Combination of CP/MAS NMR spectroscopy and X-ray crystallography: Structure and dynamics in molecular crystals of hydrogen, lithium, sodium, rubidium, and cesium penicillin V

Article abstract of DOI:10.1021/ja9640657

¹³C CP/MAS NMR spectra of the free acid of phenoxymethyl penicillin (penicillin V) and of its lithium, sodium, rubidium, and cesium salts (C₁₆H₁₇N₂O₅SM, M = H, Li·H₂O, Na, Rb, and Cs) have

Full text of DOI:10.1021/ja9640657

J. Am. Chem. Soc. 1997, 119, 9793-9803
9793
Combination of CP/MAS NMR Spectroscopy and X-ray
Crystallography: Structure and Dynamics in Molecular Crystals
of Hydrogen, Lithium, Sodium, Rubidium, and Cesium
Penicillin V
Michaela Wendeler,† Jamila Fattah,† J. Mark Twyman,† Alison J. Edwards,‡,§,#
Christopher M. Dobson,*,†,§ Stephen J. Heyes,*,† and Keith Prout*,‡,§
Contribution from the Inorganic Chemistry Laboratory, South Parks Road, Oxford OX1 3QR,
U.K., the Chemical Crystallography Laboratory, 9 Parks Road, Oxford OX1 3PD, U.K., and the
Oxford Centre for Molecular Sciences, South Parks Road, Oxford, OX1 3QT U.K.
ReceiVed NoVember 25, 1996. ReVised Manuscript ReceiVed June 4, 1997X
Abstract: 13C CP/MAS NMR spectra of the free acid of phenoxymethyl penicillin (penicillin V) and of its lithium,
sodium, rubidium, and cesium salts (C16H17N2O5SM, M ) H, Li‚H2O, Na, Rb, and Cs) have been studied at a range
of temperatures between 180 and 400 K. The 13C CP/MAS NMR spectra at ambient temperature indicate that the
structures have one, two, one, four, and one molecule(s), respectively, in the crystallographic asymmetric unit. This
is confirmed by the known crystal structure of the free acid and the single-crystal X-ray diffraction studies reported
here of the various salts. The variable-temperature 13C CP/MAS NMR spectra indicate that the phenyl rings of all
the molecules perform 180° flips about their local C2 axes, with measured or extrapolated rate constants in the range
5 × 10-12 to 8 × 109 s-1 in the temperature regime 200-380 K, and spanning 11 orders of magnitude at ambient
temperature. The atomic displacement parameters of the X-ray structural model suggest that these rings also undergo
significant in-plane librations. The Arrhenius activation parameters for the ring flip motions have been determined,
by analysis of exchange and dipolar broadened NMR line shapes and with 1-D (and 2-D) magnetization transfer
experiments, for the free acid and all four salts. The activation barriers have been interpreted in terms of the degree
of entrapment of the phenyl groups and the nature of the packing interactions in the different crystal structures.
Introduction
solution of single-crystal X-ray diffraction data cannot be
determined directly, even for small, highly crystalline molecular
systems, from typical 13C CP/MAS NMR spectra of micro-
crystalline samples. Instead, 13C CP/MAS NMR is well suited
to providing qualitative and semiquantitative information about
local structure and only by inference from the local structural
details can the nature of the extended structure be proposed.4
The adps from X-ray diffraction5 contain information about
translational, vibrational, and rotational (librational) oscillations
of the molecule or groups within the molecule, and also on the
nature of any positional disorder of atoms away from their mean
positions (which may be static or dynamic in nature) and, on
occasion, on the goodness of fit of the model to the diffraction
data. The time scale of the diffraction technique (∼10-18 s) is
too short for direct detection of atomic motions that are slow
on the time scale of bond vibrations. These slow dynamic
processes will only be detected if they result in new atomic
positions and therefore cannot be distinguished from “static”
molecular disorder. When such disorder is detected the X-ray
structure solution contains no direct information about the
relative rates of any dynamic processes. In contrast to diffrac-
tion techniques, the utility of NMR in the elucidation of
molecular dynamics arises because various NMR spectroscopic
interactions are directly sensitive to motions over a wide range
of time scales (∼102-10-10 s) that are slow compared with that
of bond vibrations; no other technique possesses such a wide
range of rates over which motion may be examined explicitly.6
Motions with time scales in the range 101-10-7 s have been
A particularly powerful approach to the detailed characteriza-
tion of local and extended structure and molecular motion in
the solid state has proved to be a combination of solid-state
NMR and diffraction methods. The combined use of the two
techniques in the characterization of several molecular systems
has, for example, facilitated a full analysis of structure and
dynamics in situations where either technique in isolation would
have failed to yield even a small fraction of the information
provided by the combination of the two techniques(See ref 1
and references therein).
The information provided by X-ray diffraction and by NMR
techniques is fundamentally different and highly complementary.
Structure solution from single-crystal X-ray diffraction data2
yields information on intramolecular structure including bond
distances and conformational angles, and on intermolecular
contacts and packing relationships. Similar information is
encapsulated in NMR shielding parameters,3 but the detailed
information in the form typically obtained from structure
* To whom correspondence should be addressed.
† Inorganic Chemistry Laboratory.
‡ Chemical Crystallography Laboratory.
§ Oxford Centre for Molecular Sciences.
# School of Chemistry, University of Melbourne, Parkville, Vic. 3052,
Australia.
X Abstract published in AdVance ACS Abstracts, September 15, 1997.
(1) (a)Fattah, J.; Twyman, J. M.; Heyes, S. J.; Watkin, D. J.; Edwards,
A. J.; Prout, K.; Dobson, C. M. J. Am. Chem. Soc. 1993, 115, 5636-5650
and references therein. (b) Brouwer, E. B.; Enright, G. D.; Ripmeester, J.
A. Supramol. Chem. 1996, 7, 7-9.
(4) Clayden, N. J.; Dobson, C. M.; Lian, L.-Y.; Twyman, J. M. J. Chem.
Soc., Perkin Trans. II 1986, 1933-1940.
(2) Glusker, J. P.; Lewis, M.; Rossi, M. Crystal Structure Analysis for
Chemists and Biologists, VCH: Weinheim, 1994.
(3) Harris, R. K. Nuclear Magnetic Resonance Spectroscopy, rev. ed.;
Longman: London, 1986.
(5) Dunitz, J. D.; Maverick, E. F.; Trueblood, K. N. Angew. Chem., Int.
Ed. Engl. 1988, 27, 880-895.
(6) Muetterties, E. L. Inorg. Chem. 1965, 4, 769-771.
S0002-7863(96)04065-6 CCC: $14.00 © 1997 American Chemical Society

9794 J. Am. Chem. Soc., Vol. 119, No. 41, 1997
Wendeler et al.
studied in the present case through the direct sensitivity of 13
C
NapenV. A solution of HpenV (2.00g, 5.71 mmol) in THF (8 mL)
and distilled H2O (3 mL) was adjusted to pH ∼5 with 2.5% NaOH,
and the excess THF was removed under reduced pressure. The sample
was lyophilized to give NapenV, 1.83 g, 86%. NapenV (500 mg) was
dissolved in distilled water (0.8 mL), and 2-propanol (0.15 mL) added
dropwise. The flask was left to stand in an acetone bath, and crystal
growth in the form of fine needles was observed after 1 h. The crystals
(440 mg) were collected by filtration after three days and dried in air.
Elemental analysis: Found: C 51.50, H 4.64, N 7.46, S 8.52, Na 5.91
C16H17N2NaO5S, requires C 51.61, H 4.60, N 7.52, S 8.61, Na 6.17.
Microanalyses were performed by the analytical service of the
Inorganic Chemistry Laboratory. C, H, and N contents were determined
by combustion analysis and metal analyses by atomic absorption
spectroscopy.
For each compound selected crystals were used for the single-crystal
X-ray diffraction analysis. The same batch of the compound, ground
into a finely divided powder, was used for the CP/MAS NMR
experiments. The X-ray powder diffractograms obtained from the
samples used for the CP/MAS NMR experiments were compared with
those computed (LAZY PULVERIX8) from the cell dimensions and
atomic parameters observed by single-crystal diffraction work and
confirmed that the X-ray diffraction and NMR experiments were carried
out on the same polymorphic form.
ii. Single-Crystal X-ray Diffraction. Relevant details of the
crystallography for LipenV, NapenV, and CspenV are given in Table
1 and more extensively in the Supporting Information Table S1. Data
reduction included Lorentz and polarization corrections. All three
structures were solved by direct methods9 which yielded all non-
hydrogen atom positions for each of the three salts. Positional and
first isotropic then atomic displacement parameters for the non-hydrogen
atoms were refined to convergence by the least-squares method,
minimizing ∑w(Fobs - Fcalc)2 for all observed reflections. The full
normal matrix was used together with a Chebyshev polynomial
weighting scheme,10 the parameters for which are given in Table S1.
For all three structures an empirical absorption correction (DIFABS11)
was applied prior to the final refinement cycles. Hydrogen atoms were
included in the models at geometrically idealized positions. Final
residuals are given in Table 1. For the lithium salt, which is a hydrate,
the lithium ion and the water molecule could not be distinguished in
the interpretation of the electron density maps until after the beginning
of the anisotropic refinement.
CP/MAS NMR experiments to magnetization transfer, exchange-
broadening, and dipolar broadening time scales. NMR is
particularly useful in its sensitivity to these relatively slow
dynamic processes. We applied this approach to a study of the
potassium penicillin V (KpenV) system. The combination of
13C CP/MAS NMR with X-ray crystallography was fundamental
to the elucidation of the complexities of the low-symmetry
structure and its dynamically induced phase transition.1a In this
work we extend the study of the penV system to the crystalline
Li-, Na-, Rb-, and CspenV and the free acid (HpenV). The
penV series was chosen for study because particular questions
are raised from examination of the information already available.
The structure of KpenV is unusual with four molecules in the
asymmetric unit and, although related to the previously known
HpenV structure,7 the differences appear to be due to the
different coordination demands of the cation; the present study
considers systematically the effect that the size of the cation
has on the crystal structure. A feature of previous 13C CP/MAS
NMR studies of a variety of penicillins is the detection of
dynamic averaging effects, notably those associated with 180°
flips of the phenyl rings in side chains about their C2 axes. The
present study enables the frequency of these ring flips to be
characterized in a closely related set of packing environments.
The methodical comparison offers an opportunity to examine
the balance of effects on the rate of phenyl ring flips, not just
the influence of direct interactions with local environment but
also of the flexibility in that local environment conferred by
features of the long-range ordering motif. From this study
insight will be gained into the general aspects of spatially
demanding, slow motions in molecular solids.
Experimental Section
i. Preparation and Characterization of H+, Li+, Na+, Rb+, and
Cs+ Salts of 6-(Phenoxyacetamido)penicillanic Acid (PenV). The
lithium (LipenV), sodium (NapenV), rubidium (RbpenV), and cesium
(CspenV) salts of 6-(phenoxyacetamido)penicillanic acid (HpenV) were
prepared by ion exchange from the free acid and recrystallized by vapor
diffusion.
HpenV. HpenV (Sigma Chemical Co.) was recrystallized by the
vapor diffusion of acetone into an aqueous alcohol solution. Crystal
formation was observed after 5 h, and the crystals were collected by
filtration after 2 days and dried in air.
Li-, Rb-, and CspenV. Solutions of HpenV (3.50 g, 10 mmol) in
THF (12 mL) and distilled H2O (1 mL) were adjusted to pH ∼5 with
the dropwise addition of 4% LiOH/5% RbOH/3% CsOH. The samples
were lyophilized to give LipenV, 2.93 g, 78.3%/RbpenV 3.34 g, 76.8%/
CspenV, 4.259 g, 88.3%. For recrystallization, MpenV (300 mg) was
dissolved in distilled water (Li, 0.6 mL; Rb, 0.3 mL; Cs 0.25 mL).
Ethanol (Li, 0.6 mL; Rb, 0.3 mL; Cs, 4 mL) and 2-propanol (Li, 2
mL; Rb, 0.2 mL; Cs, 0.8 mL) were added dropwise. Flasks were left
to stand in ethanol (Li and Cs) or acetone (Rb) baths, and crystal growth
in the form of fine needles was observed after ∼2 d. The material
was then subjected to two further similar recrystallizations, and the
final crystals were collected by filtration and air-dried. In contrast with
other penV salts the samples of LipenV contain water of crystallization;
all procedures used to prepare and recrystallize LipenV resulted in one
crystalline form, the monohydrate. The water is lost progressively on
heating above 340 K, but the crystals rehydrate on standing in the
atmosphere, (Figure S1). (Figures S1-S11 are in the Supporting
Information). Elemental analyses: (LipenV) Found C 51.54, H 5.44,
N 7.49, S 8.48, Li 1.90 C16H17LiN2O5S‚H2O; requires C 51.34, H 5.12,
N 7.48, S 8.57, Li 1.85. (RbpenV) Found C 44.8, H 3.95, N 6.53, Rb
17.82 C16H17N2O5RbS; requires C 44.19, H 3.94, N 6.44, Rb 19.65.
(CspenV) Found C 39.82, H 3.48, N 5.79, Cs 25.17 C16H17CsN2O5S;
requires C 39.84, H 3.55, N 5.81, Cs 27.56.
Except where indicated to the contrary, the Oxford CRYSTALS12
system was used for all crystallographic calculations and CAMERON13
for crystallographic drawings. Scattering factors were taken from
Cromer and Weber.14 The atomic coordinates, atomic displacement
parameters, bond distances and angles, and interionic contact and
H-bond distances, Tables 2-4, are available as Supporting Information.
An X-ray structure analysis of RbpenV is in progress.15
iii. Solid-State NMR Spectroscopy. 13C CP/MAS NMR spectra
were acquired on a Bruker MSL 200 spectrometer equipped with an
Oxford Instruments 4.7 T wide-bore (98 mm) superconducting solenoid
(8) Yvon, K.; Jeitschko, W.; Parthe, E. LAZY PULVERIX. J. Appl.
Crystallogr. 1977, 10, 73-74.
(9) Sheldrick, G. M. in Crystallographic Computing 3; Sheldrick, G.
M., Kruger, C., Goddard, R., Eds.; Oxford University Press: Oxford, 1985;
pp 175-189.
(10) Carruthers, J. R.; Watkin, D. J. Acta Crystallogr. 1979, A35, 698-
699.
(11) Walker, N.; Stuart, D. Acta Crystallogr. 1983, A39, 158-166.
(12) Watkin, D. J.; Carruthers, J. R.; Betteridge, P. W. CRYSTALS User
Guide, Chemical Crystallography Laboratory, Oxford. 1985.
(13) Pearce, L. J.; Prout, K.; Watkin, D. J. CAMERON User Guide,
Chemical Crystallography Laboratory, Oxford, 1994.
(14) Cromer, D. T.; Weber, J. T. International Tables for X-ray
Crystallography, Ibers, J. A., Hamilton, W. C., Eds.; Kynoch Press:
Birmingham, 1974; Vol. 4.
(15) The diffraction patterns of crystals of RbpenV at 100 K (a )
9.460(2), b ) 12.485(3), c ) 15.105(3) Å, R ) 92.88(2), â ) 99.16(2), γ
) 90.40(2)°, V ) 1758.8 Å3), 293.4 K (a ) 9.505(2), b ) 12.622(3), c )
15.363(3) Å, R ) 93.81(2), â ) 98.85(3), γ ) 90.19(2)°, V ) 1817.0 Å3),
and 370 K (a ) 9.530(4), b ) 6.335(1), c ) 15.488(3) Å, R ) 94.18(2),
â ) 98.86(2), γ ) 90.16(2)°, V ) 921.3 Å3) have been recorded. A detailed
examination of these diffraction patterns shows that RbpenV is essentially
isomorphous and isostructural with KpenV both above and below the phase
change at ∼365 K.
(7) Abrahamsson, S.; Hodgkin, D. C.; Maslen, E. N. Biochem. J. 1963,
86, 514-535.

Studies of Molecular Crystals of Penicillin V
J. Am. Chem. Soc., Vol. 119, No. 41, 1997 9795
Table 1. Crystal Data and Refinement Parameters
crystal and experimental data
LipenV
NapenV
CspenV
formula
Mr
crystal system
temperature, K
a, Å
b, Å
c, Å
C16H17LiN2O5S‚H2O
374.34
orthorhombic
295(2)
7.425(1)
21.777(2)
22.677(2)
90
90
90
3667
8
C16H17N2NaO5S
372.37
monoclinic
293(2)
8.841(1)
6.129(7)
15.929(2)
90
97.43
90
856
2
C16H17CsN2O5S
450.23
orthorhombic
295(2)
6.5436(4)
9.6181(7)
30.011(2)
90
90
90
1888.8
4
R, deg
â, deg
γ deg
cell vol deg Å3
Z
Dc, g cm-3
space group
radiation
λ, Å
1.356
1.45
1.583
P212121 (no. 19)
Cu KR
1.5418
2011
18.29
P21 (no. 4)
Cu KR
1.5418
1746
21.6
0.038, 0.043
P212121 (no. 19)
Cu KR
1.5418
2330
158.7
no. observed (I g 3σI)
µ, cm-1
b
final R,a Rw
0.049, 0.055
0.036, 0.043
2
a R ) ∑(|Fo| - |Fc|)/∑|Fo|. b Rw ) {∑[w(|Fo| - |Fc|)2]/∑[w|Fo| ]}1/2
.
magnet operating at frequencies of 50.32 and 200.13 MHz for 13C and
1H, respectively. CP/MAS spectra were recorded using a double-
bearing magic-angle sample spinning probe (Bruker Z32-DR-MAS-
7DB), utilizing dry nitrogen for all gas requirements. Approximately
250 mg of sample was packed into 7 mm zirconia rotors with Kel-F
caps for MAS at typical rates of 1-4 kHz. A single contact spin-
lock cross polarization (CP) sequence16 was used with alternate cycle
spin-temperature inversion17, with flip-back of 1H magnetization18 and
with proton rf fields of ∼1.7 mT, resulting in a 90° pulse length of
3.5-4.5 µs (ν1H ≈ 55-75 kHz). Temperature measurement and
regulation, utilizing a Bruker B-VT1000 unit equipped with a copper-
constantan thermocouple and digital reference, was of the bearing gas.
Temperature calibration was achieved as described previously.1a
Spectra were generally recorded with 3-5 K temperature increments
and ∼30 min was allowed for equilibration at each new temperature
before commencement of spectral acquisition. Free induction decays
were defined by ∼2-4 K data points over a spectral width of 20 kHz
and were zero-filled to 16 K data points prior to Fourier transformation.
Typically 100-1000 transients, with a CP contact time of 0.75 or 2.5
ms and a relaxation delay of 3.5-8 s, were accumulated for each
spectrum. Chemical shifts are reported with respect to δ(TMS) ) 0
ppm and were referenced externally to the upfield resonance of solid
adamantane at 29.5 ppm.19 Nonquaternary Supression (NQS) dipolar
dephasing experiments20 were carried out using delays of 20-200 µs.
simulated using the program DNMR4 (QCPE no. 466).26 Dipolar
broadening27 was analyzed through the deconvoluted line width, ∆ν1/2′,
the difference between the measured and limiting line widths. A plot
of ln(∆ν1/2′) vs (1/T) was fit, through a least-squares, error-weighted
procedure, to the theory of Rothwell and Waugh.27 Identification of
the temperature of maximum broadening, as indicated by the best fit
equation, at which the modified rate of the two-site ring flip exchange
process is 4πν1H, enables the line widths to be converted to rate
constants, and these constants are included in the full Arrhenius
analyses. Rate data from magnetization-transfer experiments, exchange
broadening simulations, and dipolar broadening analysis are all quoted
in the form of modified rate constants, k () 2kAB for the equipopulated
two-site exchange processes studied in this work),25 and were used to
determine Arrhenius activation parameters from best fits for ln k vs
(1/T).
Results
a. X-ray Structure Analysis. The crystal structures of
LipenV, NapenV, and CspenV have been determined in this
work. Structures of KpenV at temperatures above and below
the phase change occurring between 356 and 366 K have been
reported by us in a previous paper.1a The crystal data15 for
RbpenV above and below the apparent phase change shown by
NMR indicate that both forms are isomorphous with the
corresponding forms of KpenV. The crystal structure of HpenV
was determined by Abrahamsson et al.7 The HpenV coordinates
were transformed to give the correct absolute configuration. The
chemical formula of penV is given in 1; the notation, which is
used in the discussion of the NMR results, is in accordance
with conventions used in penicillin chemistry.
Rotationally asynchronous, TPPI phase-sensitive CP/MAS 2D-
exchange NMR experiments with mixing times in the range 2-3 s
were performed as described previously.23 1D-magnetization-transfer
measurements used the rotationally synchronized, selective experiment
of Conner et al.24 and were analyzed following the procedure of Jeener
et al.25 Exchange-broadened 13C CP/MAS NMR lineshapes were
The crystal and molecular structures with atomic displacement
ellipsoids are shown for CspenV, NapenV, and LipenV in
Figures 1-3. Similar diagrams for KpenV are given in ref 1a.
The crystal structure of HpenV is redrawn in Figure 4 from the
transformed coordinates of Abrahamsson et al.7 The asym-
metric units of the crystal structures of HpenV, NapenV, and
CspenV each contain one anion and one cation, that of LipenV
two anions, two cations, and two water molecules per asym-
metric unit, and those of KpenV and RbpenV four anions and
four cations. All five structures are similar in the molecular
packing motif within the crystals, but the only isomorphism is
(16) Pines, A.; Gibby, M. G.; Waugh, J. S. J. Chem. Phys. 1973, 59,
569-590. Yannoni, C. S. Acc. Chem. Res. 1982, 15, 201-208.
(17) Stejskal, E. O.; Schaefer, J. J. Magn. Res. 1975, 18, 560-563.
(18) Tegenfeldt, J.; Haeberlen, U. J. Magn. Res. 1979, 36, 453-457.
(19) Earl, W. L.; VanderHart, D. L. J. Magn. Reson. 1982, 48, 35-54.
(20) Opella, S. J.; Frey, M. H. J. Am. Chem. Soc. 1979, 101, 5854-
5856.
(21) Harbison, G. S.; Smith, S. O.; Pardoen, J. A.; Courtin, J. M. L.;
Lugtenburg, J.; Herzfield, J.; Mathies, R. A.; Griffin, R. G. Biochemistry
1985, 24, 6955-6962.
(22) Torchia, D. J. Magn. Reson. 1978, 30, 613-616.
(23) Twyman, J. M.; Dobson, C. M. Magn. Reson. Chem. 1990, 28, 163-
170.
(24) Conner, C.; Naito, A.; Takegashi, K.; McDowell, C. A. Chem. Phys.
Lett., 1985, 113, 163-170. Takegashi, K.; McDowell, C. A. J. Am. Chem.
Soc. 1986, 108, 6852-6857.
(25) Jeener, J.; Meier, B. H.; Bachmann, P.; Ernst, R. R. J. Chem. Phys.
1979, 71, 4546-4553.
(26) Bushweiler, C. H.; Letendre, L. J.; Brunelle, J. A.; Bilotsky, H. S.;
Whalon, M. R.; Fleischmann, S. H. Quantum Chemistry Program Exchange
Program, Program No. 466, DNMR4.
(27) Rothwell, W. P.; Waugh, J. S. J. Chem. Phys. 1981, 74, 2721-
2732.

9796 J. Am. Chem. Soc., Vol. 119, No. 41, 1997
Wendeler et al.
Figure 2. The crystal structure of NapenV shown in projection down
a, together with a drawing of the molecule in the same orientation, in
which the atoms are represented by their 50% atomic displacement
ellipsoids.
Figure 1. The crystal structure of CspenV shown in projection down
a, together with a drawing of the molecule in the same orientation, in
which the atoms are represented by their 50% atomic displacement
ellipsoids.
R-CO2H group axial. The fused-ring â-lactam/thiazolidine
system appears very rigid so that the only significant differences
between the nine molecules lie in the orientation of the phenoxy
side chain with respect to the bicyclic ring system. These
differences are emphasized in Figure 5 and Table 2, which lists
for each molecule the five torsion angles which describe the
side chain conformations. Three of these torsion angles, C5-
C6-N14-C15, C6-N14-C15-C17, and N14-C15-C17-
O18, are very similar at 180 ( 15° for all nine conformers.
Thus the amide bond is trans, and the N-H proton must form
an intramolecular hydrogen bond with the ether oxygen atom.
Ignoring the H atoms, the chain of atoms from C5 to O18 is
coplanar with two trans linkages and a final cis linkage.
Relative to this planar extended chain of atoms the phenyl group
occupies a wide variety of positions, indicating that there is no
single conformation with a significantly lower intramolecular
energy than the rest. In the structure of LipenV the two
independent molecules in the asymmetric unit have very similar
side chain conformations, but in KpenV (RbpenV) there are
four very different orientations of the phenyl group.
ii. Molecular Packing within the Crystal Structures. The
crystals have layer structures composed of stacks of molecular
sandwiches. Each sandwich has a hydrophilic interior with a
layer of cations (and, in the case of LipenV, water molecules)
at the center coordinated by the penV oxygen atoms. The
hydrophobic phenyl groups of the side chains constitute the
surfaces of these sandwiches. In each crystal these sandwiches
are parallel to the ab plane. The c axis repeat is either one
sandwich (HpenV, NapenV, KpenV, and RbpenV) or two
sandwiches (LipenV and CspenV) related by a 2-fold screw
axis. The interlayer separations fall into two groups, depending
that of KpenV and RbpenV. CspenV is in fact isostructural
with potassium benzylpenicillin (KpenG)28,29 and NapenV is
isostructural with sodium benzylpenicillin (NapenG).28
i. The Molecular Structure of the PenV Anion. The
bonded distances and interbond angles in the penV anions in
the Li, Na, and Cs salts are presented in the Supporting
Information and in most respects show no significant differences
from those reported for KpenV1a and HpenV.7 In all nine
independent molecules of the five MpenV crystal structures the
thiazolidine ring has a C-3 conformation;4 that is with the
(28) Crowfoot, D.; Bunn, C. W.; Rogers-Low, B. W.; Turner-Jones, A.
In The Chemistry of Penicillin; Clarke, H. T., Johnson, J. R., Robinson, R.,
Eds.; Princetown University Press, New Jersey, 1949; pp 310-366.
(29) Prout, K.; Baird, P.; Crook, S. M.; Watkin, D. J.; Twyman, J. M.;
Heyes, S. J.; Dobson, C. M. Unpublished results.

Studies of Molecular Crystals of Penicillin V
J. Am. Chem. Soc., Vol. 119, No. 41, 1997 9797
Figure 3. The crystal structure of LipenV shown in projection down a, together with a drawing of the two molecules in the same orientation, in
which the atoms are represented by their 50% atomic displacement ellipsoids. Molecule 1, the molecule associated with Li1, is drawn in black in
the structure diagram, and molecule 2, associated with Li2, is drawn with open circles.
on whether the phenyl groups of adjacent sandwiches just touch
(CspenV, d002 ) 15.005 Å; RbpenV, d001 ) 15.145 Å; KpenV,
d001 ) 15.150 Å and NapenV, d001 ) 15.795 Å) or interleave
(LipenV, d002 ) 11.338 Å; HpenV, d001 ) 12.06 Å) (Figures
1-4).
The orientation of the fused-ring â-lactam/thiazolidine system
is determined by the number of oxygen atoms required to satisfy
the cation coordination. In the structure of CspenV the Cs+
ion is coordinated by seven oxygen atoms, one â-lactam, two
amide, and four carboxylate oxygens from six neighboring penV
anions, Figure 1, and is very similar to but not the same as the
coordination of K+ in KpenV.1a In NapenV the Na+ ion is
coordinated by five oxygen atoms (Figure 2). In LipenV the
penV anions interact with a mixed layer of Li+ ions and water
molecules (Figure 3). Each Li+ ion is coordinated by four
oxygen atoms, one amide oxygen, two carboxylate oxygens from
three penV anions, and an oxygen atom of a water molecule,
and each water molecule coordinates one Li+ ion. The water
molecule oxygen atom O(1) H-bonds to the carboxylate O(102)
and the â-lactam oxygen O(103), O(2) H-bonds only to
carboxylate oxygen atom O(201). The HpenV structure follows
the same general principles of a molecular sandwich with a
hydrophobic exterior and a hydrophilic middle. The basic unit
is, however, a chain with the carboxylate hydroxyl of one
molecule H-bonded to the side chain amide oxygen of the
molecule translated one unit cell along b. C-centering extends
the chains into sheets and the 2-fold axes the sheets into
sandwiches.

9798 J. Am. Chem. Soc., Vol. 119, No. 41, 1997
Wendeler et al.
345 K is given in Figures 7a and S3; other regions of the
spectrum are essentially temperature independent.34 In the
spectrum at 213 K separate C2′ and C6′ aromatic carbon
resonances are clearly resolved, but the C3′,C5′ resonances are
degenerate. As the temperature is increased above 213 K the
resonances of the inequivalent ortho carbons C2′,C6′ of the
phenyl ring display classic two-site exchange broadening, with
coalescence at 235 K and subsequent resharpening to give a
single well-defined resonance. For the resonances of the C3′,C5′
sites, whose frequency separation is too small to be observed,
only slight line-broadening can be seen to occur over this
temperature range. Despite initial sharpening of the averaged
C2′,C6′ resonance above 237 K, it and the C3′,C5′ resonance
both broaden dramatically above ∼244 K, reaching maximum
line width at 282 K before resharpening to their limiting high-
temperature line width by 340 K. This behavior is interpreted
as the motional rate of the ring flip process moving initially
into the MAS broadening regime35 and then subsequently at
higher temperatures into the dipolar broadening regime.27
The observation of the temperature-dependent behavior of
the aromatic resonances in the 13C CP/MAS NMR spectrum
arising as a consequence of the phenoxy ring reorientation
provides an opportunity to extract quantitative kinetic informa-
tion and activation parameters. The results of three independent
and complementary techniques, chemical exchange lineshape
analysis, dipolar broadening analysis and magnetization transfer
experiments, were obtained from studies of CspenV. Between
205 and 225 K exchange is too slow to result in significant
Figure 4. The crystal structure of HpenV shown in projection down
a. The coordinates used are those of Abrahamsson et al.,7 inverted to
give the correct enantiomer. The molecular packing arrangement leads
to hydrogen-bonded molecular chains parallel to b.
iii. Anisotropic Displacement Parameters. The atomic
displacement ellipsoids (Figures 1-3 for the Cs, Na, and Li
salts) of the atoms of the phenoxy groups show no evidence
for any disorder. The ellipsoids may be viewed both in
projection on to and perpendicular to the plane of the phenoxy
group, Figure S2. In two structures (KpenV and CspenV) larger
amplitude in-plane librational motion is suggested, but there is
little significant difference (except perhaps for molecule 3 of
the KpenV structure) in the out-of-plane motion. For LipenV
the ring of anion 1 has larger in-plane components to its thermal
motion than that of anion 2.
lineshape changes, but the kinetic behavior was studied by 13
C
CP/MAS magnetization transfer experiments24 between the C2′
and C6′ resonances.
Between 213 and 244 K chemical exchange broadening of
the C2′,C6′ resonances is evident, and rate constants were
determined by comparison of the experimental spectra with
calculated lineshapes (DNMR4, QCPE 46626) (Figure S4). The
rate data fit an Arrhenius rate law with an activation energy of
84.7 kJ mol-1 and a frequency factor of 7.5 × 1021 s-1. The
rates at 241 and 244 K deviate from linearity, as the contribu-
tions to the line widths from MAS broadening become
significant.
Above 282 K in the short correlation time regime of the the
dipolar broadening interaction, where the effects of exchange
broadening are negligible, the line broadening was analyzed to
provide additional rate data. The deconvoluted line widths,
∆ν1/2′ ) ∆ν1/2 - (∆ν1/2)o, of both exchange-averaged isotropic
resonances, (C2′,C6′) and (C3′,C5′), were measured, and
ln(∆ν1/2′) for each was plotted against the inverse temperature.
Fits to the theory of Rothwell and Waugh27 yield activation
barriers of 68.8 and 65.5 kJ mol-1 for C2′,C6′ and C3′,C5′
respectively, (Figure 8). These values are consistent and indicate
well-defined Arrhenius behavior.
b. 13C CP/MAS NMR Spectroscopy. 13C CP/MAS NMR
spectra of HpenV, LipenV, NapenV, RbpenV, and CspenV at
ambient temperature are shown in Figure 6. The spectra show
well-resolved, widely dispersed resonances of good signal-to-
noise ratio; bands of resonances from carbon atoms of distinct
functional groups can be identified immediately. Detailed
assignments were made as discussed previously for KpenV,1a
by identifying 13C resonances of quaternary and mobile carbon
atoms from NQS experiments,20 and carbon atoms bonded
directly to nitrogen by their characteristic asymmetric doublets,30
and by comparison with spectra of structurally related com-
pounds,4,31 particularly KpenV.1a
i. CspenV. In the ambient temperature spectrum of CspenV
(Figure 6f) the very sharp well-resolved lines in the upfield
region, notably those of the 2R- and 2â-methyl groups,4 are
clearly evident. Only a single resonance is apparent for each
resolved carbon site, indicating that the crystals contain one
molecule in the asymmetric unit. The aromatic region shows
only two sharp resonances, which are assigned to the ipso C1′
and para C4′ carbons, and two very broad resonances which
are assigned to the remaining four aromatic carbon sites. Similar
spectra have been observed before for penicillins, specifically
KpenV,1a sodium carfecillin,32 and KpenG.29 This pattern is
interpreted as indicative of a 180° flip motion of the phenyl
rings about their local 2-fold axes and has been discussed in
detail in respect of KpenV.33
The rate constants determined from the three independent
analyses are plotted together in Figure 9 and combined analysis
yields activation parameters of Ea ) 72 ( 10 kJ mol-1 and Ao
) (1.3 ( 1) × 1019 s-1
.
(33) The pattern of nuclear exchange indicated by magnetization transfer
experiments; the observation of pairwise exchange broadening and coales-
cence of certain phenyl resonances; the extent of the averaging of chemical
shift anisotropy (CSA) indicated with rise in temperature; the observation
of an extensive 13C-1H dipolar broadening regime; the absence of any
positional disorder in the X-ray crystal structuressall these observations
are consistent with 180° ring flip motions as the single most significant
dynamic process in these crystals.
(34) On lowering the temperature below 220 K the methyl resonances
broaden progressively. This is interpreted as the methyl group C3-
reorientation slowing into the dipolar broadening regime.27
(35) Suwelack, D.; Rothwell, W. P.; Waugh, J. S. J. Chem. Phys. 1980,
73, 2559-2569.
The variation with temperature of the aromatic region of the
13C CP/MAS NMR spectrum in the temperature range 210-
(30) Hexem, J. G.; Frey, M. H. Opella, S. J. J. Am. Chem. Soc. 1981,
103, 224-226. Hexem, J. G.; Frey, M. H. Opella, S. J. J. Chem. Phys.
1982, 77, 3847-3856.
(31) Twyman, J. M.; Dobson, C. M. J. Chem. Soc., Chem. Commun.
1988, 786-788. Twyman, J. M. D. Phil Thesis, University of Oxford, 1989.
(32) Twyman, J. M. D. Phil Thesis, University of Oxford, 1989.
Wendeler, M.; Heyes, S. J. Unpublished results.

Studies of Molecular Crystals of Penicillin V
J. Am. Chem. Soc., Vol. 119, No. 41, 1997 9799
Figure 5. The nine penV anion conformations observed in the crystal structures of (from left to right) HpenV, LipenV (two molecules), NapenV,
KpenV (four molecules), and CspenV, illustrating the observed conformational variability of the phenoxy portion of the side chain and the similarity
and apparent conformational rigidity of the penam nucleus. The relevant torsion angles are listed in Table 2.
Table 2. Selected Torsion Angles Referred to Atom Numbering
on the NMR Convention Presented in 1
atoms
angles
1
2
3
4
A
B
C
D
E
C5
C6
N14
C15
C17
C6
N14
C15
C17
O18
C1′
C15
C17
O18
C1′
N14
C15
C17
O18
C6′
molecule
A, deg
B, deg C, deg D, deg E, deg
HpenV
LipenV
-155.6
-178.6
-161.2
170.5
176.5
179.5
5.1
5.1
160.8
174.4
15.3
16.7
2.3
1
2
6.5 -168.0
NapenV
KpenV (293 K)
162.4 -165.2 -14.3
161.2 -14.7
1
2
3
4
173.6 172.3 -1.2 -160.0 -51.5
168.7 -176.2
165.5 -178.0
178.5
8.4
11.0
12.8
13.3
-86.7
161.0
87.0
0.4
77.3
3.2
171.6
CspenV
172.3 -179.7
89.1
25.0
ii. RbpenV. The ambient temperature 13C CP/MAS NMR
spectrum of RbpenV (Figure 6e) is very well-resolved, but is
considerably more complex that that of CspenV, resembling
closely the spectrum of KpenV.1a The changes in the spectrum
with temperature (Figure S6) also parallel those observed and
analyzed in detail previously for KpenV.1a They indicate that
RbpenV, like KpenV, has a low temperature phase with a
complex crystal structure having four molecules in the asym-
metric unit. At ∼355 K a phase change occurs to a structure
with two molecules in the asymmetric unit.
The aromatic region of the spectrum at 202 K (Figures 7b
and S7) shows C2′,C6′ and C3′,C5′ aromatic carbons of all four
molecules in the asymmetric unit to be averaged into the fast
limit of the exchange-broadening time scale by ring flip
processes. As the temperature is raised above 202 K, these
resonances broaden gradually to the limits of observability at
∼240 K and then begin to narrow, becoming fully resharpened
by 296 K. In contrast to CspenV, therefore, the aromatic ring
motions of RbpenV could not be frozen out into the slow
exchange limit at the lowest temperature accessible to us. This
fact, and the observation that maximum dipolar broadening
occurs at ∼240 K in comparison with 282 K for CspenV, is
indicative of a faster rate of ring flip at a given temperature for
RbpenV than CspenV. RbpenV shows similar rates of ring flip
to KpenV,1a but in contrast to KpenV the identical broadening
behavior of the different resolved C2′,C6′ resonances indicates
the phenoxy rings of the four crystallographically inequivalent
molecules to have essentially identical dynamic behavior.
Analysis of the dipolar line broadening in the short correlation
time regime above 278 K indicates an Arrhenius activation
barrier of Ea ) 51 ( 4 kJ mol-1. Substitution of the known
rate of exchange (k ) 9.4 × 105 s-1) at the temperature of
Figure 6. Ambient temperature (296 K) 13C CP/MAS NMR spectra,
with annotated assignment, for crystalline samples of (a) HpenV, (b)
LipenV, (c) NapenV, (d) KpenV, (e) RbpenV, and (f) CspenV (s
denotes spinning sidebands).
maximum broadening (240 K) yields a preexponential factor
of Ao ) 9.1 ( 3 × 1016 s-1
.

9800 J. Am. Chem. Soc., Vol. 119, No. 41, 1997
Wendeler et al.
Figure 7. Parts of the aromatic regions of the 13C CP/MAS NMR spectra of (a) CspenV, (b) RbpenV, (c) NapenV, (d) LipenV, and (e) HpenV
at selected temperatures. For LiPenV care was taken to ensure that the salt was in the fully hydrated form.
iii. NapenV. The ambient temperature spectrum of NapenV
(Figure 6c) like that of CspenV shows broadening effects due
to aromatic ring dynamics and also indicates the presence of
only one molecule in the asymmetric unit. The 13C CP/MAS
spectrum for the aromatic region in the temperature range 232-
386 K (Figures 7c and S5) shows discrete resonances at e268

Studies of Molecular Crystals of Penicillin V
J. Am. Chem. Soc., Vol. 119, No. 41, 1997 9801
Figure 8. Plots of ln(∆ν1/2′) vs (1/T), where ∆ν1/2′ is the difference
between the measured and limiting (200 K) line widths, for C2′,C6′
and C3′,C5′ resonances of the 13C CP/MAS NMR spectra of CspenV.
Both resonances show behavior typical of a reintroduction of dipolar
broadening, and at lower temperatures (longer correlation times) both
sets of line width data show the affects of MAS broadening, and the
C2′,C6′ additionally, and more dramatically, exchange-broadening
phenomena. Solid lines represent for each data set a fit, by a least-
squares, error-weighted procedure, to the theory of Rothwell and
Waugh,27 restricted to data in the region of maximum broadening and
at higher temperatures into the short correlation time regime. In principle
the entire C3′,C5′ data set could be fit to a combination of dipolar and
MAS broadening expressions, but in practice there is little to gain from
this approach. The dipolar broadening data are incorporated into the
Arrhenius fit to the composite rate data from the various experiments
responding to the different timescales (Figure 9) by conversion of the
line widths to rate constants, with an accuracy influenced primarily by
the accuracy of the estimation of the temperature of maximum dipolar
broadening achieved through the above fit.
Figure 10. Principal aromatic region of a rotationally asynchronous
phase-sensitive 13C CP/MAS 2-D exchange experiment (MAS rate νr
of 3.95 kHz, H rf decoupling field ν1H ) 55 kHz, 512 t1 increments
1
of 108 µs, and a mixing time of 1.0 s) on LipenV, performed at 343
K, and showing strong magnetization transfer between the inner pair
of C2′, C6′ resonances and less intense cross peaks between the outer
pair of C2′,C6′ resonances. This suggests that the rate of ring flips is
different for the two inequivalent molecules in the structural asymmetric
unit. The C3′,C5′ region also shows cross peak structure, but the smaller
spread in chemical shift leads to poorer resolution of the individual
cross peaks.
were determined from exchange-broadened lineshape simula-
tions in the temperature range 268-309 K (Figure S4) and
magnetization transfer data measurements in the range 232-
261 K for the C2′,C6′ resonances. Arrhenius behavior is found
over the entire temperature range with an activation energy for
the phenoxy ring flips of 68 ( 3 kJmol-1 and a frequency factor
Ao of (7.1 ( 5) × 1014 Hz.
iv. LipenV. In the ambient temperature spectrum of
LipenV‚H2O (Figure 6b), the methyl carbon resonances in the
upfield region of the spectrum show particularly clearly two
peaks for each of the 2R- and 2â-methyl carbon atoms. This
indicates that the crystals contain two molecules in the crystal-
lographic asymmetric unit. Similar splittings are evident in the
aromatic region; two pairs of C2′,C6′ aromatic resonances
arising from the two molecules are clearly resolved at 296 K.
At g377 K36 the inner pair of the C2′,C6′ lines broadens and
loses intensity (Figures 7d and S8) while the outer two peaks
remain sharp; the C3′,C5′ resonance region also shows some
broadening. Even at the maximum temperature studied, 384
K, the rate of aromatic ring motion is evidently too slow to
result in coalescence and eventual resharpening of exchange-
averaged resonances, as seen in the spectra of the other salts
discussed above.
Figure 9. Combined Arrhenius plots for HpenV, LipenV, NapenV,
RbpenV, and CspenV, with rate constant data derived from magnetiza-
tion transfer experiments (b), chemical exchange line shape simulations
(4) and dipolar broadening analysis (0). The filled boxes (9)
correspond to the points of maximum broadening. The solid lines
indicate Arrhenius relationships derived from fits, weighted according
to the estimated experimental errors (not indicated on this plot). Li(I,II)
indicates the two independent molecules in LipenV. There was no
differentiation in the dipolar broadening of the various C2′,C6′
resonances of the four independent molecules of RbpenV, so a single
set of rate data is shownsthat from the most upfield of the resonances,
at ca. 114 ppm.
A
13C CP/MAS 2D exchange experiment23 with a mixing
time of 1 s performed at 343 K (Figure 10) reveals strong
magnetization transfer between the inner pair of C2′,C6′
resonances and less intense cross peaks between the outer pair
of C2′,C6′ resonances. These observations are consistent with
slow ring flip motions, occurring at different rates for the two
molecules in the asymmetric unit. The dynamic behavior was
probed by variable mixing time 1D magnetization transfer
experiments over the temperature range 325-365 K and by
lineshape analysis from 365-384 K. The derived rate data
K for each of the six inequivalent aromatic carbon atoms. As
the temperature is increased above 268 K, both the C2′,C6′ and
C3′,C5′ carbon resonances of the phenyl ring display classical
two-site exchange lineshape changes with coalescence at 285
K for C2′,C6′ and at 305 K for C3′,C5′. Above these
temperatures the averaged resonances initially sharpen, but
beyond 324 K they broaden again progressively with increase
in temperature until at 386 K they are barely resolved from the
baseline. Rates of the ring flip exchange process (Figure 9)
(36) On heating LipenV‚H2O much above 340 K for significant periods
of time an additional set of peaks became evident in the 13C CP/MAS NMR
spectra (Figure S1) at increasing intensity with time. Thermogravimetric
analysis (TGA) under a flowing argon atmosphere indicated loss of crystal
water above the same temperature, to an extent dependent on the duration
of the heating and the sample temperature. Placing material that had been
heated in this manner in air at ambient temperature overnight resulted in
full recovery of crystal water molecules, as shown by the 13C CP/MAS
NMR spectrum, which is identical to that of the sample prior to heating.

9802 J. Am. Chem. Soc., Vol. 119, No. 41, 1997
Wendeler et al.
Table 3. Arrhenius Parameters and Structural Data for Phenyl Ring Flips in PenV Salts
PenV
Salt
Ea, kJ
C-C short,
(C-C short)av,
phenyl layer
width/phenyl units
mol-1
Ao, s-1
Å
Å
H
89 ( 3
111 ( 8
102 ( 15
68 ( 3
(4.4 ( 5) × 1015
(4.6 ( 5) × 1017
(2.7 ( 5) × 1015
(7.1 ( 5) × 1014
(9.1 ( 3) × 1016
(1.3 ( 1) × 1019
3.544
3.387
3.465
3.672
3.636
3.567
3.728
3.7485
0.75
0.0
0.0
1.25
1.75
1.5
Li(1)
Li(2)
Na
Rb
Cs
51 ( 4
72 ( 10
3.720
3.7665
(Figure 9) fit simple Arrhenius models; activation parameters
of Ea 102 ( 15 kJ mol-1 with Ao ) (2.7 ( 5) × 1015 s-1, and
Ea ) 111 ( 8 kJ mol-1 with Ao ) 4.6 ( 5 × 1017 s-1 for the
two distinct molecules.
spatially demanding motion such as a phenyl ring flip, and the
activation barrier should be temperature dependent.
X-ray structure analysis shows that the Li+, Na+, K+, Rb+,
and Cs+ salts of penV and the free acid exhibit a wide variety
of crystal structures (only KpenV and RbpenV are isomorphous).
The penam unit is structurally rigid and adopts an almost
identical C-3 conformation consistently throughout the penV
structures. The phenoxymethyl side chain shows an extremely
wide range of conformations with respect to the penam unit.
Although by this measure the side chain seems to be remarkably
flexible, only in the high temperature forms of K and RbpenV
is there significant motion of the side chain between alternative
conformations. The adoption of a specific side-chain conforma-
tion appears to be driven by the need, in a given structure, for
the packing arrangements to satisfy other structural criteria, such
as a preferred coordination environment of the cation (see
below). The activation barrier to ring flips is not expected to
have a significant intramolecular contribution, and indeed there
is no obvious direct relationship between the side chain
conformation (any or all of the torsional angles listed in Table
2) and flip rates or Arrhenius activation energy to ring flip.
However, the molecules with the most compact conformations,
CspenV and two of the K(Rb)penV conformers, are those that
show the most rapid ring flips at typical temperatures, and the
most elongated conformations (e.g., molecule 1 of LipenV) tend
to show the highest activation barriers to ring flips. It should
be noted that the four different molecules of the RbpenV
structure show widely differing torsion angles, but NMR
indicates very similar ring flip dynamics for each.
v. HpenV. The ambient temperature spectrum of the free
acid HpenV (Figure 6a) indicates a single molecule in the
asymmetric unit. Two narrow, well-resolved resonances are
observed from the C2′,C6′ phenoxy sites, between which strong
magnetization transfer is observed in a 13C CP/MAS 2D
exchange experiment23 with a mixing time of 2 s performed at
297 K (Figure S10). This indicates phenoxy ring flips in HpenV
at a rate in the slow limit of the exchange broadening regime at
this temperature. At g 340 K (Figures 7e and S9), the (C2′,
C6′) phenoxy line widths increase, in the manner of a classical
exchange broadening, with coalescence at 365 K, initial
sharpening of the averaged resonance before broadening again
at the highest temperature measured, 380 K. Simulation of the
exchange-broadened line shapes between 329 and 365 K (Figure
S9), and variable-temperature 1D-magnetization-transfer experi-
ments in the temperature range 307-335 K yield ring flip rates
which follow Arrhenius behavior with an activation barrier of
Ea ) 89 ( 3 kJ mol-1 and a frequency factor of Ao ) (4.4 (
5) × 1015 s-1
.
Discussion and Conclusions
One hundred eighty degree flip motions of the aromatic rings
have been detected by 13C CP/MAS NMR in the temperature
range 200-380 K for all the penV salts studied here. In each
case the rate data can be fitted successfully to Arrhenius
relationships (Figure 9 and summarized in Table 3). At any
given temperature the flips occur at very different rates in the
different salts. For example, at 340 K the same fundamental
motion, 180° ring flips, is measured in CspenV with k ) 3.6 ×
108 s-1 through dipolar broadening effects and in LipenV with
k ) 0.5 s-1 through magnetization-transfer experiments. At
ambient temperature measured and predicted (from Arrhenius
behavior) rates cover 11 orders of magnitude. Only for CspenV
are full data in three motional regimes (magnetization transfer,
exchange broadening and dipolar broadening) available. Lin-
earity is, however, observed for the NapenV data over a similarly
wide temperature range, and the fits over more restricted
temperature regimes for HpenV, LipenV, and RbpenV are no
less convincing.
The observation of Arrhenius behaviour at all temperatures
and timescales is very striking. Given the changes that occur
in a typical molecular crystal through this temperature range,
including overall expansion of the crystal (estimated, from
typical organic molecular crystal data as indicated by Dunitz37
and from our own variable-temperature single-crystal X-ray
study of KpenG,29 to be of the order of a 2% change in volume)
it is remarkable that simple Arrhenius behavior is observed.
Expansion of the unit cell with increase in temperature would
be imagined to lead to a decrease in the energy barrier to a
The crystal structures of all the salts and the free acid have
the same basic packing motif, the exact form of which appears
to be determined by the coordination requirements of the cation
(or, for the free acid and partially for LipenV hydrate, H-
bonding). The structural theme is that of a molecular sandwich
with a hydrophilic core and a lipophilic surface. All the
structures can be considered as comprising of layers that are
two molecules in thickness. All structures contain regions in
which electrostatic interactions between the cations and oxygen
atoms dominate intermolecular interactions and other regions
with primarily van der Waals interactions. It is within the layers
that strong Coulombic forces dominate the structural arrange-
ment, but adjacent layers interact only through much weaker
dispersion forces. The principal change in the series of crystals
is a systematic variation in the radius of the atomic countercation
to the penV anion from H+, through Li+, Na+, K+, and Rb+ to
Cs+. With the increase in size of the cation the number of
ligator atoms at the cation increases from 2 to 7, thus increasing
the number of penV oxygen atoms that engage the cation and
so changing the orientation of the anion with respect to the plane
of the sandwich structure. In one respect these structures are
alike: they all have a rigid ionic core with increasing molecular
motion toward the lipophilic surface. The increased motion is
indicated by increasing XRD adps toward the lipophilic surface
and the consistent observation of 180° flips of the phenyl rings
(37) Dunitz, J. D. X-Ray Analysis and the Structure of Organic Molecules,
2nd ed., 2nd repr.; VCH: Weinheim, 1995.

Studies of Molecular Crystals of Penicillin V
J. Am. Chem. Soc., Vol. 119, No. 41, 1997 9803
from the 13C CP/MAS NMR, albeit at flip rates that span orders
of magnitude at a given temperature.
The ring flip event does not lead to positional disorder and
so is essentially diffraction invisible. The adps of the phenyl
ring atoms in the crystal structures relate to the extent of the
faster (t ≈ 10-15 s) vibration of the atoms and librational motions
of the phenyl groups. Previous studies have shown that the
phenoxy and benzyl rings of penicillin V benzyl ester or of its
â-sulfoxide have very different ring flip rates and Arrhenius
activation parameters.31 Yet, single-crystal X-ray studies at
ambient temperature and also at 143 K for the â-sulfoxide38
show very similar adps for phenoxy and benzyl rings at a given
temperature, thus suggesting that there is no specific correlation
between librational amplitudes and the rates of the slower,
activated ring flip process. TLS analyses39 of the adps of the
phenoxy groups of NapenV, and CspenV and molecule 2 of
LipenV are consistent with rigid body motion, but that of
molecule 1 of LipenV is not. On the grounds of the nature of
its closest contacts the phenyl ring of molecule 1 may be judged
the ring most “entrapped” in any of these structures, and is thus
the least likely to undergo ring flip motion. The maximum
librational amplitude is greatest for CspenV and is approximately
equal for LipenV (molecule 2) and NapenV. The libration
analysis for CspenV is consistent with that for the isostructural
KpenG (benzyl penicillin) for which variable-temperature single-
crystal XRD data are available and have been analyzed in
detail.29
Looking in detail at the local environment of the structures,
in LipenV, which has the lowest flip rates and the highest
activation barriers, the phenyl groups lie between pairs of amide
groups of penV anions from the other molecule type in the
asymmetric unit, in the neighboring layer. Deformation of the
surroundings is likely to be difficult, explaining the relatively
slow rates of ring flips. HpenV also has a low rate of and a
high activation barrier to ring flips, and in this structure it is
the C2-methyl groups of molecules from a neighbouring layer
that provide the most rigid obstacle to flips. In the NapenV
structure, the extended conformation of the side chains leads
to the closest contacts to a phenyl ring being to atoms of the
phenyl rings of molecules in the adjacent layer. In addition,
the C2 axes of adjacent rings are oriented at ∼90° to each other
in a plane orthogonal to the layers. This offers adjacent rings
a better opportunity to be displaced to enable a neighboring
ring flip event to occur. In the K(Rb)penV and CspenV
structures the compact conformation of the penicillin V sidechain
leads to closest contacts with methyl groups of the penam unit,
either intramolecularly or with those of a molecule in the
adjacent layer but the increased size of the cation opens up the
contact distances between adjacent phenyl groups within a layer,
giving a generally more open structure. There is, therefore a
qualitative correlation between the rates of aromatic ring flips
and the structure of the crystals in which they are located.
A variety of structural parameters from each penV crystal
were examined for possible correlation with the observed ring
flip parameters. Some of these parameters do correlate surpris-
ingly well with the propensity for ring flips in the different salts
(Figure S11). However, ring flip events clearly occur when
the structure has temporarily “opened up” around the ring. It is
not surprising then that a single parameter from the equilibrium
structure determinations is inadequate to summarize the quality
of the local environment of a ring in regard of its encouragement
to the nonequilibrium ring flip events. We sought nevertheless
Figure 11. Correlation of the Arrhenius activation barriers, Ea, to
phenyl ring flips with a measure of the width of the layer motif of the
phenyl groups. The measurement unit is the width of the phenyl ring
projected onto the c axis and so differs between the various penV salts
as a result of different orientations of the plane of the phenyl ring with
respect to the lamellar motif. Values are defined only to the nearest
0.25 of a unit. This measurement unit has been adopted with the aim
of encapsulating the effects of layer width and relative phenyl ring
orientations within a single structural parameter tailored for prediction
of ease of ring flip.
to assess the qualities of the molecular packing arrangements
that might indicate the relative ease of ring flips in the closely
structurally similar set of penicillin crystals. We note that when
the phenyl groups of adjacent sandwiches interleave (LipenV
and HpenV) the flip rates are low but when the phenyl groups
of adjacent sandwiches just touch (CspenV, RbpenV, KpenV,
and NapenV) the rate of ring flips is much faster, and of the
latter group the ring flips for RbpenV and CspenV, the most
open structures, are fastest throughout the temperature range.
A similar pattern is found for the activation barriers to ring flip,
which are larger for the intercalated than the touching structures.
The energy barriers are lower for the more complex, and open,
structure of the isomorphous KpenV and RbpenV than for the
simpler structures of NapenV and CspenV. Although our
conclusion is that these slow, spatially demanding motions are
determined by a complex interplay between different structural
factors, we find a pleasing correlation of the activation energy
to ring flips with the phenyl layer spacings in the structure
(Figure 11).
We believe that slow motions in molecular crystals of the
type discussed in this paper represent a widespread phenomenon
and that analysis of the underlying determinants of these
motional properties is possible using a combination of NMR
and diffraction techniques. The observation of activation
parameters for a range of related structures offers an invaluable
information source with which to attempt to model the ring flip
processes. Such analyses could have substantial implications
for our understanding not just of intermolecular interactions
within the crystal, but of interactions within and between
macromolecular species including synthetic and natural poly-
mers and macromolecules, for which similar motions have been
observed.
Acknowledgment. J.M.T. thanks the EPA Cephalosporin
and Educational Trusts and Pembroke College for financial
assistance. S.J.H. thanks the Glasstone Benefaction for a
Fellowship, M.W. thanks the Studienstiftung Deutschlands for
a Scholarship, and A.J.E and J.F. thank the SERC for financial
support.
Supporting Information Available: Tables of atomic
coordinates, atomic displacement parameters, and bond distances
and angles and 11 figures referred to in the text as Figures S1-
S11 (25 pages). See any current masthead page for ordering
and the Internet access instructions.
(38) Baird, P; Twyman, J. M.; Dobson, C. M.; Prout, K. Unpublished
results.
(39) Wendeler, M.; Edwards, A. J.; Heyes, S. J.; Prout, K. Unpublished
results.
JA9640657