2.3.1. Schlieren Imaging

The schlieren technique is based on the fact that when a light ray travels through a medium with refractive index gradients, it suffers a deflection due to the refraction phenomenon [25]. Accordingly, any variations of refractive index such as those produced by density variations at the injection of fuel can be recorded as different gray levels in an image. Consequently, this technique allows to observe the local density variations that the fuel air mixing, auto ignition or flame development provoke. In this

experiments, a high-speed single-pass schlieren imaging configuration was implemented to visualize the spray and flame boundaries at any operating conditions, with an optical arrangemen<sup>t</sup> similar as that described in [26].

On the illumination side, light from a xenon lamp is driven with a liquid light guide into an iris diaphragm, to generate a point light source at the focal length plane of a parabolic mirror (f = 610 mm, D = 150 mm) so that the measurement area is illuminated with a collimated beam. In addition, to avoid interference and restrict the spectrum of the xenon lamp in the DBI and NL images commented later, a BG18 band pass filter was used. On the other side of the chamber, a spherical lens (f = 750 mm) was placed very close to optical access. This lens focusses the light onto the Fourier plane where an iris diaphragm with a cut-o ff diameter of around 3 mm was located. Additionally, another BG18 band pass filter was placed just before the Photron Fastcam SA-5 camera (Photron, San Diego, CA, USA) to minimize undesired light from sources other than the schlieren illumination lamp. The camera was equipped with a Carl Zeiss Makro-Planar T 100 mm f/2 ZF2 camera lens (Carl Zeiss, Aalen, Germany). Images were recorded at 25 kfps. The shutter time was 4.24 μs and it was kept constant throughout all the experiments. The resolution was 800 × 320 pixel with a total magnification of 6.9 pixel/mm. 20 injection cycles per test were recorded in all cases. The schlieren images have been used to describe the spray tip penetration under nonreactive, as well as the penetration under reactive conditions and the ignition delay. Image segmentation from the background has been performed by using the standard methodology developed by ECN [20].

In this paper, ignition delay (ID) is obtained from schlieren images as the time at which the derivative of the accumulated pixel intensity within the spray boundaries is 50% of range between the minimum and maximum ones. This criterion was defined and explained in [27].

#### 2.3.2. High-Speed OH\* Chemiluminescence (OH\*) Imaging

Excited hydroxyl radicals (OH\*) are good tracers of high temperature combustion regions in a flame [28]; therefore, visualization of OH\* chemiluminescence at the base of the flame allows to quantifying the lift-o ff length (LOL). Moreover, if an intensified high-speed camera is used, the ID can be measured too. In these experiments, a high-speed image intensifier (Hamamatsu C10880 by Hamamatsu, Hamamatsu city, Japan) was used, with gain set between 700 and 850 depending on fuel and operating condition. It was coupled to a Photron Fastcam SA5 camera with a 1:1 relay lens. The system was equipped with a UV f/4 100mm focal length lens. An interference filter centered at 310 nm (10 nm FWHM) was placed in front of the camera to remove most of the radiation of the flame while keeping OH\* chemiluminescence. As shown in Figure 1, a dichroic mirror, which reflects UV spectrum and transmit the visible light spectrum, is used to reflect the UV flame radiation to the intensified camera.

Images were taken at 25 kfps. The resolution was 704 × 416 pixels and the magnification was 5.43 pixel/ mm. The shutter time was adapted between 19.97 μs and 39.75 μs to accommodate the camera dynamic range to the flame radiation intensity as a function of the fuel and operating condition. Again, 20 injection cycles per test were recorded in any case. ID was obtained determining the first frame with detectable light intensity, and it was used as a check on the results obtained with the schlieren technique.

The algorithm used to obtain the LOL is based on the procedure described in [29] and recommended by the engine combustion network (ECN). Thus, the LOL was as the distance between the injector tip and the first axial locations above and below the spray centerline with intensity greater than 50% of the intensity peak of that zone.

#### 2.3.3. Di ffused Back Illumination (DBI) Extinction Imaging

Di ffused back-illumination (DBI) is a technique based on measuring the amount of light attenuated by liquid droplets or by soot particles within the flame, which is related to the liquid and soot concentration, respectively. A red LED (λ = 660 nm) was used in these experiments as the light source

to create short flashes synchronized with the camera frames. An engineered di ffuser (EDC-20 by RPC Photonics, Henrietta, NY, USA) was placed in front of the LED to create a di ffused Lambertian intensity profile [30]. On the collection side, the transmitted light from the LED and the flame radiation went through a beam splitter with a 50% reflection rate. Then, half of the light was collected by the Photron SA-X2 camera. The exposure time was 1 μs and the resolution was 896x384 with a magnification of 6.85 pixel/mm. The sampling frequency was 25 kfps.

The images taken were analyzed considering that the total light registered by the camera has two contributions: the transmitted LED light intensity and the flame radiation. Due to the use of a bandpass filter centered at 660 nm (FWHM = 10 nm), the crosstalk of flame radiation into the DBI signal is minimized. However, the flashing frequency of the LED was set as quarter of the camera frame rate to capture a LED image between every three consecutive dark images. This configuration was used to deal with the non-ideal CMOS sensor behavior, that occurs when the camera is exposed to a sudden change in light intensity between two consecutive frames [18]. The flame luminosity from dark image before the LED image (third dark image) was quantified and used to isolate the transmitted LED light from the total registered radiation. Then, the light attenuation can be related with the optical properties of the soot cloud by means of Lambert-Beer's law, as described in Equation (1):

$$\frac{I\_{\rm ou} - I\_{off}}{I\_{\rm o}} = e^{-\rm KL} \tag{1}$$

where *Ion* is the light intensity recorded by the camera when the LED is on, i.e., the sum of the transmitted LED intensity and the flame luminosity. *Ioff* is the intensity of the flame acquired when the LED is o ff. *IO* is the LED light intensity obtained from images recorded before the start of injection (SOI). K is the soot dimensional extinction coe fficient and L is the light beam path length through the soot cloud. Thus, the product KL represents the integral value of the soot extinction coe fficient along the light path, which is related with the soot concentration [24].

Besides determining soot production through the KL factor, DBI is also suggested as an experimental standard to measure the liquid length (LL) of fuels by ECN [20]. The method uses the extinction produced by the spray droplets to provide a quantitative parameter related to the liquid volume fraction along the path of the light. ECN recommends also taking care with vapor phase beam steering from temperature gradients, which could disturb the measurement. In the current work, the liquid length has been determined only at 800 K and 900 K. It was observed that at higher temperature the measurement is not reliable because the flame lift-o ff zone is close to the liquid jet zone and more intense beam steering exists there.

The image processing method to ge<sup>t</sup> KL was based on creating a mask from the dark image *Ioff* with the aim of collecting the information only corresponding to flame radiation. Based upon this mask, attenuation (*Ion* − *IOff*) is calculated, from which Equation (1) is applied. Figure 2 shows a KL profile along the spray axis at a time instant where the flame is well developed (3500 μs), within the quasi- steady period chosen in this study (between 3000 μs and 4000 μs). The confidence interval at 95% (red shadow) for the measurement of the ensemble averaged KL value and the standard deviation (blue shadow) have been represented. The operating condition shown is the nominal case and the fuel chosen is the diesel.

Similarly, the total soot mass (*smass*) at a given time was determined from Equation (2) as the sum of the values of over all the pixels of the average image taken at that time, and corrected with the other factors indicated. In this equation, ρ*soot* corresponds to the soot density defined as 1.8 g/cm<sup>3</sup> by Choi [31], λ is the wavelength used in the current work (660 nm), *r* is the pixel-mm ratio (6.85 in this work) and *ke* is the dimensionless extinction coe fficient equal to 7.27 determined in this study through the ratio of scattering and absorption cross-sections which is used in small particle Mie theory:

$$s\_{\text{mass}} = \frac{\sum KL \cdot \rho\_{\text{sout}} \cdot \lambda}{kc \cdot r^2} \tag{2}$$

**Figure 2.** KL profile on the spray axis for diesel at nominal case in the time instant 3500 μs.

In Figure 3 the temporal evolution of total soot mass for diesel at nominal condition is shown. The standard deviation (blue shadow) and confidence interval at 95% for the ensemble average soot mass value (red shadow) have been represented too. In the results section, these values will be shown for the other fuels and operating conditions tested.

**Figure 3.** Total soot mass for the diesel at nominal condition.

#### **3. Results and Discussion**

Before entering the quantification and detailed analysis on the effect that fuel has upon the spray characteristics, the combustion process and soot formation, a comparison between results for the dodecane fuel from current study and from similar studies available in the ECN database [20] is presented. Table 4 summarizes the mean values obtained under the Spray A condition (15% of Oxygen concentration, 900 K of temperature and 1500 bar of injection pressure) by different research centers members of the ECN. It can be observed that results obtained for liquid length (LL), ignition delay (ID), and flame lift-off length (LOL) in this work are very close to those previously obtained. This makes

possible to confirm the reliability of current results and extending the methods in the current work for the experiments with the other fuels tested.

**Table 4.** Comparison of main spray parameters between database available in the ECN [20] and current work for dodecane under Spray A conditions (900 K, 1500 bar and 15% O2).


#### *3.1. Maximum Liquid Length*

Spray liquid length is the maximum axial penetration of the liquid phase fuel. Previous studies have demonstrated that a shorter spray liquid length leads to a better air-fuel mixing, because the fuel vaporization is completed before the fuel reaches the combustion region [32]. On the other hand, too long liquid length leads to fuel wall impingement on the cylinder, which could produce large soot emission with reduced engine e fficiency [33]. Figure 4 shows the liquid length (LL) at 800 K and 900 K for the tested fuels. The injection pressure is 1500 bar and the oxygen concentration is 15%. The LL is constant in time; therefore, the values correspond to an average between 800 μs and 4000 μs. From this figure, it can be seen that the spray liquid length of OME1 is shorter, which could be due to the lower distillation temperature of OME1 (37.40 ◦C) compared to the other fuels. Additionally, as Kook indicated in a previous study [34], low viscosities and high densities lead to shorter liquid lengths. This could explain why OMEx (υ = 1.082 mm<sup>2</sup>/s; ρ = 1057.1 kg/m3) shows shorter LL than diesel and HVO (υ = 2.7 mm<sup>2</sup>/s; ρ = 779.1 kg/m3).

**Figure 4.** Liquid length for 15% O2 and 1500 bar of injection pressure.

## *3.2. Ignition Delay*

Ignition delay (ID) is defined as the time elapsed from the start of injection (SOI) to the start of combustion. Figure 5 shows the ignition delay comparison obtained from schlieren technique and from OH\* chemiluminescence imaging. It can be seen that ID values for both techniques are almost identical, therefore, any ID value used in the current work is valid. Hereafter, the represented ID values correspond to those from the schlieren technique.

**Figure 5.** Comparison between ignition delay measured from schlieren and from OH\* chemiluminescence images.

In Figure 6, ID for dodecane, HVO, OMEx and OME1 have been compared against diesel. All the operation conditions have been presented with their standard deviation. The results attend to the trend found in previous studies [23,35]. When the air temperature, injection pressure, and ambient oxygen concentration are increased, the ID values decrease. It is a common behavior for all the tested fuels.

**Figure 6.** Comparison between ignition delay of diesel and the ignition delay for the other fuels tested.

Furthermore, Figure 6 shows that dodecane, HVO, and OMEx ignite before diesel does. The three fuels have a higher cetane number than diesel (see Table 1), and the results confirm the relevance of this parameter on autoignition. Based on this, one could expect OMEx to ignite later than HVO and dodecane. However, its molecular structure has high oxygen content, which makes it more reactive. This explains that OMEx ignites earlier than HVO and dodecane, despite its lower cetane number. For OME1, although its molecular structure also contains oxygen, its cetane number is much lower when compared to the other fuels tested. Thus, it is the last one to ignite as shown in Figure 6. It is important to mention that the OME1 at 800 K and 15% of O2 does not ignite and at 21% of O2, the ignition occurs very late (around 2400 μs).

## *3.3. Lift-O*ff *Length*

An important parameter in the combustion and soot production processes is the flame lift-o ff length (LOL). This is strongly related with the amount of fuel–air mixing process upstream the combustion region. Enough air entrained reduces the average equivalence ratio at the LOL [29] and this result in less soot production. Consequently, flame LOL is worth to be studied for the di fferent fuels tested in current work.

As previously done with the ignition delay, the LOL for the diesel compared to the other fuels is shown in the Figure 7, for the whole test matrix. The interval time to average the LOL corresponds to that in which the flame is "quasi steady" (between 3000 μs and 4000 μs). The HVO and dodecane present a very close behavior, showing smaller LOL than diesel. However, the tendency is not so clear for OMEx since at 800 K its LOL is slightly longer that diesel, at 900 K it is similar but at 1000 K is smaller, close to HVO and dodecane. It is also possible to observe that at 800 K and 15% of O2 the LOL of OMEx has the highest standard deviation, indicating a lower combustion stability. The OME1 cases show longer LOL than the other fuels because of its lower reactivity (higher ignition delay). However, at 1000 K the di fference between LOL of diesel and LOL of OME1 is smaller compared with the other operating conditions.

**Figure 7.** Comparison between lift-o ff length (LOL) of diesel and the other fuels tested.

#### *3.4. Spray Tip and Flame Penetration*

The temporal evolutions of vapor penetration and LOL, as well as the ID for all the fuels at nominal condition have been shown in Figure 8. The vapor penetration and ID have been measured from schlieren images and LOL from OH\* chemiluminescence, as was mentioned previously. The standard deviation for each parameter is represented by shaded areas. A first stage from start of injection until ignition can be observed, where vapor penetration is identical for all the fuels. Only at certain time after ignition di fferences appear. This first stage corresponds to the ignition delay phase, which extends for each fuel until combustion starts. The ignition delay time has been marked in the figure as a dashed vertical line for each fuel. During the period until ignition, there are not e ffects on the spray behavior, as expected because the penetration is governed by the momentum flux at the nozzle [34] which only depends on the pressure drop across the nozzle and on the orifice area, fuel e ffects are negligible. In addition, in Figure 8 the e ffect of the di fferent fuels upon the indicated parameters can be observed. The dashed-dot horizontal lines correspond to the temporal evolution of LOL. After the

ignition, the fuel that ignites first penetrates faster than the others do. In general, it is possible to see that while OMEx has similar behavior to the paraffinic fuels, OME1 is quite far from them.

**Figure 8.** Spray tip penetration and lift-off length at 900 K 15% O2 and 1500 bar (Spray A condition by ECN). Vertical dashed lines represent ignition delay.

On the other hand, from OH\* images, the flame tip penetration can be measured also from the high-speed OH\* chemiluminescence imaging as the axial distance from tip injector until the flame front. In Figure 9, the temporal evolution of the flame penetration for all fuels at 900 K and 1500 bar and both oxygen concentrations has been represented. The solid lines correspond to the case of 15% of oxygen concentration and the dotted lines to 21% case

**Figure 9.** Flame penetration at 900 K and 1500 bar. The solid lines correspond to 15% of Oxygen and dotted lines to 21%.

From this figure, flame tip penetration is seen to undergo a similar time evolution as spray tip penetration during most of the injection period. However, for all cases flame tip becomes constant after a given period, which depends on fuel type and oxygen concentration. For oxygenated fuels, this occurs between 1000 to 1500 μs, while for all other paraffinic fuels it happens close to the end of injection event (around 4200 μs). During that period, the stoichiometric reacting surface stabilizes, and hence the maximum distance where OH\* chemiluminescence is recorded does not change with time. For constant ambient and injection conditions this distance has been shown to scale inversely with the stoichiometric mixture fraction, i.e., directly with the stoichiometric A/F ratio [36,37], which depends on oxygen concentration and fuel composition. Table 1 shows stoichiometric air-fuel ratios, from which one can observe that para ffinic fuels have similar values, resulting in the similar long stabilization distance, while the oxygenated fuels approximately a 50% lower (A/F)st, resulting in a much shorter flame length.

Furthermore, di fferences among Diesel/HVO/Dodecane are small, in agreemen<sup>t</sup> with the small (A/F)st di fferences, OME1 shows a consistently roughly 10% shorter penetration compared to OMEx. Figure 10 has been used to depict the di fferences in flame structure for each fuel at the nominal operating condition in an instant of time where the combustion is quasi steady (3500 μs). Such images confirm the previously plotted trends among fuels in stabilized flame length, with para ffinic fuels having a similar length, and a shorter one is observed for oxygenated fuels. On the other hand, observed di fferences in lift-o ff length also point out at interesting features. As shown above, this quantity is quite similar for all para ffinic fuels and OMEx, while it is far longer for OME1. The equivalence ratio at the lift-o ff length (Φ*LOL*) can be estimated for all fuels using Equation (3), and the values are shown in the Figure 10:

$$
\Phi\_{\rm LOL} = \frac{f\_{\rm LOL}}{1 - f\_{\rm LOL}} \cdot A / F\_{\rm st} \tag{3}
$$

**Figure 10.** Flame images from OH\* chemiluminescence for all fuels at the nominal operating condition at 3500 μs ASOI.

In the Equation (3), the term *fLOL* represents the fuel mixture fraction along spray axis and it is calculated using the equation (4), where K is a constant equal to 7, do is the nozzle diameter and ρf and ρa correspond to the density of the fuel and the ambient respectively.

$$f\_{\rm LOL} = \frac{K \cdot d\_{\rm o^{-}} \sqrt{\frac{\rho\_f}{\rho\_x}}}{\rm LOL} \tag{4}$$

Results show that the flame is rich at the lift-off for all fuels except for the OME1, and hence a typical lifted diffusion flame stabilized at stoichiometric conditions is observed for all fuels except for the latter one. OME1 flame stabilizes at lean conditions, where all needed air is already available, which explains the previously observed differences in flame length between OMEx and OME1. Similar lean stabilized flames were observed for an oxygenated fuel (70% tetraethoxypropane 30% heptamethylnonane) with a 100 um orifice at 21% O2, 850 K and 14.8 kg/m<sup>3</sup> ambient conditions in [38].

## *3.5. Soot Production*

As was indicated in methodology section, the DBI technique was used to determine the soot production. The DBI images were collected after schlieren. However, OH\* chemiluminescence was recorded with the two sets. To verify that both sets are consistent and there are no experimental discrepancies among them, the results obtained for ID from OH\* images measured with the first and in the second set are compared in Figure 11. The grey shadow in the bisector represents the uncertainty associated to the time interval between consecutive frames. It can be observed that ID differences between both sets are always less than the time elapsed between two consecutives frames (40 μs). Thus, it is possible to conclude that both sets are consistent and results can be analyzed together.

**Figure 11.** Comparison of ignition delay (ID) obtained in the first set and in the second one.

DBI images were processed to calculate the flame soot in terms of the optical thickness (KL). KL maps were constructed to depict soot evolution throughout the combustion event, by calculating the KL for each axial distance and time-step. A good way to simultaneously evaluate the KL evolution in time and space is transforming the KL map at any given instant into a 1D vector where KL values at any given cross section from the nozzle are accumulated into a single value. Thus, each KL map at any instant is converted into a vector of "accumulated KL" values along the flame axis. By compiling those vectors at every time step (i.e., at every frame), a plot such as those of Figure 12 can be obtained. In any of those plots, the abscissa axis reads for time, the ordinates axis indicates axial distance from

the nozzle and the color at any point (X,Y) in the plot indicates the accumulated KL at a cross section at distance Y from the nozzle, of frames taken at time X after SOI.

**Figure 12.** Accumulated KL map at the operating condition with the greatest tendency to soot formation (1000 K of temperature, 500 bar of injection pressure and 15% of O2).

In Figure 12, the accumulated KL has been represented for the operating condition with greater tendency to soot formation, that is, at the highest temperature (1000 K), the lowest injection pressure (500 bar) and the lowest oxygen concentration (15%). Dashed lines have been used to depict ID, vapor penetration and the flame lift off length. In this figure, it is possible to observe that soot production for OME1 and OMEx is below the detection threshold. The color scale for these two fuels has been modified in other to enhance this fact. Therefore, a sootless flame seems to be established. This holds for all operating conditions and, consequently, KL values for those fuels cannot be analyzed further. Dodecane, HVO and OMEx show smaller KL values than diesel. As for the time evolution, in all three cases it can be observed that initially the largest soot amount is located at the spray tip, in the head vortex area, but when reaching a distance around 70 mm, this soot disappears due to the establishment of the quasi-steady flame The highest KL region appears for the three fuels from 40 to 50 mm from the nozzle.

For a given fuel composition, the soot production is related to the amount of air entrainment that occurs upstream of the lift-off [29]. Longer LOL suggests less soot formation, but this depends also on fuel composition. This can be observed in Figure 12, where diesel has the longest lift-off length but also the largest soot formation. The stoichiometric air/fuel ratio (ma/mf) for dodecane and OME1 is quite different. At 15% of oxygen concentration, stoichiometric air-fuel ratio for dodecane is 20.72 versus 10.03 for OME1 indicating that oxygenated fuels require less air to oxidize. This explains that OMEx, although with LOL closer to diesel, does not produce soot. This statement is supported by the equivalence ratio (Φ) at LOL. For dodecane, Φ is 4.9 at baseline operating conditions, while for the OME1 is 1.0. Furthermore, OME fuels do not have carbon-carbon (C-C) bonds, which contribute to the absence of soot.

In the case of HVO, its aromatic free composition suggests that less polycyclic aromatic hydrocarbons (PAH) are formed [39]. This could explain the lower soot production for HVO compared to diesel, as they can be considered as the building blocks for particulates in flames [40].

The total soot mass production was calculated from instantaneous KL, following the procedure described in Section 2.3.3. Figure 3 depicts the total soot mass for diesel, dodecane and HVO at di fferent operating conditions with the aim of showing the e ffect of decreasing temperature, increasing injection pressure and increasing the ambient oxygen content. The shown temperatures are 1000 K and 900 K, from top to bottom; the injection pressures are 500 bar and 1000 bar, from left to right; and 15% and 21% of oxygen have been represented for 900 K and 500 bar at the bottom of the figure.

It can be seen that when decreasing the temperature from 1000 K to 900 K (from top to bottom, at the left side of the Figure 13), the maximum value of soot production is reduced 40% for diesel and 30% for dodecane and HVO. When the injection pressure is increased from 500 bar to 1000 bar (from left to right at the top of figure), the soot reduction is around 30% for all fuels. When the oxygen concentration is increased from 15% to 21% (from left to right at the bottom of figure) the total average soot production is reduced 30% for diesel, 40% for HVO and 50% for dodecane. Results for all conditions show that the parametric variation in terms of soot tendency always holds, with diesel showing the largest soot production, while HVO and dodecane present similar soot production. In fact, considering all results presented in this work regarding combustion and soot formation characteristics, it can be concluded that n-dodecane is a good surrogate to model the HVO due to its similarities.

**Figure 13.** Influence of the variation of injection pressure, temperature and oxygen concentration on the total soot mass.
