P

VOLUME 26, NUMEEIt 2 APRIL, 1954

se oI: .Xota1ling Coorc inates in .V. .agnetic

:resonance . '.ro ~. .ems
I. I. RA3r, Columbia Urfiversity, Sm Fork, Xm York

N. F. RAMszv AND J. ScHwlNczR, Harvard University, Cambridge, Massachusetts

The use of a rotating coordinate system to solve magnetic resonance problems is described. On a coordi-
nate system rotating with the applied rotating magnetic field the e6'ective Geld is reduced by the Larmor
Geld appropriate to the rotational frequency. However, on such a coordinate system problems can more
readily be solved since there is no time variation of the field. The solution in a stationary frame of reference
is then obtained by a transformation from the rotating to the stationary frame. This procedure is equally
valid in classical and in quantum-mechanical problems. The method is applied both to the molecular beam
magnetic resonance method and to resonance absorption and nuclear induction experiments.

I. INTRODUCTION onance calculations is so great that a detailed descrip-

HE theory of the va, rious molecular beam magnetic tion of it is needed. This need is sufficiently great that
resonance methods and of the resonance absorp- several authors' ' have had to include a partial descrip-
tion and nuclear induction experiments is usually tion of the rotating coordinate method in order to
chieQy concerned with the calculation of the eGect of
describe their experiments electively.
weak oscillating or rotating magnetic fields on nuclear The rotating coordinate system method is equally
Inagnetic moments in the presence of a strong constant applicable to classical and quantum-mechanical systems.
magnetic field. Some of the simplest problems of this Because of the great simplicity and extensive appli-
sort were first solved by Rabi, ' Schwinger, ' and Bloch' 4 cability of the classical description it will be given first
in detail while the quantum-mechanical case will be
by a straightforward quantum-mechanical calculation
of transition probabilities or by related methods. Al- considered in the last section of this paper.
though these methods are consistent with and closely II. CLASSICAL FORMULATION OF ROTATING
related to the one described here, they are not as well COORDINATES PROCEDURE
suited to the simplified analysis of many more com-
Consider a system consisting of one or more nuclei or
plicated problems. The extensive and explicit use of
atoms all of which have the same constant gyromag-
the rotating coordinate system was first developed by
netic ratio y. Then, if I is the nuclear angular momen-
successive contributions of Bloch4 and the present
tum in units of k, the nuclear magnetic moment is
authors' over eight years ago. Since the method was
yhl and the equation of motion of the system in a sta-
originally considered chieQy as a new technique of
tionary coordinate system is
calculation rather than as an intrinsically new result, no
attempt was made to describe the method in the dI
literature. However, in subsequent time it has become h, — = q&&H=~hi&&H.
dt
apparent that the value of the method in nuclear res-
' I. Rabi, Phys. Rev. 51, 652 (193'7). But if 8/itt represents differentiation with respect to
' J. I.Schwinger, Phys. Rev. 51, 648 (1937); Rabi, Ramsey, and a coordinate system that is rotating with angular
Schwinger, Private communications and lecture notes (1945 on).
' F. Bloch and A. J. Siegert, Phys. Rev. 57, 522 (1940); L. ' ¹ Bloembergen, iVttctear Magnetic Retaxation (Schotanus and
Alvarez and F. Bloch,
Phys. Rev. 57, 111 (1940). Jens, Utrecht, Holland).
'F. Bloch, Phys. Rev. 70, 460 (1946); R. K. Wangsness and ' H. C. Torrey, Phys. Rev. 76, 1060 (1949).
F. Bloch, Phys. Rev. 89, 728 (1953). ' E. L. Hahn, Phys. Rev. 80, 580 (1950).
167
 RA B I, RAMSEY, AND SCHWINGER
system is much simpler in the rotating coordinate sys-
tem than in the stationary system rom Eq. (6) it F.

= QJ k/p follows that the magnitude of the effective magnetic

QJ/y field is

j
~„. = $(IIp
~
'+H is]l = a/y,
a/p)—

with ~p by defininition being pHp. Likewise the angle

0 of H, „relative to Hp is given by
cosO'= (&up
—pp)/a sinO= (a»Hi/Hp)/a. (9)
From this it is apparent that when rp=Np 0=90' and
a magnetic moment initially parallel to Hp can precess
about H„until it becomes antiparallel to Hp. In other
words, such a moment can have its orientation relative
to Hp changed most completely when co=cop so Mp can
FIG. 1. Fffective magnetic held in the rotating be considered as the resonance frequency of the system.
coordinate system. If one next goes to a second rotating coordinate sys-
tem which rotates about H, „with a suitable angular
velocity ~, velocity, the effective field H„„ in the doubly rotating
dI BI
— =— + ppXI, (2)
system can be reduced to zero. In this doubly rotating
coordinate system the problems become trivial since
dt Bt
there is then no magnetic field and consequently no
where I on both sides of the equation is the angular change in the orientation of I. If a is the angular velocity
momentum as measured by the stationary observer, about H„which reduces H, „, to zero, a must be deter-
but the BI/Bt represents how a rotating observer would mined by
find the stationary observer's I to vary as a function
a
of time. By rearrangement Eq. (2) becomes 0 = Herr= Her+ (10)
7
h —=ykIX (H+ep/y)
eli
=ykIXH, „, Hence, from (7) if n is a unit vector parallel to H, „,
Bt
a= —yH, „=—an,
where H, „ is the effective field in the rotating coordinate
system and is defined by where a is the quantity previously defined in Eq. (8).
As is shown in a later section of the present report,
H, „=H+ pi/y. (4) the above considerations apply to a quantum-me-
In other words, the eGect of the rotation of the coordi- chanical as well as to a classical system. Consequently,
nate system is merely to change the eGective field by either classically or quantum mechanically on the
the added term ep/y. doubly rotating coordinate system with the two rota-
This result can readily be applied to interpret the tional angular velocities ~ and a, there is no effective
effect of the rotating magnetic fields used in the various resultant field and the state of the system remains con-
nuclear resonance experiments. In most of these there stant in time.
is a constant field Hp about which another (usually
much weaker) field Hi perpendicular to Hp rotates with
III. CLASSICAL INTERPRETATION OF NUCLEAR
RESONANCE EXPERIMENTS
angular velocity — ~. However, from the point of view
of a coordinate system rotating with Hi, none of the The above can be directly applied to various problems
magnetic fields are changing as a function of time. arising in nuclear resonance experiments. Although
Therefore, the axes of the rotating coordinate system oscillating instead of rotating fields are usually used in
can be selected so that these experiments the problems can usually be treated
—Hpk, H, = EE,~, as ones involving a rotating field since an oscillating
Hp p~ = —&uk. field is equivalent to two opposite rotating fields and
Then on the rotating coordinate system, only the component rotating in direction to be able
to give a resonance in Eq. (8) has an important effect
H„= (Hp —(u/y)k+Hii (6) '
in most problems. As a first application the method
as is shown schematically in Fig. 1. Since this field is can be used to demonstrate the criterion for the rate
constant in time the solution to the motion of the ' F. Bloch and A. J. Siegert, Phys. Rev. 57, 522 (1940).
 ROTATI N 6 COOR 0 I NATES I N MAGNET I C RESONANCE P ROB LE M S 169
"
of change of a field to be "adiabatic, i.e., to be such the expectation values satisfy the same equation. There-
that a nuclear moment preserves its magnetic quantum fore, a correspondence between the classical and
number (classically its angle) relative to the field as quantum-mechanical solutions can be established by
the field is moved. Let the field be rotated with angular requiring them to agree on their predicted average s
velocity — orI. Then for this problem, in the notation of component and on the total probability. If I'~~ is the
the previous section, Ho is zero and HI is the full field probability of a system of spin ~ being in the state of
H. On the rotating coordinate system then magnetic quantum number m equal to &-, these re- '
quirements are
H, „=( —~,/p)k+Hi. (12) I'~ — I' ~=cosQ,
The nuclei will then preserve their orientation relative (16)
to H provided H„ is approximately equal to Hi or that
Pt+P-;=1.
Therefore,
~", ~«ap,
1—cosa G(ts tr)—
which can be written alternatively as = sin'0" sin'

HXHi/a «~e. (14)

a(ts —
i

(~oai/EZs)' tr)
The use of the rotating coordinate system also sin' . (17)
provides a simple pictorial interpretation of the transi- (cop
—co)'+ (copHr/EIp)'
tion process that occurs in the molecular beam magnetic
resonance method originally introduced by Rabi. ' A As proved in Sec. V, this is exactly the same result that
is obtained from a pure quantum-mechanical calculation.
singly rotating coordinate system rotating with the
oscillator frequency — or can be used throughout. Prior Likewise the rotating coordinate system analysis
to the molecule reaching the rotating field region H~ procedure is applicable to the molecular beam resonance
equals zero and H„has the value (Hs —
method with separated oscillating fields introduced by
&u/p)k. As the
molecule enters the transition region where the rotating
"
Ramsey. In this case, the description through the first
field is being established H, „changes. Conditions are oscillating held is just the same as in the preceding
usually such that near resonance the condition of Eq. paragraph. After leaving the oscillating field in this
method the nucleus enters an intermediate region
(14) applied to H, „ is violated. Consequently the transi-
tion under such circumstances is not adiabatic and the where there is only IXO so the magnitude of H& is zero.
nuclear moments do not follow H, „as Hr is established. Relative to the singly rotating coordinate system, the
After Pr achieves its full value H, „ is (Ps — nucleus in this region precesses about (Hp — cv/p)k
"/y)k+Hri
and the nuclei precess about this effective field. When until it reaches the second oscillating field. As a result
the molecules leave the rotating field region B„again of this precession the nucleus will in general have a
changes too rapidly for the nuclei to follow and they are
diBerent orientation relative to H, „ in the second
left with the orientation relative to the s axis to which rota, ting field region than it did in the first. On the
other hand, if the average value of Hs — a&/y in the in-
they have precessed in the region of the rotating field.
At exact resonance this precession is about a field Hr termediate region is zero the orientation of the nucleus
which is perpendicular to the original direction of the relative to H, „ is the same in the second oscillating field
field and consequently the change of orientation can region as in the first. This will be true regardless of the
be large. velocity of the molecule. However, if the average of
The qualitative analysis of the preceding paragraph
8', —co/y has any value other than zero, the orientation
can be also expressed quantitatively. Assume the I of the nucleus relative to H, „ in the second field is
is initially parallel to Hp. Then in therotating coordinate difterent from that in the first and the magnitude of
system I will precess about H, „with the precession the difference will depend upon the velocity of the
angle 0' and at an angular velocity a. If a is the angle molecule. When the combined eGects of the two rota, ting
between Hs and I, the simple geometry of the above field regions are averaged over the velocity of the mole-
precession is such that after a time interval t2 — tI,
cules, it is therefore found that the transition prob-
ability is a maximum for or equal to the average value
cosa=cos'0~+sin'O~ cosa(ts —tr) of yIIO in the intermediate region.
The rotating coordinate system procedure is also of
a(ts —tr) value in interpreting the various nuclear resonance ab-
=1—2 sin'O~ sin'
sorption and induction experiments based on the original
experiments of Purcell" and Bloch. Experiments of"
On the other hand, as shown in the next section, the these types have by now been carried out in a number
quantum-mechanical angular momentum operators of different ways. One of the marked differences has been
satisfy the same Eq. (1). Since this equation is linear ' N. F. Ramsey, Phys. Rev. 78, 695 (&,950).
'Kusch, Miilman, and Rabi, Phys. Rev. SS, 666 (1939); "Purcell, Torrey, and Pound, Phys. Rev. 69, 37 (1946).
Kellogg, Rabi, Ramsey, and Zacharias, Phys. Rev. 55, 318 (1939). ~ Bloch, Hansen, and Packard, Phys. Rev. 69, 127 (1946).
170 RAB I, RAMSEY, AND SCH WINGER
in the extent' to which the adiabatic condition of Eq. Other limiting cases with the various relaxation
(14) is satisffed for H, „. In some experiments a large processes have been discussed in the literature. 4 ' As "
value of the rotating 6eld HI has been used and the has been discussed by Hahn, ' the rotating coordinate
6eld H p has been varied about the resonance value at system procedure is particularly well suited to the de-
a sufficiently slow rate that Eq. (14) was satisfied. In scription of the spin echo phenomena.
this case the original magnetism,
IV. QUANTUM-MECHANICAL FORMULATION OF
Mo ——
xHp, ROTATING COORDINATES PROCEDURE
simply retains its orientation relative to H, „ in the Relation (4) derived above can be proved equally
singly rotating coordinate system and the dominant well quantum mechanically as classically. As indicated
eGect of the small changes in Hp ls to change the orienta- in the previous section, one procedure is simply to say
tion of H, „and hence Mo. Therefore the x component that Eq. (1) applies to the quantum-mechanical angular
of this in the rotating system varies with Hp as momentum operators. Equation (1) can in fact easily
be proved to follow from the standard operator relation
Mgv pPi/Pp
M =Mpcose=
— (19) (k/i)dI/dt= Px, Ij=hei —IK, (23)
t (a 0 (o)'+ ((ooPi/Ho)']&
where the Hamiltonian 3C is taken as
3f, then goes through a resonance maximum at co=~p.
However, as the coordinate system and hence M are x= —ykI H. (24)
rotating with frequency co relative to the laboratory Alternatively, from a wave mechanical point of view,
system, the resulting rotating magnetic moment will the Schrodinger equation for the problem relative to
induce a signal in a suitably placed coil. a stationary coordinate system is
Actually the preceding case rarely applies since the
local magnetic 6elds from neighboring molecules cause iVi4 =X%= —ykI H%. (25)
the nuclei to return to the thermal equilibrium relative
However, in quantum mechanics the Rnite rotation
to Ho and to lose their phase coherency about Ho. The operator'4 for the coordinates to be rotated an angle i
characteristic time constants for these processes are con-
about an axis along which the system's angular momen-
ventionally designated T& and T&.4 From the point of
tum, I, has the component Ir is the unitary operator
view of the singly rotating frame of reference, TI is the
exp(ifIr). Let 4' and H be the wave function and
characteristic time for resuming thermal equilibrium field relative to a stationary coordinate system whereas
about the z axis, and T2 is the characteristic time con-
V„and H„are the same quantities relative to coordinates
stant for losing coherency about that axis, i.e., for the
rotating with angular velocity u. These quantities are
g and y components of the magnetization to average
related by the unitary transformation so that"
to zero. Although the preceding results are not directly
applicable when these relaxation eGects are important, %=exp( —i~ It)%„, (26)
they are nevertheless of value in many approximate
calculations even then. For example, if T&«T&, Ti«1/
I H„=exp(im It)I. H exp( —ie It). (27)
yH, , and (T,Ti)'«1/yHi, the rate of absorption of If Eq. (26) is substituted in Eq. (25) and if Eq. (27) is
energy at resonance in a nuclear resonance absorption used to simplify the resulting expression one immedi-
experiment with nuclei of spin 2 can roughly be approxi- ately obtains
mated by the assumption that the transition goes as in
Eq. (17) but is stopped by the loss of all coherency ik+„= —~kI (H„y~/7)e„= —~hI H„e„, (28)
after a time T~. At exact resonance and for a time T., where H„ is given by Eq. (4), with the understanding
Eqs. (17) and (8) give for the transition probability that H is to be expressed relative to the rotating coordi-
P;= sin'(yHiTi/2) = 4y'BPT~'. (20)- nate system. This result justifies the application of the
previous discussions to quantum-mechanical as well as
The average rate TV of transition per nucleus is then classical systems.
approximately given by
OF
—'y'BPTg. (21) —-
V. QUANTUM-MECHANICAL CALCULATION
W =P *,/Ti TRANSITION PROBABILITY
Therefore, the average rate R of net absorption of The probabilities for transition of the system from
energy is approximated by a state of one magnetic quantum number to another
P= Wn k(o= (1/8)y'HPTp:Vk'oP/kT (22) can readily be calculated with the above procedure.
Consider the case discussed earlier in which the mag-
where n, =-,'Eke&/kT is the number of surplus nuclei
in the lowest energy state, Ã is the total number of "Bloembergen, Purcell, and Pound, Phys. Rev. 73, 679 {1948).
'4 E. C. Kemble, J Nndementel I'rirIczples
nuclei present, ken is the energy separation of the two of Qmentmm 3feclzunzcs
{McGraw-Hill Book Company, Inc. , New York), pp. 247, 307,
orientation states, and T is the absolute temperature. and 532.
 ROTATING COORDINATES IN MAGNETIC RESONANCE PROBLEMS 171

netic 6elds are given by Eq. (5) and assume that up In general, the numerical evaluation of (33) is some-
to time tI the magnitude of H~ is zero after which it is what cumbersome because of the noncommutativity of
HI until time t2. As use will be made of the stationary, the terms in the exponent. However, a series expansion
singly, and doubly rotating coordinate systems pre- of the exponential may be used. In the special case of
viously discussed, wave functions relative to these three spin i~, it becomes much simplified for then I=~n,
systems will be designated as %(t), 'k„(t), and 4, „(t), where e is the Pauli spin matrix. Since for the Pauli
respectively. Since H, „, for the doubly rotating coordi- spin matrix (n n)' equals one, the series expansion of
nates is zero, the above exponential together with the series expansion
for sine and cosine m erely give
4„(t2)=4'„(ti) =@,(ti). (29)
However, as the doubly rotating system between times ex
exp —
z —I ne
ti and t~ has rotated an angle — a(t~ — j
Z q
ti) relative to the & 2
singly rotating system one must use the previous finite
rotation operator to relate 4'„„(t2) and %, (t2) with the a(t, -t,) a(t, —t, )
result that
= cos +in n sin
2
4'„(t ) = exp (ia[t, —t, $n. I)4„„(t,). (30) a(t2 —ti) a (t, —t, )
=cos +i(g, sinO. ~+0., cosO) sin — . (34)
Hence from (29), 2 2
4', (t2) =exp(ia[t2 —t&)n I)@„(ti).
Therefore
a(t, —t, )
In a similar fashion, this can be reduced to a non-
rotating coordinate system with the result P;;=sin'0' sin'
2
(35)

@(t2) =exp(ia&t2k. I)exp(ia[t2 —tijn I)

which is the desired transition probability that is
Xexp( —icatik I)+(ti). (32) applicable to the conventional molecular beam res-
It should be noted, however, that because of the com- onance method. ~ It should be noted that the above
ponents of I not commuting, one cannot perform all agrees with the classically derived expression in Eq. (17).
the operations appropriate to exponentials of ordinary This procedure may also be used to calculate the
transition probability applicable to the molecular beam
numbers; instead, the exponentials may be taken as
de6ned by their series expansion. resonance method with separated oscillating 6elds. In "
From the above, the transition probability from a this case, Eq. (32) can be applied separately to the first
state m to a state m' can be calculated by taking 0 (ti) oscillating 6eld region, to the intermediate region, and
in which case to the second oscillating field region. The resulting three
equations can then be combined to express the anal
P ~
=~ (4, @(t)—
~2= (m'j exp(i~dt2k. I)
—
state of the system in terms of the initial state and
Xexp(ia[t2 ti]n I)exp( icotk I) (m) (' from this the transition probabilities can be calculated
= (m'[ exp(ia[t2 —ti]n I) (m) ('. (33) as in Eq. (25).
The authors wish to thank Professor F. Bloch and
t

The last simplifying step is a consequence of 0 and Professor E. M. Purcell for many stimulating dis-
4' being eigenfunctions of k. I.
~ cussions of this subject.