tions reveals well-preserved plant organs seed of possible conifer origin with a surface are not suitable in this

is case as
and tissues, with conifer wood possess- well-preserved embryo and several well- candidates for coorditnate clocks of an
ing distinctive secondary xylem predomn- preserved fern annuli. inertial space. Nevertheless, the relative
inating. Young roots with well-devel- FRANcIs T. C. TING timekeeping behavior of terrestrial clocks
oped tracheids and wood rays reveal a Department of Geology, can be evaluated by reference to hypo-
tissue surrounding the xylem, which University of North Dakota, thetical coordinate clocks of an under-
may be phloem with definite alternating Grand Forks 58201 lying nonrotating (inertial) space (6).
rows of thick-walled fibers (Fig. IE). References and Notes For this purpose, consider a view of
The cortex is usually not well preserved 1. J. M. Schopf, 111. State Geol. Surv. Rep. Invest. the (rotating) earth as it would be
(Fig. IC), whereas the epidermis when 75 (1941); S. H. Mamay and E. L. Yochelson, perceived by an inertial observer look-
present is sharply delineated by its U.S. Geol. Surv. Prof. Pap. 345-1 (1962), p.
193. ing down on the North Pole from a
rather wavy cuticular layer (Fig. IC). 2. Silicified beds of possible lignite origin have great distance. A clock that is station-
The tracheids of the primary xylem been reported in the past, but none possess
well-preserved plant structures. ary on the surface at the equator has a
possess a helical secondary wall; those 3. R. W. Brown, U.S. Geol. Surv. Prof. Pap. 375 speed RQ relative to nonrotating space,
of the secondary xylem have uniseriate, (1962); E. Dorf, Carnegie Inst. Wash. Pubi.
508 (1942). and hence runs slow relative to hypo-
bordered pits with crassulae on the 4. J M. Schopf, Science 169, 274 (1970). thetical coordinate clocks of this space
S. I thank V. Facey for critically reviewing the
radial section (Fig. 1G). The rays are material and offering valuable suggestions. in the ratio 1 - R2Q2/2c2, where R
uniseriate, ranging from two to eight This work was partly supported by a faculty
research grant from the University of North is the earth's radius and Q its angular
cells (Fig. 1I). Dakota. speed. On the other hand, a flying clock
Other structures observed include a 17 April 1972 circumnavigating the earth near the
surface in the equatorial plane with a
ground speed v has a coordinate speed

Downloaded from http://science.sciencemag.org/ on March 12, 2019

RQl + v, and hence runs slow with a
Around-the-World Atomic Clocks: corresponding time ratio 1 - (RD +
V)2/2C2. Therefore, if T and To are the
Predicted Relativistic Time Gains respective times recorded by the flying
and ground reference clocks during a
Abstract. During October 1971, four cosium beam atomic clocks were flown complete circumnavigation, their time
on regularly scheduled commercial jet flights around the world twice, once east- difference, to a first approximation, is
ward and once westward, to test Einstein's theory of relativity with macroscopic given by
clocks. From the actual flight paths of each trip, the theory predicts that the
flying clocks, compared with reference clocks at the U.S. Naval Observatory, T - To=- (2RV + v),ro/2c2 (1)
should have lost 40 ± 23 nanoseconds during the eastward trip, and should have Consequently, a circumnavigation in the
gained 275 + 21 nanoseconds during the westward trip. The observed time differ- direction of the earth's rotation (east-
ences are presented in the report that follows this one. ward, v > 0) should produce a time
loss, while one against the earth's rota-
One of the most enduring scientific ferences are compared with our ob- tion (westward, v < 0) should produce a
debates of this century is the relativistic served time differences in the following time gain for the flying clock if Ivj
clock "paradox" (1) or problem (2), report. RQ.
which stemmed originally from an A brief elementary review of the General relativity predicts another
alleged logical inconsistency in predicted theory seems appropriate, particularly effect that (for weak gravitational fields)
time differences between traveling and because of some confusion about the is proportional to the difference in the
reference clocks after a round trip. capacity of such experiments to pro- gravitational potential for the flying and
This seemingly endless theoretical de- duce meaningful results (5). Special ground reference clocks. If the surface
bate, which has flared up recently with relativity predicts that a moving stan- value of the acceleration of gravity is g
renewed vigor (2, 3), begs for a con- dard clock will record less time com- and the altitude for the circumnaviga-
vincing empirical resolution with macro- pared with (real or hypothetical) co- tion is h < R, the potential difference is
scopic clocks. A simple and direct ex- ordinate clocks distributed at rest in gh, and Eq. 1 then reads
perimental test of the clock problem an inertial reference space. For low
with portable atomic clocks is now pos- coordinate speeds (U2 < c2), the ratio T-To = [gh/c' - (2RQv + v')/2c2]ro
sible because of the unprecedented of times recorded by the moving and (2)
stability achieved with these clocks (4). reference coordinate clocks reduces to The gh/c2 term, which is related to the
In this first of two reports, we pre- (1 - u2/2c2), where c is the speed of gravitational "red shift," predicts a time
sent relativistic time differences calcu- light. Because the earth rotates, stan- gain for the flying clock irrespective of
lated from flight data for our recent dard clocks distributed at rest on the the direction of the circumnavigation.
around-the-world flying clock experi- For typical aircraft speeds and altitudes,
ments. The theory predicts a detectable both the gravitational and kinematic
effect with cesium beam clocks if they Table 1. Predicted relativistic time differences terms in Eq. 2 are comparable in abso-
are flown around the world at typical (nsec). lute magnitude, and v2/2c2 is small
jet aircraft speeds (4). Moreover, it Direction compared with Rav/c2. For a westward
predicts an interesting asymmetry in the Effect circumnavigation (v < 0) both terms
East West
time difference between the flying are positive and they add to give a large
clocks and a ground reference clock, Gravitational 144 + 14 179 + 18 net time gain, but for an eastward cir-
depending on the direction of the cir- Kinematic -184 ± 18 96 ± 10 cumnavigation (v > 0) they tend to
cumnavigation (4). Predicted time dif- Net -40 ± 23 275 ± 21 cancel and produce a net time differ-
166 SCIENCE, VOL. 177
ence that may be positive or negative T-,0 (nsec)
depending on details of the flight. Fig. 1. Predicted relativis-
We can compare the predicted time tic time gain for a flying
differences with detection thresholds. If clock after a nonstop
the circumnavigation is nonstop, to a equatorial circumnaviga-
first approximation the trip time r,, - tion of the earth at var- Westward Eastward
ious altitudes. The label-
2-rR/lv1. Substitution of this value for ed dots correspond to
,, in Eq. 2 gives cruising ground speeds Concorde
and altitudes for the in- Velocity (m/sec)
T- T, [gl/lvi - Rv/1vJ/J11 - i!|/2] dicated aircraft. The area 500 11000
within the hatched lines ,
(3) is below approximate de-
This relationship is graphically illus- tection thresholds with a
trated in Fig. 1 over the range of portable cesium beam
clock.
ground speeds and altitudes of interest.
The area within the hatched lines in
Fig. 1 is below detection thresholds
estimated from past experience at the
U.S. Naval Observatory with portable and lasted 80.3 hours with 48.6 hours in so forth) in our theoretical approxima-
cesium beam clocks (7). The labeled flight. Flight data necessary for numeri- tions isfully justified, small but perhaps
cal evaluation of Eq. 4 were provided not entirely negligible first order effects
points correspond to cruising altitudes
and ground speeds for the indicated by the various flight captains. In most may arise from the presence of the

Downloaded from http://science.sciencemag.org/ on March 12, 2019

aircraft (8). Figure I illustrates that cases they traced their flight path on moon and sun. In fact, the center-of-
circumnavigations with cesium beam an appropriate flight map and recorded mass of the earth-moon system, not the
clocks at jet aircraft speeds should pro- the time and aircraft grouLnd speed and center of the earth, is in free fall
duce measurable relativistic time dif- altitude at various navigation check around the sun, and a more precise
ferences. Furthermore, the mere ex- points along the flight path. This in- calculation should include this effect. It
istence of a definite east-west directional formation divided the eastward trip into is unlikely, however, that the precision
asymmetry in the observed time dif- 125 intervals and the westward trip into of our experiments permits detection of
ferences would give strong evidence for 108 intervals. The latitude and longi- any effects other than the dominating
the validity of the kinetic term in Eq. 2. tude for each check point, read di- ones retained in Eq. 4 (10, 11).
These predictions for equatorial cir- rectly from the flight maps, combined J. C. HAFELE*
cumnavigations will be modified some- with the time (U.T.) over each check Departmnent of Physics, Washington
what for our actual flights because point permits calculation of an average University, St. Loutis, Missoluri 63130
commercial around-the-world jet flights, ground speed, latitude, and eastward RICHARD E. KEATING
of course, do not follow equatorial azimuth for each interval. The average Titne Service Division, U.S. Naval
paths, nor do they maintain constant altitude for each interval was taken as Observatory, Washington, D.C. 20390
altitude, ground speed, or latitude. In the average of the altittudes at the end References and Notes
this case, however, it is only necessary points. This information then permits l. L. Marder, Titme and the Space-Traveller
to integrate the appropriate differential numerical evaluation of the integral in (Allen & Unwin, London, 1971). This book
form of Eq. 2 along the actual flight Eq. 4. Table I gives the predicted time gives an excellent historical review of debate
on the clock paradox and serves well as an
path: differences resulting from these calcu- introduction to relativity theory.
2. Letters, "The clock 'paradox'-majority view,"
lations. Phys. Today 25, (No. 1), 9 (1972).
T - To 3. M. Sachs, ibid. 24, (No. 9), 23 (1971).
We conclude this report with a word 4. J. C. Hafele, Amer. J. Phys. 40, 81 (1972);
r[gh1c'- (2R!?vcos0cosX + v-)/2c-`] dT about uncertainty in these predictions. Nature 227, 270 (1970).
(4) 5. News and Views, Natiure 233, 519 (1971); R.
Possible errors stem from two sources: Schlegel, Natiure Phys. Sci. 229, 237 (1971).
This expression contains a slightly mod- (i) errors and deficiencies in the flight 6. It is important to emphasize that special
-elativity purports to describe certain physical
ified directionally dependent term, data, and (ii) theoretical approximations phenomena only relative to (or from the
which for nonequatorial flights becomes used in the derivation of Eq. 4. We es- point of view of) inertial reference systems,
and the speed of a clock relative to one
proportional to both the eastward com- timate the maximum possible fractional of these systems determines its timekeeping
ponent of the ground speed, v * cos 0, uncertainty from flight data errors to behavior [G. Bulilder, Autst. J. P/hys. 11, 279
(1958)]. Although inertial systems are highly
and the cosine of the latitude, cos A be less than 10 percent for each term of specialized, they have an objective physical
(4, 9). Because only lowest order rela- Eq. 4 after numerical integration. This relationship with the universe because they
have no acceleration or rotation relative to
tivistic time differences can be detected estimate includes both systematic and the universe. The difference in the times
in our experiments, only lowest order random errors in the flight data. If it indicated by two clocks locateg at the same
place is a physically observable quiantity that
terms need be retained in the calculated is assumed that the errors from these is invariant. Therefore, a correct derivation
predictions, and, to this order of ap- of a relativistic time difference after a round
terms add in quadrature, the maximum trip using a particular (inertial) reference
proximation, it is immaterial whether fractional uncertainty in the net value system is independent of that system. This
means that a subsequent coordinate transfor-
the differential time for 7 or for T,, iS for the eastward circumnavigation is mation into the (noninertial) rest system for
used in Eq. 4. about 60 percent, while that for the the clocks is unnecessary. Appropriate trans-
formations in this case wouLld be those of
The eastward trip began on 4 October westward circumnavigation is only 8 the general theory. The application of such a
1971 at 191,3011, U.T. and lasted 65.4 percent. These uncertainties are listed in transformation would not be incorrect, but
one is not required for a correct calculation;
hours with 41.2 hours in flight. The Table I. it would only unnecessarily complicate the
westward trip began during the follow- Although neglect of higher order theoretical description [J. C. Hafele, Nature
Phls. Sci. 229, 238 (1971)].
ing week on 13 October at 19140n1 U.T. terms (proportional to C -4, C -6, and 7. Previous trips with cesium beam clocks (HP-
14 JULY 1972 167
 5061A) lasting several days to weeks gave U.S. Naval Observatory, Washington, D.C. approach the ideal standard clock of
normally distributed, zero-center closure times 20390).
with a spread of about 60 nanoseconds per 11. "Nanosecond" is abbreviated as "nsec" in relativity theory.
day of trip [G. M. R. Winkler, in Proceed- this and the following report because this However, no two "real" cesium beam
ings of the Second Annual Precise Time and is the usage dictated by the editorial policy
Time Interval Strategic Planning Meeting, 10 of Science. The U.S. Naval Observatory ad- clocks keep precisely the same time,
to 11 December 1970, vol. 1 (available from heres to the "International System of Units" even when located together in the labo-
Technical Officer, U.S. Naval Observatory, in which the abbreviation for "nanoseconds"
Washington, D.C. 20390)]. is ns." ratory, but generally show systematic
8. H. J. Schaefer, Science 173, 780 (1971). 12. We thank Pan Am, TWA, and AA for their rate (or frequency) differences which in
9. Equation 4 includes the effect discussed by cooperation, and particularly their flight crews
W. J. Cocke [Phys. Rev. Lett. 16, 662 (1966)]. for providing the necessary flight data. We extreme cases may amount to time dif-
Clocks at rest on the earth's surface (at also thank R. Agricola of Pan Am and J. ferences as large as 1 ,tsec per day. Be-
average sea level) keep the same relativistic Clay of TWA, who were sent as escorts at
time independently of latitude differences. The no extra cost. The Office of Naval Research cause the relativistic time offsets expect-
effect of the difference in surface speed at provided financial support. H. N. Acrivos and
different latitudes is canceled to lowest order J. McNeece of the Time Service Division of
ed in our experiments are only of the
by a corresponding effect from the difference the U.S. Naval Observatory gave advice and order of 0.1 Msec per day (1, 4), any
in surface potential owing to theoblate figure assistance in scheduling the flight. We thank
of the earth. the Washington University Computer Center such time divergences (or rate differ-
10. Preliminary results of this work have been for computer time. ences) must be taken into account.
reported: J. C. Hafele, in Proceedings of the * Present address: Research Department, Cater-
Third Annual Precise Time and Time Interval pillar Tractor Co., Peoria, Illinois 61602. A much more serious complication is
Strategic Planning Meeting, 16 to 18 November, caused by the fact that the relative rates
1971, vol. 1 (available from Technical Officer, 24 March 1972; revised 8 June 1972
for cesium beam clocks do not remain
precisely constant. In addition to short
term fluctuations in rate caused mainly
Around-the-World Atomic Clocks: by shot noise in the beam tubes, cesium
beam clocks exhibit small but more
Observed Relativistic Time G s

Downloaded from http://science.sciencemag.org/ on March 12, 2019

or less well defined quasi-permanent
changes in rate. The times at which
Abstract. Four cesium beam clocks flown around the world on commercial these rate changes occur typically are
jet flights during October 1971, once eastward and once westward, recorded separated by at least 2 or 3 days for
directionally dependent time differences which are in good agreement with predic- good clocks. Some clocks have been ob-
tions of conventional relativity theory. Relative to the atomic time scale of the served in the laboratory to go as long as
U.S. Naval Observatory, the flying clocks lost 59 ± 10 nanoseconds during the several months without a rate change
eastward trip and gained 273 + 7 nanoseconds during the westward trip, where the (2, 5).
errors are the corresponding standard deviations. These results provide an un- These unpredictable changes in rate
ambiguous empirical resolution of the famous clock "paradox" with macroscopic produce the major uncertainty in our
clocks. results. Because of the nature of these
changes, however, their effect on the
In science, relevant experimental facts theory with flying clocks. The funda- observed time differences can be re-
supersede theoretical arguments. In an mental unit of time interval, the sec- moved to a large extent in the data
attempt to throw some empirical light ond, is now by definition equal to analysis. Under normal conditions
on the question of whether macroscopic 9,192,631,770 accumulated periods of changes in relative rates occur inde-
clocks record time in accordance with the frequency of the atomic transitions pendently, that is, there are no known
the conventional interpretation of Ein- of an "ideal" cesium beam frequency systematic correlations between rate
stein's relativity theory (1), we flew standard (2, 3). Because these clocks changes of one clock and those of an-
four cesium beam atomic clocks around are regulated by the frequency of a other. Consequently, the chance that
the world on commercial jet flights, natural atomic transition, a particularly two or more clocks will change rate by
first eastward, then westward. Then we well defined hyperfine transition in the the same amount in the same direction
compared the time they recorded dur- ground state of the 133Cs atom, they at the same time is extremely remote.
ing each trip with the corresponding Because of the random and independent
time recorded by the reference atomic character of these rate changes, the
time scale at the U.S. Naval Observa- Table 1. Observed relativistic time differences long-term average rate of an ensemble
from application of the correlated rate-change
tory, MEAN(USNO) (2). As was ex- method to the time intercomparison data for of clocks is more stable than the rate
pected from theoretical predictions (1), the flying ensemble. Predicted values are listed of any individual member.
the flying clocks lost time (aged slower) for comparison with the mean of the ob- Starting at Oh U.T. on 25 September
served values; S.D., standard deviation.
during the eastward trip and gained 1971, we recorded more than 5000 time
time (aged faster) during the westward Clock Ar (nsec) differences during the data period.
serial
trip. Furthermore, the magnitudes of No. Eastward* Westward Figure 1 shows the time difference data
the time differences agree reasonably 120 -57 277 relative to MEAN(USNO) for the en-
well with predicted values, which were 361 - 74 284 tire data period, which lasted 636 hours.
discussed in the preceding report (1). In 408 - 55 266 The labels in Fig. 1 are the serial num-
this second report, we present the time 447 - 51 266 bers of the corresponding clocks, and
difference data for the flying ensemble, Mean the traces give the measured differences
and explain the methods by which the + S.D. -59 10 273 7 in time between the corresponding
relativistic time differences were ex- Predicted clock and MEAN(USNO). Of course,
tracted. +.Error est. -40 23 275 21 no comparisons with MEAN(USNO)
The development of compact and * Negative signs indicate that upon return the were possible during the trips. Exactly
portable cesium beam atomic clocks (3) time indicated on the flying clocks was less than the same electronic arrangement was
the time indicated on the MEAN(USNO) clock
permits a terrestrial test of relativity of the U.S. Naval Observatory. used for all time intercomparison mea-
168 SCIENCE, VOL. 177
Around-the-World Atomic Clocks: Predicted Relativistic Time Gains
J. C. Hafele and Richard E. Keating

Science 177 (4044), 166-168.

DOI: 10.1126/science.177.4044.166

Downloaded from http://science.sciencemag.org/ on March 12, 2019

ARTICLE TOOLS http://science.sciencemag.org/content/177/4044/166

REFERENCES This article cites 10 articles, 1 of which you can access for free
http://science.sciencemag.org/content/177/4044/166#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the
Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights
reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S.
Government Works. The title Science is a registered trademark of AAAS.