Nuclear Magnetic Resonance

1 Background

What we call “nuclear magnetic resonance” (NMR) was developed simultaneously but indepen-
dently by Edward Purcell and Felix Bloch in 1946. The experimental method and theoretical
interpretation they developed is now called “continuous-wave NMR” (CWNMR). A different ex-
perimental technique, called “pulsed NMR” (PNMR), was introduced in 1950 by Erwin Hahn.
Pulsed NMR is used in magnetic resonance imaging (MRI). Purcell and Bloch won the Nobel Prize
in Physics in 1952 for NMR; more recently NMR was the subject of Nobel Prizes in Chemistry in
1991 and 2002. We have both NMR setups in the advanced labs: one is a variation of the CWNMR
method, and the other is a pulsed NMR system.
The physics underlying NMR is the same for both the continuous and pulsed methods, but the
information obtained may be different. Certainly, the words used to describe what is being done in
the experiments are different. In continuous-wave NMR one tunes a weak radio-frequency oscillator
to match the natural frequency of nuclear magnetic moments precessing in a static magnetic field.
When the frequencies match, the moments “resonate” and energy transfer between the RF field and
the sample is optimized; this resonance creates an electrical signal in a pickup coil. By sweeping
the magnetic field, one can leave the RF frequency fixed but tune the resonance itself to excite the
sample and then see how quickly the the magnetization dies away by watching a “beat pattern.”
In pulsed NMR, one applies a sequence of strong RF bursts or “pulses” to the sample that are very
short in duration. The resulting signal, called the “free induction” is then optimized to tune the RF
oscillator. The pulses are strong enough and short enough to reorient all of the spins in a coherent
way. One can flip the magnetization upside down with a “π-pulse” (180 degrees) or sideways
with a “π/2-pulse” (90 degrees). With the a sequence of pi-pulses, one can create the fascinating
phenomenon of “spin echoes.” More on the two methods is given later in this introduction.
Below, we give a very simplified introduction based on classical ideas to the physics of NMR.
More thorough discussions, focusing on the CWNMR technique, may be found in in the books by
Preston and Dietz [1] and Melissinos [2] (see references). In particular, the chapter in Preston and
Dietz gives a nice description of the connection between quantum-mechanical and semi-classical
approaches to NMR physics. For a comprehensive treatment of NMR, see the book by Slichter [3].

1.1 Semiclassical ideas

To observe NMR, one needs nuclei with a non-zero angular momentum I~ and magnetic moment
~ . The relationship between these two quantities is
µ

~ = γ I~ ,
µ (1)

where γ is called the gyromagnetic ratio (a scalar that may be positive or negative).
A simple classical calculation would give γ = q/2M , where q is the charge of the nucleus and M
is its mass. But quantum mechanics requires this dimensionally correct result to be modified. In
practice, we specify γ in units of the nuclear magneton µn ≡ eh̄/2mp times a dimensionless factor
g called the “spectroscopic splitting factor” (often just “g factor”):
gµn ge
γ= = , (2)
h̄ 2mp

1
where e is the elementary charge and mp is the mass of one proton. The g factor is on the order of
unity. It is positive for some nuclei, and negative for others. For the proton, g = +5.586; a table
of various g factors may be found in Preston and Dietz.
When a nucleus of moment µ ~ 0 it will experience a torque causing
~ is placed in a magnetic field B
a change in the angular momentum following Newton’s second law:
~
τ =µ
~ ~ 0 = dI = 1 d~
~ ×B
µ
. (3)
dt γ dt
Figure 1 shows some of the vectors defined with B ~ 0 lying along the z axis (which is horizontal in
most lab setups, including ours). The rate of change of µ ~0
~ is, by Eq. (3), both perpendicular to B

Figure 1: (a) Vectors defined in semiclassical picture of NMR. (b) Tip of angular momentum vector
~ viewed opposite to B
I, ~ 0.

and µ
~ itself; hence µ ~ 0 . From Fig. 1 it is easy to derive from
~ precesses about the direction of B
geometry that the vector magnitudes obey
µ ¶
µ dφ
µB0 sin θ = sin θ ; (4)
γ dt
thus
dφ
ω0 ≡ = γB0 , (5)
dt
which is the Larmor angular frequency.
When we deal with nuclei of spin I = 1/2 (e.g., protons), quantum mechanics tells us that in a
~ 0 the ground state splits into two sublevels, as shown in Fig. 2. The energies U of
magnetic field B

Figure 2: Splitting of nuclear energy levels due to applied magnetic field.

the two sublevels are given by

U = −~ ~ 0 = −µz B0 = −γh̄m B0 ,
µ·B (6)
I

2
 ~ 0 , m = ±1/2.
where mI is the angular momentum projection of I along the z direction defined by B I
The difference in energy between the two sublevels is thus

∆U = U− − U+ = γh̄B0 , (7)

where U+/− is the energy of the state corresponding to mI = ±1/2. Transitions between the two
energy sublevels may be induced by photons carrying one unit of angular momentum (h̄) and energy
h̄ω0 such that
h̄ω0 = γh̄B0 , (8)
or ω0 = γB0 , which is the same as the Larmor frequency, Eq. (5).
For free protons, the gyromagnetic ratio is (according to NIST, https://physics.nist.gov/cuu)

γ = 2.6752 × 108 radians/(seconds-Tesla) ,

so for fields in the 0.1–1 T range the frequencies will be in the megahertz range. These are
radio frequencies (RF), and both of our setups—pulsed and continuous—work in this MHz range.
(Historical note: There was an advantage to having NMR work in the RF range, as existing radio
equipment and techniques were used for detection and modulation of NMR signals.) Given that
magnetic fields in these experiments are frequently measured in kilogauss, a useful relationship to
remember for protons is,
f0 (MHz) = 4.2577 × B0 (kilogauss) , (9)
where f0 = ω0 /2π.
In order to excite the magnetic resonance, the applied RF must be circularly polarized, optimally
~ 1 (t) lies
with the rotation axis along the z axis. This is shown in Fig. 3, where an applied field B
in the x-y plane and rotates about the z axis at the Larmor frequency and in the same direction as
the Larmor precession. Now imagine observing the field from the reference frame that rotates along

~ 1 field rotates with the precessing µ

Figure 3: The RF B ~ in the semiclassical picture.

with the Larmor precession; the field B ~ 1 would be seen as a steady field relative to the moment µ ~,
~ ~
and also perpendicular to B 0 . The effect of B 1 (t) would be to exert a torque in that rotating frame.
However, if the RF were applied at a frequency different than the Larmor frequency, the torque on
~ in the rotating frame would vary in direction and strength over time, so the net work done by
µ
the field would be less, tending to zero as the RF angular frequency moves away from ω0 . A little
mathematics shows that this situation follows the standard picture of resonance phenomena—a
resonance which involves the interaction between a nucleus with a magnetic field, hence the name
nuclear magnetic resonance.
In the quantum-mechanical picture for a single nucleus we can say that the probability of a “spin-
flip” between the mI = +1/2 and mI = −1/2 state is maximized when the RF photon has energy

3
equal to h̄ω0 . One may well wonder whether a complete quantum mechanical approach would
give different answers than the classical picture we have presented. It is somewhat surprising
to discover that Bloch showed that the classical equations governing NMR can be derived from
quantum mechanics. See the text by Slichter for a readable and thorough presentation of this
result [3].

1.2 Net magnetization and relaxation

Clearly, any physical sample will consist of many atoms and the NMR signal measured by an
experiment will be due to the combined effect of the magnetic field on them all. As an example,
consider an assembly of N protons, say, in water, glycerin, mineral oil, or animal tissue. The net
magnetization M ~ of the sample is the vector sum of all of the individual moments µ ~ . For the
example of spin- 2 protons, the net magnetization along the z-axis, Mz would be given by
1

X 1
Mz = γh̄mI i = γh̄ (N+ − N− ) , (10)
2
i

where mI i denotes the state of the ith proton, and N+ , N− are the numbers of protons in the + 21
and − 12 state, respectively. When an RF field B ~ 1 (t) is applied, some protons have their spin flip
to align against B ~ 0 ; these absorb energy from the RF field. Other protons experience a spin-flip to
~
align with B 0 ; these give energy back to the RF field. Thus, if the populations of the two sublevels
were equal, Mz = 0 and no net energy would be absorbed from or given up to the field.
At room temperature with no RF field but with the steady B ~ 0 field there will be a small difference
between the number of protons with spins aligned in the direction of B ~ 0 and the number aligned
opposite to it; the magnetic sublevel with mI = + 2 will have a slightly larger population N+ than
1

the population N− in sublevel mI = − 21 . The ratio of the populations is proportional to the ratio
of the associated Boltzmann factors, according to the principles of statistical mechanics:
N+ exp(−U+ /kT ) exp (+ 12 γh̄B0 /kT )
= = = eγh̄B0 /kT . (11)
N− exp(−U− /kT ) exp (− 12 γh̄B0 /kT )

~ 1 (t) to this equilibrium population, energy goes into the system, since there
If one applies the RF B
is an imbalance in the number of up (+ 12 ) spins versus down (− 21 ) spins. In a fairly short time,
however, this imbalance will vanish, because the torque experienced by each moment will cause the
populations N+ and N− to become equal. At this point, the net energy transferred between the
sample and the field drops to zero as Mz = 0, and the sample is said to be saturated.
There is still a tendency for the system to recover its equilibrium configuration, even while it is
subject to the oscillating field. The collection of protons could continue to absorb energy from an
oscillating magnetic field B~ 1 (t) if the rate of equilibrium population recovery is greater than the
rate of energy transfer from the field. The rate of Mz recovering its equilibrium (nonzero) value is
due to energy exchanges between the proton magnetic moments and their local environments. The
rate of energy transfer from the RF field depends on the amplitude of the RF field and how close
its frequency matches the resonance frequency.
Let us consider the magnetization a bit more generally. The net magnetization M ~ is a vector,
and thus can be broken into components. The component along the direction of the static field
Mz is called the longitudinal magnetization, and the components along two orthogonal directions
perpendicular to B ~ 0 are called the transverse magnetizations Mx and My . Clearly, each of these

4
components is the sum of the individual component moments µz , µx and µy , respectively, e.g.,
~ 0 . It
P
Mx = i µxi . Now look again at Fig. 3 which depicts a single moment precessing about B
should be obvious that for a collection of µ
~ s at a particular time, each precessing independently
about B~ 0 , one could have a nonzero Mz but a (possibly) zero Mx or My .

The equilibrium nonzero Mz exists because the moments aligned parallel to B ~ 0 are at a lower
energy than those aligned antiparallel. In order to return to equilibrium following absorption from
~ 1 (t) field, the moments would need to give energy to their surroundings, conventionally called
the B
the “lattice” (even if we are dealing with liquid or gaseous materials). The relaxation of Mz to its
equilibrium value is called the “longitudinal” or “spin-lattice” relaxation, and is associated with
a characteristic time called T1 . In liquids, T1 is typically short: a few milliseconds to as long as
a second or two in water or water-based ionic solutions. But in solids, T1 may be very long, e.g.,
T1 ≈ 10, 000 s in ice. This is one reason why we use liquid samples in the continuous NMR setup,
as the short T1 allows for a continuous input of energy from the oscillator coil.
A nonzero Mx or My is harder to obtain, as it requires a majority of the moments to be precessing
in lock-step with each other. Even so, Mx and My cannot be constant in time, but at best oscillate
between positive and negative values at the Larmor frequency. If one can contrive to force the
moments to have a net projection onto, say, the x-axis (and indeed one can—we’ll see how), then
the resulting oscillating Mx would decay because variation in the local magnetic field, some of which
is due to the field of neighboring moments, would cause different moments to precess at different
rates. This is called “spin-spin dephasing”, and the characteristic time for this process is called T2 ,
the “spin-spin” or “transverse” relaxation time. In most cases, T2 is shorter than T1 ; in solids it is
much shorter, but in low viscosity liquids (like water or light oils) T2 approaches T1 .
With these two relaxation processes in mind, consider how one would detect the signal of these
precessing and relaxing moments. Since they are magnetic, an obvious choice would be to use
a pickup coil. The coil also needs to be oriented transverse to the static field in order to sense
~ 0 , the coil would only be sensitive to a slight
the precession, since in an orientation parallel to B
change in flux from a non-oscillatory Mz ; whereas the transverse component Mx or My will oscillate
between positive and negative values at the Larmor frequency ω0 .
This arrangement is depicted schematically in Fig. 4, which is taken from the article by Hahn [4].
The top of the figure shows that at t = 0, all of the moments are aligned along the +x direction and
~ 0 points out of the page toward the
the pickup coil has its axis along the y-axis. The static field B
viewer, thus the moments will precess in the clockwise direction (following the sense of M ~ ×B ~ 0 ).
The right side of the figure depicts the voltage measured across the pickup coil. Faraday’s law
requires that this voltage be proportional to the rate of change of the flux through the coil, and it
is a simple matter to prove that the maximum signal from the coil will occur at t = 0 in this setup.
~ which is rotating about the z-axis at ω0 , and calculate the
(Try it: let the field in the coil be µ0 M
rate of change of the flux, dΦ/dt from this.)
As the moments precess, the induced voltage oscillates at the Larmor frequency (νLARMOR in the
figure), and also decays. The decay is due to the spin-spin dephasing, shown as a spreading of the
moments away from the principal rotating x-axis—the T2 process.
While the overall decay of the signal from the pickup coil comes from spin-spin relaxation, exper-
imentally a third effect is often more predominate: spin-spin dephasing due to inhomogeneities in
~ 0 over the volume of the sample. Taken together, the effects lead to a combined
the external field B

5
Figure 4: Induction signal measured by a pickup coil in the transverse plane following orientation
of moments along the x-axis. Taken from the article by Hahn [4].

relaxation time T2∗ defined as

1 1
= + γ∆B0 , (12)
T2∗ T2
where ∆B0 is the variation in B0 . In effect, γ∆B0 is equal to a spread in Larmor frequencies ∆ω0
which, all else being equal, would cause spin-spin dephasing in a characteristic time of 1/∆ω0 .
Similarly, one can assign an effective resonance width ∆ωeff ∼ 1/T2∗ .
In summary, two processes are involved in the relaxation of the net magnetization: 1) spin-lattice
relaxation, which involves energy exchange with the environment, and is characterized by the time
constant T1 ; 2) spin-spin relaxation, characterized by a time constant T2 , which is caused by loss of
the phase relationships among the various moments and is due to the fact that each spin experiences
a slightly different local magnetic field. In a practical setup, the measured signal comes from the
oscillating transverse magnetization, characterized by a time constant T2∗ that is related to T2 , but
is typically dominated by inhomogeneities in the static field B ~ 0 . It is important to note that the
recovery to equilibrium following saturation follows T1 , but the decay of the detected signal follows
T2∗ .
The two relaxation times T1 and T2 are related to each other. For non-viscous liquids, in which
the molecules undergo very rapid motion on a small scale, theory predicts that T1 ≈ T2 . When
the material becomes more viscous (like heavy oil or grease), there is a divergence between T1 and
T2 , with T2 becoming shorter than T1 [3, pp. 210-214]. As required by Eq. (12), T2∗ < T2 , and in
practice T2∗ ≪ T1 .

6
1.3 The two experimental methods

In the continuous NMR method, the RF excitation producing the rotating B ~ 1 field is applied all
~
of the time. The resonance is created by having the B 0 field swept slowly through the value which
satisfies Eq. (5). A pickup coil surrounding the sample detects the resonance, and this signal is
mixed with the fixed RF signal to create “beats” (the term for the modulation of two signals which
have nearly the same frequency). The effect of the beats can be seen on an oscilloscope. Once the
resonant signal is found, the relationship between the magnetic field and the resonant frequency is
fixed, and depending on the givens, the information can be used to extract B0 itself, γ, or related
quantities. One can also crudely measure T2∗ by looking at the decay of the beat signal.
In the pulsed NMR method, the RF excitation is applied to the sample in a series of short bursts,
or pulses. The application of the RF field for a short time (the “pulse width”) allows the applied
torque to rotate the net magnetization M ~ by a specific amount. For example, one can apply a pulse
of RF field to rotate all moments by 90 . If this pulse is applied to a sample initially at equilibrium
◦

with a net magnetization M ~ = Mz k̂, then M ~ will become Mz ı̂ (i.e., the same magnitude of
magnetization now pointing in an orthogonal direction), which will then precess about the z-axis.
The pickup coil will see a signal like that shown in Fig. 4 as the individual magnetic moments µ ~i
get out of phase with each other and as the magnetization decays back to its equilibrium state. A
pulse which accomplishes this trick is called a “π/2 pulse”, and the signal seen as a consequence
is called the “free induction decay”. A pulse of a longer duration can flip the net magnetization
completely: Mz k̂ → −Mz k̂; this type of pulse is called a “π pulse”. Interestingly, the free induction
decay signal immediately following a π pulse is zero since there is no net transverse component of
the magnetization available to induce such a signal.
The real utility of the pulse method comes from using a sequence of pulses. By such sequences,
one can measure accurately and independently T1 and T2 , and also compensate for the effects of
field inhomogeneity. The discussion of pulse sequences and their effects are found in the theoretical
introduction to the pulsed NMR experiment write-up.

References

[1] Preston, D. W., and E. R. Dietz, The Art of Experimental Physics, John Wiley & Sons, New
York, 1991, pages 264–285.

[2] Melissinos, A. C., Experiments in Modern Physics, Academic Press, San Diego, 1966, pages
340–374.

[3] Slichter, C. P., Principles of Magnetic Resonance, 3rd Ed., Springer-Verlag, Berlin, 1990.

[4] Hahn, E. L., “Free nuclear induction”, Physics Today, November 1953, pp. 4-9.

7
2 The continuous-wave NMR experiment

In this experiment you will be looking for the magnetic resonance condition for protons (also known
as 1 H nuclei) and for fluorine (19 F) nuclei. As discussed previously, our apparatus brings the sample
into resonance by adding a small sweep magnetic field to the steady state field, B0 . A diagram of
the apparatus is shown in Fig. 5. The sweep field has a triangular waveform (frequency≈ 80 Hz) as

Figure 5: The continuous-wave NMR (CWNMR) apparatus.

shown in Fig. 6, is symmetric about zero, and has a magnitude of several gauss peak to peak (which
is very small compared to the steady state field generated by the electromagnet). Figure 6 also
shows the resonance signal as the sample passes through the resonance condition on both sides of
the triangular waveform. A more detailed picture of the resonance signal is shown in Fig. 7, which
was acquired using the averaging feature of the digital oscilloscope. It is a signal similar to this
that you will be looking for, first with the X-Y analog scope, and later with the digital scope. A
gaussmeter is used to provide an independent measure of magnetic field strength. The gaussmeter
probe is attached to the shaft holding the NMR head, and the sensing region of the probe is right
at the tip, which has been positioned as close to the sample as possible. The gaussmeter is capable

8
Figure 6: Resonance signal compared to field sweep drive.

Figure 7: Closeup of resonance signal.

9
of measuring either AC or DC fields, and you will find specifications for this device available in the
lab or on the class web site. This gaussmeter uses the Hall effect to measure the magnetic field,
but almost all Hall effect based gaussmeters are factory calibrated against NMR standards.

2.1 Procedure

The sample of choice for seeing your first resonance is H2 O + CuSO4 as it has a strong, easily
observed signal. The main effect of CuSO4 in solution is to shorten the relaxation time, as discussed
below. The CuSO4 ions do not appreciably shift the resonance conditions (frequency or magnetic
field) for the protons in the H2 O molecules.

1. As a precaution, remove watches, credit cards, computer disks, or any other magnetically
sensitive material from your clothes and body, and set them aside on a table. Don’t forget to
pick them up when you are done!

2. Turn on the cooling water for the magnet and plug in the circulating pump (or turn on the
power strip that the pump may already be plugged into). The TA or Professor will assist you
with this.

3. Place the H2 O + CuSO4 sample in the “NMR head” and position the NMR oscillator/detector
box so the head is approximately centered in all directions between the electromagnet pole
pieces.

4. Check that the magnet current is set to 0 (current adjust control on front of magnet power
supply should be set to 0), and turn on the supply by pushing the black button. You should
hear a click and see the power light come on. (Note: the only controls you will use on the
magnet current supply are the on/off buttons and the current adjust knob.)

5. Turn on the rest of the electronics: both oscilloscopes, the gaussmeter, the HP frequency
counter, the EG&G 5113 pre-amplifier, the Exact model 7050 function generator and the RF
oscillator/detector box that has the NMR head on it.

6. After the scopes have warmed up, look at the trace produced by the output of the Exact
function generator on the digital scope. It should be set to make a triangle wave of about 80
Hz with an amplitude of about 8 volts peak-to-peak, similar to triangle wave seen in Fig. 6.

7. On the analog oscilloscope in the instrument rack, you should see a horizontal trace centered
on the screen. With both inputs set to 1V/div, you should see a line that is about 8 divisions
wide. Adjust the trace position controls to center this line.

8. Set the EG&G preamplifier so that it has a gain of 10 and a filter bandwidth of 3 Hz to
300 kHz. You may want to change these settings as you start to make measurements, but
these are a good place to start.

9. Adjust the frequency of the RF oscillator to about 9.4 MHz which is near the “50” mark on
the large dial. If you do not see noise wiggles on the scope screen and/or if the HP frequency
counter is not responding, adjust the Oscillator Current Control (small knob below and left
of the frequency control) in order to bring the circuit into oscillation. Please ask for help if
you cannot get it to work (the batteries may be dead).

10
 10. Once you see evidence that the oscillator is working, steadily increase the magnet current.
When you get close to 35 amps slow down your adjustment: you are looking for the the
resonance signal to pass by on the analog scope “X-Y” display. After finding the resonance
(at about 36 amps or 2.19 kG), adjust the magnet current so that the two resonance bumps
are symmetrically located on the X-Y display.

11. Carefully move the NMR head around in the gap between the magnet pole pieces to find the
position where the resonance signal reaches a maximum. You are looking for the strongest
signal with the most “wiggles”. It may be necessary to shim the oscillator box on the bottom
with some foam strips to keep it in the optimum tilt position.

12. Adjust the Oscillator Current Control to optimize the combination of signal amplitude and
background noise. Note that turning this knob clockwise increases the signal (and noise!)
amplitude. If this knob is turned too far in either direction, the oscillator circuit may shut
down, as noted by a flat line on the scope screen, or it may latch-up into spontaneous oscilla-
tion, which will look like an excessively large signal. Ideally, you want the oscillator current
drive to be just a little above the lowest value needed to see the noise plus NMR resonance
signal.

Observe how the position of the resonance signals on the X-Y display is shifted by changing either
the oscillator frequency or the magnetic field. For this part of the experiment, the gaussmeter
is set up to measure only the static field B0 . To associate a resonant frequency with a particular
gaussmeter reading, the resonance must occur at the zero crossing of the sweep field. This condition
is met when the two resonance signals are symmetrically located on the X-Y display.
NOTE: It is common for the signal to drift, especially when the equipment is just turned on, until
the temperature of both the oscillator circuit and magnet coil stabilizes.
The signal “wiggles” or “ringing” that one sees as the sample goes through resonance are the beats
between signals at the (fixed) frequency of the oscillator, fosc , and the signal at the frequency of
precession, fprec , of the net magnetic moment of the protons in the total magnetic field, which is
the static field plus the sweep field. If resonance occurs as the sweep field crosses zero, then right
at resonance we have the condition

fprec = fosc = γB0 . (13)

As the sweep field changes from zero, fprec changes as

fprec = γ[B0 + Bsweep (t)] . (14)

The oscillator has a detection system which outputs the difference or “beat” frequency between
the fixed oscillator frequency and the changing precession frequency. The “beat” signal observed
on the scope is in the audio frequency range, and you can see the beat frequency increase as the
sweep field moves further away from zero.
Note: the EG&G PAR 5113 preamplifier has been added to the electronics in to filter out some
of the noise on the NMR oscillator output. To give a cleaner signal you can raise the low pass
frequency up to 1 kHz, and drop the high pass frequency down to 10 kHz if you need to. In order
to avoid overloading the preamp input, it is necessary to attenuate the oscillator signal by sending
it through the 1:4 attenuator box before connecting it to the preamp. Once you have found the
resonance, you may want to adjust the preamp settings to give a clean signal on the scopes.

11
Exercise I: Measurement of γ for protons in water.

The gyromagnetic ratio γ is defined as the proportionality constant between the magnetic moment
~ and the angular momentum I~ of the nucleus, µ
µ ~ As shown in section 1.1, γ can also be
~ = γ I.
expressed as
2πfresonance
γ= ,
B0
where B0 is the value of the steady state field at resonance and fresonance is the resonant frequency.
Place the H2 O + CuSO4 sample in the NMR head and center the head between the pole pieces
so as to obtain the strongest resonance signal. For a series of B0 -values between approximately
1.9 and 2.6 kilogauss, measure the frequency and magnetic field at which resonance occurs. It is
important that the resonance be symmetrically centered on the horizontal trace, and this is most
easily accomplished by first centering the trace on the scope screen, making sure you can see both
ends of the trace, and then centering the “bat-wings” of the resonance on the screen itself.
As there are typically fluctuations due to noise in both the frequency and magnetic field readings,
take at least 5 independent readings of each and compute the average. A good strategy is to hold
the frequency steady and recenter the resonance with the current control for each reading. Plot the
averages of resonant frequency vs. magnetic field, and from the slope of this graph, derive a value
for γproton . Compare your value to the accepted value given Eq. 9. Also, calculate the g factor for
the proton according to Eq. (2).
You should be able to measure over the whole range of the RF oscillator: from about 6.4 MHz up
to a little over 12 MHz. The relationship is linear, so 10 to 15 data points (with their uncertainty)
is sufficient. You will need to adjust the RF Oscillator Current Control as you go to optimize the
signal. Notice how the control also affects the oscillator frequency.
When you are above 10 MHz the magnet current will be rather large (> 40 amps). Do
NOT leave the current at such high values for an extended period of time, and when
you are not making measurements turn the current down to zero. If the current supply
gets too hot it will shut itself off.

Exercise II: Precise measurement of magnetic fields.

In terms of a bona fide measurement, exercise I is something of a ruse: the γ for protons is extremely
well known. In fact, gaussmeters are typically calibrated against NMR resonances, since frequency
is easy to measure to a high accuracy, all protons are exactly alike, and the response of Hall probes
varies with temperature and construction which makes them unsuitable as a primary standard. In
effect, exercise I should be better thought of as a measurement of the accuracy of the gaussmeter.
In this exercise we’ll use the sensitivity of the NMR resonance to study the magnetic field and its
variation between the pole pieces of the magnet.
Adjust the magnet current to a value of 36–38 amps, and allow it to stabilize for a minute or so.
Using the H2 O + CuSO4 sample, find the frequency at which resonance occurs in different regions
between the pole pieces. The field values may be calculated from
fresonance (MHz)
magnetic field = .
4.2577 MHz/kilogauss

Another effect should be obvious as you move the sample around—the shortened relaxation time and
diminished magnitude of the resonance signal. These effects are due to the increasing inhomogeneity

12
of the magnetic field as you move the sample away from the center of the gap between the pole
pieces. When B0 is not uniform over the sample volume, different parts of the sample go through the
resonance condition for different values of the sweep field. The signals from these various resonances
tend to cancel each other out, thus shortening the observed relaxation time and diminishing the
signal magnitude.
Estimate the variation in B0 from the center of the gap between the pole pieces to positions on
either side of the center, both along the axis of the magnet (shift the head from side to side)
and perpendicular to the axis (shift the head forward and back) by noting the change in resonant
frequency at these non-optimum positions. You will see two effects: (1) the resonance position will
shift slightly on the scope screen and (2) when the head is near the edges of the pole pieces the
resonance will vanish suddenly as the field becomes highly nonuniform.
Also keep an eye on the gaussmeter as you move the head around. Does the gaussmeter follow
what you measure from the resonance shift? Try to estimate how much B0 changes over the extent
of the sample itself.
Then, measure the magnitude of the sweep field produced by the Exact function generator and
the sweep coils mounted on the NMR head by noting how much frequency shift is needed to move
the resonance from one side of the trace to the other side of the trace. (It takes a steady hand
to control the resonance position with the RF frequency control knob!). In your report, estimate
these variations in both relative (percent) and absolute (e.g., kilogauss or tesla) units.
Don’t sweat too much over these measurements. You only need to give a rough calculation for each
part, not a full uncertainty analysis.

19
Exercise III: Measurement of γ for F nuclei

In order to measure γ for fluorine nucleus 19 F, repeat Exercise I with a twist: instead of comparing
the resonance frequency to the magnetic field measured by the gaussmeter, compare the frequency
of the fluorine resonance to the the frequency of the proton resonance at the same magnet current.
Then use the ratio of these frequencies, and the known value of γ for the proton to obtain γ for
19 F.

Specifically, make the following measurements. For a series of different magnet currents (i.e.,
different field strengths), tune the oscillator to center the resonance on the X-Y display for the
H2 O+CuSO4 sample. Then, without changing the magnet current, replace the sample with the
perfluoropolyether oil sample (a high-quality lubricant containing fluorine) and re-tune the oscillator
to center its resonance on the X-Y display. Be careful when changing the sample in the oscillator
probe while the magnet is on. Remove watches and jewelry from the hand you use to reach into the
magnet gap.
Hint: at a given field strength, the fluorine resonance frequency is about 94% of the proton resonance
frequency.
Plot the resonance frequency of the proton versus the resonance frequency of the fluorine, fit the
result to a line, and use the fit to calculate γ and g of the fluorine in terms of the known values
for the proton. The accepted value of g for 19 F is +5.2952. However, fluorine is known to be
susceptible to relatively large “chemical shifts:” the actual resonant frequency depends on the
particular electron environment that the fluorine atom resides in. Based on the accepted value of
g and your measurement, how much (if any) chemical shift do you see?

13
Finally, see if you can find the resonance of the Teflon sample. Teflon (polytetrafluoroethylene) is
a polymer containing long chains of carbon atoms with two fluorine atoms bonded to each carbon
atom. The signal from the Teflon sample is considerably smaller than that from the proton sample,
so you will have to look for it very carefully. You may also be able to see a chemical shift between
Teflon and fluorinated oil.

Exercise IV : Effect of paramagnetic ion concentration on T∗2 of protons in water

Relaxation, in general, means the return of a system to an equilibrium condition after an external
stimulus has been applied to the system. In this experiment, relaxation means the re-establishment
of the net magnetization of the sample parallel to B ~ 0 (equilibrium condition) after the sample has
gone through resonance (external stimulus). If the sample goes through the resonance condition
again before it has had sufficient time to relax, the resonance signal will be reduced in amplitude.
Why this is so is more readily understood after doing the pulsed NMR experiment, where the
relationship between observed signal strength and relaxation time can be investigated in detail.
In this exercise you will qualitatively investigate the relationship between the relaxation time of a
sample of water with varying concentrations of copper sulfate in solution. CuSO4 ions have large
magnetic moments due the unpaired electron spins in the copper atoms. These magnetic moments
perturb the local field seen by individual protons with the result that the equilibrium configuration
is recovered more rapidly (T1 is shorter), and the resonance signal decays more rapidly (T2 is also
shorter) than in the absence of the CuSO4 ions. Thus T2∗ is shorter and we see a stronger resonance
signal since the oscillator can continuously put energy into the sample.
Distilled water, saturated CuSO4 solution and vials are available in the lab for making up samples.
Start by putting some distilled water into a vial and sealing the top with a piece of masking tape.
Do not overfill the vial—notice how the pre-made samples have about 1 cm depth of liquid. Note
how weak the resonance signal is: in fact you may not see it at all. This is because protons in pure
water have a T1 relaxation time on the order of a second or more, while we are sweeping through
the resonance condition at approximately 160 Hz (i.e., 2 × fsweep ). Now make up a sample with half
pure water and half saturated CuSO4 solution, again to a depth of about one centimeter, sealing
the top with masking tape. Note the dramatically shorter T1 relaxation time as evidenced by the
strong signal every time the sample goes through resonance.
To make a quantitative estimate of T2∗ , it is convenient to capture the resonance signal on the
digital scope, similar to that shown in Fig. 7. Ask for help if you have trouble configuring the
scope to capture this signal. Once you have the resonance signal displayed on the scope, you
can use the averaging feature to eliminate some of the noise on the signal. After capturing a
satisfactory waveform, print it out by pressing the HARDCOPY button on the scope, and from
this copy, estimate the relaxation time T2∗ of the sample, where the envelope of the signal decays
∗
approximately as e−t/T2 .
Repeat the scope grabs for several samples of varying concentrations of copper sulfate in solution,
starting with the saturated solution, and then diluting it by a factor of 2 until your signal is
indistinguishable from that of pure water. The easiest way to do this is to use two eyedroppers,
one with saturated CuSO4 and one with pure water. For each sample you want to have the same
overall volume because the signal strength depends on the size of your sample. About 4 drops in
a vial gives a depth of about 1 cm. You can mix the pure water and the saturated solution in a
small plastic cup and then transfer a few drops into a vial.

14
Your goal in these measurements is to see the number of wiggles (as the trace decays) increase
as the concentration of CuSO4 decreases. Unfortunately, because of the inhomogeneity of the
magnetic field, the effect is fairly weak: the observed relaxation time is shorter than that due to
just the interactions between the spins themselves and any other magnetic moments in the sample
environment.
The limitations of this apparatus do not allow one to separate out the contribution to the relaxation
time due to inhomogeneities in the external field, but you will see how this can readily be done
using the more sophisticated pulsed NMR technique.

2.2 Shut Down

After you have made your measurements on the various solutions, please empty and rinse out the
vials for use by the next group of students.
Then in this order:

• Turn off the electronics,

• Turn off the NMR oscillator box, as the batteries will run down overnight.

• Turn the magnet current to zero before turning off the magnet power supply.

• Close the Supply and Return valves for the cooling water.

• Unplug the circulating pump.

Prepared by J. Stoltenberg, D. Pengra, R. Van Dyck and O. Vilches

cwnmr_2018.tex -- Updated 25 June 2018

15