Envlron. Sci. Technol. 1083, 17, 183-186
Carbon Balances in Simulated Atmospheric Reactlons: Aromatic Hydrocarbons Robert J. O’Brien,* Patrick J. Green, Ngan-Lien Nguyen, Richard A. Doty, and Bruce E. Dumdei Department of Chemistry and Environmental Sciences, Doctoral Program, Portland State University, Portland, Oregon 97207
Carbon balances for aromatic hydrocarbon photooxidation systems, calculated as the sum of detected products, have been poor. With use of measurements of total gas-phase carbon during the course of the reaction, direct evidence for heterogeneous removal of intermediate products has been obtained. Experiments with toluene show 63% of the carbon leaves the gas phase during the reaction. A comparison of normalized CO data is made between our chambers, the SAPRC chamber, and the UNC chamber, to show their similarity. It is concluded that heterogeneous processes have important implications for the understanding of aromatic systems.
conditions. To our knowledge, this type of study has never previously been attempted-in part because it is difficult or even impossible in many smog chambers, due to high background levels of organic compounds and/or carbon dioxide. Our results indicate that ignoring heterogeneous processes in reactions of all aromatic hydrocarbons, at least in smog chambers, may be a serious error. For instance, recent detailed mechanisms for the reactions of toluene in a smog chamber make no mention of wall loss of reaction products (8,9).However, these studies required invoking a “wall”source for free radicals, thus indicating the likelihood of condensed products.
Introduction The chemistry of polluted air has been studied in the laboratory for almost 30 years (1). Much effort has been devoted to the identification of the intermediate hydrocarbon oxidation products. For alkanes and alkenes, common atmospheric hydrocarbons, these products are often aldehydes and ketones, but peroxides such as PAN (peroxyacetyl nitrate) are formed as well (2). Organic nitrates are also found as products, but the yield of these compounds is proportional to the NO, concentration and is not expected to be very significant for NO, concentrations found in polluted atmospheres. In the case of cyclic alkenes (e.g., cyclohexene or the pinenes) and for dialkenes, multifunctional products are formed, and acids are found among the products. These compounds have been well demonstrated to form products of sufficiently low volatility that they condense out of the gas phase to form aerosols (3-5). However, the suspended aerosol yield, even for the most prolific aerosol formers, is only a fraction of the total hydrocarbon consumption. Considerable mention has been made of the carbon balance in smog chamber experiments. The goal, of course, is to account for all the reacted hydrocarbon in terms of intermediate product molecules, with the ultimate gasphase products being carbon monoxide and carbon dioxide (CO is slowly oxidized to C02 as well). In the past, this carbon balance could only be attempted by adding up the sum of all the concentrations of the detected products. Carbon balances of this type for the lower molecular weight alkanes and alkenes are fairly good, with the identified products accounting for most of the reacted hydrocarbon over the course of the reaction. The case of carbon balances for the higher molecular weight alkanes and alkenes has been studied to a very limited extent. Carbon balances for the aromatic hydrocarbons, although the subject of considerable study, have always been extremely poor (6). Lack of adequate product data in a sense makes the job of chemical mechanicians easier, since it reduces the constraints put upon their mechanisms. However, the reliability of these models must of necessity remain in doubt until better data are obtained. Our own experience in studying the products of the reactions of aromatic hydrocarbons under simulated atmospheric conditions (7) has convinced us of the importance of determining carbon balances before undertaking explicit model development. Consequently, we have carried out a direct study of the yield of gas-phase hydrocarbon products formed under simulated atmospheric
Experimental Section The reactions were carried out in evacuable Pyrex vessels of 22- and 239-L volumes (IO). The flasks were spherical and were fitted with Teflon-plug glass stopcocks. The smaller vessels could be cleaned by heating to 723 K in an annealing oven. The large vessel was cleaned by radio frequency glow discharge at a pressure of about 1torr of oxygen, a technique similar to that used in the semiconductor industry for cleaning substrates (11). Both these processes served to remove condensed hydrocarbon products of previous experiments from the walls. The cleanliness of the vessels was monitored by irradiating air-zero gas (50% relative humidity) with about 1part per million (ppm) of nitrogen dioxide. Oxides of nitrogen were measured by a chemiluminescent NO, analyzer (Therm0 Electron Corp.). During this preirradiation, which lasted several days or weeks if necessary, CO, C02, and total gas-phasecarbon (TGC) were measured (TGC in the 239-L flask only). When the background rate had been characterized sufficiently, the hydrocarbon was added to initiate the reaction. Background TGC were generally less than 2 ppmC, C02 was generally less than 1ppm, and CO was less than 0.05 ppm. The flasks were irradiated with fluorescent lights to simulate the ultraviolet portion of the solar spectrum. The large flask had an equal combination of fluorescent black lights and fluorescent sun lamps. The overall light intensity gave an NO2photolysis rate (k,)of about 0.2 min-’ (12). The absolute spectral intensity distribution was not measured but was probably deficient at the short-wavelength end compared to actual sunlight. The 22-L flasks were irradiated with black lights only. We obtained the gas-phase carbon measurements by passing the air sample for analysis over an oxidative catalyst heated to 923 K (13). This oxidized all hydrocarbons and carbon monoxide to C02,which was then measured quantitatively by gas chromatography. We call this measured quantity total gas-phase carbon. Since no filtering of the air sample was used, TGC includes suspended aerosol when present. The total carbon analysis was calibrated quantitatively by pressure/volume expansion into the large reaction vessel. Known amounts of various hydrocarbons and carbon monoxide were compared in their GC response to carbon dioxide to verify their quantitative conversion to COB Then, in each experiment, the hydrocarbon added initially was calibrated as C02 after combustion in the flow system. The sample lines leading from the flask to the combustion chamber were heated to 325 K. In addition to TGC, CO, COP,and hydrocarbon
4
0013-936X/83/0917-0183$01.50/0
0 1983 American Chemical Society
Environ. Sci. Technol., Vol. 17,No. 3, 1983
183
*
O
50%. All these yields are uncorrected for the background CO and C02generation rate and for the suspended aerosol formed from the oxygenated aromatics. Thus they represent upper limits. The yield of gas-phase products other than CO and C02may be seen in Figure 1 as the difference between the two uppermost curves. Only a fraction of this was accounted for as measured products. At longer reaction times, the TGC deficit decreased as the material on the walls and/or its gas-phase vapor component continued to react and form CO and C02. Figure 1 indicates this at times after 1800 min. Several reactions were carried out in the small 22-L flasks for extended periods, up to a maximum of 40 days preirradiation and 40 days after toluene was added. In this case a complete recovery of the added toluene as C02 and CO was obtained. However, in these experiments it was not possible to measure TGC.
A
15
u
E
a
g
IO
t 4
t
w z 0
5
0 0
1200 REAClION 7ItlE
M)(l
le00
2400
3000
nln
Figure 1. Typical experimental data for reaction carried out in our 2 3 9 1 reaction vessel. The TGC deficit is indlcated by the upper curve. The concentration of gas-phase products other than CO and COP is represented by the difference between the two uppermost curves. The initial concentrations of CO, COP,and TGC have been subtracted out (corrections of 10-20%). All reported concentrations are in parts per million as carbon atoms (ppmC).
were measured by gas chromatography as a function of time. Carbon monoxide has often been measured as a reaction product, but measurement of carbon dioxide is not frequently carried out, a major hindrance in determining the fate of atmospheric hydrocarbons, we believe. Gas-phase carbon balances were measured for several hydrocarbons in addition to the aromatics. Hydrocarbons such as cyclohexene that are known to be prolific aerosol formers were not studied. In the case of the low molecular weight hydrocarbons, current mechanisms predict very good gas-phase carbon balances. We observed this to be the case for propene, acrolein, formaldehyde, acetaldehyde, butane, and hexane, all of which gave essentially 100% TGC yields.
Results Experiments were carried out with two aromatic hydrocarbons, toluene and o-xylene, and with two known toluene reaction products, benzaldehyde and o-cresol. All these compounds gave large TGC deficits. The aromatics did this without formation of any condensation nuclei while both benzaldehyde and o-cresol generated large amounts of condensation nuclei at the time the lights were turned on. Toluene was studied in considerable detail since it is one of the most abundant atmospheric hydrocarbons, and services as a prototype for the reactions of the aromatic hydrocarbons as a class. This class of hydrocarbon presents the last major area of uncertainty about polluted-air photochemistry, and major efforts are currently underway to learn the mechanism of its atmospheric reactions (14). Figure 1 shows typical carbon data that indicate that during the first 30 h of continuous irradiation, toluene reaction gave a gas-phase carbon (TGC) yield of only 37%. Other reactions of toluene in the 239-L reaction vessel gave similar results. The initial concentrations ranged from 0.2 to 4.3 ppm, and the TGC yields were not significantly higher at lower concentrations. In the case of o-xylene, the TGC yield was 53 % . The TGC yield from o-cresol was 184
Environ. Sci. Technol., Vol. 17, No. 3, 1983
Discussion Since experiments of this type have apparently not been carried out previously, we cannot compare our TGC results to data from other reaction vessels. However, we can compare the carbon monoxide data from both our large and small reaction vessels with the extensive data available from the large, evacuable smog chamber at the Statewide Air Pollution Research Center at the University of California, Riverside, and with data from the outdoor Teflon chamber at the University of North Carolina. These data are shown in Figure 2. The data presented in Figure 2 can be simulated with a simple kinetic scheme representing CO in terms of prompt and delayed yields as follows (IO):
-
+ HO k l al(CO) + other products ki toluene + HO a2(intermediate)
toluene
intermediate
+ HO
ka
(1) (2)
b(C0)
(3) Here, kl and k2 are rate constants for the overall reaction with HO; reaction 1 gives a direct CO yield and reaction 2 produces an intermediate which ultimately produces some CO as well; ul, u2, and b are the fractional yields for the various steps. For simplicity, photolysis of the intermediate and overall dilution have been ignored here. The effect of dilution was minimal since the chemical lifetime of toluene was much shorter than the dilution lifetime (about 500 and 5000 min, respectively) in our large reaction vessel. Photolysis of the intermediate can be incorporated into an effective value of k2. Analytical solution of these kinetic differential equations on a time-independent basis yields the following equation for the reaction product, carbon monoxide, expressed as a function of the concentration of its parent hydrocarbon, toluene:
Here, [TI is the toluene concentration; the subscript 0 indicates initial concentration-all other concentrations are at time t. R is the rate constant ratio k2/kl. The first term is the direct CO yield in reaction 1. The second term is the indirect CO yield from the intermediate. The actual mechanism is undoubtedly more complex than reactions 1-3, with several intermediates being important and probably multistep CO formation as well. Neverthelbss, this simple model is seen in Figure 2 to be adequate for comparison of data from different reaction vessels. Note that the equation is independent of the hydroxyl radical
0.8
b 0.2
r 8
0.15 O
n
w
Z 3 J
0.1
0
t; \
n 0
: 0.06 0 0
0.2
0.4
0.8
0.8
I0
0.2
0.4
0.6
d
C 0.2
i
0.8
'
0
0.2
0.4
0.6
0.8
I
0
0.2
0.4
0.6
0.8
I
Figure 2. Dimensionless plot of carbon monoxide data for toluene reaction in four reaction vessels: (a) five experiments in our 239-L flask; (b) three experiments in our 2 2 1 flasks; (c) nine experiments In the SAPRC chamber (2)(EC-79, 80, 81, 82, 83, 84, 85, 166, 169); (d) two experiments In the UNC chamber (75) (UNCB 81678, UNCR 81678). The line In the figure was produced from eq 4 by using arbitrary values of a , = 7.5%, a 2b = l o % , and R = 0.9. The sum of a 4- a & Is the overall yield of GO in the reaction, since CO is itself oxidized so slowly, and is the intercept of the curve. The overall curvature is a function of the value of R-for lower values the curve rises more rapidly as it approaches the vertical axis.
,
concentration as well as time and that it can be made dimensionless by dividing by the initial toluene concentration. This dimensionless data analysis procedure shows promise in normalizing results from any type of smog chamber reaction carried out at any light intensity and at any initial concentration, provided only that ozone reactions are of minimal importance. If dilution or product photolysis are important, more complete equations are available (IO). Data from the 239-L flask and the small 22-L flasks are shown in Figure 2, parts a and b, respectively. A line is shown with arbitrary values of 7.5% of
the carbon appearing as CO shortly after hydroxyl attack on toluene and 10% appearing somewhat later from the reaction of intermediate compounds. There is an apparent difference in the overall yields of carbon monoxide in these two reaction vessels, with about 20% overall yield in the small reaction vessel and about 14% overall yield in the larger reaction vessel. Data from the SAPRC chamber, Figure 2c, is virtually identical with data from our 239-L vessel, although the reactions did not run long enough to umambiguously obtain the overall yield. Data from the UNC chamber, Figure 2d, is similar as well. Environ. Sci. Technol., Vol. 17, No. 3, 1983
185
The nominal surface-to-volume ratio is most favorable in the UNC chamber and is worst in the 22-L flasks. However, the UNC Teflon walls and the Teflon coating on the SAPRC chamber probably have a high surfaceroughness factor (porosity), so it is difficult to compare these directly. We do find that our Teflon-plug stopcocks become discolored and difficult to clean. We believe these, or the O-rings, may be the source of background C02 generation in our 239-L reaction vessel by deabsorbing products of previous reactions which are then oxidized to COP We plan to replace these with metal plugs to try and reduce the background. We find that when the 239-L flask is first filled with air-zero, the TGC reading is quite low but that it rises in several hours to 1-2 ppmC, apparently due to desorption. Thus the commonly observed chamber “wall effects” may in fact be due to photolysis of gas-phase material that is desorbed from the walls. Geometrical considerations indicate that in all but the UNC reaction vessel, less than a monolayer can be formed by deposition of reaction products from a single reaction at normal concentrations of reactant hydrocarbon. Thus the overall wall deposition process is probably controlled by adsorption and/or absorption rather than by condensation to the bulk phase. This is confirmed by the insensitivity of the TGC yields to the starting toluene concentration. The similarities between the CO data from the other chambers and our 239-L flask are so great that we believe the overall chemistry of the toluene intermediates that yield CO cannot be greatly different in the two reaction vessels. Comparison of C02data would be even more informative, but unfortunately no COz data are available from the other chambers. It is also significant that a recent toluene mechanism (8),based upon totally gas-phase processes, apparently overpredicts the CO yields by a factor of 3. It is probable that the low CO yields observed experimentally are due to wall deposition of most of the intermediates, rather than to extensive fragmentation as proposed. In conclusion, we strongly recommend that if at all possible, COPyields be obtained routinely in smog chamber experiments. It would of course be preferable to measure TGC yields as well, but this may be far more difficult. Even the COz data, however, would be indicative of the overall process. Finally, the CO data, which are widely available, should be given their due weight in developing overall explicit chemical models, since these data may provide considerable information about the intermediates. The overall goal of the study of pollution chemistry is to develop a greater understanding of both the homogeneous and heterogeneous processes occurring in real atmospheres and not in smog chambers. The bearing of the results presented here on the situation in real atmospheres is not entirely clear. Certainly, the atmospheric surface to volume ratio is not greatly different from that in large smog chambers, due to the presence of suspended aerosols. On the other hand, smog chamber experiments carried out
186
Environ. Sci. Technol., Vol. 17, No. 3, 1983
even in large reaction vessels do not show aromatic hydrocarbons to give suspended aerosol yields as large as the TGC deficits we report here. Thus the actual atmospheric case remains somewhat enigmatic.
Acknowledgments We thank J. J. Huntzicker of the Oregon Graduate Center for the oxidizing furnace used in our lab. Registry No. Toluene, 108-88-3.
Literature Cited (1) Haagen-Smit, A. J. Ind. Eng. Chem. 1952, 44, 1342. (2) Finlayson-Pitts, B. J.; Pitts, J. N., Jr. Adu. Enuiron. Sei. Technol. 1977, 7 , 75. (3) O’Brien, R. J.; Holmes, J. R.; Bokian, A. J. Environ. Sei. Technol. 1975,9, 568. (4) O’Brien, R. J.; Crabtree, J. H.; Holmes, J. R.; Hoggan, M. C.; Bokian, A. J. Adv. Environ. Sci. Technol. 1980,10,367. (5) Grosjean, D.; Friedlander, S. K. Adv. Environ. Sei. Technol. 1980, 10, 435. (6) Altshuler, A. P.; Kopcynski, S. L.; Lonneman, W. A.; Sutterfield, F. D.; Wilson, D. L. Enuiron. Sei. Technol. 1970, 4 , 44. (7) OBrien, R. J.; Green, P. J.; Doty, R. A. In “Chemical and Biological Implications of Nitrogeneous Air Pollutants”; Grosjean, D., Ed.; Ann Arbor Science: Ann Arbor, MI, 1979; p 189. (8) Atkinson, R.; Carter, W. P. L.; Darnall, K. R.; Winer, A. M.; Pitts, J. N., Jr. Int. J. Chem. Kinet. 1980, 12, 779. (9) Hendry, D. G.; Baldwin, A. C.; Barker, J. R.; Golden, D. M. “Computer Modelling of Simulated Photochemical Smog”; EPA-600/3-78-059; U.S. EPA, Research Triangle Park, NC, 1978. (10) OBrien, R. J., Portland State University; unpublished data, 1981. (11) Lehmann, H. W.; Widmer, R. J. Vac. Sei. Technol. 1980, 17, 1177. (12) Holmes, J. R.; O’Brien R. J.; Crabtree, J. H.; Hecht, T. A.; Seinfeld, J. H. Environ. Sei. Technol. 1973, 7 , 519. (13) Huntzicker, J. J.; Johnson, R. L. Conference on Carbonaceous Particles in the Atmosphere, Mar 1978; National Science Foundation and Lawrence Berkeley Laboratory: Berkeley, CA; paper no. 2. (14) O’Brien, R. J.; Johnson, H. E.; Wilkerson, C. In “The Alkylbenzenes“; Peter, F. M., Ed.; National Academy of Sciences: Washington, D.C., 1981; Chapter IV. (15) Whitten, G. Z., Killus, J. P.; Hogo, H. EPA-600/3-80-028a, Systems applications, Inc., San Rafael, CA, 1980.
Received for review December 17, 1981. Revised manuscript received June 28,1982. Accepted November 22,1982. This work was supported in part by the U.S. Environmental Protection Agency, Grant R-804764, Marcia Dodge, project officer. Although the research described in this article has been funded in part by the U.S. Environmental Protection Agency, it has not been subjected to the Agency’s required peer and administrative review and therefore does not necessarily reflect the view of the Agency, and no official endorsement should be inferred.
