Energy Fuels 2009, 23, 5338–5348 Published on Web 10/26/2009
: DOI:10.1021/ef900492g
Generation of Silica Nanoparticles in Turbulent Non-premixed Flames with Oxygen Enrichment Soo Ho Bae and Hyun Dong Shin* Department of Mechanical Engineering Korea Advanced Institute of Science and Technology 373-1, Guseong Dong, Yuseong Gu, Daejeon 305-701, Republic of Korea Received May 20, 2009. Revised Manuscript Received July 13, 2009
Silica nanoparticles were made via the gas-to-particle conversion of tetraethylorthosilicate (TEOS) in co-flowing turbulent methane-oxygen-enriched non-premixed flames. The effects of fuel velocity, oxygen concentration, flame residence time, and temperature distribution on the characteristics of the silica nanoparticles were investigated. The flame length was measured by OH chemiluminescence, using an intensified charge-coupled device (ICCD) camera. The primary particle diameter of the silica nanoparticles was quantitatively measured by transmission electron microscopy (TEM). Particle number concentration, as well as the geometric mean diameter and standard deviation, were measured by a scanning mobility particle sizer (SMPS) spectrometer. Adiabatic temperatures according to oxygen concentration and mixture fraction were calculated from thermodynamic equilibrium calculations, considering chemical species using CHEMKIN EQUIL code. Flame temperature, velocity, and turbulent intensity distributions were calculated using Fluent software adapted to the presumed probability density function (PDF) model, considering chemical species in the turbulent non-premixed methane-oxygen-enriched flame burner without including the particles. For each oxygen concentration condition, the flame residence time, average primary particle diameter, and geometric mean diameter decreased as the fuel velocity increased. Typically, agglomerates of silica nanoparticles were made on averaged primary particle diameters of 9-15 nm and geometric mean diameters of 60-100 nm, according to the oxygen concentration, which varied from 50% to 100% for first jet velocities of 40 and 70 m/s.
“white carbon” as a reinforcing filler for rubber, although precipitated silicas were eventually used for that application. Today, fumed silica is found in optical fibers, filler in silicone rubber, thickening, antisettling, and reinforcing agents, catalyst carriers, and microelectronics polishing, to name a few.6 Gas-phase methods using a tube reactor, flame reactor, plasma reactor, and a laser ablation reactor have been investigated.5-8 Gas-phase combustion synthesis offers significant advantages over other processes. It is an inexpensive processing method that can provide an ideal environment for the high-temperature flame that is needed for the production of pure particles. In addition, flame reactors are known to be easily scaled up for practical applications and, therefore, can be used for continuous processing.9 However, because of the harsh environment, steep temperature gradients, and rapid particle evolution, the fundamentals of particle formation and growth are not well-understood.8 The characteristics of flame-synthesized particles are controlled by many factors. These include the mixing of the reactants and precursor, the overall composition, the flame temperature, the time-temperature behavior, and the rapid quenching of the gas-particle flow.1,3,10-12 Material
1. Introduction In recent decades, interest in synthesizing particles has increasingly focused on the size effects exhibited by nanostructured materials. Particles with a very narrow size distribution and a well-controlled phase composition and morphology (called ‘‘functional nanoparticles’’) have become desirable products.1 The size effects exhibited by nanoparticles compared to the bulk material are a consequence of their large surface-to-volume ratio. For example, in a 4-nm-sized particle, half of the molecules that form the nanostructure are actually located at the surface. As a result, the lattice structure is affected and dramatic changes in the physical and chemical properties are observed, when compared to the bulk material. For a given material, the melting temperature, band gap, and catalytic behavior may all exhibit changes, along with other mechanical, magnetic, and optical properties.1,2 A variety of technologies for synthesizing powders and films has been discussed in previous research.3 Recently, gas-phase synthesis of various types of nanoparticles (including silica, titania, aluminum oxides, iron oxides, zinc oxides, and metal nanoparticles) has been conducted.4-6 With regard to fumed silica, the original objective was to develop a
(7) Stark, W. J.; Pratsinis, S. E. Powder Technol. 2002, 126 (2), 103– 108. (8) Choi, M. S. J. Nanopart. Res. 2001, 3, 201–211. (9) Wegner, K.; Pratsinis, S. E. Chem. Eng. Sci. 2003, 58 (20), 4581– 4589. (10) Pratsinis, S. E.; Wenhua, Z.; Vemury, S. Powder Technol. 1996, 86 (1), 87–93. (11) Jang, H. D. Aerosol Sci. Technol. 1999, 30, 477–488. (12) Kusters, K. A.; Pratsinis, S. E. Powder Technol. 1995, 82 (1), 79–91.
*Author to whom correspondence should be addressed. Tel.: 82-42350-8821. Fax: 82-42-350-8820. E-mail:
[email protected]. (1) Roth, P. Proc. Combust. Inst. 2007, 31, 1773–1788. (2) Preining, O. J. Aerosol Sci. 1998, 29 (5/6), 481–495. (3) Kodas, T. T.; Hampden-Smith, M. J. Aerosol Processing of Materials; Wiley-VCH: New York, 1999. (4) Rosner, D. E. Ind. Eng. Chem. Res. 2005, 44 (16), 6045–6055. (5) Wooldridge, M. S. Prog. Energy Combust. Sci. 1998, 24 (1), 63–87. (6) Pratsinis, S. E. Prog. Energy Combust. Sci. 1998, 24 (3), 197–219. r 2009 American Chemical Society
5338
pubs.acs.org/EF
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
OH chemiluminescence is free from background luminosity, because it is located in the ultraviolet (UV) range.22,23 Therefore, the OH chemiluminescence can be properly used as a marker for deciding the flame length behavior. However, note that obtaining quantitative information on the concentration of OH in the present study is impossible. This article represents a study performed to elucidate the controlling parameters of silica formation in coaxial turbulent nonpremixed flames with oxygen enrichment. Specifically, the effects of fuel jet velocity, oxygen concentration in the oxidant stream, flame residence time, and flame temperature on the characteristics of silica nanoparticles were analyzed. These particle characteristics include the average primary particle diameter and its standard deviation, the geometric mean diameter and its standard deviation, and the particle number concentration. Diameter was measured through transmission electron microscopy (TEM), while particle number concentration was ascertained by a scanning mobility particle sizer (SMPS) spectrometer. The flame length and the structure of the reaction zone have been analyzed through images of the OH chemiluminescence. Definitions for the global flame residence time and the global strain rate were examined for coaxial turbulent jet flames in general, and the assumptions used in various simplified expressions were reviewed. Furthermore, adiabatic temperatures according to oxygen concentration and mixture fraction were calculated from thermodynamic equilibrium calculation, considering chemical species using CHEMKIN EQUIL code. From these results, it was necessary to include the most important species;i.e., CH4, O2, CO2, H2O, N2, CO, H2, OH, O, and H;in the combustion of oxygen-enriched and pure-oxygen methane. If these species are not considered, the flame temperature will be significantly overestimated,26 which also influences the predictions of the characteristics of the particles. The flame temperature was calculated using Fluent, which is a commercially available computational fluid dynamics (CFD) software package. It was adapted to the geometry and conditions of the turbulent non-premixed methane-oxygen-enriched flame burner without including the particles.
properties, residence time, and process temperature play an especially key role in determining the extent of particle coalescence or sintering, and eventual powder morphology and crystallinity.6,13,14 Kusters and Pratsinis12 showed that the elevation of the operation temperature generally increases the reaction rate but decreases the surface tension. This provided an excellent way to improve monodispersity. However, they also showed that a high reaction rate and a low surface tension resulted in larger aerosol concentrations and smaller particle sizes. Using silica property data, Xiong and Pratsinis15 determined the characteristic sintering time of silica based on a viscous flow mechanism.16 Ehrman et al.17 calculated the characteristic sintering time of silica (for viscous flow) from property data while accounting for the effect of hydroxyl groups.15-17 Their results examined the effects of temperature on the characteristic sintering time of the silica particles. For turbulent jet flames, Becker and Liang18 suggested a global flame residence time (τR), which is given by mass of flame F Vflame Vflame L3 L ≈ ¥ 2 ≈ 2 τR ¼ fuel mass flowrate m_ f =fsto df Uf df Uf Uf ð1Þ where F¥ is the density of the surrounding fluid, fsto the stoichiometric value of the mixture fraction, df the diameter of the fuel jet, Uf the average fuel jet velocity, L the visible flame length, and Vflame the flame volume, which is defined as the volume inside the mean stoichiometric contour. Use of the flame volume in defining the residence time may be undesirable, because of difficulties in the direct measurement of Vflame. Since Becket et al. used only scaling laws, the simplification shown in eq 1 is possible if Vflame scales with some physical dimensions of the flame. For example, L is a convenient quantity that is easily accessible from experimentation.18-20 Several methods to define the flame length have been used by previous researchers. These techniques include direct still or video camera photography, luminous flame length obtained by an intensified charge-coupled device (ICCD) camera, and average flame length obtained from accumulated chemiluminescence images using an adequately filtered ICCD camera. Among those methods, the detection of chemiluminescence has been widely used by previous researchers to describe various aspects of the combustion process qualitatively.21-25 Chemiluminescence intensities are roughly proportional to the reaction intensity.
2. Experimental Section 2.1. Experimental Apparatus. The experimental setup for the synthesis of silica nanoparticles from a precursor in turbulent nonpremixed flames with oxygen enrichment is shown in Figure 1. This apparatus consists largely of a co-flowing burner, a saturator, a mass flow controller (MFC), a camera, an ICCD camera, a thermophoretic sampling device, and a SMPS spectrometer. The precursor used in this work is tetraethylorthosilicate (TEOS:SiO4C8H20, 98%, SAM). Nitrogen (N2) and methane (CH4: 99.999%) were used as the carrier gas and fuel, respectively. The co-flowing turbulent non-premixed flame burner is composed of stainless steel and consists of two concentric tubes. The inner/outer diameter of the first tube is 1 mm/1.5 mm, whereas the inner/outer diameter of the second tube is 20 mm/ 20.5 mm. The precursor-laden carrier gas and the fuel are delivered through the first (center) tube, and the oxidant, which is a mixture of O2 and N2, is delivered through the second tube. Saturation of the carrier gas by the precursor is controlled by a saturator that is comprised of a temperature controller (Autonics, TZ4SP), a T-type thermocouple (Omega
(13) Mueller, R.; Kammler, H. K.; Pratsinis, S. E.; Vital, A.; Beaucage, G.; Burtscher, P. Powder Technol. 2004, 140 (1/2), 40–48. (14) Tsantilis, S.; Briesen, H.; Pratsinis, S. E. Aerosol Sci. Technol. 2001, 34, 237–246. (15) Xiong, Y.; Pratsinis, S. E. J. Aerosol Sci. 1993, 24, 283–300. (16) Kingery, W. D.; Bowen, H. K.; Uhlmann, D. R. Introduction to Ceramics; John Wiley & Sons: New York, 1976. (17) Ehrman, S. H.; Friedlander, S. K.; Zachariah, M. R. J. Aerosol Sci. 1998, 29, 687–706. (18) Becker, H. A.; Liang, D. Combust. Flame 1982, 44, 305–318. (19) Chen, J. Y.; Kollmann, W. Combust. Flame 1992, 88, 397–412. (20) Kim, S. H.; Yoon, Y. B.; Jeung, I. S. Proc. Combust. Inst. 2000, 28, 463–471. (21) Schefer, R. W. Combust. Sci. Technol. 1997, 126, 255–270. (22) Tsushima, S.; Saitoh, H.; Akamatsu, F.; Katsuki, M. Symp. (Int.) Combust. 1998, 27 (2), 1967–1974. (23) Renard, P. H.; Rolon, J. C.; Thevenin, D.; Candel, S. Symp. (Int.) Combust. 1998, 27 (1), 659–666. (24) Sautet, J.; Salentey, L.; DiTaranto, M. Int. Commun. Heat Mass Transfer 2001, 28 (2), 277–287. (25) Docquier, N.; Candel, S. Prog. Energy Combust. Sci. 2002, 28 (2), 107–150.
(26) Ji, Y.; Sohn, H. Y.; Jang, H. D.; Wan, B.; Ring, T. A. J. Am. Ceram. Soc. 2007, 90 (12), 3838–3845.
5339
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Figure 1. Schematic diagram of the experimental apparatus.
Engineering), and heating tape. The carrier gas is assumed to be saturated with the precursor. The saturator, which contains TEOS liquid, is maintained at 120 C. The vapor pressure of TEOS, which is used to calculate the flow rate, is taken from the following empirical equation in the National Institute for Standards and Technology (NIST) standard reference database:27 1561:277 Þ ð2Þ log10 ðPÞ ¼ 4:17312 - ð T - 67:572
Table 1. Material Properties for TEOS and SiO2 property
TEOSa
SiO2b
molecular weight (g/mol) boiling point (K) density (g/cm3) surface tension (J/m2)
208.33 442 0.936
60.08 3070 ( 75 2.16 0.3
a
Data taken from ref 11. b Data taken from ref 28.
speed of 10 ms. The time-averaged images were processed to determine the flame length. The flame residence time was calculated from the flame length and the first jet velocity, as defined in eq 1. For this study, the OH chemiluminescense signals were integrated along the line of sight, because this adequately provided for the determination of flame shape and flame length. Characterization of the silica particles was initially performed using an ex situ method. A transmission electron microscopy (TEM) system (Philips, Model Tecnal F20 with a magnification of 115000) was used to ascertain the mean primary particle diameter and the morphology. Silica particle samples were collected using a thermophoretic sampling device that rapidly moves a TEM grid (EMS, Carbon Only, cu 300 mesh, 3 mm diameter) to a precisely defined position (at a height above burner (HAB) distance of 30 cm). A carbon-coated copper TEM grid, mounted on the tweezers, was connected to a pneumatic cylinder (TPC, Model TANHAY-SMC, 9 cm stroke) that was driven by a solenoid valve (TPC, Model TVF3130). The solenoid used a relative pressure of 0.1 MPa and was equipped with a timer (OMRON, Model H3CR).29 For each experimental condition, a typical sampling of 200-350 primary particles was counted with the image processing technique and then evaluated statistically. Following the initial ex situ characterization of the particles, an in situ method was applied. The technique involved a SMPS spectrometer that was equipped with a vacuum pump (ULVAC, Model DA-15D) and a similar dilution (N2, 10 L/min) sampling
where P is the vapor pressure (expressed in units of bar) and T is the liquid temperature (given in degrees Kelvin), which has values in the range of 289-441.6 K. The physical properties of TEOS and SiO2 are shown Table 1. Visible flame photographs are widely used to observe the approximate flame geometry. Direct flame images were taken using a camera (Nikon, Model D70s) that was equipped with a lens (AF Nikkor, 50 mm, F1.4). These images are acquired under the conditions of F16 with an exposure time of 100 ms. The detection of chemiluminescence has often been used in previous research to describe various aspects of the combustion process qualitatively.21-25 The chemiluminescence intensities are roughly proportional to the reaction intensity. The OH chemiluminescence is free from background luminosity because it relaxes to the ground state in the UV range, with the most intense bands being the (0,0) at 306.4 nm and the (1,0) at 281.l nm.22,23 Therefore, the OH chemiluminescence can be properly used as a marker for deciding the flame-tip behavior. However, as stated in the Introduction, it is impossible to get quantitative information on the concentration of OH in the present study. The OH chemiluminescence signals were collected using an ICCD camera (Princeton Instrument, Model PI-MAX2:512RB) with 512 512 pixels. This camera was equipped with UG11 and WG305 filters through a UV Nikkor (F4.5, FL = 105 mm) camera lens. For each measurement, the time-averaged OH chemiluminescence image consisted of 300 images obtained under the condition of F4.5 with a shutter
(28) Grayson, M. Encyclopedia of Glass, Ceramics and Cement; John Wiley & Sons: New York, 1985. (29) Oh, K. C.; Shin, H. D. Fuel 2006, 85, 615–624.
(27) Stull, D. R. Ind. Eng. Chem. 1947, 39, 517–540.
5340
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
methane flames with oxygen enrichment. The size of the computation domain is 300 mm in length and 100 mm in radius. The grid setup of the length and radius is 520 164. These dimensions apply to the finer grids placed near the nozzle of the burner to resolve the complex flow patterns in the burner. The time-averaged conservation equations for mass, momentum, energy, mean mixture fraction, and variance were calculated using a finite control volume approach. This was accomplished using an axisymmetric 2D solver in the Fluent32 software without including the precursor, because of the lack of data for TEOS. However, the jet velocity remained unchanged as the precursor was replaced with methane. For turbulent modeling, a standard k-ε turbulent model was used. To model the reaction, the beta probability density function (β-PDF) with the equilibrium chemistry model was employed. The discrete ordinate (DO) method for radiative heat transfer has been adapted. The fixed velocity and mass fraction boundary conditions for each species are applied at the inlets, whereas the pressure boundary conditions are specified at the outlet and far-field walls. At the walls, no heat-flux boundary conditions were applied. A second-order discretization scheme was used to solve all governing equations. The equilibrium model assumes that the chemistry is rapid enough for chemical equilibrium. This model, which is based on the minimization of Gibbs free energy, is used to compute species mole fractions and adiabatic temperature from the mixture fraction. The PDF model does not require a detailed chemical reaction mechanism to obtain reasonable results that involve very high temperature (e.g., combustion systems using preheat, oxygen enrichment, and pure oxygen). However, it is necessary to consider the influence of thermal dissociation and radical species in those cases. Kim et al.33 used the chemical mechanism of GRI 3.0 to investigate air-fuel (CH4) and oxyfuel (CH4) flames, while another group of researchers34 showed the effects of thermal dissociation on oxy-fuel (C3H8) and air-fuel (C3H8) combustion using the EQUIL code in the CHEMKIN packages. Brink et al.35 applied the presumed PDF for the reaction model to oxy-natural gas combustion and compared it with that for the EDC model. In this work, to investigate the influence of thermal dissociation, radical species, and the adiabatic temperature were calculated for 10 cases using the CHEMKIN EQUIL code with GRI 3.0. The species applied for thermodynamic equilibrium calculations at a stoichiometric equivalent ratio in the combustion of oxygen-enriched and pure-oxygen methane are listed in Table 3. Species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalent ratio using the set that consisted of 53 species. Figure 2 shows the resulting adiabatic temperatures that have been obtained according to oxygen concentration from thermodynamic equilibrium calculations, considering chemical species in the combustion of oxygen-enriched and pureoxygen methane. In the combustion of air methane, the case of 5 species for equilibrium chemical reaction calculation leads to reasonable simulation results. However, in the
Table 2. Experimental Conditions for the Synthesis of Silica Nanoparticles in Turbulent Non-premixed Flames with Oxygen Enrichment flow rate velocity of of methane Q_ CH4 first jet, V1 (m/s) (cm3/min)
flow rate flow rate oxygen of of fraction oxygen, nitrogen, Q_ N2 Q_ O2 in second jet, XO2 (cm3/min) (cm3/min)
total flow rate of oxidant, Q_ oxidant (cm3/min)
40 40 40 40
1231 1231 1231 1231
0.5 0.6 0.8 1.0
10000 12000 16000 20000
10000 8000 4000 0
20000 20000 20000 20000
50 50 50 50
1702 1702 1702 1702
0.5 0.6 0.8 1.0
10000 12000 16000 20000
10000 8000 4000 0
20000 20000 20000 20000
60 60 60 60
2174 2174 2174 2174
0.5 0.6 0.8 1.0
10000 12000 16000 20000
10000 8000 4000 0
20000 20000 20000 20000
70 70 70 70
2645 2645 2645 2645
0.5 0.6 0.8 1.0
10000 12000 16000 20000
10000 8000 4000 0
20000 20000 20000 20000
probe, which was developed and tested by Ahn et al.30 and placed 30 cm above the burner. Ahn et al.30 compared the geometrical standard deviation obtained from SMPS with the TEM results. From these results, they showed a confidence on the dilution sampling system. In their experiment, the dilution sampling probe performs well enough to freeze the particle-size changes after the sampling and the probe inserted in the flame does not significantly alter the flame characteristics. In the evaluation of particle size distributions, the SMPS spectrometer is the most widely used instrument and is considered to be the standard against which all other techniques are judged.30,31 As such, the SMPS spectrometer was utilized in this study to ascertain the particle number concentration distributions as well as the geometric particle diameter and its standard deviation. The SMPS spectrometer used during the course of these experiments (TSI, Model 3936) incorporated a traditional long classification column (Model 3081) and a TSI Model 3776 condensation particle counter (CPC). The operating parameters used (unless stated otherwise) were as follows: aerosol flow rate, 0.3 L/min; sheath flow rate, 3.0 L/min; CPC inlet flow rate, 0.3 L/min; CPC sample flow rate, 0.05 L/min (for the size range of 14.3-673.2 nm); and scan time, 120 s. The minimum time required between consecutive scans is 30 s. 2.2. Experimental Conditions. The experimental conditions for the synthesis of silica nanoparticles are shown in Table 2. A total of 16 experimental conditions are given. For all experiments, the flow rate of the precursor (TEOS) and the carrier gas (N2) were fixed at 154 cm3/min and 500 cm3/min, respectively. This TEOS feed rate is equivalent to 4.74 10-3 mol/min if the N2 carrier gas is saturated at the exit of the saturator. The velocity of the first jet (which is a mixture of TEOS, N2, and CH4) ranged from 40 m/s to 70 m/s, whereas the velocity of the second jet (which is a mixture of O2 and N2 with a fixed flow rate of 20 L/min at 25 C and 1 atm) was fixed at 1.07 m/s.
3. Numerical Evaluation For this work, two-dimensional (2D) axisymmetric system meshes were constructed for the turbulent non-premixed
(32) Fluent V6 User’s Guide, December 2001. (33) Kim, H. K.; Kim, Y. M.; Lee, S. M.; Ahn, K. Y. Energy Fuels 2007, 21, 1459–1436. (34) Kim, H. J.; Choi, W.; Bae, S. H.; Shin, H. D. Energy Fuels 2009, 23, 1456–1463. (35) Brink, A.; Hupa, M.; Breussin, F.; Lallemant, N.; Weber, R. J. Propul. Power 2000, 16 (4), 609–614.
(30) Ahn, K. H.; Jung, C. H.; Choi, M.; Lee, J. S. J. Nanopart. Res. 2001, 3, 161–170. (31) Zhao, B.; Uchikawa, K.; McCormick, J. R.; Ni, C. Y.; Chen, J. G.; Wang, H. Proc. Combust. Inst. 2005, 30, 2569–2576.
5341
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
combustion of oxygen methane, the adiabatic temperature of the case of 5 species at stoichiometry reaches a value of >5000 K, because the thermal dissociation effect is not considered at high temperatures. This value decreases to 3384 K when 7 species are considered. Furthermore, this maximum temperature reaches 3052 K when 10 species are considered. This temperature is similar to that calculated in the case of 53 species. Therefore, it is necessary to consider a minimum of 10 species for the chemical equilibrium calculation in the combustion of oxygen-enriched and pure-oxygen methane. Figure 3 shows the resulting adiabatic temperatures as a function of mixture fraction obtained from thermodynamic equilibrium calculations for considering chemical species (as shown in Table 4) in the combustion of pure-oxygen methane. The reason for the overestimation of the adiabatic temperature in the case of 4 species (which is not included the influence of thermal dissociation) is shown. The temperature in the case of 9 species is similar to that calculated in the case of 15 species. Therefore, it is desirable to consider a minimum of 9 species for the chemical equilibrium calculation in the combustion of pure-oxygen methane. Consequently, considering CH4, O2, CO2, H2O, N2, CO, H2, H, O, and OH can produce realistic temperature predictions in the combustion of oxygen-enriched and pure-oxygen methane, and these results can be used in the β-PDF combustion model.
most common definition of the flame length is the distance between the burner exit plane and the visual flame tip. This tip is located at the farthest downstream point at which flaming gas is seen with an appreciable frequency. Often, the tip is measured by visual observation of the flame photographs.18 A definition based on the visible emission gives more or less reliable results, depending on the experimental conditions. Such conditions include the fuel and oxidant used, the surrounding environment, and the actual observers. However, the spontaneous emission of OH is free from background luminosity because, as stated earlier, it is located in the UV range.22,23 It can be properly used as a reaction zone marker for deciding the flame tip behavior, because it clearly ceases outside the reaction zone. Therefore, we have applied a method that relies on the spontaneous emission of OH. The flame length is defined as the distance from the burner exit plane to the axial position where the OH chemiluminescence intensity is 95%. This definition is reliable, regardless of the flame and the surrounding conditions. Figure 4 shows representative visible flame images of the coaxial turbulent non-premixed jet flames with oxygen enrichment. Parameters of these flames include a constant TEOS/N2 flow rate of 654 cm3/min, oxidant flow rates of 20 L/min (oxygen concentration of 50%), and a methane flow rate that varied from 289 cm3/min to 3116 cm3/min. The visible flame images were acquired under the condition of F16, with an exposure time of 100 ms. As shown in Figure 4, the flame took on a strong, luminous white-yellow appearance when TEOS was added. This can be explained by the formation and oxidation of carbon from the combustion of both methane and TEOS hydrocarbons. This is accompa-
4. Results 4.1. Flame Characteristics. The flame length is one parameter that can be difficult to define, especially when the flame tip is randomly fluctuating. Nevertheless, it is considered to be one of the representative flame characteristics. The Table 3. Species Applied for Equilibrium Calculations case
species considered
5 species 7 species 10 species 53 species
CH4, O2, CO2, H2O, N2 CH4, O2, CO2, H2O, N2, CO, H2 CH4, O2, CO2, H2O, N2, CO, H2, H, O, OH H2, H, O, O2, OH, H2O, HO2, H2O2, C, CH, CH2, CH2(S), CH3, CH4, CO, CO2, HCO,CH2O, CH2OH, CH3O, CH3OH, C2H, C2H2, C2H3, C2H4, C2H5, C2H6, HCCO, CH2CO, HCCOH, N, NH, NH2, NH3, NNH, NO, NO2, N2O, HNO, CN, HCN, H2CN, HCNN, HCNO, HOCN, HNCO, NCO, N2, AR, C3H7, C3H8, CH2CHO, CH3CHO
remarks a a a a
a species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0.
Figure 3. Adiabatic temperature as a function of mixture fraction obtained from equilibrium calculations for considering chemical species in the combustion of pure-oxygen methane.
Figure 2. Adiabatic temperature according to oxygen concentration from thermodynamic equilibrium calculation, considering chemical species at a stoichimetric equivalent ratio.
5342
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Table 4. Species Applied for Equilibrium Calculation in the Combustion of Pure Oxygen Methane cases
species considered
4 species
CH4, O2, CO2, H2O
5 species
CH4, O2, CO2, H2O, CO
6 species
CH4, O2, CO2, H2O, CO, H2
7 species
CH4, O2, CO2, H2O, CO, H2, OH
9 species
CH4, O2, CO2, H2O, CO, H2, OH, O, H
15 species
CH4, O2, CO2, H2O, CO, H2, OH, O, H, CHO, H2O2, HCO, HO2, HOCO, O3
remarks species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0 species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0 species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0 species that have a high concentration were selected from the result of the calculation for a stoichiometric equiva lence ratio using EQUIL code with GRI 3.0 species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0 species that have a high concentration were selected from the result of the calculation for a stoichiometric equivalence ratio using EQUIL code with GRI 3.0
Figure 4. Representative flame photographs for the oxygen-enriched flames at a constant TEOS/N2 flow rate of 654 cm3/min (from left to right, the first jet velocity varies over a range of 2080 m/s; the oxygen concentration is 50%). Figure 6. Flame length and residence time upon varying the first jet velocity over a range of 40-70 m/s at oxygen concentrations of (9) 50%, (b) 60%, (2) 80%, and (1) 100%.
conditions as described for Figure 4. The OH chemiluminescence images are an accumulation of 300 images, obtained under the condition of F4.5, with an exposure time of 100 ms. As the first jet velocity is increased from 20 m/s to 40 m/s, the flame length increases rapidly, and then is almost constant over the first jet velocity of 40 m/s. Qualitatively, the flame appears longer in the images shown in Figure 4 than in the OH chemiluminescence images of Figure 5. This is due to the presence of soot radiation that surrounds and extends above the reaction zone. Consequently, the visible flame length somewhat overestimates the length of the reaction zone. Because of this scenario, the flame length was determined by the OH chemiluminescence images for each condition (see Table 2). Figure 6 shows the flame length and residence time given by eq 1, as a function of first jet velocity, which varied over a range of 40-70 m/s. The oxygen concentrations are 50% (represented by rectangles (9) in the figure), 60% (represented by circles (b) in the figure), 80% (represented by triangles (2) in the figure), and 100% (represented by inverted triangles (1) in the figure). These results were
Figure 5. OH chemiluminescence images for the oxygen-enriched flames at a constant TEOS/N2 flow rate of 654 cm3/min (from left to right, the first jet velocity varies over a range of 20-80 m/s; oxygen concentration is 50%).
nied by the formation of silica. As the first jet velocity increases to 40 m/s, the flame length increases and then is almost constant. The flame width also broadens with increasing first jet velocity. Figure 5 shows representative OH chemiluminescence images of the coaxial turbulent nonpremixed jet flames with oxygen enrichment under the same 5343
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Figure 8. Flame temperature (top), velocity (middle), and turbulent intensity (bottom) along the center line of flames according to the height above burner (HAB) for oxygen concentrations of 50%, 60%, 80%, and 100% and a first jet velocity of 40 m/s.
Figure 7. Flame temperature (top), velocity (middle), and turbulent intensity (bottom) along the center line of flames according to the height above burner (HAB) for first jet velocities of 40, 50, 60, and 70 m/s and an oxygen concentration of 50%.
70 m/s, the flame length remained virtually unchanged, according to oxygen concentration. Under all oxygen concentration conditions, the flame residence time decreased as the first jet velocity varied from 40 m/s to 70 m/s. At a first jet velocity of 40 m/s, the residence time is ∼4.3 ms at oxygen concentrations of 80%-100% and ∼4.0 ms at oxygen concentrations of 50%-60%. When the first jet velocity was 70 m/s, the residence time was almost constant at 2 ms for all oxygen concentrations. Figure 7 shows the flame temperature, velocity, and turbulent intensity along the center line of flames according to HAB for first jet velocities of 40, 50, 60, and 70 m/s at an oxygen concentration of 50%. The flame temperature increased rapidly as the HAB increased, reaching a maximum temperature of ∼2640 K. The flame temperature then slowly decreased as the HAB increased. As the first jet velocity was
obtained from the OH chemilumincesce images attained under the experimental conditions (see Table 2). For all oxygen concentration conditions, the flame length decreased slightly as the first jet velocity increased from 40 m/s to 70 m/s. At a first jet velocity of 40 m/s, the flame length underwent a slight increase as the oxygen concentration increased from 50% to 100%. The following trends in flame length were observed when the first jet velocity was varied from 40 m/s to 70 m/s: flame length was almost constant at 15 cm when O2 = 50% (represented by rectangles (9)), flame length decreased from 16.1 cm to 14.3 cm at O2 = 60% (represented by circles (b)), flame length decreased from 17.4 cm to 14 cm at O2 = 80% (represented by triangles (2)), and flame length decreased from 17.5 cm to 14.5 cm at O2 = 100% (represented by inverted triangles (1)). When the first jet velocity was 5344
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Figure 9. TEM images for the oxygen-enriched flames with TEOS (from left to right, the oxygen concentration was varied from 50% to 60%, 80%, and then 100%; from top to bottom, the first jet velocity was varied from 70 m/s to 60 m/s, 50 m/s, and then 40 m/s).
position of the maximum flame temperature moved slightly toward the nozzle. However, velocity and turbulent intensity were almost constant for all oxygen concentrations. 4.2. Characteristics of Particles. TEM images of the silica nanoparticles are shown in Figure 9. These particles were collected by thermophoretic sampling at HAB = 30 cm with an oxygen concentration that varied over a range of 50%100% and a first jet velocity that varied over a range of 4070 m/s. The particles are observed to be aggregated and partially sintered. The average primary particle diameters are obtained by measuring each primary particle diameter of agglomerators. Increasing the first jet velocity from 40 m/s to 70 m/s resulted in smaller primary particle diameters. An increase in the oxygen concentration from 50% to 100% led to the formation of smaller primary particle diameters at higher oxygen concentrations. As shown in Figures 7 and 8, a higher oxygen concentration subsequently increased the flame temperature at the same oxidant flow rate. Also, a higher first jet velocity at the same oxygen concentration increases the turbulent intensity in the flame. Figure 10 shows the influence of the oxygen concentration on the average primary particle diameter for first jet velocities of 40 m/s (represented as rectangles (9) in the figure), 50 (represented as circles (b) in the figure), 60 (represented as triangles (2) in the figure), and 70 m/s (represented as inverted triangles (1) in the figure). Increasing the oxygen concentration from 50% to 100% led to the following decreases in the average primary particle diameter: from 14.9 nm to 9.5 nm at a first jet velocity of 40 m/s (represented as rectangles (9)), from 14.1 nm to 9.2 nm at 50 m/s (represented as circles (b)), from 12.0 nm to 8.9 nm at 60 m/s (represented as triangles (2)), and from 11.2 nm to 9.0 nm at 70 m/s (represented as inverted triangles (1)). At an
Figure 10. Primary particle diameter at oxygen concentrations varying over a range of 50%-100% for first jet velocities of (9) 40, (b) 50, (2) 60, and (1) 70 m/s.
increased, the position of the maximum flame temperature moved slightly downstream. However, the maximum flame temperature was almost constant for all first jet velocities. As the first jet velocity was increased, the velocity and turbulent intensity was increased. Figure 8 shows the flame temperature, velocity, and turbulent intensity, according to HAB, for a first jet velocity of 40 m/s and oxygen concentrations of 50%, 60%, 80%, and 100%. The flame temperature increases rapidly as the HAB increases, reaching a maximum temperature of ∼2640 K at [O2] = 50%, 2730 K at [O2] = 60%, 2830 K at [O2] = 80%, and 2900 K at [O2] = 100%. The temperature then slowly decreased as the HAB increased. As the oxygen concentration was increased, the 5345
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Figure 11. Evolution of primary particle size distributions (PPSDs) for oxygen concentrations of (9) 50%, (b) 60%, (2) 80%, and (1) 100% at first jet velocities of (a) 40, (b) 50, (c) 60, and (d) 70 m/s.
oxygen concentration of 100%, increasing the first jet velocity from 40 m/s to 70 m/s resulted in minimal change in the average primary particle diameter. The standard deviation in the primary particle diameters exhibited the following decreases as the oxygen concentration increased from 50% to 100%: from 5.5 to 3.8 for 40 m/s (represented as rectangles (9)), from 5.5 to 2.8 for 50 m/s (represented as circles (b)), from 5.0 to 2.3 for 60 m/s (represented as triangles (2)), and from 3.6 to 2.2 for 70 m/s (represented as inverted triangles (1)). Figure 11 shows the evolution of the primary particle size distributions (PPSD) obtained from the TEM images (shown in Figure 9) at oxygen concentrations of 50% (represented as rectangles (9) in Figure 11), 60% (represented as circles (b) in Figure 11), 80% (represented as triangles (2) in Figure 11), and 100% (represented as inverted triangles (1) in Figure 11) at first jet velocities of (a) 40 m/s (Figure 11a), 50 m/s (Figure 11b), 60 m/s (Figure 11c), and (d) 70 m/s (Figure 11d). This fraction is obtained from the ratio of the number of primary particles per unit size interval to the total number of primary particles. As shown in Figure 11, the fraction of primary particle diameters is shifted to smaller sizes with increasing oxygen concentration at all first jet velocities. As the first jet velocity was increased from 40 m/s (Figure 11a) to 70 m/s (Figure 11d), the fraction of primary particle diameters in the range of 0-10 nm increased for oxygen concentrations of 50% (rectangles (9)), 60% (circles (b)), and 80% (triangles (2)). However, as the first jet velocity increased, the fraction did not change significantly at an oxygen concentration of 100% (inverted triangles (1)). Figure 12 shows the particle number concentrations as a function of particle size for first jet velocities of 40 m/s (represented as rectangles (9) in the figure), 50 m/s (represented as circles (b) in the figure), 60 m/s (represented as triangles (2) in the figure), and 70 m/s (represented as inverted triangles (1) in the figure) at oxygen concentrations
of 50% (Figure 12a), 60% (Figure 12b), 80% (Figure 12c), and 100% (Figure 12d). At each oxygen concentration, the mode of the particle number concentrations is shifted to smaller sizes as the first jet velocity increases. The mode of the particle number concentrations decreased as follows: from 102 nm to 71 nm at O2 = 50% (Figure 12a), from 98 nm to 71 nm at O2 = 60% (Figure 12b), from 109 nm to 69 nm at O2 = 80% (Figure 12c), and from 85 nm to 57 nm at O2 = 100% (Figure 12d). The effect of the first jet velocity on the particle number concentrations is more significant than the impact of oxygen concentration. Figure 13 shows the geometric mean diameter, with respect to the first jet velocity, for oxygen concentrations of 50% (represented as rectangles (9) in the figure), 60% (represented as circles (b) in the figure), 80% (represented as triangles (2) in the figure), and 100% (represented as inverted triangles (1) in the figure). As the first jet velocity increased, the geometric mean diameter of the particles roughly decreased for all oxygen concentrations. As the first jet velocity increased from 40 m/s to 70 m/s, the geometric mean diameter decreased steadily from 104 nm to 70 nm for an oxygen concentration of 50% (rectangles (9)), from 97 nm to 69 nm for an oxygen concentration of 60% (circles (b)), from 98 nm to 67 nm for an oxygen concentration of 80% (triangles (2)), and from 90 nm to 59 nm for an oxygen concentration of 100% (inverted triangles (1)). At all oxygen concentrations, smaller flame residence times with increasing first jet velocity allows less time for particle growth. This results in the formation of smaller particles. Moreover, the higher oxygen concentration leads to smaller particles at each first jet velocity. As the first jet velocity increases from 40 m/s to 70 m/s, the geometric standard deviation (which describes the width of the geometric particle diameter distribution) is almost constant, at 1.6, or even deviated slightly. In this region, the geometric standard 5346
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
Figure 12. Particle number concentrations with particle size for first jet velocities of (9) 40, (b) 50, (2) 60, and (1) 70 m/s at oxygen concentrations of (a) 50%, (b) 60%, (c) 80%, and (d) 100%.
Figure 13. Geometric mean diameter according to the first jet velocity for oxygen concentrations of (9) 50%, (b) 60%, (2) 80%, and (1) 100%.
Figure 14. Geometric particle diameter scaled by the flame residence time with global strain rate. Oxygen concentrations are (9) 50%, (b) 60%, (2) 80%, and (1) 100%.
deviation attains the so-called “self-preserving distribution” (SPSD). This is consistent with the result of Yu et al.36 Figure 14 depicts the trend in which the ratio of the geometric particle diameter to flame residence time increased as the global strain (the ratio of the difference of between first jet velocity and second jet velocity to diameter of first nozzle) increased. Oxygen concentrations are 50% (represented as rectangles (9) in the figure), 60% (represented as circles (b) in the figure), 80% (represented as triangles (2) in the figure), and 100% (represented as inverted triangles (1) in the figure). The global strain rate was obtained from the ratio of the difference between the first jet velocity and second jet velocity to the first nozzle diameter. The geometric particle diameter was measured by SMPS (see Figure 13). The flame residence time was calculated from the flame
length (see Figure 6), which is obtained from OH chemiluminescence results and the first jet velocity. Although an increasing trend of geometric mean diameter scaled by flame residence time with global strain rate is investigated, the results at each oxygen enriched conditions do not collapse to a single line. This is the effect of increasing the flame temperature as the oxygen concentration in the oxidant stream increases (as shown in Figure 8). The shortened flame residence time and the higher turbulent intensity with increasing first jet velocity at all oxygen concentrations (as shown in Figures 6 and 7) allow less time for the particles to grow, resulting in smaller particles (see Figures 10 and 13). The higher flame temperature with increasing oxygen concentration at each first jet velocity (as shown in Figure 8, the flame temperature increases but the turbulent intensity is almost constant) results in an increase
(36) Yu, M.; Lin, J.; Chan, T. Powder Technol. 2007, 180, 9–20.
5347
Energy Fuels 2009, 23, 5338–5348
: DOI:10.1021/ef900492g
Bae and Shin
in the chemical reaction rate and a decrease in the surface tension of silica nanoparticles. This results in an increase in the nucleation rate.12 Therefore, smaller silica nanoparticles can be obtained (as shown in Figures 10 and 13).
the first jet velocity had a tendency to decrease as the first jet velocity increased at all oxygen concentrations. (3) The higher flame temperature generally increased the reaction rate but decreased the surface tension. This resulted in larger aerosol concentrations and smaller particle sizes. Therefore, the averaged primary particle diameter and its standard deviation decreased as the flame temperature increased at the same residence time. Increasing the first jet velocity at each oxygen concentration enhanced the turbulent intensity and shortened the flame residence time at high temperature. This led to smaller primary particles and standard deviations. However, the geometric mean particle diameter decreased, but its standard deviation did not significantly change with decreases in the residence time. (4) In particular, the result showed that an increasing trend of geometric mean diameter scaled by residence time with global strain rate was investigated.
5. Conclusions The characteristics of a turbulent non-premixed flame with oxygen enrichment in a co-flowing jet burner for synthesizing silica nanoparticles were investigated numerically and experimentally, and the characteristics of generating silica nanoparticles were also experimentally investigated. The following results were obtained: (1) The adiabatic temperatures were obtained from considering sets of the most important species, i.e., CH4, O2, CO2, H2O, N2, CO, H2, OH, O, and H in the combustion of oxygen enriched methane. The flame temperature could be calculated using the β-PDF combustion model, and reasonable results were obtained for the combustion of oxygen enriched methane. It is also expected that these results will be very useful in investigating the flame characteristics to an industrial oxy-fuel combustor. (2) The flame residence time obtained from the flame height, measured from OH chemiluminescence results, and
Acknowledgment. This research was supported by the Combustion Engineering Research Center (CERC) of the Korea Advanced Institute of Science and Technology (KAIST) in Korea.
5348