PHASE CHANGES IN SALT VAFOURS BY E. R. BUCKLE Dept. of Chemical Engineering and Chemical Technology, Imperial College of science and Technology, London, S.W.7 Received 20th June, 1960 A high-temperat ure cloud chamber technique devised for studying the freezing of supercooled ionic melts also provides data on phase changes involving salt vapours. Descriptions are given of the condensation, growth and evaporation of liquid and solid particles in clouds of alkali halides. Comparison is made with other work on salt aerosols and the results are examined in the light of recent knowledge of vapour constitution and the theory of nucleation in condensation and crystal growth. A method for studying the spontmeous freezing of ionic melts has recently been described.1 In this method molten salt is prepared in a state of fine sub- division by condensing the vapour.Salt vapour at the saturation pressure is generated, in the presence of argon, by heating the solid in the central chamber of a resistance furnace lined with carbon. A salt bead formed on a small platinum heater coil in the roof of the chamber functions as a supersaturating device. On passing a current of about 2 A through the coil the temperature of the bead is raised by 3QQ-4QOo@ and the vapour in its vicinity rapidly becomes supersaturated. Clouds of solid or liquid particles are formed, depending on the furnace tem- perature, and these are studied as they circulate in the supporting gas (fig. 1). Movement of individual particles is followed by examining the strongly-illuminated clouds with a telescope, the presence of crystalline particles being revealed by twinkling.By studying clouds at various furnace temperatures below the melting point Q, a point is located where twinkling first becomes apparent during the lifetime of a cloud. In this way, values are obtained for the threshold of freezing Ts and the critical supercooling 8 = Tf-.Ts. A full account of the method together with results for alkali halides will be given elsewhere? An interpretation of the observed supercoolings, based on the theory of homogeneous crystal nucleation, is also in progress.3 Certain qualitative information concerning the condensation, growth and evaporation of cloud particles was obtained during the course of these experiments which received only brief discussion in previous papers.In the present paper this is set out more fully, together with additional evidence, and compared with other work relating to alkali halide aerosols. The results are discussed with reference to recent studies on the constitution of alkali halide vapours and the theory of nucleation in condensation and crystal growth. RESULTS OF OBSERVATIONS ON CLOUDS The most profound changes in the characteristics of cloud particles for any salt were brought about by changes in the furnace temperature. This provides a convenient basis for the description of clouds, FURNACE TEMPERATURES ABOVE Ts At furnace temperatures above the freezing threshold the Airy diffraction patterns seen in the telescope consisted of a bright central disc of about 0 . 5 m diam.with occasionally a faint concentric ring separated from the disc by a dark 46E. R. BUCKLE 47 region. The patterns showed no abrupt periodic changes in intmsity and it was concluded that the particks were molten. Confirmation o€ this has now been obtained by examining fall-out from clouds with the microscope. Impingement on a glass slide placed in the cloud chamber results in spreading and co.alescence of the droplets, which thereupon freeze to a glass. For the majority of the salts studied, droplets formed rapidly on operating the supersaturator. Further growth was also rapid under continued supersaturation, droplets falling out from the cloud when their size exceeded about 4,u diam. When the initial degree of supersaturation was not maintained dxoplets quickly evaporated.Because of the rapid growth or evaporation of droplets the lifetimes of clouds were comparatively short. A lii'etirne of 5-10 sec was typical at tem- peratures within a few degrees of T , but this diminished quite sharply when the temperature was raised still further. Preferential growth of the larger droplets and their rapid precipitation on achieving a diameter of about 4 p probably explains the essentially monodisperse appearance of clouds very soon Gter their formation (fig. 1). The behaviour of sodium and lithium fluorides was rather different. Heats of evaporation are high for these compounds and more drastic use o€ the super- saturator was necessary to induce condensatlan. Droplets grew slowly under prolonged supersaturation but re-evaporated quickly when this was discontinued.Clouds had a polydisperse appearance immediately on formation and this per- sisted throughout their lives. Lifetimes €or lithium. fluoride clouds were sub- stantially longer than those for other slkali halides due to its low density. FURNACE TEMFERATUWS AT OR JUST WLOW Ts The freezing of droplets at T, was apparent in the onset of twinkling. This was visible in the telescope as flickerings of the Airy discs, the frequency of in- tensity change being always about 5 sec-1, Although the extent of intensity change could not be judged reliably by @ye, photographs showed that even where the effect was most readily seen it involved only partial extinction of the scattered light. The alkali halides so far examined may be classified into two series according to the reproducibility of their T,-values.z Critical supercoolings ("C) derived from T,-values and compiled data on melting points 4 are as follows : series A : NaCI, 168; NaBr, 166; KC1, 171; KBr2 168; KI, 159; RbCl, 156; CsF, 132.Series B : LiF, 232 ; LiCl, 190 ; LBr, 94 ; NaF, 281 ; CsCl, 152 ; CsBr, 161 ; CsI, 193. salt material was mainly of A.R. or laborqtory reagent purity and was examined without further purification. Compounds in series A showed clear twinkling and T,-values were reproducible to f3"C. As the furnace temperature was lowered the onset of twinkling ocqyrred earlier in the lifetime of a cloud until at 5-10" below T,, particles appewed to twinkle from the outset. Twiilkling was more difficult to discern for salts of series B, and the onset of twinkling was less definite.In spite of a number oE lengthy experiments with these compounds, it was not possible to obtain consistently reproducible values for T,, but in at least two experiments on each salt results were within f3" of the given values. Microscopical examination revealed that provided strict precautions were taken to exclude moisture, fall-out from cbuds of Kx or RbCl was composed, of trans- parent particks of about 5,u &am. In many preparations these were spherical and presumably glassy, although they showed no tendency to crystallize spon- taneously on keeping. Pravided, however, sufficient time was allowed for the condensate to cool before settling out on the slide particles were mainly crystalline. Crystals were often cubic, sometimes to a high degree of perfection, but mostly showed (1 1 I> facets at the cube corners.In some cases the (1 1 I} surfaces were more prominent, crystals showing various combinations of the cube and octa- hedron (fig. 2). Particles containing inclusions of lower refractive index were frequently found in RbCl fall-out. These inclusioqs disappeared readily when the48 PHASE CHANGES I N SALT VAPOURS particles were deliberately moistened. It appeared in some instances that crystal- lization of droplets was arrested at a very late stage, the last traces of melt having solidified to glass (fig. 2). A noticeable feature of all RbCl preparations was that incipient crystallization of droplets occurred at a single centre in an overwhelming majority of cases.Similarly, inclusions were usually found to occur singly, and their size was fairly uniform. The crystallinity of CsBr particles was much more difficult to discern on exam- ination with the microscope. It was apparent only in the surface roughness of the particles, which were invariably of spherical shape. Sediment from clouds formed at temperatures quite close to T, also contained many glassy particles with inclusions of lower refractive index than the matrix. The size of inclusion was much more varied than for WbCl and in some cases represented 10 % or more of the total volume of particle. Although, as with RbCl, the inclusions were apparently water-soluble the particles were non-hygroscopic. Glassy particles were evidently still fluid on hitting the microscope slide as many showed flattened surfaces. Con- glomerates of three or more glassy particles were occasionally found which showed flat interfaces.In so far as it was possible to watch droplets in clouds formed at these tem- peratures growth and evaporation appeared to take place readily, as at higher tem- peratures. Crystalline particles, however, grew more slowly, even at high super- saturation. Evaporation was also slow for crystals, and they were lost from clouds mainly by sedimentation. The marked change in the stability of clouds as the furnace temperature passed through Ts served to locate this roughly in pre- liminary experiments. As at higher temperatures, clouds of NaF and LiF were polydisperse, and the slow growth and rapid evaporation typical of droplets was also observed for crystals.FURNACE TEMPERATURES WELL BELOW T, At temperatures 50" or more below the freezing threshold the faintness of the Airy discs and their marked translational Brownian motion indicated that par- ticles of much smaller dimensions were formed. Confirmation was obtained on examining fall-out from clouds of RbCl and CsBr, which were found to consist entirely of sub-micron particles. In most cases growth of particles became in- creasingly difficult as the furnace temperature was lowered. A point was eventu- ally reached where prolonged use of the supersaturator merely resulted in multi- plication of particles, dense smokes being formed which dispersed very slowly on switching off the supersaturator. The behaviour of LiCl and LiBr was strikingly different, however.With these compounds growth and evaporation occurred readily at low temperatures, and the dense smokes observed with other salts did not form even at a temperature 400" below the melting point. DISCUSSION It is believed that for alkali halides of series A the nucleation and growth of the vast majority of cloud particles were spontaneous processes, since the general behaviour of clouds at any temperature was entirely reproducible. Any effects due to non-volatile chance impurities were either entireIy absent or confined to a very small proportion of particles. The materials responsible for inclusions in sediment particles of RbCl and CsBr have not yet been identified. Unless the inclusions were merely voids, it would seem that the impurities existed in the samples at the start and were volatilized and condensed along with pure salt.Inclusions were not visible in droplet preparations from clouds formed above T,, and their appearance at lower temperatures seemed to have no influence on the crystal habit of the pure compound.FIG. 1.-Molten salt particles in the cloud chamber (RbCl ; 1/25 sec exposure). FIG. 2.-Particles from RbCl clouds formed at furnace temperatures [To face page 48 about 15°C below the freezing threshold.E. R. BUCKLE 49 CONDENSATION AND RE-EVAPORATION OF DROPLETS The temperature and degree of supersaturation at which nucleation of droplets took place cannot be estimated very accurately. It is known 2 that temperatures in the vicinity of the supersaturator rose to levels where the saturation vapour pressure of the salt is several mm of mercury.Calculations using the Becker and Doering theory of homogeneous nucleation in vapours 5 and conservative estimates of the supersaturation predict nucleation rates adequate for the observation of clouds at temperatures above the melting point of the salt.2 As observations on the further growth of molten salt droplets and their re-evaporation were only possible over a narrow range of temperatures relatively little was learned about these processes. On comparing the behaviour of droplets with that of solid particles, however, it seems likely that for most alkali halides the condensation and evaporation of melt are non-activated changes (see also below). MECHANISM OF SOLIDIFICATION, AND THE GROWTH AND EVAPORATION OF CRYSTALS GROWTH IN THE MELT It has been shown elsewhere 2 that at furnace temperatures close to Ts crystals are formed in clouds of alkali halides by the freezing of droplets.Microscopical examination reveals that droplets of salts with the rock-salt structure crystallize to single cubes, although these later develop octahedral facets. In interpreting the critical supercoolings it is postulated 193 that the cubes grow from single three-dimensional nuclei and that the rate, and thereCore the tem- perature, of freezing is governed by the nucleation process. This implies that the time required for complete crystallization of the molten droplet following nucleation is negligible compared with the observation time.Microscopy of fall-out from RbCI. clouds might appear to contradict this assumption since cubes are occasion- ally seen which have a thin envelope of glass preserving the spherical outline of the original droplet (fig. 2). This is probably due to the sudden cooling of droplets which were swept on to the slide by convection currents. This interpretation is supported by the experiments of Jones, Burgers and Amis 6 who obtained spherical particles 70-300 ,u in diameter, and with smooth surfaces, by pouring molten NaCl on to a metal disc rotating at 4000rev/min. X-ray diffraction showed that the bulk of the material was crystalline. The formation of crystalline particles in aerosols of NaCl or KCl has been reported by a number of investigators.7-12 Correlation with present results is made difficult by lack of knowledge concerning the mechanism of solidification in these earlier experiments. The genera1 features of the products in all cases show close similarities. Where adequate precautions were taken to exclude moisture, particles usually consisted of cubic crystals of 0.01-5 p diam.9-11 Exposure to moist air resulted in rapid re-crystallization and sintering of particles,7-11 with at least partial modification of the cubic habit, although in one investigation 13 re- crystallization and growth of NaC1 particles exposed to moist air apparently pro- duced cubes of greater perfection.The degree of perfection of cubes in NaCl and KCl aerosols may vary for other reasons, however. Balk and Benson12 have observed rounding of cube corners in anhydrous preparations which they believe may be due to the presence of high-index facets.Support is given by experimental measurements on the heats of solution of the products. These lead to surface enthalpies which are anomalously high compared with theoretical estimates for { 100) surfaces. High surface energies for NaCl were also obtained by Lipsett, Johnson and Maass7 from heats of solution, and by Craig and McIntosh 9 from the pressures of water vapour in equilibrium with products.59 PHASE CHANGES IN SALT VAPOURS GROWTH IN THE VAPOUR If the theory of crystal growth developed by Burton, Cabrera and Frank for van der Waals’ crystals is applicable to ionic crystals,l4 spontaneous roughening of (100) and (111) surfaces of alkali halides with the rook-salt structure is likely to be absent at temperatures below the melting point.Growth of perfect crystals in the vapour must therefore depend on surface nucleation, which is a process requiring a high degree of supersaturation. In the experiments descyibd here crystals grew fairly easily at temperatures close to Ts. Thqs, once crystals had formed by the freezing of droplets they continued to grow in the vapour at levels of supersaturation well below that necessary for the nucleation of fresh droplets. Crystals formed from the melt may therefore hqve contained numbers of im- perfections, such as screw dislocations, which enhanced. the growth rate. The slowness of growth at furnace temperatures well b@ow the frqezipg threshold suggests that here a surface nucleation mechanism was operative.These particles were probably formed by sublimation,2 and were presumably too small to contain even a single screw dislocation. Calculations based on the translational Brownian motion were not informative as they indicated particle sizes of sub-atomic leve€,2 but the finest smokes, which gave rise to barely perceptible light scattering, possibly contained crystals little larger than three-dimensional nuclei. The preference for particle multiplication at low temperatures may be explained as follows. Three-dimensional nucleation is possible only in the region of high supersaturation in the immediate vicinity of the salt bead. These nuclei can grow very little before being transferred by convection to more remote parts of the cloud chamber where the degree of supersaturation is too low to support further growth by repeated surface nucleation.Because of the turbulence induced by the high excess temperature of the supersaturator, further exposures of particles to high supersaturations near the hot bead are infrequent and transitory. Work by various authors 15-24 on the constitution of the vapours of alkali halides has shown that most of these contain polymeric species. These might be expected to influence the rates af processes such as Condensation, growth and evaporation? In crystal growth, for example, condensation of polymeric Specks might involve dissociation to monomers or the attainment of critical orientations prior to their incorporation into a growing step. These requirements could con- ceivably result in an activation free energy not necessarily involved in the converse process of evaporation.Such considerations may explain the growth and evaporation behaviour of solid and liquid particles of LiF and NaF at high temperatures since there is an exceptionally high content of dimers in the vapours of these compounds.19, 22 Zarzycki’s studies on X-ray diffraction in molten alkali halides 25 show that the degree of short-range order is higher in fluoride melts than in the corresponding chlorides. Restrictions on the orientation of vapour molecules during condensa- tion may therefore be more severe for molten fluorides than for other alkali halides, where comparatively free interchange of small polymers would be expected between melt and vapour.26 EVAPORATION OF CRYSTALS Interesting differences in the evaporation process for various alkali halides are revealed by previous experimental studies.Bradley and Volans 27 found that evaporation in vacuo took piace at more OF less equal specific rates from (100) and (1 1 l } faces of KCl monocrystals grown from the melt. Evaporation coefficients M; of 0.5-0-7 were obtained for (loo), (110) and (1 1.1) faces. As the values were almost independent of temperature over a range of about 100°C they were thought to be due to an orientation factor28 It was later shown by Miller and Kusch 16 that the vapour of KCI contains appreciable quantities of dimers, although their experi- mental temperatures were about 200°C higher than those of Bradley and Volans.E. R. BUCKLE 51 Rotliberg, Eisenstadt and Kusch 22 observed activated evaporation from (100) faces of LiF and NaCl crystals in evacuated enclosures, a varying with temperature in the range 0.2-1.0, The vapours were shown to contain large proportions of dimers.In the same investigation it was found that the vapours of CsBr and Csl showed little or no evidence of polymers, and a was about 0-3. The orientation of the crystal faces was not known, however, and no information on the tem- perature dependence of a was obtained. In the present experiments the stability OP clouds at low temperatures indicates that evaporation of the smallest particles is an activated process. The exceptional behaviour of LiCl and LiBr is difficult to explain from considerations of the heats of sublimation or the structure of crystal and vapour since these do not differ sharply from those of most other alkali halides.No information is yet available on the crystal form of these particles. The development of octahedral facets on cubic RbCl crystals formed from the melt (fig. 2) suggests that these subsequently evaporate preferentially at { 1 1 1) planes, in contrast to the behaviour of KCl crystals of much larger size studied by Bradley and Volans.27 The rounding of cube corners in the NaCl and KCI preparations of Balk and Benson 62 may also have been due to evaporation. It is not easy on the basis of existing experimental evidence to trace any definite correlations between vapour constitution and the processes of crystal growth and evaporation of alkali halides.More data, particularly of a quantitative nature, are needed to test current theories of homogeneous nucleation in phase changes 29 and of the kinetics of growth and evaporation.26 1 Buckle and Ubbelohde, I. U.P.A. C. Symp. IFhermodynamics (Wattens, 1959). 2 Buckle and Ubbelohde, Proc. Roy. SUC. A, 1960,259,325. 3 Buckle and Ubbelohde, to be published. 4 Kubaschewski and Evans, Metallurgical Thermochemistry (Pergamon, London, 5 BwIcer and Doering, Am. Physik., 1935, 24, 719. 6 Jones, Xlurgers and h i s , 2. physik. Chem., 1955, 4, 220. 7 Liipsett, Johnson and Maass, J. Amer. Chenz. Soc., 1927, 49, 925, 1940. 8 Keenan and Holmes, J. Physic, Chem., 1949, §3, 1309. 9 Craig and McIntosh, Can. J. Chem., 1952,3Q, 448. 10 Young and Morrison, J. Sci. Instr., 1954, 31, 90. 11 Harrison, Morrison and Rose, J. Physic. Chem., 1957, 61, 1314. 12 Balk and Benson, J. Physic. Chem,, 1959,63, 1009. 13 McLauchlan, Sennett and Scott, Can. J. Res. A,. 1950, 28, 530. 14 Burton, Cabrera and Frank, Phil. Trans. A , 1951, 243, 299. 15 Friedman, J. Chem. Physics, 1955, 23, 477. 16 Miller and Kusch, J. Chem. Physics, 1956, 25, 845. 17 Pugh and Barrow, Trans. Farday SOC., 1958, 54, 671. 18 Berkowitz and Chupka, J. Chem. Physics, 1958, 254 453. 19 Eisenstadt, Rothberg and Kucb, J. Chem. Physics, 1958, 29, 797. 20 Porter and Schoonrnaker, J. Chem. Physics, 1958,29, 1070. 21 Kusch, J. Chern. Physics, 1959, 30, 52. 22 Rothberg, Eisenstadt and Kusch, J. Chem. Physics, 1959, 30, 517. 23 Schoonmaker and Porter, J. Chem. Physics, 1959,30, 283. 24 Datz and Smith, J. Physic. Chem., 1959, 63, 938, 25 Zarzycki, J. Physique Rad. (Suppl. Phys. Appl.), 1958, 19, 13A. 26 Hirth and Pound, J. Physic. €hem., 1960, 64, 619. 27 Bradley and Volans, Proc. Roy. SOC. A , 1953, 217, 508. 28 Bradley, Proc. Roy. SOC. A , 1953, 217, 524. 29 See, e.g., Dunning, Chemistry ofthe Solid State, ed. Gamer (Buttenvorths, London, 2nd ed., 1956). 1955, chap. 6).