Every clock does the same job: it counts something that repeats at a steady rate, then turns that count into hours, minutes and seconds. A pendulum swings, a quartz crystal flexes, and an atom absorbs microwave energy at one exact frequency and no other. Understanding how atomic clocks work comes down to that last part, because what is being counted decides how good a clock can ever be.
This article explains what is actually counted, the caesium-133 hyperfine transition that now defines the second itself, why atoms outperform pendulums and quartz oscillators, and the accuracy a modern atomic clock reaches. If you would rather see the result than the mechanism, the atomic time display shows server-synchronised time alongside the live drift of your own device in milliseconds.
How Atomic Clocks Work: What Is Being Counted?
They count oscillations of microwave radiation, not the movement of any physical part. The atom never ticks. It serves as a reference, absorbing energy at exactly one frequency, and the electronics tune an oscillator until it matches.
That frequency comes from a property called a hyperfine transition. Both the electron and the nucleus behave like tiny magnets, and in certain elements the lowest energy state splits into two levels depending on whether those magnetic moments are aligned or opposed. An atom crosses the gap between them only when struck by radiation of precisely the right frequency. Nothing about that gap is manufactured: every caesium-133 atom in existence has the identical one, so there is no tolerance to hold and nothing to wear out.
The Caesium-133 Hyperfine Transition
Since 1967 the SI second has been defined by caesium-133. One second is the duration of 9,192,631,770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of that atom. The figure looks arbitrary because in a sense it is: it was chosen to match the astronomical second already in use, so that redefining the unit would not visibly shift the world’s clocks. Count 9,192,631,770 cycles of that microwave signal and one second has passed, by definition. There is no more fundamental ruler to check the result against.
Caesium was picked for practical reasons as much as theoretical ones. It has a single outer electron, so its hyperfine structure is simple to read; it melts a little above room temperature, so producing a beam of atoms is easy; and its transition sits in the microwave band, mature engineering by the 1950s. Rubidium and hydrogen have useful transitions too, but caesium is the atom the unit is anchored to.
How Atomic Clocks Work, Step by Step
A working atomic clock is a feedback loop. A quartz oscillator generates a microwave signal, that signal is fired at a sample of caesium atoms, a detector measures how many atoms changed state, and the result steers the oscillator back onto frequency.
In a classical caesium beam standard the stages run in this order:
- Source: caesium is warmed in a small oven, producing a thin beam of atoms travelling down an evacuated tube.
- State selection: a magnet or a laser strips out atoms already in the wrong hyperfine level, so the beam enters the cavity in a known state.
- Interrogation: the beam crosses a microwave cavity fed by the quartz oscillator, and atoms flip state only if the frequency is right.
- Detection: a second selector and a detector count how many atoms actually flipped.
- Correction: the count peaks at the true transition frequency, so the electronics nudge the oscillator until it sits on that peak, then hold it there.
The seconds you eventually read are produced by dividing down that disciplined oscillator, exactly as in a quartz watch. The atoms supply no motion at all. They answer one question, over and over: is the oscillator fast or slow?
The Caesium Fountain
Modern primary standards such as NIST-F2 in Boulder, Colorado use a fountain rather than a horizontal beam. Lasers cool a cloud of caesium atoms to a fraction of a degree above absolute zero, then toss it upward through a microwave cavity; the cloud rises, slows, and falls back through the same cavity. Cold, slow atoms linger in the field far longer than a hot beam can, and a longer interrogation gives a sharper reading of the transition frequency, much as a long held musical note is easier to tune than a short one. Every primary atomic clock in service today uses this arrangement.
Why Atoms Beat Pendulums and Quartz
Mechanical and quartz clocks depend on the properties of a manufactured object, and manufactured objects vary, age and react to their surroundings. An atomic clock avoids all of that.
- Pendulum clocks: the finest kept time to about a second a day, but the rod lengthens with temperature and the swing responds to air pressure, altitude and vibration in the building.
- Quartz oscillators: a watch crystal cut to vibrate 32,768 times a second is typically held to around twenty parts per million, which is a second or two of quartz drift a day, and its rate shifts with temperature in a shallow curve.
- Atomic clocks: the caesium transition is identical in every atom, immune to ageing, and residual temperature effects can be measured and subtracted rather than merely tolerated.
- Reproducibility: two independently built caesium standards agree without ever having been compared, which is precisely what makes a single international timescale possible.
Quartz drift is less a defect than a property of the material, and temperature effects dominate it: a crystal that is perfect on a desk runs differently in a warm server room. Atoms sidestep the problem, which is why the devices around you still keep quartz for the ticking and take their truth from a network signal. That division of labour is the subject of why computer clocks drift.
How Accurate Is an Atomic Clock?
The accuracy of an atomic clock is quoted as a fraction: how far its rate may differ from the true second. A good commercial caesium standard holds about one part in a million million, roughly a second in thirty thousand years.
How well an atomic clock performs depends chiefly on how long the atoms can be watched, which is why fountains beat beams and why cold atoms beat hot ones. A caesium fountain such as NIST-F2 is characterised at around one part in ten thousand million million, so it would neither gain nor lose a second over hundreds of millions of years. Optical lattice clocks, which interrogate strontium or ytterbium with laser light at hundreds of terahertz rather than microwaves, push that a further hundredfold. The three families are compared in caesium, rubidium and optical clocks.
From the Laboratory to Your Screen
No single instrument defines the time you see. National laboratories run ensembles of clocks and send their measurements to the BIPM outside Paris, which combines several hundred of them into International Atomic Time, or TAI. Coordinated Universal Time is that scale with an integer number of leap seconds applied so it keeps step with Earth’s rotation, and UTC is what reaches your devices over the Network Time Protocol, over longwave radio, and from the atomic clocks aboard GPS satellites. The institutions behind it are described in national time standards. You can read the plain figure on the current UTC clock, or set the scales side by side on the TAI and GPS time comparison.
Conclusion
How atomic clocks work is, at heart, a simple arrangement: a quartz oscillator does the counting, and a cloud of caesium-133 atoms says whenever it is wrong. Because the 9,192,631,770-cycle hyperfine transition is a property of nature rather than of a machined part, the same correction is available in every laboratory on Earth, and it never drifts or wears out. That is why the second was handed to caesium in 1967, and why every accurate clock you own is ultimately traceable to one. To see atomic time applied to your own device, open the live atomic clock and check your drift, or start from the atomicclock.now homepage.