Вы находитесь на странице: 1из 43

Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Recent advances in experimental solid state NMR methodology for


half-integer spin quadrupolar nuclei
M.E. Smith a,*, E.R.H. van Eck b
b

a
Department of Physics, University of Warwick, Coventry CV4 7AL, UK
School of Physical Sciences, University of Kent, Canterbury, Kent CT2 7NR, UK

Received 11 November 1998

Contents
1.
2.
3.
4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nuclear interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Background experimental principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
One-dimensional experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1. Static broad line experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2. Magic angle spinning observation of the central transition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3. Magic angle spinning and spinlocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4. Magic angle spinning observation of satellite transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5. Variable angle spinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6. Double angle spinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5. Multiple resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1. Cross-polarisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2. SEDOR, REDOR and TEDOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3. TRAPDOR and REAPDOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6. Two-dimensional experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1. Nutation NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1.1. Off-resonance nutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2. Dynamic angle spinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3. 2D MQMAS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.1. Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4. 2D XY correlation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Keywords: Experimental NMR; Materials characterisation; Solid-state; Quadrupole nuclei

* Corresponding author. Tel.: +44 1203 522 380; fax: +44 1203 692 016.
E-mail address: M.E.Smith.1@warwick.ac.uk (M.E. Smith)
0079-6565/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved.
PII: S0079-656 5(98)00028-4

160
161
163
169
169
171
179
181
183
183
185
185
186
188
190
190
191
191
193
195
197
199
199
199

160

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

1. Introduction
NMR is applicable to any nucleus that possesses a
magnetic moment and has made a significant impact
on branches of science and technology as diverse as
measuring spin gaps in magnetic materials at less than
4 K to clinical diagnosis through imaging techniques.
Although NMR has, in principle, widespread application to the Periodic Table, the vast majority of studies
have been limited to relatively few of the NMR-active
nuclei. One of the main reasons for the limited range
of nuclei studied has been low sensitivity, with 1H
being the most studied nucleus because of its large
gyromagnetic ratio, although with ever higher applied
magnetic fields available the range of nuclei being
routinely studied is expanding. Many of the other
nuclei that are now commonly studied, such as 13C,
15
N, 31P and 29Si, are spin-1/2 nuclei which have the
attraction that their spectra are largely determined by
chemical shift effects, especially if averaging techniques such as magic angle spinning (MAS) and decoupling are applied, so that spectral interpretation is
relatively straightforward. However most NMRactive nuclei have a spin-quantum number I 1/2
and consequently have a non-spherically symmetric
electrical charge distribution within the nucleus that
gives rise to a nuclear electrical quadrupole moment.
This quadrupole moment interacts with gradients in
the electric field producing splittings of the nuclear
energy levels. As outlined in the next section these
effects can be very large such that first-order broadening can spread the intensity over a very significant
frequency range. Quadrupolar nuclei fall into two
major categories depending on whether or not a
nucleus possesses integer spin. Non-integer spin
nuclei experience no first-order broadening of the
central (1/2, 1/2) transition (vide infra) and for
NMR of powder samples containing such nuclei it is
the central transition that is usually observed. The
non-central, or so-called satellite transitions, can be
spread over a frequency range of many MHz, and
even though this makes their observation by pulse
techniques difficult it has recently been realised they
can be measured, and provide useful information.
The importance of non-integer spin quadrupole
nuclei can be gauged by considering the 120 nuclei
usually quoted in NMR Tables with 31 spin-1/2, 9
with integer spin and 80 with non-integer spin 1/2

(3/2 (32), 5/2 (22), 7/2 (18) and 9/2 (8)). Hence with
around two-thirds of NMR-active, stable nuclei being
non-integer spin quadrupolar their study merits
serious attention. In solids the major difficulty for
such nuclei is that for the more easily observed central
transition the quadrupole interaction can be sufficiently
strong that second-order quadrupolar effects (vide infra)
have to be considered, and these can cause broadening
of the central transition of tens of kHz that causes overlap of the resonances in the NMR spectra from powders.
The commonly applied approach of MAS, used so
successfully for averaging anisotropic dipolar and
chemical shift effects can only partially reduce
second-order quadrupole effects. The residual anisotropy, and indeed the presence of isotropic secondorder quadrupole effects has led to a great deal of
theoretical and experimental endeavour, particularly
over the last decade, to produce alternative NMR
approaches for examining such nuclei in the solid state
to give better resolution. This has resulted in a whole
range of methods available for the study of such nuclei,
some being very sophisticated. These methods include
two-dimensional (2D) approaches, complex sample reorientation during the experiment and excitation of
multiple quantum transitions. The main categories are:
1. Static, MAS and variable angle spinning for observing the central transition.
2. Static and MAS for observing the satellite transitions.
3. Spatial reorientation; dynamic angle spinning
(DAS) and double angle rotation (DOR).
4. Multiple resonance including cross-polarisation
(CP) and indirect detection (e.g. TRAPDOR).
5. Nutation.
6. Multiple quantum (MQ).
There are some excellent review articles giving
detailed background information and applications of
many of these techniques individually and these will
be extensively referenced here. This article seeks to
give an overview of the physical background of these
approaches, and in particular to compare their relative
merits. It is a goal that this article should be accessible
to the non-specialist who simply wants to use solid
state NMR of a non-integer spin quadrupole nucleus
to better characterise samples. It will seek to answer
practical questions, such as how readily the different
techniques can be implemented, what information

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

they provide, what are their limitations and how can


they be combined to give a complete NMR methodology for studying quadrupole nuclei in the solid state.

2. Nuclear interactions
NMR interactions are represented by a spin Hamiltonian that will have parts that correspond to the
experimental conditions, the so-called external part,
and those parts that result from the sample itself, the
internal part, which provide information about the
structure of the sample. The total interaction energy
of the nucleus may be expressed as a sum of individual Hamiltonians given in Eq. (1), which are
discussed in detail several excellent books [14].
Htot vo Iz gB1 =2I eivt I eivt HD
HCS Hq1 Hq2

The basis of the NMR experiment is the non-degenerate nuclear energy levels created by the Zeeman
interaction ( v oIz, Fig. 1(a)) of the nuclear
magnetic dipole moment m ( g I, where g is the
gyromagnetic ratio of the nucleus) with an applied
magnetic field Bo. This field is taken to define the zaxis in the laboratory frame and Iz is the z-component
of I with eigenvalues mz ( I mz I), and v o is the
Larmor frequency. In this article we will only
consider the high field limit whereby the nuclear
spin states are well described by the Zeeman energy
levels and that all the other interactions can be
regarded as perturbations of these spin states. The
actual NMR experiment involves measurement of
the energy separation of these levels by application
of a time-varying orthogonal magnetic field B1 (term 2
in Eq. (1)). B1 excites transitions (through I and I,
the conventional spin raising and lowering operators)
when its frequency (v ) is close to v o, typically in the
rf region 10 MHz1 GHz.
For spin-1/2 nuclei in diamagnetic insulating
solids the important interactions experienced are
dipolar (HD) and chemical shielding (HCS). Given
the tensorial nature of these interactions, for a
powder containing a random distribution of particle orientations relative to the main magnetic
field, these interactions give rise to broadening
of the NMR spectra. Fortunately, to first-order,

161

all these interactions (HD, HCS and Hq (I)) have similar


angular dependencies of (3cos 2 u 1 h sin 2 u
cos 2u ) where h is the asymmetry parameter of the
interaction tensor in the molecular principal axes (h
0 for axial symmetry).
The interaction between the electric quadrupole
moment (eQ) of the nucleus and the electric field
gradient at the nuclear site can have a profound effect
on the NMR spectrum. The electric field gradient is
again a tensor interaction that in its principal axis
system (PAS) is described by the three components
Vx 0 x 0 , Vy 0 y 0 and Vz 0 z 0 , where 0 indicates that the axes are
not necessarily coincident with the laboratory axes
defined by the magnetic field. Although the tensor is
completely defined by these components it is conventional to recast these into the electric field gradient
eq Vz 0 z 0 and the asymmetry parameter hq Vy 0 y 0
Vx 0 x 0 =Vz 0 z 0 : The electric field gradient is set up by the
charge distribution outside the ion (e.g. Al 3) but the
initially spherical charge distribution of inner shells of
electrons will become polarised to lower their energy
in the electric field. This polarisation produces an
electric field gradient at the nucleus itself of eqn
eq(1 g ) where (1 g ) is the Sternheimer antishielding factor which is a measure of the magnification of eq caused by distortion of the core electrons
[5]. Full energy band structure calculations of electric
field gradients show how important the contribution of
the electrons on the ion itself are compared to the
lattice [6]. Although the quadrupole interaction is an
electrical interaction it depends on the orientation of
the nuclear spins and therefore affects the nuclear
energy level splittings [7]. The background of the
quadrupole interaction is given in the classic article
by Cohen and Reif [7]. The quadrupole Hamiltonian
(considering axial symmetry for simplicity) in the
laboratory frame with u being the angle between the
z 0 -axis of the quadrupole PAS and Bo is
Hq hCq =8I2I 13cos 2 u 13Iz a
3sin ucos uI2 I I I I Iz
2
2
I

3sin2 u=2I

where Cq e2 qQ=h1 g and a II 1. In


the limit HZ q Hq a standard perturbation expansion
using the eigenstates of HZ is applicable. The firstorder term splits the spectrum into 2I components

162

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 1. The effects of the quadrupolar interaction for an I 5/2 nucleus showing (a) the energy level diagram and the corresponding (b)
complete first-order spectrum and second-order (c) static and (d) MAS spectra for h q 0 for the central transition. A II 1 3=4n2q =n0 .
2

(Fig. 1(a)) 1 of intensity m 1Ix m / a mm


1 at frequency n1
m away
2
n1
m 3Cq =4I2I 13cos u 1mz 1=2

This perturbation can cause the non-central transitions (i.e. mz 1/2) to be shifted (Fig. 1(b)) sufficiently far from the Larmor frequency that these
1
The angular terms for the second-order broadening correct those
given in Fig. 2(a) of M.E. Smith, Appl. Magn. Reson., 4 (1993) 1.

transitions become difficult to observe with conventional pulse techniques. Fortunately for the central
(mz 1/2) transition, n1
m 0 and the dominant
perturbation is to second-order only Eq. (4) which
gives a characteristic lineshape (Fig. 1(c) for axial
symmetry).
2
2
2
2
n2
m 9Cq =64no I 2I 1 a 3=41 cos u

 9cos2 u 1

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

163

Fig. 1. (Continued).

As h q varies, the lineshapes of both the central and


satellite transitions change (Fig. 2) which can provide
useful structural information as h q is related to the
local symmetry of the site. Other practical aspects of
interpreting quadrupolar perturbed NMR spectra are
discussed later.

3. Background experimental principles


As is well known in NMR, a pulse of duration Tp
applied exactly on resonance produces a resultant
magnetic field orthogonal to Bo in the rotating frame
[24] which causes a coherent oscillation of the
magnetisation such that it is tipped by an angle u p
( gB1 Tp ) away from the direction defined by Bo.
This brings the first important practical consideration

for quadrupolar nuclei in that the presence of the


different transitions and the extent to which they are
excited will change the effective pulse lengths and
observed intensities which can be explained within
the fictitious spin-1/2 formalism [8]. For excitation
of just the central transition, the 90-pulse is reduced
from when all the transitions are excited, or equivalently from a spin-1/2 nucleus with the same gyromagnetic ratio, by a factor of (I 1/2) [8]. The
values of this factor for all the transitions are
summarised in Table 1. In the extreme cases of n1 q
nq (n1 gB1 =2p and the quadrupole frequency nq
3Cq =2I2I 1 is a more direct comparison), the
non-selective regime, and n1 p nq , the selective
regime, the observed magnetisation displays simple
sinusoidal oscillations at n 1 and (I 1=2n1 ,
respectively. However in the intermediate regime

164

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 2. The effects of the asymmetry parameter (h q) on (a) the first-order satellite and (b) the second-order central transition lineshapes.
(A II 1 3=4n2q =n0 ).

i.e. n1 nq then a complex oscillation of the magnetisation occurs [9] as illustrated by the AlO6 resonance
of Y3Al5O12 (Fig. 3) and is the basis of two-dimensional quadrupole nutation spectroscopy (vide infra).
If the rf-pulses used are too long the lineshape can
show severe distortion because of this nutation effect,
as has been calculated [10] and observed experimentally [10,11]. Also for the most accurately quantitative
spectra from samples with sites that can have widely
differing quadrupolar coupling constants short, small
flip pulses should be used whose length (Tp) should
satisfy (I 1=2v1 Tp p=6 for less than 5% error in
intensity [12].
As the spectrum is produced by Fourier
transformation of the time domain signal (the free
induction decay (FID)), distortions can occur for the

broad lines that can be encountered for quadrupole


nuclei in solids, because of non-uniform excitation
of the spectrum, the frequency response of the probe
and corruption of the initial part of the time domain
signal. The power spectrum Pn of a pulse applied at
n o is given by a sine-square function [13]
Pn sin2 pTp no n=no n2

This function describes the best possible case, with


often a much stronger frequency dependence of power
output delivered at the probehead. (It should be noted
here that other excitation schemes are possible for
quadrupolar nuclei, such as adiabatic passage [14]
and stochastic excitation [15] and their development
should be monitored carefully.) The pulse is then

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

165

Fig. 3. Comparison of the pulse response of 27Al in sites with different quadrupole frequencies using an applied rf field of 5 mT (n 1 56 kHz).

incident on the probe circuit which itself has a


frequency response so that in detecting the signal
intensity at different frequencies, a broad line will
not only experience non-uniform irradiation but the
intensity detected per spin at a different frequency
offset will depend on this probe response, which
depends on the quality factor (Q) [13]. The width of
the frequency response decreases as the Q increases so
that typically for a Q of 100 the halfwidth of the
frequency response is about 1.5 MHz. The Q can be
reduced but the other requirements for the probehead
such as sensitivity and generation of the rf-field have
different dependencies on Q [16] which mean that
there is no single Q-value that will optimise all
requirements and some compromise has to be
reached. Hence direct FT-pulse observation of broad
spectral lines becomes impractical with pulse techniques for linewidths greater than 200 kHz. Broader
spectral lines can be reproduced provided that corrections are made for the rf-irradiation and probe
responses but this requires careful calibration and

correction. This has been most extensively done for


examining satellite transition spectra with the correction factors for an MAS probe shown in Fig. 4 [17].
The overall correction factor is then applied to the
observed intensities to get the true intensity distribution. The corrections can have a large effect on the
lineshape as is shown in Fig. 5 for a 27Al MAS satellite
transition spectrum. The raw experimental data (Fig.
5(a)) shows some structure with a hint of a singularity
at around 200 kHz, but as the intensity is significantly
attenuated by instrumental factors at higher frequencies the singularity is indistinct. However, once the
correction factors have been applied (Fig. 5(b)) the
singularity is very clearly defined and there is also
significant intensity at even higher frequencies.
Also with many spectrometers the start of the FID is
often corrupted because of various instrumental deadtimes [16] which lead to intensity problems in the
spectrum. The spectrometer deadtime is made up of
a number of sources that can be apportioned as electronic and probe. The loss of the initial part of the FID

166

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 4. Experimentally determined correction factors that need to be applied to broad spectra as a function of the offset frequency. (a) A
polynomial function of the overall correction function, (b) probehead response, (c) squared probehead response and (d) transmitter response
(Reproduced from work of Dr G. Kunath-Fandrei (Ref. [17]) with permission).

is manifest in the spectrum as a rolling baseline (Fig.


6). In the best cases the deadtime is 2ms, but even
this can still lead to severe distortion of broad spectral
lines. Baseline correction can be achieved by use of
either simple spline fits using spectrometer software
or the use analytical functions which effectively
amount to full intensity reconstructions [18], and
back prediction techniques.
Now that the basic pulse requirements have been
outlined, a practical question to ask is what is the
comparative ease with which different quadrupole
nuclei can be observed. The practising solid state
NMR spectroscopist is often confronted with a question from the enthusiastic chemist such as Well we
have been observing 27Al easily what about 95Mo?.
In comparing nuclei a first consideration is the sensitivity which is given by the receptivity g3 CII 1
where C is the natural abundance [3]. Table 2 lists
these relative to 27Al as this is a quadrupole nucleus
that is widely studied so that an appreciation of the

Fig. 5. MAS sideband intensities for 27Al in CsAlTiO4 at 7.05 T (a)


directly measured and (b) corrected for experimental factors.
(Reproduced from work of Dr G. Kunath-Fandrei (Ref. [17]) with
permission).

intensity of the signal from different nuclei can be


readily gained. A further factor that needs to be
included is that if, as is common it is only the central
transition that is being observed, the central transition
intensity for different values of I (Table 1) needs to be
taken into account. After sensitivity the width of the

Fig. 6. NMR spectrum taken at 17.64 MHz and Fourier transformed


immediately after the pulse showing the baseline roll caused by the
initial deadtime.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

167

Table 1
Some useful NMR parameters for the different transitions of non-integer spin quadrupolar nuclei
I

mz

p Dmz a

90-pulse b

Int (%)

(2) c
d q,iso

(2) d
Dq,iso

C0(p) e

3/2

1/2
3/2
1/2
3/2
5/2
1/2
3/2
5/2
7/2
1/2
3/2
5/2
7/2
9/2

1
3
1
3
5
1
3
5
7
1
3
5
7
9

0.500
0.578
0.333
0.354
0.447
0.250
0.258
0.389
0.378
0.200
0.204
0.218
0.250
0.333

30.0
60.0
25.7
45.6
28.6
19.0
35.8
28.6
16.6
15.2
29.0
25.4
19.4
11.0

3/4
2.0
9/50
0.125
3.500
15/196
0.400
1.400
4.400
1/24
0.625
0.500
2.375
5.000

3/4
0.889
9/50
0.292
1.183
15/196
0.622
0.511
2.400
1/24
0.764
0.056
1.125
2.778

3
9
8
6
50
15
27
15
147
24
54
30
84
324

5/2

7/2

9/2

Co(p)/Co(1) QIS
3
3/4
25/4
9/5
1
49/5
9/4
5/4
7/2
27/2

C4(p) f
54
42
144
228
300
270
606
330
966
432
1092
1140
168
2232

C4(p)/C4(1) QA
7/9
19/12
25/12
101/45
11/9
161/45
91/36
95/36
7/18
31/6

Here p coherence order of ^ mz transition.


90-pulse length relative to the non-selective case.
c
(2)
Here d q,iso
isotropic second-order quadrupolar shift in ppm in units of 1 h2q =3Cq2 =3n20 105 for the central transition and then relative
to this for the other transitions.
d (2)
Dq,iso second order quadrupolar broadening in units of (25 22h q h q2)Cq2/144n 0 for the central transition and then relative to this for
other transitions.
e
Here Co p 2ma 3m2 .
f
Here C4 p 2m18a 34m2 5.
b

line is important and will determine not only the


observability but also which technique should be
applied. The most important factor in determining
the linewidth in general, for a given electric field
gradient will be the quadrupole moment and these
are listed in Table 2. This factor will directly determine the width of the first-order quadrupole
broadened satellite transitions. However, as it is the
central transition that is often observed, it is secondorder broadening that will determine its linewidth
which depends on the factor Q2 a 3=4=g2I2I
12 which is again listed in Table 2 relative to 27Al.
The actual linewidth will be given by this factor multiplied by the electric field gradient. There will always
be cases where a nucleus finds itself in a cubic local
environment (i.e. n q 0) and lines from such
compounds are narrow, making them ideal for examining a particular nucleus. Although definitive statements about the field gradients for a given nucleus
cannot be made it is true that some nuclei tend to
have larger field gradients than others and this is to
some extent a reflection that the field gradient set up

by the surroundings is amplified by the Sternheimer


antishielding factor, and this also needs to be taken
into account when comparing nuclei. Hence to get
some feel for the observability of a central transition,
the second-order broadening factor should be multiplied by the Sternheimer antishielding factor (Table 2)
squared.
There is also the question of the choice of magnetic
field with higher magnetic fields giving better sensitivity. For non-integer spin quadrupole nuclei there is
the additional advantage that second-order quadrupole broadening scales inversely with the applied
magnetic field thereby decreasing the linewidth as
the magnetic field is increased. The advantage of
higher magnetic fields is shown in the simulation of
the 27Al MAS NMR spectrum of kyanite (Al2SiO5),
where at a field of 135.3 T perfect resolution of the
four sites is seen in the spectrum (Fig. 7). Typical
fields for solid state NMR are 7.059.4 T, although
14.1 T widebore magnet systems are now just becoming available and these offer distinct advantages for
observing quadrupolar nuclei.

168

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Table 2
Summary of properties of non-integer quadrupole nuclei
Nucleus

Natural
abundance (%)

n o (MHz) at
7.05 T

Relative
receptivity a

Q/fm 2b

Quadrupole
broadening c

Sternheimer antishielding
factor (g ) d

3/2
3/2
3/2
5/2
3/2
3/2
5/2
5/2
3/2
3/2
3/2
3/2
3/2
7/2
7/2
5/2
7/2
7/2
3/2
5/2
7/2
3/2
3/2
3/2
5/2
3/2
3/2
9/2
3/2
3/2
3/2
9/2
5/2
3/2
9/2
5/2
9/2
5/2
5/2
5/2
5/2
5/2
9/2
9/2
5/2
7/2
5/2
3/2
7/2
3/2
3/2

92.5
100
80.1
0.037
0.27
100
10.0
100
0.75
75.77
24.23
93.26
6.73
0.135
100
7.28
5.51
99.75
9.50
100
100
1.14
60.11
39.89
4.11
60.1
39.9
7.73
100
50.69
49.31
11.5
72.16
27.84
7.00
11.22
100
15.92
9.55
12.7
17.0
22.33
4.3
95.7
57.36
42.64
100
21.2
100
6.59
11.23

16.67
42.20
96.32
40.71
23.71
79.44
18.39
78.27
23.06
29.44
24.51
14.02
7.70
20.23
73.03
16.95
16.95
79.05
17.00
74.56
71.04
26.87
79.65
85.32
18.82
72.24
91.79
10.50
51.57
75.46
81.34
11.59
29.08
98.56
13.06
28.02
73.69
19.65
20.07
13.78
15.45
13.80
66.02
66.17
72.30
39.15
60.47
24.79
39.64
30.02
33.58

1.31
6.71 10 2
6.40 10 1
5.22 10 5
3.21 10 5
4.48 10 1
1.30 10 3
1
8.23 10 5
1.73 10 2
3.19 10 3
2.30 10 3
2.60 10 5
4.19 10 5
1.46
7.41 10 4
1.00 10 3
1.85
4.18 10 4
8.64 10 1
1.35
1.98 10 4
3.12 10 1
1.71 10 1
5.69 10 4
2.03 10 1
2.76 10 1
5.27 10 4
1.23 10 1
1.95 10 1
2.37 10 1
1.06 10 3
5.13 10 2
2.38 10 1
9.20 10 4
5.14 10 3
2.36
2.52 10 3
1.61 10 3
6.94 10 4
1.31 10 3
1.22 10 3
7.30 10 2
1.64
4.52 10 1
9.60 10 2
4.61 10 1
2.89 10 3
2.34 10 1
1.59 10 3
3.80 10 3

4.01
5.29
4.06
2.56
10.16
10.89
19.94
14.03
6.78
8.17
6.44
6.01
7.33
4.08
22
29
24
5.2
15
33
42
16.2
22.0
20.4
15.0
17.0
10.0
17.3
31.4
33.1
27.6
25.3
27.4
13.2
33.5
20.6
32
2.2
25.5
7.9
45.7
66.0
79.9
81
36
49
78.9
12.0
0.37
16.0
24.5

2.28 10 1
1.10
2.83 10 1
6.40 10 2
7.21
2.47
8.60
1
3.30
3.75
2.80
4.27
11.6
1.39 10 1
1.12
19.73
5.74
5.78 10 2
21.9
5.81
4.20
16.18
10.06
8.08
4.76
6.62
1.80
2.62
31.67
24.05
15.51
5.08
10.27
2.93
7.91
6.02
1.28
0.98
12.88
1.80
53.75
125.5
8.9
9.13
8.98
10.36
40.93
9.62
5.84 10 4
14.12
29.60

0.2
0.2
0.19
13.8
9.5
5.5
4.1
3.6
52.2
42.0
42.0
21.8
21.8
18.8
23.1
9.0 e
9.0 e
7.6 e
6.6 e
5.8 e
4.7 e

Li
Be
11
B
17
O
21
Ne
23
Na
25
Mg
27
Al
33
S
35
Cl
37
Cl
39
K
41
K
43
Ca
45
Sc
47
Ti
49
Ti
51
V
53
Cr
55
Mn
59
Co
61
Ni
63
Cu
65
Cu
67
Zn
69
Ga
71
Ga
73
Ge
75
As
79
Br
81
Br
83
Kr
85
Rb
87
Rb
87
Sr
91
Zr
93
Nb
95
Mo
97
Mo
99
Ru
101
Ru
105
Pd
113
In
115
In
121
Sb
123
Sb
127
I
131
Xe
133
Cs
135
Ba
137
Ba
9

25.2
25.2
21.9
17.0
17.0
8.7 e
80
80
85.5
52.8
52.8
47.8
26.6 e
23.0 e
20 e
20 e

28 e
28 e
162
110
110
110

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

169

Table 2 (continued)
Nucleus

Natural
abundance (%)

n o (MHz) at
7.05 T

Relative
receptivity a

Q/fm 2b

Quadrupole
broadening c

Sternheimer antishielding
factor (g ) d

3/2
7/2
5/2
7/2
7/2
7/2
7/2
5/2
5/2
3/2
3/2
3/2
5/2
5/2
7/2
7/2
5/2
7/2
7/2
9/2
7/2
5/2
5/2
3/2
3/2
3/2
3/2
3/2
9/2
7/2

92.5
99.91
100
12.18
8.30
15.0
13.8
47.8
52.2
14.8
15.65
100
18.9
24.9
100
22.95
16.12
97.41
18.606
13.629
99.988
37.40
62.60
16.1
37.3
62.7
100
13.18
100
0.72

16.67
42.73
91.90
16.35
10.07
12.51
10.31
74.63
32.95
9.28
12.18
72.12
10.33
14.46
64.07
8.66
14.61
34.28
12.18
7.65
36.40
68.50
69.21
23.64
5.4
5.87
5.22
20.07
49.09
5.85

1.31
2.93
1.62
2.00
3.17
1.10
5.67
4.14
3.90
1.07
2.50
3.36
4.38
1.56
9.87
5.55
1.05
1.47
1.26
3.62
1.81
2.51
4.33
1.90
5.41
1.21
1.45
9.49
6.98
5.22

4.01
20
5.9
63
33
26
9.4
90.3
241
127
135
143.2
247
265
358
357
280
497
336
379
317
218
207
85.6
81.6
75.1
54.7
38.6
50
493.6

2.28 10 1
1.58
1.51 10 1
41.02
18.28
9.13
1.45
43.45
700.9
2878
2478
470.9
2348
1931
338.0
2487
2134
1218
1566
1728
466.5
275.9
246.1
513.3
2042
1591
949
122.9
4.69
7038

0.2
71 e

Li
La
141
Pr
143
Nd
145
Nd
147
Sm
149
Sm
151
Eu
153
Eu
155
Gd
157
Gd
159
Tb
161
Dy
163
Dy
165
Ho
167
Er
173
Yb
175
Lu
177
Hf
179
Hf
181
Ta
185
Re
187
Re
189
Os
191
Ir
193
Ir
197
Au
201
Hg
209
Bi
235
U
139

10

10 3
10 4
10 3
10 4
10 1
10 2
10 4
10 4
10 1
10 4
10 3
10 1
10 4
10 3
10 1
10 3
10 4
10 1
10 1
10 1
10 3
10 5
10 4
10 4
10 4
10 1
10 6

73.8

Relative receptivity the relative receptivity given by g3 CII3 and normalised to 27Al where C is the natural abundance of the isotope.
Q The quadrupole moment taken from CRC Handbook, D.R. Lide (Ed.), CRC Press (1996) pp. 985.
c
Quadrupole broadening The second-order quadrupolar broadening factor which is proportional to Q2 =ga 3=4=2I2I 12 and has
been normalised to 27Al.
d
Sternheimer antishielding factor is given where meaningful calculations have been carried out for ions with closed shell structure e.g. Al 3,

Cl . Given for a lattice when possible (Schmidt et al., Phys. Rev. B, 22 (1980) 4167) and if not for free ions.
e
Atomic and Nuclear Data Tables, Johnson et al., 28 (1980) 333.
a

4. One-dimensional experiments
4.1. Static broad line experiments
Static powder patterns offer one way of characterising a material, and if spectral features can be
observed, and the line simulated then accurate
determination of the NMR interaction parameters is
possible. For quadrupole nuclei this could either be
observation of just the central transition, which would
be applicable to those sites/nuclei with larger

quadrupole interactions, or the whole satellite transition manifold. If the singularities of the satellite transition can be observed then an estimate of Cq can be
made, but an estimate of h q can be difficult. The
simplest experiment to carry out is the one pulse
experiment and despite deadtime effects as the singularities are relatively narrow features they are still
recorded. This is illustrated in Fig. 8(a) for 27Al
from a -Al2O3. For the outer satellite transition only
the first singularity is recorded but nevertheless Cq can
be estimated. Several practical points should be noted

170

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

in that significant first-order phase and subsequent


baseline correction are necessary to form this spectrum. Also all the spectral intensity that is expected
between the singularities is effectively lost in the
initial system deadtime.
A common way to overcome deadtime problems is
to form a signal with an effective time zero point
outside the deadtime, i.e. an echo. There is a huge
multiplicity of methods for forming such echoes.
Most methods are two-pulse sequences with the classic spin echo [19] consisting of 90 t 180 with refocusing at t after the second pulse, whose effect can be
explained for non-interacting spin-1/2 nuclei using a
classical vector model [3,13]. In the case when the
quadrupole interaction is present the classical description of the echo formation can no longer be used and
echo formation has to be described by quantum
mechanics with the original work done by Solomon
[20]. It is convenient to consider two cases depending
on the spacing t relative to the length of the FID (Tf).
A so-called Solomon echo forms if t p Tf , whereas
the Hahn type of echo forms if t Tf . Early calculations of the behaviour assumed the hard pulse
regime so that n1 nq , so that evolution during the
pulse is solely governed by the rf-pulse and effects of
the quadrupole interaction could be ignored. Much of
the early work is summarised in the review by Mehring and Kanert [21]. With the advent of computers,
numerical calculation of echo formation became
possible and can be readily extended to the regime
where the first-order quadrupole interaction is significant so that evolution during the pulse occurs under
both the rf and first-order quadrupole interactions.
Calculations have been made for a variety of twopulse sequences and for varying rf-field strengths
[2225]. The general conclusions of this work are
that echoes will generally be formed but that to get
quantitatively reliable information great caution has
to be exercised. For the quantitatively simplest case
weak (n1 p nq ) rf-excitation of only the central transition should be used. This is rarely the case adopted
practically but nevertheless useful structural information can still be obtained. It should also be noted that
n q can be deduced from the intensity variation as a
function of the length of the second pulse [24,25].
As already stated the main practical reason for
employing an echo approach is to overcome deadtime
problems and allow accurate determination of the

spectral lineshape. Hard rf-pulses are then preferred


for uniform excitation over broad lines. Our own work
has tended to use an echo sequence with phase cycling
first proposed by Kunwar et al. [26], which combines
quadrature phase cycling with further cycling
designed to cancel direct magnetisation (the remaining FID) and ringing effects (Fig. 9). The rotation
produced by the second pulse in the two-pulse echo
experiment is not critical. In practice, the best choice
is with the second pulse twice the length of the first,
with the actual length a trade-off between sensitivity
and uniformity of the irradiation. The results from a
spin-echo for 27Al from a -Al2O3 are shown in Fig.
8(b). The excitation still has only sufficient bandwidth
to allow detection of the first two singularities of the
satellite transitions, but compared to the one-pulse
experiment, records much better the intensity of the
satellite transitions between the singularities, although
it is still below that expected theoretically. In recording echoes there is an important practical consideration in that the point of applying the echo is to move
the effective t 0 position for the FID to being
outside the region where it is corrupted. However in
order that phasing problems do not re-emerge, the
data sampling rate used should be sufficient to allow
this point to be accurately defined. If T2 is sufficiently
short such that an echo time can be used that allows
the whole echo (both before and after the maximum)
to be accurately recorded without an unacceptably
large loss of intensity, there is no need to accurately
define the new t 0 position. Fourier transformation
of the whole echo (which effectively amounts to integration between ^ ) followed by magnitude calculation removes phasing errors producing a pure
absorption lineshape with the signal-to-noise 2 larger
than that obtained by transforming from the echo
maximum.
Once lines become broader than about 200 kHz it
becomes difficult to record the resonances accurately
using direct pulse techniques. To overcome these
difficulties several approaches have been adopted
based on the philosophy that although the line is
broad it can be recorded using a series of narrow
banded experiments that overcome the distortions
introduced into a broad spectrum. One of these
approaches is to carry out a spin-echo experiment
using relatively weak rf pulses, recording only the
intensity of the on-resonance magnetisation and

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 7. Simulations of the 27Al MAS centreband of kyanite


(Al2SiO5) as a function of the applied magnetic field.

repeating the experiment at many frequencies to map


out the lineshape. This approach has been successfully
used in a series of studies, most recently for example
on high temperature superconductors [27], and 91Zr in
ZrO2 from which it was shown that the polymorphs
could be readily distinguished [28] (Fig. 10). This
approach works but is extremely tedious because
each frequency step requires accurate retuning of the
probe. An alternative is to sweep the main magnetic
field. There are several examples of sweeping the
main magnetic field for solids dating from the earliest
days of NMR but only a limited number reported
using superconducting magnets [29,30], with a recent
example being for 27Al in a -Al2O3 [31]. There is no
doubt that this approach is universally applicable but
in all the work to date the superconducting magnets

171

that have been used are not suitable for high resolution
liquid state NMR spectroscopy, and hence this has
been regarded as a specialist experiment. However it
has been recently demonstrated that it is possible to
have a single NMR spectrometer that is capable of
both conventional high resolution spectroscopy and
also field sweep operation at relatively little extra
cost [32]. The field sweep in the 7.05 T instrument
described by Poplett and Smith [32] is limited to
^0.5 T which is sufficient to cover many broad
lines. Control of the field is completely automated
and integrated with the pulse programme. As with
the stepped frequency experiment, relatively soft
pulses are applied, and although strictly the on resonance part of the magnetisation should be used,
experience shows that using the spin-echo intensity
directly accurately reproduces the lineshape. For
comparison, the field-sweep 27Al spectrum of a Al2O3 is shown in Fig. 11(a). Two major differences
are noticeable compared to both the one pulse and
echo spectra (Fig. 8). First the intensity between the
singularities is much higher and much more closely
matches theoretical expectation, and second as there
is essentially no bandwidth limit on the experiment
the outer singularities are easily recorded. The total
frequency width of this lineshape is 1.4 MHz but
this approach can be applied to much broader lines as
is demonstrated for 27Al in Al3Zr where the frequency
width is well beyond 3 MHz (Fig. 11(b)) and three
sites can be identified with values of Cq of 3.84,
9.05 and 11.33 MHz [33]. The sensitivity of the satellite transitions to spectral parameters is well illustrated by the intermediate Cq site having h q 0.03
which is easily detected from the shape of the singularities. There is little doubt that this field sweep
approach could be extremely useful as an alternative
way of examining nuclei, especially low-g nuclei
where narrowing techniques do not yet offer real
widespread opportunities for improved resolution.
4.2. Magic angle spinning observation of the central
transition
Magic angle spinning (MAS) is the single most
used technique in solid state NMR [34]. Its
popularity stems from its success in improving
resolution by removal of the anisotropic
broadenings. In an isotropic powder it is the

172

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 8. Static

27

Al NMR spectra from a -Al2O3 using (a) single pulse and (b) spin-echo acquisition.

(3cos 2 u 1 h sin 2 u cos 2u ) angular terms that


lead to the broadening of the powder spectrum,
thus degrading spectral resolution. Spinning the
sample container at an angle b to Bo rapidly at a
frequency n r imparts a time dependence to the interactions so that the Hamiltonian in the laboratory
frame becomes
Hlab t H H*t

The laboratory frame can be rotated into the rotor


frame by the Euler angles (0,b ,v rt) and the rotor
frame into the PAS of a particular crystallite by
(f ,u ,c ). Wigner rotation matrices can be used to
make this double transformation from the PAS to
the laboratory frame to give the time-dependent
laboratory Hamiltonian H(t), which is limited to

secular terms [4]. Then in Eq. (6)


H Aiso 31=2 A2 FI; m3cos2 b 1=421=2
 3cos2 u 1d hdsin2 ucos2 c

where A2 is a constant, F is a function of I and m, and


d and h are the anisotropy and asymmetry of the
tensor respectively. H*(t) contains sinusoidally timedependent terms so that rapid sample spinning will
average H* to zero leaving H. For a powder with all
crystallite orientations present, all values of u and c
will occur but they have a common modulation factor
of (3cos 2 b 1). By choosing b 5444 0 8 00 , the socalled magic angle, the Hamiltonian is reduced to the
isotropic part just as in a liquid. Quadrupolar nuclei
provide a particular challenge for study by MAS as
the first-order broadening that occurs causes the (^m,

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

173

Fig. 9. Spin-echo sequence where the second pulse has twice the tip angle of the first with extended phase cycling to give both quadrature
selection and cancellation of ringing and direct magnetisation effects.

^ (m 1)) transition to be spread over (2m 1)n q.


Until recently this was regarded as too broad for
useful application but now offers some exciting possibilities in satellite transition spectroscopy (vide infra).
However the significant second-order perturbation of
the central transition produces isotropic and anisotropic terms, the latter being proportional to the second
and fourth Legendre polynomials P2(cos u ) and
P4(cos u ). There is no direction that will allow both
P2(cos u ) and P4(cos u ) contributions to be simultaneously completely removed by spinning around a
single axis. MAS removes the P2(cos u ) term and
causes partial averaging of the P4(cos u ) term producing spectra (Fig. 1(d)) that are a factor of 34
(depending on h ) narrower than static spectra [35].
The residual second-order effects under MAS can
result in characteristic lineshapes whose widths, like
the static central transition are proportional to n2q =no : If
such lineshapes can be resolved and simulated, as
with static lineshapes, not only can the isotropic
chemical shift be deduced but so can the quadrupolar
parameters. The residual second-order quadrupolar
width Dn2
q under MAS Eq. (8) determines the spinning rate necessary to cause narrowing.
2
2
2
Dn2
q Cq II 1 3=46 hq =224no I

 2I 12 :

The maximum value of Cq that can be narrowed


because n r must exceed Dn2
depends on Dn2
q
q
for efficient line narrowing to occur. At nr
Dn2
q the centreband and spinning sidebands (which
are not simple replicas of the centreband) strongly

overlap giving a complex lineshape. This is shown


for 23Na in Amelia albite (NaAlSi3O8) where below
4 kHz an extremely complicated lineshape is obtained
which is very difficult to simulate (Fig. 12). Above
4 kHz there is a recognisable centreband but even up
to 11 kHz the lineshape does not have a classical
second-order quadrupole form and it is only by spinning at speeds in excess of 11 kHz that a lineshape
that could be readily simulated is observed. This can
be compared to the 11 kHz spectrum on the same
sample at 14.1 T where there is a very clear distortion-free lineshape (Fig. 12(b)). The maximum Cq that
can be narrowed for I 5/2 (assuming h 0) is
8.82(n rn 0) 1/2 which has encouraged the use of faster
MAS at higher applied magnetic fields. The available
spinning speeds and applied magnetic fields have
greatly increased the range of sites which can be
observed from quadrupolar nuclei and the importance
of fast MAS in improving the clarity of MAS spectra
from quadrupolar nuclei cannot be underestimated.
For example, for 27Al the maximum Cq that can be
narrowed varies from 4.6 MHz using 3.5 kHz at
7.05 T, which was typical of the conditions available
15 years ago, to 12.1 MHz at 9.4 T using 18 kHz
which are fairly standard conditions now available,
to 23.2 MHz that could be narrowed using 32 kHz at
18.8 T. Two practical points should be made here.
First, even if the residual second-order quadrupolar
linewidth is exceeded so that n r is only just greater
then although a distinct centreband is
than Dn2
q
observed the lineshape can be noticeably different
from that calculated in many simulation packages
which are written for the infinite spinning speed

174

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 10. Static 91Zr frequency stepped NMR spectra showing the
lineshapes of the central transition of the tetragonal (t), monoclinic
(m) and cubic (c) polymorphs of ZrO2.

limit. Secondly, although the line will narrow even if


the magic angle is not accurately set, even quite small
deviations of angle can lead to changes in the lineshape that will again make it impossible to accurately
simulate the centreband.
From Fig. 1(d) it is immediately apparent that the
centre of gravity of the line does not occur at the
isotropic chemical shift. Second-order quadrupolar
effects will shift it by
2
n2
q;iso m 3Cq a 9mm 1 3

 1 h2q =3=40no I 2 2I 12 :

When only a featureless, usually asymmetric


line is observed the peak position is often quoted
and this can be significantly different from the
isotropic chemical shift and this effect can be so
large that a peak is shifted into an isotropic chemical
shift range usually associated with another structural
unit, and there are numerous examples in the literature

of mis-assignment based on this effect. For featureless


resonances the field dependence of the peak position,
or more correctly the centre of gravity of the central
transition, can be used to estimate the strength of the
quadrupolar interaction from a plot of peak position
27
(in ppm) against B2
o ; shown for the Al MAS NMR
from the mineral kaolinite in Fig. 13, producing a
good straight line. The gradient is proportional to
Cq2 1 h2=3
q ; whose square root is sometimes termed
the second-order quadrupolar effect parameter. Extrapolation to B2
o 0 simulates carrying out the NMR
experiment at very high magnetic field where secondorder quadrupolar effects are suppressed and the peak
position becomes diso;cs :
In materials where there is either atomic and/or
structural disorder a range of interactions will be
present and distinct second-order quadrupole structure
is usually lost. The actual lineshape depends on which
interaction is dominant, with a distribution of isotropic chemical shifts producing a Gaussian line,
whereas a distribution of quadrupole interactions
produces a distinctive tail to low frequency [36]
(Fig. 14). Direct extraction of Cq from such a lineshape, especially at a single magnetic field is essentially impossible. Measurement of the shift of the
centre of gravity and the linewidth as a function of
applied magnetic field can give some useful information but such an approach should be adopted
cautiously. In particular, it is important that a sufficient spinning speed is used. In both 23Na and 27Al
NMR in glasses even at modest spinning rates (e.g.
5 kHz) spectra with good S/N are obtained. However,
as the spinning speed is increased in glasses it is often
observed that the linewidth increases and the centre of
gravity changes. This can be readily understood
because of the range of Cq that is present. As shown
in Eq. (8) only those sites with Dn2
q below n r will
narrow, as the spinning speed is increased, lines from
larger Cq sites also narrow and these have larger Dn2
q ;
hence the broader line and also a greater second-order
quadrupole shift and consequent change in the centre
of gravity. This is illustrated for two albite glasses at
7.05 T (Fig. 15), with the dry glass having a much
broader distribution of Cq The rapid fall in the linewidth to a minimum at n r 3 kHz corresponds to
where MAS averaging first becomes effective. As
the spinning speed increases to 18 kHz the linewidth
increase from 2.7 to 3.7 kHz. Over the same range of

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 11. Field sweep NMR spectra of 27Al in (a) a -Al2O3 and (b) Al3Zr.

175

176

Fig. 12.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

23

Na MAS NMR spectra from crystalline Amelia albite (NaAlSi3O8) at (a) 7.05 T up to 16 kHz and (b) at 14.1 T at 11 kHz.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

177

Fig. 12. (Continued).

spinning speeds there is a decrease in the centre


of gravity of the line from 19 to 36.5 ppm. It
is probable that much of the early 27Al and in
particularly 23Na MAS NMR work which was
performed at moderate magnetic fields with relatively modest spinning speeds needs to be treated
with caution as probably only a subset of all the sites

in the glass were recorded in these MAS NMR


spectra.
Quantification of MAS NMR spectra of quadrupolar nuclei posed many problems 1015 years ago,
with a number of 27Al studies on both crystalline
and glassy powders noting that the integrated signal
intensity failed to agree with the known aluminium

178

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 13. Position of the centre of gravity of the 27Al MAS NMR spectra from kaolinite (Al2Si2O5(OH)4) plotted against Bo2.

distribution in different sites and with the total aluminium content [37]. However both the understanding
and the available experimental technology have
improved markedly over the last few years. A combination of using the correct experimental conditions,
together with higher magnetic fields and faster MAS
rates has removed many of the difficulties that gave
rise to the idea of NMR invisible aluminium.
Using small flip angle pulses so that work is
performed in the linear regime (this condition presents
a minor problem in that with typical values of B1 of
75 kHz for 27Al this would require only a 0.4 ms pulse
to be used!) means that sites with widely differing Cq
produce the same intensity response, but even after
this precaution is taken, quantitative interpretation
of quadrupole perturbed NMR spectra needs to be
approached cautiously. One of the sources of error
is that it is often only the prominent centreband that
can be recorded accurately. The intensity will be
distributed between the centreband and the spinning
sidebands. The spinning sidebands could be integrated
as well but with more than one transition for quadru-

pole nuclei, partitioning this magnetisation can be


problematic, and full scale simulation of the complete
manifold can be time consuming and requires that a
good S/N be obtained on the spinning sidebands as
well. A more pragmatic approach is to record the
centreband and to correct the observed intensity by
a factor that corresponds to the amount of magnetisation that appears in the centreband [38]. For the (1/2,
1/2) transition, this depends on the parameter
n2q =no nr while for the satellite transitions it is a function of 1 2mnq =nr ; with both being only a weak
function of h q (Fig. 16(a),(b) respectively). Once
these factors are taken into account, even for sites
with widely differing Cq values, the measured integrated intensity can be corrected to give the actual
distribution in compounds to a high degree of accuracy [3840]. Y3Al5O12 makes an interesting case
study which has AlO4 and AlO6 sites in the structure
in the ratio of 3:2 but integration of the centrebands of
the 7.05 T, 7 kHz 27Al NMR spectrum gives a ratio of
1 : 1 (Fig. 16(c)). Here, for the AlO4 site, the centreband effectively only has intensity from the central

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

179

Fig. 14. The effect of a distribution in the isotropic chemical shift


and the quadrupole coupling constant on the central transition MAS
lineshape. For these simulations the average Cq is 2 MHz, h q 0
and the isotropic chemical is 0 ppm. The Larmor frequency is
80 MHz so A 2343.8 Hz. (a) No distribution, (b) Gaussian distribution of the isotropic chemical shift of FWHM 400 Hz (5 ppm)
or 0.17 A, (c) Gaussian distribution of FWHM 0.8 MHz and (d)
Gaussian distribution of both the isotropic chemical shift and Cq
which are the same as in (b) and (c). (A II 1 3=4n2q =n0 ).

transition, with 78% of this central transition intensity


present and no contribution from the outer transitions
(Table 3). However for the AlO6 site all of the central
transition intensity is in the centreband which also has
a significant contribution from the outer transitions
(Fig. 16(c), Table 3). Once these factors are corrected
for, the centreband intensity comes out to be 1.44 : 1,
which is close to the 1.5 : 1 expected from the crystal
structure.
4.3. Magic angle spinning and spinlocking
It is well known that rotationally induced spinlocked magnetisation can appear when an rf-field is
applied to a quadrupolar nucleus in conjunction with
(magic angle) spinning [41]. As a result of the spinning, the orientation of the quadrupole nucleus
changes continuously. Hence, the splitting between
the energy levels which depends on the crystallite

Fig. 15. Changes in the position of (a) the centre of gravity (d cg) and
(b) linewidth (fullwidth half maximum) of 23Na MAS NMR spectra
from two albite (NaAlSi3O8) glasses.

orientation will vary and level crossings between the


different spin levels will occur. During one rotor
period spin density is transferred from the outermost
spin levels to a spinlocked state and back again and
vice versa. Depending on the orientation of the quadrupole tensor with respect to the spinning axis there
are either 2 or 4 level crossings occurring per rotor
period. There are three spinlock regimes depending on
the so-called adiabaticity parameter a n21 =nq nr : For
a 1 the level crossings are adiabatic and spin
density will be transferred from one level to another.
An MAS quadrupole spinlock experiment will show
an oscillatory behaviour with either two minima per
rotor period or two maxima depending on the initial
state of the density matrix, respectively r (0) Cx and
r (0) Iz [41]. For a p 1 the change is sudden and no

180

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 16. Theoretical plots of the fraction of magnetisation in the centreband as a function of n q and n r for (a) the central transition and (b) the
satellite transitions. (c) The experimental MAS NMR spectrum form Y3Al5O12 at 7.05 T and 7 kHz MAS together with simulation of (d) the
complete spectrum as well as components from the AlO6 (e) (1/2, 1/2), (f) ( ^ 3/2, ^ 1/2) and (g) ( ^ 5/2, ^ 3/2) transition and (h) for
the (1/2, 1/2) transition of the AlO4. (Taken from Ref. [38] with permission of Academic Press).

transfer of population occurs, and in this case the


normal spinlock behaviour is expected. In the intermediate regime the transfer of spin density is not efficient but still occurs, giving rise to very short effective
spin-lock times. For 27Al when Cq 2 MHz, spinning
at 5 kHz and an rf-field strength of 60 kHz gives a
2.4, which means it is in the adiabatic regime. So for
effective spinlocking one should choose the spinning

speed and the rf-field strength in such a way that one is


well into the adiabatic or sudden regime. The rotationally induced level crossings adversely affect the
performance of various NMR experiments such as
cross-polarisation and nutation NMR. It also enables
the TRAPDOR effect (vide infra). A more detailed
and thorough discussion of MAS and spinlocking of
quadrupole nuclei has been provided by Vega [41].

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

181

Fig. 16. (Continued).

4.4. Magic angle spinning observation of satellite


transitions
Although the majority of NMR on non-integer spin
quadrupolar nuclei has concentrated on the central
transition, routine observation of the satellite transition, in fact, plays a central role in MAS. The most
convenient method for setting the magic angle uses
compounds containing non-integer spin quadrupole
nuclei in nominally cubic environments but where
defects cause some small quadrupole interaction at
some sites [42]. For MAS, a small deviation of the
angle db from the exact value causes a residual broadening of 3n qcos b sin b db . Hence the width of the
spinning sidebands is sensitive to angle. Nominally
cubic compounds are usually chosen to set the angle

as the first-order quadrupole broadening of the satellite transition is not too severe and hence can be
narrowed over a small but not too restrictive range
of angles about the magic angle. Usually the bromine
resonance of KBr is used, but for probes that will only
tune to lower frequencies, then 85Rb in RbCl is also
useful. However, for observation of the non-central
transitions in many compounds much more accurate
setting of the angle is required which demands the use
of compounds with larger Cq that are consequently
more sensitive to the angle. Essentially any compound
that has a quadrupole interaction can be used but
compounds which have been more widely used practically are 23Na in NaNO3 and the favourite of the
authors is 27Al in Y3Al5O12. The AlO6 site has a
small Cq that allows setting of the angle better than

182

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Table 3
Intensity correction factors for different transitions for the 27Al
MAS centrebands of Y3Al5O12 at 7.05 T spinning at 7 kHz corresponding to Fig. 15

n q (KHz)
d cs, iso (ppm)
Experimental relative
integrated intensities of the
centrebands
(1/2, 1/2) transition
n q2/n on r
Inherent intensity
Fraction contributed to the
centreband
( ^ 3/2, ^ 1/2)
1 2mnq =nr
Inherent intensity
Fraction contributed to the
centreband
Intensity contributed to the
centreband
( ^ 5/2, ^ 3/2)
1 2mnq =nr
Inherent intensity
Fraction contributed to the
centreband
Intensity contributed to the
centreband
Total contributed intensity to
centreband
Corrected centreband intensity

Site
AlO4

AlO6

900
76.0
1

90
0.8
1

1.48
1
0.78

0.015
1
1

257
1.78
0

25.7
1.78
0.045

0.08

514
1.11
0

51.4
1.11
0.304

0.039

0.78

1.12

1.28(1.44)

0.89(1)

when using KBr, and then, if required the larger Cq at


the AlO4 site allows even more accurate setting of the
angle. The other advantage is that as the AlO4 site
shows a second-order quadrupole lineshape it
provides a check that the pulse length chosen is not
distorting the lineshape. With this type of work the
probe manufacturers have to seriously consider their
angle setting mechanisms. In current probes these are
too coarse for routine, reproducible setting of the
angle for compounds with relatively large quadrupole
interactions.
Observation of the satellite transitions under MAS
has distinct advantages. If the quadrupolar interaction
is small but non-zero the second-order quadrupolar
structure of the central transition can be obscured by
other broadening mechanisms. However, if the
powder pattern of the outer transitions can be recorded

Cq and h q can be deduced. As detailed before this can


be done on static samples but with fast, accurate (
0.005) MAS this pattern splits up into a set of sharp
sidebands that closely follow the powder pattern
envelope. The MAS approach is less susceptible to
initial deadtime problems (Section 3) and provides
sensitivity advantages. The sidebands can be simulated [43,44] to deduce Cq and h q which on conventional pulse spectrometers for solids this method is
applicable for values of Cq up to 2 and 3.3 MHz for
spins 3/2 and 5/2 respectively [43]. For amorphous
solids the change in the extent of the sideband manifold and the shape of the envelope can be used to
estimate the average Cq and the distribution. The
work of the group of Jaeger has greatly extended the
practical implementation and application of this technique with a comprehensive review given in Ref. [45].
The main practical requirements are to have fast,
stable spinning precisely at the magic angle and accurate probe tuning. Any variation of the spinning speed
or mis-setting of the angle will degrade the resolution
seen in the sidebands. In accumulating the data the
corrections necessary because of rf-irradiation and the
probe, can be made by semi-theoretical modification
for smaller Cq values, or for larger Cq values by
measuring the spectrum at several frequencies and
combining these after applying correction factors.
As single pulse excitation is used there are also
often baseline distortions present but provided the
spinning sidebands are separated, so that intermediate
baseline can be observed, normal spline fitting with
spectrometer software is adequate. Of course there is
no reason why the field sweep method outlined before
for accumulating broad static lines cannot be used for
such MAS satellite spectra with it being convenient to
make the step size in the field equal to n r.
What gave real impetus to satellite transition spectroscopy was the work of Samoson [46] who showed
that both the second-order quadrupolar linewidths and
isotropic shifts are functions of I and m (Table 1).
Some combinations of I and m produce smaller
second-order quadrupolar effects on the satellite
lines than for the central transition, thus offering better
resolution and more accurate determination of diso;cs :
The two cases where there are distinct advantages of
this approach over using the central transition are ( ^
3/2, ^ 1/2) for I 5/2 and ( ^ 5/2, ^ 3/2) for I 9/2.
The ( ^ 3/2, ^ 1/2) transition for I 5/2 is likely to be

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201
27

of most use, being applicable to nuclei such as Al


and 17O, and providing only 30% of the width and
12.5% of the second-order isotropic quadrupolar
shift compared to the central transition. 27Al has
been the focus of much of the satellite transition
work, but two examples from 17O are those for boehmite [44] and SiO2 [47]. 27Al has been applied to a
range of compounds, including ordered, crystalline
materials, atomically disordered solids, and amorphous solids. A general problem in 27Al NMR spectra
from highly disordered solids is that the centreband
peaks tend to be broad and asymmetric with often
extensive overlap of resonances that makes the
complete lineshape difficult to simulate. However
the satellite ( ^ 3/2, ^ 1/2) transition sidebands for
I 5/2 are better resolved and just as importantly
significantly more symmetric as second-order quadrupolar broadening is reduced, allowing more
straightforward deconvolution of the spectra [48].
Even for nuclei where the satellite transitions are
broader than the central transition (Table 1), if their
second-order quadrupolar structure can be observed,
they act as an independent check of quadrupolar parameters deduced from the central transition [45]. The
m-dependence of the shift for a particular value of I is
similar to carrying out the experiment at a different
magnetic field, because for each transition there is a
different net isotropic shift, and this allows the isotropic chemical shift to be determined from a single
spectrum. The larger shift of the satellite transitions
for I 3/2 has been extremely useful in identifying
different boron species using 11B NMR in borate
glasses [45].

183

At 43.5 the linewidth is independent of h q. For h q


0, spinning at 36 and 75, the line consists of an
extremely intense narrow component with a broader
component, so that although the total linewidth is not
at its narrowest at this point, in a complex material the
resolution provided in practice by the intense component may be an advantage. This advantage becomes
increasingly less evident for higher h q. Calculations
have also been performed for the sidebands of this
central transition formed by spinning off the magic
angle [51,52]. If the quadrupolar interaction is dominant, and dipolar and chemical shielding effects can
be neglected, VAS can be effective, but as these interactions are not normally negligible then VAS is useful
only in a limited range of applications.
4.6. Double angle spinning
Although the quadrupolar information is itself
useful, often it is desirable to produce better resolution
by complete removal of second-order quadrupolar
effects. As spinning about a single axis is unable to
remove P2(cos u ) and P4(cos u ) effects simultaneously, a more complex time dependence needs to

4.5. Variable angle spinning


Spinning at any angle will introduce a scaling of the
NMR lineshape and for the central transition, when
the second-order quadrupolar broadening is dominant,
there are angles that are more efficient at reducing
second-order effects than 54.7, although there is no
single angle at which second-order broadening can be
completely eliminated. This approach has been
termed variable angle spinning (VAS) [49]. When
second-order effects dominate, the total width of the
centreband of the central transition has its minimum
width in the range 6070, depending on h q with the
line being around twice as narrow as under MAS [50].

Fig. 17. 17O DOR NMR spectra of siliceous zeolite Y at three static
applied magnetic fields. (* indicate spinning sidebands). (Taken
from Ref. [65] with permission of the American Chemical Society).

184

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

be imposed upon the sample. The direct solution to


this problem is to make the spinning axis a continually
varying function of time. A scheme that has been
proposed to achieve this uses a spinning rotor (termed
the inner rotor) which moves bodily as a function of
time by enclosing it in a spinning outer rotor so that
the axis of the inner rotor describes a complicated but
continuous trajectory as a function of time [53,54].
This is termed double angle rotation (DOR) and the
second-order quadrupolar frequency of the central
transition experiences a double modulation in the
laboratory frame (similar to the single modulation of
MAS Eq. (7)) of the form
2
t hCq2 =8I2I 1no K0 I; m; hq
Hq;lab

h2q 3=283cos2 u 1
21:5=7cos 2fsin2 u3cos2 b1 13cos2 b2 1
 K1 I; m=4 318 h2q =1120
 35cos4 u 30cos2 u 3
9hq =7cos2fsin2 u7cos2 u 1
3h2q =32cos4fsin4 u
35cos4 b1 30cos2 b1 3
 35cos 4 b2 30cos2 b2 3K2 I; m=64
terms / cosv1 t g2

10

where
K0 I; m; hq a 9mm 1 33 h2q =5; 11
K1 I; m 8a 36mm 1 15

12

and
K2 I; m 6a 34mm 1 13:

13

b 1 and b 2 are the angles between the outer rotor and


the magnetic field and the angle between the axes of
the two rotors respectively, and f and u describe the
orientation of the principle axis system of a crystallite
in the inner rotor. b 1 and b 2 can be chosen so that
P2(cos b 1) 0 (54.74) and P4(cos b 2) 0 (30.56 or
70.15) [54,55]. DOR works well if the quadrupolar

interaction is dominant and the sample is highly crystalline with some extremely impressive gains in
resolution observed [56,57]. The technique is
mechanically ingenious with the stable spinning of
such a system limited to certain ranges of the ratio
of the two spinning speeds [58]. One of the major
limitations is the relatively slow rotation speed of
the large outer rotor. The slow speed can lead to difficulties in averaging strong homogeneous interactions
and produces many closely spaced sidebands in the
spectrum. In disordered solids where there is a distribution of chemical shifts, quite broad sidebands can
result that may coalesce at the slow rotation rates
used. A commonly used trick is to recognise that the
phase of the sidebands depends on g 2, the orientation
of the outer rotor in the magnetic field. If expansions
of expressions for the sidebands are examined it turns
out that for even-order sidebands the phase of the
signal is unchanged but that odd-order sidebands are
inverted between the g 2 0 and g 2 p positions
[59]. Hence if the acquisition of the signal is triggered
synchronously at these two positions for consecutive
transients then the odd-order sidebands are cancelled,
effectively doubling the spinning speed. Currently the
maximum actual spinning speed that can be routinely
obtained in the latest system with active computer
control of the gas pressures is 1500 Hz and
undoubtedly the technology associated with the technique will continue to improve leading to increased
spinning speeds and thus expanding the application of
the technique. The other main problems with DOR are
that, as the rf coil encloses the whole system, the
filling factor is small which leads to sensitivity
being low, and the large coil size also means that
the rf that can be generated is quite low and also
that double tuning for cross-polarisation is difficult
although such an experiment has been performed [60].
Simulation of the complete DOR spectrum (centreband plus the spinning sidebands) will yield the NMR
interaction parameters [61,62]. However, it is most
usual to perform the experiment to give improved
resolution and simply quote the measured peak position which appears at the sum of the isotropic chemical and second-order quadrupole shifts. DOR
experiments at more than one applied magnetic field
[6365] (Fig. 17) will allow these different contributions to be separated and hence provide an estimate of
the quadrupole interaction. This approach is similar to

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

that using the field variation of the centre of gravity of


the MAS centreband but has the advantage that the
narrower, more symmetric line makes determination
of the correct position more precise. For experiments
carried out at two magnetic fields where the Larmor
frequencies are n 01 and n 02 for the measured DOR
peak positions (in ppm) at the two magnetic fields
of d dor1,2 then

diso;cs n201 ddor1 n202 ddor2 =n201 n202

14

and
3a 3=4=40I 2 2I 1Cq2 1 h2q =3
n201 n202 ddor1 ddor2 =n201 n202 :

15

In practice DOR requires a specialised, complex


probe. There are cases where the significantly
enhanced resolution obtained has provided the key
to understanding the NMR spectrum. However the
expense of the probehead and small range of
compounds where significant narrowing is observed
means that the technique has only found limited application to date.
5. Multiple resonance
In principle, multiple resonance methods such as
cross-polarisation (CP), Rotational Echo Double
Resonance (REDOR), and Transferred Echo Double
Resonance (TEDOR) can be applied to systems involving quadrupolar nuclei. There are, however, certain
aspects that are different from the application of these
experiments to spin-1/2 nuclei. These may prohibit
the straightforward application of these experiments
to cases where quadrupolar nuclei are involved, or
they can alter the interpretation of the results. There
are also benefits however, as illustrated by the Transfer of Populations by Double Resonance (TRAPDOR)
experiment. This does not work on spin-1/2 systems
but it is easily implemented and can provide qualitative information about the structure of materials.
5.1. Cross-polarisation
The main difference between cross-polarisation
(CP) involving spin-1/2 nuclei only and systems
with one or two quadrupolar nuclei is the altered

185

HartmannHahn match condition. The Hartmann


Hahn matching condition is fulfilled when the energy
splitting in the rotating frame of both nuclei is the
same. Whereas with spin 1/2 this is simply the same
as the rf-field strength in kHz, with quadrupole nuclei
it is the nutation frequency. This will pose a problem
in the intermediate rf-field strength range of n q/n 1
0.051. In this case there is a whole range of nutation
frequencies for the quadrupole nucleus, ranging
between n 1 and (I 1/2)n 1, and only a small fraction
of the nuclei will have a HartmannHahn match with
the other nucleus and signal to noise will be poor. In
contrast, there will always be nuclei for which the
HartmannHahn match holds, which can be an advantage. As an example, if one wanted to cross-polarise
from 1H to 27Al, and the aluminium nucleus had a
quadrupole coupling constant of 2 MHz, then n q
would be 300 kHz. To perform CP in the strong rffield limit, the rf-field on both 1H and 27Al should be
larger than n q, which is not feasible. For the weak rffield limit, n 1 for 27Al should at most be 15 kHz and
the 1H rf-field should then equal 45 kHz. This might
lead to very short T1r for both nuclei, which does not
favour the CP process. Finally, the intermediate range
will give a poor signal-to-noise as a result of the
limited number of 27Al nuclei for which the match
condition holds.
MAS is usually applied to increase spectral resolution. This has a pronounced effect on the CP dynamics
for quadrupolar nuclei [41]. Bearing in mind the
previous discussion on MAS and spinlocking, and
the fact that for most practical cases the weak to intermediate rf-field strength limit applies, one should aim
to use as high spinning speeds and relatively low rffields so as to come into the sudden spin-lock regime
(see Section 4.3). To circumvent the problem of the
time dependency of the effective quadrupole interaction with MAS, variable angle spinning can be
used where the cross-polarisation takes place at a
spinning axis angle of 0 with the magnetic field
and subsequent detection of the signal at the
magic angle. The only limitation for this technique
is that the time needed for the spinning axis to change
orientation may prevent its use for samples with short
T1 values.
Fig. 18 shows an example of a static cross polarisation experiment from 27Al to 17O in a -alumina [66],
which shows a huge signal gain of the CP spectrum

186

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

the perturbing S spin 180 pulse. By measuring a


SEDOR curve as a function of echo time or pulse
position one can determine the IS distance through
the value of the dipolar coupling constant using:
D

Fig. 18. Comparison between (a) 27Al- 17O cross-polarisation spinecho experiment and (b) 17O spin-echo experiment. The same
number of scans and the same recycle delay of 10 s were used for
both spectra. (Taken from Ref. [66] with permission of Elsevier
Science).

compared with the direct excitation 17O spectrum.


Only very weak rf-fields were used in this case and,
as both nuclei are spin-5/2, the rf-field strengths were
the same (10 kHz) for both nuclei.
5.2. SEDOR, REDOR and TEDOR
The Spin Echo Double Resonance (SEDOR) [67
69], REDOR [7072] and TEDOR [7274]
experiments are all multiple resonance experiments
that exploit effects dependent on the heteronuclear
dipolar coupling interaction. With these experiments
it is possible to derive qualitative information, i.e. are
spins I and S close to each other, or quantitative information, the IS distance. First of all the general principles behind these experiments and then effect of the
quadrupole coupling on them will be discussed.
First of all, in the SEDOR experiment [6769], in
which the sample is static, an echo pulse sequence is
applied to one nucleus, say spin I (Fig. 19(a)). The
echo refocuses heteronuclear dipolar coupling and the
chemical shift anisotropy (CSA). During the echo
period a single 180 pulse is applied to the other
nucleus, say spin S. This inverts the sign of the dipolar
coupling which perturbs the dipolar refocusing
process and consequently, the echo intensity is diminished. The amount of signal intensity that is lost (the
SEDOR fraction) depends on the magnitude of the I
S dipolar coupling, the echo time, and the position of

1 g1 gs m0

in Hz
3
4p
2p rIS

16

The magnitude of the dipolar coupling that can be


readily determined by this method is limited by T2.
The REDOR [70,71] experiment is in principle a
SEDOR experiment in combination with MAS. Spinning the sample greatly enhances sensitivity and resolution and allows the echo to be measured for longer
times. There are various pulse schemes for REDOR.
In one of them a rotor synchronised echo sequence is
applied to spin I which is detected after a time 2t
equalling an even number of rotor periods (Fig.
19(b)). Two dephasing 180 pulses per rotor period
are applied on spin S, ensuring that the sign of the
dipolar coupling at the start of each rotor period is
the same so that the dipolar dephasing adds up for
each rotor period. In contrast to the static SEDOR
experiment, where the echo was necessary to refocus
the CSA and dipolar dephasing, the REDOR experiment employs MAS which induces rotational echoes
which will take care of the refocusing of the CSA and
the dipolar interaction. The refocusing 180 pulse is
still needed as it ensures the detection of an in-phase
signal. The dipolar coupling can be obtained by
measuring the intensity loss as a function of the
number of rotor periods (echo time) or as a function
of the position of the dephasing 180 pulses on spin S.
Dipolar couplings as small as 25 Hz have been
measured using the REDOR experiment [75].
REDOR, like SEDOR, is also a difference experiment
where the attenuated signal is subtracted from the full
echo to obtain the REDOR fraction.
Quantitative distance information is easy to extract
for isolated 2 spin systems. This often requires specific labelling of samples. The interpretation of the
results becomes much more complex when more
spins are involved, e.g. IS2, IS3 or IS4 spin systems.
It is possible to obtain distance information for these
more complex systems but it is necessary to assume a
structural model beforehand [69]. To circumvent this
problem Gullion has developed theta-REDOR [76],
an elegant variation of the REDOR experiment.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

187

Fig. 19. Schematic representation of some double resonance pulse sequences; (a) SEDOR with t1 2t /3, (b) REDOR, (c) TEDOR and (d)
TRAPDOR. In all sequences the tip angle of the narrow pulses is p /2 and the wide pulses is p . For the sequences that employ MAS the number
of rotor cycles (Nc) are shown along the bottom.

Instead of applying a dephasing 180 pulse, a pulse


with a small flip angle is applied. In a multispin
system, for instance IS4, this means that the probability of all 4 S spins flipping and hence affecting the
dipolar dephasing of spin I is very small. In contrast,
the probability of only one spin undergoing a transition is highest (for small flip angles). In this way one
does not have to consider a complicated 5-spin system
but analyse four isolated 2-spin systems.
The TEDOR experiment [72,73] is capable of
transferring magnetisation between heteronuclear
spins, and consists of two consecutive REDOR
experiments, first on spin I and then on spin S,
which is observed (Fig. 19(c)). Using the spin density

matrix formalism one can easily show that the first


REDOR sequence creates 2IySz coherence which is
transferred to 2IzSy coherence by the application of
two 90 pulses on both spins. The subsequent
REDOR sequence on spin S transfers this coherence
to observable Sx coherence. Therefore, instead of
determining the amount of signal that is lost on spin
I, as in SEDOR and REDOR, one measures the signal
that is gained on spin S. The dipolar coupling constant
between I and S can be obtained by measuring the
signal intensity as a function of the number of rotor
cycles (from either the first or second REDOR period)
or as a function of the position of the dephasing
pulses. Each experiment has its own advantages and

188

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

disadvantages. The SEDOR experiment is easiest to


set up but lacks the resolution of MAS and is limited
in its range of accessible dipolar couplings by the
often short static T2. The REDOR experiment is
more elaborate to set-up and requires stable spinning
but the increased resolution enables multisite systems
to be explored while the longer T2 enables longer
distances to be determined. TEDOR is less sensitive
than REDOR as it involves a transfer step with a
theoretical maximum efficiency of 50%. In contrast,
it is not a difference experiment, and is therefore less
prone to experimental errors. A more important application of TEDOR is its use in 2D-correlation spectroscopy, which will be discussed in a later section.
All these experiments perform well for spin-1/2
systems and when an isolated spin pair is present
they allow the accurate determination of the IS
distance. They all rely on the ability to accurately
invert the spins by applying 180 pulses. With quadrupolar nuclei it is usually not possible to invert the
spin population with a 180 pulse. Therefore, in using
REDOR and SEDOR in a system with a quadrupolar
spin and a spin-1/2, one should apply the echo
sequence to the quadrupolar nucleus and the dephasing pulses on the spin-1/2. In this way, accurate
distance information can still be obtained, otherwise
it can only be qualitative. A further advantage in
performing SEDOR and REDOR in this manner is
that T1 of the quadrupolar nucleus will usually be
shorter than that of the spin-1/2, thereby reducing
the experimental time. If it is possible to apply an
accurate 180 pulse to the quadrupolar nucleus
which mainly inverts the population of the central
transition and which will leave the other spin levels
relatively unperturbed, only a fraction (given by the
intensity column of Table 1) of the quadrupolar nuclei
will be affected. This can provide a convenient way of
creating isolated spin pairs, as in the theta-REDOR
experiment. When using TEDOR for qualitative
information the best choice could be to start out
with the nucleus with the shortest T1, usually the quadrupole. To obtain distance information one can vary
the number of rotor periods of either REDOR section
in the TEDOR experiment. It would be preferred to
vary the REDOR section in which the dephasing 180
pulses are applied to the spin-1/2 system.
If, however, two quadrupolar nuclei are to be investigated one should be very careful in the quantitative

interpretation of the results. Information whether two


nuclei are in each others vicinity is still very useful
though and can be provided with these double resonance methods.
Fig. 20 shows the SEDOR curve of a 27Al/ 1H
SEDOR experiment on zeolite ZSM-5 [77]. An alumi was obtained
nium proton distance of 2.43 ^ 0.03 A
in this way. This agrees very well with the distance
that was determined using the more indirect method of
moments to analyse the 1H spectrum. As aluminium
was observed, it could be treated as a pseudo spin-1/2
and the SEDOR equations remained the same.
5.3. TRAPDOR and REAPDOR
The previous double resonance experiments were
developed for spin-1/2 systems. The TRAPDOR
[78,79] and Rotational-Echo Adiabatic Passage
Double Resonance (REAPDOR) [80,81] experiments
are designed specifically for quadrupolar nuclei and
do not work on spin-1/2 only systems. These MAS
experiments still rely on the modulation of the heteronuclear dipolar coupling (as in the REDOR experiment) to prevent an echo from refocusing but the
modulation is no longer achieved by the application
of 180 pulses.

Fig. 20. The aluminium SEDOR fraction as a function of the pulse


position. The data is corrected for the proton homonuclear dipolar
coupling. The theoretical curve is for an aluminiumproton distance
. (Taken from Ref. [77] with permission of the American
of 2.43A
Chemical Society.)

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

In the TRAPDOR experiment the observed nucleus


can be either spin-1/2 or a quadrupolar spin but the
dephasing spin has to be a quadrupolar nucleus.
During the rotor synchronised echo on the observed
nucleus the dephasing quadrupolar spin is continuously irradiated (Fig. 19(d)). As was discussed before,
continuous irradiation of a quadrupolar nucleus in
combination with MAS leads to rotationally induced
level transitions. So an individual spin will sample
different spin states (m) each rotor period. As each
spin state is affected differently by the dipolar
coupling (2mD), and the level transitions occur either
2 or 4 times per rotor period, the level transitions
cause dipolar dephasing which is additive for each
rotor period. Further, in contrast to REDOR which
mainly affects the 1/2 and 1/2 levels, the spin
locking affects all transitions so the dephasing in a
TRAPDOR experiment will be greater than in the
REDOR experiment [82]. As it is difficult to know
the spin states of the dephasing quadrupole nucleus
at each instant of the rotor period and the fact that the
efficiency of the spin-locking strongly depends on the
adiabaticity parameter a (see Section 4.3), it is difficult to calculate the dephasing effect on the other
nucleus exactly. For very small values of a , no
TRAPDOR effect will be visible while for completely
adiabatic spin-locking the dephasing effect will be the
largest. Therefore one cannot get accurate values of
the dipolar coupling using the TRAPDOR experiment, and it should only be used as a qualitative
experiment. Another interesting feature of the TRAPDOR experiment is its use for the estimation of Cq
values [83]. When the irradiation frequency of the
quadrupole falls outside the first-order quadrupole
spectrum the TRAPDOR effect is zero. Hence, the
position of the cut-off frequencies for the TRAPDOR
effect gives a direct result for n q from which the Cq
can be calculated. So for a spin I 5/2 the cut-off
frequencies will be at ^ 2n q.
Fig. 21 shows a 31P/ 27Al TRAPDOR experiment to
probe the binding of trimethylphosphine (TMP) on
dehydroxylated zeolite HY [84]. The spin-echo spectrum shows 4 resonances, of which the one at 3 ppm
originates from TMPH , the other 3 resonances were
thought to be associated with Lewis-bound TMP
adducts. The TRAPDOR experiment shows that
only one of the [31] P resonances (46 ppm) is
clearly in close contact with aluminium and can

189

Fig. 21. 31P/ 27Al TRAPDOR experiment of trimethylphosphine on


dehydroxylated zeolite HY. Without 27Al irradiation (a), with irradiation (b) and the TRAPDOR spectrum (c) which is the difference
between (a) and (b). The spinning speed was 5.5 kHz, the 27Al rffield strength was 60 kHz and irradiation during only one of the
echo periods for 725 ms. (Taken from Ref. [84] with permission.)

consequently be assigned to a TMPaluminium


Lewis acid complex. Using the offset effect of the
TRAPDOR experiment, the Cq of the aluminium associated with the 46 ppm signal is estimated at
11 MHz. The resonance at 60 ppm has only a very
small TRAPDOR fraction which indicates that this
phosphorus in not directly bound to aluminium
atoms or that the aluminium Cq is extremely large
( 15 MHz).
The REAPDOR [80,81] experiment can be considered as a variation of the TRAPDOR experiment
where a train of rotor-synchronised 180 pulses is
applied on the spin-1/2 nucleus, spaced half a rotor
period apart. The experiment uses an even number of
rotor periods. At the centre of the train of 180 pulses,
the dipolar evolution period, the 180 pulse is omitted.
In the first half of the dipolar evolution period the

190

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

spins will dephase as a result of the chemical shift


anisotropy and the heteronuclear dipolar coupling.
The second half will refocus the magnetisation and
an echo is formed at the end of the dipolar evolution
period. A short rf-pulse, the adiabatic passage pulse, is
applied for a duration t to the quadrupolar nucleus at
the centre of the dipolar evolution period. The spin
states of the quadrupolar nucleus will change (see
Section 4.3) and hence the value of the dipolar
coupling between the quadrupolar nucleus and the
spin-1/2 will change. The second half of the dipolar
evolution period will still refocus the chemical shift
anisotropy but no longer the dipolar dephasing, just as
in the REDOR experiment. The optimum duration of
the adiabatic passage pulse was determined for 14N to
be 1/3 of a rotor period. As in the TRAPDOR experiment, attention has to be given to the adiabaticity of
this pulse. An advantage of the REAPDOR experiment over the experimentally less demanding TRAPDOR experiment is that does not need continuous
irradiation of the quadrupole nucleus, so the chances
of the probe breaking down are reduced. Further, it
will be easier to calculate the dephasing effect of the
REAPDOR pulse sequence as one only needs to simulate the behaviour of the quadrupole spins with MAS
and continuous irradiation for a time t and not for the
entire (or half) the echo period.
It should be noted that in both the TRAPDOR and
REAPDOR experiments, the rf-field on the dephasing
quadrupolar nucleus is left on for a considerable time.
This can result in a BlochSiegert shift [85] which
manifests itself as a miss-set in the zero-order phasing
of the reduced echo as compared to the full echo
spectrum.

6. Two-dimensional experiments
6.1. Nutation NMR
Quadrupole parameters can often be obtained by
the measurement of the static or MAS spectra of a
powdered sample as outlined before. Unfortunately,
the distinct features of a quadrupolar lineshape are
often obscured by the existence of a spread in chemical shifts, residual dipolar broadening and/or the
quadrupole interaction. The idea behind nutation
NMR is to study the quadrupole interaction in an rf-

field in the rotating frame whereby the influence of the


chemical shifts is dramatically reduced as they scale
linearly with the applied field while one gains the
sensitivity benefit of measuring in a high static
magnetic field. The 2D nutation experiment consists
of monitoring the magnetisation as a function of the
nutation pulse (t1). During t1 the spin system is
governed in the rotating frame by the Hamiltonian.
H1 nrf Ix nq u; f3Iz2 I 2

17

with

nq u; f nq 3cos2 u 1 hq sin2 ucos 2f

18

In order to calculate the nutation spectrum the


density matrix formalism is used in which the Hamiltonian is diagonalised either numerically [11,86] or
analytically [87,88].
As was mentioned in Section 3, the nutation
frequency of a quadrupolar nucleus is (I 1/2)n 1
when n 1 p n q, and n 1 when n 1 n q, with n 1
g B1/2p . It follows then that the regime in which the
nutation spectrum displays distinct features is in the
region where n 1 is of the order of n q. Simulation of
nutation behaviour has shown that the useful range of
rf-field strengths lies within the ratio n q/n 1 0.051.
Within this range the distinct features in the nutation
spectrum allow the quadrupole parameters to be
obtained. This means that only quadrupole frequencies up to a certain limit are accessible with currently
available rf-fields of 500 kHz in specially dedicated
probeheads, and around 100 kHz in most commercially available systems. This especially limits the
investigation of spin I 3/2 nuclei such as 23Na.
In cases where more then one resonance could be
observed one would like to increase the resolution in
the F2-dimension by applying MAS. However, as was
discussed before, rotationally induced spin-locked
magnetisation can appear when an rf-field is applied
in conjunction with spinning. This process severely
distorts the nutation spectrum as it gives rise to strong
dispersive components in the 2D spectrum and a large
signal at zero frequency. The spinning speed should
therefore be kept low (24 kHz) so that the nutation
signal has decayed within a quarter rotor period at
most. It is possible however to simulate the nutation
behaviour of quadrupole nuclei under the influence of

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

191

MAS and from these the quadrupole parameters can


be extracted [89].
6.1.1. Off-resonance nutation
To increase the upper limit of quadrupole couplings
that are accessible by nutation NMR, off-resonance
nutation has been developed [90]. In this experiment
the nutation behaviour of the spins in the effective
field is monitored rather then in the rf-field, where
the effective field is the vector sum of the rf-field
and the resonance offset. This greatly increases the
upper limit of accessible Cq values as the resonance
offsets can be chosen at will and is only limited by the
Q of the probe. For increased sensitivity and for the
reduction of the zero frequency signal it is beneficial if
the magnetisation nutates perpendicular to the effective field. Therefore, one can either apply a frequency
stepped adiabatic half passage [14, 91] or a soft 90 to
bring the magnetisation into the xy-plane, of which the
former gives the best results. Then, the rf-phase is
shifted 90 and the frequency is switched to the
required resonance offset. The useful range of resonance offsets is between 1 and 8 times n 1, depending
on the ratio n q/n 1. After a time t1 the signal is detected
on resonance. Phase coherent frequency switching
(switching times of around 200 ns) of the rf-field is
required in this experiment which is within the
capabilities of most modern day spectrometers. A
recent extensive review article has appeared that
deals with many of the details of off-resonance nutation spectroscopy [92].
Whereas the on-resonance nutation spectra are
amplitude modulated and can be phased properly in
the F1 dimension, the off-resonance nutation spectra
are phase modulated and can no longer be phased
properly. Therefore, a magnitude calculation is
performed after Fourier Transformation. Further,
because the magnetisation evolves during t1 around
the effective field, the sine and cosine components
of the modulation are different and the amplitudes
of positive and negative F1 signals are unequal.
Fig. 22 gives a good example of the strength of the
off-resonance nutation technique [93]. Off-resonance
nutation experiments were performed on various
samples of g -alumina impregnated with phosphorus
and/or molybdenum. The figure shows the off-resonance nutation spectrum of Mo(12)P(2)/g -Al2O3. The
nutation spectrum of the tetrahedral resonance is very

Fig. 22. Off-resonance nutation experiments on Mo(12)P(2)/g Al2O3 with an rf-field strength of 41 kHz. (a) The single pulse
27
Al spectrum which looks identical to the spectrum of pure g Al2O3. (b) The off-resonance nutation spectrum for the tetrahedral
peak (top) and its simulation (bottom), Cq 4.6 MHz, h q 0.3,
n off 155 kHz. (c) The off-resonance nutation spectrum of the
octahedral peak (iv) and the simulation of subspectrum of g Al2O3 (ii), Cq 4.5 MHz, h q 0.3, n off 144 kHz, the subspectrum of Al2(MoO4)3 (i), Cq 1 MHz, h q 1, n off 144 kHz and
(iii) their weighted sum. (Taken from Ref. [93] with permission of
the American Chemical Society.)

similar to the nutation spectrum of g -Al2O3. The octahedral resonance however, shows a nutation spectrum
that can only be simulated using two sets of subspectra, one would be the nutation spectrum of g -Al2O3
and the second subspectrum has quadrupole parameters similar to those of Al2(MoO4)3. This led to
the conclusion that upon impregnation of g -Al2O3
with molybdenum the molybdenum reacts with the
g -Al2O3 surface to form Al2(MoO4)3.
6.2. Dynamic angle spinning
Dynamic Angle Spinning (DAS) [55, 94] is one of
the techniques that can be used to improve the resolution of solid state NMR spectra from quadrupolar

192

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

nuclei. DAS is a 2D experiment in which the sample is


spun about a different axis during each evolution
period. During the first evolution time t1 the sample
is spun at an angle of u 1 degrees. The magnetisation is
then stored along the z-axis and the angle of the spinning axis is changed to u 2. After the rotor is stabilised
(in the order of tens of milliseconds) the magnetisation is brought into the xy-plane again and a signal is
acquired. The second-order quadrupole frequency of
an individual crystallite depends on the angle of the
spinning axis. So during t1 the quadrupole frequency
will be n 1, and n 2 during t2. If n 2 is of opposite sign to
n 1 the signal from the crystallite will be at its starting
position again at some time during t2. One can choose
both angles in such a way that the signals from each
individual crystallite will be at the starting position at
exactly the same time. In other words, an echo will
form and the effect of the second-order quadrupolar
broadening is removed. The angles should fulfil the
following equations simultaneously:
P2 cos u1 kP2 cos u2
P4 cos u1 kP4 cos u2

19

where P2 and P4 are the second- and fourth-order


Legendre polynomials, and k is the scaling factor.
There is a continuous set of solutions for u 1 and u 2,
the so-called DAS complementary angles, and each
set has a different scaling factor. For these solutions,
the second-order quadrupole powder pattern at u 1 is
exactly the scaled mirror image of the pattern at u 2.
And an echo will form at t2 kt1. For the combination
u 1 30.56, u 2 70.12 the P4(cos u ) terms are zero
and the scaling factor k 1.87. For the combination
u 1 37.38, u 2 79.19 the scaling factor k 1 and
the spectra are exact mirror images and an echo will
form at t1 t2. Finally, the combination u 1 0, u 2
63.43 (k 5) is special because it allows one to use
efficient cross-polarisation for signal enhancement
prior to the DAS sequence. There are several ways
in which the DAS spectra can be acquired; one can
acquire the entire echo, which means that the resulting
2D spectrum will be sheared. Some additional processing is then required to obtain an isotropic spectrum in
F1. The acquisition could also start at the position of
the echo so that an isotropic spectrum in F1 is
obtained directly. This has the disadvantage that the

signal to noise ratio is less than if the complete echo is


acquired. A third possibility is to carry the experiment
out as a pseudo 1D experiment where only the top of
the echo is acquired as a function of t1. In this case the
isotropic spectrum is acquired directly but there is no
saving in the duration of the experiment. It is however
the best way to combine DAS with 2D heteronuclear
correlation.
The spinner axis is usually changed using a stepper
motor and a pulley system [55]. Angle switching
needs to be as fast as possible and reproducible. The
rf-coil can be fixed to the rotor so it changes orientation as the angle is switched. This gives a good filling
factor, but the rf-field will vary with the angle and the
rf-leads need to be flexible and resistant to metal fatigue. The coil may also surround the whole assembly
and remain static. In this case the filling factor will be
much worse but pulse lengths will remain relatively
constant, and most importantly, cross-polarisation at
an angle of 0 with the static field is now possible, in
contrast to the other design. A limitation of the DAS
technique is that it cannot be used on compounds with
a short T1 because of the time needed to reorient the
spinning axis. Further, a dedicated probe is needed to
carry out the experiment.
Fig. 23 shows a 2D 17O DAS spectrum of zeolite
Sil-Y and its isotropic projection [65]. DOR spectra at

Fig. 23. Two dimensional 17O DAS spectrum of zeolite Sil-Y,


obtained at a field of 11.7 T. Projections of the anisotropic spectra
for the separate resonances are shown at the right, the isotropic
spectrum is shown on top. The total duration of the experiment
was 2 weeks. (Taken from Ref. [65] with permission of the American Chemical Society.)

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

different fields allowed the isotropic shifts and Cq(1


h q2/3) to be determined. There are four different 17O
resonances present with the same intensity, as
expected from the structure. Simulation of the anisotropic slices from the 17O 2D DAS spectrum for each
peak, using the constraints obtained from DOR,
allowed the extraction of Cq and h q for each resonance. The experimentally determined values for
each site for the isotropic chemical shift (35
47 ppm), Cq (5.145.39 MHz) and h q (0.10.3)
were all very close. Quantum-chemical calculations
were used to obtain the electric field gradients and
chemical shifts for each oxygen site, leading to the
assignment of all four lines.
6.3. 2D MQMAS
In the past three years, a new experiment has
emerged that has had a great impact on solid state
NMR spectroscopy of quadrupolar nuclei. The 2D
Multiple Quantum Magic Angle Spinning (2D
MQMAS) experiment [95, 96] greatly enhances
resolution of the spectra of half-integer spin quadrupolar nuclei. Basically, this experiment correlates the
m; m multiple quantum transition to the
1=2; 1=2 transition. The resolution enhancement
stems from the fact that the quadrupole frequencies
for both transitions are correlated. At specific times
the anisotropic parts of the quadrupole interaction are
refocused and an echo forms. The frequency of an
m; m transition is given by:

np C0 pnQ0

7
C pnQ
4 u; f
18 4

Hz

21

where p 2m is the order of the coherence, p 1 for


the 1=2; 1=2 transition, p 3 for the 3=2; 3=2
transition, etc. The coefficients are defined as:


3
C0 p p II 1 p2
4


17 2
P 5 :
22
C4 p p 18II 1
2
Table 1 gives a list of the values of the coefficients
for the various coherences up to spin I 9/2. The
isotropic part nQ
0 of the quadrupole frequency is:

nQ0 n2q 3 h2q =90n0

Hz;

23

and the anisotropic part

nQ4 u; f

193

is given by:

nQ4 u; f n2q =112n0 7=183 hq cos 2f2 sin4 u




2hq cos2f 4 2=9h2q sin2 u
2=45h2q 4=5

24

Numerous schemes exist that are used to obtain 2D


MQMAS spectra. The most simple form of the experiment [96] (see Fig. 24(a)) is when the MQ transition is
excited by a single, high power rf-pulse, after which
the MQ-coherence is allowed to evolve for a time t1.
After the evolution time, a second pulse is applied
which converts the MQ coherence into a p 1
coherence which is observed during t2. The signal is
then acquired immediately after the second pulse and
the echo will form at a time t2 QAt1 , where QA is
the quadrupole anisotropy as given in Table 1. Both
pulses are non-selective and will excite all coherences
to a varying degree. Selection of the coherences of
interest is done by cycling the pulse sequence through
the appropriate phases. Phase cycles can be easily
worked out by noting that an rf-phase shift of f
degrees is seen as a phase shift of pf degrees for a
p-quantum coherence [97]. After a 2D Fourier transformation the resonances will show up as ridges lying
along the Quadrupole Anisotropy (QA) axis. The
isotropic spectrum can be obtained by projection of
the entire 2D spectrum on a line through the origin
(n 1 n 2 0) perpendicular to the QA-axis (vide
infra). Fig. 24 shows a selection of the many different
pulse sequences and the coherence pathways that can
be used for the 2D MQMAS experiment. Fig. 24(a) is
the basic pulse sequence as discussed before; note that
the coherence transfer pathway is asymmetric. Without proper weighting of the two different pathways
one obtains dispersive components in the 2D spectra.
Fig. 24(b) shows the pulse sequence that the authors
use most which includes a z-filter pulse (selective, of
low rf-power) and has the advantage of having a
symmetrical coherence transfer pathway [98]. Fig.
24(c) shows a split-t1 experiment in which the isotropic spectrum is directly obtained, no shearing of the
data is necessary here [99]. However, in many amorphous and non-crystalline materials a distribution in
chemical shifts exists. Suppose that the quadrupole
interaction in such a material is zero, in that case the

194

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 24. Four different MQMAS pulse sequences and their coherence pathways. (a) Two pulse sequence. (b) Z-filter type MQMAS experiment.
(c) Split-t1 experiment with z-filter and (d) RIACT(II) sequence. Sequences (a)(c) rely on powerful rf-pulses to excite the multiple quantum
transitions and to convert the coherences back to p 1 or 1 coherence. For a triple quantum experiment the conversion pulse is roughly 1/3 of
the duration of the excitation pulse.

frequency during t1 will be p times the frequency


during acquisition and an echo will form at t2 3t1.
This clearly illustrates a disadvantage of the split-t1
experiment as it relies on the echo formation that is
solely because of the quadrupole interaction. As a
result, intensity will be lost. Fig. 24(d) displays the
RIACT(II) sequence [100] in which the excitation and
conversion of the coherences is done by a spin-lock of
duration Tr/4 instead of a hard pulse, and has the
advantage of more uniform excitation and conversion
which is essential in obtaining quantitative

information. Each experiment has its own advantages


and disadvantages and new developments are still
being made.
There are two methods that are usually employed
for the phase sensitive detection in the F1 dimension,
both depend on amplitude modulation. A straightforward TPPI [101] can be incorporated in the pulse
sequence where the phase of the first pulse is incremented with 360/4p degrees for each t1 value. This
means that each t1 value contains alternately real (in
phase) and imaginary (out of phase) intensities. The

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

effective spectral width in the F1 dimension has to be


halved in this case. The other method is the hypercomplex States TPPI [102] where for each t1 value a real
and imaginary data point is obtained, and now the
effective spectral width in the F1 dimension is still
the inverse of the t1 dwell time. The difference
between the two methods does not seem large.
However, when one wants to apply rotor-synchronisation in the F1 dimension the difference in effective
dwell time reduces the maximum spectral width. For
example, with a rotor speed of say 20 kHz, rotor
synchronisation demands a dwell time of t1 Tr
which would mean a spectral width of 2/Tr,
(10 kHz) for TPPI but 1/Tr (20 kHz) for the States
TPPI method. As the former might be too small, the
latter is clearly preferred.
6.3.1. Data processing
In the three years since the first MQ report from
non-integer spin quadrupole nuclei there has been a
lot of research activity and anyone coming new to the
subject will find a number of schemes for processing
and presenting the data. The relationship between the
measured peak positions and the NMR interaction
parameters crucially depends on the processing and
referencing conventions adopted. There is one main
distinction between the two main approaches and
depends on whether the MQ evolution is regarded as
having taken place only in the evolution time (t1)
which is the convention adopted for example in
Refs. [96] and [103], or if the period up to the echo
is also regarded as being part of the evolution time,
which is then (1 QA)t1 (QA is defined in Table 1) as
adopted in Refs. [104] and [105]. A detailed critique
of these two approaches and the consequences of
adopting each has recently been given by Man
[106]. Our work has used the former direct definition
of evolution time which together with our preferred
processing philosophy is outlined in the following
text.
The isotropic shift and the quadrupole induced shift
(QIS) can easily be obtained from the data. First of all,
the 2D spectrum needs to be referenced correctly. To
obtain the correct ppm scale in the F1 dimension one
needs to set the spectrometer frequency in the F1
dimension to pn 0. The shift reference is most easily
set at the carrier frequency, i.e. the shift in ppm at the
carrier frequency in F1, the multiple quantum

195

dimension, is the same as the shift in ppm at the


carrier frequency in the F2, the single quantum dimension. This is valid for data obtained with or without
delayed acquisition or processed with or without
shearing. The isotropic chemical shift is the same
for both the single quantum and the multiple quantum
dimensions. The QIS is different in both dimensions,
however, and is given by

dpqis

C0 pnQ
0
106
pn0

in ppm:

25

The position of the centre of gravity dpcg with p 1


for the single quantum dimension and p Dm for the
multiple quantum dimension, is given by:

dpcg diso dpqis :

26

Hence the isotropic shift can be easily retrieved


using:

diso

p
C0 pdp1
cg pC0 1dcg
C0 p pC0 1

in ppm

27

and the isotropic quadrupolar shift:

nQ0

p
dp1
cg dcg
n
C0 1 C0 p=p 0

in Hz:

28

A graphical method to obtain the isotropic shift and


QIS from unsheared 2D spectra can give a quick
assessment of these parameters. Through the estimated centre of gravity one draws a line with a
slope equal to the QIS axis (see Table 1) crossing
the isotropic chemical shift line, d (F1) d (F2). The
intersection of these two lines gives the isotropic shift
and the difference in the shift between the centre of
gravity and the isotropic chemical shift gives QIS for
both transitions. Note that the slope of the QIS axis is
in Hz per Hz and needs to be adjusted for the ppm
scale, i.e. it is a factor 1/p less steep. The key point in
determining the quadrupole parameters is the accuracy in the position of the centre of gravity. Both
the excitation efficiency as well as significant intensity
in the spinning sidebands can adversely affect the
accuracy. It is therefore advisable to obtain the quadrupole parameters by simulation of the 1D MAS spectrum using the constraints on Cq and h q, dictated by
QIS, and the isotropic chemical shift as starting parameters.
Shearing of the data is done to obtain isotropic

196

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

spectra in the F1 dimension and to facilitate easy


extraction of the 1D slices for different peaks. Shearing is a projection of points that lie on a line with a
slope equal to the anisotropy axis onto a line that is
parallel to the F2 axis. The lines intersect at the F2
zero frequency (i.e. carrier frequency) line. In effect,
this means that a point that had a frequency of (n 1,n 2)
will now lie at frequency (n 1 kn 2), with k equal to
QA (see Table 1). If we now take the absolute
frequencies relative to the carrier frequency in the
F1 and F2 dimension:

n1 C0 pnQ0

extraction of QIS and d iso values. Moreover, shearing


introduces an extra processing step, which may introduce artefacts.
Fig. 25 illustrates the power of the 2D MQMAS

7
C pnQ
4 u; f pniso noff
18 4
29

and

n2 C0 1nQ0

7
C 1nQ
4 u; f niso noff : 30
18 4

then the sheared frequency n 1 0 n 1 kn 2 in F1


becomes

n 01 C0 p kC0 1nQ0 p kniso noff


in Hz:

31

The scale of the F1-axis remains unaltered and the


frequency of the isotropic peaks can readily be
obtained. Converting the F1-axis from frequency
units to ppm, such that 1 ppm equals [p k]n 0 Hz,
will give a peak position of:

d diso

C0 p kC0 1 Q
n0
p kn0

in ppm

32

when referenced correctly (i.e. d at carrier frequency


in F1 is d at carrier frequency in F2).
Shearing can be achieved in two ways. After the 2D
Fourier Transform one recalculates the position of the
data points to obtain the sheared spectrum. The disadvantage of this method is the need for interpolation.
Or, prior to Fourier Transformation with respect to t1
one applies first-order phasing to the data in the t1
dimension. That is, all points are multiplied by
exp( 2p in 2t1k), where n 2 is the frequency offset
from the carrier frequency in F2. Shearing essentially
achieves the same as the split t1 experiment or delayed
acquisition of the echo. Although sheared spectra may
look more attractive, they do not add any extra information and they are certainly not necessary for the

Fig. 25. (a) 17O Triple quantum 2D MQMAS spectrum of albite


glass. (b) Single pulse spectrum and the fit of the single pulse
spectrum using the slices obtained from the sheared 2D spectrum.
The MQMAS experiment is clearly able to resolve the 1D spectrum
into two lines. SiOSi: Cq 5.1 ^ 0.3 MHz, h q 0.15 ^ 0.05
and d iso 49 ^ 1 ppm; AlOSi: Cq 3.5 ^ 0.2 MHz, h q
0.05 ^ 0.05 and d iso 33 ^ 1 ppm. (Taken from Ref. [107]
Elsevier Science.)

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

197

17

technique. Here a O 2D triple quantum MAS


experiment was performed on dry albite glass. In
this glass one would expect to see SiOSi, AlO
Si and possibly even AlOAl bridging oxygens.
However, from the single pulse spectrum it is clear
that the lines overlap strongly. The 2D MQMAS spectrum separates the two lines which can be identified as
SiOSi and AlOSi species from their quadrupole
parameters [107]. No evidence of AlOAl species
was found. By shearing the data and adding all the
spinning sidebands to the centreband intensity the
subspectra for SiOSi and AlOSi were used to
fit the normal 1D MAS spectrum and the relative ratio
of the species were determined as 55% ^ 5% and
45% ^ 5%, respectively. In this way differences in
MQ excitation efficiencies between the two resonances were not influencing the results.
Fig. 26 illustrates the difference in resolving power
between five quantum and triple quantum 2D
MQMAS experiments [108]. It should be noted,
however, that the excitation of 5 quantum coherences
is less efficient than for 3 quantum coherences
[109,110], but as the 12Al Cq values for the tetrahedral
aluminium atoms are relatively small in AlPO-40
(1.62.7 MHz) [63] this posed no problem in this
case.
6.4. 2D XY correlation methods
Heteronuclear XY correlation experiments are very
useful in determining connectivities between different
nuclei when many resonances are present. In order to
measure a 2D heteronuclear correlation (2D
HETCOR) spectrum one needs to be able to transfer
magnetisation between the heteronuclei. This can be
achieved by CP or TEDOR. A variable time t1 is
inserted prior to the magnetisation transfer from say
spin I to spin S. By measuring the FID of spin S as a
function of the time t1 the S spin FID is modulated by
the signal intensity of spin I at t1. A 2D Fourier Transform will give the IS 2D correlation spectrum. Phase
sensitive detection in F1 can be achieved by for
instance TPPI. Various (2D HETCOR) experiments
involving quadrupolar nuclei have appeared in the
literature [111,112]. The main difference with the
spin-1/2 only case is the attention that has to be
given to the transfer of the magnetisation (as was
discussed in previous sections). Overlapping lines in

Fig. 26. (a) 27Al triple quantum 2D MQMAS spectrum of aluminophosphate molecular sieve AlPO-40 and (b) the 27Al five quantum 2D MQMAS spectrum. Both spectra display only the
tetrahedral aluminium region. The eleven resonances that are
present are better resolved in the five quantum spectrum than in
the triple quantum spectrum. Note that the slopes of the quadrupole
anisotropy and the QIS axes are different in each spectrum. (Taken
from Ref. [108] with permission of Elsevier Science.)

the 1D spectrum of the quadrupolar nucleus can be


resolved in the 2D correlation spectrum. This is
however not always the case as a result of the
second-order quadrupole broadening and might
hamper the assignment of the correlation peaks. One
would therefore like to incorporate a line narrowing

198

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Fig. 27. 23Na 31P 2D-HETCOR spectra of sodium trimetaphosphate (Na3P3O9) (a) obtained with a straightforward HETCOR pulse sequence
(90t1 CPt2) and (b) a MQMAS pulse sequence is incorporated in the experiment, leading to much higher resolution. (Taken from Ref. [114]
with permission of Academic Press.)

method such as DOR, DAS or MQMAS into the 2D


HETCOR experiment.
DAS has been used to obtain a high resolution 2D
HETCOR spectrum of sodium trimetaphosphate
[113]. It involved spinning at two different angles
(79.19 and 37.38) during t1 such that the isotropic
echo is formed at the end of the t1 evolution period.
The magnetisation was stored along the z-axis and
then transferred from 23Na to 31P at an angle of 0

and the 31P signal was detected at the magic angle.


Three reorientations of the spinner axis were used
during the experiment which can be a drawback for
short T1 compounds. However, the number of reorientations during the experiment can be reduced to only
two by selecting a different set of DAS complementary angles. In that case, u 1 63.43 and u 2 0,
which eliminates the need for an extra reorientation
for the CP sequence.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

Another possibility is to incorporate the MQMAS


experiment into the 2D HETCOR sequence [114]. In
this experiment one performs a split-t1 MQMAS experiment during the t1 evolution period, again such that at the
start of the magnetisation transfer the isotropic echo is
formed. Fig. 27 shows the results of a MQMAS/
HETCOR experiment on sodium trimetaphosphate
with the resolution in the 23Na dimension greatly
increased compared to the straightforward HETCOR
experiment. This sequence has recently been improved
by changing the position in the sequence at which CP
occurs [115]. It should be noted that in both types of
experiments the results are not quantitative. To our
knowledge HETCOR experiments on a DOR probe
have not been performed yet.
7. Conclusions
This article has presented the physical background
to many of the techniques that are now available for
the study of non-integer quadrupolar nuclei using
solid state NMR. There are other reviews which the
interested reader is referred to for different perspectives on the same area of research [116118]. The aim
of this article was to outline the relative merits of the
different approaches and make them accessible, particularly with regards to the practical implementation
for the non-specialist.
Acknowledgements
The authors would like to thank all the people and
copyright holders who gave permission to use figures
from their original work to illustrate this article, and
all those workers who have shared their expertise. Ms
G. Roch is thanked for running the spectra of Fig. 14.
MES thanks EPSRC for funding research on developing NMR to characterise materials (Grants GR/
K74867 and GR/L28647). Also HEFCE is thanked
for funding high field solid state NMR research
through a JREI grant.
References
[1] A. Abragam, Principles of Nuclear Magnetism, Oxford
University Press, Oxford, 1983.

199

[2] C.P. Slichter, Principles of Magnetic Resonance, SpringerVerlag, Berlin, 1990.


[3] R.K. Harris, NMR Spectroscopy, Pitman, London, 1984.
[4] B.C. Gerstein, C. Dybowski, Transient Techniques in NMR
of Solids, Academic Press, New York, 1985.
[5] R.M. Sternheimer, Phys. Rev. 95 (1954) 736.
[6] P. Blaha, K. Schwarz, P.H. Dederichs, Phys. Rev. B 37
(1988) 2792.
[7] M.H. Cohen, F. Reif, Solid State Phys. 5 (1957) 321.
[8] V.H. Schmidt, Proc. Ampere Int. Summer School II, 75
(1971).
[9] D. Fenzke, D. Freude, T. Frohlich, J. Haase, Chem. Phys.
Lett. 111 (1984) 171.
[10] A. Samoson, E. Lippmaa, Phys. Rev. B 28 (1983) 6567.
[11] F.M.M. Geurts, A.P.M. Kentgens, W.S. Veeman, Chem.
Phys. Lett. 120 (1985) 206.
[12] E. Lippmaa, A. Samoson, M. Magi, J. Am. Chem. Soc. 108
(1986) 1730.
[13] E. Fukushima, S.B.W. Roeder, Experimental Pulse NMR,
Addison-Wesley, Reading, MA, 1981.
[14] A.P.M. Kentgens, J. Magn. Reson. 95 (1991) 619.
[15] M.Y. Liao, B.G.M. Chew, D.B. Zax, Chem. Phys. Lett. 242
(1995) 89.
[16] M.E. Smith, J.H. Strange, Meas. Sci. Technol. 7 (1996) 449.
[17] G. Kunath-Fandrei, PhD Thesis, Univerity of Jena, ShakerVerlag, Aachen, 1998.
[18] L.B. Alemany, D. Massiot, B.L. Sherriff, M.E. Smith, F.
Taulelle, Chem. Phys. Lett. 177 (1991) 301.
[19] E.L. Hahn, Phys. Rev. 80 (1950) 580.
[20] I. Solomon, Phys. Rev. 110 (1958) 61.
[21] O. Kanert, M. Mehring, in: P. Diehl, E. Fluck, R. Kosfeld
(Eds.), NMR Basic Principles and Progress, Springer-Verlag,
Berlin, 3 (1971) 1.
[22] P.P. Man, Appl. Magn. Reson. 4 (1993) 65.
[23] J. Haase, E. Oldfield, J. Magn. Reson. A101 (1993) 30.
[24] P.P. Man, Phys. Rev. B 52 (1995) 9418.
[25] J. Haase, E. Oldfield, J. Magn. Reson. A 104 (1993) 1.
[26] A.C. Kunwar, G.L. Turner, E. Oldfield, J. Magn. Reson. 69
(1986) 124.
[27] Y. Furukawa, S. Wada, J. Phys.: Condensed Matter 6 (1994)
8023.
[28] T.J. Bastow, M.E. Smith, Solid State NMR 3 (1994) 17.
[29] A. Narath, Phys. Rev. 162 (1967) 320.
[30] H. Ebert, J. Abart, J. Voitlander, J. Phys. F 16 (1986) 1287.
[31] X. Wu, E.A. Juban, L.G. Butler, Chem. Phys. Lett. 221
(1994) 65.
[32] I.J.F. Poplett, M.E. Smith, Solid State NMR 11 (1998) 211.
[33] T.J. Bastow, C. Forwood, M.J. Gibson, M.E. Smith, Phys.
Rev. B 58 (1998) 2988.
[34] E.R. Andrew, Int. Rev. Phys. Chem. 1 (1981) 195.
[35] J. Bohm, D. Fenzke, H. Pfeifer, J. Magn. Reson. 55 (1985) 197.
[36] C. Jaeger, G. Kunath, P. Losso, G. Scheler, Solid State NMR
2 (1993) 73.
[37] R. Dupree, M.H. Lewis, M.E. Smith, J. Appl. Crystal. 21
(1985) 109.
[38] D. Massiot, C. Bessada, J.P. Coutures, F. Taulelle, J. Magn.
Reson. 90 (1990) 231.

200

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201

[39] M.E. Smith, S. Steuernagel, Solid State NMR 1 (1992) 175.


[40] T.J. Bastow, M.E. Hobday, M.E. Smith, H.J. Whitfield, Solid
State NMR 3 (1994) 49.
[41] A.J. Vega, J. Magn. Reson. 96 (1992) 50.
[42] J.S. Frye, G.E. Maciel, J. Magn. Reson. 48 (1982) 125.
[43] H.J. Jakobsen, J. Skibsted, H. Bildsoe, N.C. Nielsen, J.
Magn. Reson. 85 (1989) 173.
[44] J. Skibsted, N.C. Nielsen, H. Bildsoe, H.J. Jakobsen, J.
Magn. Reson. 95 (1991) 88.
[45] C. Jaeger, in: B. Blumich, R. Kosfeld (Eds.) NMR Basic
Principles and Progress, Springer Verlag, Berlin, 31 (1994)
135.
[46] A. Samoson, Chem. Phys. Lett. 119 (1985) 29.
[47] C. Jaeger, R. Dupree, S.C. Kohn, M.G. Mortuza, J. NonCryst. Solids 95 (1993) 155.
[48] G. Kunath-Fandrei, T.J. Bastow, J.S. Hall, C. Jaeger, M.E.
Smith, J. Phys. Chem. 99 (1995) 15 138.
[49] S. Ganapathy, S. Schramm, E. Oldfield, J. Chem. Phys. 77
(1982) 4360.
[50] F. Lefebvre, J.P. Amoureux, C. Fernandez, E.G. Derouane, J.
Chem. Phys. 86 (1987) 6070.
[51] J.P. Amoureux, C. Fernandez, F. Lefebvre, Magn. Reson.
Chem. 28 (1990) 5.
[52] S. Ganapathy, J. Shore, E. Oldfield, Chem. Phys. Lett. 169
(1990) 301.
[53] A. Llor, J. Virlet, Chem. Phys. Lett. 152 (1988) 248.
[54] A. Samoson, E. Lippmaa, A. Pines, Molec. Phys. 65 (1988)
1013.
[55] J. Zwanziger, B.F. Chmelka, in: B. Blumich, R. Kosfeld
(Eds.), NMR Basic Principles and Progress, Springer-Verlag,
Berlin, 31 (1994) 202.
[56] Y. Wu, B.F. Chmelka, A. Pines, M.E. Davis, P.J. Grobet,
P.A. Jacobs, Nature 346 (1990) 550.
[57] K.T. Mueller, Y. Wu, B.F. Chmelka, J. Stebbins, A. Pines, J.
Am. Chem. Soc. 113 (1991) 32.
[58] Y. Wu, B.Q. Sun, A. Pines, A. Samoson, E. Lippmaa, J.
Magn. Reson. 84 (1990) 297.
[59] A. Samoson, E. Lippmaa, J. Magn. Reson. 89 (1989) 410.
[60] Y. Wu, D. Lewis, J.S. Frye, A.R. Palmer, R.A. Wind, J.
Magn. Reson. 100 (1992) 425.
[61] B.Q. Sun, J.H. Baltisberger, Y. Wu, A. Samoson, A. Pines,
Solid State NMR 1 (1992) 267.
[62] E. Cochon, J.P. Amoureux, Solid State NMR 2 (1993) 105.
[63] M.P.J. Peeters, J.W. De Haan, L.F.M. Van De Ven, J.H.S.
Van Hofe, J. Phys. Chem. 97 (1993) 5363.
[64] M. Hunger, G. Engelhardt, H. Koller, J. Weitkamp, Solid
State NMR 2 (1993) 111.
[65] L.M. Bull, A.K. Cheetham, T. Anupold, A. Reinhold, A.
Samoson, J. Sauer, B. Bussemer, Y. Lee, S. Gann, J.
Shore, A. Pines, R. Dupree, J. Am. Chem. Soc. 120 (1998)
3510.
[66] J. Haase, E. Oldfield, Solid State NMR 3 (1994) 171.
[67] M. Ehmswiller, E.L. Hahn, D.E. Kaplan, Phys. Rev. 118
(1960) 414.
[68] P.K. Wang, C.P. Slichter, J.H. Sinfelt, Phys. Rev. Lett. 53
(1984) 82.
[69] E.R.H. Van Eck, W.S. Veeman, Solid State NMR 1 (1992) 1.

[70] T. Gullion, J. Schaefer, J. Magn. Reson. 81 (1989) 196.


[71] T. Gullion, J. Schaefer, Adv. Magn. Reson., Academic Press,
San Diego, 13 (1989) 57.
[72] E.R.H. Van Eck, W.S. Veeman, Solid State NMR 2 (1993)
307.
[73] A.W. Hing, S. Vega, J. Schaefer, J. Magn. Reson. 96 (1992)
205.
[74] A.W. Hing, S. Vega, J. Schaefer, J. Magn. Reson. A103
(1993) 151.
[75] M.E. Merritt, J.M. Goetz, D. Whitney, C.P.P. Chang, L.
Heux, J.L. Halary, J. Schaefer, Macromolecules 31 (1998)
1214.
[76] T. Gullion, C.H. Pennington, Chem. Phys. Lett. 290 (1998)
88.
[77] N.P. Kenaston, A.T. Bell, J.A. Reimer, J. Phys. Chem. 98
(1994) 894.
[78] C.P. Grey, W.S. Veeman, Chem. Phys. Lett. 192 (1992) 379.
[79] C.P. Grey, W.S. Veeman, A.J. Vega, J. Chem. Phys. 98
(1993) 7711.
[80] T. Gullion, Chem. Phys. Lett. 246 (1995) 325.
[81] T. Gullion, J. Magn. Reson. A117 (1995) 326.
[82] E.R.H. Van Eck, A.P.M. Kentgens, H. Kraus, R. Prins, J.
Phys. Chem. 99 (1995) 16 080.
[83] C.P. Grey, A.J. Vega, J. Am. Chem. Soc. 117 (1995) 8232.
[84] H. Kao, C.P. Grey, Chem. Phys. Lett. 259 (1996) 459.
[85] F. Bloch, A. Siegert, Phys. Rev. 57 (1940) 522.
[86] A.P.M. Kentgens, J.J.M. Lemmens, F.M.M. Geurts, W.S.
Veeman, J. Magn. Reson. 71 (1987) 62.
[87] L. Pandey, S. Towta, D.G. Hughes, J. Chem. Phys. 85 (1986)
6923.
[88] R. Janssen, W.S. Veeman, J. Chem. Soc. Faraday Trans. 84
(1988) 3747.
[89] N.C. Nielsen, H. Bildsoe, H.J. Jakobsen, J. Magn. Reson. 97
(1992) 149.
[90] A.P.M. Kentgens, J. Magn. Reson. A104 (1993) 302.
[91] E. Van Veenedaal, B.H. Meier, A.P.M. Kentgens, Molec.
Phys. 93 (1998) 195.
[92] A.P.M. Kentgens, Prog. NMR Spec. 32 (1998) 141.
[93] H. Kraus, R. Prins, A.P.M. Kentgens, J. Phys. Chem. 100
(1996) 16 338.
[94] K.T. Mueller, B.Q. Sun, G.C. Chingas, J.W. Zqanziger, T.
Terao, A. Pines, J. Magn. Reson. 86 (1990) 470.
[95] L. Frydman, J.S. Harwood, J. Am. Chem. Soc. 117 (1995)
5367.
[96] A. Medek, J.S. Harwood, L. Frydman, J. Am. Chem. Soc.
117 (1995) 12 779.
[97] R.R. Ernst, G. Bodenhausen, A. Wokaun, Principles of NMR
in One and Two Dimensions, Clarendon Press, Oxford, 1987.
[98] J.P. Amoureux, C. Fernandez, S. Steuernagel, J. Magn.
Reson. A123 (1996) 116.
[99] S.P. Brown, S. Wimperis, J. Magn. Reson. A 124 (1994) 279.
[100] G. Wu, D. Rovnyak, R.G. Griffin, J. Am. Chem. Soc. 118
(1996) 9326.
[101] D. Marion, K. Wuthrich, Biochem. Biophys. Res. Commun.
113 (1983) 967.
[102] D.J. States, R.A. Haberkorn, D.J. Ruben, J. Magn. Reson. 48
(1982) 286.

M.E. Smith, E.R.H. van Eck / Progress in Nuclear Magnetic Resonance Spectroscopy 34 (1999) 159201
[103] M. Hanaya, R.K. Harris, J. Phys. Chem. A101 (1997) 279.
[104] D. Massiot, B. Touzo, D. Trumeau, J.P. Coutures, J.
Virlet, P. Florian, P.J. Grandinetti, Solid State NMR 6
(1996) 73.
[105] S.H. Wang, Z. Xu, J.H. Baltisberger, L.M. Bull, J.F.
Stebbins, A. Pines, Solid State NMR 8 (1997) 1.
[106] P.P. Man, Phys. Rev. B 58 (1998) 2764.
[107] P.J. Dirken, S.C. Kohn, M.E. Smith, E.R.H. Van Eck, Chem.
Phys. Lett. 266 (1997) 568.
[108] J. Rocha, J.P. Lourenco, M.F. Riberio, C. Fernandez, J.P.
Amoureux, Zeolite 19 (1997) 156.
[109] J.P. Amoureux, C. Fernandez, L. Frydman, Chem. Phys.
Lett. 259 (1996) 347.
[110] J.P. Amoureux, C. Fernandez, Solid State NMR 10 (1998)
211.

201

[111] C.A. Fyfe, K.T. Mueller, H. Grondey, K.C. Wong-Moon,


Chem. Phys. Lett. 199 (1992) 198.
[112] E.R.H. Van Eck, W.S. Veeman, J. Am. Chem. Soc. 115
(1993) 1168.
[113] T.P. Jarvie, R.M. Wenslow, K.T. Mueller, J. Am. Chem. Soc.
117 (1995) 570.
[114] S.H. Wang, S.M. De Paul, L.M. Bull, J. Magn. Reson. 125
(1997) 364.
[115] S. Steuernagel, Solid State NMR 11 (1998) 197.
[116] D. Freude, J. Haase, in: P. Diehl, E. Fluck, H. Guenther, R.
Kosfeld, J. Seelig (Eds.), NMR Basic Principles and
Progress, Springer Verlag, Berlin, 29 (1993) 1.
[117] R.E. Youngman, U. Werner-Zwanziger, J.W. Zwanziger, Z.
Naturforsch. 51A (1996) 321.
[118] A.P.M. Kentgens, Geoderma 80 (1997) 271.

Вам также может понравиться