yoday
lU9LUdO|9A9Q PUB
ivoiiAivNvoiava
NiioaiNOO Ainvno
HVOI1A1VNV HOd XOOaQNVH
09f02 'D'O 'uoiuiqseM
Ajjsnpui pue s|BJ8U]|/\i 'A6J8U3 p aoi^o
iueiudo|9Aaa pue qojessay jo ao|^o
AouaBv
p UOISIAJQ
-------
E-EP/77-1*
EPA-600/7-77-088
August 1977
HANDBOOK FOR ANALYTICAL QUALITY CONTROL
IN RADIOANALYTICAL LABORATORIES
by
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 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
U.S. Environmental Protection Agency
Region 5, Library {5PL-16}
230 S. Dearborn St-eet, Room 1670
Chicago. IL 60604
-------
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.
ii
-------
ABSTRACT
A basic responsibility of operating a nuclear power program is to
establish an integral program for demonstrating the reliability and
validity of analytical data taken in connection with the nuclear program.
This handbook has been written for those persons who are responsible for
producing radioanalytical laboratory data. The primary objectives are
to identify the factors that can invalidate data and to provide guide-
lines for instituting a quality control program to monitor these factors.
This handbook 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, Subagreement 2, with the Environmental Protection Agency. Work
was completed as of May 1976.
ill
-------
-------
CONTENTS
Abstract .....................
Figures ............. ................. vii
............................. vii
Acknowledgment ......................... £x
1. Introduction ...................... i_i
2. The Radioanalytical Laboratory ............. 2-1
General ..................... 2-1
General laboratory arrangement ............. 2-1
Facilities and services ................ 2-2
Glassware ....................... 2-3
Laboratory instruments ................. 2-4
Extraneous counts ................... 2-4
Sampling ........................ 2-6
Sample handling and treatment ............. 2-6
Personnel training ................... 2-7
Other items ...................... 2-8
References ....................... 2-9
3. Quality Control of Counting Equipment ......... 3-1
General ........................ 3_1
Types of counting equipment .............. 3-1
Standard reference sources ............... 3-3
Instrument control charts ............... 3-3
Alpha and beta counting ................ 3_4
Gamma counting ..................... 3_y
Calibration ..................... 3-11
References ....................... 3-11
4. Analytical Performance
General ........................ 4_j
Selection of analytical methods ............ 4-1
Reagents and chemicals ................. 4_2
Terminology of quality control of data ......... 4-3
Performance criteria .................. 4_4
Internal quality control—demonstrating precision . . . 4-4
Internal quality control—demonstrating accuracy. . . . 4-11
External quality control—demonstrating accuracy. . . . 4-14
Summary ................. 4-14
References .................... 4-15
-------
5. Data Handling and Reporting 5-1
General 5-1
Analytical process 5-1
Data 5-2
Errors caused by computational processes 5-3
Data storage 5-4
References 5-5
6. Statistics and Counting Data 6-1
General 6-1
Counting statistics 6-1
Propagation of errors. 6-3
Limits of detection 6-6
References 6-9
7. Handling Standards of Radioactivity 7-1
General 7-1
Storage of standard solutions of radioactive materials . 7-1
Dilution of standard solutions 7-1
Using calibrated sources • • 7-2
References 7-2
VI
-------
FIGURES
Number
1 Control chart for daily measurements of background
radiation
2 Control chart for the daily counting of a standard
reference source ................ 3_o
3 Example of a typical record or log of quality
control duplicates ................. 4.7
4 Plot of duplicate range vs. duplicate mean shows
the effect of increasing concentration on range
value ................... 4_o
5 Means control chart ................. 4-12
6 Individual results control chart .......... . 4-13
TABLES
1 Laboratory Precision 4.5
2 Factors for Calculating Range Control Chart Lines '. '. 4-6
3 Gross Beta in Water (pCi/£) Duplicate Analysis Data . 4-10
4 Formulae of Propagation of Error for Some Simple
Functions ,_
VII
-------
-------
ACKNOWLEDGMENT
The author wishes to thank Mr. Arthur Jarvis of the Environmental
Protection Agency for his encouragement and advice during preparation
of this document. Appreciation is also extended to Mr. E. A. Belvin
and Dr. R. L. Doty for their editorial assistance and Mr. B. B. Hobbs
for his technical advice.
IX
-------
-------
SECTION 1
INTRODUCTION
The role of an analytical laboratory is to provide qualitative
and quantitative data that will assist in decision making. This is
especially true in a program in which (1) laboratory measurements are
used to indicate whether a given activity has significantly impacted
the environment or (2) information is provided to determine whether
operation of a specific facility complies with applicable regulations.
To be of value, analytical data must describe accurately the com-
position of samples submitted to the laboratory for analysis. A poor
or incorrect result is often worse than no result at all. Further, a
laboratory that is operated with no knowledge of its sources of varia-
tion and no procedures for action based on its results is not effective.
Therefore, it is imperative that quality control be an integral part of
any analytical laboratory program.
Although quality control is practiced to some degree in most radio-
analytical laboratories, it occasionally is subordinated to pressures generated
by a heavy workload and the need for rapid solutions to immediate problems.
To be effective and useful, quality control must be built into a laboratory
program to such an extent that it is a routine part of all other laboratory
activities.
A quality control program should (1) ensure the accuracy and preci-
sion of data produced in the laboratory so that laboratory management can
assess laboratory results in light of values known to be appropriate to
the methodology and (2) maintain the quality of laboratory data continu-
ously. Such a quality control program should cover personnel training,
methods selection, equipment operation, and data handling procedures. A
quality control program should provide sufficient information to demonstrate
that all activities conducted by the laboratory fulfill the functions above.
This manual has a twofold purpose: (1) To introduce people of all
backgrounds, who are new to radiochemistry and radiological counting instru-
ments, to (a) special problems found in a radioanalytical laboratory (as
well as good laboratory practices in general) and (b) advantages of quality
control in their work; and (2) to provide a practical guide for professional
radioanalytical personnel that will demonstrate methods for maintaining a
higher degree of control in performing analyses and producing data.
Detailed in this manual are (1) good analytical operating practices, (2)
methods for evaluating analytical data to ensure accuracy and precision,
(3) methods for identifying problems and improving data quality, (4) methods
for ensuring that laboratory equipment is operating within specified control
limits, with emphasis on alpha, beta, and gamma counting systems, and (5)
methods of applying simple statistical techniques to document data quality.
-------
-------
SECTION 2
THE RADIOANALYTICAL LABORATORY
GENERAL
Section 2 describes the variables that an analyst must consider and
control before he attempts to produce quantitative data. All laboratories
have a large number of facilities, services, reagents, glassware, and
sampling procedures that, although always considered quality control items
per se, significantly affect the quality of data produced.
In the radioanalytical laboratory, the analyst must consider several
unique factors if he is to perform radioanalytical determinations reliably,
accurately, and precisely. Although certain factors, such as special air
conditioning and reagent purity, normally are accounted for in any analy-
tical laboratory, routine factors in the radioanalytical laboratory, such
as natural background radiation, shielding, and floor loading, are not
serious problems in other types of laboratory.
GENERAL LABORATORY ARRANGEMENT
Because a radioanalytical laboratory handles radioactive material in
every aspect of its program, the laboratory must be designed to minimize
employee exposure and cross-contamination of samples. The laboratory
should be arranged so that radioactive materials are confined to one area
or building, clearly designated as a "hot" area, to which access is
restricted to authorized users of radioactive materials. Environmental
laboratories can be protected from inadvertent contamination from adjacent
areas of higher activity by (1) prescribed traffic flow patterns, (2) proper
ventilation design to move air positively from low- to high-activity areas,
and (3) good working rules and procedures. Laboratory employees must always
remember that, although a spill of radioactive material may disrupt only one
analysis, it could also cause a mandatory shutdown of the entire facility.1
Another important consideration is radioisotope storage. Radioactive
sources should be stored so that the external radiation dose does not
exceed applicable limits; normal storage is in a lead or lead-lined cave
(vault). All dilution of radioactive materials to working concentrations
should be performed in an isolated area. Each container should be labeled
with the name and quantity of material contained and the appropriate dating
information.
Counting instruments should be located in a room isolated from all
other laboratory activities. This room should have sufficient shielding
to provide a consistent radiation background at a level low enough to
permit accurate determination of activities at the required limits of
detection for the laboratory program. Generally, individual instruments
rather than whole rooms are shielded. Floor loading due to heavy
shielding and other operating considerations usually requires that a
radioanalytical laboratory, or at least the counting room or rooms, be
located on the lowest level of the building.
-------
2-2
FACILITIES AND SERVICES
Distilled water is used in the laboratory for all routine labora-
tory operations.2 However, the purity of ordinary distilled water is
generally not sufficient for use in the radioanalytical laboratory since
most counting techniques used in the radioanalytical laboratory are
sensitive enough to detect very minute traces of radionuclide impurities.
Specifically, water is often used for background determinations in gamma
spectroscopy. Radionuclide impurities commonly found in distilled water
could seriously impair analysis of low-activity samples. For determining
background activity and other applications, distilled water must be further
purified by passing it through a mixed-bed ion exchange column.2'3
However, this process does not remove dissolved air, which may contain
radon, or tritium, which is incorporated in the water molecules them-
selves. The deionized water used for background determinations in gamma
spectroscopy must be stored for about 30 days to allow the short-lived
daughter products of radon to decay.
The quality of compressed air required in the laboratory needs to
be very high. Clean air should be produced at the compressor and main-
tained to the point of its use in the laboratory. Oil and water must be
removed from the air supply. For radioanalytical purposes, the presence
of radon gas in the compressed air must be considered. Although using
tanks of compressed air that have been stored for 30 days or longer will
reduce the radon concentration, the tank materials may contain small
amounts of uranium or thorium, which by decay may produce a small but
steady radon concentration in the compressed air. Of concern in some
applications may be the presence of carbon-14 in the form of carbon
dioxide in compressed air. This problem may be eliminated by using air
manufactured by recombining nitrogen and oxygen.
An adequate electrical system, including both 115- and 230-V service,
is indispensable for any laboratory. Many instruments require relatively
constant voltage to maintain reliable operation. Lines that supply power
to the counting room must be as noise-free as possible and should be used
only for counting equipment. As necessary, individual line filters may be
placed on various instruments. Laboratory lighting must be sufficient to
allow detailed work and to provide comfort for employees.4
Counting rooms require special attention in the radioanalytical labora-
tory.5 During initial construction only low-background materials should be
used. Control of air conditioning and humidity is necessary: The tempera-
ture should be kept well below 30°C and should vary by no more than ±3°C,
and humidity should be kept between 35 and 70%. High humidity may cause
deterioration and arcing between components. Because fluorescent lighting
can produce noise and interference in certain types of counting, incandes-
cent lighting may be preferred; however, incandescent lights produce a
large amount of heat. Counting rooms should be kept at a positive pressure
relative to the laboratory and other parts of the building to minimize dust
and fume concentrations and to protect against possible airborne contamina-
tion. In addition, the air conditioning system to the counting room
should exchange the air five to ten times each hour.
-------
2-3
GLASSWARE
Laboratory glassware is used in innumerable ways in every labora-
tory, but generally it is used to store reagents, measure solution
volumes, and serve as reaction vessels. The principal glasswares used
are Pyrex and Kimax, both of which are made from borosilicate glass,
which is generally satisfactory for any analytical work. Recently, the
use of Teflon, polyethylene, and polypropylene has become more common,
but the limitations and advantages of these materials must be considered
before use. For specific information on the properties of these materials,
the reader should refer to the catalogs of the manufacturers.
Volumetric glassware and its proper use are important factors in
many laboratory operations. For accuracy, the apparatus must be read
correctly; that is, the bottom of the meniscus should be tangent to the
calibration mark. Because fluctuations in temperature can cause expan-
sion and contraction of glassware and solution volumes, glassware should
be used only at the temperature for which it has been calibrated unless
volume correction factors are known. The analyst should never assume
that newly received glassware is accurate to the limits marked on it;
new volumetric glassware should be thoroughly checked to ensure that it
does meet specifications before it is used in routine work.6
Any good text on quantitative analysis7 describes the techniques
particular to various pieces of glassware. The analyst should be aware,
however, that the notations "TC, to contain" and "TD, to deliver" that
commonly appear on analytical glassware mean exactly what they say; if
the two types are used interchangeably, laboratory results may be greatly
reduced in quality. A "TC" pipet holds the exact amount of solution
stated on the calibration and therefore should be emptied completely for
accurate transfer. A "TD" pipet, when emptied, delivers the exact volume
shown on the calibration; any residual solution should not be forced out
of a "TD" pipet.
Glassware used for measuring liquids must be clean so that the film
of liquid never breaks at any point when the glassware is being emptied.
The required amount of solution will not be delivered nor will the amount
delivered be reproducible if this precaution is not taken.
The method of cleaning will depend on the use to which the glassware
is to be put and on the residue present from previous use. The first rule
in keeping glassware clean is to rinse it thoroughly with water immediately
after use since cleaning becomes much more difficult after the glassware
has dried.
Volumetric glassware can be cleaned by several methods. One solution
for normal cleaning is 30 g sodium hydroxide, 4 g sodium hexametaphosphate,
and 8 g trisodium phosphate in 1 liter of water.6 For more stubborn residue,
a chromic acid solution may be used. Caution should be taken since this is
a very powerful oxidizing mixture. Chromic acid solution is prepared by
adding 1 liter of concentrated sulfuric acid to 35 ml of saturated sodium
dichromate solution; this addition should be made slowly and with stirring
-------
2-4
because a large amount of heat is evolved. The solution is then added
to dirty glassware and allowed to stand for about 15 minutes. Glass-
ware must be rinsed thoroughly with tap water, then rinsed with distilled
water, to remove all traces of chromic acid. The chromic acid solution
can be returned to a storage bottle for reuse. When the mixture becomes
too dilute or turns a greenish color, it is no longer effective.
Greasy residue can be removed by soaking the glassware in acetone
or by allowing warm sodium hydroxide solution to stand in the vessel.
Again, care must be taken since sodium hydroxide can etch glassware. To
dry glassware, rinse with acetone and blow or draw clean air, free of
grease and oil, through it.
In a radioanalytical laboratory, it may be advantageous to use dis-
posable pipets in procedures that may result in contamination. For small
volumes, the automatic pipets with disposable tips are often used. Glass-
ware used in work involving (1) high levels of activity or (2) environmental
or low-level activity should be well segregated and clearly labeled to
prevent cross-contamination; for example, this glassware should be washed
in different laboratory stations. At times it may be best to throw away
glassware used for work with high levels of activity rather than risk
possible contamination of other samples.
LABORATORY INSTRUMENTS
Not as many common instruments are used in radioanalytical labora-
tories as in general laboratories. Equipment such as balances and pH meters
must be checked and standardized regularly. The pH meters should be
standardized daily with two buffer solutions that bracket the pH range to
be used in laboratory work. Buffers obtained from chemical supply houses
are satisfactory for routine use.
If balances are to be used only for measurements of difference, they
need only be stable. A schedule of routine measurements on a set of known
weights will indicate the proper stability. Normal precautions and mainte-
nance must be used to achieve true weighings and ensure proper balance
function. Certainly, an instrument logbook, containing records of all
test weighings and balance servicing, should be maintained for each balance.
EXTRANEOUS COUNTS
The most difficult problem of the radioanalytical laboratory is to
isolate the radiation emitted by a specific radionuclide from' all other
sources (extraneous counts). Extraneous counts can originate from
background radiation or interferences from other radionuclides in the
sample.
Background radiation usually is accounted for by measuring a simulated
sample or source that is identical to an actual sample except for the rela-
tive absence of radioactivity. This technique can simulate those counts
arising from environmental radioactivity (e.g., 40K and decay products of
the 238U and 232Th series), radioactivity in the detectors themselves,
-------
2-5
cosmic rays, or electronic noise. The basic assumption in this technique
is that background is stable (constant over a period of time) and that the
only fluctuations that occur result from the statistics involved in the
radioactive decay process. Actually, background often has a larger
variability than predicted solely by counting statistics.
To reduce fluctuations and further stabilize background contri-
butions, shielding is necessary. Thick shields of selected lead or
steel with graded liners will reduce measurably the background arising
from environmental radioactivity. Background can be reduced further by
using anticoincidence counting techniques.
Background contributions from environmental sources are exemplified by
radon daughters, decay products of the 238U series. Radon is always pre-
sent in the laboratory in concrete block walls, compressed air, water, and
often in the samples themselves. Therefore, the laboratory can attempt
only to reduce the effects of radon fluctuations as much as possible.
Interferences can be caused by other radioisotopes in the sample
that are present originally or introduced during sample processing.
Errors encountered from this type of contamination may be reduced by
carrying a "blank sample," a sample having no known activity, through
the total analysis. However, for the situation in which more than one
radionuclide in the sample may be of interest, the components or their
decay products, which cannot be eliminated, may interfere with each
other. If the interferents are different elements, chemical separation
is possible; if they are isotopes of the same element, the interferents
may be distinguished by a physical technique.
Other techniques that can be used to resolve interferents are selected
on the basis of the half-lives of the radionuclides or the type and energy
of radiation. For example, in a strontium analysis, the determination of
85Sr in the presence of other strontium nuclides would present no difficulty
because it is a gamma emitter, whereas 89Sr and 90Sr are beta emitters.
Further selectivity can be exercised by using coincidence techniques. For
example, because 131I emits beta and gamma radiation, in cascade, it can be
selected from other iodine radionuclides by applying coincidence techniques
that choose the related loll emissions. Use of such a technique also yields
sensitivities much greater than those possible by normal counting methods
because the background is reduced to near zero.
Overall, extraneous counts generated by interfering radioactivity can
limit the accuracy attainable in any analysis. Corrections depend on the
degree of separation possible and the reproducibility of the separation.
Nevertheless, statistical fluctuations from the interferent will introduce
errors in the final result just as do background variations.
-------
2-6
SAMPLING
The sample collector must ensure that samples are collected properly
and transmitted to the laboratory in a condition acceptable for producing
meaningful results. Therefore, the sample collector must start with a
clear understanding of the purpose of the samples he is collecting and
the manner in which the sampling program has been designed to meet this
objective. From this starting point, one can determine what samples
must be taken and what frequency of sampling is necessary to yield a
true picture.
Samples collected by poor procedures always produce poor results.
Accuracy and precision of the analytical procedures in the laboratory
cannot compensate for an inaccurate sample (i.e., one that does not reflect
the actual situation). The analyst must identify unsatisfactory samples
when they are received and discuss, with those persons responsible for
sample collection, the reasons for discarding the sample.
No attempt to discuss the varying methods of sample collection and
treatment will be made since everyone experiences slightly different situa-
tions. Each sampling program may require different methods. However,
procedures that are standardized and accepted by other laboratories in the
field of study should be used. Sources of sampling information include
HASL Procedures Manual,8 Environmental Radioactivity Surveillance Guide,9
Nuclear Regulatory Commission (NRC) Regulatory Guidelines,10 American
Society for Testing Materials (ASTM) Guidelines,11 and Recommended Methods
for Water-Data Acquisition - Preliminary Report of the Federal Interagency
Work Group on Designation of Standards for Water-Data Acquisition.12
SAMPLE HANDLING AND TREATMENT
Once a representative sample is collected and delivered to the labora-
tory, the laboratory staff is responsible for properly treating and storing
the sample. Often samples collected in an environmental survey require
treatment because they are not physically ready for analysis. Treatment of
the sample after it is received depends on the sample itself and the analyses
to be performed on it. Most treatment and handling techniques have been
established and well known for many years. Again, certain precautions must
be taken in the radioanalytical laboratory.
Water Samples
Generally, water samples should be acidified when collected. Under
certain conditions this procedure should be modified; for example, if
total and dissolved fractions are to be analyzed, the samples should be
filtered before acidification. If tritium or carbon-14 analyses are to
be performed, portions for these analyses should be separated before acid
is added. Samples for tritium analyses should not be stored in polyethyl-
ene bottles for more than 3 or 4 months because water can evaporate through
the polyethylene. If the samples are to be stored for any length of time,
-------
2-7
carrier or complexing agents should be added to prevent adsorption of
trace metals on the storage containers. Adsorption is known to occur
quite rapidly under some conditions.
Air Filters
The air filter should be handled with care when dust loading is
observed because particulate matter is easily removed from the filter,
thus invalidating the analysis. Air filters are often received by the
laboratory in envelopes; some extremely low-level analyses may require
analysis of the envelope in which the sample arrived as well as the
sample itself.
Milk
Milk samples should be refrigerated until analyses can be performed.
If the analyses will be delayed for more than a few days, a preservative
(formalin or merthiolate) should be added to inhibit bacterial growth and
retard spoilage. Milk samples that are to be analyzed for 131i should
not have formalin added because formalin is thought to cause increased
complexing of the iodine.
Soil
Soil samples should be dried, pulverized, and sieved to pass a 200
mesh (this will vary according to the analyses to be performed) sieve
before analysis. Further thorough mixing is required to ensure a
homogeneous sample.
Other Samples
Perishable samples should be preserved by refrigeration or freezing.
Vegetation and other samples may need to be dried, pulverized, or ashed
before analysis.
PERSONNEL TRAINING
Although the degree of skills and training necessary for laboratory
personnel naturally corresponds to their job responsibilities, all labora-
tory personnel must be thoroughly acquainted with basic laboratory
operations. A new laboratory analyst must learn (1) how-the sample pro-
cessing system of the laboratory works so that he is familiar with the
laboratory's unique sequence of work; (2) how each type of sample is to
be treated when it 'arrives in the laboratory; (3) how to use routine
laboratory equipment such as volumetric glassware and analytical
balances; (4) how to clean glassware properly; and (5) how to maintain
routine equipment of the laboratory such as pH meters and balances.
-------
2-8
Before a new analyst is assigned to independent work, a gradual on-
the-job training program should be conducted. The trainee should first
observe experienced personnel at work and study the laboratory manual
thoroughly. The trainee should then perform the analysis under close
supervision of an experienced analyst. Finally, the analyst should be
allowed to work independently, but his work should be checked frequently
at first. Provision should also be made to retrain each analyst regularly,
especially in any areas in which he does not perform often.
The work in the radioanalytical laboratory ranges from sample pre-
paration to simple analyses, such as a gross beta analysis, to complex
analyses, such as strontium or plutonium separations. The time necessary
to train an analyst properly increases with the difficulty of the
analyses expected of him. Several months of training may be necessary
before a new analyst can perform independently all the types of work
required of him.
The most difficult task in the radioanalytical laboratory is to
operate the counting equipment properly. This job is complicated because
the operator must be able to detect invalid data early to avoid large
wastes of counting time and analytical effort. For example, the examina-
tion of raw gamma spectroscopy data to ensure an adequate measurement is
a very subjective art. The operator will have to see hundreds of samples
under all conditions before he can judge data effectively; this will
require many months of experience.
OTHER ITEMS
Another factor that must be considered when planning the radioana-
lytical laboratory program is the half-life of radioactive materials.
Samples must be analyzed soon after they are received to ensure accuracy
and low error estimates for radionuclides that have half-lives of only a
few days. Half-life considerations require that the laboratory have a
program for continuously renewing its supply of calibrated material.
Most standards are assumed to be invalid after passing through four half-
lives from the date of certification because of compounding of errors and
assumptions made in the original determinations of half-lives for the
various radionuclides.
Quantities of material needed or available for analysis have a signi-
ficant role in the work of the laboratory. Large samples are necessary
for working with low-level radioactivity to provide sufficient activity to
count; however, large samples reduce the efficiency of a gamma counting
system because of the poor counting geometry and introduce possible inter-
ferents into the analysis. Therefore, the sample size must be considered
when computing the overall analytical error. The problem of sample size
often is more pronounced for the radioanalytical laboratory because of the
isotopic abundances of some radionuclides. For instance, the abundance of
40K is about 1/100 of one percent of the total natural abundance of potas-
sium. Whereas an analyst in a general laboratory would analyze for total
potassium, the analyst in a radioanalytical laboratory may determine only
the 40K content of the sample.
-------
2-9
Samples containing high levels of radioactivity may require dilutions
to avoid counting errors in gamma spectroscopy caused by dead time of the
multichannel analyzer system. Here again the radioanalytical laboratory
has a more pronounced version of the problem since a sample may have a
very high activity, but a very low initial concentration of the nuclide
of interest. Dilution simply amplifies the problem. Usually, carrier
must be added to the solutions to keep the radionuclides of interest
from absorbing on the container walls at their low concentrations.
Chemical separations can always be performed, but if the analysis can be
performed by gamma spectroscopy without chemical separations, a significant
cost savings can be effected.
REFERENCES
1. American Institute of Chemical Engineers, Design Guide for a
Radioisotope Laboratory (Type B). Science Press, New York, 1964.
47 pp..
2. American Public Health Association. Standard Methods for the
Examination of Water and Wastewater. 13th ed. New York, 1971.
874 pp.
3. Applebaum, S. B., and G. J. Grits. Ind. Water Eng., September-
October 1964.
4. Steere, N. V., Ed. Handbook of Laboratory Safety. The Chemical
Rubber Co., Cleveland, Ohio, 1971. 854 pp.
5. International Commission on Radiation Units and Measurements.
Measurement of Low-Level Radioactivity. ICRU Report 22, Washington,
DC, June 1972. 66 pp.
6. U.S. Environmental Protection Agency. Handbook for Analytical
Quality Control in Water and Wastewater Laboratories. Cincinnati,
Ohio, June 1972.
7. Rieman, W., J. D. Neuss, and B. Naiman. Quantitative Analysis.
McGraw-Hill Book Co., New York, 1951.. 657 pp.
8. U.S. Atomic Energy Commission. HASL Procedures Manual. HASL-300,
New York, 1972.
9. U.S. Environmental Protection Agency. Environmental Radioactivity
Surveillance Guide. ORP/SID72-2, Washington, DC, June 1972.
10. U.S. Nuclear Regulatory Commission. Regulatory Guidelines.
Washington, DC.
11. American Society for Testing Materials. ASTM Guidelines.
12. U.S. Dept. of the Interior. Recommended Methods for Water-Data
Acquisition—Preliminary Report of the Federal Interagency
Work Group on Designation of Standards for Water-Data Acquisition.
Washington, DC, 1972. (A newer edition is being prepared.)
-------
-------
SECTION 3
QUALITY CONTROL OF COUNTING EQUIPMENT
GENERAL
Most chemical laboratories can switch to a strictly wet chemical
approach to avoid instrumentation problems, but virtually all measure-
ments in a radioanalytical laboratory eventually require the use of
counting equipment for final determinations. A number of quality
control measures are necessary to ensure proper instrument performance.
Specific quality control procedures to be followed depend on the type
of radiation being analyzed and the methods of detection and signal
processing used. Such procedures will be discussed in this section.
TYPES OF COUNTING EQUIPMENT
Virtually all analyses in the radioanalytical laboratory eventually
require the measurement of alpha, beta, or gamma radiation. In general,
each type of radiation is analyzed with different types of equipment of
a highly specialized nature, all of which can detect a low-level signal,
amplify it, process it, and store or record it as digital information.
Alpha radiation can be counted in several different ways, depending
on the type of information and degree of sample handling desired. For
a sample that has undergone chemical separation to eliminate interfering
radionuclides and deposition as a thin layer in a counting geometry,
counting can be performed using either an internal proportional counter
(IPC) or a thin-window gas-flow proportional counter. (In an IPC,
the sample to be counted is actually introduced into the counting
chamber of the detector itself. The sample is outside the detector in
a thin-window gas-flow proportional counter; therefore, the beta or alpha
emissions must penetrate through the detector window to enter the counting
chamber.) Selection of the instrument to be used for a particular analysis
is based on the counting efficiency and sample-processing speed desired.
The IPC can normally achieve a counting efficiency approaching 50 percent
for alpha and beta radiation, but it is relatively inconvenient for
processing samples rapidly because most IPCs have no automatic sample
changers.
In general, the thin-window gas-flow proportional counter is preferred
for most gross counting of alpha and beta radiation. Most newer models
are dual-detector instruments operated in the anticoincidence mode. The
instruments are capable of achieving efficiencies (corrected for self-
absorption) of greater than 20 percent for alpha (210Po) radiation and
greater than 40 percent for beta (90Sr) radiation while maintaining a
background of less than 1.4 counts per minute. Also, most modern instru-
ments have automatic sample changers with capacity for at least 50 samples.
-------
3-2
Isotopic analyses of alpha-emitting radionuclides can be made by
alpha spectroscopy. Alpha spectroscopy is especially useful for analyzing
uranium milling and mining materials, which have a large number of alpha-
emitting radionuclides present. The nuclides of interest must be
chemically separated and then electroplated onto a small metallic disc.
The alpha emissions from thin sources are detected by surface barrier
detectors, and the data are collected with a pulse-height analyzer
(PHA); however, all sources must be electroplated to achieve any
reasonably reproducible resolution for the instrument.
Activities of samples that emit low-energy beta particles such as
tritium and carbon-14 are usually determined by liquid scintillation
counting. Liquid scintillation counting can be a rapid and sensitive
technique, but it usually requires some type of purification of the
sample to prevent quenching. Some alpha and X-ray counting can be done
by liquid scintillation.
A scintillation system composed of a nylon disc coated with an
activated zinc sulfide (ZnS) phosphor also may be used to count alpha and
beta radiation. The sample is placed on the disc and counted with a
photomultiplier tube. The counting system must be in a lighttight container.
Radium-226 is often determined by the technique of radon emanation.
In this technique, a radium-226 solution is stored in a sealed container
for 15 to 30 days to allow its daughter product, radon-222, to grow into
the sample. The radon is then transferred to a ZnS (silver-activated)
coated counting cell called a Lucas cell. The alpha emissions of radon-222
and its daughters produce light pulses, which are counted with a photo-
multiplier tube. This technique has a counting efficiency approaching
70 percent. Again, the counting must be performed in a lighttight container.
Gamma radiation can be counted with many different instruments;
however, the two basic categories are single- and multichannel analyzers.
If a specific photopeak of a particular radionuclide is easily isolated
and if the sample contains no other radionuclide that can affect the
counting, a single-channel analyzer and sealer can provide all the
information needed for one analysis.
The more common analytical problem is the analysis of a sample
that possibly contains a wide range of radionuclides, any or all of
which may be of interest. In this type of application, gamma spectro-
scopy with a PHA system yields the most informative estimate of the
activities of several radionuclides. There are basically two types
of detectors now being used for gamma spectroscopy: the thallium-doped
sodium-iodide [Nal(Tl)] detector and the lithium-drifted germanium
[Ge(Li)] or the intrinsic or hyperpure germanium (HPGe) detector.
An analysis system based on a 10- by 10-cm Nal(Tl) crystal typically
requires an analyzer with at least 256 channels of memory and at least
one readout device. A shield with a minimum of 10 cm of steel or its
equivalent is necessary to achieve acceptable background levels for
environmental monitoring.
-------
3-3
A germanium-based analysis system requires a much more sophis-
ticated multichannel analyzer, with at least 2000 channels of memory
and one high-speed readout device. Normally, a 4000-channel analyzer
is a better selection to fully utilize the improved resolution of newer
germanium detectors. Many analyzer systems available today for germanium
analysis are computer-based systems having 24 to 128K of usable memory.
Because these systems have their own analysis software capability, the
user is completely independent of external computer requirements.
The activity levels being determined in different laboratories
or within the same laboratory can vary extensively. Counting samples
that contain low levels of activity requires counting equipment with
extremely low noise and high selectivity, and counting of high-activity
samples requires equipment with exceptional response and low distortion.
Therefore, the instruments must be of the highest quality and versatility.
STANDARD CHECK SOURCES
A specific check source should be used with each counting system.
A source chosen as a check will contain a nuclide or nuclides whose
type and energy of radiation correspond to the type of analysis for
which the counting system is to be used. This source will be counted
routinely to determine general performance of the system and to ensure
that the efficiency of the system has not changed. Although knowledge
of the absolute activity of the source is not necessary, the source must
be sufficiently active to provide adequate counting statistics over the
time for which it is to be counted. However, the source must not be
so active as to cause pulse pileups, dead time that is significantly
different from that to be expected from routine samples, or gain shift
in the case of PHA systems. For example, a check source for a PHA system,
equipped with a modern analog-digital converter (ADC), should not count a
total rate exceeding about 1000 counts per second. Any source to be used
as a check source should be either sealed or encapsulated to prevent loss
of the source and contamination of the counting system. The check source-
to-detector geometry must be known and held constant. The geometry is
extremely important because small changes can have large effects that
overshadow normal statistical counting variations.
Because alpha and beta sources are subject to leakage after a few
years, they should be replaced periodically even though the nuclide may
have a long half-life. Also, alpha and beta sources should be surveyed
(smeared) routinely for possible leakage.
INSTRUMENT CONTROL CHARTS
The statistical nature of radioactive decay will result in
uncertainties in the determinations of check and background count
rates. Despite this effect, the analyst is able to detect deviations
-------
3-4
from the "true" values that result from instrument malfunctions by
recording routine check-source and background determinations on control
charts. Quality control charts initially were devised to follow graphi-
cally the quality control of manufactured products by testing a limited
sample to determine whether reproducible results were being obtained. This
same principle can be applied to any type of repetitive measurements.
A control chart consists of a graph showing time on the abscissa
and count rate, or total counts, on the ordinate. Lines are drawn parallel
to the time axis at values (corrected for decay if necessary) for the "true"
count rate and for values of ±2 and ±3 standard deviations. The "true"
count rate is determined by averaging at least 20 countings whose individ-
ual statistics are acceptable, or from a single measurement for a longer
period, such as an overnight count. When multiple measurements are to
be made and averaged, the limits of the quality control chart must be
corrected appropriately. For example, for a control chart based on the
average of three measurements, the ±2 and ±3 standard deviation limits
must be divided by the square root of three to yield appropriate values.
Quality control charts must be interpreted objectively. When a
point goes outside the limits, instrument service may or may not be
needed. To determine whether service is necessary, the analyst should
run a series of repeated measurements and apply another statistical test,
such as a Chi-square test, to determine whether the variation was
nonstatistical.4
Trends from control charts may show other information. For example,
if regular measurements of the check source are daily moving in one
direction, one can infer that some system variable is changing. This
variation may not always require instrument service; instead, reevaluation
of the values of the standard deviation and other related limits of the
control chart may be necessary.
ALPHA AND BETA COUNTING
On a daily basis (or before each use), the check source for each
system should be counted for a predetermined time. The count rate is
entered in the logbook and plotted on the control chart established for
the specific system. This value is compared with the ±2 sigma (warning)
limits and the ±3 sigma (out-of-control) limits, and the procedure is
repeated if the ±2 sigma boundary is exceeded. Sustained values above
the warning levels require appropriate action.
On a daily basis (or before each use), background for each system
should be counted for the standard counting time (the time for which
samples normally are counted) if overnight background counts are not
made. Background measurements obtained during the same period in which
actual sample measurements are being made are preferable. This value
is entered in the logbook and plotted on the control chart established
for the specific system (Figure 1). The value is compared with the ±2
and ±3 sigma limits, and appropriate action is taken if indicated.
-------
Figure 1. Control chart for daily measurements of background radiation.
-------
3-6
Appropriate action for background or efficiency changes includes:
1. A check of the source positioning or for possible contamination.
2. A check of the gas supply for possible variation in the gas pressure.
3. A check of the gas itself; bad gas usually is indicated when several
instruments operating off the same supply change drastically at the
same time.
4. A check of the high-voltage (HV) setting to determine whether drift
or other factors have caused the HV to change from the correct
operating potential; also check the HV actually delivered to the
detector.
5. A thorough cleaning of the detector to remove possible contamination
from the window or sample chamber.
6. A check of the line volage to determine whether noise is entering
from the power source.
7. A check of the instrument preamplifiers and other signal processing
circuits for proper functioning.
8. A check of the instrument guard detector (if a dual-detector system)
for proper functioning.
9. A check of cabling and grounding to determine whether the HV is
arcing.
Certainly this list is not all-inclusive, and other checks should be
identified by the laboratory staff as part of a routine check-off
procedure when malfunctions occur.
The control charts should be reviewed each week for possible
trends. If trends are present, the cause should be determined and
appropriate action should be taken.
On a quarterly basis or after electronic repair or modification,
the detector plateau for gas-discharge devices should be determined and
plotted. Some laboratories prefer to redetermine plateaus after each
change of counting gas, although this may not be necessary if the check
source counts normally. •'All pertinent instrument settings, the source
used, and the rate of gas flow should be recorded on the plateau graph,
which should be attached permanently to the logbook. From this plateau,
the operating voltage is selected or verified and the plateau slope at
the operating point is calculated. The slope should generally not
exceed 2 percent per 100 volts for a 90Sr source. The operating potential
is usually selected as the midpoint of the plateau. Thereafter, the HV
setting should be checked for drift once every two months.
The percentage of alpha-beta crossover should be determined periodi-
cally for those systems in which alpha radiation is differentiated from
beta radiation by pulse-height discrimination. Window settings are then
adjusted as necessary to minimize crossover.
-------
3-7
GAMMA COUNTING
Due to differences in signal processing, methods for use in gamma
counting are divided into two classes: single-channel analysis and
multichannel analysis.5'6
Single-Channel Analysis
The selected check source should be counted over the energy window
of interest for a predetermined time each day or before each use. The
amplifier fine gain is adjusted for optimum response before this measure-
ment. The total counts in the window are then entered in the logbook
and plotted on the control chart established for the specific system.
This value is compared with the ±2 sigma and ±3 sigma limits, and the
measurement is repeated if the ±2 sigma limit is exceeded. Sustained
values above the warning limits require appropriate action.
A background count should also be made over the same energy
window of interest for the standard time required for sample counting
if overnight background counts are not made. The total counts in the
window are entered in the logbook and plotted on the background control
chart established for the specific system. This value is compared with
the ±2 and ±3 sigma limits, and appropriate action is taken if the limits
are exceeded.
Occasionally, or when anomalous background measurements are observed,
the analyst may perform a Chi-square test to determine whether the
performance is statistically acceptable (see Section 6). Too small a
value of Chi-square indicates some regular interference such as 60-hertz
noise, and too large a value of Chi-square indicates malfunctioning
electronics.
Appropriate action for background or efficiency changes includes:
1. A check of source positioning or for possible contamination.
2. A check of energy window settings, including possible changes
of window size that are not reflected on the actual dial settings.
3. A check of gain settings.
4. A check of the preamplifier for noise and malfunction.
5. A check of the HV settings of the detector and HV delivery to
the detector.
6. A check of the HV supply for noise.
7. A check of the check source geometry.
Because other possibilities exist, a good checklist should be prepared
and consulted x\rhenever necessary.
-------
3-i
The control charts should be reviewed each week for possible trends,
If trends are present, the cause should be determined, and appropriate
action taken.
Multichannel Analysis
Before instrument efficiency or background counting rates are
determined, the instrument must have a proper energy calibration; that
is, a multiline reference source should be counted for a time sufficient
to provide acceptable statistics. The channel positions of the spectral
lines are adjusted with the appropriate PHA controls until the spectral
lines correspond as closely as possible to the known line positions. A
laboratory may wish to have more than one source for this check for
Nal(Tl) detectors: (1) a check source, such as 207Bi, that has several
emission lines spread widely throughout the major regions of interest
and (2) a source that has several lines in the low-energy (<300-keV)
area, where linearity is most critical. For germanium detector systems,
a check source, such as a multinuclide point source traceable to the
National Bureau of Standards (NBS), can be used. Such a source has lines
from about 300 keV, spaced at even intervals throughout the energy spectrum,
to 1800 keV. From the channel positions of two or more specified lines
appropriate to the energy region of interest, the energy by channel
(keV/channel) and zero intercept are calculated and entered in the logbook.
After energy calibration, the selected check source should be counted
for a predetermined time each day or before each use by using a selected
energy window. The amplifier fine gain is adjusted to center a specified
photopeak on a selected channel. The window can be set to measure either
the total number of counts in the specified photopeak or the total number
of counts summed over the number of channels, symmetrically disposed
around the photopeak channel, that is equal to the full width at 1/10
maximum. The total number is entered in the logbook and plotted on the
efficiency control chart established for the specific system (Figure 2).
This value is compared with the ±2 and ±3 sigma limits, and the process
is repeated if the ±2 sigma limit is exceeded. Sustained values above
the warning limit require appropriate action.
Background should be counted for the standard counting time. The
total counts in the energy window defined above for the efficiency test
are entered in the logbook and plotted on the background control chart
established for the specific system. This value is compared with the
±2 and ±3 sigma limits, and appropriate action is taken if the limits
are exceeded.
Appropriate action for background or efficiency changes includes:
1. A check of source positioning or for possible contamination.
2. A check of the energy/channel value to determine whether a gain
shift has occurred.
3. A check of gain settings.
4. A check of the preamplifier for noise or malfunction.
-------
OJ
I
Figure 2. Control chart for the daily counting of a standard reference source; the
chart is corrected for decay.
-------
3-10
5. A check of the HV settings of the detector and HV delivery to
the detector.
6. A check of the HV supply for noise.
7. A check of the counting geometry of the check source.
8. A check on the ADC performance.
Again, these suggestions do not constitute a complete list of
possible checks; however, personnel in the counting room should compose
a checklist for reference purposes when problems arise.
The control charts should be reviewed weekly for possible trends.
If trends are present, the cause should be determined and appropriate
action should be taken.
The resolution for a specified photopeak should be determined and
recorded at least once a month. Changes of about 2 keV for Nal(Tl)
systems require corrective action. At the same time the peak-to-Compton
ratio should be determined and recorded in the logbook.
The integral energy linearity of the PHA system should be determined
periodically with a multinuclide source covering the energy range of
interest. The differential linearity of the multichannel analyzer should
be determined with a sliding pulser at least twice a year.
Time bases should be checked regularly for proper functioning in
all the time periods that will be used for analysis.
An energy efficiency curve should be determined annually for each
germanium detector system for each geometry with a multiline reference
source calibrated by the NBS. The curve for the most frequently used
geometry should be checked frequently during the year. Complete
information concerning the half-lives and decay schemes can be obtained
from the nuclear data tables7'8 if it is not provided by NBS with the
source. For comparability, these same values should be used in any
similar work. A check on the energy efficiency curve is not necessary
for every geometry for quality control purposes. The EPA Environmental
Monitoring and Support Laboratory in Las Vegas also offers well-
characterized standardization material, which is available regularly.
By using a gamma ray source or sources that have energies ranging
from about 100 to 1500 keV, the operator can estimate the status of a
Nal(Tl) or germanium detector. Any degradation over time of the low-
energy line would indicate that the surface of an Nal(Tl) detector is
deteriorating or that the depth of the dead layer of a germanium
detector is increasing. The status of the bulk volume of the detector
can be estimated from the higher-energy lines.
A thorough and up-to-date discussion of the problems of quality
control in gamma-ray spectrometry has been compiled by Zeigler and Hunt.9
Specific tests and evaluation procedures, with sample data, are discussed
at length.
-------
3-11
To achieve meaningful data, these routine measurements of quality
control must be made in an identical fashion every time. This will
require formulation of extremely reproducible geometries, explicit
instructions, and meticulous obedience to those instructions by all
operators. Small daily changes in methods can result very quickly
in values that may differ from earlier results by 50 percent.
Use of these routine procedures and others specifically drawn
up by each laboratory will allow the operators to state with confidence
whether a particular system is operating properly at any moment.
CALIBRATION
Instrument calibration with good standards should be performed
regularly. Only well-defined standards should be used, and care must
be taken to prepare samples homogeneously and in accurately defined
geometries. Replicate samples, multiple measurements, and good counting
statistics should be stressed as important for every geometry used.
If multiple detectors are run simultaneously by one multichannel
analyzer, the operator must avoid acquiring data on high-activity samples
from more than one detector at a time through the same ADC. Even though
the analyzer corrects for dead time, the samples may contribute to the
dead time by different percentages. Also, use of samples that have
sufficient activity to cause spectral distortion should be avoided by
using dilution as necessary.
REFERENCES
References 1, 2, and 3 discuss the construction of control charts,
and references 4, 5, and 6 discuss some specific parameters that should
be monitored by control charts.
1. Bennett, C. A., and N. L. Franklin. Statistical Analysis in
Chemistry and the Chemical Industry. John Wiley & Sons, Inc.,
New York, 1954. 724 pp.
2. Enrick, N. L. Quality Control and Reliability. 6th ed. Industrial
Press, Inc., New York, 1972. 306 pp.
3. Grant, E. L. Statistical Quality Control. 9th ed. McGraw-Hill
Book Co., New York, 1972. 694 pp.
4. U.S. Atomic Energy Commission. HASL Procedures Manual. HASL-300,
New York, 1972.
5. Price, W. J. Nuclear Radiation Detection. 2nd ed. McGraw-Hill
Book Co., New York, 1964. 430 pp.
-------
3-12
6. Crouthamel, C. E., Ed. Applied Gamma-Ray Spectroscopy. Pergamon
Press, New York, 1960. 443 pp.
7. Lededer, C. M., J. M. Hollander, and I. Perlman. Table of
Isotopes. 6th ed. John Wiley & Sons, Inc., New York, 1967.
594 pp.
8. Nuclear Data Sheets. Academic Press, New York. Monthly publication.
9. Zeigler, L. H. , and H. M. Hunt. Quality Control for Environmental
Measurements Using Gamma-Ray Spectrometry, U.S. Environmental
Protection Agency (in press).
-------
SECTION 4
ANALYTICAL PERFORMANCE
GENERAL
Maintaining a flow of accurate data from the laboratory in a pro-
gram that involves a large volume of samples is difficult. A laboratory
must first analyze the scope of its projected program to determine those
factors that will influence the final results. Facilities, services,
methods, equipment, reagents, chemicals, and training of personnel all
have an impact on data quality.
Once a laboratory has competent analysts, reliable, properly operat-
ing equipment, and well-known tested procedures, quality control procedures
can be used to (1) minimize systematic error in the total analysis and (2)
document the quality of data produced by the laboratory. Although quality
control cannot ensure absolute accuracy, it can help produce consistent data.
SELECTION OF ANALYTICAL METHODS
A number of chemical procedures and counting procedures are available
for use in any radiochemical analysis. This multiplicity of techniques
makes method standardization difficult and often indescribable since varia-
tions of a single procedure may be used in a laboratory for different types
of samples. Also, this array of choices makes method selection difficult
and requires that the analyst be able to document his choice of method and
the validity of the results obtained with that method.
The analyst can be more confident of a particular method if it has
been generally accepted throughout the industry. Therefore, when the
laboratory staff plans to adopt new procedures, it should consider standard
methods, particularly from ASTM, EPA, NRC, the American Public Health
Association, and the Association of Official Analytical Chemists. These
procedures should be studied to help determine, for a particular analy-
sis, (1) the levels that can be measured precisely and accurately in the
presence of known interferences; (2) the equipment and skills that are
required to perform the analysis; (3) the size of sample load that can
be analyzed by the method; and (4) the information that is available to
indicate the validity of the method.
After the laboratory staff has accepted a method that fulfills the
needs of the laboratory in terms of the above criteria, it should docu-
ment that method carefully before it is used routinely. Proper documenta-
tion includes a write-up in a standard procedural format used for all
analytical methods in the laboratory. Once the method is accepted,
documented, and placed into use, it should be checked periodically to
ensure that it continues to meet the needs of the laboratory. As these
needs change over time, the adoption of newer, quicker, or more sensitive
methods may be necessary.
-------
4-2
Acceptance of a procedure should not stop investigations that might
lead to improvements in the specificity or sensitivity of the method.
Before such a modification is made, it should be worked out carefully
and data should be accumulated to document the superiority expected from
the modification. Before the new procedure is used routinely, it should
be rewritten in the laboratory's standard procedural format to ensure
that the procedure is followed correctly by the staff.
REAGENTS AND CHEMICALS
If all other factors in a laboratory are under control, the most
important source of error remaining is the chemicals and reagents used
to process samples. Control of chemicals and reagents is extremely
important to any program.
The quality of reagents is of great importance. The type of reagent
for a particular use is often determined by the impurities that can be
found in the reagent. Generally, the best quality of reagent available
should be employed. Chemicals that bear the classification "ACS grade"
meet the specifications set forth by the American Chemical Society. Other
chemicals are classified as "analytical reagent grade" or "spectral grade
organic solvent." Methods for determining impurities of suspect reagents
can be found in Refs. 1-3.
Because even materials of analytical reagent grade vary in quality
from lot to lot and manufacturer to manufacturer, reagents must be checked
carefully before they are used in the laboratory. Once in service, a
reagent should be tested frequently to detect deterioration as early as
possible. To maintain reagent quality for as long as possible, the manu-
facturer's directions for storage and use should be followed closely.
Radioanalytical laboratories require purer chemicals than do other
laboratories for many purposes; for example, liquid scintillation requires
extremely pure solvents and scintillators to prevent quenching. Trace
amounts of impurities that alter counting data without interfering with
any chemical procedure must be eliminated (e.g., all traces of radium
must be removed from barium used in coprecipitation of radium).
Another difficulty encountered in the radioanalytical laboratory is
obtaining radionuclide standards to be used for tracing, testing, and
calibration. These standards should be of the highest quality and trace-
able to the NBS if possible. The EPA standards program mentioned in
Section 3 is equally acceptable and is probably better for routine radio-
nuclides. A certificate of calibration must describe the standard
adequately and should accompany each radionuclide. The calibration
certificate should provide (1) a description of the solution (principal
radionuclide, mass or volume, and chemical composition); (2) the
reference time and date; (3) the measurement result (activity of prin-
cipal and possible daughter radionuclides per gram of solution); (4) the
measurement method; (5) a statement of purity (list of known or suspected
radionuclide impurities, their activities, and how they were measured);
-------
4-3
(6) the decay information (statement of the assumed half-life and other
decay information); and (7) an estimate of errors (includes errors from
the measurements themselves and those created by the decay assumptions).
Calibration standard solutions are shipped in flame-sealed glass
ampoules. Once opened, these solutions will deviate rapidly from their
calibration values if not properly stored and diluted.
TERMINOLOGY OF QUALITY CONTROL OF DATA
mean: The sum of the test results divided by the number of results
taken; that is, X = IX. /n, where X = mean, X. = individual result,
and n = number of results.
precision: A measure of the reproducibility among replicate observations
variance: The sum of the squares of deviations of the test results from
the mean after division by one less than the total number of
n - 2
results; that is, VAR = I. (X - X. ) /(n - 1) .
standard deviation: The square root of the variance; that is,
n
a = (VAR)2 =
I (X - X.r/(n - 1)
range: The difference between the highest test result and the lowest
test result in a set of observations.
accuracy: A measure of the agreement between observed and accepted
values.
systematic error: Errors that may be traced to the personal errors of
the experimenter, the instrumental errors of his measuring devices,
the errors that repose in the method of analysis he employs, or a
combination of these. Accuracy describes this type of variability
or error.
random error: The necessity for making estimations is inherent in the
process of collecting data for the measurement of any quantity.
For this reason, any measurement will be uncertain, in an amount
that depends on the relative magnitude of the estimations involved
in its evaluation. Careful experimental design can reduce this
uncertainty; however, small irreducible variations will remain.
Since radioactive decay is a random process, any counting measure-
ment will have a random error associated with it. Precision describes
this type of variability or error.
bias: The difference between the average of a set of test results and
the accepted value. Bias usually is indicated only when a consistent
difference is observed over time and can be corrected for by the
application of appropriate correction factors. Bias is a measure
of the systematic error.
-------
4-4
An excellent discussion of this terminology and its significance
to an analytical chemist can be found in Ref. 4.
PERFORMANCE CRITERIA
For a successful quality control program, acceptable and attainable
performance criteria must be selected for precision and accuracy. These
criteria must reflect the capabilities of the laboratory and the purposes
for which the data are to be used.
These criteria can be drawn up initially from experience with the
analytical method or from criteria set by other laboratories using the
same procedure. A tabulation of allowable deviations used by the EPA in
their Environmental Radioactivity Laboratory Intercomparison Studies Pro-
gram is given in Table I.5 As can be seen, the criteria are a function
of the particular analysis under study. These values certainly are not
the only ones that could be used; however, a laboratory might use these
until enough data can be compiled to set its own criteria from experience.
INTERNAL QUALITY CONTROL--DEMONSTRATING PRECISION
A laboratory must be able to reproduce its analytical results inter-
nally within acceptable limits. If the laboratory cannot show such
consistency, all its results can be considered unsatisfactory. To
demonstrate this internal consistency, the laboratory staff should set
up a program of duplicate analysis on a portion of the actual sample work-
load, with properly kept records6"10. This technique will also direct the
staff to any problems in analytical procedures very quickly.
The best type of duplicate (or replicate) analysis program is one
based on blind samples, which denies the analyst any foreknowledge concern-
ing the sample that might bias a second determination. However, conducting
such a program in the laboratory is not always easy, especially in those
smaller laboratories in which the analyst is aware of every laboratory
activity.
In such a case, an alternative approach to a duplicate analysis pro-
gram can be used, in which one sample out of every ten submitted for a
particular analysis is randomly selected for duplicate analysis. All
eleven samples, ten unknowns and one duplicate, are analyzed simultaneously
so that the data from the analyses are then free of any foreknowledge con-
cerning the expected result. Of course, if less than ten samples are to be
analyzed, one of those available is selected as the duplicate.
The laboratory staff must ensure that all duplicates are taken from
a homogeneous matrix so that no additional variance is introduced into the
duplicate determinations. However, the sample size and the physical and
chemical characteristics of the duplicate samples should closely approxi-
mate those of samples routinely analyzed singly. Because true duplication
of all types of samples is not always possible, spiked samples may have to
be substituted.
-------
4-5
TABLE 1. LABORATORY PRECISION
One Standard Deviation Value for Various Analyses
Nuclide
Level
Standard deviation
(single determination)'
131-
140
137
Ba
Cs
89Sr
90Sr
40K
Gross alpha
Gross beta
3H
226Ra
239pu
5-100 pCi/1 or kg
>100 pCi/1 or kg
5-100 pCi/1 or kg
>100 pCi/1 or kg
5-100 pCi/1 or kg
>100 pCi/1 or kg
5-100 PCi/l or kg
>100 pCi/1 or kg
2-30 pCi/1 or kg
>30 pCi/1 or kg
>0.1 g/1 or kg
<20 pCi/1
>20 pCi/1
<100 pCi/1
>100 pCi/1
<4000 pCi/1
>4000 pCi/1
>0.1 pCi/1
>0.1 pCi/1, g, or sample
5 pCi/1
5%
5 pCi/1
5%
5 pCi/1
5%
5 pCi/1
5%
1.5 pCi/1
5%
5%
5 pCi/1
25%
5 pCi/1
5%
la = 16985 x (pCi/1)
10%
15%
10%
0.0933
These limits must be corrected appropriately if multiple determinations
are being made.
Source: Environmental Protection Agency. Environmental Radioactivity
Laboratory Intercomparison Studies Program, FY 1977.
EPA-600/4-77-001; Las Vegas, Nev., January 1977.
-------
4-6
Selection of a sufficient number of samples is necessary to determine
whether the data meet the established criteria. This selection must be a
compromise between the number of results necessary for statistical evalua-
tion and the amount of additional workload the laboratory can perform.
The ideal procedure would be to perform duplicate analyses on about 10
percent of the samples processed by the laboratory. This goal may be too
high for a laboratory that has a large workload; therefore, 20 sets of
duplicate data of each sample analysis type should be considered the
minimum amount of data to be statistically analyzed during a reporting
period. All laboratories should try to perform duplicate analysis on at
least five percent of its samples.
Accurate records should be kept for duplicate analyses. All factors
that might affect the analytical performance should be recorded. Informa-
tion should include at least the date, analyst, and instruments used.
Figure 3 is an example of a typical record or log of quality control
duplicates.
After these data are collected, a statistical range analysis can be
performed. One of the most common analytical techniques currently in use
is that suggested by Rosenstein and Goldin,6 in which a mean range (R)
between duplicate analyses is calculated from the standard deviation of
the analysis; this range is a function of concentration (see Table 1).
The formula for this calculation is R = d%(J, where d2 is a function of the
number of replicates involved (see Table 2) and a is the standard deviation,
as_derived from Table 1. The control limits can be calculated by R + 3a =
D,R = D4doCr, where
-------
4-7
QUALITY CONTROL REPLICATE ANALYSIS RECORD
Determination No. 1
Sample
Date
Counter
Analyst
Result
Determination No. 2
Date
Counter
Analyst
Result
Figure 3. Example of a typical record or log of quality control duplicaU
-------
-p-
CO
MEAN OF DUPLICATES
Figure 4. Plot of duplicate range vs. duplicate mean shows the effect of
increasing concentration on range values.
-------
4-9
Once the range limits are set up, results of the duplicate analyses
can be evaluated. The observed range between duplicates is classified as
<(R + O, <(R + 2a_). <(R + SOL,), or >(R + 3(0 (see Figure 1). The number
- i\ - K - K K
of duplicate analyses that fall into each category indicates the laboratory's
performance. Theoretically, the distribution should be as follows:
Range
<(R +
<(R +
<(R +
>(R +
QR)
2aR)
3aR)
3aR)
Percent of
84
97
100
0
results
.5
Using gross beta data from Table 3 and statistical parameters from
Tables 1 and 2, one can perform a sample evaluation. The standard devia-
tion for a single determination is 5 pCi/H; the standard deviation for a
duplicate determination is 5/^2 = 3.54 pCi/j£. Therefore, R = d£Cr =
1.128(3.54) = 3.99 pCi/£; and dR = R(D4 - l)/3 = (3.99)(3.267 - l)/3 = 3.01
pCi/£. From this the range limits may be determined:
<(R + aR) = <7.00 pci/JJ; (1)
; (2)
D
K
D) = <13.02 pci/£; (3)
-
>(R + 3a ) = <13.02 pci/£. (4)
Therefore, 29 duplicate measurements result in the following distribution:
Results
Number
Range (n=:29) Percentage
<(R + (O 26 90
<(R + 2aR)
<(R + 3cO
- K
>(R + 3(O
K
2
1
0
97
100
0
This agrees very favorably with the theoretical distribution shown earlier.
-------
4-10
TABLE 3. GROSS BETA IN WATER (pCi/£) DUPLICATE ANALYSIS DATA
First determination
Sample
number
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
pt
Date Counter
10/22
10/22
10/28
11/4
11/17
11/19
12/3
12/22
1/4
1/10
1/12
1/18
1/18
1/20
1/25
1/31
2/10
2/13
2/15
2/27
3/15
3/17
3/20
3/25
3/30
4/2
4/4
4/8
4/17
LB
LB
LB
LB
LB
LB
WE
LB
WB
WB
LB
LB
WB
WB
LB
LB
LB
LB
WB
WB
WB
LB
WB
LB
LB
WB
LB
LB
LB
Analyst
B
B
C
A
B
B
C
C
C
B
A
A
C
C
B
C
C
C
B
B
C
A
B
B
A
C
A
A
C
Result
1.7
2.3
1.9
9.7
4.5
3.3
3.5
3.7
1.0
23.7
1.8
1.9
4.3
7.7
0.9
1.0
54.7
2.7
6.9
2.1
7.2
2.7
5.8
2.3
2.4
11.1
4.1
1.9
52.0
Second determination
Date Counter
10/22
10/22
10/28
11/4
11/17
11/19
12/3
12/22
1/4
1/10
1/12
1/18
1/18
1/20
1/25
1/31
2/10
2/13
2/15
2/27
3/15
3/17
3/20
3/25
3/30
4/2
4/4
4/8
4/17
LB
LB
LB
LB
LB
LB
WB
LB
WB
WB
LB
LB
WB
WB
LB
LB
LB
LB
WB
WB
WB
LB
WB
LB
LB
WB
LB
LB
LB
Analyst
B
B
C
A
B
B
C
C
C
B
A
A
C
C
B
C
C
C
B
B
C
A
B
B
A
C
A
A
C
Result
2.0
2.7
2.2
1.4
3.1
3.1
3.2
4.4
3.1
12.2
2.1
2.5
4.5
6.8
0.6
1.6
55.5
0.9
0.1
2.3
6.8
2.6
4.2
2.5
3.3
4.0
4.1
4.0
51.1
0.3
0.4
0.3
8.3
0.2
0.2
0.3
0.7
2.1
11.5
0.3
0.6.
0.2
0.9
0.3
0.6
0.8
1.8
6.8
0.2
0.4
0.1
1.6
0.2
0.9
7.1
0.0
2.1
0.9
LB = Beckman Low Beta II; WB = Beckman Wide Beta II.
-------
4-11
Interpretations of observed results can yield several possibilities:
1. An acceptable distribution will either be close to that pre-
dicted by theory or shifted toward the (?„ and the 20^. categories.
K K
2. An unacceptable distribution will be shifted toward the 2aR,
3a_, and ^30,, categories.
3. An "outlier" distribution follows the theoretical distribution,
but has a few results in the >3o_ category. This indicates
acceptable precision with occasional poor results. However,
these others should be examined closely to determine source
because important knowledge of weaknesses in the analytical
system may be discovered.
These results may be used as a basis for further study of the analysis
system. However, the occurrence of an unacceptable distribution requires
that immediate steps be taken to correct the situation causing the poor
analytical results.
INTERNAL QUALITY CONTROL--DEMONSTRATING ACCURACY
A laboratory can establish the accuracy of its data by preparing special
standard samples containing known concentrations (spikes) of the nuclides of
interest and statistically studying the relation between the resulting
analytical values and the known concentrations. These spikes should be
analyzed regularly so that the laboratory has a running account of its
analytical ability.
This type of study requires quadruplicate analyses. The mean of the
determinations is calculated from the collected data, and the mean values
and individual values then are plotted on control charts.
Means Control Chart
The means control chart enables one to determine the amount of bias
between the mean and the expected value. A basic assumption is that the
resulting data will resemble a normal distribution over time. When pre-
paring such a control chart (Figure 5), one assumes the known activity
to be the mean. The allowable standard deviation for the individual
determinations is selected as for the duplicate studies (see Table 1);
the standard deviation of the mean is calculated by dividing the standard
deviation of the individual measurements by the square root of the number
of determinations (a = a/Vn). Acceptable results should fall within
±2am for 95 percent of the results. Any bias will show up as a consistent
displacement of results from the mean.
Individual Results Control Chart
The control chart for individual results (Figure 6) is essentially
the same as the means chart except that individual values, rather than
the mean, are used. This chart can show the actual effect of the indi-
vidual results on the mean values. In this chart the standard deviation
-------
m
m
X
m
m
Figure 5. Means control chart. X is mean, a = a
the abscissa is time.
n, the ordinate is concentration, and
-------
Figure 6. Individual results control chart. X is the known concentration (ordinate), a is the
standard deviation, and the abscissa is the run number.
-------
4-14
for the individual results, rather than for the mean, is used. Again,
the known value is assumed to be the central value.
A control chart for range also could be developed for use in these
statistical analyses; however, the range values determined from the
duplicate analyses program should describe the precision of the
laboratory adequately. Control charts for individual analysts and
counters may also be set up to break down the variables even further.
EXTERNAL QUALITY CONTROL--DEMONSTRATING ACCURACY
An internal program for testing accuracy should not be considered
sufficient for documentation. Participation in collaborative testing
programs with other laboratories is the best way for a laboratory to
evaluate its performance with respect to that of others.
The EPA conducts an Environmental Radioactivity Laboratory Inter-
comparison Studies Program in which all environmental laboratories should
participate. This program provides a large number of different crosschecks
regularly. Results from crosschecks are sent to EPA within four weeks
after receipt of the samples. After all results are tabulated, the EPA
prepares a statistical analysis of the data for each participant that shows
its performance with respect to the known values and the performance of
other participants. At this time, over 100 laboratories participate in
these studies to identify problems with procedures and instruments and to
demonstrate analytical accuracy.
Although other agencies, such as the International Atomic Energy
Agency, conduct crosscheck programs, most are not nearly as comprehensive
as the EPA program.
Programs of split-sample analysis between two or more independent
laboratories can also help establish analytical credibility and identify
analytical problem areas. In such a program, actual environmental samples
are collected and processed; thus, portions of the samples are sent to one
or more other laboratories for analysis. This type of program is the only
way for an environmental laboratory to estimate the accuracy of its results
at true environmental levels. Programs such as those conducted by EPA
cannot be carried out at the low levels found in the environment and there-
fore do not substitute for a split-sample program.
SUMMARY
If a laboratory conducts routine assessments of its precision and
accuracy, documentation of the quality of its work is possible. Also,
attention can be focused rapidly on analytical problems that otherwise
would go undetected. Therefore, quality control must be built into the
laboratory program just as any other aspect of analysis is incorporated
into the program.
-------
4-15
REFERENCES
1. American Chemical Society. Reagent Chemicals - American Chemical
Society Specifications. 4th ed. Washington, D.C., 1968. 651 pp.
2. Rosin, J. Reagent Chemicals and Standards. 5th ed. D. Van Nostrand Co.,
Princeton, N.J., 1967. 557 pp.
3. Silvey, W. D. Concentration Method for the Spectrochemical Deter-
mination of Minor Elements in Water. Water Supply Paper No. 1540-B,
U.S. Geological Survey, Washington, B.C.
4. Skoog, D. A. Fundamentals of Analytical Chemistry. Holt, Rinehart,
and Winston, Inc., New York, 1963. Chapter 3, 786 pp.
5. U.S. Environmental Protection Agency. Environmental Radioactivity
Laboratory Intercomparison Studies Program, FY 1977. EPA-600/4-77-001,
Las Vegas, Nev., January 1977. 17 pp.
6. Rosenstein, M., and A. S. Goldin. Statistical Technique for Quality
Control of Environmental Radioassay. AQCS Report Stat-1, U.S. Public
Health Service, Winchester, Mass., November 1964. 20 pp.
7. U.S. Environmental Protection Agency. Handbook for Analytical Quality
Control in Water and Wastewater Laboratories. Cincinnati, Ohio, 1972.
8. Bennett, C. A., and N. L. Franklin. Statistical Analysis in Chemistry
and the Chemical Industry. John Wiley & Sons, Inc., New York, 1954.
724 pp.
9. Enrick, N. L. Quality Control and Reliability. 6th ed. Industrial
Press, Inc., New York, 1972. 306 pp.
10. Grant, E. L. Statistical Quality Control. 9th ed. McGraw-Hill
Book Co., New York, 1972. 694 pp.
-------
-------
SECTION 5
DATA HANDLING AND REPORTING
GENERAL
After the laboratory has established control over sample collection,
instrumentation, and analytical methods, the remaining aspect of produc-
ing useful and accurate data on all samples is the handling and reporting
of data produced in the laboratory.
Prescribed procedures must be set up and controlled for the passage
of samples through the laboratory to ensure that the proper analyses
are assigned and completed. Control measures should be adopted to
ensure that correct values are reported for each analysis; the values
themselves must have a consistent form and be free of mathematical
error.
ANALYTICAL PROCESS
A system for controlling the passage of samples through the
laboratory must be established. Because the procedure described
here may not answer the needs of every laboratory, each laboratory
should develop its own methods according to its own needs and docu-
ment these methods in written procedures. A general outline of a
sample-handling process is described as follows.
1. As they arrive, samples are recorded in a logbook. At this time,
a laboratory number is assigned to the sample, and all descriptive
information concerning the sample, including sample source, sample,
quantity, collection time, and sample collector is recorded.
2. An analytical request sheet, which lists the sample number and
descriptive information about the sample, is prepared for the
sample. All necessary pretreatments and analyses to be performed
should appear on the sheet. Space is provided for completion
dates.
3. As results are completed, more information, such as the name of
the analyst who performed the work, his calculations for the
sample, and all pertinent data that would allow recalculation
of results at a later time if necessary, is noted on the result
sheets.
4. As various analyses are performed and results are reviewed by the
analyst, the request sheet is checked off and the result sheets
are attached to the analytical request sheet.
5. After all analyses are completed, the results are sent to the
laboratory supervisor or person responsible for reporting results.
After the results are reviewed, they are transferred to a final
report form.
-------
5-2
6. The analysis request form and result sheets are appropriately filed
for future reference. Several years of results should be kept
for ready reference.
This procedure will allow data to be tested for adequacy at any
point during the analysis and to be reexamined at any future time. The
data are reviewed thoroughly several times by designated personnel,
starting with the senior analyst, before being reported in final form.
In any analytical program, a senior analyst should be responsible for
recognizing anomalous results and discussing these with the appropriate
persons so that a sample can be reanalyzed before it is discarded or a
significant amount of time has passed.
DATA
Although calculations, whether performed by hand, calculator, or
computer, usually can generate more digits than are actually needed,
valid digits often are thrown away. For this reason the proper use of
significant figures should be emphasized.1'2 A brief list of rules
can be given for these evaluations.
1. Final zeros in a whole number may or may not be significant. In the
measurement 1800 mm, do the zeros signify that the length was measured
to 1 mm or do they merely locate the decimal point (i.e., to distinguish
1800 from 18 mm)? To avoid confusion in cases of this type, use a
larger unit. If one chooses the meter as the unit, then 1800 mm
becomes 1.8, 1.80, or 1.800 m, depending on the accuracy of the
measurement.
2. Zeros before a decimal point with other preceding digits are signi-
ficant. With no preceding digits, a zero before the decimal point
is not significant. For example, in the measurement 150.12, the
zero is significant, but in the measurement 0.12, the zero is not
significant.
3. If there are no digits preceding a decimal point, the zeros
following the decimal point but preceding other digits are not
significant; these zeros only indicate the position of the decimal
point. Thus, in the measurement 0.050015 kg, the first two zeros
are not significant and serve only to locate the decimal point;
however, the zeros between 5 and 1 are significant.
4. Final zeros after a decimal point are always significant figures.
A weight such as 7.530 g indicates that the measurement was made
to the nearest milligram.
5. To round off a figure, if the digit following those to be retained
is less than 5, the digit is dropped and the retained digits are
not changed.
-------
5-3
6. To round off a figure, if the digit following those to be retained
is greater than 5, the digit is dropped and the last retained
digit is raised by 1.
7. To round off a figure, if the digit following those to be retained
is 5 and there are no digits following it except zeros, then the
last retained digit is raised by 1 if it is odd and kept unchanged
if it is even.
Good books concerning quantitative analysis should be consulted for
discussion of how to propagate rounding through arithmetic calculation
as well as detailed discussion on the rules above. Generally, analytical
results should not be reported to more than four significant figures.
ERRORS CAUSED BY COMPUTATIONAL PROCESSES
Two types of errors, rounding and truncation, may occur in any
numerical process.
Truncation errors result from the necessary termination of an
otherwise naturally infinite (or very lengthy) process. In most compu-
tations, it is neither possible nor necessary to carry an infinite number
of significant figures. Therefore, certain classes of numbers, functions,
or constants will never be used with exact accuracy. Included are
irrational numbers (V2), transcendental numbers (TT or e) or functions
(logarithm, exponential, sine), and fractions that have no terminating
representation. With regard to this last category, fractions that have
an exact representation in one number base may not possess this property
in another. For example, in the decimal system used in manual calculations
and by the majority of hand-held and desktop calculators, the fraction 1/10
is accurately expressed as 0.1. In the binary system used by most large-
scale computers, however, the fraction has no terminating representation
and cannot be expressed exactly. Therefore, the sum of 10 numbers, each
expressed as binary representation of 0.1, will not necessarily result
in 1.0.
Rounding and chopping (truncation) errors are generally considered
to be a troublesome problem when using a computer. By rounding, we refer
to the symmetrical rounding procedure discussed previously. Rounding may
best be compared with the chopping process by an example: As the result
of some numerical operation, we obtain the number 0.41877 and wish to
express it to four significant digits; by rounding, 0.41877 becomes
0.4188, but by chopping, 0.41877 becomes 0.4187. In chopping, the digits
beyond those that have been declared significant are ignored and have no
effect on the final resulting number.
Because computers generally have a fixed word length and cannot
express any number with infinite precision, some decision must be made
as to what to do when the limits of precision within the computer are
reached. A large number of FORTRAN compilers do, in fact, set up the
object program to use chopping, which introduces less error in the
-------
5-4
calculated result than the familiar rule for rounding. In addition, the
use of the familiar symmetrical rounding procedure in every arithmetric
operation, including the many places in a program in which it is not
really necessary, would waste computer time.
For computers and calculators, rounding and chopping may introduce
errors not only in internal calculations, but also in displayed or printed
output. While the machine internally may carry out arithmetic operations
with precision ranging from 8 to 16 decimal digits, this degree of pre-
cision is generally not necessary for display or printout of final results.
For straight-line calculations, rounding or chopping the final answer to
four significant figures may introduce a relative error that is greater
by several orders of magnitude than any error accumulated during the
calculation. Also, chopping of the final displayed result by the computer
does introduce more error in the displayed decimal than would rounding.
Recursive or iterative operations suffer more from the effects of
rounding or chopping than do straight-line calculations. This problem
is typified in the resolution of multicomponent gamma ray spectra, where
a set of simultaneous equations must be solved. In general, there are
two types of numerical techniques for solving simultaneous equations:
direct methods, which are finite (Gaussian elimination), and indirect
methods, which are infinite (iterative techniques). Obviously, no prac-
tical technique can actually be infinite; what is meant is that the direct
methods will produce, in principle (neglecting rounding errors), an exact
solution, if one exists, in a finite number of arithmetic operations. An
indirect method, on the other hand, would in principle require an infinite
number of arithmetic operations to produce an exact solution. That is,
an indirect method has a truncation error, whereas a direct method does
not.
Rounding errors may not be neglected. In a large, ill-conditioned
system, the rounding errors in a direct method may make the solution
meaningless; the magnitude of the final error may, in severe cases, be
larger than the derived result. Therefore, in spite of its truncation
error, an indirect method may be much more desirable because the problem
of accumulated rounding error is minimized.
DATA STORAGE
If the analytical data reported by the laboratory have as their final
destination a computerized data base, then a great deal of care must be
exercised to ensure that no transcription errors are generated. Because
every transposition of data increases the possibility of such errors, the
number of intermediate transcriptions of data should be limited. Procedures
should be set up for random checking of the data base to ensure the quality
of the stored data.
-------
5-5
REFERENCES
1. Skoog, D. A., and D. M. West. Fundamentals of Analytical Chemistry.
2nd ed. Holt, Rinehart, and Winston, Inc., New York, 1969. 853 pp.
2. Pierce, W. C., D. T. Turner, and E. L. Haenisch. Quantitative
Analysis. 4th ed. John Wiley & Sons, Inc., New York, 1958. 647 pp.
-------
-------
SECTION 6
STATISTICS AND COUNTING DATA
GENERAL
Statistics, as a science, deals with probability. Probability is
a useful concept in all areas of science because no two analyses or
measurements give identical answers. Measurements of radioactivity
are bound even more closely to statistical evaluation because of the
random nature of the disintegration process. The basic laboratory
question is how to determine, from a few measurements or a single
measurement, the best approximation to a "true" value.
The goal of this manual has been to examine areas and to demonstrate
methods that can be used to reduce systematic errors to the point that
only the small unavoidable fluctuations of random error are present. Of
course, every analytical result still has a degree of uncertainty. These
uncertainties arise from all the measurement processes involved in the
analysis, such as weighing, volumetric processes, and counting error. An
analyst must try to estimate this error (uncertainty) for each single
analysis by propagating the random error in all measurements for each
individual sample. Measuring radioactivity involves a statistical evalua-
tion of counting measurements to estimate these uncertainties and reduce
them to as small a figure as possible.
The lower the level of activity in a sample, the more difficult it
is to distinguish the activity of the sample measurement from statistically
allowable fluctuations in the background. To evaluate sample measurements,
one must be able to estimate the sensitivity of the analysis, which is
sometimes called lower limit of detection (LLD) or minimum detectable
activity (MDA). This section describes how sensitivities can be estimated
for some analytical methods.
This section by no means is a treatise on statistics.
COUNTING STATISTICS
Two types of statistical distributions are used most often in the
radioanalytical laboratory: (1) The Gaussian, or normal, distribution
is used to describe continuous variables, and (2) the Poisson distribu-
tion applies to discrete variables.1"4
The form of the Gaussian distribution for n measured values is
f(x) =
where
a
exp
(x -
(5)
[j = mean of the measurement values (x) ,
a = standard deviation = [I(x - [j)2/(n -
X
-------
6-2
This equation describes a bell-shaped curve, any area under which
is related to the probability of a particular result. This area is
often divided into ranges of a about the mean, (J. Thus,
Area
(J ± a
x
H ± 2a
x
(j ± 3a
x
Probability
0
0
0
.683
.955
.997
This division into regions is the source of terms such as 2a, 2 sigma
level, and 95 percent confidence level. Gaussian statistics are used in
the radioanalytical laboratory to describe the behavior of multiple
measurements of a single value.
Multiple measurements are seldom possible for most work performed in
the radioanalytical laboratory. Availability of instruments for repeated
countings is too expensive to maintain. For this reason, the use of Poisson
statistics can allow an estimate of behavior from a single measurement.
The form of the Poisson distribution is expressed as
f(x) = (e'X)(x)X) , (6)
x!
where x = 0, 1, 2,...,
x = estimated mean.
The standard deviation, s, for the estimated mean is
EXAMPLE 1: A standard check source was counted to determine the
counter efficiency. The following one-minute counts were observed:
7747, 7738, 7840, 7785, 7705, 7667, 7812, 7827, 7623, and 7739. The
total count was 77483, and the mean was calculated to be |J = 77483/10
7748. Thus, for the Gaussian distribution,
a = [Z(x - M)2/(n - l)2 - 70, (7)
and for the Poisson distribution,
s = Vx~- 88. (8)
The agreement of the two estimates indicates that the counting pro-
ceeded satisfactorily. The Gaussian estimate and the Poisson estimate
should be approximately the same. The agreement of the results also shows
that the Poisson estimate is a quick but satisfactory measure of counting
data .
-------
6-3
For situations in which the standard deviation is of the same order
of magnitude as the mean, the distribution of data will be markedly non-
Gaussian, and Gaussian probability levels and also propagation of errors
give grossly biased answers. The only way to eliminate the bias is a
very tedious trial-and-error use of the Poisson distribution itself.
The Poisson standard deviation often appears as slightly different
expressions from that shown above when converted to a counts-per-minute
(counts/min) figure:
s . / • = Vx/t
counts/mm
= V57t , (9)
where t •= counting time,
R = counting rate.
PROPAGATION OF ERRORS
If all radioanalytical analyses involved only one counting measurement,
estimating the uncertainty of a final result by applying Poisson statistics
to obtain a standard deviation would be quite simple. However, this is
seldom, if ever, the case since almost all counting data must be corrected
for background contributions. Both the sample and background counts have
uncertainties that must be reflected in the final result. The theory of
propagation of errors can be applied to estimate the reliability of the
final calculated result.5'6
If Q = f(X, ¥,...)> where X,Y,... are independent, normally distributed
variables, the asymptotic uncertainty (variance) in Q resulting from uncer-
tainties in X,Y,... is given by the expression
a 2 = OQ/9X)2aY2 + (9Q/3Y)2a 2 + ... (10)
v x y
The standard deviation for Q is found by taking the square root of the
expression. For the specific case of the error of difference between two
counting determinations (Q = X - Y) , the proper equation is
where an = error of the difference,
a = estimated standard deviation for the sample,
S
a = estimated standard deviation for the background.
0
Formulas for other simple functions are shown in Table 4.
-------
6-4
TABLE 4. FORMULAE OF PROPAGATION OF ERROR
FOR SOME SIMPLE FUNCTIONS
Function Error formula
Q - X ± Y an = (Q 2 + a 2) %
Q x y
Q = aX ± bY an = (a2a 2 + b2a 2) \
Q x y
Q = XY a = XY(ax2/X2 + a 2/Y2)^
Q = XY a = x/Y(a 2/X2 + a 2/Y2)'
Q x y
Q = Xn aQ = n(Xn-\)
Q = In X a = a /X
x X
Q = log X a = 0.434ax/X
Sources: Overman, R. T., and H. M. Clark. Radioisotope Techniques.
McGraw-Hill Book Co., New York, 1960. 476 pp.
Ku, H. H. J. Res. Nat. Bur. Stand. Sect. C, 70: 263, 1966
-------
5-5
EXAMPLE 2: A sample counted for 50 minutes gave a gross count of
123 counts; the background for the same period was 68 counts. The error
of the net count is shown below:
a = 123^ = 11.09 (a 2 = 123) ; (12)
s "
0D = 68^ = 8.25 (0B2 = 68) ; (13)
D D
a = (68 + 123)^ = 13.82 . (14)
Therefore, the net count is 55±14.
A more common expression for this error term, when both a sample
measurement and background measurement are involved, is
where R = gross sample counting rate,
S
R
R = background counting rate,
t = sample counting time,
s
tR = background counting time.
EXAMPLE 3: A sample counted for 50 minutes gave a gross count of
137 counts. The background for a 10-minute count was 15 counts.
_ /137/50 + 15/10\ h
CTQ - \~50~~ "TO")
- (0.055 + 0.15)^ = (0.205)^
= 0.45 counts/min.
Therefore, the net counting rate is 1.74±0.45 counts/min (neglecting signifi-
cant figures). Because the gross counting rate is the sum of the background
and sample counting rates, the background will dictate the precision of the
final answer if background and sample counting times are equal. Therefore,
counting times must be selected to minimize the background contribution to
the error term.
Several possibilities exist for reducing the counting error for a
particular analysis: (1) increasing the counter efficiency; (2) increasing
the sample size; (3) increasing the counting time; and (4) lowering the back-
ground. Increasing the efficiency or sample size gives the greatest benefit
since these terms are applied directly in all final calculations. Possibili-
ties (3) and (4) vary with the square root and therefore yield a lesser
reduction in the size of the error term.
-------
6-6
When using background measurements in analytical calculations, a
figure representing the background over a long period of time, rather
than figures from daily measurements, is preferable. For example, back-
ground of 1.2710.10 counts/min averaged over a year is a much better value
for calculation purposes than a value of 1.33 counts/min determined for a
particular day. Although using a long-term average in calculations will
tend to smooth out fluctuations, an average figure should not, of course,
be used when a radical change in background has occurred.
A test often used to evaluate instrument background measurements to
determine whether the instrument is performing as statistically expected
is the Chi-square test, in which Chi-square (x2) is calculated by
.X2 = [ I (x, - x)2]/x, (17)
i=l
where x. = individual result,
x = average of all the measurements.
Values of Chi-square have been calculated and tabulated for the number of
values measured.
Number of
measurements
5
10
15
20
30
Allowed x^
limits
0.3 -
2 -
4 -
7 -
14 -
13
22
29
36
50
A value outside the limits indicates that the instrument is not performing
as expected from statistical considerations.
LIMITS OF DETECTION
An examination of equation (15) shows that the percent error increases
as the gross counting rate approaches the background counting rate. Neither
the gross counts nor the background counts are hard, fast numbers; instead,
they are members of two distributions that begin to overlap when the sample
has very low amounts of activity above background. Estimating the reliability
of the difference between the two measurements becomes a complex problem when
overlap occurs.
-------
6-7
Several methods for estimating the LLD or MDA in radioanalytical
measurements have been proposed.6'7 The method most commonly used assumes
that, unless the gross counting rate is greater than the background count-
ing rate plus two standard deviations, the amount of activity contained in
the sample is less than detectable.
EXAMPLE 4: A low-background beta counter has an average background of
64 counts in 50 minutes measured over a period of a year.
s = x = V64 = 8,
2s = 16 counts.
(18)
In this example, the gross counts must exceed 80 counts in 50 minutes to
give 95 percent confidence that there is actually any activity present in
the sample.
Another approach to this problem is presented in Refs. 8 and 9. This
procedure uses a statistical technique known as hypothesis testing, in which
two types of error are assumed to be possible: (1) concluding that there is
sample activity when there is none (Type I error) and (2) concluding that
there is no sample activity when there is some (Type II error). The terms
alpha (a) and beta (p) represent Type I and Type II, respectively; alpha is
usually allowed to be 5 percent (0.05) and beta is usually set at 5 percent
(0.05). The LLD is then approximated as
LLD = (ka+kp)so
(19)
where
k = the value of the upper percentile of the standardized
normal variate corresponding to the preselected a,
k,, = the corresponding value for the predetermined degree
of confidence for detecting the presence of activity
(1 - P),
s = the estimated standard error for the net sample count-
ing.
If the values of a and p are set at the same level (0.05) and the sample
and background counts are close, as would be the case at the LLD, other
approximations also may be made. If
net
'B'
(20)
and
k = k = k
a
(21)
then
LLD = 2^2
(22)
-------
6-8
a
0.01
0.02
0.05
0.10
0.20
1 - p
0.99
0.98
0.95
0.90
0.80
k
2.327
2.054
1.645
1.282
0.842
For the previous example, we can calculate the LLD by this method using
a = p = 0.05 to find the background variance:
LLD - 2,/2 ksB ; (23)
s = /~x = 764 = 8 ; (24)
LLD = 2^2 (1.645) (8) - 37 counts. (25)
This value is the number of net counts above background that must be
observed before one can report a result above LLD. Note that this is
a more conservative estimate of the LLD than that shown in example 4.
This method of LLD calculation is becoming the most accepted procedure.
This 95 percent confidence criterion is a very rigorous one. A more
liberal criterion might be to allow a 20 percent chance of erroneously
reporting activity when none is actually present and to keep the 5
percent requirements for not reporting activity when it is actually
present. This change in criteria would lead to the following expression:
LLD = (k + k ) V2 SD (26)
ri
= (1.645 + 0.842) ^2 sro
B
= (2.487) V2 SB
= 3.52sB .
For the previous example, the LLD calculated by this method x^ould be 28
net counts .
The extraction of results from allowable variations can be difficult
and always introduces a degree of uncertainty in the final result. This
final result must be measured in light of the LLD levels before attaching
meaning or significance to the value. Allowable variations and LLD levels
also must be considered when formulating quality control analysis of
laboratory data.
-------
6-9
Many other authors10 12 have approached the problem of LLDs with
reference to radioanalytical measurements.
REFERENCES
The first four references are general in nature and apply to all areas
of statistics involved in radiochemistry.
1. Dixon, W. J., and F. J. Massey. Introduction to Statistical Analy-
sis. 3rd ed. McGraw-Hill Book Co., New York, 1969. 488 pp.
2. Evans, R. D. The Atomic Nucleus. McGraw-Hill Book Co., New York,
1955. 952 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. Jarrett, A. A. Statistical Methods Used in the Measurement of
Radioactivity with Some Useful Graphs and Monographs. AECU-262,
U.S. Atomic Energy Commission, Oak Ridge, Tenn., 1961. 42 pp.
5. Crumpler, T. B., and J. H. Yoe. Chemical Computations and Errors.
John Wiley & Sons, Inc., New York, 1940. 247 pp.
6. Currie, L. A. Anal Chem., 40:586, 1968.
7. Pasternack, B. S., and N. H. Harley. Nucl. Instrum. Methods,
91:533, 1961.
8. U.S. Atomic Energy Commission. HASL Procedures Manual. HASL-300,
New York, 1972.
9. Altshuler, B., and B. Pasternack. Health Phys., 9:293, 1963.
10. Hartwell, J. K. Detection Limits for Radioisotopic Counting Tech-
niques. Document No. ARH-2537, Atlantic Richfield Hanford Company,
Hanford, Wash., June 22, 1972. 23 pp.
11. Donn, J. J., and R. L. Wolke. Health Phys., 32:1-14, 1977.
12. Corley, J. P., D. H. Denham, D. E. Michels, A. R. Olsen, and
D. A. Waite. A Guide for Environmental Radiological Surveillance
at ERDA Installations. ERDA 11-Ik, Energy Research and Development
Administration, Washington, D.C., March 1977.
-------
-------
SECTION 7
HANDLING STANDARDS OF RADIOACTIVITY
GENERAL
Every laboratory involved in radioanalytical determinations uses
radioactive standards for calibration and procedures testing. The
first requirement is to acquire standards that are well characterized
and documented. After standards are received, they should be properly
stored and diluted to ensure that the standardization is not invalidated
by improper handling.
STORAGE OF STANDARD SOLUTIONS OF RADIOACTIVE MATERIALS
There are several ways in which the calibration of a radioactivity
standard can change by 10% or more. Because degradation of a standard
is often difficult to detect, it should be avoided by compliance with
proper handling procedures. The most common ways by which the standard-
ization may change are (1) adsorption of trace metals on the walls of
the container, (2) precipitation, (3) loss of water by evaporation,
(4) biological activity in the solution, and (5) radioactive decay.
Adsorption or precipitation is prevented by the addition of acid,
complexing agents, or carriers to the solutions. The supplier of the
standard should furnish the chemical composition of the nuclide solu-
tion with his certification. Radionuclide standards should not be stored
in plastic bottles for more than a few days since water can evaporate
through the bottle walls; storage in glass bottles is preferred. To
prevent the growth of biological material in standards, add a few drops
of either formaldehyde or chloroform. Standards should not be used beyond
four half-lives of the radionuclides.
DILUTION OF STANDARD SOLUTIONS
Standard radionuclide solutions usually are calibrated by weight.
Since weighing is more accurate and normally more convenient than measure-
ments by volume, weighing is the preferred method for diluting standards.
An additional benefit of this method is the fact that weighing eliminates
the necessity of making density corrections when acid or carrier is
present.
All normal weighing precautions should be exercised when preparing
standards. Special consideration should be taken to ensure that thermal
equilibrium and constant weights are obtained.
The volumetric flask used for the dilution should be rinsed care-
fully with the diluting solution containing carrier, acid, or complexing
agent and then partially filled. After the standard solution is pipetted
accurately into the flask, additional diluting solution is added to bring
-------
7-2
the flask to the mark. The flask then should be sealed with a stopper
of ground glass, and the solution should be mixed thoroughly. When
the solution is not being used, the stopper should be kept in place to
prevent evaporation.
Pipets used for these standards can be either lambda (|j£) pipets
made of glass or fixed-volume pipets with disposable tips. The glass
pipets should be stored and cleaned separately from other pipets used
in the laboratory to prevent contamination. Each pipet should be labeled
to identify the radionuclide for which it has been used in the past.
USING CALIBRATED SOURCES
The accuracy with which the half-life and purity of the source are
known will affect the accuracy of calculations when certified emission
rates are used. Also, knowledge of the calibration geometry, instrumenta-
tion, and efficiencies occasionally may be necessary.
The rate of emission from the surface of a solid source may not be
the same as the disintegration rate. The possible existence of a complex
decay scheme for the nuclide and the phenomena of adsorption and scatter-
ing of radiation within the source both affect this relationship.
Variations in shape, size, or composition between a calibrated
source and the source to be measured may imply different detection
sensitivities for the two sources because of the possible differences in
adsorption or scattering of radiation. These effects must be minimized
and accounted for in estimating the accuracy of the unknown source.
REFERENCES
1. National Academy of Sciences, National Research Council. Users'
Guides for Radioactivity Standards. NAS-NS-3115, Washington,
B.C., February 1974. 85 pp.
2. Mann, W. B., and H. H. Seliger. Preparation, Maintenance, and
Application of Standards of Radioactivity. NBS Circ. 594,
National Bureau of Standards, Washington, B.C., 1958. 47 pp.
3. U.S. Public Health Service. Recommendations for the Storage
and Dilution of Radioactivity Standard Solutions - Notes on
Pipettes and Pipetting. Technical Report 69-TEC-l, January
1969. 10 pp.
4. International Commission on Radiation Units and Measurements.
Measurement of Low-Level Radioactivity. ICRU Report 22,
Washington, B.C., June 1, 1972. 66 pp.
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-60Q/7-77-088
3. RECIPIENT'S ACCESSION-NO.
4. TITLE ANDSUBTITLE
HANDBOOK FOR ANALYTICAL QUALITY CONTROL IN
RADIOANALYTICAL LABORATORIES
5. REPORT DATE
August 1977
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
L. G. Kanipe
8. PERFORMING ORGANIZATION REPORT NO.
E-EP/77-4
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Division of Environmental Planning
Tennessee Valley Authority
Chattanooga, TN 37401
10. PROGRAM ELEMENT NO.
INE-625C
11. CONTRACT/GRANT NO.
78 BDI
12 SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research & Development
Office of Energy, Minerals & Industry
Washinaton, B.C. 20460
13. TYPE OF REPORT AND PERIOD COVERED
Milestone
14. SPONSORING AGENCY CODE
EPA/600/17
15. SUPPLEMENTARY NOTES
This project is part of the EPA-planned and coordinated Federal Interagency
Energy/Environment R&D Program.
16. ABSTRACT
Quality control in the radioanalytical laboratory is discussed. The
discussion includes laboratory operating practices, analytical methodology,
instrument quality control,Nand data handling and reporting. Two other
sections on handling radioactive materials and counting statistics are
included; these sections are brief and serve only as an introduction to
the subjects.
The handbook provides methods for conducting internal and external quality
control programs. Topics such as control charts, duplicate analyses, and
routine spiked analyses are brought out.
17.
(Circle One or More)
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Ecology
^Environments'^
gfnvironmcntal Engineering)
Geography
Hydrology. Limnology
Biochemistry
Earth Hydrosphere
Combustion
Refining
Energy Conversion
Physical Chemistry
Materials Handling
Inorganic Chemistry
Organic Chemistry
Chemical Engineering
6F 8A 8F
8H 10A 10B
7B 7C 13B
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSIFIED
21. NO. OF PAGES
60
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
------- |