Trapping and aerogelation of nanoparticles in negative gravity hydrocarbon flames

hydrocarbon flames Rajan K. Chakrabarty, Igor V. Novosselov, Nicholas D. Beres, Hans Moosm€ uller, Christopher M. Sorensen, and Christopher B. Stipe Department of Energy, Environmental & Chemical Engineering, Washington University in St. Louis, St. Louis, Missouri 63130, USA Laboratory for Aerosol Science, Spectroscopy, and Optics, Desert Research Institute, Nevada System of Higher Education, Reno, Nevada 89512, USA Department of Mechanical Engineering, University of Washington, Seattle, Washington 98195, USA Enertechnix Inc., Maple Valley, Washington 98068, USA Condensed Matter Laboratory, Department of Physics, Kansas State University, Manhattan, Kansas 66506, USA TSI Incorporated, 500 Cardigan Rd, Shoreview, Minnesota 55126, USA

We report the experimental realization of continuous carbon aerogel production using a flame aerosol reactor by operating it in negative gravity (Àg; up-side-down configuration).Buoyancy opposes the fuel and air flow forces in Àg, which eliminates convectional outflow of nanoparticles from the flame and traps them in a distinctive non-tipping, flicker-free, cylindrical flame body, where they grow to millimeter-size aerogel particles and gravitationally fall out.Computational fluid dynamics simulations show that a closed-loop recirculation zone is set up in Àg flames, which reduces the time to gel for nanoparticles by %10 6 s, compared to positive gravity (upward rising) flames.Our results open up new possibilities of one-step gas-phase synthesis of a wide variety of aerogels on an industrial scale.V C 2014 Author(s).All article content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 Unported License.
[http://dx.doi.org/10.1063/1.4884057]Aerogels are volume spanning, semirigid networks of solid nanoparticles (NPs). 1 Owing to their unique material properties such as ultralow density and high surface area, 2 these mesoporous materials have found extensive applications ranging from catching space dusts to purifying air and water supplies. 2,3However, aerogel synthesis via the sol-gel process is non-continuous and requires supercritical point drying, hence is time-consuming and expensive. 1This has prevented their mass production and widespread application. 2Cost-effective and continuous synthesis routes using gas-phase flame aerosol reactors (FARs), 4 which have been widely adopted by industries for production of nanostructured materials, 5 are yet to be demonstrated as viable options for producing aerogels.
In the gas-phase, aerogels form from a dispersion of NPs via the process of irreversible diffusion-limited cluster aggregation (DLCA). 6DLCA starts out in the limit of spatially uncorrelated binary cluster collisions (cluster dilute regime 6 ) to partially relieve the system of its non-equilibrium condition 7 and leads to formation of fractal NP clusters with fractal dimension %1.8.This continues until the cluster volume fraction (f clus )-the ratio of cluster separation to sizeincreases to the point where clusters no longer encounter each other in a spatially uncorrelated manner, that is, the clusterdense regime. 6In this regime, when the f clus reaches unity, the clusters jam together to form a gel.The phase of high f clus is an absolute necessary condition for gelation to occur.The time for a gel to form, t gel , can be approximated as where a is the radius of the basic NP repeating unit (or monomer) constituting a cluster, f v is the monomer volume fraction, and K is the aggregation kernel, which specifies the cluster aggregation rate. 8The approximate Eq. ( 1) is both simplistic and instructive in laying down the strong functional dependence of t gel on f v and a.
Conventional sol-gel routes allow sufficient residence time t res (several hours to days) for NPs in a solution to reach threshold t gel . 1 On the contrary, NPs experience t res on the order of submilliseconds 4 in an upward rising (þg; positive gravity) FAR.In such rapid kinetics conditions, an initial monomer f v ! 10 À4 is needed to yield a t gel that is shorter than t res . 6Such high f v values have been observed to exist locally in the annular region of a heavily sooting þg diffusion flame's tip. 9 However, the high shear stress accompanying the buoyant tip challenges the structural stability of a gel-like cluster network, and tears it apart. 10The fragmented NP clusters, after exiting the tip, encounter a high dilution zone in the flame's over-fire region, where high f v conditions are impossible to achieve.Thus, unsustainable cluster-dense f v conditions, very low t res , and buoyancy-induced instabilities of a þg flame have rendered FARs unsuitable for gel synthesis.
In this Letter, we demonstrate that for FARs operated in an up-side-down (Àg; negative gravity) configuration, the necessary gelation conditions of cluster-dense f v and enhanced t res are steadily maintained in a large crosssectional area of the flame body.A "U" shaped, closed but non-tipping zero-acceleration plane of stagnation is created by the effect of buoyancy opposing the fuel and air flow fields 10,11 (Fig. 1).This plane is stable and non-flickering, and traps NP clusters from convectionally flowing out of the flame body.As a result, the trapped clusters experience sufficient t res > t gel , which enable them to cross-over to the cluster-dense regime and gel.The clusters gel continuously in the flame and fall out gravitationally once they grow above a threshold gel size.
We operated a basic co-flow, Burke-Schumann arrangement diffusion FAR 10 in both Àg and þg configurations to systematically study the differences in emitted NP cluster properties.The flame was housed in a 48.3 cm long and 4.75 cm diameter quartz tube.Hydrocarbon fuel flowed through a 1.9-cm diameter inner tube, while combustion air was injected into the annular region between the fuel jet and the outer quartz tube.We fed the burner with two hydrocarbon fuels of varying threshold soot index (TSI), 12 namely, ethylene and acetylene.The flow rates of fuel and air were varied to tune the net fuel-to-air equivalence ratio (u). 13For ethylene, we varied the fuel flow rates between 1.2 and 2.0 l min À1 and the co-flow air rates between 16.5 and 17.5 l min À1 .For acetylene, we fixed the air flow rate at 105 l min À1 and varied the fuel flow rates between 0.5 and 0.8 l min À1 .Under these flow conditions, u varied between 0.92-1.54and 0.12-0.2for ethylene and acetylene, respectively.We observed the Àg flame morphologies to be significantly longer, wider, and cylindrical in shape, compared to þg flames operated under same flow conditions (Fig. 1).Gel particles were gravitationally settling out from flame bodies and were collected on petri dishes placed below the flame for structural analysis.The quantity of gel production increased with increasing u and fuel TSI.
Figure 2 shows electron microscopy (EM) and optical microscopy images for acetylene aerosol gel particles.Here, we specifically chose to present results from our acetylene combustion experiments.Acetylene is a fuel widely used in industries and research laboratories for manufacturing carbon black powder.Our burner produced a maximum of 5 g h À1 of acetylene gel particles.Figures 2(a)-2(d) graphically demonstrate that a macroscopic aerosol gel particle, greater than a millimeter in size, forms from the aggregation of radius a % 10 nm carbon NPs (monomers).In contrast, the average size of an aggregate emitted from a þg FAR is sub-micron ( 1 lm) with a % 20 nm. 14 The EM pictures show that the aerosol gels are ramified fractal structures with pores trapped inside.The specific surface areas (SSAs) and surface porosity of these gel particles were determined using the Brunauer-Emmett-Teller analysis technique with nitrogen gas as an adsorbate. 15The gel particles were found to have a mean SSA of 208 m 2 /g, which is about four times higher than that of soot aggregates generated by þg FARs. 15These particles were found to have a specific mesopore volume of 0.24 cm 3 /g for a mean pore size of 1.7 nm.We determined the gel (effective) density of the particles, by measuring the mass of a known volume of the sample, to be as low as 4.5 mg/cm 3 .
We performed detailed computational fluid dynamics simulations for an axisymmetric laminar acetylene-air diffusion flame operated in both Àg and þg configurations to gain an accurate understanding of the gel formation mechanism.ANSYS v14 computational package was used for performing the simulations.Aerosol formation in the flames was modeled by a comprehensive approach as described by Brooks and Moss, 16 which includes terms for carbon NP nucleation, surface growth, coagulation, and oxidation.Our numerical model solves the time-dependent, twodimensional equations for conservation of mass density, momentum, energy, individual species concentration, f v and t res distribution inside the FAR, and rates of reaction.The computational grid included the full burner geometry including the extension tube, located downstream of the quartz tube.Acetylene and air flow rates were maintained at 0.65 and 105 l min À1 , respectively, for Àg and þg configurations, to match our experiments.
Figure 3 shows instantaneous images of axial velocity, t res , and f v at one time step for our flame system in Àg configuration.The "U" shaped stagnation plane of the Àg flame is well replicated in our simulation results.The Àg flame reaches a maximum axial velocity of 1.5 cm/s, much lesser than the velocity of þg flames under same fuel-air flow conditions.Due to the significant reduction in axial velocity and buoyancy-driven convection in Àg flames, diffusion becomes the dominant mechanism of transport. 11As a result, these flames are much longer and wider than þg flames with thicker diffusion layers.
By tracing the axial velocity vectors in Fig. 3(a), our simulation results suggest that a particle originating at the center of a Àg flame gets trapped in a closed-loop recirculation zone, which has a turn-around time of 1.3 s.Under ideal conditions, a trapped particle would be unable to escape and encounters an unlimited t res in this zone.In reality, we observed gravitational forces ejecting gel particles out of the recirculation zone when it grew above %1 mm in size.
The area-averaged f v in the recirculation zone was calculated to be 1 Â 10 À5 , which is around 100 times greater than the average f v (9.8 Â 10 À8 ) in þg flames.Because of the increase in f v in this zone, the particles emit and absorb larger quantities of thermal radiation.The resulting radiative heat loss causes a decrease in net flame temperature (1900 K), compared to a þg flame (2100 K).
With the knowledge of f v and a, we estimated the aggregation kernel K in both Àg and þg acetylene flames.The average f v in the recirculation zone can be expressed as where n is the number of monomers, V p is the volume of a monomer ( 4 3 p a 3 ), A is the view area, and h is the depth of view.For Àg flame systems, using Eq. ( 2), a ¼ 10 Â 10 À7 cm and f v ¼ 1 Â 10 À5 , we get n ¼ 2.4 Â 10 12 cm À3 .For þg flame systems, using Eq. ( 2), a ¼ 20 Â 10 À7 cm and f v ¼ 9 Â 10 À8 , we get n ¼ 2.9 Â 10 9 cm À3 .The growth of DLCA clusters is governed by the Smoluchowski equation, 17 and the mass to linear size (radius of gyration) relationship of the clusters made up of monomers can be expressed as a power-law relationship where N is the number of monomers constituting a cluster, D f (¼1.8 for a cluster-dilute aggregate) is the mass fractal dimension, 17,18 R g is the aggregate radius of gyration and k 0 is the fractal pre-factor.For millimeter-size aerosol gel clusters, we assumed monodisperse and spherical (D f ¼ 3) clusters as per past recommendation, 9 and calculated their R g from their perimeter radius R using the equation The average R g of a gel cluster was calculated to be 0.34 mm.For þg flame-generated sub-micron size clusters, R g ¼ L max /3, where L max is the maximum projected length of a cluster. 19We calculated the average R g of þg flame clusters to be 333 nm.For Àg flames, using Eq. ( 3), we calculated average N ¼ 1.7 Â 10 8 per aerosol gel cluster.For þg flames, we calculated average N ¼ 316 per aggregate.By dividing n by N, we calculated the density of clusters g for Àg and Àg flames to be 1.36 Â 10 3 and 9.2 Â 10 6 cm À3 , respectively.The equation connecting K, aggregation rate kernel, and g is given as 9 We used t ¼ 0.1 s, representative of average time for cluster formation in a hydrocarbon flame, 9 and Eq. ( 5) to calculate K for Àg and þg flames to be 7.3 Â 10 À4 and 1.1 Â 10 À6 cm 3 s À1 , respectively.Using Eq. ( 1) and K, we calculated t gel % 60 ms in the recirculation zone, which is more than eight orders of magnitude lower than t gel % 10 7 s of a þg flame.This factor of >10 8 decrease in t gel coupled with the zone's non-converging t res facilitates the continuous structural arrest of NPs to form gel clusters.We speculate that this zone represents a deep local minimum in the energy landscape. 20However, verifying this is beyond the scope of this study and a topic of future research.
In conclusion, we demonstrated a facile, highthroughput (grams per hour) carbon aerogel synthesis method via the flame aerosol route.It overcomes the complexities involved in synthesizing carbon aerogels using the non-continuous and time-consuming wet sol-gel process that requires supercritical point drying.Unlike solution phase methods, the flame aerosol route allows better control of the shape and size of monomers. 5While the SSA of carbon aerogels produced by our technique is lower than those produced via the sol-gel route, past research has shown that particle SSA in flames could be significantly increased by applying external electric fields across the flame flow or by adding additives to the fuel mixture. 21Our aerogelation mechanism also has the potential of finding applications in the production of advanced functional structures requiring high surface area per unit volume, such as cathode materials for Li-ion batteries 22 and photoactive thin films. 23

FIG. 1 .
FIG. 1. Gravitational effects on emitted particle morphology from flames.(a) and (b) Photographs of our diffusion FAR operated in downward (Àg) and upward (þg) configurations under same flow conditions.The fuel used was acetylene.The dark region in the middle of the Àg flame body is the recirculation zone, from where gel particles fall out.(c) and (d) Schematic of particle formation in these flame systems.In Àg flames, cluster-dilute particles get trapped in a deeply metastable recirculation zone.This zone is formed due to the opposing effects of buoyancy on fuel-air flow forces.With time, the particles cross-over to cluster-dense conditions and get structurally arrested to form gel particles.In contrast, a þg flame buoyantly convects out cluster-dilute particles from the flame body.While these particles encounter (for a fraction of a second) a cluster-dense zone near the tip of the flame, the shear stress associated with the flickering flame front does not allow gel formation.

FIG. 3 .
FIG. 3. Two dimensional contour plots of simulated acetylene flame parameters for downward (Àg) configuration.(a) Particle path lines colored by axial velocity.(b) Path lines colored by particle residence time.(c) Soot monomer volume fraction distribution.