Effect s of Carbon Dioxide Addition on the Soot Particle Sizes in an Ethylene / Air Flame

Carbon dioxide (CO2) is an important compound for the inhibition of combustion flame. This work investigated the effects of CO2 addition on the soot particle sizes in axisymmetric laminar co-flow ethylene/air diffusion flames, as determined by using temporal laser-induced incandescence (LII). Additionally, the flame height and absorption function of soot particles were also studied. The results showed that the size and absorbency of soot particles decreases with the addition of CO2. The total flame height also decreases with CO2 addition, while the height of the dark region grows. These results show that CO2 addition dilutes the gas stream, compresses the nucleation of soot particles and makes them less mature. This study provides useful information for the control of soot particle formation during combustion processes.


INTRODUCTION
Reducing the emission of soot particles is of great importance, since soot particles present a significant threat to human health and accelerate the greenhouse effect (Das et al., 2015;Jose, 2016;Lin et al., 2016;Nagyet al., 2016).CO 2 is one of the main components used in flue gas recirculation, which is a highly effective technique used for lowering NO x emissions from burners.Therefore, it is necessary to find out how CO 2 dilution affects the soot particles in such flames.
Previous studies have been conducted to analyze the effects of CO 2 dilution.Generally speaking, CO 2 dilution can reduce soot formation.Earlier works mainly focus on the three effects of adding CO 2 , which are dilution, temperature and direct chemical participation (Mclintock, 1968;Schug et al., 1980;Angrill et al., 2000).In more recent years most researchers have studied the influence of CO 2 dilution on soot formation directly, without analyzing the three specific effects, since Liu et al. (2001) claimed that it was impossible to totally isolate these three effects experimentally.Instead, they investigated the nucleation, surface growth and oxidation of soot particles.Most results showed that CO 2 dilution exerted negative effects on the nucleation of incipient soot and had nearly no effect on the oxidation of soot particles.As for the surface growth of soot particles, Oh et al. (2005) found no changes in the surface growth rate, while Gu et al. (2016) concluded in their simulation results that CO 2 dilution on the fuel side strongly reduced surface growth.Consequently, an experimental study is needed to make clear the influence of CO 2 addition on soot formation.Moreover, few studies have been carried out to investigate the influence of CO 2 dilution on the flame structure and optical properties.
Flame height has been used to test models of flame structure (D'Anna et al., 2002), and calculate the residence times of soot particles (Oh et al., 2007).It was studied by Roper (1977) and a correlation was put forward to predict the stoichiometric height of flames.Zhao et al. (2014) conducted their experiment by measuring flame heights with the addition of hydrogen and nitrogen to the fuel side, and found that the Roper correlation could well predict the changes in flame heights.They concluded that the flame heights should not be influenced by the addition of gases which did not participate in the chemical reaction.Wu et al. (2015) studied the effects of hydrogen addition on the flame heights of the dark region, and a significant increase in the soot-free length was observed with the addition of hydrogen, indicating that this reduced the carbon loading.However, to date there are no similar studies examining CO 2 dilution.
Most of the properties of soot particles are very different from those of the other kinds of particles that are generated simultaneously (Pu et al., 2015;Zhan et al., 2016), and closely related to their sizes (Wu et al., 2015;Fan et al., 2016).Soot particle sizes can be measured by laser-induced incandescence (LII), which has become a powerful optical diagnostic technique.In a previous study conducted by Oh and Shin (2006), the measurement of soot particle sizes was performed by applying the LII technique and sampling methods, with the results showing that the sizes measured by both methods were in good agreement.They divided diffusion flames into three regions according to the changes in soot particle sizes.With the growth of soot particles, soot particle sizes stayed nearly the same in the inception region, increased in the surface growth region and decreased in the oxidation region.These findings are now commonly used to analyze soot formation.
If we are to analyze the incandescence signal or make use of other soot optical diagnostic measurements, then it is essential to first know the optical properties of soot.For the LII measurements, determination of the soot absorption function, E(m), is significant, since this can cause significant uncertainties with regard to the simulation results of LII models (Bladh et al., 2011).Previous studies found that the absorption function is positively related to the maturity of soot particles (Wang, 2011;Barone et al., 2003), thus indicating that the change in soot maturity could be obtained by studying the absorption function.In order to accurately analyze this previous works plotted LII fluence curves, which reflect the relation between the heating laser fluence and peak LII signals.Olofsson (2015) observed LII fluence curves plotted from a premixed flat flame, and concluded that soot particles were more transparent at lower flame heights.Migliorini et al. (2015) pointed out that the initial temperature of soot particles in flames exerted an influence on the fluence curves, which was also noted by Liu et al. (2006).However, to date no studies has been performed to analyze the effects of CO 2 dilution on the optical properties of soot.
In the current work, the study of CO 2 dilution was conducted according to the three aspects mentioned above.
The luminous flame height as well as the soot-free region height were thus measured to find out how CO 2 dilution influences the flame structure.The optical properties and sizes of soot particles were analyzed by LII measurements to investigate the effects on soot formation.

METHODS
The soot generation system is shown in Fig. 1.Laminar diffusion co-flow flames were produced by a Santoro burner consisting of two concentric tubes with inner diameters of 11.1 and 101 mm.The mixture fuel of CO 2 and ethylene passed through the inner tube while the air provided by a compressor passed between the two tubes, which contained packed beds of glass beads and porous ceramic honeycomb.The burner was placed on a three-dimensional translating stage system so that the position of the flame produced by the burner could be located and altered precisely.The flow rate conditions for the laminar diffusion co-flow flames fueled with ethylene and CO 2 of different volumetric fractions, as presented in Table 1, were controlled by the mass flow controllers (Alicat Scientific, Inc.).The flow rates of air and ethylene remained at 43 L min -1 and 0.15 L min -1 during the process of all experiments.

Flame Height Measurements
Pictures of flames were recorded using a digital camera (Nikon, D700).The ISO was set to 200 and the white balance was set to "direct sunlight" to reduce the influence on the flame luminosity.Before taking pictures of flames, a picture of a ruler placed right above the burner was taken to figure out the spatial resolution.To make sure that flames could be clearly seen in the pictures, a suitable shutter time was set for each flame, as shown in Table 1.Flame heights were determined by finding the corresponding pixel to the threshold value of luminance from each flame, which will be demonstrated later.
1/500 30 0.064 1/500 50 0.15 1/250 Note: X CO 2 is the volumetric fraction of CO 2 in the fuel.Q CO 2 is the volumetric flow rate of CO 2 .Q C 2 H 4 is the volumetric flow rate of ethylene.Q air is the volumetric flow rate of air.T s is the shutter time set in the camera.

LII Measurements
The experimental setup and the analysis of the LII signal were based on the work conducted by Chen et al. (2017).The experimental setup is shown in Fig. 2.
The laser-based measurement platform consisted of laser source system, soot generation system and signal detection system.In the laser source system, a Nd/YAG laser (1064 nm) with a pulse width of 7 ns and a repetition rate of 10 Hz, was used as the excitation source to heat the soot particles in the flames.An attenuator was applied at the front of the laser to control the laser energy, which was monitored by a power meter behind the flames.Subsequently, a diaphragm was used to select the top-hat region of the laser beam where laser fluence was almost uniformly distributed.The diameter of the selected region of the laser beam was reduced to nearly 1 mm when the beam went through a telescope system consisting of two spherical lenses with diameters of 500 and 200 mm.In the signal detection system, the cylindrical probe zone with a diameter of 1 mm and a width of 1 mm in the flame was determined by a slit.The LII signals were recorded by a photomultiplier (PMT, Hamamatsu H10721-01) connected to a digital 6 GHz oscilloscope (DSO80604B).A band-pass filter at 450 nm was placed in front of the PMT to reduce the influence of flame luminosity on the LII signals.
In order to obtain LII signals with satisfying signal-tonoise ratios, the position detected in the flames should have high soot volume fraction so that the signals are strong enough.Consequently, the LII signals to be analyzed were detected at between 18 to 40 mm above the burner (HAB).The LII signals used for measurements of soot particle sizes were averaged over 128 laser shots, and the peak LII signals used for plotting fluence curves were averaged over 50 laser shots.

Flame Heights
The flame heights of diffusion flames can be divided into two parts, which are the yellow luminous flame height and dark region height.The yellow color of the diffusion flames is considered to be caused by soot particles in the flames, while there are rarely soot particles in the dark region (Wu et al., 2015).Information about soot formation can thus be obtained through changes in the two heights.
Measurement of heights was performed by finding the uppermost pixel to cross a threshold value of luminance from each flame.The thresholds of the luminous height and dark region height were set to be twenty times and one hundred times as much as the luminance of background at the edge of each picture, respectively.Fig. 3 presents the results of the measurement of flame heights.
As shown in Fig. 3, with addition of CO 2 to the fuel side, the total flame height slightly decreased.To explain this behavior, the Roper correlation was used to simulate the total height of the flame (Roper, 1977).This correlation can predict the stoichiometric height of flames correctly, as the related equation is written as follows: where H is the stoichiometric height (cm), Q F is the fuel gas volumetric flow rate (cm 3 s -1 ), D ∞ is the diffusion coefficient at initial temperature (cm 2 s -1 ), S is the stoichiometric volume ratio of air to fuel, T ∞ is the initial temperature (K), which is regarded as the ambient temperature, and T f is the temperature of the flame (K).
The stoichiometric height, also known as the height of the blue reaction region, is defined as the height of the position where the fuel and oxidizer are in stoichiometric proportions.Since it is hard to detect the blue reaction region in normal diffusion flames without using measurements of OH PLIF, researchers replace the stoichiometric height with the yellow luminous height, as they are considered to be approximate to each other (Sunderland et al., 2004).In this work, the stoichiometric height was thus regarded as the yellow luminous height.
When T ∞ was set to be the diffusion coefficient of ethylene in air (0.166 cm 2 s -1 ) and T f was set to 1700K in Roper's correlation (Zhao et al., 2014), the height of the flame fueled with pure ethylene was calculated to be 5.18 cm, which well predicted the height of the flame, which is measured to be 5.09cm.In Fig. 3, a slight decrease in the total flame height with the addition of CO 2 was observed, and this most likely resulted from an increase in the diffusion coefficient D ∞ of the fuel, since the temperature of the flame T f was predicted to increase, as found by Gu et al. (2016) A clear trend is that the dark region height increased while the luminous region height decreased with the addition of CO 2 .Since few soot particles existed in the dark region, the increase in the dark region height indicated that the formation of nascent soot began at a higher position in the flame, which suggests that the CO 2 dilution slowed down the nucleation rate.Since the total heights and shapes of the flames stayed nearly the same, the following results of the LII measurements of different flames were compared at the same position in the flame.

LII Fluence Curves
Having discussed the influence of CO 2 addition on the nucleation rates in the last section, the LII fluence curves are plotted in this section to gain information about the optical properties of soot particles.The measurements of the LII fluence curves were carried out at four different positions in each flame.The position with 5 mm radial distance to the axis of flames at 30mm HAB was considered to be the edge of flames.The results of the measurements are shown in Fig. 5, where the curves were normalized to the strongest peak LII signal from each height.
It can be seen from Fig. 4 that all the presented curves followed the same trend that the peak LII signal increased rapidly with fluence and reached a plateau at a laser fluence higher than ~0.3 J cm -2 .A slight decrease in the LII signals at higher fluence can be seen in the conventional shape of the LII fluence curves, because of the sublimation of soot particles (Olofsson et al., 2013;Migliorini et al., 2015;Olofsson et al., 2015).However, this decrease was not found in our results, which might be caused by the fact that the top-hat profile of the laser was still not equally distributed enough, with a standard deviation of 15% observed by a beam profiler.The three presented plots showed the same trend that higher laser energy was required for the signal to reach the plateau at a lower position in the flame, indicating that higher laser energy was required for soot particles to sublimate at lower heights.This behavior could be explained by the fact that soot particles became more mature and less transparent at higher flame heights in diffusion flames, which was also found in previous studies (Smooke et al., 2005;Michelsen, 2017).
By comparing the fluence curves of different flames, it was found that the peak LII signal shifted slightly to a higher laser fluence with carbon-dioxide addition on the fuel side, as shown in Fig. 5.This behavior indicated that soot particles  became more transparent with carbon-dioxide dilution.This was in good agreement with the speculation made in the previous section that CO 2 dilution on the fuel side slowed down the growth of soot particles, making them became more nascent and with weaker absorbency at the same position in the flames.

Soot Particle Sizes
The laser fluence needs to be properly selected in order to measure the soot particle size correctly.It can be seen in Fig. 4 that the rising trend of the curves began to slow down around the fluence of 0.25 J cm -2 , when the sublimation of soot particles started.This fluence has two advantages for measuring particle sizes, as Sun et al. (2015) pointed out.The influence of sublimation on LII signals stayed low at this fluence, and the LII signals were also quite steady, since the peak temperature of the soot particles was reached around this fluence.Based on these results, the laser fluence was chosen to be 0.254 J cm -2 during the measurement of soot particle sizes.
A comparison of the measured LII signal decays τ derived from the experimental data with the simulated ones based on an LII model is needed to acquire the primary particle sizes (Will et al., 1998;Lehre et al., 2003).The LII model we used in this work was demonstrated in detail in our previous work (Wu et al., 2017).However, while the simulated LII signals agreed well with the measured ones in the low fluences region, when applied at high fluence a difference can be found between them.This difference in the high fluence region is mainly caused by the inadequacy of the modeling of sublimation, which is still in the research stage.
The selection of the accommodation coefficient and refractive index is of great importance.In this work, the frequently used value of 0.3 is employed for the thermal accommodation coefficient, while the value of 1.56-0.46i is used for the refractive index (Schulz et al., 2006;Michelsen et al., 2007).After setting down the input parameters, including the flame temperature (1700 K), laser fluence (0.254 J cm -2 ), excitation wavelength (1064 nm) and detection wavelength (450 nm) in the model, simulated LII signals corresponding to different primary particle sizes were obtained, as shown in Fig. 6.Previous studies pointed out that the LII decay signal was similar to an exponential function, shown as Eq. ( 2) (Schulz et al., 2006).
where S is the LII decay signal, and S 0 is the peak LII signal.
As mentioned above, the excitation laser fluence of this measurement was set at 0.254 J cm -2 , which is high enough for soot particles to sublimate.In this condition, the LII signal was mainly controlled by the rates of evaporative heat loss and mass transfer shortly after the excitation laser pulse, so it could not be fitted by an exponential function.
To avoid the influence of sublimation and obtain the correct signal decays, the range of the LII decay signal for exponential fit was selected to be from 150ns to 2000ns after the excitation laser pulse, when the sublimation was almost over and the soot particle sizes were considered to be constant (Vander et al., 1998).Consequently, the relation between the particle size d p and the signal decay was revealed.As shown in Fig. 7, a linear function well fitted this relation: The measured LII signal decays τ were derived from the LII signals which were recorded by a photomultiplier connected to a digital oscilloscope.To study the effects of CO 2 addition to fuel on soot particle sizes, measurements were carried out at four different positions in the flames.Fig. 8 shows the normalized LII signals at 40 mm HAB in different flames.Signal decays were obtained in the same way as mentioned above.Applying the linear relation between signal decays and soot particle sizes, as shown as Eq.(3), the soot particle sizes were calculated, and the results are shown in Fig. 9.
Fig. 9(a) shows that adding CO 2 to the fuel side had a significant impact on decreasing particle size.This might be caused by the decrease in the nucleation rate and increase in the oxidation rate.In the previous two sections it was demonstrated that CO 2 dilution slowed down the nucleation rate of soot particles, and the effects of CO 2 dilution on the oxidation rate still need to be discussed.
The relation between soot particle sizes and their heights above the burner are revealed in Fig. 9(b).With the height increasing from 18 mm to 40 mm above the burner, the soot particle sizes increased first and then decreased.Since the soot particle sizes stayed nearly the same in the inception region, increased in the surface growth region and decreased in the oxidation region, the boundary between surface growth This boundary was found to become higher with the addition of CO 2 , indicating that the oxidation of soot particles started at a higher position.Since the LII signal was hard to detect and the signal-to-noise ratio was too low in positions below 18 mm HAB, the boundary between the inception region and surface growth region, as well as the surface growth rate, could not be determined.It can be seen from Fig. 9(b) that the decreasing parts of the three curves stayed nearly parallel, indicating that the addition of CO 2 had little effect on the oxidation of soot particles.Consequently, the decrease in particle sizes with CO 2 dilution was mainly caused by the decrease in the nucleation rate.

CONCLUSIONS
Addition of CO 2 to the fuel side of a laminar co-flow ethylene/air diffusion flame was studied experimentally in three regards in the current work: the flame heights, particle sizes and absorption function of soot particles.Both the luminous flame height and the dark region height were determined by using image processing.Laser-induced incandescence was applied to investigate the absorption function and sizes of soot particles.Information about the The luminous flame height decreased with the CO 2 addition, while the height of the dark region increased.It is thus concluded that CO 2 addition to the fuel side slowed down the nucleation rate and thus compressed the formation of nascent soot A slight decrease in the absorbency of soot particles was observed.This indicates that CO 2 dilution made soot particles less mature.From the results of the particle size measurement, it was concluded that adding CO 2 to the fuel side reduced the particle sizes and changed the boundary between the surface growth region and the oxidation region of flames.

Fig. 5 .
Fig. 5.The fluence curves plotted at 30 mm HAB in flames fueled with gases containing different volumetric CO 2 fractions.

Fig. 9 .
Fig. 9. Calculated particle sizes in different flames.(a) The relation between particle sizes and volumetric CO 2 fraction of CO 2 .(b) The relation between particle sizes and the height above the burner.

Table 1 .
Flow rate conditions Fuel side.