Elemental solution standards

can be produced that are

almost “clones” of NIST SRMs.

TRACEABILITY
of Single-Element
Calibration Solutions
ommercially available single-element solutions are used of measurement results to SI ensures their comparability over

C to calibrate the instruments with which millions of ele-

mental determinations are performed each day. Atomic
spectrometric techniques rely on calibration against these solu-
time and space.
To provide suitable references for the millions of atomic spec-
trometric determinations made daily, the National Institute of
tion standards to determine the tens of elements in hundreds of Standards and Technology (NIST) disseminates elemental solu-
samples in thousands of labs. The solutions are, or are prepared tion standards through its standard reference materials (SRMs)
from, certified reference materials (CRMs), which are the basis program; SRMs are CRMs distributed by NIST. We will describe
for the accuracy of the measurement results. A certificate that as- the concept of traceability and one approach by which traceabil-
serts a claim of “traceability” usually accompanies these CRMs. ity to these references can be achieved by directly comparing so-
The claim of traceability aims to ensure that the certified lutions against them. The approach we describe is based on work
value of the CRM is linked to appropriate national and inter- from our laboratory using a method we refer to as high-per-
national references, which are presumably stable and accurate formance inductively coupled plasma-optical emission spectrom-
and of higher metrological order. Properly established, this etry (HP-ICP-OES; 1–4).
linkage couples the information content of the traceable value
to the information content of the reference, and it can add im- Traceability
mensely to the acceptance of the CRM. This linkage also con- The concept of traceability predates our contemporary global
nects the measurement results calibrated from the traceable metrology system. Originally, the king’s foot was a measure of
value of a CRM to the larger network of measurements, which length; then, traceable artifacts the same length as his foot were
ultimately stems from the International System of Units (SI). disseminated to permit measurements to be made with ease, thus
The maintenance of the base and derived units of SI, together saving the king’s foot a lot of wear and tear. The International Vo-
with traceability to SI, underpins contemporary metrology. SI cabulary of Basic and General Terms in Metrology defines trace-
provides a uniform, stable measurement scale, and traceability ability as the “property of the result of a measurement or the value

136 A A N A LY T I C A L C H E M I S T R Y / A P R I L 1 , 2 0 0 5 © 2005 AMERICAN CHEMICAL SOCIETY

 Marc L. Salit
Gregory C. Turk
National Institute of Standards and Technology

of a standard whereby it can be related to stated references, usu- that reference is itself traceable to an external reference (to which
ally national or international standards, through an unbroken other references may also be traceable), then traceability to that
chain of comparisons all having stated uncertainties” (5). This external reference may be established. In this manner, measure-
definition, together with agreed-upon units of measure, ensures ment results can then be compared among laboratories. This net-
that a traceable result (a value and its associated uncertainty) work rises from standards held at individual laboratories through
measured in the absence of bias can be compared with other those held at national and international levels.
traceable results, also measured in the absence of bias. Such re- Comparability between national and regional measurement
sults may arise from comparison to the value of the same stable systems is being established in part through the 1999 Comité
artifact in different places at different times, or from comparison International des Poids et Mesures (International Committee for
to the values of different artifacts, whose values are also traceable. Weights and Measures; CIPM) multilateral mutual recognition
Comparability of the results is ensured by traceability. The arrangement. This agreement, signed by the 38 member nations
ability to compare a result to other results, expectations, or spec- of the Meter Convention, was a response to the need for an
ifications is essential. A measurement result that cannot be com- open, transparent, and comprehensive scheme to give users reli-
pared is of no value. Even if results are to be compared only able quantitative information on the comparability of national
within a given laboratory, those results are useful only by virtue metrology systems (www.bipm.fr/enus/8_Key_Comparisons/
of their values relative to other results. mra.html). Through this horizontal comparability between
The extent of space and time over which results can be com- nations, the vertical traceability of a measurement to a national
pared depends on the scope of the reference to which they are reference may link it to international comparability, which is po-
linked or to which they are traceable. Local traceability ensures tentially important for international trade, commerce, and regu-
local comparability. A measurement result traceable to a reference latory affairs, as well as global science.
held in a particular laboratory can be meaningfully compared Of increasing relevance to chemical analysis is laboratory ac-
only to other measurement results traceable to that reference. If creditation based on International Organization for Standard-

A P R I L 1 , 2 0 0 5 / A N A LY T I C A L C H E M I S T R Y 137 A
ization Guide 17025 standards (4), and the characterization
that require a clear demonstra- of an LiAlO2 ceramic material
tion of traceability. In the Unit- (2). The method was used in
ed States, government regula- an international key compari-
tions or laboratory quality son (K8: Monoelemental Cali-
systems often require that the bration Solutions) of the
results of an analysis be “trace- Comité Consultatif pour la
able to NIST,” an ill-consid- Quantité de Matière (Consul-
ered phrase that, more properly tative Committee for the
stated, means “traceable to the values on NIST measurements or Amount of Substance) of the CIPM, with excellent results that
standards.” NIST policy on traceability explicitly states: “Other validated it against a number of other analytical techniques (6).
organizations are responsible for establishing the traceability of A similar approach using ICPMS for the multielement analysis of
their own results or values to those of NIST or other stated ref- environmental CRMs has been reported (7 ).
erences” (www.nist.gov/traceability). Additionally, the policy Pairing of the 3100 series with a rugged and practical com-
states that NIST “[a]sserts that providing support for a claim of parison method enables the manufacture of traceable calibra-
traceability of the result of a measurement or value of a standard tion solutions. This traceability linkage couples the information
is the responsibility of the provider of that result or value, content of the SRM to the traceable solution. Thus, a manu-
whether that provider is NIST or another organization, and that facturer of a CRM could leverage this approach, relying on
assessing the validity of such a claim is the responsibility of the comparison against the SRM, and avoid duplication of the
user of that result or value.” This caveat warns that the users of work done at NIST. The 3100 series solutions are themselves
materials that claim traceability must be able to evaluate those traceable to the mole, the SI unit for chemistry. The well-char-
claims, and thus it gives tremendous importance to the trans- acterized primary materials, against which the SRMs are com-
parency of a traceability claim. pared, are realizations of the mole for the elements. Compari-
The analyst must recognize that traceability only requires two son against the SRM can yield a value for a CRM, also now
things: a suitable reference to compare to and a measurement traceable to the mole. Production of a rhodium solution CRM
method for the comparison. Of course, an essential part of a is a good example. Rhodium metal is extremely difficult to dis-
traceable result is its uncertainty, which must consist of the un- solve, and soluble salts of rhodium are difficult to assay with
certainty of the value of the reference combined with the uncer- small uncertainty. A commercial or in-house CRM solution
tainty of the comparison. An assessment of traceability evaluates could be prepared from a soluble salt, and a traceable value de-
the validity and suitability of the reference, the appropriateness rived by comparison against SRM 3144, the rhodium standard
and reliability of the comparison, and whether the uncertainty of solution SRM (8, 9).
the traceable value is valid. Unquestionably, the traceable mass fraction of analyte in the
CRM could not have smaller uncertainty than the SRM. The
Elemental solution standards 3100 series solutions are optimized to have stable mass fractions
In the United States, designating a calibration solution for ele- with small uncertainties. However, the quality of a calibration so-
mental analysis as traceable usually means that the certified value lution is derived from a number of parameters, not only the un-
of that solution is traceable to the certified value of the matching certainty of the mass fraction. Solution purity, stability, packag-
elemental solution from the NIST 3100 series of SRMs. These 69 ing, and concomitant anions all play a role in fitness for purpose,
single-element solution SRMs serve as national standards for ele- or quality. A traceable CRM could be of either higher or lower
mental solution mass fraction. These SRMs are prepared in bulk quality than the SRM.
and are batch-certified for the mass fraction of the constituent el- For example, many of the 3100 series solutions are packaged
ement (typically 10 mg/g). The target range for the relative un- in glass ampoules that enhance stability by limiting transpiration
certainty in the certified mass fraction for these SRMs is ±0.3%. but could induce contamination. A CRM could have a certified
As part of our process to certify the 3100 series, candidate value traceable to the SRM and have higher purity, making it
SRM materials are compared against well-characterized primary more suitable for multielement analysis. Or, a solution CRM
materials. Through this comparison, traceability of the SRM val- might be prepared from a poor grade of material containing
ues to the SI is established. HP-ICP-OES has been our compar- many impurities. The traceable certified value might be excel-
ison method of choice for this process since 1997. This method lent—accurate, with small uncertainty—yet the material may be
uses unmodified, readily available, mature, reliable commercial inappropriate for its application and of low quality.
instrumentation; it has been demonstrated to have very small
comparison uncertainty, ruggedness with respect to bias (small How to do it
changes in conditions do not bias the results), economy, suffi- HP-ICP-OES exploits two strengths of array-detector-based
cient linear dynamic range, and a high degree of inter-element spectrometers: simultaneous measurement of the analyte and in-
selectivity. The HP-ICP-OES method has been applied to the ternal standard, and time-correlated off-line background correc-
analysis of single-element solutions for 64 elements (3), the cer- tion (3). Moderately high analyte mass fractions and signal levels
tification of the major constituents of a high-temperature alloy provide practical immunity from blank and spectral interferences,

138 A A N A LY T I C A L C H E M I S T R Y / A P R I L 1 , 2 0 0 5
and a drift-correction approach mitigates low-frequency noise. recorded. These solutions should be well within reasonable oper-
These strategies typically yield results with variability <0.1%, ating levels for the instrument; for most elements, 10 µg/g is ap-
which is on the order of that associated with sample handling. propriate. Once the sensitivity ratio is established, a mass ratio
Sample-handling variability is quantified with an experiment that (mass of analyte per mass of internal standard) that yields unity in-
includes the analysis of replicate preparations. The HP-ICP-OES tensity ratio can be calculated. Then, an internal standard spike so-
method, or one like it using a different measurement tool for lution of appropriate mass fraction can be prepared.
comparison, might be used by those wishing to prepare solutions Once the spike is added to the analyte, the mass ratio is estab-
traceable to NIST SRMs—in particular, by commercial CRM lished. This is the only sample preparation data value used in cal-
manufacturers whose products are designed to ensure the accu- culating the result, and it is unaffected by any subsequent dilu-
racy and comparability of most elemental analysis results. tions. Typically, the spiked samples are diluted when introduced
We have developed a spreadsheet tool that leads the analyst to the plasma, in order to yield an appropriate signal level that is
through the HP-ICP-OES comparison to establish traceability of comfortably within the working linear response range of the in-
the value of a CRM to the value of a NIST 3100 series SRM. The strument and sufficiently larger than any anticipated blank or in-
essential elements of this comparison include a fixed experimental terference level. This dilution can be performed by eye using a
design; a method development experi- digital pipette, and it is not necessary
ment that establishes the sensitivities to record any values associated with
for the analyte and an internal stan- A measurement result this dilution.
dard; a gravimetric sample preparation All masses whose uncertainties
scheme; a fixed set of emission intensi- that cannot be compared propagate to the results are >2.5 g.
ty measurements of both elements; and This amount permits the use of com-
to other results,
the calculation of the traceable value mon top-loading balances, which typ-
and its uncertainty. The spreadsheet expectations, or specifications ically have 1.5-mg-repeatability stan-
also includes useful diagnostic charts dard deviations. This keeps the relative
is of no value.
that permit visualization of the instru- uncertainty of weighing to <0.1%,
ment drift and the noise correlation be- which is on the order of our ability
tween the analyte and internal stan- to prepare duplicate preparations.
dard. The spreadsheet tool, an example data set, and instructions Aliquots of ~2.5 g are used for both the SRM and the test sam-
are available in Supporting Information. ples. The scheme assumes that the nominal mass fractions of the
Four separate preparations are made of the desired element SRM and the test samples are within an order of magnitude of
from the 3100 series SRM and the single-element solution to be each other. Aliquot masses are adjusted to deliver the same mass
compared (the test sample). The 4 preparations of the SRM are of analyte from each solution.
designed to consume ~10 g of solution (many of the 3100 series An internal standard solution is prepared so that the spike
solutions are deployed in packages of 5 single-use 10-g glass am- masses are ~20 g. The mass fraction of this solution need not be
poules). Ideally, the four preparations of the test sample are se- accurately known, as only the relative masses of spike weighed
lected from different portions of the batch, so the population of into the reference and test solutions appear in the calculation of
results from the test sample will include a measure of any het- results. The mass used need not be precisely 20 g, and the solu-
erogeneity. This replication permits evaluation of several uncer- tion is typically delivered with a graduated cylinder. The 20 g of
tainty components associated with the preparation of the solu- spike solution is easily weighed with negligible uncertainty and
tions and acts to validate the method and its results. The analyst typically dilutes the high-mass-fraction (10-mg/g) SRM. When
must take proper care to develop the instrumental method two solutions with high mass fractions are mixed, a precipitate
(wavelengths, spectral-background-correction approach and pa- might form; dilution helps mitigate this. Final dilution from the
rameters, and integration times) such that the measured emission spiked sample is also facilitated by this initial dilution. Moreover,
intensities have high S/ N and selectivity. A useful reference for the lower the mass fraction of the spike, the less the relative loss
method development is the study reporting HP-ICP-OES analy- (from missed droplets, evaporation, or aerosolization) during so-
ses for 64 different elements (3). lution handling.
Sample prep. Analyte and internal standard mass fractions are Intensity measurements. Intensity measurements are per-
selected to yield an emission intensity ratio near unity. For some formed sequentially on the eight samples, alternating the SRM
ICP-OES instruments, this facilitates simultaneous integration, in and test samples (the run order is SRM sample 1, test sample 1,
which time is used to extend the dynamic range of a spectrometer, SRM sample 2, test sample 2, etc.). This sequence of measure-
and limits the impact of deviation from linear intensity response. ments is repeated five times. Each is typically the mean of sever-
Because the analyte and internal-standard spectral lines are like- al replicate intensity measurements. The short-term noise of the
ly to have different sensitivities, a preliminary experiment is per- intensities, expressed as the relative standard deviation of the
formed to roughly determine the sensitivity ratio. Solutions of the replicate measurements, is a good diagnostic measure of the sam-
analyte and internal standard, whose mass fractions are approxi- ple input performance. The short-term noise of the ratios, and its
mately known (10% uncertainty is adequate), are measured under improvement over the intensity noise, helps to differentiate
the selected operating conditions, and the emission intensities are among the sources of measurement noise; the correlation of an-

A P R I L 1 , 2 0 0 5 / A N A LY T I C A L C H E M I S T R Y 139 A
 Table 1. Sensitivity test.
Target
Sensitivity working
solution Measured mass
mass emission Sensitivity fraction Sensitivity
fraction (µg/g) intensity [(µg/g) –1] (µg/g) ratio
cate measurements of the 8 samples took place in <2 h. Five re-
Cobalt 10 1,670,869 167,086.9 5.98491 0.037901 peat integrations of 8 s each were measured for each replicate
Scandium 5 22,042,376 4,408,475 0.226836 measurement of each sample. Total photon integration time was
26 min 40 s. At every sample change (40 of them), 90 s elapsed
to permit the new solution to traverse to and equilibrate in the
alyte and internal standard signals is an indicator of performance spray chamber and reach the plasma. The analyte and internal
and sets the expectation for uncertainty. standard signals were highly correlated. A sevenfold reduction in
Calculations. The straightforward calculations are based on the noise in the ratio compared to the analyte signal can be ob-
Equation 1 in Reference 6. Intensity ratios are calculated for each served in Figure 1.
repeat measure of each solution; these ratios are drift-corrected A straight-line calibration model is fitted and forced through
and summarized for each sample. Mass ratios are calculated from zero for the four SRM solutions. The slope of this line is esti-
the preparation data, and the ratio of the intensity ratio to the mated as the mean of the slopes calculated for each of the four
mass ratio is calculated for each sample. A single calibration factor solutions, resulting in a single calibration factor. Mass fractions
is determined from the summary of the SRM measurements and for all solutions (treating each SRM solution as an unknown) are
is used to calculate the traceable mass fraction of the test sample. then calculated from this calibration factor; Figure 2 presents re-
The uncertainty budget includes components for dispersion sults before and after drift-correction and reference lines for the
of measurements of the SRM and the test sample (due to vari- mean mass fraction and uncertainty of the SRM. The center of
ability of the intensity ratio measurements and preparation and the gray reference bar is the reference value for our synthetic test
weighing) and for the uncertainty of the certified value of the sample, and the thickness is its uncertainty. The uncertainty of the
SRM. A coverage factor is selected to expand the standard un- SRM is excluded from all results graphed for the test sample to
certainty to a 95% confidence interval. allow only the comparison to be assessed. The dashed black lines
represent the mean and standard uncertainty of the HP-ICP-OES
Test experiment results (9.950 ± 0.0022 mg/g before drift-correction and 9.950 ±
Two ampoules of cobalt SRM 3113 (lot 000630, certified mass 0.0026 mg/g thereafter).
fraction 9.996 ± 0.024 mg/g) solution were selected to demon- Several observations can be made directly. When the data for
strate the approach. One of these ampoules was diluted by ~0.5% the individual solutions are drift-corrected, the dispersion is
(45.76 mg of 2% volume fraction HNO3 was added to 10.30459 smaller. However, the dispersion among the results of the four
g of SRM solution) to prepare the synthetic test sample. The test solutions (for either the SRM or the test sample) is essentially un-
sample mass fraction was 9.952 ± 0.0043 mg/g. Cobalt mass affected by drift-correction. The dispersion among the four
fraction was determined by HP-ICP-OES comparison of the di- preparations of a solution dominates the dispersion of the mea-
luted sample against the unadulterated SRM. Details of the ex- surements. This preparation effect has been documented and
periment can be found elsewhere (3). Table 1 reports the results arises from uncertainty in solution handling and weighing, limi-
of the sensitivity test: the mass fractions of the sensitivity solu- tations of the balance, and evaporation of the solution while it is
tions, measured intensities, sensitivities,
working mass fractions (nominal mass 1.0
fractions of analyte and internal stan-
dard to be introduced to the plasma),
and sensitivity ratio. The working mass 0.8
fractions were calculated to yield an op-
erating signal level of 106 counts/s.
The target and actual weights for 0.6
the eight solutions prepared can be
found in Supporting Information. The
RSD

target weights are based upon the sen- 0.4

sitivity test, an SRM aliquot of 2.5 g,
and a minimum spike weight of 20 g.
Target weights are nominal values. 0.2
Variations of as much as tens of percent
are well within expected norms when
0.0
the solutions are prepared. Typical vari-
ability is observed in these data. Ap-
proximately 60 g of solution were pre-
pared for analysis. Based upon the Intensity Intensity ratio
working mass fraction, the nominal
weight of spiked solution to dilute to FIGURE 1. S/N improvement expressed as relative standard deviation with time-correlated
60 g is calculated as 0.32 g. The 5 repli- internal standardization.

140 A A N A LY T I C A L C H E M I S T R Y / A P R I L 1 , 2 0 0 5
being weighed. The mag- 10.02
nitude of this effect is
Drift-corrected
~0.03% (3). The result for
the synthetic test sample 10.00
from the comparison is
Mass fraction (mg/g)
consistent with the refer-
ence value. 9.98
The uncertainty budget
before and after drift-cor-
Drift-corrected
rection for the comparison
is presented in Table 2. 9.96
Reference value for
This simple yet rigorous test sample
budget includes compo-
nents of uncertainty for the 9.94
measurement of the test
sample and SRM 3113.
These standard uncertain- 9.92
ties are evaluated as the 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
standard deviation of the SRM SRM Test sample Test sample
mean of the results from
the four preparations. The FIGURE 2. Results of test–experiment comparison.
variability of the intensity
ratios is convolved into the dispersion of the replicate prepara- sult for the test solution. The results of the test experiment
tions, as is the variability of preparation. This “method-valida- demonstrate that establishing traceability can be done without
tion” approach to uncertainty evaluation is a useful feature of the inflating uncertainty and in less than half a workday. HP-ICP-
experiment, bundling together the evaluation of individual un- OES is an effective tool to link the value of a test solution, com-
certainty components into evaluation via replication. mercial CRM, or in-house standard to global comparability.
The dominant uncertainty in the traceable value calculated When traceable materials are used to calibrate a valid method,
from the comparison arises from the uncertainty of the certified the results of analyses using that method can be compared with
value of the SRM. This value is ~4
larger than the uncertainty other results, thereby giving them real value.
due to the comparison. It is a general rule of thumb that uncer-
tainty components 3
smaller than the largest components can Marc L. Salit is a research chemist and Gregory C. Turk heads the spec-
be regarded as insignificant (11). This rule of thumb applies here: trochemical methods group at the National Institute of Standards and
We observe an insignificant difference between the uncertainty Technology. Address correspondence about this article to Turk at
of the certified value of the SRM and the uncertainty of the re- [email protected].

Table 2. Results and uncertainty budget. References

(1) Salit, M. L.; Turk, G. C. Anal. Chem. 1998, 70, 3184–3190.
Before drift-correction After drift-correction
(2) Salit, M. L.; Vocke, R. D., Jr.; Kelly, W. R. Anal. Chem. 2000, 72, 3504–3511.
Degrees of Degrees of (3) Salit, M. L.; et al. Anal. Chem. 2001, 73, 4821–4829.
mg/g freedom mg/g freedom
(4) National Institute of Standards and Technology. SRM 1775/2175 Certifi-
Observed value of cate of Analysis. NIST: Gaithersburg, MD.
test solution 9.9496 9.9501
(5) International Organization for Standardization. International Vocabulary of
Uncertainty sources Basic and General Terms in Metrology, Section 6.10, ISO: Geneva, Switzer-
SRM 3113 0.0014 3 0.0021 3 land, 1993.
Test sample 0.0017 3 0.0016 3 (6) Metrologia 2002, 39, Tech. Suppl. 08002.
(7) Turk, G. C.; et al. Fresenius’ J. Anal. Chem. 2001, 370, 259–263.
Combined uncertainty
of measurements only 0.0022 0.0026 (8) Beck, C. M., II; et al. Anal. Chem. 1993, 65, 2899–2902.
(9) Gordon, C. L.; Schlecht, W. G.; Wichers, E. J. Res. Natl. Bur. Stand. (U.S.)
Uncertainty of certified
1944, 33, 363–381.
value of SRM 3113 0.0100 8 0.0100 8
(10) Taylor, B. N.; Kuyatt, C. E. Guidelines for Evaluating and Expressing the Un-
Combined standard certainty of NIST Measurement Results; NIST Technical Note 1297, 2nd ed.;
uncertainty 0.0102 9 0.0103 9
National Institute of Standards and Technology: Gaithersburg, MD, 1994;
Coverage factor 2.26 2.26 http://physics.nist.gov/cuu/Uncertainty/index.html.
Expanded uncertainty 0.0230 0.0232 (11) Eurachem/CITAC Guide: Quantifying Uncertainty in Analytical Measurement,
2nd ed.; Laboratory of the Government Chemist: London, 2001.

A P R I L 1 , 2 0 0 5 / A N A LY T I C A L C H E M I S T R Y 141 A