Search for a predicted hydrogen bonding motif - A multidisciplinary investigation into the polymorphism of 3-azabicyclo[3.3.1]nonane-2,4-dione

Article abstract of DOI:10.1021/ja0687466

The predictions of the crystal structure of 3-azabicyclo[3.3.1]nonane-2,4- dione submitted in the 2001 international blind test of crystal structure prediction (CSP2001) led to the conclusion that crystal structures containing an alternative hydrogen bond

Full text of DOI:10.1021/ja0687466

Published on Web 03/01/2007
Search for a Predicted Hydrogen Bonding Motif - A
Multidisciplinary Investigation into the Polymorphism of
3-Azabicyclo[3.3.1]nonane-2,4-dione
Ashley T. Hulme,† Andrea Johnston,‡ Alastair J. Florence,‡ Phillipe Fernandes,‡
Kenneth Shankland,§ Colin T. Bedford,† Gareth W. A. Welch,† Ghazala Sadiq,|
,#
Delia A. Haynes, W. D. Samuel Motherwell, Derek A. Tocher,† and
Sarah L. Price*,†
Contribution from the Christopher Ingold Laboratory, Department of Chemistry, UniVersity
College London, 20 Gordon Street, London WC1H 0AJ, U.K., Strathclyde Institute of Pharmacy
and Biomedical Sciences, 27 Taylor Street, UniVersity of Strathclyde, Glasgow G4 0NR, U.K.,
ISIS facility, CCLRC Rutherford Appleton Laboratory, Chilton, Didcot, Oxon OX11 1QX, U.K.,
Molecular Materials Centre, School of Chemical Engineering and Analytical Sciences,
UniVersity of Manchester, SackVille Street, Manchester M60 1QD, U.K., and Cambridge
Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, U.K.
Received December 6, 2006; E-mail: s.l.price@ucl.ac.uk
Abstract: The predictions of the crystal structure of 3-azabicyclo[3.3.1]nonane-2,4-dione submitted in the
2001 international blind test of crystal structure prediction (CSP2001) led to the conclusion that crystal
structures containing an alternative hydrogen bonded dimer motif were energetically competitive with the
known catemer-based structure. Here we report an extensive search for a dimer-based crystal structure.
Using an automated polymorph screen a new catemer-based metastable polymorph (form 2) and two new
catemer-based solvates were found, and concurrent thermal studies reproduced form 2 and identified a
plastic phase (form 3), whose powder X-ray diffraction pattern was consistent with the cubic space group
I23 (a ) 7.5856(1) Å). Computational studies on the monomer showed that the imide N-H was a weak
hydrogen bond donor, rationalizing the occurrence of the plastic phase which involved the breaking of all
hydrogen bonds, and modeling of small clusters showed that dimers could easily reorganize to give the
catemer. FTIR spectra confirmed the weakness of the hydrogen bond, with the solute showing no self-
assembly in solution. It is concluded that the weakness of the N-H donor, coupled with the globular shape
of the molecule, allows unusually facile transformation between alternative hydrogen bonding motifs during
aggregation and nucleation.
Introduction
was observed that for two of the test molecules the predicted
structures consistently showed a hydrogen bonding motif
Computational methods of organic crystal structure prediction
can only be definitively evaluated when all experimental
polymorphs are known. The potential benefits of such compu-
tational methods to the pharmaceutical1 and speciality chemicals
industries, as well as fundamental scientific interest, have led
the Cambridge Crystallographic Data Centre to organize three
international blind tests of crystal structure prediction.2-4 In the
participants’ discussion after the 2001 blind test3 (CSP2001) it
alternative to that found in the released experimental crystal
structures. This led to the belief that there was the distinct
potential for new polymorphs of these molecules to be discov-
ered. For one of the two test molecules, the flexible molecule
6-amino-2-phenylsulfonylimino-1,2-dihydropyridine, a new poly-
morph was subsequently discovered with the expected alterna-
tive hydrogen bonding motif.5 Here we report an extensive
search for polymorphs of the second of these test molecules,
the rigid molecule 3-azabicyclo[3.3.1]nonane-2,4-dione (I,
Figure 1a).
From the results of CSP2001 it was found that the imide
group of I could principally form two hydrogen bond motifs,
an R22(8) dimer (Figure 1b) and a catemer (Figure 1c). In
CSP2001 many of the participants had the same dimer-based
structure within their three submitted crystal structure predic-
tions, and it was even the first ranked structure submitted by
† University College London.
‡ University of Strathclyde.
§ CCLRC Rutherford Appleton Laboratory.
| University of Manchester.
Cambridge Crystallographic Data Centre.
# Current address: School of Chemistry, University of KwaZulu-Natal,
Howard College Campus, Durban, 4041, South Africa.
(1) Price, S. L. AdV. Drug DeliVery ReV. 2004, 56, 301-319.
(2) Lommerse, J. P. M.; Motherwell, W. D. S.; Ammon, H. L.; Dunitz, J. D.;
Gavezzotti, A.; Hofmann, D. W. M.; Leusen, F. J. J.; Mooij, W. T. M.;
Price, S. L.; Schweizer, B.; Schmidt, M. U.; van Eijck, B. P.; Verwer, P.;
Williams, D. E. Acta Crystallogr., Sect. B 2000, 56, 697-714.
(3) Motherwell, W. D. S. et al. Acta Crystallogr., Sect. B 2002, 58, 647-661.
(4) Day, G. M. et al. Acta Crystallogr., Sect. B 2005, 61, 511-527.
(5) Jetti, R. K. R.; Boese, R.; Sarma, J.; Reddy, R. S.; Vishweshwar, P.;
Desiraju, G. R. Angew. Chem., Int. Ed. 2003, 42, 1963-1967.
9
10.1021/ja0687466 CCC: $37.00 © 2007 American Chemical Society
J. AM. CHEM. SOC. 2007, 129, 3649-3657
3649

A R T I C L E S
Hulme et al.
Table 1. Outline of Conditions Included in the Automated Parallel
Crystallization Search on Ia
crystallization
conditionb
mass of
crystallization
method
agitation
(rpm)
solid added
1
2
3
excess
100 mg
100 mg
coolingc to 288 K
cooling to 288 K
evaporation
1000
850
500
Figure 1. (a) Molecular diagram for 3-azabicyclo[3.3.1]nonane-2,4-dione
(I); (b) a representative R22(8) dimer motif taken from a predicted structure;
(c) the chain motif (found in form 1).
a Conditions 1 and 2 used the complete library of 67 solvents. Condition
3 used a reduced library of 48 solvents. b Solutions were prepared by adding
solid to 2 mL of each solvent at the selected temperature (Tprep), prior to
being filtered automatically into a clean crystallization vessel. Postdisso-
lution, in-line filtration excludes undissolved particulates larger than 10 µm.
A short heating cycle is then applied to the postfiltration solutions to remove
smaller seeds before crystallization is induced. Condition 1, an excess of
solid (typically >500 mg) was added such that the solution was saturated
at Tprep (see Supporting Information for full details of individual crystal-
lizations). Condition 2, 100 mg of I were added to each 2 mL of solvent at
both of the participants who submitted the correct catemer-based
structure (form 1) as their second and third choices, respec-
tively.3 The majority of submissions were based on a compu-
tational search for the global minimum in the lattice energy,
which found a large number of both dimer- and catemer-based
structures to be energetically competitive leading to the sup-
position that a polymorph of I could well exist based on the
R22(8) dimer hydrogen bonding motif. An alternative crystal
structure prediction method6 compared a set of computationally
generated crystal structures with a carefully chosen subset of
the experimental crystal structures from the Cambridge Structure
Database (CSD)7 that contained the same imide functional group
and concluded that the correct catemer-based structure was more
likely to occur than dimer-based structures. With the implicit
influence of any kinetic factors on the structures present in the
CSD, this suggests6 that there may be a kinetic reason why the
catemer-based structure is observed.
I was synthesized in sufficient quantities (ca. 35 g) to perform
an automated parallel solvent crystallization screen to discover
alternative polymorphs of I, which could be based on the R22(8)
dimer motif. We report the crystallization conditions that
produced two solvates of I and the thermal methods that yielded
two new polymorphs of I: a metastable catemer-based poly-
morph related to form 18 and a high-temperature plastic phase.
Given that the extensive experimental searches failed to find
the alternative dimer-based hydrogen bonding motif, further
investigations were undertaken to determine the kinetic factors
that preclude the formation of the dimer-based polymorphs for
this molecule, while both dimer- and catemer-based polymorphs
can be observed in other systems. The formation of the plastic
phase suggests that hydrogen bonding in I is unusually weak
for a NsH‚‚‚OdC conventional hydrogen bond and easily
disrupted, which was confirmed by examination of the electro-
static potential which showed that the NsH in the imide
functional group is a weak hydrogen bond donor. Cluster
calculations were used to explore the stability of the R22(8)
dimers with respect to disruption by the introduction of further
molecules, and from these it was concluded that the barrier to
rearrangement from the dimer to catemer motifs is negligible.
FTIR spectra of I in both polar and nonpolar solutions showed
no self-assembly of molecules of I, for which strong hydrogen
bonds would be required. Consequently, a rationalization for
the lack of a dimer-based structure is developed by a combina-
tion of experimental and computational studies.
T
prep for each solvent group, varying the rate and extent of supersaturation
obtained during cooling. For Condition 3, 100 mg of I were added to 2 mL
of solvent at Tprep and heated solutions were transferred to clean glassware
and held at elevated temperature with agitation. Recrystallization was then
induced by evaporation. Evaporation temperatures for each solution ranged
from 313 to 348 K depending on the solvent boiling point (see Supporting
Information). c The cooling rate used for Conditions 1 and 2 was ap-
proximately 3.5 K min-1. The solvent library (67 solvents, see Supporting
Information for details) was ranked according to solvent boiling point and
divided into four groups of ca. 16 solvents each. Within each solvent group,
the temperature at which solutions were prepared (Tprep) is equivalent to
Tprep ) (minimum boiling point within group - 10) K.
Perkin9 was used to enrich mixed cis/trans 1,3-cyclohexanedicarboxylic
acid to the pure cis acid. A modified version of the method of Hall10
was used to give the final product. The powder X-ray pattern of the
synthesized product was compared to the simulated powder pattern from
the single-crystal structure8 of form 1, and they were found to match.
Full details are given in the Supporting Information.
Thermal Analysis and Capillary Variable-Temperature XRPD.
The thermal properties of I were studied using a combination of hot-
stage microscopy (HSM; Meiji Techno ML9430 polarizing microscope,
with JVC color video camera and Linkam Scientific Instruments hot-
stage with TMS 94 controller) and simultaneous differential scanning
calorimetry (DSC) and thermogravimetric analysis (TGA) thermal
analysis (STA; Netzsch STA 449-C Jupiter instrument). For STA
analysis samples were placed in pierced 10 µL aluminum pans and
heated from 298 to 473 K at 10 K min-1
.
Data for structural analysis were collected using variable-temperature
capillary XRPD. Samples were lightly ground with an agate mortar
and pestle and filled into 0.7 mm borosilicate glass capillaries prior to
being mounted and aligned on a Bruker-AXS D8 Advance powder
diffractometer. The instrument was equipped with transmission capillary
geometry, a primary monochromator (Cu KR1; λ) 1.540 56 Å), and a
Vantec position sensitive detector. The plastic phase of I was prepared
in situ by heating a sample of form 1 to 420 K using an Oxford
Cryosystems 700 Plus Series cryostream device. Form 2 was obtained
by rapidly cooling the plastic phase from 420 to 250 K inside a 0.7
mm borosilicate glass capillary.
Automated Solvent Crystallization Screen. I was recrystallized
from solution, using an automated parallel crystallization approach
involving a total of 182 individual crystallizations implemented on a
Chemspeed Accelerator SLT100 platform.11,12 The search utilized a
library of 67 solvents covering a diverse range of physicochemical
properties. Crystallization was induced in filtered solutions by either
controlled cooling or evaporation (Table 1). Upon recrystallization,
suspended samples were reclaimed by filtration and transferred to a
Method
Synthesis. A total of 37.7 g of 3-azabicyclo[3.3.1]nonane-2,4-dione
was synthesized in a two-step process. The procedure of Goodwin and
(9) Goodwin, W.; Perkin, W. H. J. Chem. Soc. 1905, 87, 841-855.
(10) Hall, H. K., Jr. J. Am. Chem. Soc. 1958, 80, 6412-6420.
(11) Florence, A. J.; Johnston, A.; Price, S. L.; Nowell, H.; Kennedy, A. R.;
Shankland, N. J. Pharm. Sci. 2006, 95, 1918-1930.
(12) Florence, A.; Johnston, A.; Fernandes, P.; Shankland, N.; Shankland, K.
J. Appl. Crystallogr. 2006, 39, 922-924.
(6) Sarma, J. A. R. P.; Desiraju, G. R. Cryst. Growth Des. 2002, 2, 93-100.
(7) Allen, F. H. Acta Crystallogr., Sect. B 2002, 58, 380-388.
(8) Howie, R. A.; Skakle, J. M. S. Acta Crystallogr., Sect. E 2001, 57, o822-
o824.
9
3650 J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007

Search for a Predicted Hydrogen Bonding Motif
A R T I C L E S
multiposition sample holder for sample identification using multisample
foil transmission XRPD13 (full experimental details are provided in the
Supporting Information).
Solid State and Solution Infrared Spectroscopy. Solid state spectra
of both I and succinimide were recorded at room temperature using a
Perkin-Elmer Spectrum One FTIR-ATR spectrometer fitted with a
diamond ATR crystal. Spectra of saturated solutions of I and succin-
imide in pentane, chloroform, water, and methanol were recorded at
room temperature in ATR mode using an Atavar 360 ESP spectrometer
fitted with a diamond composite ATR crystal in conjunction with
Nicolet’s OMNIC software. All solvents were spectroscopic grade.
Computational Modeling. The molecular structure was ab initio
optimized at the MP2 6-31G(d,p) level, using Gaussian03.14 This
electron density was analyzed using GDMA2.215 to provide a distributed
multipole description which was used to calculate the electrostatic
contribution to the intermolecular potential, by including all terms in
the atom-atom multipole series up to Rik-5. All other terms in the
intermolecular potential were modeled by an isotropic atom-atom
empirical potential:
Figure 2. Differential scanning calorimetry data exhibiting the phase change
to the plastic phase and the melting event. Heating was from 303 to 473 K
at 10 K min-1 (red), followed by cooling back to 303 K at a rate of 10 K
min-1 (blue).
Sadlej basis set,27 to test the sensitivity to variations in theoretically
reasonable electrostatic models.28
Results
6
U ) Σ((AιιAκκ)1/2 exp(-(Bιι + Bκκ)Rik/2) - (CιιCκκ)1/2/Rik
)
Thermal Analysis. DSC on form 1 identified an endotherm
(410 K) prior to melting (463 K) which corresponds to a
transition to a crystalline plastic phase, form 3 (Figure 2). HSM
showed a loss of birefringence at ca. 418 K accompanying the
transition with the crystallites retaining their morphology in the
plastic state until a further increase in temperature gave rise to
plastic flow (Figure 3). XRPD data for form 3 were indexed to
give a cubic lattice (DICVOL-9129) which was confirmed by a
Pawley-type fit30 in TOPAS31 to be consistent with a cubic
lattice, space group I23, a ) 7.5856(1) Å, V ) 436.49(1) Å3,
Rwp ) 0.016 (Figure 4). This represents a volume per molecule
(218.25 Å3) equivalent to a 15% increase on that for form 1 at
298 K. Cubic symmetry is characteristic of plastic crystals,32,33
and the transition from an ordered crystal into a plastically
crystalline state is usually accompanied by a decrease of 10-
15% in density,34 as such the body centered cubic structure is
physically consistent with form 3 being a plastic crystal.
where atom i in molecule 1 is of type ι and atom k in molecule 2 is of
type κ, using a set of parameters16 for atomic types C, N, O, HC, and
HN which had been derived by empirical fitting to a range of crystal
structures and heats of sublimation.
This intermolecular potential energy surface was used in cluster
calculations performed using ORIENT.17 This program was also used
to calculate the electrostatic potential arising from the distributed
multipole model on the solvent accessible surface of I, defined as
1.4 Å above the van der Waals surface,18 as defined by the Bondi radii19
with a zero radius on the polar hydrogen atom (HN).20
A representative crystal structure prediction search (see Supporting
Information) was also carried out with this intermolecular potential, to
confirm that, as in SLP’s entry in CSP2001, it generated approximately
equi-energetic dimer- and catemer-based crystal structures. The program
MOLPAK21 was used to generate densely packed hypothetical crystal
structures using the same rigid ab initio optimized molecule conforma-
tion. The physical properties of this small, illustrative set of predicted
crystal structures were also calculated: the elastic tensor and intermo-
lecular phonons22,23 were calculated in the rigid-body harmonic ap-
proximation and used to estimate24 the free energy at 298 K; the
morphologies of the low-energy crystals were estimated using the
attachment energy model and used to estimate25 the relative rates of
growth of the different crystals from the vapor.
DSC analysis of the form 1 f form 3 transition based on
seven repeats shows an average onset temperature of 408.6 (
0.4 °C, giving ∆Htrs ) 16.3 ( 0.7 kJ mol-1 and ∆Strs ) 39.9
( 1.9 J K-1 mol-1. For the form 3 melt, the average onset
temperature from six measurements is 463.6 ( 0.3 °C, with
∆Hfus ) 3.27 ( 0.2 kJ mol-1 and ∆Sfus ) 7.05 ( 0.43 J K-1
mol-1. Thus, this phase transformation satisfies a key thermo-
Most of the calculations were repeated using the distributed
multipoles obtained from a PBE026 charge density calculated with a
dynamic criterion for plastic-crystal formation,34 i.e., ∆Sfus
21 J mol-1 K-1, and has Σ∆trsS/∆fusS ) 5.7 > 1.
<
(13) Florence, A. J.; Baumgartner, B.; Weston, C.; Shankland, N.; Kennedy,
A. R.; Shankland, K.; David, W. I. F. J. Pharm. Sci. 2003, 92, 1930-
1938.
(14) Frisch, M. J., et al. Gaussian 03; Gaussian Inc.: Wallingford, CT, 2003
(15) Stone, A. J. J. Chem. Theory Comput. 2005, 1, 1128-1132.
(16) Coombes, D. S.; Price, S. L.; Willock, D. J.; Leslie, M. J. Phys. Chem.
1996, 100, 7352-7360.
(17) Stone, A. J.; Dullweber, A.; Engkvist, O.; Fraschini, E.; Hodges, M. P.;
Meredith, A. W.; Nutt, D. R.; Popelier, P. L. A.; Wales, D. J. ORIENT,
version 4.6; University of Cambridge: 2006
Quench cooling a sample of form 3 held in a capillary from
425 to 245 K using a cryostream device produced a new
metastable crystalline modification (form 2). The crystal
structure was solved at 1.67 Å resolution by simulated annealing
from laboratory powder data collected at 250 K. Subsequent
Rietveld refinement yielded an Rwp of 0.070 to 1.54 Å
(18) Lee, B.; Richards, F. M. J. Mol. Biol. 1971, 55, 379.
(19) Bondi, A. J. Phys. Chem. 1964, 68, 441-451.
(20) Buckingham, A. D.; Fowler, P. W. Can. J. Chem. 1985, 63, 2018-2025.
(21) Holden, J. R.; Du, Z. Y.; Ammon, H. L. J. Comput. Chem. 1993, 14, 422-
437.
(27) Sadlej, A. J. Collect. Czech. Chem. C 1988, 53, 1995-2016.
(28) Misquitta, A. J.; Szalewicz, K. J. Chem. Phys. 2005, 122, art-214103.
(29) Boultif, A.; Loue¨r, D. J. Appl. Crystallogr. 1991, 24, 987-993.
(30) Pawley, G. S. J. Appl. Crystallogr. 1981, 14, 357-361.
(31) Coelho, A. A. Topas user manual, version 3.1 ed.; Bruker AXS GmbH:
Karlsruhe, Germany, 2003.
(22) Day, G. M.; Price, S. L.; Leslie, M. J. Phys. Chem. B 2003, 107, 10919-
10933.
(23) Day, G. M.; Price, S. L.; Leslie, M. Cryst. Growth Des. 2001, 1, 13-26.
(24) Anghel, A. T.; Day, G. M.; Price, S. L. CrystEngComm 2002, 4, 348-
355.
(25) Coombes, D. S.; Catlow, C. R. A.; Gale, J. D.; Rohl, A. L.; Price, S. L.
Cryst. Growth Des. 2005, 5, 879-885.
(26) Adamo, C.; Barone, V. J. Chem. Phys. 1999, 110, 6158-6170.
(32) Dunning, W. J. J. Phys. Chem. Solids 1961, 18, 21-27.
(33) Dunning, W. The Crystal Structure of Some Plastic and Related Crystals.
In Plastically Crystalline State: Orientationally Disordered Crystals;
Sherwood, J. N., Ed.; John Wiley & Sons: Chichester, 1979; pp 1-37.
(34) Bazyleva, A. B.; Kabo, G. J.; Blokhin, A. V. Physica B 2006, 383, 243-
252.
9
J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007 3651

A R T I C L E S
Hulme et al.
Figure 3. Hot stage microscopy: (a) crystals of form 1 at room temperature; (b) plastic phase at 418 K; (c) on further heating plastic flow is observed; (d)
melting of plastic phase at 468 K; (e) liquid droplets condense to plastic phase below 463 K; (f) plastic phase converts to form 1 at 381 K.
resolution.35 Form 2 contains two molecules in the asymmetric
unit and has a unit cell that closely approximates that of form
1 doubled along the c-axis and may well be an example of a
metastable form which represents a stage in the pathway from
the plastic phase toward the more stable form 1.36 Indeed, on
standing, form 2 transforms fully to form 1 within an hour at
room temperature, showing this to be an example of a higher
Z′ metastable polymorph, with pseudo-symmetry, that converts
readily to a more stable lower Z′ form.37 Form 2 was also
recrystallized from solution, though from only two of the
solvents used in the search (see below).
Automated Solvent Crystallization Screen. The automated
solvent screen produced 172 crystalline samples from the total
of 182 crystallization experiments. The condition and outcome
of each individual crystallization is given in the Supporting
Information. In summary, the majority (162/172) of crystal-
lizations gave the known form 1, two crystallizations (from CCl4
and CHCl3 under condition 1, Table 1) gave the metastable form
2, and acetic acid (3/172) and 1-methylnapthalene (2/172) each
yielded novel solvates. The structure of the acetic acid solvate38
was solved by single-crystal X-ray diffraction from block
crystals obtained from one of the automated crystallizations.
However, the 1-methylnapthalene solvate grew as fine needles
and was only suitable for structure solution by simulated
annealing from laboratory capillary XRPD data,39 collected at
room temperature. Subsequent Rietveld refinement, using data
collected to 1.51 Å resolution, yielded an Rwp value of 0.057.
The hydrogen bonding motifs in the two solvate structures
and polymorphic forms 1 and 2 are shown in Figure 5. The
chain in form 2 is visually similar to that in form 1, but the
glide symmetry that propagates the form 1 chain is lost in the
form 2 chain with two symmetry independent molecules in the
unit cell. The chains lie side by side forming layers in both
forms, but form 1 has an ABAB layer repeat, whereas form 2
has an ABCD repeat, with CD related to AB by a shift of
approximately half a unit cell (Supporting Information Figure
S8). The chain of molecules of I in the 1-methylnapthalene
solvate is very similar to that found in both polymorphs, with
six chains surrounding channels of 1-methylnaphthalene mol-
ecules. The acetic acid solvate crystal structure has the solvent
incorporated into the catemer motif with the carboxylic acid
group mimicking the role of the CO-NH group of I in the chain
structure.
Discussion of the Experimental Polymorph Screen. The
automated crystallization search found a metastable polymorph
(form 2) and two solvates; however all three crystal structures
are based on the hydrogen bonded catemer motif, and no
evidence of R22(8) dimer-based structures was found. The
search was designed to test the effect of a wide range of solvent
properties, supersaturation, temperature, and agitation on the
crystallization of I while eliminating seeding by the raw material
(form 1). Rapid crystallization was obtained by controlled
cooling (Table 1) and may favor the formation of metastable
forms (i.e., kinetic conditions), while the variation in solvent
and concentration may influence the mode of association of
molecules of I in solution prior to nucleation. However, if a
(35) Hulme, A. T.; Fernandes, P.; Florence, A.; Johnston, A.; Shankland, K.
Acta Crystallogr., Sect. E 2006, 62, o3046-o3048.
(36) Mondal, R.; Howard, J. A. K. CrystEngComm 2005, 7, 462-464.
(37) Anderson, K. M.; Afarinkia, K.; Yu, H. W.; Goeta, A. E.; Steed, J. W.
Cryst. Growth Des. 2006, 6, 2109-2113.
(38) Hulme, A. T.; Johnston, A.; Florence, A. J.; Tocher, D. A. Acta Crystallogr.,
Sect. E 2006, 62, O545-O547.
(39) Hulme, A. T.; Fernandes, P.; Florence, A.; Johnston, A.; Shankland, K.
Acta Crystallogr., Sect. E 2006, 62, o3752-o3754.
9
3652 J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007

Search for a Predicted Hydrogen Bonding Motif
A R T I C L E S
Figure 4. Final observed (O), calculated (s), and difference (yobs - ycalc) profiles for the Pawley-type fit to capillary XRPD data collected from a sample
of the plastic crystalline phase (form 3) in the range 9°-60° 2θ. Inset, zoomed plot showing the fit to the data in the range 20°-60° 2θ. The pattern is
consistent with that expected from a disordered cubic plastic phase with relatively few observed peaks across the full data range and considerable background
scattering. The observed misfit in the final refinement is attributable to small inaccuracies in modeling both the background and the peak shape for reflections
belonging to the cubic cell.
solvents were likely to promote the formation of R22(8) dimers
highly metastable dimer-based structure was formed as the first
emergent phase under any of the conditions tested, it either did
not survive past critical-nucleus size40 or transformed via either
solution-mediated or solid-solid transformation prior to XRPD
detection. Variable temperature XRPD studies plus comple-
mentary sublimation and high-pressure crystallization from
solvent41 were also unsuccessful in producing any evidence of
additional polymorphs. From this extensive experimental search,
it appears most unlikely that an R22(8) dimer-based polymorph
of I can be formed by the growth of crystals from the gas phase
or solution.
Infrared Solid State and Solution Spectra. Infrared spectra
of I in various solutions and the solid state were carried out to
attempt to exploit the emerging link between structural synthons
present in the crystal structure and the growth synthons present
in solution from self-association equilibria.42 An indication of
whether a dimer-based structure of I could ever be found would
be the observation of self-assembled dimer units in solution,
most likely in nonpolar solutions. The R22(8) carboxylic acid
dimer motif has been shown by infrared spectroscopy to be
present in chloroform solutions of tetrolic and benzoic acid,
which crystallize to give structures containing the dimer motif,
and absent in the ethanolic solutions of tetrolic acid which give
rise to the catemeric â-polymorph of tetrolic acid.43 Solutions
of I in pentane and chloroform were studied because these
in the solution. For comparison, spectra of solutions in methanol
and water were also recorded, as these solvents could be
expected to solvate I more strongly and disrupt any self-
assemblies. Given the solid state spectrum of I, it was expected
that a strongly hydrogen bonded carbonyl bond would absorb
at around 1680-90 cm-1. However, the infrared spectra of I in
pentane, methanol, and water showed no significant differences
from one another with the carbonyl frequency recorded at
approximately 1740 cm-1 in all solutions indicating that I is
monomeric in all cases. In chloroform, a shift of the carbonyl
frequency to 1706 cm-1 is only suggestive of solvation rather
than dimerization. Hence, it seems that I does not associate in
solution and the R22(8) motif remains unobserved. The infrared
spectra of succinimide, a molecule that forms the R22(8) dimer
in the crystalline state,44 were measured for the solid state and
the same four solvents, and it also appears to be monomeric in
solution. Since succinimide crystallizes as a dimer, it can only
be concluded that the lack of dimers of I in solution does not
preclude the formation of a dimer in developing crystal nuclei.
The solution and solid state infrared spectra are included in the
Supporting Information.
Cluster Modeling and Crystal Structure Prediction.
Further evidence of the weakly interacting nature of molecules
of I in R22(8) dimers comes from energy minimization of
clusters of three or four molecules using the distributed multipole
intermolecular potential. Figure 6 shows one particularly sug-
(40) Davey, R. J.; Allen, K.; Blagden, N.; Cross, W. I.; Lieberman, H. F.; Quayle,
M. J.; Righini, S.; Seton, L.; Tiddy, G. J. T. CrystEngComm 2002, 4, 257-
264.
(41) Wood, P. A.; Parsons, S.; Pidcock, E. 2005. Personal Communication.
(42) Hunter, C. A. Angew. Chem., Int. Ed. 2004, 43, 5310-5324.
(43) Davey, R. J.; Dent, G.; Mughal, R. K.; Parveen, S. Cryst. Growth Des.
2006, 6, 1788-1796.
(44) Mason, R. Acta Crystallogr. 1961, 14, 720-724.
(45) Schaftenaar, G.; Noordik, J. H. J. Comput. -Aided Mol. Des. 2000, 14,
123-134.
9
J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007 3653

A R T I C L E S
Hulme et al.
Figure 6. An optimization of the intermolecular energy of three molecules
of I. The inset molecular diagrams show the conformation at the indicated
steps, starting from a dimer and a monomer, formation of a hydrogen bond
between the monomer proton and a free oxygen in the dimer, and further
rearrangement of the trimer to reach a minimum on the intermolecular
potential energy surface. Dotted red lines indicate hydrogen bonds defined
by 1.5 < H‚‚‚O < 3.15 Å and N-H‚‚‚O > 145° as given by MOLDEN.45
clusters with chain hydrogen bonds of lower energy than those
with the R22(8) hydrogen bonding motif.
Both of the intermolecular potentials predict approximately
equienergetic dimer- and catemer-based crystal structures, with
a dimer-based global minimum (Table 2). Forms 1 and 2 are
only about 0.5 kJ mol-1 above the global minimum, although
this increases to 2 kJ mol-1 for the alternative molecular charge
density model,28 which has a larger electrostatic contribution
to the lattice energy of approximately 10 kJ mol-1. A compari-
son with a selection of published computational models for the
lattice energy of predicted structures of I (Table 2) shows that
form 1 is generally only slightly less stable than hypothetical
dimer structures. Although not one of these intermolecular
potential models explicitly includes the induction energy, the
success of this type of model in crystal structure predictions of
related molecules46 and some preliminary estimates of the
induction contribution47 show that the distortion of the charge
density within the crystal lattice is unlikely to make the dimer-
based structures thermodynamically unfeasible. The calculated
lattice energies refer to static crystal structures at 0 K; however
the harmonic rigid-body crystal lattice modes and elastic
constants of the dimer- and chain-based low-energy structures
do not give a significant difference in the estimated free energies
at room temperature (Table 3). Thus given the uncertainties of
Figure 5. Hydrogen-bonded chains of I in form 1 (blue) overlaid with
those in (a) form 2; (b) the methylnaphthalene solvate and (c) the acetic
acid solvate.
gestive trajectory that provides some understanding of the
crystallization pathway of I: a third molecule readily hydrogen
bonds to the unused acceptor of a previously optimized R22(8)
dimer (Figure 6 step 1 to 2). The energy is lowered by
rearrangement of the molecules (to step 3), and finally a small
energy lowering is achieved by tilting the hydrocarbon moieties
into better van der Waals contact (step 3 to 4), although this
involves breaking a hydrogen bond, resulting in a hydrogen-
bonding chain motif for the trimer (4). The optimized trimer
has gained 12.1 kJ mol-1 in dispersion energy and lost 6.5 kJ
mol-1 in electrostatic energy relative to the intermediate trimer
cluster with the R22(8) motif (3), which is only 2.0 kJ mol-1
less stable.
The tendency for the approach of other molecules to disrupt
or severely distort the hydrogen bonding of the R22(8) dimer
occurred in 5 of the 20 optimizations attempted, including one
where two R22(8) dimers were allowed to approach one an-
other. When the alternative potential energy surface with a
stronger electrostatic interaction was used,28 the occurrence of
trajectories that led to the breakage of the dimer hydrogen
bonding was reduced, though not precluded. This is consistent
with there being a subtle balance between the electrostatic and
dispersion forces between molecules of I that produces some
(46) Day, G. M.; Motherwell, W. D. S.; Jones, W. Cryst. Growth Des. 2005, 5,
1023-1033.
(47) Welch, G. W. A. 2006. Personal Communication.
(48) Williams, D. E. J. Mol. Struct. 1999, 486, 321-347.
(49) Williams, D. E. J. Comput. Chem. 2001, 22, 1154-1166.
(50) Mayo, S. L.; Olafson, B. D.; Goddard, W. A. J. Phys. Chem. 1990, 94,
8897-8909.
9
3654 J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007

Search for a Predicted Hydrogen Bonding Motif
A R T I C L E S
Table 2. Comparison of the Relative Stability of the Known and Hypothetical Crystal Structures of I in a Range of Crystal Structure
Predictions by Lattice Energy Minimizationa
rank,
∆
E
1
kJ mol-
repulsion
−
dispersion
electrostatic
model
global minimum
motif
model
reference
form 1
form 2
FIT16
multipoles
(MP2, 6-31G**)
multipoles
this work
dimer
dimer
5, 0.6
4, 0.5
FIT16
this work
5, 1.9
9, 2.1
(PBE0, Sadlej)
multipoles
(DFT B3P91)
charges
(DFT B3P91)
charges
multipoles
Will9948,49
Will9948,49
46
46
chain ) dimer
(∼ak98)
3, 0.15
7, 3.37
dimer (∼ak98)
CVFF
Leusen3
Mooij3
dimer (∼ak98)
3, 0.5
2, 0.2
Dreiding50
dimer (∼ak98)
a Only the two submissions in CSP2001 which found the correct chain structure are included, but seven other submissions with different energy models
had a dimer-based structure that approximates ak98 (Table 3) in their three submissions (see Table 9 in ref 3).3
Table 3. Properties of Crystal Structures of I Corresponding to the Lowest Lattice Energy Minimaa
lowest
shearc
(GPa)
free
lattice
energy
(kJ/mol)
space
group
density
(g/cm3)
growth
rateb
energyd
structure
Z
a (Å)
b (Å)
c (Å)
â (deg)
(kJ/mol)
cb33
aq15
ak98
Pbca
8
4
4
8
4
4
4
4
11.5221
7.6543
10.1991
8.0295
7.8925
6.0716
6.2792
6.1566
12.0423
12.2517
7.8806
10.3377
10.5992
11.5367
11.6268
11.4942
11.4634
8.344
90.0
90.0
76.8
95.9
94.3
76.9
80.3
80.5
1.2793
1.3002
1.3095
1.3095
1.3059
1.2567
1.2683
1.2803
376.26
309.47
216.94
2.9237
5.7371
5.746
0.4119
2.4932
1.6561
0.3696
5.2291
-106.146
-104.934
-104.968
-106.034
-105.041
-106.692
-105.656
-105.203
-88.91
-88.49
-88.46
-88.40
-88.35
-87.69
-87.66
-87.49
P212121
P21/c
P21/c
P21/a
P21/n
P21/n
P21/n
9.9281
18.8224
9.3397
11.8665
11.1493
11.3872
form 2e
form 1e
am63
am22
am10
232.19
237.67
228.72
253.93
a Structures based on the R22(8) dimers are given in italics, and experimental structures, in bold. b The relative volume growth rate of the crystals
calculated by assuming that the growth rate of every face is proportional to the attachment energy. This is not calculated for form 2 which has Z′ ) 2 and
so a different growth unit. c The lowest eigenvalue of the shear submatrix of the elastic stiffness tensor. d Estimated Helmholtz free energy at 298 K as the
sum of the thermal energy at 298 K, the lattice energy, and the zero-point vibrational energy, derived from second derivative properties for rigid body
motions. e The experimental cell parameters which lead to this lattice energy minimum are a, b, c (Å) ) 7.676, 10.594, 19.020; R, â, γ (deg) ) 90.00,
94.125, 90.00 (P21/c) for form 2 and a, b, c (Å) ) 7.677, 10.584, 9.311; R, â, γ (deg) ) 90.00, 94.968, 90.00 (P21/a) for form 1. The relationship between
the cells is given in the Supporting Information, Figure S8.
current computational models, it is plausible that form 1 is the
thermodynamically most stable form but dimer-based structures
are unlikely to be outside the energy range of possible
polymorphism.
crystal structures for I does not give any reason why the
catemer- and not the dimer-based structures are actually
observed.
Discussion
A prediction of the relative growth rate of the low-energy
crystals from the vapor does show some variation between the
structures (Table 3) but does not suggest that the surfaces of
the dimer-based crystals, such as ak98, have particularly low
attachment energies. Hence once the crystallites have formed,
their relative growth rates, according to the simple attachment
energy model, do not discriminate between chain- and dimer-
based structures. The mechanical stabilities of the possible
structures of I vary considerably (Table 3), with both chain-
and dimer-based structures showing a wide range of anisotropy
in their elastic constants. Form 2 is predicted to be particularly
readily distorted by shearing forces, which is consistent with
the ease of transformation of form 2 to form 1 by shifting of
the layers. The lowest eigenvalue of the shear submatrix of 0.4
GPa for form 2 of I is somewhat larger than that of 0.15 GPa
for a predicted structure of aspirin51 that was subsequently found
as a highly metastable polymorph.52 However, the predicted
dimer-based crystal structures are generally more mechanically
stable. Thus, consideration of the properties of low-energy
Form 3 is an example of a novel plastic crystalline phase.
Molecules with similar globular shapes,53 such as bicyclo[3.3.1]-
nonane,54 bicyclo[3.3.1]nonan-9-one,55 and 3-oxabicyclo[3.3.1]-
nonane-2,4-dione, also form plastic phases, though not one of
these molecules is capable of hydrogen bonding. Relatively few
compounds containing hydrogen bond donors and acceptors are
known to form plastic crystals with examples including cyclo-
hexanol, trimethylacetic acid, and pentaerythritol derivatives.33
In plastic crystals, it has long been accepted53 that the molecules
have the freedom to rotate while maintaining a coherent crystal
lattice, until this is finally disrupted by melting, but only the
most spherical molecules can approximate free rotation in the
plastic phase. In most cases, the plastic crystal is “considered
to be disordered: in different sites the molecules occupy one
of n possible orientations and perform librations in the potential
well accompanied by jumps from one equilibrium position to
another”.56 In form 3, the body centered cubic structure implies
a spherical molecule radius of 3.28 Å. The van der Waals surface
(51) Ouvrard, C.; Price, S. L. Cryst. Growth Des. 2004, 4, 1119-1127.
(52) Vishweshwar, P.; McMahon, J. A.; Oliveira, M.; Peterson, M. L.;
Zaworotko, M. J. J. Am. Chem. Soc. 2005, 127, 16802-16803.
(53) Timmermans, J. J. Phys. Chem. Solids 1961, 18, 1-8.
(54) Laszlo, I. Recl. TraV. Chim. Pays-Bas 1965, 84, 251.
(55) Mora, A. J.; Fitch, A. N. Z. Kristallogr. 1999, 214, 480-485.
9
J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007 3655

A R T I C L E S
Hulme et al.
of I ranges from 2.1 to 4.3 Å from the center of mass, and so
the molecules in form 3 must correspond to the jumping
molecular disorder model. This model57 could be used to
estimate the number of discrete orientations in the plastic and
ordered phases, ∆Strs ) R ln(nplastic/nordered), with form 3
calculated to have 120 ( 30 discrete orientations. This is
comparable with the estimated number of states for heptacy-
clotetradecane.58 While there is little evidence that the entropies
of transitions for molecular crystals correspond well to this
model59 and more sophisticated treatments are possible,58 the
large transition entropy is not compatible with a limited number
of hydrogen bonding orientations. Thus, the structure and
thermodynamics of formation of form 3 imply that the hydrogen
bonding in the crystalline form 1 is significantly disrupted on
forming the plastic phase.
If a dimer-based structure of I were ever to be formed,
subsequent transformation to a more stable catemer-based
structure would require a significant change in the hydrogen
bonding. There is generally expected to be a significant energy
barrier for the rearrangement of hydrogen bonds within the solid
state, as many pairs of polymorphs which differ in their
hydrogen bonding occur (e.g., tetrolic acid,60 5-fluorouracil,61
famotidine,62-64 and cimetidine65). However, for I, the thermal
transition of form 1 to form 3 in which the hydrogen bonded
chain cannot remain intact, allied to the globular shape of the
molecule which allows its easy rotation, suggests that the barrier
to hydrogen bond rearrangement in I may be unusually low.
The cluster calculations provide further evidence to support
this hypothesis. The R22(8) dimer has an interaction energy of
-36.3 kJ mol-1, which is rather smaller than would be
traditionally associated with two independent NsH‚‚‚OdC
hydrogen bonds.66 This can be attributed to the proximity of
the two carbonyls to the NsH group in the imide group reducing
the electrostatic potential in the region of the donor (Figure 7)
so that it is comparable with the potential in some of the purely
hydrocarbon regions. This interplay of adjacent NsH and Cd
O groups has been shown to reduce the hydrogen bond strength67
for barbituric acid, cyanuric acid, parabanic acid, and alloxan,
which, when combined with steric considerations, leads to
unused hydrogen bond acceptors in the crystal structures. Figure
7 also shows that I is also approximately spherical, and so, with
the weak positive electrostatic potential around the imide proton,
neither the steric nor electrostatic forces would produce a
significant barrier to the rearrangement of the hydrogen bonds
of I. Thus, loose prenucleation clusters of I are expected to have
considerable freedom to rearrange to give the most stable crystal
Figure 7. Electrostatic potential on the water-accessible surface of I. The
color scale is from -60 to +60 kJ mol-1; the minimum potential on this
surface is -56.27 kJ mol-1, and the maximum is +38.62 kJ mol-1. The
molecule is viewed (a) with the imide proton pointing directly out of the
plane of the page; (b) from the side of the cyclohexane ring with the imide
group vertical and slightly behind; and (c) perpendicular to the plane of
symmetry.
structure, almost certainly the catemer-based form 1, with the
metastable form 2 a trapped intermediate showing the later
stages of crystal organization36 with chains forming early on in
the nucleation process.
While formation of a chain trimer shown in Figure 6 only
approximates a possible path for molecular association from a
gas at 0 K, the kinetic energy and possible solvent involvement
in molecular association in real crystallization processes are
likely to readily facilitate rearrangement of any R22(8) dimers
into the more stable chains. The lack of FTIR spectroscopic
evidence for dimers in even pentane solutions is consistent with
the hydrogen bonding being so weak that R22(8) dimer forma-
tion of I is a rare event under any crystallization condition.
(56) Fouret, R. Diffuse X-Ray Scattering by Orientationally-Disordered Crystals.
In Plastically Crystalline State: Orientationally Disordered Crystals;
Sherwood, J. N., Ed.; John Wiley & Sons: Chichester, 1979; pp 85-121.
(57) Guthrie, G. B.; Mccullough, J. P. J. Phys. Chem. Solids 1961, 18, 53-61.
(58) Kabo, G. J.; Blokhin, A. V.; Charapennikau, M. B.; Kabo, A. G.; Sevruk,
V. M. Thermochim. Acta 2000, 345, 125-133.
(59) Clark, T.; Mckervey, M. A.; Mackle, H.; Rooney, J. J. J. Chem. Soc.,
Faraday Trans. 1 1974, 70, 1279-1291.
(60) Leiserowitz, L.; Nader, F. Angew. Chem., Int. Ed. 1972, 11, 514-&.
(61) Hulme, A. T.; Price, S. L.; Tocher, D. A. J. Am. Chem. Soc. 2005, 127,
1116-1117.
(62) Ferenczy, G. G.; Parkanyi, L.; Angyan, J. G.; Kalman, A.; Hegedus, B.
THEOCHEM 2000, 503, 73-79.
(63) Golic, L.; Djinovic, K.; Florjanic, M. Acta Crystallogr., Sect. C 1989, 45,
1381-1384.
(64) Shankland, K.; McBride, L.; David, W. I. F.; Shankland, N.; Steele, G. J.
Appl. Crystallogr. 2002, 35, 443-454.
Conclusion
(65) Hegedus, B.; Gorog, S. J. Pharm. Biomed. Anal. 1985, 3, 303-313.
(66) Desiraju, G. R. Acc. Chem. Res. 2002, 35, 565-573.
(67) Lewis, T. C.; Tocher, D. A.; Price, S. L. Cryst. Growth Des. 2005, 5, 983-
993.
The extensive experimental screen undertaken has covered
solvent and thermal crystallization methods without finding any
9
3656 J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007

Search for a Predicted Hydrogen Bonding Motif
A R T I C L E S
evidence of a dimer-based polymorph but cannot exclude further
polymorphs being found by other crystallization methods that
have recently been shown to generate new forms, such as
polymer templating,68 crystallization additives,69,70 or in the
course of attempted cocrystallization experiments.71 This screen
produced two solvates, the known unsolvated form, a metastable
polymorph, and a plastic crystalline phase. While the plastic
phase is stable above 145 °C, the catemer-based form 1 is the
most stable known polymorph at ambient temperatures, and is
almost certainly the most stable possible structure down to
0 K.
This study clearly demonstrates the synergy between experi-
mental polymorph screening and the computational prediction
of the energetically feasible crystal structures and reinforces the
recognition40 that a more complete understanding of the kinetic,
thermodynamic, and structural aspects of nucleation is required
for further development of computational polymorph prediction
methodologies.
Acknowledgment. G. M. Day, R. J. Davey, and G. Dent are
thanked for valuable discussions, and G. M. Day is thanked for
providing the results of his computational study. The authors
acknowledge the Basic Technology Program of the Research
Councils UK for supporting “Control and Prediction of the
Organic Solid State” (www.cposs.org.uk Grant Number GR/
S24114/01 to S.L.P.).
The relative energies of potential crystal structures of I implies
that such a dimer-based polymorph would be within the possible
energy range of potential polymorphs, though no evidence of
such a form has been found. This can be rationalized in terms
of the unusual balance between the hydrogen bonding and van
der Waals interactions of this nearly spherical molecule,
allowing the facile rearrangement of nucleating clusters contain-
ing the dimer into the more stable hydrogen bonded chain-based
form. The evidence that there is at most a low kinetic barrier to
the rearrangement of clusters of dimers into a catemer-based
motif reduces the likelihood that any crystallization method
could achieve a new dimer-based polymorph of I.
Supporting Information Available: Computed crystal struc-
tures are stored on CCLRC e-Science Centre dataportal and are
available from the authors on request. The following Supporting
Information is available: S1 Details of synthesis; S2 Spectro-
scopic investigation of solution aggregation; S3 Crystal Structure
Prediction results; S4 Principal Component Analysis of solvent
molecules; S5 Crystallization details, results, and XRPD
analysis; S6 Overlay diagram contrasting forms 1 and 2; S7
Ab initio optimized geometry and energy for I; S8 Complete
refs 3, 4, and 14. This material is available free of charge via
the Internet at http://pubs.acs.org.
(68) Price, C. P.; Grzesiak, A. L.; Matzger, A. J. J. Am. Chem. Soc. 2005, 127,
5512-5517.
(69) Weissbuch, I.; Addadi, L.; Lahav, M.; Leiserowitz, L. Science 1991, 253,
637-645.
(70) Torbeev, V. Y.; Shavit, E.; Weissbuch, I.; Leiserowitz, L.; Lahav, M. Cryst.
Growth Des. 2005, 5, 2190-2196.
(71) Day, G. M.; Trask, A. V.; Motherwell, W. D. S.; Jones, W. Chem. Commun.
2006, 54-56.
JA0687466
9
J. AM. CHEM. SOC. VOL. 129, NO. 12, 2007 3657