Characteristic Evaluation of Gas Chromatography with Different Detectors for Accurate Determination of Sulfur Hexafluoride

Sulfur hexafluoride (SF6), which survives in the atmosphere for an extremely long period of time, is the most potent greenhouse gas regulated under the Kyoto Protocol. So, the accurate monitoring of atmospheric SF6 plays an important role in the study of the control policies for reducing greenhouse gas emissions. The instruments for SF6 measurement are typically calibrated using certified reference materials. The concentrations of the commercially available SF6 reference materials usually have a broad range, from 1 μmol/mol to 6000 μmol/mol. Some characteristics including sensitivity, linear range, relative standard deviation, and accuracy are crucial for the determination of SF6 in such a broad concentration range. Therefore, the selection of a proper detector for the accurate determination of SF6 with such a broad range is extremely important to establish a gas chromatography (GC) method for developing SF6 reference materials. In this paper, several typical GC methods with different detectors, including a thermal conductivity detector (TCD), a pulsed discharge helium ionization detector (PDHID), and a flame photometric detector (FPD), were carefully established for the accurate determination of SF6 with different concentrations. The results show that an FPD detector has a relatively narrow linearity range, thus a quadratic equation should be established for building a calibration curve. The PDHID and TCD have good linearity with coefficients of 1.0000 in the concentration range of 10–100 μmol/mol (using a PDHID), and 100–1000 μmol/mol (using a TCD), respectively. Further considering the measurement errors of indication results, the PDHID is suitable for SF6 measurement when the concentrations are below 100 μmol/mol, whereas the TCD is suitable for SF6 measurement when the concentrations are over 100 μmol/mol. These results provide useful guidance in choosing an appropriate GC detector for the accurate determination of SF6, which are especially very helpful for developing SF6 reference materials.

Of all the greenhouse gases regulated under the Kyoto Protocol [7], SF 6 is the most potent greenhouse gas [8] and persists in the atmosphere for an extremely long time, roughly 800-3200 years [9][10][11].
The earliest measurements of SF 6 reported a mole fraction of <1 pmol/mol (or ppt, parts per trillion) [12][13][14].The mass spectrometer (MS) and electron capture detector (ECD) are most often employed as gas chromatographic detectors for measuring trace SF 6 in the atmosphere.Most ambient SF 6 measurements using ECD typically have a precision of 2%.Pre-concentration may provide more precise results, although it is less frequent than loop-injection air samples and comes with more complexity [11].SF 6 emissions from electrical equipment occur during production, routine maintenance, refill, leakage, and disposal [15,16].Random failure or deliberate or accidental venting of equipment may also cause high levels of emissions.For instance, in a single incident in 2013, a broken seal resulted in the emission of 113 kg of SF 6 [17].Such gas leakage incidents could be prevented through using an alarm or other gas leakage detection system.
All the instruments for SF 6 measurements are typically calibrated against certified reference materials (CRMs) [18][19][20][21][22][23].The concentrations of commercially available SF 6 reference materials usually have a broad range from 1 µmol/mol to 6000 µmol/mol.Choosing a proper detector for the accurate determination of SF 6 is extremely important to establish a GC method for developing SF 6 reference materials.In this study, typical GC methods with different detectors, including TCD, PDHID, and FPD were carefully researched and established for the accurate determination of SF 6 .The detectors' performance, including precision and accuracy, is comprehensively discussed.

Determination of SF 6 Using GC-FPD
A typical chromatogram of SF 6 measured using an FPD is shown in Figure 1.The main peak at 1.317 min corresponds to SF 6 .For repeatability, the RSD value of the peak area obtained from six standard injections of 10.0 µmol/mol was calculated to be 0.62%.parts per trillion) [12][13][14].The mass spectrometer (MS) and electron capture de (ECD) are most often employed as gas chromatographic detectors for measuring tra in the atmosphere.Most ambient SF6 measurements using ECD typically have a pre of 2%.Pre-concentration may provide more precise results, although it is less fre than loop-injection air samples and comes with more complexity [11].
SF6 emissions from electrical equipment occur during production, routine m nance, refill, leakage, and disposal [15,16].Random failure or deliberate or accidenta ing of equipment may also cause high levels of emissions.For instance, in a single in in 2013, a broken seal resulted in the emission of 113 kg of SF6 [17].Such gas leakag dents could be prevented through using an alarm or other gas leakage detection sy All the instruments for SF6 measurements are typically calibrated against ce reference materials (CRMs) [18][19][20][21][22][23].The concentrations of commercially available S erence materials usually have a broad range from 1 µmol/mol to 6000 µmol/mol.Cho a proper detector for the accurate determination of SF6 is extremely important to est a GC method for developing SF6 reference materials.In this study, typical GC me with different detectors, including TCD, PDHID, and FPD were carefully researche established for the accurate determination of SF6.The detectors' performance, incl precision and accuracy, is comprehensively discussed.

Determination of SF6 using GC-FPD
A typical chromatogram of SF6 measured using an FPD is shown in Figure main peak at 1.317 min corresponds to SF6.For repeatability, the RSD value of the area obtained from six standard injections of 10.0 µmol/mol was calculated to be 0.6 The broad range from 10 µmol/mol to 6000 µmol/mol commonly exists for SF6 ence materials.Establishing a single equation describing the response of the FPD a SF6 concentrations in such a wide range is impossible.GC-FPD was unsuitable f quantitative analysis of SF6 concentration if both the nonlinear response and the qu ing effect were not considered [24], particularly for sulfur concentrations in a rela wide linear range.If the calibration curves have a relatively narrow range, a qua equation, y = a + bx + cx 2 , can be used to establish a calibration equation.The peak are SF6 concentrations were denoted as y and x in the calibration curves.Six narrow co tration ranges were divided among the wide concentration range.Thus, six calib curves were constructed via plotting the peak areas against the concentrations, as s The broad range from 10 µmol/mol to 6000 µmol/mol commonly exists for SF 6 reference materials.Establishing a single equation describing the response of the FPD against SF 6 concentrations in such a wide range is impossible.GC-FPD was unsuitable for the quantitative analysis of SF 6 concentration if both the nonlinear response and the quenching effect were not considered [24], particularly for sulfur concentrations in a relatively wide linear range.If the calibration curves have a relatively narrow range, a quadratic equation, y = a + bx + cx 2 , can be used to establish a calibration equation.The peak area and SF 6 concentrations were denoted as y and x in the calibration curves.Six narrow concentration ranges were divided among the wide concentration range.Thus, six calibration curves were constructed via plotting the peak areas against the concentrations, as shown in Figure 2. The coefficient of determination (R 2 ), used to express linearity, was calculated to be ≥0.9995 for all equations.The accuracy, reflecting the performance characteristics, of the calibration curve was also evaluated.The relative measurement error was used to evaluate the accuracy of the calibration curve.A sulfur hexafluoride in nitrogen CRM (GBW08124), with a certified value of 10.0 µmol/mol was chosen as an "unknown" and analyzed using GC-FPD under the same analytical conditions.The measurement value of this SF6 CRM was calculated using the established calibration equation: y = 0.2884x 2 + 20.432x − 88.378.As a result, the measurement value was calculated to be 9.983 µmol/mol.The uncertainty of the measurement was evaluated to be 1.8% (k = 2).Equation ( 1) was used to evaluate the relative measurement error, where Δ is the relative measurement error; x is the measurement value based on the established equation; and s x is the certified value.Finally, the relative measurement error was calculated to be −0.17%,which is far smaller than the expanded uncertainty of the used CRM, which certified 1% (k = 2), and shows very good accuracy of GC-FPD through building up the equation for the narrow concentration range.The accuracy, reflecting the performance characteristics, of the calibration curve was also evaluated.The relative measurement error was used to evaluate the accuracy of the calibration curve.A sulfur hexafluoride in nitrogen CRM (GBW08124), with a certified value of 10.0 µmol/mol was chosen as an "unknown" and analyzed using GC-FPD under the same analytical conditions.The measurement value of this SF 6 CRM was calculated using the established calibration equation: y = 0.2884x 2 + 20.432x − 88.378.As a result, the measurement value was calculated to be 9.983 µmol/mol.The uncertainty of the measurement was evaluated to be 1.8% (k = 2).Equation ( 1) was used to evaluate the relative measurement error, where ∆ is the relative measurement error; x is the measurement value based on the established equation; and x s is the certified value.Finally, the relative measurement error was calculated to be −0.17%,which is far smaller than the expanded uncertainty of the used CRM, which certified 1% (k = 2), and shows very good accuracy of GC-FPD through building up the equation for the narrow concentration range.

Determination of SF 6 Using a PDHID
A typical chromatogram of an SF 6 CRM using PDHID is shown in Figure 3.The peak at 5.888 min corresponds to SF 6 .For repeatability, the RSD value of the peak area obtained from six injections of 10 µmol/mol was calculated to be 0.18%.

Determination of SF6 Using a PDHID
A typical chromatogram of an SF6 CRM using PDHID is shown in Figure 3.The at 5.888 min corresponds to SF6.For repeatability, the RSD value of the peak area obta from six injections of 10 µmol/mol was calculated to be 0.18%.The PDHID has a good linearity for the range of 10-100 µmol/mol, showing a d mination coefficient (R 2 ) of 1.0000 (Figure 4).The same SF6 CRM (GBW08124) with a tified value of 10.0 µmol/mol was chosen as an "unknown" and was analyzed using PDHID.The measurement value of this SF6 CRM was calculated according to the e lished calibration equation: y = 491.6591x + 268.6090.As a result, the measurement v was calculated to be 10.019 µmol/mol.The uncertainty of the measurement was evalu to be 1.2% (k = 2).The relative measurement error was calculated to be 0.19%.The extent of the calibration curve of SF6 studies on 100-1000 µmol/mol is show Figure 5.It was clear that the calibration was linear, as a result of a coefficient of dete nation (R 2 ) of 0.9997.The linearity is not quite as good as that from the linear calibra range of 10-100 µmol/mol.To evaluate the accuracy for this calibration curve, anothe CRM (GBW08124) with a certified value of 100 µmol/mol was chosen as an "unkno and was analyzed.The measurement value was calculated through using the establi equation: y = 415.8977x + 9408.4270.The measurement value was calculated to be 9 µmol/mol.The uncertainty of the measurement was evaluated to be 1.3% (k = 2).In a case, the relative measurement error between the measurement value and the cert value was calculated to be −0.69%.The relative measurement error slightly increases the increased concentration, indicating that the PDHID is not suitable for the analys SF6 with a concentration >100 µmol/mol, when high accuracy is required.The PDHID has a good linearity for the range of 10-100 µmol/mol, showing a determination coefficient (R 2 ) of 1.0000 (Figure 4).The same SF 6 CRM (GBW08124) with a certified value of 10.0 µmol/mol was chosen as an "unknown" and was analyzed using GC-PDHID.The measurement value of this SF 6 CRM was calculated according to the established calibration equation: y = 491.6591x + 268.6090.As a result, the measurement value was calculated to be 10.019 µmol/mol.The uncertainty of the measurement was evaluated to be 1.2% (k = 2).The relative measurement error was calculated to be 0.19%.The PDHID has a good linearity for the range of 10-100 µmol/mol, show mination coefficient (R 2 ) of 1.0000 (Figure 4).The same SF6 CRM (GBW08124) tified value of 10.0 µmol/mol was chosen as an "unknown" and was analyzed PDHID.The measurement value of this SF6 CRM was calculated according t lished calibration equation: y = 491.6591x + 268.6090.As a result, the measure was calculated to be 10.019 µmol/mol.The uncertainty of the measurement wa to be 1.2% (k = 2).The relative measurement error was calculated to be 0.19%.The extent of the calibration curve of SF6 studies on 100-1000 µmol/mol Figure 5.It was clear that the calibration was linear, as a result of a coefficient nation (R 2 ) of 0.9997.The linearity is not quite as good as that from the linear range of 10-100 µmol/mol.To evaluate the accuracy for this calibration curve, CRM (GBW08124) with a certified value of 100 µmol/mol was chosen as an and was analyzed.The measurement value was calculated through using the equation: y = 415.8977x + 9408.4270.The measurement value was calculated µmol/mol.The uncertainty of the measurement was evaluated to be 1.3% (k = a case, the relative measurement error between the measurement value and value was calculated to be −0.69%.The relative measurement error slightly inc the increased concentration, indicating that the PDHID is not suitable for the SF6 with a concentration >100 µmol/mol, when high accuracy is required.The extent of the calibration curve of SF 6 studies on 100-1000 µmol/mol is shown in Figure 5.It was clear that the calibration was linear, as a result of a coefficient of determination (R 2 ) of 0.9997.The linearity is not quite as good as that from the linear calibration range of 10-100 µmol/mol.To evaluate the accuracy for this calibration curve, another SF 6 CRM (GBW08124) with a certified value of 100 µmol/mol was chosen as an "unknown" and was analyzed.The measurement value was calculated through using the established equation: y = 415.8977x + 9408.4270.The measurement value was calculated to be 99.31 µmol/mol.The uncertainty of the measurement was evaluated to be 1.3% (k = 2).In such a case, the relative measurement error between the measurement value and the certified value was calculated to be −0.69%.The relative measurement error slightly increases with the increased concentration, indicating that the PDHID is not suitable for the analysis of SF 6 with a concentration >100 µmol/mol, when high accuracy is required.

Determination of SF6 using a TCD
A typical chromatogram of an SF6 CRM using a TCD is shown in Figure 6 peak at 2.458 min corresponds to SF6.For repeatability, the RSD value of the obtained from six injections of 100 µmol/mol was calculated to be 0.60%.A good linearity was obtained in the range of 100-1000 µmol/mol.The l efficient of determination (R 2 ) was calculated to be 1.0000, as shown in Figur cated that the TCD allows for the higher detection of ranges from 100 µmol/m µmol/mol.To evaluate the accuracy of the calibration curve, an SF6 CRM (G with a certified value of 100 µmol/mol was chosen as an "unknown" and wa The measurement value was calculated through using the established equation x − 0.7074.The measurement value was calculated to be 100.43µmol/mol.Th ment uncertainty was 1.7% (k = 2).In such a case, the relative measurement err the measurement value and the certified value was calculated to be 0.43%.

Determination of SF 6 Using a TCD
A typical chromatogram of an SF 6 CRM using a TCD is shown in Figure 6.The main peak at 2.458 min corresponds to SF 6 .For repeatability, the RSD value of the peak area obtained from six injections of 100 µmol/mol was calculated to be 0.60%.

Determination of SF6 using a TCD
A typical chromatogram of an SF6 CRM using a TCD is shown in Figure 6.The peak at 2.458 min corresponds to SF6.For repeatability, the RSD value of the peak obtained from six injections of 100 µmol/mol was calculated to be 0.60%.A good linearity was obtained in the range of 100-1000 µmol/mol.The lineari efficient of determination (R 2 ) was calculated to be 1.0000, as shown in Figure 7.It cated that the TCD allows for the higher detection of ranges from 100 µmol/mol to µmol/mol.To evaluate the accuracy of the calibration curve, an SF6 CRM (GBW0 with a certified value of 100 µmol/mol was chosen as an "unknown" and was anal The measurement value was calculated through using the established equation: y = 0 x − 0.7074.The measurement value was calculated to be 100.43µmol/mol.The mea ment uncertainty was 1.7% (k = 2).In such a case, the relative measurement error bet the measurement value and the certified value was calculated to be 0.43%.A good linearity was obtained in the range of 100-1000 µmol/mol.The linearity coefficient of determination (R 2 ) was calculated to be 1.0000, as shown in Figure It indicated that the TCD allows for the higher detection of ranges from 100 µmol/mol to 1000 µmol/mol.To evaluate the accuracy of the calibration curve, an SF 6 CRM (GBW08124) with a certified value of 100 µmol/mol was chosen as an "unknown" and was analyzed.The measurement value was calculated through using the established equation: y = 0.3167 x − 0.7074.The measurement value was calculated to be 100.43µmol/mol.The measurement uncertainty was 1.7% (k = 2).In such a case, the relative measurement error between the measurement value and the certified value was calculated to be 0.43%.

Discussion
The detailed characteristics of the relationships between sulfur hexafluoride concentrations and the response from different detectors of gas chromatography are shown in Table 1.An adequate calibration equation is necessary to describe the calibration curves for the SF6 analysis.Linear and second polynomial equations were used to verify the adequacy of the equations.The coefficient of determination (R 2 ) is the sole criterion, and all the numerical values are higher than 0.9995.The relative measurement error was used to evaluate the accuracy of the calibration equation.The results show that the relative measurement error can be decreased when an adequate calibration equation is adopted.

Instruments
The TCD experiments were performed on a gas chromatograph 7890B (Agilent, Santa Clara, CA, USA).A column HayeSep Q (0.91 m, ID 2 mm, mesh size 80/100) was used for GC separation.The sampling volume was set at 1 mL.The sample was injected using a split mode with a split ratio of 10:1.The temperature was set as follows: column oven, 100 °C; inlet, 250 °C; and TCD, 180 °C.Effects of the column oven and TCD temperature on

Discussion
The detailed characteristics of the relationships between sulfur hexafluoride concentrations and the response from different detectors of gas chromatography are shown in Table 1.An adequate calibration equation is necessary to describe the calibration curves for the SF 6 analysis.Linear and second polynomial equations were used to verify the adequacy of the equations.The coefficient of determination (R 2 ) is the sole criterion, and all the numerical values are higher than 0.9995.The relative measurement error was used to evaluate the accuracy of the calibration equation.The results show that the relative measurement error can be decreased when an adequate calibration equation is adopted.

Instruments
The TCD experiments were performed on a gas chromatograph 7890B (Agilent, Santa Clara, CA, USA).A column HayeSep Q (0.91 m, ID 2 mm, mesh size 80/100) was used for GC separation.The sampling volume was set at 1 mL.The sample was injected using a split mode with a split ratio of 10:1.The temperature was set as follows: column oven, 100 • C; inlet, 250 • C; and TCD, 180 • C. Effects of the column oven and TCD temperature on SF 6 (100 µmol/mol) response are shown in Figures S1 and S2.Helium gas was selected as a carrier gas, and the flow rate was set at 45 mL/min.GC-PDHID experiments were conducted with a 7890B (Agilent, Santa Clara, CA, USA), which has been custom configured by Wasson-ECE, and includes the columns.This system was designed for the analysis of trace impurities in gas.It is also suitable for the analysis of SF 6 .The temperature was set as follows: column oven, 50 • C, firstly maintained for 8 min, then increased to 100 • C with a gradient of 20 • C/min, maintained for 5.5 min; inlet, 60 • C; and PDHID, 250 • C. Helium gas was chosen as a carrier gas, and the flow rate was set at 30 mL/min.FPD experiments were performed on gas chromatograph TP-2090E (Tianpu Instrument Co., Guangzhou, China).A column GDX-502 (3 mm ID × 4 m, inert and porous solid particles which are suitable for SF 6 analysis) was used for GC separation.The temperature was set as follows: column oven, 60 • C; inlet, 60 • C; and FPD, 140 • C. The gas flow rate for the FPD was set as follows: nitrogen, 30 mL/min; hydrogen, 140 mL/min; air, 80 mL/min and 168 mL/min.

Methods
Repeatability was used to assess measurement precision.The relative standard deviation (RSD) affects the uncertainty of the measurement results.The results were expressed in terms of RSD for the peak areas of SF 6 .Six consecutive injections of SF 6 CRMs were used to calculate the repeatability, and then to evaluate the precision of the method.
Calibration curves were obtained using the SF 6 concentration against the peak area from GC with different detectors.Lower concentrations of SF 6 in the calibration curve were diluted directly through a high concentration SF 6 CRM, using mass flow controllers (Alicat Scientific, Tucson, AZ, USA).The mass flow controllers were calibrated by the Institute of Metrology, according to the National Verification Regulation.The expanded measurement uncertainty for mass flow controllers' calibration was in the range of 0.51~0.54%(k = 2).An SF 6 CRM with a certified value was chosen as an "unknown" to verify the accuracy of the calibration curve made using the diluting method with mass flow controllers.
The measurement uncertainty of SF 6 was evaluated according to the Guide to the Expression of Uncertainty in Measurement (GUM).On the basis of GUM, the standard measurement uncertainty of SF 6 was combined with the uncertainty of the SF 6 CRM, the mass flow controllers' calibration, and repeatability.The expanded uncertainty was obtained via multiplying the standard uncertainty by a coverage factor (k = 2).

Conclusions
In this paper, GC detectors, including an FPD, a PDHID, and a TCD, analyzing SF 6 were carefully evaluated.For the FPD, the calibration curves have a relatively narrow range.Second polynomial equations were used, fitting a calibration equation.Both the PDHID and TCD have a good linearity, and coefficients of determination (R 2 ) are equal to 1.0000 in the concentration range of 10-100 µmol/mol, and 100-1000 µmol/mol, respectively.The relative measurement error indicated that the PDHID is more suitable for analyzing SF 6 with concentrations lower than 100 µmol/mol, whereas the TCD is more suitable for analyzing SF 6 with concentrations larger than 100 µmol/mol.In summary, an appropriate gas chromatographic detector for the accurate determination of SF 6 , especially for developing SF 6 reference materials, was crucial to establish a GC method.The results from this study provide guidance for choosing an appropriate detector for the accurate determination of SF 6 , which would subsequently contribute to the development of SF 6 reference materials used in greenhouse gases' control.

Table 1 .
Calibration equations for the sulfur hexafluoride concentration and the peak area for GC with different detectors.

Table 1 .
Calibration equations for the sulfur hexafluoride concentration and the peak area for GC with different detectors.