Atmospheric Carbon Tetrafluoride: A Nearly Inert Gas

An analysis of existing thermodynamic, photochemical, and kinetic data indicates that the dominant sinks for atmospheric carbon tetrafluoride (CF4) are in and above the mesosphere. Theoretical calculations predict an atmospheric residence time for CF4 of over 10,000 years, about 100 times that for dichlorodifluoromethane (CF2Cl2) and monofluorotrichloromethane (CFC13). It is predicted that CF4 will be well mixed through the stratosphere and mesosphere; only one or two parts of hydrogen fluoride in 1012 are predicted in the high stratosphere as a result of the decomposition of CF4. Although natural sources of CF4 cannot be ruled out, there are several likely industrial sources that may account for its present concentration. The principal environmental effect of CF4 could be the trapping of outgoing planetary infrared energy in its intense bands near 8 micrometers.

Carbon tetrafluoride (CF4) was detected in the air of several European countries by Gassman in 1973 (1) and in both hemispheres of the troposphere by Rasmussen et al. (2) in 1978. This compound is probably the most stable fluorocarbon gas. In this report I examine the potential of atmospheric and environmental processes to decompose CF4 and conclude that it has an atmospheric lifetime of over 104 years. Its principal sinks are vacuum ultraviolet (UV) radiation in the high mesosphere and ionosphere, possible (but unlikely) reactions with electronically excited oxygen atoms and with vibrationally excited (v = 9) OH molecules in the stratosphere and above, several ionospheric processes, and pyrolysis in high-temperature combustion. Its dominant sources are uncertain, but I identify several likely industrial processes. Such a long lifetime (residence time) ensures that relatively small sources will increase the atmospheric CF4 burden and that CF4 will remain ubiquitous in the atmosphere.
In an effort to understand the atmospheric behavior of CF4 as quantitatively as possible, I have used available data on its chemical properties, estimated certain unknown parameters, and performed numerical calculations to simulate its atmospheric transport and photochemistry. Table 1 lists possible CF4 sinks and the reaction rate parameters I adopted. Solar vacuum UV photons with wavelength A < 103 nm photodissociate and photoionize CF4 (3). Although no absorption has been observed for X > 103 nm (3), I have performed calculations with photodissociation cross sections, a-, of 3 x 10-19 and 3 x 10-20 cm2 at X = 122 nm because of the intense solar 121.6-nm line and with ar (122 nm) = 0. These data yield photodissociation rates, J (per second) as follows (3): the total J above the entire atmosphere (altitude, z = cc) is 6.8 x 10-7 due to absorption over all wavelengths, that is, 122 nm and 60 nm < X < 103 nm. With the maximum a-at 122 nm, that is, SCIENCE, VOL. 206, 5 OCTOBER 1979 3 x 10-19 cm2, J (122 nm) = 10-7 at z = o. As a result of absorption by 02 and N2, the only significant photodissociation below an altitude of 90 km is due to the solar lines at 122 and 103 nm. With the maximum a-, J (122 nm) -4.7 x 10-8, 2.6 x 10-8, 1.5 x 10-9, and 3.6 x 10-14 at 90, 80, 70, and 60 km, respectively, when averaged diurnally; J (103 nm) = 9 x 10-9, 3 x10"-1, and 1.2 x 10-23 at z = xc, 90, and 80 km, respectively. Consequences of these rates are discussed below.
Chemical reactions between CF4 and atmospheric constituents are limited by C-F bond strengths. I estimated upperlimit bimolecular rate constants ki through k13 (Table 1) as follows. Reaction 1, an 0 atom insertion, should be slower than the observed (upper limit) rate (4) for 0 + CF3H, an H abstraction.  Table  1 are presumed effective at their upper limit rates. The ground-level source of CF4 was assumed to begin in 1935 at 5.7 x 109 g year-' increasing (at about 6 percent per year) to 4.8 x 1010 g year-l in 1975 and remaining constant thereafter at the 1975 rate. The 1980 profie approximates available data (1, 2). Eventually, the CF4 mole fraction would rise to 4 x 10-8. Also shown is (100 times) the predicted 1980 HF profile due to CF4 decomposition. I assumed an Arrhenius expression of the form A exp(-EaIRT), where A is a constant, Ea is the activation energy, R is the gas constant, and T is the temperature. For a generous upper limit on k1, I took Ea for 0 + CF3H and 0 + CF4 as their respective AH (enthalpy of reaction) values (5), although generally Ea> AH. Multiplying this maximum k, by concentrations of 0 at each altitude from 10 to 110 km, one finds that the maximum product, 10-11 sec-', occurs near the stratopause (50 km) (6).
The only exothermic reaction of CF4 with O('D) other than quenching to O(3P) is the insertion reaction in Table 1; for OCS), two exothermic paths exist. In my calculations I let k4, k5, and k6 = 0 and 10-12 cm3 sec1I separately (7). Reaction with H atoms, although exothermic, requires at least 10 kcal per mole of activation energy (8). Reactions of groundstate OH with CF4 are strongly endothernmic and thus negligible. Vibrationally excited OH (v = 1,2, * * 9) is present in the stratosphere and mesosphere (9). It is possible for CF4 to react with OH (v = 9); I estimated reaction rates kl and kl12 in Table 1. Dissociative electron capture by CF4 is strongly endothermic (10) so that free electrons in the lower ionosphere are insignific4nt.
Above 150 km ionospheric photoelectrons with energies > 2 eV would dissociate CF4, but this process would be very slow globally because the CF4 available for attack there is 10-9 of that in the troposphere.
The nonphotochemical processes in Table 1 were evaluated as follows. If all rainfall is saturated with CF4 with respect to its atmospheric partial pressure, pCF4, and 1 m of water falls annually, the ratio of annually precipitated CF4 to the atmospheric content is 2 x 10-6 for any pCF4, according to its solubility (5). Moreover, because the rate of hydrolysis of CF4 is immeasurably small (11), the capacity of the oceans to assimilate-CF4 is limited by its solubility; if all ocean waters (surface and deep) were saturated with CF4, less than 0.2 percent of the atmospheric burden of CF4 at any time would be in the oceans. The removal time due to pyrolysis in Table 1 is based on annual 02 consumption rates in hightemperature combustion (11). Finally, there are nQ indications of biological processes that can break C-F bonds in CF4 (12).
With the photochemical reactions of Table 1, I performed numerical calculations (13) for the time and altitude dependence of CF4 concentrations. A constant ground-level source of CF4 natural-0036-8075/79/1005-0059$00.50/0 Copyright K 1979 AAAS Table 1. Bimolecular rate constants (in cubic centimeters per second) and process rates for the decomposition of atmospheric CF4. Additional reactions between CF4 and NO, NO2, CO, 03, HO2, SO2, Cl, CIO, HCl, other gases, and positive ions were considered but no exothermic paths exist. Although the reaction H2 + CF4 -HF + CF3H is exothermic, it is expected to have a prohibitive activation energy as is the case for NO3 + CF4 -* NO2 + CF30F. Other natural sinks, for example, solar x-rays, ionospheric photoelectrons, lightning discharges, and surface reactions with atmospheric aerosols are active but unimportant quantitatively (11   years, but this is unrealistic because nonphotochemical processes such as combustion would limit the CF4 residence time (at least until fossil fuels are depleted). I also performed calculations in which a time-varying, presumably industrial CF4 source was used. Figure 1 shows altitude profiles of the CF4 mole fraction at four dates. All photochemical destruction rates for CF4 in Fig. 1 were the maximum values of Table 1. The present atmospheric burden of CF4, about 1012 g (mole fraction = 6 x 10-"1), could have materialized thus. Because the CF4 sinks are in the mesosphere and above except for O('D) and OH* (minor compared to 122-nm photodissociation), only about 10-3 to 10-2 of the atmospheric CF4 is exposed to the sinks at any time. The local photochemical removal processes have characteristic time constants of over 100 60 years, and r = 14,000 years. Thus the CF4 concentration increases with time, and the altitude profile approaches a straight line. At steady state, the ratio of the 90-km mole fraction to the 0-km mole fraction is 0.97. Thus the release of four F atoms in CF4 decomposition leads to very little HF (Fig. 1). The dominant inorganic F compound in the stratosphere and mesophere should be HF (14). If CF4 is a recent addition to the air, its mole fraction, f, will be less in the stratosphere and above than at z = 0, for example, the 1980 profile in Fig. 1. If, instead, CF4 was injected into the air over 100 years ago, it should be well mixed (like the A.D. 2105 profile). In any case, an atmospheric residence time of > 14,000 years implies that, even with no sources, a given CF4 profile will decay by only 7 percent in 1000 years (e-"11), as I have confirmed by additional numerical computations.
With such a long atmospheric lifetime, CF4 concentrations will grow because CF4 has industrial sources, for example, CF4 released in the electrolytic reduction of alumina (15). Indeed, the liberation of CF4 is not entirely limited to relatively brief anode-effect intervals. Japanese studies (15) suggest that the estimate of Rasmussen et al. (2) of 6 x 109 g of CF4 per year released during electrolytic cell anode effects are supplemented by CF4 released during normal cell operation.
Release of as much as 1010 g year-' is easily possible, but changing industrial practices, technology, and output imperil such estimates; direct measurements are needed. Several other industrial processes are likely sources of CF4, either because CF4 has been detected in the process or because C, F, and heat are available: the electrolytic generation of F2, especially in carbon or graphite electrode systems, the analogous reduction of UF4 or UF6 (16), the use of fluorspar in steelmaking (17, p. 298; 18) (although CF4 production from CaF2 seems disfavored thermodynamically), the burning of polyfluoroethylenes, rocket fuel combustion (18), and inadvertent production in fluorocarbon manufacture. Direct, intentional industrial production of CF4 amounts to only 10 to 60 tons per year.
Sources of CF4, especially anthropogenic, require measurements to quantify, as does the possible CF4 photoabsorption at 122 nm. With these data, more accurate predictions of CF4 concentrations and effects will be feasible.
Note added in proof: A recent stratospheric measurement by Goldman et al. (19) found 7.5 x 10-"1 (mole fraction CF4) at 25 km, as is roughly predicted by the curve for 1980 in Fig. 1  10. For a C-F bond energy of 129 kcal mole-' (5.61 eV), e + CF4 -* F-+ CF3 is endothermic by 2.27 eV because the electron affinity of F is 3 Abstract. From the echoes of elastic waves incident on inclusions in solids, one may extract certain resonance features. These "spectral lines" and their widthsform a code identifying the material composition of the inclusion in a way that resembles spectroscopy. This idea finds applications in geophysics, materials science, and any field dealing with materials containing inclusions.
The amplitudes of backscattered waves returned by inclusions in viscoelastic solids, when plotted as a function of frequency, exhibit so many rapid oscillations and complicated features that until very recently it was not possible to extract the physical information contained in them. The amplitudes of these waves can be analyzed in light of our new resonance theory of scattering from cavities in solids (1, 2) and can be used to identify, for a given shape of the cavity, the material composition of the filler substance. When a (spherical) filler is set into oscillation by elastic (say, compressional) waves incident upon it, a set of modal resonances (fundamental and overtones) gets excited in it; these resonances characterize the filler as if they were its signature. Since incident shear waves excite the same resonances in the filler, we will limit this analysis to incident compressional waves and we shall consider fluid fillers only. From the usual spectral plots of the backscattered wave amplitudes versus nondimensional frequency x kda, it is possible to obtain these resonances, which manifest themselves as narrow lines or wider "'spikes" (kd = (/Cd, where X is the circular frequency of the incident wave, Cd is the speed of the compressional waves, and a is the cavity radius). These plots display a quantity which, for simplicity, we will call "the echo." The way the resonances of an unknown filler are thus being used for purposes of material discrimination resembles the way chemical elements are identified from their optical spectra. The resonances obtained from the (heretofore physically incomprehen-sible) echo plot lead directly to a deciphering of the code, indicating the composition of the filler material that is contained in the echo.
Plane p (that is, compressional) elastic waves incident on fluid-filled spherical cavities in solids produce two scattered waves, one compressional and the other shear (that is, s). The scattering amplitudes fPP or fP5 of both these scattered waves could be analyzed, but, since all the main points of this report can be shown from either one of these, we choose f"PP(). This nonmode-converted, normalized amplitude can be shown (1) to be f _(ft) =_ fnPP() a n=O a ac(2n +l) A5P5(cos 0) (1) n0 ikda where the coefficients An are given by ratios of two 3 x 3 determinants whose elements contain products of the fillerto-matrix density ratio (that is, pf/p) with various spherical Bessel and Hankel functions and their derivatives, of arguments kda and ksa, and of order n. These elements are determined from the boundary conditions of the problem and are given in (1). In the backscattering direction 0 ti, the Legendre polynominals are simplified by means of the relation Pn (cos 7r) = (-_ )n. Figure 1 shows the plot of the modulus of this summed backscattered amplitude for a cavity filled with ethyl alcohol in an aluminum matrix. This is the "echo" containing the rapid oscillations and complex features mentioned above. 0036-8075/79/1005-0061$00.50/0 Copyright G) 1979 AAAS