore, small A- and B-site cations are
partial pressure, such as BSCF, BCF, and DP membranes. To wanted to obtain a high oxygen ionic conductivity for perovskite
model the oxygen permeation process for oxygen separation by MIEC materials. The FV is defined by the difference between
MIEC membranes, the model equations can be modified to show the volume of a single cell and that of the overall ions, and it
the process taking place in a MIEC membrane reactor. The inter- decreases as the B-site cationic radii increase. A larger FV pro-
facial exchange resistances are replaced by the reaction polariza- vides more room for lattice relaxation; therefore, increasing the
tion resistances, which can be derived from the equation for the FV facilitates the diffusion of the oxygen ions in the membrane
catalytic reaction rates.[17] However, the reaction conditions need bulk and reduces the transport activation energy of the oxygen
to be in accordance with the assumptions used to derive these ions. The ABE indicates the level of difficulty to break the MO
models to avoid incorrect results and conclusions. bonds in the perovskite structure. Cations located at the A and
B-sites with low oxidation states enhance the conductivity of the
oxygen ions; however, low-valent cations with larger ionic radii
3. Perovskite MIEC Membrane Materials decrease the critical radius. Thus, changes in the critical radius,
free volume, and ABE cannot simultaneously favor an increase
Perovskite oxides with the formula ABO3 are typical MIEC in the ionic conductivity.
membrane materials. The critical radius (rc), free volume (FV), In addition, the stability of the perovskite structure should
and average bonding energy (ABE) of the MO bonds are fre- be considered during the design of MIEC materials. The sta-
quently considered when discussing the relationship between bility of a perovskite oxide depends on two factors, i.e., ther-
the structure and oxygen ionic diffusion for perovskite MIEC modynamics and kinetics. The ABE and tolerance factor are
materials. In a cubic perovskite oxide, the oxygen ionic diffusion the two most important factors influencing the thermodynamic
follows the ⟨110⟩ direction via a trigonal window constructed by stability of a perovskite oxide. High ABE values correspond to
a B-site cation and two A-site cations (Figure 3a). The apothem high stability but low mobility of oxygen ions in the perovskite.
of the trigonal window, which is known as the “critical radius,” For a tolerance factor close to 1, the BO6 octahedron has less
decreases as the A- and B-site cations become larger, assuming distortion in the perovskite structure, and the corresponding
that the perovskite maintains its cubic structure.[5,6] The devi- perovskite therefore has a higher stability. The kinetic stability
ations in the positions of the cations from the normal lattice of a perovskite MIEC membrane is related to the diffusion
positions under the squeezing action of oxygen ions allow the rate of the A- and B-site cations in the material. As a chemical
oxygen ionic diffusion across the window. This kind of lattice potential gradient of oxygen is applied on a MIEC membrane,
relaxation is temperature-dependent, so it is intimately con- not only oxygen ions but also cations are forced to direction-
nected to the diffusion activation energy of the oxygen ions in ally move from one side to the other side. Since the diffusion

Figure 3.  a) Diffusion of oxygen ions in a perovskite oxide. b) Oxygen permeation fluxes of BaxSr1−xCo0.8Fe0.2O3−δ membranes with different Ba contents.
b) Reproduced with permission.[21] Copyright 2001, Elsevier B.V. c) Oxygen permeation flux of an asymmetric BSCF membrane. Reproduced with per-
mission.[23] Copyright 2011, Elsevier B.V. d) Effect of the cooling–heating cycle on the oxygen permeation fluxes of the BaFeO3−δ and BaCe0.05Fe0.95O3−δ
membranes. Reproduced with permission.[29] Copyright 2006, Elsevier B.V. e) Possible oxygen ion transport paths. f) Oxygen ionic migration barrier
energies of oxygen ions via different paths in a Ba8Fe7GdO24 supercell. e,f) Reproduced with permission.[27] Copyright 2016, Royal Society of Chemistry.

Adv. Mater. 2019, 1902547 1902547  (5 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

coefficient of oxygen ions is typically five orders of magnitude In addition to the production of pure oxygen from air,
higher than that of cations,[6] there is no observable change another important application of MIEC membranes is the inte-
in the elemental compositions of the membrane surfaces in gration of catalytic reactions with oxygen separation into mem-
a short-term oxygen permeation test. However, the inevitable brane reactors. The common behaviors of these reactions are
kinetic demixing of cations is expected after a long-term opera- inputting or producing strong reducing gases at elevated tem-
tion. The rate of kinetic demixing largely depends on the dif- peratures. Therefore, membrane materials that can sustain the
fusion rates of the cations. The diffusion rates of the cations perovskite structure under a reducing atmosphere are needed
depend on the temperature, atmosphere, chemical potential for these reactions. BaFe1−xMxO3−δ (M = Y, Zr, Ce, Gd, Sm, etc.)
gradient of oxygen, oxygen exchange rates at the interfaces, is another typical series of cobalt-free MIEC perovskite mem-
and membrane microstructure. Usually, the lower ABE value branes.[25–29] The as-prepared BaFeO3−δ has a complex phase
for a perovskite material corresponds to higher diffusion coef- composition at room temperature, comprising monoclinic,
ficients for both the oxygen ions and the cations. Therefore, it rhombohedral, orthorhombic, hexagonal, and cubic phases.
is difficult to combine the excellent permeability and stability The phase transformation to cubic at 800 °C is accompanied
simultaneously in a membrane material. Thus, there is a need by a sharp increase in the electronic and ionic conductivities. In
for further study of the design of MIEC materials according to addition, the BaFeO3−δ maintains its perovskite structure even
the target applications. at oxygen partial pressure of 10−17 atm and 827 °C.[29] We found
that the most effective strategy to maintain the cubic high-tem-
perature phase at room temperature is to dope large and high-
3.1. Improvement of Stability and Permeability valent cations in the B-site with a low doping amount (5–10%),
by Adjusting the Crystal Structure as shown in Figure 3d. Although doping high-valent cations
in the B-site increases the ABE value, the oxygen permeation
The cobalt-based perovskite oxides, in particular flux is significantly enhanced, and the activation energy of per-
SrCo0.8Fe0.2O3−δ, show high oxygen permeability. However, for meation decreases to as low as 30–40 kJ mol−1,[26,27,29] which is
oxygen partial pressures lower than ≈0.1 bar and temperatures even lower than that of BSCF. This abnormal phenomenon was
lower than ≈790  °C, the cubic SrCo0.8Fe0.2O3−δ is not stable, explained by a first-principles calculation.[27] A 2 × 2 × 2 super-
and it transforms to an orthorhombic brownmillerite structure cell of Ba8Fe7GdO24 with an Fe atom replaced by a Gd atom
(A2B2O5) with ordered oxygen vacancies.[18–20] Partial substi- was constructed to fit the first-principles calculation on the
tution of Sr by La in the A-site is effective to enhance the sta- Gd-doped BaFeO3−δ, as shown in Figure 3e. Based on the dif-
bility of the membrane materials. Similarly, doping high-valent ference in the distances from Gd cations, the anions/vacancies
and less reducible cations (Ti4+, Zr4+, Nb5+, etc.) in the B-site in the supercells have three types, i.e., O1/VO1, O2/VO2, and
is effective to maintain the cubic phase. However, the stability O3/VO3. A fast path of oxygen ion diffusion in the supercell was
increase is accompanied by a permeability decrease. Because observed such that the migration barrier for O2 to VO1 was as
the substitution reduces the oxygen vacancy concentration and low as 0.17 eV (Figure 3f), although the migration energies of
free volume, the ABE value increases as well. All these changes oxygen from the O1 site to the adjacent VO1 (1.62 eV) and VO2
decrease the oxygen diffusion coefficients of materials. (0.99 eV) were higher than that in the cubic BaFeO3 pristine
However, we found that the stability and permeability lattice (0.91 eV). This result indicates that the oxygen migration
were simultaneously improved by Ba doping in the Sr-site in was more difficult for those near the large, high-valent doped
SrCo0.8Fe0.2O3−δ. The BaO bond had a similar bond energy cations and less difficult for those far from the dopants. Too
to that of the SrO bond, so the partial substitution of Sr by much dopant would inhibit the oxygen diffusion because of the
Ba had little effect on the ABE value. The oxygen vacancy con- dopant binding to the oxygen anions. The BaFe1−xMxO3−δ mate-
centration was increased after the Ba doping. As expected, rials even maintained the cubic perovskite structure after being
the Ba0.5Sr0.5Co0.8Fe0.2O3−δ (BSCF) exhibited a higher oxygen exposed to 5% H2/Ar at 900 °C. However, the perovskite struc-
permeation flux than did the SrCo0.8Fe0.2O3−δ (Figure 3b).[21] tures of the Ba0.5Sr0.5Co0.8Fe0.2O3−δ and La0.5Sr0.5Ga0.3Fe0.7O3−δ
Although the BSCF membrane was first reported by Yang’s were destroyed after reduction under the same conditions.[29]
group in 2000,[22] it has the highest known oxygen permeability Therefore, the BaFe1−xMxO3−δ materials are promising for con-
to date. A high oxygen permeation flux of 50 mL cm−2 min−1 structing membrane reactors if the reaction atmosphere con-
was obtained on a BSCF membrane with a dense separation tains little or no CO2.
layer of 70 µm at 900 °C (Figure 3c).[23] The incorporation of
Ba in the A-site led to a decrease in the trapped oxygen vacan-
cies in the lattice, which inhibited the order–disorder transi- 3.2. Degradation Induced by Microstructural
tion occurring on the SrCo0.8Fe0.2O3−δ.[18] BSCF has widely and Interfacial Changes
gained the attention of researchers from not only the field of
MIEC membranes but also the fields of solid oxide fuel cells The high operating temperature (700–1000 °C) is a distin-
and electrocatalysis because of its outstanding catalytic activity guishing feature of MIEC membranes. With the improvements
toward the oxygen activation reactions. Long-term operation of in the fabrication technology for asymmetric membranes, the
BSCF membranes has been proofed up to 17 000 h.[24] Thus, dense layer for oxygen permeation has been reduced from
BSCF is considered a promising material for pure oxygen pro- the typical 0.5–2 mm to >0.01 mm. As a result, it is pos-
duction due to its high permeability and stability at elevated sible to achieve commercial-valued permeation fluxes at ILT
temperatures. (350–700  °C). The operation of MIEC membranes at ILT has

Adv. Mater. 2019, 1902547 1902547  (6 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

many significant advantages compared with its high-temper- 3.2.1. Gas–Solid Interfacial Degradation
ature operation. The creep rate drops sharply with decreases
in the temperature since the activation energy of perovskite Recently, we reported that nonmetallic impurities containing
ceramic creep is as high as 300–500 kJ mol−1.[6,30] The creep sulfur and silicon elements, would diffuse from the membrane
induced by the pressure loading on a MIEC membrane would bulk region to the gas–solid interfaces and lead to a significant
result in a macroscopic deformation at elevated temperatures decrease in oxygen reduction and evolution rates on both of
during a long-term on-stream operation;[31] thus, failure of the the gas–solid interfaces.[8,33–35] Three perovskite membranes,
MIEC membranes is inevitable. However, reducing the run- BaCe0.1Co0.4Fe0.5O3−δ (BCCF),[36] BaZr0.2Co0.4Fe0.4O3−δ (BZCF),[37]
ning temperature to ILT would greatly reduce the creep rate and BaCo0.7Fe0.22Nb0.08O3−δ (BCFN),[38] reported by different
and thus prolong the lifetime of the MIEC membranes. Less authors, have stable phase structures for oxygen permeation but
energy is needed to heat air to high temperatures and less show quick performance degradation at 600 °C (Figure 4a,b).[33,34]
expensive stainless steel and other materials can be used to For example, a significant degradation of the permeation flux
construct the membrane separation unit if the MIEC mem- by 80% after operating for 480 h at 600 °C was observed on the
branes are operated at ILT. In addition, the great challenge in BCCF membrane. After a long-term stability test, BaSO4 crystals
maintaining perfect sealing of MIEC membranes under large were observed on the surfaces of all the membranes, and the
pressure gradients (<2 MPa) and long running periods (>5 amount of BaSO4 crystals on the permeation-side surface was
years) becomes less difficult at ILT. Because of these advan- more than that on the feed-side surface (Figure 4c,d). In addi-
tages, MIEC membranes used for pure oxygen production tion to BaSO4, Zr(SO4)2 was detected on the permeation-side
have become more competitive compared with the traditional surface of a BZCF membrane. The chemical precursors, i.e.,
technologies. nitrates, oxides and carbonates, were found to be the sources
However, the quick degradation of the permeation flux at of the sulfur impurities after a series of characterizations and
ILT is inevitable for MIEC membranes. This problem occurs analyses. The elemental sulfur content was ≈ 20 ppm in the
in many of the high-performance membranes, such as BSCF, perovskite oxide, even though high-purity chemical precursors
La1−xSrxCo1−yFeyO3−δ, (Ba,Sr)Co1−x−yFexMyO3−δ, and BaFe1−xMx were used. However, this low sulfur content caused a significant
O3−δ. In general, the performance degradation becomes more degradation of the permeation flux at 600 °C. Silicon impurities
severe at higher temperatures for materials. The abnormal commonly appearing in the starting chemical precursors and the
phenomenon occurring on the MIEC membranes at lower ceramic vessel, and dust were found in the perovskite materials
temperatures indicates the degradation mechanisms are dif- as well. Elemental silicon enriches to the surfaces of membranes,
ferent from that described by the conventional wisdom. Three showing behavior similar to that of elemental sulfur (Figure 4a,f).
mechanisms were suggested to explain the degradation at ILT Therefore, membranes containing silicon impurities have degra-
for perovskite MIEC membranes, i.e., oxygen vacancy ordering, dation mechanisms similar to those containing sulfur impurities.
phase transformation and kinetic demixing. Kruidhof et al. Slater and coworkers found that oxyanions, such as sulfate,
treated La0.6Sr0.4CoO3−δ at 700 °C for 100 h and found ordered silicate, and phosphate groups, are good dopants to stabilize
oxygen vacancies in small microdomains (5–20 nm).[19] An the hexagonal perovskite phases and orthorhombic brown-
endothermic peak was observed at 910 °C during the heating millerite phases to cubic perovskite phases with low doping
of a disk-type La0.6Sr0.4CoO3−δ in air, indicating there was a amounts (<5 at%).[39,40] They reported that the perovskite lat-
change from an ordered defect structure at lower tempera- tice allowed high doping amounts of the silicate group of up
tures to a disordered defect structure at higher temperatures. to 15 at%.[40] High temperature calcination in air was required
Phase transformations of membranes such as SrCoO3−δ, to ensure the incorporation of these oxyanions into the perov-
Ba0.5Sr0.5Co0.8Fe0.2O3−δ, and BaFeO3−δ, from the cubic perov- skite lattice. However, the oxyanions tended to move outside
skite phase to phases with ordered oxygen vacancies result of perovskite lattices, as the as-sintered membranes were
in their degradation at lower temperatures. The performance annealed in air or helium at 600 °C. More impurities were
degradation of phase-stable MIEC membranes usually was found on the samples treated in helium.[8,33] The above results
regarded following the kinetic demixing mechanism. Kinetic indicate that at a low temperature, the chemical potentials of
demixing of cations is caused by the chemical potential gra- oxyanions in perovskite lattices are greater than those on the
dient of oxygen through the MIEC membrane and results in gas–solid interfaces, and the chemical potentials in air are
a change of phase structure in both the bulk and the surfaces. higher than those in helium. The membrane was operated
The degradation of many membranes with stable perovskite with one side fed with air and the other side fed with helium.
structures has been attributed to the kinetic demixing mecha- Impurities were enriched on the surfaces of both sides during
nism. The kinetic demixing process is related to the activated the initial stage but only on the surface of the permeation side
diffusion of cations. Thus, the degradation rate is probably after a long-term operation. The migration of oxyanions is indi-
faster at high temperatures than at low temperatures. Demixing rectly related to the oxygen chemical potential gradient. The
occurs when an oxygen chemical potential gradient is applied chemical potential of the oxyanions at the gas–solid interface is
across a membrane, but alkaline earth element segregation a function of the oxygen partial pressure. At the initial stage of
occurs when the perovskite materials are treated in an oxi- the oxygen permeation, the chemical potential of the oxyanions
dizing atmosphere.[32] Therefore, the above three mechanisms in the membrane bulk was higher than those on the air–solid
cannot explain the degradation of most MIEC membranes with and helium–solid interfaces, so the diffusion of oxyanions to
stable crystal structures, i.e., no ordering of oxygen vacancies both sides was observed with the air side enriched with fewer
and phase transformation at ILT. impurities than the helium side (Figure 4g). However, after a

Adv. Mater. 2019, 1902547 1902547  (7 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 4.  a) Time-dependent oxygen permeation fluxes of the uncoated BCCF perovskite membranes and the same BCCF perovskite membranes
coated by SSC porous catalyst layers at 600 °C. b)Time-dependent oxygen permeation fluxes of BZCF (red) and BCFN (blue) membranes with both
surfaces coated and uncoated by SSC porous catalyst layers at 600 °C. c) Feed-side surface. d) Permeation-side surface containing sulfur impurities
of the BCCF membrane after 480 h of operation at 600 °C. e) Feed-side surface. f) Permeation-side surface containing silicon impurities of the BCCF
membrane after 477 h of operation at 600 °C. h) Top view and cross-section view of the porous SSC catalyst layer. The arrows mark BaSO4 grains.
g) Schematic graphs of the chemical potential of elemental sulfur/silicon and the transport of sulfate/silicate group across the membranes under
different atmospheres. Data from ref. [33] and ref. [34] are replotted in (a) and (b). c,d,h) Reproduced with permission.[33] Copyright 2013, Wiley-VCH.
e,f,g) Reproduced with permission.[34] Copyright 2015, Elsevier B.V.

long-term operation, the chemical potentials of the oxyanions At low temperatures, sulfates tend to form on surfaces, so sul-
on both surfaces and in membrane bulk were equal, indicating fates accumulate on the membrane surfaces and result in a sig-
a net diffusion of oxyanions from the air side to the permea- nificant decrease in the oxygen exchange rates, while at high
tion side; as a result, the content of impurities decreased to temperatures, the formation rate is slow, so no sulfate accumu-
near zero on the air side (Figure 4e–g). lates and the membrane can keep stable oxygen permeation.[33]
At high temperatures, oxyanions migrate from the mem- For membranes containing silicon impurities, the stabiliza-
brane bulk to the two surfaces, but performance degrada- tion mechanism at high temperatures is attributed to the sin-
tion was not observed for the MIEC perovskite membranes. tering effect of the accumulated silicates on the membrane
The results of the thermodynamic calculations reveal that surfaces.[34] The resultant large silicate particles occupy only a
increasing the temperature inhibits the formation of sulfates. small portion of the membrane surfaces, so they have a weak

Adv. Mater. 2019, 1902547 1902547  (8 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

effect on the oxygen exchange rate on the gas–solid interfaces. exchange reactions at ILT. The permeation flux is close to
Thus, the different surface microstructures of the impurities 1 mL cm−2 min−1 at 600 °C under an oxygen partial pres-
induced by the different operating temperatures govern the sta- sure gradient of 21 kPa/2 kPa through a 0.5 mm thick BSCF
bility of the oxygen permeation. membrane. The permeation flux meets the frequently referred
The above interfacial degradation mechanism for perovskite threshold (>1 mL cm−2 min−1) for commercial applications.[41]
MIEC membranes with good phase stability indicates that a trace However, many studies have shown that the performance deg-
amount of oxyanions moves out of the membrane bulk to the radation occurs on the BSCF membrane at ILT due to the com-
surfaces controlling the permeation stability of the membranes. plex phase transformation. The 2H-type hexagonal perovskite
According to the degradation mechanism, there are three possible structure was formed after the sintered powder was treated for
solutions to sustain the permeation flux at ILT. The first is the pre- 240 h at 750 °C.[42] The real driving force of the phase transfor-
process control in which the oxyanion-containing impurities are mation from the cubic to the hexagonal perovskite structure is
removed by purifying the starting chemical precursors in advance the oxidation of Co2+ with a high spin (HS) state to Co3+ with a
and the process of membrane fabrication is strictly controlled low spin (LS) state, which corresponds to a significant decrease
to avoid introducing oxyanion-containing impurities during the in the cationic radii from 0.0745 nm (Co2+, HS) to 0.0545 nm
preparation of the membranes. Unquestionably, this process (Co3+, LS).[43] Müller et al. observed that the content of the hex-
would sharply increase the cost of MIEC membranes. The second agonal phase increased slowly with the treating time. In addi-
solution is the in-process control in which the migration of oxy- tion to the hexagonal phase, plate-like regions were observed
anions is blocked or slowed by changing the composition of the when the BSCF was annealed at 700–900 °C.[44] The plate-like
MIEC membranes. However, the migration path or mechanism regions contained three phases, i.e., the cubic perovskite phase,
of oxyanions in the perovskite lattice is still unclear. Therefore, hexagonal phase, and Ba3Co10O17 phase.[45] The Ba3Co10O17
there is no clear direction for the development of the materials. phase comprised a periodic arrangement of two (111)-planes
The third is the post process control in which the enriched oxy- of O octahedrons with a cubic stacking order and three planes
anion-containing impurities are removed from the membrane with a CdI2-type structure and two interface planes, as shown
surfaces to avoid the enrichment on the gas–solid interfaces. in Figure 5a. The generation of plate-like regions is more rapid
We proposed a simple method to remove oxyanion-containing than that of the hexagonal phase at the GBs of BSCF. Recently,
impurities from the membrane surfaces by changing the surface we found that the Ba3Co10O17 phase with a composition of
microstructures.[8,33,34] Porous layers with high catalytic activity approximately Ba0.45Sr0.07Co1.6Fe0.18O2.8 was formed at the GBs
toward oxygen exchange reactions were coated on both sides of after the BSCF membrane was operated for 500 h at 600 °C
the membranes, which cause the migration of oxyanions from the (Figure 5b,c), but we did not find the hexagonal phase.[10] The
membrane bulk into the porous layers; thus, the impurities accu- Ba3Co10O17 phase is not stable at temperatures above 800 °C
mulate not on the surfaces but in the porous layers (Figure 4h). and tends to decompose to Co3O4 or CoO (depending on the
SSC perovskite was used to create ≈10 µm thick porous layers on temperature) and the 2H-type hexagonal phase.[46] Therefore,
the two surfaces of a BCCF membrane. The flux of the SSC-coated the Ba3Co10O17 phase was observed only at ILT, while the Co3O4
BCCF membrane was only 5% lower than the initial value after (or CoO) phase and the 2H-type hexagonal phase were found
the 480 h on-stream process at 600 °C (Figure 4a). For the SSC- at 700–850 °C for the BSCF material. Müller et al. reported
coated BCFN and BZCF membranes, stable permeation fluxes that plate-like regions tend to form around the CoO grains.[45]
were observed during the 100 h operation at 600 °C (Figure 4b). However, another possible explanation to this phenomenon is
After the long-term operation, the postmortem analysis indicated the decomposition of the Ba3Co10O17 phase producing the CoO
that the sulfur and silicon impurities migrated to the outer skin of impurities.
the SSC layers. Therefore, no elemental sulfur or silicon impuri- From the above discussion, the BSCF phase transformation
ties remained on the surface of the membrane. The porous layers is different from those of SrCoO3−δ, BaFeO3−δ, SrCo0.8Fe0.2O3−δ,
had a much higher surface area than did the membrane surface, etc. For SrCoO3−δ and BaFeO3−δ, the phase transformation
so the enrichment of impurities did not significantly inhibit the from cubic to 2H hexagonal is a martensitic transformation
oxygen exchange reactions. The success in realizing stable fluxes that occurs in the temperature range of 800–900 °C in air.[35]
through the three perovskite MIEC membranes suggests that this This process is a diffusionless transformation with rapid rear-
method may extensively work well for the MIEC membrane fami- rangement of atoms and thus has a very low thermal activation
lies to prevent the permeation degradation at ILT. Therefore, the energy and an extremely fast rate. However, the generation of
third solution is the simplest and the most effective. However, a crystal nuclei requires the long-distance diffusion of the atoms
slight reduction in the catalytic activity toward the oxygen reduc- in the grain to reach the proper sites for the phase transforma-
tion and evolution reactions is inevitable when the oxyanion- tion of the BSCF, and then the crystal nuclei can grow to new
containing impurities are enriched in the porous layers. There- phases. Therefore, this process is a diffusion-controlled phase
fore, there was a weak degradation (<5%) in the permeation flux transformation. New phases are frequently observed at GBs,
during the initial stage.[33,34] because GBs are active sites for the generation of crystal nuclei
due to their loose arrangement of atoms and their large amount
of defects.
3.2.2. Phase Transformation at the Perovskite GBs A convenient method to stabilize the perovskite structure of
BSCF is the substitution of cations with high and stable oxidation
BSCF is the most attractive material because of its high states. However, a low doping amount cannot completely stabilize
ionic diffusion rate and high catalytic activity toward oxygen the perovskite structure, while a high doping amount leads to a

Adv. Mater. 2019, 1902547 1902547  (9 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 5.  a) Structure of Ba3Co10O17. b) BSCF membrane after oxygen permeation for 500 h at 600 °C. “R” denotes the trigonal phase precipitated at the GBs.
c) Elemental mapping of the area marked in the red frame in (b). d) Time-dependent oxygen permeation through Zr-, Nb-, Y-, Ce-doped BSCF membranes
at 600 °C. e) Zr-doped BSCF membrane after operating for 432 h at 600 °C. f) GB of Y-doped BSCF with nanoparticles after the 500 h oxygen permeation
testing. g) GB of Y-doped BSCF without nanoparticles after the 500 h on-stream process. h) GBs of the as-prepared Ce-doped BSCF membrane. i) Time-
dependent bulk permeation resistances of the Ce-doped BSCF membrane at 600 °C. j) Oxygen permeation flux of the asymmetric Ce-doped BSCF mem-
brane. a) Reproduced with permission.[6] Copyright 2017, Springer Nature. b–j) Reproduced with permission.[10] Copyright 2015, American Chemical Society.

significant reduction in the oxygen permeability. In addition to the In the above equation, the first term on the right side is the
lattice doping method to enhance the perovskite stability, a new inherent grain boundary energy, and the second term is the
strategy was proposed to prevent a phase transformation in BSCF Zener pinning energy. To prevent the phase transformation at
by introducing nanoparticles at the GBs.[10] A thermodynamic the GBs, the grain boundary energy must be larger than zero,
analysis indicated that the appearance of nanoparticles at the GBs which gives the following relationship
produces a Zener pinning pressure. Thus, after the introduction
of nanoparticles at the GBs, the grain boundary energy is f VRg
sin2θ ≥ 1 (8)
rNP
3γ Vm 3γ Vm f V sin2θ 3  1 f 
∆GGB = − + = − γ Vm  − V sin2θ  (7)
2Rg 2rNP 2  R g rNP  The maximum value of sin2θ appears at θ = π/4; there-
fore, for a fVRg/rNP greater than one, the grain boundary
where γ is the grain boundary energy per unit area, Vm is energy is reduced to zero at equilibrium. Thus, the driving
the molar volume of the material, Rg is the grain radius force of the heterogeneous nucleation at GBs is reduced after
of the material, fV is the volume fraction of the nanoparticles the introduction of nanoparticles at the GBs. The increase in
in the material, rNP is the radius of the nanoparticles, and θ is the volume fraction of nanoparticles results in an increase in
the angle between the grain boundary surface and the surface the particle size during the sintering at high temperatures.
at the point where the grain joins the nanoparticle (0 ≤ θ ≤ π/2). In addition, the oxygen permeation flux is reduced when there is

Adv. Mater. 2019, 1902547 1902547  (10 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

a high volume fraction of nanoparticles in membranes. The were observable at the GBs. The oxygen permeation flux
grain size of BSCF membranes typically changes in the range was stable during a 500 h test, as shown in Figure 5c. No
of 10–50 µm, and the increase in the Rg decreases the strength Ba3Co10O17 phase appeared at the GBs, but many BaCeO3
and creep resistance of the ceramics.[6] Therefore, it is effective nanoparticles were found at the GBs with sizes of 50–100 nm
to reduce the size of the nanoparticles to obtain fVRg/rNP ≥ 1. and at crystal edges with sizes of 100–200 nm. The depend-
A kinetic analysis indicates that the heterogeneous nuclea- ence of the bulk diffusion resistance on time was determined
tion and new phase growth at GBs obey a stepwise growth by using Zhu’s model to provide an analysis of the inhibition
mechanism. As nanoparticles are introduced at the GBs, the of the phase transformation by the nanoparticles at the GBs.
driving force of the step motion includes two parts, i.e., the The resistance to bulk diffusion remained constant at 600 °C
inherent grain boundary pressure and the Zener pinning pres- (Figure 5i). The volume fraction of the BaCeO3 nanoparticles
sure. Thus, the motion velocity of the steps is was 4%, and the average size of the nanoparticles was 100 nm.
The average size of the BSCF grains was 15 µm. Therefore, the
3γ km 3γ kmf V sin2θ 3γ km  1 f  value of fVRg/rnp was 6, which was higher than the critical value
v= − =  − V sin2θ  (9) of 1. Therefore, the heterogeneous nucleation and Ba3Co10O17
2hR g 2hrNP 2h  R g rNP 
phase growth can theoretically be inhibited. The zig-zag nature
of the GBs (Figure 5h) supports the stepwise-motion mecha-
where h, k, and m are the height of the step, the distance nism, because the mobility of GBs close to nanoparticles is
between the two steps, and the boundary migration mobility, more heavily reduced than that of GBs far from nanoparticles.
respectively. In the above equation, the first term on the right In contrast, smooth GBs were found in samples without nano-
side represents the inherent motion velocity of the steps, and particles at the GBs. An asymmetric Ce-doped BSCF membrane
the second term represents the motion velocity inhibited by the with a 90 µm thick dense layer showed a high oxygen permea-
Zener pinning effect due to the introduction of nanoparticles tion flux up to 5.2 mL cm−2 min−1 at 600 °C, indicating that it
at the GBs. The inhibition of the step motion indicates that the is promising to use Ce-doped BSCF membranes for practical
atomic net diffusion fluxes at the GBs are limited by the nano- oxygen production at ILT (Figure 5j).
particles at the GBs; thus, there is no accumulation of cations
for the heterogeneous nucleation and new phase growth.
The above theory has been described by our group.[10] We 4. Dual-Phase Membranes
attempted to use four cations (Y3+, Zr4+, Nb5+, and Ce4+) with
different valence states and ionic radii to form nanoparticles Although the BaFe1−xMxO3−δ and other cobalt-free perovskite
at the GBs of BSCF membranes. At the doping amount of membranes are stable in strong reducing environments, decom-
3 mol%, the Zr4+ and Nb5+ entered the BSCF lattice completely. positions of the perovskite structure up to tens of microns deep
Oxygen permeation experiments revealed that the BSCF mem- are inevitable after long-term operations in catalytic membrane
branes doped with Zr4+ and Nb5+ showed degradation behavior reactors. Kinetic demixing of cations for the perovskite oxides
similar to that of the BSCF parent (Figure 5d). The surfaces of is accelerated when they are exposed to a high oxygen partial
all membranes were covered with the SSC porous catalyst layers pressure gradient or atmospheres containing CO2 and steam.
to eliminate the negative effects of the oxyanion-containing Because the responses of the A- and B-site cations of perov-
impurities enriched from the membrane bulk to both the sur- skites to the surrounding atmospheres are different, i.e., the
faces. The Ba3Co10O17 phase was still detectable at the GBs A-site cations easily lose their lattices sites under the erosion
after a 432 h oxygen permeation test of the Zr-doped sample of CO2 and steam, while the B-site cations tend to be reduced
at 600 °C (Figure 5e). A higher doping amount of the Zr4+ and to metals or low-valent cations by reducing gases. In addition,
Nb5+ would inhibit the phase transformation, and a significant the diffusion rates of cations in the A- and B-sites through the
decrease in the oxygen permeability would be inevitable. perovskite lattice are different, so the kinetic demixing is inevi-
Y3+ partially entered the BSCF lattice and produced a table even though the perovskite material can sustain the ero-
small amount of nanoparticles at the GBs (Figure 5f). The sion of CO2 and steam.
degradation rate of the Y-doped BSCF was much slower than Mazanec and Frye proposed a concept of “dual-phase” MIEC
that of the BSCF and the Zr- and Nb-doped BSCF, indicating membranes in a patent in 1990[48] that may solve the difficulty
that the nanoparticles at the GBs effectively inhibited the encountered in “single-phase” membranes. In the DP mem-
phase transformation (Figure 5d). Other researchers verified branes, one phase is an oxygen ionic conductor and the other
that the BSCF stability can be enhanced by a slight Y-doping phase is an electronic conductor (Figure  6a). The oxygen ionic
in the B-site.[47] However, the amount of nanoparticles in the conductor is usually chosen from the fluorite-type oxides (such
sample was too low to fully inhibit the phase transforma- as Y2O3 stabilized ZrO2 (YSZ) and CeO2 doped by rare earth
tion. An analysis of the Y-doped sample after a 500 h test at elements), which have high oxygen ionic conductivities and
600 °C for the oxygen permeation revealed that the Ba3Co10O17 structure stability but low electronic conductivities at elevated
phase did not appear at the GBs where the nanoparticles were temperatures. The development of DP materials experienced
inlayed (Figure 5f,g). three periods based on the type of electronic conductor. In
However, for the Ce-doped sample, the Ce4+ did not enter the first period, precious metals, such as Ag, Au, and Pt, were
the BSCF lattice but combined with the Ba2+ to form BaCeO3 adopted as the electronic conductors because they can maintain
nanoparticles at the GBs (Figure 5h). Even when the doping their metallic states under redox atmospheres at elevated temper-
amount was reduced to 0.1 mol%, the BaCeO3 nanoparticles atures.[49] A volume fraction of no less than 30% is required for

Adv. Mater. 2019, 1902547 1902547  (11 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 6.  a,b) DP membranes made of an electronic conducting phase and ionic conducting phase (a) and a mixed conducting phase and an ionic
conducting phase (b). c) Arrhenius plots of SDC–SSF and SDC–SSCr DP membranes with a thickness of 0.5 mm. c) Reproduced with permission.[51]
Copyright 2012, Elsevier B.V. d) Schematic illustration of the oxygen chemical potential drop and conductivity change across a MIEC membrane in a
POM reactor, where L represents the thickness of the membrane and σe is the electronic conductivity including p-type and n-type. e) Dependence of the
total conductivities on the oxygen partial pressure of SDC, SDC–SSAF and SDC–SSCF at 950 °C. d,e) Reproduced with permission.[57] Copyright 2016,
Elsevier B.V. f) Comparison of oxygen fluxes in oxygen permeation and POM reactions. Data in (f) are replotted from ref. [57].

the electronic conductor to form a continuous conducting net- fraction of the oxides for ionic transport equals unity because
work across membranes. The oxygen permeation fluxes of the the mixed conducting phase has a high ionic conductivity. The
DP membranes consisting of fluorite-type oxides and precious DP membranes 75 wt%Ce0.85Sm0.15−25 wt%Sm0.6Sr0.4CrO3−δ
metals were far lower than the threshold value for industrial (SDC–SSCr) and 75 wt%Ce0.85Sm0.15−25 wt%Sm0.6Sr0.4FeO3−δ
applications. Both the high material cost and the low perme- (SDC–SSFe) were prepared and tested to determine the differ-
ability discouraged the development of this type of materials. In ence in oxygen permeability between the second and third types
the second period, to decrease the material cost, electronic con- of DP membranes (Figure 6c).[51] The DP membrane, SDC–SSF,
ducting oxides were suggested to be substituted for the precious made of ionic conductors and mixed conductors, i.e., the third
metals in the DP membranes.[50] The electronic conducting type, has much higher oxygen permeability and lower per-
oxides included perovskite oxides (such as La1−xSrxMnO3 and meation activation energy than those of the second type of DP
La1−xSrxCrO3), spinel oxides (NiFe2O4, MnFe2O4, etc.) and other membranes consisting of ionic conductors and pure electronic
materials. These DP membranes had high stability and high conducting oxides. After a series of studies reported by our
tolerance to CO2 but low oxygen permeability. The electronic group,[51–57] considerable effort by other researchers was devoted
conducting phase, which had extremely low ionic conductivities, to developing the third type of DP membrane.[58–61]
occupying at least 30% of the membrane bulk volume blocked
the ionic transport in the membrane bulk, thus resulting in low 4.1. Structure Design of DP Membrane Materials
oxygen permeability. Therefore, in the third period, to enhance
the oxygen permeability of DP membranes, we proposed that The design theory of DP membranes is different from that
mixed ionic–electronic conductors be adopted to substitute for of the single-phase perovskite membranes introduced in
the traditional electronic conductors in the DP membranes. Section 3. Although many researchers have argued that the
Therefore, the oxygen transport paths in the membrane bulk ionic and electronic transport paths of DP membranes can be
were extended from those between the ionic conducting phase separately designed according to their potential applications,
grains to the whole membrane bulk (Figure 6b). The volume the interactions (such as chemical reaction, interdiffusion

Adv. Mater. 2019, 1902547 1902547  (12 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

of elements, thermal expansion mismatch) between the two Zener pinning action induced by the ionic conducting phase in
phases do not occur in single-phase perovskite materials but the DP membrane protects the MIEC phase from decomposi-
must be considered during the design of the DP membranes. tion. In addition, the creep rate of the materials is limited by
There are many types of ionic conductors, such as fluorite-type the DP structure.
(including zirconia-based, ceria-based, and bismuth-based),
perovskite-type, La2Mo2O9 series, Aurivillius-type, apatite-type,
pyrochlore-type, and many MIEC materials. The number of 4.2. Microstructural Effects
pairwise combinations of the ionic conducting phase and the
MIEC phase is very large, even without considering the changes The effects of changes in compositions on the oxygen trans-
in the ratios between the two phases and the element ratios in port properties of the ionic conductors and the use of MIEC
each phase. Therefore, it is difficult to obtain a DP membrane perovskite oxides as the electrodes of solid oxide fuel cells and
material with ideal performance. To quickly obtain materials oxygen separation membranes have been widely studied. There
with the targeted performance, a design principle needs to be have been many reviews published that have provided good
constructed for the DP membrane according to the working summaries of elemental doping.[5,6] However, the DP system
environment of the potential applications. is much more complicated than the single-phase system. Many
For a catalytic membrane reactor, the stability of the mem- new phenomena were observed during the investigation of DP
brane material under the reaction conditions is more important membranes. Here, the microstructural effects on the oxygen
than its oxygen permeability. Thus, the stabilities under both permeation through DP membranes are of concern.
oxidative and reductive environments are considered prior For DP membranes, increases in the permeation fluxes with
to considering the permeability during the design of the DP time during the initial tens of hours of operation have been fre-
membrane. For example, three principles (i.e., the structural quently reported,[55,56,59,62] but this phenomenon rarely occurs
stability under catalytic reaction conditions, the ionic and elec- with perovskite MIEC membranes. Experiments have shown
tronic conductivities under reaction conditions, and the chem- that the increase in the permeation flux with time corresponds
ical compatibility between the two phases) should be consid- to the oxygen exchange reactions at the gas–solid interfaces. For
ered for the selection of materials to devise a DP membrane a DP membrane, its surface microstructure changes from the
material for the syngas (H2+CO) production via the partial oxi- as-prepared state to the state being facile to oxygen exchange
dation of natural gas. Following the above three principles, the during the oxygen permeation process. This microstructure
selection of the ionic conductors and the MIECs is not difficult. change is related to the redistribution of the two phases on
We have given a detailed description of the selection of the two the surfaces at elevated temperatures. Not only does the
phases.[6,51] The resultant DP membranes have the composition unsteady permeation disappear, but the flux can also be sta-
Ce1−xLnxO2−δ–Ln1−ySryFeO3−δ. The perovskite stability under bilized at a higher value if both sides of the membrane are
reducing conditions can be improved by doping appropriate coated with porous catalyst layers for oxygen activation. Cobalt-
amounts of cations such as Al3+, Ga3+, Cr3+/4+, Mn3+/4+, Ti4+ containing perovskite oxides are usually used as the catalyst.
and Zr4+ into the B-site. Ba2+ and Ca2+ cations may be appro- Ostensibly, the function of the porous catalyst layer coated
priate doping cations for the perovskite phase, depending on on DP membranes is similar to that coated on perovskite
the systems of DP membranes being considered. Here, the membranes operated at ILT, but the stabilization mechanisms
design of DP membranes is based on existing knowledge, while are different. When the porous layer is coated on a DP mem-
more advanced strategies are expected with the development brane, it speeds up the oxygen exchange reactions, and thus
of ionic conducting materials. As an example of the material the changes in the surface microstructure have little effect on
design, 75 wt%Ce0.85Sm0.15−25 wt%Sm0.6Sr0.4Al0.3Fe0.7O3−δ the total permeation resistance of the DP membrane. However,
(SDC–SSAF) exhibits a high oxygen permeation flux, which when the porous layer is coated on a perovskite membrane,
is ≈10 times higher than those of traditional membranes.[6] A it mainly acts as a sponge to accommodate the impurities
1100 h on-stream process at 950 °C for syngas generation in enriched from the membrane bulk.
a SDC–SSAF membrane reactor showed that the oxygen per- DP membranes have many new and unique microstructural
meation flux reached 4.2 mL cm−2 min−1, and the phase struc- characteristics, such as the grain size ratio of the two phases,
tures of the SDC–SSAF membrane surface remained intact grain distribution in the bulk and on the surfaces and the GBs
after the long-term on-stream.[52] Moreover, there were no ele- between the two phases. Oxygen exchange and bulk transport
ments enriched on the membrane surfaces. However, a stable are sensitive to these microstructural characteristics, which
surface is not exhibited on perovskite MIEC membranes. The depend on the preparation method, sintering temperature, ele-
DP structure increases the difficulty of the cationic transport mental composition, etc. The effects of the powder synthesis
through the membrane bulk; thus, the special microstructure method on the microstructures, chemical composition, crystal
can prevent the kinetic demixing of the MIEC phase under lattice defects, ionic and electronic conductivity, and oxygen
the large gradient of the oxygen chemical potential. Elemental permeability of perovskite MIEC membranes have been well-
strontium, with its large size, does not easily enter the ceria studied in the past 20 years. For DP membranes, the factors that
lattice; thus, the transport of Sr from one perovskite grain to affect perovskite materials also exist in DP materials to some
another is difficult if there is a fluorite oxide grain between extent. We compared four methods to prepare the powders
the two perovskite grains. From a geometric point of view, the of DP membranes and found that the distributions of the two
diffusion paths are significantly prolonged for the strontium phases in the membranes are different.[63] The DP membrane
element in the SDC–SSAF membrane. More importantly, the derived from a one-pot method (i.e., mixing all the chemical

Adv. Mater. 2019, 1902547 1902547  (13 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

precursors together to prepare the composite powder) has SDC–SSCF DP membrane.[57] The Cr-substitution led to a great
the most homogeneous distribution of the two phases, while improvement of the electronic conductivity in the oxygen par-
the DP membrane derived from the traditional method (i.e., tial pressure range of 10−2–10−15 atm, as shown in Figure 6d. A
mixing the two powders to prepare the composite powder) high oxygen permeation flux of up to 7.8 mL cm−2 min−1 was
shows an inhomogeneous distribution of the two phases. The achieved at 950 °C for the syngas generation, which is nearly
oxygen permeation flux is closely related to the distribution two times that of the SDC–SSAF under the same conditions,
of the two phases, since the more homogeneous distribution although the SDC–SSCF exhibited a lower oxygen permeation
of the two phases results in more active sites for the oxygen flux than did the SDC–SSAF under an oxygen partial pressure
exchange reaction. However, when the volume fraction of the gradient of 0.21 atm/0.005 atm (Figure 6f).
MIEC phase is lower than the percolation threshold (≈30 vol%),
the inhomogeneous distribution of the two phases improves
the oxygen permeation flux if the surfaces are covered with 5. Catalytic Membrane Reactors
MIEC porous catalytic layers to speed up the oxygen reduction
and evolution reactions. The inhomogeneous distribution can Partial oxidation of methane and methane coupling are two
increase the possibility of the formation of an electronic con- widely investigated reactions in MIEC membrane reactors. In
ducting network across the DP membrane. An extreme case these membrane reactors, one side is fed with air and the other
of the inhomogeneous distribution of the two phases is when side is fed with methane. Many reviews have given comprehen-
an electrolyte membrane outside or inside is short-circuited by sive summaries of the two reactions.[5,6,69] With improvements
a wire.[64] Compared with a DP membrane containing an Fe- in the stabilities of membrane materials, new membrane reac-
based MIEC phase, a DP membrane containing a Co-based tors in which both sides of the MIEC membranes are exposed
MIEC phase needs a larger volume fraction of the MIEC phase, to strong reducing atmospheres have been reported in recent
since the Co-based MIEC phase acts as a sintering promoter to years. The thermodynamic limited reactions, for instance
cause the rapid growth of the grains to a large size. The large hydrogen production from water splitting[70–72] and CO2 reduc-
MIEC grains have a lower probability of forming a continuous tion to CO,[73] can be promoted by coupling with a thermody-
conducting network than do small grains at the same volume namic spontaneous reaction in MIEC membrane reactors.
fraction. Therefore, an excessively high sintering temperature Balachandran et al. proposed new MIEC membrane reactors,
disrupts the electronic conducting network if the volume frac- in which the hydrogen generation via water decomposition was
tion of the MIEC phase is not high enough.[65,66] In addition, an combined with syngas production via the partial oxidation of
over high sintering temperature induces the accumulation of methane,[70] and the membrane reactor was experimentally ver-
ionic conducting phases on the surfaces of DP membranes.[65] ified by Jiang et al. using a perovskite hollow fiber membrane
There are many factors that determine the oxygen permeability reactor.[71] The combination of the selective oxidation of ethane
of DP membranes, and some factors interact with each other, to ethylene with the water decomposition reaction was also con-
so it is difficult to study the actions of each factor separately. ducted in the hollow fiber membrane reactor.[72] Recently, our
The microstructure effects of DP membranes are very compli- group proposed two reaction–separation–reaction processes in
cated; therefore, there is no standard explanation for the micro- MIEC membrane reactors for coproducing two types of syngas
structure effects, and a case-by-case analysis is needed. and hydrogen separation.[74,75]
Several DP membranes consisting of ceria-based ionic con-
ductors and Fe-based perovskite-type mixed conductors were
employed for the production of syngas,[54,55,57,58,61,67] and no 5.1. Coproduction of Two Types of Syngas
degradation of phase structure was observed, indicating good
structural and permeation stability under redox environments. Ammonia, one of the most widely produced chemicals in
However, the oxygen permeation fluxes through 0.5 mm thick the world, is synthesized via the classical Haber–Bosch pro-
DP membranes were all ≈4 mL cm−2 min−1 at 950 °C for syngas cess, in which six main steps (i.e., steam reforming, air
production, although the DP membranes had different compo- reforming, water–gas shifting at high and low temperatures,
sitions in both phases.[52,54,55,61] We found that the extremely low CO2 separation and methanation) are performed to produce
electronic conductivity in the central region of the membrane ammonia synthesis gas (ASG, H2/N2  = 3) in which natural
bulk limited the ambipolar diffusion of the oxygen ions and gas is used as the feed. A large amount of CO2, i.e., ≈400 mil-
electrons (Figure 6d), thus increasing the bulk diffusion resist- lion metric tons annually, is discharged from the ammonia
ance. In a relatively high oxygen partial pressure system (>10−10 synthesis industry. More than 80% of the energy consump-
atm) MIEC materials exhibit a p-type electronic conduction, tion of the entire ammonia synthesis process is contributed
while in a relatively low oxygen partial pressure system (<10−15 by the long-route ASG production process. Therefore, new
atm), they exhibit n-type electronic conduction. However, in the processes for ammonia synthesis have attracted the attention
intermediate-low oxygen partial pressure range, i.e., 10−10–10−15 of many researchers. On the other hand, three main steps
atm, they usually have a minimum value corresponding to a (i.e., steam reforming, air separation, and mixed reforming)
p–n transfer.[57,68] Therefore, the oxygen permeation flux can are involved in the production of liquid fuel synthesis gas
be increased by enhancing the electronic conductivity in the (LFSG, H2/CO = 2), and >60% of the energy consumption and
intermediate-low oxygen partial pressure range for syngas investment in the gas-to-liquid process is contributed by the
production in MIEC membrane reactors. Elemental Cr was three steps. The air separation step contributes a large por-
used to replace the elemental Al in SDC–SSAF to form the tion of the energy consumption (30–40%). However, these

Adv. Mater. 2019, 1902547 1902547  (14 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

two multistep and high-energy-consumption processes can conversion were obtained in the asymmetric DP membrane
be united in one MIEC membrane reactor for coproducing reactor (Figure 7e). The membrane reactor was stable during
ASG and LFSG with natural gas, water, and air in one step the steady operation (Figure 7f). The conversion of the water
(Figure 7a). decomposition to hydrogen was up to 55% at a H2O/CH4 ratio
To carry out the process, air and steam in an appropriate of 1.25 (Figure 7g) and corresponded to an energy savings
ratio are introduced into one side (side I) of the membrane larger than 60% compared to the current industrial processes
reactor, and natural gas is introduced into the other side for the two types of syngas production. The success in the stable
(side II). At elevated temperatures, the oxygen species in the air operation of membrane reactors indicates that the new concept
and steam are driven by the oxygen chemical potential gradient for coproducing two types of syngas in membrane reactors is
across the membrane from side I to side II where they react promising and may trigger innovations in the natural gas and
with natural gas to produce LFSG. The electrical neutrality of chemical industries.
the whole process is achieved by the transfer of electrons from
side II to side I. After the oxygen in the air and steam is con-
sumed by the MIEC membrane, ASG with a H2/N2 ratio of 3:1 5.2. MIEC Membranes for Hydrogen Separation
after drying is produced on side I. Thus, the resultant mixture
is ready for ammonia synthesis. The syngas produced on side Hydrogen is a high-quality, clean and sustainable energy car-
II with a H2/CO ratio of 2 is suitable for liquid fuel synthesis. rier. Its production, separation, storage, and utilization have
The highly intensified process has many advantages, as illus- been widely investigated in the past 20 years. CO, CO2, N2, and
trated in Figure 7b. The greatest advantages of the catalytic H2S are the common impurities in the as-produced hydrogen.
membrane reactor are its simple process (i.e., the nine steps These impurities should be removed before hydrogen is sup-
being shortened to one step) and its high energy efficiency. plied to protonic exchange membrane fuel cells and used for
Membrane reactors containing BCF perovskite membranes and the fabrication of semiconductors and solar energy cells. The
SDC–SSAF DP membranes were employed to illustrate the fea- current technologies are limited by their low separation selec-
sibility of the process.[67,74] Considering the high stability but tivity (for the pressure swing adsorption and polymer mem-
low permeability of SDC–SSAF compared to BCF, the SDC– brane separation technologies) or high price and low stability
SSAF was fabricated into asymmetric membranes with a dense (for Pd-based membranes).
layer of 40 µm. Ru/SDC was used as the catalyst coating on In the previous investigations of water splitting in
the surface with a thickness of 20 µm and was infiltrated into MIEC membrane reactors, if the driving force of the water
the porous support layer (Figure 7c,d). High production rates of decomposition reaction is provided by the hydrogen oxidation
the two types of syngas as well as high CO selectivity and CH4 reaction, the whole process does not seem to warrant further

Figure 7.  a) One-step production of the two types of synthesis gas for ammonia (H2/N2 = 3:1) and liquid fuel (H2/CO = 2:1) synthesis from water,
air, and natural gas in a membrane reactor. b) Merits of the membrane reactor compared with those of the industrial processes. a,b) Reproduced with
permission.[74] Copyright 2016, Wiley-VCH. c,d) Top-view (c) and cross-sectional view (d) scanning electron microscopy (SEM) images of the asym-
metric SDC–SSAF membrane. e) Effects of temperature on the performance of the membrane reactor. f) Stable operation of the membrane reactor at
900 °C. g) Energy saving in the membrane reactor. d–g) Reproduced with permission.[67] Copyright 2019, Elsevier Ltd.

Adv. Mater. 2019, 1902547 1902547  (15 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 8.  a) Oxygen-permeable membrane for hydrogen separation; b,c) cross-sectional view of the asymmetric SDC–SSAF DP membranes with both
sides coated with Ru catalysts before (b) and after (c) the long-term operation. d) The dependence of the H2 separation rate and the corresponding PO2
for both sides on temperature. e) Stable operation for the hydrogen separation. f) Effect of the H2S concentration on the hydrogen separation rate. The
concentration of H2 to side I is 50% for the H2S concentration of 0–100 ppm and is 99.98% for the H2S concentration of 200 ppm. a–f) Reproduced
with permission.[75] Copyright 2017, Royal Society of Chemistry.

study.[70,71] However, the process using low-purity hydrogen to performances of the dual-phase membrane reactors were
produce high-purity hydrogen can be operated as a process of stable for hydrogen separation when 50% H2/N2 and 90%
hydrogen separation if steam is fed to side I while a low-purity H2O/He were used as the feed gases (Figure 8e). Although
hydrogen source is introduced into side II (Figure  8a).[75] The the low-purity hydrogen source contained impurities such
high-purity hydrogen is obtained by cooling and drying the as CO2[53] and H2S (Figure 8f),[75] the dual-phase membrane
effluent gas of side I. In this process, when a hydrogen mol- reactors were stable for hydrogen separation. However, Pd-
ecule is produced via the water-splitting reaction on side I, based membranes and MEPC membranes are sensitive to
there is a hydrogen molecule consumed simultaneously on side the poisoning of H2S.[76] Therefore, hydrogen separation
II. The net chemical reaction is zero in the whole process, but using DP membranes is feasible in principle. The mem-
a mass exchange is achieved between the two sides. The fed brane thickness and porous supports have a strong influ-
hydrogen has a low purity while the obtained hydrogen has a ence on the ambipolar diffusion resistance in the membrane
high purity; thus it is a process of hydrogen separation with the bulk and on gaseous reactions at the gas–solid interfaces.
help of steam. In theory, the selectivity of the hydrogen separa- These effects were observed on the MIEC DP membranes
tion is up to 100%, because only oxygen is allowed to permeate for hydrogen separation. Concentration polarization induced
through the dense MIEC membranes. by the thick porous support of the asymmetric membranes
The hydrogen separation process was achieved on DP was not observed compared with membranes with the same
membranes with both surfaces covered with Ru or Ni cata- thickness of the dense layer but different structural types.[77]
lysts to speed up the water-splitting and hydrogen oxidation However, the concentration polarization has been verified on
reactions (Figure 8b,c).[75] At 950 °C, the hydrogen separation asymmetric membranes for oxygen separation from air.[23,78]
rate was as high as 16 mL cm−2 min−1 (Figure 8d), which is This abnormal phenomenon is caused by the diffusion rate of
100–1000 times higher than those of mixed electronic–pro- the hydrogen molecules being faster than that of the oxygen
tonic conducting (MEPC) ceramic membranes. This high flux molecules. The hydrogen separation rate is jointly controlled
is on the same level as those of Pd composite membranes. by the bulk diffusion and interfacial reactions compared with
The hydrogen separation rate was 1.7 mL cm−2 min−1 even membranes with the same structure types but different thick-
at a temperature as low as 600 °C, indicating it is possible to nesses of the dense layer.
extend the running temperature to ILT for hydrogen separa- The concept of using oxygen-permeable MIEC membranes
tion using asymmetric DP membranes. Although the defects for hydrogen separation is different from the traditional con-
in the membrane and sealing are inevitable, the hydrogen cept of membrane separation. In the traditional approach, there
separation factor was achieved on the order of 103–104. The are two mechanisms for the separation of a target molecule,

Adv. Mater. 2019, 1902547 1902547  (16 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

i.e., permeation and rejection. However, in the new method, the the pressure vessel were tested under a variety of operating
hydrogen separation is realized via a reaction–separation–reac- conditions. However, the oxygen permeation flux of a mem-
tion process. This special separation concept may be extended brane module containing 12 wafers decreased by ≈10% after
to other systems in the future. operating for 15 000 h. The degradation was attributed to the
ceramic creep leading to the deformation of the microchannels
under elevated pressure and temperatures (Figure 9c,d).[31] A
6. Technological Development of MIEC transient liquid phase method was invented by APCI to seal
the MIEC membranes. The ceramic sealant has the same or
Membranes for Pure Oxygen Production
similar chemical composition as the MIEC membrane; thus,
and Catalytic Membrane Reactors there is no chemical and thermal incompatibility between the
sealant and the membrane. The innovative sealing technology
In the past 20 years, APCI has led the technological develop- is the key to the APCI’s success in planar MIEC membranes.
ment of MIEC membranes for pure oxygen production. APCI The APCI and its partners were motivated by the success of the
began research and development in 1989 and focused on 5 TPD prototype to construct a 100–150 TPD intermediate-scale
the development of planar ceramic wafers (Figure  9a) with a test unit. However, no further news on the intermediate-scale
composition of LnxA1−xCo1−y−zFeyCuzO3−δ (where Ln and A test unit has been released to date.
are selected from the lanthanide block and the alkaline earth Different from APCI using the planer membranes, the
elements, respectively).[6,31,79] In 2006, a 5 ton-per-day (TPD) Institut für Keramische Technologien und Systeme (IKTS) of
subscale engineering prototype containing six independent Fraunhofer in Germany built a membrane module with BSCF
modules was successfully operated over 1000 days to produce capillaries (Φ 3 mm) with an oxygen production capability of
oxygen with a purity >99% (Figure 9b).[79] Next, full size MIEC 0.38 TPD in 2017 (Figure 9e,f).[24] With the advanced design
modules with an oxygen production capability of ≈1 TPD in of heat exchange between the inlet and outlet air, rubber was

Figure 9.  a) Construction steps of the commercial modules. b) Subscale engineering prototype capable of testing up to six 144-wafer ITM Oxygen
modules. a,b) Reproduced with permission.[79] Copyright 2009, Springer. c) 12-wafer module after testing for over 15 000 h. d) Example of deformation
and decreased cross-sectional area of the channels due to creep. c,d) Reproduced with permission.[31] Copyright 2016, Elsevier B.V. e) BSCF capability
prepared by IKTS. f) Pilot plant for the production of 10 m3 (STP) O2 h−1. e,f) Adapted with permission.[24] Copyright 2018, ICIM.

Adv. Mater. 2019, 1902547 1902547  (17 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

successfully used to seal the capillaries with a long-lasting seal. design of dual-phase materials by considering the critical factors
The module design scale-up to the oxygen production capa- such as phase composition, phase ratio, phase distribution, grain
bility of hundreds of tons per day needs further verification. size, and gas–solid interfaces. Based on the perfect membranes,
IKTS evaluated the total energy demands for oxygen separation newly catalytic membrane reactors were developed for reaction–
including the heat needed to keep the modules at elevated tem- separation–reaction coupling to produce two types of synthesis
perature and the energy requirement for gas compression, and gas in one step and to use the oxygen-permeable membranes
found that the energy costs can be reduced to 18.39 or 14.01 for hydrogen separation. In the common membrane separation
Euro ton−1 oxygen if the gas combustion or waste heat is used process, the separation of the target molecules is realized via
to heat the modules. The energy costs is only 54% and 41% that permeation or rejection by the membranes; however, hydrogen
of cryogenic separation technology. separation with oxygen-permeable membranes is realized via a
Praxair Inc. focuses on developing MIEC membrane reac- coupling process of reaction–separation–reaction. In the future,
tors for natural gas conversion to syngas.[80] Tubular dual-phase the development of MIEC oxygen-permeable membranes should
membranes with a dense layer thickness of 20–30 µm were be focused on the following points:
used to construct the membrane reactors. An innovative design
of an integrated reactor was proposed that contained membrane 1) Membrane materials. It is possible to explore the operating
reactors and steam reformers. The membrane tubes were par- temperatures of dual-phase membranes for intermediate-
allelly arranged in a panel, and the steam reformer tubes were low temperatures by optimizing the microstructures and
arranged in another panel. Then, the membrane tube panels and preparation methods. The processes of oxygen separation
the steam reformer panels were alternately installed in a vessel. and reaction–separation coupling have more potential to be
The natural gas and steam were fed into the steam reformer commercialized using dual-phase membranes since they
tubes and then into the membrane tubes. Air was fed outside the have higher stability than do single-phase perovskite mem-
membrane tubes and steam reformer tubes. This design allows branes at similar permeabilities. The recent development of
for the thermal exchange between the steam reformer tubes dual-phase membrane materials is focused on the change of
(endothermic) and the membrane tubes (exothermic). This the electronic conducting phase. The ceria-based ionic con-
integrated reactor combines the primary steam methane per- ductors are the most often used ionic conducting phase in
former, secondary autothermal reformer, and air separation unit dual-phase membranes, but the remarkable chemical expan-
together. Compared with the traditional reactors, the integrated sion of ceria due to the reduction of Ce4+ to Ce3+ increases
reactor is expected to decrease CO2 emissions by 29% and the the breakage risk of membrane reactors. Zirconia-based ionic
capital investment by 18% while increasing the net income >30% conductors may be good alternatives due to their high stabil-
for a 3000 ton day−1 methanol plant. Six-panel arrays, containing ity. However, it is challenging to develop a mixed conduct-
216 membrane tubes, with a capability of up to 190 Nm3 h−1 ing phase with good chemical compatibility with zirconia.
syngas were tested for 44 weeks without failure during >100 The microstructures and interfaces (including gas–solid and
start-up/shut-down cycles. Praxair Inc. reported the achieve- solid–solid interfaces) may play important roles in the per-
ment of commercial ceramic performance targets for permea- meability of the zirconia-based dual-phase membranes.
tion flux, strength, creep life, degradation, and sealing. A project 2) Permeation model and simulation. The mechanisms of
was started in 2019 to build an integrated reactor with a capa- oxygen exchange at the gas–solid interfaces are not fully
bility of up to 9940 Nm3 h−1 syngas for a methanol plant. understood. The dependence of the permeation resistances,
including the resistances on both the interfaces and the bulk,
on the oxygen partial pressure is not known. Thus, the cur-
rent models cannot provide an accurate analysis of the oxygen
7. Conclusions and Perspectives
permeation process. The applicability of models to different
In this paper, the designs of the microstructure and gas–solid materials and processes needs to be explored in detail to avoid
interfaces of MIEC oxygen-permeable membranes to improve the misleading conclusions. Little research has been carried out
permeability and stability were summarized. The understanding on the simulation of catalytic membrane reactors. The rate-
of the permeation mechanism with the help of a modeling anal- determining step of membrane reactors is rarely discussed in
ysis is important for the development of membrane materials, the literature. The challenge is that all the current models for
oxygen exchange catalysts, and the degradation mechanism. The oxygen permeation are not suitable for membrane reactors,
enrichment of impurities from the membrane bulk to the surface because the conditions of the gas–solid interfacial reactions
decreases the oxygen exchange rate on the gas–solid interfaces, and bulk transport process do not meet the assumptions of
while the new phase appearing at the grain boundaries blocks the the permeation models. In addition, little effort has been put
ambipolar diffusion. Thus, the degradation of MIEC membranes into simulating the membrane modules to understand the
at intermediate-low temperatures is closely related to the changes mass transport in the membranes and modules; thus, the
in the gas–solid interfaces and bulk microstructures. Accordingly, current stage of the design of membrane modules is guided
the strategies to overcome the degradation are focused on the by experiments with a low efficiency.
designs of interfaces and bulk microstructures. The microstruc- 3) Reaction–separation coupling. There is a need to explore new
tural and interfacial effects are more pronounced in dual-phase reaction–separation coupling processes in MIEC oxygen-per-
membranes due to the complicated chemical and physical rela- meable membrane reactors. Decreasing the operating tem-
tionships between the two phases. We have obtained high-per- perature from the high-temperature range (800–1000 °C) to the
formance dual-phase membranes through the comprehensive intermediate-low temperature range (350–700 °C) would allow

Adv. Mater. 2019, 1902547 1902547  (18 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

many oxygen-related catalytic reactions in the MIEC oxygen- J. A. Kozinski, Energy Sci. Eng. 2016, 4, 99; c) Q. Fu, Y. Kansha,
permeable membrane reactors, while the challenges of the C. Song, Y. Liu, M. Ishizuka, A. Tsutsumi, Appl. Energy 2016, 162,
permeation flux of membranes and activity of catalysts would 1114.
[2] a) B. Sikder, A. Chanda, Adv. Eng. Mater. 2016, 18, 1529; b) J. Zhu,
increase sharply. Because it is different from traditional hetero-
G. Liu, Z. Liu, Z. Chu, W. Jin, N. Xu, Adv. Mater. 2016, 28, 3511;
geneous catalysis, the transport of oxygen ions and electrons
c) J. Zhu, G. Zhang, G. Liu, Z. Liu, W. Jin, N. Xu, Adv. Mater. 2017,
from the membrane bulk to the catalysts warrants further study. 29, 1606377; d) W. Fang, F. Liang, Z. Cao, F. Steinbach, A. Feldhoff,
Common catalyst supports, such as alumina, silica, and zeo- Angew. Chem., Int. Ed. 2015, 54, 4847; e) I. García-Torregrosa,
lites, which cannot transport oxygen ions and electrons simulta- M. P. Lobera, C. Solís, P. Atienzar, J. M. Serra, Adv. Energy Mater.
neously, are not effective for membrane reactors. In the past 20 2011, 1, 618; f) J. Sunarso, S. Liu, Y. S. Lin, J. C. Diniz da Costa,
years, research has focused on the development of membrane Energy Environ. Sci. 2011, 4, 2516; g) H. Wang, C. Tablet, A. Feldhoff,
materials and new reaction–separation coupling processes, but J. Caro, Adv. Mater. 2005, 17, 1785.
few studies have been conducted on catalyst development for [3] a) P. A. Armstrong, D. L. Bennett, E. P. Ted Foster, V. E. Stein,
membrane reactors. The experiences and knowledge from solid presented at Gasification Technologies Conf., San Francisco, CA,
oxide cells are useful for the development of customized cata- October 2002; b) P. A. Armstrong, D. L. Bennett, E. P. Ted Foster,
V. E. Stein, Gasification Technologies Conf., Washington DC, USA,
lysts for MIEC oxygen-permeable membrane reactors.
October 2004.
4) Technologies. Tape-casting technology is the most frequently [4] a) W. Chen, L. van der Ham, A. Nijmeijer, L. Winnubst, J. Membr.
used approach for the fabrication of flat asymmetric mem- Sci. 2015, 492, 461; b) M. Puig-Arnavat, S. Soprani, M. Søgaard,
branes. However, the tubular shape is ideal for a MIEC oxygen- K. Engelbrecht, J. Ahrenfeldt, U. B. Henriksen, P. V. Hendriksen,
permeable reactor operating at a high pressure. Pores and cracks RSC Adv. 2013, 3, 20843; c) S. Smart, C. X. C. Lin, L. Ding,
are difficult to avoid if the tubular asymmetric membranes K. Thambimuthu, J. C. Diniz da Costa, Energy Environ. Sci. 2010, 3,
are prepared via a traditional dip-coating method. Therefore, 268.
new methods are needed for the highly effective preparation [5] a) W. S. Yang, H. H. Wang, X. F. Zhu, L. W. Lin, Top. Catal. 2005,
of tubular asymmetric membranes. Membrane sealing has 35, 155; b) J. Sunarso, S. Baumann, J. M. Serra, W. A. Meulenberg,
not been adequately developed in the past 20 years. Sealants S. Liu, Y. S. Lin, J. C. Diniz da Costa, J. Membr. Sci. 2008, 320, 13;
c) Y. Wei, W. Yang, J. Caro, H. Wang, Chem. Eng. J. 2013, 220, 185;
such as glass, ceramics, and metals have been reported in the
d) P. M. Geffroy, J. Fouletier, N. Richet, T. Chartier, Chem. Eng. Sci.
literature; however, their reliabilities need to be carefully evalu- 2013, 87, 408; e) C. Zhang, J. Sunarso, S. Liu, Chem. Soc. Rev. 2017,
ated under operating conditions. Sealing membranes with rub- 46, 2941; f) C. Li, J. J. Chew, A. Mahmoud, S. Liu, J. Sunarso, J.
ber gaskets as they come out of the high-temperature zone is Membr. Sci. 2018, 567, 228; g) A. A. Plazaola, A. C. Labella, Y. Liu,
easily accomplished; however, its feasibility at the industrial N. B. Porras, D. A. P. Tanaka, M. V. S. Annaland, F. Gallucci,
scale needs carefully to be evaluated when scale, energy con- Processes 2019, 7, 128.
sumption, costs, and lifetime are all considered. In addition, [6] X. Zhu, W. Yang, Mixed Conducting Ceramic Membranes, Springer,
the MIEC membrane process and the economic outlook are Berlin 2017.
competitive but have yet to be verified by a practical case. [7] X. Zhu, H. Liu, Y. Cong, W. Yang, AIChE J. 2012, 58, 1744.
[8] Y. Liu, X. Zhu, M. Li, Y. Zhu, W. Yang, AIChE J. 2016, 62,
2803.
[9] Y. Liu, X. Zhu, W. Yang, AIChE J. 2015, 61, 3879.
Acknowledgements [10] Y. Liu, X. Zhu, M. Li, R. P. O’Hayre, W. Yang, Nano Lett. 2015, 15,
The authors appreciate the financial supports from the National Natural 7678.
Science Foundation of China (91545202 and U1508203), the Strategic [11] H. J. M. Bouwmeester, H. Kruidhof, A. J. Burggraaf, Solid State
Priority Research Program of the Chinese Academy of Sciences (CAS) Ionics 1994, 72, 185.
(Grant No. XDB17000000), grants of Dalian National Laboratory for [12] a) T. H. Lee, Y. L. Yang, A. J. Jacobson, B. Abeles, M. Zhou, Solid
Clean Energy (DNL) (DNL180203 and DICP&QIBEBT UN201708), and State Ionics 1997, 100, 77; b) T. H. Lee, A. J. Jacobson, Y. L. Yang,
the Youth Innovation Promotion Association of CAS. B. Abeles, S. Milner, Solid State Ionics 1997, 100, 87.
[13] Y. Lin, W. Wang, J. Han, AIChE J. 1994, 40, 786.
[14] S. J. Xu, W. J. Thomson, Chem. Eng. Sci. 1999, 54, 3839.
Conflict of Interest [15] Y. Zhu, W. Li, Y. Liu, X. Zhu, W. Yang, AIChE J. 2017, 63, 4043.
The authors declare no conflict of interest. [16] a) X. Li, T. Kerstiens, T. Markus, J. Membr. Sci. 2013, 438, 83;
b) Z. Yáng, J. Martynczuk, K. Efimov, A. S. Harvey, A. Infortuna,
P. Kocher, L. J. Gauckler, Chem. Mater. 2011, 23, 3169.
Keywords [17] W. Wang, Y. S. Lin, J. Membr. Sci. 1995, 103, 219.
[18] E. G. Babakhani, J. Towfighi, L. Shirazi, A. N. Pour, J. Membr. Sci.
interfaces, membrane reactors, microstructure, mixed ionic–electronic 2011, 376, 78.
conductors, oxygen permeation [19] H. Kruidhof, H. J. M. Bouwmeester, R. H. E. Doorn, A. J. Burggraaf,
Solid State Ionics 1993, 63–65, 816.
Received: April 22, 2019
[20] S. McIntosh, J. Vente, W. Haije, D. Blank, H. Bouwmeester, Solid
Revised: June 11, 2019
State Ionics 2006, 177, 833.
Published online:
[21] Z. Shao, G. Xiong, J. Tong, H. Dong, W. Yang, Sep. Purif. Technol.
2001, 25, 419.
[22] Z. Shao, W. Yang, Y. Cong, H. Dong, J. Tong, G. Xiong, J. Membr.
Sci. 2000, 172, 177.
[1] a) Q. Fu, Y. Kansha, C. Song, Y. Liu, M. Ishizuka, A. Tsutsumi, [23] S. Baumann, J. M. Serra, M. P. Lobera, S. Escolástico,
Energy 2016, 103, 440; b) S. Nanda, S. N. Reddy, S. K. Mitra, F. Schulze-Küppers, W. A. Meulenberg, J. Membr. Sci. 2011, 377, 198.

Adv. Mater. 2019, 1902547 1902547  (19 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

[24] M. Bernhardt, R. Kircheisen, R. Kriegel, presented at 15th Int. Conf. J. Membr. Sci. 2006, 286, 22; d) J. Yi, Y. Zuo, W. Liu, L. Winnubst,
on Inorganic Membranes, Dresden, Germany, June 2018. C. Chen, J. Membr. Sci. 2006, 280, 849; e) C. Yang, Q. Xu, C. Liu,
[25] a) X. Zhu, H. Wang, W. Yang, Chem. Commun. 2004, 1130; J. Liu, C. Chen, W. Liu, Mater. Lett. 2011, 65, 3365; f) H. Luo,
b) X. Zhu, Y. Cong, W. Yang, J. Membr. Sci. 2006, 283, 38; H. Jiang, K. Efimov, J. Caro, H. Wang, AIChE J. 2011, 57, 2738.
c) K. Watanabe, D. Takauchi, M. Yuasa, T. Kida, K. Shimanoe, [51] X. Zhu, M. Li, H. Liu, T. Zhang, Y. Cong, W. Yang, J. Membr. Sci.
Y. Teraoka, N. Yamazoe, J. Electrochem. Soc. 2009, 156, E81; 2012, 394–395, 120.
d) T. Kida, A. Yamasaki, K. Watanabe, N. Yamazoe, K. Shimanoe, [52] X. Zhu, Q. Li, Y. Cong, W. Yang, Catal. Commun. 2008, 10, 309.
J. Electrochem. Soc. 2010, 183, 2426; e) Y. Lu, H. Zhao, K. Li, X. Du, [53] a) Q. Li, X. Zhu, W. Yang, J. Membr. Sci. 2008, 325, 11; b) X. Zhu,
Y. Ma, X. Chang, N. Chen, K. Zheng, K. Świerczek, J. Mater. Chem. H. Liu, Y. Cong, W. Yang, Chem. Commun. 2012, 48, 251; c) L. Cai,
A 2017, 5, 7999. S. Hu, Z. Cao, H. Li, X. Zhu, W. Yang, AIChE J. 2019, 65, 1088.
[26] Y. Lu, H. Zhao, X. Cheng, Y. Jia, X. Du, M. Fang, Z. Du, K. Zheng, [54] X. Zhu, W. Yang, AIChE J. 2008, 54, 665.
K. Świerczek, J. Mater. Chem. A 2015, 3, 6202. [55] X. Zhu, Q. Li, Y. He, Y. Cong, W. Yang, J. Membr. Sci. 2010,
[27] Y. Lu, H. Zhao, X. Chang, X. Du, K. Li, Y. Ma, S. Yi, Z. Du, K. Zheng, 360, 454.
K. Świerczek, J. Mater. Chem. A 2016, 4, 10454. [56] X. Zhu, Y. Liu, Y. Cong, W. Yang, Solid State Ionics 2013, 253, 57.
[28] W. Li, Z. Cao, X. Zhu, W. Yang, AIChE J. 2017, 63, 1278. [57] L. Cai, W. Li, Z. Cao, X. Zhu, W. Yang, J. Membr. Sci. 2016,
[29] X. Zhu, H. Wang, W. Yang, Solid State Ionics 2006, 177, 2917. 520, 607.
[30] a) B. Rutkowski, J. Malzbender, T. Beck, R. W. Steinbrech, [58] a) J. Xue, L. Chen, Y. Wei, H. Wang, Chem. Eng. J. 2017, 327, 202;
L. Singheiser, J. Eur. Ceram. Soc. 2011, 31, 493; b) V. Stournari, b) J. H. Park, Y.-i. Kwon, G. D. Nam, J. H. Joo, J. Mater. Chem. A
S. F. P. ten Donkelaar, J. Malzbender, T. Beck, L. Singheiser, 2018, 6, 14246.
H. J. M. Bouwmeester, J. Eur. Ceram. Soc. 2015, 35, 1841. [59] Y. He, L. Shi, F. Wu, W. Xie, S. Wang, D. Yan, P. Liu, M.-R. Li,
[31] L. L. Anderson, P. A. Armstrong, R. R. Broekhuis, M. F. Carolan, J. Caro, H. Luo, J. Mater. Chem. A 2018, 6, 84.
J. Chen, M. D. Hutcheon, C. A. Lewinsohn, C. F. Miller, [60] a) H. Luo, K. Efimov, H. Jiang, A. Feldhoff, H. Wang, J. Caro, Angew.
J. M. Repasky, D. M. Taylor, C. M. Woods, Solid State Ionics 2016, Chem., Int. Ed. 2011, 50, 759; b) M. Anderson, Y. S. Lin, Energy Environ.
288, 331. Sci. 2013, 8, 3675; c) W. Fang, Y. Zhang, J. Gao, C. Chen, Ceram. Int.
[32] B. Koo, K. Kim, J. K. Kim, H. Kwon, J. W. Han, W. C. Jung, Joule 2014, 40, 799; d) E. Rebollo, C. Mortalò, S. Escolástico, S. Boldrini,
2018, 2, 1476. S. Barison, J. M. Serra, M. Fabrizio, Energy Environ. Sci. 2015, 8, 3675;
[33] Y. Liu, X. Zhu, M. Li, H. Liu, Y. Cong, W. Yang, Angew. Chem., Int. e) W. Fang, F. Steinbach, Z. Cao, X. Zhu, A. Feldhoff, Angew. Chem.,
Ed. 2013, 52, 3232. Int. Ed. 2016, 55, 8648; f) J. Garcia-Fayos, M. Balaguer, S. Baumann,
[34] Y. Liu, X. Zhu, M. Li, W. Li, W. Yang, J. Membr. Sci. 2015, 492, 173. J. M. Serra, J. Membr. Sci. 2018, 548, 117; g) Z. Du, Y. Ma, H. Zhao,
[35] Y. Liu, X. Zhu, W. Yang, J. Membr. Sci. 2016, 501, 53. K. Li, Y. Lu, J. Membr. Sci. 2019, 574, 243; h) Y.-i. Kwon, J. H. Park,
[36] Q. Li, X. Zhu, Y. He, W. Yang, Sep. Purif. Technol. 2010, 73, 38. S. M. Kang, G. D. Nam, J. W. Lee, J. H. Kim, D. Kim, S. M. Jeong,
[37] J. Tong, W. Yang, B. Zhu, R. Cai, J. Membr. Sci. 2002, 203, 175. J. H. Yu, J. H. Joo, Energy Environ. Sci. 2019, 12, 1358.
[38] Y. Cheng, H. Zhao, D. Teng, F. Li, X. Lu, W. Ding, J. Membr. Sci. [61] H. Luo, H. Jiang, T. Klande, Z. Cao, F. Liang, H. Wang, J. Caro,
2008, 322, 484. Chem. Mater. 2012, 24, 2148.
[39] a) C. A. Hancock, P. R. Slater, Dalton Trans. 2011, 40, 5599; [62] a) W. Fang, F. Steinbach, C. Chen, A. Feldhoff, Chem. Mater. 2015,
b) J. M. Porras-Vazquez, P. R. Slater, J. Power Sources 2012, 209, 27, 7820; b) K. Partovi, M. Bittner, J. Caro, J. Mater. Chem. A 2015,
180; c) A. Jarvis, P. R. Slater, Crystals 2017, 7, 169. 3, 24008; c) W. Yang, F. Li, Q. Li, Chem. Eng. Sci. 2019, 199, 210.
[40] J. M. Porras-Vazquez, R. I. Smith, P. R. Slater, J. Solid State Chem. [63] X. Zhu, H. Wang, W. Yang, J. Membr. Sci. 2008, 309, 120.
2014, 213, 132. [64] a) K. Zhang, Z. Shao, C. Li, S. Liu, Energy Environ. Sci. 2012, 5, 5257;
[41] B. C. H. Steele, Curr. Opin. Solid State Mater. Sci. 1996, 1, 684. b) K. Zhang, L. Liu, Z. Shao, R. Xu, J. C. Diniz da Costa, S. Wang,
[42] S. Svarcova, K. Wiik, J. Tolchard, H. J. M. Bouwmeester, T. Grande, S. Liu, J. Mater. Chem. A 2013, 1, 9150; c) L. Wang, S. Imashuku,
Solid State Ionics 2008, 178, 1787. A. Grimaud, D. Lee, K. Mezghani, M. A. Habib, Y. Shao-Horn, ECS
[43] M. Arnold, T. M. Gesing, J. Martynczuk, A. Feldhoff, Chem. Mater. Electrochem. Lett. 2013, 2, F77.
2008, 20, 5851. [65] Q. Li, X. Zhu, Y. He, Y. Cong, W. Yang, J. Membr. Sci. 2011, 367,
[44] P. Müller, H. Störmer, L. Dieterle, C. Niedrig, E. Ivers-Tiffée, 134.
D. Gerthsen, Solid State Ionics 2012, 206, 57. [66] H. Li, X. Zhu, Y. Liu, W. Wang, W. Yang, J. Membr. Sci. 2014, 462,
[45] P. Müller, H. Störmer, M. Meffert, L. Dieterle, C. Niedrig, 170.
S. F. Wagner, E. Ivers-Tiffée, D. Gerthsen, Chem. Mater. 2013, 25, [67] W. Li, Z. Cao, H. Li, X. Zhu, W. Yang, Int. J. Hydrogen Energy 2019,
564. 44, 4218.
[46] J. Sun, M. Yang, G. Li, T. Yang, F. Liao, Y. Wang, M. Xiong, J. Lin, [68] a) M. V. Patrakeev, J. A. Bahteeva, E. B. Mitberg, I. A. Leonidov,
Inorg. Chem. 2006, 45, 9151. V. L. Kozhevnikov, K. R. Poeppelmeier, J. Solid State Chem. 2003,
[47] M. Meffert, L.-S. Unger, L. Grünewald, H. Störmer, S. F. Wagner, 172, 219; b) M. V. Patrakeev, I. A. Leonidov, V. L. Kozhevnikov,
E. Ivers-Tiffée, D. Gerthsen, J. Mater. Sci. 2017, 52, 2705. V. V. Kharton, Solid State Sci. 2004, 6, 907.
[48] T. J. Mazanec, J. G. Frye, Eur. Patent Appl., 0399 833 A1 1990. [69] a) Y. S. Lin, Y. Zeng, J. Catal. 1996, 164, 220; b) Y. Y. Liu, X. Y. Tan,
[49] a) C. Chen, B. A. Boukamp, H. J. M. Bouwmeester, G. Z. Cao, K. Li, Catal. Rev. 2006, 48, 145; c) W. Deibert, M. E. Ivanova,
H. Kruidhof, A. J. A. Winnubst, A. J. Burggraaf, Solid State Ionics S. Baumann, O. Guillon, W. A. Meulenberg, J. Membr. Sci. 2017,
1995, 76, 23; b) C. Chen, H. Kruidhof, H. J. M. Bouwmeester, 543, 79; d) V. O. Igenegbai, R. J. Meyer, S. Linic, Appl. Catal., B
H. Verweij, A. J. Burggraaf, Solid State Ionics 1997, 99, 215; c) J. Kim, 2018, 230, 29.
Y. Lin, J. Membr. Sci. 2000, 167, 123; d) J. Kim, Y. S. Lin, AIChE J. [70] a) U. Balachandran, Int. J. Hydrogen Energy 2004, 29, 291;
2000, 46, 1521. b) B. U. T. H. Lee, C. Y. Park, S. E. Dorris, ECS Trans. 2008, 13,
[50] a) V. V. KhartonA. V. Kovalevsky, A. P. Viskup, F. M. Figueiredo, 379; c) C. Y. Park, T. H. Lee, S. E. Dorris, Y. Lu, U. Balachandran,
A. A. Yaremchenko, E. N. Naumovich, F. M. B. Marques, J. Elec- Int. J. Hydrogen Energy 2011, 36,9345; d) C. Y. Park, T. H. Lee,
trochem. Soc., 2000, 1472814; b) U. Nigge, H.-D. Wiemhofer, U. Balachandran, Int. J. Hydrogen Energy 2013, 38, 6450.
E. W. J. Romer, H. J. M. Bouwmeester, T. R. Schulte, Solid State [71] H. Jiang, H. Wang, S. Werth, T. Schiestel, J. Caro, Angew. Chem., Int.
Ionics 2002, 146, 163; c) B. Wang, J. Yi, L. Winnubst, C. Chen, Ed. 2008, 47, 9341.

Adv. Mater. 2019, 1902547 1902547  (20 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

[72] H. Jiang, F. Liang, O. Czuprat, K. Efimov, A. Feldhoff, [77] W. Li, Z. Cao, X. Zhu, W. Yang, J. Membr. Sci. 2019, 573, 370.
S. Schirrmeister, T. Schiestel, H. Wang, J. Caro, Chem. - Eur. J. 2010, [78] a) X. Feng, J. Ivory, V. S. V. Rajan, AIChE J. 1999, 45, 2142; b) H. Li,
16, 7898. U. Schygulla, J. Hoffmann, P. Niehoff, K. Haas-Santo, R. Dittmeyer,
[73] a) W. Q. Jin, C. Zhang, P. Zhang, Y. Q. Fan, N. P. Xu, AIChE J. 2006, Chem. Eng. Sci. 2014, 108, 94.
52, 2545; b) X. Y. Wu, A. F. Ghoniem, ChemSusChem 2018, 11, 483. [79] A. C. Bose, G. J. Stiegel, P. A. Armstrong, B. J. Halper, E. P. Foster,
[74] W. Li, X. Zhu, S. Chen, W. Yang, Angew. Chem., Int. Ed. 2016, 55, 8566. in Inorganic Membranes for Energy and Environmental Applications
[75] W. Li, Z. Cao, L. Cai, L. Zhang, X. Zhu, W. Yang, Energy Environ. Sci. (Ed: A. C. Bose), Springer, New York 2009, p 3–25.
2017, 10, 101. [80] a) J. Li, presented at DOE/NETL Carbon Capture Meet., Pittsburgh,
[76] a) M. Kajiwara, S. Uemiya, T. Kojima, Int. J. Hydrogen Energy 1999, PA, USA, June 2015; b) J. Li, presented at DOE/NETL Gasification
24, 839; b) Z. Tao, L. Yan, J. Qiao, B. Wang, L. Zhang, J. Zhang, Systems Program Portfolio Review Meet., Pittsburgh, PA, USA, March
Prog. Mater. Sci. 2015, 74, 1. 2017.

Adv. Mater. 2019, 1902547 1902547  (21 of 21) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

You might also like