1958 twoPhaseSystems Zimmerman Brittin
1958 twoPhaseSystems Zimmerman Brittin
1958 twoPhaseSystems Zimmerman Brittin
ZIMMERMAN
AND W. E. BRITTIN Vol. 61
Further evaluation studies of nuclear resonance relaxation phenomena of the hydrogen nuclei of HzO adsorbed on silica
gel observed by means of radio frequency pulse techniques are described. A stochastic theory for multiphase systems some-
times encountered in relaxation phenomena is presented. Special emphasis is directed toward a theoretical understanding
of two phase systems observed in the study of water vapor adsorbed on silica gel, Adsorption experiments are discussed
which directly demonstrate, for a s ecific x / m value the simultaneous existence of (a) two phase behavior for transverse
( T t )relaxation measurements and (i)
single hase behavior for longitudinal relaxation ( T I )data. Finally, a close estimate
is obtained of the life-time of a water molecuk in an adsorbing phase at low coverage on silica gel.
Since D is a finite, say n X n, matrix, e-Dt can be phase, and since all final phases are possible, it fol-
expressed as a polynomial in D. Hence lows that
n-1 ~3)C
i,i Iij(t)Pi (14)
e-DL E fmDrn (8)
m=O
The quantities Iij(t) satisfy equations which are
where fm are functions of t. If D has distinct eigen- similar to the Chapman-Kolmogoroff equations;
values XI,. .
. ,A,, then representations of D and i.e.
e-Dt with respect to the basis of eigenvectors of D
are given by
1 where
+ l/Ti) -
Slj (Cl Clplj
Flj
+ Slj/TI = + Elj
= Dlj Dlj (16)
respectively. Hence
e-Xnt =
n-1
fmhkm (94
60 = li]
r:1
and 40 = [PI,
Pa, . . ., P.] (18)
m=0
then the average value, 1 3 , may be simply ex-
If one sets pressed as
1 Hence
I(t) = &I
&o = Q d 3
is used to signify that (40
(19)
(20)
and D = [&] = [(Xi)-l]a-l, i,k = 1, . . . . ,n, then &) form a conjugate pair of eigenvectors of D.
(9a) may be expressed as the matrix equation I n fact, it is easily verified thats
e ( t ) = slf (9b)
Ds0 = 0 and &D = 0 (21)
which has the solution
Asymptotic Expressions for Very Slow and Very
f = a-1 e(l) (10) Rapid Exchange. (a) Very Slow Exchange.-For
Theory of Relaxation.-If a spin is in the ith the case of a very slow exchange, the relaxation is
phase, certain quantities, &: associated with the relatively rapid; and (71, which is the reciprocal,
relaxation process decay at a rate lifetime for the Ith phase, is considered to be very
3
dt
= -Q/Ti
L
Q ( t ) = Qoe-Jotdt/T(t)
where T ( t )is a random variabletaking on thevalues,
(12)
Hence e-t’Tnl i (22)
J
n
T i , as the system6jumps from phase to phase. The
relaxation process is completely described by the Z(t) N PleWTi
i-1
average value of the integral
Equation 22 shows that, for very slow exchange
Z ( 1 ) = e-fil/T(t) (13) the relaxation is the weighted average of effects
taking place separately in each phase.
The method of Kubo’ is very convenient for find- (b) Very Rapid Exchange.-For this case, CI,
ing the average value,-), of I@). the reciprocal lifetime for the Ith phase is very
Let Iij(t) be the average value of I(t) on the con- (8) The result that &D = 0 follows from the fact that, indepenient
dition that the system is initially in the ith phase and
at time, t, is in the jth phase. Since P i is the prob-
(5) -
ability that the system be found initially in the ith
For TYphenomena, Q MI:for TIphenomena, Q = M I - M I ,
of the value i. lim
t 4m
Pij(t) = Pj. Therefore, Iim [Pil(t)] =
1 4m
L& J
(6)
nique
(7)
Here “system” refera to a single spin, and the averaging tech-
gives results for the entire ensemble or multiphase system.
R. Kubo, J . Phryr. Boa. Japan, S, 939 (1954).
The result that D+p = 0 follows directly from X D i j - 0.
1330 J. R. ZIMMERMAN
AND W. E. BRITTIN Vol. 61
large compared to T I , the relaxation time associ- Equation 29 shows that for very rapid exchange the
ated with the Ith phase; i.e., CI >> TI. At first, relaxation is described by a single relaxation time,
one might be tempted to argue that for this ex- whose reciprocal is a weighted average of the re-
change rate limit, D >> E, and hence that ciprocals of the various relaxation times; i.e.
- n
~ ( t ~) Z O e - D t d o = 1 (23)
(1/T)sv = PilTi (30)
However, equation 23 neglects relaxation com- i=l
pletely and is not the desired result. I n order to Stochastic Theory of Relaxation Times for Two-
obtain a correct limiting result, first-order pertur- phase Systems.-The Chapman-Kolmogoroff equa-
bation theory is used to calculate an approxima- tion dP/dt = -PD are particularly simple for a
tion for e-". The perturbation theory presented two-phase system. Since plz = 1
is merely a generalization of the customary per-
turbation theory employed in atomic physics.
Let (+k, G k ) be the normalized eigenvectors of
D; i.e.9
=
c1
[-cz
D 2 = (C, + C2)D I
1 (31)
and
1* k
(26) with X = C1 + Cz. The matrix F is given by
klk = hk f gk-@dk
The matrix, 8, which diagonalizes F is, to first or-
der, given by The eigenvalues (p1, p z ) of F are the solutions of
det [F - pl] = 0. These are given by
8= [$a, $1,
I>) = ias
or
the result becomes
where
a1 -- PZ
1
- P1 IP2 - PIP1 - pzPz] = PI
and
a2 = -PP
Theref ore I I
+ Pze-t/Ta
I T ) = Ple-t/h (42)
For very rapid exchange, c k >> 1 / T k ,
2~ - (CI + Cz ((CI
+ TI + 1/Td =t
+ Cz)' - 2(Cz - Cl)(l/Tz - l/T')}'/z
and
E 200-
E
z
2
Therefore T
0
I(t) = e-t/Tav y loo-
with & 90-
67 80-
i- Pz/Tz
l/T&v= P J T , 70 -
The special two-phase results expressed by (42) 60-
and (45) are in complete agreement with the gen- 50 -
eral results of equations 22 and 30.
Experimental 40-
b
In the preceding theory no distinction was made between T1 30-
and Tz phenomena, but rather it was pointed out that the
theory should apply equally well to both T I and T zmeasure-
ments. For transverse relaxation the quantities &(t) and 201 I I I I I 1 1 1 1 1
Qo to be associated with nuclear resonance relaxation proc- 0 02 04 06 08 IO 12 14 1.6 I8 20
eases are given by
and
-
Q(t) = M I
&o = MI0
and
r1103- SECDNDS.
MI1 = Mo (1 - 2
i=l
Pie-VTii ) (54)
Fig, 3.-Typical transverse relaxation time plot for If free decay signals which immediately follow each of two
HzO adsorbed on silica gel illustrating single-phase be- successive r-f pulses are used to determine T I values, then
havior. the experimental amplitude expression can be conveniently
written as
R ( T ) = (ZPis-dTii)-1
where (55)
R(T) -
Mo/(Mo Mli)
and where T is the time between r-f pulses. A theoretical
plot of log R ( T )va. T for a longitudinal relaxation two-phase
system is shown in Fig. 4. I n respects similar to two- hase
T Z data described earlier, both the Ti’s and the refative
population ratio, Pi/Pj, are available in principle from such
a plot.
i (b) Very Fast Exchange.-Again from equation 45, the
longitudinal moment decay expression is of the form
-
MI1 = Mo(1 e-t/T1)
where
Fig. 4.-Theoretical
T-
longitudinal relaxation plots for two- 1/Ti P d T u + P2/T12 (56)
phase systems. The corresponding experimental expression for determining
T Ivalues by free decay signals is simply
R(T)= M,/(Mo - Mil) = e++/Ti (57)
A plot of log R( T ) vs. T would, of course, yield a straight lime
curve. Figure 5 illustrates such effective single phase be-
havior for two separate x / m values of water vapor adsorbed
on silica gel.
Lifetime of a Molecule in an Adsorbing Phase on Silica
Gel.-The average lifetime of a spin in a given phase char-
acterized by a single relaxation time Ti is simplyali = 1/Ci.
The problem to be answered is whether a limiting value
might be placed on the lifetime of a water molecule in an ad-
sorbing phase on silica gel. Two assumptions enter into
an such lifetime estimate: (a) The lifetime of a water mole-
C U in ~ an adsorbing phase is to be identified m t h the Me-
time of a hydrogen nuclear spin in a given relaxation phase.
(b) If a two-phase system exists for transverse ( T d relaxa-
tion phenomena, then a corresponding two-phase system
0 20 40 60 80 100 I20 140 160 180 200 exists for longitudinal ( T I )relaxation phenomena.
Assumption (b) might be more clearly stated as follows.
f‘ lo3 SECONDS. If a two-phase system exists for transverse relaxation, then
Fig. 5.-Typical longitudinal relaxation time plots for there are two distinct average nuclear correlation times,
HzO adsorbed on silica gel at low coverages illustrating 7111 and.Twj, associated with the relaxation processes. These
single-phase behavior. correlation times will then be expected to give rise to not
only two transverse relaxation times but also two longitu-
The spin-echo amplitude signal is given by dinal relaxation times. The existence of such multiple
phase systems is no guarantee that experimental measure-
A(%) = A,e-WTS (51) ments will resolve multiple Tsi’s and T ~ i ’ s . I n other words,
where T is the time between r-f pulses. A plot of log 4% if the average lifetime, 1/Ci, of the molecule in the ith phase
va. T would, of course, be a simple straight line. It should is very short compared with the time required t o obtain an
be emphasized that, regardless of the number of p b e e in- experimental measurement, averaging processes between
Oct., 1957 NUCLEAR
MAGNETIC
RESONANCE PHASESYSTEMS
IN MULTIPLE 1333
0.20
I
of x/m = 0.0434.
-
within an order of approximation, 3 X IO-' sec.<<1<<29
X lo-' sec. or t 3 X lo-' sec. for the particular coverage
single phase behavior for longitudinal (TI) relaxa-
tion measurements. From a combination of this
experimental data and theory of stochastic proc-
esses in relaxation phenomena, the lifetime of a
Conclusion water molecule in an adsorbing phase on silica gel
A stochastic theory of relaxation times for mul- is estimated to be t N 3 X second. One as-
tiple phase systems has been developed which ac- sumption in the evaluation of this lifetime should
counts explicitly for the essential features of very not be overlooked-namely, if the time required to
SIOW (Ci << 1/Ti) and very fast (Ci >> I/Ti) obtain relaxation data could be made sufficiently
exchange; i.e., for very slow exchange the relaxa- short for both TI and Tz measurements, such that
tion is the weighted average of effects taking place the time of measurements would be much less than
separately in each phase; and for very fast ex- the lifetime of the molecules in the adsorbing
change, the relaxation is described by a single re- phases, then both transverse and longitudinal re-
laxation time, whose reciprocal is a weighted laxation data would describe two-phase systems.
average of the reciprocals of the various relaxation Investigations are underway which do verify this
times. assumption for specific systems; these will be re-
From the stochastic multiple phase theory, de- ported on a t a later time.
cay expressions are obtained for the special case of
two-phase systems. This theory describes the es- DISCUSSION
sential features of the relaxation processes en- J. J. BIKERMAN.--IS
it possible to compare the relaxa-
countered in adsorption studies not only for the tion times found for the hydrogen nuclei in the "adsorbed
asymptotic conditions of slow and fast exchange but state" with their relaxation times in, say, water and in a
i.e., for Ci -
also for the overlapping region of exchange rates;
1/Ti. It is clear from the decay ex-
pressions for a two-phase system that the relaxa-
molecule such as Si(OH)d?
J. R. ZIMMERMAN.-Certainly qualitative comparisons
are possible. For example, the transverse relaxation time
of water at room temperature is a few seconds; the relaxa-
tion times, Ti and Tj,which are evaluated from the tion times of adsorbed water in these experiments are of
data, are not necessarily the relaxation times cor- the order of milliseconds. For ice the transverse relaxation
responding to the asymptotic conditions of ex- time of the hydrogen is of the order of microseconds. How-
change (see expression 40). ever, for quantitative comparisons, the simplified BPP
model which relates a single nuclear correlation time to
Adsorption studies of water vapor on silica gel the relaxation processes is generally not an adequate model
by means of nuclear magnetic resonance tech- for adsorption phenomena.