Purification of Structurally Similar Compounds by the Formation of

Compounds by the Formation of Impurity Co-Former Complexes in Solution ... 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States ...
1 downloads 0 Views 457KB Size
Download PDF
Article pubs.acs.org/crystal

Purification of Structurally Similar Compounds by the Formation of Impurity Co-Former Complexes in Solution Kay H. Hsi,† Meghan Kenny,† Allison Simi,‡ and Allan S. Myerson*,† †

Novartis-MIT Center for Continuous Manufacturing and Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States ‡ Department of Chemical and Biological Engineering, Princeton University, A221 Engineering Quadrangle, Olden Street, Princeton, New Jersey 08544, United States ABSTRACT: The purification of structurally similar compounds was investigated using selective impurity complex formation in solution followed by crystallization of the target compound. Two systems of structurally similar compounds, benzamide/benzoic acid (BAM/BA) and cinnamamide/cinnamic acid (CAM/ CA), were chosen. Three reported co-formers that form co-crystals with both BA and CA were selected: isonicotinamide (INA), 2-amino-4,6-dimethylpyrimidine (DMP), and dimethylglyoxime (DMG). The addition of DMG to the BAM/BA system demonstrated the largest improvement of BAM purity. The amount of BA did not decrease with increasing amount of DMG added. For the CAM/CA system, adding DMP resulted in the largest CAM purity increase. Phase solubility diagrams were measured to calculate binding constants for 1:1 complexes. These binding constants were used as indications of the complexation level in the solution. For the CAM/CA system, the more complex formed in solution, the purer the CAM. This result was not seen in the BAM/BA system. The results indicate that, for difficult to purify compounds where the structurally similar impurities substitute in the crystal lattice, the use of additives which have the potential to complex with the impurity in solution can result in an increase in crystal purity. However, the correlation between the level of complexation and the purification result needs further verification.

■

INTRODUCTION In the pharmaceutical industry, separation processes are employed throughout the active pharmaceutical ingredients (APIs) synthesis processes. It is important to purify intermediates prior to the next synthetic step as well as for final APIs to meet U.S. Food and Drug Administration specifications. The main difference between purifying intermediates and APIs is that the allowed impurity level for APIs often is much lower than that for intermediates. It is often possible for an impurity to substitute into an API crystal lattice because of their structural similarity. In Hsi et al., purifying structurally similar compounds using selective impurity cocrystal formation was demonstrated.1 In that work, the impurity co-crystals formed were less soluble than the impurity itself and thus were crystallized, leaving the compound of interest in solution. The goal of this work is to purify structurally similar compounds using selective impurity complex formation in the solution thus preventing lattice substitution of the impurity in the target compound when it is crystallized. For this work, we chose two well-studied systems in the field of tailor-made additives: the benzamide/benzoic acid (BAM/ BA) system and the cinnamamide/cinnamic acid (CAM/CA) system. Tailor-made additives are typically designed to alter API properties such as morphologies.2 They can also be used as nucleation promoters and growth inhibitors.3−5 Tailor-made additives are designed (or selected) to be structurally similar to the API. Because of its structural similarity, it is easy for a tailormade additive to substitute into the API crystal lattice and © 2013 American Chemical Society

disturb the growth of the API crystal when the API was crystallized from an API/impurity mixture. Berkovitch-Yellin et al. used BA to change the morphology of BAM.2 The low impurity level (reported between 0.5% and 1%) made the BAM/BA system a good system for the purpose of this work. Similarly, CAM/CA was chosen to be our second system because of its similarity to the BAM/BA system. The structures of BAM, BA, CAM, and CA are shown in Figure 1. Both systems were investigated for purification. The aim of this work was to prevent the impurity from substituting into the crystal lattice of the target compound. To achieve this, we added a co-former that can form a co-crystal with the impurity to the target compound/impurity mixture. In the 1950s, Higuchi conducted research on complexes formed in solution by caffeine.6−10 The stoichiometries of these complexes and the mechanisms of complex formation were studied. It is proven that complexation can affect solid properties. In our work, we assumed that (1) the co-formers that can form co-crystals with the impurity have a high possibility of forming complexes with the impurity in the solution and (2) because of steric effects, the impurity complexes can no longer fit into the crystal lattice of the target compound. By adding co-formers, the amount of Received: December 11, 2012 Revised: March 7, 2013 Published: March 8, 2013 1577

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582

Crystal Growth & Design

Article

Figure 1. Structures of benzamide, benzoic acid, cinnamamide, and cinnamic acid (from left to right). 99%) were purchased from Sigma-Aldrich and were used as received. Anhydrous ethanol (200 proof) was USP grade and was purchased from VWR. Methanol (MeOH, CHROMASOLV for HPLC ≥99%) and water (H2O, CHROMASOLV Plus for HPLC) were purchased from Sigma-Aldrich and used for HPLC. Co-Former Selection. A search in Cambridge Structural Database (CSD) was conducted to select the co-formers reported to form cocrystals with both impurities: BA and CA. This search led to the selection of three co-formers: (1) isonicotinamide (INA), (2) 2amino-4,6-dimethylpyrimidine (DMP), and (3) dimethylglyoxime (DMG). Their structures are shown in Figure 2. The reference codes of these reported co-crystals in CSD are presented in Table 1.

impurity incorporated into the crystal lattice is decreased, purifying the compounds of interest. To confirm that the purification was due to complexation in solution, we estimated the binding constants of the complexes and used them as indications for the level of complexation. Phase solubility diagrams for co-crystal systems were used to find the stoichiometry and the binding constants for these complexes in equilibrium with co-crystals.11−16 With different combinations of co-crystal stoichiometry, complex stoichiometries, and the binding constants of the complexes, the solubility of the co-crystal is a unique function of the co-former concentration.12,13 Nehm et. al showed that, for a 1:1 cocrystal AB, the concentration of A, the compound of interest, can be expressed as different functions of the concentration of B, the co-former, depending on whether or not the 1:1 complex forms.12 Similarly, if A and B form a 2:1 co-crystal, the solubility of this binary co-crystal A2B in a pure solvent is affected by the lack of complex formation, the formation of only the 1:1 complex, only the 2:1 complex, and both complexes. However, in the case where A and B form 2:1 co-crystals, the binding constants cannot be solved analytically. Instead, they have to be fitted using the concentrations of A and B obtained experimentally. Phase solubility diagram data measured for a co-crystal system with known co-crystal stoichiometry can be used to determine if the co-crystal components form complexes in the solution as well as the binding constants of the complexes. The following co-formers were chosen to form complexes with the impurities BA and CA: (1) isonicotinamide (INA), (2) 2-amino-4,6-dimethylpyrimidine (DMP), and (3) dimethylglyoxime (DMG). Separation experiments were performed; the purities of the target compounds, BAM and CAM, under different experimental conditions were measured; and the results are presented in Table 2, Table 3, and Table 4, respectively. Two systems showed promising purification results: (1) BA−DMG and (2) CA−DMP. Phase solubility diagram data for these two systems were measured and used to find the binding constants of the complexes formed in the solution. Phase solubility diagram data for two other systems that did not show significant purification were measured as comparisons. The purification results were explained using the level of complexation in the solution. In this work, we were able to purify BAM and CAM by adding co-formers reported to form co-crystals with the impurities, BA and CA, respectively. Although it is premature to predict purification results based on the level of complexation, it is confirmed that, by adding these co-formers, we can prevent the impurities from substituting into the crystal lattice.

■

Figure 2. Structures of (a) isonicotinamide (INA), (b) 2-amino-4,6dimethylpyrimidine (DMP), and (c) dimethylglyoxime (DMG).

Table 1. Reference Codes of the Reported Co-Crystals INA

DMP

BA

BUDWEC

CA

LUNMAI

VOCZET, VOCZET01, VOCZET02, VOCZET03 DAYWAC

DMG NABCAV, NABCAV01 NABCEZ

These six co-crystals were made successfully. Four co-crystals (BA− INA, BA−DMP, CA−INA, and CA−DMP) were prepared using solution crystallization with reported co-crystal stoichiometries (1:1 molar ratio). BA−DMG and CA-DMG co-crystals were prepared using different stoichiometries (1:1 molar ratio) from the co-crystal stoichiometry (2:1 molar ratio) because of the low solubility of DMG in ethanol. The obtained co-crystals were confirmed to be the reported ones using X-ray powder diffraction (XRPD). Crystallization Experiments. Three sets of experiments were performed to determine if the co-former could be used to purify the target compound. In the first experiment, BAM and CAM were crystallized with co-formers to determine if co-formers substitute into their crystal lattices. The co-formers incorporated into their crystal lattices were no longer investigated in later steps. In the second experiment, BAM and CAM were crystallized with only the impurity, BA and CA, respectively, to examine how much impurity was incorporated into their crystal lattices. In the third experiment, BAM and CAM were crystallized with both the impurity and the co-formers that were not excluded in the first experiment to examine the purity of BAM and CAM after the addition of co-formers. The initial solution contained the saturation concentration of target compound at 40 °C and the impurity at a weight ratio 1:9 and 1:4 to the compound of interest. The amount of co-former added was the stoichiometric amount to form the co-crystal with the impurity. That is, the impurity to co-former ratio was same as the stoichiometry used to make the impurity co-crystal. The solution was heated in a water bath to 50 °C until the target compound dissolved. The solution was cooled to 30 °C at a rate of 1 °C/min and then to 20 °C at a rate of 1 °C/6 min. The slurry was then set for an hour at 20 °C. The solids obtained from the crystallization were collected and washed using iced ethanol. The purities of the resulting products were determined using high performance liquid chromatography (HPLC).

MATERIALS AND METHODS

Materials. Benzoic acid (BA, ACS reagent ≥99.5%), benzamide (BAM, 99%), trans-cinnamic acid (CA, ≥ 99%), cinnamamide (CAM, predominately trans, 97%), 2-amino-4,6-dimethylpyrimidine (DMP), dimethylglyoxime (DMG, ACS reagent), and isonicotinamide (INA, 1578

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582

Crystal Growth & Design

Article

[A]T,exp. The method of least-squares was applied to find out the best solution that gives the minimum sum of squared errors, which was defined as

Phase Solubility Diagram Construction. Different compositions of A and B solids (A-B combinations: BA−DMP, BA−DMG, CA−DMP, CA−INA) were added to ethanol to make supersaturated solutions with respect to all components and the co-crystal. The slurries were stirred with magnetic stir bars in 20 mL glass vials overnight at constant temperature (20 °C) maintained using a circulating water bath to reach equilibrium. The liquid phase was filtered using 0.45 μm PTFE syringe filters and diluted using the same solvent as the HPLC mobile phase (30/70 water/methanol with 0.1% trifluoroacetic acid). The concentration of the liquid phase was determined using the HPLC method described below. The solid phase was collected using vacuum filtration with filter paper and dried at room temperature overnight. The dried solids were confirmed to be the desired co-crystal using X-ray powder diffraction. High Performance Liquid Chromatography (HPLC). The HPLC instrument (Agilent 1260 Infinity) was equipped with a UV diode array detector (Agilent Technologies G1315D). The column used was a YMC-Pack ODS-A 150 × 4.6 mm i.d. column packed with 3 μm particles with 12 nm pore size (YMC America Inc.). The maximum wavelength for absorbance was set at 230 nm. The concentrations were analyzed using a 5 min isocratic method with a 30/70 water/methanol mobile phase containing 0.1% trifluoroacetic acid. X-ray Powder Diffraction (XRPD). X-ray powder diffraction patterns were obtained using a PANalytical X’Pert PRO Theta/Theta powder X-ray diffraction system using a monochromatic Cu Kα radiation source with nickel filter (λ = 1.5418 Å) generated at 45 kV and 40 mA, using an X’Celerator high-speed detector. The intensities were measured at 2-theta values from 5° to 40° at a continuous scan rate of 5°/min. Aluminum sample holders with a zero background silicon plate were used to carry out the measurements. Model Fitting. Different approaches were used to calculate the binding constants of the complexes for the 1:1 co-crystal systems and the 2:1 co-crystal system. For each 1:1 co-crystal system, the plot of the total impurity concentration versus the reciprocal of the total coformer concentration was analyzed using least-squares regression. If linear dependence was shown, we can find the solubility product, Ksp from the slope and the binding constant for the 1:1 complex, K11, from the intercept. If the intercept is not significantly different from 0, there was no complexation and K11 was 0. Otherwise, the binding constants of the complexes, K11, can be determined using the following equation:12

[A]T =

K sp [B]T

+ K11K sp

sum of squared errors =

K sp [B]T − K11K sp

■

RESULTS Purification Results. The purities of BAM obtained from the crystallization experiments are presented in Table 2. The Table 2. Purities of Benzamide in Different Crystallization Experiments 9:1 BAM/BA with DMP with 1× DMG with 1.5× DMG with 2× DMG

BA (wt %)

DMP (wt %)

DMG (wt %)

decreased (%)

± ± ± ±

0.087 ± 0.020 -

0.00 0.00

8.18 18.79 7.71

-

0.00

11.00

0.35 0.32 0.28 0.32

0.033 0.049 0.025 0.024

0.31 ± 0.029

results using INA are not shown in the table because it substituted into the BAM crystal lattice and was not used to purify BAM. The original mass percentage of BA in BAM crystal was 0.35(±0.033) wt %. Once DMP was added (1:1 molar ratio of BA:DMP), the amount of BA in BAM crystal decreased to 0.32(±0.049) wt % with 0.087(±0.02) wt % of DMP present in BAM crystals. With the addition of DMG (1:1 molar ratio of BA:DMG), the amount of BA in BAM crystal decreased to 0.28(±0.025) wt % with no trace of DMG detected. The effect of co-former stoichiometry was studied by varying the amount of DMG (1:1.5 and 1:2 molar ratios of BA:DMG). With the addition of 1:1.5 and 1:2 molar ratios of DMG, the amount of BA decreased to 0.32(±0.024) wt % and 0.31(±0.029) wt %, respectively. To determine whether the decreases are statistically significant at a 95% confidence interval, all of the amounts of BA after the addition of coformers were compared to the original amount of BA using a t test. The decreases of the amount of BA after the addition of DMG at all molar ratios were statistically significant. On the other hand, the decrease of the amount of BA after the addition of DMP was not statistically significant. In addition, a t test was performed to determine whether the amount of BA decreases with the increasing amount of DMG added or not. The amounts of BA with the addition of 1:1.5 and 1:2 molar ratios of DMG were compared to that with the addition of 1:1 molar ratio of DMG, respectively. The results showed that increasing the amount of DMG did not statistically significant decrease the amount of BA substituted into the crystal lattice. The purities of CAM obtained from the crystallization experiments are shown in Table 3. The original mass percentage of CA in CAM crystal was 0.85(±0.078) wt %. Once DMP was added (1:1 molar ratio of CA:DMP), the amount of CA in CAM crystal decreased to 0.65(±0.101) wt %. After the addition of INA (1:1 molar ratio of CA:INA), the amount of CA decreased to 0.76(±0.142) wt %. However, the addition of DMG increased the amount of CA incorporated into CAM crystal lattice to 0.92(±0.153) wt %. No trace of coformers was detected in any resulting CAM solids. Similarly, a t

(1)

+ K11K sp

(4)

Various Ksp, K11, and K21 values were examined to find the best set of constants that minimized the sum of squared errors. This range was chosen based on theoretical estimation, literature value, and the fact that it should be ≥0.

If no linear dependence was shown, then the following equation was used to find the solubility product and the binding constant of the complex:12

[A]T =

∑ ([A]T,exp − [A]T,fitted )2

(2)

Equation 2 can be rearranged and expressed as ([A]T × [B]T ) = K11K sp([A]T + [B]T ) + (K sp − K112K sp2) (3) The plot of ([A]T × [B]T) versus ([A]T + [B]T) was used to find the solubility product and the binding constant. For the 2:1 co-crystal system, the solubility product and binding constants cannot be found using linear regression. Instead, the phase solubility diagram data obtained for the BA−DMG system was fitted to four different models: the 2:1 co-crystal system with no complexation (model 1), the system with the 1:1 complex in the solution (model 2), the system with the 2:1 complex in the solution (model 3), and the system with both 1:1 and 2:1 complexes (model 4), respectively. The total concentrations of BA, [A]T,fitted, were calculated using experimentally measured concentrations of DMG and values for Ksp and the binding constants, if any. [A]T,fitted, were then compared to the total concentrations of BA obtained experimentally, 1579

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582

Crystal Growth & Design

Article

plotted against the reciprocal of the total co-former concentrations, [DMP]. Linear regression with least-squares was applied, and no linear dependency was shown. Linear regression was again applied using eq 3. The results are shown in Figure 3. The solubility products and the binding constants

Table 3. Purities of Cinnamamide in Different Crystallization Experiments 9:1 CAM/CA with DMP with DMG with INA

CA (wt %)

decreased (%)

± ± ± ±

23.79 −7.88 10.93

0.85 0.65 0.92 0.76

0.078 0.101 0.153 0.142

test was performed to determine whether the amount of CA decreases statistically significantly at the 95% confidence interval with the addition of co-formers. The amounts of CA decreased statistically significantly with the addition of DMP but not significantly with the addition of DMG and INA. The effect of initial impurity concentration was investigated using the CAM/CA system. CA was added at a weight ratio of 1:4 to CAM. DMP (1:1 molar ratio of CA:DMP) was added to purify CAM crystals. The purities of CAM obtained from the crystallization experiments are shown in Table 4. The original

Figure 3. Linear regression results of the BA−DMP system.

were calculated using the slope and intercept obtained from Figure 3. The solubility product was 0.0053, and the binding constant for the 1:1 complex was 6.67. For the CAM/CA system, the addition of DMP significantly decreased the amount of CA incorporated into CAM crystal lattices. CA and DMP is known to form 1:1 co-crystals. Similar to the BA−DMP system, no linear dependence was shown between [CA] and (1/[DMP]). Linear regression was applied using eq 3, and the results are presented in Figure 4. Ksp (0.0018) and K11 (64.86) were calculated using the slope and intercept.

Table 4. Purities of Cinnamamide in Different Crystallization Experiments 4:1

CA (%)

decreased (%)

CAM/CA with DMP

2.01 ± 0.21 1.58 ± 0.20

21.56

mass percentage of CA in CAM crystal was 2.01(±0.21) wt %. Once DMP was added (1:1 molar ratio of CA:DMP), the amount of CA in CAM crystal decreased to 1.58(±0.20) wt %. No trace of DMP was detected in any resulting CAM solids. A t test was performed, and the amount of CA decreased statistically significantly with only the addition of DMP. To summarize, two systems showed statistically significant increases in the purity of the target: the BA−DMG system and the CA−DMP system. The addition of other co-formers did not have a significant effect on the amount of impurities incorporated into the crystal lattices of target compounds. In the BA−DMG system, the amount of BA did not decrease with increasing amount of DMG added. Complexation Measurement. To examine whether the level of complexation has a correlation to purification results or not, we estimated the binding constants of complexes in our systems of interest. For the BAM/BA system, DMG was an effective co-former for BAM purification. Since BA and DMG are known to form a 2:1 co-crystal, the stoichiometries and the binding constants of complexes were determined using the method described above for 2:1 co-crystals. Model 3 had a large error (0.4393) compared to other models (0.0065 for model 1 and 0.018 for models 2 and 4) and was not considered. The facts that K21 is 0 in model 4 and the values of Ksp and K11 are identical in models 2 and 4 (Ksp = 0.4413 and K11 = 0.3469) indicated that the assumption of having 2:1 complex did not change the behavior of model 2. That is, no 2:1 complex was present in the solution and model 4 was not considered. Model 1 had a smaller sum of square error than model 2 did. However, it is highly unlikely to form 2:1 co-crystals without complex formation in the solution. Therefore, the model with 1:1 complex formation was determined to be the best model. The BA−DMP system was also examined for complexation behavior. BA and DMP are known to form a 1:1 co-crystal. Following the method described above for systems forming 1:1 co-crystals, the total impurity concentrations, [BA], were

Figure 4. Linear regression results of the CA−DMP system.

The complexation behavior in the CA−INA system was examined as a comparison to that in the CA−DMP system. Linear regression between [CA] and (1/[INA]) was performed because CA forms 1:1 co-crystals with INA. A linear dependence was shown in Figure 5. Examination of the linear regression analysis revealed that the y-intercept was not statistically different from zero. This fact suggested that no complex was formed in the solution and the solubility product of the CA−INA system was the slope of the regression line, 0.023. The solubility products and the binding constants for the 1:1 complex for the BA−DMG, BA−DMP, CA−DMP, and CA−INA systems are summarized in Table 5. To summarize, we estimated the binding constants in four systems: BA−DMG, BA−DMP, CA−DMP, and CA−INA. The solubility product, Ksp, and the binding constant for the 1:1 complex, K11, were estimated by comparing the experimental 1580

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582

Crystal Growth & Design

Article

py, ultraviolet−visible spectroscopy, and infrared spectroscopy) and nuclear magnetic resonance (NMR). For spectroscopies, impurity peaks overlap with co-former peaks and it is difficult to identify the contribution from the impurity, the co-former, and their complex. For NMR measurements, the complexation was not observed due to the concentration limit allowed in NMR. Additionally, the solvent we used in purification experiments, ethanol, is a polar solvent that can interfere with hydrogen bonding formation between the impurity and the co-former. Hence, binding constants were the best indications for the level of complexation in the solution even if we only considered the interaction between the impurity and the co-former. With these observations, we know that it is possible to purify structurally similar compounds by adding co-formers which have the potential to complex with the impurities in solution. Further work will be required to determine whether the purification is due to the formation of impurity−co-former complexes and to develop a predictive tool for the selection of co-formers for purification.

Figure 5. Total CA concentration in equilibrium with the co-crystal, CA−INA, as a function of the reciprocal of total INA concentration in ethanol at 20 °C.

Table 5. Solubility Products and Binding Constants for (1) BA−DMG, (2) BA−DMP, (3) CA−DMP, and (4) CA−INA Ksp K11

BA−DMG

BA−DMP

CA−DMP

CA−INA

0.4413 0.3469

0.0053 6.67

0.0018 64.86

0.023 0

■

CONCLUSIONS In this paper, we described the purification and complexation behavior of the BA/BAM and CAM/CA systems after adding co-formers to the solution. The results demonstrated the possibility of purifying mixtures of structurally similar compounds by adding co-formers to their solutions. The addition of DMG significantly decreased the amount of BA in BAM crystals. The purity of CAM increased significantly after the addition of DMP. However, no clear correlation was shown between the level of complexation and the purification result. For the systems studied, the purification observed cannot be predicted using the binding constants between the impurity and the co-former.

data to the theoretical models. K11 was used to indicate the complexation level in the solution.

■

DISCUSSION Results for the BAM/BA system did not show a positive correlation between the complexation level in the solution and purification results. The addition of DMG decreased the amount of BA the most, but the binding constant of the BA− DMG complex was smaller than that of the BA−DMP complex. This result can be explained by the fact that DMP substituted into BAM crystal lattice. As shown in Table 2, after the addition of DMP, DMP was detected in CAM crystals (0.087(±0.020) wt %). It is likely that some BA−DMP complexes substituted into BAM crystal lattice. Therefore, despite the high complexation level in the BA−DMP system, the purity of BAM enhanced the most with the addition of DMG. A positive correlation was found between the level of complexation in the solution and purification results in the CAM/CA system. The amount of CA decreased the most with the addition of DMP and did not decrease significantly with the addition of INA. In addition, the binding constant of the CA− DMP complex was much larger than that of the CA−INA complex. We can conclude that, for the CAM/CA system, the more complexes formed in the solution, the better the purification results are. This positive correlation was observed when we compared the results of the two systems that showed significant decreases: the BAM/BA system with the addition of DMG and the CAM/ CA system with the addition of DMP. Although the original impurity levels were different in the BAM/BA and CAM/CA systems, we can compare purification results by calculating the decreases of impurity amounts in mol/g solids. The decrease of the CAM−CA−DMP system was 1.3 × 10−5 mol/g solids while the decrease of the BAM−BA−DMG system was 5.3 × 10−6 mol/g solids. The binding constant of CA−DMP complex was significantly larger than that of the BA−DMG complex. This positive correlation implies that the purification is due to the impurity complex formation in the solution. The complexation in purification experiments was not measurable using spectroscopies (including Raman spectrosco-

■

AUTHOR INFORMATION

Corresponding Author

*Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, 66-568, Cambridge, MA 02139. Phone: 617-452-3790. Fax: 617-253-2072. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS The support of Novartis for this research is gratefully acknowledged. ABBREVIATIONS BAM, benzamide; BA, benzoic acid; CAM, cinnamamide; CA, cinnamic acid; INA, isonicotinamide; DMP, 2-amino-4,6dimethylpyrimidine; DMG, dimethylglyoxime

■

REFERENCES

(1) Hsi, K. H.; Chadwick, K.; Fried, A.; Kenny, M.; Myerson, A. S. Separation of Impurities from Solution by Selective Co-crystal Formation. CrystEngComm 2012, 14, 2386−2388. (2) Berkovitch-Yellin, Z.; Addadi, L.; Idelson, M.; Lahav, M.; Leiserowitz, L. Controlled Modification of Crystal Habit via “TailorMade” Impurities. Application to Benzamide. Angew. Chem., Int. Ed. Engl. 1982, 21 (Suppl.8), 1336−1345. (3) Addadi, L.; Berkovitch-Yellin, Z.; Weissbuch, I.; van Mil, J.; Shimon, L. J. W.; Lahav, M.; Leiserowitz, L. Growth and Dissolution of Organic Crystals with “Tailor-Made” InhibitorsImplications in

1581

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582

Crystal Growth & Design

Article

Stereochemistry and Materials Science. Angew. Chem., Int. Ed. Engl. 1985, 24, 466−485. (4) Weissbuch, I.; Addadi, L.; Lahav, M.; Leiserowitz, L. Molecular Recognition at Crystal Interfaces. Science 1991, 253, 637−645. (5) Weissbuch, I.; Popovitz-Biro, R.; Lahav, M.; Leiserowitz, L. Understanding and Control of Nucleation, Growth, Habit, Dissolution and Structure of Two- and Three-Dimensional Crystals using “Tailormade” Auxiliaries. Acta Crystallogr. 1995, B51, 115−148. (6) Higuchi, T.; Zuck, D. A. Investigation of Some Complexes Formed in Solution by Caffeine. II. Benzoic Acid and Benzoate Ion. J. Am. Pharm. Assoc. 1953, 42, 132−138. (7) Higuchi, T.; Zuck, D. A. Investigation of Some Complexes Formed in Solution by Caffeine. III. Interactions Between Caffeine and Aspirin, p-Hydroxybenzoic Acid, m-Hydroxybenzoic Acid, Salicylic Acid, Salicylate Ion, and Butyl Paraben. J. Am. Pharm. Assoc. 1953, 42, 138−145. (8) Higuchi, T.; Lach, J. L. Investigation of Some Complexes Formed in Solution by Caffeine. IV. Interactions between Caffeine and Sulfathiazole, Sulfadiazine, p-aminobenzoic acid, Benzocaine, Phenobarbital, and Barbital. J. Am. Pharm. Assoc. 1954, 43, 349−354. (9) Higuchi, T.; Lach, J. L. Investigation of Some Complexes Formed in Solution by Caffeine. V. Interactions between Caffeine and paminobenzoic acid, m-hydroxybenzoic acid, Picric acid, o-phthalic acid, Suberic acid, and Valeric acid. J. Am. Pharm. Assoc. 1954, 43, 524−527. (10) Higuchi, T.; Pitman, I. H. Caffeine Complexes with Low Water Solubility: Synthesis and Dissolution Rates of 1:l and 1:2 CaffeineGentisic Acid Complexes. J. Pharm. Sci. 1973, 62, 55−58. (11) Nehm, S. J.; Rodríguez-Spong, B.; Rodríguez-Hornedo, N. Phase Solubility Diagrams of Cocrystals Are Explained by Solubility Product and Solution Complexation. Cryst. Growth Des. 2006, 6, 592− 600. (12) Jayasankar, A.; Reddy, L. S.; Bethune, S. J.; Rodríguez-Hornedo, N. Role of Cocrystal and Solution Chemistry on the Formation and Stability of Cocrystals with Different Stoichiometry. Cryst. Growth Des. 2009, 9, 889−897. (13) Alhalaweh, A.; George, S.; Boström, D.; Velaga, S. P. 1:1 and 2:1 Urea−Succinic Acid Cocrystals: Structural Diversity, Solution Chemistry, and Thermodynamic Stability. Cryst. Growth Des. 2010, 10, 4847−4855. (14) Alhalaweh, A.; Sokolowski, A.; Rodríguez-Hornedo, N.; Velaga, S. P. Solubility Behavior and Solution Chemistry of Indomethacin Cocrystals in Organic Solvents. Cryst. Growth Des. 2011, 11, 3923− 3929. (15) Xu, S.; Dang, L.; Wei, H. Solid-Liquid Phase Equilibrium and Phase Diagram for the Ternary Carbamazepine-Succinic Acid-Ethanol or Acetone System at (298.15 and 308.15) K. J. Chem. Eng. Data 2011, 56, 2746−2750. (16) Grossjohanna, C.; Ecclesb, K. S.; Maguirec, A. R.; Lawrenceb, S. E.; Tajbera, L.; Corrigana, O. I.; Healya, A. M. Characterisation, Solubility and Intrinsic Dissolution Behaviour of Benzamide:Dibenzyl Sulfoxide Cocrystal. Int. J. Pharm. 2012, 422, 24−32.

1582

dx.doi.org/10.1021/cg301814z | Cryst. Growth Des. 2013, 13, 1577−1582