Conversion of Glucose to Hydrogen-Rich Gas by Supercritical Water

May 21, 2008 - ... to load: https://cdn.mathjax.org/mathjax/contrib/a11y/accessibility-menu.js .... available by participants in Crossref's Cited-by L...
0 downloads 0 Views 1MB Size
4106

Ind. Eng. Chem. Res. 2008, 47, 4106–4114

Conversion of Glucose to Hydrogen-Rich Gas by Supercritical Water in a Microchannel Reactor Aaron K. Goodwin and Gregory L. Rorrer* Department of Chemical Engineering, Oregon State UniVersity, CorVallis, Oregon 97331

Microchannel reactors offer the intensification of heat transfer to endothermic chemical reactions. Microchannel reactors consist of a parallel array of micron-scale channels that are integrated into one device using microfabrication techniques. The gasification of glucose by supercritical water was studied in a stainlesssteel microchannel reactor at 250 bar and 650-750 °C. The microchannel reactor architecture consisted of a parallel array of 21 rectangular microchannels (75 µm × 500 µm), each of 100 cm length, that were packaged into a serpentine pattern of 25 layers. At 750 °C, glucose was completely converted to gas products within a 2.0 s residence time, yielding an average gas composition of 53% H2, 35% CO2, 10% CH4, and 0.5% CO and a H2 yield of 5.7 ( 0.3 mol H2/mol glucose. At 650 °C, the intermediate products from the decomposition of glucose prior to their gasification and reforming were characterized. Routes for glucose transformation included the decomposition to acetic and propanoic acids, acid-catalyzed dehydration to 5-hydroxymethylfurfural and 2,5-hexanedione, and conversion to phenol. This study has shown that microchannel reactors have considerable promise for intensifying the thermochemical conversion of biomass constituents to useful chemicals and fuels. Introduction Thermochemical process technologies are a promising route for the conversion of renewable biomass feedstocks to useful chemicals and fuels, as they have many features in common with petrochemical processing technologies.1 Candidate thermochemical conversion technologies include the partial oxidation of biomass to a mixture of carbon monoxide and hydrogen (synthesis gas), pyrolysis to liquid bio-oils, and reforming to hydrogen-rich gas. Unlike petroleum-based hydrocarbon feedstocks, renewable biomass feedstocks such as starch and lignocellulose cannot be reformed to hydrogen with steam. However, it is well established that water in the supercritical state at temperatures greater than 374 °C and pressures greater than 221 bar converts these renewable feedstocks to hydrogenrich gas. Gasification of biomass carbohydrates in near- and supercritical water was recently reviewed by Matsumura et al.2 Advantages of supercritical water gasification include the direct processing of wet feedstocks, additional production of H2 through reforming, and generation of a compressed gas product. Glucose is an excellent model feedstock for benchmarking new thermochemical biomass conversion technologies because starch and cellulose are polymers of glucose. The theoretical reaction for reforming of glucose in supercritical water is C6H12O6 + 6H2O(s.c.) f 6CO2(g) + 12H2(g)

(1)

A technical challenge associated with the gasification of biomass carbohydrates in supercritical water is heat transfer to the reactor.2 Previous studies have shown that increasing the heat transfer rate increases the gasification efficiency.3–5 In order to provide sufficient heat transfer under continuous flow conditions, the gasification of carbohydrates and lignocellulosic biomass in supercritical water at temperatures ranging from 500 to 750 °C has been studied in laboratory-scale tubular reactors.6–13 These studies showed that complete gasification typically required a reactor residence time of 10-40 s at reactor inner diameters of around 2-6 mm. * To whom correspondence should be addressed. E-mail: rorrergl@ engr.orst.edu. Phone: (541) 737-3370.

The endothermic reforming of biomass constituents to hydrogen-rich gas in supercritical water could potentially benefit from the process intensification offered by microchannel reactor technology. In this regard, a recent review on hydrogen gas production from carbon-based feedstocks, including fossil fuels and biomass, stated that microchannel reactors are one of the most attractive options for process intensification.14 In contrast to conventional tubular reactors, microchannel reactors consist of a parallel array of micron-scale channels that are integrated into a single device using scaleable microfabrication techniques.15,16 The microchannel reactor architecture intensifies heat transfer to the fluid so that reactants are rapidly brought up to temperature and the endothermic reaction is maintained at the reaction temperature. This present study has two objectives. The first objective is to describe the fabrication and performance of a novel microchannel reactor for the gasification of glucose to hydrogen-rich gas by supercritical water at temperatures of 650 and 750 °C. Prior to this study, there were no reports of developing microchannel reactor architectures for carrying out any type of chemical reaction in supercritical water at high pressures and temperatures. The second objective of this study is to characterize the formation kinetics of the intermediate organic products of glucose decomposition in supercritical water in the microchannel reactor at 650 and 750 °C. Most of the previous studies on the gasification of biomass carbohydrates in supercritical water at temperatures above 500 °C focused only on the characterization of the gas products. However, analysis of intermediate product formation will provide fundamental insights on the reaction network underlying the conversion of glucose to hydrogen-rich gas in supercritical water at temperatures of 650-750 °C. Microchannel Reactor Design and Fabrication The microchannel architecture is detailed in Figure 1. The microchannel section of the reactor was sealed at the top and bottom with 304 stainless-steel square plates of 1.875 in. per side (4.76 cm/side) and 3/16 in. (0.48 cm) thickness. The

10.1021/ie701725p CCC: $40.75  2008 American Chemical Society Published on Web 05/21/2008

Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008 4107 Table 1. Microchannel Reactor Design Parameters design parameter

value and units

microchannel width (Wc) microchannel height (Hc) microchannel length (Lc) number of parallel channels (Nc) spacing between channel centerline number of channel layers (Np) effective channel length (Le) microchannel reactor height (HR) total reactor fluid volume (VR) fluid inlet/outlet header height and width fluid header length (LH) overall width and length total thickness (H)

500 µm 75 µm 4.0 cm 21 0.209 cm 25 100.0 cm 0.390 cm 0.800 cm3 250 by 500 µm 4.5 cm 4.76 cm 1.4 cm

a given channel. In the current design, 25 channel shims and 26 interlayer shims yielded an effective channel length of 100.0 cm (1.0 m) and an effective reactor height of 0.39 cm. The final microchannel design parameters are summarized in Table 1. Fluid was delivered to each channel through the bottom shim interlayer ports via an inlet header channel which ran perpendicular to the microchannel array. The inlet header channel, which was 250 µm deep, 500 µm wide, and 4.5 cm in length, was cut into the bottom plate of the reactor. Likewise, the fluid exiting the parallel array of microchannels was collected in an outlet heater channel of the same dimensions cut into the top plate of the reactor. The total header volume was 0.01125 cm3, only about 1.4% of the total 0.800 cm3 fluid reactor volume. The serpentine microchannel reactor was fabricated at the Nano/Micro Fabrication Facility of the Microproducts Breakthrough Institute, Corvallis, OR. A Toolcrafter CNC mill cut the header into the top and bottom plates. The stainless-steel shims were through cut by an ESI 5330 laser with a wavelength of 355 nm. A total of nine passes per channel at cutting speed of 150 mm/sec was used to cut each 500 µm width line for the channel shims or 500 µm square ports for the interlayer shims. The cleaned shims were placed into a mold to align all of the channels for bonding. The mold was placed into a Thermal Systems diffusion hot press and pressed at 850 °C and 555 psi for 150 min. After bonding, the inlet and outlet ports were micro-tig welded onto the reactor. Finally, the perimeter of the reactor was micro-tig welded. A representative microchannel reactor was cut into sections by wire EDM, and the cross sections were imaged by light photomicrography, as shown in Figure 2. Experimental Section

Figure 1. Microchannel reactor architecture: (a) Serpentine arrangement of 25 layers, cross section side view (75 µm channel height); (b) parallel array of 21 channels, top view (500 µm channel width).

microchannel section of the reactor consisted of a parallel array of 21 channels assembled from 304 stainless-steel channel shims and interlayer shims. The thickness of each shim plate was 75 µm, and the outer dimension was cut to a square of 1.875 in. (4.76 cm) per side. Each channel on a given channel shim was 500 µm in width, 75 µm in height, and 4.0 cm in length. Between each channel shim was an interlayer shim which consisted of an array of 21 holes, each 500 µm square, aligned in parallel array with the end of each 500 µm wide channel on the channel shim. The interlayer shims were arranged to weave the microchannel vertically up the reactor in a serpentine pattern by connecting one channel shim to the channel shim above it (Figure 1a). This arrangement increased the effective length of

Microchannel Reactor Test Loop. The continuous-flow microchannel reactor test loop is presented in Figure 3. All wetted parts, including thermocouples and pressure gauges, were constructed of 316 stainless steel. The microchannel reactor described in Figure 1 and Table 1 was sandwiched between two 375 W flat plate ceramic heaters. The heater block assembly was insulated. Conduction from the ceramic heating plates to the upper and lower plates of the microchannel reactor was the primary mode of heat transfer. The reactor temperature was controlled by a PID controller with a type J thermocouple imbedded into the bottom plate of the microchannel reactor. A feed solution containing 0.100 mol/L (18 g/L) glucose was fed to the reactor with an Agilent 1100 series dual-piston isocratic high-performance liquid chromatography (HPLC) pump. The hot effluent fluid exiting the reactor was cooled to 25 °C with a shell-and-tube heat exchanger using water as the coolant. An adjustable precision back-pressure regulator (KHB1WOA6C2P6000, Swagelok,

4108 Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008

Figure 2. Photomicrographs of microchannel reactor cross sections: (a) Axial cross section of microreactor, showing serpentine channel array; (b) transverse cross section of microchannels and fluid inlet header; (c) overall footprint of microreactor, showing cross-section cut lines; (d) perspective view of microchannel array, showing reactor, top plate, and bottom plate. Table 2. Reactor Conditions process parameter

value and units

temperature (T) pressure (P) density of water FR (T, P) glucose feed solution concentration (CG,o) liquid feed rate residence time (τ)

Figure 3. Continuous flow reactor test loop.

Inc., stainless steel) stepped down the pressure from 250 to 1.03 bar. The gas and liquid products separated. The liquid products were collected in the gas-liquid separator. The gas products were dried in-line, metered with a mass flow-meter (Omega, Inc. FMA 1800 series, 0-20 sccm, and 0-100 sccm, aluminum/brass body), and then collected into a 2.0 L gas collection bag. The volumetric flow rate obtained from the mass flow-meter reading was corrected for the gas composition.

650, 750 °C 250 bar 0.063 g/cm3 (650 °C) 0.055 g/cm3 (750 °C) 0.100 mol/L 0.1-1.0 cm3/min 2.0-15.1 s (650 °C) 2.6-26.3 s (750 °C)

Gasification Experiments in the Microchannel Reactor. The reactor conditions considered by this study are presented in Table 2. The reaction temperature was taken at the thermocouple position, and the pressure was taken at the pump head. The fluid residence time (τ) was estimated by τ)

VRFR(T, P) VoFo

(2)

where VR is the reactor fluid volume (cm3), Vo is the volumetric flow rate of the liquid feed to the reactor (250 bar, 25 °C), Fo is the density of the liquid feed (1.08 g/cm3), and FR is the density of the fluid at the reactor temperature (T) and pressure (P). Equation 2 assumes the fluid stream is at the reactor setpoint

Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008 4109

temperature and pressure and does not account for density changes as the water is being heated from the subcritical liquid state to the supercritical state. Since the concentration of glucose in the feed solution was dilute (0.100 mol/L, or 18 g glucose/ L), water was the dominant species in the reactor. Therefore, the formation of gas products in the reactor was not factored into the residence time calculation, but since more gas products would be formed relative to water consumed, the residence time calculated by eq 2 would be longer than the actual residence time. Prior to the first set of experiments with the new microchannel reactor, 0.5 g/min of water was pumped through the reactor system at 600 °C and 250 bar for 48 h to passivate all wetted surfaces. Typically, after approximately 24 h of continuous operation at 750 °C with a 0.1 mol/L glucose feed solution, the reactor back-pressure increased by 20 bar, which suggested clogging of the reactor due to coke formation. When the overpressure at the pump exceeded 20 bar (270 bar inlet pressure), the reactor was set to 750 °C and back-flushed with water at 0.5 g/min until the delivery pressure fell back to 250 bar. The reactor was then flushed again with air at 9 bar for 1.0 h. Detailed analysis of coking on the reactor walls was beyond the scope of this study. Analytical Procedures. Gas and liquid products were analyzed by gas chromatography (GC) and HPLC, respectively. Gas products were quantitatively analyzed by a Hewlett-Packard HP 5890 GC instrument with thermal conductivity detection using an Alltech Associates Carbosphere 80/100 1/8 in. by 6 ft (0.3175 cm by 183 cm) packed column. The gases CO2, CO, and CH4 were analyzed using He as the carrier gas at 28 mL/ min, a constant oven temperature of 80 °C, an injector temperature of 120 °C, and a detector temperature of 120 °C. The injection volumes were 250 or 500 µL. These gas compounds were identified by retention time and quantified by external calibration against a gas standard mixture (Alltech Associates, Inc., gas standard no. 19792). The H2 in the gas products was analyzed separately using N2 as the carrier gas at 34 mL/min and an oven temperature of 85 °C. Glucose in the dissolved water condensed from the reactor effluent was analyzed by carbohydrate HPLC with refractive index detection using a BioRad Aminex HPX-87P column at 85 °C with 0.6 mL/min HPLC-grade water as the eluent. Organic compounds in these liquid-phase products were analyzed by a Dionex DX-300 HPLC instrument with UV detection at 210 and 290 nm using a BioRad HPX-87H column at 65 °C with 0.5 mM sulfuric acid eluent at 0.6 mL/min. The injection volume was 25 µL. Compounds were identified by retention time and UV spectrum against standard compounds obtained from SigmaAldrich. The following organic compounds were quantified by the external standard method at 210 nm: acetic acid (Rf ) 8.44 × 103 µV · s/µmol), propanoic acid (Rf ) 9.74 × 104 µV · s/ µmol), propenoic acid (Rf ) 1.15 × 106 µV · s/µmol), 2,5hexanedione (Rf ) 2.88 × 104 µV · s/µmol), and phenol (Rf ) 1.42 × 106 µV · s/µmol). Thermodynamic Equilibrium Calculations. Equilibrium thermodynamic calculations were performed on ChemCad 5.4 (Chemstations, Inc.) by the Gibbs free energy minimization approach17 using physical properties, reference state enthalpies, and reference state Gibbs free energies for the reactants (glucose and water) and products (H2, CO2, CH4, and CO). Simulations were performed at the state variables (initial reactant concentrations, temperature, pressure) associated with the reactor experiments. The assumed reactions included glucose reforming to H2 and CO2 (eq 1), the water gas shift reaction, and the

Figure 4. Comparison of gas production rate at reaction temperatures of 650 and 750 °C as a function of glucose feed rate into the microchannel reactor. The feed was 0.100 mol/L glucose solution, and the pressure was 250 bar.

methanation reaction. Thermodynamic parameters for glucose were obtained from the NIST Chemistry WebBook.18 Heat Transfer Coefficient Calculations. The convective heat transfer coefficient for supercritical water within a single microchannel in laminar flow at low Graetz numbers (Gz) was estimated from the general form of Hausen’s correlation,19 given by Nu )

K1Gz hd ) Nu∞ + k 1 + K2Gzb

(3)

with K1 ) 0.023, K2 ) 0.0012, and b ) 1. In eq 3, Nu is the Nusselt number, Nu∞ is the asymptotic Nusselt number, h is the heat transfer coefficient, k is the thermal conductivity of the fluid, and d is the effective channel diameter. The effective channel diameter was estimated by d ) 2HCWC/(HC + WC), where HC ) 75 µm (channel height) and WC ) 500 µm (channel width). The Graetz number (Gr) was estimated by Gz ) Re × Pr × d/L, where L is the channel length, Pr is the Prandtl number, and Re is the Reynolds number. The Reynolds number was estimated by Re ) ν∞d/V, where V∞ is the bulk fluid velocity through one channel at the temperature and pressure of the fluid and V is the kinematic viscosity. For water at 750 °C and 250 bar, k ) 0.12 W/m · K, V ) 7.15 × 10-7 m2/sec, and Pr ) 0.893.18 For a rectangular duct with WC/HC ) 6.67, Nu∞ ) 5.33 under constant wall temperature.20 Results The reaction of glucose with supercritical water in the stainless-steel microchannel reactor produced gaseous products and organic compounds dissolved in the aqueous phase of the condensed reactor effluent. The total gas production rates (referenced to 1.0 atm and 25 °C) vs the glucose feed rate at reaction temperatures of 650 and 750 °C are compared in Figure 4. At both temperatures, the gas production rate increased linearly with glucose feed rate. However, the gas production rate at 750 °C was consistently about 2.2 times higher than the gas production rate at 650 °C. The effect of fluid residence time on product gas composition at reaction temperatures of 650 and 750 °C is presented in Figure 5. Similarly, the effect of fluid residence time on hydrogen yield (mol H2 formed/mol glucose reacted) and total carbon recovery in the products at 650 and 750 °C is presented in Figure 6. At 750 °C, gas composition, H2 yield, and percent carbon recovery were not dependent on residence time. The average gas composition was 53% H2, 35% CO2, 10% CH4, and 0.5% CO. No other higher hydrocarbons (C2, C3, etc.) were detected by

4110 Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008

Figure 5. Gas composition versus residence time in the microchannel reactor. The feed was 0.100 mol/L glucose solution, and the pressure was 250 bar. Key: (a) Reaction temperature of 650 °C and (b) 750 °C. Product gas compositions were averaged over the residence time range.

Figure 7. HPLC profiles of liquid-phase products at two representative residence times. The feed was 0.100 mol/L glucose solution, and the pressure was 250 bar. Key: (a) Reaction temperature of 650 °C and (b) 750 °C. Compound identification: lactic acid (5), formic acid (6), acetic acid (7), propanoic acid (8), propenoic acid (9), acetone (11), 2,5-hexanedione (12), 5-hydroxymethylfurfural (14), furfural (21), phenol (22), and 3-methyl-2cyclopenten-1-one (23).

the carbon recovery in the gas-phase products was only 46.1 ( 6.8%. The CO concentration also was much higher at 650 °C (7.1%) and decreased with increasing residence time. Therefore, the results from Figures 4 and 6 suggested that glucose gasification was not complete at 650 °C. The product gas composition at 750 °C was used to predict the following empirical overall reaction stoichiometry for glucose (C6H12O6) reforming in supercritical water: C6H12O6(s) + 3H2O f 4.5CO2(g) + 1.5CH4(g) + 6H2(g) (4)

Figure 6. Hydrogen yield and carbon recovery versus residence time in the microchannel reactor. The feed was 0.100 mol/L glucose solution, and the pressure was 250 bar. Key: (a) Reaction temperature of 650 °C and (b) 750 °C.

gas chromatography with thermal conductivity detection. Over the range of fluid residence times shown in Figure 6, the average H2 yield was 5.7 ( 0.3 mol H2/mol glucose, and the average total recovery of carbon in the gas-phase products was 81.0 ( 3.4% based on CH4, CO2, and CO. However, at 650 °C, the average H2 yield was only 2.6 ( 0.5 mol H2/mol glucose, and

The predicted line at 750 °C in Figure 4 is based on eq 4, assuming complete glucose conversion. The predicted line matches the data. If the CH4 product is subsequently reformed in water according to the reaction CH4 + 2H2O T 4H2 + CO2, then the H2 yield would be 12 mol H2/mol glucose. Equation 4 is endothermic with a standard enthalpy of reaction equal to +114 kJ/mol based on gas-phase water or +246 kJ/mol based on liquid-phase water, estimated by Hess’ Law using known standard enthalpies of formation.18 No glucose was found dissolved in the liquid water condensed from the reactor effluent at all fluid residence times tested for both 650 and 750 °C. However, dissolved organic compounds were found in the liquid-phase products. The liquid products were clear and free of particulates. Representative HPLC analysis profiles of the liquid phase at 650 °C are compared in Figure 7a. A total of 23 products were found in the liquid phase by HPLC under UV detection at 210 nm. The dominant products were acetic acid (7), propanoic acid (8), 2,5-hexanedione (12), and phenol (22). Other organic compounds, including lactic acid (5), formic acid (6), propenoic acid (9), acetone (11), 5-hydroxymethylfurfural (5-HMF, 14), furfural (21), and 3-methyl2-cyclopentenone (23) were also found. Peaks 1 and 2 were salts, and all other unknown compounds (3, 4, 13, and 15-20)

Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008 4111

Figure 9. Comparison of measured and predicted gas composition (Gibbs equilibrium) at 650 and 750 °C for the gasification 0.100 mol/L glucose solution to H2, CO2, CH4, and CO in supercritical water at 250 bar.

Figure 8. Selected liquid-phase products and total carbon recovery in liquid versus residence time in the microchannel reactor. The feed was 0.100 mol/L glucose solution, and the pressure was 250 bar. Key: (a) Acetic acid and total carbon recovery in the liquid at 650 °C; (b) propanoic acid, propenoic acid, and 2,5-hexanedione at 650 °C; and (c) phenol at 650 and 750 °C. The solid line in panel a represents a curve fit to the empirical equation CA ) k1τ exp-k2τ, where CA is acetic acid concentration (mM) and τ is residence time (s), with fitting constants k1 ) 41.3 ( 3.3 s-1 and k2 ) 0.246 ( 0.017 s-1.

were not identified or quantified, as their HPLC analysis peaks were relatively small. At 750 °C, only acetic acid and phenol were identified by HPLC (Figure 7b), but the acetic acid concentration was not quantified. The effect of fluid residence time on acetic acid, propanoic acid, propeneoic acid, 2,5-hexanedione, and phenol concentration in the liquid-phase products at a reaction temperature of 650 °C is presented in Figure 8. At 650 °C, the pH of the liquid products increased from 3.2 to 3.5 over the range of residence times tested, which reflected the dissolved organic acids concentration. At longer residence times, the concentration of the organic acids and 2,5-hexanedione decreased, suggesting that they were further converted to gaseous products. Total carbon recovery in the liquid products obtained at a reaction temperature of 650 °C decreased from 31 to 9% as the residence time increased, as shown in Figure 8a. The effect of fluid residence time on phenol concentration in the liquid-phase products at reaction temperatures of 650 and 750 °C is compared in Figure 8c. At 650 °C, the phenol

concentration increased as residence time increased, whereas at 750 °C, the phenol concentration decreased as residence time increased, suggesting that it was an intermediate product. At 750 °C, the pH was 4.2. This acidity was due mainly to CO2 from the gas phase that dissolved into the liquid phase and formed carbonic acid. On the basis of the data presented in Figures 7 and 8, it appears that the phenol intermediate product was the most recalcitrant to gasification, but was ultimately gasified as temperature and residence time were increased. A Gibbs free energy minimization approach17 was used to estimate the equilibrium composition of gas products resulting from the reaction of glucose with water to the observed gas products H2, CO2, CO, and CH4. For a feed solution of 0.1 mol/L glucose in water, the calculated equilibrium composition of the gas products at 650 and 750 °C is presented in Figure 9. At these conditions, the equilibrium calculation did not predict that CH4 was a significant product, due in part to the large stoichiometric excess of water in the reactants. However, the measured composition of CH4 from the reactor experiment was 8.1% at 650 °C and 10.3% at 750 °C. As a consequence, the Gibbs model predicted a higher H2 concentration in the product gas. Discussion This study has described the fabrication and performance of a novel microchannel reactor that promotes the efficient gasification of glucose to a hydrogen-rich gas mixture by supercritical water. Prior to this study, there were no reports of developing microchannel reactor architectures for carrying out any type of chemical reaction in supercritical water at high pressures and temperatures. Microchannel reactors consist of a parallel array of micronscale channels that are integrated into a single device using scaleable microfabrication techniques.15,16 In this study, we developed and fabricated a novel, serpentine microchannel reactor architecture that packaged a parallel array of 21 rectangular microchannels (75 × 500 µm), each of 100 cm (1.0 m) effective length, into one compact laboratory reactor that is only 4.75 cm/side and 4 mm thick (Table 1, Figures 1 and 2). Microchannel reactors hold considerable promise for the intensification of endothermic steam reforming reactions because the micron-scale channel dimensions greatly increase the heat flux to the reacting fluid.15,16 This is especially true under laminar flow conditions at low Graetz numbers, where the asymptotic Nusselt number dominates and the heat transfer coefficient increases inversely to the channel dimension.19,21

4112 Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008

Figure 10. Suggested scheme for decomposition of glucose to organic intermediates in supercritical water at 650 °C. All numbered compounds were found by HPLC.

The endothermic reforming of glucose to hydrogen-rich gas in supercritical water benefited from process intensification offered by microchannel reactor technology. Previous studies have shown that increasing the heat transfer rate to the gasification of biomass in supercritical water increases the gasification efficiency.3–5 In order to provide sufficient heat transfer under continuous flow conditions, the gasification of glucose and biomass in supercritical water has been studied in laboratory scale tubular reactors6–13 at residence times ranging from 10 to 40 s. It is difficult to compare reactor performance in these studies because the reaction conditions and definitions of residence time were all different. However, as the tube diameter decreased, the residence time (based on eq 2) required for complete gasification generally decreased. For example, Lee et al.9 reported that the complete gasification of a 0.6 mol/L glucose solution to H2-rich gas required at least 19 s residence time at 700 °C and 280 bar in a Hastelloy C276 tubular reactor of 6.22 mm inner diameter in laminar flow. Holgate et al.7 reported that the complete gasification of 0.0010 mol/L glucose solution required at least 6 s of residence time at 600 °C and 246 bar in a Hastelloy tubular reactor of 1.71 mm inner diameter in laminar flow. Similar results were obtained by Byrd et al.13 at 700 °C and 248 bar with an Inconnel 600 tubular reactor of 3.0 mm inner diameter packed with Ru/Al2O3 catalyst. In contrast, this study showed that a microchannel reactor with a channel dimension of 75 µm in the direction of heat transfer converted a 0.100 mol/L glucose solution to a hydrogen-rich gas within a residence time of only ∼2 s at a reaction temperature of 750 °C and pressure of 250 bar (Figures 5 and 6). In supercritical water, glucose decomposed to organic compound intermediates, which subsequently reformed in water to a gaseous mixture of H2, CO, CO2, and CH4 (Figures 7 and 8). These organic intermediates were dissolved in the aqueous-phase products condensed from the reactor effluent. The major organic intermediates were acetic acid, propanoic acid, propenoic acid, 2,5-hexanedione, and phenol, whereas the minor organic

intermediates were 5-HMF, 3-methyl-2-cyclopenten-1-one, and furfural. To date, only one previous tubular reactor study considered the effect of residence time on the organic intermediate product distribution. Holgate et al.7 identified many of the key compounds found in this present study (acetic acid, propanoic acid, 5-HMF, and 2,5-hexanedione), but focused on products obtained at a reaction temperature of 500 °C. In this study, we characterized the liquid-phase products as a function of residence time and temperatures at 650 and 750 °C as a means to shed insights on glucose decomposition and reforming at these temperatures (Figure 8). The products identified from the decomposition of glucose by supercritical water in the stainless-steel microchannel reactor at 650 °C and 250 bar are assembled into the scheme presented in Figure 10. Several studies have characterized the glucose decomposition products in hot compressed water or supercritical water at or below 410 °C, which is close to water’s supercritical temperature of 374 °C.22–26 More fundamental aspects of the organic reaction chemistry possible in supercritical water have been reviewed by Savage.27 The decomposition of glucose to organic compound intermediates in the stainless-steel microchannel reactor at 650 °C most likely proceeds by three parallel pathways that are consistent with many elements of these previous studies. In the first pathway, glucose thermally decomposes to acetic acid and propenonic acid in supercritical water. Acetic acid accumulates as the predominant organic acid intermediate (Figure 8a), consistent with the observation that acetic acid is recalcitrant to gasification in supercritical water.6,28 In the second pathway, the organic acids catalyze the partial dehydration of glucose to 5-HMF. The acid-catalyzed dehydration of glucose to 5-HMF in water is a well-known reaction.29 This dehydration reaction is preferred at ionic conditions where the temperature is below 374 °C, the critical temperature of water.22 The 5-HMF intermediate is converted to 2,5-hexanedione, a known product of 5-HMF decomposition in nearsupercritical water.22,23 This diketone, which dehydrates to

Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008 4113

3-methyl-2-cyclopenten-1-one under hydrothermal reaction conditions, was also found by HPLC (compound 23, Figure 7. In the third pathway, glucose breaks down to phenol. Phenol can be gasified in supercritical water to a mixture of H2, CO, CO2, and CH4.30 To date, no mechanistic schemes have been reported to describe how glucose decomposes to phenol in supercritical water, although a speculative pathway proposed by Williams and Onwudili23 suggests that phenol is formed from glucose in near-supercritical water via a Diels-Alder cycloaddition reaction of 3-hydrofuranone and diene intermediates. At 650 °C, glyeraldehyde, pyruvaldehyde, and dihydroxyacetone were not found in the liquid-phase products, even at the lowest residence time of 2.0 s. Previous work by Kabyemela et al.25,26 showed that these compounds were the primary decomposition products of glucose at 300-400 °C and reaction times of 2.0 s or less. Future work could consider glucose decomposition experiments at 300-400 °C to determine if these products are also formed in the stainless-steel microchannel reactor. The organic intermediates from all three pathways subsequently decomposed to H2 and CO, which then underwent the well-known water gas shift and methanation reactions: water gas shift: methanation:

CO + H2O T CO2 + H2

(5a)

CO + 3H2 T CH4 + H2O

(5b)

The water gas shift reaction did not achieve the predicted Gibbs equilibrium at 650 °C (Figure 9). Detailed reaction kinetic and heat transfer modeling of the microchannel reactor for the reforming of glucose in supercritical water was beyond the scope of this study and is reserved for future work. However, estimation of the heat transfer coefficient in the microchannel may give insight into the ability of the microchannel reactor to deliver a high heat flux to the endothermic glucose reforming reaction. Specifically, the heat transfer coefficient (h) for supercritical water at 750 °C in the microchannel reactor was 4917 W/m2 · K at a Reynolds number (Re) of 69, which corresponded to at a glucose solution feed flow rate to the reactor of 1.0 mL/min. In contrast, if this feed flow rate was fed to a single tubular reactor of the same total reactor volume (0.800 cm3) but with diameter of 1.0 mm and length of 102 cm, similar in dimensions to previous laboratoryscale tubular reactor studies mentioned above,7,13 then the heat transfer coefficient would be 442 W/m2 · K at a Re value of 528, over a factor of 10 lower. The microchannel reactor described in this study was fabricated from 304 stainless steel, which nominally contains 18 wt % Cr and 8 wt % Ni.31 There are two potential issues with using stainless steel as the reactor material for carrying out reactions in supercritical water. First, supercritical water under continuous-flow conditions is known to corrode stainless steel, which exposes Cr and Ni to the reactor surface.32 Previous studies with biomass gasification in supercritical water reported similar phenomenon with tubular reactors made from Hastelloy.8 Exposed Ni clusters could promote “unintentional catalysis” of the gasification reactions.24,33 We speculate that unintentional catalysis may be the source of CH4 formation in the product gas, because an equilibrium thermodynamic calculation (Figure 9) did not predict any significant amount CH4 in the product gas. But the alternative explanation might simply be that the water gas shift and methanation reactions (eqs 5a and 5b) were not at equilibrium. Second, the maximum allowable stress of 304 stainless steel at 750 °C, when defined as 1% creep in 10 000 h, is 250 bar.31 This maximum allowable stress is right at the reactor operating pressure, and corrosion would likely lower this value. However, no obvious mechanical damage to

the reactor was observed over the time course of the experiments, as evidenced by the reactor cross sections shown in Figure 2. Detailed stress analysis of the microchannel reactor at high temperature and pressure under static and flow conditions was beyond the scope of this present study and is reserved for future work. Future work will also consider microchannel reactors fabricated from high-nickel metals such as Hastelloy which resist corrosion in supercritical water and also avoid the potential for material failure at 250 bar and 750 °C. In conclusion, we have described a novel microchannel reactor that promotes the rapid conversion of glucose to hydrogen-rich gas by the use of supercritical water. Furthermore, the intermediate products of the glucose decomposition reactions were characterized. This study has shown that microchannel reactors have considerable promise for intensifying the thermochemical conversion of biomass constituents to useful chemicals and fuels. Acknowledgment This research was supported by the U.S. Army under the Tactical Energy Systems program administered through the Oregon Nanoscience and Microtechnology Institute. The authors thank Brian Paul and Todd Miller of the Micro/Nanofabrication Facility of the OSU/PNNL Microproducts Breakthrough Institute at Oregon State University for assistance with the design and fabrication of the microchannel reactor. Literature Cited (1) Huber, G. W.; Corma, A. Synergies between Bio- and Oil Refineries for the Production of Fuels from Biomass. Angew. Chem., Int. Ed. 2007, 46, 7184–7201. (2) Matsumura, Y.; Minowa, T.; Potic, B.; Kersten, S. R. A.; Prins, W.; van Swaaij, W. P. M.; van de Beld, B.; Elliott, D. C.; Neuenschwander, G. G.; Kruse, A.; Antal, M. J. Biomass Gasification in Near- and Supercritical Water: Status and Prospects. Biomass Bioenergy 2005, 29, 269–292. (3) Sinag˘, A.; Kruse, A.; Rathert, J. Influence of Heating Rate and the Type of Catalyst on the Formation of Key Intermediates and on the Generation of Gases During Hydropyrolysis of Glucose in Supercritical Water in a Batch Reactor. Ind. Eng. Chem. Res. 2004, 43, 502–508. (4) Watanabe, M.; Aizawa, Y.; Iida, T.; Levy, C.; Aida, T. M.; Inomata, H. Glucose Reactions within the Heating Period and the Effect of Heating Rate on the Reactions in Hot Compressed Water. Carbohydr. Res. 2005, 340, 1931–1939. (5) Matsumura, Y.; Harada, M.; Nagata, K.; Kikuchi, Y. Effect of Heating Rate of Biomass Feedstock on the Carbon Gasification Efficiency in Supercritical Water Gasification. Chem. Eng. Commun. 2006, 193, 649– 659. (6) Yu, D.; Aihara, M.; Antal, M. J. Hydrogen Production by Steam Reforming Glucose in Supercritical Water. Energy Fuels 1993, 7, 574– 577. (7) Holgate, H. R.; Meyer, J. C.; Tester, J. W. Glucose Hydrolysis and Oxidation in Supercritical Water. AIChE J. 1995, 41, 637–648. (8) Antal, M. J.; Allen, S. G.; Schulman, D.; Xu, X.; Divilio, R. J. Biomass Gasification in Supercritical Water. Ind. Eng. Chem. Res. 2000, 39, 4040–4053. (9) Lee, I.-G.; Kim, M.-S.; Ihm, S.-K. Gasification of Glucose in Supercritical Water. Ind. Eng. Chem. Res. 2002, 41, 1182–1188. (10) Hao, X. H.; Guo, L. J.; Mao, X.; Zhang, X. M.; Chen, X. J. Hydrogen Production from Glucose Used as a Model Compound of Biomass Gasified in Supercritical Water. Int. J. Hydrogen Energy 2003, 28, 55–64. (11) Yoshida, T.; Oshima, Y. Partial Oxidative and Catalytic Biomass Gasification in Supercritical Water: A Promising Flow Reactor System. Ind. Eng. Chem. Res. 2004, 43, 4097–4104. (12) Lu, Y. J.; Guo, L. J.; Ji, C. M.; Zhang, X. M.; Hao, X. H.; Yan, Q. H. Hydrogen Production by Biomass Gasification in Supercritical Water: A Parametric Study. Int. J. Hydrogen Energy 2006, 31, 822–831. (13) Byrd, A. J.; Pant, K. K.; Gupta, R. B. Hydrogen Production from Glucose using Ru/Al2O3 Catalyst in Supercritical Water. Ind. Eng. Chem. Res. 2007, 46, 3574–3579.

4114 Ind. Eng. Chem. Res., Vol. 47, No. 12, 2008 (14) Navarro, R. M.; Pena, M. A.; Fierro, J. L. G. Hydrogen Production Reactions from Carbon Feedstocks: Fossil Fuels and Biomass. Chem. ReV. 2007, 107, 3952–3991. (15) Ja¨hnisch, K.; Hessel, V.; Lo¨we, H.; Baerns, M. Chemistry in Microstructured Reactors. Angew. Chem., Int. Ed. 2004, 43, 406–446. (16) Holladay, J. D.; Wang, Y.; Jones, E. Review of Developments in Portable Hydrogen Production Using Microreactor Technology. Chem. ReV. 2004, 104, 4767–4790. (17) Tang, H.; Kitagawa, K. Supercritical Water Gasification of Biomass: Thermodynamic Analysis with Direct Gibbs Free Energy Minimization. Chem. Eng. J. 2005, 106, 261–267. (18) NIST Chemistry WebBook; U.S. National Institute for Standards and Technology (NIST). http://webbook.nist.gov, accessed 2007. (19) Celata, G. P.; Cumo, M.; McPhail, S. J.; Zummo, G. Single-Phase Laminar and Turbulent Heat Transfer in Smooth and Rough Microtubes. Microfluid. Nanofluid. 2007, 3, 697–707. (20) Haji-Sheikh, A. Fully Developed Heat Transfer to Fluid Flow in Rectangular Passages Filled with Porous Materials. Trans. ASME 2006, 128, 550–556. (21) Lee, P.-S.; Garimella, S. V.; Liu, D. Investigation of Heat Transfer in Rectangular Microchannels. Int. J. Heat Mass Transfer 2005, 48, 1688– 1704. (22) Kruse, A.; Gawlik, A. Biomass Conversion in Water At 330410 °C and 30-50 MPa. Identification of Key Compounds for Indicating Different Chemical Reaction Pathways. Ind. Eng. Chem. Res. 2003, 42, 267–279. (23) Williams, P. T.; Onwudili, J. Composition of Products from the Supercritical Water Gasification of Glucose: A Model Biomass Compound. Ind. Eng. Chem. Res. 2005, 44, 8739–8749. (24) Kersten, S. R. A.; Potic, B.; Prins, W.; Van Swaaij, W. P. M. Gasification of Model Compounds and Wood in Hot Compressed Water. Ind. Eng. Chem. Res. 2006, 45, 4169–4177.

(25) Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai, K. Kinetics of Glucose Epimerization and Decomposition in Subcritical and Supercritical Water. Ind. Eng. Chem. Res. 1997, 36, 1552–1558. (26) Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai, K. Glucose and Fructose Decomposition in Subcritical and Supercritical Water: Detailed Reaction Pathway, Mechanisms, and Kinetics. Ind. Eng. Chem. Res. 1999, 38, 2888–2895. (27) Savage, P. E. Organic Reactions in Supercritical Water. Chem. ReV. 1999, 99, 603–621. (28) Calvo, L.; Vallejo, D. Formation of Organic Acids during the Hydrolysis and Oxidation of Several Wastes in Sub- and Supercritical Water. Ind. Eng. Chem. Res. 2002, 41, 6503–6509. (29) Lourvanij, K.; Rorrer, G. L. Reactions of Aqueous Glucose Solutions with Y-Zeolite Catalysts at 110 to 160 OC. Ind. Eng. Chem. Res. 1993, 32, 11–19. (30) DiLeo, G. J.; Neff, M. E.; Savage, P. E. Gasification of Guaiacol and Phenol in Supercritical Water. Energy Fuels 2007, 21, 2340–2345. (31) American Society for Metals (ASM). Metals Handbook, Properties and Selection: Iron, Steels, and High-Performance Alloys, 10th ed.; ASM International: Materials Park, OH, 1990; Vol. 1. (32) Hayward, T. M.; Svishchev, I. M.; Makhija, R. C. Stainless Steel Flow Reactor for Supercritical Water Oxidation: Corrosion Tests. J. Supercrit. Fluid 2003, 27, 275–281. (33) Davda, R. R.; Shabaker, J. W.; Huber, G. W.; Cortright, R. D.; Dumesic, J. A. A Review of Catalytic Issues and Process Conditions for Renewable Hydrogen and Alkanes by Aqueous Reforming of Oxygenated Hydrocarbons over Supported Metal Catalysts. Appl. Catal., B. 2005, 56, 171–186.

ReceiVed for reView December 18, 2007 ReVised manuscript receiVed March 25, 2008 Accepted March 28, 2008 IE701725P