You are on page 1of 44

Modification andrEvaluation of the WINS Impactor

Final Report

prepared by:
,--.
Mr. Thomas M. Peters
Robert W. Vanderpool, Ph.D.

Center for Engineering and Environmental Technology


Research Triangle Institute
PO Box 12194
Research Triangle Parþ NC 27709

Prepared for:

Project Officer: JackH. Shrefler, Ph.D.


Task Manager: Russell W. Wiener, Ph.D.

National Exposure Research Laboratory


Atmospheric Processes Research Division
U. S. Environmental Protection Agency
Research Triangle Parþ NC 277II

EPA Contract Number: 68-D5-0040


RTI Project Number: 6360-007

September 1996
ABSTRACT

The EPAwell-type impactor ninety-six (WNS) impactor was modified and evaluated to
serve as a separation device for a proposed federal reference method for particulate matter under
2.5 pm. Desþ modification to the \ryINS focused on the requirement of sharpening the
penetration versus particle aerodynamic diameter curye. The ratio of the throat length over the jet
widtb the T/Iil ratio, was reduced from 5.5 to to sharpen the shape of the penetration versus
1.6
particle aerodynamic diameter curve. The diameter ofthe well was increased from 30 mm to 37
mm to accommodate the larger nozzle necessary for the TÄil alteration. The jet-to-plate distance
over the jet width (S/!Ð was increased from 2 to 3 with the intention that higher loadings would
atrect the shape and position of the penetration curve to a lesser degree. Additionall¡ the screw
threads of the anti-spill ring on the well were changed to an o-ring for simplicity, ease of use, and
to prevent oil leakage. The performance curve of the redesigned \ryINS has a geometric standard
deviation of 1.18 as comparedto 1.24 of the original desþ.
Two candidate oils (Neovac and Dow Corning 704 diffiision pump oils) and filters (glass
fiber filters, drain discs, and nuclepore membrane filters) were evaluated for use as impaction
substrates in the WINS. The two oils, the glass fiber filters, and the drain discs showed similar
penetration characteristics; however, the nuclepore filters had s[ghtly elevated penetrationvalues.
As a result of this testing, the original WINS configuration, a Gelman Type AÆ glass fiber filter
immersed inDow Corning 704 diffi¡sion pump oil, was maintained as the impaction substrate for
the modified \ryINS.
Further testing showed that the penetration curve was essentially the same with I or 3 mL
quantities of oil. Carry-over of the oil to the afterfilter was not detected in significant quantity by
either gravimetric means or by XRF over 1 and 5 day sampling periods. Tests to obsen¡e the
impactor performance as a function of loading showed that the conical build-up of coarse mode
aerosol on the impaction surface increases the impactor's collection efficiency. It is estimated that
this change in performance would result in a maximun¡ negative bias in measured fine particulate
mass concentration of 5 percent following continuous sampling of a coarse mode aerosol
concentration equivalent to 165 pgnf for six consecutive days.
TABLE OF CONTENTS

ABSTRACT .......ü

TABLEOFCONTENTS.... ... iii

SECTION

1.0 INTRODUCTION ........I


2.0 PENETRATIONMEASUREMENTPROCEDURES... ........3
3.0 DESIGNOFTHEWINSIMPACTOR .......5
4.0 EVALUATION OF TTIE WINS IMPACTION SUBSTRATE . . . . 17

5.0 EVALUATION OF TTIE \ryINS IMPACTOR AFTER COARSE MODE


AEROSOLCOLLECTION. ....25
6.0 STJMMARY A}{D CONCLUSIONS . . . 38

7.0 REFERENCES... ......39

APPENDD(

A Machine Drawings of the Modified EPA WINS Impactor . . . 4l

u
INTRODUCTION

The Environmental Protection Agency (EPA) is currentþ developing a federal reference


method (FRlvf) for measuring particulate matterunder 2.5 microns (PÀ[.r). The proposed
approach of the FRM involves inertial separation and gravimetric quantitation of collected Pì4.,
atmospheric aerosol. It is intended that this FRMbe defined in 40 CFRPart 50 AppendixL and
consist of a hybrid of design and performance specifications. The desþ ofthe sampling train
hardware upstream'of the filter, those components intended to representively aspirate and
separate the Pltt., aerosol from the environment, are to be manufactured from detailed machine
drawings which will be provided in Appendix L. The flow control and temperature control
systems are to be based on performance criteria documented in Appendix L and tested by the
procedures prescribed in 40 CFRPart 50 Subpart E.
Those sampling train components specified by design are the sampler inlet, downtubq
fractionator, filter cassette, and the inner dimensions of the upper portion of the filter holder. The
inlet desþ is identical to the Graseby-Andersen (Atlantq GA) Model 5A2468 which is designed
to aspirate particulate matter under l0 ¡rm (PÀ[J from the atmosphere at 16.7 al-pm and has
successfully passed wind tunnel PMro aspiration tests at wind speeds up to 24lallÍn (RTI, 1990).
The downtube is simply a straight section of tubing which transports the aerosol from the inlet to
the fractionator. A single-stage, impactor-type fractionation device, the EPA well-impactor
ninety-six (WINS, RTI, 1996a), is designed to separate the Plvfr., from the aspirated PMro
aerosol. The filter cassette houses a Teflon filter to collect the fine particulate matter for
subsequent conditioning and gravimetric analysis.
In the WINS, size separation is achieved by a single-jet, round hole impactor. Particles
greater than2.5 ¡rm have sufficient inertia to be impacted into a well containing a circular 3 cm
diameter glass fiber filter immersed in I mL of low volatility oil. The oil and filter combination is
intended to minimize substrate overloading and subsequent particle bounce-offexperienced by
some conventional impactors. The desþ of the impactor was based upon classical impaction
theory (Murpl. and Willeke,1976). Performance ofthe prototype WINS desþunder static
conditions was evaluated using monodisperse, solid, ammonium fluorescein aerosol (RTI, I996a)
to have a particle aerodynamic 50 percent cut-point diameter (Dro) of 2.5 ¡rm with a geometric
standard deviation (GSD) of 1.24. Because of historical convention and conveniencg GSD was
used to represent the shape of the penetration curve even though the curve is not symmetric
2l
j:

above and below the particle D, diameter. GSD was calculated as the square root of the particle
diameter vrith 16 percent penetration divided by the particle diameter \ilith 84 percent penetration
i

During public comment on the progress of the FRlvf, the shape of the curve was critiqued
i

to be too shallow when compared with theoretical impactor curyes and other existing impactors t.
I

designed for other purposes (Marple, 1996 and Lundgrer¡ 1996). In response to these concerns, i

the clean air scientific advisory committee (CASAC) recommended that the size fractionator
maintain a cutpoint of 2.5 ¡rm with a GSD of less thanl.Z.
The intent of this Work Assignment was twofold: 1.) modifr the desþ of the WINS
fractionator so that its performance meets the requirements of CASAC, and 2.) test this design to
ensure that the WINS fractionator adequately performs as intended for use in the FRM for Plvfr.r.
Section 2 of this report describes modifications to the static test apparatus necessary to facilitate
rapid evaluation of impactor penetration necessary in the evaluation of the WINS. Section 3 of
this report presents the steps taken and the results of the impactor design modification effort.
Section 4 describes the evaluation of the impactor substrate, oil carry-over to the afterfilter, and
operation of the impactor well with one or tlree mL of diftision pump oil. an
Section 5 presents l

evaluation of the WINS after being loaded with a high concentration of coarse mode aerosol. I
2.0 PEI\TETRATION MEASI]REMENT PROCEDI]RES

Penetration versus particle aerodynamic diameter is the fundamental "yard sticlC' in the
evaluation of aerosol separation devices and may be measured in a number of ways. RTI (1996a)
presented a static testing apparatus and procedures used to evaluate penetration for the original
WINS impactor. Briefly, this protocol challenged the separator with monodispersed aerosols
created with a vibrating orifice aerosol generator (VOAG). The VOAG aerosols were composed
of solid ammonium fluorescein with a known densþ of 1.35 glc
t. The number concentration of
primary particles downstream and upstream of the separation device was measured with an
Aerosizer (Model Aerosizer-LD, API, Amherst, lvfA). The Aerosizer wan¡ used to verify the
aerodynamic size of the primary particles as well. Penetration for a given size was calculated by
dividing the number concentration of primary particles in the downstream by that in the upstream.
An entire penetration curve was then developed by stepping through discrete particle sizes until
an adequate number of sizes were evaluated so that the Dr and GSD of the separator could be
determined.
The tests involving the modification to the WINS impactor, presented in Section 3, were
performed using these identical procedures and will, therefore, not be presented here. The
requirements of testing impaction substrates and further evaluation ofthe \ryINS, which are
presented in Section 4 and Section 5, required a more rapid technique to measure the penetration
curve. To accomplish this, the most labor-intensive and time-consuming component of the above
procedure, the production of monodispersed VOAG aerosols and the need for testing only one
size at a time, was replaced with generation of an aerosol composed of multiple sizes of
polystyrene latex (PSL) spheres.
A pneumatic nebulizer @rofessional Medical Products, Inc., Greenwood, SC, 29646)was
used to nebulize suspensions containing deionized water and the PSL spheres of the size and type
presented in Table 1. The nebulizer was fed filtered, compressed ar at a gauge pressure of
103 liPa . Two suspensions (identified as A and B in Table 1) were prepared with each
suspension containing th¡ee discrete sizes ofPSL. These suspensions were prepared such that the
relative number concentration of each PSL size were roughly equivalent and such that the overlap
between sizes was minimat or non-existent. The other components ofthe static test apparatus and
procedures to determine penetration were identical to that presented in RTI (1996a).
Table I displays the calculated mean aerodynamic diameter based on the vendor-reported
t
physical mean diameter and a PSL density of 1.05 ú" as well as the mean aerodynamic
diameter as measured by the Aerosizer. The measured aerodynamic mean diameter is within
Table 1. Polystyrene Latex particle reported and measured particle size.
Measured
Reported Reported Calculated Mean
PSL Physical St¿ndard Aerodynamic Aerodynamic
Yendor Solution I)iameter. um' I)eviation. um' Diameter. umt I)iameter. um'

Polysciences A 1.531 0.056 r.57 1.59

Polysciences B 2.092 0.095 2.t4 2.r9

Bang's Labs A 2.300 NA 2.36 2.34

Polysciences B 2.836 0.136 2.9t 2.91

Polysciences A 3.189 0.054 3.27 3.2r

Polvsciences B 4.392 0.269 4.M 4.22

" Vendor reported.


b
Calculated based on reported physical diameter and particle density of 1.05 gl"t f .

o
Measured by Aerosizer.'
NA- not available.

0.05 ¡rm of the calculated diameter for all sizes excluding the two larger sizes. For both of the
larger diameters, the measured diameter was observed to be slightly smaller than that calculated.
This discrepancy may be due to losses of the larger particles in the bends of the static test
apparatus; however, it is not a significant consequence due to the fact that these sizes are
substantially larger than the WINS 2.5 pmcutpoint and the fact that all of the measurements made
with the PSL aerosol are in relationship to one another.
DESIGN OF THE WINS IMPACTOR

The design ofthe originalWINS was documented inRTI, I996a and is shown
schematically in Figure 1. The impaction assembly was designed to operate downstream of the
Graseby-Andersen Model 2468 tnlet at a volumetric flowrate of 1 m3/hr. The unit consists of an
upper housing, an impaction reservoir (or well), and a lower housing. Separation of the aerosol is
provided by a single-jet round hole which is integral to the upper housing. Particles with
sufficient inertia are impacted upon a circula¡ 3 cm diameter glass fiber filter which is immersed in
1 mL of a low volatility diffiision pump oil. The PMro aerosol enters from the top of the unit and

the PMr., fraction of this aerosol exits through the bottom of the unit for collection on an after
filter.
Conventional impactordesþ guidelines presented inMarple and Rubow (1986) were
closely considered during the desþ of the original \ryINS ¡szzls section. These guidelines
recommended that the minimum S/\il be 1.0, the jet Reynolds number be between 500-3000, and
the entrance of the ¡sz'ls be tapered or conical. Additional guidelines from Marple and Willeke
(L976) suggested that the throat of the nozzlebe of constant width and possess a TAil ratio of
greater than 1.0. Table 2 presents the critical dimensions of the originat desþ of the \ryINS
impactor. The jet-to-plate distance over thenozzlewidtt\ the SAil ratio, was 1.8, the nozzle
throat length over the noz.zfewidth, the TAil ratio, was 5.5, and the jet Reynolds number was
5200. The design jet Reynolds number was outside the recommended 500-3000 range due to the
use of a single hole with the 1 m3/Tr flow rate. The single hole was determined to be desi¡able
because of its simplicþ of construction, lower tendency to clog, and compatibility with the well-
t¡pe impaction surface.
The impaction plate and substrate were designed to minimize overloading and subsequent
particle bounce-offexperienced by some conventional impactors. The use an oil immersed in a
porous surface for an impaction substrate was selected because of the abilþ of an oil to wick
through previously deposited particulate and present a continuousþ wetted surface to the
incoming aerosol (Marple et al., 1987 and Turner and Hening, 1987). The non-traditional well-
type impaction plate geometry was designed to hold a relatively large quantþ of oil (up to 3 mL)
in the impaction area. Additionally, the well is intended to retain any "blow-off' of the tip of the
conical build-up that is typical after the collection of large quantities of particulate matter with
diameters above the cutpoint ofthe impactor.
The original \ryINS penetration versus particle aerodynamic diameter, shown rnEigare 2,
possessed a cutpoint diameter of 2.5 ¡rm and a GSD of 1.24. Areview of the performance ofthis
PM-10 Aerosol
from Inlet

Nozzle

Collection Cup
withAntispill Ring

Tmpaction Surface:
Filter Immersed in I mL
Dow Corning 704
Diffi¡sion Pump Oit

PM-2.5 Aerosol
to Filter

Figure 1. Schematic View of the Original Design of the \ryINS Impactor.


Table 2. Critical dimensions of impactors tested.

Jet-to-
Throat plate Nozle
Length Distance lvidrh Jet
'lV. cm s/rv T/W Re
Confisuration T. cm S. cm

Original WINS 2.5 0.80 0.452 1.8 5.5 5200

Bored \4ryNS 0.56 0.80 0.452 1.8 t.2 5200

3-Hole WINS 0.38 0.44 0.300 1.5 1.3 2600

Modified \ryINS 0.61 1.22 0.391 3.1 r.6 6000


100

90

80

70

s 60
io
(ú 50

o
c
o 40
fL
30

20

10

2.0 3.0 4.0

Aerodynamic Diameter, pm

Bigure 2. Penetration Versus Aerodynamic Diameter for Selected Tests in the Impactor
Slope Investigation.
9

unit in response to public comments prompted CASAC to recommend that the size separation
cutpoint be maintained at a Dro of 2.5 ¡rrq but that its performance curve be improved to possess
a GSD of less than 1.2. This section presents the steps taken to meet this goal for the WINS
impactor: 1) identification of the cause of the shallow penetration curve, 2) redesþ of the unit,
and 3) a complete penetration evaluation ofthe new desþ. Ammonium fluorescein solid aerosol
was used for all tests presented in this section.

3.1 Experimentation to Improve Penetration Curve Sharpness

Departures in the destgn of the WINS impactor when compared with that of an ideal
impactor were investigated to better understand the factors which affect the sharpness of the
penetration curve. As discussed above, the original desþ ofthe \ryINS impactor departs from
traditional impactor design 6, îsz"Jegeometry Reynolds number, and impaction substrate
geometry and type. Each ofthese factors was considered individually as well as the possibility
that acombination of them could contribute to impactor performance curve shape. The original
WINS and several prototlrye lszzles were tested with various impaction surface configurations to
aid in this effort.
Fþre2 displays the percent penetration versus aerodynamic diameter for selected
configurations. Figure 3 displays the same selected configurations expressed as penetration in
percent versus the aerodynamic diameter divided by the cutpoint diameter. This figure is useful
for penetration slope comparison. Table 3 presents the cutpoint diameter, the 84ú percentile
diameter, the 16ü percentile diameter, and GSD for each configuration tested.
As a starting point, the original \ryINS was operated with the well impaction substrate
replaced with a flat greased plate. It was conjectured that the extra bends experienced by the
airflow due to the presence of the well could create the opportunity for particle loss leading to a
degradation in penetration slope. Figure 2 shows that losses do occur in the well section with the
cutpoint shifted upwards from 2.55 to 3.05 pm; however, the shape of the penetration curye was
affected only minimatly with the GSD remaining high at 1.22. Thts suggested that the presence of
the well was not the major influence on the shape of the curve.
The performance of two original WINS in series was evaluated as a potential way to
sharpen the size separation characteristics in a manner simila¡ to that presented in Marple et al.
(1987). In this configuratior¡ the cutpoint was reduced to Dro: 2.4 ¡tm and the GSD was
improved to 1.18. This approach did not of[er an attractive solution to a sharper cut due to
increased machining costs, extra maintenance, and the desire of EPA to maintain a single stage
device.
10

100 l

l.
\MNS lmoactor Static Tests l

t-
90 OdginalWNS Design I
L

tr OriginalWlNS Ì
I

H
* Two WINS in Series I
I

80 X Flat Plate
I- Bored WNS
Apiezon Coated Surfaces
70 O Flat Plate
A with Well
s 60
E Large Diam Well
OilSoaked GF Filter
co O Flat Plate
4
(ú Well
50
I^ Large Diam Well
o
co 40
fL
30 J

20

10
J

0.5 1.0 1.5 2.0

D/D5O

Figure 3. Penetration Versus Aerodynamic Diameter Divided by the Cutpoint Size for
Selected Tests in the Impactor Slope Investigation.
11

Table 3. Summary of results oftesting various desþs and configurations.

I)so Ds4 I)ro


Confisuration DIn ü,m um GSI)

Original \ryINS
oiVfilter inwell 2.55 2.05 3.15 L.24
flat greased plate 3.05 2.55 3.8 r.22
two WINS in series 2.4 2.0 2.8 1.18

BoredWINS
Greased surface:
flat plate 3.2 2.75 3.55 r.t4
3.0 cmwell 2.8 2.4 3.5 r.2t
3.7 cmlargewell 3.0 2.6 3.6 1.18
Oil with glass fiber filter:
flat plate 2.8 2.45 3.5 1.20
3.0 cmwell 2.7 2.2 3.2 t.2r
3.7 cmlargewell 2.9 2.35 3.3s 1.19

Three-Hole WINS
flat greased plate 2.4 1.8 3.05 1.30
oiVfilter inwell 2.3 1.65 2.9 1.32

ModifiedDesien 2.45 2.05 2.85 1.18

The next effort was to observe the effect of noz.zle desþ changes onthe impactor slope.
It was conjectured that the long throat length of the originat design could perhaps encourage the
development of a large boundary layer atthenozzlewalls where the particles would have lower
velocities than in the middle of the jet of air. If this was the case, the different particle velocities
would give the particles different inertial energies and hence a cut that was not sharp. Fþre 4
(A) displays prototypetestîoz.zle, referred to as the bored WINS, in which the throat length
a
was reduced dramatically by boring out the interior section of the nozzle. The exterior
T2

A. Bored WINS

Three-hole WINS

-ì l<-
0.38

View from below looking at nonle ip


0.96

D¡ill three holes


1¡¡s diam = 0.300

All dimensions in cenfÍmeten

Figure 4. Experimental WINS Nozzle Desþs: A.) the Bored WINS and B.) the Three-Hole
WINS.
13

dimension of the nozzle had to be increased from the original design to accommodate the bored
noz.zhe. Table 2 shows that the TAA/ ratio was reduced from 5.5 in the original desþ to 1.2 in the
bored WINS ¡¡szzls. This nozzle was evaluated with three impaction plate geometries: a flat
plate, the original well (3 cm diameter), and a larger diameter well (3.7 cm diameter). Each of
these impaction plate configurations was evaluated with a greased surface and with the oiVfilter
substrate used in the original desþ.
The bored WINS with a flat greased plate had a GSD of 1.14 which \ilas a significant
improvement over that ofthe originalWINS nozzlewiththe same impaction surface (GSD:
1.22). With a greased well, the GSD for the 3.0 cm and 3.7 cmwell was 1.21 and 1. 18,
respectivelS indicating that the presence and size of the well with this nozzle does have an impact
on the sharpness of the cut. The GSD was greater with oil and a glass fiber filter impaction
substrate than that with a greased surface in all cases except for the small well. Close examination
ofFigure 3 shows that the presence ofthe filter lowers the penetration curve for particle sizes
smaller than the Dro of the impactor and, hencæ, the GSD is increased. The reason for this is most
likely attributable to surface roughness ofthe glass fiber filter media.
A second ¡szzls,the three-hole \UINS, was designed and constructed as presented in
Figure 4 (B). This unit had three holes as opposed to one in the original WINS and, as a result,
the Reynolds number was reduced from 5200 in the original WINS to 2600. The hole size, W:
0.300 cnL was calculated from classicat impaction theory (Marple and Willeke,1976) to have an
Dro cut size of approximately 2.8 ¡rm for a traditional flat plate impaction surface. The throat
length was such that the TAil ratio was equal to 1.3 and the jet-to-plate distance was such that the
S/JM ratio was equal to 1.5. Tests of the three-hole WINS showed an extremely high GSD for
:
both a flat grease plate impaction surface (GSD 1.30) and a small well with an oiVfilter surface
:
(GSD I.32). The deposits of particles directly under the impaction nozzle were observed to be
irregular in shape indicating that the nszzlss were most likely too closeþ spaced. The multþle
hole desþwas abandoned because ofthe potential for nozzfe clogging with moisture, the
increased complexity of multiple hole construction" and the incompatibility with the well desþ.

3.2 Redesþ ofthe WINS Impactor

Based on the above tests, the WINS was redesigned with the intention of satis$ing the
goal of having a GSD below 1.2, while maintaining the desirable features of the original WINS
desþ. This desigr¡ referred to as the modified WINS and shown schematically inFþre 5,
maintains the single-hole geometry. A conical entrance was adopted to have a smooth transition
from the inner nozzle bore to the final nozzle width to minimize disruption of the air streamlines
moving through the noz-zie and encourage a more uniform velocity profile at the jet exit. A
T4

PM-10 Aerosol
from Inlet

Collection Cup
with Antispill Ring

Impactìon Surface:
Filær Immersed in 1 mL
DowCorningT04
Diffi¡sion Pump Oil

PM-2.5 Aerosol
to Filter

Figure 5. Schematic View of the Modified Desþ of the \ryINS impactor.


l5

3.7 cmwell was designed that has a less intrusive anti-spill ring. Additionally, the screw threads
for this spill-ring were replaced with an o-ring to minimize the possibility of oil leakage.
Appendix A presents detailed machine drawings for this design.
Table 2 displays the critical dimensions of this modified \ryINS. Thenozzlewidthwas
0.391 cm to yield a Reynolds number of 6000, slightly greater than that of the original WINS.
This smaller nozzle diameter was necessary due to lower losses associated with the larger well
and smaller anti-spill ring. The jet-to-plate distance, S, was increased from 0.8 cm in the original
desþ to I.22 cm in the modified design. This change, resulting in a SÂil of 3.1, was adopted in
an effort to minimize interaction between the impactor air jet and the potential conioal build-up of
particles on the impaction surface. The throat lengtt¡ T = 0.61 crn, was maintained short as in the
bored WINS desþ to prevent a large velocþ profile gradient from being formed. The impaction
substrate ofDow Corning 704 otl and Gelman Type AÆ filter was not changed to maintain the
desirable loading characteristics of the original desþ.
Fþre 6 displays the measured penetration versus particle aerodynamic diameter for the
modified and original WINS. The shape of the penetration curves below their respective cutpoint
diameters were similar in both cases; however, a significant reduction in penetration was achieved
with the modified WINS for diameters above its cutpoint diameter. The modified WINS has a Dro
cut size of 2.48 and a GSD of 1.18.
l6

100

90

80

s 70
>
(J
co 60
(J

TU 50
co
ct 40
o
co 30
fL
20

t0

0
3.5
2.0 3.0 4.O

Aerodynamic Diameter, Um

Figure 6. Comparison of the Penetration Curve for the Original WINS and the Modified
WINS.
t7

4.0 EVALUATION OF THE WINS IMPACTION SUBSTRAIE

Several experiments were performed to further evaluate the \{INS impaction substrate.
First, the WINS penetration curye was measured using two types of diffilsion pump oils and th¡ee
t¡pes of filter media. Next, the penetration curve for the \,.NS was measured using two
quantities of oil in the well. Last, an experiment was conducted to determine ifthe oil used in the
well was transported to the afterfilter in either liquid forn¡ via migration, or in vapor fornU via
volatilization.

4.1 Evaluation of the Impaction Substrate Oil and Filter

The WINS was originally designed to use a glass fiber filter immersed in a low volatility
diffi¡sion pump oil. In this configuratior¡ it was intended that the oil be wicked through deposited
particulate matter up into the impaction area to present incoming aerosol with a continuousþ
wetted surface. Alimited survey of available oil and filter candidates was performed fromwhich
two oils and three filters were selected. A qualitative observation of the oil movement through
the filter was performed followed by a quantitative assessment of the penetration curve for each
combination of oil and filter.
Selection of appropriate oils and filters was based on the desired characteristics for each.
The criteria for selection of the oil were: cos! low vapor pressure to minimize volatilizatior¡ and a
viscosþ great enough to prevent migration from the well while maintaining the ability to wick
through the particulate deposit. The study by Turner and Herring (1987) identified that oils with
viscosities in the 30-300 cSt range behaved in a desirable manner in that they could be wicked up
into the particulate deposits while an oil with viscosþ of 1000 cSt performed similar to a grease
a

because they could not adequately coat the depositing aerosol. Selection of the oil was limited to
diffirsion pump oils because of their low vapor pressures (t¡'pically less than 10-7 torr).
The two diffi¡sion pump oils selected as the candidate oils were Neovac (P/N 4120-
K6948-301, Varian Vacuum Products, Woburn, lvfA) and Dow Corning 704 (P/ìI704 dift¡sion
pump oil, Dow Corning Corporatioq Mdland, MI). Table 4 shows that these two single-
component oils have similar characteristics with the exception of composition. The Neovac oil is
hydrocarbon based while the Dow Corning 704 otlis silicon based. The estimated cost per mL
(the oil required in the well of the WINS) is 8 cents and 17 cents for the Neovac and Dow
Corning oils, respectively.
The selection criteria for the filter were: cost, the ability to wick the selected oil into the
central region of the filter, and the ability to maintain structural integfity when immersed in the
18

Table 4. Critical Parameters of Oils Tested.

Neovac I)ow Corning T04

Vendor Matsumura OilResearch Dow Coming


Corp.

Viscosity, cSt 25 @40 "C 39 @25 "C

VaporPressure, mmHg 1x 10{ @25 "C 2 x 10{ @25 "C


Composition Hydrocarbon Single-component Silicon
(mono-N-alþldiphenylether) (tetramethyltetraphenyltri-
siloxane)

Cost per impactor oil change 8 cents 17 cents

selected oil.Table 5 displays the vendor information and other pertinent information for the
candidate filters selected for this series of experiments. Glass fiber filters were originally selected
to be used with the original WINS desþ. Two flatter type media were selected as candidates
because of the possibility that the surface features of glass fiber filters could affect the penetration
for particle sizes below the cutpoint of the impactor. Nuclepore poþarbonate membrane filters
offered an extremely flat surfacg but a higher cost (66 cents each as compared to 14 cents each l
i

for glass fiber filters). Drain discq a media typically used under membrane filters to improve the I
I

spatial uniformity of particulate matter deposites, were selected as the third candidate due to their
I
l-:
I

flat surface (more like the Nuclepore filter) and more porous poþster weave. Porous sintered
I
I

mgtal was considered as a replacement for the filter media; however, was eliminated from
consideration because of the difficulties associated with periodic cleaning such a surface.
A qualitative evaluation of each candidate oil and filter combination was performed by
visually inspecting the movement of the oils in the well as a jet of air impinged on the surface of
the impaction substrate. The jet diameter was roughly equivalent to the WINS impaction noz.zle,
I mL of oil was used in all cases, ild the air flow was metered at 16.7 aLpm. There rüas no
visible difference in the behavior of the two oils; however, the filter media had an effect on the
retention of the oil. The dense weave ofthe glass fiber filter retained a bulk of the oil in the center
of the well with a slight ring of excess oil pushed to the edge of the filter by the force of the
impinging air jet. The drain disc demonstrated simila¡ dynamic qualities but with a larger

Table 5. Critical Parameters ofFilters Tested.


19

Glass Fiber Nuclenore Drain Disc

Vendor Gelman" Costar Corp. Costar Corp.

Catalog or model no. Tlpe AÆ cat# 1108110 cat#23llÛ0


Composition glass fiber polycarbonate spun-bonded
membrane poþster
Pore size, pm 1.0 1.0 not available

Thickness, pm 457 10 not available

Qualities dense \¡/eave of fibers flat, featureless flat, thin, porous


surface

Approx. price each $0.14 $0.66 $0.36

" Whatman type 934AII is an equivalent media.

quantþ of excess oil due to the lower retention qualities associated with this media. The
nuclepore filter appeared to have an even lower holding capacity with a more pronounced ring of
oil with the visible oil cleared from the surface of the filter.
A more quantitative assessment oJthe oiVfilter combinations was performed by measuring
the penetration curve for each when used in the WINS. The static test setup and PSL aerosol, as
described in section 2,wasused for this purpose. Figure 7 displays the results of this experiment.
There was no consistent difference observed between the two oils. The drain disc and the glass
fiber filter behaved with no noticeable differences as well. The only observable difference in
substrate performance rilaÍ¡ detected for the nucþore filter. The penetration measured lvith this
media was consistentþ greater than that measured with the other filters. From this experiment it
was decided that the impaction substrate used by the original'WINS, a glass fiber filter immersed
in Dow Corning 704 difrision pump oil, was an acceptable choice of substrates.
20

100

90

80

70

60
s
c
o 50

o
c 40
q)
fL
30

20

l0

Aerodynamic Diameter, pm

Figure 7. Performance Evaluation of Candidate Oils and Filters for Use as an Impaction
Surface in the \ryINS.
2t

4.2 Operation ofthe impactor at two oil levels

The performance of the impactor was evaluated at I mL and 3 mL oil levels in the well to
gain more insight into the critical nature of this paraÍieter. Penetration for the two conditions was
measured using PSL aerosol. Fþre 8 displays that the penetration is virtually identical for the
two oil levels. The deviation ofthe point at approximately 2.2 ¡rm is believed to be a slight
experimental aberration.

4.3 Evaluation of Oil Carry-over to the Afterfilter

An experiment was performed to determine if the Dow Corning impaction oil could be
carried out of the well and to the after filter resulting in positive measurement biases. Although
the FRM for Plrdr., is exclusively a gravimetric method, research studies will likely include x-ray
fluorescence Q<RF) measurements of the after filter. For this reasorL detection of the oil on the
after filter was performed by )(RF as well as gravimetric means. Gravimetric detection was
simply performed by obsenring differences in pre and post weights of the Teflon after filter. For
XRF detection, a fingerprint of the elemental composition of the oil was first obtained from which
a single element was selected as a tracer.
Gravimetric measurements \üere made with an electronic microbalance (Serial Number
76896, Model C-31, Calm Instruments, Inc., Cerritos, CA). Teflon filters @art Number
R2P1047, Gelman Sciences, Ann Arbor, MI) were pre and post conditioned in a relative humidity
@H: 47 +3%) and temperature (T :75 + 2"C) controlled environment for a minimum of 24
hours prior to being weighed. Pre and post weights were performed by the same operator. The
uncertainty associated with the gravimetric measurements presented in this section is estimated to
be approximately * 10 pg.
All )(RF measurements were performed using the energy dispersive x-ray spectrometer
owned and operated by the EPA at their National Exposure Resea¡ch Laboratory. A fingerprint
ofDow Corning 704 dift¡sion pump oil was taken in preparation for XRF detection. A single-jet
nebulizer (Reorder Number 5207, Professional Medical Products, Inc., Greenwood, SC) was
operated at 15 psi to spray the oil into a mist. The oil mist was then diluted with I cfrn of filtered
compressed air. The diluted mist was then sampled onto three Teflon filters at a flow rate of
L6.7 aLpm over a 1 minute time period. These filters were then analyzed by )(RF to obtain the oil
fingerprint. Table 6 shows that silicon was detected in each of the three filters at a level rangrng
from 5.1 to 6.7 times the background detection limit and was therefore used as the fingerprint
element.
22

100

90

80

70

60
s
c
o 50
(Il
(l)
c(¡) 40
fL
30

20

l0

Aerodynamic Diameter, pm

X'igure 8. WINS Penetration Curve with 1 mL and 3 mL Oil.


23

Table 6. Results ofthe Oil Carry-overExperiment.

Gravimetric Analysis )GX'Analysis


Sample
llme lVeig[t Masg Elements ¡bove 3
(hours) Gain Conc.. tines)ßF )(RF SiMass
lmol fuolm3l uncertaintrf lnølcmî¡

Dow704 16.93 S(5.1), zûQ.1) 95,000 + 18700

Dow704 7.46 5(6.4), Cu(3.8), 67888 + 10625


zri(6.7)

Dow704 5.54 s(6.6) 53204+7960

Neovac none

Neovac c(3.3)
Neovac none

FieldBlank 1" 24 4.012 <0 none 27.1+13.0

FieldBlank 2 24 {.013 <0 none 28.8 * 13.9

FieldBlank 3 24 {.016 <) none 34.1+L4.4

FieldBlank 4 119 {.009 <0 none 0.6 + 13.3

Sample I 24 0.021 0.88 s(3.s), s(5.0), 54.9 + 15.5


K(9.0), Fe(3.9)

Sample 2 24 0.018 0.75 none 2I.8+14.3

Sample 3 24 4.002 <0 s(4.2) -5.7 t13j


Sample 4 119 0.011 0.09 Al(4.0), si(4.8), 19.6 +L6.6
517.9). Fel3.1)

" Numbers in parenthesis are number of times greater than the analytical uncertainty.
b
Area of deposition on filter is approximately 13.5 cm2.
"A field blank is a Teflon filter transported to the field with the sample filter with the
corresponding number. The field blanfrs were used to gauge gravimetric repeatability and to
provide blank filters for XRF analysis.
24

Three l-day samples and one 5-day sample were obtained by sampling air through a high-
efficiency particulate air (IIEPA Model Ultra Filter Cartridge, TWe Itr Part Number 95302,
MS.\ Pittsburgt¡ PA) filter prior to entering the WINS impactor. This filter has a reported
efficiency of 99 .98Yo for particles of 0.3 Fm in size at a 20 Lpm flowrate. A pump system with a
variable speed motor controlled with a laminar flow element (Serial Number 00711, Model SP-
280, ADE, Inc., Naples ME) was used to pull air through the entire system at 16.7 alpm. The
samples were collected on the roof of the EPA Annex Building (79 Alexander Dr., Resea¡ch
Triangle Parþ NC) during early July. The temperatures during sampling approached the upper
limit of the proposed FRlvf, rangng from approximately 30'C at night and 40 'C during the
daytime hours. Performing this experiment at these elevated temperatures, as opposed to
laboratory conditions, was done to produce a worst case oil migration situation with the oil vapor
pressure approaching a morimum and the viscosity approaching a minimum for this application.
Table 6 displays the results ofthis experiment. The field blanks had negative weight gains
with the XRF silicon content of the blank filters all below three times the experimental
uncertainty. The sample filters had weight gains ranging from 21 ¡tgto -2 ¡rg. The mass
concentration corresponding to these weight gains was determined less than I lrdmt for all four
samples. The silicon content measured on these filters was slightþ detectible for sample 1 of the
one-day samples and sample 4,the five-day sample. These values are considered to be negligible
indicating that carry-over of the diffi.¡sion pump oil is not a significant source of measurement
bias.
25

5.0 EVALUATION OF THN WINS IMPACTORAFTER COARSE MODE


AEROSOL COLLECTTON

A criticat aspect of a size separation device that is commonly overlooked is its ability to
maintain its performance cha¡acteristics for an extended period of time. As discussed earlieq the
well impaction surface of the WINS was designed with the intention of minimizing the change of
performance during periods of high concentration coarse particle sampling such as during a wind
blown dust storm. In this experiment, the entire FRM sampling train (the Graseby-Andersen
2468 inlet and the WINS impactor) were exposed to a laboratory-generated, highJevel, coarse-
mode aerosol over an extended period of time to challenge this desþ. Measurement of the
penetration curve were then performed after discrete loading levels.
A special pu{pose toading apparatus was designed and constructed underWA 10 (EPA
contract number 68-D5-0040) for the purpose of conducting these experiments. Abrief
description of this systenr" shown in Fþre 9, is presented here; more detail may be obtained from
RTI, 1996b. Aventuri-type aerosol generator \ila¡¡ used to aerosolize a standard test dust. The
aerosolized dust was then charge neutralized, diluted to a known concentration, and injected into
the tunnel. The aerosol was passed through a mixing baffie and then transported at a low velocþ
(nominally 20 cm/s) to the test section. Air flow was provided by a high-volume blower motor
which operated at approximately 60 cfrtt.
Anzonatest dust (ATD, 0-10 pnr, Powder Technology Incorporated, Burnsville, MN)
was selected for this experiment because it is considered to represent the greatest challenge to the
impactor substrate. The size distribution of this dust was measured in the loading tunnel with an
Mdwest Resea¡ch Institute cascade impactor to have a mass median aerodynamic diameter of
approximately 5 pm and a GSD of approximately 2. Anattempt was made to maintain the
concentration of the dust at 300 pglm3 throughout the loading experiments.
The upper section of the 2468 inlet was installed into the test section so that the inlet
sampled aerosol exclusively from inside the test section. AWINS impactor with a freshly
prepared well substrate followed by a filter assembly was connected to the 2468 inlet to complete
the FRM sampling train. Atotal filter assembly was collocated in the test section to determine the
total mass concentration in the tunnel. The flowrate through the FRM sampling train was 16.7
al,pm and that for the total filter was maintained at flow rate to provide for isokinetic sampling of
the tunnel aerosol. After each 24hour sampling period, the average mass
26

Side View of Tunnel

Hieh
HEPA Tunnel lnside Dimensions: Volume
Filter 34 cm by 34 cm HEPA Blower
Filær

Charge Neuüalizer c__._\


Toúal Filær
FRMSampling
Train

Top View of Dust Generator

DustCloud
into Tunnel
l

Compressed Air Inlet

Sonic Nozzle
with Venturi
DnstPick-up Tumtable with
ArizonaRoadDust

Figure 9. Test Apparatus Used to Load the WINS with Arizona Test Dust.
27

concentration wrls determined from the tot¿l filter weight gain and the sampling flow rate. The
WINS impactor was transported to the static test apparatus for determination of penetration
versus aerodynamic particle size. Penetration measurements were conducted using PSL as
described in Section 2 ofthis report.
There were three 24-hour sampling periods conducted with the corresponding mass
concentration in the tunnel 370,301, and326 pgln,3. The impaction of the ATD onto the well
substrate created a visiblg symmetrical cone of material on the surface of the glass fiber impaction
filter directly underneath the impactor nozzle. This deposited ATD was observed to be a wet-
looking, dark grey-brown color as opposed to the dry, light grey bulk AID because of the
capillary action which pulled the oil into the cone. The surface of the deposit was also observed
to be wet to the touch. Fþre 10 displays that the presence of this cone increases the collection
efficiency and causes the penetration curve to shift to the left. This is unlike most greased-plate,
conventional impactors in which the penetration curve shifts to larger sizes due to particle bounce
as the substrate loads with dry material.

5.3 Numerical Estimation of WINS Performance Versus Substrate Loading

As described in the previous section, evaluation of the WINS impactor under loaded
conditions revealed that an approximately L}Yo reduction in sampler cutpoint could be expected
following sampling of a coa¡se mode aerosol (mean concentration equaled 332 pglms' for three
continuous days. It is important to note, however, that a l0% reduction in cutpoint following this
loading level does not directly translate into a l0% reduction in measured mass concentration. In
reality, biases in expected mass concentrations are a complex function of the aspirated arnbient
particle size distribution as well changes in the position and shape of the fractionation curve. This
section describes numerical techniques which were developed for quantitatively estimating
changes in expected man¡s concentration as a function of substrate overloading. While these
techniques were specifically designed to apply to effects of substrate loading, the same approach
can be used to perform sensitivity analysis of sampler performance with respect to other important
variables such as flow rate biases, machining intolerances, shape and position of candidate
equivalent fractionator performance curves, and a¡nbient temperature and pressure variations.
For a specific sampler operating at a fixed sampling flow rate, the expected fine particle
mass concentration is primarily a function of the size distribution of the arnbient aerosol at the
sampling site. Intuitively, if the site possessed a¡nbient aerosols exclusively in the sub-micron
regiorç the expected fine particle concentration would equal that of the anrbient total
concentration. Moreover, the fine particle measurement would be virtually independent upon
28

às
c
o
ct
(t)
c(l)
fL

Aerodynamic Diameter, pm

Figure 10. \ryINS Performance Clean and After being Loaded with Arizona Test Dust over
MultipleDays.
29

minor changes in the shape and position of the effectiveness curye. Conversely, if the site
aerosols consisted exclusiveþ of a¡nbient aerosols significantly larger than the fractionator's
cutpoint, the expected fine particle concentration would represent only a small fraction of the
total aerosol mass concentration. Agarr, moderate changes in the fractionator performance would
have a relatively minor effect on the measured fine particle mass concentration.
In reality, of course, a¡nbient aerosols are rareþ confined to size ranges significantþ above
or below the adopted 2.5 ¡rm cutpoint for fine particles. Ambient field studies conducted during
the last thirty years have consistentþ confirmed that ambient aerosols typically cover a wide range
of sizes @urton and Lundgreq 1987, Lundgren and Paulus,1975, Whitb¡ 1978). These size
distribution studies typically identify multimodal peaks as a function of particle size with
differences in peak locations due primarily to differences in particle formation mechanisms. In
general, submicron aerosols are formed primarily by processes of combustion, condensation, and
gas-to-particle conversion. Conversely, supermicron aerosols result from the mechanical breakup
of larger parent material and a¡e thus produced by processes such as grinding, crushing, and
mechanical abrasion. The locations and relative heights of the fine and coarse mode peaks vary
significantþ as a fiinction of location, time, and local and regional emission activities. In general,
therefore, the presence of both fine and coarse aerosol formation mechanisms can cause fine and
coarse modes to overlap and are typically found between 1 ¡rm and 3 ¡rm aerodynamic diameter.
Since the 2.5 pmcutpoint of the fine particle FRM sampler generally exists near the minimum in
the intermodal region between fine and coarse mode peaks, variations in the FRM's performance
due to loading can be expected to create rather minimal changes in expected mass concentration.
To predict the magnitude ofthese changes in measured mass concentrations, it was first
necessary to select some readily accepted particle size distributions for the a¡rrbient aerosol.
Because no single aerosol (either measured or hlpothetical) can accurately model the arnbient air
at all sampling sites and in all sampling situations, it was necessary to adopt different distributions
which encompas¡s the expected distribution range under most sampling situations. Table 7
presents critical distribution parameters ¡ot ¡dsalized "coarse", "qpical", and "find' arnbient
aerosols. For purposes of comparability, each distribution was normalized to a total mass
concentration of 100 ¡rglm3. As a qualitative measure of the fine particle fraction of each
distribution" Table 7 also includes calculated Pllr,/PMro ratio. For purposes of this calculatiorq
PMr., fractions were calculated using a step function and PMro fractions were calculated using the
Chan-Lippman regression curye. In increasing order of Plvf""rPMro ratios, the three distributions
were:
Table 7. Critical parameters ofidealized ambient particle size distributions.

trïne Particle Mode Coar¡e Partlcle Mode ¡l¡¡y garnpler


Idealiz,ed PMr/PM,o Expected
Distribution Ratlo Ma¡¡ Conc.
MMI) GcG Conc. MMI) G€o. Conc.
0un) Std.Dev. ú€/d) Std.Dev (lrgln) Ggln)
Grn)

Coarse 0.50 2 t2.o 10 2 88.0 0.n 13.8

rTypical' 0.50 2 33.3 10 2 66.7 0.5s 34.3

Fine 0.85 2 85.0 l5 2 15.0 0.94 78.5


31

ldealized Coarse Ambient Distribution

MMD = 0.5 pm
GSD = 2.
Conc. = 12

co
E
8o-

40

S-s
cn4
CD<
(u(Ð

ob
S
:= =t' 20

E
t-
zo

0.01 0.10 1.00 10.00 100.00

Aerodynam¡c Diameter, pm

Figure 11. Idealized Si2e Distribution Representing a Site ofPredominantly Coarse


Ambient Aerosols.
fi^ n, A¿"-=*e
trxeÉ L Ù,4"^^ _fvLaSS ),il"r,^
_ _

/-)
(-/
32

ttCoa'rsett (Fþre 11) was developed based upon recent


- This idealized distribution
trichotomous sampler measurements performed in Phoenix during the
sunìmer (Lundgreq et al., 1996). The calculated Pìlf"'/Pì{' value for this
distribution isO.27 which is close to the monthly average measured by the
tricot. Note from the figure that the distribution has a intermodal minimum
of approximately 1.5 ¡rm aerodynamic diameter.

"T5pical" - This distribution presented in Figure 12 is somewhat similar to that used


during effectiveness calculations ofPlvfro samplers (EP,\ 1987). Although
no single distribution represents all sampling sites, the PI\[ /PMto ratio of
this distribution is 0.55 which is close to the 0.57 value reported using
2269PI:N'{2.JPIo'[.L. data points for the conterminous United States (EP,\
1996). This distribution has a minimum at approximately 2 pm.

ttFinet'- This distribution (Fþre 13) was idealized based on \il'alter John's "droplet
mode" measurements (Iohn et al., 1990) and is used to determine if the
go\¡¡th of hygroscopic aerosols above 0.5 ¡rm has an effect on the
sensitivity analysis. In this distribution" the droplet mode peak is located
near its ma<imum theoretical limit. Note that the slope of the droplet mode
is fairly sharp at 2.5 pm. APM,./PI\{' ratio of 0.94 was oaloulated for this
distribution.

Predicting the expected mass concentration measured by a given size selective sampler
involves numerically integrating a sampler's effectiveness curve with the sampled arnbient particle
size distribution. For integrating effectiveness with ambient distributions, a spreadsheet was
developed which divides each anrbient distribution into 380 discrete, geometrically spaced size
intervals rangmg from 0.005 pmto 10.0 pm. For fitting experimentally determined sampler
ef[ectiveness curves, it was determined that most commonly-used routines (lognormal, cubit,
logistic, sigmoidal, etc.) tended to work adequately if the effectiveness curve was symmetrical
about the cutpoint. Significant ñtting errors were found to occur with these algorithms, however,
if the effectiveness curve displayed any appreciable asymmetry. In evaluating these curves, a five-
parameter, asymmetrical sigmoid function was adopted and appeared to fit both symmetrical and
asymmetrical functions well. Once the effectiveness function was knowrU the function was¡
divided into the same discrete intervals used for the a¡nbient distributions. The expected mass
JJ

ldealized Typical Ambient Distribution


Fine Mode: Coarse Mode:
MMD = 0.5 pm MMD = 10 Um
GSD=2 GSD=2
Conc. = 33.3 pg/m3 Conc. = 66.7 pg/m3

c
.9

(l)
c)o-
crì
Icn
(Jo
üi€
G'
CU
20

tt tt
o=L
.N i

(u
E I
I

zo
I

0.01 0.10 1.00 10.00 100.00

Aerodynamic Diameter, pm

Figure 12. Idealiz.ed Ambient Particle Size Distribution Considered to be "Tlpicali' of


Conterminous AIRS Sites.
34

ldealized Fine Ambient Distribution


Coarse Mode:
MMD = 15 pm
GSD=2
Conc. = 15 pg/m3

co

Þ
c
8o-
caì 40
l¿
c)oct)
sd
o<È

ft=
:=
(E
20

E
zo

0.01 0.10 1.00 10.00 100.00

Aerodynam¡c Diameter, pm

FÍgure 13. Idealized Size Distribution Representing a Site of Predominantþ Fine Ambient
Aerosol.
35

for each range was then calculated as the mass associated with each interval times the
effectiveness value for that interval. The expected fine particle mass concentration was then
calculated as the sum of all the interval mass concentrations. Sensitivity analysis ofthis technique
revealed that increasing the number of discrete intervals over the same size range did not
appreciably improve the accuracy of the model.
Table 8 summa¡izes the results ofthe numerical modeling of expected fine particle mass
concentration versus coarse aerosol loading time. As was presented in Figure 10, the effeø of
accumulated mass deposits on the impactor stage is to cause a gradual increase in the stage's
collection efficiency which corresponds to a reduction in fraotional penetration. This behavior
intuitively reduces the mass penetrating to the after filter and thus results in a negative bias in the
fine particle concentration measurement. As Table 8 indicates, however, the magnitude of this
predicted bias is relatively minor even at these high loading values. Following three consecutive
days sampling this hypothetical, high concentration coarse aerosol, a ma:rimunç negative bias of
orúy 5Yowould be predicted for any ofthe three idealized distributions. Since the effect of
substrate loading is primarily a function of deposited mass, this same change in performance
would be expected in sampling half the 330 pglm3 test concentration for twice the original
sampling duaration (i.e. 165 pglmt for 6 consecutive days).
These quantitative estimates can be easily visualized by considering the resulting loading
curyes superimposed on the idealized "typical" anrbient particle size distribution (Fþre 14).
Since the expected Pì4., concentration represents the integrated area of the anrbient mass
distribution under the effectiveness curve, one would prediø a relatively small change in expected
mass as a function of loading. These expectations are verified by the results of the numerical
modeling presented above.
Table 8. Predicted Measurement Bias as a Function of Coarse Aerosoll Sampling Time

After ls Day at After 2nd Day at After 3rd Day at


CIean 370 uslms 301 uelm3 326 nslms
Ambient
Aerosol Size Predicted Predicted Predicted Predicted
Predicted Predicted Predicted Predicted
Distribution PMr., PMr.. PMr.t Ptr4.,
Bias vs. Bias vs. Bias vs. Bias vs.
Conc. Conc. Conc" Conc.
Clean Unit Clean Unit Clean Unit Clean Unit
luslm3ì luslm3ì lusy'msl llslm3l
Coarse 13.81 0.0% 13.57 -1.8% 13.30 -3.8% 13.t2 -5.3Yo

"Tlpical" 34.28 0.0% 33.92 -r.t% 33.63 -r.9% 33.43 -2.60/o

Fine 78.54 0.0% 77.26 -r.7% 76.36 -2.9% 75.66 -3.8o/o

I Mean coarse aerosol concentration for these tests was 332 pglfif .
37

co
ReroSá for 3 Days y
g
c(1)
s rJ o-
ctì
io ldealized Typical" Mass SE,
.^J
(ú Distribution 20 üíp
(Úò
o
co >E
fL -\eo)
q)f-
N
:=
at
E
o
z.

0.01 0.1 I 10 100


i

Aerodynam¡c Diameter, Um
i

Figure 14. WINS Clean and Loaded Performance Superimposed on an Idealized "Tlpicaf' i

Mass Distribution. i

t.
i
38

6.0 SI]MII{ARY AND CONCLUSIONS

The WINS was successfully modified to sharpen the penetration versus particle
aerodynamic diameter curve in response to recommendations by CASAC that the GSD be less
than 1.2. A reduction in the impactor nozzle TAil ratio from 5.5 to 1.6 and increasing the
diameter ofthe impaction well from 3.0 cm to 3.7 cm effectively reduced the GSD from 1.24 to
1.18. Further modifications lvere made to increase the SAil ratio and to make the anti-spill ring
less intrusive.
Neovac and Dow Corning diff¡sion pump oils demonstrated similar performance
characteristics when used as \ryINS impaction substrates. Glass fiber filters and drain discs
performed similarly; however, nuclepore filters had a slight increase in penetration perhaps due to
particle bounce. The original WINS impaction substrate, a Gelman T¡pe AÆ glass fiber filter
immersed in Dow Corning 704 diftision pump oil, was not modified based on these results.
Further tests indicated that the penetration curve was the same with the well containing 1 mL or 3
mL of difr¡sion pump oil. Additionally, the diffiision pump oil was not carried over to the a.fter
filter in significant quantities when measured gravimetrically and arnlyzedby )RF.
Lastly, the penetration characteristics of the WINS was evaluated after being loaded with
a laboratory generated coarse mode aerosol. The coarse aerosol was observed to create a conical
build-up ofmaterial on the surface ofthe impaction substrate. This served to make the WINS
more efficient at collecting particles, thereby reducing the cutpoint ofthe unit. Numeric
modeling however, indicated that this shift in separator performance due to loading would
produce relativeþ minor biases in fine particle measurement.
39

7.0 RETERENCES

Burtorq R.M. and Lundgren" D.A. (1987). "Wide Range Aerosol Classifier: A Size Selective
Sampler for Large Particles", Aerosol Sci. Technol. 6:289-301.

Johr¡ W., Wall, S.M., Ondo, J.L., WinHmayr, W. (1990). "Modes in the Size Distribution of
Atmospheric Inorganic Aerosof', Atmos. Environ. 9:2349-23 59.

Lundgrer¡ D.A. and Paulus, H.J. (1975). *The Mass Distribution oflarge Atmospheric
Particles", IAPCA 25 :1227 -L23 l.

Lundgrer¡ D.A, Hlaing, D.N., Rictr" T.4., and Marple, V.A (1996). '"M'/PM /PNf, Data from
a Trichotomous Samplef', Aerosol Sci. Technol. 25 :3 53 -3 57 .

Lundgreq D.,A. (1996). "Review and Comment on: Three EPADraft Documents and Discussion
at the March I, 1996 Public Meeting of the CASAC Technical Subcommiuee for Fine
Particle Monitoring". Letter to CASAC.

Marplq V.A and Willekq K. (1976). "Impactor Design", Atmospheric Environment, 10:891-
896.

Marplg V.A. and Rubow, K.L. (1986). "Theory and Desþ Cruidelines". In "Cascade Impactor"
edited by J.P. Lodgg Jr and T.L. Chan. pp 79-101, American Industrial Hygiene
Association, Akron, OH.

Marple, V.4., Rubow, K.L., Turner, W., and Spengler, J.D. (1987). "LowFlowRate Sharp Cut
Impactors for Indoor Air Sampling: Desþ and Calibrationl', JAPCd 37 1303-1307.

Marple, V.A. (1996). "Critique ofEPA's Pld* pm Impactor Samplef'. Letter to CASAC.

RTI Project Report (1990). "Test of the Sierra-Andersen 2468 Dichotomous Sampler lurúet at2,
8, and ?4I<NIIII', Wind Tunnel Test Report, EPA Contract No. 68-02-4550,
Environmental Protection Agency, Research Triangle Parþ NC.
40

RTI Project Report (1996a). "Development and Evaluation of Sampling Components for
Measuring Particulate Matter Under 2.5 Mcrometers", Task I-OOI,EPA Contract No.
68-D5-0040, Environmental Protection Agency, Research Triangle Parþ NC.

RTI Project Report (1996b). Report to be writter¡ Task tr-010, EPA Contract No. 68-D5-0040,
Environmental Protection Agenry, Research Triangle Parþ NC.

Turner, J.R. and Hering S.V. (1987). "Greased and Oiled Substrates as Bounce-free Impaction
I
Surfaces". Aerosol Sct, 18:215-224.

U.S. Environmental Protection Agency: Revisions to the National Ambient Air Quality Standards
forParticulate Matter (1987). Fed. Reg. 52:24634-24750.

U.S. Environmental Protection Agency: Air Quality Criteria for Particulate Matter, Volume I of
m (1996), Document # EPA/600/P-95/001aF.
*The
Whitby, K.T. (1978). Physical Characteristics of Sulfur Aerosols", Atmos. Environ. 12:135-
159.
t-

4t

APPENDD( A
Machine Drawings of the Modifie.d EPAWINS Impactor

These drawings were reprinted from a draft of the FRM for fine particulate matter dated
September 24,1996 (Appendixl 40 CFRPart 53). The drawing figure numbers have not been
modified to avoid confusion.

You might also like