Investigation of an N?H hydrogen bond in a solid benzoxazine dimer by 1H-15N NMR correlation techniques under fast magic-angle spinning

Article abstract of DOI:10.1002/mrc.931

The N?H distance within the unusual hydrogen-bonding arrangement adopted by a pair of methyl-substituted benzoxazine dimers (C₆H₃(OH)₂CH₂)₂N(CH ₃) has been determined by solid-state NMR to be 194±5 pm. This indicates that the proton is shared between the nitrogen and oxygen atoms, with a preference for an O-H rather than an N-H bond character. It is to be noted that a previous X-ray single crystal study was unable to localize the position of this hydrogen-bonded proton. The advanced solid-state NMR methods employed utilize REDOR-type recoupling under fast magic-angle spinning to recouple the heteronuclear ¹H-¹⁵N dipole-dipole interaction, such that rotor-encoded spinning-sideband patterns are obtained, the analysis of which yields the ¹H-¹⁵N dipole-dipole coupling and hence the N?H distance. Different designs of recoupling pulse sequences are discussed, which allow the experiment to be adapted to the system under investigation in terms of the required ¹⁵N or ¹H chemical shift resolution, conventional (¹⁵N) or inverse (¹H) detection as well as the importance of the perturbing influences of further spins. The chosen recoupling scheme employs inverse, i.e. ¹H, detection, because it provides a dramatic increase in signal sensitivity, resulting in savings in measurement time by a factor of at least 20, as well as ¹H chemical-shift resolution in the directly detected spectral dimension. This is the method of choice for cases such as this, where chemical shift resolution is not required in the ¹⁵N dimension. In addition, the perturbing effect of further protons on the N?H coupling of interest is minimized, such that a relatively long N?H distance can be determined despite the presence of several other couplings of comparable strength. Copyright

Full text of DOI:10.1002/mrc.931

MAGNETIC RESONANCE IN CHEMISTRY
Magn. Reson. Chem. 2001; 39: S5–S17
· · ·
Investigation of an N H hydrogen bond in a solid
benzoxazine dimer by 1H–15N NMR correlation
techniques under fast magic-angle spinning
Gillian R. Goward,1 Ingo Schnell,1 Steven P. Brown,1† Hans Wolfgang Spiess,1∗
Ho-Dong Kim2 and Hatsuo Ishida2
1
Max-Planck-Institut fu¨ r Polymerforschung, Postfach 3148, D-55021 Mainz, Germany
Department of Macromolecular Science and Engineering, Case Western Reserve University, Cleveland, OH 44106, USA
2
Received 3 May 2001; Revised 18 May 2001; Accepted 1 June 2001
· · ·
The N H distance within the unusual hydrogen-bonding arrangement adopted by a pair of methyl-
substituted benzoxazine dimers (C6H3(OH)2CH2)2N(CH3) has been determined by solid-state NMR to
be 194 5 pm. This indicates that the proton is shared between the nitrogen and oxygen atoms, with a
preference for an O–H rather than an N–H bond character. It is to be noted that a previous X-ray single
crystal study was unable to localize the position of this hydrogen-bonded proton. The advanced solid-state
NMR methods employed utilize REDOR-type recoupling under fast magic-angle spinning to recouple the
heteronuclear 1H–15N dipole–dipole interaction, such that rotor-encoded spinning-sideband patterns are
obtained, the analysis of which yields the 1H–15N dipole–dipole coupling and hence the N H distance.
· · ·
Different designs of recoupling pulse sequences are discussed, which allow the experiment to be adapted
to the system under investigation in terms of the required 15N or 1H chemical shift resolution, conventional
(15N) or inverse (1H ) detection as well as the importance of the perturbing influences of further spins.
The chosen recoupling scheme employs inverse, i.e. 1H , detection, because it provides a dramatic increase
1
in signal sensitivity, resulting in savings in measurement time by a factor of at least 20, as well as H
chemical-shift resolution in the directly detected spectral dimension. This is the method of choice for
cases such as this, where chemical shift resolution is not required in the 15N dimension. In addition, the
· · ·
perturbing effect of further protons on the N H coupling of interest is minimized, such that a relatively
· · ·
long N H distance can be determined despite the presence of several other couplings of comparable
strength. Copyright  2001 John Wiley & Sons, Ltd.
determine the length of a comparatively long hydrogen bond
uniquely.
INTRODUCTION
Investigations into hydrogen-bonded structures, including
self-organized and biologically relevant molecules, are being
pursued in many disciplines of science.1,2 The motivating
factor behind these studies is the structure stabilizing, or
even structure-directing role played by hydrogen bonds,
which is as yet not fully understood, and represents one of
the current frontiers of scientific endeavour. The question
addressed in this paper concerns the characterization of an
Ð Ð Ð Ð Ð Ð
1H solid-state NMR has developed into a versatile tool for
the investigation of hydrogen-bonding structures in organic
solids, and also in peptides.3–5 Compared with alternative
1
methods used to study hydrogen bonds, H NMR has the
unique advantage of directly detecting the nuclei forming
the hydrogen bonds via the 1H chemical shift. The degree to
which a 1H resonance is shifted to higher frequency has been
correlated to the strength of the hydrogen bonding in which
it is involved. This correlation has been established for a
Ð Ð Ð Ð Ð Ð
unusual N
H
O hydrogen bond, through assessment of
15
1
Ð Ð Ð
N
the
H bond length, using advanced solid-state NMR
range of O
H
O hydrogen bonds, e.g. in carboxylic acids,
methods. Whereas other studies focus on the measurement
1
Ð Ð Ð
through comparison of the H chemical shift to the O
O
Ð Ð Ð
of short N H bonding interactions in the region of 110 pm,
distances,6 or O H distances, determined by X-ray and
neutron diffraction respectively. Another study of the hydro-
gen bonding protons in carboxylic acids using the combined
rotation and multiple pulse sequence (CRAMPS) to achieve
better 1H chemical shift resolution was presented by Harris
et al.8 Consistently, the 1H resonance frequency increased as
the distance between the heteroatoms decreased. Recently,
a study of amide protons in hydrogen-bonded structures of
glycine-containing peptides and poly-peptides in the solid
7
Ð Ð Ð
our study demonstrates the ability of magic-angle spinning
(MAS) and heteronuclear dipolar recoupling techniques to
ŁCorrespondence to: H. W. Spiess, Max-Planck-Institut fu¨r
Polymerforschung, Postfach 3148, D-55021 Mainz, Germany.
E-mail: spiess@mpip-mainz.mpg.de
†Present address: Laboratoire STIM, UMR 5532 CNRS/ENS, Ecole
Normal Superieure de Lyon, 69364 Lyon, France.
Contract/grant sponsor: DFG.
DOI: 10.1002/mrc.931
Copyright  2001 John Wiley & Sons, Ltd.

S6 G. R. Goward et al.
state correlated the N (H) O distances with 1H chemi-
cal shifts obtained at a high magnetic field and under fast
MAS.9 A trend towards high-frequency shifts for shorter
using fast MAS and 1H–1H DQ spectroscopy.13 Benzoxazine
dimers serve as model compounds for the polybenzoxazines,
a new class of phenolic resins, which have been investigated
extensively in the past 10 years.25–27 Studies have focused
both on the synthesis of new members of this family
of compounds and on the characterization of the unique
properties of these materials. This substantial research
effort has been motivated by the observation that the
polybenzoxazines exhibit a number of unusual properties,
such as low volumetric shrinkage or expansion upon
curing,28 lack of water absorption and excellent resistance
to chemicals and UV light,29,30 as well as surprisingly
high Tg values given the low crosslinking density,31 all of
which make them attractive candidates for many commercial
applications. The key structural feature that is presumed to be
responsible for the remarkable behaviour of these polymers
Ð Ð Ð
Ð Ð Ð
Ð Ð Ð
N
Ð Ð Ð
(H) O distances was observed and confirmed by ab ini-
tio calculations, although there was significant scatter in the
experimental data. However, this latter study did not attempt
to place the proton between the two heteroatoms, and thus
the hydrogen-bonding structures were not fully determined.
Another approach to investigate hydrogen bonds involves
Ð Ð Ð Ð Ð Ð
studies of dynamics in 15
N
(L) 15N low-barrier hydrogen
bonds, where L represents either hydrogen or deuterium,
15
Ð Ð Ð
which have examined the two
tances using solid-state 15N NMR and quantum chemical
calculations.10
N
H interdependent dis-
In solid-state NMR the 1H chemical shift informa-
tion is usually obscured by substantial line broadening
effects caused by strong homonuclear 1H–1H dipole–dipole
couplings. Various approaches to the problem of gain-
Ð Ð Ð Ð Ð Ð
is the formation of N
Ð Ð Ð Ð Ð Ð
H
O and O
H
O hydrogen
bonds.32 A better understanding of these hydrogen-bonding
structures is therefore of much importance.
1
ing sufficient H spectral resolution have been presented,4
e.g. the application of multiple-pulse sequences, such as
WAHUHA,11 to single-crystal samples, as well as the com-
bination of such multiple-pulse sequences with MAS in
the CRAMPS approach.12 Recently, significant technological
and methodological progress in 1H NMR has been afforded
by extending MAS to spinning frequencies in excess of
30 kHz. The improvement in the 1H resolution of hydrogen-
bonded protons afforded by MAS at such a fast rotation
frequency is illustrated by Yamauchi et al.,9 Schnell et al.,13
and Brown et al.14 It is to be noted that significantly narrower
1H linewidths can be obtained in experiments involving
the application of ‘windowless’ homonuclear decoupling
sequences, such as the Lee–Goldburg (LG) technique15 and
refinements, namely the frequency-switched and phase-
modulated LG (FSLG16,17 and PMLG18) sequences, as well
as the computer-optimized sequence DUMBO-1.19 Recently,
Lesage et al. have shown that a linewidth as low as 60 Hz can
be obtained for the aliphatic resonances in L-alanine using a
constant-time (CT) CRAMPS experiment.20
The 1H–1H DQ studies of the benzoxazine dimers
resulted in a proposal of two unique intermolecular
hydrogen-bonding arrangements: one in which the dimers
arrange pairwise, and a second in which a twist of half
of the dimer unit about the central nitrogen atom allows
for a ladder-like extended structure.13 Whereas the former
structure was first determined by X-ray crystallography,33
the latter was originally proposed exclusively on the basis
of the 1H–1H DQ NMR spectra, and has subsequently been
verified, in the case of the propyl dimer by a single-crystal
X-ray diffraction study.34 This illustrates the power of the
NMR methods to provide accurate structural information
in the absence of crystallographic data. Beyond the deter-
mination of structural elements that can also be accessed
through diffraction methods, NMR offers the ability to locate
protons accurately. In general, proton positions are difficult
to determine via diffraction methods, where this can only be
accomplished successfully through Fourier difference map-
ping in X-ray scattering studies of small organic molecules,
or with the assistance of a nuclear reactor and deuterium
labelling in neutron scattering.35,36 By comparison, solid-
state NMR is routinely available, offers superior sensitivity,
and specific isotope enrichment is required only in special
cases.
In addition to achieving high resolution, and hence
1H chemical shift information, it is desirable to access the
1
structural and dynamic information inherent to the H–1H
dipole–dipole coupling. One approach by which this can be
achieved is the combination of MAS with 1H–1H double-
1
quantum (DQ) spectroscopy.3,4,21,22 Using the H DQ MAS
Here, we present such a special case, in which we make
use of 15N isotopic enrichment, and extend our previous
study of hydrogen-bonded benzoxazine dimers. Heteronu-
clear 1H–15N approaches, based on REDOR-type37,38 dipolar
method, it has been shown that valuable insight into the
hydrogen-bonded structures adopted by different alkyl-
substituted benzoxazine dimers,13 as well as the change
in the quadrupole hydrogen-bonding arrangement formed
by dimers of 2-ureido-4-pyrimidone units upon a keto-
enol tautomerism,23 could be obtained. In addition, it has
recently been demonstrated that an analysis of 1H DQ MAS
sideband patterns allows the quantitative determination of
proton–proton distances for the complex hydrogen-bonded
arrangement adopted by the biologically important molecule
bilirubin.14 Furthermore, the kinetics of the making and
breaking of hydrogen bonds could be followed by analysing
the loss of signal in a DQ-filtered experiment.24
Ð Ð Ð
recoupling methods, were used to determine the N H dis-
tance in the dimers. Relying on fast MAS (ꢀR typically 30 kHz)
for dipolar decoupling and 1H resolution enhancement,
dipolar recoupling is used to generate heteronuclear single-
quantum correlation (HSQC) and to observe rotor-encoded
MAS sideband patterns, which serve as a sensitive mea-
sure for the underlying N–H dipole–dipole coupling.39–41
In addition, different variants of the pulse sequence are
discussed, which employ conventional (15N) or inverse (1H)
detection and suppress perturbing influences of further N–H
couplings to varying degrees. The sensitivity enhancement
As noted above, in a previous study, hydrogen-bonded
benzoxazine dimers have been investigated by 1H NMR
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N H hydrogen bond study by fast MAS S7
1H–15N dipole–dipole couplings involving the labelled
15N site need to be considered. Apart from the proton in
the hydrogen bond, there are four CH2 protons and the
and optimization of experimental time achievable by an
appropriate design of the pulse sequence are demonstrated.
Besides the heteronuclear correlation techniques pre-
Ð Ð Ð
sented here, CRAMPS can also be used to measure X
distances in the solid state through the analysis of dipolar
sideband patterns obtained at low spinning speeds. In a
H
Ð Ð Ð
three protons of the methyl group, whose N H distances
can be reliably estimated from the crystal structure to be
¾200 pm (CH2) and ¾190 pm (CH3, the N–H dipole–dipole
couplings are partially averaged, though, due to fast methyl
rotation). Hence, it is to be expected that the perturbing N–H
recent example, the N H distance in 15N labelled poly(l-
Ð Ð Ð
aniline) was measured using 1H CRAMPS NMR.42 Differing
bond lengths were determined for the ˛-helical morphology
(112 pm) compared with ˇ-sheet (109 pm), and attributed to
a stronger hydrogen bond in the latter case. These results
show that dipolar 1H–15N methods are capable of resolving
Ð Ð Ð
Ð Ð Ð
dipole–dipole couplings of N CH2 and N CH3, as well
Ð Ð Ð
as the N H coupling of interest, are of comparable strength.
Therefore, the distinction of the proton species (CH2, CH3
and hydrogen bond) interacting with the nitrogen is vital,
Ð Ð Ð
very fine differences in close N H distances, and in the
Ð Ð Ð
Ð Ð Ð
and perturbing effects of N CH2 and N CH3 couplings
Ð Ð Ð
following we show that longer N H distances can also be
Ð Ð Ð
on the N H coupling need to be minimized.
investigated, even in the presence of further protons, by tak-
ing advantage of the 1H resolution enhancement achievable
using very fast MAS.
1H fast MAS and 1H–1H DQ spectra
First we consider 1H spectra of the dimer obtained under
fast MAS. Whereas the 1H MAS spectrum provides chemical
shift information and, in this way, primary evidence for
the presence of hydrogen bonds, the 1H–1H DQ spectrum is
well suited for obtaining insight into the proximities between
proton species and for identifying the structure of multiple
hydrogen bonds.3,4 In 1H–1H DQ NMR spectroscopy, the
generation of DQ coherence (DQC) requires the existence of a
dipolar coupling between two protons. Thus, the observation
of particular DQ peaks implies the existence of a sufficient
dipolar coupling between the respective nuclei. In this way,
detailed information about internuclear proximities can be
readily obtained from the signals in two-dimensional (2D)
DQ spectra, which correlate a DQ spectral dimension (t1) with
a single-quantum (SQ) dimension, with the latter being used
for signal detection. In the DQ dimension, each observed
resonance frequency ωAB corresponds to the sum frequency
of the two nuclei, A and B, involved in the coherence:
ωAB D ωA C ωB. Hence, 2D DQ spectra show signals at the
positions (ωA, ωAB) and (ωB, ωABꢁ, and such signal patterns
straightforwardly allow the identification of internuclear
proximities.3,13
PRIMARY SAMPLE CHARACTERIZATION
Crystal structure
Figure 1 shows a schematic drawing of the chemical structure
of a benzoxazine dimer pair, together with a portion of the
crystal structure of the methyl benzoxazine dimer, with the
33
Ð Ð Ð Ð Ð Ð
N
H
O interaction of interest indicated with arrows.
Owing to different averaging effects of vibrational motions,
scattering methods tend to yield shorter bond lengths
than NMR methods based on dipole–dipole couplings.13
In Fig. 1b, the respective hydrogen atoms are placed on the
trajectory between the oxygen and nitrogen, but without
Ð Ð Ð Ð Ð Ð
a defined position. The position of these N
H
O and
Ð Ð Ð Ð Ð Ð
O
H
O protons is expected to depend significantly on
the character and the strength of the hydrogen bonds. It
Ð Ð Ð Ð Ð Ð
is relevant to mention that the N
H
O bond is not
expected to be linear, but likely is part of a distorted six-
membered ring. This intramolecular hydrogen-bonded ring
has been shown to be stable both in the solid at elevated
temperatures, and in solution using FT-IR studies.32,43
While investigating the hydrogen-bond N H distance in
The 1H MAS spectrum and the 2D 1H–1H DQ MAS spec-
trum of the dimer are shown in Fig. 2a and b, respectively.
Ð Ð Ð
the following, potentially perturbing influences of further
a)
b)
CH3
H
H3C
CH3
N
H
N
O
O
O
H3C
H3C
CH3
CH3
H
H
N
O
H
H
N
CH3
H3C
CH3
Ð Ð Ð Ð Ð Ð
Figure 1. (a) The molecular structure and (b) the crystal structure of a pair of methyl benzoxazine dimers. The N
hydrogen-bonding interaction is highlighted.
H
O
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

S8 G. R. Goward et al.
an AC cross peak, which was weakly evident in the pre-
E
1
vious spectrum, and arose from a H–1H contact between
D
a)
Ð Ð Ð Ð Ð Ð
(N
Ð Ð Ð Ð Ð Ð
H
O) and (O
H
O), attributed to the presence of
a second conformation. The absence of the AC cross peak is,
therefore, taken as evidence of the high purity achieved in
the synthesis of the labelled compound, and indicates a lower
degree of structural dispersion present in this sample. Based
on the previous study of the series of benzoxazine dimers,
including methyl-, ethyl- and propyl-substituted materials,
both the ethyl and propyl variants were found to exhibit a
ladder-like structure, achieved by twisting the dimer about
the central nitrogen atom and thus creating an extended
hydrogen bonding arrangement. The primary evidence for
this arrangement was the presence of a strong AC peak in
the 2D DQ spectra of these compounds.
C
B
A
12 10
8
6
4
2
0
−2
ppm
b)
−5
0
EE
CE
5
1H–13C recoupled-dipolar HSQC spectrum
10
15
20
25
CC
AD
1H–1H DQ spectra provide much information concerning
inter- and intra-molecular proximities, and the experiment
is also facile to implement and inexpensive in terms of
experimental time. However, important structural details
are only available through the heteronuclear correlation
methods, as described above; thus we extended our study to
AB
−2 −4
12 10
8
6
4
2
0
include a 1H–13C correlation spectrum. In Fig. 3, the 1H–13
C
ppm
correlation spectrum of the methyl benzoxazine dimer is
shown, which was obtained in a rotor-synchronized fashion
using recoupled polarization transfer (REPT) under MAS.39,40
The pulse sequence is described below (see the chemical shift
resolution in heteronuclear dipolar correlations section and
Fig. 6a). The 1H and 13C chemical shifts are observed in
the indirect (F1) and direct (F2) dimensions respectively. Of
interest is the observation that all the 13C resonances are split
except for that of the methyl group at 40.8 ppm, which is
bound to the nitrogen. The expanded aliphatic and aromatic
regions, depicted in Fig. 3b and c, clearly show the splitting
of the peaks in the 13C dimension, whereas the frequencies
of the 1H resonances essentially remain single peaks. A
comparison of the 13C resonances obtained by solution- and
solid-state NMR is given in Table 2.
Figure 2. (a) 1H MAS spectrum and (b) 2D rotor-synchronized
1H–1H DQ MAS spectrum of the 15N-labelled methyl
benzoxazine dimer acquired with 1/2ꢂR excitation time
(ꢀR D 30 kHz and 700 MHz 1H Larmor frequency).
Table 1. Assignment of proton chemical shift resonances in
single quantum and double quantum dimensions of the 2D DQ
spectrum (Fig. 2a), acquired with 1/2ꢂR excitation time
Double
Single quantum
chemical
Double
quantum
coherences
quantum
chemical
shifts/ppm
Proton
shift/ppm
N–H (A)
11.2
AB
AD
18.3
13.2
The splitting of the 13C resonances in the solid state
indicates a degree of asymmetry in the dimer, pivotal about
the central nitrogen, since the resonance of the attached
methyl group is not split. Increasing the temperature
incrementally in steps of 10 K, to a maximum of 380 K, did
O–H (B)
Aromatic H (C)
7.2
6.0
CC
CE
12.0
7.3
CH2 (D)
CH3 (E)
2.0
1.3
not cause the lines in the corresponding one-dimensional 13
C
EE
2.6
REPT spectra to coalesce (data not shown). The observed
splitting is consistent with the crystal structure,33 in which
a C2 symmetry element in the centre of the dimer pair
dictates that the two dimers of the pair are equivalent,
but the two halves of the dimer, i.e. the phenol units, are
crystallographically distinct.
A splitting of 1H resonances in the solid state has recently
been observed in several cases, for example L-tyrosine,40,44
hexabenzocoronenes,45 and molecular-tweezer host–guest
complexes.46 These splittings emerge due to intermolecular
interactions in the solid state, namely aromatic ꢃ-electron
effects on nearby protons. The electronic environment of the
proton is strongly influenced by its orientation with respect
to proximal aromatic units and their ring currents, which
1
The H SQ and DQ chemical shifts observed in the spectra
are listed in Table 1. Three SQ resonances and a shoulder
Ð Ð Ð Ð Ð Ð
Ð Ð Ð Ð Ð Ð
are evident, and are assigned to N
H
O (A), O
H
O
(B), aromatic protons (C), and CH3 protons (E). Based on the
resolution achieved in the DQ spectrum, a fifth resonance is
differentiated, and is attributed to CH2 protons (D). Both the
presence and absence of DQ peaks provide valuable infor-
mation for structure determination. A comparison between
this spectrum and that originally documented by Schnell
et al.13 confirms that the intermolecular hydrogen-bonded
structure described previously is also present in the labelled
sample. In fact, in this spectrum there is no evidence for
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N
H hydrogen bond study by fast MAS S9
b)
a)
−1.0
0.0
1.0
2.0
3.0
4.0
−2
0
2
23 21 19 17 15
ppm
c)
4
4.0
5.0
6.0
7.0
8.0
6
8
140
120
100
80
60
40
20
13C chemical shift [ppm]
134 132 130 128 126
ppm
Figure 3. (a) 1H–13C REPT-HSQC rotor-synchronized 2D correlation spectrum of the methyl benzoxazine dimer, together with F2
and F1 skyline projections, corresponding to the 13C and 1H dimensions respectively. On the right, insets corresponding to the
expanded (b) methyl and (c) aromatic CH regions are shown.
Table 2. Comparison of chemical shifts in solution- and
solid-state 13C spectra of 15N-labelled methyl benzoxazine
dimers
dipolar correlations under fast MAS. As revealed by the
1
1
Ð Ð Ð
crystal structure and the H– H DQ spectrum, the N
H
distance of interest is embedded among a multitude of
other N–H proximities. Though the 15N nucleus can easily
and selectively be accessed using isotopic enrichment, 1H
chemical shift resolution is required to distinguish the
different proton species and to select the hydrogen-bonded
Chemical shift Chemical shift
13C
centre
in
in the solid
state/ppm
solution/ppm
Ð Ð Ð
one. In addition, the N H coupling of interest needs
to be independently measured in the presence of several
further N–H dipolar interactions of comparable strength.
Thus, an approach is required that combines N–H coupling
measurement with 1H spectral resolution and which is as
insensitive as possible to perturbing N–H interactions.
Aromatic methyl CH3 (o-OH)
Aromatic methyl CH3 (p-OH)
N–CH3
15.8
20.4
41.1
16.1
20.1
16.6
21.2
40.8
CH2
59.3
59.7
63.2
Aromatic C–H (o-N)
Aromatic C–H (p-N)
128.2
131.3
127.5 129.9
130.9 132.3
Rotor-encoded MAS sideband patterns
Whereas MAS at 30 kHz provides homonuclear dipolar
decoupling and resolution enhancement among the protons,
heteronuclear dipolar recoupling is used for generating the
1H–15N correlations. Recoupling is accomplished by trains
of radio frequency (RF) ꢃ-pulses in a REDOR fashion37,38
applied during two periods of the pulse sequence (see
Figs 4 and 6). It is preferable to apply the ꢃ-pulses on
the channel, where there is no evolution of transverse
magnetization, although the application of a central ꢃ-pulse
on the other channel is necessary in order to compensate
for chemical shift or off-resonance effects. The recoupling
approach is a dedicated fast MAS technique with a well-
defined average Hamiltonian,44 which ensures coherent
polarization transfer. Cross polarization (CP) approaches,
in contrast, need to be adapted to fast MAS conditions50 and
tend to non-coherent transfer processes, an exception being
the recently proposed FSLG-CP approach.51 In this sense,
heteronuclear dipolar recoupling bears a direct analogy
give rise to the shifts of the 1H resonance frequencies. These
effects have been quantitatively investigated by means of
quantum chemical calculations.47–49 In the case of the methyl
1
benzoxazine dimer, no clear splitting of the H resonances
is observed, although the CH2 resonances are very broad,
spanning 4 ppm. The latter is most likely due to ring current
effects. For the CH3 and aromatic protons, the experimental
absence of a splitting of the 1H resonances is consistent
with the X-ray crystal structure, where significantly different
exposure of protons to ꢃ-electrons, or even ꢃ-stacking, is not
present.33
· · ·
METHODS FOR N H DIPOLAR DISTANCE
DETERMINATION
In this study, the heteronuclear dipole–dipole coupling
is exploited for N–H distance determination by means
of recoupling techniques, which generate heteronuclear
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

S10
G. R. Goward et al.
Excitation
Reconversion
Detection
t1
t2
I
I
I
I
or
or
or
S
S
S
S
Figure 4. Schematic representation of the possible structure of heteronuclear dipolar recoupling pulse sequences, where different
options for the pathways followed by the transverse coherences are indicated with arrows.
to the INEPT technique52 used in solution-state NMR,
with dipole–dipole and J-couplings serving as mediating
interactions respectively.
The heteronuclear correlations generated are based on the
dipole–dipole coupling between two spins, denoted I and
S. The average Hamiltonian describing the evolution during
the interval (t; t C NꢂR) of dipolar recoupling is given by40
where Nexc and Nrec denote the number of rotor periods
used for excitation and reconversion respectively, and the
brackets hÐ Ð Ði indicate that an average is taken over all
orientations for powdered samples. Fourier transformation
of this signal yields characteristic sideband patterns that, for
the ideal case of an isolated spin pair, consist of odd-order
sidebands only and can serve as a sensitive measure for
DIS, because the appearance of the pattern depends strongly
on the parameter Nexc/recDIS/ωR. It is to be noted that such
sideband patterns are observed for both homonuclear and
heteronuclear spin pairs.3,45,56 The larger the value of the
parameter, the broader the span of the pattern, and the more
accurate the determination of the coupling from the pattern.
With longer recoupling times (i.e. with increasing Nexc/rec),
however, perturbing effects of additional interactions and
relaxation come increasingly into play. Consequently, the
duration of the recoupling periods must be optimized.
Figure 5 shows the generalized build-up curve for a
dipolar IS two-spin correlation calculated as a function
Nt
ꢂR
O
O
2IzSz
O
HIS t; t C NꢂRꢁ D
1ꢁ
where t is the dipolar phase factor, which is given by
p
DIS
ωR
t D ꢀ
2
2 sin 2ˇ sin ωRt C ꢄꢁ
2ꢁ
The angles ˇ and ꢄ relate the principal axes system of the
dipole–dipole coupling tensor to the rotor-fixed frame, and
ωR is the rotor frequency. DIS is the dipolar coupling constant
in units of angular frequency, and depends on the internu-
clear distance rIS and the respective magnetogyric ratios of
the nuclei:
of DISꢂrcpl/2ꢃ (where ꢂrcpl D NrcplꢂR and Nrcpl D Nexc
D
ꢅ0 ꢄIꢄSh¯
4ꢃ
Nrec), together with the rotor-encoded sideband patterns
observed for three choices of DISꢂrcpl/2ꢃ. Since a short
excitation time favours the first-order sidebands, it does
not produce a sideband pattern from which a valid
dipolar-coupling constant can be determined. To generate
higher-order sidebands, data must be acquired beyond
DISꢂrcpl/2ꢃ D 0.8 in the build-up curve, but excessively
long excitation times eventually obliterate signal intensity
due to relaxation effects. Empirically, the relaxation typically
observed for dipolar two-spin correlation build-up curves
can be described by an exponential decay, as is also displayed
in Fig. 5. Such build-up curves and rotor-encoded sideband
patterns have been extensively studied and used for the
determinations of homonuclear and heteronuclear dipolar
interactions.3,45,56–59
DIS
D
Ð
3ꢁ
r3IS
The recoupled heteronuclear dipolar Hamiltonian given in
Eqn (1) implicitly depends on the rotor phase via the dipolar
phase [Eqn (2)]. Hence, by combining two such recoupling
periods with different phases, corresponding to different ini-
tial rotor orientations, the signal becomes rotor encoded.53–55
The difference in phases can easily be created by introducing
a ‘waiting time’ t1 between the two recoupling periods. This
leads to a 2D experiment consisting of four periods, which
will henceforth be referred to as the excitation, t1, reconver-
sion, and detection periods. A schematic representation of
the experiment and its variants is depicted in Fig. 4. The par-
ticular placement of the ꢃ/2-pulses allows the pathway of the
polarization to be selected such that either the I-spins or the
S-spins are the evolving species during each of the four peri-
ods. In order to subject the spins to a dipolar evolution under
the Hamiltonian given in Eqn (1), transverse I or S spin states
need to be created by applying an RF ꢃ/2-pulse before the
ꢃ-pulse train accordingly. After two such periods of dipolar
evolution, a signal can be detected (in a final detection period
t2), the amplitude of which is modulated in the form
Chemical shift resolution in heteronuclear dipolar
correlations
In the previous section, a sensitive and efficient way of
measuring dipole–dipole couplings under MAS conditions
has been described. In general, and in particular for the
benzoxazine sample investigated here, the distinction of
different couplings with respect to the spins involved is
of major importance. For this purpose, spectral dimensions
S / hsin Nexc0ꢁ sin Nrect1ꢁi
4ꢁ
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N H hydrogen bond study by fast MAS S11
(ii)
(i)
(iii)
0.5
0.4
0.3
0.2
0.1
0
(i)
(ii)
(iii)
−5 −3 −1
1
3
5
0
1
2
3
ω/ωR
DISτrcpl / 2π
Figure 5. The build-up of intensity in a heteronuclear experiment of the type illustrated in Fig. 4 as a function of the unitless
parameter, DISꢂrcpl/2ꢃ, for the ideal case of an isolated IS spin pair. Exponential damping, which mimics relaxation effects, is
included for decay time constants of 2 ms (dashed line) and 0.5 ms (dotted line). Simulated frequency-domain sideband patterns for
the indicated DISꢂrcpl/2ꢃ values (i)–(iii) are shown to the right.
Preparation
x
Excitation
t1
Reconversion
Z-Filter Detection
y
I
IxSz
DD
a)
b)
c)
x
x
x
S
y
y
x
I
CP
x
IzSx
S
CP
y
y
y
y
x
x
x
I
IzSx
IxSz
S
y
y
y
y
I
DD
CP
x
x
x
d)
S
CP
Figure 6. A comparison of REPT and DIP pulse sequences: (a) directly detected REPT-HSQC, (b) inversely detected REPT-HSQC,
(c) inversely detected DIP-HSQC, and (d) directly detected DIP-HSQC.
interaction
providing chemical shift resolution of the I and S spins
can straightforwardly be introduced into the t1 and t2
periods of the basic scheme for generating rotor-encoded
sideband patterns (see the pulse sequences depicted in
Fig. 6). The product operator formalism will be used to
describe the relevant spin states during their time-dependent
evolution.60 Starting the excitation period with a transverse
I state (generated by an initial ꢃ/2-pulse on I with
phase x), the system evolves under the recoupled dipolar
Nexc12IzSx
ꢀIy ꢀꢀꢀꢀꢀꢀꢀ! ꢀIy cos Nexc0 C IxSz sin Nexc0
5ꢁ
In this way, an antiphase correlation, IxSz, is created. When
no ꢃ/2-pulse is applied at the end of the excitation (as in
Fig. 6a), the transverse Ix component evolves under the I-spin
chemical shift during the subsequent t1 period as follows (the
residual Iy magnetization is removed by phase cycling, and
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

S12
G. R. Goward et al.
Regarding the 15N-labelled benzoxazine sample investi-
gated here, it is important to note that spectral resolution
of resonances in the H dimension is vital in order to dis-
will not be considered further):
ωCI St1Ix
1
2IxSz sin Nexc0 ꢀꢀꢀꢀ! 2IxSz sin Nexc0 cos ωCI St1
tinguish the different proton species present in the sample
and to identify the N–H coupling of interest. In the 15N
dimension, in contrast, only a single line will be observed
due to the selective isotopic enrichment and the presence
of a single nitrogen site in the crystal structure.33 Generally,
to optimize the use of measuring time for 2D data sets, it
is advisable to record data via the nucleus making higher
demands on resolution and spectral width in the direct
dimension.63 Here, the NH correlation experiment should
therefore be designed in the inverse fashion, such that the
single 15N signal is recorded in the indirect (t1) dimension,
whereas the 1 H chemical shifts are resolved in the direct (t2)
dimension. Furthermore, under on-resonance conditions for
15N, procedures ensuring phase-sensitive detection can be
omitted. Thus, major advantages in terms of signal sensitiv-
ity, as well as minimum experiment times, are expected for
inverse, i.e. 1H, detection.
C 2IySz sin Nexc0 sin ωCI St1
6ꢁ
Following the t1 evolution period, simultaneous ꢃ/2-pulses
are applied to the I and S spins, interchanging transverse
and longitudinal I and S components and, in this way,
preparing the polarization transfer from I to S. By alternating
the phase of the ꢃ/2-pulse on the I-spins in subsequent
experiments, both cosine and sine components of the I-spin
t1 signal are acquired, allowing for sign discrimination in the
indirect dimension. Considering only the cosine component
in Eqn (6), the IzSx antiphase state evolves to Sy during the
reconversion period, which is finally detected, following a
z-filter.
N
rect 2IxSz
1
2IzSx sin Nexc0 cos ωCI St1 ꢀꢀꢀꢀꢀꢀꢀ! Sy sin Nexc0
ð sin Nrect cos ωC1 St1
7ꢁ
1
Multi-spin effects
Another important point besides the distinction of reso-
nances in the spectra is the perturbing effect that other
dipolar interactions have on the measurement of the cou-
pling of interest. So far, heteronuclear dipolar correlations
were derived for the simple case of a single IS spin pair. In
many cases, particularly when protons are involved, further
couplings need to be taken into account, for example by
considering an InS multi-spin system with abundant I-spins
and rare S-spins. Whereas, in the IS spin-pair case it is irrel-
evant which of the two spins occupies a transverse state and
evolves under the heteronuclear dipolar interaction to the
other, the situation is critically different for InS systems.
The heteronuclear couplings in an InS system behave
‘inhomogeneously’, following the terminology introduced
by Maricq and Waugh.64 In other words, the Hamiltonians
for the n different IS couplings commute, such that the
overall propagation of the spin system under the combined
action of all n couplings can be simplified to the sequential
action of the individual couplings. Thus, when a transverse
state of the I spins, i.e. Ix,y, is subject to the heteronuclear
couplings to the S spin, the situation can simply be viewed
In the course of the reconversion, the signal acquires one
more dipolar phase factor t , which differs from 0 unless
1
t1 is incremented in full rotor periods, i.e. t1 D nꢂR. In this
particular case, the resulting 2D spectrum corresponds to a
heteronuclear correlation spectrum, where t1 is modulated
only by the I-spin chemical shift.
The experiment described above yields an HSQC spec-
trum, familiar from solution-state studies,52b where the I- and
S-spin chemical shifts occur in the indirect and direct dimen-
sions, respectively, and was performed in order to obtain the
1
1H–13C HSQC spectrum discussed above (see the H–13
C
recoupled dipolar HSQC spectrum section and Fig. 3). Effec-
tively, I-spin polarization is transferred to the S-spins via
dipolar recoupling, with an I-spin spectral dimension divid-
ing the ꢃ-pulse train into two periods. Such an experiment
has been termed REPT.39,40 Its one-dimensional version
(t1 D 0 in Fig. 6a) is identical to the TEDOR (transferred-
echo, double resonance) sequence, which has been shown to
be useful for extracting heteronuclear dipolar couplings for
the case of selectively pairwise labelled systems.58 However,
the focus of this paper is on the determination of dipolar
coupling constants from rotor-encoded spinning sideband
patterns, which are observed in the indirect (t1) dimension
of two-dimensional experiments. The differences between
REPT and TEDOR experiments have also been discussed in
a)
b)
I
I
1
Ref. 39. Although it appears natural for H–X experiments
to start with the excitation period on the abundant I (1H)
spins, the experiment can of course be performed inversely,
as depicted in Fig. 6b. The initial S-spin polarization can, for
example, be created by a preceding CP step, while the detec-
tion takes place on the I-spins. In the case of 1H, a significant
advantage is obtained for the inverse variant on account of
detecting the nuclei with the higher magnetogyric ratio and,
hence, the higher sensitivity. In combination with fast MAS,
inverse detection has not been widely applied in solid-state
NMR, and has only recently attracted renewed attention,23,61
whereas it is routinely used in solution-state NMR.62
S
S
I
I
I
I
Figure 7. The local dipolar fields experienced by the 15N (S)
and 1H (I) nuclei.
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N
H hydrogen bond study by fast MAS S13
ꢃ
ꢀ
ꢁ
sin Nexc iꢁꢁ
cos Nexc jꢁ sin Nrec iꢁꢁ
as each I-spin experiencing only one coupling, because the
spin I iꢁ is unaffected by couplings HISjꢁ / IzjꢁSz for j D i. This
scenario is depicted schematically in Fig. 7a. Consequently,
in the InS spin system the evolution of the transverse state
of each I-spin is identical to the IS spin-pair case described
by Eqn (5). However, when the single S-spin occupies a
transverse state, Sx,y, it evolves in the local dipolar fields
of all I-spins, because the S-spin is part of all heteronuclear
dipolar Hamiltonians. During the excitation period, an Sy
state evolves into heteronuclear antiphase states in which
successively more and more I-spins are incorporated:41
S /
t
1
0
0
i
jDi
ꢄ
ꢁ
jꢁ
1
ð
cos N

14ꢁ
rec
t
iDj
depending on the particular choice of transverse I- or S-states
during excitation and reconversion. In the experiments,
the ‘pathway’ of transverse states can be selected by
placing the ꢃ/2-pulses at the start and/or at the end of
the recoupling periods accordingly (see also Fig. 4). The
pulse sequences giving rise to the signals described by
Eqns (11)–(14) are shown in Fig. 6a–d. Note that the pulse
sequences depicted in Fig. 6a and b, which have been
discussed above, include a polarization transfer from I to
S or vice versa respectively. The propagation of the spin
states in the two recoupling blocks occurs asymmetrically,
in that the transverse and longitudinal components are
interchanged between the two blocks. The variants shown
in Fig. 6 c and d, in contrast, are symmetrically designed,
and the polarisation in not transferred from one spin
species to the other, i.e., the same spin species occupies
a state of transverse polarisation during both excitation
and reconversion. This type of sequence (Fig. 6 c and d)
is termed DIP, taken from dipolar.40 DIP sequences can also
be considered 2D REDOR experiments, where the central
ꢃ-pulse is split into two ꢃ/2-pulses and the t1-dimension
is inserted in between. A previous example of such a type
of REDOR experiment (termed 3D REDOR) was proposed
by Michal and Jelinski, from which multiple distances could
be extracted based on the REDOR curves.63 In that case,
chemical shift information in two dimensions provided
ꢀ
N
exc iꢁ2IziꢁSz
0
ꢁ
i
iꢁ
Sy ꢀꢀꢀꢀꢀꢀꢀꢀꢀꢀ! Sy
cos N

exc
0
i
ꢀ
ꢁ
jꢁ
ꢀ 2
I jꢁSx sin Nexc iꢁꢁ
cos N

exc
0
z
0
i
jDi
ꢀ
C 4
I iꢁI jꢁSx sin Nexc iꢁꢁ sin Nexc jꢁꢁ
0
0
z
z
iDj
ꢁ
kꢁ
0
ð
cos N

š Ð Ð Ð
8ꢁ
exc
kDi,j
As demonstrated by Saalwa¨chter and Spiess,41 this growth
process can be followed by spin-counting experiments, or IS
multi-spin states can be selected with respect to the number
of I-spins involved by the use of appropriate phase cycles.
For simplicity, we will restrict our considerations here to
the leading terms, i.e. two-spin IzSx antiphase states, which
acquire the dipolar phases of all heteronuclear couplings in
the following way:
the necessary resolution for measuring several 13C–15
N
ꢀ
N
exc iꢁ2IziꢁSz
0
distances. The important difference between the REDOR
and the REPT/DIP approaches is how the information about
the heteronuclear dipolar coupling is accessed. While in
REDOR experiments the spectral intensity is recorded as a
function of the recoupling time (yielding REDOR curves), in
REPT/DIP experiments a rotor-encoded sideband pattern is
observed in the indirect (t1) dimension. From this pattern, the
dipolar coupling strength can be determined in a single two-
dimensional REPT/DIP experiment (which can in addition
be performed for various recoupling times, vide infra), while
the REDOR approach is based on a dephasing behaviour
which is observed for a series of experiments with different
recoupling times.
Additional complications in the InS spin system could
obviously arise from homonuclear dipolar interactions
among the I spins. Such interactions cause the system to
become ‘homogeneous’, and the overall propagation of the
spin system under the combined action of all couplings can
no longer be fully separated, but approximate approaches
through series expansions become necessary. Under the
fast MAS conditions used here, however, the homonuclear
couplings have only minor effects because they are not
recoupled, but are sufficiently suppressed by spinning
frequencies exceeding 25 kHz.41,65
ꢀ
ꢁ
i
jꢁ
Sy ꢀꢀꢀꢀꢀꢀꢀꢀꢀꢀ! ꢀ2
I iꢁSx sin Nexc iꢁꢁ
cos N

exc
0
z
0
i
jDi
9ꢁ
Equivalently, during reconversion the IzSx antiphase state
picks up the same dipolar phases when evolving back to a
transverse Sy state:
ꢀ
N
exc iꢁ2IziꢁSx
0
ꢀ
ꢀ
ꢁ
i
jꢁ
2
IziꢁSx ꢀꢀꢀꢀꢀꢀꢀꢀꢀꢀ! ꢀSy
sin Nrec iꢁꢁ
cos N 
rec
0
0
i
i
jDi
10ꢁ
Hence, considering heteronuclear two-spin states, the
important difference between the evolution of an IS and
an InS spin system is the presence of the product term
ꢂ
jꢁ
0
cos N

in the InS case in addition to the
exc/rec
iꢁ
jDi
sin N

term, which is similar from the two-spin case.
exc/rec
0
Hence, for an InS spin system the rotor-encoded sideband
patterns of IS two-spin correlations arise from one of the
following four expressions:
ꢃ
ꢃ
ꢃ
ꢄ
ꢀ
ꢁ
jꢁ
1
S /
S /
S /
sin Nexc iꢁꢁ sin Nrectiꢁꢁ
cos N

11ꢁ
rec
t
0
1
i
jDi
ꢄ
ꢀ
ꢁ
sin Nexc jꢁꢁ
cos Nexc jꢁ sin Nrec iꢁꢁ 12ꢁ
t
1
0
0
i
jDi
Recalling the spin system encountered in the benzoxazine
sample, there are eight protons proximate to the nitrogen,
only one of which is involved in the hydrogen bond, but
ꢄ
ꢀ
sin Nexc iꢁꢁ sin Nrectiꢁꢁ
13ꢁ
0
1
i
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

S14
G. R. Goward et al.
which can be identified from its chemical shift. Hence,
...
N H
Ð Ð Ð
in order to measure the N H pair coupling of interest
CH
CH3
OH
CH2
by means of rotor-encoded sideband patterns, influences
of further N-H couplings need to be avoided as much
as possible. This can be accomplished by performing the
1H–15N dipolar recoupling experiment given in Fig. 6c,
which gives rise to a spin-pair sideband pattern in the
F1 t1ꢁ dimension [corresponding to Eq. (13)], and which
provides resolution of the 1H chemical shift in the F2 t2ꢁ
dimension.
−100
−50
0
From the consideration of different possible 1H–15
N
dipolar recoupling experiments, it is, thus, concluded that
an experiment consisting of two recoupling periods with
50
1
1H transverse states and detecting the H signal during t2
is the optimum choice for the complex spin system under
investigation.
100
16
12
8
4
0
I-spin (1H) chemical shift (
RESULTS AND DISCUSSION
ppm)
Figure 8. The 2D 1H–15N spectrum of the 15N-labelled methyl
benzoxazine dimer obtained using the inversely (1H) detected
DIP-HSQC sequence (Fig. 6c), with ꢂrcpl D 18ꢂR. F1 and F2
skyline projections, corresponding to the (DIP-HSQC-filtered)
1H spectrum and the rotor-encoded spinning-sideband
patterns, respectively, are shown. The 1H resonances in the F2
dimension are assigned according to the 1H MAS spectrum.
The primary objective of this work is to determine the
Ð Ð Ð
dipolar coupling constant of the N H interaction and,
thereby, the distance between these two atoms. Based on
the phenolic character of the benzoxazines, and on the
anticipated six-membered ring geometry, a relatively weak
Ð Ð Ð
hydrogen bond, corresponding to a long N H distance,
was anticipated.43 A standard 1H–15N CP MAS spectrum
acquired at a slow spinning frequency of 4 kHz (not shown)
shows a single 15N resonance at ꢀ342.5 ppm with only first-
order sidebands of approximately 1% intensity with respect
to the isotropic resonance. The 15N shift is comparable to
1H resonances in the range of 10–15 ppm were assigned
Ð Ð Ð Ð Ð Ð
Ð Ð Ð Ð Ð Ð
to protons involved in N
bonds. Similar chemical shifts are also observed among
H
O or N
H
N hydrogen
C
the NH4 shift of the reference compound, NH4NO3, at
Ð Ð Ð Ð Ð Ð
imidazole-containing enzymes, in which the N
interaction has been extensively studied using solid-state
H
O
ꢀ358.4 ppm, and the small chemical shielding anisotropy is
also characteristic for a tetrahedral symmetry at the nitrogen
atom.66 Hence, both the chemical shift and its anisotropy
are indicative of a tetrahedrally coordinated nitrogen in
the methyl benzoxazine dimer, requisite for the hydrogen-
Ð Ð Ð Ð Ð Ð
NMR.5 These observations give qualitative assurance of the
Ð Ð Ð
hydrogen-bonding nature of the N H interaction in the
Ð Ð Ð
methyl benzoxazine dimer. The N O distance, determined
crystallographically,33 is 265.4 pm, corresponding to
a
bonded N
H
O arrangement to be present.
Ð Ð Ð
Ð Ð Ð
comparatively close interaction. For such N O (or O O)
distances, typical 1H chemical shifts range from 12 to
17 ppm.5 These examples, however, examined an N–H
bonding interaction, with a hydrogen ‘bridge’ to oxygen.
Thus, the relatively low chemical-shift value observed in
the benzoxazine dimer points toward a preferred O–H
As has been outlined above, REPT and DIP approaches
can be carried out such that a sideband pattern is observed, in
which the dipolar coupling information is encoded. Figure 8
shows an example of a rotor-encoded 1H-detected DIP-
HSQC spectrum (using the pulse sequence given in Fig. 6c).
The displayed F2 and F1 projections correspond to the (DIP-
HSQC-filtered) 1H spectrum and the rotor-encoded 15N–1H
sideband patterns, respectively. Note that the pulses were
applied on-resonance on the 15N channel, and the t1 signal
was detected as phase-insensitive cosine data. Since a pure
dipolar rotor-encoding of the t1 signal, as observed in our
case, gives rise to a periodic amplitude modulation, the
acquisition can be restricted to a single rotor period, i.e.
0 < t1 ꢁ ꢂR. However, to prevent correlated noise from being
superimposed with the sideband pattern, signal acquisition
for 0 < t1 ꢁ 2ꢂR is preferable. In the course of data processing,
the t1 signal is catenated before Fourier transformation
in t1.
Ð Ð Ð
bonding and, therefore, a relatively long N H hydrogen
bonding distance. This is consistent with the expected phe-
nolic character of the benzoxazine, in which the proton is
Ð Ð Ð Ð Ð Ð
involved in a continuous, bent O
six-membered ring.
H
N bond in a distorted
Of vital importance is the observation that the 1H
resonance of interest is well resolved from the other 1H
resonances in Fig. 8. Moreover, this resonance experiences
the strongest coupling to the 15N nucleus, since only this res-
onance exhibits third-order sidebands in the 15N dimension.
For longer recoupling times, third-order sidebands for the
CH3 and CH2 groups become evident, too, but these are not
analysed here.
Ð Ð Ð
The N H proton chemical shift of 11.2 ppm corre-
Ð Ð Ð
The build-up of rotor encoded N H sideband patterns
observed at the 11.2 ppm 1H shift are presented in Fig. 9
for a range of recoupling times ꢂrcpl, together with the cor-
responding fits, generated by numerically evaluating the
sponds to values expected for hydrogen-bonded structures,
for example in bilirubin14 as well as in ureido-pyrimidinone
complexes studied using high-resolution H MAS,23 where
1
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N H hydrogen bond study by fast MAS S15
slightly, but noticeably, larger than the values resulting from
experiments with shorter recoupling times. This indicates
that, at recoupling times around 1 ms, the experimental
scenario begins to deviate from the assumptions on which
the simple analysis of the patterns in terms of an NH spin-
pair is based. It is expected that, at longer recoupling times,
additional effects such as weak residual 1H–1H dipolar
interactions, long-range 15N–1H couplings, as well as RF
pulse imperfections and B1 imhomogeneities, gradually
come into play, such that the resulting patterns can no longer
Experimental Results
Fits of Experimental Data
rIS = N/A
τrcpl = 10 τR
a)
b)
τrcpl = 14 τR
rIS = 193 pm
c)
d)
e)
f)
τrcpl = 18 τR
τrcpl = 22 τR
τrcpl = 26 τR
τrcpl = 30 τR
rIS = 192 pm
r
IS = 195 pm
IS = 202 pm
Ð Ð Ð
be interpreted solely on the basis of a single N H coupling
and ideal hard RF pulses. Among others, these imperfections
reduce higher-order sideband intensities, which are expected
for stronger dipole–dipole couplings. For recoupling times
<1 ms, however, the simple analysis is appropriate and
yields distances with an estimated error of <2%, as
demonstrated by the striking consistency of the results for
ꢂrcpl < 26ꢂR, with an optimum recoupling time of 22ꢂR.
Moreover, in the inversely, i.e. 1H, detected HSQC
experiments, the signal sensitivity is dramatically enhanced
compared with conventional 15N detection experiments. In
the experiments presented here, the typical data acquisition
time was about 1 h, whereas comparable experiments using
15N detection methods, carried out over a 12 h period,
provided significantly inferior signal intensity.
r
rIS = 205 pm
−100 −200
0
−100 −200
200 100
0
200 100
kHz
kHz
Figure 9. Spinning-sideband patterns extracted from 1H–15
REPT inversely (1H) detected DIP-HSQC spectra with
(a) ꢂrcpl D 10ꢂR, (b) ꢂrcpl D 14ꢂR, (c) ꢂrcpl D 18ꢂR,
N
(d) ꢂrcpl D 22ꢂR, (e) ꢂrcpl D 26ꢂR, (f) ꢂrcpl D 30ꢂR. Experimental
data are shown on the left, together with the corresponding
best-fit simulated spectra.
expression given in Eqn 13.‡ The expected trend is observed,
as the relative intensity of the third-order sidebands increases
with increasing recoupling time. The relative intensities,
together with the dipolar coupling constants, and corre-
SUMMARY AND OUTLOOK
Ð Ð Ð
The goal of this investigation was to measure the N
H
Ð Ð Ð Ð Ð Ð
distance in an N
H
O hydrogen bond of methyl-
Ð Ð Ð
sponding N H hydrogen-bond lengths are given in Table 3.
substituted benzoxazine dimers, which are presumed to
contain the hydrogen bond as a part of a distorted six-
membered ring. To determine relatively weak dipole–dipole
couplings, special care had to be taken to select the
most informative experiment. In particular, in the methyl
benzoxazine dimer the single 15N nucleus is subject to several
As anticipated, the measured distance is consistent with a
Ð Ð Ð
long N H hydrogen bond of 194 š 5 pm. In Fig. 9a, only
first-order sidebands are evident; therefore, no dipolar cou-
pling constant can be obtained. In Fig 9b–d, the relative
intensity of the third-order sidebands increases uniformly,
Ð Ð Ð
Ð Ð Ð
N
H couplings of comparable strength. Therefore, the
from which dipolar coupling constants and N H distances
are accurately determined.
chemical shift resolution in the 1H dimension is essential
Ð Ð Ð
to the study in order to identify and access the N
H
For sideband patterns recorded with longer recoupling
times, i.e. ꢂrcpl D 26ꢂR and ꢂrcpl D 30ꢂR, as illustrated in Fig. 9e
coupling relevant to the hydrogen bonding. Moreover, the
perturbing influence of further couplings to other spins
must be minimized. To this end, 1H–15N dipolar recoupling
methods under fast MAS have been considered with respect
to 1H or 15N signal detection and the pathway of transverse
1H or 15N spin states. Of the methods described, the inversely
(1H) detected symmetric DIP-HSQC sequence was chosen,
on the basis of its ability, firstly, to provide 1H chemical shift
resolution, and, secondly, to avoid possible heteronuclear
multi-spin effects in the case of the methyl benzoxazine
Ð Ð Ð
and f, the fitting procedure yields N H distances that are
Table 3. Dipolar coupling constants and internuclear N–H
distances determined from the inversely detected DIP-HSQC
rotor-encoded sideband patterns in Fig. 9
Ratio of
3rd/1st
Ratio of
5th/3rd
Nexc/ꢂR sidebands (%) sidebands (%) DIS/kHz RIS/pm
Ð Ð Ð
dimer, such that a relatively long N H hydrogen bonding
distance could be determined despite the presence of several
other NH couplings of comparable strength. Using this
10
14
18
22
26
30
0
4
16
37
81
83
0
0
0
0
4
7
N/A
N/A
1.69 š 0.1 193 š 5
1.71 š 0.1 192 š 5
1.63 š 0.1 195 š 5
1.48 š 0.1 202 š 5
1.43 š 0.1 205 š 5
Ð Ð Ð
experiment, a long N H distance of 194 š 5 pm was
determined, which indicates that, in the hydrogen bond,
the proton is proximal to the oxygen, while being shared to
some extent with the nitrogen. Thus, the dimer exhibits a
predominant phenolic character. Furthermore, with regards
to the experimental time necessary to achieve comparable
spectral quality, the 1H-detected variants are known to
provide significant sensitivity enhancement.23,61
‡Numerical evaluation and fitting of the sideband patterns (using a
Levenberg-Marquardt algorithm) was performed using a CCC
routine written by K. Saalwa¨chter.
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

S16
G. R. Goward et al.
In conclusion, the inversely detected DIP-HSQC experi-
ment has achieved three major goals. Firstly, the resolution
in the 1H dimension allows us unambiguously to identify the
1H–15N coupling of interest. Secondly, the effect of dipolar
interactions between the 15N and other protons, though of
comparable strength, could be suppressed by using fast MAS
Solid-state NMR
Data were collected on a 16.4 T magnet with a Bruker
DRX spectrometer, using a double-resonance MAS probe
supporting rotors of 2.5 mm outer diameter. The resonance
1
frequencies of H, 13C and 15N are 700 MHz, 176 MHz, and
71 MHz respectively. The sample was spun at 30 kHz. The
ꢃ/2 pulse lengths were set to 2 µs for 1H and 1H–13
1
and selecting a coherence transfer pathway with H trans-
C
experiments, and to 2.5 µs for 1H–15N experiments. A
recycle delay of 3 s was used. The spectra are referenced
to 15NH4NO3 (ꢀ358.4 ppm, 15N), adamantane (1.63 ppm,
1H) and alanine (52 ppm, 13CH).
The rotor-synchronized 1H–1H DQ spectrum was
recorded using a five-pulse sequence with ꢂexc D 1/2ꢂR.
States-TPPI was applied for phase-sensitive detection70 in t1,
and 64 slices were detected in the indirect dimension, with
16 transients averaged per slice. The experimental time was
about 1 h.
verse states during both recoupling periods. In this way,
the 1H–15N coupling of interest could be measured selec-
tively. Thirdly, by employing 1H detection signal, sensitivity
is enhanced and experiment times are minimized by more
than an order of magnitude.
It should be noted that the 1H–15N dipolar NMR methods
described here focus on elucidating average structural fea-
tures, and maximizing the spectral intensity. The dynamics
Ð Ð Ð Ð Ð Ð
of proton transfer within N
Ð Ð Ð Ð Ð Ð
O and N
H
H
N hydro-
gen bonds is also a topic of intense investigation, particularly
for low-barrier hydrogen bonds.67,68 However, in our inves-
tigation of benzoxazine dimers, we have assumed a rigid
structure, which does not include possible proton-transfer
effects. Examination of the influence of dynamics on dipo-
lar recoupling NMR experiments and the development of
methods to access dynamical information on different time
scales in the millisecond range is in progress (see also Ref. 70).
Moreover, further methodological developments will
exploit the different ways by which the variants of the HSQC
recoupling experiment discussed here are sensitive to multi-
spin interactions. The combination of different 15N- and
1H-detected variants of REPT and DIP-HSQC approaches
offers promising potential for studying both a spin-pair
interaction of interest as well as for exploring, e.g., the
geometry of the multi-spin environment.
All heteronuclear dipolar recoupling experiments were
carried out with equal excitation and reconversion times.
1
For the rotor-synchronized H–13C spectrum, an excitation
time of ꢂrcpl D ꢂR was used, and the number of slices in the
indirect dimension was 24, with 512 transients averaged per
slice. Phase-sensitive detection in t1 was achieved through
States-TPPI, and the experiment time was ¾8 h. The time
signal for the 1H–15N HSQC spinning sideband patterns
was recorded at excitation times of ꢂexc D 2, 6, 10, 14, 16,
18, 20, 22, 26 and 30ꢂR, with t1 being incremented in steps
of 1.67 µs. 20 slices were collected in the indirect dimension,
with 128 transients averaged per slice The experimental time
was approximately 90 min. The 1H transmitter frequency
was set equal to ¾6 ppm, and the on-resonance transmitter
frequency for the 15N signal was selected.
1
For heteronuclear experiments employing H detection
in the solid state, an adequate phase cycle is of particular
importance to ensure pure signal.71 The basic phase cycling
scheme starts with a sign alternation through a phase
EXPERIMENTAL
Synthesis
The synthesis of a 15N-labelled methylamine-based benzox-
azine dimer, N,N-bis(3,5-dimethyl-2-hydroxybenzyl)methy-
lamine-15N, was carried out according to a previous
study33 with the exception that methylamine-15N hydrochlo-
ride was substituted for methylamine. To produce the
methylamine-15N-based benzoxazine monomer, the mixture
of (methylamine-15N hydrochloride 99%ꢁ : NaOH 98%ꢁ :
inversion of one of the two 90 pulses applied to 15N
°
in every other transient. In this way, the signal of 1H
involved in NH correlations is selected after two steps.
A further sign alternation is performed on the initial 1H
polarization by changing the phase of the first 1H pulse
accordingly, as familiar from CP experiments (two more
steps in the cycle). In addition to these fundamental selection
procedures, the suppression of contributions from possible
1H higher-order coherences to the signal is important. This
was achieved by implementing two nested four-step phase
°
water D 1 : 1 : 10) was stirred at 10 C for 1 h in order
to obtain methylamine-15N solution. Subsequently, a 2 : 1
ratio of p-formaldehyde:2,4-dimethylphenol was added to
the methylamine-15N solution with chloroform as a reaction
solvent, following which the reaction was refluxed hetero-
geneously for 2 h. The reaction product in the chloroform
layer was washed by 2 M NaOH solution several times, and
finally washed using distilled water. 1 M of methylamine-
15N-based benzoxazine monomer was reacted with 1 M of
1
cycles of the excitation and reconversion blocks on the H
channel. In the case of 15N detection, simple two-step cycles
usually suffice, because dilute 15N spins do not usually
form homonuclear double or triple quantum coherences.
Heternonuclear pulse sequences employing 1H detection, in
contrast, need to eliminate spurious signal of homonuclear
1H double, and potentially higher, quantum coherences to
ensure undistorted patterns of purely heteronuclear origin.
Consequently, in our experiments the overall phase cycle
consisted of 2 ð 2 ð 4 ð 4 D 64 steps.
°
2,4-dimethylphenol at 105 C for 30 min without solvent,
and the final product was recrystallized three times using
n-hexane, with an overall yield of 81%. White, irregular crys-
tals were obtained. 1H NMR (200 MHz, CDCl3, 298 K) υ :
2.21, 2.22 (15H, –CH3), 3.64 (4H, Ar–CH2 –N), and 6.73, 6.87
(4H, Ar–H). 13C–NMR (50.1 MHz, CDCl3, 298 K) υ : 15.74,
20.41 (4C, Ar–C), 41.06 (1C, N–C), 59.25 (2C, Ar–C–N), and
121.80, 124.24, 128.24, 130.84, 152.00 (12C, Ar).
Acknowledgements
Financial support from the DFG is acknowledged. HI and GRG
thank the Alexander von Humboldt Stiftung and NSERC (Canada)
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17

Ð Ð Ð
N H hydrogen bond study by fast MAS S17
for a senior scientist award and a post-doctoral fellowship respec-
tively. Stimulating discussions with Kay Saalwa¨chter are gratefully
acknowledged.
36. Jeffrey GA, Yeon Y. Acta Crystallogr. Sect. B 1975; 31: 2345.
37. Gullion T, Schaefer J. J. Magn. Reson. 1989; 81: 196.
38. Gullion T. Concepts Magn. Reson. 1998; 10: 277.
39. Saalwa¨chter K, Graf R, Spiess HW. J. Magn. Reson. 1999; 140:
471.
40. Saalwa¨chter K, Graf R, Spiess HW. J. Magn. Reson. 2001; 148:
398.
41. Saalwa¨chter K, Spiess HW. J. Chem. Phys. 2001; 114: 5707.
42. Kimura H, Shoji A, Sugisawa H, Deguchi K, Naito A, Saito H.
Macromolecules 2000; 33: 6627.
REFERENCES
1. Jeffrey GA, Saenger W. Hydrogen Bonding in Biological Structures.
Springer-Verlag: New York, 1991.
2. Sijbesma RP, Beijer FH, Brunsveld L, Folmer BJB, Hirschberg
JHK, Lange RFM, Lowe JKL, Meijer EW. Science 1997; 278: 1601.
3. Schnell I, Spiess HW. J. Magn. Reson. 2001; 151: 153.
4. Brown SP, Spiess HW. Chem. Rev. 2001; submitted for publica-
tion.
43. Kim HD, Ishida H. J. Phys. Chem. 2001; submitted for publication.
44. Van Rossum B-J, Fo¨rster H, de Groot HJM. J. Magn. Reson. 1996;
124: 516.
5. Wei Y, McDermott AE. In Modeling NMR Chemical Shifts: Gaining
Insights into Structure and Environment, ACS Symposium Series
732. ACS: 1999; 177–193.
6. Berglund B, Vaughan RW. J. Chem. Phys. 1980; 73: 2037.
7. Jeffery A, Yeon Y. Acta. Crystallogr. Sect. B 1986; 42: 410.
8. Harris RK, Jackron P, Merwin LH, Say BJ. J. Chem. Soc. Faraday
Trans. 1988; 84: 3649.
45. Brown SP, Schnell I, Brand JD, Mu¨llen K, Spiess HW. J. Am.
Chem. Soc. 1999; 121: 6712.
46. Brown SP, Schaller T, Seelbach UP, Koziol F, Ochsenfeld C,
Kla¨rner FG, Spiess HW. Angew. Chem. 2001; 40: 717.
47. Ochsenfeld C, Head-Gordon M. Chem. Phys. Lett. 1997; 270: 399.
48. Ochsenfeld C, Brown SP, Schnell I, Gauss J, Spiess HW. J. Am.
Chem. Soc. 2001; 123: 2597.
9. Yamauchi K, Kuroki S, Fujii K, Ando I. Chem. Phys. Lett. 2000;
324: 435.
10. Benedict H, Limbach HH, Wehlan M, Fehlhammer WP, Golbuev
NS, Janoschek R. J. Am. Chem. Soc. 1998; 120: 2939.
11. Waugh JS, Huber LM, Haeberlen U. Phys. Rev. Lett. 1968; 20:
180.
49. Ochsenfeld C. Phys. Chem. Chem. Phys. 2000; 2: 2153.
50. (a) Hediger S, Meier BH, Ernst RR. Chem. Phys. Lett. 1995; 240:
449. (b) Hediger S, Signer P, Tomaselli M, Ernst RR, Meier BH.
J. Magn. Reson. 1997; 125: 291.
51. Van Rossum B-J, de Groot CP, Ladizhansky V, Vega S, de
Groot HJM. J. Am. Chem. Soc. 2000; 122: 3465.
12. (a) Gerstein BC, Chou C, Pembleton RG, Wilson RC. J. Phys.
Chem. 1977; 81: 565. (b) Scheler G, Haubenreisser U, Rosen-
berger H. J. Magn. Reson. 1981; 44: 134. (c) Maciel GE, Bron-
nimann CE, Hawkins B. Adv. Magn. Reson. 1990; 14: 125.
13. Schnell I, Brown SP, Lee HL, Ishida H, Spiess HW. J. Am. Chem.
Soc. 1998; 120: 11 784.
14. Brown SP, Zhu XX, Saalwa¨chter K, Spiess HW. J. Am. Chem. Soc.
2001; 123: 4275.
15. Lee M, Goldburg WI. Phys. Rev. A 1965; 140: 1261.
16. Bielecki A, Kolbert AC, Levitt MH. Chem. Phys. Lett. 1989; 155:
341.
17. Bielecki A, Kolbert AC, de Groot HJM, Griffin RG, Levitt MH.
Adv. Magn. Reson. 1990; 14: 111.
18. Vinogradov E, Madhu PK, Vega S. Chem. Phys. Lett. 1999; 314:
443.
19. Sakellariou D, Lesage A, Hodgkinson P, Emsley L. Chem. Phys.
Lett. 2000; 319: 253.
52. (a) Morris GA, Freeman R. J. Am. Chem. Soc. 1979; 101: 760.
(b) Bodenhausen G, Ruben DJ. Chem. Phys. Lett. 1980; 69: 185.
(c) Bax A, Griffey RH, Hawkins BL. J. Magn. Reson. 1983; 55: 301.
53. Geen H, Titman JJ, Gottwald J, Spiess HW. J. Magn. Reson. A
1995; 114: 264.
54. Gottwald J, Demco DE, Graf R, Spiess HW. Chem. Phys. Lett.
1995; 243: 314.
55. Friedrich U, Schnell I, Brown SP, Lupulescu A, Demco DE,
Spiess HW. Mol. Phys. 1998; 95: 1209.
56. Fechtenko¨tter A, Saalwa¨chter K, Harbison M, Mu¨llen K,
Spiess HW. Angew. Chem. Int. Ed. Engl. 1999; 38: 3039.
57. Graf R, Heuer A, Spiess HW. Phys. Rev. Lett. 1998; 80: 5738.
58. Dollase T, Graf R, Heuer A, Spiess HW. Macromolecules 2001; 34:
298.
59. Schmedt auf der Gu¨nne J, Eckert H. Chem. Eur. J. 1998; 4: 1762.
60. Sørensen OW, Eich GW, Levitt MH, Bodenhausen G, Ernst RR.
Prog. NMR Spectrosc. 1983; 16: 163.
20. Lesage A, Duma L, Sakellariou D, Emsley L. J. Am. Chem. Soc.
2001; in press.
21. Geen H, Titman JJ, Gottwald J, Spiess HW. Chem. Phys. Lett. 1994;
227: 79.
22. Lee YK, Kurur ND, Helmle M, Johannessen OG, Nielsen NC,
Levitt MH. Chem. Phys. Lett. 1995; 242: 304.
23. Schnell I, Langer B, Soentjens SHM, van Genderen MHP,
Sijbesma RP, Spiess HW. J. Magn. Reson. 2001; 150: 57.
24. Brown SP, Schnell I, Brand JD, Mu¨llen K, Spiess HW. Phys.
Chem. Chem. Phys. 2000; 2: 1735.
61. (a) Ishii Y, Tycko R. J. Magn. Res. 2000; 142: 199. (b) Ishii Y,
Yesinowski JP, Tycko R. J. Am. Chem. Soc. 2001; 123: 2921.
62. (a) Maudsley AA, Muller L, Ernst RR. J. Magn. Reson. 1977;
28: 463. (b) Muller L. J.Am. Chem. Soc. 1979; 101: 4481.
(c) Bax A, Griffey RH, Hawkins BL. J. Magn. Reson. 1983; 55:
301. (d) Bodenhausen G, Ruben DJ. Chem. Phys. Lett. 1980; 69:
185. (e) Redfield AG. Chem. Phys. Lett. 1983; 96: 537. (f) Otting G,
Wu¨thrich K. J. Magn. Reson. 1988; 76: 569.
63. Ernst RR, Bodenhausen G, Wokaun A. Principles of Nuclear
Magnetic Resonance in One and Two Dimensions. Claredon Press:
Oxford, 1997.
64. Maricq MM, Waugh JS. J. Chem. Phys. 1979; 70: 3300.
65. Filip C, Hafner S, Schnell I, Demco DE, Spiess HW. J. Chem. Phys.
1999; 110: 423.
25. Ning J, Ishida H. J. Polym Sci. Chem. Educ. 1994; 32: 1121.
26. Shen BS, Ishida H. Polym. Comp. 1996; 17: 710.
27. Lui JP, Ishida H. In Polymeric Materials Encyclopedia. CRC Press:
Florida, 1996; 484–494.
28. Ishida H, Low HY. Macromolecules 1997; 30: 1099.
29. Kim HD, Ishida H. J. Appl. Polym. Sci. 2001; accepted for
publication.
30. Macko J, Ishida H. J. Polym. Sci. Polym. Phys. 2000; 38: 2687.
31. Ishida H, Rodriguez Y. Polymer 1995; 36: 3151.
32. Wirasate S, Dhumrongvaraporn S, Allen DJ, Ishida H. J. Appl.
Polym. Sci. 1998; 70: 1299.
66. Mason J. In Multinuclear NMR. Plenum Press: New York, 1987;
Chapter 12.
67. Takeda S, Inabe T, Benedict C, Langer U, Limbach HH. Ber.
Bunsenges. Phys. Chem. 1998; 102: 1358.
68. Jimenez JA, Claramunt RM, Mo O, Yanez M, Wehrmann F,
Buntkowsky G, Limbach HH, Goddard R, Elguero J. Phys. Chem.
Chem. Phys. 1999; 1: 5113.
33. Dunkers J, Zarate EA, Ishida H. J. Phys. Chem. 1996; 100: 13 514.
34. Chirachanckai S, Laobuthee A, Ishida H, Tashiro K. J. Am. Chem.
Soc. 2001; submitted for publication.
69. Schnell I, Watts A, Spiess HW. J. Magn. Reson. 2001; 149: 90.
70. Marion D, Ikura M, Tschudin R, Bax A. J. Magn. Reson. 1989; 85:
393.
35. Inabe T, Luneau I, Mitani T, Maruyama Y, Takeda S. Bull. Chem.
Soc. Jpn. 1994; 67: 843.
71. Bodenhausen G, Kolger H, and Ernst RR. J. Magn. Reson. 1984;
58: 370.
Copyright  2001 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2001; 39: S5–S17