Aerosol Science 35 (2004) 1105 – 1118

www.elsevier.com/locate/jaerosci

Theoretical and experimental study of an axial %ow cyclone

for &ne particle removal in vacuum conditions
Chuen-Jinn Tsaia;∗ , Daren-Ren Chenb , HungMin Cheinc , Sheng-Chieh Chena ,
Jian-Lun Rotha , Yu-Du Hsuc , Weiling Lib , Pratim Biswasb
a
Institute of Environmental Engineering, National Chiao Tung University, No. 75, Poai St., Hsin Chu 300, Taiwan
b
Environmental Engineering Science, University of Washington in St Louis, One Brookings Dr.,
St Louis, MO 63130, USA
c
Industrial Technology Research Institute, Center for Environmental, Safety and Health Technology Development,
Chutung, Hsin Chu 310, Taiwan
Received 17 November 2003; accepted 19 March 2004

Abstract

An axial %ow cyclone to remove &ne particles at low pressure conditions (6–23 torr) was designed and
tested. The inner diameter of the cyclone is 3:0 cm and tested %ow rates are 0.455 and 1:0 slpm. A theoretical
evaluation was &rst carried out to calculate the particle collection e;ciency and cuto< diameter for the axial
%ow cyclone. Experiments were then conducted to test the theory at di<erent %ow rates, pressures and cyclone
designs. It was demonstrated that at su;ciently low pressure of 6 torr (%ow rate = 0:455 slpm), the axial %ow
cyclone is able to remove particles below 100 nm e;ciently, and the %ow Reynolds number was found to
have a great e<ect on the particle collection e;ciency.
? 2004 Elsevier Ltd. All rights reserved.

Keywords: Axial Flow cyclone; Air pollution control; Particle control

1. Introduction

A cyclone separator is a well-known dust separator or particle sampling device which is based
on the particle centrifugal force created by vortex %ow in the cyclone. There are two main de-
signs: tangential and axial %ow cyclones are the two main types. In a tangential %ow cyclone, the
%ow enters tangentially in a cylinder creating a vortex inside. At the end of the outer vortex, the %ow

∗
Corresponding author. Tel.: +886-3-5731880; fax: +886-3-5727835.
E-mail address: [email protected] (C.-J. Tsai).

0021-8502/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jaerosci.2004.03.007
1106 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

Nomenclature

B pitch of vanes
C slip correction factor
dp particle diameter
dpa50 cuto< aerodynamic diameter
e charge of an electron
D cyclone diameter (tangential %ow cyclone)
De exit tube diameter (tangential %ow cyclone)
 particle collection e;ciency
 a factor = V=Ua
Ic electric current at the outlet of the cyclone
It electric current at the outlet of the by-pass tube
k number of charges
L e<ective body length, distance from the tip of the outlet tube to the bottom of the
spindle
 mean free path of air molecules at the inlet of the cyclone when pressure is Pcyc
0 mean free path of air molecules at standard condition
N number of vanes
n number of vane turns
noc (dpk ) particle concentrations at the outlet of the cyclone, the particle carries k charges
ni (dpk ) inlet particle concentration, the particle carries k charges
not (dpk ) particle concentrations at the outlet of the by-pass tube, the particle carries k charges
Pen overall penetration
Pcyc pressure at the inlet of the cyclone
P760 standard pressure, or 760 torr
Pen(dpk ) particle penetration through the cyclone for a particle carries k charges
Pent (dpk ) particle penetration through the by-pass tube for a particle carries k charges
50 dimensionless cuto< aerodynamic diameter
Q volumetric %ow rate
Q0 standard volumetric %ow rate
r new particle radial position
ri particle initial radial position
rmax inner radius of the cyclone
rmin radius of the vane spindle
Ref
%ow Reynolds number (for tangential %ow cyclone)
Ref %ow Reynolds number (this study)
St Stoke number de&ned as St = vt =(rmax − rmin )
Stb Stokes number de&ned in the body section, Stb = Ua =B
tv particle transit time in the vane section
 particle relaxation time
Ua mean axial gas velocity in the body section
Ui inlet gas velocity (tangential %ow cyclone)
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1107

vr particle radial velocity

vt tangential gas velocity in the vane section
V mean gas velocity in the vane section
Vol e<ective volume
w vane thickness
 a &tting constant

is reversed and &nally exits axially at the top. The %ow in the tangential cyclone is usually highly
turbulent and non-stationary, which together with the %ow reversal in the cyclone result in a relatively
high pressure drop across the cyclone (in Maynard, 2000). The inlet and outlet of the tangential
cyclone are perpendicular to each other. If the particle separator needs to be positioned in a straight
section of a %ow system, an axial %ow cyclone may be used in which the tangential motion of the
%ow is induced by guiding vanes installed in line with the %ow direction. The axial %ow cyclone
has a lower pressure drop than that of the tangential cyclone since the %ow is in line with the
axial direction of the cyclone, and the %ow is not reversed and much less disturbed (Nieuwstadt &
Dirkzwager, 1995).
Kao and Tsai (2001) reviewed many existing theories of the tangential %ow cyclone including
Lapple (1950), Stairmand (1951), Barth (1956), Leith and Licht (1972), Dietz (1981), Dirgo and
Leith (1985), Iozia and Leith (1989), Li and Wang (1989), and Iozia and Leith (1990) and compared
them with the existing data. They found that most theories are applicable in a certain range of %ow
Reynolds number only. The %ow Reynolds number is based on the di<erence of the cyclone radius
and that of the exit tube. At small %ow Reynolds number, all theories under-predict the cuto<
aerodynamic diameter and discrepancy increases as the %ow Reynolds number is decreased.
Tsai, Shiau, Lin, and Shih (1999a, b) found that the cuto< aerodynamic diameter (the diameter
corresponding to 50% penetration) of the 10 mm nylon cyclone to be lower than expected presumably
due to build-up of deposited particles inside the cyclone. The nylon cyclone is for respirable particle
collection. A similar cyclone with a bigger body diameter was developed, which was shown to have
a less shift of the cuto< aerodynamic diameter and was also demonstrated to be more accurate in
measuring respirable particle concentrations in the &eld study. Kenny, Gussman, and Meyer (2000)
developed a sharp-cut tangential %ow cyclone which was shown to have a sharp cut for ambient
PM2:5 sampling. The cuto< diameter of the cyclone was less a<ected by particle loading than the
WINS (Well Impactor Ninety-Six) impactor (USEPA, 1997). In the laboratory dust loading study,
Kenny et al. (2000) found WINS shifted steadily from 2.5 to 2:15 m after sampling about 6 mg of
PM2:5 particles, while Peters and Vanderpool (1996) found a shift to 2:25 m after sampling 24 mg
of dust. The e<ect of deposited particles on the cuto< diameter was also studied earlier by Tsai
and Cheng (1995) for impactors. Unless the body diameter is very small (such as 10 mm nylon
cyclone), it is expected that the axial %ow cyclones will be less a<ected by the loaded particles on
the cuto< diameter than impactors since the particle deposit in the cyclones is more di<use.
In contrast to the tangential %ow cyclone, only few researchers have studied the axial %ow
cyclone, such as the experimental work of Liu and Rubow (1984), Weiss, Martinec, and Vitek
(1987) and Vaughan (1988), and theoretical work of Maynard (2000). These researches studied the
axial %ow cyclone operating at ambient condition. For example, an axial %ow cascade cyclone at
1108 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

a design %ow rate of 30 l=min was developed by Liu and Rubow (1984) for sampling high con-
centration of particles. The cuto< aerodynamic diameter of the &ve stages are 12.2, 7.9, 3.6, 2.05
and 1:05 m. Total particle loss in the system, including the loss in the body, vane insert and exit
tube of the collection cup, was shown to be signi&cant. It ranged from 15% for particles of 1 m
in diameter to 33.3% for particles of 8 m in diameter. Therefore, using this cascade cyclone to
measure particle size distributions requires complete recovery of all particles lost in the cyclone.
Maynard (2000) derived the particle penetration of the axial %ow cyclone based on the assumption
that particle collection mainly occurs in the vane and body sections only. The overall penetration,
Pen, is combined from the penetrations derived separately for the vane and body sections as
42 n2 2 Stb2 2nStb
 1=2
Pen = (1−162 Stb L∗ )1=2 + ∗2 ∗2 − ∗2 −r ∗2
1 + 4 2 ∗2
r min +4 2
1 − 16 2 St L∗
b ;
rmax −rmin rmax min
(1)
where n is the number of vane turns, rmax is the inner radius of the cyclone, rmin is the radius of
the vane spindle and L is the e<ective body length (refer to Fig. 1 in Maynard, 2000). The relative
dimensions of these three variables, denoted by the variables with asterisk, is the actual dimension
divided by the pitch of vanes, B. Stokes number Stb is de&ned in the body section as Stb = Ua =B,
in which  is the particle relaxation time. The factor  = V=Ua , in which V is the mean gas velocity
in the vane section and Ua is the mean axial gas velocity in the body section. If there are N vanes
and each has a &nite thickness w, then  in Eq. (1) is calculated as
∗

rmax ∗2
1 + 42 rmax
= ∗ ∗ ): (2)
2(1 − Nw∗)(rmax − rmin
The form of Eq. (1) does not allow one to obtain a simple analytical form for St50 (Stokes number
with 50% penetration), it must be calculated by numerical iteration.
In this study an axial %ow cyclone operating at low pressure conditions (¡ 20 torrs) is designed
and tested. The cyclone is expected to have submicron cuto< diameter and less particle bounce prob-
lems than impactors. In the following, &rst the theoretical collection e;ciency and cuto< aerodynamic
diameter of the axial %ow cyclone is derived under low pressure condition. Then the experimental
data obtained by using monodisperse oleic acid particles are used to validate the theoretical results.
Finally, to match with the data, the need to adjust the cuto< aerodynamic diameter with the %ow
Reynolds number is demonstrated.

2. Theoretical method

In this study, the particle radial velocity and hence the collection e;ciency is calculated theoret-
ically based on the gas volumetric %ow rate, properties of the cyclone, carrying gas and particles.
The transit time of an aerosol particle in the vane section, tv , is given by the ratio of the e<ective
volume in the section (Vol) to the gas volumetric %ow rate (Q). The e<ective volume in the vane,
Vol, is
2 2
Vol = (rmax − rmin )n(B − Nw); (3)
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1109

where the other variables were described in the last section. Therefore the particle (or gas) transit
time in the vane section, tv , is
2 2
(rmax − rmin )n(B − Nw)
tv = : (4)
Q
The tangential gas velocity in the vane section, vt , can be calculated as
2 rmin nN 2rmin QN
vt = = 2 2
(5)
tv (rmax − rmin )(B − Nw)
Due to centrifugal force, vt2 =r, the particle will move in the radial direction. Neglecting the
transient state of particle motion, the steady state particle radial velocity, vr , is
2
dr vt2 4rmin Q2 N 2
vr = = = 2
: (6)
dt r r(rmax
2 − rmin )2 (B − Nw)2
Assuming that the particle (or the vortex) continues its spiral motion in the body section downstream
of the vane such that it makes additional turns n( − 1), where  is a &tting constant which gives
the best agreement between the theoretical particle collection e;ciencies and the published data in
the literature. Adding the number of turns in the vane section, n, the total number of turns of the
particle is n in the cyclone. Rearranging Eq. (6) and integrating, the new particle radial position,
r, at the end of spiral motion can be calculated based on the particle’s initial radial position, ri , at
the inlet of the cyclone as
2
8nQrmin N2
r 2 − ri2 = 2
: (7)
(rmax
2 − rmin )(B − Nw)
Assuming plug %ow and uniform particle concentration at the inlet of the cyclone, the particle
collection e;ciency (or 1 − Pen), , can be calculated as
r 2 − ri2 2
8nQrmin N2 1
= 2
= 2
= 4nSt ; (8)
rmax − rmin
2 (rmax − rmin ) (B − Nw)
2 2 1 + rmax =rmin
where St, the Stoke number, is de&ned as St = vt =(rmax − rmin ). If  = 1, Eq. (8) is similar to
Eq. (12) in Maynard (2000) for calculation, the penetration through the vanes of the cyclone.
Setting  = 0:5 in Eq. (8), the theoretical cuto< aerodynamic diameter, dpa50 , can be calculated as

2 2
9 (rmax − rmin )2 (B − Nw)
dpa50 = 2
; (9)
8nQrmin N 2 C
where the slip correction factor C is calculated as (Hinds, 1999)
  
 dp
C =1+ 2:34 + 1:05 exp −0:39 : (10)
dp 
At high vacuum, Eq. (10) is found to give the same friction coe;cient as the molecular drag
formula described in Rader and Geller (1998). Furthermore at high vacuum, C can be shown to be
directly proportional to  and indirectly proportional to dp as
1:695
C= : (11)
dp =2
1110 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

This equation is accurate at high vacuum condition. For example, the di<erence in Eqs. (10) and
(11) is less than 1% for particles smaller than 300 nm at 10 torr. Since the relationship between 
and the pressure at the inlet of the cyclone, Pcyc , at the isothermal condition is
P760 0
= ; (12)
Pcyc
then the slip correction factor can be re-written as
1:695P760 0
C= : (13)
Pcyc dp =2
At a &xed standard %ow rate Q0 , the actual %ow rate Q = Q0 (P760 =Pcyc ). Therefore, substituting
Eq. (13) into Eq. (9), the cuto< aerodynamic diameter can be written as
 
Pcyc 2 (rmax
2 2
− rmin )2 (B − Nw)
dpa50 = 0:106 2
: (14)
P760 "po nQ0 rmin N 2 0
That is, at high vacuum, the cuto< aerodynamic diameter is proportional to the pressure at the inlet
of the cyclone to the second power given that the standard %ow rate is &xed.
However, at low pressure, the %ow Reynolds number is usually very low and the cuto< aerody-
namic diameter will be underestimated. For the tangential %ow cyclone, Kao and Tsai (2001) found
that the dimensionless cuto< aerodynamic diameter, 50 , is a unique function of the %ow Reynolds
number, Ref
, as follows:
ln 50 = −0:72 ln Ref
− 3:46; (15)
where
√
Cdpa50
50 = ; (16)
D
"(D − De )Ui
Ref
= ; (17)
2
where D and De are the cyclone and exit tube diameter, respectively. Similar correlation was found
in Moore and McFarland (1996). It is expected that for the axial %ow cyclone, the theoretical cuto<
aerodynamic diameter, Eq. (9), has to be adjusted also by using the %ow Reynolds number.

3. Experimental method

The experimental system is shown in Fig. 1. Monodisperse oleic acid particles of 35–460 nm
("p = 894 kg=m3 ) were generated by the atomization and electrostatic classi&cation technique. Oleic
acid particles were &rst generated by atomizing 0.01 to 5% (v/v) oleic acid containing alcohol, dried
by a silica gel drier, and &nally classi&ed by the TSI model 3080 electrostatic classi&er with the TSI
model 3081 long DMA. To obtain particles less than 100 nm, dried particles were further evaporated
in a furnace and subsequently grown by condensation before electrostatic classi&cation.
The axial %ow cyclone was designed with rmax = 1:5 cm; rmin = 1:0 cm; B = 0:5 cm (design 1)
or B = 1:0 cm (design 2). The e<ective cyclone body length (or the distance between the tip of the
outlet tube and the bottom of the vane spindle), L, is 2:0 cm, and the inside diameter of the outlet
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1111

TSI 3081
LDMA
TSI 3080
Electrostatic
Classifier

TSI CPC

clean orifice
air
furnace
ni

diffusion
dryer
pressure
transducer #1

clean
air pressure
transducer #2 low
atmoizer pressure
cyclone
power supply and
readout

by-
electromet
elect meter pass
line
vacuum
vacuum nco nto
pump

Fig. 1. Present experimental setup.

tube is 0:776 cm in this study. Two di<erent vanes were designed as shown in Fig. 2. Design 1 has
one vane which makes 3 complete turns; and design 2 has 3 vanes each makes half of a turn. Two
di<erent %ow rates, 0.455 and 1:0 slpm were tested. Inlet cyclone pressure tested was 13, 14, 15,
16, 17 or 20 torr at 1 slpm, or 6, 8, 10, 13 torr at 0:455 slpm.
The steadiness of the concentration of classi&ed monodisperse particles was monitored by the
TSI model 3025A CPC (condensation particle counter) before the aerosol entered a critical ori&ce
(O’Keefe Controls Co., V-14, 1 slpm; or V-9, 0:455 slpm). The TSI model 3068 electrometer or a
home-made Faraday cage with the Keithley model 6514 electrometer was used to measure aerosol
current after the cyclone, and after the by-pass tube for calculating the particle collection e;ciency.
The vacuum pump (SV16 10981, SOGEVAC, France, nominal pumping speed: 16 m3 =h) was used
to create the %ow and the pressure inside the cyclone was measured by the MKS Baratron type,
626A13TAE (1000 torr max:) or MKS Baratron type, 626A11TAE (10 torr max.)
Since the collection e;ciency is determined by current measurement, the penetration of a singly
charged particle with the diameter of dp1 will be in%uenced by the penetration of particles with k
multiple charges (diameter dpk ), especially for particles greater than 100 nm. The electric currents,
1112 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

design 1 design 2

Fig. 2. Vanes of design 1 and 2.

Ic and It , were measured at the outlet of the cyclone and by-pass tube, respectively. The currents
are related to %ow rate Q, and inlet particle concentration ni (dp ) as
∞
Ic = Qe kni (dpk )Pen(dpk ) ; (18)
k=1
∞
It = Qe k ni (dpk )Pent (dpk ) ; (19)
k=1

where k is the number of charges, e is the elementary charge. The particle penetration of the cyclone,
Pen(dp ), and the particle penetration through the by-pass tube, Pent (dp ), are de&ned as

Pen(dp ) = noc (dp )=ni (dp ); (20)

Pent (dp ) = not (dp )=ni (dp ); (21)

where noc (dp ) and nto (dp ) are particle concentrations at the outlet of the cyclone and by-pass tube,
respectively.
It can be shown by simple algebraic manipulation of Eqs. (18)–(21) that Pen (dp1 ) is
∞ ∞
Ic kni (dpk )Pent (dpk ) kni (dpk )Pen(dpk )
Pen(dp1 ) = Pent (dp1 ) + − : (22)
It ni (dp1 ) ni (dp1 )
k=2 k=2

In this study, Pent (dpk ) was found to be close to that of the cyclone without vanes. Therefore
Pent (dpk ) was assumed to be 1.0. The above equation was used to correct for multiple charge
e<ect on the penetration of a single charge particle with diameter dp1 . When using Eq. (22), one
normally has to measure the penetration of the largest particles &rst, which is determined by the
cuto< diameter of the impactor at the DMA inlet. This penetration is known to be accurate since
there is no multiple charge e<ect for the largest particles entering DMA. Then the penetration of
smaller particles is measured and corrected for the multiple charge e<ect by Eq. (22).
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1113

4. Results and discussion

The theoretical cuto< aerodynamic diameter of the present study is &rst compared with the previous
experimental data in the literature. As shown in Fig. 3, at 1 atm the theoretical cuto< aerodynamic
diameters, Eq. (9), compare very well with the previous experimental data of Liu and Rubow (1984)
(symbol: circle), Weiss et al. (1987) (symbol: square), and Vaughan (1988) (symbol: triangle).
One exception is the experimental data point near 4:8 m which has a large deviation from the
present prediction. The experimental cuto< aerodynamic diameter ranges from 1.05 to 12:2 m. It
was found that best agreement was obtained if the number of turns in the cylindrical body after the
vane was assumed to be half of the number of vane turns. That is,  was assumed to be 1.5 in
Eq. (9).
The theoretical prediction by Maynard (2000), Eq. (1), also gives similar agreement, with larger
deviation for some data points of large particles.
The pressure at the inlet (before the vane) and outlet (after the vane) of the cyclone were measured
and shown in Fig. 4. Design 1 with one 3-turn vane has higher pressure drop than design 2 with
3 half turn vanes, due to lower air %ow velocity in the vane in the latter compared to that in the
former. At 1 slpm, the lowest pressure achieved is 13 torr at the cyclone inlet, and the pressure drop
(inlet pressure minus outlet pressure) is 4 and 2:9 torr for design 1 and 2, respectively. The pressure
drop decreases with an increasing inlet pressure. At the inlet pressure of 23 torr, the pressure drop
becomes 1.8 and 1:4 torr for design 1 and 2, respectively. Higher vacuum is achieved at lower
volumetric %ow rate. At 0:455 slpm, the highest vacuum achieved is 6 torr at the cyclone inlet, and
the pressure drop for design 1 and 2 is 3.3 and 2:32 torr, respectively. The pressure at the stagnation

16
filled symbols: present study
open symbols: Maynard (2000)

12
Aero. diameter, µm (predicted)

4 Weiss et al. (1987)

Liu and Rubow (1984)

Vaughan (1988)
0
0 4 8 12 16

Aero. diameter, µm (experiment)

Fig. 3. Comparing the cuto< aerodynamic diameter of the present theory with previous experimental data.
1114 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

6
design 1, 1 slpm
design 2, 1 slpm

pressure drop cross cyclone, Torr

5
design 1, 0.455 slpm
design 2, 0.455 slpm
4

0
0 5 10 15 20 25
pressure before cyclone, Torr

Fig. 4. Pressure in the cyclone.

point should be used to calculate the e;ciency (Biswas & Flagan, 1984). As this cannot be readily
obtained, in this study, the average of the inlet and outlet pressure was used to calculate particle
collection e;ciency.
The experimental particle collection e;ciency curves are shown in Figs. 5(a)–(d) for two %ow
rates and two vane designs. In general, a lower pressure results in a higher particle collection
e;ciency, and design 1 has higher collection e;ciency than design 2. For design 1, at the %ow rate
of 0:455 slpm and the inlet pressure of 6 torr, the collection e;ciency increases from 32% to nearly
100%, as the particle diameter of oleic acid is increased from 35 to 100 nm (Fig. 5(a)). The cuto<
diameter is 48:7 m. As the inlet pressure is increased from 6 to 8, 10 or 13 torr, the collection
e;ciency drops considerably and the curves look less like S-shaped. The cuto< diameter is increased
to 153.4 and 330 nm for the inlet pressure of 8 and 10 torr, respectively.
For design 2, at the same %ow rate of 0:455 slpm, and the same inlet pressure as design 1,
the corresponding collection e;ciency is lower (Fig. 5(b)). Lower pressure drop of design 2 is
mainly responsible for the lower particle collection e;ciency, which agrees with the observation of
traditional cyclones. The cuto< diameter of design 2 is much larger than design 1, which is 107.6
and 343:2 nm for the inlet pressure of 6 and 8 torr, respectively.
As the %ow rate is increased to 1 slpm, the minimum inlet pressure, 13 torr, is higher than that of
0:455 slpm. The particle collection e;ciency curves shown in Figs. 5(c) and (d) follow the similar
trend as in the previous &gures. That is at the same inlet pressure, design 1 has higher collection
e;ciency than design 2, and inlet pressure has a very important e<ect on particle collection e;ciency.
When the inlet pressure is increased from 13 to 20 torr, the cuto< diameter increases from 120.5 to
411 nm for design 1, and from 145.5 to 600 nm for design 2.
When compared to the present theory, Eqs. (8) and (9), it is found that experimental collection
e;ciency is much lower, and the experimental cuto< aerodynamic diameter is much larger for
all particle diameters. This is due to the Reynolds number e<ect (Kao & Tsai, 2001; Moore &
McFarland, 1993) as pointed out in the previous section. Thus, the cuto< aerodynamic diameter has
to be correlated with the %ow Reynolds number, in a form similar to the right-hand side of Eq. (15)
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1115

100 100

80 6 torr 80
8 torr
Collection efficiency, %

Collection efficiency, %
6 torr

60 60

8 torr
40 10 torr 40

20 20 10 torr
13 torr

0 0
0 100 200 300 400 0 100 200 300 400
(a) Particle diameter, nm (b) Particle diameter, nm

100 100
13 torr 13 torr
14 torr
80 80 14 torr
15 torr
16 torr 15 torr
Collection efficiency, %

Collection efficiency, %

17 torr
16 torr
60 60

17 torr
20 torr
40 40
20 torr

20 20

0 0
0 100 200 300 400 500 0 100 200 300 400 500
(c) Particle diameter, nm (d) Particle diameter, nm

Fig. 5. Particle collection e;ciency versus particle diameter (a) Q = 0:455 slpm, design 1; (b) Q = 0:455 slpm, design 2;
(c) Q = 1 slpm, design 1; (d) Q = 1 slpm, design 2.

but with di<erent empirical constants. It is found that the following cuto< aerodynamic diameter,
dpa50 , adjusted empirically from Eq. (9) by the %ow Reynolds number gives good agreement with
the experimental data:

2 2
9 (rmax − rmin )2 (B − Nw)
dpa50 = 2
× exp(−0:276 ln(Ref ) + 1:18); (23)
8nQrmin N 2C
1116 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

800
design 1 Q = 0.455 slpm
exp. data
design 2

cutoff aerodynamic diameter, µm

design 1
theory
600 design 2
Ref adjusted

400

200
original

0
6 8 10 12 14
(a) pressure before cyclone, torr

600
design 1 Q = 1 slpm
design 2 exp. data
design 1
cutoff aerodynamic diameter, µm

theory
design 2

400 Ref adjusted

200
original

0
4 8 12 16 20
(b) pressure before cyclone, torr

Fig. 6. Comparison of theoretical cuto< aerodynamic diameter with experimental data. (a) Q =0:455 slpm; (b) Q =1 slpm.

where the %ow Reynolds number is now de&ned as

"(rmax − rmin )Ua
Ref = (24)

and Ua is the average axial velocity through the cyclone vane.

Fig. 6 shows the comparison of theoretical cuto< aerodynamic diameter with the experimental
data. As can be seen, the original theory, Eq. (9), underestimates the cuto< aerodynamic diameter
substantially at low %ow Reynolds number of Ref = 6:4 for 0:455 slpm, and Ref = 14:1 for 1 slpm.
At Ref = 6:4 and 14.1, the experimental cuto< aerodynamic diameters are 2.7 and 3.4 times the
theoretical values. After correlating with the %ow Reynolds number, it is found that Eq. (23) is
 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118 1117

able to predict the cuto< aerodynamic diameter at di<erent inlet pressures, %ow rates and cyclone
designs.

5. Conclusions

An axial %ow cyclone with an inner diameter of 3:0 cm was designed and tested to remove &ne
particles at low pressure (6–23 torr) and low %ow rate (0.455 and 1:0 slpm) conditions. Pressure
inside the cyclone was found to have a considerable e<ect on the particle collection e;ciency. At
lower pressure, the particle collection e;ciency was also found to be higher. The cyclone with
higher pressure drop has higher collection e;ciency. At a &xed inlet pressure, higher pressure drop
of design 1 with one 3-turn vane was found to have a higher collection e;ciency than design 2
with three half-turn vanes. At lowest achievable pressure of 6 torr at the %ow rate of 0:455 slpm in
this study, the axial %ow cyclone of design 2 is able to remove particles below 100 nm e;ciently,
and the cuto< aerodynamic diameter is 43:3 nm.
Theoretical prediction of cuto< aerodynamic diameter was found to agree with published experi-
mental data in the literature at ambient conditions. However, at vacuum and low %ow rate conditions,
experimental collection e;ciency was found to be much lower than theoretical e;ciency. Based on
the experimental data, a semi-empirical equation incorporating the %ow Reynolds number was de-
veloped to predict the cuto< aerodynamic diameter with good accuracy at di<erent inlet pressures,
%ow rates and cyclone designs.
This low %ow rate axial %ow cyclone is applicable for &ne particle removal in the low pressure
exhaust gas of reaction chambers, such as in the semiconductor and optoelectronic industries. In the
future, it is worthwhile to study the e<ect of particle loading in the cyclone and particle material on
the particle collection e;ciency.

Acknowledgements

Authors would like to thank for the &nancial support of this project by the Industrial Technology
Research Institute (ITRI) in Taiwan in 2003, and Taiwan National Science Council
(NSC 92-2211-E-009-016).

References

Barth, W. (1956). Design and layout of cyclone separator on the basis of new investigations. Brennst Warme Kraft,
8, 1–9.
Biswas, P., & Flagan, R. C. (1984). High-velocity inertial impactors. Environmental Science and Technology, 18,
611–616.
Dietz, P. W. (1981). Collection e;ciency of cyclone separtors. A.I.Ch.E. Journal, 27, 888–892.
Dirgo, J., & Leith, D. (1985). Cyclone collection e;ciency: Comparison of experimental results with theoretical predictions.
Aerosol Science and Technology, 4, 401–415.
Hinds, W. C. (1999). Aerosol technology (2nd ed.) (p. 127). New York: Wiley.
Iozia, D. L., & Leith, D. (1989). E<ect of cyclone dimensions on gas %ow pattern and collection e;ciency. Aerosol
Science and Technology, 10, 491–500.
1118 C.-J. Tsai et al. / Aerosol Science 35 (2004) 1105 – 1118

Iozia, D. L., & Leith, D. (1990). The logistic function and cyclone fractional e;ciency. Aerosol Science and Technology,
12, 598–606.
Kao, K. Y., & Tsai, C. J. (2001). On the theory of particle collection e;ciency of cyclones. Journal of Aerosol and Air
Quality, 1, 47–56.
Kenny, L. C., Gussman, R., & Meyer, M. (2000). Development of a sharp-cut cyclone for ambient aerosol monitoring
applications. Aerosol Science and Technology, 32, 338–358.
Lapple, C. E. (1950). Gravity and centrifugal separation. Industrial Hygiene Quarterly, 11, 40–48.
Leith, D., & Licht, W. (1972). The collection e;ciency of cyclone type particle collectors: a new theoretical approach.
A.I.Ch.E. Symposium Series, 126, 196–206.
Li, E., & Wang, Y. (1989). A new collection theory of cyclone separators. A.I.Ch.E. Journal, 35, 666–669.
Liu, B. Y. H., & Rubow, K. L. (1984). A new axial %ow cascade cyclone for size characterization of airborne particulate
matter. In: Liu, B. Y. H., Pui, D. Y., & Fissan, H. J. (Eds.), Aerosols (pp. 115–118). Amsterdam: Elsevier.
Maynard, A. D. (2000). A simple mode of axial %ow cyclone performance under laminar %ow conditions. Journal of
Aerosol Science, 31, 156–167.
Moore, M. E., & McFarland, A. R. (1993). Performance modeling of single-inlet aerosol sampling cyclone. Environmental
Science and Technology, 27, 1842–1848.
Moore, M. E., & McFarland, A. R. (1996). Design methodology for multiple inlet cyclones. Environmental Science and
Technology, 30, 271–276.
Nieuwstadt, F. T. M., & Dirkzwager, M. (1995). A %uid mechanics model for an axial cyclone separator.
Industrial Engineering and Chemical Research, 34, 3399–3404.
Peters, T. M., & Vanderpool, R. W. (1996). Modi&cation and evaluation of the WINS impactor. R. T. I. Report
No. 6360-011.
Rader, D. J., & Geller, A. S. (1998). Showerhead-enhanced inertial particle deposition in parallel-plate reactors. Aerosol
Science and Technology, 28, 105–132.
Tsai, C. J., & Cheng, Y. H. (1995). Solid particle collection characteristics on impaction surfaces of di<erent designs.
Aerosol Science and Technology, 23, 96–106.
Tsai, C. J., Shiau, H. G., Lin, K. C., & Shih, T. S. (1999a). E<ect of deposited particles and particle charge on the
penetration of small sampling cyclones. Journal of Aerosol Science, 30, 313–323.
Tsai, C. J., Shiau, H. G., & Shih, T. S. (1999b). Field study of the accuracy of two respirable sampling cyclones. Aerosol
Science and Technology, 31, 463–472.
USEPA, 1997, Ambient Air Monitoring Reference and Equivalent Methods. United States Environmental Protection
Agency, Federal Register 40 CFR Parts 50.
Vaughan, N. P. (1988). Construction and testing of an axial %ow cyclone pre-separator. Journal of Aerosol Science,
19(3), 295–305.
Weiss, Z., Martinec, P., & Vitek, J. (1987). Vlastnosti Dulnibo Prachu A Zaklady Protiprasne Techniky. SNTL: Prague.