Ind. Eng. Chem. Process Des. Dev. $902, 21. 404-409
404
Carbon Deposition Studies Using Nickel and Cobalt Catatysts Michael P. Mannlng,' James
E. Garmlrlan, and Robert C. Reid
Department of C%emical Engineerlng, Massachusetts InstifUte of Technolcgy, Cambridge, MassaChuS8tts 02 139
I n the NASA-developed Bosch process, GO2 is reduced with hydrogen to form carbon and water. Cb and methane are also formed. Carbon accumulates in the reactor arid water is removed in the recycle gas stream. Normally an iron catalyst is used. I n this study, cobalt and nickel catalysts were studied to overcome the limitation that iron readily oxidizes to a noncatalytlc form and this limits the extent of reaction. I t was shown that cobalt and nickel catalysts allow much hlgher exit water concentrations, but still it is not possible to attain the equitibrium predicted from assuming the product carbon is @-graphite. I t is postulated that carbon deposits as a metastable metal-carbide intermediate and, therefore, water yields are limited by the metal-carbide equilibrium. The effect of this carbide route to carbon formation is very pronounced at 700 K and below, but it is less important at higher temperatures.
The Bosch process is being considered by NASA as a means of recovering oxygen from metabolic carbon dioxide. The ultimate objective is to employ this process in a life support system capable of supplying oxygen in a cyclic process. The cycle begins by reducing carbon dioxide with hydrogen to form solid carbon. COZ 2H2 C 2Hz0
+
-+
+
An iron catalyst is normally used. The water formation reaction is followed by electrolysis to yield hydrogen and oxygen. 2H20 2Hz + 0 2 -+
The overall process results in the reduction of metabolic carbon dioxide to carbon which accumulates on the iron catalyst, and oxygen which is returned to the spacecraft. In the past, NASA scientists have run Bosch reactors using a steel wool catalyst and recycling unreacted and byproduct gases to obtain complete reaction (Holmes et al., 1970). One problem which they encountered was that the maximum water concentration obtained in the reactor was significantly less than the thermodynamically predicted equilibrium water concentration. Previous investigations (Manning and Reid, 1977; Sacco and Reid, 1979) were able to show that the limited water yield resulted from oxide inhibition rather than slow reaction rates. That is, in addition to the reactions which occurred between carbon monoxide, carbon dioxide, methane, hydrogen, water, and carbon, oxidation of the iron catalyst also occurred. Above 860 K the reaction Fe + HzO e FeO + H2 occurred, and below 860 K 3Fe + 4H20 e Fe304+ 4H2 These reactions, coupled with the discovery that iron oxide does not catalyze carbon deposition reactions, led to an explanation for the low water concentrations present in the NASA reactors. Experimental data on the Bosch system were in good agreement with iron-iron oxide-gas equilibrium calculations based on tabulated Gibbs energy data. Preliminary theoretical calculations have revealed that water production limitations associated with the ironcatalyzed Bosch process may not be encountered when the iron is replaced by nickel and cobalt. The regions of simultaneous metal oxide formation and solid carbon formation which are thermodynamically expected in the iron system are thermodynamicallyimpossible in the nickel and 0196-4305/82/ 1121-0404$01.25/0
cobalt systems. This suggested that a study of the nickel and cobalt catalyzed Bosch process may lead to a more efficient reactor. Experimental Section
The equipment and operating procedures used in this study have been described in previous publications (Manning and Reid, 1977; Sacco and Reid, 1979). Additional details are available in a thesis (Garmirian, 1980). In brief, the apparatus is composed of a feed-gas delivery system, a thermogravimetric reactor, and a gas-sampling system. In the feed-gas delivery portion, chemically pure gases were individually metered and mixed to synthesize desired gas compositions. In this work, water was added to the feed gas by metering liquid water and vaporizing to provide the requisite rate of steam flow. The reactor was composed of a vertical quartz tube within an isothermal furnace. The catalyst was held in a special carrier between plugs of glass wool. Weight measurements could be taken periodically by suspending the assembly from an analytical balance. Weight changes within f l mg could be detected. Three catalysts were employed in the study. The first was a Ni/A1203catalyst powder supplied by the Alfa division of Ventron Corp. It had a BET surface area of 200 m2/g. It was reported to contain 60-65 wt 5% Ni. A second catalyst, cobalt powder, was supplied by Bram Metallurgical Co. The powder was formed from 1-5-pm spheres fused to form approximately 40-pm aggregates. The purity was reported to be 99.9% and had a BET surface area of 0.7 m2/g. A third catalyst used was prepared by drying a 5 wt ?% solution of cobalt nitrate on quartz wool at 423 K. This was then calcined in air and reduced in hydrogen at 700 K to form metallic cobalt. A gas sampling system, which consisted of a vacuum system, heated gas-sample valve, and gas chromatograph, allowed the determination of inlet and outlet compositions. Approach
As noted earlier, in the Bosch process of interest to NASA, the ultimate goal is to reduce carbon dioxide with hydrogen to form water and solid carbon. In the process carbon monoxide and methane are also formed. For a given temperature, pressure, and gas phase O/H ratio, if any three independent reactions are chosen, such as H2 + C02 s CO + H2 (A) CHI H2O + 3H2 + CO (B) 2co s c + c02 (C)
+
0 1982 American Chemical Society
Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 3, 1982
405
Table I. Effect of Carbon Type on the Equilibrium Gas Compositions on the Bosch Processa equilib solid p-graphite Dent etal. p-graphite Dentetal. 0-graphite Dent etal. a
AG,,
partial pressure, bar T, K P H , PH,O
kcal/
Pco Pco- PCH. mol
900 0.41 0.18 0.18 0.18 0.05 0.0
900 800 800 700 700
0.40 0.28 0.23 0.14 0.08
0.13 0.32 0.17 0.45 0.11
0.22 0.05 0.09 0.01 0.03
0.17 0.23 0.29 0.23 0.40
0.08 0.11 0.22 0.17 0.38
0.85 0.0 1.69 0.0 2.86
O/H= 0.5;pressure = 1.0bar. -20
then the equilibrium mole fractions of each component can be calculated. Typically, a solid carbon phase of pure @-graphiteis assumed in making the calculations. However, results of Dent et al. (1945) and Rostrup-Nieben (1972) indicate that such a simplification can lead to erroneous results. The equilibrium constant for reaction C was measured by both Dent et al. and Rostrup-Nielsen, while the reaction CH4
+ 2H2 + C
(D)
was studied by Rostrup-Nielsen. They found that only at temperatures of about 1000 K would the assumption of a solid phase of @-graphitebe justified. A t lower temperatures the Gibbs energy change for the carbon forming reactions is larger than that for the corresponding reaction assuming a solid phase of graphite. This difference was about 1.5 and 3.0 kcal/mol at 800 and 700 K, respectively. In Table I we show the equilibrium gas composition at various temperatures, 1 bar, and an O/H ratio of 0.5 assuming a solid phase of either graphite or one which deviates from graphite by the amount reported by Dent et al. The O/H ratio employed here is the atomic ratio of oxygen to hydrogen in the gas phase. As can be seen, the results are quite different depending upon the equilibrium solid carbon phase assumed in the calculation. The water concentration changes from 18 to 13%,32to 17%,and 45 to 11% ,at temperatures of 900,800, and 700 K, for a solid phase of @-graphitevs. “Dent” carbon. These calculations demonstrate the effect that the Gibbs energy differences can have on equilibrium water concentrations obtainable in a Bosch process and emphasizes the importance of experimentally determining the carbon-gas equilibrium. In Table I there is a column entitled AG,. This parameter was obtained from experimental data by Dent et al. where in nickel catalysts were employed in a methane synthesis reactor containing CO and C02. For reaction C, the thermodynamic equilibrium constant is Kp = Pco,/Pco2
(1)
AG, = -RT In [K,/K,(@-graphite)]
(2)
We define
where K,(@-graphite) is the equilibrium constant for reaction C calculated from Gibbs energy data for CO and COz and assuming the carbon phase was @-graphite. Clearly, if the solid carbon phase in the experimental study were &graphite, AG, would be zero. AG, values may also be obtained from experimental gas compositions of CH4 and Hzand considering reaction D in a similar manner. In the experimental program, the parameter AG, was varied in the following manner. For studies with either a nickel or cobalt catalyst, the temperature and the atomic ratio of O/H in the feed-gas were selected. With a pressure of 1 bar, one could then calculate the gas compositions of
-10 00 10 20 GIB B S E N E R G Y DlFF ERENC E AGc ( k c a l / m o l e )
Figure 1. Rate of carbon deposition as a function of AGc for an O/H ratio of 0.5.
H2, H20, CO, COP,and CHI in equilibrium with a solid @-graphitephase. By assuming that AG, was nonzero, i.e., that the solid phase had a different activity or Gibbs energy from @-graphite,different equilibrium gas compositions corresponding to different values of AG, could be computed (as was done in Table I). With AG, values ranging from about -5 to +5 kcal/mol, gas feed mixtures corresponding to the calculated equilibrium compositions were prepared and passed over the catalyst. Gravimetric observations which indicated whether carbon was deposited or removed were used to determine the equilibrium point experimentally. Thus each feed-gas mixture was characterized by its temperature, O/H ratio, pressure (1 bar), and the AGc value. The AG, range chosen extended beyond the values reported by Dent et al. (1945), Rostrup-Nielsen (1972), and Schenck (1927) to ensure that the gas compositions employed would enclose the equilibrium point. Results hxperiments were run at temperatures of 700,750,800, and 900 K and a pressure of 1 bar. Measurements were made for the binary gas systems of CO-C02 and CHI-H2 over carbon and catalyst. Multicomponent experiments with H2, CO, CHI, C02,H20, and C were also carried out at gas phase O/H ratios of 0.17, 0.5, and 1.5. Preceding every experimental run where fresh catalyst was put into the reactor and reduced, carbon was deposited from a 1:l mixture of hydrogen and carbon monoxide. This was done at the reaction temperature of the experiment. In each case about 0.2 g of C was deposited. During any experiment, the rate of carbon deposition (or removal) was determined by intermittent weighing of the catalyst. In cases where rates were high (>1 mg/min), four measurements were taken, one every 10 min. For slower rates, the time between weighings was increased. By plotting the catalyst weight vs. time, the slope (which was essentially constant) yields the rate of carbon deposited (or removed). Typical data for a Ni/Al2O3catalyst are shown in Figures 1 through 3 for an O/H ratio of 0.5 and at temperatures of 900, 800, and 700 K, respectively. In these figures, the rate of carbon deposition (or removal) is plotted as discrete points, as a function of AG,. Also shown are the equilibrium gas phase concentrations as a smooth function of AG,. The gas compositions shown at a specific value of AG, was used experimentally to obtain the rate of carbon deposition at that AG,. For gas compositions corresponding to a large, positive AG,, deposition is rapid at all temperatures. The absolute rate of deposition is not significant as this value would depend on the area of the catalyst, the specific activity,
406
Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 3, 1982
.E
-
15
. P E E
800K, lobar
la
10
10
08
08
0 6
06 r
z 0
E
v1
x z x
01
C a
00 m
u
I
-0 5
i -
00
Y
2-0 2 00 20 40 60 GIBES ENERGY D I F F E R E N C E AGc (kcol /mole1
-20
00 10 20 30 GIEES ENERGY D I F F E R E N C E AG, ( k c a l / m o l e l
-10
Figure 2. Rate of carbon deposition as a function of AG, for an O/H ratio of 0.5.
2
02 m
02
U
k" L
n 2 V
04
w 0 4 0
LL
2
Figure 5. Rate of carbon deposition as a function of AG, for an O/H ratio of 0.5.
I.
Symbol
1
O/ H
7 0 0 K , I O bar Ni/AI2Oj
10 50091
: -
0 8
t
05 W
z 04
m
iT
5
02
iL
w 00 U
-20 00 20 40 60 GIBES ENERGY DIFFERENCE A Gc ( k c o l / m o l e 1
Figure 3. Rate of carbon deposition as a function of AGc for an O/H ratio of 0.5.
700
9W
BOO
TEMPERATURE ( K )
Figure 6. Carbon deposition and removal boundary for various O/Hratios over nickel. Symbolswith diagonal slash indicate carbon removal boundary; others indicate carbon deposition boundary.
: i 2 -0 I
-20
-10 00 I O 20 GlBBS ENERGY DIFFERENCE AGc ( k c o l / m o l e )
Figure 4. Rate of carbon deposition as a function of AG, for an O/H ratio of 0.5.
etc. As the gas composition is changed to yield lower AG, values, the rate decreases and would be expected to be zero at a AG, value of 0 if the carbon phase were @-graphite. What was found, however, was that the rate became zero at a positive value of AG,, and this value of AGc increased with decreasing temperature. Further, we note that carbon removal was only affected at negative values of AG,, an obvious observation if one assumed that the carbon phase was @-graphite. (At low temperatures, approximately 700 K, the rate of carbon removal was extremely slow.) Data for cobalt catalysts were similar to those obtained using nickel. In Figures 4 and 5 we show plots at 800 and 700 K again for a constant O/H ratio of 0.5.
1
-
"
L 7 00
-
0
_--_ 800
TEMPERATURE
900
(K)
Figure 7. Carbbn deposition and removal boundary for various O/Ef ratios over cobalt.
The data indicating the carbon depositiaa boundary for botli nickel and cobalt catalysts are summarized in Figures 6 and 7. The data for an O/Hratio of zero were obthined from experiments using only H2 and CHI as feed gases. Similarly, for an O/H ratio of infinity, the gas phase contained only CO and COz. At any given temperature, the difference in the value of the AG, between carbon deposition and removal can be approximated. Note that this difference increases with decreasing temperature. For
Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 3, 1982 407
cobalt at 900 K, no difference exists; that is, the rate of carbon deposition becomes zero when AG, is zero. Also, as a further example, from Figure 2 (nickel, 800 K, O/H = 0.5),carbon deposition ceases at a AG, approximately 1.5 kcal/mol and carbon removal begins when AG, is 0. These two limits are shown in Figure 6.
Discussion From the experimental data developed in this work, as well as those from earlier investigations, it is seen that, except a t high temperatures, carbon deposition requires a more reducing gas (higher concentrations of CO, H2,CH4) than if one simply assumed /3-graphite were formed. In prior work with the same system, but using iron catalysts, Manning and Reid (1977) and Sacco and Reid (1979) also came to the same conclusion. In this case, however, the cause was the formation of iron oxide which was shown to be noncatalytic for carbon deposition. The fact that the @-graphiteequilibrium cannot be attained would also mean that less water would be present in any reaction gas mixture. For example, in Figure 3, the mole fraction water expected if /3-graphite were present is about 42% (700 K). However, since carbon deposition ceases when AG, = 3 kcal/mol, the maximum water content is only about 10%. The apparent thermodynamic anomaly for carbon at low temperatures may, in fact, be due to the formation of a metastable carbide intermediate. That is, one may have a reversible reaction to form carbide from the metal and methane such as CHI + nMe + Me,C + 2H2 (E) where Me can be either nickel (where n = 3) or cobalt (where n = 2 or 3). The formation of carbide can also occur from other reactions, such as H2 + CO + nMe s Me,C + H 2 0 and 2CO + nMe
F!
Me$
+ COz
-
The carbide can also decompose to free metal and carbon Me,C C + nMe It is suggested that the formation of carbon occurs by this reaction series, Le., via an intermediate carbide. The net result, using methane-hydrogen-carbon reaction as one example, is CH4 C + 2Hz
-
In our experimental results, the observed catalyst weight increase is much greater than could be accounted for by the formation of a stable carbide. Thus, if a carbide formed, it must have continuously decomposed to account for the net mass increases observed in the experiments. Carbides have, in fact, been shown to exhibit a metastable behavior, with increasing instability at higher temperatures. Hofer et al. (1949,1950) have shown that both nickel and cobalt carbides decompose readily at temperatures of 630 K; the time required was less than 1 h. Nagakura (1957, 1961) reported that in a temperatureprogrammed study, decomposition of nickel carbide did not occur until 704 K. He also indicated that the maximum net rate of carbide formation occurred at 623 K. For the cobalt system, the decomposition of CozCoccurred at 723 K and at 750 K for CO~C.The maximum net rate of formation of both carbides occurred at temperatures between 723 and 773 K. It should also be noted that Renshaw et al. (1971) observed Ni& at temperatures as high as 823 K. In Nagakura's study, only free metal and carbon
7
.
Ax 0 Dent R o s t rel u p01 - N t(19451 e lrs e n 119721A
V
0
-2 6.0 d a
g
Browning ond Emmett
11952)
v
$;
0 This s t u d y
I
\
I
101
0
500
,
l x
8
lx\y
600 700 800 900 T E M P E R A T U R E (K)
1000
Figure 8. Gibbs energy difference for the CH4-H2-C or CH4-H2Ni-NisC equilibrium. Open symbols indicate a measurement approached from the hydrogen rich side of equilibrium;closed symbole indicate approach from the methane rich side.
were observed at temperatures above 773 K for Ni and 823 K for Co. Thus, the available experimental evidence indicates that the transient existence of both nickel and cobalt carbides at temperatures between 700 and 800 K is highly likely. X-ray diffraction experiments did not reveal the presence of a carbide phase in our samples. This is reasonable on two bases. First, the time required to cool down and remove a solid sample from the reactor is much longer than the time required to decompose the carbide. Second, as the carbide is an unstable intermediate at these temperatures, the expected steady state concentration of carbide in the metal is very low. Browning and Emmett (1952) studied the reaction 3Ni + CHI F! Ni3C + 2H2 between 500 and 600 K at subatmospheric pressure. At these lower temperatures, X-ray analysis confirmed that the solid phase contained nickel carbide. From the reported equilibrium constants (K, = pH2'/pCH4) values of AG, could be determined. Note that AG, would then be equivalent to the Gibbs energy change for the reaction 3Ni + C ~t Ni3C where C represents @-graphite.The AG, values computed from the data of Browning and Emmett are shown in Figure 8. Also shown are the high-temperature AG, values determined from the experiments of Dent et al. (1945) and Rostrup-Nielsen (1972). The experimental results from the present study, using a nickel catalyst, are also shown in Figure 8. There appears to be a definite trend of the AG, values with temperature and this suggests that the Dent et al. and Rostrup-Nielsen experiments may also have involved nickel carbide as the solid phase. From the proximity of the extension of our data with the reported nickel carbide equilibrium data, we suggest that the carbon deposition boundary corresponded not with the graphite-gas equilibrium, but with a different solid phaseprobably carbide. The cobalt studies result in similar conclusions. However, as noted earlier, there was normally a range of AG, values where deposition had ceased and carbon removal was not observed. This indicates that either the method of weight detection was not sufficiently sensitive, or below the point at which carbon deposition ceased the catalyst was converted to a less active form. In almost all cases carbon removal occurred only if AG, of the gas mixture was less than 0 kcal/mol. This would also be expected for a system where graphite was present. In
408
Ind. Eng. Chem. Process Des. Dev., Vol. 21, No. 3, 1982 I \
I OIH.05
Graphite
0O
W
700
800
1 900
TEMPERATURE ( K )
Hydrogen
Oxyger
H 20
-
Upper C u r v e Nickel C a r b i d e Gas E q b i i l b r i u m Lower Curve G r o p h i t e G a s E q u i l i b r i u m
-
Figure 9. Carbon deposition boundaries for the Bosch process over nickel.
Figure 9 the position of the graphite-gas and nickel-nickel carbidegas equilibrium lines are shown at 700 and 800 K. The diagram for cobalt is similar. As temperature is increased the graphite and carbide lines are seen to move closer together. Consider the equilibrium diagram shown in Figure 9 and the two equilibrium boundaries at 700 K. The carbide phase line indicates the points at which the gas phase is in equilibrium with both metal and metal carbide. Above the line, for example at point X, the free metal will react to form carbide; below the line, at point Y, the carbide will react to form free metal. The graphite equilibrium is different since there is only one solid phase which is considered along with the gas phase. Gas mixtures, which are represented by a point such as X or Y above the line, should deposit carbon, while ones which are below the line such as Z should remove carbon. In the experiments studied in this work, carbon deposition was observed only in the region of the phase diagram above the metal-metal carbide equilibrium curve and, therefore, above the graphite-gas equilibrium curve. Since the upper boundary is in fact a carbide phase line, in the region above, metal carbide is present. For most of the experiments, there existed a region of no weight change. Although carbon removal was not always observed below the graphite phase line (AG, = 0.0 kcal/mol), very rarely was it observed above. In the region below the carbide phase line, no carbide would be expected to be present. If carbon formation only occurred through a carbide intermediate, no carbon deposition would be expected below the carbide equilibrium curve. The reaction Me,C
+ 2H2
-
nMe
+ CHI
is not observed in that region either because the carbide decomposes to carbon quickly, or the amount present is a relatively small fraction of the total mass of metal plus carbon. Carbon removal was in fact observed under some conditions. This must have occurred via the reaction C + 2Hz CHI -+
which will only occur for gas mixtures where AGc is less than 0 kcal/mol. For the most part, Figures 6 and 7 support this thermodynamic argument. Alternatively, one might invoke the dissolution-precipitation types of mechanisms proposed by Trimm (1979) and RostrupNielsen and Trimm (1977) to explain the carbon deposition phenomena. In these mechanisms, the carbon, which appears as whiskers or filaments, is formed in a series of steps involving surface reaction, carbon diffusion through a catalytic crystallite, and carbon precipitation at the filamentlcrystallite interface. As indicated in Figure 3 of Rostrup-Nielsen and Trimm (1977), the carbon activity
Figure 10. Equilibrium water concentration of a Bosch reactor for graphite, iron, cobalt, and nickel systems.
in the gas phase gives rise to a carbon concentration gradient in the catalytic metal crystallite. An extension of this reasoning, however, leads one to expect carbon gasification if the carbon activity in the gas were maintained below the activity in the filament. This is not experimentally observed as indicated by our experiments in Figures 6 and 7, and one can only conclude that the model is incomplete. In summary, carbon deposition is believed to have occurred via a metal carbon intermediate, upon reaching the metal-metal carbide-gas equilibrium carbon deposition ceases. In the region between the carbide and graphite equilibrium curves where no carbide is present, no carbon deposition or removal was observed. The rate, however, was slow and became negligible as the temperature of the experiments decreased.
Application of Results To date, iron, nickel, and cobalt have been examined as possible catalysts for the Bosch process. The maximum water concentration obtainable in the iron system was shown by both Manning (1976) and Sacco (1977) to be determined by the iron-iron oxide-gas equilibrium at low temperature and by the graphite-gas equilibrium at high temperature. The equilibrium water concentrations obtainable in the graphite, cobalt, nickel, and iron system are shown in Figure 10. For the iron system, the maximum water concentration obtainable occurs at 915 K, the temperature at which the iron-iron oxide-gas equilibrium and the graphite equilibrium intersect at an O/H ratio of 0.5. This result, originally predicted theoretically, has been confirmed in a recent experimental investigation by Heppner et al. (1980). For the cobalt system, the maximum water concentration is obtained at 800 K, the point at which the carbide-gas equilibrium begins to have an effect. In the experiments run over nickel, the maximum water concentration occurred at 825 K. The water concentrations were 16,32, and 25% for iron, cobalt, and nickel, respectively. In a simple recycle reactor, where the feed is hydrogen and carbon dioxide (in the ratio of 2:l) and water is removed from the recycle gas, the recycle ratio is calculated as r = .
2(
&)
- 1 (mol of gas recycled/mol of C 0 2 fed)
For the iron, cobalt, and nickel system, the minimum recycle ratios are 10.5, 4.3, and 6.0 respectively. For an equilibrium reactor, cobalt yields the lowest recycle ratio. Literature Cited Browning, L. C.; Emmett, P. H. J . Am. Chem. SOC. 1952, 7 4 , 1680. Dent, J. F.; Moignard, L. A.; Blackburn, W. H.; Herbden. D. “An Investigation into the Catalytic Synthesis of Methane by Town Gas Manufacturlng”:49th Report of the Joint Research Committee of the Gas Research Board of
Ind. Eng. Chem. Process Des. Dev. 1982, 21, 409-475 the University of Leeds, GRB 20 (1945). GermMan, J. E. W.D. Thesis, Massachusetts Insmute of Technology, Cambridge, MA, 1980. Heppner, D. B.; Halick, T. M.; Schubett, F. H. “Performance Characterlratbn of a B o b C02 Reduction Subsystem”: NASA CR-152342 (Feb 1980). Hofer, L. J. E.; Cohn, E. M.; Peebles, W. C. J . phvs. COlkM Chem. 1949, 53, 861-669. Hofw. L. J. E.; Cohn. E. M.; Peebies, W. C. J . phvs. Cdbkj Chem. 1950, 54. 1161-1169. Holmes, R. F.; Keller, E. E.; King, C. D. “A Carbon Dioxide Reduction Unit Usina Bosch Reactbn and Expendable Catalyst Cattridges”; NASA CR1682 (Nov 1970). Manning, M. P. Sc.D. Thesis. Massachusetts Instltute of Technology. Camb*. MA, 1976. Manning, M. P.; Reid, R. C. Ind. Eng. Chem. Process Des. Dev. 1977, 16, 356. Nagakura, S. J . Wys. SOC.Jpn. 1957, 12(5),482-494. Nagakura, S. J . h y s . Soc.Jpn. 1901, 16(6), 1213-1219.
409
Renshaw, G. D.; Roscoe, K.; Walker, P. L. J . Catai. 1971, 22, 394. RostrupNielsen, J. J . Catai. 1972, 27. 343. RostrupNieisen, J.; Trimm, D. L. J . Catal. 1977, 48, 155-165. Sacco, A. Sc.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1977. Sacco, A.; Reid, R. C. AIChE J . 1979, 25, 839. Schenck, R. 2.Anorg. Chsm. 1927, 164, 315-324. Trimm, D. L. Catai. Rev. 1077, 16. 155-189.
Received for review September 15, 1980 Revised manuscript received January 27, 1982 Accepted February 18, 1982 The authors gratefully acknowledge the support of the National Aeronautics and Space Administration under Grant No. NGR22-009-123.
Thermodynamic Model for Solvating Solutions with Physical Interactions Eugene E. Spala‘ and Ne11 L. Rlcker’ Depam“
of Chemical Engineering, University of Washington, Seattk, Washington 98 195
A method for the correlation of phase equllibrlum data for solvating, multicomponent liquid solutions is proposed. Chemical equilibrium constants are used to calculate the extent of formation of discrete solvation complexes, and the UNIFAC group-contrlbution theory is wed to predict the physical interactions between species in solution. The method is applied to example binary and multicomponent solvating systems including a quaternary trioctylaminelacetic acid/sohrent/water system from a developmental llquid extraction process that exhibits unusually complex phase equllibrla. The proposed method gives a much better representation of such systems than has been reported previously. Potential shortcomings of the approach are also discussed.
Introduction Many separation processes in the chemical industry exploit solvation effects, i.e., specific chemical interactions between two or more components in liquid solution. Common examples of such processes include extractive and azeotropic distillation, acid-gas absorption, and the separation of metal ions by liquid extraction. Recently, Ricker et al. (1980a,b) proposed a liquid extraction process for the recovery of carboxylic acids from aqueous wastes;the process was based on an extractant mixture of a long-chain tertiary aliphatic amine and an organic diluent. Wardell and King (1978) and Ricker et al. (1979) showed that under the optimal extraction conditions acetic acid and similar solutes were highly solvated in this extractant phase, resulting in favorable equilibrium distribution coefficients. They found, however, that the phase-equilibrium behavior was a complicated function of the amine structure, the type of diluent, the relative amounts of amine and diluent in solution, and the equilibrium concentration of the acidic solute(s). I t was clear that the optimization of the process would require either a great deal of experimentation or a calculational model that one could use to predict the equilibrium distributions of the various components. In the present paper, we develop such a model and present several example applications. Stauffer Chemical Co., Richmond, CA 94804. 0196-4305/82/1121-0409$01.25/0
The Calculational Model Traditional theories of liquid solutions interpret solution nonidealities exclusively in terms of either strong “chemical”intermolecular interactions or weak “physical” interactions (see, e.g., Prausnitz, 1969). These distinct viewpoints lead to relatively simple thermodynamic models that can be used to correlate data for solutions in which one type of interaction dominates over the other. Unfortunately, neither of these simple viewpoints adequately represents the equilibria of the acid/amine/diluent systems studied by Ricker et al. (1979). In these and in analogous systems, a representative model must allow for the simultaneous effects of chemical and physical interactions in solution. In other words, a unified treatment is required. Work along these lines has been performed by a number of previous authors, including, e.g., Harris and Prausnitz (1969),Renon and Prausnitz (1967),Nagata (1973,1977), Nagata and Kawamura (1979a,b), and Chen and Bagley (1978). The most common approach has been to express the excess Gibbs free energy as the s u m of a chemical and a physical contribution The chemical contribution results from the dependence of the “true” composition of the solution on the degree of solvation and/or association of the various species. Many of the past investigations have involved solutions in which one component, usually an alcohol, was assumed to self@ 1982 American Chemical Society
