Environ. Sci. Technol. 1982, 16, 492-497
slow rates of the chemical reactions under these mild biological conditions. More stringent chemical conditions would presumably enhance the rate of decomposition of nitrosoguanidine and change the number and the types of products found in this study. The demonstrated sensitivity of both nitroguanidine and nitrosoguanidine to UV light suggests this treatment as an alternative in alleviating pollution hazards associated with nitroguanidine-laden waste streams. Acknowledgments
We thank Carmine DiPietro for his MS analysis. Literature Cited Bissett, F. H.; Levasseur, L. A. Technical Report TR-76/47, U.S. Army Natick Res. Dev. Com. 1976. Small, M. J.; Rosenblatt, D. H. Technical Report 7404,U.S. Army Bioengineering Res. Dev. Lab. 1974. Kemula, W.; Kalinowski, M. K.; Kryogowski, T. M.; Lewandowski, J. A.; Walasek, A. J. Bull. Acad. Pol. Sci., Chem. Ser. 1970,18,445-461. Davis, T. L.; Rosenquist, E. N. J . Am. Chem. SOC.1937, 15, 2112-2115.
Cloak, L. H. British Patent 869 306,1961. Diebner, R. L. U S . Patent 3860594, 1975. Ames, B. N. Supplement to Method Paper, University of California, Berkeley, 1979, 1-10. Ames, B. N.; McCann, J.; Yamasaki, E. Mut. Res. 1975,31, 347-364. M h , J. E.; Janes, R. H. Anal. Biochem. 1956,28,846-849. Korn, E. D. “Methods in Enzymology IV, Purines and Pyrimidines”; Academic Press: New York, 1957;631-632. Knappe, E.; Rohdewald, I. Zeit. Anal. Chem. 1966,223, 174-181. Ishidate, M.; Odashima, S. Mut. Res. 1977,48,337-354. Sax,M. I. “Dangerous Properties of Industrial Materials”; Van Nostrand Reinhold: New York, 1975. Wvensch, A,; Amberger, A. A. 2.Pflanzenphysiol. 1974, 72,359-366. (15)Iwasaki, K.;Weaver, R. J. J. Am. SOC.Hort. Sei. 1977,102, 584-587. (16) National Cancer Institute, “Bioassayof Calcium Cyanamide for Possible Carcinogenicity”; Nat. Tech. Inf. Ser. PB293625,1979. (17) Iskandarov, T. I. Gig. Primen., Toksikol. Pestits. Klin. Otravlenii 1970,8,337-340. Received for review July 30, 1981. Accepted April 14, 1982.
Experimental and Theoretical Study of a Two-Dimensional Virtual Impactor Larry J. Forney” Schools of Chemical and Civil Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332
Davld G. Ravenhall Department of Physics, University of Illinois, Urbana, Illinois 61801
Seung S. Leet Department of Civil Engineering, University of Illinois, Urbana, Illinois 6180 1
Experimental results are presented on the operating characteristics of a versatile two-dimensional virtual impactor. The influence of geometry, including the throat angle (38O IO I 58.2O) and normalized void width (0.7 I h/w I 2.0), on the particle cutoff diameter, efficiency curve steepness, and properties of the internal particle loss factor are presented. Theoretical solutions to Navier-Stokes equations are used to correlate the data for fixed instrument Reynolds number (Re = 1540) and bleed flow (Q/Qo = 0.1). The theory, supported by trends in the empirical data, predicts that internal particle losses reduce to zero as the normalized void width increases to h / w = 1.4 f 0.1, while the data show a minimum of 7% loss at h / w = 1.6 f 0.1. Increasing the void width, however, is shown to substantially reduce the steepness of the particle efficiency curves. Introduction
The potential health hazard represented by micron-sized suspended particles in the ambient air has focused attention on aerosol sizing instruments. One of the simplest devices capable of size-classifying aerosols (e.g., respirable vs. nonrespirable) are inertial impactors. Traditional particle-surface interaction problems associated with inertial impactors such as particle bounce, reentrainment, ‘Current address: US.Army Corps of Engineers, Baltimore, MD 21133. 492
Environ. Sci. Technol., Vol. 16, No. 8, 1982
and collection-surface overload have been largely eliminated recently by the replacement of the solid impaction surface with a slowly pumped stagnate air void. Progress has been rapid on these new designs called virtual impactors in recent years (1-€9,and several instruments that utilize the virtual impaction concept are now available commercially. Most of the research and development on virtual impactors has focused on axisymmetric geometries (round jets), and these designs have been incorporated into commercial instruments with excellent results. However, there are a number of operational difficulties associated with the use of axisymmetric jets. Because these devices are of fiied geometry, it is necessary to use a number of test aerosols of varying particle diameter to calibrate the instrument. In addition, constant air-flow rates restrict the sizing capabilities of the device to a single particle cutoff diameter. Moreover, it is necessary to limit the volume flow rate through a round jet in order to prohibit jet instabilities. Nevertheless, experimental data taken with single-stage axisymmetric virtual impactors demonstrate the attractive features of steep particle-efficiency curves and small internal particles losses (6-8). Recent attempts to theoretically predict axisymmetric virtual impactor performance have had only limited success (9). While it has been shown that one can adequately predict the total efficiency curves for these devices (fraction of particles projected into the void and lost on internal surfaces), attempts to predict internal particle losses have
0013-936X/82/0916-0492$01.25/0
0 1982 American Chemical Society
I
I
/
*Outflow
4
Excess
Air Movable
Figure 1. Schematic of twfxiimensional virtual Impactor where aerosol inflow = Qo,bleed flow = 0,outflow = Qo-Q,and excess air = Q,.
been unsuccessful. Earlier theoretical work with two-dimensional virtual impactors by Forney et al. (10) and Ravenhall et al. (11)has provided fundamental information on general flow field properties and sizing capabilities of virtual impactors such as the fluid flow patterns, onset of instabilities, steepness of the efficiency curves, and properties of the internal particle loss spectrum. In the present paper extensive experimental results with a prototype two-dimensional virtual impactor are presented and correlated with theoretical predictions. In particular, this study documents the effects of the throat
angle and, more important, the void width on the magnitude of the particle cutoff diameter, the slope of the efficiency curve, and the number of particles lost internally. Water model studies are also used to suggest limitations on the device geometry and Reynolds number to prevent a breakdown of the fluid flow field. These features, including the operating characteristics of the two-dimensional device, are common to axisymmetric impactors, and these results should provide useful information concerning the performance of virtual impactors of all geometries. The present device, which represents a modification of an earlier impactor by Forney (5),uses a variable slit width for convenience similar to that developed by Cooper and Spielman (12)and Delany and Dolan (13). The variable-slit feature allows one to easily calibrate the instrument with a single test aerosol and would also provide a cumulative particle size distribution on a real-time basis with the integration of either optical or electrical sensors attached to the primary air flow. In addition, the present geometry minimizes the number of dimensionless groups necessary to characterize instrument performance.
Instrument Design The basic design of the prototype device used in this study and shown in Figure 1is identical with the geometry presented earlier by Forney (5). Modifications have been included, however, to eliminate unnecessary internal particle losses. These modifications consist of the addition of an excess clean-air inlet to reduce particle losses on the inside front wall of the device as shown in Figure 1 and the addition of filter holders to the exhausts as indicated in Figure 2. As described in the earlier work of Forney, the instrument is constructed of brass and aluminum stock and is
&Micrometer / /
Micrometer
-c@
- - - - -------E x h a u sFti l tFei rt t i n g
Flgure 2. Side view of virtual impactor. Envlron. Sci. Technol., Vol. 16, No. 8, 1982 493
designed to operate at fixed total volume flow rate Qo and throat angle 8. The two movable plates shown in Figures 1 and 2 are constructed such that both intersect at a common point on the left throat wall. This feature minimizes the number of geometric ratios in the problem and allows one to preserve geometric and dynamic similarity as the slit dimensions h and w are simultaneously changed with the two micrometers indicated in Figure 2. I t can be shown for the geometry of Figure 1that the total particle collection efficiency E of the device, defined as the percentage of particles that project into the void volume or impact with the horizontal fluid deflecting plate (movable), must be a function of five dimensionless groups 01
E = E($, s / w , h / w , Re, Q/Qo) Here, $ is the dimensionless Stokes number, $ = ( p Qo/ 1 8 p l ) ( d / ~ ) ~where C , pp is the particle density (g/cmg, Qo is the total volume flow rate (cm3/s), p is the air viscosity (g/(cm 8 ) ) . 1 is the impactor dimension perpendicular to the flow (cm), d is the particle diameter (cm), and C is the slip correction factor. Also, besides the device dimensions w, s, and h as shown in Figure 1, Q is the bleed flow or secondary flow (cm3/s) and Re = p Q o / l p is the Reynolds number. Moreover, since s / w = cot 0 and one can fix the geometric ratio h / w for any jet width w by proper adjustment of the micrometers, the particle efficiency reducea to E = E($) for constant throat angle B and instrument flow rate Qo in which case $ = ( d / w ) 2 . It is therefore possible to calibrate the instrument with a single test aerosol and to determine a cumulative size distribution of a polydispersed aerosol by proper adjustment of the slit widths w and h while preserving both dynamic similarity (constant Re) and geometric similarity (constant s / w and h/w).
Model Studies A scaled-up water model of the impactor was constructed of Plexiglas. Dye injection flow visualization techniques were used to determine general properties of the flow field. Results of similar studies for moderate Reynolds numbers and device geometries are extensively documented in the earlier work of Forney, et al. (IO). In the present study, however, limitations are suggested for the usable range of device Reynolds number, Re, and mechanisms are discussed that can lead to a total hreakdown of the fluid flow field. Because excessive fluid flux into the void volume prohibits effective particle sizing with a virtual impactor, a systematic study was conducted on the causes of intermittent fluid loss to the void. Two problem areas were identified, and their effects were observed with the water model. The fmt is the existence of a fluid boundary layer along the throat wall on the left of Figure 1, while the second is the onset of fluid instabilities in the jet core at sufficient distance from the throat. So that the effects of both problems could be measured, visual observations were made of the percentage loss of an ink tracer positioned at a distance of 0 . 2 from ~ the left wall of the impactor throat (see Fimre 1 ) . and the results of these measurements are shown% Figure 3. For small-iet Revnolds numbers in the ranee Re < 700 a large loss of tracer fluid to the void was observed as shown in Figure 3. This results from the fact that fluid in the boundary layer along the wall of symmetry on the left of Figure 1 does not have sufficient momentum to overcome the adverse pressure gradient. A t large Re (>1600) or large geometric ratios, jet-core instabilities I
4S4
Envlron. Scl. Technol.. Vol. 16.
No. 8. 1982
R.
Flgure 3. Percent of Ink traCBr lost 10 void vs. jet-Reynolds number. Innial tracer posklon b w15 from impactor wall.
F w r e 4. Photograph of flow field illustratlng jet-cae lnstabllkies fa s I w = 3.0, h l w = 2.8, and Re = 1500.
developed as shown in Figure 4. These instabilities can also lead to excessive fluid flux into the void as demonstrated, and the results of measurements that indicate both effects are shown in Figure 3. AU prototype tests described in this study were conducted for Re = 1540,which is within the indicated acceptable range of 700 < Re < 1600.
Prototype Tests Laboratory tests were conducted with the prototype device. The instrument was suspended in a 600-L aerosol holding changer. The test aerosol was a monodispersed 2.5-pm diameter, uranine-tagged diodyl phthalate (DOP) oil droplet. The aerosol was generated in prefiltered air with a vibrating orifice generator, and the droplet charge was neutralized with a Thermc-Systems Model 2054 =Kr source. Particle efficiency and internal losses were determined at a total volume flow rate of Qo= 30 I/min (Re = 1540). The aerosol was collected on 0.8-pm pore Millipore fdters at the primary and secondary (bleed flow) exhausts of the instrument as indicated in Figure 2. The particulate was washed from the filter and analyzed fluorimetrically, and the results were compared with that collected from a third fdter mounted near the device whose total volume flow rate
1.0
ct 3
Q.JQ,
o,8
h / w = 1.3
o.6 E. L
I
Re = 1540
Re = 1540 QiQ,= 0.1
=
(
I
0 0.02
A
!
f
0.05
siw = 1.0 h / w = 1.3
t
N 0
0.10 0'05
/a
I
& : ' 0.2 O
U
I 0
;2
04
I 0.6
I
I 08
I
I 1.0
I 1.2
0
0.4
0.2
0.6
0.8
1.0
1.2
+'/a "72
Flgure 5. Particle efflclencies for varying clean excess alr ratios, Qe/Qo. Solid symbols are particle losses; open symbols are particle collection efflclencies.
matched that of the impactor. A typical experimental procedure was to sample the aerosol for a period of 30 min at each of several values of the particle Stokes number $ while holding the geometric ratios s / w and h / w , the total volume flow rate Qo, and the ratio of flow rate Q/Qo constant. Thus, the collection efficiency E of the prototype sampler was easily determined over a range of Stokes numbers $ with the same test aerosol of fixed particle diameter by simply adjusting the slit widths w and h. Earlier experimental work (5) demonstrated that the present instrument was subject to internal particle losses on the left front wall shown in Figure 1. Section 2 describes modifications to reduce the magnitude of these losses. The effect of the addition of clean excess air Q, into the void space along the left front wall on the magnitude of internal particle losses is demonstrated in Figure 5. It is clear from these data that the magnitude of Qe/Qo only affects internal losses at large $1/2. All internal particle 0.9 losses for values of $1/2 smaller than that near $lI2 of Figure 5 are assumed to be a result of particles impacting on the top of the horizontal fluid deflecting plate (see Figures 1and 2). The data of Figure 5 suggested an excess bleed flow of Qe/Qo = 0.15 to minimize internal losses. Moreover, it was concluded that the complete elimination of all internal losses at large $1/2 would require some redesign of the void space, but this was outside the scope of the present work. The effect of the magnitude of secondary flow into the void was investigated,and these results are shown in Figure 6. Clearly, increasing Q/Qo decreases internal particle losses. A value of Q/Qo = 0.1 was chosen, which is equal to the magnitude of the bleed flow used on most commercial axisymmetric devices. In conclusion, all experimental data with the prototype device were taken with typical fixed values of secondary flow Q/Qo = 0.10, Reynolds number Re = 1540 (Q = 30 L/min), and excess air flow Qe/Qo = 0.15. One typical set of experimental data is shown in Figure 7.
Data Correlation So that the influence of geometry on impactor performance could be fully characterized, important operating parameters are introduced similar to those defined earlier by Ravenhall et al. (11). These parameters have been modified here where appropriate to account for the secondary or bleed flow into the impactor void. The first
Figure 6. Particle efficiencies for varying secondary flows, Q / Q o Solid symbols are particle losses; open symbols are partlcle collection efficiencies. 1.0.
-
1
1
1
1
1 U
1
u1 t1
~,/a,=o.i5
-
s/w = 1.0 h/w = 1.1 Data
0
-
-
.6-
!
-
Re=1540 QiQ, = 0.1
-
E-O L-A
$Y2
Flgure 7. Typical experimental data for total efficiency and loss factor.
parameter is the value of $lI2 for which the total efficiency is E = (1 f)/2 (1)
+
where f = Q/Qois the fraction of total air flow passing into the void. Thus for f = 0.1, we define a parameter $1/2 ( E = (1 + f ) / 2 ) = $551/2 proportional to the particle cutoff diameter. This characterizes the cutoff diameter for the primary flow. Also of interest is a second parameter S, which we call the relative dispersion. This represents a measure of the spread of the particle efficiency curve E about the particle cutoff diameter. Formally, S-' is an estimate of the slope of the efficiency curve, where S is defined by d(ln $lI2)/dE evaluated at E = (1+ f)/2. Since it is desirable to maximize the normalized slope of the efficiency curve for an impactor, one seeks to minimize S. Here we define S in the segment of the efficiency curve away from the influence of the secondary flow as $1/2(E= 0.84 0.16f) - G1l2(E=0.16 0.84f) S= (2) 0.68(1 - f)$'12(E = 0.5 + 0.5f)
+
+
where, in this study S(f = 0.1) =
$s561'2
- $244lI2
0.612$551/2
(3)
Environ. Sci. Technol., Vol. 16, No. 8, 1982 495
Re = 1540
woo = 0.1 O&lo
= 0.15
Data s/w ( 8 )
TheDv
.-
1.28 1380) 1.0 145") 0.62 I58.ZEl -
0
x
-
1.4
0
0
A 0
1.2
i
S
0
Additional parameters defined are the peak value L, and area A under the loss factor L. The loss factor L represents the fraction of particles of a given #ll2 that make contact with internal surfaces of the device and thus do not appear at either the primary or secondary instrument exhaust. Hence with either experimental or theoretical data, we define a normalized area under the loss factor as
.6
.E
1.0
1.2
1.4
:-t
.3
Results and Discussion
496
Envlron. Scl. Technoi., Voi. 18, No. 8, 1982
1.8
2.0
Flgure 9. Relative dispersion vs. normalized void width.
which represents the total fraction of particles lost to internal surfaces. Theoretical efficiency and loss factors were determined numerically by using relaxation techniques similar to those described elsewhere (9). The details of the computations for the present application are described by Ravenhall et al. (14). Variation of the operating characteristics, #551/2, A, L,, S with changes in the jet-to-plate spacing s / w , and the normalized void width hl w were determined from the experimental data. These results were compared with data derived from the numerical computations and are discussed in the next section.
Experiments with a scaled-up water model operating within a normal range of geometries indicate that the flow field in the present two-dimensional device is subject to irregularities for both small (1600) Reynolds numbers. The lower limit in Re in the present device is due to boundary-layer development on the plane of symmetry (left throat wall of Figure l), but this would not inhibit the operation of axisymmetric devices. The upper limit in Re for the present two-dimensional device was due to the development of jet-core instabilities (see Figure 4). More research is necessary, however, to clarify the potential problem of jet-core instabilities in round jets at large Re. Normally, it would not be necessary to operate rectangular slit impactors at large Re to achieve greater flow rates since this can be accomplished with longer slits at fixed Reynolds numbers. Experimental and theoretical data in Figure 8 indicate that the particle cutoff diameter #551/2 is independent of the normalized void width h l w . Theory demonstrates a 25% increase in $551/2 as the jet-to-plate spacing s / w increases from 0.62 to 1.28. While the latter trend was not evident from the experimental results, theory correlates empirical results to within the scatter of the data of 12%. The dispersion S of the total particle efficiency about the particle cutoff diameter is plotted in Figure 9. Theory
1.6
hlw
Lm
0
.2
hlw
Figure 10. Peak in loss factor vs. normalized void width.
predicts the trend of increasing S vs. the normalized void width h l w in the data. However, the theoretical predictions are roughly 50% lower than experimental results. As in Figure 8, differences in experimental results for the dispersion S due to changes in jet-to-plate spacing over the range 0.62 5 s / w 5 1.28 were not demonstrated. Theoretical and experimental predictions of properties of the particle loss factor L are shown in Figures 10 and 11. Theory and experiment for both the peak in the loss factor L, and the area under the loss factor A representing the total fraction of particles lost show a marked decrease with increasing void width h l w . The experimental value of A was determined for each instrument geometry from the measurements of particle losses (e.g., Figure 7). Since the experimental data representing L did not approach zero for large values of the Stokes parameter 9, the data were extrapolated such that L 0 as shown by the dashed
-
.7
I
1
I
I
l
I
.6
1
I
Re = 1540 Q/Q, = 0.1 Qe/Qo = 0.15
-A
I
I
'
Acknowledgments
-
The research reported here was supported by the US. Department of Energy Grant EE-77-S-02-4319 to the University of Illinois and subcontracted to the Georgia Tech Research Institute. We thank R. L. Butenhoff and R. W. Beadle, who acted as contract officers. The findings, opinions, and recommendations expressed are those of the authors and not necessarily those of the University or the U.S. DOE.
Literature Cited (1) Hounam, R. F; Shenvood, R. J. Am. Znd. Hyg. Ass. J. 1965,
hlw
Flgure 11. Number fraction lost vs. normalized void width.
line in Figure 7. This procedure provided a finite value of A. Clearly, all of the particles would project into the void for large 9 or L 0. However with the present instrument, the modifications described in Section 2 were not extensive enough to reduce all particle contact on the left front wall (see Figures 1 and 2) of the void space, and further modifications to the instrument design were beyond the scope of the present work. Theory indicates that L, and A should become negligible for large h / w (51.41, while experimental data demonstrate a minimum of A N 7% at a similar value of hlw N 1.6. Theoretical predictions of L , and A in Figures 10 and 11 are lower than the empirical results by as much as 50%. However, theoretical predictions of decreasing L , and A with decreasing jet-to-plate spacing s / w are correctly correlated with the experimental data.
-
26, 122. (2) Conner, W. D. J . Air Pollut. Control Assoc. 1966,16, 35. (3) Loo, B. W.; Jaklevic, J. M.; Goulding, F. S. Lawrence Berkeley Laboratory Publ. LBL-3854, Berkeley, CA, 1975. (4) Dzubay, T. B.; Stevens, R. K. Environ. Sci. Technol. 1975, 9, 663. (5) Forney, L. J. Rev. Sci. Instrum. 1976, 47, 1264. (6) McFarland, A. R.; Ortiz, C. A,; Bertch, R. W., Jr. Environ. Sci. Technol. 1978, 12, 679. ( 7 ) Loo, B. W.; Adachi, R. S.; Cork, C. P., Goulding, F. S.; Jaklevic, J. M.; Landis, D. A.; Searles, W. L. Lawrence Berkeley Laboratory Publ. LBL-8725, Berkeley, CA 1979. (8) Forney, L. J.; Ravenhall, D. G. U.S. Department of Energy, Final Report COO-4319-6, 1980. (9) Marple, V. A,; Chien, C. M. Enuiron. Sci. Technol. 1980, 14, 976. (10) Forney, L. J.; Ravenhall, D. G.; Winn, D. S. J. Appl. Phys. 1978,49, 2339. (11) Ravenhall, D. G.; Forney, L. J.; Jazayeri, M. J. Colloid Interface Sci. 1978, 65, 108. (12) Cooper, D. W.; Spielman, L. A. Atmos. Environ. 1974,8, 221. (13) Delay, A. C.; Dolan, G. J. Rev. Sci. Instrum. 1975,46,1650. (14) Ravenhall, D. G.; Forney, L. J.; Hubbard, A. J. J. Colloid Interface Sci. 1982, 85, 508.
Received for review September 26, 1980. Revised manuscript received July 30, 1981. Accepted April 15, 1982.
Envlron. Sci. Technol., Vol. 16, No. 8, 1982 497
