Spectrochimica Acta Part B 66 (2011) 193–241

Contents lists available at ScienceDirect

Spectrochimica Acta Part B

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s a b

Review

Neutron activation analysis: A primary method of measurement☆

Robert R. Greenberg a,⁎, Peter Bode b, Elisabete A. De Nadai Fernandes c
a
National Institute of Standards and Technology, Gaithersburg, MD 20899-8395, USA
b
Delft University of Technology, Delft, The Netherlands
c
Centro de Energia Nuclear na Agricultura, Universidade de São Paulo, Piracicaba, SP, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Neutron activation analysis (NAA), based on the comparator method, has the potential to fulfill the
Received 12 September 2010 requirements of a primary ratio method as defined in 1998 by the Comité Consultatif pour la Quantité de
Accepted 23 December 2010 Matière — Métrologie en Chimie (CCQM, Consultative Committee on Amount of Substance — Metrology in
Available online 27 January 2011
Chemistry). This thesis is evidenced in this paper in three chapters by: demonstration that the method is fully
physically and chemically understood; that a measurement equation can be written down in which the values
Keywords:
Neutron activation analysis
of all parameters have dimensions in SI units and thus having the potential for metrological traceability to
Metrology these units; that all contributions to uncertainty of measurement can be quantitatively evaluated,
Primary method of measurement underpinning the metrological traceability; and that the performance of NAA in CCQM key-comparisons of
Uncertainty budget trace elements in complex matrices between 2000 and 2007 is similar to the performance of Isotope Dilution
Metrological traceability Mass Spectrometry (IDMS), which had been formerly designated by the CCQM as a primary ratio method.
Published by Elsevier B.V.

Background methods in the sense that a complete mathematical description of the

The Comité Consultatif pour la Quantité de Matière — Métrologie
en Chimie (CCQM, Consultative Committee on Amount of Substance —
Metrology in Chemistry) has defined [Minutes from the Fifth Meeting
(February 1998) of the CCQM, held at the Bureau International des
Poids et Mésures (BIPM), Sèvres, France] a primary method as: “A
primary method of measurement is a method having the highest
metrological properties, whose operation can be completely de-
scribed and understood, for which a complete uncertainty statement
can be written down in terms of SI units”. More specifically, a primary
direct method was defined as “a primary method of measurement
that measures the value of an unknown without reference to a
standard of the same quantity”; and a primary ratio method as “a
primary method of measurement that measures the value of a ratio of
an unknown to a standard of the same quantity; its operation must be
completely described by a measurement equation.”
The draft report of a CCQM external peer review panel for the
National Institute of Standards and Technology (NIST) Chemical
Metrology Program (October 1999) stated: “… The discussion of
sources of error … provided strong evidence that, under certain
circumstances, nuclear methods have the potential to be “primary”
☆ With contributions by: Richard M. Lindstrom, Elizabeth A. Mackey, Rolf Zeisler and
Márcio Arruda Bacchi.
⁎ Corresponding author. Tel.: +1 301 975 6285; fax: +1 301 208 9279.
E-mail addresses: robert.greenberg@nist.gov, bob@greenbe.net (R.R. Greenberg),
p.bode@tudelft.nl (P. Bode), lis@cena.usp.br (E.A. De Nadai Fernandes).
method, and a complete uncertainty budget, can be developed. This
hypothesis should be further tested at the international level, e.g.,
through a series of CCQM “pilot” exercises which might eventually
lead to one or more key comparisons. In CCQM discussions of so-
called “primary” methods, the only technique relevant to the
production of certified reference materials for inorganic trace analysis
is Isotope Dilution Mass Spectrometry (IDMS). There would be great
benefit in having another method, in particular one which does not
require prior dissolution of solid samples.”
The metrological basis for neutron activation analysis (NAA) was
firmly established in the mid-to-late 1990s, although the fundamental
research was largely completed earlier. NAA was proposed by P. Bode,
R. R. Greenberg and E. A. De Nadai Fernandes as a primary ratio
method to the CCQM as part of the Primary Methods Symposium at
the April 2000 meeting of the CCQM. However, despite previously
identifying 5 potentially primary methods, the CCQM declined to
identify and add additional primary methods at that time. Since April
2000, various forms of NAA have been used in a large number of CCQM
key comparisons and pilot studies.
In a meeting of the Inorganic Analysis Working Group (IAWG) of
the CCQM on April 17, 2007, P. Bode, R. R. Greenberg and E. A. De
Nadai Fernandes sustained their thesis that NAA based on the
comparator method has the potentials fulfilling the requirements of
a primary ratio method with evidence on the methods' metrological
fundamentals including the measurement equation, the evaluation
and quantitation of all sources of uncertainty, the aspects of
metrological traceability and the performance of NAA in CCQM pilot
studies and key comparisons between the years 2000 and 2007.

0584-8547/$ – see front matter. Published by Elsevier B.V.

doi:10.1016/j.sab.2010.12.011
194 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Subsequently, the IAWG agreed accepting the thesis and proposed the
CCQM recognizing NAA as having the potentials of a primary ratio
method. The CCQM, in its meeting of April 19, 2007, has followed this
suggestion, and in the minutes of that meeting, approved on March
31, 2008, it is explicitly stated that “…The President said that although
the CCQM had not discussed the list of ‘potentially primary methods’
Chapter 1. Neutron activation analysis: Principles and analytical characteristics
Peter Bode ⁎
Delft University of Technology, Faculty of Applied Sciences, Department of Radiation, Radionuclides and Reactors, Section Radiation and Isotopes for Health, Reactor Institute Delft,
Mekelweg 15, 2629JB Delft, The Netherlands
1.1. Introduction
Activation analysis is a method for the determination of elements
based upon the conversion of stable nuclei to other, mostly
radioactive nuclei via nuclear reactions, and measurement of the
reaction products. In neutron activation analysis (NAA) the nuclear
reactions occur via bombarding the material to be analyzed with
neutrons. The reaction products to be measured are either the
radiation, released nearly instantaneously upon neutron capture; or, if
the resulting new nuclei are radioactive, the induced radioactivity by
which they decay. Only the latter mode will here be discussed since it
is the most common way to perform NAA.
All of the stable elements have properties suitable for production of
radioactive isotopes albeit at different reaction rates. Each radionuclide
is uniquely characterized by its decay constant – the probability for the
nuclear decay in unit time – and the type and energy of the emitted
radiation. Among the several types of radiation that can be emitted,
gamma-radiation offers the best characteristics for the selective and
simultaneous detection of radionuclides and thus of elements.
The activation will result in a mixture of radioactivities, which can
be analyzed for individual contributions by two approaches:
(i) The resulting radioactive sample is decomposed, and through
chemical separations it is divided into fractions with a few
elements each: Destructive or Radiochemical Neutron Activa-
tion Analysis (RNAA).
(ii) The resulting radioactive sample is kept intact, and the elements
are determined by taking advantage of the differences in decay
rates via measurements at different decay intervals utilizing
equipment with a high energy resolution: Non-destructive or
Instrumental Neutron Activation Analysis (INAA).
A procedure in INAA is characterized by (i) activation via
irradiation with reactor neutrons, (ii) measurement of the gamma-
radiation after one or more decay times and (iii) interpretation of the
resulting gamma-ray spectra in terms of radionuclides, associated
elements and their mass fractions.
1.2. History
After the discovery of the neutron by Chadwick in 1932, Frédéric
and Irène Joliot Curie discovered in 1934 the induced radioactivity.
Enrico Fermi was one of the first in the 1930s using neutrons to
produce artificial radioactive isotopes with George de Hevesy
following this direction of research for studies on the rare earth
elements. Results of these experiments on the action of neutrons on
the rare earth elements by George de Hevesy and Hilde Levi in 1935
* E -mail address: p.bode@tudelft.nl.
during recent years, it was recognised that NAA had claims to a similar
status to that of the five methods listed originally by the CCQM and
that NAA will be added to that list….”
The considerations leading to the three presentations given on
April 17, 2007 and the thesis of NAA being a primary ratio method are
described in the three chapters in this paper.
marked the onset of neutron activation analysis. They made their own
neutron source by pulverizing beryllium in an agate mortar by hand
and mixing it in a glass ampoule with 222Rn (half life 3.8 days) from a
solution of radium salt, which was available at the cancer hospital
in Copenhagen [1]. The α particles (4He nuclei), emitted by 222Rn,
cause a nuclear reaction in beryllium with the emission of neutrons:
4
He + 9Be → 12C + 1n. Once they observed the large differences in
induced radioactivity between the various (pure) lanthanides, and
especially the very high induced radioactivity of dysprosium, De
Hevesy realized that even the smallest amount of dysprosium might
be detected as impurity in any other of the rare earth elements. The
proof of principle was given in their 1936 publication [2] by
demonstrating the ability of determining the dysprosium content of
yttrium preparations. This was followed by, according to Levi [1], the
really decisive analytical experiment of determining the europium
impurities in gadolinium samples [3]. George De Hevesy and Hilde
Levi stated in their first publication [2]:
“… The usual chemical methods of analysis fail, as is well known, for
most of the rare earth elements and have to be replaced by
spectroscopic, X-ray, and magnetic methods. The latter methods can
now be supplemented by the application of neutrons to analytical
problems by making use of the artificial radioactivity and of the great
absorbing power of some of the rare earth elements for slow neutrons….”
In 1949 Boyd [4] suggested to term the procedure “… the method
of radioactivation analysis …”, or, more succinctly, “… activation
analysis…” and he discussed the use of the “… chain-reacting pile …”
as a source of neutrons and an example of an analysis was given. In the
same year Brown and Goldberg [5] gave a new impulse to this method
by demonstrating that with the “… neutron pile (nuclear reactor)…
very high specific activities …” could be obtained – as compared with
isotopic neutron sources used so far – thereby extending the
capabilities of NAA. Kohman reported already also in 1949 on errors
in the measurement of radioactivity [6], identifying problems such as
sample self-absorption and measurement geometries.
The potentials of NAA method were explored in the 1950s and
1960s, initially by the analysis of the decay and characteristic
absorption of the radiation emitted, and/or radiochemical separations.
Gamma-ray spectrometry with scintillation detectors gave a great
impulse to the improvement of the radionuclide – and thus element –
selectivity of the technique. Chemical separations of the mixture of
radioactivities were in those years still a necessity in many
applications for selective determination of the element(s) of interest.
The introduction of the semiconductor detector in the early 1960s
increased the selectivity in gamma-ray spectrometry with orders of
magnitude, and to such an extent that the individual radionuclides
could be identified in a mixture of radioactivities directly, without the
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
need for chemical separations. It marked the onset of INAA as
principal form of neutron activation analysis which development was
further stimulated by the availability of multi-channel pulse height
analyzers and laboratory computers.
In the 1956 bi-annual review ‘Nucleonics’ [7] of the journal
Analytical Chemistry, the first papers are cited dealing with sources of
error in activation analysis, some of them already published in 1954
[8–11]. Aspects of neutron fluence rate perturbation, geometry errors,
counting efficiency are discussed and it is described how to minimize
errors. The first practical solutions to e.g. effects of neutron self-
shielding and gamma-ray self-attenuation came also in the 1950s era,
e.g. by the introduction of the internal standard method by Leliart et.
al. [12]. In 1960, at the Vienna Radioactivation Analysis Symposium
[13], Cook presented, among others, many potential sources of error
and the documented lively discussion indicates the already estab-
lished quality assurance awareness.
“Errors in activation analysis” is one of the chapters in the first
book, published in 1963, on activation analysis [14]. About a decade
later, probably the most elaborate book on this technique by De Soete,
Gijbels and Hoste was published [15], with a very extensive treatment
of systematic errors in activation analysis.
The use of NAA for the development and certification of reference
materials at organizations such as the former National Bureau of
Standards (NBS, now the National Institute of Standards and Technol-
ogy, NIST) and the International Atomic Energy Agency (IAEA) [16,17]
has strongly contributed to the awareness and evaluation of sources of
error, the quantization of their contribution and impact, and the
development of methods to minimize their occurrence. It is therefore
not surprising that the 3rd International Conference in Modern Trends
in Activation Analysis, held at the NBS in 1968, had already a session on
accuracy and precision, the predecessor on current conference sessions
like Quality Control and Quality Assurance. These terms became familiar
to the community since the end of the 1970s, though they were – and
still are – often used loosely to their definition. The awareness of a
structured implementation of quality assurance practices culminating
into quality management [18] was introduced in the 1990s and many
NAA laboratories, some of them supported by the IAEA [19] have
successfully implemented such management systems.
All these developments contributed to the very thorough under-
standing of the principles of NAA, its sources of error and
methodologies to account for them and/or quantify their contribu-
tions to the uncertainty of measurement. This forms the basis of the
demonstrated capability of operating NAA at the highest metrological
level, as a primary ratio method of measurement.
INAA has found its usage in many fields of science. Particularly
advantage is taken of the fact that the samples do not have to undergo
any chemical treatment, neither prior, nor after the activation: INAA is
‘non-destructive’.1 In addition, low atomic number elements such as
H, C, N, O, Si which in many materials belong to the major matrix
components, do not produce radioactive products upon neutron
activation that interfere with the determination of the other activities.
It enables the observation of trace elements, often at amounts in the
order of microgram to nanogram or even less in matrices composed of
the low Z elements. The high selectivity of gamma-ray spectrometry
allows for simultaneous determination of many radionuclides. Hence,
INAA is a technique for multi-element determination.
The non-destructive character makes INAA attractive for applica-
tion in e.g. geochemistry and related sciences. Because of the limited
sample handling operations, there is also a lower risk of contamina-
tions compared to other element determination techniques in which
the sample has to be dissolved. This advantage has been exploited in
1
The description ‘non-destructive’ is commonly used to emphasize that the test
portion does not have to be dissolved prior to the analysis. As will be further described
in Sections 1.3 and 1.6, under certain conditions also in INAA the test portion may be
considered as ‘damaged’.
195
many biological applications of INAA and for the analysis of minute
quantities of material such as atmospheric and cosmic dust.
Since the signals in INAA are related to the properties of the atomic
nucleus, the results in INAA are not affected by the chemical and
physical state of the elements.2 The method is well described by physical
law and specificity is unambiguous for all elements because of the
combination of the nuclear properties (i) decay constant (often
converted to the half-life, i.e. — for a single radioactive decay process,
the time required for the activity to decrease to half its value) and (ii) the
energies and intensities of the gamma-radiation, uniquely characteristic
for each radionuclide.
Although many forms of NAA have been identified by various
practitioners, the discussions in this presentation will be limited to
only the standard forms of INAA and RNAA for normal size samples (a
few grams or less). Other forms of NAA such as prompt gamma NAA
[20], epithermal NAA [21], preconcentration NAA [22], cyclic NAA
[23], large sample NAA [24], molecular NAA [25], etc. are considered
beyond the scope of this presentation since only the standard forms of
INAA and RNAA have been directly compared with other primary
methods in CCQM comparisons.
1.3. Fundamentals of neutron activation analysis
Activation
The activation with neutrons is, upon preparation of the test portion,
the first stage in an INAA procedure. Its purpose is to convert some of the
stable nuclei in radioactive nuclei emitting radiation that can be used for
analytical purposes. Insight into the reactions that may take place during
activation facilitates the identification of the relation between the
observed radioactive nucleus, its target nucleus and associated element.
Insight into the reaction rates is of importance for the quantitative
analysis and for a priori estimates of the feasibility of an analysis.
Each atomic nucleus can capture a neutron during irradiation,
resulting in a nuclear reaction in which often the nuclear mass
changes; immediately (typically 10− 14 s) after the capture excess
energy in the form of photons and/or particles will be emitted.3 The
newly formed nucleus may be unstable. If unstable after activation, it
starts decaying to a stable state by the emission of radiation through
one or more of the following processes: α-decay, β−-decay, electron
capture, β+-decay, or internal transition decay. In most cases γ and
X-radiation will be emitted too.
The capture of a neutron by an atomic nucleus and the resulting
reaction may be illustrated, in the case of a cobalt target nucleus, by
59 1 60
Co þ n→ Co + ðpromptÞγ  radiation:
The resulting 60Co nucleus is radioactive and decays (and thereby
converts) by emission of β− radiation to excited nuclear levels of the
stable 60Ni nucleus, followed by the emission of γ-radiation from the
internal transition of the nucleus from these excited levels to its
ground state; see Fig. 1.1. Commonly, the reaction is written in the
shorthand notation as
59 60
Coðn; γÞ Co
in which ‘γ ’refers to the prompt emitted radiation, not to the gamma-
radiation following the β− decay.
2
When a radionuclide decays by emission of gamma-radiation in cascade and
fulfills certain constraints, in principle the intensity of sum-peaks in the gamma-ray
spectrum can be affected by the chemical form of the element due to perturbation of
the directional correlation between the gamma-rays involved. The extent of this effect
can be neglected in INAA considering the way the measurements are carried-out and
the relative unimportance of sum-peaks.
3
This prompt radiation can also be used for analytical purposes. It requires
measurement during activation. This method is not considered here further.
196 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Fig. 1.1. Decay scheme of 60Co [27].
The most common reaction occurring in NAA is the (n,γ) reaction,
but also reactions such as (n,p), (n,α), (n,n′) and (n,2n) are important.
The neutron cross section, σ, is a measure for the probability that a
reaction will take place, and can be strongly different for different
reaction types, elements and energy distributions of the bombarding
neutrons. Some nuclei, like 235U are fissionable by neutron capture
and the reaction is denoted as (n,f), yielding fission products and fast
(highly energetic) neutrons.
Neutrons are produced via
– Isotopic neutron sources, like 226Ra(Be), 124Sb(Be),241Am(Be),
252
Cf. The neutrons have different energy distributions with a
maximum in the order of 3–4 MeV; the total output is typically
105–107 s− 1 GBq− 1 or, for 252Cf, 2.2 · 1012 s− 1 g− 1.
– Particle accelerators or neutron generators.4 The most common
types are based on the acceleration of deuterium ions towards a
target containing either deuterium or tritium, resulting in the
reactions 2H(2H,n)3He and 3H(2H,n)4He, respectively. The first
reaction, often denoted as (D,D), yields monoenergetic neutrons of
2.5 MeV and typical outputs in the order of 108–1010 s− 1; the
second reaction (D,T) results in monoenergetic neutrons of
14.7 MeV and outputs of 109–1011 s− 1.
– Nuclear research reactors. The neutron energy distribution
depends on design of the reactor and its irradiation facilities. An
example of an energy distribution in a light water moderated
reactor is given in Fig. 1.4 from which it can be seen that the major
part of the neutrons has a much lower energy distribution that in
isotopic sources and neutron generators. The neutron output of
research reactors is often quoted as neutron fluence rate in an
irradiation facility and varies, depending on reactor design and
reactor power, between 1015 and 1018 m− 2 s− 1.
In the majority of INAA procedures thermal reactor neutrons are
used for the activation: neutrons in thermal equilibrium with their
environment (see Section 1.4). Sometimes activation with epithermal
reactor neutrons (neutrons in the process of slowing down after their
formation from fission of 235U) is preferred to enhance the activation
of elements with a high ratio of resonance neutron cross section over
thermal neutron cross section relatively to the activation of elements
with a lower such a ratio.
In principle materials can be activated in any physical state, viz.
solid, liquid or gaseous. There is no fundamental necessity to convert
solid material into a solution prior to activation; INAA is essentially
considered to be a non-destructive method although under certain
conditions some material damage may occur due to thermal heating,
4
Neutrons at very high energies are also produced in spallation sources. Although
mainly tuned for realizing neutron beams, some spallation sources have also facilities
for NAA (e.g. at Paul Scherrer Institute in Switzerland).
radiolysis and radiation tracks by e.g. fission fragments and α-radiation
emitting nuclei.
During activation the material is heated by primarily the
absorption of both reactor gamma rays (uranium fission, the fission
and activation products in the construction materials) and prompt
gamma-rays released upon neutron capture. Radiolysis is another
effect caused by the gamma-rays. H2, O2 and H2O2 are formed from
e.g. hydrate water, proteins or other organic molecules. Fission
fragments and α-particles may produce small holes in dedicated
plastic foils; these holes can be made visible under a microscope after
chemical treatment of the foil.
Decay
Radioactive nuclei are instable and decay resulting in product(s)
with mass(es) less than that of the parent. The decay involves emission
of various types of ionizing radiation towards the ground state of a stable
nucleus. Radioactive decay is a statistically random process; the
probability that a given nucleus will decay in a certain time interval
depends only on the time of observation. It is not possible to predict
when a given nucleus will decay, but the decay characteristics can be
described by the physical laws of radioactive decay, which are
comparable to first order chemical kinetics. This is further elaborated
in the paragraph on the derivation of the measurement equation.
Decay schemes provide the details of the decay of radionuclides, such
as, e.g., the energy levels and half lives involved, the transitions, spins,
multipolarities and branching ratios, types of radiation emitted, relative
intensities, conversion coefficients [26]. Decay scheme data are based on
direct experimental information with full references towards the
original literature. An example of such a decay scheme is given in Fig. 1.1.
Measurement
The radioactivity induced is measured by the detection of the
radiation, emitted during the decay of the radionuclide. In principle
both the beta radiation from the nuclear transformation and the often
following gamma-radiation from the isomeric transition to the ground
state can be used for this; moreover, the decay rate of the radionuclide can
be measured. In NAA, nearly exclusively the (energy of the) gamma-
radiation is measured because of its higher penetrating power of this type
of radiation, and the selectivity that can be obtained from distinct energies
of the photons — differently from beta radiation which is a continuous
energy distribution. The interaction of gamma- and X-radiation with
matter results, among others, in ionization processes and subsequent
generation of electrical signals (currents) that can be detected and
recorded. A radiation detector therefore consists of an absorbing material
in which at least part of the radiation energy is converted into detectable
products, and a system for the detection of these products.
In ionization detectors such as gas-filled detectors and semicon-
ductor detectors, the ions or the electrons are collected directly by
applying an electric field over the absorbing medium and subsequent
collection of the free charge carriers on the electrodes. There are two
great advantages of using a solid material above a gas for detection of
radiation. Firstly, radiation is much more effective being absorbed in
solid than in gas due to the much higher density of the material.
Secondly, the energy required to form a single ion pair in a solid is
equal to the forbidden energy gap between the valence and
conduction energy bands, which is usually much smaller than the
ionization potential of an isolated gas atom. As a result, more ion pairs
are formed upon interaction and the energy resolution is improved.
Solid state ionization detectors must have a very low leakage current
to make the signals from ionizing radiation detectable, and contain
little impurities that might trap the charge carriers before detection.
Crystals of semiconductor materials as such best meet these
requirements. Intrinsic silicon and germanium crystals are the most
widely applied. The detector is constructed in such a way that it acts as
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
a diode rectifier so that a high voltage can be applied in one direction
without discharging current. Cooling the detectors to at least − 30 °C
(Si) or − 160 °C (Ge) reduces the leakage current.
In scintillation detectors the ionization is measured indirectly
through the secondary detection of light scintillations to which the
absorbing material is transparent. The difference between direct and
indirect detection of the ionization products results directly in a
considerable difference in statistical distribution of the observed
events and thus of radiation energy resolution capability.
Si, and especially Ge semiconductor detectors are currently the
most applied detectors in NAA because of their energy resolution,
which is superior to the resolution of scintillation detectors.
The performance of the various types of semiconductor and
scintillation detectors in terms of detection efficiency and energy
resolution depend strongly on the size and construction of the
detector, and the energy of the radiation to be measured. The output
of a radiation detector is a current pulse. The amount of energy
deposited by ionization is reflected by the integral of this pulse with
time. It is generally assumed that the shape of the output pulse of a
radiation detector does not change from one event to another. The
integral is directly proportional to the amplitude or pulse height of the
output pulse. The relation between the energy of the incident radiation
and the total charge deposited in the detector crystal or pulse height of
the output signal is referred to as the response of the detector. For
semiconductor detectors, the response is linear for different energies.
The shape of the response function or gamma ray spectrum reflects the
three different interaction processes of photons with matter: photo-
electric effect, Compton scattering and pair production. If an
interaction results in photoelectric effect, the total energy Eγ of the
photon is transferred to an atomic electron. This photoelectron will
lose its energy Eγ − Eb in the absorber. During filling of the vacancy in
the electron shell, Eb is released through emission of X-rays or Auger
electrons. The energy carried by the low energy Auger electron will be
transferred completely to the absorber. As a result of the subsequent
process, the energy deposited in the absorber totals Eγ − Eb + Eb = Eγ
which becomes visible in the gamma-ray spectrum as the so-called full
Fig. 1.2. Gamma-ray spectrum of 60Co as measured with a Ge semiconductor detector showing full energy photopeaks at 1173 and 1332 keV, the Compton continuum, annihilation
peak, single (S.E.) and double (D.E.) escape peaks and backscatter peak. [28].
197
energy photopeak. Compton scattering interaction results in a
photoelectron and a scattered Compton photon, which may interact
again via one of the three interaction mechanisms. In these successive
scattering processes the scattered gamma ray contains less and less
energy until all photoelectric absorption becomes inevitable. In that
case, the original event will again contribute to the full energy
photopeak. However, there is a distinct probability that the scattered
Compton photon will escape from the crystal, taking the remaining
energy with it. This probability depends on the size of the active
volume of the crystal and the surface area of the detector, and the
energy of the photons. The detector response to the absorption of the
Compton scattered photoelectrons only is referred to as the Compton
continuum. Above a threshold of 1022 keV, the incident photon may
interact with the electric field of the atomic nucleus. It results in the
conversion of the photon into an electron-positron pair, and the
process is called ‘pair production’. The photon is completely absorbed
in this process. The threshold energy arises from the energy,
required to create the electron–positron pair, which is equivalent to
2mec2 = 1022 keV. The excess energy of Eγ-1022 keV is carried by the 2
electrons as kinetic energy. Initially this energy is transferred to the
absorber. When the positron has lost its kinetic energy it will meet an
electron and annihilate, resulting in two annihilation photons of
511 keV each. These photons can escape or deposit their energy
partially or totally in the absorber. The net energy absorbed as a
consequence of pair production is therefore:
if 1 annihilation photon escapes: Eγ minus 511 keV
if 2 annihilation photons escape: Eγ minus 1022 keV
if both annihilation photons undergo photoelectric effect and
complete absorption: Eγ keV.
The peaks in the gamma-ray spectrum resulting from escape of the
annihilation photons are denoted as the single and double escape
peak, respectively. An illustrative gamma-ray spectrum as measured
with a Ge semiconductor detector is depicted in Fig. 1.2, showing the
photopeak as the result of photoelectric absorption, the ‘Compton
198 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
continuum’ reflecting the distribution of the energies deposited in the
crystal after photons resulting from Compton interaction have
escaped from the crystal and the single and double escape peaks
which reflect the energy deposited in the crystal due to escape of one,
or both 511 keV quanta after pair production.
Since the probabilities of all three interacting processes are
dependent on the photon energy, the detection efficiency is also
energy dependent. The detection efficiency is the fraction of all the
radiation emitted from a source which produces some recorded
interaction in the sensitive volume of the detector. The detection
efficiency is composed of the following principal factors:
(i) The interaction efficiency, i, which is the fraction of the
radiation striking the sensitive volume of the detector and
producing a recorded event.
(ii) The geometrical factor, Gf, which is the fraction of all emitted
radiation which are emitted in the direction of the detector. The
geometrical factor is related to the solid angle Ω: Gf = Ω / (4π).
(iii) The peak-to-total ratio, pt, which is the fraction of recorded
events, which result in the photopeak corresponding with the
energy of the emitted radiation.
In calculations of the detection efficiency also has to be included a
term for the attenuation of the radiation in the source itself, and in the
materials between source and sensitive volume of the detector.
The product of interaction efficiency and geometrical factor is
denoted as total efficiency T : T = iGf while the product of total
efficiency and peak-to-total ratio results in the full energy photopeak
efficiency p : p = T pt.
The energy resolution of a gamma-ray detector is a function of the
energy deposited in the detector. Fluctuations in the number of
ionizations and excitations in the detector causes that the peaks in the
spectrum have a finite width normally considered as a Gaussian
distribution. The probability that an incident radiation produces
electron–hole pairs and thus charges carriers is such a process that, in
a first approximation, Poisson statistics can be applied. The actual width
of the peaks in the spectrum is determined by the intrinsic resolution of
the detector — which is determined by the number and statistical
distribution of the charge carriers, created in the crystal upon interaction
of the gamma-radiation-, and the extrinsic resolution, e.g., in semicon-
ductor detectors the leakage current through or around the active
volume and the broadening effects of all electronic components
following the detector. The intrinsic resolution is also related to the
size and type of the detector crystal, and its temperature. In addition, the
actual width of the peaks depends on the type of radiation. Nuclear
energy levels involved in the emission of gamma-radiation have very
small widths with a negligible impact on the shape of the peaks in the
spectrum, differently from natural line width of X-ray emissions. The
shape of the 511 keV peak in the spectrum results from annihilation
radiation, which is Doppler broadened.
The gamma-spectrum is analyzed to identify the radionuclides
produced and their amounts of radioactivity in order to derive the
target elements from which they have been produced and their
masses in the activated sample. The spectrum analysis starts with the
determination of the location of the (centroids of the) peaks.
Secondly, the peaks are fitted to obtain their precise positions and
net peak areas. The positions – often expressed as channel numbers of
the memory of a multi-channel pulse height analyzer – can be
converted into the energies of the radiation emitted; this is the basis
for the identification of the radioactive nuclei. On basis of knowledge
of possible nuclear reactions upon neutron activation, the (stable)
element composition is derived. The values of the net peak areas can
be used to calculate the amounts of radioactivity of the radionuclides
using the full energy photopeak efficiency of the detector. The
amounts (mass) of the elements may then be determined if the
neutron fluence rate and cross sections are known. In the practice,
however, the masses of the elements are determined from the net
peak areas by comparison with the induced radioactivity of the same
neutron activation produced radionuclides from known amounts of
the element of interest.
The combination of energy of emitted radiation, relative intensities
if photons of different energies are emitted and the half life of the
radionuclide is unique for each radionuclide, and forms the basis of
the qualitative information in NAA. The amount of the radiation is
directly proportional to the number of radioactive nuclei produced
(and decaying), and thus with the number of nuclei of the stable
isotope that underwent the nuclear reaction. It provides the
quantitative information in NAA. This is described in more detail in
Sections 1.4 and 1.5.
Measurand
The measurand in NAA – the quantity intended to be measured,
[29] – is the total mass of a given element in a test portion of a sample
of a given matrix in all physico-chemical states. The quantity ‘subject
to measurement’ is the number of disintegrating nuclei of a
radionuclide. The measurement results in the number of counts in a
given period of time, from which the disintegration rate and the
number of disintegrating nuclei is calculated; the latter number is
directly proportional to the number of nuclei of the stable isotope
subject to the nuclear reaction, and thus to the number of nuclei of the
element, which finally provides information on the mass and amount
of substance of that element.
Practice of NAA
In the practice, a NAA procedure consists of the simultaneous
irradiation of test portions of the unknown sample and a known
amount of a comparator of the same element — serving as a calibrator,
and the sequential measurements of the induced radioactivities. An
INAA procedure has the following steps:
– Preparation (e.g. drying) and weighing of test portions of sample
and comparator; no need for dissolution
– Encapsulation of the test portions in e.g. plastic foil, plastic
capsules or quartz ampoules
– Activation via simultaneous irradiation of test portions and
comparators with neutrons
– No chemical separations, material remains intact
– Sequential measurements of the induced radioactivities in each test
portion and comparator by gamma-ray spectrometry after one or
more decay periods. This is illustrated in Fig. 1.3 which shows the
changes with time of the shape of the gamma-ray spectrum. Thus,
complementary information on various elements can be obtained.
– Interpretation of the gamma-ray spectra towards elements and
their masses.
Like any other analytical method, test portions of internal quality
control materials (including blanks) are following the same steps as
the samples and comparators, and are simultaneously irradiated and
measured separately.
An RNAA procedure differs from the steps in the above only by the
chemical separations of the test portion of the sample after the
irradiation: so the various steps are:
– Preparation (e.g. drying) and weighing of test portions of sample
and comparator; no need for dissolution
– Encapsulation of the test portions in e.g. plastic foil, plastic
capsules or quartz ampoules
– Activation via simultaneous irradiation of test portions and
comparators with neutrons
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 199

Fig. 1.3. Gamma-ray spectra, recorded at different intervals after neutron activation [30].

– Chemical separations of element(s) of interest (and related radio- in which

nuclides) from other elements in the sample test portion; the test
portion of the comparator (e.g. in case of a single element standard) N0 number of target nuclei.
may be left intact. N number of radioactive nuclei.
– Sequential measurements of the induced radioactivities in each λ the decay constant, s− 1; λ = ln2/t1/2 with
test portion by gamma-ray spectrometry after one or more decay t1/2 the half-life of the radionuclide, s.
periods.
– Interpretation of the gamma-ray spectra in terms of elements and The disintegration rate of the produced radionuclide at the end of
their masses. the irradiation time ti follows from


An example of typical ranges of experimental conditions is given in λt
Dðti Þ = N ðti Þλ = N0 R 1  e i ð1  3Þ
Table 1.1.
with
1.4. Derivation of the measurement equation
D disintegration rate, Bq, of the produced radionuclide,
The reaction rate R per nucleus capturing a neutron is given by assuming that N = 0 at t = 0 and N0 = constant
∞
R = ∫ nðvÞvσ ðvÞdv ð1  1Þ The cross section and the neutron fluence rate are neutron energy
0 dependent. In nuclear research reactors – which are intense sources of
neutrons – three types of neutrons can be distinguished (see Fig. 1.4):
where
(i) Fission or fast neutrons released in the fission of 235U. Their
−1 energy distribution ranges from 100 keV to 25 MeV with a
v the neutron velocity (m s ).
σ(v) the neutron cross section (in m2)5 for neutrons with maximum fraction at 2 MeV. These neutrons are slowed down
velocity v; by interaction with a moderator, e.g. H2O, to enhance the
n(v)dv the neutron density (m− 3) of neutrons with velocities probability of them causing a fission chain reaction in the 235U.
between v and v + dv, considered to be constant in time.
The production of radioactive nuclei is described by
Table 1.1
Example of typical ranges of experimental conditions of an INAA procedure.
dN
= RN0 −λN ð1  2Þ Test portion mass: 5–500 mg
dt
Neutron fluence rates available 1016–1018 m− 2 s− 1

Irradiation Decay Measurement

5–30 s 5–600 s 15–300 s
5
1–8 h 3–5 days 1–4 h
In many tabulations and for sake of simplicity, the term ‘barn’ (notation for dimension:
20 days 1–16 h
‘b’) is still used and preferred by many practitioners; 1 barn (b)= 10−28 m2.
200 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Fig. 1.4. Schematic representation of the neutron fluence rate spectrum in a nuclear Fig. 1.5. Relation between neutron cross section and neutron energy for the 59Co(n,γ)
reactor [31]. 60
Co reaction[32].

(ii) Epithermal neutrons. These are neutrons in the process of in which

slowing down by collisions with the nuclei of the moderator.
vCd
Epithermal neutrons have energies between approximately
n = ∫ nðvÞdv ð1  6Þ
0.5 eV and 100 keV.
0
(iii) Thermal neutrons, i.e. neutrons in thermal equilibrium with the
atoms of the moderator. The energy distribution of these is called the thermal neutron density and, when Φth = nv0,
thermal neutrons is Maxwellian, with a most probable velocity
v0 of 2200 m s− 1 at 20 °C, corresponding to an energy of Φth the ‘conventional’ thermal neutron fluence rate, m− 2 s− 1,
0.025 eV. for energies up to the Cd cut-off energy of 0.55 eV.
Typically thermal neutron fluence rates in a nuclear research σ0 the thermal neutron activation cross section, m2, at 0.025 eV.
reactor are in the order of 1016–1018 m− 2 s− 1. As can be seen from υ0 the most probable neutron velocity at 20 °C: 2200 m s− 1.
Fig. 1.4, the thermal neutrons have the highest fluence rate. The
epithermal and fast neutron fluence rates strongly depend on the The second term is re-formulated in terms of neutron energy
configuration of the reactor, particularly on the choice of moderator. rather than neutron velocity and the infinite dilution resonance
The epithermal neutron fluence rate in the irradiation facilities of a integral I0 – which effectively is also a cross section (m2) – is
light water moderated reactor is typically a factor 40–50 lower than introduced:
the thermal neutron fluence rate.
Reactions of the (n,γ) and (n,f) type have the highest cross section ∞ E max
(typically in the order of 0.1–100 b) for thermal neutrons whereas the σ ðEn Þd En
∫ nðvÞvdv = Φepi ∫ = Φepi I0 ð1  7Þ
other reactions ((n,p), (n,α), (n,n′), (n,2n)) mainly occur with fast vCd ECd
En
neutrons at cross sections 2 or 3 orders of magnitude lower. In several
cases nuclear reactions result into the conversion of a stable nucleus
with
into another stable nucleus.
The cross section for thermal neutrons is often inversely
E max
proportional to the neutron velocity; in the epithermal region the σ ðEn Þd En
neutron cross section can be very high for neutrons of a discrete I0 = ∫ ð1  8Þ
ECd
En
energy and the neutron cross section versus neutron energy
relationship shows ‘resonance peaks’ (see Fig. 1.5).
The dependence of the activation cross section and neutron in which
fluence rate to the neutron energy can be taken into account in
Eq. (1-1) by dividing the neutron spectrum into a thermal and an Φepi the ‘conventional’ epithermal neutron fluence rate per unit
epithermal region; the division is made at En = 0.55 eV (the so-called energy interval, at 1 eV.
cadmium cut-off energy). This approach is commonly known as the
Høgdahl convention [33]. The integral in Eq. (1-1) can then be From this definition of I0 it can be seen that it assumes that the
rewritten as energy dependency of the epithermal neutron fluence rate is
proportional to 1/En. This requirement is fulfilled to a good
approximation by most of the (n,γ) reactions.
vCd ∞
In the practice of nuclear reactor facilities the epithermal
R = ∫ nðvÞvσ ðvÞdv + ∫ nðvÞvσ ðvÞdv: ð1  4Þ
0 vCd
neutron fluence rate Φepi is not precisely following the inverse
proportionality to the neutron energy; the small deviation can be
accounted for by introducing an epithermal fluence rate distribution
The first term can be integrated straightforward: parameter α:

vCd vCd Emax

α σ ðEn Þd En
∫ nðvÞvdv = v0 σ 0 ∫ nðvÞdv = nv0 σ 0 ð1  5Þ I0 ðαÞ = ð1 eVÞ ∫ ð1 + α Þ
: ð1  9Þ
0 0 ECd En
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 201

The expression for the reaction rate can thus be re-written as with

C net counts in the γ-ray peak of Eγ

R = Φth σ0 + Φepi I0 ðαÞ ð1  10Þ NAv Avogadro's number, mol− 1
θ isotopic abundance of the target isotope
Expressing the ratio of the thermal neutron fluence rate and the mx mass of the irradiated element, g
epithermal neutron fluence rate as f = Φth/Φepi and the ratio of the Ma atomic mass, g mol− 1
resonance integral and the thermal activation cross section as Q0(α) = Γ the gamma-ray abundance, i.e. the probability of the
I0(α) / σ0, an effective cross section can be defined: disintegrating nucleus emitting a photon of Eγ (photons
disintegration− 1)

Q ðαÞ
 ε the full energy photopeak efficiency of the detector, i.e. the
σ eff = σ 0 1 + 0 : ð1  11Þ probability that an emitted photon of given energy will be
f
detected and contribute to the photopeak at energy Eγ in
the spectrum
It simplifies the Eq. (1-10) for the reaction rate to

Although the photons emitted have energies ranging from tens of

R = Φth σ eff :
This reaction rate applies to infinite thin objects. In objects of
defined dimensions, the inside part will experience a lower neutron
fluence rate than the outside part because neutrons are removed by
absorption. Corrections have to be applied for this self-shielding
to thermal and epithermal neutrons; this is further described in
Chapter 2.
Decay and measurement
ð1  12Þ
keVs to MeVs and have high penetrating powers, they still can be
absorbed or scattered in the sample itself depending on the sample
size, composition and photon energy. This effect is called gamma-ray
self-attenuation. Also, two or more photons may be detected
simultaneously within the time resolution of the detector; this effect
is called summation. The corrections to Eq. (1-15) for gamma-ray self-
attenuation and summation effects are described in Chapter 2.
Eq. (1-15) can be simply rewritten towards the measurement
equation of NAA, which shows how the mass of an element measured
can be derived from the net peak area C:

The nuclear transformations are established by measurement of the

number of nuclear decays. The number of activated nuclei N(ti,td) λ Ma
mx = C  ⋅ : ð1  16Þ
1−e−λti ⋅e−λtd ⋅ 1−e−λtm ⋅Φth ⋅σeff ⋅Γ⋅ε θ⋅NAv
  
present at the start of the measurement is given by6

RN0
−λt

−λt
Nðti ; td Þ = 1e i e d ð1  13Þ
λ 1.5. Metrology

and the number of nuclei ΔN disintegrating during the measurement Nuclear reaction
is given by
All possible nuclear reactions can be completely enumerated
RN0
−λt

−λt


−λt
 by knowledge of the incoming particles (thermal neutrons, epithermal
ΔNðti ; td ; tm Þ = 1−e i e d 1−e m ð1  14Þ
λ neutrons, and fast neutrons), their energy and the chart of nuclides.

in which Nuclear decay

td the decay or waiting time, i.e. the time between the end of
the irradiation and the start of the measurement
tm the duration of the measurement
Additional correction resulting from high counting rates may be
necessary depending upon the gamma-ray spectrometer hardware
used. This is further elaborated in Chapter 2.
Replacing the number of target nuclei N0 by (NAvm)/M and using
the Eq. (1-12) for the reaction rate, the resulting net counts C in a peak
in the spectrum corresponding with a given photon energy is
approximated by the ‘activation formula’:
The decay schemes of the radionuclides are well established
and unambiguous. Gamma-ray energies, absolute decay intensities and
half-lives are known [26]. The combination of decay scheme and half-life
is a unique characteristic of each decaying radionuclide.
The decay process is a property of the atomic nucleus. The binding
electrons, chemical species and temperature do not influence the decay
process.7

7
There are a few exceptions: the chemical binding, in electron capture and internal
−λtm conversion decay, may have an effect to the half-life of the radionuclide since the


NAv θmx
−λt

−λt
1−e electron capture rate is inversely proportional to the electron density at the nucleus,
C = ΔNγε = Φth σeff 1−e i e d Γε ð1  15Þ which is almost entirely determined by s-electrons. If the removed or added
Ma λ
outermost electron is in the s-state, there will be an effect on the capture rate.
However, the measured effects to the decay constant are in the order of 10− 2% and
only measurable for selected, metallic or crystalline chemical species [34]. Secondly,
the chemical binding may also have a measurable effect to the nuclear decay if this
decay – under very specific conditions – includes a cascade of gamma-rays along an
intermediate state of nanosecond half life. Differences of 1–2% in the intensity ratios
6
The representation given here is correct only for the simplest of cases. between the sumpeak of the cascade and the peaks of the individual gamma-rays have
Complications may arise when the activated nucleus decays to another unstable been observed for different chemical species of 111In [35]. This phenomenon also
nucleus of which the activity is measured, e.g. 46Ca(n,γ)47Ca(β−)47Sc; here 47Ca is the causes the observed intensity of full energy photopeaks of certain radionuclides to
activation product, but 47Sc is the radionuclide assessed. appear less than the true intensities [36].
202 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Measurand
The area of a photopeak is a measure of the current transfer
in the detector, resulting from interaction of gamma-radiation of
corresponding energy in the radiation detector. The current is directly
proportional to the number of decaying radioactive nuclei.
The number of decaying radioactive nuclei is directly proportional
to the number of stable nuclei of the isotope that underwent a nuclear
reaction.
The number of stable nuclei of the isotope is directly proportional
to the total number of nuclei of the corresponding element and,
assuming given abundances of the stable isotopes, to the total mass of
that element in the irradiated test portion.
Measurement equation and metrological traceability
The metrological traceability of the values of the masses obtained
by an NAA procedure, can be derived from the measurement equation
[Eq. (1-16)]. The dimensional units of the physical constants in this
equation are:
λ decay constant (= ln 2/t1/2) s− 1
σeff spectrum averaged cross section, m2
Γ fraction of decays producing γ-ray of Eγ, dimensionless
NAv Avogadro's number, mol− 1
θ isotopic abundance of the target isotope, dimensionless
Ma atomic mass, g mol− 1
of the experimental parameters are:
ϕeff neutron fluence rate during irradiation, s− 1 m− 2
ε detection efficiency for γ-ray of Eγ, dimensionless
and of the measured quantities
C net counts in the γ-ray peak of Eγ, dimensionless
ti irradiation duration, s
td decay time to start of count, s
tm duration of the measurement, s
mx mass of the irradiated element, g
Hence, the values of all constants, parameters and measured
quantities in the measurement equation have the potential of being
metrologically traceable to the S.I.
Calibration
Calibration is based on the determination of the proportionality
factors F that relate the net peak areas in the gamma-ray spectrum to
the amounts of the elements present in the sample under given
experimental conditions:
C instead of individual calibrators for each element.
F= : ð1  17Þ
m The single comparator method for multi-element INAA was based
Both absolute and relative methods of calibration exist.
Absolute calibration
The values of the physical parameters determining the propor-
tionality factor θ, NAv, M, σeff Γ, λ, are taken from literature. The
parameters σeff respectively Γ, λ are not precisely known for many
(n,γ) reactions and radionuclides, and in some cases θ is also not
accurately known. Since the various parameters were often achieved
via independent methods, their individual uncertainties will add up in
the combined uncertainty of measurement of the elemental amounts,
leading to a relatively large combined standard uncertainty. More-
over, the metrological traceability of the values of the physical
constants is not known for all radionuclides. The other parameters C,
mx, Φ, ε, ti, td, tm are determined, calculated or measured for the given
circumstances and uncertainties can be established.
Relative calibration
Direct comparator method
The unknown sample is irradiated together with a calibrator8
containing a known amount of the element(s) of interest. The calibrator
is measured under the same conditions as the sample (sample-to-
detector distance, equivalent sample size and if possible equivalent in
composition). From comparison of the net peak areas in the two
measured spectra the mass of the element of interest can be calculated:
!
C
tm :e−λtd · 1−e−λtm
 
mxðunkÞ = mxðcalÞ · !unk ð1  18Þ
C
tm :e−λtd · 1−e−λtm
 
cal
in which
mx(unk), mx(cal) mass of the element of interest, in unknown sample
and calibrator, respectively; in g
In this procedure many of the experimental parameters — such as
neutron fluence rate, cross section and photopeak efficiency cancel
out at the calculation of the mass and the remaining parameters are
all known. This calibration procedure is used if the highest degree of
accuracy is required.
The relative calibration on basis of element calibrators is not
immediately suitable for laboratories aiming at the full multi-element
powers of INAA. It takes considerable effort to prepare multi-element
calibrators for all 70 elements measurable via NAA with adequate
degree of accuracy in a volume closely matching the size and the
shape of the samples.
Single comparator method
Multi-element INAA on basis of the relative calibration method is
feasible when performed according to the principles of the single-
comparator method. Assuming stability in time of all relevant
experimental conditions, calibrators for all elements are co-irradiated
each in turn with the chosen single comparator element. Once the
sensitivity for all elements relative to the comparator element has
been determined (expressed as the so-called k-factor, see below),
only the comparator element has to be used in routine measurements
on the ratio of proportionality factors of the element of interest and
of the comparator element after correction for saturation, decay,
counting and sample weights. Girardi et al. [38] defined the k-factor
for each element i as
ðMa Þi;cal γcomp εcomp θcomp σ eff ;comp
ki = : ð1  19Þ
Ma;comp γi;cal εi;cal θi;cal ðσ eff Þi;cal
8
A calibrator is defined as a measurement standard used in calibration [37].
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 203

Masses for each element i then can be calculated from these ki- elsewhere measured ratio(s) of the activation rate(s) of element(s)
factors; for an element determined via a directly produced radionu- and the standard element, Au. Some of these values and parameters
clide the mass mx(unk) follows from are still regularly updated. The neutron fluence rate parameters f
(ratio of thermal to epithermal neutron fluence rate) and α no longer
cancel out in mass calculations and must be measured in each
!
C
irradiation facility, preferably even for each irradiation and sample
1−e−λti ⋅tm ⋅e−λtd ⋅ 1−e−λtm unk
   
mxðunkÞ = mxðcompÞ ⋅ ! ⋅ki ð1  20Þ [42]. At least three isotopes must be activated and measured
C to determine these parameters. A composed fluence rate monitor
1−e−λti ⋅tm ⋅e−λtd ⋅ 1−e−λtm comp
   
containing adequate quantities of Au and Zr is suitable for this
with purpose; in a single measurement the induced activities of
198
Au, 95Zr and 97Zr can be assessed; recently an alternative set of
mx(comp) mass of element x in comparator, in g. Mo, Cr and Au was suggested for monitoring of well-thermalized
facilities [43]. As such, the k0-method is not a single comparator
These experimentally determined k-factors are often more but e.g. a triple comparator method. The photopeak efficiencies of the
accurate than when calculated on basis of literature data as in the detector must also be (experimentally) determined in this approach.
absolute calibration method. However, the k-factors are only valid for The k0-method requires tedious characterizations of the irradia-
a specific detector, a specific counting geometry and irradiation tion and measurement conditions and results, like the single
facility, and remain valid only as long as the neutron fluence rate comparator method, in relatively high uncertainties of the measured
parameters of the irradiation facility remain stable. values of the masses. Moreover, metrological traceability of the
The single comparator method requires laborious calibrations in currently existing k0 values and associated parameters to the S.I. is not
advance, and finally yield relatively (compared to the direct yet transparent and most probably not possible.
comparator method) higher uncertainties of the measured values. Summarizing, relative calibration by the direct comparator
Moreover, it requires experimental determination of the photopeak method renders the lowest uncertainties of the measured values
efficiencies of the detector. Metrological traceability of the measured whereas metrological traceability of these values to the S.I. can easily
values to the S.I. may be demonstrated. be demonstrated. As such, this approach is often preferred from a
metrological viewpoint. The measurement equation can be further
The k0-comparator method simplified, substituting

At the Institute for Nuclear Sciences in Ghent, Belgium, an attempt λCeλtd

A0 =  ð1  23Þ
1−eλtm ⋅ 1−eλti
 
has been made to define k-factors which should be independent of
neutron fluence rate parameters as well as of spectrometer char-
acteristics. In this approach, the irradiation parameter (1 + Q0(α)/f) to
(Eq. (1-11)) and the detection efficiency ε are separated in the
expression (1-19) of the k-factor, which resulted at the definition of A0ðunkÞ
the k0-factor [39] munk = mcal R R R R ð1  24Þ
A0ðcalÞ θ Φ σ ε

Q0;comp ðαÞ in which

1 1+ f εcomp Mcomp θcal σ0;cal γcal
k0 = Q0;cal ðαÞ
= ð1  21Þ
ki 1 + εcal Mcal θcomp σ0;comp γcomp
f Rθ ratio of isotopic abundances for unknown sample and
calibrator.
Rϕ ratio of neutron fluence rates (including fluence gradient,
and the mass, again for an element determined via a directly produced neutron self shielding, and scattering) for unknown sample
radionuclide, is found from and calibrator.
Rσ ratio of effective cross sections if neutron spectrum shape
! differs from unknown sample to calibrator.
Q0;comp ðαÞ C
1+ 
1−e−λti ⋅tm ⋅e−λtd ⋅ 1−e−λtm ⋅m unk 1
   Rε ratio of counting efficiencies (differences due to geometry
f εcomp
mxðunkÞ = mxðcompÞ
Q0;unk ðαÞ εunk
! ⋅k : and γ-ray self shielding) for unknown sample and calibrator.
1+ C 0
f 
1−e−λti ⋅tm ⋅e−λtd ⋅ 1−e−λtm ⋅m comp
  

ð1  22Þ
1.6. Analytical characteristics

Total sample analysis

All in the above is based on the assumption that the cross section in
the thermal energy region varies with the inverse of the neutron
energy. For a small group of nuclear reactions (151Eu(n, γ) 152,152mEu,
168
Yb(n, γ)169Yb, 176Lu(n, γ)177Lu) this assumption results in
significant deviations, requiring the use of the Westcott formalism
instead of the Høgdahl convention for describing the relation between
activation cross section and neutron energy, thereby introducing the
neutron temperature as a parameter. The impact of the equations is
well documented [40].
The k0-factor thus has become a purely nuclear parameter for
the thermal neutron spectrum. In the k0-convention, Au is often
proposed as comparator element. The k0 values and associated
parameters such as Q0 are tabulated [e.g., 41] and based on
Neutron activation analysis provides information on the total mass
of elements in the sample processed. Since INAA, originally denoted as
‘non-destructive NAA’ does not require dissolution of the sample,
there are in principle no losses by incomplete digestion, precipitation
or wall adsorption effects nor does the technique suffer from
contaminations that might be introduced by the solvents used. It is
well known that special measures have to be taken to prevent losses
by volatilization of some elements during irradiation (e.g. Hg and Br);
in such cases often sealed quartz ampoules are preferred above
polyethylene vials. The degree of volatilization depends whether the
element is bound to an inorganic or organic ligand, and the irradiation
conditions (neutron and gamma-ray fluences and temperature).
204 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
In an INAA procedure the element mass in the entire (bulk) sample
is determined, which distinguishes the technique from X-ray
fluorescence analysis, which provides information only on the surface
composition of a sample.
In an RNAA procedure the dissolution takes place after the
irradiation. The chemical yield determination can be done with high
degree of accuracy by either addition of a non-radioactive spike of
the same element(s) to be separated, and or by quantitation of the
recovery of a radiotracer (same element(s), produced separately)
added after irradiation.
No effects from the chemical state and physical form of the element to be
determined
All phenomena (nuclear reaction, decay, and emission of radia-
tion) are related to the properties of the atomic nucleus. The valence
electrons are with some exceptions (see footnote 7) not involved in
these processes. Hence, the determination of an element via
measurement of its induced radioactivity is in principle independent
of its chemical state or physical form; there is no difference whether
an element is bound to an inorganic compound or an organic
compound, or if it is present as a pure metal, provided the degrees
of gamma-ray self-attenuation in sample and calibrator are equivalent
and can be quantified (cf. parameter Rε in Eq. (1-24)). Chemical
matrix effects, known to be significant sources of error in some other
types of instrumental chemical analysis, are insignificant in NAA or
can be quantified and corrected for.
This characteristic implies that the calibrators (standards) in NAA
do not have to be identical in composition than the measurands. It
allows for use of single element comparators of the highest purity and
known stoichiometry (e.g. metal foils) with relatively small contri-
bution to the combined standard uncertainty of measurement. The
differences in neutron and gamma-ray self-shielding, and in irradi-
ation and counting geometry between calibrator and sample
are described by physical parameters, and can be calculated and/or
experimentally determined.
Self-validating characteristics
The nuclear physical principles of NAA also provide many self-
validating opportunities of the measurement results, contributing
favorably to the degree of accuracy.
Most of these result from the fact that many elements have
multiple stable isotopes from which more than one radionuclide can
be produced upon neutron irradiation (e.g. 122Sb and 124Sb).
Moreover, these multiple radionuclides from one element have
different half-lives (e.g. 122Sb: 64 h and 124Sb: 60 days), and emit
several gamma-rays with different energies with accurately known
emission probabilities (intensity ratios) e.g. 122Sb: 564 keV (63%
emission probability) and 693 keV (3.3%), and 124Sb: 602.7 keV (98%),
723 keV (11%) and 1691 keV (52%).
The combination of half-life and gamma-ray energy makes
confirmation of identity of the radionuclide 100% and confirmation
of identity of the element (on basis of known nuclear reactions) in
most cases also 100%. In addition, quantitative results derived from
the different peak areas of one radionuclide, as well as from different
radionuclides of the same element should be in agreement. Equally so,
results from the same radionuclide, registered after different decay
intervals, should agree with each other.
Interferences during irradiation may occur. As an example,
the radionuclide 28Al may result from the nuclear reactions 27Al
(n,γ)28Al, 28Si(n,p)28Al or 31P(n,α)28Al. Such interferences can be
identified, quantified and corrected for. Similarly, interferences
during measurement may occur (e.g., the gamma-ray line at
320 keV may result from either the decay of 51Ti or 51Cr to levels in
51
V) but identification and correction is possible taking advantage
of the differences in half lives of the radionuclides (e.g., 51Ti: 5.8 m,
51
Cr: 27.7 days) and/or the information obtained from other radio-
nuclides formed from the element of interest and intensity ratios
from different gamma-ray lines (e.g., in the decay of 51Ti also gamma-
rays of 605 keV and 928 keV is emitted, which does not occur in the
decay of 51Cr).
The impact of some of these interferences is mass dependent
but such interferences can easily be identified and quantified by
measurements after different decay intervals. Multiple measure-
ments after different decay intervals also serve for assessing pile-
up effects and interference corrections, and to test all counting-
related nonlinearities. The degree of gamma-ray self-attenuation
can be derived from the intensity ratios of peaks in the spectrum
or calculated after experimental determination of the measured
transmission through the sample of an external beam of gamma-
rays.
The peak areas can be determined in several ways – e.g., with
fitting routines or via channel-by channel counting and background
subtraction – thus providing an additional validation of this step.
Finally, as will be demonstrated in detail in Chapter 2 (Evaluation
of Uncertainties for NAA Measurements Using the Comparator
Method of Standardization) of this paper, the type A evaluations of
contribution to uncertainty of measurement are mass dependent
and can be measured. Most of the type B evaluations of contribution
to uncertainty of measurement are not mass dependent, and affect
all elements in the same irradiation or measurement to the same
extent.
Many adjustable experimental parameters for optimization of
experimental design
A NAA procedure has many adjustable experimental para-
meters, which is reflected by the measurement (Eq. (1-16)).
These can be used for optimizing the design of the NAA procedure
towards e.g. minimization of uncertainty of measurement, mini-
mization of interfering effects during irradiation and/or measure-
ment, minimization of the limit of detection or reduction of the
total analysis (turnaround) time. Sample and calibrator mass may
be optimized for e.g. reduction of geometric effects during
irradiation and measurement, for reduction of self attenuation
effects or for increasing the signal-to-noise ratio. Irradiation time
and neutron fluence rate have an effect to the amount of induced
radioactivity of the radionuclide of interest but irradiation time
may also be varied for changing the ratio of induced radioactivity of
this radionuclide and interfering ones. Cadmium or boron filters in
the irradiation facility remove the thermal neutrons and thus
change the neutron energy spectrum. This is a well-established
approach for selective activation of elements with epithermal and
fast reactor neutrons.
It was illustrated already by Fig. 1.3 that complementary
information can be obtained from measurements after different
decay periods. This is also an approach for reduction of spectral
interferences such as closely spaced gamma-ray peaks, or for
improvement of the detection limit of small amounts of induced
radioactivity of radionuclides with long half lives in the presence of
high amounts of radionuclides with shorter half lives. Moreover, such
multiple measurements are often preferred for taking full advantage
of the self-validating opportunities of the various radionuclides
produced from activation of a given element.
Elements such as H, C, N, O, and Si do not affect the determination of
other elements
The neutron activation of a number of low Z elements (H, He, Be, Li,
B, C, N, and O — which belong often to the most abundant in many
materials-) and of a few high Z elements (Bi, Tl, and Pb) is
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 205

characterized by either one, or a combination of (i) very low Table 1.2

activation cross sections, (ii) activation products with very short Detection limits for elements in mg kg− 1 as observed in a NAA procedure of a plant material
and a soil material [33]. Experimental conditions: plant: ti =4 h at 5⁎1016 m− 2 s− 1,
half-lives (in the order of seconds) and (iii) the emission of radiation td =5 days, tm =0.5 h at coaxial detector, followed at td =3 weeks and tm =2 h at well-type
which is not recorded in the gamma-ray spectrum and therefore does detector; sample size=200 mg; soil: ti =1.5 h at 5⁎1016 m− 2 s− 1, td =5 days, tm =1 h at
not interfere with the measurement of the gamma radiation emitted coaxial detector, followed at td =3 weeks and tm =1 h at well-type detector; sample
by the other activation products. As such, these elements are virtually size=200 mg.
absent in gamma-ray spectrometry in NAA, making the sample Plant Soil Plant Soil
‘transparent’ for the signals from the radionuclides.
Na 2 10 K 200 1500
Ca 700 4000 Sc 0.001 0.02
Suitable for measurement of total element mass in the order of 10− 6 Cr 1 1 Fe 8 100
to 10− 9 g, or less Co 0.02 0.3 Ni 2 30
Zn 0.4 6 Ga 2 10
As 0.2 0.8 Se 0.1 1
The detection limit in NAA is based on the signal-to-noise ratio, the
Br 0.3 0.8 Rb 0.4 6
selectivity of determining, with a certain degree of confidence, a peak in Sr 5 60 Zr 5 80
the gamma-ray spectrum. The photopeak in the spectrum of the Mo 4 10 Ag 0.2 2
gamma-ray emission by the radionuclide of interest is the ‘signal’. The Cd 3 8 Sn 10 20
Sb 0.02 0.2 Te 0.3 3
‘noise’ results from the detection of photons from the ambient
Cs 0.02 0.3 Ba 10 40
background, from the Compton continuum due to the interaction of La 0.1 0.3 Ce 0.2 1
higher energy γ-rays, as well as from γ-ray spectrum interferences from Nd 0.7 8 Sm 0.01 0.03
e.g. the blank from pre-irradiation treatment, from packing materials Eu 0.006 0.05 Tb 0.008 0.1
and from (partly) overlapping peaks. The detection limit depends on the Yb 0.03 0.2 Lu 0.004 0.02
Hf 0.01 0.1 Ta 0.01 0.2
irradiation, the decay and the counting conditions. It is quite common to
W 0.3 1 Re 0.08 0.2
estimate the detection limit using Currie's formula [44]: Os 0.1 0.6 Ir 0.0006 0.004
pffiffiffi Au 0.003 0.01 Hg 0.05 0.4
DL = 2:71 + 4:65 B ð1  25Þ Th 0.01 0.1 U 0.2 2

where DL is the detection limit and B is the background under a γ-ray

peak. This approach is valid only when the γ-ray background
(counting statistical error) is the major interference.
The detection limits depends in the practice of INAA on:
– The amount of material to be irradiated and to be counted: often
set by availability, sample encapsulation aspects and safety limits
both related to irradiation and measurement, and possibly because
of neutron self-shielding and gamma-ray self-absorption effects.
For these reasons in the practice of INAA the sample mass is often
limited to approximately 250 mg.
– The neutron fluence rates: set by available irradiation facilities.
– The duration of the irradiation time: set by practical aspects, such
as the limitations in total irradiation dose of the plastic containers
because of radiation damage, or by reactor operation cycles.
– The total induced radioactivity that can be measured is set by the
state-of-the-art of counting and signal processing equipment, with
additional radiation dose and shielding considerations.
– The detector size, measurement geometry and background
shielding.
radionuclides produces such as half life and energies of the gamma-
ray emission. As was mentioned in the above, the very low Z elements
(like H, He, B, Be, C, N, and O) are therefore not suitable for detection
by NAA, but also a few other elements cannot be determined too (e.g.,
Tl and Bi) or at least not at a low level. The activation of lead (Pb), for
instance, results only at milligram quantities to measurable activities
which is for many applications an inadequate sensitivity.
The sample matrix should not have both high density (resulting in
strong gamma-ray self-attenuation), high atomic number (also
affecting gamma-ray self attenuation) and extremely high neutron
absorbing properties. With respect to the latter, high mass fractions of
B, Cd, Gd, and Dy will cause neutron self-absorption of the thermal
neutrons inside the sample, resulting in a neutron energy distribution
different from the distribution in the standard. Moreover, it may cause
a neutron fluence rate depression outside the sample, affecting the
neutron fluence in the calibrators which surround the sample. Large
amounts of the elements with strong neutron absorbing properties
are also unwanted in NAA if their neutron capture results in the
emission of charged particles such as α-radiation (either prompt or
delayed). This occurs mainly with the elements B, Li and U and may
cause excessive thermal heating during the irradiation.

It all illustrates that the detection limit for a given element by INAA
may be different for each individual type of material, and analysis
conditions. In Table 1.2 [45] are given, as an indication, typical detection
limits as derived from the analysis of a plant and a soil material.
1.7. Applicability of NAA
Neutron activation analysis, like any other trace element analysis
technique, is not a panacea for any trace element determination in any
type of material. The selection of NAA as a method of choice may be
based on the fact that element of interest and sample material
(matrix) should have specific chemical properties, physical forms and
physical characteristics for analysis by NAA.
Preferred conditions for selection of NAA
The nuclear properties of the element of interest are of
importance: the activation rate and the decay characteristics of the
Samples particularly suited for NAA
The analytical characteristics of neutron activation analysis are
employed at full advantage for the analysis of three major categories
of materials:
• Solid materials that are difficult to bring completely into a solution
Examples are:
– soils, rocks, minerals, ores, cosmic material, air particulate matter,
zeolites, and new composite materials
– materials in with C, H, N, and O as major elements: plants and
similar biological material, plastics
• Solid materials that are easy to contaminate during preparation of
the test portion, if e.g. digestion would be needed for a different
analytical technique
Examples are:
– ultra pure substances: Si semiconductor material and carbon fiber
– ultra small quantities: air particulate matter and cosmic dust
206 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
– biological tissues and fluids (blood, serum, and urine)
• Solid materials that are unique and should keep their integrity
Examples are:
– materials from forensic studies
– archaeological, cultural and art objects
• Solid materials of which the bulk composition has to be determined
and for which surface techniques such as XRF and solid-state
spectroscopic techniques (e.g. LIBS and laser ablation ICP) are
therefore inadequate.
Examples of such samples, within a selected group of disciplines, are:
(1) Archaeology — amber, bone, ceramics, coins, glasses, jewelry,
metal artifacts and sculptures, paintings, pigments, pottery,
soils and clays, and stone artifacts.
(2) Biomedicine — animal and human tissues, blood and blood
components, bone, drugs and medicines, gallstones, hair,
implant corrosion, kidney and kidney stones, medical plants
and herbs, milk, nails, placenta, snake venom, teeth, dental
enamel and dental fillings, urine and urinary stones.
(3) Environmental science and related fields — aerosols, atmo-
spheric particulates (size fractionated), fossil fuels and their
ashes, animals, birds, insects, fish, aquatic and marine biota,
seaweed, algae, lichens, mosses, plants, trees, household and
municipal waste, soils, sediments, and sewage sludges.
(4) Forensics — bomb debris, bullet lead, explosives detection, glass
fragments, paint, hair, gunshot residue swabs, and shotgun pellets.
(5) Geology and geochemistry — asbestos, bore hole samples, bulk
coals and coal products, coal and oil shale components, crude
oils, cosmo-chemical samples, cosmic dust, coral, diamonds,
exploration and biogeochemistry, meteorites, ocean nodules,
rocks, sediments, soils, glacial till, ores and separated minerals.
(6) Industrial products — alloys, catalysts, ceramics and refractory
materials, coatings, electronic materials, fertilizers, fissile
material detection and other safeguard materials, graphite,
high purity and high-tech materials, integrated circuit packing
materials, oil products, pharmaceutical products, plastics,
semiconductors, pure silicon and silicon processing.
(7) Nutrition — composite diets, foods, honey, seeds, spices,
vegetables, milk and milk formulae, and yeast.
Additional aspects of the use of NAA
Some additional operational aspects of NAA are worth mentioning
here for completeness. NAA is not available as a stand-alone push-
button apparatus with a fully integrated software package to be
operated ‘on-the-spot’ at just any analytical laboratory. First, access to
a nuclear research reactor is essential, obviously. Though most NAA
laboratories are found on-site of the reactor, there are quite a few
other cases in which the NAA laboratory is not on-site and samples
(before and after irradiation) are transported over distances up to
more than 100 km.9 This implies limitations with respect to the
measurement of radionuclides with (very) short half lives. The
reactor's irradiation facilities define the available neutron fluence
rates, neutron energy distributions and set technical constraints with
respect to sample encapsulation, maximum and minimum irradiation
duration.
The required instrumentation for measurement of the induced
radioactivity, viz. the gamma-ray spectrometer with semiconductor
detector, is build up with modular units that all are commercially
available. Sample changers for NAA, versatile with respect to be used
9
The NAA laboratory of one of the authors of this paper (EADNF) is at
approximately 150 km from the research reactor in São Paulo.
with sample vials of specific dimensions and different detector types
(e.g. well-type detectors) are limited commercially available. In most
cases a laboratory has to develop and build its own automation.
Spectrum analysis and interpretation software is nowadays available
with all major suppliers of gamma-ray spectrometers as well as the
International Atomic Energy Agency. However, getting started with NAA
still requires considerable effort, especially with respect to mastering
sources of error in the conduct of the technique and with the
interferences in the spectrum analysis and interpretation. The merge
of a physical method and chemical analysis requires simultaneous
expertise from both disciplines for an effective operation.
The NAA laboratory has to meet the legal requirements for
radiological safety. This may imply high investment costs if a non-
radiological laboratory has to be adapted to such requirements. A
radiological health officer may be required (which could be a part-
time function of an employee of the NAA laboratory) and provisions
have to exist for storage and disposal of radioactive waste. Employees
have to be trained in the radiological safety and practical aspects of
handling radioactivity. If the research reactor is not on-side of the NAA
laboratory, the safety in transportation of the radioactive samples
requires attention too.
The fundamentals of NAA are further described in detail in e.g.
reference [15].
1.8. References for Chapter 1
[1] H. Levi, Semicentennial Lecture, held on June 23, 1986 at the opening of the
7th International Conference Modern Trends in Activation Analysis, Copenha-
gen, Denmark.
[2] G. Hevesy, Hilde Levi, The Action of Neutrons on the Rare Earth Elements, Det.
Kgl. Danske Videnskabernes Selskab, Mathematisk-fysiske Meddelelser XIV, 5
(1936) 3–34.
[3] G. Hevesy, Hilde Levi, Artificial Activity of Hafnium and some other Elements,
Det. Kgl. Danske Videnskabernes Selskab, Mathematisk-fysiske Meddelelser
XV, (1938) 11–21.
[4] G.E. Boyd, Method of Activation Analysis, Anal. Chem. 21 (1949) 335–347.
[5] H. Brown, E. Goldberg, The Neutron Pile as a Tool in Quantitative Analysis;
The Gallium and Palladium Content of Iron Meteorites, Science 109 (1949)
347–353.
[6] T.P. Kohman, Measurement Techniques of Applied Radiochemistry, Anal.
Chem. 21 (1949) 352–364.
[7] W. Wayne Meinke, Nucleonics, Anal. Chem. 28 (1956) 736–756.
[8] W.B. Lewis, Isotope production, how to choose irradiation times, Nucleonics
12 (1954) 30–33.
[9] W.B. Lewis, Flux perturbations by material under irradiation, Nucleonics 13
(1955) 82–88.
[10] R.C. Plumb, J. E. Lewis, How to minimize errors in neutron activation analysis,
Nucleonics 13, (1955) 42–46.
[11] L.H. Rietjens, G.J. Arkenbout, G.F. Wolters, J.C. Kluyver, Influence of distance
between source and crystal on detection efficiency of gamma-ray scintillation
spectrometers, Physica 21 (1955) 110–116.
[12] G. Leliart, J. Hoste, J. Eeckhaut, Activation analysis of Vanadium in high
alloy steels using Manganese as internal standard, Anal. Chim. Acta 19 (1958)
100–107.
[13] G.B. Cook, Radioactivation analysis in a nuclear reactor, Pure and Applied
Chemistry 1(1960) 15–30.
[14] H.J.M. Bowen, D. Gibbons, Radioactivation Analysis, Oxford University Press,
Oxford, UK, 1963, 295 pp.
[15] D. De Soete, R. Gijbels, J. Hoste, Neutron Activation Analysis, Wiley-
Interscience, London, 1972.
[16] Advisory Group of the International Atomic Energy Agency, Quality
assurance in biomedical neutron activation analysis, Anal. Chim. Acta
165 (1984) 1–29.
[17] H. Kawamura, R.M. Parr, H.S. Dang, W. Tian, R.M. Barnes, G.V. Iyengar, Analytical
quality assurance procedures developed for the IAEA's Reference Asian Man
Project (Phase 2), J. Radioanal. Nucl. Chem. 245, (2000) 123–126.
[18] P. Bode, J.P. van Dalen, Accreditation: a prerequisite, also for neutron activation
analysis laboratories, J. Radioanal. Nucl. Chem.179 (1994)141–148.
[19] M. Rossbach, J. Gerardo-Abaya, A. Fajgelj, P. Bode, P. Vermaercke, M. Bickel,
Quality system implementation in Member States of the IAEA, Accred. Qualit.
Assur. 10 (2006) 583–589.
[20] G.L. Molnar (Ed). Handbook of prompt gamma activation analysis with
neutron beams. Kluwer Academic Publishers, Dordrecht, (2004) ISBN
1402013043, 423 pages.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 207

[21] E. Steinnes, D. Brune, Determination of uranium in rocks by instrumental 1.9. List of symbols used in Chapter 1
activation analysis using epithermal neutrons, Talanta. 16 (1969) 1326–1329.
[22] R.R. Rao, A. Chatt, Microwave acid digestion and preconcentration neutron
activation analysis of biological and diet samples for iodine, Anal. Chem. 63
A0 decay corrected count rate
(1991) 1298–1303. B background counts under a γ-ray peak
[23] A. Egan, S.A. Kerr, M.J. Minski, Determination of selenium in biological C net peak counts
materials using Se-77 m (T = 17.5 s) and cyclic activation analysis, Radio- a as subscript denoting element of interest in sample
chem. Radioanal. Lett. 28 (1977) 369–378.
c as subscript denoting element of interest in comparator
[24] P. Bode, R.M.W. Overwater, Trace element determinations in very large
samples — a new challenge for neutron activation analysis, J. Radioanal. Nucl. DL detection limit
Chem. 167 (1993) 169–176. Dti disintegration rate (Bq) at the end of the irradiation time ti
[25] L.R. Opelanio, E.P. Rack, A.J. Blotcky, F.W. Crow, Determination of chlorinated d as subscript denoting decay
pesticides in urine by molecular neutron activation analysis, Anal. Chem. 55 Eγ gamma-ray energy
(1983) 677–681.
[26] Table of Isotopes, Eight Edition, R.B. Firestone, V.S. Shirley, Eds., J. Wiley, New
ECd neutron energy at the cadmium cut-off point, 0.55 eV
York, 1996. Emax maximum energy of epithermal neutrons
[27] D.D. Sood, A.V.R. Reddy, N. Ramamoorthy, ‘Fundamentals of Radiochemistry’, En neutron energy
IANCAS, Mumbai, India, 2000. F C/m
[28] Gamma-Ray Spectrum Catalogue, CD-ROM 1999, Dr. R.G. Helmer. Idaho
f Φth/Φepi
National Energy & Environmental Laboratory, Idaho Falls, ID, USA.
[29] International Vocabulary of Metrology — Basic and General Concepts and I0 resonance integral
Associated Terms (VIM) 3rd edition, 2007. I0(α) resonance integral, taking into account the not perfect
[30] Adapted from: P. Bode, J.J.M. de Goeij, Activation Analysis, in; Encyclopedia of inverse proportionality of Φepi to the neutron energy
Environmental Analysis and Remediation, R.A. Meyers, Ed., J. Wiley, New York i as subscript denoting irradiation
(1998), ISBN 0-471-11708-0, pages 68-84.
k Girardi's proportionality factor for calibration in NAA
[31] Adapted from: G. Erdtmann, H. Petri, Neutron Activation Analysis: Fundamentals
and Techniques, in: Treatise on Analytical Chemistry, Part 1, Theory and Practice, k0 proportionality factor for the k0-method of calibration in
Volume 14, P.J. Elving, Ed., J. Wiley, New York (1986), ISBN0-471-80648-X, 795 NAA
pages. Ma atomic mass
[32] Reference neutron Activation Library, IAEA Technical Document TECDOC-1285 m mass
(2002) Vienna, Austria.
m as subscript denoting measurement
[33] O.T. Høgdahl, Neutron Absorption in Pile Neutron Activation Analysis,
N number of radioactive nuclei produced
Michigan Memorial Phoenix Project, Univ. of Michigan, Ann Arbor, USA,
Report MMPP-226-1 (1962).
NAv Avogadro's number
[34] H. Daniel, Influence of Chemical Environment on Lifetimes in Nuclear N0 number of target nuclei
Physics, in: The Use of selective Nuclear Techniques for the Elucidation of n neutron density
Chemical Bonding, Topical Issue of Atomic Energy Review 17 2 (1979), n(v)dv neutron density of neutrons with velocities between v and v +
287–344.
dv, considered to be constant in time
[35] M. de Bruin, P.J.M. Korthoven, The Influence of the Chemical Form of
Radionuclides on the Shape of the Gamma-Ray Spectrum, Radiochem. Q0(α) I0(α)/σ0
Radioanal. Lett. 21, 5 (1975) 287–292. R reaction rate per nucleus capturing a neutron
[36] R.J. Gehrke, R.H. Helmer, R.C. Greenwood, Precise relative χ-ray intensities Rj ratio of parameter j (= θ, Φ, σ, ε) for unknown sample and
for calibration of Ge-semiconductor detectors, Nucl. Instr. Meth. 147 (1977)
calibrator
405–423.
[37] International vocabulary of metrology — Basic and general concepts and
t time
associated terms (VIM), JCGM200: 2008. t1/2 half-life
[38] F. Girardi, G. Guzzi, J. Pauly, Reactor Neutron Activation Analysis by the Single v neutron velocity
Comparator Method, Anal. Chem 37 (1965) 1085–1092. v0 most probable neutron velocity at 20 °C: 2200 ms− 1
[39] A. Simonits, F. de Corte, J. Hoste, Single Comparator Methods in Reactor Neutron
vCd neutron velocity at the cadmium cut-off point, ECd = 0.55 eV
Activation Analysis, J. Radioanal. Chem. 24 (1975) 31–46.
[40] F. de Corte, A. Simonits, F. Bellemans, M.C. Freitas, S. Jovanovic, B. Smodis, mx mass of the irradiated element
G. Erdtmann, H. Petri, A. de Wispelaere, Recent advances in the k0- α epithermal fluence rate distribution parameter
standardization of neutron activation analysis: extensions, applications, Γ fraction of decays producing γ-ray of energy Eγ
prospects, J. Radioanal. Nucl. Chem. 169 (1993) 125–158.
γ gamma-ray abundance
[41] F. de Corte, A. Simonits, Recommended nuclear data for use in the k0-
standardization of neutron activation analysis, Atomic Data Nuclear Data
ε photopeak efficiency
Tables 85 (2003) 47–67. λ decay constant
[42] P. Bode, M. Blaauw, I. Obrusník, Variation of Neutron Flux and related Parameters in σ(v) neutron cross section, m2, for neutrons with velocity v
an Irradiation Container, in Use with k0-Based Neutron Activation Analysis, σ0 thermal neutron activation cross section at 0.025 eV
J. Radioanal. Nucl. Chem. 157 (1992) 301–312.
σeff effective neutron cross section during irradiation
[43] M.J.J. Koster-Ammerlaan, M.A. Bacchi, P. Bode, E.A. De Nadai Fernandes, A new
monitor for routine thermal and epithermal neutron fluence rate monitoring θ isotopic abundance
in k0 INAA, Appl. Rad. Isotop. 66 (2008) 1964–1969. Φth ‘conventional’ thermal neutron fluence rate for energies up
[44] L.A. Currie, Limits for qualitative detection and quantitative determination — to the Cd cut-off energy of 0.55 eV
application to radiochemistry, Anal. Chem., 40 (1968) 586–593. Φepi ‘conventional’ epithermal neutron fluence rate equivalent
[45] P. Bode, Instrumental and organizational aspects of a neutron activation
analysis laboratory Ph.D. dissertation, Delft University of Technology, Delft.
to the epithermal neutron fluence rate density per unit
(1996). energy interval at 1 eV
208 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Chapter 2. Evaluation of uncertainties for neutron activation analysis measurements

using the comparator method of standardization
Robert R. Greenberg *, Richard M. Lindstrom, Elizabeth A. Mackey, Rolf Zeisler

National Institute of Standards and Technology, Analytical Chemistry Division, 100 Bureau Drive, Gaithersburg, MD 20899-8395, USA

2.1. Introduction fP correction for pulse pileup (correction method depends

upon the actual hardware used)
This chapter describes the individual uncertainty components fltc correction for inadequacy of live time extension (correction
associated with measurements made with neutron activation analysis method depends upon the actual hardware used)
(NAA) using the comparator method of standardization (calibration),
as well as methods to evaluate each one of these uncertainty Note that the “R” values are normally very close to unity, and all
components. This description assumes basic knowledge of the NAA units are either SI-based or dimensionless ratios. Thus an uncertainty
method, and that experimental parameters including sample and budget can be developed using only SI units and dimensionless ratios
standard masses, as well as activation, decay, and counting times have for an NAA measurement by evaluating the uncertainties for each of
been optimized for each measurement. It also assumes that the the terms in Eqs. (2-1) and (2-2), and for any additional corrections
neutron irradiation facilities and gamma-ray spectrometry systems required (e.g., interferences, dry mass conversion factors, etc.).
have been characterized and optimized appropriately, and that the Uncertainties for some of the terms in Eq. (2-1) have multiple
choice of irradiation facility and detection system is appropriate for components. If we sub-divide the uncertainty for each term in the
the measurement performed. Careful and thoughtful experimental above equations into individual components, add terms for potential
design is often the best means of reducing uncertainties. corrections, and separate into the four stages of the measurement
The comparator method involves irradiating and counting a known process, including: pre-irradiation (sample preparation); irradia-
amount of each element under investigation using the same or very tion; post-irradiation (gamma-ray spectrometry), and radiochemis-
similar conditions as used for the unknown samples. The NAA process try, we arrive at the complete list of individual uncertainty
can be described by the following measurement equation (Eq. (2-1)): components for NAA listed in Table 2.1. Only uncertainties from the
first three stages should be considered for instrumental neutron
  activation analysis (INAA) measurements, while all four stages
A0ðunkÞ
mxðunkÞ = mxðstdÞ   Rθ Rϕ Rσ Rε −blank ð2  1Þ should be considered for radiochemical neutron activation analysis
A0ðstdÞ
(RNAA) measurements.
where:
2.2. General irradiation and gamma-ray spectrometry
facility considerations
mx(unk) mass of element x in the unknown sample
mx(std) mass of element x in the comparator standard (calibrator)
A brief discussion of the facility and equipment characterization
A0(unk) decay corrected count rate in the unknown sample
that is necessary before complete uncertainty budgets can be
A0(std) decay corrected count rate in the comparator standard
evaluated is provided here as a precondition for the uncertainty
Rθ ratio of isotopic abundances for unknown and standard
evaluation process addressed later in this chapter. This characteriza-
Rϕ ratio of neutron fluences (including fluence drop off, self
tion addresses the general facility parameters that do not substantially
shielding, and scattering)
change from one measurement to the next; the characterization
Rσ ratio of effective cross section differences (if neutron
required to evaluate components of uncertainty specific to an
spectrum shape changes)
individual measurement is addressed point-by-point in subsequent
Rε ratio of counting efficiencies (differences due to geometry
sections.
and γ-ray self-shielding)
Neutron field characterization: gradients and energy spectrum —
blank mass of element x in the blank
characterization of the neutron irradiation field is necessary to correct
for any potential differences in neutron exposure among samples and
and:
standards, to correct for potential interferences caused by reactions with
high-energy (fast) neutrons, and to evaluate the uncertainties for such
λCeλtd corrections. It is important to characterize both the neutron fluence rate
A0 =   fP fltc ð2  2Þ
1−e−λtL 1−e−λti and the neutron energy spectrum as functions of location and time

within each irradiation facility. Although specific corrections may differ
from measurement to measurement, an initial characterization of each
with: irradiation facility allows the analyst to select the best facility to use for a
particular measurement, and to estimate the magnitudes of potential
A0 decay corrected count rate, irradiation interferences and other effects due to potential variations in
C net counts in γ-ray peak, the neutron fluence rate and spectrum shape. The initial characteriza-
λ decay constant ln 2/t1/2 tion allows the analyst to decide in advance for each analysis how best to
td decay time to start of count position samples and standards within an irradiation container (e.g., a
tL live time of count “rabbit”), and whether precise determinations of correction factors
ti irradiation time (with uncertainties) for fluence-rate variations and/or fast neutron
interferences will be necessary, or whether the likely effects are
⁎ E-mail address: robert.greenberg@nist.gov. insignificant and uncertainty estimates only are sufficient.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Table 2.1
Complete list of individual uncertainty components for NAA measurements using the
comparator method of standardization; line numbers in this table reflect subsection
where the uncertainty component is discussed.
2.3 Pre-irradiation (sample and standard preparation) stage
2.3.1 Elemental content of standards (comparators)
2.3.2 Target isotope abundance ratio — unknown samples/standards
2.3.3 Basis mass (or other sample basis) — including drying
2.3.4 Sample and standard blanks
2.4 Irradiation stage
2.4.1 Neutron fluence exposure differences (ratios) for unknown samples
compared to standards (comparators)
2.4.1.1. Physical effects (fluence gradients within a single irradiation)
2.4.1.2. Temporal effects (fluence variations with time)
2.4.1.3. Neutron self shielding (absorption and scattering) effects within a
single sample or standard
2.4.1.4. Neutron shielding effects from neighboring samples or standards
2.4.2 Irradiation interferences
2.4.2.1. Fast (high energy) neutron interferences
2.4.2.2. Fission interferences
2.4.2.3. Multiple neutron capture interferences
2.4.3 Effective cross section differences between samples and standards
2.4.4 Irradiation losses and gains
2.4.4.1. Hot atom transfer (losses and gains by recoil, nanometer movement)
2.4.4.2. Transfer of material through irradiation container
2.4.4.3. Sample loss during transfer from irradiation container
2.4.4.4. Target isotope burn up differences
2.4.5 Irradiation timing and decay corrections during irradiation (effects of half
life and timing uncertainties)
2.5 Gamma-ray spectrometry stage
2.5.1 Measurement replication or counting statistics (depending on number of
replicates) for unknown samples
2.5.2 Measurement replication or counting statistics (depending on number of
replicates) for comparator standards
2.5.3 Corrections for radioactive decay (effects of half life and timing
uncertainties for each measurement)
2.5.3.1. From end of irradiation to start of measurement
2.5.3.2. Effects of clock time uncertainty
2.5.3.3. Effects of live time uncertainty
2.5.3.4. Count-rate effects for each measurement (either 2.5.2.3.1 and 2.5.2.3.2
or 2.5.2.3.3 but not all three)
2.5.3.4.1. Corrections for losses due to pulse pileup for conventional analyzer
systems
2.5.3.4.2. Effects due to inadequacy of live-time extension for conventional
analyzer systems
2.5.3.4.3. Uncertainties due to hardware corrections for Loss-Free or Zero
Dead Time systems
2.5.4 Corrections for gamma-ray interferences
2.5.5 Corrections for counting efficiency differences (if necessary), or uncertainty
for potential differences
2.5.5.1. Effects resulting from physical differences in size and shape of samples
versus standards
2.5.5.2. Corrections for gamma-ray self absorption
2.5.6 Potential bias due to peak integration method
2.5.7 Potential bias due to perturbed angular correlations (γ-ray directional
effects)
2.6 Radiochemical stage (only if radiochemical separations are employed)
2.6.1 Losses during chemical separation
2.6.2 Losses before equilibration with carrier or tracer
Neutron fluence-rate monitoring (e.g., [1]), is typically used for the
initial characterization of the spatial neutron fluence rate gradient,
and the temporal differences in neutron exposure for each irradiation
facility. Appropriate metal wires or foils are rigidly positioned at
different locations within an irradiation container or containers,
irradiated, and counted under identical conditions. The decay
corrected count rate per unit mass for each piece of wire or foil is
proportional to the monitor's neutron exposure; a comparison of
these values can be used to establish a “map” of the relative neutron
intensity as a function of position within each facility. Repeating this
process at different points in the reactor fuel cycle provides
information on temporal variability. When such measurements
indicate that the fluence rate gradient is stable for a given irradiation
facility, it may be useful to fully characterize that gradient and use
correction factors derived from those measurements for subsequent
209
analyses. The uncertainties associated with these corrections can be
estimated based on the results of characterization measurements. If
this method is used, it is a good practice to include at least a few
fluence-rate monitors with each experiment to verify that the facility
is still stable. If the initial characterization indicates that the fluence-
rate gradients are not stable, a sufficient number of fluence-rate
monitors (or some other method) must be used to provide corrections
for each analysis (see Section 2.4.1).
Initial characterization of the neutron energy spectrum provides
valuable information regarding the general magnitude of effects such
as interferences from fast neutron reactions (see Section 2.4.2), the
potential for neutron energy spectrum changes from neutron
scattering interactions within the sample, and resonance self-
shielding (see Section 2.4.3). This initial characterization allows the
analyst to evaluate whether more precise measurements are needed
(for a particular measurement) to fully evaluate a correction factor
(with an uncertainty), or whether the likely contribution is insignif-
icant and an uncertainty estimate alone is sufficient. For example, if
the irradiation facility is very highly thermalized and the fast neutron
component is very small, a shift in the energy spectrum due to
neutron scattering is not likely to occur and therefore not likely to
significantly contribute to measurement uncertainty. Similarly, con-
tributions from threshold reactions may be less important, depending
on the matrix and relative amounts of the elements of interest (see
Section 2.4.2). Activation foils are commonly used, e.g., [1], to
characterize the neutron energy spectrum by determining the
absolute thermal, epithermal, and fast (fission spectrum) neutron
fluence rates in reactor irradiation facilities and other neutron fields. A
known mass of the appropriate monitor element is irradiated, and the
induced activity is measured with a detector of known efficiency. The
fluence rates are then calculated through use of the appropriate cross
section values (see e.g., [2]).
Gamma-ray spectrometry system characterization: stability over
dynamic range — the uncertainty components associated with
gamma-ray spectrometry cannot be evaluated in any meaningful
way without first assuring that the spectrometry systems are properly
calibrated and optimized for the energy and count-rates ranges to be
used. Gamma-ray spectrometry system optimization and character-
ization are performed to achieve the best resolution and most
symmetric peak shapes, and to determine the dynamic range over
which the system is stable. System stability is needed to assure the
accuracy of software pileup corrections (based on determinations of
the pileup constant), or the accuracy of corrections provided by
hardware to compensate for losses during system dead times, as well
as to assure that peak shapes and resolution are optimal over the
range of count-rates and system dead-times used. After the system is
optimized and the dynamic range for a given system is established,
the analyst should verify that the system is stable over this range at
appropriate intervals (possibly at each use). These procedures help to
reduce uncertainties associated with corrections for pileup, loss-free
or zero dead-time system corrections, and peak integration.
2.3. Uncertainty evaluation — pre-irradiation
(sample preparation) stage
2.3.1 Elemental content of comparator standards
The method used to evaluate this uncertainty will depend upon the
nature of each standard used [3,4]. For highest accuracy work, a high-
purity material may be used directly, or more typically, a standard
solution prepared from a high-purity material is deposited onto filter
paper. In this case there will be uncertainties associated with:
• the purity and/or stoichiometry of the high-purity material used
• potential losses during dissolution
• gravimetric or volumetric standard solution preparation
210 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

• gravimetric or volumetric deposition onto individual filter papers and
• contamination during the standard preparation process
• losses of solvent during the standard preparation process
• determination of the element content of the filter paper, acids and
water or other diluent used in preparation, or “blank” [5,6].
If a high-purity material CRM is used as the starting material, the
uncertainty for the elemental content of the material is given on the
certificate supplied by the CRM provider. If a high-purity CRM solution
is used, the uncertainty for dissolution losses is included in the
uncertainty of the certified value. If multielemental standards are
prepared, special caution must be taken to evaluate impurity
contributions of one element to another, since typically multi-
elemental standards for NAA contain widely differing amounts of
the different elements [3]. In some cases, matrix CRMs are used as
standards for lower-accuracy NAA measurements, however, care
must be taken to include the uncertainty of the certified value of the
content of each element in the CRM, since these are often quite large.
2.3.2. Target isotope abundance ratio — unknowns/standards
NAA is an isotopic technique and Eq. (2-1) considers differences in
the isotopic abundances of elements (Rθ = ratio of isotopic abun-
dances for unknown and standard). While no corrections or
uncertainties are needed for the determination of monoisotopic
elements (Rθ = 1 with absolutely no uncertainty), corrections and
uncertainties may be necessary for elements with more than one
isotope and whose abundances vary in nature or are perturbed by
man. Table 2.2 lists the elements relevant for NAA for which IUPAC [7]
identified uncertainties in their atomic weights due to natural or man-
made variations in their isotopic composition.
The isotopic abundance variability influences the NAA results in
widely varying degrees depending on the isotope used in the
measurement. In many instances the uncertainty of the abundance
ratio is insignificant, in particular for determinations based on a highly
abundant isotope. However, if a minor isotope is used, the isotopic
abundance ratio between unknown and standard may differ

Table 2.2
Elements relevant in NAA with natural or man-made uncertainties in their isotopic composition; abundance values are expressed as a range or as a single value with an uncertainty
[9]. Values of u(Rθ ) given in the last column were calculated assuming a triangular distribution for the ratio of isotopic abundances of unknowns and standards.

Atomic number Element Name Atomic weight Note NAA target Abundance (%) u(Rθ)
18
8 O Oxygen 15.9994(3) a,b O 0.2217–0.1877 4.8%
22
10 Ne Neon 20.1797(6) a,c Ne 9.96–9.20 2.3%
30
14 Si Silicon 28.0855(3) b Si 3.102–3.082 0.19%
36
16 S Sulfur 32.065(5) a,b S 0.019–0.013 10.8%
37
17 Cl Chlorine 35.453(2) c Cl 24.356–24.077 0.33%
40
18 Ar Argon 39.948(1) a,b Ar 99.6003 (6) 0.0009%
41
19 K Potassium 39.0983(1) a K 6.73022 (292) 0.06%
48
20 Ca Calcium 40.078(4) a Ca 0.188–0.186 0.31%
51
23 V Vanadium 50.9415(1) a V 99.7513–99.7498 0.0004%
63
29 Cu Copper 63.546(3) b Cu 69.338–68.983 0.15%
65
Cu 31.017–30.662 0.33%
86
36 Kr Krypton 83.798(2) a,c Kr 17.29835(26) 0.002%
85
37 Rb Rubidium 85.4678(3) a Rb 72.1654 (132) 0.03%
84
38 Sr Strontium 87.62(1) a,b Sr 0.58–0.55 1.5%
88
Sr 82.75–82.29 0.16%
94
40 Zr Zirconium 91.224(2) a Zr 17.380 (12) 0.98%
98
42 Mo Molybdenum 95.96(2) a Mo 24.19 (26) 1.52%
102
44 Ru Ruthenium 101.07(2) a Ru 31.55 (14) 0.63%
110
46 Pd Palladium 106.42(1) a Pd 11.72 (9) 1.1%
109
47 Ag Silver 107.8682(2) a Ag 48.161 (8) 0.02%
114
48 Cd Cadmium 112.411(8) a Cd 28.73 (42) 2.1%
116
50 Sn Tin 118.710(7) a Sn 14.54 (9) 0.88%
121
51 Sb Antimony 121.760(1) a Sb 57.21 (5) 0.12%
123
Sb 42.79 (5) 0.17%
130
52 Te Tellurium 127.60(3) a Te 34.08 (64) 2.7%
134
54 Xe Xenon 131.293(6) a,c Xe 10.4357 (21) 0.03%
139
57 La Lanthanum 138.90547(7) a La 99.910 (1) 0.001%
140
58 Ce Cerium 140.116(1) a Ce 88.449–88.446 0.001%
146
60 Nd Neodymium 144.242(3) a Nd 17.35–17.06 0.49%
152
62 Sm Samarium 150.36(2) a Sm 26.75 (16) 0.85%
153
63 Eu Europium 151.964(1) a Eu 52.19 (6) 0.16%
152
64 Gd Gadolinium 157.25(3) a Gd 0.20 (1) 7.1%
160
Gd 21.86 (19) 1.2%
164
66 Dy Dysprosium 162.500(1) a Dy 28.260 (54) 0.27%
170
68 Er Erbium 167.259(3) a Er 14.910 (36) 0.34%
168
70 Yb Ytterbium 173.054(5) a Yb 0.13 (1) 11%
176
Yb 12.76 (41) 4.5%
176
71 Lu Lutetium 174.9668(1) a Lu 2.59 (2) 1.1%
180
72 Hf Hafnium 178.49(2) a Hf 35.100–35.076 0.02%
184
76 Os Osmium 190.23(3) a Os 0.02 (1) 71%
192
Os 40.78 (19) 0.66%
235
92 U Uranium 238.02891(3) a,c,d U 0.7207–0.7198 0.036%
238
U 99.2752–99.2739 0.0004%
a
Geological specimens are known in which the element has an isotopic composition outside the limits for normal material. The difference between the atomic weight of the element
in such specimens and that given in the Table may exceed the stated uncertainty.
b
Range in isotopic composition of normal terrestrial material prevents a more precise value being given; the tabulated value should be applicable to any normal material.
c
Modified isotopic compositions may be found in commercially available material because it has been subject to an undisclosed or inadvertent isotopic fractionation. Substantial
deviations in atomic weight of the element from that given in the table can occur.
d
Element has no stable nuclides. However a characteristic terrestrial isotopic composition and an atomic weight are tabulated.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
significantly from unity. An extreme example in which the isotopic
ratio between unknown and standard can be significantly different
from unity is the determination of sulfur by NAA using the reaction 36S
(n,γ)37S [8]. In this example, Rθ can range from 1.4 to 0.7 depending
whether the sample or the standard, or both, exhibit an extreme
variation in the abundance of 36S, resulting in a bias of up to 40% if
these abundance differences are ignored. However, most isotopes
used for NAA vary only slightly in nature.
Corrections and uncertainties associated with isotopic abundance
differences between samples and standards may be directly included
in the uncertainty budget if they are known (or have been measured).
However, this type of information is rarely available. When the actual
abundances are not known, no corrections can be applied and Rθ must
be assume to be unity. However, the uncertainty for Rθ = 1 can be
estimated from the degree of natural variability for samples and
standards that have not been isotopically manipulated using the data
from Table 2.2.
For a simple case where accurate values for the isotopic
abundances in samples and standards are not known, it is reasonable
to assume that the abundance values for samples and standards lie
within the range given in Table 2.2. The relative range of potential
values for the isotope abundance for either sample or standard would
be:
! of the components of that correction factor, i.e., the wet mass to dry
θH −θL
rangeð%Þ = 1  100; ð2  3Þ
2 ðθH + θL Þ
where θH = highest abundance, and θL = lowest abundance in
Table 2.2.
If we assume that the actual (“true”) value of sample and standard
isotope abundances can lie anywhere between the two extremes of
this range, but the actual (“true”) isotopic abundances of both sample
and standard most likely lie at the midpoint of the two extremes, our
best estimate of Rθ is 1. The uncertainty for Rθ will be 1/2 the relative
range divided by √6 (for a triangular distribution) multiplied by √2 to
account for the uncertainty of both the sample and standard. On the
other hand, if we have no reason to believe that both sample and
standard abundances most likely lie at the midpoint of the two
extremes, or if we wish to be more conservative, our best estimate of
Rθ is still 1, however, to evaluate the uncertainty for Rθ, we should
divide 1/2 the relative range by √3 (for a rectangular/uniform
distribution) instead of √6 and multiply by √2 to account for the
uncertainty of both the sample and standard. If an abundance
uncertainty in Table 2.2 is given for the isotope abundance, instead
of an abundance range, the uncertainty in Rθ, can be estimated by
dividing the abundance uncertainty by the isotope abundance, and
multiplying by √2 (to account for differences in samples and
standards).
In the case of the determination of S via the 36S(n,γ)37S reaction
discussed above, if we wished to make the measurement (despite the
large potential bias resulting from a potential Rθ range of 40%), and we
assume that the actual (“true”) isotopic abundances of both sample
and standard most likely lie at the midpoint of the two extremes of
0.019% and 0.013% (e.g. there are multiple sources of S composing
both samples and standards), our best estimate of Rθ would be 1, and
we would assume a triangular distribution. The relative standard
uncertainty for Rθ would be 38%/2* √6*√2, or 10.8%. However, if we
wish to be more conservative, or we have no reason to believe that
both sample and standard abundances most likely lie at the midpoint
of the two extremes, our best estimate of Rθ would still be 1, however,
we would assume a uniform/rectangular distribution, and evaluate
the relative standard uncertainty for Rθ as 38%/2* √3*√2, or 15.3%. The
choice between these two possibilities should be based on chemical
knowledge of the samples and standards.
For practically all NAA applications the mid-point of the range or
the stated isotopic abundance is a valid assumption. The notes in
211
Table 2.2 are intended to draw attention to the possibility of
deviations in rare cases of special geological materials, or samples
from industrial materials where elements with enriched or depleted
abundances are used. Obtaining assurance of the true isotopic
abundance of the unknown sample will be necessary for a reliable
NAA determination based on isotopes with abundances exhibiting
significant variability in the types of samples analyzed or standards
used.
2.3.3. Determination of the basis mass (or other sample basis)
The uncertainty associated with determining the mass basis of the
sample may be simply the uncertainty associated with use of an
analytical balance. For the analysis of many biological or environ-
mental samples such as tissues or sediments, the element mass
fractions of the sample are typically reported on the basis of dry mass.
The sample used for NAA may have been dried in advance or may be
analyzed as received and the dry mass basis determined using
separate portions. In either case the uncertainty associated with the
determination of the correction factor will be that associated with the
mass determinations. Similarly, in the analysis of biological tissues,
the mass fraction values may be reported on a fresh tissue basis, a lipid
basis, or protein basis but the material will be analyzed as a dried
material. The uncertainty is calculated from the determination of each
mass, the lipid to wet mass, or protein to wet mass ratios. Changes in
the mass basis that occur as a result of irradiation are discussed in a
separate section below.
2.3.4. Sample blanks
Due to time constraints in the determination of elements for which
the products of neutron irradiation are short-lived isotopes, or due to
the nature of the samples, it may not be practical or possible to
transfer each sample from the irradiation container or packaging
material into a clean (unirradiated) container prior to performing
gamma-ray spectrometry. In these cases, a complete characterization
of the element content of the packaging material (and its variability
within a lot) may be required to account for its contribution to the
total amount of element measured. Where possible, it is preferable to
transfer the samples from the packaging material into non-irradiated
containers to eliminate the need for this type of correction. Where
transfer is not practical or preferred, a statistically valid number of
blanks must be analyzed to determine the amount of element per unit
mass of packaging material in order to derive appropriate corrections
for those elements. Evaluation of the uncertainty will include the
variability of amount of element in the blank and the uncertainty
associated with the analytical determination of amount content.
Uncertainty associated with analyte losses (or gains) during irradi-
ation is discussed in a separate section below.
2.4. Uncertainty evaluation: irradiation stage
2.4.1. Neutron fluence exposure differences (ratios) for samples
compared to comparator standards
As induced activity is directly proportional to the neutron
exposure, the ratio of decay-corrected activities of unknown samples
and comparator standards is also directly proportional to the neutron
exposure of the samples versus the standards. Assuming that they are
irradiated at nominally the same neutron fluence rate in the same
reactor facility, for the same amount of time, there are four factors that
result in unequal neutron exposures: temporal and spatial neutron
fluence-rate variations, and neutron absorption and scattering effects.
Spatial neutron fluence rate variations or irradiation geometry effects
result from differences in neutron fluence rate at the different
positions of each sample and standard in the irradiation facility.
Temporal neutron fluence rate changes are also common within many
212 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
facilities and must be evaluated when more than one irradiation is
required. Neutron self-shielding can be significant when the outer
portion of a sample or standard absorbs a significant fraction of the
neutrons, and so the interior of that sample or standard gets a reduced
neutron exposure. Neutron scattering effects, although very small for
NAA, may perturb the neutron exposure of some samples in a manner
similar to absorption effects. Although it is possible to evaluate the
absorption and scattering effects separately, it is most convenient to
evaluate their combined effect as described below.
2.4.1.1. Differences in neutron exposures for unknown samples compared
to standards (comparators) due to physical effects (fluence gradients
within a single irradiation)
In a typical irradiation facility, the reactor fuel is not symmetri-
cally located around the irradiation position, and therefore the
neutron fluence rate differs in different locations within the rabbit
leading to potential differences in neutron exposures among and
between standards and unknowns within one irradiation. In general,
the neutron fluence rate decreases with distance from the reactor
core and if the rabbit's axis is perpendicular to the flux contours, the
gradient is nearly linear. The analytical bias introduced by this
gradient can be compensated by inverting the rabbit midway
through the irradiation time (i.e., by removing the rabbit midway
through the irradiation, rotating (flipping) it 180°, and reinserting
the rabbit for the second half of the irradiation). The results of using
this procedure to minimize the linear gradient at the NBSR are shown
in Fig. 2.1. This procedure greatly decreases but does not entirely
eliminate effects of the neutron gradient within the irradiation
container.
Other physical manipulations can be used to equalize the neutron
exposure within the irradiation container. Many reactors have a
“Lazy-Susan” facility for long irradiations where the samples are
rotated throughout the irradiation. In either case, residual differ-
ences can be corrected, and uncertainties evaluated, as described
under the flux-monitoring approach, or if corrections are not
necessary, uncertainties can be evaluated as for the historical-data
approach.
2.4.1.2 Differences in neutron exposures for unknown samples compared
to standards (comparators) due to temporal effects
Over the course of a reactor cycle, neutron fluence rates within a
single facility vary as a function of time so that differences in neutron
exposures exist from one irradiation to the next. During the first hours
of reactor operation, the moderator temperature gradually reaches a
1.0
0.9
Relative fluence rate
0.8
0.7
Flipped
0.6
0.5
0.4
0 1 2 3 4 5 6 7
Axial position, cm
Fig. 2.1. Neutron fluence exposure variations for samples irradiated in one irradiation
facility in the NBSR.
steady state and thus the neutron spectrum will change during this
time. As the fuel is consumed, regulating rods move in the lattice,
changing the spatial distribution of the neutron field. It is essential
that each irradiation facility be fully characterized for both spatial
neutron gradients within a given irradiation, and temporal differences
in neutron fluence rates from one irradiation to the next for NAA
experiments that involve more than one irradiation.
Corrections for both physical and temporal differences in
neutron exposure (described in Sections 2.4.1.1 and 2.4.1.2), and
their associated uncertainties can be evaluated using historical data
and/or fluence monitoring measurements. Historical data is partic-
ularly useful for facilities that are stable over time, especially when
no corrections are necessary. In this case, a large number of identical
samples are irradiated and counted (preferably far from the
detector) with very good counting statistics for one or more
gamma rays. The relative uncertainty due to irradiation geometry
is the observed relative standard deviation (rsd), reduced in
quadrature by the uncertainty due to counting statistics. As an
example, an irradiation container filled with identical samples is
irradiated. The samples are individually counted, and the observed
rsd of the decay corrected count rate per unit sample mass of a
radionuclide is 0.5%. If the typical counting uncertainty for each
sample is 0.3%, and all other sources of variation are negligible, the
uncertainty due to irradiation geometry effects would be the square
root of [(0.5%)2–(0.3%)2], or 0.4% for each sample (or standard). If
there are additional significant sources of variation, such as counting
geometry differences, those uncertainties can also be subtracted (in
quadrature) from the observed rsd. This uncertainty of 0.4% can be
used to characterize each sample and standard irradiated in all
future irradiations in this facility, as long as nothing is done to
change the flux distribution. If one sample and one standard are
irradiated together in this facility, the combined irradiation
geometry uncertainty for sample and standard would be the square
root of [(0.4%)2 + (0.4%)2], or 0.57%. If multiple samples (of the same
material) and/or standards are irradiated together, the sample
counting geometry uncertainty and the standard counting geometry
uncertainty (for the mean value) would be reduced by the square
root of the number of samples and/or standards irradiated.
Neutron fluence monitors (also known as flux monitors) are often
used to determine the static spatial gradient, the temporal gradient,
and neutron spectrum changes. Results from NAA experiments are
often normalized by irradiating fluence monitors of known mass and
uniform composition along with the samples. A number of monitors
have been used: foil and wire of iron, copper, nickel, zinc, or other
metals, standard solutions pipetted onto filter paper, or specially
fabricated alloys such as Au/Al (IRMM-530R) or Co/Al (NIST SRM
953). For relative measurements it is not necessary to determine the
absolute activity of the monitors. After irradiation, the monitors are
removed and gamma rays counted to determine the relative neutron
exposure for each sample, and corrections applied, if necessary.
Uncertainties for these corrections can be determined from the
counting statistics observed for the monitors. As with the historical
data case, if multiple samples (of the same material) and/or standards
are irradiated together, the sample counting geometry uncertainty
and the standard counting geometry uncertainty (for the mean value)
would be reduced by the square root of the number of samples and/or
standards irradiated.
2.4.1.3. Neutron self-shielding: neutron absorption and scattering effects
within a single sample or standard
When a sample or standard is immersed in a neutron field, its
8
interior will be exposed to a smaller neutron fluence rate than the
exterior due to neutron absorption and scattering by the sample. This
effect, neutron self-shielding, varies with the macroscopic scattering
and absorption cross-sections and the size and shape of the sample.
The correction factor for neutron self-shielding, fϕ, is the ratio of the
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 213

1.000
mean fluence rate over the entire sample volume, compared to the
fluence rate incident on the sample. For an absorbing and scattering
sample surrounded by an isotropic neutron field, the value of fϕ is 0.999

Self-shielding factor f
always less than unity [10], unless the sample contains fissile
materials. 0.998

2.4.1.3.1. Monte Carlo methods and calculated values 0.997

A rigorous calculation of self-shielding correction factors requires
knowledge of the neutron energy spectrum inside and outside the
sample, and the absorption and scattering cross sections as a function
of energy. Monte Carlo methods may be used to determine values
for fϕ, but few calculations performed to date have included both
changes in the neutron energy spectrum due to neutron scattering
and spectrum averaged cross-section values for the nuclides of
interest.
Scattering affects the history of neutrons in the sample even if
there is no change in neutron energy. Copley [10] studied the effects
of neutron scattering and absorption within spherical samples
immersed in an isotropic neutron field and calculated the values of
fϕ using an iterative solution to the neutron transport equation and
Monte Carlo methods. Results from both approaches were consis-
tent and indicate that neutron scattering may decrease fϕ values,
contributing to further self-shielding, for samples that both scatter
and absorb neutrons within an isotropic neutron field. Similar
results are expected and have been demonstrated for other sample
shapes. Copley's results for spherical samples in an isotropic
neutron field were approximated by Blaauw [11] using a modifica-
tion of the Stewart and Zweifel [12] formulation. Subsequently
Salgado et al. [13] modified Blaauw's formulation with similar
agreement.
Analytic expressions for fϕ as a function of the scalar cross sections
and sample composition have been derived for a few specific
geometries. A frequently cited summary paper is that of Fleming,
[14], who collected the equations for a sphere, an infinite slab, and an
infinite cylinder, irradiated in either an isotropic mono-energetic
neutron field or a monoenergetic parallel neutron beam. By employ-
ing a geometrical argument [15], these equations have been extended
to finite cylinders, a common configuration in many irradiation
experiments. In addition, Fleming tabulated macroscopic cross-
section values for a large number of materials and mixtures. For
most materials, where the comparator standard and standard have
similar scattering properties, an approximation based on calculations
for absorption only using a slab geometry introduces little or no bias,
e.g., when comparing biological samples with comparator standards
prepared using filter paper matrices. The approximation of fϕ for a slab
of thickness T, taking into account only absorption is
0.996
0.995
Sphere
0.994
0.1 1 10
Length/Diameter
Fig. 2.2. Calculated self-shielding factor fϕ for 250-mg cylinders and a sphere of SRM
1632c bituminous coal. For this material Σa = 0.015 cm− 1 and Σs = 2.7 cm− 1: the
major absorbers are hydrogen, boron, and nitrogen, and the major scattering elements
hydrogen and carbon.
cross sections between 10 and 20 b. Based on Copley's results for
spheres within a range of macroscopic absorption and scattering
cross-sections corresponding to the ranges expected of most
biological and environmental samples, the bias introduced by
ignoring scattering is typically between ≤0.1% and 0.3%, depending
on the diameter of the sphere and the values for those cross-sections
[10]. The bias introduced would be worse for very strong absorbers
(∑a ≥ 2 cm− 1), though there are few examples of materials this
strongly absorbing.
Spreadsheets for calculating fϕ for the Fleming geometries are
available on request from one of the authors (richard.lindstrom@nist.
gov). An example for a 200-mg pellet (radius = 0.65 cm) of Bovine
Liver (SRM 1577c) is given in Fig. 2.3. Note that the calculated
correction for scattering is significantly smaller for the pellet (0.995)
compared to the sphere (0.991), which is listed in Table 2.3.
When these spreadsheets are used to calculate fϕ values for
samples and standards, an uncertainty estimate of about 10% of the
correction is reasonable in cases where the mass fractions of all
major absorbing and scattering components reasonably well-known.
When the neutron spectrum contains epithermal neutrons then
resonance self-shielding must also be taken into account [16]. A
quasi-universal self-shielding expression has been formulated [17,
18] and shown to be adequate for most work in a mixed neutron
field [19].

2.4.1.3.2. Experimental fϕ evaluation

1 = 2−E3 ð∑a T Þ
fϕ = ; ð2  4Þ The consequence of thermal scattering by hydrogen on induced
∑a t
activity for NAA has been determined experimentally [20] by
where ∑a is the macroscopic absorption cross section and E3 is the irradiating sandwiches of iron foils in the well-thermalized NIST
exponential integral of order 3. Corrections for neutron self-shielding reactor with and without polyethylene spacers. No significant
for typical 200-mg biological or environmental samples are normally difference was detectable at a precision of order 0.1%. One may best
on the order of several tenths of a percent, and are rarely greater than
a percent or two. Strongly absorbing materials such as some metals or
some ceramics may require more careful evaluation. The shape of the Table 2.3
sample may be important. For a cylindrical sample fϕ as a function of Calculated self-shielding factors for 200 mg spherical samples of representative
the length/diameter ratio is shown in Fig. 2.2. The maximum materials.
absorption occurs when the shape is most nearly spherical.
Material Self-shielding factor Principal absorbers
Self-shielding factors (including both absorption and scattering)
Quartz 0.999 Si
have been calculated for 200-mg spherical samples of some represen-
SRM 2709a Soil 0.997 Si, Fe
tative materials typically analyzed by NAA and are given in Table 2.3. SRM 1632c Coal 0.994 H, C
In practice, hydrogen is the most important scattering element, SRM 1633b Fly Ash 0.998 B, Fe
with a free-atom absorption cross-section of 20.5 b and a maximum Cellulose 0.996 H
bound atom cross-section of 82 b. Other important scattering SRM 1577c Bovine Liver 0.991 H, N
Iron 0.970 Fe
elements in typical samples are N, Cl, Fe, Ni, and Pb, with scattering
214 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Fig. 2.3. Calculated self-shielding for a 200-mg pellet of SRM 1577c bovine liver.
estimate the value of fϕ for a given material experimentally by
performing NAA on a series of samples of varying thicknesses and
determining the correction factors based on comparison with a thin
sample for which the fϕ value is essentially 1. Indeed, for a material of
unknown composition this is the only possible method.
2.4.1.3.3. Uncertainty in fϕ estimates
Correction factors may be determined by appropriate Monte Carlo
methods, calculated using one of the approximations, or determined
experimentally. Where determined experimentally or by Monte Carlo
simulation, the uncertainty associated with the correction factor may
be determined based on statistical considerations of the method used.
Where formulaic approximations are used, the uncertainty of the
correction factor may be estimated as 10% of the correction. Note that
for either Monte Carlo simulations or calculated corrections based on
either ∑a or ∑a and ∑s, additional uncertainty may need to be added
when the sample composition is partially unknown.
2.4.1.4. Neutron shielding effects from neighboring samples or standards
When a set of multiple samples and standards are irradiated
together, neutron absorption and scattering effects on individual
samples/standards resulting from absorption and scattering by neigh-
boring ones should be considered, especially when they are of large size,
closely packed, and/or are composed of highly absorbing constituents.
Correction for fluence depression by neighboring samples/standards
can often be calculated if the sizes, shapes and compositions of all
materials are known; however, corrections are often most conveniently
accomplished by normalizing to embedded fluence monitors, at least for
samples with sizes and compositions typically analyzed by NAA. Large
samples (i.e., much greater than 1 g), especially those with a high
content of neutron absorbing materials, may require additional
corrections. In addition to fluence monitoring corrections for effects
external to the samples/standards, neutron fluence monitors can be
used to determine and correct for the effects of static spatial fluence
gradient, the temporal gradient, and neutron spectrum changes (as
discussed in Sections 2.4.1.1 and 2.4.1.2). However, the effects of
neutron self-shielding within each sample (or standard) need to be
addressed separately (as discussed in Section 2.4.1.3).
The number and placement of fluence monitors will depend upon
the number, mass, shape, composition, closeness (or packing density)
of the samples and standards, as well as knowledge of the fluence
gradients in the irradiation facility when no samples or standards are
present. For example, in an irradiation facility with a linear drop-off
perpendicular to the samples/standards, which are all relatively small,
well-separated and not particularly absorbing, one fluence monitor at
each end of the irradiation container and one in the middle of the
sample/standard pack may be sufficient. However, for relatively
absorbing samples that are closely packed, a fluence monitor between
every sample and/or standard may be necessary. The average decay
corrected activity for the two surrounding fluence monitors provides
the correction factors for each sample and/or standard. Uncertainties
for the corrections will depend upon the counting statistics for each of
surrounding monitors, combined in quadrature.
The NAA measurement with the perhaps the highest demonstrat-
ed accuracy and smallest uncertainty to date was the measurement of
Cr in stainless steel (SRM 1152a) by Zeisler et al. [21]. Iron fluence
monitors were used to correct for neutron exposure differences at the
surface of the samples and standards, and neutron self-shielding
corrections were calculated as described in Section 2.4.1.3. The
measured Cr value (for 300 mg samples) with an expanded relative
uncertainty of 0.2% agreed within 0.05% relative to the certified value,
which also had an expanded relative uncertainty of 0.2%. Chromium
calibration was accomplished with pure metal chips and slabs
(approx. 1.5 mg and 80 mg each). The uncertainty from the neutron
self-shielding corrections actually dominated all other uncertainty
components (approx. 90% of the combined uncertainty) due to the
difficulties in characterizing the actual dimensions and shapes of the
metal pieces used. Most of the samples and standards each required
relative self-shielding corrections of several percent.
2.4.2. Irradiation interferences
During the irradiation step, three types of interferences are
possible for some of the elements frequently determined by NAA.
Most nuclear reactions used for NAA are based on reactions with
thermal or epithermal neutrons. However, high-energy neutrons can
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 215

knock out a proton or alpha particle from a nucleus, and interfere with of magnitude greater. Even in RT-2, there are several significant
the determination of some elements. For example, Al is typically interferences that are routinely encountered: the (n,p) interference of
determined via the reaction 27Al(n,γ)28Al. 28Al can also be produced Al on the determination of Mg in geological and related materials, and
by Si and P with high-energy neutrons via the reactions 28Si(n,p) 28Al the (n,α) interference of 31P on the determination of Al in biological
and 31P(n,α) 28Al [22]. The relative size of the interference will materials. Other interferences have been significant in a few selected
depend upon both the mass fraction of Al compared to that of each applications, such as the (n,α) interference of 54Fe on the determi-
interfering element in the sample, as well as the fraction of fast nation of Cr in whole blood.
neutrons in the irradiation facility relative to the amount of thermal These interferences were calculated using published fission-
neutrons. Fortunately, these (n,p) and (n,α) reactions cannot occur spectrum cross sections [23] to calculate products that have not been
unless the incoming neutron carries enough energy to equal the mass measured directly. Most values were normalized to the measured 28Si
difference between reactants and products, as well as the additional (n,p) 28Al interference. It should be noted that although these
energy needed to overcome the Coulomb barrier for emission of interferences were normalized to the (n,p) reaction on 28Si, the
charged particles. Only a few such reactions are typically encountered measured 51Cr interference for the (n,α) reaction on 54Fe was within
in neutron activation analysis. In a similar manner, thermal fission of 10% of the similarly calculated value in RT-4, which has a neutron
235
U will produce (as fission fragments) some of the same nuclides spectrum very similar to RT-2. The measured 28Si (n,p) 28Al) interference
used to determine other elements by NAA, e.g., 99Mo is one of the in RT-4 was measured to be 43 μg-Al/g-Si versus 46 μg-Al/g-Si in RT-2.
most commonly produced fission products and is also used to As mentioned above, interferences in a light water reactor are
determine Mo via the reaction 98Mo(n,γ) 99Mo. Finally, in rare cases considerably higher. To estimate the severity of interferences, the fast-
interferences can be caused by double neutron capture. neutron fluence rate in the irradiation facility can be determined [1] by
irradiation of any of the pure nuclides in the preceding table and scaling to
2.4.2.1. Fast (high energy) neutron interferences the fast/thermal ratio in RT-2. It is relatively easy to determine the (n,p)
The more significant fast neutron interferences for the RT-2 interference of 28Si on 28Al, by irradiating some high purity Si (e.g., a piece
pneumatic tube facility of the NIST (NBSR) reactor are listed in of semiconductor grade Si metal) with a small piece of Al foil, and
Table 2.4, where the thermal and fission fluence rates are 3.4 × 1013 determine the apparent Al content of the Si. The scaling factor for other
and 1.4 × 109 cm− 2s− 1, respectively. The interferences (X0) listed fast reactions will be the ratio of the measured Si interference (in μg/g) to
represent the apparent μg of the interference per gram of the the value of interference (46 μg/g) determined for RT-2. However, the
interfering element irradiated. Note that this facility is exceptionally calculated interference values should only be used as a guide to determine
well thermalized, since the NBSR is a D2O moderated reactor and RT-2 which interferences may be significant. If interference is judged to be
is located far from the reactor core. Fast neutron interferences in most significant, it should be measured and corrected. The uncertainty for the
light water moderated reactors would typically be two to three orders corrected value is determined from the uncertainty of the interference
ratio measurement, the uncertainty of the content of the interfering
element in the sample of interest, and the uncertainty of the
determination of the element of interest before the correction is applied.
Table 2.4
Fast neutron interferences for the RT-2 facility of the NBSR. 2.4.2.2. Fission interferences
Fission of 235U will produce (as fission fragments) some of the same
Analyte Activation Half life Interfering σ(fast), mb X0, μg/g
element product nuclide nuclides that are used to determine other elements by NAA. Some of the
more important interferences are for molybdenum and barium. These
(n,p) interference
Na 24
Na 15 h 24
Mg 1.53 4.0
and other potentially significant ones are listed in Table 2.5, with the
Mg 27
Mg 9m 27
Al 4.0 1566 apparent interference in μg of element per μg of natural uranium in the
Al 28
Al 2m 28
Si 6.6 46 sample. Since the interferences are due to 235U fission, they will be
31 31
Si S 3h P 36 17,353 different for samples with non-natural isotope abundances of uranium.
32 32
P P 14 days S 69 626
35 35 Two of the interferences listed need further discussion. Although 131Ba is
S S 88 days Cl 78 7811
S 37
S 5m 37
Cl 0.218 2657 not produced by fission, the 496 keV gamma ray from 131Ba, frequently
Sc 46
Sc 80 days 46
Ti 12.5 0.064 used for the determine barium by NAA, can be interfered with by the
497 keV gamma ray of 103Ru, which is a fission product. This can be an
51 51
Ti Ti 6m V 0.87 157
52 52
V V 4m Cr 1.09 0.33 important interference, especially in geological type materials. Due to
56 56
Mn Mn 3h Fe 1.07 0.13
Fe 59
Fe 40 days 59
Co 1.42 653
the difference in half-lives of 131Ba (11.5 days) and 103Ru (39 days), the
Co 60
Co 5y 60
Ni 2.1 0.026 apparent interference increases with decay time. In addition, 140La
65 65
Ni Ni 3h Cu 0.48 15.8 (40.3 h) is not produced directly from uranium fission, but as a daughter
64 64
Cu
66
Cu 13 h
66
Zn 31 8.8 of the 140Ba (11.7 days) fission product. Since most geological materials
Cu Cu 5m Zn 0.62 0.4
76 76 have much higher mass fractions of lanthanum compared to uranium,
As As 26 h Se 0.17 0.007
Rb 88
Rb 18 m 88
Sr 0.15 7.2 this interference is rarely significant about a week or so after irradiation,
Sr 89
Sr 50 days 89
Y 0.31 113.5 but becomes increasingly more important as the decay time increases.
In view of the potential for fission interferences, it is important to
(n,α) interference know the approximate level of uranium in samples when determining
24 27
Na Na 15 h Al 0.725 2.1
Mg 27
Mg 9m 30
Si 0.155 1.8
elements that have fission interferences. If the amount (or mass
Al 28
Al 2m 31
P 1.9 13.0 fraction) of uranium is known, the potential interferences can be
P 32
P 14 days 35
Cl 8.8 57.6 approximated using the information in Table 2.5 (data from [24]),
52 55
V V 4m Mn 0.11 0.038 however, if the uranium content is not known, and if elements with
51 54
Cr Cr 30 days Fe 0.6 0.09
56 59 potential interferences are under investigation, it is important to
Mn Mn 3h Co 0.156 0.02
Fe 59
Fe 40 days 62
Ni 0.09 1.49 determine uranium in the samples under investigation, subtract any
Co 60
Co 5y 63
Cu 0.50 0.02 relevant interferences, and calculate uncertainties. This can be
65 68
Ni Ni 3h Zn 0.05 1.02 accomplished by preparing single-element uranium standards,
76 79
As As 26 h Br 0.02 0.004 irradiating these standards with the samples under investigation,
89 92
Sr Sr 50 days Zr 0.01 0.86
and counting them with the samples. The standards are first used to
216 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Table 2.5 given neutron energy E. However, if an irradiation facility is not well
Typical uranium fission interferences — apparent μg of element per μg uranium. thermalized, neutron scattering within a sample may change the
Element Isotope Thermal neutrons Epithermal neutrons neutron energy spectrum within a sample, with quantitative
99 consequences. With a harder neutron spectrum (a larger epithermal
Molybdenum Mo 1.4–31 0.1
Barium2 131
Ba 3.4 e0.0402D component), scattering of high-energy neutrons within the sample
Barium 139
Ba 0.6 shifts the spectrum downward in energy. Because neutron cross
141
Cerium Ce 0.3 sections typically are inversely proportional to neutron velocity (the
140
Lanthanum La 0.02 (after 6 days decay)3
147 “1/v law”), this leads to an increase in reaction rate, and results in an
Neodymium Nd 0.2
Ruthenium 103
Ru 0.1 increase in apparent element mass fraction in highly scattering
Tellurium 131
Te 0.8 0.5 samples when compared to a standard that scatters less. While this
95
Zirconium Zr 10 3 may be significant in a poorly thermalized facility, there is no large
97
Zirconium Zr 20 change in effective cross section for a well-thermalized irradiation
1
Dependent on thermal to epithermal ratio due to high resonance integral for Mo facility, since thermal neutrons dominate the reaction rate and
activation. scattering overall will not result in a net change in the average
2
Contribution to the 496.2 keV line from 103Ru. D signified decay time (td) in days
energy of the neutrons. Although most irradiation facilities in light
from end of irradiation to start of count. Greater decay intervals result in greater
apparent interferences. water or heavy water reactors are well thermalized, the degree of
3 140
La daughter (40.3 h) fed by longer lived 140Ba (11.7 days). Interference can be thermalization should be evaluated during the initial reactor
very significant after long decay. characterization measurements. An example has been given by St.
Pierre and Kennedy [28].
determine the uranium mass fractions in the samples. The U In a few very rare cases, a change in the neutron energy spectrum
standards are then are processed as samples themselves to determine may be caused by resonance self-shielding. This effect refers to
the interference ratios, i.e., the apparent amounts of the interfered removal of a significant portion of resonance energy neutrons by a
elements per μg (or μg/g) of uranium (in the standard). The material with a large resonance absorption cross-section, for instance
uncertainty for each fission interference correction will include the gold, indium, samarium and gadolinium. Because activation is
uncertainty of the determination of the uranium content in each proportional to the product of the neutron fluence rate and the
sample (predominantly from the counting statistics of the sample and energy-spectrum averaged absorption cross-section, any change in
the standard), as well as the uncertainty of the interference ratio the energy spectrum will affect the amount of activity produced. The
determined from the counting uncertainty of the individual fission magnitude of the effect is greatest for facilities with poorly
products in the uranium standard. It should be noted that the thermalized neutron spectra, and samples/standards with a large
interferences listed in Table 2.5 are for bulk samples, i.e., a mg or more amount of a strong neutron absorber with a sharp low resonance
total mass. Samples consisting of thin films, surface deposits, or energy. Correction factors for epithermal neutron resonance self-
having very small total masses may appear to have lower levels of shielding have been calculated using Monte Carlo methods [29].
interferences due recoil losses of fission fragments from surfaces, Where very highly absorbing samples with significant resonance
especially if samples are transferred after irradiation. absorption are analyzed, these effects should be determined exper-
imentally through irradiation of successively diluted portions and
2.4.2.3. Multiple neutron capture interferences determination of the correction factors needed to account for
Interferences due to double neutron capture are very rare, since high differences between the samples and comparator standards with
cross sections are usually necessary for both the target and daughter respect to effective neutron exposures.
isotopes. The classic example is the determination of Pt via the 198Pt
(n, γ) 199Pt, which decays to 199Au, and is used to quantify Pt. Gold 2.4.4. Irradiation losses and gains
interferes with this determination via the double neutron capture of
197
Au [25]. This interference is highly significant, since both 197Au and There are several mechanisms that have the potential to produce
198
Au have very high (n,γ) cross sections, and Au and Pt are often found losses and or gains during the irradiation process. Although in the vast
at similar mass fractions in many materials [26]. Double neutron capture majority of cases these effects are very small, it is important to
by 45Sc can interfere with the determination of Ca via the reaction 46Ca consider them since they can be significant in a few cases. These
(n,γ) 47Ca, which decays to 47Sc. This interference may be important in include: transfer of material to or from the sample through the
geological type materials mainly due to the extremely low isotope irradiation container; hot atom transfer to or from the sample
abundance and cross section of 46Ca and the high (8 b) cross section of container itself; target isotope burn up differences when samples
46
Sc. The measurement of 129I/127I ratios in nature by NAA via 130I is and standards are irradiated under different conditions, sample mass
limited by triple neutron capture in stable 127I [27]. change during irradiation [30,31], and loss of sample material upon
When double capture interferences are potentially significant, they transfer of sample out of the irradiation container. Such gains and
must be measured and corrected with the uncertainty propagated. Since losses are typically extremely small, and thus uncertainties are
the double-capture interference rate depends on (ϕ•ti)2, while the typically negligible. However, in the rare cases where gains or losses
activation rate of the element of interest will increase as (ϕ•ti), the are significant, corrections should be applied and uncertainties
relative interference will increase as a function of irradiation time. For the evaluated.
197
Au interference on the determination of Pt via 199Au, the authors [25]
determined the apparent Au interference as a function of irradiation time 2.4.4.1. Hot atom transfer to and from the surface of a sample during
in each facility used, determined the Au content of each sample, and then irradiation as a result of recoil
subtracted the interference. The uncertainty for the interference correc- Upon activation, nearly all nuclides emit gamma or other radiation
tion included the uncertainty in the interference ratio combined in and recoil with enough energy to break the chemical bonds between the
quadrature with the uncertainty of the Au determination for each sample. atoms within the molecules that comprise the sample or standard
(Szilard–Chalmers effect). There is often enough recoil energy to
2.4.3. Effective cross section differences between samples and standards transfer some of the activated nuclides from the surface of a sample to
the walls of the material used to encapsulate the sample. This will result
A central tenet of activation analysis is that in dilute samples the in losses of a portion of the activated material if the sample is transferred
probability of neutron capture σ(E) per target atom is a constant for a out of the irradiation container. However, since the maximum recoil
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
distance is typically much less than about 10 nm, the amount of material
lost through hot atom transfer is rarely significant for bulk samples
unless the element of interest is predominantly at or very near the
surface of the sample or the sample encapsulation material. For most
samples, this effect is negligible. For example, if a 1 mm thick
homogeneous sample is irradiated, 0.0005% of a nuclide recoiling
10 nm will be lost. A similar loss would be expected from a 1 mm thick
standard. Although maximum recoil energy can be obtained for each
nuclear reaction, and the corresponding distance in various matrices
calculated [32], this potential loss of material can be considered
negligible (with a negligible uncertainty) for bulk samples. However,
if thin film samples are analyzed, potential losses should be calculated,
or samples should not be transferred from the irradiation container
before counting, eliminating the possibility of losses.
Hot atom transfer can also result in unexpected transfers (gains)
from the surface of the container (blank) to the sample or standard if
an element is concentrated at or near the surface of the encapsulation
material, resulting in a higher than expected blank if the sample is
transferred. This has been frequently observed for the determination
of low-level Cr in biological samples encapsulated in polyethylene
bags. Polyethylene films used to prepare bags are often left with a
surface deposit of Cr during the manufacturing process that is difficult
to remove by acid washes. Up to about 20% of the total (activated) Cr
in a polyethylene bag may be transferred to the sample during
irradiation. Of course this will depend on the distribution of Cr in the
polyethylene, as well as the amount of surface area of the bag that is in
direct contact with the sample. It is thus important to assess the
amount of each element to be determined in the encapsulation
material compared to the sample, and to assess the amount of
potential transfer via recoil. Alternatively, acid-washed, high-purity
quartz can be used to encapsulate samples with very low mass
fractions of Cr and other common elements, where recoil transfer of
surface deposits is extremely unlikely.
One way to assess the amount surface transfer (blank) from
polyethylene or another plastic is to prepare samples of a test material
with very low, but known, mass fractions of the element(s) of interest.
These samples can then be encapsulated in the irradiation containers
to be tested, irradiated, counted after sample transfer, and mass
fractions of the elements of interest determined. The transfer blank is
just the difference between the observed elemental content compared
to the known content. Uncertainties can be calculated by combining
the uncertainty of the known value and the measurement uncertainty
for the test material.
2.4.4.2. Transfer of material through irradiation container
Upon activation, volatile Hg species can be formed as a result of
recoil, and can pass through polyethylene under certain conditions
that may occur during sample irradiation. This can result in a loss or a
gain of activated mercury in each sample when only polyethylene
encapsulation is used for irradiation. It is thus a good practice to
encapsulate samples in quartz when Hg is to be determined, thus
eliminating the possibility of gains or losses. Smaller losses have been
observed for selenium from standards prepared using selenium metal
in dilute nitric acid dried on filter paper, but the selenium is retained
by the polyethylene film packaging. When Se-containing standards
are transferred from the inner-most irradiation container, it is a good
practice to count the empty container to check for Se loss. Corrections
with uncertainties should be applied if appropriate.
In a similar manner, activated forms of mercury and the halogens
are sometimes found in the vapor phase of an encapsulated sample,
and may be lost upon transfer. This is a rare occurrence, and to the
best of our knowledge, has only been observed when reducing liquids
(e.g., urine, oil, etc.) have been irradiated. If reducing liquids are
analyzed for Hg or halogens, it is probably best to leave the sample in
the original encapsulation, if possible, and count the samples in
several different orientations to check for non-homogeneous spatial
217
distribution of these elements within the container [33]. If samples
must be transferred, e.g., to perform a radiochemical separation for
Hg, extreme care must be taken to prevent losses. Freezing in liquid
nitrogen prior to opening the sample, followed by a very rapid
transfer, has been effectively used for this purpose.
Mass loss may also be observed due to radiolysis or exposure to
heat in the irradiation position generated by gamma rays from the
reactor and the sample itself. This is often the case for materials that
have appreciable moisture content providing an added complication
to the determination of the mass basis. Iyengar [30] and Kiem et al.
[31] have studied weight losses in blood samples during irradiation as
a function of irradiation and other experimental conditions and
discussed the effects of weight loss on analytical uncertainty.
Corrections for this type of mass loss may be determined experimen-
tally by determination of the differences between individual or
combined masses of the containers and samples before and after
irradiation. The standard uncertainty is then calculated from
measurement replication, i.e., s/√n.
2.4.4.3. Sample loss during transfer from irradiation container
This loss mechanism corresponds to the bulk material left behind
when a sample is transferred out of an irradiation container such as a
polyethylene bag. Powders, even when pressed into pellets, often
leave a small surface film behind upon transfer. This is rarely
significant for bulk samples when no visible residue is left in the
sample container, however, it is important to visually inspect each
container for residue if the samples are transferred. If residue is
visible, it is important to assess whether a significant amount has
been left behind. The simplest way to do this is to record the mass
of the irradiated sample in the container before and after transfer
and to record the mass of each empty container prior to irradiation
and reweigh after sample transfer post irradiation. Note that although
the mass of the irradiation container is not likely to change during
irradiation, one must record the mass of the sample in the irra-
diation container after irradiation because the mass of the sample
may change during the irradiation. This procedure is necessary to
determine the proportion of mass loss that is attributable to
incomplete transfer in addition to that attributable to loss during
irradiation.
2.4.4.4. Target isotope burn up differences (when samples and standards
are irradiated under different conditions)
Unlike many analytical methods, consumption of the analyte
during the INAA process is entirely negligible, although for the most
exacting work it should be considered when standards and samples
are irradiated for different times or in different facilities [34]. The
extreme case is the determination of gadolinium, which has the
largest neutron capture cross section σ of any element, 47,700 b; the
next, samarium, is nearly an order of magnitude less. If a sample is
irradiated for ti = 12 h in a neutron fluence ϕ = 1013 cm− 2s− 1, the
fraction of Gd transmuted is σϕti = 2.1%. However, Gd is determined
through 153Gd, produced from 152Gd with a cross section of 811 b, so
the error in INAA even under these extreme irradiation conditions is
0.03%. The uncertainty of this correction may be estimated as a
significant fraction of the correction, such as 10%. However, when
standards and samples are irradiated for the same time in same
facility, the fraction of each isotope lost in the standard(s) will be
identical to that for the sample(s), and so the activity ratio will be not
be affected.
2.4.5. Irradiation timing and decay corrections
When samples and standards are irradiated separately, the actual
start times of irradiation will have uncertainties, as will the lengths of
irradiation. Although these are rarely significant, they should be
considered for the irradiation systems are used.
218 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
2.5. Uncertainty evaluation: gamma-ray spectrometry
2.5.1. Measurement replication or counting statistics (depending on
number of replicates) for unknown samples
Uncertainties for measurement replication are common to all
analytical measurement methods. Typically uncertainties need to be
considered separately for samples and standards. The most common
way to evaluate the standard uncertainty for measurement replica-
tion of the unknown samples is to divide the observed standard
deviation by the square root of the number of independent samples
measured. However, such an evaluation is not possible if only one
sample is analyzed. In addition, if only a few samples are analyzed,
the observed standard deviation is not very meaningful (i.e., there is
a high degree of uncertainty for the standard deviation). In these
cases, it is often more appropriate to compute the uncertainty for
measurement replication using the uncertainties from counting
statistics, since radioactive decay follows a Poisson distribution. Thus
the replication uncertainty of a single measurement would just be
the square root of the net peak area combined with twice the
background area under the peak if equal numbers of peak and
background channels are used [35].
2.5.2. Measurement replication or counting statistics (depending on
number of replicates) for standards
Discussion for the standards is essentially the same as for the
unknown samples in the preceding section (2.5.1). The uncertainty
for measurement replication of the standards may be evaluated either
as the observed standard deviation of the mean A0 value for the
standards (typically A0 per µg of each element) when multiple
standards for are used each element, or as the propagated uncertainty
from counting. If only a limited number of standards are measured, it
is often more appropriate to use the observed uncertainty due
counting statistics. If only a limited number of standards are
measured, it is often more appropriate to use the propagated
uncertainty due to counting statistics. Note that when counting
statistics uncertainties are used for the measurement replication
uncertainty for the standards, it is extremely important to make sure
that all other standard preparation, irradiation, and counting
uncertainties are considered and fully evaluated.
2.5.3. – Corrections for radioactive decay (effects of timing and half-life
uncertainties)
Radioactive decay strictly follows A = A0e− λt, where λ = ln(2)/t1/2.
If the error in the decay time td is δ, then the error in the decay
correction fd is
δðfd Þ = e−λtd −e−λðtd + δÞ : ð2  5Þ
For example, a 1% error in either td or λ corresponds to a relative
error of 0.01λt.
Since quantification of NAA measurements depends upon
comparing the decay corrected count rates of unknown samples to
comparator standards, uncertainties due to the corrections for
radioactive decay of samples and standards during irradiation,
prior to gamma spectrometry and during the gamma spectrometry
process itself must be considered. Uncertainties due to the half-lives
of each radionuclide, as well as uncertainties in establishing the
various decay times and counting intervals (live time, clock time) all
contribute to the overall uncertainty of the measurement. Fortu-
nately, half-lives of the radionuclides used in NAA are generally so
well known that the decay corrections are rarely significant for the
total uncertainty budget. Some exceptions have been noted,
however. In a high-precision analysis of arsenic via 76As, the use of
a recently published and compiled half-life gave a serious and
readily detected bias in the analytical results. Re-determination of
that half-life [36] and of three other nuclides [37] that had been
measured by the same author gave values compatible with the
consensus of the earlier literature. More recently, a determination of
Ge in Ge–Si alloy via 77Ge was hampered by a poor half-life in the
literature, and again a new determination was necessary to achieve
satisfactory results [38].
2.5.3.1. Decay corrections from end of irradiation to start of each
measurement
Since timing information for irradiation and counting is usually
determined from different instruments, timing errors resulting from
improperly synchronized clocks are possible. However, in most
cases, the potential biases, and uncertainties for corrections due to
radioactive decay from the end of irradiation to start of measure-
ment are typically extremely small (and insignificant). If samples
and standards are irradiated together, any effects resulting from
timing errors will cancel when ratios of A0s are calculated. If half-
lives of the nuclides under investigation are relatively long
compared to possible timing errors of the different clocks, biases
will also be insignificant. Only when the half-life of the nuclide
under investigation is very short and the samples and standards are
irradiated separately is there any possibility of significant bias or
uncertainty. In such cases, the bias can be evaluated (and corrected
if necessary) by comparing the times of the different clocks. If no
difference is observed, the maximum possible difference can be
estimated, and an uncertainty calculated by calculating the effect of
the maximum possible difference on A0 (when samples and
standards are irradiated separately). A standard uncertainty can
then be assessed by assuming a uniform distribution and dividing by
the square root of 3.
2.5.3.2. Clock time uncertainty
Early multichannel analyzer systems (MCAs) often had timing
resolutions of one second. When measurements were performed to
a preset live time, the clock time could be off by nearly a full second.
This could result in minor biases in decay corrections for nuclides
with very short half-lives. However, current MCAs typically have
time resolutions of 0.01 s or better, making any potential bias
insignificant for nuclides with half-lives more than about one
second. The potential bias can be evaluated by calculating the effect
on A0 by adding an additional increment of analyzer time (i.e., an
additional 0.01 s) to the clock time of a count. That would represent
the maximum possible bias. A standard uncertainty can be evaluated
by assuming a uniform distribution and dividing by the square root
of 3.
2.5.3.3. Live time uncertainty
In a manner similar to that described above for clock time
uncertainty, live time measurements could be in error by up to one
second for early MCAs when measurements were performed to preset
clock times. However, most measurements are made using preset live
times rather than preset clock times, eliminating this error. As above,
biases are typically insignificant with current MCAs, but can still be
evaluated by calculating the effect of adding an extra unit of time
resolution.
2.5.3.4. Count-rate effects for each measurement
Depending upon the type of analyzer system used the corrections
and uncertainties described in Sections 2.5.3.4.1 and 2.5.3.4.2, or those
described in Section 2.5.3.4.3 should be considered, but not all three.
2.5.3.4.1. Correction for pulse pileup losses
Pulse pileup is the result of two or more gamma rays arriving at
the detector during a time period too short to be resolved by the
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 219

amplifier, and results in lost counts. Hardware corrections for these Table 2.6
losses can be made using: pileup rejection/live-time correction Differences in corrections for inadequacy of live time extension.

circuitry in the amplifier; a high-accuracy pulser where the pulser 28

Al (t1/2 = 2.24 min)
loss rate equals the loss rate of all counts in the spectrum; or loss-
Live time Clock time Dead time Correction (Cs) Correction (Cc) Ratio
free or zero dead time hardware to assess count losses as they occur (s) (s) (%) for single for constant of Cs/Cc
and correct the spectra by adding a number of counts for the time radionuclide dead time
that the counting system is unavailable to record gamma rays or 300 360 20 1.172 1.119 1.047
calculating a correction factor for each channel. Uncertainties for the 600 720 20 1.385 1.174 1.179
hardware corrections depend on the actual hardware used. Uncer-
tainties for corrections using a high-accuracy pulser can be
determined by combining in quadrature the relative counting Equations [41] for the correction of live-time extension losses
statistics uncertainty for the number of counts in the pulser peak using conventional counting for single radioisotope and for constant
with the relative uncertainty of the pulser count rate when dead time are given below:
radioactive source is being counted. Uncertainties for the loss-free
or zero-dead-time systems depend on the linearity of these systems, eλðtc −tL Þ −1
FltcðsÞ = Single radioisotope ð2  7Þ
which can be evaluated by collecting multiple counts of a decaying λðtc −tL Þ
source over the appropriate range of system count-rates; the effects
of the corrections on the counting statistical uncertainty are further

−λt

discussed in Section 2.5.3.4.3. 1−e L ðtc Þ
FltcðcÞ = Constant dead time ð2  8Þ
Software solutions are also possible. For example, corrections for 1−e−λtc ÞðtL Þ
pileup losses (FP) can be made via Eq. (2-6):
where: Fltc = the correction factor of the inadequacy of live time
P ðtc −tL Þ
FP = e tL
ð2  6Þ correction (the ratio of “true” count rate compared to the observed
count rate; tc = clock or real time; tL = live time and λ = the decay
constant for the nuclide under investigation. Table 2.6 shows the
where: FP = Pileup Correction, tc = Clock (or Real) Time, tL = Live differences for the two correction factors for cases where the dead time
Time, and P is an experimentally determined pileup constant [39,40]. is 20% and the count time is nominally two and four times the half
The constant P is usually evaluated by a least squares regression of life of the radionuclide under investigation, 28Al (t1/2 = 2.24 min)
count rates of a single (fixed) source at varying dead times (often in this case. There is almost a 5% difference in the correction factors
through the use of a second radioactive source at varying distances to for a live time of 300 s, and an 18% difference for a live time of 600 s.
the detector). Using the uncertainty of P from the least squares Unless the analyst has recorded separately the dead time at the start
regression, the standard uncertainty for the count rate is calculated by and/or end of the count, it can be difficult to determine which
comparing the ratios of FP's for a sample/standard pair calculated correction to apply. However, if shorter counting intervals are used,
using P, versus the ratios of FP's calculated using P + uP, where uP is the the difference between the two corrections becomes smaller, and
uncertainty in determining the pileup constant P. therefore it is less important which one is used. As a general rule of
thumb, when the clock (real) time of a count is less than half of the
2.5.3.4.2. Inadequacy of live-time extension (for conventional counting half-life of the radionuclide of interest, the difference between the two
systems) corrections becomes insignificant in all but the highest accuracy
Conventional gamma-ray spectrometry systems extend the applications. Although shorter counts will result in poorer counting
counting time to make up for time that the analogue to digital statistics, multiple short counts can provide nearly the same counting
converter (ADC) of the MCA was unavailable (system dead time). statistics uncertainty as a longer count without the problem of
This does not fully correct for lost events due to radioactive decay choosing which correction factor to use. Table 2.7 shows actual
because the probability of observing an event after the original measurement data for multiple counts of the same sample and the
counting period is lower during the extended counting period. values of the two different correction factors for 28Al. The first count
This loss can be corrected either with hardware (loss-free or zero was started when the 28Al activity was approximately the same as the
dead time systems), or through software. If hardware corrections are next most intense peak, and the sample was counted for a clock time of
used, no additional uncertainty is necessary since there is no 72 s, just over 50% of the 28Al half-life. As shown in Table 2.7, the
extension of counting time. Any residual uncertainty will be included difference for the two correction factors was only 0.2%. In addition,
in the uncertainty evaluated as described in Section 2.5.3.4.3. since the 28Al activity did not dominate, and all other significant
However, if software corrections are used, information about how sources of activity had longer half lives, the midpoint of the two
the dead time varies with time may be necessary for an accurate correction factors was the most appropriate to use. The relative
correction and uncertainty evaluation. Since this information (variation standard uncertainty for this correction would be 0.1% (the difference
of dead time as a function of time) is typically not recorded by most between the correction used and the two boundary conditions)
gamma spectrometry systems, an exact solution is not often possible. A divided by the square root of 6 (triangular distribution), or 0.4%.
complete discussion of these effects using conventional gamma Although the difference between the two correction factors became
spectrometry systems can be found in De Soete et al. [41]. Normally,
Eqs. (2-7) and (2-8) for the corrections of two boundary cases are Table 2.7
considered: the case of a single radioisotope and the case where dead Corrections for 28Al in multiple counts of one sample of SRM 1633b (fly ash) following a
time is constant over the duration of the count. The correction for a short irradiation.
single radioisotope is the maximum correction necessary for nearly all Decay Live Clock Dead Correction (Cs) Correction (Cc) Ratio
NAA measurements. The correction for constant dead time is the (min) time time time for single for constant of Cs/Cc
minimum correction necessary, and rapidly goes to unity as the half-life (s) (s) (%) radionuclide dead time
of the isotope under investigation greatly exceeds the counting interval. 19 60 72 17 1.0316 1.0297 1.0019
However, when the counting interval approaches or exceeds the half- 21 120 134 10 1.0370 1.0327 1.0041
life of the nuclide under investigation, the constant dead time correction 24 300 319 6 1.0506 1.0371 1.0130
31 900 937 4 1.1018 1.0394 1.0601
may significantly underestimate the losses.
220 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Table 2.8 where dW/W is the relative statistical uncertainty of the weighting
Validation of pulse pileup and live — time extension corrections. Data for multiple
factors. Under the reasonable assumption of Poisson statistics for the
counts of one sample of SRM 1633b (fly ash) following a short irradiation.
number N of analog-to-digital conversions into a channel the total
Count Dead Live Al (%) V (μg/g) Mn (μg/g) Na (μg/g) relative uncertainty dM/M may be rewritten as:
time time t1/2 = 2.24 min t1/2 = 3.75 min t1/2 = 2.58 h t1/2 = 14.9 h
(%) (s)
2 1=2
h i
–1 = 2
#1 17 60 14.90 ± 0.101 305 ± 7 130.7 ± 2.1 2033 ± 157 dM = M = N 1 + ðdW =W Þ ; ð2  10Þ
#2 10 120 15.03 ± 0.11 294 ± 5 134.2 ± 1.4 1930 ± 77
#3 6 300 14.97 ± 0.13 291 ± 5 132.8 ± 0.8 1979 ± 39
#4 4 900 14.48 ± 0.46 313 ± 9 133.9 ± 0.5 1994 ± 23 illustrating the increase over the basic relative uncertainty which is
#5 3 600 – – 133.6 ± 0.7 1991 ± 27 caused by the correction procedure.
Red. Chi. Sq. 0.615 2.089 0.858 0.196
The relative uncertainty dW/W of weighting factors is determined
Probability of 60.5% 9.9% 48.8% 94.1%
exceeding for every peak in the spectrum by direct comparison of simulta-
1
neously measured loss-corrected and non-corrected spectra in the
Uncertainties represent 1 sigma counting statistics.
respective regions of interest, according to the formula for the
variance of grouped data:
significantly greater for subsequent counts, since the 28Al activity
constituted a minor contribution to the total activity, the correction for
1 Mi
 
2
constant dead time was appropriate, and uncertainties for this 2
σW = ∑ Ni −W ; ð2  11Þ
ðn−1Þ n Ni
correction were insignificant.
Another advantage to the use of multiple shorter counts
(instead of one longer one) is that multiple counts provide the ∑ Mi
n
opportunity for quality assurance checks of the pileup and W= ; and ð2  12Þ
∑ Ni
inadequacy of live time extension corrections. Table 2.8 shows n

the mass fraction results for Al, V, Mn and Na based on the counts
described in Table 2.7. Additional data for Mn and Na are also listed
for a fifth count of the same sample. Both Mn and Na have half-
lives that are many times longer than any of the counting intervals,
so the choice of correction factor for live time extension is
insignificant. Therefore, any trend in the data for Mn and Na as a
function of dead time would be due to bias in the pileup correction.
Looking at the data for these two elements in Table 2.8, it is apparent
that there is no increasing or decreasing trend. It is often instructive to
calculate the probability that a random sample of n points taken from a
pffiffiffi
normal distribution of mean xw and standard deviation sẋ˙ = n will
exceed the observed value of reduced chi-squared [42,43]. The reduced
chi squared values for Mn and Na in Table 2.8 indicate that all variation
can be explained by counting statistics alone. Because the half-lives of
the isotopes used for both Al and V are relatively short, corrections for
both pileup and live time extension can be both significant and difficult
to separate. However, since pileup affects all peaks in a spectrum to the
same degree, and has been shown not to be a significant problem for Mn
and Na in these data, it is not a significant problem for the Al and V
determinations listed in Table 2.8. Thus the lack of any trend in the data
for Al and V, combined with reduced chi squared values indicating that
all observed variations can be explained solely by counting statistics
(95% confidence), provide a great deal of confidence in the corrections
applied.
2.5.3.4.3. Uncertainties due to hardware corrections for Loss-Free10 or
Zero Dead Time10 (ZDT) systems
Commercially available systems for loss-free data acquisition [44]
may overcome the inadequacies of live-time extension and pulse
pileup calibrations. The uncertainty associated with the use of these
real-time correction modes is experimentally evaluated by compar-
ison of corrected and uncorrected spectral data.
The statistical uncertainty of the channel content M of a spectrum
measured with real-time correction of counting losses essentially
depends on the number N of analog-to-digital conversions into that
channel and has been described by Westphal [45] as:
2 2 2 larger uncertainties the additional uncertainty may gain significance.
ðdM =MÞ = ðdN =NÞ + ð1 = NÞðdW =W Þ ;
10
Certain commercial equipment and materials are identified in this work for the
purpose of adequately describing experimental procedures. Such mention does not
imply endorsement by the National Institute of Standards and Technology.
dW σ
= W; ð2  13Þ
W W
where Mi and Ni are the contents of corresponding loss-corrected and
non-corrected channels. The weighting factors for each channel are
calculated and the average weighting factor W is determined. The
standard uncertainty σ W of W over each peak region is calculated
and Eq. (2-9) is applied to the calculation of the standard uncertainty
of each peak area.
An alternate method to determine the uncertainty is implemented
in the DSPECPLUS ™ digital spectrometer where a variance spectrum
represents the statistical uncertainty of the loss correction process. In
the interpretation of Pommé [46], the variance spectrum ∑n2 is
formed by multiplication of the loss corrected channel content with
its weighting factor (Eq. (2-14), using the same symbols as above):
∑n2 = wM: ð2  14Þ
Integration of the variance spectrum in the regions of interest (peak
areas) provides the uncertainty of the peak area in the usual manner.
Table 2.9 gives an example for the uncertainties encountered with
loss-free data acquisition. A 152Eu source was counted together with a
137
Cs source placed at different distances. Peak areas N and M are
calculated from the channel contents in the spectrum, the number of
channels for the background is set equal to the number of channels in
the peak. Columns in the table show data according to Eq. (2-9). The
next to last column of Table 2.9 shows the relative increase over the
counting statistics uncertainty due to the ZDT process, and the last
column gives an estimate of the total uncertainty using the model of the
variance spectrum, which was artificially created from the example data
using Eq. (2-14) for each channel. The estimates are slightly different
because of the experimental uncertainty in W. The increases are nearly
insignificant in peaks with excellent counting statistics. Here the
uncertainty in the correction factor W is small as well. For peaks with
ð2  9Þ
For practical purposes, an upper limit can be estimated by determining
the variance of the correction factor W in the high-energy region of the
spectrum and using the applicable part of Eq. (2-9). If dW/W were as
high as 10% relative, the additional uncertainty would not exceed 0.5% of
the uncertainty of the uncorrected peak counts. Thus if dW/W is ≤ 10%,
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 221

Table 2.9
Example data for loss-free counting of a 152Eu source with a 137
Cs source placed at different distances to create different count rates.

Energy (keV) N dN/N (%) M W dW/W (%) dM/M Relative increase (%) uZDT (%)

Count rate 76 kHz

121 142,134 0.72374 1,639,975 11.551 0.60 0.72376 0.0018 0.72482
661 8,557,932 0.03520 98,827,818 11.511 2.50 0.03521 0.031 0.035204
1408 14,134 0.92784 161,797 11.48 5.74 0.92937 0.16 0.93409

Count rate 15 kHz

121 262,799 0.286775 399,800 1.523 0.57 0.286779 0.0016 0.28707
661 1,982,894 0.071777 3,011,938 1.517 1.55 0.071786 0.012 0.071784
1408 26,422 0.62465 40,199 1.521 7.57 0.62644 0.29 0.62452

the total uncertainty (including ZDT-related uncertainties) for a peak interference ratio for the 265/1221 keV lines is obtained by counting a
with a counting statistics uncertainty of 1.000% would be ≤1.005%. 182
Ta source (with the same gamma attenuation characteristics) on
the counting system used for the unknown samples. The number of
2.5.4. Corrections and uncertainties for gamma-ray interferences counts of the 1221 keV peak in each unknown sample is multiplied by
the 182Ta 265/1221 interference ratio to determine the contribution of
182
Although there are many interference-free gamma lines, there are Ta to the 265 keV peak. This interference is then subtracted from
more interferences than most NAA practitioners expect. Examples of the gross 265 keV peak to obtain the net 75Se peak area. The absolute
interferences observed in three geological materials are listed in uncertainty (in counts) for the interference contribution is obtained
Table 2.10. by multiplying the absolute uncertainty of the 1221 keV peak in each
However, rarely (if ever) do interfering nuclides have half-lives unknown sample by the uncertainty of the interference ratio, which is
similar to that of the nuclide of interest. Therefore, agreement of determined by combining in quadrature the (relative) counting
results from replicate measurements made at different decay intervals statistics of the 265 keV and 1221 keV peaks used to determine
(with no apparent trends) gives added confidence that there are no interference ratio. The absolute uncertainty of the corrected (net)
significant, unrecognized interferences, since the relative contribution 265 keV peak of 75Se is determined by combining in quadrature the
from interfering nuclides would increase or decrease with decrease absolute counting uncertainty of the 265 keV peak with the absolute
with decay time depending upon the whether the half-life of the uncertainty of the interference contribution.
interfering nuclide was greater or smaller than the nuclide of interest. This method for interference correction and uncertainty evalua-
Similarly, in cases where interference corrections are applied, tion depends upon the sample and 182Ta source having the same
agreement of results from replicate measurements made at different gamma-ray attenuation characteristics. This is especially important
decay times gives added confidence that interferences have been when the energy of the gamma ray of interest (in this case 265 keV)
accurately corrected. The authors have seen about 40 significant and the gamma ray indicating the interference (in this case 1221 keV)
interferences, listed in Table 2.11, however, none of these were is relatively far apart. Although a 182Ta source with matching
from radioisotopes with half-lives similar to the isotope under characteristics to the sample can be readily prepared by pipetting
investigation. and mixing a liquid 182Ta spike onto unirradiated sample material (of
Correcting interferences can be relatively simple when they are the same size and shape as actual samples under investigation, it is
recognized since most potentially interfering nuclides typically have often more convenient to choose an interference indicator with an
additional gamma rays that can be used for identification and energy closer to the gamma ray of interest. For example, 182Ta
correction. For example, in many geological/environmental samples, has a 222 keV gamma ray in addition to the 265 keV and
182
Ta often interferes with the determination of Se when using the 1221 keV gammas, which is close enough to the 265 keV peak
265 keV gamma ray from 75Se. The presence of 182Ta is indicated by to minimize gamma attenuation differences for most materials.
the presence of a 1221 keV gamma ray in the unknown spectrum. An Although 82Br (t1/2 = 35 h) has a 222 keV gamma ray, most measure-
ments of Se via 75Se are performed after a decay period of at least a
month (N20 half-lives of 82Br), at which time essentially all the 82Br
222 keV gamma rays are gone.
Table 2.10 In cases where an interference is possible (or expected), but the
Typical errors that would be observed in three geological certified reference materials if indicator gamma ray is not observed, an upper limit value (at 95%
interferences were not corrected. confidence level) for the indicator gamma ray in the spectrum of
Nuclide Energy Interfering Brick clay Buffalo River Sed. Fly ash interest can be calculated [47], and an upper limit on the interference
(keV) nuclide SRM 279 SRM 2704 SRM 1633 can be calculated via an interference ratio determined as above. One
46
Sc 1120.5 182
Ta 0.5% 0.4% 0.4%
possible approach would be to make no correction and use a standard
51
Cr 320.1 147
Nd 1% 0.3% 0.1% uncertainty equal to 1/2 of the upper limit value. An alternative
51
Cr 320.1 177m + g
Lu 0.5% 0.2% 0.05% approach would be to subtract 1/2 of the upper limit value from the
65 160
Zn 1115.6 Tb 3% 1% 2% potentially interfered peak. Assuming a uniform distribution, the
75 181
Se 136.0 Hf N 100%1 500% 90%
75 182 standard uncertainty would then be 1/2 of the upper limit value
Se 264.7 Ta N 100%1 150% 30%
122
Sb 564.2 76
As 5% 2% 7% divided by the square root of 3.
122 134
Sb 564.2 Cs 15% 4% 2%
131 103
Ba 496.3 Ru – 50% 40% 2.5.5. Corrections and uncertainties for counting efficiency differences
153 239
Sm 103.1 Np 1.5% 2% 3%
153 233
Sm 103.1 Pa 0.5% 2% 0.3%
153
Gd 103.2 233
Pa 60% 70% 40%
The NAA measurement equation (Eq. (2-1)) listed earlier in this
177
Lu 208.4 239
Np 3% 2% 10% section includes a factor (Rε) to account for any differences in
198
Au 411.8 152
Eu 100% 100% 100% counting efficiency between samples and standards. Although this
1
Within counting statistics, the 136 and 265 keV peaks observed for brick clay were ratio is usually very close to unity, two factors are may contribute to
due entirely to 181Hf and 182Ta interferences to 75Se. significant deviations. The first results from differences in the physical
222 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Table 2.11
Interferences observed in several common matrices.

Analyte nuclide γ-ray energy Interfering Interference energy Indicator energy Typical ratio of interference Range of relative Comment codes4
(keV) nuclide (keV) (keV) to indicator peak2 errors if uncorrected3

Geological materials
46 182
Sc 1120.5 Ta 1121.3 1221.4 1.4 0–75%
51 147
Cr 320.1 Nd 319.4 531.0 0.2 0.5–1000% M,S
51 177m + g
Cr 320.1 Lu 321.3 208.4 0.02 0.2–100% M
51 177m
Cr 320.1 Lu 319.0 208.1 0.001 0–2% M
65 160
Zn 1115.6 Tb 1115.1 879.4 0.05 1–10%
75 182
Se 264.7 Ta 264.1 222.1 0.4 10–200%
75 181
Se 136.0 Hf 136.55 133.1 0.15 20–500% D
110m 152
Ag 657.8 Eu 656.5 1408.1 0.01 0–20%
122 76
Sb 564.2 As 563.2 559.1 0.03 3–15%
122 134
Sb 564.2 Cs 563.2 795.8 0.13 5–30% S
131 103
Ba 496.3 Ru6 497.0 None – 2–100% X,S
152 152m
Eu 964.0 Eu 963.5 841.6 0.8 0–100% D
153 239
Sm 103.1 Np 103.7 106.37 0.8 0.5–5%
153 233
Sm 103.1 Pa 103.8 312.0 0.03 0.2–3% S
153 153
Sm 103.1 Gd 103.2 None – 0.1–100% X,S
153 233
Gd 103.2 Pa 103.8 312.0 0.03 50–500%
177 239
Lu 208.4 Np 209.8 106.3 0.1 1–20%
177 183
Lu 208.4 Ta 209.9 246.4 0.1 0–1%
198 152
Au 411.8 Eu 411.0 1408.1 0.4 10–100% S

Botanical materials
51 147
Cr 320.1 Nd 319.4 531.0 0.2 5–100% D
75 181
Se 136.0 Hf 136.55 133.1 0.15 10–50% D
75 182
Se 264.7 Ta 264.1 222.1 0.4 5–30%
110m 152
Ag 657.8 Eu 656.5 1408.1 0.01 10–1000%
113 160
Sn 391.7 Tb 392.5 298.6 0.03 100–1000%
122 76
Sb 564.0 As 562.8 559.1 0.01 1–10%
122 152
Sb 564.0 Eu 564.0 1408.1 0.3 0–5% S
153 239
Sm 103.1 Np 103.7 106.37 0.8 1–10%
198 152
Au 411.8 Eu 411.0 1408.1 0.4 1–10% S
203 75
Hg 279.1 Se 279.5 264.7 0.4 20–100% S

Biological tissues
46 182
Sc 1120.5 Ta 1121.3 1221.4 1.4 0–75%
64 24
Cu 511 Na 5118 1368 0.1 10–500%
64 65
Cu 511 Zn 511 1115 0.1 0–5%
122 76
Sb 564.0 As 562.8 559.1 0.01 1–10%
122 134
Sb 564.2 Cs 563.2 795.8 0.13 0–20%
124 134
Sb 602.7 Cs 604.6 795.8 1 0–300%
203 75
Hg 279.1 Se 279.5 264.7 0.4 20–100%

Additional miscellaneous interferences observed

60 82
Co 1173.2 Br9 1173.4 554.3 Varies
60 82
Co 1332.5 Br9 1330.8 554.3 Varies
64
Cu 511.0 cosmic10 511.0 1460 5
64
Cu 511.0 β+, PP11 511.0 Various Varies
80 80m
Br all Br all half-life Varies
85
Sr 513.9 β+, PP11 511.0 Various Varies
116m 41
In 1293.3 Ar 1293.6 None –
140 131
La 487.0 Ba 486.5 496.3 0.04
181 131
Hf 133.1 Ba 133.6 496.3 0.1
197 198
Hg 68.8, 80.2 Au 69.0, 80.1 411.8 0.04, 0.02
199 47
Au 158.3 Sc 159.4 None –
199 177
Au 208.2 Lu 208.4 None –
1
Nominally clean indicator peak used to identify interfering nuclide.
2
Typical ratio of interfering peak to interference indicator observed for the interfering nuclide.
3
Relative range of errors observed if interference is not corrected.
4
Comment codes:
M – Multiple peaks often present – can be difficult to integrate accurately.
D — Interfering nuclide has significantly shorter half-life than analyte nuclide. Better results obtained after additional decay.
S — Interfering nuclide has significantly longer half-life than analyte nuclide. Shorter decay intervals often provide better results.
X — Cannot be accurately subtracted by interference ratio technique; track half-life or use other correction method.
5
Includes equal intensity 181Hf peaks at 136.3 and 136.9 keV.
6
Fission product.
7
Includes equal intensity 239Np peaks at 106.1 and 106.5 keV.
8
Result of pair production events, which often occur in the detector shield. Other sources of high-energy gamma rays may similarly interfere.
9
Interferences from 82Br to 60Co are sum peaks; γ-ray ratios will vary with counting geometry.
10
Although the 1460 keV peak is from 40K, it can be used as an indicator for natural radiation.
11
Resulting from pair production (PP) events, normally in the Pb detector shield.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 223

size and shape of the samples, and can be highly significant since analytically [35]. For the most common geometry in INAA, a
radiation decreases as distance squared. In fact, ignoring these effects homogeneous plane source of thickness T (in cm), the self-absorption
is frequently a common source of significant bias for NAA measure- factor is:
ments. The second results from differences in the fraction of the
gamma rays internally absorbed by the sample and standard due 1−e−μT
Fγ = ð2  15Þ
difference in size, shape and composition. μT

2.5.5.1. Effects resulting from physical differences in size and shape of
samples versus standards
Since radiation decreases as a function of distance squared, small
differences in sample size and shape relative to the standards can cause
significant differences in the fraction of radiation reaching the detector,
and lead to biased results if ignored. Although it is possible to calculate
the solid angle differences for samples and standards to the active region
of the detector, the bulk of this effect in most cases is largely due to
differences in sample/standard thickness. A simple model based upon
average sample/standard distance to the detector works well for small
and thin samples and standards counted more than several cm from the
detector, i.e., the samples and standards are smaller than about half the
diameter of the detector, and are less than about 0.5 cm thick. Using this
model, the average distance to the active region of the detector is the
sum of half the thickness of the sample or standard, the distance from
the sample/standard to the surface of the detector surface, and an
estimate of the distance from the detector surface to the average
distance below the surface that the interaction between gamma ray and
the crystal occurs. A value of 3 cm works well for the detector surface to
interaction distance for most modern coaxial Ge(HP) detectors with
relative efficiencies of about 25% to 100% of a 3 in.× 3 in. Na(Tl)I
detector. A value of 2 cm may be more appropriate for detectors with
relative efficiencies less than 25%, and a value of 4 cm may be more
appropriate for detectors with relative efficiencies greater than 100%.
However, at typical sample-to-detector counting distances for activated
samples, the value chosen has only a small effect on the calculation for
the efficiency correction, which is obtained by squaring the ratio of
average distances to the detector for sample to standard. As an example,
assume that samples are 0.3 cm thick, standards are 0.2 cm thick,
gamma-ray spectrometry measurements are made 10 cm from the
surface of the detector, and the average interaction of the gamma ray
occurs 3 cm below the surface of the detector endcap. Then the average
distance for the samples is 10 + (0.3/2) + 3 cm or 13.15 cm, and the
average distance for the standards is 10 + (0.2/2) + 3 cm or 13.10 cm.
The sample count rates need to be multiplied by (13.15/13.10)2, or
1.00765 to account for the thickness differences. Estimating a standard
uncertainty of 1 cm (33%) for the interaction depth, the correction
would become (12.15/12.10)2 or 1.00828. The relative uncertainty for
the efficiency correction would be the difference between the two
corrections, or 0.06%.
2.5.5.2. Corrections for gamma ray self absorption
A major advantage of INAA is that both the probing (neutron) and
indicating (gamma) radiations are highly penetrating. Neutron self-
shielding, generally the larger term, was discussed in Section 2.4.1.3.
Most gamma rays employed in INAA have energies on the order of
1 MeV. At this energy, the transmission through 1 mm of wood, bone,
and steel is 99.6%, 99.3%, and 96%, respectively.
Gamma-ray absorption is strictly exponential with thickness. For a
pencil beam through a material of density ρ (g/cm3) and thickness T


(cm), the attenuation factor Iout = Iin = exp − μρ ⋅ρT , where the mass
attenuation coefficient μ/ρ (dimensionally cm2/g) depends on the
composition. Values of μ/ρ for many elements and composite
materials have been tabulated [48], but are most conveniently
calculated by the widely available XCOM program [49]. To calculate
the attenuation for a finite source, integration over the sample volume
is necessary. While in the general case a Monte Carlo calculation may
be necessary, many geometries of interest have been solved
where the density ρ is incorporated into the linear attenuation
coefficient μ. The standard uncertainty may be difficult to assess
quantitatively, since μ/ρ depends on the composition which, for the
sample is not fully known. However, the major elements can usually
be estimated within reasonable limits, and minor elements have a
relatively small effect on self-absorption. In addition, varying the
sample size will give empirical information about the importance of
this effect. Since the physics of XCOM is so complete, and the
program is so widely accepted, the uncertainty in composition
usually dominates. Therefore it is usually reasonable to estimate a
relative standard uncertainty of 5% to 10% of the correction
depending upon how well the major constituents of the sample
are known.
Because gamma radiation is highly penetrating, NAA practitioners
often do not consider the effects of gamma ray absorption. However,
significant bias can be introduced in many cases by ignoring this effect,
especially when samples and standards have significantly different
densities, thicknesses, and/or contents of higher-Z elements (i.e., mass
fractions greater than about several g/kg). Thus it is important to be
familiar with methods to calculate gamma-ray self-shielding factors.
One method is to use the NIST web resource for XCOM at http://www.
nist.gov/physlab/data/xcom/index.cfm (updated November 2010, last
accessed December 17, 2010). First, use the “Database Search Form” to
calculate the gamma-ray mass attenuation coefficients for each gamma
ray of interest for the material under investigation. Elements,
compounds or mixtures can be specified in the search form. Total
mass attenuation coefficients without coherent scattering should be
used for most matrices; however, total mass attenuation coefficients
with coherent scattering (that is, including diffraction) may be more
appropriate for single crystals.
The next step is to calculate the gamma-ray attenuation factor for
each energy of interest in the samples and standards using Eq. (2-15).
The mass attenuation coefficients need to be multiplied by the sample
or standard density to determine each attenuation coefficient (μ).
Examples of the calculated gamma-ray self-absorption correction
factors for 2 mm thick pellets of the above matrices (assuming
densities of 1) are given in Table 2.12. Corrections are applied by
dividing the observed gamma-ray intensity by the calculated
correction factor.
Note that there is very little difference among the correction
factors for these common matrices as long as the thicknesses and
densities are the same; thus the ratio between samples and standards
would be very close to unity. A small admixture of heavy elements, for
example 0.1% Pb, makes little difference. However, when there are
differences in either density or thickness, significant bias may occur.
Table 2.12
Calculated gamma-ray self-absorption correction factors for 2 mm thick pellets (with
densities of 1) for some typical matrices and energies.
Energy Matrix
(keV)
SiO2 Cellulose Cellulose (99.9%) Al2O3 Na2SiO3
(CH2O) with 0.1% Pb
100 0.9844 0.9844 0.9839 0.9849 0.9849
300 0.9895 0.9888 0.9887 0.9897 0.9896
500 0.9914 0.9908 0.9908 0.9915 0.9915
1000 0.9937 0.9933 0.9933 0.9938 0.9938
1500 0.9948 0.9945 0.9945 0.9949 0.9949
224 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Table 2.13
Comparison of calculated self-absorption corrections for 2 mm thick pellets of CH2O
with density 1 compared to 2 mm thick pellets of SiO2 with density 2.
Energy Matrix Matrix Correction ratio for Potential bias
(keV) CH2O SiO2 CH2O w/density = 1 Resulting from
Density = 1 Density = 2 versus SiO2 density differences
w/density = 2
100 0.9844 0.9691 1.0158 1.6%
300 0.9888 0.9791 1.0099 1.0%
500 0.9908 0.9828 1.0081 0.8%
1000 0.9933 0.9874 1.0060 0.6%
1500 0.9945 0.9897 1.0048 0.5%
For example, Table 2.13 shows the differences in self-absorption
correction factors and the potential biases for 2 mm thick samples of
SiO2 compared to dilute standard solutions pipetted onto cellulose
(i.e., CH2O).
2.5.6. Potential bias and uncertainty due to peak integration method
The method used to integrate the areas of gamma ray peaks can lead
to measurement bias if peak size, shape and background regions are
significantly different for samples and standards. Although this bias is
typically negligible for large, well-resolved peaks, it is important to look
for potential integration biases for multiplets, small peaks, and peaks
overlying complex background regions. Often the peaks for the
comparator standard are well defined, without complicated background
and without nearby interferences that may affect peak shape, whereas
the peaks for samples may be much more difficult to integrate.
One method that may be used to estimate the uncertainty associated
with the determination of the peak areas involves comparing different
peak integration methods for singlets including: peak search with
channel-by-channel summation; standard peak fitting using automatic
routines; interactive peak fitting in which peak boundaries and numbers
of channels used for background subtraction are selected manually. It is
often useful to vary the number of peak and background channels to
observe potential effects. If significant differences are observed for the
ratio of samples to standards, the most appropriate integration method
should be selected for use, and a standard uncertainty estimated from
the distance between sample/standard ratio of the selected method to
that of the other method(s). In cases where one method is obviously
more appropriate than the others, it may be appropriate to assume a
triangular distribution and divide this distance by √6 to determine the
standard uncertainty. In other cases, a rectangular/uniform distribution
may be more appropriate, and the distance from the selected method to
other methods divided by √3.
Evaluating the uncertainty for peak integration methods for
multiplets may be more difficult since fewer methods of integration
are available. However, comparison of two or more peak fitting
routines is essential, unless the estimated uncertainty can be obtained
by other methods, e.g., evaluation of residuals. Many commercially
available routines display the residuals. Alternatively, other mathe-
matical models can be used. The different parameters used for the
fitting routines should be varied to observe the effects of small
changes. An interactive peak fitting routine is particularly useful in
that the effects of small changes in peak and background channel
selection demonstrate how the net peak area changes. Again it is
important to compare the sample/standard ratios of the different
methods to estimate the standard uncertainty.
2.5.7. Potential bias due to perturbed angular correlations (γ-ray
directional effects)
Although gamma rays from irradiated samples are emitted
isotropically (in the absence of strong magnetic fields), the directions
of two gamma rays emitted in rapid sequence in the decay of a single
nucleus may be correlated. This perturbed angular correlation is
exploited in materials science to study the microscopic environment
of certain tracer nuclides, of which only a few are encountered in
NAA [50]. Because angular correlation is measurable with good
efficiency only by gamma-gamma coincidence with two or more
detectors, this effect is not experienced in the usual one-detector
configuration used in INAA unless a cascade sum peak were to be
used for analysis, or if loss of counts via summing is significant. In
practice, this is the case only when the counting efficiency is very
high, for example within a well detector. Furthermore, in comparator
NAA the effect cancels in the ratio of sample to standard if both are
counted using the same geometry, with the caveat that a subtle
dependence on chemical form has been observed in a few rare cases
[51, 52].
2.6. Uncertainty evaluation: radiochemical stage
(only if radiochemical procedures are used)
2.6.1. Losses during chemical separation
Loss of analyte can present a significant bias for RNAA measure-
ments, especially if samples and standards are not processed in an
identical manner. To avoid such problems, carriers or radiotracers are
typically used to evaluate the chemical yield [41] for each element
determined in every sample, unless the separation is known to be
quantitative, and both samples and standards are processed in an
identical manner (standards are dissolved with unirradiated sample
material). Based on the yield determination, corrections can be
applied to all samples and standards. Uncertainties are derived from
the determination of the chemical yield, and will depend on the
method used. For example, when radiotracers are used, uncertainties
will be largely due to a combination of counting statistics (before and
after separation), and the pipetting precision. When gravimetric
measurement of carrier is used for yield determination, the
uncertainty of the chemical yield will depend largely on mass
determination of the carrier before and after separation, and purity
of the separated carrier, i.e., moisture content, stoichiometry and
impurities.
2.6.2. Losses before equilibration with carrier or tracer
Losses prior to equilibration with carrier and/or tracer are often
not detected by yield determinations and can lead to measurement
bias. Similar problems are possible for IDMS measurements if losses
occur prior to equilibration of sample and isotopic spikes. The most
common types of these losses include incomplete sample decompo-
sition (leaving residual undissolved material), and volatility of specific
chemical species found in the unknown samples. Losses due to
undissolved material can be evaluated by filtering representative
sample solutions and counting the filters. Volatility losses are often
more difficult to detect; however, a closed-vessel dissolution is useful
when analyzing new matrices, or determining analytes that have
potentially volatile species.
2.7. Methods to evaluate complete uncertainty budgets for NAA
2.7.1. Manual approach general considerations
Once the standard uncertainties for each uncertainty component
have been evaluated, a combined uncertainty for the can be
determined manually from the measurement equation (Eq. (2-1)).
 
A0ðunkÞ
mxðunkÞ = mxðstdÞ   Rθ Rϕ Rσ Rε −blank: ð2  1Þ
A0ðstdÞ
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Note that mx(unk) and mx(std) are the masses of the element of
interest in sample and standard. If mass fractions are desired, this
equation becomes:
wxðstdÞ mstd A0ðunkÞ
! first order for the linear axial fluence rate gradient along the rabbit.
blank
wxðunkÞ = Rθ Rϕ Rσ Rε − ð2  16Þ
munk A0ðstdÞ munk
where wx,unk and wx,std are mass fractions of element x in the unknown
and standard, munk and mstd are the (total) masses of sample and standard,
and blank/munk is the relative blank.
Since A0 is determined from a complex equation, the uncertainty
of the A0 ratio of unknowns and standards should first be evaluated
from the standard uncertainties of each term in the A0 equation
(Eq. (2-2)):
λCeλtd
A0 = fP f :
ð1−eλtL Þð1−eλti Þ ltc
The most straightforward method to combine the uncertainties for
the terms of the A0 equation (Eq. (2-2)) is to increase or decrease each
parameter (individually) by its standard uncertainty and evaluate the
relative effect on the sample/standard A0 ratio. In many cases it is
appropriate to use the average value of the samples compared to the
standards. These relative effects are then combined in quadrature to
establish the relative A0 ratio uncertainty.
The relative uncertainty for the A0 ratio is then combined in
quadrature with the relative uncertainties for the other terms (except
for the blank) in the measurement equation (Eq. (2-1)), and
converted to an absolute uncertainty. This absolute uncertainty
(calculated from the uncertainties of all terms except blank) is then
combined in quadrature with the absolute uncertainty of the blank to
establish the combined standard uncertainty.
Alternatively, a term of (1 − blank/munk) can be substituted for the
relative blank subtraction in the mass fraction equation (Eq. (2-16)),
and a standard uncertainty evaluated for (1 − blank/munk). This makes
all operations multiplicative, allowing a combination in quadrature of
all the relative standard uncertainties in this form of the measurement
equation (Eq. (2-17)).
! As2O3 solution was conservatively assumed to be equal to that of
wxðstdÞ mstd A0ðunkÞ blank
 
wxðunkÞ = Rθ Rϕ Rσ Rε 1− ð2  17Þ
munk A0ðstdÞ munk
2.7.2. Manual approach — example of a calculation of a complete
uncertainty budget
2.7.2.1. Measurement description
Arsenic in Silicon – Standard Reference Material (SRM) 2134
(Arsenic Implant in Silicon) was produced at the National Institute of
Standards and Technology (NIST) to serve as a calibrant for secondary
ion mass spectrometry, and its arsenic content was certified using
INAA as a primary method of measurement [53]. Ten samples of SRM
2134 and two blank chips of silicon were heat-sealed into envelopes
of conventional polyethylene that had been cleaned in high-purity
HNO3. A dilute solution standard was prepared by gravimetrically
diluting a 10-mg/g solution of As2O3 (stated purity 99.999%) in dilute
ammonia to about 2 mg/kg. Another solution was prepared similarly
from SRM 3103a Arsenic Spectrometric Solution. Aliquants of both
solutions were weighed from a polyethylene pipette onto 1-cm disks
of acid-leached filter paper and dried to produce standards with about
200 ng As each. Two standards from each solution were irradiated
with the samples. A blank filter paper was shown in a separate
experiment to contain no significant arsenic (≪0.01% of the amount
in the standards).
Samples, standards, and blanks were stacked in the irradiation
container (rabbit) in a quasi-random sequence: standard, blank, five
samples, two standards, five samples, blank, and standard. The stack was
225
irradiated for four hours in the RT1 pneumatic tube of the NIST research
reactor (NBSR) at a thermal fluence rate of 7.7 × 1013 cm− 2s− 1,
inverting the rabbit halfway through the irradiation to equalize to the
After irradiation the gamma-ray activity was assayed at 10 cm
from a Ge detector with a 1.71 keV resolution and 40% relative
efficiency at 1333 keV, with the ion-implanted surface toward the
detector. Pulse pileup rejection/live time correction was performed by
hardware (in the amplifier). Spectra were acquired with a conversion
gain of 0.4 keV/channel. The order of sample counting was random-
ized for the first count, and the sequence reversed for the second
count in order to equalize precision from counting statistics. On
average more than 200,000 counts were collected in the 76As peak in
each sample and standard spectra, with a mean Poisson relative
standard deviation of 0.25%. A few samples were re-counted as long as
ð2  2Þ
six days after irradiation to check the half-life and search for long-
lived radioactive impurities. The combined 559 + 563 keV doublet of
76
As in each spectrum was integrated using a channel-by-channel
summation routine.
2.7.2.2. Pre-irradiation (sample preparation)
2.7.2.2.1. Elemental content of standards (comparators) [3]
Two solution standards were used: SRM 3103a and one prepared
gravimetrically from As2O3 (stated purity 99.999%). The certified
(expanded) uncertainty for SRM 3103a was 0.3%, which approx-
imates a 95% confidence interval. Therefore a 1 s relative standard
uncertainty was essentially 0.15%. An uncertainty for the second
standard was difficult to evaluate directly. If an uncertainty were
calculated from the uncertainties in gravimetric preparation com-
bined with the stated purity, the resulting value would be a few
hundredths of a percent. However, stated purities often do not
consider all contaminants, or non-stoichiometric behavior of
compounds. Since no differences were observed (over the counting
statistics) between the decay-corrected count rates (per unit mass of
As) for the standards prepared from the As2O3 solution, and those
prepared from SRM 3103a, the relative standard uncertainty the
SRM 3103a, or 0.15%. The combined relative standard uncertainty for
concentrations of the two standard solutions (1 s) was approximat-
ed as 0.15%/√2, or 0.106%. The standard solutions were deposited
gravimetrically onto filter papers. Since ≈ 100 mg of solution was
transferred, and readings were made to 0.1 mg, the uncertainty per
weighing is 0.1/√3 (rectangular distribution) or 0.058%. Two mass
determinations were required for each standard prepared (i.e.,
before and after deposition), and four individual filters standards
were used in this analysis. Therefore the overall relative standard
uncertainty for transfer of standard solutions is 0.058% · √2/√4 or
0.041%. The uncertainty for the standards concentration was
combined in quadrature with that of the uncertainty for standards
deposition to obtain standard uncertainty of the As content of the
comparator standards of 0.114%.
2.7.2.2.2. Target isotope abundance ratio (unknowns/standards) — since
As is monoisotopic the standard uncertainty was 0%.
2.7.2.2.3. Basis mass (in this case sample area)
All samples were cut with nearly identical sizes from the original
wafer. The micrometer used to measure the length and width of the
samples had a resolution of 0.001 mm. This micrometer was checked
with a calibrated gauge, and found to be accurate to within 0.001 mm.
Assuming a uniform (rectangular) distribution, the 0.001 mm
absolute accuracy was divided by √3 to obtain the standard
uncertainty, or 0.00058 mm. Therefore the relative standard uncer-
tainty for the accuracy basis for computing sample area (≈1 cm2) was
equal to [(1.000058 cm/1 cm)2 − 1]/100, or 0.012%. An additional
uncertainty resulted from the reproducibility of measuring the
226 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
dimensions of the wafers. Three replicate measurements were made
of the area of nine of the wafers, yielding an average relative standard
deviation of 0.046% for each measurement. Dividing by the square
root of 27 yielded a measurement replication uncertainty of 0.009%,
which was combined in quadrature with the micrometer accuracy
uncertainty of 0.012% to obtain the total uncertainty for sample area of
0.015%.
2.7.2.2.4. Sample blanks
No arsenic was seen in the blanks, with a detection limit of 50
counts. The samples each had approximately 200,000 counts in the
559 keV peak of 76As, and so the blank detection limit (at 95%
confidence) was 0.025% of the As content of the samples. The blank
relative standard uncertainty was estimated as half of the detection
limit (2σ), or 0.013%.
2.7.2.3. Irradiation stage
2.7.2.3.1. Neutron fluence exposure differences (ratios) for unknown
samples compared to comparator standards (4 individual components)
2.7.2.3.1.1. Neutron fluence exposure differences (ratios) for unknowns
compared to standards due to physical effects (fluence gradients)
This uncertainty was approximated from previous measurements
in the same irradiation facility from the previously observed relative
standard deviations (RSDs) observed for samples (of the same
material) irradiated within a single irradiated rabbit. Since observed
RSDs also include effects of other sources of variation, such as
counting statistics and material heterogeneity, the previous measure-
ment with the smallest observed RSD among a relatively large
number of samples (to minimize fortuitous agreement) was selected
for evaluation. Previously, the smallest variation observed for samples
irradiated in RT-1 was 0.2% (1 s relative) for Br among 12 aliquants of
a coal SRM that entirely filled a rabbit. For the current (As)
measurement, approximately one third of the length of the rabbit
was filled and so we divide the previously observed variation value by
3. The overall uncertainty in the mean value due to geometry
differences for the 10 samples for the As in silicon measurement was
therefore 0.2%/3/√10 or 0.021% relative. The overall uncertainty for
the mean of the 4 standards was 0.2%/3/√4 or 0.033%, and the
combined relative standard uncertainty (in quadrature) for samples
and standards was 0.039%.
2.7.2.3.1.2. Neutron fluence exposure differences (ratios) for unknowns
compared to standards due to temporal effects (fluence variations with
time)
All unknown samples, standards and blanks were irradiated
together, so there are no temporal effects, and the uncertainty was 0%.
2.7.2.3.1.3. Neutron self shielding (absorption and scattering) effects –
the difference in the self shielding for each sample compared to each
standard
The self-shielding for a finite cylinder [14] was calculated via:
R⋅fslab + L⋅f∞ cyl R⋅L
ffinite cyl: ≅ ; with x ≡ ∑ ð2  18Þ
R+L R+L
where R and L are the radius and the length of the cylinder, ∑ is the
macroscopic cross section, and
x x 3
 
x
2 section of 1000 b (10− 21 cm2). Assuming that the entire resonance
fslab = 1− lnð1 = xÞ− −γ − ð2  19Þ
2 2 2 6
4x x2 x2 5
 
f∞ cyl = 1− + lnð2 = xÞ + −γ ð2  20Þ
3 2 2 4
and γ (Euler's constant) is 0.577216. Note that if using Eq. (2-19) to
calculate the self-shielding factor for a slab, the definition of x changes
to T∑, where T is slab thickness. In a similar manner, if Eq. (2-20) is
used to calculate the self-shielding factor for an infinite cylinder, x
becomes R∑ [14].
The resulting self-shielding factor (absorption) for each sample
was 0.9985, and for each standard was 0.9986. A correction factor of
1.0001 (0.01% difference) was applied, and a standard uncertainty
equal to 10% of the correction, or 0.001% was approximated.
Scattering effects contributing to self-shielding were not significant
since the cellulose standard used was only 0.2 mm thick, and
scattering effects go to zero as thickness goes to zero. In addition,
there are no significant scattering effects from the silicon matrix of the
samples. Because this measurement employed nearly isotropic
irradiation conditions, any potential changes in activation due to
scattering within the sample stack were undetectable [20] 0.001%.
2.7.2.3.1.4. Neutron shielding (absorption and scattering) effects — effects
on the average sample versus the average standard due to combined
effects of neighboring samples and standards
The combined cumulative self-shielding effects for the multiple
samples and standards depended upon actual path length for each
neutron passing through samples or standards. Since the standards
were placed on the ends of the stack of samples as well as in the center
of the samples, we considered two boundary conditions: neutrons
passing parallel and perpendicular to stack of samples and standards.
The self-shielding for the two boundary conditions was calculated
using a simple exponential function for neutron transmission. In one
case, the samples see 0.033% greater neutron exposure than do the
standards, and in the other case the samples see 0.05% less. A
correction factor of 1.00009 (midpoint of two cases) applied, and the
uncertainty calculated by dividing the range from the correction to
either limiting case (0.042%) by √3, or 0.024%.
2.7.2.3.2. Irradiation interferences (3 individual components)
2.7.2.3.2.1. Fast (high energy) neutron interferences
Possible interferences are from 76Se (n,p) or 79Br (n,α). No
activation products of Se or Br were observed; Se and Br detection
limits were calculated (95% confidence level) with fast neutron
fluence rates and appropriate fast neutron cross sections. The
maximum interference from Se = 0.00069%, and the maximum
interference from Br = 2 × 10− 8%. Using half of maximum interfer-
ences based on the 95% detection limits for the (1 s) standard
uncertainties and combining in quadrature, the uncertainty for the
interference from fast neutron reactions was 0.0003%.
2.7.2.3.2.2. Fission neutron interferences
76
As cannot be made by fission reactions, therefore the standard
uncertainty was 0%.
2.7.2.3.2.3. Double neutron capture interferences
76
As cannot be made by a double neutron capture of a stable
nuclide. Double neutron capture on 74Ge produces 76Ge, which is
stable; therefore the standard uncertainty was 0%.
2.7.2.3.3. Effective cross section differences between samples and
standards
The uncertainty component and potential correction are due to
differences in the resonance integral due to removal of narrow, high-σ
resonance peaks. The largest resonance peak of 75As has a cross
cross section for arsenic has a value of 10− 21 cm2, which is a gross
overestimation, and that 14% of activation comes from resonance
neutrons, self-shielding calculation were performed in same manner
as for normal neutron self-shielding (except for zero contribution
from the matrix). In one case the samples see 0.0000013% greater
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
neutron fluence than standards, and in the other case the samples see
0.000008% less. Since this is an overestimation no correction was
applied, and the standard uncertainty was (estimated by dividing the
larger correction by √3 (uniform distribution between 0 and
0.000008) or 0.000005%.
2.7.2.3.4. Irradiation losses and gains (4 individual components)
2.7.2.3.4.1. Hot atom transfer (losses and by recoil,
nanometer movement)
Since the As was present as a near surface implant, evaluation of
the potential transfer during irradiation via recoil was necessary. Any
potential gains were evaluated as part of the blank determination. For
potential losses, the maximum possible recoil energy for the
absorption of a neutron by 75As is 380 eV. The maximum range of
As with this energy in silicon was calculated by the TRIM program [32]
to be 2.2 nm + 1.0 nm for straggling. Since the implant depth was
nominally 100 nm, both losses and uncertainty were considered to be
0%.
2.7.2.3.4.2. Transfer of material through irradiation container
As is not volatilized during neutron irradiation of a solid sample
under the irradiation conditions used, therefore the uncertainty was
0%.
2.7.2.3.4.3. Sample loss during transfer from irradiation container
Since element of interest was below the surface, there is no
possibility of loss, and the uncertainty was 0%.
2.7.2.3.4.4. Target isotope burn up differences
Samples and standards were irradiated for the same time so there
is no burn up differences and the uncertainty was 0%.
2.7.2.3.5. Irradiation timing and decay corrections
Since the samples and standards were irradiated together no decay
correction was necessary. The standard uncertainty was 0%.
2.7.2.4. Gamma-ray spectrometry stage
2.7.2.4.1. Measurement replication or counting statistics) for unknown
samples
The standard uncertainty was just the observed relative standard
deviation of the mass fractions of the 10 samples divided by √10 or
0.081%.
2.7.2.4.2. Measurement replication or counting statistics) for standards
Since only 4 individual standards were measured, the propagated
counting statistics uncertainty (1σ) for the 7 counts of the standards
was used, which was equal to 0.073%.
2.7.2.4.3. Corrections for radioactive decay for each measurement (effects
of timing and half-life uncertainties — 4 components)
2.7.2.4.3.1. Decay correction from end of irradiation to start of each
measurement
The half-life used for 76As was 1.097 day with a standard
uncertainty (1 s) of 0.0011 day [54]. The unknown samples measured
in this study decayed from 0.81 day to 2.96 days prior to start of
counting and no trend was observed for samples as function of decay
time during this interval. The standards decayed from 1.77 day to
3.66 days prior to start of count. The difference between samples and
standards for the midpoints of these counts is 0.83 day. The decay
correction for 0.83 day for a t1/2 of 1.097 day = 1.6895, and the decay
correction for 0.83 day for a t1/2 of (1.097 + 0.0011) day = 1.6886
which is equivalent to a relative difference of 0.053%. Thus the
measurement uncertainty due to the half-life uncertainty was 0.053%.
227
2.7.2.4.3.2. Clock time uncertainty
Since all measurements were made to a preset clock time, the
effect of clock inaccuracy was 0%.
2.7.2.4.3.3. Live time uncertainty
Since all measurements were made to a preset clock time,
uncertainties for live time may be significant for counts of short
duration. In this case, the multichannel analyzer system used records
the live time to the nearest centisecond. Thus each count has a live time
uncertainty of 0.005 s. Since live times for all counts were 7200 s (or
more in a few cases), the relative uncertainty for each count is 0.00007%.
The combined live time uncertainty for samples (20 counts) and
standards (7 counts) was [(0.00007%/√20)2 + (0.00007%/√7)2]1/2, or
0.00003%.
2.7.2.4.3.4. Uncertainties due to count rate effects for conventional
counting systems (2 individual components)
2.7.2.4.3.4.1. Inadequacy of live-time extension
The dead time for this measurement was controlled entirely
by 76As after the 2.62 h 31Si decayed away. The samples decayed
from 0.81 to 2.96 days prior to start of counts, corresponding to
7.5 to 27 half lives of 31Si. The single nuclide decay correction
was applied to samples and standards. This may provide a very
slight underestimation for the first few counts of unknowns, and
a very slight overestimation for the last few counts of the
unknowns. The corrections ranged from factors of 1.00025 to
1.00285. Using a conservative estimate of 10% for each
correction, and an average correction of 0.067%, the standard
uncertainty for the 20 counts of the samples was (.067%) (10%)/
√20 or 0.001%.
2.7.2.4.3.4.2. Correction for pulse pileup losses
Pileup rejection/live time correction was performed via hard-
ware, however, it became apparent through a comparison of
counts of the same samples/standards at different dead times that
a small degree of overcompensation was applied. This is not
unexpected, since various types of electronic noise can be
interpreted by the rejection circuit as gamma rays arriving at the
amplifier at the same time as real gamma rays. The overcompen-
sation for the measurement was evaluated through subsequent
counts at different dead times using a high-accuracy pulser and
counting. The samples and standards for the As measurement were
both counted over the same fractional dead time range of 0.5% to
10%, and so no bias was introduced; however some additional
measurement uncertainty needed to be considered. The pulser
count rate was observed to increase by 0.52% over the dead-time
range of 0.5% to 10%. The midpoint of the overcompensation range
was 0.26%. If we treat the overcompensation for each count of
sample or standard as a uniform (rectangular) distribution and
divide by √3 to obtain 1 s, we obtain a relative standard
uncertainty of 0.150% per count. Thus the combined uncertainty
for the mean of 20 counts of samples is 0.150%/√20 or 0.034%, and
for the mean of 7 counts of standards the uncertainty is 0.150%/√7
or 0.057%. Combined in quadrature the total relative standard
uncertainty was 0.066%.
2.7.2.4.4. Corrections for gamma-ray interferences
The only possible significant interference for the 559 keV peak
of 76As in this measurement was from the 564 keV peak of 122Sb,
which also has a 697 keV gamma ray. No 697 keV peak from
122
Sb was observed in the spectrum for the sample that was
counted for longest time. The detection limit at 95% confidence
corresponds to an interference of 0.000007% for the 76As peak.
The uncertainty was estimated as half of this 2σ detection limit,
or 0.000004%.
228 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

2.7.2.4.5. Corrections for counting efficiency differences (2 individual nents in quadrature. Based on the effective degrees of freedom, a
components) k value of 2.04 was chosen, and the expanded uncertainty was 0.38%.
Results are summarized in Table 2.14.
2.7.2.4.5.1. Effects resulting from physical differences in size and shape of
samples versus standards 2.7.3. Kragten spreadsheet approach (for one sample and one standard) –
All counts were made at a distance of 10 cm from the detector. The general considerations and example
sample thickness was 0.00002 cm (surface implant) and the standard
thickness was 0.02 cm with the As distributed uniformly. The average When a measurement equation f(x1 … xn) is at all complicated,
gamma-ray interaction with the detector used was estimated to occur calculating the propagation of uncertainties can require substantial
1 cm below the detector endcap. The correction for the (midpoint) skill in calculus to construct all the required derivatives. Kragten [56]
counting geometry difference between samples and standards was has proposed a straightforward spreadsheet method in which
(11.01/11.00001)2 or 1.00182, which was applied. This geometry standard uncertainties Δx replace derivatives dx. This numerical
estimate was tested by counting one sample in two orientations: with
the implanted face up and with the implanted face down. The Table 2.14
observed ratio of corrected counting rates was 1.014, with a 1σ Uncertainty evaluation for the determination of arsenic in silicon, SRM 2134.
uncertainty of 0.004 from counting statistics alone. The expected ratio
Uncertainty component ui (%)
from a 1/r2 model of counting efficiency, as described above, gives a
ratio 1.013, in good agreement with the observed value. If the Pre-irradiation stage
standard uncertainty on the interaction depth were estimated to be Elemental content of standards
Elemental content of standard solutions 0.106
50%, or 0.5 cm, the correction factor would become (10.51/
Amount of standard solutions used 0.041
10.500001)2 or 1.00191, 0.009% different. Thus the uncertainty on Target isotope abundances 0
geometry correction was 0.009%. Basis mass (or other basis unit) 0.015
Blank (1 — relative blank) — see Eq. (2-16) to permit 0.013
2.7.2.4.5.2. Corrections for gamma-ray self-absorption combination in quadrature

The calculated self-absorption factor for the 559 keV gamma ray
Irradiation stage
within the 0.02 cm cellulose standard is 0.99962. The applied Neutron fluence exposure differences (ratios) between
correction was therefore 0.038%. If we estimate the relative samples and standards
uncertainty as 10% of the correction it is equivalent to 0.004%. The Irradiation geometry differences (ratios) between samples 0.039
and standards
attenuation factor for the 559 keV gamma-ray through the
Neutron fluence exposure differences due to temporal 0
0.00002 cm silicon layer on top of the As in the samples (surface effects
implant) is 0.9999985. The correction is therefore ≈10− 4%, and if we Neutron self shielding differences (for individual samples 0.001
again estimate the uncertainty for the correction as 10% of the and standards)
correction it is equivalent to 10− 5% relative. Combining these two Neutron shielding differences (effects of neighboring 0.024
samples/standards)
uncertainties in quadrature, the relative standard uncertainty for
Irradiation interferences
gamma-ray self-absorption was 0.004%. Fast neutron 0.0003
Fission neutron 0
2.7.2.4.6. Potential bias due to peak integration method Double neutron capture 0
Effective neutron cross section differences between samples 0.000005
The spectra for samples and standards were nearly identical
and standards
(essentially only 76As). Peak integration was accomplished by a Irradiation losses and gains
channel-by-channel summation using the same peak and baseline Hot atom transfer (losses and gains by recoil, nanometer 0
regions (background channels) for both the samples and the movement)
standards. This summation method was essentially the same as Transfer of material through irradiation container 0
Sample loss during transfer from irradiation container 0
used previously for high-accuracy gamma-spectrometry measure-
Target isotope burn up (for samples and stds. irradiated 0
ments of uranium isotopic compositions in bulk U3O8 standards [55]. under diff. conditions)
In that study, the mean relative difference between the gamma Irradiation timing and decay corrections (if samples and 0
spectrometry measurements and the ultimately certified values for 5 standards irradiated separately)
sets of materials (at different isotopic ratios) was 0.033% (0.012% to
Post irradiation
0.090%). We conservatively estimated the total peak integration Measurement replication (or counting statistics) of samples 0.081
uncertainty for the current arsenic study as equivalent to entire Measurement replication (or counting statistics) of standards 0.073
(average) difference for the U isotopic study. The relative standard Corrections for radioactive decay (due to timing and half-life
uncertainty was calculated assuming a rectangular distribution and uncertainties for each measurement)
Decay correction to start of count 0.053
dividing by √3 to obtain a value of 0.019%.
Clock time uncertainty 0
Live time uncertainty 0.00003
2.7.2.4.7. Potential bias due to perturbed angular correlations (γ-ray Count-rate effects for each measurement
directional effects) Inadequacy of live-time extension for very short lived nuclides 0.001
76
As is not subject to perturbed angular correlation effects, and Corrections for pulse pileup 0.066
Corrections for gamma-ray interferences 0.000004
both samples and standards were counted under identical geometries, Corrections for counting efficiency differences
therefore the uncertainty was 0%. Physical differences in shape and thickness of samples and 0.009
standards
2.7.2.5. Potential losses during chemical decomposition and processing Gamma-ray self absorption differences between samples 0.004
and standards
Since radiochemistry was not performed for this measurement,
Peak integration biases 0.019
the uncertainty was 0%. Perturbed angular correlation (almost never encountered) 0
Losses during chemical decomposition and separation (for RNAA) 0
2.7.2.6. Combined standard uncertainty and expanded uncertainty Relative combined uncertainty (uc) 0.19%
The combined standard uncertainty for this measurement was k-value 2.04
Relative expanded uncertainty 0.38%
calculated by combining each of the individual uncertainty compo-
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 229

Fig. 2.4. Example of uncertainty calculation for mass fraction using the Kragten procedure.

Fig. 2.5. Calculation of A0 uncertainty for the unknown sample in the example of the Kragten procedure.

procedure is especially useful in studying the sensitivity of individual
sources of uncertainty to the final result [57].
As an example, in Fig. 2.4 the Kragten procedure has been applied to
the simplest form of the INAA equation, in which the mass fraction wx is
calculated from the fundamental measurement equation.
Fig. 2.4 illustrates the uncertainty calculation for one sample of
arsenic implant in silicon, and one comparator of an evaporated
standard arsenic solution taken from the example of the manual
calculation described in Section 2.7.2. The calculation of wx is performed
five times, once with all four variables as measured and again with each
parameter incremented by one standard uncertainty u. Examination of
the four variance contributions shows clearly that the counting
precision of the unknown sample is the dominant contributor to the
overall uncertainty, as it should in a well-designed measurement. The
combined relative standard uncertainty (shown in Fig. 2.4) is 0.35%.
Calculations of uncertainties for both the A0 value of the sample
and the A0 value of the standard are approached in a similar manner,
conveniently on separate pages of the spreadsheet. Fig. 2.5 shows the
uncertainty calculation of A0 for the unknown sample.
In this version of the A0 equation there are seven parameters, so the
calculation is performed eight times. For this example, significant
uncertainty in A0 once again comes from only counting statistics of the
sample. Note that the A0 equation above is just the general equation given
previously with FP and Fltc(s) replaced with an explicit pileup correction,
and using the live-time extension correction for a single decaying nuclide.
For all its convenience in identifying the most important sources of
uncertainty in an analytical process, the Kragten method has several
drawbacks. Compensating errors and other kinds of covariance are
ignored, and the experimental design that is implicitly modeled (one
variable changed at a time) is not the most efficient one for optimizing
a multivariate experiment. When a large numbers of parameters are
involved, or when multiple samples and standards are involved, the
spreadsheets become unwieldy. The method is further discussed in
the Eurachem quality manual [58]. A detailed uncertainty budget for
k0 NAA has been constructed by this procedure [59].
2.7.5 Partial derivatives approach (for one sample and one standard) –
general considerations
Calculation of uncertainty
The combined standard uncertainty of a quantity of interest, the
measurand y, is the effect of all variables' uncertainties on the
uncertainty of the measurement function based on them. Thus the
230 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

combined standard uncertainty of y, uc(y), is calculated in terms of tainty uc is a function of the uncertainties of the independent
component uncertainties, u(xi): parameters M.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
n ∂y
ðuðxi ÞÞ2 ;


uc ðyðx1 ; …xn Þ = ∑ ð2  21Þ uc = f wxðstdÞ ; A0ðunkÞ ; A0ðstdÞ ; Rθ ; Rϕ ; Rσ ; Rε ; B; mstd ; munk ; αunk ð2  23Þ
i=1 ∂xi

∂f
M
 2
where y(x1,x2,…xn) is a function of several parameters x1,x2,…xn. 2 2
uwxðunkÞ = ∑ uxi ð2  24Þ
Each variable's contribution is just the square of the associated i=1 ∂xi
uncertainty expressed as the standard deviation multiplied by the
!2
square of the associated partial derivative. ∂f

u2wxðunkÞ = A0ðunkÞ ; A0ðstdÞ ; Rθ ; Rϕ ; Rσ ; Rε ; B; mstd ; munk ; αunk u2wxðstdÞ
The above Eq. (2-21) is valid only if the quantities xi are ∂wxðstdÞ

independent (uncorrelated), the distribution of each xi is a Gaussian

!2
∂f
2
+ wxðstdÞ ; A0ðunkÞ ; A0ðstdÞ ; Rθ ; Rϕ ; Rσ ; Rε ; B; mstd ; munk ; αunk uA0ðunkÞ + …
distribution, and if u(xi) ≪ xi. In most cases Eq. (2-21) is used even ∂A0ðstdÞ
when the aforementioned restrictions are not met. (Deviations of ð2  25Þ
parameter distributions from the Gaussian distribution are not so
important. According to the central limit theorem of statistics, the λt
C⋅λe d
distribution for the values of the derived quantity y approaches the with A0 = , and t L = live time of count with appropriate
Gaussian distribution as the number of individual quantities xi 1−e−λtL
corrections applied.
increases, independent of the details of the distribution.)
The expanded uncertainty u is obtained by multiplying the λe−λtL
combined standard uncertainty by a suitable coverage factor kα. The Let ftL = , then
1−e−λtL
unknown value of the measurand is believed to lie in the interval y ± u
 2
with a confidence level of approximately α. uA0
u 2
C

u 2
λ 2
= + ð1 + λtd −tL ⋅ftL Þ + u2td λ2 + u2tL ft2L : ð2  26Þ
Applying this calculation to a slightly modified form of activation A0 C λ
equation (Eq. (2-22))
Quantification of uncertainty
!
A0ðunkÞ mstd Not all from the above-mentioned uncertainty components will
wxðunkÞ = wxðstdÞ ⋅ ⋅Rθ ⋅Rϕ ⋅Rσ ⋅Rε −B ⋅ ; ð2  22Þ
A0ðstdÞ αunk ⋅munk contribute significantly to the combined uncertainty. An initial
evaluation should be made to identify the relevant uncertainties in
with B = blank, and α = appropriate corrections to mass leads to accordance with the experiment that is to be evaluated, for example is
the following expressions for the uncertainty. The combined uncer- it an individual measurement or a series of similar measurements that

Table 2.15
Partial-derivative equation evaluations of relative standard uncertainties, and an evaluation of the relative combined standard uncertainty for the arsenic in silicon measurement
using one unknown sample and one comparator standard.

Parameter Source Standard uncertainty Type A/B Description ui

Gravimetry
up
wx(std) (%) Purity uwxðstdÞ = pffiffiffi B Mass fraction in primary assay standard 0.033%
2 3
u
mu (kg) Balance um = balance
pffiffiffi A Mass of sample used 0.015%
k n
ms (kg) A Mass of primary assay standard used 0.121%
udry
α Between units um = pffiffiffi A Standard uncertainty of n dry mass determinations n/a
k n
A0 calculations pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Cx,u Poisson statistics uN = N + 2B A Number of counts in indicator gamma-ray peak for sample 0.251%
Cx,s A Number of counts in indicator gamma-ray peak for standard 0.189%
δλ
λ (s− 1) Literature values uλ = pffiffiffi ð1 + λtd −tL ftL Þ B Decay constant 0.026%
2 3
δutd
td(unk) (s) Time tag utd = pffiffiffi λ B Decay time until start of measurement for sample 0.00021%
2 3
td(std) (s) B Decay time until start of measurement for standard 0.00021%
δutL
tL(unk) (s) Time tag utL = pffiffiffi ftL B Elapsed time of measurement for sample 0.0044%
2 3
tL(std) (s) B Elapsed time of measurement for standard 0.0090%
Irradiation and counting
δθ
Rθ Literature values uRθ = pffiffiffi B Ratio of isotopic abundances for sample and standard n/a
2 6
δϕ
Rϕ Absorption, flux gradient uRϕ = pffiffiffi B Ratio of neutron fluences (including fluence gradient, 0.033%
2 3 self absorption and scattering)
δσ ϕ
Rσ Neutron spectrum uRσ = pffiffiffi B Ratio of effective cross section if neutron spectrum shape 0.0004%
2 3 differs between sample and standard
δσ ε
Rε Geometry, density uRε = pffiffiffi B Ratio of counting efficiencies (including geometry, gamma-ray 0.069%
2 3 self shielding and counting effects)
Measurement blank vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u n  2
u ∑ yB;i −B
u
ti = 1
B Measured uB = B Amount of analyte in blank in g; most INAA procedures can 0.013%
ðn−1Þ
be carried out essentially blank free. An estimate from the limit
of detection is included.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
are combined into one result. Table 2.15 assumes the calculation of a
combined uncertainty from a single measurement of an unknown
sample that is compared to a single measurement of a standard. For
each component a standard uncertainty is calculated based on likely
distributions of the value.
2.7.6. Example of the partial derivatives approach
Data from a single unknown sample and a single standard from the
arsenic in silicon experiment described in Section 2.7.2 were
evaluated using the partial derivative approach and the results are
listed in Table 2.15. The combined relative standard uncertainty was
calculated by combining, in quadrature, all ui values from the last
column of Table 2.15, and was found to be 0.35%, the same as found
using the Kragten spreadsheet approach described in Section 2.7.3.
2.7.7. — Monte Carlo approach
The use of a Monte Carlo method for estimating the uncertainty of
measurement is described in Supplement 1 [60] to the GUM [61]. This
approach is especially useful if, e.g., it is difficult or inconvenient to
provide the partial derivates of the model, or if the probability density
functions of the input quantities are asymmetric. These (and other
conditions, see [59]), are not applicable with the use of comparator
NAA; therefore this approach will not been discussed in detail here.
However, the Monte Carlo method can be useful if the measurement
equation is complicated, as for the k0-NAA method, since the
uncertainty of the parameter α appears as an exponent.
2.8. References for Chapter 2
[1] ASTM, Standard Practice for Determining Neutron Fluence, Fluence Rate, and
Spectra by Radioactivation Techniques (E261). ASTM International, West
Conshohocken, PA, 1998.
[2] R.M. Lindstrom, R. Zeisler, E.A. Mackey, P.J. Liposky, R.S. Popelka-Filcoff, R.E.
Williams, Neutron irradiation in activation analysis: a new rabbit for the NBSR,
J. Radioanal. Nucl. Chem. 278 (2008) 665–669.
[3] R.R. Greenberg, Accuracy in standards preparation for neutron activation
analysis, J. Radioanal. Nucl. Chem. 179 (1994) 131–139.
[4] J.R. Moody, R.R. Greenberg, K.W. Pratt, T.C. Rains, Recommended inorganic
chemicals for calibration, Anal. Chem. 60 (1988) 1203A–1218A.
[5] K. Heydorn, E. Damsgaard, Gains or losses of ultratrace elements in
polyethylene containers, Talanta 29 (1982) 1019–1024.
[6] C. Yonezawa, M. Hoshi, Blank analysis of quartz container for instrumental
neutron activation analysis, Bunseki Kagaku 39 (1990) 25–31.
[7] M.E. Wieser, Atomic weights of the elements 2005, Pure & Appl. Chem. 78
(2006) 2051–2066.
[8] R.F. Fleming, R.M. Lindstrom, Limitations on the accuracy of sulfur
determination by NAA, Trans. Am. Nucl. Soc. 41 (1982) 223.
[9] J.R. de Laeter, J.K. Böhlke, P. De Bièvre, H. Hidaka, H.S. Peiser, K.J.R. Rosman, P.D.
P. Taylor, Atomic weights of the elements. Review 2000, Pure & Appl. Chem. 75
(2003) 683–800.
[10] J.R.D. Copley, Scattering effects within an absorbing sphere immersed in a field
of neutrons, Nucl. Instrum. Methods A307 (1991) 389–397.
[11]M. Blaauw, The derivation and proper use of Stewart's formula for thermal neutron
self-shielding in scattering media, Nucl. Sci. Eng. 124 (1996) 431–435.
[12] J.C. Stewart, P.F. Zweifel, Self-shielding and Doppler effects in the absorption of
neutrons, Prog. Nucl. Energy, Ser. 1 3 (1959) 331–359.
[13] J. Salgado, I.F. Gonçalves, E. Martinho, Development of a unique curve for
thermal neutron self-shielding factor in spherical scattering materials, Nucl.
Sci. Eng. 148 (2004) 426–428.
[14] R.F. Fleming, Neutron self-shielding factors for simple geometries, Int. J. Appl.
Rad. Isot. 33 (1982) 1263–1268; R. M. Lindstrom and R. F. Fleming, Neutron Self-
Shielding Factors for Simple Geometries, Revisited, Chem. Anal. (Warsaw) 53
(2008), 855-859.
[15] J. Gilat, Y. Gurfinkel, Self-shielding in activation analysis, Nucleonics 21 (8)
(1963) 143–144.
[16] C. Chilian, J. St-Pierre, G. Kennedy, Dependence of thermal and epithermal
neutron self-shielding on sample size and irradiation site, Nucl. Instrum.
Methods A564 (2006) 629–635.
[17] E. Martinho, I.F. Gonçalves, J. Salgado, Universal curve of epithermal neutron
resonance self-shielding factors in foils, wires and spheres, Appl. Radiat.
Isotop. 58 (2003) 371–375.
231
[18] E. Martinho, J. Salgado, I.F. Gonçalves, Universal curve of the thermal neutron
self-shielding factor in foils, wires, spheres and cylinders, J. Radioanal. Nucl.
Chem. 261 (2004) 637–643.
[19] C. Chilian, J. St-Pierre, G. Kennedy, Complete thermal and epithermal neutron
self-shielding corrections for NAA using a spreadsheet, J. Radioanal. Nucl.
Chem. 278 (2008) 745–749.
[20] R.M. Lindstrom, R.R. Greenberg, D.F.R. Mildner, E.A. Mackey, R.L. Paul, Neutron
scattering and neutron reaction rates. in: B. Smodis (Ed.), Second International k0
Users Workshop, Jozef Stefan Institute, Ljubljana, Slovenia, 1997, pp. 50–53.
[21] R. Zeisler, R.M. Lindstrom, R.R. Greenberg, Instrumental neutron activation
analysis: a valuable link in chemical metrology, J. Radioanal. Nucl. Chem. 263
(2005) 315–319.
[22] R. M. Lindstrom, R.F. Fleming, in: J.R. Vogt (Ed.), Proc. 3rd Int. Conf. Nucl.
Methods in Envir. and Energy Res. (CONF-771072), U. S. Dept. of Energy,
Washington, 1977, pp. 90–98.
[23] A. Calamand, Handbook on Nuclear Activation Cross-Sections, IAEA, Vienna,
1974, pp. 273–324.
[24] S. Landsberger, Update of uranium fission interferences in neutron activation
analysis, Chem. Geol. 77 (1989) 65–70.
[25] R. Zeisler, R.R. Greenberg, Ultratrace determination of platinum in biological
materials via neutron activation and radiochemical separation, J. Radioanal.
Chem. 75 (1982) 27–37.
[26] R. Zeisler, R.R. Greenberg, in: P. Brätter, P. Schramel (Eds), Trace element analytical
chemistry in medicine and biology, de Gruyter, Berlin, 1988, pp. 297–303.
[27] R.M. Lindstrom, G.J. Lutz, B.R. Norman, High-sensitivity determination of
iodine isotopic ratios by thermal and fast neutron activation, J. Trace
Microprobe Tech. 9 (1991) 21–32.
[28] J. St-Pierre, G. Kennedy, Effects of reactor temperature and sample mass on the
activation of biological and geological materials with a SLOWPOKE reactor,
Biol. Trace Elem. Res. 71–72 (1999) 481–487.
[29] I.F. Gonçalves, E. Martinho, J. Salgado, Monte Carlo calculation of epithermal
neutron resonance self-shielding factors in foils of different materials, Appl.
Radiat. Isotop. 56 (2002) 945–951.
[30] G.V. Iyengar, Loss of sample weight during long time reactor irradiation of
biological materials, J. Radioanal. Chem. 84 (1984) 201–206.
[31] J. Kiem, G. Rumpf, K. Beyer, G. Koslowski, Weight loss as error source in neutron
activation analysis of biological samples, Fresen. J. Anal. Chem. 328 (1987) 221–225.
[32] J.F. Ziegler, J.P. Biersack, U. Littmark, The stopping and range of ions in solids,
Pergamon, New York, 1985.
[33] R.R. Greenberg, R.F. Fleming, R. Zeisler, High sensitivity neutron activation
analysis of environmental and biological standard reference materials,
Environ. Internat. 10 (1984) 129–136.
[34] S. Pommé, A. Simonits, R. Lindstrom, F. De Corte, P. Robouch, Determination of
burnup effects in 197Au(n,γ)198Au prior to reactor neutron field characterisa-
tion, J. Radioanal. Nucl. Chem. 245 (2000) 223–227.
[35] K. Debertin, R.G. Helmer, Gamma- and X-ray spectrometry with semiconduc-
tor detectors, North-Holland, Amsterdam, 1988.
[36] R.M. Lindstrom, M. Blaauw, R.F. Fleming, The half-life of 76As, J. Radioanal.
Nucl. Chem. 257 (2003) 489–491.
[37] R.M. Lindstrom, M. Blaauw, M.P. Unterweger, The half-lives of 24Na, 43K, and
198
Au, J. Radioanal. Nucl. Chem. 263 (2005) 311–313.
[38] R.M. Lindstrom, R. Zeisler, R.R. Greenberg, Accuracy and uncertainty in radioactivity
measurement for NAA, J. Radioanal. Nucl. Chem. 271 (2007) 311–315.
[39] A. Wyttenbach, Coincidence losses in activation analysis, J. Radioanal. Chem.
8 (1971) 335–343.
[40] R.M. Lindstrom, R.F. Fleming, Dead time, pileup, and accurate gamma-ray
spectrometry, Radioact. Radiochem. 6(2) (1995) 20–27.
[41] D. De Soete, R. Gijbels, J. Hoste, Neutron activation analysis, Wiley-
Interscience, London, 1972.
[42] P.R. Bevington, Data reduction and error analysis for the physical sciences,
McGraw-Hill, New York, 1969.
[43] R.M. Lindstrom, SUM and MEAN: standard programs for activation analysis,
Biol. Trace Elem. Res. 43–45 (1994) 597–603.
[44] G.P. Westphal, Review of loss-free counting in nuclear spectroscopy,
J. Radioanal. Nucl. Chem. 275 (2008) 677–685.
[45] G.P. Westphal, On the performance of loss-free counting — a method for real-
time compensation of dead-time and pileup losses in nuclear pulse
spectroscopy, Nucl. Instrum. Methods 163 (1979) 189–196.
[46] S. Pommé, A plausible mathematical interpretation of the ‘variance’ spectra
obtained with the DSPECPLUS digital spectrometer, Nucl. Instrum. Methods
A482 (2002) 565–566.
[47] L.A. Currie, Limits for qualitative detection and quantitative determination:
application to radiochemistry, Anal. Chem. 40 (1968) 586–593.
[48] J.H. Hubbell, S.M. Seltzer, Tables of X-ray mass attenuation coefficients and
mass energy-absorption coefficients 1 keV to 20 MeV for elements Z = 1 to 92
and 48 additional substances of dosimetric interest, NISTIR 5632, Natl. Inst.
Stds. & Tech., Gaithersburg, 1995, p. 111.
[49] M.J. Berger, J.H. Hubbell, S.M. Seltzer, J. Chang, J.S. Coursey, R. Sukumar, D.S.
Zucker, XCOM: Photon cross section database. NIST, Gaithersburg, MD.,
1998, available online at http://www.nist.gov/pml/data/xcom/index.cfm.
232 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

[50] M. de Bruin, P.Bode, Application of gamma–gamma directional correlation μ/ρ mass attenuation coefficient
measurements in trace element research, Analusis 11 (1983) 49–54.
N0 Number of target atoms originally present
[51] M. de Bruin, P.J.M. Korthoven, The influence of the chemical form of
radionuclides on the shape of the gamma-ray spectrum, Radiochem. Radio- Ntrans Number of target atoms transmuted
anal. Lett. 21 (1975) 287–292. P Pileup constant for software correction
[52] K. Yoshihara, H. Kaji, T. Mitsugashira, S. Suzuki, Chemical and environ- Rθ ratio of isotopic abundances for unknown and standard
mental effects on γ-ray sum peak intensity of 111In due to perturbed Rϕ ratio of neutron fluences (including fluence drop off, self
angular correlation of cascade γ-emission, Radiochem. Radioanal. Lett. 58
(1983) 9–16.
shielding, and scattering)
[53] R.R. Greenberg, R.M. Lindstrom, D.S. Simons, Instrumental neutron activation Rσ ratio of effective cross section differences (if neutron
analysis for certification of ion-implanted arsenic in silicon, J. Radioanal. Nucl. spectrum shape changes)
Chem. 245 (2000) 57–63. Rε ratio of counting efficiencies (differences due to geometry
[54] J.F. Emery, S.A. Reynolds, E.I. Wyatt, G.I. Gleason, Half-lifes of radionuclides—
and γ-ray self-shielding)
IV, Nucl. Sci. Eng. 48 (1972) 319–323.
[55] R.R. Greenberg, B.S. Carpenter, High accuracy determination of 235U in ρ density
nondestructive assay standards by gamma-ray spectrometry, J. Radioanal. σ neutron cross section
Nucl. Chem. 111 (1987) 177–197. T thickness of sample or standard
[56] J. Kragten, Calculating standard deviations and confidence intervals with a tc clock time of count
universally applicable spreadsheet technique, Analyst 119 (1994) 2161–2166.
[57] C.E. Rees, Error propagation calculations, Geochim. Cosmochim. Acta 48 (1984).
td decay time to start of count
[58] S.L.R. Ellison, M. Rosslein, A. Williams (Eds.), Quantifying Uncertainty in ti irradiation time (duration)
Analytical Measurements (2nd ed.; QUAM:2000.P1), Eurachem/CITAC, 2000. tL live time of count
[59] P. Robouch, G. Arana, M. Eguskiza, S. Pommé, N. Etxebarria, Uncertainty t1/2 half-life
budget for k0-NAA, J. Radioanal. Nucl. Chem. 245 (2000) 195–197.
θ isotope abundance
[60] Joint Committee for Guides in Metrology, Evaluation of measurement data –
Supplement 1 to the “Guide to the expression of uncertainty in measurement” – ui standard uncertainty for uncertainty component (i)
Propagation of distributions using a Monte Carlo method (2008) JCGM 101:2008 wx mass fraction of element x in the unknown sample or
(ISO/IEC Guide 98-3-1). standard (wx(unk), wx(std))
[61] Joint Committee for Guides in Metrology, Evaluation of measurement data —
Guide to the expression of uncertainty in measurement (2008) JCGM 101:2008
(ISO/IEC Guide 98-3. Symbols used for evaluation of neutron self-shielding corrections
in the example of the determination of arsenic in silicon

2.9 List of Symbols used in Chapter 2 L length of cylinder

R radius of cylinder
A count rate ∑ macroscopic cross section
A0 decay corrected count rate T thickness of slab
C net counts in γ-ray peak γ Euler's constant
fp correction for pulse pileup (general case)
fltc correction for inadequacy of live time extension (general case) Symbols used for discussion of hardware corrections for Loss-Free
fϕ correction for neutron self shielding (general case) or Zero Dead Time systems
FP calculated correction factor for pulse pileup
Fltc calculated correction factor for inadequacy of live time extension B blank
Fγ gamma-ray attenuation factor M channel contents of a multichannel analyzer
ϕ neutron fluence rate N analog-to-digital conversions into that channel and
λ decay constant (ln 2/t1/2) W weighting factors
m total mass of the unknown sample standard (munk, mstd) dM/M statistical uncertainty of the channel with content M
mx mass of element x in the unknown sample or standard (mx(unk), dN/N statistical uncertainty for the number of analog-to-digital
mx(std)) conversions into channel with content M
μ gamma-ray attenuation coefficient dW/W relative statistical (uncertainty) of the weighting
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241
Chapter 3. Performance of neutron activation analysis laboratories in CCQM key
comparisons and pilot studies
Elisabete A. De Nadai Fernandes *, Márcio Arruda Bacchi
Centro de Energia Nuclear na Agricultura, Universidade de São Paulo
Avenida Centenário 303, 13416-000 Piracicaba, SP, Brazil
3.1. Introduction
The CCQM, after a Primary Methods Symposium in April 2000 at the
BIPM in Sèvres, France, concluded that insufficient evidence was
available on the actual performance of NAA at the highest metrological
level (as reflected by key comparisons) to accept the claim that NAA
meets all the requirements in the definition of a primary ratio method.
Both the CCQM and its Inorganic Analytical Working Group (IAWG)
questioned whether multiple laboratories could make NAA measure-
ments with the highest metrological properties, and in particular submit
results in CCQM comparisons that demonstrated agreements and
uncertainties similar to those demonstrated by IDMS; IDMS had been
designated by the CCQM in 1997 as one of the potentially primary ratio
methods for determining amounts of substance. Shortly after the
Primary Methods Symposium, the Inorganic Analytical Working Group
began organizing pilot studies and key comparisons on matrices and
measurands that were appropriate for NAA measurements, and a
number of NAA laboratories were encouraged to participate.
NAA is not available at most national metrology institutes (NMIs),
which distinguishes it somewhat from other methods for determining
amount of substance (such as IDMS or AAS). This is largely due to lack
of access to nuclear research reactors, radiological laboratories, and
trained personnel. It would be difficult drawing conclusions about
NAA's performance from results of only the few NMIs with such NAA
laboratories in their facilities. However, the participation of selected
NAA laboratories in pilot studies, and in particular those pilot studies
organized in parallel to many key comparisons, enabled a direct
comparison of the results of NAA and IDMS laboratories. It should be
noted that unless these laboratories were officially designated by an
NMI, their results could not be included in the BIPM MRA Appendix B
nor used for establishing the key comparison reference value (KCRV).
In the period 2000–2006, 13 NAA laboratories have participated in
pilot studies and key comparisons (Table 3.1). Except for a few cases,
which are identified below, the laboratories used the instrumental (non-
destructive) variety of NAA (instrumental neutron activation analysis or
INAA). These NAA results have been used to evaluate the method's
potential as a primary ratio method by comparing NAA's performance
with the performance of a previously-recognized primary ratio method,
i.e. IDMS. The comparison is based on agreement with accepted values
(i.e., the KCRV if available), magnitude and completeness of the
estimated uncertainty budgets, and dispersion of the results among
laboratories operating the same method.
3.2. NAA performance of various laboratories: mono-isotopic
elements
The first opportunity for NAA laboratories to demonstrate the
performance of the method occurred in 2001, via the CCQM Pilot
Study P11 (CCQM-P11) “Arsenic in Shellfish” [1] followed in 2002 by
the associated CCQM Key Comparison K31 (CCQM-K31) “Arsenic in
Shellfish” [2]. Selection of arsenic as a measurand was largely due to
its importance as a common, toxic element, and the lack of a primary
method available for its determination. It is not possible to determine
arsenic by IDMS since the element is mono-isotopic.
⁎ Email address: lis@cena.usp.br.
233
In total eleven laboratories participated in CCQM-P11, six of them
used NAA. In the CCQM-K31, a total of eight laboratories participated.
Three laboratories used NAA including two NMIs, and one university
affiliated laboratory (the authors’ laboratory, CENA/USP), as an
unofficial participant. The results are presented in Figs. 3.1 and 3.2,
respectively. Dashed lines in these and subsequent figures represent
the best estimate of the central value uncertainty (k≈2) for each
study.
The results from the CCQM-P11 demonstrated good agreement
among the six NAA laboratories. The observed relative standard
deviation was 1.5%, compared to 5.9% for the five non-NAA
laboratories. The reported measurement uncertainties for the NAA
laboratories were in general similar to the uncertainties reported by
laboratories employing other methods (ICP-AES, HR-ICP-MS, GFAAS
and FI-HGAAS).
The observed relative standard deviation among the three NAA
laboratories participating in CCQM-K31 was 1.4%, and considerably
smaller than the observed relative standard deviation of 8.2% for the
results reported by laboratories employing other methods (GFAAS,
ICPMS, and FI-HG-GFAAS). In addition, the three NAA values appear
towards the middle of all reported values. It can also be seen from
Fig. 3.2 that the estimated measurement uncertainties of the three
NAA laboratories were similar to, and in some cases smaller than, the
ones reported by the non-NAA laboratories.
The results from these first two studies demonstrated that a significant
number of NAA laboratories, distributed among three continents, each
with different levels of metrological expertise, different nuclear reactors,
and different NAA instrumentation can provide results that are in good
Table 3.1
Overview of all pilot studies and key comparisons (with reference to their organizer) in
which NAA laboratories participated since the year 2001. The authors' laboratory
(CENA/USP) participated in pilot studies and key comparisons listed in this table,
except for P33, P62, K44, P74, and P86.
Year study CCQM key comparison (K) or pilot study (P) and organizing NMI
began
2001 –P11 arsenic in shellfish, NIST
–P29 cadmium and zinc amount content in rice, IRMM
–K24 cadmium amount content in rice, IRMM
–P34 constituents in Al alloy, BAM
2002 –K31 arsenic in shellfish, NIST
2003 –P39 As, Hg, Pb, Se and methylmercury in tuna fish, IRMM
–P33/P56 boron in Si, PTB
–K33/P56 minor elements in steel, NMIJ/NIST/BAM
2004 –P34.1 constituents of an aluminum alloy, BAM
–K42 constituents of an aluminum alloy, BAM
–P39.1 As, Hg, Pb, Se and methylmercury in salmon, IRMM
–K43 As, Hg, Pb, Se and methylmercury in salmon, IRMM
–P62 trace analysis of high purity nickel, BAM
–P65 chemical composition of clay, CENAM
2005 –P64 nonfat soybean powder, NRCCRM
–P66 key metals in fertilizer, NIST
–K44/P70 trace metals in sewage sludge, IRMM
–P74 composition of fine ceramics, NMIJ
2006 –K49/P85 essential and toxic elements in bovine liver, NIST
–P86 total Se and Se–methionine in yeast, LGC/NRC
234 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

20
As
0.15

Percentage difference (%)

10
0.14
INAA INAA INAA

mmol kg-1
INAA
0.13 0

INAA INAA
0.12
-10

0.11

-20

Fig. 3.1. Results from CCQM-P11 “Arsenic in shellfish”, adapted from [1].

20
As
0.13

Percentage difference (%)

10
0.12 INAA
mmol kg-1

INAA
0
0.11
INAA

0.10 -10

ISS C NA T A-1 GC IJ A-2

KR NR CE NIS IAE IM NM IAE
0.09 -20

Fig. 3.2. Results from CCQM-K31 “Arsenic in shellfish”, adapted from [2].

agreement with each other and with other analytical methods. In addition,
the observed relative standard deviations among the NAA laboratories, as
well as the magnitudes of their reported uncertainty statements, appeared
to be similar to those observed for IDMS measurements in other CCQM
studies of trace element content in complex matrices.
3.3. NAA performance compared to IDMS performance:
initial studies of the determination of trace element content in
complex matrices
The next step was to directly compare the performance of NAA and
IDMS in the same study. This first became possible via CCQM-K24/P29,
“Cadmium Amount Content in Rice” and “Cadmium and Zinc Amount
Content in Rice”, respectively. In addition, CCQM-P39, “As, Hg, Pb, Se and
Methylmercury in Tuna Fish” and CCQM-K43, “As, Hg, Pb, Se and
Methylmercury in Salmon” allowed comparisons of Se mass fractions
reported by similar numbers of INAA and IDMS laboratories.
The NAA and IDMS results reported for CCQM-K24/P29 are shown
in Figs. 3.3 and 3.4. These figures have been adapted from the original
reports to facilitate the comparison between performance of NAA and
IDMS. It can be seen from these figures that the mean values of the Cd
and Zn amount contents determined by NAA and IDMS, as well as the
variability among the NAA and IDMS laboratories and the estimates of
measurement uncertainty are similar for the two techniques. Note

20
17 Cd
Percentage difference (%)

16
10
IDMS INAA

15
nmol g-1

0
14

13 -10
AM

E
IIM

S
PA

IJ

N L
1

i
M
N MM

2
i-T A

ft
M
AR

T-

T-

el
LN

IS

PT

EN
LG

R
SI
M

BA

R
N
EM

D
VN

N
EN

IS

IS
C
KR

IR

12
N

U
C

N
C

M
N

-20

Fig. 3.3. Cadmium amount content in rice, results from CCQM-K24/P29, adapted from [3, 4].
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 235

0.42 20
Zn
0.40

Percentage difference (%)

10
IDMS INAA
0.38

µmol g-1
0.36
0
0.34

0.32 -10

0.30

M
AM

-1

B
L

ft

2
IJ

A
C
AR

T-

T-
el
IS

PT

LN

EN
LG

R
M

M
BA
IIM

D
N
EN

IS

IS
KR

IR
N

U
VN

N
i- T
C

M
N
-20

Fig. 3.4. Zinc results of CCQM-P29 “Cadmium and zinc amount content in rice”, adapted from [4].

that, in general, results from pilot studies are not publically available. K43). A numerical comparison of the results for the two methods
However, in a few cases agreement was obtained from all participants (NAA and IDMS) is given in Tables 3.2 and 3.3. Note that in Fig. 3.6, the
to make the results public. When such agreement was reached, and Se result from the authors' laboratory (CENA/USP) is included since
the data from the pilot studies are publically available, the the laboratory participated as an invited expert laboratory in
participating laboratories will be identified here. When pilot study the associated pilot study CCQM-P39.1 that was carried out in parallel
data is not publically available laboratories will not be identified, to CCQM-K43, as well as CENA/USP's long history of successful
except for the authors' own laboratory (CENA/USP). participation in CCQM studies. CCQM-P39.1 involved the determina-
The Se data from CCQM-P39, “As, Hg, Pb, Se and Methylmercury in tion of the same measurands in the same salmon material as CCQM-
Tuna Fish” and CCQM-K43, “As, Hg, Pb, Se and Methylmercury in K43. However, CENA/USP's results have not been included in the
Salmon” are presented in Fig. 3.5 (CCQM-P39) and Fig. 3.6 (CCQM- calculation of the Key Comparison Reference Value since the

20
95 Se

90

Percentage difference (%)

IDMS INAA 10
85
µmol kg-1

80 0

75

-10
70
M
S

C
E

K
A
IJ

T
IS
IS

C
EN
LG

G
M

BA
-L

-S
IM
N
KR

C
M

M
BN

65
IR

-20

Fig. 3.5. Selenium results from CCQM-P39 “As, Hg, Pb, Se and methylmercury in tuna fish”, adapted from [5].

20
Se
8.5
Percentage difference (%)

IDMS INAA 10
8.0
µmol kg-1

7.5
0

7.0

-10
6.5
IA

A
C

IJ

T
IS
LN

EN
LG

G
M

R
M

BA
M

IM

ET

N
N

IR
N

6.0
IN

-20

Fig. 3.6. Selenium results from CCQM-K43 “As, Hg, Pb, Se and methylmercury in salmon”, adapted from [6]. Also includes results from CENA in the parallel Pilot Study CCQM-P39.1.
236 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

Table 3.2 those of the NAA laboratories in the same study (CCQM-K43), despite
Comparison of results of Se determination in CCQM-P39 “Hg, Pb, Se, As and a greater number of NAA labs reporting results (5 versus 4). The
methylmercury in tuna fish” by NAA and IDMS.
difference between the IDMS results for the two studies seems, at
NAA IDMS least partially, to be due to the order of magnitude lower Se mass
Mean1 79.7 80.1 fractions for the key comparison. In conclusion, the results of the two
U (2 s/√n)1 1.7 0.94 studies of Se in fish indicate that for similar number of laboratories,
# of labs 5 4 the NAA results were similar, or perhaps a little better, than the IDMS
1 s% 2.4 1.2
results.
Range (%) 6.0 2.7
1
Units are 10−6 mol kg− 1. 3.4. NAA performance compared to IDMS performance: additional
complex matrices

Table 3.3
Comparison of results of Se determination in CCQM-K43 “As, Hg, Pb, Se and
methylmercury in salmon” by NAA and IDMS.
NAA
Mean1 7.36
U (2 s/√n)1 0.17
# of labs 5 4
1 s% 2.4
Range (%) 5.4
1
Units are 10−6 mol kg− 1.
laboratory is not affiliated to an NMI or an NMI designated laboratory;
hence CENA/USP's results were not included in the data of Table 3.3.
The mean Se content reported by the NAA laboratories in CCQM-
P39 was very close to the mean Se content reported by the IDMS
laboratories (79.7 μmol kg− 1 versus 80.1 μmol kg− 1), and well within
expanded uncertainties calculated as 2 s/√n for each method
(1.7 μmol kg− 1 and 0.94 μmol kg− 1). The relative standard deviation,
and the range of results for the NAA laboratories in CCQM-P39 (2.4%
and 6.0%) were about twice as large as the corresponding values for
the IDMS laboratories (1.2% and 2.7%). However, this may be partially
due to the larger number of NAA laboratories (5 versus 4). In addition,
the NAA RSD and range appear relatively typical for most CCQM
comparisons of trace elements in complex matrices, and are quite
small for most non-CCQM comparisons.
The mean Se content reported by the NAA laboratories in CCQM-
K43 was very close to the mean Se content reported by the IDMS
laboratories (7.36 μmol kg− 1 versus 7.33 μmol kg− 1), and well within
expanded uncertainties calculated as 2 s/√n for each method
(0.17 μmol kg− 1 and 0.43 μmol kg− 1). The relative standard deviation
and the range of results for the NAA laboratories in CCQM-K43 (2.4%
and 5.4%) were quite similar to the NAA results for CCQM-P39 (2.4%
and 6.0%); however, the corresponding values for the IDMS
laboratories in CCQM-K43 (5.8% and 13%) were significantly greater
than those of CCQM-P39 (1.2% and 2.7%). In addition, the dispersion
values were more than twice as large for the IDMS laboratories as
Subsequent key comparisons and pilot studies provided additional
opportunities to compare the performance of IDMS and NAA for more
measurands in other complex matrices. Additional CCQM compar-
IDMS isons including CCQM-P70 “Trace Elements in Sewage Sludge”; CCQM
7.33 P64 “Nonfat Soybean Powder”; CCQM-P65/K57 “Chemical Composi-
0.43 tion of Clay”; and CCQM-K49/P85 “Essential and Toxic Elements in
Bovine Liver” will be discussed.
5.8
Sewage sludge offers analytical challenges since it can be quite
13
difficult to dissolve, and the high levels of many elements provide
many opportunities for potential interferences. The results of four
NAA laboratories and four IDMS laboratories for Zn in sewage
sludge pilot study (CCQM-P70) are compared in Fig. 3.7. Note that
although the mean of the NAA and IDMS appears to be relatively
similar, there is a much greater interlaboratory dispersion among
the IDMS values. Although two of the IDMS values have much
smaller uncertainties than the NAA values, there is little overlap of
any of the IDMS values within reported uncertainties compared to
strong overlap of the uncertainties of all NAA values. Clearly this
measurand/matrix combination provided unrecognized challenges
to the IDMS method despite the relatively high amount content of
Zn.
Results for three NAA laboratories for Fe and Zn in CCQM-P64
“Nonfat Soybean Powder” are compared with those of six (Fe) or
seven (Zn) IDMS laboratories in Figs. 3.8 and 3.9. Although all NAA
values agree with all IDMS values within expanded uncertainties, one
NAA laboratory reported results with significantly larger uncertainties
for both Fe and Zn compared to the other two NAA laboratories, as
well as to all the IDMS laboratories. The large uncertainty appears to
be a result of lack of metrological experience, as this laboratory was a
relative newcomer to the field of metrology, and had very limited
experience participating in high-level intercomparisons.
Some clays can be quite difficult to totally dissolve, and thus provide
analytical challenges for methods requiring sample dissolution. CCQM-
P65, “Chemical Composition of Clay” allowed the comparison of Ca, Fe
and K results from one NAA and one IDMS laboratory. Results from all
methods are included in Figs. 3.10, 3.11 and 3.12. The results of all three

20
Zn
55
Percentage difference (%)

10
IDMS INAA

50
µmol kg-1

45

-10

40
-20

Fig. 3.7. Zinc results from CCQM-P70 “Trace metals in sewage sludge”, adapted from [7].
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 237

30
Fe
190
IDMS INAA 20
180

Percentage difference (%)

170
10
160

mg kg-1
0
150

140
-10
130

120 -20

110
-30

Fig. 3.8. Iron results from CCQM-P64 “Nonfat soybean powder”, adapted from [8].

30
56
Zn
IDMS INAA 20

Percentage difference (%)

52

10
48
mg kg-1

0
44

40 -10

36 -20

32
-30

Fig. 3.9. Zinc results from CCQM-P64 “Nonfat soybean powder”, adapted from [8].

elements determined by the NAA laboratory (the authors' laboratory at
CENA/USP) fall well within the range of results reported by laboratories
using other methods, as do the results from the IDMS laboratory. The
uncertainties of the NAA measurements appear to be similar to those
reported by most other laboratories and smaller than those reported by
the IDMS laboratory. The uncertainty reported for Ca by IDMS appears to
be the largest uncertainty reported.
Quite a few laboratories participated in CCQM-K49/P85 “Essential
and Toxic Elements in Bovine Liver”, including some with limited
experience in such high-level, metrological comparisons. In fact, some
laboratories were participating for the first time in a CCQM
comparison. In view of this lack of metrological experience, the
results of only the key comparison will be compared in Figs. 3.13 to
3.17. However, results from the authors' laboratory (CENA/USP) that
were submitted to the CCQM-P85 pilot study have been included in
these figures in view of CENA/USP's long history of successful
participation in CCQM studies (since 2001).
CCQM-K49/P85 included measurands of varying degrees of difficulty
in this matrix. Iron and zinc were considered to be relatively easy due to
their high levels, and the ease of matrix decomposition. The results for
NAA and IDMS participants in the key comparison are compared in
Figs. 3.13 and 3.14. Note that the dashed lines in these figures,
representing the relative key comparison reference uncertainties, are
both about 1%. Overall, the INAA and IDMS mean values agree quite well.

20
Ca
95
Percentage difference (%)

10
90 IDMS
CaO (mg g-1)

85
0

80 INAA

75 -10

70

-20

Fig. 3.10. Calcium results from CCQM-P65 “Chemical composition of clay”, adapted from [9].
238 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

20
63
Fe
60

Percentage difference (%)

10
57

Fe2O3 (mg g-1)

INAA
54
0
51
IDMS

48 -10

45

-20

Fig. 3.11. Iron results from CCQM-P65 “Chemical composition of clay”, adapted from [9].

20
30 K

Percentage difference (%)

28 10
K2O (mg g-1)

26
0

INAA IDMS
24
-10

22

-20

Fig. 3.12. Potassium results from CCQM-P65 “Chemical composition of clay”, adapted from [9].

The magnitudes of the expanded uncertainties of the NAA laboratories
appear quite similar to those of the IDMS laboratories.
Selenium, Cd and Cr were considered to be more challenging due
to their lower content and the higher potential for matrix interfer-
ences. The IDMS and NAA results from the key comparison are
compared in Figs. 3.15 (Se), 3.16 (Cd), and 3.17 (Cr). Results from the
three laboratories using INAA and two laboratories using IDMS to
determine Se in CCQM-K49 are compared in Fig. 3.15. All five results
were within the uncertainties of the key comparison. The magnitudes
of the reported expanded uncertainties for the two techniques are
similar, with the uncertainty from one IDMS laboratory larger and one
smaller than those of the three NAA laboratories.
Six IDMS results and one RNAA result were reported for Cd in
CCQM-K49. One laboratory reported both IDMS and RNAA results. As
can be seen in Fig. 3.16, the reported Cd mass fraction and uncertainty
determined by RNAA was indistinguishable from the IDMS values.

20
Fe
230
Percentage difference (%)

220 IDMS INAA

10

210
mg kg-1

200
0

190

180 -10

170
S

IM
L
IA

IJ

T
C

A
IM

B
KG
IS

IS

N
LN

PT
LG

M
M

R
N

CE
KR

N
N
N

IN
H

160 -20

Fig. 3.13. Iron results from CCQM-K49 “Essential and toxic elements in bovine liver”, adapted from [10]. Also includes results from CENA in the parallel Pilot Study CCQM-P85.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 239

20
Zn
210

Percentage difference (%)

200 IDMS INAA 10

190

mg kg-1
0
180

170
-10
160

IM
L

IA

A
T

IJ

T
C
IM

E
KG

IS
IS

IS
EN
LN

LG

BA
M

R
N

KR
N

N
N
N

IN
H

C
150
-20

Fig. 3.14. Zinc results from CCQM-K49 “Essential and toxic elements in bovine liver”, adapted from [10]. Also includes results from CENA in the parallel Pilot Study CCQM-P85.

20
2.4 Se

2.3

Percentage difference (%)

10
2.2 IDMS INAA
mg kg-1

2.1 IDMS
0
2.0
INAA INAA
1.9
-10
1.8
IM

A
IA
T

IM

EN
IS

M
R

N
N

N
IN

1.7
-20

Fig. 3.15. Selenium results from CCQM-K49 “Essential and toxic elements in bovine liver”, adapted from [10]. Also includes results from CENA in the parallel Pilot Study CCQM-P85.

Three IDMS laboratories and one INAA laboratory reported results for
Cr in CCQM-K49. Note that this measurement is extremely challeng-
ing with Cr present at approximately 50 μg/kg in a solid, complex
matrix. As can be seen from Fig. 3.17, both mean values and expanded
uncertainties from the individual laboratories were similar.
3.5. Discussion
Neutron activation analysis differs from other methods of chemical
analysis in that there is an absence of integral, commercially-available
instruments and standardized software. Nuclear research reactors, as
sources of neutrons, have different neutron intensities and energy
spectra available for activation; the gamma-ray spectrometers are
composed of individual detectors, amplifiers and multichannel analy-
zers, and can therefore not be compared with e.g. a standard
(catalogued) AAS or ICP instrument. In addition, software for gamma-
ray spectrum analysis and interpretation is also often not standardized,
but developed in-house. The very good agreement among the NAA
laboratories submitting results for the CCQM studies demonstrates the
robustness of the method, and the typically good agreement (within
uncertainties) of the NAA values with the reference values of the
comparisons indicates that the metrological principles of NAA are well

20
0.115 Cd

0.110
Percentage difference (%)

IDMS 10
0.105
mg kg-1

0.100 IDMS
0
0.095 IDMS IDMS
RNAA IDMS
0.090 IDMS
-10
0.085
1

-2

IA
IJ

M
E

IM
T-
LN

BA

M
N
IS

IS
N

0.080
N

-20

Fig. 3.16. Cadmium results from CCQM-K49 “Essential and toxic elements in bovine liver”, adapted from [10].
240 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241

20
IDMS
0.060 Cr

Percentage difference (%)

0.057 INAA
10

0.054

mg kg-1
0
0.051

IDMS
0.048
-10
0.045 IDMS

IJ

T
B

IM
0.042

IS
PT

N
N
N
-20

Fig. 3.17. Chromium results from CCQM-K49 “Essential and toxic elements in bovine liver”, adapted from [10].

Table 3.4
Comparison of IDMS and NAA characteristics for trace element determination; the symbols mark the characteristics in which a technique differs favorably from the other.

IDMS NAA

Instrumentation availability ++
Expanded uncertainties (U) — g/kg level ++
Expanded uncertainties (U) — mg/kg level (+) (+)
Blank ++
Dissolution losses ++
Elemental coverage Missing mono-isotopic elements Missing some key elements (Pb etc.)
Metrological expertise required yes yes
Agreement in comparisons Quite good, but dependent on level, Quite good, but dependent on level,
element, and matrix element, and matrix

understood and practiced in many NAA laboratories; even in ones that
are not affiliated with an NMI. However, metrological experience is
extremely important, and laboratories with less experience are more
likely to provide less reliable results and have problems in developing
complete uncertainty budgets.
It should be noted that the NAA laboratories participating in the
studies presented here applied the comparator method of standard-
ization, in which a calibrator consisting of a known amount of each
element to be determined in the sample of interest was used. The so-
called k0 method for standardization, often applied in many field
laboratories for direct multi-element determinations, was not used in
these comparisons because of the relatively high uncertainties of
some of the physical constants needed, and the still existing lack of
clarity on the traceability of the results obtained by this method of
standardization.
The niches and limitations of neutron activation analysis for trace
element determinations have been described and discussed in Chapter
1. During the last decade, NAA laboratories have participated in key
comparisons and pilot studies in which both the matrices and the
elements of interest were suitable for this method. A number of these
studies, beginning between 2001 and 2006, allowed for direct
comparison of NAA and IDMS determinations of amount content,
uncertainty budgets, and dispersion of reported results for each method.
As a result, it may be concluded that, in general, the metrological
properties of NAA and IDMS appear to be similar for the determination
of trace elements in complex matrices. Dissolution aspects may explain
that in some cases the range in NAA results was smaller than in IDMS
results. Lack of metrological experience may explain some of the results
with larger uncertainty budgets and deviations from the accepted study
values.
Although only a limited type of matrices and measurands were
involved in the key comparisons and pilot studies, a general
comparison of the characteristics and potential metrological perfor-
mance of both methods could be made, and is given in Table 3.4.
3.6. Conclusion
It has been demonstrated, via participation in CCQM key comparisons
and pilot studies starting between 2001 and 2006, that results
determined by metrologically-experienced NAA laboratories have
properties similar to those obtained by metrologically-experienced
IDMS laboratories for the determination of many trace elements in
complex matrices. INAA, because of the non-destructive nature, its
freedom from blank, and its elemental coverage provides a good
complement to IDMS methods. Neutron activation analysis, based on the
comparator method for calibration, can therefore meet all requirements
of a primary ratio method, and has been accepted as such by the CCQM.
The CCQM, in its meeting of April 19, 2007, decided subsequently “…that
NAA had claims to a similar status to that of the five methods listed
originally by the CCQM and that NAA will be added to that list….” [11,12].
Acknowledgements
The authors wish to express their thanks to all laboratories that
participated between 2001 and 2006 in the CCQM key comparisons
and pilot studies discussed above.
3.7 References for Chapter 3
[1] R. R. Greenberg, E.A. Mackey, CCQM-P11 Pilot Study Arsenic in Shellfish, Final
Report November 28, (2001) NIST, Gaithersburg, USA.
[2] R.R. Greenberg, E.A. Mackey, CCQM-K31 Key Comparison: Arsenic in Shellfish,
Metrologia 43, (2006) 08003.
[3] Y. Aregbe P. Taylor, CCQM-K24 Key Comparison. Cadmium Amount Content in
Rice, Metrologia 40, (2003) 08001.
 R.R. Greenberg et al. / Spectrochimica Acta Part B 66 (2011) 193–241 241

[4] Y. Aregbe P. Taylor, CCQM-P29 pilot study. Cadmium and Zinc Amount Content HR-ICP-MS high resolution inductively coupled plasma mass
in Rice, Metrologia 40, (2003) 08002.
spectrometry
[5] Y. Aregbe, C. Quétel, P.D.P. Taylor, CCQM-P39: As, Hg, Pb, Se and
Methylmercury in Tuna Fish, Metrologia 41, (2004) 08004. ICP inductively coupled plasma
[6] Y. Aregbe, P.D.P. Taylor, CCQM-K43: As, Hg, Pb, Se and Methylmercury in ICP-AES inductively coupled plasma atomic emission spectrometry
Salmon, Metrologia 43, (2006) 08005. ICPMS inductively coupled plasma mass spectrometry
[7] Y. Aregbe, P.D.P. Taylor, presentation at CCQM-IAWG Meeting, Oct. (2005), IDMS isotope dilution mass spectrometry
CCQM-K44 and P70: Trace Metals in Sewage Sludge.
[8] M. Liandi et al., 2006, Report of the CCQM-P64 & APMP.QM-P07, Cu, Zn, Fe and
IMGC Institute of Metrology Gustavo Colonnetti, Italy
Ca in Nonfat Soybean Powder. INMETRO Instituto Nacional de Metrologia Normalização e Qualidade
[9] J.A. Salas, E.M. Ramírez, (2006) Report of CCQM-P65 Study, Chemical Industrial, Brazil
Composition of Clay. INRIM Instituto Nazionale di Metrologia delle Radiazioni
[10] R.R. Greenberg, Final Report on Key Comparison CCQM-K49: Essential and
Ionizzanti, Italy
Toxic Elements in Bovine Liver, Metrologia 45, (2008) 08016.
[11] Report of the 13th meeting of the CCQM, 19–20 April (2007), http://www. IRMM Institute for Reference Materials and Measurements,
bipm.org/utils/common/pdf/CCQM13.pdf (last accessed August 19, 2010). European Commission
[12] Report of the 14th meeting of the CCQM, 3–4 April (2008), http://www.bipm. KRISS Korean Research Institute of Standards and Science,
org/utils/common/pdf/CCQM14.pdf (last accessed August 19, 2010). Republic of Korea
LGC Laboratory of the Government Chemist, United Kingdom
3.8 Acronyms used in Chapter 3 LNE Laboratoire National de Métrologie et d'Essais, France
MRA Mutual Recognition Arrangement
NARL National Analytical Reference Laboratory, Australia
AAS atomic absorption spectroscopy NIM National Institute of Metrology, China
BAM Bundesanstalt für Materialforschung und -prüfung, Germany NIST National Institute of Standards and Technology, United
BIPM Bureau International des Poids et Mesures States of America
BNM-LNE Bureau National de Métrologie – Laboratoire National d’Essais, NMi Nederlands Meetinstituut, The Netherlands
France NMI national metrology institute
CCQM Consultative Committee for Amount of Substance – Metrology NMIA National Measurement Institute, Australia
in Chemistry NMIJ National Metrology Institute of Japan, Japan
CENAM Centro Nacional de Metrología, México NMI-TUDELFT Nederlands Meetinstituut, Delft University of
CENA/USP Centro de Energia Nuclear na Agricultura – Universidade de Technology, The Netherlands
São Paulo, Brazil NRC National Research Council of Canada, Canada
CSIR Council for Scientific and Industrial Research, South Africa NRCCRM National Research Centre for Certified Reference Materials,
EMPA Swiss Federal Laboratories for Materials Testing and Research, China
Switzerland PTB Physikalisch-TechnischeBundesanstalt, Germany
FI-HGAAS flow injection hydride generation atomic absorption SCK Belgian Nuclear Research Centre, Belgium
spectroscopy VNIIM D.I. Mendeleyev Scientific and Research Institute for
FI-HG-GFAAS flow injection hydride generation graphite furnace Metrology, Russia
atomic absorption spectroscopy
GFAAS graphite furnace atomic absorption spectroscopy
HKGL Hong Kong Government Laboratory, Hong Kong