International Comparison CCQM K53 Oxygen in Nitrogen

Transcription

1 CCQM-K53 Final Report International Comparison CCQM K53 Oxygen in Nitrogen Jeongsoon Lee 1, Jin Bok Lee 1, Dong Min Moon 1, Jin Seog Kim 1*, Adriaan M.H. van der Veen 2, Laurie Besley 3, Hans-Joachim Heine 4, Belén Martin 5, L.A. Konopelko 6, Kenji Kato 7, Takuya Shimosaka 7, Alejandro Perez Castorena 8, Tatiana Macé 9, Martin J.T. Milton 10, Mike Kelley 11, Franklin Guenther 11, Angelique Botha 12 1 Korea Research Institute of Standards and Science (KRISS), Division of Metrology for Quality Life, P.O.Box 102, Yusong, Daejeon, Republic of Korea 2 VSL, Thijsseweg 11, 2629 JA Delft, the Netherlands 3 National Measurement Institute, Australia (NMIA), Bradfield Road, West Lindfield, Lind-field, NSW 2070, Australia 4 Bundesanstalt für Materialforschung und prüfung (BAM), Abteilung I, Unter den Eichen 87, D Berlin, Germany 5 Centro Espanol de Metrologia (CEM), C/ del Alfar, 2, Tres Cantos (Madrid), Spain 6 D.I. Mendeleyev Institute for Metrology, Rostekhregulirovaniye of Russia (VNIIM), Department of State Standards in the field of Physical Chemical Measurements, 19, Moskovsky Prospekt, St- Petersburg, Russia 7 National Metrology Institute of Japan (NMIJ), Tsukuba Central 3, Tsukuba , Japan 8 CENAM, Km. 4,5 Carretera a los Cues, Municipio del Marques C.P , Queretaro, Mexico 9 Laboratoire national de metrologie et d'essais (LNE), rue Gaston Boissier Paris Cedex 15, France 10 National Physical Laboratory (NPL), Teddington, Middlesex, TW11 0LW, UK 11 National Institute of Standards and Technology (NIST), Chemical Science and Technology Laboratory, 100 Bureau Drive, Gaithersburg MD, USA 12 National Metrology Institute of South Africa (NMISA), Meiring Naude road, Building 4, Room W14, 0001 Pretoria, South Africa 1

2 Field Amount of substance Subject Oxygen 100 level in Nitrogen Participants Total 12 NMIs have taken part in this comparison. The participants are listed in table 1. Table 1: List of participants Acronym Country Institute NMIA AU National Metrology Institute of Australia, Linfield, Australia BAM DE Bundesanstalt für Materialforschung und prüfung, Berlin, Germany CEM ES Centro Espanol de Metrologia, Madrid, Spain LNE FR Laboratoire National d'essais, Paris, France NMIJ JP National Metrology Institute of Japan, Tsukuba, Japan KRISS KR Korea Research Institute of Standards and Science, Daejeon, Republic of Korea CENAM MX Centro Nacional de Metrologia, Queretaro, Mexico VSL NL VSL, Delft, the Netherlands VNIIM RU D.I. Mendeleyev Institute for Metrology, St. Petersburg, Russia NPL UK National Physical Laboratory, Teddington, Middlesex, United Kingdom NIST US National Institute of Standards and Technology, Gaithersburg, United States of America NMISA ZA National Metrology Institute of South Africa, Pretoria, South Africa Organizing body CCQM Introduction Gravimetry has been used as the primary method for the preparation of primary standard gas mixtures in most of NMIs [1], which requires the complete abilities of the 2

3 purity assessment, weighing technique and analytical skills. At the CCQM GAWG meeting in October 2005, it was agreed that KRISS coordinates a key comparison on the gravimetric preparation of gas, at a level of 100 of oxygen in nitrogen. KRISS compared the gravimetric value of each cylinder by an analytical instrument. A preparation for oxygen gas standard mixture requires a particular notice to be accurate, because it is a major component of atmosphere. Key issues for this comparison are related to (1) gravimetric technique which needs at least two steps for dilution, (2) oxygen impurity in nitrogen, and (3) argon impurity in nitrogen. Support for CMC claims This key comparison will support claims for capability to perform a gravimetric preparation, including purity assessment, on primary gas mixtures from 10 to 10 mmol/mol of oxygen in nitrogen. As a result of the comparison, CMCs can also be recognised for stable gas species, such as argon, neon, helium, hydrogen, in nitrogen because of similarity in the preparation and purity measurement processes. Schedule Jan. 15, 2007 Shipment of cylinders from participants to KRISS Feb. 1 ~ Mar. 31, 2007 Measurement of oxygen and argon amount in cylinders by KRISS 2007 CCQM/GAWG April meeting Draft A report 2008 CCQM/GAWG Fall meeting Discussion on a KCRV determination Design of the key comparison For this key comparison, participants were divided in three groups according to the nominal mole fraction of oxygen in nitrogen and prepared their standard mixtures of the appointed amount of substance fraction of oxygen in nitrogen, as shown in table 1. The amount-of-substance fractions of oxygen had to be in the range of 98 to 102. Table 2: the three groups of participants and their nominal mole fraction of oxygen in nitrogen appointed for CCQM-K53. Nominal Oxygen amount Participating NMI fraction in Nitrogen [] 3

4 Group A 99 NMIJ, NPL, BAM Group B 100 CEM *, NMIA, NIST, NMISA, LNE ** Group C 101 CENAM, KRISS, VNIIM, VSL * CEM announced to have an error in their data, so the CEM result was not used for the evaluation of the KCRV. ** LNE was appointed to prepare 100, but they sent the gas cylinder with the amount of the oxygen fraction of 101 to the coordinating lab. Participating laboratories had to report weighing method in detail and all of related data including a detailed description of the uncertainty budget and send the cylinders to the coordinating laboratory with a sufficient amount of gas left (pressure of at least 5 MPa). After receiving all cylinders from participants, the coordinating laboratory compared each cylinder. Verification of analytical procedures KRISS has reported a measurement capability at a mole fraction of 100 oxygen in nitrogen measured with a gas chromatography (GC)/thermal conductivity detector (GC/TCD) and a GC/atomic emission detector (AED) including assessment of argon impurity during CCQM/GAWG meetings. A modified GC/TCD was used as an analyzer for this comparison. During this comparison, the measurement reproducibility with the GC/TCD was kept within 0.06 % (relative standard deviation). In order to check an internal consistency, a series of 4 mixtures with mole fractions between 98 and of oxygen in nitrogen were prepared gravimetrically. One out of 4 cylinders (A, B, C, D) was chosen as reference cylinder (A), the others were analyzed against the reference. Each cylinder was measured three times against the reference cylinder, according to the following sequence (for cylinder B) : A-B-A-B-A- B-A to compensate the drift of the GC/TCD analyzer. It is well known that GC-TCD shows a good linearity within a wide range. Introducing an oxygen null gas into a sample inlet, the response of a chromatogram recorded with a GC/TCD analyzer is negligible enough that it is extrapolated to origin. Therefore, a calibration curve was forced to go through origin with 4 points. The curve including origin prevents KCRV from getting into some accidental deviation caused by any wrong cylinder. Hence the predicted mole fraction (x pred ) in table 3 has been calculated on the basis of an unweighted ordinary linear least squares (OLS) regression including origin, as the uncertainty of y is the same for all points. The measurement data (x prep, y) are in Table 3 and the result is shown in Eq. 1 and Fig. 1. The value of y-intercept in Eq. 4

5 1 shows a deviation of ± , which is much less than the value of reproducibility of the analyzer. y = ( ± ) x + ( ± ) (R 2 = , N=5). (Eq. 1) In table 3, the following data are summarized x prep mole fraction, from preparation (10-6 mol/mol) u (x) standard uncertainty of xprep (10-6 mol/mol) y response ratio against the reference cylinder (ME5662) with GC/TCD u (y) standard uncertainty of y x pred x calculated value using linear regression equation (Eq. 1) (10-6 mol/mol) difference between x pred and x prep (10-6 mol/mol). Table 3: analysis of the four primary reference cylinders of oxygen in nitrogen prepared by KRISS Cylinder # x prep [] u (x) using a GC/TCD analyzer [] y (1) u (y) x pred [] x (=x pred - x prep ) [] ME ME ME ME (1) Response ratio against the reference cylinder, ME5662. Two standard deviations in table 3 are taken as the uncertainty of preparation (0.01 %) and the uncertainty of the analysis (0.06 %), respectively. This linear regression analysis looks reasonable in that the maximum deviation value ( x = ) is sufficiently less than the value of the preparation uncertainties (0.022 ). Based on the residuals satisfy the following requirements based on ISO 6143[4]. x k u x ), (Eq. 2) i ( i where k denotes coverage factor and u(x i ) the standard uncertainty associated with x i. As a result of the regression analysis for 5 points including origin, the regression analysis indicates that any influential systematic error during this work may not be engaged. 5

6 Accordingly, this regression analysis demonstrates the internal consistency between the standard mixtures prepared by KRISS. Figure 1: Linear regression of 4 calibration points for quality control Measurement of oxygen in nitrogen The measurement of oxygen in nitrogen was carried out using a HP6890 GC analyzer, equipped with a modified TCD detector and a sampling valve line without an injection port. The sampling valve is directly connected to the column to remove possible dead volume. The column used was a Molsieve 5A, 4 m, OD 1/8", mesh. Helium was used as carrier gas, with a flow of 28 ml/min. Samples passed through a Multi Position Valve (MPV) directly connected to each cylinder, then through a mass flow controller to be finally introduced in the sample loop (Fig. 2). Two pressure restrictors were adopted to provide damping effect and to maintain a constant pressure in the TCD detector and in the sample loop before venting to the atmosphere. For the purpose of checking possible leakages and optimizing the sensitivity of the system, we had operated it over a week with the MPV connected to the cylinders. As a result, the optimized experimental conditions in the GC system were fixed at a sample flow rate of 250 ml/min with a volume of 5 ml. In case that the peak of argon can not be separated from that of oxygen under these conditions, other experimental conditions were sought by lowering the temperature of the GC column in oven. Two examples of experimental conditions which enable the oxygen peak in the chromatogram to be resolved from the argon peak are presented in table 4. 6

7 Figure 2: Schematic diagram of the experimental setup During the measurement of each cylinder, the pressure restrictors were adjusted to a pressure of 1.9 bar, which ensured the pressure in the sample loop to be constant. The sample loop was continuously rinsed (except when injecting). As a rule, it takes 20 minutes to obtain one data point and most of the time was consumed in the process of heating and cooling the oven to separate argon and oxygen. Table 4: Analytical conditions for two different cases Condition 1 Condition 2 Detector TCD (modified) " Dec. temp. 275 " Carrier flow 28 ml/min (50 psi) " Ref. flow 45 ml/min " Column Resteck Molesieve 5A 4m, 1/8", SS " Oven temp. -15, 13.5 min 30 /min 225, 7 min -33, 19 min 30 /min 225, 7 min Oven equilibrium time 5 min " Valve box temp. 75 " Sample flow 250 ml/min " Sample loop 5 cc " 7

8 In addition, in between each cylinder the reference cylinder A was measured to correct for the drift of the analyzer during the time of the comparison (sequence A-B-A-C-A-...). Two standard mixtures (CC and PC6082, volume of 30 L), of a nominal mole fraction of 100 prepared by KRISS with the gravimetric method were used during the entire comparison period as working standards. To assess the repeatability, the measurement of each sample was repeated three or five times. Considering repeatability, drift and reproducibility, the analyzer response ( y, ) of i-th NMI cylinder is expressed as Eq 3. y J = ( yrespons, i, S /( y y J j ref, i 1, S + j ref, i 1, S ) / 2), (Eq. 3) j responscor r, i + / S j = 1 responscorr i where y, respons i S j, is measurement response of i-th NMI during S j th series measurement; y ref i 1, S j, and y ref i 1, S j, + denote the response of standard reference cylinder of CC (or PC6082) before and after measurement of a i-th NMI to control a drift. S j denotes j-th series measurement and j ranges from 1 to J. It means that a series of measurement was repeated J times for reproducibility before obtaining the final response value. Actually all measurements for NMI cylinders performed 3-4 times from first of March to 13 th of March. From Eq. 3. the analysis uncertainty is obtained. For example of KRISS cylinder, the standard uncertainties were ranged from to during four sequent measurements and its final uncertainty for J=4 was calculated as listed in Table 7. 8

9 Results 1. Argon impurity The mole fraction of argon in nitrogen was measured in all of standard gas mixtures submitted by NMIs so as to examine the level of impurities. (Argon amounts in each cylinder ranged from 0.21 to ) The amount of argon deduced from the ratio between the peak of argon and the peak of oxygen in each chromatogram, as thermal conductivity detectors generally show a very stable linear response. Because the high mole fraction of argon present in the cylinders caused an interference between the argon and oxygen peaks, during experiment by means of GC, the peak of oxygen was explicitly separated by that of argon (Fig. 3) with two analytical conditions as summarized in table u V T C D 2 B, ( \ M E D ) KRISS, Ar = u V T C D 2 B, ( \ N M I A D ) NMIA, Ar = Nitroge O O Ar N2-15, 13.5 min m i n Ar N2-33, 19 min m i n Figure 3: Chromatograms for two gas mixtures (KRISS and NMIA) measured with GC/TCD Table 5 shows the Ar measurement results. Nine cylinders were analyzed with condition I and Argon was measured with mole fractions in the range of , while three cylinders were with condition II and Argon was found with mole fractions greater than 100. During all analysis, the standard deviation (1σ) on four repeats was within the maximum of 0.68 (Table 5). The results of the measurement show complete separation of argon and oxygen by the GC method, so we can conclude that the results for oxygen are not affected by the presence of argon irrespective of the amount of argon present. 9

10 Table 5: Summary for argon amounts in each mixture Cylinder Ar conc.() received from average σ NMIJ NPL BAM CEM NMIA NIST NMISA CENAM LNE KRISS VNIIM VSL

11 2. Results of Oxygen comparison In the comparison a degree of equivalence (D i ) is defined [3,4] as the difference of a measurement result with respect to the key comparison reference value (KCRV) [17,18] and its associated uncertainty. x i = D i = x i,prep x i,kcrv (Eq. 4) where x i,kcrv and x i,prep denotes the key comparison reference value and the preparation value of i-th NMI, respectively. Here a key comparison reference value, x i,kcrv is replaced with x i,ref in this comparison as a matter of convenience. When considering the non independence of the reference value on the contributing laboratories, the covariance between the value submitted by the laboratory and the reference value has to be taken into account for the calculation of u(d i ). Here a contributing laboratories indicates a dataset which participate the KCRV determination. In this case a degree of equivalence is expressed as Eq. 5. x i,adj is completely different from the following x i,pred (Eq. 10) physically in that x i,adj and x i,pred provide the difference based on orthogonal and horizontal projection, respectively. D i = x i,prep - x i,adj. (Eq. 5) In order to obtain x i,adj, a general least square method such that the corresponding sum of squared deviations between each preparation data point ( x y ) and an adjusted data point ( x ˆ, yˆ ), weighted by the inverse squares of the associated uncertainties, is a i i minimum is used as a rule. i, i n x ˆ ˆ i xi yi yi χ = + = min 2 2 (Eq. 6) i= 1 u ( xi ) u ( yi ) Here it is expected the coordinates ( xˆ, yˆ ) in the equation (6) to satisfy the regression relation (Eq. 7). yˆ i a + 1 a2xˆ i i i =, (Eq. 7) which has been chosen as model in this case. Minimizing χ 2 leads to the values for the model parameters a 1 and a 2, which are obtained by computation of Eq. 6 with GLS (general least square). In ISO 6143 [4] the procedures on parameter evaluation are explained. The functional relationship (Eq. 6) indicates that the smaller uncertainty of the submitted value (x prep and u(x prep )), the more weight in the least-square fit. Uncertainty associated with degrees of equivalence can be obtained from the uncertainties of the adjusted and prepared point, taking also into account the covariance 11

12 ( u x i, xˆ ) ) between them. Its expression is: ( i 2 2 u( x) = u ( x ) + u ( xˆ ) 2 u( x, xˆ ). (Eq. 8) i i Only 8 among 12 participants were chosen to contribute to the KCRV calculation, with the consequence that the following data points have been removed: (1) The CEM value, as it had been declared by them to be in error, (2) the NMISA, VNIIM, NMIA values, as they were inconsistent with the others owing to their ambiguity in impurity analysis. i i Table 6 summarizes the results for this comparison. In the table, the mole fraction (x prep ) together with its expanded uncertainty (U(x prep )) which was prepared by each NMI and the analyzer response value (y) along with its standard uncertainty (u(y)) which were measured by coordinating Lab. are presented. The response values have no need to be corrected for any pressure influence, because the pressure of the measurement system has been kept constant by means of the pressure restrictor. The column of U( x) denotes the expanded uncertainty with 95 % confidence level. The adjusted mole fraction (x adj ) of Table 6 has been obtained on the basis of a general linear least square (GLS, XGenLine [5]) regression with the origin point, that is (0 ± 0.01, 0 ± ), where 0.01 and are from table 3. This selection of regression line with the origin point, which was also supported by the internal validation results, helps the regression line not to be affected by some points which are suspected to be highly deviated from the line. Accordingly, the 9 values including both the KCRV subset and an origin are participated in the process of KCRV determination. The results of KCRV subset laboratories are given in bold in table 6. The regression model line of the computation reads as follows: y = ( ± ) x + ( ± ), (N=9, including origin). (Eq. 9) The numbers in the parenthesis represent the standard deviations of the slope and the intercept, respectively. The covariance between parameters is Figure 4 shows the least square regression line and the data points (x prep, y). All bars in all data points in the figure present x-direction and y-direction expanded uncertainties (k = 2), respectively. On the contrary, the degree of equivalence of the other four laboratories, which were not included among the KCRV subset, is derived from the vertical deviation from the regression line of Eq. 9. The results of these laboratories 12

13 are written in table 6 where the uncertainty of the difference is given as 2 2 u ( xi ) = u ( xi, prep) + u ( xi, pred ), where x i, pred y a a i 1 =. (Eq. 10) 2 The deviations between the adjusted points or the predicted values and the preparation values originally submitted and the corresponding uncertainties of their deviations calculated for both contributing and the non-contributing laboratories are shown in Fig. 5. Their uncertainties are given as 95% confidence intervals, as shown in table 6. The result shows that most of the KCRV dataset are consistent with KCRV considering their uncertainties, that is, the criterion (ISO 6143 [4]) x k u x ) (k = 2) (Eq. 11) i ( i is satisfied for 7 data points. The parameter En in the last column of table 6, the deviation divided by its expanded uncertainty, is introduced to examine the agreement with the KCRV. En of table 6 shows that most of the KCRV subset is within the limit of unity, only LNE deviate from the KCRV slightly. In Fig. 5, the uncertainties of deviations of NMIJ and KRISS come to be fairly smaller than the submitted uncertainties because of the covariance term of Eq. 8. Fig. 6 shows the relative deviation and its uncertainty (k=2) of each NMI. The figure indicates that the contributing laboratories (the KCRV subset) agree well with the KCRVs compared to the others. In addition, the figure shows that all residuals in the KCRV subset (except LNE) well satisfy the above consistency criterion (Eq. 11) of ISO 6143 [4]. The degree of equivalence and its uncertainty obtained by this approach can give a typical representation of a preparative and analytical capability when evaluating CMCs in the same way. 13

14 Figure 4: Linear regression of oxygen data; error bar denote U of the point. Figure 5: Graph of equivalence for CCQM-K53 14

15 Table 6: Full data set of CCQM-K53 results x prep U(x prep ) x adj u(x adj ) x pred u(x pred ) x ( D i ) U( x), k=2 NMI y u(y) [] [] NMIJ NPL BAM CEM NMIA NIST NMISA CENAM LNE KRISS VNIIM VSL E n 15

16 Figure 6; Relative degrees of equivalence for CCQM-K53 Conclusion This key comparison focuses a comparison to evaluate the capability for the preparation and the analysis of the stable gas species in nitrogen. The target amount of substance is 100 of oxygen in nitrogen. The oxygen primary standard gas mixtures were prepared gravimetrically through at least two or three steps for dilution with the range of preparation uncertainty, to 0.15 (1 σ). As a result they are consistent with each other within the maximum deviation of The reference value is obtained from the linear regression line (with origin) of a selected set of participants. The KCRV subset except one agree with each other. The standard deviation of the x- residuals of them (which consists of NMIJ, VSL, NIST, NPL, BAM, KRISS, and CENAM) is and well consistent with the uncertainties given to their standard mixtures. The standard deviation of the residuals of all participating laboratory is

17 With respect to impurity analysis, overall Ar amounts of the cylinders are in the range of about 3, however four cylinders showed Ar amount fraction over 10. Two out of them are inconsistent with the KCRV subset. The explicit separation between two peaks of oxygen and argon in the GC chromatogram is essential to hold analytical capability. Additionally oxygen impurity analysis in nitrogen is indispensable to ensure the preparative capability. References [1] Alink A., The first key comparison on Primary Standard gas Mixtures, Metrologia 37 (2000), pp [2] International organization for standardization, ISO 6142, Gas analysis, Preparation of calibration gas mixtures, Gravimetry method, ISO, Second edition, 2001(E) [3] Van der Veen A.M.H., Brinkmann F.N.C., Arnautovic M., Besley L., Heine H.-J., Lopez Esteban T., Sega M., Kato K., Kim J.S., Perez Castorena A., Rakowska A., Milton M.J.T., Guenther F.R., Francy R., Dlugokencky E., International comparison CCQM-P41 Greenhouse gases. 2. Direct comparison of primary standard gas mixtures, Metrologia 44 (2007), Techn. Suppl [4] International organization for standardization, ISO 6143, Gas analysis, Comparison methods for determining and checking the composition of calibration gas mixtures, ISO, Second edition, 2001(E) [5] Milton M.J.T., Harris P.M., Smith I.M., Brown A.S., Goody B.A., Implementation of a generalized least-squares method for determining calibration curves from data with general uncertainty structures, Metrologia 43 (2006), pp. S291-S298 17

18 Annex 1: Measurement raw data of Oxygen Table 7: Raw data of Oxygen for CCQM-K53 Response of measurements NMIs mean σ Conc. prep [] Unc. prep (k=2) [] NMIJ NPL E BAM E CEM E NMIA NIST NMISA CENAM LNE KRISS VNIIM VSL

19 Report form CCQM-K53 (Oxygen in Nitrogen) Date: 18 Oct Laboratory: National Metrology Institute of Japan Prepared 99 oxygen gas Cylinder Number: CPC00641 Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: Results of determinations Purity of nitrogen: mol/mol Component Amount fraction Standard Assumed 10-6 mol/mol Uncertainty distribution Oxygen Normal Argon Rectangle Carbon monoxide Rectangle Carbon dioxide Normal Methane Rectangle Nitrous oxide Rectangle Water Rectangle Nitrogen Normal Purity of oxygen: mol/mol Component Amount fraction Standard Assumed 10-6 mol/mol Uncertainty distribution Nitrogen Rectangle Argon Rectangle Carbon monoxide Rectangle Carbon dioxide Normal Methane Rectangle 19

20 Nitrous oxide Rectangle Water Rectangle Oxygen Normal Gravimetric Preparation Data We prepared 99 Oxygen gas by three steps dilution of pure oxygen gas by pure nitrogen gas with the gravimetric method based on ISO6142. In the first dilution, we prepared oxygen gas (Gas A), which was diluted to 2155 (Gas B) in the next step. In the final step, we prepared 99 (Gas C) by diluting the Gas B. Specification of Balance (Model No., Readability, Resolution, etc.,) Model No. : Mettler-Toledo Model: KA10-3/P Resolution : 1 mg Pooled experimental standard deviation for weighing, s p =3.9 mg, Weighing method (A-B-A, Substitution method, etc.,) A-B-B-A method was used. A and B corresponds to a reference and a sample cylinders, respectively. The sample and reference cylinders were 10 L Aluminum cylinders. The scheme A-B-B-A was automatically repeated 6 times per 1 measurement. The difference of readings between the reference and sample cylinders was always kept within ±1 g by using calibrated standard mass pieces (OIML E 2 class), which were traceable to the National Standards at NMIJ s mass force standard group. Coverage factor: 2, because of ISO

21 In preparation of oxygen gas (Gas A) Weight of oxygen source gas: mg Weight of nitrogen source gas: mg Uncertainty source Mass of nitrogen gas (mg) Mass of Oxygen gas (mg) Molar mass of N 2 (mg/mol) Molar mass of O 2 (mg/mol) Purity of nitrogen gas () Conc. of Oxygen in nitrogen gas () Conc. of nitrogen in Oxygen gas () Purity of Oxygen gas () Estimate, Assumed x i distribution Contribution to Standard Sensitivity standard uncertainty, coefficient uncertainty, u(x i ), c i u(y i ) Normal Normal Rectangle Rectangle Normal Normal Rectangle Normal Other impuritires ca 0.5 in nitrogen and oxygen gas Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: 4.9 Negligible the uncertainty for 21

22 In preparation of 2150 oxygen gas (Gas B) Weight of Gas A: mg Weight of nitrogen source gas: mg Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: 0.31 Uncertainty source Mass of nitrogen gas (mg) Mass of gas A (mg) Molar mass of N 2 (mg/mol) Molar mass of O 2 (mg/mol) Purity of nitrogen gas () Conc. of Oxygen in nitrogen gas () Conc. of nitrogen in Gas A () Conc. of Oxygen in Gas A() Other impuritires in nitrogen and A gas Estimate, x i Assumed Standard distributio uncertainty, n u(x i ) Sensitivity coefficient, c i Contribution to standard uncertainty, u(y i ) Normal Normal Rectangle Rectangle Normal Normal Rectangle Normal ca 0.5 Negligible for the uncertainty 22

23 In preparation of 99 oxygen gas (Gas C) (Cylinder Number: CPC00641) Weight of gas B: mg Weight of nitrogen source gas: mg Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: Uncertainty source Mass of nitrogen gas (mg) Mass of gas A (mg) Molar mass of N 2 (mg/mol) Molar mass of O 2 (mg/mol) Purity of nitrogen gas () Conc. of Oxygen in nitrogen gas () Conc. of nitrogen in Gas B () Conc. of Oxygen in Gas B () Other impuritires in nitrogen and B gas Estimate, x i Contribution to Assumed Standard Sensitivity standard distributio uncertain coefficient uncertainty, n ty, u(x i ), c i u(y i ) Normal Normal Rectangle Rectangle Normal Normal Rectangle Normal ca 0.5 Negligible for the uncertainty 23

24 Report form Laboratory: NPL Cylinder Number: Step One - Preparation of pre-mixture Parent 1= Pure oxygen Amount fraction Uncertainty CH CO CO C3H H H2O N O Parent 2 = Metrology grade nitrogen Amount fraction Uncertainty Ar CO CO CxHy H2O N NO SO Masses Added Parent Mass Uncertainty 24

25 Pure oxygen Metrology grade nitrogen Daughter=Pre-mixture Amount fraction Uncertainty CH CO CO C3H H H2O N O Ar CxHy NO SO Step Two (Preparation of final mixture) Parent1=Pre-mixture (as above) Amount fraction Uncertainty CH CO CO C3H H H2O N O Ar CxHy NO

26 SO Parent2=BIPplus grade Nitrogen Amount fraction Uncertainty Ar CO CxHy N CH Masses Added Parent Mass Uncertainty 1 Pre-mixture Nitrogen Final mixture Amount fraction Uncertainty CH CO CO C3H H H2O N O Ar CxHy NO SO All uncertainties given above are expanded with k=2. 26

27 Final result = / (k=2) oxygen in nitrogen Expanded uncertainty = 0.097% (relative) 27

28 Report form Laboratory: Bundesanstalt fuer Materialforschung und pruefung (BAM) Cylinder Number: BAM 6047 Amount fraction of oxygen: 99,17 Balance gas Nitrogen Purity of nitrogen: 10-6 mol/mol Component Amount fraction 10-6 mol/mol Uncertainty O 2 0,18 15 %rel. Ar <0,5 (not determined) H 2 0 <0,5 (not determined) CO <0,1 (not determined) CO 2 <0,1 (not determined) KW <0,1 (not determined) Nitrogen 99,9999 2*10-4 %rel Purity of oxygen: 10-6 mol/mol Component Amount fraction 10-6 mol/mol Uncertainty N 2 0,18 15 %rel. Ar <1 (not determined) H 2 0 <0,5 (not determined) CO <0,1 (not determined) CO 2 <0,1 (not determined) KW <0,1 (not determined) 28

29 Oxygen 99,9999 2*10-4 %rel Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Voland HCE25, Weighing method (A-B-A, Substitution method, etc.,) : According to ISO 6142 using a two arm balance (difference weighing: Reference- and Calibration gas cylinder. 3 pre-mixtures were used. Weight of oxygen source gas: each circa 150 g Weight of nitrogen source gas: each circa 1600 g Amount fraction of oxygen: 99,17 Balance gas Nitrogen Coverage factor: 2 Expanded uncertainty: 0,1 (without the uncertainty arises from the verification process and stability 0,25 (included the uncertainty arises from the verification process and stability) 29

30 Preparation scheme: O 2 N 2 210,827 g 1651,775 g Pre mixture 1 / BAM O N 2 181,098 g Pre mixture 2 / BAM O N ,841 g 185,601 g Pre mixture 3 / BAM O ,292 g N 2 182,778 g Final mixture BAM ,17 O 2 Bilance: N 2 + Impurities 1684,492 g 30

31 Uncertainty source Estimate Assumed x i distribution Standard uncertainty u(x i ) Sensitivity codfficient c i Contribution to standard uncertainty u(y i ) Balance (A) 6 mg n 6 mg 1 6 mg Handling of cylinder (B) 4 mg n 4 mg 1 4 mg Imputities in O 2 (B) 2*10-4 %rel rectangle - O 2 in N 2 (A) 0,03 rectangle 1 0,06 mg Impurities in 2*10-4 %rel rectangle - N 2 (B) Composition 3*10-4 %rel n 1 0,5 mg of the pre mixtures to 6*10-4 %rel to 0,1 mg Verification (A) 0,2 n 1 0,1 Stability (B) <0,1 rectangle 1 0,1 / 2 years (not verifiable) 31

32 Report form Laboratory: Centro Español de Metrología (CEM) Cylinder Number: Purity of nitrogen: GC (Quality) Component Uncertainty Amount fraction mol/mol mol/mol (K=2) CO 0,50 0,58 CO 2 0,50 0,58 H 2 1,00 1,16 O 2 0,01 0,01 H 2 O 0,02 0,02 CH 4 0,10 0,12 Component Amount fraction 10-2 mol/mol Uncertainty 10-6 mol/mol (K=2) Nitrogen 99,9998 1,42 Purity of oxygen: Component Uncertainty Amount fraction mol/mol mol/mol (K=2) CO 0,50 0,58 CO 2 0,50 0,58 H 2 1,50 1,73 N 2 4,00 4,62 H 2 O 1,50 1,73 HC 0,50 0,58 Component Amount fraction 10-2 mol/mol Uncertainty 10-6 mol/mol (K=2) Oxygen 99,9992 5,32 32

33 Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Mass comparator, Mettler Toledo PR10003, maximum range: g; repeatability: 2 mg; resolution: 1 mg; linearity: 10 mg; E2 masses are used. The cycle of weighing is as follows (two differents cases): If the weight of the reference is bigger than the weight of sample: 1.- Reference + 1 g 2.- Reference 3.- Sample + weights 4.- Reference 5.- Sample + weights 6.- Reference 7.- Sample + weights 8.- Reference 9.- Reference + 1 g If the weight of the reference is smaller than the weight of sample: 1.- Reference + 1 g + weights 2.- Reference + weights 3.- Sample 4.- Reference + weights 5.- Sample 6.- Reference + weights 7.- Sample 8.- Reference + weights 9.- Reference + 1 g + weights The weighing data are collected directly by software, with temperature, pressure and humidity values. 33

34 This software also collect the raw material data (O 2 and N 2 ) and with all those date the software calculates the uncertainty of the mixture following ISO 6142, but as it is calculated by the software, it is not possible to fill the table of each uncertainty contribution. Weight of oxygen source gas: 69,963 8 g Weight of nitrogen source gas: 628,334 6g Amount fraction of oxygen: 100,04 Coverage factor: k = 2 Expanded uncertainty: 0,023 34

35 Report form Laboratory: National Measurement Institute Australia (NMIA) Report for CCQM K53: Oxygen in Nitrogen Cylinder Number: ME2630 Source Gases: Purity of nitrogen: mol/mol Component Amount fraction 10-6 mol/mol Uncertainty Oxygen Water Methane Nitrogen Purity of oxygen: mol/mol Component Amount fraction 10-6 mol/mol Uncertainty Argon Nitrogen Water Methane Oxygen Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Mass Comparator: Sartorius CC10000S. Resolution 0.1mg. Readability 0.1mg Reproducibility 0.25mg Uncertainty of 0.5mg 35

36 Weighing method (A-B-A, Substitution method, etc.,) An ABBA weighing scheme is used to minimise the effect of instrument and environmental drift. This involves weighing a reference mass (cylinder) prior to and after two weighings of the sample mass. During the weighing process, the air temperature, pressure and relative humidity are measured during each reading on the mass comparator to calculate the air density. The air density is used to correct for changes in the buoyancy of cylinders and rings. Dilution Stages: First stage dilution (cylinder ME2635): Mass of oxygen source gas: g Mass of nitrogen source gas: g Amount fraction of oxygen: mmol/mol Coverage factor: 2 Expanded uncertainty: mmol/mol Second stage dilution (cylinder ME2633): Mass of oxygen source gas: g Mass of nitrogen source gas: g Amount fraction of oxygen: mmol/mol Coverage factor: 2 Expanded uncertainty: mmol/mol Final Stage dilution (cylinder ME2630): Mass of oxygen source gas: g Mass of nitrogen source gas: g Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty:

37 37

38 Report form Laboratory: NIST Cylinder Number: CAL Purity of nitrogen: 10-6 mol/mol Component Amount fraction 10-6 mol/mol Uncertainty Oxygen Argon 1 1 Water 4 4 CO Nitrogen Purity of oxygen: 10-6 mol/mol Component Amount fraction 10-6 mol/mol Uncertainty Nitrogen Argon 5 5 CO CO Water 4 4 Oxygen

39 Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) 1) Mettler B5C1000, Readability 0.1 mg, Capacity 1000 g, uncert (k=1) 0.2 mg 2) Voland HCE 10, Readability 1 mg, Capacity 15 Kg, uncert (k= 1) 9 mg 3) Mettler Readability 0.01 g, Capacity 10.1 Kg, uncert 0.03 g Weighing method (A-B-A, Substitution method, etc.,) Substitution method, reference cylinder (A-B-A) Mixture 1: Weight of oxygen source gas: g Weight of nitrogen source gas: g Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: 1.1 Uncertainty source Estimate Assumed x i Distribution Standard uncertainty u(x i ) Sensitivity coefficient c i Contribution to standard uncertainty u(y i ) Mass Parent Gas normal Mass Balance Gas normal MW O normal MW N normal Impurities in balance 69 µmol Rectangular 57 µmol Impurities in oxygen 0.13 µmol Rectangular 0.09 µmol x 10-7 Mass O2 from N2 2.2 µg Rectangular 2.2 µg Mass N2 from O2 3.9 µg Rectangular 3.9 µg x

40 Mixture 2: Weight of oxygen gas: g Weight of nitrogen source gas: g Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: 0.12 Uncertainty source Estimate Assumed x i Distribution Standard uncertainty u(x i ) Sensitivity coefficient c i Contribution to standard uncertainty u(y i ) Mass O2 from Parent Gas Normal Mass N2 from Balance Gas Normal Mass N2 from Parent Normal Total N2 added Normal MW O Normal MW N Normal Impurities in Balance 470 µmol Rectangular 12 µmol x 10-8 Impurities in Oxygen 3.9 µmol Rectangular 3.1 µmol x 10-9 Mass O2 from N2 4.4 µg Rectangular 4.4 µg

41 Report form Laboratory: NMISA Cylinder Number: 8396 Purity of nitrogen: ,375 x 10-6 mol.mol -1 Component Amount fraction x 10-6 mol.mol -1 CO 0,0215 0,0248 CO 2 0,050 0,019 H 2 0,5 0,58 H 2 O 0,01 0,012 HC (Hydrocarbons) 0,05 0,058 O 2 0,005 0,0058 Nitrogen ,375 0,822 Uncertainty (k=2) Purity of oxygen: ,75 x 10-6 mol.mol -1 Component Amount fraction x 10-6 mol.mol -1 CO 0,25 0,29 CO 2 0,5 0,58 H 2 O 1,5 1,73 HC (Hydrocarbons) 0,25 0,29 N 2 21,77 3,31 Oxygen ,75 5,377 Uncertainty (k=2) Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Mettler Toledo PR 10003, Readability: 0,01 mg, Resolution: 0,01 mg Weighing method (A-B-A, Substitution method, etc.,) Substitution method Weight of oxygen source gas: 67,51552 g 41

42 Weight of nitrogen source gas: 608,84153 g Amount fraction of oxygen: 100,585 x 10-6 mol.mol -1 Coverage factor: 2 Expanded uncertainty: 0,022 x 10-6 mol.mol Preparation Data - Weights of nitrogen and oxygen in the dilution series First dilution step: 10% mol.mol -1 Component Weight (g) Combined Standard uncertainty (mg) O 2 77, ,93 N 2 611, ,78 Second dilution step: 1% mol.mol -1 Component Weight (g) Combined Standard uncertainty (mg) O 2 68, ,42 N 2 604, ,55 Third dilution step: 1000 x 10-6 mol mol -1 Component Weight (g) Combined Standard uncertainty (mg) O 2 68, ,55 N 2 612, ,70 Final dilution step: 100 x 10-6 mol mol -1 Component Weight (g) Combined Standard uncertainty (mg) O 2 67, ,15 N 2 608, ,11 42

43 2. Example of uncertainty sources in gravimetric preparation Weighing data for 100 x 10-6 mol mol -1 mixture i. Vacuum weighing Standard Uncertainty Type Degrees uncertainty Sensitivity contribution (c x A/B of Parameter Estimate (u) coefficient (c) u) freedom Sensitivity 0, , , , A 1 Weighing difference 0, , , , A 2 Mass pieces 94, , , ,42738 x 10-5 B infinity Air density 1, , B infinity Volume expansion 0, ,52548 x , ,70434 x 10-6 B infinity Density of mass pieces (stainless B steel) ,002-1,53242 x ,06485 x 10-9 infinity Mass (g) 95, Standard uncertainty (mg) 2, ii. Weighing after oxygen addition Uncertainty Type Degrees Standard Sensitivity contribution (c A/B of Parameter Estimate uncertainty (u) coefficient (c) x u) freedom Sensitivity 0, , , , A 1 Weighing difference 0, , , , A 2 Mass pieces 162, , , ,72203 x 10-5 B infinity Air density 1, , B infinity Volume expansion 0, ,25559 x , ,51141 x 10-6 B infinity Density of mass pieces (stainless B steel) ,002-2,61727 x ,23453 x 10-9 infinity Mass (g) 163,

44 Standard uncertainty (mg) 2, iii. Weighing after nitrogen addition Standard Uncertainty Type uncertainty Sensitivity contribution (c x A/B Degrees of Parameter Estimate (u) coefficient (c) u) freedom Sensitivity 0, , , , A 1 Weigh difference 0, , , , A 2 Mass pieces 771, , , , B infinity Air density 1, , , ,05545 x 10-5 B infinity Volume expansion 0,015 0,003 1, , B infinity Density of the B mass pieces (stainless steel) ,24288 x ,48576 x 10-8 infinity Mass (g) 772, Standard uncertainty (mg) 4, iv. Purity table of 100 x 10-6 mol.mol -1 mixture Component Amount fraction Standard uncertainty Expanded uncertainty (x10-6 ) mol.mol -1 (x10-6 ) mol.mol -1 (x10-6 ) mol.mol -1 CO 0, , , CO 2 0, , , H 2 0, , , H 2 O 0, , , HC (Hydrocarbons) 0, , , N , , , O 2 100, , ,

45 Coverage factor : 2 ; level of confidence = 95,45% ; ν = Expanded uncertainty : 0,02247 (x10-6 ) mol.mol -1 eff 45

46 Report form Laboratory: CENAM Cylinder Number: FF39563 Purity of nitrogen: , mol/mol Component Amount fraction 10-6 mol/mol Uncertainty O 2 0,15 0,10 CO 0,11 0,08 CO 2 0,30 0,13 CH 4 0,100 0,058 H 2 O 0,50 0,31 Nitrogen ,84 0,19 Purity of oxygen: , mol/mol Component Amount fraction 10-6 mol/mol Uncertainty N 2 0,54 0,15 CO 1,0 0,58 CO 2 1,0 0,58 CH 4 0,292 0,086 H 2 O 4,90 2,80 Ar 0,243 0,046 THC 0,1 0,058 H2 0,1 0,058 Kr 1,0 0,58 Xe 1,0 0,58 He 0,1 0,058 Oxygen ,77 3,04 46

47 Gravimetric Preparation Data The preparation of the standard gas mixture was carried by at least three steps dilution as show in the diagram: O 2 N 2 O 2 N Mezcla 1,5 cmol/mol O 2 /N N 2 Mezcla 1,5 cmol/mol O 2 /N 2 1 Mezcla 800 Mezcla Mezcla 1200 O 2 /N 2 O 2 /N 2 O 2 /N N 2 Mezcla 80 O2/N2 Mezcla 90 O2/N2 Mezcla 101 O2/N2 Mezcla 110 O2/N2 Mezcla 101 O2/N2 Mezcla 120 O2/N2 Preparation of the cylinders before the filling The used cylinders to prepare the mixtures are aluminum made type ALS, they were provided by the Praxair Mexico (manufacturer LUXFER). Praxair Mexico carry out routine tests on the cylinders: Vent hydrostatic pressure, leaks, humidity analysis, passivation, vacuum level, and cleaning. Before weighing each cylinder The external surface and valve of each cylinder was cleaned using a cotton cloth with alcohol to eliminate dust or remainders of the cylinder. It was carried out a vacuum to each cylinder < 0,7 Pa, using the filling panel of the gas mixture preparation system. The cylinders with low pressure and cleaned were introduce to the gas mixture preparation lab to let acclimate to reach the laboratory temperature (24 h). Filling of the mixtures Previous to the filling it was come to weigh the clean cylinders, selecting as a reference cylinder the one that had similar weight. Later was made the filling in agreement with CENAM internal procedures. Determination of mass by means of weight 47

48 The calibration standards for the measurements were primary standards (primary standard mixtures, PSMs), prepared by weigh, the cylinders were weighted after each compound addition and thermal equilibrium with the room. The method used for the preparation of PSMs was the gravimetric method following the guidelines of the ISO/DIS The procedure for weighing was a Borda weighing scheme (RTRTRTR). CENAM experts prepared the gas mixtures at the gas producer facilities (Praxair). Specification of Balance The instrument for weighing was a Mettler balance model KB-50 (60 kg capacity and 0,05 g resolution) and sets of weights class E2 (serial number , from 1 to 5 kg 4 pieces) and E2 (serial number , from 1 mg to 1 kg 25 pieces) according to the R 111 of OIML, all of them traceable to SI by CENAM s Standards. Weight of oxygen source gas: 95,90 g Weight of nitrogen source gas: 854,01 g Amount fraction of oxygen: 100,97 Coverage factor: k = 2 *Expanded uncertainty: 0,30 *Expanded uncertainty: It was obtained by the product of the combined standard uncertainty and a coverage factor of 2 at 95% level of confidence Uncertainty budget submitted of the gas mixture. The uncertainty of the standard mixture has been calculated applying the law of propagation of uncertainties to the equation 3 of the ISO 6142 (2001). The uncertainty contributions taken in account are following: A.- The uncertainty in molar mass B.- The uncertainty on the weighing C.- The uncertainty in the purity analysis. A. The uncertainty of molar masses has been taken as a type B evaluation of uncertainty from the IUPAC assuming a uniform distribution. B. The uncertainty of weighing has been calculated using the law of propagation of 48

49 uncertainties to the model for single substitution weighing: m T = m R + ρ a * (V T - V R ) + Ιω / S where: m T = mass of the tested cylinder m R = reference mass ρ a = air density V T = weights used in the tested cylinder V R = weights used in the reference cylinder Ιω = difference in balance readouts S = balance sensitivity C. The uncertainty in the purity analysis. Their uncertainties were calculated by type A evaluation or/and type B evaluation. Uncertainty budget mixture 49

50 Uncertainty source Estimate Standard uncertainty Assumed Distribution Sensitivity coefficient Contribution to standard uncertainty X I x I u(x i ) C I u I (y) MH2O Rectangular -9.02E E-33 MN Rectangular 3.70E E-26 MTHC Rectangular -1.62E E-30 MH Rectangular -3.23E E-41 MCO Rectangular -5.75E E-32 MCO Rectangular -1.52E E-31 MAr Rectangular -7.85E E-38 MCH Rectangular 3.23E E-33 MKr Rectangular -3.23E E-35 MXe Rectangular -3.23E E-34 MHe Rectangular -3.23E E-44 MO Rectangular -3.23E E-25 mo Normal 9.45E E-15 mn Normal -1.06E E-17 XH2O,O Normal -5.83E E-24 XN2,O Normal -9.06E E-20 XTHC,O Rectangular -1.42E E-22 XH2,O Rectangular -6.52E E-32 XCO,O Rectangular -9.06E E-23 XCO2,O Rectangular -1.42E E-23 XAr,O Normal -1.29E E-30 XCH4,O Normal -5.19E E-30 XKr,O Rectangular -2.71E E-27 XXe,O Rectangular -4.25E E-26 XHe,O Rectangular -1.29E E-31 XO2,O Rectangular E-14 XO2,N Normal E-15 XH2O,N Normal 5.83E E-23 XCO,N Normal 9.06E E-23 XCO2,N Normal 1.42E E-23 XCH4,N Rectangular 5.19E E-24 XN2,N Rectangular 9.06E E-22 CENAM Participants List: Francisco Rangel Murillo, Victor M. Serrano Caballero, Alejandro Pérez Castorena 50

51 Report form Laboratory: LNE Cylinder Number: Purity of nitrogen: mol/mol Component Amount fraction 10-6 mol/mol Uncertainty O H 2 O CO+CO THC H Nitrogen Purity of oxygen: mol/mol Component Amount fraction 10-6 mol/mol Uncertainty H 2 O N C n H m CO H CO NO x Oxygen

52 Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Mettler balance : - ID5 15kg at 2 mg/ PR g at 0.1 mg Weighing method (A-B-A, Substitution method, etc.,) -weighing empty small cylinder + standards mass (10 weighing) -weighing small cylinder with parent mixture (10 weighing) -weighing empty cylinder + standards mass (10 weighing) -weighing cylinder with parent mixture + nitrogen (10 weighing) Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: 0.10 Uncertainty source Weight of small cylinder with O2/N2 Weight of empty small cylinder with standard weight Weight of cylinder with O2/N2 and N2 Weight of cylinder with standard weight Estimate Assumed x i distribution Standard uncertainty u(x i ) Sensitivity codfficient c i Contribution to standard uncertainty % σ(exp) σ(exp) σ(exp) σ(exp) N2 molar mass U=ku O2 molar mass U=ku N2 purity U=ku Standard weight U=ku Amount fraction U=ku Standard weight U=ku Air density U=ku Air density U=ku Resolution balance 1 0 uniform Resolution balance 2 0 uniform

53 CCQM-K53 (Oxygen in Nitrogen) Report form Date: 16 Dec 2006 Laboratory: Korean Research Institute of Standards and Science (KRISS) Prepared 101 oxygen gas Cylinder Number: ME5697 Amount fraction of oxygen: Coverage factor: 2 Expanded uncertainty: Results of determinations Nitrogen source gas: %mol/mol Component Amount fraction (10-6 mol/mol) Standard uncertainty (10-6 mol/mol) Assumed distribution Hydrogen Rectangular Oxygen Normal Carbon monoxide Normal Carbon dioxide Rectangular Methane Normal Argon Normal Water Normal Nitrous oxide Rectangular Hydrocarbons(CxHy) Rectangular Neon Normal Nitrogen Normal 53

54 Oxygen source gas: %mol/mol Component Amount fraction (10-6 mol/mol) Standard uncertainty (10-6 mol/mol) Assumed distribution Hydrogen Rectangular Nitrogen Normal Carbon monoxide Normal Carbon dioxide Normal Methane Rectangular Argon Rectangular Water Normal Oxygen Normal Gravimetric Preparation Data Primary standard gas mixture was prepared gravimetrically through three steps dilution according to ISO6142. Specification of Balance (Model No., Readability, Resolution, etc.,) Model No. : Mettler-Toledo Resolution : 1 mg, Capacity : 10 Kg Uncertainty (k=2) : 3.2 mg Weighing method (A-B-A, Substitution method, etc.,) Substitution method, tare cylinder (A-B-A) Fig. 1 shows the diagram where the preparation of the standard gas mixture was carried by at least three steps of dilution. 54

55 Pure N2 Pure O2 YA YA YA Pure N2 Pure N2 YA YA YA Pure N2 Pure N2 ME5662 ME5704 ME5697 ME5648 Figure 3 Diagram of the preparation of the standard gas mixture Preparation of 4 % mol/mol oxygen in nitrogen (Parent gas A) Weight of oxygen source gas : g Weight of nitrogen source gas : g Amount fraction of oxygen : %mol/mol Coverage factor : 2 Expanded uncertainty : 2.80 Uncertainty source Estimate Standard uncertainty Assumed distribution Sensitivity coefficient Contribution to standard uncertainty Mass of O 2 (g) Normal Mass of N 2 (g) Normal Purity of pure O 2 gas (mol/mol) Purity of pure N 2 gas (mol/mol) Impurity of O 2 in N 2 gas () Atomic weight of O (g/mol) Normal Normal Normal Normal

56 Atomic weight of N (g/mol) Uncertainty of balance(mg) Normal Normal Preparation of 0.2 %mol/mol oxygen in nitrogen (Parent gas B) Weight of parent gas A : g Weight of nitrogen gas : g Amount fraction of oxygen : %mol/mol Coverage factor : 2 Expanded uncertainty : Uncertainty source Mass of parent gas A (mg) Mass of N 2 diluent gas (mg) Conc. of O 2 in parent gas A (%mol/mol) Conc. of N 2 in parent gas A (%mol/mol) Purity of pure N 2 gas (mol/mol) Impurity of O 2 in pure N 2 gas () Impurity of pure O 2 gas () Estimate Standard Assumed Sensitivity Contribution to uncertainty distribution coefficient standard uncertainty Normal Normal Normal Normal Normal Normal Rectangular Impurity of pure N gas except O Rectangular () Molecular weight Normal

57 of O 2 (g/mol) Molecular weight Normal of N 2 (g/mol) Uncertainty of Normal balance (mg) Preparation of 101 oxygen in nitrogen (Standard gas) Weight of parent gas B : g Weight of nitrogen gas : g Amount fraction of oxygen : Coverage factor : 2 Expanded uncertainty : Standard Assumed Sensitivity Contribution to Uncertainty source Estimate uncertainty distribution coefficient standard uncertainty Mass of parent gas Normal 0.28 B(g) Mass of N Normal diluent gas (g) Conc. of O 2 in parent gas B Normal (%mol/mol) Conc. of N 2 in parent gas B Normal (%mol/mol) Purity of pure N Normal gas (mol/mol) Conc. of O 2 in pure N 2 gas Normal 1.0 () Impurity of pure Rectangular O 2 gas () Impurity of pure N 2 gas except O Rectangular () Molecular weight Normal

58 of O 2 (g/mol) Molecular weight of N 2 (g/mol) Uncertainty of balance (mg) Normal Normal

59 Key Comparison CCQM-K53 Oxygen in Nitrogen Report Laboratory: VNIIM, Research Department for the State Measurement Standards in the field of Physico-Chemical Measurements Cylinder Number: D Purity of nitrogen: , mol/mol Component Standard Amount fraction 10-6 uncertainty mol/mol 10-6 mol/mol Hydrogen 0,05 0,006 Oxygen 0,37 0,023 Argon 2,8 0,12 Water 2,0 0,12 Carbon dioxide 1,1 0,17 Carbon monoxide 0,25 (< 0,5) 0,14 Methane 0,25 (< 0,5) 0,14 Nitrogen ,2 0,3 Purity of oxygen: , mol/mol Component Standard Amount fraction 10-6 uncertainty mol/mol 10-6 mol/mol Hydrogen 0,015 0,0052 Argon 1,0 0,06 Nitrogen 4,2 0,35 Krypton 0,06 0,006 Methane 1,2 0,06 Carbon dioxide 0,5 (< 1) 0,17 Xenon 0,005 (< 0,01) 0,

60 Water 3,8 0,23 Oxygen ,2 0,5 Gravimetric Preparation Data Specification of Balance Model: 81-V-HCE-20kg, hnu-voland, USA; Readability: 1 mg; Experimental standard deviation for 10 l cylinder: 5,7 mg. Weighing method - Substitution method. Weight of oxygen source gas: 0,13259 g Weight of nitrogen source gas: 1144,002 g (Preparation of the gas mixture was carried out in 2 stages. Amount fraction of oxygen in the pre-mixture X=0, mol. %; u(x)=0, mol. %) Amount fraction of oxygen: 101,08 Coverage factor: 2 Expanded uncertainty: 0,14 Uncertainty source Assume Standard Estimate d uncertain x i distributi ty on u(x i ) Contribution Sensitivity to standard coefficient uncertainty c i u(y i ), m O2 Preparation 12,908 g Normal 5,7 mg 7, ,0439 of the premixture Preparation of the final m N2 m pre-mixture 1125,392 g 5,7 mg 8, ,0005 Normal 11,590 g Normal 5,7 mg 8, ,0490 mixture m N2 1132,543 g Normal 5,7 mg 1, ,0009 Purity of O 2 in N 2 0,37 0, ,023 gases Normal 60

61 Other impurities 6,45 Normal - - 0,00001 Molar masses Combined uncertainty in N 2 Impurities 10,8 in O 2 standard are not taken into account Normal - - 0, ,

62 Report form Laboratory: VSL Cylinder Number: D Purity tables pure gases: 62

63 Gravimetric Preparation Data Specification of Balance (Model No., Readability, Resolution, etc.,) Sartorius 10 kg comparator scale, repeatability 1 mg. Use pooled estimate for 63