2636
Ind. Eng. Chem. Res. 2004, 43, 2636-2642
Production of Hydrogen from the Noncatalytic Partial Oxidation of Ethanol D. O. Christensen, P. L. Silveston, E. Croiset, and R. R. Hudgins* Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada N2J 3L2
An experimental study on the noncatalytic partial oxidation of ethanol to produce hydrogen is presented along with modeling of the oxidation kinetics. Experiments were conducted in a quartz integral reactor at temperatures ranging from 600 to 800 °C, for different oxygen/ethanol ratios and residence times. The experimental results were compared with those generated using the Marinov model for high-temperature ethanol oxidation from the ChemKin software package. For the temperature range studied, the Marinov model predicts trends but does not reproduce the experimental data accurately. Results suggest that the noncatalytic partial oxidation of ethanol is not an attractive method for hydrogen production. Introduction Different methods of hydrogen production must be developed in order for hydrogen to be a feasible alternative to fossil fuels. One solution is to produce hydrogen from an alcohol, such as ethanol. Ethanol is a renewable resource because it can be made from waste biomass. When burned, its greenhouse gas emissions are far lower than fossil fuels because, over the life cycle of the biomass and the fuel, the carbon is trapped in a virtually closed loop. Ethanol also has the advantages that it is already in use as an additive to gasoline and that worldwide production is increasing. The issue becomes how to efficiently extract the hydrogen from this fuel. Over the last decade, several techniques have been discussed in the literature. Most of the publications present research on steam reforming1-6 and autothermal reforming (also referred to as oxidative steam reforming),10 while there appear to be only a few articles on catalytic partial oxidation.7-9 Noncatalytic partial oxidation is mentioned only in passing, and there appear to be no published results and, hence, no baseline with which to compare the catalytic studies. This paper presents results from an experimental study on the behavior of noncatalytic partial oxidation of ethanol. It also provides a comparison of the results with those predicted using the Marinov model for high-temperature ethanol oxidation.11 Experimental Apparatus and Procedure Overview of the Apparatus. The schematic of the experimental apparatus used in this study is shown in Figure 1. Air and ethanol flow through separate preheaters. In the ethanol preheater, the liquid ethanol is vaporized. The streams are then combined just before the reactor, which is located in a tubular furnace. Because the objective of this study was to compare the performance of a noncatalytic partial oxidation reactor at various operating conditions with the results obtained from the Marinov model, an integral reactor was used. The quartz reactor was 38 cm long with a volume of 180 mL. Quartz was used to avoid catalytic effects that are common with metal-walled reactors. * To whom correspondence should be addressed. E-mail:
[email protected].
Product gas flowed through heated tubing to prevent condensation of vapors prior to the gas chromatograph [GC; SRI model 8610, thermal conductivity (TCD) and flame ionization (FID) detectors]. A valve on the main tubing line was adjusted manually to control the system pressure to 69 kPa (g). This was done to ensure that product gas flowed through the sample loop, which was made of much smaller diameter tubing and had a higher pressure drop than the main tubing. The GC was connected to a PC for data collection and treatment. Sample Analysis. Two packed columns (a Porapak Q and a molecular sieve 5A) were used simultaneously in the GC in order to detect all of the compounds of interest. Because CO2 permanently adsorbs on the molecular sieve column, the GC columns were connected to a 10-port Valco valve with a 50-µL sample loop in order that their relative sequence could be reversed. Thus, a gas sample was first injected from the sample loop into the Porapak column. The light compounds (H2, CH4, etc.) passed quickly through and onto the molecular sieve. Just as CO2 was about to elute from the Porapak, the sequence of the columns was reversed. This column sequence reversal allowed CO2 to pass directly to the detectors without passing through the molecular sieve. Thus, the first response on the chromatograms was always CO2. The light compounds that had already passed on to the molecular sieve left that column and passed through the Porapak column once more before reaching the detectors. The retention time of ethanol was excessive, so for our experiments, temperature programming was used. Figures 2 and 3 show sample chromatograms for the TCD and FID, respectively. As can be seen in Figure 2, water and nitrogen coelute. Therefore, a calculation was developed to estimate their concentrations. We assumed that all carbonbearing species were measured. Consequently, measurement of both air and ethanol flows permits the calculation of the molar flow rate of the product gas. The water concentration could now be obtained from an oxygen balance. Because a number of assumptions are involved in this calculation, its accuracy was verified experimentally. About half of the experiments were run for an extended duration. During the extended time, GC conditions were modified to measure directly the nitrogen concentration.
10.1021/ie0303653 CCC: $27.50 © 2004 American Chemical Society Published on Web 04/17/2004
Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004 2637
Figure 1. Experimental apparatus.
Figure 2. Sample TCD chromatogram.
Figure 3. Sample FID chromatogram.
At these conditions, only nitrogen could be determined because of co-eluting peaks. The calculated nitrogen concentrations were compared to those measured directly. It was found that they agreed to within (2 vol % (absolute). Water concentrations were calculated as described above. Procedure. The furnace, preheaters, and GC were set to their respective desired temperatures with the ethanol pump and the air mass flow controller in their off positions. GC detectors were turned on and allowed to stabilize, after which the desired flow rates were set on the air mass flow controller and the ethanol pump. The system pressure was manually controlled at 69 kPa
(g). Data collection began after the system reached steady state. For sampling, the sample loop valve was switched to the intake position, while the GC data acquisition software was simultaneously activated. At 48 s, this valve was switched to the sampling position. The FID was then re-ignited because the flame was extinguished by the valve changes. At the completion of a sample injection, the chromatograms were saved in data files and the peak areas and retention times recorded. A sample could be taken every 45 min. Five samples were taken for each experiment to confirm that the system
2638 Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004
Figure 4. Sample temperature profiles.
was at steady state and to indicate the reproducibility of the samples collected. Each experiment was replicated. The data from each set of replicates were pooled for each species and standard deviations calculated. Data reported are average values. Further details of the experimental apparatus and procedure are given by Christensen.12 Reactor Temperature Profiles. During preliminary studies, it was discovered that a thermocouple inside the reactor, to control the furnace temperature, caused significant carbon formation on the thermocouple itself, likely as a result of the catalytic activity of its nickel-containing sheath. To avoid this problem, the thermocouple was moved outside of the reactor. Thus, it was not possible to measure temperatures directly within the reactor. Instead, temperature profiles were estimated by controlling the furnace with the thermocouple on the outside of the reactor while incrementally moving a second thermocouple through the inside of the reactor in the absence of a chemical reaction. The temperature, pressure, and total flow gas flow rate were in the appropriate range for actual experiments, but no ethanol was present; however, the air flow was increased appropriately to account for differences in the volumetric flow rate due to reaction. This estimation method should work well as long as flows are dilute so that the heat release or consumption is small. A sample temperature profile is given in Figure 4. It shows that the temperature profile is not uniform. There are two reasons for this nonuniformity. First, because of equipment limitations and to prevent cracking ahead of the reactor, the preheaters raised the temperature of the reactants to only about 175 °C. Second, heat-loss end effects caused the temperature to drop near the reactor exit. Thus, the temperature profile of Figure 4 shows that, at 800 °C, the gases are slow to heat and quick to cool. This means that the temperature is within (10 °C of 800 °C over about one-third of the reactor length starting about halfway through the reactor. Thus, the temperatures reported in Figures 5-11 represent the target temperatures of the furnace. Experimental Variables. The variables studied were the reactor temperature (T; which was taken to be the measured furnace temperature), the residence time (τ), and the oxygen/ethanol molar ratio (r). Reactor temperatures were 600, 700, and 800 °C. Residence
times were 2 and 4 s, and the oxygen/ethanol molar ratios were 0.5 and 1. The residence time was defined as the time the reactants remained in the reactor if the reactor were at a uniform temperature and with no chemical reaction. Results Observations. During experiments conducted at 600 °C, it was observed that oscillating blue (and occasionally orange) flames were visible and caused pressure fluctuations. At longer residence times, these flames stabilized and pressure fluctuations disappeared. At shorter residence times (i.e., higher flow rates), the flames were unsteady and pressure fluctuations were more severe. These fluctuations are thought to be the primary reason for the variability in the data taken at low temperatures. This variability becomes progressively less with increasing temperature. Flames were not observed for experiments at T ) 700 or 800 °C, and pressures were stable. Species detected in the product gas were CO2, H2, H2O, CH4, CO, C2H4, C2H6, and C2H5OH. No solid carbon formation was observed during any experiments. Mass balances were somewhat inaccurate. From 30 different runs, the hydrogen estimated from the identified products was underestimated at 91.6% with two standard deviations of 8.4%. This is likely due to the fact that the two unknown peaks were components containing hydrogen. Carbon balances were within about (10%. Experimental Results. Figure 5 shows experimental results for varying reactor temperatures and residence times for an oxygen/ethanol ratio (r) of 0.5. Figure 6 shows experimental results for an oxygen/ethanol ratio of 1.0. The maximum hydrogen concentration of 14% occurred at T ) 800 °C, τ ) 2 s, and r ) 0.5. In most cases, ethanol was completely converted. The lowest conversion of ca. 98% occurred at the lowest temperature, T ) 600 °C, and the lowest oxygen/ethanol ratio, r ) 0.5. The residence time, whether τ ) 2 or 4 s, seemed to have no significant effect on ethanol conversion. A statistical analysis was performed on the experimental results, the results of which are summarized in Table 1. A designation of “positive” means that increasing the variable increases the given species concentra-
Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004 2639
Figure 5. Experimental results for r ) 0.5 and τ values of 2 and 4 s.
Figure 6. Experimental results for r ) 1.0 and τ values of 2 and 4 s. Table 1. Summary of Qualitative Dependence of the Experimental Variables on the Species Concentrations variable species
temperature
residence time
oxygen/ethanol ratio
hydrogen carbon monoxide carbon dioxide methane water ethylene ethane
positive positive positivea not significanta not significanta not significanta negative
negative negative not significant not significant not significant not significant not significant
negative negative positive negative positive negative negative
a This result is unexpected, probably because of experimental error or variability.
tion, while “negative” means the opposite. Hydrogen concentrations were reduced by increasing the residence times and oxygen/ethanol ratios. It is noteworthy that temperature did not have a significant effect on H2O, CH4, and C2H4 concentrations. Equilibrium considerations, discussed below, suggest that the temperature should have a negative effect on these species. Perhaps variability in the data may have masked the effect.
Discussion Equilibrium. Concentration curves assuming thermodynamic equilibrium were constructed using minimization of free energy.3,13 As shown in Figures 7 and 8, ethane, ethylene, and ethanol should not be observed at equilibrium and, indeed, were not detected. These figures compare the experimental data at r ) 0.5 and 1.0, respectively, to the equilibrium curves. Experimental data are plotted for the temperature at the exit of the reactor as given by the temperature profile discussed earlier. As can be seen, the experimental data generally do not follow the equilibrium curves. The ChemKin kinetic package was used to model the results for residence times much greater than those used in the actual experiments. The results indicate that there is a rapid reaction within the first second of the residence time in which oxygen is consumed, followed by a much slower re-equilibration among the product species. The measured products are therefore not at equilibrium. It was observed from simulation results between 600 and 800 °C that CO and hydrogen mole fractions pass through local maxima as a function of the residence time. This might explain in part why some of
2640 Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004
Figure 7. Experimental data compared to equilibrium at r ) 0.5 and τ values of 2 and 4 s.
Figure 8. Experimental data compared to equilibrium at r ) 1.0 and τ values of 2 and 4 s.
the species cross over the equilibrium curves. For a species such as CO2, which is relatively close to the equilibrium curves, experimental error might be invoked as an explanation of crossover. Yield. The maximum possible hydrogen yield is 3 mol/ mol of ethanol. This is given by stoichiometric equations (1) and (2) for ethanol partial oxidation.
1 C2H5OH + O2 f 2CO + 3H2 2
(1)
3 C2H5OH + O2 f 2CO2 + 3H2 2
(2)
The yields obtained experimentally are only about onetenth of the maximum even though the ethanol conversion is always close to 100%. The average yields as a percentage of the maximum for the various experimental conditions are given in Figure 9. There are two major reasons for the low hydrogen yield. In the first place, ethylene, ethane, and methane are present in the product gas mixture. These are, effectively, hydrogen “sinks”, and although these concentrations are small, there are four hydrogen atoms per ethylene molecule,
six for ethane, and four for methane, which means that there is a significant amount of hydrogen bound up in these species. Second, the amount of water present in the product gas mixture is significantly above that predicted by equilibrium. This implies that a significant portion of the hydrogen produced is consumed by further oxidation. These two observations indicate that the noncatalytic partial oxidation of ethanol is not very selective to hydrogen production. As a result, noncatalytic partial oxidation is not attractive for use as a technique for hydrogen production. A significant amount of CO is generated, as the above figures indicate, so that noncatalytic partial oxidation can produce a CO-rich synthesis gas. The partial oxidation of ethanol could also generate a gaseous fuel that might be used for cold startup of an ethanol-fueled vehicle. Startup of such vehicles is difficult in cold climates.12 Modeling. We attempted to use the Marinov model for high-temperature ethanol oxidation to predict the results of partial oxidation. The Marinov model, described in detail elsewhere,11 is comprehensive: it incorporates 57 species and over 370 elementary reactions. Using the ChemKin suite of applications, the Marinov model was solved at each of the various
Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004 2641
Figure 9. Percentage of stoichiometric yield.
Figure 11. Model comparisons for CH4, C2H4, and C2H6. Figure 10. Model comparisons for CO2, H2, CO, and EtOH.
experimental conditions, using either a plug flow or a premixed flame model. The reactor temperature was described in different cases by an isothermal, adiabatic, and customized model using the temperature profile measured experimentally. We observed that the experimental data are best predicted by a plug-flow reactor model with either an isothermal temperature profile or the temperature profile measured experimentally. Figures 10 and 11 compare the model results at r ) 0.5 and τ ) 4 s with experimental data. As can be seen from these figures, the models predict the trends quite well but do not accurately reproduce the experimental data. Also, they tend to underpredict H2 and CO while overpredicting CH4 and C2H4. From the equilibrium curves, H2 and CO are expected to increase with temperature, while CH4 and C2H4 are expected to decrease. This suggests the possibility that the models are under- or overpredicting the concentrations because the temperatures taken for the data are higher than those assumed for the models. A sensitivity analysis was conducted on the rate of hydrogen production using the SENKIN subprogram of the ChemKin modeling software. Hydrogen production is most heavily influenced by dehydrogenation reactions. The reactions favored tended to differ for different r values but were consistent across the temperature range. Predominant reactions were dehydrogenation of
formaldehyde and acetaldehyde. At r ) 0.5, dehydrogenation of ethanol was important, while at r ) 1.0, dehydrogenation of ethylene was important. Considering the predicted concentrations of these precursors suggests that the aldehydes react very quickly and are always completely consumed, whereas ethanol and ethylene react more slowly and, as a result, appear in the product gas. Conclusions During the partial oxidation of ethanol, the following species were detected: CO2, H2, H2O, CH4, CO, C2H4, C2H6, and C2H5OH. The major species were CO, H2, and CH4, while CO2, C2H4, and C2H6 were present in small amounts. The selectivity to hydrogen was low even at nearly complete conversions of ethanol. A cause of low selectivity is further oxidation to water. Thus, noncatalytic partial oxidation of ethanol is not an attractive method for hydrogen production. Detailed reaction kinetics using the high-temperature Marinov model for ethanol oxidation between 600 and 800 °C predicts only general trends in the experimental data. According to the model, hydrogen arises through dehydrogenation of oxygenated species and ethylene. Acknowledgment The authors thank the Natural Sciences and Engineering Research Council (NSERC) of Canada for their
2642 Ind. Eng. Chem. Res., Vol. 43, No. 11, 2004
support. The authors also thank Drs. S. and A. Therdthianwong from King Mongkut’s University of Technology Thonburi for their valuable assistance and advice in this project. Nomenclature EtOH ) ethanol r ) oxygen/ethanol molar ratio T ) furnace temperature (°C) Greek Symbol τ ) ideal residence time (s)
Literature Cited (1) Vasudeva, K.; Mitra, N.; Umasankar, P.; Dhingra, S. C. Steam Reforming of Ethanol for Hydrogen Production: Thermodynamic Analysis. Int. J. Hydrogen Energy 1996, 21, 13. (2) Ioannides, T. Thermodynamic Analysis of Ethanol Processors for Fuel Cell Applications. J. Power Sources 2001, 92, 17. (3) Garcia, E. Y.; Laborde, M. A. Hydrogen Production by the Steam Reforming of Ethanol: Thermodynamic Analysis. Int. J. Hydrogen Energy 1991, 16, 307. (4) Breen, J. P.; Burch, R.; Coleman, H. M. Metal-Catalysed Steam Reforming of Ethanol in the Production of Hydrogen for Fuel Cell Applications. Appl. Catal., B 2002, 39, 65. (5) Galvita, V. V.; Semin, G. L.; Belyaev, V. D.; Semikolenov, V. A.; Tsiakaras, P.; Sobyanin, V. A. Synthesis Gas Production By Steam Reforming of Ethanol. Appl. Catal., A 2001, 220, 123.
(6) Emonts, B.; Bøgild Hansen, J.; Schmidt, H.; Brube, T.; Ho¨hlein, B.; Peters, R.; Tschauder, A. Fuel Cell Drive System with Hydrogen Generation in Test. J. Power Sources 2000, 86, 228. (7) Mitchell, W. L.; Thijssen, J. H. J.; Bentley, J. M.; Marek, N. J. Development of a Catalytic Partial Oxidation Ethanol Reformer for Fuel Cell Applications. Proceedings of the 1995 SAE Alternative Fuels Conference; SAE Technical Paper Series 952761; SAE: Warrendale, PA, 1995; p 209. (8) Recupero, V.; Pino, L.; Di Leonardo, R.; Lagana, M.; Maggio, G. Hydrogen Generator, via Catalytic Partial Oxidation of Methane for Fuel Cells. J. Power Sources 1998, 71, 208. (9) Ahmed, S.; Krumpelt, M.; Kumar, R.; Lee, S. H. D.; Carter, J. D.; Wilkenhoener, R.; Marshall, C. Catalytic Partial Oxidation Reforming of Hydrocarbon Fuels, Conference Paper, Fuel Cell Seminar, Palm Springs, CA, Nov 16-19, 1998. (10) Rampe, T.; Heinzel, A.; Vogel, B. Hydrogen Generation from Biogenic and Fossil Fuels by Autothermal Reforming. J. Power Sources 2000, 86, 536. (11) Marinov, N. M. A Detailed Chemical Kinetic Model for High-Temperature Ethanol Oxidation. Int. J. Chem. Kinet. 1999, 31, 183. (12) Christensen, D. O. Production of Hydrogen from the NonCatalytic Partial Oxidation of Ethanol. MASc Thesis, University of Waterloo, Waterloo, Ontario, Canada, 2003. (13) Smith, J. M.; Van Ness, H. C.; Abbott, M. M. Introduction to Chemical Engineering Thermodynamics, 5th ed.; McGrawHill: New York, 1996.
Received for review April 28, 2003 Revised manuscript received February 18, 2004 Accepted March 1, 2004 IE0303653