Research On-Line Analysis of Gas-Phase Composition in the Combustion Chamber and Particle Emission Characteristics during Combustion of Wood and Waste in a Small Batch Reactor T . F E R G E , †,‡ J . M A G U H N , † K . H A F N E R , † F. MU ¨ H L B E R G E R , †,‡ M . D A V I D O V I C , § R. WARNECKE,| AND R . Z I M M E R M A N N * ,†,‡,⊥ Forschungszentrum fu ¨ r Umwelt und Gesundheit, Institut fu ¨r O ¨ kologische Chemie, Ingolsta¨dter Landstrasse 1, D-85764 Neuherberg, Germany, Analytische Chemie, Lehrstuhl fu ¨ r Festko¨rperchemie, Institut fu ¨ r Physik, Universita¨t Augsburg, Universita¨tsstrasse 1, D-86159 Augsburg, Germany, Clausthaler Umwelttechnik-Institut GmbH, Leibnizstrasse 21+23, D-38678 Clausthal-Zellerfeld, Germany, Gemeinschaftskraftwerk Schweinfurt, Hafenstrasse 30, 97424 Schweinfurt, Germany, and Abteilung Umwelt-und Prozesschemie, Bayerisches Institut fu ¨ r Angewandte Umweltforschung und-technik, Am Mittleren Moos 46, D-86167 Augsburg, Germany
The emission of particulate matter and gaseous compounds during combustion of wood and refuse-derived fuel in a small batch reactor is investigated by laser massspectrometric on-line measurement techniques for gasphase analysis and simultaneous registration of physical aerosol properties (number size distribution). The gas-phase composition is addressed by a laser-based mass spectrometric method, namely, vacuum-UV single-photon ionization time-of-flight mass spectrometry (VUV-SPI-TOFMS). Particle-size distributions are measured with a scanning mobility particle sizer. Furthermore, a photoelectric aerosol sensor is applied for detection of particle-bound polycyclic aromatic hydrocarbons. The different phases of wood combustion are distinguishable by both the chemical profiles of gas-phase components (e.g., polycyclic aromatic hydrocarbons, PAH) and the particle-size distribution. Furthermore, short disturbances of the combustion process due to air supply shortages are investigated regarding their effect on particle-size distribution and gas-phase composition, respectively. It is shown that the combustion conditions strongly influence the particle-size distribution as well as on the emission of particle-bound polycyclic aromatic hydrocarbons. * Corresponding author phone: +49 89 3187 4544; fax: +49 89 3187 3510; e-mail:
[email protected]. † Institut fu ¨r O ¨ kologische Chemie. ‡ Universita ¨ t Augsburg. § Clausthaler Umwelttechnik-Institut. | Gemeinschaftskraftwerk Schweinfurt. ⊥ Bayerisches Institut fu ¨ r Angewandte Umweltforschung undtechnik. 10.1021/es049493o CCC: $30.25 Published on Web 01/22/2005
2005 American Chemical Society
Introduction The investigation of incineration processes is an important task in environmental and analytical chemistry, because the combustion of biomass and fossil fuels represents the largest source of anthropogenic air pollution. Emissions produced by combustion processes are composed of gaseous pollutants as well as particulate matter. Therefore, the gas phase and emitted particles should be considered in order to obtain a comprehensive characterization of the emission profile of combustion processes. Several epidemiological studies in the past decades establish a correlation between the concentration of particulate matter in the ambient air and adverse effects on human health (1, 2). In this context, the fine (d < 2.5 µm) and ultrafine (d < 0.1 µm) particles are of special concern due to their ability to penetrate deep into the lung upon inhalation (3). In industrialized areas, anthropogenic combustion processes are the major source of emissions of fine and ultrafine particles into the atmosphere. In particular, industrial incineration plants and traffic emissions (4-6) are the major sources of the anthropogenic particle load of the atmosphere. The new insights into particle-mediated health effects motivate the investigation of particle emissions from anthropogenic sources, especially focusing on the fine and ultrafine particle-size range. In this context physical characterization (i.e., particle number concentrations and size distributions) as well as chemical characterization is necessary for a better understanding of formation mechanisms as well as for an estimate of possible adverse effects on human health. For example, soot particles from combustion processes are known to be inhalable carriers of highly carcinogenic compounds such as polycyclic aromatic hydrocarbons (PAH) (7, 8). To reduce aerosol emissions by means of process-integrated primary measures, it is necessary to learn how and why particle properties are correlated with plant operating conditions such as air supply and boiler load. In general, there are two main sources for generation of particles from combustion processes: Particles can be formed from (1) inorganic material in the fuel ash and (2) incomplete combustion of the organic fraction of the feeding material (9, 10). In the first case, the particles contain mainly inorganic salts (mixture of alkali compounds, sulfates, nitrates, and silicates), whereas the particles of the latter type are composed predominantly of soot and condensable unburnt organic material. These two particle types, however, can occur as complex internal and/or external mixtures. The emission of particulate matter is correlated to the operating conditions of the combustion process. This is particularly true for particles of the latter type, which are suspected to exhibit a strong impact on human health (11). In several studies the effects of process control measures on the emission of aerosol particles into the flue gas of incineration processes were investigated. In particular, this applies to the case of combustion of biomass, which nowadays is more and more discussed as a CO2-neutral fuel. Particles emitted from wood combustion are found to be primarily in the submicrometer size range (12). For instance, the number concentration of particles in the raw gas of largescale biomass combustion has been reported to be (1-2) × 107 particles/cm3 and the maximum of the number-size distribution to be 150-300 nm (13). Other studies revealed similar results with maxima of the size distribution around 100 nm, exhibiting a shift to larger particle sizes during decreased oxygen concentration and thus more incomplete VOL. 39, NO. 6, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1393
combustion (14-16). Generally, large-scale boilers can be operated under more favorable, that is, less nonstationary combustion conditions. Due to larger combustion air and feeding material volumes, the influence of local fluctuations in combustion performance is attenuated. Hence, particles originating from large-scale combustion facilities usually contain a relatively small organic fraction (13, 17-19). The submicrometer particles emitted are mainly composed of potassium chloride and potassium sulfate, while supermicrometer particles contain species such as calcium, magnesium, silicon, or aluminum. This chemical composition shows that these combustion particles originate from the fuel ash (10, 17, 18, 20, 21). In small batch-operated boilers, the particle-number distribution shows a larger variation, ranging from 107 to 1010 particles/cm3 (22-24). The level of excess air seems to affect the particle-number concentration. In automatically and permanently operated wood chip burners, the particle-number concentration decreases when the oxygen concentration in the flue gas is reduced (22, 23). Furthermore, the number distribution changes according to the different combustion phases. For example, during the start-up phase of the combustion the particle-size distribution is reported to exhibit a maximum at the largest particle size around 200 nm, during the intermediate phase at lower particle sizes in the range of 150 nm, and during burn-out at the lowest particle size with a maximum around 50 nm (24, 25). As in large-scale combustion, the main elemental composition of the particles is due to potassium, chlorine, sulfur, and oxygen. The concentration of organic carbon can be as low as 1-10% in the fly ash under conditions of complete combustion (26). Investigations on the direct effects of changes in process parameters on the emission of particles are important to increase knowledge of the formation mechanisms and particle emission levels. Lillieblad et al. (15) report in a recent publication on the influence of boiler operation on the emission of submicrometer particles and several gaseous compounds (e.g., PAH) from biomass combustion. They find a dependency of the particle-number concentration on boiler load, resulting in higher particle concentrations with increasing load. Potassium and sulfur dominated the submicrometer particle composition for all loads, making up 6987% of the analyzed mass. The concentration of elemental carbon was higher for low load compared to medium load. This is in agreement with an increase in CO and total hydrocarbons at low load, indicating conditions of incomplete combustion. The concentration of carbon-containing particles can be influenced by process control measures (27). For example, by firing the back-up oil burners, a very pronounced additional maximum in the size distribution around 30 nm shows up. Such newly formed particles predominantly are due to unburnt carbon (soot) (28, 29). A correlation of combustion conditions with emission of gas-phase and particulate constituents needs the application of on-line techniques for a time-resolved analysis of gas and particle phases. In recent papers it was demonstrated that laser-based mass spectrometric methods are feasible for the on-line detection of gas phase components in the flue gas of industrial incineration plants, especially for PAH (30-40). On the other hand, the photoelectric aerosol sensor (PAS) can be used for the on-line detection of particle-bound PAH (PPAH) as a sum value, giving information about the chemical composition of combustion particles (24, 41-43). In this paper, we report on the correlation of gas phase and particle emissions during combustion of wood and mixtures of wood with refuse-derived fuel in a small batchoperated boiler. The gas-phase composition was measured by vacuum-UV single-photon ionization time-of-flight mass spectrometry (VUV-SPI-TOFMS) (38, 44). Particle emissions 1394
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 6, 2005
FIGURE 1. Schematic drawing of the combustion reactor: (A) fuel bed; (B) sampling point for VUV-SPI-TOFMS measurements; (C) sampling point for particle measurements; (D) combustion chamber. were investigated regarding the size distribution with a scanning mobility particle sizer (SMPS). Furthermore, the PAS signal was monitored. Several measurements were performed during stable combustion conditions as well disturbed combustion conditions, namely, periods of incomplete combustion due to air-supply shortages (CO peaks).
Experimental Methods Fuel Characteristics. Wood combustion was performed with wood chips from waste wood that had a grain size of approximately 3 cm. The fuel had an average caloric value of 16 MJ/kg and a humidity of 28% (45). The experiments dealing with waste combustion were performed with a mixture of wood chips and refuse-derived fuel (RDF). RDF is made of domestic waste, which is first sorted to remove incombustibles, dried, and eventually mixed and pelletized. The pellets of the RDF used in this study had an approximate size of 2 × 4 cm. The RDF had an average caloric value of 22 MJ/kg and a humidity of 7%. The ash content was about 18% (45). For each experiment 4 kg of wood chips out of a much larger batch of waste wood was used. The repeatability of the experiments was therefore good from one experiment to the next. Combustion of RDF was performed by thorough mixing of 2 kg of wood chips with 2 kg of RDF pellets to result again in a fuel batch of 4 kg. Combustion Process. The measurements were carried out in a small batch-operated combustion reactor with a thermal capacity of 50-150 kW (Figure 1). In the described experiments the reactor was fired with wood chips. The advantage of the small reactor is that air supply and boiler load can easily be adjusted. Therefore several operation conditions as well as disturbed combustion processes can be simulated and monitored. Figure 1 shows a schematic drawing of the reactor. The insides of the combustion chamber, which are covered with firebricks, are preheated to approximately 1000 °C by a methane gas burner. Sub-
FIGURE 2. Schematic drawing of the sampling system for VUVSPI-TOFMS measurements and the particle sampling and analysis system. sequently, the burner is removed and the grate with the combustion fuel is installed below the combustion chamber. Then the fuel is heated and dried by the heat radiation from the chamber walls. Shortly after, flaming combustion starts. The primary air supply is located below the grate, the air being blown through the fuel bed from below. Additionally, secondary air is blown into the combustion chamber through nozzles, which are located 1.0 m above the burning bed. All measurements were started when the fuel was inserted into the combustion chamber (t ) 0 s) and were not stopped until all material was burnt and no smoldering could be detected by visual inspection through a small window located at the top of the combustion chamber. Combustion experiments performed in the batch reactor can exemplify the temporal behavior of a distinct grate zone in industrial incineration plants, where feeding material is moved continuously down the grate. In this context, one complete combustion cycle in the batch reactor represents a complete passage over the grate zone. VUV-SPI Measurements. The sampling point for the SPI measurements was located below the nozzles of the secondary air supply about 30 cm above the flames and therefore in the primary combustion zone. Figure 2 shows a scheme of the sampling system. A quartz-glass tube (i.d. 5 mm) shielded by a stainless steel tube was inserted into the combustion chamber and a sample flow of 5 L/min was applied by a small sampling pump. The sampled gas passed a heated quartz-fiber filter to prevent dust from entering the further sampling line. Behind the filter, the tip of a quartz capillary (i.d. 0.3 mm) was located in the center of the tube and a small amount of the sample flow (approximately 15 mL/min) was split, flowing to the mass spectrometer for analysis. The whole sampling line was heated to 250 °C to prevent condensation of tarry substances and thus clogging of the capillary. A detailed description of the inlet system is available in the literature (35, 38).
The sampled flue gas was subsequently analyzed by vacuum-UV single-photon ionization (VUV-SPI) mass spectrometry (38, 44, 46-49). In the measurements described here, VUV photons with a wavelength of 118 nm were used for a single-photon absorption/ionization process. Ions were subsequently analyzed in a time-of-flight mass spectrometer (TOFMS). The 118 nm radiation is generated via a rare-gas cell for frequency tripling of the 355 nm radiation of a Nd: YAG laser (50-53). As only compounds with an ionization potential (IP) below the photon energy can be ionized, selectivity of the ionization process is obtained by virtue of the ionization potential. The photon energy for SPI is 10.5 eV (118 nm). VUV-SPI is a very soft (i.e., nearly free of fragmentation) ionization method. In most cases only the molecular ion peaks of detected compounds are present in the mass spectra. Therefore, the method is suited to analysis of gas mixtures. A more detailed description of the technique is given in references 38, 40, and 44. All gas-phase measurements were performed with a mobile TOFMS, which is described in detail in a previous publication (38). The instrument is equipped with a Nd:YAG laser, which provides 355 nm radiation with a pulse energy of 7 mJ. However, the results showed that the ionization efficiency is highly dependent on the energy of the 355 nm laser radiation (∼I3). Therefore, a more powerful laser (Surelite II, Continuum, CA) was coupled to the instrument and used for pumping of the rare-gas cell. This laser supplies 355 nm pulses with a pulse energy of 20 mJ into the rare-gas cell. With this setup, limits of detection of 85 parts per billion (ppb) for benzene, 120 ppb for toluene, and 134 ppb for xylene were obtained (S/N ) 2). Limits of detection were determined according to literature (31). In a recent instrument the limits of detection achieved here could be lowered due to further improvement of the laser setup and are as low as 1.5 ppb for benzene and 2 ppb for toluene (54). Particle Measurements. The particle measurements were performed at the beginning of the flue gas duct above the combustion chamber (Figure 2) at a flue-gas temperature of about 250 °C. Particle sampling was performed isokinetically. The outer part of the sampling line was heated to temperatures above 120 °C to avoid condensation of water and other condensable gases. The aerosol passed a heated preimpactor for precipitation of particles with diameters larger than 40 µm. The aerosol then was immediately diluted with particlefree air prior to analysis by a factor of 100 by use of two ejector diluters in line (model VKL-10E, Palas GmbH, Karlsruhe, Germany). The first diluter was also heated to 120 °C to avoid condensation effects. A detailed description of the sampling system can be found in the literature (27). The diluted aerosol was subsequently analyzed regarding its size distribution with a scanning mobility particle sizer (SMPS, model 3936, consisting of a DMA model 3071 and a CPC model 3025A, TSI Inc., St. Paul, MN). Additionally, information about the chemical nature of the particles could be obtained with a photoelectric aerosol sensor (PAS 2000, EcoChem, U ¨ berlingen, Germany). The PAS is working on the principle of photoionization of particle-bound PAH. Carbonaceous aerosol particles carrying adsorbed PAH molecules emit electrons when irradiated with UV light and thus are positively charged. Subsequently, the particles are collected on a filter inside an electrometer, which detects the charge deposited on the filter. The analyzer signal therefore is a rough measure of total PAH adsorbed on carbon particles (42, 43). Note that some interferences due to metal oxides and soot particles occur and therefore the signal cannot be considered as quantitative measure of PAH-rich particles in combustion processes. The SMPS was operated in the size range from 28 to 430 nm; scanning times were adjusted to a scan-up time of 30 s and a retrace time of 10 s, so that a complete size distribution VOL. 39, NO. 6, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1395
FIGURE 3. Typical VUV-SPI time-of-flight mass spectrum obtained during combustion of wood chips (a) and wood/RDF (b); lower spectra in each panel are 10-fold amplified. Major peaks are due to benzene (78 m/z), naphthalene (128 m/z), and phenanthrene (178 m/z). Complete assignment of peaks is given in Table 1.
TABLE 1. Peak Assignment for VUV-SPI Spectra Measured during Combustion Experiments with Wood and Wood/RDF as Feeding Material m/z [amu]
total molecular formula
assigned compound
17 28 30 40 52 67 78 93 103 117 128 153 167 178
NH3 C2H4 NO C3H4 C4H4 C4H5N C6H6 C6H5NH2 C6H5CN C8H7N C10H8 C10H7CN C12H9N C14H10
ammonia ethylene nitrogen oxide propyne vinylacetylene pyrrole benzene aniline benzonitrile indole naphthalene cyanonaphthalene carbazole phenanthrene
could be recorded every 40 s. One PAS measurement was performed every 10 s. For correlation with the SMPS data these values were averaged in the appropriate time intervals to yield comparable 40 s values.
Results and Discussion Wood Combustion. The first experiments were performed during burning of wood chips under stable process conditions. Figure 3a shows a typical sum spectrum obtained by VUV-SPI-TOFMS at a wavelength of 118 nm. Aromatic compounds such as benzene (78 m/z), naphthalene (128 m/z), and phenanthrene (178 m/z) are clearly visible in the spectrum. A tentative assignment of the observed peaks is given in Table 1. The VUV-SPI technique allows the simultaneous on-line real-time detection of several target compounds with high time resolution. In principle, 10 spectra/s can be recorded; here a time resolution of 2 Hz was achieved as five spectra were internally averaged on the transient recorder. The signal intensity is directly correlated to the absolute concentration of the target compound in the sample gas; therefore the timeintensity profile is a direct measure of the time-dependent concentration changes of several target compounds (35, 55, 56). However, profiles of different compounds cannot be 1396
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 6, 2005
FIGURE 4. (a) Time-intensity profiles of benzene and phenanthrene recorded with VUV-SPI-TOFMS during a measurement with stable combustion conditions. The different phases of wood combustion are indicated by the vertical lines. (b) Time profiles of the number concentration of selected particle sizes (30, 60, 100, and 200 nm). compared directly in terms of absolute concentration without previous calibration due to substance-specific differences in the ionization cross sections. Figure 4a shows the timeintensity profiles of benzene and phenanthrene as exemplary compounds present in the gas phase, which were detected by VUV-SPI during a typical wood combustion experiment. In general, the course of wood combustion can be summarized as follows (57, 58): When wood is heated, its constituents start to hydrolyze, oxidize, dehydrate, and pyrolyze (drying/outgassing phase). With further increasing temperature, combustible volatiles as well as tarry substances and reactive carbonaceous char are formed. When the ignition temperature of the volatile and tarry substances is reached, the exothermic oxidative reactions start (flaming combustion). During flaming combustion, char formation continues until the flux of combustible volatiles drops below
the level needed to maintain flaming combustion. At this time, the afterglow or smoldering combustion phase starts, best described as solid-phase combustion of the residual char (59). The different combustion phases are marked in Figure 4a. During the start-up phase (drying and outgassing), no SPI-TOFMS signal can be detected except for a short peak of the benzene concentration right at the beginning of the process. Flaming combustion initiates at t ) 600 s. During this phase, the signal of benzene shows a quite high intensity. Obviously it is released in relatively high concentrations whereas higher mass aromatic hydrocarbons (like phenanthrene) show only weak signals. The relatively constant level of benzene (and naphthalene, not shown here) is typical for the combustion experiments with stable conditions performed in this study. Because the sampling point was located below the secondary air nozzles, the data describe the primary combustion zone. The transition from flaming combustion to the smoldering phase takes place at t ) 1150 s (also observed by visual inspection). Immediately, the pattern of aromatic compounds changes completely. The profiles of the small compounds decrease considerably, whereas the large PAH increase in concentration (high signal intensity). The emission profile gets dominated by the aromatic hydrocarbons of higher mass. Pyrolysis of the remaining fuel is the most prominent decomposition reaction during the smoldering combustion phase. The primary products of the pyrolytic decomposition of wood are cracked. Under the still high temperature conditions, the growth of polynuclear aromatics is due to low molecular weight hydrocarbons, which are formed directly from the gas-phase cracking of remaining wood compounds (60). The shift observed in the hydrocarbon pattern, which is indicating the transition from flaming to smoldering (that was also observed by visual inspection), was consistently found in all combustion experiments performed in this study. Figure 4b shows the corresponding SMPS time profiles of particle-number concentrations of particles of a selected size (30, 60, 100, and 200 nm). No significant change in number concentrations can be observed at the transition from the start-up phase to flaming combustion. Particles with a diameter of 30 nm exhibit high number concentrations of (2-5) × 109 particles/cm3. Not until the transition from flaming to smoldering combustion do particles of larger diameters start to increase in number concentration. The 30-nm particle concentration starts to decrease immediately after the transition; larger particles (60 and 100 nm) peak at the beginning of the smoldering phase and decrease again in concentration subsequently. Figure 5 shows a comprehensive view of the particle-size distribution during this combustion experiment measured with the SMPS. In the contour plot and the three-dimensional size distribution plot, the boundaries of the different combustion phases are plotted as white lines. During the outgassing phase, the particle-size distribution exhibits a maximum around 30 nm. With the onset of the flaming combustion, the particle-size distribution still exhibits a clear maximum at 30 nm with a higher number concentration of 7 × 109 particles/cm3. This corresponds to the concentration range stated in the literature (10). Our finding of larger particle diameters during the smoldering phase of combustion appears to contradict literature data, since we observe an increase in particle size as well as particle-number concentration. As is shown in Figure 5, the maximum of the particlesize distribution during the smoldering phase is around 60 nm with a concentration of 1.1 × 1010 particles/cm3. This effect of increasing particle size with transition to smoldering combustion was consistent throughout all measurements performed. The chemical nature of the particles was investigated with the aid of the PAS with respect to particle-bound PAH (PPAH), which can be used as indicators of incomplete
FIGURE 5. (a) Particle number size distribution for a wood combustion experiment during stable combustion conditions. (b) Contour plot of the particle-size distribution during wood combustion with stable combustion conditions. (c) Three-dimensional representation of the particle-size distribution during wood combustion with stable combustion conditions. combustion. However, in experiments with stable combustion conditions, the PAS showed no relevant photoemission (PE) signal. This indicates that the particles emitted from the combustion chamber are predominantly of inorganic nature (mainly potassium salts), as was typically observed for biomass combustion (10, 19, 22, 23). The fuel-inherent inorganic material is released as vapor and subsequently nucleates in the upper region of the burning chamber forming particles in the ultrafine particle-size range. During the smoldering phase, the temperature at the top of the combustion chamber drops from 750 to 600 °C. The temperature in the subsequent flue gas duct is consequently reduced even further, causing semivolatile gas-phase constituents to condense on particles, thus increasing their size. VOL. 39, NO. 6, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1397
FIGURE 6. Temperature and PE-signal profile during a measurement with hot CO peak. The period of the CO peak (410-510 s) is indicated by shading. Note that due to technical reasons, the temperature refers to the top of the combustion chamber and not to the actual particle sampling point in the flue gas channel. Also the formation rate of PAH in the gas phase in the primary combustion zone is increased. This agrees well with studies on pyrolysis of wood and formation of PAH in the gas phase (60). Despite an increase in PAH formation, no significant increase of PAH-containing soot particles could be observed by photoemission. The increase in size and number concentration of inorganic particles containing chloride and sulfate is of particular importance in large industrial scale incineration plants as well as in coal and biomass combustion. Alkali species often serve as carriers of chlorine and sulfates, whose deposition is a persistent cause for intermittent fouling and corrosion problems (61, 62). The growth in particle size enhances the deposition probability inside boiler sections such as, for example, the heat exchanger tubes. These particle deposits are therefore suspected to play a major role in corrosion, which constitutes a severe problem, especially in large-scale waste incineration plants. Influence of Combustion Conditions: Impact of ShortTime Disturbances. One objective of this work is to investigate the influence of combustion conditions on the emission profile in the gas and particle phase. Due to increasing importance of biomass combustion in energy production as well as domestic heating, profound knowledge of the influence of combustion conditions on the emission profile is needed. Likewise, the investigation of combustion processes in reactors with well-defined conditions can be used as model experiments for large-scale industrial incinerators. Conditions of disturbed combustion can be easily simulated by decreasing the air flow through the secondary air nozzles in the upper part of the burning chamber (see Figure 1), causing an insufficient oxygen supply in the secondary burning zone. This is referred to as “hot CO peak” due to the constant high-temperature level in the primary combustion zone. During a wood combustion experiment, a CO peak was induced by reducing the secondary air supply. The resulting temperature profile in the secondary combustion zone of the combustion chamber along with the observed PE signal is shown in Figure 6. The period of the hot CO peak is marked. During this period, the temperature at the top of the combustion chamber drops from 850 to 750 °C. At the same time the PE signal increases to values of up to 16 pA, indicating an emission of large amounts of PAH-rich particles (again, note that there can be interference by metal oxides). Figure 7 panel a shows the gas-phase profiles of benzene and phenanthrene and panel b shows the time profiles of the particle-number concentrations. Again, the combustion 1398
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 6, 2005
FIGURE 7. (a) Time-intensity profiles of benzene and phenanthrene recorded with VUVSPI-TOFMS during a measurement of wood combustion with hot CO peak. (b) Time profiles of the number concentration of selected particle sizes during a measurement of wood combustion with hot CO peak. phases are indicated as well as the period of the CO peak. As was observed for stable combustion conditions, a shift in the emission pattern of gas-phase compounds occurs at the transition from flaming to smoldering combustion. In the same way, the particle concentrations show similar behavior; that is, the 30-nm particles exhibit high number concentrations of 3 × 109 particles/cm3 during flaming combustion and a decrease at the transition to smoldering, whereas the particles of larger diameters show an increasing number concentration at this time. Note that the number concentration for flaming combustion refers to the period before the CO peak, as particle emissions stay affected also after the CO peak has vanished (see below).
FIGURE 8. (a) Particle-size distribution for a wood combustion experiment with hot CO peak. (b) Contour plot of the particle-size distribution during wood combustion with hot CO peak. (c) Threedimensional representation of the particle-size distribution during wood combustion with hot CO peak. The particle-size distributions as well as contour and three-dimensional plots are shown in Figure 8. The CO peak itself occurred between 410 and 510 s. At this time the particlesize distribution shows a considerable increase in number concentration as well as the rise of a second maximum. During stable flaming combustion, before the secondary air supply was decreased, the particle-size distribution exhibits a maximum of 30 nm with a particle-number concentration of 3 × 109 particles/cm3. During the CO peak the maximum of the ultrafine mode is slightly shifted to 35 nm and the number concentration increases to 1.5 × 1011 particles/cm3. Due to the lower temperature, nucleation and condensation reactions take place, leading to higher particle concentrations
by gas to particle conversion. Furthermore, these high number concentrations lead to coagulation of particles, resulting in a second maximum around 130 nm (dN/d log D ) 4 × 1010 particles/cm3). Additionally, the chemical nature (and thus the photoelectric properties) of the particles is influenced as is shown in Figure 6. Obviously, reduced postcombustion of volatiles and particulate matter produces large amounts of photoelectric active particles. It is likely that these particles generated by incomplete combustion are predominantly soot particles containing large amounts of PAH. The time-intensity profiles of gas-phase constituents measured with VUV-SPI are influenced (e.g., in the case of benzene, Figure 7a), although the sampling point of the VUVSPI measurement was located below the secondary air nozzles (Figure 1). This effect can be explained by the reduced heat radiation from the secondary combustion zone during the period of secondary air shortage and demonstrates that changes in the secondary combustion zone conditions affect the chemical reactions in the primary zone (i.e., by thermal coupling of primary and secondary combustion). After a return to the normal operation mode, the PE signal shows a steep decrease, but stays at elevated levels at 3-4 pA. Similarly, the particle-size distribution still exhibits its bimodal shape (admittedly less pronounced) in the early post-CO peak period. Obviously, more photoelectrically active particles are emitted after the disturbance compared to normal flaming wood combustion. Therefore, it has to be assumed that the emission characteristics of combustion processes depend not only on the actual state of the combustion process but also on the “process history” (63). The transition to smoldering combustion occurs at 1050 s. This is obvious from the change in the emission profiles of the gas-phase constituents. At the same time, the particlesize distribution shows the characteristic, pronounced shift to larger particle diameters with a new maximum at 60 nm and a particle-number concentration of 1.0 × 1010 particles/ cm3. The PE signal is still slowly decreasing. This shows that photoelectrically active particles are still emitted. One interesting feature of the transition from flaming to smoldering is seen in Figure 6. At t ) 1040 s the PE signal as well as the temperature shows a short peak. At this moment a second ignition could be observed of some feeding material lasting only for several seconds. Therefore the temperature rises again in the combustion chamber. The PE signal shows that this second flaming takes place under conditions of incomplete combustion, which becomes obvious when one compares the gas-phase signals (Figure 7a) of this time period (t ) 1040 s) with the CO peak. A similar secondary ignition event was described in a study on biomass ignition and explained by inhomogeneities in the sample so that a coexistence of different combustion zones is possible (64). As the ignition depends on the diffusion of oxygen to the fuel surface and reactions with volatiles, it is imaginable that in some region of the grate unburnt material remained and a second ignition could be observed. However, this effect was observed only once, so that a more complete explanation of this behavior still needs further investigation. Combustion of Refuse-Derived Fuel. In Figure 3b a typical mass spectrum of an experiment with RDF/wood mixture as burning material is depicted. Besides the pure hydrocarbon compounds, also nitrogenous compounds can be detected by VUV, namely, benzonitrile (103 m/z), indole (117 m/z), naphthalenecarbonitrile (153 m/z), and carbazole (167 m/z). A complete assignment of observed peaks is given in Table 1. Nitrogenous compounds play a role in the formation of NOx and are a useful indicator for flame chemistry. A detailed discussion on studies of nitrogen-containing compounds in combustion processes will be given in a forthcoming publication (65). VOL. 39, NO. 6, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1399
FIGURE 9. Temperature and PE-signal profile during a measurement with wood/RDF as burning material. The period of the occurring cold CO peak (350-410 s) is indicated by shading. In combustion processes, disturbances arising from the primary combustion zone can be observed accidentally. The occurrence of this so-called cold CO peak (contrary to hot CO peaks, see above) is caused by fuel inhomogeneities (like agglomeration of wet feeding material) leading to locally poor oxygen supply in the primary combustion zone and resulting in decreased temperature in the fuel bed. Due to such inhomogeneities in the feeding material, a cold CO peak could be observed during a measurement. In Figure 9 the temperature profile and the PE signal during a measurement with a cold CO peak is depicted. Again, the period of the CO peak is marked. During the CO peak the temperature at the top of the combustion chamber drops from 890 to 790 °C, resembling the temperature drop during the hot CO peak (i.e., the distinction between the two types of CO peaks refers to the temperature in the fuel bed and not to the temperature in the secondary combustion zone). The PE signal again shows large values of up to 17 pA. Figure 10 panel a depicts the gas-phase profiles of benzene and phenanthrene, whereas in panel b the time profiles of the particle-number concentrations of particles with diameters of 30, 60, 100, and 200 nm are shown. Again, the combustion phases are indicated as well as the period of the CO peak. In accordance with the experiments on wood combustion, a shift in the emission profile of gas-phase compounds occurs at the transition from flaming to smoldering combustion (Figure 10). However, in comparison with wood combustion (Figures 4 and 7), some differences can be observed. The decrease of particles with 30 nm diameter exhibits a steady decrease over the whole sampling period. The increase of larger particles during smoldering combustion is most obvious for particles with 200 nm diameter, whereas particles around 60-100 nm show a short-term decline shortly after the transition from flaming to smoldering and a new rise in concentration thereafter. Obviously, the shift in the particlesize distribution is not as incisive as in the case of combustion of pure wood chips. During the period of disturbed combustion conditions, one difference from the so-called hot CO peak becomes obvious: The emission of gaseous compounds in the primary combustion zone exhibits changed behavior. While the benzene signal is reduced, phenanthrene becomes a prominent compound in the gas phase of the primary combustion zone. In this context, the cold CO peak resembles the smoldering combustion phase. This finding is plausible due to the mechanism responsible for the appearance of a cold CO peak due to locally insufficient oxygen supply in the fuel bed and thus emerging conditions of incomplete combustion. Figure 11 depicts the particle-size distributions along with contour and three-dimensional plots of the temporal varia1400
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 6, 2005
FIGURE 10. (a) Time-intensity profiles of benzene and phenanthrene recorded with VUV-SPI-TOFMS during a measurement with wood/ RDF as burning material with cold CO peak. (b) Time profiles of the number concentration of selected particle sizes during a measurement with wood/RDF as burning material with cold CO peak. tion of the particle-number distribution. Before the CO peak (during normal flaming combustion) the size distribution exhibits a maximum at around 50 nm with a number concentration of 2.5 × 1010 particles/cm3. These larger particle diameters are due to the use of RDF, the combustion of which generally exhibits larger particle diameters compared to wood combustion. Again, a slight shift in the particle-size distribution can be observed at the transition from flaming to smoldering combustion. The maximum of the particle-size distribution during smoldering is around 80 nm; in contrast to the measurements performed with wood as feeding material, the number concentration is decreased to 1.5 × 108 particle/cm3.
Therefore, conditions of incomplete combustion are present in the primary combustion zone, resulting in the observed change in the emission pattern detected with VUV-SPI right above the flame front. In the post-CO phase, the particle-size distribution immediately returns to a unimodal size distribution with a maximum at 64 nm, which is a slightly larger value compared to the maximum found before the disturbance (53 nm). Besides this small shift of the size distribution, there is no strong influence of the disturbance on the particle emission (regarding particle size) after return to normal operation.
FIGURE 11. (a) Particle-size distribution for a wood/RDF combustion experiment with cold CO peak. (b) Contour plot of the particle-size distribution during wood/RDF combustion with cold CO peak. (c) Three-dimensional representation of the particle-size distribution during wood/RDF combustion with cold CO peak. During the disturbance (lasting from 350 to 410 s), a bimodal particle-size distribution is observed similar to the bimodal distribution of the hot CO peak. The maximum in the ultrafine range is slightly shifted to smaller particle sizes from 50 to 30 nm. The second maximum appears at 160 nm. Here again, the decreased temperature leads to nucleation and condensation effects, which cause the bimodal particlesize distribution. Likewise, a high PE signal (up to 18 pA) is observed (Figure 9). However, some differences are visible in the emission profile of the gaseous species (Figure 10a) when compared to the hot CO peak (Figure 7a). In the period of the CO peak, a short-term increase of phenanthrene as well as a considerable decrease in the benzene signal is apparent. The disturbance is caused by inhomogeneities in the fuel bed, leading to decreased oxygen supply in the fuel.
For both measurements of disturbed combustion conditions (wood as well as wood/RDF mixture) a different emission characteristic could be observed for the post-CO period compared to combustion under “normal” combustion conditions. As was pointed out above, it has to be assumed that the emission characteristics of combustion processes depend not only on the actual state of the combustion process but also on the “process history”. The observation of a changed emission characteristic after CO peaks (here the particle-size distribution and photoelectric activity) resembles the emission memory effects, which were observed in an incineration plant after periods of disturbed combustion. After unstable combustion conditions, high emission levels of PAH and PCDD/F (gas phase and particle phase combined) typical for incomplete combustion were observed for several hours (63, 66) even though the actual combustion conditions were again stable. A possible explanation for memory phenomena is the formation of surface deposits on the internal surface of the boiler, most probably an effect of ash deposition. After return to normal combustion conditions, that is, high temperatures in the postcombustion zone, these wall deposits start to pyrolyze and evaporate, causing sustained release of PAH and soot particles. In the early postCO peak phase (in the case of the hot CO peak), the number concentration still shows a bimodal distribution. At t ) 800 s (i.e., approximately 300 s after the end of the hot CO peak), the maximum at 130 nm has vanished and the particle-size distribution again resembles the unimodal distribution as observed before the CO peak (single maximum around 30 nm). However, the particle-number concentration still is increased (8 × 109 instead of 3 × 109 particles/cm-3) at t ) 800 s (Figure 8). Additionally, the PE signal also shows higher levels than before the hot CO peak (Figure 6). In the case of the cold CO peak, an immediate return to the unimodal size distribution but with a larger maximum can be observed (Figure 11). Similarly, the PE signal shows a slow decrease after the return to normal operating conditions (Figure 9). When compared to combustion disturbances in larger-scale waste incineration processes (63), these results might be an indication for similar processes resulting in memory emissions. Upon comparison of the PE signals of both types of CO peaks, a slower decline and therefore an extended memory was observed for the cold CO peak. However, more work is required to address these issues.
Acknowledgments This study was carried out within the scope of the GSF-Focus “Health relevance of aerosols”, which coordinates aerosolrelated research within the GSF Research Center. Funding from the BMBF, Germany, and the GSF Research Center as well as from the BStMUGV and the European Union (Grant UGV08030604128) is gratefully acknowledged. We thank M. Blumenstock, K. Neuer-Etscheid, and T. Hauler for contributions during the measurement campaign as well as T. Adam, S. Mitschke, T. Streibel, W. Welthagen, and S. Gallavardin for support and discussions. T.F. thanks the Deutsche Bundesstiftung Umwelt for a Ph.D. scholarship. VOL. 39, NO. 6, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1401
Literature Cited (1) Dockery, D. W.; Pope, C. A.; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G.; Speizer, F. E. New Engl. J. Med. 1993, 329, 1753-1759. (2) Pope, C. A.; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W. Am. J. Respir. Crit. Care 1995, 151, 669-674. (3) Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. Am. J. Respir. Crit. Care 1997, 155, 1376-1383. (4) Jahresbericht 1999, Umweltbundesamt: 1999. (5) 1998 Emissions. In National Air Pollution Emission Trends 19001998; United States Environmental Protection Agency: Washington, DC, 2000. (6) Avakian, M. D.; Dellinger, B.; Fiedler, H.; Gullet, B. K.; Koshland, C. P.; Marklund, S.; Oberdo¨rster, G.; Safe, S.; Sarofim, A. F.; Smith, K. R. Environ. Health Perspect. 2002, 110, 1155-1162. (7) Denissenko, M. F.; Pao, A.; Tang, M.; Pfeiffer, G. P. Science 1996, 274, 430-432. (8) Schnelle, J.; Gebefu ¨ gi, I.; Welzl, G.; Ja¨nsch, T.; Kettrup, A. Atmos. Environ. 2001, 35, S71-S81. (9) Lighty, J. S.; Veranth, J. M.; Sarofim, A. F. J. Air Waste Manage. 2000, 50, 1565-1618. (10) Johansson, L. S.; Tullin, C.; Leckner, B.; Sjo ¨ vall, P. Biomass Bioenergy 2003, 25, 435-446. (11) Andersson, P.; Marklund, S. Chemosphere 1998, 36, 1429-1443. (12) Gullet, B. K.; Touati, A.; Hays, M. D. Environ. Sci. Technol. 2003, 37, 1758-1765. (13) Christensen, K. A.; Stenholm, M.; Livbjerg, H. J. Aerosol Sci. 1998, 29, 421-444. (14) Strand, M.; Pagels, J.; Szpila, A.; Gudmundsson, A.; Swietlicki, E.; Boghard, M.; Sanati, M. Energy Fuels 2002, 16, 1499-1506. (15) Lillieblad, L.; Szpila, A.; Strand, M.; Pagels, J.; Rupar-Gadd, K.; Gudmundsson, A.; Swietlicki, E.; Bohgard, M.; Sanati, M. Energy Fuels 2004, 18 (2), 410-417. (16) Kaufmann, H., Chlorine-compounds in emissions and residues from the combustion of herbaceous biomass. Ph.D. Thesis, Swiss Federal Institute of Technology, Zu ¨ rich, Switzerland, 1997. (17) Valmari, T.; Kauppinen, E. I.; Kurkela, J.; Jokiniemi, J. K.; Sfiris, G.; Revitzer, H. J. Aerosol Sci. 1998, 29, 445-459. (18) Lind, T.; Kauppinen, E. I.; Sfiris, G.; Nilsson, K.; Maenhaut, W. Environ. Sci. Technol. 1999, 33, 496-502. (19) Obernberger, I.; Brunner, T.; Jo¨ller, M. Proceedings of the International Seminar of IEA Bioenergy Task 32: Aerosols from Biomass Combustion, Zu ¨ rich, Switzerland, 2001; pp 69-74. (20) Olsson, J. G.; Ja¨glid, U.; Petterson, J. B. C. Energy Fuels 1997, 11, 779-784. (21) Davidsson, K. O.; Stojkova, B.; Petterson, J. B. C. Energy Fuels 2002, 16, 1033-1039. (22) Oser, M.; Nussbaumer, T.; Schweizer, B.; Mohr, M.; Figi, R. Proceedings of the International Seminar of IEA Bioenergy Task 32: Aerosols from Biomass Combustion; Zu ¨ rich, Switzerland, 2001; pp 59-64. (23) Gaegauf, C.; Wieser, U.; Macquat, Y. Proceedings of the International Seminar of IEA Bioenergy Task 32: Aerosols from Biomass Combustion, Zu ¨ rich, Switzerland, 2001; pp 81-85. (24) Hueglin, C.; Gaegauf, C.; Ku ¨ nzel, S.; Burtscher, H. Environ. Sci. Technol. 1997, 31, 3439-3447. (25) Svane, M.; Stojkova, B.; Hagstro¨m, M.; Petterson, J. B. C. Proceedings of the Sixth International Aerosol Conference, Taipei, Taiwan, 2002; pp 263-264. (26) Nussbaumer, T.; Hasler, P. Holz Roh Werkstoff 1999, 57, 13-22. (27) Maguhn, J.; Karg, E.; Kettrup, A.; Zimmermann, R. Environ. Sci. Technol. 2003, 37, 4761-4770. (28) Zimmermann, R.; Van Vaeck, L.; Davidovic, M.; Beckmann, M.; Adams, A. Environ. Sci. Technol. 2000, 34, 4780-4788. (29) Zimmermann, R.; Ferge, T.; Ga¨lli, M.; Karlsson, R. Rapid Commun. Mass Spectrom. 2003, 17, 851-859. (30) Cool, T. A. In Lasers and Mass Spectrometry; Lubman, D. M., Ed.: New York, 1990. (31) Williams, B. A.; Tanada, T. N.; Cool, T. A. In 24th Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, PA, 1992; pp 1587-1596. (32) Oser, H.; Thanner, R.; Grotheer, H.-H. Combust. Sci. Technol. 1996, 116-117, 567-582. (33) Gittins, C. M.; Castaldi, M. J.; Senkan, S. M.; Rohlfing, E. A. Anal. Chem. 1997, 69, 286-293.
1402
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 6, 2005
(34) Boesl, U., Nagel, H., Weickhardt, C., Frey, R., Schlag, E. W., Eds. Vehicle Exhaust Emission Time-resolved Multicomponent Analysis by Laser Mass Spectrometry; John Wiley & Sons: 1998. (35) Heger, H. J.; Zimmermann, R.; Dorfner, R.; Beckmann, M.; Griebel, H.; Kettrup, A.; Boesl, U. Anal. Chem. 1999, 71, 46-57. (36) Zimmermann, R.; Heger, H. J.; Kettrup, A.; Nikolai, U. Fresenius’ J. Anal. Chem. 2000, 366, 368-374. (37) Boesl, U. J. Mass Spectrom. 2000, 35, 289-304. (38) Mu ¨ hlberger, F.; Zimmermann, R.; Kettrup, A. Anal. Chem. 2001, 73, 3590-3604. (39) Grotheer, H.-H.; Nomayo, M.; Pokorny, H.; Thanner, R.; Gullett, B. K. Trends Appl. Spectrosc. 2001, 3, 181-206. (40) Cao, L.; Mu ¨ hlberger, F.; Adam, T.; Streibel, T.; Wang, H. Z.; Kettrup, A.; Zimmermann, R. Anal. Chem. 2003, 75, 5639-5645. (41) Hu ¨ glin, C. New Applications of Aerosol Photoemission: Characterization of Wood Combustion Particles and Time-Resolved Thermal Desorption Studies. Ph.D. Thesis, Swiss Federal Institute of Technology, Zu ¨ rich, Switzerland, 1996. (42) Niessner, R.; Wilbring, P. Anal. Chem. 1989, 61, 708-714. (43) Burtscher, H.; Schmidt-Ott, A.; Siegmann, H. C. Aerosol Sci. Technol. 1988, 8, 125-132. (44) Dorfner, R.; Ferge, T.; Yeretzian, C.; Kettrup, A.; Zimmermann, R. Anal. Chem. 2004, 76, 1368-1402. (45) Hafner, K., Untersuchungen zur Bildung brennstoffabha¨ngiger Stickoxide bei der Abfallverbrennung mittels on-line analytischer Messmethoden. Ph.D. Thesis, Fakulta¨t Wissenschaftszentrum Weihenstephan fu ¨ r Erna¨hrung, Landnutzung und Umwelt, Mu ¨ nchen, Germany, 2004. (46) Butcher, D. J. Microchem. J. 1999, 62, 354-362. (47) Sykes, D. C.; Woods, E.; Smith, G. D.; Baer, T.; Miller, R. E. Anal. Chem. 2002, 74, 2048-2052. (48) Oktem, B.; Tolocka, M. P.; Johnston, M. V. Anal. Chem. 2004, 76, 253-261. (49) Butcher, D. J. Microchem. J. 2000, 66, 55-72. (50) Pallix, J. B.; Schu¨hle, U.; Becker, C. H.; Huestis, D. L. Anal. Chem. 1989, 61, 805-811. (51) Steenvoorden, R. J. J. M.; Kistemaker, P. G.; Vries, A. E.; Michalak, L.; Nibbering, N. M. M. Int. J. Mass Spectrom. Ion Processes 1991, 107, 475-489. (52) Steenvoorden, R. J. J. M.; Hage, E. R. E.; Boon, J. J.; Kistemaker, P. G.; Weeding, T. L. Org. Mass Spectrom. 1994, 29, 78-84. (53) Nir, E.; Hunziker, H. E.; Vries, M. S. Anal. Chem. 1999, 71, 16741678. (54) Mu ¨hlberger, F.; Hafner, K.; Kaesdorf, S.; Ferge, T.; Zimmermann, R. Anal. Chem. 2004, 76 (22), 6753-6764. (55) Zimmermann, R.; Heger, H. J.; Kettrup, A. Fresenius’ J. Anal. Chem. 1999, 363, 720-730. (56) Zimmermann, R.; Heger, H. J.; Blumenstock, M.; Dorfner, R.; Schramm, K.-W.; Boesl, U.; Kettrup, A. Rapid Commun. Mass Spectrom. 1999, 13, 307-314. (57) Simoneit, B. R. T.; Schauer, J. J.; Nolte, C. G.; Oros, D. R.; Elias, V. O.; Fraser, M. P.; Rogge, W. F.; Cass, G. R. Atmos. Environ. 1999, 33, 173-182. (58) Simoneit, B. R. T.; Rogge, W. F.; Lang, Q.; Jaffe, R. Chemosphere: Global Change Sci. 2000, 2, 107-122. (59) Ohlemiller, T. J. In SFPE Handbook of Fire Protection Engineering, 3rd ed.; DiNenno, P. J., Drysdale, D., Beyler, C. L., Walton, W. D., Eds.; National Fire Protection Agency (NFPA): Qunicy, MA, 2002; pp 2/200-202/210. (60) Evans, R. J.; Milne, T. A. Energy Fuels 1987, 1, 123-137. (61) Bryers, R. W. Prog. Energy Combust. Sci. 1996, 22, 29-120. (62) Pyyko¨nen, J.; Jokiniemi, J. K. Fuel Process. Technol. 2003, 80, 225-262. (63) Zimmermann, R.; Blumenstock, M.; Heger, H. J.; Schramm, K.W.; Kettrup, A. Environ. Sci. Technol. 2001, 35, 1019-1030. (64) Grotkjaer, T.; Dam-Johansen, K.; Jensen, A. D.; Glarborg, P. Fuel 2003, 82, 825-833. (65) Streibel, T.; Hafner, K.; Mu ¨ hlberger, F.; Adam, T.; Zimmermann, R. Manuscript in preparation. (66) Wikstro¨m, E.; Ryan, S.; Touati, A.; Gullet, B. K. Environ. Sci. Technol. 2004, 38, 2097-2101.
Received for review April 2, 2004. Revised manuscript received October 25, 2004. Accepted November 12, 2004. ES049493O
