XA0200221

REFERENCE MATERIALS AND THEIR ROLE IN

QUALITY ASSURANCE IN ENVIRONMENTAL MONITORING

Eduardo Cortes
Chemical Metrology Unit, Nuclear Applicatios Department
Chilean Nuclear Energy Commission
P.O. Box 188-D, Santiago, Chile

Abstract

The importance of a good and routine quality control procdures for the analyses of
environmental samples is presented. The use of Reference Materials as one simple procedure
for validating analytical methodologies and determine the accuracy of analytical data is
emphasized. The quality of the rference materials is also discussed as it is their selection and
correct use. The convenience of preparing "in-house" reference materials is discussed and
attention is called to relevant aspects to be consider. An example of the preparation of a
reference material is presented and some aspects of the procedure are discussed.

1: INTRODUCTION

It was after the United Nations' Summit on the protection of the environment held in
Rio de Janeiro (Brazil) in June 1992, that many countries decided to give special attention to
the problems related to the protection of the environment. This concern has been confirmed
with the implementation of several projects or international conventions aiming at a
sustainable development and the conservation of the environment Among such actions it can
be mentioned the Paris and Oslo convention for the North Sea, or the Barcelona convention
for the Mediterranean Sea. Some of the activities considered in these plans require a
permanent monitoring of the geographical areas under consideration. These monitoring
programmes implies the measurements of various parameters necessary to evaluate the
situation and follow their trends. Decisions to be taken by the authorities and actions to
mitigate possible damage to the environment are based on the results of such measurements.
On the other hand, monitoring programmes implies long term sampling and the respective
analysis of the collected specimens. Thus, the analytical process plays a very important role in
this context and requires to be carry out under well defined procedures and strict quality
control programmes.

The need for reliable information on chemical substances in the environment is of

most importance. There is a continuos emission of compounds and chemical elements into
the environment from different sources. Some of these substances require profound study to
determine possible effect on the flora and fauna and, if necessary, to give a scientific basis to
the health and environmental authorities to take measures to control or limit their use.

Chemical measurements in the different compartments of the environment are necessary for
several purposes [I]:

• Legislation may require the checking of the concentrations with respect to the
standards in products, environmental compartments and/or in waste materials.

1
 Corrective actions are necessary if the standards are exceeded.
• Legislation may entail that a penalty is imposed, on the basis of the
concentrations of a chemical parameter (e.g. the chemical oxygen demand in
waste waters)
• The method of disposal for a certain material (e.g. dredged material) may have
to be established on the basis of its degree of contamination. The method of
disposal of the material is assessed on the basis of the concentrations of a
number of priority pollutants.
• Environmental monitoring programmes may be started to assess temporal
trends and spatial distributions of potential pollutants. These programmes
provide feedback with respect to the effectiveness of the measures taken.

Reliable data are the result of a chain of actions which starts with the proper definition of the
problem to be solved. This should lead to a clear identification of the parameter(s) concerned.
It includes the target samples to be selected, the sampling strategy and the proper sampling
technique to be applied. When adequate samples have been analysed, the results should be
reported in such a manner that those who have to draw the conclusions and to take action, can
do so with all possible assurance. [2]

Analytical steps Estimation of errors

Defining the scientific problem

Discussion by experts ; Production of
Benefits/ use calculation : " data cementery"
Panning the analyses

-0-
j Representative sampling j Up to 1000%

-0-
Sample preparation

1. Physically
washing
drying
homogeneization
Between 100 - 300%
2. Chemically
ashing
decomposition
enrichment
speciation
__
•IS
Normally between
Instrumental measurement 2 - 20%

Data evaluation
Up tp 50%
Solving the scientific problem
I.

FIG. 1. Estimation of errors in different steps of a

chemical analytical process

X"
The analytical process comprises several steps, each one having their own relevance and
importance when determining the reliability of the final result. Figure 1 shows the different
stages during a chemical process and an estimation of their contribution to the total
uncertainty of the final result. [3]

However, analysts pay great attention to the instrumental measurement since it is the
step under their own control and responsibility. Correct determinations of chemical
parameters implies precise (repeatable and reproducible) and accurate results. To achieve the
most reliable results the laboratory has to have a qualified infrastructure and personnel, an
established periodical maintenance of apparatus, proper chemicals, adequate management, etc
according to a quality assurance programme.

2. ACCURACY IN THE ANALYTICAL LABORATORY

Precision may be achieved within the laboratory by working under strict QA and QC
rules and by properly validating the analytical method, while accuracy is demonstrated by
external means. Accuracy is necessary to produce comparable data which are comparable
among laboratories. Environmental monitoring is a good example for which accurate data are
needed, e.g. for evaluation of contamination trends or assessment of environmental quality.
Accuracy can be demonstrated by three different means, (i) comparison with a different
method, (ii) comparison with other laboratories, and (iii) use of certified reference materials.
A short description of the first two approaches and a more comprehensive discussion on the
third procedure will follow

2.1. Comparison with a different method

Each analytical method has it own sources of errors. Results provided by atomic
absorption spectrometry (AAS) may suffer from error which may occur during the digestion
step, while for instrumental neutron activation analysis (INAA) measurements may be
influenced by the resolution of the gamma ray detection system, e.g. insufficient separation of
gamma-ray peaks. If results obtained on similar samples by both methods are in good
agreement, it may be considered that the values found are accurate.

For this approach it is recommended to use only methods validated in detail and under
statistical control. It has been demonstrated that a good quality assurance programme is
essential to the accuracy and precision of analytical results. Further, the cross-comparison of
two methods for the evaluation of their accuracy should not be done within one laboratory but
preferable participating in interlaboratory exercises.

2.2. Comparison with other laboratories

Participating in interlaboratory round robins helps to improve the quality and

performance of a laboratory. These intercomparisons consist in exercises in which several
laboratories determine one or more substances in a similar sample provided by a central
organizing body. Comparison of different methods applied in different laboratories enables
detection of sources of errors as a result of a specific procedure or a given laboratory.
 Global and Local
Comparability
arbfifrd reference molenofc.
ltcoricv l e ^ i and accreditation

Internal Comparability of
Laboratory Data
MuffialioralGjc*^ccn!tc(uwiiQn-tiauso
"canl-ol sanities" afKi opptop^ilocaf

FIG. 2. How to obtain results close to the

true value bt means of traceability

By participating in properly organized interlaboratory studies it is possible to evaluate

the accuracy of the laboratory and compare the performance of given methods and
techniques., The organization of interlaboratory studies involves the production and
distribution of reference materials (RMs) of verified homogeneity and stability, the collection,
treatment, and presentation of the results, and, highly desirable, the organization of a technical
meeting with all the participants. The statistical treatment of the data will not enable outlying
results to be explained nor give information about accuracy; it is only useful to analyse the
data population and inform on its statistical properties.

2.3. Use of certified reference materials

Modern practices of analytical chemistry have introduced concepts which have to be

applied in every day routine work, in the laboratory. Among those concepts, traceability and
uncertainty are closely related to the quality and reliability of analytical results.

By definition, traceability of a measurement is achieved by an unbroken chain of

calibrations connecting the measurement process to the fundamental units, known as SI Units.
In the vast majority of chemical analyses, the chain is broken because in the treatment the
sample is physically destroyed by dissolutions, calcinations, etc. To approach full traceability
it is necessary to demonstrate that no loss or contamination has occurred in the course of the
sample treatment. The only way for any laboratory to ensure traceability in a simple manner is
to verify the analytical procedure by means of a so-called matrix RM certified in a reliable
manner. The laboratory which measures such an RM by its own procedure and finds a value in
disagreement with the certified value is thus warned that its measurement includes an error, of
which the source must be identified. Figure 2 shows the required steps to demonstrate
traceability of the measurements performed in an analytical laboratory.

On the other hand, as indicated in the International Vocabulary of Basic and General
terms in Metrology [4], uncertainty is a parameter associated with the result of a
measurement, that characterises the dispersion of the values that could reasonable be
attributed to the measurand. Such parameter could be, for instance a standard deviation, a
multiple of it or the width of a confidence interval. Uncertainty of measurements comprises,
in general, many components. Some of these components may be evaluated from the
statistical distribution of the results of series of measurements and can be characterised by
standard deviation. Other components, which can also be characterised by standard deviation,
are evaluated from assumed probability distributions based on experience or other
information. The ISO Guide refers to these different cases as Type A and Type B estimations
respectively. Ideally, all laboratories should try to identify all possible sources of uncertainly
in their analytical process, determine the uncertainty of their measurements, and, therefore,
report all results with their respective associated uncertainty.

The reader is advice to consult the very useful publications on Quantifying Uncertainty in
Analytical Measurement, First and Second editions, published by Eurachem [5]

3. REFERENCE MATERIAL AND THEIR ROLE IN ENVIRONMENTAL

MONITORING

The use of RMs is the easiest way to achieve accuracy. Certified reference materials
link the user's results to those of the international scientific community. Additionally, they
enable the user to verify its performance at any desired moment. ISO GUIDE 33 has given the
following definitions for CRMs and RMs [6,7].

Reference material (RM): a material or substance one or more properties of which are
sufficiently well established to be used for the calibration of an apparatus, the assess-
ment of a measurement method, or for assigning values to materials.

Certified reference material (CRM): a reference material one or more of whose prop-
erty values are certified by a technically valid procedure, accompanied by or traceable
to a certificate or other documentation which is issued by a certifying body.

Reference materials are used for a number of purposes in the analytical laboratory:

• for the verification of the accuracy of results obtained,

• for monitoring the performance of the method,
• to calibrate equipment which requires a calibrant similar to the matrix,
• to demonstrate equivalence between methods and
Q to detect errors in the application of standardized methods

The conclusion on the accuracy obtained on the unknown sample is always a

conservative one: if the laboratory finds wrong results on a RM it is by no means certain of a
good performance on the unknown. If, however, the laboratory finds a value in agreement
with the certified value (according to ISO Guide 33 [6]), owing to discrepancies in
composition between RM and unknown, there is still a risk that the result on the unknown
may be wrong. The use of as many as possible relevant RMs is therefore necessary for a good
QA.

3.1. Availability of reference materials

There are a number of suppliers of matrix CRMs for almost all .type of sample
analysed currently in analytical laboratories. Some of them specialize in a certain field of
interest: the International Atomic Energy Agency (IAEA) through its Marine Environment
Laboratory located in Monaco produces marine RMs as well as the National Research Council
of Canada, NRCC, Two main bodies, the National Institute of Standards and Technology
(NIST, USA) and the BCR of the Commission of the European Communities, cover several
fields and ensure long-term availability of RMS, owing to the large batches of materials
produced. The IAEA Seibersdorf Laboratory, near Vienna, Austria, mainly provides certified
materials for nuclear measurements but has available some RMs for non-nuclear analysis.

There are a number of publications about the availability of RMs. The iSO Council on
Reference Materials (REMCO) of ISO has available a Directory for Reference Materials
which may be consulted [8]. The International Union for Pure and Applied Chemistry
(IUPAC) issues a catalogue of available CRMs [9J. The IAEA has been concerned with the
exchange of information to the scientist in its Member States and also produced a first survey
on RMs in the early 1980s, which was followed for a second one in 1990 on Biological and
Environmental Reference Materials for Trace Elements, Nuclides and Organic
Microcontaminants [10]. At present, the IAEIA has a full data base on RMs at its Web site on
the INTERNET.

Despite the apparently large number of suppliers and of available RMS, there is still a
lack of RMs of different matrices and analytes, such as F, I, Si and Sn in biological CRMs and
a multielemental RM for airborne particulate matter. A problem which may also arise with
RMs. as happened in the past, is that some certified values may be erroneous and the RM
needed to be re-certified. Examples of this situation was the change in the certified value for
arsenic in NBS SRM-1571 Orchard Leaves and in the IAEA-A-11 Milk Powder and H-4
Animal Muscle, were new values had to be assigned to As, Cd, Cr, Cu, Hg, I, Mn, Mo, Ni and
Pb. However, the proper use of available CRMs remains one of the most effective ways to
validate an analytical method.

It is recommended to select RMs as similar as possible to the samples being analysed

in both respects: to the type of matrix (in order to be subject to the same levels of
interferences) and with regard to the concentration levels of the analyte(s) of interest. This is
not always possible since it could happen that the matrix is not available or the levels of the
analyte(s) of interest are different from those found in actual samples. On the other hand, the
cost of RMs produced by internationally well known suppliers limits drastically the possibility
of buying such materials These facts have motivated that many laboratories, in particular in
developing countries, are preparing their own control materials to satisfy their requirements.
This situation is more dramatic when there are economical implications due to trade of goods,
special with exported food. As an example, large amounts of seafood are exported from the
author's country to Europe, North America and Asia. The importing countries have set
maximum permissible levels for a number of heavy metals, such as Cd, Hg and Pb. in
molluscs. This requires an accurate determination of these elements, with appropriate
detection limits and a solid technical basis to assure a complete harmonization of the results
produced in both the exporting and the importing country. A search for a suitable RM of the
matrix and analytes adequate to the samples, none was found. It was the decided to look into
the possibilities of preparing RMs of the highest possible quality at the author's institution.

Two advantages were found in this respect. The first one was the fact that it would be
possible to prepare a material "a la carte", with respect to the matrix, according to the needs of
the laboratories. The second one was that a previous study carried out allowed the
characterization of several banks of molluscs in the coast, thus, the levels of the elements of
interest was known for each potential source of raw material. Taken into consideration these
and others prerequisites, the appropriate mollusc (matrix) and sampling site were selected and
the material collected.

However, the preparation of reference materials is not a trivial matter. There are a
number of requisites that the product has to have in order to fulfill the requirements. The
experience of the author in the preparation of one of the RMs, namely CCHEN-MR-002, clam
{Venus antiqua) produced by its institute, is described in the following paragraphs.

3.2. Preparation of reference materials

Reference materials can be prepared from several sources depending on the purpose
they are issued for:

• pure substances or solutions for calibration and/or identification or test of

procedures in improvement schemes
Q materials of known matrix composition for the calibration of comparative
methods
• matrix materials which as far as possible represent the matrix being analysed
by the user, either used in interlaboratory studies or for the verification of a
measurement process
As far as possible, the RM to be used should be similar to those samples routinely
analysed both with regard to the type of matrix (in order to be subject to the same levels of
interferences) and with regard to the concentration levels of the analyte(s) of interest. It is rare
to find an RM matrix exactly the same as the samples to be analysed, but the same type of
material should be used where possible. It is important that the concentration of analyte in the
RM should be of the same order of magnitude as that in the majority of the samples, otherwise
the technique is not adequately tested.

3.3. Selection, collection and preparation of the raw material

For environmental monitoring, where the matrices of the samples being studied are
very important, it is highly relevant to look for the appropriate raw material and, if possible,
perform a survey to find out the most convenient collection place.

The amount of raw material to be prepared should be such that the supply of the RM
will last for several years. Therefore, the amount of raw material should be several hundreds
kilograms in the case of molluscs, since once the shell and some parts are eliminated the
original weight can be reduced to as low as 20% of the original weight. The laboratory' has to
be properly equipped to handle large amounts of material.

In collecting the material, efforts has to be placed to unsure a clean, non-contaminated

material and that no losses of any of the analytes of interest occur. This is particular relevant
when the material is collected in places far from the laboratory or place were it will be
processed, as normally happen. Frozen of the material at the collection site is highly
recommended and it should be transported in such state.

7
 The collection of molluscs is normally done by professional divers and this operation
has to be closely supervised by a technical person. Care had to be placed in selecting
individuals as uniform as possible as regard size to ensure a similar age and, therefore,
analogous content of the analytes of interest.

Once in the laboratory, the material has to be further clean to assure that does not
contain extraneous substances such as sand or rest of shell. In the case of CCHEN-MR-002,
this cleaning was performed in a large portable fume hood with HEPA filters and materials
made out of Teflon and titanium. The hepatopancreas of the clam was also eliminated and the
parts of interest were again frozen. The material was then milled, mixed and storage at -80 °C.
All the material was treated in the same way. The material was then lyophilised in a large
capacity freeze drier. Once all the material was dried it was grinded and milled again and
homogenised in a large size "Y" shaped homogeniser. Most of the equipment used for the
preparation of the material was made at the institute mechanical and electrical shops.

The material candidate to RM was then sieve an distributed into clean glass bottles in
amount of about 25 g and sent for sterilization with 60Co gamma-rays

3.4. Homogeneity and stability

The RMs prepared from biological and environmental materials are heterogeneous by
nature. They should be subjected to tests to assure a certain degree of homogeneity to allows
them to fulfill the requirements of the analysts.

In theory, a material is perfectly homogeneous, with regard to a given characteristic, if

there is not a difference among the value that presents that characteristic in different sub-
samples of the same material. However, in practice it is accepted as homogeneous material,
with regard to a given characteristic, if experimentally a difference cannot be detected in the
characteristic measure in different sub-samples of the material.

The quantity of material in an individual bottle is, normally, enough to allow several
determinations of the analyte(s) of interest. Therefore, the homogeneity of the RM, with
respect to a particular analyte has to be verified to ensure that, within a bottle and from one
bottle to another, the content is the same (within- and between bottle-homogeneity). Another
important information for the analysts, which is derived from the determination of the
homogeneity refers to the minimum sample size which should be taken for analyses. This is,
the amount of sample for which it can be guarantee that the uncertainty caused by
inhomogeneity does not exceed the uncertainty of the assigned [11,12].

Therefore, the practical concept of homogeneity involves a determination of the

characteristic of interest and a parameter to estimate the dispersion of results (standard
deviation or variation coefficient). Also, in many cases, the precision of a measuring method
is affected by the size of the portion taken for analysis. Therefore, the grade of homogeneity is
defined for a specific sample size.

It is important to keep in mind that a material can be sufficiently homogeneous with

regard to the characteristic of interest to be useful as a RM although it is inhomogeneous with
regard to other characteristics, provided this lack of homogeneity doesn't exercise an influence
detectable on the accuracy and precision of the methods of determination used for the
characteristic of interest.

In a candidate to RM it is fundamental to carry out a main test to determine the

homogeneity, after it has been packed in their final form as usable units (bottles), selecting
randomly "h" units and carrying out "n" measurements of the characteristic of interest in each
one of them. The objective of this is not simply to detect a possible heterogeneity, but rather,
to estimate its magnitude.

The heterogeneity can be shown from two points of view:

• Different subsamples of one unit of the candidate to reference material can

differ in the property of interest,
• It can have differences among the units from the candidate to reference
material.

In statistical terms the experimental design should satisfy the following objectives:

• To detect if the variation inside the units is statistically significant in

comparison with the well-known variation of the measuring method.
• To detect if among the units, the variation is statistically significant in
comparison with the variation inside the unit, and
• When detecting variation, to determine if the magnitude of this is enough as to
disqualify the candidate to reference material.

3.5. Constant of homogeneity.

The use of small samples for chemical analysis could result in obtaining big errors
[13], This discovery motivated it to develop the concept of sampling error SH that is the
variation coefficient due to heterogeneity that implies the variation of the distribution of an
analyte in taken samples of a material lot. The magnitude of SH was considered clerk of the
mass of the sample, the distribution of the particle size, the density of the particles, the
number of components, and the distribution of the analyte among the components.

In microanaliticos methods, the sampling error SH is the biggest contributor to the total
variation, because they use very small test portions of the material, and the probability of
finding in two different portions of same mass the same distribution of particles with identical
composition, diminishes.

2
10Q.(CA-CB) f A .f B d A .4-p.r
S H (%)= •i
C V d-3-m

where C, CA and CB are the concentrations of the element in the total sample and in their
component phases A and B respectively; d, dA and dB are the densities of the sample and of
their component phases A and B, respectively; r is the mean radius of particle, m it is the mass
of the sample, and fA and FB they are the mass fractions of the components A and B,
respectively.
 In practice, SH diminishes when increasing the mass of the sample. This happens
because the materials in general are subjected to a process of quite efficient homogenization,
and when using big masses it exists the probability that significant differences are not
detected, within the limits of precision of the used method. Also, the particles with high
content of the element in question, are compensated with those that have low content,
obtaining a smaller variation in the results. This variation will be dominated by the random
errors characteristic of any mensuration, being insignificant the variation caused by
heterogeneity. In the practical SH it also diminishes when diminishing the particle size by
means of the pulverization of the sample.

To calculate the sampling error, it is required information about the particle size,
density and composition of the sample. In 1973 Ingamells and Switzer postulated that a well
mixed material can be characterized by the sampling constant (KJ [14,15] term defined as:

The relative sampling error can also be expressed by the constant of relative
homogeneity HE defined for Kurfurst [16], where there HE represents the error of relative
sampling if a sample of unitary mass is used.
Hg
SH{%) = -j=
\'m
The reason to use HE instead of Ks, is that Ks gives very big numbers for moderate
masses of sample. On the other hand, HH is particularly appropriate to describe small masses
of sample, giving moderate values for the constant.

4. BASIC STATISTICS TO DETERMINE THE HOMOGENEITY OF

MATERIALS.

4.1. Number of units and subsamples to analyse

To measure the homogeneity among the units of a material prepared from a bulk
sample where, in principle, there is no reason to define subsamples from the total, it is
recommended to choose units randomly; to give each unit an opportunity similar of selection,
from the total of units "UT". A number between 10 and 30 units is acceptable in most cases, or
one can use the following formula to determine
3 •^UT = h

To prove the homogeneity inside the units, it is enough to analyse 5 or 6 samples for each 2
units. However, it is preferable to analyse 2 or more subsamples of each selected unit and to
carry out the comparison calculations using analysis of variance (ANOVA) and the Fisher F
test.

4.2. Statistical interval of tolerance.

The uncertainty of a certified value for a reference material ideally should be

manifested by several components, some supported by data and other not:

10
 • A statistical interval of tolerance giving limits for the heterogeneity of the
material based on data and statistical calculations.
• An interval of confidence giving limits for the error of the measurements based
on data and statistical calculations.
Q Components of the measurement uncertainty caused by the variation between
laboratories and/or operators and measuring methods.
• A combination of estimated limits for "well-known" sources of possible
systematic errors based on the experience and the experience (in other words,
there are not data or these they are insufficient to make a statistical calculation).

The statistical interval of tolerance estimates the interval that should contain a certain
percentage of the population's individual measurements.

To establish an interval of tolerance (p), first the average of the measured property is
calculated (X) and then a measure of the variation (s), that can be the standard deviation or
coefficient, variation.

I = X±p
The value of p is calculated as a multiple of s
p = k'-s

The K' value is tabulated in statistical books (19SUSANA) and depends on three
parameters:

• The total number of samples measures.

Q The proportion of the population's total to be contained (95%).
Q The specified level of probability (95%).

In the case of the CCHEN-MR-002, the material candidate to RM was carefully and
thoroughly tested to determine its homogeneity.

4.3. Homogeneity testing of the CCHEN-MR-002

The main determination of the homogeneity of the CCHEN-MR-002 was performed

using a Solid Sampling Atomic Absorption Spectrometry (SS-AAS) for the analyses of a
large number of samples [17-19]. The SS-AAS is a technique already established within the
family of AAS techniques. It allows the introduction of small amounts of samples, typically
about 1 mg, of solid material for electrothermal atomization in a graphite furnace and single
element determination. One important task for this instrument though, as also for many others,
is a good calibration. This was performed using certified reference materials of matrices
similar to that of the sample of interest. The technique is quite fast and no chemical treatment
(digestion) of the sample is necessary. Typical analysis time is about 2 minutes. In addition, it
has a low cost and good sensitivity for a large number of elements. This paper describes the
results obtained for the determination of Cd and Pb in the material candidate to reference
material CCHEN-MR-002.
4.4. Methodology for lead determination

The development of the analytical procedure is elsewhere [20]. The validation of the
methodology was carried out using the certified reference material NRC Dolt-2 Dogfish Liver
with a content of 0,22 ± 0,02 ^g/g of lead. Table I shows the data obtained during four days of
measurements of the CRM. Figure 3 shows a Shewhart control chart for those measurements.
The average analytical mass was 0,616 [mg].

TABLE I. Results of the analysis of the CRM for Pb determinations

Sample mass [ingj Mass fraction [fig/g]

Day 1 0.656 0.225

0.638 0,199

0.641 0.222

0.613 0,232

0,609 0,225

0.641 0.241

Day 2 0.606 0.230

0.590 0.220

0,630 0,212

0,602 0.232

0.596 0.211

0.632 0.217

Day 3 0,600 0.207

0.600 0.206

0,618 0.178

0,609 0.259

0,596 0.197

0.628 0,192

Day 4 0.613 0.187

0.632 0.216

0.590 0.244

0.629 0.241

0.625 0.239

0.596 0.258

12
 The statistical analysis of the evaluation of the methodology is presented in the Table II. Also in
the Figure 4 a Shewart control chart shows the obtained results, which allows to evaluate any tendency of
the methodology during its use.

TABLE II. Statistical parameters from the analytical results of the determination of Pb in the CRM

Number of measurements 24
Average mass fraction 0,220 [|ig/g]
Standard deviation 0.021
Coefficient of variation 9,55%
Test de Student
(for a 95% confidence level and 23 d.f
t calculated 0
t tabulated 2,07
Test de Fischer
(for a 95% confidence level and 23 d.f
F calculated 1,10
F tabulated 2,02
Control limits
Warning limits 0,262 ; 0,178
Action limits 0,283; 0,157

The statistical parameters obtained confirm that the methodology doesn't show evidences
of systematic errors and the calculated concentration is very close to the certified concentration.

——— UCL 0.270

Number of measurement

FIG. 3. Shewart control chart for the determinations ofPb in the

NRC Dolt-2 Dogfish Liver CRM
4.5. Methodology for cadmium determination

The development of the analytical procedure is elsewhere [20].For the validation of the
methodology the NIST SRM 1566 Oyster Tissue certified reference material was used. The
content of Cd in this CRM is 4,15 ± 0,38 (J-g/g. Table III shows the data obtained during four
days of measurements of the CRM. Figure 4 shows a Shewhart control chart for those
measurements. The average analytical mass was 0,128 [mg].

TABLE III: Results of the analyses of the CRM for Cd determinations

Sample mass (ing] Mass fraction [(ig/g]

Day 0.129 4.38

0.128 4.10

0.128 4.46

0.123 3.94

0.13! 4.36

0.123 4.04

Dav 2 0,127 4.13

0.128 4.38

0.129 4.22

0.129 3.89

0.132 3.96

0,127 4.31

Day 3 0.129 3.96

0.128 4.03

0.125 4.20

0.132 4.14

0.126 4.46

0.125 4.04

Day 4 0.127 4.18

0.128 3.99

0.130 4.32

0.127 4.17

0.126 3.87

0.130 4.03

14

-14-
 The statistical analysis of the evaluation of the methodology is presented in the Table IV.
Also in the Figure 6 a Shewart control chart shows the obtained results, which allows to
evaluate any tendency of the methodology during its use.

TABLE IV. Statistical parameters from the analytical results of the determination of Pb in the CRM

Number of measurements 24
Average mass fraction 0,220 [mg/g]
Standard deviation 0.021
Variation coefficient 9,55%
Test of Student
(for 95% level of confidence and 23 d.f
t calculated 0
t tabulated 2.07
Test of Fischer
(for 95% level of confidence and 23 d.f.
F calculated 1.1
F tabulated 2.02
Control limits
Warning limit 0,262; 0,178
Action limit 0,283:0,157

Number of sample

FIG. 4. Shewart control chart for the determination of Cd in

the NIST CRM 1566 Oyster tissue CRM

As in the case of Pb, the statistical parameters obtained confirm that the methodology
doesn't show evidences of systematic errors and the calculated concentration agrees with the
certified concentration.
5. DETERMINATION OF THE DISTRIBUTION OF Pb AND Cd IN THE CCHEN-
MR-002

To determine the homogeneity of the material candidate to reference material

MR-CCHEN-002, with regard to the lead contents and cadmium the methodology used was the
described previously. Thus, 6 units of the packed material were selected at random and 17
subsamples from each bottle were taken and analysed by SS-AAS.

5.1. Determination of the total variation:

The determination of the total standard deviation was made considering the results of the
102 measures of lead concentration and cadmium obtained for the material in study using a mass
average of 0,916 mg to determine lead and of 0,893 mg to determine cadmium,

Figures 5 and 6 show the plots of the mass fraction for Pb and Cd, respectively, for the
102 measurements.

0 6 12 18 24 30 36 42 48 $4 60 66 72 78 84 90 96 102

Number of sample

FIG. 5. Results of the determination of Pb in the material

candidate to reference material CCHEN-MR-002

0 6 12 18 24 30 36 42 46 54 60 66 72 78 84 90 96 102
Number of sample

FIG. 6. Results of the determination ofCdin the material 16

candidate to reference material CCHEN-MR-002
 An analysis of variance (ANOVA) was performed for the data obtained to detect any
significant difference between the units (bottles) or within a given unit. The result of this
analysis is presented in Table V.

TABLE V. Results of the analysis of variance (ANOVA) of the results for Pb and Cd in the CCHEN-MR-
002

ANOVA (95% confidence level)

Parameters Lead Cadmium
Between Between
Origin of the variations Within units Total Within units Total
units units
Sum of squares 0,024 0,169 0,193 0,013 0,174 0,187
Degrees of freedom 16 85 101 16 85 101
Average of the squares
0,0015 0,0020 0,0008 0,0020
(variances)
F calculated 0,74 0,41
Critical value for F 1,76 1,76

5.2. Calculation of the minimum mass for analysis

With the results obtained, and following with the example for Pb and Cd in the CCHEN-
MR-002, the minimum mass to be taken for analysis or the minimum mass at which the
homogeneity of the material with respect to the analytes of interest can be assured with a given
confidence level, were calculated following the procedures described before. Table VI shows the
results of these calculations.

TABLE VI. Minimum mass to be taken for the analysis of CCHEN-MR-002 at a given level of
heterogeneity

Parameters Lead Cadmium

Relative homogeneity constant I \v 30,44 %[mg"2] 6,86 %[mg"-|

Tolerance factor k' 2,233 2.233

Minimum mass at which homogeneity can be guarantee

For a variation coefficient due to a

5% heterogeneity 184,81 [mg] 9.39 [nig]

For a variation coefficient due to a

10% heterogeneity 46,2 [mg] 2.35 [mg]

17
 From Table VI it can be concluded that Cd is more homogeneously distributed as
compare to Pb. This means that one wants to determine Pb with a 5% contribution to the
uncertainty due to the heterogeneity of the material, it is necessary to take about 185 mg, while
for Cd is necessary to take 9,5 mg only.

5.3. Establishing assigned or certified values

Procedures for assigning or certifying levels in RMs or CRMs are very similar.
Assigning a value can only be performed by measuring the final content of the material, which
should be carried out by highly experienced laboratories using properly validated procedures.
Assigned values are more reliable when they are established by a group of laboratories using
different validated techniques.

The certification of reference material follows strict rules which are describes in detail in
a special ISO Guide [21] which states that "the certified value should be an accurate estimate of
the true value with a reliable estimate of the uncertainty compatible with the end use
requirements". The approach followed for certification may differ according to the objectives of
the study. Whereas pure compounds or calibrant solutions may be prepared gravimetrically, as
stressed above, the so-called "matrix CRMS", which are the most appropriate materials for
environmental related analyses, present much more difficulties to certifying bodies [3].
Certification may be carried out either within a single laboratory or institute, or through an
interlaboratory study.

5.3.1. Certification by a single laboratory approach

Two approaches may be followed for certification within a single laboratory:

Q by definitive method, or
• by two or more reference methods (applied by two or more independent analysts).
The laboratory should always be of proven proficiency. This single laboratory approach
does not eliminate the risk of systematic errors In addition, certification of matrix materials
using one single definitive method, e.g. IDMS, does not give the user who does not apply such
technique routinely a fair estimate of the uncertainty achievable by more classical methods.
Furthermore, such definitive method are limited in their field of application and still do not exist
for a variety of organic compounds, or elements with different oxidation states. For these
analytes an interlaboratory approach is the only satisfactory solution.

5.3.2. Interlaboratory certification studies

This should be done in the same way as an interlaboratory studies, but involving
laboratories with a recognised high quality performance. The organiser should also be recognize
as capable of organizing such studies and should provide the participants with detailed
instructions for participation and reporting forms (for results and method description). An
interlaboratory certification study should involve several methods applied by different
laboratories. This type of study for assigning values to the analytes of a material candidate to
reference material has proven to be very powerful and, in addition, helpful for detecting
systematic bias of a single technique.

The CCHEN-MR-002, has been analysed using this second approach. A total of 21
laboratories participated in the intercomparison study using different analytical methods, from
nuclear based ones to more conventional.

The evaluation and assignation of value to each analyte is another task that has to be
carry out very carefully. In addition to the purely statistical procedure to determine the most
probable values, it is necessary to evaluate the performance of each technique for the
determination of the respective analyte and to estimate the actual contribution of a given result tp
the "grand average". Tables VII shows the results reported for the determination of As and Cd in
the CCHEN-MR-002

TABLE VII: Results of the determination of As in the CCHEN-MR-002. Average are based on six
individual determinations

Lab.
Lab Method Sample average Abs. Rel. Std.
code code size (g) (Ug/2) std. dev. std. dev. (%) error

48 HAAS 2,0 1,20 0,01 0,9 0,01

33 INAA 0,5 5,40 0,76 13,9 0,31
15 INAA 0,22 5,75 0,32 5,5 0,13
45 INAA 0,14 5,79 0,49 8,5 0,20
17 ICPAES 0,5 5,83 0,05 0,9 0,02
36 ICPMS 0,2 5,93 0,17 2,9 0,07
25 INAA 0,35 5,97 0,27 4,5 0,11
30 TXRF 0,5 6,18 0,20 3,3 0,08
39 PIXE 0,35 6,23 2,83 45,4 1,16
26 INAA 0,2 6,30 0,24 3,8 0,10
50 HAAS 0,2 6,37 0,14 2,2 0,06
27 TXRF 0,36 6,44 0,26 4,1 0,11
40 ICPAES 0,5 7,06 0,47 6,6 0,19
TABLE VIII: Results of the determination of^fin the CCHEN-MR-002. Average are based on six
individual determinations '

Lab.
Lab Method Sample average Abs. Rel. Std.
code code size (g) (U2/g) std. dev. std. dev. (%) error

17 ICPAES 0,5 0,43 0,02 4,8 0,01

32 VOLT 0,5 0,45 0,11 25,1 0,05
22 FAAS 0,2 0,48 0,01 2,4 0,01
48 FAAS 2 0,50 0,02 4,2 0,01
14 SSAAS 0,001 0,50 0,01 1,6 0,01
36 ICPMS 0,25 0,50 0,0! 2,3 0,01
29 IDTIMS 0,2 0,55 0,02 4,1 0,01
20 FAAS 2,5 0,57 0,01 2,1 0,01
42 ZETAAS 0,2 0,59 0,01 1,3 0,01
43 ID-MS 0,5 0,66 0,01 1,1 0,01
11 FAAS 0,5 • 0,78 0,12 15,4 0,05

From this data, the final, value assigned to these elements as well as to others are finally
determined. However, one important parameter is rather difficult to establish when the assigned
value is determine in this way. Unless each laboratory determine a complete uncertainty budget
for their measurements, this parameter can be expressed as standard deviation or a confidence (at
a given confidence level) calculated from all the accepted or valid values.

CONCLUSIONS

Certified reference materials are fundamental for several actions in the analytical
laboratory. Most important, they help to achieve reliable and accurate results which are the basis
for environmental monitoring. Accurate and reliable results are needed for controlling and
supervising compliance with established regulations. The use of CRMs in laboratories ensures
the possibility of providing results which are also traceable to recognised international references
(standards) and therefore can be compared with the results from any other laboratory having the
same traceability.

Despite the high quality of available RMs from recognised international suppliers, in
many occasions it is not possible to find the most suitable RM for a given type of sample
matching, as close as possible, the matrix and the concentration levels of the analytes of interest.
This fact, in addition to the cost of internationally available RMs, purse analytical laboratories,

20
in particular in developing countries, to prepare their own RMs, complying with established
guidelines to be recognised and accepted by other laboratories.

However, this effort may not be fully successful since the RMs produced may not have
been characterised by a single laboratory or by a limited intercomparison study and may not be
traceable due to lack of a properly determined uncertainty. Nevertheless, developing countries
need, at least, reasonable well prepared and characterised control materials for their routine work
to cope with the every-day routine analysis and RMs for seldom calibration of the instruments,
re-evaluation of the analytical procedure and for checking the accuracy of the data.

REFERENCES

[1 ] COFINO, W.P., "Quality assurance in environmental analysis", in Environmental

Analysis: Techniques, Applications and Quality Assurance, Ed. D. Barcelo, pag. 359,
Elsevier, 1996
[2] MAIER, E.A., "Certified reference materials for the quality control of measurements in
environmental monitoring", in Environmental Analysis: Techniques, Applications and
Quality Assurance, Ed. D. Barcelo, pag. 383, Elsevier, 1996
[3] MARKERT, B., "Instrumental analysis of plants", in Plants as Biomonitors, Ed. B.
Markert, pag. 65, VCH, 1993
[4] ISO; International Vocabulary of Basic and General Terms in Metrology, second Edition,
International Organization for Standardization, Geneva, Switzerland 1993
[5] EURACHEM, Quantifying uncertainty in analytical measurements, First Edition 1995,
Second Edition 2000
[6] ISO, Uses of certified reference materials, ISO Guide 33, revised version, International
Organization for Standardization, Geneva, Switzerland (1989)
[7] ISO, Terms and Definitions Used in Connection with Reference Materials, ISO Guide
30, Internacional Organization for Standardization, Geneva, Switzerland (1981)
[8] REMCO/ISO, Directory of Certified Reference Materials, Secretary for REMCO, ISO,
Geneva, Switzerland
[9] Physicochemical Measurements: Catalogue of Reference Materials from National
Laboratories, Pure. Appl. Chem., 48 (1976) 503
[ 10] CORTES, E., PARR, R.M., CLEMENTS, S.A:, Biological and Environmental Reference
Materials for Trace Elements, Nuclides and Organic Micro-contaminants, A Survey,
IAEA/RL/128 (Rev. 1), IAEA, Vienna (1990)
[11] SCHRAMEL, P., SCHMOLK, J., MUNTAU, H., J. Radioanal. Chem., 50 (1989) 179
[12] GRIEPINK, B., COLINET, E., GUZZ1, G., HAMERS, I., MUNTAU, H., Fresenius Z.
Anal. Chem. 315(1983)20
[13] CHATT, A., RAO, R., JAYAWICKREME, C. K. Y MC DOWELL, L. S.,
"Determination of sampling constants for selenium in biological reference materials by
neutron activation", Fresenius' Z. Anal. Chem, 338(4) (1990) 399-407
[14] INGAMELLS, CO., SWITZER, P., Talanta 20 (1973) 547
[15] INGAMELLS, CO., Talanta 21 (1974) 141
[16] KURFURST, U., "Solid sample analysis", Springer, Alemania, pp. 61 - 66 (1998).

21
[ 17] PAU WELS, J., DE ANGELIS, L. Y GROBECKER, K-H., "Solid sampling zeeman
atomic absortion spectrometry in production and use of certified reference materials",
Pure & appl. Chem., 63(9) (1991) 1199 - 1204
[18] PAUWELS, J. Y VANECASTEELE, C , "Determination of the minimum sample mass
of solid CRM to be used in chemical analysis", Fresenius' Z. Anal. Chem., 345(2-4)
(1993)121-123
[ 19] ROSSBACH, M. Y GROBECKER, K-H., "Homogeneity Studies of reference materials
by solid sampling - AAS and INAA", Accreditation and quality assurance, 4 (1999) 498-
503
[20] SEPULVEDA, S., "Determinacion de la homogeneidad de muestras para analaisis
quimico y materiales de referenda", Thesis, University of Santiago,, Santiago. Chile
(2000)
[21 ] ISO, '"Certification of Reference Materials. General and Statistical Principles (ISO/IEC
Guide 35), International Organization for Standardization, Geneva (1985)

22