Is Quantum Mechanics

Necessary for
Understanding Magnetic
Resonance?
LARS G. HANSON
Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark

ABSTRACT: Educational material introducing magnetic resonance (MR) typically con-

tains sections on the underlying principles. Unfortunately the explanations given are
often unnecessarily complicated or even wrong. MR is often presented as a phenomenon
that necessitates a quantum mechanical explanation whereas it really is a classical effect,
i.e. a consequence of the common sense expressed in classical mechanics. This insight is
not new, but there have been few attempts to challenge common misleading explana-
tions, so authors and educators are inadvertently keeping myths alive. As a result, new
students’ first encounters with MR are often obscured by explanations that make the sub-
ject difficult to understand. Typical problems are addressed and alternative intuitive
explanations are provided. Ó 2008 Wiley Periodicals, Inc. Concepts Magn Reson Part A 32A:
329–340, 2008.

KEY WORDS: magnetic resonance imaging; education; quantum mechanics; classical

mechanics; tutorial; spin; myths

INTRODUCTION interactions between electrons and nuclei, atoms

would not exist as they would collapse in fractions of
Since the beginning of the twentieth century it has a second because orbiting electrons radiate energy
been known that classical physics as expressed in and hence loose speed according to classical mechan-
Newton’s and Maxwell’s equations do not form a ics. The phenomena not explicable by classical
complete description of known physical phenomena. mechanics inspired the formulation of the fundamen-
If, for example, classical mechanics described the tal laws of quantum mechanics (QM). They have
been tested very extensively for almost a century and
no contradictions between experiments and the pre-
Received 5 April 2008; revised 20 July 2008; dictions of QM are known.
accepted 21 July 2008 The QM theory is probabilistic in nature, i.e., it
Correspondence to: Lars G. Hanson. E-mail: [email protected] only provides the probabilities for specific observa-
Concepts in Magnetic Resonance Part A, Vol. 32A(5) 329–340 (2008) tions to be made. This is not a surprising aspect of a
Published online in Wiley InterScience (www.interscience.wiley. physical law as a system cannot generally be pre-
com). DOI 10.1002/cmr.a.20123 pared in a state precisely enough to ensure a specific
Ó 2008 Wiley Periodicals, Inc. future outcome (the uncertainty of the initial condi-
329
330 HANSON

tions must generally be reflected in uncertainty of the It was argued earlier that QM has bizarre aspects
future). What is bizarre and nonintuitive, however, is that must be acknowledged to apply and interpret the
that QM is not generally reducible to a nonprobabil- theory. It is important to note, however, that most
istic theory, even when initial conditions can be con- aspects of QM are not surprising. In particular, the
trolled perfectly, unless other equally bizarre addi- so-called correspondence principle must hold true. In
tions to the theory are made (1). Hence, according to the macroscopic limit QM typically reduces to classi-
QM, measurements are associated with some intrin- cal mechanics, i.e., give similar predictions to those
sic uncertainty, even when the state of the system is of Newton’s and Maxwell’s equations (macroscopic
not. This indeterminism of nature has been tested quantum phenomena exist, but they are few or non-
extensively and experimentally verified. obvious). Luckily the consequence in the context of
That a complete description of the world has MR is that a classical description is adequate, and
aspects that are considered bizarre by humans is not overwhelmingly so in tutorials for nonphysicists.
surprising, as phenomena encountered during species Typically neither students nor teachers of MR have
evolution all fall within a very narrow range. Until the background for meaningful discussions of QM. It
recently, no creature made detailed observations of is fortunate that they can refrain from engaging in
phenomena on other length and time scales than their such, because quantum phenomena are difficult to
own macroscopic scale, humans being the first observe with MR hardware, and since QM play no
known exception. The laws of classical mechanics role for the vast majority of MR measurements. In
that are based on macroscopic observations describe addition to challenging myths, it is therefore a pur-
most phenomena on this scale well, but typically fail pose of this article to suggest alternative, yet correct
when applied to atomic and cosmological length and explanations and graphs based on classical mechan-
time scales. Hence, it is not surprising that QM ics only.
occurs as a rather difficult theory to learn and under- QM is here used to show that classical mechanics
stand. In fact, even physicists perfectly capable of is fully adequate for almost all purposes related to
applying the laws of QM to make right predictions MR. Using one formalism to demonstrate that the
about results of experiments may make misleading same formalism should be avoided in favor of some-
interpretations of the same experiments. QM is, in thing simpler, may seem counter-intuitive. QM is the
other words, easier to apply than to understand and more complete theory, however, and only by demon-
explain, probably because little emphasis is put on strating that QM reduces to classical mechanics in
interpretation in most contexts including education the relevant situations, can the case be made rigor-
that is typically rooted in pragmatism. ously. Consequently, this text contains outlines of
This problem is unfortunately evident in the field calculations that require QM knowledge to be under-
of magnetic resonance (MR), and it is amplified by stood, even though the target audience is people in
the diversity of people who teach and write books need of explaining or understanding MR of which
about this subject. Physicists, medical doctors, many are nonphysicists. The rigor is needed, espe-
radiographers, electrical engineers, and chemists are cially because the subject covered is potentially con-
among the most common authors of books that troversial, considering the large number of authors
include sections on the basic physics of MR. Many and educators that may feel targeted. The aim of this
of these people are not trained in QM. Hence, even article is not to warn against typical presentations,
excellent books and lectures on MR may contain however. The tutorials referenced for problematic
statements that are misleading, overly complicated, propositions are, for example, all excellent in other
or downright wrong. Examples can be found in early respects. Rather it is the aim to avoid the continuous
MR literature, and some are repeated so often that repetition of misleading arguments in MR literature
alternative formulations are not given sufficient con- and to avoid the confusion it causes among students
sideration. Precise formulations of MR basics exist, that are already sufficiently challenged without such.
e.g. as presented by Levitt (2) or advocated on the An article on this matter is consequently considered
ReviseMRI web site (3), but they are unfortunately a long due. The references were chosen among many
minority. Many texts aimed at physicists and other similar to exemplify that the problem is neither new
people trained in QM do not make the mistakes nor seems to be diminishing.
pointed out here, but they often fail to mention that The theory section provides examples of common
most aspects of MR are perfectly understandable misconceptions, prove them wrong or misleading,
from a classical perspective. It is a purpose of this ar- and gives alternative explanations. The possible
ticle to challenge some of the myths and misleading origins and consequences of the myths are also dis-
explanations appearing in MR tutorials. cussed. Sections requiring a detailed knowledge of

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
 UNDERSTANDING MAGNETIC RESONANCE 331

QM are relegated to appendices that may be skipped

by readers who accept the given arguments without
reading the proofs.
The discussed phenomena are common to most MR
effects, e.g., electron spin resonance. All examples
will be drawn from nuclear MR, however, as this phe-
nomenon is commonly explained for nonphysicists in
introductions to magnetic resonance imaging (MRI).

THEORY
Figure 1 These figures illustrating the same situation are
Myth 1: According to QM, Protons Align
frequently seen in MR-tutorials but they do not contribute
Either Parallel or Antiparallel to the much but confusion. They illustrate the spin eigenstates
Magnetic Field which are of little relevance to MR, as the state reduction
This myth reflects a misinterpretation of QM and it is induced by measurement is only partial and does not
found in numerous texts on MR, e.g. (4–9). The bring single nuclei into eigenstates. [Color figure can be
viewed in the online issue, which is available at www.
problem is realized by making a classical analogy. If
interscience.wiley.com.]
a collection of noninteracting compasses were sub-
ject to the earth magnetic field, and they behaved as
described, we would be surprised: Some would swing (the Greek letter c pronounced psi is typically used
to the north as expected and some would swing in this context):
south, which is not seen experimentally. QM is not
v ¼ vx x^ þ vy y^ [1]
classical mechanics, and as argued in the introduc-
tion, we do expect surprises, but this is not one of jci ¼ c" j"i þ c# j#i [2]
them, neither for compasses nor for nuclei.
The possible states are weighted sums of the eigen-
From a technical point of view, it is easy to track
the origin of the misconception. According to QM, a states which indicate that there are many more states
proton in a magnetic field has only two spin-states available to the protons than spin-up and spin-down.
with a well-defined energy (energy eigenstates). The weights c: and c; are complex numbers that
These are typically called spin-up and spin-down express the direction of spins as precisely as nature
where up and down refers to parallel and antiparallel allows in accordance with QM. Considering the prop-
to the magnetic field. These eigenstates are written as erties of the weights, it can be shown that there are two
j"i and j#i by physicists. Despite their name, these degrees of freedom for the spin of a proton (azimuthal
states have elements of magnetization perpendicular and polar angle) just as there are in classical mechan-
to the magnetic field in addition to longitudinal com- ics, i.e. a spin can point in any direction in three-dimen-
ponents. Hence the spin-up and spin-down states are sional space, although, as described in the ‘‘Intro-
often illustrated by two cones as shown in Fig. 1, e.g. duction,’’ the directions are associated with some
in references (5–7, 10). intrinsic uncertainty. When the magnetizations of iso-
The energy eigenstates form a so-called basis for chromats (groups of protons experiencing the same
all possible states. Spin orthogonal to the field can, magnetic field) are considered rather than of individual
for example, be written as a weighted sum of spin-up nuclei, the relative uncertainty vanishes. This is the
and spin-down. To explain the concept of a basis, a case for samples with more than a few atoms.
highschool-level example will be given: consider a It is important for the understanding of QM that
particle moving in the two-dimensional xy-plane. addition of states differs from addition of spin vec-
The orthogonal unit vectors x^ and y^ form a basis for tors. Adding equal amounts of spin-up and spin-down
the two-dimensional vector space, so any velocity in the sense expressed by Eq. [2], for example, does
vector v, for example, can be decomposed into veloc- not lead to cancellation, but amounts to a state of
ity along the x-direction and velocity along the y- transversal magnetization.
direction. Just as any two-dimensional vector in the Probably any physicist would agree to the above
xy-plane (for example velocity) can be written as a so this does not explain how myth 1 occurred. The
weighted sum of x^ and y^, any spin-state jci can be origin is the following: if the spin of an individual
written as a weighted sum of spin-up and spin-down proton is measured along the direction of the mag-

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
332 HANSON

Figure 2 In the absence of magnetic field, the spins are pointing randomly hence giving a
spherical distribution of spin orientations. This is illustrated to the right by a large number of
example spins in an implicit magnetization coordinate space similar to that of Fig. 1. [Color fig-
ure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

netic field, it will be found to be either in the spin-up made: when subject to a magnetic field, an oblong pi-
or spin-down state, no matter which mixed state jci ece of magnetizable material have a strong tendency
it was in before. Furthermore, it will stay in that new to align itself in one of two opposite directions paral-
state until the proton is subject to more interactions lel to the field (in contrast to permanently magnetized
with environment (e.g. another measurement). This material that orient itself in one direction only). De-
so-called collapse into an eigenstate is a consequence spite a superficial resemblance, this well-known phe-
of QM. It apparently implies that a measurement of nomenon has nothing to do with the effect expressed
the net magnetization (e.g. by MRI), will force each in myth 1. Rather it is a consequence of reorientation
proton into either the spin-up or the spin-down state of magnetic constituents inside the metal. This gives
in agreement with myth 1. This is wrong, however. rise to the existence of two low-energy states for the
The emphasized word individual above is important orientation of the metallic piece, parallel and antipar-
in the present context, as we can only infer from QM allel to the field. The magnetic constituents are in ei-
that the protons are forced into single-spin eigen- ther case parallel to the field, because they have only
states, if we measure their magnetization one-by-one one low-energy state. Similarly, the proton spin has
as can be done with a Stern-Gerlach apparatus, for only one low-energy state. Nothing but MR-irrele-
example (11). In contrast, that is never done in MR vant single-proton measurements give spins a tend-
spectrometers or scanners: to get a measurable MR- ency to align antiparallel to the field.
signal the total magnetization of many nuclei is Consequently, spins can point in any direction and
always measured, and myth 1 does not follow. It the energy eigenstates are not more relevant to MR
could be true nevertheless, but in fact it is not, which than any other state (the eigenstates form a conven-
is shown in appendix (proposition 1) by employing ient basis for computations, but they are irrelevant
the QM formalism: A measurement of the net mag- for the understanding). Hence Fig. 1 that illustrates
netization causes a perturbation of the system that is the nature of spin eigenstates, do not contribute much
insufficient to affect the individual protons signifi- but confusion in an MR context. QM is later shown
cantly. In particular, they are not brought into their to imply that the spin-evolution of individual protons
eigenstates by the measurement process. happens as expected classically unless perturbed,
It is worth noting that even though the arguments e.g., by a single-spin measurement.
mentioned above may occur complicated for the non- Finally, replacements for Fig. 1 are discussed.
technical reader, they are what many students of MR According to both classical and QM, spins are
more or less implicitly lay ears to, and for no good expected to point in all directions in the absence of
reason, as QM is not needed for understanding basic field as shown in Fig. 2. Except for precession, the
MR. Moreover, the students often hear the wrong situation does not change much when the polarizing
version of the argument. B0-field used for MR is applied as shown in Fig. 3.
The lifetime of myth 1 may have been prolonged The energies associated with the orientation of the
by an observation that many working with MR have individual spins are much smaller than the thermal

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
 UNDERSTANDING MAGNETIC RESONANCE 333

It is important to understand that precession of the

individual nuclei starts as soon as the sample is
placed in the field (not only after excitation by RF
fields, as frequently stated). The nuclei therefore emit
radio waves at the Larmor frequency as soon as they
are placed in the field. Similarly, however, they
absorb radio waves emitted by their neighbors and
surroundings. Because the distribution of spin direc-
tions is even in the transversal plane, the net trans-
versal magnetization is zero, and there is no net
exchange of energy between the sample and its sur-
roundings. The exchange of radio waves within the
sample is nothing but magnetic interactions between
neighboring nuclei. These are responsible for
relaxation.

Myth 2: MR Is a Quantum Effect

A quantum effect is a phenomenon that cannot be
Figure 3 Better alternative to Fig. 1 showing the spin adequately described by classical mechanics, i.e., one
distribution in a magnetic field. The spins will precess as where only QM give predictions in accordance with
indicated by the circular arrow and a longitudinal equilib- observations. In the introduction, it was stated that
rium magnetization (large vertical arrow) will gradually atom formation is a quantum effect because atoms
form as the distribution is skewed slightly toward mag-
are not expected to be stable according to classical
netic north by T1-relaxation (uneven density of arrows).
mechanics, whereas experiments have proven that
The equilibrium magnetization is stationary, so even
though the individual spins are precessing, there is no net they are. This does not imply that all phenomena
emission of radio waves in equilibrium. [Color figure can involving atoms are defined as quantum effects, since
be viewed in the online issue, which is available at such a broad definition would be quite useless.
www.interscience.wiley.com.] Instead, phenomena are hierarchically divided into
classical and quantum phenomena, so a classical phe-
energies so the spins only have a slight tendency to nomenon can involve atoms that are themselves inex-
point along the direction of the field (and no plicable by classical mechanics. Similarly, proton
increased tendency to point opposite the field, neither spin is a quantum effect but MR is not since the latter
classically, nor quantum mechanically). The situation is accurately described by classical mechanics. This
can be compared to the one described earlier involv- is the subject of the present section.
ing a hypothetical collection of compasses placed in It is often and correctly said that spin is a quantum
the earth magnetic field. All compasses will swing to effect. From a classical perspective it cannot be
the north, if they are noninteracting and not dis- explained why protons apparently all rotate with the
turbed. The situation changes if the compasses are same constant angular frequency, which result in an
placed in a running tumble-dryer or similar device observable angular momentum (spin) and associated
increasing the collisional energies above those asso- magnetic moment. Despite the fact that this is really
ciated with changing the direction of the compass mind-boggling, it is usually not perceived so by
needles. The bouncing and interacting compasses students of MR. Just as atoms are taken for granted,
will no longer all point toward north but there will it is typically accepted without argument that protons
still be a slight tendency for them to do so. If the net appear to be rotating and that they as a result behave
magnetization is measured, it will point toward north. as small magnets with a north and a south pole, i.e.
The situation is like that of the moving protons in a they have angular and magnetic moments. Most
liquid sample where the magnetic interactions books state this correctly and there is no apparent
between neighboring nuclei cause reorientation of reason to elaborate, as a deeper understanding of spin
the magnetic moments (relaxation). In the absence of is typically of little use in the context of MR.
a magnetic field, the angular distribution of spins is Even though spin is a quantum effect, MR is not,
spherical in either case. When a field is added, the according to the definition given earlier, as it does
distribution is skewed slightly toward the field direc- not necessitate a quantum explanation. Classically, a
tion by relaxation. magnetic dipole M with an associated angular mo-

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
334 HANSON

mentum M/g (spin) will precess in a magnetic field

B0 at the Larmor frequency f ¼ gB0. The gyro-mag-
netic ratio g is specific for the type of nucleus. If sub-
ject to an additional, orthogonal, magnetic field rotat-
ing at the Larmor frequency, the magnetization will
also precess around the rotating field vector B1(t).
This is most easily appreciated in a rotating frame of
reference (12), where B1 appears stationary and the
effect of B0 is not apparent except for its influence
on relaxation. This is classical MR, as visualized for
Figure 4 This figure sometimes seen in MR literature is
example in published animations (9, 13, 14).
misleading. It shows how spins in the eigenstates (left)
It was shown by the famous physicist and Nobel can be reoriented as to form a transversal magnetization
laureate Richard Feynman and coworkers (15) that (right). However, a homogeneous RF field never changes
the class of phenomena called two-level quantum the relative orientations of spins which contradicts the
dynamics can be understood in the light of classical validity of the figure. [Color figure can be viewed in the
MR. Specifically, the paper showed that these phe- online issue, which is available at www.interscience.
nomena are described by the same math that applies wiley.com.]
to classical MR, and that an abstract vector quantity
(the Bloch vector) descriptive of the quantum state tions between spin-up and spin-down states (16) in
moves like a magnetic dipole in a magnetic field. contrast to the flips or jumps that are often high-
For the special case of the spin-up/spin-down two- lighted in MR-tutorials (4–6) but are not occurring
level system, the Bloch vector is indeed proportional because the protons are not forced into eigenstates by
to the expected magnetization, which was shown to MR measurements.
move as predicted by classical mechanics. Hence
Feynman and Vernon pointed out that the dynamics
Myth 3: RF Pulses Brings the Precessing
of a proton in a time-varying magnetic field is easy
Spins into Phase
to understand because it behaves classically.
In that light, it is difficult to understand the ration- It is sometimes said that the effect of a resonant RF
ale of many introductory MR books that explain MR field is to bring the precessing spins into phase (4, 6,
to students with nontechnical backgrounds by the use 17) as indicated in Fig. 4 that has no basis in reality.
of quantum mechanical concepts, i.e., do the opposite It is a result of the wrong belief that the spins can
of what Feynman suggested. The remaining part of only be in the energy eigenstates shown in Fig. 1,
this paragraph summarizes a typical QM-inspired, combined with an attempt to explain how a magnet-
unsatisfying explanation of MR. The spin-up and ization can nevertheless be transverse.
spin-down states have an energy difference propor- In contrast, it is easy to demonstrate using either
tional to the magnetic field (Zeeman splitting). In classical or QM (see appendix, proposition 3) that a
equilibrium, they are almost equally populated with a homogeneous RF field can never change the relative
small surplus of nuclei in the low-energy spin-up orientation between noninteracting proton spins.
state. If the two-level system is subject to RF fields Hence RF fields can only rotate the spin-distribution
and if the photon energy matches the energy differ- as a whole. This immediately explains why it is suffi-
ence, transitions between the spin-up and spin-down cient to keep track of the local net magnetization in
state will be induced according to QM. Hence, the MR experiments and why RF fields cannot be used
population of the low-energy state can be excited to to change the size of this, even though Fig. 4 wrongly
the high-energy state. seem to indicate that this might be possible. Another
Even if this superficially sounds like a simple ex- immediate consequence is that RF fields can never
planation, it is not. First of all, it requires familiarity change coherence if this is defined as nonrandom
with concepts such as energy eigenstates, Zeeman phase relations generally (order). The concept of co-
splitting, and photons, concepts that are notoriously herence, however, is typically used for nonrandom
difficult to understand. Furthermore, the explanation phase relations in the transverse plane only, a some-
does not give any hint of the importance of coherent what unfortunate definition that will nevertheless be
evolution, which is crucial for the QM understanding adopted here (see appendix, proposition 3, for expla-
of MR. So the above has character of a pseudo-expla- nation). Only the combination of a polarizing field
nation, unless the reader is familiar with the QM and additional inhomogeneous fields associated with
equations of motion. These describe smooth transi- nuclear interactions create the skewness of the spher-

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
 UNDERSTANDING MAGNETIC RESONANCE 335

Figure 5 This figure shows how an RF field on resonance can rotate the spin distribution of
Fig. 3. As all spins precess, the distribution and the net magnetization rotates around the B0 field
direction. So does the orthogonal magnetic field vector associated with a resonant, circularly
polarized RF field. Seen from a frame of reference rotating at the common frequency, all appear
stationary except for a slow rotation of the spin distribution around the RF field vector. If this is
pointing toward the reader, the magnetization will be rotated in the direction indicated. [Color
figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

ical spin-distribution needed for having population DISCUSSION

differences and coherences. For the QM literate, it is
worth noting that coherence and population differen- QM and other laws of nature cannot necessarily tell
ces are two aspects of order: what appears as popula- us what really happens—they only describe our per-
tion differences in one basis are coherences in ception of nature. Some interpretations of QM imply
another and vice versa. Hence there is no conceptual that one should, in principle, not speak about what is
problem in field-assisted T1-relaxation to be the real really happening but only speak about past and future
source of coherence, though relaxation is normally outcomes of measurements (experience and predic-
considered a source of coherence loss. tions, (1)). This can be used to argue that the two
Figure 4 is plain wrong, while Fig. 1 is not when views expressed in Figs. 1 and 3 are equally good as
interpreted as a visualization of the somewhat irrele- the predictions they give rise to, are the same. It
vant eigenstates. Because RF fields can only rotate could even be argued that no such mental representa-
the spin-distribution as a whole, a better alternative tion should be made. The latter extreme view, how-
to Fig. 4 is one showing a rotation of the distribution ever, does not help the MR student in establishing an
shown in Fig. 1 so the cones end up in the horizontal intuitive feel of how MR works or what results are
plane. This too would seem highly nonintuitive, but expected. Even though such pictures (and all other
would nevertheless not be wrong in the sense that mental representation of the world) represent simpli-
experimental observations match predictions. Yet fications of reality and should be interpreted as such,
another—and much better—alternative to Fig. 1 is a they can still be immensely useful. The quality crite-
nearly spherical, precessing spin-distribution some- ria are intuitiveness, simplicity, and prediction accu-
what skewed toward magnetic north, as predicted by racy of which the latter is most important. In the
both quantum and classical mechanics, and shown in present case, the prediction accuracy of the mental
Fig. 3. The corresponding Fig. 5 replacing the mis- representations shown in Figs. 1 and 3 are the same,
leading Fig. 4 is similar except rotated so the slight if the user has sufficient insight to understand both.
overweight of spins, and therefore the net magnetiza- In particular, coherent evolution as expressed in the
tion, is pointing in a new direction. Schrödinger equation must be understood rather than

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
336 HANSON

just the semirandom spin-flipping that is wrongly and leave QM to the few, who can appreciate both
associated with the resonance phenomenon in some the subtle differences and the overwhelming agree-
tutorials. In contrast, classical mechanics as ex- ment with classical mechanics.
pressed in Figs. 2, 3, and 5 give intuitive and correct QM is considered somewhat exotic and intriguing
predictions understandable by most people. by many, and should this motivating factor not be
It is often said that the classical explanation is exploited? Should students not get a glimpse of the
adequate for most purposes but insufficient to really underlying physics, even if not needed? Even though
understand MR. This disregards the fact that QM QM is underlying classical mechanics, the physics
translates directly into classical mechanics in the underlying MR are classical. If MR is not sufficiently
present context. Such statements are often followed challenging for the students, or if they are sufficiently
by misinterpretations of QM based on Figs. 1 and 4 capable, they may indeed benefit from learning the
that raise more questions than they answer: Why are quantum physics of MR and how the classical limit
spins only oriented parallel or antiparallel to the field? is approached as expressed in the correspondence
In particular, why would nearly half the spins align principle (MR provides an excellent example of that).
opposite to the field? Why do RF fields reduce the But it is not justified to take any odd explanation and
phase spread on the two cones? How does the phase call it QM. Physics students, for example, may defi-
of the RF field translate into an azimuthal phase? nitely benefit from a QM derivation, especially if it is
How does the figure look for flip-angles different preceded by a classical explanation, that in addition
from 90 degrees? These questions cannot be answered to being intuitive express the same physics in the
satisfactorily from the figures that are neither consist- case of MR. A good example of this approach is pro-
ent with QM, nor with classical mechanics. The prob- vided by Levitt (2), who also advocate some of the
lems mentioned typically only occur on the very first views expressed in this article.
pages of MR introductions. The explanations get back It is also important to note that QM plays a role in
on track after a (semi-)classical picture is introduced MR, especially when described quantitatively. QM
and it is stated (typically without argument) that RF governs the nuclear interactions that are responsible
fields in a classical picture rotate the net magnetiza- for relaxation, for example. Although relaxation is
tion, which is inconsistent with Fig. 4. consistent with classical mechanics, the observed
In this article, it was argued that the spins are not sizes of relaxation rates are not. Only if these are
forced into eigenstates by their interactions with calculated based on QM do they match experiment.
environment. It is interesting to note, however, that Normally, however, relaxation rates are measured
if, by some means, the individual spins were brought rather than calculated from first principles. Hence,
into eigenstates before an MR-experiment as indi- examples like this do not warrant the use of QM for
cated by Fig. 1, subsequent observations would not nonphysicists. The differences are subtle and detailed
be changed. The expectation is, in other words, inde- knowledge is required to acknowledge them.
pendent of whether they are based on the nonintuitive
Fig. 1 or the preferred counterpart Fig. 2. This is not
surprising, but it raises a question: in which sense is CONCLUSION
the latter picture more correct? First of all, there is
nothing in QM telling us that the overall state collap- QM often get the blame for basic MR appearing
ses into a single-spin eigenstate as argued in the ap- complicated or even incomprehensible. This is dou-
pendix (proposition 1). Second, the appendix shows bly unjust: MR is not as complicated as it is often
that the relative orientation is not changed by RF claimed to be, and QM is not responsible because
fields (proposition 3), so even if Fig. 1 does not seem MR is a classical phenomenon.
all that nonintuitive, it certainly does after rotation of It is not only a matter of the classical description
the shown distribution around a horizontal axis. Such being preferable to the quantum counterpart for educa-
a rotation is induced by a 908 excitation pulse. tional purposes. It is also an issue that QM-inspired
Another blow against Fig. 1 and excessive use of descriptions often include nonintuitive interpretations
QM is delivered by Occam’s Razor that can be that are not supported by QM. In particular, there is lit-
described as follows: if there are two explanations tle basis in QM for the nonintuitive proposition that
for the same set of observations, choose the simpler. spins are forced into the spin-up and spin-down states
Using a scanner, it is extremely difficult to do experi- during MR experiments. As MR is a classical phenom-
ments that reveal the quantum aspects of MR. Hence, enon, MR educators fortunately do not have to engage
the natural consequence is to acknowledge that MR in QM-inspired descriptions that raise more questions
is accurately described as a classical phenomenon than they answer when presented in simplified forms.

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
 UNDERSTANDING MAGNETIC RESONANCE 337

APPENDIX The terms correspond to the two ways that a total

spin of zero can result from protons being in eigen-
This appendix contains sections that involve too states, but jc0 i is not itself a single-particle eigen-
much QM formalism to be understandable by most. state: it cannot be factorized and the states of particle
QM is here used to show that MR is adequately 1 and 2 are said to be entangled. Also it is seen that
described by classical mechanics. Each section head- the coherence is partially preserved. If more particles
ing below is a true proposition appearing in the main are present, there are more possible states with a
text, and it is followed by a QM-based justification. polarization near zero, and hence the measurement-
The basis for the calculations can be found in several induced loss of coherence becomes insignificant
textbooks including (2, 18, 19). when the total spin of many particles is measured
(the dimensionalities of the associated subspaces
Proposition 1: An MR Measurement increase). Consequently the individual protons are
Does Not Make the State of an never forced into their eigenstates by MR, and myth
Ensemble Collapse into 1 is not supported.
Single-Particle Eigenstates
It is shown that a measurement of the total spin (or
Proposition 2: QM and Classical
magnetization) of an ensemble of protons does not
Mechanics Give the Same Predictions
force the individual particles into their energy eigen-
states unless a polarization of 61 is measured (full The state vector formalism used so far becomes
alignment), which never happen in MR where the impractical when more protons are considered as the
polarization is close to 0. For the sake of clarity the dimensionality of the problem increases as 2N where
argument is made for just two protons, but it is N is the number of particles. A highly appropriate al-
straightforward to extend to more. ternative to the vector approach is the density opera-
The combined state vector for a system of two tor formalism that has significant advantages when
noninteracting protons is introduced. ensembles of identical systems are described and
when classical uncertainty and quantum indetermin-
jci ¼ c"" j""i þ c#" j#"i þ c"# j"#i þ c## j##i ism occur simultaneously (18). It is beyond the scope
of this article to describe this commonly used formal-
The four-dimensional state space is spanned by the ism in detail, but a few important points must be
product-states spin-up and spin-down for each of the made in this context.
two particles. The total spin operator is the sum of The density operator defined for a pure state jci
the individual spin operators for the two particles, S as r ¼ jcihcj, is descriptive of the QM state just as
¼ S1 þ S2. Any measurement will project the state the state vector itself. Coherent evolution under the
vector onto the eigenspace associated with the mea- influence of a Hamiltonian H is described by the
sured eigenvalue of the measurement operator. A Liouville equation, which is the density operator ana-
measurement of the spin of particle 1 along a direc- log of the Schrödinger equation [18]:
tion will therefore force the state vector into a corre-
sponding eigenstate of the associated operator. How- qr 1
ever, a measurement of the total spin along the same ¼ ½H; r
[A1]
qt i
h
direction will not necessarily force the system into an
eigenstate of the individual corresponding spin oper- The expectation value of an operator A is given by
ators. Looking at the equation above, it is seen only the trace of the product rA:
to be the case if a polarization of 61 is measured,
corresponding to parallel spins. Depending on the h Ai ¼ TrðrAÞ [A2]
measured value, the state vector will collapse into
j""i or j##i after such a measurement. These are On the basis of spin-up and spin-down states, the
indeed eigenstates of both spin operators. A measure- density operator has the matrix representation
ment of zero total magnetization, however, will not !
force the system into an eigenstate of any single jc" j2 c" c#
spin-operator. Introducing a renormalization constant r¼
c" c# jc# j2
k, the new state is
It is an important advantage of density matrices over
jc0 i ¼ kðc#" j#"i þ c"# j"#iÞ state vectors that they can be averaged over statistical

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
338 HANSON

ensembles in a meaningful way, i.e. so that Eqs. [A1] qhli

and [A2] are still valid. The density operator for an ¼ ghli  B [A6]
qt
ensemblePof N particles labeled i is defined as
r ¼ 1=N Ni¼1 ri . For individual nuclei, the probabil- This equation is remarkable: The individual mag-
ities of measuring energies corresponding to the netic moments and the macroscopic magnetization of
energy eigenstates are present in the diagonal ele- a sample evolve according to the classical equations
ments of the operator, whereas the complex phase of of motion. In particular, hli will precess around B at
the off-diagonal holds the information about the the Larmor frequency. The equation is equally valid
direction of the transverse magnetization. The aver- for a single proton, but in that case it must be
aged density operator is independent of whether it is acknowledged that it describes the mean of expected
calculated for an ensemble of nuclei prepared with outcomes of magnetization measurements rather than
random phase as indicated in Fig. 2 or of nuclei each a definite nuclear magnetization, as the existence of
being in spin-up or spin-down as indicated in Fig. 1. the latter is not consistent with QM.
In the first case, the off-diagonal elements average to
zero, whereas each nuclear density matrix is itself
diagonal in the latter. As predictions only depend on
Proposition 3: A Homogeneous RF Field
the density matrix, the two situations cannot be
Preserves the Relative Orientation of Spins
distinguished by experimental observations alone.
If the density operator is diagonal, the state is said It is shown that a homogeneous magnetic field never
to be incoherent. The distinction between coherent changes the relative spin-orientation of noninteract-
and incoherent states is somewhat arbitrary, however, ing protons. This is true for both static and RF fields.
as population differences (differing diagonal elements It implies that RF fields can only rotate the spin dis-
of the density operator) can be exchanged for off- tribution as a whole. This proposition follows from
diagonal coherence terms if a simple change of basis commutator relations: If a Hermitian operator com-
is performed by applying a unitary transformation. mutes with the Hamiltonian, the corresponding
The components of the proton magnetic moment observable is constant in time. For two nuclei with
l are conveniently expressed in terms of raising and spins S1 and S2, the Hamiltonian in a magnetic field
lowering operators Sþ ¼ Sx þ iSy and S ¼ Sx  iSy. B(t) is H ¼ g(S1 þ S2)  B(t). This expresses that
Because the operator associated with magnetic the energy is low when the nuclei are parallel to the
moment along the x direction is mx ¼
h2g ðSþ þ S Þ, field. The scalar product S1  S2 is proportional to the
the expectation value of mx is cosine of the angle between the two spins. Hence it
suffices to show that the commutator [H, S1  S2] is

hg zero. This follows from relations for products of the
hmx i ¼ Traceðmx rÞ ¼ ðr þ r"# Þ [A3]
2 #" components of spin components [19]:

The Liouville equation provides the associated time h
2 i
h

evolution: S2x ¼ S2y ¼ S2z ¼ and Si Sj ¼ Sk [A7]
4 2


q h mx i
hg qr#" qr"# g where indices i,j,k ¼ x,y, or z, in cyclic order.
¼ i þi ¼ ð½H; r
#" þ ½H; r
"# Þ
qt 2i qt qt 2i Because the relation is true for any two nuclei, it is
g true generally, as expected on classical grounds also.
¼ ððH#"  H"# Þðr##  r"" Þ
2i The axis of precession may fluctuate, but the shape
þ ðr#"  r"# ÞðH##  H"" ÞÞ ½A4
of the spin distribution remains unchanged as it is
merely rotated at any instant in time. In contrast, the
Consequently, for the dipolar Hamiltonian, H ¼ inhomogeneous fields created by the nuclei them-
l  B, selves change the relative orientation of nuclear spins
and hence also the magnitude of the net magnetiza-

 tion. These fields are responsible for relaxation.
q h mx i By ðr""  r## Þ Bz ðr#"  r"# Þ
¼
hg2 þ Starting from the incoherent equilibrium state,
qt 2 2i
  application of a 908 pulse makes the state coherent in
¼ gBy hmz i þ gBz my ½A5
the above mentioned sense. So the 908 pulse can be
perceived as the source of the coherence. This is true
Cyclic permutation finally provides the general in a trivial sense, but is misleading. What appears as
formula. a population difference in one basis is coherence in

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
 UNDERSTANDING MAGNETIC RESONANCE 339

another. The 908 pulse rotates the state vector so as less reduces to the sum of just two terms. In contrast,
to transform the population difference in the spin-up/ the classical partition function remains an infinite
spin-down basis into coherence. With another choice sum (integral). The energy E(y) ¼ mB0cosy
of basis, the same situation will be interpreted quite depends on the angle y between the field B0 and the
differently, e.g. opposite. The real source of the magnetic moment l. The angular distribution again
coherence is therefore not the 908 pulse but the field- obeys Boltzmann statistics:
associated longitudinal relaxation that created the
population difference—the 908 pulse only made the expðEðyÞ=kTÞ
PðyÞ ¼ R p [A10]
population difference detectable as expressed in the expðEðyÞ=kTÞ sin ydy
0
coherence terms of the density operator. The typical
pffiffiffi
use of the word coherence as referring to azimuthal, The magnetic moment of a proton is m ¼ 3=4
hg
nonrandom phase relations thus unfortunately implies which follows from Eq. [A7]. These properties are
an apparent ability for RF pulses to change coherence used to calculate the equilibrium magnetization:
whereas homogeneous RF fields in reality only can Z p
rotate the ensemble as a whole and therefore never hmz i ¼ PðyÞðm cos yÞ sin ydy [A11]
can change the coherence generally defined as non- 0
random phase relations (azimuthal or polar angles).
R1
expðmB0 u=kTÞu du h
2 g2 B0
Proposition 4: Classical and QM Predict ¼ m R11  [A12]
expðmB0 u=kTÞdu 4kT
the Same Equilibrium Magnetization for 1

Small Degrees of Polarization The last approximation is valid for small degrees
The equilibrium magnetization is first calculated of polarization. For high polarizations, e.g. for
using QM and Eq. [A2]: When expressed based on
hgB0 > kT), the classical and quantum predictions
energy eigenstates, the expectation value of the differ, which is easily appreciated: Classically, the
energy is the sum of eigenenergies each weighted by maximum net magnetization is reached at zero tem-
the probability of measuring that energy. From an perature when the nuclei are perfectly aligned and
energy accounting point of view, it therefore appears each contributes a magnetization of m. But even at
as if all nuclei are in their eigenstates, even zero temperature, the spins are not perfectly aligned
when they are not. Hence the relative populations in agreement with QM. Hence, each nucleus only con-
expressed in the diagonal elements of the equili- tributes a longitudinal magnetization of
hg=2. At tem-
brium density operator are given by the Boltzmann peratures and fields relevant for liquid state nuclear
distribution: MR, polarizations are small (e.g., 106), and quantum
and classical predictions are equal as shown.
expðE" =kTÞ Interestingly, the quantum derivation appears sim-
P" ¼ pler than the classical counterpart, which can be used
expðE" =kTÞ þ expðE# =kTÞ
as an argument for choosing a QM approach to MR-
expð
hgB0 =2kTÞ teaching. Whereas the quantum derivation is easier
¼ [A8]
hgB0 =2kTÞ þ expð
hgB0 =2kTÞ
expð
with respect to the use of calculus, it requires more
insight a priori. It is unfortunate that the math may
The similar expression for P; differs only by the seem to indicate that all spins are in eigenstates,
sign of the numerator exponent. The net longitudinal which is not the case as explained in the appendix
magnetization per nucleus mz is calculated from Eq. (proposition 1). Another unfortunate aspect of the
[A2]: quantum derivation is that it implicitly relies on the
validity of classical mechanics since classical Boltz-

g
h h 2 g 2 B0

mann-statistics are employed rather than Fermi-sta-
hmz i ¼ ðP"  P# Þ  [A9] tistics that describes the properties of ensembles of
2 4kT
half-integer spin particles (20). Hence, there is no
The last approximation is valid when the thermal guarantee that the quantum derivation is valid in the
energies far exceed the energies associated with spin domain where classical mechanics fail.
orientation, i.e. when the degree of polarization is For MR-tutorials aimed at nonphysicists, it is con-
small. sequently suggested that2 the expression for the result-
g2 B0
All spin orientations are possible according to QM ing net magnetization,
h 4kT that is common to quan-
but as argued earlier, the partition function neverthe- tum and classical mechanics, is presented as a result

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a
340 HANSON

of the slight skewness of the field-generated orien- 12. Rabi II, Ramsey NF, Schwinger J. 1954. Use of rotat-
tational spin-distribution. This is logical and true ing coordinates in magnetic resonance problems. Rev
in both cases. Neither derivation contributes much Mod Phys 26:167–171.
clarity for nonphysicists anyway. 13. Hargreaves B. 2005. MRI movies.http://www-mrsrl.
stanford.edu/brian/mri-movies/. Retrieved: 2008.
14. Hanson LG. 2007. Graphical simulator for teaching
ACKNOWLEDGMENTS basic and advanced MR imaging techniques. Radio-
Graphics 27:e27.
Drs. Lise Vejby Søgaard and Jens E. Wilhjelm are 15. Feynman RP, Vernon FL, Jr., Hellwarth RW. 1957.
gratefully acknowledged for comments that led to Geometrical representation of the Schrödinger equa-
numerous improvements in the manuscript. tion for solving MASER problems. J Appl Phys 28:
49–52.
16. Rabi II. 1937. Space quantization in a gyrating mag-
netic field. Phys Rev 51:652–654.
REFERENCES 17. Schild HH. 1990. MRI Made Easy. Berlin, Germany:
Schering AG.
1. Bell JS. 1987. Speakable and Unspeakable in Quan- 18. Callaghan PT. 1983. Principles of Nuclear Magnetic
tum Mechanics. Cambridge, England: Cambridge Resonance Microscopy. Oxford: Clarendon Press.
University Press. 19. Bransden BH, Joachain CJ. 1989. Introduction to
2. Levitt MH. 2008. Spin Dynamics: Basics of Nuclear Mag- Quantum Mechanics. Harlow, England: Longman Sci-
netic Resonance, 2nd ed. Chichester, England: Wiley. entific and Technical.
3. Higgins DM. 2003. ReviseMRI web site, http://www. 20. Landau LD, Lifshitz E. 1980. Course of Theoretical
ReviseMRI.com/. Retrieved: 2008. Physics: Statistical Physics, 3rd ed., Part 1. Oxford,
4. Bushong SC. 2003. Magnetic Resonance Imaging. England: Pergamon Press.
Missouri, USA: Mosby Inc.
5. Rinck PA. 2003. Magnetic Resonance in Medicine,
5th ed. Berlin, Germany: ABW Wissenschaftsverlag
GmbH.
BIOGRAPHY
6. Westbrook C, Roth CK, Talbot J. 2005. MRI in Practice,
3rd ed. Oxford, England: Blackwell Publishing Ltd.
7. Gadian DG. 1982. Nuclear Magnetic Resonance and Lars G. Hanson, M.Sc., Ph.D. is a Senior
Its Applications to Living Systems. New York, USA: Scientist at the Danish Research Centre
Oxford University Press. for Magnetic Resonance (DRCMR) at Co-
8. Hornak JP. 1996. The Basics of MRI. Web book, penhagen University Hospital Hvidovre in
http://www.cis.rit.edu/htbooks/mri/. Retrieved: 2008. Denmark. LGH did graduate work at the
9. Hoa D, Micheau A, Gahide G. 2006. Creating an interac- Max Planck Institute for Quantum Optics
in Germany on atomic physics and
tive web-based e-learning course: a practical introduction
obtained his Ph.D. from University of Co-
for radiologists. Radiographics 26:e25; quiz e25.
penhagen on fast spectroscopic MR imag-
10. Farrar TC, Becker ED. 1971. Pulse and Fourier ing in 2000. The work was partly per-
Transform NMR. New York, USA: Academic Press. formed at the DRCMR, partly at Stanford University, California.
11. Gerlach F, Stern O. 1922. Der experimentelle nach- Current interests include metabolic imaging, MRI sequence
weis der richtungsquantelung im magnetfeld. Zeit- design, magstripe EEG-fMRI, and educational aspects including
schrift für Physik 9:349–352. visualization software.

Concepts in Magnetic Resonance Part A (Bridging Education and Research) DOI 10.1002/cmr.a