Pp

Xa

b

c

a

ARRAA

KOHPM

1

ebphpLbihsttot

b

0d

Journal of Membrane Science 352 (2010) 189–196

Contents lists available at ScienceDirect

Journal of Membrane Science

journa l homepage: www.e lsev ier .com/ locate /memsci

ilot-scale production of oxygen from air usingerovskite hollow fibre membranes

iaoyao Tana,b,∗, Zhigang Wangb, Bo Mengb, Xiuxia Mengb, K. Li c

School of Environmental and Chemical Engineering, Tianjin Polytechnic University, Tianjin 300160, ChinaSchool of Chemical Engineering, Shandong University of Technology, Zibo 255049, ChinaDepartment of Chemical Engineering and Technology, Imperial College London, South Kensington, London SW7 2AZ, UK

r t i c l e i n f o

rticle history:eceived 19 October 2009eceived in revised form 1 January 2010ccepted 4 February 2010vailable online 11 February 2010

eywords:xygen productionollow fibre

a b s t r a c t

La0.6Sr0.4Co0.2Fe0.8O3−˛ (LSCF) hollow fibre membranes prepared by a phase-inversion/sintering tech-nique were assembled into a membrane system to produce oxygen of high purity (>99%) from air atelevated temperatures. The separation performances, stability, scaling-up effect and the energy consump-tion of the membrane system were investigated both theoretically and experimentally. The membranesystem containing 889 hollow fibres could yield maximum 3.1 L(STP) min−1 oxygen with the purity of99.9% at 1070 ◦C and 98.5 kPa vacuum degree, but the temperature higher than 1070 ◦C would lead tothe system failure. It showed the potential of much higher production rates only if the high-temperaturesealing problem could be solved. When operated at around 960 ◦C, the system exhibited more than 1167 h

−1

erovskiteixed conducting membrane

longevity with the oxygen production rate of 0.84 L(STP) min and oxygen purity of 99.4%. The energyconsumption of the system increased with operating temperature but the energy consumption per unitoxygen product decreased with increasing the operating temperature and the effective membrane areas.In order to reduce the oxygen cost to commercial level, heat exchangers have to be integrated in themembrane system to recover the heat energy in both the exhaust gas and the oxygen product. The oxy-gen recovery should be limited within 20–40% for the sake of both energy and membrane area savings.

. Introduction

Oxygen production from air is of great importance in bothnvironmental and chemical industries. It is usually achievedy cryogenic distillation, pressure swing adsorption (PSA) or byolymeric membrane separation. These processes are either ofigh energy consumption or unable to produce oxygen of highurity. Alternately, dense mixed conducting membranes such asa1−xSrxCo1−yFeyO3−˛ (LSCF) perovskite which exhibit apprecia-le oxygen ionic and electronic conductivity have become of great

nterest as a potentially economical, clean and efficient means ofigh pure oxygen production [1–7]. When an oxygen partial pres-ure gradient is imposed across such dense membranes at a highemperature (usually >700 ◦C), the oxygen may be transferred fromhe high partial pressure side to the low partial pressure side with-

ut the need of electrodes and external electrical loadings, makinghe membrane system and operation much simplified.

In recent years, perovskite hollow fibre membranes haveeen successfully prepared by a phase-inversion/sintering process

∗ Corresponding author. Tel.: +86 533 2786292; fax: +86 533 2786292.E-mail address: cestanxy@yahoo.com.cn (X. Tan).

376-7388/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2010.02.015

© 2010 Elsevier B.V. All rights reserved.

[8–12]. Such hollow fibre membranes can provide much largerareas per unit volume than other configurations such as flat sheetor tubular, thus it is possible to reduce the membrane systemsize remarkably. Furthermore, the hollow fibre configuration alsomakes the high-temperature sealing less problematic in fabri-cating membrane modules [13–15]. In this work, we developeda perovskite hollow fibre membrane system that could produceover 3 L(STP) min−1 oxygen gas with the purity of >99%. The sep-aration performances, the stability, the scaling-up effect and theenergy consumption of the perovskite membrane systems havebeen investigated both theoretically and experimentally.

2. Experimental

2.1. LSCF powder and hollow fibre membranes

La0.6Sr0.4Co0.2Fe0.8O3−˛ (LSCF) hollow fibre membranes wereprepared by a phase-inversion/sintering technique using the pow-

ders synthesized through a sol–gel combustion process which weredescribed elsewhere [8,12]. The resultant hollow fibre membraneswere usually ca. 28–32 cm in length and around 0.12/0.18 cm ini.d./o.d. One end of the hollow fibres was closed with the sameLSCF material and the gas-tight property was achieved after sin-

190 X. Tan et al. / Journal of Membrane Science 352 (2010) 189–196

hollo

tcaacc

2

imqtM3lttTpot(wblr

suiswauTmpA

troller with a K-type thermocouple positioned in the middle of thefurnace. Meanwhile, a temperature gauge fitted with another K-type thermocouple was used to monitor the temperature at thesealing points. In the operation, the oxygen product flow rate andthe concentration of the oxygen product were recorded with time

Fig. 1. Photos of (a) the LSCF hollow fibre membranes; (b) the

ering. Surface modification was conducted on the hollow fibres byoating catalysts on the outer surface so as to improve the perme-tion flux of the hollow fibre membranes. The catalyst consisted ofperovskite with high Co content and an Ag paste with the parti-

le size <7 �m and was mixed in weight ratio of 6:1. The detailedoating procedures were described elsewhere [16].

.2. Hollow fibre membrane module and system

In practical applications, the hollow fibres must be assemblednto a membrane module. For the sake of sealing and facile replace-

ent of broken fibres, we first bundled 7 fibres together in auartz tube (8 mm in diameter and 70 mm in length) with a high-emperature silicone sealant (1592, purchased from Tonsan New

aterials and Technol. Co., Beijing) that is able to withstand up to50 ◦C. The fibre bundles as required were then placed in a stain-

ess steel holder. Fig. 1 shows the LSCF hollow fibre membranes,he fibre bundles and the membrane module. It is very importanto avoid any leakage in order to obtain highly pure oxygen product.herefore, each hollow fibre was individually tested to be gas-tightrior to bundling and every bundle was again tested to be freef leakage before integrating them into a module. In this work,hree modules containing 1 bundle (7-fibre module), 9 bundles63-fibre module) and 129 bundles (889-fibre module) respectivelyere fabricated to study the scaling-up effect. The effective mem-

rane areas by subtracting the sealing length from the whole fibreength for these modules were ca. 78 cm2, 702 cm2 and 9914 cm2,espectively.

The hollow fibre membrane system for oxygen production ischematically shown in Fig. 2. The hollow fibre membrane mod-le was placed under a vertically positioned tubular furnace. An

nsulator was placed between the bottom inlet of furnace and thetainless steel holder so that the temperature at the sealing pointsas lower than that the sealant can withstand. A vacuum was

pplied to the membrane module using a claw rotor oil-free vac-

um pump (2ZBL, purchased from Beijing Huaxiajia Sci & Tech Inc.).he operating vacuum degree was controlled using a frequencyodulator (Suny3200, from Sunye Electric Co.) fitted to the vacuum

ump and was measured with a digital pressure sensor (PSA, fromutonics Co.). The oxygen concentration and the flow rate of the

w fibre bundles; and (c) the hollow fibre membrane module.

product were measured using an oxygen analyzer (TG-J310, fromXi’an Taige analysis instrument Inc.) and a digital film flow meter(SF-1U/2U, purchased from Horiba Stec Co. Ltd.), respectively. Theoperating temperature was controlled using a temperature con-

Fig. 2. Hollow fibre membrane system for oxygen production (A) flow chart; and(B) photo of the setup.

brane Science 352 (2010) 189–196 191

ufla

3

3

mtm(otsshntlawtogtdAaocf2oidAtowtpft11

Ft

X. Tan et al. / Journal of Mem

nder different operating conditions. All the values of volumetricow rates in this study were calibrated to the standard temperaturend pressure (STP).

. Results and discussion

.1. Production capacity of the membrane system

Production capacity of a membrane system is mostly deter-ined by the membrane’s effective permeation areas as well as

he permeability of single hollow fibre membrane. In this work, weainly used the membrane module containing 889 hollow fibres

called as 889-fibre system) to study the separation performancesf the perovskite membrane systems because it can reflect most ofhe problems involved in practical applications. On the other hand,ince the silicone sealant can only withstand maximum 350 ◦C, theealing points of the hollow fibre module or the joints between theollow fibres and the quartz tube have to be placed out of the fur-ace. As the hollow fibre module is inserted deeper in the furnaceube, the effective fibre length for oxygen permeation would getonger and the oxygen production rate could be increased due tolarger permeation area provided. Meanwhile, the sealing pointsould also be closer to the furnace’s bottom inlet leading to higher

emperature and failure of sealing on these points. Fig. 3 shows thexygen production rate of the 889-fibre system with different airaps, which is referred to the distance from the sealing points tohe bottom inlet of the furnace. As can be seen, the oxygen pro-uction rate was increased noticeably by decreasing the air gap.t the air gap of 1.5 cm, the maximum oxygen production ratet 1070 ◦C reached to 3.10 L min−1 with the oxygen concentrationf 99.88%. However, the temperature at the sealing points in thisase reached to as high as 235 ◦C and the operation only lastedor about 2 h. A fast decrease in oxygen purity was observed afterh operation. The post-mortem analysis indicated that the leakageccurred at the sealing points of several fibre bundles. In addition,t was also found that there were several hollow fibres collapsedue to the sucking force of vacuum under such a high temperature.fter the air gap was increased to 3.0 cm, the highest tempera-

ure at the sealing points decreased to 173 ◦C but the maximumxygen production rate at 1070 ◦C also decreased to 2.36 L min−1

ith the oxygen concentration of 99.71%. Even now, the separa-ion lasted only for about 20 h after which leakage at the sealingoints occurred again. In order to prevent the membrane system

rom failure due the leakage, the air gap was set to 4.5 cm so thathe temperature at the sealing points would not be higher than30 ◦C. Under this air gap, the maximum oxygen production rate at070 ◦C only reached 1.79 L min−1 with the oxygen concentration

ig. 3. Oxygen production rate of the 889-fibre membrane system as a function ofemperature with different air gaps (vacuum degree = 98.9 kPa).

Fig. 4. Performance of the membrane system at different temperatures (889-fibremodule, vacuum degree = 97.4 kPa) (�,� for the first cycle; �, � for the second cycle).

of 99.58%. Since there existed a noticeable temperature gradientfrom the center to the bottom inlet of the furnace tube, the effec-tive heating length of the fibres for oxygen permeation could bemuch shorter than the apparent length of the fibres. As the air gapwas increased from 1.5 cm to 4.5 cm, the effective heating lengthwould also be shortened by 3 cm. Consequently, the actual per-meation area was significantly decreased leading to the noticeabledecrease in oxygen production rate although the air gap was smallcompared to the total fibre length. In the following experiments,the operating temperature would not be higher than 1060 ◦C so asto avoid the membrane breakage due to over-sintering.

3.2. Effect of operation parameters

Performances of the 889-fibre membrane system at differenttemperatures are shown in Fig. 4 where the applied vacuum waskept at 97.4 kPa. Two operating cycles, which is defined as a wholeoperating process from heating up the membrane system fromroom temperature, collecting experimental data at high temper-atures to cooling the system back to ambient temperature, wereconducted under the same conditions. As can be seen, the twocycles exhibited the same increasing trend with temperature forboth the production rate and the product purity, and the data forthe same temperatures had almost the same values. This indi-cated that the repeatability of the system operation was good.As is expected, the oxygen production rate and the product con-centration increased with increasing temperature since a highertemperature facilitated both the surface exchange reactions andthe bulk diffusion. When the system was operated at 960 ◦C, theoxygen production rate and the oxygen purity reached 0.86 L min−1

and 99%, respectively. As the temperature was increased to 1060 ◦Cthe oxygen production rate also increased to 1.66 L min−1 with thecorresponding oxygen concentration of 99.67%. The oxygen con-centration in product did not reach to the theoretical value (100%)because of minor leakage of the membrane system which led tothe entry of some nitrogen into the product stream. At room tem-perature, oxygen permeation in the perovskite membrane wouldnot occur but there was still some gas could be collected from theoutlet of the membrane module under vacuum operation. This gaswas mostly the air entering into the product side due to the systemleakage. The flow rate of this air leaking was measured at room tem-perature using a gas bubble flow meter to be about 4.7 mL min−1.But in operation, the amount of air leaking to the product streamwas calculated from the production rate and the product concen-

−1

tration as 6–15 mL min depending on the operating temperature.The leakage may be produced either from the tiny defects pre-sented on the hollow fibre membranes or from the silicone sealingjoints. For the present membrane system the oxygen productionrate and the product concentration could reach above 99% and

1 brane Science 352 (2010) 189–196

aw

atpTgtviipu9ettt

3

b9dttbatppaup

3

m

Fs

92 X. Tan et al. / Journal of Mem

bove 1.1 L min−1, respectively, only if the operating temperatureas higher than 1050 ◦C.

Fig. 5 shows the oxygen production rate and the product puritys a function of vacuum degree under different temperatures. Bothhe oxygen production rate and product purity at a given tem-erature increased with increasing the applied vacuum degree.his result can be expected because the driving force for oxy-en permeation was increased. At lower temperatures (<900 ◦C),he product concentration increased rapidly with increasing theacuum degree. But the promotion of product purity by increas-ng vacuum degree would be weakened as the temperature wasncreased. On the other hand, at lower vacuum degrees, the oxygenroduction rate can be improved noticeably with increasing vac-um degree. However, when the vacuum degree was higher than9 kPa, the increase in oxygen production rate would be lessenedspecially at higher temperatures. Therefore, it is not necessaryo apply the vacuum degree higher than 99 kPa because it con-ributes little to improving the oxygen production rate but leadso appreciable increase in operation cost.

.3. Long-term test

In order to measure the stability of the hollow fibre mem-rane system, it was operated continuously for 1067 h at around60 ◦C and 97–98 kPa applied vacuum degree with the recordedata shown in Fig. 6. It should be mentioned that it had alreadyaken totally about 101 h with more than 20 cycles in 30 days forhe performance measurement prior to the longevity test. It cane seen that the production rate was kept at around 0.84 L min−1

nd the oxygen concentration in product was around 99.4%. Whenhe operating temperature was a bit lowered, the oxygen producturity and the production rate also decreased. But once the tem-erature was recovered to 960 ◦C, both the oxygen product puritynd the production rate were also recovered to the original val-es. Therefore, the present LSCF hollow fibre membrane systemossessed a good stability.

.4. Scaling-up effect

In order to examine the scaling-up effect of the hollow fibreembrane system, three modules containing 7 fibres (1 fibre bun-

ig. 5. Effect of vacuum degree on the performance of the hollow fibre membraneystem (889-fibre module).

Fig. 6. Performance of the 889-fibre membrane system as a function of operationtime.

dle in ˚20 mm×280 mm furnace tube), 63 fibres (9 bundles in˚60 mm×350 mm furnace tube) and 889 fibres (129 bundles in˚180 mm×600 mm furnace tube) were fabricated respectively.Fig. 7 shows the product concentration of the membrane systems atdifferent temperatures. It can be seen that the larger module couldyield purer oxygen product at the same temperature. For example,the oxygen product concentration of the 7-fibre membrane mod-ule only reached maximum 97.14% at 1060 ◦C, but the maximumproduct concentrations of the 63-fibre and the 889-fibre mem-brane system reached 98.9% and 99.71% at 1060 ◦C, respectively.This implies that the relative leakage to oxygen production rate inthe larger membrane system was lower than that in the smallerone. Furthermore, for all the three modules the product concentra-tion always increased with increasing temperature. Therefore, it ispossible to obtain highly pure oxygen products by increasing theproduction rate of the perovskite membrane systems.

However, the efficiency of the hollow fibre membrane systemwould be reduced remarkably due to the use of larger membranemodules. That is to say, the scaling-up effect of the hollow fibremembrane system was significant. As can be seen from Fig. 8, theaverage oxygen production rate of a single hollow fibre membranein the 889-fibre system is much lower than that in the 7-fibresystem especially at higher temperatures. For example, the aver-age oxygen production rate per hollow fibre in the 7-fibre systemwas 10.37 mL min−1 at 1060 ◦C but in the 889-fibre system it wasonly 1.86 mL min−1 at the same temperature, which was only 18%of the 7-fibre module’s average production rate. It is noted thateven in the 7-fiber system the average production rate was still

much lower than that calculated by the single fibre permeation test(1.85 mL cm−2 min−1 at 1000 ◦C [16]). Such significant decrease inthe average oxygen production rate in a single fibre mostly orig-inated from the uneven temperature distribution in the furnace

Fig. 7. Oxygen product concentration as a function of temperature for differenthollow fibre membrane systems.

X. Tan et al. / Journal of Membrane Science 352 (2010) 189–196 193

Fi

ttfitdtkiipprwauatibalt1rctufipBa

3

mtpeltsiitaoOt

with heat exchangers for heat recovery where the parameters forenergy calculation are also presented. The oxygen balance in themembrane system is given by:

0.21FAir = FOx + (FAir − FOx ) · xe (1)

ig. 8. Average oxygen production rate of a single fibre as a function of temperaturen different hollow fibre membrane systems.

ube. Due to the unavoidable temperature gradient from the centero the bottom inlet of the furnace tube, the length of the hollowbres served for oxygen permeation is actually much shorter thanhe total fibre length. Such loss of the effective permeation lengthue to the uneven distribution of temperatures would get more ashe furnace tube is enlarged. On the other hand, it is difficult toeep all the hollow fibre membranes in a large module contain-ng hundreds of fibres to have the same length, which can be seenn Fig. 1b. The shorter fibres actually contributed little to oxygenermeation especially in the large furnace having a significant tem-erature gradient. Consequently, the average oxygen permeationate over the whole fibre length in the large hollow fibre moduleould be much lower than that in the smaller one. In addition,

lthough the furnace tube was increased with the membrane mod-le size, the top outlet for exhaust gas was not enlarged as muchs the bottom inlet so as to achieve and keep the high tempera-ure in the furnace tube for permeation. As the permeation ratencreases, the oxygen concentration in air stream in the large mem-rane system may be lower than that in the smaller one leading toreduced permeation flux. Theoretically, if the whole hollow fibre

ength can be utilized for oxygen permeation, the oxygen produc-ion rate of the 889-fibre module can reach up to 22.9 L min−1 (or.38 m3 h−1) at 1000 ◦C, which is 19.5 times the present productionate, 1.16 L min−1. Therefore, in order to improve the productionapacity of the system, it is essential to eliminate or at least to lessenhe uneven distribution of temperatures for the membrane mod-le. A feasible way to utilize the whole membrane area in a hollowbre module for oxygen permeation is to preheat the air feed to theermeation temperature before it is introduced into the system.ut the high-temperature sealing will become a great challenge tossemble intact hollow fibre membrane modules.

.5. Energy consumption

The energy consumption of the membrane system which deter-ines the cost of oxygen product mainly includes the power for

he furnace to heat air feed to permeation temperature and theower for the vacuum pump to yield oxygen partial pressure gradi-nt. Generally, the power consumed by the vacuum pump is muchower than that by the furnace. This was also found to be true inhe operation of the present hollow fibre membrane system. Fig. 9hows the heating power of the furnace as a function of operat-ng temperature. As can be seen, the heating power of the furnacencreased with increasing operating temperature. This is becausehe oxygen production rate increased with increasing temperature

nd hence more energy was required to heat the air feed. More-ver, the heat loss also increased at a higher operating temperature.n the other hand, the higher oxygen production rate at a higher

emperature would lead to decrease in oxygen concentration in the

Fig. 9. Heating power of the furnace in the 889-fibre membrane system at differentoperating temperatures.

exhaust gas. Consequently, the volume of the exhaust gas as well asthe heat loss to produce a volume of oxygen product also decreasednoticeably. Therefore, the heating power per unit oxygen productdecreased with operating temperature as shown in Fig. 9. It sug-gests that increasing the operating temperature favors reducingthe cost of oxygen product. However, the energy consumption toproduce 1 Nm3 O2 for the present hollow fibre membrane systemwas still much higher (∼12 kWh even at 1050 ◦C) than that by thePSA process (∼0.7 kWh/Nm3). It contradicts to the general sensethat the ceramic membrane process can save a lot of oxygen pro-duction cost [18]. The primary reason for that is due to the failureto recover the heat energy from exhaust gas which will be analyzedin details as follows.

Fig. 10 shows the operation of the membrane system combined

Fig. 10. Diagram of the membrane system with heat exchange (A) between the airfeed and the oxygen stream; (B) between the air feed and the exhaust stream; and(C) between air feed and both the exhaust and oxygen streams.

194 X. Tan et al. / Journal of Membrane Science 352 (2010) 189–196

Table 1Parameters for calculating the energy consumption of oxygen product.

Membrane area Am = 1 m2

Outer diameter of the fibre membrane 0.18 cmInner diameter of the fibre membrane 0.12 cmEnhancement factor by catalytic modification ˛ = 2.0Operating vacuum degree 98.3 kPaAir feed flow rate 0.2 Nm3 min−1

Air feed temperature 25 ◦CTemperature of exhaust gas 50 ◦COperating temperature 850–1050 ◦CEfficiency of vacuum pump 0.8

−2(

8852.5)

2

wdtm

F

wppvtadte

t

(

(

(

p

W

w

Using the parameters in Table 1, the energy consumption of themembrane system with different membrane areas to produce onemolar oxygen product was calculated and plotted against operat-ing temperature in Fig. 11. For comparison, the oxygen cost by thePSA process is also presented with a dash line in the figure. As can

Diffusion coefficient of oxygen vacancy [7]

Forward reaction rate constant of Eq. (3) [7]

Reverse reaction rate constant of Eq. (3) [7]

here FAir and FO2 are the air feed flow rate and the oxygen pro-uction rate, respectively; xe is the oxygen fraction in the furnaceube and exhaust stream. For the catalytically modified hollow fibre

embrane, the oxygen production rate may be given by [17]:

O2 = Am · ˛ · kr [(p1xe)0.5 − (p2)0.5]

[(Rm/Ro) · p0.52 +(Rm/Rin) · (p1xe)0.5]+[((2kf (Ro − Rin))/DV ) · (p1xe · p2)0.5]

(2)

here Am is the membrane area of the hollow fibre module; ˛ theermeation enhancement factor by catalytic modification; p1 and2 are the upstream and downstream pressures (p2 = pa−operatingacuum degree where pa is atmospheric pressure), respectively; Rm

he logarithmic radius, Rm = (Ro−Rin)/ln(Ro/Rin), in which Ro and Rinre respectively the outer and the inner radius of the fibre; DV is theiffusion coefficient of oxygen vacancy; kf and kr are, respectively,he forward and the reverse reaction rate constants for the surfacexchange reaction:

12 O2 + V

••O

kf /kr←→OxO + 2h

•(3)

With the assumption of 10% heat loss of the membrane system,he heating power of the furnace may be given, respectively by,

A): heat exchange between the air feed and the oxygen productonly

Q1 =FO2 CpO2 · (50− 25)+ [(0.21FAir − FO2 )CpO2 + 0.79FAir CpN2 ] · (T − 25)

0.9

(4a)

B): heat exchange between the air feed and the exhaust streamonly

Q2 =FO2 CpO2 · (T − 25)+ [(0.21FAir − FO2 )CpO2 + 0.79FAir CpN2 ] · (50− 25)

0.9

(4b)

C): heat exchange between the air feed and both the exhaust andoxygen streams

Q3 =FO2 CpO2 · (T ′ − 25)+ [(0.21FAir − FO2 )CpO2 + 0.79FAir CpN2 ] · (50− 25)

0.9

(4c)

where Cpi is the specific heat capacity of gas species i; T′ isthe temperature of the air feed after heating by the exhauststream,

T ′ = (0.21FAir − FO2 )CpO2 + 0.79FAirCpN2

(0.21CpO2 + 0.79CpN2 )FAir· (T − 50)+ 25 (5)

The power of the vacuum pump corresponding to the oxygen

roduct rate may be estimated by:

= paVO2 · ln(pa/p2)�

= 2.48× 103FO2

�· ln

(pa

p2

)(6)

here � is the efficiency of the vacuum pump.

DV = 1.58× 10 exp − T (cm /s)

kf = 1.85× 104 exp(− 27291

T

)(cm/Pa0.5 s)

kr = 2.07× 104 exp(− 29023

T

)(mol/cm2 s)

Fig. 11. Energy consumption of the oxygen product by the membrane system withheat exchange (A) between the air feed and the oxygen stream; (B) between the airfeed and the exhaust stream; and (C) between air feed and both the exhaust andoxygen streams.

X. Tan et al. / Journal of Membrane

Fa

bngfpbmsugtabpw

�

peremnigts

4

sh39fpa1vt9otobm

ig. 12. Plot of the energy consumption of oxygen product and membrane areagainst oxygen recovery (air feed = 0.2 Nm3 min−1).

e seen, the oxygen production cost of the membrane process can-ot be reduced to the PSA cost level unless the heat in exhaustas is recovered by performing heat exchange between the aireed and the exhaust stream. In addition, the energy consumptioner unit oxygen product by the membrane system can be reducedy either increasing the operating temperature or increasing theembrane areas. Since the energy consumption of the membrane

ystem is mainly the residual heat along with the exhaust and prod-ct streams, it can be reduced by decreasing the amount of exhaustas for a given air feed. In another word, it is necessary to increasehe oxygen recovery so as to reduce the oxygen cost. This can bechieved by increasing the operating temperature and the mem-rane areas. Fig. 12 plots the energy consumption per unit oxygenroduct and the required membrane area against oxygen recoveryhich is defined as:

= FOx

0.21FAir× 100% (7)

It can be seen that the energy consumption per unit oxygenroduct decreases remarkably with increasing the oxygen recov-ry at lower recovery levels (<20%). However, after the oxygenecovery is higher than 40%, further increasing the oxygen recov-ry only leads to a slight decrease in energy consumption, but theembrane areas to achieve this recovery have to be increased sig-

ificantly especially for a lower operating temperature, resultingn the remarkable increase of membrane cost. Therefore, the oxy-en recovery should be limited within 20–40% so as to reducehe overall oxygen production cost of the perovskite membraneystems.

. Conclusions

A LSCF perovskite hollow fibre membrane system has beenuccessfully scaled-up to produce large quantity of oxygen withigh purity. The maximum oxygen production rate reached to.1 L min−1 with the corresponding oxygen concentration of above9.8%. The system possessed a good stability and can be operatedor more than 1167 h at around 960 ◦C to produce 99.4% oxygenroduct in production rate of 0.84 L min−1. The operating temper-ture for the LSCF hollow fibre membrane system cannot exceed070 ◦C otherwise the system breakage would occur. Increasingacuum degree favors the increase in oxygen production rate, buthis effect is alleviated after the vacuum degree is higher than9 kPa. For the commercialization of ceramic membrane technol-

gy, heat exchangers have to be integrated in the membrane systemo recover the heat energy in exhaust gas and oxygen products. Theperation cost of the perovskite membrane system can be reducedy either increasing the operating temperature or increasing theembrane areas. In order to reduce the overall oxygen produc-

[

Science 352 (2010) 189–196 195

tion cost of the perovskite membrane systems, the oxygen recoveryshould be limited within 20–40%.

Acknowledgements

The authors gratefully acknowledge the research funding pro-vided by the National High Technology Research and DevelopmentProgram of China (No. 2006AA03Z464), the National Natural Sci-ence Foundation of China (No. 20676073) and EPSRC in the UnitedKingdom (EP/E032079/1).

Nomenclature

Am effective membrane area for oxygen permeation(cm2)

Cp specific heat capacity of gas species (J/(mol K))DV effective diffusivity of oxygen vacancy (cm2/s)FAir air feed flow rate (mol/s)FO2 oxygen permeation rate (mol/s)kr reverse surface exchange reaction rate constant

(mol/(cm2 s))kf forward surface exchange reaction rate constant

(cm/(Pa0.5 s))pa atmospheric pressure (1.013×105 Pa)p1, p2 upstream and downstream pressure (Pa)Q1, Q2, Q3 heating power of the furnace (W)Rm algorithmic radius of fibre,

Rm = (Ro−Rin)/ln(Ro−Rin)Rin, Ro inner and outer radius of hollow fibre (cm)T operating temperature (K)T′ temperature of the air feed after heating by the

exhaust stream (K)W power of the vacuum pump (W)xe oxygen fraction in the furnace tube and exhaust

stream˛ permeation improvement factor due to the surface

modification� efficiency of the vacuum pump� oxygen recovery (%)

References

[1] V.V. Kharton, A.A. Yaremchenko, A.V. Kovalevsky, A.P. Viskup, E.N. Naumovich,P.F. Kerko, Perovskite-type oxides for high-temperature oxygen separationmembranes, J. Membr. Sci. 163 (1999) 307–317.

[2] J. Sunarso, S. Baumann, J.M. Serra, W.A. Meulenberg, S. Liu, Y.S. Lin, J.C.D. daCosta, Mixed ionic-electronic conducting (MIEC) ceramic-based membranesfor oxygen separation, J. Membr. Sci. 320 (2008) 13–41.

[3] M. den Exter, J.F. Vente, D. Jansen, W.G. Haije, Viability of mixed conductingmembranes for oxygen production and oxyfuel processes in power production,Energy Procedia 1 (2009) 455–459.

[4] P.N. Dyer, R.E. Richards, S.L. Russek, D.M. Taylor, Ion transport membrane tech-nology for oxygen separation and syngas production, Solid State Ionics 134(2000) 21–33.

[5] S. Li, W. Jin, N. Xu, J. Shi, Synthesis and oxygen permeation properties ofLa0.2Sr0.8Co0.8Fe0.2O3−ı membranes, Solid State Ionics 124 (1999) 161–170.

[6] D. Stefan, J.V. Herle, Oxygen transport through dense La0.6Sr0.4Fe0.8Co0.2O3−ı

perovskite-type permeation membranes, J. Eur. Ceram. Soc. 24 (2004)1319–1323.

[7] S.J. Xu, W.J. Thomson, Oxygen permeation rates through ion-conducting per-ovskite membranes, Chem. Eng. Sci. 54 (1999) 3839–3850.

[8] X. Tan, Y. Liu, K. Li, Preparation of La0.6Sr0.4Co0.2Fe0.8O3−˛ hollow fibre mem-

branes for oxygen production by a phase-inversion/sintering technique, Ind.Eng. Chem. Res. 44 (2005) 61–66.

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

10] S. Liu, G. Gavalas, Oxygen selective ceramic hollow fibre membranes, J. Membr.Sci. 246 (2005) 103–108.

1 brane

[

[

[

[

[

[

[perovskite hollow fibre membranes by surface acid-modification, J. Membr. Sci.

96 X. Tan et al. / Journal of Mem

11] T. Schiestel, M. Kilgus, S. Peter, K.J. Caspary, H. Wang, J. Caro, Hollow fibreperovskite membranes for oxygen separation, J. Membr. Sci. 258 (2005) 1–4.

12] Z. Wang, N. Yang, B. Meng, X. Tan, Preparation and oxygen permeation prop-erties of highly asymmetric La0.6Sr0.4Co0.2Fe0.8O3−˛ (LSCF) perovskite hollowfibre membranes, Ind. Eng. Chem. Res. 48 (1) (2009) 510–516.

13] X. Tan, Y. Liu, K. Li, Mixed conducting ceramic hollow fibre membranes for airseparation, AIChE J. 71 (2005) 1991–2000.

14] X. Tan, K. Li, Oxygen production using dense ceramic hollow fiber membranemodules with different operating modes, AIChE J. 53 (2007) 838–845.

15] X. Tan, Z. Pang, K. Li, Oxygen production using La0.6Sr0.4Co0.2Fe0.8O3−˛ (LSCF)perovskite hollow fibre membrane modules, J. Membr. Sci. 310 (2008) 550–556.

[

Science 352 (2010) 189–196

16] X. Tan, Z. Wang, H. Liu, S. Liu, Enhancement of oxygen permeation throughLa0.6Sr0.4Co0.2Fe0.8O3−ı hollow fibre membranes by surface modifications, J.Membr. Sci. 324 (1/2) (2008) 128–135.

17] Z. Wang, H. Liu, X. Tan, Y. Jin, S. Liu, Improvement of the permeation through

345 (1/2) (2009) 65–73.18] P.A. Armstrong, E.P. Foster, D.A. Horazak, H.T. Morehead, V.E. Stein, Ceramic

and coal: ITM oxygen for IGCC, in: Proceedings of the 22nd InternationalPittsburgh Coal Conference, Pittsburgh, Pennsylvania, 2005, September 11–15,2005.