Article pubs.acs.org/JPCA
Terahertz Spectroscopy and Computational Investigation of the Flufenamic Acid/Nicotinamide Cocrystal Sean P. Delaney and Timothy M. Korter* Department of Chemistry, Syracuse University, 1-014 Center for Science and Technology, Syracuse, New York 13244-4100, United States S Supporting Information *
ABSTRACT: Terahertz spectroscopy probes the low-frequency vibrations that are sensitive to both the intermolecular and intramolecular interactions of molecules in the solid state. Thus, terahertz spectroscopy can be a useful tool in the investigation of crystalline pharmaceutical compounds, where slight changes in the packing arrangement can modify the overall effectiveness of a drug formulation. This is especially true for cases of polymorphic systems, hydrates/solvates, and cocrystals. In this work, the cocrystal of flufenamic acid with nicotinamide was investigated using terahertz spectroscopy and solid-state density functional theory. The solid-state simulations enable understanding of the low-frequency vibrations seen in the terahertz spectra, while also providing insight into the energetics involved in the formation of the cocrystal. The comparison of the cocrystal to the pure forms of the molecular components reveals that the cocrystal has better overall binding energy, driven by increased intermolecular hydrogen bond strength and greater London dispersion forces and that the trifluoromethyl torsional potential is significantly different between the studied solids.
1. INTRODUCTION Within the pharmaceutical industry, the solid-state assembly of molecules is important to the overall effectiveness of the drug product, where the preferred molecular solid must be achieved before marketing can begin.1−3 The investigation into the varying solid-state arrangements of pharmaceutical molecules is driven by the dependency of the physical properties on the solid-state structure. Polymorphism, the ability of a molecule to exist in two or more different solid-state forms, leads to a group of chemically identical molecules but with different physical properties based on changing intramolecular and intermolecular interactions. The altered physical properties, such as solubility, solid-state chemical/physical stability, and compressibility, can significantly impact the quality of pharmaceutical products. Rather than relying on polymorph discovery alone, attempts have been made to customize these physical properties through the addition of a solvate or a salt into the crystalline lattice. These additions change the overall structure of the molecular solid and have become very common methods for administering medicines (salts are estimated to be used in over half of the drug products currently on the market4). However, for a salt compound to be stable, the active pharmaceutical ingredient (API) must contain a suitably ionizable site. Therefore, other methods for obtaining solidstate structures with the desired physical properties have been investigated. Although salts and solvates have been part of drug development for many years, more recently, systematic approaches to the development of pharmaceutical cocrystals © 2015 American Chemical Society
have been investigated. A broad definition of a cocrystal is a crystal that contains two different molecules (mixed crystal).5 This can be further refined by saying that cocrystals are made from reactants that are solids at room temperature, which would exclude hydrates and other solvates, and clathrates or inclusion compounds. 5 Interest in cocrystals has been increasing, with potential applications in nonlinear optics,6 solvent-free organic synthesis,7,8 and host−guest chemistry.9−11 The use of cocrystals has also risen in the pharmaceutical community, where the addition of dissimilar molecules in the solid state can lead to unique physical properties in the same way that salt compounds can affect the physical properties. Flufenamic acid (FFA) is a nonsteroidal anti-inflammatory drug (NSAID), with analgesic properties suitable for long-term therapy,12 that can exist as nine unique polymorphs.13 Of the eight explicitly solved polymorphs of flufenamic acid, only two of them are used in the actual drug formulation (forms I and III).14 Flufenamic acid is an effective drug, but it has a low water solubility (0.009 09 g L−1) hindering its bioavailability.15 Recently, attempts have been made to cocrystallize FFA with other organic molecules, including nicotinamide, theophylline, 4,4′-bipyridine, and 2-pyridone, to increase the solubility of the drug.15 The pharmaceutical cocrystal focused on in this work is the flufenamic acid/nicotinamide (FFA/NCTA) system (Figure 1). Nicotinamide (NCTA) is particularly suitable Received: December 16, 2014 Revised: March 13, 2015 Published: March 18, 2015 3269
DOI: 10.1021/jp5125519 J. Phys. Chem. A 2015, 119, 3269−3276
Article
The Journal of Physical Chemistry A
Figure 2. Molecular structure and unit cell molecular packing of the pure NCTA crystal with atomic labeling scheme.
ization of pharmaceuticals.22−26 Relevant to the current work, both polymorphs of pure FFA have been previously investigated with this technique.27 In this study, terahertz spectroscopy is used to investigate the sub-90 cm−1 vibrations of the FFA/NCTA cocrystal and the pure NCTA, both are then compared to the pure FFA data previously published.27 In addition to aiding pharmaceutical detection and identification, the low-frequency terahertz spectra also provide an excellent method for validating the computational techniques being utilized in the investigation of the structure and energetics of these crystalline solids. Although terahertz investigations can yield new quantitative insight into the solid-state crystal packing changes in, the physical understanding of the low-frequency vibrations found in the terahertz region is more difficult to achieve. Solid-state density functional theory (DFT) incorporating periodic boundary conditions is able to predict the intramolecular and intermolecular vibrational motions that give rise to the observed terahertz spectra,28−30 enabling the unambiguous assignment of each experimental feature. The correlation of the simulated and experimental spectra provides validation of the computational model and results in a more accurate accounting of the relative energies of polymorphs to be reached, including the contributions of the conformational energy and individual intermolecular forces. In FFA/NCTA, it is noteworthy that the molecular conformation of FFA is distinctly different from the pure FFA polymorphs, a fact which has a direct impact on the magnitudes of the intermolecular forces present in the cocrystal.
Figure 1. Asymmetric unit and unit cell molecular packing of the FFA/NCTA cocrystal with atomic labeling scheme.
because of its high water solubility (500 g L−1)16 and the likelihood of cocrystal formation due to strong hydrogen bonding between the carboxylic acid groups in FFA and the nitrogen atom in the heteroaromatic ring of NCTA.17 The presence of the NCTA molecule (Figure 2, pure NCTA) in the unit cell affects the solid-state positioning and the molecular conformation of the flufenamic acid (compared to pure FFA). The variation in the solid-state arrangement allows for the FFA molecules to exhibit conformational disorder (torsion of the −CF3 group, with a population ratio of 0.91:0.09) that is not present in pure FFA. Such disorder can be problematic for pharmaceuticals because it may result in crystals having different physical properties (especially relative stability) based on the ratio of disordered conformations present.3,18−21 Therefore, the FFA/NCTA cocrystal represents a challenging species that exists as a cocrystal with a complex hydrogen bonding network, yielding favorable physical properties, but simultaneously exhibits undesirable disorder within the solidstate structure. Information about the intermolecular hydrogen bonds that drive the cocrystal formation and the potential energy surface governing the trifluoromethyl disorder in the solid can be gathered by direct investigation of the low-frequency vibrational motions in the samples. Changes in solid-state packing results in variations of the low-frequency lattice vibrations exhibited by the molecules, such as external rotations and translations or internal torsions. Terahertz (far-infrared) spectroscopy, which measures the intermolecular and intramolecular vibrations of molecular solids in the sub-200 cm−1 range, directly probes these vibrations, making it an ideal method for the character-
2. METHODS 2.1. Experimental Section. Flufenamic acid and nicotinamide were purchased from Sigma-Aldrich (product numbers: F9005 and 72340, respectively). The pure 1:1 cocrystals with FFA and NCTA were produced using the method described by Fabian et al. in which the two substances were dissolved in ethanol and mixed together (1:1 molar solutions) using a 3270
DOI: 10.1021/jp5125519 J. Phys. Chem. A 2015, 119, 3269−3276
Article
The Journal of Physical Chemistry A syringe filter.17 The formation of the FFA/NCTA cocrystal was verified using powder X-ray diffraction methods and comparison with the known structure (Supporting Information Figure S1). The crystal structure of nicotinamide was re-evaluated by single crystal X-ray diffraction, and the 90 K data used in this work can be found in the Supporting Information. The relevant experimental data for pure FFA has been reported previously.27 For terahertz sample preparation, each crystalline material was mixed with a powdered polytetrafluoroethylene (PTFE) matrix (Spurlock Specialty Tools, ∼3 μm particles) at a concentration that was dependent on the substance. The cocrystal used 4.3% and pure NCTA used 1.6% by mass to ensure that measurements were made within an optimal absorption range. The mixtures were then pulverized using a stainlesssteel grinder/mixer (Dentsply Rinn 3110-3A) to minimize particle size and reduce both Mie scattering and crystal anisotropy.31 Approximately 0.55 g (total) of the sample mixtures were pressed, at 2000 psi, into pellets with diameters of 13 mm and thicknesses of 2.0 mm. Pure PTFE was pressed into a pellet of equal mass to be used as a blank reference. The experimental spectra were obtained using a time-domain pulsed terahertz spectrometer based on an amplified Ti:sapphire femtosecond laser system. Zinc telluride crystals were used for both generation of terahertz radiation by optical rectification32 and detection by free-space electro-optic sampling.33 A detailed description of the terahertz spectrometer has been reported elsewhere.34 The samples and blank for measurement were held under vacuum in a cryostat with data acquired at both 293 and 78 K. Both were scanned 32 times for each individual data set over a time window of 32 ps (chosen to exclude all pulse reflections) consisting of 3200 data points, which was then symmetrically zero-padded to a total of 10 000 data points. The ratio of the Fourier-transformed data sets of the sample and blank resulted in a terahertz spectrum over the range 10−90 cm−1 with a spectral resolution of approximately 1.0 cm−1. Each data set was replicated four times at both temperatures and then averaged to obtain the final spectra reported here. Spectral intensities are reported in units of ε (M−1 cm−1) where molarity is expressed in terms of the concentration of crystallographic unit cells (Z) rather than individual molecules. 2.2. Theoretical Details. The solid-state simulations in this work were performed using the CRYSTAL0935 software package utilizing the PBE36 density functional in combination with the atom-centered Gaussian-type 6-311G(2d,2p)37 basis set. The total energy convergence criteria were ΔE < 10−8 hartree for geometry optimizations and ΔE < 10−11 hartree for frequency calculations. All structural optimizations were performed without limits on atomic positions or unit cell dimensions, other than those imposed by space group symmetry, and began using starting structures obtained by experimental X-ray diffraction measurements. The pure FFA I, FFA III, and NCTA structures were measured for this work at 90 K using single crystal X-ray diffraction. Because of the difficulty in obtaining large single crystals of the FFA/NCTA cocrystal suitable for X-ray diffraction, the previously reported X-ray structure (at 100 K) was used as the starting point for its simulation.17 The radial and angular distributions for DFT integration were defined by a pruned (75, 974) grid. Truncation tolerances for Coulomb and HF exchange integrals were defined as 10−8, 10−8, 10−8, 10−8, 10−16 hartree (TOLINTEG command35,38). A shrinking factor of 5 (27 k points in the irreducible Brillouin zone) was determined after
sampling and monitoring of the total energy convergence as a function of k-point count in reciprocal space according to the Pack−Monkhorst method.39 Normal-mode frequencies and infrared intensities were then calculated for the optimized structures. The frequency of each normal mode was calculated within the harmonic approximation by numerical differentiation of the analytical gradient of the potential energy with respect to atomic position.38 The infrared intensities for each normal mode were calculated from the dipole moment derivatives (dμ/ dQ) determined using the Berry phase technique of calculating Born charges as polarization differences between equilibrium and distorted geometries.35,40 Mode descriptions and assignments were made by visual inspection of the eigenvector atomic displacements for each normal mode. Typical DFT functionals do not account for the weak noncovalent interaction found in molecular solids, which can be a significant factor in achieving valid simulations of crystalline structure and dynamics.41,42 Therefore, the solid-state DFT approach used in this study was supplemented with corrections for London-type dispersion forces using a semiempirical method proposed by Grimme,43 and then later modified for use in the solid state by Civalleri et al.41 Through a comparison of the calculated unit cell parameters (a, b, c, α, β, γ, and volume) and the experimental cryogenic X-ray data taken for both the cocrystal, pure FFA, and pure NCTA, a global scaling factor (s6) of 0.50 was determined and employed for all of the solids. Only comparisons to low-temperature structures were made due to the sensitivity of the scaling factor to experimental temperature.42 Basis set superposition error (BSSE), an error imparted to the final energies by finite basis sets, must be estimated and removed from the calculated energies. The counterpoise method,44 utilizing a spatial cutoff, was chosen for this purpose. In the BSSE calculation, a single molecule was extracted from the already optimized solid-state unit cell and evaluated using the same theoretical method (PBE/6-311G(2d,2p)) as for the periodic calculations. It was found, through monitoring of the BSSE convergence, that 250 atoms within 5.0 Å of the molecule being evaluated were sufficient spatial cutoff limits, and that level was used in all calculations to gauge BSSE effects. Previous studies have found that using this approach reveals the majority of the BSSE energy.45,46
3. RESULTS AND DISCUSSION 3.1. Terahertz Spectroscopy. The 78 and 293 K terahertz spectra from 10 to 90 cm−1 of the FFA/NCTA cocrystal and the pure NCTA sample are shown in Figure 3, and the observed vibrational frequencies are listed in Table 1. The two materials have unique and identifying terahertz spectra, with the cocrystal having more spectral features because of the inclusion of two different molecules in the solid-state asymmetric unit, resulting in lowered space group symmetry with more IR-active normal modes of vibration. The low-temperature spectra exhibit sharpened features compared to the room-temperature data due to the decrease in the populated vibrational states of the molecules in the solid state, making the 78 K spectra essential for the proper assignment of the vibrational modes. All of the 78 K spectral features for the FFA/NCTA cocrystal exhibit a slight shift to higher energy versus their 293 K positions, as is commonly observed in the terahertz spectroscopy of cooled solids.22,34,47 In contrast, one of the 78 K spectral features observed in NCTA (19.43 cm−1) appears at a lower energy than found at room temperature. This unusual 3271
DOI: 10.1021/jp5125519 J. Phys. Chem. A 2015, 119, 3269−3276
Article
The Journal of Physical Chemistry A
hydrogen atoms. The quality of the unit cell parameters (a, b, c, and volume) was also evaluated and the results are listed in Table 3. Analysis of the structural reproduction of FFA/NCTA Table 3. Comparison of the 90 K Experimental Unit Cell Dimensions and the Simulated Unit Cell Dimensions of the FFA/NCTA Cocrystal and Pure NCTAa NCTA a (Å) b (Å) c (Å) α (deg) β (deg) γ (deg) V (Å3) % error in unit cell parameters
Table 1. Observed Terahertz Vibrational Frequencies (cm−1) of the FFA/NCTA Cocrystal and Pure NCTA at 293 and 78 K and the Approximate Correlations between the Two Temperatures NCTA 293 K
78 K
19.99 35.54 57.75 64.97 79.97 92.74
19.43 36.09 60.53 69.41 86.07 96.63
38.32 46.64
40.54 49.42 54.42 59.98 70.53 81.08 86.08 91.63 98.85
55.53
phenomenon has been observed before in the terahertz spectroscopy of sucrose48 and gabapentin,46 but it is rare. To better understand the nature of the spectral features and to study the origins of the energetic differences between the cocrystal and its pure components, solid-state DFT has been employed in the spectral analyses. 3.2. Solid-State DFT Simulations. 3.2.1. Structural Reproduction. Structural simulations were based on full optimizations using PBE/6-311G(2d,2p) without limits on the unit cell dimensions. Root-mean-squared-deviations (RMSDs) of bond lengths, bond angles, and dihedral angles (Table 2) were used to determine the quality of the reproduced structures. The RMSD values represented here exclude all Table 2. Evaluation of the Structural Reproduction Using PBE/6-311G(2d,2p) by Root Mean Squared Deviations (RMSDs) of Both the Cocrystal and Pure NCTA, Excluding Hydrogensa NCTA FFA/NCTA
bonds
angles
dihedrals
0.0104 0.0121
0.272 0.837
1.067 1.325
theoretical
experimental
theoretical
3.88 15.60 9.38 90.00 98.45 90.00 560.86
4.02 15.68 9.23 90.00 97.28 90.00 577.20 2.15
5.11 15.96 22.12 90.00 90.47 90.00 1802.36
5.09 15.89 22.43 90.00 91.59 90.00 1814.41 0.69
is complicated by the presence of disorder in the −CF3 substituent that is reported as 0.91:0.09 populations of two conformations.17 The 0.91 populated state of the FFA/NCTA cocrystal readily optimized with low RMSD values, and unit cell parameters differing by only 0.69% on average from experiment. The 0.09 populated cocrystal, however, did not optimize well, with the −CF3 group immediately relaxing into the higher populated 0.91 cocrystal conformation. This finding prompted a more thorough theoretical investigation into the disorder (vide inf ra) that suggests the disorder is not physically real, but likely is a result of the solving approach used in the analysis of the single crystal X-ray diffraction experimental data.17 The pure NCTA crystal also optimized well, with similar but slightly better RMSD values as compared to the 0.91 FFA/ NCTA cocrystal for bond lengths, bond angles, and torsion angles. Though it was found that the internal molecular structure of NCTA was well calculated, the external crystal packing reproduction was less successful. The optimized unit cell parameters for NCTA showed an average error that was increased by a factor of 3 (2.15%) as compared to the cocrystal results (0.69%). The relatively large error in the unit cell parameters stems from the specific error in one axis, the a-axis, which changes from 3.88 to 4.02 Å, a 3.6% increase. The impact of this deviation in the a-axis is discussed in the analysis of the terahertz spectral features to follow. The average unit cell errors in pure FFA I and FFA III are 0.83% and 0.45%, respectively. Further details on the optimized structures of the pure FFA can be found elsewhere.27 3.2.2. Vibrational Analyses. The simulated terahertz spectra of the FFA/NCTA and NCTA are displayed in Figure 4. The calculated vibrational modes (frequencies and intensities for modes
