Oxidation of Simple Alcohols in Supercritical Water III. Formation of

---

86

Ind. Eng. Chem. Res. 2001, 40, 86-93

Oxidation of Simple Alcohols in Supercritical Water III. Formation of Intermediates from Ethanol Steven F. Rice* and Eric Croiset† Combustion Research Facility, Sandia National Laboratories, MS-9052, P.O. Box 969, Livermore, California 94551-0969

Raman spectroscopy is used as an in situ diagnostic to measure the oxidation of ethanol by oxygen in supercritical water. An elementary reaction mechanism based on the work of Marinov is shown to predict accurately many of the experimental observations. Experimental measurements are reported at 24.5 MPa over a temperature range of 410-470 °C in supercritical water with reaction times ranging from 0.5 to 3.0 s. Concentrations of ethanol, acetaldehyde, formaldehyde, methanol, carbon monoxide, carbon dioxide, and hydrogen peroxide are measured as functions of time and temperature. The data show that the formaldehyde is the primary stable organic intermediate. An elementary reaction mechanism, modified for supercritical water conditions and supplemented with key methylperoxyl reactions, is used to interpret the observations. The experimental data are consistent with the purely radical chain oxidation process represented by this mechanism. Analysis of the mechanism identifies the primary oxidation pathway proceeding through acetaldehyde with oxidation routes involving initial abstraction of the hydroxyl hydrogen or a hydrogen atom from the secondary carbon. A pathway originating from H-abstraction from the methyl group of the ethanol molecule contributes to the overall conversion to a lesser degree. Introduction The application of supercritical water oxidation (SCWO) as a waste treatment technology is in a transition from laboratory development to pilot-scale testing at a number of installations around the world. This progress has been driven by technological innovations in reactor design that have occurred, in part, because of a better understanding of the kinetics of the oxidation process. Most kinetics studies in this field have generally employed chromatographic analysis of liquid and gaseous effluents from small tubular reactors and subsequent reduction of these data into global rate expressions. Over a decade ago, a significant result of these early studies pointed to the potential applicability of elementary reaction kinetic modeling of the complete oxidation process (Helling and Tester, 1987). Progress in the understanding of oxidation mechanisms has ranged from the success of a pioneering detailed approach on very simple species such as CO and H2 (Holgate and Tester, 1994a,b) to broad-based pathway descriptions for complicated organic compounds (Savage, 1999). Building on parallels to combustion systems, several groups have refined methane and methanol oxidation elementary kinetic models using established methods and kinetic parameters by adapting key elementary reactions to higher density (Alkam et al., 1996; Dagaut et al., 1996; Brock and Savage, 1995; Brock et al., 1996; Savage et al., 1998). These mechanisms generally produced good agreement for feed conversion on limited data sets but until recently (Savage et al., 2000) were typically not tested beyond reproducing disappearance rates of the initial feed species. * Corresponding author. † Present address: Dept. of Chemical Engineering, University of Waterloo, Ontario, Canada.

Three published papers describing the use of Raman spectroscopy as an in situ method for measuring kinetics and refining mechanistic descriptions under supercritical water conditions established a capability for measuring the concentration of stable intermediates at concentrations as low as 10-3 mol/L (Steeper et al., 1996; Rice et al., 1996; Hunter et al., 1996). This paper extends these methods to monitor a range of intermediates present during the oxidation of ethanol in supercritical water. These experimental results are compared to the predictions of an elementary reaction mechanism recently developed by Marinov (1999) specifically for ethanol oxidation, modified for SCWO conditions. The details of ethanol reaction kinetics in supercritical water oxidation systems are important for several reasons. First, an understanding of ethanol kinetics can provide a link between the now well-established methanol reaction pathways and empirical models of higher alcohols, alkanes, and aromatics for which some quantitative data are now available. A second reason is that, because ethanol is among the more reactive species to have been examined to date, it is well-suited as an initiating fuel for autothermal reactor designs (Haroldsen et al., 1996; Kodra and Balaktaiah, 1994). In addition, low-grade ethanol is inexpensive and therefore can result in significant operational savings as a feed supplement for SCWO applications for wastes that have low heating values. Finally, simple alcohols present a good test system for examining the effect of a small amount of a reactive compound on the overall conversion rate of kinetically robust species such as halogenated aromatics (Lin et al., 1998) and nerve agent components such as alkyl phosphonates (Bianchetta et al., 1999). Ethanol may prove to be an important additive to the waste feed in treatment systems designed for these applications.

10.1021/ie000372g CCC: $20.00 © 2001 American Chemical Society Published on Web 12/08/2000

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 87 Table 1. Species and Raman Line Positions Examined molecule

Raman line position (cm-1)

C2H5OH CH3OH CH2O CH3CHO CO CO2 HCOOH O2 H2O2

890, 2928 2843, 2944 2780 1718, 2730, 2917 2165 1389 2960 1571 874

Experimental Section The experimental methods and equipment used in this work are very similar to those detailed in earlier papers on methanol and 2-propanol (Hunter et al., 1996; Rice et al., 1996) with one distinction. In this work, ethanol is introduced into the oxidizing mixture using the injection method employed in our lab to examine hydrogen peroxide decomposition kinetics in supercritical water (Croiset et al., 1997). The reactor has been modified such that one preheated supercritical water line (0.477 cm i.d.) contains water and dissolved oxygen. The organic reactant (e.g., ethanol) is injected as a pure material directly into the oxidizer flow through a smalldiameter capillary. The fuel is less than 1% of the total flow and is preheated to the reaction temperature within milliseconds prior to being introduced into the bulk flow. The data were recorded with a system pressure of 24.5 MPa and reactant concentrations of 0.39 mol % ethanol and 1.41 mol % oxygen in the high-temperature flow of supercritical water. The fuel equivalence ratio, Φ, (defined as the molar ratio of ethanol to oxygen initially in the premixed system divided by the ratio of ethanol to oxygen needed for full conversion of the fuel to CO2 and water) was 0.83, indicating a slight excess of oxygen in the system. The methods used to convert the Raman scattering intensities determined experimentally to concentrations are identical to those reported previously (Hunter et al., 1996). Table 1 lists the Raman peak positions of the species discussed here. Raman data were recorded in a number of spectral regions chosen to monitor ethanol, hydrogen peroxide, oxygen, carbon monoxide, carbon dioxide, acetaldehyde, formaldehyde, methanol, and formic acid. The Raman intensity and spectral profile of all of the stable species were recorded independently at 430 °C and 24.5 MPa in supercritical water. All of the carbon-containing species, except formic acid (Yu and Savage, 1998), were determined to be stable relative to hydrolysis or pyrolysis on the reactor residence time scale of approximately 0.5 s needed to record good calibration data. Hydrogen peroxide is not stable on this time scale at this temperature; however, absolute concentrations of H2O2 could be determined on the basis of our earlier work (Croiset et al., 1997). Although evidence for the presence of formic acid was sought, none was found, making difficulties calibrating it irrelevant. Reaction times were calculated as described previously (Rice et al., 1996). Mass flow rates were maintained sufficiently high to produce Reynolds numbers in the range of 3000-12000, ensuring turbulent plug flow conditions. Experiments using slower flow rates produced erratic Raman measurements. Because of this restriction on the reactor’s available flow rates, to cover an adequate range of reaction times, the data were

recorded with the spectroscopic cell positioned at both 27.3 and 83.8 cm relative to the reactant mixing point. The major sources of error in all of the measurements are: (1) Raman intensity measurement accuracy; (2) Raman scattering calibration accuracy; and (3) calculated mass flow rate stability. The Raman intensity measurement accuracy is determined primarily by the signal strength and the magnitude of the dominant spectroscopic noise source, which in this case is the dark count on the CCD detector. However, it is important to recognize that the fractional statistical error in the integrated intensity of a given Raman band is much less than the RMS noise associated with a single pixel. Typically, the statistical error in an integrated peak was approximately 1000 counts, which corresponds to a few percent of the total integrated signal associated with the unreacted ethanol feed. The scatter in the data presented below is consistent with this estimate. Because the same error is present in determining the calibration to convert from Raman intensity to concentration, the absolute fraction reported for an individual species has a systematic error relative to the other species of up to a few percent of the magnitude of the measurement. These error sources are minor within the context of the conclusions of this paper. The greatest source of error in the data presented in the concentration vs time figures presented here originates from the instability of the total mass flow rate during the time required to make each species measurement. The system pressure was nominally fixed at 24.5 MPa by a downstream spring-actuated pressure regulator. However, over the course of a Raman measurement of a particular species lasting several minutes, this pressure can fluctuate as much as (0.8 MPa. This introduces a density fluctuation in the flow and an error in the calculated flow rate nearly proportional to this drift. Therefore, the greatest source of scatter in the data is not a result of error in intensity measurements, but rather, it originates from an error in the calculated reaction time than can be as great at (6% at 430 °C and 24.5 MPa. Fortunately, the results in this paper benefit from numerous systematic measurements, such that the point-to-point noise in the data resulting from reaction time accuracy does not significantly affect the interpretation. Results Figure 1a shows a Raman spectrum of the 890 cm-1 feature of ethanol in supercritical water at 24.5 MPa and 430 °C in the absence of an oxidizer. This feature is primarily due to the C-C stretching motion. Figure 1b shows the spectrum in the same region in the presence of oxygen after 1.5 s of reaction time. Much of the ethanol has been consumed; however, there remains an asymmetrical feature reflecting the presence of an intermediate, which was identified as hydrogen peroxide on the basis of previous work (Croiset and Rice, 1998). Figure 2 shows the Raman spectrum of unreacted ethanol in the 2900 cm-1 region and evidence of a mixture of species when oxygen is present at 410 °C after 3.5 s. The figure shows the presence of formaldehyde (2780 cm-1) and methanol (2843 cm-1) as intermediates in the reacting system. The band at 2880 cm-1 due to ethanol is still present but is weaker, and the feature at 2930 cm-1 is distorted and shifted, indicating the presence of bands appearing at approximately 2915 and 2945 cm-1. The latter is due to methanol.

88

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

Figure 3. Experimental concentration of ethanol as a function of time and temperature with the results of the SENKIN calculation at the four experimental temperatures. The notations “long “ and “short” refer to two positions (83.8 and 27.3 cm) of the Raman spectroscopic cell in the flow reactor. Note that there is good agreement between points recorded at similar reaction times at both rapid and slow flow rates corresponding to the long and short detection positions.

Figure 1. (a) Raman spectrum of ethanol in supercritical water at 24.5 MPa and 430 °C in the 900 cm-1 region. (b) Raman spectrum of a feature at 879 cm-1 produced during the oxidation of ethanol.

Figure 2. Raman spectrum in the 2900 cm-1 region of ethanol in supercritical water at 410 °C and after a reaction time of 3.5 s.

A complete set of tables listing reaction times and measured concentrations collected on this oxidation system over a range of temperatures is contained in this paper’s Supporting Information. The most important general observation from these tables is that nearly all of the carbon can be accounted for over the course of the reaction, within experimental error, by considering CO, CO2, methanol, formaldehyde, and acetaldehyde as the reaction products, except for several points at early times at 450 °C. Figure 3 shows the measured time dependence of the concentration of ethanol during oxidation by dissolved oxygen in supercritical water based on the intensity of the 890 cm-1 Raman band at four different reaction temperatures. Figure 3 also includes the results from the elementary reaction mechanism discussed later in this paper. The data are plotted as normalized concentrations defined as the observed concentration of ethanol

divided by the concentration in the initial feed. The conversion rate of ethanol is comparable to that of 1-propanol, faster than those of methanol and 2-propanol, and much faster than that of methane under the same conditions. As a point of reference, 1-propanol is 50% converted at 430 °C in 1.1 s, as is shown in the Supporting Information, and ethanol appears to be 50% converted at 430 °C in 1.25 s. 2-Propanol requires 1.6 s to be 50% converted at 430 °C (Hunter et al., 1996), and methanol requires more than 3.0 s for 50% conversion (Rice et al., 1996). Methane oxidation by oxygen requires approximately 6 min for comparable conversion at this concentration, temperature, pressure, and equivalence ratio in supercritical water (Steeper et al., 1996). These data on ethanol are not consistent with those reported by Helling and Tester (1988), which show a much lower overall conversion rate. However, those data were recorded at a much lower feed concentration of reactants. It is likely, but not proven here, that this oxidation mechanism, a branching radical chain propagation, will not exhibit first-order kinetics over a wide range of fuel feed concentrations in a fashion similar to our previous results on methanol. The higher feed concentrations used in this work are likely to react much more rapidly (Rice et al., 1996). Our earlier work has shown that formaldehyde is the only organic intermediate that is formed in appreciable concentration during methanol oxidation and that only a small amount of hydrogen peroxide is accumulated (Rice et al., 1996; Croiset and Rice, 1998). The oxidation of 2-propanol produces large amounts of acetone as the initial and dominant carbon-containing intermediate formed by the abstraction of the secondary hydrogen and subsequent loss of the hydroxyl hydrogen. This earlier work did not explore the presence of other intermediates; however, carbon balances determined in subsequent experiments reveal that only CO and CO2 appreciably accumulate in the 2-propanol system. These data on 2-propanol and results supporting the statements above on 1-propanol oxidation rates are contained in the Supporting Information accompanying this paper. Here, the goal was to follow the concentration of all of the major carbon-containing species over the entire oxidation process to form a more complete picture for ethanol as a prototypical system.

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 89

ing a careful quantitative assessment of the initial hydrogen abstraction rates by various reactive radicals including H, OH, HO2, and CH3. Both of these papers indicate that abstraction of the hydroxyl hydrogen atom results primarily in the production of formaldehyde and methyl radical as intermediates. Abstraction of one of the two secondary H atoms produces acetaldehyde and H atoms, and abstraction of one of the three primary H atoms produces ethylene and OH radicals. In all cases, these reactions produce intermediates having distinct structural differences from ethanol that are easy to identify in the Raman spectra. The measurements presented here in supercritical water reflect the relative importance of these different pathways under SCWO conditions. By starting with the Marinov mechanism, appropriately augmenting it to be suitable for SCWO conditions, and comparing with the data set, these details can be evaluated. All calculations were conducted using the SENKIN code (Lutz et al., 1988) that functions as a driver for the widely used CHEMKIN (Kee et al., 1996) suite of subprograms. SENKIN produces concentration vs time profiles for all of the species present in the system under isothermal conditions, as well as other useful interpretive information such as sensitivity coefficients and reaction pathway mass flux. Assuming that the experimental system undergoes ideal plug flow with no axial diffusion, SENKIN provides an appropriate representation of the evolution of the composition of the flow as a function of time, calculated at a fixed sampling position using the simple bulk linear velocity. Each individual reaction is represented by a rate constant expressed as Figure 4. (a) Evolution of methanol concentration as a function of time and temperature during the oxidation of ethanol for the same conditions as Figure 3. (b) Evolution of formaldehyde concentration as a function of time and temperature during the oxidation of ethanol.

Figure 4 shows the evolution of formaldehyde and methanol as functions of time and temperature in supercritical water for the same conditions as shown in Figure 3. It is clear that a significant amount of formaldehyde is formed prior to methanol in this system, suggesting that there is a mechanistic route to formaldehyde in ethanol oxidation that does not include a step that first forms methanol. This indicates that, although there are similarities in the fuel consumption profile for the higher alcohols relative to methanol, most notably an induction period followed by more rapid reaction, the reaction pathways are more complicated. Specifically, as Li et al. (1991) suggests in a qualitative sense, there are important C-C bond cleavage steps that can occur on the same time scale as H-atom abstraction. Reaction Mechanism In 1992, Norton and Dryer (1992) developed an elementary reaction model for the oxidation of ethanol and compared the predictions of the model to experimental results in a flow reactor at approximately 820 °C and atmospheric pressure. In that work, they establish the basis of a detailed mechanism and recognize that variation in the product spectrum depends on the site of initial hydrogen-atom abstraction. Recently, Marinov (1999) reported results using an elementary reaction mechanism to describe the oxidation of ethanol under high-temperature combustion conditions, includ-

k ) ATb exp(-Ea/RT)

(1)

where A is a preexponential factor having units of cm3 mol-1 s-1 K-b for bimolecular reactions and s-1 K-b for unimolecular decomposition reactions written in the high-pressure limit appropriate for the conditions addressed here. T is in Kelvin, b is unitless, and Ea is expressed in cal/mol. Note that this code uses an ideal gas equation of state for all of the species including water. In this temperature and pressure range, the true water density is significantly higher than is predicted using an ideal gas equation of state (EOS). Consequently, species volumetric concentrations from the known input mole fractions will be too small. To account for this, following Brock et al. (1995), these calculations use a modified system pressure. The assumed pressure is the pressure pure water would require using an ideal gas EOS at the experimental temperature to result in the density of real water. For example, at 430 °C and 24.5 MPa, water’s density is 6.55 mol/L, which corresponds to 38.3 MPa for an ideal gas. As a starting point, the predictions of the Marinov model are compared with the experimental conversion of ethanol at 430 °C and 24.5 MPa. The calculation predicts a conversion rate that is about a factor of 3 slower than observed. In addition, it shows the primary products of the reaction on this time scale to be CO and methane, which is contradictory to our experimental observations showing that the primary intermediates are methanol, formaldehyde, and CO, with considerable amounts of CO2 produced at later times. No methane is observed experimentally. Note that Marinov’s mechanism is designed to accommodate much higher temperature chemistry. Key

90

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

Table 2. Changes and Additions to Marinov Mechanism reaction

A

b

Ea

H2O2 ) OH + OH CH3 + O2 (+M) ) CH3O2 (+M) LOW/5.81E+25 -3.30 0.00 / CH3O2 + H2 ) CH3O2H + H CH3O2 + H ) CH3O + OH CH3O2 + O ) CH3O + O2 CH3O2 + OH ) CH3OH + O2 CH3O2 + HO2 ) CH3O2H + O2 CH3O2 + H2O2 ) CH3O2H + HO2 CH3O2 + CH3O2 ) CH3O + CH3O + O2 CH3O2 + CH3O2 ) CH3OH + CH2O + O2 CH3O2H ) CH3O + OH CH3OH + H ) CH3 + H2O CH3CO3 + CH3HCO w CH3CO3H + CH3CO CH3CO3H + CH3CO w CH3CO3 + CH3HCO CH3HCO + CH3O w CH3CO + CH3OH CH3CO + CH3OH w CH3HCO + CH3O CH3HCO + CH3O2 w CH3CO + CH3O2H CH3CO + CH3O2H w CH3HCO + CH3O2 CH3CO + O2 w CH3CO3 CH3CO3 w CH3CO + O2 CH3CO3 + HO2 w CH3CO3H + O2 CH3CO3H w CH3CO2 + OH CH3CO3H w CH3 + CO2 + OH CH3CO3 + CH3O2 w CH3CO2 + CH3O + O2 CH3CO3 + CH3O2 w CH3CO2H + CH2O + O2 CH3CO3 + HO2 w CH3CO2 + OH + O2 CH3CO3 + CH3CO3 w CH3CO2 + CH3CO2 + O2 CH3CO2 (+M) w CH3 + CO2 (+M) LOW/1.20E+15 0.0 12518.0 / CH3O2 + CH3 w CH3O + CH3O CH3O + CH3O w CH3O2 + CH3 CH3O2 + HO2 w CH3O + OH + O2 CH3O2H + OH w CH3O2 + H2O CH3O2 + H2O w CH3O2H + OH CH3O2H + OH w CH2O2H + H2O CH2O2H + H2O w CH3O2H + OH CH3O2H + CH3O w CH3O2 + CH3OH CH3O2 + CH3OH w CH3O2H + CH3O CH3O2H + CH3O w CH2O2H + CH3OH CH2O2H + CH3OH w CH3O2H + CH3O CH2O + CH3O w HCO + CH3OH HCO + CH3OH w CH2O + CH3O CH2O + CH3 w HCO + CH4 HCO + CH4 w CH2O + CH3 CH2O2H w CH2O + OH CH2O + OH w CH2O2H HCOOH + HO2 ) HOCO + H2O2 HCOOH + OH ) HOCO + H2O HCOOH + H ) HOCO + H2 HCOOH + CH3 ) HOCO + CH4 HOCO ) OH + CO HOCO ) H + CO2 HOCO + O2 ) CO2 + HO2 HOCO + HO2 ) CO2 + H2O2 HOCO + CH3O2 ) CO2 + CH3O2H CH2O + HO2 ) HCO + H2O2

2.28E+13 7.83E+08

0.00 1.2

43019.0 0.0

Croiset, 1997 Baulch et al., 1992

3.01E+13 9.64E+13 3.61E+13 6.03E+13 4.63E+10 2.41E+12 5.48E+10 2.19E+09 6.00E+14 1.00E+13 1.20E+11 1.99E+10 1.15E+11 3.02E+11 3.55E+09 5.02E+09 1.00E+10 2.88E+16 1.00E+12 3.98E+15 2.00E+14 1.81E+12 3.01E+11 1.00E+12 4.78E+12 3.00E+12

0.0 0.0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.0

26032.0 0.0 0.0 0.0 -2583.0 9936.0 -835.0 -3580.0 42327.0 5500.0 4900.0 10000.0 1280.0 18160.0 5050.0 10100.0 -2700.0 37300.0 0.0 40000.0 40150.0 0.0 0.0 0.0 0.0 16722.0

Brock and Savage, 1995 Brock and Savage, 1995 Brock and Savage, 1995 Brock and Savage, 1995 Brock and Savage, 1995 Brock and Savage, 1995 Baulch et al., 1992 Baulch et al., 1992 Baulch et al., 1992 Brock and Savage, 1995 Kaiser et al., 1986, 8 8 Kaiser et al., 1986, 9 9 Kaiser et al., 1986, 10 10 Kaiser et al., 1986, 13 Kaiser et al., 1986, 14 Kaiser et al., 1986, 15 Kaiser et al., 1986, 16 Kaiser et al., 1986, 17 Kaiser et al., 1986, 18 Kaiser et al., 1986, 19 Kaiser et al., 1986, 20 Kaiser et al. 1986, 21b

3.80E+12 2.00E+10 1.00E+12 3.23E+13 3.02E+13 2.51E+13 3.01E+13 7.07E+11 3.01E+13 7.07E+11 3.01E+13 1.15E+11 3.02E+11 1.00E+10 2.09E+10 3.98E+15 3.16E+13 2.40E+19 1.85E+07 6.06E+13 3.90E-07 4.09E+12 1.74E+12 8.73E+11 1.00E+12 1.00E+12 4.11E+04

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.50 0.5 0.00 1.00 -2.2 1.5 -0.35 5.8 0.53 0.307 0.00 0.00 0.00 2.5

-2600.0 0.0 0.0 1000.0 32800.0 1000.0 32800.0 4000.0 32800.0 4000.0 32800.0 1280.0 18160.0 6000.0 21140.0 23000.0 14570.0 14030.0 -962.0 2988.0 2200.0 33981.0 32928.0 0.0 0.0 0.0 10205.0

Kaiser et al., 1986, 23 23 Kaiser et al., 1986, 24 Kaiser et al., 1986, 30 30 Kaiser et al., 1986, 31 31 Kaiser et al., 1986, 32 32 Kaiser et al., 1986, 33 33 Kaiser et al., 1986, 83 83 Kaiser et al., 1986, 86 86 Kaiser et al., 1986, 153 153 Marinov, 1998 Marinov, 1998 Marinov, 1998 Marinov, 1998 Brock et al., 1996 Brock and Savage, 1995 Brock et al., 1996 Brock and Savage, 1995 Brock and Savage, 1995 Eiteneer et al., 1998

a

ref

Units are cal/mol, mol/cm3, s, K. b High-pressure limit by analolgy to CH3CO w CH3 + CO in Marinov, 1998.

reactions involving hydrogen peroxide and methylperoxyl radical are not incorporated because reactions involving these species play no role at higher temperatures. Therefore, six additions or modifications to the mechanism were made, including: (1) The H2O2 decomposition rate is determined by experimental measurements under supercritical water conditions using a highpressure limit with A ) 2.28 × 1013 s-1, b ) 0.00, and Ea ) 43019 cal/mol (Croiset et al., 1997). (2) The CH3O2 chemistry used by several authors for methane and methanol oxidation in supercritical water is included (Brock and Savage, 1995; Brock et al., 1996). (3) The direct reaction of CH3 with H2O to form methanol and H is included from Brock and Savage (1995). (4) All of

the reactions identified as key reactions from the intermediate-temperature acetaldehyde oxidation reaction mechanism of Kaiser et al. (1986) are included if not already considered by Marinov or in the CH3O2 reactions in point 2 above. (5) The H-abstraction reactions involving formic acid that are condensed in the original mechanism are expanded to include HOCO as a distinct intermediate. The conversion reactions of HOCO to CO and CO2 are taken from Brock et al. (1996). (6) The kinetic parameters for the removal of hydrogen from formaldehyde to form HCO are updated to those recommended by Eiteneer et al. (1998). Table 2 shows a complete list of changes and additions made to the Marinov mechanism. The new data reported in

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 91

Figure 6. Diagram showing the major pathways of the organic species during the oxidation of ethanol in supercritical water. The numbers refer to the reactions listed in Table 3.

Figure 5. Evolution of the relative concentration of key intermediates produced during the partial oxidation of ethanol in supercritical water by oxygen vs ethanol conversion. The concentrations are expressed as a fraction of the initial ethanol concentration. Temperature is 450 °C.

this paper were not used to generate any of the rate parameters used in these modifications. The sources of the parameters are listed in Table 2. Included in Figure 3 are the results of the new mechanism comparing the conversion rate of ethanol over the temperature range examined. The agreement for conversion between the calculations and the experimental data is significantly improved by the changes in the mechanism relative to the results of the original mechanism. The ethanol conversion rate is near the observed rate, and there is no significant production of methane. Figure 5 shows the concentrations of all of the species detected at 450 °C vs ethanol conversion compared with the model predictions. The results are plotted vs conversion of ethanol to remove confusion originating from the small discrepancy in the predicted and observed conversion rates. Figure 5a shows that the maximum concentrations predicted for methanol and formaldehyde are in fair agreement with the data, with a slightly greater amount of formaldehyde and a lesser amount of methanol predicted. However, the model predicts that formaldehyde is formed earlier in the oxidation process than is observed. The model and the data agree very well for the formation of acetaldehyde. The mechanism does not predict significant accumulation of any species other than those that are observed. Figure 5b shows the production rates of CO2, CO, and H2O2 at 450 °C. The model predicts a greater amount of H2O2 than is observed, but the production and

consumption time scales agree well. The model predicts that both CO and CO2 are formed later in the oxidation process. However, the model underpredicts the conversion rate of CO to CO2 under these conditions, with the data indicating that, at 98% conversion of ethanol, most of the carbon has also been fully converted to CO2. The model indicates that, at this temperature, a large amount of carbon remains as CO and does not oxidize to CO2 on this time scale. The rate of final conversion of CO to CO2 is characteristic of CO oxidation by oxygen in supercritical water with no organic present (Holgate and Tester, 1994a). The primary reaction pathways of the major carbon containing species as predicted by the elementary reaction mechanism at 450 °C are shown in Figure 6. The reactions noted by number in Figure 6 are listed in Table 3, where the relative carbon atom fluxes for the maximum ethanol conversion rate at t ) 1.13 s are listed. In Table 3, the total carbon flux is normalized to 1.0. Most of the ethanol is converted in the formation of acetaldehyde from abstraction of the secondary hydrogen by OH and HO2 followed by removal of the hydroyxl hydrogen by oxygen to form HO2. Some additional acetaldehyde is formed from the ethoxy radical that results from initial abstraction of the hydroxyl hydrogen. Acetaldehyde loses the aldehydic hydrogen to HO2 forming H2O2 and CH3CO, which decomposes to CO and CH3. A secondary pathway originates with the abstraction of the hydroxyl hydrogen by OH to form ethoxy radical, which decomposes to form CH3 and formaldehyde. A minor pathway results from the abstraction of a primary hydrogen by OH and capture by O2 to form HOC2H4O2. This species undergoes decomposition to form two formaldehyde molecules and OH. At this temperature, the formation of ethylene does not compete

92

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

Table 3. Key Reactions and Mole Flux

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

reaction

flux (mole fraction C)

C2H5OH + HO2 w CH3CHOH + H2O2 C2H5OH + OH w CH3CHOH + HO2 C2H5OH + OH w CH3CH2O + HO2 C2H5OH + OH w C2H4OH + HO2 CH3CHOH + O2 w CH3CHO + HO2 CH3CH2O + M w CH3CHO + H + M CH3CH2O + M w CH3 + CH2O C2H4OH + O2 w HOC2H4O2 HOC2H4O2 w 2CH2O + OH CH3CHO + HO2 w CH3CO + H2O2 CH3CHO + HO2 w HCOOH + CH3 CH3CO (+M) w CH3 + CO CH3 + O2 (+M) w CH3O2 (+M) CH3O2 + H2O2 w CH3O2H + HO2 CH3O2 + HO2 w CH3O +OH +O2 CH3O2H w CH3O +OH CH3O (+M) w CH2O + H (+M) CH2O + OH w HCO + H2O CH2O + HO2 w HCO + H2O2 HCO + O2 w CO + HO2 HOCO + O2 w CO2 + HO2 CH3CO2 (+M) w CO2 + HO2

0.28 0.24 0.33 0.13 0.52 0.09 0.13 0.13 0.13 0.37 0.09 0.36 0.33 0.20 0.11 0.20 0.26 0.26 0.21 0.47 0.02 0.03

with capture by O2. Thus, nearly all of the carbon is converted to CO, CH3, and CH2O. CH3 is not identified in the system because the CH3 oxidation pathway in this temperature range proceeds through rapid formation of CH3O2 and CH3O2H by association with O2 and subsequent hydrogen-atom exchange with H2O2. Both of these species form CH3O, which is mostly converted to CH2O. Formaldehyde is converted to CO and CO2, as has been discussed in the methanol SCWO mechanism literature by a number of authors. Some of the CH3O reacts with water, CH3O2H, acetaldehyde, or formaldehyde to form methanol, which accumulates in the system at early times until the supply of CH3O begins to fall. We conclude that the present model captures most of the important chemistry. We reemphasize that the greatest difference between the model and the data is that the model predicts that formaldehyde is formed earlier in the reaction and that less methanol and slightly more formaldehyde are formed than is observed experimentally. This may be due to a greater reactivity of water in this unusual environment in the reverse of the reaction CH3OH + OH ) CH3O + H2O than is provided in the mechanism. This is one of the main routes to CH3OH and would simultaneously slow the production of CH2O. The fact that the observed production rate of acetaldehyde is well matched by the model predictions suggests that the branching ratios for the hydrogenabstraction reactions in the original Marinov mechanism are appropriate for these conditions. Because of this good agreement, this work provides support for the quantitative accuracy for the rates of hydrogen abstraction by OH and HO2 for hydroxl, secondary, and primary hydrogen during the oxidation of organics in supercritical water. In future SCWO mechanism development, these rate parameters can be extended within the more complex schemes that have been proposed over the past decade to produce quantitative mechanisms for the oxidation of larger organic molecules. It is widely held by many that an important and stable intermediate in SCWO is acetic acid. If this were the case, then the oxidation of ethanol should probably form it. Acetic acid has strong Raman bands at 892 and

2944 cm-1. Unfortunately, the 2944 cm-1 band is obscured by the second methanol band at the same position, and the 892 cm-1 band is indistinguishable from the ethanol 890 cm-1 band. However, there is no persistent band at either position at times when most of the ethanol has been consumed but there is still a large amount of organic carbon in the system as evidenced by the presence of formaldehyde, methanol, and peroxide. In addition, the carbon balances are near unity. The data indicate that acetic acid is not formed during SCWO of ethanol in the temperature range that has been explored, but rather that the radical chain mechanism as described above represents the oxidation path. This mechanism does not provide a path to acetic acid. These observations lead us to suggest that, in the SCWO of large organics, conversion proceeds through aldehydes and alkyl peroxides. Carbon chains are broken stepwise by reactions analogous to reaction 13 in Table 3 and not by the formation of carboxylic acids. Conclusions In situ Raman spectroscopic measurements are used to identify all of the major carbon-containing species present during the oxidation of ethanol by oxygen in supercritical water. Overall conversion rates of ethanol to products are predicted by a detailed elementary reaction mechanism that also predicts many of the quantitative measurements of the formation and consumption of intermediates. The model predictions show that the primary carbon-carbon bond cleavage route involves the formation of acetaldehyde that subsequently forms CO and CH3 following abstraction of the aldehydic hydrogen and decomposition of CH3CO. It is possible that a generalization of these pathways can describe the oxidation of much more complicated species. Acknowledgment The authors acknowledge the support of the DoD/ DOE/EPA Strategic Environmental Research and Development Program (SERDP). Supporting Information Available: Complete set of tables listing reaction times and measured concentrations collected on the oxidation of ethanol by oxygen in supercritical water over a range of temperatures. This material is available free of charge via the Internet at http://pubs.acs.org. Literature Cited (1) Alkam, M. K.; Pai, V. M.; Butler, P. B.; Pitz, W. J. Methanol and Hydrogen Oxidation Kinetics in Water at Supercritical States. Combust. Flame 1996, 106, 110-130. (2) Baulch, D. L.; Cobos, C. J.; Cox, R. A.; Esser, C.; Frank, P.; Just, T.; Kerr, J. A.; Pilling, M. J.; Troe, J.; Walker, R. W.; Warnatz, J. Evaluated Kinetic Data for Combustion Modeling. J. Phys. Chem. Ref. Data 1992, 21, 411-429. (3) Bianchetta, S.; Li, L. X.; Gloyna, E. F. Supercritical Water Oxidation of Methylphosphonic Acid. Ind. Eng. Chem. Res. 1999, 38, 2902-2910. (4) Brock, E. E.; Oshima, Y.; Savage, P. E.; Barker, J. R. Kinetics and Mechanism of Methanol Oxidation in Supercritical Water. J. Phys. Chem. 1996, 100, 15834-15842. (5) Brock, E. E.; Savage, P. E. Detailed Chemical Kinetics Model for Supercritical Water Oxidation of C1 Compounds and H2. AIChE J. 1995, 41, 1874-1888.

Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 93 (6) Croiset, E.; Rice, S. F.; Hanush, R. G. Hydrogen Peroxide Decomposition in Supercritical Water. AIChE J. 1997, 43, 23432352. (7) Croiset, E.; Rice, S. F. Direct Observation of H2O2 during Alcohol Oxidation by O2 in Supercritical Water. Ind. Eng. Chem. Res. 1998, 37, 1755-1760. (8) Dagaut, P.; Cathonnet, M.; Boettner, J.-C. Chemical Kinetic Modeling of the Supercritical Water Oxidation of Methanol. J. Supercrit. Fluids 1996, 98, 33-42. (9) Eiteneer, B.; Yu, C. L.; Goldenberg, M.; Frenklach, M. Determination of Rate Coefficients for Reactions of Formaldehyde Pyrolysis and Oxidation in the Gas Phase. J. Phys. Chem. A 1998, 102, 5196-5205. (10) Haroldsen, B. L.; Ariizumi, D. Y.; Mills, B. E.; Brown, B. G.; Rousar, D. C. Transpiring Wall Supercritical Water Oxidation Test Reactor Design; Technical Report SAND96-8213; Sandia National Laboratories: Albuquerque, NM, 1996. (11) Helling, R. K.; Tester, J. W. Oxidation of Carbon Monoxide in Supercritical Water. Energy Fuels 1987, 1, 417-423. (12) Helling R. K.; Tester, J. W. Oxidation of Simple Compounds and Mixtures in Supercritical Water: Carbon Monoxide Ammonia and Ethanol. Environ. Sci. Technol. 1988, 22, 13191324. (13) Holgate, H. R.; Tester, J. W. Oxidation of Hydrogen and Carbon Monoxide in Sub- and Supercritical Water: Reaction Kinetics, Pathways, and Water-Density Effects. 1. Experimental Results. J. Phys. Chem. 1994a, 98, 800-809. (14) Holgate, H. R.; Tester, J. W. Oxidation of Hydrogen and Carbon Monoxide in Sub- and Supercritical Water: Reaction Kinetics, Pathways, and Water-Density Effects. 2. Elementary Reaction Modeling. J. Phys. Chem. 1994b, 98, 810-822. (15) Hunter, T. B.; Rice, S. F.; Hanush, R. G. Raman Spectroscopic Measurement of Oxidation in Supercritical Water II. Conversion of Isopropyl Alcohol to Acetone. Ind. Eng. Chem. Res. 1996, 35, 3984-3990. (16) Kaiser, E. W.; Westbrook, C. K.; Pitz, W. J. Acetaldehyde Oxidation in the Negative Temperature Coefficient Regime: Experimental and Modeling Results. Int. J. Chem. Kinet. 1986, 18, 655-688. (17) Kee, R. J.; Rupley, F. M.; Meeks, E.; Miller, J. A. CHEMKIN-III: A Fortran Chemical Kinetics Package for the Analysis of Gas-Phase Chemical and Plasma Kinetics; Technical Report SAND96-8216; Sandia National Laboratories: Albuquerque, NM, 1996.

(18) Kodra D.; Balakotaiah, V. Autothermal Oxidation of Dilute Aqueous Wastes Under Supercritical Conditions. Ind. Eng. Chem. Res. 1994, 33, 575-580. (19) Li, L.; Chen, P.; Gloyna, E. F. Generalized Kinetic Model for Wet Oxidation of Organic Compounds. AIChE J. 1991, 37, 1687-1697. (20) Lin, K. S.; Wang, H. P.; Li, M. C. Oxidation of 2,4-Dichlorophenol in Supercritical Water. Chemosphere 1998, 36, 20752083. (21) Lutz, A. E.; Kee, R. J.; Miller, J. A. SENKIN: A Fortran Program for Predicting Homogeneous Gas-Phase Chemical Kinetics and Sensitivity Analysis; Technical Report SAND87-8248; Sandia National Laboratories: Albuquerque, NM, 1988. (22) Marinov, N. M. A Detailed Chemical Kinetic Model for High-Temperature Ethanol Oxidation. Int. J. Chem. Kinet. 1999, 31, 183-220. (23) Norton, T. S.; Dryer, F. L. An Experimental and Modeling Study of Ethanol Oxidation Kinetics in an Atmospheric Pressure Flow Reactor. Int. J. Chem. Kinet. 1992, 24, 319-344. (24) Rice, S. F.; Hunter, T. B.; Ryde´n, A° . C.; Hanush, R. G. Raman Spectroscopic Measurement of Oxidation in Supercritical Water I. Conversion of Methanol to Formaldehyde. Ind. Eng. Chem. Res. 1996, 35, 2161-2171. (25) Savage P. E.; Yu, J.; Stylski, N.; Brock, E. E. Kinetics and Mechanism of Methane Oxidation in Supercritical Water. J. Supercrit. Fluids 1998, 12, 141-153. (26) Savage, P. E. Organic Reactions in Supercritical Water. Chem. Rev. 1999, 99, 603-621. (27) Savage, P. E.; Rovira, J.; Stylski, N.; Martino, C. J. Oxidation Kinetics for Methane/Methanol Mixtures in Supercritical Water J. Supercrit. Fluids 2000, 17, 155-170. (28) Steeper, R. R.; Rice S. F. Kinetics Measurements of Methane in Supercritical Water. J. Phys. Chem. 1996, 100, 184189. (29) Yu, J.; Savage, P. E. Decomposition of Formic Acid under Hydrothermal Conditions. Ind. Eng. Chem. Res. 1998, 37, 2-10.

Received for review March 30, 2000 Revised manuscript received August 21, 2000 Accepted September 29, 2000 IE000372G