- Email: [email protected]
Structural evolution of soot particles during diesel combustion in a single-cylinder light-duty engine
Combustion and Flame 162 (2015) 2720–2728
Contents lists available at ScienceDirect
Combustion and Flame j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e
Structural evolution of soot particles during diesel combustion in a single-cylinder light-duty engine Renlin Zhang, Sanghoon Kook ⇑ School of Mechanical and Manufacturing Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
a r t i c l e
i n f o
Article history: Received 8 April 2015 Received in revised form 9 April 2015 Accepted 9 April 2015 Available online 23 April 2015 Keywords: Diesel combustion Soot morphology TEM Jet–wall interaction
a b s t r a c t The structural evolution of soot particles in a single-cylinder, light-duty diesel engine has been investigated by conducting thermophoretic soot sampling and subsequent transmission electron microscope (TEM) imaging. The location of TEM grid with respect to a diesel flame is varied so that soot particles are sampled at three different combustion stages including (1) when the impingement of flame on the wall occurs, (2) after the flame impingement, and (3) during the late-cycle burn-out. For comparison purposes, engine-out soot particles are also collected at the same operating conditions. The results show that diesel soot particles are aggregates of varying numbers of primary particles. It is found that the flame impingement on the wall makes a significant impact on the soot aggregate structures, evidenced by the decreased mean radius of gyration of the aggregates from 38 to 26 nm between the wall-impinging and post-impingement stages. This was due primarily to the fragmentation of large soot aggregates while the mean diameter of primary particles remains the same. From the post-impingement to late-cycle burn-out stages, most of the soot aggregates disappear due to the oxidation leaving only highly agglomerated substructures. As a result, soot aggregates show higher fractal dimension in the late-cycle burnout stage than that in the previous stages. The intensive soot oxidation also reduces primary particle diameter from 19 to 15 nm through surface oxidation. The engine-out soot samples show similar particle size and structure to the late-cycle samples, suggesting that the late-cycle soot particles experienced little oxidation before exiting through the exhaust. This leads to a conclusion that the highly agglomerated substructures of soot aggregates can survive the late-cycle burn-out and become a major contributor to exhaust soot emissions. Ó 2015 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
1. Introduction Soot particles are a major portion of particulate matter (PM) emissions from diesel engines [1,2] that are strictly regulated due to their negative impacts on the environment and human health [3–7]. While modern common-rail diesel engines achieve very low soot emissions by mass, ultra-fine soot particles from these engines are suspected to be more toxic due to the increased defects in particle nanostructures and higher surface reactivity [8,9]. This issue presents the need for an improved understanding of the size and structure of soot particles during diesel combustion. The improved knowledge about the soot fractal structures would also help develop soot models to clarify underlying physics [10,11]. The size distribution of diesel soot particles is widely investigated using a scanning mobility particle sizer (SMPS) [12–15]. Also, there are many fine papers reporting the morphology of ⇑ Corresponding author. Fax: +61 (0)2 9663 1222. E-mail address: [email protected] (S. Kook).
exhaust soot particles from a tail-pipe particle sampling and transmission electron microscopy (TEM) imaging [16–22]. However, the soot particles in the exhaust stream represent only the product of complex soot processes that involve multi-stage formation and oxidation steps occurring inside the engine cylinder. For example, Tree and Svensson [1] summarised that during diesel combustion, soot is formed from the super-saturated gas-phase precursors in fuel-rich reaction zones and then undergoes nucleation, coalescence, agglomeration as well as oxidation inside engine cylinder before exiting through the exhaust. Therefore, the exhaust soot particles provide limited information about the structural evolution of soot particles during formation and oxidation processes. The in-flame soot is widely investigated using optical/laserbased diagnostics for the measurement of soot area, optical thickness (i.e., KL value), soot volume fraction, and the size of soot particles [23–27]. However, the information about particle structures was very limited until a direct soot particle sampling from a quasi-steady diesel jet flame and subsequent TEM imaging was implemented in a constant-volume combustion vessel [28–31],
http://dx.doi.org/10.1016/j.combustflame.2015.04.008 0010-2180/Ó 2015 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
2721
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
which provided an improved understanding of size and structures of in-flame soot particles. For instance, soot precursor-like structures are found immediate downstream of the flame base while aggregates of soot primary particles with complex fractal structures dominate at the peak soot volume fraction location. In the jet head region, smaller size and simpler structure aggregates are observed, evidencing soot oxidation. Soot morphology was also studied in a working diesel engine using a bulk-gas sampling technique [32–35], which reported the decreased primary particle size and the increased aggregate fractal dimension during diesel combustion. However, this bulk-gas sampling approach raises a question about whether or not the structures of soot particles were affected by the sampling process. Our previous study addressed this issue by placing a TEM grid within the flame for direct sampling of soot particle via thermophoresis (i.e., positive thermal diffusion) [36]. This sampling technique has been used to understand structural changes of in-flame soot particles for various engine operating conditions [37,38]. The present study further utilises this in-flame soot sampling technique to better understand the structural evolution of soot particles occurring inside the cylinder of a diesel engine. Of particular interest is how the flame impingement on the wall impacts the soot particle morphology. The flame–wall interaction is well known to influence local fuel/air mixing and flame temperature significantly which in turn affects combustion and soot processes [39–47]. This question was addressed by sampling the in-flame soot particles at two different combustion stages including (1) when the diesel flame impinges on the wall and (2) after the flame–wall impingement. Moreover, the location of the TEM grid with respect to the flame is varied so that soot particles are sampled in the late-cycle burn-out stage. For comparison purposes, engine-out soot particles are also collected at the same operating conditions.
2. Experiments 2.1. Engine specifications and operating conditions The engine and soot particles sampling system are illustrated in Fig. 1. The specifications and operating conditions are summarised in Table 1. Soot sampling experiments were carried out in an optically accessible, single-cylinder, small-bore diesel engine. Figure 1 shows two soot sampling probes that were used to hold a TEM grid; one probe was installed on the cylinder liner by replacing one of four quartz windows and the other probe was installed in the exhaust manifold. To enable in-flame soot sampling while the engine was running, a portion of the piston bowl-rim (30mm wide) was removed. This was to avoid the potential crash between the sampling probe and fast-moving engine parts such as the piston and intake/exhaust valves. This piston modification resulted in a reduced compression ratio of 15.2, which is still relevant to the production engines of today. The swirl ratio of the engine was fixed at 1.4. Heated water of 90 °C temperature constantly flowed through the cylinder liner and engine head to simulate a thermally-stable, warmed-up engine condition. The engine was naturally aspirated and the air temperature at the intake port was measured at 30 °C throughout the experiments. All tests were conducted at fixed engine speed of 1200 revolution per minute (rpm) using a 37-kW AC motor. A second-generation Bosch common-rail injection system was used to deliver ultra-low-sulphur diesel fuel with cetane number of 51. The original injector had a 7-hole nozzle with the same inter-jet spacing and a 150° included angle. Similar to our previous studies [36,37], the nozzle was modified for single-hole injection by blocking six holes using a laser-welding technique. This
Engine Head
Exhaust Manifold Pressure Sensor
Injector
Valves
Diesel Flame
Quartz/Metal Liner Window
Drop-down Liner Quartz Piston Window Extended Piston
Exhaust Pipe Sampling Probe
TEM Grid Diesel Flame
Bowl-rim cut-out
Exhaust Gas
Sampling Probe
TEM Grid
Fig. 1. Cross-sectional sketch of the diesel engine (top) and the close-up views of the soot sampling regions inside the cylinder (bottom left) and exhaust (bottom right).
Table 1 Engine specifications and operating conditions. Displacement volume Bore/stroke Compression ratio Swirl ratio Coolant temperature Intake air temperature Engine speed Injection system Fuel Cetane number Nozzle hole diameter Mass per injection Injection pressure Injection timing
498 cm3 (single cylinder) 83 mm/92 mm 15.2 1.4 90 °C 30 °C 1200 rpm Bosch second-generation common-rail injector Ultra-low-sulphur diesel 51 134 lm (nominal) 9 mg 70 MPa 7°CA aTDC
approach was to isolate a single diesel jet from complex jet–jet interactions and at the same time to allow for long injection duration while keeping the in-cylinder pressure below the burst pressure of the quartz windows. The injection duration of 2.34 ms (actual) was selected for all tested conditions in the present study. At 70-MPa injection pressure, this injection duration corresponds to 9 mg of diesel fuel per injection for a single hole. If all 7 holes were used, the injected mass would be relevant to upper-mid to high-load conditions where soot emissions are most problematic. The injection timing was fixed at 7 crank angles after the top dead centre of the compression stroke (°CA aTDC). Throughout the experiments, the in-cylinder phenomena were monitored by measuring in-cylinder pressure at various crank angle locations using a piezo-electric pressure transducer (Kistler 6056A). The measured in-cylinder pressure traces were used to calculate
2722
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
apparent heat release rate (aHRR) traces using a simple energy balance equation [48]. 2.2. Soot particle sampling Soot particles were sampled using two in-house designed probes, each holding a 3-mm diameter, 400-mesh, carbon-coated TEM grid [36]. Soot particles were deposited onto the TEM grid via thermophoretic force created by a temperature gradient between hot soot laden gas and the carbon film on the TEM grid. For each sampling experiment, a new TEM grid was placed on the tip of the sampling probe prior to the engine running. The engine was motored for about 90 s to achieve thermally stable in-cylinder conditions that are evaluated by the peak motored pressure. Then the fuel injection was executed every 10th cycle (i.e., 9 motoring and 1 injection cycles) to reduce the thermal loading on quartz windows and to expel the residual gases from a previous firing cycle. The engine was run continuously until the predetermined number of fuel injections was conducted. The total number of fuel injections was varied depending on the soot concentration of the sampling location. For example, five injections were executed for the flame–wall impingement and post-impingement stages whereas twenty injections were used for the late-cycle and exhaust samples. These injection numbers were enough to sample many soot particles for statistically meaningful data analysis while avoiding a potential soot overloading issue. Detailed discussions about the effect of number of injections as well as the cyclic dispersion on the sampled soot particles are found in our previous studies [36,38]. Once this pre-determined number of fuel injections was reached, the engine was stopped and then the TEM grid was removed from the sampling probe. The grid was carefully stored before it was brought to a TEM to obtain multiple images for various on-grid locations. Figure 2 (top row) illustrates four different combustion stages investigated in the present study. These include (1) when the impingement of flame on the wall occurs (flame–wall impingement), (2) after the flame–wall impingement (post-impingement) and (3) at the late-cycle burn-out stage (late-cycle). Soot particles were also collected in the exhaust pipe: (4) exhaust. As mentioned previously, the sampling probe location was fixed within the bowlrim cut-out region due to the risk of crash. Therefore, the nozzlehole orientation of the fuel injectors was adjusted so that the fuel jet trajectory was varied with respect to the sampling probe. This approach is illustrated in Fig. 2 (bottom row). For example, the flame–wall impingement case utilised the 9 o’clock hole in the field of view and thus the TEM grid location was close to the wall-impingement region of the sooting flame. This was a high sooting region [45,46] where soot particles were expected to be mature solid particles with various fractal structures [28,29]. The post-impingement soot particles were sampled by using the 10 o’clock hole so that the sooting flame impinges on the wall and then travels along the liner wall before the particles deposit on the TEM grid. Due to the improved mixing associated with a turbulent ring-vortex formed in the wall–jet head region [46,47,49], the soot particles were expected to show evidence of strong oxidation. For the late-cycle soot particles, the same 10 o’clock hole was used while the piston was rotated 180° so that the TEM grid was exposed to the sooting flame only during the late-cycle burn-out. The late-cycle and exhaust particle sampling experiments were conducted simultaneously. It should be noted that the fixed sampling probe approach of this study does not provide information about soot particles at a certain instant but time-integrated information for the duration that the TEM grid is exposed to the sooting diesel flame. However, for the flame-impingement and post-impingement cases, the exposure time estimated by the high-speed movie of soot
luminosity was very short at 1–3 ms [37], which is comparable to a quick-insertion-type sampling approach widely used in open flame burners. The corresponding still images from high-speed soot luminosity movies are shown in Fig. 3. The figure shows the temporal evolution of hot soot luminosity during flame–wall impingement, post-impingement and late-cycle combustion stages. Images were taken using two high-speed CMOS cameras (VisionResearch Phantom v7.3) at 36,000 frames per second for the bottom-view and side-view orientations. The timing of each image set is noted as °CA aTDC and °CA aEOI (after the end of injection). The locations of bowl-rim cut-out and the soot sampling probe are illustrated in each image. The leftmost images for the flame–wall impingement and postimpingement combustion stages show the first frame with visible soot luminosity. The following images show the time when the soot clouds reach the location of the TEM grid (short red bar) indicating the start of soot particles collection. It is observed that the soot sampling for the flame–wall impingement and post-impingement stages started roughly at 10.4 and 12.4°CA aTDC. As for luminosity images during late-cycle burn-out stage, the first two side-view images are completely dark due the obstruction of the piston bowl-rim (the imaging was done from the opposite end of the bowl-rim cut-out). Judging from the bottom-view images and the third side-view image, the time when the TEM grid is first exposed to the soot luminosity should be between 24 and 28°CA aTDC. 2.3. TEM imaging and imaging post-processing Soot particles imaging was performed using a JEOL 1400 TEM with 0.38 nm point resolution and 100 kV acceleration voltage. An 11-mega pixel Gatan CCD camera was used to digitise the magnified soot particle images. For each soot sample, multiple TEM images were taken at various on-grid locations considering the fluctuations of sampled soot particles due to the inhomogeneous nature of fuel–air mixture. In addition, multiple TEM images allowed for high number of soot particles and thereby increasing the credibility of statistical analysis. The magnification was set at 50,000 times considering both the number of particles per image and the image quality. The TEM images were post-processed using an in-house-developed Matlab code [50] for the particle number count, the projection area of soot particles, the primary particle diameter (dp) and the radius of gyration of soot aggregates (Rg). These data were used to calculate the fractal dimension (Df) and prefactor (kf) of soot aggregates. More details about the data processing and uncertainty analysis are found in our previous work [36]. Depending on the sampling stage, the number of processed soot aggregates varied from 129 to 996 while the number of processed primary particles ranged from 3640 to 15,579 in the present study. 3. Results and discussion 3.1. In-cylinder pressure and apparent heat release rate The averaged in-cylinder pressure traces and corresponding aHRR for three different sampling locations are shown in Fig. 4. Due to three different nozzle-hole orientations (Fig. 2), the soot particles sampling was conducted for different engine runs, which raised a question on run-to-run as well as hole-to-hole variations. However, Fig. 4 shows that the in-cylinder phenomenon does not vary significantly. The level of variations are well within cycleto-cycle variations [36], suggesting that the soot particles were sampled at fixed in-cylinder conditions. On the aHRR traces (Fig. 4 bottom), the start of soot sampling is illustrated for each case which confirms that the flame–wall impingement and
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
2723
Late-cycle
Post-impingement
Flame-wall Impingement
Fig. 2. Illustrations of sooting diesel flame evolution (top) and corresponding sampling approaches (bottom).
Distance from nozzle [mm] Fig. 3. Temporal evolution of hot soot luminosity during the diesel combustion event corresponding to the soot sampling experiments at various combustion stages of the present study. Both bottom-view (top) and side-view (bottom) images are shown at various °CA after TDC (aTDC) and °CA after the end of injection (aEOI). The field of view is indicated by dashed-line circles. Yellow dashed-lines indicate the jet centre line. The TEM grid positions (red bars) and limits of bowl-rim cut-out (green bars) are shown in both the bottom view and side view images. Numbers (mm) indicates the distance from the tip of injector nozzle.
2724
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 4. Averaged in-cylinder pressure and corresponding apparent heat release rate traces for various crank angle locations. Three traces correspond to three separate sampling experiments that designed to collect soot at various combustion stages. The approximate crank angle locations of the start of soot collection are marked as circles on the heat release rate traces.
post-impingement samplings were conducted during the main combustion event. For the late-cycle sampling, the soot particles were collected during the burn-out stage of diesel combustion. The exhaust soot sampling was conducted between 128 and 366°CA aTDC when the exhaust valves were opened.
3.2. TEM images Selected TEM images of soot particles collected from various combustion stages are shown in Fig. 5. Shown on the bottom-left corner of each image is a 200-nm scale bar. It is seen that the majority of soot particles are aggregates comprised of many primary particles with near-circular shape, consistent with previous studies [32,36,37]. Particularly in the flame–wall impingement image, many agglomerated particles with various structures are observed. This was expected because fuel rich mixtures would be formed near the flame impingement region due to the limited fuel–air mixing [45,46] and thereby promoting the soot formation [51]. It is however noted that the maximum soot coverage ratio over the TEM image was around 15%, which is close to the value suggested by Megaridis and Dobbins [52] (10%) to avoid the particle overloading and associated issue of the particle agglomeration on deposited particles. The change in the number concentration of soot particles and overall soot amount between earlier and later stages is striking. Figure 5 shows that late-cycle and exhaust soot images present much lower number of soot aggregates than the flame–wall impingement and post-impingement images. Specifically, soot aggregates that are larger than the 200-nm scale
bar are easily seen in the flame–wall impingement and postimpingement samples but do not exist in the late-cycle and exhaust soot samples. Furthermore, single-primary particles and small aggregates with only a few primary particles shown in the flame–wall impingement and post-impingement samples are rare in the late-cycle and exhaust samples. Structural difference is also significant. In the first two combustion stages, large soot aggregates appear to be comprised of many primary particles forming stretched chain-like branches. It is also noticeable that the branches are attached to highly concentrated core regions of the aggregates. Such structures are common for peak soot location of a free (no wall impingement) diesel jet [29]. By contrast, most of the soot aggregates at later stages show compact and agglomerated aggregate structures. The variations in soot particles between each combustion stage are also observed. For instance, a closer look at the TEM images for the flame–wall impingement and post-impingement samples suggests the reduction in soot particles, particularly large soot aggregates. This was explained by the evolution of soot particles along the flame trajectory (i.e. increasing distance from the nozzle). Previous studies conducted in a constant-volume combustion chamber where the soot particles sampling was conducted for a free (no wall) flame reported a similar trend of reduced large soot aggregates further downstream of the flame due to soot oxidation [28–30]. Indeed, the estimated flame travel distance between the nozzle hole and TEM grid was 42 and 60 mm for the flame–wall impingement and post-impingement, respectively, which might suggest soot particles oxidation. However, the increasing distance from the nozzle was not the only cause for the reduction of large soot particles. It was likely that the flame–wall interactions also impacted the soot particles through the enhanced mixing in the jet head-vortex region [45], which would promote the soot particles oxidation. Unlike the evident changes found between the first two stages, the soot particles in the exhaust appear to be similar to the latecycle sample. This might suggest the origin of exhaust soot particles such that those particles survived the oxidation processes during the main combustion would exit through the exhaust. An interesting observation from the late-cycle and exhaust samples is transparent objects with a size of tens of nanometres, which are illustrated by rectangles in Fig. 5. The cause for these structures is not entirely clear but they could be condensed aliphatic compounds as a result of decreased gas temperature in the expansion stroke [53]. In this study, these crystalline-like particles were not included in the image processing. It should be noted that the TEM images in Fig. 5 are a result of various numbers of fuel injections depending on the sooting level of each combustion stage. This was to guarantee high enough number of soot particles on the TEM images for statistical analysis. For instance, a total of 20 injections were executed for the late-cycle and exhaust soot samples whereas only 5 injections were applied to the flame–wall impingement and post-impingement samples. Despite fourfold higher number of fuel injections, the soot particles in the late-cycle and exhaust samples are much less, evidencing that the soot reduction shown in Fig. 5 is real. 3.3. Size distribution of soot particles Figure 6 shows histograms of the radius of gyration (Rg) of soot particles (including single-primary particles and multi-primary aggregates) at various combustion stages. Noted next to each histogram are the total number of soot particles, the number of soot particles per TEM image, and the projection area of soot particles on the TEM images. Between the flame–wall impingement and post-impingement samples, it is shown that the number of soot particles is higher but the projection area is lower for the
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
2725
Fig. 5. Example TEM images of soot particles sampled at various combustion stages. The scale bars of 200 nm are shown as red horizontal bars in the bottom-left corner of each image. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
post-impingement stage. As mentioned previously, the observed trend was expected because the soot particles would be oxidised due to the increased distance from the nozzle [28–30] and the enhanced mixing in the jet head-vortex region [45,46,54,55]. As a result, the overall soot amount decreased in the post-impingement sample. Also, it could be interpreted that those large soot aggregates observed in the flame–wall impingement sample would disintegrate into multiple smaller aggregates due to oxidationinduced fragmentation [56], leading to the increased number of small soot aggregates (Rg < 25 nm). In comparison, late-cycle and exhaust soot samples show low particle number counts and reduced projection area. Although some small particles (Rg < 10 nm) were still observable, a significant portion of the soot particles at these two stages have Rg higher than 10 nm. This supports the earlier interpretation of the TEM images in Fig. 5 that the small particles were easily oxidised and those highly-concentrated
core regions of large aggregates remained in the late-cycle and exhaust stages. Due to significantly lower sample size, one might question the effectiveness of thermophoretic sampling for exhaust soot particles in the present study. For the first three combustion stages, the soot luminosity images confirmed the direct exposure of the TEM grid to a sooting flame (Fig. 3). However, the exhaust gas temperature could be much lower than the in-cylinder flames and thus resulting in significantly weaker thermophoresis compared to the in-flame sampling. Indeed, significant reduction in a temperature gradient between exhaust gas and TEM grid is expected due to cylinder expansion and exhaust valve opening. However, the overall particle size distribution and particle number density of the exhaust soot sample are similar to that of the late-cycle sample, suggesting that the reduction in the temperature gradient does not have significant impacts on particle deposition process in the
2726
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 6. Histogram of Rg of soot particles at various combustion stages. Total particle number counts, number counts of particles per TEM image and particle projection area per TEM image are also noted. Error ranges are estimated with 95% confidence.
exhaust. Previous exhaust soot studies reported large aggregates with complex fractal structures that are similar to the soot aggregates shown in the flame–wall impingement or post-impingement samples [17,19–21]. It was believed that this discrepancy was due to very low-sooting operating conditions of this study. More specifically, the single-cylinder operation, the use of single-hole nozzle injector, skip-firing and no exhaust gas recirculation (EGR) in the present study produced much less soot particles than multicylinder engines used in the exhaust soot studies. Furthermore, Fig. 6 shows that the number concentration and projection area of the exhaust sample is roughly 0.5% of the flame–wall impingement sample, if four times more injections were considered. This is very consistent with the previous study [33] suggesting that more than 99.5% of soot formed inside the engine cylinder is oxidised before they exit through the exhaust. Therefore, it can be concluded that the low sample size in the exhaust was not due the experimental error but a result of low soot level of the tested engine and selected operating conditions. Figure 7 shows the probability density function (pdf) of Rg and dp for all four combustion stages. The mean values and error ranges are also annotated next to the corresponding plot. Figure 6 (top) shows that the mean Rg is around 38 nm for the flame–wall impingement soot particles, which decreases to about 26 nm for the post-impingement soot particles. As previously mentioned, this shift towards lower Rg regime can be explained by the increased number of small aggregates due to the fragmentation of large aggregates. By contrast, the pdf of dp shows identical distribution with a mean value of about 20 nm between the flame–wall impingement and post-impingement soot particles (Fig. 7 bottom). This suggests that the primary particles at these combustion stages are mature solid particles due to the balance between the surface growth and oxidation. Compared to the post-impingement sample, the late-cycle soot particles show higher mean Rg and a significant decrease of small soot aggregates (Rg < 50 nm), implying intensive soot oxidation during late-cycle burn-out. Also, the mean dp of the late-cycle sample is 15.5 nm, 4.2 nm lower than the post-impingement sample, suggesting surface oxidation of the primary particles [32]. The exhaust soot particles appear to be further oxidised from the late-cycle soot particles, evidenced by slightly lower Rg and dp. Overall, the Rg and dp of the late-cycle and exhaust samples suggest that those highly concentrated substructures of soot aggregates [40,51,57] which could not be completely oxidised during the late-cycle burn-out stage would exit through the exhaust. It should be noted that the size of exhaust soot particles in the current
Fig. 7. Probability density functions of radius of gyration (Rg) of aggregates (top) and diameter (dp) of primary particles (bottom) for soot particles at various stages. The mean values and error ranges (95% confidence) are noted.
studies is smaller than the ones found in previous studies [17,19–21]. As discussed earlier in this section, this could be the results of low-sooting operating conditions that implemented in the present study. 3.4. Fractal morphology of soot aggregates The fractal morphology of soot aggregates can be characterised by the mass fractal relation shown in Eq. (1) [58]:
N ¼ kf
Rg
!D f ð1Þ
dp
where N is the number of primary particles within a soot aggregate and dp is the mean diameter of primary particles in each aggregate. The fractal dimension Df represents the structural compactness of a fractal aggregate such that higher Df means more compacted fractal morphology. The kf is the fractal prefactor to estimate the optical and transport properties of soot particles [59]. As suggested in Eq. (1), Df and kf can be obtained by reading the slope and the intercept of the least square fit line of the logarithm plot ln(N) over ln(Rg/dp ). In this study, N was determined using Eq. (2):
N ¼ ka
Aa Ap
a ð2Þ
where Aa is the projection area of a soot aggregate, and Ap is the mean projection area of primary particles within the aggregate.
2727
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 8. Statistical determination of soot aggregate fractal dimension (Df) and prefactors (kf) using the number of primary particles (N), the radius of gyration (Rg) and the mean primary particle diameter of each soot aggregate (dp ). Error ranges are estimated with 95% confidence.
Flame-wall Impingement
Post-impingement
Late-cycle
Exhaust
Fig. 9. Illustration of soot particles evolution during diesel combustion in a light-duty engine.
The empirical constant ka and primary particle overlap factor a equal to 1 and 1.09, respectively [52]. Figure 8 shows the log–log plot of Df and kf for all four combustion stages of the present study. The Df and kf values together with error ranges (95% confidence) are noted in each plot. It should be noted that Df was calculated only for the soot aggregates with more than three primary particles (N > 3) because monomers (N = 1) are not aggregates and the aggregates with only two or three primary particles (N = 2 or 3) hardly have fractal structures. It is observed from Fig. 7 that the fractal dimension measured in this study ranges between 1.7 and 1.9, which is consistent with previous studies conducted in both engine exhaust [17,19–21] and constant-volume combustion chambers [29–31]. This suggests that soot formation in diesel combustion falls into the diffusion limited cluster–cluster aggregation regime [52]. It is also seen that fractal dimensions are very similar between the flame–wall impingement and post-impingement samples. Considering the increased number of small soot particles (Figs. 5 and 6) in the post-impingement samples, the similarity in fractal dimension was not expected. It could be explained that the decrease in Df associated with the stretched chain-like branches cancelled out the increase in Df due to the compact structures of the remaining cores. The late-cycle and exhaust samples show the same Df that
are higher than the flame–wall impingement and post-impingement samples. The increased fractal dimension implies that the aggregates became more compact, likely due to the soot oxidation occurring in the branches as well as in the primary particles at the outskirt of the remaining cores [32]. Some studies [52,60] suggested that the aggregation of soot particles begins with monomer–monomer interactions which are followed by the monomer-cluster interactions, leading to small and compact clusters (Df 3). Thereafter, the clusters join other clusters through cluster–cluster aggregation to form larger aggregates with long stretched fractal morphology (Df < 2). Surface growth would occur throughout the process to add mass and dimension to the aggregates. Also, one study [61] reported that large clusters tend to breakdown into smaller aggregates during the oxidation process and shrink in size. The results of the present study are consistent with the previous studies, suggesting fragmentation and de-generation process of soot aggregates due to the flame–wall interaction and late-cycle burn-out. 4. Conclusion An experimental investigation of the structural evolution of soot particles has been conducted in a single-cylinder light-duty
2728
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
diesel engine. Thermophoretic soot sampling was used to collect soot particle samples from various combustion stages. Collected soot samples were imaged using a TEM. The TEM soot images were post-processed to obtain soot particle number concentration, size distribution and fractal morphology. The main findings of the present study are summarised by illustrating the soot particle structures at various diesel combustion stages as shown in Fig. 9. It is understood that some vulnerable soot particles (light grey circles) such as single-primaries, small aggregates and some weak linking primary particles in large aggregates are oxidised during flame–wall interactions, leading to the fragmentation and shrinking of soot aggregates. This results in the increased number of small particles in the post-impingement stage. The partially oxidised soot particles then undergo further oxidation in the latecycle burn-out stage, leading to more compacted soot particles through the detachment of the chain-like branches from the large aggregates. The de-generation also reduces the size of primary particles therein. However, the large aggregates comprised of highly-concentrated primary particles (black circles) survive the soot oxidation. Between the late-cycle and exhaust stages, there is not much of further oxidation due to significantly reduced gas temperature. Finally, those remaining soot particles from the late-cycle burn-out stage exit through the exhaust. Acknowledgments Experiments were performed at the UNSW Engine Research Laboratory, Sydney, Australia. Support for this research was provided by the Australian Research Council via Discovery Project. References [1] D.R. Tree, K.I. Svensson, Proc. Combust. Inst. 33 (2007) 272–309. [2] I. Glassman, Symp. (Int.) Combust. 22 (1989) 295–311. [3] J.S. Lighty, J.M. Veranth, A.F. Sarofim, J. Air Waste Manage. Assoc. 50 (2000) 1565–1618. [4] V. Muzyka, S. Veimer, N. Schmidt, Sci. Total Environ. 217 (1998) 103–111. [5] D.M. Broday, R. Rosenzweig, J. Aerosol Sci. 42 (2011) 372–386. [6] R. Zhang, A.F. Khalizov, J. Pagels, D. Zhang, H. Xue, P.H. McMurry, Proc. Natl. Acad. Sci. USA 105 (2008) 10291–10296. [7] L. Benbrahim-Tallaa, R.A. Baan, Y. Grosse, B. Lauby-Secretan, F. El Ghissassi, V. Bouvard, N. Guha, D. Loomis, K. Straif, V.M. Arlt, Lancet Oncol. 13 (2012) 663– 664. [8] B. Frank, R. Schlögl, D.S. Su, Environ. Sci. Technol. 47 (2013) 3026–3027. [9] D.S. Su, A. Serafino, J.O. Müller, R.E. Jentoft, R. Schlögl, S. Fiorito, Environ. Sci. Technol. 42 (2008) 1761–1765. [10] M.E. Mueller, G. Blanquart, H. Pitsch, Proc. Combust. Inst. 32 (2009) 785–792. [11] T. Li, H. Ogawa, SAE Int. J. Engines 5 (2011) 94–101. [12] I.S. Abdul-Khalek, D.B. Kittelson, B.R. Graskow, Q. Wei, F. Bear, SAE Technical Paper 980525, 1998. [13] S.S. Gill, D. Turner, A. Tsolakis, A.P.E. York, Environ. Sci. Technol. 46 (2012) 4215–4222. [14] H. Kim, Y. Sung, K. Jung, B. Choi, M. Lim, J. Mech. Sci. Technol. 22 (2008) 1793– 1799.
[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61]
R. Zhu, H. Miao, X. Wang, Z. Huang, Proc. Combust. Inst. 34 (2013) 3013–3020. K. Al-Qurashi, A.L. Boehman, Combust. Flame 155 (2008) 675–695. A. Neer, U.O. Koylu, Combust. Flame 146 (2006) 142–154. U. Mathis, M. Mohr, R. Kaegi, A. Bertola, K. Boulouchos, Environ. Sci. Technol. 39 (2005) 1887–1892. J.P. Shi, D. Mark, R.M. Harrison, Environ. Sci. Technol. 34 (2000) 748–755. J. Zhu, K.O. Lee, A. Yozgatligil, M.Y. Choi, Proc. Combust. Inst. 30 (2005) 2781– 2789. K.O. Lee, J. Zhu, SAE Technical Paper 2004-01-1981, 2004. K. Yehliu, O. Armas, R.L. Vander Wal, A.L. Boehman, Combust. Flame 160 (2013) 682–691. L.M. Pickett, D.L. Siebers, Combust. Flame 138 (2004) 114–135. E. Cenker, G. Bruneaux, L.M. Pickett, C. Schulz, SAE Int. J. Engines 6 (2013) 352– 365. J. Manin, L.M. Pickett, S.A. Skeen, SAE Int. J. Engines 6 (2013) 1908–1921. R. Ryser, T. Gerber, T. Dreier, Combust. Flame 156 (2009) 120–129. U. Aronsson, C. Chartier, O. Andersson, B. Johansson, J. Sjöholm, R. Wallander, M. Richter, M. Alden, SAE Int. J. Engines 3 (2010) 137–155. S. Kook, L.M. Pickett, Proc. Combust. Inst. 33 (2011) 2911–2918. S. Kook, L.M. Pickett, SAE Int. J. Fuels Lubr. 5 (2012) 647–664. K. Kondo, T. Yamaguchi, H. Nishigai, S. Takano, T. Aizawa, SAE Technical Paper 2011-24-0068, 2011. T. Yamaguchi, K. Kondo, H. Nishigai, S. Takano, T. Aizawa, SAE Int. J. Fuels Lubr. 5 (2011) 316–328. Z. Li, C. Song, J. Song, G. Lv, S. Dong, Z. Zhao, Combust. Flame 158 (2011) 1624– 1630. A. Pungs, S. Pischinger, H. Bäcker, G. Lepperhoff, SAE Technical Paper 2000-011999, 2000. H. Bäcker, O. Hild, G. Lepperhoff, SAE Technical Paper 970875, 1997. Y. Aoyagi, T. Kamimoto, Y. Matsui, S. Matsuoka, SAE Technical Paper 800254, 1980. S. Kook, R. Zhang, K. Szeto, L.M. Pickett, T. Aizawa, SAE Int. J. Fuels Lubr. 6 (2013) 80–97. R. Zhang, S. Kook, Environ. Sci. Technol. 48 (2014) 8243–8250. R. Zhang, K. Szeto, S. Kook, SAE Int. J. Fuels Lubr. 6 (2013) 605–614. G. Bruneaux, Int. J. Engine Res. 9 (2008) 249–265. J.E. Dec, D.R. Tree, SAE Technical Paper 2001-01-1295, 2001. A. Rusly, M.K. Le, S. Kook, SAE Int. J. Fuels Lubr. 6 (2013) 615–626. A. Rusly, M.K. Le, S. Kook, E.R. Hawkes, Fuel 125 (2014) 1–14. F.R. Held, S. Bornschlegel, S. Riess, T. Vogel, M. Wensing, SAE Technical Paper 2013-24-0063, 2013. L. Song, J. Abraham, J. Fluids Eng. 125 (2004) 997–1005. G. Bruneaux, SAE Technical Paper 2005-01-2097, 2005. L.M. Pickett, J.J. López, SAE Technical Paper 2005-01-0921, 2005. M.K. Le, S. Kook, E.R. Hawkes, Fuel 140 (2015) 143–155. J.B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill Book Co., USA, 1988. R. Solsjo, M. Jangi, C. Chartier, O. Andersson, X. Bai, SAE Technical Paper 201401-1127, 2014. T. Aizawa, H. Nishigai, K. Kondo, T. Yamaguchi, J.G. Nerva, C. Genzale, S. Kook, L.M. Pickett, SAE Int. J. Fuels Lubr. 5 (2012) 665–673. D.R. Tree, J.E. Dec, SAE Technical Paper 2001-01-1926, 2001. C.M. Megaridis, R.A. Dobbins, Combust. Sci. Technol. 71 (1990) 95–109. H. Wang, Proc. Combust. Inst. 33 (2011) 41–67. G. Bruneaux, SAE Technical Paper 2005-01-2100, 2005. K. Inagaki, S. Takasu, K. Nakakita, SAE Technical Paper 1999-01-0508, 1999. M.E. Mueller, G. Blanquart, H. Pitsch, Proc. Combust. Inst. 33 (2011) 667–674. J.E. Dec, P.L. Kelly-Zion, SAE Technical Paper 2000-01-0238, 2000. R. Botet, R. Jullien, Ann. Phys. Fr. 13 (1988) 153–221. C.M. Sorensen, G.C. Roberts, J. Colloid Interface Sci. 186 (1997) 447–452. P. Meakin, J. Colloid Interface Sci. 102 (1984) 505–512. H. Jung, D.B. Kittelson, M.R. Zachariah, Combust. Flame 136 (2004) 445–456.
Contents lists available at ScienceDirect
Combustion and Flame j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e
Structural evolution of soot particles during diesel combustion in a single-cylinder light-duty engine Renlin Zhang, Sanghoon Kook ⇑ School of Mechanical and Manufacturing Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
a r t i c l e
i n f o
Article history: Received 8 April 2015 Received in revised form 9 April 2015 Accepted 9 April 2015 Available online 23 April 2015 Keywords: Diesel combustion Soot morphology TEM Jet–wall interaction
a b s t r a c t The structural evolution of soot particles in a single-cylinder, light-duty diesel engine has been investigated by conducting thermophoretic soot sampling and subsequent transmission electron microscope (TEM) imaging. The location of TEM grid with respect to a diesel flame is varied so that soot particles are sampled at three different combustion stages including (1) when the impingement of flame on the wall occurs, (2) after the flame impingement, and (3) during the late-cycle burn-out. For comparison purposes, engine-out soot particles are also collected at the same operating conditions. The results show that diesel soot particles are aggregates of varying numbers of primary particles. It is found that the flame impingement on the wall makes a significant impact on the soot aggregate structures, evidenced by the decreased mean radius of gyration of the aggregates from 38 to 26 nm between the wall-impinging and post-impingement stages. This was due primarily to the fragmentation of large soot aggregates while the mean diameter of primary particles remains the same. From the post-impingement to late-cycle burn-out stages, most of the soot aggregates disappear due to the oxidation leaving only highly agglomerated substructures. As a result, soot aggregates show higher fractal dimension in the late-cycle burnout stage than that in the previous stages. The intensive soot oxidation also reduces primary particle diameter from 19 to 15 nm through surface oxidation. The engine-out soot samples show similar particle size and structure to the late-cycle samples, suggesting that the late-cycle soot particles experienced little oxidation before exiting through the exhaust. This leads to a conclusion that the highly agglomerated substructures of soot aggregates can survive the late-cycle burn-out and become a major contributor to exhaust soot emissions. Ó 2015 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
1. Introduction Soot particles are a major portion of particulate matter (PM) emissions from diesel engines [1,2] that are strictly regulated due to their negative impacts on the environment and human health [3–7]. While modern common-rail diesel engines achieve very low soot emissions by mass, ultra-fine soot particles from these engines are suspected to be more toxic due to the increased defects in particle nanostructures and higher surface reactivity [8,9]. This issue presents the need for an improved understanding of the size and structure of soot particles during diesel combustion. The improved knowledge about the soot fractal structures would also help develop soot models to clarify underlying physics [10,11]. The size distribution of diesel soot particles is widely investigated using a scanning mobility particle sizer (SMPS) [12–15]. Also, there are many fine papers reporting the morphology of ⇑ Corresponding author. Fax: +61 (0)2 9663 1222. E-mail address: [email protected] (S. Kook).
exhaust soot particles from a tail-pipe particle sampling and transmission electron microscopy (TEM) imaging [16–22]. However, the soot particles in the exhaust stream represent only the product of complex soot processes that involve multi-stage formation and oxidation steps occurring inside the engine cylinder. For example, Tree and Svensson [1] summarised that during diesel combustion, soot is formed from the super-saturated gas-phase precursors in fuel-rich reaction zones and then undergoes nucleation, coalescence, agglomeration as well as oxidation inside engine cylinder before exiting through the exhaust. Therefore, the exhaust soot particles provide limited information about the structural evolution of soot particles during formation and oxidation processes. The in-flame soot is widely investigated using optical/laserbased diagnostics for the measurement of soot area, optical thickness (i.e., KL value), soot volume fraction, and the size of soot particles [23–27]. However, the information about particle structures was very limited until a direct soot particle sampling from a quasi-steady diesel jet flame and subsequent TEM imaging was implemented in a constant-volume combustion vessel [28–31],
http://dx.doi.org/10.1016/j.combustflame.2015.04.008 0010-2180/Ó 2015 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
2721
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
which provided an improved understanding of size and structures of in-flame soot particles. For instance, soot precursor-like structures are found immediate downstream of the flame base while aggregates of soot primary particles with complex fractal structures dominate at the peak soot volume fraction location. In the jet head region, smaller size and simpler structure aggregates are observed, evidencing soot oxidation. Soot morphology was also studied in a working diesel engine using a bulk-gas sampling technique [32–35], which reported the decreased primary particle size and the increased aggregate fractal dimension during diesel combustion. However, this bulk-gas sampling approach raises a question about whether or not the structures of soot particles were affected by the sampling process. Our previous study addressed this issue by placing a TEM grid within the flame for direct sampling of soot particle via thermophoresis (i.e., positive thermal diffusion) [36]. This sampling technique has been used to understand structural changes of in-flame soot particles for various engine operating conditions [37,38]. The present study further utilises this in-flame soot sampling technique to better understand the structural evolution of soot particles occurring inside the cylinder of a diesel engine. Of particular interest is how the flame impingement on the wall impacts the soot particle morphology. The flame–wall interaction is well known to influence local fuel/air mixing and flame temperature significantly which in turn affects combustion and soot processes [39–47]. This question was addressed by sampling the in-flame soot particles at two different combustion stages including (1) when the diesel flame impinges on the wall and (2) after the flame–wall impingement. Moreover, the location of the TEM grid with respect to the flame is varied so that soot particles are sampled in the late-cycle burn-out stage. For comparison purposes, engine-out soot particles are also collected at the same operating conditions.
2. Experiments 2.1. Engine specifications and operating conditions The engine and soot particles sampling system are illustrated in Fig. 1. The specifications and operating conditions are summarised in Table 1. Soot sampling experiments were carried out in an optically accessible, single-cylinder, small-bore diesel engine. Figure 1 shows two soot sampling probes that were used to hold a TEM grid; one probe was installed on the cylinder liner by replacing one of four quartz windows and the other probe was installed in the exhaust manifold. To enable in-flame soot sampling while the engine was running, a portion of the piston bowl-rim (30mm wide) was removed. This was to avoid the potential crash between the sampling probe and fast-moving engine parts such as the piston and intake/exhaust valves. This piston modification resulted in a reduced compression ratio of 15.2, which is still relevant to the production engines of today. The swirl ratio of the engine was fixed at 1.4. Heated water of 90 °C temperature constantly flowed through the cylinder liner and engine head to simulate a thermally-stable, warmed-up engine condition. The engine was naturally aspirated and the air temperature at the intake port was measured at 30 °C throughout the experiments. All tests were conducted at fixed engine speed of 1200 revolution per minute (rpm) using a 37-kW AC motor. A second-generation Bosch common-rail injection system was used to deliver ultra-low-sulphur diesel fuel with cetane number of 51. The original injector had a 7-hole nozzle with the same inter-jet spacing and a 150° included angle. Similar to our previous studies [36,37], the nozzle was modified for single-hole injection by blocking six holes using a laser-welding technique. This
Engine Head
Exhaust Manifold Pressure Sensor
Injector
Valves
Diesel Flame
Quartz/Metal Liner Window
Drop-down Liner Quartz Piston Window Extended Piston
Exhaust Pipe Sampling Probe
TEM Grid Diesel Flame
Bowl-rim cut-out
Exhaust Gas
Sampling Probe
TEM Grid
Fig. 1. Cross-sectional sketch of the diesel engine (top) and the close-up views of the soot sampling regions inside the cylinder (bottom left) and exhaust (bottom right).
Table 1 Engine specifications and operating conditions. Displacement volume Bore/stroke Compression ratio Swirl ratio Coolant temperature Intake air temperature Engine speed Injection system Fuel Cetane number Nozzle hole diameter Mass per injection Injection pressure Injection timing
498 cm3 (single cylinder) 83 mm/92 mm 15.2 1.4 90 °C 30 °C 1200 rpm Bosch second-generation common-rail injector Ultra-low-sulphur diesel 51 134 lm (nominal) 9 mg 70 MPa 7°CA aTDC
approach was to isolate a single diesel jet from complex jet–jet interactions and at the same time to allow for long injection duration while keeping the in-cylinder pressure below the burst pressure of the quartz windows. The injection duration of 2.34 ms (actual) was selected for all tested conditions in the present study. At 70-MPa injection pressure, this injection duration corresponds to 9 mg of diesel fuel per injection for a single hole. If all 7 holes were used, the injected mass would be relevant to upper-mid to high-load conditions where soot emissions are most problematic. The injection timing was fixed at 7 crank angles after the top dead centre of the compression stroke (°CA aTDC). Throughout the experiments, the in-cylinder phenomena were monitored by measuring in-cylinder pressure at various crank angle locations using a piezo-electric pressure transducer (Kistler 6056A). The measured in-cylinder pressure traces were used to calculate
2722
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
apparent heat release rate (aHRR) traces using a simple energy balance equation [48]. 2.2. Soot particle sampling Soot particles were sampled using two in-house designed probes, each holding a 3-mm diameter, 400-mesh, carbon-coated TEM grid [36]. Soot particles were deposited onto the TEM grid via thermophoretic force created by a temperature gradient between hot soot laden gas and the carbon film on the TEM grid. For each sampling experiment, a new TEM grid was placed on the tip of the sampling probe prior to the engine running. The engine was motored for about 90 s to achieve thermally stable in-cylinder conditions that are evaluated by the peak motored pressure. Then the fuel injection was executed every 10th cycle (i.e., 9 motoring and 1 injection cycles) to reduce the thermal loading on quartz windows and to expel the residual gases from a previous firing cycle. The engine was run continuously until the predetermined number of fuel injections was conducted. The total number of fuel injections was varied depending on the soot concentration of the sampling location. For example, five injections were executed for the flame–wall impingement and post-impingement stages whereas twenty injections were used for the late-cycle and exhaust samples. These injection numbers were enough to sample many soot particles for statistically meaningful data analysis while avoiding a potential soot overloading issue. Detailed discussions about the effect of number of injections as well as the cyclic dispersion on the sampled soot particles are found in our previous studies [36,38]. Once this pre-determined number of fuel injections was reached, the engine was stopped and then the TEM grid was removed from the sampling probe. The grid was carefully stored before it was brought to a TEM to obtain multiple images for various on-grid locations. Figure 2 (top row) illustrates four different combustion stages investigated in the present study. These include (1) when the impingement of flame on the wall occurs (flame–wall impingement), (2) after the flame–wall impingement (post-impingement) and (3) at the late-cycle burn-out stage (late-cycle). Soot particles were also collected in the exhaust pipe: (4) exhaust. As mentioned previously, the sampling probe location was fixed within the bowlrim cut-out region due to the risk of crash. Therefore, the nozzlehole orientation of the fuel injectors was adjusted so that the fuel jet trajectory was varied with respect to the sampling probe. This approach is illustrated in Fig. 2 (bottom row). For example, the flame–wall impingement case utilised the 9 o’clock hole in the field of view and thus the TEM grid location was close to the wall-impingement region of the sooting flame. This was a high sooting region [45,46] where soot particles were expected to be mature solid particles with various fractal structures [28,29]. The post-impingement soot particles were sampled by using the 10 o’clock hole so that the sooting flame impinges on the wall and then travels along the liner wall before the particles deposit on the TEM grid. Due to the improved mixing associated with a turbulent ring-vortex formed in the wall–jet head region [46,47,49], the soot particles were expected to show evidence of strong oxidation. For the late-cycle soot particles, the same 10 o’clock hole was used while the piston was rotated 180° so that the TEM grid was exposed to the sooting flame only during the late-cycle burn-out. The late-cycle and exhaust particle sampling experiments were conducted simultaneously. It should be noted that the fixed sampling probe approach of this study does not provide information about soot particles at a certain instant but time-integrated information for the duration that the TEM grid is exposed to the sooting diesel flame. However, for the flame-impingement and post-impingement cases, the exposure time estimated by the high-speed movie of soot
luminosity was very short at 1–3 ms [37], which is comparable to a quick-insertion-type sampling approach widely used in open flame burners. The corresponding still images from high-speed soot luminosity movies are shown in Fig. 3. The figure shows the temporal evolution of hot soot luminosity during flame–wall impingement, post-impingement and late-cycle combustion stages. Images were taken using two high-speed CMOS cameras (VisionResearch Phantom v7.3) at 36,000 frames per second for the bottom-view and side-view orientations. The timing of each image set is noted as °CA aTDC and °CA aEOI (after the end of injection). The locations of bowl-rim cut-out and the soot sampling probe are illustrated in each image. The leftmost images for the flame–wall impingement and postimpingement combustion stages show the first frame with visible soot luminosity. The following images show the time when the soot clouds reach the location of the TEM grid (short red bar) indicating the start of soot particles collection. It is observed that the soot sampling for the flame–wall impingement and post-impingement stages started roughly at 10.4 and 12.4°CA aTDC. As for luminosity images during late-cycle burn-out stage, the first two side-view images are completely dark due the obstruction of the piston bowl-rim (the imaging was done from the opposite end of the bowl-rim cut-out). Judging from the bottom-view images and the third side-view image, the time when the TEM grid is first exposed to the soot luminosity should be between 24 and 28°CA aTDC. 2.3. TEM imaging and imaging post-processing Soot particles imaging was performed using a JEOL 1400 TEM with 0.38 nm point resolution and 100 kV acceleration voltage. An 11-mega pixel Gatan CCD camera was used to digitise the magnified soot particle images. For each soot sample, multiple TEM images were taken at various on-grid locations considering the fluctuations of sampled soot particles due to the inhomogeneous nature of fuel–air mixture. In addition, multiple TEM images allowed for high number of soot particles and thereby increasing the credibility of statistical analysis. The magnification was set at 50,000 times considering both the number of particles per image and the image quality. The TEM images were post-processed using an in-house-developed Matlab code [50] for the particle number count, the projection area of soot particles, the primary particle diameter (dp) and the radius of gyration of soot aggregates (Rg). These data were used to calculate the fractal dimension (Df) and prefactor (kf) of soot aggregates. More details about the data processing and uncertainty analysis are found in our previous work [36]. Depending on the sampling stage, the number of processed soot aggregates varied from 129 to 996 while the number of processed primary particles ranged from 3640 to 15,579 in the present study. 3. Results and discussion 3.1. In-cylinder pressure and apparent heat release rate The averaged in-cylinder pressure traces and corresponding aHRR for three different sampling locations are shown in Fig. 4. Due to three different nozzle-hole orientations (Fig. 2), the soot particles sampling was conducted for different engine runs, which raised a question on run-to-run as well as hole-to-hole variations. However, Fig. 4 shows that the in-cylinder phenomenon does not vary significantly. The level of variations are well within cycleto-cycle variations [36], suggesting that the soot particles were sampled at fixed in-cylinder conditions. On the aHRR traces (Fig. 4 bottom), the start of soot sampling is illustrated for each case which confirms that the flame–wall impingement and
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
2723
Late-cycle
Post-impingement
Flame-wall Impingement
Fig. 2. Illustrations of sooting diesel flame evolution (top) and corresponding sampling approaches (bottom).
Distance from nozzle [mm] Fig. 3. Temporal evolution of hot soot luminosity during the diesel combustion event corresponding to the soot sampling experiments at various combustion stages of the present study. Both bottom-view (top) and side-view (bottom) images are shown at various °CA after TDC (aTDC) and °CA after the end of injection (aEOI). The field of view is indicated by dashed-line circles. Yellow dashed-lines indicate the jet centre line. The TEM grid positions (red bars) and limits of bowl-rim cut-out (green bars) are shown in both the bottom view and side view images. Numbers (mm) indicates the distance from the tip of injector nozzle.
2724
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 4. Averaged in-cylinder pressure and corresponding apparent heat release rate traces for various crank angle locations. Three traces correspond to three separate sampling experiments that designed to collect soot at various combustion stages. The approximate crank angle locations of the start of soot collection are marked as circles on the heat release rate traces.
post-impingement samplings were conducted during the main combustion event. For the late-cycle sampling, the soot particles were collected during the burn-out stage of diesel combustion. The exhaust soot sampling was conducted between 128 and 366°CA aTDC when the exhaust valves were opened.
3.2. TEM images Selected TEM images of soot particles collected from various combustion stages are shown in Fig. 5. Shown on the bottom-left corner of each image is a 200-nm scale bar. It is seen that the majority of soot particles are aggregates comprised of many primary particles with near-circular shape, consistent with previous studies [32,36,37]. Particularly in the flame–wall impingement image, many agglomerated particles with various structures are observed. This was expected because fuel rich mixtures would be formed near the flame impingement region due to the limited fuel–air mixing [45,46] and thereby promoting the soot formation [51]. It is however noted that the maximum soot coverage ratio over the TEM image was around 15%, which is close to the value suggested by Megaridis and Dobbins [52] (10%) to avoid the particle overloading and associated issue of the particle agglomeration on deposited particles. The change in the number concentration of soot particles and overall soot amount between earlier and later stages is striking. Figure 5 shows that late-cycle and exhaust soot images present much lower number of soot aggregates than the flame–wall impingement and post-impingement images. Specifically, soot aggregates that are larger than the 200-nm scale
bar are easily seen in the flame–wall impingement and postimpingement samples but do not exist in the late-cycle and exhaust soot samples. Furthermore, single-primary particles and small aggregates with only a few primary particles shown in the flame–wall impingement and post-impingement samples are rare in the late-cycle and exhaust samples. Structural difference is also significant. In the first two combustion stages, large soot aggregates appear to be comprised of many primary particles forming stretched chain-like branches. It is also noticeable that the branches are attached to highly concentrated core regions of the aggregates. Such structures are common for peak soot location of a free (no wall impingement) diesel jet [29]. By contrast, most of the soot aggregates at later stages show compact and agglomerated aggregate structures. The variations in soot particles between each combustion stage are also observed. For instance, a closer look at the TEM images for the flame–wall impingement and post-impingement samples suggests the reduction in soot particles, particularly large soot aggregates. This was explained by the evolution of soot particles along the flame trajectory (i.e. increasing distance from the nozzle). Previous studies conducted in a constant-volume combustion chamber where the soot particles sampling was conducted for a free (no wall) flame reported a similar trend of reduced large soot aggregates further downstream of the flame due to soot oxidation [28–30]. Indeed, the estimated flame travel distance between the nozzle hole and TEM grid was 42 and 60 mm for the flame–wall impingement and post-impingement, respectively, which might suggest soot particles oxidation. However, the increasing distance from the nozzle was not the only cause for the reduction of large soot particles. It was likely that the flame–wall interactions also impacted the soot particles through the enhanced mixing in the jet head-vortex region [45], which would promote the soot particles oxidation. Unlike the evident changes found between the first two stages, the soot particles in the exhaust appear to be similar to the latecycle sample. This might suggest the origin of exhaust soot particles such that those particles survived the oxidation processes during the main combustion would exit through the exhaust. An interesting observation from the late-cycle and exhaust samples is transparent objects with a size of tens of nanometres, which are illustrated by rectangles in Fig. 5. The cause for these structures is not entirely clear but they could be condensed aliphatic compounds as a result of decreased gas temperature in the expansion stroke [53]. In this study, these crystalline-like particles were not included in the image processing. It should be noted that the TEM images in Fig. 5 are a result of various numbers of fuel injections depending on the sooting level of each combustion stage. This was to guarantee high enough number of soot particles on the TEM images for statistical analysis. For instance, a total of 20 injections were executed for the late-cycle and exhaust soot samples whereas only 5 injections were applied to the flame–wall impingement and post-impingement samples. Despite fourfold higher number of fuel injections, the soot particles in the late-cycle and exhaust samples are much less, evidencing that the soot reduction shown in Fig. 5 is real. 3.3. Size distribution of soot particles Figure 6 shows histograms of the radius of gyration (Rg) of soot particles (including single-primary particles and multi-primary aggregates) at various combustion stages. Noted next to each histogram are the total number of soot particles, the number of soot particles per TEM image, and the projection area of soot particles on the TEM images. Between the flame–wall impingement and post-impingement samples, it is shown that the number of soot particles is higher but the projection area is lower for the
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
2725
Fig. 5. Example TEM images of soot particles sampled at various combustion stages. The scale bars of 200 nm are shown as red horizontal bars in the bottom-left corner of each image. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
post-impingement stage. As mentioned previously, the observed trend was expected because the soot particles would be oxidised due to the increased distance from the nozzle [28–30] and the enhanced mixing in the jet head-vortex region [45,46,54,55]. As a result, the overall soot amount decreased in the post-impingement sample. Also, it could be interpreted that those large soot aggregates observed in the flame–wall impingement sample would disintegrate into multiple smaller aggregates due to oxidationinduced fragmentation [56], leading to the increased number of small soot aggregates (Rg < 25 nm). In comparison, late-cycle and exhaust soot samples show low particle number counts and reduced projection area. Although some small particles (Rg < 10 nm) were still observable, a significant portion of the soot particles at these two stages have Rg higher than 10 nm. This supports the earlier interpretation of the TEM images in Fig. 5 that the small particles were easily oxidised and those highly-concentrated
core regions of large aggregates remained in the late-cycle and exhaust stages. Due to significantly lower sample size, one might question the effectiveness of thermophoretic sampling for exhaust soot particles in the present study. For the first three combustion stages, the soot luminosity images confirmed the direct exposure of the TEM grid to a sooting flame (Fig. 3). However, the exhaust gas temperature could be much lower than the in-cylinder flames and thus resulting in significantly weaker thermophoresis compared to the in-flame sampling. Indeed, significant reduction in a temperature gradient between exhaust gas and TEM grid is expected due to cylinder expansion and exhaust valve opening. However, the overall particle size distribution and particle number density of the exhaust soot sample are similar to that of the late-cycle sample, suggesting that the reduction in the temperature gradient does not have significant impacts on particle deposition process in the
2726
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 6. Histogram of Rg of soot particles at various combustion stages. Total particle number counts, number counts of particles per TEM image and particle projection area per TEM image are also noted. Error ranges are estimated with 95% confidence.
exhaust. Previous exhaust soot studies reported large aggregates with complex fractal structures that are similar to the soot aggregates shown in the flame–wall impingement or post-impingement samples [17,19–21]. It was believed that this discrepancy was due to very low-sooting operating conditions of this study. More specifically, the single-cylinder operation, the use of single-hole nozzle injector, skip-firing and no exhaust gas recirculation (EGR) in the present study produced much less soot particles than multicylinder engines used in the exhaust soot studies. Furthermore, Fig. 6 shows that the number concentration and projection area of the exhaust sample is roughly 0.5% of the flame–wall impingement sample, if four times more injections were considered. This is very consistent with the previous study [33] suggesting that more than 99.5% of soot formed inside the engine cylinder is oxidised before they exit through the exhaust. Therefore, it can be concluded that the low sample size in the exhaust was not due the experimental error but a result of low soot level of the tested engine and selected operating conditions. Figure 7 shows the probability density function (pdf) of Rg and dp for all four combustion stages. The mean values and error ranges are also annotated next to the corresponding plot. Figure 6 (top) shows that the mean Rg is around 38 nm for the flame–wall impingement soot particles, which decreases to about 26 nm for the post-impingement soot particles. As previously mentioned, this shift towards lower Rg regime can be explained by the increased number of small aggregates due to the fragmentation of large aggregates. By contrast, the pdf of dp shows identical distribution with a mean value of about 20 nm between the flame–wall impingement and post-impingement soot particles (Fig. 7 bottom). This suggests that the primary particles at these combustion stages are mature solid particles due to the balance between the surface growth and oxidation. Compared to the post-impingement sample, the late-cycle soot particles show higher mean Rg and a significant decrease of small soot aggregates (Rg < 50 nm), implying intensive soot oxidation during late-cycle burn-out. Also, the mean dp of the late-cycle sample is 15.5 nm, 4.2 nm lower than the post-impingement sample, suggesting surface oxidation of the primary particles [32]. The exhaust soot particles appear to be further oxidised from the late-cycle soot particles, evidenced by slightly lower Rg and dp. Overall, the Rg and dp of the late-cycle and exhaust samples suggest that those highly concentrated substructures of soot aggregates [40,51,57] which could not be completely oxidised during the late-cycle burn-out stage would exit through the exhaust. It should be noted that the size of exhaust soot particles in the current
Fig. 7. Probability density functions of radius of gyration (Rg) of aggregates (top) and diameter (dp) of primary particles (bottom) for soot particles at various stages. The mean values and error ranges (95% confidence) are noted.
studies is smaller than the ones found in previous studies [17,19–21]. As discussed earlier in this section, this could be the results of low-sooting operating conditions that implemented in the present study. 3.4. Fractal morphology of soot aggregates The fractal morphology of soot aggregates can be characterised by the mass fractal relation shown in Eq. (1) [58]:
N ¼ kf
Rg
!D f ð1Þ
dp
where N is the number of primary particles within a soot aggregate and dp is the mean diameter of primary particles in each aggregate. The fractal dimension Df represents the structural compactness of a fractal aggregate such that higher Df means more compacted fractal morphology. The kf is the fractal prefactor to estimate the optical and transport properties of soot particles [59]. As suggested in Eq. (1), Df and kf can be obtained by reading the slope and the intercept of the least square fit line of the logarithm plot ln(N) over ln(Rg/dp ). In this study, N was determined using Eq. (2):
N ¼ ka
Aa Ap
a ð2Þ
where Aa is the projection area of a soot aggregate, and Ap is the mean projection area of primary particles within the aggregate.
2727
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
Fig. 8. Statistical determination of soot aggregate fractal dimension (Df) and prefactors (kf) using the number of primary particles (N), the radius of gyration (Rg) and the mean primary particle diameter of each soot aggregate (dp ). Error ranges are estimated with 95% confidence.
Flame-wall Impingement
Post-impingement
Late-cycle
Exhaust
Fig. 9. Illustration of soot particles evolution during diesel combustion in a light-duty engine.
The empirical constant ka and primary particle overlap factor a equal to 1 and 1.09, respectively [52]. Figure 8 shows the log–log plot of Df and kf for all four combustion stages of the present study. The Df and kf values together with error ranges (95% confidence) are noted in each plot. It should be noted that Df was calculated only for the soot aggregates with more than three primary particles (N > 3) because monomers (N = 1) are not aggregates and the aggregates with only two or three primary particles (N = 2 or 3) hardly have fractal structures. It is observed from Fig. 7 that the fractal dimension measured in this study ranges between 1.7 and 1.9, which is consistent with previous studies conducted in both engine exhaust [17,19–21] and constant-volume combustion chambers [29–31]. This suggests that soot formation in diesel combustion falls into the diffusion limited cluster–cluster aggregation regime [52]. It is also seen that fractal dimensions are very similar between the flame–wall impingement and post-impingement samples. Considering the increased number of small soot particles (Figs. 5 and 6) in the post-impingement samples, the similarity in fractal dimension was not expected. It could be explained that the decrease in Df associated with the stretched chain-like branches cancelled out the increase in Df due to the compact structures of the remaining cores. The late-cycle and exhaust samples show the same Df that
are higher than the flame–wall impingement and post-impingement samples. The increased fractal dimension implies that the aggregates became more compact, likely due to the soot oxidation occurring in the branches as well as in the primary particles at the outskirt of the remaining cores [32]. Some studies [52,60] suggested that the aggregation of soot particles begins with monomer–monomer interactions which are followed by the monomer-cluster interactions, leading to small and compact clusters (Df 3). Thereafter, the clusters join other clusters through cluster–cluster aggregation to form larger aggregates with long stretched fractal morphology (Df < 2). Surface growth would occur throughout the process to add mass and dimension to the aggregates. Also, one study [61] reported that large clusters tend to breakdown into smaller aggregates during the oxidation process and shrink in size. The results of the present study are consistent with the previous studies, suggesting fragmentation and de-generation process of soot aggregates due to the flame–wall interaction and late-cycle burn-out. 4. Conclusion An experimental investigation of the structural evolution of soot particles has been conducted in a single-cylinder light-duty
2728
R. Zhang, S. Kook / Combustion and Flame 162 (2015) 2720–2728
diesel engine. Thermophoretic soot sampling was used to collect soot particle samples from various combustion stages. Collected soot samples were imaged using a TEM. The TEM soot images were post-processed to obtain soot particle number concentration, size distribution and fractal morphology. The main findings of the present study are summarised by illustrating the soot particle structures at various diesel combustion stages as shown in Fig. 9. It is understood that some vulnerable soot particles (light grey circles) such as single-primaries, small aggregates and some weak linking primary particles in large aggregates are oxidised during flame–wall interactions, leading to the fragmentation and shrinking of soot aggregates. This results in the increased number of small particles in the post-impingement stage. The partially oxidised soot particles then undergo further oxidation in the latecycle burn-out stage, leading to more compacted soot particles through the detachment of the chain-like branches from the large aggregates. The de-generation also reduces the size of primary particles therein. However, the large aggregates comprised of highly-concentrated primary particles (black circles) survive the soot oxidation. Between the late-cycle and exhaust stages, there is not much of further oxidation due to significantly reduced gas temperature. Finally, those remaining soot particles from the late-cycle burn-out stage exit through the exhaust. Acknowledgments Experiments were performed at the UNSW Engine Research Laboratory, Sydney, Australia. Support for this research was provided by the Australian Research Council via Discovery Project. References [1] D.R. Tree, K.I. Svensson, Proc. Combust. Inst. 33 (2007) 272–309. [2] I. Glassman, Symp. (Int.) Combust. 22 (1989) 295–311. [3] J.S. Lighty, J.M. Veranth, A.F. Sarofim, J. Air Waste Manage. Assoc. 50 (2000) 1565–1618. [4] V. Muzyka, S. Veimer, N. Schmidt, Sci. Total Environ. 217 (1998) 103–111. [5] D.M. Broday, R. Rosenzweig, J. Aerosol Sci. 42 (2011) 372–386. [6] R. Zhang, A.F. Khalizov, J. Pagels, D. Zhang, H. Xue, P.H. McMurry, Proc. Natl. Acad. Sci. USA 105 (2008) 10291–10296. [7] L. Benbrahim-Tallaa, R.A. Baan, Y. Grosse, B. Lauby-Secretan, F. El Ghissassi, V. Bouvard, N. Guha, D. Loomis, K. Straif, V.M. Arlt, Lancet Oncol. 13 (2012) 663– 664. [8] B. Frank, R. Schlögl, D.S. Su, Environ. Sci. Technol. 47 (2013) 3026–3027. [9] D.S. Su, A. Serafino, J.O. Müller, R.E. Jentoft, R. Schlögl, S. Fiorito, Environ. Sci. Technol. 42 (2008) 1761–1765. [10] M.E. Mueller, G. Blanquart, H. Pitsch, Proc. Combust. Inst. 32 (2009) 785–792. [11] T. Li, H. Ogawa, SAE Int. J. Engines 5 (2011) 94–101. [12] I.S. Abdul-Khalek, D.B. Kittelson, B.R. Graskow, Q. Wei, F. Bear, SAE Technical Paper 980525, 1998. [13] S.S. Gill, D. Turner, A. Tsolakis, A.P.E. York, Environ. Sci. Technol. 46 (2012) 4215–4222. [14] H. Kim, Y. Sung, K. Jung, B. Choi, M. Lim, J. Mech. Sci. Technol. 22 (2008) 1793– 1799.
[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61]
R. Zhu, H. Miao, X. Wang, Z. Huang, Proc. Combust. Inst. 34 (2013) 3013–3020. K. Al-Qurashi, A.L. Boehman, Combust. Flame 155 (2008) 675–695. A. Neer, U.O. Koylu, Combust. Flame 146 (2006) 142–154. U. Mathis, M. Mohr, R. Kaegi, A. Bertola, K. Boulouchos, Environ. Sci. Technol. 39 (2005) 1887–1892. J.P. Shi, D. Mark, R.M. Harrison, Environ. Sci. Technol. 34 (2000) 748–755. J. Zhu, K.O. Lee, A. Yozgatligil, M.Y. Choi, Proc. Combust. Inst. 30 (2005) 2781– 2789. K.O. Lee, J. Zhu, SAE Technical Paper 2004-01-1981, 2004. K. Yehliu, O. Armas, R.L. Vander Wal, A.L. Boehman, Combust. Flame 160 (2013) 682–691. L.M. Pickett, D.L. Siebers, Combust. Flame 138 (2004) 114–135. E. Cenker, G. Bruneaux, L.M. Pickett, C. Schulz, SAE Int. J. Engines 6 (2013) 352– 365. J. Manin, L.M. Pickett, S.A. Skeen, SAE Int. J. Engines 6 (2013) 1908–1921. R. Ryser, T. Gerber, T. Dreier, Combust. Flame 156 (2009) 120–129. U. Aronsson, C. Chartier, O. Andersson, B. Johansson, J. Sjöholm, R. Wallander, M. Richter, M. Alden, SAE Int. J. Engines 3 (2010) 137–155. S. Kook, L.M. Pickett, Proc. Combust. Inst. 33 (2011) 2911–2918. S. Kook, L.M. Pickett, SAE Int. J. Fuels Lubr. 5 (2012) 647–664. K. Kondo, T. Yamaguchi, H. Nishigai, S. Takano, T. Aizawa, SAE Technical Paper 2011-24-0068, 2011. T. Yamaguchi, K. Kondo, H. Nishigai, S. Takano, T. Aizawa, SAE Int. J. Fuels Lubr. 5 (2011) 316–328. Z. Li, C. Song, J. Song, G. Lv, S. Dong, Z. Zhao, Combust. Flame 158 (2011) 1624– 1630. A. Pungs, S. Pischinger, H. Bäcker, G. Lepperhoff, SAE Technical Paper 2000-011999, 2000. H. Bäcker, O. Hild, G. Lepperhoff, SAE Technical Paper 970875, 1997. Y. Aoyagi, T. Kamimoto, Y. Matsui, S. Matsuoka, SAE Technical Paper 800254, 1980. S. Kook, R. Zhang, K. Szeto, L.M. Pickett, T. Aizawa, SAE Int. J. Fuels Lubr. 6 (2013) 80–97. R. Zhang, S. Kook, Environ. Sci. Technol. 48 (2014) 8243–8250. R. Zhang, K. Szeto, S. Kook, SAE Int. J. Fuels Lubr. 6 (2013) 605–614. G. Bruneaux, Int. J. Engine Res. 9 (2008) 249–265. J.E. Dec, D.R. Tree, SAE Technical Paper 2001-01-1295, 2001. A. Rusly, M.K. Le, S. Kook, SAE Int. J. Fuels Lubr. 6 (2013) 615–626. A. Rusly, M.K. Le, S. Kook, E.R. Hawkes, Fuel 125 (2014) 1–14. F.R. Held, S. Bornschlegel, S. Riess, T. Vogel, M. Wensing, SAE Technical Paper 2013-24-0063, 2013. L. Song, J. Abraham, J. Fluids Eng. 125 (2004) 997–1005. G. Bruneaux, SAE Technical Paper 2005-01-2097, 2005. L.M. Pickett, J.J. López, SAE Technical Paper 2005-01-0921, 2005. M.K. Le, S. Kook, E.R. Hawkes, Fuel 140 (2015) 143–155. J.B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill Book Co., USA, 1988. R. Solsjo, M. Jangi, C. Chartier, O. Andersson, X. Bai, SAE Technical Paper 201401-1127, 2014. T. Aizawa, H. Nishigai, K. Kondo, T. Yamaguchi, J.G. Nerva, C. Genzale, S. Kook, L.M. Pickett, SAE Int. J. Fuels Lubr. 5 (2012) 665–673. D.R. Tree, J.E. Dec, SAE Technical Paper 2001-01-1926, 2001. C.M. Megaridis, R.A. Dobbins, Combust. Sci. Technol. 71 (1990) 95–109. H. Wang, Proc. Combust. Inst. 33 (2011) 41–67. G. Bruneaux, SAE Technical Paper 2005-01-2100, 2005. K. Inagaki, S. Takasu, K. Nakakita, SAE Technical Paper 1999-01-0508, 1999. M.E. Mueller, G. Blanquart, H. Pitsch, Proc. Combust. Inst. 33 (2011) 667–674. J.E. Dec, P.L. Kelly-Zion, SAE Technical Paper 2000-01-0238, 2000. R. Botet, R. Jullien, Ann. Phys. Fr. 13 (1988) 153–221. C.M. Sorensen, G.C. Roberts, J. Colloid Interface Sci. 186 (1997) 447–452. P. Meakin, J. Colloid Interface Sci. 102 (1984) 505–512. H. Jung, D.B. Kittelson, M.R. Zachariah, Combust. Flame 136 (2004) 445–456.