Products and Kinetics of Glucomannan Pyrolysis - Industrial

Mar 14, 2013 - Mohd Asmadi , Haruo Kawamoto , Shiro Saka. Journal of Analytical and ... Yoshiaki Kato , Ryohei Enomoto , Minami Akazawa , Yasuo Kojima...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/IECR

Products and Kinetics of Glucomannan Pyrolysis C. Branca,† C. Di Blasi,*,‡ C. Mango,‡ and I. Hrablay‡,§ †

Istituto di Ricerche sulla Combustione, CNR, P.le V. Tecchio, 80125 Napoli, Italy Dipartimento di Ingegneria Chimica, dei Materiali e della Produzione Industriale, Università degli Studi di Napoli “Federico II”, P.le V. Tecchio, 80125 Napoli, Italy

‡

S Supporting Information *

ABSTRACT: The pyrolysis of particles of glucomannan, the main component of softwood hemicelluloses, is investigated in a fluidized-bed reactor. A first set of experiments is carried out for temperatures in the range 530−690 K to determine yields and composition of products. The most abundant are char, water, and carbon dioxide. The condensable organic fraction mainly consists of acetic acid, formic acid, hydroxypropanone, hydroxyacetaldehyde, and furfuryl alcohol. The second set of experiments is made to determine the weight loss characteristics of small samples exposed in the expanded bed at temperatures of 503−593 K that are then used to develop one- or two-stage pyrolysis mechanisms.

1. INTRODUCTION Hemicelluloses are structural polysaccharides of the plant cell wall in close association with cellulose and lignin, forming the lignocellulosic biomass. These polysaccharides are generally heterogeneous and are built up of different hexoses (C6 sugars) and pentoses (C5 sugars) sometimes in addition to uronic acids.1 They have a lower degree of polymerization than cellulose, are largely soluble in alkalis, and are also more easily hydrolyzed. They are usually between 20 and 30% of the dry matter of wood and up to 35% for some agricultural residues.2 A summary of the main properties of wood hemicelluloses is reported in ref 3. Hemicelluloses in softwoods are mainly galactoglucomannan, containing mannose/glucose/galactose residues in a ratio of 3/1/1, and glucomannan with mannose/ glucose residues in the ratio 3/1. Glucomannan amounts to about 10−15 wt % and galactoglucomannan to about 5−8%.4 Arabinoglucuronoxylans are also present in softwoods (7−8 wt %), whereas in hardwoods acetylglucuronoxylans are present (15−30 wt %) in association with variable percentages of galactose, arabinose, rhamnose, and methylglucuronic acid units and acetyl groups. A small percent of glucomannan with a mannose to glucose ratio of 2/1 is also reported in hardwoods. In herbaceous biomass, such as straw, arabinoxylans are the predominant hemicellulose polysaccharides and those containing galactose and mannose are only minor ones.1 Hemicelluloses are the least thermally stable among biomass components, and their pyrolytic behavior is of paramount importance for improving the knowledge of process fundamentals and the design of effective conversion systems. In particular, for the torrefaction process, carried out at temperatures of 473−593 K,5 the decomposition of hemicellulose is the main chemical reaction.6,7 A significant number of studies is available concerning the chemical kinetics and products of pyrolysis of xylan, used as a model compound for hardwood hemicelluloses (for instance, refs 8−14.). On the contrary, glucomannan and galactoglucomannan, the main components of softwood hemicelluloses, have never been studied in detail, although the wood pyrolysis characteristics are known to be affected by the hardwood or softwood species.15−17 © 2013 American Chemical Society

Data on the degradation of hemicellulose isolated from Douglas fir wood were examined in ref 18 over a temperature range of 383−493 K with observation times between 2 h and 64 days. An activation energy of 112 kJ/mol was evaluated for the degradation process with a rate that, at 423 K, is about 4 times as fast as that of wood. However, such data are not of interest for practical applications. Single thermogravimetric curves for glucomannan and xylan pyrolysis were compared for samples extracted from wood according to standard procedure19 and commercial samples.20 Both measurements indicate that the degradation of glucomannan is anticipated at lower temperatures, although the two samples show highly different dynamics of weight loss. No attempt was made to develop a kinetic model for the process. Finally, glucomannan was isolated from Japanese cedar wood, with subsequent demineralization (hydrolyzable sugar contents amounted to 16.5% (glucose), 68.1% (mannose), 4.7% (galactose), 6.4% (xylan), and minor amounts of arabinose).21,22 The sample (50 mg) was pyrolyzed at 1073 K, and the condensable products included hydroxyacetaldehyde, acetic acid, hydroxypropanone, furfural, levoglucosan, 5-hydroxymethylfurfural, levomannosan, and some furan and cyclopentenonone compounds. Of course a single experiment is not sufficient to characterize glucomannan pyrolysis, and on the other hand, the temperature investigated is excessively high for the typical conditions established in torrefaction and pyrolysis. This brief analysis about the state of the art shows that the pyrolytic behavior of glucomannan, which is of interest for the torrefaction and pyrolysis of softwood species, is largely unknown. In particular, the degradation kinetics and the yields and composition of products, for the temperature range where primary degradation of hemicelluloses takes place, have never been analyzed. In this study a laboratory scale fluidized-bed reactor is applied to investigate the pyrolytic behavior of commercial Received: Revised: Accepted: Published: 5030

January 15, 2013 March 8, 2013 March 14, 2013 March 14, 2013 dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

metal plate supporting the bed. The bed material is calcined sand (at 1000 K for 3 h) sieved to 240−280 μm size range, with a static bed of 0.11 m height and 1650 kg/m3 density, and a minimum fluidization velocity at atmospheric pressure of 0.04 m/s. The volumetric gas flow rate is varied with the bed temperature to achieve in all cases a superficial velocity about 2.5 times higher than that at minimum fluidization conditions, which is typical of laboratory scale pyrolyzers based on bubbling fluidized beds.25−27 For velocities greater than that for the onset of fluidization the pressure drop across the whole bed remains nearly constant, while the additional gas flow produces greater expansion of the bed.28Temperature profiles along the reactor axis are measured by four thermocouples (chromel−alumel type, 500 μm diameter), with their tips exiting from a protective steel tube, at chosen distances from the flow distributor. The expanded bed (about 0.13 m) is isothermal at a temperature determined by a proper set point of the furnace (PID controller), but gradients in the upper part are high. However, this is not a drawback given that only the expanded bed characteristics are of interest. The first set of experiments is carried out to determine the yields and composition of products of glucomannan pyrolysis. Once the desired temperature of the expanded bed is attained (indicated in the following as heating temperature, Th), the sample (about 40 g) is instantaneously (about 7−10 s) fed at a distance of about 0.15 m from the flow distributor by means of a removable pipe (diameter 0.021 m). In this way, possible elutriation of the smaller particles during feeding is avoided. Heating temperatures are varied in the range 530−690 K. Nitrogen and volatile pyrolysis products pass through a condensation train consisting of two water/ice cooled condensers (with a catch pot, where the largest fraction of liquids is collected and chemically characterized), two wet scrubbers, three cotton wool traps, and a silica gel bed (all connected in series). Gas sampling and analysis are made at selected times allowing the exit volumetric flow rate and mass of each gaseous species to be determined. Gas analysis is carried out through a gas chromatograph (Perkin-Elmer Auto-System XL), equipped with a thermal conductivity detector (TCD) and a packed column (Supelco 60-80 Carboxen 1000, 15 ft) with helium as the carrier gas. The liquid products are stored at a temperature of 277 K with no light exposure. After filtration with microfilters (Millex-Gx of 0.45 μm), chemical analysis is made by means of gas chromatography/mass spectrometry (GC/MS; Focus GC-DSQ, Thermo Electron) with a quadrupole detector and a DB-1701 capillary column (60 m × 0.25 mm i.d., 0.25 mm film thickness). Gas-chromatographic conditions are the same as in previous work of this group.2,29−31 The water content of the liquid product is determined by means of Karl Fischer titration according to the standard test method ASTM E203-96. The organic fraction is computed as the difference between the measured total liquids and the water so determined. The char yield is determined as the difference between the bed inventory before feeding and that at the conclusion of the pyrolysis test. The second set of experiments is made to determine the weight loss curves of glucomannan pyrolysis. Once the desired thermal conditions are achieved, a small sample mass is instantaneously suspended in the expanded bed (at a distance of about 0.11 m the from flow distributor). Indeed, the reactor closing cap is substituted by another one equipped with a thin steel rod acting as a support for a sample holder and a thermocouple (in direct contact with the sample). The sample holder, made of Pyrex, is cylinder-shaped with a 0.01 m

glucomannan at several temperatures. This system allows the process to be conducted under isothermal conditions, in this way producing data that can be applied for various reactor configurations.

2. MATERIALS AND METHODS Commercial glucomannan AS (Dr. Behr GmbH), also indicated as Konjac glucomannan, was isolated from the corm of Amorphophallus konjac and used in this study. It consists of β-1,4-linked D-mannose and D-glucose residues in a molar ratio of about 1.6/1.23 There are some branches linked to the backbone, but the exact branched position is still under debate. The glucomannan backbone of the sample possesses 5−10% acetyl substituted residues, and it is widely accepted that the presence of this group confers water solubility. According to the producers, the sample as received presents contents of moisture, ash, and protein of 9.8, 3.8, and 2.9%, respectively. Particle size distribution is as follows: 0.075 mm (1%), 0.09 mm (1.3%), 0.125 mm (21.6%), 0.18 mm (48%), 0.25 mm (20%), and 0.3 mm (8%). Before the experiments the sample is ovendried at 343 K for 15 h. The experiments are performed in a laboratory scale system,24 schematically represented in Figure 1, consisting of

Figure 1. Schematic of the fluid-bed pyrolysis reactor: (1) furnace, (2) reactor, (3) isolation valve, (4, 5) water/ice condensers, (6) wet scrubber, (7) cotton trap, (8) silica gel bed, (9) acquisition data set, and (10) thermocouples.

a fluidized bed (stainless steel reactor with 0.063 m internal diameter and 0.45 m length), electrically heated by a furnace. Nitrogen, fed through a jacket (internal diameter 0.089 m) at the reactor top, is also preheated and distributed by a perforated 5031

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

diameter and 0.06 m height. The top surface is open, while the bottom surface acts as a support for a thin sample layer. The height is appropriate to permit the position of the sample over the bottom surface and to hinder the possible inlet of sand when the sample is exposed in the hot bed. A series of tests has indicated that a sample mass of about 15 mg (particle sizes below 80 μm) is sufficiently small to produce weight loss characteristics not affected by heat and mass transfer limitations for temperatures in the range 503−593 K. For a given heating temperature, the weight loss curve is constructed by a series of experiments which provide the current sample weight at selected times. For each observation time, the sample is extracted from the bed and the reactions are quenched by immersion of the sample holder in cold water under a continuous nitrogen flow at ambient conditions. As already outlined, the thermocouple is supported by a thin steel rod connected to the reactor closing cap. To minimize heat conduction across its length, the thermocouple (and the support) is insulated by means of a ceramic tube. In this set of experiments, given the very small sample mass, volatile products are not collected.

3. RESULTS Results are presented about the pyrolysis of glucomannan with emphasis on the primary decomposition reactions. In section 3.1 the heating dynamics, the rate of gaseous species formation, and the physical changes undergone by the sample while reacting are presented. Then the yields of the lumped classes of products and the composition of the gaseous and condensable fractions are given. In section 3.2, weight loss curves, measured for several heating temperatures, are subjected to a kinetic analysis leading to reaction mechanisms with the estimation of the related kinetic constants. In all cases the tests have been made in duplicate, showing good repeatability. In particular, for the first set of experiments, the mass closure is around 91−92%. 3.1. Products. Experiments to determine the yields and composition of the products of glucomannan pyrolysis are made for temperatures of the expanded bed, Th, in the range 530−690 K. The examination of the temporal profiles of temperature, at four positions along the reactor axis, and the rate of gas release (defined as the time derivative of the produced gas mass fraction) from the time of the batch feed to complete conversion indicates that two main behaviors can be identified. These are summarized by the results reported in Figure 2 for heating temperatures of 605 and 690 K. For the less severe heating condition (Th = 605 K, Figure 2A), it appears that sample feeding is associated with a rapid diminution in the temperature at the positions 0.11−0.15 m, followed by an equally rapid increase toward the initial bed temperature. The temperature at 0.06 m is initially not affected and then shows the same qualitative profile as those of the upper zone, even though the minimum value is considerably higher. It can be observed that highly variable temperature dynamics, established from the feeding time up to about 40 s, are not associated with gas production. Therefore, they can be essentially attributed to feed preheating from ambient conditions to temperatures sufficiently high for the beginning of the pyrolysis reactions. Indeed, the same qualitative trends are observed when, instead of glucomannan, inert sand (in the same amount and comparable particle sizes) is fed into the hot bed. Thus during the initial transients, glucomannan, fed at about 0.15 m, mixes with the hot sand, first in the upper zone (0.11−0.13 m) and then in the more lower zones (0.06 m), and is heated. For times longer than 40 s, the temperature profiles become flat with barely increasing values. The very small differences between the temperatures recorded at the various positions suggest

Figure 2. Time profiles of the bed temperature, at different heights, and the rate of gas release (time derivative of the gas mass fraction) for heating temperatures of (A) 605 and (B) 690 K during glucomannan pyrolysis.

that the single glucomannan particles preserve their identity during heating and degradation and that the temperature of the solid phase (sand and glucomannan particles at positions below 0.13 m) and that of the gas phase (volatile products of glucomannan pyrolysis and nitrogen at a position of 0.15 m) are approximately the same during the conversion process. Also, all the particles react under comparable thermal conditions independently of their position in the bed. In reality, glucomannan conversion does not take place under truly isothermal conditions but the actual temperature of the reacting particles slowly increases with time and always remains below the initial bed heating temperature. To refer product yields to the actual reaction temperature, a time-integral value can be introduced with reference to the conversion time32 with the definition of a reaction temperature, Tr, that is an average temperature. The value recorded by the thermocouple positioned at a bed height of 0.11 m is used for such a purpose, taking as initial and final times those corresponding to amounts of gas released of 2 and 75%, respectively (for a heating temperature of 605 K, the computed reaction temperature is 577 K). The conditions that reproduce heating dynamics qualitatively similar to those discussed here are established for heating temperatures between 530 and 605 K, corresponding to reaction temperatures, Tr, lower by 13−28 K (see details in Table S1 in the Supporting Information). For this range of heating temperatures, the curves of the rate of gas (and vapor) release show that a large part of the conversion process occurs over a rather narrow time period (the maximum is attained at a time of 76 s for the case of Th = 605 K) followed by a slow tail. The temperature profiles measured for more severe heating conditions (heating temperatures above 605 K, such as for the case of 690 K reported in Figure 2B) are representative of completely different dynamics. Apart from the anticipation in the beginning of the reaction process (at about 25 s), it can be 5032

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

Figure 3. SEM photographs of a glucomannan char agglomerate at magnifications of (A) 46, (B) 200, and (C, D) 800.

An experiment, to investigate the softening behavior of glucomannan, has been made by exposing a 20 mg sample, by means of the sample holder used for the construction of the weight loss curves, over the expanded bed height for a heating temperature of 580 K and examining the solid during the initial conversion stage at different times by means of SEM images (Figure 4). The measurement of the temperature versus time profile allows the different images to be associated with well-defined thermal conditions. For temperatures below approximately 500 K, no significant change takes place with respect to the virgin sample, but at higher temperatures (in good agreement with the glucomannan softening point previously reported33) single particles clearly show the existence of a molten phase with the appearance of surface bubbles, in part exploded due to the beginning of volatile species release. Also, at higher temperatures clusters of particles are formed linked together by filaments. The chief parameters characterizing the conversion dynamics can be gained from the gas release rate versus time, for various heating temperatures, reported in Figure 5 and include the conversion time, tc, defined as the time when the 75% of the total gas has been released to avoid the uncertainty associated with the wide tailing zone,2 the maximum gas release rate, dYmax, and the corresponding time, tmax (see details in Table S1 in the Supporting Information). It can be observed that the two characteristic times first decrease significantly (tmax from 324 to 76 s and tc from 1300 to 137 s) and then increase (up to 95 and 283 s) as a function of the heating temperature. In correspondence with these two zones, the maximum gas release rate is characterized by increasing (range 0.015−0.19 s−1) and decreasing values (up to 0.06 s−1). These results can be explained by taking into account the changes in the mixing/heating dynamics of glucomannan particles induced by material softening which, as already discussed above, is attained at successively shorter times as

observed that pyrolysis takes place under highly variable thermal conditions that are dependent on the bed height. Moreover, the lowest part of the bed (thermocouple positioned at a bed height of 0.06 m) remains at the initial conditions, indicating that it is not the site where feed particles are heated and degraded. To explain these dynamics, it should be taken into consideration that glucomannan powder undergoes thermal softening (for pine glucomannan a softening point of 454 K is reported, defined as the temperature at which the powder, when compressed under constant load in a glass capillary, collapses into a solid plug.33) Given the high heating temperatures, feed particles reach the softening point before they mix with the hot sand, leading to partial agglomeration of the sample where the identity of the single particles is, for a large part, lost. It is likely that the sample remains mainly positioned at the top of the bed with the actual heating rates depending on the size/shape of particle agglomerates. Therefore, they may become slower than those observed at lower heating temperatures when good mixing of the feed with sand takes place before the softening point is reached and the glucomannan particles preserve their identity during the conversion process. In fact, the comparison between the curves of the gas release rate reported in Figure 2 for the heating temperatures of 605 and 690 K reveals that, for the latter case, as a consequence of lower heating rates, conversion takes place over a wider time interval with a lower peak in the gas release rate. Particle agglomerates, which form at the higher heating temperatures, can still be observed in the final solid charred residues as shown in Figure 3 by means of scanning electron microscope (SEM) images. A charred agglomerate shows that the shape of the single particles is lost to give rise to a molten phase that solidifies, when charring reactions become active, and preserves some of the features typical of a molten phase. 5033

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

Figure 4. SEM photographs of glucomannan particles exposed above the expanded bed for a heating temperature of 580 K at different times corresponding to sample temperatures of (A) 300 K (t = 0), (B) 530 K (t = 150 s), (C) 535 K (t = 160 s), (D) 538 K (t = 165 s), and (E, F) 543 K (t = 180 s).

Instead, as shown by the heating and mixing dynamics, it may result from a decrease in the actual reaction temperature when the heating temperature is increased above a certain limit. The water yields show some scatter, but they remain approximately constant over the range of temperatures investigated. From the quantitative point of view, the yields of solid product vary from a maximum of about 53% to a minimum around 31% which is attained for heating temperatures of 660 K and above. The yields of gas and organics increase from about 10 and 4%, respectively, to maximum values around 15%. However, it can be reasonably assumed that the contribution of about 8−9%, representing the unsuccessful complete mass closure, should be ascribed to organic compounds (this would lead to yields of organic compounds varying between 12 and 23%). Water yields are around 27%. The gas consists of CO2 (yields between 7 and 11%) and CO (yields between 3 and 4%). The organic fraction includes carboxylic acids, alcohols, aldehydes, ketones, furans, sugars, and other minor compounds

the heating temperature is increased. Also, when particle agglomerates are formed (heating temperatures above 605 K), the region of high gas release rates may show local maxima due to nonuniform heating and conversion. The main information on the pyrolysis products of glucomannan is shown in Figure 6 for the yields of the lumped classes of pyrolysis products (char, gas, organics, and water) and gaseous species (CO and CO2) versus the heating temperature. A typical chromatogram of the pyrolysis liquid (water and organic fraction) is shown in Figure 7 with the identified compounds listed in Table 1. As for the lumped product classes, the same qualitative trends typically observed for lignocellulosic fuels34,35 are again established with the yields of solid residue decreasing to the advantage of gaseous and organic compounds as the reaction temperature increases. A barely visible maximum in the yields of organics appears for heating temperatures around 640 K, followed by a decrease. Given the low temperatures examined, it is not likely that this trend can be attributed to secondary degradation reactions. 5034

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

(Figure 7, Table 1) which are quantitatively dependent on the heating temperature. The yields of the main species, including acetic acid (AA), hydroxyacetaldehyde (HAA), hydroxypropanone (HP), formic acid (FA), levoglucosan (LG), acetoxyacetone (AAone), 2-methyl-2-cyclopentenone (2M2C), 3-methyl-2-cyclopentenone (3M2C) 2-ethyl-2-hydroxy-2-cyclopentenone (2E2HC), propionic acid (PA), hydroxybutanone (HB), butyrolactone (BL), furfural (FF), acetylfuran (AF), furfuryl alcohol (FAL), 2(5H)furanone (2(5H)Fone), and 5-methylfurfural (5MF), have been quantified and are reported versus the heating temperature in Figure 8 and Table S2 of the Supporting Information. The majority of compounds (HAA, AA, HP, PA, HB, 2E2H2C, FA, FF, 2(5H)Fone) tend to increase with the reaction temperature. In other cases approximately constant yields are observed. The most abundant products are AA, HAA, HP, FA, and FAL. Significant values are also measured for PA, HB, AF, and 2(5H)Fone. Very small yields are observed for the remaining compounds. These findings are in agreement with the analysis about the torrefaction products of wood/biomass which are the result of hemicellulose degradation.36−40 CO2 is always the chief gaseous product, independently of the hemicellulose properties. As for the condensable products, water is the most abundant. In particular for the torrefaction of pine wood,40 comprising galactoglucomannan hemicellulose, it corresponds to about 75−52% of the liquid produced over the temperature range 513−593 K, which compares well with the figures of this study (86−64% of the liquid produced). The composition of the organic fraction is affected by the hemicellulose properties. In agreement with the results found for pine wood torrefacion,40 AA is the main organic product of glucomannan pyrolysis. The other reported products (HP, PA, FF, and FA) are also found here. A comparison is available36 for willow and larch (hardwood and softwood, respectively). It is outlined that, in the latter case, the yields of condensable products are smaller and that the main acid formed is FA. This is also reported for high temperature (1073 K) pyrolysis of wood extracted glucomannan.21 Pyrolysis of glucomannan AS, investigated here, also indicates that FA is one of the most important products. HAA, which shows increasing yields with the reaction temperature, is also reported in ref 21. 5-Hydroxymethylfurfural (5-HMF), which is known to be a product of softwood hemicellulose2,41 and is reported at high temperature,21 is not detected. It can be hypothesized that the significant amount of ash, present in the glucomannan sample, catalyzes its decomposition, giving FAL, or that it is preferentially formed at higher temperatures. Finally, the production of anhydro sugars (in particular LG), at the low temperatures investigated, is very small, but it can be expected to become significant at high temperatures.21 3.2. Chemical Kinetics. Weight loss curves of glucomannan under negligible heat and mass transfer limitations have been measured for seven temperatures of the expanded bed in the range 503−593 K. The measured temperature profiles of the sample show maximum heating rates of 300−400 K/min so that, in principle, weight loss could be considered to occur under isothermal conditions. However, in the kinetic analysis, the actual sample temperature is considered. The method of parameter estimation makes use of a numerical solution for the mass conservation equations (with the temperature evolution provided by the measured profile) and a direct method for the minimization of the integral form of the objective function taking into account the differences between the model predictions and the experimental measurements for all the thermal conditions. More precisely, the evaluation is carried out in the framework of the Matlab platform (subroutines ode15s and fminsearch).

Figure 5. Curves of the gas release rate versus time for heating temperatures of (A) 530−605 K and (B) 620−690 K.

Figure 6. Yields of product classes (char, gas, organics, and water) and CO and CO2 from glucomannan pyrolysis versus heating temperature.

Figure 7. Total ion chromatogram of the pyrolysis liquid obtained for Th = 620 K. The peak numbers correspond to the species listed in Table 1. 5035

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

Table 1. Pyrolysis Liquid Compounds Identified by GC/MS with Retention Times (RT) Reported in Figure 7a no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 a

species oxygen formaldehyde acetaldehyde methyl alcohol ethyl acetate 2,3-butanedione 2-butanone 3-pentanone formic acid hydroxyacetaldehyde 2-butenal acetic acid 2,3-pentanedione hydroxypropanone 3-hydroxy-2-butanone propanoic acid 2-propenoic acid cyclopentenone 2-methylpyrimidine 1-hydroxy-2-butanone 1,2-ethanediol, monoacetate 2-methylpropanoic acid 3-furaldehyde butanedial 2-cyclopentenone furfural 4-hydroxy-4-methylpentanone 2,5-dimethylpyrazine 2,4-dimethylpyrazine 2-methyltetrahydro-2-furanol furfuryl alcohol acetoxyacetone 2-methyl-2-cyclopentenone 2-methylpropanoic anhydride 2-butanone acetylfuran vinylcrotonate

RT [min]

match [%]

4.13 4.33 4.58 4.68 7.00 7.15 7.21 7.57 8.46 8.59 9.36 9.85 10.47 11.51 13.15 14.15 15.08 15.77 16.03 16.11 16.36 16.70 17.86 18.59 19.22 19.29 20.01 20.34 20.99 21.25 21.65 22.12 22.21 22.40 22.53 23.00 23.50

99.9 96.7 91.7 94.7 92.6 96.8 87.6 83.1 i i 91.8 i 86.0 i 90.6 i 83.4 87.9 89.8 i 79.1 84.0 93.8 91.4 90.3 i 85.3 85.0 80.0 89.0 i i i 88.0 88.3 i 86.0

no.

species

RT [min]

match [%]

38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73

2-hydroxy-2-cyclopenten-1-one furfurylalcohol, acetate 5-methyl-2-furaldehyde 1-acetoxy-2-butanone 3-methyl-2-cyclopentenone butyrolactone 2(5H)-furanone 3-methyl-2,5-furandione 2-hydroxy-1-methylcyclopenten-3-one 3-methyl-2(5H)-furanone 2,3-dimethyl-2-cyclopenten-1-one 2-furanone, 2,5-dihydro-3,5-dimethyl phenol 3-furancarboxylic acid, methyl ester o-cresol 3-ethyl-2-hydroxy-2-cyclopenten-1-one 3-hydroxy-2-methyl-4H-pyran-4-one (maltol) 2-hydroxy-γ-butyrolactone p-cresol m-cresol 2-methylpropanoic anhydride 5-(hydroxymethyl)dihydro-2(3H)-furanone unknown sugar unknown sugar 1,4:3,6-dianhydro-α-D-glucopyranose 5-acetoxymethyl-2-furaldehyde 5-hydroxymethyl-2-furaldehyde 1,2-benzenediol 5-(hydroxymethyl)dihydro-2(3H)-furanone 2-methyl-5-hydroxybenzofuran hydroquinone 2-methyl-1,4-benzenediol 1,6-anhydro-β-D-glucofuranose unknown sugar unknown sugar levoglucosan

24.97 25.72 26.48 26.84 27.28 27.75 28.26 29.49 30.02 30.41 30.66 30.98 31.76 33.96 34.10 34.32 34.57 34.90 35.80 35.94 36.09 36.44 37.26 40.06 42.74 45.01 45.58 45.84 46.26 49.52 50.60 52.21 56.45 58.54 58.64 59.06

97.0 86.2 i 84.5 i 95.0 i 83.3 95.6 80.6 88.1 i 95.4 i i 89.3 86.3 i i 91.5 88.8 − − 91.1 84.6 i 91.4 91.8 84.7 i 84.3 − − − i

The percentage indicates the match between the species spectrum and the NIST library; “i” indicates injected species.

The weight loss curves are interpreted by means of either a one-stage mechanism consisting of two competitive reactions for the formation of volatile species (V) and char (C):

The reaction rates present the usual Arrhenius dependence on temperature and a first-order dependence on the solid mass fraction. It is worth observing that one-stage mechanisms of components can be useful for formulating multistep pyrolysis mechanisms42,43 which can be applied to describe with sufficient accuracy the degradation behavior of different biomass fuels once the chemical composition is known. One-stage mechanisms, where the global kinetic constant is determined, can also be combined with the information on the yields of the product classes dependent on the temperature to determine their respective formation rates.32 More detailed mechanisms, such as the two-stage mechanism proposed here, can also be useful, although the complexity of transport models incorporating such kinetics will increase. However, in the mathematical modeling of torrefaction processes, where only the decomposition of the hemicellulose component is of interest, a two-stage mechanism is expected not to introduce significant difficulties. Results in terms of kinetic parameters and deviations between predictions and measurements (these defined as in ref 44) are summarized in Table 2. Predicted and measured weight loss curves are compared in Figure 9. In all cases, in agreement with

or a two-stage mechanism describing the competitive formation of volatiles (V1) and a solid intermediate (B) and the successive degradation of the latter into volatiles (V2) and char (C) by two competitive reactions:

5036

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

for char formation and 96 kJ/mol versus 114−143 kJ/mol for volatile formation). These differences can be attributed to the different nature of the hardwood hemicellulose with respect to glucomannan, the possible interaction between components, and the changes introduced in the component structure by the extraction method. Moreover, although isothermal curves of weight loss are not available for softwoods for a comparison, it can be hypothesized that the different origin/properties of glucomannan may also play a role. Even higher activation energies are reported in ref 42 for the reaction of hemicellulose decomposition (202 and 146 kJ/mol for tar and char formation, respectively), but it is observed that, in this case, decomposition begins significantly later than observed in the experiments. The two-stage mechanism is the same as that proposed for xylan degradation.11 Similarly, it foresees a first very fast stage (ratios of the characteristic time to the total conversion time of about 15 to 3) where a large part of the volatiles (yields between 50 and 70%) is released. The kinetic parameters, in particular, the activation energies, are comparable with those already estimated for the onestage mechanism. The second stage is much slower and essentially describes the transformation of the solid-phase reaction intermediate into char. On the whole, the estimated activation energies for the reactions of the two stages are comparable with those obtained for xylan degradation.

Figure 8. Yields of some organic compounds generated from glucomannan pyrolysis as functions of the heating temperature (acetic acid (AA), hydroxyacetaldehyde (HAA), hydroxypropanone (HP), formic acid (FA), propionic acid (PA), hydroxybutanone (HB), acetylfuran (AF), furfuryl alcohol (FAL), and 2(5H)-furanone (2(5H)Fone)).

Table 2. Kinetic Parameters for Glucomannan Pyrolysis Described by Means of One-Stage and Two-Stage Mechanisms one-stage mechanism

two-stage mechanism

EV = 96.2 kJ/mol

EV1 = 98.5 kJ/mol

AV = 4.42 × 106 s−1

AV1 = 1.57 × 107 s−1

EC = 52.5 kJ/mol AC = 3.28 × 102 s−1 dev % = 2.35

EB = 54.5 kJ/mol AB = 2.64 × 103 s−1 EV2 = 70.1 kJ/mol

4. CONCLUSIONS The pyrolysis of commercial glucomannan has been investigated by means of a small-scale, batch-operated fluidized bed reactor over a range of heating temperatures of interest for the torrefaction process. The actual heating rates and temperatures experienced by the sample have been found to be affected by the ratio of the mixing time (hot sand and glucomannan particles) and the time needed by the glucomannan particles to reach the softening point. For relatively low heating temperatures (up to about 605 K), the mixing time is shorter than the time needed to achieve the softening point. The glucomannan particles preserve their identity, are distributed across the expanded bed, and undergo pyrolysis at about the same temperature independently from their position. The process is not rigorously isothermal, but the bed temperature weakly increases with time and remains always below the initial heating temperature. For heating temperatures above 605 K, glucomannan particles reach the softening point before mixing with the hot sand occurs. Consequently, agglomerates are formed which are positioned in the upper part of the expanded bed and react in the presence of highly variable temperatures depending on the bed position and time. The characteristic process size is related not to the original particles any longer but to the agglomerates with actual heating rates and reaction temperatures that decrease as the heating conditions are made more severe. The yields of the lumped classes of products, char, water, organics, and gas show the usual trends versus the reaction temperature already reported for lignocellulosic fuels. Char and water are the most abundant products. The gas consists of CO2 and CO. The organic fraction is constituted of carboxylic acids (AA and FA), carbonyl compounds (HAA, HP), furan compounds (FF and mainly FAL), and minor amounts of carbohydrates. Weight loss curves of glucomannan have been measured under kinetic control and isothermal conditions (temperatures in the range 503−593 K). A one-stage pyrolysis mechanism provides an acceptable description of the process with activation energies for the formation of char and volatile species of 52 and 96 kJ/mol, respectively. A better description is obtained with a two-stage

AV2 = 2.10 × 104 s−1 EC = 48.2 kJ/mol AC = 5.21 × 101 s−1 dev % = 1.63

Figure 9. Measured (symbols) and predicted (dashed and solid lines for a one-stage mechanism and a two-stage mechanism) mass loss curves of glucomannan pyrolysis at various temperatures.

measurements, the yields in the solid residue progressively diminish as the reaction temperature is increased, indicating that the formation of solid-phase products is associated with a lower activation energy compared with that of volatile products. As expected, the two-stage mechanism, based on a higher number of parameters, produces a better agreement with measurements, but the one-stage mechanism also gives acceptable results. In this case, the estimated activation energies for the formation of volatile species and char are lower than those previously estimated for the first step of hardwood degradation45,46 that can be associated with hemicellulose decomposition (values of 52 kJ/mol versus 76 kJ/mol 5037

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

(15) Di Blasi, C.; Branca, C.; Santoro, A.; Gonzalez Hernandez, E. Pyrolytic behavior and products of some wood varieties. Combust. Flame 2001, 124, 165−177. (16) Di Blasi, C.; Branca, C.; Santoro, A.; Perez Bermudez, R. A. Weight loss dynamics of wood chips under fast radiative heating. J. Anal. Appl. Pyrolysis 2001, 57, 77−90. (17) Asmadi, M.; Kawamoto, H.; Saka, S. Pyrolysis reactions of Japanese cedar and Japanese beech woods in a closed ampoule reactor. J. Wood Sci. 2010, 56, 319−330. (18) Stamm, J. Thermal degradation of wood and cellulose. Ind. Eng. Chem. 1956, 48, 413−417. (19) Ramiah, M. V. Thermogravimetric and differential thermal analysis of cellulose, hemicellulose and lignin. J. Appl. Polym. Sci. 1970, 14, 1323−1337. (20) Alen, R.; Rytkonen, S.; McKeough, P. Thermogravimetric behavior of black liquors and their organic constituents. J. Anal. Appl. Pyrolysis 1995, 31, 1−13. (21) Hosoya, T.; Kawamoto, H.; Saka, S. Cellulose−hemicellulose and cellulose−lignin interactions in wood pyrolysis at gasification temperature. J. Anal. Appl. Pyrolysis 2007, 80, 118−125. (22) Hosoya, T.; Kawamoto, H.; Saka, S. Pyrolysis behavior of wood and its constituent polymers at gasification temperature. J. Anal. Appl. Pyrolysis 2007, 78, 328−336. (23) Gao, S.; Wu, C.; Nishinari, K. In situ pH-decrease-induced gelation of sodium alginate/carboxymethylated Konjac glucomannan. J. Appl. Polym. Sci. 2008, 108, 2825−2832. (24) Di Blasi, C.; Branca, C. Temperatures of wood particles in a hotsand bed fluidized by nitrogen. Energy Fuels 2003, 17, 247−254. (25) Scott, D. S.; Majerski, P.; Piskorz, J.; Radlein, D. A second look at fast pyrolysis of biomassthe RTI process. J. Anal. Appl. Pyrolysis 1999, 51, 23−37. (26) Di Blasi, C. Modelling the fast pyrolysis of cellulosic particles in fluidized bed reactors. Chem. Eng. Sci. 2000, 55, 5999−6013. (27) Di Blasi, C. Modelling intra- and extra-particle processes of wood fast pyrolysis. AIChE J. 2002, 48, 2386−2397. (28) Gogolek, P. E. G.; Grace, J. R. Fundamental hydrodynamics related to pressurized fluidized bed combustion. Prog. Energy Combust. Sci. 1995, 21, 419−451. (29) Di Blasi, C.; Branca, C.; Galgano, A. Effects of diammonium phosphate on the yields and composition of products from fir wood pyrolysis. Ind. Eng. Chem. Res. 2007, 46, 430−438. (30) Di Blasi, C.; Branca, C.; Galgano, A. Products and global weight loss rates of wood decomposition catalyzed by zinc chloride. Energy Fuels 2008, 22, 663−670. (31) Di Blasi, C.; Galgano, A.; Branca, C. Influences of the chemical state of alkaline compounds and the nature of alkali metal on wood pyrolysis. Ind. Eng. Chem. Res. 2009, 48, 3359−3369. (32) Di Blasi, C.; Branca, C. Kinetics of primary product formation from wood pyrolysis. Ind. Eng. Chem. Res. 2001, 40, 5547−5556. (33) Goring, D. A. I. Thermal softening of lignin, hemicellulose and cellulose. Pulp Pap. Mag. Can. 1963, 64 (12), T517−T527. (34) Di Blasi, C. Modeling chemical and physical processes of wood and biomass pyrolysis. Prog. Energy Combust. Sci. 2008, 34, 47−90. (35) Di Blasi, C. Combustion and gasification rates of lignocellulosic chars. Prog. Energy Combust. Sci. 2009, 35, 121−140. (36) Prins, M. J.; Ptasinski, K. J.; Janssen, F. J. J. G. Torrefaction of wood Part 2. Analysis of products. J. Anal. Appl. Pyrolysis 2006, 77, 35−40. (37) Bridgeman, T. G.; Jones, J. M.; Shield, I.; Williams, P. T. Torrefaction of red canary grass, wheat straw and willow to enhance solid fuel qualities and combustion properties. Fuel 2008, 87, 844− 856. (38) Wannapeera, J.; Fungtammasan, B.; Worasuwannarak, N. Effects of temperature and holding time during torrefaction on the pyrolysis behavior of woody biomass. J. Anal. Appl. Pyrolysis 2011, 92, 99−105. (39) Wannapeera, J.; Worasuwannarak, N. Upgrading of woody biomass by torrefaction under pressure. J. Anal. Appl. Pyrolysis 2012, 96, 173−180.

mechanism similar to that previously proposed for xylan degradation.

■

ASSOCIATED CONTENT

* Supporting Information S

Table S1, parameters characterizing glucomannan pyrolysis in a fluidized-bed reactor; Table S2, yields of the main organic compounds generated from glucomannan pyrolysis as a function of the heating temperature. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Tel.: 39-081-7682232. Fax: 39-081-2391800. E-mail: diblasi@ unina.it. Present Address §

Institute of Chemical and Environmental Engineering, Faculty of Chemical and Food Technology, Slovak University of Technology, Radlinskeho 9 812 37 Bratislava, Slovak Republic. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Theander, O. Cellulose, hemicellulose and extractives. In Fundamentals of Thermochemical Biomass Conversion; Overend, R. P., Milne, T. A., Mudge, L. K., Eds.; Elsevier Applied Science Publishers: London, 1985; pp 35−60. (2) Di Blasi, C.; Branca, C.; Galgano, A. Biomass screening for the production of furfural via thermal decomposition. Ind. Eng. Chem. Res. 2010, 49, 2658−2671. (3) Moreira, L. R. S.; Filho, E. X. R. An overview of mannan structure and mannan-degrading enzyme systems. Appl. Microbiol. Biotechnol. 2008, 79, 165−178. (4) Sjostrom, E. Wood Chemistry: Fundamentals and Applications; Academic Press: New York, 1981. (5) van der Stelt, M. J. C.; Gerauser, H.; Kiel, J. H. A.; Ptasinski, K. J. Biomass upgrading by torrefaction for the production of biofuels: a review. Biomass Bioenergy 2011, 35, 3748−3762. (6) Branca, C.; Albano, A.; Di Blasi, C. Critical evaluation of wood devolatilization mechanisms. Thermochim. Acta 2005, 429, 133−141. (7) Brostrom, M.; Nordin, A.; Pommer, L.; Branca, C.; Di Blasi, C. Influence of torrefaction on the devolatilization and oxidation kinetics of wood. J. Anal. Appl. Pyrolysis 2012, 96, 100−109. (8) Simkovic, I.; Varhegyi, G.; Antal, M. J.; Ebringerova, A.; Szekeley, T.; Szabo, P. Thermogravimetric/mass spectrometric characterization of the thermal decomposition of (4-O-methyl-D-glucurono)-D-xylan. J. Appl. Polym. Sci. 1988, 36, 721−728. (9) Bilbao, R.; Millera, R. A.; Arauzo, J. Kinetics of weight loss by thermal decomposition of xylan and lignin. Thermochim. Acta 1989, 143, 137−148. (10) Varhegyi, G.; Antal, M. J.; Szekely, T.; Szabo, P. Kinetics of the thermal decomposition of cellulose, hemicellulose and sugar cane bagasse. Energy Fuels 1989, 3, 329−335. (11) Di Blasi, C.; Lanzetta, M. Intrinsic kinetics of isothermal xylan degradation in inert atmosphere. J. Anal. Appl. Pyrolysis 1997, 40−41, 287−303. (12) Nowakowski, D. J.; Woodbridge, C. R.; Jones, J. M. Phosphorus catalysis in the pyrolysis behaviour of biomass. J. Anal. Appl. Pyrolysis 2008, 83, 197−204. (13) Nowakowski, D. J.; Jones, J. M. Uncatalysed and potassiumcatalysed pyrolysis of the cell-wall constituents of biomass and their model compounds. J. Anal. Appl. Pyrolysis 2008, 83, 12−25. (14) Collard, F. X.; Blin, J.; Bensakhria, A.; Valette, J. Influence of impregnated metal on the pyrolysis conversion of biomass constituents. J. Anal. Appl. Pyrolysis 2012, 95, 213−226. 5038

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039

Industrial & Engineering Chemistry Research

Article

(40) Zheng, A.; Zhao, Z.; Chang, S.; Huang, Z.; He, F.; Li, H. Effect of torrefaction temperature on product distribution from two-staged pyrolysis of biomass. Energy Fuels 2012, 26, 2968−2974. (41) Branca, C.; Di Blasi, C.; Galgano, A. Pyrolysis of corncobs catalyzed by zinc chloride for furfural production. Ind. Eng. Chem. Res. 2010, 49, 9743−9752. (42) Miller, R. S.; Bellan, J. A generalized biomass pyrolysis model based on superimposed cellulose, hemicellulose and lignin kinetics. Combust. Sci. Technol. 1996, 126, 97−137. (43) Manya, J. J.; Arauzo, J. An alternative kinetic approach to describe the isothermal pyrolysis of micro-particles of sugar cane bagasse. Chem. Eng. J. 2008, 139, 549−561. (44) Branca, C.; Di Blasi, C.; Horacek, H. Analysis of the combustion kinetics and the thermal behaviour of an intumescent system. Ind. Eng. Chem. Res. 2002, 41, 2104−2114. (45) Branca, C.; Di Blasi, C. Kinetics of the isothermal degradation of wood in the temperature range 528−708K. J. Anal. Appl. Pyrolysis 2003, 67, 207−219. (46) Prins, M. J.; Ptasinski, K. J.; Janssen, F. J. J. G. Torrefaction of wood Part 1. Weight loss kinetics. J. Anal. Appl. Pyrolysis 2006, 77, 28− 34.

5039

dx.doi.org/10.1021/ie400155x | Ind. Eng. Chem. Res. 2013, 52, 5030−5039