Solid State NMR Basic Principles and Practice

SOLID-STATE NMR
SOLID-STATE NMR:
BASIC PRINCIPLES & PRACTICE
DAVID C. APPERLEY, ROBIN K. HARRIS & PAUL HODGKINSON
Logo.eps
Solid-State NMR: Basic Principles & Practice
Copyright © Momentum Press®, LLC, 2012.
All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system, or transmitted in any form or by any means—electronic,
mechanical, photocopy, recording, or any other—except for brief
quotations, not to exceed 400 words, without the prior permission of the
publisher.
First published by Momentum Press®, LLC

222 East 46th Street, New York, NY 10017
www.momentumpress.net
ISBN-13: 978-1-60650-350-8 (hard back, case bound)

ISBN-10: 1-60650-350-2 (hard back, case bound)
ISBN-13: 978-1-60650-352-2 (e-book)
ISBN-10: 1-60650-352-9 (e-book)
DOI: 10.5643/9781606503522
Cover design by Jonathan Pennell

Interior design by Exeter Premedia Services Private Ltd.,
Chennai, India
10 9 8 7 6 5 4 3 2 1
Printed in the United States of America

CONTENTS
PREFACE
ABOUT THE AUTHORS
1 INTRODUCTION
1.1 The utility of NMR
1.2 A preview of solid-state NMR spectra
1.3 The solid state
1.3.1 Introduction
1.3.2 Symmetry in the crystalline state
1.3.3 Effects of crystal structure on NMR
1.3.4 Types of solids
1.4 Polymorphism, solvates, co-crystals & host:guest systems
1.5 NMR of solids & the periodic table
2 BASIC NMR CONCEPTS FOR SOLIDS

2.1 Nuclear spin magnetization
2.2 Tensors
2.3 Shielding
2.4 Indirect coupling
2.5 Dipolar coupling
2.6 Quadrupolar coupling
2.7 Magic-angle spinning
2.8 Relaxation
3 SPIN - FMEQN003.WMF NUCLEI : A PRACTICAL GUIDE

3.1 Introduction
3.2 The vector model & the rotating frame of reference
3.2.1 Pulse angle
3.3 The components of an NMR experiment
3.3.1 Recycle delay
3.3.2 Acquisition time
3.3.3 Receiver gain
3.3.4 Spectral width
3.3.5 Dead-time
3.3.6 Spinning sideband suppression
3.3.7 Decoupling
3.3.8 Spectral referencing
3.3.9 Temperature calibration
3.4 Cross polarization
3.4.1 The cross-polarization experiment
3.4.2 Contact time
3.4.3 Direct excitation or cross polarization? A summary
3.5 High-resolution spectra from 1H (& 19F)
4 QUANTUM MECHANICS OF SOLID -STATE NMR

4.1 Introduction
4.2 The Hamiltonians of NMR
4.2.1 Spin operators
4.2.2 Secular & non-secular terms
4.2.3 Coupling Hamiltonians
4.2.4 Radiofrequency & the rotating frame
4.3 The density matrix
4.4 Density operator treatments of simple NMR experiments
4.4.1 The basic NMR experiment
4.4.2 Echoes & coherence pathway diagrams
4.5 The density matrix for coupled spins
4.5.1 Example 1: a dipolar-coupled homonuclear spin pair
4.5.2 Example 2: Cross-polarization
4.6 Euler angles & spherical tensors
4.6.1 Magic-angle spinning
4.7 Additional analytical tools
4.7.1 Average Hamiltonian theory
4.7.2 An overview of Floquet theory
4.7.3 Introduction to irreducible spherical tensor operators
5 GOING FURTHER WITH SPIN - FMEQN003.WMF SOLID -

STATE NMR
5.1 Introduction
5.1.1 Spin- Eqn003.wmf NMR at high magnetic fields
5.1.2 Advanced heteronuclear decoupling
5.1.3 Advanced cross polarization
5.2 Linewidths in solid-state NMR
5.3 Exploiting indirect (J) couplings in solids
5.4 Spectral correlation experiments
5.4.1 Basic principles of two-dimensional NMR
5.4.2 Transfer of magnetization via dipolar couplings
5.4.3 Heteronuclear correlation
5.4.4 Homonuclear correlation
5.5 Homonuclear decoupling
5.5.1 Overview of homonuclear decoupling sequences
5.6 Using correlation experiments for spectral assignment
5.7 Further applications
5.7.1 Labeled systems
5.7.2 Quantitative applications
6 QUADRUPOLAR NUCLEI
6.1 Introduction
6.2 Characteristics of first-order quadrupolar spectra
6.3 First-order energy levels & spectra
6.4 Second-order zero-asymmetry cases
6.4.1 Transition frequencies
6.4.2 Central-transition spectra: Static samples
6.4.3 Central-transition spectra: Rapid sample spinning
6.5 Spectra for cases with non-zero asymmetry: Central
transition
6.6 Recording one-dimensional spectra of quadrupolar nuclei
6.6.1 Nutation
6.7 Manipulating the quadrupolar effect
6.7.1 Variable-angle spinning
6.7.2 Double rotation
6.7.3 Dynamic-angle spinning
6.7.4 Multiple quantum magic-angle spinning
6.7.5 Satellite transition magic-angle spinning
6.7.6 Summary for spectroscopy of half-integer quadrupolar
nuclei
6.8 Spectra for integral spins
7 RELAXATION , EXCHANGE & QUANTITATION
7.1 Introduction
7.2 Relaxation
7.2.1 The cause of relaxation
7.2.2 Proton relaxation times
7.2.3 Lineshapes & linewidths for non-spinning samples
7.2.4 Relaxation & high-resolution measurements combined
7.2.5 Measuring relaxation times
7.3 Exchange
7.3.1 Positional exchange
7.3.2 Hydrogen exchange
7.3.3 Reorientation without a change in isotropic chemical
shift
7.3.4 Diffusive motion
7.3.5 “Soft” solids
7.3.6 Interference
7.4 Quantitative NMR
7.4.1 Relative intensity
7.4.2 Absolute intensity
7.5 Paramagnetic systems
7.5.1 Relaxation effects
7.5.2 Shift effects
8 ANALYSIS & INTERPRETATION

8.1 Introduction
8.2 Quantitative measurement of anisotropies
8.2.1 General
8.2.2 Quantitation of powder lineshapes
8.2.3 Quantitation of spinning sideband manifolds
8.2.4 Single-crystal vs. polycrystalline samples
8.2.5 Resolving anisotropy information by isotropic shift
8.3 Measurement of dipolar couplings
8.3.1 Measurement of heteronuclear couplings
8.3.2 Measurement of homonuclear couplings
8.4 Quantifying indirect (J) couplings
8.5 Tensor interplay
8.6 Effects of quadrupolar nuclei on spin- Eqn003.wmf spectra
8.7 Quantifying relationships between tensors
8.7.1 From one-dimensional spectra
8.7.2 From correlation spectra
8.7.3 Specialized experiments
8.8 NMR crystallography
8.8.1 Chemical examples
8.8.2 Computation of NMR parameters
APPENDICES
A THE SPIN PROPERTIES OF SPIN - FMEQN003.WMF

NUCLIDES
B THE SPIN PROPERTIES OF QUADRUPOLAR NUCLIDES
C LIOUVILLE SPACE , RELAXATION & EXCHANGE

C.1 Introduction to Liouville space
C.2 Application to relaxation
C.3 Application to chemical exchange
D INTRODUCTION TO SOLID -STATE NMR SIMULATION

D.1 Specifying the spin system
D.2 Specifying the powder sampling
D.3 Specifying the pulse sequence
D.4 Efficiency of calculation
INDEX
PREFACE
Nuclear magnetic resonance (NMR) has proved to be a uniquely powerful
and versatile spectroscopy, and no modern university chemistry department
or industrial chemistry laboratory is complete without a suite of NMR
spectrometers. The phenomenon of nuclear spin may seem an odd basis for
an analytical tool, but it is the relative isolation of the nuclear spin from its
surroundings that makes it an ideal noninterfering probe of the electronic
environment. Different sites are clearly identified by their chemical shifts,
while J couplings in 1 H spectra provide connectivity information. The
combination of these two complementary interactions, plus the formidable
array of different NMR experiments developed since the arrival of Fourier
transform NMR in 1966, has revolutionized the practice of chemistry.
While the original discovery of NMR involved both solids and liquids, the
application of NMR to materials in solid form developed at a markedly
slower rate than its solution-state counterpart until relatively recently. The
most obvious explanation for this difference is the fact that molecular
mobility in isotropic solutions averages anisotropic interactions (such as
shielding) to their isotropic values. In particular, for solids it is necessary to
consider a number of NMR interactions (notably the dipolar and
quadrupolar interactions) that can generally be ignored for the solution
state. These can significantly reduce spectral resolution and complicate
interpretation of spectra. On the other hand, the direct effect of these
interactions means that solid-state NMR is potentially a much richer
information source than NMR of solutions.
A second, perhaps less appreciated, reason for the relatively tardy progress
of solid-state NMR is the nature of the solid state itself. Solids, and
especially the samples the solid-state NMR spectroscopist is asked to deal
with, are rarely the simple monocrystals beloved by diffraction
crystallographers. Solid-state NMR may well give broad, featureless lines
for amorphous or heterogeneous samples, but this reflects the underlying
nature of the system—the chemical shift of a given nucleus may be a
distribution and not a single value as it would be if the sample were
dissolved. As a result, solid-state NMR may be providing too much
information, so that the art of solid-state NMR spectroscopy lies in finding
the right approach to refining the information content.
It is this complex interaction between the nature of the sample and the
multiple anisotropic NMR interactions that makes solid-state NMR a
challenging technique to master, and the treatments in undergraduate texts
do little to dispel the image of a dark and troublesome technique. The
purpose of the current text is to provide a bridge between the familiar world
of isolated molecules relevant to solution-state NMR and the subtle world
of solid materials. We hope that this introduction and survey will be of
value to both established researchers wishing to learn what solid-state
NMR can (and cannot) do for their systems and graduate students starting
work in this area. We have deliberately avoided detailed mathematical
treatments in the early chapters, concentrating on providing that essential
qualitative “feel” for the different aspects of solid-state NMR. The
theoretical treatment, which will be of most relevance to those starting a
career in the field, is delayed to chapter 4. This survey of the theoretical
tools used in solid-state NMR should allow the reader to tackle the more
specialized literature with increased confidence. The advanced experiments
described in chapters 5–8 can only be completely understood in terms of
the underlying theory, but again a basic feel for the experiments and what
information they provide should still be clear without such knowledge.
Solid-state NMR is now too vast and varied to try to present a

comprehensive review. Our goal is rather to provide a survey of the
technique and the essential tools for exploring further, together with a
practical guide on the application and use of solid-state NMR. A selection
of review articles is listed at the end of each chapter for those wishing to
explore individual topics in more detail. Similarly, to keep the text to a
manageable size, we are assuming a basic familiarity with 1 H and 13 C
solution-state NMR and the Fourier transform NMR experiment (see the
“Further reading” for chapter 1 for some excellent introductory texts).
We have tried to emphasize the interactions between solid-state NMR and

other techniques. A very significant one of these is the cross-fertilization
between the solid-state and solution-state branches of NMR. The recent
past has seen a proliferation of multidimensional techniques, often initially
developed in solution, being applied to solid-state NMR. Solids are
intrinsically complex materials and we cannot expect a single technique to
provide all the answers, so we also point out where the interpretation of
NMR results is aided by other experimental techniques, such as X-ray
diffraction, or tools from theoretical chemistry, such as ab initio calculation
of chemical shifts. Moreover, we emphasize the additional information
(much of it complementary to that obtained by diffraction measurements)
that NMR brings to our knowledge of solid-state structure.
In summary, making the best use of solid-state NMR involves a good

understanding of the materials under study, familiarity with a broad range
of techniques, and a sense for the complementary information that can be
provided by other techniques. This may seem a daunting challenge for a
newcomer, but the subtle interactions between sample, equipment, and
theory are key as to why solid-state NMR is such an intellectually
rewarding field in which to work.
Many people have contributed to the appearance of this book. Members of

our research group have read and commented on the various chapters as
they were written, for which we thank them. We are also very grateful for
detailed comments on chapter 6 by Dr. Sharon Ashbrook. A number of
research colleagues in other locations have supplied the basis for some of
the diagrams, as is acknowledged at appropriate places, and we thank these
friends for such help. Finally, we thank our publishers, Momentum Press,
for the attractive style and efficient production of the book.
David C. Apperley, Robin K. Harris, Paul Hodgkinson

October 2011, Durham
ABOUT THE AUTHORS
David C. Apperley studied chemistry at the University of East Anglia,
Norwich, and gained a Ph.D. for research on dipolar coupling in solids
from the Open University, Milton Keynes, in 1986. He further developed
his interest in solid-state NMR while working as a senior experimental
officer in the Durham Solid-State NMR Research Service, at first in the
Industrial Research Laboratories and now in the Department of Chemistry,
Durham University. For many years he has been the manager of the facility,
which serves both industry and other universities (the latter operation
funded by the Engineering and Physical Sciences Research Council). His
role with the NMR Service is to provide access to, training in, and
interpretation of results from solid-state NMR measurements for scientists
in industry and in the UK universities. As well as providing support to
organic, organometallic, inorganic, and physical chemists, he specializes in
the experimental application of solid-state NMR techniques to
characterization or problem solving in a wide range of solid (and
sometimes not so solid) materials, including pharmaceuticals, catalysts,
ceramics and glasses, polymers (synthetic and natural), soils (and related
materials), and household products. He has co-authored 150 publications in
the scientific literature.
Robin K. Harris is an emeritus professor at the University of Durham, UK,

where he previously served as professor of chemistry, head of the Physical
and Theoretical Chemistry section, and chairman of the Chemistry
Department. He obtained his first degree from Cambridge University and
undertook research in NMR there, supervised by Norman Sheppard, for his
Ph.D. After 2 years as a postdoctoral fellow at the Mellon Institute,
Pittsburgh, he joined the (then new) University of East Anglia (Norwich,
UK) as a lecturer, winning promotion to a readership and then obtaining a
personal chair. He moved to the University of Durham in 1984. His early
research was on solution-state NMR, but, starting in 1976, he has carried
out pioneering research in solid-state NMR, using cross-polarization and
magic-angle spinning, for a wide variety of elements and in an extensive
range of chemical systems, including organometallics, polymers, ceramics,
and inorganics. In the last 20 years, much research has been directed
toward problems involving pharmaceutical compounds and systems,
especially relating to polymorphism and quantitative studies. Current
interests center around the concept of “NMR crystallography” and include
both sophisticated NMR experimental methods and computations of
chemical shifts using crystallographic repetition. Professor Harris was
awarded an Sc.D. degree by Cambridge University for his research in 1978.
He is a Fellow of the Royal Society of Chemistry and has won its awards in
Chemical Instrumentation (1985) and in Analytical Spectroscopy (1998).
He was the director of the UK National Research Service in Solid-State
NMR (1986–2004). He has been an author of over 500 research articles,
mostly on NMR, and has acted as author or editor of several books on the
subject. He is the senior editor-in-chief of the Encyclopedia of Magnetic
Resonance .
Paul Hodgkinson studied chemistry at Queen’s College, Oxford,

completing his Ph.D. in 1995 on the sampling of NMR data. His interests in
solid-state NMR developed during postdoctoral research at UC Berkeley
(Royal Society/NATO fellowship with Professor Alex Pines) and at the
Ecole Normale Supérieure de Lyon (Marie-Curie fellowship with Professor
Lyndon Emsley). He was appointed to a research fellowship in the
Chemistry Department, University of Durham, UK, in 1998 and is
currently a reader in magnetic resonance at Durham, where he directs the
Durham Solid-State NMR Research Service. His research combines
interests in technique development and methodology in solid-state NMR as
well as applications to chemical problems. His group develops NMR theory
and numerical simulation software to explore the dynamics of large
coupled spin systems and applies solid-state NMR in the area of structural
chemistry, particularly of systems with mobility such as soft solids and
solvates, and to pharmaceutical solids. A particular interest is in combining
information from diffraction-based experiments, NMR and computation of
NMR parameters (using DFT codes), and dynamics (molecular dynamics
simulations). He has authored/coauthored over 60 research articles in the
area of NMR.
CHAPTER 1
INTRODUCTION
1.1 THE UTILITY OF NMR
This chapter gives a brief introduction to solid-state nuclear magnetic

resonance (NMR). However, in order to understand its applications,
practitioners need to be aware of a number of facets of solid-state structure,
so these are described. The final section introduces the three key techniques
of high-resolution NMR of solids and the types of nuclei for which
different approaches are needed.
NMR spectroscopy is undoubtedly one of the most powerful techniques for

determining molecular-level structure and dynamics. Three main factors
contribute to this situation. Firstly, it can be applied to the vast majority of
samples. Secondly, nearly all elements have spin-active nuclides that can be
accessed—and the resonance regions do not generally overlap (i.e., the
spectra are isotope specific). Thirdly, under suitable experimental
conditions the resolution is extremely high so that small differences in the
electronic environment of atoms result in observably different resonance
frequencies.
The first successful NMR experiments on condensed phases (conducted by

physicists in late 1945) encompassed both solids and solutions. However, it
soon became clear that (i) resolution fo r solutions was orders of magnitude
better than for solids and (ii) the solution-state spectra gave molecular
structure information of great value to chemists. This stimulated the
production of commercial spectrometers for solution-state work. For
several decades, the overwhelming majority of NMR applications were
carried out for solutions by chemists. The use of solid-state NMR remained
the preserve of a handful of physicists and physical chemists.
In order to understand why NMR of solids is so different from solution-

state NMR, it is necessary to discuss the nature of the solid state and the
various factors affecting NMR spectra, as in the sections below. Many
books describe solution-state NMR (see Further reading), and graduate
chemists will certainly have received lectures on the topic, so the present
book assumes a certain amount of background knowledge. The current
chapter will discuss the nature of the solid state and will introduce the ways
in which solid-state NMR is distinguished from its solution-state analog, in
spite of the fact that the same factors are at work.
NMR spectra are generally obtained from samples in strong magnetic fields
by recording the response to radiofrequency (RF) radiation, which is
normally applied in the form of short pulses. The detailed behavior of the
nuclear spins depends on the RF pulsing regime, on the spin properties, and
on the interactions of the spins. Thus, one must evaluate the terms
governing the energy of the spins in the magnetic field and subject to
irradiation. The Zeeman effect of the spin magnets in the applied magnetic
induction field (B 0 —conventionally in the z direction) almost always
forms by far the largest contribution to the energy. The full Zeeman energy,
E Z , involves a sum over all the nuclear species in the sample. For a given
spin j (ignoring shielding), it is expressed as:
CH_01-01.wmf 1.1
where h is Planck’s constant, γ j is the magnetogyric ratio of the nucleus

concerned (i.e., a measure of its magnetic strength), and m j is the spin
component quantum number of nucleus j.
Magnetic fields arising from RF radiation form small perturbations in

energy and cause transitions . The remaining interactions, which are all
internal to the sample and therefore give chemical information, are of four
main types,1 one of which (shielding) modifies the Zeeman effect, while
the other three involve spin couplings (indirect, dipolar, and quadrupolar)
and mostly lead to splittings in the spectra. These four interactions give rise
to energies designated here as E S , E J , E D , and E Q , respectively. Thus
the total energy of an ensemble of spins, E NMR , is the sum of the several
effects:
CH_01-02.wmf 1.2
where E RF refers to the interaction of the spins with applied

radiofrequencies. The “internal” interactions, namely E S , E J , E D , and E
Q , are discussed in separate sections in chapter 2 . The important feature to
note here is that these internal interactions all depend on the orientation of
the relevant molecular-level fragment in the applied magnetic field of the
NMR experiment (i.e., they are anisotropic). Thus for isotropic solutions,
where the molecules are rapidly and chaotically tumbling, the relevant
energies are averaged over all orientations. However, for solids, with
relatively static molecules (and certainly not the complete motional
disorder of solutions), the internal interactions depend strongly on the
orientation of the solid in the magnetic field, giving a significantly more
complex situation than for solutions.
An extra term is required in equation 1.2 when paramagnetic systems are

involved because the unpaired electrons can couple with nuclei. While this
book concentrates on diamagnetic systems, which attract the vast majority
of NMR studies of solids, section 7.5 discusses some of the special effects
seen in paramagnetic systems.
1.2 A PREVIEW OF SOLID-STATE NMR SPECTRA
In order to give a taste of what is to come in this book, a few examples of

spectra are given at this point. Figure 1.1 compares 13 C spectra for a
solution of alanine and for a microcrystalline sample, with low-power
proton decoupling for the former, as normally used for solution-state
spectra. The overarching reason for the difference is, as described above,
the extensive averaging of various interactions in the solution state, caused
by rapid random isotropic tumbling of the molecules. Thus the effects of
anisotropies in the interactions (i.e., of variations with the molecular-level
or sample orientation in the applied magnetic field B 0 ) are removed from
the spectra, causing considerable simplification (and therefore high
resolution)—but simultaneously losing information. Some solid-state
situations give better resolution, for instance, molecular solids when the
molecules are highly mobile, as is the case for plastic crystals such as
adamantane (see inset 7.8). Inorganic compounds lacking hydrogens can
give better-defined spectra, as for the 207 Pb spectrum of lead nitrate,
shown i n figure 1.2 . Figures 1.1 (upper trace) and 1.2 (upper trace) are for
microcrystalline samples, yielding what are referred to as powder-pattern
spectra . It is possible, though unusual, to record NMR spectra of single
crystals (a rather large crystal is required!). A single crystal of lead nitrate
would give a relatively sharp singlet 207 Pb spectrum rather than a powder
pattern, but its chemical shift will depend on the orientation of the crystal in
the magnetic field, as discussed in sections 2.3 and 8.2.4. For a
microcrystalline powder, there will be a random (statistical) distribution of
orientations, so the spectrum consists of a summation of many such sharp
lines at different shifts, which overlap to form the powder pattern.
Figure 1.1. Carbon-13 spectra of alanine in the solid state (upper trace) and
in solution (lower trace). The latter was obtained using proton decoupling,
as is usual for solutions. The former involved no special techniques.
ch01-F02.ai
Figure 1.2. Lead-207 NMR spectra of microcrystalline lead nitrate under

static ( upper trace ) and magic-angle spinning (lower trace) conditions .
The former is an example of a powder pattern. The latter is discussed in
section 1.5. Magic-angle spinning (including spinning sidebands, indicated
in this figure by asterisks) is discussed in section 2.7.
1.3 THE SOLID STATE
1.3.1 INTRODUCTION
The principal way in which solids differ from fluids lies in the relative lack
of molecule-level motion. In isotropic solution, molecules and ions tumble
rapidly, randomly, and chaotically. Therefore, any properties that depend on
molecular orientation become averaged to their isotropic values. The
internal NMR interactions of importance come into this category. Such
interactions each involve products of two vectors, so they are classed as
tensors (see chapter 2 ). In fact, two of the most important features, namely
the electric quadrupolar and magnetic dipolar interactions, are averaged to
zero by isotropic tumbling and therefore do not affect the NMR transition
frequencies of solutions. Simple interpretations of spectra can ignore their
existence, though they do affect relaxation times. Moreover, the other two
important interactions (shielding, which gives rise to chemical shifts, and
indirect coupling) are simplified so that for solutions a single scalar
chemical shift results for each chemical site and a scalar coupling constant
(symbol J ) exists for every pair of nuclear spins.
There are many types of solids and solid systems. The simplest (and most
commonly met with in chemistry research) is the crystalline state. In
“ideal” crystalline samples, the atoms (whether in molecular or in
framework solids) are effectively locked into positions. The structures as a
whole are described by unit cells wherein the atomic positions can be
determined by diffraction methods. The orientation of a crystal in B 0 will
determine the resonance frequency for each NMR-active nucleus.
1.3.2 SYMMETRY IN THE CRYSTALLINE STATE
The concept of a unit cell underlies the translational symmetry inherent in

crystalline solids. NMR spectra reflect the contents of the unit cell so that
resonances can be interpreted in terms of only the unit cell. Of course,
atoms in neighboring unit cells will influence the chemical shift of a
nucleus in a given cell. However, such influences die away very rapidly
with distance. NMR thus responds to short-range effects, in contrast to
diffraction techniques, which largely rely on long-range order.
For NMR purposes, it is important to distinguish between the effects of

different symmetry elements on the equivalence or otherwise of molecules
or atomic groupings. For rigid fully-ordered structures, translational
symmetry means that each unit cell is identical and therefore gives the
same NMR spectrum, that is, spectra directly reflect the contents of the unit
cell. However, there may be several chemically identical sites for a given
nucleus, related by symmetry other than translation. Consider molecular
systems (similar matters arise for framework structures). Molecules related
by a center of symmetry (figure 1.3 (a)) are fully equivalent and give
identical spectra. However, molecules related by other symmetry elements,
such as planes or axes of symmetry (figure 1.3 (b)), may give different
spectra in some experiments (see below) and so are only conditionally
equivalent. In other situations, there may be two or more whole molecules
in a unit cell which are not related by symmetry (figure 1.3 (c)) and are
therefore classed as independent . In fact, the minimum group of atoms
which, when repeated by all the symmetry of the crystal, gives the contents
of the whole unit cell, forms what is known as the crystallographic
asymmetric unit . Figure 1.4 presents a practical case, that of the α form of
testosterone, for which the asymmetric unit consists of two molecules,
labeled U and V. There are analogous situations for network structures of
solids; again, one can define an asymmetric unit.
ch01-F03.ai
Figure 1.3. Schematic illustration in two dimensions of (a) fully equivalent

(under the influence of a center of symmetry), (b) conditionally equivalent
(under the influence of a mirror plane), and (c) nonequivalent molecules of
the same chemical species in a unit cell. The symbol Z is conventionally
used for the number of molecules in the unit cell, whereas Z ′ refers to the
number of molecules in the crystallographic asymmetric unit (see the text).
ch01-F04.ai
Figure 1.4. Left: The two independent molecules (labeled U and V) in the
asymmetric unit of the a form of the steroid testosterone. These have
different environments and slightly different geometries. Right: Carbon-13
spectrum of α -testosterone (high-frequency region only), obtained under
conditions of proton decoupling and magic-angle spinning (see section 1.5)
to show crystallographic splittings. The asterisk indicates a spinning
sideband (see section 2.7).
1.3.3 EFFECTS OF CRYSTAL STRUCTURE ON NMR

As stated above, solid-state NMR spectra reflect the situation of all the
nuclei in the unit cell (in contrast to solution-state spectra, which reflect the
situation of the isolated molecule). There are at least four distinguishable
ways in which the internal structure of a static crystal will give rise to a
larger number of resonances than for a solution of the same compound:
For a static single crystal, atomic sites which are in identical chemical
positions will, in general, be in different orientations with respect to B
0 and will therefore give rise to different resonance frequencies. Then
more peaks are obtained than for the solution state, forming
crystallographic splittings . However, such splittings may be
eliminated by the technique of magic-angle spinning (MAS),
discussed later (section 1.5).
Independent molecules in an asymmetric unit will differ in both their
intermolecular (packing) and intramolecular (geometry) arrangements.
Therefore the environment of chemically analogous spins in (say) two
independent molecules will be different, and so they will give rise to
different resonance frequencies—another (more commonly observed)
cause of crystallographic splittings. This type of splitting cannot be
eliminated by MAS. A partial 13 C spectrum of α-testosterone is
shown in figure 1.4 to illustrate this phenomenon. It is difficult to
disentangle experimentally the influences of packing and geometry on
NMR spectra.
Moreover, a molecule that has symmetry in its isolated state (as in
solution) may well lose its symmetry in the crystalline state. Thus, for
instance, if the two sides of a paraphenylene group (related by
symmetry in the isolated molecule) have different local environments
in the crystal, the relevant spins will have different chemical shifts. In
such a case, the whole paraphenylene group is (part of ) the
asymmetric unit, though if the symmetry of the isolated molecule is
retained, the asymmetric unit may be half the molecule (plus any atom
on the effective mirror plane or symmetry axis relating the two sides
of the ring). The two situations are illustrated schematically in figure
1.5 .
The pharmaceutical drug barbital provides an example (figure 1.6 ). In
this case the intramolecular lack of symmetry induced by the
molecular environment in the crystal also extends to the two ethyl
groups in some forms. The structure of polymorphic form III (figure
1.6 (a)) consists of molecular chains in which the two halves of the
ring are related by symmetry, as are the two ethyl groups, so the
asymmetric unit is half a molecule. This means that only single
resonances are seen for the methyl and methylene carbons (as would
be the case for a solution-state spectrum). However, form I has a
double-chain structure (figure 1.6 (b)) in which there is no such
relationship, so an entire molecule is the asymmetric unit. Thus the
methyl and methylene signals are split into two. Note, however, that
the ring carbon signal is unsplit.
Finally, molecular-level motion, which is fast on the NMR timescale

(see inset 1.1) in the solution state, may well be slowed in the
crystalline state, causing further nonequivalences to be revealed. Thus
the methyl carbon nuclei of a C–C(CH 3 ) 3 group may be effectively
equivalent in solution because of rapid internal motion about the C–C
bond, therefore giving a single 13 C signal, but when the t-butyl group
is in a fixed position in a crystal, the environments of the methyl
groups may differ, resulting in three 13 C signals.
ch01-F05.ai
Figure 1.5. Schematic diagram to show the equivalences and

inequivalences for carbon nuclei in a phenylene ring in different situations.
The rectangles beside the aromatic rings are simply indicative of the local
environment. In the general case (only) two resonances will be seen for the
two carbons ortho to substituent X (and similarly for those ortho to Y). The
dotted line in the right-hand case refers to a twofold rotation axis or a plane
of symmetry.
ch01-F06.ai
Figure 1.6. Left: Structures of barbital polymorphs, showing the molecular

chains that affect the relationships between the two halves of the ring.
Right: The corresponding 13 C spectra (omitting the C=O signals). (a)
Barbital III. (b) Barbital I. The three signals (or groups of signals) are
assigned to (right to left) the methyl, methylene, and ring carbons.
Inset 1.1. NMR timescales

There are a number of timescales involved in the NMR experiment:
(a) The acquisition time to record a single free-induction decay (from tens
of milliseconds to several seconds; see inset 3.2). It is this range that is
generally referred to as the NMR timescale . If molecules are essentially
static (apart from vibrational motion) over such times, then the spectrum
will be a superposition of those for all molecular positions (e.g.,
conformations) present. However, if there is rapid motion, that is over times
shorter than this, an averaged spectrum will be obtained (see section 7.3).
Typically, the critical motional rates are those comparable to chemical shift
differences between corresponding nuclei for the different molecular
positions.
(b) The total accumulation time taken to record a spectrum, which may be
from minutes to hours or even days, depending on the number of free
induction decays that are recorded. If the compound is stable over this time,
its spectrum will be uniquely obtained. If reaction occurs, the spectrum will
be a superposition of all the compounds present.
(c) Times relating to relaxation, ranging from microseconds to seconds (or

even up to hours in some cases; see section 2.8).
It is actually a misconception to think that atoms and molecules in solids

are static. There is, for example, always vibrational motion. Moreover,
many crystals have much more motion. Whole molecules or parts of them
may be rotating rapidly. Thus C-methyl groups commonly rotate rapidly on
the NMR timescale about the C–CH3 axis at ambient temperatures,
rendering the protons equivalent. Phenyl and phenylene groups often
undergo rapid 180° ring-flips about their axes. Ring inversion may occur
for cyclohexyl groups. Such processes lead to averaging of resonances for
solids just as they do for solutions, though usually the motional rates are
much higher in the latter state. Significant motion of molecules as a whole
leads to the plastic state. An example is adamantane, which has a roughly
spherical molecular shape (see inset 7.8) and so can readily rotate
isotropically in its crystalline form (though the molecules do not translate).
In other cases, atoms, molecules, and ions can translate through the crystal
lattice, as happens in ion conductors such as lithium-based battery
materials. Systems with channel or tunnel structures, such as zeolites and
urea complexes, may contain mobile guest molecules. In all these cases,
there may be substantial averaging of the NMR spectrum.
1.3.4 TYPES OF SOLIDS
There are a number of types of solid state that are not crystalline or not
perfectly crystalline. Some solids exhibit disorder. This may take a number
of forms; in particular, it may be static or dynamic (spatial or temporal
disorder). In either situation, diffraction techniques report occupancy
factors at the possible sites, implying that there are fractional atoms there.
In the former case, adjacent “unit cells” may have some atoms in different
positions, which do not change with time. In the latter case, there may be
exchange (over a particular timescale) of atoms between two or more
possible positions for any given unit cell; the rate of exchange will
determine the appearance of the NMR spectrum (see section 7.3). An
example is the 3:2 complex of phenol and triphenylphosphine oxide
(TPPO). A partial structure is shown in figure 1.7 . The phenol molecule is
rotationally disordered (and was thus solved in a CH_In_01-01.wmf
space group), with arrows indicating two notional oxygen half-atoms. With
this molecule fixed in one of its two arrangements, the phosphorus atoms of
the two TPPO molecules are nonequivalent. At the lowest temperature this
is the case: the molecules are static on the NMR timescale, so two 31 P
signals appear. However, as the temperature increases, rotation of the
phenol molecule becomes rapid, causing the signals to merge because an
average situation is observed by NMR. The NMR proves that the disorder
is dynamic (temporal, not spatial).
ch01-F07.ai
Figure 1.7. Left: Partial structure of the unit cell contents of a 3:2 adduct of
phenol and triphenylphosphine oxide. The other two phenol molecules in
the unit cell, which are not disordered, are omitted for clarity. Right:
Phosphorus-31 spectra as a function of temperature.
The different sites in disordered systems do not need to occur with rational
occupancy ratios. Many inorganic crystalline systems exist in which there
is a fractional excess of some types of atom (over and above the nominal
stoichiometric amount). In other cases, there is a deficiency of some types
of atom. Such systems are known as defect structures, and their nature has
consequences on the NMR spectra (e.g., on relative intensities of
resonances).
Yet other structures are classified as incommensurate . They contain

different chemical entities with translational repetition distances that do not
bear any rational relationship. An example is discussed in section 1.4.
Many pure (homogeneous) systems are amorphous. These may be glassy

(i.e., relatively rigid at the molecular level) or rubbery (much more mobile).
Some organic systems readily produce a supercooled (rubbery) state when
cooled rapidly to below their freezing points. As temperature is lowered
further, they may pass into the glassy condition at the glass transition
temperature , T g . All amorphous systems contain the same atoms and
chemical groups in a range of local environments, and so they give rise to a
corresponding range of NMR transition frequencies, thus causing line
broadening, as shown in figure 1.8 . Actually, apparently amorphous
polymers and fibers (including naturally occurring systems such as silk)
can have a degree of order in terms of a preferred orientation . Thus
polymers drawn through a hole may have chains preferentially oriented
along the draw direction. Such preferences can be studied by NMR using
static samples.
ch01-F08.ai
Figure 1.8. Carbon-13 NMR spectra of the organic compound nifedipine,

obtained under conditions of high-p ower proton decoupling and magic-
angle spinning, showing the line broadening for the amorphous state (lower
spectrum) compared with the crystalline state (upper spectrum).
Moreover, even chemically homogeneous solids may be physically

heterogeneous, some with relatively simple biphasic structures (for
instance, many semicrystalline homopolymers have amorphous and
crystalline domains). Chemical heterogeneity is exceedingly common (e.g.,
wood and other natural products), resulting in highly complex structures
and correspondingly crowded NMR spectra. Crystalline compounds may
have many defects—and naturally the nuclear spins near such defects or at
the surfaces of microcrystalline particles have different environments to
those in the core of a perfect single crystal and so give rise to different
chemical shifts in an NMR experiment.
A final type of physical system that needs special consideration is the liquid
crystalline state, which can be considered as a half-way house between the
isotropic solution and crystalline solid states. Such systems have
considerable mobility analogous to that of a liquid but possess order
analogous to that of a solid in some dimensions. The behavior of liquid
crystals in NMR experiments is a specialized topic which will not be
discussed in this book. Suffice it to say here that dipolar and quadrupolar
interactions are not averaged to zero for liquid crystals and therefore
influence the resonance frequencies.
Thus questions of both spatial order/disorder and temporal disorder

(mobility) strongly affect the experimental methods used in solid-state
NMR and/or the quality of the spectra obtained. Figure 1.9 illustrates the
relationships between the two types of disorder for a variety of condensed-
phase situations.
ch01-F09.ai
Figure 1.9. Classification of different types of condensed-phase matter

according to their spatial and temporal disorder.
1.4 POLYMORPHISM, SOLVATES, CO-CRYSTALS & HOST:GUEST

SYSTEMS
Chemical compounds frequently exist in more than one crystalline “form.”

This may affect only the external appearance or habit of the crystals, with
no change in the underlying structure. However, in many cases, the internal
structure differs in some way, resulting in the phenomenon of
polymorphism (the corresponding situation for structures of elements is
known as allotropy). Carbon is a classical example of allotropy since it can
exist as diamond or graphite or the series of fullerenes such as C60 (with
the soccer football shape). Polymorphism is, in fact, ubiquitous in
chemistry, occurring for organic and inorganic compounds, as well as
polymers and natural products (e.g., cellulose). The example of barbital has
already been mentioned (see figure 1.6 ). The classification of polymorphs
represents some difficulties, but here network systems (such as the various
forms of silica) and molecular polymorphs, which are of considerable
importance in the pharmaceutical industry, can be distinguished.
All properties of polymorphs for a given compound differ, at least in

principle. For instance, dissolution rates of the polymorphs and their
equilibrium concentrations will differ. However, their solutions will behave
identically. Polymorphs are obtained in a variety of ways (for instance, by
crystallization from different solvents). A pair of polymorphs can be either
monotropic or enantiotropic. In the former case only one of the forms can
be stable (under a range of temperatures and pressures), whereas the other
is always metastable (i.e., not thermodynamically stable but often with a
long lifetime). In enantiotropic cases, however, the two (or more) forms are
each stable, under different conditions, and there will be a transition
temperature at which they will interconvert (depending on the pressure).
The situation may be expressed in a phase diagram (see figure 1.10 ),
which shows the regions of pressure–temperature space for the various
states of a pure compound. However, transition rates between polymorphs
can be extremely slow so that it is common to be able to obtain several
polymorphs (even in monotropic cases) at, say, ambient temperature, which
clearly makes their study easier. For more information on phase diagrams
and the thermodynamics of polymorphic systems, see Further reading.
ch01-F10.ai
Figure 1.10. Schematic phase diagram for an enantiotropic system. The

regions of stability for the two solid forms, A and B (the latter in gray), are
indicated. The dashed lines show the boundaries of form A existence in the
absence of form B (e.g., if transitions from A to B are slow). L = liquid
phase; V = vapor phase.
In addition, many organic compounds form stoichiometric solvates or co-

crystals, for example as clathrates, in which a host system accepts a guest,
either with a change of crystal structure or with retention of the host
structure with minimal adjustment to the presence of the guest.
Sulfathiazole, for instance, has been shown to participate in over a hundred
solvates. When the guest molecule is solid in its pure state, the term co-
crystal is frequently used instead of solvate. The most important class of
solvates involves water as the guest, in which case the term hydrate is
appropriate. The water molecules often act by hydrogen bonding to hold
the structure together. In some cases, hydrates with different
stoichiometries (and different crystal structures) exist.
NMR proves to be one of the best ways of characterizing polymorphs and

solvates, though it is always wise to use it in conjunction with other
techniques such as vibrational spectroscopy and X-ray diffraction, together
with thermal methods.
In many cases, the basic structure of the host compound contains cages or
channels that can be occupied by a range of guest molecules. Zeolites, for
instance, often contain both cages and channels, which result from the use
of organic molecules as templates during synthesis. Another example is
provided by the complexes of small organic molecules with urea. The urea
provides a host structure (unstable in the absence of guest molecules!)
containing linear tunnels into which elongated molecules such as alkanes
and their simple derivatives can fit. However, the repeat distance of the
urea host in the tunnel direction is not in general the same as the repeat
distance of the guest molecules, which is governed by their length. Such
structures fall into the class of incommensurate systems mentioned in
section 1.3.4. Frequently, the guest molecules in such complexes are highly
mobile in the host tunnels and high-quality spectra can therefore be readily
ob tained (figure 1.11 ). Diffraction te chniques struggle to precisely locate
the guest molecules in the tunnels because of their high degree of dynamic
disorder.
ch01-F11.ai
Figure 1.11. Top left: Structure of the inclusion compound between urea
and 1-fluorononane as determined by X-ray studies at ambient temperature,
viewed down the tunnels formed by the urea host. Note that the disordered
guest molecules do not give rise to Bragg scattering and so the tunnels
appear to be empty in the reported structure. Bottom: The low-frequency
part of the 13 C spectrum (showing the signals arising from the guest
molecule only), obtained under conditions of high-power proton
decoupling and magic-angle spinning (see section 1.5).
Sometimes guest molecules can be incorporated into host crystal structures

for a wide range of guest:host ratios, that is, in non-stoichiometric fashion.
Such situations are highly likely when suitably sized channels or tunnels
are present in the host chemical structure, allowing ready ingress or egress
of guest molecules. For instance, the compound sildenafil citrate (the active
principle of Viagra ® ) readily accepts water molecules—to a variable
extent depending on the humidity of the environment.
Chiral molecules may form crystals that each contain only one enantiomer,
but in such cases usually as a mixture of crystals of the two forms (a
conglomerate ). The NMR spectra of the two pure enantiomers (and of the
corresponding conglomerate) will be identical. Alternatively, crystals of an
internal racemate may form, with the enantiomers in a unit cell related by a
center or a plane of symmetry. The spectrum of a racemate will in general
differ from that of the related enantiomers, thus providing a powerful way
of distinguishing between the two situations.
1.5 NMR OF SOLIDS & THE PERIODIC TABLE
As discussed above, the nature of the crystalline state implies that NMR
spectra will be more complex than those of isotropic solutions. Additional
considerations arise when microcrystalline powders are examined by NMR
(as is the normal case), since each crystallite will be separately oriented in
the applied magnetic field. This results in powder-pattern spectra for static
samples, as illustrated earlier. Three particular techniques, which are
discussed in more detail later, are of considerable importance in obtaining
high-quality, well-resolved spectra, especially from powdered samples.
Two of these relate to improvements in resolution, while the third is
concerned with increasing signal intensities:
High-power proton decoupling (HPPD) . A typical problem for observing

spectra (of, say, 13 C or 15 N) of organic or organometallic compounds
arises because they contain many protons. Heteronuclear dipolar
interactions such as (13 C,1 H) or (15 N,1 H) can be very strong (up to ~ 50
kHz), which causes extensive line broadening. The difficulty may be
overcome by high-power heteronuclear proton decoupling (see sections
3.3.7 and 5.1.2) analogous to proton decoupling for solution-state NMR,
though the latter requires only low-power irradiation since it is required
merely to eliminate “scalar” (J ) coupling (which is much weaker than
dipolar coupling for solids). High-power decoupling for solids eliminates
scalar as well as dipolar coupling.
Magic-angle spinning (MAS). Another problem of spectral resolution

arises because of the existence of shielding anisotropy, which, as mentioned
earlier, implies that identical nuclei in different crystalline particles (i.e., at
different orientations to B 0 ) will have different chemical shifts, resulting
in line broadening (up to hundreds of parts per million for 13 C and even
more for heavy-metal nuclei). Very rapid sample spinning about an angle of
54°44′ (the “magic” angle) suffices to average orientation effects and to
produce relatively sharp lines. It may be noted that in principle this also
averages dipolar interactions, but higher rotation rates are usually required,
which are not readily achieved.
Cross polarization (CP) . The third problem, which affects signal intensities
rather than resolution, is that relaxation times for dilute spins in solids tend
to be long (generally tens to hundreds of seconds, but in some cases as
much as hours), rendering the usual pulse-and-acquire technique for
recording free induction decays very inefficient (i.e., requiring long inter-
pulse delays; see section 3.3.1). This difficulty is ameliorated by the use of
CP from protons to the dilute spins (achieved by a pulse sequence as
described in section 3.4).
The difficulties involved in obtaining good-quality spectra of powdered

solids also depend on the type of nuclear spin involved. Since the
properties of nuclides (spin quantum number I , natural abundance x ,
magnetogyric ratio γ , and quadrupole moment Q ) are all-important for
obtaining NMR spectra, they are listed in appendices A and B. Roughly
speaking, four categories may be distinguished:
“Dilute” or “rare” spin- CH_In_01-02 bOLD.wmf nuclides . In these

cases, all three problems mentioned above are important. Some atoms are
“dilute” because their natural isotopic abundances are low. Typical cases
are 13 C, (1.07%) and 15 N (0.368%). These can be made more abundant
by isotopic enrichment, which is commonly practiced for biochemical
NMR studies. However, it is not only isotopically dilute nuclides that fall
into the “dilute” category. A nucleus such as 31 P, which is present in 100%
natural abundance, may be chemically dilute, as, for example, for a large
biomolecule containing a single phosphorus atom. In fact, one can also
envisage a physically dilute situation as for molecules in an inert matrix,
though this would probably not be a crystalline situation. In practice in
most situations nearly all spin- CH_In_01-02A.wmf nuclei fall into the
“dilute” category as far as the application of NMR techniques is concerned.
Dilution limits sensitivity, so CP is usually vital, but for reasons discussed
above so are HPPD and MAS. Such experiments form what is usually
termed CPMAS NMR (the high-power proton decoupling being assumed).
The combination of CP with MAS and high-power proton decoupling was
first successfully implemented (for 13 C in organic systems) in 1976, which
may be regarded as the birth of modern solid-state NMR, though the three
techniques had already been known individually for several years.
“Abundant” spin- CH_In_01-02 bOLD.wmf nuclides . Typically, such

cases involve protons (because these are not only isotopically abundant, but
also form a high proportion in number of atoms in most organic and
organometallic compounds). For highly fluorinated compounds, 19 F nuclei
may also act as “abundant” spins. In relatively rare situations, other nuclei
may fall into the abundant category, for example, by isotopic enrichment.
These situations involve strong homo nuclear dipolar interactions, which
cause very substantial line broadening. In principle, MAS can overcome
this problem also, but until the beginning of the 21 st century, the required
spin rates were not available, and the “classical” technique for overcoming
this problem is so-called multiple-pulse decoupling (see section 5.5), which
is a technically demanding method of homonuclear decoupling. Of course,
if protons are chemically dilute for some reason, this difficulty does not
arise and MAS at modest rates may suffice. Heavy deuteration can be used
to artificially dilute the proton content of compounds.
Quadrupolar nuclides with half-integral spin quantum numbers . When

nuclides have spin quantum numbers CH_In_01-02B.wmf they have not
only magnetic dipole moments but also electric quadrupole moments,
which interact with electric field gradients at the nuclei to give quadrupolar
coupling. In fact the majority of the elements in the periodic table have at
least one quadrupolar nuclide. The quadrupolar coupling constants can be
very large (up to several hundred megahertz), so it is usually not possible to
eliminate these by MAS. Fortunately, when the spin quantum number is an
odd multiple of CH_In_01-02A nEW.wmf the central
CH_In_01-02C.wmf transition is unaffected by first-order quadrupolar
coupling, as will be explained further in section 2.6. The broadening
influences of proton dipolar coupling and chemical shift anisotropy can be
removed and spectra involving this transition sharpened by using high-
power proton decoupling and MAS respectively.
Quadrupolar nuclei with integral spin quantum numbers . Such nuclei are in
principle the most difficult to deal with, since they have no central
transitions. Fortunately, there are very few such nuclides, the principal ones
being 2 H, 6 Li, 10 B, and 14 N. In each case, there is an alternative
nuclide. However, it is sometimes valuable to have spectra of one or other
of these four (though not normally of 10 B). In fact, the quadrupole
moments of 2 H and 6 Li are small, so their spectra present no problem and
only 14 N can be said to still represent a difficulty (see section 2.6).
FURTHER READING
GENERAL TEXTS ON NMR
“Nuclear magnetic resonance: Concepts and methods ”, D. Canet, John

Wiley & Sons Ltd. (1996), ISBN 0 471 94234 0.
“Understanding NMR spectroscopy ”, J. Keeler, John Wiley & Sons Ltd.

(2005), ISBN 0 470 01786 4.
“Magnetic resonance in chemistry and medicine ”, R. Freeman, Oxford

University Press (2003), ISBN 0 19 926225 X.
GENERAL TEXTS ON THE SOLID STATE
“Crystalline solids ”, D. McKie & C. McKie, Thomas Nelson & Sons Ltd.
(1974), ISBN 0 17 761001 8.
“Fundamentals of crystallography ”, Ed. C. Giacovazzo, Oxford University
Press (1992), ISBN 0 19 8555 79 2.
“Molecular crystals ”, 2nd edition, J.D. Wright, Cambridge University

Press (1995), ISBN 0 521 46510 9.
“Solid state chemistry ”, L. Smart & E. Moore, Chapman & Hall (1992),
ISBN 0 412 40040 5.
“Basic solid-state chemistry ”, A.R. West, John Wiley & Sons Ltd. (1988),
ISBN 0 471 91797 4.
“Polymorphism in the pharmaceutical industry ”, J. Bernstein, Oxford

University Press (2002), ISBN 0 19 850605 8.
SPECIAL TEXT ON LINKING NMR AND THE SOLID STATE
“NMR crystallography ”, Eds. R.K. Harris, R.E. Wasylishen & M.J. Duer,
John Wiley & Sons Ltd. (2009), ISBN 978 0 470 69961 4.
NOTES
1 Spin-rotation interactions, which do not generally affect spectra of solids,

will be ignored herein.
CHAPTER 2
BASIC NMR CONCEPTS FOR SOLIDS
2.1 NUCLEAR SPIN MAGNETIZATION
This chapter is designed to give readers an understanding of the nature of

tensor properties, in particular those that affect NMR. The present section
will cover some basic aspects of the behavior of nuclear spins. The material
can be found in many introductory NMR textbooks and it will be familiar
to anyone with an undergraduate knowledge of NMR. For this reason, the
section is short and the reader is encouraged to explore other texts for
additional details.
A moving (spinning) charge, such as that associated with a nucleus of non-

zero magnetic quantum number, I , generates a magnetic moment. In the
presence of a strong magnetic field, as used in an NMR experiment, this
nuclear magnetic moment is quantized into missing image file directions;
that is, there are different spin states defined by the magnetic component
quantum number, missing image file The discussion at this point is
restricted to spin I = missing image file nuclei so that there are just two
such states. The low-energy state (when γ is positive) has a component of
magnetic moment aligned with the magnetic field and is labeled α
missing image file while the high-energy state (with a component
opposed to the field) is designated β missing image file . The energies of
these states are governed by the Zeeman effect, as mentioned in chapter 1
and expressed in frequency terms in equation 2.1:
missing image file 2.1
The transition between these states is the basis of NMR spectroscopy. The
energy difference between them is
and this equates to a resonance frequency (ν NMR ) through

so that missing image file 2.4
This frequency is often referred to as the Larmor frequency. In order to

understand or even describe (in pictures) the effect of an NMR experiment,
it is helpful to introduce some tools. Key to these is the behavior of a bulk
sample (as opposed to the quantum mechanical treatment of individual
nuclei). This description of NMR comes from the classical physics of
electromagnetism. For an ensemble of nuclear spins, such as is contained in
a typical NMR sample, the population (number of nuclei) in each spin state
is missing image file and missing image file respectively. At thermal
equilibrium, there is a Boltzmann distribution between the two states:
where k is the Boltzmann constant and T is the absolute temperature. This

ratio depends both on the applied magnetic field and on the nuclide in
question (see inset 2.1). Equation 2.5 predicts that there will be slightly
more nuclei (magnetic moments) aligned with the applied magnetic field
than opposing it. Thus, there will be a net (bulk) magnetization aligned
with the field. The magnitude, missing image file , of this bulk
magnetization for a sample containing N nuclei is given by
It is the behavior of this bulk magnetization that provides a convenient way

of describing an NMR experiment (see section 3.2).
The resulting spectra are influenced by the interactions mentioned in

section 1.1. Consequently, a more detailed description of the “internal”
interactions, which ultimately give useful chemical information about the
sample, is now given.
Inset 2.1. NMR and sensitivity

For 13 C, from equations 2.4 and 2.5 with γ = 6.728 × 10 7 rad s –1 T –1 at
a magnetic field of 9.4 T and at a temperature of 294 K, p β / p α =
0.999983. Put another way, for every 1 million nuclei in the β state, there
are only 17 more in the α state. This is the reason why NMR is, as
analytical techniques go, relatively insensitive.
2.2 TENSORS
All the internal interactions involved in NMR are orientation dependent, so

they are properly expressed as tensor properties. While this section will be
written in terms of a general tensor, R , the shielding tensor is chosen when
a specific example is needed. A tensor may be thought of as the link
between two vectors. Thus the shielding tensor, σ , links the magnetic field
arising from shielding of the nucleus by the electrons, missing image file
, to the applied magnetic field, missing image file :
Since vectors have three components each, any tensor R (in its Cartesian
representation) involves a 3 × 3 matrix, that is, nine components:
Thus, if we suppose that B 0 is directed along z , then the shielding field

has the following components:
In other words, the shielding field is not in general parallel to the applied
field (figure 2.1 ).
However, tensors are often symmetric, that is, R ji = R ij and, even when
this is not the case, the antisymmetric contributions have little influence on
NMR spectra and are usually ignored. In such circumstances, there are only
six distinct components to a tensor, and it is possible to choose axes
(forming the principal axis system, PAS) in which R is diagonal:
The terms R XX , R YY , and R ZZ are called the principal components of

the tensor (and the use of capital X , Y and Z indicates this). They provide
information of interest, as does the orientation of the PAS in an axis system
fixed in the molecule or crystal (involving the other three necessary
parameters). The transformation of a tensor from a general frame of
reference to the PAS (or vice versa) is carried out by a process of rotation
(see section 4.6 and Further reading).
missing image file
Figure 2.1. Schematic relationship of a shielding field to the applied field:

(a) isotropic site; (b) anisotropic site. Shielding ellipsoids are shown on the
right—ellipsoids indicate the magnitude of the tensor in different
orientations.
Actually, instead of R XX , R YY , and R ZZ being used, it is more

common to define three alternative quantities. One of these is the isotropic
average :
which is the quantity that would be observed in a solution-state experiment,

since it is invariant to rotational transformation. In many high-resolution
spectra of solids, this is also the only parameter that is measured. The other
two parameters are the anisotropy and the asymmetry , which are defined
(together with the choice of axis labeling) in inset 2.2.
Inset 2.2. Definitions of tensor anisotropy and asymmetry

In the case of shielding, there are two alternative definitions (with different
symbols) for anisotropy:
The relation between them is missing image file Either definition may be
used for shielding. However, ΔJ is the normal form for indirect coupling.
The asymmetry, η , is a dimensionless quantity lying between 0 and 1. It is

defined by:

Actually, as can be seen, the asymmetry is also a type of anisotropy. This
definition is common to shielding, indirect coupling, and quadrupolar
coupling. When a nucleus is at a site of axial symmetry, R XX and R YY
are necessarily equal, so η is zero, and this is normally the case for dipolar
coupling. In such a situation, the notations missing image file and
missing image file may be used for the unique (Z ) component and the
ones perpendicular to it respectively. The term missing image file is the
isotropic average (as defined in equation 2.11), while missing image file
is the anisotropy ΔR .
Ordering of X , Y , and Z is defined herein (for the NMR tensors listed in

table 2.1 , except the quadrupolar coupling (see page 29 )) by:
This choice looks a little odd until it is realized that the X , Y , and Z
components are in alphabetical order in the spectrum (see figure 2.2 ),
either R ZZ ≥ R YY ≥ R XX or R ZZ ≤ R YY ≤ R XX (depending on the
system studied).
The asymmetry becomes 1 when the Y component is midway between the

X and Z components.
The following sections give details of the four tensor properties most
important for solid-state NMR, namely shielding, indirect coupling, dipolar
coupling, and quadrupolar coupling. Table 2.1 gives some of the general
properties of these tensors. The equations in this book are written such that
the coupling parameters (and their anisotropies) appear as frequencies
(other texts use angular frequencies).
Table 2.1. Properties of tensors involved in NMR
Figure Missing
2.3 SHIELDING
When a molecule is placed in a magnetic field, the nuclei are magnetically

shielded by the presence of electrons, resulting in a shielding field as
expressed in equation 2.7. Shielding leads to the well-known phenomenon
of chemical shift . Isotropic chemical shifts have undoubtedly been the
most important items of information in solution-state NMR, allowing
chemical structures to be determined. Isotropic chemical shifts can also be
obtained for solids, as described in this book, and are similarly used,
though there are also special features that apply only to the solid state.
In general, however, electrons are not spherically distributed around any

given nucleus, so therefore such a field will be anisotropic. Thus, shielding
will be a tensor property, σ . It is a single-spin quantity since the primary
field, B 0 , is external to the system. The lead nucleus in lead nitrate, for
instance, is differently shielded when the local symmetry axis at the lead
nucleus is perpendicular to B 0 than when it is parallel to B 0 (see figure
1.2 ). These extreme tensor components are labeled missing image file
and missing image file respectively. When the molecular axis is at an
angle θ to B 0 , the shielding in the z (B 0 ) direction can be shown to be:1
Since the shielding field, B S , is orders of magnitude less than B 0 , any

components of B S perpendicular to B 0 can be ignored, so the total field
considered is:
Equation 2.1 must therefore be modified by the shielding energy, E S , and

the energy of an ensemble of spins of type j in B 0 therefore becomes:
for a given orientation θ , assuming shielding is the only important

interaction other than the Zeeman effect. The transition (Larmor) frequency
is correspondingly modified from equation 2.4, thus giving rise to a
chemical shift that depends on the electronic environment of the nucleus in
question. The total spin energy of a sample will involve summation over all
types and numbers of spins, j.
For a polycrystalline sample, there will be a complete spread of angles θ ,
so a range of resonances, constituting a shielding powder pattern , will be
observed. Such a spectrum for an axially symmetrical case was shown in
figure 1.2 . The resonance frequency depends on missing image file and
therefore becomes a measure of θ so that any preferred orientation (such as
occurs in a drawn polymer sample) can be observed as a departure from the
expected intensity distribution, recognized qualitatively and determined
quantitatively. The intensity of the signal at a particular frequency will
depend on the probability of the relevant angle θ . Clearly, θ = 90° is much
more common than θ = 0° (think of the many directions from the center of
the earth to the equator versus the unique directions to the poles), so the
intensity at missing image file will be much higher than that at
missing image file as seen schematically in figure 2.2(a) (and
experimentally in the upper spectrum of figure 1.2 ). The more intense edge
of the powder pattern will depend on whether missing image file which
will vary with the chemical system in question.
For a nucleus at a site of cubic symmetry (as for 19 F in crystalline CaF2 ,

for example), the three tensor components will be equal, so the shielding
will be independent of the sample orientation and a single relatively sharp
resonance will be seen.
In the more general case, there will be three different principal components
of the shielding tensor, as mentioned in section 2.2 above, and the powder
pattern will have, correspondingly, three turning points as in the three lower
spectra in figure 2.2 . The relevant equation for the effective shielding (to
be used in equation 2.18 for deriving energy-level diagrams and hence
transition frequencies) will be:
where θ j is the angle between the shielding principal axis j and the
magnetic field. This equation may be rewritten in terms of the anisotropy, ζ
, and asymmetry, η , as:

where θ and φ are the spherical angles defining the orientation of the Z
principal axis of shielding in the magnetic field B 0 .
missing image file
Figure 2.2. Schematic powder patterns of shielding for (a) an axially

symmetric case with negative shielding anisotropy, (b) a general case with
negative shielding anisotropy, (c) an analogous case with positive shielding
anisotropy, and (d) a case with an asymmetry of one. The bottom spectrum
also shows the narrow lines obtained from three particular crystallite
orientations, with the X and Y tensor components aligned along B 0 and for
an intermediate situation. Note that as the scale uses the normal convention
of frequency increasing right to left, the orientation of the lineshapes will
match experimental spectra.
For powder patterns of the types shown in figure 2.2 , the turning points
give the values of the principal components (but not the orientation of the
PAS in the molecular/crystallographic frame). To obtain reasonably
accurate data, a bandshape-fitting procedure is required, as discussed in
section 8.2.1. However, in situations of nearly axial shielding,
differentiation of σ XX from σ YY is difficult, that is, values of η below 0.2
are not determined accurately.
Shielding anisotropies are rather small for 1 H, but are up to ∼ 200 ppm for
13 C and can be massive for some metallic nuclides.2 Their values,
together with those for shielding asymmetries, can be used to assign
resonances to chemical sites. They also give important crystallographic
information and frequently distinguish signals far better than do isotropic
shifts. They help to understand electronic structure, but this is properly
obtained only when the experimental measurements are combined with
quantum mechanical calculations of shielding. One could say that
measurement of the full shielding tensor (including its orientation in B 0 )
provides an order of magnitude more information than the isotropic value
alone.
2.4 INDIRECT COUPLING3

This type of spin–spin coupling operates via electrons and so is classed as
“through bonds.” It therefore gives information on molecular-level
connectivity and questions of molecular conformation. The coupling
parameter is given the symbol J and is highly important for solution-state
NMR, but somewhat less prominent in solid-state NMR studies. This is
mainly because it is generally the smallest interaction in magnitude, being
normally less than 1 kHz for light elements (though for some heavy metal
nuclides, it can be in the region of 10 kHz). Solution-state NMR depends
only on its isotropic value (and hence it is often referred to as “scalar”
coupling), but in fact it is actually a tensor property, J . Any asymmetry in J
is usually neglected since it is expected to be generally small and difficult
to determine, but anisotropies ΔJ need to be taken into account for solid-
state spectra in some (relatively unusual, at least for 13 C NMR) cases. The
contribution of indirect coupling to NMR energies for a pair of nuclei j and
k in a sample is:
For the full coupling contribution to the energy, this term must be summed
over all pairs of spins. For molecular solids, indirect coupling is generally
confined to intramolecular interactions, but it can extend across hydrogen
bonds between molecules and therefore it can be used to recognize the
existence of such bonding. Its relationship to connectivity is particularly
valuable for solid-state NMR.
2.5 DIPOLAR COUPLING
Dipole–dipole coupling between spins corresponds to the classical

interaction between two magnets, with dipole moments μ j and μ k , and
thus operates “through space.” The dipolar energy for such a pair is given
by:
where r is the distance between the two dipoles (see figure 2.3 ).
missing image file

Figure 2.3. Interaction between two dipoles.
For a heteronuclear spin pair in the high magnetic field approximation, this
expression becomes (to first order in energy):
where θ is the angle between the internuclear vector and the magnetic field,
while D jk is the dipolar coupling constant between spins j and k. The latter
is given (in frequency units) by:4
Typical values of dipolar coupling constants are given in table 2.2 . The
tensor D has no antisymmetric component and, unless there is anisotropic
motion involving the two nuclear spin sites involved, it is also axially
symmetric. The isotropic component of D is zero, so dipolar coupling does
not affect resonance frequencies for solutions, although dipolar interactions
are a primary cause of relaxation.
For a single crystal of a substance with an isolated pair of spin-

missing image file heteronuclei, A and X, and a unique direction r AX ,
the A spectrum will be a doublet of spacing missing image file , which
varies in position and splitting magnitude with the crystal orientation (and
is zero if θ is 54 ° 44′). The X spectrum behaves identically to the A
spectrum. The angle θ can be evaluated by varying the orientation of the
single crystal in B 0 . Hence, D can be measured and r AX determined. For
a typical C–H bond distance (say, r AX = 0.11 nm), the dipolar coupling
constant is 22.7 kHz. Homonuclear spin pairs have more complicated spin
energies (see section 4.5.1) so that equation 2.23 is no longer adequate.
However, a similarly simple result is obtained, the only difference being
that a single doublet is seen, with a spacing of missing image file , that is,
there is an extra factor of missing image file .
Table 2.2. Some typical bond distances and corresponding dipolar coupling
constants
Table Missing
For polycrystalline powders, the distribution of orientations must be taken
into account, as shown above for shielding. In fact for a pair of isolated
heteronuclear spins (AX), the A spectrum will look like two superimposed
patterns of the type shown in figure 2.2(a) , one for each of the doublet
resonances, with reversal of direction for one of them. This is illustrated in
figure 2.4 . Such a powder pattern is known as a Pake Doublet ,5 the outer
limits of which correspond to θ = 0° and the intense “horns” to θ = 90° .
The separation of the horns is D , so the distance r AX can be obtained even
in the case of a powder. The X spectrum will be identical. The powder
pattern for an isolated pair of homonuclear spins will be a single similar
shape, but with a scaling factor of missing image file .
In fact, the classic case of a dipolar coupling is that of the proton spectrum
of gypsum, CaSO4 · 2H2 O, for which the protons in the water molecules
form relatively isolated spin pairs. Since this case involves homonuclear
interactions, the factor of missing image file , mentioned above, is
required for the spectrum. The unit cell contains two water molecules with
different orientations, so two doublets are generally observed for a single
crystal in a fixed setting of the crystal in B 0 , though for some special
settings only one doublet is observed. For one setting, a single line is
observed, showing that r HH for the two different water molecules can be
nearly at the magic angle θ = 54° 44′ to B 0 simultaneously. A full analysis
shows that the dipolar coupling constant is 30.7 kHz, so r HH is determined
to be 0.158 nm. Given the traditional difficulties of determining hydrogen
positions by X-ray diffraction, such NMR experiments are potentially
valuable.
missing image file
Figure 2.4. A schematic heteronuclear Pake Doublet for the A spin of an

AX spin system. The dotted lines show the two subspectra (for different
values of m I for the X spin). ν 0 refers to the transition frequency under the
influence of shielding.
In general, spectra involving dipolar interactions are much more complex

because they are affected by the influence of all the spin pairs in the
sample, with a huge variety of internuclear distances and orientations in the
applied magnetic field. It is difficult to adequately take this into account,
but suffice it to say that dipolar coupling causes extensive broadening of
spectra, as is especially true for proton NMR of solids, where a single
broad featureless line (over 50 kHz in width at half-height) is generally
seen for rigid organic systems in static samples (figure 2.5 ). It is also
typically the case for proton-coupled 13 C spectra, as illustrated in the
upper spectrum in figure 1.1 . A common aim of many attempts to
determine interatomic distances by NMR is the production of samples with
isolated spin pairs, for example by selective isotopic enrichment.
missing image file
Figure 2.5. Proton spectrum of a static sample of microcrystalline alanine.
As already stated, molecular mobility in the liquid phase averages dipolar

interactions to zero (except if the motion is anisotropic, as for liquid
crystals). Any motion at the molecular level in solids will cause some
averaging also, which will narrow resonances, as shown in figure 2.6 for
the proton spectrum of amorphous indomethacin (which is an active
pharmaceutical ingredient). In this case, motional narrowing occurs as the
sample passes through its glass transition temperature (see section 1.3.4).
(The motion will also simultaneously partially average shielding tensors.)
Obviously, measurement of dipolar coupling constants supplies values of

internuclear distances and so is an important feature of NMR
crystallography. It is even more valuable for amorphous systems, where
such information is otherwise difficult to obtain. However, the effects of
molecular-level mobility must be taken into account. If the crystal structure
is known from diffraction experiments, details of mobility may be inferred.
missing image file
Figure 2.6. Proton spectrum of indomethacin at various temperatures,

showing motional narrowing, which becomes significant above the glass
transition temperature.
2.6 QUADRUPOLAR COUPLING

For nuclides with spin quantum numbers greater than missing image file
(only), another form of coupling exists. This is an electric interaction, not
magnetic, and involves the nuclear electric quadrupole moment , eQ , of the
nuclide. An introduction to this topic will be given here, but details are
presented in chapter 6 . Values of Q are listed in tabulations of nuclear spin
properties (see table 6.1 and appendix B). This property can be positive (as
for 27 Al) or negative (as for 17 O), and its value depends on the nature of
the charge distribution within the nucleus. The quadrupole moment couples
with the electric field gradient (EFG), eq , at the nucleus, which originates
from the surrounding electrons. The EFG is a tensor property and therefore
so is quadrupolar coupling. While this effect is electric in nature, it does
affect the spin energy and so influences NMR spectra. Now q is a traceless
tensor,6 so the isotropic average, missing image file is zero (like that of D
). Thus, resonance frequencies for solutions are basically unaffected by the
quadrupolar interaction. However, it still contributes significantly to
relaxation and therefore to line broadening—so much so that for nuclei
with large values of Q , it may be very difficult to observe the signal for the
solution state.
For solids, the quadrupolar effect can be very important. Indeed, it often
dominates the appearance of spectra, since the quadrupole coupling
constant may be hundreds of megahertz. On the other hand, when the
nucleus in question is at a site of cubic symmetry, the EFG is in principle
zero. An example is the EFG at 33 S in a sulfate ion; however, it will only
be negligibly small if the ion retains its tetrahedral symmetry in the solid
state, whereas frequently the local environment lowers the symmetry.
Only two parameters are in principle needed to characterize the magnitude

of q in any given situation, namely its anisotropy and asymmetry (see
section 2.2). Since q is traceless, the former is simply given by the largest
component in magnitude. This may be either positive or negative, the
principal axes being denoted by the order |q ZZ | ≥ |q YY | ≥ |q XX | .7
However, it is more normal to use, instead of the EFG anisotropy itself, a
factor proportional to it, known as the nuclear quadrupole coupling
constant ,8 χ . This parameter and the quadrupolar asymmetry, η , are
defined as:
where χ is expressed in frequency units but η is dimensionless (0 ≤ η ≤ 1) .
If the quadrupolar energy is much smaller than the Zeeman energy, it can
be treated as a first-order perturbation (see chapter 6 for details). In such a
case, the energy, missing image file for nuclear spin I is given by:
where m I is the spin-component quantum number and the parameters θ

and ø are the polar angles of B 0 in the principal axis system of the tensor.
The full quadrupolar energy term must be summed over all the types of
quadrupolar nuclei present. However, unlike the situation for dipolar
interactions (which occur for spins pairwise), the quadrupolar effect
involves single spins acting independently, which simplifies theoretical
treatments.
If the asymmetry is zero, the first-order quadrupolar contribution to the

resonance frequency for a transition from m I to m I − 1 is:
Since there are 2I + 1 energy levels (figure 2.7 shows the case for a spin-
missing image file nucleus), there will be 2I nondegenerate lines in the
spectrum for each value of θ other than 54° 44′.
Most quadrupolar nuclides have spin quantum numbers that are odd
multiples of missing image file (i.e., missing image file
missing image file , missing image file …). In such cases, the central
transition missing image file is unique because the first-order quadrupolar
effects on the two energy levels are equal. This is true even including cases
of nonzero asymmetry, since the quadrupolar energy depends on
missing image file , not m I (see equation 2.27). Therefore, the
quadrupolar contribution to the transition frequency is zero and is thus
independent of the orientation. The other NMR transitions (satellite
transitions ) are, however, strongly dependent on θ , so the intensity for a
polycrystalline sample will be spread over a wide frequency range. The
central transition resonance will thus stand out like a sore thumb. Figure 2.8
shows a powder pattern for an I = missing image file hypothetical case.
The spacing of the prominent horns (indicated by the arrow in figure 2.8 )
depends on the spin quantum number I ; in the I = missing image file
case, it is missing image file . The satellite transitions form subspectra as
indicated. Figure 2.9 provides an experimental example for an I =
missing image file system.
missing image file
Figure 2.7. Energy levels for a quadrupolar nucleus. Left: zero quadrupole
coupling constant; right: schematic introduction of a small quadrupole
coupling constant (first-order effect).
missing image file
Figure 2.8. Simulated quadrupolar powder pattern for an I =

missing image file case. The three types of transition (subspectra) are
shown separately.
If χ is large, the satellite transitions may not be readily visible, in which

case only 40% of the intensity (for I = missing image file ) will be
detected (see inset 6.1). On the other hand, when χ is small, only a single
line (though usually rather broad) may be seen, comprising all transitions.
This is often the case for 133 Cs, which has a small value of Q (− 0.343
fm2 ) and is usually ionic (q ZZ small). Contrast this with the cases of 23
Na (Q = 10.4 fm2 ) and 35 Cl (Q = − 8.165 fm2 ).
However, when χ is large, as is often the case, the simple theory given
above is not adequate and second-order effects become prominent, as will
be described in chapter 6 . The case of nuclides with integral spin is also
different since there is no central transition (see inset 2.3). The only
nuclides in this category usually considered to be important are 2 H, 6 Li,
10 B, and 14 N. Of these, the first two do not generally give broad spectra,
since they have small quadrupole moments, and 11 B is usually preferred to
10 B, so only 14 N causes special difficulties.
missing image file
Figure 2.9. Sodium-23 spectrum of a static microcrystalline sample of

sodium nitrate, which has a relatively small quadrupole coupling constant,
χ , of ~ 335 kHz and zero asymmetry. The upper spectrum is a vertical
expansion to show the turning points more clearly.
Inset 2.3. Spin-1 nuclei

Equation 2.27 shows that, for a spin-1 nucleus in an axially symmetric site,
the energy levels m I = +1 and m I = −1 are affected in the opposite
direction to the m = 0 level by the quadrupolar interaction. Therefore, one
of the two possible Δm I = ±1 transitions increases in frequency, while the
other one decreases. There can be no “central transition.” For a powdered
sample, the transitions form subspectra that are influenced by (3cos 2 θ − 1)
in opposite ways so as to give a total bandshape, which is identical to that
of the homonuclear Pake Doublet, but with a scale governed by χ /2 instead
of D , that is with the separation of the “horns” of 3χ /4. Figure 2.10 shows
schematically a typical deuterium spectrum for a zero-asymmetry case with
a quadrupole coupling constant of 170 kHz.
missing image file
Figure 2.10.
Quadrupolar data are useful in a number of ways, including acting as

resonance assignment aids and as enhancements for understanding
electronic structure. Accounting for quadrupolar effects is necessary to
obtain chemical shifts of the relevant nuclei, with implications for structure
determination, as is described in chapter 6 .
2.7 MAGIC-ANGLE SPINNING
All the internal NMR interactions have a common mathematical form. In

particular, there is a universal dependence on orientation of the
molecular/crystallographic frame in the applied magnetic field, which, at
least to first order as given in the above sections, is of the form:
The average value of this function for isotropic motion is zero. Thus
molecular motion in solutions means that the anisotropies in σ and J do not
affect the resonance frequencies. Since the isotropic values of D and q are
zero, these interactions also have no significant influence on solution-state
resonance frequencies.
For solids, the anisotropies cannot be ignored, as mentioned above, and

they give rise to substantial line broadening. However, it was realized in the
1950s that physical rotation of macroscopic samples about an axis inclined
at an angle β to the applied magnetic field would average equation 2.29 to
some extent and therefore diminish the broadening effects of anisotropic
interactions on solid-state spectra. In fact, it can be readily shown, by
simple geometric arguments,9 or in terms of quantum mechanics (see
section 4.6.1), that:
where Θ is the angle between a relevant molecular-level direction (e.g., an

internuclear distance in the case of a dipolar interaction) and the axis of
rotation (see figure 2.11 ).
missing image file
Figure 2.11. This shows the angles θ , β , and Θ , relevant to magic-angle

spinning (see the text). The heavy black arrow represents a vector such as
an interatomic distance, while the dashed arrow is the axis of rotation of the
sample.
For a polycrystalline sample, the angle Θ can take all possible angles, just
as θ can. However, the angle β is fixed by the experimentalist. Therefore,
the factor missing image file acts to scale the powder pattern produced by
the range of values of θ . If β = 0 ° (rotation about the z axis), there is no
scaling. At the other extreme, β = 90 ° , the spread of the powder pattern is
reduced by a factor of 2, and if β = 54 ° 44′, the effects of anisotropies
vanish since the scaling factor is zero missing image file The width of the
signal is reduced as if by magic, so the term “ magic-angle spinning ”
(MAS) was devised to describe the phenomenon. The MAS technique is
now ubiquitous in solid-state NMR.
The right-hand side of figure 2.12 shows an example for 13 C. The spin
rate should be significantly greater than the static linewidth in order to
provide a fully narrowed single signal. Normally, spin rates in the region 3
− 30 kHz are used, though faster rates are now feasible. At the lower end,
these suffice to average 13 C shielding anisotropies, but higher speeds are
necessary to remove the effects of H,H dipolar interactions, and it is
impossible to completely average most quadrupolar effects.
At this point, it is important to appreciate that interactions can be classified

as either homogeneous or inhomogeneous and that these have distinctly
different responses to MAS.10 In the latter case (shielding anisotropy, for
example), a single crystal would give sharp lines and the broadening found
for polycrystalline samples occurs because of the variety of orientations
that are present. Therefore, MAS rates only need to exceed the linewidths
which would be found for a single crystallite, so averaging by sample
rotation is easy. MAS at rates substantially less than the static bandwidth
for a polycrystalline sample still gives full line sharpening. However, in
such a case, spinning sidebands are seen—resonances spaced at intervals of
the spin rate from the centerband at the isotropic resonance frequency (see
the right-hand side of figure 2.12 ). Spinning sideband manifolds can be
both a bane (complicating the appearance of spectra) and a blessing
(allowing determination of shielding tensor parameters, for instance). The
distribution of intensities for a shielding case depends on the relevant
anisotropy and asymmetry. Computer programs exist to extract these
parameters from the spinning sideband intensities (see figure 2.13 and
section 8.2).
missing image file
Figure 2.12. Spectra of solid dicyclopentadienyl zirconium dichloride,

(cp)2 ZrCl2 , at various magic-angle spinning rates. Left: proton spectra;
right: carbon-13 spectra (with high-power proton decoupling).
This procedure is often an improvement on using static spectra because the
MAS spectra allow resolution of sideband manifolds of chemically
different nuclei whereas the broad resonances for static samples would
contain severe overlapping. Spin rates in excess of the static bandwidth of
the spectrum are needed to eliminate spinning sidebands for spectra subject
to inhomogeneous broadening. This is achievable when shielding
anisotropies are low, as is generally the case for 13 C spectra but not
usually the situation for heavy-metal spin- missing image file nuclei such
as 119 Sn.
missing image file
Figure 2.13. Fitting the 31 P spinning sideband manifold for compound I .

The simulated spectrum is shown in black slightly shifted to the left from
the experimental spectrum (in grey) for clarity of comparison. The
computation shows that the shielding anisotropy, ζ , is 77.5 ppm and the
asymmetry is 0.77. A 65 Hz line broadening has been applied to the
simulated spectrum. The isotropic chemical shift is 42.0 ppm.
On the other hand, for homogeneous interactions (typically homonuclear

H,H dipolar coupling), single crystals would still give broad lines because
of the great variety in the magnitude and orientations of internuclear
distances. Slow spinning would give no sharpening at all and high
resolution can be obtained only at spin rates greatly exceeding the static
linewidth (see the 1 H spectra in figure 2.12 ).
There is a close relation between the case of homogeneous line broadening

and the phenomenon of spin diffusion (see inset 7.3). This involves spatial
dispersion of spin orientation (without any physical diffusion of
molecules). It is caused by the flip-flop of the orientations of neighboring
spins under the influence of homonuclear dipole–dipole interactions.
Successive flip-flops can diffuse a spin orientation in a random-walk
fashion. Such spatial dispersion takes time, so the phenomenon can be used
to assess the size of domains in heterogeneous materials such as
semicrystalline polymers (see inset 7.3).
For most spectra of quadrupolar nuclei, the total resonance bandwidth is so

large that no feasible MAS rate will be sufficient to give a single peak;
therefore, numerous spinning sidebands will usually be seen. As is the case
for sidebands arising from shielding anisotropy, it is feasible, by the use of
suitable computer programs, to extract (quadrupolar) anisotropies and
asymmetries from the intensity distributions in spinning sideband
manifolds.
2.8 RELAXATION
NMR spectra are obtained from samples in strong magnetic fields by

recording the response of bulk spin magnetization to radiofrequency (RF)
radiation. Since the 1970s, pulses of radiation have been used, which excite
a range of frequencies (to cover the relevant spectral range). The resulting
signal is a time response, referred to as a free-induction decay (FID).
By convention, the strong applied magnetic field is said to be applied in the

z direction. The RF tilts the magnetization towards the xy plane, with
detection also occurring in the xy plane. Generally, signals are summed
over the FIDs following many pulses in order to increase the signal-to-
noise ratio (S/N) to an acceptable level (see chapter 3 ). During the whole
NMR experiment (or whenever the system is perturbed), the net
magnetization of the ensemble of spins, M , will experience a tendency to
revert to its equilibrium situation. The process of regaining equilibrium is
known as relaxation . However, there are a number of different relaxation
processes, as discussed in chapter 7 , depending on the experimental
situation. Solid-state NMR is particularly concerned with three cases
briefly described as follows:
(a) Spin–lattice relaxation (also sometimes known as longitudinal

relaxation). This describes the process of regaining equilibrium of the z
component of M (i.e., along B 0 ) following a perturbation (or immediately
after the sample is placed in B 0 ). For homogeneous solid samples, this
process is generally exponential, that is, describable as single exponential,
with a relaxation time constant, T 1 , ranging from seconds to kiloseconds,
as given in equation 2.31:

where M z (0) is the magnetization in the z direction after the perturbation.
The time T 1 is the inverse of a first-order rate constant for the relaxation.
In multipulse experiments, T 1 is of particular importance because ample
allowance for spin–lattice relaxation is normally vital between pulses (see
section 3.3.1).
(b) Spin–spin relaxation (better described as transverse relaxation). This

relates to the xy component of M (i.e., perpendicular to B 0 ), which is zero
at equilibrium, but which can become nonzero as a result of applying RF
pulses (see chapter 3 ). Since M z (parallel to B 0 ) and M xy
(perpendicular to B 0 ) are orthogonal, transverse relaxation is entirely
distinct from spin–lattice relaxation, though the mechanisms may be the
same. In general, for solids, transverse relaxation is unlikely to be single
exponential. The typical relaxation time constant is given the symbol T 2 ,
which, for protons in a rigid solid, will be very short (tens of microseconds)
—a markedly different situation from mobile solutions, for which T 1 = T 2
. The value of T 2 is directly related to the linewidth or lineshape of
resonances. Further information on lineshapes and on spin–spin relaxation
may be found in sections 5.2 and 7.2.3.
(c) Spin–lattice relaxation in the rotating frame. This describes the return to
equilibrium of transverse magnetization in the presence of an RF magnetic
field, B 1 , in the same direction. In this situation, the magnetization is said
to be spin-locked , because the relaxation time constant, T 1 ρ , is greatly
extended beyond transverse relaxation. It may be likened to T 1 but as
appropriate for a low magnetic field (B 1 ) rather than a high field (B 0 ).
Typically, values of T 1 ρ for 1 H in solids are 1− 20 ms (whereas for
mobile solutions they are equal to T 1 ). Values of T 1 ρ are of particular
importance in cross-polarization experiments (see section 3.4).
Several mechanisms contribute to relaxation. They all require a time-

dependent interaction involving the nuclear spins, so the interactions
(shielding, dipolar, and quadrupolar) discussed above can be involved.
Naturally, the larger the energy of interaction, the more efficient is the
relaxation mechanism. Thus, when present, the quadrupolar interaction will
often dominate. However, both dipolar and shielding anisotropy
interactions can be very important. The requisite time dependence arises
from motion at the molecular level. Thus, measurement of relaxation times
gives information about dynamics at the molecular level, especially if
variable-temperature (or variable-field) experiments are carried out.
For solids, motions can be very complex and are often poorly understood.
In simple cases, for example molecular motion which is either isotropic or
involves rotation about a unique axis, T 1 and T 1 ρ pass through a well-
defined minimum as the motional rate increases, with a tendency to
increase toward infinity for very fast motions (as in mobile liquids) or very
slow motions (as for rigid solids). For most solids, mobility at the
molecular level is complex and the relaxation behavior as a function of
temperature is poorly understood. However, internal rotation about C–CH3
bonds tends to be relatively facile and so its effects can be well separated
from those of other motions. In such situations, relaxation times may show
several minima as a function of temperature. Figure 2.14 shows the case for
trimethylphosphine sulfide, which has two T 1 minima corresponding to
distinct motional processes, namely rotation of the methyl groups and
rotation of the whole molecule about the symmetry axis.
missing image file
Figure 2.14. Proton spin–lattice relaxation time of solid trimethylphosphine

sulfide as a function of temperature, showing minima for two types of
molecular motion. Minimum A occurs because of overall rotation about the
molecular axis, while minimum B corresponds to internal rotation of the
methyl groups about the C- P bonds.
Minima in T 1 versus T plots relate to motions at the Larmor frequency,

that is, tens or hundreds of megahertz, whereas T 1 ρ responds to motions
at frequencies related to the RF power, that is, missing image file which
are generally tens of kilohertz. Transverse relaxation, on the other hand, is
caused by very low-frequency motions. Thus the three relaxation types give
complementary information. However, even together, they cover only three
frequency regions, though of course B 1 can be varied to some extent and
different spectrometers may provide several Larmor frequencies. A better
strategy is to vary B 0 over a large range. This requires special equipment,
now commercially available (but not common), and constitutes the subject
of relaxometry .
Any molecular-level mobility at rates comparable to the inverse linewidth
of a spectrum will also lead to partial averaging of anisotropic interactions
and hence to sharpening of spectra, as already illustrated in figure 2.6 .
FURTHER READING
GENERAL REVIEWS OF THE PRINCIPLES OF SOLID-STATE NMR
“Transient techniques in NMR of solids” , B.C. Gerstein & C.R. Dybowski,

Academic Press Inc. (1985), ISBN 0 12 281180 1.
“Nuclear magnetic resonance spectroscopy: A physicochemical view” ,

R.K. Harris, Pearson Education Ltd. (1987), ISBN 0 582 44653 8.
“High resolution NMR in the solid state” , E.O. Stejskal & J.D. Memory,
Oxford University Press (1994), ISBN 0 19 507380 0.
“Introduction to solid-state NMR spectroscopy” , M.J. Duer, Blackwell

Publishing Ltd. (2004), ISBN 1 4051 0914 9.
NOTES
1 Note the use of lower-case zz to indicate that σ zz in this equation does

not refer to the ZZ principal component.
2 Metal samples have so-called Knight shifts (which can be very large) in
their resonances. These will not be dealt with in this book. Paramagnetic
samples are also subject to special shift effects (see section 7.5).
3 This coupling phenomenon is frequently referred to by the clumsy

designation “J coupling.” This is so common in the literature that the
terminology is sometimes used in this book, though on other occasions the
more meaningful expression “indirect coupling” is employed.
4 Some texts define D jk include a negative sign in its definition. The key
point is that the energy expressed in equation 2.23 is lower when the
nuclear magnetic dipoles are parallel (if γ values for both nuclei are
positive).
5 The term “Pake Doublet” originally referred to the homonuclear case.
6 The symbol V is often used in the literature instead of eq for the EFG
tensor.
7 This (traditional) definition puts the Y component between the X and Z

components, contrary to the usage for the other NMR tensors.
8 The symbol C Q is often used for this quantity, though χ is the IUPAC-
approved symbol.
9 See appendix 5 of the book by Harris given in Further reading.
10 The quantum mechanical origin of this distinction is discussed in inset

4.5.
CHAPTER 3
SPIN - NUCLEI : A PRACTICAL

GUIDE
3.1 INTRODUCTION
This chapter gives a beginner’s guide to the experimental solid-state NMR

of spin - nuclei. It is a break from consideration of the underlying
principles of solid-state NMR, concentrating instead on what it is necessary
to know in order to obtain one-dimensional spectra, such as those shown in
figure 3.1 . These spectra are often valuable on their own and they are also
the starting point for any of the more complex experiments described later
on. Quadrupolar nuclei are treated separately in chapter 6 .
Figure 3.1. Examples of high-resolution solid-state NMR spectra: (a) a carbon-13 spectrum from a
steroid; (b) a silicon-29 spectrum from a zeolite.
There are 24 elements (31 isotopes) with non-radioactive spin-

nuclides (and some of these elements have quadrupolar nuclei as well); see
inset 3.1. Carbon, nitrogen-15, silicon and phosphorus are the “bread and
butter” nuclei for solid-state NMR and, under the right experimental
conditions, their spectra are relatively easy to obtain. 77 Se, 89 Y, 111/113 Cd,
117/119 Sn, 125 Te, 195 Pt, 199 Hg and 207 Pb present few technical challenges,
other than those associated with extensive spinning sideband manifolds

arising from large shielding anisotropies. 57 Fe, 103 Rh, 183 W, 187 Os and
107/109 Ag have low re sonance frequencies and slow relaxation also can be a
great handicap. This situation poses severe instrumental difficulties at low

magnetic fields (≤ 11.7 T), and their nuclei are really accessible only to high-
field instruments. In the solid state, 1 H and, to some extent, 19 F are special
cases and they are dealt with separately in section 3.5. Tritium has special
problems because of its radioactivity and helium-3 is not important
chemically. This leaves 129 Xe, which has some rather specialist
applications, and 203/205 Tl, which is an awkward element as its resonance
frequencies fall between 31 P and the high-band nuclei (19 F, 1 H, and 3 H) and
are out of the range of most solid-state probes.
Inset 3.1. Elements with spin- isotop es
H He C
N F Si
P Fe Se
Y Rh Ag
Cd Sn Te
Xe Tm Yb
W Os Pt
Hg Tl Pb
3.2 THE VECTOR MODEL AND THE ROTATING FRAME OF

REFERENCE
A vector model of the magnetization is useful for the description of the

NMR experiments introduced in this and subsequent chapters. One aspect of
this, the bulk magnetization vector, was introduced in chapter 2 and a
second, the rotating frame of reference, is described briefly here. The text by
Keeler (see Further reading) goes into more detail on this and the material
introduced in section 3.2.1.
Precession is the movement of the axis of a spinning body about a second
axis (see figure 3.2 ). If the bulk magnetization is displaced from the
laboratory z axis by a resonant radiofrequency pulse, it will precess about B 0
with a frequency equal to ν NMR given by equation 2.4. The effect of the
pulse is easiest to describe in a coordinate frame rotating at the RF
transmitter frequency ( ν RF )—the rotating frame of reference (figure 3.3 ). In
such a frame, the oscillating magnetic field associated with the pulse will
appear to be static (in effect ν RF is subtracted out of the response to the
pulse) and the precession of the bulk magnetization about B 0 is
“eliminated”. This is the way the NMR experiment is treated in chapter 4
and more details about the rotating frame can be found there. The
elimination of the precession is equivalent to removing the magnetic field B
0 and that makes it easier to describe the interaction of the bulk
magnetization with the much weaker radiofrequency field B 1 .
Figure 3.2. The axis of a spinning body precesses around another axis (the motion of a gyroscope).
In a coordinate frame rotating at the same rate as a

rotating point, the position of the point is constant.
Figure 3.3. The rotating frame of reference.
Inset 3.2. A note on Fourier transform NMR spectroscopy
Precessing magnetization (see main text), generated in response to a pulse of

electro-mag netic radiation, induces a voltage in a surrounding coil. It is this
voltage that is the NMR signal, interferogram , or free induction decay (FID), recorded
over the acquisition time .
This time domain signal is generally compli-cated because it contains many
different precession frequencies and is rarely analyzed directly. The NMR signal is
more usually presented as a frequency spectrum—the frequency domain . The usual
mathematical method for interconverting the time and frequency domains is Fourier
trans-formation (FT).
3.2.1 PULSE ANGLE

Pulses are usually described as being applied relative to the axes of the
rotating frame.1 In this frame, a pulse applied with phase x rotates the
magnetization vector, at a rate γ B 1 (the nutation rate; see inset 3.3), from z ,
about x , toward the y axis as shown in figure 3.4 .2 The angle through which
it is rotated is called the pulse (or tip) angle (θ ) and is proportional to the
duration of the pulse, τ p :
3.1
Figure 3.4. A pulse applied along x rotates the bulk magnetization vector about x .
where θ is in radians. Only the component of magnetization that is

perpendicular to B 0 is detectable, so the greater the magnetization in the xy
plane, the more intense will be the NMR signal. The maximum signal
intensity will thus be obtained for a pulse angle of 90° (a 90° pulse ). Figure
3.5 shows a typical nutation curve (the signal as a function of pulse duration).
Figure 3.5. Experimental 31 P nutation curve. The first spectrum was recorded with a pulse of 1 μ s
duration, the pulse duration then increases successively in steps of 0.4 μ s to the right. The overall loss
of signal with increasing pulse duration arises from RF inhomogeneity.
A nutation curve is routinely recorded to determine the optimum 90°

pulse duration and to calibrate the RF field strength. Because of
experimental features such as a variation in RF field strength across the
sample (RF inhomogeneity ), it is not unusual to find that the observed 90° pulse
is slightly different from (often less than) half of the 180° pulse duration and
it may be difficult to pinpoint exactly the times for which the signal is zero.
Nevertheless, it is easier to visually determine or estimate the zero-signal
points rather than the signal maxima, and half the difference between the
180° and 360° pulse durations often gives a better indication of the true
magnitude of the RF field strength (see inset 3.3).
Inset 3.3. Pulse angle and field strength
The RF field strength associated with a pulse is a magnetic flux density and has
the unit tesla (T). The field associated with a pulse, B 1 , typically has a magnitude
of several milliteslas. If, say, this is 10 mT , a pulse duration, τ p , of 3 μ s would
produce a pulse angle of 90° for silicon-29 (from equation 3.1).
The strength of the resonant RF field is usually discussed in terms of an
equivalent rotation rate (the nutation rate of the affected spins):
3.2
where τ 90 is the 90° pulse duration. So a 3 m s 90° pulse corresponds to an 83.3

kHz RF field. Note the distinction between ν 1 (the nutation rate produced by the
on-resonance pulse) and ν RF (the frequency at which it is applied).
In FT-NMR, the pulse of RF irradiation excites a range of frequencies as

illustrated in figure 3.6 (this is the whole point of using a pulse). Exciting a
large spectral width is often important in solid-state NMR, so short (often
termed hard ) pulses are generally required as these give a wide frequency
coverage; it is commonplace to work with 90° pulses with a duration of 4 μ s
or less. A 90° pulse can be expected to excite a frequency range roughly
equivalent to 1/(4τ 90 ) with a reasonable degree of uniformity (so the 3 μ s
90° pulse described in inset 3.3 would excite a range of about 85 kH z
reasonably uniformly).
Figure 3.6. The y component of the magnetization resulting from an RF pulse along x as a function of
transmitter offset (expressed relative to the nutation frequency). A 90° pulse excites more efficiently
but does so over a small range: a range of approximately ± 0.5ν 1 (the nutation frequency) is within
10% of the maximum excitation. A shorter pulse, of the same amplitude, with a smaller tip angle,
generates less signal but excites over a wider range (±2v 1 is within 10% of the maximum excitation
for this pulse).
Off-resonance
So far it has been assumed that the pulse has been applied on-resonance and
ν RF = ν NMR . In the general case, and the frequency of the
signal of interest, ν , depends on interactions in addition to the Zeeman one,
for example, shielding (see equation 2.18), so When
(it has to be so for a spectrum containing more than one
resonance), instead of the magnetization being rotated about B 1 , it rotates
about an effective field along a tilted axis , B eff , which is the resultant from B
1 and a residual static field B r = 2 π Δ ν /γ (B 0 is no longer effectively zero
in the rotating frame):
3.3
This of f-resonance situation has two consequences. Firstly, relative to the

on-resonance case, it takes a longer pulse to rotate the magnetization to the xy
plane and, secondly, when it arrives in that plane, it will have a phase error
with respect to on-resonance magnetization. Thus, for a 13 C spectr um, an
RF field equivalent to 80 kHz giving a 90° pulse at 100 ppm on a 9.4 T
spectrometer (100 MHz for 13 C) would give an 83° pulse at 0 and 200 ppm.
This may have consequences for the more advanced experiments described
later, where precision in the pulse duration is important. The phase error in
the spectrum can be removed with a first-order (frequency-dependent)
correction.
3.3 THE COMPONENTS OF AN NMR EXPERIMENT
There are two main methods for exciting nuclei and generating spectra in
solid-state FT-NMR. The simplest experiment consists of a pulse followed
by the detection of an FID (an acquisition ). It is variously known as pulse-
acquire, single-pulse excitation (SPE), direct polarization, or direct
excitation. Direct excitation is the term used here. Alternatively,
magnetization can be transferred from an abundant spin such as 1 H to a
dilute one. This cross-polarization technique is described in section 3.4. The
previous section dealt with the excitation part of the direct-excitation
experiment. The other components and some of the parameters under the
control of the spectroscopist are introduced next.
3.3.1 RECYCLE DELAY

It was noted earlier that NMR is an insensitive technique. In experimental
terms this means that the amount of signal generated by each radiofrequency
pulse, or sequence of pulses, is usually small (of the order of microvolts)
relative to the level of the electronic noise inherent in the spectrometer. To
impr ove sensitivity it is usual to repeat an experiment many times, adding
each FID to the sum of the previous ones. The magnitude of the NMR signal
is proportional to the number of repetitions (n ). The noise, which is random,
increases only as so the higher the number of repetitions, the higher the
signal-to-noise ratio (see inset 3.4).
Inset 3.4. Signal-to-noise ratio
The signal-to-noise ratio (S/N) is a useful measure of the quality of a spectrum and
the performance of the spectrometer. The noise is usually characterized by its
standard deviation and that, relative to the height of the signal in question, gives
the signal-to-noise ratio. Invariably, the spectroscopist seeks the highest possible
signal-to-noise ratio.
Repeating an experiment raises the issue of how long to wait between

repetitions. The physical characteristic that relates to this is the spin–lattice
relaxation time, T 1 (which was introduced in section 2.8 and is discussed
further in chapter 7 ). In solution-state NMR, there is often no need to wait
between finishing one acquisition and starting the next because T 1 ≈ T 2
(and so is of the order of the acquisition time; see the next section): as soon
as the FID has decayed to zero, the pulse sequence can be repeated. In the
solid state, however, T 1 is usually several orders of magnitude longer than
the acquisition time, so even after the FID has decayed to zero, it is
necessary to wait until the magnetization has returned to equilibrium before
applying another pulse.
The time between successive repetitions of the NMR experiment (the
experiment is defined here as the excitation period plus the acquisition time)
is called the recycle delay or pulse delay . As already noted, the optimum recycle
delay depends on T 1 , so is sample-, nucleus- and environment-dependent
and it cannot be readily predicted (see chapter 7 ). Using an inappropriate
delay can lead to signals being missed or their intensities being wrongly
represented (see figure 3.7 ) or it can waste spectrometer time.
Figure 3.7. Two 31 P spectra acquired from the same sample but with different recycle delays. At the
shorter recycle, the broad, low-frequency shoulder on the narrow line is barely detectable and the
high-frequency signal is significantly underrepresented.
So what is the best recycle delay to choose? That depends on whether

the experiment needs to be quantitative or whether a high signal-to-noise
ratio is of most importance. It also depends on the type of experiment: as
will be discussed in section 3.4.3, cross-polarization experiments tend not to
be quantitative, so here the key issue is generating the highest signal-to-noise
ratio.
Suppose a direct-excitation experiment is being carried out and the
sample has been in the magnet long enough for equilibrium (a Boltzmann
distribution of populations) to have been reached. At the beginning of the
experiment (time t = 0 ), the magnetization along the z axis, Mz , is the
equilibrium magnetization, M 0 , and the magnetization in the xy plane, Mxy ,
is zero. After a 90 ° pulse, Mz = 0 and Mxy = M 0 . (It is the magnetization Mxy
that is detected in the FID.) After the pulse, spin–lattice relaxation begins to
return magnetization to the equilibrium state (i.e., to the z axis), as described
by equation 3.4 and shown graphically in figure 3.8 .
3.4
Figure 3.8. The amount of z magnetization at time t (relative to T 1 ) after a 90° pulse.
Now suppose the experiment is repeated and a second 90° pulse is

applied. If the recycle delay is short compared with T 1 so that the second
pulse is applied before M z recovers to the value M 0 , the pulse will produce
a reduced amount of signal. Conversely, if the recycle delay is very long, the
condition M z = M 0 might have been reached some time before the pulse is
applied and spectrometer time has been wasted. To observe the fully relaxed
signal after each pulse, the optimum recycle delay is 5 × T 1 (which recovers
99.3% of the signal).
Such a recycle delay is appropriate when quantitative intensity
information is required, so for any sample that gives multiple signals (with
differing T 1 values), setting the recycle delay to five times the longest T 1 will
ensure that the signal intensities will be quantitative. (This was the recycle
delay used for the spectrum shown in figure 3.1 (b).) However, if accurate
intensity information is not an issue, then a trade-off can be made; reducing
the recycle delay reduces the amount of signal obtained per repetition, but
allows more repetitions in a given time, potentially increasing the signal-to-
noise ratio. As figure 3.9 shows, the highest signal -to-noise ratio is obtained
when the recycle delay is 1. 26 × T 1 (~72% of the full signal is obtained on
each repetition ). It would appear, then, that the choice of recycle delay
requires prior knowledge of T 1 . However, if a signal can be observed within
a few minutes, a short sequence of experiments with differing recycle delays
is usually sufficient to determine an appropriate recycle delay (see figure
3.10 ).3
Figure 3.9. The signal-to-noise ratio, for a given total experiment time and 90° pulses, as a function of
the recycle delay (expressed as a multiple of T 1 ).
Figure 3.10. Spectra recorded with different recycle delays. The horizontal line is drawn at 72% of
full intensity. The maximum intensity is obtained with a 30 s recycle, although little signal is
sacrificed with a 10 s delay. In terms of the signal-to-noise ratio in any given total time, the optimum
recycle is around 3 s.
3.3.2 ACQUISITION TIME
The acquisition time is the time during which the signal (the FID) is
detected. Getting this right is important to the appearance, and information
content, of the spectrum. If the acquisition time is too long, the unnecessarily
acquired noise in the time domain results in additional noise in the spectrum.
If it is too short the FID is truncated . Fourier transformation of a truncated
signal results in “wiggles” at the base of the lines in the spectrum (figure
3.11 ).4 Although it is possible to compensate for a truncated FID with
additional mathematical processing (as is commonly done for two-
dimensional data sets) and an overlong acquisition can be multiplied by an
apodization function (see Further reading), it is far better to optimize the
acquisition time to start with, if the signal is sufficiently visible.
missing image file

Figure 3.11. (a) A free induction decay (FID) acquired for 40 ms so that the signal decays to the level
of the noise. (b) Part of the spectrum from the Fourier transformation of the full FID shown in (a). (c)
The same part of the spectrum from the Fourier transformation of an FID obtained with a 10 ms
acquisition time (but zero-filled to the same total number of data points as (b)). This truncation leads
to characteristic baseline wiggles on either side of the affected peaks and a general loss of resolution .
3.3.3 RECEIVER GAIN
The receiver gain is the amount of amplification applied to the raw signal. The
receiver (the part of the spectrometer that digitizes the signal detected in the
NMR coil) has a finite range. Very-high-amplitude signals will overflow the
receiver and so will not be properly digitized. Significant clipping of the signal
leads to baseline artifacts in the spectrum (figure 3.12 ). Usually, the receiver
gain can be adjusted to avoid such problems, but in extreme cases (e.g., 1 H
observation) it may be necessary to add some attenuation to the signal
pathway. Too low a gain can result in artifacts when digitizing very weak
signals (digitization noise ).
missing image file
Figure 3.12. The “clipped” top and bottom of this free induction decay (FID) are characteristic of too
much signal prior to digitization. The subsequent Fourier transform results in the dips at the foot of the
line and the artifacts in the baseline.
3.3.4 SPECTRAL WIDTH
This is the frequency range under observation. The spectral width is

determined by the rate at which the data points of the FID are sampled and the
time between sampling points is called the dwell time . The maximum spectral
width is determined by the fastest available sampling rate. Too high a
spectral width is not usually a problem given that the pulse has a sufficient
excitation range (see section 3.2.1) and that a large number of data points
can be handled.5 Too small a spectral width, however, can result in signals
that fall just outside the spectral width being folded back into the spectrum
(which is a potential trap for the unwary since the apparent chemical shift
values for folded peaks will be incorrect). This situation is most likely to be
encountered, for spin- nuclei, with the heavy metals (cadmium, tin, lead,
mercury, and platinum) when large spinning sideband manifolds are
recorded. An example is shown in figure 3.13 . The only remedy is to repeat
the experiment with a more appropriate spectral width.
missing image file

Figure 3.13. Too small a spectral width can result in signals being folded back into the spectrum. The
extent of this 119 Sn spinning sideband manifold has caught out the spectroscopist. The way folded-
back signals appear in the spectrum tends to be instrument specific.
3.3.5 DEAD-TIME
Signals extending over large spectral widths can be a problem in solid-state
NMR—that is, if the spectrum is to have a flat baseline.6 In any NMR
experiment, excitation by relatively high- voltage RF pulses (typically
hundreds of volts in solid-state NMR) is followed by the detection of a low-
voltage (typically microvolt) response. Ideally, acquisition needs to start
immediately after the pulse. However, if it is started too soon, the beginning
of the FID can be distorted by pulse breakthrough (real pulses are not perfectly
rectangular and the decaying “tail” of the pulse can be detected by the
receiver). The same spectrometer circuitry also has to cope with both high-
and low-voltage conditions and it tends to ring from the shock of the high-
voltage pulses. The time that it takes for the pulse to decay and the circuitry
to settle down is called the dead-time (see figure 3.14 ). No meaningful data
points can be obtained during the dead-time so it is not possible to follow the
evolution of the signal from its start. In practical terms, this results in a
distortion to the phasing of the spectrum (spectral phasing is determined by
the way the real and imaginary parts of a complex Fourier transformation are
combined, but that is something that is covered in other texts; see Further
reading). This is particularly noticeable when the spectrum extends over a
large frequency range as it does in the case shown in figure 3.15 . A first-
order phase correction can be applied to the spectrum to produce signals that
are in-phase, but that introduces a roll to the baseline. Matching the dead-time
to an integer number of dwell periods, together with numerical prediction of
the early part of the signal decay, can minimize the baseline roll.
Alternatively, the baseline correction routines in the spectrometer software
can be used to produce a flat baseline. Another technique for cases involving
large spinning sideband manifolds is to carry out the Fourier transform from
the top of the first rotary echo . Under magic-angle spinning (MAS),
orientation-dependent interactions give rise to rotary echoes. Echoes form
because the resonance frequency from a given crystallite changes as the
rotor position changes during spinning and each time the rotor returns to its
original position the resonance frequency returns to its original value. Rotary
echoes are illustrated in figure 3.15 (d). An echo can also be used (in fact it
is usual to do so, see inset 3.5) to record the broad lines from non-spinning
samples (the refocused signal at the top of an echo is a good approximation
to observing the signal from zero time).
missing image file

Figure 3.14. After a pulse there is a dead-time, while the spectrometer circuitry settles down, when no
data points can be recorded. The missed points are represented here by the open circles. Typically, the
receiver is switched on a few microseconds before it is instructed to acquire data (so avoiding any
artifacts associated with the switch on).
missing image file missing image file
Figure 3.15. The 199 Hg magic-angle spinning spectrum from Hg(SCN)2 acquired with a 500 kHz
spectral width and at a spin rate of 9.78 kHz. There was a 15 m s dead-time delay between the end of
the pulse and the first data point. (a) With no first-order phase correction. Note the flat baseline but
extensive distortion to the phases of the spinning sidebands. The centerband is the in-phase signal at -
1300 ppm. (b) The same data set but with a large first-order phase correction (- 3529°). The sidebands
now all have the same phase but the baseline rolls. (c) The same data set but transformed from the top
of the first rotary echo shown in (d). Only a small (96°) first-order phase correction was necessary to
produce this spectrum.
Inset 3.5. The solid (or quadrupolar) echo
A significant amount of signal can be lost in the spectrometer dead-time for the
broad lines encountered with non-spinning samples. When shielding anisotropy is
negligible most of the missing signal can be recovered using a solid (or
quadrupolar) echo pulse sequence (see also section 4.4.2):
missing image file
1
The H spectra on the right were obtained (a) with and (b) without a solid echo.
The integrated intensity for the band in (b) is 23% of that shown in (a) from the
same sample. In (b) the dead-time between the pulse and the first point of the FID
was 10 μ s.
missing image file
The dead-time is usually a small number of microseconds—but it is

probe and frequency dependent. Acquiring data while the probe is still
ringing results in the corruption of the NMR signal and that leads to an
irregular distortion of the spectral baseline. Probe ringing (and hence the
dead-time) increases at low observation frequencies and is a particular
problem for low-gamma nuclei (loosely defined as anything with a
magnetogyric ratio less than that of 15 N), although it does depend on the
strength of the magnetic field and the study of low-gamma nuclei is one
reason for going to high field strengths. Some probes, for experiments where
pulses and data acquisition at the same frequency are interleaved, are
designed to give very short dead-times (at the expense of sensitivity).
3.3.6 SPINNING SIDEBAND SUPPRESSION
The spectrum shown in figure 3.15 (c) consists of a centerband at the

isotropic chemical shift and a manifold of spinning sidebands. But which line
is the centerband? As the position of the centerband is invariant to the spin
rate, the easiest way to find out is to change the spin rate and rerecord the
spectrum. Any line that changes position must be a sideband. The
centerband is at −1300 ppm relative to Hg(CH3 )2 in this spectrum. In this
case the sideband manifold does not obscure a second centerband (this was
evident from a spectrum obtained at a second spin rate) so, other than
effectively diluting the signal, it does not pose a problem. Indeed, computer
fitting of the sideband intensities gives information on the shielding tensor
and hence on the electronic environment of the nucleus (see section 8.2.3). It
would take a spin rate in excess of 100 kHz to produce a sideband-free
spectrum for this sample (in the 7 T magnet used for this measurement).
Sidebands are not always so extensive. They are usually barely
detectable in 29 Si spectra such as the one shown at the beginning of this
chapter (figure 3.1 ) and they account for the lowest intensity signals in the
carbon spectrum in that figure. In the latter case, their identity is easily
confirmed by a change in spin rate and they do not interfere significantly
with any of the centerbands in the spectrum. However, sometimes sidebands
can be a problem and it may not always be desirable or possible to spin the
sample fast enough to remove them from the spectrum. In such cases,
spinning sideband suppression techniques can be useful. Two common ones
are TOtal Sideband Suppression (TOSS), although the word total is
something of a misnomer, and Sideband ELimination by Temporary
Interruption of Chemical Shift (SELTICS). Both techniques suppress
sidebands rather than refocus them so their intensity is lost from the
spectrum (which might be important if intensity information is required).
Both techniques are reasonably good at removing first-order sidebands
(centerband ± 1 × ν r where ν r is the spin rate) but neither is suitable for
removing large manifolds of sidebands such as those shown in figure 3.15 .
3.3.7 DECOUPLING
As was mentioned in chapter 1 , a hydrogen-rich sample requires 1 H

decoupling if a high-resolution spectrum of any other nuclide is to be
obtained. Effective decoupling, at least at modest magnetic fields and sample
spin rates (≤ 10 T and 20 kHz, say), can be obtained by turning on a
continuous-wave (CW)7 decoupling pulse that lasts for the duration of the
acquisition.
Given that the acquisition time is often many milliseconds, a decoupling
pulse will be orders of magnitude longer than a typical excitation pulse.
Experimentally, then, some care is required when decoupling: the RF field
strength (decoupling power) and duration should not exceed the
specification for the probe and the total RF on-time also should not be too
high a proportion of the total experiment time. It is often useful to specify a
duty-cycle limit of approximately 10%.
It is sometimes necessary to use a very long acquisition time (> 100 ms)
to avoid truncating an FID. This situation may occur for samples with high
molecular mobility such as gels or rubbers. As these also tend to have short
relaxation times (allowing short recycle delays), it is easy to reach the duty-
cycle limit. However, such samples also tend not to require very high
decoupling power (the molecular motion partially averages the dipolar
interactions involving 1 H) so the duty-cycle limit can be exceeded providing
the decoupling power is reduced.
For organic materials, it is well worth using as much decoupling power
as is available, up to the limit the probe can withstand, as figure 3.16
illustrates. However, setting too high a decoupling power can result in a
spectacular electrical breakdown in the probe (known as arcing ). Minor
arcing can result in an increased and nonuniform noise pattern in the
spectrum. Serious arcing can obliterate the signal that is being detected (see
figure 3.17 ).
missing image file

Figure 3.16. The line-narrowing effect of increasing the continuous-wave (CW) decoupling power,
expressed as the 1 H nutation rate, is illustrated for the backbone carbon in crystalline polyethylene.
missing image file

Figure 3.17. Catastrophic arcing such as that shown here is the result of an electrical breakdown in the
probe, for example, between the turns of the coil, and destroys the spectrum. Apart from using too
high a decoupling power, arcing can occur when there is an electrical problem in a probe (such as a
badly soldered contact), when the air supply is dirty (or wet), or if there is sample leakage.
Continuous-wave decoupling has the advantage of simplicity, but more

elaborate decoupling schemes are becoming commonplace, particularly for
the carbon-13 NMR of organic materials. Two-pulse phase-modulated (TPPM)
decoupling is one such scheme that can be used to increase decoupling
efficiency, especially at high magnetic field, high sample spin rate, or when
decoupling power is limited. The subject of decoupling is discussed in more
detail in section 5.1.2. The factors governing linewidth are discussed in
section 5.2.
3.3.8 SPECTRAL REFERENCING
Most of the high-resolution experimental spectra in this book are presented

with a scale that has 0 ppm properly defined. That is, 0 ppm is the chemical
shift of a recognized standard compound. The primary reference compounds
are liquids or solutions, and while they can be put into a solid-state probe,
there are good reasons why this is not always desirable: (i) they cannot be
used to calibrate the class of experiments introduced in the next section
(which work only on a solid sample) and (ii) some of them are volatile,
poisonous, or reactive (and sealing them, intact, into a glass tube small
enough to fit into a solids probe is not easy).
In solid-state NMR, referencing is usually external— that is, a separate
reference spectrum is obtained from a standard sample. Solid-state spectra
are not affected by solvent susceptibility as solutions are, so external
referencing is reliable—although it should be carried out on a regular basis
to counter drift in the applied magnetic field. If a liquid sample is
impractical, a secondary solid standard is required. Ideally, this will give an
easily detectable, narrow line (so th at its position can be defined accurately)
and will have a known chemical shift with respect to the appropriate
recognized standard compound. However, because the solid-state spectrum
from a quadrupolar nucleus may be magnetic field dependent in a way a
solution is not (as explained in chapter 6 ), solutions are used routinely to
calibrate and reference experiments on such nuclei in solids. Table 3.1 lists
commonly used chemical shift reference materials for solid-state NMR.
Table 3.1. Common chemical shift reference compounds (including some for
quadrupolar nuclei) relative to recognized standard compounds (see the reference in
footnote f ).
Nuclide Reference for solid-state NMR (chemical shift / ppm)
1
H Adamantane (1.9) a
Glycine (8.5) b ,c
13
C Adamantane (38.5, CH 2 ) c ,d
Glycine (176.5, COO – ) c
15
N NH 4 15 NO 3 (–5.1, NO 3 ) c
Glycine (–346.8)
19
F C 6 F 6 ( –164.9) e
29 tetrakis(trimethylsilyl)silane (–9.8,–135.4)
Si
31 CaHPO 4 · 2H 2 O (1.0)
P
77 (NH 4 ) 2 SeO 4 (1040)
Se
119 Sn(C 6 H 12 ) (–97.4)
Sn
199 [Hg(dmso) 6 ][O 3 SCF 3 ] 2 (–2313)
Hg
7
Li LiCl (0) f
11 BF 3 /O(CH 2 CH 3 ) 2 (0)
B
17 H 2 O (0)
O
23
Na NaCl (0) f
27
Al Al(NO 3 ) 3 (0) f
45
Sc Sc(NO 3 ) 3 (0) g
51 b -NaVO 3 (–519) (solid)
ν
59
Co K 3 Co(CN) 6 (0) f
a
At high (>10 kHz) spin rates (when the centerband is well separated from the
spinning sidebands).
b
When used under high-resolution, multipulse conditions. Literature values vary
considerably so this value should be used with caution.
c
This is the highest frequency centerband in the spectrum.
d
Ξ = 25.145 970 and 25.145 743% for the CH2 and CH, respectively (see text).
e
There is some debate about this value. cfc l3 is no longer freely available so
resolving this issue is a problem.
f
In aqueous solution; details given in Pure Appl. Chem. 80 (2008) 59.
g
0.11 M in 0.05 M HCl.
Referencing methods that use documented absolute frequencies for
different nuclei, defined through the parameter Ξ , relative to the 1 H
resonance from tetramethylsilane in dilute solution in chloroform (Ξ = 100%
)8 can be used in the solid state, although this may require some ingenuity in
the absence of the deuterium lock used to do this in solution-state NMR.
(With the possible exception of biochemical solid-state applications, solid-
state experiment times are generally too short and lines are too broad for any
drift in magnetic field to be noticed, so a deuterium lock is not usually used
in solid-state measurements.)
3.3.9 TEMPERATURE CALIBRATION
Carrying out a solid-state NMR measurement at any temperature leads to the

question: What is the true sample temperature?
Due to the mechanics of MAS, it is not feasible to insert a temperature-
monitoring device directly into the sample. Instead, it is usual to have a
thermocouple in the heated/cooled gas stream close to the sample. But the
temperature of the nearby gas is not necessarily an accurate indicator of the
sample temperature because, even in the air bearing used for MAS, there is
frictional heating of the rotor (and so the sample). Some sort of temperature
calibration is therefore necessary if the true sample temperature is to be
known. Two of the various methods for doing this are mentioned here.
In solution-state NMR, neat methanol is a well-known “thermometer”
and the chemical shift difference in parts per million, Δ , between the CH3
and OH proton signals as a function of temperature is well documented:
For solid-state NMR it can be used either neat (if it can be contained) or on a
solid support. It is useful for calibrating temperature as a function of spin
rate (see figure 3.18 ) under “ambient” conditions and for giving an absolute
temperature. Because most probes can operate over a wide temperature
range, other temperature-sensitive substances have to be used for
temperatures at which methanol is not useful (below − 90 °C or above +60
°C). One such is solid lead nitrate (see figure 3.18 ). The chemical shift of
the 207 Pb signal is very sensitive to temperature (0.70 ppm / °C) and can be
used to determine the temperature difference from a known point (there is no
accepted absolute relationship between temperature and the chemical shift of
the single lead resonance), for temperatures up to about 250 °C (it
decomposes at 290 °C). The shape of the line (see figure 3.19 ) also gives an
indication of the temperature variation in the sample (this increases with
extremes of temperature and can be several degrees depending on the shape
and size of the sample space). The only major drawback with using lead
nitrate is its toxicity.
missing image file missing image file

Figure 3.18. (Left) Sample temperature as a function of spin rate for a 4 mm (rotor outside diameter)
magic-angle spinning (MAS) probe determined using neat methanol and with no temperature control.
At 15 kHz the sample temperature exceeds 60 °C. (Right) Temperature calibration graph for an MAS
probe at constant spin rate. Under the conditions used for this calibration, the true sample temperature
is about 8 °C above the temperature set on the spectrometer.
missing image file
Figure 3.19. 207 Pb spectrum from lead nitrate at nominal temperatures 25 and 150 °C. In the high-
temperature spectrum, the parts per million spread of the signal equates to a temperature range from
around 133 to 158 °C.
With all measurements at non-ambient temperature, it is important to

wait (usually at least 10 minutes ) after changing the temperature (or spin
rate) for a proper thermal equilibrium to be established. Two consecutive
measurements will determine whether the temperature is stable.
The effects of heating from the RF irradiation cannot always be
neglected. High duty cycles and long, high-power pulses can cause sample
heating. The extent of this tends to be more sample dependent than spinning-
induced heating and it is, therefore, more difficult to calibrate the effect on
the sample temperature. Samples with high dielectric constants (those that
are highly ionic or have a high water content) may need special care.
Measurements on delicate (e.g., biomolecular) systems should be carried out
at reduced temperatures to decrease the risk of sample degradation.
3.4 CROSS POLARIZATION
Cross polarization (CP) involves the transfer of magnetization (or

polarization) from the nuclei of one element to those of another. The most
commonly encountered cross polarization experiment involves transfer of
magnetization from abundant 1 H (or, occasionally, 19 F) spins to dilute X ones,
where X is any other spin- nucleus. It is worth noting, however, that cross
polarization is not limited to this combination.
The cross-polarization experiment is more complicated than that for
direct excitation (see section 3.4.1), so what are its advantages? There are
two main ones:
Because the magnetization originates from 1 H, the recycle delay is

limited by the recovery of the 1 H magnetization and not that of the X
spins. That is, the recycle delay depends on T 1 H and not T 1 X .
Usually T 1 H << T 1 X (see inset 3.6), so this means that the pulse
sequence can be repeated much more rapidly than in a direct-excitation
experiment, so significantly increasing the signal-to-noise ratio in the
spectrum.
In the limit of 100% magnetization transfer, there is a signal
enhancement by a factor equal to missing image file due to the
difference in equilibrium populations (see equation 2.6). For 13 C the
maximum enhancement is ~4 and for 15 N it is ~10, which represent
16- or 100-fold reductions, respectively, in experiment time. The
enhancement actually achieved depends on the efficiency of the cross-
polarization process which varies with the nature of the sample and the
experimental conditions.
The impact that these two factors might have on experiment time is
exemplified in inset 3.6 and illustrated in figure 3.20 .
Inset 3.6. Cross polarization vs. direct excitation

For 3-methoxybenzoic acid, the contrast between T 1 H and T 1 C is striking. Due
to rapid spin diffusion between the protons, there is a single value for T 1 H for the
molecule of 1.7 s. Although T 1 C for the methyl group is short (~1 s) because of
the mobility of that group, the average value for the other carbons is extremely
long (190 s). To record the same number of repetitions takes 112 times longer by
direct excitation than by cross polarization (given that an appropriate recycle delay
of 1.2 times T 1 H or T 1 C for cross-polarization and direct-excitation
experiments, respectively, is used). Suppose also that cross polarization gives the
maximum signal enhancement of a factor of 4 over that obtained by direct
excitation. It therefore takes 1792 (4 2 × 112) times as long for the direct-
excitation experiment to reach the same signal-to-noise ratio as the cross-
polarization one. Put another way, to match a cross-polarization experiment lasting
1 h, a direct-excitation one would take over 10 weeks! Cross-polarization and
direct-excitatio n spectra from a complex organic molecule are compared in figure
3.20 .
missing image file

Figure 3.20. Three carbon-13 magic-angle spinning spectra from the plant sterol stigmasterol. The top
spectrum was obtained using a cross-polarization experiment in 22 minutes (448 repetitions with 3 s
recycle delay). The middle spectrum was obtained using a direct-excitation experiment also in 22
minutes (448 repetitions with 3 s recycle delay). Note the lower signal-to-noise ratio and selectivity of
the experiment—only the signals from fast-relaxing carbons are obtained. The signals from the other
carbons in the molecule are observed only when the recycle delay is increased. The bottom spectrum
was obtained using a direct- excitation experiment with 120 s recycle delay (448 repetitions, total
experiment time nearly 15 h). The broad line near 110 ppm in the direct-excitation spectra arises from
the CF 2 carbon in the Teflon ® used as rotor caps. Because it contains no protons, t he Teflo n ®
signal is not detected in the cross-polarization experiment.
3.4.1 THE CROSS-POLARIZATION EXPERIMENT
The cross-polarization pulse sequence is shown in figure 3.21 . The initial 1

H 90° pulse directed along x , say, rotates the 1 H magnetization onto the y
axis. The second, 90°-phase-shifted 1 H pulse spin-lock s the magnetization
(see section 2.8) . During this spin-lock period, a pulse is applied
simultaneously at the X frequency. The time for which these two pulses are
applied is called the contact time and the X pulse will be referred to as the
contact pulse. Under the right conditions (see below), magnetization will
transfer from 1 H to X, and X magnetization will build up during the contact
time. The NMR signal of the X spins is then measured during the acquisition
time, which, as the sample contains 1 H, is carried out under 1 H high-power
decoupling. The contact time might last for up to tens of milliseconds (if the
probe can take it) for the polarization transfer to reach its peak. This is in
contrast to the pulses in a direct-excitation experiment which, typically, have
durations of a few microseconds.
missing image file

Figure 3.21. The cross-polarization pulse sequence.
The right conditions for polarization transfer occur when the 1 H and X
RF fields fulfill the Hartmann–Hahn match condition (the fields are then said to be
matched ):
missing image file or, from equation 3.2, missing image file
3.7
In other words, cross polarization occurs when the H and X nutation

frequencies are equal. It is also a prerequisite of cross polarization that there
is an interaction (generally dipolar coupling) between the two types of
nucleus.
The match condition is typically determined experimentally by
observing the X signal as a function of the amplitude (or power) of either the
1 H spin-lock pulse or the X contact pulse (while keeping the other RF field
constant). It is usually safer if the former is varied while the contact pulse is
kept constant as this is less likely to result in too much RF power being
applied to the probe, since missing image file at the matching
condition (inset 3.7 relates to the measurement of RF power). Figure 3.22
shows a typical 13 C match profile from hexamethylbenzene (HMB).9 At a
modest spin rate (3 kHz), a well-define d maximum is observed (this is
relatively uncommon behavior—the match profile is broader and flat-topped
for more typical materials).
Inset 3.7. Volts, watts and decibels
missing image file or missing image file , where P is the power in watts and V
pp is the peak-to-peak voltage (at an impedance of 50 W ).
A decibel value is the ratio of two powers: P(dB) = 10log10 (P1 /P2 ) When
quoted as dBm P 2 has a reference value of 1 mW.
So 1 mW ≡ 632 mV (0 dBm), 100 W ≡ 200 V (50 dBm), and 1 kW ≡ 632 V (60
dBm).
missing image file

Figure 3.22. The match profile from the methyl signal for hexamethylbenzene at a spin rate of 3 kHz.
The solid line is a visual guide only. The amplitude of the 1 H spin-lock pulse was varied while the 13
C contact pulse had a fixed amplitude.
For HMB the match behavior becomes more complicated at higher spin
rates (see section 5.1.3) and in other materials similar complexity can arise at
low spin rates when hydrogen is dilute in the sample or when there is a high
degree of molecular motion (as there is for adamantane, which is commonly
used in setting up experimental conditions for a spectrometer).
Any X pulse, including a contact pulse, will result in a signal being
generated by direct excitation as well as, and independent of, that originating
from cross polarization. This is most noticeable for nuclei with high natural
abundance such as 31 P or for isotopically labeled samples where the direct-
excitation signal has a high intensity. This unwanted signal is suppressed
using a phase cycle .
Phase Cycling
The phase cycle is a key part of most solid-state NMR experiments. As has
already been discussed, the pulses in a pulse sequence are labeled with a
phase. The receiver has a phase too (see inset 3. 8). Over a cycle of
successive repetitions of the pulse sequence, it is usually the case that these
phases change—the phase cycle. In the simplest of cases, the phase cycling
uses the quadrature phases x , y , – x , and – y (or 0, 90, 180, 270) to
compensate for any missetting of the phases (exactly how this works is not
imp ortant here). The receiver phase simply needs to follow the pulse. So the
pulse-acquire experiment might be phase cycled as:
Inset 3.8. Phase-sensitive (quadrature) detection
The “ raw ” NMR signal is split into two to give “ real ” and “ imaginary ”
components that have a phase difference of 90°. This is a necessary precursor to
the complex Fourier transform that converts the time-domain signal into the
frequency spectrum. The details of the way the signal is split are not important
here–and in any case tend to change with developments in the electronics within
the spectrometer. The important point is that a receiver nominally with phase x will
detect both the x and y components of the signal.
Table Missing
The phase cycle in a cross-polarization pulse sequence is designed to do

a further job—to select only signal that is generated by cross polarization
(i.e., to suppress the signal that is generated directly from the contact pulse
on the X-channel). A typical phase cycle might be:
Table Missing
Over the four repetitions of the pulse sequence in this phase cycle, the cross-
polarization signal is always detected as positive (relative to the receiver
phase) but the sum of the direct-excitation signal is zero over the phase
cycle. The alternating reversal of the phase of the 90° pulse is sometimes
referred to as spin-temperature inversion .
Designing complex phase cycles (like the one illustrated in inset 3.9) is
an advanced skill and takes some practice (and understand). An error in the
phase cycle is a common cause of unexpected behavior from a new pulse
sequence. See the texts in the Further reading for detailed information on
devising phase cycles.
3.4.2 CONTACT TIME
During the contact time, the X magnetization builds up toward a steady state
corresponding to an equilibrium between the 1 H and X magnetizations. At
the same time the spin-locked 1 H and X magnetizations decay through spin–
lattice relaxation in the rotating frame (section 2.8) at rates of
missing image file a nd missing image file respectively. These processes
are summarized in figure 3.23 . If missing image file then
missing image file
Inset 3.9. Phase cycles for coherence selection
The phase cycle is also a key part of the coherence selection in multiple quantum
experiments. Here the quadrature phases are not the only ones that can be used; 45,
90, 135 … are found in experiments where double quantum coherences are
selected and 30, 60, 90 … can be used for triple quantum coherence selection.
Figure 6.16(a) shows, a two-pulse, triple-quantum/single-quantum correlation
experiment that has the following 24-repetition phase cycle:
Table Missing
Table Missing
As this case illustrates, phase cycles can be many repetitions long. So that the
phase cycle can do the job it is designed for, it is always a good idea to acquire an
integer multiple of the number of repetitions in a phase cycle. That may not be as
trivial as it sounds for a complex two-dimensional experiment with a many-
repetition phase cycle—particularly if a long recycle delay is needed and the total
time required becomes prohibitive.
can be safely ignored and the amount of signal (S ) obtained with a contact
time t c is given by equation 3.8
wh ere S 0 is the maximum obtainable intensity in the absence of relaxation

and T XH is loosely defined as the time constant for cross polarization.
missing image file

Figure 3.23. The relationship of the parameters influencing a cross polarization experiment.
missing image file is usually much longer than the other time constants and is often ignored.
Strongly dipolar-coupled hydrogen nuclei in rigid solids tend to have a

common missing image file value because spin diffusion is fast on the
NMR timescale. However, T XH is determined by the local strength of the
dipolar coupling between X and H, and so is specific to the X environment.
For example, T SiH for (Si O)3 SiOH and (Si O)4 O environments in silicates
are very different as they have different relationships to the hydrogen.
Similarly, T CH for a CH2 carbon will be different from that for a quaternary
carbon (e.g., around 30 and 440 m s, respectively, in isoleucine). For a
complex sample, there is no guarantee that the signal intensities at a
particular contact time will be proportional to the number of nuclei giving
rise to that signal. This is a disadvantage of cross polarization, but is
analogous to the situation encountered in solution-state NMR where the
relative intensities of slowly relaxing quaternary carbons are often not
quantitative. In the solid state, quantitivity can be restored at the expense of
recording multiple spectra as a function of contact time and fitting the result
to equation 3.8 to obtain S 0 (see section 7.4). Experience suggests, however,
that equation 3.8 is often too simplistic (see figure 3.24 ).
missing image file

Figure 3.24. Contact time behavior for the carboxylic acid signal from L -isoleucine (left). The
observed signal (crosses) can be fitted (the solid line) using equation 3.8 with T XH = 0.44 ms and
missing image file S 0 (from the intercept on the signal axis indicated by the dashed line) is 160%
of the maximum observed intensity. The short-contact time behavior for the nitrogen in 15 N labeled
proline is shown on the right. Equation 3.8 is no longer useful for simulating the observed behavior,
although the transient oscillation can be used to measure the 15 N, 1 H dipolar coupling (see section
8.3.1) .
Occasionally, missing image file is long (tens of milliseconds)

and, providing the 1 H spin-lock is maintained to the end of the acquisition
time,10 a 1 H 90° flip-back pulse (of opposite phase to the initial one) can
return 1 H magnetization to the z axis and circumvent the need for a long
recycle delay, hence speeding up experiments. This is illustrated in figure
3.25 .
missing image file
Figure 3.25. (a) The flip-back modification to the cross-polarization pulse sequence and (b) 13 C
spectra from mannitol obtained with a 60 s recycle delay with (on the left) and without (right) flip-
back (all other acquisition conditions being equal). Without flip-back, a 300 s recycle delay is required
to produce the equivalent result—increasing the experiment time by a factor of five.
Despite the potential loss of quantitation and the added complexity of

the experiment, cross polarization is an essential tool for the study of dilute
spins at natural abundance. It makes 13 C an d 15 N spectroscopy feasible in
the solid state. It provides an extra tool for studies involving 29 Si and is a
useful aid to speeding up experiments on the heavy metals 77 Se, 113/111 Cd,
117/119 Sn, 195 Pt and 199 Hg. But does cross polarization always work
effectively? The short answer is no. Cross polarization relies on dipolar

coupling between 1 H and X and, to a lesser extent, between 1 H and 1 H,11
so anything that reduces or removes this or gives rise to a very long T XH or
very short missing image file can interfere with it. Molecular motion
in a gel or rubbery material can sufficiently interrupt the dipolar coupling to
render the sample invisible to a cross-polarization experiment. This is a case
where a direct-excitation experiment may be the only viable one. There is
also a problem if missing image file is very short (in the microsecond
range) since there is not enough time for a cross-polarization signal to build
up before the 1 H magnetization is lost. This situation often arises in
materials that are paramagnetic or contain a paramagnetic impurity.
3.4.3 DIRECT EXCITATION OR CROSS POLARIZATION? A

SUMMARY
With no hydrogen, 1 H–X cross-polarization

Does the sample contain hydrogen ?
is not possible, so direct excitation is the only option (this is likely to be the
case with a very low hydrogen content too). This may mean very long
accumulation times as (and hence the recycle delay) is likely to be long.
It may prove difficult to obtain a high signal-to-noise ratio for low
abundance nuclei.
Is the sample a rigid solid? If it is, cross polarization is the only realistic
choice for obtaining a spectrum with a high signal-to-noise ratio for natural
abundance 13 C or 15 N. For 31 P and, often, 29 Si, direct excitation can also
yield useful information even though long recycle delays may be required
(2–5 minutes are typical). For the heavy metals, the situation is complicated
by potentially large spinning sideband manifolds (which disperse the signal;
see section 3.3.4) or low resonance frequencies (which present some
instrumental challenges) and it is more difficult to generalize.
Is the sample a soft or mobile solid (rubber or gel like) ? Under these conditions cross
polarization may be ineffective, so direct excitation may be the only feasible
experimental method. However, in such materials, T 1 X is likely to be
relatively short so long recycle delays may not be required.
Will intensity information be used quantitatively? If so, a direct-excitation
experiment can be made quantitative easily (by setting the recycle delay to 5
× T 1 X ). If this is not feasible for dilute nuclei, then a series of cross-
polarization spectra, as a function of contact time, followed by numerical
analysis, will be required (see section 7.4).
Are low-intensity signals of particular interest? If they are, then cross polarization
(if it is effective) is likely to give a spectrum with the highest signal-to-noise
ratio.
Is the sample heterogeneous? Depending on the nature of the components
(rigid/mobile, 1 H rich/poor), heterogeneous samples may present some
special challenges and a combination of cross-polarization and direct-
excitation experiments may be required to characterize the sample. Given an
appropriate recycle delay, the direct-excitation measurement will give the
most quantitative picture of the composition of the sample.
Is the sample paramagnetic? Wholly paramagnetic samples cause special
problems. Nuclei close to a paramagnetic center may be undetectable (their
resonance lines being too broad to distinguish them from the baseline). More
distant ones may be observable but probably only with a direct-excitation
experiment ( missing image file is likely to be too short to allow
efficient cross polarization). Often a small amount of a paramagnetic
impurity can be tolerated and in some cases can cause a helpful shortening
of the T 1 X relaxation time (see section 7.5).
3.5 HIGH-RESOLUTION SPECTRA FROM 1 H (& 19 F)
It was noted at the beginning of this chapter that 1 H and, to some extent, 19
F are special cases. This is because, despite the most advanced experimental
techniques available at the time of writing, 1 H and 19 F linewidths, for true
solids, remain stubbornly high (often several 100 Hz) and this fact limits the
chemical information available from a spectrum. This is in stark contrast to
the solution state, where a proton NMR spectrum is usually the starting point
for the characterization for any soluble, proton-containing material.
Solution-state linewidths are typically of the order of 1 Hz. In the solid state,
strong homonuclear dipolar coupling is the main contributor to high
linewidths. For 1 H, the problem is compounded by its small chemical shift
range (~20 ppm) and that results in poorly resolved spectra (see inset 3.10).
The chemical information from a 1 H spectrum may be limited to the general
identification of signals as aliphatic, aromatic or, when present, strongly
hydrogen bonded (which usually have a distinct high-frequency shift). The
situation is slightly better for 19 F because the chemical shift range is over
400 ppm and useful chemical information can be usually obtained. Examples
of solid-state 1 H and 19 F spectra are given in figure 3.26 .
Inset 3.10. Spectral resolution
Spectral resolution is a measure of the degree to which two closely spaced signals
in a spectrum can be separately identified. The resolution depends on linewidth
and line separation as illustrated in the figure (right), which shows two lines with
Gaussian shape and half-height width Δ ν , (a) at the same chemical shift, (b) 0.8Δ
ν apart—unresolved (the flat top is characteristic of this situation), (c) Δ ν apart—
just resolved (the limit of resolution), and (d) 2Δ ν apart—resolved.
missing image file
A related issue is digital resolution. This is the frequency difference of successive

data points in the spectrum. It is inversely proportional to the number of points in
the Fourier transform. Fourier transformation of the acquired data points only (i)
may result in limited resolution. Adding zeros to the FID (zero-filling) so that the
number of points transformed is twice that acquired (ii) gives the maximum
resolution. Further zero-filling simply improves the aesthetic appearance of the
spectrum (iii).
missing image file

To obtain “high-resolution” proton spectra, it is necessary to overcome
the homonuclear dipolar coupling. High sample spin rates can achieve this to
a certain extent—the higher the better (usually above 25 kHz before a
significant impact is observed). For crystalline solids, high magnetic fields
are also a distinct advantage. However, even with the highest spin rates
achievable with traditional MAS technology, resolution is still disappointing
from the point of view of routine sample characterization.
The alternative to mechanically rotating the sample is to manipulate the
sample with a carefully designed sequence of RF pulses and interleave these
with data acquisition. The background to this homonuclear decoupling is discussed
in chapter 4 and is visited again in section 5.5. Combining multiple-pulse
homonuclear decoupling with two-dimensional methods, often detecting on
a nucleus such as carbon and obtaining a heteronuclear correlation spectrum
(see section 5.4.3), is generally a more promising approach.
missing image file
Figure 3.26. Examples of solid-state 1 H and 19 F spectra. (a) Naturally resolved 1 H spectrum from
the mineral octosilicate (spin rate 5 kHz). (b) 1 H spectra from 3-methoxybenzoic acid, magic-angle
spinning (MAS) at 15 kHz (upper trace) and with homonuclear decoupling (a windowed phase-
modulated Lee-Goldburg, wPMLG, spectrum) and spinning at 10 kHz. (c) 1 H spectrum from an
organic compound in a silica matrix recorded at a spin rate of 60 kHz at a field of 21.1 T. (d) 1 H high-
resolution (HR) MAS spectrum from an organic compound bound to a swollen polystyrene bead. (e)
19 F spectrum from polyvinylidenedifluoride. (f) 19 F spectrum from octafluoronaphthalene . Note the
different scale ranges. Figure 3.26 (c) supplied courtesy of Dr Anne Lesage.
These multipulse techniques can also be applied to fluorine, although

such extreme measures are probably needed only in systems with strong
homonuclear dipolar coupling such as inorganic fluorides or perfluorinated
organics. Sample spin rates of the order of 10–20 kHz are often adequate for
obtaining useful fluorine spectra. In organofluorine systems, fluorine atoms
may be dilute within a molecule and in such cases homonuclear coupling is
not a significant issue; instead it is efficient decoupling of the protons that
becomes important. However, decoupling 1 H while observing 19 F is not
trivial because of the similarity in their resonance frequencies
missing image file and a specially designed probe and RF filters are
needed to do this successfully. An added complication at low field (less than
11.7 T) is the Bloch–Siegert effect,12 which needs to be taken into account
when quoting chemical shift values and it may also have an impact on
linewidths under 1 H decoupling.
Cross polarization from 1 H to 19 F is possible. Because of the similarity
in their magnetogyric ratios there is no significant signal enhancement;
however, if the relaxation behaviors are different (proton significantly
faster), then there is a potential gain in signal-to-noise ratio by carrying out
the cross-polarization experiment. The potential selectivity of the experiment
can also be exploited. The dynamics of the cross-polarization process may
not have the relative simplicity of equation 3.8 in systems where both 1 H
and 19 F are strongly coupled (equation 3.8 is based on assumption that the X
spins are dilute). Cross polarization from 19 F to 1 H is also feasible and may
be useful for spectral editing.
Hydrogen- and fluorine-containing polymers are often used in the
construction of NMR probes and in the materials used to cap rotors (the
rotors themselves are usually made from a ceramic such a zirconia or silicon
nitride and, when clean, contain no hydrogen or fluorine). These components
will contribute a signal to the spectrum and that can become a problem if the
signal from the sample under study has a low intensity. It is usual to choose
materials so that this background signal is minimized . For example, Teflon®
(tetrafluoroethylene) or Kel-F® (trifluorochloroethylene) contain only
residual traces of hydrogen so are ideal for use when recording proton
spectra but should be avoided when observing fluorine.13 Instead, a hard-
wearing polyimide such as Vespel® is ideal for fluorine. In practice, some
compromises may have to be made and it is not always possible to fully
avoid background signals. In such cases it is advisable to record a spectrum
from the empty rotor and, if necessary, subtract it from the spectrum
obtained from the sample under study. Pulse sequences are also available for
reducing the intensity of background signals (although it is still advisable to
record a spectrum from the empty rotor).
1 H (or 19 F) do not always experience strong homonuclear coupling and
combinations of molecular mobility and relative isolation may mean that
well-resolved spectra can be obtained without any special experimental
methods. Soft solids, such as gels or tissue samples, fall into this category.
Resolved 1 H spectra can be obtained from these with MAS probes designed
for work with true solids. However, these are often not optimized for the
highest resolution in the same way that solution-state probes are and
resolution can be improved by using a specifically designed high-resolu tion
MAS (HRMAS) probe (see figure 3.26 (d)).
At the other extreme, the broad lines obtained from a non-spinning,
hydrogen-rich material yield no chemical information directly but can give
useful insights into the physical characteristics of the sample. This is a topic
that is examined in chapter 7 .
In this chapter, basic experimental NMR methods for the study of solids
have been introduced. Although there is a vast range of possible solid-state
NMR measurements, one or other of these methods is central to most of
them. The components of the experiment that control the appearance of the
spectrum it produces have also been introduced.
In the next chapter the theme changes from practical issues to the
quantum mechanics of solid-state NMR. This chapter should give the reader
an insight into the theory of solid-state NMR. It is not essential to the
understanding of the subsequent chapters, so if this insight is not required
yet, go on to the rest of the book and come back to it later.
FURTHER READING
DESCRIPTION OF THE NMR EXPERIMENT
“Understanding NMR spectroscopy” , J. Keeler, John Wiley & Sons (2005), ISBN 978 0 470 01787 6.
PULSE NMR AND PRACTICAL ISSUES RELATING TO SIGNAL
PROCESSING
“Modern NMR techniques for chemistry research” , A.E. Derome, Pergamon Press (1987), ISBN 0 08
032513 0.
“High-resolution NMR techniques in organic chemistry” , T.D.W. Claridge, Elsevier (2009), ISBN
978 0 08 054818 0.
NOTES
1 The axes of the rotating frame should be labeled differently to those of the fixed laboratory frame
(say, x ′ , y ′ , and z ′ instead of x , y, and z ). However, it is customary to omit the distinction. Phases in
this sense are relative labels and have no absolute physical meaning.
2 This rotation is what is expected from the classical physics of two interacting orthogonal magnetic
fields (the bulk magnetization and B 1 ; in the rotating frame and on-resonance, B 0 disappears).
Conventions for the direction of rotation vary—consistency is the important factor in describing an
experiment.
3 There is also a trade-off between pulse angle and recycle delay. If T is known, then the pulse angle,
1
θ , that results in the maximum amount of magnetization in the xy plane for a chosen recycle delay is
given by cosθ = exp(-recycle delay /T 1 ). This angle is known as the Ernst angle.
4 These sinc wiggles appear only if the signal is zero-filled (as is usually the case, see inset 3.10) prior
to Fourier transformation.
5 In fact, modern spectrometers automatically oversample (which is equivalent to using a very high
spectral width) and then digitally filter and downsample to give a more manageable number of data
points (all usually hidden from the spectrometer operator). This process is unlikely to result in folded
back signals.
6 Sometimes this may simply be a case of aesthetics, but if intensity information is required a flat
baseline is usually essential.
7 “Continuous wave” is a term in widespread use, but it should be noted that for solid-state NMR the
decoupling is not actually continuous. Here, “continuous wave” is merely a label for decoupling with
no amplitude or phase modulation.
8 For details see Pure Appl. Chem. 80 (2008) 59.
9 Hexamethylbenzene (HMB) is often used as a set-up compound for 13 C because it gives a well-
defined match profile irrespective of other instrument variables such as decoupling efficiency or
shimming. Adamantane (see inset 7.8) is a highly mobile solid that gives very narrow lines and is a
good check for lineshape.
10 In figure 3.21 the spin-lock and decoupling pulses have the same phase so the spin-lock is
maintained even when there is an increase in RF field for decoupling. The spin-lock is lost if the 1 H
irradiation is turned off, as it would be for a dipolar dephasing experiment (see section 7.2.4), or if
more complex decoupling schemes than simple CW irradiation are used.
11 1 H,1 H coupling enables spin diffusion to be efficient among the protons and this leads to
increased polarization transfer, making it more likely that a signal enhancement closer to the
theoretical missing image file maximum will be achieved.
12 When missing image file 1 H decoupling has no detectable impact on the chemical shift of the
observed X nuclei. If this condition does not hold (as it might not for 19 F observation with 1 H
decoupling), then a small shift in the X resonance frequency may be observed. This is known as the
Bloch–Siegert effect.
13 Because these materials contain no 1 H, they are also ideal for use when observing 13 C using 1 H
to 13 C cross polariz ation (in effect they are invisible). They will, however, contribute signal to a
direct-excitation spectrum.
CHAPTER 4
QUANTUM MECHANICS OF SOLID -STATE NMR
4.1 INTRODUCTION
The previous chapters have deliberately avoided delving deeply into the
theory underlying solid-state NMR. Indeed many solid-state NMR
experiments can be understood in terms of the vector model familiar from
the solution state (and discussed in section 3.2), so that most experiments
can be interpreted using straightforward qualitative pictures of the different
NMR interactions and their effects on the spectrum. As the subject has
progressed, however, the complexity of experiments and the associated
theoretical background have considerably increased. This chapter aims to
provide an introduction to the various theoretical tools used in solid-state
NMR. The Further reading lists some more detailed and thorough texts, but
this concise overview should give the reader sufficient insight to be able to
explore further and to tackle the NMR literature with confidence. Those
only interested in using the results of solid-state NMR will probably prefer
to skip over this chapter and go to the applications of solid-state NMR,
ignoring the references to theory when some of the more complex
experiments are discussed.
The theory of solid-state NMR necessarily involves quantum mechanics

which often appears to bear little relationship to the quantum mechanics
covered in undergraduate chemistry courses. However, as will be seen
below, the quantum mechanics of NMR is in many ways simpler and more
elegant than typical quantum chemistry.
Ultimately, all quantum mechanics reduces to solving the Schrödinger

equation missing image file where ψ is a wavefunction for the system, E
is an allowed energy corresponding to the wavefunction ψ , and
missing image file is the Hamilton operator , or Hamiltonian , which
describes the energy of the system. We can write missing image file
schematically for a system of nuclear spins and the surrounding “system”:
missing image file

where missing image file contains terms that only depend on the nuclear
spins, missing image file refers to terms with no dependence on the
nuclear spin state (such as molecular vibration and rotation), while
missing image file contains terms that involve both the nuclear spins and
the surrounding environment, for example, the J coupling, which is
mediated by the electronic wavefunction.
This total Hamiltonian is tremendously complex, with degrees of freedom

(independent variables) for the nuclear and electron spins, the positions of
the atoms, etc. Fortunately the “coupling” (degree of interaction) between
the nuclear spins and their environment is extremely small, which allows
the problem to be vastly simplified. The largest of the NMR interactions
introduced in chapter 2 , the Zeeman interaction, is only of the order of
100s of MHz1 and all other energy terms in missing image file (vibration,
rotation, electronic, etc.) are associated with much higher frequencies. The
nuclear spin energy has negligible impact on the other degrees of freedom
of the total wavefunction and these other components evolve on much
faster timescales. The external degrees of freedom can therefore be
integrated out, leaving a Hamiltonian that only depends on the nuclear
spins. In this spin Hamiltonian , missing image file is a constant and so
can be ignored, while the terms involving both nuclear spin and external
coordinates, missing image file , can be simplified to terms involving
nuclear spin operators and constants that depend on the external system.
For instance, the (isotropic) J-coupling Hamiltonian between two nuclear
spins, I and S , over the NMR timescale, becomes
missing image file
So while the full J-coupling Hamiltonian involves both the spin coordinates
and the entire electronic wavefunction, the corresponding spin Hamiltonian
just involves the states of the two relevant nuclear spins and a coupling
constant, J IS , which results from integrating out the other degrees of
freedom.
This “decoupling” of spin and space coordinates greatly simplifies the

analysis of NMR. While the wavefunction for a multi-electron system has
no analytical solution, so that deriving increasingly accurate solutions is the
main challenge of quantum chemistry, the Schrödinger equation for a
number of interacting nuclear spins is generally embarrassingly easy to
solve. The drawback of splitting off the nuclear spin Hamiltonian from the
complete system is that terms involving the external degrees of freedom are
reduced to empirical constants, such as coupling constants and chemical
shifts. Calculating these constants from first principles ( ab initio quantum
chemistry) is difficult since it requires the full multi-electron Hamiltonian.
Historically, NMR parameters have generally been treated as empirical
quantities that are compiled, and empirical correlations are then derived
between their values and chemical/ structural features. Increasingly,
however, quantum chemistry is able to provide robust estimates of NMR
parameters, particularly shielding and quadrupole coupling constants,
allowing the NMR parameters to be tied in a more direct and meaningful
way to the underlying molecular structure (see section 8.8.2).
4.2 THE HAMILTONIANS OF NMR
NMR theory is further simplified by working at sufficiently high magnetic

field that the Zeeman interaction dominates all the other interactions. As
discussed in chapter 6 on quadrupolar NMR, this is not always achievable
for nuclei with large quadrupole moments and low NMR frequencies, but,
even here, the quadrupole interaction can usually be treated as a
perturbation on the dominant Zeeman term. As there is little chemical
interest in performing NMR at low magnetic fields, we will work
exclusively within this high field approximation .
If we consider a single spin with nuclear spin quantum number I , the

different spin (angular momentum) states can be conveniently represented
in the Dirac bracket notation, equ missing , where m I is the magnetic
component quantum number (see section 2.1). The Hamiltonian for the
Zeeman interaction is
where the magnetic field, B 0 , has its conventional alignment along the z
axis, and Î z is the operator for the z component of the nuclear spin angular
momentum. The Schrödinger equation is trivial to solve for this
missing image file since the 2I + 1 spin states are already eigenfunctions
of Î z
Hence missing image file 4.3
This recovers the simple expression for the energy of a nuclear spin in a
magnetic field presented in chapter 2 :
4.2.1 SPIN OPERATORS
The term Î z used to represent the z magnetization operator in equation 4.1

deserves further comment. This is one of a set of operators that extract
information about the nuclear spin angular momentum from the nuclear
spin wavefunction (see equation 4.2). Ultimately all Hamiltonians
involving nuclear spins are defined in terms of such operators.
The Î x and Î y operators correspond to x and y components of the angular

momentum respectively. It is often simpler, however, to express the x and y
operators in terms of the so-called raising and lowering operators
These names are derived from their effect on the Zeeman eigenstates:
If we consider, for example, a spin- missing image file then

missing image file or, using the conventional shorthand
missing image file and missing image file then missing image file
that is, Î + “raises” the missing image file state to missing image file
Note from equations 4.6 and 4.7 how trying to raise a state of maximum m
I or lowering a state of minimum m I fails: Figure Missing
Many NMR problems can be expressed simply using such symbolic

representations for spin operators. For example, the quantum mechanical
description of the NMR experiment in section 4.4.1 uses spin operators to
express the state of the system. Particularly when dealing with multi-spin
systems with strong couplings between them (as commonly encountered in
solid-state NMR), however, operator treatments can become unwieldy. In
these cases it is often simpler to work with matrix representations of
operators. Numerical simulations as discussed in appendix D, for example,
invariably make use of matrix representations.
The matrix representation of an operator, Ô , is obtained by evaluating all

possible combinations of bras and kets, Figure Missing Thus for spin-
missing image file
Note that the factor of missing image file has been dropped from the spin
operators. This corresponds to the normal NMR convention of expressing
the spin Hamiltonian and its components (such as coupling constants) in
terms of frequency rather than energy.
It follows from equation 4.5 that Î x = (Î + + Î − )/2 and Î y = − i(Î + − Î −

)/2, hence
Inset 4.1 shows an example of deriving matrix representations for a spin

with a higher value of I .
4.2.2 SECULAR & NON-SECULAR TERMS
Because the other interactions are usually small in comparison with the
Zeeman interaction, the eigenbasis of the total nuclear spin Hamiltonian is
the same as the Zeeman eigenbasis figure missing (except when large
quadrupole interactions are present, as discussed in chapter 6 ). Moreover,
components of the interaction Hamiltonian that do not commute with the
Zeeman interaction can be neglected .
Inset 4.1. Matrix representations of spin operators

Matrices for general spin quantum numbers can be generated using
equations 4.2–4.7. For example, a spin- missing image file has four spin
states, missing image file and missing image file and the associated
representations are 4 × 4 matrices:
missing image file
The non-zero matrix elements for Î + will be figure missing and

missing image file that is, figure missing Hence
missing image file
This is a subtle but recurring theme in NMR quantum mechanics and is

worth considering in detail. In classic perturbation theory, we consider the
effect of a perturbing Hamiltonian, figure missing on the eigenvalues and
eigenstates of a dominant Hamiltonian, figure missing Working in the
eigenbasis of figure missing , with the set of states, figure missing the
first-order correction to the energy of level n is just figure missing In
other words, the off-diagonal elements of figure missing (expressed in the
figure missing eigenbasis) have no effect to first order on the energy. In
more formal terms, the perturbing Hamiltonian can always be divided into
a component that commutes with figure missing (see inset 4.2),
figure missing and a component that does not, figure missing The
secular approximation involves discarding figure missing since it has no
effect on the NMR frequencies to first order. We will see an alternative, but
equivalent, way of looking at this problem when considering the rotating
frame of reference in section 4.2.4.
The shielding Hamiltonian, responsible for the chemical shift, provides a

concrete example of the distinction between “secular” and “non-secular”
terms. The full shielding Hamiltonian is
where B is the magnetic field vector, Î is the vector (Î x , Î y , Î z ), and σ is

the shielding tensor. If the field is along z , that is, Β 0 = (0, 0, B 0 ), this
simplifies to
However Î x and Î y do not commute with Î z (a basic principle of quantum

angular momentum), and so these terms are non-secular with respect to the
Zeeman interaction and can be neglected. Only the term in Î z will add to
the diagonal and so contribute directly to the energies of the Zeeman
eigenstates. Hence the shielding Hamiltonian can be simplified to ν NMR σ
zz Î z , as previously assumed in section 2.3.
Inset 4.2. Reminder of some basic matrix terminology

Commuting and non-commuting matrices The order in which two
matrices, A and B , are multiplied is important since frequently AB ≠ BA .
Such matrices are said to be non-commuting . On the other hand, if AB −
BA = 0 , then A and B are said to commute , and the order of matrix
multiplication can be freely interchanged. The commutator between A and
B , AB − BA , is often written in the shorthand form [ A , B ].
Matrix trace The trace of a matrix is the sum of its diagonal elements
For instance, the expectation value of an operator Ô given the density

operator missing image file is missing image file .
Handy relation : The trace of a product is invariant to cyclic permutation,

tr( ABC ) = tr( CAB ).
Matrix inverse The inverse of a square matrix A is denoted by A −1 and

has the property
where 1 is an identity matrix.
Handy relation : ( AB ) −1 = B −1 A −1
Transpose and conjugate transpose Transposing a matrix A (common
symbol A T ) involves swapping matrix elements across the diagonal. The
conjugate transpose , A † , involves taking the conjugate of the matrix
before or after transposition:
Handy relation : ( AB ) T = B T A T
Alternatively, the matrix representation of the total Zeeman and shielding

Hamiltonians for a spin- missing image file would be
The diagonal terms of the shielding Hamiltonian give the first-order

correction to the eigenvalues, hence the overall NMR frequency (given by
the difference of the eigenvalues) is ν 0 = ν NMR (1 − σ zz ), as expected.
The off-diagonal terms would contribute to a second-order correction to the
energy, but this can be readily neglected; the effect of the first-order term
on the NMR frequencies is measured in parts per million, and so the
second-order correction due to the Î x and Î y components of the shielding
Hamiltonian is of the order of 1:1012 .
4.2.3 COUPLING HAMILTONIANS
The full Hamiltonians for NMR couplings have a common form:2
where Î j indicates the vector of spin operators (Î x , Î y , Î z ) for spin j ,

and R is a tensor (see section 2.2 for an introduction to tensors and section
4.6 for a more detailed treatment). As previously discussed in section 2.2, R
can always be decomposed into a rank-0 term (isotropic), a rank-1 term
(which can be ignored), and a rank-2 term.
The final form of the equation 4.18 depends on the nature of the
interaction. For instance, the rank-2 component of the J (or indirect dipole–
dipole) coupling is either eliminated by molecular tumbling in the solution
state or can be merged into the much larger direct dipole–dipole interaction
(see equation 8.3). Hence the J coupling between two spins I and S is
normally reduced to its isotropic component
where the scalar (dot) product between the spin operators expands to
figure missing In the homonuclear case, I and S have the same Larmor
frequency, ν I , and the Zeeman Hamiltonian is
In this case, the secular approximation does not provide further

simplification to figure missing since it commutes with figure missing
(exercise for the reader!).
If, however, the difference in NMR frequencies is large in comparison with

J IS , either because they form a heteronuclear pair or because their
chemical shifts are very different, then further terms can be discarded. The
dominant component of the Hamiltonian is:
In this case, only the figure missing terms of figure missing commutes
with figure missing and the coupling Hamiltonian can be reduced to
This corresponds to the weak coupling limit familiar from solution-state

NMR spectra.
The full dipole–dipole coupling Hamiltonian between I and S in the terms

of equation 4.18 is
where D is the dipolar coupling tensor. D in its principal axis system is

given by3
where D IS is the dipolar coupling constant between I and S (see section

2.5). Note how D has zero trace (i.e., no isotropic component) and zero
asymmetry missing image file as previously presented in table 2.1.
The algebra of the reduction to the secular component is considerably more

involved and we will jump straight to the result (see Further reading for
details):
where P 2 (x ) is the second-order Legendre polynomial, P 2 (cos θ ) =

(3cos2 θ − 1)/2. The second term of equation 4.25 is often referred to as the
“flip-flop” term since it has the effect of interconverting the states
figure missing (see equation 4.76).
As with the J-coupling Hamiltonian above, this Hamiltonian commutes

with the Zeeman Hamiltonian for a like-spin pair, but only the zz term
survives for a heteronuclear spin pair:
which is equivalent to the expression for the nuclear spin energies given in
equation 2.23.
The presence of the additional flip-flop term in the homonuclear

Hamiltonian of equation 4.25 is responsible for the difference between the
spectra for a homo- and a heteronuclear spin pair referred to in section 2.5
and discussed in detail in section 4.5.1.
As mentioned above, the secular approximation begins to break down for

the quadrupolar coupling, and so discussion of the quadrupolar coupling
Hamiltonian is deferred to chapter 6 .
4.2.4 RADIOFREQUENCY & THE ROTATING FRAME
The final term required for this quantum treatment of the NMR experiment
is the Hamiltonian for the radiation used to excite the nuclear spins. The
Hamiltonian for a nuclear spin interacting with electromagnetic radiation is
where B 1 is the intensity of the oscillating magnetic field and ø is its phase
(with zero conventionally corresponding to initial x phase). ν RF denotes
the frequency of the radiation from the radiofrequency (RF) transmitter.4
The reason for the leading factor of 2 will appear later.
The time dependence of figure missing greatly complicates solving the

Schrödinger equation, and the normal solution is to move into an
interaction frame rotating about the z axis at frequency ν RF , as introduced
qualitatively in section 3.2. The transformation from the normal laboratory
frame to this rotating frame corresponds to the following transformation of
the axis system
that is, the xy axes rotate around the z axis with angular frequency ω RF
(the use of angular frequencies avoids factors of 2π).
After multiplying out and applying some trigonometric identities, equation

4.27 reduces to
that is, one component of the RF is now independent of time in the new
(rotating) frame, while the other is rotating at an angular frequency of 2ω
RF .
The frequency of this second component of the oscillating magnetic field is

very much faster than the NMR interactions (in the rotating frame) and so
averages away over the NMR timescale.5 Hence the RF Hamiltonian
reduces to

Note how the factor of 2 present in equation 4.27 has disappeared along
with the discarded component of the magnetic field. The primes used to
denote the rotating frame have been dropped since the rotating frame will
be used from now on.
The key feature of equation 4.31 is that the RF Hamiltonian is time

independent in this frame of reference, in contrast to the laboratory frame
Hamiltonian (equation 4.27). This greatly simplifies the task of determining
the evolution of the nuclear spins. Even when the implicit approximation
that the Zeeman interaction is much larger than the other components of the
nuclear spin Hamiltonian becomes slightly questionable, for example, when
the quadrupolar interactions are very large, it is normal practice to add
correction terms using second-order perturbation theory rather than
abandon the rotating frame altogether.
The transformation into the rotating frame also affects the Zeeman /
chemical shift Hamiltonian . The Zeeman + shift Hamiltonian in the
laboratory frame is
(for a single spin), where ν 0 refers to the Larmor frequency (including the
shielding term). The effect of this Hamiltonian is to cause the net nuclear
magnetization to precess about z at its NMR frequency. The effective
precession rate in the rotating frame will be missing image file Hence the
effective Hamiltonian in the rotating frame is
The total Hamiltonian will then be
where missing image file 4.35
By definition the RF needs to be close to the Larmor frequency, ν 0 , for

nuclear magnetic resonance to be observed, and so the offset term, − Δ Î z ,
will be negligible in comparison with the RF term. Hence the net
magnetization vector will rotate around the rotating-frame B 1 vector at a
rate ν 1 . This is the nutation frequency discussed in chapter 3 . If, however,
the RF is far from resonance and/or very weak, the Δ Î z term will dominate
and the spins will precess around z , unaffected by the RF irradiation.
4.3 THE DENSITY MATRIX
Understanding spin dynamics in solution-state NMR is relatively

straightforward largely because the only significant interactions directly
affecting the spectrum are chemical shifts and J couplings. Moreover, these
interactions commute with each other in the weak coupling approximation
(that the J couplings are much smaller than the differences in Larmor
frequencies between inequivalent spins). This allows the effects of the shift
and coupling Hamiltonians to be considered separately, and relatively
simple rules can be derived to describe the effects of the NMR interactions
and an RF pulse sequence. See Further reading for books that describe such
product operator treatments. This rarely applies in solid-state NMR, and so
a more thoroughly quantum mechanical treatment is generally required.
This is outlined below (again see the Further reading for more gentle
introductions to the quantum mechanics of NMR).
The state of an individual system is determined by its wavefunction, Ψ.

This can always be expressed as a linear combination of basis functions
(generally just the Zeeman eigenbasis) that is,
and so the “value” of an operator, Ô , (its expectation value ) will be
If we have an ensemble of identical systems (i.e., systems with the same

Hamiltonian), then the ensemble-averaged expectation value of Ô is
The matrix elements, figure missing are identical between the different
systems, hence the measured expectation value reduces to
that is, the only quantities we need to define are the ensemble-averaged
coefficients missing image file Hence it is sufficient to define the state of
the system using the matrix, ρ , with elements
ρ is referred to as the density matrix and is the matrix representation of the

density operator , missing image file (although ρ and missing image file
are often used interchangeably). The expectation value of Ô is then simply
where tr is the matrix trace (see inset 4.2). We lose information about the
wavefunction of an individual system, but this is not a great loss as the
NMR experiment invariably involves measurements on large ensembles.
For example, the equilibrium state of the nuclear spins within a magnetic
field involves net magnetization along the z axis, with net excess of spin
magnetization, M 0 /2, in, say, the α state. The density matrix6 for the
equilibrium state would therefore be
We can then obtain expectation values for the different angular momentum
operators of equation 4.9, for example:
that is, the average state corresponds to pure magnetization along z .
Note how the diagonal elements of the density matrix, figure missing ,
directly correspond to the populations of the states. The physical
significance of the off-diagonal elements is less immediately obvious, but
consider the density matrix
Clearly the average populations of the α and β states are zero, but the
system is still in a well-defined state. This is impossible classically since a
two-level system must be in one of the two available states at any one time.
Quantum mechanics, however, allows different states of existence,
coherences , which involve mixtures of the eigenstates. Hence we need a
matrix to hold the populations of the states (diagonal elements) and the
coherences (off-diagonal elements) that can be created between the
different states.
Figure 4.1 illustrates different ways of representing the state of the set of
nuclear spins. The vector model, figure 4.1(a) , is useful when there is a
single spin type or when couplings between spins can be ignored. It is
difficult, however, to represent coherences other than simple x , y , and z
magnetization in this way. Especially when problems involve an isolated
pair of spins, it can be useful to consider an energy level diagram such as
figure 4.1(b) . The relative populations of the four possible states are
indicated by the occupancy of each level while the arrows mark possible
coherences between the states. Coherences can be illustrated with such
diagrams, for example, the single-headed arrow would indicate a coherence
(of order + 2) between the figure missing and figure missing states, but
this could be confused with a transition or a relationship between the
populations of the states involved. The density matrix representation, figure
4.1(c) , provides the most compact representation of the different possible
states of the system.
missing image file
Figure 4.1. Different ways of representing coherences in NMR: (a) the

vector model of a collection of spins in a magnetic field and the resulting
net magnetization; (b) schematic representation of the energy levels for two
spin- missing image file nuclei. Filled circles represent relative
populations of the spin states and double-headed arrows correspond to
potential coherences between the states. (c) The density matrix for this
system. Numbers give the coherence order, p , for the blocks, see section
4.5.1.
For instance, the population distribution in (b) would correspond to entries
of figure missing figure missing and figure missing down the
diagonal of the matrix, while the + 2 coherence would correspond to a non-
zero value for the matrix element connecting figure missing Note how the
density matrix can be divided into blocks of a given “coherence order ”.
This important concept is explored in section 4.5.1.
The NMR experiment can now be described in this formalism, starting with
the density operator that describes the initial equilibrium state of the
system. At equilibrium there are no net coherences between the states, and
the relative populations of the eigenstates are given by the Boltzmann
distribution. The equilibrium density operator is7
where missing image file is the total nuclear spin Hamiltonian and Z is a
normalization factor (the partition function). Since the nuclear spin energies
are so small in comparison with thermal energies, the exponential can be
expanded
where Z equals the number of spin states, 2I + 1, in this limit (see inset 4.3
for an explanation of how to take exponentials of matrices such as
missing image file ). missing image file is the identity operator and so
represents an equal distribution of population between the different states.
This term is of no interest (since it does not evolve under
missing image file ) and can be dropped. The second term corresponds to
the tiny changes to the populations caused by the lifting of the degeneracy
of nuclear spin states by the spin Hamiltonian missing image file .
Hence we will work with the reduced density operator ,

missing image file which gives the difference of the full density operator
from the situation of equal population. The expression for the equilibrium
density matrix is further simplified using the high-field approximation, that
is, missing image file The equilibrium density operator for a single spin
can then be written simply
where M is a constant of proportionality that effectively expresses the size

of the nuclear spin magnetization. For instance, missing image file for a
single spin- missing image file in these terms would be
which corresponds to net positive population of pure α states and a net

negative population of β states.
Applying the Schrödinger equation to the density operator gives the

Liouville–von Neumann equation
for the evolution of the density operator under the Hamiltonian,

missing image file (again expressed in frequency units).
Integrating equation 4.50 gives the density operator at time t
where missing image file is the density operator at time t = 0 and Û (t , 0)

is the propagator —the matrix that “propagates” the density operator from
time zero to time t . For a time-independent Hamiltonian, the propagator is
given by
4.4 DENSITY OPERATOR TREATMENTS OF SIMPLE NMR

EXPERIMENTS
The density operator approach comes in to its own when multiple spins are
involved, but it is useful to start with some simple experiments that can be
easily visualized in vector model terms.
4.4.1 THE BASIC NMR EXPERIMENT

Consider a system of identical spins at thermal equilibrium. From equation
4.48, the density operator at time zero is
The system Hamiltonian (in a frame rotating at the same rate as the RF)
consists of a small chemical shift offset term
Because missing image file and missing image file commute (they are
simply proportional to each other), the Liouville–von Neumann equation,
equation 4.50, reassuringly predicts no evolution of the density operator
missing image file
If we apply RF of phase x , the Hamiltonian is now
Inset 4.3. More advanced matrix algebra

Matrix diagonalization In matrix terms, solving the Schrödinger equation,
H ψ = Ε ψ , involves finding the eigenvalues of the Hamiltonian matrix H
(which correspond to the allowed energies, E ). The eigenvectors of H
describe the transformation from the initial basis set into the basis that
diagonalizes H . This is written symbolically:
where V is the matrix of eigenvectors and Λ is a diagonal matrix with the

eigenvalues, Λ i , along the diagonal.
For example, considering a Hamiltonian matrix containing a dominant

diagonal term and weak off-diagonal component:
The eigenvalues and eigenvectors are

V is a rotation matrix, with transformation V -1 HV generating the diagonal
matrix Λ containing the eigenvalues (energies).
Handy relation : if two matrices commute, then the same matrix V will
diagonalize both matrices, that is, they share a common eigenbasis.
Matrix exponentials Exponentiation is defined for matrices as for simple

scalars:
where k is a scalar.
In practice this series converges slowly and the exponentials in equation

4.52 are typically evaluated via the matrix eigenbasis:
where the elements of the exponential of the diagonal matrix Λ are simply
[exp( Λ )] ii = exp( Λ i ).
Handy relations : (i) if matrices A and B commute, then e A+B = e A e B =

e B e A . If A and B represent Hamiltonians, for example, then this means
that the propagators (see equation 4.52) for A and B can be evaluated and
applied to the density operator independently; (ii) [e A ] −1 = e − A .
where ν 1 is the nutation rate. As discussed in section 4.2.4, the offset, Δ,

can be neglected if figure missing hence the propagator for such a “hard”
(or ideal) pulse of duration τ is (equation 4.52)
where the tip angle is θ = 2π ν 1 τ . This propagator corresponds to rotation

about the x axis, and so the density operator at the end of the pulse will be8

missing image file 4.6 4
that is, missing image file after a missing image file pulse. This
corresponds to the nuclear spin magnetization now being aligned with the y
axis.
When the RF is turned off, the Hamiltonian returns to equation 4.59 and the
propagator for the following period of free precession (evolution under the
system Hamiltonian in the absence of RF) is
This simply corresponds to rotation about the z axis at rate Δ. Hence the
density operator is
Note how in periods of free precession, the evolution of the density

operator does not change the order, p , of coherences; initial y ( p = ± 1)
magnetization evolves into a mix of x and y (also ± 1). By contrast,
evolution under RF mixes different coherence orders, for example, the
interchange between z ( p = 0) and y ( p = ± 1) magnetization in equation
4.64.
To determine the NMR signal, we need to evaluate the expectation values

of Î x (nominally x magnetization in the “real” channel) and Î y operators (
y magnetization in the “imaginary” channel). These are combined to give
the complex NMR signal, entity image missing allowing positive and
negative precession frequencies to be distinguished (see inset 3.8).9 This
neatly corresponds to determining the expectation values of the Î + = Î x +
iÎ y operator. Here we can read out the coefficients of Î x and Î y directly
from equation 4.66 to give

that is, an NMR signal of frequency Δ (relative to the transmitter frequency
used for the rotating frame of reference). The factor of i corresponds to a
phase shift of 90° relative to x phase, since the magnetization at t = 0 is
aligned with the y axis.
Note that evaluating the NMR signal as tr( missing image file Î + )
corresponds to detecting p = − 1 coherences of the density matrix. For
example, if the density matrix is a hypothetical pure − 1 coherence,
missing image file = MÎ − , the observable NMR signal at this point is
The comparison between the vector model and density operator treatments
of this problem is illustrated in figure 4.2 .
missing image file
Figure 4.2. Comparison of (top) vector model and (bottom) density

operator treatments of the pulse-and-acquire NMR experiment (see text for
further details).
4.4.2 ECHOES & COHERENCE PATHWAY DIAGRAMS
Echoes of various kinds form the central building blocks of many NMR
experiments. Two of the most important of these, the simple (or Hahn )
echo and the solid (or quadrupolar ) echo , are illustrated in figure 4.3 .
The simple spin echo is used to “refocus” the effect of Hamiltonian terms
that are “linear” in the spin of interest, such as the chemical shift and
heteronuclear couplings. After initial excitation, in this case using a
missing image file pulse to create y magnetization from the equilibrium
state, the xy coherences (or equivalently + 1 and − 1 coherences) evolve
during the free precession period, τ . Including just linear terms, the free
precession Hamiltonian might be:
missing image file

Figure 4.3. Pulse sequences for (a) a simple spin echo and (b) a solid (or
quadrupolar) echo, and (c) the coherence pathway diagram that applies to
(a). The bold solid line in the coherence pathway diagram marks the path
taken by the coherences that are detected (−1). The lighter line traces out a
mirror pathway which is not detected (see text).
where Δ is the offset from the origin of the rotating frame, and J is a scalar
coupling between I and S spins. The problem can, however, just be
expressed in terms of the I spin, since the sub-spectra corresponding to
different states of the S spin are fully independent. For example, if S is a
spin- missing image file , then we can consider an effective offset, Δ′ = Δ
± J /2, reducing the Hamiltonian to missing image file
Using the same arguments discussed above, initial y magnetization,

missing image file evolves into
at the end of the first τ period and before the inversion pulse.
The missing image file pulse will invert the sign of the y magnetization
and leave the x component unchanged, that is, the density operator after the
pulse is:
The second period of free precession again gives rises to a “rotation” of

2πΔ′τ about z . Applying this to the Î y and Î x operators gives the final
density operator
Using the trigonometric identity cos2 θ + sin2 θ = 1, the density operator at

2τ simplifies to

that is, the density operator corresponds to magnetization along – y . In
other words, the effect of the chemical shift and heteronuclear couplings is
refocused irrespective of the effective offset Δ′. In typical situations where
there is a spread of offset frequencies, the xy magnetization will decay as
the spins precess at different frequencies during the initial τ period, but this
dephasing is reversed during the second τ to create an “echo” at 2 τ when
all the spin packets are aligned along – y .
Figure 4.3(c) shows the coherence pathway diagram that applies to the
spin-echo (and also the quadrupole-echo) experiment. These diagrams
summarize the changing state of the spin system during the experiment in
terms of a “pathway” showing the order of coherences (zero quantum,
single quantum, etc.) present at different steps of the experiment. The
pathway always starts at zero order, corresponding to equilibrium z
magnetization, and finishes with observable p = –1 coherence
magnetization during the acquisition period.
During free precession periods (no pulses), the different coherences evolve
but do not change order. For instance, initial Î x magnetization will evolve
under the influence of a heteronuclear coupling to a spin S to create mixed
coherences of the form figure missing These are still, however, single
quantum coherences, and so the coherence pathways are horizontal during
such delay periods. By contrast, RF pulses mix coherences between the
different orders that the spin system can support, allowing the pathway to
jump between coherence orders. It is important to note that the coherence
pathway only summarizes the (desired) coherences present at a particular
point and does not predict the detailed state of the system. For instance, the
presence of an echo at 2τ must be deduced from analytical treatment of the
density operator.
Considering the coherence pathway diagram for the spin-echo experiments,

figure 4.3(c) , the excitation pulse converts initial z magnetization into y
magnetization (a sum of + 1 and –1 coherences). The 180° pulse then
interchanges + 1 and –1 coherence. Finally, the –1 coherence is detected
during the acquisition period. Hence the pathway of interest, marked in
bold, corresponds to 0 → + 1 → –1.
As previously referred to in section 3.4.1, the purpose of phase cycling is to
ensure that unwanted pathways are suppressed. For example, an imperfect
initial 90° pulse will leave some residual zero-order (population)
coherence, which will translate to unwanted additional signals if these
coherences are propagated through to –1 coherence by later pulses. The key
concept behind any form of phase cycling is that changing the phase, ø , of
the RF pulse has a different effect on different coherence orders.
Specifically for a given change in coherence order due to a pulse, Δp ,
changing the phase of the pulse by Δø will change the phase of the relevant
coherences in the density matrix by –Δø Δp . The phase cycle is designed
so that unwanted coherences acquire different phases over the course of the
cycle such that the unwanted terms cancel when the overall signal is added
together. The pulse phases in the spin-echo experiment of figure 4.3(a) , for
example, lead to magnetization being refocused on the –y axis. Shifting the
phase of the refocusing π pulse by 90° results in refocusing along the + y
axis. Hence the desired coherence, which has passed through the Δp = –2
pathway, has acquired a phase shift of 180° . If the receiver is shifted by
180° between the two experiments, then the desired signal will add
constructively, while unwanted signals with Δp = 0 (corresponding to
components of the density matrix that have not been inverted) will be
cancelled. Since achieving perfect inversion across the full spectral
width/sample is often difficult, this phase cycling is very important in spin-
echo experiments. See Further reading for more details on the construction
of phase cycles.
Figure 4.3(b) shows the pulse sequence for the solid (or quadrupolar) echo
experiment. This has strong similarities to the simple spin echo (indeed the
coherence pathway diagram is identical), but the details are different since
the goal is to refocus the evolution under “bilinear” Hamiltonians, such as
the quadrupole or homonuclear coupling Hamiltonians.10 In particular, the
refocusing pulse is a 90° pulse; a 180° pulse has no effect on the evolution
under bilinear Hamiltonians, see section 4.7.3.
4.5 THE DENSITY MATRIX FOR COUPLED SPINS
This may seem a great deal of effort to describe the basic NMR experiment,
and, indeed, the density operator treatment for a single spin-
missing image file is equivalent to the much simpler description of the
NMR experiment provided by the vector model of NMR. The density
operator approach comes into its own, however, for treating systems with
more than two levels.
4.5.1 EXAMPLE 1: A DIPOLAR-COUPLED HOMONUCLEAR SPIN

PAIR
It was noted without detailed explanation in section 2.5 that the splittings
due to homonuclear couplings were 50% larger than corresponding
heteronuclear couplings. This first example shows how this factor arises.
The matrix elements of equation 4.25 first need to be evaluated in the

“product” eigenbasis for a pair of spin - missing image file nuclei I and S
, which consists of the four different combinations of α and β states for two
spins: figure missing The first quantum state in the ket refers to I and the
second to S .11
The matrix representation for terms such as figure missing is simply the
product of the matrices12 for Î + and figure missing Alternatively, the
matrix elements can be determined directly by considering the effect of the
individual operators on the appropriate spins. For example,
figure missing involves applying the lowering operator to S in the α state
and the raising operator to the I spin in the β state
The evaluation order is not significant since operators involving different

spins must, by definition, commute figure missing Similarly
Working through the combinations gives the matrix representation of the

dipolar coupling Hamiltonian:

where d = D IS P 2 (cos θ ) is used as shorthand for the dipolar coupling
constant between I and S for a particular orientation θ . In the heteronuclear
case, where the off-diagonal terms can be dropped, the NMR frequencies
can be directly read off as ±d by inspection of the Hamiltonian matrix. This
corresponds to a splitting of 2d in the spectrum.
In the homonuclear case, we need to diagonalize equation 4.78, and the

NMR frequencies are less obvious. To determine the NMR spectrum, we
also need to evaluate the sum angular momentum operators
figure missing etc. These sum operators are conventionally denoted
missing image file etc., and they are readily derived using equations 4.2–
4.7 (but dropping the factors of figure missing ). The non-zero matrix
elements are:
H ence the matrices are
and so missing image file 4.83
Note the block structure of all the matrices, where the blocks correspond to
the different values of m I + m S , that is, the eigenvalues of the sum z
operator, missing image file The Hamiltonian matrix of equation 4.78, for
example, has non-zero elements in a 1 × 1 block corresponding to
missing image file = 1, a 2 × 2 block corresponding to the two states with
missing image file and missing image file and a final 1 × 1 block
corresponding to missing image file This block structure also applies to
the density matrix, with a block linking states with missing image file and
missing image file corresponding to coherences of order p = k – l . Hence
the system of two spin - missing image file nuclei, as previously
illustrated in figure 4.1 (c), can support coherences of order 0 (the
populations of the different eigenstates plus zero-quantum coherences
between, e.g., missing image file and missing image file detectable
NMR coherences of order ±1, plus multiple-quantum coherences of order
±2 between the αα and ββ states.
Rather than model the complete NMR experiment as in the previous

section, the NMR frequencies and amplitudes can be calculated directly. If
we start with x magnetization and acquire data points at integer multiples, n
, of the dwell time, Δt , then the NMR signal will be
where Û is the propagator over the dwell time, Û ≡ Û (Δ t , 0). Since the
propagator for each dwell time is the same, we can write Û ( n Δ t , 0) = Û
(Δ t , 0) n . Equation 4.84 is most easily evaluated in the eigenbasis of the
Hamiltonian. The matrix of equation 4.78 is readily diagonalized thanks to
its block structure—only the central 2 × 2 block will have nontrivial
eigenvalues and eigenvectors:
Transforming missing image file and missing image file into the new
eigenbasis gives
Since Λ is diagonal, the propagator, equation 4.52, is easily evaluated in the

Hamiltonian eigenbasis:
Expanding out the matrix products of equation 4.84 gives (another exercise
for the reader):
If we create a new matrix with elements missing image file we can write

that is, the NMR signal is the sum of oscillations with frequencies Λ jj – Λ
kk (which correspond to differences of the eigenvalues of the Hamiltonian)
and amplitudes given by the matrix elements A jk .
Evaluating the “amplitude matrix” gives
The two non-zero elements of A correspond to two allowed transitions of

unit intensity, giving
Hence a pair of lines with a frequency difference of 3d will be seen in the

spectrum. As predicted, this splitting is 50% larger than the splitting of 2d
observed in the heteronuclear case.
4.5.2 EXAMPLE 2: CROSS-POLARIZATION
Although cross-polarization (CP) generally involves transfer of

magnetization from an abundant spin species, such as 1 H, the basic
principles can be seen in a simple two spin picture. This example also
illustrates the use of different frames of reference to simplify analysis.
The Hamiltonian for a heteronuclear spin pair, subject to RF irradiation

along the x axis on both spins, is
where ν I and ν S are the nutation frequencies due to the RF on the I and S
spins respectively. For simplicity we assume that the RF is on-resonance
for both spins. The Hamiltonian matrix in the product Zeeman basis is
Note how the ν I elements connect states in which the first (I ) spin flips,
while the ν S elements correspond to allowed NMR transitions of the S
spins.
Despite its elegance, the eigenvalues and vectors of this matrix are not
obvious, and there are large terms (those involving the RF) that are off the
diagonal. The situation is improved using a basis that puts the dominant
terms along the diagonal by rotating the axis system so that the new z axis
is along the RF spin-lock axis. The required 90° rotation about y simply
permutes the labels on the operators (x → z → − x → − z → x ) to give
The significance of matching the nutation rates of the two nuclei, ν HH = ν

I = ν S , is then easier to appreciate (HH denotes the Hartmann–Hahn
matching condition). At this condition
Since the RF field strength must significantly exceed the dipolar coupling
for a good spin-lock, figure missing the secular approximation can be
invoked to drop the off-diagonal terms connecting different diagonal
elements, leading to
The off-diagonal terms linking the approximately degenerate middle two

states cannot be neglected, and so the dipolar coupling will “mix” these
states.
Considering the matrix representations for I and S magnetization along the

spin-lock axes (i.e., z )
the mixing of the central states can be seen to interconvert figure missing
and figure missing It is left as an exercise to show, using the approach set
out in section 4.5.1, that the evolution of the density operator in the
conventional basis will be
In other words, the magnetization will oscillate between I and S spin

magnetization, with the frequency of oscillation set by the strength of the
dipolar coupling.13
It is also useful to consider this problem in terms of spin operators. Starting

from equation 4.95, the Hamiltonian in the tilted frame at the Hartmann–
Hahn condition can be written
where missing image file is the sum magnetization operator. The second
term is a flip-flop term interchanging states figure missing and
figure missing while the third interchanges figure missing and
figure missing . If the spin-lock is strong, that is, figure missing then
the secular approximation can be invoked to discard terms that do not
commute with ν HH missing image file The Hamiltonian then reduces to
The initial density operator (in the tilted frame) is missing image file This
commutes with the RF Hamiltonian (i.e., the magnetization is spin-locked).
Hence the density operator will evolve simply under the influence of d
missing image file . As discussed above, this flip-flop term has the effect
of interconverting I and S spin magnetization.
The two formulations are exactly equivalent; discarding non-commuting

spin operator terms is equivalent to neglecting off-diagonal terms
connecting non-degenerate eigenstates in the matrix representation.
Operator representations often have the advantage of generality (they are
not specific to a spin- missing image file pair), but the necessary operator
algebra can be distracting. The matrix representation is not strictly general
but can be a more visual way to tackle nontrivial problems.
4.6 EULER ANGLES & SPHERICAL TENSORS

So far, we have mostly treated the NMR parameters, such as dipolar
couplings, as simple empirical constants. In practice, the NMR interactions
are orientation dependent and we need to consider their full tensor nature.
As set out in section 2.2, it is relatively straightforward to write down
expressions for NMR parameters, such as the dipolar coupling, in terms of
the polar angles, θ and ø , defining the orientation of the principal axis
system for the tensor in the laboratory frame defined by the external
magnetic field.
In more complex problems, however, it is necessary to introduce additional

intermediate frames. Considering a collection of nuclear spins in a
crystalline framework, we first need to relate the principal axis systems of
the individual interactions (dipolar, CSA, etc.) to a common molecular
frame (M) (typically defined in terms of the crystallographic lattice). This
molecular frame must then be related to the laboratory frame (L) by a set of
angles that specify the orientation of the crystal.
In general three angles are needed to define a transformation in three-

dimensional space. These three Euler angles are conventionally denoted α ,
β , and γ , while the set of angles specifying a transformation, (α , β , γ ), is
often collectively indicated by figure missing For instance, the
transformation14 from the molecular frame to the laboratory frame may be
represented by figure missing Inset 4.4 explains how these angles are
commonly defined. In some cases, specification of all three Euler angles is
unnecessary. For example, the α angle is irrelevant when transforming a
dipolar coupling from its principal axis system to another frame since the
dipolar interaction is symmetric about its Z principal axis (the internuclear
vector) in the absence of molecular motion. On the other hand, α must be
specified if the interaction has a non-zero asymmetry, η . Similarly, the γ
angle is unimportant when transforming into the final laboratory frame,
since the nuclear spin Hamiltonian is unaffected by rotations around the
magnetic field axis (this follows from the high-field approximation
discussed in section 4.2).
As discussed in inset 4.4, the simple 3 × 3 matrix representation of rank-2

tensors is not well adapted to expressing the effect of rotations and usually
leads to extremely unwieldy expressions when a series of frame
transformations is required. The remainder of this section discusses an
alternative representation of tensors that better expresses symmetry under
rotation and leads to more elegant and compact descriptions of anisotropic
NMR interactions.
Inset 4.4. Euler angles and tensor rotation

missing image file
Figure 4.4. Definition of the Euler angles (Rose convention) * for the
transformation of the original axes (x , y , z ) to a new axis system (x ′ , y ′ ,
z ′ ). The axes are first rotated by α about the z axis to bring the y axis onto
the nodal line defined by xy and x ′ y ′ planes. This is followed by rotation
about the new y axis through β and then a final rotation through γ is applied
about the final z axis (z ′ ).
Figure 4.4 shows the most common definition of the three Euler angles, α ,
β, and γ , specifying the rotation of one coordinate system to another.
Unfortunately, the definition of the angles is not unique and different
systems have been used in the literature. Care is needed if reporting Euler
angles or when compiling orientational information from different sources.
If we consider a typical rank-2 tensor, such as a shielding tensor, in its

conventional matrix (Cartesian) form:
The tensor in the new basis is given by A ' = RAR -1 where the rotation
matrix R is
The final matrix A' is not shown as it is rather cumbersome and not
particularly informative. The main text discusses an alternative
representation of tensors that is better adapted to frame transformations.
* Elementary theory of angular momentum, M. E. Rose, Wiley (1957).

In this spherical tensor representation, the tensor is decomposed using a
basis with the same properties under rotation as the spherical harmonics
familiar from atomic orbital or angular momentum theory. A rank-2 tensor
decomposes into an isotropic (rank-0) term, R 0 , 0 (compare with an s
orbital), three rank-1 terms, R 1 , 0 , R 1 , ±1 (compare with p orbitals), and
five rank-2 terms, R 2 , 0 , R 2,±1 , R 2,±2 (compare with d orbitals). The
rank-1 components correspond to the antisymmetric component of the
tensor, that is, terms involving R xy − R yx , etc. These have no observed
effect on NMR spectra and can be safely ignored.
The relationship between the spherical tensor components and the

Cartesian representation is simple for a tensor in its principal axis system,
that is, with the matrix just containing the diagonal elements:
where R iso is the isotropic component, and ζ and η are the anisotropy and
asymmetry respectively (see inset 2.2). Note how the spherical tensor
representation captures the rotational symmetry: the rank-0 isotropic term
is independent of orientation, while the anisotropy is contained within the
rank-2 terms.
At high field, the only components that survive the secular approximation
are the entity image missing and entity image missing components of
the tensor in the laboratory frame. The isotropic term is, by definition,
independent of the axis system, and so the problem of finding the high-field
Hamiltonian involves determining entity image missing after applying a
series of transformations to a tensor starting in its principal axis system.
The effect of rotation through Euler angles figure missing = (α , β , γ ) on

the rank-2 components is given by
where missing image file are elements of the rank-2 Wigner rotation
matrix . These are given by
where the “reduced” matrix elements missing image file (β ) are solely
functions of the β angle.15
4.6.1 MAGIC-ANGLE SPINNING
As an example, consider a sample under magic-angle spinning. As

illustrated in figure 4.5 , the various NMR interactions need to be expressed
in a rotor frame (R), which is aligned along the rotation axis and rotates at
the spinning frequency. The transformation from this frame to the final
laboratory frame is given by the Euler angles
missing image file
Figure 4.5 . Hierarchy of frames of reference for sample spinning problems.

Individual principal axis systems (P1 and P2 for the two dipolar
interactions indicated by solid arrows) are related to a common molecular
or crystal reference frame (M). The Euler angle set figure missing
describes the transformation of the molecular frame to a frame fixed with
respect to the rotor (R). Finally the rotor frame is related to the laboratory
frame (L) via the Euler angle set figure missing (see equation 4.109).
where the rotor position, α , is time dependent ( α 0 and ω r are the initial
rotor angle and spinning frequency respectively). The β angle corresponds
to the angle of the rotor with respect to the magnetic field, that is, the magic
angle θ m . The γ angle can be arbitrarily set to zero, as discussed above.
By combining equations 4.108 and 4.109, missing image file is given by
If there is a single time-dependent interaction (or the different terms of the

Hamiltonian commute with each other), then the overall Hamiltonian is
said to be “inhomogeneous” and we can consider the Hamiltonians for the
individual terms independently; this important distinction between
“inhomogeneous” and “homogeneous” Hamiltonians is discussed in inset
4.5. Average Hamiltonian t heory (see section 4.7.1) can then be used to
find the average Hamiltonian over a complete rotor period. This will
involve the average of missing image file
missing image file ( β ) is the familiar second-order Legendre

polynomial, (3cos 2 β – 1)/2. Hence at the magic angle,
missing image file that is, the average Hamiltonian vanishes and there is
no evolution of the density matrix.
Unless the spinning rate is much larger than the interaction (in which case
the oscillatory terms corresponding to n = ±1 and n = ±2 are negligibly
small), the n = ±1 and n = ±2 in equation 4.110 will lead to some
oscillatory evolution of the density matrix. In terms of the NMR signal, this
corresponds to the formation of rotary echoes , in which the signal evolves
during the rotor period, but is “refocused” to create a rotary echo at each
complete rotor period (as required by equation 4.112). This series of
echoes, with a period equal to the rotor period, corresponds to a set of sharp
peaks separated by the spinning frequency in the spectrum, that is, a
manifold of spinning sidebands. See figure 3.15 for a practical example of
rotary echoes.
4.7 ADDITIONAL ANALYTICAL TOOLS
The approaches outlined above of following the evolution of the density

operator using a combination of matrix or symbolic representations of the
spin operators are effective for understanding many solid-state NMR
problems. There are, however, a large class of experiments where these
simple approaches become intractable, and it is necessary either to resort to
numerical simulation (which is discussed in appendix D) or to use more
sophisticated analytical tools. This section reviews some of these more
advanced theoretical tools.
4.7.1 AVERAGE HAMILTONIAN THEORY

Average Hamiltonian Theory (AHT) has been widely used in solid-state
NMR to derive exact analytical expressions, but also more qualitatively
when the problem is too complex to be reduced to a tractable expression.
AHT is useful whenever the Hamiltonian is time-dependent and periodic,
that is, the time-dependent component repeats itself after a characteristic
“cycle time”, τ c . This may correspond to a repeated RF sequence or
sample rotation.
AHT applied to sample rotation has already been discussed in section 4.6.
In this simple case, it was sufficient to average over the time dependence to
obtain the average Hamiltonian to first order. By contrast, when dealing
with strong time-dependent RF, it is generally necessary to transform the
Hamiltonian into an interaction frame that allows the average Hamiltonian
to be readily computed. The total Hamiltonian is first separated into a term,
figure missing , containing the NMR interactions, etc., and a time-
dependent component, figure missing for the RF:
Inset 4.5. Homogeneous and inhomogeneous Hamiltonians

As previously discussed in section 2.7, there is an important distinction
between “inhomo-geneous” interactions, such as the chemical shift
anisotropy, which lead to sharp spinning sidebands under magic-angle
spinning, and “homogeneous” interactions, notably homonuclear dipolar
coupling, that result in centerbands and sidebands with distinct width. The
origin of this distinction involves some subtle quantum mechanics. The
spin Hamiltonians for shift interactions and heteronuclear couplings at high
field only involve z spin operators, for example, figure missing for the
heteronuclear dipolar coupling between spins I and S . Any pair of
Hamiltonians involving only z operators must commute, since [ Î jz , Î kz ]
= 0 for all values of j and k . Hence we can consider their effects on the
evolution of the density operator independently. The MAS spectrum
associated with an individual interaction consists of sharp centerbands and
sidebands, and so the overall spectrum will also consist of sharp features
(mathematically, the spectrum is a “convolution” of the spectra associated
with the interactions taken individually).
By contrast, if the Hamiltonian involves homonuclear couplings, the flip-
flop terms from different spin pairs do not generally commute with each
other
As a consequence, the evolution under the total Hamiltonian cannot be

simply expressed in terms of the individual components. Moreover, the
evolution is not refocused over a rotor cycle and the centerbands and
sidebands have a distinct width.
The presence of a non-commuting term in the Hamiltonian tends to

“contaminate” the rest of the Hamiltonian, giving it the same
“homogeneous” character. For instance, linewidths in 13 C NMR are often
limited by the efficiency of the 1 H heteronuclear decoupling even though
the 13 C interactions (at natural isotopic abundance) only involve
inhomogeneous terms such as the chemical shift. Unless the 13 C nucleus
is very efficiently decoupled from the 1 H dipolar network, its NMR lines
are affected by homogeneous broadening.
The expression for propagators for the interval from t = 0 to t is then:
4.115
where missing image file is the Dyson time-ordering operator —

essentially a reminder that we need to evaluate equation 4.115 in small,
time-ordered steps to account for the time dependence of figure missing .
If we transform figure missing into a frame in which time dependence
due to the RF has been removed, we can factor the overall propagator into
two components:
where missing image file 4.117
is the overall evolution under the RF and

is the evolution under the system Hamiltonian in the interaction frame

(denoted by ~), where
This may seem obscure, but the rotating frame of reference provides a good
example of an interaction frame; U RF (t ) in this case is exp(iω NMR tÎ z
), representing rotation about the z axis at the Larmor frequency, ωNMR ,
and missing image file is now the familiar rotating frame Hamiltonian.
Fortunately some simplifications can now be made! First we restrict

attention to sequences that are “cyclic”, that is, sequences for which there is
no net rotation of the axis system over the course of the cycle. This means
that U RF (τ c ) is the identity matrix, and the propagator over a complete
cycle is simply figure missing Hence we can describe the effect of the RF
pulse sequence over a complete cycle in terms of an “average
Hamiltonian,” missing image file which creates this propagator,
missing image file This still leaves the problem of expressing
missing image file in terms of missing image file
To first order, however, missing image file is given by the average of

missing image file over the cycle period:
The additional terms in this series expansion (termed the Magnus

expansion ) are increasingly complex. However, the first (and successive
odd-order) correction terms, missing image file etc., evaluate to zero if
the interaction-frame Hamiltonian has time symmetry missing image file
Moreover, the correction terms should be small if the interaction frame is
appropriate.
The WHH-4 sequence for homonuclear decoupling provides a good

illustration of AHT in action (see section 5.5.1 for discussion of
homonuclear decoupling sequences). As shown in figure 4.6 , the sequence
consists of a series of 90° pulses separated by gaps which can be used to
detect the NMR signal. The interaction frame during the different periods
of the sequence is obtained by applying these 90° rotations to an initial
starting set of axes. Note how the frame returns to its starting point at the
end of the cycle, satisfying the “cyclicity” condition.
missing image file
Figure 4.6. The WHH-4 (WAHUHA) pulse sequence for homonuclear

decoupling: (top) the pulse sequence and (bottom) the axes of the
interaction frame during the different sections of the sequence. Note that x ,
y , z refer to the laboratory frame, while X , Y , Z refer to the interaction
frame. Figure adapted from Principles of magnetic resonance in one and
two dimensions (see Further reading).
Using missing image file to denote the interaction-frame Hamiltonian for

the homonuclear dipolar coupling between a spin pair at the start of the
cycle,
we can write down the Hamiltonians for the successive steps of the cycle.
The transformation into the interaction frame for the period between the
first two pulses corresponds to relabeling the axes using y → z , z → –y ,
and so the interaction-frame dipolar Hamiltonian during this period is
Similarly the Hamiltonian in the middle segment is
The average Hamiltonian over the entire period is (to first order):

that is, the dipolar interaction is suppressed to first order. It is important to

note that the average Hamiltonian is only relevant to the density matrix
sampling at integer multiples of the cycle period, τ c . For example, the
average Hamiltonian of equation 4.128 implies that the density matrix
appears not to evolve when sampled after each complete WHH-4 cycle,
even though the system will evolve under the influence of both RF and
dipolar couplings in between sampling points.
Since the expression for missing image file in equation 4.121 is linear in
missing image file we can compute the first-order average Hamiltonian
for different interactions independently, that is, missing image file Hence
equation 4.128 will be true for an arbitrary collection of homonuclear
coupled spins. This does not apply to higher-order contributions to the
average Hamiltonian, and calculating these higher-order “cross-terms”
between different components of the Hamiltonian is one of the major
challenges to applying average Hamiltonian approaches beyond first order.
4.7.2 AN OVERVIEW OF FLOQUET THEORY
AHT often provides an intuitive summary of complex time-dependent

problems. It suffers, however, from a number of restrictions. For example,
the evolution is only sampled at complete periods of the time dependence.
Just as sampling the NMR signal at too low a rate means that different
frequency components may not be distinguished (aliasing ), so different
components of the density operator evolution may be aliased. In the case of
magic-angle spinning, for example, the sampling only occurs once per rotor
period and so spinning sidebands and centerbands are not distinguished.
Hence information about the spinning sideband intensities, or differences in
lineshape between centerbands and sidebands, cannot be obtained.
Floquet theory is a widely used approach that generalizes AHT. It is

restricted to considering problems where the Hamiltonian can be expressed
as a Fourier series:
where T is the period (e.g., the rotation period in the case of MAS, or the
sequence period in the case of periodic RF) and m is the index of the
Fourier series. However, since most functions can be expanded as a
convergent Fourier series, this is not a major restriction. Moreover, more
than one time dependence (e.g., magic-angle spinning plus time-dependent
RF) can be accommodated by expanding with respect to more than one
Fourier index.
The density operator, NMR signal, etc., can then all be expressed in terms
of Fourier series; for example, the NMR signal for MAS problems becomes
where ν j is the centerband frequency for transition j and A m,j is the

intensity of sideband m for transition j .
Without going into the mathematical details, Floquet theory provides the
machinery to calculate the A m,j coefficients, starting from the Fourier
series for the Hamiltonian (plus the initial density matrix). Note how
Floquet theory imposes no restriction on the sampling of the NMR signal
(equation 4.130 is valid for all t ) and naturally distinguishes centerbands
and sidebands.
4.7.3 INTRODUCTION TO IRREDUCIBLE SPHERICAL TENSOR

OPERATORS
As part of deriving the interaction frame for the WHH-4 sequence above,
the effect of the RF pulses was described in terms of “rotations” of the spin
operator components of the Hamiltonian. While the 90° rotations of WHH-
4 can be readily expressed in terms of Cartesian axes, working with
arbitrary rotations in this framework becomes cumbersome. For this reason
most work involving average Hamiltonians for RF pulse sequences
describes the spin Hamiltonian in terms of a spherical tensor basis, which is
analogous to the spherical tensor representation of rotations in physical
space. This representation will be only briefly introduced here—see the
more advanced texts in Further reading for more information.
In terms of irreducible spherical tensors, the homonuclear dipolar coupling
Hamiltonian of equation 4.25 is
where A 2 , 0 = missing image file is the spatial component and
is the spin operator term. 16
The principles described in section 4.6 for rotations of spherical tensors in

physical space apply equally well to rotations of the spin operators by RF
pulses. For example, a π pulse corresponds to a rotation using the Euler
angles (0, π , 0). From equations 4.107 and 4.108, the rotation of the spin
operator missing image file by a π pulse will generate
The only non-zero missing image file hence the spin operator after
rotation by π is missing image file In other words, the homonuclear
dipolar Hamiltonian is unchanged by a 180° rotation. The relative ease of
expressing rotations in terms of irreducible spherical tensors is a major
incentive for their use when developing new RF irradiation schemes.
4.7.3.1 Application to Decoupling and Recoupling Sequences
One of the most common applications of AHT is for sequences designed

either to suppress homonuclear dipolar couplings or to “recouple” the
homonuclear couplings that are suppressed by magic-angle spinning.
We can always express the Hamiltonian of the spin system as a sum of

terms that are either rank 1 (linear) or rank 2 (bilinear) in spin operators
that are relevant to the RF being applied:
where missing image file and λ and λ ′ sum over the rank-1 and rank-2
terms respectively. Heteronuclear coupling terms are included in the rank-1
sum, since they contain only operators of the form Î z with respect to the
spins being manipulated by the RF.
The average Hamiltonian associated with an RF pulse sequence will also be

a sum of rank-1 and rank-2 terms. The original spin operators 17 are
replaced by new average terms:
where missing image file is a linear combination of the

missing image file operators, and α i is a scaling factor. Note how the
spin operators contained in the average Hamiltonian are not generally the
same as the original operators, missing image file and
missing image file
For recoupling , the goal is to design RF sequences that maximize the

scaling factor on the homonuclear couplings, that is, α 2 . Particularly in
quantitative applications it is also desirable to minimize the scaling factor
on the rank-1 terms, α 1 , in order to create a Hamiltonian that depends only
on the homonuclear couplings.
By contrast, homonuclear decoupling sequences should minimize α 2 . In

most applications, it is also important to retain the evolution under rank-1
terms (i.e., the chemical shift), and so it is desirable to maximize such rank-
1 terms. The spin operator terms of the rank-1 average Hamiltonians will
involve
where missing image file has been expressed in terms of the conventional
Cartesian spin operators, with missing image file This decomposition
effectively defines a new axis system. If the missing image file operator
in this frame is defined by
then the I -spin magnetization will precess about this z ′ axis. As discussed
further in section 5.5, z ′ does not usually coincide with the normal rotating
frame z axis, resulting in tilted axis precession . The scaling factor, α 1 ,
scales the rate of precession about this axis, and a small scaling factor will
result in poor spectral resolution. Hence the importance of maximizing α 1
while minimizing α 2 .
APPENDICES
Two of the appendices of the book develop themes that are relevant to this
chapter, but which are not required in later chapters. Appendix C introduces
the topic of Liouville space and how this is applied to problems involving
relaxation and site exchange. These are rather specialist topics and
understanding the theory behind them is not necessary in most applications
of solid-state NMR. Appendix D gives an introduction to the numerical
simulation of solid-state NMR spectra and experiments. Simulation is a
useful way of exploring the ideas developed in this chapter and is often
necessary in the development of new techniques and analysis of complex
problems in solid-state NMR. The software used for such simulations is,
however, subject to change and so the practical description of setting up
NMR simulations is confined to a distinct appendix.
FURTHER READING
INTRODUCTORY TEXTS: GENERAL NMR
“Understanding NMR spectroscopy” , J. Keeler, John Wiley & Sons Ltd.

(2005), ISBN 978 0 470 01786 9.
“Spin dynamics ”, M.H. Levitt, John Wiley & Sons Ltd. (2008), ISBN 978
0 470 51117 6.
MORE ADVANCED TEXTS: GENERAL NMR
“Principles of nuclear magnetic resonance in one and two dimensions ”,

R.R. Ernst, G. Bodenhausen & A. Wokaun, Oxford University Press
(1990), ISBN 0 19 855647 0.
““Tensors and Rotations in NMR” , L. J. Mueller, Concepts in Magnetic

Resonance Part A, 35A (2011) 221.
MORE ADVANCED TEXTS: SOLID-STATE NMR
“Introduction to solid-state NMR spectroscopy ”, M.J. Duer, Blackwell

Publishing Ltd. (2004), ISBN 1 4051 0914 9.
“Multidimensional solid-state NMR and polymers ”, K. Schmidt-Rohr &

H.W. Spiess, Academic Press (1994), ISBN 978 0 126 26630 6.
NOTES
1 Note that the nuclear spin Hamiltonian is typically expressed in terms of

frequency units (Hz, MHz, etc.), which are the natural units for NMR
interactions. A slightly different convention, common in theoretical work,
uses angular frequencies (rad s − 1 ). The two conventions are usually
distinguished by using ν for frequencies and ω for angular frequencies.
2 The NMR shielding Hamiltonian, equation 4.11, also has this form, but
with the external magnetic field instead of the second spin.
3 D and equation 4.23 can be formulated in different ways, but these will
expand to the same overall expression for the dipole–dipole coupling
Hamiltonian.
4 The frequency of electromagnetic radiation used to observe NMR, ν RF ,

and nutation frequency of the nuclear spin magnetization resulting from the
resonant absorption, ν 1 , must be distinguished carefully. Various notations
are used in the literature, for example, ν RF often refers to a nutation
frequency.
5 The Bloch–Siegert effect is a shift in the effective NMR frequency due to

this neglected component. Its size is inversely proportional to the NMR
frequency and so is negligible for typical high-field NMR. Analogous
effects can be seen at high field if irradiating one nucleus while observing
another with a very similar Larmor frequency. For example, shifts in 19 F
resonance frequencies due to 1 H decoupling can be readily observed since
19 F and 1 H NMR frequencies differ only by about 12%.
6 Strictly speaking this is the reduced density matrix. See later for the
distinction.
7 The factor of h follows from expressing the Hamiltonian in terms of

frequency rather than energy.
8 See Further reading for proof that missing image file
9 This is a somewhat handwaving treatment. The relationship between the

quantum mechanical evolution and the observed NMR signal is extremely
subtle. A more thorough treatment can be found in Levitt’s Spin Dynamics
(Further reading).
10 The analytical derivation of the echo response is more involved than for
the simple spin echo and can be found in the Further reading. In contrast to
the simple spin echo, the solid echo cannot be pictured in simple vector
model terms. Moreover, the phase relationship of the two pulses is also
significant. While the relative phases of excitation and refocusing pulses
only affect the phase of the simple spin echo (the direction in xy along
which the magnetization is refocused), the phase difference must be 90° to
obtain a full solid echo.
11 The ordering is not significant, but obviously must be used consistently!
12 A more formal way to construct such matrices is via the “direct” or

“Kronecker” product between the 2 × 2 matrix representations for the
individual spins. The direct product (usual symbol ⊗ ) should not be
confused with simple matrix multiplication.
13 In cases where the IS spin pair is sufficiently well isolated, it is possible

to observe such “transient oscillations” during Hartmann–Hahn cross-
polarization experimentally. As discussed in section 8.1.5, this often
provides a simple approach to measuring dipolar couplings.
14 Subtle issues such as active vs . passive rotations are being glossed over
here. See, for example, Levitt’s Spin Dynamics and Mueller’s Tensors and
Rotations in NMR (Further reading) for details.
15 Table B.2 of Schmidt-Rohr and Spiess in the Further reading contains
tables of Wigner functions.
16 Equation 4.131 is equivalent to equation 4.25, but care is needed to

avoid mixing the two representations as the normalization factors (here 1 /
missing image file ) are different. Here we have used A to denote a
spatial tensor associated with a spherical tensor representation of the spin
operators to contrast with R used in section 4.6, where the spatial tensors
were associated with a Cartesian representation of the spin operators.
17 The focus here is on the spin operators. Obviously the time dependence
of the spatial terms must also be included when determining average
Hamiltonians if MAS is involved.
CHAPTER 5
GOING FURTHER WITH SPIN -

SOLID -STATE NMR
5.1 INTRODUCTION
The basic principles of solid-state NMR for dilute spin - nuclei, such as 13
C and 15 N, were set out in chapter 3 . This chapter first considers how the
basic techniques of Hartmann–Hahn cross- polarization (CP) and high-
power proton decoupling need to be modified under the conditions of high
magnetic fields and fast magic-angle spinning (MAS) rates which are now
available and which are increasingly important for the study of complex
systems, such as microcrystalline proteins.
Although straightforward one-dimensional NMR is usually the most
efficient route to solving chemical problems via solid-state NMR, there are
inevitably occasions where more “sophisticated” techniques, such as 2D
NMR, provide vital information that cannot be obtained from simple 1D
spectra. This chapter discusses more complex experiments involving
multidimensional NMR and/or homonuclear decoupling techniques (used to
improve 1 H resolution). The focus here is on experiments that provide
answers to essentially qualitative questions, such as “Is site A close to site
B?” The more demanding quantitative experiments, for example “What is
the distance between site A and B?”, are discussed separately in chapter 8 .
5.1.1 SPIN- NMR AT HIGH MAGNETIC FIELDS
To the solution-state NMR spectroscopist, a specific section on spin -

NMR at high magnetic fields may seem unnecessary, since the effects of
increased magnetic fields on spectral resolution and sensitivity are relatively
straightforward. However, the question of how linewidths and resolution
(and in consequence sensitivity) vary with increasing magnetic field is
actually quite subtle in solid -state NMR. Linewidths in spin - NMR are
discussed in some detail in section 5.2, but some of the key issues are
illustrated in Figure 5.1 , which shows the 13 C spectrum of a solid form of
the steroid drug finasteride. The linewidths vary considerably across the
spectrum, with some peaks being so sharp that they exhibit truncation
artifacts (see section 3.3.2), while the methylene (CH 2 ) signals in particular
are much broader. These differences can be explained by the local strength
of the 13 C, 1 H dipolar couplings. Decoupling 1 H from 13 C is relatively
easy when these local fields are weak, for example for quaternary sites and
carbons affected by dynamics, such as methyl groups. By contrast,
methylene resonances are particularly difficult to decouple effectively due to
the strong homonuclear coupling between the geminal protons, leading to
noticeable broadening. Hence a trade-off needs to be made between
extending the acquisition (to limit truncation of the narrow lines) and
increasing the strength of the decoupling RF field (to narrow the
resonances).
Figure 5.1. Low-frequency section of the 13 C spectrum of a solvate form of finasteride with 1,4-
dioxane illustrating the interaction between decoupling and widths of selected lines. Two-pulse phase-
modulated decoupling (see section 5.1.2) was used during the 30 ms acquisition time (RF nutation
rate: 57 kHz). The meth yl, quaternary, and dioxane signals are slightly truncated, leading to “sinc-
wiggle” artifacts (see sectio n 3.3.2), while the CH2 resonances are markedly broader than the CH
resonances. The resonance of the dioxane is sharp due to the high mobility of the dioxane molecules.
(MAS rate: 8.5 kHz. 13 C NMR frequency: 125.7 MHz .)
Provided the efficiency of decoupling can be maintained, working at

higher magnetic fields is expected to improve both resolution and sensitivity
for dilute spins such as 13 C. However, obtaining the expected improvements
is not as straightforward as it might seem. The scaling of chemical shift
anisotropies (CSAs; in frequency units) with the NMR frequency means that
the spectra at higher magnetic fields (say over 9.4 T) can become cluttered
with excessive spinning sidebands and centerband intensities reduced.1
Hence MAS rates need to be increased. This is not itself a problem, since the
improvements in sensitivity from working at higher magnetic fields allow
smaller, faster-spinning rotors to be used. However, as MAS rates exceed ~8
kHz, the implicit assumption that the spinning rate is much smaller than the
short-range dipolar couplings (homonuclear and heteronuclear) begins to
fail, and key components of the CPMAS experiment, Hartmann–Hahn CP,
and continuous-wave (CW) decoupling, become inefficient. The following
sections discuss how more sophisticated decoupling and CP schemes allow
the full potential of higher magnetic fields to be exploited.
5.1.2 ADVANCED HETERONUCLEAR DECOUPLING
If the effects of MAS can be ignored, 1 H decoupling is straightforward in

solids. In hand-waving terms, the strong network of dipolar couplings
between the spins means that it is sufficient to apply a continuous pulse of
RF irradiation to the middle of the 1 H spectrum in order to obtain a
decoupled spectrum of the dilute spins. The width of the 1 H static spectrum
(typically ~50 kHz for rigid organic solids), which is due to these same
dipolar couplings, means that relatively high RF nutation rates are required
in order to obtain useful resolution. However, provided sufficient RF power
is used, simple CW irradiation is generally very effective. This contrasts
with the situation in solution-state NMR, where the interactions to be
decoupled (i.e., J CH couplings) are much weaker. This allows much lower
power RF irradiation to be used, but means that the sequences must be
effective over the entire width of the 1 H spectrum despite the relatively low
RF nutation rates. Hence solution-state NMR involves sophisticated phase
and/or amplitude modulations to achieve so-called broadband decoupling.
As the MAS rate increases, however, the 1 H spectrum begins to break
up into spinning sidebands (see section 2.7). Under these conditions, simple
CW decoupling starts to lose its efficiency. Since linewidths in dilute-spin
NMR are often limited by decoupling efficiency (see section 5.2), this means
that lines begin to broaden significantly as the spinning rate increases. The
difficulty of maintaining decoupling efficiency at faster spinning rates
means, somewhat counterintuitively, that increasing MAS rates tend to
degrade rather than improve resolution for dilute spins!
A full understanding of 1 H decoupling in solids remains a formidable
challenge due to the number of interactions involved; homo- and hetero-
nuclear dipolar couplings, CSAs and isotropic shifts all influence decoupling
efficiency. Fortunately a number of solutions have been found that largely
offset the unwelcome effect of MAS on CW decoupling. The simplest of
these is the TPPM (two-pulse phase modulation ) scheme, which involves a
continuous train of 1 H pulses of constant duration but whose phase switches
between two values (see Figure 5.4 ). Empirically it is found that the optimal
tip angle of the pulses is generally close to ~170° while the phase
modulation, ø , is typically of the order of 10–30° . The exact optimal
conditions depend on experimental parameters such as MAS and RF
nutation rates, but are transferable between similar samples, that is, the
decoupling can be optimized on a setup sample and then applied to other
samples under the same conditions .
These effects are illustrated in Figure 5.2 . Increasing the strength of the
decoupling, by increasing the 1 H nutation rate, systematically decreases the
13 C linewidths and improves spectral resolution. Replacing CW decoupling
with TPPM also leads to significant resolution improvements, and the

combination of TPPM decoupling with the higher RF power leads to a much
improved spectrum (Figure 5.2 (d)). As noted above, methylene (CH 2 )
resonances, such as the one marked with an arrow, are typically very
sensitive to the efficiency of the decoupling; the distinctly triangular
lineshape of the CH 2 resonances in the spectra of Figure 5.2 (a–c) is a
typical sign of inadequate decoupling.
Other decoupling schemes that have proved popular for solids are
SPINAL64 (a more complex modulation, but one that often outperforms
TPPM and is simple to optimize) and XiX decoupling (which again has a
single optimization parameter and often outperforms other decoupling
schemes at spin rates above ~30 kHz). However, the development of better
heteronuclear decoupling schemes for solids is an area of active research and
so “best practice”, especially at fast spin rates, can be expected to evolve in
the medium term. See Further reading for more information on these
sequences and heteronuclear decoupling in general.
Figure 5.2. Illustration of the effects of 1 H decoupling on the 13 C NMR spectrum of powdered
cortisone, using two different 1 H RF nutation rates and continuous-wave (CW) versus two-pulse
phase modulation (TPPM) decoupling. (MAS rate: 5 kHz. 13 C NMR frequency: 75.40 MHz.)
5.1.3 ADVANCED CROSS POLARIZATION
Sideband m atching
C ross polarization (CP ) using Hartmann–Hahn matching involves matching

the nutation rates of the abundant (H) and dilute (X) spins,
Hartmann–Hahn CP was originally developed for static (non-spinning)
samples, but the same matching condition also works for 1 H/13 C CP under
MAS conditions due to the presence of the 1 H homonuclear couplings;
spinning at modest MAS rates has negligible impact on the strong
homonuclear coupling network, and optimal polarization transfer still occurs
at the simple Hartmann–Hahn matching condition.
At faster spin rates, when the 1 H spectrum is itself breaking up into
sidebands, or in isolated H,X spin systems, the MAS must be taken into
account explicitly, and the matching condition becomes:
5.1
where is the MAS rate and n is 1 or 2. In other words, the matching

condition is no longer at the “centerband,” (when the 1 H nutation rate
is matched to that of X), but shifted to “sidebands” either side of the normal
matching condition.
This is illustrated in Figure 5.3 , which shows the matching profile for CP
from 1 H to 13 C at two different spinning rates. When the sample is spinning
at 3 kHz, the profile shows a maximum at the normal Hartmann–Hahn
matching condition (around 60 kHz). By contrast, the matching profile at 15
kHz has broken up into sidebands, and the efficiency of the CP at the
“centerband” matching condition is very poor. As a result, the RF amplitudes
need to be modified so that the match is made on one of these sidebands.
The most immediate impact of having to match on a sideband is that the
matching is dependent on spinning rate, that is, it is necessary to optimize
the CP conditions at the target spin rate. More perniciously, the matching
condition is narrower and more delicate than at slow spin rates. Even the
“width” of the matching conditions shown in Figure 5.3 gives a misleading
impression of the degree of tolerance required when matching. Much of this
variation is the result of inhomogeneity in the B 1 fields of the RF, which
results in the matching condition (particularly for a sideband match) being
different in different parts of the sample. The narrowness of the matching
condition is largely responsible for the observed decrease in maximum CP
efficiency; the largest CP signal at the faster spin rate is only about half that
obtained at the slower rate. The “delicate” nature of the matching condition,
and its variation across the sample, makes it increasingly difficult to
maintain CP efficiency at increasing MAS rates.2
Figure 5.3 . A cross-polarization matching profile for hexamethylbenzene at spinning speeds of 3
(solid line) and 15 kHz (dashed line). The profile shows the intensity of the CP signal as a function of
the 1 H spin-loc k nutation rate while the 13 C spin-lock field is kept constant. The shaded region
indicates a typical range for ramped CP (see text).
Ramped c ross p olarization
The simplest solution to these problems is to use ramped CP . In contrast to

simple Hartmann–Hahn matching in which constant amplitude RF is applied
to both channels, ramped CP varies the RF nutation rate on one of the
channels, sweeping the RF amplitude “through” the matching condition. By
arranging the ramp to sweep through a complete sideband in the matching
profile (as indicated for the sideband by the shaded region in Figure
5.3 ), the matching conditions for different parts of the sample will all be
achieved at some point in the ramp. Note that the dynamics of ramped CP
are subtly different from those of conventional CP due to the time
dependence of the RF amplitudes, so the overall duration of the CP period
(the contact time) needs to be optimized along with the magnitude of the
ramp.3 In practice, however, ramped CP is considerably more robust with
respect to variations in experimental parameters in comparison with
conventional matching at fast spin rates. This makes the use of ramped CP
particularly valuable in the context of multidimensional NMR involving CP
since the reproducibility of CP efficiency is essential in such experiments.
Figure 5.4 shows the pulse sequence for a CPMAS experiment suitable
for high magnetic field and relatively fast MAS. In comparison with the
basic experiment illustrated in Figure 3.21 , this pulse sequence uses a ramp
of the 1 H nutation frequency rather than a simple Hartmann–Hahn match,
while TPPM is used rather than CW decoupling during the acquisition.
Figure 5.4 Schematic illustration of a typical cross-polarization magic-angle spinning (CPMAS)

experiment at high magnetic field and MAS rate, which uses ramped CP and two-pulse phase
modulation (TPPM) for 1 H decoupling. As shown, TPPM involves alternating the RF phase, with a
phase modulation of ø ~10–30° , with the duration of each pulse being typically a little less than that
of a 180° pulse.
5.2 LINEWIDTHS IN SOLID-STATE NMR

The width of resonances in solution-state NMR is not a topic of particular
interest since they tend, in the absence of chemical exchange effects, to be
more or less independent of sample and magnetic field.
The situation is considerably more complex in solids. For quadrupolar
nuclei under MAS, where the resolution is limited by second-order
quadrupolar broadenings (see chapter 6 ), increased magnetic fields offer a
clear advantage, since such broadenings (expressed in frequency units) scale
inversely with magnetic field, and so the overall resolution increases as the
square of B 0 . However, the resolution for spin- nuclei increases at most
linearly with magnetic field, or may not improve at all. To understand why
this is the case, the factors that determine spin- linewidths in solids need to
be examined in some detail.
Figure 5.5 illustrates how the overall observed lineshape can be
decomposed into two fundamental types of contribution: homogeneous
contributions , which determine the underlying intrinsic linewidth, and
inhomogeneous contributions . For instance, in solution-state NMR, the
homogeneous linewidth would be determined by T 2 (spin–spin) relaxation,
while inhomogeneous broadenings would arise from instrumental factors
such as inhomogeneity of the magnetic field. This variation of the magnetic
field would lead to a distribution of NMR frequencies and hence an overall
broadening of the lineshape. NMR in the solid state is subject to additional
broadening factors contributing to both the homogeneous and
inhomogeneous linewidths.
Figure 5.5. Schematic illustration of how the underlying “homogeneous” (intrinsic) lineshape and an
inhomogeneous distribution of NMR frequencies combine to give the overall observed lineshape. The
intrinsic linewidth is generally dominated by factors, such as couplings, that are independent of
magnetic field (and hence expressed in frequency units), while inhomogeneous broadenings scale with
magnetic field and are expressed in fractional (parts per million) units. The inhomogeneous
distribution spreads the NMR frequencies over a wider range, reducing the maximum signal intensity,
while preserving its integral.
The homogeneous/intrinsic lineshape is considered first. The limit on

the intrinsic linewidth is set by T 2 just as it is in solution-state NMR.
However, decay of the NMR signal due to T 2 relaxation is difficult to
distinguish from dephasing of the NMR signal due to residual dipolar
couplings. For example, 1 H NMR signals do not decay rapidly because of
fast spin–spin relaxation but rather from rapid “dipolar dephasing” due to the
strong homonuclear couplings. Spin diffusion via the homonuclear couplings
rapidly spreads the initial magnetization into other (unobserved) coherences.
(This is the basis of the dipolar dephasing experiment discussed in section 7.2.4.)
Studies in which the rate of dipolar dephasing is modified, for example by
changing the MAS rate, suggest that the effects of T 2 relaxation are
negligible in comparison with the effects of residual dipolar couplings.4
Hence the major contribution to the intrinsic linewidth is usually this
“residual coherent linewidth” due to dipolar couplings. Heteronuclear
decoupling slows down this dephasing process, but this is never 100%
effective and dilute-spin linewidths are often limited by the efficiency of the
heteronuclear decoupling. The effects of extensive dipolar coupling usually
dominate the linewidths of abundant spins themselves, for example, 1 H, 19
F, and often 31 P, even under fast MAS. Techniques for obtaining useful
resolution in these cases are discussed separately in section 5.5.
In terms of inhomogeneous contributions, there are again factors that are
particular to solid-state NMR. Dissolving a sample means that the NMR
responses of all the molecular entities of a given type are identical, and they
perfectly sum together to give sharp resonances. For well-crystalline samples
(i.e., samples in which a negligible fraction of sites are close to surfaces
and/or defects), crystalline symmetry means this should also be true for solid
NMR. In many cases, however, solid-state NMR is being used precisely
because it can provide information from samples that are not well-
crystalline, for example, amorphous solids. In such samples, the local
environment of different molecules is thus subtly different, and so a
distribution of chemical shifts will be observed for each site rather than a
single well-defined shift.
Although poor crystallinity generally reduces the quality of solid-state
NMR spectra, it is often the case that careful sample preparation does not
eliminate inhomogeneous line broadening. This is typically observed in
solids involving π-stacked aromatic rings. The bulk magnetic susceptibility
of such samples is strongly anisotropic, since the induced magnetic field
associated with the stacked rings (due to the aromatic ring current effect) is
strongly dependent on the orientation of the stack with respect to the external
magnetic field. This induced magnetic field changes the effective field in the
sample, uniformly shifting the NMR frequencies for a given crystallite. (This
overall bulk sample effect is distinct from the local changes in magnetic field
at the nuclei due to shielding.) The anisotropy of the bulk magnetic susceptibility
(ABMS) results in shifts that depend on the crystallite orientation. While the
isotropic component of the bulk magnetic susceptibility is eliminated by
MAS, the ABMS broadening is only partially averaged, and its effect can be
significant for certain samples, for example about 1 ppm for
hexamethylbenzene (both 1 H and 13 C resonances).
As illustrated in Figure 5.5 , the final lineshape arises from the
combination (mathematically, a convolution ) of the intrinsic/homogeneous
lineshape with any inhomogeneous distribution of the NMR frequencies. As
a result, the resolution of dilute-spin solid-state NMR spectra depends
crucially on the nature of the dominant source of linewidth. For samples
such as amorphous solids, where the dominant line broadening is
inhomogeneous in nature, the lineshape simply scales in width with NMR
frequency, and so increased magnetic fields offer no advantage in resolution
and only modest improvements in sensitivity. By contrast, linewidths for
crystalline samples are often limited by homogeneous residual dipolar
coupling effects. For example, steroids generally form well-defined
microcrystalline samples with negligible ABMS broadenings. As shown
previously in Figure 5.1 , their 13 C NMR spectra typically show extremely
narrow signals for non-protonated carbons, while the signals from the CH 2
region are significantly broader and are very sensitive to the 1 H decoupling
efficiency . This is characteristic of linewidths limited by the residual
influence of dipolar coupling.
5.3 EXPLOITING INDIRECT (J) COUPLINGS IN SOLIDS
Much of the power and flexibility of modern solution-state NMR flows from
the complementarity of information provided by chemical shifts and
through-bond (indirect) couplings; the chemical shift is key to resolving and
identifying sites with different chemical functionality, while indirect
couplings (often loosely termed J couplings) provide direct information on
the chemical connectivity between the different sites.
The situation is rather different in solids. The dipole–dipole interactions
(particularly those involving 1 H) are much larger, and techniques such as
MAS and high-power proton decoupling are necessary to suppress the
dipolar interactions in order to achieve useful chemical shift resolution .
Even after this, the linewidths observed in solids are often too broad to allow
other indirect couplings to be directly observed, with the exception of
spectra involving heavy atoms where J couplings are naturally large; for
example, coupling constants between two directly bonded 195 Pt nuclei may
be ~9 kHz.
Fortunately, it is often possible to exploit J couplings in the solid state
even when they are not resolved in the spectrum. As is illustrated in Figure
5.6 , a spin-echo experiment (see section 4.4.2) refocuses the evolution
under inhomogeneous factors and is often sufficient to reveal evolution
under J couplings, either homonuclear or heteronuclear.5 This evolution
appears as an oscillation superimposed on the decay due to relaxation and
other homogeneous line-broadening factors; that is, the signal for a given
spin coupled to one other will fit to:
Figure 5.6 31 P NMR spectra of a molybdenum pyrophosphate after spin-echo periods of different
durations . The peak intensities are modulated by the 2 J pp couplings (which are not refocused by the
spin echo), and the solid lines are fits to equation 5.2. These couplings, and hence the corresponding
P–O–P angle in the pyrophosphate (P2 O7 ) units, are very different for the inner and outer pairs of
peaks, providing direct structural information. The spectral linewidths must be dominated by
inhomogeneous effects, since couplings as small as 10 Hz are measured despite linewidths being over
an order of magnitude larger. (Figure adapted from results published in Lister et al ., Inorg. Chem . 49
(2010) 2290.)
5.2
where 2 τ is the total spin-echo period, J is the coupling constant, and is

the phenomenological time constant for the decay (the prime distinguishing
it from true spin–spin relaxation) . sets the ultimate limit on the size of
couplings that can be measured; if the coupling is significantly smaller than
then the NMR signal will have decayed before evolution under the
coupling can be observed or exploited. Particularly in non-protonated
samples, where dipole–dipole couplings are relatively modest, values can
be relatively long (tens or even hundreds of milliseconds), allowing
couplings as small as a few hertz to be observed in some cases.
By contrast, observation of J couplings involving 1 H is usually very
challenging for solids due to the rapid decay of the 1 H coherences under the
influence of the homonuclear dipolar couplings. For example, naively
omitting heteronuclear decoupling in order to observe a “coupled” 13 C
spectrum does not reveal the couplings but just gives a very poor
quality spectrum. However, by applying strong 1 H homonuclear
decoupling, the dipolar dephasing can be suppressed to the point where the
effects of couplings (which are typically in the range 125–250 Hz) can
be observed. This is illustrated in Figure 5.7 , which compares the 13 C
spectra of adamantane under conditions of 1 H heteronuclear and
homonuclear decoupling. When the 1 H homonuclear couplings are
suppressed, the CH 2 and CH resonances split into a triplet and doublet
respectively
Figure 5.7 . 13 C cross-polarization magic-an gle spinning (CPMAS) spectrum of powdered
adamantane acquired with (dashed line) CW 1 H heteronuclear decoupling and (solid line) BLEW-12
1 H homonuclear decoupling.
under the effect of the couplings. Note that the apparent couplings
(about 60 Hz) are scaled by the homonuclear pulse sequence; this scaling is
discussed further in sections 4.7.3 and 5.5.1. As explained in inset 7.8,
adamantane is a somewhat special case since molecular motion strongly
reduces the magnitude of the dipolar couplings, which means they can be
suppressed relatively effectively. However, with efficient decoupling
strategies, couplings can be resolved in more typical organic solids,
allowing an important class of solution-state NMR experiments to be applied
to these samples. The resulting correlation experiments are discussed in
more detail below.
missing image file couplings, which are vital to many classic
solution-state NMR experiments such as COSY and TOCSY, are at least an
order of magnitude weaker than There is little immediate prospect of
such small couplings being resolved in the context of typical rigid organic
solids. However, 2 J couplings are often accessible: missing image file
couplings measured across hydrogen bonds have been used to probe
intermolecular interactions, and a number of researchers have investigated
the dependence of missing image file couplings on bond angle in X–
O–X units (as in Figure 5.6 ).
5.4 SPECTRAL CORRELATION EXPERIMENTS
5.4.1 BASIC PRINCIPLES OF TWO-DIMENSIONAL NMR
A key class of “advanced” solid-state NMR experiments involves spectral

correlation , that is, two (or more) dimensional experiments that correlate one-
dimensional spectra of the same or of different nuclei. The basic principles
of 2D correlation spectroscopy are identical in solid- and solution-state
NMR so just the key principles are presented here as a reminder; see Further
reading for more detailed discussions.
Figure 5.8 schematically illustrates the basic principles of 2D NMR. An
additional (variable) evolution time t 1 is inserted after the initial
“excitation,” during which the magnetization evolves. An optional mixing
time may then, say, allow exchange of magnetization between sites, etc. The
NMR signal is then detected during t 2 . The evolution during t 1 is thus
observed indirectly; crucially, the amplitude6 of the NMR signal at the start
of t 2 is “modulated” by the evolution in t 1 (plus mixing time). The two-
dimensional data set is created by repeating the experiment but regularly
incrementing t 1 in order to sample the evolution in this indirect dimension .
Double Fourier transformation of this data set (with respect to both t 1 and t 2
) results in a two-dimensional spectrum.
The nature of indirect detection means that careful thought is required
when setting up two-dimensional experiments. When the NMR signal is
detected directly, there is little penalty for sampling more data points than
strictly required, either by using very short dwell times (leading to a larger
spectral width than necessary) or acquiring far out into the free-induction
decay (FID) where the NMR signal has disappeared. When sampling
indirect dimensions, however, the duration of the experiment increases
directly with the number of sampling points, and so the dwell time and
maximum t 1 need to be chosen carefully to keep total experiment times
reasonable. A second difference between direct and indirect dimensions is
the treatment of frequencies outside the set spectral width. In the direct
dimension, filters largely suppress frequencies (both signal and noise)
outside the spectral width,7 although some frequencies outside the set width
may “fold” around (see Figure 3.13 ) . By contrast, there is no means to
suppress frequencies outside the spectral width set by the sampling rate in
indirect dimensions; all the frequencies outside the spectral width will be
folded back in. It is then often necessary to set indirect dimension spectral
widths to be a multiple of the spinning rate so that centerbands and
sidebands are folded on top of each other rather than cluttering up the
spectrum.
missing image file

Figure 5.8 . Schematic illustration of 1D and 2D NMR pulse sequences. In 1D NMR, the NMR signal
is D etected directly after an E xcitation block (such as a 90° pulse or cross polarization). Detection
involves direct sampling of the NMR signal at intervals of missing image file (the dwell time). (P
refers to P reparation—e.g., a period for the magnetization to recover toward thermal equilibrium.) In
2D NMR, an additional time period, t 1 , is added after excitation, during which the magnetization
evolves. The signal is then detected during t 2 after an optional “M ixing block” of RF pulses. The t 1
time is regularly incremented to create a 2D data set with each row corresponding to a different t 1
value.
A final point of contrast between directly and indirectly detected

dimensions is the handling of “sign discrimination”. As discussed in sections
3.4.1 and 4.4.1, both the x and y components of the precessing magnetization
are typically measured when acquiring an NMR signal, allowing frequencies
with different signs to be discriminated. There are different schemes of
achieving such quadrature detection in indirect dimensions, but a common
approach is to acquire two FIDs per t 1 increment, which are analogous to the
x and y components of the directly detected signal. Details of these schemes,
together with other practical details of implementing 2D NMR and
processing 2D spectra can be found in the Further reading.
The power of 2D NMR lies in the flexibility of the basic scheme;
changing the different “blocks” of the pulse sequence allows a wide variety
of spectra to be created. This chapter focuses on using 2D NMR to establish
correlations between one-dimensional spectra, for example peaks in the 2D
spectrum indicate resonances that are linked by indirect or dipolar couplings.
The key step in such correlation experiments is the mixing period ( M ),
during which magnetization that has evolved on one spin during t 1 is
transferred to another spin prior to detection of the NMR frequencies in t 2 .
This transfer is usually via J couplings in solution-state NMR experiments,
although the dipolar couplings are used indirectly to provide proximity
information in NOESY-type experiments (the dipolar coupling being largely
responsible for the nuclear Overhauser effect). The strengths of dipolar
couplings in the solid state make them a natural first choice for correlation
experiments, but J couplings can also be exploited in favorable cases (as
previously discussed). Often both dipolar-based and J-based experiments are
necessary when correlating complex spectra or when the underlying
structure is unknown.
5.4.2 TRANSFER OF MAGNETIZATION VIA DIPOLAR

COUPLINGS
Dipolar couplings are in general much larger than corresponding indirect (J)
couplings, and so it ought to be relatively straightforward to exploit dipolar
coupling to transfer magnetization between the spins before the
magnetization decays away. Indeed, for 1 H spins in typical organic solids, it
is sufficient to use a short mixing period of the order of milliseconds; z
magnetization is efficiently exchanged between the spins via the dipolar
couplings (i.e., spin diffusion). The rate of spin diffusion may be reduced by
MAS, but it is still effective even at the highest achievable spinning rates.
Transfer of magnetization via the dipolar couplings becomes more
difficult when the coupling network is weaker, as these couplings are then
efficiently suppressed by the MAS required for good spectral resolution. For
example, the rate of direct 13 C, 13 C spin diffusion under MAS conditions is
generally too slow to be useful even in fully 13 C labeled compounds.8 In
addition, the rate of spin diffusion is likely to depend strongly on the sites
involved—the greater the frequency difference between the sites (relative to
their linewidths), the slower the rate of spin diffusion. This complicates the
interpretation of correlation peaks, since their intensities are not simply
related to spatial proximity.
In order to get around this problem, RF pulse sequences can be used to
recouple the dipolar interactions of interest, while using MAS to suppress the
unwanted interactions. In the absence of RF, evolution under individual
dipolar couplings is refocused over a rotor period, forming rotary echoes at
multiples of the rotor period (see Figure 3.15 and section 4.6.1). Unless the
coupling is sufficiently large for spinning sidebands to be observed, it is not
then possible to observe or exploit the coupling. However, addition of RF
pulses, synchronized with the rotation, disrupts the averaging by MAS. Fully
disrupting the MAS averaging to reintroduce CSAs as well as dipolar
couplings is obviously counterproductive, and so the RF pulse sequences are
designed to selectively recouple the interactions of interest.
Many such sequences have been devised, for recoupling both
homonuclear and heteronuclear interactions. For example, radiofrequency-driven
recoupling (RFDR) involves applying a train of rotation-synchronized π pulses
to recouple (see Figure 3.12 for an example). More complex schemes such
as POST-C7 are designed to create a well-defined “average Hamiltonian”
over the rotation period that depends only on the dipolar couplings. (Note
that such sequences often require fixed relationships between the spin rate,
and the relevant RF nutation rate, missing image file for example, the
condition missing image file must be met for C7-based sequences.) In
practice, the particular sequence used is not critical for the qualitative
applications described here.
For most applications, it is desirable for these sequences to be
“broadband,” that is, the efficiency of the dipolar recoupling should be
approximately constant across the spectrum. This ensures that the intensity
of any cross peak in a correlation experiment provides a reasonable measure
of the strength of the coupling between the pair of spins involved. The
magnetization transfer will be fastest between spins with the strongest
couplings, but as the mixing period increases, magnetization will equilibrate
between all the coupled spins. Such nonselective transfer is usually
appropriate for correlation experiments. Site-selective exchange of
magnetization is usually more relevant when the goal is measurement of
specific internuclear distances. These applications are discussed in section
8.3.
5.4.3 HETERONUCLEAR CORRELATION
Heteronuclear correlation via the dipolar interaction is particularly

straightforward since the necessary magnetization transfer can generally be
achieved using CP. Figure 5.9 shows such a heteronuclear correlation, or
HETCOR, experiment applied to 1 H and 29 Si. The spectrum shows
correlation peaks between the 1 H signal at ~17 ppm and the silicate
framework, allowing this resonance to be assigned to SiOH groups. No
correlations are observed with the signal at ~5 ppm, which is assigned to
water.
Obtaining useful correlations from HETCOR and related experiments is
more difficult if abun-dant, strongly coupled spins, such as 1 H, are involved.
Firstly, homonuclear decoupling (see section 5.5) is usually necessary in
order to narrow the abundant spin linewidths.9 More subtly, any spin
diffusion occurring during the magnetization transfer step (CP) will
compromise the correlation information in the final spectrum. For example,
if dilute spin X has a significant (heteronuclear) coupling to abundant spin A
but not to B, the HETCOR would ideally only show a cross peak between A
and X. If, however, magnetization is exchanged between A and B due to
spin diffusion during the mixing time, then some magnetization starting on B
will finish on X at the end of the mixing time, leading to a potentially
misleading (but sometimes useful) additional relay peak between B and X.
Keeping the mixing time short limits the build-up of such relay peaks, as
well as cross peaks arising from longer-range (weaker) dipolar couplings,
but at the expense of overall sensitivity. Indeed, as illustrated in Figure 5.10 ,
it is useful to record HETCOR experiments at both short and longer mixing
times to distinguish between the correlations that build up quickly (and so
correspond to truly short distances) and those that build up more slowly.
Another technique for reducing the impact of spin diffusion on such
correlation experiments is to use a modified CP sequence that suppresses the
homonuclear couplings while maintaining polarization transfer via the
heteronuclear couplings. The simplest of these techniques, Lee-Goldburg (LG)
CP, is discussed in the context of homonuclear decoupling in section 5.5.1.
missing image file
Figure 5.9 . (Left) Schematic of an H/X HETCOR experiment: 1 H magnetization evolves at the 1 H
NMR frequencies during t 1 , before being transferred (usually via CP to the X spins) where it is
detected. (Right) An 1 H/29 Si HETCOR spectrum of octosilicate acquired in about 14 h. In this case,
the 1 H spins are sufficiently weakly coupled to each other that 1 H homonuclear decoupling during t
1 was unnecessary. The traces at the top and l eft-hand side are separate 1-D spectra, not projections.
missing image file
Figure 5.10 . 1 H/13 C HETCOR experiment on 3-methoxy benzoic acid using cross-polarization
contact times of (bottom) 100 μ s and (top) 400 μ s. At short mixing (contact) times, correlations are
only observed between protonated carbon sites and their associated H resonances. At longer times,
additional correlations are observed, for example, between the carboxylic acid resonance (at ~173
ppm, split by crystallographic inequivalence) and the acid protons (~13 ppm). Experiment time: ~14 h
for each spectrum.
One important application of correlation spectra is to assist the

assignment of one-dimensional NMR spectra, by connecting known peaks in
one spectrum to unknown peaks in the other. However, for complex organic
systems, the ambiguities introduced by 1 H spin diffusion often limit the
usefulness of the simple HETCOR experiment. Experiments based on
indirect couplings may provide more clear-cut answers in these cases. As
discussed in section 5.3, it is possible to resolve missing image file
couplings in rigid organic solids by applying homonuclear decoupling to the
1 H spins. This allows techniques familiar from the solution state to be
adapted to the solid state, such as the heteronuclear multiple quantum

coherence experiment illustrated in Figure 5.11 . Because the correlation
information is provided by through-bond couplings and is not compromised
by spin diffusion, such 2D experiments provide unambiguous correlations
between 1 H and dilute-spin spectra. Note how the 1D 1 H spectrum of this
system would be strongly overlapped and uninterpretable. The 2D
correlation is effective in “pulling apart” these overlapped resonances,
allowing the chemical shift of the hydrogens attached to each carbon to be
determined. On a practical note, however, these experiments involve
extended periods of continuous homonuclear and heteronuclear decoupling
and are technically demanding; careful optimization of the homonuclear
decoupling is required to ensure that the missing image file couplings
can be resolved while at the same time avoiding overstressing the probe.
missing image file

Figure 5.11 . Solid-state NMR version of the heteronuclear multiple quantum coherence experiment,
which correlates the 13 C and 1 H NMR spectra via J couplings, as applied to cholesteryl acetate. This
alkyl section of the 2D spectrum shows how each of the carbon sites in the 13 C spectrum can be
connected with the 1 H resonance of the attached hydrogens. (Figure adapted from results published in
Lesage et al. , J. Am. Chem. Soc. 120 (1998) 13194.)
5.4.4 HOMONUCLEAR CORRELATION
Homonuclear correlation experiments are particularly valuable for abundant

nuclei such as 19 F and 31 P, as illustrated in Figure 5.12 . In contrast to 1 H,
where specialized homonuclear decoupling is often necessary, simple MAS
is usually sufficient for these nuclei to resolve multiple sites, although
homonuclear recoupling sequences (such as RFDR or C7) are needed to
ensure “broadband” recoupling of the dipolar interactions.
missing image file

Figure 5.12 . (Left) Schematic of a homonuclear correlation experiment using radiofrequency-driven
recoupling (RFDR) (dipolar recoupling). X magnetization (generated from CP) evolves during t 1
before the magnetization is transferred to the z direction. Application of 180° pulses in the middle of
each rotor period “recouples” the homonuclear dipolar interactions, allowing magnetization to be
exchanged during τ mix . The magnetization is then returned to the xy plane for detection. (Right) 19 F
correlation spectrum of the semicrystalline polymer polyvinylidene fluoride. The 2D plot
unambiguously shows that the “defect” (reverse unit) sites are associated with the amorphous regions
of the polymer (experiment time: ~18 h).
As for the heteronuclear case discussed earlier, correlations derived

from J couplings are a useful complement to dipolar-based homonuclear
experiments. Figure 5.13 shows the pulse sequence for the refocused
INADEQUATE experiment which is widely used to acquire J-based
homonuclear correlation spectra in solids.10 This experiment is directly
analogous to the INADEQUATE experiment in solution-state NMR,11 but it
is more appropriately termed a double quantum/single quantum (DQ/SQ)
correlation when dealing with abundant spins such as 31 P; the Incredible
Natural Abundance part of the INADEQUATE acronym is hardly
appropriate for 100% abundant nuclei!
missing image file

Figure 5.13 . The (refocused) INADEQUATE pulse sequence as used for solids. During the time 2τ
following CP, the magnetization evolves under the influence of (J) couplings to create double-quantum
coherences between coupled spins; the 180° pulse refocuses evolution due to chemical shifts, and
synchronizing the τ period to multiples of the rotor period avoids modulation of the signal by MAS.
The double-quantum coherences evolve during the indirect acquisition time t 1 before being
“refocused” to simple single-quantum coherences prior to signal acquisition. Phase cycling is essential
to ensure that the detected signal has passed through double-quantum coherences.
Figure 5.14 (a) shows a DQ/SQ correlation spectrum which exploits the
missing image file couplings within pyrophosphate groups,
missing image file . Although this coupling is small (typically 10–30
Hz) and is not resolved in the 1D NMR spectrum, sufficient double-quantum
coherence can be created to result in a high-quality 2D spectrum. The
appearance of the spectra in Figure 5.14 is markedly different from the
previously shown single-quantum/single-quantum correlation experiments
since the indirect dimension plots double-quantum frequencies. Such coherences
involve pairs of spins and evolve at a frequency that is the sum of the
individual NMR (i.e., single-quantum) frequencies. A “cross peak” in a
double-quantum/single-quantum correlation experiment consists of a pair of
horizontal peaks linking two single-quantum frequencies (direct dimension)
with the double-quantum frequency (indirect dimension), which is the sum
of the two SQ frequencies. The presence of this correlation implies that a
double-quantum coherence was created involving the linked spins. Figure
5.14 compares the information content from J- and dipolar-based spectra on
the same sample: the J-based spectrum, Figure 5.14 (a), identifies 31 P
resonances related by bonding, that is, those part of the same pyrophosphate
unit, while the dipolar-based experiment, Figure 5.14 (b), gives information
on spatial proximity between P sites. There is the risk with dipolar-based
correlation experiments in particular, that so many correlations are observed
that little concrete information can be derived. In these cases, it is often
useful to look for absent correlations, indicated by dashed circles in Figure
5.14 (b), which identify sites that must be relatively far apart.
missing image file
Figure 5.14 . 31 P DQ/SQ homonuclear correlation experiments on a pyrophosphate material. (a) The
correlations in the J-based experiment (refocused INADEQUATE) show which pairs of resonances are
part of the same P2 O7 unit. The diagonal peak (*) is the result of a “rotational resonance” and is not
the result of J coupling (see text). (b) The dipolar-based experiment (using POST-C7 recoupling)
shows the correlations within the pyrophosphate units, and also weaker longer range couplings which
reflect interactions between nearby P2 O7 units. Dashed circles mark absent correlations and arrows
mark auto-correlation peaks (see text). (Figure adapted from results published in Lister et al ., Inorg.
Chem . 49 (2010) 2290.)
Interpretation of c ross p eaks in c orrelation s pectra
In most cases, the presence of a cross peak (or pair of cross peaks in double-
quantum/single experiments) can be interpreted straightforwardly in terms of
close spatial proximity of the spins involved (dipolar-based experiments) or
in terms of chemical bonding (experiments based on indirect (J) couplings).
There are, however, some subtle issues that must be borne in mind when
analyzing correlation spectra.
Figure 5.15 illustrates schematically the different types of DQ/SQ and
SQ/SQ correlation experiments. The information content of the double-
quantum and single-quantum variants is essentially the same (and the same
recoupling sequences can be used for both). The DQ/SQ spectra are, in
general, less crowded since peaks along the diagonal are not usually present
(but see the discussion of autocorrelation peaks below), whereas diagonal
peaks are always present in simple SQ/SQ correlation spectra.12 This makes
it easier to observe cross peaks between sites of similar NMR frequency.
missing image file

Figure 5.15 . Schematic illustration of different correlation experiments for a system of three
inequivalent spins, A–C, in which only sites A and B are linked by indirect (J) coupling. The relative
size of the spots indicates the strength of the correlation; the stronger the coupling, the more intense
the cross peak at a given mixing time.
In contrast to J-based experiments, dipolar-based experiments will tend

to show cross peaks between all sites in the same phase as the mixing time
increases, as a result of spin diffusion (see secti on 2.7). While the cross
peaks between close spins (A/B and B/C) will build up more rapidly, cross
peaks are also observed between A and C, corresponding to a combination
of slower magnetization transfer via the weak A,C coupling together with
relayed transfer of magnetization between A and C via site B. Since it is also
impossible to distinguish between transfer via inter- and intra-molecular
couplings, care is required when interpreting correlation peaks in dipolar-
based experiments. As illustrated in Figure 5.10 , it is advisable to obtain
correlation spectra at both short and longer mixing times in order to identify
the longer-range correlations. Spin diffusion can be used to good effect to
distinguish physically distinct components in heterogeneous samples, as
previously shown in Figure 5.12 .
Interpreting J-based experiments is generally more straightforward,
since (subject to an important caveat given below) the presence of a cross
peak unambiguously identifies sites with a nonnegligible indirect coupling,
which generally corresponds to pairs connected by a bonding pathway. Care
is required, however, in interpreting autocorrelation peaks , which involve pairs
of sites with the same (isotropic) chemical shift. If the sites are fully
equivalent, then no double-quantum coherence will be created via the J
coupling (just as coupling between equivalent spins has no visible effect on
the conventional NMR spectrum). This is in contrast to the dipolar coupling-
based experiment, in which double quantum coherence can be created
between both equivalent and inequivalent sites13 , for example, the arrowed
peaks in Figure 5.14 (b). Note that the equivalence of spins is a somewhat
subtle issue in solid-state NMR; this is discussed in inset 5.1.
Inset 5.1. Equivalence in solid-state NMR
In the solid state, two sites are only truly equivalent if they are related by a center
of inversion or simple translation; in this case, all the tensors for the NMR
interactions are always identical. If, however, there is a symmetry relationship
between the sites which is less than this, then only the isotropic components of the
NMR interactions are strictly identical. The sites will have the same isotropic
NMR frequency, but will be inequivalent at a general crystallite orientation as a
result, for example, of CSA. Coupling between such sites will result in a so-called
rotational resonance under sample spinning, leading to unusual NMR
lineshapes and unexpected autocorrelation peaks in 2D spectra (see Figure 8.7 for
an example of a related type of rotational resonance). Such rotational resonance
effects can be suppressed by “spinning out” the anisotropic interactions, but they
can, with care, be used to elucidate quite subtle questions of crystal symmetry.
5.5 HOMONUCLEAR DECOUPLING
The suppression of the strong dipolar couplings between protons in the solid
state has been a repeated theme in many of the experiments described above.
This section explores this important topic in greater detail.
As discussed in section 2.7 and inset 4.5, the Hamiltonian for a
collection of spins coupled by homonuclear dipolar couplings behaves
“homogeneously” under MAS; in contrast to other anisotropic interactions
such as the CSA, MAS does not break up the static lineshape into a set of
sharp spinning sidebands, but rather the sidebands (and centerband) have a
finite width that decreases only modestly with increasing spin rate. (To a
good first approximation, the linewidths decrease inversely with increasing
spin rate.)
The problem of limited resolution is particularly acute in 1 H NMR.
This is firstly due to the strength of the homonuclear interactions; 1 H has a
high magnetogyric ratio and individual dipolar couplings are large (e.g., over
20 kHz for the two protons in a rigid CH2 group), resulting in a large overall
homogeneous linewidth (typically ~50 kHz for a rigid solid). Secondly, the 1
H chemical shift range is limited to about 15 ppm, which corresponds to a
range of frequencies of only 7.5 kHz at an 1 H Larmor frequency of 500
MHz. As a result, even the fastest MAS is typically only able to resolve a
handful of distinct resonances. The situation is much easier for other
abundant spins such as 19 F and 31 P since overall coupling strengths are
lower and the range of NMR frequencies spanned is much larger,
particularly at higher magnetic fields. It is usually much easier when
working at current magnetic field strengths14 to use fast MAS rather than
homonuclear decoupling for these nuclei. The remainder of this section
focuses exclusively on homonuclear decoupling applied to 1 H.
Achieving useful resolution in a 1 H spectrum requires the effects of the
homonuclear couplings to be suppressed by more than two orders of
magnitude. This is a demanding requirement, and even small experimental
imperfections can become noticeable when such high levels of performance
are needed. Moreover, homonuclear decoupling is subject to a variety of
experimental factors that are often hard to control or characterize (e.g.,
variation of B 1 across the sample, finite rise and fall time of pulses). The
higher the resolution expected from the homonuclear decoupling, the more
critical the experimental set-up.
A significant complication associated with all the widely used
decoupling sequences (see below) is that the 1 H magnetization precesses
about a tilted axis during the homonuclear decoupling, rather than precessing
about the z axis as is the case in the absence of RF irradiation (Figure 5.16 ).
Moreover, the exact position of the precession axis is not always well
defined, as a result of experimental imperfections. Sampling the raw x and y
magnetization leads to quadrature errors when the signal is Fourier transformed,
that is, artifact peaks appear at frequency –f for each genuine signal at
frequency f . The usual solution is to perform the experiment off-resonance;
the transmitter frequency is placed to one end of the 1 H spectrum making it
unnecessary to acquire a full quadrature signal. This may degrade the
effectiveness of decoupling for peaks furthest from the transmitter, but the
limited frequen cy range of the 1 H spectrum means that off-resonance
detection is usually preferred.
missing image file

Figure 5.16 . The effective Hamiltonian under RF homonuclear decoupling in general corresponds to
precession in a plane tilted away from the z axis. This tilting means that the amplitude of the motion
has a different projection on the x and y axes, which leads to quadrature artifacts if the magnetization
is sampled directly.
Homonuclear decoupling does not remove other anisotropic components

of the Hamiltonian. Hence if homonuclear decoupling is being applied to
obtain resolved 1 H spectra, it is usually combined with MAS to suppress the
other anisotropic interactions, particularly the CSA. This combination of RF
homonuclear decoupling and MAS is known as CRAMPS—combined rotation
and multiple-pulse spectroscopy .
Figure 5.17 illustrates the resolution that can be achieved on 1 H using
either fast MAS alone to suppress the homonuclear couplings or using MAS
at a moderate rate plus RF decoupling. The simple MAS experiment
provides some useful correlation information, especially in the less crowded
high-frequency regions of the spectrum, while the experiment using RF
decoupling provides correlation information for all the 1 H resonances.
missing image file
Figure 5.17 . DQ/SQ 1 H correlation experiments on a small dipeptide using (left) 30 kHz magic-
angle spinning (MAS) and (right) 12.5 kHz MAS + homonuclear decoupling in both dimensions. The
improvements in resolution are substantial (about a factor of 5). (Figure adapted from results
published in Brown et al. , J. Am. Chem. Soc. 126 (2004) 13230.)
5.5.1 OVERVIEW OF HOMONUCLEAR DECOUPLING

SEQUENCES
Much of the early development of NMR involved abundant spins in solid

samples, and so the problem of reducing linewidths in dipolar-coupled
systems was addressed at an early stage. One of the first sequences to
emerge was the WHH-4 (or WAHUHA) sequence previously discussed in
section 4.7. 1. As shown in Figure 4.6 , the repeating element of WHH-4
involves four 90° pulses separated by delays, with the double-length delay
providing a convenient window for acquiring the NMR signal. Such windowed
sequences can be used in the direct dimension of an NMR experiment. In
subsequent developments (e.g., MREV-8, BR-24, and others), this basic
pattern was extended by “supercycling” to achieve higher quality decoupling
and make the sequences more robust with respect to experimental
imperfections.
To work efficiently, windowed sequences require both the pulses and
the delay periods to be as short as possible without overloading the probe or
causing excessive breakthrough of the tail of a pulse with the signal
acquisition (see section 3.3.5). Hence, acquiring the NMR signal during
homonuclear decoupling requires careful balancing of the experimental
parameters. These considerations are largely redundant if homonuclear
decoupling is being applied in indirect dimensions, that is, it is being applied
during a t 1 period of experiments such as 1 H/13 C HETCOR. This allows
the use of windowless sequences that use continuous RF irradiation,15 for
example, the classic BLEW-12 sequence as well as more recently developed
sequences such as PMLG (see below) and DUMBO (see Further reading).
Lee-Goldburg (LG) d ecoupling
One particularly important windowless sequence is LG decoupling . This

involves using off- resonance RF irradiation, where the RF offset has been
chosen such that the effective nutation axis is at the magic angle with respect
to the static magnetic field (see section 3.2.1 for discussion of off-resonance
irradiation). That is, if the nutation rate due to on-resonance RF is
missing image file the RF transmitter is set missing image file off-
resonance. The nutation axis is then at
with respect to the z axis. Magnetization precesses16 about this tilted axis
and the homonuclear couplings are suppressed (to first order).
Simple LG decoupling is not particularly competitive for achieving
high-resolution spectra since it only suppresses the dipolar line broadening
to lowest order. In order to suppress “higher order” terms and hence achieve
more effective decoupling, frequency-switched Lee-Goldburg (FSLG) decoupling
(Figure 5.18 ) alternates the sign of the frequency offset and switches the RF
phase. This significantly improves decoupling performance. Note that FSLG
is frequently implemented using continuous linear ramps of the RF phase,
avoiding the need for frequency switching, in which case it is usually termed
phase-modulated Lee-Goldburg (PMLG) decoupling.
Although there are close analogies to physical MAS of the sample, it is
important to note the differences. In MAS, it is the spatial component of the
Hamiltonian (the interaction tensor) that is being modulated by the spinning.
Hence isotropic components (such as the isotropic chemical shift) are
unaffected by the averaging process. In LG decoupling, however, it is, in
effect, the spin components of the Hamiltonian that are being modulated.
“Bilinear” interactions such as the homonuclear dipolar couplings are
averaged (to first order). But “linear” terms, such as the chemical shift
Hamiltonian or heteronuclear couplings, are also partially averaged—they
are scaled by a factor of missing image file for LG decoupling. Hence
if a spin would precess at frequency Ω in the absence of RF, it will precess at
a frequency missing image file under LG decoupling.
As discussed in more detail in section 4.7.3, this scaling of the
precession frequencies is an unavoidable consequence of using RF
irradiation to suppress a component of the spin Hamiltonian, and it is
important to correct for this scaling before extracting chemical shift
information from CRAMPS spectra. Although the ideal scaling factor for a
given sequence will be known, observed scaling factors are quite sensitive to
experimental factors, and it is good practice to measure the actual scaling
factor on a known set-up sample.
missing image file

Figure 5.18 . Schematic illustration of the Lee-Goldburg (LG) and frequency-switched Lee-Goldburg
(FSLG) decoupling sequences. LG decoupling involves continuous radiation of fixed phase (here x ),
but with the transmitter off-resonance such that the precession axis is at the magic angle,
missing image file with respect to z . The frequency-switched variant involves simultaneously
switching the phase by 180° and swapping the sign of the resonance offset. The switching period
usually corresponds to a full 360° rotation about the tilted precession axis.
LG irradiation is frequently used as a simple way to reduce 1 H spin

diffusion during the CP mixing time of HETCOR-type experiments (see
section 5.4.3); off-resonance irradiation is used to spin-lock the 1 H
magnetization along an axis inclined at the magic angle.17
Interaction b etween h omonuclear d ecoupling and MAS
A final important practical consideration is the interaction between

homonuclear decoupling and MAS. The classic decoupling sequences were
developed for static or slowly spinning samples and their performance
strongly degrades when the MAS rotation period becomes comparable to the
repeat period of the decoupling sequence. This is a particular problem at
higher magnetic fields since faster spinning rates are necessary to spin out
interactions such as the CSA, making it difficult to avoid destructive
interaction between the homonuclear decoupling and MAS. Hence
sequences with short repeat periods such as PMLG/FSLG have distinct
advantages over many of the classic decoupling approaches.
5.6 USING CORRELATION EXPERIMENTS FOR SPECTRAL

ASSIGNMENT
Assignment of the peaks in an NMR spectrum to chemical sites is

straightforward if the resonances are clearly distinguished and can be
identified using well-established correlations between isotropic chemical
shifts and structural groups, for example using correlation charts derived
from solution-state studies. The differences in chemical shifts between the
solution and the solid states are usually small, and so assignments of the
solution-state NMR spectrum can be useful guides when assigning solid-
state spectra. However, ambiguities often arise, either because of spectral
complexity or from substantial differences between solution- and solid-state
chemical shifts, for example due to intermolecular hydrogen bonding. As
discussed in section 8.8.2, first principles calculations can be used to predict
chemical shifts in crystalline solids starting from the atomic coordinates in
the crystallographic unit cell. Since these account for intermolecular
interactions, they are a useful complement to solution-state assignments,
where the crystal structure is known. In complex cases, however, assignment
of closely spaced peaks may still not be possible, and further experiments
may be required.
Useful assignment information for dilute spins in organic solids can
often be provided by relatively straightforward one-dimensional NMR
methods that probe the local dipolar environment. For example, dipolar
dephasing experiments, discussed in section 7.2.4, selectively observe sites
with weak dipolar interactions to 1 H, allowing CH and CH2 sites to be
distinguished from quaternary carbons and methyl groups (where molecular
motion strongly reduces the net dipolar couplings).
In more difficult cases, however, two-dimensional correlation
experiments may be necessary. For example, 13 C/1 H HETCOR experiments
are useful for assigning 13 C resonances if the 1 H spectrum can be at least
partially assigned. However, homonuclear correlation experiments involving
well-resolved spectra provide the most powerful tools for assignment. Thus
a 13 C INADEQUATE experiment can solve a common problem of
assignment in cases where there is more than one molecule in the
crystallographic asymmetric unit. In such cases, each chemical type of
carbon gives rise to multiple signals, and it is very difficult to determine
which individual signal is associated with which individual molecule in the
crystallographic asymmetric unit. Thus for two doublets, A and B say, there
is no simple way of determining whether the low-frequency A signal is for
the same molecule as the low-frequency B signal or not. However, the
INADEQUATE spectrum allows such linkages to be made.
Figure 5.19 shows part of the 13 C INADEQUATE spectrum of the a -
form of testosterone, which has two molecules in the crystallographic
asymmetric unit. The horizontal and vertical lines link signals for bonded
carbons, showing that the high-frequency C9 line is in the same independent
molecule as the low-frequency C8 line. Of course, connectivity obtained by
this method for an organic molecule is interrupted when heteroatoms
intervene (a problem that does not occur for testosterone). Similarly tracing
out the connectivity is complicated if the pairs are not clearly resolved.
missing image file

Figure 5.19 . INADEQUATE 13 C spectrum for a testosterone (see text). This experiment, performed
at natural isotopic abundance, required about 3 days of continuous spectrometer time on a 500 MHz
instrument! (Figure reproduced, with permission, from Harris et al. , Phys. Chem. Chem. Phys. 8
(2006) 137).
The major weakness of INADEQUATE-style experiments for dilute

spins is that it requires observation of signals from adjacent pairs of NMR-
active nuclei. For 13 C at natura l isotopic abundance, this involves only
~0.01% of the molecules for each atom pair. Thus this experiment is highly
demanding both of spectrometer time (often requiring several days!) and of
hardware—high-power proton decoupling is being applied for relatively
long periods. By contrast, such DQ/SQ correlation experiments are
straightforward for abundant spins such as 31 P (see Figure 5.14 ).
5.7 FURTHER APPLICATIONS
5.7.1 LABELED SYSTEMS
This chapter has concentrated on experimental protocols that are practical

for “normal” samples with natural isotopic abundance. However, for some
systems, such as large biomolecules, working at natural abundance is not
viable due to the complexity of the spectra and the small quantities of
samples that can be produced affordably. Hence NMR studies of large
molecular systems, such as biomolecules, are usually only feasible after
isotopic enrichment. Nonselective or partially selective labeling strategies
are used to enhance the concentration of spin-active nuclei, typically 13 C
and 15 N, and, increasingly, deuteration schemes also are being used to dilute
the concentration of 1 H spins. The latter reduces the strength of the proton–
proton coupling network, greatly improving resolution (particularly on 1 H
itself, but also for the dilute spins), and reduces the problems caused by spin
diffusion. Indeed, with a high-quality microcrystalline sample, the 13 C
resolution is often limited by 1 J CC couplings, which can no longer be
ignored in enriched samples; pulse sequences may need to be adapted to
include “homonuclear decoupling” of 13 C or to selectively observe
individual components of J multiplets.
Isotopic enrichment of other dilute spins has a major influence on the
viability of various solid-state NMR experiments. In particular, methods
involving double-quantum coherences between spin pairs become much
more accessible, as illustrated by the INADEQUATE experiment discussed
above. Especially if 1 H resolution has been significantly improved via
partial deuteration, the suites of techniques and protocols used for the
assignment and analysis of proteins in the solution state can then be adapted
for the solid state. More information on this important area can be found in
the Further reading.
5.7.2 QUANTITATIVE APPLICATIONS
This chapter has focused on applications of 2D NMR to provide resolved

spectra that can be used in an essentially qualitative manner to answer
chemical problems, for example, tracing connectivity in a framework
material or a molecular solid. Once these questions have been answered,
however, it is often possible to use NMR to address quantitative problems of
structure and dynamics in solid materials.
Chapter 7 discusses how molecular motion affects NMR spectra and
introduces a range of NMR experiments that can be used to quantify
parameters such as exchange rates and activation barriers. Finally, chapter 8
builds directly on the techniques introduced here in order to provide
quantitative information on NMR parameters and structure. In particular,
dipolar recoupling, often combined with selective labeling to mark
individual sites or pairs of sites, can be used to provide accurate
measurements of dipolar couplings and hence internuclear distances.
FURTHER READING
2D AND MULTI-DIMENSIONAL NMR APPLIED TO THE SOLUTION
STATE
“Understanding NMR spectroscopy” , J. Keeler, Wiley (2005), ISBN 978 0 470 01786 9.
“High-resolution NMR techniques in organic chemistry” , T.D.W. Claridge, Elsevier (2009), ISBN
978 0 080 54818 0.
HETERONUCLEAR DECOUPLING
“Heteronuclear decoupling in the NMR of solids”, P. Hodgkinson, Prog. NMR Spectry. , 46 (2005)
197. DOI: 10.1016/j.pnmrs.2005.04.002.
HOMONUCLEAR DECOUPLING AND 1 H NMR
“High-resolution 1 H NMR spectroscopy of solids”, P. Hodgkinson, Ann. Rep. NMR Spectry. , 72

(2011) 185–223.
“Probing proton–proton proximities in the solid state”, S.P. Brown, Prog. NMR Spectry. , 50 (2007)
199. DOI: 10.1016/j.pnmrs.2006.10.002.
BIOMOLECULAR APPLICATIONS
“NMR spectroscopy of biological solids”, Ed. A. Ramamoorthy, CRC Press (2005), ISBN 978 1 574
44496 4.
NOTES
1 Spinning sideband suppression techniques, as discussed in section 3.3.6, are only a partial solution
as they cancel out the sidebands without increasing centerband intensity.
2 Faster MAS also decreases 1 H spin diffusion rates (see section 2.7), slowing the rate at which
magnetization can be transferred from more distant 1 H spins to the X spin. This further diminishes
overall CP efficiency.
3 At first sight, the time dependence of the matching condition in ramped CP might be expected to
reduce the effectiveness of polarization transfer at an individual matching condition. In fact, sweeping
through a match can (in principle) result in a more complete (adiabatic ) transfer of polarization from
one spin to the other.
4 It is a common misconception that T relaxation must be fast in solids. Classic NMR relaxation
2
theory assumes that the motional processes occur on a much faster timescale than the NMR frequency
and so its results cannot be simply extrapolated to the slow-motion limit; indeed T 2 relaxation must
be slow in the absence of motional processes to drive it. At the other end of the scale, the only
difference between a plastic crystalline phase and a liquid is the absence of translational motion.
Hence it would be expected that missing image file for such “soft solids” as for the fast-motion
limit in solution-state NMR.
5 The heteronuclear case requires simultaneous π pulses on both nuclei.
6 Generally, the amplitude of the signal is modulated at the start of t , rather than its phase . “Phase
2
modulation” usually results in “twisted” lineshapes in 2D spectra, which cannot be phased into all-
positive peaks. These subtle, but important, issues in 2D NMR are discussed in more detail in the
Further reading.
7 In some cases, it can be useful to “open up” the filters and deliberately allow extensive folding in the
direct dimension, for example to superimpose all the peaks of a spinning sideband manifold.
8 Carbon-carbon proximity information must be obtained via effi cient 1 H spin diffusion. Such
experiments, such as proton-driven spin diffusion , are only viable for labeled compounds and are not
considered further.
9 The choice of probe and MAS rotor for heteronuclear experiments is generally determined by the
least sensitive nucleus. As this will be the dilute spin (for unlabeled samples), the rotor diameter will
tend to be relatively large, limiting the maximum MAS rate that can be used.
10 “Refocused” refers to the refocusing period after t which converts the anti-phase doublets
1
normally observed in solution-state INADEQUATE experiments into in-phase doublets. The lower
resolution in typical solid-state NMR spectra makes the acquisition of anti-phase doublets highly
undesirable, since the overlap of adjacent signals with opposite signs results in extensive signal
cancellation.
11 The original solution-state 13 C INADEQUATE experiment is rarely used due to its extremely low
sensitivity (the signal arises from the tiny fraction of C–C bonds in which both nuclei are 13 C).
Although an even more demanding experiment in the solid state, it may be the only means to
completely assign a 13 C spectrum. The situation is considerably eased if samples can be globally
enriched in 13 C.
12 With appropriate phase cycling, it is possible to modify SQ/SQ correlation experiments so that only
pathways involving double-quantum coherences are selected. The information content of such double-
quantum-filtered experiments is then equivalent to the DQ/SQ experiment, with the only significant
difference lying in the form of the spectra.
13 The reason for the difference lies in the different form of the Hamiltonians for the rank-0
(isotropic) and rank-2 tensor components of NMR couplings (see section 4.2.3). Figure 2.4 shows the
spectrum of a pair of equivalent spins coupled by dipole–dipole coupling (pure rank-2); the equivalent
spectrum for a pair of identical spins coupled by an isotropic J (indirect dipole–dipole) interaction
(pure rank-0) would be a simple singlet!
14 However, increased magnetic fields also mean that the contribution of CSAs and isotropic chemical
shift differences to the overall Hamiltonian can become substantial for nuclei such as 19 F. As a result,
decoupling sequences derived for 1 H NMR often perform poorly when applied to other nuclei.
15 Windowless sequences (which use continuous RF) will tend to decouple more efficiently than
windowed sequences consisting of isolated pulses. Indeed, recent work that applies homonuclear
decoupling in directly acquired dimensions has generally inserted acquisition windows into new
windowless sequences (such as PMLG, DUMBO) rather than using classic windowed sequences.
16 Precession rather than nutation is used here as the spin evolution over complete cycles of the Lee-
Goldburg irradiation rather than the rapid nutation during the LG cycle is being considered. See the
discussion of WHH-4 decoupling in section 4.7.1 for more details.
17 Note that the effective nutation rate under off-resonance irradiation is given by
missing image file which corresponds to an increase by a factor of missing image file at the LG
condition. The matching condition needs to be adjusted accordingly.
CHAPTER 6
QUADRUPOLAR NUCLEI
6.1 INTRODUCTION
This chapter is intended to give the reader an understanding of the unique

characteristics of the quadrupolar interaction, which come into play when it is
significant in magnitude compared with the Zeeman interaction, giving second-order
effects. The reader is also introduced to the special techniques that are frequently
necessary to obtain good-quality, well-resolved spectra of quadrupolar nuclei.
As briefly mentioned in chapter 1 , quadrupolar nuclides are distinguished by
the fact that they have spin quantum numbers greater than . Their NMR properties
are substantially different from those of spin- nuclides. The width of the spectrum
is usually dominated by quadrupolar coupling, the magnitude of which varies very
widely. As explained in section 2.6, this effect is caused by the existence of an
electric quadrupole moment, eQ , which is characteristic of the nuclide in question.
The value of Q ranges from –0.0808 fm2 for 6 Li to 85.6 fm2 for 189 Os and is over
200 fm2 for isotopes of Hf, Ta, and Re. Values for all stable nuclides (except those
of the lanthanides and actinides) are to be found in appendix B. Some of the more
important ones are given in table 6.1 . Note that quadrupole moments can be either
positive (prolate) or negative (oblate),1 though this is usually of little consequence
for NMR. Now the broader a resonance, the lower the sensitivity in terms of S/N, so
that it becomes more difficult to resolve signals from the same nuclide in different
chemical situations. Since Q strongly affects spectral widths, nuclides with small Q
are the easiest to study, other things being equal. The 2 H, 6 Li, and 133 Cs nuclides
have particularly small quadrupole moments and their NMR spectra are usually
relatively simple. The ratio Ξ/Q 2 (with Q in fm2 and Ξ in % —see table 6.1 ) strongly
influences the occurrence of second-order effects (see section 6.4) and so it gives a
measure of the viability of obtaining good resolution (other factors being
comparable). From this point of view, it can be seen that 2 H, 6 Li, and 133 Cs are
expected to give particularly good spectra. It is also clear that 37 Cl may be preferred
to 35 Cl (though the relative isotopic abundances indicate that the latter would give
the higher intensity).
Table 6.1. Quadrupole moments of some nuclides
Factors other than Q can be equally or more important for spectra of

quadrupolar nuclei. In particular, the electric field gradient (EFG) at the nucleus—
which is a tensor—plays a strong role. In fact, a quadrupolar spectrum depends on
the relevant quadrupolar coupling constant , χ (see equation 2.25, where it is defined in
frequency units).2 This represents the anisotropy of the quadrupolar tensor (see
section 2.6) and depends on both e Q and , where the latter is the largest (in
magnitude) principal component of the EFG.
The value of the EFG depends (unlike the quadrupole moment) on chemistry.
Simple ions, such as are frequently found for compounds of the alkali metals and
halogens, have, in principle, zero EFGs at the nucleus (though in practice, the
environment of an ion in a crystal may reduce the symmetry substantially) and
therefore give narrow resonances. This fact, combined with the low value of Q for
133 Cs, leads to this nucleus being regarded as having pseudo-spin- status! Atoms
at sites of cubic, octahedral, or tetrahedral symmetry will also have very low values
of χ . For instance, 33 S in sulfate ions gives relatively sharp resonances; the
narrowest appears to be for ammonium aluminum sulfate dodecahydrate, for which
the linewidth is 18 Hz.3 In other chemical circumstances the EFG can be large,
leading to broad lines. Finally, as shown by equation 2.27, there is a strong influence
of the spin quantum number on the first-order quadrupolar energy.
As discussed in section 2.6, nuclides with spin quantum numbers, I , which are
odd multiples of , have relatively sharp central transitions (i.e., ) , whereas
the other ( satellite ) transitions are spread over a range of frequencies for
polycrystalline samples. Each type of transition may be said to give a subspectrum . On
the other hand, nuclides with integral spin quantum numbers have no central
transitions and their spectra are inevitably spread over a range of frequencies.
Fortunately, of such nuclei only 14 N raises any real problems for the NMR
spectroscopist; its quadrupole coupling constants are frequently of the order of 3
MHz and are therefore too large to treat simply.
In general, spectra of quadrupolar nuclei will be influenced by shielding
anisotropy and asymmetry as well as the quadrupolar effects themselves. However,
in this chapter the interplay between the two tensors will be ignored and the
quadrupolar effects treated in isolation. Chapter 8 will discuss some more
complicated cases.
6.2 CHARACTERISTICS OF FIRST-ORDER QUADRUPOLAR

SPECTRA
Magic-angle spinning (MAS) is commonly used to record spectra of quadrupolar

nuclei. MAS can, in principle, average first-order quadrupolar interactions to zero,
but in most cases accessible spin rates are insufficient for the result to be a single
line for each chemically distinct site. In consequence, it is normal for a substantial
number of spinning sidebands to appear. Figure 6.1 (a) shows a complete (central
and satellite transitions) MAS spectrum of a spin- nucleus with zero asymmetry.
The quadrupole coupling constant is only modest in value (150 kHz) so that first-
order theory (section 6.3) suffices to explain the spectrum. The strong central
transition is the most prominent feature (as for the static case shown in figure 2.9 ).
To record the whole of such spectra, the maximum available spin rate should
normally be used (unless the quadrupole coupling constant is small).
Figure 6.2 shows an experimental I = 1 ( 2 H) spectrum (for a nearly axially
symmetrical situation, as is usual for C–D bonds). The slight lack of symmetry in
the 2 H spectrum illustrated here probably arises for technical reasons (e.g.,
digitization limitations), although there may be some effect from shielding
anisotropy.
Figure 6.1 . Simulated magic-angle spinning (MAS) spectra for a first-order spin - nucleus with χ = 150 kHz:
(a) η = 0, and (b) η = 1. Shielding anisotropy is ignored. The intensity of the central transition is truncated.
Figure 6.2 . Deuterium magic-angle spinning (MAS) spectrum at 7.05 T and a spin rate of 8 kHz for 2,3-d2 -
fuma ric acid. The quadrupolar parameters are χ ≅ 160 kHz and η ≅ 0.
The strong central feature of spectra for nuclides with I an odd multiple of can
be readily used to determine the isotropic chemical shift for both static and MAS
cases (but see later for complications in cases with large quadrupole coupling
constants). For nuclides with integral spin, such a measurement is more difficult and
it may be necessary to take the full bandshape into account. However, in both cases,
it is feasible to fit the total bandshape to give the quadrupole coupling constant;
suitable computer programs are available for simulating and iteratively fitting both
static and MAS spectra, in the latter case using the intensities in the spinning
sideband manifolds. The results, in turn, give information on the electronic
environment of the nucleus and thus on the bonding situation. Quantum theory is
often used to compute quadrupole coupling constants from crystal structure
information, thus providing a sensitive test for both theory and experiment.
Thus far, it has been tacitly assumed that axially symmetric sites are being
discussed. However, in general, the quadrupole coupling tensor is asymmetric.
Asymmetries affect the appearance of static spectra (and hence of MAS spectra
also). Figure 6.1 (b) shows an example for a spin - nucleus with η = 1 , for
comparison with the η = 0 case of figure 6.1 (a). The computer programs mentioned
above take account of asymmetries and thus these parameters can be derived by
fitting static spectra or spinning sideband manifolds.
The accurate setting of the magic angle is particularly important for
quadrupolar nuclei because of the frequently very wide range of the resonance. A
very small offset from the magic angle is immediately manifested in line broadening
for the spinning sidebands of the satellite transitions. In fact, the usual way of
setting the magic angle is to optimize the linewidths of spinning sidebands for the 79
Br spectrum of KBr. The optimization is best done by observing the duration of the
free-induction decay (FID), as shown in figure 6.3 .
Figure 6.3 . Bromine-79 free-induction decays (FIDs) of KBr, spinning at 3 kHz (a) off (by <1° ) and (b) on the
magic angle at a magnetic field of 9.4 T. The insets show portions of the corresponding spectra away from the
centerband.
The total intensity in a given subspectrum (central or satellite transition) is
governed by the raising and lowering operators (I + and I – , respectively; see section
4.2.1) and is, therefore, for the transition proportional to:
6.1
See inset 6.1 for subspectra intensity ratios4 as a function of I .
Inset 6.1. Intensity ratios for subspectra of different spins
I Ratio
3:4:3
5:8:9:8:5
7 : 12 : 15 : 16 : 15 : 12 : 7
9 : 16 : 21 : 24 : 25 : 24 : 21 : 16 : 9
6.3 FIRST-ORDER ENERGY LEVELS & SPECTRA
Quadrupolar coupling is unique among NMR interactions in that quadrupole

coupling constants are frequently several megahertz (and may be up to several
hundred megahertz) in magnitude, which is significant in relation to the Zeeman
energy. In many cases (especially at high magnetic fields B 0 ), the Zeeman term is
still dominant, so quadrupolar coupling can be seen as mixing the spin states
established by the Zeeman interaction. For relatively low quadrupole coupling
constants, perturbation theory is applicable. For high-spin nuclei, such theory is
adequate even for moderately large quadrupole coupling constants. Perturbation
theory allows the contribution of the quadrupolar energy to be expressed as a sum of
terms of decreasing significance. When quadrupole coupling constants are very
high, a full theoretical treatment will be necessary, but in most cases of interest,
second-order perturbation theory is sufficient. Therefore, only the first two
quadrupolar Hamiltonian terms will be retained in this book:
6.2
In many cases (with low quadrupole coupling constants), a first-order perturbation

treatment of quadrupolar effects suffices, so that equation 2.27 for the quadrupolar
energy is applicable. Such situations have been fully discussed in section 2.6 and
readers should digest the information therein before tackling the rest of the present
chapter. The most important result is that the energy levels are affected
equally by first-order perturbations and therefore the central transition remains as a
single line (independent of the EFG orientation in the magnetic field) at the
isotropic chemical shift. However, the first-order approach is often inadequate, and
the levels are shifted to different extents (see figure 6.4 ) so that the
resonance frequency of the central transition becomes influenced by quadrupolar
effects (and depends on the orientation of the quadrupolar tensor with respect to B 0
). Thus, when the quadrupolar energy is above about 5% of the Zeeman energy,
extending the perturbation treatment to second order is required.
Figure 6.4 . Energy-level diagram for a spin- nucleus in a magnetic field B 0 , showing the effects of zero
(left-hand side), first-order (center), and second-order (right-hand side) quadrupolar perturbations, the last
leading (in general) to a decrease in the resonance frequency for the central transition.
As mentioned in section 4.2.3, the full Hamiltonian for a coupling interaction

between spins I 1 and I 2 is where R is the tensor for the coupling in
question. In the case of the quadrupolar effect, only one spin is involved and the
Hamiltonian becomes:
6.3
As discussed in section 2.6, only two parameters are in principle needed to

characterize the magnitude of q in any given situation, namely its anisotropy and
asymmetry ( η ). The chosen parameters for quadrupolar interactions are the
quadrupole coupling constant, χ (which is proportional to the anisotropy of q ) and
the quadrupolar asymmetry (identical to the asymmetry of q ). These are defined in
the principal axis system (PAS) in equations 2.25 and 2.26. The anisotropy χ
(alternative symbol C Q ) is normally expressed in frequency units but η is
dimensionless (0 ≤ η ≤ 1 ). It may be shown that, in terms of these parameters,
equation 6.3 becomes (in the quadrupolar PAS):
6.4
where and are the raising and lowering operators, respectively (see section
4.2.1).
Before this equation is developed here, the reasons why the results are often
mathematically complicated should be understood. The first is that all Hamiltonian
terms must be given in the same axis system. However, equation 6.4 is expressed in
the principal axis frame of the EFG tensor, whereas the Zeeman energy is
determined by B 0 , which lies in the laboratory frame of reference. Therefore, a
conversion of the spin operators from one frame to the other is necessary, which
involves rotations, generally over two angles (see section 4.6 for details). In fact,
when MAS is employed, the rotor axis must also be considered and an additional
rotation of axes is necessary. Converting Hamiltonian terms by rotations is relatively
simple when the tensors are axial, since then only one angle is involved (e.g., that
between the Z EFG axis and B ).5 Thus, when asymmetry is neglected, equation
0
6.4 becomes, in the laboratory frame:
6.5
This is the origin of the familiar 3cos2 θ – 1 factor, which is averaged to zero by
MAS (though proper consideration involves transformation to the axes of the
rotation). However, the angular dependence in the second and third terms within the
square brackets cannot be averaged by MAS and is the source of difficulties for
high-resolution spectra of quadrupolar nuclei, as discussed in the next section.
Significant additional complications result if the quadrupole tensor is asymmetric,
since three angles are then needed to describe rotations of axes.
The second reason for complications lies in the nature of the spin operators.
Terms such as are secular (i.e., affect the energies of states but do not mix
states significantly); application of such operators gives first-order energy changes.
Terms such as missing image file and missing image file (se e sec tion
4.2.1), on the other hand, mix states (in the sense that the wavefunctions are
significantly changed), requiring substantial mathematical manipulation. This gives
rise to second-order contributions to the energies.
In what follows in this section and later sections, the text will deal with various
relatively simple situations, starting with the effects of asymmetry on first-order
spectra.
When asymmetry is included, the value of missing image file in the
laboratory frame is found to be:
(Capital letters for the axes indicate the principal components, as usual.) If only the
secular terms are retained (i.e., at relatively low quadrupole coupling), the truncated
quadrupolar Hamiltonian becomes:
Application of the Hamiltonian 6.7 gives the first-order quadrupolar correction to

the energy, including the effect of asymmetry, which has been discussed in chapter 2
(see equation 2.27). Equation 2.28 expressed the transition frequencies when the
asymmetry is zero, while figure 2.9 showed an experimental spectrum for a static
sample of a zero-asymmetry case. Equation 6.8 gives the transition frequencies
when there is non-zero asymmetry of the EFG
Figure 6.5 shows schematically the effect of asymmetry on the static powder pattern
for a spin - n ucleus. The two subspectra for the transitions missing image file
and missing image file are indicated. In all cases the spectrum is symmetrical (in
the absence of shielding anisotropy). Thus the subspectra for the satellite transitions
are mirror images for all values of the asymmetry parameter; they completely
overlap when η = 1.
missing image file

Figure 6.5 . Simulated full spectra for a static sample with missing image file and varying asymmetry,
showing the subspectra (dashed lines). The satellite transitions are identical when η = 1. The quadrupole
coupling constant used is 100 kHz. The central transition has been cut off below its full intensity.
Magic-angle spinning increases the S/N but results in the appearance of

spinning sideband manifolds, which may span moderately large ranges for the
subspectra (except for the central transition). Simulated spectra of this type were
displayed in figure 6.1 .
While equation 6.7 applies equally to nuclei with integral spin quantum
numbers and to those with quantum numbers an odd multiple of , this chapter will
mainly address spectra for the latter. However, section 6.8 will deal briefly with
nuclides of integral spin quantum number.
The first-order quadrupolar effect is independent of the applied magnetic field,
B 0 . Therefore, as the applied field increases, the ratio of quadrupolar to Zeeman
energy decreases and spectra become less complicated by the second-order effects
to be described below. Generally speaking, this results in better resolution
(enhanced because chemical shifts increase in frequency units). There are also gains
in signal intensity with increase in B 0 .
6.4 SECOND-ORDER ZERO-ASYMMETRY CASES

6.4.1 TRANSITION FREQUENCIES
When quadrupole coupling constants are significant with respect to the Zeeman
term, it becomes necessary to include perturbation to second order, which causes a
number of complications to appear in the spectra. This is because terms involving
the raising and lowering operators cause mixing of states, as mentioned above,
which means that the eigenstates are no longer pure Zeeman states. This leads to
second-order energy contributions and thence to shifts in the spectra. The evaluation
of the relevant equations can be tedious unless advanced methods (spherical
harmonics) are used (see chapter 4 ), and the results are complicated. Here only an
equation for the second-order quadrupolar energy, missing image file of a static
sample in an axially symmetrical case will be given:
where θ is the angle between the EFG symmetry axis and B 0 , while ν 0 is the
Larmor frequency, and m I is the spin component quantum number of the energy
level in question. In equation 6.9, the three terms in the square brackets are referred
to as zeroth-, second-, and fourth-rank contributions . They involve the zeroth-,
second-, and fourth-degree Legendre polynomials , respectively. The zeroth-de gree
polynomial is unity and the other two are:
These two factors arise from the requirement to rotate the spin operators between
different axis systems (in this case those of B 0 and the EFG). The second-degree
polynomial will be recognized as the one involved in MAS , as discussed in chapter
2.
Since equation 6.9 depends on m I as well as missing image file the
energies of the levels missing image file are changed differentially. In fact the
proportionality of E Q to mI shows that they are affected to equal but opposite
extents. Therefore, the splitting between these levels (the central transition) is
altered (in contrast to the zero effect of the first-order quadrupolar term)—see figure
6.4 . Moreover, the magnitude of this effect depends on the orientation of the EFG
as expressed by θ . Therefore, the central transition will be a powder pattern when
polycrystalline or amorphous materials are studied.
6.4.2 CENTRAL-TRANSITION SPECTRA: STATIC SAMPLES
The simplest form for the resonance frequency of the central transition (
missing image file ) for an axially symmetric case, which can be derived from
equation 6.9 , is:
where the spin-dependent factor, f (I ), is given by:
Equation 6.12 makes a number of spectral features of the second-order quadrupolar

effect on the central transition clear:
There is a general shift away from ν 0 for all orientations except for θ = 0 ° ,
180 ° , and missing image file ( θ = 70.53 ° and 109.47 ° ).
Since this shift depends on the orientation of the EFG in B 0 , a
microcrystalline sample will give a powder pattern, in contrast to the first-
order perturbation (which left the central transition as a single line).
The center of gravity of the powder pattern will not be at ν 0 .
The effect is inversely proportional to ν 0 and will therefore become smaller if
B 0 is increased.
The magnitude of the effect decreases with I , other things being equal (see
inset 6.2).
Since the second-order term as a whole is not proportional to
missing image file , MAS cannot eliminate the effect.
Inset 6.2. Dependence of the second-order effect on the nuclear spin quantum number
for the central transition
The value of the spin-dependent factor, f ( I ), is as follows:
INSET IMAGE MISSING
This factor means that, for comparable quadrupolar coupling constants, second-order
effects get significantly smaller as I increases, which is perhaps counterintuitive. Thus,
second-order broadening of the central transition for 87 Sr ( missing image file ) is only
0.055 that of 87 Rb ( missing image file ) in spite of the fact that the former has the larger
quadrupole moment (33.5 vs. 13.35 fm 2 ).
The upper spectrum of figure 6.6 shows an example of second-order

quadrupolar effects: the central 27 Al transition for a static sample of
microcrystalline solid aluminum acetylacetonate at 7.0 T. The effects of the first
three of the points bulleted above can be readily seen in the top spectrum of that
figure. The EFG for this compound is nearly axially symmetric ( η = 0.16), with a
modest quadrupole coupling constant ( χ = 3.0 MHz).
missing image file
Figure 6.6 . Central-transition 27 Al spectra of polycrystalline aluminum acetylacetonate. Top: static sample.
Bottom: Sample rotated at the magic angle at 9.26 kHz. The position of the isotropic chemical shift is indicated
as ν 0 . The small peak just to high frequency of ν 0 arises from the centerband of the inner satellite transition (
missing image file ).
While equation 6.12 is useful for descriptive purposes, greater insight can be
obtained by recasting it into a form obtained by use of Legendre Polynomials (see
equation 6.9):
The center of gravity of the powder pattern for the central transition is determined
by an isotropic second-order shift , given (for zero asymmetry) by the first term of equation
6.14:
Thus, as mentioned earlier, the center of gravity is shifted to low frequency, though
(for a static sample) the powder pattern extends both sides of the isotropic shielding
position.
6.4.3 CENTRAL-TRANSITION SPECTRA: RAPID SAMPLE SPINNING
When a sample is spun about an angle β to B 0 , there is some (though not complete)
averaging of the second-order contribution to the resonance frequency. Equation
6.14 shows that the average values of missing image file and missing image file
and are required. This involves another change of axes and therefore brings in new
Legendre polynomials. The necessary average of equation 6.14 can be shown to be:
where the angle β is that between the rotor axis and B 0 , while missing image file
is the angle between the Z axis of the EFG and the rotor axis (which can take any
value for a powder sample). The factors missing image file and
missing image file act to scale the relevant terms for the powder bandshape. At the
magic angle, missing image file the middle term is eliminated, while the final term
is scaled by missing image file
Thus the second-order shift under conditions of MAS (including the isotropic
part) becomes:
This term describes a powder pattern, with low-frequency shifts for all values of
missing image file (unlike the static case). The center of gravity ( equation 6.15 )
is invariant to the spinning. The lower spectrum in figure 6. 6 shows the case of
aluminum acetylacetonate under MAS conditions. Clearly, the second-order effects
depend on the ratio missing image file and thus inversely on the magnitude of the
applied magnetic field B 0 . Figure 6.7 shows such a variation of the central
transition for a zero-asymmetry case. The bottom spectrum contains sharp spinning
sidebands from the outer transitions, which become broadened as
missing image file increases. At higher missing image file ratios, spinning
sidebands of the central transition appear.
Direct measurements from spectra such as that shown in figure 6.6 can yield
information about both the isotropic shielding and the quadrupolar coupling
constant, the latter being strongly related to the electronic environment of the
nucleus in question. Inset 6.3 indicates how the relevant parameters may be derived
from the MAS powder pattern of the central transition for an axially symmetric
case. Optimum choice of operating magnetic field depends on the aim of the
experiment. Resolution is best for the highest B 0 (and this makes determination of
isotropic chemical shifts relatively easy). However, when quadrupole coupling
constants are small, it may be advantageous to obtain the quadrupolar parameters
from the central-transition bandshape at relatively low applied magnetic field so as
to increase the magnitude of the second-order broadening. Alternatively, such
parameters may be derived from analysis of the spinning sidebands from the
satellite transitions at low missing image file .
missing image file
Figure 6.7 . Variation of the central-transition region of a spin - spectrum (simulated) with the ratio of the
quadrupolar coupling constant to the Larmor frequency , missing image file . The dotted line indicates the
isotropic chemical shift.
Inset 6.3. Analysis of the central transition for a second-order MAS powder pattern
Figure 6.8 shows a schematic powder pattern for the central transition of a quadrupolar
nucleus under MAS conditions, with the positions of the three turning points indicated.
Expressions for these turning points may be obtained by differentiating equation 6.17.
They are found to be at missing image file = 0 ° , 90 ° , and 49.1 ° (with values of 5 –
18cos 2 missing image file + 21 cos 4 missing image file equal to 8, 5, and
missing image file , respectively). The positions are given in figure 6.8 in units of K ,
where:
For a zero-asymmetry case with a well-defined powder pattern, the quadrupole coupling
constant can be readily derived from the splitting between the prominent horns of the
powder pattern, which is missing image file In the case of aluminum acetylacetonate (see
figure 6.6 ), this splitting is 670 Hz, leading to a value of 2.7 MHz for missing image file
. The powder pattern may be reproduced in full by suitable computer programs and the
values of missing image file and missing image file obtained more accurately by
iterative fitting of the full bandshape. For the aluminum acetylacetonate case, a value of 3.0
MHz for missing image file was obtained from a fit of the full bandshape. Although
assumed to be zero in the approximation above, the fit gave an asymmetry value of
missing image file
missing image file
Figure 6.8 . Schematic magic-angle spinning (MA S) NMR powder pattern for the central transition of a second-order
spectrum for a quadrupolar nucleus with an axially symmetric EFG. The angle missing image file = 22. 2 ° gives rise
to a resonance at the same frequency as the turning point at 90° .
6.5 SPECTRA FOR CASES WITH NON-ZERO ASYMMETRY:

CENTRAL TRANSITION
The relevant equations for cases with missing image file are significantly more
complex and will not be quoted here. The reader is referred to other texts, as listed
in the Further reading section, for details. However, the general effects of
missing image file for the MAS case are shown in figure 6.9 . For a few purposes,
the effect of asymmetry is simply introduced by a factor of missing image file .
For example, in equation 6.15 for the isotropic second-order effect, the factor 3 must
be replaced by missing image file . Since missing image file , the influence of
this factor in such cases is rather small.
Of course, real examples do not often look like the ideal cases of figure 6.9 .
Thus figure 6.10 shows the central-transition portion of the 27 Al ( I =
missing image file ) spectrum for a natural aluminosilicate clay. The
relatively featureless spectrum is typical of those from many aluminosilicate
materials that contain some disorder. The two intense centerband resonances are
from aluminum octahedrally ( δ Al ≅ 1 ppm) and tetrahedrally ( δ Al ≅ 58 ppm)
coordinated by oxygen. The remaining signals are spinning sidebands. While the
theoretical shapes of the resonances are not apparent, the isotropic chemical shifts
(including the second-order quadrupolar contribution) can be derived from the
centers of gravity of the signals. If the spectrum is measured as a function of
magnetic field, the true isotropic shift (i.e., excluding quadrupolar effects) can be
obtained by plotting the center of gravity as a function of the inverse of the Larmor
frequency and extrapolating to zero.
missing image file

Figure 6.9 . Simulated second-order magic-angle spinning (MAS) quadrupolar powder patterns for the
centerband of a spin- nuclide as a function of the asymmetry factor for χ = 1 MHz and missing image file .
missing image file
Figure 6.10 . Aluminum-27 magic-angle spinning (MAS) spectrum, obtained at 9.4 T with a spin rate of 12
kHz, for an aluminosilicate clay.
6.6 RECORDING ONE-DIMENSIONAL SPECTRA OF

QUADRUPOLAR NUCLEI
For most samples, spin–lattice relaxation times of quadrupolar nuclei are short since
they are dominated by the quadrupolar mechanism and this is normally efficient (a
consequence of the large quadrupolar interactions!) provided there is some
molecular-level mobility. Therefore direct-excitation methods (repeated single
pulses, each followed by an FID) are normally used (for both central and satellite
transitions). Recycle delays following the FIDs do not need to be lengthy (in
contrast to the case for spin- nuclei). This implies that the optimum signal-to-noise
ratio may be obtained from pulse angles less than 90° , so it may be worth using the
Ernst angle (see section 3.3.1). Moreover, there are other reasons for keeping the
pulse angle low (see below). Proton decoupling may or may not be required,
depending on the chemical system involved.
An example of the potential value of obtaining quadrupole coupling
information is provided by the 23 Na spectrum of a mixture of sodium chloride and
sodium nitrite shown in figure 6.11 . It is immediately obvious that the high-
frequency signal has a negligible quadrupole coupling constant and so the sodium
atoms must be at a site of very high symmetry. This establishes the assignment as
that of sodium chloride. The other signal, which must be from sodium nitrite, has
the typical appearance arising from an axially symmetric site; simulation confirms
this and gives missing image file = 1.09 MHz. This is consistent with the
crystal structure and would have given structural information in the absence of
diffraction data.
missing image file

Figure 6.11 . Sodium-23 direct-excitation spectrum (centerband region), obtained at 79.34 MHz, of a mixture of
sodium chloride and sodium nitrite using a 15° pulse angle and a 1 s recycle delay. The spin rate was 4.5 kHz.
Cross-polarization (CP) is rarely required for the purpose of gaining signal

intensity, though it may sometimes be valuable to identify signals of nuclei near to
protons (i.e., to obtain selectivity). Moreover, there are some exceptions to the
general rule that T 1 is short for quadrupolar nuclei, especially for nuclides with low
quadrupole moments. In such cases, CP may be valuable in increasing signal
intensity. However, the match condition for CP to a quadrupolar nucleus does not
have the simplicity of equation 3.8 but now involves the quadrupolar coupling
constant (and therefore depends on both the sample and the nuclear environment).
Determining the optimum match condition is best done empirically on the system of
interest by varying either of the RF field strengths. The experiment is highly
selective.
When quadrupole coupling constants get significantly large (say, > 5 MHz), the
central transition becomes hundreds of kilohertz wide, even under MAS conditions,
and the S/N is reduced. This problem can be at least partially overcome by the use
of the quadrupole Carr–Purcell–Meiboom–Gill (QCPMG) method. As the name
implies, this technique is analogous to the CPMG method for spin- nuclides
described in section 7.2.3. It starts with a two-pulse sequence (
missing image file ) which causes an echo signal centered at a time τ 1
following the π pulse. Repeated π pulses then give a train of echoes, which are
detected. The experiment can be done on static samples or with MAS. In the latter
case, it is essential to have rotor synchronization throughout. In the quadrupolar
case, the pulse durations are scaled from the nominal π/2 and π angles by the Rabi
factor (see the next section). The normal powder-pattern signal is broken up into a
series of discrete spikelets (analogous to the occurrence of the sidebands caused by
MAS), separated by the inverse of the π-pulse separations (see figure 6.12 ). In
consequence, the S/N significantly increases. The quadrupolar parameters can be
obtained by analysis of the spikelet envelope. For cases with really large quadrupole
coupling constants, it is necessary to use co-added frequency-stepped experiments
in order to get a true representation of the bandshape.
missing image file

Figure 6.12 . Rubidium-87 quadru pole Carr–Purcell–Meiboom–Gill (QCPMG) spectrum of a static sample of
Rb2 SO4 , obtained at 9.4 T. There are two rubidium environments with chemical shifts 42 and 15 ppm. The
quadrupolar parameters are: χ = 2.8 and 5.7 MHz; η = 0.95 and 0.15, respectively. The spikelet separation is 625
Hz.
In practice, spectra will be influenced by shielding anisotropy and asymmetry.

Since the quadrupole coupling constant is independent of the applied magnetic field
and second-order quadrupole effects decrease as the field increases, whereas
chemical shift anisotropy increases (when expressed as a frequency), separately
determining the two tensors is facilitated by operating at two or more magnetic
fields and computer-simulating the spectra.
6.6.1 NUTATION
However, the existence of large quadrupole coupling constants raises new issues for
obtaining spectra . During solid-state experiments on spin- nuclei, the RF pulse
energy E RF (generally expressed as the equivalent angular frequency, γ B 1 ) is
sufficiently high in relation to the strengths of dipolar and shielding interactions that
they can be ignored during the pulse. However, that is not necessarily the case for
the quadrupolar interaction. In the situation where the quadrupolar energy is much
less than the RF energy, missing image file , nutation (rotation of the magnetization
under the influence of the RF; see section 3.2 .1) will occur as in a case with zero
quadrupole coupling, that is at an angular frequency γ B 1 . This is known as a hard
pulse and results in all the resonances (central and satellite transitions) being
observed.
On the other hand, in the extreme case that missing image file nutation
for the central-transition magnetization occurs at an angular frequency
missing image file where missing image file is known as the Rabi factor.
In such a case, the RF duration required for a 90° pulse is shorter than that for an
equivalent situation for a solution; for an missing image file case, for
example, the Rabi factor is 2. Such a situation is said to involve a soft pulse and the
effect is selective, which means that only the central transition of the quadrupolar
energy-level system will be fully affected. Both soft pulses and hard pulses have
uses for quadrupolar spectra. For intermediate values of the pulse power (sometimes
also referred to, rather confusingly, as “hard,” but better called simply nonselective ),6
the nutation situation is complex, and different coherences (see section 4.3) are
created. Figure 6.13 shows the development of two such coherences, namely single
quantum (central transition, SQ) and triple quantum (TQ) for a spin-
missing image file system following such a pulse. Because TQ coherence
cannot be created directly but only via the creation of SQ coherence, there is an
apparent “induction period” prior to the appearance of the former, as can be clearly
seen in figure 6.13 .
Of course, a particular pulse may be hard in respect of some nuclei but soft in
relation to other nuclei of the same isotope in the same sample. Variation of the
pulse duration, with two-dimensional Fourier transformation, gives a nutation spectrum
that can sharply reveal differences of quadrupolar interaction strengths in cases
where resonances overlap in one-dimensional spectra. However, such differences
produce problems for determining relative intensities between signals in well-
resolved one-dimensional spectra. The easiest way to obtain good results
semiquantitatively is to employ small pulse angles (≤8 ° ), since deviations between
the various cases are relatively small under such conditions. However, this does not
optimize the signal intensity for any of the nuclear sites.
missing image file

Figure 6.13 . Development of single-quantum (central transition) and triple-quantum coherences for an
missing image file s pin system by means of a RF pulse with a power-equivalent of 100 kHz for a system with
a quadrupole coupling constant of 0.5 MHz. The solid line is for SQ coherence, whereas the dashed line refers to
TQ coherence . (Figure supplied courtesy of Dr. Sharon Ashbrook.)
6.7 MANIPULATING THE QUADRUPOLAR EFFECT
Although it can be useful for the derivation of quadrupolar parameters, the broad
bandshapes caused by the second-order quadrupolar effects introduce problems of
spectral resolution, making analysis complicated if there is more than one relevant
site for the nucleus in the crystals in question. In principle, it would be useful to be
able to eliminate the second-order effects so as to improve resolution and to obtain
chemical shifts more readily. One obvious method is to use high magnetic fields, but
these are frequently not accessible and, in many cases, spectrometers operating at
sufficiently high fields do not exist. There are several more sophisticated ways
round this problem by manipulating the quadrupolar effect, as described in the next
few subsections. However, these techniques are largely confined to cases with
relatively small quadrupole coupling constants since impossibly high spin rates
would be required otherwise.
6.7.1 VARIABLE-ANGLE SPINNING

Spinning at the magic angle does not eliminate the second-order quadrupolar effect,
since it is not, as a whole, proportional to missing image file . Some narrowing
will be achieved by spinning about any angle, β , to B 0 . The effect can be seen by
examination of equation 6.16. The isotropic term will, of course remain under
variable-angle spinning (VAS). The other two terms will be scaled by the averages
of missing image file and missing image file , respectively. The first of these is
only zero for rotation at the magic angle. The term in missing image file is zero at
two angles, namely 30.56 ° and 70.12 ° . Unfortunately, the value of using these
angles is much reduced by the fact that shielding anisotropy and dipolar coupling
effects are not eliminated, nor are the contributions to the first- and second-order
quadrupolar terms involving missing image file —in fact, they are only scaled by
0.6122 and –0.3265, respectively. However, analysis of spectra obtained at different
spinning angles may enable shielding anisotropies and asymmetries to be separated
from quadrupolar effects. More importantly, such considerations form the basis of
the next technique to be discussed.
6.7.2 DOUBLE ROTATION
Spinning at an angle of 54.74° eliminates shielding anisotropy and dipolar

interactions, plus the first- and second-order quadrupolar terms that involve
missing image file ). However, use of the angles 30.56° or 70.12° removes the
second-order quadrupolar effects dependent on missing image file ) so that
simultaneous rotation about 54.74° and 30.56° , say, should remove all problems
arising from anisotropies. This will result in a single line for the central transition,
which will be at the Larmor frequency plus the isotropic second-order quadrupolar
shift. Spinning sidebands will also be seen. This experiment requires a rotor within a
rotor, with independent rotations. It is a difficult experiment to implement, but it can
be done. Figure 6.14 shows the general arrangement. The outer rotor is fixed at an
angle of 54.74° to B 0 while the angle between the two rotor axes is set to 30.56° .
Figure 6.15 compares a 17 O double rotation (DOR) experimental spectrum of l
-alanine with that of a simple MAS experiment. The former reveals the signals for
the two crystallographic sites, which overlap severely in the latter. A major
limitation of DOR lies in the low value of the feasible rates of rotation, especially
for the outer rotor; spinning sidebands can reduce the effective resolution.
missing image file
Figure 6.14 . The arrangement of rotors for double rotation (DOR).
missing image file

Figure 6.15 . Oxygen-17 spectra of l -alanine at 81.3 MHz. Top: magic-angle spinning (MAS) at ~16 kHz;
bottom: double rotation (DOR) with outer rotor rate 1.8 kHz and inner rotor rate ~8.5 kHz. Simulations of the
spectra of the two individual sites are plotted underlying the upper spectrum. (Fig ure supplied courtesy of
Professor Ray Dupree.)
6.7.3 DYNAMIC-ANGLE SPINNING
This technique provides two-dimensional spectra (see section 5.4.1 for a general
discussion of 2D operation), giving isotropic spectra in one dimension and
quadrupolar-broadened anisotropic spectra in the other. The experiment consists of
spinning at different angles during two times t 1 and kt 1 , where the constant k is
calculated so as to provide the elimination of the anisotropic effects. A CT -selective
90° pulse is applied prior to each of these times and the magnetization is stored in
the z direction (by another 90° pulse) between them while the rotor angle is changed.
The FID is recorded during time t 2 following the end of kt 1 . As is usual in two-
dimensional experiments, t 1 is incremented and the dataset S (t 1 , t 2 ) is doubly
Fourier transformed. In the f 1 dimension, the second-order quadrupolar broadening
is eliminated (though the second-order isotropic effect remains). There are many
pairs of angles that can be used, but none eliminate shielding anisotropy or dipolar
interactions, so resolution remains a problem. The best pair of angles is 0° and 63°
(which requires k = 5) , since the latter is not far from the magic angle (so that in t 2
the first-order anisotropic effects are significantly reduced). This method is also
technically demanding, since it requires rapid changes of rotor angle, but (in
contrast to DOR) there is no difficulty about obtaining high spin rates because only
one rotor is involved.
6.7.4 MULTIPLE QUANTUM MAGIC-ANGLE SPINNING

The two-dimensional multiple quantum magic-angle spinning (MQMAS) approach
now forms the most popular method of studying quadrupolar nuclei, so it will be
discussed in some detail. The operation of some of the pulse sequences involved
will be introduced and some of the practical issues will be explored in depth.
As discussed in sections 4.3, 4.4.2, and 6.6.1, coherences between energy levels
differing by more than one in m I can be created by RF pulses, though they are not
directly detectable. In practice, while MQ and SQ coherences are qualitatively
similar in operation, the former do have one distinction from single-quantum
coherences, namely in the way they are affected when the phase of a pulse is
changed. Coherence order , p , is defined as the difference between the values of m I for
the energy levels linked by the coherence. Pulses change coherence orders. When
the phase of a pulse changes by ø , say, a given pulse-induced change of coherence
order, Δp , acquires a phase shift of missing image file . This is consistent
with simple ideas of single-quantum missing image file transitions induced by
an RF pulse: In this case, inverting the phase of a π/2 (x ) pulse from + x to –x will
change the phase of the resulting magnetization vector from y to –y (i.e., phase shift
of –π).
The creation of coherences other than p = ±1 is relatively simple for
quadrupolar nuclei. All that is necessary is to apply a single nonselective (but not
truly “hard”) pulse (E Q >E RF ). The number and strengths of the various
coherences change with the pulse duration (see section 6.6.1). The difference in
phase behavior between different coherences allows phase cycling to be used to
select a particular coherence order. The coherence selected can then be allowed to
freely precess. After a period of time, a second pulse can transfer the selected
coherence into observable single quantum coherence.
Such a sequence of two pulses, with a variable time between them (as
illustrated in figure 6.1 6(a)) forms the basis of the two-dimensional MQMAS
technique.7 Phase cycling is used (section 3.4.1) to select those transitions caused
by the first pulse that are symmetrical in quantum number m I , that is,
missing image file . For instance, with missing image file a triple-
quantum transition is involved (3QMAS NMR), which is the only possibility for
missing image file As for the missing image file case, all such
symmetrical transitions are unaffected by first-order quadrupolar interactions.
Figure 6.16 involves the coherence transfer pathways for the triple-
quantum/single- quantum (3QMAS) correlation experiment, which is the one most
commonly encountered. The two pathways ( p = 0 → +3 → –1 and p = 0 → – 3 → –1 ) can
be combined to give a signal that is amplitude-modulated (with respect to t 1 ), which,
with a two-dimensional Fourier transformation, gives (under the right conditions;
see later in this subsection) a pure absorption lineshape.8 The phase cycle for this
experiment is chosen to achieve this. (See section 3.4.1 for a general discussion of
phase cycling.)
missing image file

Figure 6.16 . The pulse sequence for (a) a two-pulse and (b) a three-pulse, z -filtered MQMAS experiment along
with the coherence transfer pathways for triple-quantum experiments. t 1 is the usual evolution time in a two-
dimensional experiment and data acquisition starts in t 2 after a dead-time delay following the final pulse. In (b),
the third pulse is a low-power selective pulse (on the missing image file transition). With a phase-cycle to
select the appropriate coherence transfer, the delay t simply needs to be long enough for the spectrometer to
implement the large change in RF power level between the second and third pulses.
The effect of the first pulse depends on the size of the quadrupole coupling, so
it is sample and environment dependent. The implications of this, for samples
containing more than one environment, will be discussed later. After an evolution
time, t 1 (which can be rotor synchronized although this is not essential), the ±3
coherence order is converted to an observable –1 coherence order with the second
pulse. As usual for two-dimensional experiments, FIDs are collected over a time t 2
for various values of time t 1 to produce a two-dimensional data set involving two
time variables. Double Fourier transformation then gives a two-dimensional plot
involving two frequency dimensions, f 2 and f 1 , respectively. The direct dimension
(frequency f 2 ) may be labeled MAS or 1Q since a projection onto this axis results
in the normal one-dimensional spectrum. The indirect dimension (frequency f 1 ) is
labeled MQ (or, specifically, 3Q, 5Q, etc., as appropriate). The two coherences ( p =
±3) selected by the first ( excitation ) pulse are produced with equal efficiencies.
However, the second ( reconversion ) pulse, which is arranged to give observable p = –
1 coherence, acts unsymmetrically, that is the effectiveness of the
missing image file and missing image file coherence transfer steps is
different. If the duration of this pulse is chosen badly, then there will be an
imbalance between the signals generated by each pathway, which results in a poor
final lineshape. Conversely, for the missing image file case only, if the
duration of the pulse is chosen so that the two transfer steps produce an equal
amount of signal then a pure absorption lineshape will result. For higher spins, a
pulse duration can be selected to give a good approximation to a pure-phase
spectrum, as shown in figure 6.17 (b)9 for a spin - missing image file case.
missing image file
Figure 6.17 . Triple-quantum 27 Al MQMAS spectra from aluminum acetylacetonate. The spectra were
obtained at 9.4 T, with an excitation-pulse duration of 3.6 μ s at an RF field equivalent to 110 kH z. The
reconversion pulse durations were 0.8 and 2.0 μ s for (a) and (b), respectively. (a) Contour plot illustrating the
poor lineshape (containing a large dispersive component) obtained when the two coherence transfer pathways
are combined unequally. The dashed lines represent negative intensity. (b) The same experiment but with
coherence transfer pathways combined approximately equally.
The resonance shown in figure 6.17 (b) does not lie parallel to the axis in the
directly detected dimension. This occurs because the second-order effects influence
the spectra in both dimensions. A process known as shearing is used to compensate
for this so that the resonance will then lie parallel to the MAS dimension (figure
6.18 ). This matter was not mentioned when multidimensional spectra were first
introduced (in chapter 5 ) because it is rarely important for spin-
missing image file spectra. The Fourier transform routines of the
spectrometer cope with shearing automatically. (This matter is described in detail in
the review articles listed in the Further reading section.) After shearing, one of the
axes is of mixed 1Q and MQ character. It has become common to label this as the
“isotropic” axis, though this is potentially confusing (see later in this subsection).
missing image file
Figure 6.18 . The triple-quantum 27 Al MQMAS spectrum from aluminum acetylacetonate , as shown in Figure
6.17 (b) but after shearing has been applied.
The possible imbalance between the two coherence pathways, mentioned

above, gives rise to practical difficulties. In particular, for more complex systems
containing species with differing quadrupolar interaction strengths, it may not be
possible to produce pure absorption lineshapes for all the resonances in the
spectrum. Modifying the experiment, so that it includes a z-filter (see Further
reading), can solve this problem. The experiment, with its coherence transfer
pathway is shown in figure 6.16 (b). The second pulse in this sequence is used to
create a period with coherence order zero, while the third pulse (a selective π /2 )
finally produces observable p = –1 coherence. Because the Δ p = ± 3 coherence
transfer steps are now equivalent, the experiment produces pure absorption
bandshapes for all environments. The result, with the exception of some loss of
signal, is as illustrated in figure 6.18 . Note that a shearing transformation is still
required.
The two-dimensional character of plots such as figure 6.18 enables good
resolution to be obtained. The f 1 “isotropic” dimension gives a high-resolution
spectrum free of second-order effects. The position of the resonance along this axis
involves the strength of the quadrupole interaction (as well as being dependent on
the chemical shift). The f 2 bandshapes (slices in the “MAS” dimension), which
retain the second-order broadening (though in practice it may be somewhat
distorted), can be computer-fitted to extract the quadrupole coupling parameters and
isotropic chemical shifts for each resonance. While, in practical terms, the MQMAS
experiment is, perhaps, superfluous on a system with only one environment (except
to prove that this is the case when the bandshape is not so clear-cut as it is here),
there is potential for improving resolution in more complex systems where there is
an overlap of bandshapes in a one-dimensional spectrum. The utility of the
MQMAS experiment for deciphering complex spectra consisting of overlapping
bandshapes is illustrated in figure 6.19 . However, because the effectiveness of the
coherence transfer steps in this experiment are potentially different for different
environments within a sample, the MQMAS experiment cannot be considered to be
generally quantitative.
The practical steps required to successfully carry out an experiment of the type
discussed above are described in inset 6.4. Referencing the indirectly detected
(isotropic) axis is discussed in inset 6.5. Numerous other experiments purport to
improve the experiment. Some are classed as shifted-echo experiments (but may not
work if T 2 > is short), others are known as split-t 1 experiments and these remove the
need for shearing. Further pulse sequence modifications can be made to generally
improve sensitivity. One method involves redistributing the population of the spin
energy levels such that the population difference between the central
missing image file pair is enhanced by a process known as rotor-assisted
population transfer (RAPT). Details can be found in Further reading. High RF fields
(>100 kHz) are an advantage for any MQMAS experiment and, as is the case for
quadrupolar nuclei in general, so are high magnetic fields. The review article by
Goldbourt & Madhu (see Further reading) provides extensive detail on the MQMAS
class of experiments and includes sensitivity enhancement schemes.
missing image file
Figure 6.19 . Sodium-23 triple-quantum MQMAS spectrum from a mixture of polymorphs of sodium
tripolyphosphate (Na5 P3 O10 ). A three-pulse z -filter experiment was used at 9.4 T with pulse durations of 4.8
and 2 μ s (at an RF field equivalent to 100 kHz) for the first two pulses. The selective third pulse had a duration
of 12 μ s (at an RF field equivalent to 10 kHz). The delay between the second and third pulses was 3 μ s. The
recycle delay was 2 s and 48 (complex) increments in t 1 , with 192 repetitions for each, were utilized. The
spectral width in the isotropic dimension was 20 kHz (before scaling; see inset 6.5) and the sample spin rate was
10 kHz. Note that spinning sidebands (ssb) can occur in the isotropic dimension just as they can in the MAS
one. Cross-sections through each resonance are shown on the right.
Inset 6.4. MQMAS in practice
The practical steps necessary to set up a successful three-pulse, z -filtered 3QMAS

experiment illustrate some of the features that are specific to quadrupolar nuclei and
multiple-quantum excitation. Here, a step-by-step approach (see figure 6.20 ) to this set-up
23
is described for a spin - case (i.e., Na). The appearance of signal height vs. pulse
duration plots will be sample ( missing image file ) dependent. The system under study
in this illustration contains multiple sodium sites and the largest missing image file value
is ~2 MHz. The magnetic field is 9.4 T.
(1) Using a standard solution (0.1 M NaCl in the case described here) calibrate a 90 ° 23 Na
pulse, at the highest RF power available within the limits of the probe. Reference the zero
point on the chemical shift axis (see section 3.3.8).
(2) (Optional) On the sample of interest determine the response to a single pulse at the
higher RF power (figure 6.20 (a)). Note the shape of the nutation curve compared with that
for the spin - missing image file case shown in figure 3.5. The 90 ° pulse duration on the
solution at this RF field was 2.5 μ s. If not already known, determine an appropriate
recycle delay.
(3) Carry out a pulse calibration at low RF power and note the observed (or inherent ) 90 °
pulse duration (figure 6.20 (b)). This selective pulse behaves more like the spin -
missing image file case.
(4) Using a three-pulse z -filter pulse sequence (figure 6.16 (b)), with t 1 = 0 , set the third
pulse duration to be the low-power 90 ° calibrated in step (3) (8 μ s here). A good starting
value for the second pulse angle is 60 ° at the high RF power (relative to the 90 °
determined on the solution). Vary the duration of the first pulse and note the value that
gives maximum signal (figure 6.20 (c)). The “signal height” scale can be compared with
that in figures 6.20(a) and 6.20(b). In this case the experiment now gives about 10% of the
maximum signal height obtained in response to a single pulse. If the signal intensity is low,
change the duration of the second pulse and repeat the experiment.
(5) Set the duration for the first pulse to the value determined in step (4), then vary the
duration of the second pulse and again note the maximum (figure 6.20 (d)).
(6) (Optional) Using the values determined from the previous two steps, the duration of the
third pulse can also be optimized (figure 6.20 (e)).
(7) The final step is to choose a spectral width for the indirect dimension. This can be rotor
synchronized (although it is not essential). If it is set equal to the spin rate, it is possible
that it will not be large enough to accommodate the full spectrum (so the potential for
folded-back resonances to occur must be considered; see section 3.3.4). Choose the
number of increments in t 1 to avoid truncating the signal.
missing image file

Figure 6.20 . The steps required to set up a 3QMAS experiment.
Inset 6.5. Axis-labeling of MQMAS spectra
Often, it is the number of resonances (environments) in the indirectly detected (f 1 )

dimension of an MQMAS experiment that provides useful chemical information.
Additional information on the nature of the environments can be obtained from the shape
and position of the band in the directly detected (f 2 ) dimension. A scale for f 1 is, perhaps,
not critical but, nevertheless, one should be provided. Unfortunately, there is (currently) no
internationally recognized way of labeling the f 1 dimension in MQMAS spectra and,
confusingly, various methods are encountered in the literature.
In figure 6.17 , f 1 is a pure triple-quantum (3Q) dimension and is labeled as such.
Chemical shift values can be added in the same way as for the f 2 (single-quantum or 1Q)
axis (see inset 6.4, point (1)). This method is consistent with IUPAC recommendations but
does mean that chemical shift values on 1Q and 3Q (or 3Q and 5Q axes, etc.) axes will be
different. Here, the 1Q f 2 axis is labeled “MAS.”
After shearing (figures 6.18 and 6.19), the new f iso dimension is a linear combination
of f 1 and f 2 frequencies and is labeled “isotropic.” For a spin - nucleus, the spectral
width ( SW ) in f iso is related to that in f 1 through missing image file The
multiplication factor is missing image file for spin - missing image file , spin
- missing image file , and spin - missing image file nuclei, respectively.
Shifts in the isotropic dimension are both chemical and quadrupolar in origin (see
section 6.4.2), so the isotropic chemical shift cannot be read directly from this axis. The
upshot of this is that the isotropic chemical shift can still only be determined properly by
simulating the bandshape or by determining its center-of-gravity (equation 6.15)—either
from the one-dimensional spectrum obtained with a small pulse angle or from the sections
through an MQMAS spectrum.
6.7.5 SATELLITE TRANSITION MAGIC-ANGLE SPINNING
Up to this point, much emphasis has been placed on the central transition. However,
as stated earlier, the complete static spectrum or the intensities of the satellite
transition spinning sideband manifolds (in particular) may be computer-fitted to
determine the quadrupolar parameters. Inset 6.6 gives the basic theory for the
satellite transitions. It shows that it is possible to identify the true chemical shift by
locating the various centers of gravity. This can be a simple process for spin-
missing image file if the inner and outer satellite peaks can be resolved. In this
case, the isotropic quadrupolar shifts for the outer, inner, and central transitions are
in the ratio 28 : 1 : –8 (figure 6.21 ), with a constant factor:
In order to achieve spectra of the type shown in figure 6.21 , the magic angle must
be accurately set and rotor synchronization must be used—or else all the spinning
sidebands must be carefully added into the centerbands (as was done in the case of
figure 6.21 ).
The center of gravity of the inner satellites is actually very close to the true
isotropic chemical shift. Moreover, the second-order broadening of transitions is
proportional to the relevant A 4 (I , mI ) coefficient listed in table 6.2 . For spin -
missing image file , these show that the widths for individual spinning
sideband signals for the outer, inner, and central transitions are in the ratio:
(The negative sign implies a reversal of the shape from that shown in figure 6.9 .)
The inner satellite lines are appreciably sharper than the others.
missing image file

Figure 6.21. Aluminum-27 ( missing image file ) m agic-a ngle spinning ( MAS) spectrum of aluminum
acetylacetonate, obtained at a magnetic field of 9.4 T. All the spinning sidebands have been folded into the
displayed centerbands. The outer (ST 2 ), inner (ST 1 ), and central (CT) transitions are well resolved and their
differing quadrupolar isotropic shifts and bandwidths are clearly shown. Quadrupolar parameters: χ = 3.0 MHz,
η = 0.16.
Inset 6.6. Satellite transitions
Equation 6.14 may be expanded to the general missing image file case:
where the spin-dependent coefficients are given (for spin- missing image file and spin-
missing image file nuclei) in table 6.2 . The above equation assumes axial symmetry.
Table 6.2. Spin coeffi cients for single-quantum transitions (see equation 6.19)
Table Missing
The analogous equation to 6.19 for sample spinning can be obtained by replacing P 2 (cos θ )
and P 4 (cos θ ) by P 2 (cos β ) P 2 (cos Θ ) and P 4 (cos β ) P 4 (cos Θ ) , respectively (as in going
from equation 6.14 to equation 6.16). Thus, under MAS conditions, equation 6.19
becomes, for the missing image file transition:
missing image file 6. 20
For the specific case of the missing image file satellite transitions, the quadrupolar shift
is:
Since the A 0 (I , mI ) coefficients differ for the central and satellite transitions, so will the
centers of gravity.
Clearly it would be useful to separate the different subspectra. This can be

achieved by a two-dimensional experiment that correlates the satellite transitions to
the central transition. This is usually referred to as the satellite transition magic-
angle spinning (STMAS) experiment. To understand this experiment, one needs to
realize that a nonselective (but not truly “hard”) pulse will convert any particular
coherence into all the other coherences (see section 6.6.1). Consider then a series of
two such pulses. The first will create single-quantum coherence (for both the central
transition and the satellite transitions). Such coherences will evolve during the
interpulse time t 1 . The second pulse will spread these coherences around further.
Thus, a satellite single-quantum coherence created by the first pulse will be partially
converted into central-transition single-quantum coherence by the second pulse.
Detection of the eventual FID (over time t 2 ), followed by double Fourier
transformation over t 1 and t 2 , results in a two-dimensional spectrum that correlates
the different coherences in the two times.
As with MQMAS, there are many versions of the relevant pulse sequence, each
with its special characteristics. One possibility is the z -filtered experiment shown in
figure 6.22 . Again as with MQMAS (figure 6.16 ), this involves two nonselective
pulses and a soft pulse. A major difference lies in the phase cycling, which, for
STMAS, selects Δ p = ± 1 by the first pulse. The time t 1 must be accurately rotor-
synchronized. The second pulse converts the magnetization into the z direction and
the final pulse results in selective detection (by recording an echo) of the central
transition. The second-order effect is refocused at the center of the echo, the
appearance of which depends on time t 1 . After Fourier transformation, the resulting
two-dimensional spectrum contains separate ridge-lines for CT–CT and ST–CT
correlations.
missing image file

Figure 6.22 . Pulse sequence and coherence transfer pathway for a z -filtered satellite transition magic-angle
spinning (STMAS) experiment.
An experimental example ( missing image file ), for the 87 Rb resonance

of rubidium nitrate, is shown in figure 6.23 . Suitable processing has resulted in
horizontal lines for the ST–CT transitions along the f 2 dimension so that projection
onto the f 1 dimension will give a high-resolution (isotropic) spectrum free from
second-order quadrupolar effects. There are three crystallographic sites. The
quadrupole coupling constants for the signals at –27.4, –28.5, and –31.3 ppm are
1.68, 1.94, and 1.72 MHz, with asymmetries 0.2, 1.0, and 0.5, respectively.10 The
two giving signals at higher frequency are marginally resolved in the two-
dimensional plot but not at all in the projection onto the ST dimension.
missing image file
Figure 6.23 . Satellite transition magic-angle spinning (STMAS) 87 Rb spectrum of rubidium nitrate after
shearing. Operating parameters: the first two pulses were of duration 2.3 μ s and angle 90° (as measured for an
aqueous solution). The RF power was equivalent to 108 kHz. The z -filter pulse was of duration 6.0 μ s at a 14
kHz power equivalent. The spin rate was 10 kHz and the spectral width was 10 kHz in both dimensions.
While STMAS is somewhat more sensitive than MQMAS, it is significantly
more demanding to set up. The magic angle must be set precisely and the spin rate
must be highly stable.
6.7.6 SUMMARY FOR SPECTROSCOPY OF HALF-INTEGER

QUADRUPOLAR NUCLEI
Obtaining useful spectra of quadrupolar nuclei is more of a challenge than is the

case for spin- nuclei. The choice of experiment must be tailored to the situation in
question. When quadrupole coupling constants are small or when there is only one
crystallographic site for the nucleus of interest, simple MAS operation may suffice,
but when second-order effects are significant and resolution of multiple sites is
required, it may be useful to employ one of the special techniques described above,
though these do not work satisfactorily for substantial quadrupole coupling
constants (> 8 MHz, say). The use of DOR or DAS is limited by the requirement for
special hardware, so in most cases it is necessary to turn to either MQMAS or
STMAS. The spectrometer manufacturers supply suitable software for these
experiments, so most users will be well advised to follow the relevant instructions.
However, some understanding of the principles given in section 6.7 will be essential
to the interpretation of the spectra obtained. Detailed analysis will, in many cases,
involve fitting bandshapes using standard computer programs. It will frequently be
necessary to use both quadrupolar and anisotropic chemical shift parameters in such
fitting. An ability to obtain spectra at different applied magnetic fields will be a big
advantage in this process. Moreover, when large quadrupole coupling constants are
involved, the use of the highest possible field will be preferred.
6.8 SPECTRA FOR INTEGRAL SPINS
Mention was made in section 2.6 that there are only four nuclides of significance for
NMR with integral spin quantum numbers. Of these, 10 B has I = 3, while the others
(2 H, 6 Li, and 14 N) are spin-1. Boron-10 NMR is unimportant because 11 B is the
preferred boron nuclide. Lithium-6 has a low quadrupole coupling constant and can
usually be treated as a “pseudo spin- missing image file ” nuclide—it provides a
useful foil to the spin- missing image file nuclide 7 Li. Consequently, only
deuterium and nitrogen-14 need to be considered in any detail here.
Quadrupole coupling constants for 2 H are generally modest (< ~300 kHz), so
second-order effects are not a problem and methods analogous to those used for
spin- nuclei generally suffice (though the spin dynamics is different). An example
has been given as figure 6.2 . Static bandshapes and spinning sideband manifolds in
2 H spectra frequently reflect local molecular-level mobility and are much used in
that context. Figure 6.24 shows an example (a 2 H MAS spectrum). The spinning-
sideband manifolds for finasteride hydrate THF solvate cover only ~35 kHz,
whereas the expected quadrupole coupling constants for a static molecule are ~200
kHz. There must be rapid (though not isotropic) motion of the THF molecule in the
solvate. However, the quadrupole parameters for the two sites are still unequal—for
the high-frequency peak χ = 20.2 kHz and η = 0.99, whereas for the low-frequency
peak χ = 16.4 kHz and η = 0.93.
Nitrogen-14 is something of a problem nucleus because there is no central transition
and quadrupole coupling constants can be significant (e.g., ~3 MHz)—sufficiently
high that MAS results in innumerable spinning sidebands and thus a loss of S/N.
Various schemes have been mooted to alleviate this situation, but there is no widely
used protocol yet.
missing image file

Figure 6.24 . Deuterium magic-angle spinning (MAS) spectrum at 11.7 T of finasteride hydrate THF solvate
with fully deuterated THF. The expansion of the centerband shows the two different chemical shifts. The spin
rate was kept low (~1 kHz) in order to show the complete spinning sideband manifolds, which illustrate that the
two chemically distinguishable sites have different quadrupole coupling constants.
FURTHER READING
GENERAL TEXTS ON QUADRUPOLAR NMR
“Quadrupole effects in NMR studies of solids”, M.H. Cohen & F. Reif, Solid State Physics , 5 (1957) 321–438.
“Quadrupole effects in solid-state NMR”, D. Freude & J. Haase, in NMR basic principles and progress , Eds. P.
Diehl, E. Fluck, H. Günther, R. Kosfeld & J. Seelig, Vol. 29 , Springer (1993) 1–90.
TEXTS ON SPECIFIC ASPECTS OF QUADRUPOLAR NMR

“Nutation spectroscopy of quadrupolar nuclei”, B.C. Gerstein, in “Encyclopedia of NMR” , Eds. D.M. Grant,
R.K. Harris & R.E. Wasylishen, John Wiley & Sons Ltd. (online posting date 15 March 2007), DOI:
10.1002/9780470034590.emrstm0359.
“Solid-state NMR line narrowing methods for quadrupolar nuclei: Double rotation and dynamic-angle
spinning”, B.F. Chmelka & J.W. Zwanziger, in NMR basic principles and progress , Eds. P. Diehl, E. Fluck,
H. Günther, R. Kosfeld & J. Seelig, Vol. 33 , Springer (1994) 79–124.
“Multiple-quantum magic-angle spinning: High-resolution solid-state NMR spectroscopy of half-integer
quadrupolar nuclei”, A. Goldbourt & P.K. Madhu, Monat. Chem. , 133 (2002) 1497–1534. DOI:
10.1007/s00706-002-0502-y.
“High-resolution NMR of quadrupolar nuclei in solids: The satellite-transition magic angle spinning (STMAS)
experiment”, S.E. Ashbrook & S. Wimperis, Prog. NMR Spectry. , 45 (2004) 53–108. DOI:
doi:10.1016/j.pnmrs.2004.04.002.
NOTES
1 The existence of a quadrupole moment corresponds to a nonspherical (but axially symmetrical) distribution of
positive charge within the nucleus. If the charge is greater along the spin axis than perpendicular to it, the
distribution is prolate. If the reverse is true, the quadrupole moment is oblate.
2 Alternatively, the symbol C can be used for the quadrupole coupling constant, though χ is recommended by
Q
IUPAC.
3 T.A. Wagler, W.A. Daunch, M. Panzner, W.J. Youngs & P.L. Rinaldi, J. Magn. Reson. 170 (2004) 336.
4 These ratios are correct only for nonselective pulses (see subsection 6.7.4).
5 Since the rotor and laboratory frames involve unique directions, only the unique angle between them is
involved in changing operators between these frames.
6 For a discussion of the conditions for selective excitation, see the article by Freude and Haase in Further
reading.
7 Actually, there are innumerable variations in the pulse sequences used for MQMAS, giving different
efficiencies and having various advantages and disadvantages. The subject is not for the faint-hearted! A
detailed evaluation is to be found in the article by Goldbourt and Madhu (see Further reading).
8 An alternative approach is to select only one pathway and generate a phase-modulated signal with respect to t
1 . See the review articles in Further reading for more information on this class of experiment.
9 In some articles, the scale in the 3Q dimension is divided by three to make it comparable to the scale of the
MAS dimension. However, this practice is not followed in comparable spin- 2D spectra and we believe it is
unhelpful.
10 D. Massiot, B. Touzo, D. Trumeau, J.P. Coutures, J. Virlet, P. Florian & P. J. Grandinetti, Solid State NMR 6
(1996) 73.
CHAPTER 7
RELAXATION , EXCHANGE & QUANTITATION
7.1 INTRODUCTION
It is important not to picture a solid as a rigid entity but rather as one in

which there is motion, driven by thermal energy, albeit on a more restricted
scale than in solution (Figure 7.1 ). Motion is the driving force for
relaxation and it can have an impact on the appearance of the NMR
spectrum.
missing image file
Figure 7.1. Examples of the timescales of motions in the solid state and the
NMR properties that can be used to explore t hem.
This chapter starts by exploring the relationship between motion and

relaxation. It is shown how the relaxation behavior can be used to deduce
information on sample morphology and how relaxation data can be
combined with high-resolution measurements to carry out spectral editing.
Methods for measuring relaxation times are described.
Section 7.3 shows how motion in the form of reorientation or, for example,
exchange of hydrogen, can affect the appearance of a spectrum.
Measurements designed to determine rates of motion are introduced there.
It is demonstrated how the anisotropic interactions important in solid-state
NMR (shielding/chemical shift anisotropy (CSA), dipolar coupling,
quadrupolar coupling) are modified by anisotropic motion and how
experiments can be used to probe details of this motion.
As motion affects relaxation and relaxation influences the intensity of the

resonances in the spectrum, section 7.4 of the chapter deals with
quantitative aspects of solid-state NMR. Some of the special effects seen in
paramagnetic systems are discussed in section 7.5 at the end of the chapter.
7.2 RELAXATION
Three modes of relaxation important to solid-state NMR were defined in
section 2.8. Spin–lattice (or longitudinal) relaxation, characterized by the
time constant T 1 , describes the process of restoring equilibrium to the z
component of the net magnetization for an ensemble of spins.1 Spin–spin
(or transverse) relaxation, with time constant T 2 , relates to the xy
component of the magnetization. Spin–lattice relaxation in the rotating
frame, with a time constant T 1 ρ , describes the return to equilibrium of
transverse magnetization in the presence of spin-locking from an RF
magnetic field.
7.2.1 THE CAUSE OF RELAXATION
Relaxation in NMR, unlike the case for most other spectroscopies, is non-
radiative . That is, the dominant cause of relaxation for the sample under
study is not by emission of electromagnetic photons. Instead, fluctuations
in the local field at the excited nucleus allow it to relax. Molecular motion
causes this fluctuation (see inset 7.1). In the solid state, the main
contributions to the local magnetic field arise from shielding anisotropy and
from dipolar coupling to the magnetic nuclei of surrounding atoms. Due to
its magnitude, the quadrupolar interaction, where present, often gives rise
to very efficient relaxation, though in this case it is a fluctuation in the local
electric field that is effective.
Inset 7.1. When solids lack motion

That relaxation is related to motion is illustrated by considering the spin–
lattice relaxation time in systems where there is little or no motion. Silicon
carbide exists as a rigid, 3D network of silicon and carbon atoms. The
silicon T 1 in one polymorph has been measured as 35 minutes! In such
systems, the only relaxation mechanism is often via paramagnetic defects.
So relaxation may be extremely slow in high-purity network (3D)
materials.
A full treatment of relaxation is not something to be undertaken lightly!

The analysis of anything more than an isolated spin pair soon gets
complicated. A detailed understanding of the relaxation of a real system,
with many interacting nuclei and a number of degrees of freedom for
motion, is not generally feasible. Nevertheless, facets of the relaxation
behavior can usefully be related to the properties of the system under study.
Relaxation at a frequency ν (which might be the resonance frequency of the
protons in a sample, for example) is most efficient when the molecular
motion results in a fluctuating magnetic field of the same frequency. A
quantity called the spectral density , J (ν ),2 plays a key role in the theory of
relaxation (see inset 7.2). It can be written as:
Inset 7.2. Spectral density

In a solid, random motion produces fluctuations in the local magnetic field.
These fluctuations can be characterized by a time correlation function, G ( t
), which describes the correlation between the average local field at a
moment in time and at a time t later. In effect, this is a measure of the
persistence of the local field. An unchanging (correlated) field or one that
changes very quickly (uncorrelated) provide poor relaxation pathways.
Often, the value of the correlation function is large over short times and
decays exponentially with a time constant τ c , the correlation time, as time
increases, so:
(The proportionality constant, omitted for simplicity here, depends on the

magnitude of the local field.) The spectral density is the Fourier transform
of the correlation function:
so that, with G ( t ) defined as it is here, and ignoring the proportionality

constant, J ( ν ) is given by equation 7.1.
where τ c is called the correlation time and is the time constant that
characterizes the motion in question.3 Short correlation times imply fast
motion (as for liquids and solutions) and long ones slow motion (as for
rigid solids).
A plot of J ( ν ) against τ c puts equation 7.1 into context—Figure 7.2 . The

higher the value for J ( ν ), the more efficient the relaxation (as equations
7.4 and 7.5 will illustrate). Efficient spin–lattice relaxation requires
fluctuations in the local field at frequencies of hundreds of megahertz
(Larmor frequencies) and this equates to correlation times on the order of
nanoseconds. On the other hand, spin–lattice relaxation in the rotating
frame requires fluctuations of tens of kilohertz, equating to correlation
times on the order of microseconds. Also it can seen from Figure 7.2 that a
fluctuation in the local field at a typical RF nutation frequency (62.5 kHz,
curve (c)) can be more effective at causing relaxation than a fluctuation at a
Larmor frequency (curves (a) and (b)). This is why T 1 ρ is often much
shorter (usually milliseconds) than T 1 (often seconds) for the solid state.
missing image file
Figure 7.2. A log–log plot of the spectral density, given by equation 7.1, as
a function of correlation time for frequencies of: (a) 400 MHz, ( b) 800 MH
z , and (c) 62.5 kHz. For a given value of ν , J ( ν ) reaches a maximum
when 2π ντ c = 1 . For very slow motions (long correlation times),
indicated by the gray box, “classic” relaxation theory breaks down.
For a single, simple motion, the relaxation rates encountered in NMR can
be considered as being proportional to a linear combination of spectral
densities. For example, suppose the dominant relaxation mechanism is a
result of homonuclear dipolar coupling between spin- missing image file
nuclei, as might be the case for the protons in an organic solid. Under these
conditions, the relaxation times for a single spin species relate to the
spectral densities as follows:
The constant K depends on the magnitude of the dipolar coupling. This

constant is different for other relaxation mechanisms, as are the coefficients
of the individual terms. T 2 –1 can be expressed in a similar way but this is
rarely valid in the solid state so the equation is not given here and T 2 is
omitted from Figure 7.2 and Figure 7.3 (it is discussed later in section
7.2.3).
missing image file

Figure 7.3. A log–log plot of the relaxation times as a function of
correlation time. The curves are calculated for frequencies of: (a) 400 MHz,
(b) 800 MHz, and (c) 62.5 k Hz. (a) and (b) are appropriate for T 1 given by
equation 7.4 and (c) is for T 1 ρ given by equation 7.5.
Two limiting conditions for J (ν ) are important in NMR. If 2 π ντ c << 1

then missing image file and the relaxation rates are proportional to τ c .
This is called the extreme narrowing condition and applies for fast
molecular motion. At this condition T 1 = T 2 = T 1 ρ , which is the reason
that usually T 1 = T 2 in solution.
The opposite end of the scale is particularly important in solid-state NMR.

This is the rigid lattice limit. Here 2 π ντ c ≫ 1 so that J (ν ) becomes
proportional to missing image file (and so do the relaxation rates).
However, for very slow motions (say, with correlation times greater than
0.1 ms) this approach to interpreting relaxation behavior is no longer
appropriate.
For a given relaxation mechanism, the behavior of the relaxation times, as a

function of correlation time, defined by equations 7.4 and 7.5, is shown in
Figure 7.3 . Several observations can be made:
T 1 in the rigid lattice limit is proportional to the magnetic field of the

spectrometer.
T 1 ρ is independent of the magnetic field of the spectrometer (because
J (ν RF ) is the dominant term in equation 7.5); however, it depends on
the strength of the RF magnetic field.
The T 1 and T 1 ρ curves go through a minimum. At the minimum 1/ τ
c = 2 π ν 0 or 2 π ν RF for T 1 and T 1 ρ , respectively (when a single
spectral density is involved).
Energy of Activation
For a thermally activated process, the activation energy,

missing image file , can be obtained from an Arrhenius plot of the
correlation time as a function of temperature:

where R is the gas constant, T is the temperature, and A is a constant (the
pre-exponential factor). In the extreme narrowing limit, it follows that ln(T
–1 ) = ln(A )–E /RT , so a plot of missing image file against T -1 will
1 a
have a gradient proportional to the activation energy. The result in the rigid
lattice limit is the same, apart from a change in the sign of the gradient.
This applies equally well to T 1 ρ . An example of such a plot is shown in
Figure 7.4 . The correlation time can also be written in an Eyring form :
where missing image file is the enthalpy of activation,

missing image file is the entropy of activation, k B is the Boltzmann
constant, and h is Planck’s constant. Plotting ln( T 1 / T ) (in the extreme
narrowing limit) or ln(1/ T 1 T ) (in the rigid lattice limit) against 1/ T will
yield the thermodynamic parameters.
In general, the relaxation behavior is much more complex, since it contains

contributions from all of the motional processes present in the solid. For the
case shown in Figure 2.14 , for example, the molecule experiences two
types of motion so two minima are observed in the T 1 relaxation for the
relevant temperature range. In a complex system it can, therefore, be
difficult to interpret the relaxation behavior, so that relating a single T 1 or
T 1 ρ determination back to a particular molecular motional property is
often not feasible although such information can be used to design a filter
for more complex measurements (see section 7.2.4). Recording relaxation
times as a function of temperature can give information on activation
barriers and that may give some clue as to the nature of the molecular
dynamics. Such an approach may also highlight physical changes to the
sample under study, such as a glass transition (the point at which a polymer
changes from a rigid glass-like state to a rubbery one, thus giving an abrupt
change in T 1 ), even if a detailed interpretation is not forthcoming.
missing image file

missing image file
Figure 7.4. An Arrhenius plot of the carbon spin–lattice relaxation rate for
the methyl group labeled (a) in ibuprofen. The gradient of the fitted line
gives an activation energy for methyl group rotation (assuming this is the
cause of the relaxation) of 7.9 kJ mol–1 .
Which relaxation times, then, are useful in solid-state NMR? In principle,

they all are but some are easier to measure than others. For spin -
missing image file nuclei (other than 1 H), solid-state T 1 values tend to
be long (often tens of seconds) and measuring them is time consuming.
However, if they are measured through a high-resolution spectrum (see
section 7.2.5) and can be associated with signals from specific nuclear
species present in that spectrum, they may provide localized dynamic
information. T 1 ρ values can be tens of milliseconds and, while they can
provide useful information on slower motions, care has to be taken in
measuring them (long, high-power spin-lock pulses are required to measure
long T 1 ρ times, and these can damage probes). T 1 relaxation times for
quadrupolar nuclei vary; they can be many seconds but they are frequently
very short (milliseconds). Proton relaxation times are the easiest to measure
and these are the subject of the next section.
7.2.2 PROTON RELAXATION TIMES
These can be measured directly, often from a non-spinning (static) sample,

or they can be obtained indirectly using a high-resolution cross-polarization
(CP) method (see section 7.2.5). Spin diffusion (see inset 7.3) in a tightly
coupled network of protons is very efficient, so proton relaxation times in
solids are rarely localized to functional groups within a molecule. Usually a
molecule-wide, average relaxation time is obtained for homogeneous
samples, but it may be possible to relate relaxation times to specific
components in the system for heterogeneous materials (see below). Despite
the limitations, proton relaxation times can yield useful information, as
illustrated here with a set of examples.
Relaxation times in polyethylene
As polyethylene is formed of just a long chain of –CH2 – units (in the

absence of branching), it might be expected to show relatively simple
relaxation behavior. However, samples are usually semicrystalline (i.e.,
they contain crystalline and amorphous domains), so proton relaxation can
exhibit complicated behavior. The usual starting point for proton relaxation
studies is the measurement of T 1 . This will be useful for several purposes,
including determination of the recycle delay for the experiments discussed
later in this section and in section 7.2.4. Sensitivity is not an issue, so the
simplest approach is to make measurements on a static sample. The result
of such a measurement on a commercial polyethylene sample is shown in
Figure 7.5 (a). Plotted on a log scale, the behavior of the signal is linear,
that is, a single time constant derived from equation 7.9 describes the T 1
relaxation.
Inset 7.3. Spin diffusion and heterogeneous materials

Spin diffusion is a consequence of dipolar coupling. In a network of
dipolar-coupled spins, the redistribution of magnetization from a locally
excited site to distant sites can be viewed as a diffusion process. However,
the details are situation dependent—the 1D lamellar structures of some
polymers need a different model to the 3D case of a molecular crystal, for
example. Nevertheless, the conclusions are qualitatively the same. For the
3D case, the mean-square distance traversed via spin diffusion (assumed to
be isotropic) in a time t , missing image file is given by:
where D is an isotropic spin diffusion constant. This parameter depends on

the molecular-level mobility in the system. For the case of spin diffusion
between protons in proton-rich, relatively rigid systems, it is of the order of
10–15 to 10–16 m2 s–1 ; this will decrease as mobility increases. The
relevant time, t , over which the impact of spin diffusion needs to be
considered, is the relaxation time, that is, T 1 or T 1 ρ . For solids, T 1 >> T
1 ρ so spin diffusion has an influence over greater distances for T 1 than it
does for T 1 ρ . In this context, it is useful to talk about domains within a
material that are either chemically or morphologically distinct. Thus, in
heterogeneous materials containing such domains, spin diffusion will lead
to average T 1 values over larger domain sizes than T 1 ρ .
Suppose proton measurements are made on a material where D = 5 × 10 –

16 m 2 s –1 such that T 1 = 1 s and T 1 ρ is found to have two components:
5 and 30 ms. Substituting these values into equ ation 7.8 gives an estimate
for the limiting size of the domains. From T 1 ρ it is ~10 nm. Any smaller
than this and a single average value would have been obtained. From T 1 it
is ~55 nm. Any larger than this and T 1 values attributable to the two
domains would have been obtained (unless they were accidentally
coincident).
By generalizing, on this basis, three rules of thumb can be deduced:
One T 1 and one T 1 ρ value is observed—any domains must be

smaller than 5 nm.
One T 1 but multiple T 1 ρ values are observed—the average domain
size is between 5 and 50 nm.
Multiple T 1 and T 1 ρ values are observed—the average domain size
is larger than 50 nm.
In the case of T 2 , spin diffusion is not fast enough to equilibrate

magnetization in different domains so they have a separate T 2 for all sizes
greater than the molecule.
Proton T 1 ρ values often provide extra information about the system. The
result of such a measurement on the same polyethylene sample is shown in
Figure 7.5 (b). This time, the signal decay is nonlinear and the relaxation
cannot be represented by one time constant. Such behavior is often an
indication of physical heterogeneity within a sample. For this polyethylene
sample, the T 1 relaxation is that expected of a homogeneous material but
the T 1 ρ behavior suggests heterogeneity. As discussed in inset 7.3, it is
likely that this sample is semicrystalline and that the average domain size is
between 5 and 50 nm. Analyzing relaxation data in terms of discrete
exponentials is one approach but it is not the only one, as shown in inset
7.4.
missing image file
Figure 7.5. (a) A plot from a saturation-recovery T 1 measurement on a

commercial polyethylene sample. S ( t ) is the signal after a recovery time t
and S ∞ is the signal at t = ∞ (in practice, the signal with a recovery time of
5 × T 1 ). The solid line is the least-squares fit to the experimental data (the
circles) and gives a time constant of T 1 = 0.84 s.
missing image file
(b) The results from a T 1 ρ measurement on the same sample, showing the
signal S ( t ) as a function of spin-lock time (t ). The solid line is a least-
squares fit to the experimental data. The best fit is obtained from the sum of
three discrete exponential decays with time constants of T 1 ρ = 1.2, 4.5,
and 27.1 ms. Fitting curves like this to more than two components can be
contentious as the chemical/physical significance of the additional
component(s) may be debatable! Careful consideration of the error in each
value is essential. Both (a) and (b) were obtained with a solid echo (see
inset 3.5) and S ( t ) is the intensity of the echo maximum.
Inset 7.4. Modeling relaxation behavior

Least-squares fitting of the experimental data to one or more discrete
exponential functions is a common method found in the literature for
analyzing relaxation measurements. However, the nature of some materials
is such that distributions of relaxation times are likely to be encountered.
Then, analyzing in these terms is more appropriate. The result of such a
treatment for the decay in Figure 7.5 (b) is shown here, together with the
values from the discrete component model (the vertical lines). The major
components are treated consistently but the problem of dealing with a weak
component is illustrated in the difference for the minor one. Measurement
of pore-size distribution in oil-bearing rock is one example where this type
of approach is used (the relaxation time of the oil is pore-size dependent).
missing image file
In a system with multiple relaxation time constants, each one will have a
population (the number of 1 H nuclei relaxing with that time constant)
associated with it. If there is no interaction between the domains to which
the time constants relate, then the populations will be proportional to the
number of 1 H atoms in the domains. However, if the domains interact so
that spin diffusion can occur between them, the populations of the faster
relaxing domain will be enhanced at the expense of that relaxing more
slowly and the measured populations will no longer reliably relate to the
relative numbers of 1 H atoms in each domain. In such cases, the lineshape
is a better indicator of the composition of a sample.
7.2.3 LINESHAPES & LINEWIDTHS FOR NON-SPINNING SAMPLES
The interpretation of the relaxation results discussed in the previous section

is helped by knowledge of the 1 H lineshapes observed for non-spinning
samples. Homonuclear dipolar coupling is the main contributor to the
linewidth and the bandshape for an isolated, dipolar-coupled spin pair has
already been introduced (see section 2.5). In polyethylene the interaction
between the hydrogens in each CH 2 unit might be dominant (25.1 kHz)
but these protons cannot be considered as an isolated pair and there are
many longer-range interactions that contribute to the bandshape, resulting
in a broad, featureless line (see section 5.2).4 Any motion in the system
will tend to reduce the dipolar coupling, which results in line narrowing.
This fact leads to an explanation of the 1 H bandshape for th e p
olyethylene sample discussed in the previous section and shown in Figure
7.6 (a). The bandshape has two components, one of which is broad with a
full width at half height, missing image file of around 50 k Hz. Values of
this magnitude are typical of rigid, crystalline organic materials. The
second component is narrower, missing image file kHz, which implies
that the coupling is reduced through motion. This is typically associated
with an amorphous domain (the looser structural constraints allow more
freedom for motion). So the polyethylene sample consists of crystalline and
amorphous domains. It is possible to show, by measuring the bandshape as
a function of spin-lock time, that the broad component is associated with
the longest missing image file value (Figure 7.6 (b)). The crystallinity of
the sample can be determined from the relative intensities of the broad and
narrow components. This is not necessarily a trivial exercise, as the
components of the bandshape may not fit well to simple (Gaussian or
Lorentzian; see inset 7.9) lineshape functions.
missing image file
Figure 7.6. (a) The static 1 H spectrum from polyethylene. (b) The
spectrum from the same sample obtained after spin locking the 1 H
magnetization for 18 ms. The part of the spectrum with the narrowest line
is largely lost and therefore must be associated with the shortest proton T 1
ρ value.
missing image file

So far much has been made of the properties of T 1 and T 1 ρ but little has
been said about T 2 . As has been shown, T 1 and T 1 ρ are relatively easy
to characterize but the same is not true for T 2 . For this reason, transverse
relaxation measurements tend to receive less attention in solid-state NMR
studies.
missing image file
Figure 7.7. (a) Point-by-point Carr–Purcell–Meiboom–Gill (CPMG)

measurement for T 2 . (b) missing image file measurement from the free-
induction decay (FID). (c) missing image file measurement by varying
the echo delay in a spin-echo experiment.
For a solution, T 2 is the time constant that describes the irreversible

magnetization decay in the xy plane, in the absence of the effects of
magnetic field inhomogeneity. It can be measured using the Carr–Purcell–
Meiboom–Gill (CPMG) pulse sequence (Figure 7.7 (a)), which creates a
series of echoes, refocusing the effects of the magnetic field inhomogeneity
. In solutions this decay is generally exponential. However, the observed
decay of the NMR signal (because of the effects of magnetic field
inhomogeneity) occurs on a shorter timescale than T 2 . A related quantity
missing image file (the rate of loss or dephasing of the phase coherence
produced by the excitation) is defined as the time taken for the signal to
decay to 1/e of its initial value (Figure 7.7 (b)). For a solution, the
exponential decay of the signal leads to a Lorentzian lineshape with full
width at half-height given by missing image file
For a solid the definition of a specific time T 2 is a matter for discussion.

For a rigid solid, the echo decay from the CPMG experiment is not usually
exponential (see Figure 7.8 ) so a value for T 2 is not forthcoming in the
same way as it is in solution. However, a value of missing image file can
be measured as it is in solution. For a Lorentzian lineshape, it relates to the
linewidth as it does for a solution, or for a Gaussian shape (see section 7.4)
the full width at half-height is given by missing image file So a short
missing image file (and that can mean less than a few tens of
microseconds) implies a broad line (a static proton line, with Gaussian
shape—often a good approximation—and a half-height width of 40 kHz,
equates to missing image file = 13 μ s). A long missing image file
corresponds to a narrow line. In solid-state NMR, a slightly different
parameter, missing image file can be obtained from the decay of the echo
intensity as shown in Figure 7.7 (c). This parameter is related to the
homogeneous linewidth (see section 5.2).
missing image file
Figure 7.8. The decay of the proton spin-echo intensity for a non-spinning,
polycrystalline sample of glycine.
7.2.4 RELAXATION & HIGH-RESOLUTION MEASUREMENTS

COMBINED
Interrupted Decoupling
The interrupted decoupling experiment (also known as dipolar dephasing,

non-quaternary suppression, and protonated carbon dephasing) exploits
differences in missing image file values to simplify (or edit or filter ) the
spectrum. It is one of the most widely used editing experiments in the solid-
state NMR study of organic materials. As shown in inset 7.5, the pulse
sequence includes a short window where the 1 H decoupling is turned off.
Briefly (often 40–50 μ s for carbon or ~200 μ s for nitrogen) turning off the
decoupling connects the 13 C (say) magnetization of protonated carbon
sites to the large proton spin bath. This drains intensity away from the
observable 13 C signal and suppresses the signal from those protonated
carbons. The magnetization from quaternary carbons (or tertiary nitrogens)
will not be suppressed to the same extent as they are only weakly coupled
to protons (they have no directly bonded hydrogens). An example of such
an experiment is shown in Figure 7.9 . Molecular motion, either of a
molecule as a whole or of part of a molecule, such as methyl group
“rotation”, tends to reduce dipolar coupling (to 1 H) and
missing image file is not reduced to the same extent as for a rigid CH
grouping, so signals from mobile species (particularly methyl groups) are
rarely fully removed from the spectrum. Fortunately, methyl groups can
usually be distinguished from quaternary carbons by their chemical shift.
This experiment is not the only means of editing spectra on the basis of
multiplicity, but it is simple and reliable.
Inset 7.5. Interrupted decoupling
The interrupted decoupling experiment is used to simplify spectra from
organic materials. The delay τ is sometimes referred to as the dephasing
delay. The rotor-synchronized 180° pulse on the X-channel refocuses the
signal, which makes the result easier to process (it avoids the need for a
large first-order phase correction). τ r is the rotor period.
missing image file

missing image file
Figure 7.9. Carbon-13 cross-polarization magic-angle spinning (CPMAS)

spectra from 3-methoxybenzene carboxylic acid recorded (a) without and
(b) with 40 m s of interrupted decoupling. (c) A difference spectrum
(showing the CH carbons) which confirms the presence of a CH carbon
signal under that from carbon 1 (marked with an arrow). There are two
molecules in the asymmetric unit (note the pairs of lines) and the * indicate
spinning sidebands.
Relaxation f ilters
Any difference in relaxation times in systems exhibiting multicomponent

relaxation behavior can be exploited to simplify (or aid the interpretation of
) spectra. Most of the pulse sequences given in section 7.2.5 potentially can
be used to do this, but two specific examples are included here using the
polyethylene example discussed earlier in this chapter. Figure 7.10 (a)
illustrates the carbon-13 cross-polarization magic-angle spinning (CPMAS)
spectrum. It shows two major lines, the narrower one at 32.8 ppm and a
broader one at about 31 ppm. Delaying the contact time, in a modified CP
experiment (see Figure 7.14 (b)), allows the proton magnetization from the
component of the material with the shortest missing image file to decay.
Subsequent CP can then occur only from that part of the material with the
longer missing image file (the rigid, crystalline component). As shown in
Figure 7.10 (b), the broad line is reduced in intensity and the narrow signal
from crystalline polyethylene dominates. (Note that it is not always the case
that a crystalline component has the longer missing image file ) This is an
example of a missing image file filter.
A filter based on missing image file produces the opposite result ( Figure
7.10 (c)). Now the crystalline component, which has the shorter
missing image file , is the one that is reduced in intensity (the experiment
used to produce this result is described in the next section).
missing image file
Figure 7.10. (a) The cross-polarization magic-angle spinning (CPMAS)

spectrum from a commercial polyethylene sample. (b) The spectrum from
the same sample but recorded with a missing image file filter (18 ms
delay before contact); (c) obtained from the sample but with a
missing image file filter (a 16 μ s delay in the sequence shown in inset
7.6).
If signals in a high-resolution spectrum can be separated in this way, then it

becomes possible to explore any interactions between the components of
the sample by adding a mixing time to the experiment (as in the EX change
SpectroscopY (EXSY) experiment described in the next part of this
chapter). Suppose that signal separation has been achieved with a
relaxation filter and the spectrum is then monitored as a function of mixing
time. If the suppressed signal(s) increase in intensity, then spin diffusion
must be transferring magnetization from one component of the sample to
another and there must be an interaction between them.5 This is generally
termed a Goldman–Shen experiment, and the result of one such
measurement, based on the filter used for Figure 7.10 (c), is shown in
Figure 7.11 . From this result, it is possible to estimate a domain size,
although this is not a trivial exercise and requires knowledge (or an
assumption) of the rate of spin diffusion, together with a model for the
shape of the domains. Note that it is usually easier to monitor an increase in
the height of a narrow line (it increases more quickly than for a broad one)
so a filter that suppresses a crystalline component is preferable.
missing image file
Figure 7.11. The carbon-13 signal height for the crystalline component of a
polyethylene sample as a function of the mixing time following a
missing image file filter. A curve of this type can be used to deduce a
domain size for the amorphous component of the sample. Signal recovery
on this timescale and with this shape correlates with a 3D amorphous
domain size of the order of ~10 nm.
1 H–X correlation
There is a simple but useful heteronuclear correlation experiment (see

section 5.4.3) that can demonstrate the existence of different mobilities
(relaxation behaviors) within a sample. This is the so-called WIdeline
SEparation (WISE) experiment (inset 7.6). The 2D experiment correlates a
high-resolution X (usually 13 C) spectrum with a low-resolution 1 H one.
Low spin rates (~5 kHz) are sufficient to generate the high-resolution X
spectrum but have little impact on the 1 H bandshape, which will resemble
the static spectrum. The result of such an experiment on the polyethylene
sample is shown in Figure 7.12 .
Inset 7.6. WISE

missing image file
The WISE pulse sequence. Spectra are acquired in 2D fashion as a function

of the delay t 1 (typically in the range 0–100 μ s).
missing image file
Figure 7.12. The result of a WIdeline SEparation (WISE) experiment on

polyethylene. The narrow 13 C signal at 32.8 ppm from the crystalline
domains correlates with a broad proton bandshape while the broader 13 C
signal (31.1 ppm) from the amorphous component correlates with a narrow
proton line.
7.2.5 MEASURING RELAXATION TIMES
The pulse sequences that are used to record the T 1 and T 1ρ relaxation
times discussed in this chapter have not yet been introduced. They are
grouped together in this section and are summarized in table 7.1.
Table 7.1. Pulse sequences for measuring relaxation times
Table Missing
Spin–lattice relaxation times
Four methods for measuring spin–lattice relaxation times are shown in

Figure 7.13 . The inversion-recovery method (Figure 7.13 (a)) gives a
potentially large signal range (from approximately fully inverted to fully
positive) but has the disadvantage (particularly for slowly relaxing species)
that the recycle delay must be at least 5 T 1 to allow the system to fully
relax between repetitions. The signal recovery is fitted to the equation
missing image file where S (τ ) is the signal intensity measured with a
recovery time τ and S 0 is the signal intensity when τ = 0 . The time at
which the signal is zero (the null time ) is useful for a quick indication of T
1 : S (τ ) = 0 when τ = ln(2) T 1 = 0.693T 1 . The rectangular 180° pulse
can be replaced by a shaped pulse for selective inversion or for the more
efficient inversion of a signal from a quadrupolar nucleus.
missing image file
Figure 7.13. Pulse sequences for measuring spin–lattice relaxation times.
The saturation-recovery sequence (Figure 7.13 (b)) starts by saturating the

spins (e.g., by setting n to 64), so there is no need to wait for the system to
return to equilibrium after the acquisition (the recycle delay can be set to
zero, though if 1 H decoupling is being used while measuring T 1 X , care
must be taken to avoid overheating the probe). The signal recovery is fitted
to: missing image file
Figure 7.13 (c) and (d) incorporate CP steps. The pulse sequence in Figure
7.13 (c) is used to measure T 1 H indirectly through an X nucleus spectrum.
In a heterogeneous sample where T 1 H is not a sample-wide average (see
inset 7.3), it may be possible to relate T 1 H values to chemically
identifiable components within the sample (this information is not available
from a wideline 1 H measurement). The sequence shown in Figure 7.13 (d)
is for measuring T 1 X . Often, the delay τ has to extend to long times but
the advantages of CP discussed in section 3.4 apply here, making this
method practical for dilute and slowly relaxing nuclei. No 1 H decoupling
is applied during the variable delay to prevent damage to the probe. As a
consequence, relaxation behavior measured this way may be non-
exponential because there is the potential for cross relaxation involving
both 1 H and X nuclei and a simple analysis of the result may not be
possible.
Spin–lattice relaxation times in the rotating frame
In the basic measurement (Figure 7.14 (a)), spectra (or a single point from
the FID) are recorded as a function of the spin-lock time. The signal decay
is fitted to an equation of the form missing image file T 1ρ depends on
the RF field strength during the spin-lock (as well as on the sample) so it is
important to state this when quoting results.
missing image file can be measured indirectly through an X nucleus

spectrum by incorporating a CP step (Figure 7.14 (b)). This is often
referred to as a delayed contact experiment. As for missing image file ,
this may impart some chemical significance to the result.
missing image file can also be determined by varying the contact time in
a CP experiment and fitting the result to equation 3.8. However, the data
analysis for the delayed contact method is simpler: missing image file
missing image file
Figure 7.14. Pulse sequences for measuring spin–lattice relaxation in the

rotating frame.
Figure 7.14 (c) gives a way of measuring missing image file with the
advantages of the CP method. Proton decoupling is not applied during the
variable delay to prevent unwanted CP. Again, cross relaxation may
complicate the data analysis.
With all the spin-lock measurements, care must be taken not to damage the
probe with very long spin-lock pulses.
The sequences involving CP are invariably used in conjunction with magic-

angle spinning (MAS). This raises the issue of whether relaxation times
measured on a non-spinning sample are the same as those from one
undergoing MAS. For protons, in particular, high spin rates weaken the
homonuclear dipolar coupling and it may be the case that a molecule-wide
average relaxation time is no longer observed. Such behavior has been
noted, for example, for hydrogen-bonded carboxylic acid protons at spin
rates of 30 kHz. In this case the acid proton may have a considerably longer
spin–lattice relaxation time than the other protons in the molecule. T 1 ρ
depends on motion in the tens of kilohertz range—similar to those
associated with MAS, so the subtle effects of interference between T 1 ρ ,
the RF field strength, and the spin rate can complicate matters—but can
provide useful information on the slower motions in a sample.
7.3 EXCHANGE
In addition to causing relaxation and influencing linewidths, motion in the

form of exchange processes can also affect the appearance of a spectrum.
Here, an exchange process is defined as one that interchanges the
environment of the atoms in the system under study. The motional process
is frequently an internal rotation or a change in speciation involving the
movement of hydrogen. The examples in the next two sections illustrate
both these types of process. There is more information on dynamic tensor
averaging in section 8.2.
7.3.1 POSITIONAL EXCHANGE
An exchange process can involve two or more sites (or environments). For
example, in an organometallic coordination polymer containing –C≡N–
Sn(CH3 )3 –N≡C–, there is rotation about the N–S n–N axis and the methyl
groups interchange between three equally populated environments. The
appearance of the carbon spectrum depends on the temperature of the
measurement: at low temperature three methyl signals are observed but at
high temperature a single signal is seen. For simplicity, exchange between
only two sites (A and B, say) is considered now:
If the forward ( k 1 ) and backward ( k 2 ) exchange rates are the same, then
A and B will be equally populated and they will contribute equally to the
NMR spectrum. If the exchange rates are not equal, then the system will
spend more time in one of the sites than the other (the lifetime, τ, of a
particular state is given by τ = 1/ k 1 or 1/ k 2 ). In this case, sites A and B
do not contribute equally to the spectrum.
The drug compound formoterol (used in the treatment of asthma) illustrates

an exchange process involving 180 ° ring flips about a phenylene axis. Its
molecular formula is shown in scheme 7.1. It can exhibit a form of
positional exchange that is often observed in the solid state for molecules
containing 1,4-bonded phenylene rings. In solution in dichloromethane, the
ring labeled A undergoes fast rotation (on the NMR timescale), which
renders carbons c and c′ equivalent (and similarly d and d′), so that each
pair gives a single line in the carbon-13 spectrum (at 115.6 and 131.8 ppm,
respectively). Part of the carbon-13 spectrum from the fumarate dihydrate
form recorded in the solid state at ambient probe temperature (~25 ° C) is
shown in Figure 7.15 and consists of a number of well-resolved lines.
Comparison with the solution-state chemical shifts allows the lines to be
assigned as indicated, but there are no obvious signals for carbons c, c′, d,
or d′. This contrasts with the situation at low temperature when four
additional lines (at 107.2, 118.6, 127.5, and 130.9 ppm) are observed.
These can be attributed to the missing carbons (the average shifts of these
pairs, 112.9 and 129.2 ppm, are close to the solution-state values).
missing image file
Scheme 7.1. Formoterol (fumarate salt).
These observations can be explained if the phenylene ring is undergoing

180° flips about the 1,4-axis. At –50 ° C, the phenylene ring is only
flipping slowly on the NMR timescale (at a rate of less than 1000 s–1 ) and
the local environments of c and c′ and d and d′ are sufficiently different for
distinct resonances to be observed. As the temperature is increased, the ring
motion becomes faster and eventually one line would be observed for c and
c′ and one for d and d′ (at the average of the individual chemical shifts). At
intermediate temperatures, the sharp lines observed at low temperature
broaden (see also section 8.8.1) and eventually each pair coalesces into one
broad line before this single line narrows at higher temperatures. For
formoterol fumarate dihydrate, the coalescence temperatures for the two
pairs of lines are around ambient temperature and it is difficult to detect the
broad lines from these carbons among the other lines from the rest of the
molecule. The behavior of the signals is illustrated schematically in Figure
7.16 .
missing image file
Figure 7.15. Part of the 13 C cross-polarization magic-angle spinning

(CPMAS) spectrum from formoterol fumarate dihydrate recorded at
ambient probe temperature (a) and at –50 ° C (b). The lettering refers to
scheme 7.1.
missing image file
Figure 7.16. Schematic illustration of the coalescence of two lines, related

by exchange. The point at which one flat-topped band is observed is known
as the coalescence temperature . The vertical scales are varied as indicated.
k is the exchange rate constant.
Coalescence t emperature
It is possible to model the lineshape to determine an exchange rate over a

range of temperatures. However, it is necessary to know the shape of the
lines in a non-exchanging state. Given that the general behavior of a pair of
lines in the solid state is, in many cases, analogous to that in solution (i.e.,
they broaden, coalesce, and then narrow), then the exchange rate ( k c ) at
the temperature of coalescence, for a two-site exchange with equal forward
and reverse exchange rates, is given by:
where missing image file ν A and ν B being the frequencies of the two
lines related by exchange. For the case discussed here, the c and c′ lines
coalesce at approximately 25 ° C and Δ ν = 860 Hz, so missing image file
7.3.2 HYDROGEN EXCHANGE
The exchange process for tropolone involves the transfer of a proton

between the two oxygen atoms, together with a rearrangement of the
double bonds (scheme 7.2). There is a subsequent rotation of the whole
molecule to restore its packing in the crystal lattice so that the start and end
points of the process are identical. This is not evident in an X-ray
diffraction study, but the presence of an exchange process can be
demonstrated using a 2D NMR experiment (called EXSY). The EXSY
experiment is almost as simple as the WISE experiment; the only difficulty
comes in choosing the mixing time , τ mix . Getting that right often comes
down to trial and error unless the exchange rate is already known (see
below). The pulse sequence is shown in inset 7.7 and the experiment
applied to tropolone gives the result shown in Figure 7.17 . The peaks
along the diagonal are the ones that appear in the 1D spectrum. The off-
diagonal peaks arise because of the exchange process. They link pairs of
resonances on the diagonal that are related by exchange: 1 and 2, 3 and 7,
and 4 and 6. Only carbon 5 has no partner as its environment is unaffected
by the exchange.
missing image file
Scheme 7.2. Hydrogen exchange in tropolone.
Inset 7.7. EXSY

missing image file
The EXSY pulse sequence, shown with a CP preparation period and

decoupling during the evolution time t 1 and the acquisition time (t 2 ). The
experiment is carried out in 2D fashion by incrementing t 1 . The parameter
τ mix is the mixing time and as this is often relatively long, it is usual to
turn the decoupling off during this period. The effect of the 90° pulse at the
end of the preparation period and before the mixing time is to store the
(carbon) magnetization along the z axis. The result of any mixing
(exchange) is then “ read” by the subsequent 90° pulse. The experiment
usually works best when the mixing time required to demonstrate exchange
is short relative to missing image file (and missing image file ).
Often, it is useful to measure the exchange rate as a function of

temperature. The 2D spectrum of the quality shown in Figure 7.17 took 76
hours to obtain, so recording a series of such spectra can be prohibitively
time consuming. However, the EXSY experiment can be adapted to yield
the exchange rate more quickly. If, instead of incrementing the evolution
delay t 1 (inset 7.7), it is set to a single value determined by the frequency
difference of two lines of interest, then a spectrum of the form shown in
Figure 7.18 will be obtained at zero mixing time. As the mixing time is
increased, exchange will reduce the intensity of li ne A, while line B will
become less negative. A plot of the intensity difference against the mixing
time will yield the exchange rate, as shown in Figure 7.19 . In the tropolone
case, at 30 ° C, this procedure yields an exchange rate of 0.12 s –1 . The
advantage with this method is that it only took a total of 2 hours to record
the spectra for the 10 mixing times used to produce Figure 7.19 , so it is
entirely feasible to obtain the exchange rate as a function of temperature.
missing image file
Figure 7.17. A 13 C EXSY experiment from tropolone obtained at 30 ° C

and with a mixing time of 3 s. The proton relaxation in tropolone is quite
slow, so a long recycle delay is needed (60 s a t 9.4 T). The numbering
refers to scheme 7.2 and the dashed lines link peaks related by the
exchange process.
missing image file
Figure 7.18. The result of a 1D form of the EXSY experiment from

tropolone, obtained at 30 ° C with zero mixing time. Line A is on-
resonance and the t 1 delay (inset 7.7) is set to 1/2Δ v . This has the effect
of inverting line B.
missing image file
Figure 7.19. The difference in signal height as a function of mixing time for
the modified EXSY experiment on tropolone at 30 ° C. The time constant
of the decay is (ignoring relaxation) 1/2 k .
Energy of a ctivation
For a thermally activated exchange process, the activation energy can be

obtained from an Arrhenius plot of exchange rate as a function of
temperature (see section 7.2.1). For tropolone the literature value for the
activation energy is 109 kJ mol–1 . The activation energy for a phenylene
ring flip like that in formoterol is typically lower (<50 kJ mol–1 ).
7.3.3 REORIENTATION WITHOUT A CHANGE IN ISOTROPIC

CHEMICAL SHIFT
In general, any anisotropic interaction can be modulated by molecular

motion, so EXSY-based experiments (or bandshapes for quadrupolar
nuclei) can often give information on a change in orientation even when
there is no change in isotropic chemical shift. This contrasts with the
solution state, where exchange effects are limited to jumps between
inequivalent sites because anisotropic interactions are averaged out.
Deuterium
It is the high natural abundance of 1 H and the myriad dipolar couplings

that are responsible for the relatively featureless static 1 H bandshapes
discussed in section 7.2.3. By contrast, for 2 H, homonuclear dipolar
coupling can be neglected so the 2 H resonance from a non-spinning
sample is dominated by quadrupolar coupling (see chapter 6 ) and this may
be affected by anisotropic molecular motion. This will translate into the
shape of the spinning sideband manifold in a spectrum acquired under
MAS.
Deuterium has a low natural abundance (0.0115%) and to observe a signal

isotopic labeling is often necessary. However, if this can be done in a site-
specific fashion, deuterium can be introduced into a molecule or
macromolecular system as a localized probe of motion. In the simplest
cases, the motion results in a scaling of the quadrupolar bandshape that
would be observed when there is no motion. This is illustrated in Figure
7.20 . The examples here show the slow-exchange limit ( Figure 7.20 (a) ),
where missing image file , and the fast exchange limit (Figure 7.20 (b)),
where missing image file . Like the spin- missing image file case,
intermediate correlation times (typically 10–5 to 10–6 s) produce distorted
(and, for MAS, broadened) bandshapes.
Other types of motion give rise to different bandshapes, as illustrated in

Figure 7.21 (and figure 6.24 ). So, the appearance of the deuterium
bandshape (or spinning sideband manifold in a MAS spectrum) can give
information on the angle of reorientation involved in the motional process.
However, in practice, things might not be as straightforward as these
simulated spectra imply. In polymeric systems, for example, a well-defined
local motion is likely to be superimposed on motion(s) of the polymer
chain. This will complicate the 2 H bandshape and make it more difficult to
extract a picture of the motion. In some cases, synthesizing a system with 2
H localized to a single environment may not always be practical and then it
may be necessary to unravel complex overlapping bandshapes from
multiple deuterium environments. Double-quantum/single-quantum or
heteronuclear correlation experiments may be necessary to achieve the
required 2 H spectral resolution. Analogous effects also can be observed for
other quadrupolar nuclei (including those where I ≠ 1) .
missing image file
Figure 7.20. Schematic simulated 2 H bandshapes. For: (a) an immobile C–

D bond, missing image file and (b) a deuteromethyl group rotatin g about
the C–C bond. For (b) the motion is a hypothetical “free” rotation with
missing image file (i.e., missing image file ) and the quadrupole
coupling is scaled by a factor missing image file where ψ is the angle
between the C–D bond and the axis of rotation (109.5° here).
missing image file
Figure 7.21 . Simulated 2 H bandshape for a fast ( missing image file )

180° phenyl ring flip, missing image file
A non-spinning EXSY experiment can also be applied to deuterium

(although in this case the 90° storage pulses are replaced by 54.7° ones).
The result is a 2D spectrum with off-diagonal elliptical ridges, the precise
nature of which gives information on the angle of molecular reorientation
involved in the exchange process6 (see section 8.7.2).
For spin- missing image file nuclei, the CSA can also be used to probe
motion. Two-dimensional experiments can distinguish between different
models of the reorientational process as the pattern of off-diagonal intensity
depends on the nature of that process. For 13 C, MAS is necessary for
sensitivity, so care is needed to ensure that the intensities of the off-
diagonal spinning sidebands relate to the exchange process (and are not
modulated by the sample spinning). Although 2D experiments can
distinguish between different motional models, acquiring a series of them
to extract exchange rates can be prohibitively time consuming. For this
reason, a number of experiments, which rely only on the measurement of
centerband intensities in 1D spectra, have been developed. One such is
CODEX (Centerband Only Detection of EXchange), and from two short
series of experiments, it is possible to obtain a correlation time and
information on the type of motion present (in principle for all resolved lines
in the spectrum) (see Figure 7.22 and Further reading).
missing image file
Figure 7.22 . Carbon-13 cross-polarization magic-angle spinning (CPMAS)

spectrum from (top) polymethylmethacrylate (PMMA) together with
(middle) its pure-exchange Centerband Only Detection of Exchange
(CODEX) spectrum (for which the signal from immobile sites has been
subtracted). The graph at the bottom is the normalized exchange intensity
as a function of mixing time for the COO carbon. The low value of the final
intensity (marked “a”) implies that less than 50% of the COO groups are
undergoing large-amplitude motion. The correlation time is ~50 ms
(determined by fitting the exponential build-up of the CODEX signal).
7.3.4 DIFFUSIVE MOTION
In most single-component molecular systems in the solid state, the motion

that affects the NMR spectrum is restricted to (hindered) rotations or
hydrogen exchange. In composite systems, however, translational, diffusive
motion is possible. The techniques discussed earlier in sections 7.2.2 and
7.2.3 can be used to study this. For example, in materials designed for
lithium-based batteries, the 7 Li linewidth and relaxation properties are
strongly affected by the ability of the lithium ions to diffuse. In one such
study, the static 7 Li spectrum at low temperature showed a broad and
featureless line consistent with immobile lithium ions in a distribution of
environments within the host matrix. As the temperature was increased, the
linewidth decreased until a single narrow and symmetric line was observed.
This was attributed to lithium ions undergoing rapid isotropic reorientation
during diffusion within channels in the host matrix. Finally, at high
temperature, a scaled quadrupole bandshape was observed, which was
interpreted as arising from lithium ions in an aggregate too large for
isotropic motion but undergoing anisotropic diffusion within the matrix.
Valuable information can be obtained by studying the guest in solid host–

guest structures, such as water in membranes or hydrocarbons in porous
zeolitic structures. However, the study of diffusion is more suitable for
pulsed field gradient or imaging techniques, both of which introduce some
sense of position within the sample to the NMR experiment and are
therefore sensitive to translational motion.
7.3.5 “SOFT” SOLIDS
In all the cases introduced so far in this chapter, the sample would be
described as a rigid solid. The general molecular motion in a “soft” solid
has a major impact on the NMR spectrum. For example, in a high-
resolution 13 C spectrum from such an organic material, the partial
averaging of the 13 C,1 H (and 1 H,1 H) dipolar coupling means that the
spectrum is to some extent self-decoupled and requires relatively little RF
decoupling power to reduce the linewidths to a small number of hertz (see
Figure 7.23 and inset 7.8). The CSA is fully averaged, so spinning
sidebands are not observed even at relatively low spin rates (<5 kHz).7
Even though such materials do not necessarily possess the long-range order
of a crystalline solid, the motional averaging of each environment (akin to
that in solution) means that the broad lines associated with a rigid
amorphous solid are not observed. In consequence, as illustrated in Figure
7.23 , highly resolved spectra can be obtained that begin to approach the
appearance of those from a solution.
missing image file
Figure 7.23. (a) Carbon-13 cross-polarization magic-angle spinning

(CPMAS) spectrum from a whole hazelnut. The narrow lines and absence
of spinning sidebands are indicative of a soft solid. The unsaturated fatty
acid component of the nut is preferentially detected with a long (10 ms)
contact time. With a short contact time (0.1 ms), (b) the “hard” component
of the nut is detected.
Inset 7.8. Adamantane
Tricyclo[3,3,1,1]decane, C 10 H 16 , or adamantane. Its melting point is
270 ° C (although it readily sublimes, even at room temperature). At room
temperature, it has a disordered structure but undergoes fast reorientation
(~1.6 × 10 11 jumps per second). This pseudo-isotropic motion removes all
intra molecular dipolar interactions but inter molecular ones are only
partially averaged due to the lack of translational motion.
missing image file
Adamantane is a soft solid that gives very narrow lines at low (~2 kHz)
spin rates and low (~30 kHz ) decoupling fields (left). It is an excellent test
of shimming—recorded at 75.43 MH z, the full half-height linewidth here
is less than 2 Hz and satellites from 1 J CC can be observed. The spectrum
was obtained with CP.
7.3.6 INTERFERENCE
Molecular mobility can produce some unexpected effects in high-resolution
spectra when the rate of motion approaches that of MAS or the nutation
rate of 1 H decoupling. In both cases, the effect can be thought of as a
destructive interference so that the interactions which the spinning or
decoupling should be removing from spectra are being reintroduced.
Without going into the mathematics of the process here, it can be said that
the result is a broadening of the affected lines. This broadening will be
most extreme when the spinning or nutation and motion rates are matched,
but lines narrow again as they diverge. An example is given in Figure 7.24 .
Interference can also arise between the timing of some pulse sequences
and, for example, the sample spin rate as noted in section 5.5.1.
missing image file
Figure 7.24. Deuterium magic-angle spinning (MAS) spectra from

finasteride d8 -dioxane solvate hydrate at (a) –7 0 ° C, (b) –5 ° C, and (c)
+40 ° C. The dioxane undergoes a chair–chair inversion (slow at –70 ° C
and fast at 40 ° C). At –5 ° C the exchange rate is close to that of the spin
rate (6 kHz) and the interference between the two results in line
broadening. At +40 ° C, the motion also causes a narrowing of the spinning
sideband manifold.
7.4 QUANTITATIVE NMR
Determining the signal intensity is key to many of the experiments

discussed in this chapter. Measuring the intensity of a single resolved (not
overlapping) line in a high-resolution spectrum from a spin-
missing image file nucleus is straightforward and is normally done using
numerical integration. The precision of the result is determined by the
signal-to-noise ratio in the spectrum.
For the integration of any signal, it is important that the spectral baseline is
flat and has no offset (so the value of the integral truly represents the area
under the peak and not the shape of the baseline as well). If signals from
different samples are being compared, the signal intensity needs to be
corrected for the amount of sample in the rotor (so the sample mass should
be recorded). For a heterogeneous sample, it is important that the
components are uniformly distributed through the rotor (as the sample is
not uniformly excited/detected along the whole length of the rotor). When
signals overlap, spectral deconvolution (see inset 7.9) may be required. For
a resonance with spinning sidebands, the intensity of the latter should be
added to that of the centerband.
Inset 7.9. Deconvolution

Deconvolution is the process of obtaining information about the
components of a bandshape that contains overlapping lines. It is commonly
applied to the spectra from spin - missing image file nuclei; software for
doing this is widespread. Deconvolution of overlapping quadrupolar
bandshapes is an altogether more complicated process (due to the number
of variables involved).
Gaussian or Lorentzian lineshapes (see Figure 7.25 ), or a linear

combination of the two, often can be used to represent the lines in a solid-
state NMR spectrum (from spin- missing image file nuclei). The result of
the deconvolution of a 29 Si bandshape is shown in Figure 7.26 . Here
Gaussian shapes have been used and the position, width, and height of the
lines in a three-line model have been allowed to vary in an iterative fit to
the observed spectrum. The observed and simulated spectra should only
differ at noise level, as in this example, but such precision, using these
simple lineshapes, can be hard to achieve—at least if the number of
components is restricted to being chemically meaningful.
missing image file
Figure 7.25 . Common lineshapes.
In Figure 7.25 the Lorentzian (gray line) is characterized by the broad base:
Gaussian (black line) lineshapes are used in Figure 7.26 :

where α = 4log e (2). In Figure 7.25 the full width at half-height (FWHH),
w , is the same for both shapes. The area under each line (of most interest
in a deconvolution) is:
for the Lorentzian and Gaussian cases, respectively.
missing image file
Figure 7.26 . Deconvolution of a 29 Si bandshape into three lines with

relative intensity 66%, 31%, and 3% (low to high frequency).
Additional problems arise for quadrupolar nuclei. Resonances from species

with different χ values will have different excitation profiles and hence the
response to a single pulse may be biased. The amount of signal located in
the (possibly undetectable) satellites might also need to be considered.
Furthermore, if one type of environment has a particularly large value of χ ,
second-order effects may broaden the signal to the point where it is
unobservable. Such problems can be encountered when quantifying 27 Al
spectra, for example, from aluminosilicate materials.8
7.4.1 RELATIVE INTENSITY
When measuring the relative intensities within one spectrum, it is essential

to take proper account of relaxation times and to set the recycle delay
accordingly. This is particularly relevant when direct excitation is used,
since T 1 values for different nuclei may vary substantially and the recycle
delay must be set at a value significantly higher than the longest relaxation
time involved (see section 3.3.1/Figure 3.7 ). This is not generally a
problem for a homogeneous system when CP is used because recycle
delays depend on missing image file which is uniform when spin
diffusion is efficient. However, a different criterion then becomes important
(see section 3.4.2 and, particularly, Figure 3.20 ), since different carbons,
for instance, will cross polarize at different rates. In particular, quaternary
carbons generally show slow CP rates, as illustrated in Figure 7.27 ,
because the protons are relatively remote. In such a case, long contact times
may be necessary for successful quantitation, which can be a problem if
significant loss of carbon magnetization, due to missing image file
relaxation, has occurred by that stage. Ideally, signal intensity should be
plotted against contact time for all resonances, but this is not normally
feasible.
missing image file
Figure 7.27 . Signal intensity vs. contact time for the CH labeled A
(circles) and carboxylic acid carbon (crosses) resonances in l -isoleucine.
More substantial problems arise when it is important to determine relative

intensities for a heterogeneous system using CP, because the recycle delay
now has a significant effect if missing image file differs for different
components. Moreover, differences in missing image file may now also
need to be taken into account. Suppose a system has two components, A
and B. If missing image file then a CP plot such as is shown in Figure
7.28 may result, which means that signals for component B will be rapidly
lost as the contact time increases. Such a case occurs for styrene/butadiene
copolymers. The example here is a block copolymer, with each domain
sufficiently large that it retains its own value of missing image file At
ambient temperature, missing image file for the butadiene domains is an
order of magnitude greater than that for the styrene domains. Therefore, at
long contact times, the signals for styrene are lost, as shown in Figure 7.29
(a) and (b). However, the butadiene is slow to cross polarize and so is
underrepresented in the spectrum recorded with a short contact time.
Spectra recorded only at high contact times would indicate that little
styrene was present, with the opposite conclusion at low contact times!9
Values of missing image file are temperature dependent and at 200 K the
situation is reversed, with styrene having the longer missing image file
(see Figure 7.29 (c)). Thus, at the lower temperature and long contact time,
only styrene signals appear.
missing image file
Figure 7.28 . Schematic plot of signal intensity vs. contact time for a
heterogeneous system of two components with equal concentrations but
widely differing values of missing image file S A (max) and S B (max) is
the maximum amount of signal that can be observed for component A and
B, respectively.
7.4.2 ABSOLUTE INTENSITY
If a spectrum is used to quantify the amount of sample, or a component of

it, in the rotor on an absolute scale (i.e., in moles per rotor), an extended
strategy is necessary. There are two ways of approaching the problem of
referencing such a measurement.
A known quantity (as a %) of a reference compound can be added to

provide a reference signal. The nature and amount of the additive need to
be carefully controlled, since a clear NMR signal with a similar intensity
(with respect to the signals from the compound to be measured) is required.
Ideally, a calibration curve should be constructed from samples of known
content and this can then be used for a sample of unknown content. If this
is not possible, any differences in relaxation, CP behavior, or excitation
efficiency must be taken into account.
missing image file
Figure 7.29 . Carbon-13 spectra from a styrene/butadiene copolymer under

various conditions: (a) ambient temperature, contact time 0.4 ms; (b)
ambient temperature, contact time 10 ms; (c) 200 K, contact time 10 ms.
The polymer contains 30 wt.% styrene.
Figure 7.30 shows a test case for a pharmaceutical formulation containing

the drug substances bambuterol hydrochloride (BHC) and terbutaline
sulfate (both used in asthma therapy) in a lactose monohydrate as the
majority component, together with a small amount (1%) of magnesium
stearate. The methylene 13 C signal for magnesium stearate was used as the
reference. A calibration curve was employed to show that the amount of
BHC is linearly related to the ratio (from deconvoluted peak areas) of the
BHC signal at 147.3 ppm to that of the reference. This plot was used when
measurements were made on samples of unknown BHC content. It was
shown that the limit of detection (in 3 hours of spectrometer time) was 0.5
mol % and the limit of quantification was 1 mol %—important information
when comparing the value of NMR with other techniques for such
measurements.
The second method of referencing signal intensity (as an alternative to

using a chemical additive to the sample) is to create a signal electronically
—the so-called ERETIC method.10 The spectrum is “spiked” with a
resonance using a signal generated electronically in a spare RF channel.
This signal can be shaped so that it produces a realistic lineshape when it is
Fourier transformed. The signal is fed to the probe and detected in the
NMR coil along with the signal from the sample. Because the two signals
are simultaneously detected, they must be coherent in phase. A Fourier
transformation will yield a spectrum containing a “synthetic” peak.
Calibration of the area under this peak against standards of known
concentration allows this signal to be used to quantify spectra of unknown
concentration. The advantages of using the ERETIC method are that the
sample is not contaminated, the ERETIC peak can easily be placed at a
position in the spectrum where it does not interfere with signals from the
sample, there is no dependence on the relaxation behavior of the reference
signal, and the amplitude of the reference can be adjusted (using the pulse
amplitude control of the spectrometer) to be similar to that of the spectrum
of interest.
missing image file
Figure 7.30. Carbon-13 cross-polarization magic-angle spinning (CPMAS)

spectrum, accumulated in 3 hours, of the pharmaceutical formulation (see
the text) containing 5% of bambuterol hydrochloride (BHC) and 1%
magnesium stearate. The signals used for quantifying the content of BHC
are indicated by arrows BHC to high frequency and magnesium stearate to
low frequency ). The BHC signal at 147.3 ppm arises from two carbamic
acid carbons, that from the stearate (C36 H70 MgO4 ) is from ~30
coincident mid-chain CH2 carbons (hence the apparent mismatch in
intensity given the respective concentrations of these components).
Crystalline polysaccharides typically have very long missing image file
(often hundreds of seconds). With the recycle delay optimized for the drug
components, the signals from the lactose saturate and are therefore severely
underrepresented in this spectrum.
7.5 PARAMAGNETIC SYSTEMS
The presence of unpaired electrons has a profound effect on NMR. The

gyromagnetic ratio of the electron is ~660 times larger than that for 1 H so
the coupling between the unpaired electron spins and the nuclear spins
(hyperfine coupling) is large and must be taken into account.11
Unpaired electron density is an efficient driver of nuclear spin relaxation,

and so relaxation times for paramagnetic systems tend to be very much
shorter than those for diamagnetic ones. Shortened T 1 s can be used to
advantage (see section 7.5.1), but the hyperfine coupling of sites close to
unpaired electron density leads to significant line broadening. It is usually
impractical to obtain useful NMR spectra from paramagnetic atoms
themselves;12 however, it is often possible to obtain useable NMR signals
from sites with weaker hyperfine couplings, that is, those at greater
distances from the paramagnetic center and/or with less delocalized
electron density from the unpaired electron at the nuclear site.
7.5.1 RELAXATION EFFECTS
Short T 1 s in paramagnetic systems can be exploited, that is, paramagnetic

species can be used to shorten recycle delays for systems with prohibitively
long T 1 s. For example, it has been observed that oxygen in the pores of
zeolites acts as a relaxation agent (molecular oxygen is paramagnetic), and
the oxygen partial pressure can be deliberately increased to reduce the 29 Si
T 1 . Similarly, spectra from coordination compounds, organometallics and
ceramics can be enhanced in intensity if the system is doped with a small
concentration of paramagnetic metal ions, such as Mn2+ , to form a solid
solution. The magnitude of this effect is strongly dependent on the distance
between the unpaired electron and the nucleus and the concentration of the
relaxation agent, and so additional care is needed when choosing the
parameters for quantitative measurements (see section 7.4.1).
7.5.2 SHIFT EFFECTS
The nature of the hyperfine coupling is complex and its magnitude (and
sign) varies from nucleus to nucleus. The coupling has essentially three
components: a Fermi contact interaction (due to unpaired electron density
at the nucleus), a pseudocontact term that is equivalent to the nuclear
dipole–dipole coupling, and a spin-orbit term (see Further reading for
details). Because the electron spins generally relax very quickly, the
multiplet resulting from the coupling is collapsed into a single line by
processes analogous to those described in section 7.3.13 However, the
position of this single line may be significantly different from that observed
from an analogous diamagnetic system. This paramagnetic shift arises
because the difference in energy of the electron spin states is not negligible
in comparison with kT. Consequently, the non-negligible population
difference for the electron spin states means that the population-weighted
average frequency is not at the center of the putative doublet (in the NMR
spectrum).
The dependence of the population distribution on kT means that the

paramagnetic shift is strongly temperature dependent, which distinguishes
it from other contributions to the chemical shift. The shift can potentially
provide useful information on the distribution of electron density and, in
favorable cases, it can be experimentally measured by comparing the
spectrum of the paramagnetic compound with that of an analogous
diamagnetic compound.
missing image file
Figure 7.31 . Proton magic-angle spinning (MAS) spectra from

paramagnetic Mn(acac)3 as a function of spin rate. (Figure adapted from
data published in N.P. Wickramasinghe et al., J. Am. Chem. Soc . 127
(2005) 5796.)
The hyperfine coupling is anisotropic, and the anisotropy can be very large
(thousands of parts per million even for 1 H spectra), so although it has the
usual p 2 (cosθ ) dependency, it can be difficult to handle with MAS. Thus
the MAS NMR spectra of paramagnetic materials are often largely
unresolved and contain extensive manifolds of spinning sidebands. As
shown in Figure 7.31 , however, very fast MAS (>25 kHz) may allow
useful resolution to be obtained.
FURTHER READING
“Multidimensional solid-state NMR and polymers” , K. Schmidt-Rohr &
H.W. Spiess, Academic Press (1994), ISBN 0 12 626630 1.
“Intramolecular motion in crystalline organic solids”, P. Hodgkinson, in

“NMR Crystallography” , Eds. R.K. Harris, R.E. Wasylishen & M.J. Duer,
John Wiley & Sons, Ltd. ( 2009), 375–386 (Chapter 25), DOI: 1
0.1002/9780470034590.emrstm1002, ISBN 978 0 470 69961 4.
“Centreband-only detection of exchange (CODEX)”, K. Schmidt-Rohr,

E.R. deAvezado & T.J. Bonagamba, in “Encyclopaedia of Nuclear
Magnetic Resonance” , Volume 9, Eds. D.M. Grant & R.K. Harris, John
Wiley & Sons, Ltd. (2002), 633–642.
“Combining NMR spectroscopy and quantum chemistry as tools to

quantify spin density distributions in molecular magnetic compounds”, M.
Kaupp & F.H. Köhler, Coord. Chem. Rev ., 253 (2009), 2376–2386. DOI:
10.1016/j.ccr.2008.12.020.
“Strategies for solid-state NMR studies of materials: From diamagnetic to

paramagnetic porous solids”, V.I. Bakhmutov, Chem. Rev ., 111 (2011),
530–562. DOI: doi:10.1021/cr100144r.
“Electron paramagnetic resonance: Elementary theory and practical

applications” , J.A. Weil, J.R. Bolton & J.E. Wertz, John Wiley & Sons Ltd.
(1994), ISBN 0 471 57234 9.
NOTES
1 When energy is lost through relaxation, it has to go somewhere. The term

“spin–lattice” is indicative of where:- energy is lost to the “lattice.”
2 Note that the symbol J does not relate to indirect coupling in this context!
3 This spectral density is for a process with a single, well-defined

correlation time. It is not so appropriate for motion with distributions of
correlation times. The definition of the correlation time depends on the type
of motion. For example, for rotational motion it is often defined as the
average time for rotation through 1 radian.
4 Some texts introduce a term called the second moment ( M 2 ) to help
describe the bandshape. This is a measure of the mean, local magnetic field
generated by all the pairwise dipolar interactions experienced by a
particular nucleus. High values for M 2 equate to broad lines and are
associated with rigid systems where the dipolar coupling is at its strongest.
For a powdered sample, the second moment for heteronuclear interactions
is: missing image file and for the homonuclear case missing image file
r jk is the distance between spins j and k.
5 Assuming the mixing time is much less than missing image file (in this
case).
6 The spectrum from deuterated dimethylsulphone is a classic example of

this: C. Schmidt, B. Blümich and H.W. Spiess J. Magn. Reson. 79 (1988)
269.
7 Low-intensity spinning sidebands can be observed but are probably

attributable to other factors such as bulk susceptibility.
8 High-symmetry (small χ value) aluminum environments octahedrally or

tetrahedrally coordinated by oxygen are easy to detect and are likely to be
excited in a similar fashion, so measuring their relative intensity can be
done reasonably accurately. However, if the environments are distorted and
if their χ values diverge, this is no longer the case.
9 The pair of spectra shown in Figure 7.23 (a) and (b) illustrate a similar
selectivity.
10 The acronym stands for “Electronic REference To access In vivo

Concentrations,” but the method is now used for nonmedical spectra.
11 This requires an additional term to be added to equation 1.2 (see Further

reading for more details on the form of this term).
12 The nucleus at the site of a localized unpaired electron will have a very
large hyperfine coupling constant (potentially tens of MHz), resulting in
multiplets that are far too broad to be observed by conventional NMR.
13 The hyperfine coupling of the paramagnetic center itself is so large that
even the very quick electron spin relaxation is not efficient enough to
collapse the multiplet from the nucleus at the site of a localized unpaired
electron.
CHAPTER 8
ANALYSIS & INTERPRETATION
8.1 INTRODUCTION
This chapter takes the reader through a variety of topics related to the
interpretation of spectra and the extraction of detailed information about the
NMR interactions. The first section describes the analysis of one-
dimensional spectra to obtain, for example, chemical shift information and
so does not involve special experimental techniques. However, obtaining
reliable estimates of other parameters, such as the strength of dipolar
couplings, may require specialist experiments and/or analysis, so most of
the rest of the chapter deals with somewhat more sophisticated aspects of
solid-state NMR, especially involving tensors and their evaluation. The
final section looks to the value of NMR in crystallography, where it
complements diffraction-based techniques (diffraction crystallography ).
8.2 QUANTITATIVE MEASUREMENT OF ANISOTROPIES
8.2.1 GENERAL
As discussed in earlier chapters, all NMR interactions are essentially

anisotropic. It is often necessary to reduce the effect of these anisotropies
on the NMR spectrum by techniques such as magic-angle spinning (MAS)
in order to improve resolution. However, the anisotropies1 contain valuable
information about the local environment. This section considers how the
anisotropic components of NMR interactions, such as the shielding and
quadrupole coupling tensors, can be measured and exploited.
Measurements of the dipolar couplings are particularly significant, since
their strength directly depends on the distance between the nuclei involved.
As a result, a wide range of experimental techniques have been derived to
measure dipolar couplings (see section 8.3), and hence determine structural
information from solid-state NMR studies.
In order to measure anisotropy information, two conditions have to be met.

Firstly the anisotropy has to be sufficiently large in order to be measured
accurately, and secondly the different sites present need to be resolved in
order to obtain anisotropy information for each distinct resonance . These
conditions may conflict. For example, anisotropy information is often
characterized most accurately at very low MAS speeds or by using static
samples, since the overall profile of the bandshape can be fully modeled.
But if multiple sites are present, spectral overlap will often prevent the
resolution of the different sites. Using fast MAS rates may lead to clean,
well-resolved spectra, but when spinning sidebands are completely absent,
information about the anisotropies is lost.
In many cases it is possible to use intermediate spinning rates to obtain

resolved spectra that still contain sufficient spinning sidebands for analysis.
The fitting of such spectra to determine anisotropic information is
discussed in the following section. In more difficult cases, alternative
strategies are required. If the interactions involved are inhomogeneous (see
section 2.7 and inset 4.5), for example the shielding anisotropy, and so are
associated with sharp spinning sidebands, it is common to use two-
dimensional spectra which contain isotropic information in one dimension
and anisotropic information in the other. These are discussed in section
8.2.5. In other cases, particularly when the interaction is homogeneous
(such as is common for the homonuclear dipolar coupling), the usual
strategy is to use moderate to fast MAS to suppress the anisotropies and
obtain high-resolution spectra, and use recoupling sequences to selectively
reintroduce anisotropic information. The qualitative application of
recoupling was discussed in section 5.4.2. The quantitative use of
recoupling, particularly in the context of dipolar interactions, is discussed
in section 8.3.
Finally, it is important to be aware of potential interactions between

dynamics and NMR anisotropies; indeed the modulation of NMR
anisotropies by motional processes is a powerful tool for the
characterization of dynamic processes in solids, as previously discussed in
section 7.3. If the frequency of the dynamic process is significantly faster
than the anisotropies involved,2 then an averaged tensor will be observed.
In the case, for example, of a two-site exchange between sites 1 and 2 with
fractional populations p 1 and p 2 = 1 – p 1 , the average of a tensor A will
be:
Note that a common reference frame must be used for the tensors; here M
denotes a molecular frame of reference, rather than the unrelated principal
axis systems of the individual tensors. The parameters of the averaged
tensor can be fitted in the same way as “normal” unaveraged tensors,
although it is important to note that averaged tensors do not necessarily
have the same symmetry properties as the original tensors. For example,
averaged dipolar tensors may have non-zero asymmetries even though the
original dipolar tensors are axially symmetric by definition.
If the dynamic process is slow compared with the anisotropies (the slow
exchange limit), then tensor information can be determined independently
for the different sites (assuming that they can be resolved). In the
intermediate case, the form of the spectrum will be changed by the dynamic
process, with the exact shape depending on the tensors involved and the
nature of the dynamic process. Fitting such exchange-modulated spectra is
a powerful means of characterizing the dynamic process. The focus here is
on spectra that are unaffected by intermediate timescale dynamics.
8.2.2 QUANTITATION OF POWDER LINESHAPES
In favorable cases, it may be possible to read off tensor information from

turning points in powder spectra. For instance, the separation of the horns
of the Pake doublet from an isolated dipolar coupling (see figure 2.4 ) can
be accurately measured and provides the dipolar coupling constant directly.
When the edges of the powder pattern are defined sharply enough, the
principal components of the shielding tensor can also be read off from the
turning points, as in figure 2.2 .
In more difficult cases, and particularly when estimates of errors are

required, it is necessary to fit the powder pattern to simulated spectra; the
NMR frequencies are calculated as a function of crystallite orientation, and
the individual spectra summed to give the overall pattern. Note that a large
number of orientations may need to be accumulated in order to obtain a
smooth lineshape if the ratio of the anisotropy to the intrinsic linewidth is
very large.
Figure 8.1 (a) shows an example of fitting the central-transition bandshape
of a MAS powder pattern (obtained at 104.20 MHz) for a 27 Al spectrum
resulting from second-order quadrupolar effects (the small feature around 7
ppm is due to the satellite transitions). Exact modeling of powder
lineshapes is often difficult. For example, the underlying (homogeneous)
lineshape cannot easily be modeled as it is swamped by the inhomogeneous
lineshape, and broad patterns are often not excited uniformly by the NMR
experiment. In figure 8.1 (a), however, the turning points of the powder
pattern are well defined, and MAS has suppressed other orientation-
dependent interactions, such as the shielding anisotropy. This allows the
principal components of the quadrupole coupling tensor to be estimated
with confidence.
missing image file
Figure 8.1. (a) Fitting of a MAS powder lineshape due to second-order

quadrupolar effects on the central transition of 27 Al in aluminum
acetylacetonate. Quadrupolar parameters of χ = 3.0 MHz and η = 0.16
match the turning points of the simulated and experimental spectra. (b) and
(c) Simulated spectra for a spin- missing image file nucleus influenced by
the quadrupolar interaction (χ = 3 MHz) and shielding anisotropy (ζ = 20
kHz), both with zero asymmetry and with coaxial tensors, at (b) 14.1 T and
(c) at 9.4 T.
In other cases, it may be necessary to obtain complementary data sets in

order to determine individual tensor contributions with confidence. For
example, if the central-transition spectrum of figure 8.1 had been obtained
under non-spinning conditions (see the upper spectrum in figure 6.6 ), the
powder pattern would contain a contribution from the shielding anisotropy,
which would distort the fitting of the quadrupole coupling parameters.
Obtaining spectra at multiple magnetic fields is the best solution for
distinguishing shift interactions (which scale in frequency units with B 0 ),
from (first-order) coupling effects (independent of field) and/or from
second-order effects (which scale inversely in frequency units with B 0 ;
see figures 8.1(b) and (c) and also section 8.7.1). A sing le set of shielding
and quadrupolar parameters3 that fits spectra at different fields
simultaneously is likely to be robust.
8.2.3 QUANTITATION OF SPINNING SIDEBAND MANIFOLDS
Fitting of anisotropy information is commonly performed using data from

MAS experiments, where the anisotropy parameters are encoded in the
pattern of intensities formed by the spinning sidebands. In comparison with
fitting spectra from static samples, MAS spectra have the advantage of
being better able to discriminate multiple sites. It is also generally easier to
phase spectra and correct baselines where peaks are separated by clear
regions of baseline; flat baselines and well-phased spectra are very
important when using intensity information quantitatively (see section 7.4).
Note that the number of spinning sidebands required depends to some

extent on the information wanted; anisotropies (i.e., essentially the “width”
of the pattern) can be quantified with good accuracy with only four to five
sidebands of measurable intensity. By contrast, determination of the
asymmetry parameter (which loosely determines the shape of the pattern)
steadily improves as more spinning sidebands are included. Unless the
signal-to-noise ratio is a major problem or slower spinning rates cause
centerbands and sidebands to overlap, it is better to err on the side of using
slower spinning rates to increase the number of spinning sidebands above
the minimum number required to quantify the anisotropy parameter
effectively. This makes it easier to identify any systematic deviations
between the model and the data.
If the sidebands of interest are well resolved from other peaks, then the
most straightforward approach is to determine the peak heights (or, better,
peak areas) of the observed sidebands. The intensity pattern of the spinning
sideband manifold is then fitted numerically to obtain the anisotropy
parameters. The Further reading has information about some of the various
software tools available.
missing image file
Figure 8.2 . Fitting of the 31 P spinning sideband manifolds of the two

phosphorus sites in γ -(MoO2 )2 P2 O7 . Although obtaining the shielding
parameters from the pattern of peak heights would be quite feasible in this
case, fitting the complete spectrum (middle trace) makes the most use of
the available data. The fitted spectrum is the top trace and the lowest trace
shows the difference between the fitting and experiment. The experim ental
spectrum was obtained at 202.28 MHz with a spin rate of 5 kHz. The zero
of the frequency scale is arbitrary. (This original figure has been adapted
for publication in the Supplementary Information to Lister et al., Inorg.
Chem. 49 (2010) 2290.)
In cases where signal overlap prevents the determination of accurate

intensity information for the individual sidebands, it is necessary to fit the
complete spinning sideband pattern directly. This is a little more involved
since additional parameters are required to model the lineshape of the
spinning sidebands, but fitting the complete manifold in one step is more
robust than using deconvolution to estimate individual sideband integrals
and then fitting the resulting intensity pattern. Figure 8.2 gives an example
of such a fit where complete sideband manifolds were modeled (see also
figure 2.13, which shows an example with a single chemical shift). The
difference between fitted and experimental data sets (lowest trace) shows
features due to an impurity peak (and its associated spinning sidebands).
The residual plot also shows small features indicating that the modeled
lineshape (a mixed Lorentzian/Gaussian; see inset 7.9) does not perfectly
match the experimental shape. Such minor deviations are to be expected
and will have negligible impact on the accuracy of the overall fitting, which
here gave 31 P shielding anisotropy (asymmetry) parameters of 65.5 ppm
(0.28) and 69.3 ppm (0.69) for the high- and low-frequency resonances,
respectively. The fitting also confirms that the overall intensity ratio for the
two sites is exactly 1:1 (within the fitting error), which is not immediately
apparent from the spectrum.
As with any fitting of experimental data, it is important to note the

associated confidence bounds on the fitted parameters. Although these may
not be entirely reliable (particularly if there are significant systematic
deviations between fitted and experimental data), they will give a good
indication of the precision that can be reasonably quoted. This is
particularly important for asymmetry parameters, since they are often
difficult to fit accurately—for example, if there are too few spinning
sidebands to determine the “shape” of sideband manifold and/or the
symmetry is close to axial. Small changes in the asymmetry parameter
when it is less than about 0.1 have very little effect on the sideband
manifold, meaning that the asymmetry is not well constrained in this case.
8.2.4 SINGLE-CRYSTAL VS. POLYCRYSTALLINE SAMPLES
The drawback of using polycrystalline (“powder”) samples, as described

above, is that information about the orientation of the tensor with respect to
the crystal axes cannot generally be obtained; it is normally only possible to
determine the magnitudes of the principal components of the tensor. In
favorable cases, discussed below, it may be possible to determine the tensor
orientation with respect to another tensor, but because polycrystalline
samples contain a random distribution of all possible crystal orientations, it
is impossible to orient the tensor with respect to the crystal axes.
In principle, tensor orientation can be obtained by working with single-

crystal samples. However, there are a number of major practical obstacles
to single-crystal NMR, not least the difficulties of obtaining crystals of
sufficient volume, and the need for a specialist NMR probe containing a
goniometer (required to orient the crystal within the magnetic field). Full
characterization of NMR tensors via single-crystal experiments is much
more demanding than the experiments on polycrystalline samples described
above.
8.2.5 RESOLVING ANISOTROPY INFORMATION BY ISOTROPIC

SHIFT
Fitting anisotropy information becomes difficult when multiple tensors are

involved. The case of multiple interactions involving the same site is
discussed in sections 8.5 and 8.6, but another important case is when
spectral overlap involving distinct sites prevents the measurement of
anisotropies.
Fortunately, a variety of two-dimensional NMR experiments can be used to

resolve the problem of overlap. For example, magic-angle turning (MAT)
experiments for shielding anisotropy measurements involve very slow
MAS and result in two-dimensional spectra with anisotropy information in
one dimension and isotropic frequencies in the other; slices through the 2D
spectrum at the isotropic frequencies provide shielding anisotropy powder
patterns for the different sites. A relatively straightforward alternative for
shielding anisotropies that works under moderate to fast MAS is the phase
adjustment of spinning sidebands (PASS) experiment. This uses rotor-
synchronized π -pulses to manipulate phases of the spinning sidebands as in
the TOSS experiment for the suppression of spinning sidebands (see
section 3.3.6), except that, rather than just providing a center-band-only (m
= 0) spectrum in which sidebands have cancelled, the result is a series of
spectra containing just the peaks from a single sideband order (m = –1, m =
0, m = 1, etc.) (see figure 8.3 ). Collating the intensity information gives the
spinning sideband manifold patterns for the different resolved sites.
missing image file
Figure 8.3 . Results (obtained at 100.56 MHz) of a 13 C PASS experiment

on 3-methoxybenzoic acid: (a) spectrum using a spin rate of 10.0 kHz,
showing only weak sideband signals; (b) spectrum using a spin rate of 2.0
kHz, showing many sideband signals; (c) separate sideband signals for the
acid and ring C–O carbons, the former with a centerband at δ C = 172.8
ppm, and sidebands indicated by the vertical dashed lines. The PASS
centerband is shown for all signals. The inset shows the experimental
(black) and simulated (gray) sideband manifolds for the 172.8 ppm
centerband. The shielding parameters used for the simulation were
anisotropy ζ = 65 ppm and asymmetry η = 0.55.
The separation of anisotropy information by isotropic shift is especially

straightforward when the anisotropy involved is a coupling rather than a
shift anisotropy, since the two types of interaction can be readily
distinguished, for example, by a spin echo. Such separated local field (SLF)
experiments are often used to measure the dipolar local field at different
sites. For instance, the cross-polarization (CP) experiment discussed in
section 3.4 provides CP build-up curves for different sites resolved by their
13 C chemical shift, allowing 1 H,13 C dipolar couplings to be measured
for each resolved site.
8.3 MEASUREMENT OF DIPOLAR COUPLINGS
The direct relationship between dipolar coupling constants and internuclear

distance (equation 2.24) makes them a particularly attractive target for
quantification. In contrast to measurements of the shielding anisotropy,
however, it is rare to encounter a case where the powder spectrum or
spinning sideband manifold is dominated by an isolated dipolar coupling
(except where specific sites have been isotopically enriched). In general,
specialized experiments are required to isolate the effects of the desired
dipolar coupling from other NMR interactions. Readers are referred to the
literature (see Further reading) for information on quantitative
measurement of dipolar couplings involving quadrupolar nuclei using
sequences such as TRAPDOR and REAPDOR.
The other inherent problem with dipolar couplings (in contrast to single-
spin interactions such as the shielding and quadrupole coupling) is the
promiscuous nature of through-space couplings; in a typical organic solid,
for example, a given 1 H spin will have significant dipolar couplings to a
number of nearby 1 H spins, resulting in a broad overall spectrum that
cannot be fitted in terms of individual couplings (see figure 2.5 ). As
mentioned above, selective labeling is often necessary to isolate an
individual pair of spins. Moreover, it cannot be assumed that an observed
dipolar coupling corresponds to an internuclear distance within the same
molecule—it could correspond to a short intermolecular distance.
Particularly in small-molecule systems and/or when accurate quantitative
results are required, it is common to recrystallize specifically labeled
molecules with an excess of the corresponding unlabeled molecule in order
to reduce the impact of intermolecular couplings.
8.3.1 MEASUREMENT OF HETERONUCLEAR COUPLINGS
Figure 8.4 illustrates two possible pulse sequences that can be used to
measure the heteronuclear dipolar couplings between a pair of spins, I and
S. The simplest approach (which works well if the coupling between I and
S is relatively strong) involves exploiting the CP experiment. If spin
diffusion among the I (and the S) spins is slow (compared with the
timescale for CP), then the CP build-up, obtained by measuring the
intensity of the S spin signal as a function of τ CP , is oscillatory. This is
illustrated in figure 8.5 for 1 H → 31 P CP in SnHPO4 . A strong
oscillation due to the 1 H,31 P dipolar coupling is observed. As discussed
in section 5.1.3, normal Hartmann–Hahn matching is inefficient in cases
where MAS is effectively suppressing 1 H spin diffusion; hence it is
necessary to match on a sideband, here n = 1. The “spectrum” associated
with this build-up curve then takes the form of a Pake-like pattern4 with the
separation of the horns given by missing image file .
missing image file
Figure 8.4 . Two pulse sequences that can be used to measure heteronuclear
dipolar couplings: (a) cross-polarization (CP) signal acquired as a function
of CP contact time, τ CP , (b) a rotational-echo double-resonance REDOR
experiment, in which the S spin signal is measured after each spin echo,
with and without a train of rotor-synchronized 180° pulses applied to the I
spins. Note that phase cycling of the I spin pulses using so-called XY
schemes such as XY8 is widely used to minimize the effects of RF
imperfections.
Note that the chemical shift (including its anisotropy) is suppressed by the
spin-locking during the contact time and so the oscillation is purely a
function of the dipolar couplings. This experiment is then in effect a
separated local field experiment, since Fourier transformation with respect
to both τ CP and t 2 provides a spectrum with the dipolar coupling
information in one dimension and S spin chemical shifts in the other (the IS
coupling being suppressed in t 2 by the heteronuclear decoupling).
Additional weak IS couplings and spin diffusion will tend to broaden the
lines in the dipolar spectrum. As discussed in section 5.5.1, Lee-Goldburg
CP can be used to suppress the effects of spin diffusion to first order,
although this is unlikely to significantly sharpen the dipolar dimension for
powder samples.5
The second relevant pulse sequence shown in figure 8.4 (b) is the rotational
echo double resonance (REDOR) experiment, which uses recoupling (as
discussed in section 5.4.2) to reintroduce the effects of dipolar coupling that
are otherwise suppressed by MAS. In terms of the observed S spins, the
pulse sequence involves a spin echo, with the echo detected after an even
number of complete rotor periods (see Further reading). In the absence of
additional pulses on the I spins, the evolution due to the IS coupling
refocuses over a rotor period, while the chemical shift evolution is
refocused by the spin echo (see sections 4.6.1 and 4.4.2). Hence the only
evolution of the “reference” signal is due to homogeneous factors
(including relaxation) that are not refocused by the spin echo. If, however,
180° pulses are applied to the I spins twice per rotor period, then the
refocusing of the IS coupling is disrupted, and the observed signal is
reduced relative to the corresponding reference signal. Plotting the
(normalized) difference between the reference signal and the signal with
the REDOR pulses as a function of the dephasing time gives a build-up
curve which can be fitted to obtain the IS dipolar coupling. Figure 8.6
shows an example where NMR gives crystallographic information
complementing that from diffraction measurements. The aluminophosphate
sieve in question has cages containing a fluorine atom. Diffraction results
suggest this is centrally located in the cages, whereas the Al–F distance
determined by REDOR indicates that it is in a bonding position close to
one aluminum atom. The XRD and NMR results can be reconciled if the
fluorines are disordered over the multiple bonding sites in each cage. The
diffraction studies (erroneously) locate the fluorines at the cage centers.
missing image file
Figure 8.5 . Illustration of the use of transient oscillations in 1 H to 31 P

cross-polarization build-up curves for SnHPO4 to measure dipolar
couplings. The separation of the horns, Δ , of the resulting “spectrum”
gives the dipolar coupling ( missing image file ), and hence the 31 P,1 H
distance (here 0.203 nm) via equation 2.24. (This original figure was
adapted for publication in Amornsakchai et al., Mol. Phys . 102 (2004)
877).
Unlike the CP-based experiment described earlier, the REDOR experiment

does not involve observing magnetization transferred from, say, I to S. This
may lead to problems from “background” signals from any S spins that are
not coupled to I. Variant pulse sequences, such as TEDOR , have been
devised which observe magnetization transferred between I and S to avoid
this problem. Because they involve polarization transfer, TEDOR-type
sequences can be readily incorporated into HETCOR experiments. 3D-
TEDOR experiments have, for example, been used to quantify the 13 C,15
N couplings responsible for the cross-peaks in a 13 C/15 N HETCOR
experiment, although such experiments are only viable in isotopically
enriched samples.
Figure 8.6 . 27 Al{ 19 F} REDOR results for the Al signal (at 10 ppm) of
the aluminophosphate molecular sieve AlPO 4 -5. The open circles are the
experimental data. Calculated curves correspond to Al–F dipolar coupling
constants of (dotted line) 2.8 kHz ( r AlF = 0.219 nm) and (continuous line)
4.15 kHz ( r AlF = 0.192 nm). (This original figure was adapted for
publication in R.D. Gougeon et al., J. Phys. Chem. B 105 (2001) 12249.)
The REDOR experiment and its variants have been widely used to measure
dipolar couplings between a variety of nuclei. It can even be applied to
couplings between spin - and half-integer quadrupolar nuclei, with the
quadrupolar nucleus as the S spin, since refocusing the chemical shift
evolution at the midpoint of the recoupling sequence can be achieved
relatively cleanly via a single 180° pulse.
As the signal is only measured at (even) multiples of the rotor period, the
REDOR experiment is not suitable for strong dipolar couplings (of the
order of the spinning speed or greater) since not enough data points can be
obtained to characterize the dipolar oscillation. The CP-based experiment
of figure 8.4 (a) is better suited to these cases. In the other direction, the S
spin missing image file sets the ultimate limit on the smallest couplings
that can be measured. It is important to avoid overloading the probe when
using long recoupling times to measure weak couplings. For example, if 1
H decoupling is required, it will need to be applied continuously during
both recoupling and signal acquisition periods. It may be necessary to
truncate the acquisition (and sacrifice some resolution of the S spin
spectrum) to keep the duty cycle within acceptable limits.
8.3.2 MEASUREMENT OF HOMONUCLEAR COUPLINGS
Extracting accurate estimates of dipolar coupling constants is often more

difficult for homonuclear spins, in large part because isolated spin pairs are
relatively uncommon. Moreover, the spin dynamics are often complex if
more than one coupling is involved, making it difficult, or impossible , to
interpret the results in terms of individual couplings.
In the case of 1 H spins in typical organic solids, there are multiple strong
homonuclear dipolar couplings between nearby H atoms, making it
impossible to isolate the effect of an individual coupling. The dynamics of
magnetization transfer between the spins are best described in terms of an
overall spin diffusion rather than the kind of oscillatory build-up curves
observed for isolated couplings seen in figure 8.5 . That said, the rate of
spin diffusion between a pair of spins will be directly related to the
coupling strength and can provide a qualitative indication of the proximity
of the spins involved.
The principal difficulty with estimating dipolar couplings between 1 H

nuclei is the strength of the couplings and the resulting limited spectral
resolution; fast MAS or strong homonuclear decoupling is required to
suppress the homonuclear couplings sufficiently to resolve the different
sites. The dipolar couplings between most other spins are much weaker,
and so, by contrast, it is often necessary to use recoupling techniques (as
described in section 5.4.2) to prevent the couplings being eliminated by
MAS. In this case, there is the option of using experiments that selectively
recouple individual spin pairs. For example, the rotational resonance
experiment involves matching the difference in NMR frequencies of two
spins to an integer multiple of the spinning frequency, that is deliberately
overlapping a spinning sideband of one peak with the centerband of the
other. Under these resonance conditions, magnetization can be exchanged
between the spins (in a similar way that magnetization is exchanged
between heteronuclei under Hartmann–Hahn matching), and the dipolar
coupling extracted from the build-up curve. Figure 8.7 shows an example.
missing image file
Figure 8.7 . Nitrogen-15 spectra, obtained at 20.28 MHz, of doubly 15 N-

labeled 5-methyl-2-diazobenzenesulfonic acid hydrochloride. Bottom: Fast
magic-angle spinning case, showing sharp peaks separated by 1786 Hz.
The peak to low frequency is assigned to the α nitrogen while the high-
frequency signal is for the β o ne. Top: The centerband region of the
spectrum obtained with a spin rate to equal the frequency difference
between the two signals (n = 1 rotational resonance). Detailed analysis of
the full rotational resonance spectrum yields 0.1110 ± 0.0004 nm for the N–
N distance (see R. Challoner & R.K. Harris, Chem. Phys. Lett . 228 (1994)
589.)
Where the magnetization exchange is not unduly affected by multiple

couplings, the broadband recoupling schemes described in section 5.4.2
(and related pulse sequences) can be used to good effect. Sequences such as
POST-C7 have the advantage that the recoupling efficiency is independent
(to first order) of the chemical shift Hamiltonian. The exact spin dynamics
under other recoupling schemes, such as radiofrequency driven recoupling
(RFDR), often depends on the chemical shift anisotropies and/or the
isotropic chemical shifts of the spins involved. This is rarely a problem in
qualitative applications (such as those described in chapter 5 ), but
complicates quantitative analysis. An example of a recoupling experiment
for the 1,10-difluorodecane/urea inclusion compound (using the
MELODRAMA pulse sequence) is shown in figure 8.8 . Analysis yields a
motionally averaged intermolecular (F,F) dipolar coupling constant,
missing image file , of 995 Hz, from which deductions have been made
regarding the conformation of the guest molecule in the urea tunnels.
missing image file

Figure 8.8 . 19 F{1 H} MELODRAMA plot for the 1,10-
difluorodecane/ure a inclusion compound. The n = 4 matching condition
was used. The experimental points are shown, together with a dashed line
for the optimized simulated plot. The horizontal axis represents a variable
time interval in the MELODRAMA pulse sequence. (This original figure
was adapted for publication in “Recent advances in solid-state NMR”, R.K.
Harris, in “ Magnetic resonance in food science: A view to the future ”,
Eds. G.A. Webb, P.S. Belton, A.M. Gil & I. Delgadillo, Royal Society of
Chemistry (2001), 3–16.)
8.4 QUANTIFYING INDIRECT (J) COUPLINGS
Scalar (J) couplings are measured straightforwardly using spin-echo

experiments in which the echo delay periods are multiples of the rotation
period. The spin echo refocuses evolution under the isotropic shift, and
when the rotor angle is set sufficiently accurately that the evolution due to
anisotropic interactions is refocused over the rotation cycle, then the
intensity of the spin-echo signal is determined only by scalar couplings that
are not refocused.
As discussed in section 5.3, the spin-echo signal as a function of the time τ

when a single coupling is active will fit to:6
where 2τ is the total spin-echo period (see sections 4.4.2 and 7.2.3), J is the
coupling constant, and missing image file is the time constant used to fit
the signal decay. Figure 8.9 shows an example of fitting J couplings
between pairs of 31 P nuclei. Couplings between different NMR nuclides
are fitted in the same way, except that the refocusing 180° is applied to both
nuclei involved to select for heteronuclear rather than homonuclear
couplings.
missing image file
Figure 8.9 . Example of fitting 2 J pp couplings from 31 P spin-echo data.

See figure 5.6 for full details.
Very small J couplings can be fitted in this way with good accuracy, with
the limit set by the missing image file decay. In this case,
missing image file ms, corresponding to a homogeneous “linewidth” of
missing image file Hz, that is, the oscillation due to the 10 Hz coupling
can still be resolved, but errors on fitting couplings smaller than 8 Hz will
increase sharply.
8.5 TENSOR INTERPLAY
So far, it has mostly been assumed that only a single type of tensor affects
the spectrum, but clearly, in many (if not most) cases, several NMR
interactions contribute to defining it. Thus one must consider interplay
between tensors. Two such cases will be considered in some detail, one in
this section and a second one in section 8.6. Here the case of interplay
between the chemical shift and heteronuclear dipolar coupling will be
discussed. This will also involve indirect coupling. The situation can be
understood relatively simply because the tensors involved commute and
therefore affect the spectrum in an additive fashion. Spectra obtained at
more than one applied magnetic field can assist in understanding the
situation (since J and D are field independent to first order, whereas
chemical shifts scale linearly with the applied field), but the treatment of an
individual spectrum is considered here.
The simplest case, which illustrates the essential features, occurs for
isolated heteronuclear two-spin (AX) systems with missing image file in
which axial symmetry prevails so that asymmetry in σ A , σ X , and J AX
can be ignored. Moreover, all three tensors for, say, the A spins, will then
be coaxial (with the major principal axis along the internuclear A, X
distance, r AX ). The anisotropy in J AX has exactly the same form as the
dipolar coupling, D AX , and therefore an effective parameter
missing image file may be defined as:
where D AX is the dipolar coupling constant in frequency units,

missing image file (equation 2. 24), and missing image file is the
anisotropy in J AX expressed as missing image file .7 The theory for this
situation is outlined in inset 8.1.
Inset 8.1. Interplay between σ , D , and J for a heteronuclear two-spin
system
The NMR Hamiltonian may be written as
where missing image file and missing image file are the A and X
Larmor frequencies in the absence of shielding, missing image file and
missing image file are the isotropic A and X shielding constants, ζ A and
ζ X are the shielding anisotropies (see inset 2.2) expressed as
missing image file and missing image file , respectively, J AX is the
isotropic (A, X) coupling constant, and θ is the angle between r AX and B 0
. The transitions of the A nucleus are therefore described by:
where m X is the appropriate spin component quantum number for the X

spin. This equation reduces to:
where missing image file is an effective tensor anisotropy given by:
The equations above are given under the assumption that γ A and γ X are
both positive. Careful consideration needs to be given to signs if that is not
the case.
Equations 8.5 and 8.6 show that the A transitions form two subspectra
missing image file in one of which the shielding and dipolar tensors
reinforce one another, whereas in the other they partially cancel each other.
Thus, one static powder subspectrum is “stretched” whereas the other is
“squeezed”, as indicated schematically in figure 8.10 . This hypothetical
case is drawn for the case when J AX , D AX , and ζ A are all positive8
(involving missing image file ), with missing image file and
missing image file
missing image file
Figure 8.10 . Schematic powder pattern for the A nucleus, say, of a

heteronuclear two-spin (AX) system subject to (a) dipolar coupling only,
(b) both dipolar coupling and shielding anisotropy, and (c) both dipolar
coupling and shielding anisotropy, with the subspectra separated by an
isotropic indirect coupling constant. The isotropic positions of the
subspectra are indicated in bold; for (a) and (b) they are at the same
position. Under the conditions as described in the text, the subspectrum
when m X is + is “squeezed” (indicated by the solid lines), while that
when m X is – (indicated by the dashed lines) is “stretched.” The plot
parameters used were: ζ A = 2D AX /5 and J AX = D AX /6.
so that it is the low-frequency subspectrum that is “squeezed”. Figure 8.10

(a) shows a case with only dipolar coupling, while figure 8.10 (b) illustrates
the effect when both dipolar coupling and shielding anisotropy affect the
spectrum. The “horns” of the total pattern for figure 8.10 (b) remain
separated by | D AX |. A static powder pattern could be analyzed to give
information on both D AX and ζ A . Figure 8.10 (c) shows that the
subspectra are distinguished when indirect coupling is present. The “horns”
are now separated by | D AX + J AX |.
Under slow MAS conditions, the influence of the angle-dependent term

missing image file is translated into a distribution of intensities among
the spinning sidebands, which extend over a larger frequency range for the
stretched subspectrum than for the squeezed subspectrum. For figure 8.10
(b) (as for figure 8.10 (a)), the isotropic shift positions for the two
subspectra are the same, so while MAS would give rise to two spinning
sideband manifolds, these would overlap and so be difficult to disentangle.
The isotropic indirect coupling serves to resolve the subspectra, as shown
in figure 8.10 (c), enabling the spinning sideband manifolds to be analyzed
separately and thus yielding the two values of missing image file . One
consequence is that the centerbands (split by J AX ) of the two spinning
sideband manifolds have different intensities, as can be seen for an
experimental spectrum in figure 8.11 . The magnitudes of ζ A ,
missing image file , and J AX are readily obtained by the analysis. Since
the sign of ζ A will normally be obvious (and can be obtained separately by
spinning sideband analysis of an A spectrum obtained with X-decoupling)
sign information on J AX and missing image file are obtained, albeit with
some ambiguity. Only the relative signs of missing image file and J AX
are determined, the rule being as follows: If the high-frequency spinning
sideband manifold has the larger magnitude of missing image file , then
(i) for positive ζ A , missing image file , and J AX have the same sign,
whereas (ii) for negative ζ A , missing image file , and J AX have
opposite signs. The inverse is true if the larger magnitude of
missing image file belongs to the low-frequency subspectrum. If r AX is
known (e.g., from diffraction studies), D AX can be calculated, so Δ J can
be derived, though with an ambiguity unless the sign of J AX is known. On
the other hand, if the relative magnitudes of D AX and Δ J /3 can be
assumed, the absolute sign of J AX can be determined.
missing image file
Figure 8.11 . Spinning sidebands of the proton-decoupled 13 C cross-

polarization magic-angle spinning (CPMAS) spectrum for the ipso phenyl
carbon of solid trimethylphenylphosphonium iodide at 75 MHz. The
centerband doublet shows a splitting arising from isotropic indirect
coupling to the 31 P nucleus of 84 Hz in magnitude, but the intensities of
the two centerband peaks are markedly different. The intensities of the
doublet peaks in the spinning sidebands are mostly biased in the opposite
direction, giving equal total intensities in the two subspectra, as shown by a
summation of all the sidebands (see the inset).
Extension to a linear AX2 case is straightforward. There will now be three

spinning sideband manifolds in the A spectrum, with effective tensor
anisotropies given by:
where Σm X = 1, 0, or –1, associated with the three centerbands at

missing image file . The central sideband manifold gives ζ A directly,
because it is independent of missing image file . Clearly, this is an
advantage, particularly if X-decoupled A spectra cannot be obtained
because of spectrometer limitations. Figure 8.12 illustrates this case, for the
119 Sn spectrum (proton decoupled) of Me3 SnF, which has a structure
with F…Sn-F…Sn chains in which each tin is effectively equally coupled
to two fluorine nuclei. The arrowed lines form the centerband of the J-
coupled triplet. The spinning sideband manifold of the high-frequency
triplet line is stretched, and that of the low-frequency triplet line is
squeezed, as expected from the theory mentioned above. However, when
the respective spinning sidebands are summed, the expected 1:2:1 triplet is
obtained. Analysis of the central spinning sideband manifold gives the Sn
shielding anisotropy ζ Sn as –221 ppm and missing image file as –4.02
kHz (negative because the magnetogyric ratio for 119 Sn is negative). The
isotropic 1 J (119 Sn,19 F) coupling constant is 1290 Hz, known to be
positive.
missing image file
Figure 8.12 . Tin-119 cross-polarization magic-angle spinning (CPMAS)

proton-decoupled spectrum at 149.12 MHz of Me3 SnF. The spin rate was
7.5 kHz. The centerbands of the three subspectra are indicated with arrows.
The linked marks under the spectrum indicate the spinning sideband
manifold for the central transition of the triplet. The inset shows a
summation over the whole spinning sideband manifold. (See H. Bai, R.K.
Harris & H. Reuter, J . Organomet. Chem . 408 (1991) 167.)
The situation is more complicated when the tensors involved are not
coaxial. This will be discussed qualitatively in section 8.7.
8.6 EFFECTS OF QUADRUPOLAR NUCLEI ON SPIN- SPECTRA
Another case of tensor interplay arises for NMR spectra of quadrupolar

nuclei because of the magnitude of the quadrupolar interaction (see chapter
6 ), the effect of which can be “transferred” to spectra of spin - nuclei by
dipolar coupling. Consider a two-spin missing image file system
consisting of a spin - nucleus (I) and a quadrupolar nucleus (S). If the
spin wavefunctions are unperturbed from their Zeeman forms, then only the
secular term of the I,S dipolar interaction needs to be considered for
the energies of the spin states. Since this depends on missing image file ,
its effects are eliminated by MAS. However, any substantial quadrupolar
coupling will mix the Zeeman wavefunctions of the S spin states, as
mentioned in section 6.3. In this case, dipolar coupling terms such as
missing image file (for the notation see section 4.2.1) will now affect the
S spin in a way which will have a different angle dependence and so will
not be eliminated by MAS. However, the value of this effect will depend on
the spin state of the I spin, and so I-spin spectra will also be affected. The
case of a two-spin system comprising a 13 C nucleus and a 14 N (spin-1)
nucleus will be considered in inset 8.2 in more detail for the simplified
situation where (a) the quadrupolar and dipolar tensors are coaxial, (b) the
EFG tensor is axial, and (c) indirect (J) coupling can be ignored.
The effect on the energy levels is shown schematically in figure 8.13 . The
missing image file manifold of levels is affected by the second-order
terms in the opposite way to that of the missing image file levels. The
energy changes on the missing image file levels are equal whereas the
effect on the missing image file levels is opposite in sign and twice the
magnitude.
This means that the three allowed 13 C transitions are moved away from
the isotropic chemical shift, though the center of gravity remains
unchanged. The result is a 1:2 or 2:1 doublet, giving what is termed a
residual dipolar splitting . For the same reason that MAS does not eliminate
second-ord er effects on a quadrupolar spectrum, it cannot remove them
from the residual dipolar splitting of a spin - spectrum, though it does
cause significant averaging of the orientation-dependent factors. Therefore ,
the resonance frequencies depend on the angles between B 0 and the
dipolar and quadrupolar axes, so that the observed MAS spectrum for a
microcrystalline sample will be a powder pattern. Figure 8.14 illustrates the
spectrum for a 14 N, 15 N example, which clearly shows the splitting and
the powder-pa ttern nature of the two components. However, for the 13 C,
14 N case details of 13 C patterns are only usually seen at very low
magnetic fields; usually a broadened 2:1 or 1:2 doublet is recorded. For
such cases (under the assumptions already mentioned), the doublet splitting
is:
Inset 8.2. Second-order quadrupolar effect on the energy level of a 13 C, 14

N spin system
Perturbation theory shows that the mixing of the nitrogen spin states by
second-order terms results in (un-normalized) wavefunctions:
in which missing image file and missing image file
In these circumstances, the dipolar coupling cannot be truncated to its

secular part, but other terms need to be included, in particular those
involving 14 N step-up and step-down operators, missing image file and
missing image file . Such terms have non-zero secular parts. For
example, the nitrogen level m N = 1 now includes an energy contribution
proportional to
An equal term is required for the level m N = –1, whereas for the level m N
= 0, the analogous quantity is missing image file . The contribution to the
energy levels also involves the dipolar coupling constant and the polar
angles of the dipole direction in the magnetic field, but the net effect will be
proportional to a 2 and –2a 2 for the 14 N levels m N = ±1 and m N = 0
respectively. The sign of the second-order terms clearly depends on m C.
(However, terms in a 1 , involving the double step-up operator
missing image file and its partner, missing image file , are independent
of m C and therefore do not affect the I-spin transitions.)
Since missing image file is the Larmor frequency of the quadrupolar spin,
its value is known. Therefore, if χ is available (e.g., from NQR
measurements), D can be obtained (and therefore the internuclear distance
between the two spins derived). Conversely, if D is obtained from
diffraction measurements, χ can be estimated.
missing image file
Figure 8.13 . Energy-level diagram for a two-spin system involving one

spin- (13 C) nucleus and one spin-1 (14 N) nucleus. The parameter a is
proportional to the a 2 factor mentioned in inset 8.2.
missing image file
Figure 8.14 . Nitrogen-15 magic-angle spinning (MAS) spectrum at 20.3

MHz of 5-methyl-2-diazobenzenesulfonic acid hydrochloride, enriched in
15 N at the α position, showing the residual dipolar splitting arising from
15 N, 14 N interactions. The splitting between the two major components is
155 Hz. (This original figure was adapted for publication in R. Challoner &
R.K. Harris, Solid State NMR 4 (1995) 65.)
Of course, life is rarely that simple and a number of additional features

need to be understood:
In principle, the anisotropy in the indirect coupling tensor, Δ J ,

behaves exactly like D , and the dipolar parameter D AX in figure
8.10 must be replaced by equation 8.3 (though this is only valid if D
and J are coaxial), thus complicating analysis (though if both χ and D
are known, Δ J can then be obtained). This problem is only likely to be
significant for situations involving heavy nuclides. It can be (and is)
generally ignored for 13 C, 14 N cases.
The isotropic indirect spin–spin coupling constant, J , may also need
to be taken into account, but this is easy since simple splitting is
involved.
Whereas axiality of the EFG and its coaxiality with D are reasonable
approximations for cases with single bonds linking I and S, as when
univalent atoms such as chlorine are involved, that will rarely be true
for examples with 14 N. Thus, more complicated expressions are
frequently required for residual dipolar coupling. Strategies for
determining the angular relationships between tensors will be
discussed in the next section.
The sign of the effect, for example whether a 2:1 or 1:2 pattern is
obtained for the 13 C, 14 N case, depends on the sign of χ and the
angle between the dipolar and quadrupolar tensors. However, sign
matters may be further complicated if D and D ′ are opposite in sign
(i.e., for missing image file ) and if the quadrupolar asymmetry is
non-zero.
Nuclei with large quadrupole moments (e.g., Cl or Br) cause analogous

effects even at very high fields such that second-order perturbation theory
may be inadequate. The reader is referred to Further reading for details.
The effects described in this section can, of course, be decreased if the

applied magnetic field is increased. However, since information can be
derived from the tensor interplay in question, this may not be desirable and
occasionally it may be helpful to reduce the applied field so as to enhance
the second-order effects. Analyzing spectra obtained at two or more applied
magnetic fields increases confidence in the resulting parameters.
8.7 QUANTIFYING RELATIONSHIPS BETWEEN TENSORS
The two preceding sections have discussed how NMR spectra are affected
when a given resonance is significantly influenced by more than one tensor
interaction, for example, a strong dipolar coupling combined with a large
shielding anisotropy. However, only cases with coaxial tensors were
examined. Frequently, this is not the situation in practice. Fitting NMR
parameters to spectra that are determined not only by the parameters of the
individual interactions but also by their relative orientations can be
difficult. However, the relative orientations of tensors form a potential
useful source of additional information, particularly in the context of
powder samples where it is impossible to establish the absolute orientations
of individual tensors. A wide variety of experiments has been proposed for
different combinations of dipolar coupling, quadrupolar coupling, shielding
tensors, etc., and this section simply summarizes the different broad
strategies in qualitative terms. Details can be found in the Further reading.
8.7.1 FROM ONE-DIMENSIONAL SPECTRA
As mentioned in sections 8.5 and 8.6, if a given spin is affected by more

than one anisotropic interaction, then the overall spectrum is a function not
only of both the interactions involved but also of their relative orientations
(see figure 8.15 ). Hence, the relative orientation of tensors can be
determined in favorable cases by careful fitting of conventional NMR
spectra. Such a fitting is not necessarily straightforward, however, as the
spectrum may not be particularly sensitive to the relative orientation—often
only the key β Euler angle (see section 4.6) can be fitted with any
confidence. Simultaneous fitting of complementary data sets (e.g., for
spectra obtained at different magnetic fields) is often necessary to obtain
robust quantification of the individual interactions (see figures 8.1 (b) and
(c)), and hence their relative orientations. Analysis of such spectra yields
accurate parameters for the anisotropic interactions, which would not be
obtained robustly from a spectrum obtained at a single B 0 field.
Static vs. MAS spectra can be useful when dealing with large quadrupole
couplings. The other anisotropic interactions are averaged out by
(sufficiently fast) MAS, leaving just a second-order quadrupole-broadened
lineshape (see section 6.4) that can provide accurate estimates for the
quadrupolar interactions, allowing the other interactions (and relative
orientations) to be determined from the static spectrum.
missing image file
Figure 8.15 . Simulated spectrum (central transition only) for a spin -

missing image file nucleus influenced by the quadrupolar interaction ( χ =
3 MHz) and shielding anisotropy ( ζ = 20 kHz), both with zero asymmetry.
The tensors are coaxial for the left trace and unaligned (with the β Euler
angle set to 20°) for the right trace.
8.7.2 FROM CORRELATION SPECTRA
The principal drawback of determining relative tensor orientations from

one-dimensional spectra is that they are principally determined by the
individual interactions, and the relative orientation has only a minor effect
on the overall spectrum.
By contrast, appropriate 2D correlation spectra can be directly dependent

on relative orientations. If one tensor is “active” in the indirect dimension
and the other in the direct dimension, then the form of the correlation peak
will depend directly on the relative orientation; if the principal axes of the
tensors are coincident, then there will be a perfect correlation between the
frequencies in the two dimensions, leading to a diagonal correlation peak.
If, however, the principal axes are not the same, then off-diagonal intensity
will be observed, with the form of the pattern being directly related to the
relative orientation. These patterns are most easily obtained and interpreted
when powder (continuous) bandshapes are involved. It is, however,
possible to correlate spinning sideband patterns in two dimensions,
although specialist experiments are required to obtain clean in-phase peaks
(see Further reading).
Such 2D techniques are most commonly used in the context of correlation

spectra involving the transfer of magnetization from one site to another. For
example, correlation peaks in EXSY-type spectra (see section 7.3.2) are
often simply used to determine which sites are connected by exchange, but
provided the anisotropy information has not been suppressed by fast MAS,
the patterns of the cross-peaks can be used to determine the relative
orientations of the tensors at the different sites.9 Figure 8.16 shows an
example: Comparison of the observed and simulated spectra reveals a
substantial difference between the α and β forms of methanol. The α form
has four different molecular orientations in the unit cell, with relatively
large angles between the C–O bond vectors, whereas the β form has only
two such orientations, with a small (only 12°) angle between C–O bonds in
adjacent molecules.
missing image file
Figure 8.16. Carbon-13 spectra of static samples for two forms of

methanol: (a) one-dimensional spectra; (b) experimental EXSY-type
spectra (involving spin diffusion during the mixing time); (c) simulated
spectra. (Figure supplied courtesy of K. Zilm. It has been partly adapted for
publication in “Crystallography and NMR: An overview”, R.K. Harris,
Chapter 1 of “ NMR Crystallography ”, Eds. R.K. Harris, R.E Wasylishen
& M.J. Duer, John Wiley & Sons Ltd. (2009). ISBN 978-0-470-69961-4.)
Other correlation experiments can also be used to probe relative tensor

orientation, for example, an MQMAS experiment (section 6.7.4) can be
readily modified to insert an exchange period between indirect (multiple
quantum) and direct (single quantum) dimensions. Magnetization transfer
during this period will lead to additional correlation peaks, the shapes of
which are sensitive to the relative orientation of the quadrupolar coupling
tensors of the sites involved.
8.7.3 SPECIALIZED EXPERIMENTS
A number of experiments have been developed to probe the relative

orientations of dipolar tensors. For instance, the relative orientation of the
two 13 C, 1 H dipolar coupling tensors in a 1 H– 13 C– 13 C– 1 H
molecular fragment provides direct information on the H–C–C–H torsion
angle. This makes it unnecessary to acquire a full two-dimensional
spectrum, and such experiments are usually run in a quasi-one-dimensional
fashion; double-quantum coherence between the 13 C spins is created
(using a recoupling sequence) and allowed to evolve under the 13 C, 1 H
dipolar couplings before being reconverted and measured. The dynamics of
the evolution under the CH couplings is dependent on the magnitude of the
couplings and the H–C–C–H torsion angle. The major drawback of such
experiments is that specific isotopic labeling is required in order to isolate a
suitable spin system.
8.8 NMR CRYSTALLOGRAPHY
Throughout this book, a number of ways in which NMR reveals details of

crystallography have been mentioned. Section 1.3.3, for instance, describes
how the number of resonances seen in MAS spectra gives information
about the crystallographic asymmetric unit. At this point, it is appropriate
to enlarge on the emerging subject of NMR crystallography. While this has
only recently become generally recognized as a subject area in its own
right, NMR has given crystallographic information from the earliest days.
Thus, Pake was able, in 1948, to report (by measurement of dipolar
coupling) the distance between hydrogen atoms in water of hydration in
gypsum (CaSO4 ·2H2 O) crystals as 0.158 nm—even now locating
hydrogen by X-ray diffraction techniques is difficult.
Each of the parameters measured by NMR, including full tensor

information when relevant, is affected by crystal structure and can therefore
give information about such structure. Thus, for instance, dipolar coupling
constants give direct (through-space) distances between nuclei, indirect
coupling reveals interactions between nuclei mediated by electrons,
chemical shifts indicate both intramolecular chemical structure and
intermolecular effects, quadrupolar coupling depends on the local
electronic structure, and relaxation times are sensitive to molecular-level
motion. Moreover, different timescales are effective for NMR and
diffraction. That of diffraction experiments is the time taken to record the
diffracted beam, which may be from seconds to hours. The NMR timescale,
on the other hand, is related to the inverse of the frequency difference
between the exchanging signals, which may be of the order of milliseconds.
NMR therefore probes faster motions than diffraction.
8.8.1 CHEMICAL EXAMPLES
The simplest use of solid-state NMR in crystallography lies in its ability to

determine the chemical structure present. This is valuable when there are
structural differences between a solid and its solution. For example, 2-
aminobenzoic acid can exist as either a neutral molecule or as a zwitterion.
Carbon-13 MAS NMR shows quite clearly (figure 8.17 ) that form I
contains equal numbers of neutral molecules and zwitterions in the unit cell
(the chemical shifts for the ipso carbons being significantly different
between the two forms). In similar fashion, NMR can distinguish between a
co-crystal and a salt—the difference lying in the migration of a hydrogen
atom. NMR can also readily identify the resonances of “solvent” molecules
in a solvate and can quantify the host:solvent molecular ratio. Figure 8.18
shows the 13 C spectrum of a 1:1 solvate of phenobarbital and acetonitrile.
The signals indicated by arrows arise from the acetonitrile and occur at
chemical shifts close to those for acetonitrile in solution.
missing image file
Figure 8.17 . Carbon-13 cross-polarization magic-angle spinning

(CPMAS) spectrum of form I of 2-aminobenzoic acid. This was obtained
using the interrupted decoupling pulse sequence (see subsection 7.2.4) so
that only signals arising from quaternary carbons were observed. The unit
cells of this polymorph contain both neutral molecules and zwitterions, the
resonances of which are separately indicated . The two signals for the
carboxyl carbons overlap.
Of course, diffraction methods are generally regarded as the “gold
standard” for crystal structure determination, and this situation will
undoubtedly continue to be true. However, such methods have limitations.
For instance, any mobility at the molecular level is hard to assess.
Moreover, departure from strict three-dimensional repetition (which is
implicit in the concept of a unit cell) presents problems. Therefore, systems
with non-stoichiometric formulae or with disorder (either static or
dynamic) are difficult to tackle. Fortunately, NMR gives both
supplementary and complementary data to that supplied by diffraction, as is
increasingly being recognized by the diffraction community. Thus, NMR
can supply information to assist structure determination from diffraction
data, and it can also provide data missed by diffraction (especially for
molecular mobility). Moreover, NMR information can be obtained from
polycrystalline, amorphous, and even heterogeneous samples. A nu mber of
examples are given here which demonstrate the power of NMR
crystallography.
missing image file
Figure 8.18 . Carbon-13 cross-polarization magic-angle spinning

(CPMAS) NMR spectrum of the acetonitrile solvate of phenobarbital. The
arrows indicate the “solvent” peaks.
The multinuclear facet of NMR can allow deductions that go beyond the
determination of the crystallographic asymmetric unit to obtain molecular
symmetry information in the solid state. For example, the cyclic trimeric
organotin chalcogenides (Me2 SnX)3 , X = S or Se, give rise to two 119 Sn
signals (and two 77 Se signals in the selenide case) with intensity ratio 1:2,
showing that the molecule has symmetry. However, this could consist of
either a mirror plane or a twofold axis. In the former case, four 13 C signals
in intensity ratio 2:2:1:1 are predicted, whereas the existence of a twofold
axis would lead to only three signals (with equal intensities). The 13 C
spectrum of the sulfide shows unequivocally that the latter case is the
correct one. NMR can go further and determine the space group of a
crystalline solid. Thus, for the low-temperature phase of zirconium
phosphate, Zr2 P2 O7 , which was known to have a superstructure of the
cubic arrangement of the high-temperature phase (limiting the number of
compatible space groups), advanced NMR techniques (such as those
discussed in section 5.4.4) applied to the 31 P spectrum (figure 8.19 )
shows the existence of 27 peaks arising from 14 diphosphate ions, one (and
one only) of which contained phosphorus atoms related by symmetry. This
information established that the space group is Pbca , enabling the powder
diffraction pattern to be solved to give the full crystal structure.
This synergy between NMR and diffraction can be taken further. For
instance, solving structures from powder diffraction data by simulated
annealing procedures can be done using computer programs that introduce
constraints based on NMR information (interatomic distances from dipolar
coupling constants or knowledge of intermolecular hydrogen bonding from
chemical shifts). In the future, it may prove to be possible to
simultaneously refine structures to fit diffraction information and NMR
spectra. Already, NMR can be used to refine structures obtained from
diffraction data, as in the case of zeolite ZSM-12, where a relatively poor
structure obtained by powder X-ray diffraction was improved to give good
agreement between the principal components of the 29 Si shielding tensors
while retaining acceptable computed diffraction patterns. It has also been
shown that 13 C tensor components for crystalline naphthalene reveal
subtle distinctions between the geometry at different carbon atoms that are
consistent with the diffraction-determined space group, whereas the
diffraction-determined geometry gives no such distinctions outside
experimental error.
missing image file
Figure 8.19 . Phosphorus-31 POST-C7 two-dimensional spectrum of Zr 2 P

2 O 7 , obtained at 121.48 MHz. Horizontal links show that there are 13
dipolar coupled 31 P pairs in nonequivalent sites (once the triple intensity
of the pair at highest double-quantum frequency—indicated by a horizontal
connection—is taken into account). There is also one peak (indicated by the
arrow) arising from a dipolar-coupled pair of equivalent 31 P nuclei. The
one-dimensional projection spectrum is shown at the top. (An adapted
version of this figure appears in I.J. King, F. Fayon, D. Massiot, R.K.
Harris & J.S.O. Evans, Chem. Commun . (2001) 1766.)
An example of the ability of NMR to probe disorder is provided by the 3:2
phenol-triphenylphosphine (TPPO) complex, as mentioned in section 1.3.4.
The diffraction work was unable to determine whether the disorder is
spatial (static) or temporal (dynamic). The reason for the success of NMR
and the failure of diffraction to distinguish the nature of the disorder lies in
the different effective timescales of the two techniques, as mentioned
earlier in this section. Analysis of the 31 P CPMAS NMR spectra as a
function of temperature, as illustrated in figure 8.20 , enables the barrier to
the reorientation process to be determined: ΔH ‡ = 38 kJ mol–1 ; ΔS ‡ = –
23 J mol–1 K–1 . Other examples of the measurement of barriers to
molecular-level motions in solids by NMR have been given in chapter 7 .
missing image file
Figure 8.20 . Phosphorus-31 cross-polarization magic-angle spinning

(CPMAS) spectra of the 3:2 phenol:TPPO complex, obtained at 81.02 MHz
as a function of temperature. Left-hand side: experimental. Right-hand
side: simulated (with the rate constants to site exchange given at the right-
hand side). (Figure adapted, with permission, from D.C. Apperley, P.A.
Chaloner, L.A. Crowe, R.K. Harris, R.M. Harrison, P.B. Hitchcock & C.M.
Lagier, Phys. Chem. Chem. Phys . 2 (2000) 3511.)
Such a case, where the extreme positions are distinguishable, should be

contrasted with situations where molecular-level motion involves mutual
exchange (i.e., where the two or more extremes are indistinguishable). In
these circumstances, diffraction techniques may not detect any mobility at
all. Common examples are internal rotation of a C–C(CH3 )3 group about
the C–C direction and internal rotation of a phenyl or phenylene group
about the ipso -carbon-to-para -carbon axis. If the relevant carbon sites are
nonequivalent, then rapid motion will reduce the number of NMR signals
observed, that is, coalescence phenomena may occur. The example of
formoterol fumarate has been given in section 7.3.1. Several other NMR
techniques are available to quantify the dynamics in such situations. For
instance, if the motional rate is of the same order as the rotation rate of
MAS or of the frequency equivalent of the decoupling power, the resonance
will broaden. Either variable temperature or variable spinning
rate/decoupling power will reveal this phenomenon and can lead to
quantitative evaluation of the barrier to the molecular motion under
consideration. Relaxation techniques also allow measurement of motional
rates in crystalline materials, as already illustrated in figure 7.3 .
In summary, NMR has a significant and increasing role to play in

crystallography, in terms of both static and dynamic structure.
8.8.2 COMPUTATION OF NMR PARAMETERS
The value of NMR in crystallography is greatly increased by the use of

computation to estimate chemical shifts. Such computations have been
carried out for many years on molecules in solution. Advances in the last
decade have enabled use to be made of the spatial repetition inherent in
crystalline solids so as to include the effects of the environment of
molecules or of long-range bonding effects for network systems. For
instance, the question of identifying resonances from the same independent
molecule in α testosterone, which has two molecules in the asymmetric
unit, has been discussed in chapter 5 (see figure 5.19 ). The difficulty was
resolved by use of the INADEQUATE experiment. The major problem
with this technique, at least for 1 3 C, is that it requires observation of
signals from two bonded 13 C nuclei. With natural isotopic abundance, this
involves only ~ 0.01% of the molecules for each atom pair. Therefore,
while setting up the experiment is relatively easy, implementation takes a
lot of spectrometer time (often several days!) and requires efficient proton
decoupling.
This is an issue that computation can help to solve. The gauge-invariant

projected augmented wave (GIPAW) method for computations of
crystalline materials involves density functional theory , plane waves to
cope with the unit-cell repetition, and so-called pseudopotentials to mimic
the effect of the core electrons of atoms. The limitations imposed by the
different conditions involved in the computations and the experiment must
be borne in mind. This implies that the GIPAW results are particularly
useful for comparisons between shifts for the same chemical atom in
different situations (e.g., for two resonances for the same carbon in
different crystallographically independent molecules). For the testosterone
case discussed in section 5.6, the computations reveal which resonances
correspond to the two different molecules in the asymmetric unit. In
consequence, they predict that the separations between the pairs of signals
for C8 and C9 are of opposite sign, which was confirmed experimentally
by INADEQUATE spectra (as mentioned above), though the absolute
splitting magnitudes are not well fitted by computation . Moreover, the
GIPAW computations give extra information since they relate the signals
for the two independent molecules to their sites in the crystal structure.
GIPAW computations can also assist in assignments more generally. Thus,

for form I of finasteride (scheme 8.1), a dispute over the assignment of
signals for C3 and C20 was solved by GIPAW computations (see A.
Othman et al., J Pharm. Sci . 96 (2007) 1380). The latter gave C3 at 161.7
ppm and C20 at 166.6 ppm (a difference of 4.9 ppm), while experiment
showed signals at 164.3 and 169.3 ppm (a difference of 5 ppm in
magnitude). Thus computation showed that the high-frequency signal could
be assigned to C20. This was confirmed by an INADEQUATE experiment
—but 6 days of spectrometer time were needed!
missing image file
Scheme 8.1. Finasteride.
Computations can also establish the assignment of resonances in

ambiguous cases. For instance, form III of phenobarbital (scheme 8.2)
contains two nitrogen atoms which are made equivalent in solution by
molecular-level motion, but the symmetry is lost in the crystalline state (see
figure 8.21 ). Thus, N1 and N3 now have different environments. There are
two observed chemical shifts, at –232.6 and –226.7 ppm. Computation
shows that the latter can be assigned to N3 since the computed shifts are at
–232.7 and –226.6 ppm for N1 and N3 respectively.
missing image file
Scheme 8.2. Phenobarbital.
missing image file
Figure 8.21 . View along the axis of the aromatic group in phenobarbital
(scheme 8.2) form III to illustrate how the nitrogen atoms N1 and N3 are
differentiated in the crystal structure.
FURTHER READING AND SOFTWARE
“Quadrupolar effects transferred to spin- magic-angle spinning spectra of

solids”, R.K. Harris & A.C. Olivieri, Prog. NMR Spectry ., 24 (1992), 435–
456. DOI: 10.1016/0079-6565(92)80004-Y.
“NMR crystallography” , Eds. R.K. Harris, R.E. Wasylishen & M.J. Duer,
John Wiley & Sons Ltd. (2009), ISBN 978-0-470-69961-4.
“NMR crystallography: The use of chemical shifts”, R.K. Harris, Solid

State Sciences , 6 (2004), 1025–1037. DOI:
10.1016/j.solidstatesciences.2004.03.040.
MEASUREMENT OF DIPOLAR COUPLINGS
“Dipolar coupling: Its measurement and uses”, M.J. Duer, Chapter 3 , pp.
111–178 (DOI: 10.1002/9780470999394.ch3) and “Molecular structure
determination: Applications to biology”, O.N. Antzutkin, Chapter 7 , pp.
280–390 (DOI: 10.1002/9780470999394.ch7) of “Solid-state NMR
spectroscopy: Principles and applications” , Ed. M.J. Duer, Blackwell
(2002), ISBN 0.632-05351-8.
“Probing proton–proton proximities in the solid state”, S.P. Brown, Prog.

NMR Spectry ., 50 (2007), 199–251. DOI: 10.1016/j.pnmrs.2006.10.002.
“Measuring heteronuclear dipolar couplings for I = 1/2, S > 1/2 spin pairs
by REDOR and REAPDOR NMR”, T. Gullion & A.J. Vega, Prog. NMR
Spectry ., 47 (2005), 123–136. DOI: 10.1016/j.pnmrs.2005.08.004.
SOFTWARE TOOLS FOR THE FITTING OF NMR TENSOR

INFORMATION
Programs such as WSOLIDS, DMFIT include various models for fitting

anisotropy parameters. General simulation programs such as SIMPSON
and pNMRsim (see appendix D) can be used to fit anisotropy information
including, specialized cases that are not handled by fixed-purpose
programs.
NOTES
1 Including in this term at this point asymmetries which are a form of

anisotropy by a different name.
2 More correctly, the relevant parameter is the modulation of the tensor by

the dynamics, rather than the magnitude itself. For example, there will be
no effect on the observed spectrum if the tensor involved is unchanged by
the dynamic exchange process.
3 Note that the spectra would also be dependent on the relative orientations
of the tensors involved, which highlights the difficulties of quantifying
spectra when multiple interactions are present. Section 8.7 discusses the
question of relative tensor orientations in more detail.
4 The pattern is not the same as the normal Pake pattern illustrated in figure
2.4 for a heteronuclear spin pair (Δ = 2D ) due to the spin-lock conditions
and the n = ± 1 sideband matching. The splitting is even smaller (Δ = D/ 2)
for the n = ± 2 sidebands, making them a less desirable option.
5 More sophisticated schemes for achieving polarization transfer while

suppressing spin diffusion are sometimes used if the dipolar spectrum is
intrinsically sharp (e.g., the PISEMA experiment for biomolecules oriented
in membranes). If multiple IS couplings are present, it is advisable to use a
variant experiment, such as the proton-detected local field (PDLF)
experiment, which allows different IS couplings to be resolved.
6 The circumstances in which this expression is valid are examined in

detail by Duma et al ., ChemPhysChem 5 (2004) 833. Their conclusion was
that this fitting is generally valid even in the presence of much stronger
dipolar and shielding interactions.
7 This implies that it is not experimentally possible to separate D AX and Δ

J AX in this case . However, when D a nd J are not coaxial, equation 8.3 is
no longer valid. In principle, J generally has an asymmetry, which, in the
absence of motion, D does not, but this factor is usually ignored.
8 The way in which the signs of J AX , missing image file , and ζ A

affect the spectrum can be assessed from the equations in inset 8.1. There
are eight possible cases, but they give only two different spectral
appearances.
9 There is a close interaction between chemical exchange and relative

tensor orientation. Chemical exchange can only be observed if the NMR
frequencies are modulated by changes in isotropic frequency and/or tensor
orientation. As a result, experiments such as CODEX (see section 7.3.3)
provide information about relative orientation as well as rates of exchange.
APPENDIX A
THE SPIN PROPERTIES OF SPIN - missing image file NUCLIDES a
Table Missing
Table Missing
APPENDIX B
THE SPIN PROPERTIES OF

QUADRUPOLAR NUCLIDES a
APPENDIX C
LIOUVILLE S PACE , R
ELAXATION & E XCHANGE
C.1 INTRODUCTION TO LIOUVILLE SPACE
There are a number of important processes that cannot be described in

terms of the evolution of a density operator under a Hamiltonian, as given
by the Liouville–von Neumann equation (equation 4.50). The most
important of these are relaxation of the nuclear spin coherences and
chemical exchange. The reason why they cannot be accounted for in the
nuclear spin Hamiltonian is their dependence on the external degrees of
freedom that were discarded in reducing the total system Hamiltonian to
one involving just the nuclear spins, as discussed in section 4.1. NMR
relaxation, for instance, is driven by molecular motion, while the derivation
of the nuclear spin Hamiltonian assumes molecular motion to be much
faster than the NMR timescale and the time dependence could be ignored.
Relaxation and other effects involving the non-spin coordinates could, of
course, be included by working with the full Hamiltonian, but fortunately
there is a much easier route. This section gives an overview of the
theoretical tools required to treat these problems in solid-state NMR, with
the emphasis on providing an intuitive picture of what is involved rather
than a detailed treatment. See the Further reading for more thorough
treatments.
The key to accounting for terms that cannot be included in the nuclear
spin Hamiltonian is switching from the normal Hilbert space , where the
number of degrees of freedom is the size of the Hamiltonian eigenbasis, to a
much larger Liouville space in which the different coherences (or elements of
the density matrix) effectively become the eigenbasis. Considering a single
spin- for example, there are just two states, in the Hilbert
space, and the density matrix has four elements,
Hence the corresponding Liouville space
has four degrees of freedom, with the basis
where represents the coherence
between states j and k . A density matrix can then be represented in
Liouville space as a simple vector of coefficients, for example, (0, 1, 0, 0)
would correspond to one unit of population in the coherence.
Transformations in this Liouville space therefore involve 4 × 4 matrices,
which may seem an excessive number of matrix elements to describe the
dynamics of a single spin!
C.2 APPLICATION TO RELAXATION
The equivalent of Hamiltonian in Liouville space is the Liouvillian which is

written the double hat being used to denote a superoperator . While a
normal operator acts on a wavefunction, or (equivalently) a set of
coefficients corresponding to the eigenbasis, a superoperator acts on the
density operator itself and is therefore able to change the density operator in
more radical ways. If the Hamiltonian operator is that is, rotation
about z with frequency Δ , the corresponding Liouvillian has the matrix
representation1
C.1
The equivalent of the Liouville–von Neumann equation in Liouville space

is
C.2
which has the solution
C.3
is a propagator in Liouville space. Note how this acts directly on

rather than as a similarity transform (as is the case in Hilbert space). Since
is already diagonal, the propagator is just:
C.4
The evolution of our initial density matrix, (0, 1, 0, 0), in Liouville

space is simply
C.5
that is, the coherence oscillates with a frequency Δ . Note how the
evolution of the density matrix can be read directly from equation C.1: the
populations, and , do not evolve, while the and
coherences evolve at frequencies of Δ and –Δ , respectively.
The Liouville space treatment is a rather extravagant way of describing
free precession. Consider, however, the Liouvillian and propagator:
C.6
hence
C.7
Now the evolution of the density matrix will be
C.8
that is, the coherence decays with rate R 2 ; this is simply T 2

relaxation.2 Evolution under the Hamiltonian and relaxation can be
naturally combined
C.9
where is a relaxation matrix as in equation C.6. Note how will give rise
to oscillations of the coherence amplitudes (due to the factor of i), while

will lead to exponential decays in the coherence amplitudes.
C.3 APPLICATION TO CHEMICAL EXCHANGE
In problems involving slow exchange (i.e., dynamics that are much slower
than the scale of the NMR interactions), it is often possible to avoid
Liouville space treatments and simply consider the exchange of z
magnetization during magnetization transfer or EXchange SpectroscopY-
like experiments. If we consider, for example, a system in which z
magnetization is exchanging between three sites, A, B, and C, with the
following rates for individual transfers
where k AB is the rate of transfer from A to B. The overall rate of change of

the magnetizations can be written down using simple chemical kinetics. For
example,
C.10
The set of overall rate equations for M A , M B , and M C are elegantly

summarized in matrix form
C.11
where M is the vector containing the site magnetizations, M k , and K is the
exchange matrix . Note how the structure of K indicates which sites are
exchanging, and how mass balance means that each column must sum to
zero. Note also that K is not generally symmetrical, and care must be taken
not to confuse rows and columns, forwards vs . backwards exchange rates,
etc.
Equation C.11 has exactly the same form as equation C.2 and again has
the solution
C.12
This is an elegant solution to an apparently complex problem involving

coupled differential equations. These solutions will have the form
C.13
where Λ j are the eigenvalues of K , and the coefficients A jk are determined

by its eigenvectors and the initial populations, M (0). As a result of the mass
balance, one of these eigenvalues must be zero, and the corresponding
eigenvector gives relative populations at equilibrium, t → ∞ .
T 1 relaxation can be straightforwardly included by adding a relaxation
matrix, R , to the Liouvillian,3 L = K + R, that is,
C.14
where R k is the rate of T 1 relaxation for site k .

As a second example, we consider two-site exchange on the intermediate
timescale , in which the motion occurs on the same timescale as differences in
the NMR frequencies. The Hamiltonian and kinetic exchange matrix are
C.15
where ΔA and ΔB are the resonance frequencies of the two sites.

In this case, the xy magnetization is of interest, rather than the z
magnetization. Since the coherence orders do not mix under free
precession, we do not need the full Liouville space but only the observed
coherences for the two sites. Hence the Liouvillian for this subspace
is4
C.16
Simplifying the problem to symmetrical two-site exchange by writing k = k

AB = k BA , the eigenvalues are:
C.17
In the slow exchange limit, k ≪πΔ, the eigenvalues reduce to
C.18
which corresponds to the A and B resonance frequencies damped by a

decay rate k , that is, the lines broaden as the exchange rate increases.
FURTHER READING
Chapter 2 of “Principles of nuclear magnetic resonance in one and two dimensions” , R.R. Ernst, G.
Bodenhausen & A. Wokaun, Oxford University Press, (1990), ISBN 0 19 855647 0.
NOTES
1 Formally is the superoperator form

of the Hamiltonian. 1 is an identity matrix whose order matches the Hilbert space and ⊗ represents
the Kronecker (direct) product.
2 Some subtle details about whether the “zero” density matrix refers to the equilibrium density
matrix (equation 4.46) or its reduced version (equation 4.48) are being glossed over here.
3 We are making the common assumption that the exchange process is instantaneous on the NMR
timescale and does not directly influence the nuclear spin dynamics, that is, at one moment the
nuclear spin is in site A and evolving under its Hamiltonian and relaxation parameters, and at the
next moment it is evolving with the parameters of site B.
4 This is identical to the matrix that would be obtained by simple modified Bloch equation treatments
of the exchange of xy magnetization.
APPENDIX D
INTRODUCTION TO SOLID -
STATE NMR SIMULATION
The quantitative analysis of solid-state NMR experiments often requires the

simulation of NMR data, for example fitting a set of spinning sidebands to
anisotropy parameters, as discussed in section 8.2.3. Simulations are
invaluable when developing more complex experiments to verify that the
NMR response will be as expected.
This section sets out the general principles involved in the numerical
simulation of solid-state NMR experiments and the various parameters
involved. Simulation programs (a number of which are listed at the end of
this appendix) come in a variety of forms: some are specialized for
particular tasks, for example sideband analysis, while others are more
general, but the same principles apply, and it should not be difficult to adapt
the discussion to a particular program.
D.1 SPECIFYING THE SPIN SYSTEM
The first step in any simulation is specifying the set of nuclear spins and
their NMR interactions needed to describe the problem at hand. Calculation
times typically increase by almost an order of magnitude for each additional
spin, and using the smallest spin system that reproduces the experimental
results is vital to obtaining results in a reasonable length of time (as well as
simplifying the resulting spectra). Fortunately it is rarely necessary to
specify more than a couple of spins. For instance, the spectra of
quadrupolar nuclei are dominated by the quadrupolar interaction and the
coupling to other nuclear spins can often be ignored. Hence, a simple one-
spin simulation involving just the quadrupolar interaction (and possibly the
chemical shift anisotropy (CSA), depending on its size) should be sufficient
to model the spectrum. Similarly, when fitting the 13 C CSA pattern, it is
not necessary to consider coupling to surrounding protons. Although
residual dipolar coupling will broaden the 13 C resonances, this does not
need to be accounted for explicitly. Simulations involving abundant spins
are considerably more challenging; calculating the lineshape associated
with a set of coupled spins necessarily involves using a sufficiently large
number of spins (typically about 10) in order to reproduce experimental
results.
As an example, the spin system for a two-spin heteronuclear problem
may be expressed (SIMPSON and pNMRsim) by
spinsys {
nuclei 13C 1H
dipole 1 2 -20e3 0 0 0
shift 1 10p 50p 0.5 0 30 0
}
where dipole specifies the coupling between the spins to be 20 kHz, with its
principal axis system aligned with the molecular axis system, that is with
Euler angles Ω MP = (0, 0, 0). The shift line specifies the isotropic chemical
shift to be 10 ppm and the CSA to be 50 ppm with an asymmetry parameter
of 0.5 and with a PAS tilted by 30° away from the molecular frame. Note
how the anisotropy information, including relative orientations, must be
fully specified for the calculation to be properly defined, which is another
motivation for using the smallest possible spin system!
D.2 SPECIFYING THE POWDER SAMPLING
Most solid-state NMR experiments involve microcrystalline powder

samples. Hence, it is necessary to average the results of calculations
performed for a number of crystallite orientations. As the time taken is
proportional to the number of orientations in the powder averaging, it is
important to choose the number of sampling points appropriately,
particularly if calculating a single response is time consuming. If
insufficient points are used, however, individual subspectra from the
different orientations will be observable and the sum spectrum will have an
unphysical jagged appearance. The optimum number of powder steps is
best determined via trial calculations —increasing the number of
orientations until the calculated spectra are unchanged—but, as a rule of
thumb, the number of orientations required is roughly proportional to the
overall widths of the spectral features relative to the intrinsic linewidth.
Hence many more orientations are required to obtain a satisfactory broad
powder lineshape compared with obtaining, say, a satisfactory CSA
spinning sideband pattern under moderately fast magic-a ngle spinning. In
simulations, the intrinsic linewidth is generally added artificially by
apodization of the simulated NMR signal. Stronger damping of the NMR
signal will naturally decrease the spectral resolution, and will also reduce
the number of crystallite orientations required to obtain a smooth powder
lineshape.
There are several schemes for choosing which crystallite orientations to
consider. Like any other numerical integration, there is no single pattern
that converges most quickly to the correct integral for any function. A naive
sampling based on equal angle steps of the α and β Euler angles is not
particularly effective since equal time is spent on the small volume
elements near the pole (β ~ 0) as points near the “equator” (β ~ π /2).
Sampling schemes such as REPULSION distribute sampling points as
evenly as possible across the unit sphere and so are invariably more
efficient. The so-called ZCW sampling schemes in which the powder angles
are incremented simultaneously are also efficient and widely used.
In contrast to static cases, problems involving powders under MAS
generally require integration over all three Euler angles, Ω RM , that connect
the molecular and rotor axis frames, with the γ angle corresponding to the
position of the crystallite around the rotor axis. In sufficiently simple cases
(in particular if time-dependent radiofrequency is not involved), it is
possible to perform the γ angle integration analytically (the “γ -COMPUTE”
algorithm), reducing the requirement for explicit powder angle integration
to just α and β . When this is not the case, the full three angle integration can
be specified in terms of an integration scheme for α and β combined with a
linear increment of γ over a specified number of steps, or, better, using a
sampling scheme specifically designed to integrate over three angles. Hence
in SIMPSON or pNMRsim.
crystal_file zcw132
gamma_angles 8
might be used to specify powder integration over 132 α , β pairs using a

ZCW scheme, plus eight steps in the γ angle (over the range 0–2π ). As a
rule of thumb, the number of γ angles does not usually need to be much
greater than the number of spinning sidebands in simple problems without
RF. By contrast, the NMR response can be strongly dependent on the γ
angle when time-dependent RF is present, and more angles may be required
before the spectrum converges.
Algorithms such as γ -COMPUTE often require certain
“synchronization” conditions to be met. In this case, the number of γ angle
steps must be a multiple of the number of observation points (dwell time
steps) per rotor cycle. In the example above, setting the spectral width to
eight times the MAS frequency would ensure that the dwell time (one-
eighth of the rotor period) will divide into the rotor period (i.e., that the
spinning and observations periods are synchronized) and that the γ -
COMPUTE synchronization condition is also met.
D.3 SPECIFYING THE PULSE SEQUENCE
Different simulation systems tend to use very different approaches to

specify the pulse sequence. A simple example is given here to illustrate the
principles; the details will need adapting to the system in use.
channels 1H
…
start_operator 13C:x
detect_operator 13C:+
…
pulse 5 100e3 x
pulse 5 100e3 -x
store XiX
…
pulseq {
acq 0 XiX
}
The different RF channels are specified in the channels directive, while

the pulse directives are used to construct a pulse sequence, here the XiX
sequence for heteronuclear decoupling, which consists of repeated pairs of p
pulses with alternating x and –x phase. The RF nutation rate on the 1 H
channel has been set to 100 kHz and so duration of each p pulse is 5 µ s.
Once constructed, the sequence can then be used in pulseq , which simply
consists of an acquisition (with a zero phase shift of the receiver) in the
presence of the decoupling sequence. Note how the preparation of the initial
13 C magnetization has been skipped in favor of directly specifying the
starting density operator (start_operator ) to be x magnetization.
D.4 EFFICIENCY OF CALCULATION

As when specifying the spin system, keeping the pulse sequence as simple
as possible simplifies the set-up of the simulation and increases the chances
that it will run efficiently. For instance, ideal (delta function) pulses can be
used whenever the effects of the finite amplitude of experimental RF pulses
are not significant. Similarly there is little point in reproducing an
experimental detail, such as two-pulse phase-modulated decoupling, when
this is not of interest. (Because simple continuous-wave decoupling is not
time dependent, it can be simulated with much greater efficiency than a
modulated sequence.) In the example above, the initial state of the density
operator was explicitly set to be x magnetization rather than starting with
equilibrium (z ) magnetization. This simplified the pulse sequence and
potentially allowed the simulation to take advantage of extra symmetries
that would not be present if the initial state of the system had been created
explicitly using pulses. Interpretation of the simulation results is also
simplified by factoring out aspects of little interest.
Avoiding unnecessary time dependencies is important, as a lack of
“synchronization” between the different timescale scales present (e.g.,
spinning period, dwell time, cycle time of RF sequences) greatly
complicates the simulation. For instance, if the dwell time, Δ t , is 8 µ s and
the rotation period, τ r , is 32 µ s, then the signal is detected exactly four
times per rotor period. This means that the evolution is completely defined
by the four propagators U (Δ t , 0), U (2Δ t , Δ t ), U (3Δ t , 2Δ t ), and U (4Δ t
= τ r , 3Δ t ) calculated over one complete rotor period. Efficient algorithms
can be used to compute the free-induction decay (FID) or spectrum given
this set of propagators. By contrast, if τ r = 33 µs, it is not possible to find a
short common period, and it may be necessary to compute a new propagator
for each point of the FID. Since computing the propagators is usually the
time-limiting step, simulations where the synchronization conditions are not
met will be much slower and may also be complicated to program.
A final parameter that determines the total time required for calculation
is the time step, often called maxdt , which is used when determining
propagators for time-dependent homogeneous Hamiltonians (see inset 4.5).
In these cases, it is necessary to compute the propagators by breaking the
evolution into small steps over which the Hamiltonian is assumed to be
approximately constant. The propagators for these small steps are
multiplied together to give the overall propagators. A coarse time step
improves the speed of calculation, but too large a step will introduce
numerical errors; the time step needs to be small enough to capture the time
dependence of the density matrix. If, for instance, the evolution is driven by
a time-dependent dipolar coupling of 20 kHz, then a time step of 10 µ s
should be adequate to calculate the density matrix evolution without
significant numerical error. On the other hand, if the evolution is driven by
time-dependent RF with amplitude of 200 kHz, a much shorter time step,
say 1 µ s, would be required.
FURTHER READING
“Numerical simulation of solid-state NMR experiments”, P. Hodgkinson & L. Emsley, Prog. Nucl.
Magn. Reson. Spectrosc ., 36 (2000), 201. DOI: 10.1016/S0079-6565(99)00019-9.
“Computer simulations in solid-state NMR”, Mattias Edén, Concepts Magn. Reson ., 17A (2003),
117; 18A (2003), 1; 18A (2003), 24. DOI: 10.1002/cmr.a.10061; 10.1002/cmr.a.10064;
10.1002/cmr.a.10065.
SIMULATION PROGRAMS
Simulation software is constantly evolving, so the list here represents a

snapshot of general simulation programs at the time of writing. More
specialist programs, for example for fitting spinning sideband data and 2D
spectra are also available, such as wSOLIDS and Dmfit . All the packages can
be found straightforwardly by a web search for the package name plus the
keyword NMR.
GAMMA provides a C++ library (with Python interface) for building
NMR simulations. Unlike other systems listed here, it provides a
framework rather than a ready-to-use simulation system and requires
significant programming skills, particularly as its functionality is mostly
directed towards solution-state NMR and relaxation.
pNMRsim is the general simulation program used in the Durham NMR
group and which is optimized for multi-spin problems.
SIMPSONis a popular general simulation program for solid-state NMR.
SIMPSON is relatively easy to use and has a large, established user base.
SPINEVOLUTION is another general simulation program for solid-state

NMR. Although specifying pulse sequences can be tricky, it handles multi-
spin problems efficiently.
Index
Absolute intensity. See Quantitative
Accumulation time, 8
Acetonitrile, 240
Acquisition
time, 8, 41, 45, 47 , 53, 90, 110, 128, 225
window, 134
Activation
energy, 182 , 200
enthalpy, 182
entropy, 182
Adamantane, 8, 119, 205
Alanine, 3, 161
Aliasing, 105. See also , Folded peaks
Aluminium-27, 27 Al, 152, 156, 165, 170, 217, 225
Aluminium acetylacetonate, 152, 154, 155, 165, 170, 217
Aluminophosphate, 224, 225
Aluminosilicate, 156
Aminobenzoic acid, 240

Amorphous
component/domain/region, 11, 127, 183–185, 192–193
examples, 10–11, 127, 183–185, 192–193
preferred orientation, 11
systems/solids, 10–11, 117
relaxation of, 183–185, 192–193
Amplitude
of motion, 201
of RF or pulse, 43, 60–61, 113–114, 211
of signal, 47, 113, 210–211, 244
Analysis
of central transition, 153
of spectra, 226, 232, 237, 244
of spikelet envelope, 158
of spinning sidebands. See Spinning sideband
Angular momentum, 73, 76, 82, 92
Anisotropic
diffusion, 203–204
interactions, 2, 32, 38, 97, 132, 201, 227, 237
motion, 25, 178, 201

Anisotropy/Anisotropies, 3, 20 , 33, 102, 159, 213, 215, 218–222
of bulk magnetic susceptibility (ABMS), 117
of chemical shifts. See Chemical shift anisotropy
of hyperfine coupling, 211
of indirect (J) coupling, 24, 220, 228, 236
of quadrupolar coupling/EFG, 28 , 35, 142, 147, 217
of shielding. See Shielding anisotropy
and tensor interplay, 228–233 , 236
Apodization, 47, 260
Arcing, 53–54
Arrhenius plot, 182, 200
Artefacts in spectra, 47–48, 110, 133,
Asymmetric unit. See Crystallographic asymmetric unit
Asymmetry, 20 , 99, 102
of indirect (J) coupling, 24
of quadrupolar coupling, 28–29 , 143, 147, 154–156
of shielding/chemical shift, 21–23, 33, 216–218, 220, 260
Auto-correlation peak. See Cross peak
Average Hamiltonian Theory (AHT). See Hamiltonian
B
Barbital, 6, 7
Background signal, 69
Bandshape, 186, 192, 207
for dipolar coupling, 186
effect of motion on, 187, 201
for quadrupolar coupling, 32, 152–154, 166–167, 167, 199, 201, 217, 238
for shielding, 23
Baseline, 48–51 , 206, 218
Biomolecules, 138, 223
Bloch-Siegert effect, 69, 80
Boltzmann distribution, 18, 46, 84
Bromine-79, 79 Br, 145
Bulk magnetization, 18, 40, 42, 84
Bulk magnetic susceptibility, 117
Carr-Purcell-Meiboom-Gill. See CPMG
quadrupolar variant. See QCPMG
Centerband
identification of, 51–52 , 206, 221, 231–232
intensity/linewidth of, 111, 132, 202, 206

in Hartmann-Hahn matching, 113
at rotational resonance, 226
of quadrupolar transition, 151, 155–156, 173
Centerband Only Detection of Exchange (CODEX). See E xchange
Central transition
energy levels, 29–30, 146, 149, 159, 169
spectra, 32, 143–144, 149–157 , 159, 170, 217, 218, 237
Center of gravity, 151, 153, 169, 170, 233
Chemical shift, 4, 21, 72, 81, 85, 88, 158, 177, 212, 215, 228, 241. See also
Shielding
anisotropy (CSA), 99, 102, 110, 132–133, 159, 201, 221, 224, 227, 259–
260. See also Shielding anisotropy
computation, 244
reference materials, 142, 168, 247–251
referencing, 54
Cholesteryl acetate, 126
Clipping of the signal, 48
Coalescence, 198 , 243
CODEX. See Exchange
Computer-simulation/computation
of shielding/chemical shifts, 244–245

of spectra, 34–35, 145, 156, 165, 172
Co-crystal, 13, 242
Coherence
double-quantum (DQ), 128–131, 242
heteronuclear multiple-quantum (HMQC), 126
in Liouville space, 253–257
order, 83–84, 88, 93, 163–164
pathway, 88–91, 163–166, 172
selection, 63
single-quantum (SQ), 159–160, 163, 172
triple-quantum (TQ), 159, 163
zero-quantum (ZQ), 93
Combined Rotation and Multiple-Pulse Spectroscopy. See CRAMPS
Contact time, 59–60, 62–66 , 114, 125, 208
delayed, 193, 195
variable, 195–196, 208–210 , 223, 224
Continuous wave decoupling. See Decoupling
Correlation experiments/spectra
EXSY, 199–202
in general, 120–122
heteronuclear. See HETCOR, WISE
homonuclear. See Homonuclear correlation
MQMAS, 163, 239
for relative tensor orientation, 237, 239
Correlation function/time, 179–182 , 201, 203
Cortisone, 112
Coupling
dipolar, 20–21, 24–28 , 78, 91–94. See also Dipolar coupling
Hamiltonians, 77–79, 94–96
heteronuclear dipolar, 25–27, 102, 124, 222–225 , 229–230
homonuclear dipolar, 26, 35, 102, 107, 116, 180, 186, 196, 201, 216, 225–
228
hyperfine. See Hyperfine coupling
indirect (J), 20–21, 24, 117–120, 126, 130, 228, 231, 239
quadrupolar. See Quadrupolar coupling
CP. See Cross polarization
CPMAS, 15, 58–66. See also Cross polarization, Magic-angle spinning
CPMG, 188
CRAMPS, 133, 135
Cross peak, 123, 124, 129–131, 224, 238
auto-correlation, 129–131
Cross polarization (CP), 15, 58–66 , 95, 113–115 , 121, 123–125, 158, 191,
208, 222–223. See also Lee-Goldburg
1 H to 19 F, 67
contact time. See Contact time
contact pulse, 59
matching, 60, 113–114
phase cycling, 61–62
ramped, 114–115
spin-lock, 59–60
transient oscillations in build-up, 64, 96, 224
vs. direct excitation, 58, 66
Cross relaxation (heteronuclear), 195
Crystal
lattice, 8
structure, 6–8, 12, 145, 157, 240–242 , 244–245
Crystalline
component/domain, 11, 19, 127, 183, 184, 191
state/form, 4–6
Crystallographic asymmetric unit, 5 , 137, 190, 244
Crystallography, 27, 239–245
CSA. See Chemical shift anisotropy

CW. See Decoupling
DAS. See Dynamic angle spinning
Dead time, 49–52 , 164
Deconvolution, 207 , 219
Decoupling/Decoupling sequences, 52–54 , 60
and average Hamiltonian theory, 107
BLEW-12, 119
BR-24, 134
continuous wave (CW), 53–54 , 111, 112, 119, 260
high-power proton decoupling (HPPD), 14, 138
heteronuclear, 52–53, 111–113
homonuclear, 67–68, 104–105, 107, 119–120, 127–129, 132–136 , 138
interrupted, 189–190, 240
Lee-Goldburg (LG), 134–136
and linewidth, 110, 116
multi-pulse. See CRAMPS
power, 53, 244
SPINAL64, 112
two-pulse phase-modulated (TPPM), 54, 111–115

WHH-4 (WAHUHA), 104 , 134
XiX, 112, 262
Delayed contact, 195
Density Functional Theory (DFT), 244
Density matrix, 81–85 , 88, 91–97, 101, 105, 255, 263
Density operator, 76, 82, 85–91, 97, 102, 105, 253–254
reduced, 84
Dephasing, 90, 224. See also Dipolar dephasing
Deuteration/deuterated, 15, 138, 144, 174
Deuterium, 2 H, 16, 31–32, 141, 144, 174, 201–202
Dicyclopentadienyl zirconium dichloride, 34
Diffusion, 203–204. See also Spin diffusion
Digitization, 47–48, 143
Digital resolution, 67
Dilute nuclides, 15 , 58
Dipolar coupling. See also Coupling
coupling constant, 25–27, 78, 92, 217, 222, 229, 234
measurement of, 222–227
recoupling of. See Recoupling
Dipolar dephasing, 116, 119, 136, 189. See also Decoupling, interrupted
Direct dimension, 120–122 , 129, 134, 164
Direct excitation, 44 , 46, 58–62, 66, 157, 208
Disorder/Disordered systems, 9–11 , 156, 205, 224, 242
Domain, 11, 183, 184, 192, 209. See also Heterogeneous materials
Double quantum. See Coherence
Double rotation (DOR), 161–162 , 173
DUMBO, 134
Duty cycle, 53 , 57, 225
Dwell time, 49–52 , 93, 120, 121, 261
Dynamic-angle spinning (DAS), 162–163 , 173
Dyson time-ordering operator, 103
Echo
simple (Hahn), 88–91 , 118–119, 188, 222–224, 227
rotary, 50–51 , 101, 123
shifted, 169
solid (quadrupolar), 51 , 88, 91, 158, 172, 185
Editing experiments, 69, 189–190
Electric field gradient (EFG), 21, 28 , 147–153, 233
Electric quadrupolar coupling.

See Quadrupolar coupling
Electric quadrupole moment, 28, 141
Enantiotropic system, 12
Energy level, 22, 29–30, 32, 83
13 C, 14 N system, 234
quadrupolar, 146–147, 149, 157, 166
Energy of activation. See Activation
Equivalence, 4, 131
ERETIC, 210
Ernst angle, 47, 157
Euler angles, 97–101 , 106, 237, 260–261
Evolution, 123, 128, 224, 225, 227, 239
of the density matrix, 84, 88, 91, 101, 103, 105–106, 255, 263
in indirect dimension, 120
of signal, 50, 119, 121
time, 120, 164, 199
Exchange
chemical/positional, 9, 124, 196–200 ,

216, 245
CODEX, 202–203
EXSY, 199–202 , 256
of magnetization, 123, 226, 256–257

rates, 139, 196, 198–199, 202, 243
EXSY. See Exchange
Extreme narrowing limit, 182
Fermi contact interaction, 212
Fibres (preferred orientation), 11
FID. See Free-induction decay
Field strength (of RF), 43 , 52, 96, 195–196
Finasteride, 110, 174, 206, 244
Flip-back. See Pulse
Flip-flop term, 79, 97, 102
Floquet theory, 105–106
1-Fluorononane, 13
Folded peaks, 49, 121
Formoterol fumarate, 197, 200
Fourier transform/transformation, 41, 47–48, 50, 61, 67, 132
of correlation function, 179
double, 120 , 159, 162–164, 172, 223
Free-induction decay (FID), 8, 36, 41 , 44–50, 53, 67, 145, 157, 162, 172,
188, 195, 262
Free-precession period, 87 , 89–90, 108, 255

Frequency domain, 41
Frequency-stepped experiment, 158
Frequency Switched Lee-Goldburg (FSLG), 135
2, 3-d 2 -Fumaric acid, 144
Gauge-Invariant Projected Augmented Wave (GIPAW), 244
Glass transition temperature (T g ), 10 , 183
Glycine, 55, 189
Goldman-Shen experiment, 192
Goniometer, 220
Gypsum, CaSO4 ·2H2 O, 26, 239
Hamiltonian, 71–76
average Hamiltonian theory (AHT), 101–105
chemical shift. See Hamiltonian, for shielding
for cross polarization, 95
and the density matrix, 85–98
for dipolar coupling, 79, 94, 123
homogeneous, 100, 102, 262
under homonuclear decoupling, 132

inhomogeneous, 100, 102
for isotropic J coupling, 72, 78
and Liouville space, 253–257
for quadrupolar coupling, 146–148
for radiofrequency irradiation, 117–119
for shielding, 75, 227
Zeeman, 73 , 77–78, 80
Hard pulses. See Pulses
Hartmann-Hahn cross polarization. See Cross polarization
Hartmann-Hahn matching, 60 , 111, 222
at high spin-rate, 113
theoretical treatment, 96
Heavy metals, 14–15, 24, 34, 49, 65–66.

See also Tin-119/7
HETeronuclear CORrelation (HETCOR), 68, 123–126 , 192
Heteronuclear Multiple Quantum Coherence. See HMQC
Hexamethylbenzene (HMB), 60–61, 114
Heterogeneous materials, 10, 35, 66, 131, 192, 206, 242
relaxation in, 183, 184
Hg(SCN)2 , 51
Hilbert space, 253

Hindered rotation, 203
High-field approximation, 73 , 97
High-power proton decoupling.

See Decoupling
HMQC, 127
Homonuclear correlation, 127–131
DQ/SQ, 128, 130, 138
INADEQUATE, 128–129 , 137, 244, 245
using POST-C7, 129
using RFDR, 127
Homonuclear decoupling. See Decoupling
Host, guest systems, 13–14. See also Finasteride; Urea inclusion systems
HPPD. See Decoupling
HRMAS. See Magic-angle spinning
Hydrogen bonding, 13
intermolecular, 120, 136, 244
Hyperfine coupling, 211–213
Ibuprofen, 182
INADEQUATE. See Homonuclear correlation
Incommensurate structures, 10
Indirect coupling. See Coupling
Indirect dimension, 120–122 , 129
homonuclear decoupling in, 134
and MQMAS, 163, 167
tensor orientation from, 239
Indomethacin, 27–28
Inhomogeneity (RF/B1 ). See RF inhomogeneity
Integral spins, 16, 174
Interactions, 2, 18–32 . See also Hamiltonian
chemical shift. See CSA
dipolar. See Coupling
effect of MAS on, 14–15, 32–35 , 102
effect of motion on, 4, 27, 37. See also Exchange
and equivalence, 131
frame, 79 , 103–104
homogeneous, 33, 35, 102, 216
indirect., See Coupling
inhomogeneous, 33, 102, 216
internal, 2, 18
interplay between, 218–221, 229

in liquid crystals, 11, 27
quadrupolar. See Coupling
shielding. See Shielding
solid vs. solution state, 3, 4, 111
tensor description of, 18
time-dependency and relaxation, 37
Zeeman. See Zeeman interaction
Interference of timescales, 136, 195,

205–206
Interferogram, 41
Internuclear distance, 25
measurement of, 123, 222–226 , 236
Interrupted decoupling. See Decoupling
Inversion-recovery, 194–195
Irreducible spherical tensor operators, 106
Isoleucine, 208
Isotopic enrichment, 15, 27, 138
Isotropic average/component
of bulk magnetic susceptibility, 117
of indirect (J) coupling, 72, 131, 228,

229, 237
of interactions, 20, 77, 99, 131
Isotropic chemical shift/shielding, 21
and exchange, 199
and LG decoupling, 135
for quadrupolar nuclei, 144, 147, 152, 154, 166, 170
resolving anisotropy information by, 220
Isotropic dimension, 166, 169, 220
labelling of in MQMAS spectra, 169
Isotropic motion/tumbling, 4, 32, 37, 201
of adamantane 8, 205
Isotropic second-order (quadrupolar) shift, 153, 156, 161, 162, 172
Isotropic spin-diffusion constant, 184
J coupling. See Coupling, indirect
Knight shift, 23
Labelled systems. See Isotopic enrichment
Larmor frequency, 18 , 22, 37, 80, 180

Lead nitrate, 3, 21, 56
Lee-Goldburg, 68
cross polarization, 125, 223
decoupling, 134
Legendre polynomials, 78, 101, 150, 153
Lithium-7, 7 Li, 203
Linebroadening. See Linewidth
Lineshape, 112, 115

1 H non-spinning, 186, 189
in 2D spectra, 120, 164, 165
for adamantane, 205
under exchange, 196
Gaussian, 209, 220
Lorentzian, 209, 220
modelling of, 217, 219, 260
quadrupolar, 164–166 , 237
rotational resonance, 131
Linewidth, 110, 115–117 . See also Resolution

1 H, 66, 126
19 F, 66–69
due to angle missetting, 145
as a function of decoupling, 112
effect of (static) disorder on, 9
variation with field, 109
homogeneous, 35, 115–116 , 132, 228
inhomogeneous, 115–117
effect of interference on, 205–206
intrinsic, 116
MAS rate in relation to, 33–34, 132
effect of motion on, 203–204
for non-spinning samples, 186
arising from paramagnetism, 211
simulation of, 260
T2 in relation to, 36, 119
Line separation. See Resolution
Liouville space, 108, 253–257
Liouville-von Neumann equation, 85,

253–257
Liquid crystalline state, 11
Local field, 110
electric, 179
magnetic, 178
and relaxation, 178–180
separated (SLF), 222, 223
Longitudinal relaxation. See Relaxation
Lowering operator. See Step down operator
Low-gamma nuclei, 52
Magic angle. See also Decoupling, Lee-Goldburg
definition of, 33
setting, 145, 170
Magic-angle spinning (MAS), 6, 14–15, 32–35 , 102

1 H spectra, 67–68
2 H spectra, 201–202
CRAMPS, 133
at high field, 110–111
at high spin rate, 67, 111, 113, 115
HRMAS (high resolution MAS), 68–69
interference with decoupling, 136, 205–206
in paramagnetic systems, 213
for quadrupolar nuclei, 35, 143, 148, 150–156

rotor synchronization with. See Rotor synchronization
simulation under, 259–260
suppression of dipolar coupling, 123, 126
temperature calibration under, 56
theory for 101–106
Magic-angle turning. See MAT
Magnetic dipolar interaction. See Coupling, dipolar
Magnetic quantum number, 17
Magnetic moment, 17
Magnetization exchange. See Exchange
Magnetization transfer. See also Exchange; Recoupling
in correlation measurements, 122–123
cross polarization, 58, 124
TEDOR, 224
Magnetization vector, 41–42, 81. See also Bulk magnetization
Magnetogyric ratio, 2, 247–251
Magnus expansion, 103
Matrix. See also Density matrix
algebra, 78, 86
exchange, 256
representations of operators/Hamiltonians, 74–75
relaxation, 255
representation of tensors, 18
MAS. See Magic-angle spinning
MAT, 220
MELODRAMA, 227
Methanol, 238
3-Methoxybenzoic acid, 125, 221
5-Methyl-2-diazobenzenesulfonic acid hydrochloride, 226
Microcrystalline. See Polycrystalline
Mixing period/time, 120–125 , 131
EXSY, 199, 238
Molybdenum pyrophosphate, 118
Motion/motional averaging. See also Anisotropic, Exchange, Disorder
effect on cross polarization, 65
diffusive, 203–204
reduced dipolar coupling, impact on spectra of, 53, 187, 190, 224
interference effects on spectra, 196, 205–206
in liquid crystals, 10, 27
effect on quadrupolar spectra, 174

rate of, 8, 177–179
relaxation, 37, 116, 157, 178–183
reorientation, 201–203
in soft solids, 69, 204–205
effect on spin diffusion, 184
effect on tensors 21, 25, 27
MQMAS, 163–169
axis labelling for, 169
experimental setup, 168
Multi-dimensional NMR (in general), 120–122
Multiple-quantum coherence. See Coherence
Multi-pulse decoupling. See Decoupling
Multiple quantum magic-angle spinning.

See MQMAS
Multiple-quantum excitation, 158, 168.

See also Coherence
Naphthalene, 242
Nitrogen-14, 14 N, 142, 174–175, 234–235
Nitrogen-15, 15 N, 14, 58, 224, 235
NMR crystallography, 27, 239–244

NMR timescale, 8, 22–23, 211, 271
Non-quaternary suppression. See , Decoupling, interrupted
Non-secular term, 74–75
Non-spinning samples. See Static samples
Nuclear electric quadrupolar moment, 16, 28 , 141, 249 –251
Nuclear quadrupole coupling constant.

See Coupling, Quadrupolar
Null time (inversion recovery), 194
Nutation
axis (homonuclear decoupling), 135
curve, 42 , 168
for quadrupolar nuclei, 159
spectrum, 159
Nutation frequency/rate, 41, 43, 81 , 95, 205
cross polarization, 60, 95, 113
heteronuclear decoupling, 53, 111–112
homonuclear decoupling, 135
T 1ρ relaxation, 180
Oblate quadrupole moment, 141
Octosilicate, 124
Off-resonance, 44 , 110. See also Decoupling, Lee-Goldburg
detection, 132
Organometallics, 212
Oversampling, 49
Oxygen-17, 17 O, 161
Pake doublet/pattern, 26 , 32, 217, 223
Paramagnetic, 2, 23, 66, 178, 211–213
PASS, 221
PDLF, 223
Perturbation theory, 75, 80, 146, 234, 236
Phase Adjustment of Spinning Sidebands.

See PASS
Phase correction, 50, 51, 189
Phase cycle(s)/cycling, 61–62 , 91, 128, 163–164, 172
Phase Modulated Lee-Goldburg. See PMLG
Phase modulation,
decoupling, 111, 135
in 2D NMR, 120, 162
Phenobarbital, 240, 245

Phosphorus-31, 31 P, 9, 15, 35, 42, 61, 66, 117, 128, 132, 219–220, 222,
224, 228, 241–243
Phenol, 9
PISEMA, 223
PMLG, 135
Polarization transfer, 58, 65, 125, 223.

See also Cross polarization
Polycrystalline, 3, 13, 22, 26, 30, 33
Polymers, 9, 35, 202, 209
polyethylene, 53, 183–186, 191–193
polymethylmethacylate (PMMA), 203
polyvinylidene fluoride (PVDF), 127
preferred orientation, 10–11, 22
spin diffusion, 183
Polymorphism, 7, 12 , 167
Population-weighted average
frequency, 212
POST-C7, 123, 226, 242
Powder pattern, 3, 14
dipolar coupling, 26, 223
involving multiple interactions, 230,

236, 239
quadrupolar, 30, 153–155, 156, 217
fitting of, 217–218
shielding, 22–23, 218, 220
Precession, 40 , 80, 87–90. See also Free-precession period
tilted axis, 108, 132, 134–135
Preferred orientation. See Polymers
Principal axis system (PAS), 19 , 22, 78, 99, 100, 147–148, 216, 260
Probe ringing, 42
Prolate quadrupole moment, 141
Proton decoupling. See Decoupling
Proton-Detected Local Field. See PDLF
Proton relaxation times, 183–186
Pseudocontact term, 212
Pseudopotentials, 224
Pulse
angle, 41, 43, 47, 88, 111, 157, 159
breakthrough, 50
calibration, 43, 168
delay. See Recycle delay
duration, 42–45, 158, 159, 163, 165, 168

excitation range, 43
flip-back, 65
hard, 43, 159, 163, 172
non-selective, 146, 159, 163, 172
reconversion, 164
rotor-synchronised, 164, 189, 221, 223
selective, 158, 162, 164, 166, 172, 194
soft, 159, 172
theory of, 89–94
Pulsed field gradient, 204
Pulse sequence diagrams
CPMG, 188
cross polarization, 60, 115, 223
delayed contact, 195
EXSY, 199
flip-back, 65
FSLG, 135
HETCOR, 124
INADEQUATE, 128
interrupted decoupling, 189

inversion recovery, 194
MQMAS, 164
Lee Goldburg, 135
quadrupolar echo. See solid echo
REDOR, 223
RFDR, 127
saturation recovery, 194
solid echo, 51, 89
spin echo, 89, 188
STMAS, 172
Torchia, 194
TPPM, 115
Variable spin-lock, 195
WHH-4 (WAHUHA), 104
WISE, 192
QCPMG, 158
Quadrupolar bandshape. See Bandshape
Quadrupolar coupling, 11, 16, 20–21,

28–32 , 141–142, 143–148 , 150, 156,
178, 215, 222
asymmetry of, 147, 148, 154, 236
constant, 142 , 152
for deuterium, 174, 201
effect on intensity, 208
multiple interactions involving, 218, 337–339
quantitation of bandshape, 217
effect on spin-1/2 spectra, 233–236
strength of, 158
Quadrupolar echo, 51, 91
Quadrupolar Hamiltonian. See Hamiltonian
Quadrupolar nuclei,
properties of, 249–251
Quadrupolar shift, 161, 170
Quadrature artefact, 132
Quadrature detection, 61
in 2D NMR, 122
Quantitation
absolute intensity, 209–211
measurement of anisotropies, 215–222
measurement of dipolar coupling,

222–227
measurement of indirect coupling, 227–228
relative intensity, 208–209
of tensor orientation, 220
of relationships between tensors, 236–239
Quaternary carbons, 64, 110, 136, 189, 208
Rabi factor, 158
Raising operator. See Step up operator
Rare nuclides. See Dilute nuclides
Receiver gain, 47–48
Recoupling (dipolar), 107, 123 , 127–130, 223–227
MELODRAMA, 227
POST-C7, 123, 226
for quadrupolar nuclei, 222
REDOR, 223–225
RFDR, 123, 127
Recycle delay, 44–47 , 44, 157, 208, 209,

211, 212
REDOR. See Recoupling
Reference signal. See also Chemical shift
intensity, 208
Relative intensity. See Quantitation
Relaxation, 36–38 , 177–194
1 H and spin diffusion, 183
filtering, 191
and linewidth, 115–117
and Liouville space, 254–255
and the NMR timescale, 8
in paramagnetic systems, 212
and recycle delays, 44–47, 208
spin-lattice (longitudinal) relaxation (T 1 ), 36 , 45–47, 178–186, 193–196,

208
spin-lattice relaxation in the rotating frame (T 1 r ), 36–37 , 62–66, 180–

185, 191, 208–209
T 2 *, 188–192
T 2 ′ , 119, 188
transverse (spin-spin) relaxation (T 2 ), 36 , 115–116, 183, 188
Relaxation agent, 212
Relaxometry, 38
Relay peak, 124
Residual dipolar coupling

effect on linewidth, 116
via quadrupole effect, 233–236
Resolution, 3, 14, 67
for abundant nuclei (1 H/19 F), 66–70 , 132–136 , 138
dependence on magnetic field, 109, 115–116
digital, 67
effect of FID truncation on, 48
effect of heteronuclear decoupling on, 52, 111–113 , 134
in paramagnetic systems, 211–212
for quadrupolar nuclei, 150, 154, 160–162, 166–167, 172–173
RFDR. See Recoupling
RF inhomogeneity, 43 , 113
Rigid lattice limit, 181
Rotary echo. See Echo
Rotating frame, 40–44 , 79–81
Rotational resonance, 129, 131, 226
Rotor angle, 162, See also

Magic angle
Rotor frame, 99, 148, 149
Rotor synchronization, 123, 170, 189,

221, 223
Rubidium-87, 87 Rb, 158, 172
Satellite transitions, 30 , 143, 145, 149, 150, 154, 157, 170–172
Satellite transition magic-angle spinning.

See STMAS
Saturation-recovery sequence, 185, 194–195
Second moment, 187
Second-order quadrupolar effects, 146–175 , 217, 234, 236
Secular approximation/term, 74–77 , 96, 99, 148, 149
Selective excitation (quadrupoles), 159
Sensitivity of NMR, 18
SELTICS. See Spinning sideband

suppression
Separated Local Field. See SLF
Shearing, 165, 166, 169
Shielding, 2, 4, 19–23 , 76
Shielding anisotropy, 14, 21–23 , 33–35, 37, 159, 216, 218, 220–222, 230,
237
Shielding asymmetry. See Asymmetry
Shielding tensor. See Tensor
Sideband matching, 113–114

Sign discrimination. See Quadrature detection
Signal-to-noise ratio, 45
Silk, 11
Sildenafil citrate, 14
Simulation/Simulated spectra, 35, 162, 201, 207, 217, 237, 241, 243, 259–
263
Sinc wiggles, 47, 110
Single crystal (NMR of), 3, 6, 25–26, 33,

35, 220
Single quantum. See Coherence
SLF, 222, 223
Sodium-23, 23 Na, 55, 142, 157, 167, 168
Soft solids, 69, 116, 204–205
Solid echo. See Echo
Solvates, 13 , 110, 174, 240
Spectral density/densities, 179–181
Spectral width, 43, 49 , 121–122, 168
Spherical tensors, 97–99
Spikelets, 158
Spin echo. See Echo

Spin diffusion, 35 , 58, 63, 65, 116,
122–126, 131, 136, 138, 183–184,
186, 192, 226
Spin Hamiltonian, 72 , 84, 253
Spin-lattice relaxation (T 1 ). See Relaxation
Spin-lock, 36 , 59–62, 95–97, 114, 136,

183, 193
Spinning sideband, 33–35 , 110,

231–233, 238
and exchange, 201–202
fitting/analysis of, 35, 155, 216,

218–220 , 259
folding of, 49, 121, 174
in quadrupolar NMR, 143–145 , 154, 161, 170, 174
and rotational resonance, 226
effect of spin rate, 33–35
suppression of (SELTICS/TOSS), 52 , 221
theoretical treatment, 105–106
Spin-½ nuclei, properties of, 247–248
Spin operators, 73–74 , 106–107, 148, 151
Spin-orbit term, 212
Spin-spin relaxation (T 2 ). See Relaxation

Spin-spin coupling. See Coupling, indirect
Spin-temperature inversion, 62
Static samples, 14. See also Powder pattern, Single crystal (NMR of)
for abundant spins, 27–28, 34
acquiring wideline spectra from, 50–51
for quadrupolar nuclei, 16, 151–153 ,

158, 174
for measurement of relative tensor orientation, 237–238
for study of relaxation/dynamics, 174, 183–187 , 194, 199–200
Step-up/step-down operator, 73–74 , 234
Steroids, 110, 117. See also testosterone, finasteride
STMAS, 172–173
Styrene/butadiene block copolymer, 209, 210
Sulfathiazole, 13
T 1 . See Relaxation, spin-lattice
T 1 ρ . See Relaxation, spin-lattice relaxation in the rotating frame
T 2 , T 2 ′ , T 2 *. See Relaxation
TEDOR, 224
Temperature calibration, 56–58
Tensor, 18–21 , 78
dipolar coupling, 20–21, 24–25 , 78, 239
indirect (J) coupling, 20–21, 24
interplay, 228–233
motional averaging of, 216
orientation and reference frames, 97,

98, 220
principal components, 19
quadrupolar coupling/EFG, 20–21, 28–29 , 145, 147, 217
relative orientation, 237–239
shielding, 18–23, 76
spherical, 97–99
Testosterone, 5–6, 137, 244
T g . See Glass transition temperature
Tilted axis precession, 108, 132
Time correlation function, 179
Time domain, 41
Tin-119/7, 119/117 Sn, 40, 49, 55, 232, 233, 241
Tip angle. See Pulse angle
TOSS. See Spinning sideband, suppression of

TPPM. See Decoupling
Transient oscillations. See Cross polarization
Transverse relaxation. See Relaxation
Trimethylphenylphosphonium iodide, 231
Trimethylphosphine sulphide, 37
Trimethyltin fluoride, 232
Triphenylphosphine (TPPO), 9, 242
Triple quantum (TQ). See Coherence
Tropolone, 199
Truncation of FID, 47–48 , 53, 110
Two-dimensional NMR (in general), 120–122 . See also Correlation

experiments
Unit cell, 4
Urea inclusion systems, 8, 13–14 , 227.

See also Soft solids
Variable angle spinning (VAS), 160–161
Vector model, 40 , 83, 88
WHH-4 (WAHUHA), 104, 106, 134

Wideline NMR. See Static samples
Wigner rotation matrix, 99
Windowed sequences, 134
Windowless sequences, 134
WISE, 192
Zeeman effect/interaction, 2, 17, 73, 146–148, 233. See also Hamiltonian
z-filter, 166
Zeolites, 8, 13, 39, 212, 242
Zero-filling, 47, 67
Zero-quantum coherence. See Coherence
Zirconium phosphate, Zr2 P2 O7 , 241–242

Other Applied Science and General Titles Certain to Be of Interest
Acoustic High-Frequency Diffraction Theory, Frederic Molinet

Diffuse Scattering and the Fundamental Properties of Materials, R. I.
Barabash, G.E. Ice, P.E.A. Turchi
The Essentials of Finite Element Modeling and Adaptive Refinement:
For Beginning Analysts to Advanced Researchers in Solid Mechanics,
John O. Dow
Polymer Testing: New Instrumental Methods, M. N. Subramanian
Virtual Engineering, Joe Cecil
Social Media for Engineers and Scientists, Jon DiPietro
Professional Expression: To Organize, Write, and manage Technical
communication, M.D. Morris
The Engineering Language: A Consolidation of the Words and Their
Definitions, Ronald Hanifan
Reduce Your Engineering Drawing Errors, Ronald Hanifan
For more information, please visit www.momentumpress.net
____________________________________________
The Momentum Press Digital Library
Engineering Ebooks For Research, Classrooms, and Reference
Our books can also be purchased in an e-book collection that features…
a one-time purchase that is owned forever; not subscription-based

allows for simultaneous readers,
has no restrictions on printing
can be downloaded as a pdf
The Momentum Press digital library is an affordable way to give many

readers simultaneous access to expert content.
For more information, please visit www.momentumpress.net/library and to

set up a trial, please contact mptrials@globalepress.com
cover

Solid State NMR Basic Principles and Practice

Uploaded by

Solid State NMR Basic Principles and Practice

Uploaded by

SOLID-STATE NMR

BASIC PRINCIPLES & PRACTICE

DAVID C. APPERLEY, ROBIN K. HARRIS & PAUL HODGKINSON

All rights reserved. No part of this publication may be reproduced, stored in

First published by Momentum Press®, LLC

ISBN-13: 978-1-60650-350-8 (hard back, case bound)

Cover design by Jonathan Pennell

Printed in the United States of America

2 BASIC NMR CONCEPTS FOR SOLIDS

3 SPIN - FMEQN003.WMF NUCLEI : A PRACTICAL GUIDE

4 QUANTUM MECHANICS OF SOLID -STATE NMR

5 GOING FURTHER WITH SPIN - FMEQN003.WMF SOLID -

8 ANALYSIS & INTERPRETATION

A THE SPIN PROPERTIES OF SPIN - FMEQN003.WMF

B THE SPIN PROPERTIES OF QUADRUPOLAR NUCLIDES

C LIOUVILLE SPACE , RELAXATION & EXCHANGE

D INTRODUCTION TO SOLID -STATE NMR SIMULATION

Solid-state NMR is now too vast and varied to try to present a

We have tried to emphasize the interactions between solid-state NMR and

In summary, making the best use of solid-state NMR involves a good

Many people have contributed to the appearance of this book. Members of

David C. Apperley, Robin K. Harris, Paul Hodgkinson

Robin K. Harris is an emeritus professor at the University of Durham, UK,

Paul Hodgkinson studied chemistry at Queen’s College, Oxford,

1.1 THE UTILITY OF NMR

This chapter gives a brief introduction to solid-state nuclear magnetic

NMR spectroscopy is undoubtedly one of the most powerful techniques for

The first successful NMR experiments on condensed phases (conducted by

In order to understand why NMR of solids is so different from solution-

where h is Planck’s constant, γ j is the magnetogyric ratio of the nucleus

Magnetic fields arising from RF radiation form small perturbations in

where E RF refers to the interaction of the spins with applied

An extra term is required in equation 1.2 when paramagnetic systems are

1.2 A PREVIEW OF SOLID-STATE NMR SPECTRA

In order to give a taste of what is to come in this book, a few examples of

Figure 1.2. Lead-207 NMR spectra of microcrystalline lead nitrate under

1.3 THE SOLID STATE

1.3.2 SYMMETRY IN THE CRYSTALLINE STATE

The concept of a unit cell underlies the translational symmetry inherent in

For NMR purposes, it is important to distinguish between the effects of

Figure 1.3. Schematic illustration in two dimensions of (a) fully equivalent

1.3.3 EFFECTS OF CRYSTAL STRUCTURE ON NMR

Finally, molecular-level motion, which is fast on the NMR timescale

Figure 1.5. Schematic diagram to show the equivalences and

Figure 1.6. Left: Structures of barbital polymorphs, showing the molecular

Inset 1.1. NMR timescales

(c) Times relating to relaxation, ranging from microseconds to seconds (or

It is actually a misconception to think that atoms and molecules in solids

1.3.4 TYPES OF SOLIDS

Yet other structures are classified as incommensurate . They contain

Many pure (homogeneous) systems are amorphous. These may be glassy

Figure 1.8. Carbon-13 NMR spectra of the organic compound nifedipine,

Moreover, even chemically homogeneous solids may be physically

Thus questions of both spatial order/disorder and temporal disorder

Figure 1.9. Classification of different types of condensed-phase matter

1.4 POLYMORPHISM, SOLVATES, CO-CRYSTALS & HOST:GUEST

Chemical compounds frequently exist in more than one crystalline “form.”

All properties of polymorphs for a given compound differ, at least in

Figure 1.10. Schematic phase diagram for an enantiotropic system. The

In addition, many organic compounds form stoichiometric solvates or co-

NMR proves to be one of the best ways of characterizing polymorphs and

Sometimes guest molecules can be incorporated into host crystal structures

1.5 NMR OF SOLIDS & THE PERIODIC TABLE

High-power proton decoupling (HPPD) . A typical problem for observing

Magic-angle spinning (MAS). Another problem of spectral resolution

The difficulties involved in obtaining good-quality spectra of powdered

“Dilute” or “rare” spin- CH_In_01-02 bOLD.wmf nuclides . In these

“Abundant” spin- CH_In_01-02 bOLD.wmf nuclides . Typically, such

Quadrupolar nuclides with half-integral spin quantum numbers . When

GENERAL TEXTS ON NMR

“Nuclear magnetic resonance: Concepts and methods ”, D. Canet, John

“Understanding NMR spectroscopy ”, J. Keeler, John Wiley & Sons Ltd.

“Magnetic resonance in chemistry and medicine ”, R. Freeman, Oxford

GENERAL TEXTS ON THE SOLID STATE

“Molecular crystals ”, 2nd edition, J.D. Wright, Cambridge University

“Polymorphism in the pharmaceutical industry ”, J. Bernstein, Oxford