Article pubs.acs.org/crystal
Symmetry Breaking: Polymorphic Form Selection by Enantiomers of the Melatonin Agonist and Its Missing Polymorph Gregory A. Stephenson,*,† John Kendrick,‡ Craig Wolfangel,† and Frank J. J. Leusen‡ †
Lilly Research Laboratories, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, Indiana, 46285, United States School of Life Sciences, University of Bradford, Bradford, BD7 1DP, United Kingdom
‡
S Supporting Information *
ABSTRACT: Synthesis of a melatonin agonist for treatment of sleep disorders produced a pair of enantiomers, of which one is biologically active. Two polymorphs were discovered using the inactive enantiomer, conserving the active enantiomer for toxicological testing. Later studies with the active enantiomer yielded only the metastable form, despite more than 1000 attempts to isolate the stable form. The difficulty is surprising, since the stable form is favored by 0.7 kcal mol−1, which is toward the extreme for stability differences between organic polymorphs. Study of individual enantiomers allowed the phase behavior of polymorphs of greatly different energy to be examined without interconversion. A number of unusual features are noted. After the stable polymorph of the inactive enantiomer was nucleated, the metastable form became very difficult to isolate. The metastable form converts into a less soluble monohydrate structure in water, whereas the stable polymorph does not due to its reduced activity. Both chiral polymorphs are denser than the racemic crystalline form at low temperature, the stable form being at the extreme for chiral-racemic pairs. Free energy-temperature relations predict “spontaneous resolution” of the racemic crystalline form into a conglomerate mixture of stable polymorph at low temperature. The unusual characteristics of the system are explained by hydrogen bonding and conformational flexibility of the molecule. Ab initio calculations aid in understanding the relative contributions of these interactions to the lattice energies and the role that conformational energy differences play in the polymorphic stability. This system highlights the importance of the creation of the very first nuclei of a crystalline form. The reluctance of the stable form to nucleate is attributed to a large energy difference between polymorphic forms. The large interfacial tension for primary nucleation reduces the probability of forming clusters of size sufficient for favorable growth in the absence of heterogeneous nucleation. This study highlights how nucleation of a new form can revise the readily “accessible” region of a compound’s crystal form landscape.
■
rhythm disorders.2,3 It is a structural analogue of the endogenous hormone melatonin. During the early stages of drug discovery, synthetic routes commonly produce racemic mixtures of enantiomers that are separated prior to testing. Generally one enantiomer is more active than the other, and a single enantiomer is selected for further testing. To conserve materials for use in biological and toxicological testing, requests are sometimes made that early crystallization studies be initiated using the inactive enantiomer of a potential new drug substance. In theory, either enantiomer will produce equivalent crystalline forms, though of opposite chirality. In the case of the melatonin agonist (MA; see Figure 1) the inactive enantiomer (S-MA) was isolated on at least five different occasions, four yielding a metastable anhydrous form, S-MA Form 1, and the fifth yielding a new crystalline form, S-MA Form 2, which is much more thermodynamically stable. The
INTRODUCTION Knowledge of the physical forms in which a pharmaceutical compound may exist, their thermodynamic relations, and potential for interconversion is essential for the development of robust crystallization and manufacturing processes. Control of polymorphic form composition in the product is a requirement for product registration, where changes may affect bioavailability and chemical stability, primarily due to differences in form solubility.1 The study of polymorphism is routine in pharmaceutical development, where the most stable crystalline form is typically selected for the product to minimize risk of form conversion during production and storage. It is difficult to know when a survey of “phase space” has been sufficiently rigorous to ensure that the most stable form has been identified. No matter the rigor of the search, optimization of synthetic route and the scaling of processes during development combined with the stochastic nature of nucleation can lead to surprising results. LY156735 (R-MA) is a potent and selective melatonin agonist investigated for the treatment of insomnia and circadian © XXXX American Chemical Society
Received: March 26, 2012 Revised: June 4, 2012
A
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
resolution electron microscopy.15 However, understanding and control of primary nucleation remains one of the most fundamental, yet challenging, elements in the study of crystallization.16 Its stochastic nature is often minimized by intentional seeding of the crystallization process, whereas unintentional seeding has led to surprising results.17 Mysteries surrounding crystallization have existed for centuries, and many of these mysteries involve crystal nucleation. For instance, xylitol was first prepared in 1891 and was considered a liquid until 1941 when it first crystallized as a form melting at 61 °C.18 In 1943, it was crystallized in its present form, melting at 94 °C.19 Polymorphic form change results in commensurate change in physical properties that are often undesired. Production sites have been isolated and travel between them limited so that cross contamination by seeds of different polymorphs could be avoided. New facilities for crystallization have been constructed to ensure unwanted seeds are not present.6,17,20 Woodward and McCrone so aptly state what many experienced in crystallizations have observed at one time or another; “most interesting to us is the fact that once one laboratory has recrystallized a compound, either for the first time or in a more stable form, other labs were able to do so, as though the seeds of crystallization, as dust, had been carried upon the winds from end to end of the earth.”21,22 It has been suggested that once nuclei of a given form have been created, the world where its nuclei have not been ceases to exist and is forever changed.17,23 Clearly something changes as the result of a form’s first nucleation, likely a consequence of secondary nucleation thereafter. Yet still, disappearing, reappearing, and late appearing polymorphs warrant further explanation.23−27 Much of the challenge in studying polymorphic systems results from the relatively small energy window within which polymorphs exist and is further complicated by crossing of their energy−temperature stability relations.28,29 Nowhere has this been so evident as in the colorful system of ROY (red − orange − yellow) polymorphs, where secondary nucleation has led to the growth of even a metastable form upon the surface of a more stable polymorph.30 The system has shown that competition exists between forms of similar energy and that both nucleation and growth processes need to be considered as their combination ultimately dictates the polymorphic outcome.31 Through the melatonin agonist, we introduce a system of significant contrast to that of ROY, yet equally fascinating. A relatively large energy difference separates its polymorphic forms, leading to many interesting findings and violation of many commonly held rules of crystallization. As of the date of this publication, the stable polymorph of the active enantiomer of MA remains a “missing” polymorph.
Figure 1. Molecular Structure of R-MA, LY156735.
driving force for conversion of the metastable form to the stable form is large, in excess of a 3-fold difference in solubility, corresponding to a Gibbs free energy difference of 0.7 kcal mol−1 at room temperature. This energy difference is much larger than the average of 0.3 kcal mol−1 for polymorphic systems, being more than two standard deviations from the mean.4 Work with the active enantiomer, R-MA, produced surprising results as the more stable polymorph could not be crystallized despite thousands of attempts at small scale, as well as at least 10 multi-kilogram scale lots.5 Knowledge of the existence of a much more stable form of the active compound resulted in plans to conduct early clinical studies using the metastable form, expecting that the stable form would eventually be found. Once found, a method for its isolation and production would be developed, and a “form switch” would occur, where bridging bioavailability studies would establish the dose required to achieve equivalent exposure. Fortunately, the anticipated efficacious dose was low so that the new form could readily be formulated in a tablet of sufficiently small size once the missing polymorph was found. This is in contrast to the classic example of a late appearing polymorph, the HIV protease inhibitor ritonavir, where the stable form was identified after product launch.6 The required dose was very high and its solubility low such that the appearance of a much less soluble polymorphic form resulted in product recall, market shortage, and reformulation at considerable expense.7 Examination of the two enantiomers of MA enabled the study of the phase behavior of two polymorphs of very different energies. This is particularly advantageous in the study of S-MA Form 1, as it is now extremely difficult to isolate, a disappearing polymorph resulting from discovery of S-MA Form 2 and the presence of its nuclei. The energy relationship and crystallization behavior of Form 1 can still be studied by using R-MA, as it appears “stuck” in its metastable form, awaiting the appearance of the missing polymorph. To our knowledge, this is the first report of its kind, highlighting the importance and at times unpredictability of formation of the very first nuclei of a given form. The order of form appearance of MA follows Ostwald’s rule of stages, where upon leaving an unstable state a system does not seek the most stable state, but rather the nearest metastable state that can be reached with minimum loss of free energy.8 Accordingly, the more-soluble metastable form is expected to, initially at least, have a faster nucleation rate. Ostwald’s rule has been questioned.9−11 Though empirical, it remains one of the most frequently cited rules in crystallization science. Kinetic and thermodynamic explanations have been proposed.12−14 Recently, images of an amorphous phase converting real-time through a series of metastable intermediate phases on its way to a stable crystalline form have been observed in situ using high-
■
EXPERIMENTAL SECTION
Materials. Synthesis and Separation of the Enantiomers. R-MA was synthesized at least 37 times, with 11 lots at a multi-kilogram scale. In its most developed synthetic route, chiral resolution occurs in the next to last step, prior to acylation of R-β-methyltryptamine to form the active enantiomer. Chiral resolution is accomplished by diastereomeric salt formation of racemic β-methyltryptamine and Ltartaric acid. An example of the final steps leading to the production of R-MA involves formation of an L-tartrate salt of the racemic intermediate β-methyltryptamine and is provided in the Supporting Information. The primary lot used throughout these studies had a defined e.e. of 99.8% at the time of its production. B
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
The inactive enantiomer, S-MA, was prepared by the same synthetic route as the active enantiomer; however, it was isolated from the soluble fraction of the salt-forming reaction with β-methyltryptamine. The same procedure was used to form the base of the S enantiomer, acylate, and crystallize it. The material isolated is very pure and essentially indistinguishable from the active enantiomer by liquid chromatography, having similar purity and related substance profile. The primary lot of S-MA used throughout the studies was determined to have a 98.6% e.e. at the time of its production. The lots of both forms have very similar impurity profiles, with the presence of the same related substances. Each lot had the same four primary related substances, with the active enantiomer being slightly more pure compared to the inactive enantiomer, 99.47 and 99.28%, respectively. The only observable difference was the increased amount of the substance located with a retention time of 18.70 min for S-MA, having related substance levels of 0.44% compared to 0.25% for the active enantiomer. The chromatograms are provided in Supporting Information as SI.T1 and SI.F1. Material Characterizations. Scanning electron microscopy (SEM) powder samples were attached to an aluminum stub with a carbon sticky pad and sputter coated with gold/palladium. Samples were imaged in high vacuum mode with an FEI QuantaFEG SEM, using a backscatter detector. Powder X-ray diffraction (PXRD) patterns were obtained in Bragg− Brentano reflection mode geometry using a Bruker Endeavor D4 diffractometer equipped with a copper source (Cu Kα 1.54056 Å), and a Vantec linear detector with collimator slits. A powder diffraction pattern was acquired under ambient conditions at a power setting of 40 mA and 45 kV. The diffraction pattern was collected over the range of 4−40° 2-theta. Single crystal X-ray diffraction (SCXRD) data were collected with a single crystal mounted on a thin glass fiber at −173 °C. Data were collected using a Cu Kα radiation source (λ = 1.54178 Å) and a Bruker D8 based 3-circle goniometer diffractometer equipped with a SMART 6000CCD area detector.32 Cell refinement and data reduction were performed using the SAINT program.33 Absorption corrections were applied using SADABS.34 The structures were solved by direct methods.35 All hydrogen atoms were located in the difference Fourier maps and all atomic parameters were independently refined. Space group choice was confirmed by successful convergence of the fullmatrix least-squares refinement on F2.36 The absolute stereochemistry was determined by refinement of the absolute structure parameter. Polarized light microscopy was performed using a Leica DMIRE2 polarized light confocal microscope with a 2.5× strain-free objective. A differential interference contrast filter was used to show birefringence. Bright, light-colored areas in the visual field indicate birefringence. Thermal gravimetric analyses (TGA) were carried out on a TA TGA unit (model TGA Q5000). Samples were heated in open platinum pans from 25 to 200−350 at 10 °C/min with a nitrogen purge of 50 mL/min. The TGA temperature was calibrated with an indium/aluminum standard, MP = 156.6 and 660.3 °C. The weight calibration was performed with manufacturer-supplied standards and verified against sodium citrate dihydrate desolvation. Differential scanning calorimetry (DSC) was conducted generally at 10 °C/min in crimped Al pans using a TA DSC-Q1000 under 50 mL/ min nitrogen purge. The temperature and heat flow were both calibrated using indium. The melting and eutectic melting data reported were the average of three measurements. The melting temperatures were taken as the onsets of melting endotherms. 13 C cross-polarization/magic angle spinning (CP/MAS) solid-state NMR spectra were obtained using a Bruker Avance II 400 MHz NMR spectrometer operating at a carbon frequency of 100.622 MHz and equipped with a Bruker 4 mm triple resonance probe. TOSS sideband suppression was used along with cross-polarization employing SPINAL64 decoupling (70.8 W) and a RAMP100 shaped H-nucleus CP pulse. Acquisition parameters were as follows: 90° proton r.f. pulse width 2.5 μs, contact time 3.0 ms, pulse repetition time of 5 s, MAS frequency of 10 kHz, spectral width of 30 kHz, ranged from 107 to 1840 scans with an acquisition time of 34 ms. Chemical shifts were
referenced to adamantane (δ = 29.5 ppm) in a separate experiment. An 80 μs delay time was used in the dipolar dephasing experiment. Purity Analysis. Representative samples of Forms 1 and 2 were analyzed by a validated stability indicating method for differences in purity. A sample of each was prepared at approximately 0.1 mg/mL and assayed by HPLC utilizing a DAD detector. Analysis was performed using an Agilent 1290 Infinity HPLC equipped with a DAD detector. An Agilent RX-C8, 5 μm, 4.6 × 250 mm column controlled at 25 °C was used. The sample diluent consisted of 25% methanol in water solution. The mobile phase was A = 0.1% phosphoric acid in water; B = methanol. A gradient was employed as follows for pump conditions: 25% mobile phase B isocratic for 5 min, followed by a linear increase to 75% mobile phase B over 20 min; isocratic hold for 5 min at 75% mobile phase B, completed with a linear gradient over 1 min back to 25% mobile phase B and an isocratic equilibration of 6 min. The flow rate was 1.0 mL per minute, the injection volume is 10 μL, and the wavelength of detection set to 225 nm. Solubility Analysis. Excess amounts of R-MA Form 1, S-MA Form 2, R-MA monohydrate, and RS-MA (the racemate) were weighed into vials to which was added an appropriate amount of water to create a slurry. Samples were briefly sonicated to ensure they were well mixed and placed in temperature-controlled thermomixers set at their respective temperatures and 750 rpm. The five samples had their thermomixer unit placed inside a refrigeration unit also set at that temperature to aid in the temperature uniformity. Samples were allowed to equilibrate at temperature and pulled after 6 and 25 h. At the pull time point, great care was taken to ensure all equipment coming into contact with the sample was prewarmed or cooled to the temperature of the sample. Samples were centrifuged very briefly (approximately 5 s), an aliquot was withdrawn and filtered through a 0.45 μM PTFE membrane syringe filter, and aliquots of the filtrate were immediately diluted to HPLC analysis concentration. The solid residues from the samples had excess liquid removed from them and were assayed by PXRD for form identification. All samples were done in duplicate. HPLC analysis was conducted using an Agilent 1290 Infinity HPLC equipped with a DAD detector. An Agilent RX-C8, 5 μm, 4.6 × 250 mm column controlled at 25 °C was used. The sample diluent consisted of 25% methanol in water solution. The mobile phase was A = 0.1% phosphoric acid in water; B = methanol. A gradient was employed as follows for pump conditions: 25% mobile phase B isocratic for 5 min, followed by a linear increase to 75% mobile phase B over 20 min; isocratic hold for 5 min at 75% mobile phase B, completed with a linear gradient over 1 min back to 25% mobile phase B and an isocratic equilibration of 6 min. The flow rate was 1.0 mL per minute, the injection volume is 10 μL, and the wavelength of detection set to 230 nm. Conformational Analysis. To investigate the conformational flexibility of MA, the molecular structure was sketched into Materials Studio.37 The amide group was assumed to remain in a trans configuration, and due to steric factors the methoxy group was assumed to remain in an orientation in which the methyl group points away from the nearby chlorine atom. An unrestrained grid scan was performed and the optimized structures clustered and ordered according to energy calculated with the Dreiding force field38 and Gasteiger atomic charges.39 After clustering, the 61 lowest energy structures were reoptimized and reranked using a density functional theory (DFT) quantum mechanical (QM) method based on the BLYP functional40 with the 6-31G* basis set41 and with a Moller−Plesset second-order perturbation theory (MP2) level of theory using double and triple-ζ plus polarization basis sets.42 In the two-stage MP2 calculations the molecular conformation was first optimized using the resolution of the identity approximation for both Coulomb and exchange terms. In the second step, the geometry was further optimized without the resolution of the identity approximation. Molecular QM calculations were performed using ORCA version 2.8.43 To compare the conformational analysis results to experiment, the molecular structures were extracted from the experimental crystal structures and optimized using the molecular mechanics (Dreiding force field with Gasteiger charges) and molecular QM (ORCA DFT) C
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
approaches as detailed above. Additional molecular QM calculations were performed at the MP2 level of theory using the double42 and triple-ζ44 plus polarization basis sets, referred to as MP2(SVP) and MP2(TZVP) respectively. The MP2(SVP) conformers were initially optimized using the resolution of the identity approximation for Coulomb and exchange terms and then further optimized without this approximation. In the MP2(TZVP) calculations the resolution of the identity approximation was used throughout because of memory restrictions. In all cases the geometry of each conformer was fully optimized. Lattice Energy Calculations. The GRACE software package45 was used to optimize the crystal structures and to calculate their lattice energies. Energies and gradients are provided by a solid state DFT method with corrections for dispersive interactions DFT(D).46 GRACE uses the VASP47 program to calculate the lattice energy and its gradients using the PW91 density functional.48 The dispersive correction is provided by a damped molecular mechanical dispersive interaction. Starting with the experimental crystal structures, the unit cells and molecular geometries were fully optimized within the constraints of the experimental space group symmetry. The DFT(D) lattice energy is defined by the energy of the process of going from molecules in the “gas phase” to the crystal. It can be broken down into terms arising from conformational, configurational, van der Waals, and Coulomb contributions. In the following a “molecular calculation” refers to a DFT(D) calculation of a molecule in a unit cell which is large enough so that the molecule does not interact with images of itself. Typically unit cells with dimensions of over 23 Å were needed to ensure convergence of the molecular calculations. A “crystal calculation” refers to a DFT(D) optimization of a crystal structure. The Coulomb contribution to the DFT(D) lattice energy was estimated from the atomic charges that reproduce the isolated molecular electrostatic potential. The van der Waals contribution was estimated as the difference in the molecular and solid state dispersive energy corrections used by the DFT(D) method. The conformational contribution arises from the difference in energies between molecularly optimized structures, while the configurational energies arise from the change in energies of conformations on moving from the gas phase to the solid state. The configurational energies were estimated from molecular calculations of the energy at an optimized crystal geometry. The sum of the configurational and conformational energy terms represents the increase in internal energy of the molecule due to its presence in the crystal. The remaining contributions to the lattice energy, including terms arising from polarization and induction, are calculated from the difference between the total lattice energies and the sum of all other contributions: conformational, configuration, van der Waals, and Coulomb.
exhibited partial transformation to Form 2 upon heating. This was evident through a small broad exotherm immediately following the melt endotherm of S-MA Form 1. The conversion to Form 2 was incomplete during the DSC experiment, as the enthalpy of fusion of the higher melt ranged from between 2 and 7% of the value of samples of Form 2 that had been produced by recrystallization. The level of Form 2 crystals or seeds present in the sample before DSC analysis was very low, since they were not detected by PXRD even upon close examination of the positions where Form 2 peaks should be observed. Typically such inspection of highly crystalline materials, as in the case of MA, results in detection down to tenths of one percent. Clearly the amount present before heating in the DSC was well below the 2−7% that were created in the DSC experiment based on its enthalpy of fusion relative to the pure Form 2. Whether seeds exist within the sample prior to heating in the DSC or whether spontaneous nucleation occurred upon heating is not certain. Hot-stage microscopic examination of the melting behavior of the samples in which Form 2 crystals were not detected by PXRD showed small droplets forming at 127 °C. Some drops appeared to nucleate Form 2 spontaneously while the majority of the drops remained clear through 150 °C of heating, as well as upon return to room temperature. The drops that did crystallize upon heating were noticeably clear briefly before nucleation, indicating that any existing nuclei must be extremely small. Even in the melted droplets that nucleated Form 2 crystals, much of the drop remained clear and did not crystallize completely. The fact that only incomplete growth took place in the melt once nucleation had occurred attests to the relatively sluggish nature of the crystallization of Form 2, at least from its melt at elevated temperature. Reluctance of organic molecules to recrystallize from their melt is not uncommon for compounds capable of forming intermolecular hydrogen bonds. Random molecular diffusion occurs by breaking hydrogen bonds with adjacent molecules and then reforming them with new neighboring molecules. The nucleation process is stochastic in nature, and when the nuclei lack the hydrogen bonding of the type that exists within the crystal, the nucleus often does not achieve the critical size necessary for its successful growth in the melt.50 Forty samples of the active enantiomer, R-MA, that had been stored for more than 14 years were analyzed by PXRD and DSC. Surprisingly, none of the lots had transformed into the stable polymorphic form. None showed the slightest hint of the melt-recrystallization event observed in the aged S-MA samples. Comparative Evaporative Crystallizations. In an attempt to reconstruct the 14-year-old observations of the relative crystallization behavior of the inactive (S-MA) and active (R-MA) enantiomers, an experiment was designed where each enantiomer was dissolved separately in 24 different solvents which were then allowed to evaporate. The samples were initially distributed into individual vials by dissolving 500 mg of either enantiomer into 13 mL of methanol. A total of 0.5 mL of the solution was transferred into the individual glass vials and allowed to evaporate to dryness. Different solvents were dispensed into each of the vials. The solutions were examined by microscopy to ensure complete dissolution. In most cases the sample went into solution in the first aliquot, with exceptions of cumene, n-amyl acetate, 3-methyl-1-butanol, cyclohexanone, and anisole, for which additional aliquots were required. The solutions were evaporated and analyzed
■
RESULTS AND DISCUSSION Revisiting the Problem. With the discontinuation of the development of MA into a marketable drug, the opportunity has arisen to describe the unique nature of this system in the scientific literature. Many years have passed, along with patent expiry, since the original research into MA. In revisiting this case, the first step involved examination of small samples of each lot that remained of each enantiomer, to determine if the problem persisted. PXRD analysis of the five available lots of the inactive enantiomer, S-MA, indicated that three of the lots had remained S-MA Form 1. The fourth lot, which had many years before been identified as pure Form 1 by PXRD, had now partially converted to Form 2. It was determined to be 79% SMA Form 2 after storage under ambient lab conditions for 14 years.49 The fifth lot was S-MA Form 2 initially produced and had remained so. The DSC results were slightly different than those observed by PXRD, likely due to their relative sensitivity to very low levels of nuclei. DSC data for the three S-MA Form 1 lots differed from their original DSC traces in that they now D
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
by PXRD. Evaporation times ranged from hours to weeks, depending on solvent volatility. A few samples did not crystallize but instead formed a glass in the bottom of the vial and are noted as such in Table 1. Although the glass Table 1. Comparative Evaporative Crystallization Results for the Active (R-MA) and Inactive Enantiomers (S-MA) solvent acetone acetonitrile ethanol methanol 2-propanol 1-butanol ethyl acetate isopropyl acetate methyl ethyl ketone methylene chloride cumene n-amyl acetate 3-methyl butanol toluene cyclohexanone anisole methyl acetate n-propanol tetrahydrofuran acetic acid methyl isobutyl ketone chloroform dioxane “wet” ethyl acetate
R-MA Form Form Form Form Form Form Form Form Form Form Form Form Form Form oila Form Form Form Form oila Form Form Form Form
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
S-MA Form oila oila Form Form Form Form Form Form Form Form Form Form Form oila Form Form Form Form oilb Form Form Form Form
2
2 2 2 2 2 2 2 2 2 2 2
Figure 2. MA crystal habits; Form 2 (top left), Form 1 (top right), racemate (lower left), hydrate (lower right).
Differential Scanning Calorimetry Analysis. The thermal properties of the forms were examined by HSM and DSC. Comparison thermograms representative of the individual forms are provided in the Supporting Information. Five replicate runs of each pure-phase sample were analyzed, resulting in RS-MA having an onset of melting of 136.1 °C and enthalpy of fusion of 34.1 kJ/mol. R-MA Form 1 crystals have a melt-onset of 127.9 °C and enthalpy of fusion of 29.2 kJ/mol. Form 2 crystals of S-MA exhibit a single melt endotherm with an onset of melting at 147.0 °C and enthalpy of fusion is 36.1 kJ/mol. The R-MA monohydrate crystal form has a broad endothermic event that occurs at 77.8 °C followed by an overlapping, sharper endothermic event at 81.8 °C corresponding to a loss of water in its TGA of 6.1% by weight followed by crystallization and a subsequent endotherm of RMA Form 1 at 127.6 °C. The DSC traces and thermal data for the two polymorphs and the racemic material are provided in Supporting Information (SI.F2 and SI.T2). Power X-ray Diffraction. The diffraction patterns of R-MA Form 1 crystals and RS-MA crystals are very similar. Initially it was thought that R-MA Form 1 and RS-MA might not be discrete phases at all but rather represent related structures of a solid-solution.51 Studies of mixtures of these forms at different compositions by DSC showed melt eutectics typical of discrete phases.52 Single crystal diffraction later confirmed their differing structures as well as their similarity. Representative PXRD patterns are provided in Figure 3. Conformational Polymorphs. The molecular conformations observed in polymorphic Forms 1 and 2 differ, presumably due to the requirements of efficient packing in the different crystalline lattices. Because of the pronounced differences, the polymorphs are considered conformational polymorphs. While the racemate and the hydrate forms do not meet the necessary criteria to be considered polymorphs per se, their molecular conformations are more similar to polymorphic Form 1 and Form 2, respectively, as shown in Figure 4. In the crystal structures of RS-MA and R-MA Form 1, the conformation of the amide side chain is “folded back” with respect to the indole ring, adopting a gauche conformation by rotation about the C10−C11 bond, whereas this side chain
2 2 2 2 2 2 2 2
a
When oil was scraped from the vial, it crystallized as Form 1. Indicates that Form 2 crystallized upon scrapping from vial, based on PXRD.
b
showed no evidence of birefringence, the sample was scraped from the vial and analyzed by PXRD. Upon scraping, the samples crystallized to the form denoted with an *. Each of the R-MA samples that crystallized directly from solution resulted in polymorphic Form 1. Each sample prepared with S-MA crystallized directly from solution as the stable polymorphic Form 2. It is worth noting that three of the samples of S-MA that had initially oiled out, but had crystallized upon scraping from the vial for analysis by PXRD, resulted in production of the metastable polymorph, whereas every attempt thereafter to isolate this form directly from solution produced the stable polymorphic form. This would be consistent with slow molecular diffusion in the oil favoring its metastable form. Therefore, the characteristics of the MA polymorphic system had remained unchanged over the intervening years, which was unexpected. Crystal Habit. Scanning electron microscopy was used to illustrate general habit of the crystals as shown in Figure 2. There is similarity in the shape of each of the crystalline forms, with the exception of S-MA Form 2. R-MA Form 1, the RSMA, and the R-MA monohydrate forms are acicular to bladed in shape, elongated along {100} consistent with each having a similarly short unit cell dimension relative to their other two. Form 2 is generally observed as “equant” crystals, reflecting its more evenly distributed unit cell metrics. E
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
R-MA Form 1 and RS-MA Structures. The crystallographic details for the structures are presented in Table 3. RMA Form 1 and RS-MA are very similar in packing. Such observations have been noted before for other chiral-racemic crystalline pairs with similar unit cells.53,54 The racemic form crystallizes in space group P21/c, with a β angle of 92.95° and a single independent conformer in its asymmetric unit. Form 1 crystallizes in the chiral orthorhombic space group P212121, also with a single conformer in the asymmetric unit. The RS-MA crystal form is composed of heterochiral molecular layers, where each molecule is related by a glide plane, within which all of the hydrogen bonding occurs. Successive layers are created by inversion and translation along the unique b-axis (18.14 Å axis). In the chiral structure of R-MA Form 1, hydrogen bonding occurs within a molecular layer in the a−b plane where each molecule is related by a 2-fold axis. Again there is no hydrogen bonding between successive layers, which are formed by 2-fold translation along c. The c-axis is the metric equivalent to the unique b axis in RS-MA. Within the individual layers hydrogen bonds form a continuous chain, propagated by unit cell translation of successive homochiral molecules. This bond involves the trans-amide functionality, where the amide donates a hydrogen bond to the oxygen of the amide group of the next molecule, N(12)−H···O(13), having the motif description C1,1(4).55 The hydrogen bond is directed along the short axis in both the R-MA and the RS-MA structures and is likely to be partly responsible for the needle-like habit of the crystal forms. The second type of hydrogen bond found in both RS-MA and RMA Form 1 involves an interaction between the indole amine hydrogen and the other lone pair of electrons on the amidecarbonyl oxygen, O13. The chain motif, C1,1(9), translates within a layer where successive molecules are related by glide planes in RS-MA or by 2-fold translation in R-MA Form 1. Direct comparison of RS-MA to R-MA Form 1 structures immediately highlights similarities in packing that results in their similar PXRD patterns. By creating a supercell and projecting along the short axis, one can see that the overall molecular arrangement is very similar between the structures (see Figure 5). In the structure of RS-MA, the active Renantiomer is depicted in yellow, whereas the S-enantiomer is shown in pink. Closer inspection illustrates how the enantiomeric differences are accommodated in the two structures. By overlaying molecular layers of the two structures and then extracting molecular pairs from the structures and rotating them 90°, the differences become clear. Comparison of molecules of equivalent chirality in R-MA Form 1 and the Renantiomer (yellow) of the RS-MA structure shows that the indole moieties superimpose, but the amide side-chains are arranged in opposite directions, although both are directed such that the amide hydrogen bond occurs along the fast growing (short) axis of the lattice. Comparison of the S-enantiomer (pink) from the RS-MA structure versus the corresponding R-
Figure 3. Comparison of the PXRD patterns of the racemic crystal (top), chiral crystal Form 1 (upper-middle), Form 2 (lower-middle), and the monohydrate form (bottom).
Figure 4. Comparison of the molecular conformations observed in the crystal structures of MA. The R-isomer from RS-MA (top left), R-MA Form 1 (top right), S-MA Form 2 (lower left), and the R-MA monohydrate (bottom right).
adopts an extended configuration in the structures of the stable polymorph and hydrate forms. The orientation of the transamide functionality differs for RS-MA and R-MA Form 1, as the amide group is rotated by nearly 90° in their respective structures. In either one, however, the “folded back” conformation allows the amide groups to form a continuous hydrogen-bonded chain between N12−H and O13 in successive molecules related by translation (hydrogen bond motif C1,1(4)). In the stable polymorphic form, S-MA Form 2, and the monohydrate form of R-MA the side chains adopt the more extended trans conformation about the C3−C10−C11− N12 torsion. Torsion angles representative of the alkyl-amide side chain conformation are provided in Table 2. Table 2. Comparison of Side-Chain Torsion Angles of MA torsion T1: T2: T3: T4: a
C2−C3−C10−C11 C3−C10−C11−N12 C10−C11−N12−C13 C11−N12−C13−C14
RS-MA (°)
R-MA Form 1 (°)
S-MA Form 2a (°)
R-MA H2O (°)
87.4 52.3 78.0 175.2
99.8 64.1 −122.2 179.2
128.6 160.6 −76.3 −177.8
111.1 173.0 101.7 178.8
The sign of the torsion angles of S-MA Form 2 is inversed for ease of comparison. F
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
Table 3. Comparison of Crystallographic Data for MA Structuresa identification
RS-MA
R-MA Form 1
S-MA Form 2
R-MA hydrate
empirical formula formula weight crystal system space group a, Å b, Å c, Å β, ° density, g/cm3 (calculated) crystal size, mm3 theta range for data collection, ° reflections collected independent reflections goodness-of-fit on F2 final R indices [I > 2σ(I)]
C14H17ClN2O2 280.75 monoclinic P2(1)/c 4.93520(10) 18.1423(4) 15.6111(3) 92.9510(10) 1.336 0.40 × 0.21 × 0.18 3.74−70.22 7897 2337 1.09 R1 = 0.0450 wR2 = 0.1427 0.274 and −0.348
C14H17ClN2O2 280.75 orthorhombic P2(1)2(1)2(1) 4.96890(10) 15.5039(2) 17.9606(3) 90 1.348 0.69 × 0.11 × 0.07 3.77−65.01 4789 2075 1.066 R1 = 0.0409 wR2 = 0.1079 0.355 and −0.196
C14H17ClN2O2 280.75 orthorhombic P2(1)2(1)2(1) 9.1740(4) 11.5383(5) 12.7475(5) 90 1.382 0.20 × 0.20 × 0.20 5.17−69.09 7148 2362 1.12 R1 = 0.0419 wR2 = 0.1031 0.252 and −0.301
C14H17ClN2O2 H2O 298.76 orthorhombic P2(1)2(1)2(1) 4.61780(10) 11.1219(2) 28.1075(6) 90 1.375 0.30 × 0.07 × 0.07 3.14−58.80 4186 1911 1.008 R1 = 0.0439 wR2 = 0.1074 0.231 and −0.317
largest diff peak and hole e·Å−3 a
Z = 4, temp = 100(2) K, λ = 1.54178 Å.
geometries and descriptors is provided in the Supporting Information, SI.T3. R-MA Monohydrate Structure. In the monohydrate structure of the active enantiomer, the molecule adopts a conformation similar to that found in the stable polymorphic form, having a more extended amide side-chain. The hydrogen atoms of the water molecules were located in the differenceFourier maps and their positions refined isotropically. The hydrogen bonding potential of the water molecule is fully satisfied, donating to the amide oxygen and another water molecule while its two electron pairs serve as acceptors for the indole hydrogen and another water molecule. The amide N−H donates donates to the methoxy oxygen of an adjacent molecule. The water molecules cluster along the 2-fold of the shorter, a-axis, making a solvent tunnel that runs in the direction of fast growth. Polymorphic Form 2. The stable polymorph differs greatly from the other structures. Molecules in the structure form corrugated molecular layers parallel with the a−b plane. Successive planes propagate along the c-axis. The amine hydrogen of the indole forms a hydrogen bond to the adjacent molecule’s carbonyl oxygen, N(1)-H to O(13), description C1,1(9)a. An unusual feature of the structure is that the amide hydrogen does not form a traditional hydrogen bond; however, a close contact is made between the amide hydrogen and the electron-rich pyrrole ring in adjacent corrugated molecular layers. Such amide NH-π hydrogen bonds are considered relatively weak; however, they are prevalent in chemistry and biology.56,57 As shown later, the relatively low strength of this hydrogen bond is compensated by the denser molecular packing and thus more favorable van der Waals interactions in comparison to the other forms. Indole Hydrogen Bond Propensities and NH-π Interaction. Somewhat surprisingly, the hydrogen bond potential of the stable polymorphic form is not completely satisfied in the traditional sense. Hydrogen bond donation of the amide hydrogen to an acceptor atom is one of the most prevalent in organic chemistry and biochemistry.58 In each of the other structures this group’s donating potential is satisfied. In each of the structures the indole’s amine proton donates to either the carbonyl oxygen or to water, in the hydrate structure.
Figure 5. Comparison of the similarity of molecular layers observed in RS-MA and R-MA Form 1, top left and right, respectively. Overlay of the R-enantiomers found at equivalent positions in the layers of the two structures, lower left, and overlay of the S- and R-enantiomers found at similar positions in the layers, lower right. The R- and Senantiomers from the RS-MA crystal structure are depicted in yellow and pink, respectively.
enantiomer from the structure of R-MA Form 1 shows similarity in side chain conformation, again with amide groups forming hydrogen bonded chains propagating along the fast growing axis of the crystal lattice. However, the difference in chirality necessitates nearly orthogonal indole groups relative to one another in the respective structures. The planes of the indoles are at +45° and −45° inclines with respect to the short axes of the lattices, making them appear superimposable when viewed along the short axis in the superlattice view. Such similarities in unit cell packing and metrics result in similarity of peak position and intensity in the PXRD patterns of RS-MA and R-MA Form 1. A table with the hydrogen bonding G
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
The relative hydrogen bond propensities of donating groups of MA were assessed using the logit hydrogen-bonding propensity model developed by Galek et al.59 A collection of structures were extracted from the Cambridge Structural Database (CSD) and surveyed,60 a model was constructed using those structures and applying logistic regression, and their functional groups were analyzed for their relative participation in hydrogen bonding. On the basis of the analysis, there was clear preference for hydrogen bond donation of the indole-NH over the amideNH group, though both are exceptional hydrogen bond donors as evident by their frequency of observation in the CSD.61 Densities of Polymorphs versus Racemate. Wallach’s rule states that racemic crystals tend to be denser than their respective chiral counterparts.62 Brock and Dunitz examined the validity of the rule using crystallographic data and have shown that on average it holds true for chiral species that do not interconvert in solution; however, the rule has many exceptions.63 In their study of the differences in density of 65 non-interconverting systems, the racemic crystal was on average 0.92% denser than the chiral crystalline counterpart. In the case of MA, both chiral crystalline polymorphs are exceptions to Wallach’s rule. S-MA Form 2 crystals are most exceptional, being 3.36% more dense, whereas R-MA Form 1 is only 0.88% denser than RS-MA. In the compilation of 65 chiral racemate pairs, only three of the exceptions had density differences larger than that of S-MA Form 2, its difference being 11.7σ from the mean. The significant density difference is attributed to the directionality and strength of hydrogen bonding in the RS-MA structure, at the expense of van der Waals forces and close packing. 13 C CPMAS Solid-State NMR Spectroscopy Analysis. In the example of the late-appearing polymorph of ritonavir, it was proposed that a highly disfavored conformation in its lattice resulted in its reluctance to nucleate and even grow, despite its substantially lower solubility. Considering the similarity of the hydrogen bonding networks and molecular conformations observed in RS-MA and S-MA Form 1 structures, as well as their more extensive hydrogen bonding relative to Form 2, it is tempting to conclude that the “folded back” conformation is the preferred conformation in solution leading to the preferential nucleation of polymorphic Form 1. Could a high energy conformer be responsible for the anticipated late appearance of the missing polymorph? Solid-state NMR, when combined with X-ray crystallography, is a powerful tool for such analysis. Comparisons of the 13C CPMAS spectra of the crystalline forms were made to establish conformational relationships with chemical shifts. A spectrum was also acquired for R-MA in its “glassy” state by melting R-MA Form 1 crystals directly in the rotor. Spectral assignments and comparisons were made relative to solution-state 13C chemical shift values of R-MA in deuterated chloroform. The 13C solid-state spectra are presented in Figure 6, and chemical shifts are tabulated in the Supporting Information as SI.T4. Cursory inspection of the NMR spectra indicates that the crystalline forms contain a single molecule in their asymmetric unit, consistent with single crystal analysis. The doublet observed at approximately 148 ppm is due to splitting of the C6 indole resonance by its coupling to the two naturally abundent isotopes of the quadrupolar chlorine 35Cl37Cl nuclei. Aromatic resonances were identified based on their proximity to their assigned solution-state values along with analysis of a spectrum acquired using the interrupted decoupling pulse sequence designed to suppress the methine and methylene
Figure 6. 13C CPMAS Solid-State NMR spectra of crystalline forms and that of the melt of R-MA in its “glassy-state”, R-MA monohydrate, S-MA Form 2, R-MA Form 1, and the RS-MA racemate (top to bottom).
carbon resonances relative to quaternary and methyl carbons through dipolar-dephasing.64 The signals of the side chain were assigned with little ambiguity, whereas some of the aromatic resonances were less definitive in their assignment. The net influence of hydrogen bonding on the amide carbon resonance results in deshielding, shifting its resonance to lower field. Hydrogen bonding to the amide carbon is apparently greatest for the racemate, which resonates at 174.4 ppm, accepting hydrogen bonds from the amide groups of two adjacent molecules with distances of 2.90 and 2.91 Å (N···O). Form 2 accepts a single strong hydrogen bond from the amine of the indole, having a distance of 2.84 Å and resonating at 172.8 ppm. The amide carbon of Form 1 resonates at 171.9 ppm, making weaker hydrogen bonds with amide groups of adjacent molecules at 2.94 Å. The water of the hydrate structure forms a single hydrogen bond with the carbonyl, resulting in a carbon resonance of 171.4 ppm. The carbonyl of the melted R-MA material resonates at 171.6 ppm, indicating that it likely has some degree of hydrogen bonding in the glassy state, albeit on average weaker than in the crystalline forms. The isotropic chemical shift of the amide carbonyl in the solution-state is upfield relative to the condensed states, resonating at 169.19 ppm, indicating a lack of hydrogen bonding in deuterated chloroform. Side Chain Conformation. The C11 methylene resonance is indicative of the relationship of the acetamide and indole groups with respect to one another about the C10−C11 bond, as described by the torsion N12−C11−C10−C2. In RS-MA and R-MA Form 1, these substituents are gauche with respect to one another, whereas in S-MA Form 2 and in the R-MA monohydrate form the groups are antiperiplanar. The torsion angles correlate well with the chemical shift values in the four crystalline forms as the chemical shift varies as a nearly linear function of the torsion angle. Variation of this angle is primarily responsible for the readily apparent conformational polymorphism, where the gauche relationship of Form 1 results in the acetamide group being “folded back” toward the indole moiety, and the anti relationship observed in Form 2 results in the acetamide side-chain being “folded away” from the indole in an extended conformation. The C11 chemical shift observed in the glassy-state of the melt is 46.2 ppm and falls between that of Form 2 and the hydrate, indicating that its side-chain conformation is also “folded away”, having the extended anti relationship. This resonance in the solution-state is observed even further downfield than that of Form 2 which is likely due to an extended antiperiplanar conformation. H
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
Table 4. Correlation of Torsion Angles with 13C Chemical Shift Values of the MA Forms torsion C15−O6−C6−C5 (°) C15 methoxy (ppm) N12−C11−C10−C2 (°) C11 methylene (ppm) C16−C10−C3−C2 (°) C16 methyl (ppm)
RS-MA
Form 1 −0.1 53.4 52.3 53.4 −35.5 20.0
Form 2
3.6 54.1 64.1 49.5 −23.2 19.4
hydrate
9.5 56.5 160.6 47.0 1.8 16.5
−5.4 55.5 173 45.8 −11.1 18.5
melt
solution
55.9
58.23
46.2
45.45
17.9
18.39
Form 2 as the lower energy conformer of those found in the crystal structures. It is evident that if anything, the solutionstate conformation would likely favor nucleation of the “missing” polymorphic form. As a result, it is apparent that overcoming a conformational energy barrier is not likely responsible for the difficulty in its crystallization, a reasoning used previously to explain the late appearance of the stable polymorph of ritonavir. Thermodynamic Stability of Hydrate versus Anhydrous Forms. Higuchi and Grant state that an anhydrous crystalline form of a substance is always more soluble in water than the corresponding hydrate form.66,67,1,68 Furthermore, it is commonly stated that in general lower temperatures favor solvate formation. While this notion is widely held as true, the basis for the empirical rule lacks an authentic, thermodynamicbased explanation. That said, exceptions are infrequently reported.69 The metastable polymorph R-MA Form 1 obeys the rule and is less stable, as it converts to the monohydrate form in water at room temperature. The rate of conversion is somewhat slow due to the compound’s relatively low water solubility. On the other hand, no conversion is observed when the stable polymorph of the inactive enantiomer, S-MA Form 2, is suspended in water. Solubility analysis was performed as a function of temperature in order to understand the relationship of the forms in water; see Figure 7. Clearly, Form 2 is the most
The chemical shift of the methyl carbon attached to the chiral center, C16, is most characteristic of the orientation of the chiral methyl group with respect to the plane of the indole moiety. The torsion angle defining the orientation of the methyl group with respect to the plane of the indole group, C16−C10−C3-C2, ranges from slightly synclinal, 35.5° and 20.0 ppm at one extreme in the structure of RS-MA to synperiplanar for S-MA Form 2 at 1.8° and 16.5 ppm. In Form 2, the methyl group comes into closer contact with the hydrogen at C2 of the indole, shifting the methyl resonance upfield relative to the other structures. The methyl group in RMA Form 1 resonates nearer to the resonance of this group in the racemate, at 19.4 ppm. This resonance is observed at 17.9 ppm in the glassy state of the melt, being nearer to that of Form 2 than Form 1. On this basis, with the melt and the solutionstate spectra having resonances at intermediate values of 17.9 and 18.4 ppm, respectively, it is likely that their mean torsion angles are slightly smaller than that of the hydrate form and are just under 10° on average. Table 4 provides a comparison of the chemical shifts as a function of the torsion angles. Methoxy Group Orientation. In each of the structures, the C15 methoxy group is approximately coplanar with the indole moiety and rotated away from the adjacent chlorine atom. Among the four crystal structures reported herein, there is good correlation of the planarity of the methoxy group with the aromatic ring. The methyl resonance is shifted farthest upfield to 53.4 ppm in the RS-MA form as the C15−O6−C6− C5 torsion tends toward −0.1°, whereas in Form 2 it is furthest out of plane, 9.5°, and resonates at 56.5 ppm. Table 4 provides the correlation of torsion angles and chemical shifts for the methoxy group. A plot of chemical shift versus torsion angle is nearly linear, and as a result it is reasonable to believe that the average orientation of the methoxy group in the condensedstate of the glass formed by melting is 6−7° out of plane, having a resonance slightly greater than that of R-MA monohydrate and less than S-MA Form 2. Extending this correlation into the solution-state requires extrapolation beyond the range of chemical shift versus torsion angle observed in the crystalline structures, as it appears to have an isotropic chemical shift value that is downfield slightly at 58.2 ppm. Its methoxy group is likely out of plane by as much as 13−14°. These values are well within the range found in a search of the CSD, where structures containing methoxy groups ortho to chlorine in aromatic rings were observed to be within 11° of the plane for 24 out of 26 structures, each having the methyl group directed away from the adjacent chlorine atom.65 In general, the 13C solid-state NMR data consistently indicate that the conformation observed in the melt of R-MA in its “glassy state” as well as in solution, is most similar to the conformation of the stable polymorph S-MA Form 2, the missing polymorph. These data are supported by ab initio calculations of conformational energies that establish the extended conformation found in the stable polymorph S-MA
Figure 7. Water solubility of MA versus temperature for crystalline forms of MA.
stable form of the crystalline forms of MA in water, even more stable than the monohydrate. Hence, S-MA Form 2 can now be added to the growing list of exceptions to the hydrate rule. Supporting Information includes a table with aqueous solubility data as a function of temperature SI.T5 and an associated van’t Hoff plot, SI.F5. Thermodynamic Relationship of Form 2, Form 1, and Racemic Form, Transition Temperatures, and Potential for “Spontaneous Resolution”. A thorough review of I
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
above the melting point of either form, reflected by a “virtual transition point” occurring at approximately 209 °C. The monohydrate form is less stable than polymorphic Form 2 in water at all temperatures studied. It is more stable than polymorphic Form 1 in water at temperatures lower than 65 °C, as confirmed by slurry studies at different temperatures, where it dehydrates to R-MA Form 1 at 70 °C and greater. If both enantiomers are present in solution in equal proportion, the racemic crystalline form, that is RS-MA, is thermodynamically more stable than all other forms at temperatures greater than 17 °C. Theoretically, a conglomerate mixture of S-MA and R-MA Form 2 crystals would be favored (as a conglomerate mixture) at lower temperatures, provided the “missing” R-MA Form 2 polymorph would nucleate. A phase diagram constructed in this manner can be very informative in the design of crystallization processes that target a specific crystalline form of a compound. In principle, spontaneous resolution of the racemic solution to form a conglomerate mixture of R-MA Form 2 and S-MA Form 2 at low temperature would create seed crystals for crystallization of the missing polymorph of the pure active enantiomer. Computational Conformational Analysis. The relative energies of the molecular conformations of MA after optimization starting from the available experimental configurations are provided in the Supporting Information as table SI.T7. The results using MP2 with different basis sets are fairly consistent indicating convergence of the method with respect to basis set. The DFT(6-31G*,BLYP) method predicts that the conformation consistent with that found in Form 2 is the most stable, while that found in the racemate crystal is the third most stable conformer. All QM methods predict the conformer found in the hydrate crystal to be the least stable. The molecular mechanics method predicts the conformer found in the Form 1 crystal to be the most stable and that in the Form 2 crystal to be the least stable. The most reliable of these calculations will be those using MP2(TZVP), which, along with MP2(SVP), predicts that the conformer optimized starting from the racemic molecular geometry is the most stable of the experimental conformations. Supporting Information SI.T7 tabulates the optimized torsion angles for each method and compares them with the initial experimental values from the crystal structures. The results reported in this table are for optimized gas phase molecules, so the comparison with the experimental values from the crystal structures is not strictly valid; however, it does indicate how consistent the methods are and whether a method has predicted a change in conformation. There are large changes in the torsion angles associated with all of the methods, which is not surprising given the comments above. The DFT method predicts an extremely large change in the T1 torsion angle of over 78° for the Form 1 conformer. In general, the two MP2 methods predict similar conformational geometries and similar relative energies. For this reason, the MP2(SVP) method was used to reoptimize the structures of the 61 lowest energy conformers predicted by the molecular mechanics conformational grid scan. Table 5 summarizes the results of the conformational analysis after an unrestrained grid scan using the Dreiding force field with Gasteiger charges followed by clustering of the resulting structures to remove duplicate structures and an optimization of each resulting conformer using the MP2(SVP) method. Full results are given in SI.T8 in the Supporting Information. It is clear from Table SI.T8 that the molecular mechanical description of the potential energy surface of the molecule is
spontaneous resolutions has been completed and its industrial importance discussed.70 It has been shown that only 5−10% of enantiomeric pairs crystallize as conglomerates, whereas the majority crystallize as racemates. Considering the greater density of Form 2 crystals and its significantly higher melting point and enthalpy of fusion relative to the racemic crystal form, one might expect that at some temperature a conglomerate mixture of Form 2 crystals might be favored over a racemic crystalline form. Collet et al. derived a method for comparing the Gibbs free energy of a conglomerate mixture (a 1:1 mixture of “mirror image” crystals composed of the individual enantiomers) to that of a racemic crystal (where the crystal lattice is composed of both enantiomers related by symmetry within the unit cell in a 1:1 ratio).52 To further extend the number of data points on the free energy versus temperature plots, melt eutectics were formed with organic additives, effectively extending the range over which the free energy differences of forms were measured to well below their melting points. The method, equations, and their derivation are well described by Yu et al.71 Thermal data are provided in the Supporting Information as Table SI.T6. The temperature range over which the stability differences between crystalline forms can be measured was further extended through addition of the points from solubility analysis. Solubility measures are commonly used to relate the stability and free energy difference of polymorphs and hydrates, assuming that the activity of the solid in equilibrium with the solution is proportional to its solubility. Less common, solubility measurements are made with crystalline enantiomers and compared to a racemic crystalline form as is done in this instance.72 The temperature dependence of the difference in Gibbs free energy of the crystalline forms of MA relative to that of the racemic crystalline form is shown in Figure 8.
Figure 8. Comparison of Gibbs free energy difference versus temperature for conglomerate mixtures of polymorphic and monohydrate crystalline forms of MA relative to the racemic crystalline form at atmospheric pressure.
The free energy-temperature diagram reveals information that can be used to effectively design crystallizations targeting specific crystalline forms under conditions of thermodynamic control. First, as predicted by the heat of fusion rule and verified through the G-T plots, Form 1 and Form 2 are monotropically related with Form 2 being the more stable polymorph. Regression lines intersect at a temperature well J
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
smallest of all four structures. The hydrate has the largest contribution from van der Waals and Coulomb interactions, which is consistent with its additional hydrogen bonds and it having both water and drug molecule present in its unit cell. The “other energy” contribution is much higher for the hydrate. This contribution reflects those interactions that have not been captured elsewhere, such as polarization and induction. Table 6 shows that it is difficult in general to point to one particular term which can be identified as the cause of a crystal’s stability. The final result is due to a very fine balance between competing forces. Nucleation of the Melatonin Agonist. Classical nucleation theory breaks crystallization down into nucleation and growth phases. Three factors govern the rate of nucleation: the temperature, the supersaturation, and the interfacial energy of aggregates growing in size forming an interface between the solute in solution and the solute in the crystalline state. A large part of the energy of cluster formation is a result of the reduced entropy of the molecules creating the solid surface. The more stable the form, the more stable the surface, and the greater the entropy cost in its formation. Theoretically, the more stable the form, the greater size the critical cluster must become before crystal growth is favored. As such, the probability of primary nucleation occurring decreases with increasing critical cluster size, leading to longer induction times and reduced nucleation rate. Like ritonavir, the energy difference between polymorphic forms of MA exists at the extreme for polymorphs, likely responsible for nucleation induction times that are prohibitively long in the absence of heterogeneous nucleation. Heterogeneous nucleation is thought to be more commonly operative than primary nucleation.17,73 Our explanation for the different crystallization behavior of the two enantiomers lies in their chirality and heterogeneous nucleation. Both enantiomers of MA share a common synthetic route until the next to final step where they are separated. Each material has been exposed to the same chemicals, experimental conditions, and laboratories. As a consequence, it is expected that S-MA has come into contact with a chiral substance or surface capable of successfully nucleating its Form 2 crystals. Thereafter, secondary nucleation and the large energy difference between forms resulted in disappearance of S-MA Form 1 crystals. The proposed chiral substance is not effective at nucleating R-MA Form 2 due to its opposing chirality; consequently, it remains in its metastable state.
Table 5. Relative Energies and Torsion Angles of the Most Important Conformations from the Dreiding Force Field Conformational Grid Scan after Reoptimization with the MP2(SVP) Methoda MP2 (SVP) rank
MM rank
ΔE (kcal·mol−1)
T1 (°)
T2 (°)
T3 (°)
corresponding experimental conformer
1 2 3 18
6 52 4 12
0.00 1.53 1.77 5.67
−84.5 −110.1 −79.8 −30.8
−45.3 −162.7 −48.8 −177.8
−72.8 81.3 91.3 −105.3
racemate Form 2 (S) Form 1 (R) hydrate (R)
a
Torsion angles as defined in Table 2.
much rougher than the MP2(SVP) surface. Many of the minima found with the Dreiding force field collapse to a single minimum on optimization with the QM method. When the MP2(SVP) optimized structures are clustered, the lowest energy conformer corresponds to the conformer found in the racemate crystal and the second lowest energy structure, 1.53 kcal·mol−1 above the global minimum, corresponds to that found in Form 2. Similarly, the conformer found in the Form 1 crystal is ranked third and that in the hydrate ranked 18th, 1.77 and 5.67 kcal·mol−1, respectively, above the global minimum. The rankings for the observed experimental conformers predicted by the Dreiding force field are also shown in Table 5 and are quite different, with Form 1, Form 2, the racemate, and the hydrate conformers being found at ranks 4, 52, 6, and 12, respectively. Calculated Lattice Energies. Two methods were used to calculate the lattice energies and the geometries of the conformations in the solid state, starting from the experimental crystal structures. These results are tabulated in Supporting Information, Table SI.T9, which also details the percentage deviations of the optimized structures from their starting, experimental structures. The Dreiding force field results show deviations in the optimized unit cell parameters and torsion angles as large as 13.9% and 18.5%, respectively. The DFT(D) method predicts the crystal structures and molecular geometries well, with maximum deviations in the cell parameters and the torsion angles of 1.1% and −3.0%, respectively. The DFT(D) predicted lattice energies of the anhydrous structures indicate an order of stability Form 2 > racemate > Form 1. The calculated energy difference between Forms 1 and 2 is 0.6 kcal/ mol, which is in good agreement with the experimental stability difference of 0.7 kcal/mol derived from solubility measurements. Note that the calculated lattice energies do not take into account entropies, thermal vibrations, or zero point energies. Table 6 shows a breakdown of the contributions to the DFT(D) lattice energies of the optimized crystal structures. Form 2, which is the most stable polymorph, has the most stable conformer according to the DFT(D) method. The van der Waals contribution for this polymorph is larger than for the other two nonhydrates, which is consistent with its higher density. However, the Coulomb stabilization of Form 2 is the
■
CONCLUSIONS The factors that govern formation of the very first nuclei of a crystalline form can be complex; a priori prediction of the existence of a given form does not constitute its discovery in practice. In the case of the melatonin agonist, knowledge of the existence of the stable polymorphic form through crystallization of its inactive enantiomer did not enable the crystallization of its active enantiomer in the most stable form. It was shown that the difference in the crystallization behavior of the two
Table 6. Breakdown of Contributions to the Total DFT(D) Lattice Energies in kcal·mol−1
Form 1 Form 2 racemate hydrate
conformational
configuration
van der Waals
Coulomb
other
lattice
2.20 0.00 0.05 2.94
3.73 3.37 2.50 3.29
−28.93 −32.15 −28.99 −36.55
−14.05 −12.46 −12.54 −20.22
−3.73 −0.12 −2.14 −9.20
−40.79 −41.36 −41.11 −59.74
K
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
Dreiding force field conformational grid scan after reoptimization with the MP2(SVP) method. SI.T8 provides the calculated lattice energies and percentage deviations of optimized crystal structures and molecular geometries from the initial, experimental crystal values. This information is available free of charge via the Internet at http://pubs.acs.org/.
enantiomers was not likely due to related substance levels alone, as the level of related substances and their identity are low and nearly identical. The reluctance of the missing polymorph to crystallize could not be attributed to the existence of a high energy conformer in its lattice that made assembly of the crystal difficult; NMR analysis and ab initio calculations indicate that the most stable conformer is the one present in greatest population in solution, in the “glassy” melt, and in the stable polymorphic Form 2. Instead, the successful nucleation of the stable polymorph of the inactive enantiomer is attributed to heterogeneous nucleation, likely involving a chiral structure. Comparison of the crystallization behavior of pure enantiomers to one another allowed for observations and comparison of crystal nucleation and energetic measurements that would otherwise have been more difficult to make. The melatonin agonist provides a polymorphic pair with an energy difference that borders on the extreme for any organic system studied to date. As such, it appears to break many conventional, empirical rules of crystallization; a monohydrate form has been found that is less stable in water than the stable anhydrous polymorph, both chiral polymorphs are denser than the racemic crystalline form, and the relatively rare occurrence of spontaneous resolution of a racemate into its conglomerate forms is predicted to occur at low temperature. This investigation demonstrates how the appearance of a crystalline form can completely revise the “apparent” crystal form landscape. Had the stable polymorphic form appeared first, the existence of the metastable form, the monohydrate form, and it is “dehydrate” form would likely have not been realized. While in reality the forms still exist in the crystal form landscape, their thermodynamic relationship with respect to the stable polymorph makes their realization through crystallization much more difficult, requiring conditions of kinetic control. In many ways the missing polymorph of MA contrasts with the polymorphic system of ROY, for which a myriad of colorful polymorphs were observed within a relatively small energy range (six within 0.3 kcal per mol). The authors believe that the evasiveness of the “missing polymorph” of R-MA, like most late appearing polymorphs, can be attributed to a large energy barrier for primary nucleation. In the absence of heterogeneous nucleation, the probability of successful formation of stable nuclei and their crystallization is unlikely or much delayed.
■
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS
■
REFERENCES
The authors thank Jeremy Hinds for acquiring SEM images, Jennifer Runyan for acquiring DSC data, David Jackson for acquisition of 13C solid-state NMR spectra, and Marcus Neumann for providing a courtesy license to the GRACE software package. One author, G.S., would like to express his gratitude to Lian Yu for the many interesting conversations over the years regarding this subject matter.
(1) Byrn, S. R.; Pfeiffer, R. R.; Stowell, J. G. Solid-State Chemistry of Drugs, 2nd ed.; West Lafayette: SSCI, Inc., 1999. (2) Nickelsen, C.; Samel, A.; Vejvoda, M.; Wenzel, J.; Smith, B.; Gerzer, R. Chronobiol. Int. 2002, 19 (5), 915−936. (3) Flaugh, M. E.; Bruns, R. F.; Shipley, L. A.; Clemens, J. A. Lilly Research Laboratories, Eli Lilly Company, Indianapolis, IN, USA. Pineal Update: From Molecular Mechanisms to Clinical Implications; Webb, S. M., Ed.; 7th Colloquium of the European Pineal Society, Sitges, Spain, March 29−31, 1996; pp 321−330. (4) Pudipeddi, M.; Serajuddin, A. T. M. J. Pharm. Sci. 2005, 94 (5), 929−939 Calculated from the average solubility ratio reported therein of 1.7; hence 0.3 kcal mol−1 = RT ln(1.7). (5) MA was never taken through Lilly’s current, comprehensive form screening. Had the molecule progressed, intensive efforts would have been applied until the stable form was isolated. (6) Chemburkar, S. R.; Bauer, J.; Deming, K.; Spiwek, H.; Patel, K.; Morris, J.; Henry, R.; Spanton, S.; Dziki, W.; Porter, W.; Quick, J.; Bauer, P.; Donaubauer, J.; Narayanan, B. A.; Soldani, M.; Riley, D.; McFarland, K. Org. Process Res. Dev. 2000, 4 (5), 413−417. (7) Bauer, J.; Spanton, S.; Henry, R.; Quick, J.; Dziki, W.; Porter, W.; Morris, J. Pharm. Res. 2001, 18 (6), 859−866. (8) Ostwald, W. Z. Phys. Chem. 1897, 22, 289−302. (9) Hedges, L. O.; Whitelam, S. J. Chem. Phys. 2011, 135, 1−6. (10) Cardew, P.; Davey, R.; Ruddick, A. J. Chem. Soc., Faraday Trans. 2 1984, 80, 659. (11) Threfall, T. Org. Process Res. Dev. 2003, 7, 1017−1027. (12) Van Santen, R. A. J. Phys. Chem. 1984, 88, 5768. (13) Keller, A.; Hikosaka, M.; Rastogi, S.; Toda, A.; Barham, P. J.; Goldbeck-Wood, G. J. Mater. Sci. 1994, 29, 2579−2604. (14) Nývlt, J. Cryst. Res. Technol. 1995, 30 (4), 443−449. (15) Chung, S. Y.; Kim, Y. M.; Kim, J. G.; Kim, Y. J. Nat. Phys. 2009, 5, 68−73. (16) Bernstein, J. Cryst. Growth Des. 2011, 11, 632−650. (17) Mullins, J. W. Crystallization, 4th ed.; Butterworth-Heinemann: Oxford, 2001. (18) Wolfrom, M. L.; Kohn, E. J. J. Am. Chem. Soc. 1942, 65, 1739. (19) Carson, J. F.; Waisbrot, S. W.; Jones, F. T. J. Am. Chem. Soc. 1943, 65, 1777−1778. (20) Bernstein, J. Polymorphism in Molecular Crystals; Oxford University Press: Oxford, UK, 2002. (21) Woodward, G. D.; McCrone, W. C. J. Appl. Crystallogr. 1975, 8, 342.
ASSOCIATED CONTENT
S Supporting Information *
Detailed crystallographic data have been submitted to the Cambridge Structural Database. A detailed method of synthesis is described and descriptions of a number of different crystallization attempts made to isolate the specific forms are provided. The DSC traces, SI.F2, and thermal data table, SI.T2, are provided. The hydrogen bonding descriptors and symmetry relations are provided in SI.T3. The results from the solubility measurements are tabulated in SI.T4. The data obtained from measurement of the melt eutectics are presented in SI.T5. The diffraction pattern of the dehydrated form 3 is provided in SI.F3. A melting temperature versus enantiomeric purity diagram is provided in SI.F4. Detailed results of the conformational analysis and the lattice energy calculations are provided. SI.T6 gives the relative energies of MA conformers and torsion angles after optimization of the experimental molecular structures. SI.T7 gives the relative energies and torsion angles of the 61 lowest energy conformations from the L
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Article
(55) Etter, M. C.; MacDonald, J. C.; Bernstein, J. Acta Crystallogr., Sect. B 1990, 46, 256−62. Etter, M. C. Acc. Chem. Res. 1990, 23, 120− 126. (56) Vaupel, S.; Brutschy, B.; Tarakeshwar, P; Kim, K. S. J. Am. Chem. Soc. 2006, 128, 5416−5426. (57) Gohlke, H.; Klebe, G. Angew. Chem., Int. Ed. 2002, 41, 2644− 2677. (58) Gavezzotti, A.; Filippini, G. J. Phys. Chem. 1994, 98, 4831−4837. (59) Galek, P. T.; Fabian, L.; Motherwell, W. D. S.; Allen, F. H.; Freeder, N. Acta Crystallogr., Sect. B 2007, 63, 768−782. (60) Cambridge Structural Database; Cambridge Crystallographic Data Centre, 12 Union Road Cambridge, CB2 1EZ, United Kingdom, Version 5.32, November 2010. (61) Umezawa, Y.; Tsuboyama, S.; Honda, K.; Uzawa, J.; Motohiro, N. Bull. Chem. Soc. Jpn. 1998, 71, 1207−1213. (62) Wallach, O. Liebigs Ann. Chem. 1895, 286, 90−143. (63) Brock, C. P.; Schweizer, W. B.; Dunitz, J. D. J. Am. Chem. Soc. 1991, 113, 9811−9820. (64) Opella, S. J.; Frey, M. H.; Cross, T. A. J. Am. Chem. Soc. 1979, 101, 3856−3857. (65) CSD search observed one structure REFCODE: FOYYAT with an angle of 16° and the second structure REFCODE: MOMAD was 69° out of plane. (66) Grant, D. J. W.; Higuchi, T. Solubility Behavior of Organic Compounds; New York: John Wiley and Sons: New York, 1990, p 38. (67) Khankari, R. K.; Grant, D. J. W. Thermochim. Acta 1995, 248, 61−79. (68) Shefter, E.; Higuchi, T. J. Pharm. Sci. 1963, 52, 781−790. (69) Gardner, G. L. J. Colloid Interface Sci. 1976, 54, 298−310. Allen, P. V.; Rahn, P. D.; Sarapu, A. C.; Vanderwielen, A. J. J. Pharm. Sci. 1978, 67, 1087−1093. Shibata, M; Kokubo, H; Morimoto, K; Morisaka, K; Ishida, T; Inoue, M. J. Pharm. Sci. 1983, 72, 1436− 1442. Katdare, A. V.; Bavitz, J. F. Drug Dev. Ind. Pharm. 1984, 10, 789−807. Wu, L. S.; Gerard, C.; Koval, C.; Rowe, S.; Hussain, M. A. J. Pharm. Biomed. Anal. 1994, 12, 1043−1046. Zhu, H.; Khankari, R. K.; Padden, B. E.; Munson, E. J.; Gleason, W. B.; Grant, D. J. W. J. Pharm. Sci. 1996, 85, 1026−1034. Zhu, H.; Padden, B. E.; Munson, E. J.; Grant, D. J. W. J. Pharm. Sci. 1997, 86, 418−429. Hu, T. C.; Wang, S. L.; Chen, T. F.; Lin, S. Y. J. Pharm. Sci. 2002, 91, 1351−1357. ReutzelEdens, S. M.; Kleeman, R. L.; Lewellen, P. L.; Borghese, A. L.; Antoine, L. J. J. Pharm. Sci. 2003, 92, 1196−2004. Petrova, R. I.; Peresypkin, A.; Mortko, C. J.; McKeown, A. E.; Lee, J.; Williams, J. M. J. Pharm. Sci. 2009, 98, 4111−4117. (70) Perez-Garcia, L.; Amabilino, D. B. Chem. Soc. Rev. 2007, 36, 941−967. (71) Yu, L.; Huang, J.; Jones, K. J. Phys. Chem. B. 2005, 109, 19915− 19922. (72) Huang, J.; Yu, L. J. Am. Chem. Soc. 2006, 128, 1873−1878. (73) Kashchiev, D. Nucleation, Basic Theory with Applications; Butterworth-Heinmann: Oxford, 2000; p 31.
(22) The prose was credited to comments made by C. P. Saylor to Woodward and McCrone which were then included in their Letter to the Editor, as stated therein in ref 21. (23) Dunitz, J. D.; Bernstein, J. Acc. Chem. Res. 1995, 28, 193−200. (24) Barsky, I.; Bernstein, J.; Stephens, P. W.; Stone, K. H.; Cheung, E.; Hickey, M. B.; Henck, J. O. Cryst. Growth Des. 2008, 8, 63−70. (25) Bernstein, J. Polymorphism and Patents. In Polymorphism in Molecular Crystals; Oxford University Press: New York, 2002; pp 297− 307. (26) Lancaster, R. W.; Harris, L. D.; Pearson, D. CrystEngComm 2011, 13, 1775−1777. (27) Li, H.; Stowell, J. G.; Borchardt, T. B.; Byrn, S. R. Cryst. Growth Des. 2006, 6, 2469−2474. (28) Yu, L.; Stephenson, G. A.; Mitchell, C. A.; Bunnell, C. A.; Snorek, S. V.; Bowyer, J. J.; Borchardt, T. B.; Stowell, J. G.; Byrn, S. R. J. Am. Chem. Soc. 2000, 122, 585−591. (29) Yu, L. Acc. Chem. Res. 2010, 43, 1257−1266. (30) Chen, S.; Xi, H.; Yu, L. J. Am. Chem. Soc. 2005, 127, 17439− 17444. (31) Yu, L. CrystEngComm 2007, 9, 847−851. (32) Bruker-AXS: Madison, Wisconsin, USA. (33) Bruker AXS: SAINT+, Release V7.68a; Bruker Analytical Systems: Madison, WI. (34) Sheldrick, G. M. SADABS, version 2.03, Program for Area Detector Absorption and Other Correction; University of Gottingen: Germany, 2001. (35) Sheldrick, G. M. SHELXS86. Acta Crystallogr., Sect. A 1990, 46, 467−473. (36) Sheldrick, G. M. SHELXS93. Program for Crystal Structure Refinement; Institute fur Anorg Chemie: Göttingen, Germany, 1997. (37) Materials Studio, 4.1; Accelrys Inc: San Diego, 2007. (38) Mayo, S. L.; Olafson, B. D.; Goddard, W. A. J. Phys. Chem. 1990, 94, 8897−8909. (39) Gasteiger, J.; Marsili, M. Tetrahedron Lett. 1978, 3181−3184. (40) (a) Becke, A. D. Phys. Rev. A 1988, 38, 3098−3100. (b) Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. Rev. B 1988, 37, 785−789. (41) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257. (42) Schaefer, A.; Horn, H.; Ahlrichs, R. J. Chem. Phys. 1992, 97, 2571. (43) Neese, F. ORCA − An ab Initio, Density Functional and Semiempirical Program Package, Version 2.6; University of Bonn: Bonn, Germany, 2008. (44) Weigend, F.; Ahlrichs, R. Phys. Chem. Chem. Phys. 2005, 7, 3297. (45) GRACE (the Generation, Ranking, and Characterization Engine) software package is a product of Avant-garde Materials Simulation SARL, 30 bis, rue du vieil Abreuvoir, F-78100 St-Germainen-Laye, France,
[email protected]. (46) Neumann, M. A.; Perrin, M.-A. J. Phys. Chem. B 2005, 109, 15531−15541. (47) (a) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558−561. (b) Kresse, G.; Furthmüller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169−11186. (48) Wang, Y.; Perdew, C P. Phys. Rev. B: Condens. Matter Mater. Phys. 1991, 43, 8911−8916. (49) The value is considered approximate but calculated using a whole-pattern method by least-squares fit of Form 1 and Form 2 reference patterns versus the aged sample pattern, resulting in a small residual difference, where Rwp < 0.10 after refinement. Integration of the enthalpy of fusion of the Form 1 endotherm indicated that the sample was slightly more than 80% Form 2. (50) Johari, G. P. Phys. Chem. Chem. Phys. 2000, 2, 1567−1577. (51) Roozeboom, H. W. B. Z. Phys. Chem., Stoechiom. Verwandtschaftsl. 1899, 28, 494−517. (52) Jacques, J.; Collet, A.; Wilen, S. H. Enantiomers, Racemates, and Resolutions; Krieger Publishing Company: Malabar, FL, 1991. (53) Simpson, H. J., Jr.; Marsh, R. E. Acta Crystallogr. 1966, 20, 550. (54) Pedone, C.; Benedetti, E. Acta Crystallogr. Sect. B 1972, 28, 1970−1971. M
dx.doi.org/10.1021/cg300398a | Cryst. Growth Des. XXXX, XXX, XXX−XXX