Table I. Percent Recovery of Vapors Adsorbed on Tenax GC Cartridges Using Thermal Desorption. Quantity adsorbed, ngb
50
Compound
100
N.nitrosodiethylaminec p-Propiolactone Ethyl rnethanesulfonate Nitromethane Glycidalde hyde Butadiene diepoxide Styrene epoxide Aniline Bis.(chloromethyl)ether
100 100 105 100 105 100 100 100 100 100 100 100 95 95 100 100 Bis-(2-chloroethy1)ether 95 90 Tenax G C cartridge, 10.5 mm i.d. x 30 mm in
200
300
-
-
100 95 70 95 100 105 95 100 90
100 100 70 80 100 90 60 90
-
length. Synthetic airvapor mixtures were introduced onto a Tenax GC bed a t 4 Ifmin. Desorption U n i t was a t 225°C. Represents theoretical a m o u n t in synthetic air-vapor mixture. CValues an average of duplicate r u n s were calculated on basis of a ratio of pbak areas for calibration mixtu;e and f r o m thermal desorption.
*
Acknowledgments The authors are grateful to L. Retzlaff for his assistance in the machining and construction of experimental devices used in this study, R. L. Marquard and C. Cleary for the design and assembly of the temperature controller, and M . E. Wall for his interest in this program. Literature Cited (1) Pellizzari, E . D., Carpenter, B., Bunch, J., Sawicki, E., Enuiron. sei. Tech., 9,552 (1975). (2) Williams. I. H.. Anal. Chem. 37. 1724 (1965). (3) Leggett, D . C., Murrmann, RT P., Jenkins, T . J., Barriera, R., CRREL, SR176, l(1972).
(4) Williams, F . W., Umstead, M. E., Anal. Chem., 40, 2232 (1969). ( 5 ) Krumperman, P. H., J . Agr. Food Chem., 20,909 (1972). (6) Hollis, 0. L., Anal. Chem., 38, 309 (1966). (7) Zlatkis, A,, Bertsch, W., Lichenstenstein, H . A., Tishbee, A., Shumbo, F., Liebich, H. M., Coscia, A. M., Fleischer, N., ibid., 45,763 (1973). (8) Zlatkis, A,, Lichenstein, H . A,, Tishbee, A , , Chromatographia, 6,67 (1973). (9) Bertsch. W.. Chane. R. C.. Zlatkis. A.. J . Chromatoe. Sei.. (1974). (11) Saalfeld, F. E., Williams, F. W., Saunders, R. A., American Lab., 3 , 8 (1971). (12) Saunders, R . A,, Umstead, M . E . Smith, W. D., Gammon, R. H.. “The atmosDheric trace contaminant Dattern of SEALABII,” ‘Proc. 3rd Ahn. Conf. Atmos. Contahination Confined Spaces, AMRL-TR-67-200, (1967). (13) Duel. C. L.. et al.. “Collection and Measurement of Atmospheric Trace’ Contaminants,” Aerojet Electrosystems Co., Azusa, Calif., Final Report, Contract NAS 1-9814, NASA Doc. No. 71-19636. (14) Saunders, R. A,, “Analysis of Spacecraft Atmospheres,” NRL Rept ,5316 (1962). (15) Turk, A,. Morrow, J . I.. Stoldt. S. H., Baecht, W., J . Air Pollut. Contr. Assoc., 16, 383 (1966). (16) Turk, A,, Morrow, J . I., Kaplan, B. E . , Anal. Chem., 34, 561 (1962). (17) Chiantella, A. J., Smith, W. D., Umstead, M . E., Johnson, J . E., “Aromatic Hydrocarbons in Nuclear Submarine Atmospheres,” A m . Ind. Hyg. Assoc. J., 27, 196 (1966). (18) Saunders. R. A.. “AtmosDheric Contamination in SEA-LAB I,” Proc. Conf. Atmos. Contamination Confined Spaces, AMRL-TR-65-230 (1965). (19) Grob. K.. Grob. G.. J . Chromatop.. 62. l(1971). (20) Jennings, G., Nursten, H . E., A&/. Chem., 39,521 (1967). (21) Herbolsheimer, R., Funk, L., Drasche, H., Staub-Reinhalt. Luft., 32,31 (1972).
Received for review Juls 11, 1974. Accepted Dec 2, 1974. Work supported by EPA Contract Rio. 68-02-1228 from the Enuironmental Protection Agency, Health, Education, and Welfare.
Aerosol-Size Spectra by Means of Nuclepore Filters Qctavio T. Melo and Colin R. Phillips* Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto M5S 1A4, Canada
The size and chemical composition of aerosols are of great importance in air pollution, industtial hygiene, atmospheric physics, and other fields. The health hazard posed by a particulate pollutant depends on the size and chemical composition of the particles involved while various atmospheric phenomena, such as condensation and photochemical smog production, involve or lead to particles of various sizes and chemical composition. While sampling and size distribution determination of aerosols above 1 pm are straightforward tasks ( I ) , the measurement of the size spectrum of submicron aerosols, as well as the chemical characterization of aerosols of all sizes, is still a very active area of research. For a long time the electron microscope was the only means of obtaining information in the submicron range through direct counting. More recent developments include diffusion batteries (21, electronic particle counters ( 3 ) , and the scintillation spectrometer ( 4 ) that have made possible the size characterization of aerosols down to about 0.05 pm. Recently Spurny et al. ( 5 ) noticed the possibility of using membrane filters to build an inexpensive and fully portable aerosol sampler. Nuclepore filters were used 560
Environmental Science 8 Technology
rather than the more traditional membrane filters for reasons that include the reduced thickness of the Nuclepore filters (about 10 micrometers) and the uniformity of shape (circular), size (geometric standard deviation 51.10), and orientation (normal to the surface) of the pores. These structural properties ensure the existence of pronounced minima in the efficiency characteristics, and this allows the use of these filters in the size classification of aerosols. The work of these authors has been expanded upon to deduce a size spectrum from filter loadings (6). The present study is an experimental application of that theory both to aerosols consisting of a single substance and to aerosols consisting of two chemical substances. Experimental Aerosol Generation. Test aerosols were generated from dye solutions with a Collison atomizer built according to the specifications of the British Standards Institution, 1955. The coarse spray from the atomizer was removed with an impactor similar to that described by Whitby et al. (7). A separation of 2.54 mm was used between the ori-
The experimental determination of the size and chemical composition of submicron aerosols by means of membrane filters is considered. The approach taken is to utilize the retention characteristics of Nuclepore filters and to measure filter loadings. Conversion of the experimental filter loadings to a size distribution is achieved by the method of Melo and Phillips. Experiments were conduct-
A
Rotameter Filters
\
Aerosol Holding Chamber
In
Mercury Manometer Figure 1. Schematic diagram of
the membrane filter sampler
Vo, V, = needle valves; a , b , c = filter holders; and S = three-way stopcock
fice plate and the impaction plate. Aerosol generation was performed a t 30 psi to the nozzle since this pressure leads to a minimum mean particle size for the atomizer-impactor combination (7). At this pressure the gas flow was 6 Ipm . The stream from the impactor was then mixed with a dilution gas stream in a 110-cm long by 7.5-cm diameter cylindrical container. This secondary stream was used to aid in the evaporation of the aerosol droplets and to dilute the final aerosol to acceptable concentrations. Typical flow rates were 12 lpm leading to a total flow of 18 lpm. From the cylindrical container the aerosol was stored in a 15-liter hold-up tank fitted with a mercury manometer that allowed measurements of the pressure drop across the filters. Laboratory grade uranine (sodium fluorescein) and medicinal grade methylene blue, both supplied by Fisher Scientific, were used to generate the aerosols. Solutions of 5, 1, 0.5, 0.1, and 0.05% dye in methanol were prepared for this purpose. Dry nitrogen was used to generate the aerosols. This prevented ignition of the methanol vapors with sparks from the electrostatic precipitator. Aerosol Sampling. Figure 1 shows a schematic diagram of the experimental setup used to collect aerosol samples by means of Nuclepore filters. Sampling could also be carried out by using an electrostatic precipitator in place of the filters. Sampling by both means was done a t the same point so that losses in the lines would not affect the final comparison. The point-to-point electrostatic precipitator of Billings and Silverman (8) was used. The sampling rate was set a t 6 lpm and the electrodes were activated at 10.5 kV. Sampling times varied from 1-6 min, depending on aerosol concentration. Collection was on 200-mesh copper specimen grids of standard 3-mm size precoated with a carbon film of about 100 A thickness. The Nuclepore filter sampler allowed the use of 1, 2, or 3 filters in series sampling a single aerosol stream. The
ed with both homogeneous and heterogeneous aerosols generated from solution by atomization. The spectra calculated from the experimental loadings were compared with those determined from electron microscope counts of the test aerosol. Good agreement was observed for aerosols of mass median diameter between 0.1 and 1.0 pm, and good repeatability was found.
collection time varied from 30 sec to about 6 min, depending on the concentration and particle size of the aerosol as well as on the sampling velocity. At the end of the experiment, the aerosol particles were removed from the filters by washing with a known volume of distilled water, diluted, and analyzed for dye content. Analysis. The filter loading analysis was fluorimetric for uranine and spectrophotometric for methylene blue. For uranine the p H was controlled a t 6.5, the exciting radiation was the 405 and 436 nm mercury lines and the 515-nm fluorescence was measured for 15" in the backward direction from the incident light. For methylene blue, the absorption peak a t 665 nm was measured. Calibration for the interference of uranine on the methylene blue determination and vice versa were determined (9). No corrections were needed in the range of concentrations used. Viewing and photographing of the aerosol samples collected by the electrostatic precipitator were performed on the unshadowed specimens using a Philips EM 300 instrument a t a filament voltage of 60 kV. Particle counting and sizing were done from 4 x 4 in. glass negatives using an electronic graph digitizer. Most aerosol particles were spherical, but an average size based on two perpendicular diameters was calculated. From 500-1000 particles were sized. Typical mean diameters of test aerosols as a function of solution concentration are reported in Table I. Filter parameters such as mean pore size and porosity were obtained from scanning electron microscope photographs of the filters. The results are shown in Table 11. Filter thickness was not measured, but reported values were used in the evaluation of the size spectra.
Table I. Observed Mean Diameters (pm) of Aerosols Generated from Uranine Solutions Calculations used scanning electron microscope photographs of electrostatically collected samples
2
Uranine d e concn,
Arithmetic mean diam
Mass mean diam
0.05 0.10 0.5 1.0 5.0
0.14 0.14 0.18 0.27 0.39
0.18 0.18 0.25 0.38 0.68
Table II. Comparison of Measured Mean Dtameters and Porosities with Manufacturer's Supplied Values for Nuclepore Filters Mean diam,
m
0.4
0.6 1.0 3.0 8.0
Measured mean diam,
m
Porosity
Measured porosity
0.32 0.54 1.15 2.19 7.42
0.10 0.07 0.13 0.15 0.05
0.13 0.08 0.18 0.11 0.09
Volume 9, Number 6, June 1975
561
Evaluation of Spectra Evaluation of the experimental observations was done in a manner similar to the simulations of Melo and Phillips (6),except that the filter loadings and sampling rates were now available from the experiments, and the experimental error was inherent in the observations. The equation of Spumy and Madelaine (10) was used to describe the efficiency characteristics of the filters. A first set of experiments vias done with four filter arrangements, each consisting of 10 observations, and five aerosols generated from solutions of pure uranine. The results were calculated as filter loadings, sampling matrix A and deduced size spectra, using four constrained methods. Methods 1 and 2 are those used previously (6),method 3 makes use of the third differences of the g L as a measure of smoothness and method 4 uses the variance of the g, as the smoothness criterion. Table I11 shows, as an example, the results obtained in an experiment with a uranine aerosol generated from a 0.1% solution. Filter loadings are in terms of mass (grams) of particulate phase collected per volume (ml) of aerosol sampled. The negative values are an artifice of the calculation procedure a t spectral ordinates close to zero. Examination of the complete set of results of these experiments failed to indicate the superiority of any one fil-
ter arrangement (Table IIIa) over the remaining. This is also the conclusion reached from calculating the eigenvalues of the matrix A. As discussed by Twomey ( I I ) , a good sampling matrix is one with large eigenvalues since the number of independent points a t which the spectrum may be evaluated equals the number of eigenvalues, A,, satisfying:
5 €j2 c j=f
Xi>> 7 gj j=l
1.0 w
0.9 -
N
0.8 -
O
w
2 0.7 im
z 0.6 -
6 I I-
E W
0.5-
Table 111. Observations and Deduced Spectra for Experiment with 0.1% Uranine Solution (a) Observations Sam-
Filter no.
Pore size, r m
1 2 3 4 5 6 7 8 9 10
8 3 8 3 3 1 3
Sampling velocity, cm sec-1
pling time, sec
Loading, ng cm-3
18.8 18.8 3.1
60 60 90 90 60 60 60 60 45 45
0.103 0.340 0.238 0.733 0.407 1.900 0.614 1.310 1.840 2.740
1
3.1 16.5 16.5 3.5 3.5
1 0.4
10.2 7.3
Remarks
land2
in series 3and 4 in series 5 and 6 i n series 7 and 8
in series
i
0 -15
/"
O l'0,,
-125
.
f
-10
I\
I
X
-075
-05
I
1
-025
Figure 2. Comparison of the size spectra of the aerosol generated from 0 05% uranine solution A , weight spectrum by electron microscopy A weight spectrum by the membrane filter sampler and 0, n u m b e r spectrum
(b) Deduced Spectra, ng c r r 2 2, -,Method = 0.0032
Method 1, = 0.0003
X
0.0010
2.61 2.17 2.14 3.73 6.14 7.85 3.16 -2.24 -2.63 -0.07
-0.447 -0.925 -1.07 3.83 10.2 10.2 2.54 -3.29 -2.78 -1.37
-3.29 -2.17 0.177 4.46 8.84 9.82 5.18 -1.15 -5.25 -7.86
-1.09 -1.03 -0.94 -0.82 -0.67 -0.53 -0.38 -0.26 -0.17 -0.11
Method 3,
y =
Method 4, = 0.0010
*/
-7.47 -2.94 1.62 5.79 8.51 8.51 5.43 0.405 -5.17 -10.8
Table IV. Magnitudes of Largest Seven Eigenvalues of A Sampling arrangement
562
1
2
3
4
0.3928 0.0458 0.0065 0.0019 0.0017 0 * 0002 0.0002
0.3269 0.0327 0.0145 0.0029 0.0029 0.0007 0.0007
0.3399 0.0149 0.0149 0.0103 0.0014 0.0014 0.0014
0.3046 0.0149 0.0149 0.0145 0.0034 0.0006 0.0002
Environmental Science 8 Technology
I
-15
-1.25
-1.0
-075
-0.5
-0.25
X
Figure 3. Comparison of the size spectra of the aerosol generate d from 0.1% uranine solution A , weight spectrum by eiectron microscopy: A , weight spectrum by t h e membrane filter sampler: and. 0 , n u m b e r spectrum
As an example, the magnitudes of the largest eigenvalues of the matrix resulting from the experiments with the 0.1% uranine dye are shown in Table IV. The four sets are comparable, and the above criterion is satisfied by 4 or 5 eigenvalues if the experimental error is about 5%. This, then, is the maximum number of independent evaluations of g that should be expected. In view of the comparable merits of the four methods of solution and the four filter arrangements, it was decided
" -15
-1.25
-1.0
-0.5
-075 X
0
-0.25
to average all results obtained under similar conditions and plot the results as cumulative spectra for comparison with the results of the electron microscope size analysis. Here a particle density of 1.4 g/cm3 was used (12). The results are shown in Figures 2 to 6. The agreement is excellent in all cases except for the aerosol generated from the 0.05% solution. The reason for this is not certain but may be due to difficulties in obtaining a representative sample for microscopy despite the correction of Robinson
-125
Figure 4. Comparison of the size spectra of the aerosol generated from 0.5% uranine solution A , weight spectrum by electron microscopy: A , weight spectrum by the membrane filter sampler: and 0, number spectrum
-10
-075
-05
X
-025
0
025
05
Figure 6. Comparison of the size spectra of the aerosol generated from 5 0% uranine solution A , weight spectrum by electron microscopy A weight spectrum by the membrane filter sampler and 0, number spectrum
10
,
I
0'
-1.5
-1.25
-1.0
-075 -0.5
I
I
-0.25
-0
I
a25
X Figure 5. Comparison of the size spectra of the aerosol generated from 1 .O% uranine solution A , weight spectrum by electron microscopy, A , weight spectrum by the membrane filter sampler, and 0, number spectrum
X Figure 7. Weight and chemical composition spectra of the aerosol generated from a solution of 0.05% uranine and 0.25% methylene blue Volume 9, Number 6 , June 1975
563
(13) that was applied to the microscope counts to account for differences in drift velocity. A second set of experiments was performed with aerosols generated from a binary solution of uranine and methylene blue. Results for the case of a 0.05% uranine and 0.25% methylene blue aerosol are shown in Figure 7, where three spectra are included. Each spectrum gives the mass of particulate phase in a cubic centimeter of aerosol associated with particles in a given size range. This means that all three spectra should be similar since all particles are of identical chemical composition. This is, in fact, borne out in spite of the fact that methylene blue and uranine were analyzed separately and the respective spectra deduced from the results. Conclusions and Recommendations The experiments carried out here confirm the simulation study of Melo and Phillips, and indicate the feasibility of obtaining particle size and chemical composition spectra with a Nuclepore filter sampler. The agreement between the spectra obtained with the Nuclepore sampler and results obtained by electron microscopy was good. Very little difference was encountered between the four solution methods and between the four filter arrangements. The experimental error was estimated a t 6.770, and a value of y of 0.003 was often found adequate to solve the constrained equation for the size spectrum. The possibility of using the sampler for chemical composition spectra has been demonstrated. In actual practice, however, this analysis depends on the ability to collect enough sample for chemical composition determination during the period when the efficiency of filtration is constant. This requirement will constrain the choice of chemical species for analysis and of the analytical method. In any case the size spectrum should be obtainable in all cases either by measuring filter loading as was done here or by measuring particle counts before and after passage of the aerosol through the filters. This last aspect is now undergoing investigation.
Nomenclature
A = sampling matrix g = discrete size spectrum vector g j = components of vector g giving the value of the spectrum at x, x = particle size parameter (x = log D, where
D is particle diameter) c j = error associated with the j t h observation X i = ith eigenvalue of A y = Lagrange multiplier involved in the constrained solution of the sampling equation
Literature Cited (1) Silverman, L., Billings, C. E., First, M . W., “Particle Size Analysis in Industrial Hygiene,” Academic Press, New York, N.Y., 1971. (2) Fuchs, N. A,, “The Mechanics of Aerosols” (Rev. ed.), Pergamon Press, Oxford, 1964. (3) Whitby, K . T., Clark, W. E., “Tellus XVIII,” 573, 1966. (4) Binek, B., Dohnalova, B., Przyborowski, S., Ullmann, W., Staub, 27,379 (1967). (5) Spurny, K . R., Lodge, J . P., Frank, E . R., Sheesley, D. C., Enuiron. Sci. Technol., 3,453 and 464 (1969). (6) Melo. 0. T.. Phillius. C. R.. ibid.. 8. 67 (1974). . . (7) Whitby, K.’T., L;ndgren,’D. A’., Peterson, C. M., Znt. J . Air Water Pollut., 9,263 (1965). (8) Billings, C. E., Silverman, L., J . Air Pollut. Contr. Assoc., 12, 586 (1962). ~~, ( 9 ) Skoog, D. A,, West, D. M., “Fundamentals of Analytic Chemistry,” Holt, Rinehart and Winston, New York, N.Y., 1963. (10) Spurny, K . R., Madelaine, G., Coll. Czech. Chem. Commun., 36,2857 (1971). (11) Twomey, S.,J. FranklinInst., 279,95 (1965). (12) Stein, F., Esmen, N., Corn, M., A m . Ind. H y g . Assoc. J., 27, 428 (1966). (13) Robinson, M . , in “Air Pollution Control Part I,” W. Strauss, E d . , Wiley-Interscience, New York, N.Y., 1971. ~
~
Received for reuieu: May 7, 1974. Accepted December 23, 1974. Work supported by Ministry for the Environment, Ontario, and the Atkinson Charitable Foundation. O.T.M . was supported by a National Research Council of Canada Scholarship.
Inertial Collection of Aerosol Particles at Circular Aperture Terence N. Smith’ and Colin R. Phillips*
Department of Chemical Engineering and Applied Chemistry, University of Toronto, Ont., M5S 1 A4, Canada Particles of sufficient mass may be separated from an aerosol by the inertial mechanism. The aerosol is subjected to accelerations and changes of direction as it flows through the geometry of the collecting system. While the finest and lightest particles respond to the motions of the fluid and follow its streamlines as they conform to the boundaries of the system, larger and more massive particles are not able to do so. The inertia of such particles causes them to cross streamlines and perhaps to impinge upon the bounding surfaces, to adhere, and so to be collected. This principle is utilized extensively in various designs of equipment for gas cleaning and aerosol sampling. A comprehensive discussion of the mechanics of the process of inertial separation is given by Fuchs ( I ) . A collecting device of some interest for sampling and, perhaps, for size analysis of aerosols is the Nuclepore 1 Present address, Department of Chemical Engineering, University of Adelaide, Adelaide 5001, Australia.
564
Environmental Science & Technology
membrane filter. This consists of a membrane typically 10 pm in thickness perforated by uniformly sized pores which are straight, parallel, and circular in section. Filters with various pore sizes ranging down to small fractions of a micrometer can be obtained. Spurny and Pich (2, 3) consider the flow of an aerosol through such a filter and identify mechanisms of inertia, interception, and diffusion in the collection of particles. Inertial collection results if a particle is too massive to respond sufficiently to the changes in direction of the fluid streamlines as they converge to enter the filter pore. This process is illustrated in Figure 1. Interception of a particle following a fluid streamline occurs if the particle touches the surface of the filter as it moves by. Recovery by this mechanism is evidently dependent upon the size of a particle. The inertial effects on some particles in an aerosol approaching the pore of a filter may not be great enough to cause them to continue to travel across streamlines and so
