Control of Particle Morphology and Size in Vapor-Phase Synthesis of Titania, Silica and Alumina Nanoparticles

Previous studies on vapor-phase synthesis of titania (TiO2), silica (SiO2), and alumina (Al2O3) nanoparticles were reviewed. Interactions between physicochemical phenomena involved in the particle growth and the operating variables were investigated. Strategies to produce non-aggregated spherical particles of the metal oxides are discussed. Model predictions based on the sintering laws without any adjustments in the sintering parameters rarely agreed with experimental data. There remains more to be understood in reaction and nucleation kinetics, sintering, and fragmentation mechanisms until the technology is developed to the stage of designing reactors for mass production of non-aggregated spherical particles for titania, silica and alumina


Introduction
Titania (TiO 2 ), silica (SiO 2 ) and alumina (Al 2 O 3 ) nanoparticles have been in commercial production for decades.Titania nanoparticles are used as sun blocking agents and photocatalysts.Silica nanoparticles are used as reinforcing fillers for rubbers and plastics, and ingredients for polishing slurries for semiconductor substrates.Alumina nanoparticles are used as catalyst supports for high-temperature reaction and ingredients for polishing slurries for semiconductor substrates.Theses nanoparticles are largely produced by vapor-phase oxidation of metal chloride vapors in flames, as aggregates composed of primary particles, a few to a few tens of nanometer in diameter.The vapor-phase synthesis is different from the spray pyrolysis: precursors are fed as vapor for the vaporphase synthesis and as liquid droplets for the spray pyrolysis.The flame oxidation of metal chlorides dates back to 1940 when a fumed silica production was commercialized in Germany.Later in 1951, a plant for producing titania particles for white pigments by flame oxidation of TiCl 4 vapor was come on stream.The pigment-grade titania is larger in size, 0.2-0.3μm because the hiding power is low if the size is too small.Nano-sized titania particles can be obtained by controlling the operating variables of the flame oxidation of TiCl 4 .
It was not until 1980s that control of particle morphology and size for the vapor-phase synthesis of the oxides was studied in a systematic way.Tables 1, 2, and 3 show the selected studies on the investigation of morphology and particle size for titania, silica, and alumina, respectively.Not only metal chlorides but other precursors were used.Fig. 1 shows the pathway from the chemical reaction of a precursor to the formation and growth of particles.The first step is a generation of product clusters by chemical reaction.Then, nuclei are formed from the clusters.The nuclei grow to particles by coagulation, condensation, surface reaction, coalescence between neighboring particles by sintering, and agglomeration.Flame and hot wall reactors were used at various temperatures and precursor concentrations.Fig. 2 shows the types of reactors used in previous studies.In many cases, particles were not dispersed but aggregated or agglomerated.Aggregate is defined as a group of particles joined together strongly with necking between particles, while agglomerate is a group of particles or aggregates joined loosely by van der Walls force.
A number of review papers have been published.Gurav et al. (1993) reviewed the generation of particles by gasphase routes for metals, oxides, and non-oxide ceramics.This may be the first comprehensive review on the vaporphase synthesis of particles for diverse materials, although pioneering articles were published earlier by Ulrich and his coworkers on the growth of silica particles (Ulrich, 1971;Ulrich and Subramanian, 1977;Ulrich and Riehl, 1982;Ulrich, 1984).A few years later, flame synthesis of nanoparticles for ceramic materials including titania, silica, and alumina was reviewed with emphasis on the fundamentals of particle formation and growth.(Pratsinis, 1998) Other reviews followed, most of which deal with synthesis techniques, application fields, and characterization of the particles, with rather shallow discussion on the particle formation and growth (Swihart, 2003;Strobel et al., 2006;Strobel and Pratsinis, 2007;Biskos et al., 2008;Athanassiou et al., 2010).The fundamentals of particle formation for single and mixed oxides were reviewed by Roth (2007) in a way similar to that by Pratsinis (1998) but with a different viewpoint on the importance of the vapor-phase reaction and nucleation kinetics.Recently, the history of the development of commercial aerosol reactors for synthesis of carbon black, titania, silica, zinc oxide, nickel particles and the effects of process variables on particle morphology and size were discussed to conclude that for most of those materials the coagulation washes out the effect of nucleation because precursors are rapidly consumed and product particles attain self-preserving distribution by coagulation.(Pratsinis, 2011) The present review was confined to the control of particle morphology and size in vapor-phase synthesis of titania, silica and alumina nanoparticles in flame and hot wall DF: diffusion; PM: premixed 1) Estimated using precursor concentration given in mol/L or precursor flow rate.
2) Estimated using precursor flow rate in the center tube.
2) Estimated using precursor flow rate in the center tube.
3) Silane volume loading DF: diffusion; PM: premixed 1) Estimated assuming that the carrier gas was saturated with the precursor.
2) Particle volume fraction reactors.In previous reviews, little attention was paid to the chemical reaction and nucleation kinetics assuming that the reaction occurs rapidly in one step to form nuclei of oxide-monomer size.In reality, the reaction may not occur in one step, but in multiple steps to produce intermediate products significantly different in physical properties from the final oxide products.The chemical reaction and nucleation kinetics were reviewed more extensively than in others.We start by looking at the physicochemical phenomena involved in the particle generation and growth: chemical reaction, nucleation, condensation, coagulation, sintering, restructuring and fragmentation.We then discuss the effects of operating variables on particle morphology and size.Finally, control strategy of particle size and morphology is discussed with future works to be challenged.

Chemical reaction
The first step to the particle generation of a metal oxide is the homogeneous reaction of a precursor vapor by oxidation, hydrolysis, or decomposition.The reaction usually occurs in multiple steps.The oxidation of TiCl 4 was reported to occur through step-wise elimination of chlorine and addition of oxygen to form titanium oxy-chlorides, TiO x Cl y .The oxy-chlorides then collide with each other to form oligomers, (TiO x Cl y ) n , which are further oxidized toward pure titania.The reaction is homogeneous before particles are generated.In the later stage, a heterogeneous reaction or surface reaction may occur simultaneously on the particles that have been formed.The surface reaction contributes to the growth of particles.
Table 4 shows reaction kinetic parameters reported for vapor-phase synthesis of titania, silica and alumina.
Power-law type expressions were proposed for the overall reaction of TiCl 4 with oxygen.(Pratsinis et al., 1990;Kobata et al., 1991) In those rate expressions, however, the homogeneous and heterogeneous reactions were not differentiated but combined.The surface reaction rate expression for the TiCl 4 oxidation is available elsewhere.(Ghoshtagore, 1970) In the flame synthesis of titania, not only the oxidation but also the hydrolysis by the water vapor generated from fuel combustion is involved.The competition between oxidation and hydrolysis was studied in a tubular reactor to conclude that the hydrolysis was important, particularly at high H 2 O/TiCl 4 ratios.(Akhtar et al., 1994) However, no kinetic data on the sole hydrolysis of TiCl 4 are available, although the rate constants for the initial few steps of the hydrolysis were calculated using the transition-state and RRKM theories (Wang et al, 2010).Titanium tetraisopropoxide (TTIP) was used in some studies in place of TiCl 4 .TTIP is thermally decomposed into TiO 2 , hydrocarbons and water.As in the oxidation of TiCl 4 , the decomposition of TTIP may proceed in multiple steps by successive β-hydride elimination of propene and polymerization (Park et al., 2001).For the over- all reaction represented by TTIP = TiO 2 + 4C 3 H 6 + H 2 O, the reaction rates were determined from measured concentrations of propylene under varying operating condition to obtain a rate equation, 164 kJ/mol in activation energy and a half order with respect to TTIP concentration.(Kanai et al., 1985) A different rate equation was reported later in which the reaction order is one and the activation energy is 70.5 kJ/mol.(Okuyama et al., 1990) The hydrolysis rate of TTIP is known to be faster than that of the thermal decomposition due to the oxidation effect of water that leads to a considerable reduction in activation energy.(Seto et al.,1995)These rate equations for TTIP decomposition do not differentiate between homogeneous and heterogeneous reactions.While, a rate constant assigned to the surface reaction of TTIP was obtained by Tsantilis et al. (2002) by using a reaction model and the experimental data of Battiston et al. (1997).Zhang and Griffin (1995) studied the growth kinetics of a TiO 2 film under the decomposition of TTIP in nitrogen and proposed a rate expression, 4.0 × 10 11 exp (-40[kJ SiCl 4 , SiBr 4 , tetraethoxysilane (TEOS), and hexamethyldisiloxane (HMDS) were used as precursor for the vapor-phase synthesis of silica particles.In the oxidation of SiCl 4 vapor with oxygen, pure silica was not formed below 1200 K; instead chlorosiloxanes containing up to 50 Si atoms per molecule were formed.(Giesenberg et al., 2004) This implies that the oxidation of SiCl 4 to silica is a multi-step reaction.The mechanism of the multi-step reaction is not available in the literature, but is presumed to follow the step-wise elimination of chlorine and addition of oxygen to form (SiO x Cl y ) n , as in the oxidation of TiCl 4 mentioned in the beginning of this section.The oxidation rate of SiCl 4 was studied at 1100-1300 °C.(Powers, 1978) The reaction was first order in SiCl 4 and zero order in oxygen up to a 20-fold excess O 2 , and the activation energy was 96 kcal/mol.In an independent study (French et al, 1978), the reaction order was one and the activation energy was 98 kcal/mol.The rate constant of the hydrolysis of SiCl 4 vapor was measured to be 10 12 exp (-121.4/RT)cm 3 mol -1 s -1 , where the activation energy is expressed in kJ/mol.(Kochubei, 1997) Ignatov et al. (2003) studied the reaction mechanism of the gas-phase hydrolysis of SiCl 4 and calculated the rate constants of the elementary reactions in the early stage using quantum mechanics.A similar kinetic study was carried out by Hannebauer and Menzel (2003).A power-law type expression for the thermal decomposition of TEOS was proposed as 7.4 × 10 10 exp (-49500/RT) s -1 .(Chu et al., 1991) The surface reaction rates of TEOS decomposition were measured at temperatures between 923 K and 1023 K, from which a rate constant was obtained with an activation energy of 42 kcal/mol.(Satake et al., 1996) For eighteen elementary reactions composing the decomposition of TEOS, the kinetic parameters were obtained by parameter estimation to fit experimental data.(Shekar et al., 2012a) AlCl 3 , aluminum triisopropoxide (ATI), trimethyaluminum (TMA), and aluminum tri-sec-butoxide (ATSB) were used as precursor for alumina particles.Kinetic data for the alumina precursors are very rare.Recently, a rate expression for the hydrolysis of AlCl 3 vapor was reported.(Park, 2014)

Nucleation
Nucleation occurs when the supersaturation ratio of a species produced in the reaction system has increased past a critical value.The nucleus size for silica by flame oxidation of SiCl 4 was calculated to be smaller than a SiO 2 monomer using the classical nucleation theory.(Ulrich, 1971) In many studies that follow, the assumption of a product monomer being a nucleus was used in the interpretation of experimental results and in the modeling of particle growth.(Landgrebe and Pratsinis, 1989;Tsantilis and Pratsinis, 2000;Heine and Pratsinis, 2006) The surface tension of the nucleus is required to calculate the nucleus size by the classical theory.However, the surface tension is not well defined for such a small object and the use of the classical nucleation theory does not make sense.The atomistic theory developed by Hoare and Pal (1971) may be better suited for systems with clusters less than 100 molecules.(Kulmala et al., 1987) In some studies, regardless of the nucleus size, the particle size distribution approached the self-preserving size distribution.(Vemury et al., 1994;Spicer et al., 2002) In other studies, however, the nucleus size and the nucleation rate were reported to affect the final particle size distribution.(Suyama and Kato, 1976;Yoon et al., 2003, Park et al., 2014) Using Ti 5 O 6 C l8 , an intermediate on the transition to TiO 2 , as nucleus in place of TiO 2 monomer resulted in a considerably different particle size distribution in a flame synthesis of titania.(Mehta et al., 2013) At present, little data are available on the kinetics of producing the intermediate species and their physical properties including vapor pressure and surface tension that are required for nucleation calculation.

Condensation and coagulation
Once nuclei or smallest particles are formed, they collide with each other to form larger particles.The clusters condense on the particles while the coagulation between particles proceeds by Brownian motion.In the classical theory of coagulation, coalescence occurs instantaneously after two particles collide, and a new sphere is formed.The instantaneous coalescence may not be valid for metal oxides depending upon operating conditions.If the coalescence is incomplete, particles grow as agglomerates comprised of primary particles.Ulrich and Subramanian (1977) first introduced a time for complete fusion of two SiO 2 particles in contact to calculate the average number of primary particles in an agglomerate.The transition of fully coalesced particles into agglomerated particles was predicted using relative values of the collision and coalescence times.(Windeler et al., 1997b) The collision rates of the agglomerates are faster than those of spherical particles because the collision cross sections are larger.The collision cross section depends on the shapes of the agglomerates.The shape of an agglomerate is defined as a function of the fractal dimension (D f ) which is correlated to the number of primary particles (N p ) in the agglomerate , where A is a constant prefactor, R is the radius of the agglomerate and R 0 is the radius of the primary particle.(Forrest and Witten, 1979) The coagulation rate expressions for spherical particles and agglomerates are available elsewhere.(Friedlander, 2000) The primary particle size increases with increasing condensation and coagulation, but is not affected by agglomeration.

Sintering, restructuring and fragmentation
If the temperature is high enough, the primary particles composing an agglomerate are sintered to form necks between particles in contact, resulting in a decrease of the surface area.Ulrich and his coworkers (Ulrich and Subramanian, 1977;Ulrich and Riehl, 1982) reported that in the flame synthesis of silica particles the coalescence of primary particles by sintering played a dominant role in determining the primary particle size.Seto et al. (1995) showed that the titania particles obtained from thermal decomposition of TTIP started to change in primary particle and agglomerate sizes by sintering at 1000 K.They showed in a later study (Seto et al., 1997) that the temperatures effective for the coalescence of the particles by sintering were 50-70 % of the bulk melting temperature for titania and 90-100 % for silica.The coalescence of primary particles by sintering was verified in other studies.(Xing et al., 1997;Cho and Choi, 2000) Koch and Friedlander (1990) incorporated into the general dynamics equation a sintering term represented by where a is the instantaneous surface area of an agglomerate, a f is the surface area of the sphere resulted by the full coalescence of the primary particles in that agglomerate, t is the time, and τ is the characteristic time for sintering.The characteristic time is defined as the time necessary to reduce by 63 % the excess surface area.Studies followed to obtain characteristic times for titania and silica, as listed in Table 5.The presence of water vapor in the sintering process was reported to accelerate the kinetics of the coalescence.(Goertz et al., 2011) The sizes of the primary particles and the agglomerates exiting the reactor were calculated using the characteristic times, and compared with experimental data.In a heated wall reactor where titania particles were produced by reaction of TiCl 4 vapor with oxygen, the maximum primary d p is the particle diameter in cm, d p,min is the particle diameter below which the sintering is instantaneous, and T is the temperature in K.
particle size was calculated to be 15 nm using the grain boundary diffusion coefficient by Astier and Vergnon (1976), whereas experimentally observed size was 55-65 nm.(Kobata et al., 1991) In a different study by Xiong et al. (1993b), significant deviations were observed for titania and silica between calculated and experimental sizes of primary particles and agglomerates.Wu et al. (1993) produced alumina particles by oxidation of trimethylaluminum (TMA) in flames at 890-1140 K.They compared experimental data with model prediction to show that the primary particle sizes calculated based on either grain boundary diffusion or volume diffusion were smaller by several orders of magnitudes than those experimentally observed.In contrast, Windeler et al. (1997b) reported that good agreement was obtained between experimental and calculated primary particle sizes for titania and silica produced in a premixed flame.In their calculation, the sintering law of Koch and Friedlander (1990) was assumed to hold for groups of primary particles in contact, not the whole agglomerate, as proposed by Lehtinen et al. (1996).Tsantilis et al. (2001) reported that experimental data were better fitted by modifying the sintering time to take into account the particle size dependence of melting temperature.In another study, a melting diameter was introduced, which is defined as half the diameter of particles whose characteristic sintering time is identical to the reactor residence time.Particles smaller than the melting diameter were assumed to coalesce instantaneously upon collision, by which the prediction at low temperatures was improved.(Park and Rogak, 2003) The agglomerates are restructured by sintering.Not only the primary particle size increases but also the agglomerates shrink to decrease the cross sectional area.Fig. 3 illustrates the restructuring of an agglomerate.(Eggersdorfer and Pratsinis, 2014) The size of an agglomerate decreases by restructuring and fusion until the agglomerate is converted eventually into a sphere.The agglomerate sintering may result in a fragmentation to reduce the agglomerate size.The fragmentation was studied theoretically (Irisawa et al., 1995;Lando et al., 2006;Thouy et al., 1997;Hawa and Zachariah, 2007) and experimentally verified for silver (Bréchignac et al., 2002), but not yet for titania, silica and alumina.The fragmentation may result in producing smaller primary particles.

Precursor type and concentration
TiCl 4 and TTIP were used as precursor for titania particles.SiCl 4 , SiH 4 , TEOS, HMDS, and octamethylcyclotetrasiloxane (OMCTS) were used for silica particles and AlCl 3 , ATSB, TMA and ATI for alumina particles.The properties of the precursors are shown in Table 6.Under the same precursor concentration in a hot wall reactor operated at temperatures between 800 °C and 1600 °C, both the primary particles and the agglomerates of titania were larger with thermal decomposition of TTIP in nitrogen than with oxidation of TiCl 4 in a mixture of nitrogen and oxygen.(Nakaso et al., 2001) Titania particles from TTIP decomposition were spherical, while those from TiCl 4 oxidation were faceted.(Yoon et al., 2003) The faceted particles from TiCl 4 turned spherical in the presence of water.(Akhtar et al., 1994) The crystalline structure of the titania from TiCl 4 in oxygen was a mixture of anatase and rutile (Kato and Suyama, 1974); the rutile content was higher in H 2 O than in oxygen.While, the crystalline structure of the titania from TTIP was anatase in oxygen and a mixture of anatase and rutile in water vapor.(Choi and Park, 2006) Operating temperature is an important factor in controlling the ratio of anatase to rutile, as discussed in the succeeding section.Three silica precursors, SiCl 4 , HMDS, and OMCTS, were compared in a diffusion flame to study the effect of precursor on the primary particle size.(Briesen et al., 1998) At the same precursor concentration, the particle size was the largest with OMCTS and the smallest with SiCl 4 .The larger particle sizes with the organometallic precursors were attributed to the higher flame temperatures resulted from the combustion of the organic compounds.Ehrman et al. (1998) reported, for a precursor concentration lower by an order of magnitude, that there was little difference in primary particle size between SiCl 4 and HMDS.No crystalline silica was produced from any precursor.As produced alumina particles by flame oxidation of AlCl 3 vapor were amorphous or contained small amount of δ and θ phases, which are meta-stable.(Pratsinis, 1998) An increase in precursor concentration increases the amount of precursor available for surface reaction, the amount of condensable product clusters, and the number of nuclei.The increases in the amounts of condensable product clusters and the precursor vapors for surface reaction act to increase primary particle size, while the increase in the number of nuclei acts to decrease the size.The opposing effects compete in determining primary particle size.Which effect will dominate depends on the operating condition and the properties of the precursor.In those studies shown in Tables 1 to 3, primary particle sizes increased with increasing precursor concentration, but for a few.The rare events in which primary particle size decreased with increasing precursor concentration are described.In a synthesis of titania particles by thermal decomposition of TTIP in nitrogen at 500 °C, an increase in precursor concentration decreased the primary particle size and increased the number of particles.(Moravec et al., 2001) In a synthesis of silica particles by oxidation of SiCl 4 in oxygen at 1200 °C with precursor concentration varied between 1 % and 8 %, the primary particle size decreased with increasing precursor concentration in the range of 4-8 %. (Suyama and Kato, 1976) In a flame synthesis of alumina with TMA as precursor, the primary particle size was rather insensitive to the precursor concentration.(Windeler et al., 1997a) In contrast with the contradictory effects on primary particle size, aggregate size increased consistently as precursor concentration was increased.(Chung et al., 1992;Windeler et al., 1997a)

Temperature and residence time
The operating temperature may be the most influential factor in controlling particle morphology and size.An in- crease in temperature will increase the reaction rate, the nucleation rate, the collision rate and the sintering rate.
The temperature profile in a reactor for vapor-phase synthesis of particles could be represented by one of the three types shown in Fig. 4. The figure is more representative of hot wall reactors, but is adaptable in concept to flame reactors in which the temperature gradients are usually stiffer than in hot wall reactors.In Type I the temperature is below the sintering temperature throughout the reactor and the sintering of particles is neglected.If the nucleation is fast enough, the particle sizes for both primary particles and agglomerates are determined by the collision rates to exhibit self-preserving size distributions in the long run (Friedlander and Wang, 1966;Vemury et al., 1994;Spicer et al., 2002).Otherwise, the primary particle size is determined by competition between nucleation and surface growth.Examples of Type I can be found in syntheses with hot wall reactors, in most of which primary particle size decreased with increasing reactor set temperature.(Suyama and Kato, 1976;Kirkbir and Komiyama, 1987;Morooka et al., 1989;Jang and Jeong, 1995;Park et al., 2001;Park et al., 2014) By preheating the reactants before they get in contact or placing the location of reactants mixing at a distance from the inlet toward the center of the reactor, the particle size was decreased.(Jang and Jeong, 1995;Park et al., 2014) In Type II there exists a high-temperature zone in which the temperature exceeds the sintering temperature, but the residence time in the hot-temperature zone is not sufficient, resulting in aggregates in which the boundaries between primary particles are unclear.(Ulrich and Subramanian, 1977;Ulrich and Riehl, 1982;Windeler et al., 1997a;Cho and Choi, 2000) The specific surface area decreased with increasing reactor set temperature.In Type III the residence time in the high temperature zone is long enough for the primary particles in the agglomerates to fully coalesce into larger particles (Xing et al., 1997;Arabi-Katbi et al., 2001;Lee and Choi, 2002;Camenzind et al., 2008).In most studies in Type III, the effects of reaction and nucleation rates in the early stage were washed out by the dominating sintering effect and primary particle size increased with increasing temperature.
In some studies that fall in Type III, primary particle size rather decreased with increasing temperature.In a study on the effect of sintering on the growth of TiO 2 particles in a hot wall reactor, the primary particle size increased with increasing temperature, showed a maximum at 1200 °C, and then decreased as the temperature was further increased to 1600 °C.(Nakaso et al., 2001) In a synthesis of silica particles in a counter-flow diffusion reactor with silane (SiH 4 ) as precursor, the primary particle size in the hotter flame was smaller than that in the colder flame.(Zachariah et al., 1989) Similar phenomena were observed in other flame syntheses.(Cho and Choi, 2000;Akurati et al., 2006) The decrease of the primary particle size was attributed to an insufficient sintering time (Mueller et al., 2004) or to a lower precursor conversion (Akurati et al., 2006), due to a reduction in flame size.We raise a conjecture that the agglomerates admitted into the sintering-coalescence zone were probably smaller, as reported for a synthesis of iron particles by thermal decomposition of iron pentacarbonyl (Moniruzzaman et al., 2007), or the fragmentation rate of the aggregates was higher.Further studies are necessary to clarify the reason for the decrease of primary particle size under the sinteringcoalescence regime.
Aggregate sizes or the number of primary particles composing an aggregate increased with increasing reactor temperature before sintering effects set in, but decreased thereafter due to restructuring.(Camenzind et al., 2008;Chung et al., 1992) In flame reactors, the particles were rapidly cooled to prevent aggregation and growth by sintering.Wegner and Pratsinis (2003) reported that virtually non-aggregated and spherical titania particles as small as 5 nm were produced by placing a quenching nozzle.By comparison, the particle size was 20-60 nm without the nozzle.The particle size increased with increasing the distance of the nozzle from the burner mouth.The nozzle quenching effect was confirmed later in a similar study by Okada et al. (2011) in whose work nearly non-agglomerated titania particles 6.8 nm in diameter were produced by both blowing in a -70 °C argon gas and using a expansion nozzle.
The rutile content of the titania produced by oxidation of TiCl 4 vapor was reported to increase with increasing reaction temperature.(Suyama et al., 1975) In the oxidation of TiCl 4 , initially anatase particles are formed and then grow as anatase particles or transformed into rutile particles before they grow.Not only the temperature but also the time that the particles remain small in size is known to be an important factor.Morooka et al. (1989) reported that the rutile content showed a maximum at 1100 °C and then decreased as the temperature was further increased to 1200 °C.This decrease in rutile content with temperature increase was attributed to the faster growth of the anatase particles to the size at which the transformation into rutile is remarkably slowed down.By irradiating titania particles with a CO 2 laser, the rutile content decreased against the expectation that the rutile content would increase because the temperature was increased by the laser irradiation.(Lee and Choi, 2002) The silica particles produced by the vapor-phase synthesis in flame reactors were all amorphous (Formenti et al., 1972;Ehrman et al., 1998), while the alumina particles were amorphous or a mixture of amorphous and metastable crystalline phases.(Pratsinis, 1998) By calcining an amorphous alumina for two hours at 1200 o C, the phase was completely transformed into alpha.(Yoo et al., 2009)

Reaction medium
The precursors for vapor phase synthesis of the oxides were reacted or decomposed in different media: nitrogen, oxygen, water vapor or a mixture.The morphology and particle size showed differences between reaction media.In a hot wall reactor at 1000-1400 °C, titania particles were synthesized from TiCl 4 with varying ratios of H 2 O to O 2 and with five injection points of H 2 O located at various distances from the inlet of the reactor.(Akhtar et al., 1994) At a reactor set temperature of 1200 °C, the primary particles obtained with the injection point at 13 cm inward from the inlet were faceted, while those with the injection point at the inlet were spherical.With the injection point at 13 cm, there was no H 2 O present in the early stage and the oxygen determined the morphology that is consistent with those faceted shapes observed in earlier studies in pure oxygen.(Suyama and Kato, 1976;Yoon et al., 2003) With the injection point located at the inlet, H 2 O was present from the beginning and played to change the particle shape from faceted to round.As the amount of H 2 O was increased, the rutile content was invariable up to 1000 °C but decreased at higher temperatures.In a synthesis of titania particles by thermal decomposition of TTIP at 300-700 °C, three different reaction media were used: N 2 , O 2 , and a mixture of N 2 and H 2 O. (Choi and Park, 2006) As shown in Fig. 5, particle morphology and size was significantly different between reaction media.In N 2 , agglomerates composed of very small particles whose individual boundaries are not clearly distinguished were produced.In O 2 , the primary particles were spherical and larger.As the medium was changed to nitrogen plus water, much larger spherical particles were obtained.The crystalline phases were amorphous in N 2 , anatase in O 2 , and a mixture of rutile and anatase in N 2 plus water, respectively.A follow-up study may be necessary to understand how the reaction media affect the morphology, size, and crystalline structure.
In most studies of synthesis of silica particles with SiCl 4 as precursor, the flame reaction was used in which both oxygen and water vapor were present.However, in a synthesis by Park and Park (2008), SiCl 4 was reacted with H 2 O only at 150 °C to produce spherical particles, as shown in Fig. 6, which are quite different from those chain-like aggregates resulted by the flame reactions.A similar method was used to produce non-aggregated spherical particles by hydrolysis of AlCl 3 vapor at 300-700 °C.(Park et al., 2014) Those particles obtained by the low temperature hydrolysis were not pure but contained residual chlorines which were removed downstream in a calciner.The low-temperature hydrolysis of metal chloride vapor appears to be an attractive route to non-aggregated spherical particles of metal oxides when a calcination is Fig. 5 Comparison of TEM images of titania particles between reaction media.The particles were synthesized in a tubular reactor by thermal decomposition of TTIP at 700 °C with the precursor concentration at 7 mol % and the residence time at 0.7 s. (Adapted from Choi and Park, 2006.)necessary as for the transformation of crystalline phase.

Additives and external forces
The presence of additives during vapor-phase synthesis of metal oxides is known to affect the particle morphology and size significantly.(Suyama and Kato, 1985) In the oxidation of TiCl 4 vapor in a hot wall reactor, the titania particles formed in the presence of FeCl 3 , AlBr 3 , SiCl 4 or ZrCl 4 were 1/2 to 1/5 the size of those without additives.The reduction in particle size was attributed to a retardation of TiO 2 deposition on the particles.An addition of AlCl 3 in the flame oxidation of TiCl 4 increased the rutile content.(Mezey, 1966) While, other additives such as SiCl 4 and POCl 3 decreased the rutile content.(Akhtar et al., 1992) The addition of an alkali salt, sodium or potassium carbonate, to a flame synthesis of silica reduced the primary particle size.(Wu et al., 1993) The alkali salts dissociated into ions, some of which adsorbed on the particle surface to create repulsive forces between particles to retard coagulation.(Xiong et al., 1992) Ferrocene was found to be an effective additive to decrease the primary particle size of silica particles by flame oxidation of SiCl 4 .(Fotou et al., 1995) Silica particles by flame synthesis were made smaller by applying electric fields.(Hardesty and Weinberg, 1973) The presence of a gaseous electric discharge (corona) in a flame synthesis of titania particles decreased the primary particle size and the rutile content.(Vemury and Pratsinis, 1995) The corona was applied to a silica flame synthesis; it was as effective as in titania synthesis and negative electric field was more effective in reducing particle growth.(Vemury and Pratsinis, 1996) The control of particle size by external electric field was successfully applied to a scaled up process of flame synthesis of silica particles.(Kammler and Pratsinis, 2000) In parallel with experimental studies on the effects of the operating variables, numerous modeling studies have been made on the basis of first principles.(Landgrebe and Pratsinis, 1990;Xiong and Pratsinis, 1993b;Pope and Howard, 1997;Lee et al., 2001;Nakaso et al., 2001;Park and Rogak, 2003;Morgan et al., 2006;Moniruzzaman and Park, 2006;Ji et al., 2007;West et al., 2009;Sander et al., 2009;Zaitone et al., 2009;Shekar et al., 2012b;Mehta et al., 2013).However, there remains more to be understood in reaction and nucleation kinetics, sintering, and fragmentation mechanisms until the technology is developed to the stage of designing reactors for mass production of non-aggregated spherical particles for titania, silica, and alumina.

Control strategy of particle morphology and size
As discussed in the preceding sections, various physicochemical phenomena are involved in determining particle morphology and size and those phenomena are influenced by temperature and residence time in the reactor, precursor concentration, reaction medium, and additives and external forces.Since the metal oxides in aggregated form have already been in commercial production, the control strategy in this work is focused on the synthesis of non-aggregated spherical particles varying in size.
As shown in Fig. 4, non-aggregated spherical particles can be produced either with temperature and residence time maintained high enough for any aggregates formed earlier to coalesce later by fusion into spheres or with the temperature low enough for any serious necking between particles in contact not to occur.With the low-temperature scheme, the nucleation rate is a key factor in determining primary particle size; the particle size decreases with increasing reactor temperature.Smaller particles can be produced by either preheating the reactants before they get in contact or placing the location of reactants mixing at a distance from the inlet toward the center of the reactor.An additional heat treatment or calcination may be necessary if the conversion of the precursor is incomplete because of the low temperature.With the high-temperature scheme, sintering of particles in agglomerates is a dominant factor.It is a key how to control operating variables to obtain a high-temperature zone for full coalescence of the particles into spheres.
Increasing precursor concentration results in an increase in particle size except for very limited cases.Reaction medium is a factor in controlling particle morphology and size in hot wall reactors.In flame reactors, the control of particle size and morphology by reaction medium is less useful because water vapor is always present in the flame.The particle size and morphology is sensitive to the burner type, premixed or diffusion, and to the configuration of injection nozzles for fuel and oxidizer in diffusion burners.(Akurati et al., 2006) The particle size can also be controlled either by injecting additives into the reactor or by applying electric fields.The use of additives may contaminate the particles, and their uses are limited to the cases where the contamination poses no problem in product specification as in the use of AlCl 3 to control the phase of titania particles.Cooling the particles rapidly before the sintering sets in is a way to obtain small non-agglomerated particles.One needs to be cautious, however, not to freeze the conversion of precursor when using rapid cooling.
The choice of a precursor is based on the volatility, the stoichiometric mass of product per mass of precursor, and the cost of the precursor.Once a precursor is chosen, the next step is to decide reactor type, hot wall or flame reactor.The control of operating variables is simpler with hot wall reactors.In flame reactors, there exist interactions between operating variables to make it difficult to decouple the effect of a variable from those of other variables and to achieve optimal operating conditions.However, there is an advantage with flame reactor that the energy needed for the synthesis of particles is supplied in situ by combustion of fuels.

Conclusion
Previous studies on the synthesis of titania, silica, and alumina particles were reviewed with an objective to derive strategies for producing non-aggregated spherical particles in hot wall or flame reactors.Various physicochemical phenomena are involved in the particle formation and growth: chemical reaction, nucleation, condensation, coagulation, sintering, restructuring, and fragmentation.These phenomena occur in series and par-allel as well, and are influenced and controlled by various operating variables: precursor type and concentration, temperature and residence time, reaction medium, and additives and external forces.By investigating the interactions between the physico-chemical phenomena and the operating variables, two approaches were derived to produce non-aggregated spherical particles of the metal oxides.One approach is to have the temperature and residence time high enough for any aggregates formed in the earlier stage to fuse and coalesce into spherical particles before they exit the reactor.The other is to have the temperature low enough for any necking between neighboring particles not to occur.If the conversion is not enough with the low-temperature approach, an increase in residence time or an extra step of calcination may be necessary in order to complete the conversion and to remove any residuals in the particles.
The configuration of reactants mixing is an important factor in controlling the morphology and size.By preheating the reactants before they get in contact or placing the location of reactants mixing at a distance from the inlet toward the center of the reactor, the particle size can be decreased particularly for the low-temperature synthesis.In flame reactors, particle size and morphology are sensitive to the burner type, premixed or diffusion, and to the configuration of injection nozzles for fuel and oxidizer in diffusion burners.The particle size increases with precursor concentration except for limited cases.By using additives, applying electric fields or rapid cooling with a quenching nozzle, both the primary particle size and the degree of aggregation decreased significantly.
It appears that the role of nucleation has been underestimated.The conversion of precursor needs to be checked more carefully in the analysis of early-stage particle growth in flame reactors.Model predictions based on the sintering laws without any adjustments in the sintering parameters rarely agreed with experimental data.There remains more to be understood in reaction and nucleation kinetics, sintering, and fragmentation mechanisms until the technology is developed to the stage of designing reactors for mass production of non-aggregated spherical particles for titania, silica, and alumina.

Fig. 1
Fig. 1 Pathways for particle formation and growth in vapor-phase synthesis of metal oxides.

Fig. 2
Fig. 2 Schematic drawings of reactors used for vapor-phase synthesis of metal oxides.

Fig. 3
Fig. 3 Illustration of the evolution of an agglomerate toward a sphere by sintering.D f represents fractal dimension.The lines showing the variation of r m and r va with residence time or temperature were taken from the work of Seto et al. (1997).r m represents the average of the mobility equivalent diameters of the aggregates and r va represents the average diameter of the primary particles composing the aggregates.(AdaptedfromEggersdorfer and Pratsinis, 2014.)

Fig. 4
Fig. 4 Schematic drawing of axial temperature profiles for producing aggregated particles or loosely agglomerated spherical particles.Three types of temperature profile are shown.Loosely agglomerated particles are produced with types I and III, while aggregated particles are produced with Type II.

Fig. 6
Fig. 6 Scanning electron microscopic images of silica particles produced at two different SiCl 4 concentrations: (a) 10 mol % and (b) 20 mol %.The silica particles were produced in a batch reactor at 150 °C by hydrolysis of SiCl 4 vapor.The residence time in the reactor was 60 s and the initial molar ratio of H 2 O to SiCl 4 was 2.(Adapted fromPark and Park, 2008.)

Table 1
Selected studies on vapor-phase synthesis of titania particles

Table 2
Selected studies on vapor-phase synthesis of silica particles

Table 3
Selected studies on vapor-phase synthesis of alumina particles

Table 5
Expressions for sintering characteristic time Kirchhof et al. (2012) a f : radius of the completely coalesced sphere, mKirchhof et al. (2012)

Table 6
Properties of precursors for vapor-phase synthesis of titania, silica and alumina particles