Aerosol Science 38 (2007) 1220–1229www.elsevier.com/locate/jaerosci

Filtration of nanosized particles with different shapeon oil coated fibres

Lucija Boskovic, Igor E. Agranovski∗, Roger D. BraddockGriffith School of Engineering, Griffith University, Brisbane, QLD 4111, Australia

Received 10 June 2007; received in revised form 2 September 2007; accepted 20 September 2007

Abstract

In our previous work it has been shown that perfectly spherical polystyrene latex (PSL) particles have higher filtration efficiencycompared to cubic magnesium oxide (MgO) particles of the same electrical mobility as PSL particles. This disparity was ascribedto the different nature of motion of the spherical and cubic particles along the fibre surface, following the initial collision. Aftertouching the fibre surface and before coming to rest, the spherical particles could either slide or roll compared to the cubic ones,which could slide or tumble. During tumbling, the area of contact between the particle and the fibre changes significantly, thusaffecting the bounce probability, whilst for the spheres, the area of contact remains the same for any point of particle trajectory.In this project, the polypropylene filter was coated with a thin layer of mineral oil that was used to absorb the energy and, respectively,to minimize particle motion along the fibre after collision. The filtration efficiency of spherical PSL, and cubic MgO particles wasmeasured in the size range of 50–300 nm, for filtration velocity of 10 and 20 cm/s. It was found that, regardless of shape, bothparticle types have very similar filtration efficiency. The theoretical predictions are in good agreement with our experimental results.Therefore, the conclusion can be drawn that the oil coating minimizes the amount of particle motion along the fibre after initialcollision, making results for all particle shapes similar.� 2007 Elsevier Ltd. All rights reserved.

Keywords: Filter; Particle removal; Oil coated fibres; Particle–fibre interaction

1. Introduction

The capture of aerosol particles by surfaces within a filter is a process of fundamental importance in aerosol science.Despite this fundamental importance, a complete study of the capture of particles, including a detailed account ofthe particle-surface collision processes, has not been given. This lack of information is due to the fact that theoreticaltreatments of the capture of aerosol particles by surfaces are based on the assumption that a particle is always capturedupon touching a surface. The possibility that a particle may bounce on contact with the surface of the filter, has largelybeen ignored.

The possibility of the particle bouncing depends on its composition, its shape, its velocity and the type of impactionsurface (Reischl & John, 1978). When a solid particle contacts a surface at low velocity the particle loses its kineticenergy by deforming itself and surface. At higher velocities, part of the kinetic energy is dissipated in the deformation

∗ Corresponding author. Tel.: +61 7 3735 7923; fax: +61 7 3735 7459.E-mail address: i.agranovski@griffith.edu.au (I.E. Agranovski).

0021-8502/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.doi:10.1016/j.jaerosci.2007.09.003

L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229 1221

process (plastic deformation), and part is converted elastically to kinetic energy of rebound. If the rebound energyexceeds the adhesion energy—the energy required to overcome the adhesive forces—a particle will bounce away fromthe surface. The harder the materials comprising the particle and surface, the larger the particle, or the greater itsvelocity, the greater the likelihood of the particle bouncing from the surface (Hinds, 1999).

There are two classical approaches to describe particle bounce. The first defines a critical velocity Vc, above whichbounce will occur (Cheng & Yeh, 1979; Hiller & Loeffler, 1980),

Vc = �

dp= 1

dp

(1 − e2pl)

e2pl

A

�x2(6ppl�)1/2 , (1)

where � is a constant for a particular impaction surface, dp is the particle diameter, epl is the coefficient of restitution(for plastic deformation only), A is the Hamaker constant, ppl is the microscopic yield pressure, � is the particle densityand x is the distance between the particle surface and its adjacent surface. Hamaker constants are given in the literaturefor a limited number of elements and compounds (Tsai, Pui, & Liu, 1991).

The other method involves the kinetic energy (KEb) required for bounce to occur when a particle (dp) collides witha surface (Dahneke, 1971), and the magnitude of kinetic energy can be calculated as

KEb = dpA(1 − e2)

2xe2 , (2)

where e is the coefficient of restitution (plastic and elastic deformation), which is equal to the ratio of the reboundvelocity to the approach velocity. The value of e is reported to range from 0.73 to 0.81, although these values werederived using hard impaction surfaces (Wall, John, Wang, & Goren, 1990). It is reported that A and e ought to bedetermined experimentally because it is very hard to determine them theoretically.

Boskovic et al. (2005) showed that the effect of particle motion along the fibre should not be ignored. Their exper-imental results showed that polystyrene latex (PSL) particles of spherical shape in the size range of 50–300 nm areremoved from the gas carrier with significantly higher efficiencies compared to cubic MgO particles in the same sizerange, even if their densities vary by a factor of 3.6. After excluding inertial mechanism as a possible factor responsiblefor such a difference, it was concluded that its main cause is likely to be the nature of motion of cubes and spheresalong the fibre surface; the spherical particles could either slide or roll whilst the cubic ones slide or tumble.

The nature of the contact surface is also important in the bouncing processes. There has been some research into theeffect of coating filter fibres with a liquid, and its influence on particle bounce. Walkenhorst (1974) examined the effectof coating model wire filters with vegetable oils, mineral oils and Vaseline. He reported that the former two substancesincreased filtration efficiency while the latter did not. Although oil coated filters are suitable for laboratory processes,using such liquids in industry would not be advantageous. Agranovski and Braddock (1998) stated that liquid filmscoating the fibre could reduce the likelihood of particle bounce, although this has yet to be proven experimentally.Agranovski, Braddock, and Myojo (1999), developed a process in which the filter is coated with a thin layer of water,allowing collection of aerosols onto the water film rather than directly by the solid fibre surface. These technologies areindustrially applicable, as quantities of water for filter irrigation are more readily available and recyclable in industrythan oils and greases.

Mullins, Agranovski, and Braddock (2003) examined the effect of particle bounce in fibrous filter systems bycomparing the filtration efficiencies of solid PSL particles with liquid diethylhexyl sebacate (DEHS) particles of thesame size and shape factor under identical filtration conditions. The experiments were conducted firstly in dry filtersystems, then in wet systems. They concluded that for the wet filtration regime, the efficiencies do not significantlydiffer with changing particle type, but bouncing was greatly reduced in wet filter systems. The water film acts to inhibitbounce either by aiding energy dissipation or preventing the particle from being repelled from the fibre. Factors suchas surface chemistry and surface roughness of the fibres could have a significant effect on wetting and wettability.However, most models assume a homogeneous fibre substrate, since this is the least complex case (McHale, Newton,& Carroll, 2001).

The aims of this paper is to describe experiments investigating filtration of particles of different shape at differentfiltration velocities on the filter fibres coated with a thin layer of mineral oil. Coating surfaces with oil or grease increasesthe adhesion energy, the deformation, and the dissipative energy and greatly reduces the processes of particle motionalong the fibre after initial collision and bounce (Hinds, 1999) decreasing the effects of shape on the bouncing process.

1222 L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229

The results of such investigation are important for better understanding of causes of substantial difference in filtrationof particles with cubic and spherical shape, reported in our previous paper (Boskovic et al., 2005). It could give aclear answer to the question: does this difference relate to the primarily filtration mechanisms or secondary effects ofbouncing and re-entrainment. The mineral oil was selected as it gives a uniform layer on the fibre surface and does notreadily run down the fibre, as water will do, over the period of the experiment.

2. Experimental set up and measurement procedure

2.1. Test aerosol production and experimental apparatus

Two types of nanoparticles were used in this research: spherical PSL(density 1050 kg/m3), and cubic magnesiumoxide (MgO) (density 3600 kg/m3). The PSL particles were obtained in five monodisperse sizes. The sizes were50, 70, 100, 150 and 300 nanometers in diameter as supplied by Bangs Laboratories (Fisher, IN, USA). The latexsuspension was diluted with nanopure deionized water with a resistivity � > 18.0 M� and sonicated by an ultra sonicbath for 20 min to prepare the suspension for subsequent aerosolization by a 3-jet Collision nebulizer (BGI Inc.,Waltham, MA).

The production of MgO by direct metal combustion in an oxygen rich environment was undertaken according toAltman, Agranovski, and Choi (2004). The MgO nanoparticles were collected on a substrate, which was located at adistance of 10 mm above the flame, then brushed into nanopure deionized water and sonicated for 20 min before beingaerosolized by the 3-jet Collision nebulizer (BGI Inc., Waltham, MA).

The size distribution of the PSL and MgO particles was measured using a differential mobility analyzer (DMA)(Model 3080, TSI Inc., St. Paul, MN). Before reaching a DMA system consisting of an impactor, charger and classifier,the total particle number concentration was approximately 104 particles/cm3.

Fig. 1 shows a schematic of the experimental set up that was used in this project and is described in details inBoskovic et al. (2005). Before reaching the nebulizer, the air was passed through three HEPA filters arranged in seriesto remove all extraneous aerosols from the air input stream. The test particle suspensions, prepared as described above,were placed into the nebulizer and aerosolized to provide the test aerosol. After passing through the generator, theair was mixed in the mixing tube (diameter 15 mm and length 30 cm) with another portion of the dry HEPA filteredair to ensure effective drying of moisture from the particle surface. The aerosol was then collected isokinetically(Clarke, 1998) through the intake pipe located at a distance of 20 cm from the mixing tube inlet (the distance waschosen to provide complete mixing of the air and particle streams Clarke, 1998). The aerosol was then passed throughan electrostatic charge neutralizer (10-mCi85Kr, model 3012, TSI Inc., St. Paul, MN) to neutralize the particles beforetheir entry to the DMA column.

After passing through the generator and the charge neutralizer, the test aerosol was moved through the DMA column,to separate monodisperse particles of the required size for subsequent passage through the test filter. A special filterholder was used instead of a standard connection of DMA and CPC, which is normally used to measure the particleconcentration. The filter assemble consisted of two identical parallel air lines with switching valves capable to passthe aerosol stream along either one of them keeping the unused line fully inaccessible for the aerosol stream. One lineof the assemble contained the filter holder with the tested filter whilst the other was empty (to ensure the identicalgeometry of air passage, this line was also equipped with the same filter holder with no filter media installed) and servedas the bypass line. The number of particles after passing through the filter represented the downstream concentration,whilst the number of particles flowing through the bypass line, corresponded to the (initial) upstream concentration ofaerosols.

The filter efficiency ET was then calculated according to the classic equation given by Brown (1993),

ET =(

1 − CA

CB

)× 100%, (3)

where CA and CB are concentrations of the particles after passing through the filter line and the bypass line, respectively.The emission of oil from coated filter was tested to ensure that no extraneous particles would influence the experi-

mental results. It was found that the oil particle production does not exist, as was verified by “zero” emission from thecoated filter in absence of test aerosol supply.

L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229 1223

Air

Particle Intake

Mixing Tube

Air

Air

Nebulizer

Flow Meter HEPA Filter

HEPA Filter

CPC

Neutraliser

DMA

HEPA Filter

Bypass

Filter

Fig. 1. Experimental set up.

2.2. Experimental procedure

A polypropylene filter with a thickness of 3 mm, 16% packing density, and fibre size of 12 �m was used in experiments(see more details in Agranovski, Braddock, & Myojo, 2001 (Filter #2)). An oil coating was applied by submergingthe filter in mineral oil until a uniform color of the filter was achieved. After that, the media was squeezed to removeall excess oil from the filter. The coating and distribution of oil was then checked by a Zeiss (Germany) Standard 25polarizing microscope with ×10 objective lens and a graduated eyepiece. To verify that the squeezing procedure did notcause any structural changes of the filter, firstly, the thickness of the media was measured before and after squeezing(no changes of this parameter were observed). Secondly, the pressure drop across the filters was continuously monitoredthroughout all experiments and used to confirm that there were no microscopic structural changes of the filter occurredas the result of squeezing. The pressure drop values for coated and uncoated filters were in perfect agreement withcorresponding theoretically calculated (Hinds, 1999) numbers.

A FEI Quanta 200F Environmental SEM (ESEM) was used to obtain a precise measurement of the fibre diameterof the oil coated fibres. The main advantage of using the ESEM is based on a fact that it retains all of the performanceadvantages of a conventional SEM, but removes the high vacuum constraint on the sample environment. Wet, oily, dirty,non-conductive samples may be examined in their natural state without modification or preparation. A single pressurelimiting is used in a conventional SEM, while an ESEM uses multiple pressure limiting apertures to separate the samplechamber from the column. The column is still operating at high vacuum, but the chamber may sustain high pressures.

1224 L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229

At the beginning of each experiment, a fresh filter was coated with oil and hermetically sealed in the filter holder.The experimental procedure, fully described in Boskovic et al. (2005), was then followed to obtain the filter efficiencyfor the range of particle sizes used. In brief, the whole system was flushed with filtered air until no particles remainedin the system. After that, the airline was closed and the aerosol intake line was opened to allow aerosolized PSLparticles of the selected size (only one particular size was used in each experimental run) to pass firstly through theDMA column, and subsequently through the CPC. The magnitude of the voltage, which corresponds to the maximumnumber concentration, was recorded and the DMA was tuned to produce monodisperse particles of the requiredsize corresponding to the selected voltage. After measuring the particle concentration in the bypass line, the sameexperimental procedure was repeated for the flow of the selected particle size, through the filter line at one of thetwo different face velocities of 10 and 20 cm/s. Considering a fixed (critical orifice controlled) flowrate required forDMA/CPC operation, the required velocity on the filter face was achieved by placing the orifice of predetermineddiameter in the filter holder. All flow rates were monitored and controlled by flow meters equipped with adjustingvalves.

The same procedure was followed for the initially polydisperse MgO particles. After aerosolization, they weredesiccated by dry diluting air (20 ◦C and ∼ 10%RH) and then passed through the DMA column at the voltagecorresponding to a PSL particle of the same mobility. After that, monodisperse MgO particles with the aerodynamicsize identical to particular PSL size were passed through either the bypass or coated filter line before reaching the CPCto obtain CA and CB for subsequent filter efficiency calculations.

3. Results and discussion

The main scope of this project is further investigation of the process of nanoparticle–fibre interaction during filtrationprocesses.As was found in our previous research (Boskovic et al., 2005), the filter efficiency is substantially different forremoval of nanoparticles with cubic and spherical shape of the same aerodynamic size. Such difference was explainedby dissimilar nature of the particle motion along the fibre right after initial collision. It was suggested that in caseof spherical shape, particle either slide or roll, in contrast with cubic particles, which are slide or tumble. In caseof tumbling, the geometry of a contact spot varies dramatically, thus affecting the bounce probability, whilst for thespheres, the area of contact remains the same for any point of particle trajectory. It was suggested that using oil coatingcould significantly increase the adhesion forces at the particle-film interface making the possibility of any particlemotion along the fibre very limited. Then, any difference between the removal efficiency results for spherical and cubicparticles could only be related to primarily filtration mechanisms (probability of collision with the fibre) leaving thenature of particle motion along the fibre and corresponding re-entrainment to the gas carrier, due to negligibly smallcontribution, outside the consideration.

Some similar approach was used in one of our previous projects (Mullins et al., 2003) targeted to identify thedifference in bouncing of solid and liquid aerosol particles from a fibre. In that paper, we compared the efficiency ofdry and water coated fibres on removal of relatively large liquid and solid particles with substantial inertia. It was foundthat solid particles have much higher bouncing capability compared to the liquid ones when were being removed ondry fibres. However, this difference has fully disappeared when the fibres were coated with water films making thefilter performance characteristics for both types of particle identical.

In the current paper, for the particle size range of interest, the inertial effect is not substantial to cause clearly identifiedinertial bouncing corresponded to the larger aerosol sizes used in (Mullins et al., 2003). However, as discussed before(Boskovic et al., 2005), it could contribute to some particle motion of different nature along the fibre after initialcollision justifying alteration in filter performance on removal of nanoparticles of different shape. Oil coating of thefibres could give a clear answer to the question: does the difference in filter performance for removal of particles ofalternative shape occur due to some difference in efficiency of primarily filtration mechanisms or due to more efficientparticle re-entrainment for the cubes compared to spheres.

Firstly, consider the kinetic energy of the spherical PSL particle required for bounce to occur on impact with a surface.The minimum particle kinetic energy for bounce, according to Dahneke (1971), was estimated by Eq. (2) for the sphericalPSL particles for our experimental conditions. The values for the coefficient of restitution, the Hamaker constant andan average distance x, used in these calculations, were respectively: e = 0.73 (Hinds, 1999); A = 6.37 ∗ 10−20 J(Tsai et al., 1991); and x = 0.4 nm (Hinds, 1999).

L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229 1225

Fig. 2. Kinetic energy of particle required for bounce according to Cheng andYeh (1979), Dahneke (1971) and particle kinetic energy before collision.

Alternatively, the minimum velocity at contact for particle bounce to occur was obtained by Eq. (1) and usedto estimate corresponding minimal kinetic energy of spherical PSL particles required for their bounce using � =2 ∗ 10−6 m2/s (Cheng & Yeh, 1979).

The above theoretical values have been compared with the kinetic energy of PSL particles just before collision withthe fibre for both experimental filter face velocities (10 and 20 cm/s). The results of theoretical calculations accordingto both strategies, along with the values of the PSL particle kinetic energy are shown in Fig. 2. As is seen, the calculatedkinetic energy of PSL particles is much smaller compared to the theoretical values required for bounce to occur forboth models used. Unfortunately, there are no corresponding theoretical results for the minimum kinetic energy, orminimum velocity at impact, available for non-spherical particles. However, the densities of PSL and MgO particlesdiffer only by a factor of 3.6 and the minimal kinetic energy required for bouncing is 3.5 orders of magnitude higherthen the calculated results related to the particles used, which suggests that the kinetic energies of all particles arewell below the thresholds of bounce. This means that the particle bouncing could be excluded from our consideration,however some possibility that the particles can tumble or roll remains.

Obviously, the efficiency and uniformness of the fibre coating is the crucial issue for our research procedure.As discussed before, we used ESEM for precise monitoring of this procedure. Fig. 3 shows an ESEM image ofuncoated fibres. It should be noted that the surface of the fibres is not smooth and contains striations possibly fromthe manufacturing process. The average size of more than 50 measured fibres was found to be 12 �m, with a standarddeviation of 1.2.

Fig. 4 shows an ESEM image of oil-coated fibres. Notice that the surface of the fibres looks much smoother,presumably due to surface tension effects in the oil film. The presence of oil is the most obvious at the intersection ofthe fibres where oil is forming bridges between the fibres, as indicated by arrows in Fig. 4. To identify parameters offilter coated by oil, their diameters were measured from the ESEM images. More than 50 measurements were madeand it was found that the coated fibre diameter was 12.9 �m with the standard deviation of 1.4. The packing density ofthe coated filter was correspondingly increased to 18.4%.

Fig. 5 shows the measured efficiency of the oil coated fibrous filter, for all particles investigated, at the filtrationface velocity of 10 cm/s. As it can be seen from the figure, the curves are all close together and the standard deviationsoverlap, showing an insignificant discrepancy in the results compared to previous findings (Boskovic et al., 2005). Inour previous research, the filtration efficiency of a dry filter used for removal of the spherical particles was significantlyhigher compared to the filtration efficiency of cubic particles. Current results obtained for coated fibres show nodifference in filter performance for removal of particles with different shape. Considering the fact that our previousresults were obtained for a different filter media, the results of experiments utilizing currently used filter with no fibre

1226 L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229

Fig. 3. ESEM image of uncoated fibre.

coating are also shown in Fig. 5. The results for PSL particles for dry filter are very close to the results obtained for thecoated filter and are not shown in the figure. As is clearly seen, the difference in performance of dry fibres for cubesand spheres is quite substantial.

These experimental results can also be compared with the theoretical filter efficiency E estimated using the classicapproach (see for example: Lee & Mukund, 2001):

E = 1 − exp

[ −4��L

�df(1 − �)

], (4)

where � is the single fibre efficiency, � is the solidity or packing density of the filter (1-porosity), L the filter thickness,and df the fibre diameter.

In calculations for the theoretical filter efficiency, the parameters of the filter were adjusted to take into account thealterations, due to the oil coating the fibre. The increase in filtration velocity, due to shrinkage of air passages throughthe filter was calculated according to (Hinds, 1999)

V = Q

A × (1 − �), (5)

where Q is the gas flow rate and A is the filter cross-section. It was found that the velocity increase was not significant(less than 3%) but it was taken into account. The results for these theoretical calculations are also shown in Fig. 5. As isseen, they are in good agreement with our experimental results with the maximum discrepancy of less than 12% for thesmallest particle size. This discrepancy is decreasing for larger particle sizes and almost disappears when the particleexceeds 100 nm diameter.

L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229 1227

Fig. 4. ESEM image of coated fibre.

Fig. 5. Efficiency of the filter coated with light mineral oil at velocities 10 cm/s.

1228 L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229

Fig. 6. Efficiency of the filter coated with light mineral oil at velocities 20 cm/s.

Fig. 6 shows similar data for filtration face velocity of 20 cm/s. The experimental results for uncoated fibres show bigdiscrepancy for removal of cubic and spherical particles, which agrees with our findings discussed above. On the otherhand, all tested particles, regardless of shape, have similar filtration efficiencies when the fibres are coated, showingthat motion of the particles along such fibres is limited. The theoretical predictions where also made for this velocity,and the corresponding results are shown in Fig. 6 along with experimental curves. The theoretical results are relativelyclose to the experimental values with the largest variation observed within the range of 75–200 nm. However, withincrease of particle size the theoretical and experimental curves are getting close to each other, which is also observedfor the smallest particle size region.

An ANOVA single factor test was used to determine whether the results for these particle sizes are significantlydifferent in a statistical sense. According to this analysis, with confidence level � = 95%, there is no statisticallysignificant difference between the filtration efficiency of PSL and MgO particles for all particle sizes investigated inthis work for the oil coated fibres. However, the same analysis to compare efficiencies of PSL and MgO particles wasapplied in a case when the filter was not coated with the oil, and the results indicated that there is statistically significantdifference in filtration efficiencies between MgO and PSL.

4. Conclusion

The influence of particle shape and filtration velocity on the filtration efficiency of oil coated fibre filters at twofiltration velocities (10 and 20 cm/s) has been investigated. It was found that in both cases the filtration efficiencies ofthe particles of different shape are very similar indicating that the motion of the particles along an oil coated filter fibreis limited.

The experimental results have been compared with the theoretical filter efficiency predictions according to the classicapproach (Hinds, 1999). To perform these calculations, all parameters of the filter were altered to take into accountthe layer of oil on the fibres of the filter. Based on these findings the experimental results are in good agreement withtheoretical predictions and there is no significant difference in filtration efficiency for all tested particles at both filtrationvelocities used.

The findings were confirmed by ANOVA single factor analysis, which showed that there is no statistically significantdifference between efficiencies of PSL and MgO particles in case when the filter was coated with the oil.

On this basis, it is clear that the oil coating minimizes the amount of particle motion along the fibre after the initialcollision, making the filter performance characteristics similar for all particle shapes.

L. Boskovic et al. / Aerosol Science 38 (2007) 1220–1229 1229

Acknowledgements

The authors are grateful for financial support provided by Smorgon Steel Tube Mills Pty Ltd, Australia. The authorswould like to thank Dr Thor Bostrom for his help with ESEM.

References

Agranovski, I. E., & Braddock, R. D. (1998). Filtration of mists on wettable fibrous filters. American Institute of Chemical Engineering, 4(12), 2775.Agranovski, I. E., Braddock, R. D., & Myojo, T. (1999). Removal of aerosols by bubbling through porous media. Aerosol Science and Technology,

31(4), 249–257.Agranovski, I. E., Braddock, R. D., & Myojo, T. (2001). Comparative study of the performance of nine filters utilised in filtration of aerosols by

bubbling. Aerosol Science and Technology, 35(4), 852–859.Altman, I. S., Agranovski, I. E., & Choi, M. (2004). On nanoparticle surface growth: MgO nanoparticle formation during a Mg particle combustion.

Applied Physics Letters, 84, 5130–5132.Boskovic, L., Altman, I. S., Agranovski, I. E., Braddock, R. D., Myojo, T., & Choi, M. (2005). Influence of particle shape on filtration processes.

Aerosol Science and Technology, 39, 1184–1190.Brown, R. C. (1993). Air filtration: An integrated approach to the theory and applications of fibrous filters. Oxford: Pergamon Press.Clarke, A. G. (1998). Industrial air pollution monitoring. London: Chapman & Hall.Cheng, Y. S., & Yeh, Y. C. (1979). Particle bounce in cascade impactors. Environmental Science and Technology, 13(11), 1392–1396.Dahneke, B. (1971). The capture of aerosol particles by surfaces. Journal of Colloid and Interface Science, 37, 342–353.Hiller, R., & Loeffler, F. (1980). On the degree of impact and the degree of adhesion on particle separation on filter fibres. Staub-Reinhaltung der

Luft, 40(9), 405–411.Hinds, W. C. (1999). Aerosol technology: Properties, behaviour, and measurement of airborne particles. New York: Wiley.Lee, K.W., & Mukund, R. (2001). Filter collection. In: P.A. Baron, & K.Willeke (Eds.), Aerosol measurement: Principles, techniques and applications

New York: Wiley.McHale, G., Newton, M., & Carroll, B. (2001). The shape and stability of small liquid drops of fibers. Oil and Gas Science and Technology, 26(1),

47–54.Mullins, B. J., Agranovski, I. E., & Braddock, R. D. (2003). Particle bounce during filtration of particles on wet and dry filters. Aerosol Science and

Technology, 37, 1–14.Reischl, G. P., & John, W. (1978). The collection efficiency of impaction surfaces. Staub-Reinhart der Luft, 38, 55.Tsai, C.-J., Pui, D. Y. H., & Liu, B. Y. H. (1991). Elastic flattening and particle adhesion. Aerosol Science and Technology, 15, 239–255.Walkenhorst, W. (1974). Investigations on the degree of adhesion of dust particles. Staub-Reinhaltung der Luft, 34, 149–153.Wall, S., John, W., Wang, H.-C., & Goren, S. (1990). Measurements of kinetic energy loss for particles impacting surfaces. Aerosol Science and

Technology, 12, 926–946.