PH YSI CAL REVI EW VOLUME 104, NUMBER 6 DECEMBER 1S, 1956

Spin Relaxation Processes in a Two-Proton System*

N. BLOEMBERGEN
Crit tuboratory, Zarpard Urseperstty, Caetbredge, Massaehttsetts
(Received September 11, 1956)

The general theory of nuclear spin relaxation, based on the Boltzmann transport equation for the density
I
matrix, is applied to the very simple, nontrivial system of two identical spins = ~. A proton pair undergoing
hindered rotation around one axis shows a resolvable doublet. The relaxation processes in this doublet are
investigated. Explicit formulas for the longitudinal and transverse relaxation times are derived, which
display a dependence on the angles of the axis of rotation with the external magnetic Geld and the radius
vector. Experimental evidence for this dependence is discussed. General expressions for the Overhauser
and saturation e6'ects with two applied radio-frequency Gelds are given, which may also be applied to a
I
magnetic ion or nucleus with = 1 and crystalline Geld splitting.

I. INTRODUCTION II. THEORY OF SPIN RELAXATION IN

' A PROTON PAIR
JAKE' has shown that a static pair of protons gives
rise to a resonance doublet. Gutowski and Pake' A pair of protons with dipolar and exchange inter-
extended the theory to the case of a proton pair rotating action is embedded in a constant magnetic Geld Ho
rapidly about an arbitrary axis. Again a doublet results, taken in the z direction. The Hamiltonian is given by
but with a different splitting. The theory for a static
X Xm+ 5Cdip+ Xexy
or rotating proton triangle was developed by Andrew
and Bersohn. 3 Numerous experimental investigations of 3e„=yh(It'+Is')Zp,
the magnetic resonance of proton pairs, triangles, and
SC. =Alt Is,
tetrahedra have subsequently been made. 4
The spin-lattice relaxation in these structures has, 3cd' Y It rls jI1' Is 3(ll' rip) (Is' rls)rls }. (4)
however, received little attention. Solomon' ' has dis-
cussed the relaxation for a pair of nonidentical spins Since the Hamiltonian is symmetric in the two spins,
which undergo completely random motions and have the antisymmetric singlet state is a constant of the
motion. No transitions to the triplet state can occur.
no resolved fine structure. In this paper the relaxation
The singlet state has zero magnetic moment and is unob-
effects in a doublet are discussed, when motion around
servable in the context of this paper. The dynamical
only one axis is present. Two identical spins ', with
—, I=
dipolar interaction and a restricted motion, constitute
behavior of the three triplet states, denoted by 0, +,
—,corresponding to a magnetic quantum number
perhaps the simplest nontrivial system, to which the Ii'+Is' 1, 0 or —1, is d— —escribed by the quantum-
general theories of nuclear spin dynamics may be mechanical equation of motion for the three-by-three
applied. These have recently been formulated by spin density matrix o'.
various authors, ~ ' and are based on the Boltzmann
r)o/r)t = —i)t—'[X,o.7. (~)
transport equation for the density matrix, introduced
"
by Bloch. We shall follow a particularly concise for- Since the exchange interaction is a multiple of the
mulation due to Redfield. ' This exercise in the appli- unit matrix and commutes with the Hamiltonian (1),
cation of the general theory will show that the experi- it does not affect the motion of the triplet spin system
and is henceforth omitted. The dipolar interaction
mental study of relaxation of proton pairs, and other
consists of a part which depends explicitly on the time
groups of nuclear spins, may give additional information
and a time-independent part. The completely random
on the structure and internal motion in solids.
motion in a liquid leads to an averaging-out of the
* Supported the Joint Services. dipolar interaction with a single resonance line, and
' G. E. Pake,byJ. Chem.
Phys. 16, 327 (1948). has been discussed extensively. The case of interest
' H. S. Gutowski
and G. E. Pake, J. Chem. Phys. 18, 162 (1950).
here is a restricted motion of the proton pair around
' E. R. Andrew and R. Bersohn, J. Chem. Phys. 18, 159 (1950).
4 See references quoted in H. S. Gutowski, Ann. Rev. Phys. a single axis, which we shall designate the z axis.
Chem. 5, 333 (1954). Two resonance lines are distinguishable with this
s I. Solomon, Phys. Rev. 99, 559 (19SS).
6I. Solomon and N. Bloembergen, J. Chem. Phys. 25, 261 restricted motion. Let the axis of hindered rotation
(1956). make an angle fp with the external magnetic field Hp.
s R. Kubo and K. Tomita, J. Chem. Soc. Japan 9, 888 (1954). Let the radius vector rI~ make a constant angle 8' with
' F. Bloch,
Phys. Rev. 102, 104 (1956). the axis of rotation. The variable azimuth is denoted by
A. G. Redfield, I.B.M. Journal (to be published). The author
p
p'(t). The dipolar interaction can be expanded into
is indebted to Dr. Red6eld for communication of his results before
publication. spherical harmonics with respect to the z axis, and
"R. K. Wangsness and F. Bloch, Phys. Rev. 89, 728 (1953). these in turn can be transformed to spherical harmonics
i542
 SPIN RELAXATION PROCESSES j.543

in the primed coordinate system. The coefficients in this Here 0~~) is the diagonal spin density matrix corre-
transformation are the irreducible representations sponding to thermal equilibrium at the lattice tem-
n&2&(o, fp, o) of the group of three-dimensional rota- perature T.
"
tions, described by Wigner. With the introduction of
the raising and lowering operators I+=I,&I„, one can o &r&=exp( —E /kT)/P exp( —E /kT). (13)
write
Equation (12) gives a complete dynamical description
xd v +f k r12 (I1 I2 3I1 I2 )I ' p (0) (6a) for any magnitude of the external field Ho, including
zero. It contains all interference and nonsecular per-
7 k r12 (Il I2 +Il I2 )++1 (0 4') (6b) turbation eBects between the five independent elements
—1y h, r12 2I1+I2+Py21 (8 $) ~
of the density matrix, whose trace is normalized to
(6C)
unity. The five independent coupled differential equa-
with
tions, represented by the operator relation Eq. (12),
2
z„'(e,y)= p n„„. would still be difficult to solve, but the problem is con-
(O, p„o)z . (s', y'). (7)
mf 2 siderably simpler than the complete motion of two
nonidentical spins, which is described by nine coupled
The P&') are the unnormalized spherical harmonics
equations. "
Pp&2& =-'(3 cos28 —1), Considerable simplification results if nonsecular per-
turbations are neglected and the time-independent
P~~&" = ~ sin8 cosine+'& Hamiltonian is diagonal in the representation chosen.
P &') =3 sin'Oe+"& The latter situation occurs when the external field is
large, 3CO&&K~;„and only first-order perturbation
In this form all matrix elements in the (+0 —) repre- theory to the time-independent part of Kp'p is applied.
sentation can be written down at once, and the time- One obtains the result of Gutowski and Pake':
independent part is clearly separated. It consists of
the terms 222'=0 which do not depend. on p'. The time-
—4ypkpr 12—'(3 cos2fp —1),
Ey = yhHp
dependent matrix elements occur in the form Ep +2y'k'r12 —
'(3 cos'Pp —1),
.
ac. .(t)=P, K. .' H(pt), E = —ykHp —41y2kpr12 '(3 cospp —1).
which is used in the theories of Bloch and Redfield. Redfield' shows that application of second-order per-
The E& are Hermitian spin operators which do not turbation theory with the random time-dependent part
contain the time explicitly and the H'(t) depend only leads to the following equation for the density matrix:
on the time-dependent lattice coordinate. The Hp(t) can
be chosen as the four real functions cosp', sin&', 2 cosp'
and sin2tt '. With random time variations of p', introduce
= —2k(E- —E- )o- +Z R- pp opp
Bt pp'
the generalized correlation functions and spectral densi-
x(~, ~', p, p'=+, 0, —). (15)
ties of the functions Hp(t) by
Only secular perturbations will be retained. The secular
k„(~)=-', ~ p'-(H (t)H*"(t —r))dr. elements of the relaxation matrix, which satisfy the
condition
i
E Ep —
R p— E+Ep
p (16)
i

The correlation function is essentially zero for r«r„

e correlation time. Redfield has shown that for short are time-indePendent constants, given by
—
correlation times that is, for
R- =Z Dk«(~-
pp ~p)+k«—
(~- ~p))K-p'K—- p"
r.It;"(E —Ep)«1 —8 P p k (p1 p1 )K p'K-
where E —Ep
represents the diGerence between any
(~v ~-)Kpv'K-"7
&-pZvk-— (17)
pair of energy levels of the time-independent Hamil-
tonian, the equation of motion can be written in the provided the correlation time satisfies the condition
operator form
R pp r, «1.
—= —2k-'[X +Xg; '"' o7
Ba
If R pp r. 1, a transition fr'om time-dependent to
at time-independent character of the perturbation takes
—Q [KP, [KP', o —o &r&77k«. (0). (12) place. This case, which has been discussed extensively
in the literature, ' will not be pursued here.
"E. P. Wigner, Gruppentheorie (E. Vieweg, Braunschweig, "A. Zahlen and E. P. Gross, Bull. Am. Phys. Soc. Ser. II, 1,
1931), Chap. 15. 216 (1956),
 N. BLOEM BERGEN
The nonsecular elements which do not satisfy (16) ++~=+——~= —2E. 00~= —y ~ r
will be neglected. In general they give rise to very small (q)
(0 y, 0) p (q)
(0 )
oscillatory components in the density matrix element.
Unless there are partially or completely overlapping K q=K q= —y'5'r '2 &S (')(0$ 0)P &')(0')
resonance lines, the condition (16) is only satisfied for —K ——y&))tpy
q pq p
p2 kQ i (p)(p fp 0)p (&)(gt)
E Ep —E—
+Ep =0. . (19) (25)
K q= 'r -—
'y'h— 'n &')
(0 f 0)P &') (0')
This observation is important. It implies that diagonal
matrix elements of 0. relax independently of the values K ~q= —y'ttt'rip
-', 'X) q, (') (Opf p, p)P, (') (8').
of the off-diagonal elements. OG-diagonal elements
relax independently of the diagonal elements and oG- The P, (q)(t)') functions are given by Eqs. (8), if the
diagonal elements with another frequency. This state- exponential factor depending on ttt' is omitted.
ment justifies the procedure, generally adopted, to Substitution of (24) and (25) into (17) and sub-
treat longitudinal and transverse relaxation inde- sequently into (15), ignoring nonsecular terms with
pendently. Solomon's treatment' for the relaxation E E E—(t+E— tt &0, leads to the following set of
processes in HF is therefore justified, and corresponds relaxation equations:
exactly to solving Eqs. (15) and (17) for that case. = —(+wi+wq)
Here the general theory will be applied to a proton ~g++/~t (g++ — g++"')
pair. Assume first that the proton pair gives rise to a +wl(g pp g ()()(r)) +wq(tr g—(») (26a)
Bgpp/() t = wi(o++ —g++(r)) —2wi(o —
well-resolved doublet. This is the experimental criterion gpp(») pp
that the condition of secularity (16) is only satisfied for
E —E =Ep —Ep . Some simplification results in our +wi(o= — o (r)), (26b)
particular case, if complex functions Hq(t) = exp(Witt ') Bg /Bt=wp(g„~ o„+ ')+— Wi(opp opp )). —
and exp(&2+') are used. Two simple models for the
random motion of the tt ' coordinate will be investigated.
—(w&+wq) (o o— (r'), (26c)
(a) Random jumps between three equilibrium posi- Bop/(tt = (wp+ pw1+ pw2+&&o+p) g+pt (27a)
tions pp', tt)p'&120' occur at an average rate of (3r) '
transitions per second to either of the two adjacent Bgp /Bt= —(wp+ pwi+ pwq+z(t)p —)o'p —, (27b)
positions. The correlation functions are in this case ()g+ /Bt= —(+wi+wp+pp)+ )o+ t (28)
(Hq (t)H*"(t+ r)) = e "'t')„,
- (20) with

and the nonvanishing spectral densities are wp —(pP) Py4N'iq 'r {2 simp cosQp sin'8' cos'0'
+8 simp sin'8'), (29)
k„(&p) = r/(1+to'r'). (21)

(b) There are a very large number of equilibrium ~P'5 rip '(1 cosQp) —
(
positions and a stochastic diGusion process describes
wi sin'8'~
&1+ '&o — p'r'&
the motion in azimuthal angle. The probability to find —3 cosQp+1)
the pair at an angle tt)p'+t)It' at time t, when it was at tttp' + ,'py4htrip '—(4cosQp

for t=0, is given by the Gaussian distribution

&&sin'8' cos'8'~
( 2r

p(y', t) =
2(~tr ')'-
The correlation functions are in this case
exp( @'r/4t)dy'. —(22)
= same
9
expression with co+0
E1+~ r
replaced by oro,
)
(30)

w, = y'h, 'rip '(1+6 cos'Pp+cosQp)

(e+'&'("e+'&'"') =
p+00
p(p' t)e'p'dtjt'= e " (23a)
128
9
I+—
(e+pip'(t)eTpip'(p)) e 4tlt— (23b)
Xsintt)'I
& 1+ 'pp)+
—, 'r'&
y4It rip '(1 —

&(sin't)' cos'0'
32
cosgp)

( 2r
(31)
Again the spectral densities vanish for q &q'. ~

(1+&p+ 'r') ~.
kq, ; =~i r/(1+to' r'), — (24a) These formulas hold for the case of stochastic di6usion
&q, q =+q=-'r/(1+ 's~'r') — (24b) of the azimuthal angle. They can immediately be
changed to the case of random jumps between three
The nonvanishing spin matrix elements E p& are equilibrium positions by using Eq. (21) rather than
obtained from Eqs. (6), (7), (8) and (9). (24). The small difference between (o~p and top in the
 SP I N RELAXATION P ROCESSES
expression for m~ has been ignored. The 5)&') functions This corresponds to a partial application of the general
have been taken from Wigner. The normalized constants operator formulation, Eq. (12), which contains all
have been checked by direct trigonometric transfor- interference and nonsecular effects. Equations (35)
mation of the functions (8), which led to the results describe the frequency pulling and damping by cross-
about as quickly as the use of Wigner's general formula relaxation of off-diagonal elements which have nearly
in this case. the same frequency. The coupled system can readily
The set of Eqs. (26) describe the longitudinal relaxa- be solved for the two normal modes. This situation,
tion. There are two characteristic times, as could be which illustrates the principles and limitations involved
expected for a problem with effective spin 1. The in the distinction between secular and nonsecular per-
"magnetic moment relaxation"" is characterized by turbations is rather academic. The overlap of resonance

~( ++ — ) — 1
(~++ —o —)=0++Ti (32)
lines is the experimental counterpart of this theory of
cross-relaxation. In the case of a proton pair in a solid.
2fop+ tffi lattice the e6ect described by Eqs. (35) will be obscured
completely by the interaction from neighboring moments
The "quadrupole moment relaxation" is described by outside the pair.
ft(o+++o —2o pp) 1
(o'+++ 0' —20'pp) GI. SATURATION AND OVERHAUSER EFFECT
3'Ny Consider the case that two radio-frequency magnetic
o'+~+0 20'pp
fields are applied. Hi(v+p ) has a frequency close to the
(33)
resonant frequency v~p, H, (vp ') is close to the resonance
frequency vp of the other line. Assume that the am-
The intensity of the magnetic resonance absorption
—
lines is proportional to 0-++ C.pp ol Opp — and will
in general approach an equilibrium according to a
0, plitudes of both fields satisfy the relation

pkHi«h(v+p vp )«hv+p (36)

linear combination of two exponentials, a exp( t/Ti) —
+ffexp( —t/Di)+c, where the coefTicients a and b are The situation can be treated in the same manner as
determined by the initial conditions, and c by the the "weak external field" case, considered by Bloch. "
lattice temperature. The terminology "weak" refers to the condition (36).
The two lines have the same transverse relaxation It will be shown here that Bloch's considerations can
time according to Eq. (27), readily be generalized to include the case of more than
one applied frequency. The radio-frequency amplitudes
Tp =Wp+ pWi+ pWp. can each have saturating strength. Because of the
The 0.+ component decays with another characteristic assumptions made at the beginning of this section, the
time, but does not correspond to an observable quantity eGect of each frequency on all transitions but one that
in a magnetic resonance experiment. is near resonance, can be disregarded as nonsecular
The transverse and longitudinal relaxation times are perturbations.
in general not equal even in the case of very short cor- The circularly polarized fields add the following term
j.
relation times LEq. (11) They have a diferent angular to the Hamiltonian:
dependence on Pp in this nonisotropic case.
For 1 — 3cos'Pp=O, the two resonance lines will 5Crf p&MIi(vip') ((Ii++Ip+)e '"+"'
coincide according to Eq. (14). In this case perturbation +(Ii—+Ip )e+'" "+) +'yh H(ivp')
theory should be carried to a higher order to determine X ( (I,++Ip+)e *"o "+(I +I --)e+-'"o-' ) (37)-
the position of the energy levels, but the splitting will
certainly be very small. When the frequency of the two The equation of motion for the off-diagonal components
lines becomes nearly equal, the nonsecular nature of
of the density matrix in the stationary state with
the term Jt+pp becomes questionable. In other words, driving 6elds become
there is a term for which the left-hand side of Eq. (16) —$M+p o'+p= ffdypfryp +p/Tp-
0— '

is of the order of unity. A changeover from nonsecular

+ ', f'yHi(v~p')-42(opp o++), . (3—
8a)

Ep- op/Tp'
to secular character takes place. The term can be

''
=

op-
carried along in a rigorous manner provided the cor- ffop o'p ZMp o'p
relation time r, is short compared to k(E+ Ep — -opp)
+ ', iyHf(vp ')V2(o (38b)— .
+E ) '. A set of two coupled equations for the off-
diagonal elements 0+p and o=p should be solved These equations replace Eqs. (27). There are corre-
sponding complex conjugate equations for 0-p+ and 0. p.
a~+ p/at=icoypo'yp+~~po'~p+~pp The steady-state solution for the diagonal com-
(35)
80'p /Bt=zÃp op +Ep p op+Re +po+p. ponents is obtained by putting the left-hand side of
'P F. Lurcat, Compt. rend. 240, 2402, 2517 (l.955). '4 F. Bloch, reference 8, Sec. III.
 1546 N. BLOEM BERGEN
Eqs. (26) equal to zero and adding a term "
state Maser. In general, transitions between the +
— and —level 'could also occur. They can readily be
ss7H)(v+p')v2(o~ ~+p)
incorporated in the analysis.
to the right-hand side of (26a), The treatment presented here has to be modified if
one of the radio-frequency fields becomes larger than
,'i—
y—Hr (vp+')%2(o p a—
p) the intrinsic line width, yII~&T2 . A rigorous dis-
to (26c) and subtract the sum of these terms from the cussion has been given for the case than one radio-
right-hand side of (26b). Solution of the set of steady- frequency Geld is very large, while the other is relatively
state equations leads, by elimination of the off-diagonal weak. Transform to a rotating coordinate system to
elements, to eliminate the explicit time dependence of the strong
field. Red6eld" has discussed the transformation of the
dipolar interaction and relaxation terms in the rotating
system. The magnetic resonance response to the weak
field is split into components. Their spacing and relative
amplitudes depend on the value of the effective field

(1+Sp ) (o pp —o. ) = o.pp'r' —o. in the rotating system. ' '

For dipolar pairs in a solid lattice, an integration over
a distribution of local fields must be carried out. The
individual components will not be resolved. The
smoothed-out absorption will follow at least quali-
tatively the equations derived in this section, unless the
(r+ y+ «p+(r —— = 1. (39c) condition (36) is violated and the saturating Geld spans
The two saturation parameters are defined by both lines.

IV. DISCUSSION AND COMPARISON

S~p y Hr (v+p ) Tp7o] {(vip' v+p)'Ts'+—1) ', (40a) WITH EXPERIMENT
Sp =y Hr (vp )Tp'(or { — ) Ts +1j .
(-vp vp (40b) The ideal situation of an isolated spin pair with
The signal intensities of the two resonance lines are hindered rotation cannot be found in actual solids.
proportional to (o~~ — opp) and (opp —
There will be interaction with neighboring pairs or
o ), as can be other dipole moments in the crystalline lattice. This
seen from Eqs. (38). In the important special case that
interaction can cause transitions between the singlet
only one of the radio-frequency amplitudes is of satur-
ating strength, e.g. , So =0, m~S+0 — +~, one obtains for and the triplet states and its rigorous inclusion into the
the Overhauser effect, i.e., the factor fo by which the theory would be quite cumbersome. Its inQuence is most
line at vo is changed on application of a very strong prominent, however, on the observed line width, since
6eld at v+o, it produces a wide distribution of resonant frequencies.
'(or+ 2wp The observation of T2 calculated above is only possible
V
()— (41) if the effect of the static fields from neighboring pairs
'(or+'(gp is eliminated. This can be done in principle by applying
a radio-frequency 6eld which is strong enough to produce
The very good approximation has been made that complete saturation of an individual component, but
g (&) —g (&) 2(gpp(&) g P')) does not span both components of the pair of resonance
lines simultaneously. One moves into the center of reso-
The Overhauser effect depends on the orientation fp nance of one component in an adiabatic rapid passage
through Eqs. (25) and (26). There is no Overhauser variation of the external magnetic 6eld and then stops,
effect, if +2=0. The enhancement factor reaches a while leaving the radiofrequency 6eld on. The magneti-
maximum value 2 for wr=0. Equations (39) are, of zation will then decay to zero with a characteristic time
course, much more general than the special case, Eq. T2. Its signal may be picked up, for example, in a second
(41). They describe in conjunction with Eqs. (38) the crossed coil. The presence of the focusing radio-frequency
intensity of both signals and include completely all
field prevents loss of phase memory due to a distribution
saturation eGects on the application of two radio-
in static local fields. If the external magnetic 6eld is
frequency fields.
Similar considerations apply to any system with modulated with a frequency co = T~— ' in the presence
three unequally spaced levels, in particular to a nucleus
'
of a strong radio-frequency field, the phase shift of
with I=1 and quadrupole interaction, or a magnetic "N. Bloembergen, Phys. Rev. 104, 324 (1956).
ion S=i in a crystalline field. Khereas the case that A. G. Redfield, Phys. Rev. 98, 1787 (1955). The T2 in this
both 6elds are of saturating strength is not of particular paper corresponds to T~, in this reference, if a small contribution
of time-dependent interaction with outside neighbors is neglected.
interest for a proton pair, it is very important in the
analysis of the operation of a recently proposed solid- "'7 F. Bloch, reference 8, Sec. VII.
S. H. Antler and C. H. Townes, Phys. Rev. 100, 703 (1955)
 SPI N RELAXATION P ROCESSES
the observed signal with respect to the modulation will A detailed investigation of the angular dependence Po
also make a measurement of T2 possible. No such experi- of the saturation for a proton pair in a single crystal
mental data are available at present. rotating with 8'= ~/2, would provide a complete
Steady-state saturation e8ects should be modified analysis of the intra- and intermolecular contributions
according to Redfield's theory" to take into account to the relaxation time. The relaxation time should
the distribution of local fields. This is especially im- reach a maximum for $0 —0. For the same orientation.
portant for the dispersive, "in-phase" part of the radio- the intensity of the other line would become a maximum
frequency magnetization. Longitudinal relaxation times through the Overhauser eGect. The theoretical increase
can be measured by steady-state saturation experiments of a factor two will not be reached because other
or by direct observation of the recovery of magnetiza- neighbors will prevent z» from becoming exactly zero.
tion after saturation. The interaction with the other The detailed nature of motion around the preferred
spin of a proton pair, undergoing hindered rotation, axis is irrelevant for the angular dependence of the
may very well be the dominant factor for longitudinal relaxation mechanism. It could be a hindered rotation,
relaxation. Equations (30) and (31) show a marked torsional oscillations interrupted by tunneling, etc.
angular dependence. An important special case will The case 0'=0 is trivial. When the axis of rotation is
now be considered in more detail. parallel to the radius vector, there is essentially no
8'=m. /2, $0 —0. The pair rotates around a perpen- relative motion in the pair: zv~ — m2
—0. Relaxation can
dicular axis, which is parallel to the external magnetic only arise from interaction with outside neighbors.
field. In this case one has m& — 0, m2)0. Steady-state If the correlation time v, becomes long and the con-
saturation of each of the components should occur very dition (18) is invalidated, the splitting will not be given
readily according to Eqs. (39). In practice the field by Eq. (14). In the limit of the quasi-static case, ' where
intensity required for saturation will not be zero, as the correlation time is much longer than the inverse
relaxation by outside interactions will take over. The dipolar interaction, (3 cos'Po — 1) has to be replaced by
field required for saturation will, however, go through a 2 (3 cos'8 — 1). There will be a pair of resonance lines
minimum for $0 ——0. It follows from Eqs. (10) that this for each occurring value of 8. For $0 — 0 and 8'=~/2,
orientation corresponds to a maximum splitting of the one has 8=m./2.
two lines. In a polycrystalline sample, the tails will It is still possible that the slow angular variation
correspondingly saturate more readily than the center around the equilibrium 0, or the occasional jump
of the resonance absorption. This argument holds not between two equilibrium values of 0, provides the
only for a pair, but for any planar configuration of dominant relaxation mechanism. Then Eqs. (30) and
spins rotating around a perpendicular axis. For every (31) for the angular dependence of wi and m, still hold.
individual dipolar interaction has the angles 8'=~/2 The relaxation times would become long in view of the
and Po — 0, and consequently wi —0. The tails of the very large value of r. It is more probable that in the
case of slow motion, other relaxation mechanisms based
absorption by proton triangles in a polycrystalline
on interactions with other neighbors, take over. Then
specimen also correspond dominantly to the orientation
there would be no Overhauser eGect and the angular
$0=0. Richards" has observed preferential saturation dependence of saturation would be different. Experi-
of the tails. Saturation of the tails could also account ments on the proton resonance in a gypsum single
for too small values of the second moment of the absorp- crystal could confirm whether intra- or intermolecular
tion line reported by Andrew' and by Powles and
Gutowski. " relaxation is the more important.

ACKNOWLEDGMENTS
' R. E. Richards (private communication). The author is
indebted to Dr, Richards for discussions on the experimental The author is indeted to Professor E. M. Purcell and
situation, which led to this work. Dr. A. G. Redfield for reading the manuscript and
20
J. G. Powles and H. S. Gutowski, J. Chem. Phys. 21, 1695
(1953). making valuable suggestions for improvement.