&EPA
United States
Environmental Protection
Agency
Environmental Sciences Research EPA-600/4-78-034
Laboratory July 1978
Research Triangle Park NC 277 1 1
Research and Development
Aerosol Analysis for
the Regional
Air Pollution Study
Interim Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/4-78-034
July 1978
AEROSOL ANALYSIS FOR THE REGIONAL AIR
POLLUTION STUDY
Interim Report
by
F.S. Goulding, J.M. Jaklevic, and B.W. Loo
Lawrence Berkeley Laboratory
Berkeley, California 94720
Interagency Agreement No. EPA-IAG-D6-0670
Project Officer
Thomas G. Dzubay
Atmospheric Chemistry and Physics Division
Environmental Sciences Research Laboratory
Research Triangle Park, N.C. 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, N.C. 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Approval does not signify that the contents necessarily
reflect the views and policies of the U.S. Environmental Protection
Agency, nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
This report was prepared as an account of work sponsored by the
United State Government. Neither the United States nor the Department of
Energy, nor any of their employees, nor any of their contractors,
subcontractors, or their employees, makes any warranty express or
implied, or assumes any legal, liability or responsibility for the
accuracy, completeness or usefulness of any information, apparatus,
product or process disclosed, or represents that its use would not
infringe privately owned rights.
This report was done with support under an interagency agreement
with the Department of Energy and Environmental Protection Agency. Any
conclusions or opinions expressed in this report represent solely those
of the author(s) and not necessarily those of The Reagents of the
University of California, the Lawrence Berkeley Laboratory or the
Department of Energy, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
ii
-------
ABSTRACT
An aerosol sampling and analysis program was conducted as part of
the Regional Air Pollution Study in St. Louis. Ten automatic dichotomous
samplers were operated in the field for two years and collected 35,000
samples. The procedures used for analyzing these samples for total mass
and elemental composition are described in detail. The characteristics
of the betagauge mass measurement and energy dispersive x-ray fluorescence
analyses are discussed, together with the factors that affect the
precision and accuracy of the data.
111
-------
CONTENTS
Abstract iii
Figures vi
Tables vii
Acknowledgements viii
1. Introduction i
2. Aerosol Sampling Procedures 2
3. Mass Measurements 5
Relative humidity effects g
Atomic number dependence g
Filter porosity effects 10
4. X-ray Fluorescence Measurements 12
Sensitivity 12
Precision and accuracy 12
Calibration accuracy 23
Reproducibility of results 29
5. Data Handling 35
References 39
-------
FIGURES
Number Page
1 Map of the St. Louis area showing the location of 10 RAMS
sites equipped with dichotomous samplers 3
2 Cut and loss characteristics of the virtual compactor 4
3 Cross section of (3-gauge apparatus 7
4 Mass gain of 4 mg/cm2 cellulose membrane as a function of %
relative humidity 9
5 Minimum detectable limit for x-ray fluorescence analysis
compared with average aerosol concentration ranges 15
6 Voltage dependence of x-ray yield in x-ray fluorescence unit.
Vertical low on curve represents normal operating voltage. ... 17
7 Cross section of secondary fluorescence geometry 18
8 Scans of sensitive region of x-ray fluorescence unit ... 19
9 Diagram of spectrum stripping procedure 22
10 Illustration showing matrix absorption effect (A) and large
particle attenuation (B) 28
11 Flow chart showing the overview of sampling, analysis and
data merging 35
VI
-------
TABLES
Number
1 Atomic Number Dependence of Beta-gauge Mass Attenuation
Coefficient 10
2 Operating Conditions for Sampling and Analysis 13
3 Sensitivity for Energy Dispersive X-ray Fluorescence
Analysis 14
4 Calibration Solutions for Thin Aerosol Deposited Standards 25
5 Summary of Calibration Measurements 27
6 Particle Size Attenuation Corrections for the Light Elements
(Adapted from Reference 11) 30
7 Reproducibility of Air Samplers Measured by Side by Side
Sampling 31
8 Root Mean Square Deviations of Analyses of Identical Samples
Over a Three-Month Period 32
9 Comparison of our Analyses with Independent XRF Measurements
by R. Giauque 34
10 Calculations Performed During Data Analysis 38
vn
-------
ACKNOWLEDGMENTS
There are a number of people in the Instrumentation Development Group who
have made significant contributions to the project over the past two years.
R. Gatti and W. Searles are responsible for the day to day operation of the
analysis equipment. D. Landis, R. Adachi, N. Madden, J. Meng, B. Jarrett and
others have made significant contributions to the design and implentation of
the hardware. J. Llacer and A. Thompson are responsible for much of the com-
puter programming used in the data handling.
We acknowledge the technical advise and assistance of T. Dzubay and
R. Stevens of EPA. R. Giauque and his associates at LBL have assisted the pro-
gram by providing assistance in the calibration of the x-ray fluorescence
instrumentation. R. Fischer and his associates have done an admirable job of
attending to the around the clock operation of the analysis equipment.
viii
-------
SECTION 1
INTRODUCTION
The activities of the past year have focused on our participation
in the St. Louis RAPS program. As of March 1, 1977, a total of 35,000
air particulate samples have been collected at the 10 selected RAMS
sites. These samples consist of membrane filters on which the fine and
coarse particles have been collected separately, using automatic dicho-
tomous air samplers. The samples are returned to Lawrence Berkeley
Laboratory (LBL) for analysis after collection in the field.
The program within the laboratory involves the mass measurement of
the deposits, using beta particle attenuation measurements which are per-
formed before and after particle collection. Elemental analysis of the
sample for 28 elements is obtained with pulsed excitation X-ray fluores-
cence analysis. At present, these measurements are continuing until the
complete set of 35,000 samples has been analyzed. As of April 1, 1977,
mass determinations have been made for 10,000 aerosol samples and com-
plete X-ray fluorescence analyses have been performed on 15,000 samples.
The complete analysis for total mass and elemental composition should
be completed for all the samples by August 1977.
As the analytical data are generated, the output tapes are being
processed and submitted to the data bank at Research Triangle Park (RTP).
This data processing includes the corrections for particle size effects
in XRF analysis, cross contamination of the fine and coarse particle
sizes in the sampler, and other systematic effects in the sampling and
analysis. The mass data and elemental analyses are then merged for the
individual samples and the data set is ordered chronologically for the
specific stations. This data processing has been performed for approxi-
mately one calendar year of sampling and the results transmitted to RTP.
At present, a more efficient disk-oriented data processing system is
being developed for processing the remaining data. It is planned that
the entire data set will be reprocessed shortly after the August 1977
completion of the analysis and made available in a condensed format.
This reprocessing will also include correction of the sulfur data to
account for the penetration of particles into the filters which has been
observed under some conditions.
The following sections of this report will discuss, in detail, the
procedure used in the sampling, analysis, data processing and validation
of the data.
-------
SECTION 2
AEROSOL SAMPLING PROCEDURES
The sampling network consists of ten selected RAMS sites which are
equipped with automatic dichotomous samplers developed and constructed
by LBL in previous programs1*2). Figure 1 is a map showing the location
and station number of the selected sites. The sampling sites are sta-
tions in the St. Louis Regional Air Monitoring System (RAMS) which have
been modified to accept the dichotomous samplers. The inlets of the
samplers are connected to a 10 cm diameter aerosol sampling manifold
whose inlet is situated 4 m above ground level. A flow of 1093 £/m is
drawn through this tube and 50 H/m is diverted isokinetically into the
inlet of the automatic dichotomous sampler.
The sampler is equipped with a two-stage virtual impactor which
separates the incoming particles into two size fractions above and be-
low 2.4 ym mass median diameter. The coarse and fine particle size frac-
tions are then collected separately onto 1.2 ym pore size cellulose ester
membrane filters. These 37 mm diameter filters are individually mounted
in 5.0 x 5.0 cm plastic holders which are carried in a standard 36-slide
projector cartridge. The use of the cartridge for the automatic handling
of the samples eliminates contamination and reduces possible sources of
operator error in the sequencing and ordering of the samples for ship-
ment to and from the analysis laboratory.
The characteristics of the dichotomous sampler are illustrated in Fig. 2.
The cutpoint DSQ (particle diameter at which 50% of the particles are
collected on the coarse particle filter) was measured to be 2.4 ym.
The cut characteristics are sharp with a ratio of DQI+ to DSQ of 1.10.
The solid particle losses are quite low. The maximum losses occur near
the 2.4 ym cutpoint, which corresponds to the normal minimum in the ur-
ban aerosol size distribution3).
As evident in the characteristics shown in Fig. 2, a small fraction
(5%) of the fine particle mass is collected as a part of the coarse par-
ticle fraction. The remaining 95% of the fine particles are collected
on the other filter. Correction for this 5% interference between size
fractions is made at the time of the final data processing (see Table
10).
Constant flow rate within the sampler is maintained by monitoring
the pressure differential between the inlet and the second stage of the
virtual impactor, (i.e., across the inlet orifices). This pressure dif-
ferential is maintained constant by adjusting a variable impedance ori-
fice included in the pump circuit as part of a feedback-loop. The flow
calibration has an accuracy of ± 2% for room temperature air, with a
repeatability of 0.5%. Periodic checks on the samplers indicate that
the flow calibration has remained constant to within 1% for nearly two
years.
-------
XBL 764-1126
Figure 1. Map of the St. Louis area showing the location of 10 RAMS sites
equipped with dichotomous samplers.
-------
-ft-
A/(A+B)
LIQUID PARTICLE LOSS
SOLID PARTICLE LOSS
4567
PARTICLE SIZE (pm)
8
10
XBL 751-124
Figure 2. Cut and loss characteristics of the virtual impactor.
-------
The temperature dependence of the flow control system results in a
mass flow which is inversely proportional to the square root of the ab-
solute temperature. Since short-term fluctutations in the pollutant
level generally exceed effects due to the temperature variations, no
corrections are currently applied. If one is calculating monthly aver-
ages from the data, it would be appropriate to apply a temperature cor-
rection at that time.
The normal sampling schedule in St. Louis consisted of 12-hour
sample periods at all stations, except #103 and #105, where 6-hour sam-
ples were standard. These latter stations routinely experienced higher
particulate concentrations, which caused filter clogging in the longer
sampling intervals. During an intensive study period during the summer
of 1976, the schedule was modified to accommodate 6-hour samples at most
stations, with 2-hour samples for stations #103, #105 and #112.
The increased flow impedance, caused by particle loading on the mem-
brane filter, is normally automatically compensated by a reduction in the
impedance of the flow control valve. In cases where the particle loading
becomes excessive, the range of the flow control valve may be insufficient
to compensate for the clogged filter. The full range of the flow control
valve could accommodate an increase in impedance to twice the normal
value. For the 1.2 ym membrane filter used in the study, fine particle
mass loading of 200 mg/cm2 or greater results in 70% increase in the fil-
ter impedance which approaches the clogged condition.
The procedures followed in the sampling program were carefully con-
trolled to ensure valid data. Individually numbered clean filters were
loaded into separately numbered 36-sample cartridges at LBL. The sample
identification number and the corresponding tare weight were recorded on
magnetic tape at the time of the initial beta-gauge measurement. The
samples were then shipped to St. Louis for exposure in the samplers.
The location and time of sampling were recorded both on the cartridge
labels and in a separate sampling log. This log also contained checklists
for site visits and notes of any irregularities in the sample routine.
The exposed filters were returned to LBL, together with the sampling
information. Final weights were recorded and entered for each individual
slide number. At the time of the X-ray fluorescence measurement, the
data pertinent to the sampling conditions are entered into the computer
system for subsequent data processing. The redundant sampling information
allowed an accurate reconstruction of the sampling condition.
-------
SECTION 3
MASS MEASUREMENTS
The total mass of particles collected in each size fraction is raeas-
sured using an automated beta-particle attenuation method. The technique
relies on the exponential dependence upon mass which the intensity of a
continuous beta-particle spectrum exhibits when a variable thickness is
placed between the radioactive beta-source and a suitable detector.
Figure 3 is a schematic of the beta-gauge showing a 300 vie
source, mounted in the upper vacuum chamber with a 2.5 cm diameter Si
semiconductor in the lower chamber. The large area detector and large
source to detector distance result in a uniform sensitivity over a large
sample area. The measurement consists of inserting the membrane filter
into the region between the source and detector and observing the change
in total counting rate. When the system has been properly calibrated,
the observed counting rate of pulses above a fixed threshold level can
be related to the filter mass using the relationship
I = I0 e-yx CD
where Io and y are the previously determined source intensity and mass
absorption coefficent, respectively. I is the observed counting rate
and x is the mass per unit area of the filter.
Although the measurement is straightforward in principle, the use of
beta-attenuation in the present study is complicated by the high precision
required. A typical measurement consists of determining the mass accumu-
lated on a 4 mg/cm2 filter to a precision of ± 10 yg/cm . This requires
that each mass measurement be accurate to ~ 0.1% The elapsed time be-
tween the measurements of the tare weight and the exposed weight might
be several months, during which time it is likely that the measuring
apparatus has been subject to deliberate or accidental changes. Further-
more, it is known that the exponential behavior of Equation 1 is not a
fundamental characteristic of beta-particle attenuation, but is the for-
tuitous result of certain properties of the spectral shape of the beta-
particles. Small departures from ideal behavior are expected which can
contribute to errors at the 0.1% level.
Our procedures employ frequent calibrations to eliminate many of the
problems associated with system instability. A series of carefully
weighed polycarbonate film standards which span the mass region of inter-
est are measured with the beta-gauge and the resulting count rates deter-
mined. To achieve adequate statistical accuracies, counts are accumu-
lated for 100 second intervals at a counting rate of ~ IO5 counts/second.
-------
300 mC C14 SOURCE
VACUUM CHAMBER
^%%%^%%%%%%%^%%%^^^^%%^^%^^^^^^e^j^%^
MOUNTING YOKE
VACUUM CHAMBER
SIGNAL TO
DISC/SCALER
SAMPLE
ALUMINUM
WINDOWS
o
3 mg/cm
DETECTOR
XBL 752-327
Figure 3. Cross section of B-gauge apparatus.
-------
The resulting data of mass (mg/cm2) versus counts/second are then fitted
to Equation 1 by the least squares method. The resulting calculated
values of I0 and y are then used to calculate the unknown masses of the
membrane filters. To ensure an accurate fit to the data, the mass
region spanned by the thin film standards is limited from 3 mg/cm2
to 6 mg/cm2. By performing a least square fit of the data over this
interval to the function of
T T yx + vx2 ,-,
I = Io e (2)
it is possible to show that the function approximates a pure exponential,
since it is observed that vx2 « yx. The deviations of gravimetric
masses from the least square calculations are typically 3 yg/cm2.
Additional corrections, which cannot be eliminated by frequent cali-
brations , are as follows:
RELATIVE HUMIDITY EFFECTS
Although the uniform, thin film standards used for calibration
purposes are immune to changes in the ambient relative humidity, the
cellulose ester filters used in the samplers are very susceptible to
such variations. To permit corrections for this phenomenon, a series
of 70 clean membrane filters were exposed to variable relative humidity
and the effects on total mass were determined. Figure 4 shows the re-
sults. The fitted slope obtained from these data is 1.80 ± 0.02 yg/cm2/%
change in relative humidity. The range of relative humidity encountered
in our laboratory is 45 to 65%. The relative humidity is noted when
filters are measured and the mass correction is applied to data at the
time of computer analysis. No correction is made for humidity effects
in the aerosol deposits, since over the limited range of relative hu-
midities experienced in the laboratory, such a correction should be
negligible compared to that for the substrate.
ATOMIC NUMBER DEPENDENCE
The rate of energy loss of electrons traversing a material of
atomic number Z and mass number A is a complex combination of ionization,
nuclear and electronic scattering and radiation losses. The actual rate
of ionization loss decreases slightly with Z/A as Z is increased.
However, the increase in scattering with increasing Z results in increased
resultant path length. At low beta energies, the effect of scattering
overcompensates the effect of having fewer electrons per gram in high Z
material. This results in a dependence of the absorption on the Z/A of
the samples. This has been experimentally measured for a lf*7Pm beta
source by using a series of thin standards of various elements. A least
square fit to the experimental values gives
y = [7.04 - 10.77 |Jx lO'* cm2/yg (3)
-------
45 50 55 60
RELATIVE HUMIDITY
XBL 776-9209
Figure 4. Mass gain of 4 mg/cm2 cellulose membrane as a function of %
relative humidity.
-------
Table 1. Atomic Number Dependence of Beta-gauge Mass Attenuation Coefficient
MATERIAL
Polycarbonate
Carbon
Calcium
(NHtt)2 SO^
Pb
]
Z/A
0.527
0.500
0.499
0.530
0.396
P (10 4cm2/pg)
1.37
1.66
1.67
1.34
2.78
This empirical formula can then be used to estimate the effect of aerosol
composition on mass measurements by beta gauging. Table 1 shows the calcu-
lated absorption coefficients for some selected substances. The use of
polycarbonate films as calibration standards causes the mass of heavier
compounds, such as (NH^) SO^ to be underestimated by about 3%; oxygen
and most other elements from C to Ca will be overestimated by about 20%.
Thus, the light hydrocarbons and the sulfates will have compensating
effects with the more abundant elements below Ca. Since the heavier
elements are usually present in trace quantities, they will contribute
negligible errors. Even 10% by weight of Pb will introduce an error of
about 5% in the accuracy of mass measurement.
FILTER POROSITY EFFECTS
As noted earlier, the calibration of the beta-gauge is achieved
through the use of carefully weighed polycarbonate film standards.
Since the cellulose ester membrane filters consist of a microscopically
nonhomogeneous and porous medium, there arises a discrepancy when their
mass is determined using the calibration obtained from the continuous
thin film standards. Since this discrepancy is constant for a given
filter mass, it has no effect on the calculated mass difference before
and after exposure. However, there are situations in which this discrepancy
10
-------
can become significant. We have observed that the magnitude of this
effect depends on the amount of material which the beta-particles have
traversed before reaching the sample itself. For example, if the de-
tector vacuum chamber window is changed from 2.40 mg/cm2 to 4.00 mg/cm2,
an apparent change in measured mass of a 4 mg/cm2 membrane filter of as
much as 38 yg/cm2 is observed when both systems were calibrated in the
same manner.
A possible explanation for this phenomenon is that both the average
energy and the angular distribution of the beta-spectrum are changed by
varying the amount of material traversed by electrons before reaching
the sample. The change in angular distribution is brought about by
multiple scattering and would be expected to affect the results more
for a porous filter medium than for a uniform filter. Certain changes
in the source-detector geometry might also be expected to produce sim-
ilar effects upon the results.
In practice, the beta-gauge apparatus is maintained in as constant
a configuration as possible. Unavoidable changes, such as punctured
vacuum windows do occur, however. A standard set of membrane filters
is used to compare the porosity effect when any beta-gauge modifications
occur. Once the effect has been calibrated, subsequent thin film stan-
dards are adequate until the next change in the system geometry occurs.
In spite of the necessity for such corrections, the beta-gauge tech-
nique still has the advantage of automatic operation for mass measurement
in large-scale sampling programs. A total of 40,000 mass determinations
have thus far been performed and a total of 70,000 will be completed
by August 1977.
Using a computer controlled automatic sample change, the filters
are individually counted for 30 seconds. The precision of this measure-
ment is 4.3 yg/cm2. After the difference between two such measurements
is calculated and the various correction factors applied, the precision
is estimated to be 10 yg/cm2- This converts to an accuracy of aerosol
mass determination of 12 yg/m3 for a two hour sample and 2 yg/m3 for a.
12 hour sample.
11
-------
SECTION 4
X-RAY FLUORESCENCE MEASUREMENTS
The elemental composition of the particulate deposits were measured,
using an energy dispersive X-ray spectrometer. Descriptions of this
method have been reported extensively in the literature4*5) and will be
treated only briefly here. Among the advantages of energy dispersive
XRF are its multiple element capability, high sensitivity for elements
of interest, ease of automation and stability of calibration.
SENSITIVITY
The particular X-ray spectrometer employed is an LBL-constructed
pulsed X-ray excitation system with computer controlled sample sequenc-
ing and analysis. The X-ray excitation is provided by a series of three
secondary fluorescence targets, which are irradiated with the output of
a pulsed X-ray tube. The advantages of the pulsed excitation are in-
creased sensitivity for analysis and elimination of certain systematic
artifacts which result from pulse pile-up in the conventional X-ray
fluorescence spectrometer systems6). Using the three secondary targets,
we routinely analyze 28 elements, although a larger number of elements
could be monitored, if desired. A summary of the operating conditions
are given in Table 2. Table 3 is a list of the elements measured and
the associated sensitivity and minimum detectable limits for the analysis
times normally employed. For a comparison of these detectabilities with
other XRF methods, see Reference 4. Figure 5 is a plot taken from that
paper, showing a comparison of the energy dispersive XRF method with
typical elemental compositions for the atmospheric aerosol.* These data
indicate that the method is sufficiently sensitive for the analysis of
most elements of interest in air particulate analysis.
PRECISION AND ACCURACY
The precision and accuracy of the XRF method are dependent upon many
components in the system, each of which must be carefully controlled.
This is particularly true in a large-scale, automated study where the
system operates unattended for extended periods. For convenience, a
discussion of errors can be separated according to the following areas:
1) excitation source instability, 2) reproducibility of sample geometry,
3) spectrometer stability, 4) errors in spectral analysis, and 5) syste-
matic errors in calibration and data analysis. A discussion of each of
these follows.
*To convert from ngm/cm2 to ngm/m3, one can assume that the dichotomous
samplers used in the study sample at the rate of 1 m3/cm2 in a two-hour
period.
12
-------
Table 2. OPERATING CONDITIONS FOR SAMPLING AND ANALYSIS
Excitation:
- Pulsed x-ray tube, W anode with secondary targets
- 85 watts average power
Detector:
- Lithium drifted silicon guard ring detector operated
in anti-coincidence mode
- 30 mm2 area, 195 eV resolution at 6.94 keV
- Maximum counting rate at 14,000 cp
- Pulsed optical feedback amplifier
Secondary targets, operating tube voltages, and analysis periods:
Ti, 50 kV, 1.57 rain.
Mo, 60 kV, 1.39 min.
Sm, 75 kV, 2.56 min.
Sampler characteristics:
Flow — 50 2,/min.
Area -- Approx. 7 cm2
Outpoint -- 2.4 um
lixcitation Source Instability
X-ray spectrometers are normally calibrated in terms of the yield of
fluorescent X-rays obtained with a constant current flowing in the X-ray
tube for some specified period of time. If the yield of excitation X-rays
per electron at the anode is constant, and if the current and time are
carefully measured, then the precision of the excitation is maintained.
In the pulsed X-ray system, the current and time measurements are
replaced with an anode current integrator. Each analysis is then normal-
ized to the calibration data according to the total charge which flowed
13
-------
TABLE 3. SENSITIVITY FOR ENERGY DISPERSIVE X-RAY FLUORESCENCE ANALYSIS
ELEMENT
Ala>
Si
P
S
Cl
K
Ca
Tib>
V
Cr
Mn
Fe
Ni
Cu
Zn
Ga
As
Se
Br
Rb
Sr
Hg
Pb
CdC>
Sn
Sb
Ba
ATOMIC
NUMBER
13
14
15
16
17
19
20
22
23
24
25
26
28
29
30
31
33
34
35
37
38
80
82
48
50
51
56
BACKGROUND^
(counts/sec)
19.0
19.2
21.6
52.3
95.8
53.2
86.4
6.3
5.4
5.2
5.0
6.3
3.7
5.3
4.8
3.1
3.2
2.9
3.7
5.6
11.3
3.3
7.6
2.8
5.0
5.3
52.4
SENSITIVITY
(counts/sec
per ygm/cm2)
7.40
25.3
48.0
83.8
125
272
411
28.8
37.8
49.3
59.8
76.4
112
128
148
166
209
234
258
304
320
109
109
75.5
75.7
74.6
62.1
MINIMUM
DETECTABLE LIMIT
(ngm/cm2)
200
58.9
32.9
29.4
26.6
9.14
7.7
31.3
22.2
16.6
13.5
11.8
6.2
6.5
5.3
3.8
3.1
2.6
2.7
2.8
3.8
6.0
9.1
5.9
7.8
8.1
31
a) These elements
b) These elements
c) These elements
d) The background
were analyzed for 93.6 seconds, using a Ti secondary target.
were analyzed for 83.4 seconds, using a Mo secondary target.
were analyzed for 153 seconds, using a Sm secondary target.
was obtained using a blank membrane filter.
14
-------
105
n
103
u
2
O
U
10-1
10-2
c
r fl Si
y n c n
H NNaMAl !CI T-
rO n hnflMfl"- n«
"H U ' o n n«
a - • pu • r
- F U n U
r . I \ \I
D \ >JJ
LIB
r n
H
U
™
_
~
en \*0'''
1
U
L
Fa
Cun
/In
1 Br
P
U Ni
As Rbj!.r 2
f] .Se
•.. GaGe nUtl 7
"^S^-ft , , , .— ^ n
U ••"• B
u
J
n
j-j TYPICAL VALUE
r fl
•*•
Wh
I*U
n
H
0
0 Cd Sr
Ag
D..-"
In
r
Pd
F
RuRh
TT
Sb Ba
/ r
' ncs '
Te
n
j
• L
] Nd
nflsm-.
H n
n r
J
T ' u
T
ud
Quy
fl Yh
• n n "
1 y 11
11
^v
-i
Pb
-
-;
•
—
-i
-
"--^
Hg
w
Hf
rJa
flr
-
Pt
F"
n
Ti
1
I
1 I
rhf] '
~ n — •
J j
• • '
.
1
105
1031
c
102 t
5
3
^\
u
UJ
ul
10" °
10-1
10-2
ATOMIC NUMBER
XBL 776-9213
Figure 5. Minimum detectable limit for x-ray fluorescence analysis compared with average aerosol
concentration ranges.
-------
in the X-ray tube during the measurement. This technique has distinct
advantages over the conventional method, since corrections for pile-up
and system dead time effects need not be considered.
The reproducibility of the charge integrator was checked by observing
the linearity of the X-ray yield with the integrated charge. A more
stringent test involved the measurement of the stability of the calibra-
tion under varying count rate conditions. A thin-film standard of Fe
was first counted for a fixed amount of collected anode charge and the
Ka intensity recorded. A thick scatterer was then placed behind the Fe
standard and the measurement repeated. The effect of the scatterer was
to reduce the average direct current in the tube by a factor of 20.
However, the system compensated by increasing the counting time, result-
ing in the same total integrated charge. The extent to which the Fe Ka
intensity remained unchanged is a measure of the accuracy of the current
integrator method. A 3% agreement was observed.
Variations in the X-ray yield per unit charge at the anode are
caused principally by the energy dependence of the X-ray production
cross sections in the X-ray tube and secondary fluorescence targets.
This dependence can be measured by observing the yield of fluorescent
X-rays from a standard sample as the tube anode voltage is changed.
Figure 6 is a series of plots, showing the measured voltage dependence
for each of the three secondary fluorescers. Combining these results
with the fact that the X-ray tube voltage is regulated to be stable to
±5%, we predict X-ray output variations of < 1% for the worst case.
Other possible sensitivity variations due to changes in the tube-
fluorescer geometry have been checked and found negligible under normal
operating conditions.
Reproducibi1ity of Sample Geometry
Variations in the system sensitivity could be caused by the fact
that the excitation and detection efficiencies are functions of the posi-
tions of the sample in the detector collimator field of view. Figure 7
is a view of the X-ray tube-secondary target detector geometry. The
sample is introduced horizontally into this region by means of an auto-
matic sample changer. A scan of the sensitive area for the case of the
Mo secondary fluorescer is shown in Fig. 8. These curves were obtained
by moving a point Cu specimen horizontally across the region where the
membrane filters are normally placed. Curve a) is a scan of the axis,
perpendicular to the view of Fig. 7. Curve b) is a scan from right to
left in the same view. The detector collimator opening was deliberately
chosen to restrict this sensitive region to an area much less than the
typical 30 mm diameter deposit on the filter. For this reason, the
fluorescence intensity is not sensitive to small displacements of the
sample in the horizontal plane, providing that the sample is uniform
over the exposed area.
16
-------
10
9
8
7
6
! 5
OC
H
CD
Q£
<^
Q
_l
LU
10
F3 oc v29 AT 75kV
F2 oc V17 AT 60kV
Fl oc V-19 AT 50kV
I
I
F2
I I I I I
F3
_L
20 30 40 50 60
X-RAY TUBE VOLTAGE (kV)
80 100
XBL 776-9194
Figure 6.
Voltage dependence of x-ray yield in x-ray fluorescence unit.
Vertical low on curve represents normal operating voltage.
17
-------
00
SAMPLE HOLDER
WITH FILTER
X-RAY TUBE
Ta Shielding
Al Shielding/
Collimator
SECONDARY TARGET
SCALE: Inches
Figure 7. Cross section of secondary fluorescence geometry.
XBL 731-102
-------
CO
ce
LU
O
o
V)
in
(K
O
\ I I I
a)
I
I
1 I I
b)
I
20 10 0 10 20 20 10 0 10
DISTANCE FROM CENTER OF SAMPLE (mm)
20
XBL 776-9193
Figure 8. Scans of sensitive region of x-ray fluorescence unit.
19
-------
Vertical displacement of the sample in the sensitive region can po-
tentially have a larger effect on the observed intensity. The membrane
filters are not always perfectly flat, resulting in a variation in verti-
cal distance from point to point across the filter. We have measured
the change in fluorescence output for a uniform thin sample as a function
of the vertical displacement of the sample and a change of 16% was ob-
served for a 1 mm displacement. Assuming a maximum departure from flat-
ness of 0.5 mm, then the output variation would be less than 8%. In
practice, the observed error is much less than this.
Spectrometer Stability
The characteristics of the Si(Li) semiconductor spectrometer system
which can potentially affect the results are changes in the absolute
efficiency, peak-to-background ratios, system resolution, and peak loca-
tions. In a typical spectrometer system, the absolute efficiency and
peak-to-background ratios are easily stable within the limits of con-
cern for the present analytical applications. Although some variations
in these parameters have been observed in the past, they were attributed
to artifacts in the system design which have been eliminated.
The energy resolution of the system is also quite stable if problems
due to local electrical noise are eliminated. To reduce these problems,
our system is operated from a regulated ac source and care is taken to
eliminate local sources of noise, such as drill motors, etc. Frequent
checks of the system resolution are made; to date, no significant changes
have been noted.
Stability of the peak location in the multichannel spectrum is
affected by long-term drifts in the amplifier baseline and gain. These
are normally associated with variations in the ambient temperature. In
the present case, the temperature is maintained reasonably constant by
using a room air conditioner. Nevertheless, it is difficult to maintain
stability below the limits of observation. For a peak with 200 eV reso-
lution at 20 keV, a stability of ± 2 eV represents changes in amplifier
gain of one (1) part in 104. For this reason, the gain and baseline are
checked weekly and fine adjustments made where necessary to maintain them
at their reference value. The corrections are made by calculating the
centroid of the Ar Ka (2.95 keV) and In Ka (24.2 keV) peaks and adjusting
controls to position these peaks at their reference position. Root mean
square deviations of these centroids, which have been observed over sev-
eral one week periods, are ± 5.3 eV for the 2.95 keV line and ± 4.8 eV
for the 24.2 keV line.
The effect of these variations on the accuracy of the complete
analysis is difficult to assess. Changes in peak locations are partially
compensated for in the spectral analysis program. However, such shifts
introduce errors into the analysis of small peaks which are close to
very intense lines, since residuals are left after stripping the large
peaks from the spectra. Since there is no systematic way to predict
such circumstances, the best check of such errors is in the repeated
20
-------
analysis of standards which replicate typical air particulate samples.
Errors in_ Spectral Analysis
Energy-dispersive X-ray fluorescence analysis requires data reduc-
tion procedures to convert the multielement spectral data into peak in-
tensities of the individual elements. The problems of the subtraction
of background and possible overlap of peaks from different elements must
be handled by the computer algorithms.
There are many methods of analyzing spectral data7'8). All are
capable of extracting peak intensities for clearly resolved major
constituents with an accuracy limited only by statistical errors in the
peak and background integrals. However, the intensities of smaller
peaks which overlap larger peaks are much more difficult to extract,
since they can be affected by small shifts in peak position and by peak
and background shapes. Furthermore, specification of the error margins
in these cases is very difficult. The present discussion focuses on
the specific method of spectral analysis used by us, and avoids discus-
sion of sources of error beyond the normal statistical considerations.
The best practical estimate of total errors is obtained by determining
the reproducibility of a large number of measurements, made on a few
samples, whose compositions are typical of those encountered in normal
environmental samples.
The on-line spectral analysis program used in the present study is
a straightforward stripping procedure illustrated in Fig. 9. A back-
ground spectrum produced by a blank filter and standard spectra for all
individual elements are stored in the computer memory. The unknown spec-
trum is then reduced by subtracting the stored background, which is nor-
malized appropriately by comparison over a selected region and then by
sequentially stripping out the contributions to the spectrum due each
element using the stored spectral line shapes for the elements. The a-
mount of each standard elemental spectrum subtracted from the unknown
spectrum is chosen to best fit the intensity of the unknown X-ray. Peak
areas are then converted to concentrations in ng/cm2 by applying the
appropriate excitation and detector efficiency factors. These calculations
are described in greater detail below.
This method of spectral analysis works particularly well for air
particulate samples, because a membrane filter consists of a thin homo-
geneous substrate on which the elements to be measured are deposited.
These properties of the substrate are almost constant from one sample to
the next and are identical, essentially, to those of the blank filter
whose spectrum is stored as a background standard. Since the samples
are thin, absorption effects will not alter the spectral response of the
system to a given set of characteristic X-rays.
This method neglects the effect of an X-ray from one element over-
lapping those of another. This effect can be handled after the spectral
analysis, however, by using a simple formula which assumes that the
21
-------
STEP #1
STEP #2
STEP #3 • • •
ORIGINAL
SPECTRUM
i
to
Is)
BLANK FILTER
ELEMENT
#1
ELEMENT
#2
A/
»
\
i
i
t
t
t
t
t
t
t
f
1
'
*
t
t
t
IA
! STORED IN
MEMORY
XBL 731-86
Figure 9. Diagram of spectrum stripping procedure.
-------
intensity of a given element can be expressed as
!.=!.- Z. C-. I. (4)
when Cij is a measure of the interference of element j with the principal
line of element i. In sophisticated analysis programs, these expressions
can be iterated to achieve self-consistency. However, since the number
of overlapping lines in a typical environmental sample spectrum is small,
a simple one-step calculation involving experimental measured Cij is nor-
mally adequate.
The Cij is determined by taking thin standards containing a known
amount of element j and measuring the contribution of this standard to
the intensity of the line due to element i. At present, these factors
have been assumed from previous work by T. Dzubay at Research Triangle
Park. Final data will be processed using Cij measured in our laboratory,
using thin film standards selected so as to ensure negligible absorption
of the very low energy X-rays.
The most sensitive indicator of the effectiveness of this spectral
analysis method is observation of the residual intensities left in the
pulse-height spectrum after the stripping procedure has been completed.
Ideally, the remaining counts in the multichannel spectrum should be
evenly distributed about zero, with deviations reflecting only statisti-
cal uncertainties of the original spectrum. Any structure or residual
intensity in the form of peaks above statistics must be regarded as an
artifact resulting from incomplete stripping of the X-ray lines or from
the effect of shifts in the unknown spectrum relative to the standards.
Over several years of operating experience in the present study, it has
been observed that if the system is properly calibrated and maintained
in a stable configuration, the residual spectrum has always been free
from such structure. However, periodic checks are made to ensure that
the spectral analysis is operating properly.
CALIBRATION ACCURACY
The errors we have discussed so far are primarily those affecting
the precision of measurement, i.e., the reproducibility over extended
periods of time and under varying analytical conditions. The absolute
accuracy of the X-ray fluorescence analysis is determined almost entirely
by the calibration procedures.
In its simplest form, the determination of elemental concentrations
using a X-ray fluorescence spectrometer consists of relating the peak
intensities observed in a spectrum to the concentration of the various
elements present on the filter. We have
C. = N F.I./A. (5)
where Ci is the concentration in ng/cm2, Ii is the observed counting rate
for element i, Fi is the calibration factor appropriate for a thin spec-
imen of element i (related to the sensitivity of Table 3 as Fi = I/Si),
23
-------
Ai is a factor which corrects the results for attenuation of the fluores-
cent X-ray as it leaves the sample and N is a normalization factor which
scales the result according to the analysis time or integrated current.
The normal calibration procedure uses thin, uniform standards which
closely replicate the geometry of the air particulate filters. These
standards have been calibrated either directly by gravimetric measurement
or by reference to another standard. The term "thin" implies that Ai = 1
for these standards.
For the lower energy fluorescence X-rays produced in a filter deposit,
absorption effects can occur either because of the location of the individ-
ual particles within the filter matrix or because of attenuation of the
X-rays arising within an individual particle. Corrections for matrix and
particle size effects are contained in the factor Ai, and are applied
during the later stages of data analysis. The overall accuracy of the
measurement then includes uncertainties in the thin standard calibration
and those associated with the absorption corrections (see Table 10).
Thin Film Calibration
The use of thin film calibration has been discussed extensively in
the literature9*11). The procedure consists of measuring X-ray yields
for elemental standards distributed across the range of atomic numbers
of interest and then interpolating for elements not directly measured.
The validity of the interpolation procedure results from the smooth be-
havior of the X-ray cross sections as a function of the atomic number of
elements.
In the present calibration, a series of thin film Cu standards were
used as the primary calibration standard. These standards consists of
an evaporated layer of Cu on mylar substrate and were obtained from Micro-
matter, Inc. These standards have been extensively cross-checked with
other gravimeteric standards, using several independent X-ray fluores-
cence measurements. The standard is accurate to better than 2%.
The relative excitation efficiency for the elements were obtained
using thin film standards prepared by depositing an aerosol generated
from a carefully prepared solution in which the relative concentrations
of the elements was previously known. Where possible, the ratio of ele-
ments was established by the stoichiometry of the chemical compound. In
other cases, solutions containing individual elements are mixed in known
ratios according to a method described by Giauque11). The principal
selection criteria were compounds which did not react in solution and
whose characteristic X-rays produced no overlapping lines in the spectra.
Table 4 is a summary of the elements and compounds which were used. The
accuracy of the ratios obtained by this method are estimated to be less
than 1%.
Because of the importance of sulfur in this program, exceptional
care was taken in its calibration. Thin film standards were prepared
24
-------
TABLE 4. CALIBRATION SOULUTIONS FOR THIN AEROSOL DEPOSITED STANDARDS
Element
Compound
Solvent
Elements with
which combined
K
K,Cr20.,
HN03
Cu
Ca
CaCO,
HN03
Cu
Cr
K2Cr207
HN03
Cu
Hn
Mn
HN03
Cu
Zn
Pb
Sr
Fe
Fe
HC1/
HN03
Cu
Ni
Ni
HN03
Mn
Cu
Cu
HN03
K
Cr
Fe
Ca
Mn
Ba
Zn
Zn
HNO,
Mn
Sr
SrCOj
HN03
Mn
In
Rb
RbCl
H20
In
Ag
AgN03
H20
In
In
In
H.NO,
Sr
Ag
Ba
Pb
Rb
Sr
Ba
BaCO,
HNOj
In
Pb
Pb
HNO,
In
Mn
25
-------
by collecting an aerosol deposition for particles of 0.3 ym diameter in
order to eliminate absorption effects. The compounds used for the deposi-
tion were CuSO l^SO^, and K2Cr20. The calibrations were then referred
te the Cu standard through the following three paths: Cu->S, Cu-HC+S, and
Cir*Cr+K-»-S . The agreement between these three independent determinations
of the sulfur calibration was within 3% .
The calibrations for the very light element Al and Si are complicated
by the strong attenuation of the low energy X-rays. For these cases, thin
C- 100 ug/cm) evaporated films are used for the direct calibrations. The
inherent inaccuracies due to the attenuation effects experienced by these
elements are reflected in the increased analytical error in the deter-
minations.
Table 5 contains a list of the calibration factors and their associ-
ated errors as determined by the procedures described below. The adopted
values are obtained by demanding a smooth curve fit through the measured
data points. These calibration factors are the reciprocals of the sen-
sitivities listed in Table 3 and are normalized to unit charge collected
in the X-ray tube instead of unit time of analysis.
Absorption Effects
Absorption of fluorescence X-rays within the sample can occur in
either of the two ways illustrated in Fig. 10. The diagram of lOa)
illustrates the matrix attenuation which an X-ray produced at the depth
x experiences. The incoming beam I0 is attenuated by e~^0x where y0
is the coefficient for energy E. .
The complete matrix absorption correction for particles collected
on a filter of thickness d is calculated by integrating the absorption
expressions over the thickness of the filter and weighted according to
a. particle density distribution p(x). If we assume a surface deposition,
then the absorption correction A. = 1. If we assume a uniform deposition
of p(x) = constant, then it can o"e shown that
A. . - (6,
Typical estimates of the factor A. assuming a uniform deposition of par-
ticulates within the filter are 0*87 for Ca, 0.67 for S, and 0.30 for Al
ticulates
Ka X-rays.
Since observed particle deposition profiles indicate that a surface
deposition model is a close approximation to real samples, the values
obtained for Fig. lOa should be used as upper limits on possible matrix
absorption corrections. Present data analysis procedures assume Ai = 1.
26
-------
TABLE 5. SUMMARY OF CALIBRATION MEASUREMENTS
Element
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Ni
Cu
Zn
Ga
As
Se
Br
Rb
Sr
Hg
Pb
Cd
Sn
Sb
Ba
a")
Measured Calibration Factor
2.4 ± 0.5*
7.84 ± 0.14
38.8 ± 0.6
4.07 ± 0.07
4.94 ± 0.08
6.31 ± 0.1
9.23 ± 0.15
10.55 ± 0.19*
12.19 ± 0.16
25.1 ± 0.20
26.4
8.98 ± 0.16
5.64 ± 0.14
5.91 ± 0.6*
4.62 ± 0.11
Adopted Value
0.70
2.39
4.54
7.92
11.85
25.64
38.8
2.38
3.12
4.07
4.94
6.31
9.23
10.55
12.19
13.75
17.3
19.3
21.3
25.1
26.4
9.00
9.00
5.62
5.63
5.55
4.62
aj Errors are based on root mean square deviation of several independent
standards, except those marked (*) which were evaporated standards.
27
-------
A)
Ei I,
dx
B)
figure 10. Illustration showing matrix absorption effects (A) and large
particle attenuation (B).
28
-------
Loo, et al10) have studied the effect of surface deposition and its
effect on the analysis of sulfur in great detail. They have found that
surface deposition normally occurs except in circumstances in which a
high relative humidity of ambient air occurs in combination with high
particulate concentrations. Under these conditions, the sulfur contain-
ing particles tend to migrate into the filter and the deposit can approach
a uniform depth deposition. A method of measuring A. is proposed which
works by measuring the X-ray fluorescence spectrum on both sides of the
filter. Future sulfur analysis will reflect this correction.
The particle size effects are illustrated in Fig. lOb). Here the
intensity of I. and I! are different due to the different path lengths
in the particles. Again, the magnitude for this effect is dependent upon
the energy of the fluorescence X-ray and is worse for the light elements.
The estimation of the attenuation A. due to this effect requires
the assumption of specific model of particle composition and morphology.
Fortunately, the use of the dichotomous sampler isolates those particles
less than 2 ym diameter where such effects are minimal. The large par-
ticle attenuation, however, can't be ignored.
We use the value calculated by Dzubay and Nelson12) which assume
uniform spheres of composition approximately that of typical aerosol
particulates. Table 6 is a tabulation of those corrections for two
size ranges. It should be noted that the fine particles S determinations
are not significantly affected by this correction.
REPRODUCIBILITY OF RESULTS
The reproducibility of the aerosol samples has been checked by col-
lecting side by side samples at LBL, using three automatic dichotomous
samplers. Two of the units had been in continuous operation at St.
Louis for two years prior to the test, the third had remained at LBL.
None of the flow controllers had been adjusted since their original
checkout following initial fabrication.
Table 7 is a summary results obtained from three separate sampling
intervals as measured by the XRF analysis of the deposited particulates.
The root mean square (RMS) deviation for the coarse particle fraction is
approximately 5%, whereas the fine particles reproduce to less than 1.5%.
This result is consistent with the lower loss experienced by the small
particles in their passage through the virtual impactor.
The precision of the XRF analysis was checked by the repeated anal-
ysis of the same filters over an extended period of time. In addition to
checking the stability of the total spectrometer system, this test will
also give some indication of the accuracy and stability of the spectral
analysis program. Table 8 is a list of average concentrations and RMS
29
-------
TABLE 6. PARTICLE SIZE ATTENUATION CORRECTIONS FOR THE LIGTH ELEMENTS.
(ADAPTED FROM REFERENCE 11.)
Element
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Ni
Cu
Zn
Fine Particle Correction
Fl
0.91 ± 0.09
0.93 ± 0.07
0.95 ± 0.05
0.97 ± 0.03
0.98 ± 0.02
0.99 ± 0.01
0.99 ± 0.01
F2
1.0 ± 0.0
1.0 ± 0.0
1.0 ± 0.0
1.0 ± 0.0
1.0 ± 0.0
1.0 ± 0.0
1.0 ± 0.0
Coarse Particle Correction
Fl
0.41 ± 0.12
0.48 ± 0.15
0.58 ± 0.24
0.64 ± 0.22
0.70 ± 0.20
0.78 ± 0.15
0.81 ± 0.13
F2
0.83 ± 0.13
0.86 ± 0.10
0.87 ± 0.10
0.90 ± 0.08
0.92 ± 0.07
0.93 ± 0.06
0.94 ± 0.05
0.96 ± 0.03
0.94 ± 0.06
0.95 ± 0.05
30
-------
TABLE 7. REPRODUCIBILITY OF AIR SAMPLERS MEASURED BY SIDE-BY-SIDE
SAMPLING
Element
S
Pb
Fe
Particle Size
Fine
Coarse
Fine
Coarse
Fine
Coarse
Mean Concentration
(jig/m3)
1218
250
1154
369
208
1622
Average % Deviations
0.5%a)
2.5%
1.2%*)
2.9%
2% M
4.6%b)
a) Since S and Pb are predominantly in the small particle fraction,
these errors should be considered representative of the precision
for collection of fine particles.
b) This error should be the precision for the collection of coarse
particles.
31
-------
TABLE 8. ROOT MEAN SQUARE DEVIATIONS OF ANALYSES OF IDENTICAL SAMPLES OVER
A THREE-MONTH PERIOD
Element
Al
Si
P
S
Cl
K
Ca
Ti
Mn
Fe
Cu
Zn
Br
Pb
Sr
Cd
Sn
Sb
Ba
Average Concentration
(ng/cm2)
740
2397
217
11693
298
515
3959
123
48.1
1930
123
519
126
653
12.9
24.8
34.8
6.8
102
Deviation
(ng/cm2)
33
40
15
189
9.2
7.7
55
18
5.5
18
5.9
29
2.0
5.5
1.1
1.6
2.1
2.0
5
% Deviation
4.5
1.7
6.9
1.6
3.1
1.5
1.4
14.6
1.1
0.9
4.8
0.6
1.6
0.9
8.5
6.4
6.0
29
4.9
32
-------
deviations obtained from ten successive measurements of the same filter
carried out periodically over a three month period. The reproducibility
of the major elements indicates a stability of approximately ± 1% over
this interval as indicated by the error in the major elements. The small,
relative error for minor constituents, such as Mn and Ti, give some indi-
cations of the reproducibility of the computer spectral analysis. The
results are particularly impressive when we emphasize that these measure-
ments were performed periodically during an interval when the system was
continually analyzing 500-1000 samples per week automatically, and no
special attention was devoted to these particular analyses.
As noted in preceding sections, the accuracy of the measurement de-
pends primarily on the accuracy of the thin film calibrations for mass
and elemental concentrations and secondarily on the accuracy to which the
various absorption and interelement correction times for XRF analysis are
shown. The accuracy of the thin film standard is specified by reference
to gravimetric methods. These standards have been further validated by
intercomparison studies and found to be accurate to better than 2%13).
The accuracy of the complete analysis is verified by intercomparison
of the analyses with other laboratories and methods. This is currently
being performed as validated results from other RAPS measurements become
available. In the interim, a detailed comparison of results obtained by
Robert Giauque of LBL uses independent XRF measurements of the same
samples. As can be seen in Table 9, this agreement is to within 5% for
major elements.
It should be also be pointed out that extensive intercomparison
studies involving our laboratory and others have been carried out in
order to validate the XRF method. The studies have been published and in
general, the agreement is excellent14).
33
-------
TABLE 9. COMPARISON OF OUR ANALYSES WITH INDEPENDENT XRF MEASUREMENTS BY
R. GIAUQUE
Element
S
K
Ca
Ti
V
Cr
Mn
Fe
Ni
Cu
In
Ga
As
Se
Br
Rb
Sr
Hg
Pb
Sample 1
R. Giauque
0
723 i 132
524 ± 53
91 ± 21
31 ± 17
33 ± 12
87 ± 11
1225 ± 17
15 ± 3
46 i 4
153 ± 5
19 ± 5
63 ± 16
53 ± 3
446 ± 6
0 ± 3
15 ± 3
0 i 6
3050 ± 23
Ours
0
684 i 3
839 + 18
166 ± 9
41 ± 7
41 ± 1
122 ± 8
1327 ±11
21 ± 1
68 ± 2
173 ± 3
0
21 ± 10
55 ± 1
520 ± 5
0
14 ± 1
5 ± 2
3221 ± 22
Sample 2
R. Giauque
20400 ± 400
343 ±133
320 ± 58
124 + 24
50 ± 19
24 ± 15
27 ± 11
689 + 14
21+4
50 ± 5
306 ± 6
7 ± 4
14 * 9
17+2
194 + 4
2 ± 3
2 ± 3
6 ± 5
991 ± 15
Ours
22289 ± 440
577 ± 6
451 ± 58
0
19 ± 5
0
37 ± 3
716 ± 9
16 ± 1
50 ± 1
294 ± 6
0
0
15 ± 3
212 ± 5
0
3 ± 3
6 ± 3
962 ± 4
Sample 3
R. Giauque
30600 ± 610
739 ± 136
698 ± 62
78 ± 25
29 ± 19
27 ± 15
49 ± 12
118S + 17
15 ± 4
79 ± 5
499 ± 8
7 ± 4
43 ± 12
24 + 3
301 + 5
1 ± 3
11 ± 4
17 ± 6
1759 ± 18
OUTS
33700 ± 670
770 ± 7
820 ± 56
79 ± 9
20 ± 7
8 ± 6
76 ± 6
1190 ± 5
18 ± 3
63 ± 7
482 ± 8
0
0
24 ± 2
318 ± 6
0
5 ± 1
19 ± 3
1751 ± 32
Sample 4
R. Giauque
14100 ± 280
1264 ± 141
1853 ± 73
95 i 25
70 ± 20
27 ± 15
173 ± 14
3980 ± 29
33 ± 5
87 + 6
3634 i 19
0 ± 4
28 ± 4
8 ± 2
174 ± 4
7 ± 3
10 ± 3
0 ± 5
1012 ± 15
OUTS
15169 ± 303
1517 ± 10
2136 ± 20
43 ± 21
70 ± 8
6 ± 4
195 ± 12
3975 ± 20
47 ± 2
121 ± 8
3534 ± 14
0
9 ± 6
8 + 2
185 + 3
8 ± 1
7 ± 2
5 ± 5
992 ± 9
Values listed are in ng/cm2. Errors are 1 o for counting statistics only.
Ti, V, Cr, Mn valifes uncorrected for Ba L X-rays.
34
-------
en
10 AIR SAMPLERS
2 SIZE FRACTIONS
2-12 HOUR SAMPLES
X-RAY ANALYSIS
ELEMENTS FROM
AI-Ba, ALSO Pb, Hg
DETECTION LIMIT
FILTER MOUNTING,
DIGITAL MARKING
50 l/min=3m3/hr
DATA FROM
OTHER
SOURCES
ELEMENTAL
CONCENTRATION
SAMPLING
INFORMATION
TARE
WEIGHT
FINAL
WEIGHT
CENTRAL
DATA
PROCESSING
EPA
DATA BANK
MAGNETIC TAPE
MICROFICHE
CONDENSED LISTING
DATA ANALYSIS
PLOTS
CORRELATIONS
FACTOR ANALYSIS
XBL 776-9076
Figure 11. Flow chart showing the overview of sampling, analysis and data merging.
-------
SECTION 5
DATA HANDLING
A considerable portion of the effort involved in the program is con-
cerned with data storage and analysis. The results of mass and elemental
composition for the 30,000 samples must be sorted, processed and stored
in some reliable and, hopefully, intelligible fashion. A complete flow
chart for the operation is shown in Fig. 11.
The data handling begins with the tare weights as measured by the
beta-gauge prior to the field sampling. These beta experimental tapes
contain calibration information and filter masses and are listed accor-
ding to sample number only. On returning from the field, a second
measurement of mass is made and a second data tape generated. This data
is also catalogued according to sample number, although the sequence
of sample numbers is not the same as in the tare weights.
The samples are then sent for XRF analysis. At this time, the
sampling information is entered. The sampling station, time of sample,
sample interval and other information concerning the details of the XRF
analysis are written on the magnetic tape output along with the results
of that analysis. The information is now available to order the samples
according to time and place.
At this stage, a program is run which generates a directory of the
available data and constructs a file relating sample numbers to specific
stations and time slots. This directory is then used to merge the appro-
priate beta-gauge data for the specific sample number or time slot and
calculate the deposited mass.
The next program used this same directory to merge the XRF data
with the mass data and generates an output file of results as a function
of time and location. This same program also performs calculations of
absorption corrections, interelement interferences and adjustments to the
final results. These calculations are summarized in Table 10. The out-
put from this program is available in an abbreviated, printed form; micro-
fiche photocopy and magnetic tape output. The magnetic tape output is
then transmitted to the EPA data.
36
-------
In order to present this mass of data in a more understandable form,
selected portions have been written onto random access storage on a mag-
netic disk. In this fashion, time plots of concentrations or ratios of
concentrations, correlation plots, monthly averages and many other data
presentation formats can be easily generated.
Future plans call for using a 40 million word disk to contain the
entire two year data set. Random access of the entire study would then
be available, including data from other sources in the RAPS study.
37
-------
TABLE 10. CALCULATIONS PERFORMED DURING DATA ANALYSIS
1) Original results from X-ray fluorescence analysis:
C ± AC
2) Correction for particle size. Constant A is obtained from Table 6:
C' = C/A
AC' = c' iff—f+ (^
3) Interelement interference correction. Constants B. are obtained
from thin film measurements:
r'' - r'+VR r' a)
L - L + /, B. u.
i
AC" = C'
4) Convert to concentration per unit volume. Constant F is the flow
of the particular sampler, and T is the sampling interval:
C = C"/(F xT)
AC = C
v v
5) Interparticle interference correction:
C*(coarse) = Cv(coarse) - 0.05 Cv(fine)
C* (fine) = Cy(fine)/0.95
a) The sum is assumed over all elements (C.) except for the one of interest
38
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse be for? completing/
1. REPORT NO.
EPA-600/4-78-034
4. TITLE AND SUBTITLE
AEROSOL ANALYSIS FOR THE REGIONAL AIR POLLUTION STUDY
Interim Report
6. PERFORMING ORGANIZATION CODE
3. RECIPIENT'S ACCESSION NO.
5
7. AUTHOR(S)
F.S. Goulding, J.M. Jaklevic and B.W. Loo
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Lawrence Berkeley Laboratory
University of California
Berkeley, CA 94720
10. PROGRAM ELEMENT NO.
1AA603 AA44 (FY-77)
11 CONTRACT/GRANT NO.
IAG-D6-0760
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory — RTF, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Interim 1/1/76 to 12/31/76
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
16. ABSTRACT
An aerosol sampling and analysis program was conducted as part of the Regional
Air Pollution Study in St. Louis. Ten automatic dichotomous samplers were operated
in the field for two years and collected 35,000 samples. The procedures used for
analyzing these samples for total mass and elemental composition are described in
detail. The characteristics of the betagauge mass measurement and energy dispersive
x-ray fluorescence analyses are discussed, together with the factors that affect the
precision and accuracy of the data.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
cos AT i 1-ield/Group
*Air Pollution
*Aerosols
Samplers
*Mass
*Elements
*Chemical Analysis
13B
07D
14B
20F
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport!
UNCLASSIFIED
21. NO. OF PAGES
47
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
39
------- |