The
Atomic Clock
early work in US | NPL
enters the race | inside the clock | birth of atomic time |
hydrogen maser | home
Quartz clocks had
revealed an annual periodic variation in the time of rotation of the earth and
hence in the value of the astronomical second of time. The stability of the
clocks enabled them to smooth out these variations as well as other small
irregularities in the earth’s motion and in the errors of observation but they
could not be regarded as standards. They were adjusted to keep time with the
earth and their virtue was that once adjusted they operated very stably from
year to year providing a time scale which could be read with great precision.
Little was to be gained by improving them still further. It was the fundamental
standard itself which was no longer adequate for modern practical application
in the fields of air navigation and communication.
The only possible
alternative appeared to be a natural periodicity within the atom. The science
of optical spectroscopy had given a picture of the atom as a miniature solar
system having a central nucleus with a number of electrons revolving round it
in permitted orbits. When an electron jumped from one orbit to another, light
of a specific frequency was emitted or absorbed according to whether the jump
was to a higher or lower energy state. Well know examples are the mercury and
sodium lamps that illuminated the streets. The frequencies of these optical
spectral lines are too high to be measured directly in terms of quartz clocks
but are calculated from their wavelength which can be measured and the velocity
of light, velocity being the product of wavelength and frequency.
The development of
microwave techniques during the war provided spectroscopists with a powerful
new tool and enabled them to study the response of atoms to electromagnetic
waves covering a whole new band of frequencies. Atoms of the alkali metals were
of particular interest because they have a single electron in the outermost
orbit and, therefore, give the simplest spectrum. The results were brilliantly
interpreted and led to the assumption that the outer electron and nucleus were
spinning in either the same or opposite direction and that the two conditions
represented states having slightly different energies. Transitions between them
were accompanied by the emission or absorption of a (quantum of) radiation in
the microwave region of the spectrum. The significance of this from our point
of view is that the frequency can be measured in terms of the quartz standards
and it becomes potentially possible to define the second as the time occupied
by a certain number of cycles of an atomic spectral line. It was suggested by
Rabi that a spectral line of an isotope of caesium might be suitable. Its
frequency was near 1010Hz in a band used for radar.
Early work in US
I was naturally
interested in these developments although spectroscopy was so far out of my
field that I did not expect to take an active part, that is, until I visited
the USA in 1950 and saw the work at MIT and Columbia University. Zacharias at
the MIT was quite enthusiastic and although he was not interested in clock
making himself he was confident that his technique could be developed to form
the basis of a time standard. The microwave spectral lines are so feeble that
they cannot be detected directly, but use is made of the magnetic properties of
the atoms arising from their spins. They behave like tiny magnets which are
deflected in a non-uniform magnetic field; and if they are subjected to another
field oscillating at the frequency of the line they are induced to change to
the other energy state, the polarity of the magnet is reversed and they are
deflected in the opposite direction. He knew of our expertise in the microwave
work and thought that we would not find the atomic beam technique too difficult.
A peculiar feature of these spectral lines is that their sharpness depends on
how quickly the transition takes place, the slower the better. For a useful
clock the time would need to be much longer than that used in the MIT
experiments so that the atoms would have to travel a longer distance in a
highly evacuated enclosure. I was not sure that a sufficiently high vacuum
could be obtained but he pointed out that we had in the UK two of the world’s
best manufacturers of vacuum equipment.
At Columbia
University Kusch expressed similar views although his colleague Townes favoured
the use of ammonia which had a much stronger spectral line, using a new
technique. This became the maser which was studied in many laboratories as a
standard without much success. It did, however, lead to the laser with its
revolutionary applications. From my experience I decided that the atomic beam
using caesium was the best option. The main problem arose from the high
accuracy required. In spite of its shortcomings the astronomical second was the
most precise of all our standards and with the help of quartz clocks was made
instantly available with an accuracy to one part in a hundred million or a
thousandth of a second in the length of a day. It was reasonable to aim for at least
a ten-fold improvement and this in my view ruled out the use of the ammonia
line. On my return to the UK I suggested that the atomic clock should be added
to our programme of work. The Director Sir Edward Bullard was sympathetic to
the idea but could not provide the two or three extra people I thought would be
required as the NPL was then concentrating on the construction of a computer
and any spare hands were directed to that project. The delay was disappointing
but it enabled me to improve the velocity of light measurement and also to work
with Froome on the refractive index of air. The precise frequency at which
these measurements were made was immaterial so I chose a value near to that of
the caesium spectral line ensuring that the electronic equipment was ready if
and when it became possible to start on the atomic clock. One item which proved
to be most important was an oscillator which could be varied smoothly in the
region of the caesium frequency. The spectral line was known to be very narrow and
in order to find it the searching oscillator would need to have even greater
spectral purity. We had often used microwave oscillators stabilised by cavity
resonators in the manner described by Pound and realised what a remarkably
useful device it was. By using the large cavity resonator used for the velocity
of light measurement and adjusting the circuit conditions with exceptional
care, I was able to reduce the bandwidth of the oscillations to a few cycles.
In a demonstration
given at the Institution of Electrical Engineers two oscillators had been shown
to beat together with a clean audible note. It seemed well suited to searching
for the spectral line and measuring its exact frequency. It was realised that
when this was known it would be necessary to synthesise the frequency from the
quartz standard and the spectral line; and provision was made for this step. I
knew that it was impossible to keep the signal pure by synthesising directly
from our 100kHz standard and that an intermediate stage must be introduced. An
oscillator that appeared to be suitable was made by Marconi’s Wireless
Telegraph Co. Its nominal frequency was 5MHz but they agreed to adjust one to
the required frequency when I let them know what this was. It turned out to be
5.0069MHz which when multiplied by 1836 gave the caesium frequency. Provision
was made for altering the quartz frequency a small amount by adjusting the
electronic circuit.
A further visit to
the USA in 1953 strengthened my interest in building an atomic clock. There had
been a change in attitude at the MIT where Zacharias was now keenly interested
in time keeping. He had made arrangements with a firm to make a commercial
model and was designing a model of enormous potential accuracy to build in his
laboratory. The National Bureau of Standards in Washington was working on
another model to the design of Kusch who was acting as their consultant.
NPL enters the race
Although it now
seemed probable that an atomic clock would soon be in operation in the USA it
was important to make an independent clock at another national standardising
laboratory. Some of the technical problems had now been solved and I proposed
to start work on a clock with our existing staff and some assistance from
another scientist JVL Parry who had expressed an interest. His half-time help
soon became full time and he was ideally suited to the work not being afraid to
tackle new techniques and having a flare for experiment. We both had a good
relationship with the workshop, recognising the important part they played in
our success. Our rough sketches of the beam chamber were discussed with A
Gridley, the head of the instrument workshop and he converted them into working
drawings for the mechanics. In the remarkably short time of a few months we
were able to begin assembling the equipment in the laboratory, where a space
had been prepared. The papers published on atomic beam spectroscopy had all
stressed the need to avoid ground vibrations. I knew something about this
having hade the same problems with quartz clocks. A large slab of concrete had
been let into the ground and supported on springs adjusted to give the block a
natural period of vibration of a few cycles per second. As our design developed
we realised that the deflections of the atoms could be made so large that
ground vibration could be ignored. The springs were never unclamped although I
expect the concrete slab is still there.
Inside the clock
The atomic clock was
made possible through the brilliant theoretical and experimental work of a
number of scientists, several of whom received Nobel prizes, but the clock
itself is very simple, as can be seen from the sketch.
Atoms leave the
oven through a narrow slit and pass between the pole pieces of a powerful
magnet which is shaped to give a non-uniform field. They follow various paths
according to their initial direction, velocity and energy state. Only two paths
are shown. A few of the atoms are selected by the slit half way along the path
and continue through the pole pieces of a second magnet, which is the same as
the first and increases the deflections in the same direction and away from the
centre line. A weak radio field is applied in the space between the magnets
and, when its frequency and strength are exactly right, the atoms jump to the
other state, those initially in the low energy state absorbing energy from the
field and, strangely enough, those in the high energy state being induced to
emit energy, so that they are all reversed. Their deflections in the second
magnet are also reversed and they are deflected back to the centre line where
they strike the detector. This is a hot tungsten wire which imparts a charge to
the atoms which boil off as charged particles and are attracted to electrodes
and, after enormous amplification, measured as an electric current. Atoms which
are not in the two states concerned are not deflected at all and give a steady
signal which is useful for lining up the apparatus. The beam strength increases
by ten per cent when transition occurs. The components are all contained in a
metal pipe about 150cm long evacuated as completely as possible. Fortunately
this was at a time when enormous improvements were being made in vacuum practice
and we were able to make immediate use of them, two stage diffusion pumps,
liquid air traps, neoprene O-rings for making seals and sophisticated devices
for measuring the pressure.
Birth of atomic time
In less than two
years the equipment was installed and the beam detected. The radio frequency
oscillator was switched on and its frequency varied through the region where
the line was expected. We were incredibly lucky to find the right conditions
after searching for a few days and there was the resonance exactly as sharp as
predicted. We invited the Director to come and witness the death of the
astronomical second and the birth of atomic time. And it was indeed the birth
because much to our surprise it was another year before any clocks were working
in the USA. It was obvious from this very first moment of operation that we
could set the quartz clocks with a far greater accuracy than could be obtained
by astronomical means.
Our first task was
to make every possible test to check to what extent the frequency could be
varied by external conditions such as pressure, temperature, strength of the
electric and magnetic fields, and so on. This could only be done by
establishing a provisional atomic time scale, making use of the stability of
our quartz clocks. They were set at intervals of a week by means of the atomic
clock operating under standard conditions. These conditions were then varied
and the effect measured by the quartz clocks. The test showed that with a very
simple control of the conditions the atomic clock was enormously more accurate
than astronomical time as well as having the advantages of being far simpler to
use and being immediately available. It did not give the time of day, of
course, but this is not required accurately. It is the length of a time
interval and its inverse, frequency, that it needed ever more accurately for
modern developments in navigation, computers and communication.
A few months after
the atomic clock had been in operation The Astronomer Royal invited me to
describe it at the meeting of the International Astronomical Union to be held
in Dublin. One of the main subjects for discussion was the adoption of a new
unit of time. Astronomers knew that the unit based on the rotation of the earth
was 
no longer adequate and they were
recommending a unit, the second of the ephemeris time, based on the revolution
of the earth round the sun. Unfortunately although this unit might be expected
to be more constant than the mean solar second, it is much more difficult to
measure, and the observations would have to be averaged over years to give the
required accuracy. This rendered it useless as a unit of measurement which must
be available immediately. I pointed out that whatever advantages this unit
might have for the astronomer it was useless for the physicist and engineer,
and suggested that since an atomic unit would be needed in the future it would
be wise to defer a decision until agreement could be obtained on the definition
of such a unit. There was no support for this suggestion and the second of
ephemeris time was adopted and was later confirmed by the International
Committee of Weights and Measures, showing how even scientific bodies can make
ridiculous decisions. One useful outcome of the Dublin meeting was that with
the help of Markowitz – I was not an official delegate myself – a resolution
was passed to the effect that when the relationship between ephemeris time and
atomic time was established the atomic clock could be used to make astronomical
time available. This meant that we had international approval to introduce
atomic time when the comparisons were completed without further international
meetings. A detailed programme was arranged with Markowitz. The time interval
between certain time signals was measured at the NPL in terms of the atomic
clock and at the US naval observatory in terms of the ephemeris second. The
comparisons took longer than anticipated because of the relative inaccuracy of
the astronomical measurements but after three years it was decided that further
averaging was not likely to improve the result. The value was, therefore,
announced and was eventually accepted internationally as the unit of time.
In parenthesis it
might be explained here that before changing a unit of measurement it is
essential to establish the relationship between them so that measurements made
in the past would still be valid within the accuracy of the old unit. The
second of time presented special problems because it was known that the mean
solar second varied. That is why the atomic second was linked with the second
of ephemeris time which was believed by astronomers to equal the average value
of the mean solar second over some hundreds of years.
No results were
announced from the USA until our clock had been working for a year. It
transpired that at the Bureau of Standards there had been difficulties with the
microwave source and the work had been delayed by a move from Washington to
Boulder, and the ambitious scheme of Zacharias had turned out to be too ambitious.
The commercial development was going well but had presented technical problems
not easily solved. The work moved from the National Company, to Varians and
then to Hewlett Packard, the scientist concerned going with it and was
supported all the time by government agencies. The clock finally put on the
market some years later was a wonderfully compact and accurate instrument.
Although some
clocks in the USA were in operation before the completion of our comparisons,
Markowitz kept to our agreed programme and the published value was based
entirely on the NPL clock. A sub-committee of the International Committee of
Weights and Measures was set up to discus atomic time and it is interesting to
follow its gradual and reluctant acceptance by astronomers. The meeting in 1957
refused to accept the term atomic clock insisting that it was simply a
frequency standard for the second: the second meeting in 1961 accepted that it
was a standard of time interval but continued to stall by recommending that
further work should be done: the third meeting in 1963 recommended the adoption
of an atomic unit of time the value being that obtained at the NPL. No formal
steps were taken to implement this recommendation and the International
Scientific Radio Union, in which I was the chairman of the relevant section,
had to stress the urgency of putting the resolution into effect. It was
formally adopted as the unit of time in 1968 with only one abstention, the
representative of the Greenwich Observatory, I regret to say. It had in fact
been used since 1955 although the astronomical second was still the official
unit. The UK standard frequency transmissions were controlled by the NPL and we
had the confidence to express and publish their values in terms of the atomic
unit. Atomic time had thus been available throughout the world and we found
that it was used to correct international time signal, thus disposing of the
argument that it could not be used to measure time. There was one problem which
had to be solved before atomic time could be universally adopted. Some users
such as those at sea with rather simple equipment still navigated by the stars
and required astronomical time even if it was not uniform. A suggestion[RE1]
from the US Time Service was that astronomical time should be used for sea navigation [RE2]and
domestic purposes, and atomic time for air navigation and scientific work. My
experiences with time signals and standard frequency transmissions convinced me
that this would cause endless confusion as well as involving duplication of
equipment and I argued strongly that a method of combining all the information
in one set of transmissions must be found. The main difficulty was that,
although the two time scales could be synchronised to start with, they would gradually drift apart because of the
variations in the rate of rotation of the earth.
An excellent
compromise was achieved when the astronomers agreed that time signals could be
allowed to drift by as much as 0.5s from astronomical time. Atomic seconds
could be transmitted continuously and if, after a year say, they were found to
be approaching this difference the markers denoting the hour signal could be
moved along by 1s, without disturbing the continuity of the atomic seconds. A
record kept of such step adjustments or leap seconds can be used to give long
intervals of atomic time which may be used to measure the periodicities of the
bodies of the solar system. The transition to atomic time was made easier by
the fact that at first only the UK and the USA were involved, and we agreed to
co-ordinate and synchronise all our time signals. As other stations joined in
they synchronised their signals with the existing ones and a worldwide system
of atomic time was established without any more formalities.
Atomic clocks were
improved and several models at national laboratories including the NPL and the
Hewlett Packard commercial model were accurate to 1 part in 1012,
10,000 times more accurate than astronomical time, and time signals instead of
needing corrections published a year in arrears were immediately available with
an accuracy of one microsecond. But scientists as well as the lay public found
it difficult to disassociate time keeping from Observatories and seemed unaware
of the revolution that had occurred. The Observatories acquired the commercial
model and transmitted the signals as before. My contribution was acknowledged
by the award of the Tompion Gold medal from the Worshipful Company of
Clockmakers and the Popov Gold Medal from the USSR Academy of Sciences who were
splendid hosts when I went to receive it in Moscow.
During the
discussion on the adoption of an atomic unit the US delegate strongly advocated
the use of the hydrogen atom instead of caesium. The choice finally fell to
caesium because it was shown that the clock using hydrogen was more influenced
by the conditions of operation. However the hydrogen maser, the brainchild of
NF Ramsey at Harvard University, is a wonderful piece of equipment. Hydrogen
atoms in the higher energy state are selected by their deflection in a magnetic
field and directed into a spherical quartz bottle, the internal walls of which
were coated with Teflon. As with the caesium atoms a narrow spectral line is
obtained only if the transition from one state to the other occurs slowly.
Normally an impact with another atom or with the wall of the vessel would cause
a transition, but the Teflon coating enabled thousands of wall collisions to
occur before this happened. The bottle was enclosed in a cavity resonator tuned
to the spectral line frequency. The energy was thereby amplified sufficiently
to maintain the circuit in oscillation. It was thus similar to the ammonia
maser but far superior in frequency stability.
The hydrogen maser
The normal
retiring age in the civil service was sixty and the government at the time was
reluctant to allow any extensions as it was keen to reduce the number of civil
servants. I was fortunate, therefore, to be allowed to stay rill the age of
sixty-two and then, on a part-time basis, for a further two years. This was a
thoroughly enjoyable period for me as I had even less administrative
responsibilities than before and I was able to investigate some conflicting
result that had been obtained with the hydrogen maser. A colleague, M Bangham,
having done well to get the maser working and indeed to demonstrate it at the
Physical Society, was busy working on an improved model. I decided to use the
old one to study the discrepancies in the frequency values published by various
people in the USA. I found that the setting of the cavity resonator to the
hydrogen frequency was more difficult than had been thought but that if I took the
average of 20 settings the results were consistent to 1 part in 1012.
Another colleague, R Donaldson, became very expert at coating the bulbs with
Teflon and also with a similar material, Fluon, made by the ICI, which proved
to be just as good. A large number of bulbs were tested under different
conditions over a period of two years giving reliable results for the effect of
the coating material and size of bulb. It seemed to me that if the setting
problem could be overcome the hydrogen maser would be a worthy contender with
the caesium clock as the fundamental standard of time. It may be that with the
absorption of caesium as the standard interest in alternatives waned, since I
have not heard of any further developments.
It is rather
unfortunate that my work sometimes did not add to my popularity with the
scientific establishment. They did not like being proved wrong on the velocity
of light and the atomic clock was a blow to the astronomers, naturally enough
since it removed one of their main functions – the accurate determination of
time. Comments which I made from time to time at international meetings on
errors in the relativity theory made me still less popular.
