Batch and Tubular-Batch Crystallization of Paracetamol: Crystal Size

May 4, 2006 - Moreover, the total run time required to relieve all of the generated supersaturation (i.e., produce a fixed mass of crystals) was signi...
1 downloads 0 Views 412KB Size
Batch and Tubular-Batch Crystallization of Paracetamol: Crystal Size Distribution and Polymorph Formation Jose´ R. Me´ndez del Rı´o and Ronald W. Rousseau* School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0100

CRYSTAL GROWTH & DESIGN 2006 VOL. 6, NO. 6 1407-1414

ReceiVed January 16, 2006; ReVised Manuscript ReceiVed March 28, 2006

ABSTRACT: The crystallization of paracetamol from ethanol and methanol solutions was used to examine the possibility of using rapid cooling in a tubular flow-through apparatus to manipulate crystal size distribution and morphology. Flow through the tubular portion of the apparatus was laminar, and hence, the device, including a receiving vessel, is referred to as a laminar-flow tubular crystallizer (LFTC). Undersaturated solution entered the tubular portion of the LFTC and was rapidly cooled to temperatures far below solubility conditions. Experimental results were also obtained using a typical batch configuration with the same solute and solvents, but with significantly lower cooling rates than in the LFTC. The crystals produced in the flow-through apparatus were of smaller mean size than those obtained from batch crystallizations, and evidence was found that using the rapid-cooling technique could lead to the generation of kinetically stable polymorphs. Moreover, the total run time required to relieve all of the generated supersaturation (i.e., produce a fixed mass of crystals) was significantly less with the flow-through device than with the batch unit. Introduction Crystallization is a common operation used to separate and purify pharmaceutical compounds, specialty chemicals, and many other chemical products. In attempts to meet specific requirements on crystal size distribution or to control polymorph formation, many industrial firms have attempted to improve their processesthroughtheuseofnovelequipmentandmethodologies.1-3 Crystal size distribution and morphology are especially important in the pharmaceutical industry, where such properties are related to drug dissolution rate and tableting properties.4-7 When the characteristic crystal size (perhaps the mean or dominant size) is greater than desired, another downstream process such as milling is required to reduce it. Unfortunately, milling can have deleterious effects on the properties of the crystals, transforming them in some instances into other polymorphs or amorphous materials.8 Differing physical properties of polymorphs of the same molecular species affect bioavailability, encapsulation, and dissolution rates. Because obtaining an undesired polymorph can negatively impact the marketability of a product (see ref 9 for an example), there has been a widespread effort to control polymorphism by changes of solvents, crystallizer design,10-12 seeding,13,14 additives,15-17 and other methods. Brenek et al.12 addressed crystal size and polymorphism by the use of a laminar-flow tubular crystallizer (LFTC) for the crystallization of a pharmaceutical product. They were able to produce and isolate an unstable polymorph of a pharmaceutical product utilizing such a system without concomitant production of the stable form, while at the same time producing crystals of smaller size than those produced in a batch crystallizer. The present work examines the concepts suggested in the work of Brenek et al.12 by comparing the outcomes from paracetamol crystallizations performed in units comprised of a laminar-flow tubular crystallizer (LFTC) and batch stirred-tank vessels. The LFTC resembles a tube-in-tube heat exchanger which transports the solution of interest through the inner tube, while a coolant flows countercurrently through the outer tube. In the system used in the present work, the small size of the inner tube (3.2-mm o.d.) compared to the size of the outer tube (11.1-mm i.d.), the length of the crystallizer (7.62 m), and the

volumetric flow rates of the fluids (10.8 to 47.2 mL/min inside, 1.27 L/min outside) provide efficient heat transfer between the fluids with a minimum temperature change in the cooling fluid. It is recognized that high cooling rates of supersaturated solutions increase the width of the metastable zone. However, a very rapid cooling rate involving a large temperature reduction in the supersaturated solution can push the system to cross the metastable limit, thereby promoting uncontrolled crystallization and the formation of fines. Often in batch crystallization, this is not desired; instead, a constant supersaturation may be used to provide a narrow size distribution with a large mean crystal size, sometimes aided by seeding.18 On the other hand, the LFTC may be able to produce crystals with a smaller size than typically obtained from well-mixed batch systems because of the high supersaturation achieved shortly after the solution enters the LFTC. Paracetamol (acetaminophen) is a widely used antipyretic and analgesic often found in over-the-counter drugs. Three polymorphs of paracetamol have been identified, with the monoclinic crystal (form I) being stable at room conditions. Much attention has been focused on the production of form II, an orthorhombic crystal, whose sliding planes allow it to have much better compressibility than form I.6 The monoclinic form used in drug products lacks sliding planes and requires binding agents for the tableting process, which adds cost to its production.4,6 Form II has been obtained by crystallization from melts,6 yet this process is impractical for production on a large scale. For this reason, Nichols and Frampton4 focused on the production of form II from ethanol solutions in laboratory-scale batch crystallizers with the assistance of seed crystals obtained by crystallization from the melt. They found that an unseeded batch would yield form I and that solvent-mediated transformation from form II to I occurred readily. They recommended crystallization at temperatures below 5 °C to retard the transformation process. These difficulties led Nichols and Frampton to note the need for extensive optimization in scaling up the process. Others subsequently optimized process variables in an attempt to maximize production of form II in batch crystallization.14,19 Form III crystals have been obtained by crystallization from the melt, but their isolation for characterization has been elusive.4, 6, 20

10.1021/cg060025v CCC: $33.50 © 2006 American Chemical Society Published on Web 05/04/2006

1408 Crystal Growth & Design, Vol. 6, No. 6, 2006

Me´ndez del Rı´o and Rousseau Table 1. Reynolds Numbers Used in the LFTC ethanol

Figure 1. Laminar-flow tubular crystallizer (LFTC) apparatus. Dashed lines show flow of fluids that controlled temperatures; solid lines show flow of paracetamol solutions. The photograph at the top of the figure shows the tubular device through which solution was pumped as it was fed to the receiver.

Experimental System Materials. Paracetamol (4-acetamidophenol, 98%) was purchased from Acros Organics. It was crystallized in the present work from ethanol purchased as 200 proof from Equistar Chemicals LP and from methanol USP purchased from Chemcentral. Batch Crystallization. A jacketed 500-mL glass vessel (9.68-cm i.d.) with a curved bottom was used to carry out the batch crystallizations. A four-port head allowed insertion of a stirrer, a focused beam reflectance measurement (FBRM) probe, a thermocouple, and a condenser. The solution was stirred with a four-blade, 5.1-cm diameter stainless steel propeller with 35° pitch rectangular blades that pumped the contents upward for good contact with the FBRM probe. The stirrer speed was maintained at 400 rpm; this provided good mixing with Reynolds numbers ranging from 10 000 to over 43 000 (depending on temperature and solvent), while avoiding excessive splashing. Three 0.635-cm diameter stainless steel baffles were also used to promote mixing and minimize vortex formation. An FBRM D600 was used to determine chord-length distributions (CLDs) of crystals in the system. The scanning rate was set to 2 m/s, with a 10-s measurement duration and a moving average set to 10 measurements. The probe also served as an additional baffle in the system. A condenser with a cooling fluid at about 10 °C was attached to the vessel to return any solvent that evaporated. A thermocouple was submerged in the solution to record temperature, which was acquired through an Omega OMB-DAQ-56 connected to a computer. The desired cooling rates were programmed on a VWR 1157P circulator, which provided the temperature control to the vessel. The upper part of the vessel, between the jacketed area and the condenser, was wrapped with a heating cloth maintained at about 85 °C (higher than the highest solution temperature) to avoid condensation of the solvent against the walls and the top of the vessel. Both the solute and solvent under study were introduced at room temperature into the batch vessel. An amount was fed that ensured that the FBRM probe tip was submerged in the resulting solution. The vessel then was heated until full dissolution of all crystals was achieved; it then was cooled to generate crystals, heated again to the desired set point temperature, and held at that temperature for 30 min prior to being cooled at a specific rate for the first experiment in a campaign. FBRM data and temperatures were recorded from the constant hightemperature period until the cooled slurry reached equilibrium as determined by the FBRM mean and median chord lengths reaching constant values. A subsequent run was initiated by heating the slurry to the high-temperature set point, where it was again held for 30 min, before cooling to start the next run at a prescribed cooling rate. LFTC Equipment. Figure 1 shows the system into which the LFTC was placed. A small vessel, having an approximate volume of 75 mL, was used to contain the solvent for flushing purposes. A 250-mL baffled

methanol

flow rate (mL/min)

NRe

flow rate (mL/min)

NRe

10.8 26.7 42.9

220 530 850

15.6 30.4 47.2

470 920 1420

glass vessel contained the feed solution. The solution and/or solvent were pumped through the tubular crystallizer with a FMI QG400 positive-displacement pump that had a jacketed head which facilitated maintaining the solution temperature at a desired value. The tubular crystallizer was made of a 7.62-m-long inner PTFE tube (1.6-mm i.d., 3.2-mm o.d.) that transported the solution; the inner tube was surrounded by a 11.1-mm i.d. (12.7-mm o.d.) PTFE tube through which the heat-transfer fluid could be pumped either countercurrently or cocurrently to the solution. The LFTC was coiled in a circular fashion and insulated by two layers of 0.95-cm-thick elastomeric foam which minimized the heat transfer from the surroundings. The receiving vessel was used as the crystallizer in the batch experiments. Temperatures were adjusted through the use of thermobath circulators maintained at the desired temperatures. A heating tape maintained at the higher temperature was used between the outlet of the solution vessel and the three-way valve to maintain the solution temperature. Temperatures of the solution and receiving vessels and the inlets and outlets to the LFTC of solution and heat transfer fluid were recorded through an Omega OMB-DAQ-56 data acquisition system connected to a computer. LFTC Procedure. Prior to starting an experiment, the thermobath circulator temperatures were set at a high value for the pump head and the solution and solvent vessels, and at a low temperature for the crystallizer and receiver. Similar amounts of solvent and solute that had been used in batch experiments were added to the solution vessel after the temperature set points were reached. Solvent and solute were stirred for a period of 30 min while the solute was dissolved in its entirety. While the solution was stirred, the LFTC was flushed with warm solvent from the solvent vessel. This solvent was sent from the outlet of the LFTC to another vessel, such as a beaker, and discarded. After the solvent was pumped and no more was observed coming through the outlet of the LFTC, the pump was turned off and the outlet tube was redirected into the receiving vessel. After the solution had been stirred for 30 min, data collection was started. The solution was then pumped through the crystallizer by switching the three-way valve to the solution vessel, opening the lower valve of the vessel, and turning on the pump. Typical flow rates ranged from 10.8 to 47.2 mL/min. Table 1 shows the Reynolds numbers for flow through the LFTC for the flow rates examined. To minimize the presence of remaining solvent from the flushing stage, the first drops coming out of the LFTC were discarded through the lower valve of the vessel. Following removal of these drops of liquid, no solution flowed into the receiving vessel for a short period of time (seconds). The cold solution process solution then entered the receiving vessel and data collection from the FBRM continued while the supersaturation was consumed through crystallization. A run was considered complete when the system reached steady state, which was defined to be when the mean and median chord lengths became constant. Estimation of Nucleation Kinetics. As part of the characterization process, a Lasentec FBRM D600 and online temperature recording were used to acquire chord-length and nucleation-relevant data in both batch and LFTC operations. For the purpose of comparing crystal sizes from different runs, we will use the median and mean values of the chord-length distribution (CLD) obtained by the FBRM. No weighting was applied to the data, and therefore, the median and mean represent number average quantities. A method to approximate the nucleation kinetics from FBRM measurements was developed based on our earlier work. We define a chord-count density function c(l) as

c(l) ) lim

∆lf0

∆C(l) V h S∆l

(1)

where ∆C(l) is the number of chord counts with lengths between l and l + ∆l in the volume of slurry scanned by the FBRM probe, V h S. This volume is estimated from scanning specifications of the probe: that is, the depth that the laser beam effectively penetrates λ, its width b,

Batch and Tubular-Batch Crystallization of Paracetamol

Crystal Growth & Design, Vol. 6, No. 6, 2006 1409

Figure 2. Use of the zeroth moment of the CLD to determine nucleation rate. The given data pertain to a run in which paracetamol was crystallized from a solution in methanol solution in the LFTC. and its scan speed along the path V˘S. The volume scanned over any period of time ∆t can be approximated by

V h S ) V˘Sλb∆t

(2)

Values used in this work for the aforementioned variables were λ ) 1.50 × 10-3 m, b ) 5.80 × 10-6 m, and V˘S ) 2 m/s. Since there is great uncertainty in these values, the kinetics evaluated in the present work may be considered in relative terms; that is, the primary purpose of the kinetic measures was to compare the nucleation rates for a given set of conditions to those exhibited under other conditions. The scanned volumes in these cases should be constant, and therefore, the actual values are irrelevant in the comparisons. The zeroth moment of the population density function is identical to the total number of particles per unit volume.21 If it is assumed that the beam from the probe is scattered from all particles scanned in the volume and that none of the particles are scanned more than once, then the number of chords counted over a given time interval is identical to the total number of particles in the volume scanned. Therefore, by drawing an analogy between the chord-length density function and the population density function, the zeroth moment m j 0 of the chord-count density function is

m j0 )

∫ c dl ∞

0

(3)

which is identical to the total number of particles per unit volume Ntotal. By neglecting crystal attrition and agglomeration, the nucleation rate J0 in a well-mixed vessel can therefore be estimated as

j0 dNtotal dm J0 ) ) dt dt

Figure 3. Paracetamol crystals obtained from batch crystallization of paracetamol from an ethanol solution cooled at 0.15 °C/min. The sample does not necessarily reflect the actual size distribution of the entire batch.

(4)

By considering CLD data from the FBRM, where counts per second are relayed in specific chord lengths, a simple spreadsheet can be developed in which the zeroth moment is calculated based on the effective scanned volume. The derivatives of the zeroth moments over time would then give the nucleation rate of the system throughout the course of a run. As seen in Figure 2, the derivative of the zeroth moment increases rapidly, reaches a maximum, and then decreases rapidly to a near-zero value. In this study the characteristic nucleation rate in a given run was defined as the maximum value.

Results and Discussion Solutions of paracetamol in ethanol were prepared by dissolving 59.5 g of the solute in 175 g of the solvent (34.0 g/100 g of ethanol). Data from a recent publication22 were used to estimate the saturation temperature at this concentration as 50.6 °C. For methanol solutions, 90.4 g of paracetamol was dissolved in 162.5 g of the solvent (55.7 g/100 g of methanol). Since solubility data at this concentration were not found, extrapolation using a van’t Hoff fit of data by Granberg and Rasmuson23 resulted in an estimate of the saturation temperature to be 52.6 °C. Fitting van’t Hoff plots of the available solubility

Figure 4. Unweighted paracetamol median and mean chord lengths at steady state. Error bars reflect the standard deviation from the averages of multiple runs.

data resulted in the following correlations:

for methanol-paracetamol ln x ) -

1673.5 + 2.8888 T

R2 ) 0.9980

(5)

R2 ) 0.9942

(6)

for ethanol-paracetamol ln x ) -

1836.8 + 3.3078 T

Batch Crystallization. Batch crystallizations of paracetamol, focusing on the effects of solvent and cooling rate, were initiated at approximately 69.5 °C and cooled to approximately 11.5 °C at cooling rates of 0.15, 0.20, 0.30, and 0.40 °C/min. The crystals in Figure 3 are typical of those recovered from solutions in methanol and in ethanol. They exhibit the characteristic prismatic shape of form I paracetamol. Although it is not intended to be representative of the entire distribution, the figure also shows a typically broad size distribution. Figure 4 shows the median and mean chord lengths attained by paracetamol crystals through crystallization from ethanol and methanol solutions. Values were determined by the FBRM instrument after steady state had been achieved. Lower cooling rates led to greater median and mean chord lengths; in other

1410 Crystal Growth & Design, Vol. 6, No. 6, 2006

Figure 5. Paracetamol chord-length distributions at steady state after crystallization from ethanol solutions.

Figure 6. Relative nucleation rates determined by analysis of zeroth moments of paracetamol crystallized from ethanol and methanol solutions.

words, more rapid cooling promoted the formation of smaller crystals. It is interesting to note that the effect of cooling rate on mean and median sizes was greater when crystallization was from ethanol. For each solvent, the initial paracetamol concentration and the final steady-state temperature were the same. Therefore, for each data set with a given solvent, the final crystal mass of crystals produced should be the same. From the observations presented in Figure 4, it is clear that increased cooling rates led to smaller crystals. The final number of crystals should be greater, then, to accommodate the same crystal mass. Figure 5 shows paracetamol chord-length distributions at the conclusion of runs in which different cooling rates were used. Note that the results are consistent with larger numbers of smaller crystals being produced at higher cooling rates. However, at larger sizes, which correspond to most of the crystal mass, the chord-length distributions nearly overlap. The nucleation rates in each of the runs described were estimated from the zeroth moments of the chord-length density function using the methodology presented earlier. Figure 6 shows how increasing the cooling rates increased nucleation rates; the effect appears to have been greater for solutions in methanol than for those in ethanol. The metastable limit, which is defined as the difference between equilibrium conditions and those at which nucleation occurs, generally increases with the rate at which supersaturation is generated; in batch cooling crystallization, this means that the temperature at which nucleation occurs should decrease as the cooling rate increases. Figure 7 shows results from the systems investigated in the present study that are consistent with the expected behavior. As shown in the figure, there is about a 10 °C difference in the nucleation temperatures observed with

Me´ndez del Rı´o and Rousseau

Figure 7. Average nucleation temperatures for paracetamol crystallized from ethanol and methanol solutions.

Figure 8. Relative supersaturation at nucleation for the crystallization of paracetamol from ethanol and methanol solutions. Relative supersaturations were calculated from nucleation temperatures and solubilities estimated from the van’t Hoff fits of published data.22,23

the highest and the lowest cooling rates, irrespective of the solvent used. The thermodynamic driving force at nucleation can be approximated with the quantity known as relative supersaturation, σ, where

σ)

C - C* C*

(7)

where C is the solute concentration at the prevailing system conditions and C* is the solubility of the solute at the system temperature. The supersaturations at nucleation are shown in Figure 8 for the systems examined. The supersaturation required for nucleation of paracetamol from methanol solutions is significantly lower than that required for nucleation from ethanol solutions. Consistent with the discussion of Figure 7, the relative supersaturation required for nucleation is higher at higher cooling rates, irrespective of the solvent. Extrapolation to the axis of the exponential fits of the data indicates that a significant supersaturation (approximately 0.23) is required for crystallization of paracetamol from ethanol, even with an infinitesimal cooling rate. This contrasts with the nearly negligible supersaturation required for solutions in methanol. Figure 9 shows how the calculated relative supersaturation of paracetamol changed with time prior to nucleation in each of the batch runs. The initial conditions were slightly different in the two solvents, but these differences became less during the course of the runs. The conditions at which nucleation occurred are shown in the figure; for example, the square marker at 120 min represents the relative supersaturation at nucleation from a solution in methanol at σ ) 0.02. It should be pointed

Batch and Tubular-Batch Crystallization of Paracetamol

Figure 9. Relative supersaturation as a function of time and cooling rate for paracetamol in ethanol and methanol solutions.

Figure 10. Final chord-length distributions for paracetamol crystallized from ethanol solutions at different batch cooling rates and in the LFTC.

out that the calculations of supersaturation after nucleation do not reflect the actual conditions in the solutions; that is, supersaturation begins to be removed through nucleation and growth of paracetamol crystals. Laminar-Flow Tubular Crystallizer. As with the batch crystallization experiments, the initial solution temperature in the LFTC runs was approximately 69.5 °C. The temperature in the jackets of the LFTC and the receiving vessel was approximately 11.5 °C. As mentioned earlier, the feed solutions contained 34.02 g of paracetamol/100 g of ethanol and 55.66 g of paracetamol/100 g of methanol. Each of the LFTC runs used the same mass of starting solution as was used in the batch runs. The flow rates through the tubular crystallizer were 10.8, 26.7, and 42.9 mL/min for ethanol solutions and 15.6, 30.4, and 47.2 mL/min for methanol solutions. Calculations have estimated that the bulk temperature of the solution approached that of the wall after traveling less than 1 m in the tube.24 No crystals were detected, either by visual inspection or FBRM readings, in the solutions leaving the crystallizer tube; however, nucleation was detected by the FBRM shortly after the cold solutions entered the receiving vessel. For ethanol solutions, the median and mean chord lengths at the final steadystate conditions were recorded as 15.02 ( 0.35 µm and 19.30 ( 0.34 µm, respectively. These values were 14.71 ( 0.42 µm and 20.31 ( 0.39 µm when the solvent was methanol. The flow rate through the tubular part of the apparatus had no significant effect on the median or mean chord lengths, irrespective of solvent. Figure 10 superimposes the CLDs measured at the final conditions in a typical LFTC run on ethanol solutions onto data obtained for different cooling rates in the corresponding batch

Crystal Growth & Design, Vol. 6, No. 6, 2006 1411

Figure 11. Final chord-length distributions for paracetamol crystallized from methanol solutions at different batch cooling rates and in the LFTC.

Figure 12. Relative nucleation rates (nucleation rate in the LFTC divided by the nucleation rate in the batch crystallizer at the given cooling rate) showing the effects of solution composition and cooling rate.

crystallizations. Figure 11 provides the same information for crystallizations from methanol. The LFTC curves in these figures show an increase in count rate as well shifts to smaller chord lengths. The median and mean CLDs were about 3 µm smaller than those attained by batch crystallization from ethanol solutions at a cooling rate of 0.40 °C/min, and they were about 6 µm smaller than those obtained from methanol solutions at the same cooling rate. The differences are even greater when compared to batch crystallizations at lower cooling rates. Supersaturation was generated rapidly in the LFTC apparatus; in essence, the solution flowing through the tube was cooled rapidly to the final temperature reached in the batch experiments. Relative supersaturations at these temperatures were estimated to be 1.30 and 1.23 for ethanol and methanol solutions, respectively. These values are significantly higher than the supersaturations shown in Figure 8. Despite generating high supersaturations in the tubular crystallizer, crystals were not observed until shortly after the solution entered the receiver. It is possible that the rapid cooling expanded the metastable zone, similar to the observations on which Figure 8 is based; alternatively, the residence time in the tubular unit after nucleation may have been insufficient for crystals to grow to an observable size. In any case, the high supersaturations generated in the LFTC promoted the formation of crystals that were significantly smaller than those formed in the batch experiments. As shown in Figure 12, the relative nucleation rates were up to 1.8 times those obtained for batch crystallizations from ethanol, while they were between 2.0 and 7.6 times greater when the solvent was methanol. These variations result from operating with different cooling rates in the batch system, as there was only one value

1412 Crystal Growth & Design, Vol. 6, No. 6, 2006

Me´ndez del Rı´o and Rousseau

Figure 13. Paracetamol crystals obtained by crystallization from a methanol solution using the LFTC apparatus with a flow rate of 30.4 mL/min and a cooling temperature of 11.5 °C. Circles highlight needleshaped crystals thought to be the form II polymorph. Table 2. Run Times To Reach the Final Conditions for Paracetamol Crystallization from Ethanol and Methanol Solutions in Batch and LFTC Systems run time ethanol solutions batch crystallization cooling rate 0.15 °C/min 0.20 °C/min 0.30 °C/min 0.40 °C/min LFTC crystallization

9.0 h 6.8 h 5.8 h 4.6 h 30-40 min

methanol solutions 7.4 h 4.7 h 2.2 h 8-12 min

for the LFTC. The effect of solvent on the relative nucleation rates cannot be explained by differences in solubility. For example, if it is assumed that the solutions are cooled to the final temperature of 11.5 °C prior to nucleation, the relative supersaturations at those conditions would be 1.12 for ethanol and 1.27 for methanol. Design Considerations. An interesting outcome of the experiments involving the LFTC is that it was possible to obtain yields of paracetamol that were similar to those achieved from batch crystallizations, but with a much reduced run time. In other words, the typical protocol in batch crystallization is to cool the crystallizer contents at a relatively slow rate in order to obtain crystals of sufficient quality, which means purity, size distribution, and morphology. In the LFTC operation, however, such quality appears to have been maintained, albeit with the production of smaller crystals. Table 2 provides a comparison of the run times in the present work. Note that batch crystallizations can take hours, depending upon the cooling rate, while crystallizations in the LFTC system can be performed in a matter of minutes. Of course, the experiences with paracetamol and the two solvents examined may not extrapolate to other systems. For example, systems that contain difficult-to-remove impurities may require conditions that ensure slow growth to achieve the necessary purity. Furthermore, in some of the systems examined in the present study, it was impossible to keep the tube from being plugged with crystallizing solute. Paracetamol Polymorphism. Figure 13 shows a photomicrograph of paracetamol crystals obtained by crystallization from a methanol solution using the LFTC apparatus. The flow rate was 30.4 mL/min, and the jacket temperature was 11.5 °C. The sample was obtained about 15 min after nucleation was detected

Figure 14. Powder XRD pattern of paracetamol crystals obtained by crystallization from a methanol solution via the LFTC with a flow rate of 30.4 mL/min and a cooling temperature of 11.5 °C.

by the FBRM. Clearly, the primary morphology is that of the form I polymorph, but the circles on the figure point out crystals that have the needle-shaped morphology of form II crystals that also were found in this run. Figure 14 shows that powder X-ray diffraction (XRD) patterns (Phillips PANalytical X’Pert PRO, Cu KR radiation, 1.541 Å wavelength) obtained from the product of the same run coincide with those of form I.4 The sample from which this pattern was obtained had been vacuum-filtered and left to dry at room conditions for 24 h prior to grinding and XRD analysis. It is likely that form II peaks were not detected because of the great difference in mass fractions of the two polymorphs. Apparently, the use of the LFTC apparatus resulted in the simultaneous production of two polymorphs of paracetamol. The photomicrograph in Figure 13 indicates that form I was favored, but the sample was obtained about 15 min after nucleation had been detected, which means that there could have been significant solvent-mediated transformation of the form II crystals to the more stable form I. The presence of form II crystals in the LFTC experiments in which a cooling temperature of 11.5 °C was used stimulated an attempt to maximize the formation of these crystals by reducing the cooling temperature to 3 °C. The reduced cooling temperature was selected based on the observations reported earlier by Nichols and Frampton,4 who recommended crystallization at temperatures below 5 °C to retard the transformation process, although the solvent they used was ethanol. Figure 15 shows photomicrographs of two different samples of paracetamol crystals; the first (on the left) was of a sample obtained shortly after crystallization, well before the system reached the final steady-state conditions. It clearly shows the presence of both forms I and II of paracetamol. Form II crystals are small in comparison to some of the prismatic form I crystals present in the slurry. This suggests that form II crystals nucleated and grew before dissolving as part of a solvent-mediated transformation in which form I crystals grow at the expense of the form II crystals. This process is similar to that observed by Nichols and Frampton4 for paracetamol crystals in ethanol solutions. The photomicrograph on the right side of Figure 15 is of another sample taken from the same run as that on the left side of the figure, but it was obtained 15 min after steady-state conditions in the receiving vessel had been reached. It shows none of the form II crystals that had been present in abundance in the earlier photomicrograph. This observation confirms that

Batch and Tubular-Batch Crystallization of Paracetamol

Crystal Growth & Design, Vol. 6, No. 6, 2006 1413

Figure 15. Paracetamol crystals obtained from a methanol solution using the LFTC apparatus at a flow rate of 30.4 mL/min and a cooling temperature of 3 °C. The photomicrograph on the left, which is of a sample taken shortly after nucleation, shows a significant fraction of the crystals exhibiting the shape of form II crystals. The photomicrograph on the right was taken 15 min after steady-state conditions in the receiving vessel had been achieved.

the less stable phase II polymorph is transformed to the more stable phase I, probably through a solvent-mediated transformation mechanism. As suggested by Nichols and Frampton,4 paracetamol crystals should be harvested as rapidly as possible if the form II polymorph is to be retained. As noted earlier, there may be good reason for doing so because, as noted earlier, it has been reported to have superior processing properties. The results from the present work show that the LFTC is capable of producing the form II polymorph without the addition of seed crystals, although it has only been observed in the presence of form I crystals. Although using lower temperatures in the LFTC may provide a route to production and stabilization of form II crystals, the extremely low temperatures utilized for that purpose by Nichols and Frampton4 are likely to be unnecessary. They obtained form II from an unseeded batch experiment in an ethanol/methanol (96%/4% v/v) solvent at -75 °C. Conclusions Comparative studies were performed using two types of processes. In one process, referred to as the LFTC apparatus, solutions were rapidly cooled in a tubular device before they were collected and allowed to achieve steady state in a receiver. In the second, the solutions were cooled at different rates in a typical batch crystallizer. Model systems used in the study involved the crystallization of paracetamol from solutions in ethanol and in methanol. Both processes were operated successfully and produced equivalent masses of product crystals. There were, however, significant differences in the product crystal size distributions and the polymorphic forms produced. Rapid cooling in the LFTC resulted in the formation of smaller crystals than those that were produced in the corresponding batch cooling crystallization. The larger surface area generated with higher nucleation kinetics made it possible to grow crystals of reasonable quality at rapid rates. This meant that the operating time required to go from initial to final conditions could be dramatically reduced, often by more than an order of magnitude. Clearly, the higher heat-transfer coefficients associated with forced convection contributed to this advantage. Of course, if large crystals are essential in the product specifications, the advantage of reduced operating time is overwhelmed by the disadvantage of producing smaller crystals.

The rapid cooling utilized in the LFTC also resulted in the formation of a kinetically stable polymorphic form of paracetamol (form II). However, this polymorph was transformed in favor of form I as the system approached equilibrium. Rapid sequestration of the unstable form and perhaps further steps would be required in order to maintain form II for use. Nevertheless, the possibility of extending the observed behavior to other systems exists and in some instances it is likely to provide a feasible route to useful polymorphic crystal forms. In addition to comparisons between the two crystallizer configurations, variations of specific process variables within each crystallizer type were also investigated. For example, the mean size of paracetamol crystals produced in batch crystallizations was reduced with increasing cooling rates. Furthermore, the role of cooling rate on the metastable limits and the corresponding metastable zones was clearly demonstrated in the batch system; as cooling rates increased, the metastable zone width and resulting nucleation rates increased. Finally, during the course of this research, methodologies were developed for using the FBRM to provide data on nucleation kinetics. The approach used the zeroth moment and the derivative of the zeroth moment of the chord-length distribution function to estimate the total number of crystals per unit system volume and the nucleation rate in the system. Acknowledgment. Financial support of this research by Pfizer Inc. is gratefully acknowledged. Additional assistance was provided through the Cecil J.“Pete” Silas Endowed Chair and the GEM Consortium Fellowship. References (1) Midler, J. M.; Paul, E. L.; Whittington, E. F.; Futran, M.; Liu, P. D.; Hsu, J.; Pan, S.-H. Crystallization method to improve crystal structure and size. U.S. Patent 5,314,506, 1994. (2) Dauer, R.; Mokrauer, J. E.; McKeel, W. J. Dual Jet Crystallizer Apparatus. U.S. Patent 5,578,279, 1996. (3) am Ende, D. J.; Crawford, T. C.; Weston, N. P. Reactive Crystallization Method to Improve Particle Size. U.S. Patent 6,558,435, 2003. (4) Nichols, G.; Frampton, C. S. Physicochemical characterization of the orthorhombic polymorph of paracetamol crystallized from solution. J. Pharm. Sci. 1998, 87 (6), 684-693. (5) Beyer, T.; Day, G. M.; Price, S. L. The prediction, morphology, and mechanical properties of the polymorphs of paracetamol. J. Am. Chem. Soc. 2001, 123 (21), 5086-5094.

1414 Crystal Growth & Design, Vol. 6, No. 6, 2006 (6) DiMartino, P.; Guyot-Hermann, A. M.; Conflant, P.; Drache, M.; Guyot, J. C. A new pure paracetamol for direct compression: The orthorhombic form. Int. J. Pharm. 1996, 128 (1-2), 1-8. (7) Bu¨rger, A.; Henck, J. O.; Hetz, S.; Rollinger, J. M.; Weissnicht, A. A.; Stottner, H. Energy/temperature diagram and compression behavior of the polymorphs of D-mannitol. J. Pharm. Sci. 2000, 89 (4), 457-468. (8) Brittain, H. G.; Fiese, E. F. Effects of pharmaceutical processing on drug polymorphs and solvates. In Polymorphism in pharmaceutical solids; Brittain, H. G., Ed.; Marcel Dekker: New York, 1999; p 331. (9) Chemburkar, S. R.; Bauer, J.; Deming, K.; Spiwek, H.; Patel, K.; Morris, J.; Henry, R.; Spanton, S.; Dziki, W.; Porter, W.; Quick, J.; Bauer, P.; Donaubauer, J.; Narayanan, B. A.; Soldani, M.; Riley, D.; McFarland, K. Dealing with the impact of ritonavir polymorphs on the late stages of bulk drug process development. Org. Process Res. DeV. 2000, 4 (5), 413-417. (10) Shan, G.; Igarashi, K.; Noda, H.; Ooshima, H. Control of solventmediated transformation of crystal polymorphs using a newly developed batch crystallizer (WWDJ-crystallizer). Chem. Eng. J. 2002, 85 (2-3), 169-176. (11) Igarashi, K.; Sasaki, Y.; Azuma, M.; Noda, H.; Ooshima, H. Control of polymorphs on the crystallization of glycine using a WWDJ batch crystallizer. Eng. Life Sci. 2003, 3, 159-63. (12) Brenek, S.; am Ende, D. J.; Schofield, R.; Jerdonek, C. DeliVering the desired particle; Lasentec Users Forum; Charleston, SC, 2002. (13) Doki, N.; Yokota, M.; Kido, K.; Sasaki, S.; Kubota, N. Reliable and selective crystallization of the metastable alpha-form glycine by seeding. Cryst. Growth Des. 2004, 4 (1), 103-107. (14) Al-Zoubi, N.; Malamataris, S. Effects of initial concentration and seeding procedure on crystallisation of orthorhombic paracetamol from ethanolic solution. Int. J. Pharm. 2003, 260 (1), 123-135.

Me´ndez del Rı´o and Rousseau (15) Kitamura, M. Controlling factor of polymorphism in crystallization process. J. Cryst. Growth 2002, 237, 2205-2214. (16) Davey, R. J.; Blagden, N.; Potts, G. D.; Docherty, R. Polymorphism in molecular crystals: Stabilization of a metastable form by conformational mimicry. J. Am. Chem. Soc. 1997, 119 (7), 1767-1772. (17) Trovao, M. C. N.; Cavaleiro, A. M. V.; de Jesus, J. D. P. Preparation of polymorphic crystalline phases of D-mannitol: influence of keggin heteropolyanions. Carbohydr. Res. 1998, 309 (4), 363-366. (18) Davey, R.; Garside, J. From molecules to crystallizers; Oxford University Press: Oxford, 2000;¢¢ p 81. (19) Al-Zoubi, N.; Kachrimanis, K.; Malamataris, S. Effects of harvesting and cooling on crystallization and transformation of orthorhombic paracetamol in ethanolic solution. Eur. J. Pharm. Sci. 2002, 17 (12), 13-21. (20) Bu¨rger, A. Zur Interpretation von Polymorphie-Untersuchungen. Acta Pharm. Technol. 1982, 28 (1), 1-20. (21) Randolph, A. D.; Larson, M. A. Theory of particulate processes: analysis and techniques of continuous crystallization, 2nd ed.; Academic Press: San Diego, 1988. (22) Worlitschek, J.; Mazzotti, M. Model-based optimization of particle size distribution in batch-cooling crystallization of paracetamol. Cryst. Growth Des. 2004, 4 (5), 891-903. (23) Granberg, R. A.; Rasmuson, A. C. Solubility of paracetamol in pure solvents. J. Chem. Eng. Data 1999, 44 (6), 1391-1395. (24) Mendez del Rio, J. R.; Solubility and phase transitions in batch and laminar-flow tubular crystallizers. M.S. Thesis, Georgia Institute of Technology, 2004.

CG060025V