Accred Qual Assur (2000) 5: 224230

Q Springer-Verlag 2000 GENERAL PAPER

Paul De Bivre Traceability of (values carried by)
reference materials
Received: 11 November 1999
Accepted: 24 February 2000
Invited paper presented at the 2nd
Central European Conference on
Reference Materials (CERM.2), 910
September 1999, Prague, Czech Republic
P. De Bivre (Y)
Adviser on Metrology in Chemistry to
the Director of the Institute for
Reference Materials and Measurements,
European Commission JRC, 2440 Geel,
Belgium
e-mail: paul.de.bievre6skynet.be
Tel.: c32-14-851338
Fax: c32-14-853908
Abstract Traceability is a property
of the result of a measurement.
Since values carried by (reference)
materials must also have been ob-
tained, of necessity, by measure-
ment, the definition of traceability
also applies to reference materials.
It is extremely helpful to give the
traceability (of the origin) of a ref-
erence material a separate name,
i.e. trackability. An analysis of the
function of values carried by refer-
ence materials, shows that they can
fulfill different functions, depend-
ing on the intended use. One of
the functions located outside the
traceability chain and hence not
very relevant for establishing tra-
ceability is evaluating the ap-
proximate size of the uncertainty
of the measurement of an un-
known sample by performing a
similar measurement on a refer-
ence material, used as a simulated
sample. Another function is lo-
cated inside the traceability chain,
where the reference material is
used as an added internal stand-
ard. Then, the value carried by the
reference material is essential for
establishing the traceability of the
measured value of an unknown
sample. In the latter application,
the reference material acts as an
amount standard (the certified
value for amount is used).
Key words Traceability chain 7
Metrology in chemistry 7
Reference material 7 Amount
standard 7 Validation
Introduction
Traceability is defined in the International Vocabulary
of Basic and General Terms in Metrology, (VIM)[1] as
the
property of the result of a measurement or the val-
ue of a standard whereby it can be related to stated
references, usually national or international stand-
ards, through an unbroken chain of comparisons all
having stated uncertainties.
Traceability is therefore clearly a property of the
value of a result of a measurement. So what is a meas-
urement? The same VIM defines it as a
set of operations having the object of determining
the value of a quantity.
Again, the notion of value is very outspoken. Most-
ly, we obtain an unknown value by comparing the un-
known value to a known value (a simple definition of
how a measurement is carried out), i.e. measuring the
ratio of an unknown value to a known value.
We have now moved the problem of establishing the
traceability of one value to the establishment of the tra-
ceability of another known value, which in turn must
be compared to another known value, etc. This proc-
ess is quite acceptable, provided it stops somewhere. It
stops when we arrive at a value which we know because
we have defined it (consequently, it has an uncertainty
of zero). It is the value of the unit in which we want to
express the result of our measurement. Thus we can de-
fine a traceability chain as follows:
225
a traceability chain is a chain of values linked by
measurements which consist of comparisons of one
value to another value, ending in the comparison
with the value of the unit we have chosen to express
the result of our measurement.
At this stage it is important to stress the prime func-
tion of a reference material, i.e. to carry a value. It is
also important to remember that such a value, includ-
ing that carried by a reference material per definition,
results from a measurement. The long and still per-
sisting tradition of focusing on the material, should be
reoriented to focus on the value carried by the material.
All other characteristics of a reference material, essen-
tial as they may seem, draw their degree of importance
from the prime consideration that they must ensure and
safeguard the value carried, and that this value must be
delivered to the analyst in such a way that it is, and
remains, meaningful for the intended purpose (stabili-
ty, homogeneity, packaging, delivering sub-samples,
etc). Traceability of the value carried by a reference
material is therefore to be distinguished clearly from
traceability of a reference material. The traceability of a
reference material or of a sample has been called quite
appropriately trackability (to the place or the produc-
er it comes from) [2, 3].
The key elements of a traceability chain, values and
links between values, have already been described [4].
More complete traceability chains are presented in
Fig. 1. The symbols used are b p amount content [5, 6]
in amount (mol) per mass (kg) of element (or com-
pound) E in material X. Note that the chain is consti-
tuted by values linked by operations called measure-
ments, defined as above. The analyst could attempt the
establishment of a complete traceability chain as shown
(Fig. 1, left chain), but that would require a huge
amount of work, or may not be possible, e.g. because
the chain may be broken (in the upper part under
chemical operations). The same reasoning applies to
the value carried by a reference material (central chain
or right chain in Fig. 1). Every time we use a reference
material, two traceability chains are involved as illus-
trated in Fig. 1: one for the measurement result ob-
tained on the unknown sample (left chain), and one for
the value carried by the reference material, (either the
central or the right chain). They must, by their very na-
ture, be similar. The first one must be demonstrated by
the analyst. The second one must be demonstrated by
the reference material producer. They sell the product
and therefore must be accountable for the product.
The traceability chain of the value of a (certified)
reference material
The traceability chain of a (certified) value is illustrated
for two cases, i.e. for two different uses of a reference
material, illustrated by the central and right chains in
Fig. 1. The chains for both reference materials go via a
purity investigation of the element/compound which is
incorporated gravimetrically in the matrix. The result-
ing value takes into account the purity (with evaluated
uncertainty) at point 2, a statement (or studies) of stoi-
chiometry (with evaluated uncertainty) at point 3 and,
finally, the conversion from the mass of a pure sub-
stance (PS) into an amount at point 4. A full traceabil-
ity chain then ends in the value 1 of the mass (m) of the
unit kg as shown.
(In the future, it will end in the value of the Avogadro
constant and the value 1 of the atomic mass unit u de-
fined as the mass of 1/12th of the mass of the
12
C atom.
This will happen when the definition of the kg (now the
mass of the prototype of the kilogram) will have
changed into the mass of a number of
12
C atoms, i.e. of
the mass of {N
A
}7m(
12
C). 1000/12).
Alternatively, at point 2 in Fig. 1, the reference ma-
terial producer could have chosen another traceability
route, e.g. a coulometric measurement of the number of
atoms of the element under investigation, which would
have ended in the product I7t (electric current times
time) and hence in amperes and seconds. Ultimately,
the end of this traceability chain is then located in the
value of the Avogadro constant and the value of the
charge of the electron [7].
In many chemical measurements, however, we
(must) use measurement scales which are based on val-
ues carried by commonly agreed reference materials.
These enable the analyst to establish a calibration
curve through which observed signals from an instru-
ment used in a measurement procedure, are converted
to a concentration [8]. Any measurement result which
is shown to be traceable to such a common scale, is
comparable to any other measurement result which is
traceable to the same scale, i.e. established by the val-
ues of the same reference materials.
We now have a reference material at hand carrying a
traceable value. How do we use it in an actual measure-
ment process applied to an unknown sample?
Using the value, carried by a reference material, in the
traceability chain of an unknown value in a sample:
1. Internal standard method (value is located inside
the traceability chain)
The value measured for the unknown sample is com-
pared at some point to the value of an internal stand-
ard added to the sample (left chain, point 1 in Fig. 1).
That value comes from a reference material carrying
that value, manufactured by a producer (right chain,
point 1 in Fig. 1). Our analyst relies on the value sup-
plied by the producer of the reference material from
this point(identified by HELP from (amount stand-
226
Fig. 1 Traceability chains of: a) the result of a measurement val-
ue obtained on an unknown sample; b) a value carried by a (ma-
trix) reference material intended to reduce the uncertainty caused
by chemical operations on the unknown sample; c) of a value
carried by a reference material intended to be part of the tracea-
bility chain of the measured value in an unknown sample
227
ard) AS in Fig. 1). It is the responsibility of the refer-
ence material producer to deliver or, at least, to state
to the analyst the value of the reference which the ana-
lyst wants/needs to state as a reference. This includes
the establishment by the reference material producer of
a traceability chain for the value offered to the ana-
lyst.
(Logically it follows that a traceability chain can
equally be stated to the value of a reference measure-
ment obtained by a reference measurement producer on
the analysts sample at point 1 in Fig. 1. This concept has
the unknown sample travelling from the analyst to the
(reference measurement) producer, rather than having
the (reference) material travelling from the (reference
material) producer to the analyst. This approach will be
developed further on another occasion.)
The role of the reference material producer is to re-
duce the burden of work for the analyst, by supplying
the traceability chain of the known value of the refer-
ence material against which the analyst wants to meas-
ure the unknown value for an unknown sample. This
approach relieves the analyst from, e.g. measuring the
purity of AS at point 2 in the right traceability chain of
Fig. 1, or from doing the painstaking stoichiometric
measurements on AS at point 3 in Fig. 1, right chain. It
relieves the analyst from having to do the conversion
from mass to amount at point 4 in Fig. 1 (which re-
quires measurement of the isotopic composition in or-
der to have a valid molar mass at hand). In short, it
relieves the analyst from establishing the traceability
chain of the measured value itself, because a stated
reference can be bought from a reference value pro-
ducer. The analyst can literally jump from the left chain
on to the chain of the producer, buy the value and in-
corporate the purchased value into the traceability
chain on the left by adding a portion of the reference
material to the unknown sample.
The name amount standard (AS) is given to a ref-
erence material which has this essential function in the
traceability chain. It is one of several possible functions
of the value of a reference material.
We will now look at the traceability chain of the val-
ue carried by a matrix reference material, in order to
understand its function in the measurement process.
That is needed to produce reference materials which
are adequate for the intended use (fit-for-purpose).
2. Exploiting the value carried by a matrix reference
material in the measurement of an unknown value
in a sample (value is located outside the
traceability chain)
This is illustrated in Fig. 2. Prior to the measurement,
the value in the sample (left chain) may have had to
undergo a recovery (extraction, distillation, dilution,
or otherwise, summarized in the upper part of Fig. 2
under chemical operations). The question must
therefore be asked: did this operation change the value
in an unknown fashion thus putting the unbroken
chain under suspicion of having been broken? Simi-
larly, before the recovery operation, the value in that
sample may have had to be digested. Again the ques-
tion must be asked: did the digestion change the value
in an unknown fashion thus putting the unbroken
chain under suspicion of having been broken? And so
on. A difficult point is where the analyst calculates the
value of the claimed quantity at the end of the meas-
urement, from the value of the quantity actually mea-
sured. Equation 1 shows that a conversion factor
K(
a
E,X) is needed:
b(
a
E,X)
sample
pK(
a
E,X)7b(
a
E,X)
sample obs
(1)
To this effect, in analytical chemistry it has been
good practice that the analyst obtains a reference mate-
rial with a matrix Z, similar but of course not identical
to the one of the material X which needs to be mea-
sured. By performing the same operations on a sample
of matrix reference material Z as performed on the
sample with matrix X, an estimate can be obtained of
the overall correction factor K(
a
E,Z). The value of the
amount content b(
a
E,Z)
RM cert
of the reference material
as supplied by the reference material producer is
known. The value b(
a
E,Z)
sample obs
is observed by the
analyst Hence, a correction factor can be calculated
K(
a
E,Z) for losses during digestion and recovery etc.
(the chemical operations in Fig. 2) as determined
with the help of the reference material. It can be ap-
plied to the measurement on the unknown sample. In
short, this process can be described by Eq 2:
K(
a
E,Z)pb(
a
E,Z)
RM cert
/ b(
a
E,Z)
RM obs
(2)
As shown in the central chain, upper part in Fig. 1,
(HELP from matrix RM),the analyst can now use the
correction factor from Eq 2 and substitute it for the un-
known correction factor K(
a
E,X) in Eq 1. This enables
a correction to be made of the value observed in the
unknown sample. In summary: the analyst simply sub-
stitutes K(
a
E,X) for K(
a
E,Z).
K(
a
E,X)pK(
a
E,Z)?
However, the uncertainty of this substitution must
be evaluated.
The evaluated uncertainty of this substitution, i.e.
the uncertainty of K, will limit the uncertainty of the
measurement result of the unknown sample and must
be taken into account as a contribution to the total un-
certainty budget of the result obtained on the unknown
sample since:
a. the matrix of both samples is different.
b. the measurement conditions during the measure-
ment of the reference material and of the unknown
sample, are not identical.
228
Fig. 2 A matrix reference ma-
terial can be used to simulate
chemical operations and esti-
mate from that simulation a
collection factor. The uncer-
tainty of that collection factor
must be evaluated and taken
into account in the uncertainty
ludget
229
This uncertainty is the quantitative result of the fact
that the analyst had to jump from one traceability
chain to another.
Thus the matrix reference material has fulfilled an-
other function than the AS function for the analyst, a
function which is not located in the traceability chain,
but outside the chain: it enables the analyst to make an
independent assessment of the possible magnitude of
the conversion factor K, thus assessing possibly re-
ducing the uncertainty of the measurement by carry-
ing out a correction. But even this correction carries
an uncertainty which must be evaluated. The problem
of this correction factor has been treated elsewhere in
more technical detail under the name recovery factor
[9].
From the above, we can now logically formulate re-
quirements for traceability chains.
Requirements for traceability chains
For the value obtained by the measurement of an
unknown sample:
a transparent formulation of what is to be measured:
identification of the measurand (related to the aim
and intended use of the measurement result).
a transparent formulation of the traceability chain
wanted, including a clear and simple formulation of
the stated reference in the chain to which traceabil-
ity of the measured value will be claimed. In other
words, establishing a traceability chain is an a priori
requirement, i.e. it must be planned before the meas-
urement. It is not the result of a measurement.
1. For the value carried by a reference material, and
used as an AS (pure elements, pure compounds):
a transparent formulation of what will be claimed
(related to the intended use of the AS).
a transparent formulation of the measurand in the
AS.
a transparent formulation of the AS in the matrix
reference material.
establishment of the proper equation which relates
what is actually measured to what is purported to be
measured.
a transparent formulation of the traceability chain to
the values of a common measurement scale, i.e. to
the value of the unit [possibly to the value of a SI
unit or even to the value(s) of (a) fundamental con-
stant(s)].
The base must be the same as for the measurement
result if we want all our measurement results to be co-
herent.
2. For the value, carried by a reference material, and
used to simulate the measurement of an unknown
sample (most matrix reference materials):
a transparent formulation of what will be claimed
(related to the intended use of the matrix reference
material).
a transparent formulation of the measurand in the
matrix reference material.
a transparent formulation of the matrix of the refer-
ence material.
a transparent formulation of the traceability chain to
a common measurement scale, i.e. to the value of a
common unit [possibly to the value of a SI unit or
even to the value of (a) fundamental constant(s)].
The base must be the same as for the AS if we want all
measurement results to be coherent.
Conclusions
1. There are different categories of reference materials
according to the function they carry out in the meas-
urement process:
those which perform a function in the traceability
chain because the unknown value can be directly
measured against their value; examples are: an
amount n(E,X) of another element, added to the
unknown sample; the case of traceability of mass
values described elsewhere [10].
those which perform a function outside of the tra-
ceability chain because they are used to estimate
an approximate conection factor (with its uncer-
tainty), for steps in the measurement of an un-
known sample for which the degree of quantitativ-
ity must be evaluated, i.e. for which the size of
the breakage of the traceability chain must be esti-
mated (as it cannot be determined directly).
2. Establishing the traceability chain of a result of a
measurement is too long and too expensive a proc-
ess for an analyst to carry out; the analyst must be
able to obtain ASs from reference material produc-
ers and/or National Measurement Institutes (NMIs),
the values of which can be used as stated refer-
ences; that puts the burden of such a programme
(with justifications) on producers and NMIs.
3. Reference material producers and NMIs must pro-
vide values, which can perform two functions:
a function as part of the traceability chain because
they provide values in that chain (AS function).
the function of estimating the correction factor
(with uncertainties) for difficult or unknown steps
in the traceability chain, by creating the possibility
to simulate more or less the measurement process
of an unknown sample on a known sample; this
function lies outside the traceability chain.
230
4. Two traceability chains are involved each time a ref-
erence material is used to simulate a measurement
of an unknown sample. It is therefore useful to draw
a picture of the two chains (Fig. 2, left and right
chain): it enables one to see immediately (right half
of Fig. 2) that the simulation measurement is in fact
the measurement or an estimation of a correction
factor, at all times with associated uncertainty.
From the above, it becomes clear that a better un-
derstanding of traceability of values carried by refer-
ence materials requires a better understanding of the
measurement process to which that concept is applied
and in which the reference material is used. Not sur-
prisingly, this leads to a better identification of possible
functions of (the values of) reference materials in the
measurement process as well as in calibration [7].
It also becomes much clearer now that traceability is
a property of a value of a measurement result, or of a
value carried by a reference material. Traceability is
not related to a method, an instrument, a material or to
a (national) measurement institute, but always to an-
other value. Trackability on the contrary, is a property
of a reference material (to its origin), or of a method
(to its physico-chemical principles) or of an instrument
(to its source). A convention on this subject would help
dispel misunderstandings in international discussions.
Traceability is not an end in itself. It is the very way
of achieving what we all want and need: comparability
of the results of our measurements. Making our results
comparable so that we can indeed compare them, is a
matter of measuring these results against the same
scale. Results do not need to be concordant [11], nor
do they need to be of the same magnitude in order to
be comparable. They can have quite different magni-
tudes, but they must have been measured against the
same scale, whatever the realization of that scale.
The task of each analyst is to demonstrate the tech-
nical/scientific authority of the measurement result in
order to lead to the necessary credibility. This means
that underpinning (Untermauerung) of the measure-
ment result is necessary and leads to the necessary
credibility. That is why the analyst needs to demon-
strate the traceability of the value claimed.
The above clarification of the traceability concept is
intended to give the analyst an underpinning to what he
has been is doing anyway for a long time.
Acknowledgements The author would like to acknowledge crit-
ical remarks from A. Fajgelj (IAEA) in the final phase of drafting
this paper, as well as the very stimulating and fertile environ-
ment provided by the staff and the visiting scientists at the IRMM
laboratories working on Metrology in Chemistry under the guid-
ance of Dr. P. Taylor.
References
1. International Vocabulary of Basic
and General Terms in Metrology
(1993) BIPM, IEC, IFCC, ISO,
IUPAC, IUPAP, OIML. ISO,
Geneva
2. Fleming JW, Neidhart B, Wegscheid-
er W (1996) Accred Qual Assur 1: 35
3. De Bivre P (1996) Accred Qual
Assur 1: 109
4. De Bivre P (1999) Accred Qual
Assur 4: 342346
5. IUPAC (1993) Quantities, units and
symbols in physical chemistry. Black-
well Science, Oxford
6. Cvitas T (1996) Metrologia 33: 3539
7. De Bivre P (2000) Accred Qual As-
sur (in print)
8. De Bivre P, Kaarls R, Peiser HS,
Rasberry SD, Reed WP (1997)
Accred Qual Assur 2: 168179
9. IUPAC (1999) Pure Appl Chem
71: 337348
10. De Bivre P, Taylor PDP (1997)
Metrologia 34: 6775
11. Papadakis I, De Bivre P (1997)
Accred Qual Assur 2: 347