Journal of Membrane Science 432 (2013) 58–65

Contents lists available at SciVerse ScienceDirect

Journal of Membrane Science

0376-73

http://d

n Corr

E-m

journal homepage: www.elsevier.com/locate/memsci

Bi1.5Y0.3Sm0.2O3-d-based ceramic hollow fibre membranes foroxygen separation and chemical reactions

Nur Hidayati Othman, Zhentao Wu, Kang Li n

Department of Chemical Engineering, Imperial College London, London SW7 2AZ, UK

a r t i c l e i n f o

Article history:

Received 17 November 2012

Received in revised form

2 January 2013

Accepted 3 January 2013Available online 11 January 2013

Keywords:

BYS

Ceramic hollow fibre membrane

Dual-layer

Membrane reactor

Oxygen permeation

Catalytic reactions

88/$ - see front matter & 2013 Elsevier B.V. A

x.doi.org/10.1016/j.memsci.2013.01.004

esponding author. Tel.: þ44 207 5945676; fa

ail address: kang.li@imperial.ac.uk (K. Li).

a b s t r a c t

In this study, Bi1.5Y0.3Sm0.2O3-d (BYS), an oxide of great ionic conductivity, has been used to develop

BYS-La0.8Sr0.2MnO3-d (LSM) dual-phase ceramic hollow fibre membranes in an objective of promoting

oxygen permeation that has been considered as the controlling step of our recent dual-layer ceramic

hollow fibre membrane reactor (DL-CHFMR) for methane conversion, and subsequently lowering the

temperature needed for both oxygen separation and catalytic reaction. Oxygen permeation of

approximately 1.21 ml min�1 cm�2 (900 1C, Ar as sweep gas) was achieved by a single-layer BYS–

LSM hollow fibre membrane, which is substantially higher than the previous counterpart of

(ZrO2)0.90(Sc2O3)0.10(ScSZ)–LSM, proving the advantage of using BYS in promoting oxygen permeation.

Although the stability of BYS in strong reducing atmosphere hampers the use of BYS–LSM/BYS–Ni

DL-CHFMR for partial oxidation of methane (POM), its great ionic conductivity and catalytic activity to

oxidative coupling of methane (OCM) would lead to the further development of more efficient

DL-CHFMR design that can be operated at possibly lower temperatures and under less reducing

atmospheres.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Ceramic membranes composed of a separation layer supportedon a porous substrate layer have shown a number of importantadvantages [1–3], especially in escalating permeation flux byreducing the thickness of separation layer. However, such amembrane structure is currently achieved via a multi-step fabri-cation process that consists of repeated coating and sinteringsteps, which is complicated, and time and energy consuming.Combining the multi-step processes into a single step processwould significantly reduce the fabrication costs, besides addi-tional benefits such as good adhesion between layers and con-sistent quality control. Our previous studies have demonstratedthat a dual-layer ceramic hollow fibre membrane consisting of anouter separation layer supported on an inner catalytic substratelayer can be prepared using a reliable single-step co-extrusionand co-sintering process. When different membrane materials areused in each layer, certain functions can be integrated into suchunique membrane design. Membranes of this type have beenutilised as a compact membrane reactor for partial oxidation ofmethane (POM) [4,5] and micro-tubular solid oxide fuel cells(SOFCs) [6–10].

ll rights reserved.

x: þ44 207 5945629.

In our previously designed dual-layer ceramic hollow fibremembrane reactor (DL-CHFMR) for POM reaction, the outer thindense layer (ZrO2)0.90(Sc2O3)0.10(ScSZ)–La0.8Sr0.2MnO3-d (LSM) or(LSM–YSZ) for oxygen separation was supported on the innerporous catalytic substrate layer (Ni-ScSZ or Ni-YSZ) [4,5]. Oxygenpermeation is always the slowest or controlling step even if thethickness of separation layer can be reduced to less than 10 mm.High operating temperatures are thus needed for proper oxygenpermeation flux, good methane conversion and suppressed cokeformation, which addresses the importance of oxygen permeationof the outer separation layer to the performance of such DL-CHFMR design.

Mixed ionic and electronic conducting (MIEC) membraneshave received great attentions for several decades due to theirunique mechanism for oxygen separation. Such membranes havealso shown many potential applications related to pure oxygenproduction, from medical treatments to large scale combustionprocesses, as long as the technical challenges of permeation flux,membrane sealing and durability can be overcome [11]. Similar toMIEC materials (single phase) with typically a perovskite typestructure, dual-phase materials with ionic conducting and elec-tronic conducting in two separate phases show mixed conductingbehaviour as well. As a result, oxygen permeation of dual-phasematerials should also be affected by two factors, i.e. the thick-ness of separation layer [12] and its mixed ionic-electronicconductivity. Based on Wagner theory, oxygen permeation of

Table 1Composition of single- and dual-layer spinning suspensions.

Materials Single-layer

membrane (wt%)

Dual-layer membrane

Outer

layer (wt%)

Inner

layer (wt%)

BYS 43.69 34.16 41.90

LSM 23.74 27.84 –

NiO – – 23.10

Polymer binder, PESf 6.75 6.20 6.50

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–65 59

MIEC membrane is proportional to sise=siþse [13]. Thus, dual-phase materials with higher ionic conductivity and electronicconductivity are always preferred. Since LSM is a great electronicconducting ceramic, improving the ionic conductivity of the LSMbased dual-phase separation layer is the key to higher oxygenpermeation. According to Singhal and Kendall [14], bismuthyttrium oxide (BY) exhibits higher ionic conductivity than ScSZand YSZ, especially between 600 and 800 1C. However, because itis thermodynamically unstable, doping technique is implementedto fine-tune its stability and conductivity. Bismuth yttrium oxidedoped with samarium (Sm2O3) gives much higher oxygen per-meation and greater chemical stability over a wide range oftemperatures [15]. In a recent study [16], BYS was mixed withLSM to fabricate dual-phase hollow fibre membranes for oxygenseparation, with the oxygen permeation flux of 0.52 ml min�1 cm�2

(3.9�10�7 mol s�1 cm�2) at 850 1C using helium as sweep gas.As a result, BYS is used to replace ScSZ or YSZ in this study as

the ionic conducting phase in the oxygen separation layer, inorder to improve the oxygen permeation of DL-CHFMR for abetter reactor performance at possibly lower temperatures. Sucha reactor design offers a more structured catalytic interface whencompared to the conventional method of packing catalystparticles inside membrane reactor, therefore allowing moreefficient use of catalytic surface area. For the purpose of compar-ison, oxygen permeations of both single-layer BYS–LSM hollowfibre membranes and BYS–LSM/BYS–Ni dual-layer counterpartsare investigated, and linked to the changes in membrane micro-structures during oxygen permeation and catalytic reactions.Although the stability of BYS in strong reducing atmospherehampers the use of BYS–LSM/BYS–Ni DL-CHFMR for partialoxidation of methane (POM), it shows potentials to oxidativecoupling methane (OCM) due to the great catalytic activity ofBYS [17].

Additive, PVP 30 2.08 1.00 –

Solvent, DMSO 23.74 30.80 28.50

Density: BYS¼8.10 g/cm3, LSM¼6.6 g/cm3 and NiO¼6.7 g/cm3.

Spinning suspension

Bore fluid Single layer

Inner layer spinning suspension

Outer layer spinning suspension

Bore fluidOuter layer

Inner layer

Fig. 1. Schematic diagram of spinneret for (a) single-layer (b) dual-layer hollow

fibre membranes.

2. Experimental

2.1. Chemicals and materials

BYS (Bi1.5Y0.3Sm0.2O3-d) with a surface area of 4.4 m2/g waspurchased from PraxAir Surface Technologies, USA. LSM(La0.8Sr0.2MnO3) with a surface area of 5.8 m2/g and NiO with asurface area of 3.7 m2/g were commercially available from FuelCell Materials, USA, and were used as supplied. Dimethyl sulf-oxide, (DMSO,VWR), Polyvinylpyrollidone (PVP-K30, Fluka) andpolyethersulfone (PESf, Radal A300, Ameco Performance) wereused as a solvent, an additive and a polymer binder in spinningsuspension, respectively. Deionized water and tap water wereused as a internal- and external-coagulations, respectively, duringthe preparation of hollow fibre membranes.

2.2. Preparation of BYS–LSM single-layer and BYS–LSM/BYS–NiO

dual-layer hollow fibre membranes

Received BYS powder was calcined in static air at 800 1C for2 h, followed by ball milling for at least 8 h prior to the prepara-tion of spinning suspension. Generally, spinning suspensionswere prepared by dispersing a pre-determined amount of ceramicpowders into a solution of DMSO and additive. The mixture wasroll milled (G 91, Gladstone Engineering) for at least 24 h beforeadding PESf. The milling was continued for another 2–3 days toensure a homogeneous spinning suspension. Table 1 lists thecompositions of spinning suspensions used in this study.

The fabrications of single- and dual-layer ceramic hollowfibres have been introduced elsewhere [5,18]. Generally, thespinning suspension needs to be degassed by stirring under

vacuum prior to the spinning, in order to remove air bubbles.After degassing, the spinning suspension was transferred into a200 ml stainless steel syringe controlled by a syringe pump(Harvard PHD 22/2000 Hpsi), extruded through a spinneret andpassing through an air gap before entering into the externalcoagulant bath. In the mean time, a stream of internal coagulantcontrolled by another syringe pump flowed through the centralbore of the spinneret, forming the hollow fibre configuration. Atube in orifice spinneret and a triple-orifice spinneret were usedfor the fabrication of single- and dual-layer hollow fibre mem-branes, respectively, as shown schematically in Fig. 1.

Spinning parameters of both the single- and dual-layer hollowfibres are listed in Table 2. The air gap for single-layer fibres wasdifferent from the dual-layer membranes, due to different visc-osities of the spinning suspensions. The formed precursor fibreswere left overnight in a water bath filled with deionised (DI)water, to complete the phase inversion process. The precursorfibres were then cut to a specific length and dried at roomtemperature prior to sintering in a CARBOLITE furnace. Thefurnace was heated up from room temperature to 600 1C at arate of 2 1C/min and dwelled for 2 h to remove the organiccomponents. Then, the furnace temperature was increased to

Table 2Spinning parameters for single- and dual-layer hollow fibre membranes.

Single-layer Dual-layer

Air gap (cm) 2 25

Single layer extrusion rate (ml/min) 10 –

Inner layer extrusion rate (ml/min) – 11

Outer layer extrusion rate (ml/min) – 0.2–5.0

Bore fluid extrusion rate (ml/min) 10–13 10

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–6560

the target temperature (1040–1100 1C) at 5–10 1C/min anddwelled for another 5–12 h. Finally, the furnace was cooled downto room temperature at 3 1C/min to complete the sinteringprocess.

2.3. Characterizations

2.3.1. X-ray diffraction (XRD)

The X-ray diffraction patterns were obtained with anX’celerator detector (X’Pert PRO model) using Cu-Ka as theradiation source. The XRD scans were carried out in a 2y rangefrom 51 to 801 using a step width of 0.051. The equipment voltageand current were set at 40 kV and 40 mA, respectively.

2.3.2. Sintering behaviour

The sintering behaviour of membrane materials was analysedusing a dilatometer (NETZCH, DIL 402 C) in static air. Ceramicpowders were uniformly packed into a square-shaped stainlesssteel mould (6 mm�6 mm�6 mm) and compressed verticallyusing a hydraulic pressure of 1 t. A 30 kN force was applied to thecompacted sample by a ceramic push rod in the dilatometerduring the tests.

2.3.3. Membrane morphologies

A JOEL JSM-5610 scanning electron microscope (SEM) wasused to investigate membrane structures. Samples were cutcarefully to ensure a clean break for a clear observation. Thesamples were then mounted onto sample stubs with the surfaceto view facing up. These samples were then spur-coated with goldusing an SEM sputter coater before being viewed with thescanning electron microscope.

2.4. Oxygen permeation of BYS–LSM single-layer membrane

Oxygen permeation of BYS–LSM single layer membrane wasmeasured between 650 1C and 900 1C, using the system shownschematically in Fig. 2(a) [19]. Prior to testing, the membrane wassealed between two dense alumina tubes using high temperatureceramic sealant (Resbond 940 LE). The furnace was heated up at arate of 5 1C/min from room temperature to target temperatures.Ambient air was used as the source of oxygen and pure argon wasused as a sweep gas feeding into the lumen of BYS–LSM hollowfibre membrane. The oxygen concentration in the permeatestream was measured using an oxygen analyser (Model 572with70.01% resolution, Servomex). Possible gas leakage can bedetected by co-monitoring the flow rate of permeates andabnormal changes in oxygen concentrations. The oxygen concen-tration data were collected 10 min after the setting temperaturewas stable. The oxygen permeation flux of single-layer membranewas calculated using the following equation:

JO2¼

VtXO2

Atð1Þ

where JO2 is the permeation flux of oxygen (ml min�1 cm�2), Vt isthe flowrate of the effluent stream (ml min�1), XO2 (%) is the

oxygen concentration obtained from the oxygen analyser and At isthe effective membrane area (cm2).

2.5. Oxygen permeation and catalytic reaction of BYS–LSM/BYS–Ni

dual-layer membranes

Similar to the single-layer counterpart, the dual-layer mem-branes were tested at the temperature range of 600 to 900 1C,using ambient air as the source of oxygen Fig. 2(b). The inlet gaseswere controlled using mass flow controllers (Model 0154, BrooksInstrument). Prior to the performance tests, argon was replacedby diluted hydrogen (50% in Ar) at 400 1C to reduce NiO to Ni. Thisreduction process was required in order to make the inner layerporous and allow the permeation of oxygen. After 15 min ofreduction, the dilute hydrogen was replaced by carbon monoxide(10% in argon) to start the oxygen permeation process for duallayer membranes. CO was used to avoid re-oxidation of Ni intoNiO. The gas flowrate from the outlet of the hollow fibre wasmeasured using a bubble flowmeter. The permeate stream wason-line analysed using a gas chromatographs (Varian 3900). Theoxygen permeation flux was calculated based on mass balance ofoxygen species, i.e. the amount of CO consumed and the amountof CO2 produced.

For catalytic reactions, CH4, instead of CO, was used after thereduction. Oxygen analyser was also used to monitor possibleleaking from sealing and system. The conversion of methane XCH4

and selectivity (Si) are calculated using the following equations

XCH4¼

FCH4inlet�FCH4outlet

FCH4 inlet� 100% ð2Þ

Si ¼niFi outlet

FCH4 inlet�FCH4outlet� 100% ð3Þ

where ni is the number of carbon atoms in species i.

3. Results and discussion

3.1. Macrostructures of fresh single- and dual-layer membranes

Fig. 3 shows SEM images of the sintered BYS–LSM single-layerhollow fibre membrane. A sandwich-like structure that consists ofa central sponge-like layer and finger-like structures on both sideswas formed, due to the short air gap used. The outer diameter ofthe fibre is approximately 1.4 mm, with a wall thickness about220 mm. The thickness of the sponge layer in the central part ofthe fibre is around 63 mm, which is about 1/3 of the fibre wallthickness. In general, the formation of finger-like void is attrib-uted to the viscous-fingering phenomenon occurred from mem-brane surfaces where more viscous spinning suspension is incontact with less viscous coagulants [20]. The growth of finger-like voids proceeds with solvent–non-solvent exchange that leadsto the precipitation of polymer binder, until the central regionwhere the local viscosity increases (due to the flow-out of solventto both sides) to a level that further growth of finger-like voidscannot proceed.

Fig. 4 illustrates SEM images of dual-layer membrane. Theouter and inner diameters of the dual layer-fibre are about1.04 mm and 0.68 mm, respectively. It can be seen that, due tothe longer air gap used, finger-like structure can only be formedfrom the inner surface. However, due to the high co-sinteringtemperature that is quite close to the melting point of BYS, suchfinger-like structure was not survived. The thickness of the BYS–LSM separation layer is about 32 mm, while the inner layer isaround 150 mm. A thinner separation layer is preferred in orderfor a higher oxygen permeation flux, when bulk diffusion still

Catalytic reactionCH4 Product stream

BYS-NiO (catalytic layer)BYS-LSM(oxygen separation layer)

Arg

on

Temperature controller

Furnace

Oxygen analyzer

Soapbubble

flowmeter

Quartz wool

Silicone sealant

Ceramic sealant

b) Dual layer-hollow fibre membrane reactor

Alumina tube

10%

CH

4 /

Ar

H2

10%

CO

/ A

r

GC

Argon + O2

BYS-LSM

Argon

O2-

e-

e-

O2-

a) Single layer-hollow fibre membrane

Valve

Fig. 2. Schematic diagram of (a) oxygen permeation system for single-layer membrane, and (b) oxygen permeation and catalytic reaction system for dual-layer membrane.

Fig. 3. SEM images of BYS–LSM single-layer membrane: (a) whole view (b) cross-section.

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–65 61

affects oxygen permeation. In Fig. 4(c), there is no observable gapat the interface between the inner and outer layers, indicatinggood adhesion between the two layers. In addition, a clear dual-phase structure of the inner layer can be seen in Fig. 4(d),indicating good dispersion of the two phases (BYS and NiO).

3.2. Single-layer BYS–LSM membrane for oxygen permeation

Fig. 5 illustrates oxygen permeation of the single-layer mem-brane, using argon as the sweep gas. It can be seen that theoxygen permeation flux increases with the increase of tempera-tures up to 900 1C. The oxygen permeation at 850 1C is approxi-mately 1.0 ml min�1 cm�2, which is slightly higher than theprevious study [16] (0.52 ml min�1 cm�2) at the same operatingtemperature, and is much higher than ScSZ–LSM and YSZ–LSM(less than 0.2 ml min�1 cm�2). This demonstrates the advantage

of using BYS as a more ionic conductive phase in this study topromote oxygen permeation, besides better chemical compat-ibility with LSM when compared with ScSZ and YSZ [4,5], as noadditional impurity phase was observed.

3.3. BYS–LSM/BYS–Ni DL-CHFMR for chemical reactions

3.3.1. Sintering behaviour of membrane materials

Sintering is a crucial step in the fabrication of ceramic mem-branes, as it affects the final structure and performance of themembranes. The sintering of ceramic hollow fibre membranes ismore difficult due to the higher surface area and curvity whencompared to other membrane geometries (planar and tubular). Anumber of parameters such as sintering temperature, dwellingtime, heating and cooling rates need to be considered, especiallyfor the membranes with a dual-layer structure. Each of the

NiO

Finger-likevoids

BYS-LSMBYS-NiO

Denselayer

Fig. 4. SEM images of dual-layer membrane: (a) whole view (b) cross section (c) interface between two layers (d) inner surface.

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

650 700 750 800 850 900

Oxy

gen

perm

eatio

n (m

l·min

-1·c

m-2

)

Temperature (°C)

Fig. 5. Oxygen permeation of single-layer hollow fibre membrane.

0

200

400

600

800

1000

1200

-25

-20

-15

-10

-5

0

5

0100200300400500Te

mpe

ratu

re ( °

C)

dL/L

o (%

)

Time (min)

0

200

400

600

800

1000

1200

-18

-16

-14

-12

-10

-8

-6

-4

-2

0

2

0 100 200 300 400 500

Tem

pera

ture

(°C

)

dL/L

o (%

)

Time (min)

BYSLSMNiOSintering profile

I II III

Fig. 6. Sintering behaviour of membrane materials (- - - represents sintering

profile).

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–6562

membrane materials has a different sintering behaviour andthermal expansion when subjected to a heat treatment; thereforethe understanding of thermal behaviours of each material isimportant. An optimum sintering condition is favoured to ensurethat the dual-layer membranes can survive the co-sinteringprocess without the formation of cracks or delamination of theouter separation layer.

Fig. 6 illustrates sintering curves of the membrane materials at1100 1C. The temperature of 1100 1C was selected because it isclose to the melting point of BYS. In general, a sintering processconsists of three different regions. The first (I) region is from roomtemperature to 600 1C, in which ceramic materials expandslightly and no significant shrinking phenomenon is observable.In region (II), BYS and NiO start to shrink at a much lowertemperature (around 650 1C) than that of LSM. The final shrinkageincreases in the order of LSM (5.6%), NiO (14.2%) and BYS (15.5%).

The rate of shrinking slows down in dwelling (30 min) andcooling stage for all the materials (region III).

In term of correlating the shrinking behaviour analysis withthe co-sintering of dual-layer membranes, the gap in finalshrinkage between the inner layer and outer layer should beminimised in order to avoid the formation of cracks or defects.This means that, for the BYS–LSM/NiO–BYS system, less BYSshould be used in both two layers to match the sinteringbehaviours in a better way, as long as the functions of each layer,such as oxygen separation of the outer layer and catalytic activityof the inner layer are not affected. The importance of selectingappropriate material compositions has been highlighted in pre-vious study [5], as it is one of the factors that governs the co-sintering process of the dual-layer membrane. In this study,the outer separation layer of the membrane is composed of adual-phase material (BYS–LSM), while the inner layer consists of

0

200

400

600

800

1000

1200

-25

-20

-15

-10

-5

0

5

0 100 200 300 400 500

Tem

pera

ture

(°C

)

dL/L

o (%

)

Time (min)

BYS-LSM (50v%:50v%)BYS-LSM (60v%:40v%)BYS-NiO (40v%:60v%)BYS-NiO (60v%:40v%)Sintering profile

Fig. 7. Sintering behaviour of outer and inner layer materials (- - - represents

sintering profile).

5 15 25 35 45 55 65 75

Rel

ativ

e in

tens

ity (c

ount

s)

2 Theta (°)

Single layer

Dual layer

BYS

LSM

NiO

Fig. 8. XRD patterns of single- and dual-layer hollow fibre membranes.

0

2

4

6

8

10

12

14

16

18

20

0.00

0.20

0.40

0.60

0.80

1.00

1.20

600 650 700 750 800 850

Met

hane

con

vers

ion

(%)

Oxy

gen

perm

eatio

n (m

l min

-1cm

-2)

Temperature (°C)

O2 permeation (calculated from CH4 conversion)O2 permeation (using CO as sweep gas)CH4 conversion

Fig. 9. Oxygen permeation and catalytic reaction of dual-layer hollow fibre membrane.

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–65 63

BYS–NiO. BYS used in both layers can be considered as themedium that facilitated the matching of the sintering behavioursbetween two layers, which is similar to the role of YSZ and ScSZemployed in our previous studies [4,5].

Fig. 7 illustrates sintering curves of the outer and inner layermaterials at selected compositions (volume %). It can be seen thatthe inner and outer layers start to shrink at similar temperatures.However, final shrinkages of the inner layer were approximately4–5% higher than that of the outer layer. Since the difference insintering behaviour between 40% BYS–60% NiO and 60% BYS–40%NiO is not significant, and more NiO in the inner layer may lead totwo potential issues after being reduced into Ni, i.e. more cokeformation during methane conversion and lower mechanicalstrength, the latter one was used as the inner layer composition.In addition, a larger amount of BYS leads to a higher finalshrinkage of the inner layer, which is opposite to the outer layer.Such a difference can be a net result of several factors, such asgreen density, phase ratio, particle size and shape and interac-tions between the two phases etc. As a result, a mixture of 50%BYS–50% LSM and 60% BYS–40% NiO was used for membranefabrication. Composition of the corresponding suspension needsto be further adjusted based on viscosity (Table 1), as an appro-priate viscosity of spinning suspension is essential to ensure aproper co-extrusion process.

The XRD patterns of BYS–LSM single-layer fibre and BYS-LSM/BYS–NiO dual-layer fibre (crushed into powder) are compared asshown in Fig. 8. No additional peaks are observed in single layerfibre, apart from the one of BYS and LSM (the membrane wassintered at 1100 1C) that is in good agreement with the structureof membrane materials. This indicates a high level of chemicalcompatibility between BYS and LSM, which is another advantageover ScSZ and YSZ as they react slightly with LSM at hightemperatures. It is the same to the dual-layer membrane whereno additional peaks present apart from the membrane materials(BYS, LSM and NiO), when subjected to the sintering at 1040 1C.

3.3.2. Oxygen permeation and catalytic reactions of

dual-layer membrane

Oxygen permeation measurement of the dual-layer membranewas carried out using CO as the sweep gas to avoid re-oxidation ofnickel. The dual-layer membrane cannot not survive testingtemperatures higher than 850 1C, which may be related to thedestabilization of BYS at higher temperatures after reduction.

As can be seen in Fig. 9, the oxygen permeation flux increaseswith increasing temperatures, the value of which is very close tothe single layer membrane. For example at 850 1C, the oxygenpermeation fluxes of both single-layer (Fig. 5) and dual-layermembranes are approximately 1.0 ml min�1 cm�2, although theseparation layer of dual-layer membranes is thinner. One of thepossible reasons for this can be the different of membranestructures. For dual-layer membranes, although its separationlayer is thinner, and CO was used as the sweep gas that shouldlead to higher permeation flux due to lower oxygen partialpressure on the permeate side, NiO in the inner layer may notbe fully reduced into Ni at 400 1C (50% H2 for 15 min). This wouldgenerate additional mass transfer resistance in the inner layer foroxygen permeation, and consequently lower oxygen permeationflux. Another possible reason can be due to the instability of BYSwhen it is in contact with strong reducing gas, which will bediscussed later. The possible change in BYS structure during thereduction (prior to oxygen permeation measurement) may affectits ionic conductivity, and subsequently lowering the oxygenpermeation flux.

Concerning the catalytic reactions in the DL-CHMR, methane isused as the reactant with the operating temperature no higherthan 750 1C, above which the dual-layer membrane starts tofracture. This fracture may be related to a number of factors,

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–6564

such as thermal and chemical stability of membrane materials atcertain reaction conditions and temperature difference across themembrane layers due to catalytic reactions, and will be furtherinvestigated. As can be seen in Fig. 9, the methane conversion isaround 14% in the temperature range of 600–750 1C. The calcu-lated oxygen permeation due to methane conversion is signifi-cantly lower than the oxygen permeation discussed above. Thisindicates a significant change of membrane materials under thereaction conditions.

A further XRD analysis was carried out to investigate thestability of BYS in inert (argon) and reducing atmosphere (hydro-gen) to understand the changes of BYS in the oxygen permeationand catalytic reaction (dual-layer membranes). In Fig. 10, it can beseen that there is no evident change when BYS is exposed toArgon (800 1C). In contrast, there is a significant change when BYSwas exposed to H2 at 400 1C. In general, the exposure of BYS to H2

can cause distortion of the ideal BYS fluorite structure, althoughdiluted hydrogen and a short reduction period were used. Thissuggests relatively low chemical stability of BYS in H2, and furtherin contact with CH4 or CO. This is very similar to 25 mol% yttria-doped bismuth oxide (BY25), where its fluorite structure andcatalytic activity can be maintained as long as stoichiometricoxygen is not consumed. Otherwise, its fluorite structure will bedestroyed because of the reduction of metallic ions [21]. Due tothe chemical instability of BYS in DL-CHFMR, CO2 is the majorproduct of methane conversion, as Ni in the inner layer cannotwork together with an ionic conducting phase to produce syngas(COþH2), although both Ni and NiO exist in the inner layer(Fig. 10). Previous study on oxidative coupling methane (OCM)reaction by Akin [22] showed that, the bismuth element migratedtowards membrane surface that was exposed to the reactants,although such bismuth migration did not affect the structure andcatalytic properties of the BYS membrane. In our study, thebismuth destabilization and migration due to the contact withH2 resulted in the deactivation of the inner catalytic layer. As aresult, it is believed that the CH4 conversion of less than 15% ismainly due to combustion, which is also the major reason forlower oxygen permeation in catalytic reactions (Fig. 10).

All these indicate that strong reducing atmospheres need to beavoided when BYS is used as a membrane material. But due to thegreat ionic conductivity and oxygen permeation of BYS at inter-mediate operating temperatures, similar reactor design can beused for other catalytic reactions of great importance, such asOCM where BYS shows great catalytic activity [23,24]. The use of

5 15 25 35 45 55 65 75

Rel

ativ

e in

tens

ity (c

ount

s)

2 Theta (°)

BYS-NiO in hydrogen

BYS-NiO in argon

BYS in hydrogen

BYS in argon

NiONiImpurity phases

Fig. 10. XRD patterns of BYS powders and BYS–NiO inner layer in inert and

reducing environments.

BYS as membrane material is expected not only to improve themembrane reactor performances, but eliminates the requirementof additional catalyst such as Ni because BYS itself is a catalyst forOCM. The conversion of methane into valuable C2 products(ethane and ethylene) through OCM has been widely studiessince 1980s. Although it has been demonstrated that the catalyticmembrane reactor can improve the methane coupling reaction,research in this area is progressing slowly, hampered by mem-brane materials and reactor performance issues.

An early investigation of the catalytic activity of BYS towardsOCM reaction was carried out using the BYS-based DL-CHFMR. SinceH2 reduction can no longer be used, the inner layer maintains adense structure, which stops the permeation of oxygen from theouter layer. As a result, a mixture of CH4 and O2 (2:1) was fed intothe lumen of the dual-layer fibre and tested between 800–900 1C,without the reduction step. This time, the fibre can withstandtemperatures up to 900 1C, much higher than POM (750 1C) as thefibre was not exposed to strong reducing atmosphere (H2). The C2

yield obtained after 2 h of stabilisation at 900 1C is about 0.75%, witha C2 selectivity of 9.40%. The performance can be further improvedwhen the inner layer is porous or sintered at a lower temperature, asa fixed bed reactor packed with BYS powder gave a much higher C2

yield (11.3%) and C2 selectivity (41.4%) under the same operatingconditions. This indicates that such BYS-based DL-CHFMR designcan be used for catalytic reactions where no strong reducing gas isinvolved, although the membrane material for the outer oxygenseparation layer needs to be changed accordingly, in order to matchsintering temperatures and operating temperature for both oxygenpermeation and OCM reaction. For catalytic reactions that requirepure oxygen, this DL-CHFMR with thin separation layer and porouscatalytic layer is preferred to ensure a great membrane reactorperformance.

4. Conclusions

BYS, a great ionic conducting ceramic at intermediate tempera-tures, has been employed in this study to replace ScSZ and YSZ andfabricate dual-layer ceramic hollow fibre membranes for catalyticreactions, resulting in better chemical compatibility with LSM andsignificantly higher oxygen permeation. This demonstrates theadvantages of using BYS in fabricating DL-CHFMR, especially interms of supplying more oxygen for possibly better reactor perfor-mance at lower operating temperatures. The insufficient chemicalstability of BYS in strong reducing atmosphere changes its structureand consequently catalytic activity of the inner layer, and affects thereactor performance for methane conversion. However, the advan-tages of such reactor design have been proved and can betransferred to other important catalytic reactions, such as OCMwhere no strong reducing environment is required.

Acknowledgement

The authors gratefully acknowledge the research funding providedby EPSRC in the United Kingdom (Grant no. EP/G012679/1) and NurHidayati Othman would like to extend her gratitude to MalaysianMinistry of Higher Education (MoHE) and Universiti Teknologi MARA(UiTM) for generous study bursary.

References

[1] K. Li, X. Tan, Y. Liu, Single-step fabrication of ceramic hollow fibers for oxygenpermeation, J. Membr. Sci. 272 (2006) 1–5.

[2] Q. Li, X. Zhu, W. Yang, Single-step fabrication of asymmetric dual-phasecomposite membranes for oxygen separation, J. Membr. Sci. 325 (2008)11–15.

N.H. Othman et al. / Journal of Membrane Science 432 (2013) 58–65 65

[3] J.De Jong, Towards single step production of multi-layer inorganic hollowfibers, J. Membr. Sci. 239 (2004) 265.

[4] Z. Wu, B. Wang, K. Li, Functional LSM–ScSZ/NiO–ScSZ dual-layer hollow fibresfor partial oxidation of methane, Int. J. Hydrogen Energy 36 (2011)5334–5341.

[5] Z. Wu, B. Wang, K. Li, A novel dual-layer ceramic hollow fiber membranereactor for methane conversion, J. Membr. Sci. 352 (2010) 63–70.

[6] M.H.D. Othman, Z. Wu, N. Droushiotis, U. Doraswami, G. Kelsall, K. Li, Single-step fabrication and characterisations of electrolyte/anode dual-layer hollowfibres for micro-tubular solid oxide fuel cells, J. Membr. Sci. 351 (2010)196–204.

[7] M.H.D. Othman, N. Droushiotis, Z. Wu, K. Kanawka, G. Kelsall, K. Li, Electro-lyte thickness control and its effect on electrolyte/anode dual-layer hollowfibres for micro-tubular solid oxide fuel cells, J. Membr. Sci. 365 (2010)382–388.

[8] M.H.D. Othman, Z. Wu, N. Droushiotis, G. Kelsall, K. Li, Morphological studiesof macrostructure of Ni–CGO anode hollow fibres for intermediate tempera-ture solid oxide fuel cells, J. Membr. Sci. 360 (2010) 410–417.

[9] M.H.D. Othman, N. Droushiotis, Z. Wu, G. Kelsall, K. Li, Novel fabricationtechnique of hollow fibre support for micro-tubular solid oxide fuel cells,J. Power Sources 196 (2011) 5035–5044.

[10] M.H.D. Othman, N. Droushiotis, Z. Wu, G. Kelsall, K. Li, Dual-layer hollowfibres with different anode structures for micro-tubular solid oxide fuel cells,J. Power Sources 205 (2012) 272–280.

[11] W. Yang, Development and application of oxygen permeable membrane inselective oxidation of light alkanes, Top. Catal. 35 (2005) 155–167.

[12] Z. Wu, A. Thursfield, I. Metcalfe, K. Li, Effects of separation layer thickness onoxygen permeation and mechanical strength of DL-HFMR-ScSZ, J. Membr.Sci., 415–416 (2012) 229-236.

[13] K. Li, Ceramic Membranes for Separation and Reaction, 2007.[14] S.C. Singhal, K. Kendall, High-temperature Solid Oxide Fuel Cells: Funda-

mentals Design and Applications, 2003.[15] Y. Zeng, Synthesis and properties of copper and samarium doped yttria-

bismuth oxide powders and membranes, J. Mater. Sci. 36 (2001) 1271–1276.[16] C. Yang, Q. Xu, C. Liu, J. Liu, C. Chen, W. Liu, Bi1.5Y0.3Sm0.2O3–La0.8Sr0.2MnO3-d

dual-phase composite hollow fiber membrane for oxygen separation, Mater.

Lett. 65 (2011) 3365–3367.[17] Y. Zeng, F.T. Akin, Y.S. Lin, Oxidative coupling of methane on fluorite-

structured samarium–yttrium–bismuth oxide, Appl. Catal., A: General 213(2001) 33–45.

[18] B.F.K. Kingsbury, K. Li, A morphological study of ceramic hollow fibremembranes, J. Membr. Sci. 328 (2009) 134–140.

[19] X. Tan, Preparation of LSCF ceramic hollow-fiber membranes for oxygenproduction by a phase-inversion/sintering technique, Ind. Eng. Chem. Res.44 (2005) 61.

[20] B.F.K. Kingsbury, Z. Wu, K. Li, A morphological study of ceramic hollow fibremembranes: a perspective on multifunctional catalytic membrane reactors,

Catal. Today 156 (2010) 306–315.[21] Y. Zeng, Y.S. Lin, Stability and surface catalytic properties of fluorite-

structured yttria-doped bismuth oxide under reducing environment, J. Catal.182 (1999) 30–36.

[22] F. Akin T., Y. Lin S., Oxidative coupling of methane in dense ceramicmembrane reactor with high yields, AlChE J. 48 (2002) 2298–2306.

[23] Y. Zeng, Y.S. Lin, Oxidative coupling of methane on improved bismuth oxidemembrane reactors, AlChE J. 47 (2001) 436–444.

[24] F.T. Akin, Controlled oxidative coupling of methane by ionic conductingceramic membrane, Catal. Lett. 78 (2002) 239–242.