PCCP View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

PAPER

Cite this: Phys. Chem. Chem. Phys., 2017, 19, 16650

View Journal | View Issue

Does Z 0 equal 1 or 2? Enhanced powder NMR crystallography verification of a disordered room temperature crystal structure of a p38 inhibitor for chronic obstructive pulmonary disease† Cory M. Widdifield, ‡a Sten O. Nilsson Lill,b Anders Broo,b Maria Lindkvist,b Anna Pettersen,b Anna Svensk Ankarberg,b Peter Aldred,b Staffan Schantz*b and Lyndon Emsley *c The crystal structure of the Form A polymorph of N-cyclopropyl-3-fluoro-4-methyl-5-[3-[[1-[2-[2(methylamino)ethoxy]phenyl]cyclopropyl]amino]-2-oxo-pyrazin-1-yl]benzamide (i.e., AZD7624), determined using single-crystal X-ray diffraction (scXRD) at 100 K, contains two molecules in the asymmetric unit (Z 0 = 2) and has regions of local static disorder. This substance has been in phase IIa drug development trials for the treatment of chronic obstructive pulmonary disease, a disease which affects over 300 million people and contributes to nearly 3 million deaths annually. While attempting to verify the crystal structure using nuclear magnetic resonance crystallography (NMRX), we measured 0

13

C solid-state NMR (SSNMR)

0

spectra at 295 K that appeared consistent with Z = 1 rather than Z = 2. To understand this surprising observation, we used multinuclear SSNMR (1H,

13

C,

15

N), gauge-including projector augmented-wave

density functional theory (GIPAW DFT) calculations, crystal structure prediction (CSP), and powder XRD (pXRD) to determine the room temperature crystal structure. Due to the large size of AZD7624 (ca. 500 amu, 54 distinct

13

C environments for Z 0 = 2), static disorder at 100 K, and (as we show) dynamic

disorder at ambient temperatures, NMR spectral assignment was a challenge. We introduce a method to enhance confidence in NMR assignments by comparing experimental 13C isotropic chemical shifts against site-specific DFT-calculated shift distributions established using CSP-generated crystal structures. The assignment and room temperature NMRX structure determination process also included measurements of 13

C shift tensors and the observation of residual dipolar coupling between 0

13

C and

14

N. CSP generated

0

ca. 90 reasonable candidate structures (Z = 1 and Z = 2), which when coupled with GIPAW DFT results, room temperature pXRD, and the assigned SSNMR data, establish Z 0 = 2 at room temperature. We find that the polymorphic Form A of AZD7624 is maintained at room temperature, although dynamic disorder Received 11th April 2017, Accepted 2nd June 2017

is present on the NMR timescale. Of the CSP-generated structures, 2 are found to be fully consistent with the SSNMR and pXRD data; within this pair, they are found to be structurally very similar (RMSD16 = 0.30 Å).

DOI: 10.1039/c7cp02349a

We establish that the CSP structure in best agreement with the NMR data possesses the highest degree of structural similarity with the scXRD-determined structure (RMSD16 = 0.17 Å), and has the lowest

rsc.li/pccp

DFT-calculated energy amongst all CSP-generated structures with Z 0 = 2.

Introduction The gold standard for the experimental crystal structure determination of molecular solids remains single-crystal

X-ray diffraction (scXRD). However, many newly-prepared systems cannot be crystallized, and only powdered solids can be isolated. When sufficiently large single crystals cannot be obtained, several alternatives for structure determination exist.

a

` Tre`s Hauts Champs, Universite´ de Lyon, 69100 Villeurbanne, France Institut des Sciences Analytiques (CNRS/ENS de Lyon/UCB Lyon 1), Centre de RMN a AstraZeneca R&D Gothenburg, Pharmaceutical Technology & Development, SE-431 83 Mo¨lndal, Sweden. E-mail: staff[email protected] c Institut des Sciences et Inge´nierie Chimiques, Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. E-mail: [email protected] † Electronic supplementary information (ESI) available: Additional 1H, 13C, 15N SSNMR spectra, 1H(T1) measurements, assignment details, 13C–13C MAT NMR spectra, shift anisotropy tensor parameter definitions and measured values, DFT-calculated system energies, additional pXRD patterns, direct space heavy atom RMSD values for leading CSP structures. See DOI: 10.1039/c7cp02349a ‡ Present address: Department of Chemistry, Durham University, DH1 3LE Durham, UK. b

16650 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

This journal is © the Owner Societies 2017

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Paper

The most common alternative is powder XRD (pXRD), but this often requires advanced computational modelling1 and/or structure model guesses that are sufficiently close to the correct structure to enable a proper refinement.2 When pXRD cannot be usefully employed, electron diffraction experiments coupled with density functional theory (DFT) calculations have been used to solve the structures of nanocrystalline organics.3 An alternative to diffraction and microscopy techniques is furnished by solid-state nuclear magnetic resonance (SSNMR) methods, dubbed ‘NMR crystallography’ (NMRX) in this context.4,5 In general, NMRX techniques using powder samples rely upon a modest amount of computational modelling and yield structural solutions with a resolution that is superior to almost every other technique, and (conservatively) at least comparable to pXRD.6 Increasingly, SSNMR data are being combined with pXRD measurements to resolve crystal structures,7,8 occasionally under the moniker of SMARTER crystallography.9 NMRX can be broadly applied, and recent examples of its use include organic/co-crystalline systems,10–12 covalent-organic frameworks,13 metal–organic frameworks,14 zeolites,15 inorganics,16 proteins and biomacromolecules,17 and even glassy/amorphous materials.18 Importantly, NMRX has been used to determine complete de novo crystal structures.19 The above de novo determination relied upon the generation of candidate structures using crystal structure prediction (CSP) methods.20 However, as molecular complexity increases, the number of degrees of freedom also increases significantly, and comprehensive CSP methods cannot rapidly sample conformation space in a sufficiently complete manner (comprehensive CSP is presently limited to molecules with about 10 degrees of torsional freedom21). This is particularly relevant as a large number of important pharmaceuticals are beyond this limit, are composed of multiple building blocks (i.e., co-crystals, solvates and salts), and/or have more than one molecule in the asymmetric unit (i.e., Z 0 4 1). Typical applications of NMRX rely primarily upon 1H and 13C isotropic chemical shifts (diso),4,19,22,23 which can usually only be interpreted globally and do not provide site-specific geometry constraints. Nevertheless, several other NMR observables can be included in a straightforward manner. For example, Harper and co-workers have shown that 13 C and 15N chemical shift tensors may be used to refine the structures of small organics which have access to a few conformational modes.24 Likewise, 1H–1H spin diffusion11,25 and dipolar recoupling experiments26,27 have also been used. Here, we use an enhanced NMRX approach to verify the scXRD-determined crystal structure, and also to determine the room-temperature crystal structure, of Form A of N-cyclopropyl3-fluoro-4-methyl-5-[3-[[1-[2-[2-(methylamino)ethoxy]phenyl]cyclopropyl]amino]-2-oxo-pyrazin-1-yl]benzamide. Henceforth, the molecular building block will be referred to as AZD7624 (Scheme 1), with polymorphic Form A of AZD7624 being specified as 1. This material is an inhaled p38 inhibitor being assessed for the treatment of chronic obstructive pulmonary disease (COPD),28 a disease which affects 328.6 million people worldwide,29 and contributed to approximately 2.9 million deaths in 2013.30 To the best of our knowledge, this is the

This journal is © the Owner Societies 2017

PCCP

Scheme 1 Structural formula of the AZD7624 molecule, including the numbering scheme used here.

largest molecular unit so far subjected to NMRX methods involving structure determination (molecular weight of 491.56 g mol1). We subsequently show that 1 has two molecules in the asymmetric unit (Z 0 = 2), and possesses localised static disorder, making this system a challenge for current NMRX methods. Not only is this molecule well beyond the current capacity for complete de novo CSP methods, but it is a challenge to assign the SSNMR spectra. We incorporate a number of NMRX tools which are used to generate information about the crystallographic space group, the number of symmetryindependent molecular conformations (i.e. the Z0 value), and information about intermolecular interactions. We then introduce a novel method that allows us to remove ambiguity in the experimental assignments by comparing experimental diso(13C) values against chemical shift distributions generated from DFT calculations on various CSP-generated structures. In addition to the usual 1H and 13C isotropic chemical shifts, we measure 13 C chemical shift tensors, 15N diso values, and include 13C–14N residual dipolar coupling31 (RDC) information. We thus verify the crystal structure determined at 100 K, but importantly also outline a process to determine the room-temperature structure of 1 using NMRX, starting from the molecular conformations found in the scXRD structure. Further, we resolve the nature of the apparently contradictory scXRD and room-temperature 13 C SSNMR data, as the Z 0 value indicated at a first glance of the 13C NMR data (seemingly, Z 0 = 1 at 295 K) does not agree with scXRD (Z 0 = 2 at 100 K).

Results and discussion A

Single crystal X-ray diffraction

The crystal structure of 1 was determined using scXRD data measured at 100 K, and has two molecules in the asymmetric unit (Z 0 = 2). We note that the structure could be treated as orthorhombic (i.e., all cell angles equal to 901); however, if treated rather as a monoclinic cell, the merging R decreases from 22.7% for the orthorhombic cell to ca. 7% for the monoclinic cell. There are no plausible orthorhombic space groups indicated by the systematic absences, and a PLATON check for missing symmetry was negative.32 The crystal refines as a pseudomerohedral twin (100 010 001) with a twin fraction of 28%.

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16651

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

PCCP

Paper

The secondary amine chains in both conformers of the asymmetric unit are statically disordered under typical scXRD data acquisition conditions. The disorder in these chains was subsequently resolved by the computational approach mentioned in the CSP experimental section below. The two in silico-generated crystal structures, each with Z 0 = 2, were calculated to differ in energy by ca. 4 kJ mol1. To understand the origin of the disorder in this region, we performed a careful analysis of the crystal structure. It was found that in 1, the amine chain does not involve the amine as a hydrogen bond donor or acceptor. Rather, the amine chain and its backbone carbons are located in a hydrophobic pocket and interact non-specifically with two aromatic rings (a phenyl ring and a pyrazinone ring). It is therefore reasonable to expect that the amine chains rotate rather freely. According to a non-covalent interaction (NCI) analysis,33 the interactions between the amine chain and the two rings are weakly stabilizing (Fig. 1).

couplings, resulting in reasonably high-resolution 1H MAS NMR spectra, although similar resolution can usually be obtained at lower MAS frequencies via homonuclear decoupling methods.36 For 1, 1H MAS NMR experiments were performed at multiple MAS frequencies, and at multiple applied fields (Fig. 2 and ESI,† Fig. S1). At the highest MAS frequency available to us (nMAS E 61 kHz), several resonances are observed, whose positions are later used to calibrate the 1H frequency axis of the 1H–13C and 1 H–15N dipolar heteronuclear correlation (HETCOR) NMR spectra,37 as eDUMBO-122 homonuclear decoupling38 was used during the indirect dimension evolution period. Proton MAS NMR experiments were also used to measure a 1H spin–lattice relaxation value (i.e., T1(1H)) of 9.1 s at 298 K and an applied magnetic field (B0) of 16.4 T, which is relevant when optimising cross-polarisation (CP)/MAS NMR experiments39 using 1H nuclei as the polarisation source (additional T1(1H) data are in the ESI,† Fig. S2–S4).

B NMR crystallography structure assignment and verification

(ii) 13C CP/MAS NMR. Room temperature 13C CP/MAS NMR spectra at various MAS frequencies and two B0 values resolve 27 carbon sites (Fig. 3 and ESI,† Fig. S5 and S6). In Table 1, the observed diso(13C) values are compiled, along with all resolved 1 H and 15N sites, and the assignment is provided. Subsequently, we outline the tools and processes used to arrive at this assignment. As there are 27 carbon atoms in AZD7624, and as the scXRD structure is such that Z 0 = 2, we expected to resolve many more peaks (up to 54) in the 13C NMR spectrum than presently observed. At first sight, this implies an inconsistency between the scXRD and 13C SSNMR datasets. Interestingly, we resolve 27 carbon sites in the 13C NMR spectra, which is exactly equal to the number of carbon atoms in one AZD7624 molecule and hence based on the NMR we would be tempted to state Z 0 = 1. By performing 13C CP/MAS NMR experiments near 100 K, we find that the acquired spectrum exhibits many additional peaks relative to the ambient temperature spectrum (ESI,† Fig. S7). It is clear that almost all of the 27 13C resonances depicted in Fig. 3 are split into two at the lower temperature, so Z 0 = 2 according to 13C NMR at the lower temperature. One way of rationalising the 13C NMR data would be to postulate that a polymorphic transition has occurred between the two measurement temperatures, which has the consequence that the low

We outline here the process and tools used to verify the crystal structure of 1 determined from scXRD data collected at 100 K. We omit overly detailed discussions, and interested readers may consult the ESI† for further details. I SSNMR experiments (i) 1H magic-angle spinning (MAS) NMR. Proton NMR experiments are a cornerstone of liquid state NMR spectroscopy, but their solid state counterparts are only recently becoming routine.34 The main challenge with 1H NMR experiments using powdered organic solids at natural abundance is the loss of spectral resolution due to the strong homogeneous 1H–1H dipolar coupling network. However, outstanding developments in MAS probes (which can now reach stable spinning frequencies (nMAS) in excess of 110 kHz35) nearly average these dipolar

Fig. 1 NCI plots of a truncated portion of 1. Reduced density gradient isosurface (IS) cutoffs are 0.25. A red IS represents an unfavourable interaction, and a green IS denotes a weak favourable interaction. Atomic colour scheme: C: cyan; F: ochre; H: white; N: blue; O: red.

16652 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

Fig. 2 1H echo MAS NMR spectrum of 1, acquired at B0 = 11.75 T and nMAS = 60.94 kHz. 16 transients were averaged with a recycle delay of 35.5 s. The corrected sample temperature was 295 K. Chemical shifts for diagnostic resonances are indicated (in ppm).

This journal is © the Owner Societies 2017

View Article Online

Paper

PCCP

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Table 1

Fig. 3 Assigned 13C CP/MAS NMR spectra of 1, acquired at B0 = 16.4 T with nMAS = 16.00 kHz and T = 298 K. Spectra in (a, c and e) are collected after a typical CP contact period (2 ms), while the spectra in (b, d and f) are collected using a short CP (SCP) pulse sequence (0.5 ms 1H spin lock followed by a 40 ms contact time), which is selective for primarily CH and CH2 group carbons, while CH3 are nearly totally suppressed and quaternary carbons are absent. (a, c and e) resulted from 760 transients, while (b, d and f) resulted from 256 transients. For all, the recycle delay was 11.8 s. Asterisks denote spinning sidebands due to MAS.

temperature structure is not the same as that at physiologicallyrelevant temperatures. Another reasonable interpretation would be that there is a dynamic process occurring at room temperature which allows the truly Z 0 = 2 crystal structure to appear averaged on the NMR timescale. We discuss the temperaturedependent structural aspects of 1 in subsequent sections. To facilitate the spectral assignment process, we also acquired a short CP (SCP) MAS NMR spectrum of 1 (Fig. 3(b, d and f), with group assignments provided in the ESI,† Fig. S8). As detailed elsewhere,40 this experiment is selective for CH and CH2 group carbons, while the signals from quaternary and methyl group carbons are absent and very greatly reduced, respectively. (iii) 15N CP/MAS NMR. Nitrogen-15 CP/MAS NMR experiments were performed on 1. Due to its low natural abundance, and the low resonance frequency of 15N relative to 13C, these experiments are less common than the more traditional SSNMR results presented thus far. By performing experiments at two transmitter offsets (ESI,† Fig. S9), five nitrogen sites are observed (one with very low signal intensity indicating that the site is not only distant from 1H nuclei, but also that it possesses an exceptionally large 15N shift anisotropy, which is consistent with an sp2-hybridised nitrogen).41 These observations are in agreement with the number of nitrogen atoms in the AZD7624 molecule.

This journal is © the Owner Societies 2017

1

H,

13

C and

15

N diso values for 1 and assignmenta

Site label

diso(1H)/ppm

1 2 3 4 5 6 7 8 9 10 11 12/13 13/12 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31/32 32/31

1.32 0.60 1.77 3.35 — 6.55 8.10 7.84 8.21 — — 0.78 1.39 7.85 —b — — 6.99 7.45 — — — 1.49 — 6.63 — 7.40 —d 9.12 1.74 1.03 1.31

diso(13C)/ppm 37.65 — 51.35 67.52 159.13 110.10 130.74 121.62 129.98 130.41 33.84c 15.67 12.11 — — —

152.93c 150.74c 120.17c 123.30c

141.12c 125.39 11.50 B160.2 113.46 136.41 122.15 166.85c — 24.65c 5.66 3.18

diso(15N)/ppm — 361.52 — — — — — — — — — — — 281.28 — — 212.91 — — 128.74 — — — — — — — — 252.27 — — —

a Chemical shift measurements occurred at 295  3 K. Typical measurement errors were 0.05 ppm for 1H, and 0.03 ppm for 13C and 15N, as verified through repeat experiments. Assignments of sites 12/13 and 31/32 are ambiguous, but based upon the GIPAW DFT computations, the more probable assignment is listed with the site label in bold. b 1 H–13C HETCOR measurements using a relatively long contact time establish proximity to a 1H site where diso = ca. 7.7–7.8 ppm, which is consistent with the NH group nearby (site 14). c Nuclei associated with these chemical shifts are experimentally observed to experience RDC to 14 N. d 1H–13C HETCOR measurements using a relatively long contact time establish proximity to a 1H site where diso = ca. 9.2 ppm, which is consistent with the NH group nearby (site 29).

With basic 1D NMR information in hand, two-dimensional (2D) correlation experiments are now discussed, which will be used to arrive at a nearly complete assignment of the observed signals to sites in the molecule. (iv) 1H–13C and 1H–15N HETCOR experiments. Dipolar HETCOR SSNMR spectra37 were acquired using 1H–1H homonuclear (eDUMBO-122) decoupling38 during the indirect dimension evolution (i.e., t1) with heteronuclear 1H–X (X = 13C, 15N) decoupling (SPINAL-64)42 during the direct dimension acquisition (i.e., t2). Under these experimental conditions, eDUMBO-122 provides an excellent resolution enhancement in the indirect 1H dimension (relative to using MAS alone), but comes with a well-known scaling of the chemical shift information.43 By choosing an appropriate 1 H transmitter offset, we are confident that the scaling of the 1 H chemical shift information is essentially constant across the 1H dimension.36 To verify, we cross-referenced the indirect dimension of the 1H–13C HETCOR NMR spectrum (Fig. 4a)

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16653

View Article Online

PCCP

Paper

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Due to the short CP block at the start of the pulse sequence, it is clear that three of the five nitrogen environments are protonated, which is consistent with the AZD7624 molecular structure. (v) 13C–13C incredible natural abundance double quantum experiment (INADEQUATE). If sensitivity issues could be surmounted, the refocused 13C–13C INADEQUATE NMR experiment45 would be widely used to assign 13C NMR spectra of solids at natural abundance. While recent examples have used high-field dynamic nuclear polarisation (DNP)46 under MAS at near liquid nitrogen temperatures to collect refocused 13 C–13C INADEQUATE spectra of various organic materials at natural isotopic abundance,27,47 it remains a very challenging experiment at room temperature due to its exceptionally low sensitivity. In addition, the AZD7624 molecule, with its 27 carbon sites and reasonably long T1(1H) value at room temperature, adds further challenge due to signal dilution per unit mass of sample, and due to limitations placed on the recycle delay. There are however examples of INADEQUATE experiments carried out upon molecules with comparable complexity using conventional room temperature NMR,48 including cases with Z 0 = 2. In the room temperature natural abundance refocused INADEQUATE NMR spectrum shown in Fig. 5, we observed many correlations amongst the aromatic carbons. Unfortunately, not all expected correlations were observed, and hence only a partial assignment of the 13C SSNMR data could be made. Assignment of the remaining sites required additional spectral information, as well as computed diso values realised from gauge-including projector augmented-wave (GIPAW) DFT calculations on CSP-generated structures containing the AZD7624 molecule. Fig. 4 Heteronuclear correlation spectra of 1, acquired at B0 = 11.75 T, nMAS = 12.50 kHz and T = 294(1) K. In (a), a 1H-13C HETCOR (CP contact time of 125 ms), showing the high 13C shift range, is depicted, while in (b), a 1 H–15N HETCOR (contact time of 300 ms), showing all observed correlations, is given. Site labels are provided above peaks. An additional 1H–13C HETCOR NMR spectrum can be found in the ESI,† Fig. S10. For (a) 160 scans per t1 increment, a recycle delay of 9.2 s, and 169 t1 points (69 h experiment time), and for (b) 160 scans per t1 increment, a recycle delay of 9.0 s, and 108 t1 points (43 h experiment time). Homonuclear 1H–1H decoupling during t1 was accomplished using eDUMBO-122,38 which required scaling the indirect dimension by a constant factor of (0.54)1.

against the 1D 1H fast MAS NMR spectrum provided earlier. As a result, we increased the precision of the measured diso(1H) values, which is critical in the subsequent structure verification process. As noted, 15N CP/MAS NMR experiments are insensitive relative to 13C analogues and thus typically require significant investments of spectrometer time to arrive at a sufficient signalto-noise ratio (S/N). Likewise, solid state correlation experiments involving 15N at natural abundance and room temperature are uncommon, but not unheard of.44 The 2D 1H–15N HETCOR NMR spectrum shown in Fig. 4b was acquired using the same pulse sequence denoted earlier, and required the same care when correcting for the 1H dimension scaling factor.

16654 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

Fig. 5 High-frequency region of the refocused 13C–13C INADEQUATE SSNMR spectrum of 1 at B0 = 11.75 T, T = 295 K, nMAS = 12.50 kHz and a 1 H/13C contact time of 2.2 ms. The spectrum was acquired in ca. 6.1 days (recycle delay = 9.2 s, 512 transients averaged per t1 point and 112 t1 increments). Unambiguous correlations are given by green dashed lines, while those which contain some degree of ambiguity are given by orange dashed lines.

This journal is © the Owner Societies 2017

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Paper

(vi) Additional SSNMR information used for assignment. The set of experiments above has been sufficient to assign observed resonances to chemical sites for the vast majority of NMRX studies beforehand;8,19,23,49,50 however, with these tools we could only partially assign 1. Additional experiments and information were required to complete the assignment of 1. By performing 13C CP/MAS NMR experiments at multiple applied fields, we established those carbon resonances for atoms which were bound to nitrogen, as the 14N nucleus (natural abundance = 99.63%)51 is quadrupolar (I(14N) = 1) and hence a RDC to 13C exists even under fast MAS. This effect is always present, but it is augmented due to the B0-dependence of the RDC (see ESI,† Fig. S11, for an example). Using this process, we identified 8 13C nuclei (out of the expected 10 for one AZD7624 molecule) which were coupled to 14N. It is strongly suspected that the two unidentified sites correspond to the O–CH2–CH2–NH–CH3 region of the molecule, which is statically disordered at 100 K according to the scXRD structure. The lack of a 13C–14N RDC for these sites provides evidence that the static disorder at lower temperatures becomes dynamically disordered at ambient temperatures, and thereby reduces on average the 13C–14N RDC value. The final bit of experimental NMR information used for structure assignment comes from performing a 5p-PULSE magicangle turning (MAT) experiment52 to measure the 13C chemical shift tensors. Shift tensor information, particularly for 13C and 15 N, has been used on several occasions for various NMRX tasks.24 This 2D experiment facilitates the measurement of this information as it cleanly separates the isotropic and anisotropic chemical shift information into two dimensions (as shown in Fig. S12 and S13, ESI†). By extracting the appropriate rows, analytical data fitting procedures can be employed53 to obtain the tensor information (tensor parameter definitions are provided in Fig. S14, ESI,† a sample of the data and fits is in Fig. 6, and the measured 13C shift tensor parameters are in the ESI,† Table S1). II CSP and GIPAW DFT calculations for Z 0 = 1 enhance assignment confidence. Assignment of the observed 1H, 13C, and 15N NMR resonances to the chemical sites of 1 was completed as a result of the suite of NMR experiments above. To increase confidence in this assignment, starting from the known molecular conformers of 1 (established using scXRD data), we used CSP under the assumption of Z 0 = 1 (see Table S2, ESI,† for calculated energies) to generate candidate crystal structures. We further wish to address: (i) if the available room-temperature SSNMR data are consistent with Z 0 = 1 structure models; and (ii) how the diso(13C) values would respond to changes in the lattice environment. While lattice effects on 13C isotropic chemical shifts are not quite as large as they are for 1H,19,23 by understanding how the diso(13C) values may distribute as a function of the lattice environment, we expect to gain insights that can be used to enhance confidence in our assignment (or potentially identify mis-assigned resonances). In total, we subjected the leading 55 CSP-generated structures (as ranked by lattice energy) where Z 0 = 1 to magnetic shielding calculations, and generated site-specific distributions for the calculated 13C isotropic chemical shift and shift tensor

This journal is © the Owner Societies 2017

PCCP

Fig. 6 Typical examples of 1D extracts of the anisotropic dimension taken from the 2D 13C–13C 5p-PULSE MAT NMR spectrum of 1, itself provided in the ESI,† Fig. S12/S13. Red traces correspond to fits to the experimental data in black. The fits (which have been offset to better illustrate the agreement between the fits and the experimental data) are obtained using two variable parameters (i.e., Dd and Z), with the diso values fixed using the 1D 13C CP/MAS NMR results provided earlier.

anisotropy (Dd) values. Histograms of selected distributions are given in Fig. 7, with corresponding normal distribution parameters (mean, standard deviation, and skew) being tabulated for all 13C sites in Table S3, ESI.† To highlight the idea behind this process, we consider in detail how the computed 13C diso and Dd values are distributed for sites labelled 15 and 16 (according to Scheme 1), and in the process verify (or not) the earlier assignment. First, the experimental shift assigned to site 16 (i.e., diso = 150.74 ppm) falls within the associated computed distribution. Secondly, considering the corresponding shift distributions for all other sites in the AZD7624 molecule, only site 15 could reasonably be considered as an alternative assignment (Fig. 3a). Consistent with the various SSNMR experiments, both 15 and 16 are known to be quaternary carbons, and both are attached to nitrogen atoms (due to the appearance of a larger RDC effect at lower B0). From the 13C–13C INADEQUATE, we see that these two sites are from carbons which are chemically bound to one another. By looking at the distributions, we see that assigning the peak at 150.74 ppm to site 16 in the AZD7624 molecule is more consistent than the only other possible assignment (i.e., to site 15). Additional validation comes by considering the Dd(13C) values: experimentally, the 13C site where diso = 152.93 ppm possesses a negative anisotropy (Dd = 125.7(0.9) ppm) which is also larger in magnitude than that of the peak at 150.74 ppm (Dd = 109.1(0.8)). Across the GIPAW DFT-computed Dd values for the 55 CSP-generated structures where Z 0 = 1, we find that the carbonyl carbon in

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16655

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

PCCP

Paper

Fig. 7 Sample histograms of the GIPAW DFT-computed 13C isotropic chemical shifts (a and b) and 13C tensor anisotropy values (c and d) for site 15 (a and c) and 16 (b and d) of AZD7624 across the 55 CSP-generated crystal structures where Z 0 = 1. Green arrows indicate the bin corresponding to the experimentally measured value. To convert a computed isotropic magnetic shielding value (siso) to a chemical shift value (in ppm), the following relationship was used: diso = 169.82  siso.50

the pyrazinone ring possesses a positive Dd that is about 10 ppm smaller in magnitude relative to the carbon directly adjacent to it, which itself possesses a consistently negative anisotropy. The computational data considered here thus supports our earlier assignment. Where applicable, other sites in AZD7624 were verified in an analogous fashion. III Verification of scXRD structure and selection of best CSP structures. In addition to the CSP-generated structures with Z 0 = 1 outlined earlier, GIPAW54 DFT calculations were performed using the crystal structure determined by scXRD at 100 K, as well as CSP-generated structures where Z 0 = 2 (operating under two scenarios: one where the AZD7624 molecule is rigid once within the crystal lattice; and one where it is allowed some conformational flexibility as part of the CSP process, see the ESI,† Tables S4 and S5, for computed system energies). It is found that the agreement between the experimental diso(1H) values and those calculated with GIPAW DFT (quantified by determining the root-mean-squared difference (RMSD) between calculated and experimental chemical shifts) using CSP-generated candidate structures is very poor when Z 0 = 1, but when Z 0 = 2, the agreement is generally quite good (Fig. 8). Importantly, it is seen that: (i) the 1H RMSD value between the experimentally measured chemical shifts (Table 1) and those which were calculated using GIPAW DFT for the scXRD crystal structure is 0.31 ppm, which provides further confidence in the

16656 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

Fig. 8 Plots of 1H root-mean-squared (RMS) differences between experimentally observed (Table 1) vs. GIPAW DFT-computed isotropic 1H chemical shift values for various CSP-generated structures. (a) Z 0 = 1, and (b) Z 0 = 2 (rigid molecule) in green, Z 0 = 2 (flexible molecule) in purple, and the scXRD structure (100 K) in red. Dashed lines spanning 0.33 ppm indicate the average RMSD for a valid structure,55 while the grey band represents 1 standard deviation in the RMSDs for a valid structure (i.e., 0.19 ppm).

assignment and in fact verifies the scXRD structure from an NMRX perspective;12 and (ii) it seems important to allow Z 0 = 2, even though the experimental NMR measurements were carried out on a system that appeared to be Z 0 = 1 at ambient temperatures. As SSNMR is well-known to be able to distinguish between polymorphic forms of organics, verification of the scXRD structure using room-temperature NMR data greatly diminishes the possibility that a polymorphic change in 1 could have occurred between 100 K and room temperature. Unfortunately, if we consider only the diso(1H) RMSD values, there is not sufficient discrimination amongst most of the CSP-generated Z 0 = 2 structures (structures are considered to be unacceptable if they possess a diso(1H) RMSD value greater than the grey shaded regions in Fig. 8) to conclude at this point which CSP-generated structure is correct from an NMRX viewpoint. With the scXRD crystal structure verified, we now consider pXRD information to increase our confidence and

This journal is © the Owner Societies 2017

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Paper

potentially allow us to discriminate between the CSP-generated Z 0 = 2 structures. A pXRD powder pattern was acquired for 1 under ambient conditions, and compared against the calculated diffraction patterns for several of the CSP-generated structures having the lowest diso(1H) RMSD values, as well as the scXRD structure at 100 K. Unsurprisingly, due to the slight variation in the unit cell as a function of temperature, the agreement between the experimental pXRD pattern and one calculated using the scXRD structure is fairly good, yet clearly not optimal (ESI,† Fig. S15). However, by simply modifying the unit cell of the 100 K scXRD structure to match that determined at room temperature using pXRD (without modifying any fractional atomic positions), the agreement between the observed and calculated XRD powder patterns is sufficiently good (ESI,† Fig. S16) to further enhance the confidence in our above NMRX verification of the scXRD structure. This now very strongly indicates that there is no polymorphic phase change in the material between 100 K and room temperature. Hence, it is likely dynamics on the NMR timescale which accounts for the apparent Z 0 = 1 observed in the room temperature 13C SSNMR spectra. Further, by (very slightly) adjusting the unit cells of several of the CSP-generated structures where Z 0 = 2 such that they match the cell dimensions determined using pXRD, very good agreement between calculated and experimental pXRD patterns is seen for several of the candidate structures (particularly CSP structures 1, 2 and 3, as labelled in Fig. 8b, and as shown in the ESI,† Fig. S17). Hence, by combining the earlier diso(1H) RMSD values with pXRD data, we have enhanced our ability to discriminate amongst the Z 0 = 2 CSP-generated structures. We note that the agreement between calculated and experimental pXRD patterns using the Z 0 = 1 CSP-generated structures was universally poor, which is consistent with the very high diso(1H) RMSDs shown in Fig. 8a. Lastly, we present quantitative and qualitative direct space comparisons between the ‘best’ three CSP-generated structures (on the basis of both 1H RMSDs and computed pXRD diffraction patterns) and the scXRD structure of 1 determined at 100 K. The best CSP candidates, as mentioned above, were structures 1, 2, and 3 from the Z 0 = 2 set. Performing heavy atom RMSD calculations using the Mercury visualisation software, we are able to fully overlay 16 AZD7624 molecules for CSP structures 2 and 3 only (Table S6, ESI†). Relative to the three leading candidate structures, all other Z 0 = 2 CSP-generated structures possessed either 16 molecule RMSD values which were significantly greater (1 out of 32), or were such that all 16 molecules could not be overlaid (31 out of 32), which implies long range crystal packing inconsistencies in the latter case. Between CSP structures 2 and 3, structure 2 has the lowest direct space heavy atom RMSD value at the 16 molecule level (i.e., RMSD16 = 0.17 Å), and we provide this 16 molecule overlay in Fig. 9. Importantly, the space group of CSP structure 2 is in agreement with that determined using scXRD, which is not the case with structure 3. As we had verified earlier that no phase change in the material occurs between 100 K and room temperature, structure 2 would thus be strongly preferred over structure 3. In addition, out of

This journal is © the Owner Societies 2017

PCCP

Fig. 9 Direct space heavy atom overlay (16 molecules) of the crystal structure for 1, determined using scXRD at 100 K (carbon atoms in grey) and the overall ‘best’ CSP-generated structure (i.e., structure 2 of the rigid Z 0 = 2 CSP set; carbon atoms in green). To enhance clarity, one molecule of each structure is provided in red, and hydrogen atoms have been omitted. Colour scheme for additional atom types: red = O; yellow = F; light blue = N.

the leading candidate CSP structures, structure 2 has the (tied for) lowest diso(1H) RMSD value, and the lowest DFT energy amongst all Z 0 = 2 CSP structures. Lastly, we note in passing that additional scXRD data acquired at 300 K exactly corroborate our NMRX findings that no phase change occurs.

Conclusions Using a suite of multinuclear magnetic resonance experiments (fast 1H MAS, 13C and 15N CP/MAS, 1H–13C and 1H–15N HETCOR, 13 C–13C 5p-PULSE MAT and 13C–13C INADEQUATE), crystal structure prediction (CSP), density functional theory calculations and powder X-ray diffraction, we have determined the structure of 1 at room temperature, which had also been determined using single-crystal X-ray diffraction techniques at 100 K. To the best of our knowledge, AZD7624 is the largest organic molecule subjected to ‘NMR-crystallography’ methods without resorting to isotopic enrichment, and possesses additional complicating factors such as Z 0 = 2 and both static and dynamic forms of disorder depending upon the system temperature. CSP methods generated chemically reasonable test structures under a variety of assumptions (i.e., Z 0 = 1, Z 0 = 2, with and without allowing for conformational flexibility once in the crystal lattice). By comparing the calculated 1H isotropic chemical shift values for these test structures against those which were experimentally measured, we convincingly show that Z 0 = 2 at room temperature. This is noteworthy as at first glance the room temperature 13C CP/MAS NMR spectrum appears to indicate Z 0 = 1. By performing an analogous NMR experiment at ca. 100 K, it becomes clear that the 13C NMR

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16657

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

PCCP

signals of the two molecules in the asymmetric unit of 1 are likely averaged on the NMR timescale at room temperature. We verify the scXRD structure at 100 K using an extended NMRX protocol, and by comparing the experimental powder X-ray diffraction pattern acquired at room temperature with diffraction patterns generated using the various CSP-generated structures (after allowing for a slight augmentation in the unit cell volume), it is demonstrated that there is no phase change in the material as a function of temperature between 100 K and room temperature. Using an expanded set of solid-state NMR experiments (specifically, the inclusion of 15N chemical shifts, consideration of 13C–14N residual dipolar coupling under MAS and 13C chemical shift tensors) we have assigned the observed NMR signals to chemical sites in the AZD7624 molecule. Without this essentially complete assignment, the computed 1H RMSDs would be less representative of the total system and hence would not be as useful in the scXRD structure verification process. It is expected that the more holistic approach employed here, which makes use of CSP, density functional theory, multinuclear magnetic resonance, and powder X-ray diffraction, will be critically important in crystallographically-relevant tasks (such as verification, refinement, and determination) pertaining to larger molecules such as AZD7624, for cases where Z0 4 1 and for cases where the system displays static/dynamic disorder.

Experimental A

Single-crystal X-ray diffraction

A single clear and colourless plate-shaped crystal of 1 was obtained after recrystallisation of AZD7624 from dichloromethane by slow evaporation. This crystal, having dimensions of 0.14  0.09  0.02 mm, was mounted on a MITIGEN holder in perfluoroether oil for data collection at 100 K. Data were collected using a Rigaku AFC12 FRE-VHF diffractometer equipped with an Oxford Cryostream low-temperature apparatus. The crystal was maintained at 100(2) K during data collection. Data were collected using profile data from o-scans of 1.01 per frame for 10.0 s using MoKa radiation (rotating anode, 45.0 kV, 55.0 mA). The number of runs and images was based on the strategy calculated by the program CrystalClear (Rigaku). The achievable resolution was 0.74 Å (ymax = 28.7001). Cell parameters were retrieved and refined using the CrysAlisPro (Agilent, V1.171.37.35, 2014) software. Data reduction was also performed using CrysAlisPro, and corrected for Lorentz polarisation. Completeness was 99.80% out to 28.7001 in y. The absorption coefficient (m) of this material is 0.093, with minimum and maximum transmissions of 0.88904 and 1.00000, respectively. The structure was solved by direct methods using ShelXS (Sheldrick, 2008), and refined by least squares using a version of ShelXL (Sheldrick, 2008). The crystal structure determined under the above conditions is provided as part of the ESI,† and has been deposited in the Cambridge Structural Database (CSD), reference: 1503408. Additional crystal structure determinations were carried out at 100 K and 300 K, and deposited in

16658 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

Paper

the CSD (CSD deposition numbers: 1503409 and 1503407, respectively). B Solid-state nuclear magnetic resonance Experiments were typically performed at or near room temperature (TSSNMR = 295(3) K) using Bruker AVANCE III spectrometers operating at B0 of 11.75 or 16.4 T. To establish spectral changes that may occur between room temperature and ca. 100 K, lowtemperature 13C NMR experiments were performed between 94–99 K and B0 = 9.4 T using an AVANCE III HD spectrometer. At 9.4 T, a low-temperature 3.2 mm HXY probe was used. At 11.75 T, in order to enhance experimental sensitivity, a 4.0 mm HX probe was used for correlation experiments, while a 1.3 mm HX probe was used when very fast MAS56 was desired. For the 1.3 mm probe, due to the heating expected when rotating the sample at high frequencies (measured as ca. +26 1C), the actual sample temperature was obtained by performing calibration 79 Br NMR experiments on KBr powder.57 Experiments at 16.4 T used a 3.2 mm HCN probe to allow for reasonably rapid MAS rates and enhanced shift dispersion relative to the lower applied fields. The 1H pulse length calibrations and chemical shift reference were established with powdered adamantane (diso(1H) = 1.87 ppm) at ca. 10–15 kHz MAS and T = 295 K. Adamantane was also used for 13C pulse calibrations, establishing the 1H/13C ramped-amplitude58 CP59 (RAMP-CP)/MAS (‘CP/MAS’) conditions, and referencing carbon chemical shifts (with the highfrequency 13C peak of adamantane at 38.48 ppm). Uniformly 13 C and 15N labelled a-glycine at 6–7 kHz MAS and ambient temperature was normally used to determine the 15N chemical shift reference (with the 15N shift of a-glycine being 347.54 ppm), and to establish the 1H/15N CP conditions. On one occasion at B0 = 11.75 T, a sample of 15N labelled alanine was used to set the shift reference (diso(15N, alanine) E 338.3 ppm). All other relevant experimental details can be found in the above sections of this manuscript or in the ESI.† C

Powder XRD

Powder XRD experiments were performed on ca. 10–15 mg of powder under ambient conditions. The powder was held between low density polyethylene films and analysed in transmission mode. Experiments were carried out on a PANalytical X’Pert PRO diffractometer (Ni-filtered Cu radiation (l = 1.5418 Å) at 45 kV and 40 mA) with a PIXCEL detector (active length 3.351 in 2y). Data were collected between 3 and 401 in 2y, with a step size of 0.0131 (0.46 s per step), using variable slits. The data were neither corrected nor processed further in any fashion and are presented ‘as-is’ for comparison against the pXRD patterns calculated from the candidate crystal structures generated via CSP methods. D

Quantum chemical calculations

DFT calculations were performed using the Cambridge Serial Total Energy Program (CASTEP), version 5.5,60 from input files `mes). For all generated with Materials Studio (Dassault Syste tasks (i.e., geometry optimisations; calculation of magnetic resonance parameters), the generalised gradient approximation

This journal is © the Owner Societies 2017

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Paper

(GGA) DFT exchange–correlation functional developed by Perdew, Burke, and Ernzerhof (PBE)61 was used. Core electrons were modelled using ultrasoft pseudopotentials62 and generated on-the-fly. Valence electrons were described with a plane wave basis, and typically with a kinetic energy cutoff of 700 eV, which has been widely tested and shown to be appropriate for organics19,23 (for one set involving the leading CSP candidate structures with augmented unit cell dimensions, 500 eV was used). Reciprocal space sampling of the Brillouin zone used a fixed Monkhorst-Pack63 k-point grid of 0.05 Å1. Dispersion corrections as given by Tkatchenko and Scheffler (TS)64 were always applied. Geometry optimisations (where applicable) were performed until the energy converged to within 5  106 eV and the maximum force was less than 0.01 eV Å1. When possible, symmetry was used to increase the efficiency of the calculations. GIPAW DFT calculations of nuclear magnetic shielding tensors were performed on all CSP-generated candidate crystal structures. E

PCCP

(rigid) conformers (Table S4, ESI†), the structures were optimised using the in-house force field only. The crystal structures generated, covering the polymorph landscape, were then subjected to GIPAW DFT magnetic shielding calculations using CASTEP without any further structural optimisation, unless denoted otherwise.

Acknowledgements We acknowledge the AstraZeneca medicinal chemistry work that led to the discovery of the AZD7624 molecule, in particular Timothy Luker (current address: Eli Lilly, UK). The authors would like to acknowledge Dr Mark E. Light at Southampton Chemical and Analytical Solutions (SCAS) for solving the crystal structure and for many fruitful discussions. We acknowledge funding from the Swiss National Science Foundation, grant no. 160112, and the Natural Sciences and Engineering Research Council of Canada (NSERC) for a postdoctoral fellowship (CMW).

Crystal structure predictions

CSP runs were initialised using the molecular conformations found in the 100 K scXRD structure. The disorder in the crystal structure was initially resolved by performing CASTEP geometry optimisations for the two alternate occupations, which lead to two separate crystal structures. The molecular conformers found in these two crystal structures were then the subject of a CSP polymorph search using a force-field developed in-house65 and the GRACE packing machinery66 under both Z 0 = 1 and Z 0 = 2. Polymorph space was searched via a MonteCarlo (MC) parallel tempering method, followed by a lattice energy minimization for each polymorph. New trial crystal structures were generated until a convergence criterion of ‘0.7’ was satisfied. This means we expect to have a 70% probability to have found all possible structure solutions given the space group and Z 0 constraints. In all CSP runs, a large number of trial crystal structures were generated in the 13 most common space groups (i.e., P21/c, P1% , P212121, P21, C2/c, Pbca, Pnma, Pna21, Pbcn, P1, Cc, C2, and Pca21). Generally, we undertook a rigid conformational polymorph search, i.e., not allowing the molecular conformer to adopt changes once in the crystal structure lattice; however, for one search assuming Z 0 = 2, we allowed conformational flexibility during the search (denoted as Z 0 = 2 (flexible)). The latter approach is considerably more time-consuming, especially for Z 0 = 2, and complete coverage of conformational space is difficult to ensure. In parallel to the initial two conformers, we created several new molecular conformers from a mixed random torsional/low mode sampling of conformational space in a water continuum model using the OPLS2005 force field,67 as implemented in ¨dinger Inc.). This was followed by conforMacroModel (Schro mer clustering and gas-phase geometry optimisations using B3LYP/6-31G(d) on representatives for each cluster. These were then used for CSP studies assuming Z 0 = 1 only. For the Z 0 = 1 (Table S2, ESI†) and Z 0 = 2 (flexible) conformers (Table S5, ESI†), the structures output from CSP were optimised first using the in-house force field specified above and then optimised using CASTEP under the constraint of a fixed unit cell. For the Z 0 = 2

This journal is © the Owner Societies 2017

Notes and references ˇ erny´ and V. Favre-Nicolin, Z. Kristallogr., 2007, 222, 105; 1 R. C A. Meden and I. Radosavljevic-Evans, Cryst. Res. Technol., 2015, 50, 747. 2 H. M. Rietveld, J. Appl. Crystallogr., 1969, 2, 65. 3 T. E. Gorelik, C. Czech, S. M. Hammer and M. U. Schmidt, CrystEngComm, 2016, 18, 529. 4 R. K. Harris, Solid State Sci., 2004, 6, 1025. 5 F. Taulelle, Solid State Sci., 2004, 6, 1053; NMR Crystallography, ed. R. K. Harris, R. E. Wasylishen and M. J. Duer, John Wiley & Sons Ltd., Chichester, West Sussex, UK, 2009. ¨rnroos, ¨rgi, D. Chernyshov and K. W. To 6 T. Aree, H.-B. Bu J. Phys. Chem. A, 2014, 118, 9951; A. Hofstetter and L. Emsley, J. Am. Chem. Soc., 2017, 139, 2573. 7 R. B. Hammond, C. Ma, K. J. Roberts, P. Y. Ghi and R. K. Harris, J. Phys. Chem. B, 2003, 107, 11820; D. V. Dudenko, P. A. Williams, C. E. Hughes, O. N. Antzutkin, S. P. Velaga, S. P. Brown and K. D. M. Harris, J. Phys. Chem. C, 2013, ¨deker and G. Brunklaus, Solid State Nucl. 117, 12258; D. Lu Magn. Reson., 2015, 65, 29; A. E. Watts, K. Maruyoshi, C. E. Hughes, S. P. Brown and K. D. M. Harris, Cryst. Growth Des., 2016, 16, 1798. 8 S. M. Santos, J. Rocha and L. Mafra, Cryst. Growth Des., 2013, 13, 2390. 9 C. Martineau, A. Cadiau, B. Bouchevreau, J. Senker, F. Taulelle and K. Adil, Dalton Trans., 2012, 41, 6232. 10 R. K. Harris, Analyst, 2006, 131, 351; A. Othman, J. S. O. Evans, I. Radosavljevic-Evans, R. K. Harris and P. Hodgkinson, J. Pharm. Sci., 2007, 96, 1380; E. Salager, R. S. Stein, C. J. Pickard, B. Elena and L. Emsley, Phys. Chem. Chem. Phys., 2009, 11, 2610; M. R. Chierotti and R. Gobetto, CrystEngComm, 2013, 15, 8599; T. Pawlak, M. Jaworska and M. J. Potrzebowski, Phys. Chem. Chem. Phys., 2013, 15, 3137; R. Koike, K. Higashi, N. Liu, W. Limwikrant, K. Yamamoto and K. Moribe, Cryst. Growth Des., 2014, 14, 4510; M. Sardo, S. M. Santos, A. A. Babaryk, ´pez, I. Alkorta, J. Elguero, R. M. Claramunt and L. Mafra, C. Lo

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16659

View Article Online

PCCP

11

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

12 13

14

15

16

17

18

19 20

Solid State Nucl. Magn. Reson., 2015, 65, 49; E. Wielgus, P. Paluch, J. Frelek, W. J. Szczepek and M. J. Potrzebowski, J. Pharm. Sci., 2015, 104, 3883; P. Paluch, T. Pawlak, M. Oszajca, W. Lasocha and M. J. Potrzebowski, Solid State Nucl. Magn. Reson., 2015, 65, 2. C. J. Pickard, E. Salager, G. Pintacuda, B. Elena and L. Emsley, J. Am. Chem. Soc., 2007, 129, 8932. C. M. Widdifield, H. Robson and P. Hodgkinson, Chem. Commun., 2016, 52, 6685. J. Leclaire, G. Poisson, F. Ziarelli, G. Pepe, F. Fotiadu, F. M. Paruzzo, A. J. Rossini, J.-N. Dumez, B. Elena-Herrmann and L. Emsley, Chem. Sci., 2016, 7, 4379. M. Baias, A. Lesage, S. Aguado, J. Canivet, V. Moizan-Basle, N. Audebrand, D. Farrusseng and L. Emsley, Angew. Chem., Int. Ed., 2015, 54, 5971; J. Xu, V. V. Terskikh, Y. Chu, A. Zheng and Y. Huang, Chem. Mater., 2015, 27, 3306. D. H. Brouwer, R. J. Darton, R. E. Morris and M. H. Levitt, J. Am. Chem. Soc., 2005, 127, 10365; D. H. Brouwer, I. L. Moudrakovski, R. J. Darton and R. E. Morris, Magn. Reson. Chem., 2010, 48, S113; D. H. Brouwer, Solid State Nucl. Magn. Reson., 2013, 51–52, 37; D. H. Brouwer and M. Horvath, Solid State Nucl. Magn. Reson., 2015, 65, 89; C. Martineau, S. Vial, D. Barth, F. Quessette and F. Taulelle, Solid State Nucl. Magn. Reson., 2015, 65, 84. J. Dutour, N. Guillou, C. Huguenard, F. Taulelle, C. Mellot´re ´y, Solid State Sci., 2004, 6, 1059; Draznieks and G. Fe J. F. Colin, T. Bataille, S. E. Ashbrook, N. Audebrand, L. Le `s, J. Y. Pivan and E. Le Fur, Inorg. Chem., 2006, 45, 6034; Polle C. M. Widdifield and D. L. Bryce, Phys. Chem. Chem. Phys., 2009, 11, 7120; F. A. Perras and D. L. Bryce, J. Phys. Chem. C, 2012, 116, 19472; B. Bouchevreau, C. Martineau, C. Mellot´bosc, O. Lafon, Draznieks, A. Tuel, M. R. Suchomel, J. Tre J.-P. Amoureux and F. Taulelle, Chem. – Eur. J., 2013, 19, 5009; C. Martineau, Solid State Nucl. Magn. Reson., 2014, 63–64, 1; C. M. Widdifield, F. A. Perras and D. L. Bryce, Phys. Chem. Chem. Phys., 2015, 17, 10118; C. P. Romao, F. A. Perras, U. Werner-Zwanziger, J. A. Lussier, K. J. Miller, C. M. Calahoo, J. W. Zwanziger, M. Bieringer, B. A. Marinkovic, D. L. Bryce and M. A. White, Chem. Mater., 2015, 27, 2633. C. Luchinat, G. Parigi, E. Ravera and M. Rinaldelli, J. Am. Chem. Soc., 2012, 134, 5006; L. J. Mueller and M. F. Dunn, Acc. Chem. Res., 2013, 46, 2008; D. Tietze, S. Voigt, D. Mollenhauser and G. Buntkowsky, Appl. Magn. Reson., 2014, 45, 841. X. Xue and M. Kanzaki, J. Am. Ceram. Soc., 2009, 92, 2803; P. Guerry, M. E. Smith and S. P. Brown, J. Am. Chem. Soc., 2009, 131, 11861; D. H. Brouwer, S. Cadars, J. Eckert, Z. Liu, O. Terasaki and B. F. Chmelka, J. Am. Chem. Soc., 2013, 135, 5641; D. M. Dawson and S. E. Ashbrook, J. Phys. Chem. C, 2014, 118, 23285; S. E. Ashbrook, M. R. Mitchell, S. Sneddon, R. F. Moran, M. de los Reyes, G. R. Lumpkin and K. R. Whittle, Phys. Chem. Chem. Phys., 2015, 17, 9049. M. Baias, J.-N. Dumez, P. H. Svensson, S. Schantz, G. M. Day and L. Emsley, J. Am. Chem. Soc., 2013, 135, 17501. S. L. Price, Chem. Soc. Rev., 2014, 43, 2098.

16660 | Phys. Chem. Chem. Phys., 2017, 19, 16650--16661

Paper

21 M. A. Neumann, J. van de Streek, F. P. A. Fabbiani, P. Hidber and O. Grassmann, Nat. Commun., 2015, 6, 7793. 22 R. K. Harris, P. Y. Ghi, R. B. Hammond, C. Y. Ma, K. J. Roberts, J. R. Yates and C. J. Pickard, Magn. Reson. Chem., 2006, 44, 325. 23 M. Baias, C. M. Widdifield, J.-N. Dumez, H. P. G. Thompson, T. G. Cooper, E. Salager, S. Bassil, R. S. Stein, A. Lesage, G. M. Day and L. Emsley, Phys. Chem. Chem. Phys., 2013, 15, 8069. 24 J. K. Harper and D. M. Grant, Cryst. Growth Des., 2006, 6, 2315; J. K. Harper, J. A. Doebbler, E. Jacques, D. M. Grant and R. B. Von Dreele, J. Am. Chem. Soc., 2010, 132, 2928; J. K. Harper, D. Tishler, D. Richardson, J. Lokvam, R. Pendrill and G. Widmalm, J. Phys. Chem. A, 2013, 117, 5534. 25 B. Elena and L. Emsley, J. Am. Chem. Soc., 2005, 127, 9140; B. Elena, G. Pintacuda, N. Mifsud and L. Emsley, J. Am. Chem. Soc., 2006, 128, 9555. 26 M. Schulz-Dobrick, T. Metzroth, H. W. Spiess, J. Gauss and I. Schnell, ChemPhysChem, 2005, 6, 315; J. P. Bradley, S. P. Velaga, O. N. Antzutkin and S. P. Brown, Cryst. Growth Des., 2011, 11, 3463; P. Thureau, G. Mollica, F. Ziarelli and S. Viel, Annu. Rep. NMR Spectrosc., 2014, 82, 217. 27 G. Mollica, M. Dekhil, F. Ziarelli, P. Thureau and S. Viel, Angew. Chem., Int. Ed., 2015, 54, 6028. 28 M. Fageras, Z. Taib, D. Cunoosamy, K. Elsby, P. Jansson and N. Patel, Am. J. Respir. Crit. Care Med., 2015, 191, A2718. 29 T. Vos, et al., Lancet, 2012, 380, 2163. 30 GBD 2013 Mortality and Causes of Death Collaborators, Lancet, 2015, 385, 117. 31 S. Kaplan, A. Pines, R. G. Griffin and J. S. Waugh, Chem. Phys. Lett., 1974, 25, 78; R. G. Griffin, A. Pines and J. S. Waugh, J. Chem. Phys., 1975, 63, 3676. 32 A. L. Spek, J. Appl. Crystallogr., 2003, 36, 7; A. L. Spek, Acta Crystallogr., Sect. D: Biol. Crystallogr., 2009, 65, 148. 33 J. Contreras-Garcı´a, E. R. Johnson, S. Keinan, R. Chaudret, J.-P. Piquemal, D. N. Beratan and W. Yang, J. Chem. Theory Comput., 2011, 7, 625; S. O. Nilsson Lill, Phys. Chem. Chem. Phys., 2011, 13, 16022. 34 P. Hodgkinson, High-Resolution 1H NMR Spectroscopy of Solids, Annu. Rep. NMR Spectrosc., 2011, 72, 185. 35 Y. Q. Ye, M. Malon, C. Martineau, F. Taulelle and Y. Nishiyama, J. Magn. Reson., 2014, 239, 75. 36 M. E. Halse, J. Schlagnitweit and L. Emsley, Isr. J. Chem., 2014, 54, 136. ¨rster and H. J. M. de Groot, J. Magn. 37 B.-J. van Rossum, H. Fo Reson., 1997, 124, 516. ¨pe and L. Emsley, Chem. Phys. Lett., 2004, 38 B. Elena, G. De Pae 398, 532. 39 R. E. Taylor, Concepts Magn. Reson., Part A, 2004, 22, 37. 40 X. Wu, S. T. Burns and K. W. Zilm, J. Magn. Reson., Ser. A, 1994, 111, 29. 41 D. Schweitzer and H. W. Spiess, J. Magn. Reson., 1974, 15, 529; R. D. Curtis, G. H. Penner, W. P. Power and R. E. Wasylishen, J. Phys. Chem., 1990, 94, 4000. 42 B. M. Fung, A. K. Khitrin and K. Ermolaev, J. Magn. Reson., 2000, 142, 97.

This journal is © the Owner Societies 2017

View Article Online

Published on 02 June 2017. Downloaded by Durham University Library on 11/07/2017 16:03:31.

Paper

43 M. E. Halse and L. Emsley, Phys. Chem. Chem. Phys., 2012, 14, 9121. 44 A. Lesage, P. Charmont, S. Steuernagel and L. Emsley, J. Am. Chem. Soc., 2000, 122, 9739; D. H. Zhou, G. Shah, C. Mullen, D. Sandoz and C. M. Rienstra, Angew. Chem., Int. Ed., 2009, 48, 1253; S. M. Althaus, K. Mao, J. A. Stringer, T. Kobayashi and M. Pruski, Solid State Nucl. Magn. Reson., 2014, 57-58, 17. 45 A. Lesage, C. Auger, S. Caldarelli and L. Emsley, J. Am. Chem. Soc., 1997, 119, 7867; A. Lesage, M. Bardet and L. Emsley, J. Am. Chem. Soc., 1999, 121, 10987. 46 T. Maly, G. T. Debelouchina, V. S. Bajaj, K.-N. Hu, C.-G. Joo, M. L. Mak-Jurkauskas, J. R. Sirigiri, P. C. A. van der Wel, J. Herzfeld, R. J. Temkin and R. G. Griffin, J. Chem. Phys., 2008, 128, 052211; Q. Z. Ni, E. Daviso, T. V. Can, E. Markhasin, S. K. Jawla, T. M. Swager, R. J. Temkin, J. Herzfeld and R. G. Griffin, Acc. Chem. Res., 2013, 46, 1933; A. J. Rossini, A. Zagdoun, M. Lelli, A. Lesage, ´ret and L. Emsley, Acc. Chem. Res., 2013, 46, 1942. C. Cope ¨lder, 47 A. J. Rossini, A. Zagdoun, F. Hegner, M. Schwarzwa ´ret, A. Lesage and L. Emsley, J. Am. Chem. D. Gajan, C. Cope Soc., 2012, 134, 16899; H. Takahashi, D. Lee, L. Dubois, ¨pe, Angew. Chem., Int. Ed., M. Bardet, S. Hediger and G. De Pae 2012, 51, 11766; A. J. Rossini, C. M. Widdifield, A. Zagdoun, ¨lder, C. Cope ´ret, A. Lesage and M. Lelli, M. Schwarzwa L. Emsley, J. Am. Chem. Soc., 2014, 136, 2324. ¨pe, A. Lesage, S. Steuernagel and L. Emsley, 48 G. De Pae ChemPhysChem, 2004, 5, 869. 49 R. K. Harris, S. Cadars, L. Emsley, J. R. Yates, C. J. Pickard, R. K. R. Jetti and U. J. Griesser, Phys. Chem. Chem. Phys., 2007, 9, 360. 50 R. K. Harris, S. A. Joyce, C. J. Pickard, S. Cadars and L. Emsley, Phys. Chem. Chem. Phys., 2006, 8, 137. `vre, J. W. Gramlich, R. J. C. 51 I. L. Barnes, T. L. Chang, P. De Bie Hageman, N. E. Holden, T. J. Murphy, K. J. R. Rosman and M. Shima, Pure Appl. Chem., 1991, 63, 991. 52 Z. Gan, J. Am. Chem. Soc., 1992, 114, 8307; J. Z. Hu, D. W. Alderman, C. Ye, R. J. Pugmire and D. M. Grant, J. Magn. Reson., Ser. A, 1993, 105, 82; S. L. Gann, J. H. Baltisberger and A. Pines, Chem. Phys. Lett., 1993, 210, 405.

This journal is © the Owner Societies 2017

PCCP

53 J. Herzfeld and A. E. Berger, J. Chem. Phys., 1980, 73, 6021. 54 C. J. Pickard and F. Mauri, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 63, 245101; J. R. Yates, C. J. Pickard and F. Mauri, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 024401. 55 E. Salager, G. M. Day, R. S. Stein, C. J. Pickard, B. Elena and L. Emsley, J. Am. Chem. Soc., 2010, 132, 2564. 56 T. Polenova, R. Gupta and A. Goldbourt, Anal. Chem., 2015, 87, 5458. 57 K. R. Thurber and R. Tycko, J. Magn. Reson., 2009, 196, 84. 58 G. Metz, X. Wu and S. O. Smith, J. Magn. Reson., Ser. A, 1994, 110, 219. 59 S. R. Hartman and E. L. Hahn, Phys. Rev., 1962, 128, 2042; A. Pines, M. G. Gibby and J. S. Waugh, J. Chem. Phys., 1973, 59, 569. 60 S. J. Clark, M. D. Segall, C. J. Pickard, P. J. Hasnip, M. I. J. Probert, K. Refson and M. C. Payne, Z. Kristallogr., 2005, 220, 567. 61 J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865; J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1997, 78, 1396. 62 D. Vanderbilt, Phys. Rev. B: Condens. Matter Mater. Phys., 1990, 41, 7892. 63 H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Solid State, 1976, 13, 5188. 64 A. Tkatchenko and M. Scheffler, Phys. Rev. Lett., 2009, 102, 073005. 65 L.-E. Briggner, R. Hendrickx, L. Kloo, J. Rosdahl and P. H. Svensson, ChemMedChem, 2011, 6, 60; S. O. Nilsson Lill and A. Broo, Cryst. Growth Des., 2014, 14, 3704. 66 M. A. Neumann, F. J. J. Leusen and J. Kendrick, Angew. Chem., Int. Ed., 2008, 47, 2427. 67 J. L. Banks, H. S. Beard, Y. Cao, A. E. Cho, W. Damm, R. Farid, A. K. Felts, T. A. Halgren, D. T. Mainz, J. R. Maple, R. Murphy, D. M. Philipp, M. P. Repasky, L. Y. Zhang, B. J. Berne, R. A. Friesner, E. Gallicchio and R. M. Levy, J. Comput. Chem., 2005, 26, 1752.

Phys. Chem. Chem. Phys., 2017, 19, 16650--16661 | 16661