- Email: [email protected]
Ash characteristics of high alkali sawdust and sanderdust biomass fuels
Twenty-Seventh Symposium (International) on Combustion/The Combustion Institute, 1998/pp. 1719–1725
ASH CHARACTERISTICS OF HIGH ALKALI SAWDUST AND SANDERDUST BIOMASS FUELS BLAKE C. CHENEVERT,1 JOHN C. KRAMLICH1 and KENNETH M. NICHOLS2 1Department of Mechanical Engineering University of Washington Seattle, WA 98195-2600, USA 2Weyerhaeuser Company Tacoma, WA 98477, USA
Suspension firing of sawdust and sanderdust fuels is often used in the wood product’s industry to raise steam and provide a heat source for drying and curing operations. The unusually high alkali content of these fuels can give rise to a number of problems that affect the operation of downstream plant systems. The research reported here focuses on the physical and chemical properties of the ash generated by these unique and important biomass fuels and uses this information to identify the mechanisms that control mineral-to-ash transformations. Four sanderdust and sawdust fuels, typical of those used to fire industrial-scale suspension burners, are fired in a laboratory-scale tunnel furnace. Size distribution, morphology, and size versus composition are obtained for particles between 0.0075 and 10 lm by combining a cascade impactor and an electrical aerosol analyzer (EAA). Each of the fuels showed a dominant mode of calcium-rich skeletal particles of size $8.3 lm that are the residue from char burnout. A second, minor mode that seems to be the result of fragmentation appears at 1–3 lm. This consists of fluxed particles that, while still predominantly calcium, also contain Fe, Al, Mn, and Si. Another minor mode at 0.4 lm also appears to be the result of fragmentation. Much of the alkali mineral matter becomes submicron aerosol via the vaporization, condensation, coagulation mechanism. This large yield of aerosol (of the order of 30% of the total ash mass) appears as chlorides in high chlorine fuels and as sulfates and carbonates otherwise. In general, only a small fraction of the alkali metals are captured by the residual ash, and no metals other than Na and K are generally detected with the aerosol. The aerosol size varies between 0.01 and 0.1 lm depending on experimental conditions.
Introduction Sawdust and sanderdust are by-products from the manufacture of particle board and strand board in the wood product’s industry. They are important sources of fuel for drying and curing operations, as well as for raising process steam. Some difficulties arise, however, due to the unique mineral composition of these fuels. In particular, alkali species can constitute more than 50% of the mineral matter. This has led to unforeseen problems as new, stateof-the-art pollution-control equipment has been installed into existing plants. The problems are most evident in the degradation of downstream volatile organic oxidizers that are used to eliminate vapors generated by drying and curing. These problems appear to involve a complex interaction between the ash, the organic vapors, and the particulate control systems that precede the oxidizers. Understanding the nature of the ash generated by these fuels is clearly important to addressing these challenges. Although the biomass fuels considered here differ from coal in many respects, they share many of the same ash-transformation mechanisms. This includes
the generation of residual ash via char burnout and fragmentation and the potential for sintering or melting of the ash, depending on composition [1,2]. The high alkali mineral content can be expected to result in considerable submicron aerosol formation due to the vaporization, condensation, coagulation mechanism [3]. However, irreversible capture of alkali vapor by other minerals, particularly aluminosilicates, can reduce the aerosol yield. The efficiency of this capture depends on the composition of the nonalkali minerals, on the original form of the alkali, and on the physical proximity of the capturing minerals relative to the source of the alkali [4–7]. The alkali minerals can also promote sintering of the residual ash. Although the behavior of the minerals in biomass fuels has been considered, particularly from the standpoint of ash deposition [8], the wide variability in biomass mineral compositions and the practical importance of sawdust and sanderdust fuels requires that their specific behavior be understood. The overall goal of the work reported here is (1) the characterization of the ash generated by sawdust and sanderdust combustion and (2) the identification of the
1719
1720
OTHER POLLUTANTS
Fig. 1. Tunnel furnace schematic.
underlying mechanisms that govern the fate and behavior of the minerals in this unique fuel. Experimental Combustion experiments are performed using the down-fired tunnel furnace (15 kW) shown in cross section in Fig. 1. A high-temperature refractory forms the inner walls of the combustion tunnel, with heat loss to the surroundings minimized by three layers of progressively lower thermal conductivity insulating material. The primary flame is provided by a swirl-stabilized natural gas and air burner located at the top of the reactor. A 35.6-cm-long by 10.2cm-diameter combustion zone (prior to the point of pulverized fuel injection) holds the flame. The solid fuel is metered by a screw feeder, entrained in an air stream, and pneumatically introduced into the furnace via a stainless steel, watercooled injector. The fuel enters the furnace 2.5 cm above the venturi constriction shown in the figure, which promotes uniform dispersion of the fuel particles. After the venturi, the tunnel expands to 20.3 cm diameter. The temperature profile through the furnace can be modified by operating pairs of auxiliary natural gas–fired burners that indirectly heat
the refractory. The solid fuels are injected at 1400 K. The temperature profile in the experimental test section decreases at 200 K/s, and the overall residence time is 1.25 s. The total furnace flow rate is 9.25 g/s, and the solid fuel flow rate is 0.077 g/s, except where noted. The main burner operates at 200% theoretical air, not accounting for the solid fuel. This yields approximately 9% O2 (mole basis, dry) in the flue gas without the solid fuels. The present research examines four sanderdust and sawdust fuels. Various wood-products’ manufacturing facilities supplied the samples, which were selected to be representative of typical fuels used in suspension burners at the given facility. The fuels are briefly described as follows: Fuel 1. A milled sanderdust from composite boards made from urea-formaldehyde resin and Southern pine. Fuel 2. A fuel that uses the same resin as fuel 1, but it consists of 50% hard wood (oak, hickory) and the balance soft wood (gum, beech). Fuel 3. A fuel that includes phenol formaldehyde resin, and it consists of 60% Virginia pine, 30% poplar, and 10% miscellaneous hardwoods. Fuel 4. A material composed of particles that are too fine for strand board; these are sieved out before
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
1721
TABLE 1 Fuel composition
Ultimate analysis (% mass, oven dry basis)
Moisture (% mass, as received) Ash at 750 8C (% mass, OD basis) Major mineral components (mg/kg raw fuel)
C N H S O by D
Ca Na K Mg Mn Fe P Si
Total halogens (mg/kg raw fuel)
resin is added. Thus, except for some miscellaneous wastes, this stream is largely unresinated. It consists of 65% aspen, 15% red pine, and 20% mixed hardwoods. Ultimate analyses, mineral elemental analysis via atomic absorption, and total halogens’ analysis were performed and are summarized in Table 1. The mineral matter is predominantly made up of calcium, sodium, and potassium. Note that fuels 1 and 2 have an order of magnitude higher halogen content than fuels 3 and 4. Significant differences in nitrogen, alkali, and halogen content between the fuels are the result of manufacturing additives (i.e., binding resins), parent wood variety, and other processing sources. The as-received fuels are milled in a Thomas–Wiley laboratory mill to pass through a 0.5mm-diameter screen. This was primarily done to allow the fuels to be compared on an identical size basis. The milled fuels have an essentially identical size distribution, being entirely between 0.1 and 0.6 mm. Ash samples are collected using a water-cooled sampling probe. A porous inner wall provides a path for dilution air. The probe is designed to furnish isokinetic sampling and to minimize thermophoretic deposition of particles on the probe wall. The CO2 content of the undiluted and diluted gas streams provides an accurate tracer to verify isokinetic conditions and calculate dilution rates. The ash size distribution is determined by an eight-stage Anderson cascade impactor (stages numbered 0 through 7). This includes a backup ultimate filter as a ninth stage to catch particles below 0.4 lm. In this configuration, the impactor allows measurement of the total yield of aerosol but cannot provide its specific size.
Fuel 1
Fuel 2
Fuel 3
Fuel 4
47.5 5.2 5.5 0.033 40.9 4.2 0.81 1640 1000 674 249 94 198 65 403 1215
47.7 3.4 5.8 0.030 42.0 4.1 1.07 2410 1500 1020 281 91 43 80 263 2130
50.8 0.19 5.6 0.062 42.6 5.2 0.66 1120 480 454 233 112 308 57 241 82
49.9 0.10 6.4 0.038 42.3 4.2 1.29 3930 870 1160 405 76 75 118 200 137
For those tests in which the aerosol is sized, the last two stages of the impactor and the ultimate filter are removed, and a TSI electronic aerosol analyzer (EAA) is substituted. The aerosol analyzer sizes the material that passes the truncated impactor into “bins” that cover the range of 0.0075–1.0 lm. A system is added ahead of the aerosol analyzer to provide variable delay times before analysis to represent the various low-temperature holdup times experienced by the ash in practical systems. The impactor stages and filter are examined by scanning electron microscopy (SEM), and stagewise elemental analyses are obtained by energy-dispersive X-ray probing (EDAX). Results and Discussion Figure 2 presents size-distribution data for fuels 1 and 2. All plots that follow are mass size distributions. The assumptions used to reduce the impactor data are (1) the particles are solid spheres, (2) the particle density is the same as that of CaCO3 (2.7 g/ cc), and (3) the mass on each impactor stage is assigned the diameter that corresponds to a 50% probability of retention on that stage (d50). While the results are not strongly sensitive to assumed particle density, the shape and porosity of the particles can influence capture. The mass on the ultimate filter, which corresponds to the aerosol mode, is nominally plotted at 0.1 lm based on past work. The determination of the true size of this mode is discussed later. Carbon burnout on all stages is essentially complete. Thus, the size distribution shows three distinct modes of ash: (1) a residual mode of $8.3 lm that
1722
OTHER POLLUTANTS
Fig. 2. Impactor size distribution for fuels 1 and 2. Fuel 1 is shown by open symbols and fuel 2 by solid symbols.
Fig. 4. Ash morphology and overall elemental analyses by mass for fuel 1, 1.6-lm size cut.
Fig. 3. Ash morphology and overall elemental analyses by mass for fuel 1, 8.3-lm size cut.
represents about 40% of the total mass, (2) a minor mode at 1–3 lm that is about 20% of the ash, and (3) a mode too fine to be captured by the smallest impactor stage that includes around 30% of the ash. Figure 3 shows an SEM of a selection of material collected on stage 0 for fuel 1 (d50 4 8.3 lm). Much of the material is highly porous and suggestive of the skeletal remains of unfused ash left as a residual by char burnout. The application of EDAX to the material results in the typical composition shown as the inset in the figure, indicating that the stage is dominated by calcium and a small amount of magnesium.
The generation of a melt by CaO under these conditions is, of course, clearly not indicated. This does not, however, suggest the absence of some surface mobility that is capable of modifying the morphology of the CaO ash. Under atmospheres containing CO2, the repeated formation and decomposition of metastable CaCO3 above the nominal CaCO3 decomposition temperature of 1170 K essentially catalyzes surface mobility. At the temperatures used in this study, the effect of this catalyzed sintering is to relax the minimum radius of curvature present in any CaO structure to an asymptotic value of about 0.25 lm [9,10]. This clearly will not lead to a collapse of the skeletal structures present in Fig. 3, but it will influence the minimum size of the components that make up the structure. Figure 4 shows an example of material collected on stage 4 (d50 4 1.6 lm), which is at the center of the 1–3-lm mode shown in Fig. 2. The inset in the figure shows that the composition includes small amounts of Fe, Al, Mn, Si, and Mg. High magnification shows that the field is dominated by three types of morphology: (1) lacy structures, (2) rounded particles that show evidence of extensive fluxing, and (3) small particles of the order of 0.4 lm in size that appear as nodes on the larger rounded particles. Pointwise EDAX shows that the lacy structures are relatively high in calcium and deficient in the other elements indicated in Fig. 4; these are likely fragments of the skeletal residual ash discussed earlier. The composition of the rounded particles and the nodes are essentially identical, being approximately
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
Fig. 5. Ash morphology and overall elemental analyses by mass for fuel 1, 0.4-lm size cut.
Fig. 6. Ash morphology and overall elemental analyses by mass for fuel 1, ultimate filter.
1723
65–70% Ca with the balance being Fe, Al, Mn, Si, and Mg. Figure 5 shows material collected on stage 7, with a d50 of 0.4 lm. Although Fig. 2 shows that the material on this stage does not significantly contribute to the overall mass, the very uniform character of the material collected here suggests its source originates in a single mechanism. Also, these particles are identical in appearance and composition to the nodes associated with the rounded particles in Fig. 4. The material has an almost gravel-like appearance, which under higher magnification appears as a meltlike rounding with some angularity preserved. The elemental analysis shows the appearance of some sodium and potassium on this stage. If, however, the alkali metals are removed and the composition renormalized, the remaining elements have the same composition as the higher-diameter stages. This suggests that this material has a similar origin but that it is contaminated either by alkali aerosol or by direct condensation. The rounded appearance of these particles may be due to fluxing induced by the other elements present. As discussed earlier, however, pure CaO under a CO2 atmosphere will sinter under these temperatures to the point where the radii of curvature is of the order of 0.25 lm, which is similar to the radii of these particles. The appearance of these particles is consistent with partial sintering rather than complete melting. All four fuels showed essentially identical behavior on all stages above the aerosol filter. The nonalkali ash thus appears to include (1) a large predominance of skeletal, high-calcium structures and (2) a lesser amount of separate, smaller particles that are rich in calcium, but contain other elements. These show signs of more extensive fluxing. A similar size mode has been observed in coal ash [11,12]. In coals, this class of particles increases in importance as the coal is more finely ground, but the yield does not change when the coal is cleaned by froth flotation. This indicates the mode is not the result of fine inclusions released by grinding, as these would be removed by cleaning. It was suggested that the early stages of char combustion result in the release of fine particles, but the accumulation of minerals at the char surface prevent further release during later stages of combustion [12]. A similar mechanism may be active here, resulting in the release of fine fragments during the early stages in char combustion. These particles are primarily calcium, but they include elevated concentrations of the trace metals previously mentioned. A high magnification view of the aerosol filter for fuel 1 is given by Fig. 6, which shows borosilicate glass fibers with agglomerates of very fine particles attached. The size of the individual particles making up the agglomerates is #0.1 lm. EDAX analysis of this stage requires the removal of the element signatures arising from the borosilicate glass. This renormalization produced the elemental composition
1724
OTHER POLLUTANTS
Fig. 7. Elemental composition of material collected on the aerosol filter for each fuel.
Fig. 8. Aerosol size distributions for three delay times for fuel 1.
shown for fuel 1 in Fig. 7. The results indicate that the aerosol catch is almost entirely alkali elements. The chlorine signal is essentially that which would be generated by stoichiometric NaCl and KCl. Although it has been shown that high chlorine content enhances alkali release from solid fuels [5], there is some evidence that a fraction of the alkali species remains with the residual ash. Trace amounts of alkali, relative to the dominant Ca species, are evident by SEM in size cuts of 0.4 lm and larger. The low chlorine-to-alkali molar ratios (0.0 to 0.22) in these larger sized particles suggests that the detected alkali is not deposited via NaCl or KCl condensation, because these are the favored vaporphase species. More likely, this alkali was retained without chlorine bonding, possibly by complexing with other minerals. Fuel 2 (also containing a high chlorine content) has a behavior essentially identical to that of fuel 1. Fuels 3 and 4 are deficient in chlorine. While other features of the ash from these fuels are similar to fuels 1 and 2 (i.e., size distribution and nonaerosol composition), the morphology of the aerosol filter catch and its composition are significantly different.
Specifically, the deposits appear to be less feathery and more like a collection of fine droplets. Figure 7 shows that the absence of chlorine results in a higher concentration of sulfur appearing on the filter. Equilibrium analysis shows that alkali hydroxides and chlorides are favored to appear in the gas phase. The first species predicted to condense is Na2SO4. This condensation is, however, accompanied by a reaction, which suggests that a chemical kinetic barrier exists for the rapid formation of this compound. Alternately, NaCl(s) and KCl(s) are formed directly from identical gas-phase species and thus face no chemical kinetic barrier. The lack of sulfur in the aerosol solids for fuels 1 and 2 indicates that sulfate is not formed during the condensation event and that sulfur is not effective at displacing chlorine as an anion following condensation. For the chlorinedeficient fuels (fuels 3 and 4), the initial condensation is expected to be as a hydroxide, with subsequent pickup of CO2 or sulfur following condensation. The filter analysis shows that in the absence of chlorine, the aerosols do capture sulfur. Equilibrium suggests that the alkali condensation temperature is around 850 K. Direct measurements of the aerosol size distribution were obtained for each of the fuels. One motivation for these measurements is that, in practical facilities, a number of parameters that influence aerosol coagulation are also process variables. Most importantly, a large amount of dilution air is added at points that vary from immediately downstream of the flame to points much further downstream where condensation and coagulation have been essentially completed. Thus, the time to dilution, and the amount of dilution air used, can influence the size of the aerosol that approaches the back-end equipment. Figure 8 shows EAA size distributions for fuel 1 as a function of the delay time between the sample point and the analyzer (here, the fuel flow is 0.0283 g/s). For the conditions used here, coagulation is clearly incomplete in both the “no delay” and 21-s delay tests. For the 198-s delay, the particles are around 0.1 lm in size. This is the practical upper limit for Brownian motion-induced coagulation of aerosol particles. Other data (not plotted) show the expected behavior with solid’s firing rate, that is, that lower firing densities lead to a smaller mean size for the aerosol over the same delay time. All of these trends are expected from coagulation theory. In the practical system, delay times are much shorter than those used in the present experiment, while firing densities are higher. Thus, the two effects tend to counteract. Nucleation, condensation, and coagulation modeling will be used to evaluate both the present data and the influence of the variety of environments presented by practical systems on the final aerosol size distribution. This is, however, beyond the scope of the present work.
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
1725
Conclusions
REFERENCES
For each of the fuels examined, calcium is the dominant element in all of the nonaerosol fractions. It appears in conjunction with a small amount of Mg in skeletal residues larger than 10 lm. Other elements (Fe, Al, Mn, and Si) are found with calcium in a 1–3-lm mode that appears to result from char or mineral fragmentation. The presence of these minerals is marked by evidence of fluxing in the ash. The fact that these other elements are concentrated into this minor 1–3-lm mode suggests that these elements may be initially physically separated from the bulk of the calcium in the original mineral matter. Otherwise, one would expect them to be more uniformly distributed throughout the calcium in the ash. The data show some evidence of capture of Na or K by the residual ash, a process that does occur in coal combustion. The small amount of Na and K that appears with the 0.4-lm cut appears to be due to condensation or coagulation of the alkali onto the existing particles or stray capture of aerosol by the last impactor stage before the ultimate filter. The alkali metals thus appear predominantly as a submicron aerosol that is due to vaporization, condensation, and coagulation. When sufficient chlorine is present, the aerosols are present as NaCl and KCl. Although equilibrium favors sulfate formation, sulfur is unable to displace chlorine. In the absence of chlorine, sulfates appear in the aerosol. The aerosol size varies (between 0.01 and 0.1 lm) depending on fuel density (decreases size) and delay time before analysis (increases size) in accordance with coagulation theory. The wide variation in dilution rates and location of dilution found in practical systems makes aerosol size a potentially controllable variable.
1. Miccio, F. and Salatino, P., in Twenty-Fourth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1992, pp. 1145–1151. 2. Mitchell, R. E. and Akanetuk, A. E. J., in Twenty-Sixth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1996, pp. 3137–3144. 3. Flagan, R. C. and Friedlander, S. K., “Particle Formation in Pulverized Coal Combustion—A Review,” in Recent Developments in Aerosol Science (D. T. Shaw, ed.) Wiley, New York, 1978. 4. Neville, M. and Sarofim, A. F., Fuel 64:384–390 (1985). 5. Linder, E. R. and Wall, T. F., in Twenty-Third Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1990, pp. 1313–1321. 6. Gallagher, N. B., Peterson, T. W., and Wendt, J. O. L., in Twenty-Sixth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1996, pp. 3197–3204. 7. Kramlich, J. C. and Newton, G. H., Fuel 73:455–462 (1994). 8. Baxter, L. L., Biomass Bioenergy 4:85–102 (1993). 9. Borgwardt, R. H., Roache, N. F., and Bruce, K. R., Ind. Eng. Chem. Fundam. 25:165–169 (1986). 10. Silcox, G. D., Kramlich, J. C., and Pershing, D. W., Ind. Eng. Chem. Res. 28:155–160 (1989). 11. Helble, J. J. and Sarofim, A. F., Combust. Flame 76:183–196 (1989). 12. Kramlich, J. C. and Newton, G. H., Fuel Proc. Technol. 37:143–161 (1994).
COMMENTS S. Yousif, Imperial College, London, UK. In your conclusion, you showed that small or fine particles are the result of vaporization followed by condensation and coagulation. My question concerns the vaporization rate of the mineral matter as a function of temperature, because the fraction of fine particles is a direct function of the percentage of these minerals vaporizing in the hot region of flame. In your work, what is the percentage of mineral matter vaporized? Author’s Reply. For the present fuels and combustion conditions, almost all the sodium and potassium have vaporized; at the same time, only sodium and potassium undergo vaporization. The evidence for this is the low concentration of Na and K in the larger size cuts and the lack of anything but Na and K in the aerosol catch. The amount
of mineral matter vaporized (as shown by Fig. 1 to be ;30– 40%), is generally comparable with the fraction of Na and K in the original mineral matter from Table 1. We have no direct measure of the vaporization rate of Na and K, although evidence from artificial char studies indicates that alkali release parallels char burnout [1]. Thus, vaporization appears to be a function of temperature only insofar as char burnout is a function of temperature. This, however, may not hold for the present biomass fuels.
REFERENCE 1. Srinivasachar, S., Helble, J. J., Ham, D. O., and Domazetis, G., Prog. Energy Combust. Sci. 16:303–309 (1990).
ASH CHARACTERISTICS OF HIGH ALKALI SAWDUST AND SANDERDUST BIOMASS FUELS BLAKE C. CHENEVERT,1 JOHN C. KRAMLICH1 and KENNETH M. NICHOLS2 1Department of Mechanical Engineering University of Washington Seattle, WA 98195-2600, USA 2Weyerhaeuser Company Tacoma, WA 98477, USA
Suspension firing of sawdust and sanderdust fuels is often used in the wood product’s industry to raise steam and provide a heat source for drying and curing operations. The unusually high alkali content of these fuels can give rise to a number of problems that affect the operation of downstream plant systems. The research reported here focuses on the physical and chemical properties of the ash generated by these unique and important biomass fuels and uses this information to identify the mechanisms that control mineral-to-ash transformations. Four sanderdust and sawdust fuels, typical of those used to fire industrial-scale suspension burners, are fired in a laboratory-scale tunnel furnace. Size distribution, morphology, and size versus composition are obtained for particles between 0.0075 and 10 lm by combining a cascade impactor and an electrical aerosol analyzer (EAA). Each of the fuels showed a dominant mode of calcium-rich skeletal particles of size $8.3 lm that are the residue from char burnout. A second, minor mode that seems to be the result of fragmentation appears at 1–3 lm. This consists of fluxed particles that, while still predominantly calcium, also contain Fe, Al, Mn, and Si. Another minor mode at 0.4 lm also appears to be the result of fragmentation. Much of the alkali mineral matter becomes submicron aerosol via the vaporization, condensation, coagulation mechanism. This large yield of aerosol (of the order of 30% of the total ash mass) appears as chlorides in high chlorine fuels and as sulfates and carbonates otherwise. In general, only a small fraction of the alkali metals are captured by the residual ash, and no metals other than Na and K are generally detected with the aerosol. The aerosol size varies between 0.01 and 0.1 lm depending on experimental conditions.
Introduction Sawdust and sanderdust are by-products from the manufacture of particle board and strand board in the wood product’s industry. They are important sources of fuel for drying and curing operations, as well as for raising process steam. Some difficulties arise, however, due to the unique mineral composition of these fuels. In particular, alkali species can constitute more than 50% of the mineral matter. This has led to unforeseen problems as new, stateof-the-art pollution-control equipment has been installed into existing plants. The problems are most evident in the degradation of downstream volatile organic oxidizers that are used to eliminate vapors generated by drying and curing. These problems appear to involve a complex interaction between the ash, the organic vapors, and the particulate control systems that precede the oxidizers. Understanding the nature of the ash generated by these fuels is clearly important to addressing these challenges. Although the biomass fuels considered here differ from coal in many respects, they share many of the same ash-transformation mechanisms. This includes
the generation of residual ash via char burnout and fragmentation and the potential for sintering or melting of the ash, depending on composition [1,2]. The high alkali mineral content can be expected to result in considerable submicron aerosol formation due to the vaporization, condensation, coagulation mechanism [3]. However, irreversible capture of alkali vapor by other minerals, particularly aluminosilicates, can reduce the aerosol yield. The efficiency of this capture depends on the composition of the nonalkali minerals, on the original form of the alkali, and on the physical proximity of the capturing minerals relative to the source of the alkali [4–7]. The alkali minerals can also promote sintering of the residual ash. Although the behavior of the minerals in biomass fuels has been considered, particularly from the standpoint of ash deposition [8], the wide variability in biomass mineral compositions and the practical importance of sawdust and sanderdust fuels requires that their specific behavior be understood. The overall goal of the work reported here is (1) the characterization of the ash generated by sawdust and sanderdust combustion and (2) the identification of the
1719
1720
OTHER POLLUTANTS
Fig. 1. Tunnel furnace schematic.
underlying mechanisms that govern the fate and behavior of the minerals in this unique fuel. Experimental Combustion experiments are performed using the down-fired tunnel furnace (15 kW) shown in cross section in Fig. 1. A high-temperature refractory forms the inner walls of the combustion tunnel, with heat loss to the surroundings minimized by three layers of progressively lower thermal conductivity insulating material. The primary flame is provided by a swirl-stabilized natural gas and air burner located at the top of the reactor. A 35.6-cm-long by 10.2cm-diameter combustion zone (prior to the point of pulverized fuel injection) holds the flame. The solid fuel is metered by a screw feeder, entrained in an air stream, and pneumatically introduced into the furnace via a stainless steel, watercooled injector. The fuel enters the furnace 2.5 cm above the venturi constriction shown in the figure, which promotes uniform dispersion of the fuel particles. After the venturi, the tunnel expands to 20.3 cm diameter. The temperature profile through the furnace can be modified by operating pairs of auxiliary natural gas–fired burners that indirectly heat
the refractory. The solid fuels are injected at 1400 K. The temperature profile in the experimental test section decreases at 200 K/s, and the overall residence time is 1.25 s. The total furnace flow rate is 9.25 g/s, and the solid fuel flow rate is 0.077 g/s, except where noted. The main burner operates at 200% theoretical air, not accounting for the solid fuel. This yields approximately 9% O2 (mole basis, dry) in the flue gas without the solid fuels. The present research examines four sanderdust and sawdust fuels. Various wood-products’ manufacturing facilities supplied the samples, which were selected to be representative of typical fuels used in suspension burners at the given facility. The fuels are briefly described as follows: Fuel 1. A milled sanderdust from composite boards made from urea-formaldehyde resin and Southern pine. Fuel 2. A fuel that uses the same resin as fuel 1, but it consists of 50% hard wood (oak, hickory) and the balance soft wood (gum, beech). Fuel 3. A fuel that includes phenol formaldehyde resin, and it consists of 60% Virginia pine, 30% poplar, and 10% miscellaneous hardwoods. Fuel 4. A material composed of particles that are too fine for strand board; these are sieved out before
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
1721
TABLE 1 Fuel composition
Ultimate analysis (% mass, oven dry basis)
Moisture (% mass, as received) Ash at 750 8C (% mass, OD basis) Major mineral components (mg/kg raw fuel)
C N H S O by D
Ca Na K Mg Mn Fe P Si
Total halogens (mg/kg raw fuel)
resin is added. Thus, except for some miscellaneous wastes, this stream is largely unresinated. It consists of 65% aspen, 15% red pine, and 20% mixed hardwoods. Ultimate analyses, mineral elemental analysis via atomic absorption, and total halogens’ analysis were performed and are summarized in Table 1. The mineral matter is predominantly made up of calcium, sodium, and potassium. Note that fuels 1 and 2 have an order of magnitude higher halogen content than fuels 3 and 4. Significant differences in nitrogen, alkali, and halogen content between the fuels are the result of manufacturing additives (i.e., binding resins), parent wood variety, and other processing sources. The as-received fuels are milled in a Thomas–Wiley laboratory mill to pass through a 0.5mm-diameter screen. This was primarily done to allow the fuels to be compared on an identical size basis. The milled fuels have an essentially identical size distribution, being entirely between 0.1 and 0.6 mm. Ash samples are collected using a water-cooled sampling probe. A porous inner wall provides a path for dilution air. The probe is designed to furnish isokinetic sampling and to minimize thermophoretic deposition of particles on the probe wall. The CO2 content of the undiluted and diluted gas streams provides an accurate tracer to verify isokinetic conditions and calculate dilution rates. The ash size distribution is determined by an eight-stage Anderson cascade impactor (stages numbered 0 through 7). This includes a backup ultimate filter as a ninth stage to catch particles below 0.4 lm. In this configuration, the impactor allows measurement of the total yield of aerosol but cannot provide its specific size.
Fuel 1
Fuel 2
Fuel 3
Fuel 4
47.5 5.2 5.5 0.033 40.9 4.2 0.81 1640 1000 674 249 94 198 65 403 1215
47.7 3.4 5.8 0.030 42.0 4.1 1.07 2410 1500 1020 281 91 43 80 263 2130
50.8 0.19 5.6 0.062 42.6 5.2 0.66 1120 480 454 233 112 308 57 241 82
49.9 0.10 6.4 0.038 42.3 4.2 1.29 3930 870 1160 405 76 75 118 200 137
For those tests in which the aerosol is sized, the last two stages of the impactor and the ultimate filter are removed, and a TSI electronic aerosol analyzer (EAA) is substituted. The aerosol analyzer sizes the material that passes the truncated impactor into “bins” that cover the range of 0.0075–1.0 lm. A system is added ahead of the aerosol analyzer to provide variable delay times before analysis to represent the various low-temperature holdup times experienced by the ash in practical systems. The impactor stages and filter are examined by scanning electron microscopy (SEM), and stagewise elemental analyses are obtained by energy-dispersive X-ray probing (EDAX). Results and Discussion Figure 2 presents size-distribution data for fuels 1 and 2. All plots that follow are mass size distributions. The assumptions used to reduce the impactor data are (1) the particles are solid spheres, (2) the particle density is the same as that of CaCO3 (2.7 g/ cc), and (3) the mass on each impactor stage is assigned the diameter that corresponds to a 50% probability of retention on that stage (d50). While the results are not strongly sensitive to assumed particle density, the shape and porosity of the particles can influence capture. The mass on the ultimate filter, which corresponds to the aerosol mode, is nominally plotted at 0.1 lm based on past work. The determination of the true size of this mode is discussed later. Carbon burnout on all stages is essentially complete. Thus, the size distribution shows three distinct modes of ash: (1) a residual mode of $8.3 lm that
1722
OTHER POLLUTANTS
Fig. 2. Impactor size distribution for fuels 1 and 2. Fuel 1 is shown by open symbols and fuel 2 by solid symbols.
Fig. 4. Ash morphology and overall elemental analyses by mass for fuel 1, 1.6-lm size cut.
Fig. 3. Ash morphology and overall elemental analyses by mass for fuel 1, 8.3-lm size cut.
represents about 40% of the total mass, (2) a minor mode at 1–3 lm that is about 20% of the ash, and (3) a mode too fine to be captured by the smallest impactor stage that includes around 30% of the ash. Figure 3 shows an SEM of a selection of material collected on stage 0 for fuel 1 (d50 4 8.3 lm). Much of the material is highly porous and suggestive of the skeletal remains of unfused ash left as a residual by char burnout. The application of EDAX to the material results in the typical composition shown as the inset in the figure, indicating that the stage is dominated by calcium and a small amount of magnesium.
The generation of a melt by CaO under these conditions is, of course, clearly not indicated. This does not, however, suggest the absence of some surface mobility that is capable of modifying the morphology of the CaO ash. Under atmospheres containing CO2, the repeated formation and decomposition of metastable CaCO3 above the nominal CaCO3 decomposition temperature of 1170 K essentially catalyzes surface mobility. At the temperatures used in this study, the effect of this catalyzed sintering is to relax the minimum radius of curvature present in any CaO structure to an asymptotic value of about 0.25 lm [9,10]. This clearly will not lead to a collapse of the skeletal structures present in Fig. 3, but it will influence the minimum size of the components that make up the structure. Figure 4 shows an example of material collected on stage 4 (d50 4 1.6 lm), which is at the center of the 1–3-lm mode shown in Fig. 2. The inset in the figure shows that the composition includes small amounts of Fe, Al, Mn, Si, and Mg. High magnification shows that the field is dominated by three types of morphology: (1) lacy structures, (2) rounded particles that show evidence of extensive fluxing, and (3) small particles of the order of 0.4 lm in size that appear as nodes on the larger rounded particles. Pointwise EDAX shows that the lacy structures are relatively high in calcium and deficient in the other elements indicated in Fig. 4; these are likely fragments of the skeletal residual ash discussed earlier. The composition of the rounded particles and the nodes are essentially identical, being approximately
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
Fig. 5. Ash morphology and overall elemental analyses by mass for fuel 1, 0.4-lm size cut.
Fig. 6. Ash morphology and overall elemental analyses by mass for fuel 1, ultimate filter.
1723
65–70% Ca with the balance being Fe, Al, Mn, Si, and Mg. Figure 5 shows material collected on stage 7, with a d50 of 0.4 lm. Although Fig. 2 shows that the material on this stage does not significantly contribute to the overall mass, the very uniform character of the material collected here suggests its source originates in a single mechanism. Also, these particles are identical in appearance and composition to the nodes associated with the rounded particles in Fig. 4. The material has an almost gravel-like appearance, which under higher magnification appears as a meltlike rounding with some angularity preserved. The elemental analysis shows the appearance of some sodium and potassium on this stage. If, however, the alkali metals are removed and the composition renormalized, the remaining elements have the same composition as the higher-diameter stages. This suggests that this material has a similar origin but that it is contaminated either by alkali aerosol or by direct condensation. The rounded appearance of these particles may be due to fluxing induced by the other elements present. As discussed earlier, however, pure CaO under a CO2 atmosphere will sinter under these temperatures to the point where the radii of curvature is of the order of 0.25 lm, which is similar to the radii of these particles. The appearance of these particles is consistent with partial sintering rather than complete melting. All four fuels showed essentially identical behavior on all stages above the aerosol filter. The nonalkali ash thus appears to include (1) a large predominance of skeletal, high-calcium structures and (2) a lesser amount of separate, smaller particles that are rich in calcium, but contain other elements. These show signs of more extensive fluxing. A similar size mode has been observed in coal ash [11,12]. In coals, this class of particles increases in importance as the coal is more finely ground, but the yield does not change when the coal is cleaned by froth flotation. This indicates the mode is not the result of fine inclusions released by grinding, as these would be removed by cleaning. It was suggested that the early stages of char combustion result in the release of fine particles, but the accumulation of minerals at the char surface prevent further release during later stages of combustion [12]. A similar mechanism may be active here, resulting in the release of fine fragments during the early stages in char combustion. These particles are primarily calcium, but they include elevated concentrations of the trace metals previously mentioned. A high magnification view of the aerosol filter for fuel 1 is given by Fig. 6, which shows borosilicate glass fibers with agglomerates of very fine particles attached. The size of the individual particles making up the agglomerates is #0.1 lm. EDAX analysis of this stage requires the removal of the element signatures arising from the borosilicate glass. This renormalization produced the elemental composition
1724
OTHER POLLUTANTS
Fig. 7. Elemental composition of material collected on the aerosol filter for each fuel.
Fig. 8. Aerosol size distributions for three delay times for fuel 1.
shown for fuel 1 in Fig. 7. The results indicate that the aerosol catch is almost entirely alkali elements. The chlorine signal is essentially that which would be generated by stoichiometric NaCl and KCl. Although it has been shown that high chlorine content enhances alkali release from solid fuels [5], there is some evidence that a fraction of the alkali species remains with the residual ash. Trace amounts of alkali, relative to the dominant Ca species, are evident by SEM in size cuts of 0.4 lm and larger. The low chlorine-to-alkali molar ratios (0.0 to 0.22) in these larger sized particles suggests that the detected alkali is not deposited via NaCl or KCl condensation, because these are the favored vaporphase species. More likely, this alkali was retained without chlorine bonding, possibly by complexing with other minerals. Fuel 2 (also containing a high chlorine content) has a behavior essentially identical to that of fuel 1. Fuels 3 and 4 are deficient in chlorine. While other features of the ash from these fuels are similar to fuels 1 and 2 (i.e., size distribution and nonaerosol composition), the morphology of the aerosol filter catch and its composition are significantly different.
Specifically, the deposits appear to be less feathery and more like a collection of fine droplets. Figure 7 shows that the absence of chlorine results in a higher concentration of sulfur appearing on the filter. Equilibrium analysis shows that alkali hydroxides and chlorides are favored to appear in the gas phase. The first species predicted to condense is Na2SO4. This condensation is, however, accompanied by a reaction, which suggests that a chemical kinetic barrier exists for the rapid formation of this compound. Alternately, NaCl(s) and KCl(s) are formed directly from identical gas-phase species and thus face no chemical kinetic barrier. The lack of sulfur in the aerosol solids for fuels 1 and 2 indicates that sulfate is not formed during the condensation event and that sulfur is not effective at displacing chlorine as an anion following condensation. For the chlorinedeficient fuels (fuels 3 and 4), the initial condensation is expected to be as a hydroxide, with subsequent pickup of CO2 or sulfur following condensation. The filter analysis shows that in the absence of chlorine, the aerosols do capture sulfur. Equilibrium suggests that the alkali condensation temperature is around 850 K. Direct measurements of the aerosol size distribution were obtained for each of the fuels. One motivation for these measurements is that, in practical facilities, a number of parameters that influence aerosol coagulation are also process variables. Most importantly, a large amount of dilution air is added at points that vary from immediately downstream of the flame to points much further downstream where condensation and coagulation have been essentially completed. Thus, the time to dilution, and the amount of dilution air used, can influence the size of the aerosol that approaches the back-end equipment. Figure 8 shows EAA size distributions for fuel 1 as a function of the delay time between the sample point and the analyzer (here, the fuel flow is 0.0283 g/s). For the conditions used here, coagulation is clearly incomplete in both the “no delay” and 21-s delay tests. For the 198-s delay, the particles are around 0.1 lm in size. This is the practical upper limit for Brownian motion-induced coagulation of aerosol particles. Other data (not plotted) show the expected behavior with solid’s firing rate, that is, that lower firing densities lead to a smaller mean size for the aerosol over the same delay time. All of these trends are expected from coagulation theory. In the practical system, delay times are much shorter than those used in the present experiment, while firing densities are higher. Thus, the two effects tend to counteract. Nucleation, condensation, and coagulation modeling will be used to evaluate both the present data and the influence of the variety of environments presented by practical systems on the final aerosol size distribution. This is, however, beyond the scope of the present work.
ASH CHARACTERISTICS OF HIGH ALKALI BIOMASS
1725
Conclusions
REFERENCES
For each of the fuels examined, calcium is the dominant element in all of the nonaerosol fractions. It appears in conjunction with a small amount of Mg in skeletal residues larger than 10 lm. Other elements (Fe, Al, Mn, and Si) are found with calcium in a 1–3-lm mode that appears to result from char or mineral fragmentation. The presence of these minerals is marked by evidence of fluxing in the ash. The fact that these other elements are concentrated into this minor 1–3-lm mode suggests that these elements may be initially physically separated from the bulk of the calcium in the original mineral matter. Otherwise, one would expect them to be more uniformly distributed throughout the calcium in the ash. The data show some evidence of capture of Na or K by the residual ash, a process that does occur in coal combustion. The small amount of Na and K that appears with the 0.4-lm cut appears to be due to condensation or coagulation of the alkali onto the existing particles or stray capture of aerosol by the last impactor stage before the ultimate filter. The alkali metals thus appear predominantly as a submicron aerosol that is due to vaporization, condensation, and coagulation. When sufficient chlorine is present, the aerosols are present as NaCl and KCl. Although equilibrium favors sulfate formation, sulfur is unable to displace chlorine. In the absence of chlorine, sulfates appear in the aerosol. The aerosol size varies (between 0.01 and 0.1 lm) depending on fuel density (decreases size) and delay time before analysis (increases size) in accordance with coagulation theory. The wide variation in dilution rates and location of dilution found in practical systems makes aerosol size a potentially controllable variable.
1. Miccio, F. and Salatino, P., in Twenty-Fourth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1992, pp. 1145–1151. 2. Mitchell, R. E. and Akanetuk, A. E. J., in Twenty-Sixth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1996, pp. 3137–3144. 3. Flagan, R. C. and Friedlander, S. K., “Particle Formation in Pulverized Coal Combustion—A Review,” in Recent Developments in Aerosol Science (D. T. Shaw, ed.) Wiley, New York, 1978. 4. Neville, M. and Sarofim, A. F., Fuel 64:384–390 (1985). 5. Linder, E. R. and Wall, T. F., in Twenty-Third Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1990, pp. 1313–1321. 6. Gallagher, N. B., Peterson, T. W., and Wendt, J. O. L., in Twenty-Sixth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1996, pp. 3197–3204. 7. Kramlich, J. C. and Newton, G. H., Fuel 73:455–462 (1994). 8. Baxter, L. L., Biomass Bioenergy 4:85–102 (1993). 9. Borgwardt, R. H., Roache, N. F., and Bruce, K. R., Ind. Eng. Chem. Fundam. 25:165–169 (1986). 10. Silcox, G. D., Kramlich, J. C., and Pershing, D. W., Ind. Eng. Chem. Res. 28:155–160 (1989). 11. Helble, J. J. and Sarofim, A. F., Combust. Flame 76:183–196 (1989). 12. Kramlich, J. C. and Newton, G. H., Fuel Proc. Technol. 37:143–161 (1994).
COMMENTS S. Yousif, Imperial College, London, UK. In your conclusion, you showed that small or fine particles are the result of vaporization followed by condensation and coagulation. My question concerns the vaporization rate of the mineral matter as a function of temperature, because the fraction of fine particles is a direct function of the percentage of these minerals vaporizing in the hot region of flame. In your work, what is the percentage of mineral matter vaporized? Author’s Reply. For the present fuels and combustion conditions, almost all the sodium and potassium have vaporized; at the same time, only sodium and potassium undergo vaporization. The evidence for this is the low concentration of Na and K in the larger size cuts and the lack of anything but Na and K in the aerosol catch. The amount
of mineral matter vaporized (as shown by Fig. 1 to be ;30– 40%), is generally comparable with the fraction of Na and K in the original mineral matter from Table 1. We have no direct measure of the vaporization rate of Na and K, although evidence from artificial char studies indicates that alkali release parallels char burnout [1]. Thus, vaporization appears to be a function of temperature only insofar as char burnout is a function of temperature. This, however, may not hold for the present biomass fuels.
REFERENCE 1. Srinivasachar, S., Helble, J. J., Ham, D. O., and Domazetis, G., Prog. Energy Combust. Sci. 16:303–309 (1990).