Physical, Crystallographic, and Spectroscopic Characterization of a

Sep 13, 2006 - Frederick G. Vogt,*,† Jeffrey Brum,† Lee M. Katrincic,† Agnes Flach,† Jerome M. Socha,†. Richard M. Goodman,† and R. Curtis...
0 downloads 0 Views 570KB Size
Physical, Crystallographic, and Spectroscopic Characterization of a Crystalline Pharmaceutical Hydrate: Understanding the Role of Water Frederick G. Vogt,*,† Jeffrey Brum,† Lee M. Katrincic,† Agnes Flach,† Jerome M. Socha,† Richard M. Goodman,† and R. Curtis Haltiwanger‡

CRYSTAL GROWTH & DESIGN 2006 VOL. 6, NO. 10 2333-2354

Chemical and Pharmaceutical DeVelopment, GlaxoSmithKline plc., P.O. Box 1539, King of Prussia, PennsylVania 19406, and Cephalon, Inc., 145 Brandywine Parkway, West Chester, PennsylVania 19380-4245 ReceiVed June 1, 2006; ReVised Manuscript ReceiVed August 1, 2006

ABSTRACT: The single-crystal X-ray diffraction structure of the sodium salt of N-(3-(aminosulfonyl)-4-chloro-2-hydroxyphenyl)N′-(2,3-dichlorophenyl) urea at 173 K is reported. The structure contains 3 mol of water situated in distinct channels in the vicinity of the sodium cation. Powders of this phase undergo isomorphic dehydration, losing 0.5% w/w water between 90 and 15% relative humidity (RH) at 25 °C without changing the powder X-ray pattern. Below 15% RH and above 50 °C, additional dehydration occurs in conjunction with a reversible phase transition. A third semicrystalline dehydrated phase appears after vacuum-drying and at high temperatures and also can be reversibly rehydrated to the original form. Single crystals of the dehydrated phases could not be prepared, so a combination of methods were used to understand the structural changes occurring during the desolvation process, including thermal analysis, vapor sorption measurements, variable-humidity and variable-temperature powder X-ray diffraction, vibrational spectroscopy, and 1H, 13C, 15N, and 23Na magic-angle spinning (MAS) solid-state NMR. The uptake of water vapor into the trihydrate form was investigated by NMR and vibrational spectroscopy using isotopically labeled water. Static 2H and 17O NMR quadrupolar line shape analysis combined with changes in MAS spectra showed exchanged sites on the parent and water molecules. The results indicate that two moles of ion-associated water in the larger tunnel are more labile than a hydrogen-bonded mole of water. Entire water molecules can exchange into the lattice to a small extent, but more efficient hydrogen transfer exchange is observed in the main channel and a smaller perpendicular side channel. The exchanged water deuterons execute rapid three-site jump motions at 273 K. Introduction Organic molecular solids often incorporate water into their structure to form hydrates, which can exist in polymorphic forms with different molecular conformations, packing motifs, supramolecular interactions, and hydration states.1-3 In pharmaceutical development, dehydration of crystalline hydrates is a special concern because of potential conversion to metastable or amorphous phases with greatly reduced chemical and physical stability.2,4 Dehydrated forms fall into two categories: those that exhibit significant phase changes upon desolvation (such as a polymorphic transformation or conversion from a crystalline phase to an amorphous phase) and those that do not show such changes. The latter category requires additional clarification; although many molecular hydrates show a well-defined stoichiometric relationship between the water and the parent molecule, the water held in certain hydrates can be removed by humidity and temperature changes in a nonstoichiometric fashion without the occurrence of distinguishable phase changes, with the dehydrated structure being isomorphic with the hydrated structure.5-17 The parent crystal structure that produces such an isomorphic desolvate is often called a channel hydrate, a term that derives from visualization of its crystal structure, where water molecules are closely associated in sub-nanometer hydrophilic tunnels or channels. (If the water is not in a channel, the structure is often called an isolated-site hydrate.) In some channel hydrates, the water tunnels fill up to reach a defined stoichiometry; in others the lattice expands continuously over * Corresponding author. Phone: (610) 270-6985. Fax: (610) 270-6727. E-mail: [email protected]. † GlaxoSmithKline plc. ‡ Cephalon, Inc.

a wide range to accommodate additional water, although the structure is still considered isomorphic.3 Many channel hydrates change water content in an isomorphic fashion over a limited temperature and relative humidity (RH) range; outside these ranges they collapse to a different phase or show a polymorphic change to a higher hydrate. In the case of salt hydrates, channel water can also play an important bridging role between a small charged counterion and the ionized parent molecule. A number of pharmaceutical and nutritional compounds are being developed or are currently marketed as hydrates, and many are known to form isomorphic desolvates within certain humidity and temperature ranges. These include hydrates of erythromycin, cephalexin, ampicillin, cromolyn sodium (disodium cromoglycate), caffeine, theophylline, thiamine hydrochloride (vitamin B1), and commercially significant compounds such as sildenafil citrate.5-17 Since the hydration state can alter the final in vivo performance of drug products, its control is of importance in pharmaceutical development. Studies must be conducted to determine special requirements during crystallization, processing, tableting, and formulation, and to determine “composition of matter” for weighing factors. The solubility of drugs can be affected by their water content, as seen in a recent study that implicated the isomorphic dehydrate of erythromycin in tablet dissolution failures.18 Isomorphic desolvates contain empty space in their crystal lattice that can result in increased mobility of functional groups along the tunnel walls and reduced chemical or physical stability.12 Even more problematic is the ability of certain isomorphic desolvates to adsorb other molecules, such as organic solvents or molecular oxygen.19 Although labile solvent-filled voids cause concern in pharmaceutical development, there are many other interesting chemical applications for clathates20 and “nanovoids”; a notable example

10.1021/cg060324k CCC: $33.50 © 2006 American Chemical Society Published on Web 09/13/2006

2334 Crystal Growth & Design, Vol. 6, No. 10, 2006 Scheme 1

Vogt et al.

methods of wide-line static 2H and 17O NMR also have potential applications.15,24,25 In the present study, vapor sorption experiments with 2H and 17O-labeled water are used to track the environment, mobility, and path of water in the structure. The overall goal of this work is to characterize the dehydration process and structural role of water, in relation to the fully hydrated crystal structure, to understand the system at hand while helping to develop more effective models for desolvation. Experimental and Computational Procedures

is a crystalline zinc phthalocyanine complex engineered to contain large nanoporous cavities of 8 nm3, which can accommodate tens of water molecules per unit cell and has potential applications in catalysis and separation sciences.21 The complex physical properties of hydrates can be better understood from knowledge of their solid-state structure. Most work on hydrates to date has concentrated on either structural or thermodynamic aspects; in the structural studies, the behavior of channel hydrates is explained by the observation of channels in the solvated crystal structure. In the case of organic molecular hydrates, little has been done to understand desolvation processes, metastable “collapsed” states, or the dynamics of water inside the structure. The latter is especially significant since water almost always engages in hydrogen bonds, which can determine the ease with which it is first mobilized prior to dehydration. Once the hydrogen bonds are broken, the nanoscopic tunnel dimensions (cross-sectional area and tortuosity) determine dehydration behavior.2,5 One-dimensional tunnels along a particular cell axis are typical, although two-dimensional tunnels are also found.16 Entire families of structurally related compounds are known to exhibit variable hydration states caused by water channels in their crystal structures.5,17 After dehydration, the crystal may collapse to another phase or retain its structure; if collapse occurs, it may be reversible. A combination of physical techniques is thus needed to obtain a complete picture of solid-state structure and properties in complex hydrates. The present work applies complementary techniques to understand hydration in a developmental drug. The compound of interest is sodium N-(3-(aminosulfonyl)-4-chloro-2-hydroxyphenyl)-N′-(2,3-dichlorophenyl) urea trihydrate (I), a CXCR2 antagonist under development as an inhibitor of IL8 binding for the treatment of inflammatory diseases including rheumatoid arthritis.22 A hydrate of the sodium salt of I was found to be stable and easily manufactured and was selected for development. The chemical structure and atomic numbering scheme of I is illustrated in Scheme 1. The crystal structure of a trihydrate form of I was determined by single-crystal X-ray diffraction (SCXRD). This form behaves like a channel hydrate prior to collapsing in a complex but reversible dehydration process. Thermal and vapor sorption methods are combined with variable temperature (VT) and variable humidity (VH) powder X-ray diffraction (PXRD) studies to characterize channel hydrate behavior and phase changes during dehydration. 13C solid-state NMR (SSNMR), although sparingly used in the study of channel hydrates, can characterize the structure and its localized response to water content changes from the viewpoint of individual nuclei.9,12-15 The use of SSNMR with multiple nuclei and the latest generation of pulse sequences, beyond the typical crosspolarization magic-angle spinning (CP-MAS) experiment, gains access to even more information about the solvation environment.23 Faster spinning rates have opened up new possibilities for the use of 1H NMR in these systems and the established

Sample Preparation. Compound I was synthetically prepared using previously reported procedures.22 The crystallization of the sodium salt produced material with a water content close to a trihydrate. Samples of I were prepared at different hydration states by exposure of the crystalline sodium salt to different RH environments from saturated salt solutions.26 Each sample was stored in a preequilibrated, sealed 2 L glass chamber with 200 mL of salt solution at room temperature (22 °C) for six weeks, after which the water levels were found to have reached a steady state by Karl Fischer titration (see below). The samples were thinly spread inside glass vials. Saturated solutions of LiBr (6.5% RH), LiCl (11% RH), MgCl2 (33% RH), and NaCl (75% RH) were used in the humidity chambers.26 All salts were purchased from SigmaAldrich (St. Louis, MO) and were of 98% or greater purity. The humidity in each chamber was monitored to an accuracy of (2% RH with TM320 digital relative humidity recorders from Dickson Instruments (Addison, IL). No chamber was found to deviate from its expected humidity by more than the accuracy of the monitor. An additional nonequilibrium drying chamber containing calcium silicatebased desiccant was also used. This chamber had an observed humidity of 1-3% RH over the time period of use. Finally, samples were also dried in an oven at 60 °C under a vacuum (25 in. Hg) for 2 to 16 h to prepare nonequilibrium low water content samples. Two sets of equilibrium deuterium vapor exchange samples were prepared. The first set was made by placing a portion of the equilibrated material stored at 33% RH into a sealed 2 L chamber containing 200 mL of D2O (Isotec, Inc., USA) saturated with NaCl with a nominal humidity of 75% RH. A second set of samples was prepared in a similar manner starting from material equilibrated in the lithium bromide chamber (6.5% RH). Both sample sets were stored in the 75% RH D2O chamber for six weeks and after removal were checked by PXRD. A D2O-recrystallized sample was also prepared by evaporation of a saturated D2O/CH3CN solution. The material was quickly vacuum filtered (to avoid back-exchange from the atmosphere) and stored in a 33% RH D2O chamber for six weeks. For 17O-water exchange experiments, a similar procedure was followed, except that only ∼50 mg of the sample originally stored at 6.5% RH was placed in a 60 mL sealed glass container containing 6 mL of 40% isotopically enriched H217O (Cambridge Isotope Laboratories, USA) saturated with NaCl. The sample was stored in this miniature H217O humidity chamber for six weeks. This chamber was assumed to be near to 75% RH but was too small to allow for humidity monitoring. Single-Crystal X-ray Diffraction. Single crystals of compound I were grown by slow evaporation from a mixture of acetone, dichloromethane, and water. Microscopic examination of the crystals showed that they were beige-colored needles. The crystals were stored at ambient humidity and were later determined to be in the trihydrate state. Efforts to dehydrate single crystals were unsuccessful, with all attempts resulting in the cracking of the crystals as observed by optical microscopy. SCXRD measurements were made with an Enraf Nonius CAD-4 diffractometer using Mo KR (0.71073 Å) radiation from a finefocus sealed tube. Lattice parameters were determined from the setting angles of 25 reflections distributed in reciprocal space. Additional details of the data collection, structure solution, and refinement are summarized in Table 1, and more details can be found in the crystallographic information file (CIF) deposited in the Cambridge library.27 Three orientation controls were monitored for crystal movement during the experiment. Data were corrected for Lorentz and polarization effects, for variation in check reflections, and for absorption with a semiempirical correction.28 The SHELXS-97 and SHELXL-97 programs were used for direct solution and refinement.29 The cell parameters from the single crystal study were refined against the room-temperature powder XRD pattern using the Rietveld module in the HighScore Plus software package (PANalytical B. V., Almelo, The Netherlands).

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2335

Table 1. Summary of SCXRD Experimental Parameters and Results for Form 1a formula formula weight temperature space group crystal system unit cell dimensions a b c a b g volume molecules per cell (Z) calculated density absorption coeff (µ) F000 index ranges measured reflections independent reflections completeness to theta variation in check reflections data/restraints/parameters goodness of fit on F2 final R indices for I > 2σ(I) data final R indices for all data largest peak diff./hole a

Na+[C13H9Cl3N3O4S]-‚3H2O 486.6 173(2) K P21/c monoclinic 20.110(8) Å 6.863(3) Å 13.555(3) Å 90° 91.04(3)° 90° 1870.5(12) Å3 4 1.728 mg m-3 0.669 mm-1 992 0 e h e 22 -7 e k e 7 -14 e l e 14 5306 2607 99.9% 6.80% 2607/10/273 0.871 R1 ) 0.0372 wR2 ) 0.1029 R1 ) 0.0729 wR2 ) 0.1352 0.327 and -0.509 e Å-3

The trihydrate state of compound I.

Vapor Sorption Measurements, Thermal Analysis, and Karl Fischer Titrations. Gravimetric vapor sorption (GVS) experiments were performed on an instrument from Surface Measurement Systems, Ltd. Samples were studied over a humidity range of 0 to 90% RH at 25 °C. Each humidity step was made if less than 0.0025% weight change occurred over 10 min, with a maximum hold time of 3 h. Differential scanning calorimetry (DSC) was carried out on a TA Instruments Q1000 system. A N2 flow rate of 50 mL/min was used. Sample sizes ranged from 0.5 to 3.0 mg, and the heating rate was 10 °C/min. Samples were run in aluminum pans, and to minimize the effects of rehydration prior to the experiment, the DSC sample was stored in its aluminum pan in the RH chamber for the six week equilibration period. As soon as the chamber was opened, a lid was pressed on the pan and the sample was analyzed. Thermogravimetric analysis (TGA) was also performed on the samples using a TA Instruments Q500 system, again using a 10 °C/min heating rate. TGA sample sizes were in the range of 5-20 mg. Coulometric Karl Fischer (KF) titrations were performed at room temperature using a Metrohm 756 KF system and Hydranal Coulmat reagent (Reidel-deHaan). The analysis was started within a minute of opening each vial. The system was checked before and after the analysis using a Hydranal 0.101% suitability test (traceable to NIST SRM 2890). All reported results were the average of two titrations. Powder X-ray Diffraction. Ambient PXRD patterns were obtained using a Phillips X’Pert Pro diffractometer equipped with an X’Celerator Real Time Multi-Strip (RTMS) detector. Samples were flattened onto a zero-background silicon holder and run immediately after preparation. A continuous 2θ scan range of 2° to 40° was used with a Cu KR (1.5418 Å) radiation source and a generator power of 40 kV and 40 mA. A step size of 0.0167 deg per 2θ step was used, and individual patterns required 5 min to obtain. Samples on the X’Pert Pro system were rotated at 25 rpm. Variable humidity (VH) and variable temperature (VT) PXRD patterns were measured on a Bruker AXS Advance D8 diffractometer under similar conditions, but with a stainless steel sample holder and an Anton-Paar TTK450 variable temperature/humidity stage. Temperature was controlled to within (2 °C, while humidity was controlled to within (2% RH. Samples on the Advance D8 system were not rotated, and measurement of each pattern required 30 min. IR and Raman Spectroscopy. A Travel-IR diamond attenuated total reflectance (DATR) Fourier transform spectrometer was used to acquire

IR spectra at a resolution of 4 cm-1. Solid powders are pressed against the DATR window for this type of analysis, with no other sample preparation. The spectra are the average of 16 scans and were acquired at room temperature and ambient humidity using a DTGS detector. Raman spectra were recorded using a Nicolet FT-Raman 960 spectrometer equipped with a near-infrared laser (1054 nm wavelength), an InGaAs detector and a CaF2 beam splitter. The laser power was 500 mW. Samples were analyzed directly in their glass vials, which were sealed immediately upon their removal from either humidity chambers or the drying oven. A total of 256 scans were averaged at a resolution of 4 cm-1 for each of the Raman spectra (requiring approximately 8 min). No evidence of laser-induced sample decomposition or fluorescence was observed. Solution-state NMR Spectroscopy. Solution-state NMR spectra were obtained at 298 K using a Bruker DRX 700 spectrometer operating at a 1H frequency of 700.13 MHz. Complete 1H, 13C, and 15N assignments were made in DMSO-d6 solution at a concentration of 20 mg/mL using the following 1D and 2D experiments: magnitude-mode 1 H double-quantum filtered correlation spectroscopy (DQFCOSY), 13C gated spin-echo (GASPE), phase-sensitive 1H-13C heteronuclear singlequantum coherence (HSQC), magnitude-mode 1H-13C heteronuclear multiple-bond correlation (HMBC), and magnitude-mode 1H-15N heteronuclear multiple quantum coherence (HMQC) set for two different coherence evolution delay settings (3 and 10 Hz).30 A small amount of tetramethylsilane (TMS) was added to the sample as a 1H and 13C reference. 15N spectra were referenced to nitromethane with the unified scale method relative to TMS.31 Solid-State NMR Spectroscopy. SSNMR experiments were performed on a Bruker Avance 400 spectrometer operating at a 1H frequency of 399.87 MHz, equipped with a variable temperature system including a BCU-05 chiller, a liquid nitrogen heat exchanger, and liquid nitrogen boil-off gas system. The sample chamber temperature for MAS experiments was maintained at 273 K to limit the effects of frictional heating unless otherwise noted. A Bruker MAS-II rate controller was used to control spinning speeds to within (2 Hz of the set point. 13C spectra were externally referenced to TMS using hexamethylbenzene and 15N spectra were externally referenced to nitromethane using NH4Cl.32,33 Other references (1H and 23Na) were calculated from the experimental 13C references using the unified scale method.31 Basic single-pulse 1H experiments were run with a Bruker 2.5-mm triple-resonance probe spinning at an MAS rate (υr) of 35 kHz. A 2.5µs excitation pulse and a 30-s relaxation delay were used, and a total of 32 transients were averaged. Additional experiments with a 2-s delay and dummy pulses were used to check for variability in 1H T1 relaxation. 13 C CP spectra were measured at 100.559 MHz with a Bruker 4-mm triple resonance probe tuned to 1H, 13C, and 15N frequencies and spinning at υr ) 8 kHz. Samples were restricted to the center of the 4-mm rotors to improve RF homogeneity. A linear power ramp from 75 to 90 kHz was used on the 1H channel to enhance CP efficiency.34 Spinning sidebands were eliminated by a five-pulse total sideband suppression (TOSS) sequence.35 Heteronuclear proton decoupling was performed at an RF power of 105 kHz using the SPINAL-64 sequence.36 13 C spectral editing with the nonquaternary suppression (NQS) pulse sequence (also known as dipolar dephasing) used 625 µs of interrupted 1 H decoupling during the TOSS period and four subsequent shiftedecho rotor periods.37 The 40.525 MHz 1D 15N NMR spectra were obtained with a Bruker 7-mm double resonance probe using the basic CP-MAS pulse sequence with a 3-ms contact time, a 5-s relaxation delay, and a 1H decoupling power level of ∼65 kHz. A total of 16 384 transients were averaged for the spectra shown here. 23Na MAS spectra were acquired at 105.772 MHz on both the 2.5 mm (tuned to 1H, 31P, and 13C/23Na frequencies) and the 4-mm triple-resonance probes. 2D CP heteronuclear correlation (CP-HETCOR) experiments between 1H and 13C, 15N and 23Na nuclei made use of frequency-switched LeeGoldburg (FSLG) homonuclear decoupling at 105 kHz.38 Heteronuclear correlation experiments using J-coupling as a transfer mechanism were performed using the MAS-J-HMQC pulse sequence.39 The MAS rate was increased to 12.5 kHz for the CP-HETCOR experiments. Static 2H and 17O NMR spectra were obtained with a Bruker wideline probe. The samples were loaded into the center 2-cm portion of 5-mm outer diameter glass tubes using Teflon spacers. 2H spectra were obtained at 61.382 MHz using the quadrupolar echo pulse sequence, a pulse width of 3 µs, and a 2 MHz spectral width.40 A 20 µs echo delay was used and a 72 point left-shift was applied. A 5-s relaxation delay was used. Each spectrum is the result of 512 scans requiring 42 min.

2336 Crystal Growth & Design, Vol. 6, No. 10, 2006 17

O spectra were measured at 54.208 MHz using the basic spin-echo method with a 1.5 µs excitation pulse, a 3 µs refocusing pulse, and a 40 µs delay between the pulses. The decay was sampled at 2 MHz and was left-shifted by 142 points prior to apodization with a 500 Hz exponential function and Fourier transformation. A total of 32k points were acquired to ensure that any slowly decaying signal (i.e. from liquid-phase water) would be detected. Each 17O spectrum is the result of either 64k or 128k scans with a 500 ms relaxation delay, requiring between 9 and 18 h. Computational Methods. Mathematica Version 5.0 (Wolfram Research, Champaign, IL) was used for quadrupolar line shape calculations and simulations. Line shapes were also fitted using the solids line shape analysis and deconvolution routines in Topspin 1.3 (Bruker BioSpin, Billerica, MA). Calculations of electric field gradient (EFG) tensors were accomplished using the Gaussian 03W software package using B3LYP density functionals and the 6-31G(d) basis set.41,42

Results Single-Crystal X-ray Diffraction. Compound I crystallizes in the monoclinic space group P21/c, a commonly encountered space group for organic compounds without a chiral center. A crystal structure was determined at 173 K, in which there are three water molecules for each molecule of compound I and sodium cation. This trihydrate structure is referred to as Form 1. Attempts to dehydrate single crystals caused cracking and jumping that prevented data collection on lower hydration states. The properties of the Form 1 crystal structure are summarized in Table 1 and a view of the single parent molecule in the asymmetric unit (Z′ ) 1) with thermal displacement ellipsoids is shown in Figure 1a. The parent molecule is ionized at the hydroxyl oxygen (O4), based on the shortened C2-O4 distance of 1.298(7) Å relative to values typically observed for phenols (1.33-1.37 Å). The phenyl rings adopt a planar but slightly twisted conformation. The planes defined by rings C1-C6 and C10-C15 are rotated by 18.0(4)° and 9.0(4)°, respectively, in the same direction from the plane of the urea group. The dihedral angle between the phenyl ring planes is 9.0(4)°. One of the water molecules (O40) forms hydrogen bonds between a urea oxygen atom (O3) in one parent molecule and the urea nitrogen atoms (N7 and N9) in an adjacent molecule. The other two water molecules (O50 and O60) are coordinated to the sodium cation. The hydrogen bonding network around a single molecule is illustrated in Figure 1b, with metrical details given in Table 2. On the basis of their dimensions, the hydrogen bonds observed in this structure are moderate in strength (i.e., 4-15 kcal/mol).20 Each sodium cation is coordinated to six oxygen atoms: two from water molecules, three from SO2 groups, and one from the phenol. The Na-O distances are in the range 2.328(5) to 2.472(7) Å, and the sodium cations are 4.221(3) Å apart. There is a π-stacking interaction between the phenyl ring, C1-C6, and an adjacent C10-C15 phenyl ring. The ring centroids are separated by 4.11 Å. The perpendicular distances between the ring plane and atoms of the adjacent ring vary from 3.11 to 3.42 Å. No unusual bond distances or angles were observed in the structure. The unit cell parameters from the single crystal study were refined against the powder diffraction pattern of a Form 1 sample equilibrated at 33% RH using the Rietveld method. Peak overlap in the observed diffraction pattern makes it difficult to index and accurately measure the positions of each diffraction peak for conventional least-squares refinement of unit cell parameters. A better choice is the Rietveld method, which obtains a global least-squares fit between calculated and observed powder patterns.43 The measured powder diffraction pattern for the trihydrate state is compared to the results of the Rietveld

Vogt et al.

refinement in Figure 1c. Background, peak half-widths, and preferred orientation effects (using a single-parameter version of the March-Dollase model testing all crystallographic axes)44 were included in the refinement, and the region below 2θ ) 4° was excluded. The final refinement yielded residuals (RP) of 22.0% and weighted residuals (RWP) of 28.2%. The residual intensity observed in the difference pattern shown in Figure 1c is primarily the result of preferred orientation effects, as the presence of contaminating phases is ruled out by SSNMR. The room-temperature refinement indicates expanded unit cell dimensions of a ) 20.122 Å, b ) 6.878 Å, c ) 13.693 Å and a shift in β to 91.655° (for an overall cell volume of 1894.3 Å3, or a 1.3% increase in volume between 173 K and room temperature). Agreement is observed for the strongest reflections at d-spacings of 20.26111 Å (hkl ) 100), 10.09334 Å (hkl ) 200), 6.86069 Å (hkl ) 002), and 5.03738 Å (hkl ) 400). The results ensure that the single crystal studied was representative of the bulk material, so that spectroscopic results can be confidently interpreted in conjunction with the SCXRD structure of the trihydrate state. Visualization of the packing in Form 1 shows the presence of potential solvent tunnels running lengthwise along the b-axis of the unit cell, as shown in Figure 2a. The tunnels appear as two separate sections: a “main” tunnel containing O50 and O60 with an approximate cross-sectional area of about 16 Å2 (estimated from van der Waals radii) and a second smaller tunnel containing O40 with a cross-sectional area of about 4 Å2. A more complex arrangement is observed along the c-axis of the cell in Figure 2b. A plane runs along this axis that contains the sodium, O50 and O60. Potential connecting tunnels are also seen between this plane and the O40 sites. Analysis of the contacts made by the three water molecules observed in the X-ray structure suggests that one water (O40) could be less tightly held than the other two (O50 and O60), both of which associate with the sodium cation. Although O40 participates in several hydrogen bonding interactions, it is not bound to the sodium cation, and at first glance might be construed as a weakly bound, readily removed mole of channel water. This hypothesis is consistent with the initial water loss seen in the DSC thermogram of fully hydrated material (see below). The VT and VH PXRD experiments and the multinuclear SSNMR analysis discussed in the following sections indicate that internal water dynamics, desolvation, and external vapor exchange processes are actually far more complex. Thermal and Moisture Sorption Analysis. The GVS isotherm for Form 1 at 25 °C is shown in Figure 3a. The sample was held at a starting point of 40% RH for 24 h. The starting point of 40% RH was set to zero for the % mass change scale and was also used to correct the data so that the molar ratio is centered on a trihydrate state, based on KF titrations on material equilibrated at a similar humidity (see below). On the first cycle, the increase of relative humidity to 90% leads to a 0.3% w/w water uptake into a plateau region. The first desorption cycle is reversible through a 0.5% w/w water loss, the signature of an isomorphic desolvate. However, the isotherm also shows hysteresis effects below 20% RH. To clarify this region, a second GVS experiment was conducted using 1% RH steps and is shown in Figure 3b. This experiment was preequilibrated and started at 20% RH; the mass change was calibrated to match that of the full-range GVS experiment. The hysteresis is more easily seen with 1% RH steps and suggests a complex reversible phase change between the trihydrate state and a sub-dihydrate state, with a 4.5% w/w loss to 0% RH. The phase change that occurs between 6% RH and 12% RH appears to be reversible

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2337

Figure 1. Results of the SCXRD analysis of Form 1. (a) Thermal ellipsoids (50% probability) for the single molecule in the asymmetric unit of Form 1 and the sodium cation at 173 K. (b) Water molecules and the hydrogen bonding contacts (dashed lines) for Form 1. The hydrogen atoms on O60 are poorly defined and are shown in possible positions. (c) Comparison of an experimental flat-plate PXRD pattern for Form 1 trihydrate at 298 K with a simulated pattern from a Rietveld refinement of the SCXRD structure (RP ) 22.0% and RWP ) 28.2%). The difference between the patterns is also shown, plotted with the same vertical scaling.

through a second cycle; this was confirmed by removing the sample from the GVS instrument after these experiments and obtaining a PXRD pattern, which matched the Form 1 reference. An additional drying experiment was performed in the GVS instrument at 45 °C using dry nitrogen. After 34 h, the sample mass had decreased by 8.8% w/w, indicating that a large portion of the starting amount of water had been removed. Upon reaching this level, the sample mass changed by less than 0.02% w/w over an additional 25-h holding period. The sample was then exposed to increasing humidity up to 40% RH at 45 °C. As the sample rehydrated, a mass jump of 9% beyond the initial weight was recorded. However, PXRD examination of samples

exposed to this procedure confirmed their identity as Form 1, and their water content was found to be in the trihydrate range by KF analysis. The behavior observed in the drying experiment is attributed to the presence of a transient amorphous or semicrystalline state. On the basis of the GVS results, equilibrium hydration experiments were performed to prepare samples for spectroscopic analysis. Four samples of Form 1 were stored in preequilibrated chambers at 75% RH, 33% RH, 11% RH, and 6.5% RH for up to six weeks. KF titrations of these samples found water contents of 11.5% w/w, 11.3% w/w, 10.9% w/w, and 6.7% w/w, respectively. The additional nonequilibrium sample stored over cal-

2338 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al.

Table 2. Hydrogen Bond Metrics for the Single Crystal X-ray Structure of Form 1a

a

-z.

hydrogen bond D-H‚‚‚A

D-H distance (Å)

H‚‚‚A distance (Å)

D‚‚‚A distance (Å)

DHA (°)

N1-H1NA‚‚‚O40a N1-H1NB‚‚‚O4 N7-H7N‚‚‚O40b N7-H7N‚‚‚O4 N9-H9N‚‚‚O40b N9-H9N‚‚‚Cl2 O40-H40A‚‚‚O3 O40-H40B‚‚‚O4c O50-H50A‚‚‚N1 O50-H50B‚‚‚O60a O60-H60A‚‚‚Cl1d O60-H60B‚‚‚O60d O60-H60B‚‚‚Cl1

0.882(13) 0.877(13) 0.879(13) 0.879(13) 0.878(13) 0.878(13) 0.841(13) 0.838(13) 0.84 0.84 0.81 0.87 0.87

2.08(2) 2.11(5) 2.13(2) 2.32(6) 2.41(3) 2.45(6) 1.974(16) 2.18(4) 2.27 2.02 2.50 2.13 2.96

2.935(7) 2.798(7) 2.989(7) 2.686(6) 3.217(7) 2.927(5) 2.813(6) 2.903(6) 3.053(7) 2.857(9) 3.308(7) 2.791(13) 3.734(7)

164(6) 134(5) 167(6) 105(5) 153(6) 115(5) 176(7) 145(6) 156.3 179.2 179.4 132.1 148.8

Symmetry transformations used to generate equivalent atoms: ax, -y + 3/2, z + 1/2. bx, -y + 5/2, z + 1/2. cx, -y + 5/2, z - 1/2. d-x + 1, -y + 1,

cium silicate desiccant for the same period had a water content of 5.4% w/w. Comparison of samples pulled after four- and six-week periods did not differ appreciably in water content, indicating that equilibrium had been reached in the chambers. Additional low water content nonequilibrium samples were freshly prepared just prior to spectroscopic analyses by drying the desiccated material in a vacuum oven at 60 °C and at 25 in. Hg for either 2 or 16 h. These samples were found to have nominal water contents of 2.3 and 0.5% w/w, respectively. (A variability of ( 0.3% w/w was observed with the drying samples over three attempts at preparation and was ascribed to the nonequilibrium nature of the drying and also to rapid rehydration during material handling.) As the theoretical water content is 11.1% w/w for a stoichiometric trihydrate and 7.7% w/w for a dihydrate, it is apparent that Form 1 exists under normal ambient laboratory conditions as an approximate trihydrate. It is also apparent that the 6.5% RH chamber and the desiccated chamber are capable of producing samples with water content corresponding to a stoichiometry below a dihydrate. Drying yields significantly lower water contents that approach an anhydrous state. The results of the GVS and KF titration studies show that Form 1 only takes up a small amount of water beyond the trihydrate state. An important finding is that samples with lower hydration states are stable enough to be quickly analyzed by KF titration, indicating that other analysis methods (e.g., NMR) can also be applied. DSC thermograms for three of these samples are plotted in Figure 4. The thermogram for the trihydrate state shows two large endotherms with peaks at 81.7 and 148.1 °C. The first endotherm onsets at ∼65 °C and has an average ∆H of 31.9 J/g (from three measurements). The second endotherm appears to be a composite event, with a typical onset at 143.6 °C and an average ∆H of 101.2 J/g. The first endotherm is significantly diminished in the DSC traces of two lower hydrates, and nearly disappears when the water content is lowered to 2.3% w/w. The endotherm also shifts to a lower peak temperature of 58 °C, implying that this water is more easily removed. Also, the large composite endotherm appears to simplify, separating from the melt while also becoming smaller and moving to a lower temperature maximum, suggestive of a phase change. Like the GVS data, the DSC results hint at the complexity of dehydration in Form 1, a process that will be enlightened by the results discussed below. TGA data on the trihydrate (not shown) indicated an initial loss from ambient temperature of 1.6-1.8% w/w, most likely involving the loosely bound water observed in the GVS experiments. This was followed by a step beginning at 60 °C and corresponding to a 10.1% cumulative weight loss by 160

°C (the theoretical water content for a trihydrate is 11.1% w/w). After water loss is completed, an additional 8.2% weight loss occurs between 160 and 200 °C from decomposition reactions. Complete decomposition occurs at temperatures greater than 210 °C. Attempts at using TGA as an orthogonal check of hydration level for the lower hydrates were not successful, as samples re-hydrated significantly before the analysis could be started. (TGA can often act as helpful verification of KF titration results.)15 KF titration was therefore deemed to be the most analytically reliable test for water content in the present study. Variable-Humidity and Variable-Temperature PXRD. Attempts to study preequilibrated or dried samples of Form 1 by PXRD on an open stage were not possible because of rehydration during the analysis. This is expected because the sample is thinly spread in the holder and is exposed to the ambient atmosphere for about 10 minutes during sample preparation and measurement. The lower hydrates seen in the DSC thermograms can partially rehydrate before the scan is finished. Because of this, a series of controlled humidity and temperature X-ray diffraction experiments were performed on the Anton-Paar stage using the Advance D8 diffractometer, to ensure that representative powder patterns were obtained. Although water content could not be directly measured in situ on the VH/VT PXRD samples, the GVS results relate the observed patterns to water content in the equilibrium samples by way of the KF titration results. The results of the VH PXRD study are illustrated in Figure 5a. Each pattern was collected after a two-day holding period at the target humidity, which, although not as long as the equilibrium chamber storage periods, was still deemed sufficient for the bulk of the material to reach the necessary state based on GVS results. The pattern shown in Figure 5a obtained at 35-37% RH matches that of the phase studied by SCXRD (Figure 1) and is clearly Form 1. Between 15% RH and 75% RH, the Form 1 pattern is observed without any changes although the GVS data indicates a 0.5% weight change in this range. The 35-37% RH pattern is shown as a representative of Form 1, which can thus be classified as a nonexpanded channel hydrate.3 As the humidity is lowered from 12% RH to 5% RH, a different pattern appears. The changes in this pattern are significant enough to label it as a second phase, characterized by strong X-ray peaks at 2θ values of 20.2° (d ) 4.38 Å), 24.3° (d ) 3.66 Å), and 31.2° (d ) 2.87 Å). This pattern corresponds to water content in the 5-7% w/w range, an approximate dihydrate, and is referred to hereafter as Form 2. (Between 12% and 15% RH, a mixture of Form 1 and Form 2 was observed.) Further reduction of the humidity to near 0% causes the appearance of a third distinct phase with 2θ peaks at 13.8° (d ) 6.42 Å) and 26.5° (d ) 3.36 Å). This pattern,

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2339

Figure 2. (a) Packing in Form 1, viewed along the b-axis of the cell; two unit cells along the a and c axes are shown. The water channels along this direction are marked with dashed lines. Arrows show potential motion along the “plane” and sodium atoms are denoted by dark circles. (b) A top view of the channels along the c-axis. Potential paths for water in the tunnels are shown and the water “plane” along the c-axis is at the center.

obtained by blowing dry nitrogen over the thinly spread sample for 2 days prior to data collection, still contains a large amount of Form 2. Form 3 is the name given to this third phase. On the basis of the results of the GVS and KF analyses, the Form 3 pattern is expected to be representative of the material with water contents less than 3% w/w (approximate monohydrates or anhydrates). It is interesting to note that during the dehydration processes, the longest d-spacing reflection drops from 20.24 Å (2θ ) 4.4°) at 35% RH to 19.2 Å (2θ ) 4.6°) at 0% RH, suggesting a minor overall shrinkage of the lattice. The conservation of this reflection from Form 1 to Form 3 signifies

an overall preservation of long-range order in the crystal structure. The dehydration process, involving two phase changes and complex water loss steps, was judged to be fully reversible during the VH PXRD study. When the sample dried under nitrogen was subsequently exposed to 35% RH, its XRD pattern reverted to that of Form 1. The reconversion time was estimated at less than 5 min but was found to be dependent on factors such as sample thickness, surface area, and particle size. Variable-temperature PXRD was used to help understand the DSC thermogram of Form 1. The sample was initially heated from 30 to 120 °C in increments of 30 °C, with a pattern being

2340 Crystal Growth & Design, Vol. 6, No. 10, 2006

Figure 3. (a) GVS isotherm for Form 1 at 25 °C on a sample preequilibrated in the instrument at 40% RH for 24 h, obtained using 10% RH steps (with a 5% RH step at low humidity). The sorption and desorption cycles are super-imposable above 20% RH; below this level, deviations are observed. The % weight change is zeroed at the start of the run (40% RH). (b) GVS isotherm at 25 °C obtained using 1% RH steps, starting from a zero point of 20% RH, and showing a hysteresis effect.

measured at each temperature after a 4-h holding period. The results are shown in Figure 5b. It is apparent that heating also induces dehydration and the related phase change from Form 1 to Form 2; the distinctive PXRD pattern of this phase is clearly observed at 90 °C and matches that obtained between 5 and 12% RH, while a mixture of Forms 1 and 2 is observed at 60 °C. At a temperature of 120 °C, Form 3 appears with a pattern similar to that observed at 0% RH in the VH PXRD data. After reversing the VT cycle and cooling to 30 °C in 30 °C increments, the patterns were observed to almost entirely revert to their previous form after a 1-h holding period at each temperature. (Small regions in the sample were likely prevented from quick re-hydration by limited surface area.) The original Form 1 pattern was obtained after several hours at ambient temperature and humidity, indicating full reversibility. Finally, the sample temperature was raised to 90 °C and then 150 °C, where a Form 3 pattern was measured as shown in Figure 5b. After an hour at this temperature, the sample was returned to 30 °C and was observed to reconvert to Form 1 after several hours. Solution and Solid-State NMR Assignments. Before proceeding with the SSNMR analysis, it is necessary to assign as many resonances as possible in the NMR spectra of compound I. Solution-state chemical shift assignments are the first step in this process. The 1H, 13C, and 15N assignments for I in DMSO-

Vogt et al.

d6 solution are reported in Table 3. All 1H resonances were assigned, including the exchangeable protons (by way of their HMBC-detected interactions with aromatic carbons). All 13C assignments were confirmed except for indistinguishable 13C signals for C3 and C12 at ∼131 ppm. 15N assignments were made using two long-range HMQC experiments, from which N7 and N9 were definitively located by three-bond J-couplings with H4 and H15, respectively, and N1 was assigned by a onebond correlation with its directly attached protons. SSNMR assignments were made using the spectra of a sample of Form 1 equilibrated at the trihydrate state (33% RH, 11.3% w/w water). The 1H spectrum of this sample (υr ) 35 kHz) is shown in Figure 6a. Three components dominate the spectrum: a sharp signal at 3.8 ppm, a much broader signal at 7.9 ppm, and a shoulder peak at 9.3 ppm. This spectrum can be simulated from mixed (1:1) Lorentzian and Gaussian line shapes centered at these three frequencies, with relative intensities of 4:10:1 and relative line widths of 6:19:8, as shown in Figure 6a. These results are confirmed using deconvolution methods and allow for integral ratios to be assigned to the components. Simple assignments of the 1H spectrum of Form 1 follow based on expected chemical shifts. The peak at 3.8 ppm, corresponding to four protons, is assigned to O50 and O60 water molecules. The narrowing of this peak is ascribed to motion of the water molecules in the main channel. The broader peak centered at 7.9 ppm with the shoulder is assigned to the other 11 protons in the solid-state structure, including those attached to O40, N1, N7, and N9. It is notable that the 1H spectrum does not show any resonances above 10 ppm, which is consistent with the lack of short hydrogen bonds in the crystal structure.45 The shortest donor-acceptor distance is 2.7 Å, and the shortest hydrogenacceptor distance is 2.0 Å, both of which normally coincide with moderate-strength hydrogen bonds with energies in the range of 4-15 kcal/mol. Further insight into the Form 1 1H spectrum requires heteronuclear correlation experiments. The 13C CP-TOSS spectrum of Form 1 is shown in Figure 6b. The assignment of resonances above 130 ppm was easily made by comparison with the solution-state results. Spectral congestion in the aromatic region was resolved using the NQSedited spectrum, which allowed for the assignment of the overlapped quaternary aromatic carbons including C1, C3, C6, C11, and C12. These assignments are aided by knowledge that the chlorine-bearing carbons are broadened by residual dipolar coupling to quadrupolar 35/37Cl nuclei.46 The residual coupling effect should be especially strong for C11 and C12 since each has two quadrupolar coupling partners with different orientations of their EFG tensors. The C12 resonance clearly shows this broadening effect in the CP-TOSS-NQS spectrum. The C6 and C11 signals overlap and are difficult to assign (their solutionstate chemical shifts differ by less than 0.2 ppm). Relying on the argument that C11 should be additionally broadened by residual coupling to two 35/37Cl nuclei and should thus have a similar line shape to that of C12, we have tentatively assigned C6 to the sharper, more intense peak and C11 to the broad component visible beneath C1. The assignments of the protonated carbons in the congested aromatic region are confirmed using the one-bond 2D MAS-J-HMQC spectrum shown in Figure 6c, which identifies those directly attached carbons and protons that engage in a one-bond J-coupling. The assignments shown on the MAS-J-HMQC spectrum are straightforward except for two positions that show an interesting shift between solution-state and Form 1. Specifically, H4 and C4 are observed at 7.77 and 116.41 ppm, respectively, in DMSO solution, while H15 and C15 are observed at 8.03 and 120.95 ppm. However,

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2341

Figure 4. DSC traces for three water content levels of I. Endothermic heat flows are plotted in the upward direction. The top trace was obtained from material equilibrated at the trihydrate state. The lower-temperature endotherm peaks at 81 °C and is ascribed to loss of loosely bound channel water. The higher-temperature composite endotherm is caused by the loss of additional water and a possible phase change. When material is equilibrated at the approximate dihydrate state, the first endotherm is greatly reduced in size and peaks at 60 °C. The water loss is still apparent in the 2-h vacuum-dried sample where the water content by KF titration is below a theoretical monohydrate state.

in the solid state, the trend is reversed; that is, the more shielded proton at 6.7 ppm is attached to the more deshielded carbon at 119.69 ppm, while the deshielded proton at 9.2 ppm is attached to the more shielded carbon at 117.29 ppm. The distinctive difference between H4 and H15 in solution makes it likely that it is the carbon chemical shifts and not the proton shifts that have actually changed in the solid state. The assignment of C4 and C15 given in Figure 6 reflects this assumption. The reversal of the 13C chemical shifts in the solid state can be attributed to differences in conformational preferences in DMSO solution and the crystal structure, or to changes in π-stacking effects. MAS-J-HMQC experiments with a mixing delay of 16 ms could not resolve this ambiguous assignment, as the HMBC-like correlation between H4 and C2 was not detectable. The Form 1 solid-state assignments are further clarified by heteronuclear dipolar correlation experiments containing throughspace information about the neighbors of several heteronuclei. In Figure 7, the results of three separate CP-HETCOR experiments correlating 1H with 13C, 15N and 23Na nuclei are plotted alongside each other for comparison. The indirectly detected CP-HETCOR spectra with FSLG decoupling have sufficient resolution to locate nearly all of the protons in the crystal structure. In the 1H-13C CP-HETCOR spectra, key correlations serve to locate H7, H9, and H15 (the last of which supports the proposed assignment of the ambiguous C4/C15 shifts). A through-space correlation between C3 and H4 (2.0 Å) is also visible. A weak correlation between C8 and a signal

at ∼6 ppm is tentatively assigned to the final mole of water (O40), as the C8-H40 distance is about 2.9 Å. The broad C6, C11, and C12 signals are absent in the 1H-13C HETCOR spectrum, which helps relieve some of the congestion in the aromatic region. The protons of the NH2 group (H1NA and H1NB) are not easily detected because of broadening of the N1 signal but may contribute to the deshielded portion of the 1H-23Na correlation. The broadening of N1, discussed further below, is most likely the result of a dynamic process involving the sulfonamide protons. The other 15N sites are readily assigned via strong correlations between the nitrogen and its pendant proton in the 1H-15N CP-HETCOR spectrum. H7 is observed at 9.5 ppm, while H9 is seen at 8.2 ppm. This shift trend is consistent with H7 engaging in a slightly shorter (by 0.2 Å) hydrogen bond to O40, as shown in Table 2. The weak 1H-15N correlation is assigned to a long-range interaction of N7 with a proton at 7.0 ppm (the nearby H4 position). From examination of the crystal structure, the 23Na site is only expected to interact with protons on the surrounding O50 and O60 water molecules, which are 3 Å away. The 1H-23Na CP-HETCOR data shows this expected correlation, helping to confirm the initial assignment of the narrowed 1H signal at 3.82 ppm. Because of the evidence of dynamics in the 1H and 15N MAS spectra, attempts were made to perform these experiments at lower temperatures; however, at the lowest temperatures achievable with the 2.5and 7-mm probes (230-250 K), neither the 1H spectrum nor the N1 signal in the 15N CP-MAS spectrum were detectably

2342 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al. Table 3.

1H, 13C,

and

15N

NMR Assignments for Compound I in DMSO-d6 Solution 13C

position

1H

shift (ppm)a

C1 C2 C3 C4 C5 C6 N7 C8 N9 C10 C11 C12

7.77 (1H, d, J ) 8.3 Hz) 6.05 (1H, d, J ) 8.3 Hz)

C13 C14 C15 N1

7.25 (1H, dd, J ) 8.2, 1.4 Hz) 7.28 (1H, t, J ) 8.2 Hz) 8.03 (1H, d, J ) 8.2, 1.4 Hz) 7.97 (2H)

a

15N

shift (ppm)a

shift (ppm)b

123.75 160.07 131.45, 131.34c 116.41 108.61 121.37

-277.8 -273.8

9.26 (1H) 152.31 9.27 (1H)

b

138.58 121.20 131.45, 131.34c 123.28 127.63 120.95

-286.0 c

In ppm relative to TMS. In ppm relative to CH3NO2. Values are ambiguous because of spectral overlap.

Figure 5. (a) Variable humidity PXRD patterns at 25 °C of I obtained on samples preequilibrated for 2 days. No change was observed in the pattern between 15% RH and 75% RH (the 35-37% RH pattern is shown as a representative). Between 5% RH and 12% RH, a second phase is observed. At ∼0% RH (dry nitrogen), a third phase begins to appear. (b) Variable temperature PXRD patterns of I obtained from 30 to 150 °C. The pattern obtained at temperatures near 90 °C matches with that observed at 10-12% RH. A phase mixture is observed at 60 °C, during the first DSC desolvation event. A third phase-like that seen at ∼0% RH appears at 120-150 °C.

affected. Other avenues for study of the dynamic processes in Form 1 are discussed below. Investigation of the Dehydration Process with Raman and NMR Spectroscopy. FT-Raman spectroscopy was used to characterize the phase changes observed with different hydration states in the VT and VH PXRD experiments. A variablehumidity and temperature Raman stage was not available, but the various equilibrated and dried samples could be examined directly in their vials and linked with PXRD results through

the KF and GVS water content measurements. Expanded portions of the FT-Raman spectra of the three phases are shown in Figure 8, where the spectrum above ∼1800 cm-1 is omitted because of a lack of informative effects in this region. Forms 1 and 2 are not easily distinguished by Raman spectroscopy. A band with a maximum at 368 cm-1 and a shoulder at 357 cm-1 in Form 1 shifts to a single band at 364 cm-1 in Form 2. These low-frequency bands are readily assigned to lattice vibrations. The resemblance of the FT-Raman spectra indicates a degree of similarity in the structures of Forms 1 and 2 despite their different PXRD patterns. The lower hydrate (Form 3) shows a very different Raman spectrum. The carbonyl stretch, at 1687 cm-1 in the higher hydrates, splits into two components at 1695 and 1680 cm-1. This can be interpreted as the appearance of two different types of carbonyl groups, probably because of a transition from the single molecule in the Form 1 asymmetric unit to a Z′ g 2 situation in Form 3. The other changes seen in the Form 3 spectrum are marked with arrows in Figure 8. Despite the changes, the overall similarity of the lattice vibrations (below 600 cm-1) indicates that Form 3 retains some aspects of the structures of Forms 1 and 2. FT-IR spectra of samples equilibrated at the trihydrate state and spectrum were also obtained, but the changes between various hydration states were not as clear and the data were not as analytically useful. The FT-Raman spectra were preferable because of their simplicity, flat baseline, and superior information content at lower frequencies. As previously noted, the dehydration of Form 1 single crystals could not be controlled to the extent required for an SCXRD study, because of the occurrence of phase transitions to Forms 2 and 3. (This is unfortunate as SCXRD studies provide useful information about the response of the host molecule to the voids left in the crystalline lattice; however, these studies are often unsuccessful because of the difficulties in preserving the crystal integrity during drying.) Nevertheless, since powdered dehydrated samples could be prepared and quickly packed into a rotor, the structural aspects of the hydration/dehydration process could be studied using the four NMR-active nuclei available in the salt, each of which is positioned to report on key features of the process. The variable water content samples prepared in the equilibrium and drying chambers and assayed by KF were used for the NMR study. Since the VT and VH PXRD data indicate that Forms 2 and 3 can be prepared either by humidity

Crystalline Pharmaceutical Hydrate

Figure 6. 1H and 13C SSNMR assignments for the trihydrate Form 1 at 273 K. The assigned 1H spectrum (υr ) 35 kHz) is shown in (a) along with a simulated spectrum (see text). The assigned 13C CP-TOSS and 4-rotor-cycle NQS (υr ) 8 kHz) spectra are shown in (b). In (c), the 2D 1H-13C MAS-J-HMQC spectrum (υr ) 12.5 kHz) is shown with the 1H and 13C CP-TOSS spectra plotted along the F1 and F2 axes, respectively.

or temperature changes, the results should be indicative of either method of dehydration. Use of the equilibrium humidity samples avoids problems encountered attempting to prepare Forms 2 and 3 in situ by heating the sample in the MAS rotor, including spinning instability caused by escape of water vapor, the limited surface area of the rotor preventing equilibration, the inability to measure water content directly, and enhanced frictional heating at high spinning rates. The KF titration results demonstrated that the samples maintained water content long enough to be quickly packed in a rotor. The 13C CP-TOSS spectra of the variable water content samples are shown in Figure 9a. There is little overall change

Crystal Growth & Design, Vol. 6, No. 10, 2006 2343

in the Form 2 13C spectra in comparison to Form 1. Like the FT-Raman results, this suggests that despite the changes in the PXRD patterns of Forms 1 and 2, the local environment of the parent molecule is conserved. However, a general broadening of the spectra is observed as the water content drops and may be related to the changes in long-range order for Form 2 detected by PXRD. The two samples with 6.7 and 5.4% w/w water content are considered to be Form 2 based on KF titration results, which are linked to the Form 2 VH PXRD pattern via the GVS isotherm. (This link was not possible for the sample with 2.3% water content, so it is labeled Form 2 by inspection of its 13C spectrum.) When the water drops to 0.5% w/w, the phase change to the nearly anhydrous Form 3 is apparent. Both the broadening effect of crystalline defects and a splitting indicative of a change to Z′ g 2 are noticeable in the Form 3 spectrum. The C2 phenol signal, originally at 162.20 ppm in Form 1, splits into two components at 164.28 and 161.18 ppm in Form 3. The C8 carbonyl develops a second peak at 151.56, shielded relative to the original peak at 153.31 ppm. This upfield shift of C8 most likely indicates the loss of the O40 hydrogen bond to the carbonyl oxygen. This is a well-known trend, caused by the ability of hydrogen bonds to polarize the CdO bond, so that electron density is pulled away from the carbon, causing deshielding of carbonyl 13C sites with stronger hydrogen bonding.47 This trend can be used to help understand differences between polymorphs.48,49 In the present case, the absence of this hydrogen bond in Form 3 is consistent with the metastable nature of this phase. The increased line widths in the Form 3 spectrum are indicative of semicrystalline or disordered content and are also an explanation for its metastable behavior. The 1H fast MAS spectra obtained from samples of varying water content are compared in Figure 9b. As previously noted, the partially narrowed resonance at 3.8 ppm in the Form 1 trihydrate spectrum is assigned to the four protons on the O50 and O60 water molecules. The sample equilibrated at 11% RH yielded the same 1H spectrum as that stored at 33% RH and is not shown, as expected from the GVS and KF results. For the rest of the equilibrated and dried samples, the 1H spectrum is appreciably altered as a function of water content. The Form 2 sub-dihydrate samples show an additional 1H peak at 2.1 ppm. The shielding effect on this new peak is consistent with the breaking of O50 and O60 hydrogen bonds from known 1H chemical shift trends.45 The water molecules resonating at 2.1 ppm are no longer engaged in the original hydrogen-bonding network and are more mobile in the partially collapsed lattice. The peak at 3.8 ppm decreases in intensity as a consequence of this change in water environment. The vacuum-dried sample, representative of Form 3, shows almost no trace of these peaks, indicating that O50 and O60 water has been fully removed (in agreement with the KF results). The slight “hump” at 9.2 ppm is also missing in the Form 3 spectrum. This change is ascribed to either removal of the O40 water molecule or the effects of the form change; the lack of resolution prevents any further analysis. In Figure 10a, the 15N CP-MAS spectra as a function of water content are shown. The appearance of Form 2 as a second phase in Form 1 can be seen in the sample with 6.7% w/w water. A more distinctive 15N spectrum of Form 2, free from the interference of other phases, is observed when the water content is lowered to between 5.4 and 2.3% w/w. Several features of the Form 2 15N spectrum are significant. The most shielded peak at -288.7 ppm, assigned to N1, becomes much more pronounced in spectra of Form 2; this peak is broad and barely detectable in Form 1 (and disappears again upon the transition

2344 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al.

Figure 7. 1H-13C, 1H-15N, and 1H-23Na CP-HETCOR spectra (using a 400 µs contact time) of trihydrate Form 1 at 298 K. The full range of 1H environments can be seen. The NH2 group is too broad to be observed in the 1H-15N HETCOR spectrum (see text). The 1H spectrum (υr ) 35 kHz) is plotted along the F1 axis while the individual 13C CP-TOSS, 15N CP-MAS, and 23Na MAS spectra are shown along the F2 axes. The 1H assignments of long-range dipolar correlations are noted.

to Form 3). This can be explained by postulating that the water dynamics also involve the NH2 group and are at least partially quenched in Form 2 at intermediate water contents, possibly because cooperative effects of the water in the main channel are altered by its dilution. The N9 resonance, detected at -269.0 ppm in Form 1, shifts slightly upfield to -270.5 ppm in Form 2, while the N7 signal at -273.7 ppm in Form 1 also shifts in the same direction to -275.0 ppm in Form 2. The shielding effect observed for both nitrogen environments likely results from a weakening of the hydrogen bonds to the O40 water molecule since this water accepts hydrogen-bonds from both N7 and N9 in Form 1 crystal structure. Since both N7 and N9 act exclusively as hydrogen-bond donors, and shielding trends for nitrogen donors have been correlated with either the loss of a hydrogen bond or a geometric change in the hydrogen bond to a less favorable arrangement, it is reasonable to attribute this effect to hydrogen bonding, especially given the overall similarities of the 13C and Raman spectra of Forms 1 and 2.50-52 It is interesting to note that no change is observed in the frequency of the 13C carbonyl signal or the Raman shift of the CdO stretch between Forms 1 and 2, indicating that O40 is still donating a hydrogen bond to this site. An important clue about the disposition of Form 2 is thus obtained: the removal of O50 and O60 water is from the primary 16 Å2 channel affects the O40 water as well. The 15N CP-MAS spectrum of Form 3,

seen at 0.5% w/w water content in Figure 10a, also contains useful evidence about its structure. As previously noted, the signal from N1 again broadens beyond detection. Both the N7 and N9 resonances split into two components in Form 3, confirming the observation in the 13C spectrum that Z′ g 2 for this phase. Of the two urea nitrogens, N7 is more strongly affected by the conversion to Form 3, splitting into resonances at -275.0 and -278.3 ppm. This can be explained by again noting that N7 is engaged in the shorter hydrogen bond to O40 in the Form 1 crystal structure; removal of this last mole of water would be expected to affect its chemical shielding environment even more than in Form 2. This confirms the removal of O40 as one of the last steps in the desolvation process, which is also consistent with the DSC, GVS, and SCXRD data. The final nucleus used to monitor the dehydration of Form 1 is 23Na, which differs in being a quadrupolar nucleus that can interact with the EFG at the nuclear site. The EFG is described by a tensor V with three principal components arranged by magnitude so that |Vzz| g |Vyy| g |Vxx|. The quadrupolar coupling constant χ determines the overall strength of the interaction and is given as χ ) eQVzz/h (where Vzz ) eq). The asymmetry parameter measures the deviation of the EFG tensor from axial symmetry and is defined as η ) (|Vyy| - |Vxx|)/|Vzz| with possible values in the range of 1 g η g 0. Figure 10b shows single-

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2345

Figure 8. Expanded regions of the FT-Raman spectra of Form 1, 2, and 3 samples at hydration states measured by KF titration. Only minor differences were observed between the spectra of samples in the trihydrate (Form 1) and approximate dihydrate (Form 2) states. The arrows denote changes in spectral features observed between the various phases upon dehydration.

pulse 23Na MAS data for the different water content samples measured at a spinning rate of 25 kHz. The 1D single-pulse spectrum of Form 1 was previously shown along the F2 axis of the 1H-23Na HETCOR results in Figure 7. The 23Na resonance appears as a Lorentzian line at an apparent chemical shift of -9.3 ppm relative to NaCl with a full width at half-maximum of 388 Hz. The line shape is narrower than that typically observed for crystalline organic sodium salts and indicates a χ of 1 MHz or less, which can be attributed to approximate octahedral symmetry seen in the crystal structure and possibly to the effects of fast dynamics involving the water sites. The apparent chemical shift of the 23Na (δexp) site is affected by the second-order quadrupolar interaction and is related to the true chemical shift (δiso) by:53,54

(1)

Figure 9. (a) 13C CP-TOSS NMR spectra (υr ) 8 kHz) of samples of I with different water contents. (b) 1H MAS NMR spectra (υr ) 35 kHz) of the same samples. The peak at 3.8 ppm varies strongly with changes in water content. A shielded peak at 2.3 ppm appears in low water content samples (see text). All spectra were obtained at 273 K.

From eq 1, δiso for the Form 1 resonance is in the range of -6.3 to -7 ppm, depending on the value of η. Figure 10b shows the drastic change in the 23Na spectra that is observed upon dehydration as Form 1 transitions to Form 2 and then Form 3. An additional broad peak with a maximum at δexp ) -20.5 ppm appears next to the narrowed Form 1 line shape. This peak is assigned to sodium sites in Form 2 with a larger EFG (e.g., 2-3 MHz), caused by the removal of coordination partners O50 and O60 and the consequent loss of symmetry. An internal lattice shift of the sodium ions or a change in dynamics during

the phase changes may also play a role in this effect. In any case, a loss of coordination is expected to cause a shielding effect on 23Na chemical shifts, while the increase in χ causes an apparent shift in the same direction of similar magnitude.55 To help interpret the relative effects of these two possibilities on the data, density-functional theory (DFT) calculations of 23Na electric field gradients were used to analyze the loss of ionassociated water and the breaking of the EFG symmetry at the sodium nucleus. The results at the B3LYP/6-31G(d) level of theory are given in Table 4 for a cluster model containing a central sodium surrounded by the six nearest water molecules

δexp ) δiso -

(

)

25 000 η2 2 1+ χ 2 3 (ν0)

2346 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al. Table 4. Calculated B3LYP/6-31G(d) Quadrupolar Coupling Constants and Asymmetry Parameters for the 23Na Site in a Simple Form 1 Cluster Model with Different Water Environments

Figure 10. (a) 15N CP-MAS NMR spectra of samples of I with different water contents, obtained at υr ) 5 kHz. (b) 23Na MAS NMR spectra (υr ) 25 kHz) of the same samples. A second-order quadrupolar pattern is observed as the water content is reduced (see text). All spectra were obtained at 273 K.

and the coordinating sections of the parent molecule. EFG values were converted from atomic units to MHz using the conversion χ ) 9.717 × 1021 (eQ/h). The results are an approximation given that the structure of Form 2 differs from Form 1 but do indicate a modest increase in EFG that would result in a ∼7 ppm shift. Therefore, a significant contribution from chemical shielding is likely contributing to the observed trend. Analysis of the Vapor Exchange Process with Raman and SSNMR Spectroscopy. Although vibrational spectra are generally not as sensitive as PXRD or SSNMR to hydration state or phase changes, they can be obtained rapidly and are sensitive to isotopic substitution. FT-Raman has been used to study vapor exchange in channel hydrates using D2O to alter the frequencies

water site

χ (MHz)

η

ion-associated water (O50 and O60) missing H-bonded water (O40) missing all water present

1.8 1.3 1.3

0.57 0.30 0.35

of key vibrational modes.11 In an experiment designed to replicate the hydration process at ambient conditions, FT-Raman spectroscopy was used to monitor Form 1 material preequilibrated at 33% RH and subsequently exposed to D2O in a 75% RH chamber. Since the GVS results indicated less than a 0.5% w/w water content change between these two humidity points, it was expected that the exchange process would be fairly controlled and readily observed. FT-Raman monitoring during the 144-h exposure period showed this to be the case; ongoing vapor exchange leads to several clear changes in the spectra, shown in Figure 11a. Only a portion of the spectra are shown; the higher-frequency region exhibited a broad band at 2470 cm-1, representing the appearance of O-D stretching vibrations from incorporated water. The O-D band had a maximum signal-to-noise of 10:1 and consisted of several components with additional minor peaks at 2588 and 2655 cm-1 showing the presence of more than one O-D stretching vibration. These peaks were clearly observed in the FT-IR spectrum at the end of the 144-hour exposure (data not shown). Substitution with deuterium lowered the frequency of these vibrations by a factor of 0.7. With continued exposure, the strong Raman peak at 989 cm-1 increased at the same relative rate as the O-D stretch. This band can be tentatively assigned to either a water bending mode or possibly a rocking mode from an exchangeable position on the molecule, such as the NH2 group. The signal at 989 cm-1 was used to plot the uptake of deuterium by Form 1 shown in Figure 11b. The strongest band in the FT-Raman spectra (at 1589 cm-1) did not show a frequency or intensity change over the course of the vapor exchange experiment and was therefore used as an internal intensity reference for the uptake profile. (The relative intensity was determined by dividing the intensity of the band at 989 cm-1 by this band.) The other changes in the spectra were comparatively analyzed to check for a multistage process, but no clear evidence for this was obtained. Plots of the band intensity at 816 cm-1 or the weak O-D stretch at 2470 cm-1 led to the same trend. Several bands, such as the peak at 1040 cm-1, were found to diminish with increasing exposure, indicating the removal of hydrogen. The uptake of deuterium could not be normalized to a w/w water content by comparison to the recrystallized sample because the FT-Raman spectrum of the D2O-recrystallized sample shows evidence of “scrambling” of the deuterium over numerous exchanged sites, despite the agreement of its PXRD pattern with the original Form 1 reference.11 Besides providing an uptake profile, the FT-Raman spectra offer some additional insights into the vapor exchange process, the most notable of which is the shift in frequency of the carbonyl stretch as a function of exposure time. The multicomponent carbonyl stretching band begins the exposure experiment with a maximum at 1687 cm-1. Between 3 and 10 h, it splits into a second component at 1679 cm-1, followed by the disappearance of the first component at 1687 cm-1 at exposure times greater than 19 h. This is caused by a change in the carbonyl environment; the shift is probably the result of 2H substitution at either O40 or the nearby hydrogens on N7 and N9, or possibly a combination of both effects (as suggested by

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2347

Figure 11. (a) Effects of D2O vapor exchange at 75% RH on the FT-Raman spectra of a sample preequilibrated at 33% RH (H2O) as a function of exposure time. The spectrum of the D2O-recrystallized sample is also shown for comparison. (b) Plot of the relative intensity of the band at 989 cm-1 with increasing exposure to 75% RH D2O.

the more significant changes seen in the recrystallized sample). Another notable frequency shift occurs for a band at 1559 cm-1. This peak, which is tentatively assigned to an H-O-H scissoring vibration, shifts drastically over the same time period as the other observed changes. A split peak at 1545 and 1535 cm-1 eventually stabilizes after 19 h of exposure. This band can be interpreted as an exchange of more than one crystallographically distinct water molecule, in light of the multicomponent nature of the O-D stretching vibration. (An alternative explanation for the splitting of the signal, that mixed H/D incorporation has occurred, is also possible). Similar effects occur for bands that appear at 1459 and 1410 cm-1 in the initial spectrum of Form 1 and for several other bands in the spectra. The PXRD pattern of the D2O-exchaged sample was remeasured at the completion of the Raman study and was indistinguishable from the initial Form 1 trihydrate pattern. Exposure to D2O vapor loads a significant amount of deuterons directly into the Form 1 lattice without initiating any phase change. This behavior is fully consistent with a crystal structure containing

a large channel.11 The FT-Raman results also suggest that either deuterons or entire water molecules have penetrated down the side channels and have affected O40 and its neighbors (e.g., the two NH sites or the NH2). Nine distinct vibrational modes have been affected by 2H exchange, indicating that deuterium has been thoroughly incorporated into the lattice. The FT-Raman spectrum of the H217O-exchanged sample did not show any distinctive differences from the Form 1 spectrum. Since the effects of 17O exchange, while small, should be noticeable, it is concluded that full water molecules do not exchange very efficiently into the lattice, and that deuterium transfer plays the primary role in the uptake profile in Figure 11b. The 144-h D2O-exchanged sample was also studied by multinuclear MAS experiments. The 1H MAS spectrum (υr ) 35 kHz) of the exchanged sample in Figure 12a shows a diminishment of the O50 and O60 water signal at 3.8 ppm. This is fully expected from the exchange of these water protons. In addition, a shoulder appears at 6.6 ppm possibly indicating a loss of dipolar 1H-1H broadening from this site. The 13C CP-

2348 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al.

Figure 13. (a) 2H quadrupolar echo NMR spectra of D2O vaporexchanged Form 1, measured over a temperature range of 250 K. (b) 2 H MAS spectrum at 273 K (υr ) 10 kHz), and an expansion of the +1 sideband (arrow) to show chemical shift effects.

Figure 12. MAS NMR spectra of Form 1 compared to D2O-vapor exchanged Form 1 after 144 h of exposure, obtained at 273 K. In (a), the fast 1H MAS spectrum shows diminishment of O50/O60 proton signal and sharpening of a shoulder at 6.8 ppm (see text). In (b) and (c), 2H exchange of protons on O50 and O60, near in space to the sodium ion, reduces CP to C1, C2, C3 and the sodium ion in the 1H13 C CP-TOSS (υr ) 8 kHz) and the 1H-23Na CP-MAS (υr ) 25 kHz) spectra. In (d), diminishment of CP in the 1H-15N CP-MAS (υr ) 5 kHz) spectrum indicates exchange of the proton attached to N7.

TOSS spectrum (υr ) 8 kHz) of the exchanged sample, seen in Figure 12b, shows a noticeable loss of CP to C1 and C2. Since these sites border the main channel, loss of CP is expected from exchange of the water in the main channel. An overall loss of signal intensity estimated at about 30% was also encountered and was linked to a minor increase in 1H T1, caused by removal of the protons on O50 and O60, which act as efficient relaxation sinks. The exchange of the O50 and O60 protons also causes a 4-fold reduction in 1H-23Na CP without altering the 23Na line

shape, as shown in Figure 12c, in spectra obtained under identical experimental conditions. The crystal structure of Form 1 shows that the protons on the ion-associated O50 and O60 atoms are good sources of CP, with internuclear distances in the range of 2.80-3.06 Å, while the next nearest proton (attached to N1) is 3.67 Å distant, and no other protons are within 4 Å. No change in the direct polarization 23Na spectra was observed between Form 1 and D2O vapor-exchanged Form 1 (data not shown). Finally, the 15N spectrum of the D2O vaporexchanged sample in Figure 12d shows an interesting effect. Besides an overall loss of intensity for both the N7 and N9 signals relative to the Form 1 spectrum, a clear diminishment of signal for N7 is seen. This provides strong evidence that the attached proton has been exchanged, given the use of a 3-ms contact time and the paucity of protons near to this site (as seen in the short-contact CP-HETCOR experiment). The D2O-exchanged sample was also used to assess the environment and dynamics of the water molecules. For the spin-1 2H nucleus, the first-order quadrupolar interaction dominates the 2H NMR spectrum. The results of static 2H quadrupolar echo experiments over a 250 K temperature range are shown in Figure 13a. No sharp central peak is present as

Crystalline Pharmaceutical Hydrate

Figure 14. 2H NMR spectra of the D2O vapor-exchanged sample stored at 373 K for 40 h. The spectra show the loss of the narrowed component assigned to the mobile O50/O60 water. Between 30 and 46 h, little change was observed in the spectra, suggesting that the dehydration had largely completed. The spectrum at 40 h is assigned to exchanged ND2 (N1) sites as well as some exchanged N7 and O40 sites. At this temperature, after water escape has stabilized, the sample exists as Form 2.

would be the case for free water; a powder line shape is present at all temperatures, demonstrating that D2O vapor exchange has incorporated bound deuterons into the lattice. (This experiment is thus useful as a simple test to confirm suspected channel hydrate behavior in a pharmaceutical compound before a crystal structure is available.) The powder line shapes change significantly over the explored temperature range. At 123 K, the lowest temperature achieved, the line shape has nearly reached the rigid limit but still has some residual intensity near its center. At 173 K, motional averaging becomes apparent with the growth of intensity at the center of the line shape. These narrowed powder pattern line shapes are indicative of the motional models of the water molecules and exchanged sites, a topic discussed in detail below. The 2H chemical shift of this central narrowed component can be obtained from the room-temperature 2H MAS spectrum in Figure 13b and is seen to be about that of the narrowed signal at 3.8 ppm in the 1H spectrum. Combined with the exchange effects seen in the multinuclear MAS spectra, this assigns the narrow powder line shape at higher temperatures to deuterons attached to O50 and O60 water molecules. The chemical shift of the wider component is primarily caused by a peak at 9.9 ppm and is assigned to exchanged N1 hydrogen sites. Additional components such as O40 and N7 are not clearly visible and may be only partially exchanged or overlapped with the other components, such as the shoulder peak seen in the 2H MAS spectrum at 6 ppm. The VT PXRD study has shown that holding the sample at 373 K (100 °C) for a sufficient period should drive away at least the first mole of water and produce Form 2. This was tested within the NMR spectrometer during the course of 2H observation. Holding the sample at 373 K for 40 h led to the results shown in Figure 14. Given that the sample was contained in an unsealed 5-mm glass tube with material lightly packed around it, the escape depth of water is expected to differ significantly from that of the thinly spread PXRD sample. The dehydration rate is expected to be somewhat slower for the 2H NMR sample;

Crystal Growth & Design, Vol. 6, No. 10, 2006 2349

this was found to be the case. The narrowed central portion of the line shape disappears after 10 h as water escapes, further confirming that this signal is due to O50 and O60 deuterons and not from exchangeable sites on I. The residual spectrum after 10 h contains a central component that continues to decay very slowly. This could be remaining O50/O60 deuterons, an exchanged site or the 3rd mole of water (O40), with the slow decay rate tied to the slow desolvation of intact O40 molecules, or via a mechanism in Form 2 that slowly removes exchangeable sites through transfer to escaping water. The broad and still intense components of the final spectrum are assigned to exchanged sites on the molecule. Between 30 and 40 h, the line shape did not change significantly and the experiment was halted. With the narrowed portion of the 2H powder pattern at higher temperatures definitively assigned to the deuterons on O50 and O60, an analysis of the line shape to obtain information on the motional model adopted by these deuterons was conducted. It is assumed that the z-axis of the deuterium EFG tensor is aligned (or nearly aligned) with the O-D bond, as is almost always the case.56,57 The line shape at 123 K was fit to a spin-1 powder pattern with χ ) 208 kHz, η ) 0.05, and 15.7 kHz of Gaussian broadening. The broadening accounts for unresolved components from multiple water molecules, different hydrogenbonding environments for the individual deuterons on water, and the contributions from exchanged sites. The value of χ ) 208 kHz should thus be considered an average, but nevertheless is typical for hydrates with moderate-strength hydrogen bonds.57 Since the line shape at 123 K could be fit to a spin-1 pattern with a reasonable χ and η, the line shape is considered to be in the slow motion range as it would appear in a rigid lattice, with a motional correlation time greater than 10-3 s. When motional averaging of a 2H EFG tensor occurs, a complete average is typically achieved when the correlation time of the motion is shorter than 10-7 s. In this fast limit, the spectrum is defined by an effective EFG tensor, which is an average of the rigid tensors at each point in the geometry of the motion. In the intermediate range, which typically occurs between correlation times of 10-3 and 10-7 s, the static 2H line shape is partially averaged and information about the rate of motion can be extracted.58 Unique line shapes are observed in this region that cannot be modeled by an effective first-order quadrupolar tensor. However, in the present case, the narrowed central portion of the spectra in Figure 13a at temperatures greater than 273 K can be fit using an effective χ and η, indicating that the analysis of the narrowed central pattern can be accomplished in the fast motion limit. Intermediate range behavior may occur between 173 and 223 K but is difficult to confirm because of spectral overlap. The environment of the water in the Form 1 crystal structure, shown in Figure 15a, is helpful for understanding the type of fast motion seen in the 2H spectra. The water “stack” is shown with the sodium cations and the hydrogen bonding interactions. One set of interactions reaching outside the main channel to O40 and N7 is also depicted. Figure 15a shows the similarity between O50 and O60 in terms of hydrogen bonding, as one hydrogen on each is engaged in a clearly-defined interaction while the other is not. With this in mind, the potential motional models can be considered. The most commonly used model in crystalline hydrates is rotation of the water molecule about its C2 axis (bisecting the H-O-H bond angle) through 180° jumps.56,57,59 In the fast limit, this type of motion results in patterns of similar overall width to those obtained in the rigid limit, even when the H-O-H angle deviates from the ideal

2350 Crystal Growth & Design, Vol. 6, No. 10, 2006

Vogt et al.

fast limit would appear as trigonal jumps over three positions on a tetrahedron defined by the H-O-H angle θ and the swing angle φ, with an unoccupied fourth position. To model the line shape under fast trigonal jumps, the 2H EFG tensor must be rotated from its principal axis system to three averaged positions. The first-order quadrupolar Hamiltonian expressed in terms of irreducible spherical tensor coordinates is:54

H(1) Q )

x6 2 eQ [3Iz - I(I + 1)] Vpas 20 4I(2I - 1)p 3

(2)

where V pas 20 is a component of the quadrupolar tensor in its principal axis system and Iz is a spin operator. Rotations of the real space portion of this interaction into a molecular trigonal jump frame can then be accomplished using Wigner rotations (D) around an angle representing the H-O-H angle (θ) and an angle φ representing the bottom face of the triangle (the swinging portion of the water molecule), as shown in Figure 15b: +2

mol ) V 2m

k Vpas ∑ 2q Dqm q ) -2

(3)

After this step, a second rotation of the form of eq 3 is employed to bring the system into the laboratory frame using a powder average over polar (R) and azimuthal (β) angles. The resulting expression for the real space component of the Hamiltonian defines the line shape (where η has been set to 0 for simplicity): 2 V lab 20 ) (3/16)χ([1 + 3 cos 2β]cos θ +

2 cos φ cos[2R + φ]sin2β sin2 θ 2 cos[R + (φ/2)] cos(φ/2) sin 2β sin 2θ) (4) Figure 15. (a) A view of the water channel from the Form 1 crystal structure showing the hydrogen bonding environment of O50 and O60; one hydrogen on each participates in the O50-H50A‚‚‚N1 or O60H60A‚‚‚Cl1 interactions, while the others are in assumed positions. One aryl ring from the parent is shown with the hydrogen bonding network from N1 to O40 and N7. (b) 2H static line shapes simulated from a three-site exchange of 2H as a function of the angles θ and φ (see text) compared with the experimental patterns, reflecting jumps around the O50-H50A...N1 or O60-H60A...Cl1 interactions combined with C2-like jump motion.

tetrahedral angle of 109.5°. Therefore, while reasonable in light of the crystal structure, a C2 jump model is unlikely given the NMR data. Similarly, 3-fold rotational models (similar to those seen in NH3 groups) also do not usually produce the narrowing effect seen in the present case, although they do allow for a wider range of effective χ and η.57,60 A model involving tetrahedral jumps is also possible, and can also lead to very narrow line shapes with a variety of effective η values.61 In the case of Form 1, a physically reasonable model can be constructed from a combination of 3-fold “trigonal” jumps over equally weighted positions on a skewed tetrahedron with the fourth position unused, as shown schematically in Figure 15b. The trigonal jump model is reasonable in the context of the O50/O60 hydrogen bonding network because a hydrogen on each water is “tethered” in a clearly-defined hydrogen bond interaction. The other hydrogen could then “swing” over an angle φ between two positions, as it reaches extrema defined by electrostatic repulsion from another water molecule or the sodium ion. In addition, C2-like jumps involving the tethered hydrogen would still occur; so that the combined picture in the

eq 4 is then evaluated over a spherical powder average of the angles R and β. Using this model, and the previously determined rigid χ of 208 kHz, a variety of different spectra were simulated as a function of θ and φ. The angle θ in crystalline water donating two hydrogen bonds is known to prefer the range of 102.0 to 114.6° (with a mean of 107.7°) from neutron diffraction studies, and thus can be restricted for the present analysis.62 In contrast, the “swing” angle φ can be varied over a wide range of 10° to 170°. Only the range of φ ) 70 to 115° and θ ) 103 to 106° offered line shapes narrow enough to match those observed at the center of the 2H spectra between 273 and 373 K. Within this subset of angles, the line shapes simulated at φ ) 83° and θ ) 106° were the only combination found to reproduce the experimental results at 273 K. As the temperature is raised, the spectrum remains in the fast motion limit but the angles involved in the motion begin to deviate. At 323 K, two components (probably O50 and O60) contributing to the line shape are separately resolved as seen in Figure 13a. Heating to 373 K causes the components to again overlap. The spectrum at this temperature is modeled by a simulation with at φ ) 90° and θ ) 103° as seen in Figure 15b. This indicates a compression of H-O-H water angle as the lattice dehydrates. The fast trigonal jump model thus accurately reproduces the observed 2H patterns, although more complex models are possible. Proton transfer in the main channel is not considered in the simple model used here; however, the effects of such a process would be averaged 2H χ values over different environments. A final set of experiments were performed on the 75% RH H217O-exchanged sample, to learn if vapor exchange of entire water molecules was possible and if so, to study the environment

Crystalline Pharmaceutical Hydrate

Crystal Growth & Design, Vol. 6, No. 10, 2006 2351

by evaluation of the standard expression over a range of powder angles R and β, in the same manner as for 2H:54 (2) ω-1/2,1/2 )-

[

]

2 3χ 1 [I(I + 1) - (3/4)] × 6ω0 2I(2I - 1)

[A(R,η) cos β + B(R,η) cos2 β + C(R,η)] (5) 4

where functions A, B, and C are given by:

A(R,η) ) - (27/8) + (9/4)η cos 2R - (3/8)(η cos 2R)2 B(R,η) ) (30/8) - (1/2)η2 - 2η cos 2R + (3/4)(η cos 2R)2 C(R,η) ) -(3/8) + (1/3)η2 - (1/4)η cos 2R (3/8)(η cos 2R)2 (6)

Figure 16. (a) 17O spin-echo NMR spectra of H217O vapor-exchanged Form 1, acquired over a temperature range of 200 K. (b) Simulation of the line shape at 123 K from the sum of two second-order centraltransition quadrupolar patterns of equal intensity with η ) 0.95, χ ) 7.7 MHz and η ) 0.54, χ ) 8.2 MHz. The two components are assigned to exchanged O50 and O60 water molecules.

of the water oxygen. Although FT-Raman did not detect incorporation of 17O-labeled water in this sample, 17O static spin-echo NMR spectra of bound water were obtained as shown in Figure 16a, although an 18 h experiment was needed. The comparative signal-to-noise of the 17O and 2H spectra is consistent with the FT-Raman results and is a consequence of both the lower isotopic enrichment and the very limited exchange achieved by “intact” water molecules. The results demonstrate the ability of water molecules to enter the Form 1 lattice at ambient conditions. The 17O central-transition line shape is broadened by second-order quadrupolar effects from bound water. The spectrum at 323 K consists of a motionally averaged component centered at 0 ppm and a broad component with an approximate χ of 8 MHz and η ≈ 1. The 17O line shape changes as the temperature is reduced until a second peak appears at 123 K, in agreement with the 2H VT results, although motional averaging may not be fully quenched at this temperature given the wide frequency span of the 17O signal. A simulated second-order quadrupolar line shape can be used to fit the rigid spectrum at 123 K. This line shape is obtained

and the spin I ) 5/2 for 17O. In practice, eq 5 is also convoluted by a line-broadening function. The use of eq 5 for 17O line shape simulation results in an idealized spectrum; potential nutation effects from finite RF power that could cause quantitative differences are not considered. The 17O powder pattern for Form 1 at 123 K is well-reproduced by the composite simulated pattern shown in Figure 16b. This was created from the sum of two equally contributing patterns, one with χ ) 7.7, η ) 0.95 MHz and the other with χ ) 8.2 MHz, η ) 0.54. The two components of the simulated line shape are also shown in Figure 16b. On the basis of the findings of the other NMR experiments, these two nonequivalent sites are assigned to O50 and O60. Nevertheless, the 123 K pattern is reproduced extremely well by this simple two-component model without any average over rotated positions, suggesting that the water molecules are truly rigid at this temperature from the point-of-view of both 2H and 17O NMR. It should be noted that while 2H first-order quadrupolar line shapes can be simulated with an effective η value, this is not the case for second-order line shapes.63 The two components of the 17O line shape reflect different environments for O50 and O60. The quadrupolar parameters for water in hexagonal ice are reported to be χ ) 6.66 ( 0.10 MHz and η ) 0.935 ( 0.01.64 As the oxygen atom in water accepts weaker hydrogen bonds, the value of χ generally increases, which is consistent with the ∼8 MHz couplings in Form 1. The variation of η between the two components is especially interesting. The 17O spectra of most hydrates show η values of 0.85 to 1, but lower values of η are possible in some cases.65 The H-O-H angle for the O50 water is not accurately measured by SCXRD. However, a fairly typical angle between O50 and its hydrogen bond acceptors (O‚‚‚Ow‚‚‚O) of 102.6° is observed.62 This site is therefore assigned to the η ) 0.95, χ ) 7.7 MHz pattern. The O60 site is then assigned to the more unusual pattern with η ) 0.54, χ ) 8.2 MHz. In the crystal structure, the angle between the hydrogen bond acceptors and O60 (Cl1-O60-O60) is narrowed to 74.9°. This would be expected to compress the H-O-H angle (for example to ∼103°) and cause a lower η value given the theoretical relationship between η and H-O-H angle.66 The unusual η value for O60 might also be caused by electronic effects on the oxygen environment (from differences in hydrogen-bonding partners for the two water sites), or by residual contributions from motional averaging in the intermediate regime. Interestingly, a low η of about 0.5 was observed for water in another channel hydrate system.25 Discussion Because Form 1 loses 0.5% w/w water without a phase change between 15% and 90% RH, it can be classified as an

2352 Crystal Growth & Design, Vol. 6, No. 10, 2006

isomorphic desolvate over this humidity range. This behavior is empirically explained by the observation of water tunnels in the crystal structure. However, further explanation is needed for the two distinct phase changes observed as the humidity drops below 15% RH or the temperature exceeds 60 °C. First, it is apparent from the conservation of features in the Raman and 13C SSNMR spectra of Forms 1 and 2 that short-range structural similarities (especially with the parent molecule) are maintained even as changes in long-range order alter the PXRD pattern. The differences between the 15N spectra of Form 1 and Form 2 are related to changes involving the N7-H7N...O40 hydrogen bond and the O40 water molecule. The significant differences in the PXRD pattern of Form 2 relative to Form 1 are also manifested in the 23Na NMR spectra and the sodium environment, which shows that the dehydration begins in the larger channel (with O50 or O60). The 23Na results, combined with the sensitivity of PXRD to long-range order, the hystersis in the GVS isotherm, and the lower temperature endotherm in the DSC, suggest a partial collapse of the main channel. The appearance of shielded water at 2.1 ppm in the 1H spectrum of Form 2 (which has lost its hydrogen bond) shows that the water remaining in channel is not strongly interacting with the host lattice, which offers an explanation for the metastable character of Form 2 at ambient conditions. In contrast to the relatively minor changes between Forms 1 and 2, Form 3 shows evidence of a significant collapse affecting the structure of the parent molecule, the cation, and the water. After the removal of ion-associated water in the main channel, further dehydration and a second phase change have a large effect on the 15N spectrum; N7 and to a lesser extent N9 are shifted as the final water (O40), hydrogen bonded to these atoms, is pulled away. From the 13C SSNMR and FT-Raman spectra, Form 3 is found to have Z′ g 2, a semicrystalline nature, and also appears to be missing the hydrogen bond into C8 (again, resulting from loss of O40). The assertion that Form 3 is a metastable semicrystalline phase is supported by its low DSC melt and the uptake of additional temporary water seen in the 45 °C GVS drying experiment. Form 3 shows weaker intensity and generally broader reflections in its PXRD pattern and has lost its crystallographic symmetry along with most of its hydrogen bonds as the remnants of Form 1 and 2 collapse. However, the ability of Form 3 to rehydrate suggests that key intermolecular arrangements are still maintained, as seen in some of the spectra. The MAS experiments on D2O-exchanged Form 1 trace the path of exchanged water through the lattice by the loss of CP. Although all three moles of water can be forcibly removed (with O40 being the final mole lost), the exchange experiments also show that the protons on O40 and a neighboring amide proton can be exchanged with deuterons without the need for dehydration. This can be rationalized as shown in Figure 15a, where N1 acts as a “bridge” between O50 and O40. The dynamics of the NH2 motion, detected in the 15N CP-MAS spectra, are probably the key to eventual hydrogen transfer to O40 and N7 through the hydrogen-bonding chain of O50, N1, O40, and N7. The interesting 2H motional model observed in the main channel, modeled by rapid three-site trigonal jump motion of deuterons between 273 and 373 K, offers an explanation for hydrogen transfer through the main channel. The model also suggests that concerted motion of several water molecules would occur; for example, as a hydrogen on O60 swings over the angle φ, a neighboring hydrogen on O50 would also move, and so on. When combined with proton transfer, this concerted motion helps explain the efficient deuterium incorporation network in

Vogt et al.

Form 1. An orthogonal view is provided by the 17O-exchange experiments, which show that the main channel is much less accessible for the exchange of whole water molecules. The two mechanisms (proton transfer and water transfer) have different time scales and efficiencies, as evidenced by the FT-Raman results and the relative intensities of the 2H and 17O NMR spectra. The limited isomorphic hydration of Form 1 between 15 and 75% RH is probably related to the low 17O exchange efficiency; that is, the small amount of water that can be pulled from the Form 1 lattice without causing a phase change is likely the same water exchanged in the 17O experiment. Intact water might penetrate to a moderate depth from the surface of individual crystals, after which the hydrogen exchange mechanism rapidly distributes deuterons through the lattice. Further work would be needed to confirm this hypothesis, such as NMR diffusion experiments on a single crystal. Finally, it should be noted that the physical properties of Form 1, while complex, do not prevent its development as a pharmaceutical compound. Many marketed drugs behave in a similar fashion. The reversibility of the phase transitions bodes well for drying, although careful process engineering would still be needed to successfully handle Form 1 to avoid producing Forms 2 and 3. This could include humidified drying of the drug substance at manufacturing scale, careful particle size reduction if needed, and solid-state monitoring of drug product blending, tableting and stability. Conclusions The crystal structure of sodium N-(3-(aminosulfonyl)-4chloro-2-hydroxyphenyl)-N′-(2,3-dichlorophenyl) urea trihydrate (Form 1) contains a network of water tunnels. The water content of the crystalline powder varies over a wide range with complex behavior below 15% RH and above 50 °C. VH and VT PXRD experiments and SSNMR and FT-Raman spectroscopy of samples at reduced water levels showed an intricate dehydration process with two distinct reversible phase changes to the subdihydrate Form 2 and the nearly anhydrous semicrystalline Form 3. The structures of these phases as observed by NMR and vibrational spectroscopy were compared with the Form 1 crystal structure. Form 2 is structurally similar to Form 1, with changes limited to the loss of water molecules in the main channel and effects on the dynamics of the remaining water, and minor effects on the water in the smaller channel. Form 3 is a collapsed, disordered structure with Z′ g 2. Its structure is marked by missing hydrogen bonds, low EFG symmetry around the sodium (caused by loss of water), and broad NMR resonances characteristic of semicrystalline content, all of which help explain its metastable character. Monitoring four NMRactive nuclei during the desolvation process provided complementary information about the structure and information beyond that available from diffraction methods. In particular, 23Na NMR was found to be a sensitive probe of hydration state in Form 1, and would be capable of detecting this directly in pharmaceutical formulations at low drug loads. The network of hydrogen bonds between channel water and the parent molecule in Form 1 also leads to interesting exchange processes that were investigated by FT-Raman and NMR spectroscopy through water vapor exchange at higher humidity, to determine an uptake time scale and the distribution of deuterium relative to the parent molecules. A three-site fast jump model was used to model the dynamics of the channel water as observed by 2H NMR. Exchange of 17O-labeled water confirmed that entire water molecules can enter the main channel, albeit to only a small extent, while individual deuterons can transfer much more efficiently inside

Crystalline Pharmaceutical Hydrate

the lattice. The use of a combined characterization approach improves the understanding of channel hydrate systems and their physical properties, including vapor sorption and phase transition behavior. Acknowledgment. The authors thank Dr. Alan Benesi (Department of Chemistry, The Pennsylvania State University, USA) for assistance with the quadrupolar line shape simulations. Supporting Information Available: An X-ray crystallographic information file (CIF) for Form 1 of sodium N-(3-(aminosulfonyl)-4chloro-2-hydroxyphenyl)-N′-(2,3-dichlorophenyl) urea trihydrate is available free of charge via the Internet at http://pubs.acs.org.

References (1) Bernstein, J. Polymorphism in Molecular Crystals; Oxford University Press: New York, 2002. (2) Byrn, S. R.; Pfeiffer, R. R.; Stowell, J. G. Solid-State Chemistry of Drugs; SSCI: West Lafayette, IN, 1999. (3) Morris K. R.; Rodriguez-Hornedo, N. In Encyclopedia of Pharmaceutical Technology; Swarbrick, J., Boylan, J. C., Eds; Marcel Dekker: New York, 1993; Vol. 7, pp 393-440. (4) Yu, L. AdV. Drug DeliVery ReV. 2001, 48, 27-42. (5) Perrier, P. R.; Byrn, S. R. J. Org. Chem. 1982, 47, 4671-4676. (6) Khankari, R. K.; Law, D.; Grant, D. J. W. Int. J. Pharm. 1992, 82, 117-127. (7) Stephenson, G. A.; Stowell, J. G.; Toma, P. H.; Xu, W.; Pfeiffer, R. R.; Byrn, S. R. J. Pharm. Sci. 1997, 86, 1239-1244. (8) Stephenson, G. A.; Groleau, E. G.; Kleeman, R. L.; Xu, W.; Rigsbee, D. R. J. Pharm. Sci. 1998, 87, 536-542. (9) Reutzel, S. M.; Russell, V. A. J. Pharm. Sci. 1998, 87, 1568-1571. (10) Chen, L. R.; Young, V. G.; Lechuga-Ballesteros, D.; Grant, D. J. W. J. Pharm. Sci. 1999, 88, 1191-1200. (11) Ahlqvist, M. U. A.; Taylor, L. S. Int. J. Pharm. 2002, 241, 253261. (12) Te, R. L.; Griesser, U. J.; Morris, K. R.; Byrn, S. R.; Stowell J. G. Cryst. Growth Des. 2003, 3, 997-1004. (13) Reutzel-Edens, S. M.; Kleemann, E. L.; Lewellen, P. L.; Borghese, A. L.; Antione, L. J., J. Pharm. Sci. 2003, 92, 1196-1205. (14) Apperly, D. C.; Basford, P. A.; Dallman, C. I.; Harris, R. K.; Kinns, M.; Marshall P. V.; Swanson, A. G. J. Pharm. Sci. 2005, 94, 516523. (15) Vogt, F. G.; Dell’Orco, P. C.; Diederich, A. M.; Su, Q.; Wood, J. L.; Zuber, G. E.; Katrincic, L. M.; Mueller, R. L.; Busby, D. J.; DeBrosse, C. W. J. Pharm. Biomed. Anal. 2006, 40, 1080-1088. (16) Kennedy, A. R.; Okoth, M. O.; Sheen, D. B.; Sherwood, J. N.; Teat, S. J.; Vrcelj, R. M. Acta Crystallogr. 2003, C59, 650-652. (17) Go¨rbitz, C. H. Acta Crystallogr. 2004, C60, 810-812. (18) Bauer, J. F.; Dziki, W.; Quick, J. E. J. Pharm. Sci. 1999, 88, 12221227. (19) Byrn, S. R.; Sutton, P. A.; Tobias, B.; Frye, J.; Main, P. J. Am. Chem. Soc. 1988, 110, 1609-1614. (20) Jeffrey, G. A. An Introduction to Hydrogen Bonding; Oxford University Press: New York, 1997. (21) McKeown, N. B.; Makhseed, S.; Msayib, K. J.; Ooi, L. L.; Helliwell, M.; Warren, J. E. Angew. Chem. Int. Ed. 2005, 44, 7546-7549. (22) Podolin, P. L.; Bolognese, B. J.; Foley, J. J.; Schmidt, D. B.; Buckley, P. T.; Widdowson, K. L.; Jin, Q.; White, J. R.; Lee, J. M.; Goodman, R. B.; Hagen, T. R.; Kajikawa, O.; Marshall, L. A.; Hay, D. W. P.; Sarau, H. M. J. Immunol. 2002, 169, 6435-6444. (23) Tishmack, P. A.; Bugay, D. E.; Byrn, S. R. J. Pharm. Sci. 2003, 92, 441-474. (24) Brown, S. P.; Spiess, H. W. Chem. ReV. 2001, 101, 4125-4155. (25) Ueda, T.; Nakamura, N. Z. Naturforsch. 2000, 55a, 362-368. (26) Nyqvist, H. Int. J. Pharm. Technol. Prod. Manuf. 1983, 4, 47-48. (27) The crystal structure data for sodium N-(3-(aminosulfonyl)-4-chloro2-hydroxyphenyl)-N′-(2,3-dichlorophenyl) urea trihydrate has been deposited with the Cambridge Crystallographic Data Center under deposition number CCDC 616424. This material can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, by emailing [email protected], or by contacting the Cambridge Crystallographic Data Centre, 12, Union Road, Cambridge CB2 1EZ, UK.; fax: +44 1223 336033. (28) Stout, G. H.; Jensen, L. H.; X-ray Structure Determination. A Practical Guide; Wiley: New York, 1989. (29) Sheldrick, G. M., SHELXS-97 and SHELXL-97, 1997.

Crystal Growth & Design, Vol. 6, No. 10, 2006 2353 (30) Braun, S.; Kalinowski, H. O.; Berger, S. 150 and More Basic NMR Experiments, 2nd ed.; New York: Wiley-VCH, 1998. (31) Harris, R. K.; Becker, E. D.; Cabral de Menezes, S. M.; Goodfellow, R.; Granger, P. Pure Appl. Chem. 2001, 73, 1795-1818. (32) Earl, W. L.; Vanderhart, D. L. J. Magn. Reson. 1982, 48, 35-54. (33) Ratcliffe, C. I.; Ripmeester, J. A.; Tse, J. S. Chem. Phys. Lett. 1983, 99, 177-180. (34) Metz, G.; Wu, X.; Smith, S. O. J. Magn. Reson. A 1994, 110, 219227. (35) Antzutkin, O. N. Prog NMR Spectrosc. 1999, 35, 203-266. (36) Fung, B. M.; Khitrin, A. K.; Ermolaev, K. J. Magn. Reson. 2000, 142, 97-101. (37) Opella, S. J.; Frey, M. H. J. Am. Chem. Soc. 1979, 101, 5854-5856. (38) van Rossum, B. J.; Fo¨rster, H.; de Groot, H. J. M. J. Magn. Reson. 1997, 124, 516-519. (39) Lesage, A.; Charmont, P.; Steuernagal, S.; Emsley, L. J. Am. Chem. Soc. 2000, 122, 9739-9744. (40) Davis, J. H.; Jeffrey, K. R.; Bloom, M.; Valic, M. I.; Higgs, T. P. Chem. Phys. Lett. 1976, 42, 390-394. (41) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03, ReVision C.02. Gaussian, Inc., Wallingford CT, 2004. (42) Schwerdtfeger, P.; Pernpointner, M.; Nazarewicz, W.; Calculation of Nuclear Quadrupole Coupling Constants. In Calculation of NMR and EPR Parameters; Kaupp, M., Bu¨hl, M., Malkin, V. G., Eds.; Wiley-VCH: Weinheim, 2004; pp 279-292. (43) Post, J. E.; Bish, D. L. Rietveld refinement of crystal structures using powder X-ray diffraction data. In Modern Powder Diffraction (ReViews in Mineralogy, Vol. 20); Bish, D. L., Post, J. E., Eds.; Mineralogical Society of America: Washington, DC, 1989; pp 277306. (44) Dollase, W. A. J. Appl. Crystallogr. 1986, 19, 267-272. (45) Jeffrey, G. A.; Yeon, Y., Acta Crystallogr. 1986, B42, 410-413. (46) Harris, R. K.; Olivieri, A. C. Prog. NMR Spectrosc. 1992, 24, 435456. (47) Stothers, J. B. Carbon-13 NMR Spectroscopy; Academic Press: New York, 1972; p 299. (48) Stephenson, G. A.; Stowell, J. G.; Toma, P. H.; Dorman, D. E.; Greene, J. R.; Byrn, S. R. J. Am. Chem. Soc. 1994, 116, 57665773. (49) Vogt, F. G.; Cohen, D. E.; Bowman, J. D.; Spoors, G. P.; Zuber, G. E.; Trescher, G. A.; Dell’Orco, P. C.; Katrincic, L. M.; DeBrosse, C. W.; Haltiwanger, R. C. J. Pharm. Sci. 2005, 94, 651-665. (50) Naito, A.; Tuzi, S.; Saito, H. Eur. J. Biochem. 1994, 224, 729. (51) Saito, H.; Tanaka, Y. Nukada, K. J. Am. Chem. Soc. 1971, 93, 10771081. (52) Saito, H.; Nukada, K. J. Am. Chem. Soc. 1971, 93, 1072-1076. (53) Almond, G. A.; Harris, R. K.; Franklin, K. R.; Graham, P. J. Mater. Chem. 1996, 6, 843-847. (54) Man, P. P. In Encyclopedia of Nuclear Magnetic Resonance; Grant, D. M., Harris, R. K., Eds.; New York, Wiley: 1996; pp 4581-4591. (55) Koller, A.; Engelhardt, G.; Kentgens, A. P. M.; Sauer, S. J. Phys. Chem. 1994, 98, 1544. (56) Weiss, A.; Weiden, N. In AdVances in Nuclear Quadrupole Resonance; Smith, J. A. S., Ed.; Heyden & Son Ltd.: London, UK, 1980; Vol. 4, pp 149-248. (57) Barnes, R. G. In AdVances in Nuclear Quadrupole Resonance; Smith, J. A. S., Ed.; Heyden & Son Ltd.: London, UK, 1974; Vol. 1, pp 335-355. (58) Wittebort, R. J.; Olejniczak, E. T.; Griffin, R. G. J. Chem. Phys. 1987, 86, 5411-5420. (59) Larsson, K.; Tegenfeldt, J.; Hermansson, K. J. Chem. Soc., Faraday Trans. 1991, 87, 1193-1200.

2354 Crystal Growth & Design, Vol. 6, No. 10, 2006 (60) Long, J. R.; Sun, B. Q..; Bowen, A.; Griffin, R. G. J. Am. Chem. Soc. 1994, 116, 11950-11956. (61) Benesi, A. J.; Grutzeck, M. W.; O’Hare, B.; Phair, J. W. J. Phys. Chem. B 2004, 108, 17783-17790. (62) Steiner, T. Acta Crystallogr. B 1998, 54, 464-470. (63) Kristensen, J. H.; Farnan, I. J. Chem. Phys. 2001, 114, 9608-9624.

Vogt et al. (64) Spiess, H. W.; Garrett, B. B.; Sheline, R. K.; Rabideau, S. W. J. Chem. Phys. 1969, 51, 1201-1205. (65) Gosling, P.; Rabbani, S. R. J. Mol. Struct. 1987, 158, 89-97. (66) Vega, S. J. Chem. Phys. 1974, 60, 3884.

CG060324K