Perfluorinated compounds impact the Earth's radiative balance. Perfluorotributylamine (PFTBA) belongs to the perfluoroalkyl amine class of compounds; these have not yet been investigated as long-lived greenhouse gases (LLGHGs). Atmospheric measurements of PFTBA made in Toronto, ON, detected a mixing ratio of 0.18 parts per trillion by volume. An instantaneous radiative efficiency of 0.86 W m−2 ppb−1 was calculated from its IR absorption spectra, and a lower limit of 500 years was estimated for its atmospheric lifetime. PFTBA has the highest radiative efficiency of any compound detected in the atmosphere. If the concentration in Toronto is representative of the change in global background concentration since the preindustrial period, then the radiative forcing of PFTBA is 1.5 × 10−4 W m−2. We calculate the global warming potential of PFTBA over a 100 year time horizon to be 7100. Detection of PFTBA demonstrates that perfluoroalkyl amines are a class of LLGHGs worthy of future study.
 Many perfluorinated compounds are trace atmospheric constituents of radiative significance. Their unique physical-chemical properties result in a propensity to partition to the atmosphere, where they can act as long-lived greenhouse gases (LLGHGs). Perfluoroalkyl amines (PFAms, CxF2x+1N(CyF2y+1)CzF2z+1) are a class of thermally and chemically stable liquids marketed for use in applications, including electronic testing, and as heat transfer agents [3M, 2000]. These compounds are synthesized using electrochemical fluorination (ECF), which yields mixtures of structural isomers [Kissa, 2001].
 Perfluorinated compounds have no known sinks in the lower atmosphere. As a result, their lifetimes are on the order of hundreds of years and typically dominated by destruction in the upper atmosphere [Prather and Hsu, 2008; Ravishankara et al., 1993]. Compounds containing C-F bonds can impact climate because they exhibit strong absorption bands in the optically thin spectral region of the atmosphere (750–1250 cm−1). Despite their low overall volume mixing ratios (VMRs), halocarbon compounds contribute 0.337 W m−2 or 13% of the total radiative forcing (RF) of LLGHGs [Forster et al., 2007]. There are perfluorinated chemicals, including perfluoroalkyl amines, for which atmospheric fate and concentration data are not available. These chemicals may contribute to radiative forcing of climate change.
 Here we present experimental results that suggest that one member of the previously uncharacterized compound class of PFAms, perfluorotributylamine (PFTBA, N(C4F9)3), is likely a LLGHG. The historical, current, and projected production rates and inventories of PFTBA have not been disclosed by the manufacturer(s). However, inventory use reports submitted by a manufacturer to the United States Environmental Protection Agency High Production Volume (HPV) Challenge Program reveals that PFTBA was produced in, or imported to, the U.S. at HPV rates (≥1 × 106 lb yr−1). More relevantly, the emission levels of PFTBA governed by industrial use practices are unknown, leading to high uncertainty regarding its environmental loading and impact. The objective of this study was to assess the potential of PFTBA to act as LLGHG. The LLGHG potential of PFTBA was determined by the following: (i) assessing its properties and environmental fate, (ii) measuring its atmospheric mixing ratio, (iii) measuring its IR absorption cross section and calculating the radiative efficiency (RE), and (iv) estimating the climate impacts of PFTBA.
2.1 Determination of Physical Properties and Fate
 We examined the structural composition of neat PFTBA using a Bruker Avance III 400 nuclear magnetic resonance (NMR) spectrometer equipped with an automatic tune and match probe tuned to 19F (376.14 MHz) with a deuterated water capillary insert.
 Most perfluorinated compounds partition exclusively to the gas phase. This was examined using a high-resolution multiple-species (HR-MS) model developed by Cahill and Mackay . The model includes a separate aerosol phase in each of the air phases. This HR-MS model can describe the fate of interconverting chemical species, which allows for modeling of processes such as ionization. Details of the physical property estimation and modeling can be found in the supporting information. Briefly, the few environmentally relevant physical property measurements for PFAms available were used. Additional properties for PFTBA were estimated using an online calculator [Hilal et al., 2004] and were verified through comparison with experimentally measured values for other perfluorinated compounds. Physical properties for the protonated analog of PFTBA, PFTBAH+, were determined from the properties of PFTBA using the ratio of experimentally measured properties for perfluorooctanesulfonic acid to perfluorooctanesulfanate. Although the pKa of PFTBA is unknown, the pKa of perfluorotripentyl amine has been determined to be less than −0.5 [Boswell et al., 2005]. The pKa of PFTBA is likely similar, but as the pKa could not be determined explicitly, values from 2 to −2 were used in the model to determine the effect of ionization.
2.2 Gas Standard Preparation
 Gaseous primary stock standards were prepared in-house by volumetric, serial dilution with a known volume concentration of neat PFTBA using the method developed by Sin et al. . Briefly, the stock standard was prepared in a clean electropolished Summa® canister by weighing a known amount of PFTBA with a microsyringe. The canister was pressurized with humidified zero air accurately controlled by a mass flow controller (MFC). The stock standard was allowed to mix at room temperature for 1–2 weeks prior to final dilution with humidified zero air using a MFC. Two primary stock standards were prepared in this way, achieving a final mixing ratio of 8.75 and 20 parts per billion by volume (ppbv) (accuracy of 10%).
2.3 Gas Handling Manifold and Analysis
 PFTBA was quantified by gas chromatography mass spectrometry (GC-MS) (Agilent Technologies 6890 GC, 5973 N MS) interfaced directly with a custom-designed and built gas-handling manifold. The MS was equipped with a chemical ionization source and acquired mass spectral data with methane (4.0 grade, Linde) as the reagent gas in electron capture negative ionization (ECNI) mode. ECNI exhibits high sensitivity and selectivity for fluorinated compounds that possess inherently high electron affinity. The spectrometer used in these experiments was tuned using perfluoro-5,8-dimethyl-3,6,9-trioxidodecane.
 Calibration was achieved by injecting small volumes of the 20 ppbv standard into a sample loop cooled to −185°C by liquid Ar. PFTBA was delivered to the GC-MS by thermal desorption using a stainless steel block heated to 340°C via a heated transfer line that was directly coupled to the analytical column. Separation on a GasPro column (50.0 m × 0.32 µm, Agilent Technologies) was achieved using an initial temperature of 100°C held for 20 min, then by increasing the oven temperature at 10°C min−1 to 240°C, followed by a 3°C min−1 increase to 260°C, which was held for 13 min. Operating in single-ion monitoring mode, ion fragments with the mass-to-charge ratios (m/z) of 633 (M–2F), 452 (M–C4F9), and 414 (M–C4F11) were used to detect PFTBA (Figure 1a).
2.4 Atmospheric Sampling
 Ambient air samples were collected during November and December 2012 through a sampling port in the Lash Miller Chemical Laboratories at the University of Toronto (43°39′42.67″N and 79°23′51.66″W). Using a vacuum pump and a calibrated MFC, ambient air was pulled for 60–90 min through ascarite (Thomas Scientific) and drierite (W. A. Hammond Drierite Company) scrubbers at a rate of 70–200 standard cubic centimeters per minute (sccm) prior to flowing into stainless steel sample loops (20 inch × 1/4 inch OD) cooled to −185°C. Cryofocused ambient air samples were analyzed by GC-MS in the same manner as calibration standards as described above. Quantification was achieved by external calibration using the extracted ion chromatogram of the major fragment, with a m/z of 633.
 The extraction efficiency of our sampling method was characterized by sampling a known amount of PFTBA in the laboratory. The 8.75 ppbv PFTBA standard was diluted with nitrogen gas using calibrated MFCs. From this, the majority of the flow was vented, while an exactly measured flow close to 70 sccm was pumped through two sample loops connected in series. The total extracted mass of PFTBA collected in the first loop was compared to the second to describe the amount of breakthrough.
2.5 Infrared Absorption Measurements
 Measurements were performed in a 140 L Pyrex reactor interfaced to a Mattson Sirus 100 Fourier transform infrared spectrometer. Infrared spectra were derived from 32 coadded interferograms with a spectral resolution of 0.25 cm−1 and an analytical path length of the infrared beam of 27.1 m. Spectra were recorded at 296 K using 1.54–4.26 mTorr of PFTBA in N2 at 760 torr. The manufactured compound was obtained from a commercial source (Fluorinert FC-40, undisclosed purity, Sigma Aldrich, Oakville, ON) and was subjected to repeated freeze-pump-thaw cycles to remove volatile impurities (e.g., air) in the sample. Four separate spectra were recorded, with one spectrum taken while warming from −196°C; no differences in the spectra were discernible, suggesting that the sample was free of significant impurities. Peak absorbances were in the range 0.05–0.7 and scaled linearly with the PFTBA concentration.
2.6 Determination of Radiative Efficiency and Radiative Forcing
Pinnock et al.  presented a method to estimate the RE directly from IR absorption spectra. Briefly, the experimentally measured spectra are divided into 10 cm−1 wide bins, and the averaged absorption cross section in each bin is multiplied by the instantaneous cloudy-sky RF per unit cross section for the global and annual mean atmosphere as derived by Pinnock et al. . This method has been demonstrated for a series of hydrofluorocarbons and hydrochlorofluorocarbons to yield an error of less than 1% when compared to RE values calculated with a more rigorous narrowband absorption model [Pinnock et al., 1995].
2.7 Determination of Global Warming Potential
 The global warming potential (GWP) of PFTBA can be calculated from the time-integrated radiative forcing (RF) over a specified time horizon (H) following a pulse unit mass release [absolute global warming potential (AGWP)] relative to that for the reference gas CO2, i.e.,
 Assuming an exponential decay of PFTBA, integration of the radiative forcing gives
where REPFTBA and τ are the radiative efficiency and atmospheric lifetime of PFTBA, respectively. We report global warming potentials over a 100 year time horizon (GWP(100)) because this is a widely used (e.g., in national and international regulations) metric for comparing the contribution of different greenhouse gases to climate change.
3 Results and Discussion
3.1 Properties and Environmental Fate of PFTBA
 The presence of branched and linear isomers was confirmed in neat PFTBA using 19F NMR (Figure 1b). The presence of structural isomers was expected, given the known products of ECF synthetic methods. The chemical shift observed at −183.43 ppm is assigned to the fluoromethylidyne group (-CF-) present in the PFTBA isomers containing an isoperfluorobutyl chain. The chemical shift at −76.167 ppm is associated with two equivalent perfluoromethyl groups (-CF3) from the isoperfluorobutyl chain. The mixture of structural isomers observed with NMR is consistent with gas chromatographic behavior, which shows total PFTBA as the sum of multiple peaks that we ascribe to the presence of isomers.
 Perfluorinated compounds are unreactive toward tropospheric oxidants. Thus, any oxidation would likely occur in the upper atmosphere. It is possible that PFTBA could protonate in the atmosphere, leading to different fates through wet or dry deposition. Given the expected low pKa of PFTBA and using an equilibrium partitioning model described in the supporting information, we demonstrate that protonation of PFTBA is extremely unlikely and has a negligible effect on the overall environmental fate of PFTBA. As such, PFTBA can be assumed to be a neutral molecule in the gas phase, subject to gas-phase loss pathways.
 In the stratosphere, reaction with singlet oxygen (O(1D)) can be an important loss mechanism. Although reactions of O(1D) with PFTBA have not been examined, reactions with a number of other fluorinated compounds have been characterized. The primary result of O(1D) reaction with perfluorinated alkanes is physical quenching of O(1D) to O(3P) rather than chemical reaction [Ravishankara et al., 1993]. In many cases, the yield of O(3P) is unity, within error. In contrast, reactions of O(1D) with both NH3 and NF3 lead to products other than O(3P) [Sanders et al., 1980; Sorokin et al., 1998], with NF3 more readily proceeding through reactive channels. It is unlikely that PFTBA is more reactive than NF3, given the low bond strength of the N-F bond (276 kJ mol−1) [Molina et al., 1995] compared to the N-C and C-F bonds in PFTBA. Another potential loss pathway in the upper atmosphere is the reaction with mesospheric free electrons, which have been shown to be important in the atmospheric fate of SF6 [Ravishankara et al., 1993]. Studies have not examined the potential of this fate for amine compounds, but it is a thermodynamically feasible destruction pathway for PFTBA. Photolysis by Lyman-α radiation (121.6 nm) in the mesosphere has been demonstrated to be the dominant fate of perfluorinated alkanes and an important fate for SF6 [Ravishankara et al., 1993]. It is probable that PFTBA is also subject to photolysis at this wavelength. Determining the lifetimes and the relative importance of reaction with O(1D), free electrons, and photolysis by Lyman-α radiation is difficult given the limited information. An estimate can be made for the overall lifetime of PFTBA by comparison to other perfluorinated compounds. The perfluorinated compound with the shortest lifetime is NF3, where loss is dominated by stratospheric photolysis, with some contribution from reaction with O(1D) (as discussed above) [Prather and Hsu, 2008]. The lifetimes of other perfluorinated compounds shown in Table 1 depend primarily on reactivity in the mesosphere and, consequently, are much longer. It is unclear whether PFTBA would undergo photolytic degradation in the stratosphere. The n → σ* transitions drive photolysis for NF3, and it is possible that a similar mechanism could occur for PFTBA. Assuming that PFTBA could undergo photolysis in the stratosphere and have some reactivity in the mesosphere, by analogy to NF3, we adopt the shortest lifetime over a 2σ range to account for uncertainties in input kinetic and photochemical parameters [Papadimitriou et al., 2013], and we estimate a lower limit for the lifetime of PFTBA of 500 years.
Table 1. Radiative Efficiencies and Global Warming Potentials for Selected Non-CO2 Long-Lived Greenhouse Gases
 In all atmospheric samples (n = 16), PFTBA was positively identified based on the match of retention times and presence of ions (m/z of 414, 452, and 633, Figure 2a). We observed a slightly dissimilar chromatographic profile in the ambient atmospheric samples compared to the gas standard. It is possible that the isomeric composition of the atmospheric samples is different from the neat PFTBA in the standard. Batch-to-batch variability in isomer composition is well documented for ECF synthesis of perfluorinated species [Benskin et al., 2010]. The developed method is highly specific, based on the unique physical properties of PFTBA. Because of the relatively low vapor pressure of PFTBA, the inherent selectivity of ECNI, and the selective detection of massive ions, false detection is highly improbable.
 In the measurements presented here, we assign an uncertainty of 15% to account for variability in the use of external calibration and ionization variability by ECNI. Other sources of uncertainty include standard preparation, environmental variability, and sampling error. We estimate that the total random uncertainty is approximately 30%. Results from breakthrough experiments indicate that PFTBA is captured efficiently by the sampling method (~90%). However, unidentified losses (likely upstream of the sampling loop) were also observed. As such, the concentrations we report here can be considered a lower limit of the true VMR. The composite mean VMR of PFTBA was determined to be 0.18 (±0.01) parts per trillion by volume (pptv) (n = 16, P < 0.05, t = 2.131).
 Back trajectories were obtained using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (R. R. Draxler and G. D. Rolph, HYSPLIT—Hybrid Single-Particle Lagrangian Integrated Trajectory Model, 2013, http://www.arl.noaa.gov/HYSPLIT_info.php; G. D. Rolph, READY—Real-time Environmental Applications and Display System, 2013, http://www.ready.noaa.gov/) to predict the path of the air mass from which the samples were obtained. We do not observe any correlation between the VMR and the air mass trajectory (Figure 2b). Samples from two sites (43°41′25.04″N, 79°23′15.11″W and 43°39′53.42″N, 79°18′48.45″W) located several kilometers upwind of the sampling port at the Lash Miller Chemical Laboratories were obtained during method development. The measurements were statistically indistinguishable from those obtained from the Lash Miller site. The absence of any discernible effect of air mass trajectory or sampling site suggests, but does not prove, the absence of significant local sources. Assuming that the concentration observed in Toronto (0.18 pptv) is representative of the global VMR, the atmospheric burden is approximately 4.7 × 107 lbs (2.1 × 107 kg) PFTBA, which we attribute entirely to anthropogenic activity since the Industrial Revolution. Further work is needed to confirm our assumed present-day and historical global background levels.
3.3 Infrared Spectrum and Radiative Efficiency
 The absolute absorption spectrum of PFTBA is shown in Figure 3. The total integrated cross section is 7.08 × 10−16 cm2 molecule−1 cm−1 (see supporting information). Uncertainties in the cross section measurement arise from the following sources: sample concentration (2%), sample purity (2%), path length (1.5%), spectrum noise ±(10−20 cm2 molecule−1), and residual baseline offset after the subtraction of the background (1.5%). From these individual uncertainties, the total (random) uncertainty in the integrated absorption cross section is ±4%. We prefer to quote a conservative uncertainty of 5%, yielding an integrated cross section for PFBAm of (7.08 ± 0.35) × 10−16 cm2 molecule−1 cm−1. Using the method of Pinnock et al.  and the measured spectrum, the instantaneous, cloudy-sky RE of PFTBA was determined as 0.86 W m−2 ppb−1. It should be noted that this instantaneous RE neglects the effect of the stratospheric adjustment. For halocarbons, this underestimates the adjusted RE by 5–10% [Hodnebrog et al., 2013].
3.4 Climate Implications
 The RE determined above for PFTBA of 0.86 W m−2 ppb−1 is the highest for LLGHGs (Table 1). Previous to this study, the compound with the highest RE that had been detected in the atmosphere was SF5CF3, with a RE of 0.57 W m−2 ppb−1 and a concentration of 0.15 pptv [Sturges et al., 2012]. Proceeding on the assumption of a global average atmospheric concentration of 0.18 pptv (with the caveats noted above), the estimated RF of PFTBA is 1.5 × 10−4 W m−2. This is higher than the RFs of SF5CF3 and C4–C6 perfluorinated alkanes (Table 1), equivalent to approximately 3% and 38% of the RFs of the Kyoto-regulated gases HFC-134a and HFC-152a, respectively, but approximately 10,000 times lower than the 1.66 W m−2 RF from the increase in atmospheric CO2 since the Industrial Revolution (1750) [Forster et al., 2007].
 Using the data from Hodnebrog et al. , we derive AGWPCO2(100) = 0.715 W m−2 ppm−1 yr. Using REPFTBA = 0.86 W m−2 ppb−1 and τ = 500 years and accounting for the differences in molecular masses needed to place the REs on a per mass basis, we calculate that PFTBA has a GWP(100) of 7100. This is lower than the GWP estimate provided by the manufacturer of 9020 [3M Company, 2011]. Given that the lifetime of PFTBA greatly exceeds the time horizon in the GWP calculation, the GWP value is relatively insensitive to uncertainties in the lifetime (varying the lifetime by 20% leads to a 4% change in the GWP).
 Two other PFAms, perfluorotripropylamine (N(C3F7)3) and perfluorotripentylamine (N(C5F11)3), are on the EPA list of HPV chemicals and may be present in the atmosphere at levels similar to that of PFTBA. We have demonstrated the potential for PFAms to act as LLGHGs. Further measurements in remote locations are needed to firmly establish the global background concentration of PFAms.
 The Universty of Toronto authors are grateful to Frank Wania and James Armitage for their modeling support. We thank the Chemistry department machine, glass, and electronics shops, along with John Sagebiel, Mike Keith, and Trevor VandenBoer for their assistance with the instrument development. We thank Daniel Wang for the preparation and donation of gas standards. We thank Jamie Donaldson for his helpful discussions while preparing the manuscript. This work was funded by a Natural Science and Engineering Research Council (NSERC) Discovery grant and NSERC CGS-M and CGS-D scholarships to A.C.H. and C.J.Y. A.C.H. and C.J.Y. contributed equally toward the research presented in this paper.
 The Editor thanks James Franklin and an anonymous reviewer for their assistance in evaluating this paper.