1

2007-01-0472

Particulate and Hydrocarbon Emissions from a Spray Guided Direct Injection Spark Ignition Engine with Oxygenate Fuel

Blends

Philip Price, Ben Twiney and Richard Stone University of Oxford

Kenneth Kar University of Auckland

Harold Walmsley Shell Global Solutions

Copyright © 2006 SAE International

ABSTRACT

The blending of oxygenated compounds with gasoline is projected to increase because oxygenate fuels can be produced renewably, and because their high octane rating allows them to be used in substitution of the aromatic fraction in gasoline. Blending oxygenates with gasoline changes the fuels’ properties and can have a profound affect on the distillation curve, both of which are known to affect engine-out emissions.

In this work, the effect of blending methanol and ethanol with gasoline on unburned hydrocarbon and particulate emissions is experimentally determined in a spray guided direct injection engine. Particulate number concentration and size distribution were measured using a Cambustion DMS500. These data are presented for different air fuel ratios, loads, ignition timings and injection timings. In addition, the ASTM D86 distillation curve was modeled using the binary activity coefficients method for the fuel blends used in the experiments.

In general, unburned hydrocarbon emissions were reduced at low load but increased at high load for the alcohol blends. The effect on particulate emissions was dependent on the operating point: for rich mixtures the accumulation mode number concentration and count median diameter were reduced with the oxygenate blends. However, blending gasoline with oxygenates also caused the nucleation mode number concentration to increase, particularly for M85. The distillation curve modeling showed that blending oxygenates affects the distillation curve much more than would be expected from a linear blending relationship: the front end volatility is reduced a little, whilst the mid range volatility is increased significantly, particularly for methanol blends.

INTRODUCTION

Spray Guided Direct Injection (SGDI) engines are a key enabler to reducing CO2 emissions and improving fuel efficiency. Methanol and ethanol can be blended with gasoline to make fuels that are suitable for automotive use, and they can be made by renewable and sustainable means. This research explores the effect on the emissions of ultra-fine Particulate Matter (PM) and unburned hydrocarbons when such fuels are used in a SGDI engine. Much of the historical work in this area has focused on the measurement of soot, smoke or particulate mass emissions from diesel engines, and how they can be reduced by using fuels with oxygen in the fuel molecule. This literature generally suggests that the reduction in PM mass emissions is correlated with the amount of oxygen in the fuel molecule. Whilst PM mass is an important metric, some recent literature suggests that PM size and number may be more important as a predictor of the health effects of PM [1]. In the present work, the number concentration of PM is measured for particles with equivalent diameters1 between 5 and 1000 nm, and data are presented as PM size distributions.

In a diesel application with methanol fuel, Kaskavaltzis et. al. [2] reported an order of magnitude reduction in PM mass compared to conventional diesel fuel. Emissions of NOx and CO were also significantly reduced. Similar results were found by King and Prakash [3]. According to Black [4], the reduction in NOx emissions for methanol is due to the lower flame

1 Electrical mobility equivalent diameter is the diameter of a spherical particle with the same electrical mobility as the particle being measured.

temperature, high heat of vaporization and fast flame speed. Adiabatic flame temperatures and heats of vaporization are given in Table 3.

Combustion of alcohols is also known to produce emissions of aldehydes, particularly formaldehyde and acetaldehyde, which are highly reactive in ozone formation. High formaldehyde emissions are mentioned in references [2-4], and also by Gardiner et. al. [5]. This is not surprising because aldehydes are partial oxidation products of alcohols.

As pure fuels, methanol and ethanol are not volatile enough to be used under cold conditions. Brinkman et. al. [6] and Gardiner et. al. [5] also reported cold starting difficulties with pure alcohol fuels. Partly to alleviate this problem, methanol or ethanol can be blended with 15 percent gasoline to provide the light-end components that vaporize at low temperatures. These blends are known as M85 (85 percent methanol and 15 percent gasoline by volume) and E85 (85 percent ethanol and 15 percent gasoline). M30 and E30 were also tested as these might be used in ‘standard’ engines without the need for significant fuel system modifications.

The spray guided prefix in SGDI refers to the mixture preparation strategy used in these engines. Through a combination of higher volumetric efficiency and compression ratio due to charge cooling, reduced wall heat losses and stable operation over a large stratified charge range, these engines have significantly higher fuel economy than port fuel injected engines. For example Frohlich and Borgmann [7] demonstrated 20 percent higher fuel economy for a spray-guided Direct Injection Spark Ignition (DISI) combustion system with piezoelectric injectors than for a port fuel injected engine. Further details can be found in references [8-10].

In all DISI engines, the time available for mixture preparation is limited to the window between injection and ignition, and this is usually less than the time needed for the charge to become a homogeneous mixture and even for all the fuel to evaporate. Consequently, charge heterogeneity and droplet combustion are two mechanisms causing PM emissions in DISI engines. In addition to emissions measurements, fast response in-cylinder temperature measurements were made using cold-wire thermometry. From these data, the change in the charge internal energy between injection and ignition was calculated as an indicator of mixture preparedness with the different fuel blends.

2

EXPERIMENTAL APPARATUS

ENGINE

The engine was a single cylinder spray guided direct injection spark ignition engine, with the geometry shown in Table 1. As can be seen from Figure 1, the injector and spark plug are mounted close together in the top of a pent roof combustion chamber. Both are inclined at a small angle to the longitudinal axis of the cylinder. The piston was a conventional flat-top design.

Figure 1. Diagram showing the combustion chamber geometry and the location of the fuel injector and spark plug in the engine used here.

Type 4-Stroke Combustion System Spray Guided DISI Bore x Stroke 89 x90 mm Swept Volume 562 cm3

Compression Ratio 11:1 Engine Speed 1500 rpm Injector Multi-hole Nozzle Fuel Pressure 150 bar

Table 1. Engine Specification.

INSTRUMENTATION AND SAMPLING – PM EMISSIONS

PM number concentration and size distribution was measured with a Cambustion Differential Mobility Spectrometer (DMS500), as described in [11] and [12]. Size discrimination is achieved by classifying particles according to their charge to aerodynamic drag ratio, yielding an electrical mobility equivalent diameter. PM number concentration is computed from electrometer currents that arise due to the flux of electrically charged

3

particles arriving at electrometer surfaces within the particle classification column.

The sample point was in the feed-gas, with 10:1 dilution at the sample point made using a choked flow dilution system. In these experiments the diluent was molecular nitrogen at ambient temperature. After dilution, the sample was transported to the DMS500 through a short length of conductive tubing.

INSTRUMENTATION – UNBURNED HYDROCARBON EMISSIONS (UBHC)

UBHC measurements were made with a Cambustion fast response Flame Ionization Detector (fastFID). A detailed description of this instrument can be found in [13]. The fastFID was setup to sample from the same position as the DMS500, in the feed-gas. As previously described in [14], a dynamic calibration system was used so that a two-point calibration was performed in-situ before and after every measurement. This method has the advantage that the calibration is done under operating conditions and so the sample probe is exposed to the same pressure and temperature as during the measurement. Drift-corrections are made automatically as the instrument sensitivity is defined as the mean of the offset corrected span calibration points.

For most hydrocarbons, the FID response is proportional to the number of C atoms in each molecule. For alcohols, the C bonded to O in an R-O-H group where R is an alkyl radical gives a response of about 50 to 85 percent of a C atom [13]. The same is true for the FID response to aldehydes.

INSTRUMENTATION – IN-CYLINDER TEMPERATURE MEASUREMENTS

The in-cylinder temperature varies vastly and rapidly from ambient temperature to combustion temperature in millisecond timescales. In order to accurately measure the temperature changes during the fuel injection and evaporation process, a sensor must have a response time in the range of milliseconds. A cold wire resistance thermometer was used to measure the in-cylinder temperature. The principle of operation is discussed in detail in [15]. The sensor was a fine tungsten wire with a platinum coating. It was 5 μm in diameter and 1.25 mm long. The small diameter ensures a minimal thermal inertia, and hence a fast response time. Based on the well known Collis-Williams heat transfer correlation [16], it has been estimated that the frequency response (-3dB point) ranges from 250 to 500 Hz at a flow velocity between 5 m/s to 40 m/s. The velocity range is considered typical of the intake and compression strokes. This is sufficiently fast to study temperature changes in the cylinder.

Since the sensor is so fine, it is too fragile to be used under firing conditions. It was only used when the engine was running with injection, but without spark. Other

conditions such as injection timing and duration, coolant temperature and operating points were matched to the firing tests as closely as possible. Since the main interest is on the temperature during the intake and compression strokes, where the effect of combustion is minimal, the results should be directly relevant to the firing tests.

The sensor was installed in place of a spark plug hole using a specially designed adaptor. The adaptor was designed to seal the gas from escaping through the sensor body, and provide mechanical support to withstand the pressure force. The measurement point was 4 mm above the combustion chamber surface, which was where the arc would have struck if the spark plug was there. The adaptor had a shield concentric to the spark plug centerline. This was to ensure that the gas temperature was accurately measured without the effect of fuel impingement.

FUEL COMPOSITION AND DISTILLATION CURVES

The composition of the base Unleaded Gasoline (ULG) is given in Table 2. In addition to this, some properties of the alcohols used for blending are given in Table 3.

Density @15°C (Kg/L) 0.729 Reid Vapor Pressure (mbar) 561 Research Octane Number (RON) 95.0 Motor Octane Number (MON) 86.3 C 6.58 H 12.5 O 0 Paraffins (%V) 7.18 Iso-Paraffins (%V) 53.3 I + N Paraffins (%V) 60.5 Olefins (%V) 4.18 Naphthenes (%V, Saturated) 6.46 Aromatics (%V) 28.8 Oxygenates (%V) 0.00 Benzene (+ Me Cy.Pentene) (%V) 0.60 Stoichiometric Air Fuel Ratio 14.6 Enthalpy Of Combustion (MJ/Kg, Gas) -43.6 Enthalpy Of Combustion (MJ/Kg, Liq.) -43.2 Thus, Heat Of Vap. (MJ/Kg) 0.368 H/C 1.90 O/C 0.000 Sulphur Content (Mg/Kg) 9.00

Table 2. Base Fuel Composition. Analysis by Shell Global Solutions.

Methanol Ethanol Density @15°C Kg/L 0.797 0.794 Reid Vapor Pressure 317 Mbar 159 Mbar Research Octane Number (RON) 106 107 Boiling Point 65 78.5 Stoichiometric Air Fuel Ratio 6.45 9.0 Enthalpy Comb (MJ/Kg, Liq. -19.9 -26.8 Thus, Heat Of Vap. (MJ/Kg) 1.167 0.925 H/C 4 3 O/C 1 0.5 Adiabatic Flame Temperature (K) 2243 2258

4

Table 3. Some properties of pure Methanol and Ethanol. Source: Owen and Coley [17], and Goodger [18].

The ASTM D86 distillation curves have been predicted for blends of ULG95 with 10% and 30% by volume of ethanol and methanol, and can be seen in Figure 2. The UNIFAC method was used to obtain activity coefficients to account for liquid phase interactions. The accuracy of this method has previously been checked by comparison with experimental data. It generally gives good agreement, although the transition region where the temperature with the alcohol blends rises sharply towards the temperature with the base fuel is generally slightly less steep in the calculated distillation curve than is found by measurements.

Figure 2. Distillation curves for Gasoline – Alcohol blends.

The distillation curve modeling (Figure 2) shows that blending alcohols with ULG affects the distillation curve much more than would be expected from a linear blending relationship: the front end volatility is reduced a little, whilst the mid range volatility is increased significantly, particularly for the methanol blends.

Note that the enthalpy of vaporization of the alcohol-ULG blends is not a linear combination of the values of the constituents due to the enthalpy of mixing. The mixing reaction is endothermic, and therefore the enthalpy of vaporization of the mixture is less than that calculated by linear combination.

CHARGE TEMPERATURE AND MIXTURE PREPARATION In direct injection engines, there is only a limited time for the fuel to vaporize and mix with the air to form a combustible mixture. The quality of mixture preparation has a direct impact on PM and UBHC emissions. Piston and cylinder wall wetting, incomplete evaporation and incomplete mixing lead to charge heterogeneity, pool fires and droplet combustion which are known to cause PM formation.

When liquid fuel is being injected into the cylinder, it evaporates and absorbs heat from the surroundings, which includes the bulk gas, piston crown and other combustion chamber surfaces. This will lower the in cylinder temperature. In general, better mixture preparation will have more fuel evaporated before ignition, and hence a higher temperature drop in the gas is expected. It follows that in-cylinder temperature can be an indicator of how well the mixture is prepared.

Two temperature parameters are considered. The first is the temperature change before and after the fuel injection event (also known as charge cooling). As the fuel is injected into the cylinder in the intake stroke, some of it vaporizes, absorbing energy from the surroundings (mostly from air). Hence the charge temperature decreases. So, the amount of fuel evaporated can be inferred from the extent of charge cooling. The second temperature parameter is the temperature rise from Inlet Valve Closure (IVC) to Ign (TTTc). Dodge [17, 20] found that most of the fuel evaporation happens in the compression stroke. This is because the rate of evaporation increases exponentially with temperature. So, TTTc may be a better indicator of fuel evaporation and the quality of mixture preparation.

0

20

40

60

80

100

120

140

Types of fuel

Tem

pera

ture

Diff

eren

ce (K

)

Charge Cooling (K) 30.1 35.9 38.4 33.8 36.8Temp. rise due tocompression, Tc (K)

116 108 104 110 107

ULG E85 M85 E30 M30

Figure 3. Comparison of charge cooling effect and temperature rise due to compression (Tc) for five different fuels.

Figure 3 shows the charge cooling and temperature rise from inlet valve closure (IVC) to ignition (Ign) for high load baseline conditions (Pinlet = 0.80 bar, Ign = 30 Crank Angle Degrees (CAD) bTDC, SOI = 80 CAD aTDC, λ = 1). The results are the averages over 230 cycles. It can be seen that the highest amount of charge cooling occurs when the fuel is M85, and the charge cools the least when ULG is used. However, a completely opposite trend is shown in the temperature rise due to compression.

5

It is important to note that the temperature measurements were made in the region of the spark plug electrode gap. These data are local measurements, and therefore might not be representative of the temperature at different locations in the cylinder. The problem is particularly severe during the fuel injection event, because the fuel is not distributed uniformly in the cylinder. In fact, the charge cooling observed is lower than the theoretical maximum attainable. For example, had all the M85 fuel evaporated after injection, the charge would be cooled by 100 K, assuming the enthalpy of vaporization was solely supplied by the enthalpy change of the charge. It is thus clear that the evaporation process is either limited by the saturation of the fuel, and/or by diffusion and mixing within the cylinder. In contrast, the temperature rise due to compression (TTTc) may be considered more like a global quantity. This is because fluid motion (mixing and turbulence) will tend to even out any temperature gradients and the droplet distribution in the cylinder during the compression stroke.

Given that evaporation happens mostly in the compression stroke and the measured charge temperature in the compression stroke is then more representative of the whole cylinder, only TTTc will be used to assess the fuel evaporation process. TTTc indicates the change in internal energy of the charge (ΔU). This is based on the assumption of a uniform air fuel mixture, i.e. no residual gas fraction. From Eq. 1, ΔU and the

heat absorbed for evaporation (L) is balanced by the work done on the charge (W) and the net heat transfer to the surroundings (Q).

Q W U L− = Δ + (1)

The heat transfer is mainly to the cylinder wall, which is not significant as the average charge temperature from IVC to Ign is only 15 K higher than the coolant temperature. So, the work done by the piston is essentially used to raise the internal energy of the charge, and to supply energy for evaporating the fuel. If more energy is used to evaporate the fuel, it should indicate more complete mixture preparation. This occurs when:

• There is a smaller increase in internal energy with the same work done.

• There is a greater work done on the charge with the same change in the internal energy.

These two aspects have to be assessed together to determine which fuel has the best mixture preparation.

Note that TTTc between different fuels cannot be used in place of ΔU. Firstly, the alcohol blends have a lower stoichiometric air-fuel ratio, so for a stoichiometric mixture and identical air mass flow rate (the control conditions in these baseline tests), more fuel mass is injected than ULG as shown in Table 4. Secondly, the constant volume specific heat capacities of the different fuels are significantly different.

ULG M85 M30 E85 E30

Cvf 1.0 0.67 0.87 0.78 0.91

mf 1.0 1.90 1.22 1.48 1.14

Table 4. Comparison of the constant volume specific heat and mass injected per cycle between different fuels. Values are normalized to the ULG’s values.

There are two competing effects in the alcohol blends: the mass in the cylinder is greater compared to the mass of ULG, but their constant volume specific heat is also lower than ULG. It was therefore decided that the best indicator is to calculate the change in internal energy directly.

PARTICULATE MATTER EMISSIONS

The data presented in this section are PM size distributions measured with the DMS500 (described earlier). The quantity on the abscissa is the electrical mobility equivalent diameter, dp, and the ordinate is the PM number concentration, normalized with the differential interval of the logarithm of particle size, dN/dlogdp. A logarithmic scale is used on both axes because the data span several orders of magnitude. The distributions shown are time-averaged and were sampled under steady-state operating conditions. Figures are presented and discussed in turn for the PM response to the relative air fuel ratio, λ, ignition timing and injection timing, with the data for the five different fuels shown as sub-figures within each figure.

In addition to PM size distributions, each figure also contains a sixth sub-figure in the lower right-hand corner. This shows the ‘total’ PM number concentration,

N. The quantity N is the PM number concentration for particles in the measured size range (5 < dp < 1000 nm), calculated by numerically integrating the PM size distributions with respect to dp. In the figures, the logarithm of N is presented to permit the comparison of data with a range of orders of magnitudes.

PM size distributions are often bi-modal, with the two characteristic modes being known as the nucleation mode and the accumulation mode. These modes are distinguished by the particle equivalent diameter, with the nucleation mode generally being 5 < dp < 40 nm and the accumulation mode being 40 < dp < 500 nm. A third mode, known as the coarse particle mode, can sometimes be observed. It is usually negligible, especially when the metric is PM number concentration. A more detailed description of these modes and a discussion of their significance can be found in [21].

Figure 4. Response of PM emissions to relative air fuel ratio. The curves are for different relative air fuel ratios in the range 0.8 < λ < 1.2. Operation point: 1500 rpm, inlet manifold pressure = 0.8 bar, ignition timing = MBT, SOI = 80 CAD aTDC.

Figure 4 shows PM size distributions for all five fuels at a series of different relative air fuel ratios in the range 0.8 < λ < 1.2. The λ ratio was set using an ECM AFRecorder, model 1200A (calibrated heated universal exhaust gas oxygen sensor). The ignition timing was set to the Minimum advance for Best Torque (MBT) by sweeping the ignition timing for each fuel and finding the maximum Indicated Mean Effective Pressure (IMEP).

For all the fuels tested, the response of PM number concentration to the relative air fuel ratio was to increase with decreasing λ ratio. So the highest PM emissions were measured for rich mixtures (λ = 0.8). For the base fuel (ULG), the response was strong. For example the accumulation mode PM number concentration at a

lambda ratio of 0.8 is over two orders of magnitude higher than at a lambda ratio of 1.2. For M85 and E85, the magnitude of the accumulation mode was smaller (~105 particles/cm3 for λ = 1.0). In addition, the accumulation mode PM response to λ for M85 and E85 was much less pronounced than for the ULG. However, M85 and E85 still showed a significant response of the nucleation mode PM to λ.

Inspection of N in the lower right hand sub-figure allows the emitted PM number concentrations for the different fuels to be ranked in order of magnitude. In general, the lowest PM number concentrations (across the measured range of dp) were measured for E85, followed by E30. After the ethanol blends, M30 and M85 gave the lowest

6

numbers for lean mixtures but ULG was lower for rich mixtures.

A possible explanation for the invariance of the accumulation mode with λ for the fuel blends with 85 percent alcohol may be drawn from the work of Santoro and co-workers [22], which is also cited in [23]. They suggest that the presence of oxygenated compounds in combustion reduces the concentrations of key intermediate species that are required for the

formation of the initial aromatic ring that is known to be a pre-cursor to soot particle inception.

A consequence of the smaller accumulation mode for the 85 percent alcohol blends is that the surface area available for adsorption of condensing organic species is reduced, and thus this material is more likely to nucleate homogeneously, forming high concentrations of nucleation mode PM, as can be seen in Figure 4.

Figure 5. Response of PM emissions to ignition timing. The curves are for different ignition timings in the range 15 < Ignition < 35 CAD bTDC. Operation point: 1500 rpm, inlet manifold pressure = 0.8 bar, relative air fuel ratio = 1.0, SOI = 80 CAD aTDC.

Figure 5 shows the PM response to ignition timing for the five fuel blends. The limits on the logarithmic scales used are the same as those used in Figure 4 and Figure 6.

The magnitude of the PM number concentrations measured generally increases with ignition advance, although the response over the range of ignition timings tested was weaker than for the relative air fuel ratio. For example the accumulation mode PM number concentrations measured when the fuel was ULG varies over about an order of magnitude, compared to two orders of magnitude when changing λ. This dependence of PM emissions on ignition timing is to be expected because the temperature in the expansion and exhaust strokes is highest for late ignition timing, and so the amount of post-flame oxidation of organic compounds before exhaust valve opening is increased. In addition to post-flame oxidation, advancing the ignition reduces the time available for mixture preparation. This change may be far greater than expected from the small change in the mixture preparation time. This is because the fuel evaporation rate is highly temperature dependent; thus a

few extra crank angle degrees late in the compression stroke may have a far greater effect on the fuel droplet size and mass fraction of fuel evaporated than the same number of crank angle degrees in the intake stroke.

Of the fuels tested, it was consistently found that the lowest PM number concentrations were when the fuel was E85. E30 and ULG were similar, and M30 followed by M85 gave the highest PM number concentrations. It should be noted that PM number concentration does not necessarily agree with PM mass. Inspection of the size distribution of the emitted PM for M85 in Figure 5 shows that most of the PM number is distributed in the nucleation mode, and that the accumulation mode number concentrations are lower than for the base ULG.

For E85, the accumulation mode PM is very low (similar to ambient background levels at (~104 particles/cm3) when the ignition advance was 15 and 20 CAD. It is currently unclear why it is the case for E85 more than the other fuels. The remaining PM is nearly all nucleation mode, and shows a strong dependence on ignition timing (expansion temperature) which suggests

7

that it may be composed of a significant volatile organic fraction. This is consistent with high UBHC emissions as shown later in Figure 9. If this is the case, the nucleation mode is probably formed by condensation nucleation of volatile organic compounds that become super-saturated as the exhaust cools. It is hoped that further

research will test this hypothesis by testing the sample volatility using a sample conditioning device such as a thermo-denuder.

Figure 6. Response of PM emissions to injection timing. The curves are for different Start of Injection (SOI) timings in the range 40 < SOI < 200 CAD aTDC. Operation point: 1500 rpm, inlet manifold pressure = 0.8 bar, relative air fuel ratio = 1.0, ignition timing = MBT.

Figure 6 shows the PM response to injection timing. It should be noted that the injection timings shown here are for the start of injection (SOI), and that the injection duration depends on the operating point and fuel being used. The blends with high alcohol concentrations require the longest injection durations due to their lower stoichiometric air-fuel ratio compared to ULG. For M85 at the operating point used in Figure 6 the injector duration was 3.8 ms, which corresponds to 34 CAD at 1500 rpm. As for the data in Figure 4, MBT ignition timing was set for each fuel by selecting the minimum advance corresponding to the highest IMEP.

The lowest accumulation mode PM number (dp > 40 nm) were measured when injection was earliest (SOI = 40 CAD aTDC), presumably due to there being the longest time available for mixture preparation. Hence PM emissions increase as the time for mixture preparation is reduced. This argument holds for ULG and E30, but does not describe the observed PM emissions for M30, M85 and E85, which are more volatile. In fact, PM emissions for M30, M85 and E85 for late injection (160, 200 CAD aTDC) are comparable if not lower than the earliest injection (40 CAD aTDC). The effect of injection timing on PM emissions is confounded by the following two factors:

1. Time for mixture preparation - this includes time for the fuel to evaporate and mix to form a homogeneous charge.

2. Piston and wall wetting - for early intake stroke injection, fuel impinges on the piston crown creating a fuel film. This film can burn diffusively and as a pool fire. Both cases are particularly effective at forming primary carbon particles that later aggregate and are measured as PM emissions.

The effect of SOI on the feed-gas PM emissions is quite strongly dependent on fuel composition and volatility [14]. If the fuel is sufficiently volatile to evaporate before ignition, the minimum PM emissions will occur at the SOI where piston and wall wetting are also at a minimum. In this case late injection (~BDC) is favorable. Conversely, if the fuel has a significant ‘heavy’ fraction, the time for mixture preparation is likely to be dominant over the effects of piston and wall wetting, so an early injection is desired to minimize PM emissions. The in-cylinder temperature measurements made here are used to verify this theory.

8

0.93

0.94

0.95

0.96

0.97

0.98

0.99

1.00

1.01

0 40 80 120 160 200 240SOI (CAD aTDC)

Nor

mal

ized

hea

t abs

orbe

d fo

r ev

apor

atio

n

Figure 7. Normalized heat absorbed for evaporation as a function of fuel injection timing for ULG.

0.920.93

0.940.95

0.960.970.98

0.991.00

1.011.02

0 40 80 120 160 200 240SOI (CAD aTDC)

Norm

aliz

ed h

eat a

bsor

bed

for

evap

orat

ion

Figure 8. Normalized heat absorbed for evaporation as a function of fuel injection timing for E85.

Figure 7 and Figure 8 show the normalized heat absorbed for evaporation as a function of SOI. The results are normalized with the values at 40 CAD aTDC. With ULG, the heat absorbed for evaporation decreases approximately monotonically with injection retard through about 6 percent compared to the value at SOI = 40 CAD aTDC. With E85, the result was quite different: the relationship between the heat absorbed for evaporation and SOI was non-linear, with a minimum at SOI = 120 CAD aTDC, and maxima at both early and late SOIs. This trend is in agreement with the accumulation mode PM emissions for E85 as seen in Figure 6.

The impact of SOI on PM emissions can be related by how readily the fuels evaporate in the compression stroke. A volatile fuel (E85) will give low PM emissions with late injection because of reduced piston and wall wetting. For a less volatile fuel (ULG) that has insufficient time to evaporate in the compression stroke, the time for mixture preparation is more important than piston wetting. In this case, late injection will result in droplet combustion, which may then lead to high accumulation mode PM emissions.

UNBURNED HYDROCARBON EMISSIONS

9

10

Figure 9. Response of Unburned Hydrocarbon Emissions to (from left) relative air fuel ratio, ignition timing and injection timing (SOI). Operation point: 1500 rpm.

The data presented in this section were measured with a fast Flame Ionization Detector (fFID), as described earlier. A four-point calibration (two before and two after) was performed for every test point, and then the fFID sensitivity was calculated as the mean of the offset corrected span calibration points. For the steady state results presented here, the data were time averaged. Whilst this is sufficient for an approximate analysis, the correct approach requires a transit time corrected and mass flow-rate weighted average. The mass flow-rate is currently not known but may be obtained from one dimensional steady flow simulations of the flow through the exhaust valves.

A further limitation of the application of this technique to measure emissions from alcohol combustion is the fFID response to alcohols and aldehydes (see Cheng et. al. [13]). Emissions of aldehydes are an inevitable consequence of alcohol combustion, particularly with rich air fuel mixtures. A more rigorous approach would be to speciate the UBHC emissions using a gas chromatogram and/or a mass spectrometer.

Figure 9 shows the UBHC response to the parameters investigated. The results are presented for two different loads, as indicated by inlet manifold pressures of 0.5 and 0.8 bar in the figure. The lower row of sub-figures (inlet pressure = 0.8 bar) correspond to the tests for which PM measurements were presented earlier.

Possibly the most significant observation is that the ranking of UBHC emissions by fuel is not the same at low (inlet pressure = 0.5 bar) and high loads (inlet pressure = 0.8 bar). At low load, the lowest UBHC emissions were measured when using the alcohol blends. The highest UBHC emissions were when using the ULG, particularly with rich air fuel mixtures (λ < 1.0). At high load, the difference in UBHC emissions between the fuels was more pronounced. With the exception of very early injection, E85 gave the highest UBHC emissions at high load. M30, M85 and E30 were in the middle and ULG was usually the lowest.

UBHC emissions decrease with increasing relative air fuel ratio in the range 0.8 < λ ~< 1.1, and then increase. The increase is coincident with a sharp increase in the coefficient of variation of indicated mean effective pressure (CoV of IMEP), which suggests an increase in partial burns and the approach of the lean flammability limit. For the relative air fuel ratio tests there is a systematic difference in the ranking of the UBHC curves by fuel. At low load, the highest UBHC emissions were measured for ULG, and the alcohol blends were comparatively similar. For example, the UBHC emissions were about 60 percent higher at a λ ratio of 0.8 for ULG compared to the alcohol blends. There was

no discernable difference in the CoV of IMEP between the fuels for λ <=1.1. At high load, ULG gives the lowest UBHC emissions, and also the lowest CoV of IMEP. E85 gave the highest, followed by the other alcohol blends. At a λ ratio of 0.8, the UBHC emissions from E85 were 65 percent higher than for ULG. The difference decreased with increasing λ ratio. At a λ ratio of 1.1 (still high load) the UBHC emissions were almost exactly the same. The low load reduction in UBHCs when using alcohol blends is consistent with results in the literature for different types of engines, and is to be expected with fuel bound oxygen. The reversal for high load is unusual, but can possibly be explained by considering the larger volume of fuel that must be injected at high load for the alcohol blends due to their lower stoichiometric air fuel ratios. In addition to this, the mixture preparation analysis showed some saturation of the charge for the 85 percent alcohol blends. This suggests incomplete evaporation at the time of ignition may be a significant contributor to UBHC emissions in these cases.

At low load (inlet pressure = 0.5 bar), UBHC emissions increase with ignition advance, consistent with decreasing expansion temperature and post-flame oxidation. At high load (inlet pressure = 0.8 bar), the UBHC emissions are significantly higher, and their response to ignition timing is quite flat for the alcohol blends, suggesting that UBHCs out-gassing from the crevice volumes in the expansion stroke is not the dominant source of UBHC emissions. If the hypothesis of saturation of the fuel vapor in the compression stroke is assumed, then the dominant mechanism is probably incomplete combustion due to the presence of liquid droplets in the charge.

For most of the alcohol blends tested, UBHC emissions increased with fuel injection retard, i.e., they increase with decreasing time for mixture preparation. When ULG was used, the UBHC response was quite flat. A possible explanation for this is that the contribution to UBHC emissions from the fuel film on the piston crown (following injection) is less significant for the alcohol blends than for the ULG due to their higher mid-range volatility. Several authors have suggested that piston wetting is a significant source of UBHC emissions [24-28], and it is suggested here that the heavy gasoline fraction will remain on the piston crown for much longer than the rest of the fuel, and therefore contribute more to UBHC emissions.

11

CONCLUSION

Measurements of emitted particulate matter and unburned hydrocarbons were made in a single cylinder spray guided direct injection spark ignition engine with the following five fuels: ULG, M30, E30, M85 and E85. Based on these measurements, the following conclusions were drawn:

• Blending 10 and 30 percent concentrations of methanol or ethanol has a profound effect on the ASTM D86 distillation curves. Compared to ULG, the front-end volatility is reduced a little, whilst the mid-range volatility is significantly increased.

• It has been shown that the heat absorbed for evaporation of the fuel can be inferred using a combination of fast in-cylinder temperature data and compression work calculated from cylinder pressure data. Further, it was shown that the temperature rise due to compression can be used to indicate the change in charge internal energy because of the tendency of the effects of different fuel masses and specific heat capacities of the alcohol blends approximately cancel out.

• For stoichiometric air fuel mixtures, the lowest PM number concentrations were measured for E85. ULG, E30 and M30 were generally similar and the highest PM number concentrations were consistently found when using M85. This is a generalization to which there are some outliers. Most notably when SOI was 120 CAD aTDC.

• Whilst the PM number concentrations measured for M85 were high, the PM size distributions were dominated by nucleation mode PM. The accumulation mode was consistently smaller than for ULG, suggesting lower PM mass, which is consistent with the literature reviewed in the introduction.

• The PM response to the relative air fuel ratio was much less pronounced for the 85 percent alcohol blends than for the 30 percent blends or the ULG, and the accumulation mode number concentrations were significantly lower, particularly for rich air fuel mixtures. It was postulated, with support from the literature, that the presence of oxygen in the fuel molecule reduces the concentrations of key intermediate species required for the formation of aromatic soot precursors.

• For ULG and E30, accumulation mode PM emissions increase with injection retard. This is only partially true for M30, M85 and E85; in these cases, the PM emissions were similar for

late injection timing. Analysis of the measured temperature data confirms that fuel volatility dictates whether the time for evaporation or piston and wall wetting is the dominant source of PM emissions. A fuel (e.g. E85) that can readily evaporate in the compression stroke would give low PM emissions with late injection because piston and wall wetting are insignificant and vice versa.

• The response of unburned hydrocarbon emissions to the fuels tested was found to be dependent on load. At low load, the highest emissions were generally found when using ULG, particularly with rich air fuel mixtures. At high load the situation was somewhat reversed, with the highest UBHC emissions being from E85, and ULG giving the lowest. Based on fast in-cylinder temperature measurements during the compression stroke, it was suggested that the additional volume of fuel that must be injected to achieve the same relative air fuel ratio for the 85 percent alcohol blends can cause the charge to become saturated. The high UBHC emissions were then attributed to incomplete combustion due to the presence of liquid droplets in the charge.

ACKNOWLEDGMENTS

This work was funded by the Engineering and Physical Sciences Research Council and Ford Motor Company Ltd. Tony Collier and Marcus Davies of Ford Motor Company are also acknowledged for their continuing support. Kenneth Kar would like to acknowledge the financial support from the Foundation for Research, Science and Technology of New Zealand.

REFERENCES

1. Oberdorster, G., Oberdorster, E and Oberdorster J. Nanotoxicology: An Emerging Discipline Evolving from Studies of Ultrafine Particles. Environmental Health Perspectives. Vol 113. No. 7. (2005).

2. Kaskavaltzis, C., Elliott, D. and Topaloglu, T. A Demonstration of Methanol-Powered Buses in Windsor Ontario. SAE Technical Paper 942314. (1994).

3. King, L.J. and Prakash, C.B. Field Trials of Ethanol in Transit Buses. SAE Technical Paper 922272. (1992).

4. Black, F. An Overview of the Technical Implications of Methanol and Ethanol as Highway Motor Vehicle Fuels. SAE Technical Paper 912413. (1991).

12

5. Gardiner, D.P., Rao, V.K., Bardon, M.F., Dale, J.D., Smy, P.R., Haley, R.F., Dawe, J.R. and Battista, V. Sub-Zero Cold Starting of a Port-Injected M100 Engine Using Plasma Jet Ignition and Prompt EGR. SAE Technical Paper 930331. (1993).

6. Brinkman, N., Halsall, R., Jorgensen, S.W. and Kirwan, J.E. The Development of Improved Fuel Specifications for Methanol (M85) and Ethanol (Ed85). SAE Technical Paper 940764. (1994).

7. Frohlich, K., Borgmann, K. (2003) Potentials of Future Fuel Consumption Technologies. 24 Internationales Weiner Motorensymposium (Vienna), May 15-16. In German.

8. Wirth M., Zimmermann, D., Friedfeldt R., Caine, J., Schamel A., Davies, M., Peirce, G., Storch A., Ries-Müller K., Gansert, K.-P., Pilgram, G., Ortmann, R.,Würfel G., Gerhardt, J. (2004) A Cost Optimised Gasoline Spray Guided Direct Injection System for Improved Fuel Economy. IMechE Fuel Economy and Engine Downsizing. Seminar Proceedings (London), May 13.

9. Szekely, G.A. and Alkidas, A.C. Combustion Characteristics of a Spray-Guided Direct-Injection Stratified-Charge Engine with a High-Squish Piston. SAE Technical Paper 2005-01-1937. (2005).

10. Kawamoto, M., et. al. A Study of Center and Side Injection in Spray Guided DISI Concept. SAE Technical Paper 2005-01-0106. (2005).

11. Reavell, K., Collings, N. and Hands T. (2002) A Fast Response Particulate Spectrometer for Combustion Aerosols. SAE Paper 2002-01-2714. (2002).

12. Reavell, K. Fast Response Classification of Fine Aerosols with a Differential Mobility Spectrometer. 13th AGM of The Aerosol Society (Lancaster), April 11-12. (2002).

13. Cheng W. K., Summers, T., and Collings, N. The Fast-Response Flame Ionization Detector. Prog. Energy Combust. Sci. 24, 89-124. (1998).

14. Price, P., Stone, R., Collier, T. and Davies M. Particulate Matter and Hydrocarbon Emissions Measurements: Comparing First and Second Generation DISI with PFI in Single Cylinder Optical Engines. SAE Technical Paper 2006-01-1263. (2006).

15. Lawton, B. and Klingenberg, G. Transient Temperature in Engineering and Science. Oxford University Press, Oxford. (1996).

16. Collis, D. C. and Williams, M. J. Two-dimensional convection from heated wires at low Reynolds numbers. J. Fluid Mech. 6, 357-384. (1959).

17. Owen, K. and Coley, T. (1995) Automotive Fuels Reference Book. Second Edition. Society of Automotive Engineers, Warrendale.

18. Goodger E. M. (1975) Hydrocarbon Fuels. Macmillan, London.

19. Dodge, L. G. Fuel preparation requirements for direct-injected spark ignition engines. Proceedings of ILASS-America, p 120-124. (1996).

20. Dodge, L. G. Fuel preparation requirements for direct-injected spark ignition engines. SAE Paper No. 962015. (1996).

21. Kittelson, D.B. Engines and Nanoparticles: A Review. J. Aerosol Sci. Vol. 29, No. 5/6, pp. 575-588, (1998).

22. Ni, T., Gupta, S. and Santoro, R. Suppression of Soot Formation in Ethene Laminar Diffusion Flames by Chemical Additives. In 25th Symposium (International) on Combustion, pp 585-592. The Combustion Institute. (1994).

23. Litzinger, T., Stoner, M., Hess, H. and Boehman, A. Effects of Oxygenated Blending Compounds on Emissions from a Turbo-Charged Direct Injection Diesel Engine. Int J Engine Research. Vol 1. No. 1. (2000).

24. Li, J., et. al. Further Experiments on the Effects of In-Cylinder Wall Wetting on HC Emissions from Direct Injection Gasoline Engines. SAE Technical Paper 1999-01-3661. (1999).

25. Park, J. et. al. Characteristics of Direct Injection Gasoline Spray Wall Impingement at Elevated Temperature Conditions. SAE Technical Paper 1999-01-3662. (1999).

26. Lindgren, R. et al. Modelling Gasoline Spray-wall Interaction - A Review of Current Models. SAE Technical Paper 2000-01-2808. (2000).

27. Habchi, C. et. al. Influence of the wall temperature on the mixture preparation in DI gasoline engines, Oil & Gas Science and Technology - Rev. IFP, 54, no. 2, pp. 211-222. (1999).

28. Block, B. et. al. Luminosity and Laser-Induced Incandescence Investigations on a DI Gasoline Engine. SAE Technical Paper 2000-01-2903. (2000).

13

CONTACT

Philip Price, Department of Engineering Science, University of Oxford, Parks Road, Oxford, OX1 3PJ, UK.

E-mail: Philip.Price@eng.ox.ac.uk

DEFINITIONS, ACRONYMS, ABBREVIATIONS

AFR: Air Fuel Ratio

BDC: Bottom Dead Centre

CAD: Crank Angle Degrees

CI: Compression Ignition

CoV: Coefficient of Variation

DI: Direct Injection

DISI: Direct Injection Spark Ignition

DMS500: Differential Mobility Spectrometer

fFID: Fast Flame Ionization Detector

Ign: Ignition timing in crank angle degrees before TDC.

IMEP: Indicated Mean Effective Pressure

IVC: Inlet Valve Closure

MBT: Minimum ignition advance for Best Torque

PFI: Port Fuel Injection

PM: Particulate Matter

SGDI: Spray Guided Direct Injection

SI: Spark Ignition

SOI: Start of Injection

TDC: Top Dead Centre

UBHC: Unburned Hydrocarbons

ULG: Unleaded Gasoline