Design and characterization of a linear Hencken-type burner

Growing concerns regarding energy security and climate change have prompted increased interest in combustion efficiency, chemistry, and emissions.^1–3 There are also long-standing concerns related to the negative impact of combustion emissions on air quality and human health.^4–6 Soot-formation chemistry and its impact on soot emissions are of particular interest because (1) soot particles are linked to a wide range of adverse health effects^4–6 and (2) soot emitted into the atmosphere absorbs solar radiation over a broad wavelength range, leading to enhanced global warming.^6–8 To gain insight into soot formation and particulate chemistry-related combustion processes, it is essential to design experiments involving controlled and well-characterized flames with the goal of probing important reactions, species, and thermodynamic properties.^3,9–11 Such efforts have recently resulted in increased engine combustion efficiency,¹² new renewable fuel blends,⁹ and changes in public policy.⁵ 

Two important considerations for experimental studies of flame chemistry and soot formation are (1) the selection of sensitive combustion diagnostics that elucidate key aspects of the chemical system under study and (2) the use of accessible flame geometries that exhibit relevant combustion attributes, enable the use of appropriate combustion diagnostics, and facilitate computational simulations. In terms of diagnostics, non-intrusive in situ measurement techniques are advantageous because they allow the combustion environment to be probed with minimal disturbance. Hence, temporally and spatially resolved laser diagnostics are commonly employed for in situ investigations of flames. When using such techniques, it is desirable that these laser diagnostics be independent of flame geometry and environment.¹³ Occasionally, however, flame geometries need to be tailored to the application of a particular diagnostic approach. For example, particle extinction or gas-phase species absorption is generally used in a path-integrated, line-of-sight configuration. Measurements of species with low densities or small extinction or absorption cross sections may require enhanced path lengths within the flame, which is usually accomplished by multi-passing the beam through the flame or by increasing the flame dimensions.^10,14–19 If flame dimensions are increased, care is required to minimize flame inhomogeneities over the extinction or absorption path length.

The flames studied in combustion laboratories can be roughly divided into three general geometric classes. Three-dimensional flames are those in which conditions are inhomogeneous in X-Y-Z space and are also often variable in time. Examples include turbulent swirl flames or pool fires. Although these flames most closely approximate practical combustion environments, such as jet or diesel engines or industrial boilers, their spatially inhomogeneous and time-varying nature makes them difficult to study without spatially resolved and temporally resolved laser-based techniques.^20,21 At the opposite extreme, one-dimensional flames, such as premixed flat flames, produced by McKenna burners or laminar diffusion flames generated by counter-flow burners exhibit changes in one dimension only (usually vertically). These flames provide excellent environments for the study of reaction kinetics by a wide variety of laser diagnostics, but their simplicity makes them far-removed from many practical combustion devices.²² Between these two ends of the spectrum are flames with two-dimensional geometries, such as cylindrical flames (R-Z environment) or linear flames (X-Z environment). These flames are often operated in regimes in which they are laminar and steady and thus provide a relatively simple environment in which diffusion, kinetics, and particle formation can be investigated. When using optical diagnostics, linear flames are especially advantageous because they provide a substantial interaction distance along their nearly homogeneous long axis.^23,24 This flame configuration simultaneously allows for imaging laser-induced signal or measuring laser-scatter signal with limited signal reabsorption from a detection angle perpendicular to, or off-axis from, the laser beam.^25–27 

In order to capitalize on the advantages of a linear flame geometry, we have developed a linear Hencken-type burner to be used for studies of soot formation, graphitization, and oxidation chemistry. This burner produces a 38-mm long (1.5-in.) quasi-two-dimensional laminar partially premixed diffusion flame. This paper provides details of the design and construction of this burner and presents characterization results for the flames that it produces.

Hencken-type burners consist of a hexagonal (hex, honeycomb) mesh with small-diameter tubes inserted through the mesh pores.^28–30 Fuel flows through the tubes while air flows through the honeycomb. Conventional arrangements involve a two-dimensional array of tubes in the center of the burner, surrounded by a co-flow of inert gas (usually nitrogen). The result is rapid mixing of the fuel and oxidizer streams downstream of the burner surface, yielding a planar reaction zone, surrounded by a non-reactive region. Hencken flames are often used as calibration reference sources in spectroscopic studies because of their ability to produce a steady stream of equilibrium products above the reaction zone.^28,29 Moreover, operating Hencken burners with sufficiently high gas flow rates results in a slight lifting of the reaction zone from the burner, yielding nearly adiabatic performance.²⁸ These burners can also function safely over a wide range of mixture fractions because they do not require a premixed gas blend.²⁸ Another important benefit of the Hencken honeycomb with embedded small-fuel-tube design is that these small passageways promote laminar flow by reducing the gas’s Reynolds number (⁠$Re=Udν$⁠, where U is the gas velocity, d is the channel diameter, and ν is the gas kinematic viscosity). These elements also result in a slight pressure drop from the inside of the burner to the burner surface, promoting more uniform flow over the entire area. For the above reasons, Hencken burners have become relatively commonplace in combustion studies.^28–35 

Whereas Hencken burners produce flat, uniform reaction zones, linear burners generate thin, long, vertical flame sheets. A common burner design for linear flame production is the slot burner (an example of which is a Wolfhard-Parker burner).³⁶ These devices often consist of three parallel rectangular shaped orifices in which fuel flows from a central rectangular tube with an air co-flow on both long sides and sometimes the short ends of the rectangular fuel tube.^25,37 In many cases, two curved screens (“gulls”) located above the reaction zone are used to stabilize the flame and prevent flickering.^25,37 Some Wolfhard-Parker burners also incorporate an inert gas co-flow in slots at the ends of the fuel tube; these are used to suppress combustion at the ends of the flame sheets.^24,37,38 When designed in this fashion, slot burners produce two thin diffusion flame fronts at the long edges of the fuel slot.^{15,16,23–25,27,36–46} Slot burners can also be configured for premixed flame operation in which case the co-flow of air and inert gases is often not used.^47–53 

The advantages of the linear flame/slot burner concept are numerous. Ideally, slot burners provide a roughly two-dimensional flame that is uniform along the long axis of the fuel slot (the Y axis). As pointed out by Wolfhard and Parker,³⁶ the slot design provides a broad reaction zone along the length of the fuel slot (compared to cylindrical flames), with good optical depth for absorption measurements or emission observations. This configuration presents the opportunity for a long path length for line-of-sight measurements, reduced signal reabsorption for imaging, and decreased interference for off-axis scattering measurements. Measurements can also be performed by orienting line-of-sight detection perpendicular to the long slots.^15,16 As pointed out by Smyth et al.⁴⁴ and Norton et al.,⁴⁵ slot burners facilitate precise lateral species profile measurements across the flame front parallel to the long burner axis, i.e., with the laser beam parallel to the long axis. Finally, the homogeneous environment along the long axis of linear flames simplifies data interpretation compared to studies of cylindrical diffusion flames. Cylindrical flames have inhomogeneous radial profiles in temperature, soot concentration, and other flame parameters and species, which complicate retrievals of spatial distributions of these parameters and species, particularly for line-of-sight measurements.^54,55 Attempts to account for variability in the radial profiles of co-annular flames generally involve the use of Abel inversions.^56,57 The linear flame/slot burner concept thus offers the advantages of a long interaction distance for optical diagnostics and simplified data interpretation.

The versatility offered by slot burners has led them to be used to make a variety of combustion measurements. Kent, Jander, and Wagner³⁷ performed light scattering and extinction measurements of soot in a flame from a Wolfhard-Parker burner, orienting the probe laser along the long axis of the burner and detecting the scattering perpendicular to this axis. Smyth et al.²⁵ performed light scattering and laser-induced fluorescence (LIF) measurements in a flame from a similar Wolfhard-Parker burner, also imaging perpendicular to the fuel slot axis. This group went on to use multiphoton ionization spectroscopy to measure methyl radicals, positioning the probe laser parallel to the long axis of the burner while inserting the tungsten detection electrode 1-2 mm above the beam.⁴⁰ Miller and Taylor⁴² also measured methyl radicals with this burner by employing mass spectrometry, positioning the quartz sampling probe in the center of the flame and orienting it parallel to the long axis of the burner in order to minimize pyrolysis of the sampled gases inside the probe; they achieved good agreement with the multiphoton ionization results.⁴⁰ Several studies were conducted using linear flame slot-type burners for atomic absorption spectroscopy measurements, taking advantage of the long path length to increase signals despite being restricted to low fuel flow rates by the required ultrasonic nebulizers.^47–49 Mercier et al.^15,16 measured OH and CH radical concentrations in a flame from a Wolfhard-Parker burner using Cavity Ring-Down (CRD) spectroscopy, making use of both the long axis (the Y axis, parallel to the fuel slot) and the transverse axis (the X axis, perpendicular to the slot). Shaddix et al.³⁸ measured laser-induced incandescence (LII), LIF, and soot thermal emission in inverse diffusion flames using a Wolfhard-Parker burner by flowing air in the central slot and fuel from the flanking rectangular apertures. Azzoni et al.⁵¹ used a Wolfhard-Parker design to produce premixed methane/air triple flames by introducing a premixed gas to all three of the burner’s slots. The authors probed these flames using laser Doppler velocimetry (velocity measurements) and laser interferometric holography (temperature measurements). In a rectangular Bunsen burner (premixed methane/air mixture), Echekki and Mungal⁵⁰ measured flame speeds using particle-tracking velocimetry, flame tip curvatures using direct flame photography along the long axis of the burner, and temperatures using Rayleigh scattering. Finally, Selle, Poinsot, and Ferret⁵³ also measured flame speeds using a premixed slot burner, correcting their results with cold-gas-velocity measurements. Their cold-gas velocity results showed that the fuel-velocity profile emerging from the fuel slot in their burner was not uniform, but rather was slightly lower at the ends of the slot (Azzoni et al.⁵¹ also detected velocity non-uniformity in their slot burner). The above studies demonstrate the advantages provided by the linear flame/slot burner configuration with respect to the ability to access the reaction zone from directions both parallel with, and perpendicular to, the long axis with a wide variety of in situ and ex situ techniques.

The linear Hencken burner described here is a hybrid of the traditional Hencken burner and slot burner designs. It combines the benefit of the conventional Hencken burner’s laminar flame with the advantage of the slot burner’s long-interaction distance geometry. The use of small-diameter fuel tubes for fuel gas delivery to the burner surface ensures that the area-averaged fuel gas velocity is approximately constant along the length of the flame, in contrast to the spatially dependent profile observed by others in slot burners.^51,53 Also, since the flame produced by the burner is short (<2 cm tall), no special efforts are required for flame stabilization, such as the curved screens sometimes used with Wolfhard-Parker burners.^25,37 In addition, the required fuel flow rates are relatively low (∼0.2 standard liters per minute) compared to conventional slot burners (1-10 standard liters per minute^25,37,38), reducing variability and down time associated with gas-bottle changes. The linear Hencken hybrid design described here yields a steady, partially premixed flame that produces reasonable conditions for studies of soot-formation chemistry and/or particle generation and growth.

A computer aided design (CAD) rendering of the assembled linear Hencken burner is shown in Fig. 1 (see also CAD drawings in the supplementary material). The overall dimensions of the burner are 71 mm long (excluding the 0.125-in. copper fuel/air inlet tubes and the water inlet/outlet tubes), 61 mm wide (excluding mounting hardware), and 28 mm tall. The primary burner pieces are fabricated from brass in order to increase heat conductivity and machinability, while the fuel tubes and hexagonal mesh are produced from stainless steel because of its chemical integrity and high melting point.

The burner consists of three brass pieces, sealed together using three Viton O-rings (a list of important components, together with relevant dimensions and supplier/part number information, is provided in Table I). Fuel enters the bottom plate and is delivered to the burner surface through a series of twenty-five 23-mm-long fuel tubes (#22 gauge); these tubes are sealed to the middle brass plate using room temperature vulcanizing silicone (RTV) in order to prevent mixing between fuel and air. Thus, a combustible mixture occurs only downstream of the outlets of the fuel tubes. In the immediate vicinity of the fuel tube outlets, some mixing of the fuel and air from the hexagonal mesh occurs, which results in partial premixing of the fuel and oxidizer.

Air enters the bottom plate on either side of the fuel inlet and is allowed to flow vertically through a series of 0.035-in.-diameter holes (#65 drill size) that distribute the flow evenly along the length of the burner (some technical design dimensions are provided using British Imperial units to improve clarity). The air is then forced through five 0.001-in. fine-wire stainless-steel screens that are sandwiched between 0.014-in. perforated-metal sheets. This sandwich assembly is clamped between the middle and top brass components and serves as a flow diffuser, promoting uniform air velocity across the entire burner surface. For the air-flow rates chosen for the flames discussed in this publication, five screens produced adequate flow diffusion. A greater number of screens should be used for higher air-flow rates. Finally, the air emerges through a half-inch-tall hexagonal mesh that is clamped within the top brass component with set screws and sealed about its perimeter with RTV (see Fig. 2 for important dimensions of the hexagonal mesh).

A line of 0.031-in. diameter holes in the sheet-screen assembly allows the fuel tubes to penetrate through the screen. The tips of the fuel tubes rise above the surface of the hexagonal mesh about 0.025 in. (0.64 mm) in order to facilitate alignment of the burner for optical diagnostics. In all experiments, the Height Above Burner (HAB) zero value is defined to be at the top of the fuel tubes.

Eighth-inch outer-diameter copper tubes, soldered to the brass parts, are used to deliver fuel and air to the burner, and cooling water to and from the burner. Cooling water enters the middle plate, passes through the burner, moves outside the burner in a U-shaped section (also made from 0.125-in. copper tubing), and then passes back through the opposite side of the burner. In order to achieve high tolerances for proper seals (e.g., between the sheet-screen assembly and the fuel tubes), all parts were cut using wire electrical discharge machining (EDM) and then milled further to specified dimensions. O-ring grooves were finished to 64-RMS and brass plate surfaces to 32-RMS.

The linear Hencken design described here does not incorporate inert gas flows at the ends of the line of fuel tubes.^24,38 This feature was not included because the co-flow of air through the honeycomb surrounding the individual fuel tubes prevents suppression of flame fronts at the burner ends. Additionally, having an air shroud completely encircling the line of fuel tubes rather than nitrogen/argon on the ends avoids concerns regarding inert gas entrainment from these end flows, as discussed by Kent et al.³⁷ The effects of flame fronts in slot burners are addressed by Miller et al.²³ and will not be discussed here. As pointed out by Selle et al.,⁵³ despite flame fronts on the burner ends, the central portion of the flame can be used for many diagnostics that employ perpendicular detection, such as laser scattering and LIF. Many slot-burner studies have been successfully performed using burners that do not have end-flame suppression.^{15,16,23,27,43–45}

The external-gas distribution and water-circulation systems employed to support this burner are important because they improve safety, flame uniformity, and burner stability. A diagram of the water- and gas-handling systems for this burner is provided in Fig. 3.

House air at 35-psig pressure is fed through a 0.5-μm filter (Swagelok model SS-4F-05) and delivered to a mass flow controller (MKS Instruments, Inc., model GM50A). The air-flow rate through the burner is fixed at 14.0 SLM (Standard (0 °C, 1 atm) Liters per Minute), yielding an exit velocity of approximately 22 cm/s at the burner surface. Representative flow rates for flame-gas components are provided in Table II. Henceforth, flames will be referred to by the descriptors used in this table (e.g., E1, H1, etc.). The flow rates for C₂H₄, C₂H₂, CH₄, H₂, CO₂, and N₂ correspond to flow rates of fuel and species mixed with the fuel, whereas air-flow rates refer to the co-flow.

For the measurements presented here, research-grade gases (C₂H₄, C₂H₂, CH₄, H₂, and CO₂) were supplied by Matheson Tri-Gas, Inc., and N₂ was taken from the boil-off of the house liquid nitrogen supply. The hydrocarbons, H₂, CO₂, and N₂ were directed through 0.5-μm filters and mass-flow controllers and allowed to combine in a mixing tube (McMaster-Carr Supply, Co. model 3529K51). The mixture then passed through a flashback arrestor (WITT-Gasetechnik GmbH & Co. KG model RF53N/H) and was directed to the burner. This flashback arrestor was included as an additional precaution; since no oxidizer was included in the fuel mixture, flashback was actually not possible under normal operation. All mass-flow controllers were calibrated (Sierra Instruments, Inc., model SL-500) prior to use and have an accuracy of better than ±3% at full scale. A supply pressure of at least 35 psig was provided on each gas line prior to the 0.5-μm filters. Measurements of the flowing-gas pressure in the copper-inlet tubes showed a slight pressure drop (1-2 psi), attributable to the tubes’ small 0.061-in. inner diameter. The small-diameter fuel tubes and diffuser screen/porous metal sheet assembly also contribute to this pressure reduction.

Cool distilled water is circulated through the burner to maintain the burner’s temperature in the range of 20–25 °C. A chiller (NESLAB Instruments, Inc., model RTE-111) provides a constant flow rate of ∼0.26 l/min of water at 20 °C. This water-flow rate ensures that the temperature increase of the water through the burner is less than 1 °C. The water-cooling capability of the burner was included because of the low gas-flow rates, which cause the flame to be stabilized close to the burner surface and hence operate in a non-adiabatic fashion. The surface temperature of the honeycomb during operation of Flame E1, as measured using a bead-type thermocouple (Omega Engineering, Inc., model 5SRTC-GG-K-30-72), is approximately 220 °C at a distance of about 2 mm from the fuel tubes and decreases rapidly to less than 100 °C by a distance of about 3 mm from the flame. These temperatures are likely inflated because of radiative-heat transfer from the flame to the thermocouple wires and therefore represent upper-bound values for the hexagonal mesh’s actual temperature.

Figure 4 shows a photograph of Flame E1. Since the flame is laminar and steady, the vertical centerline in the flame correlates roughly with the overall combustion history. The blue region seen at low heights corresponds primarily to emission from CH and C₂ radicals,^54,58 and the yellow/orange region at larger HABs results from soot luminescence.⁵⁹ Soot produced in the upper part of Flame E1 is highly graphitic in character, i.e., mature (see Section III E).

The radiant zone of Flame E1 is seen to extend approximately 10-11 mm above the burner and to be approximately 38 mm in length (however, similar to other linear flames,^37,39 the length is slightly shorter at higher heights). At low heights, the combustion area consists of two flame sheets, each ∼1 mm thick, which begin to combine about 2-3 mm above the burner to ultimately form a unified reaction zone that is about 3 mm thick. This flame’s small dimensions, while advantageous in some respects, such as reducing gas consumption, place constraints upon the choice of components for optical diagnostics. We summarize several previous measurements on slot burner flames here in order to provide a comparison for this linear Hencken-type burner. Smyth et al.^25,40,44,45 conducted laser absorption and laser Doppler velocimetry measurements using a beam focused to a spot size about 0.2 mm in diameter and directed parallel to a fuel slot with size 8 × 41 (W × L) mm²; each flame sheet produced by this burner had a thickness of approximately 2-3 mm. Kent et al.^37,39 used a 0.5-mm focused diameter laser beam in a similar configuration over a 8 × 50-mm² slot that produced flame fronts of 3-4-mm thickness. Also Mercier et al.^15,16 used laser beams with diameters of 1 mm and 0.3 mm to measure OH and CH concentration profiles, respectively, in 1-2-mm-thick flame sheets stabilized over a 4 × 30-mm² slot. The flame thicknesses listed above are similar to the overall thickness of the reaction zone in Flame E1, suggesting that beams focused to diameters of 1 mm or smaller will produce sufficient resolution for spectroscopic measurements.

We used a hot-wire anemometer^60,61 (TSI, Inc., model IFA-300, with probe model 1210-20) to extract relative flow-field velocities of the air emerging from the hexagonal mesh at a height of 2 mm above the mesh (with Flame E1 unlit), as shown in Fig. 5. This figure displays several profiles in the direction parallel to the length of the burner at several locations perpendicular to the length of the burner.

Across the surface of the burner, the measured air-flow velocity was homogenous with maximum deviations of ±10% from the average. Typical variations of approximately ±5% were observed along the length of the burner parallel to the line of fuel tubes (i.e., standard deviation of a single profile shown in Fig. 5). Variations of ±6% were observed for measurements performed perpendicular to the line of fuel tubes. The relative gas velocities measured for fuel emerging from the 25 fuel tubes were consistent to within ±7% of the average. If desired, the uniformity in the relative velocity above the hexagonal mesh can be further improved by using a greater number of wire screens in the diffuser assembly and at the fuel outlets by extending the small-diameter fuel tubes.

The characteristic time required for fuel/air mixing can be estimated according to Fickian binary diffusion because the flow near the burner is laminar. Using a binary diffusion coefficient D = 0.1655 cm²/s (obtained from Elliott and Watts⁶² for ethylene in air) and a diffusion length of L = 0.889 mm (the center-to-center separation distance for two adjacent hexagonal flow channels according to Fig. 2), this time is calculated to be τ ≅ 24 ms according to the following equation:

We can relate this value to the gas’s vertical displacement distance from the small-gauge fuel-tube outlets Z using the initial gas velocity U₀ together with a standard constant buoyant acceleration value of A = 25 cm/s² (see Santoro et al.⁶³) according to

This equation yields values for Z of 12-23 mm HAB for Flame E1, depending on whether the initial air-outlet velocity (U₀ ≅ 22 cm/s) or fuel-outlet velocity (U₀ ≅ 66 cm/s) is used; these distances agree well with the observed flame height for Flame E1.

We performed direct one-dimensional coherent anti-Stokes Raman spectroscopy (CARS) imaging to extract high-fidelity flame temperatures along the centerlines of Flames E1, H1, and H3. Here we provide only a short description of this diagnostic, as the complete details are available elsewhere.^64–66 CARS is a non-linear optical four-wave-mixing technique in which three incident photons with energies ω_pump, ω_Stokes, and ω_probe are mixed with the internal energy levels of the probed molecules to generate a fourth photon at energy ω_CARS according to Eq. (3),

The currently employed technique uses a crossed two-beam phase-matching scheme,⁶⁴ delivering the two photons, ω_pump and ω_Stokes, in a single broadband femtosecond pump/Stokes excitation pulse (45 fs), while the ω_probe photon is supplied from a spatially overlapped and time-synchronized narrowband picosecond probe pulse (70 ps). The inherent bandwidth of the 45 fs coherence driving pulse (>500 cm⁻¹) enables the interrogation of a Raman-active window covering the rotational transitions of most molecules with ∼0.3 cm⁻¹ spectral resolution. For the measurement geometry employed here, the two beams are formatted into vertically oriented sheets that are overlapped at the center of the flame giving one-dimensional spatially resolved temperature information in the overlap region. The spatial information is retrieved with a line-spread function of <40 μm. Because the flame was taller than the laser sheets, we translated the flame vertically in several increments to obtain a full height profile.

CARS thermometry in air-fed flames has mainly been achieved on N₂ because of its high abundance and inert properties, enabling CARS measurements across a wide range of operational conditions. The extreme sensitivity of the CARS spectrum to the relative population of molecular rovibrational states, which follows a Boltzmann thermal distribution, allows for benchmark thermometry measurements of high accuracy and sensitivity. The thermometry is based on a standard contour spectral fitting procedure, here performed with integrated line-intensities, where the experimental spectrum is compared with a library of pre-calculated theoretical spectra at different temperatures of the prevailing conditions. We implemented the routine with a non-linear interpolating algorithm that minimized the sum of squares of the residuals between the experimental and theoretical spectra.

Soot luminosity imposed a strong broadband background on the CARS spectra, posing a challenge for the imaging CARS measurements in Flame E1. This soot luminosity was collected within the solid angle of the imaging optics used to collect the CARS image and could not be rejected simply by moving the detector away from the source or by spatially filtering the CARS beam using a pinhole, as would be the common approach for a point-measurement CARS technique. Instead, we introduced a spinning wheel slit, synchronized with the 20-Hz CARS signal acquisition rate, at the position of the entrance plane of the imaging spectrometer (Horiba, Ltd. model iHR550). This method rejected ∼90% of the interfering background and substantially improved the signal-to-background ratio for the measurement. We ultimately achieved high-throughput single-shot CARS signals (recorded at ∼1000 counts across the ∼2 to 3 mm central part of the ∼5 to 6 mm field-of-view), leading to robust spectral fits. An example of a one-dimensional spatially correlated CARS signal is displayed in Fig. 6(a), which was recorded along the center of the long axis of Flame H1.

The resulting high-fidelity extracted temperature profiles for Flames E1, H1, and H3 are shown in Fig. 7. Each of the profiles was assembled from 20 measurement segments in steps of ∼2 mm with 500 single-shot measurements averaged at each step. This approach provides robustness in the estimated ∼2% to 3% level of accuracy across the entire spatial domain of the current measurements.

The gas temperature in all three flames increases to a maximum with increasing HAB because of increasing heat release and then declines as air from the co-flow region mixes with the reaction products at higher HABs. The maximum ethylene temperature (2038 K) is comparable to that measured in cylindrical co-flow ethylene/air diffusion flames by others.^67–70 As expected, the hydrogen flames, which are diluted in nitrogen (see Table II), are colder than the undiluted ethylene flame. (Note that undiluted hydrogen flames can also be stabilized on this burner; their temperatures are significantly hotter and a faint red/orange glow, attributable to water emission,⁷¹ becomes visible at the outlets of the fuel tubes.) Tabulated temperature values for these flames are given in the supplementary material in Table S1.

Figure 8 provides a graphical comparison of the temperature profile of Flame E1 with an end-view photograph of the luminous reaction zone. The maximum centerline flame temperature occurs near the flame tip. Measurements in ethylene/air cylindrical co-flow diffusion flames have shown that the location of this maximum temperature approximately coincides with the locations of maxima in the mole fractions of water and carbon dioxide as well.⁶⁹ 

We conducted laser-induced incandescence (LII) and laser extinction measurements to infer soot volume fractions as a function of the HAB along the centerline of Flame E1. The laser setup for these experiments is detailed elsewhere,^72,73 and we will only include the most relevant information here. For the LII measurements, soot particles in the flame were irradiated using the fundamental (1064 nm) output from an injection-seeded Nd:YAG laser (Spectra-Physics, Inc., Quanta-Ray Lab Series) with a repetition rate of 10 Hz and pulse duration (FWHM) of approximately 10 ns. We optimized the spatial and spatio-temporal profiles of the laser beam using a 2 × 2-mm² square ceramic aperture positioned at the exit of the laser to select the central portion of the beam, and relay-imaged the aperture to the detection region of the flame using a 2:1 reducing telescope.⁷⁴ The laser beam was oriented parallel to the line of small-gauge fuel tubes, and the beam size in the flame was approximately 1 × 1-mm² square. We chose a fluence of 1 J/cm² for all of the LII-based volume fraction experiments described here to ensure that soot sublimation temperatures were obtained and that peak LII signal values were independent of the exact laser fluence used. The LII signal was detected at 90° to the incident beam through an achromatic telescope, a series of neutral density filters, a bandpass filter (681.8 ± 10 nm), and a 100-μm circular aperture using a gated photomultiplier tube (PMT; Hamamatsu Photonics K.K.). We recorded signals on an oscilloscope (Tektronix, Inc., model TDS 694C) with a 3-GHz bandwidth that was optically triggered by a InGaAs photodiode. LII temporal profiles were averaged for at least 500 laser shots with a sampling rate of 10 GS/s.

The laser configuration for the extinction measurements was similar to that for the LII experiments, except that a 1-mm diameter circular ceramic aperture was used instead of the square aperture, resulting in a circular beam with an approximate 0.5-mm diameter within the flame. For extinction measurements, laser wavelengths of both 1064 nm and 532 nm (the second harmonic of the laser) were used. At each HAB, the laser fluence was chosen to avoid soot sublimation; fluence values ranged from 0.05 J/cm² to 0.2 J/cm². Transmittance (T_λ) is related to extinction and is given by the ratio of the intensity of the laser beam after passing through the flame (I) relative to the intensity of the beam before passing through the flame (I₀). In this case, I was measured on a pyroelectric detector (Electro-Optics Technology, Inc., models ET-3500 and ET-4000 for 1064 and 532 nm, respectively) with the flame on (and background corrected), and I₀ was measured with the same detector with the flame off.

We inferred soot volume fraction from transmittance measurements using Beer’s Law, i.e.,

where T_λ is the transmittance, $σabsagg$ is absorption cross section for the aggregate, $σscatagg$ is the scattering cross section for the aggregate, C_agg is the concentration of aggregates, and l is the optical path length in the flame. Substituting the Rayleigh-Debye-Gans (RDG) expressions for the aggregate absorption and scattering cross sections⁷⁵ and rearranging to express the transmittance in terms of the volume fraction gives

where E(m) is the dimensionless refractive-index function for absorption, F(m) is the dimensionless refractive-index function for scattering, G is a factor that accounts for aggregate structure,⁷⁵ λ is the laser wavelength (1064 nm or 532 nm), N is the number of primary particles per aggregate, d is the primary particle diameter, and f_v is the soot volume fraction given by

Equation (4) is often expressed in terms of the dimensionless extinction coefficient K_e as

where

The second term in this expression is attributable to the scattering contribution to the extinction and is often ignored. The factor G can be expressed as

where d_f is the fractal dimension of the aggregate and R_g is the radius of gyration. We estimated the contribution from scattering by assuming a radius of gyration of 50 nm and a fractal dimension of 1.8, which yields a value of 0.818 for 532 nm and 0.945 for 1064 nm. Based on values of the ratio F(m)/E(m) given by Michelsen et al.,⁷² we assumed that F(m) is 2.9E(m) for 532 nm and 3.3E(m) for 1064 nm. We assumed values for d of 11 nm and for N of 100. Although these parameters will depend on particle maturity and HAB, we have kept them constant for these calculations because the scattering contributions are small, and this assumption is anticipated to have little effect. We have measured the dispersion exponent using LII with laser wavelengths of 1064 and 532 nm (Johansson et al., in preparation), which were used to estimate particle maturity and values for E(m), as outlined in López-Yglesias et al.⁷⁶ with gas temperatures from Fig. 7.

Values for E(m) range from 0.446 to 0.489 for 1064 nm and from 0.443 to 0.451 for 532 nm. Calculated values at 1064 nm were 8.47 < K_e < 9.28, and values at 532 nm were 8.72 < K_e < 8.87 (values increased slightly with HAB). The values agree to within about 15% with the K_e given by Williams et al.⁷⁷ in an ethylene/air slot burner flame. We estimated the path length l by measuring the length of the luminous region using the image in Fig. 4. The transmittance measurements at both laser wavelengths are shown in Fig. 9 and are also listed in the supplementary material in Tables S2 and S3.

Based on the methodology summarized in a recent review by Michelsen et al.,⁷⁸ we used the peak LII signal intensity values at each position in the flame to obtain soot volume fractions from the LII data. We used a scaling factor, derived at an HAB of 6 mm, to match the volume fractions from the 1064-nm LII data to that of the 1064-nm laser extinction data. The Root-of-Sum-of-Squares (RSS) uncertainty in the reported volume fractions (based on the 1064-nm extinction calibration) is approximately ±17%, stemming primarily from uncertainty in the values of the measured transmittance (1.5% uncertainty in T_λ leading to 13% uncertainty in f_v), refractive index function (10% in $Em$ leading to 10% in f_v), and path length (6% in l leading to 6% in f_v).

The volume fraction results are reported for measurements along the burner centerline as a function of the HAB in Fig. 10 and demonstrate good agreement between the LII and laser extinction measurements with respect to the overall profile shape. Differences between the LII and extinction volume fractions at low and high HABs can likely be attributed to changes in the value of K_e with soot maturity and morphology (e.g., N should be larger for higher HABs, as shown below), which are likely to affect the extinction measurements, particularly for 532 nm, more than the LII data at the LII fluences used here.^77,78 Differences are also attributable to interferences by polycyclic aromatic hydrocarbon (PAH) absorption, which perturb the extinction measurements, particularly low in the flame and at the shorter laser wavelength of 532 nm. The peak soot volume fractions in Flame E1 are about one order of magnitude lower than those reported in an ethylene/air slot-burner flame,^37,39 in an ethylene/air Santoro-burner flame,⁷⁹ and in a Bunsen-type ethylene/air co-flow diffusion flame.⁸⁰ The lower soot volume fractions observed here are likely attributable to the effects of partial premixing. Tabulated volume fraction values for Flame E1 are given in the supplementary material in Tables S4–S10.

Measurements of the volume fraction at three HABs are provided in Fig. 11 as a function of position perpendicular to the long axis of the burner. At an HAB of 4 mm, soot forms preferentially at the flame fronts on the flame edges. Higher in the flame the soot distribution peaks in the center of the flame.

Figure 12 shows measurements of the volume fraction at an HAB of 6 mm as a function of position along the long axis of the burner (see orientation of Fig. 4). The soot volume fraction is relatively constant along the length of the flame; over the central 30 mm, it varies by <10%. The path length measured using the LII signal in Fig. 12 matches that obtained using the photograph in Fig. 4 to within 1%. In addition, there is good agreement between the transmittance derived at an HAB of 6 mm by integrating the f_v values shown in Fig. 12, according to

and the path-averaged transmittance at 6 mm HAB given in Fig. 9 to within about 0.4%.

We used a rapid thermophoretic sampling technique^81–85 to obtain soot samples in Flame E1 on 3-mm copper grids (Ted Pella, Inc., #01824 or #01830). A double-acting pneumatic cylinder (Parker Hannifin Corp. model B511BB549C) with a 24-mm stroke propelled the grids into and out of the flame, and the rig was oriented such that the actuation direction was perpendicular to the line of fuel tubes and was centered upon the center tube in this line. Depending on the HAB and soot concentration, flame exposures ranged from 40 to 100 ms. We used a JEOL USA, Inc., model JEM-1200EX microscope that was fitted with an 11-megapixel digital camera (Gatan, Inc., model ES1000W) to image the grids, and processed the images using ImageJ software.⁸⁶ 

Figure 13 shows transmission electron microscopy (TEM) images of soot sampled at HABs of 5 mm, 7 mm, and 9 mm from Flame E1. The images of soot sampled at 5 mm appear to have low contrast, possibly indicating that the soot particles may be composed of condensed semivolatile organic species.⁸⁷ Also, no aggregate particles can be seen at this location. The soot primary particles here have a diameter of approximately 10.3 nm on average. At 7 mm above the burner, the soot shows significant necking between the primary particles in the aggregates. The primary-particle diameters at this location are slightly larger, roughly 11.5 nm on average. At 9 mm, the images exhibit more distinct, spherical primary particles with average diameters of roughly 8.4 nm, and the aggregates show a characteristically fractal morphology. These morphological changes may be attributable to the oxidation (carbonization) of the soot as it ages during combustion in our partially premixed diffusion flame.⁸⁸ 

We have designed and constructed a modification of the Hencken burner that provides a linear laminar partially premixed flame that is ∼38 mm long. To characterize the burner’s performance, we measured its cold-gas relative velocity profiles using a hot wire anemometer, obtained the centerline temperature profiles of one ethylene/air flame and two hydrogen/air flames using coherent Raman imaging, measured soot volume fractions in one ethylene/air flame using LII/laser extinction, and obtained TEM images of soot in one ethylene/air flame. Tabulated temperature profiles, tabulated transmittance values, tabulated soot volume fractions, and CAD (mechanical) drawings are provided as the supplementary material.

See supplementary material for tabulated temperature profiles recorded using CARS measurements for Flames E1, H1, and H3 shown in Fig. 7, which are given in Table S1. Tabulated transmittance values along the centerline parallel to the long axis of the flame shown in Fig. 9 are given as a function of HAB for 1064 nm in Table S2 and for 532 nm in Table S3. Values for volume fraction inferred from these transmittance data are given in Table S4 for 1064 nm and Table S5 for 532 nm, and the corresponding values for volume fraction inferred from LII at 1064 nm are given in Table S6. These data are shown in Fig. 10. The data shown in Fig. 11 for the side-to-side soot volume fraction distributions inferred from LII measurements are given in Table S8 for an HAB of 4 mm, in Table S9 for an HAB of 6 mm, and in Table S10 for an HAB of 9 mm. The soot volume fraction as a function of the position along the length of the flame at an HAB of 6 mm as shown in Fig. 12 is given in Table S7. CAD (mechanical) drawings are also provided as part of the supplementary material in a file named SupplementaryCADDrawings.pdf.

We thank Ken R. Hencken for inspiration and advice and William C. Birdsey for useful discussions and for help in manufacturing the burner. This work was supported by the Division of Chemical Sciences, Geosciences, and Biosciences, the Office of Basic Energy Sciences (BES), the U.S. Department of Energy (DOE). K.O.J. was funded by DOE BES under the Single Investigator Small Group Research (SISGR), Grant No. DE-SC0002619. Sandia National Laboratories is a multi-mission laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under Contract No. DE-AC04-94AL85000.