Atmospheric histories and growth trends of C4F10, C5F12, C6F14, C7F16

. Atmospheric observations and trends are presented for the high molecular weight perﬂuorocarbons (PFCs): decaﬂuorobutane (C 4 F 10 ), dodecaﬂuoropentane (C 5 F 12 ), tetradecaﬂuorohexane (C 6 F 14 ), hexadecaﬂuoroheptane (C 7 F 16 ) and octadecaﬂuorooctane (C 8 F 18 ). Their atmospheric histories are based on measurements of 36 Northern Hemisphere and 46 Southern Hemisphere


Introduction
Perfluorocarbons (PFCs) are powerful greenhouse gases regulated under the Kyoto Protocol to the United Nations Framework Convention on Climate Change (UNFCCC). Due to their long lifetimes and strong absorption in the infrared, PFCs are considered to have a permanent effect on the Earth's radiative budget and have global warming potentials (GWPs) 3 to 4 orders of magnitude higher than that of carbon dioxide (CO 2 ), see Table 1, (Forster et al., 2007). Observations and atmospheric-observation-based emission estimates are available for carbon tetrafluoride (CF 4 ), hexafluoroethane (C 2 F 6 ), octafluoropropane (C 3 F 8 ) and octafluorocyclobutane (c-C 4 F 8 ) Oram et al., 2012;Saito et al., 2010). CF 4 is the most abundant PFC and has a significant natural abundance (Deeds et al., 2008;Harnisch et al., 1996a,b;Mühle et al., 2010). The predominant anthropogenic emissions of these lower molecular weight PFCs are from the production of aluminum, usage in the semiconductor industry and as refrigerants Oram et al., 2012). Both the aluminum and semiconductor industries have made efforts to reduce emissions of the lower molecular weight PFCs to the atmosphere, although global bottom-up inventories show discrepancies with atmospheric-observation-based ("top-down") emission estimates International Aluminium Institute, 2011;Semiconductor Industry Association, 2001;World Semiconductor Council, 2005). Currently, there is much less information available on the higher molecular weight PFCs: decafluorobutane (C 4 F 10 ), dodecafluoropentane (C 5 F 12 ), tetradecafluorohexane (C 6 F 14 ), hexadecafluoroheptane (C 7 F 16 ) and octadecafluorooctane (C 8 F 18 ). These PFCs have emission sources similar to other halocarbons, e.g. their usage as refrigerants, solvents, fire suppressants and foam blowing agents; they were initially suggested as replacements for ozone depleting substances (ODS) that are regulated under the Montreal Protocol (UNEP Technology and Economic Assessment Panel, 1999;Tsai, 2009). Moreover, C 5 F 12 -C 8 F 18 (liquids at room temperature), have a first-of-a-kind emission source for fluorinated compounds from their use as heat transfer fluids in the semiconductor industry (where previously deionized water and a mixture of glycol and deionized water were used) (Tsai, 2009;Tuma and Tousignant, 2001). The main sink for the PFCs is photolysis by Lyman-α radiation and a minor destruction pathway is reaction with O( 1 D) (Ravishankara et al., 1993). In this study, atmospheric observations of C 4 F 10 , C 5 F 12 , C 6 F 14 , C 7 F 16 and C 8 F 18 are presented based on measurements of Northern Hemisphere (NH) and Southern Hemisphere (SH) archived air samples. These samples cover a 39-year period, from 1973 to 2011, and include 36 NH and 46 SH separate samples. Additionally, long-term growth trends are presented based on the atmospheric histories for the high molecular weight PFCs.

Instrumentation
The cryogenic preconcentration gas chromatography-mass spectrometry (GC-MS: Agilent 6890-5973/5975) "Medusa" systems (Arnold et al., 2012;Miller et al., 2008) were used to measure the heavy PFC mole fractions in archived air samples at the Scripps Institution of Oceanography (SIO), University of California, San Diego (San Diego, CA) and at the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Division of Marine and Atmospheric Research (CMAR, Aspendale, Australia). These instruments are part of the Advanced Global Atmospheric Gases Experiment (AGAGE) network. For each measurement, the condensables in a 2-l air sample are preconcentrated onto a micro-trap and then cyrofocused onto a second micro-trap (both micro-traps are initially held at −160 • C and subsequently heated for desorption) before injection onto a capillary column. Currently, the Medusa systems in AGAGE use a CP-PoraBOND Q fused silica PLOT column (25 m, 0.32 mm ID, 5 µm, Agilent Technologies) as the main column for separation of all analytes (except CF 4 and NF 3 ) (see Arnold et al. (2012);Miller et al. (2008) for details). However for the measurements at CSIRO, the Medusa was fitted with a GS-GasPro column (60 m, 0.32 mm ID, Agilent Technologies) as the main column, which had improved separation for these analytes. To maximize the measurements' precisions, the quadrupole MS was operated in selective ion mode (SIM). Additionally, only a select number of species were measured in this experiment, as compared to the more than 50 species typically measured as part of the AGAGE network, to minimize the number of acquired ions and further improve sample precisions. Each sample measurement was bracketed by a reference gas analysis, allowing for correction of short-term instrumental drift (Prinn et al., 2000). A small Atmos. Chem. Phys., 12, 4313-4325, 2012 www.atmos-chem-phys.net/12/4313/2012/ PoraBOND 5975 959 GasPro 5973 1194 * Additional compounds are shown for reference. Parameters for the oven are: ramping from 40 • C to 200 • C over 7 minutes for both the GasPro and PoraBOND with the ovens subsequently held at 200 • C. The pressure is ramping at the same time to maintain a constant flow from 6.1 to 15.1 psig for the PoraBOND column and from 9 to 18 psig for the GasPro and subsequently held constant. Standard precisions refer to the 1-σ standard deviation taken on the working standard used, which is a NH 2010 sample for the SIO Medusa and a 2010 sample filled at Cape Grim for the CSIRO Medusa. Species in bold are those discussed in this study. Non-bolded species are added for reference. blank was detected for C 6 F 14 , C 7 F 16 and C 8 F 18 (0.008 and 0.005 ppt for C 6 F 14 , 0.012 and 0.013 ppt for C 7 F 16 , 0.017 and 0.016 ppt for C 8 F 18 on the CSIRO and SIO instruments, respectively), most likely due to the Nafion dryers used in the Medusa, and the observations were corrected accordingly. The detection limits for each species on both instruments were estimated as three times the baseline height of the noise of the target ion immediately preceding and following the elution of the species and are presented in Table 2. The differences in the detection limits between the two instruments are due to the improved separation of analytes on the GS-GasPro column used on the CSIRO Medusa and the newer model of MS (5975) used at SIO. Two experiments were done to assess whether the instrument responses of the Medusa systems were linear over the required range of mixing ratios for each species. In general, the instrument response, R, is related to the sample mole fractions, χ, by R ∝ χ (1− ) , where is the nonlinearity parameter (Prinn et al., 2000). A whole-air sample from a 2010 Cape Grim Air Archive (CGAA) tank was decanted into a 35-l stainless steel tank (Essex Cryogenics), and spiked with a small volume of high-purity methane (CH 4 ) to increase the CH 4 mole fractions from ambient to ca. 8 parts per million. From this spiked air sample, six subsamples were prepared in 6-l stainless steel SilcoCan flasks (Restek Inc.) using a vacuum manifold. Each subsample was then diluted by adding "zero-air". The zero-air was measured on the Medusa and found to be analyte free for the PFCs studied here. The amount of zero-air added was varied to give a range of dilution factors (nominally from 6.25 % to 75 %). The actual dilution factors for each subsample were determined by precisely measuring the CH 4 mole fraction (including that of the spiked parent sample) on a gas chromatography-flame ionization detector (GC-FID) system with a known linear response (Francey et al., 2003). These subsamples were subsequently measured on the Medusa systems to determine the linearity of the Medusa measurements for each species. All but the lowest concentration subsample of 6.25% were above the detection limit for C 4 F 10 , C 5 F 12 , C 7 F 16 and C 8 F 18 . The lowest concentration subsample was slightly above the detection limit for C 6 F 14 .
The second experiment to characterize the instrument response involved sampling different volumes from a single air sample . The range of relative volumes sampled was from 6 % to 200 % of the standard 2-l sample. The volume method has the advantage of characterizing instrument responses at mole fractions above present day background levels, which could not be easily achieved through the dilution subsamples unless more concentrated samples were prepared. However, as the atmospheric samples measured in the air archives were all below current atmospheric background mole fractions, the volume method served only to complement the dilution method's instrument response experiment. Generally, the systems exhibited a linear response over much of the required range of mole fractions, with departures from linearity at mole fractions corresponding to those of the oldest archive samples, which have the lowest mole fractions. Based on these dilution experiment measurements, a nonlinearity parameter, , was estimated for each PFC on each instrument and was used to correct the observations. These nonlinearity parameters were relatively small and ranged from 0 to 0.047, with the largest nonlinearity correction required for C 8 F 18 .

Calibration
Four primary standards were prepared at SIO to identify and quantify the heavy PFCs on the Medusa. The primary standards were prepared following the bootstrap method by stepwise dilution, with dichlorodifluoromethane (CCl 2 F 2 , CFC-12) used as the bootstrap gas (Prinn et al., 2000). The highpurity compounds were purchased from Synquest Laboratories with purities of: C 4 F 10 (98 % min.), C 5 F 12 (99 % min.), C 6 F 14 (98.5 % min.), C 7 F 16 (98 % min.) and C 8 F 18 (99 % min.); the nitrous oxide (N 2 O) was purchased from Scott Specialty Gases and had a purity of 99.9997 %, and CFC-12 had a purity of 99.99 %. Each high-purity compound was vacuum distilled for further purification by repeated cycles of freezing with liquid nitrogen, vacuum removal of noncondensable contaminants and then thawing.
First a gravimetric PFC/CFC-12/N 2 O mixture was prepared with a molar ratio of PFC to CFC-12 of 1.85-3.7 × 10 −3 , with CFC-12 being used as the bootstrap gas and N 2 O as the balance gas. Typically N 2 O is used as the bootstrap gas; however these standards were prepared as standard additions and the expected final N 2 O mole fractions would have been beyond the range of the currently available SIO calibration scale. Therefore, CFC-12 was used as the bootstrap gas, as the resulting final mole fractions of CFC-12 can be accurately measured on the Medusa.
A primary standard was prepared by spiking a real air sample filled at La Jolla, CA (32.87 • N, 117.25 • W) with the PFC/CFC-12/N 2 O mixture. The real air sample was measured on the SIO Medusa to measure its initial CFC-12 mole fraction and instrument response for each PFC. The additional mole fractions added to the real air sample from the PFC/CFC-12/N 2 O spike were ca. 2 parts per trillion (ppt) of C 6 F 14 , 1 ppt of the other PFCs and 540 ppt of CFC-12. The enhancement factor of the PFC/CFC-12/N 2 O mixture added to the real air sample was determined by measuring the final CFC-12 mole fractions on the Medusa. The final atmospheric mole fractions in the primary standard were estimated as described by Ellison and Thompson (2008).
The primary standards were measured on the Medusa to determine the retention times and mass spectra of the heavy PFCs as well as to quantify the atmospheric observations. The mass spectra of the high molecular weight PFCs agree with published spectra from the National Institute of Standards and Technology and the retention times are consistent with what is expected based on their boiling points (NIST, 2011). Table 2 shows the target and qualifier mass-to-charge ratios, as well as the retention times, used on the SIO and CSIRO Medusa systems; the target mass-to-charge ratio is used for identification and quantitation, while the qualifiers are only used for ensuring the proper identification of the species. Four primary standards were prepared and the calibration scale, referred to as SIO-2012, has estimated accuracies of 6.8 % for C 4 F 10 , 7.8 % for C 5 F 12 , 4.0 % for C 6 F 14 , 6.6 % for C 7 F 16 and 7.9 % for C 8 F 18 .

Archived air samples
The atmospheric histories of these heavy PFCs are based on measurements made at SIO and CSIRO of NH and SH archived air samples, which cover a 39-year period. In total, 36 NH separate samples (33 measured at SIO and 3 measured at CSIRO) with fill dates from 1973 to 2011 were measured. These tanks were filled either at Cape Meares, Oregon (45.50   materials, tank sizes and fill purposes of these samples varied. Two tanks were identified as outliers with atmospheric mole fractions significantly higher than present day values. One 1978 tank was rejected for C 4 F 10 and C 5 F 12 , but a second tank with an identical fill date was also measured. Additionally a 1974 tank was rejected for C 8 F 18 , as samples with similar fill dates were below the detection limit. From 2003 onward, a collection of tanks containing NH air have been maintained at R.F. Weiss's laboratory at SIO as an air archive. These samples were pumped into 35-l internally electropolished stainless steel cylinders (Essex Cryogenics) at Trinidad Head, California using a modified oil-free compressor (Rix Industries). Recent filled archive tanks agree with in situ measurements made by the Medusa system at the time of the tank filling for the high molecular weight PFCs studied here, confirming that the modified Rix compressor does not compromise the integrity of these samples for the high molecular weight PFCs. Forty-six separate SH samples filled between 1978 and 2010 were also measured (6 at SIO and 40 at CSIRO). All but three of these samples, which were from the M.A. Wahlen laboratory at SIO, are part of the CGAA collection (Krummel et al., 2007;Langenfelds et al., 1996). The CGAA samples analyzed as part of this study consist of whole-air samples cryogenically filled in 35-l internally electropolished stainless steel cylinders at the Cape Grim Baseline Air Pollution Station in Tasmania, Australia (40.68 • S, 144.69 • E), except for three which were cryogenically collected in 48l aluminum cylinders; these latter three samples were rejected for C 8 F 18 , as their values were significantly higher than present-day background mole fractions. Five archive samples, all collected in 2001 and 2002, had significantly higher than current baseline mole fractions for C 7 F 16 and these values were consequently flagged as contaminated; these included the three samples from the M. A. Wahlen laboratory and two CGAA samples. These five samples were all stored in stainless steel cylinders, therefore most likely a local source at the Cape Grim Station influenced these samples. The Wahlen samples were also rejected for C 6 F 14 , as an unidentified analyte coeluted with C 6 F 14 . Seven subsamples of the CGAA, covering the time period of 1986 to 2008, were decanted into 4.5-l internally electropolished stainless steel cylinders (Essex Cryogenics) and subsequently measured on the SIO Medusa system; the parent samples were measured at CSIRO. This was to verify whether the two instruments and the calibration propagations produced measurements that agreed within known uncertainties. Generally, the measurements on the two systems agreed well within the measurement uncertainty. The most notable difference between the two instruments was for C 8 F 18 , with a maximum difference of 0.009 ppt, which was 12 % of the sample concentration.
sample measurements for both sets of archives are provided in the supplementary material.

Results and discussion
The atmospheric histories from 1973 to 2011 for the high molecular weight PFCs studied here are shown in Figs. 1-5. Due to the sparseness of the available data set, the presented annual mean mole fractions and growth rates are based on cubic smoothed spline fits to the observations. The observations were weighted by their measurement uncertainty and a 50 % attenuation period of 4 years was used, which is slightly larger than the mean data-spacing, in estimating the smoothing splines (Enting et al., 2006). The uncertainties associated with the spline fits were estimated using a Monte Carlo approach, where the fitting was repeated 1000 times with randomly varied observations that had distributions based on their measurement uncertainty. The uncertainty on each spline fit was taken as the 1-σ standard deviation of these runs. The growth rate was also calculated for each of the 1000 runs and the uncertainty associated with the annual growth rate was estimated as the 1-σ standard deviation of the calculated growth rates. The smoothed spline fits are shown along with the observations in Figs. 1-5. The bottom panels of Figs. 1-5 show the annual hemispheric growth rates estimated from the cubic smoothed spline fits for each PFC.
The annual mean mole fractions and growth rates from the spline fits are presented in Tables 3-7. C 4 F 10 and C 5 F 12 are present in the earliest archived samples at 0.015 ppt and 0.011 ppt respectively, but these 1980 0.000 ± 0.002 0.3 ± 0.5 0.011 ± 0.008 1.9 ± 1.6 0.005 ± 0.005 1.1 ± 1.0 1981 0.000 ± 0.002 0.6 ± 0.5 0.013 ± 0.007 1.9 ± 1.3 0.006 ± 0.004 1.2 ± 0.9 1982 0.001 ± 0.002 1.0 ± 0.5 0.015 ± 0.006 2.0 ± 1.1 0.008 ± 0.004 1.5 ± 0.8 1983 0.002 ± 0.002 1.7 ± 0.5 0.017 ± 0.005 2.1 ± 0.9 0.009 ± 0.004 1.9 ± 0.7 1984 0.004 ± 0.002 2.5 ± 0.4 0.019 ± 0.005 2.2 ± 0.8 0.011 ± 0.004 2.3 ± 0.6 1985 0.007 ± 0.003 3.3 ± 0.3 0.021 ± 0.004 2.3 ± 0.8 0.014 ± 0.003 2.8 ± 0.6 1986 0.011 ± 0.003 4.0 ± 0.4 0.023 ± 0.003 2.5 ± 0.7 0.017 ± 0.003 3.3 ± 0.5 1987 0.015 ± 0.003 4.6 ± 0.4 0.026 ± 0.003 2.7 ± 0.6 0.020 ± 0.003 3.7 ± 0.5 1988 0.020 ± 0.003 5.0 ± 0.5 0.029 ± 0.002 3.0 ± 0.6 0.024 ± 0.002 4.0 ± 0.5 1989 0.025 ± 0.003 5.4 ± 0.6 0.032 ± 0.002 3.2 ± 0.5 0.028 ± 0.002 4.3 ± 0.5 1990 0.030 ± 0.003 5.6 ± 0.7 0.035 ± 0.001 3.5 ± 0.4 0.033 ± 0.002 4.5 ± 0.6 1991 0.036 ± 0.003 5.6 ± 0.8 0.039 ± 0.001 3.7 ± 0.3 0.037 ± 0.002 4.6 ± 0.6 1992 0.042 ± 0.004 5.5 ± 1.0 0.043 ± 0.001 3.9 ± 0.3 0.042 ± 0.002 4.7 ± 0.6 1993 0.047 ± 0.004 5. measurements are considered below the estimated detection limits of the instruments, see Fig. 1 and Fig. 2. Analysis of firn air samples from Greenland confirm that there is no detectable natural abundance for these PFCs. C 4 F 10 and C 5 F 12 exhibit quasi-exponential growth in the 1980s and then grow nearly linearly to present day globally averaged (taken as the average between the NH and SH spline fit data) background atmospheric mole fractions of 0.17 ppt and 0.12 ppt, respectively. The slowdown in growth rates in the the 2000s suggest that emissions are decreasing, as is supported by the decline in the inter-hemispheric gradients seen in the observations, for C 4 F 10 and C 5 F 12 . Emissions of these high molecular weight PFCs based on EDGARv4.2 are of anthropogenic origin and primarily released in the NH (ER-JRC/PBL, 2009). Therefore as is expected, we see higher atmospheric mole fractions in the NH than in the SH due to the 1 to 2 year mixing time between the two hemispheres. This is confirmed by the lag in the SH growth rate as compared to that estimated in the NH. Higher variability in the NH samples can be seen in the early years for C 4 F 10 and C 5 F 12 , as compared to the SH samples. This is attributed to sampling of less well mixed air due to emissions originating primarily in the NH, although efforts were made to fill the archive tanks during baseline conditions. C 6 F 14 and C 7 F 16 are not detectable in the archived samples until 1984-1985and grow quasi-exponentially until 1999and 1992. C 8 F 18 follows a similar trend to that of C 6 F 14 , although it is not detectable until the mid 1990s, see Fig. 5, which is most likely due to lower emission rates. C 6 F 14 is the most abundant of the PFCs studied here at a globally averaged background tropospheric mole fraction of 0.27 ppt in 2011. The globally averaged atmospheric mole fraction in 2011 is 0.12 ppt for C 7 F 16 , and C 8 F 18 is the least abundant of all of these PFCs at 0.09 ppt. The growth rate of C 6 F 14 peaks in 1999 and has since declined to a 2011 annual global average of 5.0 ppq yr −1 . The trend in C 7 F 16 's growth rates differ from the other PFCs, in that its growth has been relatively constant at 3 to 4 ppq yr −1 for the last 15 years. The atmospheric trends and growth rates of C 8 F 18 are similar to those of C 6 F 14 , although with lower absolute values, with a 2011 mean growth rate of 0.9 ppq yr −1 . As seen with C 4 F 10 and C 5 F 12 , the NH archived air samples for C 6 F 14 , C 7 F 16 and C 8 F 18 have higher atmospheric mole fractions than the SH samples with similar fill dates, suggesting that the emissions are primarily in the NH. There is one NH sample with a fill date in 1986 with anomalously low values for C 6 F 14 , C 7 F 16 and C 8 F 18 , lower than the atmospheric mole fractions found in the SH samples with similar fill dates. However, this 1986 NH air sample is below the detection limit of the SIO instrument. Furthermore, this sample was not filled for the purpose of an air archive and has been shown to have depleted mole fractions for C 2 F 6 , C 3 F 8 and sulfur hexafluoride (SF 6 ) most likely due to the fill technique. The absolute maximum growth rate is higher in the SH than in the NH for C 6 F 14 and C 8 F 18 . This is most likely a result of the lack of NH data in the 1990s to constrain the cubic smoothed spline fit and not that SH emissions dominate globally.
Recently Laube et al. (2012) also published atmospheric observations for C 4 F 10 , C 5 F 12 , C 6 F 14 and C 7 F 16 . Their   (2012) finds a 2010 mole fraction that is 12 % lower than the estimate in this study. These differences most likely can be attributed to differences in calibration scales. In particular, the calibration scale estimated by Laube et al. (2012) for C 7 F 16 was prepared using an 85 % n-isomer of C 7 F 16 and may be a contributing factor to the differences between the two studies. Based on the globally averaged 2011 atmospheric mole fractions (Tables 3-7), the global radiative forcing of each PFC can be estimated using the radiative efficiencies presented in Table 1. For C 4 F 10 , C 5 F 12 and C 6 F 14 we use the radiative efficiencies given by Forster et al. (2007), and for C 7 F 16 and C 8 F 18 , we use the radiative efficiencies given by Ivy et al. (2012). C 6 F 14 contributes the most of these high molecular weight PFCs to the global radiative forcing in 2011, and is similar to that of C 3 F 8 . The other PFCs in this study contribute approximately equally to global radiative forcing; and in total the high molecular weight PFCs 2011 atmospheric mole fractions contribute to a globally averaged radiative forcing of 0.35 mW m −2 .

Conclusions and future implications
Atmospheric histories and long-term growth trends have been presented for the high molecular weight PFCs: C 4 F 10 , C 5 F 12 , C 6 F 14 , C 7 F 16 and C 8 F 18 . These histories and trends are based on new measurements of a collection of NH archived air samples and a subset of the CGAA. The measurements were made with the Medusa systems and are calibrated against new primary standards for these PFCs. The contribution of all of the heavy PFCs studied here to global radiative forcing is 0.35 mW m −2 . While this is relatively small compared to the total radiative forcing of 2434 mW m −2 in 2008 for all species regulated under the Kyoto Protocol, the heavy PFC atmospheric mole fractions in 2011 contribute up to 6 % of the total anthropogenic PFC radiative forcing (Montzka and Reimann, 2011;Mühle et al., 2010;Oram et al., 2012). The heavy PFCs in this study exhibited the largest growth rates in the 1980s and 1990s and have since slowed, suggesting that recent emissions may be decreasing as alternative compounds, with most likely lower GWPs, are used (Office of Air and Radiation and Office of Atmospheric Programs, 2006). Based on previous studies, atmospheric observations are crucial in providing measurement-based emission estimates to verify bottom-up inventories, which often show large discrepancies Oram et al., 2012). Additionally, future observations of these high molecular weight PFCs will be important in confirming that the semiconductor industry, which primarily focus on the use of low molecular weight PFCs, are indeed reducing global PFC emissions (Semiconductor Industry Association, 2001;World Semiconductor Council, 2005). Although PFCs contribute a relatively small amount to global radiative forcing, due to their long lifetimes they are considered to have a permanent effect on the Earth's radiative budget when human timescales are considered.