EPA-R2-73-182
April 1973
Environmental Protection Technology Series
-------
-------
EPA-R2-73-182
X-RAY FLUORESCENCE SPECTROMETER
FOR AIRBORNE
PARTICULATE MONITORING
by
Fred S. Goulding and Joseph M. Jaklevic
Lawrence Berkeley Laboratory
Berkeley, California 94720
Interagency Agreement No. EPA-IAG-0089(D)/A
Program Element No. 1A1003
EPA Project Officers: Robert K. Stevens and Thomas G. Dzubay
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
April 1973
-------
This report has been reviewed by the Environmental Protection Agency and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, nor does
mention of trade names or commercial products constitute endorsement
or recommendation for use .
11
-------
TABLE OF CONTENTS
SUMMARY 1
1. INTRODUCTION 2
2. MONITORING SYSTEM DESIGN 3
A. Filter Specifications 3
B. Range of Elements Measured 3
C. Data Processing 5
D. Sampling Station 6
3. ANALYSIS STATION 7
A. Semiconductor Detector Spectrometer and Associated
Electronics 7
B. Mechanical Design 8
C. X-ray Tube - Fluorescer - Detector Geometry 9
D. Sample Changer 10
4. CONTROL ELECTRONICS 10
A. Dead Time Correction , 10
B. X-ray Tube Power Supply and Controller 12
C. Data Processing , 12
5. EXPERIMENTAL RESULTS 24
A. Calibration 24
B. Sensitivity 32
C. Particle Size and Matrix Effects 34
6. CONCLUSIONS 37
REFERENCES 37
TABLE CAPTIONS 38
FIGURE CAPTIONS 38
APPENDIX A.I
-iii-
-------
-------
-1-
X-RAY FLUORESCENCE SPECTROMETER
FOR AIRBORNE PARTICULATE MDNITORING
Fred S. Goulding and Joseph M. Jaklevic
Lawrence Berkeley Laboratory
University of California
Berkeley, California 94720
SUMMARY
The purpose of this contract was to develop an elemental monitoring sys-
tem for air participates based on analysis by an energy-dispersive X-ray spec-
trometer. The program has included the development of a carrier for cellulose
filters (that are relatively free of trace metals), of an automatic sampling
station to expose these filters to known amounts of air, and of a fully auto-
matic analysis system capable of analyzing for most of the important elements
at levels in the 10 ng/cm2 range in the filters. The basic design of the sys-
tem has been described in earlier progress reports, and particularly in the
first one (see Appendix I). This final report will mainly discuss experimental
results achieved during the final system testing, but any modification made in
this phase of the work will also be detailed.
Final assembly and testing of both the sampling and analysis station have
been completed and the units shipped to EPA at Research Triangle Park, North
Carolina. There the system has been reassembled, tested, and calibrated by
LBL personnel.
Virtually all of the design objectives outlined in previous reports have
been achieved. Recent modifications to the hardware and computer programs to
accommodate the automatic sequencing of three fluorescer targets were success-
fully completed. An extensive program of testing for accuracy and stability
has been carried out and the results of the calibration have been validated by
comparison with known standards. The analysis accuracy and the detectable
limits achieved are within previous predictions for most elements; no signif-
icant drift or instability in the calibration has been observed.
-------
-2-
1. INTRODUCTION
Previous reports have described various components of the LBL aerosol
particulate analysis system and the progress made in assembling the complete
system. The spectrometer has now been completed, tested, and delivered to
the EPA. During this final interval no important changes were made in the
basic design. Hardware changes have been limited to engineering refinements
indicated by extensive testing in the study of particulates at Pasadena dur-
ing October and November. The major effort in the last period has been
devoted to extending the capabilities of the computer program to include the
three fluorescent energies, and to developing and testing the final data
analysis program. An extensive experimental program designed to evaluate
the performance of this system and validate the analytical results has been
completed.
Since the basic design remains unchanged, the emphasis of this final
report will be on the experimental results obtained in the final stages and
their relationship to system performance. Where necessary, a brief review
of design details discussed in previous reports will be included. Documents
containing more detailed engineering data and an operator's manual are in
the process of being prepared.
-------
-3-
2. MONITORING SYSTEM DESIGN
A. Filter Specifications
No changes have been made in the filter holder. The samples consist
of 2" x 2" plastic sample holders containing 37 mm diameter filters. Mech-
anisms in the air sampler and X-ray fluorescence unit are designed to manip-
ulate and accurately position these samples. The 37 mm diameter was chosen
as the optimum size consistent with availability of suitable filters--both
Millipore and Nucleopore filter substrates are available in these sizes.
Cellulose filters can also be cut to this diameter.
Distortion of filters by cold flow of the plastic holder caused by
careless storage has caused some problems but appears to be easily corrected.
Filter holders can be returned to their original shape by pressing between
two plates heated to 50°C; subsequent storing in a tightly packed box retains
this shape.
A supply of 200 holders with mounted 0.8 y Millipore filters have been
delivered to EPA. In addition, 1000 blank filter holders have been delivered
together with the jig used to glue filters to the holders. These should be
adequate to support an initial experimental program by EPA personnel.
B. Range of Elements Measured
The range of elements measured in the completed system is the same as
that described in an earlier progress report. For convenience, this list (Table 1)
is repeated here together with the X-ray absorption edge energies and the
energies of the principal emission lines.^ '
-------
-4-
TABLE 1. Elements Included In Analysis
ELEMENT
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
Ga
As
Se
Br
Kb
Sr
Zr
Mb
Mo
Cd
In
Sn
Sb
I
Ba
ATOMIC NO.
13
14
15
16
17
19
20
22
23
24
25
26
27
28
29
30
31
33
34
35
37
38
40
41
42
48
49
50
51
53
56
K BINDING ENERGY3^
1.559
1.838
2.142
2.470
2.819
3.607
4.038
4.964
5.463
5.988
6.537
7.111
7.709
8.331
8.980
9.660
10.368
11.863
12.652
13.475
15.201
16.106
18.00
18.99
20.00
26.71
27.93
29.19
30.49
33.16
37.40
Ka ENERGY3)
1.486
1.739
2.013
2.307
2.621
3.312
3.690
4.508
4.949
5.411
5.894
6.398
6.924
7.471
8.040
8.630
9.241
10.53
11.21
11.91
13.73
14.14
15.7
16.6
17.4
23.1
24.1
25.2
26.3
28.5
32.1
Kg ENERGY^
-
-
-
-
-
3.590
4.012
4.931
5.426
5.946
5.489
7.058
7.648
8.263
8.904
9.57
10.26
11.72
12.49
13.29
14.96
15.83
17.7
18.7
19.6
26.2
27.2
28.4
29.7
32.2
36.3
ELEMENT
Hg
Pb
ATOMIC NO.
80
82
L BINDING ENERGY15^
14.21
15.21
Lctj ENERGY
9.99
10.55
Lgj ENERGY
11.82
12.61
a) Ref. 6. Energies are in keV. For heavier elements the energies are
averages of the partially resolved times.
b) The L shell binding energy quoted is the L,,. shell.
-------
-5-
The choice of secondary targets has been changed slightly since earlier
reports. Cu, Mo and Tb have now been chosen as a reasonable compromise
between uniform analytical sensitivity over a wide range of atomic numbers
(Z) and availability of pure materials. The energies of the principal lines
for these targets are 8.0 keV, 17.4 keV and 44.0 keV respectively. The proba-
bility of excitation for a given element depends strongly upon the difference
between excitation energy and the X-ray absorption edge. Using these three
excitation energies, the calculated relative probabilities for characteris-
tic X-ray generation are as shown in Figure 1. The data represents cross
sections for equivalent photon flux using each of the three targets. Each
curve gives the Z dependence of the yield for a given secondary target--the
abrupt cut-off at high energies represents the K-shell ionization threshold.
Since the characteristic X-ray yield converts directly into analytical sen-
sitivity, the reason for the choice of the three fluorescers becomes apparent.
C. Data Processing
No major changes in the data processing system have occurred. The hard-
ware consists of a TI960A computer interfaced directly to the X-ray analyzer
for control functions, and to a Northern Scientific 1024 Channel ADC for data
acquisition. Input/output to the computer is accomplished either via tele-
type or IBM compatible magnetic tape. In addition, there are a certain num-
ber of computer functions which can be controlled by means of a hard-wired
control panel. The total memory capability of the computer has now been in-
creased from 12,000 to 16,000 16-bit words in order to adequately handle the
three fluorescers.
-------
-6-
To facilitate program development we have written an assembler operat-
ing in the LBL CDC 6600/7600 central computer system. This has enabled us
to assemble programs on these computers which generate program magnetic tapes
that can be loaded directly into the TI960A computer. The use of this assem-
bly language has greatly facilitated program development in the later stages
and will continue to be useful for future program development.
D. Sampling Station
No basic changes have been made in the sampling station design in
recent months. The schematic of Figure 2 shows the basic elements. Minor
modifications of the sequencer module were implemented in order to increase
the electronic noise immunity of the control logic. A continuous testing
program has indicated that the unit operates reliably over long periods.
The air sampler has been calibrated in a series of measurements in
«
which the output flow rate has been measured for a number of filters and
compared with flow estimates based upon filter area and pressure differen-
tial. It- is important to determine the effective area of the filter so
that numbers obtained from the X-ray fluorescence analyzer, which are expres-
sed in ng/cm2 of elemental impurities on the filter, can be converted into
ng/m3 of the air sampled.
The air flow through the filters was 52.72 1/min as measured by
a compensated displacement gas meter. The average pressure differential
across the filter in these tests was 11.5 inches of Hg. By observing the
density profile of particulates on a number of exposed filters it was deter-
mined that the effective area of the filter in the sampling system was
7.96 cm2, and that a gradual decrease in deposition of particulates occurs
-------
-7-
over the outer 1.7 mm of the periphery of the opening in the filter holder.
These results are consistent with the published Ap vs. velocity relation-
ship for 0.8 y Millipore filters. For Ap = 11.5" Hg, the air velocity is
109.3 cm/sec, and, assuming an effective area of 8.00 cm2, the conversion
factor (based on measured flow rate or measured pressure drop) is 1 ng/cm2
of filter s 1.265 ng/m3 of air for a two-hour air sample. This conversion
factor holds only for 0.8 y Millipore filters. For other types of filters
or pore sizes, the conversion factor can be calculated from the pressure
differential if the Ap vs. velocity characteristic for the filter are known.
3. ANALYSIS STATION
A. Semiconductor Detector Spectrometer and Associated Electronics
Details of the X-ray spectrometer system have been extensively docu-
(2 31
mented elsewhere. ' The guard-ring detector is 6 mm in diameter and 3 mm
thick. Its energy resolution is 135 eV FWHM on a pulser and 185 eV FWHM on
5.9 keV X-rays. Some idea of the resolution of the system for various X-ray
lines can be obtained by consulting Figure 3 in conjunction with the data of
Table 1. The energy dependence of resolution for the EPA system can be obtain-
ed by interpolating a curve between those corresponding to electronic resolu-
tions of 100 eV and 150 eV. Scanning measurements performed after the assembly
of the detector in the vacuum cryostat indicate that the detector is within
0.2 mm of the axis of the cryostat enclosure and its face is 4.5 mm from the
front surface of the cap. Accuracy in alignment is important since positioning
of the collimators, sample, and fluorescer targets are all referred to this center-
-------
line. This alignment can only be measured after the system has been assem-
bled, pumped, and cooled to 90°K since the relative contraction of various
members can result in dimensional changes of several millimeters.
The efficiency of the detector was calculated using data based on pre-
vious measurements of detector window thicknesses and a compilation of X-ray
absorption cross sections. -* These data are summarized in Figure 4. Each
of the individual curves represents the effect of that particular component
on the relative efficiency. For example, the combined effect of 0.001" Be
and 2 cm air would be the product of the two curves shown. The existing
system can operate either with helium or air; the intrinsic detector window
and 25 /u Be are always present. Mbre recent measurements on high-energy X--
ray yields indicate that the roll-off of efficiency at higher energies is
somewhat more rapid than indicated in these curves. This is believed to be
due to fringing effects around the periphery of the detector and can start
to affect the efficiency at energies as low as 10 keV. The effects become
important in the later discussion concerning the calibration for higher
energies.
B. Mechanical Design
The X-ray tube-flucrescer and detector assembly has been shielded
with aluminum-lined lead absorbers to keep radiation exposure to people in
the area well below tolerable limits. Measurements performed with and with-
out the shielding have established that virtually no additional background
is caused in the measured X-ray spectrum by the Al-lined shields. The
-------
-9-
shielding, together with electrical interlocks on access doors are designed
to ensure safe operation. The maximum radiation level occurs when the tube
is operating at 75 keV and 400 uA--at the outside surface of the cabinet it
is less than 5 mr/hr at all points.
C. X-ray Tube - Fluorescer - Detector Geometry
No basic changes in design have been implimented in the final period.
However, a new X-ray tube has been installed and the secondary targets and
fluorescer changer design has been improved. Figure 5 is a cross section
showing the final tube-fluorescer-detector geometry. Extensive measurements
were performed to optimize X-ray intensity and peak/background of the excit-
ing radiation. These included measurements of the electron beam spot size
on the tube anode using X-ray photographic techniques, and emission vs.
anode voltage curves to establish cathode excitation efficiency. Upon achiev-
ing maximum beam on target in the tube, the complete geometry was then adjust-
ed to yield maximum counting rate. Measurement of spectra obtained without
samples in position were used to evaluate peak/background ratios and to in-
spect for spurious lines. Additional X-ray transmission filters were inserted
at crucial points to reduce scattered background to a minimum. It was found
necessary to redesign the collimator to include an additional shield to
absorb X-rays that could strike the inner collimator surface directly from
the tube anode. Finally, the diameter of the collimator was chosen to allow
the detector to view as large an area of the filter as possible without intro-
ducing excessive scattering from the sample holder frame. In the final sys-
tem, the area analyzed on the filter is a slightly elliptical region with a
major axis of 3 on.
-------
-10-
D. Sample Changer
With the exception of minor modifications designed to eliminate occas-
ional mechanical failures, the only modification performed on the sample
changer involved the incorporation of a helium atmosphere in the void between
sample and detector. By placing baffles in appropriate places we maintain
an enclosed volume beneath the air filter sample into which helium is con-
tinuously introduced via a channel in the shielding. This reduces X-ray
absorption in the path between sample and detector. While the use of helium
is optional for all three fluorescent energies, maximum benefit results
when using the Cu target.
Operator access to the sample stack-loader is provided through a
shielded door which is interlocked to stop X-ray output when opened. The
control program recognizes the "door-open" condition and restarts the anal-
ysis of the sample being analyzed when the door is closed. This permits the
insertion of samples at any time during an analysis sequence.
4. CONTROL ELECTRONICS
A. Dead Time Correction
Only minor modifications to the control system have been made in
this period to facilitate more convenient manual operation. The fact that
widely varying counting rates are produced for various samples and flucres-
cer targets makes accurate correction for events rejected by the pulse pile
up rejector essential if absolute values of elemental concentrations are to
be obtained. This correction is achieved by automatically extending the
-------
-11-
counting time by a factor which just compensates for the fraction of the
total counts rejected by the pile-up rejector. To accomplish this, the
timer is designed to measure live time--that is the clock pulses are routed
through an electronic gate which is closed for an adjustable period each
time a pulse is processed by the detector electronics. Adjustment of this
periodset by an adjustable one-shot in the timer--is an essential pre-
requisite if accurate analysis is to be performed.
The association between counting losses and system dead time is illus-
trated in Figure 6. This plot of input vs. output rate for a series
of different pulse dead times assumes that an ideal pile-up rejector circuit
is inhibiting the analysis of any pulses which occur close enough in time to
cause false energy information. The roll-off of output rate with increasing
input rate is a fundamental limitation of counting experiments and can be
serious for large dead times such as are used in the X-ray spectrometer sys-
tem (17 ys in the existing instrument). Thus, at 10,000 counts/second input
rate, only 65% of the pulses are ultimately stored.
The dead-time corrector used in the live time clock was adjusted by
requiring that the accumulated counts due to a peak produced by a radio-
isotope source remained constant as the X-ray tube flux was varied to give
a wide range of total counting rates. After final adjustment*the counting
rate in the radioisotope peak was found to vary by 0.5% from 500 cts/sec
to 10" cts/sec (351 dead time) and 2.5% from 500 cts/sec to 2 x 10" cts/sec
(65% dead time). The maximum counting rates obtained in our experience
with air filter samples is 101* cts/sec while operating the X-ray tube at
45 kV and 400 uA with the molybdenum target.
-------
-12-
B, X-Ray Tube Power Supply And Controller
The X-ray tube controller and high voltage supply have been thoroughly
checked out at voltages up to 80 keV. The final settings of the preset
voltage and emission current conditions for each of the three fluorescers
are as follows: a) Cu, 35 keV, 400 yA, b) Mo, 50 keV, 400 yA and c) Tb,
75 keV, 400 yA. These can easily be changed to accommodate conditions which
might occur with different samples. Reproducibility of the settings appears
to be within a few percent; including any possible long term drifts the
currents and voltages should remain within 5% of their preset values.
C, Data Processing
This final period, has been dominated by computer programming for
control and spectral data processing, and by testing the operation of the
programs. The following discussion will be divided into three sections
1) control and routine monitoring functions, 2) spectrum analysis 3) cali-
bration and output. We will describe the operation of the programs in some
detail. Validation of the results is described in the section on experi-
mental studies.
1. Control and Routine Monitoring Programs:
The analysis system is controlled by the computer in its normal
mode of operation. Once a sequence of secondary fluorescers and the
running time of each has been selected on the sequencer and timer
units, the operator is only required to load samples in the input
stack then push the "START" button on the computer. The samples are
-------
-13-
automatically loaded in sequence and the appropriate data taken from
each analysis. The spectral data is stored during the analysis period;
at the end of that time, the spectrum is written on magnetic tape, the
data are analyzed and the results (in ng/cm2) are printed on the tele-
type and written on magnetic tape. After the last sample has been anal-
yzed, the system returns to a standby condition to await the next batch
of samples.
In addition to the fully automatic mode of operation, the sequence
of steps can be controlled manually. A keyboard on the computer control
unit permits the operation of the computer as a pulse height analyzer
and operations such as "START ADC", "CLEAR MEMORY", "DISPLAY DATA", and
"WRITE or READ MAGNETIC TAPE" can all be performed by pushing appropriate
buttons. These features are used mainly during the set-up and calibra-
tion period although the versatility afforded by these options is also
useful in other situations.
In addition to executing these sequencial steps, the computer also
periodically monitors various key parameters of the analysis system in
order to inform the operator via teletype message of any malfunctions.
Some problems are considered serious enough to automatically shut down
the system and await corrective action by the operator, whereas, in
other cases, it is considered adequate to inform the operator that some
condition exists.
A partial list of messages programmed into the computer are listed
on the following page.
-------
-14-
Computer Messages:
FRONT SHIELD OPEN PROGRAM STOPPED
SAFETY INTERLOCKS DEFEATED PROCEED WITH CAUTION
BACK DOOR OPENED
NO EMISSION FROM XRAY TUBE
XRAY TUBE OVER HEATED
HIGH VOLTAGE SUPPLY FAILURE
SAMPLE CHANGER STUCK
FLUORESCER CHANGER STUCK
LIQUID NITROGEN LEVEL LOW
The program is designed to be as communicative as possible so
that an experienced operator can recognize most malfunctions and make
appropriate responses without consulting operating manuals and trouble-
shooting guides. Our experience in testing the control program has
been that we were reasonably successful in achieving this goal.
2. Spectrum Analysis:
Crucial to the problem of converting a measured counting rate into
a elemental concentration is the extraction of an accurate peak area
from the pulse-height spectrum. Since the energy resolution capabili-
ties of the Si(Li) detector does not prohibit some interference of
peaks from neighboring elements in the periodic table, it is necessary
to accurately subtract out the lines produced by each element so that
the value for an adjacent element is not affected. This is a problem
-------
-15-
that has received much attention in nuclear counting experiments.
Generalized formalisms exist for extracting the deconvoluted data
from extremely complex pulse-height spectra. Although one might argue
that X-ray fluorescence spectra are much more densely packed with peaks,
and hence more difficult to analyze, there are certain reasonable
assumptions which can be made regarding X-ray spectra which greatly
simplify the problem. Specifically, the number of possible lines in
a spectrum is limited and the energies and relative intensities of
the lines from any one element are known beforehand. Thus, to a good
approximation (virtually exact for air filters) the only free parameters
in the problem are the magnitude and shape of the background under the
peaks and the relative intensity (concentration) for each of the constit-
uent elements.
When analyzing a series of air samples acquired on the same filter
material, the continuous background due to scattering from the matrix
(filter substrate) is similar from sample to sample. Furthermore, the
variation in background due to varying matrix contributions is mainly
one of relative magnitude (rather than shape) provided the average Z
of the matrix remains unchanged--as is the case for filter media which
are predominantly hydrocarbons.
We have taken advantage of this fact by using for the background
shape a spectrum acquired by analyzing a blank air filter of the type
used for particulate collection. This has the advantage that any trace
impurities arising either from the filter medium or caused by the anal-
ysis system itself, are automatically excluded from analysis. Tests
-------
-16-
performed on heavily loaded filters showed that the background contri-
bution could be accurately subtracted out over almost all the energy
range covered by the spectrum by normalizing the blank filter spectrum
to the unknown spectrum just over the region containing the coherent
and incoherent scatter peaks of the exciting radiation. This is valid
since the scattered radiation is dominated by the matrix as long as the
filter loading does not appreciably change the average Z of the sample.
Calculation of the areas of the X-ray peaks is potentially a more
difficult problem due to the variation in relative intensities of the
elements and the possibility of interferences from spectral lines. The
usual approach to this problem is to mathematically generate a function
which simulates the response of the system to a particular photon energy.
The shape is then fitted to the experimental data, usually by a least
squares method, and the peak location and area determined. As noted
earlier the peak location for X-rays is fixed; furthermore, the com-
plexity of the X-ray lines, both in terms of Ka/K3 ratios and in regard
to partially-resolved Kal9 Ka2 lines makes the synthesis of such line
shapes difficult. Fortunately, the profile of the instrumental response
to the X-rays of an e"! ment can be established by running a spectrum
from a pure sample of the element. We have taken advantage of this by
using a stored spectrum of the instrumental response to the X-rays of
an element as a shape standard for that element. Thus the problem of
determining the area of a given peak reduces to that of comparing the
intensity of the unknown peak to that of a stored shape-standard spectrum.
Once this intensity ratio has been established, the total contribution
of that element to the spectrum, including all X-ray lines, can be sub-
tracted from the spectrum.
-------
-17-
Figure 7 illustrates the operation of the program on a schematic
spectrum as shown in the upper left hand portion of the figure. The
blank filter spectrum is first compared to the unknown sample over the
specified shaded region, normalized, and then subtracted point by point
to give the STEP #2 spectrum. The lines due to elements #1, #2, etc.
are then sequentially stripped out after first comparing intensities
over the shaded regions to derive normalizing factors. Any selected
portion of a spectrum can be specified for comparison--this is neces-
sary when considering more complex spectra where regions with over-
lapping lines must be avoided in making the comparison. After sub-
traction of the last elements, the residual counts in the spectrum
should be zero within statistics.
In practice, the problem is slightly more complicated. Samples
used for generating shape standards must be made of material suffic-
iently thin to prevent changes in the Ka/K$ ratio due to self-absorp-
tion. Also, to ensure that interference effects between elements are
minimal, it is necessary to carefully choose the order in which ele-
ments are removed from the spectrum. (For example Pb before As, Fe
before Co, etc.) In the present program, the sequence of removal of
elements is specified by including each element in one of three groups.
These three groups are analyzed in order, but the elements within a
group are removed from the spectrum in order of decreasing intensity.
In the above case, Pb and Fe might be in group 1 while As and Co could
be in groups 2 or 3. This presupposes some knowledge of possible inter-
ferences on the part of the person specifying the order. Other features
-------
-18-
of the program include a peak-shifting routine to correct for any small
shifts in peak location which might have occurred between the time that
the shape standards were stored and that when the unknown spectrum was
run. Our present experience indicates that the stability of the system
when running continuously is adequate to use the same shape standards
for several weeks of running.
The ultimate test of any such program lies in its ability to repro-
ducibly calculate the correct answer. Table 2 is a summary of measure-
ments performed on spectra acquired in 11 sequential analyses of the
same air filter. Each analysis consisted of a separate data-taking
interval with the acquired spectrum being analyzed using the described
program. The entries in the table include the areas calculated by the
program, the RMS deviation of the 11 measurements, the mean deviation
of the 11 measurements, and the KM3 deviation (variance) for any one
measurement calcualted from the equation:
Variance = \/Peak Area + Background
The good agreement between the statistical variance (calculated) and
the RMS deviation in the experimental results shows that there are no
extraneous or systematic program errors. The average ratio of MEAN/RMS
for the measured numbers is 0.82; for a Poisson distribution this ratio
should be 2/ir = 0.799. This supports the conclusion that the variations
are purely statistical.
-------
-19-
TABLE 2.
Reproducibility o£ Areas Calculated With Computer Program
ELEMENT
Mean Area Calculated
From 11 Measurements
RMS Deviation
(Measured)
Mean Deviation
(Measured)
RMS Calculated from
Areas (Theoretical)
Ratio Mean/RMS
(Measured)
Ca
1244
65
54
89
0.83
Ti
424
49
40
58
0.81
Mn
281
27
25
48
0.93
Fe
11584
184
144
130
0.78
Cu
5175
115
98
120
0.85
Zn
3102
65
50
85
0.77
Br
9727
182
137
217
0.75
Pb
14243
173
140
162
0.81
3. Calibration and Data Output:
The program elements which we loosely term "Calibration" contain
portions of both the control and peak analysis programs. We refer to
that portion of the program with which the operator interacts initially
to enter into the computer the necessary data for complete analysis of
the acquired spectra.
In the CALIBRATION mode, the program will automatically sequence
and accumulate data from a preselected set of samples containing pre-
pared mixtures of elements to be used as shape standards. The elements
included in each sample are chosen so as to avoid interelement inter-
ferences. In the present arrangement, ten samples are used to cover
the 33 elements listed in Table 1. In the same sequence the computer
-------
-20-
will accumulate background spectra on a blank for each of the secon-
dary fluorescer energies. Since this process of analyzing shape
standards and background is completely automatic, it can be performed
overnight without operator attention.
When the sequence is finished the operator enters the additional
data required by the analysis program by using a subroutine called
from the control panel. This routine asks the operator a series of
questions which must be answered for identification of components of
the spectral features associated with each element. Figure 8 is a
typical K X-ray spectrum with the relevant regions labeled. PRIMARY
PEAK is the peak whose area is calculated as a measure of the intensity
of the particular element. It is usually the Ka for light elements
and either the La or Lg for heavy elements. SECONDARY REGION is any
other region to be subtracted out of the spectrum when this element is
stripped from it, subject to the same normalization conditions as for
the primary peak. A TERTIARY REGION can also be specified to allow for
features such as Si X-ray escape peak for light elements or multiple L
lines for heavy elements. The FIT REGION is the region of the peak
which will be compared to the unknown spectrum in order to establish
a normalization factor. It is chosen for maximum statistical signif-
icance while avoiding the possibility of overlapping with lines of
other elements in the sample. The PEAK CHANNEL is required as identif-
ication of the peak to determine the sequence in which the elements
are stripped from the spectrum. The regions are specified by the opera-
tor responding to the questions asked by the computer by setting a dis-
play marker at the lower limit of the region and by then typing in the
width (in channels) of the region.
-------
-21-
The data regarding relative X-ray excitation efficiency and analy-
sis sequence group are also entered during this calibration phase.
The X-ray efficiency factor is used to convert the peak area to ng/cm2
in the final stage of the data analysis.
The series of questions asked by the computer and some typical
responses are as follows:
ELEMENT FE (or any other listed
in Table 1)
TO SKIP AN ENTRY, TYPE COMMA
SET MARKER AT BOTTOM, TYPE CHANNELS IN REGION
PRIMARY PEAK 22 (Ka peak)
SECONDARY REGION....20 (K3 peak)
TERTIARY REGION 15 (e.g. Si escape peak)
FIT REGION 8
MARK PEAK, HIT CARRIAGE RETURN (peak channel specified)
F = 835 (relative X-ray efficiency)
ANALYSIS GROUP (0-8) 2 (specifies group in which
element will be subtracted
out: 0-2 Cu target
3-5 Nb target
6-8 Tb target
The next element in the standard spectra is then displayed by the opera-
tor and the process steps repeated until the necessary information has
been initiated for all elements of interest. Following this, analysis
regions are specified for each of the three background spectra--the pro-
gram is then capable of analyzing an unknown spectrum for the elements
entered.
The only other data required to convert the peak areas to ng/cm2
is the normalization factor which multiplies the relative intensities
-------
-22-
o£ the elements by a number which takes into account the data acquisi-
tion interval, X-ray tube current, etc. This number is determined by
analyzing a known single-element intensity standard under the same con-
ditions as the filters. The answer given by the computer analysis of
this standard using a "dummy" normalization factor of unity is then
divided by its known concentration to give the proper system normaliza-
tion factor which is then entered into the program. Periodic analysis
of the intensity standard serves as a check on the continued accuracy
of the calibration.
Although this procedure may appear somewhat time consuming, it can
be completed in two to three hours of operator time for all fluorescers
and targets and need only be repeated every several weeks at the most.
Table 3 gives the results of the analysis of a typical urban aerosol
sample in the output format generated by the program. The elements are
listed in the order in which they were stripped from the spectrum. (A
later version of the program will deposit the results in an output buf-
fer and the output will be generated in a standard order.) The column
entries are:
a) total peak area
b) area of the background under peak (FWHM)
c) calculated concentration in ng/cm2 averaged over the filter area
d) statistical error in ng/cm2 (this does not take into account any
other sources of error).
The speed of analysis of a spectrum is limited by the time required to
type the results. Table 3 was typed out in about 50 seconds, spectral
analysis being achieved while typing was in progress.
-------
-23-
TABLE 3. Computer Output Of Results
PB
BR
FE
CA
K
TI
CU
SE
V
AS
CO
RB
ZN
SR
CR
NI
MN
PK INTEGRAL
22539
15385
7535
1525
2312
306
381
223
149
805
0
26
1082
485
365
244
250
BKGD INT.
6599
1694
1747
1783
3539
1054
2817
1315
1178
1348
980
4279
1649
11813
1146
994
1121
NGM/SQ CM
1466
466
739
595
1162
74
22
7
28
28
0
0
53
13
53
15
31
+ -
12
5
10
23
39
10
5
2
8
3
4
3
4
4
6
3
5
-------
-24-
5. EXPERIMENTAL RESULTS
A. Calibration:
Before proceeding with the details of the experimental tests of the
calibrated instrument, a few brief remarks concerning the theoretical basis
for the method are given.
The theoretical equation for the intensity (I.) of an X-ray line of ele-
ment i for the case of a homogeneous sample of thickness d irradiated with
photons of energy E and incident flux I is given in Figure 9. This expres-
sion is derived by integrating the absorption cross section for the incident
X-rays and the escape probability for the fluorescent X-ray over the thick-
ness of the sample. For the case of very thin specimens (such as air filters),
the expression in brackets reduces to unity, and the equation simplifies to:
Ii = [I0 G] [Ti >L 6i] Pi d (1)
where G includes all geometry factors, T. is the photoelectron cross section
for the i^n element at energy E , ox, T is the K,L shell fluorescence yield,
e- is the intrinsic detector efficiency for photons of energy E., and p. d is
the concentration of the element in gm/cm2. The first bracketted term is
independent of which element is being analyzed and represents a total normaliz-
ing factor which we call 1/N. The second term is specific to the element i
and represents a selective K or L X-ray excitation cross section which we call
F---these are the quantities plotted from theory in Figure 1. The concentra-
tion of any element is then expressed simply as:
Pi = ^N (2)
-------
-25-
The F. and N are now recognized as the relative excitation efficiency and
normalizing factor referred to in the previous section of this report describ-
ing the computer program. Obviously the most critical part of calibration is
obtaining the proper values for F.. Although reasonable values for the
quantities of interest are available in the literature, there are certain
assumptions inherent in the derivation of the equation which are not rig-
oriously satisfied in practice. However, the theoretical values do tell us
that the total yield is a simple well-behaved function of Z. With this modest
assumption it is possible to generate a complete calibration curve by measur-
ing the relative F. for a series of elements and assume that data can be inter-
polated between measured values subject to the constraints imposed by the
theoretical equation.
We have performed such a set of measurements using a series of evaporated
thin film samples of known concentration in yg/on2. Where convenient films
were not available, we have used compounds with known ratios of elements to
give data for regions not otherwise covered in the study. One consequence
of these measurements was the discovery that the relative efficiencies shown
in Figure 1 did not take adequate account of the reduction in detection effic-
iency at higher energies --particularly with reference to the fringing effects
around the periphery of the detector. After applying this correction to the
theoretical factors, the results shown in Table 4 were obtained. The excel-
lent agreement between the standard concentrations and the measured values
demonstrates that the system is capable of accurate calibration for single
element standards. Further validation was obtained from analysis of chemical
compounds in which the relative concentration of elements was well known
-------
-26-
TABLE 4. Comparison of Measured Concentration of Standard Samples
ELEMENT
Al
Si
S
Ti
Cr
Fe
Ti
Cr
Fe
Ni
Cu
Pb
Zr
Pd
Cd
Sn
Ba
FLUORESCER
Cu
Cu
Cu
Cu
Cu
Cu
MO
MO
Mo
Mo
Mo
Mo
Tb
Tb
Tb
Tb
Tb
MEASURED DENSITY^ (ug/cm2)
1550 c)
2380
2820
90
117
83 e)
100
121
94
109
55
132
66
138
92
142
122
ACCEPTED VALUED (yg/cm2)
2200 d)
2430 d)
2970 d)
101
122
83
101
120
87
100
49
131
61 f)
142
93 f)
138
124 f)
a) Statistical errors are less than 1% in all cases of evaporated films.
b) Thicknesses of evaporated films were determined by weighing. Estimated
errors are <5%.
c) The discrepancy in this comparison could easily be due to heavy elements
in the 1100 Al alloy used.
d) These densities represent the effective weight of infinitely thick samples.
e) The Cu fluorescer comparisons are normalized to the Fe value.
f) These were obtained by using samples of ZrBr^, CdBr2, and BaBr2 ; the Br
intensity was measured with the Mo fluorescer.
-------
-27-
(e.g., Ba Cr 0,, Pb Cr 0., K JMh 0,, etc.). Measurements performed over a
period of two weeks indicated that the calibration was stable to within 3%
over this interval.
Figure 10 is a plot of the above data showing the interpolation curves
used for elements not included in Table 4. The F. values used in the final
analysis program were taken from these curves. Comparison of Figure 10 with
Figure 1 shows the effect of the reduced detector efficiency at higher X-ray
energies.
Analysis of more complex samples, such as air filters, involves additional
considerations such as interference between spectral lines from two elements,
particle size and matrix effects. Spectral-line interference is a property
common to all energy-dispersive X-ray fluorescence analysis systems and will
be treated independently from the questions of particle size and matrix effects
which are inherent in any X-ray fluorescence method, and which become serious
problems in calibration for light elements.
The degree to which the complex spectrum can be accurately reduced into
its constituent characteristic X-rays depends mostly upon the sophistication
of the peak analysis program. The ultimate limits of accuracy of such spectral
analysis are set by counting statistics--accuracies for individual elements
depend upon the relative magnitudes of any possible interfering lines. Some
idea of the effect on sensitivity produced by spectral-line interference can
be obtained by considering a typical case of interference--the Fe KB/Co Ka
overlap. Let the Fe K3/Ka ratio be R, the rms deviation in a concentration
measurement N be a, and the number of background counts under a peak in the
Fe-Co energy region be R,.
-------
-28-
Then:
a2(Fe) = Npe+NB
a2(Co) = NCo + NB + R (Npe + NB)
where Np , NC , NR are the number of counts in the iron Ka peak, cobalt Ka
peak, and background (assumed constant) respectively. The value of R is
about 0.2, so
a2(Co) = NCQ + 1.2 NB + 0.2 Npe
(the detectable limit for Co is now a function of the concentration of iron).
For a large iron concentration, this equation becomes
a2(Co) = 0.2 Npe
or 30 (Co) = 1.5 a (Fe)
In other words, the detection limit for cobalt in the case of high iron
concentration is about 1.5 times the variance in the iron concentration
determination.
Table 5 is a tabulation of data showing the 3a detectability limit for
Co as a function of Fe concentration. The values for N~, Np and Np are
based on experimental measurements.
-------
-29-
TABLfi 5.
Detectibility of Co as a Function of Fe Concentration
for 5 Minute Counting Interval
Fe Concentration (ng/cm2)
0
10
102
10 3
10"
105
NFe
0
1.5 x 102
1.5 x 103
1.5 x 10*
1.5 x 105
1.5 x 106
NB
1.75 x 10
ii
ii
it
ii
ii
Co Detectibility (ng/cm2)3-1
9.1
9.2
9.6
14.1
35
109
a) Defined as 3o of the background.
The most meaningful judgement of accuracy in multiple element analysis
is to apply the program to real samples and compare the results with indepen-
dent measurements. Table 6 presents a comparison of results obtained using
the complete system on spectra obtained using Mo excitation which is generally
the most complex. The data are the results of analysis of eight separate air
filter samples with widely varying loadings. They are compared with indepen-
dent measurements by R. Giauque using the X-ray fluorescence system described
in Ref. 6 which has been accurately calibrated, and its results verified
using atomic absorption and neutron activation for many samples including air
filters. The agreement between the data is particularly impressive when one
-------
-30-
considers that the whole analysis time in the EPA system was only six minutes
including data acquisition, spectrum analysis and teletype output.
Based upon the analysis of these and other samples, the variance in a
series of measurements in which the errors, due to counting statistics can be
made arbitrarily small is better than 10%. The accuracy of results for com-
plex analysis is limited mostly by the sophistication of the computer analysis
program and the degree of ingenuity used in its application. However, it is
safe to say that overall rms errors less than 20% can be easily achieved for
most elements.
Similar data have been obtained for the Cu and Tb fluorescers. In each
of these cases the multielement interference problems are smaller for most
samples relative to Mo excitation. For the higher Z elements, the accuracy
is limited mostly by counting statistics in the peak due to the less abundant
heavy elements. For the Cu fluorescer, the absorption of low energy X-rays
<
by the sample limits the accuracy. These effects are discussed in more detail
in a later section.
-------
-31-
r<
CD
s
CD
CO
g
CD
(/I
H
ctf
U
oo
-
vo
LO
-
to
(Nl
Oi rH
w
3
CO
O CNl
(NJ LO
4-1 4-1
LO tO
tO (N]
3- LO
VO CN]
4-1 41
Ol I~~
Ol I~-
rH rH
t~~. to
rH *^f
4-1 4-1
vo CD
to o
CNl CNl
OO LT>
rH ^"
00 VO
Ol 00
CNl (N]
oo M-
rH "*
4-1 41
to to
^t oo
to to
00 LO
rH CNl
4-1 4-1
tO rH
to o
to to
00 LO
rH (NJ
4-1 41
oo to
CN) Ol
to to
t^ LO
T-H CNI
+1 4-1
rH rH
vO C^
CNl (Nl
I \
' ' * '
efl
U
(NJ LO
rH rH
4-1 4-1
T 1 VO
VD LO
1 1
4-1 4-1
01 to
T ( «vf"
vo o
CN]
4-1 4-1
VD CNl
rH vO
Ol LO
rH
4-1 4!
CXI O
rH CN]
4-1 +1
CO Ol
to *3-
o o
T-H rH
+1 4-1
LO 00
to *
CNl O
rH (NJ
4-1 4-1
to t~~
£~~ OO
, 1
41 4-1
o r--
to to
H
H
LO
4-1
(NJ
CNl
LO
LO
V
LO
4-1
00
LO
4-1
CN)
LO
4-1
CNl
LO
41
LO
V
/ (
1 '
*
00
41
CNl
(Nl
CNl
V
CNJ
CNJ
V
to
(Nl
\f
oo
+1
(X]
4-1
LO
rH
41
4-1
T-H
rH
H
00 Ol
rH
41 4-1
rH O
o oo
LO >sr
41 4-1
0 0
OO OO
T 1 rH
VD O
rH
+1 4-1
CNl tO
rH ^f
CNl CNl
t-- LO
rH
4-1 4-1
Ol rH
to to
CO tO
(Nl
41 4-1
LO CNl
LO LO
OO rH
CN)
4-1 4-1
rH rH
(NJ CN]
oo to
CNl
4-1 41
to vo
=* ^t
OO O
(NJ
41 41
O LO
rH O
^ Q
>_ >_
CD
PH
tO «d- to «if to
41 4-1 4-1 4-1 4-1
OO Ol CNJ CO VO
t 1 i 1 **
* to
rH 4-1 41
tO rH tO
V V V VO O
rH
-3- to
rH 41 4-1
tO rH tO
\f y V VO tO
T-H
tO "3- tO
rH 4-1 4-1 4-1
tO rH
\/ v *$ r-~ to
(NJ
to «* to
T-l +1 +i +1
tO rH
V V VO t-- rH
CNl CNl T-H
rH
^- to to
4-1 4-1 41
to \o
v v to LO to
OO LO «3-
to CNI ^t to to
41 41 4-1 4-1 4-1
LO CO *3" tO CNl
t~^ LO LO
tO CNl ^J- ^f tO
4-1 4-1 4-1 4! 41
CO ^t ^ rH Ol
Ol 00 LO
H 3 a
Z U Kl
to
4-1
O
to
4-1
CN)
rH
to
4-1
to
rH
to
4-1
CNl
4-1
T-H
Ol
CNl
4-1
rH
to
4-1
i-H
LO
to
4-1
00
LO
H
4-1
rt-
to
to
4-1
t~^
to
4-1
Ol
CNl
rH
to
4-1
rH
to
41
0
t--.
rH
to
4-1
T 1
T 1
T 1
to
4-1
LO
rH
to
4-1
rH
OO
,_^
^
(H
m
T-H
4-1
OO
o
4-1
0
LO
4-1
Cl
CNl
rH
4-1
Ol
rH
D~~
4-1
rH
r--
rH
LO
4-1
CNl
rH
4-1
LO
rH
LO
4-1
LO
01
/ 1
H
v-^
rH tO
rH LO
4-1 4-1
Ol O
CNl CNl
to to
rH rH
\o oo
4-1 +1
CO v£>
Ol Ol
rH rH
oo to
CNl
4-1 4-1
rH tO
rH ^D
LO LO
00 00
(XI
41 4-1
LO (N]
vO O
\O [""*-
Ol O
to
4-1 4-1
to f-
r-~ t-~
00 CNl
CN]
4-1 4-1
rH Ol
LO ^3-
CO VO
to
41 41
O vD
OO rH
vO t^
00 VO
rH
41 4-1
LO (N)
to to
,0
(/I
CD
.0
n
T-H
CD 3
CD
?H f-i
CD O
CD
rt o
& CC
1 1 ?H
H
t-H
rH O
Oo\Q
CD
statistics
an 10%, a I
system.
iauque.
o/VC U
S 4-J <£
4-J CO LLJ DH
O
c/> is rd
4_> -H CD CD
rrors reflec
uoted error
alues measu:
alues measui
w cr P- p.
Oj T-l 'H
H
-------
-32-
B. Sensitivity
There are a number of ways to express the trace-element detection limit
of an analytical method. The usual convention is to define the lower limit
of detectability as the amount of an element required to produce a signal
equal to the 3a fluctuation in background under a peak. However, in order
to be meaningful, this number must include a specification of the counting
interval, and the total size of the sample analyzed. Perhaps most important,
these limits should represent experimental measurements on realistic samples
and not extrapolations based upon idealized conditions.
Figures 11 and 12 are two logarithmic plots of X-ray fluorescence
spectra taken on the same air filter using the Cu and Mo fluorescer. (Sim-
ilar data were acquired with the Tb fluorescer but are not shown here since
a small Ba peak was the only observed elemental line). Each spectrum was
acquired in a five minute counting interval; the area of the filter analyzed
was 7 cm2. Typical concentrations are Pb - 1.32 yg/cm2, Fe - 0.5 yg/on2,
Mn - 30 ng/cm2, Ca - 0.5 yg/cm2. Counting rates were 10 K/sec for Mo, 5 K/sec
for Cu. The background beneath the peaks is associated with the coherent/
incoherent scatter peaks which are produced by air scatter and by scatter
from the 5 ng/cm2 MLllipore filter substrate. (A better peak/background
results from operating in a He atmosphere, since, in the case of Mo excita-
tion, 2/3 of the scattering is from the air surrounding this sample.)
Using these data, together with the air sampler calibration data, we
calculate the sensitivity vs. Z curves given in Figure 13. These numbers
represent the minimum detectable limit (3a above background) expressed as
ng/m3 in a two hour air sample which is then analyzed for five minutes in
-------
-33-
the X-ray system. (To convert these limits to ng/cm2 on the Millipore sample
multiply by 0.8 m3/cm2.) To a first approximation these should correlate
with the relative efficiency factors shown in Figure 10. Slight difference
in the shape of the curves arises from variations in the shape of the back-
ground for a given fluorescer; relative differences from one fluorescer to
the next are also affected by the difference in incident X-ray yield for each
tube setting. In particular, this largely accounts for the values for detect-
ibility measured for Tb excitation being higher then would be expected on
the basis of the calculated sensitivities.
Detectable limits achieved under other running conditions can easily
be obtained by scaling the numbers obtained from the above graph. The detect-
able limits are inversely proportional to the total air flow and to the square
root of the counting intervals. The background for Millipore and Nucleopore
filters are similar whereas the backgrounds from Whatman 41 filters are about
four times larger.
An important consideration in comparing these detectable limits with
results obtained for competing methods is the multiple-element detection
capability of energy dispersive X-ray fluorescence. The three curve segments
in Figure 13 represents the sensitivities for simultaneous detection of many
elements excited with each of the three fluorescers. (This statement is not
rigorously accurate since it neglects reduction in detectibility due to inter-
element interferences; however, in cases where the variation in concentration
between adjacent elements is not large, the data are accurate enough for com-
parison.) It can be argued that the sensitivity could be optimized for one
element, by filtering the X-rays. However, one of the greatest strengths of
the energy dispersive method is its multiple-element capability which would
-------
-34-
appear to be important in environmental research and monitoring. Any cost
analysis of competing analytical methods should bear this in mind.
C. Particle Size and Matrix Effects
As noted earlier, calibration problems arise for light elements due
to the absorption of the very low-energy characteristic X-rays either by the
filter matrix or by the individual aerosol particles. Since the mean absorp-
tion length for these X-rays may be short compared to particle diameters, or
to the filter thickness, the X-ray intensity reaching the detector depends
upon the microscopic location from which the X-ray is generated either within
the particle or in the filter matrix. Calculations of elemental concentra-
tions using the observed X-ray intensity must then include a correction for
this effect.
To calculate a reasonable correction factor, it is necessary to know
something about the particle size distribution and the location of the inter-
cepted particles within the filter material. Information concerning particle
size must be obtained either by restricting the size range reaching the fil-
ter (via impactors for instance), or by making some assumptions regarding the
size distribution in the original aerosols. Similarily, the absorption cor-
rection due to the filter matrix must be estimated by assuming localization
of the particles in the filter, most likely on its surface.
At the best any assumption appears to have dubious merit, so we have
limited our investigation of the problem to estimating the maximum effect
anticipated in certain limiting cases. This has been done by calculating
the difference between the observed X-ray intensity with and without the
-------
-35-
absorption effects. Referring to Figure 9 we see that the difference
between a thin film X-ray intensity and that including the absorption of
the X-rays integrated over a thickness d is given by a factor
_
A = - - (3)
(yQ + yp pd
where y~ and y, are the total absorption coefficient for the exciting and
emitted radiation respectively. If we now associate d with the diameter of
a homogeneous particle, we can calculate the absorption correction A as a
function of particle size. (This calculation will overestimate the correc-
tion for spherical particles since it assumes a constant thickness; however,
since so little is known about particle shapes the assumption is as valid as
any other.) Figures 14, 15 and 16 are the results of calculations for the
case of Al, S and Ca X-rays excited by Cu Ka radiation. The individual
curves represent various assumptions regarding particle composition; the
hydrocarbon assumes a unit density material having the absorption cross sec-
tions of carbon. The results indicate that no serious problems occur for
particles of size below 10 y except in the case of Al. Estimates of matrix
effects can also be obtained from these curves by recognizing that the 5 ng/cm2
Millipore filter is equivalent to a 50 y thick hydrocarbon sample. Thus, if
the material were uniformly distributed throughout the filter, the correction
to the intensity at its maximum would be the value of the hydrocarbon absorp-
tion correction at 50 y. Again the correction is not too serious except in
the case of Al.
-------
-36-
These families of curves represent a guess as to likely chemical con-
stituents of particles. It is possible that more difficult combinations of
elements might produce significant absorption effects (PbS is an obvious
candidate) . Again we are faced with the necessity of making some assumptions
regarding the nature of particulates in order to estimate the correction
factor.
The problems associated with these effects are of course inherent to
the X-ray fluorescence method and are the same regardless of how one excites
or detects the radiation. However, additional information can be obtained
by using a monoenergetic X-ray source to generate the characteristic radia-
tion. As noted in Eq. 3, the correction factor depends upon the absorption
coefficient for both the incident and emitted X-rays. By varying the inci-
dent X-ray energy, two measurements can be performed, one in which absorption
of the incident radiation is neglible over the particle diameter, and the
other where it is significant. Another way of looking at the problem is to
«
consider the higher energy excitation as a probe measuring the total par-
ticle volume, whereas the low-energy excitation samples the surface only.
In this way information regarding the absorption characteristics of the
particle can be obtained. To a first approximation, this measured absorp-
tion correction would be independent of any assumptions regarding particle
shape or composition. A similar arguement could be applied to the question
of matrix absorption within the filter.
-------
-37-
6. CONCLUSIONS
The EPA X-ray fluorescence analyzer is now a complete engineered sys-
tem, and our tests indicate that it is capable of processing large numbers
of air filters with adequate accuracy for this application. Some of the
experimental studies discussed in this report were designed to examine
limitations that will require further exploration at EPA in order to extend
the applications of the instrument to low-Z elements and samples other than
Millipore filters.
REFERENCES
1. C. M. Lederer, J. M. Hollander, I. Perlman, Table of Isotopes 6th Edition,
p. 570 (1967) Wiley.
2. D. A. Landis, F. S. Goulding and J. M. Jaklevic, Nucl. Instr. and Methods,
87_, 211 (1970).
3. F. S. Goulding, J. M. Jaklevic, B. V. Jarrett and D. A. Landis, Advances
in X-ray Analysis, 15_, 470 (1972) Plenum Press.
4. W. H. McMaster, N. Kerr Del Grande, J. H. Mallett and J. H. Hubbell,
Compilation of X-ray Cross Section, Lawrence Livermore Laboratory Report
UCRL-50174.
5. J. S. Hansen, H. V. Freund, R. W. Fink, Nucl. Phys., A142, 604 (1970).
6. R. D. Giauque, F. S. Goulding, J. M. Jaklevic and R. H. Pehl, Trace-
element Analysis with Semiconductor Detector X-ray Spectrometers,
Lawrence Berkeley Laboratory Report, LBL-647.
-------
38
TABLES
Page
Table 1. Elements measured by the system and their X-ray absorption and
emission energies. , 4
Table 2. Reproducibility of areas calculated with computer program. . . 19
Table 3. Computer output of results 23
Table 4. Comparison of measured concentrations of standard samples.. . . 26
Table 5. Detectability of cobolt as a function of iron concentration.. . 29
Table 6. Comparisons between measurements made on several filters by
R. Giauque and using the EPA system 31
Table A.I Elements potentially amenable to study A.4
FIGURES
Fig. 1. Calculated relative K X-ray production yields for three excitation
energies (Cu Ka, Mo Ka, Tb Ka X-rays) 40
Fig. 2. Schematic diagram of the sampling station 4]
Fig. 3. Resolution vs. energy for various X-ray systems 42
Fig. 4. Efficiency of a 3 mm thick silicon detector as a function of X-ray
energy 43
Fig. 5. Diagram of the geometry used in the final design 44
Fig. 6. Variation of output counting rate with input rate for various dead
times. The last pulses have been rejected by the pile-up rejector. 45
Fig. 7. Illustration of spectrum stripping procedure 46
-------
39
Fig. 8. A typical K X-ray spectrum due to one element. The regions identi-
fied in this figure must be defined during the CALIBRATION phase of
set up for the instrument 47
Fig. 9. Expression for the overall efficiency of the process of production,
absorption and detection of the fluorescent X-rays 48
Fig. 10. Actual relative efficiency curves for three fluorescers. These
can be compared with the theoretical curves of Figure 1 <4>9
Fig. 11. Air filters spectrum taken using the copper fluorescer 50
Fig. 12. Air filters spectrum taken using the molybdenum fluorescer 51
Fig. 13. Elemental detection sensitivity curves for the three fluorescers
(two hour sample collection time, five minute analysis time on
each fluorescer) 52
Fig. 14, Calcium X-ray attenuation vs. particle size 53
Fig. 15. Sulphur X-ray attenuation vs. particle size. ^4
Fig. 16. Aluminum X-ray attenuation vs. particle size S5
Fig. A.I Design of filter carrier A. 7
Fig. A.2 Block diagram of spectrometer A. 8
Fig. A.3 Estimated elemental abundancies and system sensitivity. (10 min.
analysis at 100 /*A for each of 3 fluorescers; samples collection at
10 liters/min-cm2) A. 9
-------
40
100
10
1 I I I I I I I I I I I I
RELATIVE PROBABILITY
FOR CHARACTERISTIC X-RAY PRODUCTION
BY MONO-ENERGETIC PHOTONS
Cu Ka
Mo Ka
Tb Ka
10 20 30 40
ATOMIC NUMBER
50
60
XBL 731-58
Fig. 1. Calculated relative K X-ray production yields for three excitation
energies (Cu K«, Mo K«, Tb KaX-rays).
-------
41
ELECTRONIC
CONTROL
UNIT AND
TIMER
FILTER
CLAMP
MECHANISM
-YZZ&L
t
FILTER BEING EXPOSED
EXPOSED
FILTER
'////////////////A
CONTROL VALVE
TEFLON PISTON
VACUUM PUMP
(GAST)
UNEXPOSED
FILTER
SAMPLE
CHANGER
AIR INLET
CHAMBER
50 MICRON
INLET SCREEN
AIR OUTLET
MUFFLER
Fig. 2. Schematic diagram of sampling station.
-------
42
1,000
900
800
700
600
500
400
300
250
200
150
100
ELECTRONIC NOISE
CONTRIBUTION
DETECTOR CONTRIBUTION
ONLY
5 10
ENERGY keV
20
50
100
Fig. 3. Resolution vs. energy for various X-ray systems.
-------
43
o
o
CO
X
o
o
1
1
p
1-1
Si
1 1 1 1 I
in
o
1 1 1 1
TO
1_
o>
c
Q)
TJ
C
o
o
o
JZ
E
E
co
CO
o
CD
o
LU
AONBIOIJdB
-------
44
CD
OC
o
*-!
I
CO
r-
PP
u
05
c
O)
03
0)
T3
"ca
c
a
0)
a>
o
CD
O)
(D
CO
I
b
-------
45
IDDK3
Ld
I-
o:
cr
ID
Q.
I-
ZD
n
DK=
IK-
MI
I III I III ITT*
17(JS
9ps
i T IT i mi i i i i 11 in i i i i i MI
D
H
25D-
S 2DD-J
LJ
cr
SO-
i i i i i nil i i i i 111 n i
1111
i i i i 11 in i ii i 111 M TIII rm
IK IDK I DDK
INPUT RflTE FESS
XBL 7010-6806
Fig. 6. Variation of output counting rate with input rate for various
dead times. The last pulses have been rejected by the pile-up rejector.
-------
46
CO
Q.
S m
O z>
u v>
CM
Q.
UJ
I-
co
£ <
i"
1 3
u v>
it
a.
UJ
I-
co
o
C«4
X
S o
£ E
O UJ
m
0)
Z!
o
(D
O
O
0)
Q_
Q.
00
E
o
0)
Q.
00
**
O
o
I
co
D)
U-
-------
47
o
o
PEAK
i I
SECONDARY
REGION
CHANNEL NUMBER
XBL 733-305
Fig. 8. A typical K X-ray spectrum due to one element. The regions identified in this
figure must be defined during the CALIBRATION phase of set up for the instrument.
-------
48
XBL, 7110-1500
Fig. 9. Expression for the overall efficiency of the process of production, absorption
and detection of the fluorescent X-rays.
-------
49
100
o
UJ
O
LL.
u.
UJ
UJ
>
<
_l
UJ
cr
10
10 20 30 40
ATOMIC NUMBER
50
60
XBL 733-296
Fig. 10. Actual relative efficiency curves for three fluorescers. These can be com-
pared with the theoretical curves of Figure 1.
-------
50
I I I I I I I I l I I I I I I I I I I I I I I I I I I I
Cu EXCITATION
i i i I i i i i i i i i i I i i i i i
XBL 733-304
Fig. 11. Air filters spectrum taken using the copper fluorescer.
-------
51
1 I I I I I I I I I I I I I I I
Mo EXCITATION
I
XBL 733-303
Fig. 12. Air filters spectrum taken using the molybdenum fluorescer.
-------
52
1000
E
60
100
10
20
30 40
ATOMIC NUMBER
50 60
XBL 733-306
Fig. 13. Elemental detection sensitivity curves for the three fluorescers
(two hour sample collection time, five minute analysis time on each fluorescer).
-------
53
pq
O
O
cz
LU
I-
LJJ
s
<
o
LU
_J
o
H o
q
o
CD
N
00
CO
Q.
O
c
0)
03
>^
CO
X
E
=3
CO
O
NOllVflNailV
-------
54
III I i
III M i I I T
m
x
o
o
O *
-» LU
I-
UJ
CD
N
CO
_CD
O
i-
CO
Q.
CO
CO
CD
c
o
1
I
t-
t
;
-
O
i_
<
LU
1-
h-
OC
X
o
1 1 1
-
LU I
M
«/) I
LU
I o
K
0£ _
< I
Q_ -
-
1 III 1 1 1 1 1 1 III 1 1 1 1 1 1
-| i-l i-l
d 9
o
^ CO
Q |
QJ
"^ 2
"* x
< i
Q_ GO
LT)
LL
t-H
O
NOIIVONBIIV
-------
55
411 I I I I | Ml II I I
ffl
X
o
o
o UJ
^ H
UJ
o
z o
q
o
CD
N
CO
O
w
CO
o
CO
>,
CD
L^
i
X
£
D
c
"i
L.
NOllVnN31iV
-------
-------
A-l
APPENDIX
EXTRACTS FROM FIRST PROGRESS REPORT
MONITORING SYSTEM DESIGN
As the design of hardware must conform to the whole monitoring system
parameters, it is essential at the outset to make decisions on the overall
system. The following factors enter into these decisions:
A. Filter Specification and Logistics
As shown in our earlier work on X-ray fluorescence analysis, the
principal background observed in X-ray fluorescence spectra is produced by
scatter of the primary radiation from the matrix of the sample into the
detector. It is therefore essential to use low-mass filters containing
only very light elements. Furthermore, the filter should be virtually
free of elemental contaminants in the range of elements to be investigated
(i.e. approximately from sulphur to barium in the periodic table, and also
the heavy elements).
The ideal X-ray spectrometer geometry examines an area about 1 inch
in diameter on a filter. Therefore there is no reason to use filters much
larger than this. To avoid the possibility of wall effects in the sampling
pipe producing a serious non-uniformity over the area examined, we have
chosen to use filters 37 mm in diameter. Such filters can be obtained from
Millipore, Gelman and Nucleopore at prices near $200/1000. Results obtained
in our laboratory and by other workers indicate that these filters satisfy
the general requirements for X-ray analysis. The availability of several
types of filter in this size will facilitate intercomparisons between their
properties at a later date.
-------
A. 2
The handling of large numbers of filters can only conveniently be
accomplished by mounting filters on a suitable carrier, which must cost
significantly less than the filter itself. Furthermore, the carrier should
be of such a size as to utilize convenient means of storage. Plastic mould-
ings offer the prospect of meeting these requirements at minimum cost. The
design chosen for the carrier is shown in Figure A.I; its overall dimensions
are 2" square (suitable for storage in common 35 mm slide-boxes), with pro-
vision for mounting filters into the 37 mm diameter recess. Keying is pro-
vided for orientation of the filter at one corner and a recessed shelf at
one edge provides a suitable area for identification markings (either
manual or automatic) . The design allows for automated manipulation of the
carriers both at the air sampler, and at the analysis facility.
Preliminary estimates for the moulding costs have been obtained; it
appears likely that the carriers will cost about 5<£ each in 10,000 quan-
tities after the first order. Initial tooling will cost approximately
$2000--this cost will be borne by the first order for 10,000 units.
B. Scope of System: Number of Filters and Elements for Analysis
The complexity of the data-processing requirements, and the need for
automatic sample handling, is determined by the range of elements to be
studied, and by the size of the whole monitoring system. After discussions
with EPA representatives, and with other authorities, the following targets
have been set:
i) A typical monitoring complex might consist of ten sampling
stations each producing an average of three filters per day.
Unattended, automatic operation of these stations for a week or
-------
A. 3
more appears desirable. This implies that the design of the
sample changer should be suitable for holding at least 21 filters--
30 has been chosen as the design objective. If used for daily
filter changes, this will permit one month of unattended operation.
One of our objectives will also be to use components (e.g. vacuum
pump) requiring routine attention no more frequently than once a
month.
ii) An analysis station must be capable of handling the analysis
of over 30 filters per day produced by the ten sampling stations.
Our design objective is 50 filters per day. By allowing 30 minutes
analysis time per filter, and by providing automatic operation on
a 24 hour per day basis, we will be able to achieve this objective.
Our experience has shown that coverage of the range of elements
of interest requires X-ray analysis with three different excita-
tion energies. Therefore the 30 minute analysis time is broken
down into three separate time intervals each of ten minutes dura-
tion. The excitation X-rays for these intervals will probably be
the K radiation of Ni, Mo, and possibly Gd. Using these exciting
radiations, the interesting elements shown in Table A.I are
potentially amenable to study. In practice, various effects will
limit the sensitivity for some of these elements, but, as one pur-
pose of the contract is to evaluate these limitations, it is essen-
tial to provide coverage for the entire range.
-------
A.4
Table A.I: Elements Potentially Amenable to Study
Element La Lg Energies
W 8.40 9.67
Hg 9.95 11.85
Pb 10.52 12.61
Bi 10.80 13.00
Element
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mi
Fe
Co
Ni
Cu
Zn
Ga
As
Se
Br
Rb
Sr
Zr
Mb
Mo
Cd
In
Sn
Sb
I
Ba
Ka Energy
1.49
1.74
2.02
2.31
2.62
3.31
3.69
4.51
4.95
5.41
5.90
6.40
6.93
7.47
8.04
8.63
9.24
10.53
11.21
11.91
13.38
14.14
15.75
16.58
17.44
23.11
24.21
25.19
26.36
28.61
32.07
-------
A.5
ANALYSIS STATION
Overall Spectrometer Design
The various components of the spectrometer are shown in the block dia-
gram (Figure A.2). Overall control of the system rests "w'itfr the T1960A com-
puter. Under program control, this instructs the X-ray spectrometer to
change samples, to change fluorescer, to turn on helium flow over the X-ray
detector for low-energy measurements, and to start an analysis of a sample.
The main sequencer in the spectrometer acts as the interface for these instruc-
tions. Control of the X-ray tube anode voltage is also exercised by this unit
to provide the optimum anode voltage for excitation of the fluorescer, which,
in turn, excites the sample.
The computer also acts as a pulse-height analyzer (together with the
1024-channel Northern Scientific ADC), sorting X-ray pulses derived from the
detector-preamplifier-amplifier combination. The counting-time is controlled
by a preset timer (operating on live-time and thereby correcting for dead
time losses in the system). At the end of the preset counting interval, the
main sequencer flags the computer, initiating computer analysis of the accum-
ulated spectrum. The results of the elemental analysis are printed out
immediately together with estimated errors. The computer also initiates the
required actions in the spectrometer preparatory to the next counting interval.
A monitor is provided on the liquid nitrogen level in its storage dewar;
if the level falls below the five day reserve level (normal capacity ~40 days),
a flashing indicator shows this, and computer printouts will warn the opera-
tor until the condition is corrected.
Filament power and high-voltage for the X-ray tube are supplied by a
high-voltage power supply which provides accurate and stable voltages under
servo control.
-------
A.6
Peripheral equipment, to be supplied with the computer, includes a
display, a teletype (including tape-reader) that prints all analysis results,
a magnetic tape used for program storage and data output, and a "program
switch panel". The design of this panel will permit the operator to select
and initiate several standard routines without requiring any familiarity
with the computer system.
Early work indicates that a detection limit better than 10 ng/cm2 can
be achieved for most elements. For an MF MLllipore 0.8 ypore size filter,
flow rate at 70 cm Hg pressure equals 10 1/min/cm2, in two hours, total
volume equals 1.2 m3/cm2 of filter. Therefore, detection limit - 10 ng/m3
of air. This quantity is inversely proportional to the sample collection
1/2
time, and is inversely proportional to (analysis time) . While the 10
1/min/cm2 is probably a somewhat higher airflow than we will use, it is a
convenient figure to use for this calculation.
It is of considerable interest to ascertain the elements likely to be
observed in air filters with these detection sensitivities. Figure A. 3 pre-
sents data on this question. The anticipated levels of elements indicated
in this figure represent best guesses based on very sparse data in the
literature. Those elements vhere special problems occur with X-ray analysis
are so indicated. From this figure, we can guess that a two hour sample
will reveal between 15 and 25 elements; an eight hour sample, 20 to 25; and
a 24 hour sample, over 25 elements. As the levels indicated for most of
the elements in this figure are their normal range, large increases above
normal will be measureable for virtually any of the 35 elements shown here.
-------
A.7
H <
33
31
a
8
o
ui
"~ U J
a xo
3*i
Q»/>»-
i hio
o>
tr
«r>
8
O
1 .
t
/2o
\ i
, t r* i / A
fe
4
1
OJ
K-
(0
5
3
o>
CO
o
o>
c
O)
CO
CD
D
O)
u.
-------
a>
0)
o
o
0)
CL
CO
O)
ro
o
_o
m
Cvj
a:
ft
IU
r
o
at
t
-------
A.9
o
CO
*2 1
II 1 1 1 1 1
If
K
i
CO
UJ
l* ,
UJ
Q3tJin03U 3c
*>XH WOd NOI1
1
5 O O
1 1 I ^ i i i 1 1 1 i i i i 1 i i
'
\
, !
»"^^^x
\
\
N
\
i _
..t. ;
\ ,
s *
\ 1*1
\
, \
T \
\ 1*
C N
I *
' $,_
1 *|
<
I
snou
~~l ,
VDS3 !S r
.03HMOO \ ,
1 f
I3S 30N3H3JM31$I 0X
\
' , t
' >
>
I
* t <
i '
I_^^_
_ 1 '
1
1 ' SH
(£uj/0u)
3A31 Q3Ad3S80
N
\
\
\
\
\
N
N
1 '
^^ ^M ^J
^9 ^^ ii i
iii ^^ ^^ S%OL
Sxco^uS
£ *£" <£ £
13
CO
H
o
.c
0
CO
CD , o
£ *"
* cf
o
£ «
ffl CO
H OT
_ S! co
W co
o. ° E
< 0 E
k.
x»w ^
o: >
CD t±
0 CO
N 1
3 -^
Z w
o
o 1
iS
C CO
« 1^
- *" c i
H* -1
o W ^^
o Si
_ C -t-^
CD
w *" ^~
o "S c
.-^ -^ O
n E tt
*- 0)
CO
W UJ o
< °
9 n a)
2 §-
LZc^
-------
-------
3. Recipient's Accession No.
BIBLIOGRAPHIC DATA
SHEET
1. Report No-
EPA-R2-73-182
4. Title and Subtitle
X-Ray Fluorescence Spectrometer for Airborne
Particulate Monitoring
5- Report Date
April 1973
6.
7. Author(s)
F. S. Goulding and J. M. Jaklevic
8- Performing Organization Kept.
No.
9. Performing Organization Name and Address
Lawrence Berkeley Laboratory
Berkeley, California 94720
10. Pro)ect/Task/Work Unit No.
11. Contract/Grant No.
EPA-IAG-0089(D) /A
12. Sponsoring Organization Name and Address
13. Type of Report & Period
Covered
Final Jan . 19 7 2- Jan . 7J3
14.
15. Supplementary Notes
16. Abstracts An automated system for determining the elemental composition
of ambient air has been developed. Airborne particulate matter is
collected on up to 36 membrane filters which are sequentially introduced
into the sample stream. The duration for the collection period can be
adjusted to range from 1 to 24 hours. In the analysis station the
elemental composition of the collected particulate matter is determined
using an energy dispersive X-ray spectrometer. Each filter is
separately analyzed using each of three secondary fluorescers in order
to optimize the sensitivity for a wide range of elements which have
atomic numbers greater than 12. After a brief analysis period the
results for each filter are printed out on a teletypewriter and
written on magnetic tape. Up to 36 filters can be accommodated with-
out operator attention.
17. Ke> Words and Document Analysis. 17a. Descriptors
Airborne particulate monitor
X-ray fluorescence
X-ray tubes
Trace element analysis
Air Sampling
17b. Identiflers/Open-Ended Terms
17c. COSATI Field/Group
18. Availability Statement Approved for public
release, distribution unlimited
19. Security Class (This
Report)
UNCLASyiFiKD
20. Security Class (This
Page
UNCLASS1F1HD
21- No. of Pages
70
22. Price
FORM NTIS-35 IREV. 3-72)
U5COMM-DC 14952-P72
-------
INSTRUCTIONS FOR COMPLETING FORM NTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for the Federal Government,
PB-180 600).
1. Report Number. Each individually bound report shall carry a unique alphanumeric designation selected by the performing
organization or provided by the sponsoring organization. Use uppercase letters and Arabic numerals only. Examples
FASEB-NS-67 and FAA-RD-68-09-
2. Leave blank.
3. Recipient's Accession Number. Reserved for use by each report recipient.
4. Title and Subtitle. Title should indicate clearly and briefly the subject coverage of the report, and be displayed promi-
nently. Set subtitle, if used, in smaller type or otherwise subordinate it to main title. When a report is prepared in more
than one volume, repeat the primary title, add volume number and include subtitle for the specific volume.
5- Report Date. Fach report shall carry a date indicating at least month and year. Indicate the basis on which it was selected
(e.g., date of issue, date of approval, date of preparation.
6- Performing Organization Code. Leave blank.
7. Author(s). Give name(s) in conventional order (e.g., John R. Doe, or J.Robert Doe). List author's affiliation if it differs
from the performing organization.
8. Performing Organization Report Number. Insert if performing organization wishes to assign this number.
9. Performing Organization Name and Address. Give name, street, c ity, state, and zip code. List no more than two levels of
an organizational hierarchy. Display the name of the organization exactly as it should appear in Government indexes such
as USGRDR-I.
10. Project/Tosk/Worle Unit Number. Use the project, task and work unit numbers under which the report was prepared.
11. Con tract/ Grant Number. Insert contract or grant number under which report was prepared.
12. Sponsoring Agency Name and Address. Include zip code.
13. Type of Report and Period Covered. Indicate interim, final, etc., and, if applicable, dates covered.
14* Sponsoring Agency Code. Leave blank.
15. Supplementary Notes. Enter information not included elsewhere but useful, such as: Prepared in cooperation with . . .
Translation of ... Presented at conference of ... To be published in ... Supersedes . . . Supplements . . .
16. Abstract* Include a brief (200 words or less) factual summary of the most significant information contained in the report.
If the report contains a significant bibliography or literature survey, mention it here.
17. Key Words and Document Analysis, (a). Descriptors. Select from the Thesaurus of Engineering and Scientific Terms the
proper authorized terms that identify the major concept of the research and are sufficiently specific and precise to be used
as index entries for cataloging.
(b). Identifiers and Open-Ended Terms. Use identifiers for project names, code names, equipment designators, etc. Use
open-ended terms written in descriptor form for those subjects for which no descriptor exists.
(c). COSATI Field/Group. Field and Group assignments are to be taken from the 1965 COSATI Subject Category List.
Since the majority of documents are multid isc iplinary in nature, the primary Field/Group assignments ) will be the specific
discipline, area of human endeavor, or type of physical object. The application(s) will be cross-referenced with secondary
Field/Group assignments that will follow the primary postmg(s).
18. Distribution Statement. Denote releasability to the public or limitation for reasons other than security for example "Re-
lease unlimited". Cue any availability to the public, with address and price.
19 & 20. Security Classification. Do not submit classified reports to the National Technical
21. Number of Pages. Insert the total number of pages, including this one and unnumbered pages, but excluding distribution
list, if any,
22. Price. Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
FORM NT1S-35 (REV. 3-721 USCOMM-DC
------- |