Chung Shin Kuo

Solubility, polymorphism, crystallinity, and crystal habit of an active pharmaceutical ingredient (API) play critical roles in the value chain of pharmaceutical development, manufacturing, and formulation (1–3). The solubility of an API in solvents and solvent mixtures has a considerable influence on the choice of solvents and the course of operation in solvent-based processes such as chemical reaction, extraction, crystallization, filter cake washing, and wet granulation (4–9). The polymorphism of an API determines its packing, thermodynamic, spectroscopic, kinetic, surface, and mechanical properties in the solid state (10). The crystallinity of an API contributes to the mechanical properties of a compact, an API's stability, and the dissolution rate (11–15). The crystal habit of an API also has profound effects on the rate at which the API can be processed in filtering, washing, and drying, and the success of the API in powder flow, blending, direct compaction, roller compaction, wet granulation, and dissolution rate (16–19). 

Because solubility, polymorphism, crystallinity, and crystal habit are all solvent dependent, solvent screening is of fundamental and foremost importance to many chemical process industries, especially the pharmaceutical industry (4, 20–22). Recently, there has been increased interest in performing high-throughput polymorph screening in miniaturized scales using solvent evaporation (23–25). Because of the course of solid generation from a supersaturated solution by evaporation and the relatively small amount of the API used, however, these methods may not always correlate directly with the scale-up conditions in crystallization (

 a relatively large volume of solvent, temperature cooling, and stirring) and do not provide direct information about other simultaneous effects brought about by the solvent such as an API's solubility, crystallinity, and crystal habit. 

This article promotes an alternative solvent-screening strategy that is tailor-made for the drug development and design of API solids processes. Under this initial screening strategy of pure-solvent systems, 23 kinds of solvent mostly useful for scale-up were chosen (26). The solubility of the API solute in each solvent at 15, 25, 40, and 60 °C was measured by gravimetric titration. The enthalpy and the entropy of solution for each solvent system were calculated. The solubility of the API in each solvent at 25 °C was plotted against the dielectric constant of various solvents, resulting in a characteristic solubility pattern that might serve as the fingerprint to identify a particular API. Although only pure-solvent systems were being considered in this study, the total "form space" for each API—that is, the total number of solid generation experiments in pure-solvent, cosolvent, and antisolvent systems— was also calculated on the basis of the number of good solvents for the API from the solubility studies and the number of miscible and immiscible solvent pairs from the miscibility investigations.  

Solid generation of the API solute in each pure solvent was achieved by gently shaking a 20-mL scintillation vial and by temperature cooling from 60 to 25 °C under an ambient condition (27). The cooling rate of a solution with a volume <20 mL was almost independent from the volume and the nature of solvent. The cooling profile could be approximated by a quadratic equation determined experimentally as 

 is the time (min). The relatively rapid decrease in temperature serves as an ideal way to induce a polymorph that normally does not occur thermodynamically. In addition, temperature cooling is a common method in crystallization scale-up. 

Differential scanning calorimetry (DSC) was used to determine the polymorphism and the crystallinity of the API solid samples (28). Optical microscopy (OM) was used for crystal habit imaging. Two well-known APIs were selected for our solvent screening strategies: acetaminophen (4-acetamidophenol) and racemic (±)-ibuprofen (α-methyl-4- [isobutyl] phenylacetic acid) (see Figure 1), because of their worldwide commercial values in analgesic and antipyretic therapy, their abundant information in the literature (4,11,17, 29–44), and the lack of extensive solvent studies on their solubility, polymorphism, crystallinity, and crystal habit (29–31). 

 Figure 1: Active pharmaceutical ingredient solute molecules.			

 The following solvents were used in this study: cetone (CH

, HPLC–Spectro grade, 99.5%, boiling point [bp] = 56 °C, molecular weight [MW] = 58.08, lot 411050), acetonitrile (CH

CN, ACS grade, 99.96%, bp = 81.6 °C, MW = 41.05, lot 812045), benzene (C

, ACS grade, 99%, bp = 80 °C, MW = 78.11, lot 310008), benzyl alcohol (C

OH, certified grade, 99.9%, bp = 205 °C, MW = 108.14, lot 809010), 

OH, ACS grade, 99.4%, bp = 117.7 °C, MW = 74.12, lot 205027), chloroform (CHCl

, HPLC–Spectro grade, 99.9%, bp = 60.5–61.5 °C, MW = 119.38, lot 401059), dimethyl sulfoxide (DMSO) ((CH

SO, HPLC–Spectro grade, 99.8%, bp = 189 °C, MW = 78.13, lot 202060), N,N-dimethylformamide (DMF) (HCON(CH

, ACS grade, 99.8%, bp = 153 °C, MW = 73.10, lot 020505), 

, HPLC– Spectro grade, 99.4%, bp = 98 °C, MW = 100.21, lot 111110), isopropyl alcohol (IPA) ((CH

CHOH, HPLC–Spectro grade, 99.8%, bp = 82.4 °C, MW = 60.1, lot. 310067), methanol (CH

OH, HPLC–Spectro grade, 99.9%, bp = 64.7 °C, MW = 32.04, lot. 411070), methyl tert-butyl ether (MTBE) ((CH

, certified grade, 99.9%, bp = 55.2 °C, MW = 88.15, lot 712032), methyl ethyl ketone (MEK) (C

, ACS grade, 99.6%, bp = 81.6 °C, MW = 72.11, lot 201021), tetrahydrofuran (THF) (C

O, HPLC–Spectro grade, 99%, bp = 65–67 °C , MW = 72.11, lot 411013), toluene (C

, HPLC–Spectro grade, 99.8%, bp = 110.6 °C, MW = 92.14, lot 406050), and xylene (C

, ACS grade, 98.5%, bp = 137–144 °C, MW =106.17, lot 305065). These solvents were obtained from Tedia Company (Fairfield, OH).  

The following were obtained from Acros Organics (Morris Plains, NJ): N,N-dimethylaniline (DMA) (C

, ACS grade, 99%, bp = 193 °C, MW = 121.18, lot  A0213203001), nitrobenzene (C

, ACS grade, 99%, bp = 210–211 °C, MW = 123.11, lot A019568301) and p-xylene (C

, ACS grade, 99%, bp = 138 °C , MW = 106.17 °C, lot 48754/2).  

The following were obtained from Showa Chemical Co., Ltd. (Tokyo, Japan): 1,4-dioxane (C

, ACS grade, 98%, bp = 100–102 °C, MW = 88.11, lot sp-3432R). Ethanol (CH

OH, HPLC–Spectro grade, 99.5%, bp = 78 °C, MW = 46.7 ) was obtained from Echo Chemical Co. Ltd. (Taipei, Taiwan). Ethyl acetate (CH

, ACS grade, 99.5%, bp = 76.5–77.5 °C, MW = 88.11, lot G43342) was obtained from Grand Chemical Co. Ltd. (Daejeon, South Korea). Reversible osmosis water was clarified with a water-purification system (Milli-RO Plus, Millipore, Billerica, MA). 

OH, MW = 151.17, lot S23272-444) and racemic (±)-ibuprofen (α-methyl-4-[isobutyl] phenylacetic acid, C

, MW = 206.3, lot 26H1368) were obtained from Sigma-Aldrich (Steinheim, Germany). 

 Approximately 10 mg of the API were weighted in a 10-mL scintillation vial. Drops of solvent were titrated carefully with a micropipette into the vial with intermittent shaking until all API solids were just dissolved. The solubility of the API at a given temperature was calculated as the weight of the API in a vial divided by the total volume of the solvent added to a vial. Solubility of the API in the same solvent at 15, 25, 40, and 60 °C was determined. All temperatures were maintained and controlled by a water bath. Despite the inherent inaccuracy of this method in volume measurements by eye, it provided a rapid and robust way for process scale-up. 

 Out of the 23 solvents, approximately 1-mL portions of each solvent comprising a pair were shaken together for ~1 min in a 20-mL scintillation vial. If no interfacial meniscus was observed after the contents of the vial were allowed to settle, the solvent pair was considered to be miscible. If a meniscus was observed without apparent change in the volume of either solvent, the pair was regarded as immiscible. 

Polymorph, crystallinity and crystal habit studies.

 Saturated API solutions of each solvent were prepared in a 20-mL scintillation vial in accordance with the solubility values at 60 °C that were determined from the experiments described in the previous section. Solids were generated by cooling from 60 to 25 °C in an ambient condition with intermittent shaking. Solids were filtered with a strong vacuum and vacuum-oven dried at 55 °C overnight. Polymorphism, crystallinity, and crystal habits were determined by DSC and optical microscopy, respectively. 

 A differential scanning calorimeter (DSC 7, Perkin Elmer, Norwalk, CT) was used to monitor the thermal events during heating. The instrument was calibrated with indium (8–10 mg, 99.999% pure, extrapolated melting onset at 156.6 °C). All samples were run in sealed aluminum pans under a constant nitrogen purge. Each sample was heated at 10 °C/min. 

 An optical microscope (SZII, Olympus, Tokyo, Japan) equipped with a CCD camera (SSC-DC50A, SONY, Tokyo, Japan) was used to take images of crystal habit. 

The solvents were categorized into five classes according to their functional group (26):								

Class 1: "protic" or hydrogen bond donating solvents (Lewis acids), including benzyl alcohol, 

-butyl alcohol, ethanol, IPA, methanol, and water;

Class 2: hydrogen bonding acceptor solvents (Lewis bases), including acetone, 1,4-dioxane, DMF, ethyl acetate, MEK, MTBE, nitrobenzene, THF, and water;

Class 3: polar aprotic solvents, better termed "nonhydroxylic solvents," including acetonitrile, DMA, DMF, and DMSO;

Class 4: chlorocarbon solvents, which include chloroform; 

C : hydrocarbon solvents, both saturated and unsaturated, including benzene, 

Acetaminophen dissolved well in 15 of the solvents (labeled as "good solvents") and only slightly dissolved in 8 of the solvents (labeled as "bad solvents") with a solubility <1 mg/mL at 25 °C. The bad solvents for acetaminophen were mainly solvents of Classes 2, 4, and 5 and included MTBE, DMA, chloroform, benzene, 

-heptane, toluene, p-xylene, and xylene. Therefore, only 15 solubility curves of acetaminophen in 15 good solvents of Classes 1 and 3 at 15, 25, 40, and 60 °C (50 °C was used in cases in which a solvent's boiling point is close to 60 °C) were constructed and grouped by their similar solubility ranges for comparison (see Figure 2). As for racemic (±)-ibuprofen, all classes of solvents were good solvents except for water at 25 °C. Therefore, solubility curves of racemic (±)-ibuprofen in 22 good solvents at 15, 25, 40, and 60 °C (either 50 or 55 °C was used in cases in which a solvent's boiling point is close to 60 °C) were constructed and grouped by their similar solubility ranges as well (see Figure 3). Using scattered published results (29, 45) to check  the accuracy of our data in Figures 2 and 3, the error percentage was within ±20%. 

Figure 2: Solubility curves of acetaminophen in 15 kinds of solvent.			

If the solute and the solvent were assumed to form an ideal solution, the solubility could be predicted from the van't Hoff equation: 

 is the mole fraction of the API solute in the solution, 

 is the gas enthalpy constant (8.314 J/mol K), Δ

 is the entropy of dissolution (46). Extracting 

 from each solubility curve and performing linear-curve fitting to ln 

 of the API in each solvent could be approximated from the slope and intercept of a straight line. The values of Δ

 for acetaminophen and racemic (±)-ibuprofen are summarized in Tables I and II.  

Table I: Dielectric constants of solvents versus enthalpies and entropies of acetaminophen and racemic (Â±)-ibuprofen solutions.			

 in Table I indicate that the energy of attraction of the API solute molecules with each other and the energy of attraction of the solvent molecules with each other are lower than the energy of attraction of the API solute and the solvent molecules in the solution. Therefore, heat was absorbed to make the API solute dissolve in the solvent. In this case, the solubility of the API solute increased with temperature. Furthermore, the positive values of Δ

 in Table I showed that the API solute–solvent systems became less ordered as the API solute dissolved into the solution. 

Table II: Solvent miscibility table, cosolvent and antisolvent systems of acetaminophen.			

When the solubility values of acetaminophen and racemic (±)-ibuprofen at 25 °C were plotted against the dielectric constant of various solvents (see Table I) in which the APIs were soluble, the APIs' characteristic "solubility pattern" was produced (see Figures 4 and 5). Nonetheless, there were no direct correlations between the dielectric constants of the solvents and the APIs' solubility values (see Figures 4 and 5) or between the dielectric constants of the solvents and the APIs' Δ

 values, nor between the dielectric constants of the solvents and the Δ

Figure 3: Solubility curves of racemic (Â±)-ibuprofen in 22 kinds of solvent.			

The implication is that the solubility was profoundly influenced by various intermolecular interactions. Besides the Coulombic (ion–ion or ion–dipole) interactions, other effects such as the Keesom force (rotating dipole–dipole force), the Debye force (dipole–induced-dipole force), the London force (induced-dipole–induced-dipole force), and the hydrogen bonding should all be considered. The high solvation power of the two "polar aprotic solvents"—DMF and DMSO—in the two solubility patterns (see Figures 3 and 4) suggested that solid generation might be difficult to carry out in extremely good solvents. 

Figure 4: Solubility spectrum of acetaminophen.			

Based on the solvent miscibility studies of the solvent pairs of the 23 kinds of solvent, there were 18 immiscible pairs (36 gray boxes divided by 2) (see Tables II and III). It was necessary to divide the number of boxes by 2 when a particular solvent 

. Again, there were no direct correlations between solvent dielectric constants (see Table I) and solvent miscibility values (see Tables II and III). For example, 1,4-dioxane with a dielectric constant of 2.2 was miscible with all other solvents in Tables II and III, but xylene (dielectric constant 2.0) was immiscible with DMF, DMSO, and water. Similar to solute–solvent solubility, the solvent–solvent miscibility also was profoundly influenced by the many aforementioned intermolecular interactions. 

Table III: Solvent miscibility table, cosolvent and antisolvent systems of racemic (Â±)-ibuprofen.			

The pure-solvent systems were represented as the 23 diagonal boxes of the solvent pair of 

 in the solvent miscibility tables (see Tables II and III). The "form space" of the pure-solvent systems for our initial solvent screening was limited to the number of good pure solvents (

 indicated by the yellow boxes in Tables II and III) for each API. Therefore, the form space for acetaminophen and racemic (±)-ibuprofen was 15 and 22 respectively (see Tables II and III). If the good co-solvent systems (

 the binary mixtures of good solvents) were taken into account, however, the form space of the good cosolvent systems was extended to the number of blue boxes in the solvent miscibility table divided by 2 (see Tables II and III). Therefore, the form space under these particular conditions for acetaminophen and racemic (±)-ibuprofen was 102 and 224 respectively (see Tables II and III).  

Table IV: Form space of acetaminophen and  racemic (Â±)-ibuprofen.			

In addition, if the antisolvent systems (

 binary mixtures of a good and a bad solvent) also were considered, the form space of the antisolvent systems was calculated as the number of green boxes in the solvent miscibility table divided by 2 (see Tables II and III). The form space under these circumstances for acetaminophen and ibuprofen was 105 and 11, respectively (see Tables II and III) and the total form space for acetaminophen and racemic (±)-ibuprofen should then be at least equal to 222 and 257, respectively (see Table IV). 

Figure 5: Solubility spectrum of racemic (Â±)-ibuprofen.			

The total form space should expand dramatically if various solvent compositions of binary mixtures, temperatures, and ternary solvent systems are taken into account (23). Solid generation by temperature cooling can be applied in the yellow and blue regions in the solvent miscibility table (see Tables II and III). Nonetheless, a constant temperature must be maintained if solid generation was achieved by the addition of an antisolvent in the green domain (see Tables II and III). In general, no attempts are made for solid generation in the regions of immiscible solvent pairs (gray boxes), bad solvents (orange boxes), and cosolvents of bad solvents (

 binary mixtures of any two bad solvents) in the solvent miscibility table (see Tables II and III). In addition, references of other miscible solvent pairs from organic solvents other than the 23 kinds of solvent listed in Tables II and III are available in the literature (48, 49). 

Figure 6: Differential scanning calorimetry thermogram of a typical sample of acetaminophen.			

Out of the 15 good solvents, acetaminophen solids were generated in only 14 of them. N,N-dimethylformamide was so good a solvent that acetaminophen was unable to crystallize out by temperature cooling. All 14 isolated white solids were analyzed by DSC, resulting in a sharp endotherm at ~171 °C, which is the melting point of Form I crystals (see Figure 6) (37). This result clearly verified that acetaminophen powders solidified from all 14 solvents were the thermodynamically stable Form I crystals at 25 °C. As for the racemic (±)-ibuprofen, the white powders could only be generated from 18 out of the 22 good solvents. Racemic (±)-ibuprofen supersaturated solutions of acetone, benzyl alcohol, dimethyl sulfoxide, and xylene were unable to produce solids by temperature cooling. Nonetheless, the 18 isolated solid samples showed a typical endotherm around the solid–liquid melting point of 77 °C (see Figure 7) (30, 32). Apparently, there was no new polymorph and the fact that the racemic (±)-ibuprofen is isomorphic once again was verified (18). 

Articles

Solubility, polymorphism, crystallinity, and crystal habit of acetaminophen and racemic (+/-)-ibuprofen were determined by initial screening of 23 solvents for scale-up. Solubility curves were constructed, and solubility at 25 degrees Celsius was plotted against the dielectric constants of various solvent as a fingerprint for solute identification. The total "form space" for acetaminophen and racemic (+/-)-ibuprofen were calculated to be 222 and 257, respectively. Various crystal habits and sizes for ibuprofen and acetaminophen were observed.

Solubility, Polymorphism, Crystallinity, and Crystal Habit of Acetaminophen and Ibuprofen by Initial Solvent Screening

The indentation-size effect forms the basis of the relationships connecting average particle size (

				) with densification, tensile strength, and hardness of one-component tableting systems. Those structure–property relationships also have been verified by quantitative data presented in the literature, and they may serve as a predictive tool for crystal engineers and formulators to design particles with the right average size for tableting at the beginning of product development. In general, densification, tensile strength, and hardness increase as the average particle size decreases, and a significant amount of compaction pressure is required for densification to occur. These effects present a potential problem when compacting particles that are very small. 

Producing pharmaceutical tablets involves combining the active pharmaceutical ingredient (API) and excipients. When a high API drug-load is desired, selection or manipulation of the excipients or process alone may not be sufficient to compact the tablet. The average particle size of an API influences compactability, but how particle size affects compactability remains unclear. Correlations between average particle size and tablet properties such as densification, tensile strength, and hardness are important because those correlations can generate predictions, which facilitate the selection and design of appropriately sized particles. 

Loose powders have a very low initial relative density (ρ

) is defined as the density of the compact (mass of a compact divided by its volume) divided by the true density, which is determined by helium pycnometry (1). The cold isostatic compaction of powders by the application of macroscopic compaction pressure (

, uniaxial or hydrostatic pressure) is divided into four irreversible densification stages as 

 increases (2, 3). Those stages are particle rearrangement, local plastic deformation at the interparticle contacts, elastic and plastic deformation of the particles, and bulk deformation of the compact by fragmentation once it reaches a state of closed porosity. 

In stage one, weak interparticle attractive forces counteract gravitational forces and produce dangling bonds (2), which form low-density aggregates of particles. In stage two, mechanical compaction begins to take place. The fractional packing density 

 increases from 0.35 to 0.74 (2). At this stage, each particle averages 4.75 contacts and begins to experience "jamming." The onset of general particle deformation starts around 

 = 0.85 (2). Beyond that threshold, only the application of very high pressure produces further densification in stages three and four.  

The densification phenomenon usually is described as the first-order kinetics of applied macroscopic pressure 

 are constants obtained from the slope and intercept of the plot ln (1/(1

,respectively. Two important mechanical properties of tablets: tensile strength 

 (5) further relate to the relative density (

 in Equation 2 are porosity and constants obtained from the slope and intercept of the plot ln 

 in Equation 3 are, respectively, the maximum hardness at zero porosity and the initial relative density needed for minimal shear strength of the powder bulk. When one substitutes Equation 1 into Equation 2 and Equation 3, tensile strength and hardness become functions of macroscopic compaction pressure 

Interestingly, empirical studies suggest that densification (Equation 1), tensile strength (Equation 2), and hardness (Equation 3) of tablets are not affected only by packing structures e and 

 and pressure, but that those properties are also affected by initial average particle size. (1, 6, 7). Although the effects of initial average particle size are implicitly included in the material characteristic constants of 

  in Equation 1 and Equation 2, this article presents three explicit relationships that link average particle radius 

 with densification, tensile strength, and hardness during stage two of densification. At a minimum, those relationships require the establishment of only two data points and serve as predictive tools to engineer the appropriate average particle size in direct tableting. 

 derived a simple relationship that assumes all powders are monosized spheres and the number of contacts per particle resembles microindentation and increases during stage-two compaction (2, 8): 

 is the indentation hardness of the particle material. Helle 

. also derived a semiempirical relationship between the contact size, 

 for a given level of density that is reached 

Effects of initial average particle size on powder compaction that involve interparticle contacts is conceptualized as the indentation-size effect (ISE) among particles. Hardness increases with decreasing indentation size. The phenomenon of ISE is explained by the strain-gradient plasticity. Under the influence of strain-gradient plasticity, plastic strengthening becomes important when a material in question is plastically deformed in very small volumes (

., at the tips of cracks or microindentations). The semiempirical ISE equation has the form (2, 9): 

 is the large-size hardness lower limit, 

Particle-size dependence of densification. 

If one combines Equations 4, 5, and 6 and assumes 

Quantitative data from the literature show strong relationships among average particle size, powder densification, tensile strength, and hardness.

Effects of Initial Average Particle Size on Tableting: Evaluating Predictive Tools for Crystal Engineers and Formulators

Polymorph farming on a chip is a promising approach for rapid polymorph discovery. It transforms the crystallization–crystal isolation–characterization process into one time-efficient step. Crystals grown and harvested by this method may also be suitable for single-crystal X-ray diffraction and seeding in batch crystallization. Polymorph farming on a chip's versatility and feasibility were successfully demonstrated in the current study through templated crystallization of sulfathiazole on four separate thin films characterized immediately by Fourier transform infrared (FTIR) microscopy with an analytical time of about 10 seconds per crystal. 

Self-assembly of molecules, atoms, or ions often is packed precisely in several different long-range orders. Each order in solid state is termed a "crystal polymorph." Because each polymorph is a unique material with its own physicochemical properties (1), the discovery and selection of a desired polymorph are essential in the food (2), explosives (3), optoelectronics (4), agricultural (5), and ceramics (6) industries and especially important in the pharmaceutical industry. Polymorph screening is highly important not only for evaluating the physicochemical aspects (7) of active pharmaceutical ingredients (APIs), but also for legal considerations (8). Polymorph farming on a chip allows for a thorough search for polymorphs on a global scale and at the same time minimizes the sample preparation time for crystallization, isolation, and characterization compared to other  methods (9, 10). 

Conventionally, several recrystallization methods can be used to grow polymorphs. One approach involves crystallization in various solvents or in a combination of solvents under a different rate of cooling of saturated solutions. Other methods entail evaporation of solvents, controlled addition of cosolvents, and a change of pH (11–14). These methods require extra time and labor because crystallization, crystal isolation, and sample preparation for different characterization techniques are deemed separate events. As is commonly known, characterization techniques such as powder X-ray diffraction and differential scanning calorimetry, require a relatively large amount of sample (4–10 mg) and long analytical time (30 min per sample). Although other nontraditional methods such as epitaxy matching (15), laser induction (16), size effect (17), and biomineralization (18) are known to induce uncommon polymorphs, these methods have been automated. Direct and routine implementation have removed the time-consuming drawback brought about by the three discrete activities of crystallization, crystal isolation, and sample preparation. 

Two automated platforms that miniaturize conventional recrystallization methods with the use of an array of glass test tubes in aluminum blocks (9) and polymer heteronuclei in the 96-well polypropylene plate (10) were reported recently. With the use of 

 optical and Raman microscopy, these two methods have apparently have removed the time-consuming crystal isolation–characterization step in the chain of events of crystallization–crystal isolation–characterization. The design of these two enabling platforms, however, does not allow for high-resolution crystal-lattice imaging or for direct measurement of the mechanical and electrical properties of polymorphs by electron microscopy, nanoindenter, atomic force microscopy, thermal microanalysis, and dielectric analysis. The lengthy preparation of combinatorial conditions of different solvents and polymer heteronuclei also has prolonged the crystallization process. A new strategy that enables a universal fabrication-step dedicated only to the preparation of combinatorial conditions and allows for a more general platform for other kinds of physicochemical analysis can effectively shorten the time-consuming activities in polymorph screening. 

Inspired by the microarrays in drug discovery (19) and built on the ideas of templated crystallization, (10, 15, 20–24), the strategy is to grow crystal polymorphs on thin-film templates of block copolymer, metal, and ceramics. Polymers, metals, and ceramics can be deposited on silicon wafers by spin casting and thermal evaporation. 

Polystyrene-polybutadiene-polystyrene (PS-PB-PS) (molecular weight of 140,000 g/mol, 30% by weight styrene, and density of 0.965 g/m

-butylene)-polystyrene (PS-PEB-PS) (molecular weight of 118,000 g/mol, 28% weight by styrene, and density of 0.91g/m

) were obtained from Sigma-Aldrich (St. Louis, MO). PS-PB-PS templates and PS-PEB-PS templates were prepared on polished silicon wafers by spin casting (2000 rpm) from 1% weight-per-volume toluene solutions. Gold templates with obtained by thermal evaporation of 1300 Å of gold on polished silicon wafers under 10

 torr. An unsaturated aqueous solution of sulfathiazole was prepared by warming 0.0432 g of sulfathiazole (Form III, Fluka, Buchs, Switzerland) in 17 mL of water at 90 °C. 

 Sulfathiazole, a well-characterized pharmaceutical compound (25, 26) (see Figure 1a), was chosen to test the polymorph-farming concept. One to two drops of the 90 °C unsaturated aqueous solution of sulfathiazole were separately introduced on the thin films of PS-PB-PS (see Figure 1b), PS-PEB-PS (see Figure 1c), gold, silicon and, mica thin-film templates at room temperature. Different sulfathiazole polymorphic crystals were grown under synergetic effects of film surface, a temperature drop from 90 to 25 °C, and water evaporation on thin-film templates as shown in Figure 2. Crystals grown on each template were then examined by FTIR microscopy. 

. IR spectra were collected using either an FTIR microscope or by attenuated total reflectance with a FTIR microscope system (AutoImage, Perkin Elmer, Shelton, CT). Typical acquisition parameters were 64 scans, a 4-cm

  resolution and a beam size of 10 μm with an acquisition time of only a few seconds. 

The optical images indicated that the sulfathiazole crystals, Form I (needle), Form II (cuboid), Form III (truncated hexagon), and Form IV (plate-like hexagon) were formed on gold (see Figure 3), mica (see Figure 4), PS-PEB-PS and PS-PB-PS (see Figure 5), and silicon (see Figure 6) thin-film templates, respectively. IR characteristic bands (26) 3354, 1406, 1500 and 937 cm

  for Forms I, II, III, and IV were found as 3350, 1337, 1496, and 938 cm

  in Figures 3–6, respectively. A characteristic IR triplet band (26) at 940 cm

  was also observed for Form III. Therefore, the experimental results suggest that a desired polymorph can be incubated through a particular template. 

The 90 °C unsaturated aqueous sulfathiazole solution with a concentration of approximately 254 mg/100 mL (0.0432 g of Form III sulfathiazole in 17 mL of water) was used to ensure the absolute dissolution of sulfathiazole in water in the very beginning because the solubility of 254 mg/100 mL of Form III sulfathiazole usually occurs at about 55 °C (27). As the 90 °C unsaturated aqueous- sulfathiazole solution droplet was introduced on a particular film surface at room temperature, several interrelated mechanisms were made possible. The hydrophilicity or hydrophobicity of the film surface could affect the aqueous droplet size, shape, contact area, and surface area (see Figure 2). The resultant interfacial area between the aqueous sulfathiazole solution and the film then could determine the number of nucleation sites for heterogeneous nucleation. On the other hand, if the homogeneous nucleation were the main driving force, the resultant aqueous droplet size, shape, contact area, and surface area would have determined the cooling rate of the entire aqueous sulfathiazole droplet and the evaporation rate. 

Although Form III sulfathiazole is a thermodynamically stable form in water from 25 to 75 °C, polymorphic forms of I, II, and IV were obtained by carrying out different kinetic pathways (1, 27, 28) with the synergetic effects of surface nucleation sites, rapid cooling, and rapid concentration-increase on the thin-film template. 

The deeper implication of these preliminary results is the feasibility of doing simultaneous polymorph farming of a given molecule on a microarray of thin films (see Figure 7) fabricated by standard photolithography and cleanroom techniques (29). Furthermore, the chip-based configuration conforms to many solid-state characterization methods (30) and can also be integrated with other techniques such as conductance measurement (31), supersaturation measurement (32), electrocrystallization (33), crystal-film fabrication (34), temperature-gradient crystallization (35), and chiral recognition (36). Because nucleation is the key for crystal engineering (37), the thin-film templates are required to have a remarkable level of control over heterogeneous nucleation and stabilization of a particular kind of polymorph. This can be accomplished by lowering (20) and controlling the total surface energy upon stereochemical recognition (38) and the effects of cooperativity (39) between the growing crystal nucleus and the template in the future work. Materials such as block copolymers, gold, silicon, and mica are ideal for template preparation. Block copolymers have diverse functional groups and predetermined group spacing (40). They even can be self-assembled into highly ordered microdomains (41), which may serve as potential nucleation sites. Unlike the method used by Matzger 

(10), block copolymers in this case were spun cast from a good solvent to cause the coil to swell so that the maximum exposure of different functional groups at the air-template interface was achieved. Gold, silicon, and mica, with preassembled crystal lattices and surface charges, are another ideal class of materials for template preparation. They previously were used successfully as templates for surfactants and self-assembled monolayers in different liquid environments (21, 22, 28). 

Other interesting implications of polymorph farming on a chip in formulation and manufacturing may include the effect of different drug-carriers' surfaces on the API's polymorphism and the effect of metallic surfaces of different unit operations on the API's polymorphism. Polymorph farming on a chip also may be used to examine the effect of processes involving fast cooling and evaporation rate (

 spray-drying and ink-jet printing) on an API's polymorphism. It also can be used in the preparation of polymorphic seeds for scale-up production. 

These results suggest that polymorph farming on a microarray with a library of various polymers, metals, and ceramics is within reach. Different from current methods, polymorph farming on a chip is a promising strategy for rapid polymorph discovery by transforming the tedious crystallization–crystal isolation–characterization into one time-efficient step. Crystals grown and harvested by this method may also be suitable for single-crystal X-ray diffraction and seeding in batch crystallization. To move forward, better template design in nucleus–template interaction will require guidance from catalyst science, electrochemistry, biosensing techniques, molecular biology, and protein crystallography.  

The authors are grateful for the National Science Council of Taiwan for its research grants NSC 93-2119-M-008-030 and NSC 94-2119-M-008-001 and would also like to thank Dr. Kelvin Chang at Perkin Elmer (Taiwan) Corporation (Taipei, Taiwan) for the optical images and FTIR spectra. 

 is an assistant professor at the Department of Chemical and Materials Engineering and at the Institute of Materials Science and Engineering, the National Central University, Chung-Li, Taiwan 320, Republic of China, tel. +886 3 422 7151, ext. 34204, fax +886 3 425 2296, 

 are graduate students at the Department of Chemical and Materials Engineering, the National Central University. 

*To whom all correspondence should be addressed. 

1. J. Bernstein, R.J Davey, and J.O Henck, "Reviews: Concomitant Polymorphs," 

2. P. Fryer and K. Pinschower, "The Materials Science of Chocolate," 

3. Y. Oyumi, T.B.Brill, and A.L. Rheingold, "Thermal Decomposition of Energetic Materials (Part 9): Polymorphim, Crystal Structures and Thermal Decomposition of Polynitroazabicyclo [3.3.1] Nonanes," 

, "Polymorphism and Luminescence Properties of CMONS Organic Crystals: Bulk Crystals and Nanocrystals Confined in Gel-Glasses," 

5. K.R.Waerstad and G.H. McClellan, "Preparation and Characterization of Some Long-Chain Ammonium Polyphosphates," 

6.  T. Lee and I.A. Aksay, "Hierarchical Structure–Ferroelectricity Relationships of Barium Titanate Particles," 

8.  "Patents in the Pharmaceutical Industry"

 Encyclopedia of Pharmaceutical Technology 

, J. Swarbrick and J.C. Boylan, Eds., (Marcel Dekker, New York, NY, vol. 11, 1995), pp. 309–328. 

., "Iterative High-Throughput Polymorphism Studies on Acetaminophen and an Experimentally Derived Structure for Form III," 

10. M. Lang, A.L. Grzesiak, and A.J. Matzger, "The Use of Polymer Heteronuclei for Crystalline Polymorph Selection," 

11. C.C Gu, V. Young, Jr., and D.J.W. Grant,  "Polymorph Screening: Influence of Solvents on the Rate of Solvent-Mediated Polymorphic Transformation," 

12. W. Beckmann, "Seeding the Desired Polymorph: Background, Possibilities, Limitations and Case Studies," 

13. C.K. Chen and A.K. Singh, "A 'Bottom-up' Approach to Process Development: Application of Physicochemical Properties of Reaction Products Towards the Development of Direct-Drop Process," 

14. L. Yu and K. Ng., "Glycine Crystallization during Spray Drying: The pH Effect on Salt and Polymorphic Forms," 

15. C.A. Mitchell, L. Yu, and M.D. Ward, "Selective Nucleation and Discovery of Organic Polymorphs through Epitaxy with Single Crystal Substrates," 

, "Nonphotochemical, Laser-Induced Nucleation of Supersaturated Aqueous Glycine Produces Unexpected γ–polymorph," 

17. F. Iwatsu, "Size Effects on the α–β Transformation of Phthalocyanine Crystals," 

18. K. Naka and Y. Chujo, "Control of Crystal Nucleation and Growth of Calcium Carbonate by Synthetic Substrate," 

19. Y. Fang, A.G. Frutos, and J. Lahiri, "Membrane Protein Microarrays," 

20. P.W. Carter and M.D. Ward, "Topographically Directed Nucleation of Orgainc Crystals on Molecular Single-crystal Substrates," 

, "Biomimetic Pathways for Assembling Inorganic Thin Films," 

Polymorph farming on a chip is a promising strategy for rapid polymorph discovery by transforming crystallization?crystal isolation?characterization into one time-efficient step. 