Production of Very Pure Hydrogen with Simultaneous Capture of

Sep 25, 2008 - Mohammad Ismail , Wen Liu , Martin S. C. Chan , Matthew T. Dunstan ... Christopher L. Muhich , Victoria J. Aston , Ryan M. Trottier , A...
0 downloads 0 Views 676KB Size
Download PDF
Ind. Eng. Chem. Res. 2008, 47, 7623–7630

7623

Production of Very Pure Hydrogen with Simultaneous Capture of Carbon Dioxide using the Redox Reactions of Iron Oxides in Packed Beds Christopher D. Bohn,† Christoph R. Mu¨ller,† Jason P. Cleeton,‡ Allan N. Hayhurst,† John F. Davidson,† Stuart A. Scott,*,‡ and John S. Dennis† Department of Chemical Engineering, UniVersity of Cambridge, Pembroke Street, Cambridge, CB2 3RA, U.K., and Department of Engineering, UniVersity of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, U.K.

A chemical looping process, which uses a packed bed of the various oxides of iron, has been formulated to produce separate, pure streams of H2 and CO2 from syngas. The process has the following stages: (1) Reduction of Fe2O3 to Fe0.947O in the syngas from gasifying coal or biomass. This stage generates pure CO2, once the water has been condensed. (2) Subsequent oxidation of Fe0.947O to Fe3O4 using steam, to simultaneously produce H2. (3) Further oxidation of Fe3O4 to Fe2O3 using air to return the oxide to step 1. Step 1 was studied here using a mixture of CO + CO2 + N2 as the feed to a packed bed of iron oxide particles, while measuring the concentrations of CO and CO2 in the off-gas; step 2 was investigated by passing steam in N2 through the packed bed and measuring the quantity of H2 produced. The third step simply involved passing air through the bed. Reduction to Fe, rather than Fe0.947O, in step 1 gave low levels of H2 in step 2 after 10 cycles of reduction and oxidation and led to the deposition of carbon at lower temperature. Step 3, i.e. reoxidizing the particles in air to Fe2O3, led to no deterioration of the hydrogen yield in step 2 and benefited the process by (i) increasing the heat produced in each redox cycle and (ii) preventing the slip of CO from the bed in step 1. The proposed process is exothermic overall and very usefully generates separate streams of very pure H2 and CO2 without complicated separation units. 1. Introduction The need to stabilize or even reduce the production of anthropogenic carbon dioxide and the recent progress made in the development of fuel cells make hydrogen a promising option as a fuel in the near future.1 However, for hydrogen to become a major energy source, it must be produced in an efficient, sustainable manner, and when used in fuel cells, it must be highly pure. Hydrogen is currently used in industry for various processes including the synthesis of ammonia, the production of methanol, the conversion of a hydrocarbon gas to a liquid fuel (FischerTropsch), hydrocracking, and metallurgy. The reforming of methane with steam is now the predominant way of producing hydrogen2 and generally relies3 on a nickel catalyst at 1073 K. If hydrogen is to be used in a fuel cell with a polymeric electrolyte membrane (PEM), the level of contamination from CO must be below ∼50 ppm, so as not to poison the platinum clusters on the anode. The effluent gas from a steam reformer, however, has a concentration of CO typically around ∼10 vol %, depending on feedstock and temperature. Subsequent reaction of this CO is achieved via the catalyzed water-gas shift (WGS) reaction at both high, 570-770 K, and low, 450-570 K, temperatures. At higher temperatures, a Fe/Cr catalyst is used; whereas at lower temperatures, a Cu/Zn/Al catalyst is employed.4 The high and low temperature WGS reactors in series return exit CO levels of ∼3 and 0.5 vol %, respectively.5,6 The amount of CO in the H2 exiting the low temperature shift reactor can be further lowered by (i) separating the H2 from the CO using in situ Pd membranes7,8 or (ii) adding, via preferential oxidation (PROX),9 a cleanup step downstream of the water-gas shift reactors. Both of these options require further input of * To whom correspondence should be addressed. Tel.: +44 (0)1223 332645. Fax: +44 (0)1223 332662. E-mail: [email protected]. † Department of Chemical Engineering. ‡ Department of Engineering.

energy, since i requires a vacuum on the permeate side and ii requires pure oxygen for the extra PROX process step. The transport and distribution of hydrogen present serious challenges. At present, hydrogen is typically produced in refinery complexes with capacities of ∼15 kte/y of H2 or greater. Local production of very pure H2 on smaller scales is currently not economical using traditional plants based on the steam reforming of methane. Therefore, a process which can be implemented at a smaller, distributed scale would be highly desirable. The reduction and oxidation reactions associated with iron oxides are capable of producing H2 with simultaneous capture of CO2 on a small scale and will be explored in this paper. The feasibility of storing CO2 on a distributed scale will not be discussed. The iron oxides are the following: haematite, Fe2O3, magnetite, Fe3O4, and wuestite, Fe0.947O. Here, wuestite is written as Fe0.947O; generally, wuestite is written10 as Fe(1-y)O, where 0.05 < y < 0.17 such that the stoichiometric ratio of Fe/O ) 1 is never achieved. The oxygen excess can be construed as iron vacancies in the lattice structure, where the appropriate proportion of iron ions must then be trivalent to maintain electrical neutrality. Iron has long been known to produce H2 when reacted with steam. In 1910, Messerschmidt11 patented the steam-iron process for producing hydrogen using the Fe3O4 to Fe0.947O transition. Soon after, Lane12 added Fe2O3 to the cycle by including an oxidation step with air to eliminate the residual sulfur and carbon impurities deposited during each cycle. Later, Reed13 developed a process with interconnected fluidized beds for circulating iron oxides; this enabled continuous production of H2. Iron oxide has also been investigated for use in the combustion of gaseous fuels using chemical looping, a process suitable for capturing CO2. Here, the fuel is reacted with the metal oxide in a fluidized bed to produce CO2 and H2O as products.14,15 Subsequent condensation of the H2O leaves CO2 sufficiently

10.1021/ie800335j CCC: $40.75  2008 American Chemical Society Published on Web 09/25/2008

7624 Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008

pure for sequestration, e.g. in depleted oil wells or aquifers. The reduced oxide particles are then conveyed to a separate reactor where they are regenerated in air before looping back to the reduction stage. By introducing the metal oxide as an effective oxygen-carrier, the large nitrogen ballast typically found in combustion gases is eliminated, giving a relatively pure stream of CO2. This concept has now been extended to solid fuels, using in situ gasification.16,17 Combining the features of hydrogen production and carbon capture with iron oxides, a cyclic redox process can be conceived for producing H2 with simultaneous capture of CO2 using the following three basic steps: (1) Production of CO2 with Oxide Reduction. Syngas, e.g. from the gasification of coal or biomass, is converted to a stream of CO2 and H2O by passing it through a packed bed of iron oxide particles. Condensing out the H2O from the off-gas leaves CO2 suitable for sequestration. The concentration gradient along the reactor enables Fe0.947O or even Fe to exist at the syngas inlet, while still maintaining Fe2O3 in the packed bed toward the gas outlet. From the thermodynamics discussed later in section 2, high levels of CO2 and H2O will be produced, provided some Fe2O3 exists at the outlet and the residence time in this area is sufficiently long to permit near equilibrium conversion of the reducing gas. (2) Production of Hydrogen with Partial Reoxidation of Metal Oxide. After the partial reduction of the iron oxide in the bed, passing steam through the reactor generates hydrogen and reoxidizes the reduced iron oxide to Fe3O4. Complete reoxidation to Fe2O3 using steam is impracticable, since at 1023 K, H2 levels as low as 20 ppm impede the transition from Fe3O4 to Fe2O3. As a result, a third step, oxidation with air, is required. (3) Complete Reoxidation of Metal Oxide with Air. The reoxidation to Fe2O3 in air produces a stream of N2 with residual O2. Once the entire packed bed has been regenerated to Fe2O3, the cycle can begin anew at step 1. It will be shown that the redox reactions in steps 1-3 above can be carried out from 873-1173 K, but special precautions must be taken with step 1 at the lower temperatures, since solid carbon may be deposited. While the above description is specific to a packed bed reactor, other schemes capable of producing separate streams of H2 and CO2 (e.g., a counter-current moving bed18 or a sequence of fluidized beds) can be used for the redox reactions. A key requirement for the above process is the development of a suitable iron oxide particle. It should have the following properties: (i) high reactivity at the process temperatures and (ii) a consistently high conversion over many cycles of oxidation and reduction. Much work has been done developing iron particles for producing hydrogen via the steam-iron process. Kindermann et al.19 reported that SiO2 and CaO helped to maintain the porosity when an iron oxide pellet was reduced with CO. The influence of metallic additives, such as Rh, Mo, Zr, Al, and Ga, has been studied by Otsuka et al.20 They characterized additives with respect to two criteria: their ability to prevent iron oxide from decaying in activity over repeated redox cycles and the ability to enhance the rate of reoxidation. Mo, Al, and Ce showed promise for the former function, whereas Rh, Ir, and Ag facilitated the latter. The combined effects of the two metals were studied by preparing particles based on iron oxide with 5 mol % each of Rh and Mo.21 These metals acted cooperatively to enable iron and its oxides to be repeatedly interconverted with significant rates of formation of

Figure 1. Phase diagram showing Kp versus temperature for the iron system for reactions with CO2/CO (- - -) and H2O/H2 (s). The triple point, where Fe3O4, Fe0.947O, and Fe coexist, is at 848 K.

H2 at lower temperatures. Finally, addition of Ni and Cu has been shown22 to enhance the rate of reduction of iron oxide in H2 and CO. Despite the abundance of literature on the redox reactions of iron oxides, it is still unclear which temperature and which transitions, viz. Fe2O3-Fe, Fe2O3-Fe0.947O, Fe3O4-Fe, or Fe3O4-Fe0.947O, are preferable for a cyclic scheme to produce hydrogen. In this paper, all four possible transitions are examined at 873, 1023, and 1173 K. The experimental results are then discussed in the light of industrial application, e.g. maximizing the yield of H2, maximizing the heat output, and maintaining the purity of the H2. 2. Thermodynamic Analysis 2.1. Reduction of Iron Oxides. The reduction of haematite, Fe2O3, by syngas (CO + H2) occurs in a stagewise process given for the case of CO by 3Fe2O3 + CO a 2Fe3O4 + CO2 ; ∆H°298K ) -43.2 kJ/mol (R1) 0.947Fe3O4 + 0.788CO a 3Fe0.947O + 0.788CO2 ; ∆H°298K ) + 37.3 kJ/mol (R2) Fe0.947O + CO a 0.947Fe + CO2 ; ∆H°298K ) -16.7 kJ/mol (R3) where the standard changes of enthalpy listed are for the reactions as written. Reduction to Fe occurs only if there is a sufficiently high partial pressure of the reducing gas. Figure 1 shows a phase diagram for the Fe-CO-CO2 and Fe-H2-H2O systems based on published thermodynamic data.23 The dashed lines in Figure 1 are values of the equilibrium constant Kp ) exp(-∆G°/RT), which for reactions R1-R3 are equal to the ratio of partial pressures, pCO2/pCO at equilibrium. The solid lines in Figure 1 show equivalent plots for the H2-H2O system, e.g. R4 and R5 given below. The equilibrium phase diagrams for both systems, Fe-CO-CO2 and Fe-H2-H2O, are in fact related by the equilibrium constant for the water-gas shift, KW ) pH2/pH2O × pCO2/pCO. Figure 1 shows that below the triple point at 848 K (575 °C), magnetite, Fe3O4, is reduced directly to iron, Fe, omitting wuestite, Fe0.947O, as the intermediate. 2.2. Oxidation of Iron. From Figure 1 (right-hand ordinate), it can be seen that the oxidation of Fe or Fe0.947O with steam to generate hydrogen is feasible, if the value of pH2O/pH2 exceeds the value of Kp (solid line marking the boundaries between Fe/ Fe0.947O and Fe0.947O/Fe3O4) for the respective reactions:

Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008 7625

0.947Fe + H2O(g) a Fe0.947O + H2 ; ∆H°298K ) -23.8 kJ/mol (R4) 3Fe0.947O + 0.788H2O(g) a 0.947Fe3O4 + 0.788H2 ; ∆H°298K ) -69.2 kJ/mol (R5) The purity of the hydrogen produced depends on (i) the purity of the steam used and (ii) whether carbon is deposited by the Boudouard reaction: 2CO(g) a C(s) + CO2(g) ; ∆H°298K ) -172.4 kJ/mol (R6) 2

This can occur for pCO2/pCO less than a particular value which depends on temperature. In practice, for one atmosphere of only CO and CO2, this condition becomes pCO2/pCO < 1 at temperatures below 973 K. Solid carbon causes contamination of the H2 because it is gasified by steam to form CO and subsequently CO2. It is possible, through the choice of an appropriate temperature and composition of reducing gas, to conduct reactions R1-R3 in a regime where the Boudouard reaction R6 is thermodynamically infeasible. As mentioned previously, owing to thermodynamic constraints, the oxidation of Fe3O4 to Fe2O3 with steam is nearly impossible. For example, at 1023 K, Figure 1 shows that H2 levels as low as 20 ppm (pH2O/pH2 < 50 000) prevent the transition from Fe3O4 to Fe2O3; therefore, air has to be used to reoxidise magnetite to haematite in 2Fe3O4 + 1/2O2 a 3Fe2O3 ; ∆H°298K ) -239.7 kJ/mol (R7) 2.3. Overall Process. As an illustration, suppose that the effluent gas from gasifying a solid, carbonaceous fuel in steam or CO2 is used to reduce Fe2O3 to Fe0.947O and that the resulting Fe0.947O is then reoxidised in steam to give hydrogen and Fe3O4, the latter being oxidized in air to Fe2O3. The overall reaction is then given by C(s) + 1.25H2O(g) + 0.375O2 a CO2 + 1.25H2 ; ∆H°298K ) -90.6 kJ/mol (R8) This equation demonstrates the potential of the proposed process and interestingly holds whether CO2 or H2O is used to gasify the carbon. Thus, solid carbon from either biomass or coal can be used to produce separate streams of CO2 for sequestration and high purity H2 for energy consumers, as well as heat. Because the process requires neither a heat input nor complicated separation units, it is suitable for operation on a distributed scale. However, the coupling of enthalpies between the process steps does represent a challenge. The heat released from the overall process can be increased by adding less steam to reaction R8 and more air to give additional oxygen. However, such an adjustment will decrease the amount of H2 produced. 3. Experimental Details The particles of iron oxide for this study were prepared from Fe2O3 powder (99 wt %. Water, which had been purified by reverse osmosis, was sprayed on to the dust to form agglomerates. The resulting mixture was mechanically stirred both during and after spraying, and then, the granules were sieved to the desired size fraction of 425-600 µm. These particles were then sintered in an oven at 1173 K for 3 h, subsequently cooled in air and then resieved. The packed bed reactor was made from 316 stainless steel tubing (i.d. 10.2 mm; total length 430 mm) as shown in Figure

Figure 2. Schematic diagram of the reactor containing a packed bed supported on a perforated plate. Resting on the plate was a 4 g plug of sand, supporting the active bed, with another 4 g plug of sand on top. The active bed was either (i) 20 g Fe2O3 particles, see section 4.1, or (ii) 1 g of Fe2O3 particles mixed with 4 g sand, see sections 4.2-4.6. The thermocouple extends to the center of the active bed; all measurements are in millimeters.

2. A stainless steel plate (1 mm thick) supported the bed: the plate contained 8 evenly spaced holes in a circular pitch, each with a diameter of 1 mm. The entire assembly was placed in a tubular furnace and the temperature of the bed was controlled by a K-type thermocouple which had been inserted into the bed. The reduction of the iron oxide particles was performed using mixtures of (i) 10 vol % CO with balance N2 and (ii) pure CO2. The flowrates of the gases, except CO2, were metered into the reactor using mass flow meters (Honeywell AWM5101N) followed by flow control valves. The flowrate of CO2 was measured with a rotameter. In all cases, the total gas flow was maintained at ∼0.12 m3/h, as measured at 101.3 kPa and 298 K; see Table 1. The reducing gas mixture was chosen, based on Figure 1, to give the desired iron species at the specified temperature. Between each reduction and oxidation stage, the reactor was briefly purged with nitrogen as shown in Table 1. In order to reoxidize the iron with steam, a syringe pump was used to feed liquid H2O through a hypodermic tube (1.6 mm o.d.) into an electrically heated chamber, which was packed with stainless steel mesh to increase the surface area for heat transfer. The water was vaporised and conveyed into the reactor by means of a N2 sweep gas. Reoxidation to Fe2O3 was achieved using air, also metered into the reactor using a mass flow meter and a control valve. Solenoid valves were used to switch gas streams. A summary of cycle times and gases is presented in Table 1 for experiments with an initial loading of 1 g of Fe2O3. The gas leaving the reactor was cooled to condense most of the water, before being passed through a tube filled with CaCl2 to remove any residual water. The composition of the dried gas was determined by (1) a nondispersive infrared (NDIR) analyzer measuring [CO2],[CO] (ABB Easyline, 0-20 vol %); (2) a more

7626 Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008 Table 1. Redox Cycling: Gas Mixtures (vol %) and Corresponding Flow Durations for Experiments with an Initial Charge of FeO of 1 ga purge time (min) flow (m3/h) flow (m3/h) Fe2O3-Fe Fe2O3-Fe0.947O Fe3O4-Fe Fe3O4-Fe0.947O a

reduction

purge

steam oxidation

purge

(a) air oxidation

12-16 0.12 N2 3 × 10-5 H2O(l) 75% N2/25% steam 75% N2/25% steam 75% N2/25% steam 75% N2/25% steam

16-17 0.12

17-21 0.12

N2 N2 N2 N2

air air

0-1 0.12

1-11 0.12

11-12 0.12

N2 N2 N2 N2

10% CO/90% N2 9% CO/9% CO2/82% N2 10% CO/90% N2 9% CO/9% CO2/82% N2

N2 N2 N2 N2

(b) purge 17-20 0.12

N2 N2

fig nos. 4-10 4-10 4-10 6a, 7, 9 4, 6b, 7, 9, 10 7 5, 7

Reoxidation to FeO required air (a).

sensitive NDIR for [CO] (ABB Easyline 0-2000 ppm vol); and (3) a thermal conductivity meter for [H2] (ABB Caldos27, 0-30 vol %). Mass balances on CO and CO2 confirmed that reduction to a specified oxidation state of the metal had occurred for a given temperature with errors less than 10%. Depending on the experiment, the bed was loaded in one of two ways. In each case, iron oxide (Fe2O3) particles were sandwiched between layers of inert quartz sand (425-500 µm), as shown in Figure 2. The upper layer of sand aided in preheating the inlet gas prior to contact with the iron particles; the lower layer prevented iron particles from passing through the holes in the perforated plate. For the experiments described in section 4.1, a packed bed of 20.0 g of iron oxide particles (425-600 µm; dp ) 513 µm) was used. The active bed had a length of L ∼ 200 mm and an inner diameter of D ∼ 10.2 mm, giving L/dp ) 390 for the bed of iron oxide and D/dp ) 20. Here, L is the bed length, dp is the particle diameter, and D is the bed diameter. The bed length of 200 mm was just shorter than the length (250 mm) of the heated section in the furnace. For experiments described in sections 4.2-4.6, the bed was made by carefully mixing 1.0 g of iron oxide particles with 4.0 g of inert sand. The 1 g of active material was mixed with sand to prevent maldistribution of reactant gas and to ensure near isothermal operation during the reduction and oxidation steps. For experiments with only 1 g of particles of Fe2O3, L/dp ) 137 and D/dp ) 20. To rule out any reaction of the reactor (i.e., stainless steel) or impurities in the sand, a series of blank experiments were performed with the reactor filled with only sand. The conversion of an equimolar mixture of CO and CO2 in this case was less than 1% of the conversion seen in a typical experiment when Fe2O3 was reduced to Fe0.947O by CO. 4. Results 4.1. CO2 Capture and H2 Production. Figure 3 shows the composition of the effluent gas as a function of time for an experiment with 20 g of Fe2O3 in a packed bed. Initially, N2 was flowing through the reactor. From t ) 90 s to t ) 810 s, the reducing gas was passed through the bed. Nearly all the entering CO was converted to CO2 and the mole fraction of CO at the exit remained consistently below 5 ppm vol. Once the Fe2O3 was exhausted, i.e. converted to Fe0.947O or Fe3O4, CO broke through between t ) 810 s and t ) 850 s. The reducing gas was then turned off, so that minimal CO escaped from the bed. Next, the N2 purge stream was turned on from t ) 850 s to t ) 1500 s. During this time, the trace mole fraction of CO decayed, as shown in Figure 3. Next, from t ) 1500 s to t ) 2400 s, steam was supplied and hydrogen of high purity was produced; the concentration of CO did not exceed 5 ppm vol during this phase. The purity of this manufactured hydrogen therefore exceeded the constraint required by PEM fuel cells, where CO levels above 50 ppm vol are known to poison the anode.

Figure 3. Reduction and subsequent oxidation in steam for the transition from Fe2O3 to Fe0.947O at 1023 K with a 20 g charge of Fe2O3. The compositions of the inlet gases are shown; the dashed vertical lines (- - -) indicate the times when the inlet gas to the reactor was changed.

Figure 3 demonstrates the feasibility of using the redox reactions (R1-R5 and R7) of iron oxides to produce both CO2 suitable for sequestration and H2 pure enough for use in PEM fuel cells. A mass balance on the effluent stream during reduction (90 < t < 850 s) shows that more CO2 is produced than is theorectically obtainable for the Fe2O3 to Fe3O4 transition alone. Using the thermodynamically consistent assumption that the charge of 20 g of Fe2O3 reacts sequentially to form Fe3O4 and then Fe0.947O, the mass balance reveals that ∼66% of the Fe3O4 formed had been reduced to Fe0.947O. In the remaining sections, the ability of the iron particles to produce hydrogen using various transitions of iron is investigated. Since a sustained CO2 effluent gas during the reduction phase is no longer the main objective, a shorter packed bed with an initial loading of 1.0 g of Fe2O3 was used. 4.2. Iron Oxide Reduction and Oxidation. One question regarding the redox reactions of iron oxide in a packed bed is whether the transitions between metal oxides occur at sharp fronts or more gradually. Figure 4 shows mole fractions in the effluent gas as a function of time for an experiment in which 1 g of Fe2O3 was reduced to Fe0.947O at 1023 K. This experiment is identical to that in Figure 3 except that 1 g, not 20 g, of iron oxide was used and the reducing gas mixture was not switched off at the point of CO slip. Even in this short bed, complete conversion of the inlet CO occurred for 15 s (such that the CO slip was less than 0.05 vol %) between t ) 60 s and t ) 75 s. A simple mass balance shows that the inflow of CO required for the theoretical conversion of 1 g of Fe2O3 to Fe3O4 is nearly identical to the amount entering the bed in this time period. The kinetics of the transition from Fe2O3 to Fe3O4, therefore, seem fast enough to reach equilibrium conversion within the residence time of gas in the bed, which for the given conditions and an assumed voidage of 0.4 was ∼0.2 s. For the transition from Fe3O4 to Fe0.947O, Figure 1 gives Kp ) pCO2/pCO ) 1.87 at 1023 K. This equilibrium value corresponds to a kink in the outlet [CO] and [CO2] in Figure 4. The

Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008 7627

Figure 4. Reduction and subsequent steam oxidation for the transition from Fe2O3 to Fe0.947O at 1023 K using a 1 g charge of Fe2O3. (A) Total conversion of the inlet CO as Fe2O3 is converted to Fe3O4; (B) Drop in the CO2/CO ratio to Kp for the transition between Fe3O4 and Fe0.947O; (C) Kink corresponding to Kp ) pCO2/pCO ) 1.87; (D) Slower kinetics for the transition from Fe3O4 to Fe0.947O. The dashed vertical lines (- · -) indicate the times when the inlet gas to the reactor was changed.

ensuing change in slope demonstrates that the reaction from Fe3O4 to Fe0.947O (R2) is no longer limited by the inlet flow of gas; instead, kinetic and mass transfer effects come into play. Figure 4 shows that the total time for reaction R2 to reach completion is t ) 200-75 ) 125 s; significantly longer than the reaction time to reduce Fe2O3 to Fe3O4. At t ) 200 s, the effluent gas concentration had reached the inlet condition of [CO2] ) 9.2 vol %, [CO] ) 8.8 vol %. A mass balance shows that if reduction occurs in a stagewise process, Fe2O3 to Fe3O4 to Fe0.947O, 93 mol % of the Fe3O4 formed was reduced to Fe0.947O. Besides examining the reaction times for the various metal oxide transitions during reduction, the overall reduction time was also compared to the overall oxidation time. Figure 5 shows results when the iron oxide was (a) reduced to Fe0.947O and (b) reoxidized to Fe3O4. In the first cycle, the oxide started as Fe2O3; subsequent cycles began with Fe3O4. Figure 5a shows the cumulative amount of CO2 released during the first five cycles of reduction, calculated by integrating the difference between the outlet and inlet mole fractions of CO2 with respect to time. The theoretical amount of CO2 generated in going from Fe3O4 to Fe0.947O can be determined from the mass of iron oxide loaded into the reactor, viz. 1 g of Fe2O3. Figure 5a shows that the cumulative amount of CO2 generated, given by the solid lines (s), was ∼90% of the theoretical value, given by the horizontal dashed line (- - -). Also shown in Figure 5a is the CO2 level for the case where all the entering CO is converted to CO2 (- · -). Only for the first cycle is the experimental curve a tangent to this line, again demonstrating that the inflow of reactant gas is rate-limiting for the initial reduction of Fe2O3 to Fe3O4. Figure 5b shows the cumulative amount of H2 produced when steam oxidized Fe0.947O back to Fe3O4 (R5). The time for the reaction with steam to reach completion is noticeably shorter than that for the reduction in the mixture of CO2 and CO; this discrepancy becomes more apparent after more cycles. Here, it should be noted that the mole fraction of steam was 25 vol % (5.2 mol/m3 at 101.3 kPa, 298 K), whereas that of CO had been 9 vol % (3.7 mol/m3 at 101.3 kPa, 298 K), so a direct comparison of the rates of reaction between reduction and oxidation is not possible.

A comparison between the final molar amounts from the curves in Figure 5a and b is then given in Figure 5c. It is clear that the amounts of CO2 released during the reduction of Fe3O4 to Fe0.947O and the H2 produced during the oxidation of the reduced oxide back to Fe3O4 are in good agreement. Thus, it seems that all the metal which is reduced to wuestite is subsequently available for reoxidation with steam. Figure 5d plots the time to reach 90% of the final value for the curves in Figure 5a and b. For hydrogen this is the time to reach 90% of the final value at t ) 320 s, while for CO2 this is the time to reach 90% of the final value at t ) 580 s. Figure 5d demonstrates a loss of reactivity from one cycle to the next for both reduction and oxidation. This loss of reactivity appears to be consistent from cycle to cycle and conforms to a linear fit. 4.3. Reduction to Fe0.947O versus Reduction to Fe. If the iron system is to be used for producing hydrogen, one important consideration which has to be addressed is how far to reduce the oxide. Reduction to either Fe0.947O or Fe depends on the composition of the syngas available and the restrictions imposed by chemical equilibrium at the chosen reaction temperature; see Figure 1. Reduction to Fe would be advantageous, since the capacity for producing hydrogen over the Fe3O4 to Fe transition is approximately four times greater than that for the Fe3O4 to Fe0.947O transition. To examine the difference between reducing Fe2O3 to Fe versus reducing it to Fe0.947O, redox experiments over both transitions were performed at 873, 1023, and 1173 K. Figure 6a shows that for the Fe2O3 to Fe transition at all three temperatures the amount of H2 produced decreases rapidly after the first cycle. The most precipitous drops are observed for particles at 1023 and 1173 K, suggesting that thermal sintering could contribute to the fall in reactivity. At 873 K, the production of H2 declines less sharply from one cycle to the next; however after ten cycles, H2 production at all three temperatures is unsatisfactorily low, having dropped below that achievable with the Fe2O3 to Fe0.947O transition. In contrast to Figure 6a, hydrogen production from Fe0.947O remains more consistent from cycle to cycle in Figure 6b. Here, the average amount of H2 produced during the tenth cycle is ∼80% of that produced during the first cycle. Thus, it is preferable to reduce Fe2O3 to Fe0.947O instead of to metallic Fe. Note that the reduction time of 10 min was kept constant for experiments (Table 1); for reduction of Fe2O3 to Fe, the outlet gas had not returned to its inlet value (10 vol % CO, 90 vol % N2) before the valve switch, indicating that the metal oxide was still undergoing reduction and that carbon was being deposited. 4.4. Oxidation to Fe2O3 versus Fe3O4. Besides the issue of how far to reduce the metal oxide, there is also the question of how far to reoxidize, i.e. whether to go back to Fe2O3 or to Fe3O4. For the proposed process, the presence of Fe2O3 near the reactor’s outlet is crucial for minimizing CO slip during the reduction of the metal oxide. Other processes for hydrogen production proposed in the literature,4 however, only reoxidize to Fe3O4. Final oxidation in air has the potential to oxidize any contaminants, e.g. carbon or sulfur, deposited on the particles, but it also causes a large temperature rise in the bed and so could result in thermal sintering. To examine the effects of oxidation in air, experiments were performed with and without additional oxidation of Fe3O4 to Fe2O3. To ensure that the particles were reacted for the same cycle times, a stream of N2 replaced the air in experiments without the additional oxidation step. Measurements indicated that the final oxidation in air caused a temperature rise as high as 50 K in a packed bed initially containing 1 g of Fe2O3

7628 Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008

Figure 5. Cycles of reduction and oxidation with a 1 g charge of Fe2O3 at 1023 K: reduction by inlet gas containing CO 9, CO2 9, and N2 82 vol %, followed by oxidation using steam 25 and N2 75 vol %. The bed was Fe2O3 at the start of the first cycle; subsequently, the oxide cycled between Fe3O4 and Fe0.947O. (a) Cumulative integrals of CO2 produced over five cycles. (b) Corresponding cumulative integrals for H2. (c) Comparison between the five final values of each graph. (d) Times to reach 90% of the final values in parts a and b. In parts a-c, the dashed horizontal line (- - -) marks the theoretical limit of H2 or CO2 for conversion of the original charge between Fe3O4 and Fe0.947O. The sloped line (- · -) in parts a or b marks the integral of the inlet flow for reactant (a) CO or (b) steam.

Figure 7. Total amounts of hydrogen produced by the transitions of iron oxide at 1023 K showing the effect of the number of cycles of reduction/ oxidation on producing hydrogen. In each case, the basis was a 1 g charge of Fe2O3.

Figure 6. Effect of the number of cycles of reduction/oxidation on the production of hydrogen at different temperatures for alternative transitions: (a) Fe2O3-Fe, (b) Fe2O3-Fe0.947O. The dashed line gives the theoretical amount of H2 starting with 1 g of Fe2O3.

particles. A simple heat balance confirmed that the theoretical temperature rise when converting Fe3O4 to Fe2O3 should be just over 50 K.

Figure 7 shows the total amount of H2 produced during the transitions from Fe2O3 to Fe, Fe3O4 to Fe, Fe2O3 to Fe0.947O, and Fe3O4 to Fe0.947O. From Figure 7, the drop in hydrogen production for the Fe2O3 to Fe transition follows the same course as that for Fe3O4 to Fe. Little difference in the capacity for producing hydrogen is observed between the transitions Fe2O3 to Fe0.947O and Fe3O4 to Fe0.947O either. Comparing the results for these transitions, Figure 7 demonstrates that the temperature rise during the additional oxidation in air to give Fe2O3 seems to have no adverse affect on the amount of hydrogen generated; a slight improvement can even be observed for later cycles when the final oxidation is with air. It should be noted that the reducing gas used in these experiments was of laboratory grade. Consequently, any benefits from the additional oxidation burning off contaminants other than deposited carbon, which might be present in industrial reducing gas, would not have been observed. 4.5. Deposition of Carbon. Thermodynamically it is possible that carbon was deposited on the particles when iron oxide was

Ind. Eng. Chem. Res., Vol. 47, No. 20, 2008 7629

Figure 8. First 60 s of the oxidation of Fe to Fe3O4 in steam at 873 K. Prior to the oxidation, 1 g of Fe2O3 had been reduced by a the mixture, CO 10 and N2 90 vol %, depositing carbon on the particles. Contamination of the H2 by CO and CO2 is apparent.

Figure 10. Hydrogen production (average CO level < 25 ppm vol) of the Fe0.947O to Fe2O3 transition at 1023 K. The initial charge of Fe2O3 was 1 g.

during the oxidation in steam of Fe0.947O to Fe3O4 at 1023 K. At this temperature, the Boudouard reaction (R6) is thermodynamically infeasible, so carbon will not be deposited during reduction of the metal oxide in CO. The average concentration of CO during the production of hydrogen was determined to be 23 ppm vol, demonstrating that a purge of N2 for 60 s between the reduction cycle and the oxidation with steam is sufficient to produce ultrapure hydrogen. The decreasing [CO] curve in Figure 10 shows residual CO from the reduction of Fe2O3 to Fe0.947O being swept from the system. Thus, it seems that the purity of the hydrogen can be specified to a desired level by adjusting the length of the N2 purge accordingly. 5. Discussion

Figure 9. Oxidation in air for the Fe-Fe2O3 and Fe0.947O-Fe2O3 transitions at 873 K for cycles 2, 3, and 4. CO2 is produced by oxidation of residual carbon deposited during reduction of the 1 g charge of Fe2O3 by CO. Even though the Boudouard reaction (R6) is thermodynamically favored, little CO2 is produced for the case of Fe0.947O.

reduced by CO to Fe and also in those experiments involving reduction to Fe0.947O at 873 K. Once deposited, the carbon could be reoxidised to CO or CO2 during oxidation in steam or in air. Figure 8 shows the mole fractions of CO2, CO, and H2 as functions of time for the first cycle, oxidizing Fe to Fe3O4 in steam at 873 K. The mole fraction of CO reached a maximum of 4600 ppm vol and exceeded the benchmark of 50 ppm vol for the duration of the oxidation stage; see Figure 8. If the particle was reoxidized to Fe2O3, any residual solid carbon, not oxidized during the inflow of steam, reacted with oxygen in the air to form CO2. Figure 9 shows the amounts of CO2 produced during oxidation with air at 873 K. It is notable that cycles 2 and 3 gave more CO2 in the off-gas than cycle 4, consistent with the fact that Fe catalyzes23 the Boudouard reaction (R6) and that the amount of reactive Fe diminishes with increasing number of redox cycles; see Figure 6a. Another interesting finding from Figure 9 is that very little CO2 was produced during oxidation in air, when Fe0.947O was the lowest form of the oxide used. This observation suggests that the rate of the Boudouard reaction (R6) is sufficiently slower than that of the reduction reactions (R1-R3) to prevent significant carbon deposition, when Fe is not present. 4.6. Purity of Hydrogen. If the hydrogen produced in the packed bed is to be used in low temperature PEM fuel cells, it must contain less than ∼50 ppm vol of CO. It has already been shown for the 20 g bed in Figure 3 that if a sufficiently long period of purging with N2 is used, H2 with the necessary purity can be produced. Figure 10 shows the mole fraction of CO

If carbonaceous fuels are to be used to produce hydrogen, capture of the byproduct CO2 is essential. Figure 3 showed that a process, which both made H2 of sufficient purity for use in PEM fuel cells and also could separate CO2 for potential sequestration, is feasible using the redox reactions of iron (R1-R5 and R7) in a packed bed. The CO slip was below the detection limit of the analysers used (