Chemical Engineering Science 61 (2006) 1195–1202www.elsevier.com/locate/ces

Sorption-enhanced steam reforming of methane in a fluidized bed reactorwith dolomite as CO2-acceptor

K. Johnsena,∗, H.J. Ryub, J.R. Graceb, C.J. Limb

aInstitute for Energy Technology (IFE), P.O. Box 40, NO-2027 Kjeller, NorwaybDepartment of Chemical and Biological Engineering, University of British Columbia, 2216 Main Mall, Vancouver, Canada V6T 1Z4

Received 22 February 2005; received in revised form 15 August 2005; accepted 18 August 2005Available online 28 September 2005

Abstract

An experimental investigation was conducted in which carbon dioxide was captured in order to shift the steam reforming equilibrium forthe production of hydrogen. An atmospheric-pressure bubbling fluidized bed reactor (BFBR) of diameter 100 mm was operated cyclicallyand batchwise, alternating between reforming/carbonation conditions and higher-temperature calcination conditions to regenerate the sorbent.Equilibrium H2-concentration of > 98% on a dry basis was reached at 600 ◦C and 1.013 × 105 Pa, with dolomite as the CO2-acceptor. Thehydrogen concentration remained at 98–99 vol% (dry basis) after four reforming/calcination cycles. The total production time decreased withan increasing number of cycles due to loss of CO2-uptake capacity of the dolomite, but the reaction rate seemed unaffected. Variation of thesuperficial gas velocity within the bubbling bed regime showed that the overall reaction rate was sufficiently fast to reach equilibrium, makingbubbling bed reactors attractive for this process.� 2005 Elsevier Ltd. All rights reserved.

Keywords: Hydrogen production; Sorption-enhancement; Fluidization; Catalysis; Reaction engineering; Separations

1. Introduction

Despite efforts to decrease energy consumption and green-house gas emissions, fossil fuels will continue to play an im-portant role in the coming decades as developing nations in-crease their standard of living and major economies are slowto adapt to change. Hydrogen is often referred to an importantpotential energy carrier, but its advantages are unlikely to berealized unless efficient means can be found to produce it withreduced generation of CO2.

Steam reforming of natural gas is the predominant productionroute to hydrogen for large-scale industrial applications. Formethane the reactions are:

CH4 + H2O ↔ CO + 3H2, �H0298 = 206.2 kJ mol−1, (1)

CO + H2O ↔ CO2 + H2, �H0298 = −41.2 kJ mol−1, (2)

∗ Corresponding author: Tel.: +47 63 80 61 26; fax: +47 63 81 55 53.E-mail address: kim.johnsen@ife.no (K. Johnsen).

0009-2509/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.ces.2005.08.022

CH4 + 2H2O ↔ CO2 + 4H2, �H0298 = 165 kJ mol−1. (3)

Steam methane reforming (SMR) is normally carried outat 800–900 ◦C and 15–30 × 105 Pa, with nickel on an alu-mina support as the catalyst. A typical industrial reformercontains an array of catalyst-filled tubes suspended in a hugefurnace, supplying the heat for the highly endothermic re-forming reactions. These fixed bed reformers suffer from anumber of limitations, making them inefficient (Adris et al.,1996; Chen et al., 2003). One of the most serious constraintsrelates to conversion of methane, which is limited by thethermodynamic equilibrium of the reversible reactions. Forconventional fixed bed reformers, reaction temperature hasto be in the region of 800–900 ◦C to achieve complete con-version of methane. At this elevated temperature the catalystsuffers deactivation due to carbon formation, also resultingin blockage of reformer tubes and increased pressure drops(Trimm, 1997).

The thermodynamic equilibrium can be shifted to give morefavourable yields by removing either hydrogen or CO2. Purehydrogen can be extracted from the reactor by perm-selective

1196 K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202

Fig. 1. Hydrogen content at equilibrium as a function of temperature for apressure of 1.013 × 105 Pa, a H2O : CH4 molar ratio of 3 and a CaO : CH4molar ratio of 2.

membranes made of palladium or its alloys. A number of exper-imental and modelling studies have been carried out to provethis concept (e.g. Adris et al., 1991, 1994; Chen et al., 2003;Dogan et al., 2003).

Another way of shifting the equilibrium is by adding a CO2-acceptor to the reactor. Carbon dioxide is then converted to asolid carbonate as soon as it is formed, shifting the reversible re-forming and water-gas shift reactions beyond their conventionalthermodynamic limits. Regeneration of the sorbent releases rel-atively pure CO2 suitable for sequestration. Sorption-enhancedsteam reforming and the use of calcium based CO2 sorbentsbeen demonstrated in previous work (e.g. Balasubramanian etal., 1999; Brun-Tsekhovoi et al., 1988; Han and Harrison, 1994;Ortiz and Harrison, 2001; Silaban and Harrison, 1995; Silabanet al., 1996). For example, Balasubramanian et al. (1999) addeda calcium-based CO2 acceptor to a commercial steam reform-ing catalyst producing > 95% H2 in a laboratory-scale fixedbed reactor. Han and Harrison (1994) used CaO to capture CO2,overcoming the equilibrium limitation and achieving completeCO conversion. In addition to reactions (1)–(3) above, the non-catalytic highly exothermic carbonation reaction is included insorption-enhanced steam reforming, i.e.,

CaO(s) + CO2(g) ↔ CaCO3(s),

�H0298 = −178 kJ mol−1. (4)

The advantages of combining steam reforming with in situcapture of CO2 can be understood from the thermodynamics.Fig. 1 shows the equilibrium hydrogen concentration as a func-tion of reaction temperature at ambient pressure and at a steam-to-carbon molar ratio of 3, with the predictions based on theHSC thermodynamic software package (Outokumpu ResearchOy, Finland).

The hydrogen concentration is predicted to reach a max-imum of ∼ 98% at ∼ 600 ◦C for a CaO/CH4 ratio of 2,whereas the equilibrium concentration of conventional steamreforming is only ∼ 74% at that temperature. Fig. 1 shows that

sorption enhancement enables lower reaction temperatures,which may reduce catalyst coking and sintering, while en-abling use of less expensive reactor wall materials. In addition,heat release by the exothermic carbonation reaction suppliesmost of the heat required by the endothermic reforming reac-tions. However, energy is required to regenerate the sorbentto its oxide form by the energy-intensive calcination reaction(reverse of Eq. (4)), which represents a challenge in terms ofheat transfer and reactor construction. The high temperaturerequired for CaCO3 regeneration could also cause sinteringof solids, affecting the long-term performance of sorbent.Previous workers (Abanades and Alvarez, 2003; Silaban andHarrison, 1995; Silaban et al., 1996) have reported that theabsorption capacity for Ca-based sorbents decays as a functionof the number of calcination–carbonation cycles. Silaban et al.(1996) found that dolomite (CaCO3 · MgCO3) was superior tolimestone as a sorbent in this respect, with better multi-cycleperformance. Its advantages were attributed to differencesbetween the structural properties of calcined dolomite and cal-cined limestone. Initial calcination produces complete decom-position of dolomite, but carbonation conditions are at suchhigh temperatures that only CaO forms carbonate. The excesspore volume created by MgCO3 decomposition is believed tobe responsible for the more favourable cycling performance.

Whereas hydrogen removal by membranes can be carriedout continuously in a single reactor, continuous reforming withCO2 removal requires either that there be parallel reactors op-erated alternatively and out of phase in reforming and sorbentregeneration modes, or that sorbent be continuously transferredbetween the reformer/carbonator and regenerator/calciner. Flu-idized bed reactors are commonly used in processes where cat-alysts must be continuously regenerated, while also facilitat-ing heat transfer, temperature uniformity and higher catalysteffectiveness factors. However, experimental investigations forenhancement of steam reforming by in situ removal of CO2have mainly been conducted in small-scale fixed bed reactors.Hufton et al. (1999) used hydrotalcite as the CO2 acceptor ina fixed-bed sorption-enhanced reaction process (SERP) for hy-drogen production. The sorbent had to be periodically regen-erated by pressure swing adsorption (PSA). Balasubramanianet al. (1999) demonstrated the concept of sorption enhance-ment with CaO as CO2-acceptor, again based on a fixed bedreactor. Only Brun-Tsekhovoi et al. (1988) have utilized afluidized bed reactor for sorption-enhanced reforming. Whilethis paper is very useful, the authors did not consider cy-cling, employed excessively large sorbent particles and op-erated for only very brief periods of time. Abanades et al.(2004a) used a pilot-scale fluidized bed reactor to investigatethe carbonation of CaO capturing CO2 from high tempera-ture combustion flue gases. They found high capture efficien-cies for a typical flue gas composition of 15% CO2 in air,comparable to the concentration of CO2 from steam methanereforming.

A schematic illustration of a continuous sorption enhancedSMR process based on parallel fluidized bed reactors is shownin Fig. 2. The fluidized beds could operate in different flowregimes, e.g. in bubbling or fast fluidization. In this concept,

K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202 1197

Fig. 2. Simplified schematic of the sorption-enhanced SMR process.

catalyst and sorbent are mixed in the reformer, where sorption-enhanced steam reforming is performed. The product gasfrom the reformer mainly consists of hydrogen and steam,with minor quantities of CO, CO2 and unconverted methane.Carbonated sorbent is transferred to the regenerator where heatis supplied for the endothermic calcination reaction, either byburning fuel in the regenerator or by indirect heating from anexternal heat source. Heating of a dense bubbling bed couldalso be accomplished by the insertion of heat transfer tubesinto the bed. Heat transfer would then occur between the flu-idized bed and the submerged tube surfaces. These tubes willat the same time act as baffles reducing the gas bubble size,enhancing interphase mass transfer. A portion of the hydrogenproduced in the reformer might be burned externally, supplyingheat to the tubes and eliminating the need for additional fossilfuel. Indirect heating has the advantage of producing pure CO2ready for sequestration. Direct burning of fuel inside the regen-erator would be a more efficient way of supplying the heat, butwould require downstream separation of CO2, unless hydro-gen is burned with pure oxygen. To avoid separation processesdownstream, CO2 and/or H2O can be used as the fluidizing gasin the regenerator. Calcium-based sorbents have the advantageof being available at low cost, but cannot maintain the cap-ture capacity upon multiple reforming/regeneration cycles. Amake-up stream of fresh sorbent must be included to maintaincapture capacity. This addition could be to the calciner, withwithdrawal in the reformer, as indicated in Fig. 2. Syntheticsorbents, such as Li2ZrO3, have better multi-cycle stability,but their cost would require them to sustain > 10, 000 cyclesto compete with natural sorbents (Abanades et al., 2004b).Coupling of two bubbling beds would have the advantage oflow rates of attrition due to low gas and particle velocities,and the relatively slow carbonation reaction rate will be fa-cilitated in this flow regime. There is a lack of experimentaldata on sorption-enhanced SMR in fluidized beds in the openliterature.

In our work, described below, a bubbling fluidized bed wasused for sequential sorption-enhanced steam reforming and re-generation of sorbent without separating the catalyst from the

sorbent between cycles. Due to the lack of experimental dataon the performance of the overall process in such reactor con-figuration, special attention was given to the multi-cycle perfor-mance of both the sorbent and the catalyst under these circum-stances. The rate of the combined reactions was also evaluatedusing different superficial gas velocities within the bubblingregime.

2. Experimental

2.1. Bubbling fluidized bed reformer (BFBR)

A schematic of the reactor system, used previously(Constantineau, 2004) in studies of fluidized bed roasting ofzinc concentrates, is shown in Fig. 3. The major componentsconsist of a pre-heater, a 0.66 m high by 0.1 m ID stainlesssteel fluidized bed reactor with expanded freeboard, a filtrationunit and a gas cooler unit. Methane was fed to the upper partof the pre-heater where it was mixed with steam. A removablestainless steel gas distributor plate, with 34 drilled 1.2 mmdiameter holes on a hexagonal grid, was placed between thepre-heater and the reactor. The preheated reactant gas passedthrough a mixture of commercial Ni-based steam-reformingcatalyst (Haldor Topsoe A/S, R-67R-7H) and calcined dolomite(Franzefoss A/S, Arctic Dolomite SHB). Dolomite was used inpreference to limestone because of initial tests indicating bet-ter ability to sustain performance in cyclical operation. Threedifferent zones of the reactor were each heated by electricalfurnaces, which could be controlled individually. Tempera-tures and pressure drops were recorded by a data acquisitionsystem. Teflon bags were used for gas sampling from a portat the outlet of the freeboard zone. The composition of thesesamples was determined using a gas chromatograph (ShimadzuGC-8A).

2.2. Sample preparation

The composition of the dolomite is provided in Table 1. Thisdolomite was chosen because it did not contain sulphur, whichis poisonous to the reforming catalyst.

Prior to the experiments, dolomite and catalyst were sievedto ensure particle sizes between 125–300 and 150–250 �m,respectively. The dolomite had to be first calcined to obtainthe desired oxide form. This was accomplished at 850 ◦Cin N2 without the catalyst present. A portable gas analyzer(Horiba PG-250) was used to determine complete calcination,corresponding to the disappearance of CO2 in the productgas. The reactor was then cooled, and part of the calcineddolomite removed and stored in a dessicator for later use,while the rest was mixed with catalyst and re-injected into thereactor.

2.3. Experimental procedure

The reactor had no feeding lines for solids, and was there-fore operated batchwise, with periodic calcination at higher

1198 K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202

Fig. 3. Schematic of reformer unit.

Table 1Analysis of arctic dolomite SHB, data from Franzefoss AS

Species CaO MgO SiO2 Al2O3 Fe2O3 Na2O TiO2 K2O Loss by ignition

Conc. [wt%] 32 20.3 0.7 0.1 0.1 0.003 0.005 0.004 46.3

temperatures to regenerate the dolomite. The experimental in-vestigation can be divided into two parts: multi-cycle testsand tests where the superficial gas velocity was varied. Freshdolomite and catalyst were used for the investigation of the ef-fect of gas velocity in order to make the results for differentgas velocities comparable. The total initial bed mass was 3.1 kgfor all runs, with a catalyst-to-calcined dolomite mass ratio of2.5. During the calcination stages of the multi-cycle tests, pureN2 was fed to the reactor. No effort was made to separate thecatalyst from the dolomite between cycles. To ensure that thecatalyst was active, reduction of the catalyst was performed ina H2/N2 mixture at 650 ◦C for 1–2 h prior to each reforming

period. This is equivalent to a continuous process where a por-tion of oxidized catalyst is returned to the reducing atmospherein the reformer, where it is reduced back to its active form. Thereforming reaction was always carried out at 600 ◦C and am-bient pressure. The experimental conditions for the BFBR unitare summarized in Table 2.

3. Results and discussion

The reaction conditions corresponded to operation in the bub-bling bed flow regime. With 0.9 kg calcined dolomite present,the time required for complete carbonation of CaO in dolomite

K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202 1199

Table 2BFBR experimental conditions

Parameters Values

Total mass of particles in bed (kg) 3.1Catalyst-to-calcined dolomite mass ratio 2.5(dimensionless)Bulk density of mixture (kg/m3) 1300Static bed height (m) 0.3Catalyst particle size range (�m) 150–250Dolomite particle size range (�m) 125–300Reforming temperature (◦C) 600Superficial velocity (m/s) 0.032a, 0.064, 0.096Steam-to-carbon molar feed ratio 3Calcination temperature (◦C) 850Calcination atmosphere N2

aGas velocity for multi-cycle test.

Fig. 4. Outlet composition (dry basis) as a function of time.

was calculated to be 170 min for a superficial gas velocity of0.032 m/s at 600 ◦C, based on the assumption that all carbonfed reacted with CaO to yield CaCO3. The total time of opera-tion was 5 h for each run. A typical response curve is shown inFig. 4, with the dry gas composition plotted as a function oftime.

The hydrogen concentration is stable at 98–99 vol% on a drybasis for a period of 150–180 min, often referred to as the pre-breakthrough period by previous authors (e.g. Han and Har-rison, 1994), before there is sudden drop in concentration to∼72–74%. The opposite trend is observed for the CO2 concen-tration, where there is a sudden increase from 0.3 to 13–14%after the same interval. This characteristic breakthrough occurswhen the amount of CO2 produced by steam reforming ex-ceeds the sorption capacity of the CaO. Fig. 4 clearly showsthe marked enhancement of hydrogen production achieved byin situ capture of CO2. The shape of the curve is typical forsorption-enhanced steam reforming. Due to the limited num-ber of sample points, the time of the breakthrough cannot begiven precisely, but it seems to correspond to a time between150 and 180 min, in good agreement with the calculated timefor complete carbonation (assuming 100% calcium utilization)

Fig. 5. Time variation of temperature in bed zones.

of 170 min. After complete carbonation, the hydrogen con-centration dropped to a value corresponding to equilibrium ofsteam methane reforming of ∼ 73 vol% on a dry basis. At thispoint, no CaO was left to react with CO2, so that the reac-tion enhancement was lost. This period is often referred to asthe post-breakthrough period, and re-calcination has to be per-formed to reactivate the sorbent. Combining the strongly en-dothermic steam reforming with the exothermic carbonationreaction makes the overall reforming reaction almost thermallyneutral. Carbonation of CaO (Eq. (4)) is a reversible reaction,and temperature control is very important to prevent the unde-sired reverse calcination reaction in the reformer. Temperatureuniformity promoted by rapid mixing of the solids makes flu-idized beds well suited for processes where temperature unifor-mity is important. Two thermocouples were placed in the densebed zone, one (T1) just above the distributor and the other (T2)0.19 m above. Typical temperature traces for one run are shownin Fig. 5.

The difference in temperature between the two positions, T1and T2, was nearly constant at 3–4 ◦C during the entire courseof reaction, confirming the excellent temperature uniformityof the bubbling fluidized bed. Another feature observed fromFig. 5 is the temperature drop after 150 min. This corre-sponds to the start of the breakthrough period also observed inFig. 4, caused by the diminishing exothermic carbonationreaction, while the endothermic reforming reaction continues.

3.1. Multi-cycling

In order to make the process continuous, the sorbent must beregenerated after completion of the carbonation stage. Severalprevious groups (e.g. Li et al., 2005; Ortiz and Harrison, 2001;Silaban and Harrison, 1995) have investigated the multi-cycleperformance of CaO-based sorbents. Abanades and Alvarez(2003) included previously published multi-cycle results whenthey reported an unavoidable decay in carbonation conversionthat was dependent on the number of cycles. Most of theseinvestigations have been carried out using thermo-gravimetric

1200 K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202

Fig. 6. H2 and CO2 concentrations (dry basis) as functions of number ofcycles.

analysis (TGA), either in pure CO2 or in simulated reformingenvironments. There is little information on how sorption en-hancement is affected by carbonation–calcination cycling in afluidized bed. Our BFBR was operated batchwise, with peri-odic regeneration of sorbent without physically separating thecatalyst from the dolomite. The solid mixture was exposed tohydrogen after each period of calcination to ensure that thenickel in the catalyst was in the reduced active form, corre-sponding to a continuous process where the solids would beexposed to a reducing atmosphere in the reformer.

Fig. 6 shows the concentrations of hydrogen and carbondioxide for different numbers of carbonation–calcination cy-cles. The first cycle is not included in this figure because thedolomite had been exposed to air, reducing its absorption ca-pacity, hence making comparison with the other cycles difficult.It is clear from Fig. 6 that the duration of the pre-breakthroughperiod is reduced somewhat with an increase in the number ofcycles. This reduction is due to loss of CaO capacity, but the hy-drogen concentration remained constant at 98–99%, suggestingthat the equilibrium concentration of the combined reactions(1)–(4) was reached for each cycle. The hydrogen concentra-tion during the post-breakthrough period was again at equilib-rium for successive cycles, indicating that the catalytic activityremained sufficiently high to reach equilibrium upon cycling.The breakthrough period was characterized by onset of the slowcarbonation reaction rate regime, where diffusion through thesolid product layer limited the rate of reaction. The slopes ofthe breakthrough curves indicate that the global reaction ratewas not significantly affected by the number of cycles. How-ever, the limited number of cycles and the low gas velocity inthis study make it hard to conclude that the rate of reactionwould in general not be affected by multi-cycling.

Silaban and Harrison (1995) reported that the loss of capacitywas associated with a change in structural properties of thesorbent. It was claimed that reduced porosity left the interior ofthe sorbent particles inaccessible to CO2. As a consequence, theCO2 uptake capacity decreased with cycling, consistent withFig. 6.

Brun-Tsekhovoi et al. (1988) employed relatively largedolomite particles (1.3 mm average), to facilitate their physicalseparation from the catalyst (250 �m) before regeneration. Aseparation stage should be avoided if possible as it would addextra components and complexity to the system, lead to ad-ditional attrition and cause extra heat losses. Khotomlyanskiiet al. (1970) studied the separation of catalyst from a heavierheat transfer agent in a fluidized bed using particles differingsignificantly in density and size. The catalyst density in thecurrent study was 2200 kg/m3, whereas the sorbent densitydepended on the degree of carbonation, with a possible rangefrom 1560 kg/m3 (fully calcined) to 2230 kg/m3 (completelycarbonated), the latter density being similar to that of the cata-lyst. Note that separation is more sensitive to different particledensities than to differences in particle size (Rowe and Nienow,1976). Since steam reforming catalysts commonly encountertemperatures similar to those employed in calcination, it shouldbe possible to expose the catalyst to the calcination conditionswithout separation. The present investigation shows that cat-alytic activity remained after 4 cycles, without any separationof the particles. Nitrogen would not be used as the fluidizinggas for calcination in an industrial application, due to dilutionof CO2 leaving the regenerator for sequestration. In that case,CO2 itself, or possibly steam, would be more realistic as themedium to avoid separation processes downstream. Silabanand Harrison (1995) reported that a CO2 atmosphere had anadverse effect on CaO sorption capacity during multi-cycling.It is therefore possible that the reduction in production time asa function of cycles, observed from Fig. 6, would have beengreater if the carbonated sorbent had been calcined in atmo-spheres other than 100% N2. However, Ortiz and Harrison(2001) report no significant difference in loss of multi-cycledurability for different regeneration atmospheres, except whenregeneration was carried out in pure nitrogen at 950 ◦C, usingdolomite as sorbent. It was beyond the scope of this work toinvestigate the effect of different calcination atmospheres andonly nitrogen was used.

3.2. Increasing superficial gas velocity

All multi-cycle runs were conducted at a superficial gasvelocity of 0.032 m/s, a very low velocity compared to normalcommercial fluidization processes. From an industrial point ofview, a higher gas throughput would be advantageous. Keep-ing the total mass of solid constant, the superficial gas velocitywas increased to see its effect. The highest gas velocity investi-gated was 0.096 m/s, again lower, though less so, than expectedcommercial velocities. However, it was desirable to investigatewhether bubbles bypassing would lead to significantly lowerconversions than reported for fixed bed reactors. Fig. 7 showsthe hydrogen concentration at three different superficial gas ve-locities.

As the gas velocity increased, the total production time de-creased, as expected. The hydrogen concentration exceeded95% on a dry basis for all three velocities. Both fresh dolomiteand fresh catalyst were used for each run, making the results

K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202 1201

Fig. 7. Hydrogen concentration (dry basis) in first cycle as a function ofsuperficial gas velocity.

comparable. It is evident from Fig. 7 that both mass transfer andreaction kinetics are fast enough to reach equilibrium withinthe given range of operation conditions. If there was any lossin catalyst activity, it was too small to reduce the conversioneither before or after the capacity of the sorbent was exhausted.Higher velocities were not tested, but high CO2 capture effi-ciencies by CaO have been reported (Abanades et al., 2004a,b)for flue gases at superficial velocities of 1 m/s, indicating thatfurther increases in gas velocity are likely to be feasible. Theexperiment at U = 0.064 m/s gave a somewhat lower maxi-mum H2 concentration than the two other runs, possibly due tothe gas analysis method. There were always nitrogen peaks inthe chromatograms, originating from the use of sampling bagsand a syringe for manual injection into the GC, making thepresence of air inevitable. The proportion of N2 in the sampleswas usually 2–4 vol%, but for U = 0.064 m/s a higher nitro-gen concentration was observed. Hydrogen is the most volatilegas, and any leakage from a syringe or GC injection port wouldintroduce air at the expense of hydrogen.

4. Conclusions

Equilibrium for sorption-enhanced steam reforming at600 ◦C and 1.013 × 105 Pa was reached for gas velocities inthe range of 0.032–0.096 m/s. Multiple reforming-regenerationcycles showed that the hydrogen concentration remained at98–99 vol% after 4 cycles. The total production time wasreduced with an increasing number of cycles due to loss ofCO2-uptake capacity of the dolomite, but the reaction rateseemed to be unaffected for the conditions investigated. Al-though the experimental results presented here were obtainedfrom a sequence of batch reforming-batch regeneration, it isbelieved that comparison can be made to coupling of twobubbling fluidized beds. The very uniform temperature withinthe bed, with maximum axial differences in temperature ofonly 3–4◦C, confirms the good temperature control offered bybubbling fluidized beds. At the given reaction conditions the

overall reaction rate was sufficiently fast that equilibrium wasapproached, making bubbling bed reactors an appealing choicefor this process.

Acknowledgements

This work was financially supported by the Research Councilof Norway (RCN) and the Natural Sciences and EngineeringResearch Council of Canada (NSERC). K.J. would also liketo acknowledge Det Norske Veritas (DNV) in association withThe Norwegian University of Science and Technology (NTNU)for the scholarship which made his stay at the University ofBritish Columbia possible. The authors also thank Dr. PierreConstantineau for assistance in modifying the reactor.

References

Abanades, J.C., Alvarez, D., 2003. Conversion limits in the reaction of CO2with lime. Energy & Fuels 17, 308–315.

Abanades, J.C., Anthony, E.J., Lu, D.Y., Salvador, C., Alvarez, D., 2004a.Capture of CO2 from combustion gases in a fluidized bed of CaO. A.I.Ch.E.Journal 50, 1614–1622.

Abanades, J.C., Rubin, E.S., Anthony, E.J., 2004b. Sorbent cost andperformance in CO2 capture systems. Industrial and Engineering ChemistryResearch 43, 3462–3466.

Adris, A.M., Elnashaie, S.S.E.H., Hughes, R., 1991. A fluidized bedmembrane reactor for the steam reforming of methane. Canaidan Journalof Chemical Engineering 69, 1061–1070.

Adris, A.M., Lim, C.J., Grace, J.R., 1994. The fluidized bed membrane reactorsystem: a pilot scale experimental study. Chemical Engineering Science49, 5833–5843.

Adris, A.M., Pruden, B.B., Lim, C.J., Grace, J.R., 1996. On the reportedattempts to radically improve the performance of the steam reformingreactor. Canadian Journal of Chemical Engineering 74, 177–186.

Balasubramanian, B., Ortiz, A.L., Kaytakoglu, S., Harrison, D.P., 1999.Hydrogen from methane in a single-step process. Chemical EngineeringScience 54, 3543–3552.

Brun-Tsekhovoi, A.R., Zadorin, A.N., Katsobashvili, Ya.R., Kourdyumov,S.S., 1988. The process of catalytic steam reforming of hydrocarbons inthe presence of carbon dioxide acceptor, hydrogen energy progress, VII.Proceedings of the World Hydrogen Energy Conference. Pergamon Press,New York, pp. 885–900.

Chen, Z., Yan, Y., Elnashaie, S.S.E.H., 2003. Novel circulating fast fluidized-bed membrane reformer for efficient production of hydrogen from steamreforming of methane. Chemical Engineering Science 58, 4335–4349.

Constantineau, J.P., 2004. Fluidized bed roasting of zinc sulphide concentrate:factors affecting the particle size distribution. Ph.D. Thesis, University ofBritish Columbia, Vancouver, Canada.

Dogan, M., Posarac, D., Grace, J.R., Adris, A.M., Lim, C.J., 2003. Modelingof autothermal steam methane reforming in a fluidized bed membranereactor. International Journal of Chemical Reactor Engineering 1 A2,1–14.

Han, C., Harrison, D.P., 1994. Simultaneous shift reaction and carbon dioxideseparation for the direct production of hydrogen. Chemical EngineeringScience 49, 5875–5883.

Hufton, J.R., Mayorga, S., Sircar, S., 1999. Sorption-enhanced reaction processfor hydrogen production. A.I.Ch.E. Journal 45, 248–256.

Khotomlyanskii, L.N., Brun-Tsekhovoi, A.R., Katsobashvili, Y.R., Petrov,V.N., Skoblo, A.I., 1970. Investigation of the separation of dissimilarparticles in fluidized bed. Chemistry and Technology of Fuels and Oils 6,15–19.

Li, Y., Buchi, S., Grace, J.R., Lim, C.J., SO2 removal and CO2 capture bylimestone resulting from calcination/sulphation/carbonation cycles. 2005,submitted for publication.

1202 K. Johnsen et al. / Chemical Engineering Science 61 (2006) 1195–1202

Ortiz, A.L., Harrison, D.P., 2001. Hydrogen production using sorption-enhanced reaction. Industrial and Engineering Chemistry Research 40,5102–5109.

Rowe, P.N., Nienow, A.W., 1976. Particle mixing and segregation in gasfluidized beds: a review. Powder Technology 15, 141–147.

Silaban, A., Harrison, D.P., 1995. High temperature capture of carbon dioxide:characteristics of the reversible reaction between CaO(s) and CO2 (g).Chemical Engineering Communication 137, 177–190.

Silaban, A., Narcida, M., Harrision, D.P., 1996. Characteristics of thereversible reaction between CO2 (g) and calcined dolomite. ChemicalEngineering Communications 146, 149–162.

Trimm, D.L., 1997. Coke formation and minimization during steam reformingreactions. Catalysis Today 37 (3), 233–238.