A Circumferential Slot In-Line Virtual Impactor

Aerosol Science and Technology, 38:664–674, 2004Copyright c© American Association for Aerosol ResearchISSN: 0278-6826 print / 1521-7388 onlineDOI: 10.1080/02786820490486015

A Circumferential Slot Virtual Impactor

John S. Haglund and Andrew R. McFarlandAerosol Technology Laboratory, Department of Mechanical Engineering, Texas A&M University,College Station, Texas

A virtual impactor aerosol concentrator has been developed thatuses circumferential slots for acceleration of aerosol particles andfor collection of the coarse fraction. This allows for accurate andeconomical machining of small slot widths, which leads to low-pressure losses for the separation process. One important applica-tion of the device is in the concentration of bioaerosols, especiallyfor military field applications where minimization of power con-sumption is necessary.

A prototype configuration of the circumferential slot virtual im-pactor (CSVI), which was designed using numerical methods, wasconstructed and tested. The device has a curvilinear slit nozzle witha diameter of 150.3 mm (5.918 in), which provides a total slot lengthof 472 mm. Its slot width was 0.499 mm (0.0197 in). According toLoo and Cork, for circular-jet virtual impactors the misalignmentbetween the axis of the acceleration jet and the receiver nozzlewill cause an increase in wall losses of about 1.6% for each 1%of misalignment. Measurements were made of the nozzle dimen-sions in the critical region of the CSVI that showed 1.8% relativemisalignment. When this prototype was operated at a flowrate of122 l/min and a flow fraction (minor air flowrate/total air-flowrate)of 10%, the cutpoint was 2.2 µm aerodynamic diameter and thecorresponding cutpoint Stokes number was 0.58. The collection ef-ficiency was greater than 72% for particle sizes larger than twicethe cutpoint, up to the largest particle size tested (≈10 µm aerody-namic diameter). The peak collection efficiency was greater than95%. For virtual impactors, a critical performance parameter isthe loss of particulate matter to the inner walls of the system. Forthe prototype system, where numerical methods had been used togenerate designs that reduced wall losses, the losses at the cutpointsize of 2.2 µm aerodynamic diameter, are approximately 3%. Foran operational condition of a total flowrate of 122 l/min and a coarseparticle flow fraction of 10%, the pressure drop across the majorflow stream (fine particle stream) was 63 Pa (0.25 in of water), withan ideal power consumption of 0.14 watts.

Received 12 November 2003; accepted 27 April 2004.Funding for the study was provided by the U.S. Army RDE-

COM, Edgewood Area, Aberdeen Proving Ground, MD under ContractDAAD 13-02-C-0064. Dr. Jerold R. Bottiger was the project officer forRDECOM. This support is gratefully acknowledged.

Address correspondence to Andrew R. McFarland, AerosolTechnology Laboratory, Department of Mechanical Engineering,Texas A&M University, College Station, TX 77843, USA. E-mail:[email protected]

INTRODUCTION

Virtual Impaction for Concentrating BioaerosolsOne of the major obstacles that must be overcome by a near–

real-time biological detector is the relatively low concentrationof biological agent particles that can constitute a serious threat.A release of bioagent particles may result in a few tens of agent-containing particles per liter of air (ACPLA) at the location ofthe intended target. The detection threshold for most near–real-time approaches is on the order of at least 103 particles (NationalResearch Council 1999). For a biological point detector to detectthe presence of a few tens of ACPLA within a response time suf-ficient to protect personnel, the aerosol sampled by the detectormay need to be concentrated several orders of magnitude.

For future biological detection systems, it is possible that theorganisms will be identified in the aerosol state using techniquessuch as fluorescence spectroscopy (Pan et al. 2003) or Ramanscattering (Wu et al. 2000). One very promising technique forbioaerosol detection is the B-Cell technology (Harper 2003),where the aerosol particles can be collected in the dry state andthen wetted with a drop of B-Cell hydrosol. It is anticipatedthat both the in situ optical analyses and the dry-deposition ap-proaches will require concentrated bioaerosol samples, if for noother reason than that the greater number of organisms detectedthe more statistically reliable the alarm. For the future, one ofthe critical tasks is the development of small, portable, biologi-cal agent detection systems, suitable for field applications, thatinclude efficient concentration of airborne particles. Such con-centration devices and their associated power supplies must alsobe lightweight and robust.

Virtual impaction is a technique widely used for concentrat-ing aerosols. The two common configurations of virtual impactorgeometry are the axisymmetric (circular or round nozzle) andthe planar-symmetric (rectangular or slot nozzle) virtual im-pactors. For both geometries, aerosol is drawn into one nozzle,the acceleration nozzle, and impinged on an opposed nozzle,the receiver nozzle, which has an aperture that is generally 50%larger than that of the acceleration nozzle (Figure 1). The mo-mentum of the fluid leaving the acceleration nozzle carries theparticle-laden airstream slightly into the receiver nozzle, where

664

CIRCUMFERENTIAL SLOT VIRTUAL IMPACTOR 665

Figure 1. A schematic demonstrating the concept of virtual impaction. The trajectories of particles with sizes near the cutpoint,which are discharged with the major flow, penetrate into the receiver nozzle and then make hairpin turns.

it doubles back out of the receiver nozzle in a hairpin turn. Mostof the flow, which contains aerosol particles with sizes less thanapproximately that of the cutpoint, is drawn off in the gap be-tween the nozzles and is carried away from the nozzle center-line. The remaining flow (typically 10%) is aspirated from thereceiver nozzle and contains the coarse aerosol fraction and fineparticles associated with the transport air flow.

For the bioaerosol concentration process, the virtual impactorshould have a cutpoint below the particle size range of inter-est, which is currently assumed to be either 2.5–10 µm aerody-namic diameter (AD), as used in the U.S. Government Biologi-cal Integrated Detection System, or 1–10 µm AD, as used in theU.S. Government Joint Biological Point Detection System. Forthe application of sampling the ambient environment for fine(respirable) and coarse particulate matter, a cutpoint of 2.5 µmAD is required (USEPA 2003). Although a specific virtual im-pactor will have a fixed flowrate for a desired cutpoint, arbi-trary throughputs can be achieved by operating several virtualimpactors in parallel, and greater concentration factors can be

achieved through multiple stages of virtual impaction operatedin series.

TheoryThe particle and gas-phase flow behavior in a virtual im-

pactor, having a fixed geometry, are governed by two dimen-sionless parameters, the Stokes number (Stk) and the Reynoldsnumber (Re):

Stk = Ccρp D2p Uo

18 µ Lc[1]

and

Re = ρ f LcUo

µ f, [2]

where Dp is the particle diameter, ρ p is the particle density, Cc isthe slip correction factor, Uo is the mean velocity at acceleration

666 J. S. HAGLUND AND A. R. MCFARLAND

nozzle exit, Lc is the characteristic dimension (for circular jetsLc is the nozzle radius (D/2) for Stk and nozzle diameter (D)for Re, and for slot jets Lc is the nozzle half-width (W/2) forStk and the full-width (W ) for Re), ρ f is the gas density, andµ f is the dynamic viscosity.

The Stokes number is the dominant parameter governingparticle behavior in the virtual impactor. The cutpoint Stokesnumber (Stk50), which is the value of the Stokes number corre-sponding to the particle cutpoint size, is weakly a function of theReynolds number owing to minor differences in the flow fieldas affected by the Reynolds number.

For bioaerosol concentration, it is also important to considerthe pressure drop (�P) incurred in moving air through the virtualimpactor. The pressure drop can be represented as a function ofthe acceleration nozzle throat velocity,

�P = Kρ f U 2

o

2, [3]

where K is a pressure coefficient, which is approximately aconstant for a limited range of Reynolds numbers (or flowratesfor fixed air properties and a fixed characteristic dimension).

The theoretical power required for operation of the virtualimpactor, i.e., the minimum power required to move air throughthe virtual impactor without consideration of blower/pump in-efficiencies or pressure losses within the flow handling system,is given by

Wideal = Qma �Pma + Qmi �Pmi, [4]

where Wideal is the the ideal power required, Qma is the majorflowrate (fine particle flow stream), �Pma is the pressure dif-ference between the entrance port of the acceleration jet andthe exhaust port of the major flow, Qmi is the minor (coarseparticle) flowrate, and �Pmi is the pressure difference betweenthe entrance plane of the acceleration jet and the exhaust planeof the minor flow. Here, it is assumed that the flow is incom-pressible, which is a valid assumption for maximum velocitiesless than about 100 m/s (White 1994). Typically, the secondterm on the right side of Equation (4) is much smaller than thefirst term on the right, not only because Qmi is much smaller(∼10%) than Qma , but the pressure drop for the minor flow isnegligible because of pressure recovery in the receiver nozzle(Romay et al. 2002). Neglecting the second term in Equation (4)gives

�Pma = Wideal

(1 − f ) Qtot, [5]

where Qtot is the total flow that enters through the accelerationnozzle and f is the flow fraction (fraction of the total flow as-pirated from the receiver nozzle). Because f is on the order of10%, the major flow pressure drop, �Pma , can be consideredto be an approximate measure of the ideal power per unit offlowrate into the system and as such provides a good method for

comparing the power performance of one fractionation systemagainst another.

Practical contemporary bioaerosol detection systems requireflowrates of 100–1000 l/min to reliably detect the concentra-tion of biological agents that can be expected from a release.For small-dimension round-nozzle virtual impactors, an arrayof many nozzles is usually required; and for slot nozzles, thetotal slot length must be sufficient to supply the required to-tal flow, either as one continuous slot or as an array of slotsof intermediate length. From the standpoint of manufacturabil-ity, either approach has difficulties, especially as the nozzle’scritical dimension approaches the level of tolerance control forthe manufacturing process. For an array of many round nozzles,the risk of defective acceleration nozzle/receiver nozzle pairsbecomes high as the nozzle size is reduced. For slot nozzles,nozzle edge straightness and parallelism becomes more diffi-cult to achieve as the nozzle width becomes smaller and thelength greater. Both designs require precise alignment mech-anisms to align the centers of the receiver and accelerationnozzles.

The objective of this study was to design and characterizethe performance of a CSVI system to demonstrate the proof-of-principle. Both experimental and numerical techniques wereused in the study, although only the experimental effort is de-scribed herein.

Literature ReviewThe original concept of the virtual impactor called for the

replacement of the real impaction surface of a classical inertialimpactor with a virtual surface formed at the inlet of a near-stagnant air cavity (Hounam and Sherwood 1965; Conner 1966).Several early virtual impactor studies (e.g., Dzubay and Stevens1975) used the concept for separating the fine mode particles(aerodynamic particle size, Dp < 2.5 µm AD) from coarse modeparticles (Dp > 2.5 µm AD) typically present in ambient back-ground aerosols. Subsequent theoretical studies sought to predictperformance and improve virtual impactor design (Ravenhallet al. 1978; Forney et al. 1978; Hassan et al. 1979). Early stud-ies relied on the assumption of potential flow to determine theflow field inside the impactor. Later studies (Marple and Chien1980) included viscous effects by numerically solving the time-averaged Navier-Stokes equations and applying subsequent par-ticle tracking to determine the particle behavior inside the virtualimpactor.

Loo and Cork (1988) conducted an extensive experimentalstudy to optimize nozzle geometry in circular jet virtual im-pactors for minimum particle wall losses. Their recommenda-tions define the benchmark for circular jet virtual impactor de-sign. The results of Loo and Cork (1988) demonstrate the needto maintain accurate alignment between the acceleration and re-ceiver nozzle to minimize nozzle wall losses. They observed a1% increase in nozzle wall losses with each 0.05 mm increase(1.6% relative to the acceleration jet diameter) in nozzle mis-alignment.


Besides theoretical studies, Forney et al. (1982) experimen-tally investigated the sensitivity of the ratio of acceleration noz-zle to receiver nozzle width by using a half-plane slot nozzlewith an infinitely variable acceleration nozzle and receiver noz-zle width. Han and Moss (1997) conducted a water analog studyof axisymmetric (circular) virtual impactors and found a regimethat results in stable laminar flow to be nozzle Reynolds num-bers from 2000 to 7000. Outside of this regime, different por-tions of the flow field would exhibit period fluctuations, even forReynolds numbers below 1000.

Masuda et al. (1979) and Chen and Yeh (1987) conductedstudies on circular jet virtual impactors with a clean core ofair introduced at the acceleration nozzle inlet to reduce the fine-particle contamination normally present in the minor flow stream(coarse particle stream). Chein and Lundgren (1995) used theclean-air core virtual impactor to generate near-monodispersetest aerosols by processing polydisperse aerosol sources with theclean-air core virtual impactor. Multistage virtual impactor sys-tems have been constructed for use as bioaerosol concentrators.In 1976, Peterson (2003, personal communication), developed aslot-nozzle virtual impactor that was later incorporated as a com-ponent in a commercial multistage virtual impactor/concentratorsystem (SCP Dynamics Model 1001, Minneapolis, MN, USA)for use in the U.S. Army Biological Integrated Detection Sys-tem point detector. The SCP Dynamics system had four stagesof virtual impaction, the second of which utilized a rectangu-lar nozzle. The system had a particle cutpoint of approximately2.5 µm AD, but was observed to have significant nozzle walllosses for particles larger than about 6 µm AD (Haglund et al.2002). Romay et al. (2002) developed a 300 l/min virtual im-

Table 1Comparison of cutpoint Stokes numbers of real and virtual impactor for studies at a minor/total flow ratio of 10%

Width, W ,Reference Nozzle type mm L/W Re Stk0.5

Hinds (1999) Circular ∗ ∗ ∗ 0.24(real impactor)

Rectangular ∗ ∗ ∗ 0.59(real impactor)

Sioutas et al. (1994b) Rectangular 0.35 43 2,370 0.30Rectangular 0.23 65 1,110 0.25Rectangular 0.23 65 1,780 0.23

Sioutas et al. (1994c) Rectangular 0.33 170 4,460 0.20Ding and Koutrakis (2000) Rectangular 3.05 2.5 1,500 0.48

Rectangular 3.05 2.5 4,400 0.50Rectangular 3.05 2.5 7,300 0.46Rectangular 3.05 2.5 12,000 0.49Rectangular 3.05 2.5 15,000 0.46

Current studyLSVI Rectangular 0.323 274 445 0.58CSVI Rectangular 0.499 946 270 0.58

∗Not available or not relevant.

pactor for bioaerosol concentration that uses multiple circularjets for concentrating particles in the size range of 2–10 µm AD.Sioutas et al. (1994a) developed a multijet, axisymmetric noz-zle virtual impactor for concentration of ambient fine particleshaving a reported cutpoint between 0.1 and 0.2 µm AD. Sioutaset al. (1994b) explored the influence of Reynolds number andflow ratio using two slot (L/W = 43, 65) virtual impactors,where L is the slot length and W is the slot width. They reportedthat the cutpoint Stokes number was between 0.3 and 0.25 fornozzle Reynolds numbers of 2370 and 1100, respectively, ata minor flow ratio of 0.1. Sioutas et al. (1994c) studied a slot(L/W = 170) virtual impactor concentrator having an accel-eration nozzle width of 0.33 mm (0.013′′) and a total flowrateof 225 l/min (7.9 ft3/min). They observed the value of the cut-point Stokes number to be 0.20 at a Reynolds number of 4460and at a minor-to-total flow ratio of 0.1. Ding and Koutrakis(2000) conducted a similar study of short-slot nozzle (L/W =2.5) virtual impactors and observed the cutpoint Stokes numberto be between 0.46 and 0.50 for most nozzle Reynolds num-bers between 4400 and 15,000. A summary of cutpoint Stokesnumbers for virtual impactors is given in Table 1 together withcomparative data for real impactors.

PROTOTYPE DESCRIPTIONTo concentrate the ambient aerosol particulate matter at the

volumetric flowrates required for practical bioaerosol detectionsystems and at low pressure losses, long total slot lengths arerequired (L/W on the order of 1000). For the prototype sys-tem, this was achieved with a circumferential slot nozzle having


Figure 2. Conceptual design of a single-stage circumferential slot virtual impactor.

a diameter of 152.4 mm (5.918′′), which provides a total slotlength of 472 mm (18.6′′), and a slot width of 0.499 mm (0.01966in). The prototype design is shown schematically in Figure 2and photographically in Figure 3. In the CSVI unit, the nozzlethroats are formed on the circumference of four annular blades.The aerosol enters the acceleration nozzle flowing radially in-ward. The major flow is then diverted axially both up and downin the impactor, and then exhausted through the gaps betweena series of ten evenly-spaced “spokes” to the vacuum annulus,as shown in Figure 3. The minor flow containing the concen-trated aerosol continues on the radially inward trajectory andis collected at the central axis of the device. For test purposes,this concentrated aerosol was drawn through a 25.4 mm (1.0′′)tube for subsequent collection on a 102 mm (4.0′′) glass fiberfilter. In this design, a second stage could be nested in the im-

Figure 3. Photograph of single-stage CSVI. The assembly issplit to show the internal structure.

pactor to further concentrate the coarse aerosol (minor flow).Also, higher flowrates could be achieved by stacking parallelimpaction stages.

On the basis of virtual impactor theory relating to the particleseparation characteristics of slot nozzle virtual impactors, a cir-cumferential slot nozzle should perform identical to linear jets,provided the total slot length and critical geometries are equiv-alent and provided the radius of curvature of the slot is muchgreater than the slot width. For the prototype constructed, that ra-tio was approximately 150 times. This equivalency is exclusiveof any end-effects of the linear virtual impactor or differencesin aerosol aspiration of the two concepts. It should also be notedthat one important advantage of the circumferential design overthe linear array design is the relative compactness for the sametotal slot length.

In this study, we also constructed and tested, for comparativetesting, a linear slot virtual impactor (LSVI). Because of prob-lems associated with alignment of the acceleration and receivernozzles and with machining tolerances, we fabricated severalLSVI units before we achieved a model that was suitable fortesting. Only data associated with that final model will be shownhere. The LSVI prototype is shown in Figure 4.

EXPERIMENTAL METHODOLOGY

Testing with Monodisperse Liquid Fluorescent AerosolFor particles with sizes greater than about 2 µm, tests to char-

acterize the collection efficiency and internal wall losses wereconducted with monodisperse liquid aerosols generated with aBerglund-Liu (Berglund and Liu 1973) vibrating orifice aerosolgenerator (TSI Model 3050, St. Paul, MN, USA). A nonvolatileliquid (oleic acid) was tagged with 10% sodium fluoresceinsalt (uranine) fluorescent tracer for subsequent determinationof aerosol mass. The mean size of the aerosol droplets was de-termined by impacting them on a glass slide coated with an oil-phobic film. The droplets were then measured with an optical


Figure 4. Photograph of a linear slot virtual impactor (LSVI).The acceleration nozzle is in the front side of the component,and an outlet port for the major airflow is at the top.

microscope. The original spherical sizes of the droplets werecalculated from knowledge of the degree of gravitational flat-tening of the oil droplets on the slide (Olan-Figueroa et al. 1982).

An aerodynamic particle sizer (APS: Model 3310, TSI, Inc.,St. Paul, MN, USA) was used to monitor the size distributionof the test aerosol during each test. Because liquid droplets areknown to distort sufficiently to produce appreciable particle siz-ing errors when accelerated in the APS measurement section(Griffiths et al. 1986; Chen et al. 1990), primary reliance for par-ticle sizing was placed on use of the optical microscope method.

The generated test aerosol was drawn into one end of anaerosol delivery duct. At the opposite end, a sample was ex-tracted and collected on a glass fiber filter after passing theaerosol through the virtual impactor. For tests with the LSVIunits, separate filters were collected for the left and right majorflow paths, and the minor flow. Next, a reference aerosol samplewas extracted and collected on a filter where the point of extrac-tion and flowrate were the same as those for the virtual impactorinlet. Setting the flowrate of the reference sample equal to that ofthe virtual impactor, and by using reference filter geometry withminimal losses, the reference aerosol sample was representa-tive of that entering the test device. For most experiments, threeor more pairs of reference and test configuration samples werecollected for each particle size studied to quantify repeatabilityof results. The filters were placed in solutions containing 50%isopropyl alcohol and 50% distilled water (v/v) to elute the flu-orescent tracer. The total solution volume varied between 20 mland 80 ml, where less liquid was used for smaller test particlesto increase the fluorescent concentration of the solution. Therelative concentrations of tracer in the solutions were measuredwith a fluorometer (Turner/Barnstead Thermolyne Model 450,Dubuque, IA, USA). By dividing the relative concentration of

fluorescent tracer in the test filter solution by the relative concen-tration of the reference filter, and correcting for any differences inthe sample volume, the percentage of particulate mass collectedin the major and minor flows relative to total test particulate wasdetermined.

The wall losses were directly measured by disassembling thevirtual impactor units between each set of three test runs at agiven particle size, and recovering deposited particles. The par-ticle losses were recovered from the critical region (accelerationand receiver nozzles) by wetting a cotton swab with isopropylalcohol and swiping the nozzle throat. Four swipes of the nozzlethroat were made to ensure maximum particle recovery. Fol-lowing the swabbing procedure, the nozzles were rinsed withisopropyl alcohol to remove any trace of unrecovered depositsand to establish a zero baseline for the next experiment. The cot-ton swabs containing the fluorescent deposits were immersed inthe same type of solution as used for analysis of the aerosol fil-ters, and the fluorescent concentration was analyzed as describedabove.

Testing with Presized Polystyrene Latex ParticlesFor tests with particles smaller than 2 µm diameter, a sus-

pension of solid monodisperse polystyrene latex (PSL) micro-spheres (Nanosphere Size Standards, Duke Scientific, Palo Alto,CA, USA; Estapore Uniform Microspheres, Bangs Laborato-ries, Inc., Fishers, IN, USA) in distilled water was atomizedusing a Collison nebulizer (BGI Inc. Model CN31I, Waltham,MA, USA). The mass concentration of PSL in the suspensionvaried between 0.02 and 0.08%, so the number concentrationwas well below the limit of 109 particles/ml suggested by May(1973) to prevent significant numbers of doublet particle mi-crospheres from being generated. The particle size spectrumgenerated by the nebulizer was periodically sampled through-out the tests to identify any problems, including generation ofunacceptable concentrations of doublet particles.

As in the liquid aerosol tests, a reference filter was usedto determine the relative concentration of aerosol entering thevirtual impactor. However, unlike the liquid aerosol tests, theduration of the aerosol sample collected was not determinedby length of time in which flow was drawn through the vir-tual impactor or the reference filter. Rather, for each test thevacuum pump was first activated on the device (either the vir-tual impactor or reference filter). Next, a constant pressure of138 kPa (20 psi) was applied to the nebulizer for a fixed du-ration (usually 10 min), thus producing an approximately con-stant aerosol generation rate. The aerosol generation processwas then halted, but the vacuum pump of the reference filteror virtual impactor was allowed to run for an additional minuteat the end of particle generation to fully evacuate the aerosoldelivery duct. APS samples taken at the test section of the de-livery duct confirmed that all test aerosols were removed fromthe delivery duct within approximately the first 10 s of shut-down of the nebulizer. Conducting experiments in this man-ner allowed approximately six to eight consecutive tests (either


Figure 5. Collection efficiency (minor flow) and wall losses of the LSVI.

reference or test configuration) to be run before the nebulizerneeded to be refilled. This equates to a liquid consumption rateof about 20 ml/h and is consistent with the observations of May(1973).

The glass fiber filters containing the fluorescent microspheres(Pall Gelman Sciences Model A/E 102 mm, Ann Arbor, MI,USA) were immersed for a minimum of 8 h in 20 ml of ethylacetate to dissolve the polystyrene latex and liberate the encap-sulated dye. The relative fluorescent concentration of the solu-tions was measured with a digital fluorometer (Turner/BarnsteadThermolyne Quantech Model FM109515, Dubuque, IA, USA)using optical filters appropriate for the excitation and emissionwavelength of the fluorescent dye of the PSL. Moderate agita-tion of the filter solution by swirling the liquid was sufficientto uniformly mix the solution. Multiple samples (∼5 ml) takenfrom the same original solution mixed in this manner were in-distinguishable from multiple fluorometric readings of the samesample. Thus, it was only necessary to measure one subsampleextracted from each mixed solution. Nozzle wall losses were notrecovered in the fluorescent PSL experiments. Wetting a cottonswab or tissue wipe with either isopropyl alcohol or water failed

to remove the PSL deposits. In principle, the wall losses couldbe recovered by complete immersion of the virtual impactor, butthe volume of ethyl acetate required in this study would haveresulted in solutions too dilute to give fluorescent readings suf-ficiently greater than the background fluorescence of the ethylacetate reagent.

RESULTS AND DISCUSSIONInitial collection efficiency tests of the LSVI with PSL

showed two distinct regions of behavior. For flowrates below35 l/min (1.2 cfm) classical virtual impactor performance wasobserved; however, for flowrates above 38 l/min (1.3 cfm) adeparture was noted, where the LSVI performance asymptoti-cally approached the behavior of a high-loss mixing element. At90 l/min (3.2 cfm total flow), which was the highest flowrate ex-amined, the aerosol concentrations in the major and minor flowpaths were approximately equal and independent of particle size;i.e., there was no enrichment of particles in the minor flow streameven at particle Stokes numbers significantly greater than unity,contrary to normal virtual impactor behavior. Furthermore, the


sum of the collection efficiencies of the major and minor flowpath was observed to decline with an increase in total flowratebeyond 38 l/min, which is evidence of increasing particle losses.This behavior was caused by the onset of an acoustic ringingassociated with air sampling flowrates above 38 l/min. Simi-lar behavior was observed with the CSVI, where the thresh-old flowrate for the onset of acoustic ringing was 140 l/min(5 cfm). As a consequence, all data reported herein are asso-ciated with operation below the onset of acoustic ringing. Thisringing is likely due to a standing wave generated through anacoustic instability (Blackstock 2000). Subsequent investiga-tions have shown the ringing can be eliminated by redesign ofthe plenum downstream of the receiver nozzle in the LSVI andmodification of plena upstream of the acceleration nozzles anddownstream of the receiver nozzles in the CSVI.

Linear Slot Virtual ImpactorThe collection efficiency in the minor flow of the LSVI at

a flowrate of 38 l/min (1.3 cfm) corresponding to a nozzleReynolds number of 445 is seen in Figure 5. The particle cut-point corresponding to 50% collection efficiency was 1.1 µm

Figure 6. Collection efficiency (minor flow) as a function of Stokes number for the CSVI and LSVI units at 10% minor/totalflow ratio. Reynolds number for the CSVI was 270, and for the LSVI it was 445.

AD. The collection efficiency peaks at a value of 95% at a par-ticle size of 3.5 µm AD and is reduced to a value of 72% for9.6 µm AD particles.

The wall loss results for the LSVI (Figure 5) show “Mea-sured Wall Losses” of between 5 and 10% for particles in thesize range of 2 to 7 µm AD. However, the wall losses are signif-icantly greater, 26%, at the largest particle size tested (9.6 µmAD). Also shown in Figure 5 are two connected data points witha curve titled “Calculated Wall Losses.” For particles larger thanabout 5 µm AD, the efficiency should be nearly 100%, were itnot for wall losses. The “Calculated Wall Losses” curve is thedifference between 100% and the measured efficiency. Com-parison between the “Measured Wall Losses” and “CalculatedWall Losses” provides an indication of the aerosol particle massbalance, i.e., the two curves would agree if all of the wall losseswere recovered. Good agreement is obtained for the LSVI in thatthe measured wall losses are within about 2% of the calculatedvalues.

The data presented on a basis of minor flow efficiency as afunction of Stokes number (Figure 6) show the cutpoint Stokesnumber to be 0.58. This is comparable to the values of 0.46


to 0.50 given by Ding and Koutrakis (2000). The Reynoldsnumbers for the Ding and Koutrakis study ranged from 1500 to15,000, while the Reynolds number for the LSVI was445.

Pressure drop across the LSVI at the flowrate of 38 l/min was530 Pa (2.1 in of water), and the average flow coefficient was1.9 over the Re range of 273–445. For this flow coefficient theideal power is 0.37 watts (Equation (5)) to achieve the cutpointof 1.1 µm AD at the flowrate of 38 l/min.

Circumferential Slot Virtual ImpactorThe collection efficiency in the minor flow of the CSVI Unit

at 122 l/min (4.3 cfm) is shown in Figure 7. The particle cut-point was 2.2 µm AD, and the collection efficiency peaks ata value of 96% at a particle size of 4.7 µm AD. Further in-creases in particle size reduced the efficiency to a value of 77%for the largest particle size tested (9.8 µm AD). Similar to theperformance of the LSVI, the reduction in minor flow efficiencyfor the larger particle sizes can be attributed to particle losses.However, for CSVI, wall losses in the critical zone (shown as

Figure 7. Collection efficiency (minor flow) and wall losses of the CSVI.

“Measured Wall Losses” in Figure 7) were significantly less thanfor the LSVI unit. These reduced losses are probably due to therelative superiority of the dimensional precision of the nozzlesin the CSVI compared with those of the LSVI. For the CSVI,the sum of the measured wall losses and minor flow collectiondoes not account for all of the particulate mass. For example, atthe particle size of 9.8 µm AD, the collection efficiency is 77%and the measured wall losses are 3%, which leaves 20% of theaerosol particles unaccounted for. A likely explanation for theobserved mass defect is that losses that occurred in regions otherthan the critical zone due to the more torturous flow paths trans-porting aerosol to and from the fractionation zone of the CSVIunit compared with the LSVI unit. Aerosol entered the CSVIdevice through a 50.8 mm (2′′) diameter tube and was then di-rected radially outward by a blunt deflection cone. The aerosolwas then aspirated into one of ten oblong cross-section ports inthe entrance annulus of the CSVI unit. Once aspirated, the flowwas again diverted radially inward through an aerodynamicallysmoothed transition and accelerated in the impinging nozzle.These losses were not recovered in the process of cleaning thecritical region.


Figure 8. Pressure drop across major flow path of the CSVI. Average pressure coefficient: K = 1.50.

The cutpoint Stokes number of the CSVI is 0.58, as may benoted from Figure 7. There is good agreement with the cutpointStokes number for the LSVI and with the results of Ding andKoutrakis (2000). The slope of the fractionation curve for theCSVI is sharper than that of the LSVI, perhaps reflecting thebetter adherence to tolerances with components fabricated on alathe (CSVI) rather than a mill (LSVI).

Pressure DropThe pressure loss incurred in moving air through the virtual

impactor is an important consideration for minimizing the sizeand weight of bioaerosol concentration systems. The pressuredrop in the major flow line of the CSVI at variable total flowratesand fixed minor/total flow ratio of 10% is shown in Figure 8,where it may be noted that at the flowrate of 122 l/min thepressure drop was 63 Pa (0.25 in of water). The pressure losscoefficient, K , averaged 1.5 for Reynolds numbers from 157to 367. For the flowrate of 122 l/min, this pressure loss, anda minor flow ratio, f , of 10% suggest that the ideal power tooperate the CSVI system (Equation (5)) is 0.14 watts. Using

Stokes number scaling and the flow coefficient of 1.5, a cutpointof 1.0 µm AD would be obtained if the slit width were 0.30 mm(0.012 in) and the flowrate was 196 l/min. The ideal power forthat configuration and operating condition would be 0.9 watts.

SUMMARYA circumferential slot impactor/concentrator was constructed

and characterized and a linear counterpart was fabricated forcomparative purposes. Both devices were longer slot (L/W ),lower Reynolds number, and lower pressure drop than previousstudies of such virtual impactors (Table 1). The measured valueof Stk50 was 0.58 for both devices, corresponding to a parti-cle cutpoint size of 1.1 µm AD for the LSVI configuration and2.2 µm AD for the CSVI. The collection efficiency was greaterthan 72% for all particle sizes larger than three times the cut-point up to the largest particle size tested (≈10 µm AD). Thepeak collection efficiency for both concentrators was greaterthan 95%.

An acoustic resonance was observed at flowrates above well-defined threshold values in both the LSVI and CSVI, and the


results reported herein are for flow conditions that did not pro-duce ringing. In future models, design provisions will be madeto preclude the ringing.

Pressure loss is very low, as the cutpoint of 2.2 µm AD wasachieved at a major flow pressure drop of 63 Pa in the CSVIand a cutpoint of 1.1 µm AD was associated with a major flowpressure drop of 530 Pa in the LSVI.

REFERENCESBerglund, R. N., and Liu, B. Y. H. (1973). Generation of Monodisperse Aerosol

Standards, Environ. Sci. Technol. 7:147–153.Blackstock, D. T. (2000). Fundamentals of Physical Acoustics, John Wiley &

Sons, New York, NY.Chein, H., and Lundgren, D. A. (1995). A High-Output, Size-Selective Aerosol

Generator, Aerosol Sci. Technol. 23:510–520.Chen, B. T., Cheng, Y. S., and Yeh, H. C. (1990). A Study of Density Effect

and Droplet Deformation in the TSI Aerodynamic Particle Sizer, Aerosol Sci.Technol. 12:278–285.

Chen, B. T., and Yeh, H. C. (1987). An Improved Virtual Impactor: Design andPerformance, J. Aerosol Sci. 18:203–214.

Conner, W. D. (1966). An Inertial-Type Particle Separator for Collecting LargeSamples, J. Air Pollut. Contr. Assoc. 1:35–38.

Ding, Y., and Koutrakis, P. (2000). Development of a Dichotomous Slit NozzleVirtual Impactor, J. Aerosol Sci. 31:1421–1431.

Dzubay, T. G., and Stevens, R. K. (1975). Ambient Air Analysis with Dichoto-mous Sampler and X-ray Fluorescence. Environ. Sci. Technol. 9:663–668.

Forney, L. J., Ravenhall, D. G., and Lee, S. S. (1982). Experimental and Theo-retical Study of a Two-Dimensional Virtual Impactor, Environ. Sci. Technol.16:492–497.

Forney, L. J., Ravenhall, D. G., and Winn, D. S. (1978). Aerosol Impactor: AStudy of a Fluid Jet Impinging Upon a Void, J. Appl. Phys. 49:2339–2345.

Griffiths, W. D., Iles, P. J., and Vaughan, N. P. (1986). The Behavior of LiquidDroplet Aerosols in an APS 3300, J. Aerosol Sci. 17:921–930.

Haglund, J. S., Chandra, S., and McFarland, A. R. (2002). Evaluation of a HighVolume Aerosol Concentrator, Aerosol Sci. Technol. 36:690–696.

Han, R., and Moss, O. R. (1997). Flow Visualization Inside a Water ModelVirtual Impactor, J. Aerosol Sci. 28:1005–1014.

Hari, S. (2003). Computational Fluid Dynamics (CFD) Simulations of DiluteFluid-Particle Flows, Ph.D. Dissertation, Department of Nuclear Engineering,Texas A&M University, College Station, Texas.

Harper, J. (2003). Canary B-Cell Sensor for Rapid, Sensitive Identificationof Pathogens. Presentation at 3rd Annual Aerosol Biodetection Technolo-gies/2003 International Symposium. Washington-Arlington, VA, 1–3 June,2003. The Knowledge Foundation, Brookline, MA.

Hassan, Y. A., Jones, B. G., and Yule, T. J. (1979). An Analytical Study of VirtualImpactor Aerosol Separators, Trans. Amer. Nucl. Soc. 33:182–184.

Hinds, W. C. (1999). Aerosol Technology, 2nd ed., John Wiley and Sons, Inc.,New York, NY.

Hounam, R. F., and Sherwood, R. J. (1965). The Cascade Centripeter: A Devicefor Determining the Concentration and Size Distribution of Aerosols, Amer.Ind. Hyg. Assoc. J. 26:122–131.

Loo, B. W., and Cork, C. P. (1988). Development of High Efficiency VirtualImpactors, Aerosol Sci. Technol. 9:167–176.

Marple, V. A., and Chien, C. M. (1980). Virtual Impactors: A Theoretical Study,Environ. Sci. Technol. 14:976–985.

Masuda, H., Hochrainer, D., and Stober, W. (1979). An Improved Virtual Im-pactor for Particle Classification and Generation of Test Aerosols with NarrowSize Distributions, J. Aerosol Sci. 10:275–287.

May, K. R. (1973). The Collison Nebulizer: Description, Performance and Ap-plication, J. Aerosol Sci. 4:235–243.

National Research Council. (1999). Chemical and Biological Terrorism: Re-search and Development to Improve Civilian Medical Response. NationalAcademy Press, Washington DC, p. 81.

Olan-Figueroa, E., McFarland, A. R., and Ortiz, C. A. (1982). Flattening Coef-ficients for DOP and Oleic Acid Droplets Deposited on Treated Glass Slides,Am. Ind. Hyg. Assoc. J. 43:395–399.

Pan, Y. L., Hartings, J., Pinnick, R. G., Hill, S. C., Halverson, J., and Chang,R. K. (2003). Single-Particle Fluorescence Spectrometer for AmbientAerosols. Aerosol Sci. Technol. 37:628–639.

Ravenhall, D. G., Forney, L. J., and Jazayeri, M. (1978). Aerosol Sizing with aSlotted Virtual Impactor, J. Colloid Interface Sci. 1:108–117.

Romay, F. J., Roberts, D. L., Marple, V. A., Liu, B. Y. H., and Olson, B. (2002).A High Performance Aerosol Concentrator for Bioaerosol Agent Detection,Aerosol Sci. Technol. 36:217–226.

Sioutas, C., Koutrakis, P., and Olson, B. A. (1994a). Development and Eval-uation of a Low Cutpoint Virtual Impactor, Aerosol Sci. Technol. 21:223–235.

Sioutas, C., Koutrakis, P., and Burton, R. M. (1994b). Development of a LowCutpoint Slit Virtual Impactor for Sampling Ambient Fine Particles, J. AerosolSci. 25:1321–1330.

Sioutas, C., Koutrakis, P., and Burton, R. M. (1994c). A High-Volume SmallCutpoint Virtual Impactor for Separation of Atmospheric Particulate fromGaseous Pollutants, Particle Sci. Technol. 12:207–221.

U.S. Environmental Protection Agency (USEPA). (2003). National Primary andSecondary Ambient Air Quality Standards for Particulate Matter, 40 CFR50.7, U.S. Government Printing Office, Washington, DC.

White, F. M. (1994). Fluid Mechanics, 3rd ed. McGraw-Hill Inc., New York,p. 511.

Wu, Q., Nelson, W. H., Elliot, S., Sperry, J. F., Feld, M., and Manogaran, R.(2000). Intensities of E. coli Nucleic Acid Raman Spectra Excited Selectivelyfrom Whole Cells with 251 nm Light, Anal. Chem. 72:2981–2986.

A Circumferential Slot In-Line Virtual Impactor

Documents

Transcript of A Circumferential Slot In-Line Virtual Impactor