Modeling and measurements of size distributions in premixed ethylene and benzene flames

Available online at www.sciencedirect.com Proceedings of the Combustion Institute 32 (2009) 705 711 Proceedings of the Combustion Institute www.elsevier.com/locate/proci Modeling and measurements of size distributions in premixed ethylene and benzene flames Carlos A. Echavarria a, Adel F. Sarofim a, JoAnn Slama Lighty a, *, Andrea D Anna b a Chemical Engineering, University of Utah, 50 South Central Campus Drive, Room 3290 MEB, Salt Lake City, UT 84112, USA b Dipartimento di Ingegneria Chimica, Università Federico II di Napoli, Naples, Italy Abstract This study demonstrates the major differences in the evolution of the particle size distributions (PSDs), both measured and modeled, of soot in premixed benzene and ethylene flat flames. In the experiments, soot concentration and PSDs were measured by using a scanning mobility particle sizer (SMPS, over the size range of 3 80 nm). The model employed calculations of gas phase species coupled with a discrete sectional approach for the gas-to-particle conversion. The model includes reaction pathways leading to the formation of nano-sized particles and their coagulation to larger soot particles. The particle size distribution, both experimental and modeled, evolved from a single particle mode (the nucleation mode) to a bimodal size distribution. An important distinction between the results for the ethylene and benzene flames is the behavior of the nucleation mode which persists at all heights above the burner (HAB) for ethylene whereas it was greatly suppressed at greater HAB for the benzene flames. The explanation for the decreased nucleation mode at higher elevations in the benzene flame is that the aromatics are consumed in the oxidation zone of the flame. Fair predictions of particle-phase concentrations and particle sizes in the two flames were obtained with no adjustments to the kinetic scheme. In agreement with experimental data, the model predicts a higher formation of particulate in the benzene flame as compared with the ethylene flame. Ó 2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved. Keywords: Soot; Particle size distribution (PSD); Premixed flame; Particle inception 1. Introduction Soot formation during combustion continues to be a major subject of experimental and theoretical study due to the impact of soot on human health and radiation forcing. However, soot formation is a complex process involving a great number of chemical and physical steps, and is still * Corresponding author. Fax: +1 801 585 9291. E-mail address: jlighty@utah.edu (J.S. Lighty). incompletely understood. It is widely accepted that the process of soot formation can be described by the steps of molecular precursor formation, particle inception, coagulation and soot growth, particle agglomeration and soot oxidation [1 3]. It is also widely accepted that the formation of soot from aliphatic fuels generally proceeds through the relatively slow conversion of the aliphatic molecule to aromatic compounds that can rapidly undergo polymerization to soot or growth via hydrogen abstraction and acetylene addition. This study focuses on particle inception 1540-7489/$ - see front matter Ó 2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.proci.2008.06.172

706 C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 and growth in premixed flames of ethylene and benzene in order to determine the impact of the added step of aliphatic to aromatic formation on the evolution of the particle size distributions (PSDs) in these two classes of flames. Most previous studies have focused on characterizing soot properties for premixed aliphatic fuels (ethylene, acetylene) as compared to those of premixed benzene flame [4 8]. The studies support the thesis that soot forms early in flames of aromatic compounds relative to those of aliphatic compounds consistent with the added time needed to form the first aromatic ring in an aliphatic flame. In addition, the structure of the compounds formed in aromatic flames are consistent with aromers formed by polymerization of aromatic compounds whereas a greater weighting is given to the polycondensed structures formed by the HACA (Hydrogen abstraction carbon addition) mechanism in flames of aliphatic fuels [7]. Traditionally, temperature profiles, soot volume fraction, particle size distribution, morphology, and structure are commonly characterized by methods such as light scattering, UV absorption and fluorescence, thermocouple particle densitometry (TPD), transmission electron microscopy, etc., [4,5,9]. Recent studies have added differential mobility analyzers (DMA) of particle size to investigate soot formation in flames [10 12]. The use of DMA provides spatially resolved, rapid, and online measurements of particle size down to 3 nm. This study applies the DMA to examine the evolution of the PSD in ethylene and benzene flames and compares the results with calculations using a sectional method in order to obtain insight on the particle inception and agglomeration. It is, of course, recognized that completely comparable conditions for the ethylene and benzene flames cannot be achieved because of their different sooting tendencies. Equivalence ratios of 0.69 and 0.89 were selected for both the ethylene and benzene flames. Although attempts were made to obtain comparable temperatures, the greater soot concentrations and temperature gradients in the benzene flames made this difficult. However, compensations for differences in conditions, soot concentrations and temperatures, are taken into account in the model comparisons. 2. Experimental section A premixed flat-flame burner was used in this study. The system consisted of a stainless steel chamber where fuel and air were injected and mixed prior to entering the burner. A 25-mm thick bed of mixing beads inside the burner chamber completed the mixing. The flame was stabilized over a tube bundle through which the mixture passed in laminar flow. To shield the premixed flame from atmospheric interference a nitrogen shroud was utilized. Two different fuel-rich conditions (C/O = 0.69 and 0.89) were run for both atmospheric-pressure ethylene air and benzene air flames. Air and ethylene (>99.99%) were fed to the burner using Brooks 5850E mass flow controllers. Benzene (HPLC grade) was delivered via a bubbler that was optimized to provide a uniform fuel flow to the burner. Temperature profiles along the centerline were measured using a 0.2032-mm uncoated Type-B thermocouple. Comparisons to uncoated versus coated yielded a difference in temperature of approximately 40 C. The thermocouple was inserted into the flame using a fast insertion mechanism. For each measurement, the transient response of the thermocouple was recorded at a sampling rate of 50 samples per second. Radiation correction was applied using the methodology developed by Rosner et al. [9]. PSDs were measured with a scanning mobility particle sizer (SMPS), which consisted of a TSI 3080 classifier with a Model 3085 Nano DMA and 3025 ultrafine condensation particle counter (UCPC). The SMPS was optimized to operate in the 3 80 nm range with a sheath flow of 10 L/min and an aerosol sample flow of 1 L/min. Corrections for penetration efficiency, into the probe and probe orifice, and diffusion losses during transport were applied following the procedure presented by Minutolo et al [13]. Details on the penetration efficiency are given in the Supplemental Material, Fig. S1. Corrections due to diffusion losses in the SMPS were conducted using the AIM software upgrade. The dilution system, similar to that of Zhao et al. [10] and Kasper et al. [14] with some minor changes in probe size (OD = 11 mm) and pinhole diameter (0.24 mm), yielded dilution ratios greater that 10 4, which minimized wall losses and quenched reactions or coagulation that would otherwise occur in the sampling system. 3. Model summary The details of the model have been reported previously [8,15]. The model includes mechanisms for PAH formation and reaction pathways responsible for nano-sized particle nucleation, i.e. the transition from gas-phase species to nascent particles, and their coagulation to larger soot particles. A discrete-sectional approach is used for the modeling of the gas-to-particle process; the ensemble of aromatic compounds is divided into classes of different molecular mass and all reactions are treated in the form of common gas-phase chemistry up to aromatics containing four rings. Particle size distribution is defined by a range of sections, each containing a nominal hydrocarbon species in order of increasing atomic mass. Two bins are assigned to each particle size, one for

C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 707 the stable species and the other for the radical. Twenty-six size sections (x2 to allow for radicals) are used in a geometric series with a carbon number ratio of 2.2 between sections. The carbon number range is 20 to 410 8 which represents a particle size range of 1 250 nm. In this approach, the molecular mass distribution of the species is obtained from the calculation and not hypothesized a priori. Premixed flame modeling was performed using the CHEMKIN software package. A modified version of the gas-phase Interpreter was used allowing the handling of molecules with molecular masses sufficiently large to follow soot particle inception. Particles are assumed to be spherical with a density of 1.2 for PAHs to 1.8 for 20 nm particles. 4. Results and discussion 4.1. Temperature profiles and particle size distributions Figure 1a (ethylene) and b (benzene) present the flame temperatures as functions of height above the burner (HAB) for C/O ratios of 0.69 and 0.89. Temperature peaks in the benzene flame were found to be higher than in the ethylene flame for the same C/O ratio. The high soot concentrations for the benzene flame for C/O = 0.89 were responsible for both the large temperature drop with the increase in C/O from 0.69 and also for the large temperature gradient resulting from radiation cooling. The temperature profiles in Fig. 1 were used as inputs for the model calculations. The size distributions determined by a number of investigators [10 12,16] using a DMA in laminar premixed ethylene flames under lightly-sooting conditions showed the presence of nucleation and agglomeration modes with the relative magnitude changing with combustion conditions and height above the burner. The results for the present study for a C/O ratio of 0.69 for ethylene and benzene are presented in Fig. 2. At lower HAB, 7 and 10 mm, the size distribution for the ethylene flame is unimodal (the nucleation mode) showing a monotonic decrease in number concentration starting at the detection limit of 3 nm of the SMPS. As the height increases above 10 mm a maximum at around 4 nm is discernible in the size distribution of the nucleation mode. At heights starting at 11 mm a bimodal distribution is observed with the larger sizes corresponding to the agglomeration mode. The number of particles in the nucleation mode however remains high throughout the flame. For the benzene flame the bimodal distribution is evident starting at the lowest HAB (7 mm). The number concentration at the lowest size of 3 nm of the nucleation mode decreases with HAB and is difficult to characterize at higher HAB. The decrease in the number of particles in the nucleation mode is a distinguishing feature of the benzene flame. Given that the soot volume fraction is higher for the benzene flame at the same C/O ratio, the discussion from here on will be conducted using the combined results of the model and experiments. 4.2. Measurements and model predictions for ethylene, C/O = 0.69 and 0.89 In Fig. 3 comparisons are provided of model and experimental results for a lightly-sooting ethylene flame (soot <5.0E-7 g/cm 3 ). The model data were shifted downstream by 2 6 mm to be compared with experimental data in order to take probe effects into account (a discussion of the effects of the perturbations by the dilution probe is provided by Zhao et al. [10], and experimental evidence of the probe effects is shown in Supplemental Material, Fig. S2). The amount of shift a 2000 b 2000 Flame Temperature, K C/O = 0.69 C/O = 0.89 Flame Temperature, K C/O = 0.69 C/O = 0.89 Fig. 1. Temperatures as a function of height above burner (HAB) in flames of (a) ethylene and (b) benzene for C/O ratios of 0.69 (circles) and 0.89 (squares).

708 C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 d Fig. 2. Particle size distributions in ethylene and benzene flames with a C/O ratio of 0.69 for (a) ethylene and (b) benzene at HAB of 7 mm (M), 8 mm (s), 9 mm (N), 10 mm (d), 11 mm (h), 12 mm (e), 14 mm ( * ). 1.E-06 N, #/cm 3 1.E+06 1.E+05 Concentration, g/cm 3 1.E-07 1.E-08 1.E-09 1.E-10 1.E-11 1.E-12 Fig. 3. Results for an ethylene Flame, C/O = 0.69: (a) Size distributions at HAB of 7 mm ( model, M data), 8 mm (- - - model, s data), 9 mm(-. -. - model, N data), and 10 mm (- - - model, d data). (b) Size distributions at HAB of 11 mm ( model, h data), 12 mm(- - - model, e data), and 14 mm ( model, * data). (c) Number concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data). (d) Mass concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data). was determined by the best match between model predictions and data for the lowest HAB. Figure 3a and b provide comparisons of the modeled flame results with experimental data for HAB of 7, 8, 9, 10 and 11, 12, 14 mm, respectively. The nucleation mode dominated the results at the smaller HAB (Fig. 3a) whereas clear evidence of the agglomeration mode was seen higher above the flame. Fair agreement was obtained between the model results (lines) and the experimental data

C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 709 (symbols) although the model showed only a point of inflexion between the nucleation and agglomeration modes while the data show a clear minimum. Particles in size intervals of 3 10 nm and greater than 10 nm were integrated to provide rough measures of the contents of the nucleation and agglomeration modes, respectively. The value of 10 nm was picked roughly to separate the nucleation and agglomeration modes for ethylene. The number concentrations in Fig. 3c show how the concentration of particles in the nucleation mode exceeded that in the agglomeration mode for all HAB. The number of particles in the nucleation model rose rapidly and approached a relatively constant level of 10 13 cm 3 particles at HAB >5 mm. These results will be contrasted later with the behavior of particles in the nucleation mode in benzene flames. The particles in the agglomeration mode (dark symbols in Fig. 3c) showed a fairly steady growth in number concentration up to HAB of about 11 mm, and remained fairly constant at higher elevations. In Fig. 3d, the mass concentrations of the nucleation and agglomeration mode are provided as a function of HAB. The mass of particles in the nucleation mode immediately increased after the flame front and approached a constant value (the predicted values passed through a peak). At the end of the flame most of the mass of the particles was in the agglomeration mode and increased with increased HAB, with satisfactory agreement between the modeled and measured values. The data and discussion for the ethylene flame at C/O = 0.89 is given in the supplemental material and Supplemental Material, Fig. S3. The results were comparable to those discussed above. 4.3. Measurements and model predictions, benzene flame, C/O = 0.69 and 0.89 Comparisons of model predictions and data for the benzene flame with a C/O ratio of 0.69 are presented in Fig. 4. Figure 4a and b provide the PSDs for HAB of 7, 8, 9 and 11, 12, 14 mm, respectively. A shift of 6 mm in the modeled data downstream of the burner was needed to obtain reasonable correspondence between data and predictions. In this case, a particle size of approximately 5 nm was used to determine the 1.E-05 N, #/cm 3 Concentration, g/cm 3 1.E-06 1.E-07 1.E-08 1.E-09 1.E-10 Fig. 4. Results for benzene flame, C/O = 0.69: (a) Size distributions at HAB of 7 mm ( model, M data), 8 mm (- - - model, s data), 9 mm(...model, N data), and 10 mm (-. -. - model, d data). (b) Size distributions at HAB of 11 mm (...model, h data), 12 mm(-. -. -. model, e data), and 14 mm ( model, * data). (c) Number concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data). d) Mass concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data).

710 C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 separation between the nucleation and agglomeration modes. In Fig. 4a, the data at 7 and 8 mm showed a very narrow nucleation mode; otherwise, most of the distribution was dominated by the agglomeration mode. Small nucleation modes are apparent in Fig. 4b, for both measurements and predictions. These results are reflected in Fig. 4c which shows a spike for particles in the nucleation mode around 7 mm HAB falling to negligible levels above 15 mm HAB. By contrast the agglomeration mode particles (bounded by the upper measurement limit of 80 nm) persist 1.E-05 1.E-06 N, #/cm 3 Concentration, g/cm 3 1.E-07 1.E-08 1.E-09 1.E-10 1.E-11 1.E+06 1.E-12 Fig. 5. Results for benzene flame, C/O = 0.89: (a) Size distributions at HAB of 7 mm ( model, M data), 8 mm (- - - model, s data), and 9 mm(-. -. - model, N data). (b) Size distributions at HAB of 10 mm ( model, d data), 11 mm (- - model, h data), 13 mm (...model, e data), and 15 mm (-. -. - model, * data). (c) Number concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data). (d) Mass concentration as a function of HAB for nucleation (- - - model, h data) and agglomeration ( model, j data). 4.5 0.9 Mole Fraction, % 3.5 2.5 1.5 0.5 x50 T, K Fig. 6. Concentration of soot precursors as a function of HAB for (a) ethylene (C/O = 0.69) (b) benzene (C/O = 0.89). Temperature Kd, C 2 H 4 s,c 6 H 6 M,C 2 H 2 h. Mole Fraction, % 0.7 0.5 0.3 0.1 T, K

C.A. Echavarria et al. / Proceedings of the Combustion Institute 32 (2009) 705 711 711 throughout the flame. Figure 4d shows the mass of the agglomeration mode particles dominate at high HAB with fair agreement between the measured and predicted values, yielding soot levels of about 1.E-6 g/cm 3. The observation to be emphasized is the essential disappearance of the mass of the nucleation mode above 11 mm HAB. Similar results are obtained for a benzene flame with a C/O = 0.89, as summarized in Fig. 5. The experimental results for the particle size distributions are shown in Fig. 5a and b. The experimental results in Fig. 5a for HAB of 7, 8, and 9 mm showed a narrow nucleation mode only at 7 mm HAB. The predictions best fit the data in Fig. 5a when the modeled data were shifted downstream of the burner by 4 mm. Figure 5b provides the experimental particle size distribution for HAB values 10, 11, 12, and 14 with no evidence of a nucleation mode. Again, the experimental results are well fitted by the predictions when the data were shifted towards the burner by 4 mm. The predicted number and mass concentrations for the nucleation mode exceed the measured values which fall to negligible levels at about 15 mm HAB in Fig. 5c and d. This may be due to the loss of smaller particles in the sampling probe. Good agreements are obtained (Fig. 5c and d) for the agglomeration mode between measurements and predictions. 5. Discussion and conclusions The most striking difference between Figs. 3 5 is the persistence of the nucleation mode with HAB in Fig. 3 and Supplemental Material, Fig. S3 for ethylene flames compared to its rapid decline in Figs. 4 and 5 for benzene flames with the same C/O ratios. Since conditions other than fuel changed between the flames, the question still remains as to whether the difference is accountable by the fuel change or by other flame conditions. Previous studies [12,16] have suggested that high temperatures, not fuel composition, determined the bimodal versus unimodal size distributions. In order to evaluate the role of flame chemistry the concentration profiles of the soot precursors acetylene and benzene are plotted for the ethylene (C/O = 0.69) and benzene (C/ O = 0.89) flames (Fig. 6) since these two flames had similar temperature distributions. In the benzene flame, benzene and, to a lesser extent, acetylene showed high concentrations near the burner (<3 mm HAB) which is consistent with the observed location of the nucleation mode in the benzene flame experiments and predictions. By contrast, in the ethylene flame, the acetylene and benzene concentrations persist at even high HAB which is again consistent with the observed persistence of the nucleation mode at larger HAB. These results suggest that benzene flames have a distinct behavior because the PAH are consumed in the oxidation zone thereby eliminating the nucleation peak in the upper regions of the flame [5]. Given the good agreement between the model predictions and experiments in this study, future modeling work should explore the separate roles of fuel chemistry and flame temperature on particle size distribution. For example, for the comparison in Fig. 6, the soot concentrations formed in the benzene flame were greater than that in the ethylene flame. The model can be used to explore the role of soot concentration on the nucleation mode since all conditions are difficult to control independently in the experiments. Acknowledgment The support of the National Science Foundation, Grant No. 0304433 is appreciated. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.proci.2008.06.172. References [1] H. Bockhorn, Soot Formation in Combustion; Mechanisms and Models, Springer-Verlag, Berlin, 1994. [2] M. Frenklach, Phys. Chem. Chem. Phys. 4 (11) (2002) 2028 2037. [3] M. Frenklach, H. Wang, Soot Formation in Combustion; Mechanisms and Models, Springer-Verlag, Berlin, 1994. [4] M. Alfè, B. Apicella, R. Barbella et al., Nanostructural and Optical Analysis of Soot in Aliphatic and aromatic Premixed Flames, Atti del 28th Meeting of The Italian Section of The Combustion Institute, Napoli, 2005. [5] M. Alfè, B. Apicella, R. Barbella, A. Tregrossi, A. Ciajolo, Proc. Combust. Inst. 31 (2007) 585 591. [6] K.H. Homann, H.G. Wagner, Proc. Roy. Soc. A 307 (1968) 141 152. [7] A. Violi, Combust. Flame 139 (2004) 279 287. [8] A. D Anna, M. Alfe, B. Apicella, A. Tregrossi, A. Ciajolo, Energ. Fuels 21 (5) (2007) 2655 2662. [9] D.E. Rosner, C.S. McEnally, U. Koylu, L.D. Pfefferle, Combust. Flame 109 (1997) 701 720. [10] B. Zhao, Z. Yang, J. Wang, M.V. Johnston, H. Wang, Aerosol Sci. Technol. 37 (2003) 611 620. [11] M.M. Maricq, Combust. Flame 137 (2004) 340 350. [12] M.M. Maricq, Combust. Flame 144 (2006) 730 743. [13] P. Minutolo, A. D Anna, A. D Alessio, Combust. Flame 152 (2008) 287 292. [14] M. Kasper, K. Siegmann, K. Sattler, J. Aerosol Sci. 28 (1997) 1569 1578. [15] A. D Anna, A. Violi, A. D Alessio, A.F. Sarofim, Combust. Flame 127 (2001) 1995 2003. [16] B. Zhao, Z. Yang, Z. Li, M.V. Johnston, H. Wang, Proc. Combust. Inst. 30 (2005) 1441 1448.