Combustion aerosols from municipal waste incineration - effects of feedstock composition and boiler operation

Combustion aerosols were measured in the 22 MWth FASAN Waste-to-Energy (WtE) boiler while applying changes in feedstock composition and grate operation conditions, respectively. The changes in feedstock composition were applied by adding (one-by-one) dedicated waste fractions, comprising PVC plastic, CCA(Copper-Chromate-Arsenate)-impregnated wood, shoes, salt (NaCl), batteries and automotive shredder waste, to a base-load waste. The changes in the operational conditions investigated included minimization or maximization of the oxygen content, redistribution of primary and secondary air, and extension of the length of the combustion zone.

COMBUSTION AEROSOLS FROM MUNICIPAL WASTE INCINERATION EFFECTS OF FEEDSTOCK COMPOSITION AND BOILER OPERATION Jacob Hjerrild Zeuthen 1, Anne Juul Pedersen *1 , Christian Riber2 , Thomas Astrup 2, J0 rn Hansen 1, Flemming Jappe Frandsen1, Hans Livbjerg1 1 2 Department of Chemical and Biochemical Engineering, Technical University of Denmark, Building 229, DK-2800 Kgs. Lyngby, Denmark Department of Environmental Engineering, Technical University of Denmark, Building 115, DK-2800 Kgs. Lyngby, Denmark * Corresponding author: Tel: [phone], Fax: [phone], e-mail: [e-mail] Abstract Combustion aerosols were measured in the 22 MWth FASAN Waste-to-Energy (WtE) boiler while applying changes in feedstock composition and grate operation conditions, respectively. The changes in feedstock composition were applied by adding (one-by-one) dedicated waste fractions, comprising PVC plastic, CCA(Copper-Chromate-Arsenate)-impregnated wood, shoes, salt (NaCl), batteries and automotive shredder waste, to a base-load waste. The changes in the operational conditions investigated included minimization or maximization of the oxygen content, redistribution of primary and secondary air, and extension of the length of the combustion zone. The aerosols were measured in the flue gas at the outlet of the boiler, upstream of the flue gas cleaning devices. Mass-based particle size distributions were measured using a cascade impactor (LPI) and the number-based size distributions were measured using a Scanning Mobility Particle Sizer (SMPS). Chemical analysis of the aerosol particles was made by Energy Dispersive X-ray Spectroscopy (EDS), and the particle morphology was investigated by Transmission Electron Microscopy (TEM). The mass-based particle size distribution was bimodal with a fine mode peak around 0.4 ^m and a coarse mode peak around 100 ^m. The aerosol mass-load was very stable when no changes in fuel composition were introduced (PM2.5: 252 ± 21 mg/m3) and increased significantly during combustion of waste when including automotive shredder waste, CCA-impregnated wood, and NaCl (PM2.5: 313, 320, and 431 mg/m3). The number concentrations of particles varied during combustion and between varying operational conditions and waste mixes (from 43 106 to 87106 #/cm3). The aerosols formed were mixtures of dense and aggregated particles in all tests. The particles were mainly composed of alkali chlorides and sulphates. In all runs the six major elements by weight were Cl, Na, K, S, Zn and Pb. The Cl content was increasing with the particle diameter. The content of alkali (Na + K) was rather constant for all runs, and the molar Na/K ratio decreased with the particle diameter indicating that the sodium salts condenses at higher temperatures than those of potassium. The ratio increased significantly when adding salt (+ 67 %) or shredder waste (+ 112 %) to the base­ load waste. The Zn content increased with particle diameter and for the runs with shredder waste, shoes, PVC and impregnated wood the content was increased significantly (> 7 times increase in case of shredder waste). The content of Pb was rather constant as a function of particle diameter and throughout the series of runs. However, for the shredder waste combustion a 40 % increase was found. Smaller amounts of Fe, Cu, Mn, Cr, and Hg were found in the finest particles in some runs with waste containing these elements. 1. INTRODUCTION The Danish strategy for waste management is still to increase recycling, and, at the same time, to reduce the volume of land-filled waste in order to avoid loss of resources. Combined heat and electricity producing waste-to-energy (WtE) boilers are an integrated part of this strategy. In 2005, 24 % of the total reported Danish waste production - primarily municipal solid waste from house holds, the service sector, and the industry - was incinerated at 29 WtE plants located all over Denmark. The main environmental concern for waste incineration nowadays is the leaching of hazardous elements from the solid residues [1]. The combustion aerosols that penetrate the flue gas cleaning devices and emit to the atmosphere carry small amounts of harmful components and therefore pose another serious health risk. However, new and advanced flue gas cleaning technologies ensure very low toxic emissions from modern WtE plants; most plants are equipped with flue gas cleaning devices such as wet-scrubbers, baghouse filters, absorbers or electrostatic precipitators. Filters and electrostatic precipitators are extremely effective in decreasing the particulate content of the flue gas when considering the total mass. Still, fine particles (defined as particles with an aerodynamic diameter below 2.5 ^m (PM2.5)) will always penetrate most cleaning devices to some extent, and the retention characteristics for sub-micron particles are not very well characterized. In addition, toxic heavy metals are condensed and concentrated in the fine particles. Particulate matter from municipal solid waste incinerator stacks has been found to contain high amounts of the heavy metals Fe, Zn, Mg, Mn, Cu, Ni, Pb, Cr, Hg, and Cd [2 - 4]. Only a few studies on the generation of submicron particles from waste incineration have been published. Maguhn et al. [5] have reported particle size distribution obtained by SMPS in a plant equipped with electrostatic precipitator, Yoo et al. [4] have reported particle size distributions in stacks from plants with cyclone separators, and Chang et al. [6 ] have reported mass-based size distributions measured by a cascade impactor in two incinerators. The objective of the present work is to analyze results from aerosol measurements from full-scale combustion experiments conducted at the "I/S FASAN" WtE plant in Nrestved, Denmark, while applying changes in feedstock composition and operating conditions, respectively. The measuring campaign was conducted as part of a larger research project dealing with the possibilities of improving the ash quality, and the overall plant performance, during waste incineration (PSO Contract No. 5784). Results from the present aerosol study have previously been reported by Zeuthen et al. [7]. The present paper sums up findings from that work, while also including some new aspects (link to fly ash and deposit chemistry) in the discussion, based on supplementary findings from related work [ 8 - 1 0 ] 2. EXPERIMENTAL 2.1. The measuring campaign The full-scale measuring campaign was conducted at Furnace Line 4 (22 MVth) at the grate-fired WtE plant "I/S FASAN", situated in the town of Nrestved, Denmark. During the different fullscale test runs, samples of flue gas, fly ash, aerosols and deposits were taken at different locations in the boiler, as indicated in Figure 1. A base-load waste consisting of 80 % municipal solid waste from households and 20 % "small combustible waste" from recycle stations was used as reference fuel in the full-scale measurements (a composition close to the mixed municipal solid waste that is incinerated during daily operation, but with the exclusion of industrial waste fractions). In order to evaluate the effects of changing waste input composition, special waste fractions were added one-by-one to the base-load waste in different test runs. The dedicated waste fractions all contained high concentrations of one or more potentially harmful element, i.e., a heavy metal and/or Cl, as compared to the base-load waste. The fractions, which were obtained from different recycling facilities in Denmark, were comprised of household batteries, automotive shredder waste, CCA (Copper-Chromate-Arsenate)-impregnated wood, PVC plastics, and (leather) shoes. Finally, to get an input of inorganically bound Cl, road de-icing salt (NaCl) was also included as a special waste fraction. Figure 1: Sketch of the FASAN WtE plant (Furnace Line 4), Nrestved, Denmark, with the points of sampling indicated. (NID: Novel Integrated Desulphurization) The special fractions were added to the base-load waste in concentrations ranging from app. 0.5 to app. 14 % (w/w); the mixing ratios were defined in order to obtain a substantial increase in a particular heavy metal and/or Cl concentration in the combustion residues, while at the same time ensuring an acceptable calorific value of the mixed fuel. The concentration of some characteristic elements in the resulting mixed input waste is listed in Table 1. For more details please refer to Pedersen et al. [8 ], and Astrup et al. [10]. PVC (5.5 %) CCA imp. Wood ( 11.1 %) Shoes (1.6 %) Salt (0.5 %) Base-load waste (100 %) ± 2STDEV )% % 5. .2 f (0. s 1 rf 1 a ei ri te ed ) ed .2 te ta £ O 2 zn. B 571± 396 475 488 830 1672 719 Pb (mg/kg d.b.) 950 Zn (mg/kg d.b.) 1613± 898 2058 1692 5574 1787 2127 1659 0.347 ± 0.152 0.393 0.480 0.379 0.875 0.442 S (wt% d.b.) 0.529 1.44 ± 0.13 2.25 2.08 1.49 1.55 1.54 3.03 Cl (wt% d.b.) 0.430 ± 0.128 0.427 0.392 0.396 0.490 0.526 0.364 K (wt% d.b.) 1.11 ± 0.33 1.34 1.15 1 .1 2 1.18 1.27 0.988 Na (wt% d.b.) Table 1: Concentrations of characteristic elements in the mixed input waste. Figures highlighted in bold indicate that the concentration is increased significantly (> 2 STDEV) as compared to the base-load waste composition. etts a £ d % Imp. wood > 7.9 7.5 7.8 7.9 8.0 14.3 176 14.7 175 14.2 176 11.7 175 13.8 177 1.1 1 .0 0.9 0.9 0.9 0.9 1.1 1.0 1.1 1.0 C P Shoes Shredder waste 7.3 8.3 7.2 9.1 7.9 7.5 7.6 O2 conc. in flue gas (vol. %, wet) 14.9 15.3 17.4 13.5 15.3 15.0 14.9 H 2O (vol. %) Tflue gas, sampling 175 173 176 184 175 175 175 point (oC) 0.9 1.1 1 .0 1.2 1.1 0 .8 1 .0 Primary Air Flow 0.8 0 .6 1.4 0.7 1.2 1 .0 1.1 Secondary Air Flow 1.0 0.9 1.2 1 .0 0.9 1 .0 1 .0 Total Air Flow 65/ 76/ 79/ 65/ 78/ 59/ 70/ Distribution 35 24 21 35 22 41 30 prim./sec. air (%) Table 2: Operating conditions during full-scale test runs at I/S plant's data collection system) Batteries Decreased secondary air Increased secondary air Salt Max. oxygen Undefined conditions Reference U9§£xO UI]/^ Test run name The changes in the operational conditions investigated included minimization or maximization of the oxygen content in the flue gas, and redistribution of primary and secondary air around the grate. It was also attempted to extend the length of the combustion zone (the visible flames) along the grate, but due to various unstable combustion conditions during this test run, it was instead termed "undefined conditions"). In order to ensure steady-state conditions in the full-scale experiments, the changes in feedstock composition or operating conditions were initiated at least 4-6 hours before the samplings were started, and each measuring period had a duration of several hours. A total of 12 test runs were conducted with complete sampling data collected for all but one of the runs. The different test runs, along with a short description of their characteristic operating conditions, are listed in Table 2. 1.1 1 .0 1.0 1 .0 1.0 74/ 26 69/ 31 68/ 65/ 35 68/ 32 32 FASAN (as registered by the 2.2. Aerosol sampling Aerosol measurements were performed before and after the flue gas cleaning. Total dust measurements were also performed before the filter. The measurements of mass-based size distributions of aerosols were made using a 10-stage Berner Low-Pressure Cascade Impactor (Hauke GmbH LPI). Number-based size distributions were obtained using a Scanning Mobility Particle Sizer (SMPS). The SMPS system consists of a Differential Mobility Analyzer (TSI 3071 DMA) and a Condensational Particle Counter (TSI 3010 CPC). The mass concentration of coarse particles (total dust) was obtained by filter methods and the particle size distribution for coarse particles was determined by Laser diffraction using a Malvern Mastersizer S long bed. The measurements were obtained from a dry sample and the measuring range was 0.49-754 ^m. Detailed results from the total dust measurements are reported in ref. [8 ]. The LPI and SMPS measurements operated with diluted flue gas withdrawn from the the duct by a gas ejector. The ejector dilutes flue gas into a flow of dry, filtered air [7, 11]. All concentrations given in the following have been corrected for the dilution. The dilution serves a three-fold purpose: Water condensation is prevented because the sample during dilution and cooling is held well above the water dew point. Coagulation in the sample line becomes effectively quenced because its rate is lowered by several orders of magnitude due to the dilution. The dilution also reduces the particle concentration to avoid overloading of the LPI and, more important, to keep the CPC of the SMPS system low to ensure counting of single particles. The overloading of the LPI can be avoided by reducing the time of sampling. However, a long sampling time reduces the influence of fluctuations in the operation of the plant. The dilution ratio varies with time due to changes in flow and pressure in the duct and clogging of the ejector capillary. To determine the precise dilution ratio during a measurement (up to 30 minutes) the dilution ratio is measured online by comparing the CO2 concentration in the diluted flue gas with the concentration in the undiluted flue gas. Each sampling line is dried and filtered continuously in gas conditioners and fed into an IR CO2-analyzer (Rosemount, NGA 2000). Changing the internal diameter of the inlet capillary of the ejector changes the dilution ratio. Dilution ratios in the range 5 to 200 can be chosen. By inserting the ejector in the duct with the capillary at an angle of 90o to the direction of the gas flow, it is ensured that the largest particles are pre-filtered from the sampling line by inertial forces while the submicron particles are sampled undisturbed. This pre-filtering is necessary in order to prevent overloading of the impactor by large particles. With the conditions used in the present measurements a cut-off diameter of ~3 ^m for the pre-filtering is obtained [7, 12]. The entire sampling setup is shown in Figure 2. Figure 2: Setup for the aerosol measurements. The two CO2 concentrations are stored by a data-logger and saved on a computer. The ejector capillary can be cleaned by "soot-blowing", i.e. by blocking the gas exit and forcing pressurized air through the capillary. The cascade impactor has an aerosynamic diameter range from 0.03 to 12.7 ^m. Deposited particles in the impactor are collected on aluminum foils. To reduce re­ bounce of the particles, the foils are coated with a thin film of Apiezon H grease using a dilute toluene solution of the grease. The weight gain from the deposited particles was determined with a Sartorius M5D-000V001 micro balance. The LPI is thermostated at 60o C before the measurements to avoid water condensation from the flue gas. Tygon tubes are used in the sampling line to avoid deposition due to static electrical charges. The sampling time for a LPI measurement was adjusted in order to achieve a suitable amount of deposits; in most cases 30 minutes proved accurate. Although plugging of the ejector capillary occurred quite rapidly in some runs, frequent soot-blowing sufficed to keep the ejector functioning on all runs. In between runs the ejector was withdrawn from the flue gas duct and cleaned with water and ethanol. The CO2 analyzers were calibrated daily. Deposits on impactor foils were analyzed by Energy Dispersive X-ray Spectroscopy (EDS) to obtain the chemical composition. In the SMPS system the particles are classified by their mobile diameter in the DMA with a range from 14 to 800 nm. The particles are counted in the CPC with high accuracy giving a finely resolved number-based size distribution. The SMPS system is able to determine the size distributions for particles at very low concentrations. In this campaign it was possible to place the instrument in a way that allowed switching between measuring particles from the ejector probe (before the flue gas cleaning) and particles from the clean gas in the stack. The cleaned gas was withdrawn through a 3 m Tygon tube. Due to the low number concentration of particles after the filter the coagulation in the tube is negligible. Particles were collected for morphology studies by Transmission Electron Microscopy (TEM). Cu-grids with carbon film were emerged in the diluted aerosol from the ejector exit for approximately three minutes to obtain an adequate density of particles by diffusional deposition. 3. RESULTS AND DISCUSSION 3.1. Mass-based particle size distribution The mass-based particle size distributions for the runs with changes in the operational conditions are shown in Figure 3, together with the average of the measurements from the reference runs. The references are based on a total of 8 successful reference measurements from 3 different days. The size distributions from runs with changes in the operational conditions are each based on three measurements. Mobile diameter (nm) Figure 3. Mass-based size distributions from runs with base-load waste measured with cascade impactor. The aerodynamic diameter of the particles are converted to a mobile diameter by assuming particle density of 2 g/cm3 (that of alkali chlorides) and using the interpolation formula given by Baron and Willeke [13], see Zeuthen et al. [7]. The results from the test-runs with addition of waste fractions are shown in Figure 4, and the overall properties of the size distributions are summarized in Table 3. The total mass concentration is given by the area under each curve and is obtained by integration. GMD refers to the geometric mean diameter, see Zeuthen et al. [7]. cu H ~t&) p D ( o Mobile diameter (nm) Figure 4. Mass-based size distributions from runs with different waste fractions measured with cascade impactor. The changes in operational conditions do not have a significant effect on the mass-based particle size distribution. The addition of salt, shredder waste, and CCA-impregnated wood increases the mass concentration, and shredder waste, batteries, and salt increases the mass-based GMD. The increased mass load of fine particles when adding salt is not found when adding PVC, in spite that in both cases high amounts of chlorine is added. However, the chlorine in PVC is not bound in the particles, probably due to lack of alkali metals or other condensable cations. The chlorine in this case is released as HCl(g). The mass concentrations calculated here do not cover the total particulate concentration. A peak, originating for non-combustible inclusions, is found above the range of the cascade impactor. The total dust concentration is found to 2.2 g/ m3 (std. T,P) at reference conditions. As the PM2.5 in this run was 0.23 g/m3 (std. T, P), the PM2.5 fraction of the total mass concentration is thus 10.5 % (w/w). 3.2. Number-based particle size distribution The size distribution for the aerosols before the flue gas cleaning, obtained by the SMPS system, is shown in Figure 5. The deviation between the measurements was larger than for the LPImeasurements. This was probably due to the short measuring time for one scan (~5 min). The curves shown here are all average values of several scans during 1 day. A very large change in the number-based size distributions may cause only a minor change in the mass-based size distribution, because the mass of the smallest particles is so small that it only gives a vanishing contribution to the mass concentration. The total number concentration is given by the area under each curve and is obtained by integration. The integrated properties of the mass- and numberbased particle size distributions are given in Table 3. cu H o 1000 Mobile diameter (nm) Figure 5. Number-based particle size distributions. The mass-based PM25 value for the runs with base-load waste is very constant (213 - 274 mg/m3) while the number concentration varies significantly (44 - 87 . 106 #/cm3). For the runs with different waste fractions added to the base-load waste, the variation in both mass and number concentration is significant. The number-based geometric mean diameter (GMD) is changed in a few runs. The reference value 167 nm is increased to 188 - 222 nm when adding the waste fractions salt, batteries, PVC and shoes. The geometric mean diameter is also increased to 191 nm for the run with increased secondary air. The particles are slightly larger than the ones reported by Maguhn et al. [5]. Reference * Undefined cond. Minimum oxygen Maximum oxygen Increased secondary air PM2.5 (mg/m3) 251.8±21.3 241.3±26.8 236.8±13.4 213.5±4.7 270.4±24.4 # Conc. (106/cm3) 50.1±7.6 86.7±5.0 62.1±18.3 61.3±7.9 50.7±2.0 Mass-GMD (nm) 405±23 380±34 404±28 368±21 363±16 Number-GMD(nm) 167±12 158±54 158±23 151±12 191±11 Waste fractions Salt Batteries Shredder Imp. wood PVC Shoes PM2.5 (mg/m3) 437.1 ±44.0 267.0±46.0 312.7±19.8 319.7±21.8 257.2±9.9 185.9±9.6 # Conc. (106/cm3) 71.2±28.4 43.8±13.7 63.0±20.5 54.6±6.7 42.8±5.1 53.9±1 Mass-GMD (nm) 488±156 446±4 479±107 371±30 410±65 394±23 Number-GMD(nm) 213±44 222±30 176±74 151±16 188±18 191±6 Operational conditions Table 3. Properties for fine particles from runs with changes in operational conditions or with addition of different waste fractions. *: Average of three average values for the three days with reference runs. The standard deviations for these data reflect the deviation between the different reference runs. All other standard deviations reflect the deviation during a single run. The SMPS measurements on the cleaned flue gas in the stack revealed that the average number concentration had decreased to 6.9 . 104 #/cm2, corresponding to an overall number-based penetration of fine particles of 0.11 % [7]. 3.3. Chemical composition The chemical composition of the fine particles from the runs with different waste fractions added are shown in Figure 6 and Figure 7. Only the runs with different add-on fuels were investigated. The operational conditions were not expected to influence the composition considerably. For each measured element the content is depicted for all 6 runs with changes in fuel composition together. The error bars on the reference contents indicate the standard deviation between 3 runs from 2 different days with reference measurements. The accuracy of EDS is ~ ± 1 % and measured concentrations below 1 % are generally not reliable. The elements Cd and Ni were also measured but the content was below 1 % (w/w) for all sizes in all runs and the results are not shown. In all runs the six major elements by weight are Cl, Na, K, S, Zn and Pb. The chlorine content is increasing with the particle diameter. The content of alkali (Na + K) is rather constant for all runs. The molar Na/K ratio decreases with the particle diameter indicating that the sodium salts condenses at higher temperatures than those of potassium. The ratio is increased significantly when adding salt (+ 67 %) or shredder waste (+ 112 %) to the base-load waste. For shredder waste this increase is also due to very low potassium content in the particles. The zinc content increases with particle diameter, and for the runs with shredder waste, shoes, PVC and impregnated wood the content is increased significantly. For the shredder waste run the content by weight is increased more than seven times with respect to zinc. The content of lead is rather constant as a function of particle diameter and throughout the series of runs. However, for the shredder waste combustion a 40 % increase is found. t -C ei £ n te no c ne el 10 100 1000 1000 3 1000 10 100 1000 Mobile diameter (nm) Figure 5: Size dependent elemental content of main components of the combustion aerosols in runs with addition of special waste fractions. The contents of Fe, Cu, Cr, As, and Mn are below the 1 % accuracy limit for the larger aerosol particles. However, for these elements a significant amount is found in the very fine particles in some runs, indicating that these elements condense at high temperatures (homogeneous condensation). Fe and Cu are found as major elements in the smallest particles during shredder waste and battery combustion and the two elements are found in the fine particles for all runs. Chromium is found in the smallest particles during battery, shoes and shredder waste combustion, while manganese is found mainly during shredder waste, battery and PVC combustion. Arsenic is found in the smallest particles during combustion of shredder waste and during one of the reference runs. 4 - Mn - R e fe re n c e - S alt B atteries 2 --S h red d er - PVC - S hoes Imp. w ood 0 100 1000 Mobile diameter (nm) Figure 6 : Size dependent elemental content of minor components of the combustion aerosols in runs with addition of special waste fractions. The content of Ca is above the accuracy limit in most cases, but no clear trends are found between particle diameter and content. The content is largest during combustion of shredder waste, PVC and shoes. The Hg-content of the particles was only analyzed for the reference, battery and shredder waste runs. For the reference run the mass content is below 0.5 in all stages. For the battery run 2.8 % is found in the smallest particles and in the shredder waste run 4.6 % is found in the smallest particles. The measurements of Ni (not shown) indicate that the element is concentrated in the finest particles. The Ni content is close to the accuracy limit during combustion of PVC, shredder waste and batteries. In general, the deviation between the three reference runs analyzed with EDS is quite low. Only the content of K, Zn, and S deviates considerably as seen on the depicted error bars. 3.4. Morphology The morphology of the particles found by TEM imaging of particles deposited on Cu-grids reveals a broad variety of shapes and structures. Some particles are polyhedral and probably crystalline, while most are close to a spherical shape with smaller particles attached to their edges. Micrographs showing particles from a reference run is shown in Figure 7. Figure 7. TEM images of particles from a reference run. The particles vary in size and morphology. Cubic shapes are seen and also some aggregates are found. In the left picture, examples of large particles with cubic-like structures and aggregate-like structures are shown. Figure 8 . TEM images of particles from run with increased secondary air. Left: Aggregates much larger than those found for other runs. Right: Graphite like structures are found on the surface of the particles in this run. Most particles are a few hundred nanometers in diameter as expected from the number-based particle size distribution. The cubic structures are generally found in the larger particles. This suggests that chlorides are condensed at a lower temperature than the one where particles are formed. EDS on single particles with cubic structures showed an increased content of alkali chlorides but also significant amounts of heavier elements reflecting that chlorides are condensed on other particles. The average size of particles does not change much between the different runs, and for the runs with addition of waste fractions the particle morphology is very similar to the reference case. Only in the runs with addition of salt, batteries, and shoes, and when increasing the secondary air, the particle size is increased. The size-increase in these runs is probably due to increased heterogeneous condensation after the particle formation. This is also reflected in the PM-values measured by the LPI for these runs. In the run with increased secondary air the morphology is changed. More and larger aggregates are formed. The aggregates have a different structure with single particles glued together. When looking at the surface of some of the particles, a layer-structure is observed. The structure looks like graphite. The aggregates and the layer structure is shown in Figure 8 A similar layer-structure is observed on some particles from the "undefined conditions" run and indicate incomplete combustion conditions. The chemical compositions of the single particles observed by TEM were very similar to the compositions determined from the LPI foils on similar particle sizes. The Zn content of single particles vary considerably more than the other elements. 3.4. Particle formation The formation of fine particles during waste combustion is caused by homogeneous nucleation of volatile matter during cooling of the flue gas. The particle number concentration described here varies considerably with time. This could be due to a non-steady release of the nucleating components of the fuel. Maguhn et al. [5] also found that the number concentration varied significantly with time. In their measurements, the number concentration correlates well with the SO2 concentration of the flue gas, suggesting that sulphur-containing compounds are responsible for nucleation. This correlation is not evident from our data. The number concentration found in biomass combustion span from 6 . 106 to 35 . 106 3/cm3 [14] and are lower than the ones found here. Even though the main composition is rather similar to the one found for aerosol particles from biomass combustion the particles are probably not formed by the same mechanism. The higher number concentration and the lack of correlation between SO2 concentration and particle number concentration suggest that the release of the nucleating components might be very unsteady due to the inhomogeneous fuel. 3.5. Link to fly ash and deposit chemistry We previously suggested that an increased mass-load of fine particles when adding salt to the feedstock, but not when adding PVC, reflects that the chlorine in PVC is released as HCl(g), while the chlorine in salt is particle bound. This is in good consistency with the findings from a related study from the same measuring campaign, dealing with the fly ash chemistry and overall element partitioning in the boiler [8 ]. In this study it was found that organically bound Cl (as in PVC and shoes) increased the concentration of HCl(g) in the flue gas, while the inorganically bound Cl (in salt) was preferably recovered in the bottom ash and fly ash fractions. Addition of salt also changed the chemical composition and particle size distribution of the total fly ash, towards an increased concentration of Na and Cl, and a smaller particle size, indicating increased condensation of alkali-chlorides [8 ]. A study of the deposit formation during the same campaign [9], however, displayed somewhat different results. Though, in the interpretation of these results one should bear in mind that the deposits were sampled at the entrance to the convective section, at a flue gas temperature close to 600 oC and with a surface temperature of the deposition probe of 400 oC, whereas the aerosols were sampled downstream of the convective section, at a flue gas temperature of app. 175 oC (see Figure 1). This gives rise for different ash transformations in the two cases. It was found in the deposition study that the rate of deposit build-up (g/m2/h), as well as the concentration of Cl in the deposits, was increased when firing PVC, while no significant effects on the deposit formation were observed when firing salt. The chemical analysis of the deposits from the PVC run revealed that the deposits were also relatively high in [Ca], while the alkali content was relatively low [9]. Hence we suggest that the increased deposition flux, and high [Cl], in the deposits from the PVC run may be explained by increased condensation of CaCl2 on the deposition probe in this specific run. Comparing the chemical composition of the aerosols with the chemical composition of the total fly ash, as measured by ICP (Inductively Coupled Plasma) methods, confirms that the main elements of the aerosols, Na, K, S, and Cl, are enriched significantly in the aerosols as compared to the total fly ash, as are the heavy metals Cu, Cd, Hg and Pb. Zn is enriched in some runs [7, 8 ]. When comparing the aerosol and total fly ash chemistry for each test run, the main trends are quite similar for the elements Na, K, S and Cl, while for Pb and Zn, some deviations are observed. [Pb] in the aerosols is increased significantly in the shredder waste run only (Figure 5), while for the total fly ash, the PVC and shoes runs gave rise to significantly increased [Pb] [8 ]. This may suggest increased heterogeneous condensation of Pb-salts at relatively low temperatures in the PVC and shoes runs (probably Pb-chlorides), while Pb-compounds with higher condensation temperatures dominate in the shredder waste run (e.g. nucleation of Pb-oxides). The increased [Zn] in the (larger) aerosols in the PVC and shoes runs (Figure 5) could indicate increased heterogeneous condensation of Zn-salts in these runs (probably Zn-chlorides). As [Zn] in the total fly ash was only increased slightly in these two runs [8 ], the results also indicate that the Zn-salts condense at relatively higher temperatures than those of Pb. 4. CONCLUSIONS Combustion aerosols from a 22 MWth WtE plant were characterized during runs with different waste inputs and operational conditions. The aerosol mass-load was very stable when no changes in fuel composition were introduced (PM25: 252 ± 21 mg/m3) and increased significantly during combustion of waste when including automotive shredder waste, impregnated wood, and NaCl salt (PM25: 313, 320, and 431 mg/m3). The particle size distribution is bimodal with approximately 10 % (w/w) of the total particle mass in the PM2.5 fraction. The number concentration of particles varied during combustion and between varying operational conditions and waste mixes (from 43 . 106 to 87 . 106 #/cm3). The particles were mainly composed of alkali chlorides and sulphates. A significant content of Zn and Pb was found in all runs, while smaller amounts of Fe, Cu, Mn, Cr and Hg were found in some runs with waste containing these elements. The aerosol characteristics were linked to the fly ash and deposit chemistry, and the results suggested different ash transformation mechanisms depending on the waste types incinerated. The results indicate increased heterogeneous condensation of Pb and Zn when firing PVC and shoes, probably due to formation of Pb- and Znchlorides. ACKNOWLEDGMENTS This specific work was carried out under PSO-Contract No. 5784, financed by I/S Vestforbranding, DONG Energy A/S, Aarhus Kommunale Vrerker, I/S Amagerforbranding, Babcock & Wilcox V 0 lund A/S, and Energinet.dk. The work was carried out in collaboration between the Combustion and Harmful Emission Control (CHEC) Research Centre at the Department of Chemical and Biochemical Engineering, Technical University of Denmark, and the Department of Environmental Engineering, Technical University of Denmark. The CHEC Research Centre is co-funded by DONG Energy, Vattenfall, the Danish Energy Research Programme, Nord_En Research, the Technical Research Council, and the European Union. We specifically want to thank people from I/S FASAN WtE Plant for participating in the measuring campaign. REFERENCES [1] Riber, C.; Frederiksen, G. C.; Christensen, T. H. Waste Manage. Res. 23, 126-132 (2005) [2] Mateu, J., de Mirabo, F.B., Forteza, R., Cerda, V., Colom, M., Oms, M., Water, Air, Soil Pollut. 112, 349-363 (1999) [3] Thipse, S.S., Schoenitz, M., Dreizin, E.L., Fuel Process. Technol. 75, 173-184 (2002) [4] Yoo, J.I., Kim, K.H., Jang, H.N., Seo, Y.C., Seok, K.S., Hong, J.H., Jang, M., Atmos. Environ. 36, 5057-5066 (2002) [5] Maguhn,J., Karg,E., Kettrup,A., and Zimmermann, R., Environ. Sci. Technol. 37, 4761-4770 (2003) [6 ] Chang, M.B., Huang, C.K., Wu, H.T., Lin, J.J., Chang, S.H., J. Hazard. Mater. 79, 229-239 (2 0 0 0 ) [7] Zeuthen, J. H., Pedersen, A. J., Riber, C., Astrup, T., Hansen, J., Livbjerg, H., Combust. Sci. Technol. 179, 2171-2198 (2007) [8 ] Pedersen, A. J., Frandsen, F. J., Riber, C., Astrup, T., Thomsen, S. N., Lundtorp, K., Mortensen, L. F., Energy & Fuels, (2009) (in press) [9] Frandsen, F. J., Pedersen A. J., Hansen, J., Madsen, O. H., Lundtorp, K., Mortensen, L., Energy & Fuels, 2009 (in press) [10] Astrup, T.; Riber, C.; Pedersen, A. J. Influence of waste input and operational parameters on municipal solid waste incinerator emissions and ashes. Environ. Sci. Technol. (submitted) [11] Nielsen, L.B., Jensen, L.S., Pedersen, C., R0 kke, M., Livbjerg,H., J., Aerosol Sci. 27, 365­ 366 (1996) [12] Nielsen, M.T., Livbjerg, H., Fogh, C.L., Jensen, J.N., Simonsen, P., Lund, C., Poulsen, K., Sander, B., Combust. Sci. Technol. 174, 79-113 (2002) [13] Baron, P.A., and Willeke, K., Aerosol Measurements - Principles, Techniques and Applications. John Wiley & Sons (2005) [14] Nielsen, M.T., Field study on combustion aerosols., Ph.D. dissertation, Department of Chemical Engineering, Technical University of Denmark, 2001