Modern Concepts of Crystallization

---

I

PROPELLER DRIVE

H. B. CALDWELL Swenson Evaporator Co., New York, N. Y.

Modern Concepts of Crystallization Crystals of commercially acceptable size can be produced b y x' removal of excess fines

v' circulation of crystals to the supersaturation zone 11

maintaining high magma density

THERE

are three concepts which distinguish the modern crystallizer from its predecessors. These are: removal of excess nuclei or fines, circulation of the grobring crystals to the zone where supersaturation is being produced, and maintenance of a high magma density in circulation. These factors are considered here in relation to a continuous crystallizer Jvhich produces supersaturation by vacuurn cooling, rather than by evaporation as in the usual salt evaporator. T h e design, kno\cn as the draft tubr and bafflc type. is used to produce crystals up to about 16 mesh of a number of commonly crystallized materials. Theoretical

Starting a t the beginning, the mechanism of the birth of a crystal is not well understood: and possibl>- it is unimportant here. It is sufficient to say that lrhen a solution is cooled below its saturated temperature. it becomes supersaturated and enters Miers' "metastable" zone (Figure 1 ) . By his definition no neic nuclei form. but existing crystals \rill gro\v in this zone which may extend u p to .jO F. or more of supersaturation. Further cooling beyond the "metastable" zone produces neir nuclei even in the presence of existing crystals, and Miers calls this the "labile" zone. one of high supersaturation. T h e theory is helpful in understanding crystallizer design and operation even though the boundary location benreen the "metastable" and "labile" zones is greatly affected by factors such as mechanical stress in the solution. \,arying amounts of crystals in suspension, viscosit!-: and impurities. More recently. Rumford and Bain (6) \corking {vith NaCl solutions expressed this general concept in a seemingly more logical way. They postulated that. because of the coniplex energy changes required to separate a new solid phase? there is. at a low supersaturation value, a limiting particle size of solid nucleus below which solids tend to redissolve. However, a t this same low supersatura-

tion larger particles will grolr, but no new nuclei will form and persist. A t a little higher supersaturation, the redissolving potential of the solution decreases; and still higher supersaturation produces overwhelming shoivers of nuclei (Figure 2). Thus they envision a continuity of nuclei formation. almost from the start of supersaturation, which increases exponentially, along Lvith decreasing dissolving power as supersaturation increases. This appears to be a more rational and orderly concept than Miers' boundary bet\veen zones. Having touched upon nuclei formation, the next consideration involves conditions favorable to their groirth inro commercially acceptable size crystals. A nucleus or a growing crystal is surrounded wzith a film of solution Ivhich approaches saturation a t the crystal interface. Through this film: solute must diffuse to the crystal surface. Therefore. the higher the supersaturation of the bulk of the solution, the greater is the concentration potential across the film and the higher is the rate of diffusion. However? in Miers' theory. too high a supersatui.ation causes excess ne\c nuclei to form a t a n increasing rate: hence close control of supersaturation is indicated. !\.hen some solute has diffused through the film and reached the surface of the growing crystal. it tries to become orierted into the lattice and produce a \\.elldefined crystal. This order1)- orientation requires a definite amount of time depending upon the temperature and concentration involved and the particular crystal. Carrier and others (-I) have recently studied this particle integration rate in the growth of citric acid. Thus. there are t\ro rates involved in crystal groirth. These are necessarily equal during steady state groirth. although either or both of the rates may be controlling. Rumford and Bain ( 6 ) . as mentioned previously, showed that the g r o w h rate of KaC1 is diffusion controlled at high temperatures (above 50" C.) and surface-reaction controlled a t lolrer temperatures.

,

DRAFl 'TUBE ,SETTl -ING AREA

MC

LI

il

SK BP

'ATING

i DISCHARGE SLURRY

Diagram of draft tube-baffle crystallizer shows features which provide proper crystallization environment

Typical commercial installation of draft tube-baffle crystallizer VOL. 53, NO. 2

FEBRUARY 1961

115

-I

m

0

40 30

20

14

w

~

MESH SIZE OF

PRODUCT

Figure 1. Miers’ concept of supersaturation includes a “metastable” zone in which no nuclei form but existing crystals can grow

Figure 2. Nucleation rate curve of Rurnford and Bain (6) shows point of mass nucleation. Below this point no nuclei will form

Figure 3. Curve shows final weight of product when 1 pound of 140-mesh KCI particles are grown

Crystallizer Design A modern vacuum crysta!lizer of the draft tube and baffle type. which incorporates the three concepts mentioned earlier, is shoivn on p. 115. This is a true circulating magma design. T h e propeller causes upward flow in the draft tube. T h e fines removal area is betTveen the cylindrical baffle and the shell. If there is no flow through this annular space because of the fines removal stream. only clear liquor is present. However, when the fines removal stream is in operation a definite upward liquor velocity is established, and this will carry with it all those small crystals whose free settling rate is less than the upward velocity. Thus. excessive nuclei or fines may be removed from the crystallizer Also, the rate of flow of the fines removal stream controls the magma density in the crystallizer. T h e first concept of modern design and operation is the ability to remove the excessive fines which are usually produced in a commercial crystallizer. To illustrate the importance of fines removal assume? as did Newman and Bennett (5), that the smallest KCI crystals which persist are 140 mesh. When 1 pound of these are grown, the final weight and mesh size of the product is shown in Figure 3. Thus, if a 20-niesh product is desired, 588 pounds of solute must be deposited upon the original l pound of seed, and for a 14-mesh product, 2860 pounds must be deposited. Therefore, when producing a ton of 14-mesh KCI. all fines in excess of 0.70 pound must be removed, and it is important to remove them a t a n early age; otherwise crystallizer capacity is reduced. I n a crystallizer with the ability to remove excess fines, a greatly increased supersaturation level may be employed, thus increasing the crystal growth rate.

The second modern concept invo!ves bringing the crystals continuously to the area where supersaturation is producedin this instance, the slurry surface in the vapor head. The upward flow through the draft tube in the circulating magma crystallizer is a convenient and inexpensive mechanical arrangement to accomplish this. This is quite a different concept and is more favorable to crystal groit.th rhan supersaturating clear solution in one part of the equipment and then causing it to flow to a suspended cr!.stal bed in another part of the equipment where the supersaturation may be deposited. If a supersaturated solution is made to travel an appreciable distance, it will usually give u p some of its supersaturation along the way. \Yhen this happens, only a part of the original supersaturation reaches the gro\ving crystals, and the balance comes out of solution as “fines” or forms salt deposits on the equipment, or both. Of course: this could be prevented by producing only a very small amount of supersaturation in the clear solution, but then the growth rate would be small. This all means that the circulating magma type can use advantageously a higher supersaturation, which results in more rapid c r y a l yroivth and smaller equipment. This idea leads logically to the third, more or less modern concept of crystallization, namely, maintaining a high magma density for a favorable grolvth 25, 30>or 4070 crysenvironment-say tals by Lveight. I t might seem unreasonable that rapid circulation of such a heavy magma or slurry can grow strong. big crystals of 20- to 16-mesh, but such is the case. Of course, there is attrition bet\veen the crystals themselves, and attrition between the crystals and the equipment, and sooner or later as magma den-

sity increases, the attrition becomes an overwhelming influence and crystal size starts to decline. However, if the crystals are kept close together. supersaturation will deposit out on them, rather than form new nuclei (Figure 4 ) . These data concern the production of 20 tons per hour of crystals of a comnlon inorganic heav: chemical made in a triple effect calandria evaporator with mechanical circulators in the downtakes. The sales department required a larger crystal size, and successful efforts were made to achieve it by increasing the magma density in the evaporator bodies. hlany runs were made a t magma densities varying from 20 to 80Yc by settled volume. T h e results were most gratifying. An arithmetic probability plot of the data shows even more clearly the increase of crystal size with magma density. These a r e the three fundamental factors which distinguish a modern crystallizer design. Regarding retention time, i t is self-evident that this plays an important role in achieving a given crystal size. Large crystals cannot be groivn in a few minutes. However, retention time per se is rather meaningless. I t must be related to supersaturation level, fines removal. magma density. circulation characteristics. and the crystal gro\vth rate or the particular solute. Retention Time. Retention time may be decreased as supersaturation is increased, but keep in mind the excessive nucleation situation. Also, it is probable that forced growth results in structurally unsound crystals and inclusions of mother liquor and impurities. Fines removal is definitely related to retention time. M’ithout fines removal, theoretically a very small amount of supersaturation would not produce excessive fines and the desired crystal size could be achieved. I n this case, the

1 16

INDUSTRIAL AND ENGINEERING CHEMISTRY

CRYSTALLIZATION CONCEPTS

PERCENT APPARENT SOLIW

Figure 4. Effect of magma density on crystal growth Numbers adiacent to curves r e f e r to mesh sizes

retention time would be extremely long because of the necessity of maintaining such a low supersaturation. Even SO, this would not be achieved in practice because new nuclei \vould be continuously formed by attrition of the crystals, localized areas of high supersaturation, and other factors. Therefore, if it were not for fines removal, retention time could be excessively long, and there would be no assurance that the crystals would ever reach the desired size. O n the other hand, varying the fines removal can be a method of crystal size control. Obviously retention time and magma density are closely linked. I n a crystallizer of given volume, a n increase in magma density results in a corresponding increase in retention time snd vice versa. With other factors being favorable for crystal growth, this double-barreled retention time-magma density effect will frequently offer another convenient control of crystal size. Magma density for design purposes may be accurately expressed as weight percentage of suspended crystals. For operational purposes, a settled volunie percentage is probably more convenient. However, neither gives a true picture of the distance between the crystals. I t is the author’s thought that the togetherness or “propinquity” of the crystals is important. Since supersaturation, particularly in the .’labile” zone, is a rather fleeting phenomenon. new nuclei will be formed if the supersaturation has to travel far before reaching a crystal on which it may deposit as grow-th. This is tantamount to saying that a higher supersaturation may be used advantageously if the crystals are closer together. A little simple arithmetic shows that a 25% (weight) magma can readily have a difference of 20 to 25y6 in distance between the crystals: depending upon the

specific gravities of the crystals and mother liquor. Hence, a 25% (weight) magma may be favorable in one case but not in another. Retention time and circulation characteristics a r e intertwined. T h e circulation rate establishes the flash range a n d supersaturation level which was mentioned previously. A large. slotvspeed propeller operating in the draft tube is a satisfactory arrangement for gentle circulation of the crystal magma to avoid crystal damage and creation of addirional nuclei. T h e velocities in the i.ube are low, the head is less than 6 inchcs, and the propeller speed is a maxiniuni of 75 r.p.rn. T h e most favorable feature of the circulation is bringing sufficient crystal magma actually to the surface where the supersaturation is being produced. This effectively eliminates any time-decay of the supersaturation a n d permits its fullest use. Other hydraulic patterns can increase retention time requirement. Another factor listed as affecting retention time was the growth rate of the particular cqstal. Some excellent studies on this subject have been published during the past feFv years as the result of careful experimental work. However, differing configurations of equipment have been used. Because of wide variations in the nucleating and growing habits of crystals and the effect of impurities, pilot-plant-scale test work by a skilled engineer is today the most reliable guide to crystallizer design. T h e physical arrangement of the crystallizer design can greatl! affect the retention time required. Generally speaking. the draft tubebaffle, circulating magma type of crystallizer will produce 16- to 20-mesh crystals of sound structure and high uniformity from a number of common chemical solutions. Potassium chloride, borax, (NH,)?SOd, and hypo are typical. If extreme uniformity is required (85 to 90% of a given mesh size), a n elutriation leg discharge arrangement may be used. This permits crystals smaller than desired to be retained in the crystallizer for a n additional length of time. However, the elutriating leg itself contributes nothing to the growth ability of the crystallizer; it is merely a classifying device. Operating Aids. For the proper operation of a crystallizer, it is well to have sufficient and accurate instruments and controls to give as true a picture as possible of everything that takes place. Equally important is steady state operation. Many of the ordinary factors to be controlled will come to mind immediately such as temperature, concentration, clarity, and rate of feed; also needed are controls for absolute pressure, level, and body temperature. Visual aids, such as

peepholes above a n d below the slurry level and a short piece of borosilicate glass pipe on the fines removal stream, a r e a big help. A differential thermometer accurate to 0.20’ F. showing the difference in temperatures between the draft tube and body is essential in determining the flash range and the capacity of the circulator, which should he provided Lvith a variable pitch propeller. A recordino; wattmeter \vi11 give a sufficient indication of magma density by reference to samples of the discharged slurry. A convenient method of operation is to meter the slurry discharge by means of a variable speed, calibrated, positivedisplacement pump, plus a magnetic flow meter and a manual throttling valve on the fines removal stream. With these tlvo streams under control, the level controller may operate on the clear liquid feed, on which stream a n indicating rotameter adds confidence. .A control scheme, as just described, makes it possible to set the slurry discharge flow to give the desired production rate. Procedures

Before describing start-up procedures, some thought should be given to the feeding situation. T h e feed stream \vi11 enter the crystallizer immediately below the draft tube and will be promptly mixed with the slurry Howing u p the draft tube. However, it is important that the feed does not Hash a t the point of its introduction, otherlvise excessive nucleation could occur. -4n analysis of the prevailing temperatures, pressures, and boiling point elevations MAI determine whether there is a possibility of flashing before the feed mixes with the slurry. If flashing is indicated, it may be advisable to mix the feed under pressure Lvith some cool mother liquor before introducing it into the crystallizer. Even this expedient should be examined (vith regard to the shape of the solubility curve. .4nother point to keep in mind Lvhen two streams of clear sarurated solutions a t different temperatures are ”mixed” or blended together is that heat transfer \vi11 occur more quickly than mass transfer, and the hotter stream may nucleate excessively. If this is indicated, it may be better to mix the feed under pressure \vith some slurry from the body, although continuously recirculating slurry through a n outside pump and throttle valve should be avoided to prevent grinding u p the crystals. Destroying the fines in the stream removed from behind the baffle is usually quite simple, particularly if they are not larger than 100 mesh. Sometimes this stream may be introduced into a n V a l . 53, NO. 2

FEBRUARY 1961

117

unsaturated liquor a t the front end of the process. Since considerable water is vaporized in the crysta!lizer, perhaps some dilution in the fines removal stream can be tolerated. If neither of these methods can be used. then heating the fines stream either by live steam injection or indirect means may be indicated. I n any event. the fines stream should be appreciably undersaturated to give adequate dissolving potential. There is one more point before discussing actual crystallizer operation. Make certain that it can be emptied somewhere other than the se\ver, in case something goes Ivrong. The ideal is a tank with a n agitator to prevent the crystals from settling into a solid mass in the bottom. T h e best ivay to ensure a successful start-up of a new crystallizer is to fill it partially to its operating level Lvith a solution or mother liquor of the same concentration and temperature Lvhich will prevail during normal operation. Then start the circulation and add sufficient crystals to produce a magma density of a t least 10 to 15% (weight). If possible, the crystals should be free of 100 mesh. ’i-acuum may then be established a t the operating value. feed admitted b!- level control regulation, and the fines removal stream started a t the designed rate. The slurry discharge should not be started until the magma density has reached its designed value of perhaps 30% (weight). This will require a number of hours. During this period, samples of the magma should be withdrawn a t 30-minute intervals and checked for magma density and screen analysis. To folloiv crystal growth progress: each screen analysis should be plotted. Arithmetic probability paper. with its expanded scale at both ends, is helpful. \Vhen proper inagma density is reached, slurr)- may be discharged at. say. 259; of design rate: gradually increasing the floiv comiiiensurate with crystal growth until the desired product size and capacity are reached. During this start-lip period. frequent checks should be made on the differential thermometer to make sure that the flash range (supersaturation) does not exceed the designed value. T h e \\orst blunder which can be made is to deplete the magma density, next is to permit \?ariation in the absolute pressure. Once the crystallizer is operating properly, temporary upsets \vi11 do no harm, particularly with large crystallizers of 10- to 12-feet diameter. T o start a crystallizer kvithout establishing a magma? as described above, may be time consuming and hazardous because of excessive nuclration. sometimes referred to as a cloud or skimmed milk. T o avoid such a catastrophe, seeding or inoculation using very accurate temperature control can be most

1 18

helpful. The crystallizer may be filled to its operating level with feed solution which is a few degrees above its saturation temperature. Very slowly its temperature is lowered well into the metastable zone? allowing sufficient time for the enrire charge to reach equilibrium after each temperature decrement. I t is assumed that the metastable zone boundary \vi11 have been previously established in laboratory glassware. The cooling is then stopped. and the charge is inoculated with a slurry at about the same temperature as the charge and containing perhaps 100 pounds of crystals. This should trigger off and, after several hours. deplete the existing supersaturation by growth on the added crystals. Then, cooling may be resumed a t a n increasing rate over the next 8 hours until the charge has reached its designed operating temperature. .4t this time feed: s!urry discharge. and fines removal may be established a t design rates. T h e presence of even a small amount of suspended solids such as silt. clay. iron hydrate, or other insoluble material in the feed solution used to be feared. as they might promote excessive nucleation. but the author’s experience has shown that this is not true. Undoubtedly? soluble impurities and additives have considerable influence on crystal habits ( 3 ) .but the author’s first hand knowledge of these effect? is very limited. Once. when starting a n epsom salt circulating magma crystallizer, only small granular crystals \\-ere made. but upon the continuous addition to the feed of a small amount of iYa.HPOn the production changed to large, chunky. needle-like crystals as demanded bv the trade. It is well known that (XH.JrSO4 crystallization benefits by the addition of soluble P 2 0 j and by maintaining the free acid a t about 0.50% (7. 2). Salt Deposits. The length of time a cri-stallizer may be maintained in continuous operation depends mainly upon the rate of accumulation of salt deposits and their locarion. I n general. salt build-up does not occur belo\v tht. slurry surface in the draft tube-baffle design when a proper magma density is maintained. During start-ups on clear KCI solution. there is evidence at the bottom of the draft tube of sufficient turbulence. cavitation, and localized areas of lo\v pressure to cause some saltins. Apparently this build-up does not continue appreciably after the magma density has reached a certain value. because the usual operating cycle is 30 days or longer. I n most crystallizers, salt deposits occur on the body walls a t the slurry-vapor boundary. O n e NH,NO, crystallizer was supplied with highly polished 304L stainless steel for the body walls. X salt band about 6 to 12 inches wide would form on the body wall, apparently from liquor

INDUSTRIAL AND ENGINEERING CHEMISTRY

splash. I t was found that if the slurry level was raised about 12 inches and slowly relowered once every 8 hours. the salt deposit would slough off. This particular crystallizer operated for more than a year without being shut doivn for removal of salt build up. I n other case$, a water wash ring installed about 3 feet above the slurry level and used intermittently is very effective. I n a KC1 crystallizer in S e w Xleuico, the ambient temperature a t the upper part of the vapor body was appreciably higher than the saturated vapor temperature inside the body. After 10 days of operation, a massive salt deposit had almost blocked o f f the vapor outlet. This was attributed to lack of any air condensing action in the bod!,; lvhen the vapor passed into the vapor line. a small amount of superheat \vas genrrated because of the pressure drop encountered a t the vapor outlet. This superheat caused evaporation of droplets entrained in the vapor. and since these droplets tvere saturated solution. salt (vas deposited. The problem was solved by a fresh water spray a t the vapor outlet. The Data Log. I n this day of recording instruments and controls, there seems to be a drift away from keeping a handwritten log of the data which are essential in keeping a crystallizer on stream and producing a satisfactory product. This is wrong technically and psychologically. For instance. there is no substitute for peering intently and intelligently into a peephole to observe the character of the circulation, its pattern, its degree of turbulence. the froth or foam, the glistening or salting of the walls, and the spatter of crystals on the sight glass; just as important are the steadiness of an absolute pressure gage, or the needle of a n ammeter, or the settling of the crystals in a graduated cylinder, or the Tyndall effect of the fines removal stream: or the teniperature of a bearing, or the flow of sealing ivater to a lantern ring. T o the operator who moves about his equipment in the course of keeping a log. these additional observations will become second nature and will mean much in anticipating trouble before it becomes serious. Literature Cited (1) .4dam, \$’. G., U. S. Patent 1,919,707 (July 25, 1933). ( 2 ) Berkhoff, G., Chem. 2 .Vet. En?. 44, j

.

366 (1937).

(3) Buckley, H. E., “Crystal Growth.” Chap. 10, Wiley, New York. 1951. (4) Cartier, K.. Pindzola. D.. Bruins, P. F., IND.ENG.CHEM.51, 1409 (1959). (5) Newman. H. H.. Iknnett, R. C., Chem. Ene. P Y O ~55. Y . No. 3, 65 (1959) (6) Rumfoyd. F., Bain, J . , Inst. Chem.

Engrs. Meeting, London, October 1959. RECEIVED for review July 2 . 1960 ACCEPTED November 14, 1960 ACS North Jersey Section, Symposium on Crystallization. Linden, N. J., April 1960.