Tennessee
Valley
Authority
Division of Environmental
Planning
Chattanooga TN 37401
TVA/EP79/06
United States
Environmental Protection
Agency
Office of Energy. Minerals, and
Industry
Washington DC 20460
EPA 600 7 79 054
March 1979
Research and Development
Application of
Germanium
Detectors to
Environmental
Monitoring
Interagency
Energy/Environment
R&D Program
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 INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from the
effort funded under the 17-agency Federal Energy/Environment Research and
Development Program. These studies relate to EPA's mission to protect the public
health and welfare from adverse effects of pollutants associated with energy sys-
tems. The goal of the Program is to assure the rapid development of domestic
energy supplies in an environmentally-compatible manner by providing the nec-
essary environmental data and control technology. Investigations include analy-
ses of the transport of energy-related pollutants and their health and ecological
effects; assessments of, and development of. control technologies for energy
systems; and integrated assessments of a wide range of energy-related environ-
mental issues.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
TVA/EP-79/06
EPA-600/7-79-054
March 1979
APPLICATION OF GERMANIUM DETECTORS TO ENVIRONMENTAL MONITORING
by
Dale W. Nix, Robert P. Powers, Larry G. Kanipe
Division of Environmental Planning
Tennessee Valley Authority
Muscle Shoals, Alabama 35660
Interagency Agreement No. D7-E721
Project No. E-AP 79BDI
Program Element No. INE 625C
Project Officer
Gregory J. D'Alessio
Energy Coordination Staff
Office of Energy, Minerals, and Industry
Washington, DC 20460
Prepared for
OFFICE OF ENERGY, MINERALS, AND INDUSTRY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, DC 20460
-------
DISCLAIMER
This report was prepared by the Tennessee Valley Authority and has
been reviewed by the Office of Energy, Minerals, and Industry, U.S.
Environmental Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the views and
policies of the Tennessee Valley Authority or the U.S. Environmental
Protection Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
-------
ABSTRACT
Gamma-ray spectroscopy is one of the most economical and wide-ranging
tools for monitoring the environment for radiological impact. This report
examines the problems involved in applying germanium detectors to the
analysis of environmental samples. All aspects of germanium spectroscopy—
equipment, system installation, quality control, energy and efficiency
calibration, spectral analysis, analytical sensitivities, and cost
considerations—are surveyed.
Germanium detectors can be used to achieve analytical sensitivities
of less than 10 pCi/L (for water) for most radionuclides, often at a
confidence level of 95 percent. Germanium detectors should be used to
analyze environmental samples that may contain a complex mixture of
radionuclides or unknown components because the resolution offered by
germanium detectors is unexcelled in these applications. However, use
of germanium detectors may not always be as economical as use of sodium
iodide [NaI(T£)J detecors.
This report was submitted by the Tennessee Valley Authority, Division
of Environmental Planning, in partial fulfillment of Energy Accomplishment
Plan 79 BDI under terms of Interagency Agreement EPA-IAG-D7-E721 with the
Environmental Protection Agency. Work was completed as of November 1978.
iii
-------
CONTENTS
Abstract iii
List of Figures vii
List of Tables vii
1. Introduction 1
2. Conclusions 2
3. Recommendations 3
4. Gamma-Ray Spectroscopy Using Germanium Detectors 4
4.1 Terminology 4
4.2 Equipment and initial setup of the system 6
4.2.1 Equipment 6
4.2.2 Initial set-up 12
4.3 Quality control 12
4.4 Calibration 13
4.4.1 Energy calibration 13
4.4.2 Efficiency calibration 16
5. Data Analysis 37
5.1 Spectral analysis 37
5.1.1 Spectral smoothing 37
5.1.2 Peak location 43
5.1.3 Peak intensity and peak centroid 48
5.1.4 Automatic analysis 60
5.2 Lower limit of detection 65
6. Comparison of Germanium and NaI(T£) Systems for
Gamma Spectroscopy 75
6.1 System sensitivities ..... 75
6.2 System costs 80
6.3 System applicability 82
References 86
Appendix A. Germanium Detector Leakage Current Test 99
Appendix B. Procedures for quality control of germanium
detector systems
-------
LIST OF FIGURES
Number page
I System check-out form 14
2 Energy calibration 18
3 Nonlinearity of a typical system 19
4 Typical gamma-ray spectrum obtained with a
germanium detector 28
5 Total detection efficiency 30
6 Typical absolute FEP efficiency curve 33
7 Deviation of log-log fit from experimental efficiency ... 35
8 Effect of spectral smoothing and differentiation on
a typical FEP 39
LIST OF TABLES
Number Page
1 Nominal Detector and Shield Specifications 10
2 Energy Calibration Data 17
3 Comparative LID Values for Various Marinelli Geometries . . 21
4 Standardization Sources 25
5 Correction Factors for Coincidence Summing 31
6 Analytical Functions Used to Fit Efficiency Data for
Germanium Detectors 34
7 Effect of Spectral Smoothing and Differentiation
on a Typical FEP 40
8 Fitting Functions 55
9 LLD Values for 3.5-L Geometry Used for Water 68
10 LLD Values for 0.5-L Geometry Used for Water 69
11 LLD Values for Geometry Used for Vegetation 70
12 LLD Values for Geometry Used for Soil 71
13 LLD Values for Geometry Used for Air Filters 72
14 Preparation of Background Standards for Determination of
LLD 73
15 Characteristics of Germanium and NaI(T£) Systems
for Gamma-Ray Spectroscopy .... 76
16 LLD Values for 3.5-L Geometry for Samples Containing
a Single Radionuclide 77
17 LLD Values for Samples Containing More Than One
Radionuclide 79
18 Cross-Check Results for Gamma Emitters in Water 83
19 Advantages and Disadvantages of Germanium and NaI(T£)
Spectrometry Systems for Gamma-Ray Spectroscopy 84
VI1
-------
SECTION 1
INTRODUCTION
The emphasis on monitoring the environment for radiological impact
is growing with the increased use of nuclear power and the still prevalent
concern over the effects of low-level radiation. One of the most economi-
cal and wide-ranging analytical tools for environmental monitoring is
gamma-ray spectroscopy. However, obtaining accurate and reliable quanti-
tative information from gamma-ray analysis requires a sophisticated and
knowledgeable approach. This report examines the problems involved in
applying germanium detectors to the analysis of environmental samples.
All aspects of germanium spectroscopy are surveyed, with emphasis on equip-
ment, installation of a system, quality control, energy and efficiency
calibration, spectral analysis, analytical sensitivities, and cost consi-
derations. We have tried to identify all significant literature in these
areas, to evaluate different approaches, and to provide additional infor-
mation on points often absent from the literature. Because the audience
of the report is expected to have a good knowledge of gamma-ray spectros-
copy and radiochemistry, no section has been included on basic aspects of
these topics; however, references 1 through 11 can be consulted for
assistance.
-------
-2-
SECTION 2
CONCLUSIONS
Germanium detectors can be used to achieve analytical sensitivities
of less than 10 pCi/L (for water) for most radionuclides. Depending on
counting time and sample size, these sensitivities can be reached at a
confidence level of 95 percent. However, a rough cost comparison with
sodium iodide [NaI(T£)J detectors indicates that the use of germanium
detectors may not always be the more economical method of performing
gamma-ray spectroscopy when factors such as initial capital outlay and
volume of work are considered.
To be effective, germanium-based gamma-ray spectroscopy must use a
method for spectral analysis that is accurate and reproducible, even
with small changes in experimental conditions. Also, the corrections
that affect each individual line of the calibration spectrum must be
considered when efficiency calibrations are performed.
-------
-3-
SECTION 3
RECOMMENDATIONS
Germanium detectors should be used to analyze environmental samples
that may contain a complex mixture of radionuclides or unknown components
because the resolution offered by germanium detectors is unexcelled in
these applications. The analyst must be aware of pitfalls that may
exist if undocumented software is used to analyze spectral data from
germanium detectors. Programs such as GAMANAL, SAMPO, and HYPERMET are
suggested as reliable tools, but newer techniques are available which
may account more satisfactorily for the frequently encountered problem
of poor counting statistics (see Section 5).
More effort should be given to establishing criteria for the evalua-
tion of results produced by computer-analyzed spectra. Information on
both fit and analysis of residuals is not commonly available to the
analyst.
-------
-4-
SECTION 4
GAMMA-RAY SPECTROSCOPY USING GERMANIUM DETECTORS
4.1 TERMINOLOGY
full energy peak (FEP): That region of a gamma-ray spectrum that is due
to the complete energy transfer of a photon within the detector and
the complete collection of the transferred energy within the detec-
tor. Full energy peak is also called total absorption peak or
photopeak.
absolute full-energy peak efficiency: The fraction of emitted gamma-rays
that are detected and recorded in the FEP of the pulse height
distribution.12
absolute gamma-ray emission rate: The number of gamma rays of specific
energy emitted per unit time from a radioactive source.
analog-to-digital converter (ADC): An electronic device used to convert
a pulse amplitude into a digital form. This conversion is generally
accomplished by converting a pulse of given amplitude into a number
of equal time intervals by means of a clock pulse generator.8
blank sample: A sample that contains either no radioactive material or
only low levels of radioactive material; such a sample is generally
used to determine the background count rate for a radiation detector.
branching fraction: The probability of emission of a specific gamma ray
per nuclear decay.
correlated photon summing: The simultaneous detection of two or more
photons originating from a single nuclear disintegration.13
counts per second: A measure of the count rate for a radiation detector
system for a given sample.
convolution: A mathematical transformation of data into a desired form.
For gamma-ray spectral data this is accomplished by replacing the
original data with data derived by multiplying the contents of
selected channels by a series of convolution integers, integrating,
and dividing the sum by a normalizing factor. The convolution and
normalizing factors chosen determine the type of data transformation
performed.14
direct current (DC) level: The reference input or output voltage for
unipolar pulses on a DC coupled instrument. For germanium spectro-
scopy systems, the DC level of an output signal from a linear
amplifier must be matched to the input requirements for DC level of
an ADC for proper processing of pulse signals.
-------
-5-
full width at half maximum (FWHM): The full width of a distribution
measured at half the maximum ordinate. For a normal distribution,
FWHM is equal to 2(2 £n 2)* times the standard deviation a.15
full width at tenth maximum (FWTM): Same as FWHM, except that measurement
is made at one-tenth, rather than one-half, the maximum ordinate.15
germanium detector: A semiconductor device that uses the production and
motion of excess free charge carriers produced by absorption of
radiation to detect and measure incident X and gamma radiation.16
Germanium detectors are of two general varieties: high-purity and
lithium-drifted. A high-purity germanium (HPGe) detector is fabri-
cated from hyperpure germanium crystals and contains an equal number
of free acceptor sites and free charge carriers throughout its volume.
Lithium-drifted germanium [Ge(Li)] detectors are fabricated by compen-
sating p-type germanium with lithium ions, which are moved through a
germanium crystal under an applied electric field in such a way as to
compensate the charge of impurities bound to acceptor sites.16 All
germanium detectors are operated at liquid nitrogen temperatures,
although HPGe detectors may be allowed to cycle to room temperature
when no bias is applied. The response of a germanium detector is
linearly proportional to incident photon energy.
germanium spectroscopy system: A gamma-ray spectroscopy system composed
of an HPGe or Ge(Li) detector, an amplifier, an ADC, a multichannel
analyzer (MCA), a high-voltage power supply, and a lead or steel
shield.
lower limit of detection (LID): The smallest amount of sample activity,
when a given measurement process (i.e., detector and chemical
procedure) is used, that will yield a net count for which there is
confidence at a predetermined level that activity is present.17
multichannel analyzer (MCA): A computer-like device used to store and
retrieve digital information on pulse amplitude and frequency
distribution. Generally, an array of magnetic or solid-state
memories are used for data storage and manipulation.
peak-to-Compton ratio: Ratio of the maximum number of counts in a
photopeak to the average height of the Compton continuum. The
energy of the FEP and the radionuclide used must be stated.15
pole-zero cancellation: Compensation of pulse shape for the under-
shooting of the DC level (baseline) of the amplifier due to pre-
amplifier fall time constant. Pole-zero cancellation is generally
adjustable on amplifiers used for germanium spectroscopy and is set
to minimize undershoot or overshoot of the pulse baseline.
random summing: The simultaneous detection of two or more photons
originating from the disintegration of more than one atom.
-------
-6-
relative efficiency (in percent): The absolute FEP efficiency for the
1332.5-keV gamma ray emitted from a 60Co source located 25 cm from
the face of the germanium detector, relative to the corresponding
efficiency of a 76- by 76-mm Nal(TA) detector.
sensitivity: The minimum difference between two sample activities that
can be statistically distinguished from each other at a determined
confidence level.
sodium iodide [Nal(TJK)] detector: A radiation detector using a crystal
of thallium-activated sodium iodide [NaI(T£)], which scintillates
when it interacts with X and gamma radiations. A generally linear
relationship exists between light output of an NaI(T£) crystal and
incident gamma-ray energy. By using sensitive photocathode and
photomutiplier units in conjunction with an NaI(T£) crystal, an X-
and gamma-radiation detector may be constructed. Typical spectral
resolution for 137Cs (661.6 keV) using a 10- by 10-cm NaI(T£)
spectrometry system is about 55 keV.
NaI(T£) spectroscopy system: A gamma-ray spectroscopy system composed
of an NaI(T£) detector, an amplifier, an ADC, an MCA, a high-voltage
power supply, and a lead or steel shield.
Type I error: The error committed in stating that the true activity is
greater than zero when, in fact, it is zero.18
Type II error: The error committed in stating that the true activity is
zero when, in fact, it is greater than zero.18
4.2 EQUIPMENT AND INITIAL SETUP OF THE SYSTEM
4.2.1 Equipment
Most basic gamma-ray spectroscopy systems using germanium detectors
are composed of a germanium detector (with its cryostat, dewar, and pre-
amplifier), an amplifier, a high-voltage power supply, a steel or lead
shield, and an MCA with an ADC and data readout device. Other electronics
can be added to the system, such as a liquid nitrogen monitor, a gain
stabilizer, or a pulse pile-up rejector with live-time corrector. This
report does not describe Compton suppression systems or any other type
of coincidence system.
Liquid nitrogen monitors activate visual or audible alarms when the
liquid nitrogen drops below a predetermined level. Because loss of
liquid nitrogen may damage the detector and preamplifier [actually the
field-effect transistor (FIT)], a liquid nitrogen monitor may be valuable
in certain laboratories. However, if a large-volume (30-L or larger)
dewar is used and filled weekly, a liquid nitrogen monitor would not be
needed.
Gain stabilization equipment should not be needed for most germanium
systems because modern, high-performance electronic systems are generally
stable (with a typical gain drift of only 0.03 percent per day). However,
-------
-7-
if a system is used to count samples for several days or weeks, gain
stabilization may be needed to compensate for long-term gain variation.
A pulse pile-up rejector with a live-time corrector is used for
samples with high count rates (e.g., greater than 5000 total counts per
second). Typical environmental samples have low count rates, and pulse
pile-up rejectors are not needed.
4.2.1.1 Germanium Detector
One principal component of the spectroscopy system is the germanium
detector. Two basic types of detectors exist: the lithium-drifted
germanium [Ge(Li)j detector and the high-purity germanium (HPGe) detector
(often called an intrinsic germanium detector). Both detectors can be
obtained commercially with high resolution and high efficiency. The HPGe
has the advantage of being able to be warmed to room temperature, after
removal of the detector bias, without damage to the detector.* The
Ge(Li) detector would be seriously damaged if it were allowed to warm to
room temperature under any circumstance. The HPGe detectors are now
considerably more expensive than Ge(Li) detectors; however, they may
begin to drop in price as they come into more common use.
High-efficiency (>15%) HPGe and Ge(Li) detectors can be fabricated
in a true coaxial geometry or in a closed-end coaxial geometry. The
closed-end coaxial detector is often preferred for counting environ-
mental samples because it presents a more uniform, efficient cross
section of active detector volume to samples that are counted a short
distance from the detector.
Placement of the detector close to the cryostat endcap window is
desirable because it is often necessary to place samples as close to the
detector as possible. Typically, one can specify that the detector be
placed within 5 mm of the cryostat endcap window and that the detector
be mounted with azimuthal symmetry around the axis of the crystal. The
symmetry is important when using a geometry such as a Marinelli beaker,
which is one of the predominant environmental geometries now in use.
To ensure that a detector will have a good efficiency for samples
counted at close geometries, the ratio of the detector diameter to the
detector length should be as large as practical, typically greater than
one.
The efficiency, resolution, and peak-to-Compton ratio basically
determine the performance and cost of germanium detectors. Efficiency is
a measure of the ability of the detector to detect a given level of activ-
ity. Resolution is a measure of the ability of a detector to differentiate
between closely spaced lines. The larger the peak-to-Compton ratio, the
better the detector can distinguish a low-energy FEP in the Compton ratio
*The detector and preamplifier will be damaged if the HPGe detector is
allowed to warm up with the high voltage applied.
-------
-8-
are generally specified in accordance with the Institute of Electrical
and Electronics Engineers (IEEE) standard test procedures for germanium
gamma-ray detectors.15
The relative efficiency in percent of the detector usually is speci-
fied for the 1332.5-keV gamma ray emitted from 6OCo at a source-to-detector
distance of 25 cm relative to the corresponding efficiency of a 76- by
76-mm Nal(TJfc) detector. The absolute FEP efficiency19 for a 76- by 76-mm
NaI(T£) detector is 1.2 x 10~3. Germanium detectors with 10 to 15 percent
efficiencies are typically used for environmental analyses.
The energy resolution of a detector is expressed in terms of the
full peak width in energy at one-haIf or one-tenth the maximum peak
height (FWHM and FWTM). The FWHM is usually specified for the 122.1-keV
gamma ray of 57Co and the 1332.5-keV gamma ray of 60Co. The FWTM is
usually specified for the 1332.5-keV gamma ray. Typical ranges of
values for the FWHM are 0.7 to 1.1 keV at 122.1 keV and 1.7 to 2.1 keV
at 1332.5 keV. The FWTM at 1332.5 keV is typically a little less than
twice the FWHM at the same energy. All resolution measurements are made
after the background has been subtracted. The portion of the peak above
the half-maximum should contain at least four channels, preferably more,
so that the resolution can be measured accurately. Interpolation between
channels is necessary to find the exact location of the half-maximum on
each side of the peak centroid. The fewer channels in the peak, the
less accurate the interpolation and thus the less accurate the value of
the FWHM determined.
The peak-to-Compton ratio at 1332.5 keV is the ratio of the full
peak height to the average height of the corresponding Compton plateau
in the region below the edge (1040 to 1096 keV).15 The peak-to-Compton
ratio increases as the efficiency and resolution increase. Typical
state-of-the-art peak-to-Compton ratios can be obtained from the vendors
on request (usually from a list of specifications for detectors that
have been recently delivered).
The cryostat, preamplifier, and dewar are generally supplied with
the detector. Vertical cryostats with endcaps 20 to 25 cm long are
generally preferred for environmental applications because Marinelli
beakers, petri dishes, and other types of sources can be easily placed
on the endcap. Longer endcaps may be necessary to ensure that the
detector will fit into its shield.
Because the preamplifier and detector are generally integrated
assemblies, specifying a given resolution will ensure that the pre-
amplifier and detector are properly matched.
Many manufacturers offer a variety of dewars. A 30-L dewar is
generally sufficient because it will hold more than a 20-day supply of
liquid nitrogen. Even though the dewar has a long holding time, it
should be filled every week as a precaution against accidental warm-up
of the detector. Care should be taken to prevent blockage of the
filling tubes with ice, which might result in the dewar appearing full
when it is not.
-------
-9-
Table 1 lists some nominal specifications for detector and shield.
These detectors and shields were used to acquire data for parts of this
report.
4.2.1.2 Amplifier
For routine sample analysis a spectroscopy* linear amplifier is
generally used. Certain specifications are adequate for most needs and
can be met by most suppliers of nuclear instruments: (1) temperature
stability—gain <0.01 percent per degree Centigrade from 0 to 50°C and
DC level <±5 pV/°C from 0 to 50°C; (2) integral nonlinearity--<±0.05
percent of full rated range of the output; and (3) noise--<5 yV referred
to input with a gain of 3000 and a 2-^is time constant. The amplifier
should be equipped with a baseline restorer. The amplifier should be
supplied with a connector on the rear panel, which is properly wired to
supply the appropriate power to the detector preamplifier. Additional
specifications may be needed for particular applications. Consult IEEE
Standard 301-196920 for amplifier test procedures.
4.2.1.3 Bias Supply for Detector
The bias supply for the detector should have a voltage range to
accommodate the detector; typically, a range of either 0 to 3 or 0 to
5 kV is adequate for most detectors. The voltage should be continuously
adjustable over the full operating range with a front-panel meter for
monitoring the output voltage. A bias supply with a temperature sta-
bility of about 0.02 percent per degree centigrade and noise and ripple
of less than 10 mV peak-to-peak from 2 Hz to 5 MHz should be adequate
for most applications. One should ensure compatibility of the connector
on the supply with the cable and connector on the preamplifier.
4.2.1.4 Shield
Shields of low-activity steel or lead are generally used. Lead has
a much higher gamma absorption coefficient than steel in the photo-
electric and pair production range, but low-activity steel is easier to
locate than low-activity lead.21
Backscatter within the shield decreases as the atomic number of the
shielding material increases. Thus, lead shields have a much lower
backscatter component than do steel shields.22 The backscatter component
also decreases as the internal dimensions of the shield increase.
Typical internal dimensions are 41 by 41 by 41 cm to 61 by 61 by 61 cm.
Most lead shields have a 0.08-cm cadmium liner to degrade lead X rays and
*Several vendors designate amplifiers that are to be used with germanium
detectors as spectroscopy amplifiers. However, this is not a generally
accepted grade designation for an amplifier and should not be solely
used in specifying an amplifier.
-------
TABLE 1. NOMINAL DETECTOR AND SHIELD SPECIFICATIONS
Specification
Detector 1
Detector 2
Detector 3
Detector
Type
Resolution-FWHM, keV
122.1 keV
1332.5 keV
Efficiency
at 1332.5 keV, %
Diameter, mm
Length, nan
Diffusion depth, mm
Active volume, cm3
p-Core diameter, mm
True coaxial
1.41
2.23
8.7
42
36
21.7
48
8.5
Closed-end coaxial
1.02
1.99
14.3
50.0
44.5
15.5
70
-
Closed-end coaxial
0.731
1.99
16.0
47.5
41.0
17.75
80
_
Shield
Shield thickness, cm
Material
Inside dimensions
(L x W x H), cm
10
Lead with
graded absorber
36 x 36 x 81
15
Steel with
graded absorber
41 x 41 x 41
15
Steel with
graded absorber
41 x 41 x 41
Relative to the corresponding efficiency of a 76- by 76-mm NaI(T£) detector with
source placed 25 cm from the detector face.
-------
-11-
a 0.04-cm copper liner to degrade cadmium X rays.22 In addition to the
cadmium and copper liners, steel shields may have a 0.6- to 1.2-cm lead
liner to remove some of the low-energy background.
All shields should be designed to prevent any direct paths for
radiation leakage. Provisions should be made for easy insertion of the
germanium detector into the shield while providing adequate space for
the preamplifier and its cables. Shock-absorbing material should be
placed around the closure opening to prevent any mechanical shock to the
detector as the shield is opened or closed.
Typically, steel shields are 15 cm thick and lead shields are 10 cm
thick. The loading capacity of the floor of the counting room should be
considered when shields are purchased. Provisions for installing the
shields should also be made at the time of purchase.
4.2.1.5 Multichannel Analyzer System (MCA)
Although a full discussion of MCA systems is beyond the scope of
this report, limited information is given. The MCA system is composed
of one or more ADCs, the MCA, and assorted input/output (I/O) peripherals.
The ADC is generally purchased with the MCA. If both are not
purchased from the same vendor, care should be taken to ensure
compatibility of the MCA and ADC.
An ADC for germanium spectroscopy of environmental samples should
meet certain specifications: (1) conversion gain—at least 8192, 4096,
2048, 1024, and 512 channels; (2) internal clock frequency—>50 MHz,
crystal controlled; (3) integral nonlinearity—£0.05 percent over 99
percent of the full range; (4) differential linearity—<1.0 percent
deviation from the mean channel width over 99 percent of the full scale;
and (5) stability—time, better than ±0.01 percent per day at stable
ambient temperature and line voltage; and temperature, better than ±0.01
percent per degree Centigrade from 15 to 40°C.
The output pulse characteristics of the amplifier and the input
requirements of the ADC should be compatible. This is no problem for
most combinations of amplifiers and ADCs. The ADC should have a live-
time corrector and a dead-time meter.
Two basic types of MCAs are available: nonprogrammable MCAs
(NP-MCA) and fully programmable MCAs (P-MCA). An NP-MCA has been termed
a hard-wired MCA, but this classification does not necessarily apply
because many newer NP-MCA units are built around a microprocessor.
The decision of which type of analyzer and what options to purchase
is very complex. An NP-MCA is relatively easy to use for collecting
data, is less expensive, and is generally more reliable. A P-MCA is
capable of executing data reduction and other programs and is generally
more flexible. A P-MCA can accept multiple ADC inputs, whereas most
NP-MCA units are only able to accept one ADC input. When several
detector systems are used, a single P-MCA system can be used for all the
-------
-12-
detector systems, whereas several NP-MCA systems would be required.
Also, the P-MCA can be used to analyze spectra. A total analysis and
reduction system using a P-MCA with several detector systems to collect
and analyze spectra may be less expensive than one using multiple NP-MCA
systems.
The choice of I/O peripherals for a system depends on the require-
ments of the individual laboratory. Most laboratories will need some
hardware for storing spectra. Magnetic tape and disk are frequently
used. Also, some type of fast printer and a plotter are usually
desirable.
4.2.1.6 Summary
Each component of the system should be specified and chosen with
care. The current state-of-the-art can be determined by studying several
vendors' catalogs, consulting representatives of the vendors, and
discussing appropriate topics with experienced spectroscopists.
In general, the detector, amplifier, and bias supply should be
bought as a system from the same vendor. The MCA, ADC(s), and I/O
peripherals usually are bought as a system from the same vendor. Buying
a complete system from a single vendor may prevent noncompatibility of
individual components and may eliminate the problem of dealing with many
vendors for equipment maintenance. One should request three copies of
all instrument manuals—one to give to the instrument shop, one to keep
with the instrument, and one to file. Also, all connectors, cables, and
nuclear-instrument-module (NIM) bins should be included in the order as
needed.
4.2.2 Initial Setup
If the detector and MCA systems are purchased from one vendor, most
vendors will install and demonstrate the system. Although many labora-
tories have trained spectroscopists capable of installing the system, it
is a good policy to allow the vendor representatives to install the
system because they can confirm any damage or nonconformance to
specifications.
If a laboratory finds it necessary to set up the system, all manuals
supplied by the manufacturer should be thoroughly reviewed. Manufac-
turers' instructions for setting up and testing the system should be
followed. Any questions about setting up the equipment should be
directed to factory representatives.
4.3 QUALITY CONTROL
After the system is set up, one must ensure that the system always
performs reliably. Reliable operation can be assured by using quality
control procedures designed to detect any system malfunction at the
earliest possible moment. American National Standards Institute (ANSI)
-------
-13-
Guide N71713 suggests tests for (1) eaergy calibration, (2) count rate
reproducibility, (3) system resolution, and (4) efficiency calibration.
Also, Zeigler and Hunt9 suggest (1) background stability, (2) MCA time
base, and (3) MCA live time checks. Tests are also recommended for
(1) detector leakage current, (2) detector peak-to-Compton ratio,
(3) system linearity, (4) amplifier adjustment (pole-zero and DC level),
and (5) system noise.
The detector leakage current test is discussed in Appendix A. Typical
procedures for other quality control tests are given in Appendix B. These
procedures have been written to exclude references to specific equipment
and instrument settings.
When using these quality control procedures, one must specifically
identify (by model number and serial number) the equipment being tested
and record the instrument settings on a form (similar to that in Figure
1), which is attached to the logbook for the specific system. When
system components or instrument settings are changed, the date of the
change is recorded on the form, and a new form is filled out and entered
in the logbook.
Control charts are used to monitor statistically the reproduci-
bility of the detector system over a period of time. Both background
and efficiency are readily monitored by use of control charts.
References 23 through 25 discuss the construction of control charts, and
references 9, 17, and 26 discuss their application to nuclear counting
instruments.
In addition, to an instrument quality control program for ensuring
stable instrument performance, an analytical quality control program
should be developed to ensure that reproducible and accurate analytical
results are obtained. The analytical quality control program should use
replicate analyses, spiked samples, and collaborative sampling to monitor
precision and accuracy. Kanipe26 gives a detailed discussion of analyti-
cal quality control. Programs such as the EPA-Las Vegas cross-check
program provide an excellent source for regular interlaboratory
comparisons.
4.4 CALIBRATION
4.4.1 Energy Calibration
An accurate energy calibration for a germanium detector system is
required for assignment of correct energies to the centroid of each FEP
in an unknown spectrum. Procedures for identifying the radionuclides
within a spectrum rely on methods for matching the energies of the prin-
cipal gamma rays in the spectrum to the energies of gamma rays emitted
by radionuclides that are likely components of a particular sample.
This procedure can be accomplished by either manual inspection or
computer analysis. Several criteria should be examined before a final
identification is accepted (Section 5.1.4); these checks include deter-
mining that all intense gamma rays are present and that the intensity
ratios are correct.
-------
-14-
Procedure:
Usage Date Begin: End:
Detector:
Power Supply:
Voltage: Polarity:
Amplifier:
Coarse Gain:
Fine Gain: As specified in the system logbook daily entry
page
Input Polarity: Shaping Microseconds
Polarity: Range: Restorer:
ADC:
Conversion Gain:
Group (or group selection, if so equipped):
Upper Level Discriminator (or ULD):
Lower Level Discriminator (or LLD): ______
Analog Zero (or zero or fine zero):
Zero Suppression (if so equipped):
Other Switches:
Interconnections:
MCA:
Experiment: ADC:
Channels: Preset Time: Live
Groups:
I/O Assign:
Input: Subselect:
Output: Subselect:
Other:
Figure 1. System check-out form.
-------
-15-
The energy calibration of a germanium detector system is generally
given by a mathematical function, the analytical form of which will
depend on the linearity of the system. The energy calibration of an
essentially linear system can be described satisfactorily by a straight
line. The appropriate expression is determined from two calibration
points by
G = (E2 - E^/(C2 - Cx) (1)
and
F = EL - (GKCp , (2)
where
E, and E2 = energies of calibration lines 1 and 2,
C.. and C2 = channels corresponding to the centroid of the
FEPs of lines 1 and 2,
G = gain of the system, in energy units/channel,
F = offset of the system, in energy units.
The centroid channel, offset, and gain can then be used to calculate
the exact energy of an FEP by the expression,
E = F + (G)(C). (3)
The FEPs selected for calibration points should be separated in energy
by as much as is practical. ANSI Guide N71713 recommends that the
energy separation of the calibration points be at least 50 percent of
the spectral range of interest. A least-squares fit of several calibra-
tion points is required when highly accurate energy assignments are
needed or when the system is nonlinear.27'28 Peak positions and energies
are usually fit to a polynomial equation of the general form,
N . ,
E = I A.COJ"1 , (4)
where
E = energy of the calibration line,
C = channel corresponding to the centroid of the FEP,
A. terms = fit parameters.
Gleason28 proposed a variation of this equation to describe certain
types of nonlinearities often found in older ADCs:
E = Ax + (A2)(C) + A3/C2 . (5)
All parameters are defined as for equation (4) . In all the equations
described above, the zero-order term, in channel number, specifies the
-------
-16-
zero-channel intercept. The first-order term, in channel number, specifies
the system gain, and the additional terms describe the system nonlinearity.
Table 2 lists energy calibration data obtained by using a National
Bureau of Standards (NBS) mixed-radionuclide standard point source29 and
appropriate energy values taken from Heath.30 A plot of the energy
calibration data using the 88.036-keV l09Cd line and the 1332.483-keV
60Co line as calibration points is shown in Figure 2. Figure 3 is a
plot of the system nonlinearity, measured as the deviation of the actual
data from the linear, least-squares-fitted data. For routine environ-
mental analyses, linear energy calibrations often can be used when
systems exhibit only small nonlinearities, such as those shown in the
example.
Complex energy calibration techniques, such as least-squares fits
to second-, third-, or higher-order polynomials, may not be justified
when the sample spectra contain many peaks that are not statistically
well defined (i.e., that contain a small number of counts). The centroid
of peaks with poor statistics cannot be defined with sufficient accuracy
to derive accurate energy data from the peaks; therefore, no real benefit
is gained from the complex calibration procedure. An energy calibration
that is accurate to ±0.1 keV may not be necessary when peaks in the
sample spectra cannot be located to better than ±1 or ±2 keV.
4.4.2 Efficiency Calibration
An accurate calibration of efficiency is necessary to quantify
effectively any radionuclides present in a sample. Great care must be
exercised in calibrating the efficiency because the accuracy of all
quantitative results depends on this calibration. Before determining
the efficiency, one should select, properly connect, and properly adjust
the components (germanium detector, shield, amplifier, ADC, and MCA) of
the system (Section 4.2). Certain settings and adjustments of the
amplifier, for example, should be maintained until a new calibration is
undertaken. Changes in amplifier time constants and baseline restora-
tion settings may have slight, but direct, effects on counting efficiency.
Adjustments in pole-zero cancellation and DC level can change the peak
shape. Changes in peak shape affect the determination of the FEP count
rate and thus produce changes in the efficiency.
One should consider several points before undertaking the actual
determination of efficiency: (1) sample counting configuration, (2)
calibration methods, (3) calibration sources, and (4) analytical
efficiency expressions.
4.4.2.1 Sample Counting Configuration
For routine, reproducible sample analysis, counting containers must
be selected on the basis of both the volume of sample available and the
optimum counting configuration. In general, the optimum geometry would
give the highest sensitivity (i.e., the lowest limit of detection).
If the method used by the Health and Safety Laboratory (HASL) (Sec-
tion 5.2) were adopted, the LID would be given by
-------
-17-
TABLE 2. ENERGY CALIBRATION DATA
Energy3
(keV)
88.036
122.060
136.471
165.854
279.190
391.689
513.998
661.63
898.00
1173.226
1332.483
1836.075
Gain
Offset
Location of
peak centroid
(channel)
176.05
244.16
272.97
331.74
558.17
783.07
1027.45
1322.83
1795.59
2346.22
2664.99
3673.72
Errorb (keV)
Using two calibration
points
0.000
0.030
0.024
0.141
-0.098
-0.149
-0.270
-0.215
-0.209
-0.125
0.000
0.764
0.5000
0.0126
Least-squares
fit
0.151
0.172
0.163
0.157
0.007
-0.072
-0.222
-0.203
-0.255
-0.238
-0.151
0.490
0.4999
0.1845
3The energy values are taken from a compilation by Heath.3
bThe error is the difference between the calculated value and the measured
value.
-------
-18-
2000
1900 —
1800 —
1700 -
1600 —
1500 -
1400
1300 -
1200 —
1100 -
moo —
Figure 2. Energy calibration. The gain of this system was 0.5000
keV/channel, and the offset was 0.0126 keV.
-------
-------
-20-
LLD = 4.66 >/Sj/(RVEfDAVt)
where
LLD = lower limit of detection,
S, = background count rate in the region of interest,
R = chemical yield or recovery for all steps in the
procedure,
V = sample volume or mass,
E "= counting efficiency,
f = number of photons of a given energy per nuclear decay
(i.e., the branching fraction),
D = radioactive decay fraction,
A = unit conversion factor,
t = length of the counting time interval.
The background count rate, counting efficiency, and sample volume
or mass vary directly with the counting configuration selected. The
other parameters are selected for a given sample or held constant for a
particular analysis. In general, changes in sample configuration produce
only small changes in the background rate and therefore produce even
smaller changes in the sensitivity, which is proportional to the square
root of the background rate. As the sample volume or »acs increases,
counting efficiency decreases, because in a physically larger sample,
emitted gamma rays will subtend a smaller solid angle at the detector.
Therefore, the product of the counting efficiency and sample size.must
be optimized within reasonable limits to achieve optimum sensitivity.
For most practical applications, Marinelli beaker geometries provide
optimum sensitivity. Table 3 compares several different counting
configurations.
The TVA Radioanalytical Laboratory at Muscle Shoals counts soil,
vegetation, and some liquid samples in 0.5-L Marinelli beakers,*17 other
liquid samples in 3.5-L Marinelli beakers,** and air filters and certain
biological samples in 60- by 15-mm petri dishes. Other commonly used
sample containers are listed by Zeigler and Hunt.9
All sample containers are centered on the detector endcap.
Positioning of the sample containers must be reproducible because a
small variation in sample geometry can cause a large variation in count
rate when the source is close to the detector. Marinelli beakers may be
*0.5-L Marinelli beakers are available from Control Molding Co.,
84 Granite Avenue, Staten Island, NY, 10303.
**3.5-L Marinelli beakers are available from 6A-MA and Associates, Inc.,
P.O. Box 480091, Miami, FL 33148.
-------
-21-
TABLE 3. COMPARATIVE LLD VALUES3 (pCi/L)
FOR VARIOUS MARINELLI GEOMETRIES
Nuclide
14
-------
-22-
allowed to touch on the cryostat endcap (with the sample weight resting
on the shield) and centered by using a collar or spacer around the
endcap. Any positioning method can be tested by repeated removal and
replacement of the source, with each replacement being followed by a
measurement of the count rate for a sufficient time to minimize
statistical fluctuations. A consistent count rate will indicate that a
reproducible counting geometry has been achieved.
4.4.2.2 Calibration Methods
Three primary methods of efficiency calibration are in use:
(1) theoretical Monte-Carlo methods, (2) theoretical efficiency models
with experimentally determined parameters, and (3) experimental
determinations.
Theoretical calculations of efficiency detection by Monte-Carlo
methods31"35 have been limited to point or disc sources because calcula-
tions for more complex sample configurations are difficult, time-
consuming, and relatively inaccurate. Also, because detectors come in
many sizes and shapes, a calculation would be necessary for each geometry
contemplated for use with each detector. Theoretical calculations of
detection efficiency are further complicated by the different inter-
actions that can occur under identical circumstances. For example,
effects such as bremsstrahlung, annihilation, escape, backscatter,
multiple internal scattering, pulse pile-up, statistical fluctuations in
the pulse-height conversion, and inhomogeneity in the detector are
difficult or impossible to account for properly in Monte-Carlo
calculations.**
To use the Monte-Carlo technique effectively, a description of the
physical geometry of the germanium detector must be available. The
precise definition of the active volume of a germanium detector is
difficult to determine because the boundaries of the active area are
diffuse. Also, the precise location of the crystal within the cryostat
must be determined by radiographic or gamma-scanning methods because the
copper cold-finger will contract when cooled to the temperature of
liquid nitrogen, thereby changing the location from that observed at
room temperature. In addition to these difficulties, certain inherent
systematic errors depend on the Monte-Carlo model chosen and on the
approximations made for the energy loss of secondary electrons.36 One
should attempt to calculate and use theoretical detection efficiencies
only after making a thorough search of the literature and considering
the difficulties. Such theoretical calculations now appear too
uncertain for routine use and too time-consuming for most laboratories
to perform regularly.
The Lawrence Livermore Laboratory has developed a model that
combines theoretical calculations with experimentally defined parameters
to describe the overall efficiency of a germanium detector as a function
of gamma-ray energy, source-to-detector distance, dimensions of the
detector, extension of the source in area and volume, composition of the
source, external absorbers, and the dead layer of the detector.27'37
This model has been applied to point, disc, and cylindrical sources.
-------
-23-
Measurement errors of 1 to 2 percent are typical for small-volume sources,
regardless of energy or source-to-detector distance. Large-volume
sources placed close to the detector are accurate to within 10 percent.
The measurement errors were evaluated by comparing the values determined
by the model with the known value of the standards. Because this model
can handle only point, disc, and cylindrical geometries, its usefulness
is quite limited. Effective use of the model also requires a sizable
computer for data analysis and a staff with considerable knowledge of
germanium detectors to make the necessary measurements. This model
should be considered when calibration techniques are being reviewed, but
the advantages and difficulties of implementation should be carefully
evaluated.
The most common approach to calibration of germanium detectors is
the experimental determination of FEP efficiency as a function of energy
for each counting configuration. Practical calibration standards for
each counting configuration must be prepared from appropriate, calibrated
radionuclides. The composition of these standards should approximate as
closely as possible, with respect to density and attenuation factors,
the actual samples that are to be analyzed after calibration. Laichter
et al.38 prepared solid calibration standards using evaporated sea water
and tabulated mass attenuation coefficients for representative solid
environmental and biological media. One can use granulated salt to
represent soil and vegetation samples. Finely shredded styrofoam is
used to adjust the density of the solid standards to correspond to the
density of the various environmental samples. Small volumes of radio-
active solutions are thoroughly mixed with the salt and styrofoam. Food
coloring is used to indicate the degree of mixing.
Water samples are easily prepared by using acidified water with
appropriate concentrations of carrier solutions.
Experimental determination of the FEP efficiency as a function of
energy for each counting configuration does not require the sophisti-
cated computer codes or the measurement of detector characteristics
(dead layers, dimensions of the active volume of the detector, position
of the detector within the cryostat endcap, etc.) that are required for
the first two methods. Also, the experimental method is not subject to
the uncertainties associated with the theoretical or semitheoretical
models chosen to calculate the detection efficiency. However, the
experimental method is time-consuming because of the number of samples
that must be prepared and counted, but the time required to prepare and
count samples may be less than that required to set up the necessary
computer programs and perform and verify the calculations suggested for
the first two methods.
Techniques for handling and storing radioactive solutions are
beyond the scope of this report, but several excellent reports are
available.26»3&>44
-------
-24-
4.4.2.3 Calibration Sources
Appropriate gamma-emitting radionuclides must be selected for use
as standards in calibration. Characteristically, calibrated solutions
of these radionuclides should be readily available, in a pure form, and
have an accurately measured, relatively long half-life (greater than 30
days).
Accurate, absolute gamma-ray emission rates must be available for
calibration standards. These activities are reported in terms of
nuclear disintegration rates (e.g., sec"1, Curie), although standards for
gamma-ray emission rates can be used directly to determine efficiency
(Section 4.4.2.4). The best source of gamma-ray emission rates and
other decay information is the Nuclear Data Sheets,45 published monthly
by Academic Press. The latest mass entry in Nuclear Data Sheets can be
used with the volumes entitled Recent References to obtain the latest
references to data on any radionuclide. A certificate of calibration,
which conforms to the recommendations of the International Commission on
Radiation Units and Measurements,46 should accompany all standards. The
information supplied must include activity, uncertainty, reference date,
description of purity, method of measurement, chemical composition,
volume or mass, and values for decay corrections such as half-life and
branching fractions.
Radionuclides that are used to determine efficiency can be
classified into two groups: (1) those radionuclides with only a few
prominent gamma rays and (2) those radionuclides with many prominent
gamma rays. The radionuclides listed in Table 4 belong to the first
category and have been extensively used for efficiency calibrations.47'49
Several radionuclides can be used either singly or in combination
to determine an efficiency curve for a germanium detector over an.energy
range of 10 to 2800 keV. The energy range of interest must be adequately
covered by calibration points so that interpolations between calibration
points are accurate. ANSI Guide N71713 suggests points every 100 keV
from 60 to 300 keV, every 200 keV from 300 to 1400 keV, and at least one
between 1400 and 2000 keV. Such a spacing of calibration points will
clearly define the region from 60 to about 300 keV if large variations
in efficiency are prevalent.
Many radionuclides of the second category have been suggested, but
the most frequently used sources are 56Co, 75Se. llomAg, 133Ba, 1°28Eu,
182Ta, and 226Ra and its short-lived daughters.3*'49"63 The approximate
ranges of energies in which each radionuclide is useful are 56Co—840
to 3500 keV: 75Se--120 to 400 keV; llomAg—650 to 1600 keV; 133Ba—30 to
400 keV; iS2*Eu--40 to 1400 keV. i82Ta__100 to 270 keV and 1000 to 1250
keV; and 226Ra and its short-lived daughters—180 to 2400 keV. For
source-to-detector distances of 10 cm or less, correlated photon-summing
corrections should be made for radionuclides that emit gamma rays in
cascade with the gamma ray of interest (Section 4.4.2.4). The com-
plexity, magnitude, and uncertainty of these summing corrections lead
most laboratories to use several radionuclides with few prominent gamma
rays. Even some commonly used radionuclides (134Cs, 94Nb, 46Sc, 8%,
-------
-25-
TABLE 4. STANDARDIZATION SOURCES
Nuclide
22Na
24Na
46Sc
51Cr
54Mn
"Co
60Co
85Sr
88y
94Nb
95Nb
l°9Cd
113Sn
131X
Half-life3
2.60y
15. Oh
83. 7d
27.8db
312. 5d
272. Odd
5.27y
64. 8d
106. 6d
2.0 x 104yb
35.15d
450.0db
115. 2d
8.02d
Radiation
Y ray
Y ray
Y ray
Y ray
Y ray
Y ray
Y ray
Y ray
Y ray
Y ray
K x-rays
Y ray
Ka x-rays
Y ray
Y ray
Y ray
Y ray
Y ray
K x-rays
Y ray
Y ray
Photon
energy3
(keV)
1274.5
1368.5
2754.0
889.2
1120.5
320. 1C
834.8
14.4
122.1
1173.2
1332.5
13.4
514.0
14.2
898. Od
1836.1
871.1
765.8
88. 0C
24.1
391. 7C
364.5
Number of
photons emitted
per decay
0.9995
1.000
0.9985
0.9998
0.9999
0.098°
0.9998
0.096
0.856
0.9988
0.9998
0.507
0.9928
0.525
0.914d
0.994
1.000C
0.9980
0.0372°
0.795
0.649°
0.824
-------
-26-
TABLE 4 (continued)
Nuclide
134Cs
137Cs
i39Ce
i4iCe
140La
198Au
203Hg
207Bi
'"Am
Half-life3 Radiation
2.06y Y ray
30. Oy K x-rays
Y ray
137. 6d K x-rays
Y ray
32. 5d Kff x-rays
40.27h Y ray
2.696d Y ray
46. 6d K x-rays
Y ray
30. Oy Y ray
433. Oyd Lff x-rays
Lg x-rays
Ly x-rays
Y ray
Y ray
Photon
energy
(keV)
604.6
31.8/32.2
661.6
33.0/33.4
165.8
35.6/36.0
145.5
1596.6
411.8
70.8/72.9
279.2
1063. 6C
13. 9d
17. 8d
20. 8d
26. 4d
59.5d
Number of
photons emitted
per decay
0.975
0.0564
0.853
0.641
0.800
0.126
0.800
0.956
0.9553
0.101
0.813
0.753C
0.135d
0.210d
0.050d
0.025d
0.359d
All data, unless otherwise noted, is from Debertin et al.48
bData from Heath.30
CData from Hirshfeld et al.47
uata from Hansen et al.49
-------
-27-
60Co, 24Na, aad 140La) require summing corrections. Typical magnitudes
of correlated photon-summing corrections for radionuclides from the
first and second categories are given in Section 4.4.2.4.
The NBS issues mixed-radionuclide, emission-rate standards in
point-source form and in solutions of different concentrations.*29'64
These standards provide eleven principal gamma rays covering the energy
range of 90 to 1800 keV in reasonably well spaced intervals. Figure 4
shows a typical gamma-ray spectrum observed from part of one solution.
The standards were originally developed for calibrating germanium
detectors for environmental analyses.65 This annual mixed-nuclide
standard is issued each September. The solutions are easy to use and
offer the added advantage of traceability to NBS. Only two radio-
nuclides, 60Co and 88Y, must be corrected for summing (see Section
4.4.2.4, Table 5). An excellent description of these standard solutions
and potential problems associated with their use has been given by
Coursey.29 Other suppliers of calibrated radioactive materials are
listed in the International Directory of Certified Radioactive Materials.65
4.4.2.4 Efficiency Determination
After an experimental method for determining efficiency has been
selected, count-rate data must be acquired by using the appropriate
calibration sources. Before actual standardization is begun, the
spectroscopy system should be thoroughly tested to ensure electronic
stability (Sections 4.2 and 4.3). The calibration sources then must be
carefully prepared with a total activity low enough that random summing
is negligible. Random summing is a function of the shaping times of the
amplifier, because the shaping times determine the pulse width, which in
turn determines the probability of random summing. Count rates of 5000
total counts per second or less can be used with shaping times of 1 MS
or less13 without appreciable summing, whereas 2500 and 1250 total
counts per second or less can be used with amplifier shaping times of 2
or 4 ps or less, respectively.
The standard source should be carefully positioned in the detector
shield, and spectra should be acquired which contain a statistically
sufficient number of counts in each FEP of interest. ANSI Guide N71713
suggests a minimum of 20,000 net counts in each FEP. If one assumes
that the background counts are negligible, compared with the FEP counts,
20,000 counts will give a 0.7 percent counting uncertainty at a 68 percent
(la) confidence level. Once the calibration spectra have been obtained,
the net count rate must be determined for each FEP of interest. Any
method that yields consistent results may be used to determine net count
rate, but the method selected should yield the highest efficiency while
minimizing the background contribution to the FEP region (Section 5.1.3).
*Amersham/Searle Co., 2636 S. Clearbrook Drive, Arlington Heights, Illinois
60005, has developed a mixed radionuclide emission rate standard that is
traceable to NBS and will be available quarterly. This standard has the
identical composition of the NBS standard.
-------
**•%,
CD
^j
c
9
O
o
8
ij
\miS
K
H
C/}
i
M
.
* f
» * * • I * 1
..::". ; ,
N<-'^-^».,:; . ;'
'*~~*"*~" "'"~v'<--^-.*^U.^.»T'';w.-M~)%^%_j^(tir ; , !•
'""^ B!w^i"f- v"w~"X-Wr *
^•.••-'.^
88 136
II 1 1 1 1 1 1 1
II 1 1 1 1 1 1 1
122 165 279 391 514 661 898 1173 1332 1836
ENERGY (keV)
Figure 4. Typical gamma-ray spectrum detector. Observed from an NBS mixed-radioauclide
gamma-ray standard.
-------
-29-
When source-to-detector distance is 10 cm or less, correlated
photon-summing corrections should be applied to the count rates of all
gamma rays being used in the calculations that are emitted in cascade
with other gamma rays. The correction factor for summing, S, is given
by
n
S = 1/0 (1 - q^^) , (7)
where
n
n = product of terms "i" to "n,"
i
n = number of gamma or x-rays in coincidence with the gamma ray
of interest.
i = identification of the coincidence photon of interest,
q. = fraction of coincidence photons "i" in coincidence with the
gamma ray of interest,
e. = total detection efficiency of the coincidence photon "i,"
3. "
W. = angular correlation factor between the photon of interest
and the "i"th coincidence photon.2
The angular correlation factor is often approximated as one or is
assumed to be equal to one when the angular correlation coefficients are
not known. This correction should be multiplied by the net FEP count
rate to obtain the corrected FEP count rate.
Some representative total detection efficiencies are given in
Figure 5. Correction factors for coincidence summing for several gamma
rays emitted in the decay of 1528Eu, 60Co, and 88Y are given in Table 5.
The absolute FEP efficiency e is determined by
£ = C/Ny , (8)
where
C = net count rate of the FEP,
N = gamma ray emission rate of the calibration standard.
The gamma-ray emission rate N of the calibration standard is
reported as nuclear disintegration^rate, which can be calculated by
N = KD , (9)
-------
0.09
0.08_
0.07_
0.06_
B
z
tj
h-t
CJ
t—<
u,
w
o
0.0! _
O.M.
-30-
A
I *
,_^ 0.5-L MARINELLI
16Z Gc(Li)
3.5-L MARINELLI
16Z Ge(Li)
,3.5-L MARINELLI
8Z Ge(Li)
o.o: _
o.o:
0.0]
0.01
0.1
ENERGY (MeV)
Figure 5. Total detection efficiency,
-------
-31-
TABLE 5. CORRECTION FACTORS FOR COINCIDENCE SUMMING
Nuclide
i528Eu
60Co
88y
Energy
(keV)
121.8
244.5
344.2
411.1
444.0
779.1
964.2
1086.0
1112.2
1408.1
1173.2
1332.5
898.0
1836.1
Detector 1 (3.5-L)a
1.02
1.03
1.01
1.03
1.03
1.02
1.01
1.00
1.01
1.01
1.02
1.02
1.02
1.02
Detector 3 (3.5-L)3
1.03
1.04
1.02
1.05
1.04
1.03
1.02
1.00
1.02
1.02
1.03
1.03
1.02
1.03
*See Section 4.2, Table 1, for specifications for the detectors.
-------
-32-
where
N = number of photons of a given energy emitted per nuclear
decay (i.e., the branching fraction),
D = disintegration rate of the calibration standard.
Log-log plots of efficiency data, with energy as the ordinate and
efficiency as the abscissa, can be used to display a wide range of ener-
gies. These plots show a fairly linear curve above about 150 to 300 keV
for most germanium detectors. Several typical curves for calibrations
for aqueous Marinelli beaker geometries are shown in Figure 6. The
decrease in efficiency shown at low energies is caused by absorption of
photons by the sample matrix, cryostat endcap, and the dead layers of the
detector itself. The gradual decrease in efficiency at higher energies
occurs because of the reduction in the cross section for interaction of
the detector material (e.g., germanium) with increasing energy of the
incident photon.
4.4.2.5 Analytical Efficiency Expressions
When efficiency data are available for a sufficient number of
energies in the energy region of interest, a method of representing the
efficiency as a function of energy must be chosen. Graphical methods
can be used; however, it is often desirable to express the efficiency in
some analytical form as a function of gamma-ray energy. Such expressions
can be readily programmed and are adaptable to automatic data analyses.
Least-squares fitting procedures are used to fit the efficiency data
to an analytical expression. Such a technique has the advantage of yield-
ing an unbiased estimate of the fitting errors. However, the analytical
expressions must be carefully selected to avoid introducing systematic
divergences from the observed data.
Several semi-empirical formulas involving the physical character-
istics of the detector have been proposed66'7** to fit germanium detector
efficiency curves. Due to incomplete charge collection, local varia-
tions in efficiency,71 and uncertainties in the actual geometry of coaxial
and closed-end coaxial detectors, empirical analytical functions are often
preferred.63'72 Table 6 lists some proposed empirical analytical functions
with appropriate literature references for each expression. The energy
ranges over which these functions give adequate fits can be found in the
listed references or in the review papers of McNelles and Campbell72 or
Singh.73 When the available degrees of freedom are sufficiently large
(i.e., the number of data points should exceed the number of fit parame-
ters by at least one, preferably more), the data can be broken into two
separate regions for fitting, if fitting the data for one extended region
leads to systematic divergences.
All the expressions shown in Table 6 are similar, and several are
special cases of the more general equations. The simple log-log fit
(entry number 1 in Table 6) can produce reasonable results in the 300-
to 1800-keV region for many detectors. Figure 7 illustrates the
deviation of three detectors from this analytical form. The reduction
-------
-33-
w
H
o
H
W
H
3
o
w
3
lor
l.C
0.1
HI0.5-L MARINELLI 167, Ge(Li)
0.01
«.-.. -^3 -5~L MARINELLI 16% Ge(Li)
X M° e T MARINELLI 8% Ge(Li)
D EXPERIMENTALLY DETERMINED DATA POINTS
• or X DATA POINTS FROM A LOG-LOG REGRESSION OF
EXPERIMENTAL DATA
JL
10
100
1000
10,000
ENERGY (keV)
Figure 6. Typical absolute FEP efficiency curve.
-------
-34-
TABLE 6. ANALYTICAL FUNCTIONS USED TO FIT
EFFICIENCY DATA FOR GERMANIUM DETECTORS
Function/formula References
£n e = a. + &2 £n E 48
£n e = a £n (i^££I + , £n [L^22 74
i \ t / z \ £. /
2
£n e = al + a2 £n E + a3(£n E) 75
£n e = a1 £n E + a2(£n E)2 - a3/E3 76
• i
£n e = I a.(£n E)J" 27
e = al exp(a2E) + a3 exp(a^E) 77
32
e = (aj/E) + a3 exp(-a^E) + a exp(-a6E) + a^ exp(-agE) 72
-3 -2 -1
£n e = a..E + a.E + a3E + a, + a_E 47
£n e = a. + a_E + a_E2 + a.E3 + acE + a,E5 78
I i 5 H 3 O
where
£n = natural log,
exp = exponential function,
e = absolute FEP efficiency,
al' a2' a3' ' " ' = 3^^t Parameters-
-------
2
O
W
16% Ge(Li)
14% Ge(Li)
3.5-L MARINELLI
,.,3.5-L MARINELLI
-2.0 .
-3.0
i
u>
U1
500
1000 1500
ENERGY (keV)
2000
Figure 7. Deviation of log-log fit from experimental efficiency.
-------
-36-
in efficiency observed at lower energies is often difficult to represent
with any analytical function. Several of the terms (e.g., the -as/E3
term in the fourth expression) in the analytical functions shown in
Table 6 have been added to improve fits of the data in the lower energy
region.
-------
-37-
SECTION 5
DATA ANALYSIS
5.1 SPECTRAL ANALYSIS
Most routine analysis of gamma-ray spectra from germanium systems
is performed with the aid of computer-automated techniques. Automated
techniques of spectral analysis can handle a large volume of spectra
rapidly, thereby reducing manpower costs and producing cost-effective
data reduction despite high initial investment in equipment and manpower.
Software for automated spectral analysis can be developed by the users,
adapted from published routines, or bought from commercial suppliers of
nuclear equipment.
Most programs for automated analysis include two phases: peak
analysis and radioisotope identification.79 Peak analysis usually
includes procedures for spectral smoothing, peak locating, and peak
fitting. Features for eliminating pseudopeaks and calculating the
energy and gamma-emission rates of the FEP are also generally included
in this phase. Peak analysis software typically compiles a table of
FEPs found in the spectrum, gamma emission rates of the FEPs, estimates
of statistical uncertainty of the emission rates, and a list of pertinent
fitting parameters determined for each FEP. The peak fit information
and the estimates of statistical uncertainty can be used to evaluate the
overall accuracy of the calculated gamma-emission rate.
In the second phase, the isotope identification, the individual
radionuclides present in the spectrum of the sample are identified and
their activities on the date of sample collection are calculated. This
phase is often set up for automatic generation of reports for use in
fulfilling regulatory requirements. After an analysis program is
selected, the user should continually review new or modified methods of
analysis to ensure that the methods used meet changing analytical require-
ments in the most cost-effective manner.
Several aspects of spectral analysis should be considered when
selecting an analytical technique for data reduction of germanium spectra.
Our discussion emphasizes automated techniques because of their preva-
lence, but all the methods described can be adapted to manual analysis
of data. This discussion is not intended to be an exhaustive report of
all available techniques for analysis of data; however, it will serve to
introduce the most commonly used techniques.
5.1.1 Spectral Smoothing
Due to the statistical nature of the nuclear disintegration and
detection processes, the counts C in each channel of a spectrum can have
a statistical fluctuation of the order -,/C. These statistical fluctua-
tions tend to produce a quite irregular spectrum, which can often present
-------
-38-
pr obi ems when trying to locate peaks. Effective spectral smoothing can
reduce these statistical fluctuations; however, the smoothing process
must not shift the centroid of the peak appreciably or alter its area
significantly.
Techniques for spectral smoothing14'80'81 and methods using Fourier
transform82"84 have been applied to germanium spectra. The more sophis-
ticated programs for gamma-ray spectral analysis smooth the data before
trying to locate or fit peaks , whereas other programs smooth the data as
an integral part of the peak location.27'79'8*'88
Techniques for smoothing data that use Fourier transform methods
have been reported,82'84 but this approach has yet to be recognized as a
routinely acceptable technique. Inouye, Harper, and Rasmussen82 reported
good results on a wide variety of gamma-ray spectra, some very complex.
Inouye83 has used a square-root transform to filter out statistical
fluctuations. Tominaga, Dojyo, and Tanaka84 investigated the effect of
Gaussian-shaped digital filters on determination of the area of a peak.
The application of Fourier transforms to gamma-ray spectral analysis
should be studied more extensively before such a technique is adopted.
This technique may be quite expensive in terms of computer time unless
the recently developed Fast Fourier Transform algorithms are used.
The most commonly used method for smoothing fluctuating data is the
method of moving averages. Weighting factors can be chosen which are
exactly equivalent to those one would use for a least-squares fit of the
data to a polynomial. Weighting factors also can be chosen which con-
volute the points into a smoothed, "n"th-order derivative. Savitzky and
Golay14 provide an excellent discussion of smoothing and differentiating
data; they also tabulate the convolution integers for many cases.
Figure 8 shows (1) part of an unsmoothed germanium spectrum, (2) the
smoothed spectrum fit to a quadratic or cubic polynomial, (3) the smoothed
first-derivative fit to a quadratic polynomial, (4) the smoothed first-
derivative fit to a cubic polynomial, and (5) the smoothed second-derivative
fit to a quadratic or cubic polynomial. The convolution integers for
smoothing and for the smoothed second derivative are the same for a quad-
ratic and cubic least-squares fit. Table 7 lists the data from which
Figure 8 was plotted.
The smoothed data for a quadratic or cubic function, S., the first
derivative for a quadratic function, D. (quad), the first derivative for
a cubic function, D^^ (cub), and the second derivative for a quadratic or
cubic function, D.2, for a five-point smoothing process are calculated by
St = (l/35)(-3Ci_2 + 12Ci_1 + 17Ci + 12C - SC) ; (10)
Di(quad) = C1/10)(-2C._2 - C.^ * OC. * C.+1 + 2C.+2) ; (lU
D. (cub) = (1/12) (C._2 - SC.^ + OC. + 8C.n - C.^) ; (12)
-------
-39-
O
o
700
600
500
400
300
200
100
0
700
600
500
AGO
300
200
100
0
200
100
0
-100
-200
200
100
0
-100
-200
200
100
0
-100
-200
I I I I I I I I I I '1 I | i T t I I I I I t |
QUADRATIC AND CUBIC
FIRST-DERIVATIVE
QUADRATIC
FIRST-DERIVATIVE
CUBIC
SECOND-DEFUVATIVE
QUADRATIC AND CUBIC
» I I I I I I I I I I I I I I I I I I
0
10
20
30
40
50
CHANNEL (RELATIVE)
Figure 8. Effect of spectral smoothing and differentiation on a typical FEP.
-------
-40-
TABLE 7. EFFECT OF SPECTRAL SMOOTHING AND DIFFERENTIATION ON A TYPICAL FEP
~"
Channel
(relative)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Raw data
(counts)
13
10
16
6
12
9
13
21
9
14
24
28
19
55
48
61
103
171
215
409
547
654
643
455
302
110
42
20
15
9
Smoothed data
(quadratic or
cubic)
12.91
12.99
11.11
10.88
8.48
10.62
14.79
15.77
13.19
13.91
23.65
22.42
31.51
42.05
52.62
62.02
107.02
151.79
247.57
389 . 19
556.59
651.59
619.77
479.51
281.68
130.65
37.79
19.05
13.19
8.65
First derivative
(cubic)
1.29
-1.50
-0.59
-0.59
-0.29
3.09
0.59
0.59
1.50
2.89
3.39
7.69
7.50
9.49
17.39
28.69
44.39
80.79
112.59
129.79
110.09
18.79
-68.89
-142.89
-154.69
-112.99
-66.39
-22 . 89
-8.49
-2.99
First derivative
*»
(quadratic)
-4.08
2.74
-2.58
-2.58
2.24
-0.58
8.24
-3.08
-5.58
9.41
8.49
-6.74
15.99
16.58
-2.99
26.99
59.41
45.66
121.66
181.08
127.66
60.16
-112.24
-181.99
-179.91
-137.08
-36.08
-9.58
-4.24
-5.83
Second deriv;
(quadratic <
cubic)
-0.71
-1.00
0.28
-0.28
2.71
1.57
-1.99
-2.57
2.99
5.28
-4.85
5.57
3.28
0.14
4.57
25.57
12.57
39.99
41.42
10.00
-62.99
-109.99
-99.57
-46.71
28.71
55.14
59.99
20.14
5.00
3.14
-------
-41-
7 (continued)
1
annel Raw data
ative) (counts)
5
10
9
7
8
16
6
10
3
5
12
7
11
8
8
7
12
8
12
9
Smoothed data
(quadratic or
cubic)
6.88
8.28
9.08
6.99
10.48
11.11
10.88
6.14
5.05
6.11
8.74
10.17
8.77
9.19
7.05
8.88
9.25
10.74
9.85
10.37
First derivative
(cubic)
-1.09
0.00
0.29
1.09
0.29
0.39
-1.59
-2.50
0.69
0.29
1.79
0.50
-0.69
-0.29
0.09
0.39
0.89
0.39
-0.50
-0.69
First derivative
(quadratic)
1.16
2.83
-2.24
-1.16
6.25
-1.58
-3.58
-1.08
-3.83
6.25
0.66
-0.91
1.00
-1.99
-0.75
2.66
0.33
-0.16
0.91
-1.83
Second derivative
(quadratic or
cubic)
2.71
-0.28
-1.28
2.99
-1.28
-1.71
-2.28
1.85
2.14
1.28
-1.14
-1.57
0.42
-1.00
2.14
-0.28
0.14
-1.14
0.14
-1.57
function in parentheses is the type of polynomial used in the least-squares
fit.
-------
-42-
D.2 = (1/7)(2C._2 - C.+1 - 2C. - C.+1 + 2C.+2)
where
C. = number of counts in channel "i,"
i = running index of the channel number in the original
data table.
The convolution integers and the normalization integers are taken from
Savitzky and Golay,14 and the correction on the second-derivative con-
volution integers are taken from Quittner.11
Yule81 noted that the tops of photopeaks are roughly quadratic and
the sides are roughly cubic, which suggests that quadratic or cubic
polynomial smoothing constants can be chosen. Fortunately, the weighting
factors for a least-squares fit to quadratic and cubic polynomials are
identical. Yule80'81 also found that the optimum number of channels to
use in a smoothing process is equal to the number of channels in the
FWHM of the FEP; the use of fewer channels will not produce adequate
smoothing, and the use of more channels produces peak distortions.
To use the weighted, moving-average technique, the product of an
appropriate weighting factor and the contents of the channel are summed
for a selected number of consecutive channels; this sum is then divided
by an appropriate normalization factor. The moving average is obtained
by repeating the above process while advancing across the region of
interest one channel at a time. In this process, the lowest energy
channel is dropped from the region, and the next higher energy channel
is added to the region to maintain a constant width. A (2m+l)-point
smoothing process can be decribed by the general equation,
-hn
S. = (1/N ) I [(a .)(€._)] ,
3. m' l^ m,j'v i+j'J '
j=-m
where
S. = smoothed value of channel "i,"
N = normalizing factor,
i,j,m = indices,
a . = convolution integer for channel "j,"
C.+. = number of counts in channel c of the raw data (i.e.,
data before smoothing).
Yule80'81 used different criteria to test the effectiveness of
smoothing: (1) visual inspection of plots of the observed and smoothed
data, (2) comparison of areas under the observed and convoluted results,
(3) behavior of the smoothed first derivative, (4) statistical tests,
and (5) the effect of multiple smoothings. Yule80'81 found that the
first four tests could not adequately detect small amounts of distortion;
-------
-43-
therefore, the fifth test is used to magnify the distortion until it is
detectable. In general, a single smoothing with the number of data
points (channels) approximately equal to the FWHM of the peaks introduces
negligible distortion. However, when overlapping peaks are present,
distortion can be introduced. These conclusions indicate that smoothed
data are preferred for locating peaks, but should not be used when
determining areas of peaks if overlapping peaks are present.
Least-squares fitting methods, when applied to previously smoothed
data, tend to extend the fitting width of peaks over the width determined
when raw data are used. In other words, smoothing broadens the peak.
Fitting width is defined as the region of data, representing a peak, on
which a least-squares technique is applied. An equivalent result is
obtained by a least-squares fit of unsmoothed data over a wider-fitting
range.84 Because smoothing data does not improve the least-squares fit
and can obscure multiplet structure, unsmoothed data should be used to
determine FEP areas.
5.1.2 Peak Location
Locating peaks by careful visual inspection of spectra can be
time-consuming and dependent on subjective judgment. This dependency
has led many authors to propose various methods for locating peaks with
calculators and computers. Much emphasis has been placed on methods
that are rapid and easily adapted to small, dedicated computers.
Three major methods, and variations of those methods, have been
proposed for locating peaks: (1) location of maxima, (2) testing the
smoothed first derivative, and (3) examination of second differences.
Statistical tests can be used to confirm peaks after the spectra are
treated mathematically. In the first method, locating peaks by finding
maxima, the data are scanned until the contents of a given channel N
satisfies the following conditions:
C(N - 2) < C(N) - S VC(N) (15)
and
C(N) - SVC(N) > C(N + 2) , (16)
where
C(N) = contents of channel N,
S = sensitivity of the peak detector.
The factor S is chosen empirically; values of 1 to 1.5 have been reported
to give good results.89 This method has been used in a program called
"INHOUD." Kemper and van Kempen89 describe this program and provide
Basic, Fortran, and Algol listings.
-------
-44-
Wood and Palms90 describe a program that uses a modification of
this general approach. The spectrum is scanned channel by channel, and
a peak is detected when
2 6 6 1/2
(1/3)1 I C(n + i)] > (l/3)[ 2 COH- j)] + S{(l/3)[ 2 C(n+j)]}1/Z, (tf,
i=0 j=4 j=4
where
C(n •*• i) = contents of channel (n + i),
S = sensitivity of the peak detector.
The value of S is chosen empirically.
The procedure for locating peaks by finding maxima is easy to
program, but often overlooks small peaks, identifies statistical
fluctuations and other fluctuations in the spectra as peaks, and cannot
separate multiplet peaks. This method of locating peaks can be aug-
mented by rejecting (1) any peak that exceeds a certain width and (2)
any peak with an area less than twice its standard deviation; these
rejections reduce the number of pseudoFEPs detected.
Several authors 79>91~93 use the first derivative of the spectrum
to locate peaks. The smoothed first derivative of a peak will show a
maximum difference in slope at the points immediately to the left and
right of the centroid of the peak. These points contain half as many
counts as the peak centroid itself. The smoothed first derivative will
have a value of zero at the peak maximum. A program can be written to
detect groups of neighboring channels such that the smoothed first
derivative D fulfills the criteria,
D(p) < 0 , (111
D(p + i) < 0 for i = 1, 2, . . ., r ; (Ml
and
D(p - i) > 0 for i = 1, 2, ...,£. (20)
The values of r and £ are chosen in accordance with the energy resolu-
tion so that significant peaks are recognized and statistical fluctuations
a re dis rega rded.l1
Barnes79 and Sasomoto et al.91 suggest that statistical tests be
performed on the derivative to ensure that only significant variations
in the derivatives are processed. An example of the first derivative of
a peak that has been fitted to a quadratic or cubic function is shown in
-------
-45-
Figure 8. Yule93 suggests that a fit to a quadratic function best
locates the centroid of the peak; but Barnes79 notes that, because there
is no obvious choice between quadratic and cubic functions for locating
single peaks and because the cubic function is more sensitive in
detecting incompletely resolved doublets, the cubic function should be
preferred.
The second derivative can also be used to find peaks. The second
derivative of discrete data can be replaced by the second difference,
defined by
DJJ = C(N + 1) - 2C(N) + C(N - 1) , (21)
where
2
DM = second difference at channel N,
N
C(N) = contents of channel N.
Large fluctuations in the second difference, resulting from statis-
tical fluctuation, Compton edges, and other irregular spectral features,
are invariably present. Therefore, some form of smoothing procedure or
other modification is required to identify peaks reliably.
Mariscotti85 proposed a method by which the data are smoothed "z" times
over "w" channels to give the generalized second difference. This
generalized second difference, S (i), and its error, AS (i), are
given by ' '
i+zm-H
S,w(i) = * C (i - j)Y(j) (22)
<£* « W , . - £* • W
j=i-2m-l
and
'i+znrH
where
* * * t* * W
j=i-zm-l '
[C (i - j)2]1"* , (23)
C (i - j) = weighting factors for the "i"th term,
z,w
Y(j) = contents of channel j.
A detailed study of the optimum selection of the averaging parameters w
and z indicated that the best results are achieved when w = 0.6F and z =
5, where T is the FWHM of the peak.85 Mariscotti85 used the condition
S (i) > fAS (i), where f is a factor of confidence, to indicate a
possible peak?'"Additional conditions, based on some distinctive peak
characteristic, are used to eliminate pseudopeaks such as Compton edges
and other irregular spectral features.
-------
-46-
Routti and Prussin86 use a modification of Mariscotti's second-
difference procedure in the computer program SAMPO to perform auto-
matic peak search. A smoothed second difference is calculated over (2k
+ 1) channels. The second difference S(i) and its standard deviation
AS(i) are given by
+k
(2
and
AS(i
j=-k J
/ +k 2
i) =( I C.V .
\j=-k J 1+J
where
c. = weighting factors for the "i"th term,
•J
Yi+j = contents of channel i+j.
The weighting coefficients are determined from the second deriva-
tive of a Gaussian. Optimum coefficients are chosen to maximize the
ratio of the second difference to the standard error of the second
difference. The weighting function may be varied to determine small
peaks or components of multiplets more accurately. Peaks are located by
comparing the statistical significance of the second difference [SS(i) =
S(i)/AS(i)] to given threshold values. An additional test is made on
the shape of the peak before it is recorded as a valid peak.
Phillips and Marlow87'88'94 use a modification of the smoothed
second difference, which was first proposed by Robertson et al.9S The
search routine in their program, HYPERMET, uses a negative smoothed
second difference, which is calculated by a sliding transform, often
called a square-wave transform. The transform for an integer width M
and its statistical variance are given by
J+2M-1
KON(J) = I [L(D][K(I)] (26)
I=J-M
and
J+2M-1
VAR(J) = I [L(I)r[K(I)] , (27)
I=J-M
where
KON(J) = smoothed second difference of the Jth channel,
VAR(J) = variance in KON(J) for the Jth channel,
K(l) = number of counts in the Ith channel,
L(I) = -1 for J-M < I < J-l
= 2 for J < ! < J+M-1
= -1 for J+M < I < J+2M-1.
-------
-47-
The expected value of variance in KON(J) can be calculated simply by
VAR(J) = 6MK (28)
where
M = width of the transform,
K = average of K(I).
A function, CONV(J) [= KON(J)/[VAR(J)]1/2], can be used to detect
peaks. This function is normally distributed about zero with unit
standard deviation, except in the vicinity of a peak or other sharply
varying feature with a width of the order of M. When the function of
CONV(J) exceeds a certain threshold value, usually 4, a peak is identi-
fied. Due to the width of the onset of Compton edges, the function
CONV(J) is found to be insensitive to Compton edges; thus, further
discrimination based on peak shape, which is necessary in most routines
when generalized second difference is used to locate peaks, is not needed.
When the threshold value for peak detection is set below 4, a
rectangular wave is used to detect peaks. The rectangular-wave transform
can be written
J+3M-1
KON(J) = X [L(I)][K(I)] , (29)
I=J-2M
where
L(I) = 1 for J-2M < I < J-l,
= 4 for J < I < J+M-1,
= -1 for J+M < I < J+3M-1,
and all other parameters are as defined previously. The rectangular-wave
transform is more sensitive than the square-wave transform for detecting
a weak peak on a high background.
Phillips and Marlow87'88'94 performed a detailed comparison of four
transforms: (1) square wave, (2) rectangular wave, (3) covariance cal-
culated by means of a zero-area Gaussian, and (4) second derivative of a
Gaussian using an optimum width. The zero-area Gaussian was determined
to be best, but the rectangular-wave transform was nearly as effective.
Performance of the transforms on gamma-ray spectra with poor statistics
was tested in detail.
The great advantage of the square- and rectangular-wave transforms
is that the calculations can be accomplished by computing devices in
fixed-point arithmetic, which is an efficient and rapid calculating mode.
This is of considerable value because of the increased application of
dedicated mini- or microcomputers to data acquisition and processing.
Several other methods of detecting peaks have been described in the lite-
rature,11'27'96'97 but none has been used as widely as the procedures
described above. In general, the square- or rectangular-wave transforms
should be adequate for most environmental applications. If greater detail
-------
-48-
about a particular method is required, the reader should refer to the
original literature as referenced in this section.
5.1.3 Peak Intensity and Peak Centroid Determination
In analyzing gamma-ray spectra of environmental samples, the ulti-
mate task is to measure the FEP intensity accurately. Many specific
approaches to determining FEP intensity have been developed from two
techniques—nonfitted methods and fitting methods.
The FEP intensity can be determined directly from the counts in the
channels of the peak by some form of summing or weighted summing of the
counts in each channel. When this method is used, the background is
generally estimated from several channels within the peak or in the
vicinity of the peak. This technique can be termed determination of
peak area by nonfitted methods.
A second method of determining the intensity of an FEP uses a
computer or programmable calculator to fit the peak to a function before
determining the area and centroid of the peak.
A detailed account of each method is beyond the scope of this
report and would be of limited value because comparisons of the methods
by various authors have yielded conflicting results.98'101 However, a
brief description of each method may be of value in selecting a method
for data analysis or to understand better the advantages and limitations
of a previously selected method.
5.1.3.1 Nonfitted Methods
Most nonfitted methods can be adapted for hand calculation or readily
programmed on a calculator or small computer. These methods have particu-
lar merits and drawbacks; of particular note is the fact that nonfitted
methods are capable of processing only singlet peaks. Fortunately, many
types of environmental samples have spectra with few multiplets. The
difficulties encountered as a result of the inability of these methods to
resolve multiplets quantitatively should be considered in light of the
relative number of multiplets that will be encountered in the routine
samples of a laboratory.
Several authors have identified and compared various nonfitted
methods of peak analyses.98'99'102 Seven basic nonfitted methods of
determining areas of peak have been proposed and used in different
laboratories:
Total peak area (TPA) method
A = 2 a - (r - SL + l)(a, + a)/2
x * r
-------
-49-
where
A = total net peak area,
ai = number of counts in channel "i" ,
S, = channel number at left limit of FEP,
r = channel number at right limit of FEP.
2. A modification by Wasson of the TPA method98
A = I at - (n + 1/2) (b + b ) , (31)
i=-n
where
n = number of channels to the left and right limits of the FEP
from the centermost channel,
bn = background in channel "n" as determined from a straight
line drawn between channels "£" and "r".
3. Covell's method103
A = I a.. - (n + 1/2) (an + a_Q) . (32)
i=-n
4. Sterlinski's method104
A = na + I (n - 2j + l/2)(a . + a.) . (33)
° "J J
5. A combination of the Wasson and Sterlinski methods98
A = na + I (n - j + l)(a . + a.) - Z (j -»- l/2)(b + b ) . (34)
° i=l "^ ^ 1=1
6. Quittner's method105'106
A = (a - C ) , (35)
• _ l. ^
i=-h
where
C. = background in channel "i" as determined from a polynomial fitted
1 to (2k + 1) channels on each side of the photopeak
-------
-50-
V
3(P,
(qr
M
(X+ i - XL)'
where
X, and X_
Xp
P£ and pr
q£ and qr
M
£ - Pr)
M3
(X
i ~ V
center channel of the regions described by the left
and right polynomials, respectively (X- = £ - k, X_
r + k),
centermost channel,
values of the polynomials at XL and XR,
slopes of the polynomials at XL and XR ,
XR - XL .
7. A combination of the Sterlinski and Quittner methods98
A = na
and
+ I (n - j + l)(a + a ) - Z (j + 1/2)(C - C )
j=l -J J j=i J J
A = I (a. - b.) ,
(31
where
a.
b.
= contents of channel "i,"
£ = left point at which the measured and the baseline spectra
intersect;
r = right point at which the measured and the baseline spectra
intersect,
= counts calculated for the baseline spectra at channel "i."
The various experimental conditions (peak separation, peak-to-
background ratio, and peak broadening due to high count rate and instru-
ment drift) are so complex that no method of determining the peak area
can be clearly shown to be superior to all other techniques under all
conditions. The different conclusions drawn by Hertogen et al." and
Baedecker98 serve to emphasize this point. The many options investigated
by Hertogen et al.,99 such as the channel width used in determining the
peak area or in determining the background, also indicate the complexity
involved in comparing various techniques. Nevertheless, a qualitative
comparison of the techniques is worthwhile.
-------
-51-
All the techniques proposed (other than the TPA method) attempt
either to minimize the effects of statistical fluctuations on the peak
or improve the estimation of the baseline (background).
The TPA method obtains the full area of the FEP by summing and
determines the background from a straight-line fit to a continuum channel
or an average of several channels on each side of the FEP. This tech-
nique is readily adaptable to either hand or computer calculation.
Errors in estimating the net area due to peak broadening are less than
those from any other technique listed above; however, the technique sums
over a region of channels that add considerably to the error of the peak
integration, but add little to the net number of counts. The results
are not strongly influenced by statistics when the FEP is relatively
intense and the background is not excessive, but are subject to large
statistical fluctuations for weak peaks on high backgrounds. The Wasson98
modification of the TPA method excludes those channels in the wings of
the FEP that add considerably to the error of peak integration while
adding little to the net number of counts. The Wasson technique will
yield greater errors when peak broadening has occurred.
Covell103 based his method on the fact that a summation over a
fixed portion of the FEP bears a constant proportionality to the total
area of the FEP. The method is simple to program or calculate by hand,
but is subject to errors from statistical fluctuations and peak broaden-
ing. The consistency and accuracy of the results of this method depend
on optimum selection of peak boundaries. Heydorn and Lada107 found that
the optimum integration limits on each side of the centroid are a func-
tion of the peak-to-background ratio and correspond to about 0.91 FWHM on
either side of the peak for those peaks that are small relative to the
background.
Sterlinski104 modified Covell's method to give increasingly greater
weight to those channels that have an increasingly greater number of
total counts. The variance of either Covell's or Sterlinski's method is
increased because the summation is carried out with the use of a relatively
high baseline, rather than one drawn across the base of the FEP.98
The Wasson-Sterlinski98 technique uses the advantage of the Sterlinski
approach: It gives increasingly greater weight to those channels that
have an increasingly greater number of total accumulated counts and
eliminates the increased variance of the Sterlinski technique resulting
from the high baseline.
Quittner105'106 attempted to improve the estimation of the baseline
by fitting a polynomial to several points on the left and right of the
peak. This technique gives a good representation of the background for
weak peaks, or peaks near the Compton edge, but requires several back-
ground channels, usually at least five, on the left and right of the
peak.
The Ralston and Wilcox108 technique, which uses an iterative method
to construct the baseline, yields precise results, but gives a system-
atically lower baseline.105 The low baseline will lead to large errors
when evaluating small peaks superimposed on large backgrounds or Compton
regions.105
-------
-52-
Baedecker98 found that the methods of Covell and Sterlinski showed
poorer precision than other methods of peak analysis, due perhaps to the
sensitivity of these two methods to loss of resolution or shifting of the
peak. The TPA method was found to be less precise than the Wasson, Wasson-
Sterlinski, Quittner, or Sterlinski-Quittner methods.98 After comparing
the four techniques, Baedecker98 and Hertogen et al." concluded that
Quittner's method gave better results for peaks with small peak-to-
background ratios; however, this method gave poorer results when the poly-
nomial could not be fitted accurately because of an inadequate number of
background channels on either side of the peak. The Ralston-Wilcox method
gave systematic errors when the peak-to-background ratio was low.99
Hertogen et al.99 found that errors in the calculated area of the peak as
a function of decreasing peak-to-background ratio were not biased for most
methods and did not increase rapidly with decreasing peak-to-background
ratio. These conclusions are very reassuring for applications in
environmental analysis.
Host methods could be applied to environmental analysis. The
advantage of the Quittner method is its good performance for small
peak-to-background ratios. However, the results of a spectrum analysis
must always be carefully scrutinized by an experienced spectroscopist,
regardless of the technique used.
The TPA method is often used as a back-up method or as a semi-
quantitative check on other methods; therefore, three computer
programs89'90'109 based on this method are described briefly.
All three computer programs locate the channels beginning and ending
the peak by statistical tests. The backgrounds are estimated by a straight
line drawn between the average of several channels on the right and left
of the peak. The centroid of the peak is located by various methods. One
computer program locates the centroid of the peak to the nearest half-
channel (statistical fluctuations will decrease the accuracy of this
method) by choosing the channel with the largest number of stored counts
as the location of the centroid of the peak. 9 Other programs locate the
centroid from its first moment M, given by
R
M = (I/A) I iN. , (38;
i=L x
where
A = net counts in the peak,
L and R = channel numbers assigned as the left and right
limits of the FEP,
Ni = net counts (counts after the background has been
subtracted) in channel "i."
The centroid can also be located by estimating the channel number at which
the first derivative of the data changes sign.
-------
-53-
The variance V for the TPA method is given by
r-1
V = I A + [(r - £ + l)/2]2 (A + A ) , (39)
i=£+l x * r
where all terms are as defined previously. The variance for each method
can be found in the reviews and references listed above or evaluated by
using standard techniques for propagation of error (see Ref. 110).
5.1.3.2 Fitting Methods
Many approaches to fitting an FEP are described in the literature.
Essentially, all approaches fit some analytical function, most often a
Gaussian function, to the FEP by means of a least-squares technique.
The forms of the analytical function and the methods for obtaining the
least-squares solution are numerous.
Several analytical functions and least-squares techniques for fitting
have been chosen for either simplicity111"113 or performance of a special-
ized task.114"117 Much recent work in this area has been concerned with
adapting both the fitting functions and fitting processes for use on small
computers.112'113'118 Some adaptations sacrifice accuracy to increase the
speed of processing and to reduce the amount of computer memory required.
5.1.3.2.1 Fitting Functions
The FEP generated by a germanium detector has a predominantly
Gaussian shape with a tail on the low-energy side. A Gaussian function
is described by
Y(X) = A exp[(Xp)2/(2o2)] , (40)
where
Y(X) = amplitude of the Gaussian at channel X,
X = channel number,
A = height of the Gaussian,
p = centroid of the Gaussian,
a = width of the Gaussian (FWHM = 2.3550).
-------
-54-
The Gaussian shape of the germanium FEP results from the statistical
spread of the electron-hole pair energies generated in the germanium
detector. The noise of the electronics system further broadens the FEP.
The tail on the low-energy side of the FEP results from the trapping and
recombination of charges, escape of electrons from the sensitive volume
of the detector, and Compton scattering into the detector from surrounding
material. The FEP also may exhibit a tail on the high-energy side of
the peak, which results from pulse pile-up at high count rates, improper
pole-zero cancellation, or gain-stabilizer effects. However, this tail
can usually be eliminated by proper adjustment of the instrument.
The true shape of the peak generally is approximated by using an
analytical function containing variable parameters. These shape parame-
ters then can provide an easy method for characterizing (i.e., obtaining
the area and the centroid of) a peak. Obviously, the results obtained
from fitting the function are most trustworthy when the analytical
function reproduces the experimentally observed shape.
Many different functions have been used to represent spectral peaks
(Table 8). The exact equation for a particular functional form can be
obtained by referring to the appropriate reference listed in the table.
The functional representation of the FEP should be as simple as
possible while still describing the peak well. More complex functions
tend to require more computer memory and computation time, while often
tending to diverge, oscillate, or converge to spurious optimal solutions.
This last problem can be minimized by placing appropriate constraints on
the fitting parameters.121
In general, the complexity of the analytical function and the
complexity of fitting the function increase directly with the number of
fitting parameters used. For this reason, the number of fitting para-
meters is held to a minimum and is either constrained (held to maximum
or minimum values relative to their initial values)87'88 or held constant
for an FEP of a certain energy.27'86'117'124'127
Routti and Prussin86 were the first to use parameters with fixed
values. They selected well-isolated, intense peaks to determine values
for these fixed parameters by a process termed peak-shape calibration.
The values are usually determined from peaks within the actual spectrum
or are determined for a given spectroscopy system and used for all
subsequent analyses. Determining the parameter values within the actual
spectrum or varying the values with constraints should yield the best
results because a spectroscopy system is often not completely stable
from sample to sample. However, determining the parameter values from
the actual sample spectrum requires more computation time and may
require more intervention by a knowledgeable spectroscopist than would
be required if the values were known from previous measurement and held
constant during the fitting process. Gunnink and Niday27 and Phillips
and Marlow87'88 present a good discussion of the calibration of peak
shapes and optimization of the values of peak-shape parameters.
-------
-55-
TABLE 8. FITTING FUNCTIONS
Functional description
Number of
fitting parameters
References
Simple Gaussian
Two Gaussians (same width)
Gaussian with an exponential
tail on the low-energy side
Two Gaussians
(independent widths)
Gaussian plus exponential tails
on low- and high-energy sides
Gaussian plus exponential
and error functions on the
low-energy side
Gaussian plus a complex
exponential tailing function
Gaussian with an exponential tail
and a dual exponential function
to describe the continuum
Gaussian plus a dual exponential
tail on the low-energy side
and an exponential function
on the high-energy side
Skewed Gaussian with a constant
tail and an exponential tail
on the low-energy side
Two Gaussians with an arc-
tangent function and a
constant on the low-energy side
3
4
6
6
85, 118-123
100
100-102
100
86
100, 124
27, 125
91, 101
78, 79, 87, 88, 121
100, 126
100, 101, 117
-------
-56-
When peak-shape calibration is used, only the position, the ampli-
tude, and possibly the width of the peak are allowed to vary. Most
fitting programs hold all values constant except the position and height
for each component of a multiplet peak. This increases the number of
fit parameters by two for each additional component of a multiplet peak.
In addition to attempting to locate components of multiplet peaks
in the peak search routine, most sophisticated programs test the good-
ness of the FEP fit (i.e., check the residuals of the fit) to find
additional weak or ill-defined components within a peak or group of
peaks.
A practical consideration, often overlooked, in obtaining a good-
quality fit is that the number of points in the FEP, not including
background channels, must be greater than the number of free (noncon-
strained) parameters in the fitting equation.122 This may require many
channels within the FEP when peak-shape calibration is not used or when
complex analytical functions are used. An energy calibration of about
0.5 keV/channel will produce an adequate number of channels in the FEP
if peak-shape calibration is to be used.
Extra terms are generally added to the functions to account for
background contribution to the shape of the FEP. The functions have
varied from the simple assumption of a constant background to the complex
treatments of adding a third-order polynomial or an exponential term to
the shape function. A quadratic function has been the most frequently
used method of compensating for the background contribution,87'*8'124'127
with a cubic function also being used often.117'126 Quadratic and cubic
polynomials generally give superior results if the slope of the back-
ground within the spectrum varies rapidly.105'106*126 The most common
cause of rapid background variation is superimposition of the FEP on a
large Compton or beta continuum.
Several authors have evaluated the effects of fitting functions on
the results of FEP analysis.100'101'121 McNelles and Campbell100 place
the functions into three classes: (1) simple forms (entry number 1 in
Table 8), (2) forms using one-component tailing (entry numbers 2 through
5 in Table 8), and (3) forms using two-component tailing (entry numbers
6 through 11 in Table 8). They found that the functions using two-
component tailing gave the best fit. Results of a study by Lederer121
agree with those of McNelles and Campbell,100 whereas Takeda et al.101
concluded that the functions did not differ significantly. Takeda et
al.101 did find a difference of 5 to 10 percent in the determined areas
of the peaks; however, no systematic tendency was observed. This
variation in peak areas does subject the conclusion of Takeda et al.,100
that all fitting functions evaluated were equally accurate, to question.
Some comparisons of the various fitting functions and data to establish
their validity are described in the references listed in Table 8.
In general, comparisons of fitting functions and the data used to
test the performance of the functions have used strong peaks and/or weak
backgrounds.100'101'121 Phillips and Marlow94 tested their fitting
function on spectra with very poor counting statistics similar to those
often encountered in environmental applications. Their functions gave
-------
-57-
good results for such peaks, but the weighting scheme normally used in
the least-squares analysis showed a definite trend toward smaller-than-
actual areas for the peaks and an increased fluctuation in mean position
of the peak.
In the usual weighting procedure, where weights are set equal to
the reciprocal of the number of observed counts, counts that are less
than their expected value are given a larger weight than counts that are
greater than their expected value. Therefore, the area of the peak is
often underestimated. This underestimation may be serious for spectra
in which many channels contain less than about 50 counts.105 Phillips
and Marlow94 used a three-channel average for weighting purposes to
remove the bias when the peak contained several channels with contents
of less than 25 counts.
Differently shaped functions provide a variety of definitions of
peak shape and, therefore, can produce a variety of different areas for
a peak. These areas may be systematically biased if the experimental
shape of the peak and the analytical function do not adequately agree.
The systematic errors may be minimized when singlet peaks are considered
if the efficiency calibration has been performed with the same shape
function48 that is used in the spectral analysis. However, this approach
is not always adequate when multiplet peaks must be considered.
5.1.3.2.2 Least-Squares Methods
A nonlinear least-squares program usually is used to obtain the
best estimates for the parameter values to be used in the fitting pro-
cess. The least-squares procedure minimizes a function, x2. with respect
to all chosen parameters simultaneously:
a-*2 = §-.* tY. - F.(p )]/AY^ = 0 , (41)
Fj *j i=l
where
Y. and AY. = measured value and error, respectively, at the
1 x "i"th data point X- >
Fi(pj) ~ value predicted by the function used in the fit.
The function X2 is considered to be a continuous function of the
"n" number of parameters p. describing a hyper-surface in "n" dimensional
space. The least-squares process searches the hyper-surface for the
actual minimum value of X2- This search is complicated by the fact that
there are often local minima, besides the actual minima, for x2 within a
reasonable range of values for the parameters p..110
J
Several methods have been formulated for performing the least-
squares fit that have been used to the FEP: (1) the Gauss* method,
*Also referred to as the Gauss-Seidel, Seidel. Gauss-Newton, Newton-Rapson,
Taylor series, or linearization method.121'1 8
-------
-58-
(2) the gradient-search or steepest descent method, (3) the variable-metric
minimization or maximum-neighborhood method, and (4) the random-search
method.
With the Gauss method, initial estimates of parameter values are
put into a first-order Taylor series expansion to obtain a linear
expression, which can, in turn, be treated by standard least-squares
theory. The improved estimates are then returned through the process
to obtain a more refined set of parameters. Programs using the Gauss
method have been developed by Booth and Peterman.129 Crosbie and
Monahan,130 Moore and Zeigler,131 and Hartley.132 The Gauss method has
several drawbacks: (1) It may converge slowly, necessitating a large
number of iterations; (2) it may oscillate wildly; (3) it may not
converge at all; or (4) it may diverge.110'128 However, if the initial
estimates are in the immediate vicinity of the minimum value of X2» it
converges relatively rapidly and reliably.
In the gradient-search or steepest-descent method, all values for
the various parameters are incremented simultaneously, so that the
relative magnitude of each parameter value is adjusted to give a resultant
direction of travel on the hypersurface toward the direction of maximum
change in x2-110 This method will approach the minimum value of x2
rapidly, but in the vicinity of the minimum x2 value, it converges very
slowly.128 Marguardt133 and Bevington110 developed programs based on
this method, and Spang134 developed several modifications to the basic
steepest-descent method.
The method of variable-metric minimization has been widely
used.86"88'122'127 In this method, an optimum interpolation between the
Gauss method and the gradient-search method is performed to combine the
best features of each. This technique was developed by Marguardt135'136
and programmed by Davidon.137 Beals138 also developed a variable-metric
routine. Phillips and Marlow87'88 also use variable-metric minimiza-
tion139"141 with modifications by Fletcher.142'143 The main advantages
of the variable-metric method are rapid convergence near the minimum value
of x2 and a relatively small amount of computer memory. Davidon1s137
formulation is slow because a relatively large number of tests and trial
steps are used to check for convergence after each iteration. Fletcher's
modification142'143 greatly increases the speed of the program.87'88
The random-search method has been used by two authors101'113 to
optimize values of the fitting parameters. This method does not require
good initial estimates of the fit parameters, and it converges rapidly
without divergence or oscillation. The random-search technique was
developed for computers that have a small memory allocation, but it
requires longer computing times than other conventional fitting tech-
niques.113 This disadvantage may not be serious if the available computer
is part of a P-MCA system, in which the program can be allowed to run in
background mode while data acquisition is carried out in the foreground
mode.
Several other methods or variations of these methods have been used
or suggested for fitting an FEP.100'112'118'121'123'126'144'145 The
preceding discussion can be considered applicable to most of them.
-------
-59-
Good initial values of the parameters used in a fit are essential
for rapid, reliable convergence. Limiting the number of fitting parame-
ters is also important for reliable results. Discussions of how to
obtain good initial values of the fit parameters and methods of limiting
the fitting parameters can be found in the references cited in Table 8.
Estimates of the statistical error in the net area of the peak and
of the fitting error are required for assessing the overall error in
estimating net counts in the FEP. The statistical error S (in percent)
evaluated to one standard deviation is calculated by
S = (Ap + 2Ab)1/2(100%)/Ap , (42)
where
A = net area of the peak,
Ab = background underneath the peak.
Calculation of the error due to the least-squares fitting process
is assumed to be the same for both the nonlinear and the linear case.
The inverse of the generated square symmetric matrix containing the
weighted derivatives of all the values of the parameters in the fitted
function also forms an error matrix containing the variances and
covariances of all values of the functional parameters. The determinant
of the error matrix represents the overall variance of the fit. Because
the standard deviation is the square root of the variance, the fit error
F (in percent) evaluated to one standard deviation is given by122
F = (100%)[S.D./( 0 |A(i)|)]1/Q , (43)
where
S.D. = standard deviation of the fit,
n = number of parameters,
A(i) = parameter number "i."
This method of calculating fit error is a qualitative measure of
the goodness of fit and is not absolute.122 For a more detailed dis-
cussion of the error in the fit, see Draper and Smith,128 Bevington,110
and Roscoe and Furr.122 A more detailed account of nonlinear least-
squares fitting can be found in the original references listed above or
in the books by Draper and Smith128 and Bevington.110
In summary, the use of two-component tailing and fitting by the
variable-metric method is preferred. However, if only a small computer
is available, simpler analytical functions, using the random-search
method of least-squares fitting to perform quantitative analysis, may be
employed .
The results of any fitting routine should always be thoroughly
inspected by a competent analyst. No program can consistently produce
reliable results for a wide range of sample data; therefore, there is no
substitute for the experienced spectroscopist.
-------
-60-
5.1.4 Automatic Analysis
Host environmental laboratories adopt some form of automatic analy-
sis of germanium spectra. The degree of automation is influenced by
many factors, such as sample load, available equipment, available man-
power, and budgets. Very few laboratories have automated their analyses
to the same degree or in the same way.
Automated analysis can provide reduced costs for analyzing samples,
improvements in analytical precision, improved methods of storing and
handling data, and other related benefits. Reduction in costs for
analyzing samples depends on actual sample load and is often less than
would be expected because initial costs of equipment and system develop-
ment are so high. The cost of training personnel to operate the system
properly is also substantial. One great benefit of a highly automated
system is the improvement in the analytical precision possible if the
system is properly developed and used.
Sophisticated programs for automatic analysis of data can rapidly
and accurately analyze germanium spectra that would be virtually
impossible to analyze manually in an equivalent amount of time. Errors
that occur frequently in manual analysis are virtually eliminated by
computer processing. Data from a fully automated system of analysis can
be output in a form that is readily adaptable to report preparation for
regulatory agencies or other purposes. The data, depending on the
system available, can also be transferred directly into a data bank,
where it is stored for later use in reports covering long periods or for
statistical examination and evaluation of the results of an environmental
monitoring program.
Most programs for automatic analysis of germanium spectra can be
separated into two parts, peak analysis and radionuclide identification.27
5.1.4.1 Peak Analysis
Peak analysis provides for input and reduction of the spectral data
into energies and intensities for each FEP. This may require several
program subdivisions, which will read in the data, smooth the data (Sec-
tion 5.1.1), locate peaks (Section 5.1.2), determine the net counts in
the FEPs (Section 5.1.3), determine the energies of the FEPs (Section
4.4.1), and determine the gamma emission rates for each FEP (Section
4.4.2).*
Transfer of spectral data to the analysis program is usually not a
significant problem. If a large computer is being used for data analysis,
the data may be entered by card or magnetic tape. For card entry, the
format of data can be arranged as desired. Magnetic tape must be for-
matted as the program requires, or a separate program must transfer the
data from the tape to the program, reformatting it in the process.
*The gamma emission rate is calculated by using Equation (8): N = C/e
(Section 4.4.2.4.) Y
-------
-61-
The use of an inhouse minicomputer for spectral analysis usually
may be handled by directly interfacing the spectroscopy system to the
computer and reading data under program control. If direct linkage is
not possible, magnetic or paper tape may be used with the spectral data
being read under program control by the computer. Usually, any process
for transferring data is a simple one for a knowledgeable programmer.
The operations of peak analysis have been fully discussed in
Sections 5.1.1 through 5.1.4. A program that performs all these analy-
tical operations in the proper sequence is usually a quite sophisticated
program, and considerable effort is often spent establishing routine
operation of the program. Several such computer programs are available
and widely used. These programs are referenced in this report in the
particular section (5.1.1 through 5.1.4) in which the various features
of spectral analysis were discussed.
5.1.4.2 Huclide Identification
The stage of nuclide identification is often omitted from analysis
programs.86"*8'94'123 Such programs provide information on only the
energy and intensity of each FEP. Nuclide identification uses the energy
and intensity of the FEP from the first stage of data analysis to deduce
the component radionuclides, determine their activities, evaluate the
error in the determined activities, and output the results in a particular
format. Of course, such an identification package can be added to one
program from another or written by the user to his own specifications.
The first item necessary for nuclide identification is a radionuclide
library containing the most probable spectral components of the sample.
Generally, a subroutine in the main program is used to initiate and to
update this library.27'29 Such a subroutine should produce a file that
can be easily and efficiently searched by other subroutines used in the
identification process. The file may typically contain data on (1) index
position (a number indicating the numerical position of the entry in the
file); (2) symbol of the element; (3) mass number; (4) energy and intensity
of selected gamma rays emitted by the radionuclide; (5) half-life of the
radionuclide; and (6) parent-daughter relationships of the particular
radionuclide. The decay scheme information used in these libraries can
be obtained from nuclear data sheets45 or other compilations.30'146
Several different radionuclide libraries may be prepared and used
for each specific type of sample. (For example, one library may be used
for spectra containing mainly fission products, and another may be used
for spectra containing mainly natural products.) Careful preparation
and use of these libraries can greatly facilitate accurate identification
and quantification. Each library should be prepared carefully, so that
all probable components of the sample are included, but no extraneous
components are present. This minimization of the library size can vastly
speed up identification. Radionuclides can be added and deleted until
optimum libraries are evolved for all sample types.
-------
-62-
Most programs tentatively identify radionuclides on the basis of
matching energies. In other words, any nuclide in the library that has
a gamma-ray energy listed within a specified range of the FEP usually is
selected as a possible candidate for that FEP. Those radionuclides that
have half-lives that are short compared with the sample age can then be
eliminated from consideration, because such radionuclides would have
decayed to a nondetectable level after several (typically six to ten)
half-lives.
The tests based on energy and half-life may be sufficient to identify
the radionuclides within environmental samples with very few components,
and these are the only tests that can be applied to radionuclides with
only one abundant gamma ray. For radionuclides with more than one abun-
dant gamma ray, a test that matches peaks in the unknown spectrum to
possible radionuclides and evaluates the agreement between computed and
reported energy values and intensity ratios for FEPs is often applied.
A close match for energy and intensity between two or more gamma rays of
an individual radionuclide usually is considered a positive identification
except for very complex spectra.
When FEPs of the unknown spectrum have been assigned to the component
radionuclides, the activities of these radionuclides can be calculated.
Many programs simply calculate the activity A of a radionuclide using
the observed gamma-ray emission rate N from a single FEP:
A =
where
P = gamma-ray emission probability per decay (i.e., the
' branching fraction),
f = factor to convert counts to the desired activity unit
(e.g., 3.7 x 10* disintegrations per second per microCurie).
This calculation fails to use a large amount of available data when
a radionuclide emits two or more intense gamma rays. In such cases, the
activity can be better estimated by calculating the final activity of
the particular radionuclide by use of a weighted average of the observed
emission rates from all FEPs belonging to the given radionuclide. Thus,
the expression for the activity A, based on a weighted-average calculation,
would be given by
n n
A = ( I A.W.)/ I W. , (45:
• t 11 _» i 1
where
n = total number of FEPs used in the weighted average,
Ai = activity calculated from equation based on the observed
ratio of gamma emission of the "i"th FEP,
Wi = weighting factor for the "i"th FEP.
-------
-63-
The weighting factors should be chosen so that110
W. = l/o.2 , (46)
where
a. = total uncertainty in the activity if it were calculated
from the "i"th FEP.
The uncertainty OT in the weighted-average activity is given by
or2 =1/1 (l/o.2) , (47)
A i=l x
where all terms are as previously defined.
The total uncertainty o. should combine both the uncertainty in the
observed rate of gamma emission of the sample (i.e., the statistical
uncertainty and uncertainties from the fitting process if a fitted
technique was used) and the uncertainty in the probability of gamma-ray
emission per decay. Often, the uncertainty in the probability of gamma-
ray emission is not known, and this contribution to the total uncertainty
is ignored.
The approach outlined above cannot account for unresolved inter-
ferences between peaks. This problem can be solved by forming a matrix
of linear equations that describe the spectral intensity of each con-
tributor to the FEP. Each peak is assumed to be made up of a linear
combination of gamma rays from one or more components.2 79'147 The
intensity of the "i"th peak I . is given by
m
I. = I e.A.P.. , (48)
x j=1 x J
where
e. = detection efficiency for the "i"th FEP,
A. = activity of the "j"th radionuclide,
P.. = gamma-ray emission probability per decay for the "i"th
1J line of the "j"th radionuclide.
Equations are written in matrix form for each FEP. Interdependent
sets of equations are isolated and solved by the method of least-squares.27
The program GAMANAL, developed by Gunnink and Niday,27 is applicable
to a wide range of samples, some of which have very complex spectra.
GAMANAL uses the most sophisticated methods for radionuclide identifi-
cation and quantification described in the literature. The program
incorporates a series of spectral tests that are used to determine the
radionuclide composition of a sample. Each test assigns a confidence
index to each candidate radionuclide. The most probable component
-------
-64-
radionuclides are chosen to reflect those with the highest final con-
fidence index, which is simply the product of the confidence-level
indices from each individual test. A matrix of linear equations is set
up and solved by the method of least-squares.
Several other computer programs for automatic radionuclide identifi-
cation have been developed for routine use.27'78'79'122'147~153 Some of
these programs are simply modifications or adaptations of other existing
software, whereas others contain entirely new approaches to the problems
of radionuclide identification.
The size and type of computer available for analysis of germanium
spectra greatly influence the method of analysis selected by the user.
Several programs are written in modular form so that they can be adapted
to either a small computer or a sophisticated, programmable calculator.
One approach to providing automatic analysis of germanium spectra
followed in many laboratories is to develop programs for each specific
task. As the work load increases or as the need for more sophisticated
or accurate methods increases, software is deleted from or added to the
existing program structure.
Many laboratories are purchasing germanium spectroscopy systems
that provide a combination of data acquisition and data reduction capa-
bilities. Many of these systems offer the advantage of automatic quantifi-
cation and identification of radionuclides without the necessity of the
user developing his own software. Also, the manufacturers often offer
complete software support, including updates and new features. Many
systems use the principle of multiple core overlay to allow a sophisti-
cated program to be executed with a relatively small computer; however,
before such a system is purchased, the software to be supplied by the
vendor should be carefully evaluated to ensure that it will adequately
meet the needs of the laboratory. In summary, several points should be
considered when the question of germanium spectrum analysis is addressed:
1. Because no fully automatic system exists that can produce accurate,
reliable results without fail, all results must be carefully reviewed
by a competent analyst.
2. Although the selection of an analytical methodology should be a
thorough process based partly on current sample load and costs, a
system should be chosen which will meet the needs of the laboratory
for several years. This selection process must consider that the
initial cost of setting up the system is a major expense, not to be
repeated frequently.
3. The technical knowledge and time required to obtain reliable results
should not be underestimated.
4. If a large computer is used, the cost for processing the spectra
should be considered. There may be a minimum charge for each
computer run so that the cost for processing a single spectrum
would be quite high; however, if many spectra are processed in each
run, the cost per spectra would decrease.
-------
-65-
5.2 LOWER LIMIT OF DETECTION
Detection limits are used to measure the detection capability of a
given measurement process. These limits are determined from considera-
tions of the procedures and equipment used in a specific analytical
technique. For radioanalytical analyses, the detection limits are
becoming increasingly important, because they are often the criteria
used by many regulatory agencies to define minimum acceptable performance
for an analytical procedure.
The confusion over proper mathematical definitions of the many
cerms used to define detection limits has been noted by Currie.154 The
implied statistical significance of these terms has also been a source
of confusion. To avoid any confusion in this discussion, we will adopt
the definition of the detection limit, the lower limit of detection
(LID), given by Harley17:
The smallest amount of sample activity, using a given
measurement process (i.e., chemical procedure and detec-
tor) that will yield a net count for which there is con-
fidence at a predetermined level that activity is present.
This particular definition was chosen because of its clear statistical
meaning and its use by other laboratories and regulatory agencies. The
LID may be approximated as
LID - (Ka + Kp)SQY , (49)
where
K = value of the upper percentile of the standardized normal variate
corresponding to the preselected risk of a type I error (i.e., con-
cluding that activity is present when in fact activity is not present),
Kft = value of the upper percentile of the standardized normal variate
" corresponding to the preselected risk of a type II error (i.e., con-
cluding that activity is not present when in fact it is present),
S = established standard error for the net counting rate of the sample,
$ = factor used to convert count rate due to a given radionuclide
to activity.
If the probability for the type I error is set at the same level as
that for the type II error, if the counting times for background and the
sample are equal, and if the gross activity and the background are
approximately equal, the equation can be written more simply as
LID - 2K142S^1 , (50)
where
K = Kff = Kp ,
S, = background counting rate.
-------
-66-
The probabilities of the type I and type II errors are usually
chosen to be 5 percent each. Therefore, two relationships result:
2KV2 = 4.66 , (;
and
LLD = 4.66VsJ~Y (5;
Although the 5 percent probability level has been widely used,
there is no reason why other probability levels could not be chosen to
establish a particular confidence level. Also, the importance of type I
or type II error can be individually determined. Values of the 2K^2
term for other probability levels are given in the HASL laboratory
manual.17 Other less common values of K and Kg can be found in most
statistics books.
In the discussion of LLD above, the variability in other parameters,
such as detection efficiency and obtaining representative samples, is
neglected, although their relative errors may be comparable to the
variability in the count rates. These other sources of error have been
addressed by Johnston155 and Bowman and Swindle.156
The value of reducing the variability in the counting procedure and
the chemical procedure depends on the precision of the system for sample
measurement.157 Low values for LIDs and low variabilities in the chemical
and counting procedures are not of value if the variability in the
sampling procedure is large.
Methods of reporting results consistent with the HASL definition of
LLD are given by Currie,154 Hartwell,158 and Lochamy.159
Regulatory agencies are beginning to specify the LLD that must be
met for a particular type of analytical result. In specifying these
requirements, the level of confidence and the value of the TX.^2 term
also are being outlined. To attain a given LLD, all terms used to
calculate the LLD must be considered. One term, Y> is given by
y = l/(RVEfDAVt) , (S
where
R = chemical yield or recovery for all steps within the procedure,
V = volume or mass of sample which is used for the analysis,
£ = absolute counting efficiency of the detector for the FEP,
f = number of photons of a given energy emitted per nuclear decay
(i.e., the branching fraction),
D = radioactive decay fraction,
A = unit conversion factor (e.g., 2.22 pCi/dpm),
t = length of the counting interval.
-------
-67-
The radioactive decay factor, which accounts for the decay from the
time of sample collection to the time the sample is counted, is given by
where
D = expt-o.egsc^ - tg)/t1/2] , (54)
t. = time at the midpoint of the counting interval,
t = time of sample collection,
s
tl/2 = nalf~life of tne radionuclide.
The value of the terms R and V are affected by the procedure chosen.
The value of the D term is affected strictly by the promptness with
which a sample is analyzed. This interval should be kept to a minimum
so that radionuclides with short half -lives can be determined accurately;
time, obviously, is not critical for radionuclides with long half-lives.
The counting efficiency and counting rate of the background are determined
by the equipment available.
The combination of procedure and equipment that can yield an accepta-
ble LLD cannot be straightforwardly determined. In practice, assembly
and testing of several trial combinations of equipment and procedures
experimentally may be necessary before an acceptable performance is
achieved.
When more than one radionuclide is present in a sample, the inter-
fering radionuclides will effectively increase the background over large
regions of the gamma spectrum. Therefore, the LLD calculated for a
particular gamma-emitting radionuclide will vary with the concentration
of other radionuclides present in the sample, and because the concentra-
tion of other gamma-emitting radionuclides within the sample are unknown
before analysis, a valid LLD for a given sample cannot be calculated
before analysis.
For comparative purposes only, one can calculate a nominal LLD for
a given radionuclide. However, all conditions under which these LLDs
are determined must be stated so that valid comparisons can be made with
other available data. Tables 9 through 13 list some nominal values for
the LLDs of various radionuclides in several counting geometries. These
LLDs have been calculated under the assumptions that no chemical separa-
tion or concentration is performed on the sample and that the time
between sample collection and counting is one week. The counting rate
of the background was determined by counting blank samples in identical
geometries. The blank samples are described in Table 14, and the detec-
tors and shields used in this determination are described in Table 1.
The counting rate of the background, S, , for detector 1 was
estimated by summing over a region corresponding to 5 keV for FEPs with
energies less than 400 keV and over 6 keV for FEPs with energies greater
than 400 keV. The counting rates of the background for detectors 2 and 3
were estimated by summing over a region corresponding to 3 keV for FEPs
with energies less than 400 keV and over 3.5 keV for FEPs with energies
-------
-68-
TABLE 9. LLD VALUES (pCi/L) FOR 3.5-L GEOMETRY USED FOR WATER
Nuclide
144Ce
51Cr
131j
106Ru
140Ba-Lab
l34Cs
13'Cs
9SZr-Nb
58Co
54Mn
65Zn
60Co
Energy
(keV)
133.5
320.1
364.5
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
Detector
24
32
5.8
20
29
18.0
4.7
3.0
3.8
19
3.4
5.7
6.9
3.1
3.1
3.0
6.1
3.1
3.0
LID values by
1 Detector
12
15
2.9
14
15
9.2
2.3
1.4
1.8
9.4
1.7
2.7
3.3
1.5
1.6
1.7
3.2
1.8
1.7
detector
2 Detector 3
15
19
3.5
16
17
10
2.8
1.7
2.0
11
2.1
3.3
4.0
1.7
1.8
1.8
3.6
2.0
1.9
Considering the variability in the branching fractions, the counting rate
of background, and the efficiency of detection, the overall variability
in the values assigned as LLD is about 10 percent. All LLDs are calculated
for a counting time of 54,000 s (15 h).
JFor the purposes of calculating the LLD values, 140Ba (half-life, 12.8 d)
and its daughter 140La (half-life, 40.2 h) are assumed to be in equilibrium-
-------
-69-
TABLE 10. LLD VALUES (pCi/g) FOR 0.5-L GEOMETRY USED FOR WATER2
Nuclide
144Ce
5lCr
13lj
106Ru
140Ba-Lab
134Cs
137Cs
95Zr-Nb
58Co
S4Mn
65Zn
60Co
Energy
(keV)
133.5
320.1
364.5
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
LLD
Detector 1
60
92
17
65
98
59
16
9.9
12
58
11
18
22
9.3
9.9
10
9.3
11
11
values by detector
Detector 2
27
44
7.9
35
40
38
7.2
4.4
5.2
28
4.9
8.0
10
4.7
4.6
4.6
9.9
4.9
4.8
Detector 3
28
37
7.3
40
37
23
7.0
3-9
4.6
25
4.8
8.0
8.8
4.1
4.2
4.3
7.8
4.4
5.0
Considering the variability in the branching fractions, the counting rate
of background, and the efficiency of detection, the overall variability
in the values assigned as LLD is about 10 percent. All LLDs are calculated
for a counting time of 54,000 s (15 h).
bFor the purposes of calculating the LLD values, 140Ba (half-life, 12.8 d)
and its daughter 140La (half-life, 40.2 h) are assumed to be in equilibrium.
-------
-70-
TABLE 11. LLD VALUES (pCi/g) FOR GEOMETRY USED FOR VEGETATION
Nuclide
144Ce
51Cr
131!
l06Ru
140Ba-Lab
*34CS
137Cs
95Zr-Nb
58Co
S4Mn
6SZn
60Co
Energy
(keV)
133.5
320.1
364.5
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
Detector
0.16
0.34
0.067
0.23
0.37
0.25
0.061
0.039
0.046
0.24
0.045
0.071
0.083
0.039
0.039
0.038
0.080
0.041
0.043
LLD values by
1 Detector
0.090
0.14
0.026
0.15
0.15
0.085
0.028
0.015
0.019
0.091
0.017
0.028
0.034
0.015
0.016
0.015
0.034
0.018
0.019
detector
2 Detector 3
0.090
0.13
0.026
0.14
0.15
0.085
0.028
0.016
0.018
0.095
0.019
0.030
0.034
0.016
0.016
0.017
0.033
0.018
0.019
Considering the variability in the branching fractions, the counting rate
of background, and the efficiency of detection, the overall variability
in the values assigned as LLD is about 10 percent. All LLDs are calculated
for a counting time of 54,000 s (15 h).
For the purposes of calculating the LLD values, 140Ba (half-life, 12.8 d)
and its daughter 140La (half-life, 40.2 h) are assumed to be in equilibrium.
-------
-71-
TABLE 12. LLD VALUES (pCi/g) FOR GEOMETRY USED FOR SOIL
Nuclide
144Ce
siCr
131j
106Ru
140Ba-Lab
134Cs
137Cs
95Zr-Nb
5«Co
54Mn
65Zn
6°Co
Energy
(keV)
133.5
320.1
364.5'
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
LLD
Detector 1
0.042
0.072
0.013
0.049
0.081
0.053
0.014
0.0087
0.010
0.055
0.011
0.017
0.019
0.0087
0.0087
0.0087
0.018
0.0087
0.0093
values by detector
Detector 2
0.027
0.042
0.0070
0.038
0.038
0.023
0.0058
0 . 0040
0.0050
0.0260
0.0047
0.0078
0.0093
0.0039
0.0047
0.0047
0.0086
0.0047
0.0039
Detector 3
0.022
0.032
0.0070
0.037
0.036
0.019
0.0069
0.0031
0.0039
0.022
0.0039
0.0062
0.0078
0.0039
0.0031
0.0039
0.0078
0.0039
0.0039
Considering the variability in the branching fractions, the counting rate
of background, and the efficiency of detection, the overall variability
in the values assigned as LLD is about 10 percent. All LLDs are calculated
for a counting time of 54,000 s (15 h).
bFor the purposes of calculating the LLD values, l40Ba (half-life, 12.8 d)
and its daughter 140La (half-life, 40.2 h) are assumed to be in equilibrium,
-------
-72-
TABLE 13. LLD VALUES (pCi/L) FOR GEOMETRY USED FOR AIR FILTERS'
Nuclide
144Ce
51Cr
131!
l06Ru
140Ba-Lab
134Cs
137Cs
95Zr-Nb
58Co
54Mn
65Zn
60Co
Energy
(keV)
133.5
320.1
364.5
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
Detector
10.8
15.0
2.8
11.0
17.0
9.8
3.2
1.6
2.1
11.0
1.9
3.0
4.0
1.8
1.9
1.7
3.6
1.9
2.0
LLD values by
1 Detector
3.2
6.2
1.1
6.5
7.3
3.9
1.3
0.76
0.92
3.8
0.85
1.3
1.6
0.69
0.78
0.83
1.5
0.89
0.89
detector
2 Detector 3
3.4
7.0
1.4
5.8
7.4
4.1
1.3
0.95
0.91
4.7
0.88
1.4
1.8
0.78
0.70
0.81
1.9
0.99
0.86
Considering the variability in the branching fractions, the counting rate
of background, and the efficiency of detection, the overall variability
in the values assigned as LLD is about 10 percent. All LLDs are calculated
for a counting time of 54,000 s (15 h).
DFor the purposes of calculating the LLD values, 140Ba (half-life, 12.8 d)
and its daughter 140La (half-life, 40.2 h) are assumed to be in equilibrium.
-------
-73-
TABLE 14. PREPARATION OF BACKGROUND STANDARDS
FOR DETERMINATION OF LLD
Geometry
Description
3.5-L for water
Clean, unused Marinelli beakers (3.7-L total
volume) were filled with 3.5 L of distilled
H20. The water was aerated in the Marinelli
beaker for about 2 h with inert N2 gas to
reduce radon content of the distilled H20.
A layer of plastic film was placed over the
mouth of the beaker, and the beaker lid was
installed. The plastic film and lid were
sealed to the beaker with masking tape.
0.5-L for water
0.6-L Marinelli beakers were prepared in a
manner similar to that described above with
0.5 L of distilled, aerated H20.
Soil
600 g of analytical grade salt (NaCl) was weighed
and placed into an unused 0.5-L Marinelli beaker.
The lid of the container was sealed to the sample
container with masking tape.
Vegetation
An unused 0.5-L Marinelli beaker was filled with
ground styrofoam chips. Analytical grade Nad
was added until the final weight was achieved;
the container lid was then sealed in place with
masking tape.
Air filter
A 5-cm-diameter, plastic petri dish with an air
filter inside was sealed closed with adhesive
tape.
-------
-74-
greater than 400 keV. The widths of these regions were selected to
optimize the factor
-------
-75-
SECTION 6
COMPARISON OF GERMANIUM AND NaI(T£) SYSTEMS FOR GAMMA SPECTROSCOPY
6.1 SYSTEM SENSITIVITIES
Two principal gamma-ray detection systems are used to analyze envi-
ronmental samples: germanium detectors and thallium-activated sodium
iodide [Nal(TA)] detectors. In recent years discussion has arisen regard-
ing the superiority of germanium vs. NaI(T£) systems for routine analysis
of gamma rays from low-activity samples: Which system yields the greatest
amount of usable data in the most cost-effective manner? For applications
of these systems to environmental analysis, this question can be approached
with regard to LID, cost, and ability of the system to produce quality data
in a reasonable length of time.
Some common characteristics of germanium (Ge(Li) and HPGe] and NaI(T£)
detector systems are presented in Table 15. Germanium systems are charac-
terized by high cost, excellent resolution, and low detection efficiency,
whereas NaI(T£) systems have relatively poor resolution, high detection
efficiency, and possibly lower cost.
Specific comparisons of germanium and NaI(T£) systems are complicated
by differences in analytical methods used in spectral analysis. For
example, one of the most common methods for NaI(T£) spectral analysis is
the least-squares method used in computer codes such as ALPHA-M.10'160~162
Analytical systems for spectra produced by germanium-based systems
generally use the fitting of a Gaussian or other higher-order function
to the spectral data (Section 5.1).
Methods for determining background also differ for various analytical
techniques, and consequently, comparisons of calculated LLD values for
systems using these different analytical methods must be generic in
nature. Erroneous conclusions can be drawn if LLDs calculated for
specific analytical techniques are applied empirically.
Table 16 is a compilation of LLD values for spectra containing a
single radionuclide (single-nuclide LLDs) calculated by the HASL-300
method for several germanium and NaI(T£) systems.158 Background for the
germanium detectors was estimated by evaluating the background in spectral
regions corresponding to an FEP with the use of procedures outlined in
Table 9 and Section 5.2. The LLD values for NaI(T£) detectors were
estimated from the output of the ALPHA-M program18'163'164 for data
entitled "least-squares" or were calculated by means of hand-determined
estimates of background and efficiency for the data entitled "FEP."
These comparisons are of actual environmental geometries rather than of
the ideal point-source configurations usually cited when such comparisons
are made.1*5
Data in Table 16 indicate that the detection system with the lowest
calculated LLD varies with radionuclide and gamma-ray energy. Although
germanium detectors have an inherently lower total detection efficiency
than do NaI(T£) detectors, LLD values for germanium detectors are not
-------
TABLE 15. CHARACTERISTICS OF GERMANIUM AND NaI(T2) SYSTEMS FOR GAMMA-RAY SPECTROSCOPY
Item
Germanium
NaI(T£)
Equipment
Characteristics of the
detector
Resolution
Efficiency
Operating temperature
Detector material
Capital cost
Detector ,
Electronics
MCA
Peripherals (disks,
tapes, printers, etc.)
Total equipment cost
Operating cost (per year)
Computer cost
Liquid nitrogen
Detector, cryostat, FET preampli-
fier, linear amplifier, high-
voltage power supply, ADC, MCA
(computer or hardware)
~2keV
1-30%
-200°C
Diode fabricated from HPGe ingot or
Ge(Li) ingot
$5,000-$30,000
$4,000
$10,000-$100,000
$2,000-$15,000
$27,000-$132,000
$0-$5,000
~$800
Detector, photomultiplier,
preamplifier, linear ampli-
fier, high-voltage power
supply, ADC, MCA, computer
~55keV
20°C
Sodium iodide with thallium
as impurity
$1,000-$2,000
~$3,000
$10,000
$2,000-$25,000
$16,000-$40,000
$5,000
Efficiency compared with a 76- by 76-mm NaI(T£) crystal, source-to-detector distance of 25 cm.
Assuming one ADC per detector.
May be included in capital cost if a computer-based MCA is purchased.
-------
-77-
TABLE 16. LLD VALUES (pCi/L) FOR 3.5-L GEOMETRY FOR
SAMPLES CONTAINING A SINGLE RADIONUCLIDE
Nuclide
i44Ce
51Cr
131Ru
106Ru
"0Ba
134Cs
137Cs
95Zr-Nb
58Co
54Mn
65Zn
60Co
Energy Germanium
(keV) detector 1
133.5
320.1
364.5
511.8
622.1
537.4
1596.2
604.7
795.8
569.3
661.6
756.0
724.0
765.0
810.6
834.8
1115.5
1173.2
1332.5
88.2
118.0
25.7
73.5
107.0
73.5
19.5
11.0
14.0
69.8
12.5
20.9
25.4
11.4
11.4
11.4
22.4
11.4
11.0
LLD
Germanium
detector 2
44.1
47.8
5.88
51.4
55.1
23.2
5.88
5.14
6.61
34.5
6.25
9.92
12.1
5.51
5.88
5.25
11.8
6.61
6.25
values by detector3
Germanium
detector 3
55.1
58.8
6.98
58.8
62.5
25.4
6.98
6.25
7.35
40.4
7.72
12.1
14.7
6.25
6.61
6.61
13.2
7.35
6.98
10- by 10-cm
NaI(T£)
least-squares
28.7
49.7
6.13
28.3
6.54
6.31
6.88
6.77
6.88
6.21
13.9
6.02
10- by 10-cm
NaI(T£)
FEP
46.2
45.8
9.97
43.9
99.25
33.3
87.3
8.31
10.3
10.2
9.85
10.18
10.24
21.9
10.0
9.26
Counting time of 4,000 s.
-------
-78-
correspondingly higher. Low LLD values for germanium-based systems
result from low background of the crystal over the regions used to
calculate the LLDs. These LLD values are system-specific and not
generally applicable.
Comparison of LLD values is further complicated when LLD values for
samples containing more than one radionuclide (multinuclide LLDs) are
considered. In Table 17 two cases are presented which show the variation
in LLD observed for multinuclide samples counted on germanium and NaI(T£)
systems. One can see that the germanium LLD values for l37Cs and 60Co
(about 20 pCi/L) in a 3.5-L Marinelli geometry were affected to a greater
extent when compared with the single-nuclide case, than those projected
by the NaI(T£) least-squares techniques. A conflicting case is found,
however, when 54Mn and 58Co are prepared in the 3.5-L Marinelli geometry
at about 20 pCi/L. The LLD values for these nuclides undergo a more
substantial increase for the Nal(TJi) least-squares technique than for
the germanium system.
This apparent discrepancy results from the analytical technique
for the two types of data. The least-squares technique for NaI(T£) uses
standard spectra for each nuclide and the background in the fitting pro-
cedure. The entire sample spectrum is evaluated rather than just the
photopeak area. The resultant LLD values for a particular nuclide in a
particular sample are produced from diagonal elements of the inverse
matrix.10'166 The LLD values increase rapidly for the least-squares
technique as concentration increases, but not as rapidly as they would
if simple FEP analysis, as used for the germanium data, were used.
However, the case of 58Co and 54Mn illustrates the weak point of
least-squares analysis. These nuclides have essentially the same gamma-
ray spectrum and, as indicated in Table 17, are so close in energy that
the NaI(T£) detector cannot resolve them. If the two nuclides had other
strong, distinguishing gamma-ray emissions, this analysis would be less
difficult; however, as it is, the resultant LLD values inflate rapidly
as a result of the interference.10 On the other hand, the resolution of
the germanium detector allows the simple FEP fit to produce significantly
lower LLD values.
As noted in Section 5.2, the ultimate lower limit of detection
depends on the entire spectroscopy system—the detector, the counting
system, and the analytical technique used to reduce data. The LLD value
generated for the NaI(T£) system is, in fact, a measure of the LID of
the system for that particular sample, background, and detector. The
LLD values quoted for the germanium system may or may not be routinely
achievable, depending primarily on the utility and accuracy of peak-
locating and peak-fitting programs used for spectral analyses. Typically,
with the relatively simple software for peak analysis supplied with many
minicomputer-based germanium systems, counting times of several times
those used for NaI(T£) systems are required to achieve realistically
comparable LLDs.
-------
-79-
TABLE 17. LID VALUES (pCi/L) FOR SAMPLES CONTAINING
MORE THAN ONE RADIONUCLIDE
LLD values by detector
Energy
(kev)
Multinuclide
germanium
detector 3
Single-nuclide
germanium
detector 3
Single-nuclide
NaI(T£)
detector
Multinuclide
NaI(T2)
detector
20 pCi/L of 137Cs and 60Co added to a 3.5-L Marinelli beaker
661.62
1173.23
1332.52
17.5
13.3
12.4
7.72
7.35
6.98
6.88
6.02
9.14
7.38
20 pCi/L of S4Mn and 58Co added to a 3.5-L Marinelli beaker
810.6
834.8
12.9
12.3
6.61
6.61
6.88
6.21
18.2
14.7
-------
-80-
6.2 SYSTEM COSTS
The cost of germanium and NaI(T£) systems are determined by initial
capital expenses and the costs of routine operation. The cost of
germanium detectors, disregarding electronics and analytical systems,
varies with the size and type of detector. Germanium detectors suitable
for environmental analyses may range in cost from about $6000 for a
small-volume, coaxial Ge(Li) (about 9 percent) detector to more than
$30,000 for a large-volume Ge(Li) or HPGe detector. The typical cost
for a 15 percent Ge(Li) detector is about $12,000, whereas NaI(T£)
detectors and associated photomultipliers range in cost from $1000 to
$2000. The cost of supportive signal-processing electronics is about 50
percent greater for germanium systems than for NaI(T£) systems. This
increased cost results from a need for lower noise and greater elec-
tronic stability in processing output pulses from germanium detectors.
The ADC cost can be considered equal for both spectroscopy systems;
however, for a single detector, an ADC conversion gain of 4096 channels
or greater is required to realize the superior resolution of a germanium
detector fully. The NaI(T£) detector can be used effectively with an
ADC conversion gain of 256 channels. A signal mixer-router (SMR) may be
used with an NaI(T£) system, allowing more than one NaI(T£) detector to
be used with a single ADC. Similarly, several NaI(T£) detectors can be
used with a single MCA using the SMR-ADC-MCA system.
A need for several germanium detectors will require several MCA
systems or a computer-based MCA with 20K (assuming 16-bit words) or more
words of memory. For practical applications for which there are high
sample loads and several germanium detectors, the computer-based MCA
system combines a large analyzer memory with capabilities for reduction
and analysis of data. Under these conditions, the computer-based MCA
appears to be the most cost-effective system for acquisition and
analysis of data for routine analysis using germanium detectors.
Analysis of a large number of NaI(T£)-generated spectra by the
least-squares technique also demands the use of a computer. Inherent in
the cost of any NaI(T£) system that uses spectral analysis by computer
is the cost of transferring the data to the computer. This transfer can
be accomplished by using microprocessors, magnetic tape, disk, teletype,
or card punch. The cost of transfer, by either system peripherals or
manpower, must be reflected in the cost of the system. The cost of the
peripheral systems for NaI(T£) spectroscopy might range from about $2000
to $25,000, depending on the sophistication and capabilities desired. A
typical least-squares analysis of an environmental sample requires
slightly more than one second of central processor time on a large
computing machine. For a typical time-sharing system, this equates to
about 25C per analysis, neglecting overhead expenses.
Another cost that must be considered for both NaI(T£) and germanium
systems that analyze data by computer is the cost of analyst time for
review and validation of results from the computer analysis. Depending
on the sophistication of the program used, this cost may or may not be
negligible. Because no program for analysis of gamma-ray spectra,
regardless of its complexity, is able to handle all data equally well,
elimination of this cost for review seems unlikely. Experience with a
-------
-81-
small, microprocessor-based MCA system for germanium detectors has shown
that up to one hour of analyst time can be required to ensure an accurate
spectral analysis on a complex or low-level spectrum. However, this
amount of time should not be required with the use of a more sophisticated
analysis program. Results from least-squares analysis of NaI(T£) spectra
are generally reliable. Most of the analyst's time is spent ensuring
that conditions are not present that could invalidate the analytical
results produced by the program; for instance, the presence of a nuclide
in the sample spectrum not accounted for by the standards in the program
library could invalidate the results.
Many factors contribute to the cost (time) required for a system to
produce quality data. Some of these factors, such as reliability of
spectral analysis by computer, have already been mentioned. Other
factors such as ease of calibration, stability of the system, and ease
of setting up the system contribute to the overall functioning of the
spectroscopy system. In this regard, germanium and NaI(T£) systems have
characteristic differences affecting their routine use.
Set-up of the system should be carefully considered when choosing a
system for gamma-ray spectroscopy. Implementation of programs, such as
ALPHA-M, for analysis of NaI(T£) spectra, is time-consuming and demands
a good working knowledge of computers and programming. Similarly,
computer-based systems using germanium detectors require some knowledge
of computer operation and programming for efficient use.
Calibration of the system cannot be separated from set-up of the
system because information on calibration is required before one can begin
analyses of sample spectra by computer. Systems based on germanium detec-
tors are fairly easily calibrated for energy and efficiency response by
measuring multinuclide, radioactive standard solutions in those geometries
to be used for sample analysis. Calibration of counting efficiency can be
based on detection efficiency for the FEP (Section 4.4.2). Energy calibra-
tion of a germanium system is easier than calibration of an NaI(T£) system
because of inherently better gain stability and linearity of the electronics
used. Because the location of a peak can be determined with great accuracy
in a germanium spectrum, an NBS-certified, multiline radioactive source is
sufficient for routine energy calibration of germanium systems.
Calibration of NaI(T£) systems can be much more tedious. For least-
squares analysis, a standard containing the nuclide of interest of known
activity must be counted in each sample geometry for each nuclide to
establish a program library of reference spectra. If one considers a
library of only twenty nuclides and five counting geometries, 100 radio-
active standards must be prepared, counted, and evaluated to build the
standard libraries for the least-squares program. Daily energy calibra-
tion [NaI(T£)] is performed with a multiline standard such as 207Bi.9
Variations in system gain must be minimized from run to run; therefore, a
spectral line such as that from 137Cs (661.6 keV) must be correctly
positioned between each analysis.10 Most least-squares programs, such as
ALPHA-M, can then make minor gain adjustments between the spectra in the
standard library and an actual sample spectrum. The relative simplicity
-------
-82-
of calibrating germanium detectors allows flexibility in choosing and
changing counting geometries quickly that is not afforded by the Nal(TjK)
system.
6.3 SYSTEM APPLICABILITY
A measure of the overall accuracy of NaI(T£) and germanium systems
can be determined by comparing results for identical samples. Yule162
presents comparative data on the quantitative and qualitative analyses
of environmental standards for a germanium system using a TPA method to
determine FEP intensity and for a 76- by 76-mm NaI(T£) detector system
using ALPHA-M for spectral analysis. Yule162 and Wrenn et al.167 con-
clude that, for a variety of sample types, the combination of NaI(T£)
detector and least-squares analysis can yield accuracy similar to that
provided by analysis by germanium systems for shorter counting times
than required for germanium analysis.
In Table 18 data are presented for the analytical results from an
NaI(T£) system using ALPHA-M analysis for EPA cross-check samples of
gamma emitters in water. Each sample was counted in a 3.5-L Marinelli
geometry. The counting time was 4000 s for the NaI(T£) system and
14,400 to 54,000 s for the germanium system. As the data indicate, the
two systems generally agree for quantitative assessment of identified
nuclides. Our experience has indicated that less sophisticated programs
of spectral analysis often used on minicomputers require longer counting
times for accurate determinations of low-level activity than do NaI(T£)
least-squares analyses. Other more sophisticated spectral programs may
be able to perform these analyses accurately with shorter counting
times.
Table 19 presents some advantages and disadvantages of germanium
and NaI(T£) systems. Although each detector system has characteristics
that are desirable for environmental analyses, neither detector system
encompasses all the necessary characteristics to enable optimum analysis
of all sample types. A logical approach to environmental analyses seems
to include a program for gamma spectrometry that uses both germanium and
NaI(T£) detectors. Each detector should be used for those sample types
for which it can perform the most complete, accurate, and rapid analyses.
For example, NaI(T£) systems appear to be particularly suited to the
analysis of routine samples that normally contain a known set of radio-
nuclides. If abnormally high activity is detected in the sample by
least-squares analysis or if review of the results of the least-squares
analysis indicates the presence of unknown nuclides or an inaccurate
determination of the concentration of routine nuclides, the sample could
subsequently be counted on a germanium detector for more detailed analysis.
Germanium detectors appear to be well suited for analysis of samples
whose spectra are known to be complex or not routine in nature, such as
soil, uranium mill tailings, or effluents from nuclear power plants.
-------
-83-
TABLE 18. CROSS-CHECK RESULTS (pCi/L) FOR
GAMMA EMITTERS IN WATER
Nuclide
51Cr
60Co
65Zn
106RU
134Cs
137Cs
51Cr
60Co
65Zn
106Ru
l34Cs
137Cs
51Cr
60Co
65Zn
l06Ru
134CS
l37Cs
Activity Nal
reported by EPA3
December 1976
104115
0±15
102+15
99115
93115
101115
February 1977b
202130
45115
37115
151123
76115
39115
April 1977
177127
35115
50115
72115
27115
21115
(TA)-ALPHA-M
(4000 s)
97140
314
100110
83119
10116
10416
193141
4714
3819
140120
7816
3915
144138
3613
5219
65116
2414
2214
Ge(Li)
(54000 s)
146132
010
11016
101115
8113
9713
171155
4719
50115
141148
7219
4118
168128
3314
5018
66124
2314
1912
a
Error limits at the 99 percent confidence level.
February Ge(Li) results are based on a 14,400-s counting time.
-------
-84-
TABLE 19. ADVANTAGES AND DISADVANTAGES OF GERMANIUM AND
NaI(T£) SPECTROMETRY SYSTEMS FOR GAMMA-RAY
SPECTROSCOPY
Advantages of germanium systems
Superior resolution; ~1.6 keV for 661.6-keV (137Cs) gamma ray resulting
in accurate determination of the energy of FEPs.
Ease of calibration and standardization; new or old geometries may be
determined at short notice.
Lack of appreciable problems with gain shift.
Most components of a gamma-ray spectrum can be identified by sight.
Nuclide libraries on the computer can be easily revised and expanded
without additional calibration.
Components of a spectrum not identified by computer can still be
identified by hand calculations.
Background can be estimated from the baseline around the FEP.
Disadvantages of germanium systems
Cost is high.
Small volume and low efficiency of detection.
Large memory required for the MCA.
Reduction of data by computer necessary for complex or low-level
analysis. Many software packages are unable to quantify nuclides
of low activity accurately. Successful operation also requires
personnel to have some knowledge of computers and programming.
Liquid nitrogen temperatures are required for the operation of both
Ge(Li) and HPGe detectors.
Long counting times.
-------
-85-
TABLE 19. (continued)
Advantages of NaI(T£) systems
Large detector volume and high counting efficiency.
Short counting times required for sample analysis with least-squares
techniques, compared with most germanium systems.
Detectors are low in cost.
Operation at room temperature.
Minimal memory required in the MCA.
Estimates of nuclide activity are based on the entire spectrum, not just
a single measurement of the FEP.
Disadvantages of NaI(T£) systems
Poor resolution; typically, 55 keV for 661.6-keV (137Cs) gamma-ray.
Gain shifts with time and temperature.
Complex procedures for calibration and standardization.
Complex spectra require computer analysis. This requires access to a
computer and personnel knowledgeable in computers and programming.
System is extremely sensitive to background fluctuations (primarily from
222Rn and daughters).
Difficulty in separating overlapping or interfering single gamma rays.
-------
-86-
REFERENCES
1. Evans, R. D. The Atomic Nucleus. McGraw-Hill Book Co., New York,
1955. 972 pp.
2. Siegbahn, K., ed. Alpha-, Beta-, and Gamma-Ray Spectroscopy, vols.
I and II. North-Holland Publishing Co., Amsterdam, 1965. 1742 pp.
3. Friedlander, G., J. W. Kennedy, and J. M. Miller. Nuclear and
Radiochemistry, 2nd ed. John Wiley & Sons, Inc., New York, 1964.
585 pp.
4. Harvey, B. G. Introduction to Nuclear Physics and Chemistry, 2nd
ed. Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1969. 463
pp.
5. Price, W. J. Nuclear Radiation Detection, 2nd ed. McGraw-Hill
Co., New York, 1964. 430 pp.
6. Chiang, H. H. Basic Nuclear Electronics. John Wiley & Sons, Inc.,
New York, 1969. 342 pp.
7. ORTEC, Inc. Experiments in Nuclear Science. AN-34. Oak Ridge,
Tennessee, 1971. 129 pp.
8. Adams, F., and R. Dams. Applied Gamma-Ray Spectrometry, 2nd ed.
Pergamon Press, New York, 1970. 751 pp.
9. Ziegler, L. H., and H. H. Hunt. Quality Control for Environmental
Measurements Using Gamma Ray Spectroscopy. EPA-600/7-77-144;
USEPA, Office of Research and Development, Environmental Monitoring
and Support Laboratory, Las Vegas, Nevada, December 1977.
10. Kanipe, L. G., S. K. Scale, and W. S. Liggett. Least-Squares
Resolution of Gamma-Ray Spectra in Environmental Monitoring.
EPA-600/7-77-089; TVA/EP-78/02, Division of Environmental Planning,
Tennessee Valley Authority, Muscle Shoals, Alabama, August 1977.
11. Quittner, P. Gamma-Ray Spectroscopy. Adam Hilger Press, Ltd.,
London, 1972. Ill pp.
12. Cline, J. E. Studies of Detection Efficiencies and Operating
Characteristics of Ge(Li) Detectors. IEEE Trans. Nucl. Sci.,
15(3):198-213, 1968.
13. American National Standards Institute. Calibration and Usage of
Germanium Detectors for Measurement of Gamma-Ray Emission Rates of
Radionuclides. ANSI N717, Draft 3, Rev. 4, May 1977.
14. Savitzky, A., and M.J.E. Golay. Smoothing and Differentiation of
Data by Simplified Least-Squares Procedures. Anal. Chem.,
36(8):1627-1639, 1964.
-------
-87-
15. Institute of Electrical and Electronics Engineers, Inc. An American
National Standard and IEEE Standard Test Procedures for Germanium
Gamma-Ray Detectors. IEEE Std. 325-1971, ANSI N42.8-1972. New
York, 1971. 19 pp.
16. Institute of Electrical and Electronics Engineers, Inc. USA
Standard and IEEE Test Procedure of Semiconductor Radiation Detec-
tors. IEEE Std. 300, USAS N 42.1-1969, New York, 1969. 12 pp.
17. Harley, J. H., ed. HASL-300 Procedures Manual and Supplements 1-4.
Health and Safety Laboratory, U.S. Atomic Energy Commission, New
York, 1972.
18. Altschuler, B., and B. Pasternack. Statistical Measures of the
Lower Limit of Detection of a Radioactivity Counter. Health Phys.,
9:293-298, 1963.
19. Heath, R. L. Scintillation Spectrometry Gamma-Ray Spectrum
Catalogue, vol. 2. U.S. AEC Research and Development Report
IDO-16880-2, Idaho Falls, Idaho, August 1974.
20. Institute of Electrical and Electronics Engineers, Inc. USA
Standard and IEEE Test Procedure for Amplifiers and Preamplifiers
for Semiconductors Radiation Detectors. IEEE Std. 301, USAS
N42.2-1969, New York, 1969. 19 pp.
21. Bartels, 0. G. Notes on Low Radioactivity Lead For Very Low Back-
ground Shields . Health Phys., 28:189-196, 1975.
22. Heath, R. L. Scintillation Spectrometry Gamma-Ray Spectrum
Catalogue, vol. 1. U.S. AEC Research and Development Report
IDO-16880-1. Idaho Falls, Idaho, August 1964.
23. Bennett, C. A., and N. L. Franklin. Statistical Analysis in
Chemistry and the Chemical Industry. John Wiley & Sons, Inc., New
York, 1954. 724 pp.
24. Enrich, N. L. Quality Control and Reliability, 6th ed. Industrial
Press, Inc., New York, 1972. 306 pp.
25. Grant, E. L. Statistical Quality Control, 9th ed. McGraw-Hill
Book, Inc., New York, 1972. 694 pp.
26. Kanipe, L. G. Handbook for Analytical Quality Control in
Radioanalytical Laboratories, EPA-600/7-77-088; E-EP/77-4, Division
of Environmental Planning, Tennessee Valley Authority, Muscle
Shoals, Alabama, August 1977.
27. Gunnink, R., and J. B. Niday. Computerized Quantitative Analysis
by Gamma-Ray Spectrometry, Vol. 1, Description of the GAMANAL
Program. UCRL-51061, Lawrence Livermore Laboratory, Livermore,
California, 1972. 75 pp.
-------
-88-
28. Gleason, G. A New Gain Equation for Multichannel Pulse Analysis.
Radiochem. Radioanal. Letters, 23(4):221-222, 1975.
29. Coursey, B. M. Use of NBS Mixed-Radionuclide Gamma-Ray Standards
for Calibration of Ge(Li) Detectors Used in the Assay of Environmental
Radioactivity. In: Proceedings of Measurements for the Safe Use
of Radiation (NBS Special Publication 456), National Bureau of
Standards, Gaithersberg, Maryland, 1976. pp. 173-179.
30. Heath, R. L. Gamma Energies and Intensities of Radionuclides. In:
Handbook of Chemistry and Physics, 55th ed., R. C. Weast, ed. CRC
Press, Inc., Cleveland, Ohio, 1974.
31- Aubin, G., J. Barrette, G. Lamoureux, and S. Monaro. Calculated
Relative Efficiency for Coaxial and Planar Ge(Li) Detectors. Nucl.
Instrum. Meth., 76:85-92, 1969.
32. Peterman, B. F., S. Hontzeas, and R. G. Rystephanick. Monte Carlo
Calculations of Relative Efficiencies of Ge(Li) Detectors, Nucl.
Instrum. Meth., 104:461-468, 1972.
33. Faria, N. V., and R.J.A. Levesque. Photopeak and Double-Escape
Peak Efficiencies of Germanium Lithium Drift Detectors. Nucl.
Instrum. Meth., 46:325-332, 1967.
34. Gaggero, G. Monte Carlo Calculations for the Photofractions and
Energy Loss Spectra of Ge(Li) Semiconductor Detectors. Nucl.
Instrum. Meth., 94:481-492, 1971.
35. Wairo, K. W., and G. F. Knoll. Monte Carlo Calculation of Ge(Li)
Efficiencies. Nucl. Instrum. Meth., 44:213-219, 1966.
36. Legland, J. Calibration of Gamma-Ray Spectrometers. Nucl. Instrum.
Meth., 112:229-238, 1973.
37. Gunnink, R., and J. B. Niday. A Working Model for Ge(Li) Detector
Counting Efficiencies. ERDA, 48:55-58, 1976.
38. Laichter, Y., A. Notea, and N. H. Shafrir. A General Approach to
the Calibration of a Ge(Li) Gamma Spectrometry System for Measurement
of Environmental and Biological Samples. Nucl. Instrum. Meth.,
113:61-67, 1973.
39. Kahn, B., G. R. Choppin, and J.G.V. Taylor. Users Guide for
Radioactivity Standards, R. H. Herbert, ed. NAS-NS-3115, National
Academy of Sciences, National Research Council, Oak Ridge, Tennessee,
1974. 87 pp.
40. Lowenthal, G. C., and H. A. Wyllie. The Storage of Radioactive
Solutions with Standardized Disintegration Rates. Nucl. Instrum.
Meth., 112:367-371, 1973.
41. Mann, W. B., and H. H. Seliger. Preparation, Maintenance, and
Application of Radioactivity. NBS Circular 594, National Bureau of
Standards, 1958. 47 pp.
-------
-89-
42. International Commission on Radiation Units and Measurements.
Measurement of Low-Level Radioactivity. ICRU Report 22, Washington,
D.C., June 1972. 66 pp.
43. Hardin, J. M. Recommendations for the Storage and the Dilution of
Radioactivity Standard Solutions—Notes on Pipets and Pipetting.
69-TEC-l, U.S. Department of Health, Education, and Welfare Public
Health Service, National Center for Radiological Health, Winchester,
Massachusetts, 1969. 10 pp.
44. Hardin, J. M. Procedures for the Storage and Dilution of
Radioactivity Standard Solutions—Relative Standardization of
Gamma-Emitting Radionuclides. 69-STD-l, U.S. Department of Health,
Education, and Welfare Public Health Service, National Center for
Radiological Health, Winchester, Massachusetts, 1969. 20 pp.
45. Nuclear Data Group, eds. Nuclear Data Sheets. Academic Press,
New York. (Monthly publication.)
46. International Commission on Radiation Units and Measurements.
Certification of Standardization Radioactive Sources. ICRU Report
12, Washington, D.C., September 1968. 10 pp.
47. Hirshfield, A. T. , D. D. Hoppes, and F. J. Schitna. Germanium
Detector Efficiency Calibration with NBS Standards. In: Proceedings
of the ERDA X- and Gamma-Ray Symposium, Ann Arbor, Michigan, 1976.
pp. 90-93.
48. Debertin, K., U. Schotzig, K. F. Walz, and H. M. Weib. Efficiency
Calibration of Semiconductor Spectrometers—Techniques and Accuracies.
In: Proceedings of the ERDA X- and Gamma-Ray Symposium, Ann Arbor,
Michigan, 1976. pp. 59-62.
49. Hansen, J. S., J. C. McGeorge, D. Nix, W. D. Schmidt-Ott, I. Unus,
and R. W. Fink. Accurate Efficiency Calibration and Properties of
Semiconductor Detectors for Low-Energy Photons. Nucl. Instrum.
Meth., 106:365-379, 1973.
50. Aubin, G., J. Barrette, M. Barrette, and S. Monaro. Precision
Measurements of Gamma-Rav Intensities and Energies in the Decay of
lS2gEu> l54Eu> 56CO) 110*^ and i2SSb Nucl. instrum. Meth.,
76:93-99, 1969.
51. Abreu, M. C., A. A. Maio, and M. J. Temes D'Oliveira. A New Approach
to Semiconductor Detector Efficiency Calibration. Nucl. Instrum.
Meth., 123:295-300, 1975.
52. Notea, A., and E. Elias. Low Energy Photons from 152gEu for Ge(Li)
Calibration. Nucl. Instrum. Meth., 86:269-272, 1970.
53. Lavi, N. 75Se as an Efficient Calibration Standard for Ge(Li)
Detectors. Radiochem. Radioanal. Letters, 14(5):385-388, 1973.
-------
-90-
54. Schotzig, U., K. Debertin, and K. F. Walz. Standardization and
Decay Data of 133Ba. (To be published.)
55. Hamilton, J. H., and S. M. Brahamavar. Energy and Intensity
Standardization of Ge(Li) Systems with 110Ag Gamma Rays. In:
Proceedings of the International Conference on Radioactivity in
Nuclear Spectroscopy, Nashville, Tennessee. Gordon and Breach
Publishing Co., New York, 1972. pp. 1323-1328.
56. Lavi, N. 169Yb and llomAg as Efficiency Calibration Standards for
Ge(Li) Detectors. Nucl. Instrum. Meth., 107:197-198, 1973.
57. Gurfinkel, Y., and A. Notea. Use of 133Ba as a Calibration
Standard for Ge(Li) Detectors. Nucl. Instrum. Meth., 57:173-174,
1967.
58. Mukherjee, P., and A. K. Sengupta. l52Eu as a Calibrating Source
for Ge(Li) Detectors. Nucl. Instrum. Meth., 68:165-166, 1969.
59. Nagpal, T. S., and R. E. Gaucher. 75Se as a Calibration Standard
for Ge(Li) Detectors. Nucl. Instrum. Meth., 89:311-313, 1970.
60. Debertin, K., U. Schotzig, and H. A. Weiss. Kalibrierung der
Nachweiswahrscheinlichkeit van Ge(Li) - Spektrometern mit 152Eu.
PTB - Mittelungen, 85:187-195, 1975. (English translation by Ralph
McElroy and Company.)
61. Walford, G., and C. E. Doust. A Method for the Rapid Calibration
of Germanium Spectrometers. Nucl. Instrum. Meth., 62:353-354,
1968.
62. Wallace, G., and G. E. Coote. Efficiency Calibration of Ge(Li)
Detectors Using a Radium Source. Nucl. Instrum. Meth., 74:353-354,
1969.
63. Zobel, V., J. Eberth, U. Eberth, and E. Eube. 226Ra as a
Calibration Standard for Ge(Li) Spectrometers. Nucl. Instrum.
Meth., 141:329-336, 1977.
64. Cavallo, L. M., B. M. Coursey, S. B. Garfinkel, J.M.R. Hutchinson,
and W. B. Mann. Needs for Radioactivity Standards and Measurements
in Different Fields. Nucl. Instrum. Meth., 112:5-18, 1973.
65. International Atomic Energy Agency. International Directory of
Certified Radioactive Materials. IAEA Press, Vienna, Austria, 1975.
102 pp.
66. Freeman, J. M., and J. G. Jenkin. The Accurate Measurement of the
Relative Efficiency of Ge(Li) Gamma-Ray Detectors in the Energy
Range 500-1500 keV. Nucl. Instrum. Meth., 43:269-277, 1966.
67. Paradellis, T., and S. Hontzeas. A Semi-Empirical Efficiency
Equation for Ge(Li) Detectors. Nucl. Instrum. Meth., 73:210-214,
1969.
-------
-91-
68. Mowatt, R. S. A Semi-Empirical Efficiency Curve for a Ge(Li)
Detector in the Energy Range 50-1400 keV. Nucl. Instrum. Meth.,
70:237-244, 1969.
69. Tokcan, G. , and C. R. Cothern. Precise Form of the Efficiency
Equation for a Ge(Li) Detector. Nucl. Instrum. Meth., 64:219-220,
1968.
70. Hajnal, F., and C. Klusek. Semi-Empirical Efficiency Equations for
Ge(Li) Detectors. Nucl. Instrum. Meth., 122:559-565, 1974.
71. Sherman, I. S., and M. G. Strauss. Local Efficiency Variations in
Coaxial Ge(Li) Detectors. Nucl. Instrum. Meth., 103:113-115, 1972.
72. McNelles, L. A., and J. L. Campbell. Absolute Efficiency Calibration
of Coaxial Ge(Li) Detectors for the Energy Range 160-1300 keV.
Nucl. Instrum. Meth., 109:241-251, 1973.
73. Singh, R. Validity of Various Semi-Empirical Formulae and Analytical
Functions for the Efficiency of Ge(Li) Detectors. Nucl. Instrum.
Meth., 136:543-549, 1976.
74. Kane, W. R., and M. A. Mariscotti. An Empirical Method for
Determining the Relative Efficiency of a fie(Li) Gamma-Ray Detector.
Nucl. Instrum. Meth., 56:189-196, 1967.
75. Nyman, B. Techniques for Fast Evaluation of Ge(Li)-Detector Gamma-
Ray Spectra with Small Computers. Nucl. Instrum. Meth., 108:237-241,
1973.
76. Willett, J. B. High Precision Relative Efficiency Curves for
Measurements of Gamma-Ray Relative Intensities and Internal Conversion
Coefficients. In: Proceedings of the International Conference on
Radioactivity in Nuclear Spectroscopy, Nashville, Tennessee.
Gordon and Breach Publishing Co., New York, 1972. pp. 1317-1321.
77. East, L. V. Precision Measurement of Gamma-Rays from 94Nb Decay.
Nucl. Instrum. Meth., 93:193-195, 1971.
78. Zimmer, W. H. Automatic Radioactive Inventory System. IEEE Trans.
Nucl. Sci., 20(1):352-360, 1973.
79. Barnes, V. Gaspan - An Advanced Computer Code for the Analysis of
High Resolution Gamma-Ray Spectra. IEEE Trans. Nucl. Sci., 15(3):
437-454, 1969.
80. Yule, H. P. Mathematical Smoothing of Gamma-Ray Spectra. Nucl.
Instrum. Meth., 54:61-65, 1967.
81. Yule, H. P. Study of Gamma-Ray Spectrum Distortion by Mathematical
Smoothing. Anal. Chem., 44(7):1245-1249, 1972.
82. Inouye, T., T. Harper, and N. C. Rasmussen. Application of Fourier
Transforms to the Analysis of Spectral Data. Nucl. Instrum. Meth.,
67:125-132, 1969.
-------
-92-
83. Inouye, T. Square Root Transform for the Analysis of Quantum
Fluctuations in Spectrum Data. Nucl. Instrum. Meth., 91:581-584,
1971.
84. Tominaga, H., M. Dojyo, and M. Tanaka. Effect of Spectrum Smoothing
on Peak Area Determinations. Nucl. Instrum. Meth., 98:69-76, 1972.
85. Mariscotti, M. A. A Method for Automatic Identification of Peaks
in the Presence of Background and its Application to Spectrum
Analysis. Nucl. Instrum. Meth. 50:309-320, 1967.
86. Routti, J. T., and S. G. Prussin. Photopeak Method for the
Computer Analysis of Gamma-Ray Spectra from Semiconductor Detectors.
Nucl. Instrum. Meth., 72:125-142, 1969.
87. Phillips, G. W., and K. W. Marlow. Automatic Analysis of Gamma-Ray
Spectra from Germanium Detectors. Nucl. Instrum. Meth., 137:525-536,
1976.
88. Phillips, G. W., and K. W. Marlow. Program HYPERMET for Automatic
Analysis of Gamma-Ray Spectra from Germanium Detectors. NRL Memoran-
dum Report 3198, Naval Research Laboratory, Washington, D.C., 1976.
126 pp.
89. Kemper, A., and G. A. vanKempen. A Simple Computer Program for
High Resolution Gamma-Ray Spectra. Radioanal. Chem., 6:461-475,
1970.
90. Wood, R. E., and J. M. Palms. A Gamma-Ray Spectrum Analysis
Technique for Low-Level Environmental Radionuclides. IEEE,
21(l):536-542, 1974.
91. Sasamoto, N. K., and S.-I. Tanka. An Analysis Method of Gamma-Ray
Pulse-Height Distribution Obtained with a Ge(Li) Detector. Nucl.
Instrum. Meth. 125:507-523, 1975.
92. Bowman, W. W. Rapid Analysis of Germanium Spectra. Nucl. Instrum.
Meth., 96:135-136, 1971.
93. Yule, H. P. Data Convolution and Peak Location, Peak Area, and
Peak Energy Measurements in Scintillation Spectrometry. Anal.
Chem., 38(1):103-105, 1966.
94. Phillips, G. W., and K. W. Marlow. Peak Search and Analysis of
Gamma-Ray Spectra With Low Statistics. IEEE Trans. Nucl. Sci.,
24(1):154-157, 1977.
95. Robertson, A., W. V. Prestwich, and T. J. Kennett. An Automatic
Peak-Extraction Technique. Nucl. Instrum. Meth., 100:317-324, 1972.
96. Slavic, I. A. Automatic Analyses of Gamma-Ray Spectra. Nucl.
Instrum. Meth., 112:253-260, 1973.
-------
-93-
97. Blok, H. P., J. C. DeLange, and J. W. Schotman. A New Peak Search
Method for an Automatic Spectrum Analysis Program. Nucl. Instrum.
Meth., 128:545-556, 1975.
98. Baedecker, P. A. Digital Methods of Photopeak Integration in Acti-
vation Analysis. Anal. Chem., 43(3):405-410, 1971.
99. Hertogen, J., J. DeDonder, and R. Gijbels. Experimental Data on
Photopeak Integration Methods in Activation Analysis. Nucl. Instrum.
Meth., 115:197-212, 1974.
100. McNelles, L. A., and J. L. Campbell. Analytic Approximations to
Peak Shapes Produced by Ge(Li) and Si(Li) Spectrometer. Nucl.
Instrum. Meth., 127:73-81, 1975.
101. Takeda, Y., M. Kitamura, K. Kawase, and K. Sugiyama. A Study of
Effects of Fitting Functions on Results of Gamma Ray Peak Analysis.
Nucl. Instrum. Meth., 136:369-371, 1976.
102. Kokta, L. Determination of Peak Area. Nucl. Instrum. Meth.,
112:245-251, 1973.
103. Covell, D. F. Determination of Gamma-Ray Abundance Directly from
the Total Absorption Peak. Anal. Chem., 31(11):1785-1790, 1959.
104. Sterlinski, S. Pulse Height Analyzer on Gamma-Ray Total Absorption
Peaks in Activation Analysis. Anal. Chem., 40(13):1995-1998, 1968.
105. Quittner, P. Peak Area Determination for Ge(Li) Detector Data.
Nucl. Instrum. Meth., 76:115-124, 1969.
106. Quittner, P. Precise Peak Area Determination for Ge(Li) Detectors.
Anal. Chem., 41(11):1504-1506, 1969.
107. Heydorn, K. and W. Lada. Peak Boundary Selection in Photopeak
Integration by the Method of Covell. Anal. Chem., 14(44):2313-2317,
1972.
108. Ralston, H. R., and G. E. Wilcox. Modern Trends in Activation
Analysis, J. R. DeVoe, ed. NBS Special Publication 312, National
Bureau of Standards, Washington, D.C., 1969. pp. 1238-1243.
109. Von Meerwall, E., and M. D. Gawlik. Automatic Peak Analysis on
Minicomputers. Comp. Phys. Comm., 5:309-313, 1973.
110. Bevington, P. R. Data Reduction and Error Analysis for the Physi-
cal Sciences. McGraw-Hill, Inc., New York, 1969. 336 pp.
111. Rester, A. C., A. V. Ramayya, and J. H. Hamilton. Economical
Programs for Computer Analysis of Gamma-Ray Data. In: Proceedings
of the International Conference on Radioactivity in Nuclear
Spectroscopy, Nashville, Tennessee. Gordon and Breach Publishing
Co., New York, 1972. pp. 1307-1315.
-------
-94-
112. Abondanno, U., A. Boiti, and F. Demanins. The Application of a
Non-Iterative Method for Fitting a Gaussian to Peaks from Ge(Li)
and Si Detectors. Nucl. Instrum. Meth., 142:605-607, 1977.
113. Kitamura, M., Y. Takeda, K. Kawase, and K. Sugiyama. A Compact
Fitting Algorithm for Accurate Analysis of Gamma Spectrum. Nucl.
Instrum. Meth., 136:363-367, 1976.
114. Pratt, T.A.E.C., and M. L. Luther. A Study of Germanium-Lithium
Drifted Gamma Spectrum Shapes. Nucl. Instrum. Meth., 92:317-323,
1971.
115. DeCastro Favia, N. V., J. P. Martin, and R.J.H. Levesque. A
Computer Code for Calculation of Absolute Area of Overlapping Peaks
in Energy Spectrum. Nucl. Instrum. Meth., 85:49-52, 1970.
116. Huang, F.C.P., C. H. Osman, and T. R. Ophel. Line Shape Analysis
of Spectra Obtained with Ge(Li) Detectors. Nucl. Instrum. Meth.,
68:141-148, 1969.
117. Robinson, D. C. A Computer Programme for the Determination of
Accurate Gamma-Ray Line Intensities from Germanium Spectra. Nucl.
Instrum. Meth., 78:120-124, 1970.
118. Mukoyama, T. Fitting of Gaussian to Peaks by Non-Iterative Method.
Nucl. Instrum. Meth., 125:289-291, 1975.
119. Seybold, G. D. GSSLRN-I, An Automatic Least-Squares Computer Code
for the Analysis of Photopeak Spectra. NEWL-1227, 1969.
120. Slavic, I. A., and S. P. Bingulac. A Simple Method for Full
Automatic Gamma-Ray Spectra Analysis. IBK-942, 1970.
121. Lederer, C. M. Computer Analysis of Spectra. In: Proceedings of
the International Conference on Radioactivity in Nuclear Spectroscopy,
Nashville, Tennessee. Gordon and Breach Publishing Co., New York,
1972. pp. 73-105.
122. Roscoe, B. A., and A. K. Furr. Algorithm for Computer Processing
and Analysis of Gamma-Ray Spectra for Large-Scale Laboratory Opera-
tions. Nucl. Instrum. Meth., 137:173-178, 1976.
123. Mills, S. J. A Computer Program for the Automatic Analysis of
Gamma-Ray Spectra. Nucl. Instrum. Meth., 81:217-219, 1970.
124. Dojo, M. A Shape Function of Photopeaks for Gamma-Ray Spectrum
Analyses with Ge(Li) Detectors. Nucl. Instrum. Meth., 115:425-429,
1974.
125. Sanders, W. M., and D. M. Holm. An Analytical Method for Unfolding
Gamma-Ray Spectra. LA-4030, Los Alamos Scientific Laboratory, Los
Alamos, New Mexico, 1969.
-------
-95-
126. Kern, J. Computer Analysis of Nuclear Spectra and Gamma Energy
Standards. Nucl. Instrum. Meth., 79:233-239, 1970.
127. Varnell, L., and J. Trischuk. A Peak-Fitting and Calibration
Program for Ge(Li) Detectors. Nucl. Instrum. Meth., 76:109-114,
1969.
128. Draper, N. R., and H. Smith. Applied Regression Analysis. John
Wiley & Sons, Inc., New York, 1966. 407 pp.
129. Booth, G. W. , and T. I. Peterson. Nonlinear Estimation. IBM Share
Program No. 687, Schenectady, New York, 1958.
130. Crosbie, E. A., and J. E. Monahan. A General Least-Squares Fitting
Procedure. AN E208, Argonne National Laboratory, Chicago, 1959.
131. Moore, R. H., and R. K. Zeigler. Report LA-2367, Los Alamos
Scientific Laboratory, 1960 (unpublished).
132. Hartley, H. 0. The Modified Gauss-Newton Method for the Fitting of
Nonlinear Regression Functions by Least-Squares. Technometrics,
3(2):269-280, 1961.
133. Marquardt, D. W. Solution of Nonlinear Chemical Engineering Models.
Chem. Eng. Progr., 55(6)-.65-70, 1959.
134. Spang, H. A. J. Soc. Ind. Appl. Math., Review 4, 1962. pp. 343-365.
135. Marquardt, D. W., and T. Baumeister. Least-Squares Estimation of
Nonlinear Parameters. IBM SHARE 1428, Schenectady, New York, 1962.
136. Marquardt, D. W. An Algorithm for Least-Squares Estimation of
Nonlinear Parameters. J. Soc. Ind. Appl. Math., 11(2):431-441,
1963.
137. Davidon, W. D. Variable Metric Method for Minimization. ANL-5990,
Argonne National Laboratory, Chicago, 1969 (unpublished).
138. Beales, E. VARMIT. Lawrence Radiation Laboratory Memo, BKY-ZO,
Lawrence Livermore Laboratory, Livermore, California, 1966
(unpublished).
139. Garbow, B. S. Variable Metric Minimization. ANL-Z013, Argonne
National Laboratory, Chicago, Aug. 9, 1965.
140. Davidon, W. C. Variable Metric Method for Minimization, rev. 2.
ANL-3990, Argonne National Laboratory, Chicago, February 1966.
141. Fletcher, R., and M.J.D. Powell. Computer J., 6:163, 1963.
142. Fletcher, R. Computer J., 13:317, 1970. (FORTRAN program included
as an appendix to ref. 10.)
143. Fletcher, R. Program VMM01. Appendix to United Kingdom Atomic
Energy Agency Research Group Report TP456, 1971.
-------
-96-
144. Powell, M.J.D. Computer J., 7:303, 1965.
145. Powell, M.J.D. Computer J., 7:155, 1964.
146. Erdtmann, G., and W. Soyka. The Gamma-Rays of the Radionuclides.
JUL-1003-AC, Nuclear Research Establishment (KFA)--Julich, West
Germany, 1973.
147. Pagden, I.M.H., G. J. Pearson, and V. N. Beck. A Semiautomated
Computer System for Gamma-Ray Spectrum Analysis of Thermal
Neutron-Activated Samples. IEEE Trans. Nucl. Sci., 17(1):211-217,
1970.
148. Philippot, J. C. Automatic Processing of Diode Spectrometry Results.
IEEE Trans. Nucl. Sci., 17(3):466-488, 1970.
149. Op DeBeeck, J. P. A Program for Complete Quantitative Neutron
Activation Analysis Using Ge(Li) Detector Gamma Spectrometry and a
Small Computer. Radioanal. Chem., 11:283-303, 1972.
150. Yule, H. P., J.P.F. Lambert, S. K. Goranson. Automatic Activation
Analysis Code for Component Identification and Quantitative Results
Computation. ANS Trans., 17:128-129, 1973.
151. Birch, M. L., R. A. Hearn, and D. M. Walker. The Use of On-Line
Programmable Calculators in Nuclear Power Plant Laboratories. IEEE
Trans. Nucl. Sci., 22:711-717, 1975.
152. Zimmer, W. H. Gamma, X-Ray Data Reduction System. In: Proceedings
of the ERDA X- and Gamma-Ray Symposium, Ann Arbor, Michigan, 1976.
pp. 86-89.
153. Fanshier, D., G. Goughran, D. Hoppe, R. Linsey, and D. Samsky. A
General Purpose Disk Based Distributed Processing Data Acquisition
and Analysis System. IEEE Trans. Nucl. Sci., 24(1):476-480, 1977.
154. Currie, L. A. Limits for Qualitative Detection and Quantitative
Determination. Anal. Chem., 40(3):586-593, 1968.
155. Johnson, J. W. An Index of Adequacy and Some Other Statistical
Considerations for an Environmental Monitoring System. BNWL-B-250
Battelle, Pacific Northwest Laboratories, Richland, WA, February
1973.
156. Bowman, II, W. B., and D. L. Swindle. Procedure Optimization and
Error Analysis for Several Laboratory Routines in Determining
Specific Activity. Health Phys., 31:355-361, 1976.
157. Forley, J. P., et al. A Guide for Environmental Radiological
Surveillance at ERDA Installations. ERDA 77-24, Energy Research
and Development Administration, Washington, D.C., 1977.
158. Hartwell, J. K. Detection Limits for Radioisotopic Counting
Techniques. ARH-2537, Atlantic Richfield Hanford Company, Richland,
Washington, 1972. 16 pp.
-------
-97-
159. Lochamy, J. C. The Minimum Detectable Activity Concept. In:
Processing of Measurements for the Safe Use of Radiation. NBS
Special Publication 456, National Bureau of Standards, Gaithersberg,
Maryland, 1976. pp. 169-172.
160. Schonfeld, E. ALPHA - A Computer Program for the Determination of
Radioisotopes by Least-Squares Resolution of the Gamma-Ray Spectra.
Nucl. Instrum. Meth., 42:213-218, 1966.
161. Schonfeld, E. ALPHA-M - An Improved Computer Program for
Determining Radioisotopes by Least-Squares Resolution of the Gamma-
Ray Spectra. Nucl. Instrum. Meth., 52:177-180, 1967.
162. Yule, H. P. On the Analysis of Low Radioactivity Level Environmental
Samples. In: Proceedings of ERDA Symposium on X- and Gamma-Ray
Sources and Applications, Ann Arbor, Michigan, 1976. pp. 198-202.
163. Pasternack, B., and N. H. Harley. Detection Limits for
Radionuclides in the Analysis of Multicomponent Gamma-Ray
Spectrometer Data. Nucl. Instrum. Meth., 91:533-540, 1971.
164. Fisenne, I. M., A. O'Toole, and R. Cutler. Least-Squares Analysis
and Minimum Detection Levels Applied to Multicomponent Alpha
Emitting Samples. Radiochem. Radioanal. Letters, 16:5-16, 1973.
165. Robertson, R., N. M. Spyrou, and T. J. Kennett. Low-Level Gamma-
Ray Spectrometry: NaI(T£) vs. Ge(Li). Anal. Chem., 47(1):65-70,
1975.
166. Stevenson, P. C. Processing of Counting Data. NAS-NS 3119,
National Academy of Sciences - National Research Council Publication,
Washington, D.C., May 1966. 167 pp.
167. Wrenn, M. E., S. M. Jinks, L. M. Hairr, A. S. Pashoa, and
J. W. Lentsch. Natural Activity in Hudson River Estuary Samples
and Their Influence on the Detection Limits for Gamma Emitting
Radionuclides Using Nal Gamma Spectrometry. In: The Natural
Radiation Environment II, CONF-720805-P2, ed. Adams et al.,
Houston, Texas, 1972.
-------
-99-
APPENDIX A
GERMANIUM DETECTOR LEAKAGE CURRENT TEST
Determination of the detector leakage current is a useful periodic
maintenance check that can be performed by either maintenance personnel
or the analyst. This check, which is quite sensitive, can be used to spot
any developing trend toward loss of detector resolution. An abnormally
high leakage current degrades the preamplifier's ability to differentiate
the exact changes caused by gamma-rays impinging on the crystal. This, in
turn, causes loss of resolution.
Most high-resolution germanium detector preamplifiers contain a test
point to which a probe from a digital voltmeter can be attached. The
initial measured voltage should be taken with a small amount of bias
(y>00 V) on the crystal. This value should be negative in the -0.1- to
-1.5-V range. As bias is increased toward the operating bias, the measured
voltage should spike sharply negative and then recover to a reading close
to the initial value. Subsequent increases up to the detector operating
voltage will yield a similar result, with the final measured voltage
slightly more negative than the initial reading. The leakage current is
then determined by the difference in the initial and final measured volt-
ages divided by the value of the feedback resistor.
As an example, assume that the measured voltage at the 500-V bias is
-0.23 V and that the measured voltage at the operating bias is -0.43 V;
then, AV = 0.20 V. Then, from E = iR, the leakage current for a feedback
resistor value of 5 x 10® ohms would be
0.20 V
5 x 209 ohms = 40 pA.
The specification sheet accompanying a new germanium detector should
state a maximum value for the leakage current. The manufacturer can
supply the nominal value along with the value of the feedback resistor.
When a new detector is received, the results of this test should be
checked and recorded when making other pre-setup tests. As values are
obtained periodically, they can be compared with the initial value, and
possible trends toward detector deterioration might be observed on a
more timely basis.
-------
-101-
APPENDIX B
PROCEDURES FOR QUALITY CONTROL OF GERMANIUM DETECTOR SYSTEMS
These procedures are generally acceptable for achieving quality con-
trol of germanium-based gamma-ray spectroscopy systems. Equivalent proce-
dures may be used to achieve the same control of a system; in other words,
these procedures point out the areas of concern and an approach to
monitoring them.
-------
-102-
B.l ENERGY CALIBRATION AND COUNT REPRODUCIBILITY CHECK
B.I.1 Purpose
To provide instructions for determining the energy calibration and
checking the count reproducibility of germanium spectroscopy systems.
Frequent determinations of the energy calibration allow a check of the
electronic stability of the system with time. Also, a precise energy
calibration is necessary for qualitative identification of gamma rays in
a sample spectrum. Count rate reproducibility check is valuable in
monitoring the stability of the counting efficiency of a detector with
time.
B.I.2 Scope
These instructions are applicable to all gamma spectroscopy systems
using germanium detectors.
B.I.3 References
None.
B.1.4 Abbreviations and Definitions*
germanium spectroscopy system: A gamma-ray spectroscopy system composed
of an HPGe or a Ge(Li) detector, an amplifier, an ADC, an MCA, and
a lead or steel shield.
slope: The slope of the line generated by plotting energy vs. channel
number.
offset: The intercept on the energy axis of the line generated by
plotting energy vs. channel number.
energy calibration: The process of determining the linear relationship
between energy and channel number.
count reproducibility: The ability of a germanium spectroscopy system
under fixed conditions of source-to-detector geometry and fixed
counting time to reproduce a given number of counts from a known
radioactive source within the limitations of Gaussian statistics.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-103-
B.1.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analyst and reviewing the
entries in the logbooks periodically.
2. The analyst is responsible for calibrating all germanium spectroscopy
systems each working day by following this procedure. The results
of this check must be recorded in the appropriate logbook, and any
abnormal results must be reported to the supervisor of the counting
room. These data should be periodically reviewed for trends.
B.I.6 Procedure
B.I.6.1 Apparatus
Check source containing 109Cd and 60Co mounted on a source holder.
B.I.6.2 Instructions
1. Establish the settings of the instrument and experiment parameters
required for normal operation of the spectroscopy system as listed
in the logbook of the system.
2. Position the source and source holder on the endcap of the ger-
manium detector.
3. Accumulate a spectrum for the time specified in the logbook and
perform an energy calibration using the energy calibrate feature of
the MCA, or calculate the slope and offset by using the equations
in Section B.I.6.3 of this procedure.
4. If the slope is within the range of 0.999 to 1.001 for a 2048-channel
spectrum or 0.499 to 0.501 for a 4096-channel spectrum and if the
offset is within the range -2 to +2, omit instructions 5 through 7,
and continue with instructions 8 and 9. If the slope or offset is
not within the appropriate ranges specified above, proceed to step
5.
5. Using the zero adjustment of the system's ADC, set the peak channel
of the 88.036-keV gamma ray of 109Cd to channel 88 for a 2048-channel
spectrum or channel 176 for a 4096-channel spectrum. This can be
accomplished by visually inspecting the peak after accumulating a
spectrum for a short period of time.
6. Adjust the fine gain on the amplifier so that the peak channel of
the 1332.483-keV gamma ray of *°Co is in channel 1332.5 for a
2048-channel spectrum or in channel 2665 for a 4096-channel spectrum.
7. Repeat instructions 5 and 6 until the peaks of the gamma lines are
positioned to within ±1 channel of the desired positions, and then
repeat instructions 3 and 4.
-------
-104-
8. Determine the areas under the 88.0-keV and 1332.5-keV peaks by
using a peak-search routine or manual calculation.
9. Record the fine gain, zero-level setting, the slope and offset of
the energy calibration, and the areas under the 88.0-keV and
1332.5-keV peaks in the logbook. Plot the areas under the 88.0-keV
and 1332.5-keV peaks on the appropriate control chart. Date and
initial the entries, and report any abnormal settings or areas to
the supervisor of the counting room.
B.I.6.3 Calculations for Energy Calibration
The slope m and offset a of the energy calibration can be calculated by
1332.483 - 88.036 keV _ 1244.457 keV (B.I)
and
C1332 " C88 Ch C1332 " C88 Ch
a = 88.036 keV mCgg , (B.2)
where
C_» = centroid or channel with the highest number of counts
of the 88.036-keV, full-energy peak for 109Cd,
C,__- = centroid or channel with the highest number of counts
of the 1332.483-keV, full-energy peak for 60Co.
-------
-105-
B.2 BACKGROUND CHECK FOR GERMANIUM DETECTOR SYSTEM
B.2.1 Purpose
To provide instructions for determining the background of a germanium
detector system. This procedure is necessary to monitor the fluctuations
of background (e.g., noise) that directly affect the sensitivity of
analyses obtainable with a germanium detector system.
B.2.2 Scope
These instructions are applicable to all germanium detector systems.
B.2.3 References
None.
B.2.4 Abbreviations and Definitions*
background: Any pulses stored by the germanium spectroscopy system in
the absence of a radioactive sample.
germanium detector system: A system composed of an HPGe or Ge(Li)
detector and a lead or steel shield.
germanium spectroscopy system: A system used for gamma-ray spectroscopy
composed of an HPGe or Ge(Li) detector, an amplifier, an ADC, an
MCA, and a lead or steel shield.
B.2.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that the above procedure is followed by the analyst and reviewing
the entries in the logbook periodically.
2. Each week an 8-h (28,800-s) background count should be run on all
germanium detectors. The results of these checks should be recorded
in an appropriate logbook, and any abnormal results should be
reported to the supervisor of the counting room. These results
should be reviewed periodically for trends.
B.2.6 Procedure
B.2.6.1 Apparatus
Germanium spectroscopy system.
*A11 definitions are operational definitions and are not necessarily generally
applicable.
-------
-106-
B.2.6.2 Instructions
1. Establish the settings of the instrument and experiment parameters
used during normal operation of the spectroscopy system as listed
in the logbook of the system.
2. Remove all sources and blanks from the shield around the detector
and close the shield.
3. Accumulate an 8-h (28,800-s) background spectrum. Save the
spectrum on a system medium (disc or tape), or print out a hard
copy of the spectrum.
4. Determine the nuclides present in the spectrum, report any back-
ground activities in the logbook, report any abnormal background
activities to the appropriate supervisory personnel, and file a
copy of those nuclides found in the spectrum in the appropriate
file. (For computer-based systems, this would entail use of the
nuclide report option for the system.)
5. Determine the total counts in the background for channels corre-
sponding to the energy ranges of 50 to 1750 keV, 50 to 150 keV, 360
to 459 keV, 700 to 799 keV, and 1300 to 1399 keV. (These regions
were chosen to obtain an estimate of the total background and the
background in regions where gamma rays from naturally occurring
radionuclides are not found.)
6. Record the sums of the counts in the specified energy ranges in
the logbook, plot selected regions on control charts, date and
initial the entries, and report any abnormal results to the
supervisor of the counting room.
-------
-107-
B.3 CHECK FOR FULL-ENERGY PEAK RESOLUTION AND PEAK-TO-COMPTON RATIO
B.3.1 Purpose
To provide instructions for determining the FEP resolution and peak-
to-Compton ratio of a germanium detector. This procedure is necessary to
ensure that the ability of a detector to discriminate photopeaks has not
degraded with time.
B.3.2 Scope
These instructions are applicable to all gamma spectroscopy systems
using germanium detectors.
B.3.3 Reference
1. Institute of Electrical and Electronics Engineers, Inc., Test
Procedures for Germanium Gamma-Ray Detectors. IEEE Std. 325-1971.
(ANSI N42.8 - 1972).
B.3.4 Abbreviations and Definitions*
full-energy peak (FEP): The peak in a spectrum due to complete inter-
action of a photon within the detector and complete collection of
energy from this interaction.
resolution: Energy resolution in this context is a measure of the
ability of a detector to separate closely spaced peaks.
full width at half maximum (FWHM): The full width of the peak at half
the maximum magnitude of the peak.
peak-to-Compton ratio: The ratio of the maximum number of counts in the
peak to the average of the flat portion of the Compton continuum.
The flat portion of the Compton continuum is the 1040- to 1096-keV
region for the 1332.52-keV gamma ray of 60Co.
germanium spectroscopy system: A system used for gamma-ray spectroscopy
composed of an HPGe or Ge(Li) detector, an amplifier, an ADC, an
MCA, and a lead or steel shield.
B.3.5. Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analyst and reviewing the
entries in the logbooks periodically.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-108-
During the last week of each month the analyst shall determine the
FEP resolution and the peak-to-Compton ratio for all germanium
detectors. The results of these checks must be entered in the
appropriate logbook, and any abnormal results should be reported to
the supervisor of the counting room.
B.3.6. Procedure
B.3.6.1 Apparatus
1. Germanium spectroscopy system.
2. 57Co and 60Co point sources.
B.3.6.2 Instructions
1. Establish the settings of the instrument of the germanium spec-
troscopy system for the resolution check at 122.1 keV. (This might
entail setting up the system for collection of a 4096-channel
spectrum, calibrated so that each channel corresponds to about
0.2 keV.)
2. Center a 57Co source on the endcap of the detector, and establish
count rate for the full spectrum of 500 to 1500 counts per second
by adjusting the vertical source-to-detector distance.
3. Acquire a spectrum with about 3000 counts in the peak channel for
the gamma ray at 122.1 keV.
4. Perform an energy calibration using the 122.060- and 136.47-keV
gamma rays.
5. Determine the FWHM of the 122.1-keV peak. This is accomplished by
locating the centroid of the 122.1-keV peak and determining its
magnitude, N (in counts). The magnitude of other channels in the
122-keV peak is determined until one channel on each side of the
peak centroid is found that has one-half the magnitude (in counts)
of the centroid channel. (Some interpolation of channel values is
generally necessary to define the FWHM of the peak.) The difference
in energy (keV) between the two channels defining the FWHM is
considered to be the resolution of the detector. Simple calculator
or computer programs may be developed to facilitate calculations of
the FWHM for computer-based spectrometry systems. (For quality
control procedures, the background need not be subtracted, because
the procedure is used to detect changes in resolution.)
6. Establish the settings of the instrument of the germanium spec-
troscopy system for the check of resolution at 1332.5 keV and
peak-to-Compton ratio. (Typically, this might entail setting up
the system for collection of a 4096-channel spectrum, calibrated so
that each channel corresponds to about 0.3 keV.)
-------
_109-
7. Center the 60Co source on the endcap of the detector, and establish
a counting rate for the full spectrum of 500 to 1500 counts per
second by adjusting the vertical source-to-detector position.
8 Accumulate a spectrum with more than 2000 counts in the peak channel
of the 1332-keV line.
9. Perform an energy calibration using the 1173.226- and 1332.483-keV
gamma lines of °°Co.
10. Calculate the FWHM of the 1332.5-keV peak by the same method
described in step 5.
11. Sum the counts found in the region 1040 to 1096 keV. Divide by the
number of channels to obtain an average count per channel.
12. Calculate the peak-to-Compton ratio by dividing the peak channel
counts in the 1332.5-keV peak by the average count per channel
found in step 11.
13. Compare the calculated peak-to-Compton ratio with the original
detector specification or the value last determined to see whether
change has occurred.
-------
-110-
B.4 CALIBRATION OF EFFICIENCY
B.4.1 Purpose
To provide instructions for determining the counting efficiency in
various counting geometries used with germanium spectroscopy systems.
B.4.2 Scope
These instructions are applicable to all gamma spectroscopy systems
using germanium detectors.
B.4.3 References
1. American National Standards Institute, Inc., Calibration and Usage
of Germanium Detectors for Measurement of Gamma-Ray Emission Rates
of Radionuclides. ANSI Guide N717 Draft 3, Revision 4, May 1977.
2. Coursey, B. M. Use of NBS Mixed Radionuclide Gamma-Ray Standards
for Calibration of Ge(Li) Detectors Used in the Assay of Environmental
Radioactivity. In: Measurements for the Safe Use of Radiation.
National Bureau of Standards, Gaithersburg, Maryland, 1976. pp.
173-179.
B.4.4 Abbreviations and Definitions*
germanium spectroscopy system: A system used for gamma-ray spectroscopy
composed of an HPGe or Ge(Li) detector, an amplifier, an ADC, an
MCA, and a lead or steel shield.
full-energy peak (FEP): The peak in a spectrum due to complete inter-
action of a photon within the detector and complete collection of
energy from this interaction.
counting efficiency: That part or fraction of the total number of gamma
rays emitted by a radioactive source that is detected and registered
in the FEP by the germanium spectroscopy system with a given counting
geometry. May be expressed as a fraction or percentage.
counting geometry: The conditions of space and volume in which a sample
is counted using a germanium spectroscopy system.
B.4.5 Responsibilities
1. The counting efficiency for all systems for all counting geometries
used by the laboratory for gamma-ray spectroscopy using germanium
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-Ill-
detectors should be determined yearly, when changes occur in count
rate reproducibility (see Section B.I), or when a new germanium
detector is received. After completion of calibration, pertinent
information (Section B.4.6.2.10) should be recorded in an
appropriate logbook and approved by the supervisor of the counting
room.
B.4.6 Procedure
B.4.6.1 Apparatus
1. Sources containing a mixture of radionuclides prepared in the
various counting geometries. When counted, the counting rate for
the total spectrum of these standards should be less than 1000
counts per second.
2. Germanium spectroscopy system.
B.4.6.2 Instructions
1. Establish the settings of the instruments as used for routine
operation of the germanium spectroscopy system.
2. Position the source on the detector endcap, and accumulate a
spectrum with at least 20,000 counts in all peaks to be used in
determining efficiency.
3. Determine the net areas of peaks to be used in the calibration of
efficiency. (This may be accomplished by a computer "peak search"
routine or by manual calculations.)
4. Areas of potential problems
a. If 60Co and 88Y are present, care should be taken when
determining the count rate of the 1332.5-keV peak of 60Co
because the single escape peak (1324.1 keV) from the 1836.1-keV
gamma ray 88Y may interfere.
b. If 85Sr is present, caution should be exercised when determining
the count rate of the 514.0-keV peak because of its close
proximity to the 511.0-keV annihilation peak.
c. If 109Cd is present, care should be taken when determining the
count rate of the 88.0-keV peak because of the high background
in this region resulting from the Compton continuum. X rays
from 203Hg or other nuclides with a high atomic number may
also interfere with the 88.0-keV peak.
d. In general, peaks in high continuum regions, in the X-ray
region (less than 100 keV), and in close proximity to back-
ground or escape peaks should be carefully analyzed.
-------
-112-
5. Calculate the efficiencies using appropriately determined values of
counts per second for each FEP. Counting efficiency is calculated
from a ratio of the observed counting rate to the value of disinte-
grations per second supplied with the source material for each
gamma-ray energy. (The observed counting rate must be corrected
for gamma-ray branching ratios. Similarly, the expected count rate
for each nuclide must be corrected to the time of counting.)
6. Compare calculated efficiencies with a model and with values of
efficiency determined in previous years. A log-log, least-squares
regression may be performed on the new data to model the relation-
ship of energy to efficiency (for energies greater than 270 keV.)
7. Compare efficiencies determined by the least-squares regression
with values determined in previous years. Unless prior efficien-
cies are known to be in error, average the new and old values.
Report the calculated values to the supervisor of the counting
room.
8. With the approval of the counting room supervisor, prepare new
tables and graphs on counting efficiency. For computer-based
germanium spectroscopy systems, enter the new efficiencies into the
appropriate efficiency file of the computer.
9. Prepare several copies of the table and file on efficiency.
10. One copy of the table and file is added to the logbook of the
appropriate germanium detector, along with the appropriate infor-
mation necessary to recreate the calculated efficiencies at a later
time: (a) the NBS standard identification number, (b) all dilution
factors, (c) the code number of the quality control (QC) logbook
for the sample and the page number in the QC logbook where the
preparation of the standard is recorded, (d) a copy of the original
certification for the standard, (e) branching fractions used to
convert from disintegrations per second to gammas per second, (f) a
list of all counting rates determined for each point on the efficiency
curve, and (g) a description of the counting geometry used. This
information should be put into tabular form (as possible) to save
space in the logbook.
B.5 CHECK ON SYSTEM LINEARITY
B.5.1 Purpose
To provide instructions for checking the linearity of the ADC, pre-
amplifier, and amplifier of a germanium spectroscopy system. This proce-
dure is necessary because precise and linear amplification and sorting of
pulses are essential for accurate gamma-ray spectroscopy.
B.5.2 Scope
These instructions are applicable to all gamma spectroscopy systems
that use germanium detectors.
-------
-113-
B.5.3 References
None.
B.5.4 Abbreviations and Definitions*
ADC: Analog-to-digital converter.
germanium spectroscopy system: A system for gamma-ray spectroscopy
composed of an HPGe or Ge(Li) detector, an amplifier, an ADC, an
MCA, and a lead or steel shield.
linearity (integral linearity): The linear relationship between the
amplitude of the pulse, which is proportional to energy, and the
channel in which the pulse is stored.
B.5.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analyst and reviewing the
entries in the logbooks periodically.
2. At the beginning of each quarter, after major repair of an instru-
ment, or when a new ADC, amplifier, or detector is received, the
analyst shall check linearity of all germanium detector systems.
The results of this check must be entered in the appropriate log-
book, and abnormal results must be reported to the supervisor of
the counting room. These data should be reviewed periodically.
B.5.6 Procedure
B.5.6.1 Apparatus
1. Germanium spectroscopy system.
2. Point source containing a mixture of radionuclides such as that
prepared by NBS. The NBS source contains 109Cd, 203Hg, 85Sr, 57Co,
60Co, 88Y, 137Cs, 113Sn, 139Ce and has gamma-ray lines at 88.036,
122.060, 165.854, 391.689, 513.998, 661.630, 898.00, 1173.226,
1332.483, and 1836.075 keV.
3. Holder for the source with a 4-cm spacing.
B.5.6.2 Instructions
1. Establish the instrument settings (listed in the logbook for the
system) to be used for the check of system linearity.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-114-
2. Position the holder for the source and the point source on the
detector endcap.
3. Accumulate a spectrum for the time specified in the logbook.
4. Perform an energy calibration on the spectrum using the 88.036-keV
line of 109Cd and 1332.483-keV line of 60Co as the calibration
points.
5. Determine the channel location and corresponding energies for the
122.060-, 165.854-, 391.689-, 661.630-, 898.00-, 1173.226-, and
1836.075-keV peaks. (This may be done by manual calculation or by
use of peak search routines.) Calculate the difference between the
determined energies for these peaks and the actual energy. Record
the results in the appropriate logbook, and report any differences
greater than ±0.8 keV to the supervisor of the counting room.
-------
-115-
B.6 CHECK OF THE MCA ELAPSED-TIME CLOCK
B.6.1 Purpose
To provide instructions for checking the elapsed-time clock of the
MCA. This procedure is necessary to ensure that accurate measurements of
counting time can be achieved.
B.6.2 Scope
These instructions are applicable to all ADC-MCA systems equipped
with live-time clocks.
B.6.3 References
None.
B.6.A Abbreviations and Definitions*
ADC: Analog-to-digital converter.
MCA: Multichannel analyzer.
ADC-MCA system: ADC and MCA that are used in combination.
elapsed time: The actual "clock time" between the start of data acqui-
sition and the termination of data acquisition.
B.6.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analytical chemist.
2. The analytical chemist is responsible for performing this procedure
on all ADC-MCA systems at the beginning of each year, after major
repair to the ADC or MCA, when a new ADC or MCA is received, or
when the ADC-MCA configuration is changed. The results of these
checks must be entered in the appropriate logbook, and any abnormal
results must be reported to the supervisor of the counting room.
B.6.6 Procedure
B.6.6.1 Apparatus
1. ADC-MCA system.
2. Amplifier.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-116-
3. Pulser with precision repetition rate.
4. Calibrated oscilloscope.
B.6.6.2 Instructions
1. Disconnect all inputs to the amplifier.
2. Establish settings of the instrument as listed in the logbook for
the check of MCA elapsed-time clock.
3. Connect the output of the pulser to the input of the amplifier, and
collect a spectrum for the time specified in the logbook.
4. Sum the counts stored in the MCA, record the sum in the logbook,
and date and initial the results. If the sum deviates from the
expected value by more than 0.3 percent, report the deviation to
the supervisor of the counting room.
-------
-117-
B.7 MCA LIVE-TIME CLOCK CHECK
B.7.1 Purpose
To provide instructions for checking the accuracy of the system live-
time clock of the ADC-MCA system. Accurate compensation for "dead time"
in the MCA due to pulse pile-up at high count rates is essential to ensure
that the analyzer will accept signal pulses from a source for the same
total time, regardless of count rate.
B.7.2 Scope
This procedure is applicable to all gamma spectroscopy systems that
use an MCA.
B.7.3 Reference
1. Crouch, D. F., and R. L. Heath. Routing Testing and Calibration
Procedures for Multichannel Pulse Analyzers and Gamma Ray Spectrometers,
IDO-16923, November 1963.
B.7.4 Abbreviations and Definitions*
ADC: Analog-to-digital converter.
MCA: Multichannel analyzer.
ADC-MCA system: ADC and MCA that are used in combination.
live time: That time during which the ADC-MCA combination is capable of
accepting and recording a pulse.
live-time clock: A clock that records only the live time.
dead time: That time during which the ADC-MCA combination is not capable
of accepting and recording a pulse.
B.7.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analytical chemist and
reviewing the entries in the logbooks periodically.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-118-
The analytical chemist shall perform a check on the live-time clock
on each ADC-MCA system annually, after major repair of an ADC or
MCA, or when a new ADC or MCA is received. The results of these
checks must be entered in the appropriate logbook, and abnormal
results must be reported to the supervisor of the counting room.
B.7.6 Procedure
B.7.6.1 Apparatus
1. Germanium spectroscopy system.
2. One 109Cd source (with an activity of about 150 |jCi)
3. One pulser with a precision repetition rate.
B.7.6.2 Instructions
1. Establish the settings of the instrument to be used for the check
on the MCA live-time clock as listed in the logbook. (This
typically might involve a 2048-channel spectrum and a preset live
time of 100 s.)
2. Connect the pulser to the input of the amplifier and establish a
repetition rate of about 100 pulses per second in the region of
channel 1700 for a 2048-channel spectrum.
3. Acquire two or more spectra for a live time of 100 s and sum the
counts due to the pulser (C ).
r po'
4. Place the 109Cd source in the shield, and establish a counting rate
for the full spectrum of 1900 to 2100 counts per second. (The
109Cd source can be located at any position within the shield, and
the count rate can be varied by changing the distance between the
source and the detector.)
5. Acquire two or more spectra for live times of 100 s, and sum the
counts due to the pulser (C ). (The region summed should be
broader than in step 3 because of the effects of counting rate and
summing.) Summing of pulses from 109Cd and the pulser will cause
certain pulses to be stored in regions above the main peak gene-
rated by the pulser, while the increase in counting rate will cause
the pulser-generated peak to broaden.
6. Acquire a 100-s count with the source removed from the shield, and
sum the counts due to the pulser. This will confirm that the
repetition rate of the pulser has not changed.
7. Repeat steps 4 through 6, but establish a counting rate for the
full spectrum of 4900 to 5100 counts per second.
8. Repeat steps 4 through 6, but establish a counting rate for the
full spectrum of 9800 to 10,200 counts per second.
-------
_H9-
In the logbook, record the average values for C , C and the
quantity 100% (C - C )/C for the three couREingPfates. The
latter quantity is the*perclntage gain or loss of the live times.
Date and initial the results, and report the results to the
supervisor of the counting room.
-------
-120-
B.8 ADJUSTMENT OF THE POLE/ZERO CANCELLATION AND THE DC LEVEL
B.8.1 Purpose
To provide instructions for adjusting the pole/zero compensation and
the DC level of an amplifier used with a germanium spectroscopy system.
Proper adjustment of these settings of an amplifier is necessary to ensure
proper processing of a signal pulse by an ADC-MCA system.
B.8.2 Scope
These instructions apply to all amplifiers used with germanium
spectroscopy systems.
B.8.3 References
1. Princeton Gamma Tech, Inc. Princeton Gamma Tech Model 340
Spectroscopy Amplifier Operating Manual. February 1976.
2. Canberra Industries. Canberra Model 1413 Spectroscopy Amplifier
Operating Manual.
B.8.4 Abbreviations and Definitions*
germanium spectroscopy system: A system used for gamma-ray spectroscopy
composed of an HPGe or Ge(Li) detector, an amplifier, an ADC, an
MCA, and a lead or steel shield.
pole/zero: A potentiometer used to compensate for the undershoot caused
by the fall-time constant of the preamplifier.
DC level: The output of the DC output voltage of the unipolar output
with no signal input to the amplifier.
B.8.5 Responsibilities
1. The supervisor of the counting room is responsible for ensuring
that this procedure is followed by the analytical chemist and
reviewing the entries in the logbook periodically.
2 At the beginning of each quarter, after a change in the time
constant, after major repair of an instrument, or when a new detec-
tor or amplifier is received, the analytical chemist must check the
pole-zero and DC-level adjustments of each amplifier used in a
germanium spectroscopy system. The results of these checks must be
recorded in the appropriate logbook, and abnormal results must be
reported to the supervisor of the counting room immediately.
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-121-
B.8.6 Procedure
B.8.6.1 Apparatus
1. Oscilloscope.
2. Jeweler's screwdriver.
3. Check source (137Cs).
4. Coaxial cable with BNC connectors (5 to 10 m long).
B.8.6.2 Adjustment of the DC-Level
1. Establish the settings of the amplifier to levels normally used
during routine operation of the spectroscopy system, as listed in
the logbook of the system.
2. Remove all radioactive sources from the shield, and close the
shield.
3. Set the oscilloscope for a deflection of 20 mV/cm and a sweep speed
of 10 |Js/cm, and connect the unipolar output of the amplifier to
the input of the oscilloscope.
4. Adjust the DC level so that the trace on the oscilloscope does not
shift appreciably when the signal cable to the oscilloscope is
connected and disconnected.
B.8.6.3 Pole/Zero Adjustment
1. Establish the settings of the amplifier to levels normally used
during routine operation of the spectroscopy system.
2 Place the 137Cs source on the detector endcap, and establish a
counting rate of about 300 counts per second in the FEP by proper
spacing of the source above the detector.
3. Set the oscilloscope for a deflection of 100 mV/cm and a sweep
speed of 10 ps/cm.
4. Connect the unipolar output of the amplifier to the input of the
oscilloscope.
5. Adjust the pole/zero potentiometer of the amplifier to bring the
noise to a flat baseline. (Consult page 20 of Reference 1, Section
B.8.3.3, for a pictorial representation of a proper pole-zero
setting.)
-------
-122-
B.9 CHECK OF THE SYSTEM NOISE
B.9.1 Purpose
To provide instructions for determining the system noise of a germa-
nium detector, preamplifier, and amplifier. Accurate documentation of
noise of the system during normal operation is useful for comparative
purposes if additional noise or malfunction in the system is suspected.
B.9.2 Scope
These instructions are applicable to all systems using a germanium
detector.
B.9.3 References
None.
B.9-4 Abbreviations and Definitions*
germanium detector system: A system consisting of an HPGe or Ge(Li)
detector, a preamplifier, and an amplifier.
noise: Any pulses other than those due to photon interaction of a photon
in the detector.
B.9.5. Responsibilities
1. The supervisor of the counting room is responsible for ensuring that
this procedure is followed by the analytical chemist and reviewing
the entries in the logbook periodically.
2. The analytical chemist is responsible for performing a check of noise
on all germanium spectroscopy systems at the beginning of each quar-
ter, after major repair to the detector, amplifier, or preamplifier,
or when a new detector or amplifier is received. The results of
these checks must be entered in the appropriate logbook, and any
abnormal results must be reported to the supervisor of the counting
room immediately.
>
B.9.6 Procedure
B.9.6.1 Apparatus
1. Germanium detector system.
2. Oscilloscope.
3. Coaxial cable with BNC connections (5 to 10 m long).
*A11 definitions are operational definitions and are not necessarily
generally applicable.
-------
-123-
B.9.6.2 Instructions
I. Remove all radioactive sources from the shield, and close the
shield.
2. Establish the settings of the amplifier to those specified in the
logbook.
3. Observe the unipolar output of the amplifier at the gain and oscillo-
scope settings listed in the logbook. Record the value of the
observed peak-to-peak noise, and sketch or photograph any apparent
waveforms if they exist. Date and initial the entries in the
logbook. Report any abnormal results to the appropriate supervisory
personnel.
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO.
EPA-600/7-79-Q54
3. RECIPIENT'S ACCESSION NO.
4 TITLE ANDSUBTITLE
APPLICATION OF GERMANIUM DETECTORS TO ENVIRONMENTAL
MONITORING
5. REPORT DATE
_ March_1212_
6. PfcRFOHMING ORGANIZATION CODE
7. AUTMOR(S)
D. W. Nix, R. P. Powers, and L. G. Kanipe
8. PERFORMING ORGANIZATION REPORT NO.
TVA/EP-79/06
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Division of Environmental Planning
Tennessee Valley Authority
Muscle Shoals, AL 35660
10. PROGRAM ELEMENT NO.
INE 625C
11. CONTRACT/GRANT NO.
79 BDI
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research & Development
Office of Energy, Minerals & Industry
Washinaton, D.C. 20460
13. TYPE OF REPORT AND PERIOD COVERED
Milestone
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
This project is part of the EPA-planned and coordinated Federal Interagency
Energy/Environment R&D Program.
16. ABSTRACT
Gamma-ray spectroscopy is one of the most economical and wide-ranging tools for
monitoring the environment for radiological impact. This report examines the
problems involved in applying germanium detectors to the analysis of environmental
samples. All aspects of germanium spectroscopy—equipment, system installation,
quality control, energy and efficiency calibration, spectral analysis, analytical
sensitivities, and cost considerations—are surveyed.
Germanium detectors can be used to achieve analytical sensitivities of less than
10 pCi/L (for water) for most radionuclides, often at a confidence level of 95
percent. Germanium detectors should be used to analyze environmental samples that
may contain a couples mixture of radionuclides or unknown components because the
resolution offered by germanium detectors is unexcelled in these applications.
However, use of germanium detectors may not always be as economical as use of
sodium iodide [NaI(T£)J detectors.
7.
(Circle One or More)
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Croup
Ecology
Environments
Earth Atmosphere
Environmental Engineering
Geography
Hydrology. Limnology
Biochemistry
Earth Hydrosphere
Combustion
Refining
Energy Conversion
Physical Chemistry
Materials Handling
Inorganic Chemistry
Organic Chemistry
Chamical Engineering
6F 8A 8F
8H 10A 10B
7B 7C 13B
3. DISTRIBUTION STATEMENT
Release to public
19. SECURITY CLASS (ThisReport/
Unclassified
21. NO. OF PAGES
123
20. SECURITY CLASS (Thispage)
Unclassified
32. PRICE
EPA Form 2220-1 (9-73)
------- |