Environ. Sci. Technol. 1985, 19, 796-804
initial liauid Dhase concentration of bound solute, mass basis, ig/kg of TOC S solids concentration, kg/L TOC total organic carbon, kg/L yw chemical activity in water yoct chemical activity in octanol ywloctchemical activity in octanol-saturated water yoctlw chemical activity in water-saturated octanol ypoc chemical activity in particulate organic carbon Registry No. 2,4,5,2’,4’,5’-Hexachlorobiphenyl, 35065-27-1; a,
_I
2,5,2’-trichlorobiphenyl, 37680-65-2; naphthalene, 91-20-3; chlorobenzene, 108-90-7.
(13) Schwarzenbach, R.; Westall, J. Environ. Sci. Technol. 1981, 15,1360. (14) Carter, C. W.; Suffet, I. H. Enuiron. Sci. Technol. 1982,16, 735. Landram, P. F.;Nihart, S. R.; Eadie, B. J.; Gardner, W. S. Environ. Sei. Technol. 1984,18, 187. Weber, W. J., Jr.; Voice, T. C.; Pirbazari, M.; Hunt, G. E.; Ulanoff, D. M. Water Res. 1983,17,1441. Weber, W. J., Jr.; Voice, T. C.; Jodellah, A. J.-Amer. Water Works Assoc. 1983,95,612. DiToro, D. M.; Horzempa, L. M.; Casey, M. M.; Richardson, W. J. Great Lakes Res. 1982,8,336. Eadie, B. J.; Robbins, J. A.; Landrum, P. F.; Rice, C. P.; Simmons. M. S.: McCormick. M. J.: Eisenreich. S. J.: Bell. G. L.; Pickett, R. L.; Johansen, K.;‘Rossman, R.;Hawley, N.; Voice, T. C. “The Cycling of Toxic Organic Substances in The Great Lakes Ecosystem-A Three Year Status Report”. National Oceanic and Atmospheric Administration, 1983, NOAA Technical Memorandum, ERL GLERL-45. (20) Rice, C. P., The University of Michigan, unpublished data, 1982. (21) Rice, C. P.; Meyers, P. A.; Brown, G. S. In “Physical Behavior of PCBs in the Great Lakes”, Mackay, D.; Paterson, S.; Eisenreich, S. J.; Simmons, M. S., Eds.; Ann Arbor Science: Ann Arbor, MI, 1983. (22) Eadie, B. J.; Rice, C. P.; Frez, W. A. In “Physical Behavior of PCBs in the Great Lakes”; Mackay, D.; Paterson, S.; Eisenreich, S. J.; Simmons, M. S., Eds.; Ann Arbor Science: Ann Arbor, MI, 1983. I
eature Cited Bailey, G. W.; White, J. L. Residue Rev. 1970,32, 39. Hamaker, J. W.; Thompson, J. M. In “Organic Chemicals in the Soil Environment”; Goring, C. M.; Hamaker, J. W., Eds.; Marcel Dekker: New York, 1972;Vol. 1, pp 49-143. Voice, T. C.; Weber, W. J., Jr. Water Res. 1983,17,1433. O’Connor, D. J.; Connolly, J. P. Water Res. 1980,14,1517. Voice, T. C.; Rice, C. P.; Weber, W. J., Jr. Environ. Sci. Technol. 1983,17,513. DiToro, D. M., Horzempa, L. M. In “Physical Behavior of PCBs in the Great Lakes”; Mackay, D.; Paterson, S.; Eisenreich, S. J.; Simmons, M. S., Eds.; Ann Arbor Science: Ann Arbor, MI, 1983;pp 89-113. Curl, R. J.; Keioleian, G. Environ. Sei. Technol. 1984,18, 916. Gschwend, P. M.; Wu, S.4. Environ. Sei. Technol. 1985, 19,90. Chiou, C. T.; Freed, V. H. Environ. Sei. Technol. 1977,11, 1220. Chiou, C. T.; Schmedding, D. W.; Manes, M. Environ. Sei. Technol. 1982,16, 4. Platford, R. F.J . Great Lakes Res. 1982,8,301. Leo, A,; Hansch, C.; Elkins, D. Chem. Rev. 1971,71,525.
.
Received for review May 5,1984. Revised manuscript received October 4, 1984. Accepted March 25, 1985. This work was supported, in part, by the Cooperative Research Program between the Great Lakes Environmental Research Laboratory of the National Oceanic and Atmospheric Administration, Department of Commerce, and The University of Michigan.
Trace Element Concentration as a Function of Particle Size in Fly Ash from a Pulverized Coal Utility Boiler Gregory R. Markowski” 2009 N. Madison Avenue, Altadena, California 91001
Roy Fiiby Nuclear Radiation Center, Washington State University, Pullman, Washington 99 164
Total and elemental mass size distributions from about 10 to 0.04 pm were obtained for the fly ash aerosol at the outlet of a pulverized coal utility boiler and upstream of the particulate control device. Elemental data were obtained by instrumental neutron activation analysis (INAA) and X-ray fluorescence analysis of cascade impactor samples. Submicrometer measurements were also made with an electrical aerosol analyzer (EAA). A submicrometer mode was evident in both the impactor and EAA size distributions. This fume was likely formed through vaporization of fly ash components during combustion. INAA results showed that the fly ash between about 0.5 and 8 pm was fairly uniform in composition and that the submicrometer fume composition was quite different. The fume was highly enriched in volatile elements such as Ga, As, Sb, and Se and depleted in refractory elements such as Al, Hf, and Sc. Introduction
Many studies have shown that the composition of fly ash from pulverized coal (PC) boilers is dependent on its particle size. In particular, the concentration of volatile 796
Environ. Sci. Technol., Vol. 19, No. 9, 1985
elements generally increases as the particle size decreases. In addition, there is strong evidence that, under conditions typical of PC boilers, some of the ash is vaporized during combustion and condenses to form a fume that is seen as a distinct submicrometer mode in the fly ash mass distribution. The peak of this mode typically occurs between 0.04 and 0.15 pm in diameter. Both the concentration dependence and the submicrometer mode have been seen in the field and in the laboratory (1-8). Laboratory studies (5, 6, 8) have indicated that the submicrometer fume has a very different composition from the larger fly ash particles, which would be expected from its likely formation mechanism. However, limited field data (7,9,10) have been published on the submicrometer fume composition from utility scale PC boilers because of the limited availability of field instruments capable of sizing the particles, the expense and difficulty of field measurements, and the difficulty of multielement analysis on the small spatially heterogeneous samples usually collected. These studies do give evidence that the enrichment trends seen in the laboratory continue into the size region of the submicrometer fume. However, the data
0013-936X/85/0919-0796$01.50/0
0 1985 American Chemical Society
Table I. Parameters for Last Five Stages of UW Mark 10 Impactor
stage
DM,Opm
pressure ratiob
hole dia, DH, mm
no. of holes
LIDhd
SIDn‘
23 24 25 26 27
0.32 0.24 0.11 0.053 0.037
0.980 0.965 0.881 0.592 0.437c
0.267 0.267 0.267 0.267 0.267
90 60 24 15 24
3.0 2.6 2.5 2.5 2.4
3.8 3.3 3.4 3.7 3.7
stage loading! pg of Fe 6.9 2.2 1.5 10.0
11.0
Actual diameter. Ratio of pressure after jet plate to pressure at impactor inlet nozzle (see text). Measured during sampling. Ratio of hole length (jet plate thickness) to hole diameter. ‘Ratio of spacing between jet plate and collection surface to the hole diameter. ’Here we show the amount of Fe measured on each stage as an indication of the relative amount of mass collected on each stage.
from one of the studies (7, IO),at the Kramer Station, may represent a lower limit of some sort since this boiler had the smallest submicrometer fume mass concentration of the six boilers on which our group made total submicrometer fume mass measurements (7). This paper presents the results of a study similar to but in more detail than that made at the Kramer Station. The work presented here was done at a 113 MW roof-fired boiler burning low sulfur western subbituminous coal. This boiler had the highest submicrometer fume mass concentration of the group of six boilers previously studied (7). The fly ash was collected with cascade impactors, one of which nominally gave size segregation down to about 0.04 pm actual diameter. The size-segregated samples were analyzed for about 35 elements by a combination instrumental neutron activation analysis (INAA) and X-ray fluorescence (XRF). The results showed (1) that fly ash size fractions between about 0.5 and 8 pm were fairly uniform in composition and (2) that the composition of the submicrometer fume was very different from that of the larger fly ash fractions. In particular, the fume was highly enriched, relative to Fe, in volatile elements such as Ga, As, Sb, and Se and depleted in refractory elements such as Al, Hf, and Sc. Experimental Section
Particle Sampling and Particle Size Determination. We used a version of the University of Washington Mark 10 low-pressure impactor and MRI Model 1502 impactors to collect, in situ, size-segregated samples of fly ash just after the air preheater of the subject roof-fired boiler and upstream of the particulate control device. The boiler was burning 0.5% (dry) sulfur western subbituminous coal with a heating value (dry) of 2.86 X lo7 J/kg (12300 Btu/lb). The Mark 10 impactor was made by Pollution Control Systems and had 27 stages and a final filter. The stage jet plate design was multiple straight holes in a flat plate. The last three stages were operated substantially below the impactor inlet pressure. Pressure reductions resulted from the cummulative pressure drops across the preceding stages; no separate pressure reduction orifice was used. The impactor flow rate was adjusted so that the pressure after the last impaction stage (and before the final filter) was 0.437 times the duct pressure. In situ gas flow rate and temperature were 7 L/min (0.25 cfm) and 118 OC (245 OF), respectively. At these operating parameters, the Dm range of the impactor stages was from about 13 to 0.04 pm. The D m ratio between the first consecutive 22 stages (from about 12.9 to 0.56 pm) was typically about 1.2; however, the spacing of the last five stages was much more variable. The D,, and associated parameters of the last five stages are shown in Table I. The pressure ratios in Table I are the ratio of pressure after the stage jet plate to the impactor inlet (duct) pressure. They were measured at about 20 OC, except for the pressure after the last stage
which was measured during the actual sampling at the in situ conditions. Pressure ratios for stages 24-26 were also calculated by assuming that the absolute pressure drops across the jet plates were directly proportional to the product of the square of the stage jet velocity and the gas density after the stage jet plate. The calculated pressure ratios agreed within 1.5% of the measured ratios. The DMvalues were computed by using a Stokes number corresponding to a 50% collection efficiency calculated by Marple (11) of 0.47. Marple’s theoretical calculations appear to be in good agreement with those determined experimentally (12). The Cunningham Correction Factor used in the DN calculation was based on the pressure above the stage jet plate as indicated by Flagan (13). Deposits on about every second stage were examined by using a scanning electron microscope (SEM) to check for collection of particles that bounced off or were reentrained from the preceding stages. Particle size distributions were also measured with a TSI Model 3030 electrical aerosol analyzer (EAA) and dilution system and Meterology Research Inc. (MRI) impactors. The raw EAA data were reduced by using the Twomey routine (14) to correct for the nonideal size selectivity of. the EAA. The EAA sampling procedures are discussed in more detail in ref 7. The MRI impactors were usually run at low sample flow rates (and near in situ pressure) to reduce the effects of particle blowoff and reentrainment (15). The last stage D50)s were typically 0.3 pm. EAA measurements were made continuously during the UW and MRI impactor runs. The MRI impactor runs were made 1-3 h before the UW run under the same boiler conditions. A density of 2.4 g/cm3 was used to convert volume to mass (EAA data) and aerodynamic diameter to actual diameter (impactor data). The UW and MRI impactor fly ash collection surfaces were greased with Apiezon L and H greases, respectively. The fly ash collection surface in the UW impactor was a thin (0.5 mil) peelable Kapton film which was attached to the stainless steel collection disk with Apiezon L grease. The UW and MRI backup filters were Teflon membrane and glass fiber, respectively. Two types of impactor blanks were used in this study. The first type, referred to simply as “blanks” hereafter, was obtained by sampling filtered flue gas with an impactor. Except for the filter preceding the impactor, operation during the blank run was the same as during a particulate sampling run. The second type of blank, referred to hereafter as a “control”, was collection disks prepared identically with those used in the particulate sampling and blank runs and taken to and from the field site with them. However, they were never removed from their Petri dishes. Petri dishes were used for labeling and protection of all collection disks during transportation and handling. Elemental Analysis: Samples and Standard Reference Materials. A composite coal sample taken at the Environ. Sci. Technoi., Vol. 19, No. 9, 1985
797
beginning of the impactor sampling period from the four coal feeders in use was analyzed by spark source spectroscopy and INAA. Several replicate samples, some taken on other days during the field work, and samples of National Bureau of Standards (NBS) Standard Reference Material (SRM) 1632a, bituminous coal, were also analyzed. UW impactor samples, blanks, and controls were analyzed by INAA for up to 30 elements. The samples were prepared for analysis by spray coating the samples and Kapton film with a thin layer of plastic fixer to prevent any sample loss, separating the Kapton film from the collection disk, and folding and sealing the film in a small polyethylene bag. Six fly ash samples, two blanks, and one control were cut into three or four pieces to check uniformity. Samples were analyzed at the Nuclear Radiation Center, Washington State University, by using INAA methods described previously (16,IO). The samples from stages 23 through 27 and the blanks and controls were analyzed a second time because of the larger relative interference from the low concentrations of elements in the polyethylene bags in which the samples were originally sealed and analyzed. During the reanalysis, the Kapton films were removed from the original irradiated bag to eliminate its interference during counting. Because of the brittleness of the samples after the two 16-h irradiation periods, the samples could not be reanalyzed for the short half-lived ellements, Ti, V, Ca, Al, and Mn. Stages 1-22 had large enough deposits so that interference from bag contamination was usually unimportant. In cases where the Dm’s of two consecutive stages were close, the samples from these stages were combined and analyzed as one, so that the equivalent of 21 different size classes were finally analyzed. We analyzed several NBS SRMs with the impactor samples analyzed by INAA to verify the accuracy of the INAA results. Bulk check samples, typically 20-30 mg, included SRM 1633 coal fly ash, SRM orchard leaves, and SRM 1632 bituminous coal. To approximate the actual size of the impactor samples, we analyzed six small check samples of SRM 1633a fly ash from 0.37 to 5 mg in weight. Four of these samples were loose powder enclosed in small polyethylene bags, and two samples were small piles on Kapton film that were similar in form to actual impactor samples. These piles were made by dabbing a slurry of the SRM 1633a fly ash, Apiezon L grease, and toluene onto the Kapton films and evaporating the toluene at about 50 “C. The ratio of fly ash to Apiezon L was about 2:l by weight. The submicrometer fume collected on the backup filter from two MRI impactor runs was analyzed by XRF (by NEA, Inc., Beaverton, OR) to determine the ratios to iron of the major matrix elements not measured by the INAA analysis. These data were also needed to calculate the fume’s total mass. Total fume mass was calculated by summing the contribution of the matrix elements measured (as their oxides) and scaling this sum by the total Fe found in the UW fume samples. Both A1 and Fe were determined rather precisely by both the XRF and INAA analyses. The A1 to Fe ratios calculated by the two methods agreed within 12%. Quality control for the XRF analysis included analyzing a check standard normally included with all batches of fine particle samples likely to contain crustal type material and additional thin film standards for Mg and Al. The check standard contained the elements Fe, Si, Ti, and Pb. All standard results were within 4 % of their expected values. NBS fly ash samples prepared as the fume samples were not used because of 798
Envlron. Sci. Technol., Vol. 19, No. 9, 1985
the difficulty in making uniform samples and determining XRF self-absorption corrections for the lighter elements. We prepared the fume samples by sonicating parts of the fiberglass backup filters in toluene and filtering the resulting suspensions through an 8-pm (pore size) Nuclepore filter before collecting the fine particles on a 0.1-pm Nuclepore filter. The 8-pm filter was needed to remove glass fibers that were resuspended along with the fume particles. Two MRI impactor blank filters were also processed identically to get fiberglass blanks that reflected the expected residual fiber contamination and its composition on the final particulate filter. After XRF analysis, the filters with fume particles and the highest blank concentration analyzed were examined on a scanning electron microscope to determine the correction for oversized fly ash particles and glass fibers on the fume samples. Their amounts were determined by measuring and counting the particles and fibers at several locations on each filter and computing the mass per unit area on the filter. Agreement between the amount of material on the fiberglass blank determined by fiber counting and XRF was good, about 15%. Corrections for fiber on the fume samples were made proportional to the fiber volume seen on these samples compared to the amount on the highest filter blank and its elemental analysis since the fiber volume on the lighter and most heavily loaded fume samples was 2.4 and 1.0 times the highest blank, respectively. The XRF data presented in this paper are based on the most heavily loaded fume sample since the fiber blank correction for any element in this sample was 15% at most, while the relative corrections for Si and A1 to the other fume sample were 4 times as large. Corrections for the larger fly ash on the fume material were made by using the elemental composition of the large fly ash mode particles. Nearly all the mass was in particles less than 2 pm in diameter, and the INAA data indicated the matrix element composition of this material was close to that of the larger particles. Corrections for the larger fly ash particles on the most heavily loaded fume sample were 8% or less for all elements. Data Reduction and Analysis: Size Distributions and Enrichment Factors. Diferential mass distributions as a function of particle diameter for total mass and for each element were calculated from the raw impactor data by fitting the cumulative distributions (as a function of the log of the diameter) with overlapping parabolas and taking the average slope over a diameter ratio of 2. This procedure is described in detail by Markowski and Ensor (17). Elemental data were corrected for blank and polyethylene bag contributions. EAA data taken under similar boiler operating conditions were used to estimate the width of the submicrometer distribution because examination of the last two UW impactor stages indicated that they were overloaded. Thus, they gave little information on the width of the submicrometer fume distribution. The D60)s of the last two stages were adjusted, and a Dso was assigned to the filter so that the Fe distribution (in the submicrometer fume) resulting from the data reduction procedure described above was approximately log normal with a geometric standard deviation of 1.4 and a peak at 0.054 pm. The peak size was chosen to be the largest consistent with the very light loading on the third to last impactor stage and the geometric standard deviation of 1.4. Assuming a geometric standard deviation of 1.4, the amount of material on the filter indicated that the peak of the distribution was larger than 0.045 pm.
Changes in elemental composition as a function of particle size, typically referred to as enrichment factors, were calculated directly from the elemental mass on each stage, thus, without reference to the size distributions. Because the total mass data were poor in the size range of the submicrometer fume, we used iron as the normalizing element since in these data and other studies (8) it followed the total mass in the submicrometer fume and our accuracy for iron was good. Choosing an element instead, of total mass may add some arbitrariness to the enrichment factors calculated if the chosen element changes its concentration with respect to total mass. (In addition, the assumption that the matrix fly ash elements such as Si, Al, and Fe are similarily distributed in the submicrometer fume and the larger fly ash is not likely to be true.) Our enrichment factors were calculated for a given element and stage as a double ratio: the mass ratio of the element to iron on the given stage divided by the same ratio for stage 6 (whose D50 was 8.3 pm). This method gives an enrichment factor of 1 for all elements on stage 6. An enrichment factor greater than 1 indicates enrichment relative to iron compared to stage 6, while an enrichment factor less than 1 indicates depletion.
Results Scanning electron microscopy (SEM) of the UW impactor stages indicated reasonable size segregation, although the distribution of particles on the larger particle stages was much broader than the ratio of 1.2 typical between the 0,’s of those stages. Stages with a D50 larger than about 0.7 pm showed about 80% of the particle mass between the D50 and twice the D50, with the distribution dropping sharply below the D, and above 2D50. Stages with Ow’s from about 0.7-0.10 pm showed somewhat poorer size segregation with about 60-70% of the particle mass between D50 and 2 0 5 0 . These latter stages showed some particles ca. 1 pm in diameter that had evidently bounced down from the preceding stages. On the final two stages and filter, about 20-40% of the particle mass visible appeared in the range of 0.3-2 pm, with the rest smaller in size. The mass distribution estimates, however, are only semiquantitative since they were made by visually examining the SEM pictures. The composition of the very small particles (less than 0.1 pm) determined by INAA indicated that collection of particles larger than 0.5 pm on the last two stages and filter was less than 20%. Visual analysis of the SEM pictures was somewhat uncertain because, generally, only the upper surfaces of the impactor deposit (i.e., those deposited last) were visible. Thus, a higher fraction of oversized particles would be seen compared to the sample deposited at the beginning of the run, which would be expected to contain fewer oversized particles. The analytical results for the NBS SRM check samples for the elements for which we show size distribution data were all within 25% of the NBS values (both certified and uncertified) for the check samples 0.87 mg and larger in weight, except for Al, Mn, Se, and Ga. A1 and Mn could not be measured in the 0.87-mg NBS standard because of Al in the polyethylene bag material and high blank values, respectively. Deviation from the NBS values for Se was from +1 to +71%. Ga was low by 20 f 5% in the first analysis. Check samples and impactor samples analyzed a second time showed about a 20% increase in Ga. Therefore, we increased Ga values from stages 1-22 by this factor. For all elements except those already noted, typical agreement with NBS values was about 12%. Results for the smallest check sample, 0.37 mg, were uniformly low by a factor of about 1.75 for all elements.
Table 11. Spark Source Spectroscopy Analysis of Ash from Bulk Coal Sample
compound
% composition
compound
% composition
1.7
SiOz
51
MgO
A1203
22
PpS TlOz K2O NazO total
Fe203
8.5
CaO
8.2
so3
8.1
1.2 0.7 0.55
0.20 102
Because of this uniformity, we believe that this result was due to SRM inhomgeneity or, more likely, an error in sample preparation. Reexamination of the check sample preparation notes gave some evidence for the latter cause. These results for small mass SRM samples are of interest in microanalysis and are presented in detail by Filby et al. (18). Table I1 shows the results of the spark source spectroscopic analysis of ash in the coal feeder sample (taken about 2.5 h before the UW run and about coincidently with the MRI impactor run for which we show fume elemental data). For several elements, the results were only semiquantitative. On the basis of the replicates and NBS coal analyzed, the relative accuracy (one standard deviation, four samples) was 10% for Si, about 15% for Ca, Fe, and K, about 20% for Al and Na, about 35% for Mg, and about 45 percent for Ti. NBS standard data were not available for P and S;precision was 15% or better. In this paper, we generally show size distribution data only for those elements for which reliable data were obtained for the submicrometer fume, Le., that material which resided nearly completely on the last UW two stages (stage 26 and 27) and filter. Size separation from the larger fly ash mode was rather good: stage 25 held only about l/, as much material as stages 26,27, and the filter. The fume material was nearly equalIy divided between stage 26, stage 27, and filter. The main causes of unreliable data were the very small amount of mass of many elements on the last four stages and filter (particularily stages 24 and 25), which were very lightly loaded (see Table I), and contamination of the Kapton from the stainless steel disk to which the Kapton was attached. The latter problem was severe for Cr, Ni, and Co. Elemental concentrations with relative standard deviations due to y or X-ray counting statistics greater than 33% and with relative standard deviations greater than about 125% including all sources of error considered have been omitted, shown as less than values, or otherwise flagged. We saw no differences between the blanks and controls for the data we present. Figures 1-4 show elemental mass distributions calculated from the INAA analysis of the UW impactor samples. Figure 1also includes total mass distributions measured by the EAA and MRI impactors. Mass distribution data from the UW run (not shown) were poor below 1pm due to unpredictable weight changes of the Kapton films; these changes could not be reliably corrected with the results from the blanks or controls. (A later study showed that the problem was caused by the hygroscopic nature of the Kapton film.) The MRI impactor data shown are the average of two impactor runs made about 1 h before the UW run under the same boiler operating conditions. The EAA size distribution was measured nearly simultaneously with the UW run. Note that the procedure used to calculate the elemental size distributions (described under Experimental Section) included all the elemental masses on the impactor filter in the submicrometer mode so that integrating the distributions in Figures 1-4 over the size Environ. Scl. Technol., Vol. 19, No. 9, 1985
799
Table 111. Total Element Cpncentrations and Percentages in Submicrometer Fume
element, mass units
total concn, mass/dscm" 340 71 5.5 110 400 11
150 93 35 22 220 3lbgC 960 5.5
f f
7.3 39 31 710' 29 5.5 6.4 3.3 69 21 270 3060
total concn in particles
