Aerosol Flame Reactors for the Synthesis of Nanoparticlest

This paper presents an overview of recent basic research on flame aerosol reactors for the gasphase synthesis of nanoparticles. Emphasis is placed on flame reactor technology as it is widely used in industry for the large-scale manufacture of oxide and carbon nanoparticles. The importance of reactant gas mixing, additives and external electric fields in flame technology is highlighted for the control of product particle properties by affecting the chemistry, temperature and collision histories. Laser-induced fluorescence (L!F), Fourier transform infrared spectroscopy (FTIRJ and thermophoretic sampling are addressed as some of the promising diagnostic tools in flame aerosol research and even for on-line process control. Recent work on aerosol dynamics modeling is presented, and the growing importance of computational fluid dynamics (CFD) aimed at better understanding the particle formation and growth mechanisms in flames is emphasized, focusing on the synthesis of non-aggregated


Introduction
Over the last few years, the scientific community has cherished the potential of nanosize clusters or particles [1,2,3].These entities have distinctly different properties to bulk material because the number of atoms or molecules on their surface can become comparable to that inside the particles [4].Some people even believe that nanosize particles may constitute another state of matter.Laboratory studies in various scientific fields show that especially non-aggregated nanoparticles can be used to develop new materials with unique characteristics such as optical, mechanical, electrical, catalytic, and heat transfer properties.With nanoparticles, the particle melting point decreases [5], light absorption increases, and magnetic, optoelectronic [6] and other material properties change compared with those in the bulk material.Their large surface area to volume ratio and high density of active sites make nanoparticles attractive for applications in catalysis.By reducing the grain size to nanometer scale, the sintering temperature of structural ceramics can be decreased while the plasticity is increased [7].Nanoparticles can also be used in thin membrane films, for protective glass, metal and polymer coatings, in the production of inorganic/polymeric nanocomposites, as electronic devices and, if assembled in chains or arrays, as novel recording and storage media for digital data.
In general, nanosized powders can be synthesized via the wet chemical route and gas-phase processes.The latter are advantageous for powder manufacture since they do not involve the tedious and expensive steps of solid-liquid separation, washing and drying of wet chemistry processes and they avoid the use of high liquid volumes and surfactants [1].Today, flame processes are by far the most widely used methods for the gas-phase manufacture of commercial quantities of nanoparticles, the most important of which are carbon blacks and fumed silica, produced for instance by Cabot and Degussa-Huls, as well as pigmentary titania, made by DuPont, Ishihara, Kerr-McGee, Millenium, Tioxide, and others.The annual production volume of the flame industry is several million metric tons, and aerosol reactors produce nanoparticles at a rate of 100 metric tons per day [8].Other commercial aerosol processes include hot-wall reactors for the industrial synthesis of filamentary (nanostructured) nickel and iron powders from decomposition of the corresponding metal carbonyls (BASF, !NCO).In addition, these reactors have been used for the commercial synthesis of nanostructured carbides, nitrides, borides, and other non-oxide ceramics (Dow, Bayer).More recently, the inert gas condensation technique has been scaled up for the manufacture of rather costly (about $100/kg) nanostructured metals and ceramic powders (Nanophase).Spray pyrolysis technologies are used primarily by small start-up companies (for example, Particle Technology, Nanochem, SSC, and others) for the manufacture of precious metal, ceramic and especially nanostructured composite ceramic powders from nitrate, organic and other solutions.
Despite the age and significance of industrial gasphase processes, their design and operation rely heavily on experience and empiricism.As a result, it is nearly impossible for existing industrial units to address the synthesis of nanosize powders without going through the laborious and expensive cycle of Edisonian research that was followed for the development of the current units and processes (e.g. the socalled "chloride" process for Ti0 2 synthesis).However, in contrast to the past state of affairs in industry, substantial work has been done on a laboratory scale toward understanding the fundamentals of nanoparticle formation and growth at high temperatures [8,9], and even the use of computational fluid dynamics for reactor design is actively investigated in industry today [10].

Process Classification
Gas-phase synthesis methods for nanoparticles can be divided into gas-to-particle and droplet-to-particle processes.Gas-to-particle conversion refers to the production of particles from individual atoms or molecules in the gas phase.Product powders generally exhibit small particle size, narrow particle size distribution, non-porous particles and high purity (Figure 1).Compared to droplet-to-particle or wet chemical processes, it is more difficult to produce multicomponent materials, and special care has to be taken in handling hazardous off-gases and precursors.Examples of these processes include flame [11], hot-wall [12,13], evaporation-condensation [14,15], plasma [16,17], laser [18] and sputtering [19,20].
In the droplet-to-particle route, solution droplets are suspended in gases by liquid atomization or by condensation of a superheated vapor.This is followed by the evaporation of solvent from droplets and solute crystallization to form a dried particle undergoing solid-state reactions and sintering upon heating.Compact nanostructured particles can be formed by spray drying of a nanoparticle slurry, as is shown in Figure 2. In general, the advantages of droplet-to-particle processes are: the ability to process organic and inorganic materials, to form a variety of multicomponent particles, simplicity, many choices for inexpensive liquid-phase precursors and a low degree of particle aggregation.Porous or hollow particles can be formed under certain conditions.However, the spread of the particle size is limited by the spread of the starting droplets.Spray drying [21] and pyrolysis [22] are typ-KONA No. 18 (2000) Fig. 1 Spherical, dense titania nanoparticles synthesized in a diffusion flame reactor by oxidation of titaniatetraisopropoxide (dp, av=35 nm).
ical industrial processes employing droplet-to-powder conversion and can be used for the manufacture of nanoparticles as well as flame spray pyrolysis [23], electrosprays [24,25] and freeze drying [26].Recent books give a detailed picture of the field [27,28].

Gas-Phase Synthesis Route
In the gas-to-particle route, nanoparticles are made by "building" them from individual molecules all the Fig. 3 Gas-to-particle conversion: Schematic of particle formation and growth.Adapted from Kodas and Hampden-Smith [27].
way up to the desired size as is shown in Figure 3.
The particle formation process is driven by the generation of molecules by chemical reaction from precursor gases or by rapid cooling of a superheated vapor.High temperatures are usually required to accomplish the reaction or to bring the vapor to the superheated state.Depending on the thermodynamics of the process, the product molecules can form particles either by uninhibited collisions (collision-controlled nucleation) or by a balanced condensation and evaporation to and from molecular clusters (condensationevaporation controlled nucleation).The newly formed particles grow further by collision with product molecules (condensation or reaction on the particle surface) and/or with particles (coagulation).In coagulation, two particles collide and stick to form an aggregate or agglomerate.The particles within this agglomerate can coalesce (fuse) by sintering into a spherical particle of the same volume and mass.However, as coagulation continues, the ratio of the sintering rate (the rate of particle coalescence) and the rate of coagulation determines the morphology of the final product particle.When the rate of coalescence is faster than that of coagulation, spherical particles are obtained.However, as particles grow, the sintering rate decreases and particle coalescence usually becomes slower than coagulation.As a result, irregularly shaped aggregate particles are formed.These are termed hard or soft aggregates (or agglomerates) depending on how easy it is to break the bonds connecting the primary particles [8].

Flame Synthesis
In flame reactors, the energy of a flame is used to drive chemical reactions of precursor compounds that result in the formation of product molecules, which then nucleate to form particles following the mechanism described earlier [8].High flame temperatures Argon . . . . ..rTIC of 1200 to 3000 K constitute a self-purifying environment for particle synthesis resulting in high-purity powders, as are needed for the manufacture of optical fiber preforms [29].Even though non-oxide ceramic powders such as silicon nitride [30] and tungsten carbide [31] have been synthesized in flame reactors, the production of carbon and metal oxide nanoparticles dominates the field.One distinguishes two types of flame reactors: diffusion and premixed.While premixed flames have the advantage of more uniform radial temperature profiles, diffusion flames are safer to operate (no flashback) and offer flexibility in product qualities by controlling the reactant gas composition over broad ranges.Turbulent co-flow diffusion flame reactors are used in industry for the large-scale manufacture of ceramic powders.Counterflow diffusion flame reactors, on the other hand, are often used in laboratory studies [8].The flow along the axial stagnation streamline of a counterflow diffusion flame can be approximated as one-dimensional, facilitating the use of optically-based diagnostics for the nonintrusive study of nanoparticle formation [9].
A typical laboratory set-up of an aerosol diffusion flame reactor is depicted in Figure 4.The co-flow reactor is made of three concentric stainless steel tubes of about 120 mm length.The inner diameter of the center tube is 5 mm, while the o.d. of the outer tube is 18 mm and the width of the two annular ports is about 1.5 mm.Oxygen is delivered through the outer tube while methane flows through the center port.High-purity argon loaded with the precursor vapor is added to the fuel stream prior to entering the burner.A small stream of nitrogen is introduced between the fuel and oxidant streams to prevent deposition of particles on the burner mouth.An evaporator or a bubbler can be used to load the inert gas stream with the precursor.When a bubbler is used, controlled amounts of precursor can be delivered into the reactor by adjusting the temperature of the liquid compound with the help of a thermostated bath.Evaporators offer the advantage of precise precursor dosing, easier handling of moisture-sensitive chemicals and the evaporation of mixtures.Industrial processes usually operate with halide precursors.In laboratory units, organometallic compounds are preferred to circumvent HCl and chlorine removal from the off-gas and to avoid post-cleaning of the powders.In order to prevent condensation of precursor vapor, the precursor delivery tubes and the reactor are heated above the temperature of the evaporator or bubbler.Product particles are collected on a filter with the aid of a vacuum pump.The flame is surrounded by a quartz glass cylinder in order to achieve stable burning.The outer tube of the reactor can be surrounded by a sheath of nitrogen for flame stabilization.
In diffusion flames, the maximum flame temperature usually occurs at the tip of the cone-shaped flame front, where aggregates fuse [32].Upon leaving the flame, the temperature drops quickly and particles continue to coagulate while sintering takes place at much lower rates.In many systems, the characteristic time for coalescence is longer or in the same order as the characteristic time for coagulation, resulting in aggregates consisting of a few to many thousands of primary particles [33].The characteristics of the product particles such as the morphology, crystallinity and size strongly depend on the precursor concentration, temperature profile and residence time distribution in the flame.

Current State of Research
The current state of research in this field has been recently reviewed by Pratsinis [8] and Wooldridge [9] so that here only the keypoints are reported.Fumed silica became industrially important in the 1940s as a substitute for carbon black [11] and was first marketed under the name "Aerosil" by Degussa AG [34].Flame technology was soon also applied in the manufacture of titanium dioxide from TiC1 4 as precursor (chloride process) to replace the wet-chemistry-based sulfate process [11].The manufacture of uranium dioxide by oxidation of uranium hexafluoride in a flame became the third industrially important flame process with high potential to replace its wet precipi-tation counterpart [11].The flame generation of U0 2 , as described by Federer eta!. [35], yields powders of high purity which can be pressed and sintered to form fuel pellets for nuclear reactors.
Paralleling the growing industrial importance of the flame processing of chemicals in the mid-20th century, research in this field was first led by industrial laboratories.Emphasis was placed on burner design and use of additives [8].The early patent race in flame technology for the synthesis of Ti0 2 and Si0 2 was superbly summarized by Mezey [36].
The establishment of flame technology in industry stimulated a number of research issues in academia.Especially Ulrich and co-workers [37,38,39,40] along with Formenti eta!. [41] pioneered research on flame synthesis of ceramic powders.The group of Ulrich studied the synthesis of Si0 2 by SiC1 4 oxidation in premixed laminar and turbulent jet flames.Thereby, they recognized that coagulation rather than nucleation was the dominant particle formation mechanism and that the size distributions of the product particles were self-preserving [42].These early studies revealed that the appearance of aggregates of primary particles results from the competition between particle collision and sintering [39].
In the mid-1980s an idea already mentioned in a 1936-patent by Corning Glass Works, New York, [43,44] further accelerated research in the field: the deposition of flame-generated silica to form transparent articles of high purity led to the introduction of flame reactors in the large-scale manufacture of optical fiber preforms [29].Today, the production of optical waveguide preforms which are used to draw hairthin fibers is one of the most profitable processes of Corning Inc.In academia, opposed jet (counterflow) diffusion flame reactors were introduced by Chung and Katz [45] for the synthesis of oxide powders.The flat and stable flame of the burner exhibits uniform temperature and species concentration distributions in the horizontal plane, thus allowing a precise tracing of particle formation.The importance of sintering in oxide particle formation was demonstrated by Helble and Sarofim [ 46] while investigating fly ash formation during coal combustion.Hurd and Flower [47] introduced fractal concepts to describe the structure of the resulting silica aggregates from a laminar premixed methane flame.Koch and Friedlander [33] presented a simple but elegant model describing the formation of non-spherical particles by coagulation and sintering.
Renewed interest in flame technology for the manufacture of advanced materials intensified research in the field since the early 1990s.Mixed-oxide systems such as Si0 2 -Ge0 2 , Alz0 3 -Ti0 2 , and V 2 0 5 -Alz03 were studied by Hung, Miquel and Katz [48,49,50], while Zachariah and co-workers [51,52] made superparamagnetic nanoparticles in a premixed methane-oxygen flame.When synthesizing submicron YBazCu 3 0 7 particles in an oxy-hydrogen diffusion flame reactor, Zachariah and Huzarewicz [53] found that the flame configuration may have a profound effect on the product powder properties.Pratsinis eta!. [54] observed that by merely altering the position of fuel and oxidant streams in diffusion flame reactors, the average particle size of product titania powders can be changed by as much as a factor of ten.
As a result, great emphasis was placed on determining the role of flame process variables such as flame configuration, temperature, oxidant composition, and precursor type.For example, Lindackers et a!.[55] made titania and silica particles in low-pressure, premixed oxy-hydrogen flat flames.This flame configuration resulted in enlarged reaction zones compared to standard-pressure flames and allowed them to observe the formation of titania particles.Of all variables, temperature probably has the most drastic effect on the process and product characteristics.Bautista and Atkins [29] found that hydrolysis is the main route for SiC1 4 conversion at low temperatures, while oxidation is the dominating mechanism at high temperatures.Hung and Katz [48] found that increasing flame temperatures resulted in high concentrations of fine particles during the synthesis of Si0 2 and Ti0 2 in a counterflow diffusion flame reactor.Large temperature gradients can generate strong thermophoretic forces on the newly formed particles, drastically altering their residence time at the decisive region where nucleation, growth, coagulation and sintering occur, thus affecting the morphology of flame-made particles [56].Today, oxides such as Si0 2 , Ti0 2 , Alz0 3 , Zr0 2 , V 2 0 5 , and most other oxides of metal elements in the periodic table and their composites have been produced in powder form in hydrocarbon flames on a laboratory scale [ 8,27].

Control of Particle Properties
The size and morphology of flame-synthesized particles depend on the structure and properties of the flame.The most important parameters are the temperature field of the flame, the particle residence time and, for diffusion flame reactors, the mixing of precursor and oxidant.These parameters depend on the reactor geometry and the gas flow rates into the burner.
The advantage of co-flow diffusion flame reactors regarding their flexibility in reactant mixing is illus-174 Fig. 5 Influence of the flame configuration on the morphology of titania particles synthesized with a diffusion flame reactor [54].The center tube diameter of this burner is 4 mm and the spacing between the successive tubes is 1 mm.Temperature profiles of the double-diffusion flames A, B, as well as those of single-diffusion flames C and D are calculated by computational fluid dynamics (CFD) and are adapted from Johannessen [57].The reactant mixing leads to control of primary particle size up to a factor of 10.
trated in Figure 5 by the example of titania synthesis from TiC1 4 [54] along with the corresponding temperature fields of the flame calculated by computational fluid dynamics [57].Different flame configurations referred to as the A, B, C, or D-flame were achieved by introducing the reactants through different ports of the burner.As can be seen from Figure 5, the particle size and morphology change drastically with the mode of mixing.Double-diffusion flame A produces the finest Ti0 2 particles with an average primary particle diameter of about 10 nm.Here, the precursor stream is diluted with air prior to its oxidation in the flame and the particles experience rather low temperatures [8,54].In double-diffusion flame B, the mixing of precursor and air streams takes place further downstream of the burner, resulting in particles larger than those of flame A In the single-diffusion flame C, mixing of precursor vapor and methane takes place much earlier than in flames A and B [57].The newly formed particles experience higher temperatures than in flames A and B with the temperature maximum being at the middle of the flame.As a result, sintering is rapid, creating large non-aggregate particles (""' 80 nm in diameter) as shown by TEM and proven by smallangle X-ray scattering [58].Titania particles synthesized in single-diffusion flame D are a bit larger than those synthesized in flame C. Flowing methane through the center tube results in a narrow flame front and high flame temperatures, leading to fast sintering rates and large particles with a low degree of aggregation.By using pure oxygen instead of air as the oxidant, the flame can even produce non-aggregated, perfectly spherical particles [59], as is shown in Figure 6.
For a given flame configuration, the oxygen flow rate as well as the choice of the precursor compound affects the size and morphology of flame-generated nanoparticles by altering the temperature field of the flame and gas mixing properties [60].Our recent work on a diffusion flame reactor operated in C-flame configuration, as in Figure 5, shows that the specific surface area of silica particles increases from 40 to about 200 m 2 I g when the oxygen flow rate is increased from 1000 to 12000 cm 3 /min (Figure 7).The average BET-equivalent particle diameters range from 63 to about 13 nm.Also, the degree of aggregation is reduced when the oxygen flow rate is decreased from 8000 to 2000 cm 3 /min.
Electric charges offer another tool for the control of the characteristics of flame-made powders.Electric fields applied across a particle-generating flame charge the newly formed nanosized particles.As a result, the coagulation rate of these particles is reduced in the high temperature region of the flame.Furthermore, particles are removed from the flame toward the external electrodes by electrophoresis.In premixed [61,62] and diffusion flames [63] making titania, tin oxide, silica, and even carbon or silicon-carbon composites [64,65], the aggregate and primary particle size decrease proportionally to the applied electric field strength.Figure 8 shows the average primary particle diameter of titania particles as a function of the applied electric field intensity across the flame together with TEM micrographs and pictures of the premixed flame.As can be seen, the average primary and aggregate size of these powders can be narrowly controlled by applying external electric fields across the flame.Recently, it was demonstrated by Kammler and Pratsinis [66] that electric fields can also be applied for the control of fumed silica particle properties when the production rate of a lab-scale burner was increased from a few grams per hour to almost 100 g/h.Kammler and Pratsinis further confirmed that the location of the electrodes with respect to the flame is a decisive factor in electrically-assisted flame aerosol synthesis [63].Thus, electric fields provide the unique opportunity for making powders of closely controlled size, composition and morphology.
The introduction of additives or dopants in particlegenerating flames is widely practiced in industry as they can have a profound effect on particle formation and growth mechanisms and subsequently on the product particle characteristics.Additives are used as a means to control the crystallinity (e.g.Si orAl in Ti0 2 for anatase or rutile) or the morphology (e.g.K in carbon blacks) of the powders.Vemury and Pratsinis [ 67] added SiC1 4 , SnC1 4 , and AlCh to a titanium tetrachloride precursor stream and investigated the effect of these additives on the phase composition, morphology and size of titania particles synthesized in a laminar diffusion flame.They found that the introduction of the silica-precursor inhibits the titania phase transformation from anatase to rutile, decreases the primary particle size, and, as a result, increases the specific surface area.These observations parallel those of Akhtar et al. [68], who had shown how Si, P or B create interstitial defects in the anatase lattice of titania made in a tubular hot-wall reactor.However, when SnC1 4 or AlCh were used as dopants [67], the phase transformation of anatase to rutile was enhanced and the specific surface area was decreased in agreement with Akhtar et al. [69], who had shown how Al creates substitutional defects in the titania lattice.
Regarding flame synthesis of silica, the role of ferrocene on product particle properties was investigated by Fotou et al. [70].The presence of ferrocene increased the specific surface area of silica up to 150% and removed the coarse tail of the silica aggregate size distribution.Retardation of coagulation by charging effects was found to be a possible explanation of these results.As was briefly pointed out, the use of additives and dopants in the flame synthesis of ceramic powders provides another means to control particle properties, especially when high purity of the product powders is not a concern.

Diagnostics
Combustion models and computational fluid dynamics (CFD) as well as state-of-the-art flame diagnostics are powerful tools for better understanding particle formation and growth mechanisms in the flame.Johannessen et al. [71] developed a computational fluid dynamics model of temperature, velocity and gas composition in a diffusion flame and combined it with a simple model for coagulation and coalescence of aerosol particles.By comparing the results with temperature measurements throughout the flame and analysis of alumina product particles and their size distribution, they gained insight into the fundamental flame synthesis mechanisms.Current efforts are toward an integration of particle dynamics models into CFD-codes as indicated in the recent works of Schild et al. [10] and Pyykonen and Jokiniemi [72].While Schild et al. integrated a monodisperse aerosol dynamics model [73] into fluid mechanics for the simulation of titania formation from TiC1 4 in a tubular aerosol flow reactor, Pyykonen and Jokiniemi [72] introduced a computational-fluid-dynamics-based sectional aerosol model to simulate aerosol formation in a laminar flow reactor.
Non-intrusive diagnostic tools for flame temperature, velocity and species concentration provide the means to both verifying the CFD simulations of flames and to providing an accurate and detailed database characterizing the environment for particle formation and growth models.Concerning flame velocities, 3phase Laser-Doppler Anemometry (LDA) has already been applied extensively to characterize laminar and turbulent flames.The results obtained from non-particle generating flames using micrometer-sized pigmentary titania or zirconia as seeds might be directly applicable to flame reactors, since their low precursor concentrations should not alter the velocity field significantly.Flame temperature measurements are usually done in the absence of precursor (e.g.Chung and Katz [ 45]) since particles either deposit on thermocouple probes or, for non-intrusive laser diagnostic methods, the interference of the response signal with particulate matter weakens the signal intensity.The oxidation especially of organometallic precursors, however, is highly exothermic and can constitute a large fraction of the overall combustion enthalpy of the flame [60].Thus, adding the precursor to a flame is expected to alter its temperature field, making the development of fast and reliable non-intrusive temperature diagnostics for particle-generating flames a necessity in flame aerosol reactor research.In particleladen low-pressure flames, spatially resolved temperature and OH species concentration were determined by Glumac eta!. [74] using laser-induced fluorescence (LIF).One of the few applications of LIF in particle synthesis at atmospheric pressures is the work reported by Zachariah and Burgess [75], who used LIF to measure OH and SiO concentrations and Mie scattering to measure particle distributions during the flame synthesis of silica powders.Such information is of fundamental importance to a better understanding of gas-to-particle conversion processes because gas-phase species determine not only the rate of particle formation, but also the chemical composition of the particles.
Fourier transform infrared (FTIR) spectroscopy has been successfully applied for in-situ temperature and concentration measurements in particle-generating flames operated at atmospheric pressure.Morrison et a!.[76] and Arabi-Katbi [77] have shown that FTIR measurements can determine particle temperature and concentrations as well as gas temperatures and concentrations in a premixed methane-oxygen flame for the synthesis of titania powders (Fig. 9).The application of FTIR spectroscopy also enables the acquisition of in-situ temperatures during flame synthesis of particles in the presence of electric fields as reported by Morrison et a!.[76].Hitherto, process temperatures have not been measured under these conditions since conventional temperature probes such as thermocouples cannot be applied.Information on the aggregate structure and size of particles during flame growth can be obtained by thermophoretic sampling on transmission electron microscopy (TEM) grids at different flame heights [78].By rapidly inserting and withdrawing a TEM grid, particles can be collected on the grid by thermophoresis.The method was applied by Arabi-Katbi [77] in a premixed flame, generating titania nanoparticles.Figure 1 Oa shows the tip of the sampler through which the grid is inserted into the premixed flame for titania production along the center axis at a height of 3.3 em above the burner.The corresponding TEM micrograph of titania particles is shown in Figure 1 Ob.An average particle size of 42 nm and a standard deviation of 11 nm are obtained by image ,. analysis.Thermophoretic sampling at different flame heights is perfect for determining the particle growth evolution in this flame (Figure lOc).

Modeling Particle Dynamics
There is a strong industrial interest in the development of simulators for the aerosol manufacture of nanoparticles.More specifically, mathematical models relating the characteristics of the product powder (size, polydispersity, specific surface area, crystallinity and morphology) to the process variables (reactant state, composition and flow rate as well as reactor geometry) are needed that are based on a sound understanding of particle formation and growth.This is best accomplished by interfacing models for computational fluid dynamics with models for particle dynamics, assuming that the chemistry is fast as it typically is with high-temperature processes dominated by coagulation.
The construction of such a simulator starts with the velocity and temperature profiles in the reactor that can be readily calculated even by commercial software such as CFX or Fluent.
Regarding particle dynamics, moment and sectional representations of the size distribution are employed depending on the required product particle specifications.Moment models make assumptions about the shape of the particle size distribution, allowing the population balance equation to be converted to ordinary differential equations [27).Lognormal models, for instance, rely on the assumption that the particle size distribution has a lognormal shape throughout the entire system.In sectional models, the particle size distribution is divided into sections in which the characteristics of the particles are described by average values [27).Typically, the specific surface area (SSA) or an average particle diameter (dp) and polydispersity are key characteristics of the product powder.Assuming that particles are monodisperse further simplifies the population balance equation, making monodisperse models quite attractive for fast calculation of the specific surface area and average particle diameter.For polydispersity, however, either a sectional or lognormal moment model is required.Typically, monodisperse models offer computational ~im plicity while sectional ones provide detail.Models that distinguish between primary and aggregate particles are attractive such as the one by Kruis et al. [73) that has been widely used.While early models for two-dimensional size (volume-area) distributions [80) required substantial computational time on a super-computer, recently developed computer codes for sectional models accounting for coagulation and sintering giving both primary and aggregate particle size distribution require far less computational effort and can be run on a personal computer without loss of accuracy [79).An introduction to particle dynamics models, discussing the advantages and disadvantages of different approaches and giving numerous literature references on the topic, has been recently published by Kodas and Hampden-Smith [27).
The availability of these simulators can break the Edisonian cycle for the process design of aerosol reactor units and can lead to optimal reactor design and operation, because it is possible to achieve better process control for existing products since the models make apparent which process variables most effectively determine the powder characteristics.Second, a simulator can accelerate scale-up for the manufacture of new products as it provides a much better starting point for selection of process conditions.Third, a simulator assists in minimizing the down time of existing production lines as it can reveal process conditions prone to particle deposition on reactor walls.Fourth, simulators can point out conditions for better process yields that translate in money and energy savings.Finally, simulators can assist in the training of new scientists and engineers for research and development as well as engineers and reactor operators at reaction sites, thus contributing to safety and environmental compliance.
Running these simulators can be quite revealing as they can identify problematic regions in the reactors involving either hot spots or dead volumes.As the volume fraction of particles in the suspension is typically up to 0.001, the fluid and particle dynamics can be separated.If chemical reactions are fast, as is usually the case with high-temperature processes, they can be neglected and the particle dynamics can be superimposed on the velocity and temperature profiles calculated by the commercial computational fluid dynamics simulators.This was successfully done for the synthesis of titania in hot-wall reactors [10), in flame synthesis of alumina [71) and titania [57) using a simple monodisperse model for aggregate dynamics by coagulation and sintering [73).Furthermore, the synthesis of Pd metal nanoparticles by evaporation-condensation in jet flow condensers has been simulated by interfacing Fluent with a monodisperse model for particle dynamics [81).These models [10) have been quite successful in industry, as was reported by A Gutsch [82), since they predicted the specific surface area of the product powder within 3%.
The early success of simulators has motivated research in the further development of diagnostic tools for verification as well as more accurate models that account for the full size distribution and the detailed chemistry.This, however, dramatically increases computational demands, motivating research for the efficient design of algorithms for these simulators.Nevertheless, the availability of data describing the detailed evolution of the particle size distribution by thermophoretic sampling, and computerized image analysis coupled with accurate temperature and velocity measurements creates the foundation for rigorous testing of models and hypotheses.

Concluding Remarks
Flame aerosol synthesis is a versatile technology to manufacture nanoparticles with well-defined properties.Reactant gas mixing, electric charges or dopants can be used to control the size, morphology, crystallinity and phase composition of product powders.Many diagnostic tools used in combustion research can be adapted for aerosol flame reactors, yielding important information for the understanding of particle formation and growth processes in flames.These data describing the flame environment and the stages of particle synthesis are also of great value for the validation of particle dynamic models.Recent studies show that the combination of flame diagnostics, computational fluid dynamics and particle dynamic models is an effective way to improve the understanding of flame synthesis of nanoparticles and to break the cycle of Edisonian research in this field.Gas-phase combustion technology produces some of the cheapest ceramic powders today and has high potential to provide tailor-made nanoparticles for new and commercially viable applications.Having a low-cost starting material (nanoparticles) will certainly accelerate the development of nanotechnology in the future.and Advanced Fuel Co.Some of the highlights of Prof. Pratsinis' research include the measurement of a fundamental reaction rate for oxidation of TiC1 4 vapor for synthesis of titania powder.He and his students developed the first simulator for manufacture of optical fiber preforms that is used now by Lucent Technologies (former AT&T Bell Labs) in manufacture of optical fiber preforms for telecommunications.Prof. Pratsinis and his students also developed, for the first time, algorithms for agglomerate formation and growth relating product particle characteristics to material properties and process conditions through rigorous population balances in particle mass and surface area accounting for gas phase reaction and sintering.Recently he has focused on synthesis of nanoparticles where he has shown how to precisely control the size and structure of these particles from perfectly spherical ones to highly ramified aggregates.Prof.

Fig. 6 Fig. 7 Fig. 8
Fig. 6 Influence of oxidant composition on the morphology of flame-synthesized titania particles by TiCL.oxidation, using: a) pure oxygen, b) 50% nitrogen, c) air as oxidant.Pure oxygen accelerates combustion leading to the highest tempera• ture and synthesis of perfectly spherical non-aggregated particles.Adapted from Zhu and Pratsinis [59].

Fig. 9
Fig.9Normalized radiance spectra of a premixed flame for titania nanoparticle synthesis recorded 5 mm above the burner mouth.The Planck function (blackbody) spectrum (smooth line envelopes) that matches the normalized radiance is also shown.Adapted from Arabi-Katbi[77].

Fig. 10
Fig. 10 a), b), c): Thermophoretic sampling in a premixed flame for titania synthesis.The tip of the thermophore tic sampler at a height of 3.3 em from the burner mouth is presented in Fig. lOa).Fig. lOb) shows a TEM micrograph of particles sampled at this height on the center axis of the flame.The evolution of particle size in the flame is shown in Fig. lOc) together with the centerline temperature profile of the flame.[Pictures courtesy of H.K. Kammler, ETH Zurich.]

I
Author's short biography I 182 Karsten Wegner Karsten Wegner graduated in Process Engineering at the University of Karlsruhe, Germany, in 1998, majoring in thermal and mechanical process technology.After carrying out his Diploma thesis on zeolite membranes for isomer separation at the University of Cincinnati, USA, he started working on a Ph.D project in Prof. Pratsinis' group for fine particle research at the Swiss Federal Institute of Technology in Zurich.His project deals with the synthesis of metal and ceramic nanoparticles in hot wall and flame reactors.The emphasis of the work is placed on formation of metal oxide nanoparticles in aerosol flame reactors.Sotiris E. Pratsinis Prof. Pratsinis received his Diploma in Chemical Engineering (1977) from the Aristotle University ofThessaloniki, Greece and his M.S. (1982) and Ph.D. (1985) from University of California, Los Angeles, USA He joined the faculty of Chemical Engineering at the University of Cincinnati as Assistant Professor in 1985.There he was promoted to the rank of Associate Professor with tenure in 1989 and to the rank of Professor in 1994 and served as Interim Head until 1998.Then he was elected Professor of Process Engineering at the Department of Mechanical and Process Engineering at ETH Zurich where he is leading a research program on Particle Technology focusing on synthesis and processing of fine powders (http:/ /www.ivuk.ethz.ch/staff/pratsinis/).He has published 100+ refereed papers, book chapters and three patents licensed to Dow Chemical and Hosokawa.His program has been funded largely by the U.S. National Science Foundation, Swiss National Science Foundation and ETH-Projekte as well as by DuPont, Dow Chemical, WR.Grace, ICI-Tioxide [England], Degussa and Bayer [Germany], Royal Gist Brocades [Netherlands], Genencor, Procter & Gamble, Particle Technology Inc.
Pratsinis' contributions have been recognized by the 1988 Kenneth T Whitby Award of the American Association for Aerosol Research, the 1989 Presidential Young Investigator Award by the U.S. National Science Foundation and the 1995 Marian Smoluchowski Award by the Gesellschaft fur Aerosolforschung for his research on Aerosol Reactor Engineering.Since 1991 he is on the Editorial Board of the journal of Aerosol Science, since 1998 in the Editorial Board of the journal of Powder Technology, since 1994 a consultant to the International Fine Particle Research Institute on Particle Formation and since 1998 Associate Editor of the journal of Nanoparticle Research.Prof. Pratsinis has received teaching awards at the University of Cincinnati and has taught courses at the Technical University of Delft in the Netherlands, at the University of Queensland in Australia, at the University of Karlsruhe in Germany, at the University of New Mexico, USA He has given also a number of short courses to industrial and academic audiences.Currently he is teaching Mass Transfer, Particle Technology and Particulate Processes at ETH Zurich in Switzerland.