II. EXPERIMENTAL

Salmeterol xinafoate was a commercial reagent (>97% purity), purchased from Key Organics Limited (batch 74 745), and was used as-received. The white powder was packed into a 1.5 mm diameter Kapton capillary and rotated during the measurement at ~50 cycles s⁻¹. The powder diffraction pattern was measured at 295 K at beam line 11-BM (Lee et al., Reference Lee, Shu, Ramanathan, Preissner, Wang, Beno, Von Dreele, Ribaud, Kurtz, Antao, Jiao and Toby2008; Wang et al., Reference Wang, Toby, Lee, Ribaud, Antao, Kurtz, Ramanathan, Von Dreele and Beno2008) of the Advanced Photon Source at Argonne National Laboratory using a wavelength of 0.413 891 Å from 0.5°2θ  to 50°2θ with a step size of 0.001° and a counting time of 0.1 s step⁻¹. The pattern was indexed on a primitive triclinic unit cell having a = 9.174, b = 9.481, c = 21.732 Å, α = 82.3, β = 85.2, γ = 62.2°, V = 1628.3 Å³, and Z = 2 using Jade 9.5 (MDI, 2014). Assuming Z = 2 yields an atomic volume of 18.5 Å³ atom⁻¹ for the 88 non-H atoms in the unit cell, and a reasonable calculated density of 1.225 g cm⁻³. Since commercial material is a racemate, the space group was assumed to be P−1 (#2), which was confirmed by successful solution and refinement of the structure. A reduced cell search in the Cambridge Structural Database (Allen, Reference Allen2002) combined with the chemistry “C H N O only” yielded 48 hits, but no structure for salmeterol xinafoate. A name search on “salmeterol” yielded no hits, as did a connectivity search on a salmeterol molecule.

A salmeterol cation and a xinafoate anion were built and their conformations optimized using Spartan ‘14 (Wavefunction, 2013), and saved as mol2 files. Manual intervention was needed to keep the alkyl chains in all-trans conformations. These files were converted into Fenske–Hall Z-matrix files using OpenBabel (O'Boyle et al., Reference O'Boyle, Banck, James, Morley, Vandermeersch and Hutchison2011). Many attempts to solve the structure with FOX (Favre-Nicolin and Černý, Reference Favre-Nicolin and Černý2002) and DASH (David et al., Reference David, Shankland, van de Streek, Pidcock, Motherwell and Cole2006) using these two fragments yielded solutions with molecular overlap and voids. Some contained linear salmeterol and others yielded bent conformations. Direct methods using EXPO2013 (Altomare et al., Reference Altomare, Cuocci, Giacovazzo, Moliterni, Rizzi, Corriero and Falcicchio2013) suggested relatively linear arrays of atoms, but did not yield enough of the structure to permit manual completion.

It would be very surprising if the positively-charged NH₂ group of the salmeterol cation and the ionized carboxylate group of the xinafoate anion did not participate in strong N–H···O hydrogen bonds. Accordingly, the salmeterol and xinafoate fragments were oriented so that the N···O distance was 2.80 Å and a hydrogen bond was linear. This was done for both carboxylate oxygens and both N–H hydrogens, and one arrangement of the four yielded a much better fit to the pattern. The N–H hydrogen was removed, and replaced by a Li atom at the midpoint of the N···O vector, to tie the two fragments into one. This “superfragment” was used to solve the structure with FOX. The maximum sinθ/λ used in the solution was 0.25 Å⁻¹ (d _min = 2.00 Å). Because the predicted morphology of most trial models was platy, with {001} as the principal faces, a March–Dollase preferred orientation model (unique axis = 001) was included in the structure solution.

Rietveld refinement was carried out using General Structure Analysis System (GSAS) (Toby, Reference Toby2001; Larson and Von Dreele, Reference Larson and Von Dreele2004). Only the 1.0–20.0° portion of the pattern was included in the refinement (d _min = 1.19 Å). All non-H bond distances and angles were subjected to restraints, based on a Mercury/Mogul Geometry Check (Bruno et al., Reference Bruno, Cole, Kessler, Luo, Motherwell, Purkis, Smith, Taylor, Cooper, Harris and Orpen2004; Sykes et al., Reference Sykes, McCabe, Allen, Battle, Bruno and Wood2011) of the molecule. The Mogul average and standard deviation for each quantity were used as the restraint parameters. Planar restraints were applied to the two benzene rings and the naphthalene ring system. The restraints contributed 4.88% to the final χ ². Isotropic displacement coefficients were refined, grouped by chemical similarity. The hydrogen atoms were included in calculated positions, which were recalculated during the refinement using Materials Studio (Accelrys, 2013). The U _iso of each hydrogen atom was constraint to be 1.3× that of the heavy atom to which it is attached. The peak profiles were described using profile function #4 (Thompson et al., Reference Thompson, Cox and Hastings1987; Finger et al., Reference Finger, Cox and Jephcoat1994), which includes the Stephens (Reference Stephens1999) anisotropic strain broadening model. The background was modeled using a 3-term shifted Chebyshev polynomial, with a 4-term diffuse scattering function to model the Kapton capillary and any amorphous component. The refinement yielded the residuals R _wp = 0.1199, R _p = 0.1011, and χ²  = 2.822. Re-starting the Rietveld refinement from the density functional theory (DFT)-optimized model led to a model yielding lower residuals (R _wp = 0.1075 and χ²  = 2.296), but with a different chain conformation at C21–C24, a different orientation of the hydroxymethyl group C7–O8, and a slightly different conformation around N13. A new DFT calculation indicated that this model was 49 kcal mole⁻¹ lower in energy than the first model. The final refinement was started from this second DFT model.

The final refinement of 165 variables using 19 116 observations (18 999 data points and 117 restraints) yielded the residuals R _wp = 0.1035, R _p = 0.0873, and χ²  = 2.120. The largest peak (0.21 Å from N13) and hole (1.19 Å from O44) in the difference Fourier map were 0.26 and −0.24 eÅ⁻³, respectively. The Rietveld plot is included as Figure 2. The largest errors are in the shapes of the low-angle peaks (particularly the strong 001 peak), and may indicate subtle changes in the sample during the measurement.

Figure 2. (Colour online) The Rietveld plot for the refinement of salmeterol xinafoate. The black crosses represent the observed data points, and the red line is the calculated pattern. The blue curve is the difference pattern, plotted at the same vertical scale as the other patterns, and the green line is the background. The vertical scale has been multiplied by a factor of 20 for 2θ > 7.0°, and by a factor of 50 for 2θ > 13.0°.

A density functional geometry optimization (fixed experimental unit cell) was carried out using CRYSTAL14 (Dovesi et al., Reference Dovesi, Saunders, Roetti, Orlando, Zicovich-Wilson, Pascale, Civalleri, Doll, Harrison, Bush, D-Arco, Llunell, Causà and Noël2014). The basis sets for the H, C, N, and O atoms were those of Gatti et al. (Reference Gatti, Saunders and Roetti1994). The calculation was run on eight 2.1 GHz Xeon cores (each with 6 Gb RAM) of a 304-core Dell Linux cluster at IIT, used eight k-points and the B3LYP functional, and took ~32 h.

III. RESULTS AND DISCUSSION

The powder pattern corresponds to that of salmeterol xinafoate Form I (Beach et al., Reference Beach, Latham, Sidgwick, Hanna and York1999; Tong et al., Reference Tong, Shekunov, York and Chow2001), the stable form at ambient conditions. The refined atom coordinates of salmeterol xinafoate are reported in Table I, and the coordinates from the DFT optimization in Table II. The root-mean-square deviation of the non-H atoms in the salmeterol cation is 0.256 Å (Figure 3). This good agreement between the refined and optimized structures is strong evidence that the experimental structure is correct (van de Streek and Neumann, Reference van de Streek and Neumann2014). The largest differences are in the conformation of the C22–C24 chain carbon atoms. The discussion of the geometry uses the DFT-optimized structure. The asymmetric unit (with atom numbering) is illustrated in Figure 4, and the crystal structure is presented in Figure 5. The large displacement coefficients of the atoms in the C25–C30 phenyl ring presumably reflect disorder in this portion of the molecule. We felt that detailed modeling of the disorder was beyond the scope of this study.

Figure 3. (Colour online) Comparison of the refined and optimized structures of salmeterol xinafoate. The Rietveld-refined structure is in red, and the DFT-optimized structure is in blue.

Figure 4. (Colour online) The molecular structure of salmeterol xinafoate, with the atom numbering. The atoms are represented by 50% probability spheroids.

Figure 5. (Colour online) The crystal structure of salmeterol xinafoate, viewed down the [−110]-axis. The hydrogen bonds are shown as dashed lines.

All of the bond distances fall within the normal ranges indicated by a Mercury Mogul Geometry check (Macrae et al., Reference Macrae, Bruno, Chisholm, Edington, McCabe, Pidcock, Rodriguez-Monge, Taylor, van de Streek and Wood2008). The C6–C5–C10 angle of 125.0° is flagged as unusual [average = 120.0(16)°; Z-score = 3.1]. The hydroxyl group O11 participates in a strong hydrogen bond to the carboxylate group of the xinafoate, so the unusual geometry can be rationalized. Similarly, the torsion angles C12–C10–C5–C4 and C12–C10–C5–C6 are unusual; the hydrogen bonds involving O11 seem to have resulted in distortions of that region of the molecule.

A semi-empirical conformation examination (RHF/PM3) using Spartan ‘14 (Wavefunction, 2013) indicated that the observed conformation of the salmeterol cation is ~27 kcal mole⁻¹ higher in energy than a local minimum. A molecular mechanics force field (MMFF) sampling of conformational space indicated that the optimized solid state conformation is 19 kcal mole⁻¹ higher in energy than the minimum energy conformation, which folded on itself to form a compact molecule. The energy difference indicates that hydrogen bonds and van der Waals forces contribute significantly to the crystal energy and to the extended salmeterol conformation observed in the solid state.

Analysis of the contributions to the total crystal energy using the Forcite module of Materials Studio (Accelrys, 2013) suggests that the intramolecular deformation energy contains about equal contributions from bond angle and torsion angle distortion terms. The intermolecular energy is dominated by electrostatic contributions, which in this force–field-based analysis include hydrogen bonds. The hydrogen bonds are better analyzed using the results of the DFT calculation.

As expected, there is a strong N13–H56···O43 hydrogen bond between the cationic portion of the salmeterol and the ionized carboxylate of the xinafoate (Table III). This is a discrete hydrogen bond, with graph set D1,1(2) (Etter, Reference Etter1990; Bernstein et al., Reference Bernstein, Davis, Shimoni and Chang1995; Shields et al., Reference Shields, Raithby, Allen and Motherwell2000). The other hydrogen atom of the cation forms an even stronger N13–H89···O8 hydrogen bond to the hydroxymethyl oxygen O8; this hydrogen bond participates in patterns with graph sets R2,2(18), C2,2(11) and larger patterns. The hydroxyl groups O9 and O11 make D1,1(2) hydrogen bonds to the ionized carboxylate, and the hydroxyl group O8 participates in R2,2(32) and larger hydrogen bond patterns. There is an intramolecular O44–H88···O43 S1,1(6) hydrogen bond in the xinafoate anion. The hydroxyl group O9 acts as an acceptor in two C–H···O hydrogen bonds. Although weak, these C–H donor hydrogen bonds probably contribute significantly to the crystal energy. These hydrogen bonds result in complex chains along the b-axis.

The volume enclosed by the Hirshfeld surface (Figure 6; Hirshfeld, Reference Hirshfeld1977; McKinnon et al., Reference McKinnon, Spackman and Mitchell2004; Spackman and Jayatilaka, Reference Spackman and Jayatilaka2009; Wolff et al., Reference Wolff, Grimwood, McKinnon, Turner, Jayatilaka and Spackman2012) is 809.50 Å³, 99.4% of half the unit-cell volume. The molecules are thus not tightly packed. The only significant close contacts (red in Figure 6) involve the hydrogen bonds.

Figure 6. (Colour online) The Hirshfeld surface of salmeterol xinafoate. Intermolecular contacts longer than the sums of the van der Waals radii are colored blue, and contacts shorter than the sums of the radii are colored red. Contacts equal to the sums of radii are white.

The Bravais–Friedel–Donnay–Harker (Bravais, Reference Bravais1866; Friedel, Reference Friedel1907; Donnay and Harker, Reference Donnay and Harker1937) morphology suggests that we might expect platy morphology for salmeterol xinafoate, with {001} as the principal faces. A fourth-order spherical harmonic-preferred orientation model was included in the refinement; the texture index was 1.076, indicating that preferred orientation was significant in this rotated capillary specimen. The powder pattern of salmeterol xinafoate has been submitted to ICDD for inclusion in the PDF as entry 00-065-1430.