Carbon monoxide oxidation in supercritical water: the effects of heat

---

Energy & Fuels 1992,6, 586-597

586

Carbon Monoxide Oxidation in Supercritical Water: The Effects of Heat Transfer and the Water-Gas Shift Reaction on Observed Kinetics? H. Richard Holgate, Paul A. Webley, and Jefferson W. Tester* Department of Chemical Engineering and Energy Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

Richard K. Helling Dow Chemical Company, Pittsburg, California 94565 Received January 14, 1992. Revised Manuscript Received June 1, 1992

Oxidation in a supercritical water environment is an efficient method for treating wastes without formation of harmful byproducts. Complete oxidation of organics can be limited by conversion of carbon monoxide to carbon dioxide. The kinetics of carbon monoxide oxidation has been studied previously (Helling, R. K.; Tester, J. W. Energy Fuels 1987,1,417),but efforts to establish kinetic parameters for the direct oxidation pathway (CO + l/zOz COz) were complicated by the reaction of carbon monoxide with water via the water-gas shift reaction pathway (CO + HzO COz + Hz) during preheating of the reactor feeds. The kinetics of carbon monoxide oxidation has been reexamined in an updated experimental apparatus with improved temperature measurement and hydrogen detection capabilities, and the studied range of Oz/CO feed ratios has been extended to the substoichiometric regime. Using the results of heat-transfer experiments, the temperature profiles within the feed preheater were established. The experimentally determined temperature profiles compared favorably with profiles predicted by conventional heat-transfer correlations. These temperature profiles were used in determining new kinetic parameters for the water-gas shift pathway and for the direct oxidation of carbon monoxide in supercritical water. The two kinetic pathways can apparently be treated as separable under the conditions of this study. Regressed kinetic rate forms were similar to those obtained in the earlier study. The new oxidation data exhibit a fractionalorder dependence on oxygen concentration, consistent with gas-phase studies of carbon monoxide oxidation.

-

Introduction Oxidation in a supercritical water environment is an innovative technology for the rapid destruction of hazardous organic wastes without the formation of harmful bypr~ducts.'-~Pure water is considered supercritical if ita temperature and pressure both exceed the critical values of 374.2 OC and 221 bar, respectively. At or above the critical point, the density of water is a strong function of both temperature and pressure, as are the solvation properties of water.5 Supercritical water acts as a dense gas, with the solvation characteristics of a nonpolar or* Author to whom correspondence should be addressed.

+ Presented

a t the AIChE National Meeting, November 17-22,1991. (1)Tester, J. W.;Holgate,H. R.;Armellini,F. J.;Webley,P.A.;Killilea, W. R.; Hong, G. T.; Barner, H. E. Presented at the Third Annual Symposium on Emerging Technologiesfor Hazardous Waste Management, Atlanta, GA, October, 1-3, 1991, paper 113.2;to appear in an ACS Symposium Series volume, Emerging TechnologiesforHazardous Waste Management III. (2)Thomason, T.B.; Hong, G. T.; Swallow, K. C.; Killilea, W. R. In Innovative Hazardous Waste Treatment Technology Series, Volume 1: Thermal Processes; Freeman, H. M., Ed.; Technomic Publishing: Lancaster, PA, 1990; p 31. (3)Swallow,K. C.; Killilea, W. R.; Malinowski, K. C.; Staszak,C. Waste Manage. 1989,9,19. (4)Modell,M. Instandard Handbook of Hazardow Waste Treatment and Disposal; Freeman, H. M., Ed.; McGraw-Hill: New York, 1989;p 8.153. (5)Franck, E. U.Pure Appl. Chem. 1970,24, 13. (6)Connolly, J. F. J . Chem. Eng. Data 1966,1 1 (l),13.

0887-0624/92/2506-0586$03.00/0

-

ganic+ organics and gases are completely miscible with supercritical water, yet inorganic salts are virtually insoluble.' In supercritical water oxidation, the unique properties of supercritical water allow oxygen and organics to be contacted in a single phase at temperatures above 400 O C and pressures above 230 bar. Under these conditions, spontaneous oxidation of the organics occurs, raising the mixture temperature to 550-650 "C,where destruction proceeds rapidly and completely. Conversions of 99.99 % or greater can be attained with reactor residence times of 1 min or less. Heteroatoms are converted to mineral acids which can be neutralized and precipitated from the mixture as salts by adding a base to the feed.0 Kinetic data for the oxidation of model compounds in supercritical water have become available in recent years. Compounds studied have included carbon monoxide,g methane,10 methano1,llJZ a"onia,12J3 and hydrogen,14 (7) Martynova, 0. I. In High Temperature, High Pressure Electrochemistryin A q w o w Solutions; Jones,D.deG.,Staehle,R. W.,Chairmen; National Association of Corrosion Engineers: Houston, TX, 1978;p 131. (8)Thomason, T. B.;Modell, M. Hazard. Waste 1984,1 (4),463. (9)Helling, R.K.;Teeter, J. W. Energy Fuels 1987,1 , 417. (10)Webley, P. A.; Tester, J. W. Energy Fuels 1991,5,411. (11)Webley, P. A,; Teeter, J. W. Supercritical Fluid Science and Technology;Johnston, K. P., Penninger, J. M. L., Ede.; ACS Symposium Series 406;American Chemical Society: Washington, DC, 1989;p 269. (12)Webley, P. A.; Tester, J. W.; Holgate, H. R. Znd. Eng. Chem. Res. 1991,30 ( E ) , 1745. (13)Helling, R.K.;Teeter, J. W. Environ. Sci. Technol. 1988,22(ll), 1319.

0 1992 American Chemical Society

Energy & Fuels, VoZ. 6, No. 5,1992 587

Carbon Monoxide Oxidation in Supercritical Water

as well as more complex compounds such as phen01.l~Of these, carbon monoxide and ammonia are important because they are intermediates in the oxidation of more complex organics and because their destruction can be the rate-limiting step in the complete conversion of the organic to C02 and N2. Carbon monoxide is also the simplest model organic, and its gas-phase oxidation kinetics has been well characterized. Helling and Tester9 were the first to report the oxidation kinetics of carbon monoxide in supercritical water. They observed that CO is converted toCO2by two global reaction pathways: direct oxidation (eq 1)and the water-gas shift reaction (eq 2). Below about 450 "C,the water-gas shift

co + l/,O, CO + H,O

-+

-

CO,

(1)

CO, + H,

(2)

accounted for more than 50 % of the total oxidation of CO, as evidenced by the amount of hydrogen produced. The duality of the oxidation pathways complicated the task of obtaining kinetic parameters. In our experimental apparatus, carbon monoxide and oxygen are delivered separatelyto the reactor as dilute aqueous solutions. Since the water-gas shift can occur without oxygen present, the aqueous carbon monoxide feed could undergo reaction as it was heated to reaction temperature, prior to its entering the reactor. Hellingls attempted to calculate the thermal history of the feeds using conventional heat-transfer correlations, but found that the predicted thermal behavior did not agree with that observed experimentally. Furthermore, temperatures at the rector inlet often differed from the surrounding temperature by as much as 10-25 OC,resulting in some degree of reactor nonisothermality. Despite these experimental limitations, a rate expression for the water-gas shift reaction pathway was developed by Helling and Tester,9 on the basis of 21 experiments in which no oxygen was fed to the reactor: rate,,,

= 101.6M3*57 exp((42.9 f 8.6)lRT)[C01°.57f0."

(3) where the rate is given in mol/L-s, the concentration is in mol/L, the activation energy is in kJ/mol, and the stated parameter uncertainties are at the 95 % confidence level. For the overall oxidation of CO, Helling and Testerg obtained the following global expression, based on 39 experiments with oxygen present at stoichiometric or superstoichiometric levels (i.e., ([021/ [COl)i,let I 0.5): rateoverall=

107.25f0.53

exp((-120 f 7.7)lRT) X

[co]1.01f0.09 1 0 2 10.03f0.04

(4)

with the same units and uncertainties as before. In deriving the above expression, all reactions were assumed to have occurred in the reactor, at the reactor inlet temperature. The effects of reaction during preheating were included in eq 3, although the thermal history of the feeds could not be reliably determined. (14) Holgate, H. R.;Tester, J. W. Fundamental Kinetics and Mechanisms of Hydrogen Oxidation in Supercritical Water. Presented at the Second International Symposium on Supercritical Fluids, Boston, MA, May 20-22,1991;to appear in Combust. Sei. Technol. (15)Thornton, T. D.; Savage, P. E. AIChE J. 1992,38 (3), 321. (16)Helling, R. K. Oxidation Kinetics of Simple Compounds in SupercriticalWater: Carbon Monoxide, Ammonia, and Ethanol. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1986.

Heal E x c h a w

II

Flowmeter

GC Analysis

Figure 1. Isothermal plug-flow apparatus for the study of reactions in supercritical water.

The objectives of the present study were 3-fold first, to reproduce (if possible) the earlier data in an improved apparatus; second, to extend the kinetic measurementsto the substoichiometricregime (with (IO,]/[COl)aet< 0.5); and third, to account for reaction during preheating in a rigorous fashion by improving temperature modeling in the preheating tubing. The reactor system of the earlier study was rebuilt and now includes greater reactor temperature measurement capabilities and an improved analytical method for hydrogen, as well as improved feed preheating to insure system isothermality. Heat-transfer experiments in the new system enabled characterization of the temperature profiles within the feed preheaters. These profiles were then incorporated in the data analysis of water-gas shift experiments to provide an improved rate expression for that pathway. The water-gas shift rate expression was then used to quantify the extent of reaction during feed preheating in experiments with oxygen present. In this way the "true" reactor inlet concentrations could be calculated and the effects of preheating accounted for in the data. Experimental Section Experimental Apparatus. A diagram of the current experimental apparatus is shown in Figure 1. Dilute aqueous feed solutions of carbon monoxide and oxygen were prepared by dissolving the gases in purified water in two high-pressure 3-L saturators. During water-gas shift experiments,both saturators were pressurizedwith carbon monoxide. Agitationwas provided by pumped recirculation of the saturator liquid. The feed solutions were delivered separately to the reactor by a duplex high-pressure feed pump. The reactor was 4.71 m of 0.635 cm 0.d. X 0.171 cm i.d. Inconel 625 tubing contained in a fluidized sand bath. The feeds were preheated separately to the reaction temperature in two 2.8-m lengths of 0.159 cm 0.d. X 0.108 cm i.d. Hastelloy C276 tubing (denoted "preheating tubing" in Figure 1)contained in the reactor sand bath. The preheated feeds met and mixed at the reactor inlet, where the oxidation reaction was initiated. The reactor temperature was taken as the average of the mixing and exit fluid temperatures; in no case did these temperatures differ by more than 8 "C, and the exit temperature was always within 1 "C of the sand bath temperature. Upon exitingthe reactor,the reactionmixture was quenchedto ambient temperature in a countercurrent shell-and-tube heat exchanger, and the pressure was reduced to ambient upon passing through the back-pressureregulator. The resulting gas and liquid phases were disengaged in a gas-liquid separator, and the flow rate of each phase was measured. Compositional analysis of the gas phase was accomplished by gas chromatographyusing a thermal

Holgate et al.

588 Energy &Fuels, Vol. 6, No. 5, 1992

conductivity detector. High sensitivityfor hydrogen was obtained by performing duplicate sample analyses in a second GC using nitrogen as the carrier gas. The present apparatus is quite similar to that used by Helling and Tester.9 There are, however, three distinctions relevant to this study. First, the earlier reactor was 4.24 m of 0.635 cm 0.d. X 0.211cm i.d. Inconel 625tubing, providinga total reactor volume about 25% greater than in the current system and a surface areato-volume ratio about 20% smaller. Second, the earlier system used only 2 m of preheating tubing in the reactor sand bath; auxiliary heating was provided by radiative preheaters which have since been removed. Finally, the reactor temperature in the earlier apparatus was measured only at the point of mixing of the feeds. This mixing temperature frequently differed from the sand bath temperature by as much as 10-25 "C. Other details regarding the current experimental apparatus and procedures are available elsewhere." Experimental Conditions. The oxidation of carbon monoxide in supercritical water was investigated in 43 experiments at an operating pressure of 246 bar and over the temperature range 420-571 "C, at residence times of 5.0-12.1 s. Inlet concentrations of carbon monoxide ranged from 3.73 X lo4 to 3.51 X lO-amol/L, while inlet oxygen concentrationsranged from 3.81 X lo4 to 3.87 X mol/L. Molar feed ratios of oxygen to carbon monoxide varied between 0.15 and 8.20,both above and below the stoichiometric value of 0.5. Twenty experiments were conducted to investigatethe watergas shift reaction in supercritical water at 246 bar and from 445 to 593 "C. Residence times in these experiments ranged from 5.1 to 11.1 8, with inlet carbon monoxide concentrationsvarying from 3.76 X 10-4to 2.88 X mol/L. Oxygen was deliberately excluded from the reactor in these runs; however, it was not possible to eliminateoxygen completely from the system, although the molar feed ratio of oxygen to carbon monoxide never exceeded 0.022. The range of conditionsfor both the oxidationexperiments and the water-gas shift experiments was very similar to that used in earlier experiments: with the important exception that in this study substoichiometricfeed ratios were also investigated in the oxidation experiments. Heat-transfer experiments were conducted in the current system to establishtemperatureprofilesin the preheating tubing. In these studies, the mixing temperature of the feeds was measured for varying flow rates and sand bath temperatures, with pure water flowing through the system. Flow rates ranged from 2.8 to 5.1 g/min, with sand bath temperatures of 400-575 "C.

Results Kinetics Experiments. Carbon monoxide oxidation experiments yielded total conversions ranging from 4 to 937%,based on the formation of carbon dioxide from carbon monoxide. Average reaction rates, defined as the change in carbon monoxide concentration divided by the residence time, ranged from 5.8 X 104 to 1.8 X mol/L.s. The conditions and results of the oxidation experiments are listed in Table I. Note that the inlet carbon monoxide concentrations have been corrected for reaction during preheating of the CO feed, as described in the Discussion section. The carbon monoxide oxidation data are also presented in the form of a first-order Arrhenius plot in Figure 2, where they are compared with our earlier data reported by Helling and Tester.9 The apparent first-order rate constant, k*, is defined as k* = -In (1 - X ) / r (5) with X the observed conversion of carbon monoxide and 7 the reactor residence time. The reactor temperatures -

~~

~

~~

~~

~

(17)Webley, P. A. Fundamental Oxidation Kinetics of Simple Com-

pounds in Supercritical Water. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1989.

in the earlier experimentsghave been adjusted to the sand bath temperature, which is believed to be more representative of the actual temperature within the reactor (see the section on heat-transfer experiments for details). Figure 3 shows the molar ratio of hydrogen to carbon dioxide in the effluent of the oxidation experiments plotted as a function of reactor temperature. This ratio is a measure of the relative contributionsof the water-gas shift and the direct oxidation pathways to the overall conversion of carbon monoxide to carbon dioxide. Previous work14 has demonstrated that the global oxidation of hydrogen proceeds relatively slowly under the conditions of these experiments, so that hydrogen formed by the water-gas shift pathway is unlikely to be oxidized to a significant extent. The H2/C02 ratio is thus an accurate indicator of the contribution of the water-gas shift in these experiments. Both in the data of Helling and Tester and in the data from the present study, the H2/C02 ratio exhibits a clear decreasing trend with increasing reactor temperature, indicating that the relative contribution of the water-gas shift becomes smaller as the temperature increases, Water-gas shift experiments yielded total conversions of 1.4-23%. Since oxygen could not be completely excluded from the reactor, some conversionof CO occurred by direct oxidation, as evidenced by effluent H&02 ratios that were less than unity. The measured H2/C02 ratios were used to calculate the conversion by the water-gas shift pathway alone; this calculated conversion varied between 1.4 and 20% in these experiments. Average water-gas shift reaction rates varied from 1.3 X 104to 3.0 X mol/L.s, significantly slower than overall oxidation rates. The Conditions and results of the water-gas shift experiments are summarized in Table 11. Again, reactor inlet CO concentrations have been corrected for reaction during feed preheating. The water-gas shift data are also presented in a first-order Arrhenius plot in Figure 4,where they are compared with the water-gas shift data presented earlier by Helling and Tester.9 As before, the earlier data have been adjusted with respect to reactor temperature. Several trends are apparent in Figures 2-4. First, the early overall oxidation data seem to indicate somewhat higher observed oxidation rates than in the present study, although the activation energies of the two data sets appear comparable. Similarly, rates in the earlier water-gas shift data also appear faster than in the current study, although again the activation energies do not seem vastly different. Finally, the effluent H2/C02 ratios in the earlier oxidation data appear consistently higher. Taken together, these three observations suggest that the water-gas shift pathway proceeded faster in the earlier experiments. One possible source of this discrepancy is a difference in the thermal history of the feeds in the two studies. This possibility, along with the importance of quantifying the extent of reaction during feed preheating, prompted an investigation of the temperature profiles in the preheating tubing in the current system. Heat-Transfer Experiments. Average overall heattransfer coefficients were determined for the feed preheating tubing in a series of heat-transfer experiments. In these experiments, pure water was used; since the true reactor feeds are very dilute, their thermophysical properties will be very similar to those of pure water. The flow rates of the two feed streams could be varied independently by varying the feed pump stroke. For a given sand bath temperature and set of flow rates, the temperature at the

Energy & Fuels, Vol. 6,No. 5,1992 589

Carbon Monoxide Oxidation in Supercritical Water

-

Table I. Exwrimental Data for the Overall Oxidation of Carbon Monoxide in Supercritical Water (246 bar) [colo,a av rate/ C mass 0 mass 10-3 moVL Rb T: s X,d % H2/COf 10-5mol/L.s In k* 8 In kmh closure,' % closure) %Y 1.36 -4.43 2.168 7.7 8.8 92.5 93.1 1.19 -1.60 0.0789 15.1 0.04 2.171 5.4 66.4 97.6 94.3 1.23 -4.01 0.5230 -3.04 7.7 13.0 91.1 97.5 2.03 2.233 1.20 -3.40 -1.16 4.19 0.0482 0.491 6.0 18.2 95.2 98.9 1.38 -3.31 -1.18 0.1652 0.440 6.1 20.0 94.9 99.2 4.50 1.37 2.91 0.1859 0.450 6.7 11.0 97.8 94.8 -4.05 -2.08 1.77 -4.20 2.72 0.1354 7.1 94.8 93.0 -2.20 10.1 1.91 0.462 -2.72 0.0603 97.0 83.6 -0.55 31.3 8.33 0.520 5.7 1.52 0.0989 -1.58 -3.69 14.5 3.98 0.479 6.3 93.2 92.9 1.73 -1.99 14.3 0.34 0.0306 5.3 94.9 96.4 51.7 1.47 0.454 -2.90 26.2 6.81 -0.66 0.0603 0.457 5.5 98.8 96.8 1.43 97.4 97.0 -5.36 1.09 -3.65 0.2960 0.471 8.7 4.0 2.38 0.3817 96.1 90.6 -5.37 0.918 -3.79 0.567 8.1 3.7 2.01 -2.20 -0.19 45.7 0.1131 5.5 99.0 96.7 8.30 0.899 1.00 0.3525 -4.71 -2.80 1.57 7.0 0.254 97.3 96.6 6.1 1.80 -2.27 0.1031 97.8 82.2 -4.65 2.22 5.9 0.182 5.5 2.38 -3.85 -1.44 3.58 5.5 0.0887 0.213 98.3 95.7 11.0 1.79 -1.71 15.2 0.78 5.1 0.1271 98.4 86.6 60.2 0.303 1.29 1.62 0.0080 98.0 97.5 -0.61 93.3 17.6 5.0 0.930 0.944 -3.67 15.5 -2.26 0.972 6.6 0.2501 6.812 91.8 95.1 0.414 7.5 0.1955 92.9 98.3 -4.99 5.0 -3.27 1.98 0.711 2.98 0.1011 97.1 91.2 -4.10 -2.29 3.22 5.8 1.OOO 9.2 2.04 -3.62 -1.60 5.5 0.0597 99.0 95.7 13.6 4.66 0.633 1.88 -1.44 0.0727 99.2 91.5 -3.61 13.8 4.76 5.5 0.406 1.89 0.0531 95.6 72.7 -2.78 28.9 -0.87 5.52 5.5 1.486 1.05 -2.40 -0.53 4.21 0.0346 2.767 82.9 75.5 39.3 5.5 0.589 0.0403 69.0 90.8 -2.37 40.2 -0.60 3.01 5.5 4.757 0.412 84.0 94.1 -2.43 -0.68 4.84 5.5 0.0653 2.925 38.4 0.694 -2.11 -0.37 0.0534 69.6 94.3 48.6 3.71 5.5 4.738 0.420 93.9 94.0 -4.42 -2.44 3.11 0.1790 7.0 6.0 0.360 2.68 -3.81 -1.54 3.72 0.1301 94.6 98.2 11.5 5.5 0.284 1.78 -2.24 0.1674 89.6 88.1 -3.82 12.0 2.54 5.8 2.537 1.23 -3.28 -1.71 0.1048 7.882 95.5 95.2 29.5 1.18 9.3 0.373 -4.31 -2.84 0.1244 98.3 96.5 14.9 0.583 12.1 8.199 0.472 0.1512 -4.04 14.8 -2.09 3.38 100.0 93.2 0.524 9.1 2.08 -5.21 1.41 0.1737 0.506 6.0 -3.32 11.4 98.7 95.1 2.65 0.0072 -1.43 86.0 0.63 92.8 97.4 16.6 0.962 1.58 8.2 0.0867 -4.71 -2.84 99.1 98.4 1.57 1.OOO 1.82 8.6 10.0 0.1708 -4.16 -2.57 14.0 1.37 2.786 9.6 94.8 99.6 0.944 -4.82 0.2306 2.802 -3.38 93.9 99.9 0.817 8.4 10.9 1.06 0.1666 -3.67 -1.34 19.4 98.6 77.7 6.49 0.161 2.84 8.5 -5.14 0.2786 -3.06 10.4 99.8 82.4 2.00 0.153 3.51 5.9 -3.84 -1.55 17.3 4.68 0.0690 0.275 8.8 97.6 98.0 2.38

T,O C 438.5 540.0 439.0 500.0 497.0 469.0 454.0 519.0 484.0 549.0 530.0 420.0 430.5 534.5 460.0 508.0 528.5 560.0

571.0 472.0 445.0 508.5 529.5 529.5 529.5 529.5 529.5 529.5 529.0 498.0 529.5 508.5 499.5 436.5 509.0 447.0 545.5 478.0 488.5 457.5 530.0 467.5 518.5 a Inlet CO concentration, corrected for reaction during preheating. * Feed ratio, [O2],J[COl0.e Reactor residence time. d Conversion of carbon monoxide in reactor. e Molar ratio of hydrogen to carbon dioxide in reactor effluent. f Average reaction rate, defined as X[CO]~T. t First-orderrata constant, &*= -In (1- X)/T. Rate constant for direct oxidation,from regressed rate expression (eq 22). Carbon mass balance closure. j Oxygen maas balance closure. 580

540

580

Reaclor Temperature. 'C 500 480 460

520 I

I

I

440

420

I

I

I

400

Helling and Tester. 1987

1 -

o Present Study

9

e

00

-2

r

b i -

e

-3 4

6 -

380 , 7 1.15

'

.

I

1.x)

.

'

'

I

1.25

'

I

1.30

'

,

~

1.35

,

,

)

1.40

145

150

1 OOW, k'

Figure 2. Apparent first-order Arrhenius plot of the overall oxidation of carbon monoxide in supercriticalwater at 246 bar. Data of Helling and Teste9 have been corrected for calculated reactor nonisothermalities (see text for details).

point of mixing of the two feeds could be measured. When the two flow rates were identical, the mixing temperature

400

420

440

460 460 500 520 Reactor Temperature, 'C

540

560

580

600

Figure 3. Molar ratio of hydrogen to carbon dioxide in effluent of carbon monoxide oxidation experiments at 246 bar.

was assumed to be the same as the feed temperatures prior to mixing, as the preheating tubing waa the same length for both feed streams. Varying one of the feed flow rates resulted in a change in the mixing temperature; the temperature of each feed immediately prior to mixing could

590 Energy & Fuels, Vol. 6, No. 5, 1992

Holgate et al.

Table 11. Experimental Data for the Water-Gas Shift Reaction in Supercritical Water (246 bar) [COIo,a

T,"C

10-3 mol/L

T? S

560.5 498.0 540.0 477.0 518.5 455.5 508.5 467.0 528.5 487.5 482.5 445.5 593.0 457.5 531.0 489.5 484.0 447.5 499.5 478.5

1.350 1.740 1.120 1.360 0.8920 1.150 0.7520 0.8640 0.3760 0.4320 2.260 2.670 0.7610 1.170 0.4080 0.4630 2.440 2.880 1.740 1.900

5.1 6.1 5.4 6.5 5.7 7.1 5.9 6.8 5.6 6.3 6.3 7.4 7.6 11.1 8.7 9.8 9.8 11.4 9.4 10.1

X,e

72

X,,,d % .-

av rate: 10-6 mol/L.s

In k* f

11.2 1.1 10.6 1.3 7.3 1.4 3.8 2.0 6.5 4.3 2.1 1.0 16.3 1.3 10.3 3.3 0.6 0.9 1.9 1.3

2.969 0.3225 2.192 0.2783 1.141 0.2296 0.4854 0.2595 0.4375 0.2975 0.7400 0.3547 1.629 0.1343 0.4832 0.1575 0.1607 0.2179 0.3425 0.2533

-3.64 -5.88 -3.80 -5.75 -4.22 -6.21 -4.87 -5.60 -4.27 -4.81 -5.52 -6.30 -3.54 -6.25 -4.22 -5.39 -6.51 -6.85 -5.95 -6.29

12.5 1.7 11.4 2.1 8.0 1.4 4.4 2.5 7.5 5.0 2.5 1.4 19.7 2.1 12.0 4.4 1.5 1.2 2.4 1.9

C mass

In k g -5.67 -8.03 -5.87 -7.85 -6.37 -8.49 -7.13 -7.82 -6.72 -7.19 -7.45 -8.19 -5.87 -8.55 -6.66 -7.90 -8.77 -8.93 -8.22 -8.46

c1osure.h -. 5%92.0 99.8 98.8 97.7 98.4 95.7 96.3 96.7 87.7 86.1 94.5 95.8 97.0 95.2 89.3 89.6 99.7 99.2 97.2 95.4

0 Inlet CO concentration, corrected for reaction during preheating. * Reactor residence time. Total conversion of carbon monoxide by water-gaa shift pathway. d Conversion of carbon monoxide in reactor by water-gaa shift pathway. e Average reaction rate, defiied as X,[CO]J T . f First-order rate constant, k* = -In (1 - X ) / T .8 Rate constant from regressed rate expression (eq 16). Carbon mass balance closure.

350 I

I

I

I

I

I

I

I

I

I

I

,

Y

E 2300 -

2 -

4

1

3 -

*

@

-

*

y p

- Now Rare = 4 5 g m m

,

c250 -

7

F

-5

I

8200 -

i d

-

1

-6

m1 5 0 g t

l

t

-9 -10 1 1.10

,

1.15

1.20

1.25

1.30

1.35

,'I

1

L

1.40

A

1.45

1.50

1 OOWT. K-'

Figure 4. Apparent first-orderArrhenius plot for the water-gas shift reaction in supercritical water at 246 bar. Data of Helling and Testers have been corrected for calculated reactor nonisothermalities (see text for details). then be determined by an adiabatic, steady-state, firstlaw analysis (conservation of enthalpy), as the temperature of the unchanged feed stream would remain constant. Using this simple technique, temperatures at the exit of the preheating tubing were determined for both feed streams. For a differential element of preheating tubing, the heat transfer from sand bath to feed stream can be described by

where rh is the mass flow rate of the feed stream, H is the specific enthalpy of the bulk fluid, U,is the overall heattransfer coefficient from sand bath to feed stream, Tfsb is the fluidized sand bath temperature, T is the fluid temperature, and A is the internal surface area of the preheating tubing. If U,is assumed to have a (constant) characteristic value ( U,)over the length of the tubing, eq 6 can be integrated to give

-

Flow Rare = 2 8 gfmin

1

inn I 350

400

450

500

550

600

Sand Bath Temperature ' C

Figure 5. Average overall heat-transfer coefficients U, for preheating tubing, as determined experimentally.

(7) where TOis the temperature at the inlet to the preheating tubing (room temperature) and Tfis the temperature at the preheating tubing exit (which was determined experimentally). Using the results of the heat-transfer experiments, average overall heat-transfer coefficients for the preheating tubing, as given by ( U,), could be calculated by numerically integrating eq 7 using the enthalpy of water given by the equation of state of Haar et al.la This method is admittedly crude, as it essentially infers the thermal history of the feed solely on the basis of the temperature at the preheater exit. On the other hand, the thermal history calculated using the experimentally derived ( Uo) will by definition correctly predict the observed temperature at the preheating tubing exit, where the details of the thermal history are most important for our purposes. Experimentally derived values for ( Uo)are shown in Figure 5 for various sand bath temperatures and feedstream flow rates. The experimental conditions were chosen as representative of those encountered in the kinetic (18)Haar,L.; Gallagher, J. S.; Kell, G. S. NBSINRC Steam Tables; Hemisphere Publishing Corp.: New York, 1984.

Carbon Monoxide Oxidation in Supercritical Water

experiments. T w o pump stroke settings were used, yielding nominal flow rates of 5 and 3 g/min for each feed stream. At a given pump setting, the oxygen-side pump head produces a somewhatlower flow rate (by about 10% ) than the organic-side head. Consequently, during an oxidation experiment, the flow rate of the oxygen feed was slightly lower than the flow rate of the organic feed. This small discrepancy in flow rates is reflected in the two pairs of curves in Figure 5. Several features are evident in Figure 5. First, (U,) increases with increasing flow rate; this behavior is expected since heat transfer should become more effective as the flow grows more turbulent (or less laminar). In the preheating tubing, the maximum attained Reynolds number increases from 1940 to 3550 as the flow rate increases from 2.8 to 5.1 g/min. Second, (U,)decreases with increasing sand bath temperature, after appearing to peak at a sand bath temperature of about 425 OC. The reasons for this behavior are less obvious. The reduced heat transfer at the higher temperatures may be the result of a decreasing internal or external heat-transfer coefficient or both. Finally, the magnitude of the (U,) values should be noted: at 125-350 W/m2.K, the coefficients are not very large, especially considering the manufacturer’s quoted value of 600 W/m2*Kfor the external sand bath coefficient. The ability to calculate an experimental value for U,in eq 6 rests on the assumption that U,is constant (and thus may be accurately represented by (U,)) over the length of the tubing. The validity of this assumptionis not readily apparent, particularly since the feed-stream properties vary drastically as the feeds are heated past the pseudocritical point. One might expect the overall heat-transfer coefficient to undergo similarly drastic variations. To test the validity of the constant U,assumption, the calculated (U,)values are compared with values predicted using establiihed empirical correlations for heat transfer to fluids flowing in horizontal tubes. These correlations predict only the internal heat-transfer coefficient and typically take the form Nu = h,d/k = f(Re,Pr,Gz Re Pr d / L , ~ ( d c ( ~(8) ) where Nu, Re, Pr,and Gz are the dimensionless Nusselt, Reynolds, Prandtl, and Graetz numbers, respectively, hi is the internal heat-transfer coefficient, d is the tubing internal diameter, k is the fluid thermal conductivity, and pi, and pw are the viscosities of the fluid in the bulk and at the tubing wall. At the critical point, the Prandtl number of a fluid diverges, and traditional heat-transfer correlations often fail in the near-critical The predictions of three different correlations were calculated. The first correlation, given by Perry et al.,lg is really three separate correlations, one for each of the laminar, transition, and turbulent flow regimes. These are the traditional Dittus-Boelter-type correlations which are meant to be generally applicable, without regard for the specific heat-transfer conditions. The second correlation used was that of Swenson et al.,20 which was developed specifically for turbulent flows of supercritical water. Swenson’s correlation accounted for the variations (19) Perry’s Chemical Engineers’ Handbook, Sixth Edition; Perry, R. H., Green, D., Maloney, J. O., Eds.;McGraw-Hill: New York, 1984. (20) Swenson, H.S.;Carver, J. R.; Kakarala, C. R. J.Heat Transfer 1965,87 (ll),477.

(21) Yamagata, K.;Nishikawa, K.;Haeegawa, S.;Fujii, T.;Yoshida, S. Int. J. Heat Mass Transfer 1972, 15, 2575.

Energy & Fuels, Vol. 6, No. 5, 1992 891 . .

500

Sand Bath Temperature

5 400 . m

-

501 .5’C

-Perry et ai., 1984 Swenson et ai., 1965

375

Yamagata et a1 , I 972 Experimentally Derived

350

I

0

05

1 15 2 Distance in Preheater Tubing, m

25

J

3

Figure 6. Comparison of calculated bulk fluid temperature in preheating tubing as a function of distance for various heattransfer models. Calculations were performed using the heattransfer correlations of Perry et al.,’e Swenson et al.,N and Yamagata et aLZ1with the external coefficient shown and using an experimentallydetermined average heat-transfercoefficient.

in heat capacity and density of the fluid between the wall and the bulk. The third correlation was developed by Yamagata et al.,21 also for turbulent flows of supercritical water. Yamagata’s correlation incorporated several correction factors which depended on the nearness of the fluid to the pseudocritical point. Swenson’s correlation was based on measurements for a wide variety of heat fluxes relative to fluid flow rates, while Yamagata’s correlation was derived from data mostly at low heat fluxes relative to flow rates. Heat transfer to supercritical water is typically less efficient than would be predicted by a traditional correlation of the form of eq 8. In other words, the heat-transfer coefficient does not increase to the same extent as the diverging Prandtlnumber in the near-critical region. At high heat fluxes, this deterioration is more significant. One should note, however, that the heattransfer coefficient always increases in the near-critical region; the deterioration affects only the magnitude of the increase. Both Swenson et al. and Yamagata et al. conducted experiments with mass velocities on the order of 300-2000 kg/m2*sand heat fluxes of approximately 100-2000 kW/ m2; by comparison, in our preheating tubing, mass velocities were about 10kg/m2*sand calculated heat fluxes were 20-60 kW/m2. Both studies consider heat flux “high” when the ratio of heat flux to mass velocity is on the order of 1 kW-s/kg. By this rough definition, heat fluxes in our system are high; however, it is not clear whether this definition can be extended from the larger scale of the earlier experiments to the small scale of our preheating tubing. Furthermore, both Swenson et al. and Yamagata et al. studied heat transfer to highly turbulent supercritical water. In our preheating tubing, the flow conditions are never far beyond the transition region (Re I3550). The direct applicability of these empirical correlations to our specific conditions is thus somewhat questionable. For the calculations using the correlations, a constant external heat-transfer coefficient was assumed, thermodynamicand transport properties were obtained from Haar et al.,l*and tubing thermal conductivitydata were obtained from the manufacturer. A differential element method was used to determine the internal and external tubing wall temperatures and the bulk fluid temperature on the basis of a radial energy flux balance. Overall heat-transfer

Holgate et al.

692 Energy &Fuels, Vol. 6, No. 5, 1992 300

cg B

8000

t-

'

1

ll

$200

P 0

b 6000

9

-

5 150

-p

Swenson et al 1965

---

E

-

q =

- - - - - _ . _ _ ._ _ _

3

0

Pt

i

-

Y

Sand Bath Temperature = 501 5°C Flow Rate 5 t gmin External Heat Transler Coefficient = 190 Wlm2-K

4000

-

2000

"0

----___, 50

100

150

ZOO 250 300 Bulk Fluid Temperature. 'C

350

400

450

Yamagata et a1 1972

100

I

I

i

Sand Bath Temperature = SO1 5% Flow Rate = 5 1 gmin External Heal Transler Coellicient = 190 WlmZ-K

t

I

I

500

Figure 7. Comparison of calculated internal heat-transfer Coefficients in preheatingtubing as a functionof fluid temperature for several heat-transfer correlations. Correlations used were those of Perry et al.,19 Swenson et al.,m and Yamagata et with the conditions and external coefficient shown. coefficients could be calculated from the sum-of-resistances formulation. Typical results from these heat-transfer calculationsare presented in Figure 6, where the predicted bulk fluid temperature is shown as a function of distance in the preheating tubing. The assumed value for the external coefficient was 190 W/m2.K for the three correlation calculations; this value was determined by fitting the calculated exit temperature to that observed experimentally. In addition to predictions based on correlations, the temperature profile predicted by the experimentally determined average overall heat-transfer coefficient is also shown. The brief plateau at the 1-m point in the tubing is the result of the fluid passing the pseudocritical point, where the fluid heat capacity reaches a maximum. Notable is the close agreement of all four profiles: at a given position, the maximum variation in predicted fluid temperature is 10 O C . Particularly encouraging is the close agreement of the experimentally derived profile with those predicted from correlations, especially the correlation of Swenson et al. Figure 7 shows the predicted internal heat-transfer coefficient from the three correlations as a function of bulk fluid temperature; conditions are the same as in Figure 6. All of the correlations predict asharp spike in the heattransfer coefficient at the pseudocritical temperature (383.5O C at 246 bar). The discontinuity in the prediction of the Perry et al. correlation occurs as the flow enters the transition region and the correlation changes. Interestingly, the Yamagata correlation doe0 not predict deteriorated heat transfer relative to Perry's correlation, while Swenson's correlation predicts significant deterioration. This discrepancy may reflect the conditions to which the correlations were fit; Yamagata fit his correlation to largely low heat flux conditions, while Swenson also included high heat fluxconditions where deterioration is more important. The predicted internal Coefficients are much higher than the fitted external coefficient, indicating that the external reistance is controlling in the heat-transfer process. Consequently,despite nearly order of magnitudevariations in estimates of the internal coefficient,the predicted overall heat-transfer coefficient is relatively insensitive to proper characterization of the details of the internal heat-transfer process. An additional important point is the magnitude of the external heat-transfer coefficient itself. Experimental

Figure 8. Overall heat-transfer coefficient as a function of distance in preheating tubing. For calculations using the correlations of Perry et al.,l9 Swenson et and Yamagata et al.?l the listed constant external coefficient was used. preheating tubing exit temperatures could be matched by the correlations only by assuming an external coefficient in the range 150-250 W/m2.K. At a given sand bath temperature, all correlations were able to match the experimental exit temperature (within - 5 "C)using the same value for the external coefficient;however, the fitted value of the external coefficient decreased with increasing sand bath temperature. At no time did the fitted value of the external coefficient reach the sand bath manufacturer's quoted value of 600 W/m2.K. One possible reason for this observed behavior is that the sand bath becomes less efficient at higher temperatures. All heat-transfer experiments were conducted at a constant sand bath fluidization rate, as measured at room temperature, of 15 L/min of air. The fluidization rate within the bath would thus increase as the bath temperature was increased. The increasing fluidization rate could result in a decreased sand bath heat-transfer coefficient at higher sand bath temperatures, although typically the sand bath heattransfer coefficient increases with increasing bath temperature.22 Figure 8 shows the calculated overall heat-transfer coefficient U, as a function of distance in the preheating tubing; conditions are again the same as in Figures 6 and 7. The experimentally derived coefficient is by definition constant over the length of the tubing. Significantly, the overall coefficients based on heat-transfer correlations are also approximately constant over the length of the tubing, within about &lo%. Predicted values of the overall coefficient are on the order of the external coefficient, further demonstratingthat the external coefficient controls the heat transfer. As a result of the controlling external resistance, U, is only weakly dependent on the internal coefficient, and the experimental assumption of a constant U , is borne out by the correlations. The technique for deriving U, experimentally thus appeared valid, and the experimental values for U, were used in calculating the temperature profiles within the preheating tubing. Typical results are shown in Figure 9 for a variety of sand bath temperatures (Tf& Fluid temperature is presented as a function of time rather than distance, as time is the more important variable for our purposes. The curves terminate when the fluid exits the (22) Kunii, D.; Levenspiel, 0.Fluidization Engineering; Robert E. Krieger Publishing Co., Inc.: New York, 1977.

Energy & Fuels, Vol. 6, No. 5,1992 593 Tfsb = 550°C 500

-

TISO= 500°C

~

.:/ :. ;

. .

,/Tts0- 450%

,

Tlro = 400°C

Inlet Temperature = 25°C Tubing Inner Diameter = 1 08 mm Heated Tubing length = 2.8m TISO Sand Bath Temperature

100

-

I

0

2

4

6

10 12 Time in Preheater Tubing s

8

i I

14

i 16

18

Figure 9. Temporalvariation of fluid temperature in preheating tubing for various fluidized sand bath temperatures. Temperature profiles are based on the experimentally determined average overall heat-transfercoefficient;termination of profile indicates fluid has exited the preheating tubing.

preheating tubing and enters the reactor. Several features of Figure 9 are noteworthy. First, the residence time in the preheater is significant (9 s or longer). At low sand bath temperatures (low heat fluxes), the time in the preheating tubing near sand bath (reaction) temperature can be on the order of the reactor residence time; e.g., for Tfsb= 400 "C, the fluid in the preheating tubing is within 20 "C of the reactor temperature for about 6 s prior to entering the reactor. Second, as the sand bath temperature increases, the fluid is heated more rapidly and accelerates through the preheating tubing, reducing both the residence time in the tubing and the time spent near the reactor (sand bath) temperature. On the other hand, the feed fluid experiences higher temperatures in the preheater at higher sand bath temperatures, such that reactions may occur to a greater extent than at lower sand bath temperatures. For example, for sand bath temperatures of 450 and 550 "C, the fluid in the preheater is above 400 "C for about 2 s; however, at the higher sand bath temperature, the fluid experiences higher temperatures during that same 2-5 interval, and more reaction is likely to occur. The degree to which these profiles affect the extent of reaction during preheating cannot be inferred solely from the profiles, but rather must be calculated using an appropriate reaction rate expression. The experimental and computational heat-transfer studies have revealed that the sand-bath heat-transfer coefficient is significantly lower than expected. Part of our earlier inability to predict correctly the thermal history of the feeds was our use of the 600 W/m2-Kvalue for the bath coefficient. Helling's16 calculations consequently predicted much faster heating of the feeds than observed experimentally. As previously mentioned, in the earlier experiments the feeds exited the preheating tubing at temperatures 10-25 "C lower than the sand bath temperature. Using the new values for the sand bath (external) heat-transfer coefficient, it is possible to calculate an approximate temperature profile for the feed mixture within the original reactor. For these calculations we used the internal heat-transfer correlations of Perry et al. For the range of sand bath temperatures employed in the earlier study, and the corresponding range of bath heattransfer coefficients,we estimate that the reaction mixture was heated to within 1-2 "C of the sand bath temperature within the first meter of the reactor's 4.24-m length. This

estimate is somewhatrough, as an older, smaller sand bath was used in the earlier studies, and we cannot be sure that the coefficients derived for our sand bath can be directly applied to the earlier apparatus. Realistically, however, it is unlikely that heat-transfer performances in the two baths would differ very significantly, and we thus feel that our estimate is a reasonable one. The consequence of this heating of the reaction mixture is that the temperature at the entrance to the reactor, considered to be the reactor temperature in the earlier study, is not an accurate estimate of the actual temperature within the majority of the reactor. Rather, the sand bath temperature is more likely to be representative of the reactor temperature. On the basis of this assumption, we have adjusted the earlier data of Helling and Testerg to incorporate the sand bath temperature as the "true" reactor temperature. All of the data of Helling and Tester presented in this paper have been revised to reflect this adjustment.

Discussion One of the main objectives of this study was the quantification of the extent of reaction during preheating of the reactor feeds. We have chosen to approach this objective by calculating the thermal history of the feeds and subsequently using the calculated history to derive kinetic parameters. This method is admittedly somewhat indirect; a more obvious method might have been to measure the conversion in the preheating tubing experimentally, by modifying the apparatus to bypass the reactor. Unfortunately, conversions in the preheating tubing alone are likely to be quite low (