Journal of the

OPTICAL SOCIETY
of AMERICA
VOLUME 28 JULY, 1938 NUMBER 7

An Experimental Study of the Rate of a Moving Atomic Clock

HERBERT E. IVES AND G. R. STILWELL
Bell TelephoneLaboratories,Inc., New York, N. Y.
(Received April 12, 1938)

INTRODUCTION direction of motion, and for that reason, has been

IN previous papers in this series,' various con- referred to as "the transverse Doppler effect."
For hydrogen canal rays for velocities obtainable
sequences of the alteration of the rate of a
clock in motion, which is an essential element in without difficulty in the laboratory, namely
the theory of Larmor and Lorentz, have been about 0.005 the velocity of light, a separation of
discussed. In these papers, this change in clock several hundredths of an Angstrom unit is to be
rate has of necessity been treated as an assump- expected in the lines due to laterally observed
tion., since, up to the present time it has not been stationary and moving particles.
the subject of independent experimental veri- The experiment, although recognized as of
fication. As was pointed out in these papers, the fundamental importance, has commonly been
Michelson-Morley and Kennedy-Thorndyke ex- described as beyond experimental feasibility,
periments do not give the necessary information chiefly for the reason that the light emitted by
to establish the existence of this change of rate. the moving canal rays in a tube of the conven-
An experimental search for this phenomenon is tional design does not consist of a single narrow
of particular interest, because the alteration of line, but of a diffuse band altogether too wide to
clock rate should be manifested as a positive permit the detection of the small displacement
effect, instead of the null effect characteristic of indicated by theory (Fig. la). Furthermore, in
all other optical experiments. the experiment as usually imagined, where the
The first suggestion as to a means by which canal rays are supposed to be observed from the
this positive effect might be observed experi- side, it would be extremely difficult to be sure
mentally came from Einstein and from Ritz overthat observation was made exactly at right angles
thirty years ago, namely that the newly dis- to the direction of the rays, and very small devi-
covered Doppler effect in canal rays involved ations from this direction would introduce shifts
velocities of the moving particles high enough to of the order of magnitude of the expected effect.
show the expected effect. The experiment, whose The first of these obstacles to the performance
crucial nature has been repeatedly emphasized, of the experiment has been removed by the
has been commonly imagined as performed by development by Dempster' of a type of canal-ray
observing the canal rays at right angles to their tube in which the velocities of the positive par-
' Herbert E. Ives, J. Opt. Soc. Am. 27, 177 (1937); 27, 2 H. F. Batho and A. J. Dempster, Astrophys. J. 75, 34
263 (1937); 27, 305 (1937); 27, 310 (1937); 27, 389 (1937). (1932).
215
216 H. E. IVES AND G. R. STILWELL

tions the displaced Doppler lines are observed

corresponding to motion toward and away from
the observer, and the effect to be observed is a
$ ~~~~a shift of the center of gravity of the displaced
lines with respect to the undisplaced line. As
shown in an earlier paper of this series' this shift
of the center of gravity is expressed by the equa-
tion X= X0 (l - V 2/c') where V is the observed or
measured velocity of the positive particles.
b
APPARATUS AND METHODS OF OBSERVATION

The canal-ray tubes

The experimental test which forms the subject
FIG. 1. Canal-ray spectra. (a) Undisplaced and displaced
lines in tube of original type. (b) Undisplaced and displaced of this paper was performed with hydrogen canal
lines in tube of Dempster design. rays, using the blue-green Ho line, of wave-length
4861 angstroms. The canal-ray tubes were de-
tides all lie within a very narrow range, so that signed along the general lines described by
the light emitted by them consists of a definite Dempster, as shown in Figs. 2, 3, and 4. The
sharp line or lines (Fig. lb). The second difficulty, positive particles were produced in a hydrogen
above mentioned, can be avoided by observing arc between the oxide coated filaments F and the
not at right angles, but in two directions, with grounded element of the double electrode Al, B1,
and against the motion of the particles; the the discharge being constricted to the desired
observations being made simultaneously by the region by the shielding electrode D. The double
use of a mirror in the tube. Under these condi- electrode between which the accelerating field is

FIG. 2. Diagrammatic representation of canal-ray tube.

 RATE OF ATOMIC CLOCKS 217

FIG. 3. Side view of canal-ray tube.

applied consists of two aluminum elements Hydrogen for the canal rays was stored in the
separated by 1.5 mm by quartz rod bushings side tube (E) containing activated charcoal. The
which provide insulation and also hold the elec- gas pressure in the tube was controlled by the
trodes accurately parallel and in alignment. Both depth of immersion of the side tube in liquid air.
electrode units are drilled with 1 mm diameter The relative gas pressures were measured by the
holes to provide a positive ray beam approxi-
mately 1 cm square. The front surfaces and all
edges of the accelerating plates are highly
polished, to minimize the tendency to sparking
which becomes quite bad when the higher accel-
erating potentials are used.
To one side of the perforations in the front
electrode is placed a small concave mirror C made
from an aluminum coated spectacle lens. The
radius of curvature of this mirror is such that it
will accurately image an object, such as the slit
of the spectrograph, upon itself about 2 centi-
meters in front of the observation window. The
observation window consists of a piece of plate
glass fused onto the end of the tube. The small
mirror is approximately 7 from the center of the
perforations as viewed from its center of curva-
ture. Because of the small rate of variation of the FIG. 4. End-on view of canal-ray tube.
apparent velocity with angle near zero, no dif- but falls on the other side of the spectrograph slit. It was
ference in apparent velocity of the canal rays as not believed that this lack of optical symmetry was im-
portant when weighed against the accuracy of alignment
directly viewed and as seen in the mirror, as rendered possible. However, to test this point a tube was
great as 1/100th of 1 percent can occur with the constructed furnished with a concave mirror which, by the
aid of cross hairs in the discharge path, could be adjusted
accuracy of imaging the spectrograph slit upon to image the discharge upon itself. This tube while more
itself which was possible in the experiment.' difficult to align accurately, gave identical results, as did
also a tube with a plane mirror, so that any suspicion that
3With this arrangement the real image which the mirror the phenomena observed could be ascribed to optical
forms of the discharge is not coincident with the discharge dissymmetry is untenable.
218 H. E. IVES AND G. R. STILWELL

C, =1/4 M.F.CONDENSER
C2 =y/2 M.F.CONDENSER
R,,R3 =100,000 OHMRESISTOR
R2 = (4) 50 MEGOHMRES.
L, = 200 HENRYS

FIG.5. Electrical circuits used with canal-ray tube.

gauge (G) which is a thermocouple gauge similar this method was found to give a more copious
to one described by G. C. Dunlap and J. G. supply of positive ions just in front of the holes
Trump ;4 this manometer was previously cali- in the accelerating electrodes with a consequent
brated against a McLeod gauge to read hydrogen brighter canal-ray beam.
gas pressures between 0.001 mm and 0.1 mm. The high negative potential applied to the
The whole tube with its attached charcoal accelerating electrode (B) was maintained by an
reservoir was baked out while being pumped to alternating current rectifier capable of delivering
eliminate any foreign gases, after which the up to 30,000 volts d.c. Fig. 5 shows the diagram
filaments were activated. Hydrogen gas to a of the electrical circuit used which is essentially
pressure of several millimeters of mercury was a voltage multiplying circuit. The transformers
admitted to the pump station and the charcoal and associated equipment were selected to have
tube was partly saturated with hydrogen by a much greater power rating than was necessary
for the performance of this experiment thereby
immersing it part way in liquid air. The tube was
then sealed from the pump station. insuring greater steadiness of operation and de-
creasing fluctuations of the output voltage.
Electrical power supply and measuring apparatus The rectified voltage was smoothed by the filter
The low voltage hydrogen arc was maintained network combination (R1), (Li) and (C2) and fed
between the cathode (F) Fig. 2 and the grounded to the canal-ray tube through the limiting re-
accelerating electrode (A) by an a.c. rectifier sistance (R 3). The applied d.c. voltage was
capable of delivering 400 milliamperes at 100 indicated on a high resistance voltmeter con-
volts. Fig. shows the electrical circuit diagram sisting of four 50 megohm resistances (R 2 ) in
and how the connections were made to the series with a zero to one hundred full scale
canal-ray tube. This arrangement differs from microammeter. The resistors were especially
that used by H. F. Batho and Dempster 2 and made physically large to have very small tem-
others, in that the hydrogen arc was maintained perature coefficient for the small current values
in a reverse direction between the first accelerat- used. The resistances were measured and the
ing electrode and cathode instead of between the meter calibrated to read within an accuracy of
cathode and a third electrode. Experimentally one percent.
I G. C. Dunlap andl J. G. Trump, Rev. Sci. Inst. 8, 37
Since it was important in this experiment to
(1937). hold the applied accelerating potentials very
 RATE OF ATOMIC CLOCKS 219

steady during long photographic exposures an dimensions to bring the Ho3line in the middle of
additional voltage reading technique was em- the photographic plate. An accurate bilateral
ployed which greatly facilitated the observation slit was mounted so as to be rotatable about its
of small voltage fluctuations. This consisted of a center point with great exactness. Behind the
Wtilf electrostatic voltmeter whose sensitivity slit was a pivoted observing telescope and ex-
could be varied with the applied voltage being tended light source behind a transparent mirror.
used. The deflected reed and illuminated scale By observation with this microscope the orien-
indicating the applied voltage was projected, tation of the canal-ray tube and its small mirror
greatly enlarged, onto a screen across the room. could be made with extreme accuracy, the
The sensitivity of this arrangement could be adjustment being complete when upon rotation
adjusted so that 1/10 of one percent of the of the slit its reflected image remained exactly
voltage indicated by the resistance voltmeter coincident with it.
could be easily detected. The whole spectrograph was mounted on a
Voltage adjustments were made by means of heavy steel framework which was arranged to be
a manual control in the primary circuit of the capable of rotation about a vertical axis, and the
high potential transformer; this control included apparatus as a whole was placed on a layer of
a variac step-down transformer and series re- anti-vibration material in a basement room
sistance. which was thermostatically controlled to hold to
a temperature to within 1/10th of a degree cen-
The spectrograph tigrade. The arrangement of the apparatus is
The dispersing element of the spectrograph shown schematically in Fig. 6, and in the
used was a metal on glass plane grating made by photograph, Fig. 7.
Professor R. W. Wood. This is approximately 7 A number of appropriate adjusting means were
inches in diameter with ruling of 15,000 lines to provided, including screw adjustments for trans-
the inch, with grooves so controlled as to throw a lation and rotation of the canal-ray tube, visible
very large part of the total light into the first in the photograph, Fig. 8. A sighting aperture
order on one side. In conjunction with the grating and telescope were arranged at the collimator
were two high quality telescope objectives of lens, by means of which the placing of the
approximately 5 feet focus. The grating was mirror in the canal-ray tube so that it was in line
mounted at right angles to the axis of the tele- with the optical axis of the spectrograph was
scope element, under which conditions a normal assured by preliminary adjustment before the
spectrum is obtained. 5 The whole spectrograph use of the microscope at the slit. A small slowly
was constructed as a rigid iron box of exactly the rotating variable sector was provided in front
of the photographic plate by means of which the
A test for the normality of the spectrum produced by exposures of the various canal-ray lines could be
the grating- lens combination was made on three lines of the
molecular spectrum of hydrogen covering a range of 250A equalized.
approximately centered about the H line, namely
4719.01A, 4849.32A and 4973.26A. The following figures Measurement of the spectrograms
were obtained tor the quantity A/mm:
A/mm The dispersion of the spectrograph was 10.87
Interval 4719.01-4849.32A
Plate A, measured upward on screw 10.870 angstrom units per millimeter. The displacement
PI, "1 "1 downward 10.875
Plate B, upward 10.876 due to the Doppler effect for H2 in the neighbor-
downward 10.877
Mean 10.874 hood of 20,000 volts is about 2 millimeters; the
Interval 4849.32-4973.26A expected shift of the center of gravity of the dis-
Plate A, measured upward on screw 10.877 placed lines (directly viewed and reflected in the
" "1 " downward 10.874
Plate B, upward 10.872 mirror) is about 0.05A or 0.005 mm. This made
downward 10.870
Mean 10.873 it desirable to have means for measuring the
A hyperbolic deviation from linearity, such as would spectrograms with an accuracy at least 0.001mm.
produce an apparent shift of the center of gravity of the The measuring microscope used was a Carl
magnitude in question in this investigation would mean a Zeiss instrument furnished with a special large
difference of 0.5 A/mm in the quantity for the two wave-
length ranges measured. diameter reading drum, on which 0.001 mm
220 H. E. IVES AND G. R. STILWELL

GPAS/NG

7o .ND Yt1OH

FIG. 6. Plan of apparatus.

could be read directly, and tenths of divisions OBSERVATIONS

estimated. The accuracy of the screw was not Characteristics of the canal rays
however to be relied upon for better than 0.001
Typical spectrograms obtained in the inves-
mm, and so a special measuring procedure was
tigation are shown in Fig. 9. In each the center,
employed, made possible by the approximately
undisplaced line is seen, accompanied at either
symmetrical character of the pattern presented side by two companions, which, by their separa-
by the displaced lines. A series of measurements
tions from the center line by distances in the
were made of separations of the displaced lines ratio \/2/V3 are identified as H 2 and H3, or
from the center, undisplaced line, the latter double and triple atomic hydrogen. No H, par-
being accurately located at a chosen point on the ticles were found in this work.
screw. The plate was then turned end for end, A fundamental assumption in this experiment
the center line again placed accurately on the is that the light emitted by the excitation of the
same chosen point of the screw, and the separa- stationary particles in the tube, and by the
tions remeasured. This procedure was then re- excitation of the multiple mass traveling par-
peated at several points on the screw, and the ticles, is (except for the predicted second-order
mean values of the several series were used. The change), of the same frequency. Put in another
probable errors of the final values so obtained way, the assumption is that in every case the
were found by analysis of the measurements to emission is that of the single excited hydrogen
be about 0.0025A. atom, to which all the particles must be assumed
RATE OF ATOMIC C L OCK S 221

FIG. 7. Photograph of apparatus.

FIG. 8. Tube in place.

222 H. E. IVES AND G. R. STILWELL

they are significantly affected by the collisionand

their processes active in the discharge.
20755 Volts Preliminary measurements and choice of con-
ditions
The first measurements made on the spectro-
grams were used to check the theoretical rela-
tionship
eE = 2M( V/c 2 )
13702 Volts
when E is the voltage between the electrode
plates, e the charge on the hydrogen atom, and
M1the mass of the particle observed. V/c as
obtained from the ratio AX/X should agree with
the value obtained from the above formula.
7859Volts In Fig. 10 are shown, bv the dashed lines, the
computed values of AX(=X0(V/c)), for H 2 and
H3, plotted against the square root of the voltage;
and by the marked points, the observed values
of AX at seven voltages, selected as described in
FIG. 9. Spectrograms obtained for several applied voltages.
a later paragraph. These "observed" values are
the mean measured displacements of the direct
to revert before emitting light. There is no experi-
and reflected Doppler lines, increased in the
mental or theoretical reason for expecting this
ratio 1/cos 70, where as previously described, 70
not to be the case, at least to a high order is the angular displacement of the concave mirror
of approximation, although very considerable from the axis of the beam. This gives the dis-
changes in the relative intensities and sharp-
placement for observation exactly along the
nesses of the canal ray lines occur over the range
beam, which is the quantity directly derived
of voltages and pressures used in this experi-
from the fundamental relationship between
ment. Thus the center, undisplaced line is very
voltage and velocity, and is the quantity with
weak, relative to the displaced lines, at voltages
below 3000, but is quite sharp. At voltages of
10,000 and over the center line becomes diffuse ',H 2
but with a sharp central core, and is much more 20 -7-I
intense than the displaced lines. (These differ-
ences in intensity do not show in the spectro-
grams, Fig. 9, as these were made with the 18

variable sector adjusted to equalize the ex- z

posures.) At low gas pressures (below 0.005 mm In
-%~~~~-
0
0
Hg) the H2 lines are brighter than the H 3 ; at
14 -I- ___ I-,
higher pressures H 3 becomes relatively more z
-0
intense. The Ho3line is actually a close doublet, z
of separation 0.09A, but always appeared single 12~~~~~1-
and symmetrical in our spectrograms. The pos- I-,
sibility of a varying distribution of intensities
between the components is not excluded. 10 ~
In default of evidence to the contrary the
assumption above stated has been taken as valid
in carrying through the experiment but with 8
80 Co
- _ _
100
_ _
110
_ _ _
120
_
130 140
recognition of the obligation to scrutinize the
final results carefully for any indications that FIG. 10. Computed and observed Doppler displacements.
 RATE OF ATOMIC CLOCKS 223

, -11-IM , "'
PUMP F [T
i II
I

I
wi~
-1~
~~]; I a

FIG. 11. (a) Molecular spectrum of hydrogen in the neighborhood of H3.

(b) Canal-ray spectrum for voltage selected to make H2 lines clear molecular
lines.

which the second-order shift is theoretically cor- 11) or upon very long exposure of the canal-ray
related. The data used in plotting Fig. 10 are spectrum, a large number of these second spec-
assembled in Table I. It will be seen that the trum lines appear. While very faint and under
observed Doppler displacements show no sys- exposed in the canal-ray pictures, these lines
tematic deviations from the computed and agree were found, if they fell close to or partially over
with these in general to within about 1 percent. the canal-ray lines, to shift the position of these
The next measurements were those to detect latter enough to vitiate the measurements.
and measure the shift of the center of gravity, if The next step was then to plot the positions
existent, of the directly observed and reflected of the second spectrum lines with the object of
lines. The first measurements revealed very finding which voltages could be used to place the
erratic and confusing displacements, which canal-ray lines in clear spaces, in Fig. 12 is shown
showed no systematic relationship with the orien- a plot of the chief second spectrum lines in the
tation of the apparatus or other known variables. neighborhood of the Ho line, the lines on both
The source of these unsystematic shifts of the red (dashed) and blue (full) sides being plotted
center of gravity of the displaced lines was together. On the same plot are shown the dis-
found to lie in the presence of faint lines of the placements of the canal-ray lines against the
second or molecular spectrum of hydrogen. A square root of the voltage as measured. Plotted
few lines of this spectrum are obviously present, in this way the Hi, H2 and H3 Doppler displace-
at the highest voltages, at positions other than ments fall on straight lines. As can be seen, the
those occupied by the canal-ray lines as shown
number of available voltages for securing pairs
of direct and reflected lines in clear spaces is
in Fig. 9. However, upon photographing the
spectrum of the arc behind the electrodes (Fig. limited.
After picking voltages corresponding to satis-
TABLE I. factorily clear spaces, the next experiments were
made with sufficiently rapid plates to make
XOV/C MEAN AX OBSERVED
PLATE VOLTAGELINE COMPUTED (AX/cos 7) possible obtaining exposures in about one hour.
169 6,788 HI 10.62A 10.35A
These plates were of too coarse grain for measure-
160 7,780 H2 14.04 14.02 ments of the highest accuracy, but the short
163 9,187 H2 15.30 15.40 exposure time was desired in order to investigate
170 10,574 H2 16.34 16.49
165 11,566 H, 13.88 14.07 whether there was any effect due to the orien-
172 13,560 H2 18.50 18.67 tation of the apparatus with respect to the
172 13,560 H2 15.05 15.14
177 18,350 H2 21.55 21.37 earth's motion. A series of plates were made, as
shown in Table II, with the apparatus pointing
224 H. E. IVES AND G. R. STILWELL

200

I 10000
80

60

40
20
0 0.2
MM

FIG. 12. Chief molecular spectrum lines of hydrogen in the neighborhood of Ho. Lines to
blue side shown full, to red side, dashed.

north, south, east and west, and using a voltage These plates gave eight points for measurement,
of approximately 18,000. These experiments three due to H 3 and five to H 2.
showed at once that there is a shift of the center The results are shown in Table III and Figs.
of the gravity of the displaced lines, that is, a 13 and 14. In the table, computations of the ex-
change of frequency of the light emitted from pected shift (A'X) are made in two ways. In
the canal-ray particles. They showed that this column 4 the quantity X0 (4 V 2/C2), which should
shift was of the order of magnitude to be expected equal A'X, is computed from the values of V/c
from the theory, and that it was independent of as derived from the voltage between the elec-
the direction of the apparatus. trodes. From the fundamental equation con-
An additional experiment made at this stage necting voltage and displacement it appears that
of the investigation was a test of the possible since V 2/c2 is proportional to voltage, A'X so
effect of a magnetic field. Due to the iron of the computed should be proportional to voltage, the
apparatus, a magnetic field existed at the tube factor of proportionality being different for the
which was fairly constant in direction with particles of different masses. In Fig. 13 the com-
respect to the tube, and in amount. This field puted values for H2 and H3 are shown by the
was neutralized by properly placed coils, and dashed lines, and the observed values, A'X (6th
then replaced by one of approximately ten times column) by the circles and dots. These lie along
its strength in the opposite direction. No effects the theoretical straight lines to within the prob-
of this variation of the magnetic field were found. able errors of the measurements.
TABLE I I.
Final results
DIRECTION
The final series of plates were then made with PLATE OF XO(1V2/C2) COMPUTED
NUMBER DISCHARGE OBSERVEDSHIFT FROMOBSERVEDV/C
the apparatus in one position, using special
Eastman type III-J plates which are of fine grain, 136 N 0.036A 0.0472A
137 N .052
but of a low speed, requiring all-day exposures. 138 S .050
139 E .045
In this series seven different voltages were used 140 W .051
ranging from 6788 to 18,356, adjusting as above 141 W .047
described to locate the displaced lines in the Mean 0.0468A 0.0472A
regions of the spectrum free from molecular lines.
 RATE OF ATOMIC CLOCKS 225

In column 5 values of A'X are computed using

the observed values of AX(= V/c). This method !- .040 -

of computation involves no knowledge of the

o .030
voltages used, and so is independent of any errors - - -

of voltage measurement. The A'X's so computed

are shown, plotted against AX, by the dashed .020 -0 -1:;-I -

- -
curves in Fig. 14, and the observed values are .3 1~00 600 120

again shown by the circles and dots. The agree- -0 4000 8000 12000 16000 20000
ment is within the probable errors of the measure- VOLTS

ments, and the results are not significantly FIG. 13. Computed and observed second-order shifts,
plotted against voltage.
different from those plotted in Fig. 13, as should
indeed be the case because of the close agreement
of the Doppler shifts (AX) with the values com-
puted from the voltages.
The conclusion drawn from these experiments
is that the change of frequency of a moving light U,

source predicted by the Larmor-Lorentz theory z

is verified.
0

DISCUSSION
z
Characteristics of results Z
The results graphically exhibited in Figs. 13 140

and 14 appear to be a satisfactory confirmation

of the Larmor-Lorentz theory. It is proper, how-
ever, to note features of the experiment and the
results which might be improved in further work. 0 5 10 15 20 25
Probably the least satisfactory feature is that the AX IN ANGSTROM UNITS

displacements of center of gravity which con- FIG. 14. Computed and observed second-order shifts,
stitute the evidence for the change in frequency plotted against first-order (Doppler) shifts.
are in every case less than the known separation
of the components of the H3 line. This leaves the be necessary to permit these without a break-
possibility that some of the observed shift could down of the gap between electrodes. A possi-
be due to a variation of relative intensity of these bility here is provision for continuous pumping
components, under the excitation conditions to hold the pressures different in different parts
holding in the tubes. This variety of error could of the tube. The same result, without the neces-
be excluded if voltages at least twice and prefer- sity for such high voltages, could be secured if
ably several times the highest here used could be the single mass particles HlI could be produced,
employed. Radical changes in tube design would as these should give twice the shift of the H 2
particles. We have, following suggestions in the
TABLE III.
literature, tried the introduction of water vapor,
2 2
X)O(1V/C ) COM- XO(IV2/C2) COM- and admixtures of helium and neon, but without
VOLT- PUTED FROM PUTED FROM AIX
PLATE AGE LINE VOLTAGE OBSERVED AX OBSERVED success.
The lack of positive assurance that the canal-
169 6,788 H, 0.0116 0.0109 0.011A
160 7,780 H2 .0203 .0202 .0185 ray particles do actually emit the same frequency,
163 9,187 H2 .0238 .0243 .0225 despite changes of voltage and pressure condi-
170 10,574 HZ .0275 .0280 .027
165 11,566 H, .0198 .0203 .0205 tions has already been dwelt on. Insofar as ex-
172 13,560 H2 .0352 .0360 .0345 plaining the shifts of the center of gravity by a
172 13,560 H, .0233 .0237 .0215
177 18,350 H2 .0478 .0469 .047 changing frequency of the undisplaced line, this
appears very unlikely when the results are
226 H. E. IVES AND G. R. STILWELL

plotted for the two kinds of particles at the same the light signals, and an independent experiment
must furnish this rate in order to completely
voltages, as is done in Fig. 13. If these shifts were
due wholly or in part to a change in the center solve the problem. The present experiment
line, the shifts for both kinds of particles would establishes this rate as according to the relation
be expected to be the same or to fall on lines
v = V( - Vl/0) 1,
inclined otherwise than this precise way. The
possibility that the stationary and traveling H2 where o the frequency of the clock when sta-
and H3 particles suffer changes of frequency due tionary in the ether, v its frequency when in
to collision processes which happen to follow motion. It follows then on combining this result
exactly the relation expected from the Larmor- with the results of the Kennedy-Thorndyke
Lorentz theory from velocity changes alone, is experiment that the dimensions of the moving
remote, even though conceivable. The consis- apparatus are contracted by the factor (1-V2/C2)
tency of our results with the prediction of the in the direction of motion, and are unaffected at
theory, over a considerable range of velocities, right angles to that direction.
and for two kinds of particles, can itself in the The Michelson-Morley experiment and Ken-
absence of any other known cause for the ob- nedy-Thorndyke experiments, yielding null re-
served effect, be taken as evidence for their cor- sults, could, of themselves, be equally well
rectness. explained-and more simply-by assuming an
ether entrained by the earth, or a ballistic char-
Significance of results
acter of light emission, instead of assuming two
This experiment forms the necessary optical concealed conspiring compensations-contrac-
complement to the Michelson-Morley experi- tions of dimensionsand of clockrates. The present
ment, and the more general form of this experi- experiment, giving a positive instead of a null
ment due to Kennedy and Thorndyke. The null result may hence be claimed to give more decisive
result of the latter experiment can be explained, evidence for the Larmor-Lorentz theory than is
on the assumption of a fixed luminiferous ether, given by the experiments which have yielded null
by contractions of the apparatus along and results.
across the direction of motion which are in a The discussion of the consequences of this
definite ratio to each other, namely (1- V2/C2): change in clock rate, the reality of which may be
where V is the velocity of the apparatus and c taken as established by this experiment, consists
the velocity of light. From these experiments the practically of the entire theory of the optics of
contractions are unfixed in absolute amount; i.e., moving bodies as developed by Larmor and
the contractions in the two directions could be Lorentz. Some of the special consequences of the
1- VI2 /c2 and (1- V2/c') , etc. For each absolute use of clocks which behave in this way when in
contraction, there is demanded in general a motion have been developed in earlier papers of
change in the rate of the clock at the origin of this series.'