Environ. Sci. Technol. 2000, 34, 2851-2858

Multicomponent Remote Sensing of Vehicle Exhaust by Dispersive Absorption Spectroscopy. 1. Effect of Fuel Type and Catalyst Performance MARC M. BAUM,* EILEEN S. KIYOMIYA, SASI KUMAR, AND ANASTASIOS M. LAPPAS Department of Chemistry, Oak Crest Institute of Science, 13300 Brooks Drive, Suite B, Baldwin Park, California 91706 HARRY C. LORD, III Air Instruments & Measurements, Inc., 13300 Brooks Drive, Suite A, Baldwin Park, California 91706

A remote sensor incorporating UV and IR spectrometers in conjunction with an innovative optical design is described. The instrument was used to noninvasively measure over 20 pollutants in the exhaust of 19 in-use vehicles powered by a range of fuelssreformulated Phase II gasoline, diesel, compressed natural gas, and methanol blended with 15% gasoline. CO2, CO, aldehydes, and aliphatic and speciated aromatic hydrocarbons were identified along with NOx, determined as the sum of NO, NO2; N2O and HONO were also measured, although their levels were typically below the instrument’s detection limit. NH3 levels in vehicle exhaust are reported for the first time on a car-by-car basis. The exhaust from gasoline- and methanol-powered cars was found to contain elevated levels of NH3, in some cases over 1000 ppm, despite near stoichiometric airto-fuel ratios, and were often significantly higher than corresponding NO levels. Catalyst efficiency is discussed as a function of NH3 and NO concentrations in the exhaust of vehicles operating “cold” and “hot”. In some of the tested vehicles, the three-way catalysts showed high reduction activity but poor selectivity resulting in the formation of NH3 and possibly other nitrogen-containing products other than N2. These observations could have significant implications on the formation of ammonium nitrate aerosol and on the acid-neutralizing capacity of urban air masses.

Introduction Emissions from mobile sources are well-known to play a central role in urban air pollution (photochemical smog formation, violation of carbon monoxide (CO) and ozone (O3) standards, and aerosol formation). In 1994, the U.S. Environmental Protection Agency (U.S. EPA) estimated that, for the previous year, U.S. on-road vehicles contributed 62%, 32%, and 26% of all CO, nitrogen oxide (NOx), and volatile organic compound emissions, respectively (1). Remote sensing of exhaust from light duty motor vehicles (LDMVs) * Corresponding author phone: (626)338-3333; fax: (626)338-7815; e-mail: [email protected]. 10.1021/es991351k CCC: $19.00 Published on Web 06/06/2000

 2000 American Chemical Society

has provided a wealth of useful information with respect to CO (2-5) and total hydrocarbon (THC) emissions (6-9). Data collected from these investigations indicate that approximately half of CO and THC emissions were generated by less than 10% of vehicles. Over the past 3 years NO emissions have been reported to follow similar trends (10-12). Moreover, remote sensing data suggest that fleet dynamometer testing significantly underestimates tailpipe emissions and contributes to errors in model predictions (e.g., U.S. EPA’s MOBILE4) (3, 13). A knowledge of the chemical composition of the exhaust plume emitted by on-road vehicles on a carby-car basis therefore is essential when developing effective pollution abatement strategies. Prior to this report, most remote sensing studies have relied on nondispersive infrared (NDIR) (3, 7) and nondispersive ultraviolet (NDUV) (10) spectroscopy. While these analytical techniques provide excellent data for CO and CO2, accurate THC data have been difficult to acquire due to modest sensitivity, 500 parts per million (ppm, 1 ppm ) 1 part in 106 by volume or moles) detection limit (3σ) and water interference (7). The sensitivity to NO is even poorer (300 ppm precision, 1σ), limiting the instrument to the identification of gross polluters or fleet evaluations (10). More recently, tunable infrared diode laser absorption spectrometers (TIDLAS) have been utilized in a remote sensing configuration to measure NO in vehicle exhaust with greater sensitivity and selectivity (11, 12); nitrous oxide (N2O) has also been measured by a similar system (14). Unfortunately, the deployment of these TIDLAS instruments is complicated by their requirement for cryogenic cooling, which is necessary to allow operation in the mid-IR (where most of the strong bands are located). They also need skilled operators and can be prohibitively expensive for multicomponent applications. Fourier transform infrared (FTIR) spectrometers have become very popular for open path monitoring but have found limited application in the remote sensing of auto exhaust. This is partially due to low signal-to-noise ratios resulting from the short averaging times (0.5-1.0 s) (15). Additionally, such systems are often too delicate for field use; sturdier and faster systems are available but can be expensive. A rugged, low-cost alternative to existing remote sensors is needed to measure criteria pollutants (CO, THC, NO) as well as CO2 with equal or increased precision. The instrument should also be capable of measuring other compounds of importance to tropospheric photochemistry. For instance, formaldehyde (HCHO) and acetaldehyde (CH3CHO) are key to the photolytic generation of hydroperoxyl and acylperoxy radicals; nitrous acid (HONO) is an important source of hydroxyl radicals; nitrogen dioxide (NO2) affords ozone upon photolysis and reacts with hydroxyl radicals to yield nitric acid; aromatic hydrocarbons (e.g., benzene, toluene, xylene) are important reaction sinks for hydroxyl radicals, often affording secondary organic aerosols (16, 17). Ammonia (NH3) is known to be emitted by vehicles equipped with three-way catalysts (TWCs) operating under fuel-rich conditions (1821). NH3 emissions play a key role in the production of fine particulate matter. Finally, a useful remote sensor should store sufficient spectral information in the “snapshot” of the exhaust plume to enable quantification of “unknown species” at a later date. The research described here utilizes dispersive ultraviolet and infrared absorption spectroscopy, in conjunction with an innovative optical design, to noninvasively probe exhaust plumes emitted by operational vehicles. The concentrations of over 20 compounds of tropospheric significance were measured for a fleet of 19 vehicles powered by a range of VOL. 34, NO. 13, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2851

FIGURE 1. Picture of the remote sensing instrument measuring the exhaust emissions from a CNG-powered vehicle. fuels. Many of these noncriteria pollutants have never been measured by remote sensing in vehicle exhaust on a carby-car basis. The principal objectives of these measurements were to test this novel remote sensor technology and to determine the level of detectability of noncriteria pollutants in vehicle exhaust. No attempts were made in this phase of our investigations to sample a statistically significant number of vehicles or to quantitatively estimate emission rates of pollutants.

SCHEME 1. Schematic of the RSIa

Experimental Details Remote Sensor. The remote sensing instrument (RSI) consists of a pair of broad-band light sources optically interfaced to a multipass system, adapted from the well-known White cell design (22), that spans the width of a single traffic lane. The two light beams are projected across the roadway and, after eight optical passes, are focused onto the entrance slits of the corresponding spectrometers. An Ocean Optics SD2000 2048 element CCD array spectrometer (10 ms duty cycle, spectral resolution 0.88 nm fwhm) is used to analyze UV-vis radiation, and a Monolight spectrum analysis system supplied by Macam Photometrics, Ltd. (7 ms spectral acquisition speed, 80 ms duty cycle, 0.89 mm slit width, spectral resolution 27 nm, 30 cm-1, fwhm) is used for analysis of IR radiation. Data are collected in the 190-520 nm and 15005000 nm (6660-2000 cm-1) spectral windows, respectively. A single PC controls both spectrometers, reads and manages the digital data, and stores the spectra in binary format for post analysis. Real-time analysis for a limited number of compounds is possible. A schematic of the instrument is shown in Scheme 1, and a picture of the device is given in Figure 1. Data Analysis and Calibration. The stored data were retrieved, processed to yield absorbance spectra, and analyzed with a least-squares fitting algorithm based on the singular value decomposition theorem (23). This signal processing approach greatly enhanced instrument precision. Finally, instrument response to a given analyte was related to optical density by using calibration curves generated in the laboratory from authentic gas samples. Synthetic mixtures of known concentration were prepared by blending certified gases using mass flow controllers. The concentrations of the calibration mixtures (usually 5 per analyte) were chosen to 2852

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 13, 2000

a L1, L2, L3, ZnSe lenses; MRS, multiple reflection system; R1, glowbar; R2, deuterium source; S1, IR spectrometer; S2, UV-vis spectrometer; SW1, SW2, Schwarzshield telescopes (dotted line, IR beam; solid line, UV beam).

FIGURE 2. Laboratory calibration curve for CO. span the measurement range at equal intervals. Calibration gases were passed through a gas cell (optical path length, 0.5 m), and the instrument response was recorded. A plot of the known concentration of the calibration sample (actual concentration) versus the concentration calculated by the analyzer based on the pattern recognition fit (observed concentration) yielded a polynomial calibration curve (see Figure 2); first- and second-order curves provided good fits for all the gases of interest. The calibration curves corrected instrument readings for deviations from ideal Beer-Lambert

TABLE 1. Band Locations (BLs) and Detection Limits (DLs) for Selected Compounds Measured by the RSI compound BL (nm) DL (ppm-m)a compound BL (nm) DL (ppm-m) NO NO2 HONO NH3 N2H4 O3 SO2 C6H6 C6H5Me C6H4Me2 c

226 435 360 208 245 254 297 250 265 270

3.0 25 20 0.5 5 0.5 5 2 2 2b

C6H5OH C6H5CHO HCHO MeCHO N2O CO CO2 THCc CH4 C2H2

275 284 325 285 4550 4650 4395 3390 3330 3040

0.5 0.3 100 25 20 250 1000 7 150d 15

a Column density of observed air column. b Based on the p-isomer. Measured as C3H8. d Not always conclusive.

behavior, which are common when the measured absorption feature is much narrower than the spectral resolution of the detector (24, 25). Field calibrations, using an in-line cell and span cylinder gases of CO, CO2, THC (as propane), NO, and NH3, were performed twice daily. These procedures are typical for vehicle remote sensing studies. However, the accuracy of the spectra measured in the RSI described here is tied to known invariant physical properties of the analyte. Thus, periodic calibrations are not required to maintain measurement precision and consistency as with nondispersive instruments. This is another major advantage of the dispersive technique. When certified cylinder gases were not available (e.g., HONO), a spectrum of an authentic sample was collected and used as a reference. Concentration was determined from published absorption cross-sections. Repeatability and Detection Limits. The repeatability of measurements made with the laboratory calibration system was typically within the noise level of the spectrometer, significantly better than 2% of full-scale. This was even true for NH3, known for its adsorptive properties. Field calibrations also exhibited high repeatability. On-road measurements were subject to larger repeatability errors. However, Bishop et al. have shown that the largest source of measurement variability in the field can be attributed to the vehicle, not the test procedure (26). Jime´nez et al. recognized that NO emissions for LDVs followed a γ distribution peaking around zero NO emission (12). Field testing of the TIDLAS remote sensor on a fleet of in-use vehicles yielded a set of negative concentration measurements that could be convoluted to produce a robust estimate of instrument detection limit. Such an approach could not be followed here: our analyzer does not produce negative readings since spectra are baselined prior to processing. Additionally, the adopted pattern recognition approach only scans absorbance spectra for positive signals, as negative signals generally have no physical meaning. In our investigation, “blind” computer fits were carried out on spectra of exhaust plumes collected for all 19 vehicles under accelerating and decelerating driving conditions, with the vehicles running “cold” and “hot”. Spectra yielding low readings (i.e., near zero) were subsequently analyzed manually to confirm whether the pollutants of interest were present in the mixture. This exercise had to be performed in order to validate the absence of these analytes in the matrix and, hence, determine the lowest concentration (σ) the remote sensor could reliably resolve from the background noise. Detection limits (3σ) are calculated from these quantities. Absorption band locations and detection limits for some of the measured compounds are given in Table 1. Note that these are values are conservative (especially for those compounds with well-defined rotational-vibrational structure) and represent measured column densities, not actual

TABLE 2. Characteristics of Vehicles Tested label

make

model

year

mileage

fuel type

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

Oldsmobile Mazda Chevrolet Isuzu Honda Honda Ford Nissan Nissan Honda Honda BMW Jeep Chevrolet Internationala CNG1 CNG2 CNG3 Ford

Cutlass Sierra 929 S10 Rodeo Accord LX Accord LX Explorer 200SX Pick-up Civic Del Sol 325i Wrangler Cavalier V8 Turbo Diesel

1984 1988 1988 1991 1994 1993 1996 1986 1985 1993 1995 1989 1998 1997 1996 1995 1995 1995 1996

221 138 147 433 128 806 87 309 50 759 16 563 55 560 157 828 175 069 87 070 55 038 114 965 12 065 48 860 22 244 117 508 284 11 508 37 500

gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline gasoline diesel CNGb CNGc CNGc M85d

Taurus

a

Medium duty motor vehicle. b Dual fuel. c Dedicated, factory CNG vehicle. d Dedicated, factory M85 vehicle.

levels in undiluted exhaust. Detection limits of measurements corrected for dilution will vary depending on the overlap of the optical envelope with the exhaust plume (see below). Field Measurements. The RSI was set up with an 8.8 m spacing between the field and objective mirrors of the multiple reflection system affording a probed air column of approximately 70 m. Beam centers were located at 30.5 cm (12 in.) above the road surface. Vehicles at rest were accelerated over a 20 m roadway through the optical beams. Speeds of ca. 20-30 km/h at ca. 3000 rpm were typically attained. After the vehicle emerged from the optical probe, 1 s of measurement data were collected and stored along with 160 ms of reference data, which were collected just prior to the vehicle entering the beams. Each vehicle was measured 3 successive times when running cold and 3 additional times when running hot. Idle measurements were also performed but will not be discussed here. The vehicles employed in this study are listed in Table 2 and make up a diverse sample array in terms of make, model, and year. Gasoline-powered vehicles used Federal Phase II reformulated gasoline (Summer fuel). One medium duty vehicle was powered by diesel fuel, three LDMVs were run on compressed natural gas (CNG), and one LDMV was powered by a methanol blend using 15% gasoline (M85). Safety. All toxic cylinder gases were handled in fume hoods using standard safety protocols and were destroyed with appropriate chemical scrubbers prior to release to the environment. Remote sensing of vehicle exhaust did not expose the instrument operator to significant levels of emissions.

Results and Discussion CO2 Levels and Plume Dynamics. A column of unpolluted air 70 m deep will show very strong CO2 absorption bands in the mid-IR due to the presence of ambient levels of CO2 (ca. 360 ppm). At these column densities, the fundamental CdO asymmetric stretch at 4260 nm (2350 cm-1) will be fully saturated (i.e., opaque to IR radiation) as will the combination-overtone band between 2670 and 2820 nm (3750 and 3550 cm-1). Only the 13CO2 CdO asymmetric stretch between 4350 and 4460 nm (2300 and 2240 cm-1) was usable for CO2 measurement within the spectral range accessible to the IR spectrometer. Kert described (15) a similar challenge with a remote sensor using an FTIR spectrometer as a detector to measure CO, CO2, and THC emitted by in-use vehicles. Due to the low sensitivity to THC, the author evaluated the VOL. 34, NO. 13, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2853

FIGURE 4. CO Levels in undiluted exhaust of hot vehicles. Three successive measurements per vehicle were made.

FIGURE 3. Absorbance spectra of 13CO2 collected in the field, illustrating how the 13CdO asymmetric stretch can be used to measure CO2 levels in vehicle exhaust. possibility of using multiple optical passes but concluded that the 13CO2 band could not be used due to a noisy baseline and interference from CO. These problems were not encountered in our investigation, and CO2 was conveniently measured using the 13CO2 band as illustrated by Figure 3. Rapid dilution of the turbulent exhaust plume complicates determination of undiluted tailpipe emissions; strong overlap between the optical envelope and the plume is desirable. Measuring CO2 levels is key in accounting for dilution and can be used to normalize other gases to levels in undiluted exhaust (3, 27). Assuming stoichiometric combustion of gasoline, the optical density of CO2 in undiluted exhaust is ca. 15% in a 4.5 cm diameter tailpipe. Using a measurement system configured for eight optical passes through the exhaust sample, an effective column density of ca. 15% in 36 cm or 54000 ppm-m can be assumed. The extent of plume overlap can then be calculated from the measured CO2 column density. To improve the precision of the dilution estimates, nonstoichiometric combustion must be factored into the calculation. The actual CO2 concentration in undiluted vehicle exhaust, [CO2]u, can be calculated from eq 1 (27)

[CO2]u ) 42/(2.79 + 2Q + 0.42Q′)

(1)

where Q ) [CO]/[CO2], Q′ ) [THC]/[CO2], and Q′′ ) [NO]/ [CO2]. The concentrations of all other analytes can now be corrected for dilution by multiplying the appropriate Q-ratio by [CO2]u. Two vehicles, the diesel van and CNG1, had exceptionally wide tailpipes (9.4 and 6.7 cm, respectively), and plume overlap was calculated separately. On average, a dilution factor of 5.2 was observed. This value is lower than the reported mean for the University of Denver “FEAT” (ca. 10) (27) but higher than the TIDLAS remote sensor (ca. 2.4, mean CO2 column density of ca. 5500 ppm-m with an optimal overlap of laser and plume of ca. 13 000 ppm-m) (11). The distinction between detection limit and instrument precision as well as the dependence of undiluted detection limits on exhaust plume characteristics was explained in detail by Nelson et al. (11); similar arguments are applied in this study. The detection limits presented in Table 2 for column density (i.e., diluted exhaust) need to be multiplied by the dilution factor to obtain undiluted exhaust detection limits. For example, the mean detection limit for NO in undiluted exhaust is 15 ppm. This is superior to the 20 ppm detection limit reported by Popp et al. for a single gas, dispersive UV remote sensor using a photodiode array as a detector (28). 2854

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 13, 2000

FIGURE 5. Spectral data corresponding to 45.4 ppm-m THC (C3H8) in vehicle exhaust. Note that the narrow spectral range (16 nm) of the prototype described by Popp et al. limited measurements to either NO or aromatic hydrocarbons. Criteria Pollutants. CO levels, normalized to undiluted exhaust, are shown in Figure 4. Mean CO concentrations observed in the exhaust of vehicles powered by gasoline, diesel, CNG, and M85 were determined to be 1.92%, 0.0%, 0.34%, and 2.01% respectively. The relationship between mean emission levels and fuel type qualitatively follows the trend reported in the literature (29). Also consistent with our observations is a report from a 1994 remote sensing study, where mean CO emissions from medium/heavy-duty diesel trucks in Los Angeles were found to be 5 times lower than for gasoline-powered cars (5). The RSI showed excellent sensitivity to trace levels of THC in the exhaust plume; the THC detection limit in undiluted exhaust was determined to be 35 ppm. This is 14 times better than the NDIR remote sensor described by Guenther et al. (7). IR spectra of vapor-phase gasoline were recorded with the RSI in our laboratories. The propane reference spectrum provided an excellent spectral match to the band at 3390 nm (2950 cm-1, C-H stretch) resulting from the many aliphatic hydrocarbons in the mixture. A typical spectral match between a reference and vehicle exhaust sample spectrum is shown in Figure 5. Figure 6 shows the THC levels recorded during the field study. In all cases, except for vehicle #8 (Nissan 200SX), THC concentrations in the undiluted exhaust were very low. Vehicle #8 was later found to have a fuel leak in the engine. Mean levels by fuel type were 126.5 ppm for gasoline, 50.1 ppm for diesel, 55.8 ppm for CNG, and 25.5 ppm for M85, measured as methanol using the corresponding reference spectrum. Without the biasing high emitter (vehicle #8), the mean for gasoline-powered vehicles was 71.2 ppm.

FIGURE 6. THC levels in undiluted exhaust of hot vehicles. Three successive measurements per vehicle were made. Vehicle #19: measurements made as methanol. Thus, the difference in THC emissions between these vehicles was not statistically significant. The concentrations observed in our study are significantly lower than mean values reported in previous remote sensing measurements of exhaust from gasoline-powered LDMVs (370 ppm) and from dieselpowered medium-duty trucks (350 ppm) in Los Angeles (5). The difference could be attributed to the various gasoline formulations that were employed. In addition to moderate sensitivity, NDIR remote sensors are also prone to a significant water interference (30, 31). In some cases, water droplets in the exhaust plume have led to erroneously high THC readings (up to 1% THC) (30). In our RSI configuration, most of the water bands in the THC measurement region were saturated due to the long optical path lengths used. Also, our dispersive instrument matches a reference spectral signature to the spectrum of vehicle exhaust. For water interference to occur, its spectral features would have to closely match the propane spectrum. The low resolution employed makes this highly unlikely as the weak, sharp water lines around 3390 nm are expected to blur into the background noise (32). No water interference on THC was observed by visually inspecting IR spectra collected during the field study. Additionally, a turbulent steam plume was introduced into the instrument’s beams, and UV-visIR spectra were recorded. No interference to the THC measurement was observed. The absence of water interference represents a considerable improvement over the NDIR remote sensor, especially when measurements are made under cold ambient conditions. Aromatic hydrocarbons were speciated as benzene, toluene, and p-xylene in the exhaust of some vehicles. Detected levels were low (toluene < 212 ppm, and p-xylene < 41 ppm), as expected based on the low observed THC levels. Peak values were observed for vehicle 8, as with THC. No phenol was observed. The concentration of aldehydes (formaldehyde, acetaldehyde, and benzaldehyde) in vehicle exhaust was typically below the instrument’s detection limit. Formaldehyde (0-15.6 ppm in undiluted exhaust) and benzaldehyde (0-10.9 ppm in undiluted exhaust), among other chromophores, were measured by Pitts et al. (23) with a multiple reflection UV differential optical absorption spectrometer (DOAS) placed 2 m behind LDMVs running on an outdoor chassis dynamometer. The sensitivity of the RSI described here is too low to detect such levels of formaldehyde but would have measured analogous benzaldehyde levels if present. The observed NO concentrations are shown in Figure 7. Emissions from vehicles 7 and 8 were exceptionally high. Mean undiluted levels by fuel type were 475.6 ppm for gasoline, 934.6 ppm for diesel, 2041.3 ppm for CNG, and 0.0 ppm for M85. In excellent agreement with our data for gasoline-powered vehicles is a remote sensing study by Zhang

FIGURE 7. NO levels in undiluted exhaust of hot vehicles. Three successive measurements per vehicle were made. et al., where emissions from over 50 000 vehicles were measured using an NDUV remote sensor affording a mean NO level of 500 ppm (10). The observed NO emission trends as a function of fuel type were consistent with those reported in the literature (29). N2O emissions were close to the instrument’s detection limit. When N2O was detected, the measured values agreed well with values reported by Jime´nez et al. (14) for high emitters (50-200 ppm) using a TIDLAS-based remote sensor. Nitrogen dioxide (NO2) and nitrous acid (HONO) were detected in isolated cases but were at the low borderline of the instrument’s sensitivity and, therefore, are not conclusive. Pitts et al. (24) observed NO2 and HONO levels up to 7.8 and 4.4 ppm, respectively, well below the RSIs detection limit in undiluted exhaust. Recent results suggest that an average NO/HONO ratio in motor vehicle exhaust of (2.9 ( 0.5) × 10-3 is realistic (33); this would yield a mean HONO concentration of 1.4 ppm in the exhaust from the gasolinepowered vehicles tested in this study, with a high of 9.6 ppm for vehicle #9. No sulfur dioxide was detected, as expected based on the low sulfur levels in Phase II reformulated gasoline (30 ppm S in the fuel affords approximately 2 ppm SO2 in vehicle exhaust). NH3 Emissions. Reports on NH3 levels in the exhaust of LDMVs are rare and date back to the late 1970s (18-21). Cadle et al. reported NH3 concentrations from 0.9 to 15.6 ppm in undiluted exhaust of nine vehicles equipped with either oxidative catalysts or TWCs (20). However, when a vehicle equipped with a TWC was tested at an air-to-fuel ratio below 13.5, NH3 emissions rose to 327.1 ppm. Similar observations have been made by Bradow et al. with the oxygen sensor disconnected (18). Despite these important observations, NH3 measurements are not typically included in emission studies of in-use vehicles. One notable exception is a recent report by Fraser and Cass (34). Gas-phase NH3 emitted by vehicles in a Southern California tunnel was collected onto open-faced stacked filter systems impregnated with oxalic acid. Ammonium levels were determined colorimetrically in aqueous extracts. Fraser and Cass found that NH3 emissions exceeded previous estimates used in atmospheric models by approximately a factor of 2. They concluded that NH3 release from vehicles is comparable to that from livestock waste decomposition at dairies, the most important traditional source of NH3 emissions. High NH3 concentrations in the exhaust of stationary vehicles were previously measured in our laboratory using an extractive DUV system (35). The RSI described in the present study is expected to provide superior measurements due to its open optical path (i.e., no adsorptive effects), allowing convenient, unequivocal detection. The sensor was found to have excellent sensitivity, ca. 2.5 ppm detection limit in undiluted exhaust, and selectivity to NH3 as illustrated by the fit result shown in Figure 8. VOL. 34, NO. 13, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2855

FIGURE 10. NH3 concentration as a function of A/F.

FIGURE 8. Spectral fit data corresponding to 16.9 ppm-m NH3 in diluted vehicle exhaust.

FIGURE 11. NH3 versus NO column densities in diluted vehicle exhaust. The rhodium portion of the TWC is primarily designed to reduce NOx to dinitrogen (N2), as shown in eq 3 (36).

2NO + 2CO f N2 + 2CO2 FIGURE 9. NH3 levels in undiluted exhaust of hot vehicles. Three successive measurements per vehicle were made. NH3 levels detected in the emissions from on-road vehicles on a car-by-car basis are shown in Figure 9. The mean levels observed in this study for gasoline-powered vehicles (317.0 ppm) compared well to the value reported by Cadle et al. for a vehicle operating under fuel-rich conditions, but vehicles 13 and 14 were found to emit much higher levels (>1000 ppm). In 18% (10 of 57) of the measurements, the NH3 to NO ratio was above 10. Mean NH3 emissions from diesel and CNG vehicles (16-18) were significantly lower, 0.0 and 12.1 ppm, respectively, while the vehicle powered by M85 emitted a mean NH3 level of 438.5 ppm. Based on the CO2, CO, and THC levels in vehicle exhaust the instantaneous air-to-fuel ratio (A/F) can be calculated, as shown in eq 2 (27).

A/F (by mass) ) 4.93 × (3 + 2Q)/(1 + Q + 3Q′) (2) As expected, a plot of NH3 emissions versus A/F (Figure 10) shows a correlation between low NH3 levels in the exhaust and lean driving conditions (A/F ca. 14.75). What is surprising, however, is our observation that NH3 levels above 300 ppm were emitted at mean A/F ratios of 13.93 ( 0.5. These A/F ratios correspond to combustion conditions ranging from rich to near-stoichiometric; high NH3 emissions are not expected above A/F ratios of ca. 13.5 (20). This result raises serious concerns with respect to the potential magnitude of vehicular NH3 emissions, previously believed to emanate primarily from LDMVs equipped with TWCs and running in the rich regime. 2856

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 13, 2000

(3)

NH3 is an undesirable reduction byproduct of this reaction and is believed to result from the reaction of NO, CO, and hydrogen (H2) in the catalytic unit (37), as shown in eq 4.

NO + CO + 1.5H2 f NH3 + CO2

(4)

Reduction is thought to proceed at the catalyst surface by the reaction of adsorbed nitrogen and hydrogen atoms (38) and by hydrolysis of gaseous isocyanic acid (eq 5) (37).

HNCO + H2O f NH3 + CO2

(5)

Thus, measuring NH3 levels in vehicle exhaust provides a means of gauging the performance of the catalyst with respect to NOx reduction. In our study, the catalyst temperature, chemical composition, coating, and inert supports were not analyzed. Therefore, it is not possible to predict the extent of N2 formation from NOx in “cold” vehicles. However, meaningful information on catalyst reduction efficiency can still be obtained from the field measurements. A plot of the measured NH3 versus NO column densities in diluted exhaust of “hot” vehicles, as shown in Figure 11, conforms to the expected inverse relationship (i.e., when NO levels are high, the corresponding NH3 levels are low, and vice versa). The changes in NO and NH3 levels as a function of operating conditions are summarized in Figure 12. A low NO ratio is indicative of efficient NOx reduction and suggests that the catalyst activity was low when the vehicle was running “cold”. High NH3 ratios are the result of NH3 production by the catalyst with the vehicle running “hot”. Therefore, a combination of low NO ratios and high NH3 ratios suggest that the catalyst is more efficient at producing NH3 than N2. This is the case for 35% of the tested fleet. On the other hand, 47%

(2) (3) (4) (5) (6) (7) (8)

FIGURE 12. Change in NO and NH3 levels as a function of catalyst temperature. (Note: values for vehicles # 15 and 19 have been omitted as no “cold” data was available.) of the fleet exhibited NO ratios close to unity and NH3 ratios below one, indicating that the reductive catalyst had already reached equilibrium under “cold” conditions. We note that six out of the eight vehicles falling under this category were high NO emitters when running “hot” (1000 ppm or higher) suggesting a low activity for the reductive catalyst. Two noteworthy differences in experimental parameters, between the current work and earlier studies, are the inherent TWC activities and sulfur content of the fuel. Current catalysts are significantly more active than older versions, and sulfur levels in Phase II reformulated gasoline are about 10 times lower than those employed in gasoline in the 1970s. The effect of fuel sulfur concentration on NOx reduction activity and selectivity has been examined in fully formulated Pt/Rh catalysts (39, 40). While low fuel sulfur has been shown to favor NOx reduction activity, it also increased the formation of NH3. The Pt portion of the catalyst is known to generate H2 via the water-gas shift reaction (eq 6) and the steam reforming reaction (eq 7) (41). The presence of H2 leads to concomitant formation of NH3 at the catalyst (42).

(9) (10) (11) (12) (13) (14)

(15)

(16) (17) (18) (19)

CO + H2O f H2 + CO2

(6)

(20)

CnH(2n+2) + 2nH2O f (3n+1)H2 + nCO2

(7)

(21) (22) (23)

At high fuel sulfur levels, the active sites for H2 formation on the Pt and/or Pd catalyst are believed to be selectively poisoned by SO2 in the exhaust. Only low levels of H2, and therefore NH3, are produced (42). At low fuel sulfur levels, however, much of the NO was found to be converted to NH3, apparently predominantly at the Pt sites (42). The implications of vehicular NH3 emissions on air quality have been discussed previously by Fraser and Cass (34). Gasphase NH3 combines with nitric acid, produced in photochemical smog, to form fine particle ammonium nitrate aerosol. The well-known Los Angeles visibility problem results from light scattering by fine particulate matter, and aerosol nitrate levels are often dominant during these episodes.

Acknowledgments This research was supported by the South Coast Air Quality Management District under contract No. AB2766/96028. We gratefully acknowledge the valuable assistance provided by this program as well as the loan of the M85 vehicle. We are also indebted to the Gas Company and Rio Hondo College for the loan of the CNG vehicles. The statements and conclusions in this paper are those of the authors and not necessarily those of the South Coast Air Quality Management District.

Literature Cited (1) National Air Pollution Trends, 1900-1995; EPA-454/R-96-007; United States Environmental Protection Agency, Office of Air

(24) (25) (26) (27) (28)

(29) (30) (31)

(32) (33) (34) (35)

Quality Planning & Standards, U.S. Government Printing Office: Washington, DC, 1996. Chaney, L. W. J. Air Pollut. Control. Assoc. 1983, 33, 220-222. Bishop, G. A.; Starkey, J. R.; Ihnenfeldt, A.; Williams, W. J.; Stedman, D. H. Anal. Chem. 1989, 61, 671A-677A. Guenther, P. L.; Bishop, G. A.; Peterson, J. E.; Stedman, D. H. Sci. Total E. 1994, 146/147, 297-302. Stephens, R. D. J. Air Waste Manage. Assoc. 1994, 44, 12841292. Zhang, Y.; Stedman, D. H.; Bishop, G. A.; Guenther, P. L.; Beaton, S. P.; Peterson, J. E. Environ. Sci. Technol. 1993, 27, 1885-1891. Guenther, P. L.; Stedman, D. H.; Bishop, G. A.; Beaton, S. P.; Bean, J. H.; Quine, R. W. Rev. Sci. Instrum. 1995, 66, 3024-3029. Stephens, R. D.; Mulawa, P. A.; Giles, M. T.; Kennedy, K. G.; Groblicki, P. J.; Cadle, S. H.; Knapp, K. T. J. Air Waste Manage. Assoc. 1996, 46, 148-158. Zhang, Y.; Stedman, D. H.; Bishop, G. A.; Guenther, P. L.; Beaton, S. P. Environ. Sci. Technol. 1995, 29, 2286-2294. Zhang, Y.; Stedman, D. H.; Bishop, G. A.; Beaton, S. P.; Guenther, P. L.; McVey, I. F. J. Air Waste Manage. Assoc. 1996, 46, 25-29. Nelson, D. D.; Zahniser, M. S.; McManus, J. B.; Kolb, C. E.; Jime´nez, J. L. Appl. Phys. B 1998, 67, 433-441. Jime´nez, J. L.; Koplow, M. D.; Nelson, D. D.; Zahniser, M. S.; Schmidt, S. E. J. Air Waste Manage. Assoc. 1999, 49, 463-470. Beaton, S. P.; Bishop, G. A.; Zhang, Y.; Ashbaugh, L. L.; Lawson, D. R.; Stedman, D. H. Science 1995, 268, 991-993. Jime´nez, J. L.; Nelson, D. D.; Zahniser, M. S.; Kolb, C. E. Remote Sensing Measurements of On-road Vehicle Emissions of an Important Greenhouse Gas: Nitrous Oxide. In Remote Sensing Instrumentation Improvement; Proceedings of the 7th CRC Onroad Vehicle Emissions Workshop, San Diego, CA, April 9-11, 1997; CRC: Atlanta, GA, 1997; pp 4/13-4/39. Kert, J. Remote Sensing of On-road Vehicle Emissions; CRC Project No. VE-8; Hughes Environmental Systems, Inc.: Manhattan Beach, CA, Jan 1993. Seinfeld, J. H. Atmospheric Chemistry and Physics of Air Pollution; John Wiley & Sons: New York, 1986; pp 111-194. Finlayson-Pitts, B. J., Pitts, J. N., Jr. Science 1997, 276, 10451052. Bradow, R. L.; Stump, F. D. SAE Technical Pap. Ser. 1977, No. 770369. Cadle, S. H.; Nebel, G. J.; Williams, R. L. SAE Technical Pap. Ser. 1980, No. 790694. Cadle, S. H.; Mulawa, P. A. Environ. Sci. Technol. 1980, 14, 718723. Urban, C. M. SAE Technical Pap. Ser. 1980, No. 790696. White, J. U. J. Opt. Soc. Am. 1942, 32, 285-288. Press: W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P. Numerical Recipes in C: The Art of Scientific Computing; Cambridge University Press: New York, 1992; Vol. 2, pp 59-70. Pitts, J. N., Jr.; Biermann, H. W.; Winer, A. M.; Tuazon, E. C. Atmos. Environ. 1984, 18, 847-854. Mellqvist, J.; Rose´n, A. J. Quant. Spectrosc. Radiat. Transfer 1996, 56, 209-224. Bishop, G. A.; Stedman, D. H. J. Air Waste Manage. Assoc. 1996, 46, 667-675. Bishop, G. A.; Stedman, D. H. Acc. Chem. Res. 1996, 29, 489495. Popp, P.; Bishop, G. A.; Stedman, D. H. Development of a Highspeed Ultraviolet Spectrophotometer Capable of Real-time NO and Aromatic Hydrocarbon Detection in Vehicle Exhaust. In Remote Sensing Instrumentation Improvement; Proceedings of the 7th CRC On-road Vehicle Emissions Workshop, San Diego, CA, April 9-11, 1997; CRC: Atlanta, GA, 1997; pp 4/1-4/12. Chang, T. Y.; Hammerle, R. H.; Japar, S. M.; Salmeen, I. T. Environ. Sci. Technol. 1991, 25, 1190-1197. Sjo¨din, A. J. Air Waste Manage. Assoc. 1994, 44, 397-404. Pierson, W. R.; Gertler, A. W.; Robinson, N. F.; Sagbiel, J. C.; Zielinska, B.; Bishop, G. A.; Stedman, D. H.; Zweidinger, R. B.; Ray, W. D. Atmos. Environ. 1996, 30, 2233-2256. Griffiths, P. R.; Haseth, J. A. D. Fourier Transform Infrared Spectrometry; John Wiley & Sons: New York, 1986. Kirchstetter, T. W.; Harley, R. A.; Littlejohn, D. Environ. Sci. Technol. 1996, 30, 2843-2849. Fraser, M. P.; Cass, G. R. Environ. Sci. Technol. 1998, 32, 10531057. Baum, M. M. Characterizing and Analyzing Components of Vehicle Exhaust Emissions by UV Spectroscopy. In Inspection/ Maintenance Remote Sensing; Proceedings of the World Car

VOL. 34, NO. 13, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2857

(36) (37) (38) (39)

Conference, Riverside, CA, Jan 19-22, 1997; University California, Riverside: Riverside, CA, 1997; pp 213-224. Kirk-Othmer Encyclopedia of Chemical Technology, 4th ed.; John Wiley & Sons: New York, 1991; Vol. 9, pp 985-994. Du ¨ mpelmann, R.; Cant, N. W.; Trimm, D. L. J. Catal. 1996, 162, 96-103. Williams, C. T.; Tolia, A. A.; Weaver, M. J.; Takoudis, C. G. Chem. Eng. Sci. 1996, 51, 1673-1682. Summers, J. C.; Baron, K. J. Catal. 1979, 57, 380-389.

2858

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 13, 2000

(40) Williamson, W. B.; Gandhi, H. S.; Heyde, M. E.; Zawacki, G. A. SAE Technical Pap. Ser. 1979, 790942. (41) Barbier, J.; Duprez, D. Appl. Catal. B 1994, 4, 105-140. (42) Gandhi, H. S.; Shelef, M. Appl. Catal. 1991, 77, 175-186.

Received for review December 8, 1999. Revised manuscript received March 24, 2000. Accepted April 25, 2000. ES991351K