Steam reforming of technical bioethanol for hydrogen production 
A 10 wt% Ni/MgAl2O4 and a 1 wt% K doped 10 wt% Ni/MgAl2O4 catalyst were prepared by the incipient wetness impregnation technique. An industrial high surface area spinel (MgAl2O4) with a pore volume of 427 mL kg_1 and a specific surface area of 72m2 g_1 received from Haldor Topsøe A/S was used as support material. The spinel tablets were crushed to a size fraction of 300–710 mm and dried for 1 h at 100 _C. The solution for the impregnation was prepared by dissolving the desired amount of the different metal compounds, as nitrate salts, in water under moderate heating. After impregnation, the catalysts were dried at 100 _C for 2 h and finally calcined at 500 _C for 4 h. A 2 wt% Ru/MgAl2O4 catalyst was prepared similarly from a nitrosyl nitrate aqueous solution [Ru(NO) (NO3)3], leaving out the calcination step to prevent any formation of the volatile and hazardous RuO4 as well as to avoid a decrease in reactivity of the catalyst due to metal sintering. The different catalysts investigated in this study are listed in Table 1.
All catalysts were characterized by BET surface area measurements and hydrogen pulse chemisorptions to determine the total surface area, metallic surface area, metal dispersion and metal particle size. The results are summarized in Table 2. The hydrogen pulse chemisorptions were made at 50 _C for the nickel-based catalysts and at 150 _C for the ruthenium catalyst. Before the actual pulse chemisorptions, the catalysts were reduced and activated in situ according to the catalytic experiments described in the next paragraph. After the reduction, the catalyst was kept at the temperature for 1 h in a nitrogen flow of 50 mLmin_1 to remove adsorbed hydrogen.
The experimental setup used in the catalytic investigation is shown in Fig. 1. Catalytic experiments were performed in a tubular fixed bed quartz reactor with an inner diameter of 6 mm. For each test, 200 mg of catalyst material was loaded into the reactor and secured by quartz wool. The reactor was placed in an electrically heated oven. Before each run, the loaded catalyst was reduced and activated by heating at 5 _Cmin_1 to 500 _C for 2 h in a gas flow of 100 mL min_1 of Formier gas (10% hydrogen in nitrogen) (the Ru catalyst was reduced at 600 _C). SR experiments, at 400, 450, 500 and 600 _C, were conducted with a helium gas flow of 80 mL min_1 and an ethanol/water liquid flow of 0.04 mLmin_1. The liquid was pumped by a Knauer K-120 HPLC pump and vaporized byheating tape at 150–200 _C. The bioethanol (fraction 3) was diluted to approximately 25 vol% ethanol, which gave a total molar gas flow ratio to the catalyst bed of ethanol–water–helium of about 1:10:20. An Agilent GC 6890 N equipped with a CP Poraplot Q-HT capillary column to the FID detector and an advanced packed column system consisting of Porapak N column and a Molsieve column to the TCD detector was used for analyzing the reaction stream. Furthermore, a Rosemount BINOS 100 was continuously measuring the CO and CO2 contents in the exit gas after condensation of any liquids in an ice bath. The condensate was mainly water and occasionally minor amounts of unreacted ethanol and trace amounts of other liquid species.
This paper investigates the steam reforming reaction using technical bioethanol of different purities. As seen in Table 3, the main contaminants found in fractions 2 and 3 are various higher alcohols, which are not expected to be detrimental for the reaction.
Steam reforming of ethanol is thought to proceed through two separate routes; either by dehydrogenation to acetaldehyde (Eq. (2)) or by dehydration forming ethylene (Eq. (3)).
These two intermediate products can then be decomposed and steam reformed to an equilibrium mixture of methane,carbon dioxide, carbon monoxide, hydrogen and water (Eqs. (4)–(6)). Fig. 2 shows the theoretical equilibrium composition, under the experimental conditions given in the experimental section. At equilibrium the ethanol conversion is 100%, and selectivities in Fig. 2 are in complete agreement with the actual results for all the catalysts in the investigated temperature interval of 400–600 _C, as long as no deactivation is observed. Furthermore, ethylene is known to polymerize into pyrolytic coke on metal surfaces (Eq. (7)), and it is assumed that it is mostly ethylene that is responsible for the carbon formation on the catalyst in this reaction although other carbon-forming reactions (Eqs. (8) and (9)) can also be important. The final two reactions for achieving equilibrium among the gasses are the methane SR reaction (10) and the water gas shift (WGS) reaction (11).
CH3CHO --> CH4 + CO (4)
CH3CHO + H2O --> 3H2 + 2CO (5)
C2H4 + 2H2O --> 4H2 + 2CO (6)
C2H4 --> coke (7)
CH4 --> 2H2 + C (8)
2CO --> CO2 + C (9)
CH4 + H2O --> 3H2 + CO (10)
CO + H2O --> CO2 + H2 (11)
For an accurate comparison between the technical bioethanol and the commonly used ethanol/water mixtures, the same catalysts (Ni, K/Ni and Ru cf. Table 1) as those previously investigated in our group were used. In our previous investigation, it was seen that most experiments showed equilibrium composition in the exit gas, which means that all ethanol was converted to a mixture of H2, CO2, CO, CH4 and H2O. Furthermore, the experiments showed that carbon formation was highly affected by the operating temperature and the choice of catalyst. Only small amounts of carbon were deposited on any of the catalysts at 600 °C. At 400 °C, it was only the Ru catalyst, which had a full ethanol conversion and a low rate of carbon formation, cf. Fig. 3a. For the current experiments, the bioethanol of fraction 3 was diluted to approximately 25 vol% ethanol to facilitate the comparison with previous results. From Fig. 3b and Table 4, it can be seen that the rate of carbon formation shows the same apparentorder and trends in the runs with technical bioethanol as in the previous ones with a pure ethanol/water mixture.
However, the rate of carbon formation is generally slightly higher for fraction 3 than for the pure ethanol/water mixture, which can be explained by the slightly higher concentration of carbon-containing compounds in the feed from the contaminants in fraction 3 compared to the pure ethanol/water mixtures. It is also evident, as in the investigation with pure ethanol/water mixtures that an increased temperature of the reaction slows down the rate of carbon formation. This behavior can most likely be explained by diffusion limitations. Consequently, the concentration of ethanol at the catalyst surface will be low and therefore the rate of carbon formation will also be lower. Another observation from the comparison between Fig. 3a and b is the apparent significantly lower rate of carbon formation on the nickel-based catalysts with pure ethanol/water at 600 °C. These runs were maintained over a week compared to 16–18 hours for the experiments with technical bioethanol. It is anticipated that the rate of carbon formation is highest in the start of the reaction and therefore the apparent rate of carbon formation will be slightly lower for these two long runs.Since the Ru catalyst had higher activity and longer lifetime at the lower temperatures, compared to the Ni-based catalysts, this catalyst was also used in a comparative study with the more contaminated bioethanol of fraction 2. Thisfraction was also diluted to 25 vol% and used as feed in the reaction at 400 _C and 500 _C. The exit gas composition as measured on the BINOS at 400 _C is shown in Fig. 4. From this it can be seen that the yield of CO2 decreases whereas the yield of CO increases after approximately12 h for fraction 2. For the run using fraction 3 as feedstock neither the CO nor the CO2 concentration changes during the duration of the run (approximately 18 h). For the run at higher temperature, 500 _C, no deactivation is seen for any of the fractions during the duration of the run. It is also apparent that the CO2 concentration at equilibrium is higher for fraction 2. Both of these observations make sense when the total concentration of steam reformable material is considered. The increased carbon concentration from the contaminants at the catalyst surface will give a higher rate of carbon formation, resulting in an earlier decrease in catalyst performance. The increased amount of steam reformable material is also responsible for the higher concentration of CO2 in the exit gas; a calculation of the gas composition to the catalyst bed from fraction 2 and 3 shows that the extra contaminants in fraction 2 correspond to about a 10 mol% increase in the carbon content, in good agreement with the results in Fig. 4. On the other hand, it seems like the larger alcohols and other contaminants in fraction 2 also contribute to the faster deactivation of the catalyst. Through GC-analysis it is evident for all the experiments that full conversion of ethanol leads to the equilibrium gas-mixture as shown in Fig. 2. Moreover all the contaminants in fraction 2 and 3 are also reformed as they do not appear in the GC-analysis. First when the catalysts deactivate; ethanol, acetaldehyde, ethylene, ethane and traces of the contaminants start to appear in the exit gas. This is observed simultaneously with a decrease in methane and CO2 and an increase in the CO concentration, and is occurring with all of the used ethanol/water mixtures. The larger alcohols, as present in fraction 2, could cause severe difficulties through catalyst deactivation at low SR temperatures of technical bioethanol. One possible alternative to decrease carbon formation rates could be to use autothermal reforming instead of SR, this would negate some of the problems with carbon formation but at the expense of a lower H2 yield.