Yan Chan Su is a graduate student at National Central University, 300 Jhong-Da Rd., Jhong-Li City 320, Taiwan, ROC.

Yan Chan Su

				Part 1 of this article, which appeared in the March 2010 issue of 

, described a robust miniaturized screening strategy to accelerate the preparation and characterization of spherical agglomerates produced by spherical crystallization (1). This approach can be readily adopted or automated using only a small amount of active pharmaceutical ingredient (API), so spherical crystallization can be performed before the submission of a new drug application. The authors described their approach for constructuring form spaces for the nonionizable APIs, carbamazepine, cimetidine, and phenylbutazone by initial solvent screening of 19 common organic solvents to evaluate the feasibility of spherical crystallization. The materials and methods for this study were outlined in Part I of this article (1). Part II of this article discusses the results.    

 The differential-scanning-calorimetry (DSC) identification-test scan of the purchased carbamazepine exhibited the first endothermic peak at 170–174 °C, which corresponded to the melting of Form III (2). An exothermic peak at 175 °C followed, which corresponded to the crystallization of Form I (2). A second endothermic peak at 193 °C corresponded to the melting of Form I (2). The DSC identification-test thermogram of the purchased phenylbutazone displayed a single endothermic peak around 105 °C, which corresponded to the melting of Form A (3). The polymorphism of the purchased carbamazepine and phenylbutazone, therefore, were Form III and Form A, respectively, at room temperature. The identification test for cimetidine was performed using Fourier transform infrared (FTIR) spectroscopy instead of DSC because the melting points of the different forms of cimetidine were very close to each other (4). 

The FTIR spectrum showed four significant infrared peaks for Form A cimetidine at 1204, 1156, 1077, and 954 cm

 and two well-resolved characteristic peaks at 1243 and 1228 cm

, which indicated the polymorphism of the purchased cimetidine at room temperature was Form A (5–7). Two cimetidine molecules formed a dimer unit in Form A with an eight-member ring through the hydrogen bonds between N(10)H of one molecule and the N(12) of the other molecule  (1). The cimetidine molecule could absorb water at the N(1)H group (8). The imidazolium peaks were at 1204 and 954 cm

  was attributed to a CH deformation and a ring deformation (8, 10). The peak at 1156 cm

 was either a coupled ring and a CH-bending mode or the deprotonation of the N(1)H group on the complexation with a water molecule (8, 10). 

 To construct the form space for an API, which contains the possible solvent combinations for discovering a new polymorph, the 19 common solvents for scale-up listed in the materials and methods section of Part I of this article were arranged from top to bottom vertically and from left to right horizontally in ascending order of the solvent's Hildebrand value (1). This arrangement resulted in a 19 × 19 matrix of 361 boxes (see Tables 1(a), 1(b), and 1(c)).  Each box represented a specific binary solvent combination except for the 19 boxes located diagonally in the matrix stepping down from the upper left-hand corner to the lower right-hand corner. The 19 diagonal boxes represented the 19 pure-solvent systems. The yellow boxes were the 

 (i.e, defined as a solvent that gave solubility of ≥ 5 mg/mL at 25 °C) and the red boxes the

 (i.e, defined as a solvent that gave solubility of ≤ 5 mg/mL at 25 °C) with respect to a particular API. The yellow and red boxes were the materials properties of the API.  The gray boxes also were determined experimentally. They represented the immiscible solvent pairs that were independent of the API. Once the locations of the red, yellow,  and gray boxes were fixed, the location of the green, blue, and white boxes representing the antisolvent system (i.e., a good solvent and a bad solvent), the good cosolvent system (i.e,  a good solvent and a good solvent) and the bad cosolvent system (i.e,. a bad solvent and a bad solvent), respectively, were deduced graphically from the diagonal boxes. Since the 19 × 19 matrix was symmetrical along the diagonal boxes, the actual possible solvent combinations were reduced from 361 boxes to a total of 190 boxes or less (i.e., derived from 19 + (361 – 19) ÷ 2 = 190) because it is impractical to perform experiments in the white-boxed areas.    

The actual form space of carbamazepine had 13 yellow, 6 red, 69 green, 78 blue, 14 gray, and 10 white boxes (see Table  I (a)). The actual form space of cimetidine contained 5 yellow, 14 red, 66 green, 10 blue, 14 gray and 81 white boxes (see Table I (b)). The actual form space  of phenylbutazone was composed of 17 yellow, 2 red, 22 green, 135 blue, 14 gray and 0 white boxes (see Table I (c)). The actual form spaces would have been expanded significantly if various solvent compositions of binary mixtures, temperatures, and ternary solvent systems were considered (10–12).

The solubility values of carbamazepine, cimetidine, and phenylbutazone in their corresponding solvents at 25 °C were listed in Table II. Most of the good solvents of Form III carbamazepine belonged to Class 1 (i.e., a protic hydrogen-bond-donating solvents [Lewis acids]), Class 2 hydrogen-bond-acceptor solvents [Lewis bases]), and Class 3 polar aprotic solvents, which were capable of disrupting the N–H•••O hydrogen bonds and the C–H•••O weak interactions of the carboxamide dimer units existing in the solid state of all forms of carbamazepine (2, 13). The good solvents for Form A cimetidine were mainly Class 1 solvents, which had the capability of interacting readily with the weakly basic imidazole ring  and disrupting the N–H•••N hydrogen bonds of the dimer units existing in the solid state of Form A cimetidine (8, 13, 14). The good solvents for Form A phenylbutazone were Classes 1, 2, and 3 solvents, Class 4 chlorocarbon solvents, and Class 5 hydrocarbon solvents. This result implied the intermolecular interactions in the crystal lattice of phenylbutazone were dominated by van der Waals forces (i.e., dipole-dipole interactions and London dispersion force) (15). 

 The spherical agglomeration (SA) method was used in spherical crystallization. Because all miscible pairs of a good solvent and a bad solvent (i.e., an antisolvent) were represented by the green boxes and all immiscible pairs of a bad solvent and a bridging liquid were designated by the gray boxes in the actual form space of each API, the total number of combinations of a good solvent  and a bad solvent and a bridging liquid was equal to (the sum of the number of green boxes in each column with a red box × the number of gray boxes in the same column) and (the number of green boxes in the bottom row with a red box × the number of gray boxes in the same bottom row), if water were also a bad solvent symbolized by a red box situated at the bottom right-hand corner of the actual form space. 

The number of combinations for the SA method based on Table 1(a) for carbamazepine is 183 (i.e., derivied from 10 × 4 + 12 × 2 + 13 × 1 + 13 × 1 + 13 × 1 + 8 × 10 = 183). The number of combinations for the SA method based on Table 1(b) for cimetidine is 100 (i.e., derived from 3 × 4 + 4 × 2 + 5 × 1 + 5 × 1 + 5 × 1 + 5 × 1 + 5 × 1 + 5 × 1 + 5 × 1 + 5 × 1 + 4 × 10 = 100). The number of combinations for the SA method based on Table 1(c) for phenylbutazone was 136 (i.e., derived from 14 × 4 + 8 × 10 = 136).    

Using the SA method, five types of behavior were observed:  					

Type 1: No solid production upon the addition of 10 mL of an antisolvent to 2 mL of an API solution at 25 °C 

Type 2: Dissolution of the solids upon the addition of the bridging liquid at 25 °C 

Type 3: No agglomeration upon the addition of the bridging liquid at 25 °C 

Type 4: Nonspherical agglomeration upon the addition of the bridging liquid at 25 °C 

Type 5: Spherical agglomeration upon the addition of the bridging liquid at 25 °C.    

Type-3 behavior was not observed with phenylbutazone. 

In Type-1 and Type-2 behaviors, the solubility and the width of a metastable zone of an API solid were profoundly related to the molecular interactions among API solutes and solvents in a real solution system, which were dependent on the molecular structure of an API, the physical nature of a solvent, temperature, and compositional ratios of the solvent combinations (10, 12, 16, 17). In Type-3, Type-4, and Type-5 behaviors, the agglomeration process was intricately related to the balance between the drag force and the liquid-bridge strength (18, 19). The drag force was dependent on the speed of agitation, the viscous force, and the growing diameter of an agglomerate (18, 19). The liquid-bridge strength relied on the low pressure in the bridging-liquid phase led by the curvature of the bridging-liquid interface. The curvature of the bridging liquid was a function of:  the interfacial tension of the bridging liquid and the mother liquor; the surface tension between the API crystals and the bridging liquid; the solubility of the API at the interface between the bridging liquid and the mother liquor; and the geometry of the API crystals (18–21).   

Only 10 solvent combinations from a total of 183 combinations produced decent spherical agglomerates of carbamazepine. Only 4 from a total of 100 solvent combinations produced decent spherical agglomerates of cimetidine and 4 of 136 solvent combinations produced decent spherical agglomerates of phenylbutazone. Water was present in 18 workable combinations as an antisolvent (i.e., derived from 10 + 4 + 4 = 18). This observation agreed with the many reported experimental conditions that water could be used as an antisolvent for nonionizable organic compounds (22–43). If only the solvent combinations with water as an antisolvent were considered,  the total number of solvent combinations would have been reduced from 183 to 80 for carbamazepine, from 100 to 40 for cimetidine and from 136 to 80 for phenylbutazone. The criterion of using water as an antisolvent, therefore, was essential to accelerate the screening activities. The aqueous-solution environment induced the formation of crystal surfaces with hydrophilic functionalities, which helped to sustain the curvature of the hydrophobic solvents (44). This result may explain why most of the bridging liquids in the 18 workable solvent combinations were Class-4 chlorocarbon or Class-5 hydrocarbon solvents.  

Figure 1 (ALL FIGURES ARE COURTESY OF THE AUTHORS)

 The 10 solvent combinations of good solvent, antisolvent, and bridging liquid for carbamazepine were: 1,4-dioxane, water, and chloroform; acetonitrile, water, and paraxylene; acetonitrile, water, and chloroform; acetonitrile, water, and 

-dimethylaniline; ethanol, water, and toluene; ethanol, water, and chloroform; ethanol, water, and 

-dimethylaniline; methanol, water, and xylene; methanol, water and toluene; and methanol, water, and chloroform. All carbamazepine spherical agglomerates (see Figure 1) were freshly made as the dihydrate, which was the thermodynamically stable form in the aqueous system (45, 46). They were first oven-dried at 100 °C for 4 h to become Form I before being subjected to characterization by a scanning electron microscope (SEM), whose high-vacuum system could be contaminated by the dehydration of the dihydrate form. Although the solubility of Form I carbamazepine was less than the solubility of Form III, it was advantageous to make Form I carbamazepine spherical agglomerates for several reasons (47). Form I transformed to dihydrate at a slower rate than Form III did in an aqueous environment, and the solubility of the dihydrate was known to be about 1.3 times less than the solubility of any of the anhydrous form (48). Also, the difference in the solubility values between Forms I and III was relatively small, and the area under the curve in bioavailability tests of the forms were in the order of Form I > Form III < dihydrate. Lastly, Form I was stable below 71 °C (47).

The four solvent combinations of good solvent, antisolvent, and bridging liquid for cimetidine were: methanol, water, and 

-heptane; methanol, water, and xylene; methanol, water, and paraxylene; and methanol, water, and toluene. All cimetidine spherical agglomerates (see Figure 2) were made of Form C crystals as verified by the FTIR spectrum with Form C characteristic bands at 1066, 1182, 1229, and 1347 cm

  (5). Even though Form A is thermodynamically less stable than Form C under normal pressure, Form A was preferable for medicine because of Form C's extraordinarily poor bulk flowability for tableting (4, 6). The feasibility of producing Form C cimetidine spherical agglomerates, however, might have made the industrial tablet production of Form C cimetidine possible. 

The four different solvent combinations of good solvent, antisolvent, and bridging liquid for phenylbutazone were: tetrahydrofuran, water, 				

-heptane. Spherical agglomerates (see Figure 2) of the first two cases were made of Form E crystals (see Figure 3), whereas the last two cases were made of Form A crystals as indicated by DSC. Form A showed only a single melting endotherm at 105 °C for Form A, but the DSC thermogram (see Figure 4) exhibited the first endothermic peak at 92.5 °C , which corresponded to the melting of Form E (3). This peak was followed by an exothermic peak at 95 °C, which  corresponded to the crystallization of Form E (3). A second endothermic peak at 104 °C corresponded to the melting of Form A (3). Because the equilibrium solubility and the dissolution rate of the phenylbutazone crystal forms were in the order of Form C > Form E> Form D >  Form B > Form A and Form C > Form E > Form B > Form D > Form A, respectively, spherical agglomeration might offer a way of tuning the dissolution profile of phenylbutazone by the change of polymorphism in addition to providing  good bulk flowability. 

 The polymorphism of the API solids remained unchanged before and after the addition of a bridging liquid in the previoulsy described 18 solvent systems. Other powder characteristics of the spherical agglomerates from the 18 solvent systems such as the percent yield, the length-mean diameter, the apparent density, the population density, the sphericity, and the friability index were fully analyzed based on the sample weight and digital-camera photographs and were calculated according to Equations 1–6 as follows (22, 26, 40): 

 is length-mean agglomerate diameter (cm); 

 is length-mean agglomerate diameter (cm) after 30 min, and (

 is length-mean agglomerate diameter (cm) at the beginning of the friability test. The volume of a bridging liquid and the angle of repose were entirely determined by experiments. The sample sizes varied greatly from numerous powder-like agglomerates to one large aggregate as revealed clearly in Figures 1, 2, and 3. No replicate was made for each test at this point due to the limitation of time and materials. All experimental and calculated results are summarized in Table III.  

Spherical agglomerates of carbamazepine dihydrate from an acetonitrile–water–paraxylene system, Form C cimetidine from a methanol–water–xylene system, Form E phenylbutazone from a tetrahydrofuran–water–

-heptane system, and Form A phenylbutazone from an acetonitrile–water–

-heptane system gave relatively high yields, high friability indices (i.e., greater strength) and low angles of repose (i.e., good bulk flowability) among the other spherical agglomerates produced by other systems. Therefore, they were first chosen as candidates for further scale-up or preformulation studies. Rotation in a cylindrical vessel alone did not encourage the formation of spherical agglomeration because a control experiment confirmed that spherical-shaped agglomerates could also be produced in a 500-mL sized crystallizer equipped with baffles and an overhead stirrer. The yield of each system was dependent on the compositional ratios of the solvent combinations. The yield can be easily raised during future process development by increasing the volume ratio of antisolvent to good solvent to values higher than 5:1. 

 SEM images of the cross-section of all spherical agglomerates of carbamazepine dihydrate, Form C cimetidine, Form E phenylbutazone, and Form A phenylbutazone were categorized into four different typical modes. All primary crystals were needles and were: tangled with fibrils (see Figure 5(a)); randomly stacked together and linked by solid nodes (see Figure 5(b)); bundled together in parallel (see Figure 5(c)) and radiated from a center core; and jointed by solid bridges to form many 

-shaped ladder structures (see Figure 5(d)) (49). 

Because the volume ratio of water (i.e., the antisolvent) to good solvent was fixed at 5:1, the presence of a small amount of bridging liquid immiscible with water also would be immiscible with the saturated aqueous API solution and the good solvent as confirmed experimentally. Under a high speed of stirring, the bridging liquid was dispersed in the saturated aqueous API solution and the good solvent as tiny suspended droplets. These tiny droplets were randomly drawn to near the points of contact between the preformed needle-like primary crystals by interfacial tension. The fibrils, nodes and joints (see Figures 5 (a)–5 (d)) suggested that the dissolution of API would take place initially at the crystal–bridging liquid interface because of the relatively higher solubility power of the bridging liquid than the surrounding saturated high water content of the mother liquor. As the saturated high water content of the mother liquor diffused into the bridging-liquid region, recrystallization of the API occurred inside the liquid bridges. 

The form space constructed from the common organic solvents by initial solvent screening provided a systematic way of finding ideal combinations of a good solvent, an antisolvent, and a bridging liquid for spherical agglomeration on  a miniaturized scale. The same working logic could be extended to other solvent techniques such as the spreading of cube-shaped particles, monodispersed double emulsions,  microemulsions,  and organic nanocrystal fabrication (50–53). Powder characteristics of spherical agglomerates such as percent yield, length-mean diameter, apparent density, population density, sphericity, friability index, and angle of repose served as a selection guide for narrowing down the final feasible solvent combinations for future scale-up. The initial solvent screening method made it possible to identify the right solvent combinations for spherical crystallization early in drug development. This strategy, therefore, increases the  possibility of direct tableting without performing granulation. 

This work was supported by a grant from the National Science Council of Taiwan, R. O. C. (NSC 98-2113-M-008-008-006). Suggestions from Jui-Mei Huang in DSC and Ching-Tien Lin in SEM, both at the National Central University's  Precision Instrument Center and High Valued Center are gratefully acknowledged. 

 were graduate students in the department of Chemical and Materials Engineering, and 

 was a graduate student at the Institute of Materials Science and Engineering, all at National Central University, 300 Jhong-Da Rd., Jhong-Li City 320, Taiwan, ROC, tel. +886 3 422 7151 ext. 34204, fax +886 3 425 2296, 

To whom all correspondence should be addressed.  

Submitted: Feb. 2, 2009; Accepted May 20, 2009.  

 3. M.D. Tuladhar, J.E. Carless, and M.P. Summers, 

Articles

In Part I of this article, which appeard in the March 2010 issue, the authors describe their approach for constructing form spaces for carbamazepine, cimetidine, and phenylbutazone by initial solvent screening to evaluate the feasibility of spherical crystallization. Part II of this article discusses their findings.

Spherical Crystallization for Lean Solid-Dosage Manufacturing (Part II)

Pharmaceutical manufacturing typically lags behind other industries in process efficiencies because production processes for new drug products are quickly developed in a laboratory for producing clinical-trial materials to enable the rapid submission of a new drug application (NDA) (1). The NDA locks in the documented manufacturing steps with a regulatory commitment and leaves no chance for the full optimization of processing steps such as crystallization, filtration, drying, dry blending, granulation, drying, and tableting (2). Granulation is the slowest of these processing steps, but by applying cross-performance relationships and mixing rules, the efficiency of granulation can be enhanced significantly (2–4). To achieve lean solid-dose manufacturing, the granulation step can be replaced with spherical crystallization in a common stirred tank at the crystallization step (5–26). This change may reduce the time-consuming and labor-intensive steps in solid-dose manufacturing for a high-dose formulation to only a few unit operations of spherical crystallization, filtration, drying, dry blending, and tableting.  

To the best of our knowledge, no literature has laid out a systematic way to identify the best solvent combinations for spherical crystallization. This article presents a robust miniaturized screening strategy to accelerate the preparation and characterization of spherical agglomerates produced by spherical crystallization. This strategy can be readily adopted or automated using only a small amount of active pharmaceutical ingredient (API), thus enabling spherical crystallization to be performed before the NDA submission. 

Spherical crystallization is a well-studied, solvent-based method that directly produces spherically agglomerated crystals to afford free flowability, uniform packability, and good compressibility for mixing, filling, direct tableting, and even burning characteristics (17–19). Spherical crystallization is categorized into two methods: emulsion solvent diffusion (ESD) and spherical agglomeration (SA) (13, 22, 24, 26). Both methods use a 

 (i.e., a solvent that gave solubility of ≥ 5 mg/mL at 25 °C ) to dissolve the API and a 

 (i.e., a solvent that gave solubility of < 5 mg/mL at 25 °C) miscible with the good solvent (i.e., the antisolvent) to generate the required supersaturation. For the ESD method, the solution containing the dissolved API is dispersed into the bad solvent to produce emulsion droplets. Crystals are generated inside the suspended droplets by the counterdiffusion of the good and bad solvents at the droplet interface (26). However, for the SA method, the formation of agglomerates is promoted by the addition of a small amount of the third solvent, namely the bridging liquid, which is immiscible with the bad solvent and drawn preferentially in the interstitial space between the newly formed API crystals (26). 

Spherical crystallization is a very versatile method. The operating conditions involve either: the recrystallization of a near-saturated solution by cooling (10, 12, 13, 15, 16, 19, 21, 22, 25, 26), by seeding (14), or by an antisolvent (7, 8, 10–13, 15, 16, 19–26) with (7, 8, 10–13, 15, 20–26) or without (19, 22) the presence of a bridging liquid or with the addition of an emulsifying agent (13, 15, 18) or a cross-linking agent (16, 23); or secondly, the dispersion of solids with the addition of a bridging solution (5, 6, 11, 17). The addition order of the antisolvent and the bridging liquid to the near-saturated solution can vary (11, 26). 

Spherical crystallization relies on the form space (or the solvent-miscibility plot) derived from an initial solvent-screening method performed in 20-mL scintillation vials (27–31). Because the composition ratio of an antisolvent to a good solvent at the time of nucleation determines the polymorphism of the API crystals, the solvents were added by a specific SA pathway (11, 12, 30). The antisolvent was first added to the near-saturated API solution to ensure the occurrence of recrystallization in a binary solvent system. The bridging liquid was introduced at the end point of recrystallization to induce the agglomeration of the existing primary crystals. The polymorphism of the primary crystals before and after spherical agglomeration was monitored by either differential scanning calorimetry (DSC) or transmission Fourier transform infrared (FTIR) spectroscopy. The volume of a bridging liquid used, the percent yield, the length-mean diameter, the apparent density, the population density, the sphericity, the angle of repose, and the friability index of spherical agglomerates were carefully measured and calculated (5, 9, 13, 23).  The internal microstructures of spherical agglomerates were examined by scanning electron microscopy (SEM). 

The feasibility of spherical crystallization by initial solvent screening was tested with three model APIs (32–42) (see Figure 1). These APIs were: carbamazepine (5H-dibenze (b, f) azepine-5-carboxamide), an anticonvulsant used to treat epilepsy and trigeminal neuralgia; cimetidine (N"-cyano-

'-[2-[[5-methyl-1H-imidazol-4-yl]methyl]thio]ethyl)guanidine), a specific competitive histamine H2-receptor antagonist used to  treat human peptic ulcers; and phenylbutazone (1,2-diphenyl-4-

-butyl-3,5-pyrazolidinedione), a nonsteroidal antiinflammatory drug with antipyretic and analgesic activity. These APIs were chosen because of  several key properties: their nonionizable nature between pH of 6 and 7; their worldwide commercial value; the abundance of characterization information in the literature; the lack of extensive solvent studies on their solubility, polymorphism, crystallinity, and crystal habits; and the variety of their naturally occurring polymorphs, hydrates,  and/or solvates. 

Figure 1: Molecular structures of (a) carbamazepine,(b) cimetidine, and (c) phenylbutazone.   (FIGURE I IS COURTESY OF THE AUTHORS.)

O), white, crystalline, and prismatic powders, molecular weight (mw) of 236.67 g/mol, melting point (mp) of 191–192 °C, reagent grade, Lot 036k1219, and cimetidine (C

S), white crystalline rod-like powders, mw of 252.34 g/mol, mp of 141–143 °C, reagent grade, Lot 088H1317 were purchased from Sigma-Aldrich (St. Louis, MO) and used as received. Phenylbutazone (C

), white, crystalline, and needle-shaped powders, mw of 308.38 g/mol, mp of 106–108 °C, purity of 99+ %, Lot A0230775 was purchased from Acros Organics (Morris Plains, NJ) and used as received. Identification tests for carbamazepine and phenylbutazone were carried out by DSC. The use test for cimetidine was performed by FTIR instead of DSC because DSC was unable to distinguish the polymorphs of cimetidine clearly (39). 

The following sovlents were obtained from Tedia (Fairfield, OH): acetone (CH

), high-performance liquid chromatographical/spectrometrical (HPLC/spectro) grade with purity of 99.5%, boiling point (bp) of 56 °C, mw of 58.08 g/mole, Lot 411050; 

OH),  American Chemical Society (ACS) grade with purity of 99.4%, bp of 117.7 °C, mw of 74.12 g/mole, Lot 205027; 

), ACS grade with purity of 99.8%, bp of 153 °C, mw of 73.10 g/mole, Lot 020505; 

), HPLC/spectro grade with purity of 99.4%, bp of 98 °C, mw of 100.21 g/mole, Lot HS-1712 ; isopropyl alcohol ((CH

CHOH), HPLC/spectro grade with purity of 99.8%, bp of 82.4 °C, mw of 60.1 g/mole, Lot 503027; methanol (CH

OH), HPLC/spectro grade with purity of 99.9%, bp of 64.7 °C, mw of 32.04 g/mole, Lot 411070; methyl ethyl ketone (C

), ACS grade with purity of 99.6%, bp of 81.6 °C, mw of 72.11 g/mole, Lot 201021; methyl 

), certified grade with purity of 99.9%, bp of 55.2 °C, mw of 88.15 g/mole, Lot 712032; tetrahydrofuran (C

O), HPLC/spectro grade with purity of 99%, bp of 65–67 °C, mw of 72.11 g/mole, Lot 411013; and xylene (C

), ACS grade with purity of 98.5%, bp of 137–144 °C, mw of 106.17 g/mole, Lot 305065. 

The following solvents were obtained from Acros Organics: 

), ACS grade with purity of  99%, bp of 193 °C, mw of 121.18 g/mole, Lot A0213203001 and 

), ACS grade with purity of 99%, bp of 138 °C , mw of 106.17 g/mole, Lot 48754/2. 1,4-Dioxane (C

), ACS grade with purity of 98%, bp of 100–102 °C, mw of 88.11 g/mole, Lot sp-3432R was obtained from Showa Chemical (Tokyo). Chloroform (CHCl

), HPLC/spectro grade with purity of 99.9%, bp of 60.5–61.5 °C, mw of 119.38 g/mole, Lot E554180 and ethanol (CH

OH), HPLC/spectro grade with purity of 99.5%, bp of 78 °C, mw of 46.7 g/mole were obtained from Echo Chemical (Taipei, Taiwan). Ethyl acetate (CH

), ACS grade with purity of 99.5%, bp of 76.5–77.5 °C, mw of 88.11 g/mole, Lot G43342 was obtained from Grand Chemical (Daejeon, South Korea). Acetonitrile (CH

CN), ACS grade with purity of 99.96%, bp of 81.6 °C, mw of 41.05 g/mole, Lot 0043X29B30, and toluene (C

), HPLC/spectro grade with purity of 99.8%, bp of 110.6 °C, mw of 92.14 g/mole, Lot B46755 was purchased from  Mallinckrodt Baker (Phillipsburg, NJ). Reversible-osmosis (RO) water was clarified by a water purification system (Milli-RO Plus, Millipore, Billerica, MA).  

 Under the initial solvent screening, the 19 solvents specified in the solvents section were chosen. The working procedures followed closely the details described in References 27–30. As previously defined, a 

was a solvent that gave solubility of ≥ 5 mg/mL at 25 °C. A 

 was a solvent that gave solubility of < 5 mg/mL at 25 °C). The form space was constructed in  a framework of 19 X 19 solvent matrix (27–30).  

 The working logics followed the procedures laid out in References 27–30.  

 Good solvents, antisolvents, and bridging liquids were chosen according to the form space of each API. For carbamazepine and phenylbutazone, approximately 30 mg of API were dissolved in 2 mL of a good solvent in a 20-mL scintillation vial. For cimetidine, 2 mL of saturated solution were prepared using a good solvent. Then, 10 mL of an antisolvent were added to all API solutions under a standardized agitation of 200 rpm by an orbital shaker (Model OS-701, TDK , Taipei). The 5:1 volume ratio of the antisolvent to the  good solvent was based on a specified cystallization method (11). Crystallization only took place in some of the solution systems. 

Two hours after the first appearance of solids, the solids were filtered, oven-dried under a mild condition at 40 °C overnight, and characterized for their purity, polymorphism, solvates, and crystallinity by DSC, thermal gravimetric analysis (TGA), and FTIR. 

Next, 0.02-mL portions of a bridging liquid were added in the remaining slurry every once in a while by a micropipette until all suspended crystals were transformed into agglomerates by sight. The amount of the bridging liquid used should be kept at minimum without causing any dissolution of the primary crystals or exceeding 80% volume (i.e., 16 mL in total solution volume) of the scintillation vial. All trials were performed at room temperature, and all agglomerates were filtered, oven-dried under mild conditions of 40 °C overnight, and characterized. For carbamazepine spherical agglomerates, oven-drying at 100 °C for 4 h was performed before any further characterization. For the solution systems that gave rise to spherical agglomerates, experiments were repeated again without the addition of any bridging liquid to produce primary crystals that were filtered, oven-dried at 40 °C overnight, and characterized for their polymorphism by either DSC or FTIR. 

 Thermal analytical data of 3 to 5 mg of solids placed in perforated-aluminum-sealed 60-µL pans were collected on a calorimeter (Perkin Elmer DSC-7, Perkin Elmer, Waltham, MA), with the following temperature scanning rates: 10 °C/min from 50 to 200 °C for carbamazepine; 5 °C /min from 50 to 170 °C for cimetidine (mainly for the crystallinity measurements); and 8 °C/min from 50 to 130 °C for phenylbutazone using nitrogen with purity of 99.990 % as a blanket gas (32, 33, 35, 39, 41). The temperature axis was calibrated with indium with purity of 99.999% (Perkin Elmer) with a melting onset at 156.6 °C. 

 Sample-weight loss as a function of temperature was monitored by a thermogravimetric analyzer (TGA 7, Perkin Elmer). The heating rate was 10 °C/min from 50 to 200 °C, 5 °C/min from 50 to 170 °C and 8 °C/min from 50 to 130 °C for carbamazepine, cimetidine, and phenylbutazone, respectively (32, 33, 39, 41). Weight loss was usually associated with either solvent evaporation close to the boiling point of a solvent, as in the case of solvates, or sample decomposition. The open platinum pan and stirrup were rinsed by ethanol and burned by a spirit lamp to remove all impurities. All samples were heated under nitrogen atmosphere to avoid oxidization. About 3 mg of sample were placed on the open platinum pan suspending in a heating furnace. 

 Transmission FTIR spectra were recorded on a spectrometer (Perkin Elmer Spectrum One, Perkin Elmer)  Approximately 1 mg of sample was ground gently with 99 mg of 50 °C oven-dried potassium bromide (KBr) in an agate mortar and pestle to avoid polymorphic transition possibly induced by extended grinding. The round KBr sample disk was prepared by a uniaxial press with a pressure of 7 tons. The disk was scanned with a scan number of 8 from 450 to 4000 cm

 A digital still camera (DSC-T7 Cyber-shot, Sony, Tokyo) was used to take photographs of millimeter-sized spherical agglomerates before and after the friability test. The number of spherical agglomerates and their individual diameters were measured based on the image analysis. 

 Because of the small amount of spherical agglomerates produced, the angle of repose was measured with the oven-dried spherical agglomerates inside the 20-mL scintillation vial. The vial was mounted on a lever that had one end pivoted to the benchtop and was originally laid flat on the benchtop surface.  The vial was tilted by raising the other end of the lever gradually. The angle made between the lever and the benchtop surface at which the spherical agglomerates inside the vial began to flow was the angle of repose. Each angle measurement was repeated at least 10 times with a protractor. 

 Spherical agglomerates harvested from each trial were placed in a 20-mL scintillation vial that was subjected to a rotating speed of 50 rpm on a ball-milling machine (MUBM-236-RTD, Shin Kwan Machinery, Taipei) for 30 min. Photographs of spherical agglomerates before and after this friability test were taken.  Image analysis was performed to calculate the friability index. 

 An SEM (Hitachi S-3500N, Hitachi Ltd., Tokyo) was used to observe the internal morphology of spherical agglomerates that were cross-sectioned by a razor blade. Because the cutting was done by hand, it might have crushed the agglomerates. To avoid imaging the artifacts, the cross-sectioned area was blown with compressed air to remove surface debris. Both secondary electron imaging and backscattered electron imaging were used for the SEM detector, and the magnification was 15- to 300,000-fold.  The operating pressure was 10

  Pa vacuum, and the voltage was 15.0 kV. All samples were mounted on a carbon conductive tape (Prod. No. 16073, TED Pella, Redding, CA) and sputter-coated with gold (Hitachi E-1010 Ion Spotter) with a thickness of about 6 nm. The discharge current used was about 0 to 30 mA, and the vacuum was around 10 Pa. 

 were graduate students in the Department of Chemical and Materials Engineering, and 

*To whom all correspondence should be addressed. 

Submitted: Feb. 2, 2009; Accepted May 20, 2009. 

 7. Y. Kawashima, M. Okumura, and H. Takenaka, 

 9. Y. Kawashima, M. Okumura, and H. Takenaka, 

 18. U. Teipel, T. Heintz, and H.H. Krause, 

In Part I of this article, the authors describe the materials and methods used in developing a screening strategy to accelerate the preparation and characterization of spherical agglomerates by spherical crystallization.

Spherical Crystallization for Lean Solid-Dose Manufacturing (Part 1)

In Part 1 of this article, which appeared in the May 2009 issue of 

 the authors described characterization methods for the functional properties and the process-dependent polymorphic flow charts of the three soft pharmaceutical materials: carbamazepine, cimetidine, and phenylbutazone. This part presents several indispensable processing properties (i.e., solubility, polymorphism, crystallinity, and crystal habit) of those three active pharmaceutical ingredients. The underlying principles can be easily extended to other nonconventional sytems such as metal–organic framework materials, infinite coordination polymer particles, and cocrystals. 

 The differential scanning calorimetry (DSC) use-test scan of carbamazepine in Figure 1 exhibited the first endothermic peak at 170–174 °C, which corresponded to the melting of Form III, followed immediately by an exothermic peak at 175 °C, which corresponded to the crystallization of Form I, and a second endothermic peak at 193 °C, which corresponded to the melting of Form I (1). The DSC use-test thermogram of phenylbutazone in Figure 2 displayed a single endothermic peak around 105 °C, which corresponded to the melting of Form A (2). Thus, the polymorphisms of carbamazepine and phenylbutazone were Form III and Form A, respectively, at room temperature. The use test for cimetidine was performed by Fourier transform infrared spectroscopy (FTIR) because the melting points of the different forms of cimetidine are close to each other (3). 

Figure 1: Differential scanning calorimetry scan of the carbamazepine. (ALL FIGURES ARE COURTESY OF THE AUTHORS) 

The FTIR spectrum in Figure 3 showed four significant infrared peaks for Form A cimetidine at 1204, 1156, 1077, and 954 cm

  and two well-resolved characteristic peaks at 1243 and 1228 cm

  (4–5). Therefore, the polymorphism of cimetidine at room temperature was Form A. The cimetidine molecule has three hydrogen-donor centers (i.e., N1H, N10H, and N13H) and three hydrogen acceptor centers (i.e., N3, N12, and N17) (see Figure 4) (6). Two cimetidine molecules formed a dimer unit in Form A with an eight-member ring through the hydrogen bonds between N10H of one molecule and N12 of the other. The cimetidine molecule absorbed water at the N1H group (6). Cimetidine was weakly basic, and its reported pKa values were 6.80 and 6.93. The imidazolium peaks were 1204 and 954 cm

  was attributed to a CH deformation and a ring deformation (6, 7). The peak at 1156 cm

  could either be a coupled ring and a CH bending mode or the deprotonation of the N1H group on the complexation with a water molecule (6, 7). Two imidazole tautomers coexisted in the cimetidine structure: one with a hydrogen atom bound to the N1 atom and the other with the hydrogen atom bound to the N3 atom (see Figure 4). However, only the N1-H form was present in the crystal states of cimetidine and its monohydrates. In the solution, cimetidine also assumed two staggered (

) configurations of the cyanoguanidine group (8). 

Figure 2: Differential scanning calorimetry scan of the phenylbutazone. 

 To construct the form space for an active pharmaceutical ingredient (API), which contains the possible solvent combinations for discovering a new polymorph, 19 common solvents for scale-up were first arranged vertically and horizontally in ascending order of the solvent's Hildebrand value. This arrangement resulted in a 19 x 19 matrix of 361 boxes (see Tables I–III). Each box represented a specific binary-solvent combination, except for the 19 boxes descending diagonally from the upper left corner to the lower right corner. These 19 boxes represented the 19 pure solvent systems. The yellow boxes were the good solvents, and the red boxes the bad solvents, with respect to a particular API. The yellow and red boxes were the material properties of an API. Besides the red and yellow boxes, the gray boxes were also determined experimentally. They represented the immiscible solvent pairs that were independent from the nature of an API. Once the locations of the red, yellow, and gray boxes were fixed, the location of the green, blue, and white boxes representing the antisolvent system (i.e., a good solvent and a bad solvent), the good cosolvent system (i.e., a good solvent and a good solvent), and the bad cosolvent system (i.e., a bad solvent and a bad solvent), respectively, could be deduced graphically. Since the 19 x 19 matrix was symmetrical along the diagonal boxes, the actual possible solvent combinations were reduced from 361 boxes to a total of 19 + (361 – 19) ÷ = 190 boxes or fewer because it is normally impractical to perform experiments in the white-boxed areas. 

Figure 3: Fourier transform infrared spectrum of the cimetidine. 

The form space of carbamazepine had 13 yellow, 6 red, 69 green, 78 blue, 14 gray, and 10 white boxes (see Table I); the form space of cimetidine had 5 yellow, 14 red, 66 green, 10 blue, 14 gray, and 81 white boxes (see Table II); and the form space of phenylbutazone had 17 yellow, 2 red, 22 green, 135 blue, 14 gray, and 0 white boxes (see Table III). The actual form spaces would have been significantly bigger if various solvent compositions of binary mixtures,temperatures, and ternary solvent systems had been considered (9–11). 

 When the locations of the 19 solvents were assigned within the 3-D Hansen plot of a dispersion-force component (δ

) and all the good solvents and bad solvents were represented by yellow and red, respectively, based on the colors of the diagonal boxes in the form space of a given API, a cluster of yellow domains was formed. The contour of these yellow domains outlined a 3-D volume of solubility in space that excluded all the bad solvents by a solubility sphere with center coordinates (δ

) (12). For the solubility sphere of Form III carbamazepine at 25 °C, δ

 . For the solubility sphere of Form A cimetidine at 25°C, δ

 . Finally, for the solubility sphere of Form A phenylbutazone at 25°C, δ

) were close to the reported values of 17.5, 12.5, and 10.7 MPa

 , respectively, measured by dissolution calorimetry (13). 

Table I: Solvent systems of carbamazepine. 

With the solubility sphere of an API at hand, the solubility power of any new solvent other than the 19 solvents used with known values of δ

 can be predicted instantly. If the coordinates of a new solvent (δ

) are located inside the solubility sphere of an API, it is a good solvent with respect to that API. Otherwise, it is deemed a bad solvent. To verify the validity of the solubility sphere, each sphere of an API was tested with 

-propanol to be located inside the Form III carbamazepine's solubility sphere (because 

  for Form III cambamazepine), inside the Form A cimetidine's solubility sphere (because 

  for Form A cimetidine), and inside the Form A phenylbutazone's solubility sphere (because 

  for Form A phenylbutazone). These calculations agreed well with the experimental observations that 

-propanol was a good solvent at 25 °C for Form III carbamazepine (solubility = 14.18 mg/mL), for Form A cimetidine (solubility = 13.7 mg/mL) and for Form A phenylbutazone (solubility = 31.46 mg/mL). 

Table II: Solvent systems of cimetidine. 

However, cyclohexane was calculated to be situated outside the Form III carbamazepine's solubility sphere (because 

  for Form III cambamazepine), outside the Form A cimetidine's solubility sphere (because 

  for Form A cimetidine), and outside the Form A phenylbutazone's solubility sphere (because 

  for Form A phenylbutazone) (12). Once again, the experimental results supported the calculations that cyclohexane was a bad solvent at 25 °C for all three APIs with a solubility of < 1 mg/mL. 

Table III: Solvent systems of phenylbutazone. 

Most of the good solvents of Form III carbamazepine belonged to Class 1 protic or hydrogen-bond donating solvents (i.e., Lewis acids), Class 2 hydrogen-bonding acceptor solvents (i.e., Lewis bases), and Class 3 polar aprotic solvents, which disrupted the N-H···O hydrogen bonds and the C-H···O weak interactions of the carboxamide dimer units existing in the solid state of all forms of carbamazepine (1, 14). The good solvents for Form A cimetidine were mainly Class 1 solvents that interacted readily with the weakly basic imidazole ring and disrupted the N-H···N hydrogen bonds of the dimer units existing in the solid state of Form A cimetidine (6, 8, 14). Finally, the good solvents for Form A phenylbutazone were Classes 1 to 3 solvents, Class 4 chlorocarbon solvents, and Class 5 hydrocarbon solvents. This result implied that the intermolecular interactions in the crystal lattice of phenylbutazone were dominated by van der Waals forces (dipole–dipole interactions and London dispersion force) (15). 

Figure 5: Solubility curves of Form III carbamazepine in 13 solvents. 

Figure 6: Solubility curves of Form A cimetidine in five solvents. 

The solubility curves of Form III carbamazepine in 13 single good solvents, Form A cimetidine in five single good solvents, and Form A phenylbutazone in 17 single good solvents are shown in Figures 5, 6, and 7, respectively. The solubility of all three APIs in different solvents increased with temperature. The enthalpy of dissolution (ΔH

) of an API in a particular good solvent could be approximated from the slope and the 

-intercept of a straight line, respectively, when the solubility curve of the API in that particular good solvent was replotted in the form of the van 't Hoff equation (16). 

 Figure 7: Solubility curves of Form A phenylbutazone in 17 solvents. 

 All API solids generated from the saturated solutions of their corresponding good solvents by temperature cooling were isolated and analyzed with optical microscopy, DSC, thermogravimetric analysis (TGA), and FTIR. Optical micrographs of carbamazepine, cimetidine, and phenylbutazone solids are illustrated in Figures 8, 9, and 10, respectively. The crystal habits, the aspect ratios, the polymorphism, and the crystallinity of carbamazepine, cimetidine, and phenylbutazone solids grown from the 13 good solvents are summarized in Table IV. 

Most of the carbamazepine solids exhibited the typical DSC thermograms for Form III and Form I crystals, respectively. Unlike Form III carbamazepine crystals, Form I crystals showed no transformation and only one sharp melting endotherm at 190 °C (17). A carbamazepine acetone solvate produced in acetone agreed with this observation (18). The TGA scan (see Figure 11) further showed that the carbamazepine acetone solvate had a 1:1 stoichiometric ratio and the DSC curve (see Figure 11) demonstrated the removal of acetone from the solvate at 50–88 °C and a phase change of the material to Form I with a melting endotherm at 190 °C (17, 19). 

In contrast with the previous observations that solvents with low dielectric constants yielded Form I crystals and those with high dielectric constants yielded Form III crystals, the authors obtained Form I needles not only from solvents such as chloroform and 1,4-dioxane with low dielectric constants, but also from acetonitrile with a high dielectric constant (see Figures 8c, 8g, and 8j) (1, 18, 20, 21). Form III prisms were also harvested from a solvent such as isopropyl alcohol with a low dielectric constant when enough aging time was given for the transformation of Form I to Form III according to Ostwald's Rule of Stages (see Figure 8i). The rest of the good solvents generated the thermodynamically stable Form III prisms (see Figures 8a, 8b, 8d, 8e, 8f, 8h, 8k, 8l, and 8m) (1, 20, 21). Isopropyl alcohol tended to give a mixture of needled and prismatic carbamazepine solids (see Figure 8i). The interplay between thermodynamics and kinetics made concomitant crystallization possible (22). The DSC curves of the physically isolated needles and prisms by a pair of tweezers were resolved to Form I (see Figure 1) and Form III crystals, respectively. 

Table IV: Crystal habits, aspect ratios, polymorphism, and crystallinity of solids of carbamazepine, cimetidine, and phenylbutazone generated from their corresponding good solvents. 

Apparently, the needle or prism crystal habit of carbamazepine was affected only by polymorphism and not by microscopic properties such as the Hansen parameters and the dielectric constants of the solvents. The dominant factor that influenced the crystallization outcomes could have been the degree of supersaturation imposed by temperature cooling, which placed the crystallization under either kinetic or thermodynamic control (22). This inference was supported by the feasibility of a one-solvent polymorph screen of carbamazepine and the appearance of the metastable Form I carbamazepine from a high degree of supersaturation, either by a sudden drop of the solubility through the introduction of an antisolvent or by cooling the system with a relatively steep solubility curve such as chloroform and acetonitrile (see Figure 5) (11). The drive for structures that could be formed more quickly than others might override the drive for settling with structures of the maximum decrease in energy. However, this speed seldom affected the crystallinity, as was seen by the relatively high percent crystallinity of Form I carbamazepine crystals. 

Figure 11: Differential scanning calorimetry and thermogravimetric analysis scans of 1:1 carbamazepine acetone solvate made from acetone. 

The cimetidine solids showed the two typical FTIR spectra in Figure 3 and Figure 13 as Form A and Form B crystals. The FTIR spectrum in Figure 13 displayed the six characteristic infrared peaks for Form B cimetidine at 1375, 1230, 1192, 1184, 1176, and 1066 cm

  (4, 23). The thermodynamically Form A cimetidine was formed in Class 1 solvents such as 

-butyl alcohol and in isopropyl alcohol as rods (see Figures 9a and 9b) (3). Isopropyl alcohol yielded Form A cimetidine (3). However, the steep slope for the solubility curves of Form A cimetidine in solvents of Classes 1 and 2 such as 

-dimethylformamide, methanol, and ethanol produced metastable Form B cimetidine needled crystals (see Figures 9c, 9d, and 9e) without the need of adding and seeding crystals because of the high degree of supersaturation created by temperature cooling (see Figure 6) (3). Apparently, the crystal habits for Forms A and B cimetidine depended on the solvents and the polymorphs, respectively. 

 Figure 12: Differential scanning calorimetry thermogram of Form I carbamazepine crystals grown from chloroform. 

Almost all of the phenylbutazone solids exhibited the typical DSC scan in Figure 2 for Form A long-needled crystals (see Figures 10a–k and 10m–q). However, the phenylbutazone solids grown from 

-butyl alcohol could be a newly discovered Form F (see Figure 10l) because its DSC thermogram in Figure 14 showed two endothermic peaks at 85 and 103 °C, and its TGA scan showed no weight loss at 85 °C. Therefore, the first and second endothermic peaks corresponded to the heat of transformation from Form F to Form A and the melting of Form A, respectively. Despite the many forms of phenylbutazone, phenylbutazone revealed only the ketoform in all polymorphs (24). Furthermore, two solvates were produced. Rather than a phenylbutazone solvate to tetrahydrofuran ratio of 2:1, a ratio of 3:1 (see Figure 10h) was obtained when the isolated solids crystallized from tetrahydrofuran were oven dried at 40 °C for 9 h instead of the crystal surface being blotted with filter paper (15). DSC and TGA thermograms in Figure 15 showed the desolvation endotherm from 70 to 90 °C and the sharp melting of Form A phenylbutazone at 103 °C under a heating rate of 8 °C/min. However, the stoichiometric ratio of 2:1 of phenylbutazone solvate with 1,4-dioxane (see Figure 10k) oven dried at 40 °C for 9 h remained the same as that prepared by blotting the crystal surface with filter paper (15). DSC and TGA thermograms in Figure 16 illustrate the desolvation endotherm from 80 to 84 °C and the sharp melting of Form A phenylbutazone at 103 °C under a heating rate of 8 °C/min (15). 

Figure 13: Fourier transform infrared spectrum of Form B cimetidine produced from methanol. 

Obviously, given the moderate slope of the solubility curve of phenylbutazone Form A in 

-butyl alcohol, the degree of supersaturation could not explain the production of Form F in 

-axis of phenylbutazone is > 5.1 Å, it belongs to the a-type structure (24). The molecule is stabilized more uniformly by a larger number of near neighbors (24). The molecular coordination is between 6 and 10. The C-H···O interaction, especially in the 

-butyl side chain of phenylbutazone, plays an important role in stabilizing the crystal packing in addition to the stacking of the benzene rings (24). The authors hypothesized that the similar molecular structures of 

-butyl side chain of phenylbutazone and the interference of the hydroxyl group of 

-butyl alcohol with the C-H···O interaction might have induced the formation of a thermodynamically metastable Form F nucleus and perturbed the crystal lattice order, as reflected by the low crystallinity of 61%. Since needle-shaped crystals were obtained from all good solvents, it seemed that solvents did not play a dominant role in the induction of the habits characteristic of the crystals, but rather the processing conditions and the polymorphism did (24). 

Figure 14: Differential scanning calorimetry thermogram of Form F phenylbutazone grown from n-butyl alcohol. 

The form space constructed from the common organic solvents by initial solvent screening provided information not only about materials properties such as solubility curves, solubility spheres, crystallinity, crystal habits, and a possible solvent combination for discovering a new polymorph, but also about a systematic way of finding the ideal combination of a good solvent, an antisolvent, and a bridging liquid for solvent-based processes in a miniaturized scale. Initial screening by common laboratory tools could be easily integrated by automation and be applied to other compounds as well. 

Figure 15: Differential scanning calorimetry and thermogravimetric-analysis scans of 3:1 phenylbutazone tetrahydrofuran solvate. 

This work was supported by a grant from the National Science Council of Taiwan, R.O.C. (NSC 95-2113-M-008-012-MY2). Suggestions about DSC from Jui-Mei Huang, assistant at National Central University's Precision Instrument Center, and about scanning electron microscopy from Ching-Tien Lin, assistant at National Central University's Precision Instrument Center, are gratefully acknowledged. 

Figure 16: Differential scanning calorimetry and thermogravimetric-analysis scans of 2:1 phenylbutazone 1,4-dioxane solvate. 

 is an associate professor in the Department of Chemical and Materials Engineering and the Institute of Materials Science and Engineering, 

 are graduate students at the Department of Chemical and Materials Engineering, and 

 is a graduate student at the Institute of Materials Science and Engineering, all at National Central University, 300 Jhong-Da Rd., Jhong-Li City 320, Taiwan, ROC, tel. +886 3 422 7151 ext. 34204, fax 886 3 425 2296, 

Submitted: July 8, 2008. Accepted: Sept. 15, 2008. 

What would you do differently? Submit your comments about this paper in the space below.

 1. A.L. Grzesiak et al., "Comparison of the Four Anhydrous Polymorphs of Carbamazepine and the Crystal Structure of Form I," 

The authors describe the importance of a rapid and an abbreviated screening strategy in initial solvent screening. This article contains bonus online-exclusive material.

Initial Solvent Screening of Carbamazepine, Cimetidine, and Phenylbutazone: Part 2 of 2

Many recent important solvent-based processes such as spherical crystallization, the spreading of cube-shaped particles, monodispersed double emulsions, microemulsions, and organic nanocrystal fabrication, involve combinations of two or three solvents (1–26). However, to the authors' knowledge, no literature has laid out a systematic way to identify the best solvent combinations, even though the composition ratio of the solution at the time of nucleation profoundly determines the solubility, polymorphism, and crystal habits of organic compounds functioning as in the case of the active pharmaceutical ingredients (APIs) (7, 8, 27). 

This article will re-emphasize the importance of a rapid and abbreviated screening strategy by initial solvent screening in 20-mL scintillation vials. The screen provides engineering information in solubility curves, solubility spheres, polymorphs, crystallinity, form spaces (also known as solvent miscibility plots), and crystal habits of a given compound for solvent-based processes (see Figure 1) (28–31). The attractiveness of this strategy is its ability to be adopted or automated by common laboratories with only a small amount of chemical compounds. The probable solvent combinations of a good solvent, an antisolvent, and a bridging liquid for solvent-based processes are summarized systematically in the form space and can be graphically deduced. 

Figure 1: Initial solvent screening in a 20-mL scintillation vial provides engineering information about (counterclockwise from top) solubility curves, polymorphism, solubility spheres, crystal habits, and form space for solvent-based processes. (ALL FIGURES ARE COURTESY OF THE AUTHORS) 

Initial solvent screening was thoroughly tested using three model APIs: carbamazepine (5H-dibenze[b,f]azepine-5-carboxamide), an anticonvulsant used to treat epilepsy and trigeminal neuralgia (see Figure 2a); cimetidine (

-imidazol-4-yl]methyl]thio]-ethyl]-guanidine), a specific competitive histamine H

-receptor antagonist used in the treatment of human peptic ulcers (see Figure 2b); and phenylbutazone (1,2-diphenyl-4-

-butyl-3,5-pyrazolidinedione), a nonsteroidal anti-inflammatory drug with antipyretic and analgesic activity (see Figure 2c). These compounds were studied because of their commercial value; their abundant characterization information in the literature (see Table Ia–c); the lack of extensive solvent studies on their solubility, polymorphism, crystallinity, and crystal habits; and the variety of their naturally occurring polymorphs, hydrates, and solvates (see Figure 3) (32–69). 

Figure 2: Molecular structures of (a) carbamazepine (5H-dibenze[b,f]azepine-5-carboxamide), (b) cimetidine (N"-cyano-N-methyl-N'-[2-[[5-methyl-1H-imidazol-4-yl]methyl]thio]-ethyl]-guanidine), and (c) phenylbutazone (1,2-diphenyl-4-n-butyl-3,5-pyrazolidinedione).  

Although carbamazepine has numerous polymorphs and solvates, it has four main polymorphs and two hydrates: Forms I–IV, and Dihydrates I and II (36, 43, 44). Forms I–III are enantiotropic with respect to each other (34, 36). The stability of carbamazepine is in the order of Form III, Form I, Form IV, Form II, based on the density rule, and the initial dissolution rate of carbamazepine is in the order of Form III, Form I, Dihydrate III (38, 43). Cimetidine has four polymorphs and three monohydrates: Forms A–D, and Monohydrates 1–3 (53). Forms A–D are monotropic (57). Although Forms A and D are virtually isoenergetic crystals, the stability of Form D is slightly higher than that of Form A (58). The solubility of Form A is greater than that of Form B, and Form C (54). Phenylbutazone has six polymorphs and six solvates: Forms α, β, γ, δ, ε, and ζ, and phenylbutazone 2:1 solvates with benzene, cyclohexane, 1,4-dioxane, tetrahydrofuran, tetrachloromethane, and chloroform (62, 63, 65, 69). Forms  β and δ are monotropic. Forms α and β, and Forms α and δ are enantiotropic (66). 

Table Ia: Characterization methods for four polymorphs and two dihydrates of carbamazepine.  

 Carbamazepine, white crystalline prismatic powders (C

O, MW: 236.67 g/mol, melting point (mp) = 191–192 °C, reagent grade, Lot: 036k1219) and cimetidine, white crystalline rodlike powders (C

S, MW: 252.34 g/mol, mp = 141–143 °C, reagent grade, Lot: 088H1317) were purchased from Sigma-Aldrich (St. Louis, MO) and used as received. Phenylbutazone, white crystalline needle-shaped powders (C

The authors describe the importance of a rapid and an abbreviated screening strategy by initial solvent screening in 20-mL scintillation vials.