Probing into Nucleation Mechanisms of Cooling Crystallization ofSodium Chlorate in a Stirred Tank Crystallizer and an OscillatoryBaffled CrystallizerCraig J. Callahan and Xiong-Wei Ni*

Centre for Oscillatory Baffled Reactor Applications (COBRA), School of Engineering and Physical Science, Heriot-Watt University,Edinburgh, EH14 4AS, United Kingdom

ABSTRACT: In this paper we report our recent studies intowhy different nucleation behaviors in cooling crystallizationswere observed in a stirred tank crystallizer (STC) and anoscillatory baffled crystallizer (OBC). Using sodium chlorateas the model compound and setting up a number ofhypotheses, we identified and verified different nucleationmechanisms for the above two systems: while under the sameoperational conditions, secondary nucleation occurred ex-clusively in the STC where the product crystals bore the samehandedness as the seeds; a combination of primary andsecondary nucleation was displayed in the OBC where a mixture of left- and right-handed crystals coexisted when a single-handedness crystal was seeded. Through further hypothesis, we established the vital link between the nucleation mechanisms andthe ways mixing is generated in the two systems and discovered that the scraping action between the baffle edge and the wall ofthe crystallizer could be the main culprit for the unexpected primary nucleation generated. All of our data in this paper confirmthat the fluid mechanical environment in crystallizers plays a profound role on nucleation and subsequent growth.

1. BACKGROUNDCrystallization is an important separation technique in thechemical and pharmaceutical industries. Although manytheories are available in the literature,1−5 the fundamentalscience behind crystal nucleation and growth is not fullyunderstood and in an industrial setting operator experiencerather than sound scientific understanding often plays a majorrole in the success or failure of a crystallization process. Thescale up of crystallization processes also poses furtherchallenges for chemical engineers and process chemists inthat although batch processes are well understood the scale upis often marred with batch to batch variations with additionalproblems not encountered in the laboratory scale. This islargely due to different fluid mechanical conditions present atdifferent scales of operation.6 It is widely accepted thatdiscrepancies in mixing and flow patterns have a profoundinfluence on scientific parameters such as metastable zonewidth (MSZW),7 nucleation, and growth mechanisms.8,9 Thisconsequently affects the final crystal specifications such as thesize distribution and morphology.10 A drive to minimize thesedifferences in fluid dynamics at different scales of operation hasled to the use of continuous crystallization by means of plugflow crystallizers such as the continuous oscillatory baffledcrystallizer (COBC).11−13 Due to the uniform mixing, plug flowcharacteristics,14 enhanced mass,15 and heat transfer rates,16

together with the ease and readiness of controlling temperatureprofiles along the length of the crystallizer, this provides aconstant fluid mechanical condition for nucleation andsubsequent growth of crystals. This has led to uniform crystal

sizes17 with significantly enhanced filtration rates.18 Thetransition of a labscale COBC of typically 10−15 mm diameterto a pilot/full-scale COBC of 40−80 mm diameter is a linearprocess,19 which means that there is little variation in fluidmechanical conditions as well as the metastable zone widths,and this facilitates a direct and smooth scale up operation. Inaddition, analytical tools and monitoring techniques can beused at all scales without modification, enabling direct transferof knowledge and know how from lab to industrial volumes.In recent studies of crystallization of active pharmaceutical

ingredients (APIs) in the OBC, some interesting results wereconsistently observed. For example, higher nucleation temper-atures and narrower MSZW was observed in an OBC than in astirred tank crystallizer (STC);20 rapid growth of crystals in anOBC without entrainment of impurities was possible;21 alsoseeding was not necessary in an OBC to obtain a certain crystalspecification, while it was essential in STC for the sameoperation.22 This current work aims to address the latterfinding.Our hypothesis is that a different nucleation mechanism

could be observed in an OBC compared to that in a STC dueto different styles of mixing or mixing mechanisms, while allother operational conditions remain constant, such as super-saturation, temperature, mixing intensity, etc. Differentnucleation mechanisms may result in different crystallization

Received: January 30, 2012Revised: March 29, 2012Published: April 3, 2012

Article

pubs.acs.org/crystal

© 2012 American Chemical Society 2525 dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−2532

specifications, for example, secondary nucleation can be thedeterminant parameter for final average crystal size distribu-tions as discussed by Mandare and Pangarkar (2003).23

Understanding and controlling these differences in nucleationin an OBC would be essential in order to engineer crystals ofdesired attributes with ease and confidence. It is however nearlyimpossible to observe nucleation directly due to the fact thatthe event of nucleation is too fast to be recorded and the size/concentration of nuclei is too small/low to be detected by anymodern measuring tools. The focus of this work was thendirected at finding a crystal compound whose enantiomorphismor morphology in the solid state could be related to thenucleation mechanism.One such a compound is sodium chlorate, a nonchiral

substance that on crystallization produces crystalline solidspossessing either a left- or a right-handed enantiomorphism.24

As such, the crystalline solid will display optical activity, aproperty that has been exploited in previous studies todetermine the handedness of the crystals based on the directionof rotation of plane-polarized light.25,26 For the purposes of thiswork we refer to the dextrorotatory crystals as “right handed”and the levorotatory crystals as “left handed”.Two relevant investigations into the nucleation mechanism

were reported in 1972 by Denk and Botsaris,27,28 where at asupercooling of 3 °C and 350 rpm stirring a crop of productcrystals had 100% right handedness when seeded with a singleright-handed crystal (unfilled circles in Figure 1). Theequivalent outcome was reported for seeding with a left-handed crystal. This would indicate that secondary nucleationtook place due to either dendrite coarsening or collisionbreeding within the crystallizer, i.e., the entire product crystalsoriginated directly from the seed crystal. Again, from Figure 1 itcan be seen that without seeding and at higher supercoolings ofgreater than 7 °C primary nucleation was observed. This wasshown from a mixture of 50:50 left- and right-handed productcrystals obtained (filled triangles in Figure 1). This wouldsuggest that if spontaneous nucleation occurred, a mixture ofboth left- and right-handed crystals would be the evidence. Thisoutcome was also supported by Martin et al. (1996).29

Qian and Botsaris (1997)30 cited the work of Sung (1973),31

which suggested that fluid shear had an effect on theproduction of secondary nuclei. This phenomenon was also

supported by Evans et al. (1974), who discussed that althoughcrystal−crystallizer collisions and crystal−crystal collisionsaccounted for around 75% of the nucleation within an agitatedcrystallizer the remaining 25% of nucleation was controlled,speculatively, by fluid shear.32 Ayazi Shamlou et al. (1990)showed how secondary nucleation of potassium sulfate crystalscan occur due to “fluid-dynamic-induced breeding”.33 A morerecent investigation by Liang et al. (2004) on the crystallizationof L-glutamic acid also highlighted the dependence ofnucleation on crystallizer hydrodynamics.34 What we learnedfrom these previous studies is that by studying the productcrystal handedness it would be possible for us to probe into thenucleation mechanism when the hydrodynamic environment ischanged in different crystallizers, provided that other processconditions, such as supersaturation, temperature, mixingintensity, etc., are kept constant. This is our strategy for thefirst study of this kind utilizing the OBC.

2. EXPERIMENTAL SETUP AND PROCEDURESOn the basis of the above explanations two types of batch crystallizerswere studied: the STC and the OBC. Each crystallizer was serviced bya water bath and temperature monitoring tools.

Stirred Tank Crystallizer. Figure 2 shows the setup of the STC.The STC consisted of a jacketed round-bottom glass vessel with avolume of 500 mL. A stainless steel two-blade paddle-type stirrer wasdriven by the motor. The stirrer was 70 mm in diameter, 15 mm inwidth, and 10 mm in clearance from the vessel floor. With a volume of500 mL, the height of the liquid in the vessel is approximately 80 mm,as indicated in Figure 2. Rotation was adjustable from 50 to around2000 rpm.

The power density of a stirred tank crystallizer is defined by35

=ρP

VP N D

VO s

3S5

L (1)

where P/V is the power density (W m−3), ρ the fluid density (1456 kgm−3 at 40 °C), Ns the speed of the stirrer (rps), DS the diameter of thestirrer (0.07 m), VL the volume of liquid in the STC (0.0005 m3), andPO the dimensionless power number of the agitator, which wasestimated as 2.7 based on data presented by Nienow and Miles36 forthis type of impellor.

Oscillatory Baffled Crystallizer. Figure 3 shows the setup of theOBC. The OBC consisted of a jacketed glass column with an internaldiameter of 50 mm and height of approximately 500 mm. This gave aworking volume of 500 mL. The vessel was held in position by a

Figure 1. Types of product crystals obtained in stirred solutions. Reprinted with permission from ref 27. Copyright 1972 Elsevier.

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322526

stainless steel flange. The baffle string was comprised of 4 PTFE baffleswith an outer diameter of 48 mm, an orifice diameter of 24 mm, and athickness of 3 mm. The baffles were spaced at 65 mm by stainless steelspacers. The baffles were oscillated in the vessel by means of a linearmotor (Copley Controls Corp.), which was held by an aluminum/plastic frame above the OBC. Frequencies of 0−10 Hz and amplitudesof 0−50 mm were possible.The power density of the OBC is estimated using the quasi-steady

flow model proposed by Baird and Stonestreet (1995)37

=ρπ

− αα

π⎛⎝⎜

⎞⎠⎟

PV

NC

x f23

1(2 )b

D2

2

2 o3 3

(2)

where Nb is the number of baffles per unit length of OBC (15.38 m−1),CD is the discharge coefficient of the baffles (taken as 0.7), xo is thecenter to peak amplitude of oscillation (m), f is the oscillationfrequency (Hz), and α is the baffle free area ratio (0.23 for this OBC).The free area ratio, α, is defined by the area of the orifice divided bythe tube cross-sectional area.

Measurement Tools. Heating and chilling were provided to thecrystallizers by use of a Grant GP200R2 heater/chiller. The jacket fluidwas water. Temperature was monitored using stainless steel T-Typethermocouples and a 4-channel temperature display, which wasinterfaced to a PC for recording.

As crystalline sodium chlorate will rotate the plane of linearlypolarized light, the method of analysis is polarimetry. A simplepolarimeter was constructed using a light source, two polarizers, and aglass slide upon which the crystals can be placed, as schematically

illustrated in Figure 4. The light passes through the first polarizer, thesample, and into the second polarizer for observation. The rotatingmount rotates the second polarizer either clockwise or anticlockwise,allowing the handedness of the crystal to be ascertained.

This setup was used for all crystals obtained. When the polarizerswere initially crossed, all of the dried crystals are transparent with apale blue color (Figure 5a). Rotating the second polarizer clockwiserevealed the right-handed (dextrorotatory) crystals as dark blue(Figure 5b); with further rotation the crystals became orange (Figure5c). The left-handed (levorotatory) crystals remained pale blue. Theleft- and right-handed crystals can then be separated into separatebeakers for counting and weighing.

Reagents. The laboratory-grade sodium chlorate used for theexperiments was sourced from Fisher Scientific UK (99+% purity).This is termed as the fresh material in later discussions. For someexperiments, recycled sodium chlorate from previous mother liquorswas also used. To obtain the recycled material, the mother liquor ofthe seeded trials was collected and evaporated; the concentratedsolution was then cooled, filtered, and dried to give what we refer to as“the recycled starting material” in later sections. Recycling motherliquor has become a common technique for not only meeting “greenchemistry” requirements but achieving the maximum possible yieldfrom a crystallization process.38 Distilled water was used in allexperiments, which had been filtered using a Whatman glass fiber filter(Grade GF/C, nominal pore size 1.2 μm). Qualitative analysis of thepurity of both the fresh and the recycled materials was attempted usingUV−vis spectrophotometry. Samples of 1 g mL−1 were made up withdistilled water and analyzed using a Hach DR/4000 spectropho-tometer. The resulting data is shown in Figure 6.

Note that it is difficult to identify the exact impurities in sodiumchlorate, but what we did learn from the recycling process and fromFigure 6 is that the impurities were reduced by around 28% in therecycled material (based on the area under the curves presented inFigure 6).

Figure 2. Image of the STC setup (not to scale).

Figure 3. Schematic of the OBC setup (not to scale).

Figure 4. Schematic setup of the polarimeter.

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322527

Using solubility data published by Seidell,39 the mass of sodiumchlorate required to prepare saturated solutions at 31 °C wasdetermined as shown in Table 1.

Seeds. Seeds were prepared by naturally cooling a quiescentsodium chlorate solution in a 250 mL conical flask. Sodium chlorateand water were first mixed in the sealed flask, which was thenimmersed in a water bath at 30 °C. Once the material had dissolved,the flask was removed and allowed to cool to room temperature. Afterleaving the solution overnight, crystals were extracted by filtrationusing a Buchner funnel and filter paper (Whatman grade 1). Crystalswere immediately washed with 20 mL of cold water to remove anytraces of mother liquor and then dried between filter papers. Thesewere the seed crystals stored and used for our trials. The handednessof the seed was confirmed by polarimetry as described previously.Experimental Procedure. In order to test our hypothesis that a

different hydrodynamic environment would alter the nucleationmechanism, the seeded trials were conducted at a constant set ofprocess conditions, such as supersaturation, concentration, cooling,etc., while only varying the mixing mechanism, i.e., in a STC and in anOBC. This would allow comparison between the two types ofcrystallizers. The conditions studied are displayed in Table 2.The mixing conditions studied are given in Table 3, with the aim of

matching power densities in the two devices.21,40

The general experimental procedure involved weighing out therequired amounts of reagent within each crystallizer (values given inTable 1). The solution was held at 40 °C for 1 h under the agitationconditions given in Table 3 to ensure that all solid particles had

dissolved. After dissolution, the solution was filtered using a Whatmanglass fiber filter to remove any foreign particles. The solution was thencooled to the supersaturation point under controlled conditions. Asingle seed crystal of known handedness was washed with cold distilledwater before being suspended by means of a wire in the crystallizingsolution for 3 min. The seed crystal was then removed from solutionand agitation stopped to allow product crystals to grow overnight at 30°C with no mixing throughout to form larger crystals. Thereafter,product crystals were collected and quickly washed with 20 mL of coldHPLC-grade water and dried on filter paper. Analysis of the crystalswas performed using the polarimeter described in Figure 4. The left-and right-handed crystals were then counted, separated, and weighed.Each experiment was repeated a further two times in order to establishthe margin of error within the data. The apparatus was thoroughlycleaned between trials using household detergent at 60 °C followed bythree rinses with hot water. The apparatus was allowed to dry prior touse.

3. RESULTS AND DISCUSSIONSBefore presenting the main results, a couple of benchmarkexperiments were performed where no mixing at all was appliedto the crystallizers. This was to verify our hypothesis that it isthe mixing mechanism that has an effect on the secondarynucleation. The first set of benchmark tests was conducted withseeds but in the absence of mixing.When no mixing was applied before, during, and after

seeding, we see that all product crystals bore the samehandedness as the seed crystal in all experiments for both theOBC and the STC. This indicates that secondary nucleationwas the only mechanism in these tests and implies that primarynucleation did not take place. If primary nucleation hadoccurred, a mixture of left- and right-handed crystals wouldhave been observed as primary nucleation yields both handedproduct crystals as shown by the triangles in the Figure 1

Figure 5. Determination of seed handedness (a) initially crossed polarizers, (b) second polarizer rotated clockwise, and (c) further rotating secondpolarizer clockwise.

Figure 6. Graphs of absorbance vs wavelength for recycled material [mean absorbance of 0.036, 374 nm with area under curve of 66.1 × 106 au](left) and fresh material [mean absorbance of 0.174, 370 nm with area under curve of 91.6 × 106 au] (right).

Table 1. Solution Recipe from Seidell39 for 500 mLSaturated Solution

TSAT(°C)

C (g NaClO3/100 mLsatd soln)

ρ(kg m−3)

mass ofNaClO3 (g)

vol. of H2O(mL)

31 73.79 1435 369 349

Table 2. Conditions Studied in This Worka

TSAT (°C) TCRY (°C) ΔT (°C) S vol. (mL)

31 30 1 1.006 500

aTSAT is the saturation temperature of the solution (°C), TCRY is theseeding (supersaturation) temperature (°C), ΔT is the supercooling(°C), S is the supersaturation, and vol. is the volume of solution in thecrystallizer (mL).

Table 3. Mixing Conditionsa

NS (RPM) f (Hz) xo (mm) P/V (W m−3)

STC 57 11.9OBC 0.4 32 12.0

aNS is the STC stirrer speed (rpm), f is the OBC frequency ofoscillation (Hz), xo is the amplitude of oscillation (mm), and P/V isthe power density of the system (W m−3).

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322528

presented earlier and some recent work presented by Vogel(2011).41

In order to further verify that no spontaneous nucleationcould occur as a result of surface phenomena, the second set ofbenchmark experiments was conducted in the same manner asthe first set, except no seeds were used, but only a blank wirewas suspended in the solution. The crystallizers containing thesupersaturated solutions were left overnight, and it was foundthat no crystals can be seen in either system, indicating that thepossibility of either the presence of a foreign body (the blankwire) or entraining foreign particles to initiate heterogeneousprimary nucleation was not possible. The same outcome wasobserved three times in each system. This suggests that seedsare the only necessary means to cause nucleation in bothsystems.Following on from the benchmark experiments, mixing with

identical power density was applied to the two systems. Tables4 and 5 display the outcome.

From the results of three repeated runs, it is clear that underthe conditions studied in this work the STC consistently gaveproduct crystals with 100% similarity to the seed handedness.On the other hand, the OBC was always ≤96%. As all of theproduct crystals from the STC bear the same handedness as theseed, it would appear that the nucleation mechanism wasexclusively secondary. In the OBC, the secondary nucleationwas still a dominant feature but some sort of primary nucleationmust have occurred in order to have the product crystals with

the opposite handedness to the seed. This is a very interestingfinding as it implies that the OBC could possess a means ofcreating primary nucleation without the need for an externalstimulus such as ultrasonic42 or laser.43 Before we examine howthis could be possible, we want to further confirm our findings.We repeated the same tests as above using the recycled materialas explained earlier, and the following data was collected(Tables 6 and 7).

The results are clear that the STC produced crystals alwaysof the same handedness as the seed while the OBC did not,with the similarity to the seed being less with the recycledmaterial (87.8 ± 9.9%) than that with the fresh material (94.2± 1.8%). This indicates that the recycled material promotedmore primary nucleation under the same conditions, aphenomenon that was also observed in other crystallizationprocesses.44 Intuitively speaking together with our UV−visanalysis shown earlier, the recycled material (produced by arecrystallization process) has around 28% fewer impurities in itthan the fresh material of the identical compound. Fewerimpurities would correspond to fewer opportunities forsecondary nucleation and in turn, for more primary nucleation.However, we do not yet have deeper explanations on this.Nevertheless, the repeated data served the purpose of validatingour finding.What has caused the primary nucleation in the OBC? There

are a number of possibilities, e.g., microcrystalline dusts or theway the mixing is generated. A microcrystalline dust on the

Table 4. Results for the Fresh Material with Right-HandedSeeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC1 8.1 0.6 8.7 93.12 5.4 0.4 5.8 93.13 2.6 0.1 2.7 96.3mean 5.4 0.4 5.7 94.2SD 2.8 0.3 3.0 1.8STC1 1.4 0 1.4 1002 4.3 0 4.3 1003 8.9 0 8.9 100mean 4.9 0 4.9 100SD 3.8 0 3.8 0

Table 5. Results for the Fresh Material with Left-HandedSeeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC1 0.4 6.3 6.7 94.02 0.6 9.1 9.7 94.03 0.3 5.6 5.9 95.0mean 0.4 7.0 7.4 94.3SD 0.2 1.9 2.0 0.6STC1 0 15.9 15.9 1002 0 0.9 0.9 1003 0 5.7 5.7 100mean 0 7.5 7.5 100SD 0 7.7 7.7 0

Table 6. Results for the Recycled Material with Right-Handed Seeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC1 0.8 0.1 0.9 88.92 6.9 0.2 7.1 97.23 4.8 1.4 6.2 77.4mean 4.2 0.6 4.7 87.8SD 3.1 0.7 3.4 9.9STC1 1.0 0 1.0 1002 0.5 0 0.5 1003 1.4 0 1.4 100mean 1.0 0 1.0 100SD 0.5 0 0.5 0

Table 7. Results for the Recycled Material with Left-HandedSeeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC1 0.4 7.2 7.6 94.72 0.4 6.3 6.7 94.03 1.3 6.9 8.2 84.1mean 0.7 6.8 7.5 90.9SD 0.5 0.5 0.8 5.9STC1 0 0.3 0.3 1002 0 0.5 0.5 1003 0 0.4 0.4 100mean 0 0.4 0.4 100SD 0 0.1 0.1 0

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322529

seed crystal could introduce either a seed of oppositehandedness to that intended or by chance a heterogeneousforeign body through which a crystal of opposite handedness tothe seed could appear. This seems highly unlikely as not onlywas the single seed crystal thoroughly rinsed with distilledwater prior to use but this effect would have been seen in boththe STC and the OBC, not exclusively in the OBC.On examination of the way mixing is achieved in the two

systems, i.e., stirring vs oscillating, one unique feature in theOBC stands out: the likely scraping action on the wall of thecolumn when the baffles are moving up and down the liquidphase. This interaction could cause sufficient friction to induceprimary nucleation which resulted in the mixture of left- andright-handed crystals in the product. Could the scraping actionbe the culprit for the primary nucleation?In order to test this hypothesis we redesigned the STC and

OBC on purpose in that the stirrer in the STC was largeenough to cause the scraping effect. Meanwhile, the bafflediameter is smaller than the tube diameter, ensuring that thescraping effect would be eliminated in the OBC. If ourhypothesis is correct then the reverse results would prevail, i.e.,100% of the product crystals from the OBC would bear theseed handedness, while less than 100% similarity to the seedwould be seen in the STC.For the consistency of material selections, both the scraping

baffles and the scraping stirrer were made of PTFE while thevessels were of glass. In this way, any potential discrepancies onthe effect of scraping due to material indifferences45 areeliminated. The results are shown in Tables 8 and 9.

From the above data it can be seen that on introduction ofthe scraping stirrer to the STC there are now instances wherethe product crystals display less than 100% similarity to theseed crystal (trials 1 and 2 in Table 8, trial 1 in Table 9),although there are still a few cases where complete similarity tothe seed is maintained.For the OBC where there was a gap between the baffle and

the tube, the results did show a shift in similarity to seed towardthe 100% level, as seen with the unscraped STC testsperformed earlier. These initial data are encouraging. In orderto further reinforce the outcome, we repeated the experimentsagain utilizing the recycled starting material. The results fromthese tests are shown in Tables 10 and 11.The results are indeed interesting; they partially verify our

hypothesis, i.e., it is the scraping action in the OBC that was the

means of generating the primary nucleation. It should be notedthat it was difficult to source a stirrer that would generate aneven scraping effect on the internal wall of the STC due to theuneven curvature. This may explain the smaller percentage shiftthan compared to the original OBC trials. In addition,increasing the gap between the baffles and the vessel wall inthe OBC effectively reduced the overall mixing;46 this mayexplain why a full 100% similarity was not achieved.Our data presented here is complementary to the recent

work of Qian and Botsaris (2004). They investigated the effectof the morphology of the seed crystal on the observed

Table 8. Results for the Fresh Material with Right-HandedSeeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC with increased gap1 2.5 0.01 2.5 99.62 4.4 0 4.4 1003 4.7 0.07 4.8 98.5mean 3.9 0.03 3.9 99.4SD 1.2 0.04 1.2 0.8STC with scraping stirrer1 0.2 0.01 0.2 95.22 0.5 0.01 0.5 98.03 5.1 0 5.1 100mean 1.9 0.01 1.9 97.7SD 2.8 0.01 2.7 2.4

Table 9. Results for the Fresh Material with Left-HandedSeeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC with increased gap1 0.2 3.8 4.0 942 0 4.0 4.0 1003 0 4.2 4.2 100mean 0.1 4 4.1 98.0SD 0.1 0.2 0.1 3.5STC with scraping stirrer1 0.01 0.4 0.4 97.62 0 5.9 5.9 1003 0 1 1 100mean 0 2.4 2.4 99.2SD 0.01 3.0 3.0 1.4

Table 10. Results for the Recycled Material with Right-Handed Seeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC with increased gap1 1.2 0.4 1.5 76.12 3.4 0.5 3.9 87.23 0.5 0.1 0.7 82.3mean 1.7 0.3 2.0 81.9SD 1.5 0.2 1.7 5.6STC with scraping stirrer1 0.1 <0.1 <0.2 71.52 0.2 <0.1 <0.3 83.03 2.0 0.2 2.2 90.8mean 0.8 0.1 0.9 81.7SD 1.1 0.1 1.2 9.7

Table 11. Results for the Recycled Material with Left-Handed Seeds

trialmass right

(g)mass left

(g)total mass

(g)similarity to seed

(%)

OBC with increased gap1 0.1 2.0 2.1 95.32 <0.1 3.9 4.0 99.13 0.1 2.1 2.2 95.0mean 0.1 2.7 2.8 96.5SD <0.1 1.1 1.0 2.3STC with scraping stirrer1 <0.1 0.5 <0.6 94.62 0.6 3.7 4.3 86.13 0.5 0.9 1.4 67.6mean 0.4 1.7 2.1 82.8SD 0.3 1.7 1.9 13.8

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322530

nucleation mechanism. Using their proposed embryo coagu-lation secondary nucleation (ECSN) model, where embryos ofboth handedness can exist in the boundary layer of the seedcrystal,30 they found that the percentage of the product crystalswith the same handedness as the seed was highly sensitive tothe crystal surface of the seed as well as depending on thesupercooling. They concluded that the degree of supercoolingrequired to switch from conventional secondary nucleation, i.e.,by attrition, to the ECSN model was dependent on the seedpreparation method.47 While Qian and Botsaris studied theseeded crystallization of sodium chlorate with regard to seedpreparation and supercooling, we investigated the effect of thehydrodynamic environment of crystallizers on crystal similarityand observed a change in the nucleation mechanism.The final set of three experiments that we performed

involved no seeds but kept the agitation running overnight inboth STC and OBC. This is to further consolidate our finding.After agitating overnight at 30 °C, the STC had no productcrystals present as we would expect. In the OBC, however, itwas clear that nucleation had occurred; crystals were present inthe system. In one instance out of the three, the OBC had notnucleated overnight; instead, it occurred after 2 days. At such alow supercooling and fairly gentle mixing, we consider thisresult reasonable. Analysis of the crystals’ handedness revealedthat there existed a combination of left- and right-handedcrystals, although actual quantification was difficult and timeconsuming due to the small size of the crystals obtained. Thisresult once again supports our hypothesis that the scraping isthe cause of nucleation in the OBC.

4. CONCLUDING REMARKSOne of the objectives for the initial work is to seek scientificunderstanding as to why seeds were not necessary in somecrystallization processes in the OBC while seeds were essentialin the STC for the same processes and at the same operationalconditions. The starting point for our research was on themixing mechanism, while the chemistry (the model compound)and other conditions such as the procedures and operationsremained unchanged. The experimental data showed that theSTC produced crystals exclusively of the handedness of theseed crystal whether the mixing was applied or not. The OBC,however, produced crystals of mixed handedness when themixing was applied. We confirmed our first hypothesis that themixing mechanism was the key driving force for such adifference.By further hypothesizing that the way mixing is achieved in

the system could be the cause for the primary nucleation, weinvestigated the presence of a scraping action in eachcrystallizer. The data generated so far are highly supportive ofour theories with scraping in the OBC being the sole cause ofnucleation in some experiments. This is just the start of ourjourney. We plan to carry out further tests at different mixingintensities, supersaturations, and seed loads and, moreimportantly, without seeds in order to authenticate our currentfindings and examine the validity of the ECSN model in theOBC. These will be addressed in our future correspondences.

■ AUTHOR INFORMATION

Corresponding Author*E-mail: x.ni@hw.ac.uk.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The authors thank the Scottish Funding Council for theSPIRIT award and Heriot-Watt University for the matchingfunding. Thanks are expressed to Chris Price and Alex Caillet ofGSK, Gerry Steele and Amy Robertson of Astra Zeneca, KevinGirard of Pfizer, Thomas Rammeloo of Johnson and Johnson,Jan Sefcik of the University of Strathclyde, and the otherAcademics of the Continuous Crystallization Discussion Group(CCDG) for the most helpful monthly meetings.

■ REFERENCES(1) Garside, J.; Davey, R. J. Chem. Eng. Commun. 1980, 4, 393−424.(2) Vekilov, P. G. Cryst. Growth Des. 2010, 10 (12), 5007−5019.(3) Myerson, A. Handbook of Industrial Crystallization, 2nd ed.;Butterworth Heinemann: New York, 2002.(4) Mersmann, A. Crystallization Technology Handbook, 2nd ed.;Marcel Dekker: New York, 2001.(5) Mullin, J. W., Crystallization, 3rd ed.; Butterworth-Heinemann:Oxford, 1992.(6) Tung, H.; Paul, E. L.; Midler, M.; McCauley, J. A. Mixing Effectson Crystal Growth. In Crystallization of Organic Compounds, AnIndustrial Perspective; John Wiley & Sons: New York, 2009; pp 122−123.(7) Ni, X.; Liao, A. Cryst. Growth Des. 2008, 8 (8), 2875−2881.(8) Liang, K.; White, G.; Wilkinson, D.; Ford, L. J.; Roberts, K. J.;Wood, W. M. L. Cryst. Growth Des. 2003, 4 (5), 1039−1044.(9) O’Grady, D.; Barrett, M.; Casey, E.; Glennon, B. Chem. Eng. Res.Des. 2007, 85 (7), 945−952.(10) Chen, P.-C.; Cheng, G. Y.; Kou, M. H.; Shia, P. Y.; Chung, P. O.J. Cryst. Growth 2001, 226 (4), 458−472.(11) Ni, X. Innovations Pharm. Technol. 2006, 20, 90−96.(12) NiTech Solutions Ltd. http://www.nitechsolutions.co.uk(March 18, 2011),(13) Ni, X.; Mackley, M. R.; Harvey, A. P.; Stonestreet, P.; Baird, M.H. I.; Rama Rao, N. V. Trans. IChemE 2003, 81 (A), 373−383.(14) Ni, X.; Jian, H.; Fitch, A. Trans. IChemE 2003, 81 (A), 842−853.(15) Hewgill, M. R.; Mackley, M. R.; Pandit, A. B.; Panu, S. S. Chem.Eng. Sci. 1993, 48 (4), 799−809.(16) Mackley, M. R.; Tweedle, G. M.; Wyatt, I. D. Chem. Eng. Sci.1990, 45 (5), 1237−1242.(17) Chew, C. M.; Ristic, R. I.; Dennehy, R. D.; De Yoreo, J. J. Cryst.Growth Des. 2004, 4 (5), 1045−1052.(18) Lawton, S.; Steele, G.; Shering, L.; Zhao, L.; Laird, I.; Ni, X. Org.Process Res. Dev. 2009, 13, 1357−1363.(19) Ni, X.; De Gelicourt, Y. S.; Baird, M. H. I.; Rama Rao, N. V.CJChE 2001, 79 (3), 444−448.(20) Ni, X.; Liao, A. Chem. Eng. J. 2010, 156, 226−233.(21) Chew, C. M.; Ristic, R. I. AIChE J. 2005, 51 (5), 1576−1579.(22) Ni, X.; Price, C.; Girard, K.; Havekost, D.; Marziano, I. Monthlymeetings of Continuous Crystallization Discussion Group. PrivateCommunications, 2010.(23) Mandare, P. N.; Pangarkar, V. G. Chem. Eng. Sci. 2003, 58 (7),1125−1133.(24) Bunn, C. W. Shapes of Crystals. In Chemical Crystallography;Oxford Clarendon Press: Oxford, 1945; p 39.(25) Pagni, R. M.; Compton, R. N. Cryst. Growth Des. 2002, 2 (4),249−253.(26) Mayer, C.; Lacmann, R. Chem. Eng. Technol. 1997, 20 (9), 633−640.(27) Denk, E. G.; Botsaris, G. D. J. Cryst. Growth 1972, 13/14, 493−499.(28) Denk, E. G.; Botsaris, G. D. J. Cryst. Growth 1972, 15, 57−60.(29) Martin, B.; Tharrington, A.; Wu, X. l. Phys. Rev. Lett. 1996, 77(13), 2826−2829.(30) Qian, R.; Botsaris, G. D. Chem. Eng. Sci. 1997, 52 (20), 3429−3440.

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322531

(31) Sung, C. Y.; Estrin, J.; Youngquist, G. R. AIChE J. 1973, 19,957−962.(32) Evans, T. W.; Margolis, G.; Sarofim, A. F. AIChE J. 1974, 20 (5),950−958.(33) Ayazi Shamlou, P.; Jones, A. G.; Djamarani, K. Chem. Eng. Sci.1990, 45 (5), 1405−1416.(34) Liang, K.; White, G.; Wilkinson, D.; Ford, L. J.; Roberts, K. J.;Wood, W. M. L. Ind. Eng. Chem. Res. 2004, 43 (5), 1227−1234.(35) Sinnott, R. K. Equipment Selection, Specification and Design. InCoulson and Richardson’s Chemical Engineering, 3rd ed.; Sinnott, R. K.,Ed. Butterworth-Heinemann: Oxford, 1999; Vol. 6 (ChemicalEngineering Design), pp 471−473.(36) Nienow, A.; Miles, D. Ind. Eng. Chem. Process Des. Dev. 1971, 10(1), 41−43.(37) Baird, M. H. I.; Stonestreet, P. Trans. IChemE 1995, 73 (5),503−511.(38) Smith, A. A. Org. Process Res. Dev. 1997, 1 (2), 165−167.(39) Seidell, A. Solubility of Sodium Chlorate. In Solubilities ofinorganic and metal organic compounds, 3rd ed.; Van Nostrand Co.:New York, 1940; Vol. 1, p 1250.(40) Ristic, R. I. Chem. Eng. Res. Des. 2007, 85 (7), 937−944.(41) Vogl, O. J. Polym. Sci., Part A: Polym. Chem. 2011, 49 (6), 1299−1308.(42) Virone, C.; Kramer, H. J. M.; van Rosmalen, G. M.; Stoop, A.H.; Bakker, T. W. J. Cryst. Growth 2006, 294 (1), 9−15.(43) Alexander, A. J.; Camp, P. J. Cryst. Growth Des. 2008, 9 (2),958−963.(44) Kaisersberger, E.; Knappe, S.; Opfermann, J.; Mohler, H. J.Therm. Anal. 1993, 40 (2), 821−829.(45) Shah, B. C.; McCabe, W. L.; Rousseau, R. W. AIChE J. 1973, 19(1), 194−194.(46) Ni, X.; Stevenson, C. C. J. Chem. Technol. Biotechnol. 1999, 74(6), 587−593.(47) Qian, R.; Botsaris, G. D. Chem. Eng. Sci. 2004, 59 (14), 2841−2852.

Crystal Growth & Design Article

dx.doi.org/10.1021/cg300135w | Cryst. Growth Des. 2012, 12, 2525−25322532