Measurement of VOC Reactivities Using a Photochemical Flow Reactor

Transcription

1 Environ. Sci. Technol. 1998, 32, Measurement of VOC Reactivities Using a Photochemical Flow Reactor MICHAEL D. HURLEY,* TAI Y. CHANG, STEVEN M. JAPAR, AND TIMOTHY J. WALLINGTON Ford Research Laboratory, Ford Motor Company, MD-3083, P.O. Box 2053, Dearborn, Michigan A commercial ambient air monitoring instrument, the Airtrak 2000, has been modified for use as a photochemical flow reactor and used to measure the absolute and incremental reactivity of 18 single test VOCs and the incremental reactivity of six multicomponent VOC mixtures. A flow technique is a useful supplement to traditional static chamber experiments. The static chamber technique involves periodic sampling of an irradiated mixture in a photochemical chamber. Under these conditions, the irradiated mixture is always in transition. Using a flow system, a steady-state condition is established within the flow reactor that is representative, in this case, of the early stages of the smog forming process in the atmosphere. The measurement technique also allows changes in the background chamber reactivity to be monitored and taken into account. The incremental reactivity of 13 of the 18 test compounds measured is compared with previously reported results from a static chamber experiment, and the two data sets are generally in good agreement. The additivity of reactivity was tested by measuring the incremental reactivity of six multicomponent mixtures, the components being compounds measured individually in this study. The measured reactivity of a mixture was compared to that calculated from the sum of the measured reactivity of the mixture s individual components. The results show that reactivity is additive for the concentration range studied. Introduction The formation of smog in urban areas continues to be a concern. As a result, there are further regulatory pressures to restrict emissions of NO x (NO + NO 2) and volatile organic compounds (VOCs), the constituents that lead to the formation of photochemical smog. In the past, the regulatory strategy has been to limit the total mass of these species. It has been recognized, however, that not all organic compounds produce comparable amounts of smog. More reactive compounds are capable of producing more smog during their photooxidation in the atmosphere. The concept of reactivity has been included in enacted emission regulations in California (1) to acknowledge the relative amounts of smog produced by VOCs. The reactivity of these compounds is given by a set of reactivity factors established by Carter (2). Carter has used a photochemical mechanism to establish a maximum incremental reactivity (MIR) scale for * Corresponding author fax: (313) ; mhurley3@ ford.com. hundreds of compounds. Incremental reactivity is the additional smog produced when a compound or mixture is introduced to a mix of organic compounds characteristic of an urban environment. These Carter factors are based on the best available kinetic and mechanistic data, but there is uncertainty in some of the data, and some of the factors are best-guess estimates. There is an ongoing effort to check these reactivity factors experimentally and to evaluate the associated chemical mechanisms (3-7). This effort has involved experiments to investigate the mechanisms involved in the atmospheric oxidation of individual hydrocarbons or to directly measure reactivity factors in the presence of a mix of hydrocarbons selected to be representative of urban atmospheres. Customarily, the approach has involved the use of smog reactors in which an atmospheric mixture is prepared and irradiated, with analysis of the contents of the reactor at specified time intervals. The conditions during irradiation in such an experiment are transitory and influenced by, among other things, the changing hydrocarbon to NO x ratio of the mixture over time, the degree of conversion of the hydrocarbon(s) to product, the nature of the hydrocarbon products, and the condition of the reactor walls with regard to the deposition and evolution of reactive species. An alternative approach is the use of a stirred flow reactor in which a sample stream containing a known and constant mix of reactants is continuously fed through the reactor. Eventually a steadystate condition will be achieved in the reactor that is representative of a specific time into the reaction sequence being probed in the static reactor. By selecting the flow characteristics of the reactor, different reaction time regimes can be probed. For example, a focus on the early portion of the reaction sequence makes it possible to probe the reaction mechanism at a time when the system is highly sensitive to radical production, either from the mixture components or involving the reactor walls. We have used this approach to study (a) the reactivities of individual hydrocarbons and hydrocarbon mixtures under conditions representative of the early stages of the conventional static photochemical reactors and (b) to probe the impact of reactor conditioning on the results. To do this, we have adapted a commercially available instrument, an Airtrak 2000 (Mineral Control Instrumentation Ltd., Unley, South Australia), to optimize it for use as a photochemical flow reactor. A second paper will report of the modeling aspects of this work, and the potential impact that this experimental approach could have on the validation of detailed mechanisms for the atmospheric oxidation of individual hydrocarbon species. Experimental Section Definition of Smog Produced and Reactivity. Smog can be most fundamentally defined in terms of the oxidation of NO and the formation of ozone (8). Smog produced (SP) in a reactor system can be defined as SP ) ([O 3 ] - [O 3 ] i ) + ([NO] i - [NO]) (1) where [O 3] i and [NO] i are the initial concentration of ozone and nitric oxide and [O 3] and [NO] are the concentrations after the formation of photochemical smog. Using this definition, smog can be measured even before photolysis yields significant amounts of ozone. During the measurement of reactivity with the photochemical flow reactor, NO is maintained in excess, minimizing the formation of ozone S X(97) CCC: $ American Chemical Society VOL. 32, NO. 13, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY Published on Web 05/15/1998

2 and [O 3] i ) 0. Consequently, SP is given by SP = [NO] i - [NO] ) [NO] (2) Reactivity is defined as the amount of smog produced by a VOC. This will be referred to as the absolute reactivity (AR) since it involves only the photochemical oxidation of the VOC of interest: AR ) [SP] ppm SP [VOC] i ( ppm C ) (3) where [VOC] i is the initial concentration of the VOC. However, when a VOC is released into the atmosphere, it coexists with other VOCs. Therefore, a more realistic measure of reactivity is the impact on smog formation caused by the addition of a VOC to an urban airshed. Incremental reactivity (IR) is defined as the additional smog a test VOC forms when added to a base mixture of VOCs characteristic of an urban airshed, divided by the amount of test VOC added, [VOC]: IR ) [SP(base VOC + VOC)] - [SP(base VOC)] ppm SP VOC ( ppm C ) (4) In this paper, AR and IR are defined using steady-state values of SP measured in flow reactor experiments. IR is defined in the limit where VOC f 0 to remove the dependence on the amount of VOC added and as such has the properties of partial derivatives (2). This definition has the additional advantage that the IR of mixtures can be calculated by summing the reactivity contributions of the components. IRs are found to be sensitive to environmental conditions, most particularly the availability of NO x (2). On the basis of a detailed chemical mechanism and a simple trajectory airshed model, Carter has derived a number of different ozone reactivity scales. One of these is the maximum incremental reactivity (MIR) scale, where reactivity is measured under VOC/NO x ratios where VOC emissions have the greatest effect on smog formation. Instrumental. The section of the Airtrak 2000 used for measuring reactivity consists of two gas streams. A sample stream is prepared external to the instrument. Mass flow controllers are used to mix the VOCs with zero air to reach the desired VOC concentration in a gas stream flowing at 7 L/min. As the gas stream flows past the inlet port of the instrument, the sample pump draws the prepared gas stream into the instrument at 6 L/min. A portion of the sample gas stream (zero air + VOC) is mixed with reagent NO such that [NO] ) 0.3 ppm. The sample gas stream (zero air + VOC + NO) is then divided into two equal streams, C1 and C2. The C1 stream flows through a 10-L dark reactor (functioning as a reference reactor) at 0.85 L/min, where any NO variation in the sample stream is monitored. The C2 stream flows through a 10-L photolytic reactor at 0.85 L/min. The reactors are located in a housing maintained at 40 C, the default control temperature of the Airtrak. Since the flow of gas through each 10-L reactor is 0.85 L/min, the residence time is 12 min. Modification of the instrument involved the addition of a mass flow controller to stabilize the NO concentration in the gas stream and mass flow controllers to balance the gas flow through the dark and photolytic reactors. The photolytic reactor is constructed of Teflon sheet, held in the approximate shape of a sphere with a baseball type metal seam. The photolytic reactor is surrounded with 24 UV fluorescent lamps (F15T8BL). These blacklights produce a Gaussian distribution of wavelengths from 300 to 400 nm. Photochemistry occurring in the photolytic reactor oxidizes the VOC and smog is formed, FIGURE 1. Decay of NO 2 photolysis rate coefficient, k 1, with time. consuming NO. The VOC concentration is kept small so that an excess of NO is always present. Downstream from each reactor is a 1-L Teflon titration reactor. Any ozone formed in the photolytic reactor is consumed by the excess NO in the gas stream by the titration reaction, O 3 + NO f O 2 + NO 2. After leaving the titration reactors, the gas streams are sampled sequentially by a chemiluminescent NO analyzer. The difference in NO concentration between the dark and photolytic reactors is the measure of smog produced by the photooxidation of the VOC in the gas stream. The sensitivity of the chemiluminescent NO analyzer was typically 80 na/ppm NO. The mass flow controllers used to prepare the calibrated VOC sample stream were computer controlled. Automated flow profiles were run overnight with calibration checks during the day resulting in 24-h data acquisition. NO 2 Photolysis Experiments. A critical experimental parameter is the light in the photolytic reactor, which is represented by the NO 2 photolysis rate coefficient, k 1. This was measured as follows. NO 2 was blended with zero air and sampled by the photochemical flow reactor. By observing the formation of NO in the photolytic reactor, the NO 2 photolysis rate can be determined by the photostationary state (9). The first-order NO 2 photolysis rate coefficient, k 1, is given by k 1 ) k 2 [NO][O 3 ] [NO 2 ] and for the flow reactor, we assume that [NO] ) [O 3]: k 1 ) k 2 [NO]2 [NO 2 ] where k 2 is the rate coefficient for reaction NO + O 3 f NO 2 + O 2. The Airtrak maintains the reference and photolytic reactors at 40 C, so the value used for k 2 is 33.7 min -1 (10). The NO 2 photolysis rate coefficient was measured every 2 weeks during the course of these experiments and showed the expected decay as the lights aged. The results of the k 1 measurements are shown in Figure 1, along with an exponential fit to the decay. A measure of the reproducibility of the system can be obtained from the average of the measurements with new lights, i.e., 0.94 ( 0.12 (2σ) min -1. Absolute Reactivity of Propene. Figure 2 shows the results of an experiment to measure the absolute reactivity (5) (6) ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 13, 1998

3 FIGURE 3. Reproducibility of propene absolute reactivity (normalized to k 1) with time. FIGURE 2. Propene absolute reactivity experiment. The top panel shows the NO concentration in the dark and photolytic reactors as the propene concentration is varied at 2-h intervals. The annotation gives the propene concentration (ppm C) during each interval. The middle panel shows the difference in NO concentration as the propene concentration is varied. The annotation gives the NO difference between the dark reactor and the photolytic reactor. The circles are the values used to calculated the steady-state NO difference, [NO] ss. The lower panel show the dependence of [NO] ss on propene concentration. of propene. The initial sample stream flowing through the reactors consisted of zero air and 0.3 ppm reagent NO. Initially there were no VOCs in the sample stream. The NO difference between the dark and photolytic reactors is due to an small instrumental offset. Once the NO concentration in the reactors stabilized, a computer-controlled flow profile systematically changed the propene concentration at 2-h intervals (runs with the aromatic compounds required 3 h for stabilization), a time sufficient for the steady-state condition to be reached in the reactor. With the addition of propene, smog formation occurs in the photolytic reactor and the NO concentration decreases. The top panel of Figure 2 shows the NO concentrations in the dark reactor (top trace) and photolytic reactor (bottom trace). The oscillations in the data are due to temperature cycling of the thermoelectric coolers on the NO analyzer. The middle panel of Figure 2 shows the NO difference between the dark reactor and the photolytic reactor. Annotations in the top and middle panels are the propene concentration (in ppm carbon) and the resulting steady-state NO consumption (in ppm), respectively. The steady-state NO consumption is determined by averaging the difference between the two reactors for the last 15 min of each 2-h step in the profile, as shown by circles and highlighted in the inset. The measured NO corresponds to SP as defined by eq 1. The bottom panel of Figure 2 shows a plot of the steady-state NO consumption versus propene concentration for k 1 ) 0.68 ( 0.10 min -1. At low propene concentrations, the data are approximately linear, and the slope is taken as the absolute reactivity, ( ppm NO/ppm C, calculated by a linear least-squares fit to the data in the linear regime. At higher propene concentrations, the data show a positive deviation from linearity. In general, NO vs [VOC] for alkenes and aromatics show a positive deviation from linearity, while those for alkanes show a negative deviation (7). Modeling results (not shown here) are consistent with these experimental results. Absolute Reactivities of Individual VOCs. The absolute reactivity was measured for 18 VOCs. This portion of the program was done primarily as a check of the performance of the flow reactor and to provide a guide for the incremental reactivity measurements, which are the main focus of this work. Propene was chosen as the calibration VOC for the experiments, and its AR was measured at least once a week over the course of the AR experiments. Not surprisingly, the measured propene AR decreased as the lights aged. Figure 3 shows all the propene AR measurements normalized for the appropriate light intensity taken from the Figure 1 regression line. For k 1 set to 1.0 min -1, the absolute reactivity of propene averaged over the course of the experimental program (n ) 27) is ( (2σ) ppm NO/ppm C. There is an analogous data set for the AR of a nine-component VOC mixture (defined in Table 2) that yields a value (n ) 51; k 1 set to 1.0 min -1 ) of ( (2σ) ppm NO/ppm C. We believe these two results fairly represent the uncertainties in the experimental system. To compare the absolute reactivities of the VOCs tested at different times, the results from the individual runs require adjustment for the decrease in light intensity over time. In Table 1, the ARs measured for the various VOCs (as in Figure 2 for propene) are reported, as is the appropriate value of k 1 taken from Figure 1. Normalizing the values to a k 1 of 1.0 min -1 provides ARs (where uncertainties reflect the 2σ value for propene) that can be compared to the grand average AR for propene. Absolute reactivities of VOCs studied, relative to propene, are also presented Table 1 (where the 2σ values are calculated from the standard propagation or errors). The results are broadly consistent with previous studies. As expected, alkenes are the most reactive compounds, alkanes are the least reactive compounds, and the aromatics are intermediate in reactivity. Incremental Reactivity of Propene. The base VOC gas mixture composition chosen in the present work is an urban VOC surrogate that consists of nine components (Table 2). The composition of this mixture was selected based on the work of Jeffries and co-workers (11), who chose eight hydrocarbons and two aldehydes as representative of urban ambient air. Since aldehydes are not stable in cylinders, VOL. 32, NO. 13, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

4 TABLE 1. Absolute Reactivities of Test Compounds AR(VOC) (ppm NO/ppm C) k a 1 (min -1 ) AR(VOC) b (ppm NO/ppm C) AR(VOC)/AR(propene) propene (2.6 ( 0.65) E ethene (1.1 ( 0.27) E ( 0.10 (1.5 ( 0.38) E ( butene (1.4 ( 0.35) E ( 0.10 (2.1 ( 0.52) E ( 0.28 isobutene (2.0 ( 0.50) E ( 0.13 (2.4 ( 0.60) E ( 0.32 trans-2-butene (3.3 ( 0.83) E ( 0.10 (4.8 ( 1.2) E ( pentene (6.5 ( 1.6) E ( 0.09 (1.0 ( 0.26) E ( 0.14 trans-2-pentene (5.0 ( 1.3) E ( 0.12 (6.6 ( 1.6) E ( 0.88 n-butane (1.2 ( 0.31) E ( 0.10 (1.8 ( 0.45) E ( ,3-dimethylbutane (1.3 ( 0.33) E ( 0.09 (2.1 ( 0.52) E ( ,3-dimethylpentane (9.2 ( 2.3) E ( 0.10 (1.4 ( 0.36) E ( n-hexane (9.3 ( 2.3) E ( 0.06 (2.3 ( 0.58) E ( n-octane (7.0 ( 1.7) E ( 0.06 (1.9 ( 0.47) E ( toluene (3.3 ( 0.83) E ( 0.10 (5.0 ( 1.2) E ( m-xylene (4.3 ( 1.1) E ( 0.07 (9.2 ( 2.3) E ( 0.12 p-xylene (2.8 ( 0.69) E ( 0.07 (5.9 ( 1.5) E ( ,2,3-trimethylbenzene (7.3 ( 1.8) E ( 0.06 (1.7 ( 0.43) E ( ,3,5-trimethylbenzene (1.4 ( 0.34) E ( 0.07 (3.1 ( 0.76) E ( 0.41 carbon monoxide (1.9 ( 0.48) E ( 0.10 (2.8 ( 0.71) E ( nine-component mix (5.8 ( 1.4) E ( 0.07 (1.3 ( 0.32) E ( 0.17 a k 1 ) exp ( days) b k 1 ) 1.0 min -1. TABLE 2. Composition of Nine-Component Mixture, Mixture Certified, Analytical Accuracy (2% compound [VOC] (ppm) [VOC] (ppm C) carbon fraction base case mix (ppm C) n-butane ,3-dimethylpentane ethene isobutene propene toluene trans-2-butene ,3,5-trimethylbenzene m-xylene total isobutene was added to the mixture as an aldehyde surrogate, consistent with the CB-4 mechanism (12). Figure 4 shows the result of an experiment to measure the incremental reactivity of propene. For all experiments, the flow system was initially flushed with zero air, followed by the addition of 0.3 ppm reagent NO, and then the addition of 2.1 ppm C of the base case VOC mixture. This provided a base VOC/NO x of 7, which is in the MIR region for conventional static chamber experiments. Once the NO consumption due to the base VOC mixture stabilized, a computer-controlled flow profile systematically changed the added propene concentration at 2-h intervals (3 h for the aromatics), a time sufficient for establishment of a photochemical steady state in the flow reactor. The consumption of NO due to the base VOC mixture alone was checked in the middle of this flow profile and again at the end of the experiment. The top panel of Figure 4 shows the NO concentration in the dark reactor (top trace) and photolytic reactor (bottom trace). The base VOC mixture was present for all propene concentrations. The middle panel of Figure 4 shows the NO difference between the dark reactor and the photolytic reactor. The annotation on the top panel gives the propene concentration in the sample stream, and the annotation on the middle panel gives the steady-state NO consumption for each propene concentration. The NO value is determined by averaging the difference for the last 15 min of each step in the profile as shown by circles in the figure. The results of this experiment are shown in the bottom panel of Figure 4, where NO is plotted against the added FIGURE 4. Propene incremental reactivity experiment. The top panel shows the NO concentration in the dark and photolytic reactors as the propene concentration is varied at 2-h intervals. The composition of the nine-component mix is constant at 2.1 ppm C. The annotation gives the propene concentration during each interval. The middle panel shows the difference in NO concentration as the propene concentration is varied. The annotation gives the NO difference between the dark reactor and the photolytic reactor. The circles are the values used to calculated the steady-state NO difference, [NO] ss. The lower panel show the dependence of [NO] ss on propene concentration. propene concentration. The slope is the incremental reactivity of propene, which for this particular run was ( (2σ) ppm NO/ppm C. The data included in the fit are the shaded symbols. Only data at the lowest VOC concentrations, where the data are linear, are used to determine the incremental reactivity. The y intercept represents the NO consumption due to the nine-component base VOC mixture ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 13, 1998

5 TABLE 3. Incremental Reactivities of Test Compounds IR(VOC) (ppm NO/ppm C) initial NO(9-C) (ppm NO) initial NO(9-C) (ppm NO) IR(VOC)/ IR(propene) a propene 1.0 ethene (2.6 ( 0.65) E ( butene (2.9 ( 0.74) E ( 0.23 isobutene (4.0 ( 1.0) E ( 0.38 trans-2-butene (1.7 ( 0.43) E ( pentene (1.1 ( 0.26) E ( 0.11 trans-2-pentene (9.5 ( 2.4) E ( 1.14 n-butane (1.7 ( 0.43) E ( ,3-dimethylbutane (5.5 ( 1.4) E ( ,3-dimethylpentane (2.1 ( 0.54) E ( n-hexane (3.1 ( 0.76) E ( n-octane (-2.9 ( 0.72) E ( toluene (5.8 ( 1.4) E ( m-xylene (2.9 ( 0.72) E ( 0.22 p-xylene (8.4 ( 2.1) E ( ,2,3-trimethylbenzene (2.5 ( 0.63) E ( ,3,5-trimethylbenzene (5.2 ( 1.3) E ( 0.66 carbon monoxide (4.2 ( 1.1) E ( three-component mix 1 (1.5 ( 0.38) E ( three-component mix 2 (6.4 ( 1.6) E ( three-component mix 3 (2.1 ( 0.53) E ( six-component mix 1 (9.8 ( 2.5) E ( six-component mix 2 (9.4 ( 2.3) E ( nine-component mix (1.3 ( 0.31) E ( 0.18 a IR(propene) ) 9.50E k 1AR(9-C). The incremental reactivity of propene was measured at least twice a week as a control and to check data reproducibility. There is a general trend toward lower measured reactivities with time, consistent with the measured reactivity being proportional to the decreasing NO 2 photolysis rate coefficient. However, scatter in the data cannot be explained by changes in the light intensity alone. The evidence suggests that this variation in propene reactivity is related to changes in the reactivity of the photolytic reactor that occur over the course of an experiment. In Figure 4, for example, NO due to the base VOC mixture increased from to to ppm during the incremental reactivity measurement of propene, suggesting that the photooxidation of propene conditioned the reactor in a way that increased its background reactivity. Analogous results were found for many of the VOCs tested (Table 3), although for some of the VOCs the NO (base VOC) was found to decrease. (More details on this subject are available in ref 7.) We believe that these changes directly demonstrate the effect of a given VOC on chamber reactivity and strongly suggest that the measured incremental reactivity depends on the chemical history of the reactor. An empirical relationship between propene incremental reactivity, the NO 2 photolysis rate coefficient, and chamber reactivity is shown in Figure 5. For each propene IR determination, the date of measurement is used to calculate k 1 (from Figure 1), and the NO value for the base VOC mixture is used to calculate the base VOC absolute reactivity ( NO/2.1 ppm C). The NO value used for the base VOC is one measured midway through the incremental reactivity experiment, i.e., ppm NO for the experiment shown in Figure 4. The fit to the data is quite good, with the correlation coefficient, r 2, equal to Incremental Reactivities of Individual VOCs and VOC Mixtures. The incremental reactivities of 18 compounds have been measured and scaled to propene to give a relative incremental reactivity. The results are presented in Table 3. The scaling is done in order to account for light decay and variable chamber effects and to allow comparison with static FIGURE 5. Dependence of propene incremental reactivity on k 1 and AR(9-C). chamber reactivity results. The procedure involves using the test date to calculate the photolysis rate coefficient, k 1 (Figure 1), and setting the baseline reactivity, i.e., AR (base VOC), for the nine-component equal to the value determined in the middle of the series of runs, as was done in Figure 4 for propene. Finally, k 1 and AR(base VOC) are used to determine the corresponding incremental reactivity for propene from the regression equation in Figure 5. Included in Table 3 are the initial and final absolute reactivities for the nine-component mix, normalized for the variation in light intensity. The incremental reactivity measurements of all compounds showed a positive deviation from linearity at high VOC concentrations. This could result from the photochemical mechanism or from an increase in reactivity of the photochemical chamber. VOL. 32, NO. 13, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

6 TABLE 4. Incremental Reactivity Results for Multicomponent Mixtures carbon fraction IR(mix)/IR(propene) (calcd) IR(mix)/IR(propene) (expt) IR(calculated)/ RIR(experimental) three-component mix ( ( ( ,3-dimethylbutane ( propene ( m-xylene ( three-component mix ( ( ( 0.46 n-butane ( ethene ( toluene ( three-component mix ( ( ( ,3-dimethylpentane ( trans-2-butene ( ,3,5-trimethylbenzene ( six-component mix ( ( ( 0.52 ethene ( propene ( n-butane ( ,3-dimethylpentane ( m-xylene ( ,3,5-trimethylbenzene ( six-component mix ( ( ( 0.61 propene ( trans-2-butene ( n-butane ( ,3-dimethylpentane ( toluene ( m-xylene ( nine-component mix 0.49 ( ( ( 0.45 n-butane ( ,3-dimethylpentane ( ethylene ( isobutylene ( propylene ( toluene ( trans-2-butene ( ,3,5-trimethylbenzene ( m-xylene ( The reactivity of a mixture of VOCs can be calculated from the reactivity of the individual components. Since reactivities are defined in the limit of VOC approaching zero, reactivity is in the linear regime. In this regime, reactivity is additive and the reactivity of a mixture is equal to the sum of the weighted reactivities of the individual components: n IR ) IR i F i (7) i)1 where IR is the incremental reactivity of the test VOC mixture, IR i is incremental reactivity of VOC i and F i is the carbon fraction of VOC i in the mixture. The property of additivity was checked by measuring the incremental reactivity of six multicomponent mixtures. Three-component and six-component mixtures were ordered, with the components being compounds that were measured individually in this study. All mixtures contained at least one alkane, one alkene, and one aromatic. The incremental reactivity of the nine-component mixture was also measured. The results are shown in Table 4, which presents the composition and the measured relative incremental reactivities of the mixtures. Also shown are the relative incremental reactivities of the mixtures calculated from the individual reactivities measured in this study (Table 3). The ratio of the calculated results to those determined experimentally range from 0.9 to 1.2. The results clearly demonstrate that, for the series of three-, six-, and ninecomponent VOC mixtures studied, the incremental reactivities of the individual components can be used to predict the incremental reactivities of the various mixtures. Reproducibility and Estimation of Errors. There are several potential sources of errors in the flow reactor measurements of reactivity. The uncertainty in the measurements of the individual points during the last 15 min of a concentration profile is nearly constant at ( ppm NO ((0.5%). This uncertainty likely reflects the inherent signal-to-noise ratio of the NO analyzer. There are uncertainties in the gas composition. The contents of all gas cylinders were certified, with an uncertainty of (2%. The most subjective source of error is the determination of the linear regime of reactivity. In general, the slopes in the low [VOC], linear reactivity regimes are quite well defined. Considering all these sources of errors, the overall 2σ error in the measurement of individual reactivity values is estimated to be about 10%. There is additional uncertainty involved in the calculation of k 1 (Figure 1) and the reactivity of propene (Figures 3 and 5). Using standard techniques for the propagation of error, the overall 2σ uncertainty is (35% for both the relative absolute reactivity and the relative incremental reactivity. Discussion Our results can be used to compare the absolute and incremental reactivities of VOCs measured under similar conditions. Comparing the values in Tables 1 and 3, there is a correlation between the absolute and incremental reactivities for the alkenes and aromatics that does not exist for the alkanes. There have been several recent static chamber studies of the incremental reactivity of many of the compounds reported in this study. These studies are all static chamber experiments where the reactants are mixed in a Teflon ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 13, 1998

7 TABLE 5. Table of Experimental Incremental Reactivities Relative to Propene this work Carter et al. (1995) flow reactor static chamber IR(VOC)/ IR(VOC)/ IR(propene) ETC IR(propene) propene ethene 0.72 ( ( 0.22 isobutene 1.09 ( ( 0.19 trans-2-butene 4.12 ( ( 0.31 n-butane ( ( n-hexane ( ( n-octane ( ( toluene 0.13 ( ( 0.17 m-xylene 0.63 ( ( 0.17 p-xylene 0.25 ( ( ,2,3-trimethylbenzene 0.74 ( ( ,3,5-trimethylbenzene 1.9 ( ( 0.26 carbon monoxide ( ( chamber, irradiated with a light source for several hours, and analyzed for smog at hourly intervals. Our results are most easily compared to those of Carter and co-workers (4, 13), in terms of light sources and VOC/NO x ratios. They measured the incremental reactivity of a number of compounds using 4 ppm C of a three-component base VOC mixture and 0.5 ppm NO x irradiated with blacklights. The temperature of the chamber with the lights on was 28 ( 3 C. The three-component base mix consisted of 35% ethene, 50% n-hexane, and 15% m-xylene. Carter and co-workers injected reactants into a Teflon chamber and irradiated the mixture for 6 h while analyzing the contents at hourly intervals. They did multiple runs for each VOC, but the run chosen for comparison of the individual VOCs is the run with smallest added VOC, consistent with definition of reactivity in the limit of VOC concentration approaching zero. Table 5 compares our results with those of Carter and coworkers for the 6-h irradiation. Comparable relative incremental reactivity results are available for 12 compounds. The results for eight of the 12 VOCs agree to within the combined reported uncertainties. The flow reactor reactivity values for trans-2-butene, n-butane, and carbon monoxide are about twice the static chamber values, while the results for n-hexane are comparable in magnitude but opposite in sign. It must be remembered that the two experiments are reporting results from very different experiments. The flow reactor generates data applicable to a very specific portion of the reaction sequence that controls the atmospheric oxidation of a VOC. Modeling of our flow reactor (to be discussed in detail in a subsequent paper) indicates that only about 5% of the VOC (in the case of propene) has been converted to product in the steady-state condition. This corresponds to a time very early in the reaction sequence that occurs in a static-chamber, multi-hour irradiation of the kind reported by Carter and co-workers. This is important because the early portion of the reaction sequence is strongly influenced by radical sources in the system, formed either in the oxidation of the VOC or from materials resident on the reactor walls. On the other hand, the end-of-run analyses from Carter and co-workers present the integrated system response during the large-scale consumption of the VOC (generally well in excess of 50%) and some of its products and are generally fairly insensitive to the specific reactor conditions early in the experimental sequence. Thus, appropriately designed flow reactor experiments should be able to provide new relative information concerning the development of VOC oxidation mechanisms: (a) they can provide unique data useful in the verification of a specific mechanism; (b) they can be used to probe the condition of the reactor and its impact on the measured oxidation process. Furthermore, a flow reactor might also be used to mimic the emission and dilution of smog components. In conclusion, a word of caution. These results cannot be used to directly predict atmospheric reactivities, since the light source used and the high VOC and NO concentrations used are not characteristic of the atmosphere. Rather, these results should be viewed as a demonstration of the utility of a complementary approach to the determination of atmospheric oxidation mechanisms. Acknowledgments We are grateful to the Environmental Research Consortium, consisting of Chrysler Corporation, Ford Motor Company, General Motors Corporation, and Navistar International Transportation Corporation for sponsoring this work with funding from the Coordinating Research Council and the National Renewable Energy Laboratory. Literature Cited (1) CARB. Proposed reactivity adjustment factors for transitional low-emissions vehicles: technical support document; California Air Resources Board: Sacremento, CA, (2) Carter, W. P. L. J. Air Waste Manage. Assoc. 1994, 44, 881. (3) Carter, W. P. L.; Atkinson, R. Environ. Sci. Technol. 1987, 21, 670. (4) Carter, W. P. L.; Pierce, J. A.; Luo, D.; Malkina, I. L. Atmos. Environ. 1995, 29, (5) Carter, W. P. L.; Luo, D.; Malkina, I. L.; Pierce, J. A. Environmental chamber studies of atmospheric reactivities of volatile organic compounds. Effects of varying ROG surrogate and NO x; Final report to Coordinating Research Council, Project ME-9; California Air Resources Board, Contract A and South Coast Air Quality Management District, Contract C91323; (6) Kelly, N. A.; Wang P. Measurement of the atmospheric reactivity of emissions from gasoline and alternative-fueled vehicles: assessment of available methodologies, Part 1sindoor smog chamber study of reactivity; First year final report for CRC Contract AQ , June 30, 1995; Available from Coordinating Research Council: Atlanta, GA. (7) Hurley, M. D.; Japar, S. M.; Chang T.; Wallington, T. J. Measurement of the atmospheric reactivity of emissions from gasoline and alternative-fueled vehicles: assessment of available methodologies, Part 2sAssessment of Airtrak as a Reactivity Analyzer; Second year final report for CRC Contract AQ , June 13, 1996; Available from Coordinating Research Council: Atlanta, GA. (8) Johnson, G. M. A simple model for predicting the ozone concentration of ambient air. Proceedings, Eigth International Clean Air Conference; Clean Air Society of Australia and New Zealand: Auckland, 1984; Vol. 2, p 715. (9) Wu, C. H.; Niki, H. Environ. Sci. Technol. 1975, 9, 46. (10) DeMore, W. B.; Golden, D. M.; Hampson, R. F.; Howard, C. J.; Kurylo, M. J.; Molina, M. J.; Ravishankara, A. R.; Sander. S. P. Chemical kinetics and photochemical data for use in stratospheric modeling; JPL Publication 87; National Technical Information Service: Springfield, VA, (11) Jeffries, H. E.; Sexton K. G.; Arnold J. R.; Kale T. L. Validation testing of new mechanisms with outdoor chamber data. Volume 2: Analysis of VOC data for the CB4 and CAL phtochemical mechanisms; EPA-600/ b; U.S. EPA: Research Triangle Park, NC, (12) Gery, M. W.; Whitten, G.Z.; Killus, J. P. Development and testing of the CB-IV for urban and regional modeling; EPA-600/ ; U.S. EPA: Research Triangle Park, NC, (13) Carter, W. P. L.; Pierce, J. A.; Luo, D.; Malkina I. L. Environmental chamber study of maximum incremental reactivities of volatile organic compounds; Final report to Coordinating Research Council, Project ME-9 and California Air Resources Board, Contract A ; Received for review September 22, Revised manuscript received March 11, Accepted March 31, ES970844R VOL. 32, NO. 13, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY