The objectives of this study are as follows: (a) to develop an experimental database of PM emissions (mass rate, composition, number and size distribution) from SI engines operating at steady and transient (cold start) conditions as a function of fuel type, operating conditions and fuel delivery technique; (b) to develop and apply a time-resolved diagnostic for the measurement of instantaneous PM concentrations during transients as a function engine operating conditions, fuel type, and fuel injection characteristics; (c) to use the developed database to characterize and quantify the important processes that control PM formation and oxidation in the engine; and (d) to develop a physical model for the PM formation and oxidation processes.
Recent studies have suggested that atmospheric particulate matter (PM) is one of the most important factors correlating with mortality rate increases in urban areas (NRDC, 1996). A substantial fraction of the total PM is contributed by internal combustion engines, and a fraction of those come from spark-ignited (SI) engines.
Only a relatively small number of tests have been done to quantify the total mass of PM emitted by representative SI automobiles and fleets, as well as to examine the chemical composition of PM from SI engines. This limited research revealed some of the basic trends in PM emission with respect to engine type and age, steady state engine operating conditions, and fuel/lubricant composition. However, there has been no systematic study to identify the mechanisms of PM formation in SI engines.
Considering the limited of information on the amount, size distribution, chemical composition and pathways leading to PM formation and emissions from SI engines, the goals of this project are to produce a systematic study of the mechanisms leading to the formation of PM in gasoline engines, addressing the following issues: (1) total amount, size distribution and chemical composition of PM emitted by a SI engines; (2) origin of PM (from oil or fuel) and mechanisms of formation; (3) Effect of fuel type and engine operating conditions; (4) effect of transients and cold-starts; (5) effect of injection technology (port fuel injection vs. premixed fuel induction); and (6) PM removal efficiency of contemporary three-way catalysts.
The investigation utilizes experimental techniques already developed for collection and analysis of PM from SI engines, and applies new techniques necessary for the measurement of PM during transient operation. The identification of the different mechanisms to PM emissions is performed by varying engine and fuel parameters so as to test the particular hypothesis, according to the following tasks:
Engine-out and tailpipe measurements PM is obtained at selected steady state and cold start conditions. Total PM mass, size distribution, soluble organic fractions (SOF) and chemical composition are determined, using techniques developed at the MIT Analytical Chemistry Laboratory (filter analysis) and at Caltech (scanning mobility particle sizing Ð SMPS). Modern four-stroke, four-cylinder engines (Saturn DOHC and Ford Zetec) are used in the study.
The chemical analysis of the PM is planned for the characterization of the measurements relative to previous measurements in production automobiles, both at steady state conditions as well as during cold start. The identification of the origin of PM matter can be performed by using chemical analysis (for SOF compounds), elemental analysis (to identify oil-originated metals, or inorganic matter from valve surface pitting).
Routine emissions measurements of in-cylinder pressure, CO, CO2, NO, equivalence ratio and total HC are made for correlations and interpretation of the PM measurements during each experiment. Examination of the catalytic removal efficiencies will be considered once the baseline measurements are established .
Development and implementation of a time-resolved total PM measurement techniques for transient emissions testing. Techniques for quantifying PM emissions under steady state operation or over long periods of time are well established, as are techniques to time-resolve the PM concentrations in diesel engines; however, there is a virtual absence of time-resolved measurements of PM formation at concentrations as low as those expected in an SI engine.
Laser Induced Incandescence (LII) is employed to measured time-resolved concentrations in the exhaust of the SI engine. The technique has been used previously in flames and in diesel exhaust studies. The LII signal is linearly proportional to soot volume fractions, even at concentrations far below those expected in the engine; consequently, it is expected that the data may be easily interpretable.
Time-resolved measurements of PM during typical engine load and air-fuel ratio transients are made to determine changes in PM emissions. Correlations of PM emissions during start-up with equivalence ratio excursions and HC emissions will aid in identifying the possible mechanisms for PM formation.
The baseline and transient PM measurements are used in conjunction with variations in engine parameters to identify the effects of fuel, injection strategy, and operating conditions and thus the possible important mechanisms for PM formation.
In order to separate the effects of the different potential mechanisms during start-up or transient tests, specific variables (fuel type, injection type and timing, coolant temperature, etc.) are isolated.
The results of this study, showed the optimum conditions for determination of PM in a dilution tunnel. Two conditions, a high flow rate and a low flow rate condition, were tested, and the dilution rate space was scanned for ranges where PM measurement is independent of dilution ratio, thus allowing accurate measurements of PM concentrations. Once the optimum dilution ratio was established, the PM losses in the dilution tunnel were quantified (about 1.5% per meter of tunnel, within the range found by other investigators under similar conditions). The repeatability of the measurements, as well as the expected uncertainty, was quantified (typically 20 percent of the total amount, with mean sizes varying by about 3%).
The effect of the air-fuel ratio and liquid fuel on the total number PM measurements in the engine had been made, with the conclusion that the total number and mass of PM are minimized at air-fuel ratios slightly fuel-lean of stoichiometric, with orders of magnitude changes of PM concentrations with small changes in air-fuel ratio or liquid fuel presence. The matrix of tasks has been greatly expanded, as described in what follows.
During 1998, the following tasks have been accomplished:
The planned steady state measurements of PM emissions from SI engines using the SMPS instrumentation in the dilution tunnel have been concluded. The engine operating variables considered in the study were: (a) air-fuel ratio (or the equivalent inverse normalized value, equivalence ratio), (b) fuel injection timing, (c) engine speed, (d) engine load, (e) spark timing, (f) coolant temperature, (g) exhaust gas recirculation.
Further, an extensive study of the effects of fuel type, lubricant type and catalytic converter temperature was also completed, including the following fuels: indolene (a stardard boiling range test gasoline), toluene, isooctane, propane and a blend of indolene and MTBE; mineral engine lubricants of two different viscosities (10W-30 and 15W-40) as well as a synthetic oil of similar viscosity as 10W-30. Efficiencies of PM removal through a standard three-way catalyst were examined as a function of catalyst temperature.
Finally, the results of all studies have been collected into an overall physically-based, but empirically calibrated model for PM formation in SI engines. The model is based on existing physical models of PM formation in combustion systems, with allowances for factors specific to SI engines (such as presence of liquid fuel). Models for PM nucleation, growth and oxidation were combined, allowing interpretation of the contribution of the different mechanisms of PM emissions formation in SI engines.
The main conclusions of these studies concluded in 1998 are summarized as follows.
The effects of operating conditions reflect the influence of the main parameters controlling PM formation in the engine (as discovered in developing the PM model, described in item 2): fuel availability, liquid fuel availability, post-flame oxygen availability, and flame temperatures.
The effect of equivalence ratio shows that the amounts of PM emitted from SI engines reach a minimum around stoichiometric conditions, and rise by orders of magnitude both at fuel-lean as well as rich conditions. Whereas the increase under rich conditions is expected, the increase under fuel lean conditions could not be easily explained. The model indicates that an important crucial factor in the formation of PM via ignition of liquid fuel is the availability of oxygen in the post-flame environment. Whereas under rich conditions the usual soot formation mechanisms for rich mixtures are at work, under lean conditions the low temperatures (leading to low fuel vaporization and lower oxidation rates) and high oxygen concentrations contribute to raise the final PM amount. Particle sizes also reach a minimum under conditions close to stoichiometric, indicating that particle growth mechanisms on either side of stoichiometric are important.
The second important effect is the presence of liquid fuel, which was varied by changing the time at which fuel is injected into the intake manifold Ð when fuel is injected during the period when the intake valve is open, additional liquid fuel is injected into the cylinder compared to closed valve injection. When fuel is injected during the open valve period, an increase of orders of magnitude in the PM concentrations results, for relatively minor changes in the amount of fuel entering the cylinder.
The effect of speed is not monotonic, since changes in both cylinder temperatures and amount of liquid fuel are a combination of the particular speeds and loads. The effect of load (or intake manifold pressure) is an increase in PM emissions. The effect of coolant temperature, as expected, is an increase in PM with decreasing coolant temperatures, the main effect being associated with poorer vaporization of the liquid fuel. The effect of spark timing is relatively minor, as changes in peak temperatures associated with spark timing changes are compensated for higher temperatures during expansion, leading to higher in-cylinder oxidation rates. Finally the effect of exhaust gas recirculation (EGR) is a combined effect of decreasing peak temperatures (which in general should lead to increased PM) with decreasing amounts of liquid fuel into the cylinder (a result of increased vaporization in the intake manifold due to hot gas recirculation), leading to a net decrease of PM concentrations.
The effects of fuel type on PM engine-out emissions were determined for number and mass concentrations as a function of equivalence ratio. From the number concentration, one may conclude that particle nucleation (which determines the total number of particles) is negligible for a gaseous fuel such as propane. In fact, PM emissions from propane exhaust are indistinguishable from (in fact lower than) those of ambient air.
Fuel differences between the paraffinic isooctane and the aromatic toluene (which have similar boiling points) are very striking: toluene PM emissions are clearly much higher than those of other fuels, and much less sensitive to equivalence ratio than those of isooctane: this is to be expected, in as much as nucleation from toluene is relatively straightforward, whereas nucleation from isooctane requires a number of steps sensitive to fuel amount as well as temperature. The behavior of indolene falls somewhere in between isooctane and toluene, as expected from a blend of paraffinic and aromatic compounds. Perhaps surprisingly, the indolene/MTBE blend did not show significant difference from indolene at the same equivalence ratio.
The mass PM emissions show that particle size is a function of equivalence ratio, even for the extremely low PM number densities. Particle size is always lowest (around 15 nm for propane, 30 nm for isoocctane and 45 nm for toluene) around stoichiometric air-fuel ratios).
Lubricant viscosity was found to have negligible effect on PM emissions. A factor of two higher PM mass and number emissions were found with the use of a synthetic base oil relative to the mineral oil.
The focus of this study is the effect on PM emissions at engine-out (prior to the catalytic converter), largely to avoid the potential difficulties associated with different behavior of different catalyst material. Nevertheless, a standard three-way Pt-Ir catalyst was tested with the Ford engine, and comparisons were made with and without the catalytic converter. A better comparison should be made with parallel monoliths, coated and uncoated, to analyze specific differences. Catalytic removal efficiencies of 20 to 90 percent on a mass basis were obtained when the converter was warmed up above 400 ¡C. Additional work is required to confirm these results.
Based on the measurements gathered, a physically-based model has been developed to combine the fundamental variables that change with operating conditions into a simpler model capable of reproducing the experimental results. The model inputs are the operating conditions (equivalence ratio, speed, load, EGR, injection timing, spark timing, intake and coolant temperatures) and fuel type, and the outputs are the PM concentrations, and a breakdown of the different mechanisms contributing to PM formation.
The model was developed based on existing models of particle nucleation (both liquid and gas phase), growth (via adsorption and adsorption) and oxidation, and accounts for different PM formation phenomena taking place in the cylinder and dilution tunnel. Clearly, approximations must be made in order to capture the essential features of the process, with the final result that 16 parameters are required (in the present case, fitted, with constraints around physically realistic values). However, many insights regarding the importance of the different mechanisms (as explained in sections A-C) can be gained from the model.
Engine transient operation has been thought to contribute disproportionally to PM emissions, yet there is little systematic information supporting the assumption. The current study submitted the engine to periodic load and equivalence ratio step transients, and the response of PM emissions was followed using the SMPS.
The transit and delay time of the instrument and the different parts of the system were measured, showing that there is a 6 second characteristic time associated with the SMPS when used in a single-particle size (non-scanning) mode.
Typical characteristic times for load transients were of the order of 20 seconds, however, so that much of the time response of PM is captured in these experiments. However, the response contains more than one time scale, as PM dynamics leads to a change in the distribution over a time of the order of minutes. At this point, we have no explanation for this behavior, but plans are underway to expand the study to better understand this behavior.
The instrumentation and set up for LII measurements of transient PM emissions at 20 Hz time resolution has been set up and tested. At this point, we have calibrated the measurements using diesel PM emissions (which contains larger concentrations of sooty material than SI) and comparing the LII signal with PM emissions. A linear correlation was expected, but dependence on PM size (and cooling characteristics) seems to affect the results. Excellent time response is obtained, for a simple step load transient in a small naturally aspirated diesel engine (results are measured downstream of the exhaust, in the dilution tunnel): a 2-second transient is well captured by the technique.
The next step in the development is the calibration for SI engine particulate matter (which contains less sooty material than diesel) and implementation of the system for measurements directly on the exhaust port rather than the dilution tunnel.
Filters were collected for a number of conditions and fuels during testing, and chemical analysis of the samples is under way at the MIT Analytical Chemistry Laboratory. The timing of the work has been somewhat hampered by the planned move of the lab to new quarters.
A comparison of filter mass measurements and SMPS data showed that there is no relationship between the integrated PM measurement from the SMPS and filter mass collected, the reason being that a substantial fraction (10-20 %) of the hydrocarbons present in the engine-out stream (of the order of 1-2% of the total mass of fuel injected) are absorbed onto the filter material, and the mass of these hydrocarbons overwhelms the total collected PM by several orders of magnitude.
No change has been made to the original plan.
David Kayes, who has been responsible for all SMPS measurements and the development of the model, will be concluding his Ph.D. thesis in December 1998. He is planning to conclude the transient analysis in a future publication.
Derek Kim will continue the work on the LII transients through May 1999. We expect the results of the filter analysis from the Analytical Chemistry Lab this Fall.
David Kayes, Derek Kim