&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

-------