Article pubs.acs.org/ac
Cite This: Anal. Chem. 2017, 89, 12068-12075
Gravimetric Standard Gas Mixtures for Global Monitoring of Atmospheric SF6 Jeong Sik Lim,*,†,‡ Jinbok Lee,† Dongmin Moon,† Jin Seog Kim,†,‡ and Jeongsoon Lee†,*,‡ †
Center for Gas Analysis, Korea Research Institute of Standard and Science (KRISS), Gajeong-ro 267, Yuseong-gu, Daejeon 34113, Republic of Korea ‡ Science of Measurement, Korea University of Science and Technology (UST), Gajeong-ro 217, Yuseong-gu, Daejeon 34113, Republic of Korea
Bradley D. Hall§ §
Global Monitoring Division, National Oceanic and Atmospheric Administration, 325 Broadway, Boulder, Colorado 80305, United States S Supporting Information *
ABSTRACT: In this study, standard gas mixtures of SF6 in synthetic air were gravimetrically developed as a suite consisting of 6 mixtures with mole fractions of SF6 ranging from 5 to 15 pmol/mol. For precision in weighing the gas fills, an automatic weighing system coupled with a high sensitivity mass balance was used and a gravimetry precision of 3 mg (2σ) was achieved. Impurity profiles of the raw gases were determined by various analyzers. In particular, sub pmol/mol levels of SF6 in the matrix components (N2, O2, and Ar) were carefully measured, since the mole fraction of SF6 in the final step can be significantly biased by this trace amount of SF6 in the raw gases of the matrix components. Gravimetric dilution of SF6 by purity-assessed N2 was performed in 6 steps to achieve a mole fraction of 440 pmol/mol. In the final step, O2 and Ar were added to mimic the atmospheric composition. Gravimetric fractions of SF6 and the associated standard uncertainty in each step were computed according to the ISO 6142 and JCGM 100:2008, respectively, and validated experimentally. Eventually, the SF6 fraction uncertainty of the standard gas mixtures combined by uncertainties of gravimetric preparation and verification measurements were found to be nominally 0.08% at a 95% confidence interval. A comparison with independent calibration standards from NOAA shows agreement within 0.49%, satisfying the extended WMO compatibility goal, 0.05 ppt.
A
when using measurements obtained by different laboratories, even small calibration differences can lead to measurement bias unless taken into account. The WMO Global Atmosphere Watch (GAW) recommends a compatibility of 0.02 pmol/mol for the remote troposphere and 0.05 pmol/mol for regional studies.9 A number of laboratories associated with the WMO GAW make routine measurements of SF6 in the troposphere. SF6 is also used as a tracer of oceanic currents, comparing SF6 measured in seawater to the atmospheric history. 10 SF6 measurements are typically calibrated against gas standards prepared by gravimetric methods (e.g., the AGAGE method).11 Calibration differences among independent groups ranged from 0.1% to 10% in one international comparison exercise.12 While that study showed excellent agreement between three independent measurement programs (NOAA Global Monitoring Division, Advanced Global Atmospheric Gases Experiment
nthropogenic sulfur hexafluoride (SF6) survives in the atmosphere for an extremely long period of time, about 900 years,1,2 and exhibits a global warming potential of 22800.3 Its contribution to the total anthropogenic radiative forcing is on the order of 0.1%, but its long environmental lifetime causes it to be one of the most potent greenhouse gases.4 Though ambient concentration levels of SF6 have been reported to be a few parts per trillion (ppt), they have been shown to steadily increase over time.5,6 As a dielectric gas fluid for high-voltage switching gears and an etching agent used for semiconductors and display production processes, the inorganic gas serves many purposes which lead to considerable emissions as a result of unintended leakage and imperfect disposal.6 Measurements of SF6 in the atmosphere have been used to estimate emissions5 and improve atmospheric model transport and as a transport “clock” to estimate the mean “age of air”.7,8 Atmospheric gradients of SF6 tend to be small in the remote troposphere (0.3−0.4 pmol/mol, pole to pole). In order to resolve small gradients, the measurement precision must be comparably small (e.g., 0.05 pmol/mol or better). Further, © 2017 American Chemical Society
Received: June 30, 2017 Accepted: October 13, 2017 Published: October 13, 2017 12068
DOI: 10.1021/acs.analchem.7b02545 Anal. Chem. 2017, 89, 12068−12075
Article
Analytical Chemistry
Rather, if the SF6 concentrations are higher than a sub pmol/ mol level then the SF6 impurity in O2, N2, and Ar can bias a gravimetrically determined mole fraction of SF6 originating from the raw SF6 gas. To carefully examine the LOD of the conventional GC-μECD for the assessment of SF6 impurity in N2, O2, and Ar, a sensitivity enhancement should be accomplished in order to reach the fmol/mol (1 fmol = 10−15 mol) detection level, since the LOD for SF6 was found to be 4.0 pmol/mol. The LOD of the dedicated GC-μECD system were significantly improved by preconcentrating the SF6 impurity in the raw gases (ST2). It is worth noting that socalled O2 interference effect in the conventional GC-μECD was strongly suppressed by using adsorptive preconcentration instrument (Figure S1), thanks to the chemical selectivity of the adsorbent Carboxen 1000 (40−60 mesh, Supelco), since the μECD is not exposed to enough oxygen to significantly influence the SF6 response. This suggests that analytical precision to O2, following the eluent such as SF6 in an activated alumina F1 (AA-F1) column and N2O in a Porapak Q (PP-Q) column, can be improved compared to the whole air injection schemes.11,22 To investigate the LOD of the preconcentrator-GC-μECD system, 6 × 10−2 pmol/mol of an SF6/air standard was dynamically diluted by adjusting the ratio of the flow rates of the standard and the SF6-free N2 (99.9999%, Deokyang Energen Co.). As shown in Figure 1,
(AGAGE), and the University of Heidelberg), the 0.02 pmol/ mol compatibility goal remains a challenge. To ensure that primary standards play a minimal role in measurement uncertainty and compatibility, a minimum 4-fold lower uncertainty is needed at the primary standard level. Even at this level, measurement bias should be investigated and documented. Recent work in gas metrology has led to improvements in compressed gas standards for major greenhouse gases such as CO2, CH4, N2O, and carbon isotopes.13−18 In this study, we report the gravimetric preparation of SF6/air standard gas mixtures ranging from 6 to 15 pmol/mol by the Korean Research Institute of Standards and Science (KRISS). A fully synthetic preparation of the standard gas mixture affords the opportunity to manipulate the compositions to design a suite encompassing background SF6 concentrations for every regional and seasonal variation. Impurities of the raw gases were precisely assessed, even for concentration-biasing components such as extremely low levels of SF6 in N2, O2, and Ar. SF6 mole fraction uncertainties were evaluated according to ISO 6142− 1:201519 and the Guide to the Expression of Uncertainty in Measurement (GUM).20 Two of the KRISS primary standards were analyzed by NOAA. The comparison revealed excellent agreement between two sets of independent primary standards.
2. EXPERIMENTAL SECTION 2.1. Impurity Assessment of Raw Gases. In order to assess the impurity profiles of raw gases used in this study, various detectors coupled to a gas chromatograph (GC) and a water vapor analyzer were employed. An atomic capture detector (AED), pulsed discharge ionization detector (PDD), and microelectron capture detector (μECD) served as detectors for recording chromatograms. Details regarding the analyzers applied in this study are given in the Supporting Text 1, ST1, and Figure S3. According to ISO 19229:2015, the purity, defined as the fractions of the dominant components in the raw SF6 gas, is determined according to eq 1: N
xpure = 1 −
∑ xi i=1
(1)
where N is the number of impurities likely to be present in the final mixture and xi is the fraction of impurity i determined by a purity assessment. Source generation and its purification process should be considered when selecting target impurities.21 If the expected impurity is not detected, its mole fraction is set to half the limit of detection (LOD, 3 × signal/noise), and the associated standard uncertainty is defined as the assigned mole fraction divided by √3 [i.e., LOD/(2·√3)], rectangular distribution, referring to the uniform probability density function ranging from 0 to the LOD. The impurity profile of raw SF6 gas is tabulated in Table S1, showing the impurity concentrations associated with the uncertainties and dedicated detectors. Considering an SF6 dilution factor of approximately 10−12 in the final dilution step (seventh), a few tens or hundreds of μmol/mol of N2, O2, and Ar impurities present in the raw SF6 gas, which are diluted to the amol/mol (1 amol = 10−18 mol) level at the final stage, cannot be distinguished from the N2, O2, and Ar matrix components within the total gravimetric uncertainty requirement, 0.05% (2σ) (vide infra). Therefore, it can be assumed that the mole fractions of N2, O2, and Ar in the final mixtures and their associated uncertainties are negligibly affected by the amount of N2, O2, and Ar present in the raw SF6 gas (Matheson Trigas).
Figure 1. Trace SF6 analysis of raw O2 (magenta), N2 (red), and Ar (green) gases. The coolant temperature (Tc) was set to −80 °C. A non-negligible amount of O2 from the diluted standard gas mixture (blue) and raw O2 (magenta) was concentrated to be injected to the GC-μECD, resulting in long tails in the chromatograms. Nevertheless, it is clearly shown that SF6 was not present in O2, N2, and Ar.
the LOD of the system equipped with the AA-F1 was found to be approximately 0.002 pmol/mol. Therefore, 0.001 pmol/mol of SF6 in N2 translates to only 0.0008 pmol/mol SF6 in the seventh dilution step. On the basis of the uncertainty evaluation of the gravimetric SF6 mole fraction, the sensitivity coefficients associated with the SF6 impurities in the raw N2, O2, and Ar gases at the final step were 7.8 × 1011, 2.1 × 1011, and 9.4 × 109, respectively (Table S3). However, owing to the improved LOD, the contributions to the gravimetric uncertainty were only 0.6, 0.4, and 0%, respectively. This assessment indicates the critical role of trace gas analysis in the preparation of gas mixtures at pmol/mol level. 2.2. Gravimetric Preparation of the Primary Standard Gas Mixture. Standard gas mixtures of SF6 in synthetic air at 12069
DOI: 10.1021/acs.analchem.7b02545 Anal. Chem. 2017, 89, 12068−12075
Article
Analytical Chemistry ambient levels, spanning a variation of common SF6 levels in the Northern and Southern Hemispheres, were prepared by gravimetry according to ISO 6142-1:2015. The inner surface of a 6.4 L aluminum cylinder (Luxfer, Australia) was electropolished, and it was cleaned under evacuation to about 10−6 Torr at 80 °C. In this way, SF6 contaminants and other potential interference materials (including H2O) that were adsorbed or dissolved into the manufacturing oil between threads on the head valve were removed. Evacuation was performed by a turbomolecular pump backed by a diaphragm pump. Purity-assessed raw SF6 (99.9892%, Matheson Trigas) and N2 (99.9999%, Deokyang Energen Co.) gases were diluted to 440 pmol/mol in six steps: 3.8 cmol/mol → 950 μmol/mol → 24 μmol/mol → 590 nmol/mol → 16 nmol/mol → 440 pmol/mol. The mixing ratios of the mother gases and balance gas were controlled by patented automatic weighing system. A comparator balance (Mettler-Toledo XP26003L) with a 15 kg maximum capacity and a 1 mg minimum readability was used for weighing the target gases. A position-controlled turntable and customized weighing pan were attached to the comparator balance for tight control of the loading position. Cylinders were automatically and sequentially loaded onto the exact same spot on the weighing pan, as monitored by position sensors; this careful automation led to an extremely low weighing uncertainty, ∼3 mg (2σ) of tens of kilogram weighing. Further details regarding the patented automatic weighing system and gravimetry procedure associated with uncertainty assessment are given in the ST3 and Table S2. The homogeneity of the gas mixtures was facilitated by coaxial rolling of the cylinders in a horizontal orientation. The dilution hierarchy is presented in Figure S2. In the first six steps, 4 gas mixtures (SF6 in N2) were prepared at similar mole fractions. In the final step, Ar and O2 were added to mimic atmospheric composition. The mole fraction of the relevant component k, yk, can be estimated by eq 2: ⎞ P ⎛ ⎞ mj ⎟/∑ ⎜ ⎟ i=1 ⎜ ∑i = 1 x × M ⎟ ⎟ ∑ × x M ⎝ ⎠ ⎝ i,j i i,j i⎠ j=1 j=1 n n P
yk =
⎛
∑ ⎜⎜
Figure 2. Uncertainty budget of the SF6 mole fraction in the gravimetrically prepared standard gas mixture (D155875). See Tables S2 and S3 for further details.
valid cylinders, of which measured mole fraction felt within the acceptance criteria, the dilution process is verified, but the outliers opt out of the suite for the next dilution step or use as a standard. Otherwise, the dilution step is repeated until the verification is accepted. For the KRISS standards, no outlier was found for any step (Figure S3). For the first four steps, a GC-TCD was used to perform the verification tests, as the GC-TCD response was sufficiently high and linear along the prepared concentrations down to 440 μmol/mol. In the fifth step, FTIR was used for the verification test. An integrated absorbance indicating a rovibrational band centered at 950 cm−1 was used for comparison. In the sixth and seventh steps, GC-μECD was used. Details regarding analytical conditions are shown with respect to the preparation step (Figure S3). For instance, detection responses of D015302, D015206, D015296, and D015319 at the sixth step were linear over the attempted mole fractions of SF6 ranging from 439.8 to 441.7 pmol/mol. In Figure S3, the μECD detection sensitivity is compared to show that the sixth preparation step met the acceptance criteria denoted in eq 3:
xk , j × mj
(2)
|ySF ,prep − ySF ,ver | ≤ 2 u 2(ySF ,prep ) + u 2(ySF ,ver ) 6
where xi,j is the amount of substance fraction of component i in gas j, Mi is the molar mass of component i, and mj is the mass added of gas j. The mole fraction of SF6 and the associated uncertainty in each dilution step were computed using the GUM Workbench program (version 2.3.6.141) (Tables S4, S5, and S6). The standard uncertainty of the gravimetric fraction of SF6 was evaluated according to JCGM 100:2008.20 In particular, the uncertainty of xSF6 in the impurity assessments of N2 raw gas contributes to the total uncertainty by as much as 5.4%. The uncertainties associated with weighing and handling comprises the remainder of the total uncertainty (Figure 2 and Table S3).
6
6
6
(3)
where ySF6, prep is the computed gravimetric fraction of SF6, and u(ySF6, prep) is the associated uncertainty, which is described in Table S4, ySF6,ver is the SF6 fraction determined by the averaged μECD sensitivity of the four mixtures (D015302, D015206, D015296, and D015319), and u(ySF6,ver) is the uncertainty associated with the analytical precision during the verification tests. Since the analytical precision for the attempted verification tests was shown to be
