Formation of Ti02Aerosol froin the Combustion Supported Reaction of TiC14and O2 BY A. P. GEORGE AND E. R. PLACE R. D. MURLEY Tioxide International Central Laboratories Portack Lane St ockt on-on-Tees Teeside Receioed 18th December 1972 The formation of particulate TiO has been studied by the addition of small quantities (10-5-10-3 mol fraction) of Ti& vapour to a lean CO +OL+Nz flame with a maximum temperature of about 1400°C. Measurements of TiCI concentration have been made as a function of height (residence time) by u,-v. absorption spectroscopy. The results demonstrate that chemical reaction is essentially complete 50 ms down stream of the CO flame-front at which stage the TiOz particles have reached a diameter of 410 A. Electron microscopic examination of samples of material from the flame shows that particle growth continues for a further 200ms by a flocculation mechanism.This is a major factor determining the final particle size (630 A). Agreement with theoretical flocculation predictions is reasonable both with respect to the development of the mean size and the size distribution. Results of sintering experiments carried out in the flame and of similar measurements carried out in the hot stage of an electron microscope demonstrate that the particles produced in this system exhibit a fusion temperature much below that of the bulk solid (1850°C). The occurrence of sintering in the flame is necessary to account for the form of the TiO particles produced in this system. The high-temperature oxidation of gaseous TiCI4 according to reaction (1) forms the basis of an industrial process for the production of pigmentary TiO TiC14+O,+TiO,(s) +2C1 (AH = -43.4 kcal/mol).(1) The pigmentary properties of the Ti0 are related directly to particle size and size distribution. Consequently it is important to understand the processes that determine these size characteristics. The relevant processes contributing to the final size are nucleation growth by chemical reaction and growth by flocculation. Al-though in principle the theory behind these processes is well understood there is little practical evidence to confirm the behaviour of high-temperature high-concentration small-particle aerosol systems in which chemical reaction is occurring. The present study provides some practical information on these factors which although obtained specifically for the TiCI4 +0 reaction system are relevant to other aerosol systems in similar regimes.The oxidation reaction only occurs at an appreciable rate at temperatures in excess of about 1000°C. Although the reaction is exothermic it does not become self- supporting in the manner of a combustion reaction. In order to establish a system amenable to study it was desirable to avoid the complications inherent in preheating and mixing the reactants. This was achieved by utilizing a lean flat laminar CO+O flame as the source of heat. In this way premixed TiCI4 was reacted with excess O2 in an essentially plug-flow system allowing residence time to be simply related to position.The use of CO as a fuel has the advantages that the flame pro- duces no water or ionization and that the products are relatively inert. Water has a pronounced effect on the reaction and the generation of charged species could affect all stages of the particle formation process. 63 FORMATION OF TiOz AEROSOL EXPERIMENTAL The complete gas delivery and burner system is shown diagramatically in fig. 1 and details of the burner are shown in fig. 2. The burner consists of an hexagonal array of 271 stainless steel hypodermic tubes 0.050" o.d. 0.006" wall thickness the outer three rows 144 in total providing the sheath flame. The upper burner body is cooled by transformer oil circulating from a water-cooled heat exchanger. Flame stability is improved by the provision of a stainless steel gauze 2 cm above the burner mouth.The flame temperature in this system can be controlled independently of the mixture composition by varying the total flowrate. All the measurements described refer to the FIG. 1.-Schematic diagram of burner system 1 2 3 Carbon monoxide oxygen and nitrogen source; 4 5 6 drying tubes; 7 8 9 10 11 12 flowmeters; 13 TiCI4 evaporator; 14 ballotini- packed water condenser ; 15 mixing vessel ; 16 absorption vessel ; 17 burner ; 18 coolant inlet/ outlet ; 19 stabilizing gauze ; 20 vent. FIG.2.-Details of multiorifice buruer. A. P. GEORGE R. D. MURLEY AND E. R. PLACE following flame conditions molar ratio C0/02/N2 = 1/0.95/1.1 ; flow to inner burner = 1 1. min-' ; flow to sheath burner = 3,3 1.min-' ; maximum flame temperature = 1650' K. Additioning of the inner flame gases with TiC14 is achieved using a by-pass system metered fractions of the gases passing through a TiCI4 saturator. The two gas streams then re-unite and flow to the burner via a mixing tube containing ballotini and glw wool plugs where TiCI hydrolysis products formed by reaction with residual water vapour in the gases are removed. The concentration of TiCl4 in the feed gases is determined at the commence- meat and end of an experiment by passing them through an absorption vessel containing tetrachloroethylene for a known period of time. This effects total extraction of TiC14 the concentration of which is subsequently determined colorimetrically in aqueous phase as the peroxo complex.This level of TiC14 is reduced with respect to the flame asa consequenceof losses incurred at the burner face where surface growth of Ti02 takes place irrespective of flame conditions or TiC1 addition level. These losses were determined by mass balance the solids produced in the flame king collected on a glass-fibre filter pad. As a check on completeness of reaction the exhaust gases from these experiments were scrubbed with tetrachloroethylene when no unraacted TiCI4 could be detected. The range of TiCl concentrations expressed as mol fraction of feed gases established using this technique was 4.5 x 10-5-3.3 Particle samples were taken from the flame by direct deposition on to electron microscope grids mounted on a brass holder.Samples were taken by sweeping the grid holder manually through the flame with a sweep time of approximately 1 s the number of passes required to give a particle number concentration sufficient for counting and size analysis varying from 1 to 4 dependent upon flame TiCI4 concentration. The grid mounting and support produces rapid quenching of the sample and this method was found to be superior to a quartz probe technique both in ease of operation and sample reproducibility. Determination of the absorption spectra of TiCI was carried out on the same burner. A deuterium arc lamp and slit collimator were mounted on an optical bench at one side of the burner in diametric opposition to a Hilger and Watt D292 grating monochromator. A constant slit width of 0.8mm was used throughout.The light intensity at the mono- chromator exit slit was measured with an RCA IP 28 multiplying phototube the output from which was displayed on a digital voltmeter. The reference spectrum for TiC14 was obtained on the burner with the flame unignited. The fusion of flame-produced Ti02 partides was examined initially by allowing them to deposit on a fine platinum wire at -9900°C removing a sample for examination and then re-introducing the wire into the flame at -1300"C,after which a further sample was taken. A more quantitative method was later applied in which the particles were deposited directly on palladium grids by brief exposure to a TiC1,-additioned flame after which the grids were placed in the electron microscope hot-stage and heated at a rate of approximately 20°C min-I while maintaining a visual check upon the particles.The temperature range within which the particle clusters underwent a sudden and marked shrinkage was noted. RESULTS AND DISCUSSION A variety of experimental techniques have been applied to the study of this particular reaction system ; those relevant to the present discussion have been detailed above. The discussion centres on the measured particle size distributions of material sampled from the flame a typical example of which is shown in plate 1. All distri- butions were obtained by sizing each particle present on the electron micrograph separately irrespective of its position with respect to other particles. Table 1 shows the change of particle size and standard deviation with residence time.The variation of mean particle size with initial TiCI4 concentration for samples taken at 2cm is shown in fig. 3. The results show clearly that the particles are growing in size as they travel down- stream from the burner and that the final size is related to reactant concentration. s7-3 FORMATION OF TiO AEROSOL TABLEVARIATION OF PARTICLE SIZE PARAMETERS AND GROWTH RATES AT TiCi4MOL FRACTION 2.0x 10-3 sampling position (cm above burner face) 0.5 1.o 1.5 2.0 estimated flame residence time/ms 50 100 160 230 d (geometric weight mean)/pm 0.041 0.055 0.062 0.063 standard deviation 1.372 1.349 1.329 1.354 d (geometric weight mean)/pm calculated -0.061 0.072 0.080 standard deviation -1.340 3..321 1.302 lo-' 5.8~ 7.1 x 10-3 apparent growth rate/(pm s-l) 4x 10-1 1.4~ growth rate/(pm s-l) from Ghoshtagore 4.4~ 3.1 x 7.1x 10-3 4~ 10-4 The main purpose of this work was to understand the processes leading to the initial distribution and the subsequent growth mechanism. The relevant stages to be considered are nucleation growth by chemical reaction and flocculation. X X X X X TiCI4 mol fraction (C,) FIG.3.-Variation of mean particle diameter with initial TiCI4 concentration. NUCLEATION By the use of similar arguments to those proposed by Ulrich,' it can be shown that the critical nucleus size under the experimental conditions which apply here is less than the size of the TiOz molecule. This implies that nucleation does not present a barrier to particle formation which is largely determined by the rate of chemical reaction.Collision processes between small particles is extremely rapid and as shown by the application of simple flocculation theory,' the concentration of particles present rapidly becomes independent of the initial concentration of nuclei. Hence providing that chemical reaction to produce new particles of TiO is rapid compared with the processes of growth by release of TiO at the surface of existing particles and growth by flocculation then nucleation need not be considered further as a factor affecting the final stage of the aerosol produced. We find in accord with Ulrich that assuming instantaneous chemical reaction the particle concentration and size are affected less than 1 % by the initial nucleus size and concentration after times of the order of lod6s.We are concerned here with events occurring at residence times of greater than 50 ms. A. P. GEORGE R. D. MURLEY AND E. R. PLACE The collision rate between small particles can be drastically reduced if they acquire an electric charge. This phenomenon has been shown 2g to occur in the formation of carbon particles but at the maximum temperatures encountered here the measured charge density is much less than that necessary to influence the collision rate of small particles. No particles smaller than 20 A have been observed in any of the samples taken from the smallest residence time position (50 ms). It is concluded that chemical reaction leading to formation of new particles is complete by this stage.CHEMICAL REACTION Evidence from several experimental results suggests that chemical reaction is rapid compared with flocculation. In a one-dimensional system such as the one studied here material deposition is proportional to particle surface area giving a rate of particle growth independent of particle size. This implies an invariant size distribution about an increasing mean diameter. This behaviour is not observed in practice. As shown later the results agree with the predicted behaviour of a floc- culating system. Chemical analysis of the gases present at the highest sampling point (230ms) showed no detectable presence of TiCI4. (The limit of sensitivity of the test gives a minimum value of 97 %for the extent of TiCI4 disappearance at this point.) The disappearance of TiC14 in the early stages of reaction was followed by u.-v.spectro-scopy. The results obtained are given in fig. 4. At a spatial resolution of 2 mm in 25 0 300 350 wavelength /nm FIG. 4.-Absorption Spectra V TiCI4 in 02+NZmixture corrected to flame conditions ; TiC14 additioned flame mol fraction 1.4~ x 2-3 mm above burner face; 0 3-4 mni above burner face ; 014-15 mm above burner face. FORMATION OF TiOz AEROSOL the flame the only position at which the absorption of TiCl could be detected was at a height of 2-3 mm. Even with no allowance for the unknown increase in the absorption coefficient of TiC14 with temperature comparison with the room temper-ature unreacted-TiCl spectrum shows an average degree of reaction of about 70 % integrated over the residence time of 30+ 10 ms.The estimate is even higher than this if reasonable allowance is made for the continuous background absorption prob-ably arising from the presence of TiO particles. Although disappearance of TiC14 cannot be related directly to the formation of TiO it seems likely that this step may well be rate controlling in the reaction. (Ti-C1 bond energy 82 kcal/mol). Values of apparent growth rate at the various sampling stations are given in table 1,assuming no nucleation and complete reaction at the 230 ms sampling point. Ghoshtagore gives the following kine& expression for the growth rate of Ti02 under the conditions appropriate to the concentration conditions used here dr 1.12x lo7 (-S.96 x lo3) dt - T__ exp p(TiC1,) pni s-T The values calculated from this expression are given for comparison in table 1.The predicted values are much slower than those observed experimentally. FLOCCULATIOK The evidence so far presented suggests that neither nucleation nor growth by chemical reaction are dominant factors in determining the final particle size char-acteristics. Consideration of the flocculation process shows that this can account for the essential features of the experimental results. The flocculation process comprises two components the collision process and the behaviour of particles after collision. The first has been widely discussed in the literature e.g. Fuc~s.~ The rate of collision is largely determined by the Brownian motion of particles.Only the presence of strong radial convective flows as for chemical reaction at the surface,6 and the effect of electric charges carried by the particle are likely to cause a marked change in the collision frequency. Chemical reaction can be neglected in this system. Application of electric fields to the aerosol system provides a simple method of estimating the total rate of charge generation by measurement of the saturation c~rrent.~The results of such measurements on this system have shown that at most only half the particles acquire a charge. The presence of charge at this level has a negligible effect on the observed flocculation rates. Equally important as the rate of collisions especially when the particle character-istics are being considered are the processes occurring after collision.A " sticking factor " is commonly employed to describe the fraction of collisions which result in the formation of a floc. However the characteristics of the floc vary widely depending on whether the particles stick and retain their individual identities or at the other extreme fuse completely to form a new " single " particle of larger size as happens with droplet suspensions. To assist in the interpretations of the results use was made of a computer flocculation model which allows predictions to be made of the developing size distribution of an aerosol system starting from any specified initial size distribution. Two classes of particle collision processes can be postulated to account for the observed growth of the individual particles.The first considers collisions between the observed particles and particles too small to be resolved by electron microscopy. If conditions at the first sampling point are considered and the limit of particle size that can be observed is put even as high as lOA then the collision rate of such par- A. P. GEORGE R. D. MURLEY AND E. R. PLACE ticles would be so rapid as to give growth to the final observed size of 630 A in a time of 1 ms. No conditions can be found which would predict steady growth over the observed period of about 180 ms. The second class of collisions concerns only the particles observed on the electron micrographs. If all the particles are considered to be present in the gas phase then the computer predictions for the development of both the mean size and of the size distribution agree closely with the observed particle sizes.With a sticking coefficient of 1.0 the predicted flocculation rate is about 10 "/o too rapid as shown in table 1. The calculation assumes a high degree of fusion between impinging particles since otherwise the collision diameter would increase at a rate much faster than that of the mean mass diameter. Close examination of the electron micrographs of particle samples shows in most cases that there are a number of particles present which appear to show signs of having been formed from two individual particles which have fused together. The particles arrowed in plate 1 have this form.Extrapolation of both theoretical and experimental lo data on the rate of sinter-ing of particles suggests that this phenomenon may allow an appreciable proportion of the particles present to sinter in the time available. Evidence is also available ' showing that particles of very small diameter exhibit properties of the liquid state at temperatures many hundreds of degrees below their bulk melting point. Experi-mental confirmation of this behaviour was found for the present system. Particulate material collected on a platinum probe and then reheated in the flame gases demonstrated a tenfold increase in particle diameter. Direct observation of the fusion process utilizing a hot-stage electron microscope with a sample of particles collected directly on the grid demonstrated a sudden change in structure when the temperature reached -840°C.Plates 2(a) and (b)show the sample before and after heat treatment. Heating effects of the microscope electron beam were shown to be negligible. On the basis of these results it appears plausible that at the temperatures of 1400-1100°Cin the flame rapid fusion of particles takes place. Further confirmation of the behaviour of the aerosol as a flocculating system has been obtained using the concept of the self-preserving size distribution.12 It has been established that given sufficient time the size distribution of a flocculating system expressed in non-dimensional terms should approach an equilibrium form. The properties of the size distribution have been established by Hidy.13 Fig.5 shows the measured size distributions obtained in these experiments compared with the self-preserving distributions. The trend in development of the non-dimensional distribution is similar to that observed by Ulrich and the final form lies close to the self-preserving distribution for high values of Knudsen number the regime which applies to the present particles. Finally results obtained for the variation of particle size at constant residence time with initial reactant concentration given in fig. 3 also demonstrate the appropriate behaviour for a flocculation controlled system. Theory predicts a relationship d = kC& where n can lie between 0.33 for small particles to 0.4 for the simple Smoluchowski equation.The value of tz found experimentally over an eighty-fold range of concentra- tions varies between 0.33 and 0.38 depending on the mean size parameter used. The predicted value of the constant k using small particle theory is 0.93 compared with the experimental value of 0.43. This discrepancy has not yet been resolved. The remaining item is related to the appearance of groups of particles in the samples which were taken. These are clearly observable in certain of the electron microscope pictures and their presence was originally taken as evidence of flocculates FORMATION OF TiO AEROSOL in the gas phase. This factor proved to be the major obstacle in the interpretation of the results but there is now some circumstantial evidence to demonstrate that the groups of particles seen in the samples are an artefact of the method of collection.lo-* Id' I 10 T FIG.5.-Self-preserving size-distribution function x flame residence time 50 ms ; 0 flame residence time 110 ms ; 0flame residence time 180 ms. -theoretical distribution Knudsen number > 10 from Ulrich ; - - - theoretical distribution Knudsen number 0.0 from Hidy. The major points in favour of this argument are (i) the crystal size distribution of the material in the groups is the same within experimental accuracy as the isolated particles; (ii) the apparent degree of flocculation is highest at the earliest sampling time which is contrary to expectation ; (iii) light-scattering measurements l4 on a similar type of system show particle sizes close to that of the single crystals rather than the groups observed on the sampling grids; (iv) flocculation theory cannot explain the apparent amount of flocculation seen on the earliest sample i.e.the maximum flocculation rate is too slow to give the observed result. CONCLUSIONS In the system considered flocculation is the process which essentially determines the particle size distribution. Although the findings refer specifically to this system flocculation will always eventually determine an aerosol-particle size-distribution unless factors such as particle charging reduce the collision rate to a negligible value. However in this system the fusion of particles after collision also plays a major role in that it determines whether the flocculation process leads to groups of small particles or simply larger particles.The fusion behaviour of small particles relevant to high temperature aerosols requires further examination before further conclusions can be made. The authors thank the Directors of Tioxide International for permission to publish this paper. The assistance of Mr. M. J. Westwood with respect to the calculation of theoretical flocculation rates and to Mr. W. Brander who carried out most of the experimental work is gratefully acknowledged. G. D. Ulrich Comb. Sci. Tech. 1971 4,47. E. R. Place F. J. Weinberg 11th Symp. Int. Combustion(The Combustion Institute Pittsburgh 1967) p. 245. J. Lawton and F. J. Weinberg Electrical Aspects of Combustion (Clarendon Press Oxford 1969) p. 247 fF.PLATE 1.-Electron micrograph of sample taken at 1.5 cm above burner face at TiCI4 mol fraction 2.0 x 10-3. [Toface page 70 PLATE2.-Sample of particulate TiOL(a)as taken from the flame ; and (6)after heating to 840’C. A. P. GEORGE R. D. MURLEY AND E. R. PLACE R. N. Ghoshtagore J. Electrochem. Soc. 1970 117 529. N. A. Fuchs The Mechanics ofAerosols (Pergamon Press Oxford 1964) chap. 7 p. 338. P. A. Tesner 7th Symp. Znt. Combustion (Butterworths London 1959) p. 546. ’J. Lawton and F. J. Weinberg Electrical Aspects of Combustion (Clarendon Press Oxford 1969) chap. 5. * M. J. Westwood private communication. W. D. Kingery and M. Berg J. Appl. Phys. 1955 26 1205. N. A. Fuchs A. G. Sutugin Highly Dispersed Aerosols (Ann.Arbor London 1970) p. 85. lo H. V. Anderson J. Amer. Cer. SOC.,1967,50,235. I2 D. L. Swift S. K. Friedlander J. Colloid Sci. 1964 19 621. G. M. Hidy J. Colloid Sci. 1965 20 123. I4 A. R. Jones (Imperial College London) private communication.