Chen1985 PDF
Chen1985 PDF
Chen1985 PDF
00+000
Printed in Great Britain. PergamonPress Ltd
INTRODUCTION
There is increasing interest in the use of size-classified aerosol samplers for atmospheric
sampling, because deposition of particles in the respiratory tract and reduction of visibility by
light-scattering strongly depend on particle size. One of the most commonly used size
classification devices is the conventional impactor, originally developed by May (1945), where
particle size fractionation is performed by inertial separation. In the impactor, a small plate is
placed such that the surface is perpendicular to the air flow at the exit of a particle
acceleration nozzle. The performance of a well-designed impactor has a sharp cut-off
efficiency curve following a theoretical prediction (Marple, 1970). However, particle--surface
interaction problems associated with the impactor such as particle bounce, re-entrainment,
break-up, and collection-surface overload have led to non-ideal performance (Rao and
Whitby, 1978; Cheng and Yeh, 1979) and introduced inaccuracy in the data interpretation
(Mercer, 1973).
In search of a technique that uses the concept of inertial impaction, but avoids
particle--surface interaction problems, the virtual or dichotomous impaction principle has
been developed (Hounam and Sherwood, 1965; Conner, 1966; Dzubay and Stevens, 1975;
Forney, 1976; Loo et al., 1976; Willeke and Pavlik, 1978; Masuda et al., 1979; Solomon et al.,
1983). In such a device, the solid impaction plate used in the conventional impactor is replaced
by a region of relatively stagnant air contained in the cavity of a collection probe. The major
air flow which carries low-inertia particles is laterally moved away similarly to the
conventional impactor, while high-inertia particles in the stagnation region are passed into
the collection probe at a relatively low flow rate (minor flow). Both size fractions can be
subsequently ducted to any desired location for further size classification or for any desired
method of analysis, including direct-reading continuous instrumentation.
In this paper extensive experimental evaluation of two prototype round-jet virtual
impactors is presented. In particular, this study documents the effects of nozzle Reynolds
number (Re), minor flow rate, and the geometries of the nozzle and probe on the magnitude
of the particle cut-off diameter, slope of the efficiency curve, and percentage of internal
particle losses. Both solid polystyrene Latex (PSL) and liquid dioctyl phthalate (DOP)
aerosols were used to determine the size separation efficiency of this virtual impactor, but
only DOP aerosols were used to quantify internal particle losses.
343
344 B. "T CHEN et al.
AEROSOL
LAMINATOR- - ~
i
--NOZZLE
SPACER-- '--'SPACER
COLLECTION/
PROBE
MINOR ~ . . ~
FLOW
)
MAJORFLOW
Fig. I. Schematic diagram of the single-stage virtual impactor.
VIRTUAL IMPACTOR
Based on conventional impactor design, two prototype virtual impactors having different
nozzle diameters (W) were designed and constructed. Figure 1 shows the basic design of the
virtual impactor. Aerosol is drawn through an entrance and laminarized by passage through a
ring of small holes (laminator). Flow is subsequently accelerated by the narrowing cross-
section of the duct, thus, increasing the momentum of the aerosol particles. Finally, particles
are separated according to their aerodynamic diameters at the gap between the nozzle and the
collection probe. The major flow which contains most of the fine particles is removed
perpendicular to the acceleration nozzle and collection probe, while the remaining flow
(minor flow) which contains the coarse particles and a small portion of the fine particles
passes across the gap and down through the collection probe. Because of the fundamental air
flow design of the virtual impactor, contamination of fine particles in the minor flow is
proportional to the ratio of minor flow rate to total flow rate. Four pressure taps are used to
monitor the pressures upstream of the nozzle, at the nozzle, and downstream of the nozzle at
both the major and minor flow paths.
Table 1 shows values of design parameters used in the prototype virtual impactors. The
nozzle throat length (T), entrance angle (0), and probe diameter (D2) were identical in both
impactors. Parameters such as nozzle to probe distance (S) and hole diameter (D~)could be
changed to investigate their effects on the size separation efficiency.
METHODS
Generation of solid particles
PSL spheres (Duke Scientific, Palo Alto, California) between 0.6 and 5.0/an in nominal
diameter were generated with a Lovelace nebulizer, dried by a cylindrical diffusion dryer
containing desiccant, and then introduced into a Kr-85 bipolar ion source to bring them to a
state of charge equilibrium (Fig. 2). Before entering the virtual impactor, the particles were
diluted and mixed with filtered compressed air in a mixing chamber.
The number concentrations of PSL spheres in both the major and minor flows were
measured individually by the same Aerodynamic Particle Sizer (AlaS) (TSI, Inc., St. Paul,
Minnesota, U.S.A.) to determine the particle separation effieieneies of both impactors
(Prototype I and II). All the valves downstream from the impaetor were used not only to
adjust the total (Qr), major (QM), and minor (Q,,) flow rates through the impactor, but
also to provide a proper amount ( ~ 5.7 l/rain) (Chen et al., 1985) of air flow through the APS.
Flow rate through a 3-way valve VI to the vacuum was also adjusted to be 5.7 l/rain to match
the total flow through the APS. Figure 2 shows the set-up of valves when the number
concentration of the major flow was measured: V2 and V4 (2-way valves) were closed, Va (2-
way valve) was open, and VI connected minor flow and make-up dilution air to the vacuum.
Normally, both make-up dilution air (Qd) in the minor flow line and bleed-off air (Q~) in the
major flow line were required, because during the ordinary operating conditions (Re ~ 4000),
the minor and major flow rates were less and greater than 5.7 l/rain, respectively. During
operation to determine the concentration of minor flow, V2 was open, Va and V4 were closed,
and V~ was rotated clockwise to connect the major flow to the vacuum. During valve°
switching periods between two measurements, V4 was open to the ambient air to reduce air
flow fluctuations that might be caused by switching valves. The advantage of using one APS
for both major and minor concentration measurements is that the results can be compared
and used to calculate the separation efficiency of the test aerosol without additional
calibration.
, , , , . FILTER
I DIFFUSION ~ BIPOLAR I j MIXING
[ DR;ER I -[ ION SOURCE :I -[CHAMBER~LTOT - L ~
FLOW
L VALVE / 10. N
T
o,LoT,oN 1 v VAOUUM
l I COMPRESSED~
ILOVELACE1 AIR I ~ MAKE-UP J
I I ~ DILUTION I
NEBULIZER Qo
i I V2 F" MINOR , ' ,
A = P, r'.A w I VIRTUAL I
AMBIENT " ~ 7 ~ V4 ,
I vALvk'-JV 3 1......
/ BLEED-OFF
15.7 I/rain ~ FLOW
T ~ Qb
AERODYNAMIC
PARTICLE
SIZER ~ VACUUM
Fig. 2. Experimental set-up when solid PSL aerosol was used in the study.
346 B T. C u E s et aL
Because the APS has good resolution tbr particles greater than 1 ~m in aerodynamic
diameter (Chen et al., 1985), it was possible to obtain more than one data point per run by
generating PSL aerosols of different nominal sizes concurrently with the same nebulizer
Vibrating - Orifice
Generator
J
~ Dilution Air
- Overflow
Minor ~
Flow .~ i
Large Particle l%'-'21 I ~ I Aerodynamic1
Co,,ection I Part,o
P' o,.er I
Mai°r FI°w
valve(~ ~ = Vacuum
Small
Vacuum Particle
Collection
Filter
Fig. 3 Experimental set-up when liquid D O P aerosol was used in this s t u d y
A novel virtual impactor: calibration and use 347
where m = the amount of fluorescein deposited on minor flow (large particle) filter, and
M = the amount of fluorescein deposited on major flow (small particle) filter.
\A/
cl
o11 I
I F ,
I !
Fig. 4. Schematicdrawing of virtual impactor parts that were washed separately to analyze wall
losses.
348 B T CHEN et aL
80
.48
0.91
! 2 . 3 5 2.79
0
I1 .
I I I
129 200 300 384
CHANNEL NUMBER
Fig. 5. Accumulation bin output o f an APS with an aerosol containing, from left to right, 0.91, 1.48.
2.35, 2.79, and 3,30/tin PSL spheres.
(i) E~ect of Re
To determine the influence of Re, experiments were run with the Re listed in Table 2, while
the other parameters were held constant. Separation efficiency curves of virtual impactors I
and II as a function of aerodynamic particle size (D~E)of PSL particles are shown in Figs 6(A)
and (B), respectively. Similarly to the characteristics in a conventional cascade impactor, the
EFFECT OF Re
VIRTUAL IMPACTOR I
7P- 100
uJ
/ [~Re=3900±50
ee
Lkt
A
~Re:2330
~.. 8O
Re=1950
~ 50
7
al
U-
ii 40
Z
0
7- 20
<
er
<
00.5 1.0I I
2.0
I I . I
4.0 5.0 10,0
1
A E R O D Y N A M I C DIAMETER (DAE), p.m
EFFECT OF Re
VIRTUAL IMPACTOR II
,00 8o
B
Sly oo
i 60 =~Fi
970e=
40
ktJ
20
I I l I I I
t.u
oo 0 0'.5 1.0 2.0 4.0 5.0 10.0
AERODYNAMIC DIAMETER (DAE), film
Fig. 6. Separation efficiency as a function of aerodynamic diameter of PSL spheres for (A) impactor
I(Q=/Qr = 10.8 + 0 . 9 9 o, S/W= 1.03, and DI/W= 1.51) and (B) impactor II (Q,/Qr= 10.9 :i: 0.9 9o,
S/W= 1.0, and Dt/W= 1.25) operated under different nozzle Reynolds numbers (Re).
data indicate that with a fixed nozzle diameter (W), a virtual impactor operated at higher Re
has a smaller 50 ~o cut-offaerodynamic diameter (Dso), and with a fixed Re, the impactor with
a larger W(Impactor I) has a larger Dso. We expected the efficiency curves for different Re to
be asymptotic to about 11 9o for smaller particles, because contamination of the fine particles
in the minor flow is equal to Qm/Qr, which is about 11 ~ for the operating conditions of both
impactors. The variation of efficiencies (Figs 6A and B) between O and • in both (A) and (B),
and between [] and • in (A) are mainly due to small variation in Q,~IQr value.
All the data shown in Figs 6(A) and (B) can be consolidated into a single curve (Fig. 7) ifDAE
is replaced by a dimensionless particle diameter, the square root of Stokes number (Stk),
which is defined as
Stk = D~ECVo
9~W "
where C is the particle slip correction factor, V0 the mean fluid velocity at the nozzle throat,
and/a the fluid viscosity. And the (Stk)~82 value corresponding to 50 ~o separation efficiency in
both impactors is about 0.67. The results shown here suggest that we could predict both the
separation efficiency curve and the Dso of any geometrically and dynamically similar virtual
350 B. T CHEN et at
100
z
UJ Re=7920 _ _ _ ~ = 3900*.50
o Re=4200 *-.50--- ~:~
rr
Lu 80
13... ~--Re=1950
Re:4200±50 -~ .~
~- 6oi Re=970~ Re=3900±50
o
7, ....... Re=2330
_c2 40
IJ.
7 Re 8180
o 2o
rr
<
O- I I I I i I
00.1 0.2 0.4 0.5 1.0 2.0
( S t k ) 1/2
Fig. 7. Separation efficiency as a function o f the square root of Stokes n u m b e r (Stk). Data from
Figs 6(A) and (B) are represented by a single hand-fitted curve.
impactor of fixed Q=/Qr if Re is between 1000 and 8000, S/W is ~ 1, and D1/W is between
1.25 and 1.5.
EFFECT OF Qm/Q T
VIRTUAL IMPACTOR I
t- 100
Z A
LtJ
O
LU 8O
g--
v
>- 60 Qm/QT •
o
z
tu ~.~-<~>--~-"/4 /
,Sld
t~
I.L 40 ,,.,,<oo,,--.
ILl
z
o 20
C-- ~'- 6.0%(PSL)
m
n I I I I i
IJJ 0 I 1 I I
0.1 0.3 0.5 07 1.0
(Stk)~12
EFFECT OF Qm/QT
VIRTUAL IMPACTOR II
100
B
~_ 8o
°"°----' IJ l
°77 i
60
E 4o
12J ~.°,,---2~ ? _/
!2o
, ~nim"~4n's' i 1,11 ° 1
~ 00.1 0!2 21.0
(Stk) 112
Fig. 8. Separation efficiency curves showing the effect of Q - / Q r in (A) impactor I (Re = 3900 + 50,
S / W = 1.03, and DI/W = 1.51) and (B) impactor II (Re = 4200 + 50, S~ W = 1.00, and DI /W = 1.25).
1 .0~
II
C•Prototype
"~= 1.2766-0.808 1 ( I O g l o X)
0.8 ~ + 0 . 1 9 2 2 ( I O g l o X) 2
tZ3 Prototype I- - - - - ' @ ~
0.6~-
0.5 i l I I I
2 3 5 10 20 30
Qm/Q T ( P e r c e n t )
Fig. 9. Et']['CCIo[" (2,,1/QT Oil 50o0 CUt*ON (Slk) 1'2 Of the virtual impactors.
352 B 'I~ CHEN et aL
100
Z
A ~~ IO0 If B J
W
~ 80 a_ go
"V /
~
~ 60 , >. /
I
Z
w D ~ 1.03 (-) S/W ]
i
t
~ 40- j .... 0.53 -- 40
U-
h
060
- --4/
W W 1 50 ........ ~ 7
l
Z -- Z oo ..... -:7
0 _o 2o
~ 2O
<
1 I I I I I i I.iJ I l
w 0
m 0.1 0.2 0.4 0.5 1.0 (,,) O0 1 0.2 ' o.'4o;s' ' ' io
( S t k ) ~j2 (Stk)~'2
Fig. 10. Selmration efficiency curves showing the effect of S I W in (A) impactor I iRe = 3900 + 50,
(2=/0. = l O . 8 ± 0 . 9 % , a n d D i / W = 1.51)and (B)imlm~or II (Re = 4200±50,Q.,/QT= 10.9±0.9",.
and Dr~W= 1.01).
PSL particles were used. The results shown in Fig. 10(B) for impactor II are very similar for
three different $/W values and are represented by a single curve (Dr / W value used here is 1,01
instead of the base value 1.25). Figure 10(A), however, shows a slisht difference between the
curve for S/W= 0.53 and that for $/W= 1.03 and 1.59. Despite the indicated difference, all
the data here suggest that the effect of S/W on virtual impactor performance is small
compared with S/W effect for conventional impactors (Marple and Liu, 1974).
U
n
g-
v
>- 60 >. 60
(3
Z 1.51- J Z
O
W
ua
,7 40 _o 4O
LL
W
Z
o_ 2o _o 20
I-- I,.-
<
he
< <
n
l I I I I I Ii,,I I I l I 1 I I I
w 0 I Ill 0 ~ 0
o9 0.1 0.2 0;4 0 5 . ~0 0.1 0.2 0.4 0.5 1.0
( S t k ) lt2 ( S t k ) w2
Fig. 11. Separation efficiency curves showing the effect of Dl / W in (A) impactor I ( Re = 3900 + 50,
Q,-/QT = 10.8 + 0.9 ?'o, and S / W = 1.03) and (B) impactor II (Re = 4200 4- 50, Q,,/QT = 0 9 + 0 9 '~,
and S / W = 1.00).
A novel virtual impactor: calibration and use 353
selection of larger Dt/Wvalues ( > 1.01) as base values is based on a trade-off, i.e. wall losses
are lower for larger values of Dt / W, though the cut-off characteristics might not be as good.
Wall losses
Liquid D O P particles tagged with uranine were used for internal wall loss analyses in
impactor I. Figure 12 shows total particle losses inside impactor I at three different operating
conditions ( S / W = 1.03 and D I / W = 1.51 were fixed), and Figure 13 shows wall losses
measured for individual parts (Fig. 4) of the impactor. For all conditions, the highest losses,
mainly on the upper surface of collection probe, are shown at the sizes approximately
corresponding to the 50 % cut-off (Stk) ~/2. For particles larger than 10/am in aerodynamic
diameter (i.e. (Stk) ~/2 > 2.0 and 3.0 in the cases of Re = 3850 and 7400, respectively), the wall
losses were high, primarily due to impaetion on the inner surfaces of the converging region of
the acceleration nozzle. It is interesting to note that around the cut-off sizes, losses at
Re = 3850 increase as Q~,/Qrdecreases, and losses decrease with increasing Re. These results
agree with those presented by McFarland et al. (1978). Figure 13 shows typical losses on
internal parts (A) nozzle, (B) collection probe, (C) spacer, and (D) transport of major flow (see
Fig. 4) of the impactor I operated at Re = 3850, and Q,,/Qr = 12.5 ~o. The amount of losses
due to parts (E) and (F) are negligible.
Wall losses on the upper surface of the collection probe of this prototype virtual impactor
were large (20 ,-, 40 ~ ) for particles near the cut-off size (Fig. 12), indicating that the hole
VIRTUAL IMPACTOR I
40
Re Qr./QT
3o ~ 3~o
3850 18.1%
7400 12.5%
Q_
~ 20
-~ ~o
0 . I
0.3 0.5 1.0 2.0 4.0 5.0
(Stk)~ ~2
Fig. 12. Total wall loss versus (Stk) 1~2 for virtual impactor 1 (S/W= 1.03, a n d DI/W= 1.51).
VIRTUAL IMPACTOR I
40 Re Qm/QT
385--0 12?5%
C ollectio, n P r o b e
Z 30
cO" 2 0
0
,m,,o, ,,ow,
.J
~ 10
I Nozzle
( S t k ) ~j2
Fig. 13. Wall losses measured for individual parts of virtual impactor I, labelled as in Fig. 4
(S/W= 1.03, a n d DI/W= 1.51).
,354 8 ][ CHENet al
diameter o f the p r o b e m a y be t o o small and, as a result, some particles near the cut-off size
impact and are deposited on the surface (like the impaction plate o f a conventional impactori.
CONCLUSIONS
Acknowledgements--The authors are indebted to many of our colleagues for technical review and suggestions;
Drs B. B. Bocckcr, C. H. Hobbs, and R. O. McClellan for encouragement and support; Mr F. J. Greybar for
constructing the impactors; Mr E. E. Goff for illustration; and Word Processing porsonnel for typing. This research
was performed under U.S. DefraYment of Energy Contract Number DE-AC04-76EV01013
REFERENCES
Chen, B. T., Cheng, Y. S. and Ye2a,H. C. (1984) J. Aerosol Sci. 15, 457.
Chen, B. T., Cheng~ Y. S. and Yeh, H. C. (1985) Aerosol Sci. Technol. 4, 89.
Cheng, Y. S. and Yeh, H. C. (1979) Environ. Sci. Technol. 13, 1392.
Clift, R., Grace, J. R. and Wel~r, M. E. (1978)Bubbles, Drops and Particles, pp. 97-141. Academic Press, New York.
Conner, W. D. (1966) Air Pollut. Control Assoc. J. 16, 35.
Dzubay, T. G. and Stevens, R. D. (1975) Environ. Sci. Technol. 9, 663.
Forney, L. J. (1976) Rev. Sci. lnstrum. 47, 1264.
Hounam, R. F. and Sherwood, R. J. (1965)Am. Ind. HYO. Assoc. J. 26, 122.
Loo, B. W., Jaklevic, J. M. and Goulding, F. S. (1976) Fine Particles (Edited by Liu, B. Y. H.), pp. 311-351. Academic
Press, New York.
Marple, V. A. (1970) Ph.D. Thesis, University of Minnesota.
Marple, V. A. and LiLt, B. Y. H. (1974) Environ. Sci. TechnoL g, 648.
Marple, V. A. and Chien, C. M. (1980) Environ. Sci. Technol. 14, 976.
Masuda, H., Hochrainer, D. and St6ber, W. (1979) J. Aerosol Sci. 10, 275.
May, K. R. (1945) .l. Sci. lnstrum. 22, 187.
McFarland, A. R., Ortiz, C. A. and Bertch, R. W. Jr. (1978) Environ. Sci. Technol. 12, 679.
Mercer, T. T. (1973) Aerosol Technology in Hazard Evaluation, pp. 222-240. Academic Press, New York.
Rao, A. K. and Whitby, K. T. (1978) ./. Aerosol Sci. 9, 87.
Solomon, P. A., Moyers, J. L. and Fletcher, R. A. (1983) Aerosol Sci. Technol. 2, 455.
Willeke, K. and Pavlik, R. E. (1978) Environ. Sci. Technol. 12, 563.