Ind. Eng. Chem. Res. 1991, 30, 1760-1770
1760
Kinetics of the A1kali-Metal-Car bonate-Cata1yzed Gasification of Carbon. 2. The Water-Gas-Shift Reaction Ronald Meijer,* Mieke Sibeijn, Minou R. B. van Dillen, Freek Kapteijn, and Jacob A. Moulijn Department of Chemical Engineering, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands
The catalyzed water-gas-shift (WGS) oxygen exchange takes place in the alkali-metal cluster anchored to the carbon surface. The WGS oxygen exchange rate shows the same dependency on the potassium loading as gasification in COPand HzO and is dependent on the partial pressures of both the reducing and the oxidizing agents. The oxygen exchange rate is low if C 0 2 is present in the gas mixture. For different alkali metals, the same rate behavior as with C 0 2 gasification is observed. For the kinetic description of alkali-metal-catalyzed oxygen exchange in H20, COz, H2, and CO containing gas mixtures, i.e. WGS, a three-step model is proposed in which only one step is considered to be in quasi-equilibrium. In this model COz,CO oxygen exchange proceeds through a C02-m intermediate, the amount of which plays a crucial role in the observed oxygen exchange activity. T h e estimated parameters of this model obey thermodynamic constraints, providing additional support for the selected model.
Introduction In spite of extensive research into the mechanism of the alkali-metal-catalyzed gasification, there is still a lot of discrepancy between the mechanisms proposed for this reaction. Due to the fact that steam gasification is more difficult to describe because several reactions take place simultaneously (eqs 1-3), most kinetic studies deal with the more simple C02 gasification (eq 4). C + H20 + H2 + CO
+ H2O + H2 + C02 2CO + 2H2 + CH, + COP c + cop + 2co CO
(1)
(2)
(3) (4)
Alkali-metal-catalyzed gasification of carbonaceous materials is generally assumed to proceed by an oxidation, reduction cycle (eqs 5, 6) in which oxygen is transferred to the carbon surface by means of the alkali-metal cluster present on this surface. Desorption of the carbon-xygen k
H2O (C02) + * 5 H2 (CO) + 0-* k,
(5)
k
0-* + C f S C ( 0 ) + *
(6)
C(0) 2L co
(7)
k-8
intermediate (eq 7) producing CO is generally considered to be the rate-determining step in this process. In this model m and 0-m represent the catalytically active "empty" and oxygen-containing alkali-metal clusters, and Cf and C(O), the active carbon site in its reduced and oxidized state. Different intermediate species have been proposed for the catalytically active alkali species, such as M2C03,M20, metallic M, peroxide, alkali-metal phenolate groups, intercalates, or oxygen-deficientalkali-metal oxides (McKee and Chatterji, 1975; Wigmans et al., 1983; Yokoyama et al., 1980; Mims and Pabst, 1983; Wood and Sancier, 1984; Kapteijn and Moulijn, 1983). The steadystate rate equation (8) that can be derived if desorption of C(0) intermediate is assumed to be rate determining is similar for both C02 and H20 gasification (Kapteijn and Moulijn, 1986).
One of the principal reactions taking place during steam gasification is the water-gas-shift (WGS) reaction (eq 21, the main practical importance of which lies in the production of hydrogen or in the tuning of the CO/H2 ratio for hydrocarbon syntheses on an industrial scale. Therefore many investigators have studied the kinetics of the WGS reaction over iron, copper, chromium, and cobaltmolybdenum catalysts (Newsome, 19801, but fewer studies have been made on the role of the WGS reaction in the alkali-metal-catalyzedsteam gasification (Huttinger et al., 1986; Wigmans et al., 1983; Mims and Pabst, 1987; Kapteijn et al., 1986; Kimura et al., 1985). In modeling the steam gasification process, for a proper prediction of the product gas composition from the gasifier, the rate expression is essential not only for the gasification reaction but also for the WGS reaction. In principle this reaction is one of the oxygen-exchangereactions, which are known to be catalyzed by the alkali-metal-carbon system (Mims and Pabst, 1987; Cerfontain et al., 1987). They can be described by elementary steps that are part of the steam gasification model (eq 5), so kinetic modeling of the WGS reaction additionally yields rate data relevant for the steam gasification, provided the rate expressions are based on a kinetic model consisting of a series of elementary processes pertaining to the overall reaction. In this kinetic study, part 2 of a series on the alkalimetal-catalyzed gasification of carbon (part 1: Kapteijn et al., 19861, results are presented of the water-gas-shift reaction, catalyzed by the alkali-metal-carbon system. In order to establish a kinetic model for the alkali-metalcatalyzed WGS reaction, both forward (CO + H20-1 and backward (C02+ H2+) oxygen-exchangerates have been investigated over a broad range of partial pressures of reactants and products for a K2C03-carbonsample. These results were compared with those for Na and Cs. Kinetic Modeling The kinetics of the WGS reaction are in a first approach described by the two-step oxygen-exchange model (eqs 9,
0888-5885f 91/263O-1160$02.5O/O 0 1991 American Chemical Society
Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991 1761 Table I. Commaition of the Denominator (R)in Rate Exwession 14 for the Different Prop06ed Kinetic Models"
Ib
8, 10
IC
9,lo
IIa
8, lo,11
IIb
9, 10, 11
IIC
9,l o , 11
IIIa IIIb
9, 1 2 , u
IIIC
9, 12, 13
IIId
9,121 13
IIIe
9 , U , 13
OThe reactions that are not assumed to be at quasi-equilibrium are underlined.
10) as earlier proposed by Wigmans et al. (1983). Two other models that account for COPinteraction with the catalytically active oxygen-exchangesites are included. In
H 2 0+ *
Hz+ 0-*
k*
In Keg = a,
(9)
+ a,[1000/T] + az[1000/T]2
(15)
in which a. = -3.49, a, = 3.563, and az = 0.313
co + 0-* 'k-,o cop + *
(10)
coz + o-* T -cos-*
(11)
=
co + 0-*& coz-*
(12)
ec02 + *
(13)
c02-*
librium. They are listed in Table I. The thermodynamic equilibrium constant (K,) is calculated from (Kapteijn and Moulijn, 1986)
k13
h-I3
model 2 (eqs 9-11), active sites are blocked through C02 chemisorption. In model 3 (eqs 9, 12, and 131, a COz-* intermediate as part of the COz,CO exchange is introduced. With these three kinetic models and under the assumption of steady-state conditions, a constant number of active oxygen-exchange sites (Nox)and up to two steps in quasi-equilibrium, 11 rate expressions are derived, for both the net forward rox(HzO,CO) and backward rox(C02,H2) WGS oxygen-exchange rates (eq 14). The denominator
rox(H20,CO) = -rox(COz,H2) = p(CO)p(HzO) - P(COZ)P(HZ) /& Nox (14) R R of the derived rate expression is determined by the selected model and the reactions assumed at quasi-equi-
Keg = P(CO~)P(HZ) /p(H2O)p(CO)
(16)
Over the temperature range studied the average overall change in entropy (ASo = -32.8 J-mol-'SK-l) and enthalpy (AHo= -35.9 kJ-mol-') for the WGS reaction was calculated from eq 17. With eqs 16 and 17 the thermodynamic equilibrium CO and COz conversion can be calculated for any given reactant composition (Appendix B). K,, = exp[ASo/R,] exp[-AHo/R,T] (17)
Parameter Estimation. After substitution of eq 14 into the differential form of the design equation of an integral reactor (eq 18) an expression is derived, describing the CO or C02 conversion as a function of space time WciIFo, and the rate expression variables: feed composition, K,, T, PO,, and the rate parameters (Appendix A). F°CO,COz
dXCO,COz d Wci =Fox = f(feed,T&,P co,co,,rate params) (18)
The parameters in the kinetic models are the reaction rate constants of the elementary steps multiplied by the total number of oxygen-exchange sites (kiNox)and, if present, the equilibrium constants (Ki) of the reactions that are assumed a t quasi-equilibrium.
1762 Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991 Table 11. Constraints for Thermodynamic Properties H,O
+ C
_/+-+ rules I I1 I11 guidelines I I1
0 < -Asoad, < SO,, J-mol-leK-' -&Yoab> 0 kJ.mol-1
E, > 0 kJ.mol-'
0.00
-ASoad,> 40 J.mo1-l.K-l -Ason& < 51.0 - 1.4&Y0,ds
0.04
0.06
0.08
K/C,
Estimation of the parameter values in the kinetic model was performed by nonlinear regression methods (Simplex and Marquardt). The sum of squares of residuals (SSR = C(Xob - X d 2 )i.e. , the difference between the observed and calculated CO or COz conversion, was minimized to obtain a best estimate for the parameter values. Discrimination between the proposed models was done by comparison of the SSR values, residual analysis, and testing the physical significance of the calculated parameters. In order to be of fundamental significance, the estimated parameter values must obey the laws of thermodynamics with respect to enthalpy and entropy changes of the elementary steps, putting constraints on their numerical values (Table 11) (Boudart et al., 1967). To reduce the computational difficulties that can arise from the strong correlation between the frequency factor ( k , J and the activation energy ( E , J the reparameterization proposed by Mezaki and Kittrell (1967) is applied (eq 19).
-
0.02
( 19)
k 6j
Initial estimates of the rate parameters were obtained by separately fitting the H,O,CO and COz,Hzoxygen-exchange data, assuming differential reactor behavior and linearization of the rate equation. These values were then used as starting values in the nonlinear parameter estimation.
Experimental Section Apparatus. WGS experiments are performed in a fixed-bed flow reactor described in detail elsewhere (Kapteijn et al., 1986; Kapteijn and Moulijn, 1985). Basically it consists of a gas mixing section, an oven (5"" = 1273 K) containing the carbon sample in a quartz tube (i.d. = 3-7 mm), and a gas chromatograph (HP 5190 A) for product analysis (dual column, He carrier, TCD). All experiments are performed in a ceramic (2-16-bar) or quartz (1.5-bar) reactor. Sample Preparation. The activated carbon used in this study is Norit RX1 Extra, an acid-washed, steamactivated peat char with a high specific surface area (1100 m2.g-' (COz adsorption at 273 K), 1500 m2.g-l (N2 adsorption at 77 K), particle size 0.254.6 mm, 3 wt % ash). Addition of catalyst is performed by pore volume impregnation with an aqueous M2CO3 (M = Na, K, Cs) solution. The catalyst loading is expressed as the atomic alkali-metal to carbon (M/C) ratio. Experimental Procedure. A reproducible sample for each WGS experiment was obtained by drying in situ (T = 473 K, He) followed by isothermal gasification (7' = 10oO
Figure 1. Steady-state reaction rates (r,) for C02 (A)and H 2 0 (+) gasification a t lo00 K (at 25% burn-off) and rox (H20,CO) ( 0 )at 833 K (after partial gasification in H20 at lo00 K to 25% burn-off) as a function of the K/C ratio of the sample.
K, Pt = 1.5 bar, Ft = 140 p m o l d , p(H20)or p(COz) = 0.5 bar, balance He) to a steady gasification level (20-30% burn-off). Subsequently the sample is cooled to 673 K, the desired gas mixture is generated, and a WGS experiment is started. The WGS oxygen-exchange rate rox is measured between 673 and 900 K, with the temperature being varied stepwise, and is expressed as (mol oxygen exchanged)-(molcarbon initially present)-'.s-'. The oxygen exchange rate of a particular reactant mixture is referred to as rox(H20,CO)for the forward shift and as rox(COz,Hz) for the backward reaction. H20,C0 oxygen-exchange experiments are performed with a feed consisting of H20,C0 and a balance of helium in which the molar reactant ratio q [=(p(H20)/p(CO))fJ was varied from 0.16 to 6.7 (Pt= 1.5 bar, Ft = 140 pmold). C02,H2oxygen-exchange experiments are performed with variation of the total pressure (1.5-16 bar) and reactant ratio [q' = (p(H2)/p(C02))fd= 0.14 - 7.01. For each q and q' value a fresh sample is used. Inhibition effects of products on oxygen-exchange rates are studied by (partially) replacing the helium in the feed (Pt= 1.5 bar) by a product. Data Handling. CO or COz conversions were calculated from respectively the CO2/CO and CO/CO2 flow ratios in both reactant and product gas. lCO2/CO10ut - ICO2/Win xco = 1 + ~co2/coJo", ICO/CO2lout - lCO/CO21in
xco* =
1 + 1co/co2lout
Within the temperature range used for kinetic modeling of the WGS reaction, it was verified that no gasification (CO + C02 balance) or methanation (CH, production) reaction occurred.
Results Unless stated otherwise the results that are presented or discussed have been obtained with a KzC03-carbon, (K/C), = 0.019, sample. At all experimental conditions the measured C02 or CO conversions were well below the thermodynamic equilibrium conversion. Oxygen-Exchange Rates. In Figure 1,WGS rate data (rox(H20,CO)at 833 K) and C 0 2 and H20gasscation data (r, at 10oO K) as a function of the initial potassium loading are shown. Clearly in all three cases the same dependency on the K/C ratio is observed. The carbon itself hardly showed any oxygen exchange reactivity. In comparing the H20,C0and C02,H2oxygen-exchange rates at 833 K with the gasification rates in both COz and H 2 0 at respectively
Ind. Eng. Chem. Res., Vol. 30,No. 8, 1991 1763 Table 111. Oxygen-Exchange( r ) Rates and Gasification (r,)Rates (mol*(molof C,)-~*S-~; in Different Gas Mixtures over M2C08Carbon(M/C = 0.019) Samples Wrn 104rOx COz,Hzat HzO,CO at COz at HzO at T = 8 3 3 K T = 8 3 3 K T=lOOOK T = 9 7 3 K Na 0.34 0.77 0.36 0.78 3.09 0.60 K 1.22 0.96 CS 3.04 12.7 0.90 0.85 PCO, 1
0
2
4
3
- 7
(PH,O/PCO),,,,
pH, = 1
Figure 2. rox(HzO,CO) as a function of the (p(HzO)/p(CO)), ratio. Influence of variation of p(C0) @(HzO) = 0.5 bar, open symbols) or p(HzO) @(CO) = 0.5 bar, closed symbols) on the HzO,CO exchange rate over a 10 wt % KzCOS-Norit RX1 Extra sample at 833 (o,.), 793 (A&, and 753 (v,r) K.
I \
0.10
0.0s
Ne
t \
-
1
--.--
1.2
1.1
1.3
- 3 -
14
-8..
6
-0-
2
1.4
1000/T ( l / K )
Figure 4. rox(COz,Hi) over a 10 wt % KzCOa-Norit RX1 Extra sample as a function of 1/T (l/K) for p(Hz) = 1bar (closed symbols) and p(Hz) = 2 bar (open symbols) at different Hz/CO2feed ratios.
"..
' I \ \ \\
\ : \
3
1
1.0
8X
.....
-*-
-*-
-*
0.5
1
I
L
0.00 ... .
Oms
I \
K
0
2
4
6
a
(PH,/PCO,),,,,
Figure 5. Measured COz conversion at different temperatures as a function of the Hz/CO2(4') ratio in the feed (at Pt = 4 bar): 753 K (+I, 793 K (A),833 K (O), 873 K (V),900 K (e),950 K (M).
0.001 0
0.00 16
1/T (1/K) Figure 3. CO conversion in H*O,CO oxygen exchange over a M&08-Norit RX1 ((M/C)i = 0.019, M = Na, K, Cs) sample as a function of 1/T (l/K). (K, 400-mg sample; Na and Cs, 100-mg samples). Molar feed ratio: HzOCOHe = 1:l:l (+); HzOCOCO2:He 661:5 (e);Hz0:CO:HZ = 1:1:1 (M).
lo00 and 973 K for different alkali metals, the same trend in the oxygen-exchange rate is observed as with gasification in COz (Table 111); In Figure 2 the influence of variation of p(C0) or p(H20) at respectively p(HzO) or p(C0) constant (0.5 bar) on rox(HzO,CO) is plotted as a function of (p(H20)/p(CO))fd. I t can be seen that rox(HzO,CO)is dependent on both the partial pressure of the reducing and the oxidizing agents.
For all alkali metals investigated (Na, K, Cs), addition of H2 to a feed, containing CO and H20, shows an inhibiting effect on the measured CO conversion at high temperatures, whereas addition of a small amount of COz shows a much stronger inhibition (Figure 3). Figure 4 shows that rox(COz,Hz)is predominantly dependent on the H2 partial (P(Hz))and total (PJ pressure. With increasing p(H2) and Pt (q' constant; open vs closed symbols), rox( C02,Hz)increases, whereas with increasing p(COz) (p(Hz) constant; open and closed symbols), a decrease in rox(C02,H2)is observed. In Figure 5, the measured COz conversion at different temperatures (Pt= 4 bar) is plotted as a function of (p(H2)/p(COz))fd.The COz conversion increases with increasing T and q'. Although not shown, this is observed at all investigated pressures (Pt = 1-16 bar). In the case of COz,H2oxygen exchange, addition of HzO to the feed decreases the measured COz conversion significantly for all alkali metals investigated, whereas only for Cs a minor effect of CO addition is observed (Figure 6). Parameter Estimation. The best parameter estimates, with respect to a minimum SSR value, from fitting the experimental data on the derived rate expressions are listed
1764 Ind. Eng. Chem. Res., Vol. 30, No. 8,1991 Table IV. Results of Non-Linear Regression of the Kinetic Data on the Derived Rate Expressions for E&O,-Norit ((K/C), = 0.019) Samples minimum SSR value no. of HzO,CO COZ,H2 complete residual trends, model Darame data data dataset remarks Ia 6 0.0045 0.1208 trend with p ( C 0 2 ) IIa 8 0.0045 0.0169 0.0864 high q'bad fit IIIa 8 0.0104 0.0161 0.066 trend in q and PW/F IIIb 8 0.0050 0.0131 0.0317 best tit IIIc 6 0.0103 0.0144 0.1455 low q bad fit IIId 8 0.3180 overall bad fit IIIe 6 0.2877 overall bad fit 783 813 799 T.1 K 185 96 no. of data points 89
0.01
Na
0"
0 X
0.00 0.2
I
1
Table V. Best Elltimates of the Rate and Equilibrium Constants for Model IIIbO
k.JVox = 1.57 X lo3 ex.[
- -72.550 R,T
K9 = ex,[
1
u] -1
exp[ 16.019
R,T
k-$Vox = 90.2 exp
-J [-:y I-lF1 J
k l g O X = 1.43 X lo' exp
k - l g o X = 546.8 exp
J
-
0.3
I
I
Cs
ON 0 X
0.00 15
0.0010
1/T (1/K) Figure 6. COO conversion in COz,Hz oxygen exchange over a MzC03-Norit RX1 ((M/C)i = 0.019, M = Na, K, Cs) sample (100 mg)as a function of 1/T (l/K). Molar feed ratio: C02:H2:He= 1:l:l (v);COP:H&O = 1:1:1 (A);COZ:H2:HzO 1:1:1 (0).
a
Unita: AHo and E, in kJ-mol-', ASo in J.mol-'.K-', ks,k+, and in bar%-', k13 in s-l, Nox in mole(mo1 of Ci)-l.
CO and H 2 0 pressures, independent of the H2 partial pressure, and is strongly inhibited by the presence of C02. in Table IV. Only seven instead of the initial eleven The disadvantages of describing kinetic data with a models are listed, because the results of model Ib, IC,IIb, power law rate expression are that it is only valid for the and IIc showed a very poor description of the experimental conditions studied, it disregards the influence of the reverse data. The lowest SSR value is obtained with expression reaction, it has no mechanistic basis, and therefore it IIIb. The other models produced higher SSR values and cannot be used for extrapolation purposes. showed trending effects in the residual distribution as a In several studies alkali-metal carbonatecarbon systems function of independent variables and/or had thermodyhave proven to be good WGS catalysts. Huttinger et al. namically unacceptable parameter values. The parameter (1986) have studied the H20,C0 oxygen-exchangerate over values for expression IIIb are given in Table V. K2C03-activated carbon. Their rate data were correlated by a power law rate expression, assuming differential reDiscussion actor behavior. They concluded that the H20,C0 oxyConventional WGS catalysts can be divided into two gen-exchange rate is dependent on both the partial CO and groups; low-temperature (LT)catalysts based on copperH 2 0 pressures ( a = 0.6, b = 0.4) and that addition of H2 zinc and high-temperature (HT) catalysh based on ironto the feed did not affect the oxygen-exchange rate (d = chromium. Relatively new WGS catalysts are based on 0). The effect of C02 was not investigated. Wigmans et sulfur-tolerant cobalt-molybdenum catalysts, which are al. (1983) studied the alkali-metal-catalyzedWGS reaction promoted by the addition of alkali-metal salts (Newsome, under steady-state gasification conditions at low steam 1980). pressures and concluded that the assumption that the Most kinetic WGS studies are predominantly performed WGS reaction is at equilibrium due to potassium catalysis on the H20,C0 reaction of which the kinetic data are is doubtful. usually fitted by an empirical power law rate expression Mims and Pabst (1987) have studied the oxygen ex(es 20). change-rates in H20,D2,H20,C0, C02,H2,and C02,CO mixtures systems and report a r = ~ [ p ~ C O ~ l a [ p ~ H ~ O ~ l b [ p ~(20) CO~~l c ~ ~ ~over H ~ ~alkali-metal-carbon ld decrease in rate in the order rox(H20,D2)>> rox(H20,CO) Generally it is found that 0 < a,b < 1, c < 0, and d = > rox(C02,H,) > rox(C02,CO). Addition of C02 to the 0, indicating that the rate is dependent on both the partial H20,D2mixture suppresses the oxygen-exchange rate to
k-13
Ind. Eng. Chem. Res., Vol. 30, No. 8, 1991 1765 Table VI. Global Partial Pressure Influence of Reactants on the Oxygen-ExchangeRates 'cot
---,
C01,CO
H2 HZO
co cop
D0.5 ~ X
(2)
but
Q/ND3< 5 x
lo-*
(3) Complete sedimentation of particles and negligible dispersion of the gas phase occurs when
Q/ND3> 5 x
(3)
Frijlink et al. (1984) have studied the suspension of solids in a 0.44-m4.d. mechanically agitated three-phase contactor with a six-bladed PTD. The blade angle waa 60'. They have studied the effect of sparger design and the distance between the sparger and the impeller on NSG. They found that reduction in power consumption is more when a pipe sparger was employed rather than a ring sparger. Further, for either of the spargers, reduction in
0888-5885/91f 2630-l77O$02.50/0 0 1991 American Chemical Society
