474 BLOCH, HANSEK, AND PACKARD

(5) As in comparing the moments of neutron, There are unquestionably more problems which
proton, and deuteron, nuclear induction can well will become tangible in further deyelopment of
be developed as a simple and practical method to the new electromagnetic effects. The fact that
calibrate and measure high magnetic fields with they are simple to obtain and require only
great accuracy, and t o apply it, for example, very modest equipment should make it pos-
in the construction of cyclotrons and mass sible for many investigators to enter this field
spectrographs. of research.

PHYSICAL REVIEW VOLlJME 70. SUMBERS 7 .IND 8 OCTOBER 1 A N D IS. 1946

The Nuclear Induction Experiment

F. BLOCH,W. LV. HANSEX,AXD M. PACKARD
Stanford 0-niniversity, California
(Received July 19, 1946)

The phenomenon of nuclear induction has been studied experimentally. The apparatus
used is described, both as to principle and detail. Experiments have been carried out in which
the signals from protons contained in a variety of substances were observed. The results show
the role played by the relaxation time, which was found to vary between about 10-6 second
and many seconds.

INTRODUCTION we can expect the polarization vector to deviate

appreciably from the z direction if the z field Ilo
IN this paper we aim to describe certain experi-
ments designed
- to study the phenomenon of
approaches the resonance value H* given by
nuclear induction, the theory of which is de- I Y /H*=wI (3)
scribed in the preceding paper. While it will be
where y, the gyromagnetic ratio, is the ratio of
assumed that this paper, hereafter referred to as
nuclear moment to angular momentum charac-
I , is familiar to the reader, we will not hesitate to
teristic of the nuclei under consideration. Quanti-
repeat those few formulae which are of immediate
tatively, the result of this deviation may be
importance for the present paper. We start, then,
described by a macroscopic polarization vector M
by giving the reasons for the selection of the
with components given by Eq. (44) of I,
general methods used and follow this by a brief
description of specific apparatus. Finally, the M
results are described in detail. M, = ------ cos ot,
(1+62)t
METHOD M
If matter be placed in a magnetic field Ho,in
the z direction, the nuclear magnetic moments
I&= F-
(1 +a2It sin wt, (4b)

will tend t o orient themselves parallel to the field. ill

1LI * = -
In matter of normal density, thermal equilibrium +
(1 62)11 (44
may be established, in which case there will be a where
paramagnetic polarization in the z direction 6 = (Ho -H*)/Hi
111, = xHo. is the deviation of z field HOfrom its resonance
If now we superimpose an oscillating magnetic value H* in units of the half-amplitude H1of the
field in the x direction oscillating x field. The ambiguous sign in (4b) is
t o allow for the possibility of either positive or
Hz= 2H1 cos w t , (2) negative values of y. The quantity M depends in
 SUCLEAR INDUCTION EXPERIMENT

a rather involved manner on the nuclear relaxa-

tion times and we will only say that, under
favorable conditions, it may be expected t o be of
the order of the equilibrium polarization Mo; a
precise expression is given in I , Eq. (46).
I t is this precessing vector M which we wish
to observe and either to design or discuss a TER

suitable experimental arrangement for this pur-

pose N e need some estimate both of the magnitude
of M and of the field I11 needed to produce the
precession of M. These quantities have the same
dimensions and so may be compared directly.
For this purpose we may use numbers taken
from the experiments as actually performed, a
I RECEIVER

typical case being the use of protons in water and

FIG. 1. Schematic diagram of coil arrangement used.
with a field IIo= 1826 gauss. With known values section in the yz plane shows the spherical sample
of p and j one finds a saturation magnetic mo- (cross hatched) surrounded by a receiver coil sensitive to
variations in flux in the y direction and outside this a
ment density of 0.6X gauss. If now we con- transmitter coil producing flux in the x direction. '4 static
sider a spherical sample with given moment field in the z direction is produced by a magnet not show11.
density, we easily find that the flux linkage
through a11 equatorial circle is equal to that shaded circle representing the equatorial section
caused by a magnetic field of magnitude (8ri3) of a spherical sample of volume Q containing the
times the moment density. Or, in the present nuclei whose total moment Mfl is to be observed.
case, we can say that the flux linkage through a A static magnetic field Ho in the z direction is
single equatorial turn about the sphere will be the produced by an electromagnet (not shown). An
same as that caused by a uniform field of oscillating field of peak value 2H1 in the x
5.1 X gauss. direction is produced by current flowing through
In considering possible methods of detecting an a "transmitter coil." This coil, though shown as
oscillating field of this general magnitude, we a single turn, in actuality consists of a plurality of
must not only consider the relative magnitudes turns. In consequence of the oscillating field 2H1
of thermal noise power and the power that may and the static field Ho, the moment MQ will
be generated by such an oscillating field, but also precess: as a result a variable flux links the "re-
must compare the magnitude of the field caused ceiver coil," whose axis is in the y direction, and
by the nuclear moments with the field H I . the resultant voltage which appears a t the term-
As remarked in I , the field H I should be notably inals is led off t o a receiver to be measured.
larger than the internuclear fields, which are of We now return to the problem of the detection
order one gauss. Thus, a value of five gauss for H I of a field of magnitude 5 X loF6 gauss oscillation
will perhaps be suitable. This specifies the magni- a t approximately the Larmor frequency, which
tude. As to the phase, we note that, by (2) and turns out to be 7.76X lo6 cyclesjsec., for protons
(4), the driving field, which is in the x direction, in a field of 1826 gauss. We may consider the
is in time quadrature with t h e y component of the problem in two stages: first, how would one
moment to be observed. measure such a field in the absence of the driving
Broadly speaking, then, the problem is the field, and second, what effect does the presence of
detection and measurement of an oscillating field the driving field have.
of, say, 5 X gauss or less in the presence of a -4s to the first, the obvious procedure is to
field of the same frequency and of magnitude amplify the voltage induced in the pick-up coil
about 10 gauss, or a t least 2 X lo6 times larger. by a suitable amount, whereupon the voltage
T o accomplish this we have used an apparatus may be measured by various means-provided it
which may be described with the aid of Fig. 1. is larger than any spurious voltages that may be
This shows a section in the yz plane, the center present. Practically, it is desirable to increase the
47 6 BLOCH, HANSEN. AND PACK:\RD

voltageasmuch as possible by using a multip1icit~- quite limited. For example, if one attains a Q of
of turns and by introducing a shunt condenser to 100, which is quite good for such small coils in the
induce resonance. In principle, such resonance presence of poor dielectric, the band width a t 7
does not alter the ratio of signal voltage to mc is 1.1 X lo5 c~cles/sec.,or much larger than
thermal noise voltage, but in practice it does alter the minimum set by the requirement that the
the ratio of signal to observed noise, because both response time be not inconveniently large. While
signal and thermal noise are raised relative to the somewhat larger Q values may be possible, it is
constant amplifier or tube noise. plain that no attainable value will make the band
The apparatus we used had 24 turns on a form as small as one might desire.
of diameter d = 1 . 7 4 cm and length h=0.84 cm, Some improvement might be obtained by
and had a Q, with the sample in place, of about making the modulation cyclic, and following the
81). The r.m.s. voltage available is then rectifying device which converts the 7-mc signal
to a unidirectional one with a filter, or amplifier,
sharply tuned to the modulation frequency. The
effectiveness of this scheme may be measured by
the effective band width, which is of the order
and, with a sample of volume 1 . 4 6 cc this comes ( A V ~ A Vwhere
~ ) ~ Avl and Ava are the band widths
out to be 3.7 X volt, so that large amounts of of the circuits before and after the rectifier. Thus
amplification are not needed. Also, this may be reducing Avz reduces the noise, but not as much
compared with the thermal noise voltage which, as one might hope, since the response time goes
for this coil, is about 0.7 X10-6 volt in a 3000- with l / A v z .
cqcle-per-sec, band. Thus, an enormous ratio of The proper way to reduce the band width is by
signal over noise is potentially available. This heterodyning the signal down to some lower
ratio is independenrof the number of turns on the frequency where a given percentage band width
coil, but could be increased for the same sample becomes a small absolute band width. This may
and coil size by increasing Q or decreasing the be clone with a detector, and following filter with
hand width. narrow pass band, if one introduces into the
' r ~ i ~further
o points nus st be noted, however. detector an additional signal of constant fre-
First, in order to know that the observed quency, usually called the local oscillator signal,
amplifier output is caused by the nuclear mo- and with magnitude large compared to either
ments and not by some unknown cause, it is signal or noise. Then the beats between local
essential that the voltage due to the nuclear oscillator and signal or noise components domi-
moments be varied, or modulated, in some known nate those between noise components, or between
and controllable manner in which case outputs are signal and noise components, and the output
real or spurious according as they do or do not noise then depends only on AvZ,being quite inde-
correspond to the known modulation. pendent of Avl. In this way the band width may
Second, the magnitude of any spurious signals be reduced by any desired amount, the only limit
or "noise" will depend on the frequency band being set by the response time 1,'Avz.
accepted by the resonant pick-up coil and the We come now to the question of the effects of
amplifier which follows. If the noise is of random the large field H I which, it will be remembered, is
type, its amplitude will vary with the square root a t least l o 6 larger than the field to be detected.
of the band width; if it is of impulse type, the This large background has two effects. First, if
amplitude will vary directly with band width. In even a moderate fraction appears in the receiver,
any case, it is desirable to reduce the band width any amplifiers that may be used will saturate long
as much as possible or convenient, for doing so before the signal is of suitable size. Second, and
reduces noise while leaving the signal unaltered. more fundamental, any fluctuation in the back-
The straightforward method of reducing the ground is indistinguishable from a signal and,
band width is, of course, t o increase the sharpness with such a large ratio of background to signal,
of resonance by reducing the losses as far as may such variations can be serious, though they could
be. Possibilities in this direction are, however, be, if necessary, controlled.
 NUCLEAR INDUCTION EXPERIMENT 47 7

evenly spaced (phase modulation) has a practi-

cally constant amplitude, and so, on passage
through an ordinary detector, the modulation
will not be reproduced. One must, therefore, shift
the phase of the leakage by some amount, prefer-
ably n/2, whereupon the variable signal will lead
to amplitude modulation which is easily de-
tected. The above ideas may also be expressed by
means of vector diagrams, in which representa-
tion the leakage is a large vector and the signal a
small vector a t right angles thereto, so that
variations in the signal make variations in the
vector sum which are quadratic in the signal. If,
on the other hand, the phase of the leakage is
shifted b ~ ra, 2 so that the vectors are collinear,
variations of signal amplitude appear linearly as
variations in resultant.
FIG.2. 'The d r a ~ i n g(.I)shows the flux threading the This desired phase shift may be achieved by a
transmitter coil, while (B) and (C) show how this fluxmay device which also serves the purpose of regulating
be "steered" by means of a semicircular copper paddle,
so as t o link the receiver coil in either the positive or the the magnitude of the leakage. More precisely, the
negative sense. device to be described regulates the quadrature
component of the leakage: when this is reduced
For these two reasons, we have used a receiver to zero, what remains must be in phase with the
coil which is nearly a t right angles to the trans- signal.
mitter coil, so reducing the effect of HI, which we Referring now to Fig. 2A, there is shown, in
call leakage, by a factor sin e z s , where e is the cross section, the transmitter coil, the receiver
deviation of the angle between transmitter and coil and the sample being omitted. The flux
receiver coils from ~ / 2 In
. this way the leakage through the coil will be roughly parallel to the
of 111 into the receiver coil may easily be reduced axis or, more exactly, will have symmetry about
by a factor of lo4or thereabouts. But a reduction the axis and will not link any coil whose plane
of lo6or more, while possible, would be difficult, passes through this axis. If now we place a
as it rvould involve adjustment of an angle to semicircular sheet of conductor across one end of
radian, or 0.2". the coil as in Fig. 2B, the currents induced therein
In fact, it is not desirable to try for such ex- will prevent the flux from passing through the
treme reduction, since once the leakage has been sheet. If the lines are going from left to right they
reduced to a point where it is only, say, LO3 times would then pass through a plane, a t right angles
the signal, it may actually be made to serve an to the paper and containing the axis, in the
important and useful purpose. upward direction. Likewise, if this sheet of con-
Specifically, we have alreadJ- pointed out the ductor, which we call a paddle, be placed as in
desirability, from a band width point of view, of Fig. 2C the linkage of a coil in the axial plane will
heterodyning to some lower frequency, and there be reversed. The change from Fig. 2B to Fig. 2C
is no reason for not choosing this frequency to be may be made continuously by mounting the
zero, in which case the leakage signal, slightly paddle on an axial shaft and rotating same.
modified, is an ideal source of local oscillator Thus, by rotating such a shaft either positive or
voltage. negative flux linkage may be obtained and, by
The modification needed is in the phase for, as continuity, the desired value of zero must be
we have seen, the leakage and the signal are ?r/2 possible. Any desired sensitivity of adjustment
out of phase so that, if one adds the leakage to an may be had by suitable choice of paddle size.
amount of signal which is varied by the modula- In this way the quadrature component may be
tion the result, while having nodes that are not adjusted a t will. One might fear that it would be
478 BLOCH. HANSEN, AND PACKARD

found that there would be no in-phase component amplitude a t a 60 c/sec. rate because of the
of leakage. Actually this is not found t o be true, variations in field Ho, are amplified along with
the reason being that the currents which flow in stronger constant amplitude leakage signals. The
the paddle are not a out of phase with the leakage and the varying voltages due to pre-
currents in the transmitter coil, because of losses cessing nuclei are mixed in the detector, the
in the paddle, and so have a component in phase output of which then contains pulsating unidi-
with the signal. rectional current, the steady component due to
I t remains only to describe the methods of leakage and the variations corresponding to the
modulation and presentation. desired signal. The d.c. is removed by the
T o cause the signal to vary in strength one blocking condensers in the following amplifier
might vary either the z magnetic field Ho or the which increases the signal voltages to a magni-
frequency. The former seemed the easier and was tude suitable for operation of the cathode-ray
adopted for the experiments here described, the tube.
field Ho being varied cyclically about a value Hde. APPARATUS
The modulation frequency of 60 cycles/sec. was
chosen as being low enough to avoid excessive We now describe in detail the equipment
band width and high enough for the persistence constituting the various blocks in the above
of vision needed for the cathode-ray tube presen- diagram.
tation used. Within limits set by these factors a The steady magnetic field Hdc is supplied by a
considerable range of frequencies is possible: of lecture-demonstration magnet which easily pro-
these 60 cyclesisec. has the additional advantage duces the necessary fields of a few thousand gauss.
that any spurious hum pick-up appears as a The poles are 3 inches in diameter and these are
recognizable distortion of the final pattern rather provided with laminated extensions of trans-
than as a blurring. former sheet which are 34 inchesX3+ inches and
Presentation was on a cathode-ray tube with 1 inch thick. The spacing between these lami-
horizontal plates giving a deflection proportional nated extensions is 1; inches and within the
to the variation in Ho and the vertical plates central volume of 2 cc, occupied by the sample,
giving a deflection proportional to the signal the field was found, by flip coil measurements, to
which in turn depends on the y component of the be uniform to within 2 gauss. Since variations in
nuclear magnetic moment. Thus one observes on field of order H1=5 gauss could be tolerated
the screen a plot of M , vs. Ho, the center of the without undue broadening of the resonance peak,
screen corresponding to the field Ha,. this uniformity is satisfactory.
We may summarize by giving, in Fig. 3, a The 60 c/sec. modulation of the field is pro-
block diagram of the complete apparatus. Here vided by current flowing through two coils of
the spherical sample is immersed in a field about Helmholtz proportions, placed between the
Hd,+H,, cos 377t produced by a magnet, not laminated pole faces and carrying 60 c/sec.
shown. The nuclear moments oriented by this current. The magnitude of this current can be
field are caused to precess by a driving field II, regulated by means of a variac to give values of
~xoduccdby current from the box labeled trans- Ha,between zero and 50 gauss. The same current
mitter. 'l'he precessing moments induce voltages which passes through the coils passes also
i t 1 the rcceiver roil. These voltages, which vary in through a resistor, and the resulting voltage drop
actuates the horizontal sweep of the oscilloscope.
As to the transmitter, the main requirement in
addition to ability to deliver enough power to
maintain the field H I , is a very high order of
output amplitude stability. This requirement
arises because variations in leakage are indis-
tinguishable from signal and because even the
smallest leakage is large compared to the signal.
FIG.3. Block diagram of apparatus. For example, if the nuclear moment per unit
 NUCLEAR INDUCTION EXPERIMENT

- I INCH

FIG.4. Scale drawing of an xy section of the r-f head. The spherical sample S is
surrounded by a receiver coil R, which is in turn surrounded by a transmitter
coil T, the whole being encased in a shield. A rotably mounted paddle P is used
to steer the transmitter flux. Leads to the receiver coil are the coaxial leads LlLz
while the transmitter leads are L3 and L4. The outer shield is split to avoid 60-cycle
eddy currents.

volume is lo-' times HI, as it might be in a Transmission lines a few inches long lead to a
typical case, then even if the leakage is reduced brass box about 4 inches by 8 inches by 8+inches,
to, say, of H I , a variation in transmitter which is divided in two by a horizontal partition.
output by a part in lo4 will cause a fluctuation The lower portion contains tuning condensers to
equal to the signal. resonate the transmitter coil ; the upper contains
The power required to maintain the field HI is a similar condenser for the receiver and also the
of the order 1 watt and this, together with con- detector and first audio stage of the receiver,
siderable excess to allow for imperfect impedance together with suitable batteries for same.
matching, etc., is easily supplied by an 815. The The receiver differs from the block diagram in
stability is achieved by deriving both filament that the r-f amplifier there shown is omitted, the
and plate power from batteries. Magnetic regu- receiver coil working directly into a diode de-
lators in the primary of an ordinary supply were tector. This has the disadvantage of a very con-
tried and found worthless because of slow re- siderable loss in potential signal/noise ratio, but
sponse. A carefully engineered and tested vacuum- in view of the large signal available from protons
tube regulator might do, but batteries were it was decided to sacrifice this in the interests of
adopted as quicker and absolutely reliable. simplicity. This simplicity appears in two ways:
The transmitter and receiver coils are best First, one replaces an r-f amplifier by a simpler
described by the drawing of Fig. 4, which shows audio stage, and second, whereas a volt or less
the coils and surrounding shield which we may might saturate an r-f stage, the diode will not
call the r-f head. The transmitter coil consists of saturate. This point is very important in tuning
seven turns of mean diameter 5" and length i", up, for a t the start the leakage voltage is usually
wound inside a threaded polystyrene shell. Inside 100 volts, or more, and it is only after con-
the transmitter coil is placed the receiver coil of siderable adjustment that this can be reduced to
24 turns, A" in diameter and A'' long, wound on about 0.1 volt which is usually used in operation.
the outside of a thin threaded polystyrene form. The diode detector is conventional as is the first
Inside this, in turn, is the sample, contained in a audio stage which acts mainly as an inverter and
glass bulb in. in diameter. The design aims a t impedance changer, giving push-pull output a t
the greatest mechanical rigidity and damping low impedance level. This low impedance output
possible, in an effort to avoid vibrations which, is carried to the main amplifier by a few feet of
by varying the coupling between transmitter and cable.
receiver, vary the leakage and so introduce The requirements on this amplifier are that it
microphonics. Such microphonics constitute the have a gain of a few thousand and pass all
biggest experimental difficulty with the equip- significant signal components without appreci-
ment here described. able amplitude or phase distortion. The fre-
480 BLOCH, HANSEN, AND PACKARD

low frequency disturbances obtained by raising

the low frequency cut-off is often a convenience.
Finally, the output of this amplifier actuates
the vertical sweep of a Dumont 208 oscillograph
and thereby traces a curve of M , vs. Ho suitable
for visual observation.

RESULTS
Using the method and apparatus described
I iI l i above, we have so far investigated nuclear induc-
1.
-A, --
1' t
. tion of protons only. The advantage of working
with protons is twofold. In the first place they
FIG.5. Schematic representation of the modulated corn- have of all known nuclei the highest gyromagnetic
ponent Ha of the magnetic field in its dependence upon ratio y so that moderate fields are sufficient to
time. As the magnet current decreases, the resonance field
H* is first reached a t the initial time t, and last at the bring theresonance frequency into the convenient
final time 1,. During the interval At a positive signal is frequency range of a few megacycles. In the
observed to travel across the oscillograph screen as indi-
cated by the arrows on the traces, sketched for three second place there is a great choice of conlpounds
different times a t the beginning, the middle, and the and solutions containing hydrogen which facili-
end of At. tates variation of the conditions of observation,
particularly with respect to the relaxation times.
quencies involved are the modulation frequency Our first experiments were carried out with
of 60 c.p.s. and various harmonics thereof. With water contained in small spherical glass bulbs
2H1 of, say, 10 gauss and A H of 100 gauss, the which could be sealed off. The very first signal
first ten harmonics will be of about equal was observed with a sample of only 100 milli-
strength and from there on up the strengths will grams, but in all later experiments the amount
decrease rapidly. About fifty harmonics should was increased to about 1 gram, the glass bulbs
then be ample and the amplifier must then pass being increased to a volume of about 1 cc. These
frequencies between 60 and 3000 cycles/sec. samples were placed into the receiver coil and,
Actually the amplitude response must be practi- after mounting the shields and the a.c. coils,
cally flat to a considerably lower frequency, say, providing the field modulation, were moved into
6 cycles/sec., if phase shiftsin the lower harmonics the center of the magnet gap. With proper tuning
are not to cause undesirable pedestal effects. and adjustment of the apparatus we could then
These requirements were met by a three-stage observe the proton signal on the oscillograph
amplifier, using push-pull twin triodes (7F7) screen as the current in the magnet was set to a
throughout. The use of push-pull amplification
eliminates cathode bypass condensers and, more
important, cancels the otherwise very annoying n All
disturbances due to power supply variation.
Likewise, screen by-passes are not needed and the
h4iller effect is not bothersome because of the
rather low value of high frequency cut-off. The
amplifier was tested on 60 c/sec. square waves,
which it passes without distortion.
Although frequency response, as specified
above, is needed to give a completely undistorted
pattern on the oscilloscope, it is found in practice
that the amount of pedestal effect introduced by
raising the low frequency cut-off to 60 or 120 FIG. 6. Time dependence of HOas the magnet current
increases. A negative signal travels across the screen a s
cycles is not objectionable, and the reduction in indicated by the arrows on the three sketched traces.
 NUCLEAR INDUCTION EXPERIMENT 48 1

value such that the resonance field was contained

within the limits of variation determined by the
amplitude of the modulating field.
The behavior of the observed signal is de-
termined by various factors, particularly the
ratio of the relaxation time to three other time
constants which enter into our experiment. The
smallest is the time taken to pass through the
resonance width which is usually a small fraction
of 1/120 sec. The next time of importance is the
time between successive transits through reso-
nance, of the order of 1/120 sec. Finally there is
the time during which Hdc is varied by an amount FIG. 8. Photographic record of the proton signal from
comparable to Ha,; this time can be chosen a t water. The three traces from top to bottom correspond to
will to be anywhere between the time constant of the situation a, b, c of Fig. 7.
the magnet of the order of one second and many
minutes. The following experiments cover various here observed to appear on the right side of the
orders of magnitude of the relaxation time rela- oscillogram a t the initial time ti, travel across the
tive to the above-mentioned times. screen during the time At, and to disappear on the
The easiest and most striking visual observa- left side a t the final time tf as indicated on Fig. 5.
tion of the protons in water was obtained by first The time dependence of Ho and the corresponding
raising the current in the magnet considerably observed oscillograms in the inverse process of
above the value necessary for resonance condi- turning the magnet current on are schematically
tions and then turning it off. This causes the field indicated in Fig. 6.
of the magnet to decrease exponentially with an 'The most interesting feature to be observed
appreciable time constant, the resonance field lies in the fact that in the case of Fig. 5, where the
entering and leaving the modulation range during field is decreasing, the observed signal voltage is
a time At of about half a second. The variation of positive while it is negative in the case of Fig. 6
the z component I& of the field with time is with increasing field. This remarkable "memory"
schematically plotted in Fig. 5. The signal was of the proton induction is in perfect agreement
with the formulae (4) and with the expression for
M, given in I, Eq. (46). I t indicates that the
relaxation time TI is long compared to the dura-
tion At of the observation; the sign of M is
determined by the sign of 6 in the past, which is
predominantly positive for decreasing and nega-
tive for increasing fields.
I t must be noted that, strictly speaking, one
cannot directly deduce from the sign of the ob-
served voltage the sign of the signal which, as was
pointed out in I, depends also upon the sign of the
proton moment. All that can be concluded is that
an inversion of the observed (rectified) signal
means a reversal of the actual (radiofrequency)
signal with respect to the in-phase leakage
voltage. To ascertain the sign of the proton
FIG. 7. Time dependence of HO with constant magnet
moment would require a determination of the
current. The a.c. modulation is superimposed upon a d.c. sign of the in-phase leakage voltage which we
value which in case a is above, in case b at, and in case c have not carried out in view of the fact that the
below the resonant field H*. The three sketched traces
represent the corresponding oscillograms. proton moment is already known to be positive.
482 BLOCH, HANSEN, A N D P-ICKARD

time above the resonance field H* a positive

signal was observed on the right side of the
oscillogram, similar to the case a of Fig. 7 and as
presented on trace a of Fig. 8. Thereupon the
field was quickly (i.e., during about one second)
lowered to a value sufficiently below resonance to
make the signal appear on the left side of the
oscillogram and then was held fixed a t this new
value. As was to be expected from the analogous
situation of Fig. 5, the signal was originally still
positive. However, during the following few
seconds it was observed to decrease in magnitude,
then to disappear and thereupon to grow again
FIG.9. Time de ndence of Ho with a d.c. value of the with negative values until after several seconds it
field being above t e resonant field H* for t<tl, decreasing
between tl and t~and being below H* fort> t 2 . The sketched had reached its full negative value, corresponding
traces a, 6, c, d represent oscillograms a t four different to the case c of Fig. 7 and a s presented on trace c
times, the arrows indicating the initial shift and subsequent
reversal of the signal. of Fig. 8.
This extraordinary reversal of the signal under
Another method of observing the proton signal fixed external conditions represents actually a
in water is indicated in Fig. 7. The current in the direct visual observation of the gradual adjust-
magnet, and so Hdc, is here kept constant. With ment of the proton spin-orientation to the
the relaxation time T1 long compared to 1/60 changed situation caused by the previous change
second, the signal voltage is, according to I , of the magnet current. The fact that it takes
Eq. (46), essentially determined by the time place during a time interval of several seconds
average of 6 which is positive in case a , zero in evidently indicates the relaxation time, likewise,
case b, and negative in case c. A photographic to be of the order of a few seconds. -4 photo-
record of the three oscillograms is shown on Fig. 8. graphic record of the process is reproduced on
The methods of observation, indicated in Fig. 10. The top trace a is taken a t an instant
Figs. 5-7, allow only a conclusion as to the lower (and with an exposure of slightly over 1/60
limit of the relaxation time. Thus the behavior of second), corresponding to the time tl of Fig. 9.
the signal
- under the conditions of Fig. - 7 merely
indicates that the relaxation time is long com-
pared to 1/60 second; the fact that with de-
creasing or increasing current, as in Figs. 5 and 6,
the signal does not reverse its sign while traveling
across the oscillograph screen during about +
second sets the lower limit of the relaxation time
higher, namely to a value large compared to 4
second. I t was observed, on the other hand, that
if the current was first decreased, a s in Fig. 5, and
then made to increase so that resonance condi-
tions, as in Fig. 6, were again reached after about
one minute, there was no "memory" retained of
the previous decrease. This allows a t least the
conclusion that the relaxation time is short corn-
pared to a minute, and in order to establish its
actual value, to be expected somewhere between
second and one minute, the method of observa- FIG. 10. Photographic record of the proton signal in
tion, indicated in ~ i9 was ~ . : starting at a water. The four traces from top to bottom correspond to
the times tl, t p , t ) , t p of Fig. 9. In the text they are referred
time t l , with Hdc held for a considerable previous to a s a, b, c, d , respectively.
 NUCLEAR INDUCTION EXPERIMENT 483

The following three traces represent snapshots explain relaxation times of the order of a few
taken during the next 15 seconds a t regular seconds. In order to verify this hypothesis we had
5-second intervals and clearly show on trace b the one of our water samples purified by distillation
decrease and on traces c and d the reversal of the in a vacuum* with a final partial pressure of
signal. The fact that on traces a, b, c , and (slightly) oxygen of probably no more than mm. While
also on trace d there appear two separated signals with the previous unpurified samples the reversal
on top of each other is caused by stray 60-cycle of the proton signal was essentially completed
pick-up which has the same frequency as the after 4 to 5 seconds, it took with the purified
modulation and, appearing on top of the actual sample about 15 seconds for completion of the
proton signal manifests itself in a separation of reversal. There is thus a noticeable effect of
the forth- and back-sweep, recorded during the shortened relaxation times by impurities but it is
exposure of the camera. by no means sufficient to explain their abnormal
I t was most surprising thus to establish the shortness. We can, of course, not discard the
proton relaxation time in water to be of the order possibility that even after distillation there re-
of a few seconds, although it must be admitted mained impurities in the water (originating, e.g.,
that it is difficult, as pointed out in I, to give a from material adhering to the glass walls) of
reliable theoretical estimate. Whereas such an sufficient amount and catalytic effectiveness to
estimate can be given for a gas and, t o some determine mainly the observed relaxation time.
extent, also, for a crystal, there is not enough However, it does not seem very probable to us
known about the molecular motion in liquids to that this is the case, since in one of our experi-
predict more than a crude order of magnitude. ments mineral oil was used instead of water. I t
Treating the liquid either as a highly compressed exhibited a relaxation time of the same order of
gas or as a crystal with thermal amplitudes of the magnitude, while the entirely different chemical
molecules, comparable to the lattice constants nature should be expected to result in a different
one would expect the relaxation time to be of the composition of impurities. Nevertheless there
order of a few hours rather than of a few seconds. may be an accidental coincidence and more ex-
We must point out here that the "relaxation periments will be necessary to clarify the role of
time," as observed in our experiments, does not impurities.
necessarily represent the usual (and in I, referred Another experiment of interest was carried out
to as the "longitudinal") paramagnetic relaxation with ice. The water was frozen in a small con-
time T Ibut that it representsrather an "effective" tainer and placed into the transmitter-receiver
relaxation time, which can be influenced by the assembly. No noticeable signal was observed
possibly different transversal relaxation time Tz, while the water was still frozen but the signal
introduced in I. I t will require a good deal more started to show up with the beginning of the
experimentation and observation under various melting process. We interpret this result as indi-
conditions to separate clearly the effects of TI cating a very long relaxation time in ice, since it
and Tz and, with the use of Eq. (46) of I, to was shown in I, Section 4, that the magnitude of
ascertain their values separately. This point, the signal decreases indefinitely for sufficiently
however, relates to the quantitative rather than increasing T I .
to the qualitative interpretation of our results I t seems reasonable, from theoretical con-
and is by no means sufficient to explain the siderations, that the relaxation time in a solid
extraordinary difference in magnitude between should be considerably longer than in a liquid and
the observed effective and the theoretically ex- in this respect there is nothing remarkable in the
pected paramagnetic relaxation time. behavior of ice. I t was the more surprising to find
We suspected the origin of this discrepancy t o another solid, namely paraffin, to behave in an
be caused by slight impurities and particularly to entirely different manner. With the same method
oxygen dissolved in water. Because of the mag- of observation as indicated in Fig. 5 and 6 for
netic moment of the oxygen molecules, acting as
catalysts, it would actually take only one mole- * W e wish to thank Dr. R. A. Ogg for kindly preparing
the sample for us and for valuable discussions relating to
cule of 0 2 in about a thousand molecules of Hz0 to the chemical aspect of our experiments.
484 BLOCH, HANSEN,

FIG. 11. Photographic record of the proton signal in

paraffin. The four traces from top to bottom are in the
text referred to as a, 6, c, and d. FIG. 12. Photographic record of the proton signal in a
concentrated solution of Fe(N0s)a in water. The four
traces from top to bottom are in the text referred to as
water it was found that even during the 3 second a, b, c, d.
which it took the signal to travel across the
screen, it showed instantaneous reversal after mmplete cycle was positive but that for the back
passing the midpoint and that there was no sweep it is no more the average value of 6 as in
observable dependence of the aspect of the signal the case of water, which determines its magni-
upon the previous values of Hdc. While this indi- tude, since this would be the same for the two
cates already a relaxation time considerably sweeps. Instead the short time interval of about
shorter than that of water, its actual value could 11240 second during which d is negative until
be directly ascertained by observation under returning to zero, suffices, not to make the back
conditions equal to those of Fig. 7. The photo- signal negative, but a t least to diminish it
graphic record is shown in Fig. 11 with the traces considerably. The identical behavior in reverse is
a, b, c referring to the same values of HOas the shown on trace c where resonance occurs near the
corresponding ones on Fig. 8. A fourth "zero" maximum of Ho with both signals negative and
trace d is shown on the plate where the resonance the one on the back sweep (above) being smaller
field H* was well outside the sweep range so that in magnitude than that on the forth sweep
there appears no signal. As on Fig. 10, there (below). With Hd,=H* and the signal therefore
appears a spurious signal which slightly separates appearing a t the center of the oscillogram, the
the forth and back sweep on the oscillograph to time interval of 1/120 second is sufficient for
make the trace d look like the "infinity," sign of complete reversal of the signal, as shown on
mathematics. This accidental separation of the trace b.
two sweeps is actually quite desirable in this case This leads to the conclusion that, contrary to
since, contrary to the case of water on Fig. 8 our expectations, the relaxation time in paraffin
where they coincide and show the complete is only of the order of 1/100 second. For a solid
identity of the signals 04 the two traces, there is this is indeed extraordinarily short and only a n -
here a marked difference. On trace a, where the unsuspectedly large amount of impurities can
resonance occurs near the minimum of Ho, both explain it, provided that the true explanation
signals are still positive, the signal on the back will not be found in a yet unknown and entirely
sweep (below) being, however, considerably different mechanism.
smaller than that on the forth sweep (above). The results of a last set of experiments are
The explanation is obviously that on the forth shown on Fig. 12. Here we were dealing with a
sweep the previous value of 6 during almost a concentrated solution of Fe(NOa), in water in
 NUCLEAR INDUCTION EXPERIMENT 485

order to obtain a case in which, because of the We have verified that the value of the reso-
atomic moments of the Fe ions, the relaxation nance field determined in this manner agrees with
time is short compared t o the time required to the gyromagnetic ratio for protons determined by
pass through resonance conditions. We obtain Kellogg, Rabi, Ramsey, and Zacharias.* After
thus the case of "slow passage," discussed in I, determining the resonance frequency and without
Section 5. The shape of the signal is here in com- altering the magnet current, the modulating a.c.
plete agreement with the theoretical formula, field was turned off, the transmitter-receiver
I (53), having a zero a t resonance and an up- and assembly removed and a carefully measured flip
down-swing for positive and negative values of 6, coil brought to the original position of the sample
respectively. Except for their position, there is between the pole pieces in order to determine the
no difference in the signals received on the three value of the resonant field H*. With a frequency
traces a, b, c which pertain to the corresponding v = 7.765 X lo6 set.-I it was found that the reso-
values of Ha,, represented in Fig. 6 . A "zero" nance field had the value H*= 1826 gauss which
trace d is again added, showing the slight leads to a value of the gyromagnetic ratio
separation of forth- and back-sweep.
With the time, required to sweep through the
signal of about 1/600 second we estimate the
relaxation time to be here of the order of to
sec. Other data, showing the dependence of
the relaxation time upon the concentration of the The agreement within one-quarter of a percent
paramagnetic salt will be published shortly. with the value y = 2.665 X lo4, obtained with
The shape of the signal obtained with Fe(N03)3 molecular beams, must be considered accidental
solutions makes it particularly easy to locate the since our field measurement can only claim an
resonance value of Hdc by setting the magnet accuracy of about 1 percent. We have also veri-
current to such a value that the zero point of the fied, although with somewhat less accuracy, that
signal is located a t the exact center of the the ratio v/H* for the resonance signal, obtained
oscillogram. With the whole resonance width from water, remains the same for v = 7.8X l o 6 ,
amounting to only about 1 percent of the total v = 8.8 X lo6, and v = 10.7 X lo6.
field, this setting can be easily adjusted to within
* J. M. B. Kellogg, I. I. Rabi, N. F. Kamsey, Jr., and
the order of one per permille. J. R. Zacharias, Phys. Rev. 56,!728 (1939).
 FIG.10. Photopraphic record oi the proton sipnal in
water. The four traces from top to bottok correspond to
the times t l , 1 9 , tr, t . of Fig. 9. In the text they are referred
to a s a, b, c, d, respectively.
 FIG.11. Photographic record of the proton signal in
paraffin. The four traces from top to bottom are in the
text referred to as o, b, 6, and d.
 FIG. 12. Photographic record of the proton signal in a
concentrated solution of Fe(NOl), in water. The four
traces from top to bottom are i s the text referred to as
a, 1, c, d .
 FIG.8. Photographic record of the proton signal from
water. The three traces from top to bottom correspond to
the situation a, b, c of Fig. 7.