3 The vector model
For most kinds of spectroscopy it is sufficient to think about energy levels
and selection rules; this is not true for NMR. For example, using this energy
level approach we cannot even describe how the most basic pulsed NMR experiment
works, let alone the large number of subtle two-dimensional experiments
which have been developed. To make any progress in understanding
NMR experiments we need some more tools, and the first of these we are
going to explore is the vector model.
This model has been around as long as NMR itself, and not surprisingly
the language and ideas which flow from the model have become the language
of NMR to a large extent. In fact, in the strictest sense, the vector model can
only be applied to a surprisingly small number of situations. However, the
ideas that flow from even this rather restricted area in which the model can
be applied are carried over into more sophisticated treatments. It is therefore
essential to have a good grasp of the vector model and how to apply it.
3.1 Bulk magnetization
magnetization
vector
magneticfield z
x y
Fig. 3.1 At equilibrium, a
sample has a net magnetization
along the magnetic field
direction (the z axis) which can
be represented by a
magnetization vector. The axis
set in this diagram is a
right-handed one, which is what
we will use throughout these
lectures.
We commented before that the nuclear spin has an interaction with an applied
magnetic field, and that it is this which gives rise the energy levels and ultimately
an NMR spectrum. In many ways, it is permissible to think of the
nucleus as behaving like a small bar magnet or, to be more precise, a magnetic
moment. We will not go into the details here, but note that the quantum
mechanics tells us that the magnetic moment can be aligned in any direction1.
In an NMR experiment, we do not observe just one nucleus but a very
large number of them (say 1020), so what we need to be concerned with is the
net effect of all these nuclei, as this is what we will observe.
If the magnetic moments were all to point in random directions, then the
small magnetic field that each generates will cancel one another out and there
will be no net effect. However, it turns out that at equilibrium the magnetic
moments are not aligned randomly but in such a way that when their contributions
are all added up there is a net magnetic field along the direction of the
applied field (B0). This is called the bulk magnetization of the sample.
The magnetization can be represented by a vector – called the magnetization
vector – pointing along the direction of the applied field (z), as shown in
Fig. 3.1. From now on we will only be concerned with what happens to this
vector.
This would be a good point to comment on the axis system we are going to
use – it is called a right-handed set, and such a set of axes is show in Fig. 3.1.
The name right-handed comes about from the fact that if you imagine grasping
1It is a common misconception to state that the magnetic moment must either be aligned
with or against the magnetic field. In fact, quantum mechanics says no such thing (see Levitt
Chapter 9 for a very lucid discussion of this point).
Chapter 3 “The vector model” c James Keeler, 2002
3–2 The vector model
the z axis with your right hand, your fingers curl from the x to the y axes.
3.2 Larmor precession
magneticfield
z
x y
Fig. 3.2 If the magnetization
vector is tilted away from the z
axis it executes a precessional
motion in which the vector
sweeps out a cone of constant
angle to the magnetic field
direction. The direction of
precession shown is for a
nucleus with a positive
gyromagnetic ratio and hence a
negative Larmor frequency.
Suppose that we have managed, some how, to tip the magnetization vector
away from the z axis, such that it makes an angle β to that axis. We will see
later on that such a tilt can be brought about by a radiofrequency pulse. Once
tilted away from the z axis we find is that the magnetization vector rotates
about the direction of the magnetic field sweeping out a cone with a constant
angle; see Fig. 3.2. The vector is said to precesses about the field and this
particular motion is called Larmor precession.
If the magnetic field strength is B0, then the frequency of the Larmor
precession is ω0 (in rad s−1)
ω0 = −γ B0
or if we want the frequency in Hz, it is given by
ν0 = −
1
2π
γ B0
where γ is the gyromagnetic ratio. These are of course exactly the same
frequencies that we encountered in section 2.3. In words, the frequency at
which the magnetization precesses around the B0 field is exactly the same as
the frequency of the line we see from the spectrum on one spin; this is no
accident.
As was discussed in section 2.3, the Larmor frequency is a signed quantity
and is negative for nuclei with a positive gyromagnetic ratio. This means that
for such spins the precession frequency is negative, which is precisely what is
shown in Fig. 3.2.
z
x y
Fig. 3.3 The precessing
magnetization will cut a coil
wound round the x axis, thereby
inducing a current in the coil.
This current can be amplified
and detected; it is this that forms
the free induction signal. For
clarity, the coil has only been
shown on one side of the x axis.
We can sort out positive and negative frequencies in the following way.
Imagine grasping the z axis with your right hand, with the thumb pointing
along the +z direction. The fingers then curl in the sense of a positive precession.
Inspection of Fig. 3.2 will show that the magnetization vector is rotating
in the opposite sense to your fingers, and this corresponds to a negative Larmor
frequency.
3.3 Detection
The precession of the magnetization vector is what we actually detect in an
NMR experiment. All we have to do is to mount a small coil of wire round
the sample, with the axis of the coil aligned in the xy-plane; this is illustrated
in Fig. 3.3. As the magnetization vector “cuts” the coil a current is induced
which we can amplify and then record – this is the so-called free induction
signal which is detected in a pulse NMR experiment. The whole process is
analogous to the way in which electric current can be generated by a magnet
rotating inside a coil.
3.4 Pulses 3–3
z
x
M0
sin β
β
Fig. 3.4 Tilting the
magnetization through an angle
θ gives an x-component of size
M0 sin β.
Essentially, the coil detects the x-component of the magnetization. We can
easily work out what this will be. Suppose that the equilibrium magnetization
vector is of size M0; if this has been tilted through an angle β towards the x
axis, the x-component is M0 sin β; Fig. 3.4 illustrates the geometry.
Although the magnetization vector precesses on a cone, we can visualize
what happens to the x- and y-components much more simply by just thinking
about the projection onto the xy-plane. This is shown in Fig. 3.5.
At time zero, we will assume that there is only an x-component. After a
time τ1 the vector has rotated through a certain angle, which we will call 1.
As the vector is rotating at ω0 radians per second, in time τ1 the vector has
moved through (ω0 × τ1) radians; so 1 = ω0τ1. At a later time, say τ2, the
vector has had longer to precess and the angle 2 will be (ω0τ2). In general,
we can see that after time t the angle is = ω0t.
τ1
ε1 ε2
ε
τ2
x
y
Fig. 3.5 Illustration of the precession of the magnetization vector in the xy-plane. The angle through which
the vector has precessed is given by ω0t. On the right-hand diagram we see the geometry for working out
the x and y components of the vector.
time
My
Mx
Fig. 3.6 Plots of the x- and
y-components of the
magnetization predicted using
the approach of Fig. 3.5.
Fourier transformation of these
signals will give rise to the usual
spectrum.
We can now easily work out the x- and y-components of the magnetization
using simple geometry; this is illustrated in Fig. 3.5. The x-component
is proportional to cos and the y-component is negative (along −y) and proportional
to sin . Recalling that the initial size of the vector is M0 sin β, we
can deduce that the x- and y-components, Mx and My respectively, are:
Mx = M0 sin β cos(ω0t)
My = −M0 sin β sin(ω0t).
Plots of these signals are shown in Fig. 3.6. We see that they are both
simple oscillations at the Larmor frequency. Fourier transformation of these
signals gives us the familiar spectrum – in this case a single line at ω0; the
details of how this works will be covered in a later chapter. We will also
see in a later section that in practice we can easily detect both the x- and
y-components of the magnetization.
3.4 Pulses
We now turn to the important question as to how we can rotate the magnetization
away from its equilibrium position along the z axis. Conceptually it is
3–4 The vector model
easy to see what we have to do. All that is required is to (suddenly) replace
the magnetic field along the z axis with one in the xy-plane (say along the
x axis). The magnetization would then precess about the new magnetic field
which would bring the vector down away from the z axis, as illustrated in
Fig. 3.7.
magnetic
field
z
x y
z
x y
Fig. 3.7 If the magnetic field
along the z axis is replaced
quickly by one along x, the
magnetization will then precess
about the x axis and so move
towards the transverse plane.
Unfortunately it is all but impossible to switch the magnetic field suddenly
in this way. Remember that the main magnetic field is supplied by a powerful
superconducting magnet, and there is no way that this can be switched off;
we will need to find another approach, and it turns out that the key is to use
the idea of resonance.
The idea is to apply a very small magnetic field along the x axis but one
which is oscillating at or near to the Larmor frequency – that is resonant with
the Larmor frequency. We will show that this small magnetic field is able to
rotate the magnetization away from the z axis, even in the presence of the very
strong applied field, B0.
Conveniently, we can use the same coil to generate this oscillating magnetic
field as the one we used to detect the magnetization (Fig. 3.3). All we
do is feed some radiofrequency (RF) power to the coil and the resulting oscillating
current creates an oscillating magnetic field along the x-direction. The
resulting field is called the radiofrequency or RF field. To understand how this
weak RF field can rotate the magnetization we need to introduce the idea of
the rotating frame.
Rotating frame
When RF power is applied to the coil wound along the x axis the result is a
magnetic field which oscillates along the x axis. The magnetic field moves
back and forth from +x to −x passing through zero along the way. We will
take the frequency of this oscillation to be ωRF (in rad s−1) and the size of
the magnetic field to be 2B1 (in T); the reason for the 2 will become apparent
later. This frequency is also called the transmitter frequency for the reason
that a radiofrequency transmitter is used to produce the power.
It turns out to be a lot easier to work out what is going on if we replace,
in our minds, this linearly oscillating field with two counter-rotating fields;
Fig. 3.8 illustrates the idea. The two counter rotating fields have the same
magnitude B1. One, denoted B+
1 , rotates in the positive sense (from x to y)
and the other, denoted B−
1 , rotates in the negative sense; both are rotating at
the transmitter frequency ωRF.
At time zero, they are both aligned along the x axis and so add up to give
a total field of 2B1 along the x axis. As time proceeds, the vectors move away
from x, in opposite directions. As the two vectors have the same magnitude
and are rotating at the same frequency the y-components always cancel one
another out. However, the x-components shrink towards zero as the angle
through which the vectors have rotated approaches 1
2 π radians or 90◦. As the
angle increases beyond this point the x-component grows once more, but this
time along the −x axis, reaching a maximum when the angle of rotation is π.
The fields continue to rotate, causing the x-component to drop back to zero
and rise again to a value 2B1 along the +x axis. Thus we see that the two
3.4 Pulses 3–5
x
x
y
y
2B1
B1
-
B1
+
time
fieldalongx
Fig. 3.8 Illustration of how two counter-rotating fields (shown in the upper part of the diagram and marked
B+
1 and B−
1 ) add together to give a field which is oscillating along the x axis (shown in the lower part). The
graph at the bottom shows how the field along x varies with time.
counter-rotating fields add up to the linearly oscillating one.
Suppose now that we think about a nucleus with a positive gyromagnetic
ratio; recall that this means the Larmor frequency is negative so that the sense
of precession is from x towards −y. This is the same direction as the rotation
of B−
1 . It turns out that the other field, which is rotating in the opposite sense
to the Larmor precession, has no significant interaction with the magnetization
and so from now on we will ignore it.
x
x
-ωRF
y
y
B1
-
-ωRF
Fig. 3.9 The top row shows a field rotating at −ωRF when viewed in a fixed axis system. The same field
viewed in a set of axes rotating at −ωRF appears to be static.
We now employ a mathematical trick which is to move to a co-ordinate
system which, rather than being static (called the laboratory frame) is rotating
about the z axis in the same direction and at the same rate as B−
1 (i.e. at
−ωRF). In this rotating set of axes, or rotating frame, B−
1 appears to be static
3–6 The vector model
and directed along the x axis of the rotating frame, as is shown in Fig. 3.9.
This is a very nice result as the time dependence has been removed from the
problem.
Larmor precession in the rotating frame
We need to consider what happens to the Larmor precession of the magnetization
when this is viewed in this rotating frame. In the fixed frame the
precession is at ω0, but suppose that we choose the rotating frame to be at the
same frequency. In the rotating frame the magnetization will appear not to
move i.e. the apparent Larmor frequency will be zero! It is clear that moving
to a rotating frame has an effect on the apparent Larmor frequency.
The general case is when the rotating frame is at frequency ωrot. fram.; in
such a frame the Larmor precession will appear to be at (ω0 −ωrot. fram.). This
difference frequency is called the offset and is given the symbol :
= ω0 − ωrot. fram.. (3.1)
We have used several times the relationship between the magnetic field
and the precession frequency:
ω = −γ B. (3.2)
From this it follows that if the apparent Larmor frequency in the rotating frame
is different from that in fixed frame it must also be the case that the apparent
magnetic field in the rotating frame must be different from the actual applied
magnetic field. We can use Eq. 3.2 to compute the apparent magnetic field,
given the symbol B, from the apparent Larmor frequency, :
= −γ B
hence B = −
γ
.
This apparent magnetic field in the rotating frame is usually called the reduced
field, B.
If we choose the rotating frame to be at the Larmor frequency, the offset
will be zero and so too will the reduced field. This is the key to how the
very weak RF field can affect the magnetization in the presence of the much
stronger B0 field. In the rotating frame this field along the z axis appears to
shrink, and under the right conditions can become small enough that the RF
field is dominant.
The effective fieldIn this discussion we will
assume that the gyromagnetic
ratio is positive so that the
Larmor frequency is negative.
From the discussion so far we can see that when an RF field is being applied
there are two magnetic fields in the rotating frame. First, there is the RF field
(or B1 field) of magnitude B1; we will make this field static by choosing the
rotating frame frequency to be equal to −ωRF. Second, there is the reduced
field, B, given by (− /γ ). Since = (ω0 − ωrot. fram.) and ωrot. fram. =
−ωRF it follows that the offset is
= ω0 − (−ωRF)
= ω0 + ωRF.
3.4 Pulses 3–7
This looks rather strange, but recall that ω0 is negative, so if the transmitter
frequency and the Larmor frequency are comparable the offset will be small.
z
x
∆B
θ
Beff
B1
Fig. 3.10 In the rotating frame
the effective field Beff is the
vector sum of the reduced field
B and the B1 field. The tilt
angle, θ, is defined as the angle
between B and Beff.
In the rotating frame, the reduced field (which is along z) and the RF or
B1 field (which is along x) add vectorially to give an effective field, Beff as
illustrated in Fig. 3.10. The size of this effective field is given by:
Beff = B2
1 + B2. (3.3)
The magnetization precesses about this effective field at frequency ωeff given
by
ωeff = γ Beff
just in the same way that the Larmor precession frequency is related to B0.
By making the offset small, or zero, the effective field lies close to the
xy-plane, and so the magnetization will be rotated from z down to the plane,
which is exactly what we want to achieve. The trick is that although B0 is
much larger than B1 we can affect the magnetization with B1 by making it
oscillate close to the Larmor frequency. This is the phenomena of resonance.
The angle between B and Beff is called the tilt angle and is usually given
the symbol θ. From Fig. 3.10 we can see that:
sin θ =
B1
Beff
cos θ =
B
Beff
tan θ =
B1
B
.
All three definitions are equivalent.
The effective field in frequency units
For practical purposes the thing that is important is the precession frequency
about the effective field, ωeff. It is therefore convenient to think about the
construction of the effective field not in terms of magnetic fields but in terms
of the precession frequencies that they cause.
z
x
ωeff
θ
Ω
ω1
Fig. 3.11 The effective field can
be thought of in terms of
frequencies instead of the fields
used in Fig 3.10.
For each field the precession frequency is proportional to the magnetic
field; the constant of proportion is γ , the gyromagnetic ratio. For example, we
have already seen that in the rotating frame the apparent Larmor precession
frequency, , depends on the reduced field:
= −γ B.
We define ω1 as the precession frequency about the B1 field (the positive sign
is intentional):
ω1 = γ B1
and we already have
ωeff = γ Beff.
Using these definitions in Eq. 3.3 ωeff can be written as
ωeff = ω2
1 + 2.
Figure 3.10 can be redrawn in terms of frequencies, as shown in Fig. 3.11.
Similarly, the tilt angle can be expressed in terms of these frequencies:
sin θ =
ω1
ωeff
cos θ =
ωeff
tan θ =
ω1
.
3–8 The vector model
On-resonance pulses
The simplest case to deal with is where the transmitter frequency is exactly
the same as the Larmor frequency – it is said that the pulse is exactly on
resonance. Under these circumstances the offset, , is zero and so the reduced
field, B, is also zero. Referring to Fig. 3.10 we see that the effective field is
therefore the same as the B1 field and lies along the x axis. For completeness
we also note that the tilt angle, θ, of the effective field is π/2 or 90◦.
x
-y
z
B1
x
-y
z
B1
(a) (b)
Fig. 3.12 A “grapefruit” diagram in which the thick line shows the motion of a magnetization vector during
an on-resonance pulse. The magnetization is assumed to start out along +z. In (a) the pulse flip angle
is 90◦. The effective field lies along the x axis and so the magnetization precesses in the yz-plane. The
rotation is in the positive sense about x so the magnetization moves toward the −y axis. In (b) the pulse
flip angle is 180◦ and so the magnetization ends up along −z.
In this situation the motion of the magnetization vector is very simple. Just
as in Fig. 3.7 the magnetization precesses about the field, thereby rotating in
the zy-plane. As we have seen above the precession frequency is ω1. If the RF
field is applied for a time tp, the angle, β, through which the magnetization
has been rotated will be given by
β = ω1tp.
β is called the flip angle of the pulse. By altering the time for which the pulse
has been applied we can alter than angle through which the magnetization is
rotated.
In many experiments the commonly used flip angles are π/2 (90◦) and π
(180◦). The motion of the magnetization vector during on-resonance 90◦ and
180◦ pulses are shown in Fig. 3.12. The 90◦ pulse rotates the magnetization
from the equilibrium position to the −y axis; this is because the rotation is
in the positive sense. Imagine grasping the axis about which the rotation is
taking place (the x axis) with your right hand; your fingers then curl in the
sense of a positive rotation.
If the pulse flip angle is set to 180◦ the magnetization is taken all the way
from +z to −z; this is called an inversion pulse. In general, for a flip angle β
3.5 Detection in the rotating frame 3–9
simple geometry tells us that the z- and y-components are
Mz = M0 cos β My = −M0 sin β;
this is illustrated in Fig. 3.13.
z
-y
M0
sin β
M0
cosβ
β
Fig. 3.13 If the pulse flip angle
is β we can use simple
geometry to work out the y- and
z-components.
Hard pulses
In practical NMR spectroscopy we usually have several resonances in the
spectrum, each of which has a different Larmor frequency; we cannot therefore
be on resonance with all of the lines in the spectrum. However, if we
make the RF field strong enough we can effectively achieve this condition.
By “hard” enough we mean that the B1 field has to be large enough that it
is much greater than the size of the reduced field, B. If this condition holds,
the effective field lies along B1 and so the situation is identical to the case
of an on-resonance pulse. Thinking in terms of frequencies this condition
translates to ω1 being greater in magnitude than the offset, .
transmitter
frequency
0510
ppm
Fig. 3.14 Illustration of the
range of offsets one might see
in a typical proton spectrum. If
the transmitter frequency is
placed as shown, the maximum
offset of a resonance will be 5
ppm.
It is often relatively easy to achieve this condition. For example, consider
a proton spectrum covering about 10 ppm; if we put the transmitter frequency
at about 5 ppm, the maximum offset is 5 ppm, either positive or negative;
this is illustrated in Fig. 3.14. If the spectrometer frequency is 500 MHz the
maximum offset is 5 × 500 = 2500 Hz. A typical spectrometer might have a
90◦ pulse lasting 12 µs. From this we can work out the value of ω1. We start
from
β = ω1tp hence ω1 =
β
tp
.
We know that for a 90◦ pulse β = π/2 and the duration, tp is 12 × 10−6 s;
therefore
ω1 =
π/2
12 × 10−6
= 1.3 × 105
rad s−1
.
The maximum offset is 2500 Hz, which is 2π × 2500 = 1.6 × 104 rad s−1.
We see that the RF field is about eight times the offset, and so the pulse can
be regarded as strong over the whole width of the spectrum.
3.5 Detection in the rotating frame
To work out what is happening during an RF pulse we need to work in the
rotating frame, and we have seen that to get this simplification the frequency
of the rotating frame must match the transmitter frequency, ωRF. Larmor
precession can be viewed just as easily in the laboratory and rotating frames;
in the rotating frame the precession is at the offset frequency, .
It turns out that because of the way the spectrometer works the signal that
we detect appears to be that in the rotating frame. So, rather than detecting an
oscillation at the Larmor frequency, we see an oscillation at the offset, .2
2Strictly this is only true if we set the receiver reference frequency to be equal to the
transmitter frequency; this is almost always the case. More details will be given in Chapter 5.
3–10 The vector model
It also turns out that we can detect both the x- and y-components of the
magnetization in the rotating frame. From now on we will work exclusively
in the rotating frame.
3.6 The basic pulse–acquire experiment
RF
acq
90˚
{
{
{
1 2 3
Fig. 3.15 Timing diagram or
pulse sequence for the simple
pulse–acquire experiment. The
line marked “RF” shows the
location of the pulses, and the
line marked “acq” shows when
the signal is recorded or
acquired.
At last we are in a position to describe how the simplest NMR experiment
works – the one we use every day to record spectra. The experiment comes in
three periods:
1. The sample is allowed to come to equilibrium.
2. RF power is switched on for long enough to rotate the magnetization
through 90◦ i.e. a 90◦ pulse is applied.
3. After the RF power is switched off we start to detect the signal which
arises from the magnetization as it rotates in the transverse plane.
The timing diagram – or pulse sequence as it is usually known – is shown in
Fig. 3.15.
x
-y
time
Mx
My
Fig. 3.16 Evolution during the acquisition time (period 3) of the pulse–acquire experiment. The magnetization
starts out along −y and evolves at the offset frequency, (here assumed to be positive). The
resulting x- and y-magnetizations are shown below.
During period 1 equilibrium magnetization builds up along the z axis. As
was described above, the 90◦ pulse rotates this magnetization onto the −y
axis; this takes us to the end of period 2. During period 3 the magnetization
precesses in the transverse plane at the offset ; this is illustrated in Fig. 3.16.
Some simple geometry, shown in Fig. 3.17, enables us to deduce how the
x- and y-magnetizations vary with time. The offset is so after time t the
vector has precessed through an angle ( × t). The y-component is therefore
proportional to cos t and the x-component to sin t. In full the signals are:
My = −M0 cos( t)
Mx = M0 sin( t).
As we commented on before, Fourier transformation of these signals will give
the usual spectrum, with a peak appearing at frequency .
3.7 Pulse calibration 3–11
x
-y
Ωt
Fig. 3.17 The magnetization
starts out along the −y axis and
rotates through an angle t
during time t.
Spectrum with several lines
If the spectrum has more than one line, then to a good approximation we can
associate a magnetization vector with each. Usually we wish to observe all
the lines at once, so we choose the B1 field to be strong enough that for the
range of offsets that these lines cover all the associated magnetization vectors
will be rotated onto the −y axis. During the acquisition time each precesses
at its own offset, so the detected signal will be:
My = −M0,1 cos 1t − M0,2 cos 2t − M0,3 cos 3t . . .
where M0,1 is the equilibrium magnetization of spin 1, 1 is its offset and so
on for the other spins. Fourier transformation of the free induction signal will
produce a spectrum with lines at 1, 2 etc.
3.7 Pulse calibration
It is crucial that the pulses we use in NMR experiments have the correct flip
angles. For example, to obtain the maximum intensity in the pulse–acquire
experiment we must use a 90◦ pulse, and if we wish to invert magnetization we
must use a 180◦ pulse. Pulse calibration is therefore an important preliminary
to any experiment.
flip angle
signalintensity
90º 180º 270º
π 3π/2π/2
Fig. 3.18 Illustration of how pulse calibration is achieved. The signal intensity varies as (sin β) as shown by
the curve at the bottom of the picture. Along the top are the spectra which would be expected for various
different flip angles (indicated by the dashed lines). The signal is a maximum for a flip angle of 90◦ and
goes through a null at 180◦; after that, the signal goes negative.
If we imagine an on-resonance or hard pulse we have already determined
from Fig. 3.13 that the y-component of magnetization after a pulse of flip angle
β is proportional to sin β. If we therefore do a pulse-acquire experiment
(section 3.6) and vary the flip angle of the pulse, we should see that the intensity
of the signal varies as sin β. A typical outcome of such an experiment is
shown in Fig. 3.18.
The normal practice is to increase the flip angle until a null is found; the
flip angle is then 180◦. The reason for doing this is that the null is sharper
than the maximum. Once the length of a 180◦ pulse is found, simply halving
the time gives a 90◦ pulse.
3–12 The vector model
Suppose that the 180◦ pulse was found to be of duration t180. Since the
flip angle is given by β = ω1tp we can see that for a 180◦ pulse in which the
flip angle is π
π = ω1t180
hence ω1 =
π
t180
.
In this way we can determine ω1, usually called the RF field strength or the
B1 field strength.
It is usual to quote the field strength not in rad s−1 but in Hz, in which
case we need to divide by 2π:
(ω1/2π) =
1
2t180
Hz.
For example, let us suppose that we found the null condition at 15.5 µs; thus
ω1 =
π
t180
=
π
15.5 × 10−6
= 2.03 × 105
rad s−1
.
In frequency units the calculation is
(ω1/2π) =
1
2t180
=
1
2 × 15.5 × 10−6
= 32.3 kHz.
In normal NMR parlance we would say “the B1 field is 32.3 kHz”. This is
mixing the units up rather strangely, but the meaning is clear once you know
what is going on!
3.8 The spin echo
x
-y
time
phase,φ
0
0
τ/2 τ τ
τ
3τ/2 2τ
2τ
180pulse
π
π/2
π−Ωτ
Ωτ Ωτ
Fig. 3.19 Vector diagrams showing how a spin echo refocuses the evolution of the offset; see text for
details. Also shown is a phase evolution diagram for two different offsets (the solid and the dashed line).
3.9 Pulses of different phases 3–13
RF
acq
90˚ 180˚
τ τ
2τ
Fig. 3.20 Pulse sequence for
the spin echo experiment. The
180◦ pulse (indicated by an
open rectangle as opposed to
the closed one for a 90◦ pulse)
is in the centre of a delay of
duration 2τ, thus separating the
sequence into two equal
periods, τ. The signal is
acquired after the second delay
τ, or put another way, when the
time from the beginning of the
sequence is 2τ. The durations
of the pulses are in practice very
much shorter than the delays τ
but for clarity the length of the
pulses has been exaggerated.
We are now able to analyse the most famous pulsed NMR experiment, the
spin echo, which is a component of a very large number of more complex experiments.
The pulse sequence is quite simple, and is shown in Fig. 3.20. The
special thing about the spin echo sequence is that at the end of the second τ
delay the magnetization ends up along the same axis, regardless of the values
of τ and the offset, .
We describe this outcome by saying that “the offset has been refocused”,
meaning that at the end of the sequence it is just as if the offset had been
zero and hence there had been no evolution of the magnetization. Figure 3.19
illustrates how the sequence works after the initial 90◦ pulse has placed the
magnetization along the −y axis.
During the first delay τ the vector precesses from −y towards the x axis.
The angle through which the vector rotates is simply ( t), which we can
describe as a phase, φ. The effect of the 180◦ pulse is to move the vector to a
mirror image position, with the mirror in question being in the xz-plane. So,
the vector is now at an angle ( τ) to the y axis rather than being at ( τ) to
the −y axis.
During the second delay τ the vector continues to evolve; during this time
it will rotate through a further angle of ( τ) and therefore at the end of the
second delay the vector will be aligned along the y axis. A few moments
thought will reveal that as the angle through which the vector rotates during
the first τ delay must be equal to that through which it rotates during
the second τ delay; the vector will therefore always end up along the y axis
regardless of the offset, .
The 180◦ pulse is called a refocusing pulse because of the property that
the evolution due to the offset during the first delay τ is refocused during the
second delay. It is interesting to note that the spin echo sequence gives exactly
the same result as the sequence 90◦ – 180◦ with the delays omitted.
Another way of thinking about the spin echo is to plot a phase evolution
diagram; this is done at the bottom of Fig. 3.19. Here we plot the phase, φ,
as a function of time. During the first τ delay the phase increases linearly
with time. The effect of the 180◦ pulse is to change the phase from ( τ) to
(π − τ); this is the jump on the diagram at time τ. Further evolution for
time τ causes the phase to increase by ( τ) leading to a final phase at the end
of the second τ delay of π. This conclusion is independent of the value of the
offset ; the diagram illustrates this by the dashed line which represents the
evolution of vector with a smaller offset.
As has already been mentioned, the effect of the 180◦ pulse is to reflect
the vectors in the xz-plane. The way this works is illustrated in Fig. 3.21.
The arc through which the vectors are moved is different for each, but all the
vectors end up in mirror image positions.
3.9 Pulses of different phases
So far we have assume that the B1 field is applied along the x axis; this does
not have to be so, and we can just as easily apply it along the y axis, for
example. A pulse about y is said to be “phase shifted by 90◦” (we take an
3–14 The vector model
-y
y
z
x
Fig. 3.21 Illustration of the effect of a 180◦ pulse on three vectors which start out at different angles from
the −y axis (coloured in black, grey and light grey). All three are rotated by 180◦ about the x axis on the
trajectories indicated by the thick lines which dip into the southern hemisphere. As a result, the vectors
end up in mirror image positions with respect to the xz-plane.
x
y y
z
B1
x
z
B1
(a) (b)
Fig. 3.22 Grapefruit plots showing the effect on equilibrium magnetization of (a) a 90◦ pulse about the y
axis and (b) a 90◦ pulse about the −x axis. Note the position of the B1 field in each case.
x-pulse to have a phase shift of zero); likewise a pulse about −x would be
said to be phase shifted by 180◦. On modern spectrometers it is possible to
produce pulses with arbitrary phase shifts.
If we apply a 90◦ pulse about the y axis to equilibrium magnetization we
find that the vector rotates in the yz-plane such that the magnetization ends
up along x; this is illustrated in Fig. 3.22. As before, we can determine the
effect of such a pulse by thinking of it as causing a positive rotation about the
y axis. A 90◦ pulse about −x causes the magnetization to appear along y, as
is also shown in Fig. 3.22.
We have seen that a 180◦ pulse about the x axis causes the vectors to
move to mirror image positions with respect to the xz-plane. In a similar way,
3.10 Relaxation 3–15
a 180◦ pulse about the y axis causes the vectors to be reflected in the yz-plane.
3.10 Relaxation
RF
acq
90º180º
τ
Fig. 3.23 The pulse sequence
for the inversion recovery
experiment used to measure
longitudinal relaxation.
We will have a lot more to say about relaxation later on, but at this point we
will just note that the magnetization has a tendency to return to its equilibrium
position (and size) – a process known as relaxation. Recall that the equilibrium
situation has magnetization of size M0 along z and no transverse (x or
y) magnetization.
So, if we have created some transverse magnetization (for example by applying
a 90◦ pulse) over time relaxation will cause this magnetization to decay
away to zero. The free induction signal, which results from the magnetization
precessing in the xy plane will therefore decay away in amplitude. This loss
of x- and y-magnetization is called transverse relaxation.
Once perturbed, the z-magnetization will try to return to its equilibrium
position, and this process is called longitudinal relaxation. We can measure
the rate of this process using the inversion recovery experiment whose pulse
sequence is shown in Fig. 3.23. The 180◦ pulse rotates the equilibrium magnetization
to the −z axis. Suppose that the delay τ is very short so that at the
end of this delay the magnetization has not changed. Now the 90◦ pulse will
rotate the magnetization onto the +y axis; note that this is in contrast to the
case where the magnetization starts out along +z and it is rotated onto −y.
If this gives a positive line in the spectrum, then having the magnetization
along +y will give a negative line. So, what we see for short values of τ is a
negative line.
increasing τ
Fig. 3.24 Visualization of the outcome of an inversion recovery experiment. The size and sign of the zmagnetization
is reflected in the spectra (shown underneath). By analysing the peak heights as a function
of the delay τ it is possible to find the rate of recovery of the z-magnetization.
As τ gets longer more relaxation takes place and the magnetization shrinks
towards zero; this result is a negative line in the spectrum, but one whose size
is decreasing. Eventually the magnetization goes through zero and then starts
to increase along +z – this gives a positive line in the spectrum. Thus, by
recording spectra with different values of the delay τ we can map out the
recovery of the z-magnetization from the intensity of the observed lines. The
3–16 The vector model
whole process is visualized in Fig. 3.24.
3.11 Off-resonance effects and soft pulses
-y
z
x
b
c
d
D
C
B
A
a
Fig. 3.25 Grapefruit diagram showing the path followed during a pulse for various different resonance
offsets. Path a is for the on-resonance case; the effective field lies along x and is indicated by the dashed
line A. Path b is for the case where the offset is half the RF field strength; the effective field is marked B.
Paths c and d are for offsets equal to and 1.5 times the RF field strength, respectively. The effective field
directions are labelled C and D.
So far we have only dealt with the case where the pulse is either on resonance
or where the RF field strength is large compared to the offset (a hard
pulse) which is in effect the same situation. We now turn to the case where
the offset is comparable to the RF field strength. The consequences of this are
sometimes a problem to us, but they can also be turned to our advantage for
selective excitation.
As the offset becomes comparable to the RF field, the effective field begins
to move up from the x axis towards the z axis. As a consequence, rather
than the magnetization moving in the yz plane from z to −y, the magnetization
starts to follow a more complex curved path; this is illustrated in
Fig. 3.25. The further off resonance we go, the further the vector ends up
from the xy plane. Also, there is a significant component of magnetization
generated along the x direction, something which does not occur in the onresonance
case.
We can see more clearly what is going on if we plot the x and y magnetization
as a function of the offset; these are shown in Fig. 3.26. In (a) we
see the y-magnetization and, as expected for a 90◦ pulse, on resonance the
equilibrium magnetization ends up entirely along −y. However, as the offset
increases the amount of y-magnetization generally decreases but imposed on
this overall decrease there is an oscillation; at some offsets the magnetization
is zero and at others it is positive. The plot of the x-magnetization, (b),
shows a similar story with the magnetization generally falling off as the offset
increases, but again with a strong oscillation.
Plot (c) is of the magnitude of the magnetization, which is given by
Mabs = M2
x + M2
y.
3.11 Off-resonance effects and soft pulses 3–17
(a) (b)
(c)
-20 -10
10 20
-1
-0.5
0.5
1
-20 -10 0 10 20
0.5
1
-20 -10 10 20
-1
-0.5
0.5
1
MxMy
Mabs
Ω / ω1
Ω / ω1
Ω / ω1
Fig. 3.26 Plots of the magnetization produced by a pulse as a function of the offset. The pulse length has
been adjusted so that on resonance the flip angle is 90◦. The horizontal axes of the plots is the offset
expressed as a ratio of the RF field strength, ω1; the equilibrium magnetization has been assumed to be
of size 1.
This gives the total transverse magnetization in any direction; it is, of course,
always positive. We see from this plot the characteristic nulls and subsidiary
maxima as the offset increases.
What plot (c) tells us is that although a pulse can excite magnetization
over a wide range of offsets, the region over which it does so efficiently is
really rather small. If we want at least 90% of the full intensity the offset
must be less than about 1.6 times the RF field strength.
Excitation of a range of shifts
There are some immediate practical consequences of this observation. Suppose
that we are trying to record the full range of carbon-13 shifts (200 ppm)
on an spectrometer whose magnetic field gives a proton Larmor frequency
of 800 MHz and hence a carbon-13 Larmor frequency of 200 MHz. If we
place the transmitter frequency at 100 ppm, the maximum offset that a peak
can have is 100 ppm which, at this Larmor frequency, translates to 20 kHz.
According to our criterion above, if we accept a reduction to 90% of the full
intensity at the edges of the spectrum we would need an RF field strength of
20/1.6 ≈ 12.5 kHz. This would correspond to a 90◦ pulse width of 20 µs. If
the manufacturer failed to provide sufficient power to produce this pulse width
we can see that the excitation of the spectrum will fall below our (arbitrary)
90% mark.
We will see in a later section that the presence of a mixture of x- and ymagnetization
leads to phase errors in the spectrum which can be difficult to
correct.
Selective excitation
Sometimes we want to excite just a portion of the spectrum, for example just
the lines of a single multiplet. We can achieve this by putting the transmitter
in the centre of the region we wish to excite and then reducing the RF field
3–18 The vector model
transmitter
offset
strong pulse
selective pulse
excitation
Fig. 3.27 Visualization of the use of selective excitation to excite just one line in the spectrum. At the top
is shown the spectrum that would be excited using a hard pulse. If the transmitter is placed on resonance
with one line and the strength of the RF field reduced then the pattern of excitation we expect is as shown
in the middle. As a result, the peaks at non-zero offsets are attenuated and the spectrum which is excited
will be as shown at the bottom.
strength until the degree of excitation of the rest of the spectrum is essentially
negligible. Of course, reducing the RF field strength lengthens the duration
of a 90◦ pulse. The whole process is visualized in Fig. 3.27.
Such pulses which are designed to affect only part of the spectrum are
called selective pulses or soft pulses (as opposed to non-selective or hard
pulses). The value to which we need to reduce the RF field depends on the
separation of the peak we want to excite from those we do not want to excite.
The closer in the unwanted peaks are the weaker the RF field must become
and hence the longer the 90◦ pulse. In the end a balance has to be made between
making the pulse too long (and hence losing signal due to relaxation)
and allowing a small amount of excitation of the unwanted signals.
Figure 3.27 does not portray one problem with this approach, which is
that for peaks away from the transmitter a mixture of x- and y-magnetization
is generated (as shown in Fig. 3.26). This is described as a phase error, more
of which in a later section. The second problem that the figure does show is
that the excitation only falls off rather slowly and “bounces” through a series
of maximum and nulls; these are sometimes called “wiggles”. We might
be lucky and have an unwanted peak fall on a null, or unlucky and have an
unwanted peak fall on a maximum.
Much effort has been put into getting round both of these problems. The
key feature of all of the approaches is to “shape” the envelope of the RF
pulses i.e. not just switch it on and off abruptly, but with a smooth variation.
Such pulses are called shaped pulses. The simplest of these are basically bellshaped
(like a gaussian function, for example). These suppress the “wiggles”
at large offsets and give just a smooth decay; they do not, however, improve
the phase properties. To attack this part of the problem requires an altogether
more sophisticated approach.
3.11 Off-resonance effects and soft pulses 3–19
Selective inversion
Sometimes we want to invert the magnetization associated with just one resonance
while leaving all the others in the spectrum unaffected; such a pulse
would be called a selective inversion pulse. Just as for selective excitation,
all we need to do is to place the transmitter on the line we wish to invert
and reduce the RF field until the other resonances in the spectrum are not affected
significantly. Of course we need to adjust the pulse duration so that the
on-resonance flip angle is 180◦.
Mz Mz
Ω / ω1
Ω / ω1
-20 -10 10 -5 520
0.5 0.5
-0.5 -0.5
-1 -1
1 1
(a) (b)
Fig. 3.28 Plots of the z-magnetization produced by a pulse as a function of the offset; the flip angle
on-resonance has been set to 180◦. Plot (b) covers a narrower range of offsets than plot (a). These plots
should be compared with those in Fig. 3.26.
Figure 3.28 shows the z-magnetization generated as a function of offset
for a pulse. We see that the range over which there is significant inversion
is rather small, and that the oscillations are smaller in amplitude than for the
excitation pulse.
This observation has two consequences: one “good” and one “bad”. The
good consequence is that a selective 180◦ pulse is, for a given field strength,
more selective than a corresponding 90◦ pulse. In particular, the weaker
“bouncing” sidelobes are a useful feature. Do not forget, though, that the
180◦ pulse is longer, so some of the improvement may be illusory!
The bad consequence is that when it comes to hard pulses the range of
offsets over which there is anything like complete inversion is much more
limited than the range of offsets over which there is significant excitation.
This can be seen clearly by comparing Fig. 3.28 with Figure 3.26. Thus, 180◦
pulses are often the source of problems in spectra with large offset ranges.
3–20 The vector model
3.12 Exercises
E 3–1
A spectrometer operates with a Larmor frequency of 600 MHz for protons.
For a particular set up the RF field strength, ω1/(2π) has been determined to
be 25 kHz. Suppose that the transmitter is placed at 5 ppm; what is the offset
(in Hz) of a peak at 10 ppm? Compute the tilt angle, θ, of a spin with this
offset.
For the normal range of proton shifts (0 – 10 ppm), is this 25 kHz field
strong enough to give what could be classed as hard pulses?
E 3–2
In an experiment to determine the pulse length an operator observed a positive
signal for pulse widths of 5 and 10 µs; as the pulse was lengthened further the
intensity decreased going through a null at 20.5 µs and then turning negative.
Explain what is happening in this experiment and use the data to determine
the RF field strength in Hz and the length of a 90◦ pulse.
A further null in the signal was seen at 41.0 µs; to what do you attribute
this?
E 3–3
Use vector diagrams to describe what happens during a spin echo sequence in
which the 180◦ pulse is applied about the y axis. Also, draw a phase evolution
diagram appropriate for this pulse sequence.
In what way is the outcome different from the case where the refocusing
pulse is applied about the x axis?
What would the effect of applying the refocusing pulse about the −x axis
be?
E 3–4
The gyromagnetic ratio of phosphorus-31 is 1.08 × 108 rad s−1 T−1. This
nucleus shows a wide range of shifts, covering some 700 ppm.
Estimate the minimum 90◦ pulse width you would need to excite peaks
in this complete range to within 90% of the their theoretical maximum for a
spectrometer with a B0 field strength of 9.4 T.
E 3–5
A spectrometer operates at a Larmor frequency of 400 MHz for protons and
hence 100 MHz for carbon-13. Suppose that a 90◦ pulse of length 10 µs is
applied to the protons. Does this have a significant effect of the carbon-13
nuclei? Explain your answer carefully.
E 3–6
Referring to the plots of Fig. 3.26 we see that there are some offsets at which
the transverse magnetization goes to zero. Recall that the magnetization is rotating
about the effective field, ωeff; it follows that these nulls in the excitation
3.12 Exercises 3–21
come about when the magnetization executes complete 360◦ rotations about
the effective field. In such a rotation the magnetization is returned to the z
axis. Make a sketch of a “grapefruit” showing this.
The effective field is given by
ωeff = ω2
1 + 2.
Suppose that we express the offset as a multiple κ of the RF field strength:
= κω1.
Show that with this values of the effective field is given by:
ωeff = ω1 1 + κ2.
(The reason for doing this is to reduce the number of variables.)
Let us assume that on-resonance the pulse flip angle is π/2, so the duration
of the pulse, τp, is give from
ω1τp = π/2 thus τp =
π
2ω1
.
The angle of rotation about the effective field for a pulse of duration τp is
(ωeffτp). Show that for the effective field given above this angle, βeff is given
by
βeff =
π
2
1 + κ2.
The null in the excitation will occur when βeff is 2π i.e. a complete rotation.
Show that this occurs when κ =
√
15 i.e. when ( /ω1) =
√
15. Does
this agree with Fig. 3.26?
Predict other values of κ at which there will be nulls in the excitation.
E 3–7
When calibrating a pulse by looking for the null produced by a 180◦ rotation,
why is it important to choose a line which is close to the transmitter frequency
(i.e. one with a small offset)?
E 3–8
Use vector diagrams to predict the outcome of the sequence:
90◦
− delay τ − 90◦
applied to equilibrium magnetization; both pulses are about the x axis. In
your answer, explain how the x, y and z magnetizations depend on the delay
τ and the offset .
E 3–9
Consider the spin echo sequence to which a 90◦ pulse has been added at the
end:
90◦
(x) − delay τ − 180◦
(x) − delay τ − 90◦
(φ).
3–22 The vector model
The axis about which the pulse is applied is given in brackets after the flip
angle. Explain in what way the outcome is different depending on whether
the phase φ of the pulse is chosen to be x, y, −x or −y.
E 3–10
The so-called “1–1” sequence is:
90◦
(x) − delay τ − 90◦
(−x)
For a peak which is on resonance the sequence does not excite any observable
magnetization. However, for a peak with an offset such that τ = π/2 the
sequence results in all of the equilibrium magnetization appearing along the
x axis. Further, if the delay is such that τ = π no transverse magnetization
is excited.
Explain these observations and make a sketch graph of the amount of
transverse magnetization generated as a function of the offset for a fixed delay
τ.
The sequence has been used for suppressing strong solvent signals which
might otherwise overwhelm the spectrum. The solvent is placed on resonance,
and so is not excited; τ is chosen so that the peaks of interest are excited. How
does one go about choosing the value for τ?
E 3–11
The so-called “1–1” sequence is:
90◦
(x) − delay τ − 90◦
(y).
Describe the excitation that this sequence produces as a function of offset.
How it could be used for observing spectra in the presence of strong solvent
signals?