Zervas, Montagne, and Lahaye TECHNICAL PAPER

ISSN 1047-3289 J. Air & Waste Manage. Assoc. 49:1304-1314

Copyright 1999 Air & Waste Management Association

The Influence of Gasoline Formulation on Specific Pollutant Emissions

E. Zervas and X. Montagne Institut Franais Petrole, Rueil Malmaison, France J. Lahaye Institut de Chimie des Surfaces et Interfaces, Mulhouse, France

ABSTRACT Many recent works have dealt with the influence of fuel composition on regulated and specific pollutant emissions from spark ignition engines. While many qualitative correlations have been already proposed, only a few quantitative ones are known (benzene remains an exception). This paper describes qualitative and quantitative correlations between fuel composition and specific pollutant emissions (individual hydrocarbons, aldehydes, ketones, alcohols, and organic acids) of a spark ignition engine. The aim of this work was to find the precursors of the main specific pollutants. Then, for each of them, a multilinear equation has been calculated, illustrating the correlation between its concentration in exhaust gases and its content in the fuel. The results of these calculations point out which initial compound favors the formation of a determined pollutant. As lean conditions are probably going to be used in future commercial engines, the fuel effect has been studied for a broad range of equivalence ratios (from 0.8 to 1.2). Two fuel matrixes were designed. The first one was obtained by introducing eight pure hydrocarbons (nC6, nC8, isoC8, 1-hexene, cyclohexane, toluene, o-xylene, and

IMPLICATIONS The topic of this paper strongly interests both refiners and car manufacturers. Besides being of scientific interest, knowing precisely which compound induces which pollutant, could enable refiners to determine which gasoline formulations provide better control of pollutant emissions, especially toxic compounds and ozone precursors. Car manufacturers could use this information to optimize engine tunings and catalyst use to reduce those emissions. Further, these works are of scientific interest, helping to improve understanding of the possible impact of regulation changes on pollutant emissions, helping refiners to improve fuel formulations, and helping all parties concerned by helping to improve air quality.

ethylbenzene) in an alkylate base. The second one was formulated to test the behavior of four oxygenated compounds (methanol, ethanol, isopropanol, and methyl tertio-butyl ether [MTBE]). Individual unburned hydrocarbons, aldehydes, alcohols, and organic acids were analyzed according to accurate methods developed in our laboratory. Results show that benzene exhaust is mainly emitted from benzene fuel, the substituted aromatics, and, to a lesser degree, from cyclohexane; 1,3-butadiene is produced from 1-hexene or cyclohexane; and isobutene is a product of isooctane and MTBE. Exhaust toluene comes mainly from toluene fuel, but also from o-xylene and ethylbenzene fuels. Isopropylbenzene exhaust seems to be produced only from ethylbenzene fuel. Exhaust concentrations of acetaldehyde and acetone showed a strong correlation with, respectively, ethanol and isopropanol fuel contents. Methacroleine exhaust concentration was strongly correlated with the fuels isooctane content. Benzaldehyde was produced by the aromatic fuels. Methanol was found in the exhaust gases of all the oxygenated fuels. High quantities of hexane or isooctane in fuels seem also to favor methane formation. Ethanol and isopropanol were found in the exhaust gases only when these components are added to the fuel. Propionic acid was produced by the aromatic fuels, while butyric acid came from o-xylene. Based on these results, this paper proposes the probable intermediate paths for formation of some of these pollutants. Finally, the ozone-forming potential of the exhaust gas of each fuel was calculated. Toluene and the oxygenated components of the fuel were shown to decrease this potential. INTRODUCTION For the last several years, the relationship between fuel formulation and pollutants emitted by spark ignition
Volume 49 November 1999

1304 Journal of the Air & Waste Management Association

Zervas, Montagne, and Lahaye

engines has been the focus of many studies, such as those of the Auto/Oil program, which produced a many articles in this field.1-20 The addition of oxygenated compounds such as alcohols or ethers (e.g., MTBE) to a fuel can change its exhausted pollutants. Many qualitative, but few quantitative, correlations have been established. Benzene, due to its major importance, has been studied more than other pollutants. Some authors, such as Kameoka,18 reported a quantitative correlation between some pollutant emissions, as well as total and individual aromatics, and the fuel content in paraffins and olefins. Pollutant formation mechanisms in chemical reactors have also been well studied. Though the modeling of the combustion of different fuel components (hydrocarbons, alcohols, ethers, etc.) and the description of the corresponding mechanisms in reactors are described in the literature,21-27 their validation in running engines has yet to be completed. This paper deals with the topic of qualitative and quantitative correlations between fuel formulation and individual hydrocarbon emissions obtained in a Cooperating Fuel Research (CFR) spark ignition engine. More precisely, it looks at the effects of fuel formulation on the pollutant emissions, qualitatively and quantitatively. Two matrices, one containing hydrocarbons only and one a mixture of hydrocarbons and oxygenated compounds, were used to feed this engine. Specific pollutants (individual hydrocarbons, aldehydes, alcohols, and organic acids) were measured using specific analytical methods. Multilinear models were used to determine the specific impact of each fuel component on each pollutant emission. A broad range of equivalence ratios (from 0.8 to 1.2) was examined and its influence on pollutant emissions analyzed. For some of these pollutants, a formation mechanism, based on the experimental results obtained, is proposed. Finally, the ozone forming-potential of these fuels emissions has been calculated and compared. EXPERIMENTAL Engine and Running Conditions A CFR spark ignition engine was used for our tests. The CFR is a monocylinder engine (displacement = 6.11 10-4 m3, bore = 0.08255 m, stroke = 0.1143 m) used for octane index determination. It is an experimental engine, found in several laboratories, which usually runs under two speeds: 10 and 15 Hz (600 and 900 min-1) and under a large range of easily adjustable air/fuel ratios. All engine parameters were kept constant during the tests (speed = 15 Hz, compression ratio = 6:1, indicated mean effective pressure (IMEP) = 4.5 105 Pa), except the equivalence ratio, which varied from 0.8 to 1.2. We wanted to explore a large range of equivalence ratios, especially under lean conditions, since we believe these will be in the future.
Volume 49 November 1999

The CFR engine is a test engine, having conventional valves, plug location, and head design. The response of commercial engines to fuel formulation may differ from that of the CFR engine. However the CFR engine allows us to determine the most important axes of future research in the fuel formulation field using commercial engines. Fuels Used Two fuel matrices were designed. The first one contained only hydrocarbons and is referred to as the synthetic fuel matrix. The second one also contained oxygenated compounds and is referred to as the oxygenated fuel matrix. In the first matrix, eight hydrocarbons were selected as variables to study their impact on exhaust pollutants. These compounds represent all the chemical families that can be found in commercial fuels: n-hexane and n-octane represent the C6 and C8 paraffins; 2,2,4-trimethylpentane represents the C8 isoparaffins; 1-hexene represents olefins and cyclohexane naphthenes. The three aromatics used were toluene, o-xylene, and ethylbenzene. Low levels of benzene (between 0.53 and 0.66%) were also present in all the tested fuels. An experimental design, specially adapted for the mixtures, is used to determine each component quantity in the blend.28 As the physical properties of each fuel must not be very different one another, an alkylate (containing basically isooctane but also 1.5% benzene) was used as the base fuel for these blends. Fuel R, containing an equal content of each of the eight compounds, was the central point of the experimental matrix and was also used to validate the quantitative models of the relationship between fuel composition and pollutant emissions. The name of each fuel was chosen to recall its major component. The two simplest fuels used are: iC8p, which is pure 2,2,4-trimethylpentane and iC8T, which is a mixture of 80% 2,2,4-trimethylpentane and 20% toluene. The oxygenated matrix was obtained by adding 5 or 20% of an oxygenated compound: methanol, ethanol, isopropanol, or MTBE to fuel R. Four fuels with low and four fuels with high oxygenated contents were studied. The chemical composition of each fuel is detailed in Tables 1 and 2. The selected alkylate is a common product. Octane number and distillation properties are detailed in Tables 3 and 4. Analytical Methods Regulated Pollutants. Carbon monoxide and carbon dioxide were analyzed by nondispersive infrared, nitrogen oxides by chemiluminescence, and total unburned hydrocarbons by a flame ionization detector (FID). Individual Hydrocarbons. Individual hydrocarbons were analyzed online by gas chromatography and FID (GC/FID),
Journal of the Air & Waste Management Association 1305

Zervas, Montagne, and Lahaye

Table 1. First fuel matrix: synthetic fuels (% vol. contents). C6 Hexane 1-Hexene Cyclohexane n-Octane Isooctane Toluene o-Xylene Ethylbenzene Alkylate 42 2 2 2 2 2 2 2 44 C6= 2 42 2 2 2 2 2 2 44 CC6 2 2 42 2 2 2 2 2 44 C8 2 2 2 42 2 2 2 2 44 iC8 2 2 2 2 42 2 2 2 44 T 2 2 2 2 2 42 2 2 44 o-X 2 2 2 2 2 2 42 2 44 ETB 2 2 2 2 2 2 2 42 44 R 7 7 7 7 7 7 7 7 44 iC8p iC8T

100

80 20

Table 2. Second fuel matrix: oxygenated fuels (% vol. contents). M5 Methanol Ethanol Isopropanol MTBE Fuel R Oxygen (% weight) 5 0 0 0 95 3.08 E5 0 5 0 0 95 1.98 IP5 0 0 5 0 95 1.45 MT5 0 0 0 5 95 0.90 M20 20 0 0 0 80 11.89 E20 0 20 0 0 80 7.75 IP20 0 0 20 0 80 5.72 MT20 0 0 0 20 80 3.60

using a 3400 VARIAN gas chromatograph under the following conditions: columnPLOT (Al2O3/KCl), length = 50 m, i.d. = 0.03210-3 m; sample loop210-6 m3 (2 mL); carrier gashelium; makeup510-7 m3sec-1 (30 mL/ min); split6.6710-7 m3sec-1 (40 mL/min); injector temperature473 K (200 C); and detector temperature523 K (250 C). The column temperature program was (1) initial temperature of 353 K (80 C) for 2 min; (2) from 353 K (80 C) to 473 K (200 C) with a 5 K/min heating rate; and (3) final temperature of 473K (200 C) for 40 min. A standard solution of 25 hydrocarbons (methane, ethane, ethylene, propane, propylene, acetylene, isobutane, 1-butene, cis-2-butene, isobutene, trans-2butene, butane, isopentane, pentane, 1,3-butadiene, cyclohexane, hexane, 1-hexene, benzene, isooctane, octane, toluene, ethylbenzene, o-xylene, and propylbenzene) was used for chromatograph calibration and for each hydrocarbon identification. This method allowed a good separation of the C1 to C8 compounds with no interference. Aldehydes and Ketones. Aldehydes and ketones were collected by passing a sample of raw exhaust gas, brought through a heated line, in impingers containing 0.3 10-5 m3 (30 mL) of a 2,4-dinitrophenylhydrazine (DNPH) solution in acetonitrile. This solution was acidified by hydrochloric acid, the impingers were placed in an ice bath to reduce the evaporative losses, and the flow rate was regulated at 1.67 10-5 m3 sec-1 (1 L/min).
1306 Journal of the Air & Waste Management Association

After collection, the samples are analyzed by highperformance liquid chromatography according to the technique used by IFP. A VARIAN 5000 liquid chromatograph was used under these analytical conditions: columnZORBAX ODS, length = 0.25 m, i.d. = 4.6 10-3 m; detectionUV at 365 nm; solvent65% vol. acetonitrile/ 35% vol. water; and flow rate1.67 10-8 m3sec-1 (1.0 mL/ min) for 22 min. A standard solution containing 12 aldehydes (formaldehyde, acetaldehyde, acroleine, acetone, propionaldehyde, crotonaldehyde, methacroleine, methylethylketone + butyraldehyde, benzaldehyde, o-toluenealdehyde, m-toluenealdehyde, and p-toluenealdehyde) was used for chromatograph calibration and identification of each aldehyde. The detection limits of this method are <5 ppb in the solution. Nitrogen dioxide can react with the DNPH solution during the aldehyde collection, 29,30 and the products formed can interfere with the formaldehyde peak. The analytical method used enabled us to separate these two products. The other aldehydes determined were acetaldehyde, acroleine, acetone, propionaldehyde, crotonal-dehyde, methacroleine, and benzaldehyde. These products did not interfere with the other exhaust gases components. The column used cannot separate methylethylketone and n-butyraldehyde. Organic acids and alcohols do not react with the DNPH solution.31
Volume 49 November 1999

Zervas, Montagne, and Lahaye

Table 3. Distillation characteristics, in K (C) and octane numbers of the synthetic fuels. C6 IBP 10 50 90 Endpoint RON MON 321 (48) 345.5 (72.5) 359 (86) 386.5 (113.5) 427.5 (154.5) 63.7 61.7 C6= 321.5 (48.5) 341.5 (68.5) 354 (81) 384.5 (111.5) 414 (141) 83.7 77.0 CC6 327 (54) 355.5 (82.5) 364 (91) 386 (113) 429 (156) 88.7 79.0 C8 318.5 (45.5) 373 (100) 387.5 (114.5) 398.5 (125.5) 417.5 (144.5) 43.6 44.4 iC8 320 (47) 364.5 (91.5) 374 (101) 386.5 (113.5) 420.5 (147.5) 93.8 91.4 T 319 (46) 368.5 (95.5) 378.5 (105.5) 388 (115) 425.5 (152.5) 101.3 90.3 o-X 318 (45) 374.5 (101.5) 396.5 (123.5) 415.5 (142.5) 430 (157) 96.6 86.4 ETB 317 (44) 375.5 (102.5) 394.5 (121.5) 411 138) 425 (152) 100 89.6 R 325.5 (52.5) 358 (85) 376.5 (103.5) 403.5 (130.5) 426.5 (153.5) 85.2 80.4

Table 4. Distillation characteristics, in K (C) and octane numbers of the oxygenated fuels. M5 IBP 10 50 90 End point RON MON 321.5 (48.5) 328.5 (55.5) 376 (103) 402 (129) 427 (154) 89.4 82.5 E5 325.5 (52.5) 341 (68) 376.5 (103.5) 402.5 (129.5) 426.4 (153.5) 90.4 83.3 IP5 325.5 (52.5) 347.5 (74.5) 375.5 (102.5) 402.5 (129.5) 425 (152) 87.5 82.3 MT5 316 (43) 354 (81) 375.5 (102.5) 399 (126) 431.5 (158.5) 86.8 81.7 M20 322.5 (49.5) 329.5 (56.6) 363 (90) 399.5 (126.5) 425 (152) 98.6 86.3 E20 329 (56) 340 (67) 346.5 (73.5) 399.5 (126.5) 425.5 (152.5) 97.4 87.1 IP20 327 (54) 344.5 (71.5) 354 (81) 399.5 (126.5) 424.5 (151.5) 96.6 87.3 MT20 321 (48) 344.5 (71.5) 367 (94) 399 (126) 426 (153) 93.1 86.4

Alcohols. Alcohol compounds were collected by passing a sample of raw exhaust gas in two impingers containing 0.2 10 -4 m 3 (20 mL) deionized water. For aldehyde collection, the impingers were placed in an ice bath. The flow rate was regulated at 1.67 10-5 m3sec-1 (1 L/min). The analysis of the final solution was performed by GC/FID, using a VARIAN 3300 gas chromatograph under the following conditions: columnCP52 CB, length = 25 m, i.d. = 0.32 10-3 m; carrier gashelium; makeup5 10-4 m3sec-1 (30 mL/min); split8.34 10-5 m3sec-1 (5 mL/min); injector temperature393 K (120 C); detector temperature523 K (250 C); column temperature353 K (80 C); and injected solution volume1 L. A standard solution of six alcohols (methanol, ethanol, n-propanol, isopropanol, n-butanol, and 2-butanol) was used for chromatograph calibration and identification of each alcohol. The detection limits of this method are under 0.5 ppm in the solution.31 Methanol, ethanol, and isopropanol are well separated under the analytical conditions selected. N-propanol, n-butanol, and 2-butanol were used for the chromatograph calibration but were not detected in the exhaust gases. Laboratory studies have shown that, with a 353 K (80 C) temperature column, methanol can interfere with butyraldehyde.31 These two compounds can be completely separated at 333 K (60 C). Apparently, butyraldehyde does not dissolve in water during collection, while several hydrocarbons,
Volume 49 November 1999

aldehydes, and organic acids do dissolve but do not interfere with the peaks of the above three alcohols. Organic Acids. Organic acids were collected together with alcohols in two impingers containing 0.2 10-4 m3 (20 mL) deionized water. The final solution was first analyzed by ionic chromatography. No chemical interfered with the formic acid determination, but the acetic acid could not be separated from propionic and isobutyric acids under our experimental conditions. These heavier acids can be analyzed by gas chromatography, but this method cannot detect formic acid which has weak FID response.32 Therefore, the final solution had to be analyzed by two methods: ionic chromatography for formic acid determination, and gas chromatography for the other acids determination. For ionic chromatography, a DIONEX Series 2000i ionic chromatograph was used under the following conditions: columnIonPac AS4; carrier liquidNa2B4O7 0.5 10-3 M aqueous solution; flow rate3.33 10-8 m3sec-1 (2 mL/ min); and regenerateH2SO4 0.013 M aqueous solution. For gas chromatography, a VARIAN 3300 gas chromatograph was used under the following conditions: column HP Megabore Innowax, 30m 0.53 10-3 m; detectionFID; carrier gashelium; makeup5 10-7 m3sec-1 (30 mL/min); split8.33 10-8 m3sec-1 (5 mL/min); injector temperature 523 K (250 C); detector temperature523 K (250 C); and injected solution volume1 10-9 m3 (1 mL).
Journal of the Air & Waste Management Association 1307

Zervas, Montagne, and Lahaye

The column temperature program was (1) initial temperature at 393 K (120 C) for 2 min; (2) from 393 K (120 C) to 473 K (200 C) with a 4 K/min heating speed; and (3) final temperature of 473 K (200 C). More details about this analysis can be found in the literature.33 A standard solution containing 12 organic acids (formic, acetic, propionic, acrylic, isobutyric, butyric, isovaleric, valeric, isocapric, capric, heptanoic, and benzoic acids) was used for chromatograph calibration and identification of each acid. The detection limits (in the solution) of this method were <10 ppb for formic acid and <0.5 ppm for the other acids.31, 33 Formic acid showed no interference with other organic acids nor with other ions such as Cl-, NO3-, PO43-, or SO42- during the ionic chromatography analysis. The acids analyzed by GC showed no interference with the other exhaust gas products dissolved in water. A number of unknown peaks appeared during the analysis; they may have been due to heavy hydrocarbons or isomers of heavier organic acids. Aldehydes and alcohols had a very low retention time compared to the first eluted acid, which was acetic acid. Data Analysis This work focused on the precursors of each specific pollutant. Correlations between exhaust pollutant concentrations and fuel composition were established to determine which component induced the formation of which pollutant. The shape of this quantitative correlation is (1) where Cgas is the pollutant concentration in exhaust gas, Ci,fuel is the content of each fuel component, and ai is its weight factor. After the determination of weight factors for the five equivalence ratios studied, the model was validated in comparison with the reference fuel R. Statgraphics statistical software (Manugistics, Rockville, MD) was used for these calculations. The coefficients for the very low concentrated pollutants (<1 ppm), were not estimated because of the high relative error of such calculation. RESULTS AND DISCUSSION Repeatability Tests The relative standard deviation (RSD) of the individual hydrocarbons analysis was calculated for five similar tests using fuel R. The RFDs of most individual hydrocarbons were under 12%, although one exceeded 20%. The repeatability of aldehyde, organic acid, and methanol measurements (fuel R does not emit other alcohols) were under 12% for most of these components.
1308 Journal of the Air & Waste Management Association

The RSD values for some equivalence ratios could reach 15% (acroleine), 20% (acetone, methacroleine, acetic, and propionic acids), or 30% (butyric acid). Qualitative Correlation Before establishing the quantitative models describing the interaction between fuel composition and pollutant emissions, qualitative correlations were looked at. These correlations show which of the fuel components favors the emission of a pollutant. The results (by pollutant) follow. Hydrocarbons. It was shown that (a) aromatics decrease the concentration of ethylene, which is mainly produced from alkanes, ethanol, and isopropanol; (b) toluene does not favor the formation of acetylene; (c) aromatics decrease the concentration of propylene, which is mainly formed from hexane, 1-hexene, octane, isooctane, and isopropanol; isobutane and isobutene are mainly produced from isooctane and MTBE; (e) 1-butene is mainly produced from hexane, 1-hexene, octane, and isooctane; (f) 2-butene (cis and trans) formation is favored from oxylene, 2-trans-butene is not produced from isooctane and both from toluene; (g) 1,3-butadiene is mainly produced from 1-hexene and cyclohexane; it is not produced from hexane; (h) isopentane is mainly produced from isooctane; (i)1-hexene is mainly produced from 1-hexene and cyclohexane; (j) hexane, cyclohexane, octane, isooctane, o-xylene, and ethylbenzene are produced only from their respective fuel components (small quantities of the last may also be produced from toluene); (k) benzene is mainly produced from aromatics, but cyclohexane also contribute to its formation; (l) toluene is produced from fuel aromatics; and (m) isopropylbenzene is produced from fuel ethylbenzene (small quantities may also be produced from toluene). For the oxygenated pollutants, we noted the following: Aldehydes. We found that (a) formaldehyde formation is favored from oxygenated fuel components, but also from octane and isooctane; (b) acetaldehyde is mainly produced from ethanol, but also from isopropanol, hexane, 1-hexene, and octane; (c) acroleine comes mainly from 1-hexene and octane; (d) acetone comes mainly from isopropanol, but hexane, octane, and isooctane also contribute to its formation; (e) propionaldehyde comes mainly from hexane and octane; (f) crotonaldehyde formation is not favored by aromatics and oxygenates, but by all the other fuel components used; (g) methacroleine is mainly produced from isooctane, but MTBE may contribute to its formation; and (h) benzaldehyde comes from fuel aromatics. Alcohols. We found that (a) methanol is mainly produced from methanol fuel, but it is also favored by the other
Volume 49 November 1999

Zervas, Montagne, and Lahaye

Table 5. Normalized weight of the different fuel components toward pollutants emission. Pollutant Isobutene =0.9 =1.0 =1.1 1-Butene =0.9 =1.0 =1.1 1, 3-Butadiene =0.9 =1.0 =1.1 1-Hexene =0.9 =1.0 =1.1 Benzene =0.9 =1.0 =1.1 Toluene =0.9 =1.0 =1.1 Benzaldehyde =0.9 =1.0 =1.1 Propionic Acid =0.9 =1.0 =1.1 Isooctane 1 1 1 Hexane 1.37 3.58 3.62 1-Hexene 1 1 1 1-Hexene 2.78 2.72 2.64 Benzene 17.48 37.72 56.72 Toluene 24.8 8.54 4.10 Benzene 1 1 1 Benzene 15.52 10.23 8.30 MTBE 1.39 1.68 1.26 1-Hexene 0.30 2.57 2.58 Cyclohexane 1.04 1.87 2.31 Cyclohexane 1 1 1 Toluene 0.8 2.27 5.20 o-Xylene 3.47 3.35 1.65 Toluene 14.04 7.00 5.10 Toluene 10.91 3.36 3.14 Fuel Component

Octane 1.34 3.89 4.26

Isooctane 1 1 1

o-Xylene 0.76 2.19 4.10 Ethylbenzene 1 1 1 o-Xylene 0.30 3.11 0.19 o-Xylene 1 1 1

Ethylbenzene 0.7 3.01 6.36

Cyclohexane 1 1 1

Ethylbenzene 1.09 5.64 4.41 Ethylbenzene 9.05 2.99 4.47

fuel oxygenate components, hexane, and isooctane; and (b) ethanol and isopropanol come only from their respective fuel components. Organic Acids. We found that (a) formic acid is favored by oxygenates, octane, toluene, and ethylbenzene; (b) acetic acid is favored by ethanol and isopropanol; (c) propionic acid is favored from fuel aromatics; (d) butyric acid is mainly produced from o -xylene; and (e) isovaleric acid is mainly produced from o -xylene and ethylbenzene. Quantitative Correlation The coefficients of the multilinear models were estimated for the pollutants with concentrations >1 ppm in the exhaust gases. The normalized parameters of these models are presented in Table 5, and the percentage relative errors of their calculation (as given by Statgraphics) are shown in Table 6. These results are discussed in detail for three different equivalence ratios (0.9, 1.0, and 1.1).
Volume 49 November 1999

Hydrocarbons. (a) Isobutene is a product of isooctane and MTBE according to the normalized equation at stoichiometry Isobutene = Isooctane + 1.7 MTBE (2)

At stoichiometry, the MTBE coefficient reached its maximum, while it decreased to .39 and 1.26 for fuel equivalence ratios of 0.9 and 1.1, respectively. Isooctane parameters had very low relative errors (23%), while those of MTBE were greater, starting from 13% for the equivalence ratio of 0.9 and reaching 23% at 1.1. (b) 1-butene was emitted from n-hexane, 1-hexene, n-octane, and isooctane. At stoichiometry, the normalized equation is 1-Butene = 3.6 Hexane + 2.6 1-Hexene + 3.9 Octane + Isooctane

(3)

The coefficients of the first three components increased with the fuel equivalence ratio (1.37 and 3.62 for
Journal of the Air & Waste Management Association 1309

Zervas, Montagne, and Lahaye

Table 6. Errors (%) of calculation of the normalized weight of the different fuel components toward pollutants emission. Pollutant Isobutene F=0.9 F=1.0 F=1.1 1-Butene F=0.9 F=1.0 F=1.1 1,3-Butadiene F=0.9 F=1.0 F=1.1 1-Hexene F=0.9 F=1.0 F=1.1 Benzene F=0.9 F=1.0 F=1.1 Toluene F=0.9 F=1.0 F=1.1 Benzaldehyde F=0.9 F=1.0 F=1.1 Propionic Acid F=0.9 F=1.0 F=1.1 Isooctane 2.17 3.35 3.47 Hexane 6.71 13.55 11.48 Hexane 20.66 24.39 23.75 1-Hexene 6.62 7.22 5.89 Benzene 18.69 36.50 29.59 Toluene 4.27 5.31 3.49 Benzene 96.15 58.14 38.94 Benzene 75.76 37.04 57.80 MTBE 13.23 16.84 23.36 1-Hexene 30.96 18.87 16.10 Cyclohexane 20.00 13.07 10.31 Cyclohexane 18.22 19.67 15.58 Toluene 21.74 26.81 14.25 o-Xylene 33.33 14.81 11.24 Toluene 3.70 4.94 3.87 Toluene 5.61 5.89 7.92 Fuel Component

Octane 7.01 12.50 9.73

Isooctane 26.39 13.68 11.68

o-Xylene 21.74 24.69 20.62 Ethylbenzene 115.23 49.61 12.50 o-Xylene 19.65 12.66 11.88 o-Xylene 69.44 22.52 28.33

Ethylbenzene 23.31 75.76 13.28

Cyclohexane 16.40 69.44 84.75

Ethylbenzene 5.41 7.41 5.09 Ethylbenzene 7.67 7.52 6.34

hexane, 0.30 and 2.58 for 1-hexene, 1.34 and 4.26 for octane, respectively). The percentage of 1-butene from each fuel component (for equal fuel content for the four components) was as follows: the percentage from hexane diminishes slightly with the equivalence ratio (from 34.16 for =0.9, to 32.44 for =1.0, and 31.59 for =1.1); the percentage that came from isooctane diminished sharply in the beginning and stayed almost constant for equivalence ratios greater than 1.0 (24.94, 9.06, and 8.73); the percentage from octane increased slightly (33.42, 35.20, and 37.17); and the percentage from 1-hexene presented its maximum value at stoichiometry (7.48, 23.51, and 22.51). The errors of the estimation of these parameters are as follows: the parameters for hexane and octane were estimated quite well, with errors from 6.7 to 13.5%, while the parameters for 1-hexene and isooctane ranged from 11.7 to 30.9%. (c) 1,3-Butadiene was generated from 1-hexene and cyclohexane, according the equation (at stoichiometry)
1310 Journal of the Air & Waste Management Association

1,3-Butadiene = 1-Hexene + 1.9 Cyclohexane

(4)

The second coefficient increased with the equivalence ratio (from 1.04 to 2.31, for the range presented here), indicating that cyclohexane favors the formation of 1,3butadiene more at higher equivalence ratios. The errors of estimation of hexane parameters were higher (20.624.3) than those of cyclohexane (10.3 20.0%). (d) 1-Hexene came from the 1-hexene and cyclohexane fuel according to the equation (at stoichiometry) 1-Hexene = 2.7 1-Hexene + Cyclohexane (5)

These coefficients remained almost stable (with differences of less than 1%) when the equivalence ratio ranged from 0.9 to 1.1. The errors of estimation were quite low in the case of 1-hexene (from 5.8 to 7.2%), higher in the case of cyclohexane (from 15.5 to 19.6%). For both fuel components, these errors were higher in the stoichiometry.
Volume 49 November 1999

Zervas, Montagne, and Lahaye

(e) Benzene was emitted from the aromatic fuels (benzene, toluene, o-xylene, and ethylbenzene) and from cyclohexane. The following equation was obtained at stoichiometry: Benzene = 37.7 Benzene + 3.0 Ethylbenzene + 2.3 Toluene + 2.2 o-Xylene + Cyclohexane ethanol and the isopropanol content of the fuel. However, these pollutants are also produced by the synthetic fuels, so we couldnt establish a quantitative correlation. (b) Methacroleine showed a good correlation with the fuel isooctane content. According to the literature,9,34 both isooctane and MTBE produce the isobutyl radical which forms methacroleine. In the present case, the addition of MTBE did not strengthen the correlation. (c) Benzaldehyde was produced by the aromatics fuel according to the equation Benzaldehyde = 7.0 Toluene + 5.6 Ethylbenzene + 3.1 o-Xylene + Benzene (8) The toluene weight factor decreased sharply with the equivalence ratio, while the ethylbenzene and xylene weight factors reached their maximum at stoichiometry. Pure isooctane did not produce benzaldehyde. The parameters of toluene and ethylbenzene were calculated quite well (with errors from 3.7 to 7.4%), those of o-xylene less well (11.819.6%), and those of benzene quite poorly (38.996.1%, probably because of its low content in the fuels used). (d) Ethanol and isopropanol were found only when they were added to the fuel (the errors were less than 5%). (e) Methanol was produced by all the oxygenated fuels. This pollutant is also produced by synthetic fuels. The two compounds that promote its formation are nhexane and isooctane. We tried establish a model containing methanol, MTBE, n-hexane, and isooctane, but it fits well only under lean conditions. (f) Propionic acid was produced by the four aromatics fuels according to the equation (at stoichiometry) Propionic Acid = 10.2 Benzene + 3.4 Toluene + 3.0 Ethylbenzene + o-Xylene (9) Benzene and toluene normalized weight factors decreased with the equivalence ratio, while those of ethylbenzene presented a minimum at stoichiometry. Isooctane did not produce propionic acid. The relative errors of these parameter estimations were weak in the cases of toluene and ethylbenzene (5.67.9%), and less in the case of o-xylene and benzene (22.569.4% and 37.0 75.7, respectively ). The order of these errors follows that of benzaldehyde model. (g) Butyric acid was emitted only from o-xylene (errors less than 5%). Validation of the Correlation Two methods were used to validate the correlations. The first one was to compare the emitted pollutant produced by the reference fuel R with those predicted by
Journal of the Air & Waste Management Association 1311

(6)

The weight factor of benzene was 12.5, 16.6, and 17.2 times higher than, respectively, the weight of ethylbenzene, toluene, and o-xylene. Nevertheless, the benzene content of gasolines is under 5% today and is expected to be about 2% around the year 2000. Therefore, the benzene fuel contribution to exhaust benzene will be the same as other aromatics contributions (ethylbenzene, toluene, and o-xylene). The specific weight of heavier aromatics increased with the equivalence ratio; for equivalence ratios from 0.9 to 1.1, an increase in the normalized parameters values from 0.8 to 5.2 for toluene, from 0.7 to 4.1 for o-xylene, and from 0.7 to 6.3 for ethylbenzene was observed. Cyclohexane also produced small quantities of benzene, its specific weight decreasing with the equivalence ratio. Most of the errors of estimation of these parameters are significant. The lowest errors were observed in the case of toluene (14.226.8%), followed by o-xylene (20.6 24.7%), benzene (18.736.5%, probably because of the low fuel content), ethylbenzene (13.275.7%), and cyclohexane (16.484.75%, probably because of its small contribution to the model). (f) Exhaust toluene came from toluene fuel, o-xylene, and ethylbenzene. At stoichiometry, the equation was Toluene = 8.5 Toluene + 3.35 o-Xylene + Ethylbenzene (7) The results indicate that the specific impact of the first two aromatics decreased with the equivalence ratio, while that of the third increased. Errors of estimation for toluene parameters were very low (3.45.3%), greater for o-xylene (11.233.3%), and significant in the case of ethylbenzene (12.5-115.2%). (g) Exhaust n-hexane, cyclohexane, n-octane, isooctane, o-xylene, and ethylbenzene were coming exclusively from unburned fuel components (the errors of these estimations were less than 5%). (h) Exhaust isopropylbenzene was produced only by ethylbenzene fuel components (once more, the errors of these estimations were less than 5%). Oxygenated Pollutants. (a) The acetaldehyde and acetone production was well correlated with, respectively, the
Volume 49 November 1999

Zervas, Montagne, and Lahaye

the calculations (Figure 1 illustrates this comparison for benzene). In this validation, and for each equivalence ratio, the emitted pollutant concentration and that predicted from the model using the fuel R are compared. Three values are presented for each concentration, the average, minimum, and maximum concentrations, as calculated from the RSD values determined by repeatability tests in the case of the measured concentrations, and by the relative errors of each model parameter in the case of the predicted ones. These ranges were calculated as follows: In the case of the measured concentration, we apply the formula x RSD t x x + RSD t (10) If the average values of the predicted concentrations are between the minimum and maximum range of the measured ones, the model is considered to be valid. All models present here passed this test. The second method of validating the correlations was to compare all the measured and predicted pollutant concentrations, charting all the measured and predicted concentrations for all fuels and all experimental points. In a good model, the measured and predicted concentrations will be the same and will form the line y = x. The dispersion of the points can suggest the validity of the model. The best line, y = ax+b, passed from these points can suggest the existence of another source (in the case of b>>0). This method of validation is presented in Figure 2 for the case of propionic acid. Table 7 shows the statistical characteristics of these models (r2, a, and b). We can note the high linearity of the points (straight lines with r2 > 0.9). In the case of pollutants emitted from only one fuel component (hexane, cyclohexane, octane, isooctane, o-xylene, ethylbenzene, isopropylbenzene, ethanol, isopropanol, and butyric acid), r2 lies above 0.95. The value of a is very near to 1, while the value of b is very small compared to the pollutant concentration range (the ratio b/concentration range is less than 3%), indicating no other source of these pollutants. Intermediate Species in the Formation of Pollutants From the previous correlations, some assumptions were made as to the reactions responsible for the formation of pollutants. Of course, it is necessary to examine them.

where x is the average concentration measured, RSD is the relative standard deviation, t is the Student distribution value for the desired confidence region (95% in our case), and mx is the pollutant concentration. The minimum and maximum concentrations predicted by the model are Cmin=ai,min*Ti and Cmax=ai,max*Ti (12) (11)

where Cmin and Cmax are the minimum and maximum concentrations, ai is the minimum and maximum coefficient values, and Ti is the fuel content.

Figure 1. Comparison of the measured concentration of benzene emitted from fuel R and that predicted from the best model for the five equivalence ratios studied.
1312 Journal of the Air & Waste Management Association

Concentration (ppm)

Figure 2. Validation of the propionic acid model.

Volume 49 November 1999

Zervas, Montagne, and Lahaye

Table 7. Statistical parameters of the model validations. Pollutant Isobutene 1-Butene 1,3-Butadiene Benzene Toluene Methacroleine Benzaldehyde Propionic Acid r2 0.963 0.943 0.912 0.951 0.976 0.898 0.972 0.975 a 1.04 0.97 1.03 0.96 0.99 1.05 1.00 0.97 b -1.07 0.02 -0.73 0.23 -0.14 -0.05 -0.01 0.66

42% toluene), with o-X being more significant, followed by C8, C6=, and iC8. Also, the addition of 20% toluene in isooctane (fuel iC8T) showed a decrease of this potential compared to the pure isooctane fuel (iC8p). Every fuel studied, except T, had an ozone-forming potential more significant than that of the reference fuel R. The addition of oxygenated compounds to the reference fuel R increased the ozone-forming potential, which was more significant at higher oxygenate contents. These results are in disagreement with those reported in the literature.37, 38

1,3-butadiene was emitted from 1-hexene and cyclohexane. 1-Hexene could be converted into 3-hexenyl radicals which, undergoing a -scission, produces ethylene and CH2 = CHCH2CH2 radicals, which form the 1,3-butadiene by losing a hydrogen atom. This pollutant could also come from the cyclohexane via an hexenyl radical (see formation of 1-hexene from cyclohexane). The cyclohexane fuel induced 1-hexene formation. The first step of the reaction is expected to be a hydrogen atom loss. Then, the cycle opens, giving a 6-hexenyl radical (CH2 = CHCH3CH2CH2CH2), which is converted into 1-hexene by the uptake of one hydrogen atom. The hexane does not produce any 1-hexene because the hexyl radical is rather decomposed in smaller molecules. Isopropylbenzene was produced only from the ethylbenzene fuel. The ethylbenzene loses a hydrogen atom and fits a methyl radical (probably during the cooling phase) to form isopropylbenzene. The formation of propylbenzene in the exhaust was not observed; it is therefore concluded that the extraction of the a hydrogen from the ethylbenzene is not favored under these conditions. The formation mechanism of the organic acids is expected to be quite complex. Aromatic hydrocarbons could first turn into the corresponding phenols, and then the ring could break into two species, producing organic acids. Ozone-Forming Potential These pollutants emitted into the atmosphere can form ozone. Literature reports, for every compound, a reactivity factor for its ability to form ozone.35 After these factors, the ozone-forming potential of each fuel was calculated at stoichiometry (Figures 3 and 4). Ozone-forming potential increases with the equivalence ratio for every fuel used, because exhaust gases contain more hydrocarbons at high equivalence ratios. Literature reports that aromatics, followed by olefins, have the most significant ozone-forming potential.36 Our results show that, at stoichiometry, the fuel with the least significant ozone-forming potential is fuel T (containing
Volume 49 November 1999

Figure 3. Quantity of ozone that could be formed from exhaust gases of synthetic fuels at stoichiometry.

Figure 4. Quantity of ozone that could be formed from exhaust gases of oxygenated fuels at stoichiometry.
Journal of the Air & Waste Management Association 1313

Zervas, Montagne, and Lahaye

CONCLUSIONS Based on relevant analytical methods capable of detecting individual hydrocarbons, aldehydes, organic acids, and alcohols in exhaust gases, this study has confirmed that gasoline composition has an impact on the nature of the pollutant emissions and, especially, the engine out emissions of spark-ignited engines. These experiments were carried out at steady state with synthetic fuels. Two matrices were used: one with eight pure hydrocarbons and an alkylate, the other with four oxygenated compounds incorporated in a reference fuel. The main conclusions of this study are Exhaust benzene is emitted from fuel aromatic and cyclohexane fuels; exhaust 1,3-butadiene is emitted from 1-hexene and cyclohexane fuels; exhaust isobutene is emitted from isooctane and MTBE fuels; exhaust 1-butene is emitted from hexane, 1-hexene, octane, and isooctane fuels; exhaust 1-hexene is emitted from 1-hexene and cyclohexane fuels; exhaust toluene is emitted from toluene, o-xylene, and ethylbenzene fuels; and isopropylbenzene comes from ethylbenzene. Methacroleine comes from isooctane; benzaldehyde and propionic acid are produced from the fuel aromatics; and butyric acid comes from o-xylene. Ethanol and isopropanol were found in exhaust gases only when they had been added to the fuel, while methanol was produced from all four oxygenated compounds tested and from n-hexane and isooctane. Based on these results, possible formation mechanisms of some of these pollutants in an internal combustion engine are proposed, although the assumptions would require examination. The influence of the equivalence ratio on emissions has been studied within a 0.8 to 1.2 range. The behavior of the oxygenated pollutants (aldehydes, alcohols, and organic acids) depends on the fuel used. Toluene decreased ozone-forming potential, while the addition of oxygenated compounds to a fuel blend increased it. REFERENCES
1. 2. 3. 4. 5. 6. 7. Williams, R.L.; Lipari, F.; Potter, R.A. J. Air & Waste Manage. Assoc. 1990, 40, 747-756. Hochhauser, A.M.; Benson, J.D.; Burns, V.R.; Gorse, R.A.; Koehl, W.J.; Painter, L.J.; Rippon, B.H.; Reuter, R.M.; Rutherford, J.A. SAE Technical Paper, 1991, Series 912322. Kivi, J.; Niemi, A.; Nylund, N.O.; Kuto, M.; Orre, K.; SAE Technical Paper, 1992, Series 922379. Jensen, T.E.; Siegel, W.O.; Richert, J.F.O.; Lipari F.; Loo, J.F.; Prostak, A.; Sigsby, J.E. SAE Technical Paper, 1992, Series 920320. Hamai, K.; Mitsumoto, H.; Iwakiri, Y.; Ishihara, K.; Ishii, M. SAE Technical Paper, 1992, Series 922380. Hirota, T.; Kojima, K.; Yakushiji, K.; Inoue, T. SAE Technical Paper, 1992, Series 922180. Reuter, R.M.; Benson, J.D.; Burns, V.; Gorse, R.; Hauchhauser, A.; Koehl, W.; Painter, L.; Rippon, B.; Rutherford, J. SAE Technical Paper, 1992, Series 920326. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. Leppard, W.R.; Rapp, L.A.; Burns, V.R.; Gorse, R.A.; Knepper, J.C.; Koehl, W.J. SAE Technical Paper, 1992, Series 920329. Petit, A.; Montagne, X. SAE Technical Paper, 1993, Series 932681. Siegel, W.O.; Richert, J.F.O.; Jensen, T.E.; Schuetzle, D.; Swarin, S.J.; Loo, J.F.; Prostak, A.; Nagy, D.; Schlenker, A.M.; SAE Technical Paper, 1993, Series 930142. DePetris, C.; Giglio, V.; Police, G.; Pratti, M.V.; SAE Technical Paper, 1993, Series 932679. Der Otter, G.J.; Malpas, R.E.; Morgan, T.D.B. SAE Technical Paper, 1993, Series 930372. Jeffrey, J.G.; Elliot, N.G. SAE Technical Paper, 1993, Series 932680. Morgan, T.D.B.; Der Otter, G.J.; Lange, W.W.; Doyon, J.; Barnes, J.R.; Yamashita, T.; SAE Technical Paper, 1993, Series 932678. Lange, W.W.; Muler, A.; McArragher, J.S.; Schafer, V. SAE Technical Paper, 1994, Series 941867. McDonald, C.R.; Shore, P.R.; Lee, G.R.; Der Otter, G.J.; Humphries, D.T. SAE Technical Paper, 1994, Series 941868. Neimark, A.; Kholmer, V.; Sher, E. SAE Technical Paper, 1994, Series 940311. Kameoka, A.; Akiyama, K.; Hosoi, K. SAE Technical Paper 1994, Series 941866. Stump, F.D.; Knapp, K.T.; Ray, W.D.; Siudak, P.D.; Snow, R.F. J. Air & Waste Manage. Assoc. 1994, 44, 781-786. Harpster, M.O.; Matas, S.E.; Fry, J.H.; Litzinger, T.A. SAE Technical Paper, 1995, Series 950740. Warnatz, J. 20th Symposium on Combustion, 1984, 845-856. Norton, T.S.; Dryer, F.L. 23rd Symposium on Combustion, 1990, 179-185. Chakir, A.; Bellimam, M.; Boettner, J.C.; Cathonnet, M. Inter. J. Chemi. Kinetics. 1992, 24, 385-410. Curran, H.J.; Dunphy, M.P.; Simmie, J.M.; Westbrook, C.K.; Pitz, W.J. 24th Symposium on Combustion, 1992, 769-776. Egolfopoulos, F.N.; Du, D.X.; Law, C.K. Comb. Sci. and Tech. 1992, 83, 33-75. Egolfopoulos, F.N.; Du, D.X.; Law, C.K. 24th Symposium on Combustion, 1992, 883-841. Zhang, H.Y.; McKinnon, J.T. Comb. Sci. and Tech.1995, 107, 261-300. Sado, G.; Sado, M.C. Les Plans dExpriences, ANFOR Ed., Paris, 1991. Gromping, A.; Karst, U.; Cammann, K. J. Chrom. A, 1993, 653, 341-347. Karst, U.; Binding, N.; Cammann, K.; Witting, U. J. Anal. Chem. 1993, 345, 48-52. Zervas, E. Ph.D. Thesis, University of Mulhouse, France, June 1996. Dietz, WA, J. Gas Chromatography 1967, 5:2, 68-71. Zervas, E.; Montagne, X; Lahaye, J. Collection and analysis of gas phase organic acids. A comparison of different methods, Atmos. Environ. in press. Bower, S.L.; Litzinger, T.A.; Frottier, V. SAE Technical Paper, 1993, Series 932707. Hamai, K.; Mitsumoto, H.; Iwakiri, Y.; Ishihara, K.; Ishii, M. SAE Technical Paper, 1992, Series 922380. Leppard, W.R.; Rapp, L.A.; Burns, V.R.; Gorse, R.A.; Knepper, J.C.; Koehl, W.J. SAE Technical Paper, 1992, Series 920329. Hirota, T.; Kojima, K.; Yakushiji, K.; Inoue, T. SAE Technical Paper, 1992, Series 922180. Noorman, M.T. SAE Technical Paper, 1993, Series 932668.

About the authors Dr. E. Zervas is a graduate of Athens Polytechnic School. He prepared a Ph.D. thesis at IFP in Dr. X. Montagnes team (Rueil-Malmaison, France). His research subject was spark ignition engine pollutant emissions. He now works at the Ecole des Mines in Nantes, France, where he continues to research this topic. Dr. X. Montagne is a manager in the Fuel, Lubricant and Emissions Department at IFP, 1-4, Avenue de Bois Prau, F92852 Rueil-Malmaison Cedex. An important part of his research deals with the impacts of conventional (gasoline and diesel fuel) and alternative fuels (LPG, NGV, oxygenated compounds, etc.) on emissions and engine running. Dr. J. Lahaye is a professor at Institut de Chimie des Surfaces et Interfaces in Mulhouse, France. He is also research manager at CNRS. He has great expertise in the area of combustion and pollutant formation.

1314 Journal of the Air & Waste Management Association

Volume 49 November 1999