Ind. Eng. Chem. Res. 1987,26, 1565-1573
1565
Effects of Hydrogen Treat Rate and Hydrogen Mass Transfer in SRC-I1 Liquefaction of Coal Chandra P.P.Singh* Exxon Research and Engineering Company, Florham Park, New Jersey 07932
A simulation for well-mixed SRC-I1 coal liquefaction reactors is developed to analyze the observed effects of hydrogen treat rate and hydrogen mass transfer on the rate of liquefaction. The validated simulation shows that the observed enhancement in the rate of liquefaction with increase in the hydrogen treat rate results from a reduction in H,S partial pressure in the reactor. The hydrogen partial pressure is relatively insensitive to the hydrogen treat rate, and therefore, its contribution to the change in rate of liquefaction is very small. At very low levels of mixing power in the reactor, the rate of hydrogen mass transfer can have a significant effect on the rate of liquefaction. However, even for the lower range of mixing powers used in the bench-scale bubble column reactors, the effect of hydrogen mass transfer on the rate of reaction is small. High solubility of hydrogen in coal liquids is the basic reason for the small hydrogen mass-transfer effect. In coal liquefaction, hydrogen treat rate refers to the hydrogen to feed slurry weight ratio. In general, it has been considered to have only a minor influence on the rate of liquefaction. Accordingly, all the SRC-I1 experimental studies in bench- and pilot-scale units used a constant hydrogen treat rate of 0.04 (McIllvried, 1982; Singh et al., 1982a,b; Singh and Carr, 1983). Most of the other major coal liquefaction processes, such as EDS (Epperly, 1980), H-coal, etc., also used the same hydrogen treat rate in most of their studies. Significance of the effect of hydrogen treat rate on the rate of coal liquefaction seems to. have been based on concepts rather than any appropriate experimental evidence. It is generally believed that any change in pressure, process gas composition, gas flow rate, gas residence time, etc., can influence the rate of coal liquefaction only through a change in hydrogen partial pressure in the reactor, whereas the experimental studies show that hydrogen partial pressure has a small effect on the rate of liquefaction (Singh et al., 1982a). Also, the hydrogen partial pressure in the reactor does not change significantly with change in the extent of conversion of coal to distillates (Singh and Carr, 1983). Thus, it is easily inferred that hydrogen treat rate would not have any significant effect on the rates of coal liquefaction reactions. Following the same logic, it is possible to conclude that the rate of hydrogen mass transfer would have an insignificant influence on the rate of coal liquefaction. However, the rate of hydrogen mass transfer is considered to have another significant influence on liquefaction of coal. Formation and deposits of coke like materials in the reactor orland downstream equipment represents a major operational problem for any of the coal liquefaction processes. The observed coke deposits in the reactor and separation vessels of the SRC-I1process development unit P99 in Harmarville, PA, are well documented (McIllvried and Gall, 1981a,b). In the SRC-I1 P99 unit, the hightemperature-high-pressure separator vessel used to be easily plugged by coke deposits. A small purge of hydrogen helped to eliminate the problem. This successful use of hydrogen purge, duplicated in the bench-scale SRC-I1 units, and the general concept based on studies of retrogressive reactions (Painter et al., 1979; Shibaoka and Ueda, 1978a,b) provided the basis to consider a strong dependence of coke formation on availability of hydrogen in the
* Present address: Chemical Process Simulations,Monroeville, PA 15146.
reaction environment. Since even in a fairly well-mixed reactor there may be some stagnant regions with limited availability of hydrogen, an understanding of the dependence of rate of hydrogen mass transfer on the process and operating variables in an SRC-I1 reactor was considered important. The key variable in the studies of the hydrogen masstransfer effects in SRC-I1 coal liquefaction reactors has been the level of mixing power (Carr et al., 1982; Singh and Carr, 1986). However, one of the studies considered several other variables, such as hydrogen treat rate, split of hydrogen feed between the slurry feed preheater and the reactor, and the effect of hydrogen mass transfer for two different coals (Carr et al., 1982). The expected effects of mixing power on the rate of liquefaction reactions and coke deposits in the reactor were observed. Also, the yields from the two coals were as expected. However, the hydrogen treat rate was observed to have an unexpectedly strong effect on the rate of liquefaction. The effect of other variables was small and could not be quantified from the yield data. These workers did not provide any explanation for the observed effect of the hydrogen treat rate on the rates of liquefaction reactions; even its inclusion as a variable appears to be speculative. Also, the analysis of the hydrogen mass-transfer effect is only qualitative. The necessary information is available to develop a rational basis for the observed effects of hydrogen treat rate and hydrogen mass transfer on the rates of coal liquefaction reactions. A kinetic model for liquefaction of coal catalyzed by inherent minerals shows that the rate of coal liquefaction reaction is influenced by partial pressures of hydrogen as well as H2S(Singh and Carr, 1987). Therefore, hydrogen treat rate can have a significant influence on the rate of liquefaction reaction due to change in H2Spartial pressure even if hydrogen partial pressure was relatively unchanged. Also, the above kinetic model shows that the observed small effect of the hdyrogen partial pressure on the rate of reaction is due to the inhibitive effect of H2S partial pressure which compensates for the effect of hydrogen partial pressure. In the absence of the inhibitive effect of H2S, the rate of reaction would be directly proportional to the partial pressure of hydrogen. An earlier study of the effect of hydrogen mass transfer in SRC-I1 coal liquefaction was limited by the small observed effect of hydrogen partial pressure on the rate of liquefaction as well as the experimental data which contained only the hydrogen consumption as a function of mixing power (Singh and Carr, 1986).
0888-5885/87/2626-1565$01.50/0 0 1987 American Chemical Society
1566 Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987 Table I. Reactor and Process Variables reactor vol, ( V R ) , m3 impeller diameter (dim),m reactor temp ( T ) ,K reactor pressure (PI,MPa slurry residence time ( 7 ) , s coal concn in feed (lOOgfcod),wt % iron concn in slurry (lOOgI), wt %
9.8 x 10-4 0.0476 728 13.76 3600 30 1.98
This work describes the development of a model for analysis of the hydrogen treat rate, and hydrogen mass transfer influenced yields from the SRC-I1liquefaction of coal. Validation of the model is obtained from data of Carr et al. (1982). The data were carefully screened before being used. This is important due to the use of extreme conditions, such as very low levels of mixing power, high gas throughput, deposition of coke in the reactor, etc. Some of these extreme conditions were used only in these experiments, and due to the mass balance problems, some data points from the reported study (Carr et al., 1982) were dropped from a subsequent publication (King et al., 1984). The consistent data set obtained after the screening is used to validate the model which is used to provide a better understanding of some of the key components of SRC-I1 conversion of coal to distillates.
Experimental Section The reactor used in these experiments was a 1-L stirred autoclave. Slurry feed to the reactor consisted of coal, process solvent, and processed slurry derived from SRC-I1 processing of the same coal on a process development unit (PDU-P99) (McIllvried, 1982). Use of such a feed allowed this once through processing of coal to simulate the SRC-I1 recycle operation. Details of the experimental setup, method of operation, and product analysis are available elsewhere (Singh et al., 1982a; King et al., 1984). These experiments were designed to study the effect of hydrogen mass transfer on the rate of liquefaction reactions. The hydrogen mass-transfer coefficient was varied by changing the stirrer speed from 2.5 to 16.67 rps. Also, the effect of hydrogen treat rate and split of the hydrogen feed between the preheater and the reactor were investigated. Ireland Mine coal was used to investigate the effect of all of the above variables, while composition of the feed and other process conditions were held constant at the values listed in Table I. Another coal (Powhatan No. 6) was used in some (four) additional runs only to confirm that the dependence of the rate of reaction on the hydrogen mass-transfer coefficient was not specific to a given
(Ireland) coal. Coal liquefaction products were classified into the 10 categories normally employed in the SRC-I1 experiments. These are (1)inorganic mineral matter, ASH; (2) pyridene-insoluble organic matter, IOM; (3) solvent refined coal, SRC; (4) heavy distillate, HD (288-482 "C); (5) middle distillate, MD (193-288 "C); (6) light distillate, LD (C5+-193 "C); (7) water; (8) byproduct gases, such as NH3, H2S,CO, and CO,; (9) C1-CI hydrocarbon gases; and (10) H2.Though three distillate fractions (HD, MD, and LD) were separately measured, the sum of the three, i.e., overall distillate yield, was reported. The distribution of products for the Ireland coal runs are presented in Table 11. The latter is a reproduction of the values reported by Carr et al. (1982). The runs are organized in a sequence of increasing stirrer speed at each level of hydrogen treat rate. The runs at the same hydrogen treat rate and the same stirrer speed are arranged in a sequence of decreasing liquid yield to check the reproducibility of data. Quality of Data. From Table 11, it is noted that with the exception of the four runs at 16.67 rps and a hydrogen treat rate of 0.06 or 6 g of H2/100 g of slurry (nos. 8-11), all other yields of distillates and SRC are reproducible within f l wt % maf coal. At the above-noted conditions, the distillate yield varies from 46.14 to 38.62 wt % maf coal and SRC yield varies from 30.08 to 38.05 w t % maf coal. A check for the overall, ash, and coal balances for these four runs shows that for the three runs with lower distillate yields (nos. 9-11 in Table 11) coal balances were 102.6%, 103.6%, and 105.2%, respectively. Coal balance for the fourth run (distillate yield 46.14 wt % maf coal) at the same condition was 101.1% , and that for all the runs at other conditions was mostly within 100-102%. The method of forcing 100% overall mass balance which results in greater than 100% coal balance suggests that the amount of heavy ends in the products for the three runs (nos. 9-11 in Table 11) was overestimated. It is also noted that the sequence of decrease in distillate yield (46.14, 41.61, 38.74, and 38.62 wt % maf coal) parallels the increase in coal balance (lOl.l%, 102.6%, 103.6%, and 105.4%). Therefore, the observed differences in distillate and SRC yields in the above-noted cases are likely to be a result of forced mass balance errors. It is very important to note that the significance of error resulting from forced (100%) overall mass balance increases with the increase in distillate yield. This is due to the forced mass balance procedure which assumes that lower or higher output of material from the system results
Table 11. Product Yields
no. 1 2 3 4
5 6 r
8 9 10 11 12
13 14
15 16 17
stirrer speed, rP5 3.33 3.33 3.33 5.00 5.00 6.67 10.00 16.67 16.67 16.67 16.67 2.50 3.33 3.33 6.67 16.67 16.67
Hz treat rate, g of H2/100 g of slurry 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 4.0 4.0 4.0 4.0 4.0 4.0
product yields, wt % maf coal c1-c4
measd 16.62 16.12 18.01 14.61 16.10 14.96 13.88 12.11
cases av
}1
1
13"' 14.80 14.31 15.86 15.401 17.31 15.15 13.54\ 15.40
16.92 15.35
13.78
16.36 14,47
byprod. cases 3.91 5.36 3.46 3.98 4.20 3.57 3.86 6.95 3.55 4.10 5.35 4.11 5.09 3.95 3.96 4.32 3.92
water 3.48 3.72 3.22 2.60 3.77 4.75 5.53 3.47 3.68 3.12 2.87 5.53 4.58 4.09 4.11 4.18 4.12
distillates (Cs-900 O F ) 37.00 36.95 36.46 40.40 38.43 43.14 44.20 46.14 41.61 38.74 38.62 33.10 32.90 31.26 37.15 37.95 36.81
SRC 35.70 34.60 36.88 36.40 35.61 33.06 31.69 30.08 36.33 37.12 38.05 39.00 40.34 41.02 37.41 37.46 37.44
IOM 7.91 5.50 5.67 6.66 6.60 5.18 6.01 5.10 6.14 5.97 5.20 6.80 5.13 5.90 5.47 5.79 5.73
H2 -4.62* -3.89 -3.70 -4.59 -4.65 -4.65 -5.19 -3.86* -5.19 -3.85* -4.40* -4.40 -3.44 -3.54 -3.32 -3.25 -3.42
Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987 1567 from changes in holdup of the heavy ends in the highpressure-high-temperature separator (Carr et al., 1982). If the product distribution is favorable toward the heavy ends, i.e., distillate yields are lower, the error due to such an assumption would be small. However, in the case of lighter products distribution, the above method can result in significant error. Also, an overestimate of heavy ends would give greater than 100% coal balance. Thus, the forced mass balance procedure appears to be the major reason for the large range of distillate and SRC yields at 16.67 rps and a hydrogen treat rate of 0.06. Also, all the runs having the mass balance problems were carried out with a H2 treat rate of 0.06. These were the only runs using a hydrogen treat rate of more than 0.04 which was used in all the previous work with this experimental system. I t is possible that the reactor operated at conditions close to hydrodynamic instability. In other words, at the higher gas throughput levels used in these experiments, the reactor could have occasionally flooded (Warmoeskerken and Smith, 1985). However, all other distillate and SRC yields appear to be reproducible, and they also present consistent variations with changes in stirrer speed and hydrogen treat rate. C1-C4 gas yields seem to differ by less than 2.2 or f 1.1 wt % maf coal under identical conditions. Differences of similar magnitudes are noted for byproduct gases and water yields. Except for the four runs marked with *, all the hydrogen consumption values appear to be consistent and reproducible within f O . l wt % maf coal, which is excellent. In view of more than an order of magnitude difference between the molecular weights as well as the mole fractions of various gaseous components, reproducibility of the gaseous component yield data obtained in these experiments is considered to be good.
Theoretical Model The theoretical description considers the influence of hydrogen mass transfer and change in the reactor gas composition on the rate of reaction given by the kinetic model (Singh and Carr, 1987). The kinetic model, the correlation for the hydrogen mass-transfer coefficient, the gas-phase mass balance, and their unification into an overall model are described in the following subsections. (a) Kinetic Model (Singh and Carr, 1987). The kinetic model describes the liquefaction of coal catalyzed by coal minerals, and it is applicable to the SRC-I1 process. The model considers coal to be instantaneously dissolved in the reactor. The dissolution splits coal into the various product components, each of which is a definite weight fraction of coal (f;Cod). The SRC fraction of the product undergoes a rate-controlled conversion to yield lighter products, each of which is a definite weight fraction of SRC (fiSRC). The rate of SRC reaction is given by rSRC= 2.7361 X lo5 X exp(-7916O/RgT)p~g~/(1 ~OPH,S)kg/(m3d (1) where R, is universal gas constant, kJ/(kmol.K); T is temperature, K; p H z and pHQSare partial pressures of H2 and H2S,MPa; and gI is mass fraction of iron in the slurry. The above expression gives the intrinsic rate of SRC reaction; Le., there is no effect of hydrogen mass transfer on the rate of reaction. In other words, the concentration of hydrogen in the reacting slurry corresponds to the solubility of hydrogen, and the gas- and liquid-phase concentrations of hydrogen are related by Henry’s constant CH21= HCHyg (2) where CHS and CHzlare gas- and liquid-phase concentra-
tions of hydrogen, kg/m3; and Henry’s constant, H, for hydrogen in coal liquids is given by (Singh and Carr, 1986) H = 1.2 exp(-3250/RgT) (3) From the ideal gas law, the concentrations and partial pressures of hydrogen are related by CHzg = 241pH2/T (4)
(b) Rate of Hydrogen Mass Transfer, RH2. Assuming the gas-side mass-transfer resistance to be negligible, the rate of hydrogen mass transfer from gas to the coal oil slurry in the reactor can be expressed as (5) RH, = k ~ a ( C * ~-, iC H ~ J where KLa is the volumetric mass-transfer coefficient, s-l; C*H21is the concentration of hydrogen at the gas-liquid interface which is related to the gas-phase concentration by Henry’s constant (eq 2); and CH 1 is the concentration of hydrogen in the bulk liquid, kg/mf. The partial pressure of hydrogen to be used in the SRC reaction rate expression (eq 1)for Hz mass-transfer-limited condition should correspond to the hydrogen concentration in bulk liquid, i.e., CHzl. With pHzand P*H~representing the Hz partial pressures corresponding to cHzl and c*H21,from eq 2,4, and 5, we have RH^ = 2 4 1 H K ~ a b -* P~H~J/T (6)
.
(c) Hydrogen Mass-Transfer Coefficient, KLa The mass-transfer coefficient (KLa) is a function of mixing power per unit reactor volume (E/VR), KLu = 0.1.1(E/VR)o’84 (7) where E, the mixing power, is the sum of the mixing powers provided by the stirrer (E,) and the feed gas (Eg), i.e., E = E, E, (8)
+
and E, and E, are given by E, = 0.1(Es2Ndim3/ &g056)0’45
(9)
E,, = 0.001Npp1N3dim5
(10)
E, = 0 . 0 0 1 V R ~ ~-~tg)g ~(l
(11)
and The various terms in the above equations are E,, is the mixing power for no gas flow through the reactor, kW/m3; N is the stirrer speed, s-l; dimis the stirrer (impeller) diameter, m; Q is the volumetric gas flow through the reactor, m3; Np is the power number (=5 for this system); p1 is the slurry density, kg/m3; V, is the volume of the reactor, m3; U, is the superficial gas velocity, m/s; tg is the gas holdup; and g is the gravitational constant, m/sz. Development of eq 7-11 is described elsewhere (Singh and Cam, 1986). (d) Hydrogen Mass-Transfer-Influenced Rate of Reaction. The rate of hydrogen reaction in the slurry must equal the rate of hydrogen transfer from gas to the slurry, i.e., rHp = RHz (12) where the rate of hydrogen mass transfer (RHP)is given by eq 6 and that for the reaction is given by (I3) rH2 = f 1 0 c o d p ~ f c o a l / 7 + flOSRCrSRC(l - €g) The first component of the above expression for rH2represents the amount of hydrogen consumed during initial dissolution of coal and the second represents the amount of hydrogen consumed in the SRC reaction. Due to the influence of hydrogen mass transfer on the rate of reaction,
1568 Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987
the concentration of hydrogen in the slurry ( C H 1) corresponds to p H 2 which is lower than its actual vafue p * ~ ~ . Therefore, p~~has to be used in the expression for FSRC (eq 1). Also, p * H and p H g have to be obtained from component mass balances for Hz and H2S, respectively. (e) Gas-Phase Mass Balance. In addition to the effects of process conditions and hydrogen mass-transfer effects, the gas-phase component balances also include the effect of hydrogen treat rate on the composition of gas in the reactor. 1. Hydrogen Balance. The molar flow rates of hydrogen and total gas in the feed to the reactor are given by NH2f = g H z f P l / ( T M W H z ) (14) NTf
(15)
NH2f/YHzf
where NHzfand Nm are molar flow rates of hydrogen and totalgas m the feed, kmol/(m3.s); gHzf iS the hydrogen treat rate; yHd is the mole fraction of hydrogen in feed gas; and MWH, is the molecular weight of hydrogen. Considering hydrogen consumption and gas generation during initial dissolution plus the rate-controlled reaction, the hydrogen and total molar flow rates at the reactor outlet are given by NH2
H, Treat Rate 50
__ Model
Data Points
.- - - -..
6 g of H,i100 g of slurry 4 g of H,/100 g of slurry
~
-
1
25 0
2
4
6
8
10
12
14
16
18
1
Stirrer Speed, rps
Figure 1. Comparison of experimental and model distillate yields. I
H, Treat Rate
Model __
Dala Points
----
-. .
6 g of H,/100 g of slurry
--
=
NH,f
4- kf~oalPlfiocd/~ 4-
fioSRCrsRc)/MWH,- NH~S (16) 9
NT = N H4-~ (NTf- N H 2 f )
g f c o a l P l C CfiC""'/MWi) / i=7 9
T
+ 8
~ ~ RCfiSRC/MWi) cC (17) i=7
Y H ~=
NHJNT
10
12
14
16
18
Stirrer Speed rps
Figure 2. Comparison of experimental and model SRC yields.
(18)
where N H 2 , N H @ , and NT are molar flow rates of hydrogen, H2S,and total gas at the reactor outlet. It is to be noted that fiocod and floSRC,i.e., distribution coefficients for hydrogen reaction during instantaneous coal dissolution and rate-controlled SRC reaction, both have negative values since hydrogen is consumed in these reactions. 2. H2S Balance. The molar flow rate of H2S in the reactor outlet is given by N H 2 S = gH,S,&fcodl/(rMWHg) (19) is mass fraction of coal converted to H2S;N H ~ % ~ % rate w of H2S out of the reactor, kmol/(m3-sf and M W H z s is the molecular weight of H2S. g H S , c is a fictitious number. It is used since reported yields were based on the feed coal mass. For a given coal, gHzS,c is almost a constant over the range of process conditions used in SRC-I1 liquefaction. For the Ireland Mine coal used in the Table I1 experiments, gH#,c = 0.032 (McIllvried, 1982). The mole fraction of H2Sin the reactor gas is given by YH2S = N H p S / N T (20) The partial pressures of Hzand H2S are given by P*H2
=
PYHz
(21)
PH2S
= PYHzS
(22)
(f) Product Yield. For any component i, the yield is given by Yi = f?"' f?'RCrsRCT/kfcoalP1) (23)
Procedure of Calculation For a given stirrer speed and hydrogen treat rate, eq 7-11 were used to obtain the magnitude of KLa. The gas
flow rate through the reactor was taken to be the same as that in the feed. p H P / p T H 1 = CH21/C*H31 = m was used as a constant whose magnitude was manipulated to obtain a balance between the rates of hydrogen mass transfer and reaction, Le., to satisfy eq 12. With an assumed value for m (=0.95 in the first iteration) and p*H = 0.8P, eq 1 and 13-22 were solved to calculate the vafue of p*Hzwhich satisfies hydrogen balance. In the case of a significant difference between the assumed and calculated values of P*~,,the average of the assumed and calculated value was the assumed value for the next iteration. The iterations were continued until the successive values of P * ~were , within 0.001%. Then a new value for m was calculated from the following expression obtained from eq 5 and 12: In the case of a significant difference between the assumed and calculated values of m, the average of the two was the assumed value of m for the next iteration. Depending on the magnitude of the mass-transfer coefficient, the number of iterations required for convergence of m within 0.001 5% varied from 5 to 16. Equation 23 was used to generate the yields of the product components.
Results and Discussion Validation of the Simulation. Comparisons of experimental yields and model yield profiles for distillates, SRC, hydrogen, and C1-CI gases as a function of stirrer speed at two levels of hydrogen treat rate are presented in Figures 1,2,3, and 4 respectively. As shown in Figure 1, the experimental distillate yields at each level of hydrogen treat rate are within f 2 wt % maf coal of the corresponding model values. The accuracy of the distillate yield measurement is considered to be f 3 w t 70maf coal. Therefore, the experimentaldata are considered to provide
Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987 1569
I
I
a
f
;
35-
s
k
/
30-
/
H, Treat Rate 25-
Data Points
~
-----
4 g of HJ100 g of slurry
1/1/
,-
2
,
,
Dal;
;ints,
~
4
6
8
1
Model
- - - -- - -
6 g of HJ100 g of slurry 4 g 01 H,/100 g of slurry
10
0
,
H, Treat Rate
Model
-.
6 g of H,l100 g of slurry
10
12
14
16
18
Stirrer Speed. rps
Figure 4. Comparison of experimental and model CI-C4 gas yields.
an acceptable validation of the model. The model shows that the rate of decline in distillate yield with decreasing stirrer speed decreases with reduction in stirrer speed below 3 rps. This is due to a sharp decline in the fraction of mixing power (E)contributed by the stirrer (Es). Since the contribution of feed gas to the mixing power (E,) is constant, a reduction in stirrer speed below 3 rps does not have a significant impact on the mixing power ( E ) and, hence, the mass-transfer coefficient (KLu). Thus, there is a lower limit to the magnitude of hydrogen mass-transfer coefficient (KLa)which corresponds to mixing by gas flow only, i.e., no mixing by stirrer ( N = 0). However, no experimental measurements were made at stirrer speeds lower than 2.5 rps; therefore, the region between 0 and 2.5 rps does not contribute to the validation of the model. Figure 2 shows the comparison of experimental and model predicted yields of SRC. All but the 2.5 rps data points are within f 3 wt % maf coal of the corresponding model values. Since the accuracy of the SRC yield measurement is about the same as that of the distillate yield (*3 wt % maf coal), the agreement in Figure 2 can be considered to provide an excellent validation of the model. It is also noted that all the measured yields at the hydrogen treat rate of 0.04 (4 g of Hz/ 100 g of slurry) are lower than the corresponding model values, whereas the model values are mostly lower at the hydrogen treat rate of 0.06 (6 g of Hz/lOO g of slurry). Based on the kinetic model, a higher yield for SRC should result in a lower distillate yield and vice versa. A comparison of the results presented in Figures 1 and 2 shows that there is no such correspondence
between the errors in distillate and SRC yields. It is likely that the appearance of bias toward lower or higher SRC yields in Figure 2 and distillate yields in Figure 1is either coincidental or due to additional reactions occurring in a hydrogen-deficient environment. Figure 3 shows that all the experimental data points are within f0.2 wt % of the model predicted values for a hydrogen treat rate of 0.06. Four data points from the set of runs for the hydrogen treat rate of 0.06 (nos. 1, 8, 10, and 11 in Table 11) were excluded from Figure 3. The exclusion of one 3.33 rps run (no. 1in Table 11)was based on the observed large difference between hydrogen consumption for this run and two other runs at identical conditions (nos. 2 and 3 in Table 11). The exclusion of the other three runs (nos. 8, 10, and 11in Table 11) from the data set in Figure 3 was based on their large deviation from the trend shown by the data points at other stirrer speeds and a large range for hydrogen consumption at an identical stirrer speed (16.67 rps) and an identical hydrogen treat rate (0.06). It is important to note that the hydrogen consumption values are based on measurement of flow rates and compositions of feed and outlet gases. The hydrogen balance is not influenced by the forced overall material balance. Therefore, the three runs (nos.9-11) excluded due to poor mass balance for distillate and SRC yields do not have to be excluded for hydrogen consumption values. However, only one of these (no. 9 in Table 11)appears to follow the trend suggested by the other runs and, hence, has been used in Figure 3. Exclusion of some of the undesirable data points is necessary for a meaningful use of the hydrogen consumption data. The exclusions are justified on the basis that only 20-30% feed hydrogen is consumed in the reactions. For example, at the high hydrogen treat level, the feed contains 20 g of hydrogen/ 100 g of coal. The maximum hydrogen consumption in Table I1 (5.19 wt 70 maf coal) represents less than 25% feed. If the flow and composition measurement errors summed up to about 5% of the amount of feed hydrogen, the error in the estimated hydrogen consumption would be 20%. Due to this experimental limitation, a f l wt % maf coal range for the hydrogen consumption data is generally acceptable. However, in this particular case, a f l wt % maf coal range would not allow a meaningful conclusion. Fortunately, a major part of the data points is in very good agreement with the model results for a hydrogen treat rate of 0.06. For a hydrogen treat rate of 0.04, the data points for stirrer speeds 3.33 and 2.5 rps show 0.6 and 1.9 wt 70maf coal differences, respectively, from the model values. The other three data points are within k0.4wt % maf coal of the corresponding model values. Since there is only one measurement at 2.5 rps which shows a very large difference between experimental and model values, no conclusion can be based on this data point alone. With the two other data points at 3.33 rps, this may constitute a trend. However, the errors for the two 3.33 rps runs are between 0.55 and 0.65 wt % maf coal, which is within the experimental error range. For the purpose of clarity, the comparison of experimental and model yields of C1-C4 gases in Figure 4 uses the average of the yields at the same conditions. It can easily be noted from the figure that hydrogen treat rate does not have any effect on C1-C4 gas yields. In fact, a single curve approximates all the average experimental yields presented in Figure 4. The experimental data clearly show that the yield of C1-C.I gases increases with a decrease in stirrer speed, whereas the model predicts an increase in its magnitude for each level of hydrogen treat rate.
1570 Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987
In case of insignificant hydrogen mass-transfer effect (stirrer speed, 16.67 rps) and a hydrogen treat rate of 0.04, the product distribution predictive capability of the model has been found to be very good over a very wide range of process conditions. For liquefaction of the same (Ireland) coal in the same experimental setup, a kinetic model with the same product distribution coefficients (Singh et al., 1982a) has been shown to correctly predict the increase in C1-C4 gas yields from 12.9 to 36.5 wt % maf coal for an increase in distillate yields from 24.9 to 62.2 wt % maf coal (King, 1984). In view of the applicability of the model over an extremely large range of C1-C4 gas yield (12.9-36.5 wt % maf coal), the observed increase in C144 gas yield with a decrease in stirrer speed (distillate yield) is exceptional. It should also be noted that the maximum difference between the predicted and measured yields of C1-CI gases over the entire experimental range is less than 5 wt % maf coal. At stirrer speeds lower than 6.67 ips, deposits of cokelike materials in the reactor indicate that some fractions of the reactor volume are not well mixed (Carr et al., 1982). It is possible that in such relatively stagnant sections of the volume, hydrogen deficiency leads to cracking or some other reactions which generate additional amounts of C1-C4 gases. The latter would add on to C1-C4 gases produced by the liquefaction reactions to give the exceptional distribution in Figure 4. Also, the measurement errors could account for a significant part of the difference. It is very important to note that the 3-5 wt 70maf coal higher C1-C4 gas yields at 2.5 and 3.33 rps stirrer speeds are also responsible for lower yields of distillates (Figure 1) as well as SRC (Figure 2) than would be predicted by the model. Overall, the model demonstrates a very good capability of predicting yields of distillate and SRC and consumption of hydrogen in liquefaction of coal by the SRC-I1 process. At low mixing levels, some uncertainties, resulting from violation of the basic assumption of wellmixed conditions in the reactor, influence the results. However, in most cases this has a relatively small effect on the yields of major components,i.e., distillates and SRC, and consumption of hydrogen. Most importantly, the model does not use any unknown or adjustable parameters in predicting the yields presented in Figures 1-3 which provide a very good validation of the model. The exceptional C144 gas yield behavior is not considered to indicate any weakness in the model since under the low levels of mixing power the existence of stagnant regions in the reactor violate the well-mixed assumption of the model. Hydrogen Treat Rate. An increase in the hydrogen treat rate increases the rate of SRC reaction which results in an increase in distillate yield (Figure I),decrease in SRC yield (Figure 21, and increase in hydrogen consumption (Figure 3). Under the reaction conditions in Table I, an increase in hydrogen treat rate from 0.04 to 0.06 is shown to result in a 5-9 wt % maf coal increase in distillate yield, a 6-11 wt % maf coal decrease in SRC yields, and a 0.5 to over 1.5 wt 70maf coal increase in H2 consumption. At 16.67 rps, the distillate yield and hydrogen consumption increase by about 22% and 42%, respectively, whereas the SRC yield decreases by about 33% due to an increase in hydrogen treat rate from 0.04 to 0.06. Percent changes in the magnitudes of these components are different due to the fact that the hydrogen treat rate influences the rate of SRC reaction only whereas a major part of each of the products is generated by the instantaneous dissolution of coal. To understand the relationship between the hydrogen treat rate and the rate of SRC reaction, we consider the
I
1
I
I
I
-
c
q
3
0.9
=
0.8
0.7
4
n*,r
I oL
1
;
€!
b Stirrer
Figure 5. a,,
1:
1;
1:
A
1:
j
Speed, rps
apH2,aCH2, and L Y H ~ Sas
functions of stirrer speed.
fractional changes in the magnitudes of the relevant variables which, for the purpose of illustration, are defined as CY,
=
(magnitude of x at hydrogen treat rate = 0.06)/ (magnitude of x at hydrogen treat rate = 0.04) (25) The subscript x of a for rSRC, p H P , CH21, and p H z S is represented by r , pHz, CHz, and H2S, respectively. As shown in Figure 5, the magnitudes of aPHB and aH,S are fairly constant at 1.03 and 0.68, respectively. Thus, there is very small (3%) contribution of the hydrogen partial pressure (apH,= 1.03) to the higher rate of SRC reaction (1.325 < CY, < 1.425) at the higher hydrogen treat rate. The 32.5-42.570 higher rate of SRC reaction which results in over 20% higher yield of distillate is almost entirely due to about 32% lower H2Spartial pressure (cYHB = 0.68) for the higher hydrogen treat rate. This large magnitude of the effect of the hydrogen treat rate on the rate of SRC reaction is surprising in view of the fact that the latter has not even been considered a significant variable in most of the coal liquefaction studies. Also, this study suggests that even larger increases in the rate of reaction and distillate yields can be achieved by using higher hydrogen treat rates. In Figure 5, the minima in the values of a,, aCH2, and aH,s result from the dependence of the rates of hydrogen mass transfer and reaction on the hydrogen treat rate. This is explained later. Mass-Transfer Effects. Figure 1shows that with the reduction in stirrer speed from 16.67 to 3.33 rps, the model predicted yields of distillate decline by 3.8 and 6.6 w t % maf coal at hydrogen treat rates of 0.04 and 0.06, respectively. These represent 10% and 14.5% of the respective distillate yields at 16.67 rps. Since a major part of the distillate (22 wt % maf coal) is generated by instantaneous dissolution of coal, it is more appropriate to consider the fractional changes in yields with respect to the amount produced by the SRC reaction. Such a fractional change in distillate yield would be the same as the change in the rate of SRC reaction. The above-referred decreases in distillate yields represent 23% and 28% of the distillates generated by the SRC reaction at 16.67 rps and hydrogen treat rates of 0.04 and 0.06, respectively. The corresponding declines in hydrogen consumptions are almost in the same proportions (Figure 3) since a very small amount of hydrogen is consumed in the instantaneous dissolution of coal. Figure 6 shows that at each level of hydrogen treat rate a reduction in stirrer speed from 16.67 to 3.33 rps results
Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987 1571 400
1.13
1.50
Mixing Power x lo', kW/m' 4.38 12.79 29.99
59.41
104.6
169.3
257.3
371.2
200
100 80 60
E
40
E
"--t
/ i
1 Rate g of H,/100 g of slurry
20
0.8
c
0.6
0
4
/
I
I
2
I
4
I
6
I
8
I
10
1
12
I
14
I
16
18
Stirrer Speed, rpr
Figure 8. Hydrogen concentration at gas-liquid interface (C*H,,) and in the bulk liquid (CH,~).
2
1.o 0.8 0.6;'
I
2
"
4
'
'
I
6
I
8
'
10
'
I
12
a
'
14
I
I
16
' 1
18
Stirrer Speed, rps
Figure 6. Mixing power as a function of stirrer speed.
0.4
0.1
1
c
//
/
I
O O
'
'
a :' '
'Ib
'1:
'1:
' 1 6
'1'8
1'
Stirrer Speed, rps
Figure 7. Mass-transfer coefficient (KLa)as a function of stirrer speed.
in a more than 100-fold decrease in the mixing power (E). The resulting decrease in the magnitudes of KLa (=0.11(E/ VR)0.s4)is also close to 2 orders of magnitude (Figure 7). However, the effect of hydrogen mass transfer is limited to lower stirrer speeds (Figures 1-3). For a hydrogen treat rate of 0.04, the mass-transfer effects are very small for stirrer speeds above 7 rps. As expected, for higher hydrogen treat rate, the mass-transfer effects appear to be weakly extended to higher stirrer speeds. Considering a stirrer speed of 7-3.33 rps, it is noted that a more than 4-fold reduction in KLa (Figure 7) results in a less than 25% decline in the rates of SRC and hydrogen reactions (Figure 3). The relatively weak influence of hydrogen mass
transfer on the rate of SRC (or H,)reaction is more clearly illustrated in Figure 8, which shows that the minimum concentration of hydrogen in the slurry (CHP1= 0.775 kg/m3) is about 50% of the maximum possible hydrogen concentration (C*H21 = 1.66 kg/m3). Thus, the weak influence of hydrogen mass transfer on the rate of SRC reaction is basically due to the high solubility of hydrogen in the coal slurry. The concentration of hydrogen at the liquid-gas interface, Le., C*HZ1, in Figure 8 is 40% of the gas-phase concentration, Le., H = 0.4 in eq 2. Since the mimimum level of mixing power used in an SRC reactor (U = 0.01 m/s) corresponds to about 6 rps in Figure 8, hyarogen mass transfer is unlikely to limit the rate of reaction in an SRC-I1 coal liquefaction reactor. Due to higher rate of SRC (or hydrogen) reaction, the hydrogen mass-transfer effects are more significant at the higher hydrogen treat rate (Figures 1-4). However, the minima for a,and a C H 2 in Figure 5 show a relatively complex influence of hydrogen treat rate on the actual rate of reaction in the region of hydrogen mass-transfer influence, i.e., stirrer speed below about 7 rps. Minima for aCH,and a, (Figure 5 ) . An increase in hydrogen treat rate increases the rate of reaction by lowering the partial pressure of H2S (Figure 5). The hydrogen treat rate also has two opposite effects on the mixing power ( E ) and, hence, the hydrogen mass-transfer coefficient (KLu). An increase in hydrogen treat rate increases the feed gas contribution, Le., E,, and reduces the stirrer contribution (E,) to the total mixing power (E). Therefore, at lower stirrer speeds when Eg accounts for the major part of E , a higher hydrogen treat rate gives higher mixing power. At higher stirrer speeds when E , is only a small fraction of E , the lower hydrogen treat rate yields higher total mixing power (E). As shown in Figure 6, the mixing power at a hydrogen treat rate of 0.06 is higher below a stirrer speed of 4.3 rps and lower for higher stirrer speeds in comparison to the respective values for a hydrogen treat rate of 0.04. Since KLa is proportional to ( E / V R ) O ,the ~~, relationship between mass-transfer coefficients at the two levels of hydrogen treat rate (Figure 7) is identical with the mixing energy relationships (Figure 6). As shown in Figure 9, only for stirrer speeds lower than 4.3 Ips, the hydrogen mass transfer coefficient is higher for the higher hydrogen treat rate (0.06) whereas the intrinsic rate of SRC reaction is always higher for the latter. As noted earlier the hydrogen mass-transfer effect is significant only at low stirrer speeds or mixing power. Also, the significance of the hydrogen mass-transfer effect de-
1572 Ind. Eng. Chem. Res., Vol. 26, No. 8, 1987
i
i i i
-
10UKL.
1 09-
-
--t_-
-L 2
4
1
1
1
6
8
10
4 a E 12
14
16
1 18
4 1
Stirrer Speed rps
Figure 9. aE and aKloas functions of stirrer speed.
creases with an increase in the stirrer speed. With no stirring ( N = 0), the mass-transfer effect has maximum influence on the rate of reaction. In relative terms, the hydrogen mass-transfer effect is greater for higher hydrogen treat rate due to the higher rate of reaction but lower due to the higher mass-transfer coefficient (aKLa = 1.42; Figure 9). The net effect is higher hydrogen concentration in slurry (acH, = 1.1; Figure 5). Also, the higher hydrogen treat rate has a positive effect on the intrinsic rate of reaction due to the lower H2S partial pressure (aHzs = 0.68; Figure 5). The higher intrinsic rate of SRC reaction and higher hydrogen concentration in slurry result in a more than 50% higher rate of reaction (a, > 1.5; Figure 5) at the higher hydrogen treat rate. An increase in stirrer speed results in a fast decline in the relative magnitude of the hydrogen mass-transfer coefficient (aKa ; Figure 9). Due to lower rate of hydrogen consumption anh higher rate of increase in mass-transfer coefficient with an increase in stirrer speed, the influence of hydrogen mass transfer on the rate of reaction declines faster at the lower hydrogen treat rate. As a result, a, as well as a C H 2 declines with an increase in stirrer speed. However, the rates of decline in the magnitudes of a, and aCH, reduce with an increase in stirrer speed. This is due to a faster reduction in the significance of hydrogen mass-transfer effect at lower hydrogen treat rate (0.04) compared to that a t the higher hydrogen treat rate (0.06). The minima in the magnitudes of CYCH,and a, result from the fast disappearance of the hydrogen mass-transfer effect at the lower hydrogen treat rate. Due to the lower rate of increase in the hydrogen mass-transfer coefficient and higher rate of intrinsic reaction, the mass-transfer effects at higher hydrogen treat level persist at higher stirrer speeds. This results in a continuous increase in aCH2 and a,, from their minima, with an increase in stirrer speed. Though necessary for understanding the CSTR results, the above-discussed complex relationship is not applicable to a bubble column reactor in which the mixing power is supplied by the flow of gas only; i.e., E = E,. Since any large size reactor would be a bubble column, the significance of minima for a C H , and a, is rather limited.
Conclusion In SRC-I1 coal liquefaction, an increase in the hydrogen treat rate reduces the partial pressure of H a in the reactor. Since the latter has a strong inhibitive effect on the rate of liquefaction, an increase in hydrogen treat rate results in a significant increase in the rate of liquefaction. At very low levels of mixing power, the rate of liquefaction in an
SRC-I1 reactor can be limited by the rate of hydrogen mass transfer. However, in the range of mixing power available in normal operation of the SRC-I1 reactors, the effect of hydrogen mass transfer on the rate of reaction is small. High solubility of hydrogen in coal liquids is the major reason for the small effect of hydrogen mass transfer on the rate of liquefaction of coal by the SRC-I1 process.
Nomenclature ASH = inorganic mineral matter C*H,, CHPl= concentrationsof hydrogen at gas-liquid interface and in the bulk liquid, kg/m3 d,, = impeller (stirrer) diameter, m E,, EsO,E,, E = mixing power input to reactor by gas only, by stirrer in absence of gas in liquid, by stirrer in presence of gas in liquid, and total, respectively, kW fICod, fiSRC = distribution coefficients for component i in the instantaneous dissolution and SRC reaction products, respectively. G = gravitational constant, m/sz g, = mass fraction of i gH,f = hydrogen treat rate, g of Hz/lOO g of slurry H = Henry’s law constant, dimensionless HD = heavy distillate (288-482 “C) IOM = pyridene-insoluble organic matter KLa = volumetric mass-transfer coefficient, s-l LD = light distillate (C5-193 “C) MD = middle distillate (193-288 “C) MW, = molecular weight of component i maf = moisture ash free N = stirrer speed, rps Np = power number N , = molar flow rate of i, kmol/(m3.s) P = pressure, MPa p , = partial pressure of i, MPa Q, = volumetric gas flow rate, m3/s RH2= rate of hydrogen mass transfer, kg/(m3.s) r, = rate of reaction of i, kg/(m3.s) R = universal gas law constant, kJ/(kmol.K) SkC = solvent refined coal, pyridene-soluble organic matter boiling above 482 “C T = reactor temperature, K Y , = yield of component i, wt % maf coal y, = mole fraction of i Greek Symbols = ratio of magnitudes of i at hydrogen treat rates of 6 and 4 g of Hz/lOO g of slurry t, = gas holdup pl = density of slurry, kg/m3 T = slurry residence time, s CY,
Subscripts f = feed H2 = hydrogen
HZS = HZS I = iron in slurry 1 = slurry SRC = solvent-refined coal T = total Registry No. H2S, 7783-06-4.
Literature Cited Carr, N. L.; King, W. E., Jr.; Moon, W. G. “Prepilot SRC-I1 Development Project: Hydrogen Mass Transfer Study”, Technical Report DOE-82010771, Feb 1982; U.S. Department of Energy, Washington, DC. Epperly, W. R. “EDS Coal Liquefaction Process Development. Phase IV Quarterly Technical Progress Report for July 1-September 30,1979”, Technical Report DOE/FE-2893-41,1980; US. Department of Energy Washington, DC. King, W. E., Jr. Fuel 1984, 63, 600. King, W. E., Jr.; Carr, N. L.; Moon, W. G. Fuel 1984, 63, 1143. McIllvried, H. G. “Summary of Results from SRC-I1 Runs Made on Process Development Unit P-99, Runs P99-15 to P99-85”, Tech-
Ind. Eng. Chem. Res. 1987,26, 1573-1578 nical Report DOE/ET/10104-42, Jan 1982; U.S. Department of Energy, Washington, DC. McIllvried, H. G.; Gall, W. ”Experience with Solids Deposit in SRC-I1 Process Development Unit P-99: Runs P99-77 to P99-85”, Technical Report DOE/ET/l0104-11, March 1981a; U S . Department of Energy, Washington, DC. McIllvried, H. G.; Gall, W. “Experience with Solids Deposit in SRC-I1 Process Development Unit P 9 9 Runs P99-11 to P99-76”, Technical Report DOE/ET/10104-32, Nov 1981b; U S . Department of Energy, Washington, DC. Painter, P. C.; Yamada, Y.; Jenkins, R. G.; Coleman, M. M.; Walker, P. L., Jr. Fuel 1979,58, 293. Shibaoka, M.; Ueda, S. Fuel 1978a,57,667. Shibaoka, M.; Ueda, S. Fuel 1978b,57,673.
1573
Singh, C. P. P.; Shah, Y. T.; Carr, N. L.; Prudich, M. E. Can. J . Chem. Eng. 1982a,60, 248. Singh, C. P. P.; Shah, Y. T.; Carr, N. L. Can. J . Chem. Eng. 198213, 60,261. Singh, C. P. P.; Carr, N. L. Znd. Eng. Chem. Process Des. Dev. 1983, 22, 104. Singh, C. P. P.; Carr, N. L. Chem. Eng. Commun. 1986, 42, 61. Singh, C. P. P.; Carr, N. L. Znd. Eng. Chem. Res. 1987,26, 501. Warmoeskerken, M. M. C. G.; Smith, J. M. Chem. Eng. Sci. 1985, 40, 2063. Received for review April 21, 1986 Revised manuscript received May 2, 1987 Accepted May 24, 1987
A Kinetic Study of Carbon Oxidation in an Alkali Carbonate Melt John H.Cameron Recovery Group, Chemical Sciences Division, The Institute of P a p e r Chemistry, Appleton, Wisconsin 54912
Experimental results demonstrate that oxidation of the carbon content of kraft char in a carbonate melt containing sulfur occurs through a sulfate-sulfide cycle. A kinetic study of air oxidation of kraft black liquor char in alkali carbonate-sulfate melts is described. Air and sulfate oxidation rates were measured for kraft and soda chars. The effects of the experimental parameters, including carbon concentration, sulfate concentration, and temperature, were determined, and suitable rate equations were developed. One of the principal steps in the kraft pulping process is the burning of the residue organic material from the pulp digester and the regeneration of the pulping chemicals. In the kraft recovery furnace, black liquor from the pulping process is burned and the sulfur compounds present are converted to sulfide. The firing of black liquor in a recovery boiler is somewhat more complex than the firing of a fossil fuel. Overall objectives are different, since recovery of the inorganic salts in a form suitable for regeneration of pulping chemicals is required as well as combustion of the organic matter for steam generation. The conversion of the inorganic material to the desired sodium carbonate and sodium sulfide takes place in the char bed and occurs simultaneously with the combustion of the carbonaceous char. Optimization of furnace performance requires a better understanding of these processes occurring in the char bed. One of the key reactions occurring in the bed of the kraft recovery furnace is the oxidation of the carbon content of the kraft char (formed from the drying and pyrolysis of black liquor). In a kraft recovery f m a c e , carbon oxidation occurs in a melt consisting principally of sodium carbonate, sodium sulfate, sodium sulfide, and carbon. Although carbon oxidation in this environment is industrially’ important, little information is available concerning its mechanism or controlling parameters. The lack of information for this reaction is due to the difficulty of studying a reaction involving three phases (solids, liquid, and gas), the high temperatures involved, and the corrosive nature of the reaction system. This paper describes oxidation of kraft black liquor char in an environment similar to that present in a kraft recovery furnace. The effects of the reaction variables are described, a rate expression is developed, and a mechanism is proposed.
Previous Research Carbon Oxidation in a Carbonate Melt. To support the molten salt coal gasification process (Trilling, 1974)
Atomics International has conducted a considerable amount of research on coal oxidation in molten sodium carbonate. This coal gasification process consists of the partial oxidation of coal in molten sodium carbonate. With this process, a low Btu gas is produced and any sulfur present in the coal is converted to sodium sulfate. A number of papers describing reactions in this process have been published. The papers of Stellman et al. (1976),Dunks et al. (1980), Dunks et al. (1981), and Dunks et al. (1982) describing carbon oxidation in carbonate melt were published over a &year period (1976-1981). The last paper published, Dunks et al. (1982), is the most comprehensive of this series. This paper described carbon oxidation by oxygen with and without sulfur present and sulfate reduction by carbon. Using two different size reactors (one containing 5500 g of melt and one containing 125 g), Dunk et al. (1982) studied carbon oxidation with either sulfate or oxygen in sodium carbonate-sodium sulfate melts. Spectrographic grade powdered graphite seived to produce a particle 0.15-0.18 mm in diameter was used as the carbon source, and the reactions were studied from 890 to 1038 “C. Depending on reactor size, the activation energy for air oxidation of carbon in a carbonate-sulfate melt varied from 209 to 318 kJ/mol. For air oxidation of carbon in pure carbonate, the activation energy varied from 117 to 134 kJ/mol. This dependence of activation energy on reactor size indicates that unifoim mixing may not have been achieved in the two reactors. For carbonate-sulfate melts, the rate dependence on oxygen was approximately zero order (0.1-0.2). Depending on sulfate and carbon level, the carbon surface area dependence varied from 0.45 to 1.0 order. This variation with carbon and sulfate levels may indicate a shift in the controlling mechanism of this reaction. The earlier paper by Stellman et al. (1976) showed that the oxidation of graphite was much faster in a sodium carbonate-sodium sulfate melt than in a pure sodium
0888-5885/87/2626-1573$01.50/0 0 1987 American Chemical Society