Accred Qual Assur (2012) 17:129–138 DOI 10.1007/s00769-011-0827-5 GENERAL PAPER Uncertainty of standard addition experiments: a novel approach to include the uncertainty associated with the standard in the model equation Anna-Lisa Hauswaldt • Olaf Rienitz • Reinhard Jährling Nicolas Fischer • Detlef Schiel • Guillaume Labarraque • Bertil Magnusson • Received: 17 June 2011 / Accepted: 26 August 2011 / Published online: 16 September 2011 Ó Springer-Verlag 2011 Abstract A new model equation for determining the measurement result in standard addition experiments was derived and successfully applied to the quantitative determination of rhodium in automotive catalysts. Existing equations for standard addition experiments with gravimetric preparation were changed in order to integrate the novel idea of including the uncertainty associated with the standard into the model equation. Using this novel equation combined with the ordinary least squares algorithm for the regression line also yielded a new formula for the associated measurement uncertainty. This uncertainty accounts for the first time for the uncertainty associated with the standard. The derivation for the model equation and the resulting associated measurement uncertainty is shown for gravimetric standard addition experiments both with and without an internal standard. Keywords Standard addition Measurement uncertainty Internal standard Gravimetric preparation Metrological traceability Electronic supplementary material The online version of this article (doi:10.1007/s00769-011-0827-5) contains supplementary material, which is available to authorized users. A.-L. Hauswaldt (&) O. Rienitz R. Jährling D. Schiel Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, 38116 Braunschweig, Germany e-mail: [email protected] N. Fischer G. Labarraque Laboratoire National de Métrologie et d’Essais (LNE), 1 rue Gaston Boissier, 75724 Paris Cedex 15, France B. Magnusson SP Sveriges Tekniska Forskningsinstitut, P.O. Box 857, 501 15 Borås, Sweden Introduction Standard addition is applied to liquid samples or solutions of solid samples in case of complex matrices and low concentrations, if external calibration methods provide wrong results (due to different sensitivities in the sample and calibration solution) and if the method of isotope dilution is too tedious and expensive or impossible (monoisotopic analytes). Standard addition can be performed based on either volumetric or gravimetric sample preparation. As the volumetric sample preparation causes a loss of accuracy by demanding simplifying preconditions, a model equation for gravimetric standard addition experiments was set up [1]. To reduce the contribution to the measurement uncertainty caused by indication drifts and fluctuation, an internal standard can additionally be used. The accordingly modified gravimetric equation was successfully tested in an international pilot study [2, 3]. A new approach to include the uncertainty associated with the standard directly into the model equation was presented in [4]. Overcoming the limitations of this latest approach, a new model equation was derived from the equations in [1] and [2] for gravimetric standard addition experiments both with (section Standard addition combined with internal standard) and without (section Gravimetric standard addition) an internal standard. Conveniently, both versions of gravimetric standard addition experiments yield the same model equation and have hence the same associated measurement uncertainty, see section Measurement uncertainty for standard addition. An example is given for evaluating the measurement result of standard addition experiments using an internal standard and gravimetric preparation. The method was applied to the quantitative determination of the mass 123 130 wz Standard z wx Sample x Solvent y fraction of rhodium (Rh) in automotive catalysts, which was the measurand of an international pilot study (CCQMP63, see [2, 3]). The experiments were performed on a high-resolution inductively coupled plasma mass spectrometer (HR-ICPMS) and a multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS), respectively, acquiring the indications of 103Rh and 115In. It was shown that the method of gravimetric standard addition with an internal standard (indium) yields better results than the gravimetric standard addition without an internal standard and much better results than the method of external one-point calibration for determining the mass fraction of Rh in automotive catalysts. For the calculation of the measurement uncertainty, the uncertainties of the following influences are considered: The uncertainty of the x-intercept, the mass fraction of the Rh standard, the weighing (including the air buoyancy correction), the dry mass correction, the homogeneity of the sample and the sample preparation (digestion). While the last four contributions have to be introduced by extending the model equation of the standard addition, the uncertainty of the x-intercept and—for the first time—the uncertainty associated with the standard are accounted for directly in the equation of the standard addition. Accred Qual Assur (2012) 17:129–138 0 a0 x0 0 tan = a1 x Fig. 1 Standard addition: Preparation of the measurement solutions consisting of sample x, standard z (with mass fractions wx and wz, respectively) and a solvent; applying the measurement results on the calculation of the calibration line with y-intercept a0 and slope a1. x, y variables of abscissa and ordinate, respectively. x-intercept x0 equals the ratio wx/wz as will be shown below sample preparation is completely conducted on a balance and all individual masses are arbitrary but well known. Gravimetric standard addition Theory Standard addition In case of matrices being difficult to analyze or monoisotopic analytes, the method of standard addition can be used instead of isotope dilution (IDMS) or other techniques, see [1, 4]. Figure 1 depicts the idea of standard addition which is an internal calibration method. At least three different blends are measured consisting of an aliquot of sample x, different amounts (starting with zero) of a standard z of the analyte concerned and a solvent. The analyte is measured in these different solutions, and the indications, together with their respective analyte concentrations, define an ascending line. If linearity is given, the absolute value of the x-intercept equals the ratio of the analyte mass fraction wx in the original sample to the mass fraction of the analyte in the standard wz. In practical analytical work, it is common to do the preparation volumetrically, see for example [5, 6]. Here, the different blends having exactly equal volumes contain exactly the same amount of sample. Unfortunately, this condition is difficult or even impossible to achieve in practical laboratory work. Hence, the gravimetric approach was added to the method of standard addition, meaning the 123 Gravimetric standard addition is explained in Fig. 2 with the symbols listed in Table 1. There are k solutions i with a mass mi, containing the masses mx,i of sample x and different masses mz,i of standard z. Measuring the ith solution yields an indication Ai. For evaluating the standard addition, the gravimetric approach results in a linear equation, which is based exclusively on mass and mass fraction and accounts, in its shown form, for the individuality of each solution prepared. Therefore, deviations from the ideal approach, which are unavoidable in practice, do not become a source of error. The evaluation in the gravimetric standard addition is done as follows. The sensitivities of the analyte from sample x and reference z have to be equal in the final solutions. This precondition is crucial and usually met after mixing and equilibrating with the matrix or after a digestion. Starting from the relation between the concentration ci of the analyte in the ith solution, the sensitivity a01 and the indication Ai (minus the blank value and background), see Eq. 1, and using the densities qi, all concentrations ci are substituted by mass fractions wi. The mass fraction wi of the analyte in the ith solution is—as shown in Fig. 2—the sum of the analyte in the mass mx,i of sample x with the mass fraction wx of the analyte in sample x, which is the measurand, and the added mass mz,i of standard z with Accred Qual Assur (2012) 17:129–138 131 Solvent i-th Solution Standard z m z,i Sample x m x,i mi i Vessel Fig. 2 Gravimetric standard addition: solution and indication Ai. The ith solution has a density qi and a mass mi which contains the masses of sample x and of added standard z (mx,i and mz,i, respectively) Table 1 Symbols used in gravimetric standard addition Hence, we have the linear equation Symbol Quantity a0, a1 y-Intercept and slope of the linear fit function a01 Sensitivity y i ¼ a1 x i þ a0 mi 1 with yi ¼ Ai mx;i qi Ai Indication of the ith solution and a1 ¼ a01 wz ; mi Mass of the ith solution mx,i Mass of sample x in the ith solution mz,i Mass of added standard z in the ith solution wx Mass fraction of the analyte in sample x wz Mass fraction of the analyte in standard z wi Mass fraction of the analyte in the ith solution ci Mass concentration of the analyte in the ith solution qi x, y Density of the ith solution Variables of abscissa and ordinate ð1Þ where ci ¼ qi wi and wi ¼ mx;i wx þ mz;i wz : mi Hence, mx;i wx þ mz;i wz Ai ¼ qi wi ¼ qi mi mi 1 m w þ m x;i x z;i wz Ai ¼ a01 mx;i qi mx;i mz;i 0 ¼ a1 wz þ a01 wx : mx;i a01 xi ¼ ð3Þ It follows for the mass fraction of the analyte in the sample x: a0 wx ¼ wz ð4Þ a1 the mass fraction of the analyte in standard z related to the total mass mi of the ith solution. This can be written as a linear equation 2, whose x-intercept yields the measurand, see Eqs. 3 and 4. Ai ¼ a01 ci mz;i mx;i a1 0 a0 ¼ a1 wx ¼ wx : wz and Note, that Eq. 1 contains the concentration ci and not the mass fraction wx of the analyte in sample x. Nevertheless, the Eqs. 2–4 can be expressed in terms of mass fraction wx by using the different densities qi of the solutions. Since almost all analytical instruments generate indications proportional to the amount of analyte in a certain volume, due to sample loops or pumps used to introduce the sample, Eq. 1 as the basis has to be written in terms of a concentration. The density qi directly connects the concentration ci and the mass fraction wi and has therefore to be considered in the model equation after substituting the concentration with the mass fraction. In practice, the densities qi are nearly equal for all solutions i and usually equal to the density of the solvent used to prepare the solutions. As all quantities are known, this approach does not contain any approximations or preconditions (with the exception of linearity), and it is independent of temperature [1]. Even though the densities qi are temperature dependent in theory, the final result is nevertheless independent of the temperature since simulations have shown, that, as long as the densities qi are equal, the value of qi does not change the result at all. Standard addition combined with internal standard a01 ð2Þ In the standard addition with an internal standard, the solutions consist of sample x, different amounts of a standard z, a solvent and additionally an internal standard y, which is contained in neither the sample x, the standard z nor the solvent, see Fig. 3. 123 132 Accred Qual Assur (2012) 17:129–138 With wy Internal standard y wz Standard z a01 ¼ b0 : b00 wy Equation 5 yields after rearranging Ri my;i ¼ a01 mx;i wx þ mz;i wz Ri wx Sample x my;i mx;i ¼ a01 mx;i wx þ a01 mz;i wz mz;i ¼ a01 wz þ a01 wx : mx;i Hence, we have again the linear equation Solvent Fig. 3 Gravimetric standard addition with an internal standard: Preparation of the measurement solutions consisting of sample x, standard z, internal standard y (with the mass fractions wx, wz and wy, respectively) and a solvent Then, not one indication Ai but the ratio Ri of the indications of the analyte and the internal standard is considered as shown in Fig. 4. The resulting linear equation yields again the measurand via its x-intercept, see Eq. 6. As in section Gravimetric standard addition the linear relation of the indication A and the respective concentration c is the base for deducing the equation of the standard addition with an internal standard. From Eq. 1 and the equations my;i wy mx;i wx þ mz;i wz wi ðXÞ ¼ ; wi ðYÞ ¼ mi mi y i ¼ a1 x i þ a 0 my;i Ai ðXÞ mz;i ; xi ¼ with yi ¼ Ri ; Ri ¼ Ai ðYÞ mx;i mx;i a1 and a1 ¼ a01 wz ; a0 ¼ a01 wx ¼ wx : wz It follows that a0 wx ¼ wz a1 with the symbols explained in Table 2. The resulting linear equation can be applied to the evaluation of gravimetric standard addition experiments, which additionally use an internal standard. In addition to the advantages of the gravimetric approach, the standard addition with an internal standard has the following advantages: 1. it follows that Ai ðXÞ b0 ci ðXÞ b0 q wi ðXÞ ¼ 00 ¼ 00 i Ai ðYÞ b ci ðYÞ b qi wi ðYÞ b0 wi ðXÞ b0 mx;i wx þ mz;i wz =mi ¼ ¼ 00 b w i ðY Þ b00 my;i wy =mi b0 mx;i wx þ mz;i wz ¼ b00 my;i wy 0 b mx;i wx þ mz;i wz ¼ 00 : b wy my;i No exact knowledge about the mass fraction wy of the internal standard is needed. The result is not changed by dilution or storage as the mass of the ith solution mi is not contained in the equation. The result is independent of temperature because the density qi was canceled from the equation. The ratio of indications is evaluated; hence, possible changes caused by the measuring device are canceled to a certain degree. 2. Ri ¼ ð6Þ 3. 4. ð5Þ Ri Solvent i-th Solution Standard z mz,i Int. std. y my,i Sample x mx,i mi i Vessel Fig. 4 Standard addition with internal standard: solution and measurement of the ratio Ri of the indications Ai(X) and Ai(Y). The ith solution has a density qi and a mass mi which contains the masses of sample x, of internal standard y and of standard z (mx,i, my,i and mz,i, respectively) 123 Accred Qual Assur (2012) 17:129–138 133 Table 2 Symbols used in standard addition with internal standard (IS) Symbol Quantity a0, a1 y-Intercept and slope of the linear fit function b 0 Sensitivity of analyte X b00 Sensitivity of IS Y a01 Ratio of the sensitivities of analyte X and IS Y divided by the mass fraction of IS solution Ri Ratio of the indications of analyte X and IS Y in the ith solution mx,i Mass of sample x in the ith solution Experimental my,i Mass of IS solution y in the ith solution mz,i Mass of added standard z in the ith solution wx Mass fraction of the analyte in sample x wy Mass fraction of the IS wz Mass fraction of the analyte in standard z x, y Variables of abscissa and ordinate Measurement uncertainty for standard addition From the model equations 4 and 6, respectively, the measurement uncertainty for the standard addition can be derived as follows (for details see Appendix: Derivation of the measurement uncertainty) using the standard approach of propagation of variances [7]: a0 wx ¼ wz yields a1 owx 2 2 owx 2 2 2 u ðwx Þ ¼ u ð a0 Þ þ u ð a1 Þ oa0 oa1 owx 2 2 u ðwz Þ þ owz owx owx þ2 uð a0 ; a1 Þ oa0 oa1 owx owx þ2 uð a0 ; w z Þ oa0 owz owx owx þ2 uða1 ; wz Þ: oa1 owz Hence, 2 2 3 wx 2 2 2 þ x wz uð w x Þ uð w z Þ S 61 7 ¼ þ 2 4 þ Pn 5 2 wx wz n a0 Þ i¼1 ðxi x ð7Þ with 2 S ¼ Basic for Applications (VBA), which is available as Electronic Supplementary Material from the journal’s website along with a corresponding Excel file containing an example data set. Pn i¼1 ½ y i ð a0 þ a1 x i Þ 2 ; n2 Pn x ¼ i¼1 xi n : To calculate the uncertainty associated with wx according to Eqs. 17 and 7, respectively, a Microsoft Excel macro (MU_StdAdd_wz()) was written in Visual The sample preparation and measurements described were part of the participation in the international pilot study CCQM-P63. The automotive catalyst sample was provided by LGC (UK), the co-ordinating laboratory of this study. From the powdered, homogenized and dried catalyst material five aliquots i ði ¼ 1; . . .; 5Þ with a mass mx,i were sampled and weighed directly into PTFE-TFM microwave vessels. After adding an appropriate mass of internal standard and standard solution my,i and mz,i, respectively, the samples were digested microwave assisted using an ETHOS-1600 (MLS, Germany) in two steps using nitric, hydrofluoric and hydrochloric acid. Three blanks were prepared in the same way without adding sample, standard or internal standard. Additional aliquots were used to determine the water content of the catalyst sample in order to calculate the dry mass correction factor wdry. All weighings were corrected for the air buoyancy. The density of HCl with w(HCl) = 0.07 g/g (q(HCl) = 1033 kg/m3 according to [8]) was used as an appropriate approximation of the density of the solutions qi. The Rh standard solution was prepared from Rh powder PMR-461 (Alfa Aesar, Germany). The powder was digested within 5 h at 290 °C (High Pressure Asher, Anton Paar, Austria) using HCl with w(HCl) = 0.3 g/g (Merck, suprapur) and chlorine (generated in situ from KClO3, Merck, for analysis). The resulting solution was diluted accordingly. The masses of 103Rh and 115In were measured using an HR-ICP-MS (Element2, Thermo Scientific, Germany). All parameters of the ICP-MS were optimized for signal intensity. The standard quartz sample introduction system was used. Ten series of measurements with increasing concentrations of Rh were evaluated using both standard addition and standard addition with an internal standard. These measurements were repeated with an MC-ICP-MS (Neptune, Thermo Scientific, Germany). Additionally, a Rh calibration solution with a Rh mass fraction of wz,cal was prepared and integrated in the measurement sequence. The Rh indication for the solution (i = 1) with no standard added (mz,1 = 0) was evaluated against the indication Az observed for the external calibration standard in order to compare the external one-point calibration to the standard addition. The one-point calibration results were calculated according to Eq. 8. Symbols not mentioned above are listed in Table 1. 123 134 Accred Qual Assur (2012) 17:129–138 wx ¼ A1 m1 wz;cal : Az mx;1 ð8Þ The actual standard addition experiment consisted of the following steps: 1. Weighing: Five sample aliquots i ði ¼ 1; . . .; 5Þ of mx;i 100 mg: (b) Internal standard (In-solution with wy & 40 lg/g) with my,i & 0.9 g. (c) Standard (Rh-solution with wz & 40 lg/g) with mz;i 0; . . .; 1:2 g. (a) 2. Sample preparation: (a) (b) (c) (d) (e) (f) (g) 3. Adding of 3 mL HNO3 with w(HNO3) = 0.65 g/g (Merck, p.a., subboiled), 9 mL HCl with w(HCl) = 0.3 g/g (Merck, suprapur). First microwave digestion, 1 h, 210 °C. Hot plate evaporation (220 °C) down to 1 mL. Adding of 6 mL HNO3 with w(HNO3) = 0.65 g/g, 3 mL HCl with w(HCl) = 0.3 g/g, 3 mL HF with w(HF) = 0.49 g/g (Fluka, trace select ultra). Second microwave digestion, 1 h, 210 °C. Hot plate evaporation (210 °C) down to 0.5 mL. Transferring to LD-PE bottle and filling up to solution mass mi & 120 g with HCl with w(HCl) = 0.07 g/g. Measurement with ICP-MS: (a) Plasma power 1200 W. (b) Mass resolution 300. (c) Counting mode monitoring the indications A(103Rh) and A(115In). 4. Evaluation (according to Table 5): (a) Calculating xi. (b) Subtracting the blank indications. (c) Calculating yi from the 103Rh indication (standard addition). (d) Calculating yi from the ratios Ri of the indications of 103Rh and 115In (standard addition with an internal standard). (e) Calculating a0 and a1 from the ordinary least squares algorithm. (f) Calculating wx and u(wx). Results and discussion Four different scenarios (A–D) were compared (see Table 4; Fig. 5): scenario A: external one-point calibration with measurements on an HR-ICP-MS, scenario B: standard 123 Fig. 5 Left-hand side: Intermediate result wx without dry mass correction calculated according to Eqs. 8, 6 and 17 yielded from four different scenarios A–D; right-hand side: final result w with (among others) dry mass correction applied according to Eq. 9, also yielded from the four different scenarios A–D. Error bars denote the standard uncertainty (left-hand side) and expanded uncertainties (right-hand side), respectively. The consensus value and its expanded measurement uncertainty from the pilot study CCQM-P63 (dotted and dashed lines, respectively) plotted as the reference. See Table 4 for details addition with measurements on an HR-ICP-MS, scenario C: standard addition combined with an internal standard measured on an HR-ICP-MS, scenario D: like C but with measurements on an MC-ICP-MS. The pilot study consensus value [3] from CCQM-P63 served as the reference. Scenario A was evaluated according to Eq. 8. The method of external one-point calibration resulted in a rhodium mass fraction, which differed from the consensus value by almost 100%. Evaluating the associated measurement uncertainty as well as applying the dry mass correction was, thus, superfluous in this case. Contrarily, with the gravimetric standard addition scenario B (presented in section Gravimetric standard addition), the deviation from the consensus value was covered by the measurement uncertainty associated with the rhodium mass fraction. However, this uncertainty was considerably large (13%). Scenario C, the additional use of an internal standard, (presented in section Standard addition combined with internal standard) resulted in a dramatically reduced uncertainty compared to scenario B (2.1%). The deviation from the consensus value was already covered by the measurement uncertainty associated with the consensus value. Obviously, the indication ratios—accomplished by the addition of the internal standard—rather than absolute indications improved the precision, even on a single-collector and therefore sequentially measuring ICP-MS. Since the analyte was not separated from the very complex matrix—except for major parts of the silicon—by means of chromatography (or other techniques), the absolute indications showed large standard deviations (drift and noise). Accred Qual Assur (2012) 17:129–138 Table 3 Symbols used in Eq. 9 135 Symbol Unit Quantity w lg/g Mass fraction of the analyte Rh in the automotive catalyst sample—measurand wdry g/g Dry mass correction—result of repeated measurements fexp 1 Sampling, sample preparation and inhomogeneity wx lg/g Result of the standard addition model equation (mass fraction of the analyte in the sample without moisture correction) dmx 1 Uncertainty contribution from the sample mass dmz 1 Uncertainty contribution from the mass of standard added dmi 1 Uncertainty contribution from the mass of solutions measured Table 4 Rhodium mass fractions without (intermediate result) and with dry mass correction applied (final result) along with their associated uncertainties (standard uncertainties u in case of the Scenario Method Device intermediate result, expanded uncertainties U with the coverage factor k (p = 0.95) in case of the final result) Intermediate result wx (lg/g) u (lg/g) A External one-point calibration HR-ICP-MS 450 B Gravimetric standard addition HR-ICP-MS 215.6 C Std. add. with internal standard HR-ICP-MS 234.6 4.8 D Std. add. with internal standard MC-ICP-MS 232.9 2.1 Final result urel (%) w (lg/g) U (lg/g) Urel (%) k (1) – – 450 – – 28 13 218 58 27 2.1 237.0 19 8.0 2.2 0.9 235.4 15 6.3 2.0 1.8 2.0 Pilot study (consensus value) 234.0 4.2 – 2.0 Four scenarios A–D represent combinations of different methods and mass spectrometers (devices) Using a simultaneously measuring MC-ICP-MS to analyze the solutions (scenario D) took full advantage of the combination of standard addition and internal standard resulting in a further reduced uncertainty (0.9%). The rhodium mass fraction wx calculated according to Eq. 6 without dry mass correction is only an intermediate result and becomes therefore one input quantity of the final rhodium mass fraction w described with Eq. 9. The meaning of the symbols used are compiled in Table 3. w¼ 1 fexp wx dmx dmz dmi : wdry ð9Þ The uncertainties associated with these final (dry mass corrected) results w were determined using the GUMWorkbenchTM [9] and Eq. 9 as the model equation. Table 4 and Figure 5 summarize the rhodium mass fractions without and with the dry mass correction applied in case of the four scenarios A–D. The dry mass correction and sample preparation increased the measurement uncertainty considerably. This way the advantage of the MC-ICP-MS (scenario D) became less obvious. Fig. 6 shows the relative contribution of all relevant input quantities to the measurement uncertainty associated with the dry mass corrected rhodium mass fraction w. In case of scenario D the Fig. 6 The measurement uncertainty associated with the dry mass corrected Rh mass fraction w according to Eq. 9 is dominated by different input quantities in case of the different scenarios B–D. Relative contributions (calculated as the fraction of the squared product of the sensitivity coefficient and the standard uncertainty of each input quantity from u2c (w), [7]) to the uncertainty associated with w arising from the following input quantities (compare Eq. 9): light gray: intermediate result/measurement wx, white: sample preparation/ dry mass correction fexp, dark gray: sample mass mx, black: masses of standard added and masses of final solutions (mz,i and mi, respectively). See Table 4 for details about the different scenarios 123 136 Accred Qual Assur (2012) 17:129–138 Table 5 Summary of equations in standard addition using a linear model yi ¼ a1 xi þ a0 m z;i xi ¼ mx;i OLS algorithm Model equation wx ¼ aa01 wz Gravimetric preparation plus internal standard yi ¼ Ai mmx;ii q1 yi ¼ Ri my;i x;i m i a0 ¼ Pn Pn 2 Pn Pn xi x y i¼1 xi i¼1 i i Pi¼1 P ¼ y a1 x 2 n n n x2 ð xÞ i¼1 i i¼1 i y i¼1 i a1 ¼ Measurement uncertainty 2 2 2 ðwx þxÞ u2 ðwx Þ ¼ w2x uðwwzz Þ þw2z Sa2 1n þ Pnwz 2 1 i¼1 n and Þ Ri ¼ AAii ððX YÞ Pn Pn Pn Pn xy y i¼1 xi ðyi yÞðxi xÞ i¼1 i i i¼1 i i¼1 P P P ¼ 2 n 2 n n n x2 ð xÞ ðxi xÞ i¼1 i¼1 i i¼1 i where Pn Pn 2 xi ½yi ða0 þa xi Þ ; x ¼ i¼1 S2 ¼ i¼1 n2 1 n ðxi xÞ All indications Ai must be blank corrected prior to the calculations measurement itself did not contribute significantly to the combined uncertainty. As already mentioned, it looks as if the use of MC-ICP-MS seems to be an inappropriate effort. But the MC-ICP-MS opens up the possibility to monitor the influence of the dry mass correction and different sample preparation procedures, which is nearly impossible using the HR-ICP-MS, because the information is hidden by the uncertainty associated with the measurement itself. The gravimetric standard addition yielded a Rh mass fraction with the relative expanded (p = 0.95) measurement uncertainty Urel = 27%. Applying the method of gravimetric standard addition with an internal standard (see section Standard addition combined with internal standard) decreased considerably the measurement uncertainty (Urel = 8%). Using the simultaneously measuring MCICP-MS causes further improvements (Urel = 6.3%). Even though these uncertainties seem to be rather large compared to IDMS, the small sample sizes (50–100 mg) have to be considered. Most of the other participants of CCQMP63 used sample sizes of 1–12 g. The relative expanded uncertainty of 0.9% associated with the uncorrected Rh mass fraction calculated from scenario D demonstrates the potential of the gravimetric standard addition in combination with an internal standard to compete with IDMS. Table 5 summarizes all equations needed to perform the data evaluation of standard addition experiments without and with the additional use of an internal standard. Note that merely the definition of yi changes in case an internal standard is used. The determination of Rh in an automotive catalyst sample in the framework of an international pilot study underpinned the potential of the gravimetric standard addition especially in combination with an internal standard to yield results with a performance close to that of IDMS even in demanding matrices. The method described can therefore easily be adopted in clinical chemistry where samples usually feature complex matrices and the applicability of IDMS is restricted to a certain number of analytes. Gravimetric standard addition will therefore help to set up reference methods without the efforts of IDMS. This way it could help to save money in the public health sector. Acknowledgments The authors wish to thank Carola Pape who tackled the tricky sample preparation. The research within this EURAMET joint research project receives funding from the European Community’s Seventh Framework Programme, ERA-NET Plus, under Grant Agreement No. 217257. Appendix: Derivation of the measurement uncertainty Conclusion A gravimetric standard addition can be modeled mathematically without using simplifying preconditions, which are difficult or even impossible to fulfill. It offers an alternative approach to the more sophisticated IDMS in case of complex matrices, monoisotopic analytes or in case no mass spectrometer is available in the laboratory. In the form presented here, even the uncertainty of the added reference is included in a closed form—to our knowledge—for the first time. This way it becomes easier to claim metrological traceability of measurement results determined with standard addition. 123 From the model equations 4 and 6, respectively, the measurement uncertainty for standard addition can be derived as follows following the guidelines in [7]. With wx ¼ a0 wz a1 the partial derivatives are: owx wz ¼ ; oa0 a1 owx wx ¼ ; oa1 a1 owx wx ¼ : owz wz Using the approach of propagation of variances and inserting the above partial derivatives yields: Accred Qual Assur (2012) 17:129–138 137 owx 2 2 owx 2 2 u ðwx Þ ¼ u ð a0 Þ þ u ð a1 Þ oa0 oa1 owx 2 2 þ u ðw z Þ owz owx owx þ2 uða0 ; a1 Þ oa0 oa1 owx owx þ2 uð a0 ; w z Þ oa0 owz |fflfflfflfflffl{zfflfflfflfflffl} ¼0 owx owx þ2 uð a1 ; w z Þ oa1 owz |fflfflfflfflffl{zfflfflfflfflffl} uð a0 ; a1 Þ ¼ Pn 2 i¼1 and 1 Xn n ð10Þ The uncertainty u2(wz) was estimated with an additional uncertainty budget, which includes all preparation steps along with the air buoyancy correction and concentration changes due to storage losses of the solvent. Possible correlations between the regression line and wz were determined by simulating the experiment. No correlations were observed; hence, we have: uða0 ; wz Þ ¼ 0 and uða1 ; wz Þ ¼ 0: The ordinary least squares (OLS) algorithm for linear regression curves yields the formulae for a0 and a1 (see Table 5) as well as for u2(a0), u2(a1) and the correlation u(a0, a1) [10]. The OLS was chosen because it is widely used and suited to the described experiment with uncertainties larger in y than in x. The derivation of the uncertainty associated with the analyte mass fraction u(wx) can be performed analogously with any other suitable fitting algorithm. Using Eq. 14, the uncertainties associated with a0 and a1 (Eqs. 11 and 12) were rearranged in the following way: Pn ð11Þ n S2 S2 u2 ð a1 Þ ¼ Pn ¼ P P 2 n n Þ2 n i¼1 x2i i¼1 ðxi x i¼1 xi ð12Þ u ð a0 Þ ¼ S2 1 ¼ Pn n Pn 2 i¼1 xi 2 2 i¼1 xi 2 Pn 2 Pn n i¼1 xi i¼1 xi i¼1 ðxi xÞ X 1 n x2 2 i i n X 2 X 1 2 ¼ 2 n i ðxi xÞ þ x i i n Pn ðxi xÞ2 ¼ i¼1 þ x2 : n x2 ¼ i¼1 i ¼0 S2 ð13Þ ðxi xÞ2 Additionally, the following identities are needed: X 2 X X 2 n x x ¼n ðxi xÞ2 ð14Þ i i i i i 2 wx ¼ u2 ðwz Þ wz wz wx þ2 uð a0 ; a1 Þ a1 a1 2 wx 2 2 wz þ u ða1 Þ þ u2 ð a0 Þ a1 a1 u2 ðwz Þ ¼ w2x w2z " # w2z a0 2 2 a0 2 þ 2 u ð a1 Þ 2 uð a0 ; a1 Þ þ u ð a0 Þ : a1 a1 a1 2 S2 x ð15Þ Using the Eq. 11 for u2 ða0 Þ; Eq. 12 for u2 ða1 Þ and Eq. 13 for uða0 ; a1 Þ; as well as Eqs. 14 and 15 the expression in square brackets from Eq. 10 can be evaluated as follows. " # a0 2 2 a0 u ð a1 Þ 2 uð a0 ; a1 Þ þ u2 ð a0 Þ a1 a1 ! 2 a0 S2 a0 S2 x ¼ Pn 2 Pn a1 a1 Þ2 Þ2 i¼1 ðxi x i¼1 ðxi x P S2 1 n x 2 þ Pn n i¼1 i 2 Þ i¼1 ðxi x " # S2 a0 2 a0 1 Xn 2 ¼ Pn þ2 x þ x i¼1 i n a1 a1 Þ2 i¼1 ðxi x ¼ Pn S2 ðxi xÞ2 "i¼1 !# Pn Þ2 a0 2 a0 2 i¼1 ðxi x þ x þ2 x þ a1 a1 n " # Pn 2 1 ð x x Þ i ¼ S2 Pn i¼1 n Þ2 i¼1 ðxi x " !# 2 1 a0 a0 2 2 þ S Pn þ2 x þ x a1 a1 Þ2 i¼1 ðxi x 2 2 3 a0 a1 þ x 61 7 ¼ S2 4 þ Pn 25 n ð x x Þ i i¼1 2 2 3 wx þ x wz 61 7 ¼ S2 4 þ Pn : ð16Þ 25 n Þ i¼1 ðxi x From Eq. 10 by using Eq. 16 it follows 2 2 3 wx 2 2 þ x wz uðwz Þ S 61 7 u2 ðwx Þ ¼ w2x þw2z 2 4 þ Pn 25 wz a1 n ð x x Þ i i¼1 ð17Þ 123 138 or written in terms of the relative uncertainty 2 2 3 wx 2 2 2 wz þ x uð w x Þ uð w z Þ S 61 7 : ¼ þ 2 4 þ Pn 25 wx wz a0 n ð x x Þ i i¼1 Accred Qual Assur (2012) 17:129–138 ð7Þ References 1. Rienitz O, Röhker K, Schiel D, Han J, Oeter D (2006) New equation for the evaluation of standard addition experiments applied to ion chromatography. Microchimica Acta 154:21–25 2. Rienitz O (2008) Uncertainty of standard addition experiments using an internal standard and gravimetric preparation—determination of Rh in automobile catalysts. In: Tagungsbericht 4. VDI Fachtagung Messunsicherheit praxisgerecht bestimmen, 12./ 13.11.2008, Erfurt, ISBN 978-3-98-12624-1-4 3. Hearn R, Wolff-Briche CSJ, Sargent M (2008) CCQM-P63: platinum group elements in an automotive catalyst, LGC/VAM/ 2006/028 123 4. Serapinas P, Labarraque G, Charlet P, Ežerinskis Ž, Juzikien_e V (2010) Method of standard additions for arsenic measurements in water by ICP sector field mass spectrometry at accuracy comparable to isotope dilution. J Anal At Spectrom 25:624–630 5. DIN 32633 (1998) Verfahren der Standardaddition. Beuth, Berlin 6. Harris DC (1998) Quantitative chemical analysis. W. H. Freeman and Company, New York 7. JCGM 100 (2008) Evaluation of measurement data—guide to the expression of uncertainty in measurement. BIPM, Paris 8. Küster FW, Thiel A (1993) Rechentafeln für die Chemische Analytik. Walter de Gruyter, Berlin 9. GUM-WorkbenchTM, Version 1.2.11.56 Win32 copyright ’96’99 by Metrodata GmbH 10. Papula L (2008) Mathematik für Ingenieure und Naturwissenschaftler 3: Vektoranalysis, Wahrscheinlichkeitsrechnung, Mathematische Statistik, Fehler- und Ausgleichsrechnung. Vieweg ? Teubner Verlag, Wiesbaden
© Copyright 2026 Paperzz