Effects of polymorphic differences for sulfanilamide, as seen through 13C and 15N solid-state NMR, together with shielding calculations

Article abstract of DOI:10.1002/mrc.1351

We recorded both carbon-13 and nitrogen-15 NMR spectra of the three solid forms of sulfanilamide most commonly known. This study led to an interpretation of the solid-state effects seen in cross-polarization magic angle spinning spectra. Relaxation times

Full text of DOI:10.1002/mrc.1351

MAGNETIC RESONANCE IN CHEMISTRY
Magn. Reson. Chem. 2004; 42: 313–320
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/mrc.1351
Effects of polymorphic differences for sulfanilamide, as
seen through 13C and 15N solid-state NMR, together
with shielding calculations
Alessia Portieri,1 Robin K. Harris,1∗ (the late) Richard A. Fletton,2 Robert W. Lancaster2
and Terence L. Threlfall3
1
Chemistry Department, Durham University, South Road, Durham DH1 4LE, UK
GlaxoSmithKline R & D Ltd, Medicines Research Centre, Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, UK
Department of Chemistry, University of Southampton, Southampton SO17 1BJ, UK
2
3
Received 29 August 2003; Accepted 06 November 2003
We recorded both carbon-13 and nitrogen-15 NMR spectra of the three solid forms of sulfanilamide most
commonly known. This study led to an interpretation of the solid-state effects seen in cross-polarization
magic angle spinning spectra. Relaxation times for the different forms were measured. These show
different behaviour for the three forms, arising from mobility variations. To obtain information on local
environments, static spectra and spinning sideband manifolds were recorded and analysed for the 15N
resonances, using isotopically enriched samples. Shielding asymmetries and anisotropies for the two
nitrogen nuclei were obtained, showing very different behaviour for the two sites. Shielding calculations
were carried out for both 13C and 15N nuclei, and the results are discussed in relation to the experimental
values. Copyright  2004 John Wiley & Sons, Ltd.
KEYWORDS: NMR; CP/MAS NMR; 13C NMR; 15N NMR; polymorphs; sulfanilamide; shielding computations; hydrogen
bonding; relaxation
linked with molecular and intermolecular structural infor-
mation.
In our laboratories, we have been studying in detail
aspects of the solid-state behaviour of the sulfa drugs:
INTRODUCTION
The term ‘polymorphism’ refers to the existence of different
molecular packing arrangements of a given compound
in the solid state.1–3 The study of polymorphism and
related solid-state phenomena has a long history, but
has been given added impetus within the past decade,
largely because of increasing awareness of its importance
in the pharmaceutical industry. Unrecognized polymorphs,
hydrates and amorphous forms, or unintended polymorphic
transitions, can adversely affect the processing, stability and
bioavailability of a drug substance in solid dosage forms.4,5
Polymorphism and hydration have been associated with
patent protection issues with major financial implications.4
Despite their importance, our ability to predict and some-
times even systematically prepare polymorphic crystal forms
is limited.6 The preparation and detection of polymorphs
is therefore largely an empirical matter. Although many
analytical methods can be used, often in complementary
fashion,1,7–10 solid-state NMR spectroscopy has been recog-
nised as a major tool in the detection and quantification
of polymorphs and in the study of polymorphism.11,12 Its
virtues include high discrimination between polymorphs,
6
5
1
4
1NH2
SO22NHR
2
3
R = H
sulfanilamide
sulfapyridine
R = 2-pyridyl,
R = 2-thiazolyl, sulfathiazole
R = 2-pyrimidyl, sulfadiazine
These derivatives of sulfanilamide, which act as antibac-
terials, are extraordinarily versatile in their ability to
crystallize in multiple solid-state structures. The polymor-
phic behaviour of the sulfa drugs has been repeatedly
investigated13–15 since their discovery in the late 1930s,
but our recent studies have revealed unusual solid-state
behaviour, new polymorphs and hundreds of solvates.16,17
Crystal structures of the three most accessible poly-
morphs of sulfanilamide itself (˛, ˇ and ꢀ), obtained from
single-crystal x-ray diffraction measurements, are to be found
in the literature.18–20 and can be retrieved from the Cam-
bridge Crystal Structure Database. However, these data are
relatively old and proved to be insufficiently accurate for
our purposes, especially since they contain no information
ŁCorrespondence to: Robin K. Harris, Chemistry Department,
Durham University, South Road, Durham DH1 4LE, UK.
E-mail: r.k.harris@durham.ac.uk
Contract/grant sponsor: UK Engineering and Physical Sciences
Research Council; Contract/grant number: GR/N05635.
Copyright  2004 John Wiley & Sons, Ltd.

314
A. Portieri et al.
about proton positions. We also used data obtained for
one form by neutron diffraction experiments.21 In addition,
recent x-ray diffraction results for the three forms, of greater
precision, have been obtained (M. J. Hursthouse and T. L.
Threlfall, unpublished results; A. Portieri, R. K. Harris and
H. Puschmann, unpublished results).
density functional theory (DFT) calculations of the shield-
ing tensors of both 13C and 15N were carried out both for
isolated molecules with the reported solid-state geometries
and for molecular clusters. The effects of conformation and
hydrogen-bonding patterns are discussed.
Otherpolymorphs(usuallyreferredtoasυ-sulfanilamide)
have been sporadically reported.14,15,22 We have recently
determined the crystal structure of a new polymorph, but
have not been able to produce a sufficient quantity of pure
material for spectroscopic characterization. Its calculated
powder XRD pattern and stability indicate that it is not
the υ polymorph reported by Sekiguchi et al.14 An unstable
anhydrate appears erratically during attempts to crystallize
sulfanilamide polymorphs from aqueous solution. Its crys-
tal structure has been reported.23 The stability relationships
between the polymorphs remain confused despite extensive
investigation. The stable form at room temperature is ˇ-
sulfanilamide. All the other forms gradually revert to this
on storage, although the lifetimes vary from sample to sam-
ple. On slow heating, ˇ-sulfanilamide changes slowly to ˛,
which then transforms to ꢀ, the highest melting form. There-
fore, the ꢀ and ˇ forms must be enantiotropically related,
but the situation in respect of ˛/ˇ and ˛/ꢀ is not clear.
Burger13 and Toscani and co-workers24,25 reported the ˛ and
ˇ forms as monotropically related, but this is partly based on
small melting enthalpy differences on samples of doubtful
polymorphic purity. Sekiguchi et al.14 suggested that they
are enantiotropes with a transition point below room tem-
perature. The usual (kinetic) transformation temperature
recorded under typical differential scanning calorimetric
EXPERIMENTAL
Commercial sulfanilamide was used for the 13C NMR
study, and the three forms were crystallized using literature
techniques.14,24 For the ˛ form, about 500 mg of the sample
were dissolved in 15 ml of n-butanol and the solution was
°
heated to 90 C and then cooled to ambient temperature.
Long, needle-shaped crystals appeared when the residual
solvent was left to evaporate overnight. For the ˇ form, about
the same amount of sample was dissolved in water, which
was then heated until the water evaporated to saturation
conditions. Just after removing the solution to cool, crystals of
the ˇ form started to crystallize. The ꢀ form was crystallized
°
from ethanol. The solution was heated to about 80 C and left
to cool rapidly by positioning the flask in an ice-bath.
The isotopically enriched samples were synthesized by
enriching in turn the two nitrogens of the molecule.
For the enrichment of N(2), the procedure followed was
similar to that given in the literature31,32 Thus, 12.5 ml of
sulfonyl chloride were added to acetanilide (5 g) and left
refluxing for about 1 h until no solid acetanilide was left. Ice
was added to the solution, which was then filtered. A mixture
of ammonium sulfate (Aldrich, 82.6% 15N; 9 mmol), the
sulfonyl chloride acetanilide previously prepared (18 mmol)
and K2CO3 (72 mmol) in acetonitrile (100 ml) was cooled
in an ice-bath. Water (72 ml) was then added, the flask
was stoppered and the mixture was stirred magnetically
overnight. The organic layer was separated, the solvent was
removed under vacuum and the residue was recrystallized
from water. Then, following reflux with acid hydrolysis
(2.5 ml of concentrated HCl and 7.5 ml of H2O), the final
product was obtained. Finally, the resulting sulfanilamide
was crystallized in the different forms.
°
heating rates is just above 100 C. Our own work suggests
°
a true transition temperature between 75 and 90 C. The
slow and variable transformation rates and the persistent
formation of concomitant polymorphs26 are the main rea-
sons for the difficulty in determining the thermodynamic
relationships.
In this paper, a detailed 13C and 15N solid-state NMR spec-
tral study is presented of sulfanilamide and, in particular, of
its three most accessible polymorphs: ˛, ˇ and ꢀ. Cross-
polarization magic angle spinning (CP/MAS) 13C NMR
spectra of sulfanilamide have been presented previously,27
although the assignment of some of the resonances was
ambiguous because of the difficulties in crystallizing pure
polymorphs, and also issues concerning the limited stability
of the ˛ form. Here we report a re-investigation of the 13C
NMR spectra of the sulfanilamide polymorphs, with a clar-
ification of the assignments, aided by a study of the proton
spin–lattice relaxation times in the laboratory and rotating
frames, T1(H) and T1ꢁ(H), respectively. We also measured
the high-resolution 15N CP/MAS spectra of 15N-enriched
samples for all the three solid forms for the first time. The
value of 15N spectra as an aid to the understanding of poly-
morphic behaviour has been shown previously in respect
of dyestuffs28,29 and pharmaceuticals.30 In the present work,
the separate enrichment of sulfanilamide at the N(1) and the
N(2) positions allowed an insight into molecular mobility.
The principal components of the nitrogen shielding ten-
sors were also determined. Finally, Hartree–Fock (HF) and
For the enrichment of N(1), 99%-enriched acetanilide
(2 g) was purchased from Aldrich and 15N-enriched sulfonyl
acetanilide was obtained as above. Ammonia was added to
it, and the solution was heated for about 1 h. The hydrolysis
and crystallization were carried out as above.
Carbon-13 CP/MAS spectra were obtained at ambient
probe temperature (ca 298 K) using a Chemagnetics CMX
200 spectrometer operating at 50.329 MHz (corresponding
to 200.13 MHz for protons). A two-channel Chemagnetics
probe was used, with 7.5 mm o.d. ‘pencil’ rotors made
of zirconia and Kel-F end-caps. The rotors contain about
350 mg of sample. Samples were packed into rotors without
further grinding in order to minimize polymorphic changes.
Operating conditions typically involved 5 µs 90 1H pulses
°
and decoupling powers equivalent to 50 kHz. The recycle
delay varied significantly, depending on the relaxation time
of the particular form measured. A contact time of 1 ms was
used for all forms and the spin rate was about 4 kHz. The
experiments were carried out using different conditions for
each polymorph. For T1(H) measurements (via 13C spectra
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

CP/MAS NMR sulfanilamide polymorphs 315
by pre-contact saturation-recovery) for the ˛ and ꢀ forms,
a recycle delay of 240 s was used and the recovery times
were arrayed as 0.1, 5, 10, 20, 50, 80, 100, 120 and 150 s. The
number of acquisitions per recovery time was 32. However,
for the ˇ form spin–lattice relaxation time measurement, the
recycle delay was 60 s, the number of transients was 32 and
the recovery times were arrayed as 0.1, 1, 2, 3, 5, 8, 10, 12, 15,
20, 30 and 40 s. For T1ꢁ(H), variable contact time experiments
were performed by arraying 18 values: 0.1, 0.2, 0.4, 0.6, 0.8,
1, 1.2, 1.5, 2, 3, 4, 5, 6, 7, 8, 10, 12 and 15 ms. Recycle delays of
50 s were used for the ˇ form, but 200 s for the ˛ and ꢀ forms.
Probe temperatures were calibrated based on the
measurement of the frequency difference ꢂꢃꢄ between
resonances for the methyl and hydroxyl protons of
methanol in the 1H NMR spectrum of a sample of
tetrakis(trimethylsilyl)silane soaked in liquid ethanol.33 The
13C spectra were referenced to the signal for adamantane
(υC D 38.4 ppm from the TMS resonance for the high-
frequency peak) by replacement. Detailed peak assignments
were made using dipolar-dephased spectra,34 which reveal
resonances from quaternary carbons only.
from both x-ray and neutron diffraction (ND) data. The
computing facilities used were mainly those of the High
Performance Computing Service at Durham University. This
particularly involved a computer cluster by the name of
hal, which consists of eight Sun Ultra 80 ‘nodes’, each
having four 450 Hz Ultra Sparc II processors with 2 Gbyte
memory. The computed shielding values were converted to
the chemical shift scale by subtracting the computed isotropic
shielding constants of TMS (tetramethylsilane) (186.4 ppm)41
for carbon and of nitromethane ꢂꢀ135.8 ppmꢄ42 for nitrogen.
RESULTS AND DISCUSSION
In this work, we consider the crystal structures of the poly-
morphs in order to explore the relationship between the
solid-state effects seen in the 13C and 15N spectra and struc-
turalvariations. The main differences that are seen in the crys-
tal structures (M. J. Hursthouse and T. L. Threlfall, unpub-
lished results; A. Portieri, R. K. Harris and H. Puschmann,
unpublished results) are associated with the dihedral angle,
C(3)—C(4)—S(1)—N(2), as shown in Table 1. The signs in
this table are defined as positive when the angle is clock-
wise as seen looking down at sulfanilamide from the amino
group.
Proton relaxation times T1(H) and T1ꢁ(H) (in the rotating
frame) for the three forms were measured by proton magne-
tization inversion–recovery and spin-locking respectively,
followed by cross-polarization to carbon. These relaxation
times proved to be very different at ambient probe tem-
perature for the three polymorphs (see Table 2), indicating
differences in molecular-level mobility.
Nitrogen-15 CP/MAS spectra at natural abundance
were recorded using a Varian Unity Plus 300 spectrometer
operating at 30.399 MHz. Spectra of the enriched samples
were also recorded on the CMX 200 spectrometer and were
obtained under approximately the same conditions as for
13
the C measurements, but the 90 1H pulse duration was
°
7 µs in this case, with typical contact times of 3 ms, and
the referencing was obtained from a replacement sample of
15NH4NO3, with the nitrogen resonance from the nitrate
group taken as υN D ꢀ5.1 ppm (relative to the signal
for nitromethane). Approximately 250 mg of sample were
packed into the 5 mm outer diameter zirconia rotors (Kel-F
caps) for the probe used.
The convention used herein for shielding parameters
is that of Haeberlen.35 The three tensor components are
designated
Extensive variable-temperature measurements of T1(H)
and T1ꢁ(H) were carried out for the ˇ form (only). It is not
appropriate to discuss these in detail here, but suffice it to
state that they showed evidence of at least two motional
processes. Furthermore, the considerable variation in the
jꢅ33 ꢀ ꢅisoj ½ jꢅ11 ꢀ ꢅisoj ½ jꢅ22 ꢀ ꢅiso
Anisotropy D ꢅ33 ꢀ ꢅiso
Asymmetry Á D ꢂꢅ22 ꢀ ꢅ11ꢄ/
j
°
Table 1. Dihedral angles ( ) for the various forms of
sulfanilamide (M. J. Hursthouse and T. L. Threlfall,
unpublished results; A. Portieri, R. K. Harris and
H. Puschmann, unpublished results)
Isotropic shielding ꢅiso D ꢂꢅ11 C ꢅ22 C ꢅ33ꢄ/3
Dihedral angle
˛
ˇ
ꢀ
υ
The sign convention means that the chemical shift, υ, is
C(3)—C(4)—S(1)—N(2)
C(5)—C(4)—S(1)—O(1)
C(3)—C(4)—S(1)—O(2)
124.2
54.8
8.2
108.2
42.4
ꢀ8.8
91.4
29.4
ꢀ24.2
97.1
30.7
ꢀ18.2
ꢅref ꢀ ꢅiso
.
Spinning sideband analysis was carried out using both an
in-house computer program,36 ssb97, and STARS, supplied
by Varian Associates, each based on the theory described by
Maricq and Waugh.37 The STARS program was also used to
fit spectra of static samples.
Table 2. Proton relaxation times, obtained via
cross-polarization to 13Ca, for the three forms of
sulfanilamide (at ca 298 K)
For NMR shielding constant predictions, the Gaussian
94 program38 was used for both HF and DFT methods,
with basis sets such as 6–31G** or 6–311G** for carbon and
D95** for nitrogen, all implemented with gauge-including
atomic orbitals (GIAO). The calculations were taken to a
high level of theory, namely triple zeta basis sets (6–311G**)
and DFT methods such as B3PW91, which is a Becke39
three-parameter functional with a Perdew–Wang gradient-
correlated correction.40 The geometries used were taken
Polymorph
T1(H) (s)
T1ꢁ(H) (ms)
˛
ˇ
ꢀ
94 š 4
8 š 1
7.2 š 0.3
7.1 š 0.1
3.0 š 0.1
82 š 5
a Values averaged over the different 13C signals are given and
the errors quoted reflect the variations.
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

316
A. Portieri et al.
observed values helps to explain the significant differences
between relaxation times for the different polymorphs at a
single (ambient) temperature.
C(3), which is bonded to C(2)] to the more shielded peak as
seen from Table 5.
There appear to be, consistently, differences in our
chemical shifts from those of Frydman et al.,27 possibly
arising in part from differing practices of referencing (but
also perhaps from different spectral quality). However, the
absolute values are not the major point of interest. The C(1)
signals are split into doublets because of residual (second-
order) dipolar coupling arising from the quadrupolar nature
of the directly bonded 14N nuclei. A discussion of these
splittings can be found in an earlier paper,43 but we report
in Table 3 only the weighted mean position of the signals
(which is the true chemical shift), whereas Frydman et al.27
quote the split peaks separately.
As can be seen, the major characteristic that distinguishes
the three forms in the 13C solid-state spectra is the splitting
seen for resonances C(2)/C(6), an observation that is not clear
in the paper of Frydman et al.27 The values we obtain are 2.2,
4.8 and 2.4 ppm for the ˛, ˇ and ꢀ forms, respectively. This
splitting has been one of our major subjects of interest and,
in this paper, we assign the origin to distortions of the amino
group, which, as will be seen from 15N relaxation times, is
rigidly bonded to the phenylene ring.
The ambient-temperature T1 measurements allowed our
experiments on 13C spectra to be optimized and revealed
to us that the first samples of the ˛ polymorph which we
examined were contaminated by the presence of the ˇ form.
It is clear that if relatively short recycle delays are used in the
cross-polarization experiment, then a small admixture of ˇ
in a predominantly ˛ sample will result in a spectrum with
the ˇ peaks significantly enhanced. Conversely, a substantial
impurity of ˛ in ˇ may go undetected. Indeed, we believe that
this problem was encountered in the work of Frydman et al.,27
so that their results for the ˛ form are probably suspect. Our
optimized 13C spectra of the three forms are illustrated in
Fig. 1. Assignment of the signals can be made unequivocally,
given the selection of quaternary carbon signals from dipolar
dephasing experiments together with the known substituent
effects on chemical shifts. The results (together with those
of Frydman et al.27) are given in Table 3. Further assignment
of the phenylene doublet C(2)/C(6) was feasible following
shieldingcalculations, astheseprovedtopredictthesplitting,
allowing us to assign C(2) [see Table 1 for the designation of
Whereas raising the temperature for forms ˛ and ˇ
resulted in only minor variations of chemical shifts, the
spectrum of the ꢀ form showed collapse of the C(2)/C(6)
°
C(3)/C(5)
doublet above about 50 C. Whereas this might arise from
accidental shift equivalence at high temperature, it is more
C(4)
C(2)/C(6)
°
likely to stem from 180 phenylene ring flips, as was
C(1)
concluded by Frydman et al., who estimated the activation
energy of this process to be ca 63 kJ molꢀ1. It is concluded
that the ꢀ polymorph exhibits more mobility than the other
two forms.
alpha
beta
Nitrogen-15 solid-state NMR spectra of the three forms
showed significant differences in the isotropic chemical shifts
only for N(2), as shown in Table 4. This shift seems to be
greatly influenced by the dihedral angle of the sulfonamide
group, since the smaller the angle the greater is the shielding.
The fact that for N(1) the chemical shift is very similar in
the solid and solution states is a sign of the small sensitivity
to possible hydrogen bonds for this particular site in the
molecule. This is confirmed from the crystal structure data,
especially for the ꢀ form, in which this nitrogen does not
participate in hydrogen bonding.
gamma
140
160
150
130
120
110
ppm
Figure 1. Carbon-13 CP/MAS spectra of the three forms of
sulfanilamide.
Table 3. Sulfanilamide 13C chemical shifts (ppm) and assignments
Polymorph
C(1)
153.7a
166.5, 157.6
153.4a
165.5, 156.0
151.0a
165.1, 156.0
152.7
C(4)
C(2) and C(6)
C(3) and C(5)
˛ (this work)
˛ (Ref. 27)b
ˇ (this work)
ˇ (Ref. 27)
ꢀ (this work)
ꢀ (Ref. 27)
Solution in DMSO (Ref. 32)
128.0
134.5
127.1
135.6
127.1
137.1
130.8
113.1, 115.3
118.8, 124.0
112.3, 117.1
118.3, 123.1
112.7, 115.1
120.0, 122.5
113.5
128.3
134.5
129.5
133.3
129.6
135.0
128.3
a The splitting magnitudes caused by residual dipolar coupling to 14N are 131, 116 and 101 Hz for forms ˛, ˇ
and ꢀ, respectively.
b Data for the ˛ form from this source are suspect (see text).
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

CP/MAS NMR sulfanilamide polymorphs 317
Table 4. Isotropic chemical shifts, shielding anisotropies and asymmetries for nitrogen-15
N(1) isotropic
N(2) isotropic
Polymorph
chemical shift (ppm)
chemical shift (ppm)
for N(1) (ppm)
Á for N(1)
for N(2) (ppm)
Á for N(2)
˛
ꢀ312.2
ꢀ312.1
ꢀ312.1
ꢀ312.4
ꢀ288.8
ꢀ284.1
ꢀ280.6
ꢀ284.0
š46
43
š45
1
0.8
1
70.0
71.2
67.3
0.4
0.4
0.3
ˇ
ꢀ
Solution (DMSO)a
a From Ref. 32.
One apparent oddity in the 15N spectra, which made work
with natural abundance samples difficult, is that all three
formsshowanintenseresonanceforN(2), whereasonlyweak
signals could be seen for N(1) at ambient probe temperature.
Enriching the samples showed substantial differences in
relaxation times for the two different nitrogen sites. For
N(1) very long spin–lattice relaxation times (measured for
15N by the cross-polarization method44) of 100 s for the ˇ
form and over 200 s for ˛ and ꢀ (values which were, for
obvious reasons, not very accurately measured) contrast
with the relatively short relaxation times for N(2): 2.6 s for
˛, 0.2 s for ˇ and 1.7 s for ꢀ. Unlike the proton relaxation
measurements, the nitrogen data can be assigned to local
motions around the two different sites. They point clearly
to significant differences in mobility in the region of the
two nitrogens. Although the different values of T1ꢂNꢄ
cannot explain the differences in signal heights for the
two 15N resonances directly, mobility effects can potentially
influence 15N linewidths and lineshapes via interplay with
MAS rates (and/or the decoupler power), perhaps giving
the observed effects. The 15N relaxation behaviour was
further investigated, although only for the ˇ form, through
variable-temperature relaxation-time measurements. These
experiments (Fig. 2) showed very different behaviour for the
two sites and explain the fact that at ambient temperature
T1 for N(1) is nearly three orders of magnitude longer than
that for N(2). In fact, the value for N(2) is very close to
a minimum for 294 K (suggesting an energy barrier of
ca 42 kJ molꢀ1), whereas for N(1) it increases steeply as
temperature decreases in that region, implying a significantly
higher energy barrier to local motion. The value of T1 for
N(1) becomes much lower [ca 1 s, comparable to the ambient
temperature value for N(2)] at higher temperatures, ca 380 K.
Values of T1ꢁ(15N) were also measured for all three
polymorphs at ambient probe temperature, using spin-lock
powers equivalent to 50 kHz. For N(1) the value for the ˛ and
ˇ forms was 0.64 ms, whereas for the ꢀ polymorph it was
1.5 ms. Values for N(2) were significantly longer, namely 6.5,
10.0 and 4.2 ms, respectively, again pointing to substantially
different mobility around the two sites.
0.50
0.45
0.40
0.35
0.30
0.25
0.20
3.0
3.1
3.2
3.3
3.4
3.5
3.6
1000/T(K)
100
80
60
40
20
0
2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4
1000/T(K)
Figure 2. Variable-temperature spin–lattice relaxation times
for 15N nuclei in the ˇ form: bottom, N(1); top, N(2).
for C(2)/C(6) for the ˇ form is greater than those for the
other forms could be because this form has a significantly
shorter C(1)—N(1) bond than in the other forms, showing
a greater double-bond character: values for this bond-length
(M. J. Hursthouse and T. L. Threlfall, unpublished results;
A. Portieri, R. K. Harris and H. Puschmann, unpublished
˚
˚
results) are 1.374 A for the ˛ form, 1.368 A for the ˇ form and
We interpret the mobility of only one part of the molecule
as the reason why we see a splitting in the 13C resonances
for the ortho carbons only on the side of the amino group
[C(2)/C(6)]. Mobility of the sulfonamide group might (at
least partially) average the influence of this group on the
carbons in the position ortho to it [i.e. C(3)/C(5)]. Small
distortions of the amino group, on the other hand, would
influence significantly the carbons C(2)/C(6) owing to its lack
of mobility. The reason why the magnitude of the splitting
˚
1.377 A for the ꢀ form.
For 15N, shielding anisotropies and asymmetries were
measured for both the nitrogens of the molecule. Spinning
sideband analysis was used for N(2), with the average values
listed in Table 4. However, for N(1) it is not feasible to obtain
an adequate number of resolved spinning sidebands for
accurate analysis since the anisotropies are relatively small,
so static spectra for N(1) were recorded and analysed instead,
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

318
A. Portieri et al.
the experimental data in Table 5. The shifts are clearly in
the correct order, although errors in individual values are
substantial in some cases.
Computations were also carried out for clusters of two,
three or four molecules, corresponding to the arrangements
in the crystal structures, in order to account for the effects of
intermolecular interactions, particularly hydrogen bonding.
˚
Relevant contacts between heavy atoms below 3.11 A were
taken into account. For example, Fig. 4 shows the tetramer
used for the ˛-form computations. However, intermolecular
effects were found to be relatively small. The most interesting
aspect of the calculated results is that they are able to
predict the splittings between carbons C(2)/C(6), which,
when intermolecular effects are included, reach values
near to the experimental data: 3.0 ppm for the ˛ form,
4.9 ppm for the ˇ form and 3.3 ppm for the ꢀ form, to
be compared with the experimental values of 2.2, 4.8 and
2.4 ppm, respectively. The results for the computations on
the hydrogen-bonded clusters are shown in Table 6. The
evidence of two distinct shielding constants for C(3)/C(5)
is present in the calculations but not in the spectra, and
this is presumably due to the fact that the sulfonamide
group is fairly mobile so that the shielding values are
averaged, as mentioned above. The best results were
obtained when neutron diffraction data21 were considered
(available for only the ˇ form), indicating that the most
important parameters required in order to obtain reasonably
accurate shielding constants are precise hydrogen positions
−200
−250
−300
ppm
−350
−400
Figure 3. Nitrogen spectrum of the ˇ form for N(2): bottom,
experimental; top, computer fitted. For the former, the spinning
speed was 587 š 3 Hz. The centreband is at ꢀ284.1 ppm.
with the results, which are of lower precision than those for
N(2), shown in Table 4. An example of experimental and
fitted spinning sideband manifolds for N(2) is illustrated
in Fig. 3. Although the values of the anisotropies and
asymmetries are different from one nitrogen to the other,
there do not seem to be major differences from one form
to another. The values for the asymmetries are near unity
for N(1), making the sign of the anisotropy difficult to
define.
Table 6. Computed isotropic chemical shifts for 13C (in ppm)
for hydrogen-bonded molecules, obtained using
B3PW/6-311GŁŁ basis functions
˛ form
ˇ form
ꢀ form
C(4)
C(3)
C(2)
C(1)
C(6)
C(5)
R.m.s.
132.9
129.6
109.6
153.2
112.6
126.2
2.9
133.8
132.0
112.5
153.4
117.4
130.8
3.2
133.5
131.8
108.9
149.5
112.2
128.7
3.8
Computational procedures
The results of the computations of the isotropic 13C chemical
shifts for the isolated molecules with geometries as given
in the crystal structures are listed and compared with
Table 5. Experimental and calculated isotropic 13C chemical shifts for the isolated molecules using the
B3PW91/6–311G** model and single-crystal geometries derived from x-ray diffraction (M. J. Hursthouse and
T. L. Threlfall, unpublished results; A. Portieri, R. K. Harris and H. Puschmann, unpublished results) and
neutron diffraction data for the ˇ form
˛ form
Model
ˇ form
Model
ꢀ form
Model
ˇ forma
Model
Atom
Expt.
Expt.
Expt.
Expt.
C(4)
C(3)
C(2)
C(1)
C(6)
C(5)
R.m.s. error
136.1
129.9
108.3
153.8
109.7
127.7
4.5
128.0
128.3
113.1
153.7
115.3
128.3
135.9
131.0
110.3
153.2
111.1
128.7
4.5
127.1
129.5
112.3
153.4
117.1
129.5
135.8
129.6
107.1
149.8
110.0
129.2
4.7
127.1
129.6
112.7
151.0
115.1
129.6
134.6
132.9
110.1
152.9
114.5
131.3
3.7
127.1
129.5
112.3
153.4
117.1
129.5
a Using geometry data from neutron diffraction results.21
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

CP/MAS NMR sulfanilamide polymorphs 319
molecule2
molecule3
molecule1
2.99Å
2.96Å
molecule4
3.07Å
Figure 4. Tetramer used for the shielding calculations of the ˛ form.
(see Table 5). Results using the neutron diffraction data
show that the errors are not due to limitations of the
basis sets but to imperfections of the positioning of the
hydrogens.
values are expected to be more reliable than the absolute
shieldings. Actually, the agreement between experimental
and computed data is reasonable. However, in each case
the experimental anisotropy is less than the computed
value, presumably because of molecular motion. It is,
though, surprising in this context that the experimental and
theoretical ratios of the anisotropies for the two nitrogens in
each form are similar.
Computations were also performed on 15N shielding
(Tables 7 and 8), and in this case the full tensor parameters
were derived for comparison with the experimental results.
Since 15N is more sensitive to intermolecular effects, it
proved to be important to include more than one molecule
in the calculation; see Fig. 4. Including hydrogen bonds
(two molecules per calculation, considering one hydrogen
bond at a time), the differences between computed and
observed 15N isotropic chemical shifts (Table 7) become
lower but they were still not completely satisfactory. For
15N the reference calculation is clearly inadequate, as has
been noted previously for theophylline.30 Since the principal
interest here lies in comparing data for the two nitrogens
and for the different polymorphs, the values have been
rebased by 37 ppm to facilitate such comparisons. In fact,
the shift differences between N(1) and N(2) are well
represented by the computations, whereas the variations
between polymorphs appear to be too small to be accounted
for by the theoretical results.
Table 8. Comparison of calculated and experimental 15
N
shielding anisotropies for the hydrogen-bonded molecules
Computed Computed Experimental Experimental
anisotropy asymmetry anisotropy asymmetry
˛(HB)—
Nꢂ1ꢄ3.07
Nꢂ2ꢄ2.99
Nꢂ2ꢄ2.96
ˇ (HB)—
Nꢂ1ꢄ3.03
Nꢂ2ꢄ3.10
Nꢂ2ꢄ3.00
ꢀ(HB)—
Nꢂ2ꢄ3.00
Nꢂ2ꢄ3.01
ꢀ65
83.9
86.1
0.61
0.48
0.47
š46
70.0
1
˚
A
0.4
0.4
˚
A
70.0
˚
A
ꢀ51
83.2
86.9
0.8
43
0.8
0.4
0.4
˚
A
0.47
0.44
71.2
71.2
˚
A
˚
A
Results from the calculations of the full shielding tensors
are shown in Table 8, where two molecules are considered
for each computation. Since anisotropies and asymmetries
are differences involving a single molecule, the computed
83.2
82.1
0.50
0.51
67.3
67.3
0.3
0.3
˚
A
˚
A
Table 7. Experimental and calculated isotropic chemical shifts (in ppm) for N(1) and N(2)
Referenced
(υ) (ꢀ135.8 ppm)
Rebased by
37 ppm
Nitrogena
B3LYP/D95** (ꢅ)
Experimental (υ)
Differenceb
˛(HB)
ˇ(HB)
ꢀ(HB)
Nꢂ1ꢄ3.07
Nꢂ2ꢄ2.99
Nꢂ2ꢄ2.96
212.7
188.3
189.8
ꢀ348.5
ꢀ324.1
ꢀ325.6
ꢀ349.0
ꢀ323.9
ꢀ327.2
ꢀ311.5
ꢀ287.1
ꢀ288.6
ꢀ312.0
ꢀ286.9
ꢀ290.2
ꢀ312.2
ꢀ288.8
ꢀ288.8
ꢀ312.1
ꢀ284.1
ꢀ284.1
0.7
1.7
0.2
˚
A
˚
A
˚
A
Nꢂ1ꢄ3.03
Nꢂ2ꢄ3.00
Nꢂ2ꢄ3.10
213.2
188.1
191.4
0.1
ꢀ2.8
ꢀ6.1
ꢀ0.6
ꢀ0.7
˚
A
˚
A
˚
A
Nꢂ2ꢄ3.00
Nꢂ2ꢄ3.01
182.4
182.5
ꢀ318.2
ꢀ318.3
ꢀ281.2
ꢀ281.3
ꢀ280.6
ꢀ280.6
˚
A
˚
A
a The subscript is the H-bonded distance used for the computation.
b Computed minus experimental.
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320

320
A. Portieri et al.
16. Apperley DC, Fletton RA, Harris RK, Lancaster RW, Taverner S,
Threlfall TL. J. Pharm. Sci. 1999; 88: 1275.
CONCLUSIONS
We have shown here that the ˛, ˇ and ꢀ forms of
sulfanilamide may be clearly distinguished by their 13C
NMR spectra, but only if detailed attention is paid to
the experimental conditions. This is especially important if
mixtures of polymorphs are to be quantified. Moreover, it is
important to combine data from 13C and 15N spectra in order
to understand the mobility of the whole system. Further
light has been thrown on the molecular-level behaviour of
the polymorphs by relaxation time measurements and, in
the case of 15N, by analyses of spinning sideband manifolds
and of static spectra, yielding shielding tensor information.
Computations of the shielding parameters using both
isolated molecule and molecular cluster approaches have
brought additional understanding of the differences between
polymorphs in the splittings of the ortho-carbon resonances
and in the ratio of 15N shielding anisotropies.
17. Bingham A, Hughes D, Hursthouse MB, Lancaster RW,
Taverner S, Threlfall, TL. Chem. Commun. 2001; 601.
18. O’Connor BH, Maslen EN. Acta Crystallogr. 1965; 18: 363.
19. Alle´aume M, Decap J. Acta Crystallogr. 1965; 18: 731.
20. Alle´aume M, Decap J. Acta Crystallogr. 1965; 19: 934.
21. O’Connell AM, Maslen EN. Acta Crystallogr., 1967; 22: 134.
22. McLachlan D. X-ray Crystal structure. McGraw-Hill: New York,
1957.
23. Alle´aume M, Decap J. Acta Crystallogr. 1968; 24: 214.
24. Toscani S, Thoren S, Agafonov V, Ceolin R, Dugue J. Pharm. Res.
1995; 12: 1453.
25. Toscani S, Dzyabchenko A, Agafonov V, Dugue J, Ceolin R.
Pharm. Res. 1996; 13: 151.
26. Bernstein J, Davey RJ, Henck J-O. Angew. Chem. Int. Ed. Engl.
1999; 38: 3440.
27. Frydman L, Olivieri AC, Diaz LE, Frydman B, Schmidt A,
Vega S. Mol. Phys. 1990; 70: 563.
28. McGeorge G, Harris RK, Chippendale AM, Bullock JF. J. Chem.
Soc., Perkin Trans. 2 1996; 1733.
29. McGeorge G, Harris RK, Batsanov AS, Churakov AV, Chippen-
dale AM, Bullock JF, Gan Z. J. Phys. Chem. A 1998; 102: 3505.
30. Smith EDL, Hammond RB, Jones MT, Roberts KJ, Mitchell JBO,
Price SL, Harris RK, Apperley DC, Cherryman JC, Docherty R.
J. Phys. Chem. B 2001; 105: 5818.
Acknowledgement
We thank the UK Engineering and Physical Sciences Research
Council for support under grant GR/N05635. Robin K. Harris is
grateful to the Leverhulme Trust for the award of an Emeritus
Fellowship.
31. Ott DG. Synthesis with Stable Isotopes of Carbon, Nitrogen and
Oxygen. Wiley; New York, 1981.
32. Pomerantz M, Chou W, Witczak MK, Smith CG. J. Org. Chem.
1987; 52: 159.
REFERENCES
33. Aliev AE, Harris KDM. Magn. Reson. Chem. 1994; 32: 366.
34. Opella SJ, Frey MH. J. Am. Chem. Soc. 1979; 101: 5854.
35. Haeberlen U. Adv. Magn. Reson. 1976; Suppl. 1: 9.
36. Ascenso JR, Merwin LH, Bai H-P, Cherryman JC, Harris RK.
unpublished (originally mentioned in the open literature in
Harris RK, Sebald A. J. Organomet. Chem. 1989; 85: 1409).
37. Maricq MM, Waugh JS. J. Chem. Phys. 1979; 70: 3300.
38. Frish MJ, Trucks GW, Schelegel HB, Scuseria GE, Robb MA,
Cheeseman JR, Zakrzewsi VG, Montgomery JA, Stratman RE,
Burant JC, Dapprich S, Millan JM, Daniels AD, Kudin KN,
Strain MC, Farkas O, Tomasi J, Barone V, Cossi M, Cammi R,
Mennucci B, Pomelli C, Adamo C, Clifford S, Ochtersky J,
Petersson GA, Ayala PY, Cui Q, Morokuma K, Malick DK, Rad-
buck AD, Raghavachari K, Foresman JB, Cioslowsi J, Ortiz JV,
Stefanov BB, Liu G, Liashenko A, Piskorz P, Komaromi I,
Gomperts R, Martin RL, Fox DJ, Fox TA, Keith TA, Al-
Laham MA, Peng CY, Nanayakkara A, Gonzalez C, Chall-
combe M, Gill PMW, Johnson BG, Chen W, Wong MW,
Andres JL, Head-Gordon M, Replogle ES, Pople JA. Gaussian 94,
Revision C.2. Gaussian: Pittsburgh, PA, 1995.
1. Threlfall TL. Analyst 1995; 120: 2435.
2. Caira MR. Top. Curr. Chem. 1998; 198: 163.
3. Yu L, Reutzel S, Stephenson G. Pharm. Sci. Technol. Today 1999;
1: 118.
4. Bernstein J. Polymorphism in Molecular Crystals, Oxford
University Press: Oxford, 2002.
5. Chemburkar SR, Bauer J, Deming K, Spiwek H, Patel K,
Morris J, Henry R, Spanton S, Dziki W, Porter W, Quick J,
Bauer P, Donaubauer J, Narayanan RA, Soldani M, Riley D,
McFarland K. Org. Proced. Res. Dev. 2000; 4: 413.
6. Gavezotti A. Acc. Chem. Res. 1994; 27: 309.
7. Threlfall T. In The Encyclopedia of Vibrational Spectroscopy,
Griffiths P, Chalmers J (eds). Wiley: Chichester, 2001; 5:
3557–3573.
8. Szelagiewicz M, Marcolli C, Cianferani S, Hard AP, Vit A,
Burkhard A, von Raumer M, Hofmeier U Ch, Zilian A,
Francotte E, Schenker J. Journal of Thermal Analysis and
Calorimetry 1999; 57: 23.
9. Bernstein J. In Industrial Applications of X-ray Diffraction,
Chung FH, Smith DK (eds). Marcel Dekker: New York, 2000;
21.
39. Becke AD. J. Chem. Phys. 1993; 98: 5648; Becke AD. J. Chem. Phys.
1996; 104: 1040.
10. Giron D. Thermochim. Acta 1995; 248: 1.
11. Christopher EA. PhD Thesis, University of Durham, 1993.
12. Bugay D. Pharm. Res. 1993; 10: 317.
40. Perdew JD, Burke K, Wang Y. Phys. Rev. B 1996; 54: 16 533.
41. Jameson AK, Jameson CJ. Chem. Phys. Lett. 1987; 134: 461.
42. Jameson CJ, Mason J. Multinuclear NMR. Plenum Press: New
York, 1987.
13. Burger A. Sci. Pharm. 1973; 41: 290.
14. Sekiguchi K, Tsuda Y, Kanke M. Chem. Pharm. Bull. 1975; 23:
1353.
43. Harris RK, Olivieri AC. Prog. Nucl. Magn. Reson. Spectrosc. 1992;
24: 435.
15. Lin HO, Baenziger NC, Guillory JK. J. Pharm Sci. 1974; 63: 145.
44. Torchia DA. J. Magn. Reson. 1978; 30: 613.
Copyright  2004 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2004; 42: 313–320