The Effects of the Catalytic Converter and Fuel Sulfur Level on Motor

nor the fuel sulfur content has a significant effect on gasoline vehicle tailpipe particulate matter (PM) emissions. For current technology, port fuel...
0 downloads 0 Views 251KB Size
Environ. Sci. Technol. 2002, 36, 276-282

The Effects of the Catalytic Converter and Fuel Sulfur Level on Motor Vehicle Particulate Matter Emissions: Gasoline Vehicles M. MATTI MARICQ,* RICHARD E. CHASE, NING XU, AND DIANE H. PODSIADLIK Ford Motor Company, Research Laboratory, P.O. Box 2053, MD 3083, Dearborn, Michigan 48121

Scanning mobility and electrical low-pressure impactor particle size measurements conducted during chassis dynamometer testing reveal that neither the catalytic converter nor the fuel sulfur content has a significant effect on gasoline vehicle tailpipe particulate matter (PM) emissions. For current technology, port fuel injection, gasoline engines, particle number emissions are e 2 times higher from vehicles equipped with blank monoliths as compared to active catalysts, insignificant in contrast to the 90+% removal of hydrocarbons. PM mass emission rates derived from the size distributions are equal within the experimental uncertainty of 50-100%. Gravimetric measurements exhibit a 3-10-fold PM mass increase when the active catalyst is omitted, which is attributed to gaseous hydrocarbons adsorbing onto the filter medium. Both particle number and gravimetric measurements show that gasoline vehicle tailpipe PM emissions are independent (within 2 mg/mi) of fuel sulfur content over the 30-990 ppm concentration range. Nuclei mode sulfate aerosol is not observed in either test cell measurements or during wind tunnel testing. For three-way catalyst equipped vehicles, the principal sulfur emission is SO2; however a sulfur balance is not obtained over the drive cycle. Instead, sulfur is stored on the catalyst during moderate driving and then partially removed during high speed/load operation.

Introduction Reports of epidemiological correlations between ambient particulate matter (PM) levels and health effects (1, 2) as well as the Environmental Protection Agency’s (EPA) subsequent introduction of the PM2.5 particle standard (3) have elicited considerable interest in the measurement and characterization of airborne particulate matter. Motor vehicle tailpipe emissions are no exception to this. Thus, there has been a resurgence of diesel vehicle PM research (4-7), which had seen significant activity two decades ago (8, 9). And there have been extensions aimed at studying gasoline vehicle PM emissions (10-13). Understanding the effect of engine operating conditions, for example, speed, load, and spark timing, on PM emissions is, of course, crucial to the design of lower emitting engines. However, emissions reduction is a systems issue that includes fuel, engine, and exhaust aftertreatment. Equally important, therefore, are the relationships between PM emissions on * Corresponding author phone: (313)594-7527; fax: (313)594-2923; e-mail: [email protected]. 276

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002

one hand and the catalytic converter and fuel structure on the other (14). With respect to gaseous emissions, this relationship is obvious; modern three-way catalysts remove well over 90% of the engine-out hydrocarbon emissions (vide infra). Sulfur, in contrast, poisons catalytic activity (15). What are the corresponding relationships with regard to PM emissions? It is conceivable that the catalyst will oxidize particles, thus removing them from the exhaust stream. Or, the particles might pass unscathed through the catalyst monolith. Sulfur in the fuel, on the other hand, has been blamed for nanoparticle formation (particles < ∼50 nm diameter), as a result of its oxidation to SO2, subsequent conversion to sulfuric acid, and nucleation into an aerosol (16). The present paper examines these questions with respect to current technology gasoline vehicles. The companion paper addresses the corresponding issues for light duty diesel vehicles. PM measurements of gasoline vehicle exhaust typically reveal a roughly 5-fold lower emission rate during the hot start phase of the Federal Test Procedure (FTP) drive cycle as compared to the cold start (11); thus, one might anticipate that the catalyst plays a significant role in PM reduction. However, as discussed below, this is not the case. The effect of fuel sulfur on PM emissions is examined via conventional dilution tunnel measurements as well as using a wind tunnel to simulate “real world” exhaust dilution. After describing the experimental method and presenting the data, we discuss the results relative to the questions that currently exist regarding nanoparticle emissions and the comparisons that are made between PM emissions from diesel versus gasoline vehicles.

Experimental Method The chassis dynamometer facility employed for these experiments is the same one as used in our previous work characterizing gasoline vehicle PM emissions (11, 17, 18). A robot driver drives the test vehicles on a 48-in. single roll, AC electric dynamometer, following the Federal Test Procedure (FTP) and US06 portion of the Supplemental FTP drive cycles. The FTP consists of three phases: (1) a cold start and moderate speed driving (up to ∼50 mph), (2) a city drive portion, and (3) a repeat of the phase 1 drive trace, but with a hot start. The US06 is an aggressive high-speed cycle with speeds up to 80 mph. The vehicle exhaust is conducted via an insulated and heated (to prevent water condensation) corrugated stainless steel hose to a 30.4 cm diameter stainless steel dilution tunnel. The hose has a diameter of 9 cm and a length of 6.5 m. Therefore, on route to the dilution tunnel, the exhaust experiences a residence time in the transfer hose that ranges from ∼1 s at a 2.5 m3/min exhaust rate typical of the heavier FTP accelerations to about 20 s at an idle exhaust rate of 0.1 m3/min. At the dilution tunnel the exhaust is diluted with heated (38 °C), filtered, low humidity (-9 °C dewpoint) air. The exhaust is introduced along the axis of the tunnel, and an orifice plate placed just downstream promotes mixing. The tunnel operates at a constant total (exhaust + diluent) flow that can be set between 10 and 30 m3/s. Therefore the dilution ratio varies during the course of the test cycle; typically the instantaneous dilution ratio varies from about 120 at idle to about 2.5 at the heaviest accelerations. It falls below 5 for about 25 s of the FTP cycle. The average dilution ratio varies from about 15-20 for phases 1 and 3 of the FTP cycle and from 20 to 30 for phase 2, depending on test vehicle. The fact that these dilution ratios may affect PM sampling is one of the motivations for the wind tunnel studies described below. 10.1021/es010961t CCC: $22.00

 2002 American Chemical Society Published on Web 12/08/2001

Particles are sampled more than 10 tunnel diameters downstream to permit thorough mixing. Standard gravimetric PM emissions measurements are performed by sampling the diluted exhaust through 2 µm Teflon filters (47 mm diameter expanded PTFE Teflon) (19) at a rate of 0.66 L/s. Automated valves switch between separate filters for each phase of the FTP drive cycle. Because of the low PM emissions rates, the exhaust from four repeated drive cycles is collected onto the same set of filters prior to making the mass determinations. Typically this results in the collection of 5-500 µg of particulate matter, with higher values recorded for the cold start phase and lower values for the hot phases. A small, < 1 µg, correction is made for the PM in the tunnel background (0.3-0.8 µg/m3). The filters are equilibrated to room conditions for a minimum of 2 h prior to pre- and post-weighings in a temperature and humidity controlled room (20). The resulting mass emission rates are accurate to about 0.1 mg/mi; the additional contribution of test-to-test variation is included in the tabulated data. Particle size distributions are measured by a scanning mobility particle sizer (SMPS, from TSI, Inc.) and an electrical low-pressure impactor (ELPI, from Dekati, Inc.). The diluted exhaust is sampled via a 3/8” i.d. stainless steel tube inserted into the tunnel flow and conducted via Tygon tubing to the sizing instruments. The ELPI charges the particles using a corona discharge and then separates them according to aerodynamic size using a Berner low-pressure cascade impactor (12 stages ranging from 35 to 10 000 nm aerodynamic diameter) (21). It provides transient second-by-second particle size distributions by monitoring the electrical current deposited on the various impactor plates as a function of time. The SMPS also charges the particles in the diluted exhaust but separates them according to their drift rate through air in an electric field. Scanning the electric potential provides a high-resolution size distribution of the particles’ mobility diameters. However, the scanning process requires > 30 s to accomplish, which is not suitable for transient testing. Instead, the SMPS is fixed to transmit a single mobility diameter throughout the drive cycle. In this mode the system responds with a time constant of about 5 s. A set of transient size distributions is then built up from a series of drive cycles each run at a different mobility diameter. As we have previously shown, this is feasible owing to a PM emissions reproducibility that averages about 20% from test to test (11). The raw particle count rates at a fixed particle diameters, C(t, Dp), are converted to size distributions via

C(t, Dp)T(Dp)qT RC dN(t) e ) dlog (Dp) qaf(+1)ic

(1)

This expression simply accounts for the particle charging efficiency and transmission function of the SMPS and corrects for inlet and particle counter nonidealities. Here N(t) represents the tailpipe particles per second, T(Dp) is the transmission function of the classifier (approximated as a triangle, 2/∆(logDp)), qT is the tunnel flow, qa is the aerosol sample flow, f(+1) is the size dependent fraction of +1 charged particles, and c and i are the efficiencies of the condensation nuclei counter and the impactor at the classifier entrance. The factor eRC represents a coincidence correction for high particle count rates (R ) 4 × 10-7 s). Integration of the transient data recorded over the drive cycle provides phase by phase size selected PM emissions. Because PM emissions are measured only at a small number of sizes (4-8), the phase averaged data are fit to log-normal distributions for data interpolation. These are used to estimate particle number emissions and average size. Semiquantitative mass emission rates are calculated from the size distributions

by assuming the particles to have a density of 1 g/cm3 and integrating the total volume of emitted particulate matter. We have previously reported nanoparticle artifacts that can grossly affect particle number measurements made using standard dilution tunnel methodology (18). The artifacts arise from hydrocarbon material deposited over time onto the walls of the transfer hose to the dilution tunnel. During high speed steady-state testing or aggressive transient drive cycles, the heat from the vehicle exhaust can be sufficient to desorb some hydrocarbon material, which subsequently nucleates into nanoparaticles during dilution and gives the appearance of high particle number emissions. Similar artifacts also appear if silicone rubber hoses, or connectors, to the tailpipe are used. The extent of the artifacts depends on the exhaust temperatures of the test vehicles and the condition of the transfer hose. We have avoided silicone rubber materials and conditions that are prone to artifacts (especially in the case of US06 tests) in the experiments reported here. Because conventional PM sampling does not reproduce real world exhaust dilution rates, the potential for nucleation mode sulfate aerosol formation is also examined by vehicle tests performed in a wind tunnel. Air enters through a 4.7 m2 nozzle and exits ∼7 m downstream. The tunnel is operated in a complete air exchange mode to ensure that particles do not accumulate in the wind tunnel air during testing. The test vehicle is operated on a chassis rolls at the upstream end of the tunnel. Particle size distributions in the exhaust plume are recorded using a SMPS centrally located 4.6 m behind the vehicle. The exhaust plume is sampled through a 0.9 cm i.d., 15 cm long stainless steel tube, and transported to the SMPS using a short (∼1/2 m) Tygon tube. The vehicle was tested at steady-state speeds of 40 mph, 70 mph, and 70 mph with a 5% grade. Transient effects were examined using repeated accelerations from 40 mph to 70 mph at 5 mph/s.

Results Active Catalyst versus Blank Monoliths. One can examine the effect of the catalytic converter on gasoline vehicle tailpipe PM emissions in a number of ways. One is to measure the PM before and after the catalyst, another is to remove the catalyst, and a third is to remove the active ingredients from the catalyst substrate. Each method has advantages and disadvantages. The first is the most straightforward in principle but poses the challenge of ensuring that particle sampling is performed identically from each location. The second method tests the combined mechanical filtering of exhaust particles by the monolith support as well as the chemical removal of soot by catalytic oxidation. But removal of the catalyst substrate from the exhaust system alters the exhaust flow dynamics, which in turn affects engine operation. Therefore, in the present experiments the third approach was taken, namely to compare PM emissions before and after replacing the active catalyst with a blank monolith support. The results inform us on the ability of the catalyst to oxidatively remove particles but not on the mechanical filtering capacity of the support. With this procedure, the data are not simultaneously measured during a single vehicle test. But, both the active catalyst and blank monolith tests are performed using the identical dilution tunnel sampling system. Two vehicles, a light duty truck and a large passenger car, were tested to examine the catalyst effect on PM emissions. They are described in Table 1. The emissions measurements are presented in Table 2. As expected, the gaseous emissions are substantially higher when the active catalyst is omitted. The drive traces for phases 1 and 3 of the FTP drive cycle are identical, except that phase 1 begins with a cold engine start and phase 3 begins with a hot start, after a 10 min engine off period. With the blank substrate, the higher gaseous emissions measured in phase 1 versus phase 3 reflect the VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

277

TABLE 1. Test Vehicles disp. vehiclea year liters # cyl

odometer miles transmission

test fuel

Catalyst Study LT1 C1

1995 3.8 1994 4.6

6 8

LT2 LT3 C2

1997 3.8 1996 4.0 1997 3.0

6 6 6

C3

1996 4.6

8

EPA cert. EPA cert.

2600 10600

automatic automatic

4200 2900 3500

automatic automatic automatic

9500

automatic

Sulfur Level Study Cal ph 2 refrm Cal ph 2 refrm Cal ph 2 refrm

Wind Tunnel Study

a

Cal ph 2 refrm & regular grade

LT ) light duty gasoline truck, C ) gasoline automobile.

effect of engine temperature on the emissions. Use of an active catalyst reduces hydrocarbon emissions by nearly ∼90% in phase 1 and by >98% in phases 2 and 3. CO and NOx reductions follow similar trends. The higher phase 2 and 3 catalyst efficiencies relative to phase 1 occur because of the catalyst light-off period in phase 1. The gravimetric filter data and the particle size distributions tell two different stories with respect to PM emissions. The gravimetric PM data in Table 2 show the emission rates to increase by 3-10 times when the active catalyst is omitted. In stark contrast, the calculated mass emissions from the particle size distributions (Table 2) indicate that the catalyst has a dramatically smaller effect on PM. This is graphically illustrated in Figure 1. In the top panel (vehicle LT1, FTP test), the catalyst has a negligible effect on particles > ∼50 nm diameter but decreases by a factor of 2-3 the number of particles < ∼50 nm. The net effect on the calculated PM mass emissions is minimal (Table 2). In the bottom panel (vehicle C1, US06 test) the catalyst has no discernible effect on exhaust particle size distribution over the entire 20-300 nm diameter range. Table 2 shows that the mass emissions calculated from the size distributions agree well with the gravimetric PM mass when an active catalyst is installed, but they are significantly lower for the blank monoliths. The large relative uncertainties in particle number, diameter, and calculated mass emissions (20-50% for particle number and 50-100% in calculated mass) originate from the necessity of interpolating the total emissions from measurements performed at a limited number of particle sizes (4-8). However, even

FIGURE 1. Comparison of particle size distributions measured with an active catalyst versus a blank brick. Top panel: phases 1 (cold start) and 3 (hot start) of the FTP drive cycle for gasoline vehicle LT1. Symbols represent ELPI data and lines represent SMPS data. Bottom panel: US06 drive cycle for gasoline vehicle C1. Lines represent log-normal fits to the data. at a ( 100% uncertainty in the calculated mass emission rates, it is clear that the particle counting data contradict the gravimetric measurements. The discrepancy between gravimetric and particle counting data can be reconciled by noting that removal of the active catalyst results in a 10-100-fold increase in hydrocarbon concentrations that are sampled through the filters along with the particles. This suggests that a fraction of the gaseous hydrocarbons adsorbs onto the filters and/or onto particles collected by the filters, and that this is responsible for the mass increase recorded by the filters in the absence of the catalyst. This explanation is supported by a decrease in of hydrocarbon concentrations on the downstream side of the filters, as measured by flame ionization detection (22).

TABLE 2. Emissions Summary - Catalytic Converter Study vehiclea LT1d LT1 C1e

C1

catalyst/ no catalyst

phase

THC, g/mi

CO, g/mi

NOx, g/mi

grav. PM,b mg/mi

calc. PM,c mg/mi

no. of particles,c ×1010/mi

mean diam,c nm

no catalyst no catalyst no catalyst catalyst catalyst catalyst no catalyst no catalyst no catalyst no catalyst catalyst catalyst catalyst catalyst

1 2 3 1 2 3 1 2 3 US06 1 2 3 US06

2.08 1.66 1.54 0.293 0.010 0.032 2.15 1.71 1.46 1.91 0.236 0.004 0.026 0.089

13.9 9.4 10.0 3.02 0.26 0.54 11.2 9.84 9.09 31.2 0.90 0.25 0.45 15.6

2.50 1.82 2.41 0.099 0.002 0.054 2.36 1.72 2.27 3.90 0.273 0.032 0.103 0.074

2.5 ( 0.4 2.1 ( 0.4 2.1 ( 0.5 0.9 0.3 0.4 3.1 ( 1.7 1.3 ( 0.6 5.0 ( 1.2

1.4 ( 1.0 0.03 ( 0.02 0.2 ( 0.2 1.1 ( 0.5 0.1 ( 0.05 0.2 ( 0.1 0.3 ( 0.2 0.03 ( 0.02 0.4 ( 0.6 0.2 ( 0.4 0.3 ( 0.2 0.01 ( 0.01 0.3 ( 0.5 0.5 ( 0.9

350 ( 100 60 ( 15 100 ( 25 170 ( 50 60 ( 20 50 ( 20 80 ( 50 5(3 40 ( 30 80 ( 20 50 ( 30 3(2 50 ( 30 80 ( 20

30 ( 15 42 ( 3 39 ( 3 61 ( 5 54 ( 3 67 ( 8 40 ( 20 80 ( 30 100 ( 40 37 ( 9 40 ( 20 80 ( 30 100 ( 40 39 ( 8

0.6 0.1 0.5

a LT ) light duty gasoline truck, C ) gasoline automobile. b Data represent averages of 1-3 mass measurements. Each measurement comprises PM emissions from four FTP tests collected onto a single filter for each phase. Error bars represent the standard deviation when 2-3 determinations were made. c Error bars are 95% confidence limits propagated from log-normal fits of the size distributions. d Gaseous emissions from two sets of four vehicle tests were repeatable to within 12%; the average deviation was 6%. e Gaseous emissions from two sets of four vehicle tests were repeatable to within 5%; the average deviation was 3%.

278

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002

FIGURE 2. Comparison of particle number emissions for an active catalyst versus blank brick during steady-state operation of vehicle C1 at various speed/load combinations. Panel A: active catalyst. Panel B: blank substrate. At each condition, the particle size distributions exhibit a single mode, which has a nearly log-normal shape and a mean diameter of about 70 nm. Hydrocarbon adsorption onto filters has also been noted by Schauer et al. (4) in the case of diesel vehicle PM measurements. For completeness, Figure 2 presents total particle number emission rates integrated from particle size distributions recorded during steady-state vehicle operation at various speed/load points. With or without an active catalyst, the steady-state emissions are extremely low (equivalent to < 0.05 mg/mi) and require subtraction of the background particles that remain in the filtered (90%) dilution air. At such low levels, the maximum observed variation of a factor of 2 between catalyst and no catalyst is inconsequential. A nanoparticle mode, potentially from semivolatile hydrocarbon nucleation, is not observed in the absence of the active catalyst. The low steady-state PM emissions are consistent with earlier reports that most gasoline particle emissions occur during transient operation (11, 17, 23). Fuel Sulfur Level. The effect of fuel sulfur content on gasoline vehicle tailpipe PM emissions was explored using the two light duty trucks and medium size passenger car that are described in Table 1. Each was equipped with a threeway catalytic converter. California reformulated gasoline, containing 35 ppm sulfur, was employed as the base fuel. The sulfur level was raised to 350 and 600 ppm by adding a cocktail of organosulfur compounds comprised of 4 wt % dimethyl sulfide, 23 wt % thiophene, and 73 wt % benzothiophene. While this produces a somewhat artificial fuel, in the sense that it is not found on the market, the other option of choosing commercially available fuels with varying sulfur content has the drawback that the fuel content then differs in other ways as well, such as the aromatic or oxygenate content. Figures 3 and 4 illustrate the effect of sulfur level on PM number and mass emissions. Table 3 lists the experimental

FIGURE 3. Gasoline vehicle particle mass emissions, as determined by filter collection, as a function of fuel sulfur content during phases 1-3 of the FTP cycle and during the US06 portion of the SFTP cycle.

FIGURE 4. Gasoline vehicle particle number emissions as a function of fuel sulfur content during phases 1-3 of the FTP cycle and during the US06 portion of the SFTP cycle. results for each phase of the FTP drive cycle, along with the FTP average gaseous emissions. Table 4 provides the analogous data for the US06 drive cycle. It is apparent that fuel sulfur content has little effect on either the number or mass of particles emitted from the tailpipe (particle size, not plotted, also remains unaffected). The only noticeable effect VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

279

TABLE 3. FTP Emissions Summary - Sulfur Study FTP 3 phase averageb vehiclea

phase 1

phase 2

phase 3

sulfur, ppm

THC, g/mi

CO, g/mi

NOx, g/mi

NMHC, g/mi

PM,c mg/mi

SO2,d mg/mi

PM,c mg/mi

SO2,d mg/mi

PM,c mg/mi

SO2,d mg/mi

35 350 600 35 600 35 350 600

0.091 0.094 0.112 0.151 0.174 0.183 0.251 0.255

0.90 0.86 1.01 1.12 1.72 0.96 1.32 1.40

0.094 0.172 0.282 0.165 0.432 0.045 0.079 0.112

0.083 0.081 0.092 0.140 0.141 0.165 0.216 0.218

4.8 ( 0.6 3.4 2.8 1.6 1.1 2.5 ( 0.1 1.7 ( 0.3 1.4 ( 0.3

0 7 12 0 85 0 4 34

0.4 ( 0.06 0.7 0.2 0.1 0.1 0.1 ( 0.07 0.1 ( 0.01 0.1 ( 0.03

0 10 47 0 51 0 0 34

1.3 ( 0.1 0.6 0.2 0.2 0.1 0.3 ( 0.05 0.1 ( 0.06 0.3 ( 0.03

0 79 180 0 230 0 41 110

LT2 LT3 C2

a LT ) light duty gasoline truck, C ) gasoline automobile. b Gaseous emissions data from 4 to 8 tests were repeatable to an average of 5-15%, with a few outliers of 50% noted. c Data are masses from 1 to 2 measurements, each comprised of the PM emissions from four FTP tests collected onto a single filter for each phase. Error bars are the standard deviation of two data points. d Values are variable, because extent of sulfur storage depends on vehicle driving history.

TABLE 4. US06 Emissions Summary - Sulfur Study US06 drive cycle vehiclea LT2

LT3 C2

sulfur, THC,b ppm g/mi 35 350 600 35 600 35 350 600

0.09 0.13 0.14 0.06 0.11 0.11 0.15 0.22

CO,b g/mi

NOx,b g/mi

grav. PM,c calc. PM,d SO2, mg/mi mg/mi mg/mi

6.5 6.8 6.6 4.7 6.9 10.3 7.4 11.1

0.21 0.35 0.45 0.87 1.22 0.06 0.075 0.11

9.6 ( 3.2 8.1 ( 4.4 8.7 ( 8.7 6.3 ( 15 14 ( 13 6.5 ( 1.7 4.4 ( 2.3 3.5 ( 1.5

8(5 7(4 8(6 0.4 ( 0.2 0.4 ( 0.2 0.1 ( 0.4 0.1 ( 0.2

57 143

a LT ) light duty gasoline truck, C ) gasoline automobile. b Gaseous emissions data derived from 3 to 6 vehicle tests were repeatable on average to ∼20%. c Error bars are 95% confidence limits from 3 to 6 repeated filter measurements. d Error bars are 95% confidence limits propagated from log-normal fits of the size distributions.

is a very small, but persistent, drop in PM emissions as the sulfur content is increased, which is consistent between the particle sizing instrumentation and the gravimetric PM mass determinations. In contrast, the gaseous emissions steadily increase with higher sulfur content, by 13-30% for hydrocarbons, 1035% for CO, and by a factor of 2-3 for NOx, upon increasing the sulfur level from 35 to 600 ppm. The explanation for this is linked to the sulfur balance recorded for each phase of the test. SO2 represents the major sulfur containing species in the exhaust. As expected and evident from Table 3, its emissions increase with higher fuel sulfur. What is unexpected, however, is the trend with FTP test phase. For the 350 and 600 ppm sulfur fuels, the SO2 emissions are smallest during phase 1, are comparable for phase 2, when emissions are expected to drop due to the lower speeds and loads in this portion of the FTP drive cycle, and then increase substantially in phase 3. Figure 5, which displays the cumulative sulfur emissions recorded along the FTP (using 800 ppm sulfur fuel), helps explain this trend. It occurs because the sulfur measured in the exhaust (SO2, H2S, and SO4, including particulate sulfate) represents approximately 10% of the sulfur consumed by combustion for phase 1, 30% for phase 2, and nearly 100% for phase 3. The remainder of the sulfur is stored on the catalyst and is released during the high accelerations run after the end of the FTP test (these are used to confirm that the dynamometer correctly simulates the vehicle load). The sulfur storage reduces catalyst efficiency and explains the observed increases in hydrocarbon, CO, and NOx emissions with increasing sulfur level. Tailpipe PM levels are unaffected since, as demonstrated above, the catalyst has a minimal effect on PM emissions. Wind Tunnel Measurements. In principle, sulfate can contribute to PM emissions in two ways: It can condense 280

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002

FIGURE 5. Cumulative gaseous sulfur emissions along FTP drive cycle and during subsequent series of high accelerations. Fuel with 800 ppm sulfur was used in this test. onto preexisting soot particles or it can nucleate into an ultrafine aerosol. The latter fate depends very sensitively on the rate of dilution and cooling that the exhaust experiences; thus, conventional dilution tunnel PM emissions measurements, where the dilution ratio varies from ∼100 at idle to < 5 during hard accelerations, were compared to tests conducted in a wind tunnel facility that mimic the more rapid dilution experienced under ambient conditions. Two sets of steady state and transient tests were performed with 30 ppm sulfur fuel, two tests with 260 ppm sulfur fuel, and one set with 990 ppm sulfur fuel. Steady-state PM emissions were barely detectable above the wind tunnel background levels. Typically the background air contained 8 × 103 ((50%) particles/cm3 in a log-normal distribution (sometimes bimodal) with a geometric mean diameter between 70 and 110 nm. In comparison, the vehicle operating at 40 mph, 70 mph, and 70 mph at a 5% grade (at a constant wind speed of 40 mph) added ∼1000 particles/ cm3 at a mean diameter of 50 nm, 1500 particles/cm3 at 50 nm, and 2000 particles/cm3 at 40 nm, respectively, for the 30 ppm sulfur fuel. Within the estimated uncertainty of (50% in particle concentration and (20% in mean diameter, the same PM emissions were measured for the 260 and 990 ppm sulfur fuels. These low PM emissions are consistent with previous work, which concludes that steady-state PM rates from PFI gasoline vehicles are very low (equivalent to