Today’s number is 9,192,631,770.

That’s how many times a cesium-133 atom oscillates between two specific energy states in one second. Or, more precisely, it’s how many times we decided a cesium atom does that in one second. The number is doing some heavy lifting that the universe didn’t ask for.

This post is about the number, how it got picked, and how we use it to measure time.

The chain

To know what a second is, you need a way to measure one. For most of human history, the answer was: “look at Earth and count” or earth++ for the programmers reading this.

One second = 1/86,400 of a mean solar day.

That worked for sundials and pendulum clocks. It stopped working in the 1930s, when good quartz clocks revealed that Earth’s rotation isn’t uniform. Yesterday, we talked about how. Earth’s rotation is not a good measurement of time. I mean good in the sense of precise, because there are so many variables that go into the rotation and impact it. We don’t have an accurate way to measure and predict how those variables change.

In 1960, astronomers tried again. Because why not? Wheel, reinvented.

They picked Earth’s orbit instead of its rotation, which is far more stable:

One second = 1/31,556,925.9747 of the tropical year for 1900.

That weird denominator was chosen specifically so the new “ephemeris second” came out the same length as the old mean solar second. Don’t break clocks … like ever.

Great in principle, terrible in practice. To know what time it was, you had to consult a 60-year-old astronomical table. It doesn’t work with the clock in the lab. The ephemeris second won the philosophy and lost the engineering.

But wait, maybe there’s a third option.

How I wish it was Cesium-123 and not Cesium-133, but alas. The 133 is the mass number: 55 protons + 78 neutrons. Cesium-133 also happens to be the only stable isotope of cesium, so it kind of picked itself.

The cesium handoff

Cesium-133 has two slightly different ground-state energies. Drop an atom from one to the other and it spits out a photon at a specific frequency, call it f. If you can build a clock that locks itself to that frequency, you have a timekeeper that doesn’t depend on the Earth, the Sun, or any astronomer’s table. Every cesium-133 atom in the universe agrees on f to absurd precision, because quantum mechanics doesn’t have local variants.

The number for f is approximately 9.192631770 GHz. So if you count 9,192,631,770 oscillations, exactly one second has elapsed.

In 1967, the General Conference on Weights and Measures (the CGPM, the body that gets to define units) voted to make that the official definition. The SI second became:

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 the cesium-133 atom.

No Earth, no Sun, no tables. Just an atom.

Why that exact number

The number 9,192,631,770 is not a fundamental constant of the universe. It was measured. By two specific guys. In 1955.

Their names were Louis Essen and Jack Parry, and they worked at the National Physical Laboratory in the UK. They had just built the first cesium clock that actually worked. The question was: how many cesium ticks fit in one ephemeris second? They measured it carefully. Their answer, published in 1958 with William Markowitz and Robert Hall:

1 ephemeris second = 9,192,631,770 ± 20 cesium periods.

Nine years later, the CGPM took that number, dropped the uncertainty, and made it the definition. From then on, the second was defined as exactly 9,192,631,770 ticks. The plus-or-minus twenty disappeared.

This means three things:

1. The cesium second was chosen so it would equal the ephemeris second.
2. The ephemeris second was chosen so it would equal the mean solar second.
3. The mean solar second was 1/86,400 of an Earth day.

The three layers of don’t break the previous standard. If Essen and Parry had been off by one in their measurement, the entire SI second would be slightly different today, and every clock and computer and GPS satellite would be calibrated to that other version.

The number is a historical fingerprint, not a physical constant.

It records, with eleven decimal-digit precision, exactly how well two physicists in 1955 could compare a cesium clock to an Earth-orbit calculation.

Every GPS satellite, every NTP server, every timestamp on every photo you’ve ever taken, they’re all calibrated to the limits of what Essen and Parry pulled off with 1955 equipment. The universe doesn’t care about your iPhone. Your iPhone cares about two guys at the NPL.

How precise are we now

Modern cesium fountains, the latest generation, where atoms are laser-cooled and tossed gently upward through a microwave cavity, falling back down under gravity like a slow ballistic juggling act, hit accuracies of about one part in 10¹⁶. That’s roughly one second of drift per 300 million years.

NIST-F2 in Boulder, the U.S. primary clock, is one of these. About a dozen others sit in metrology labs across France, Germany, the UK, Japan, and China. They all report their measurements to the BIPM in Paris, which combines them into a weighted average called TAI, International Atomic Time. TAI is what your phone’s clock is ultimately disciplined to, through a long chain of NTP servers and GPS signals.

Everything you do that involves time:

• sending a message,
• taking a photo with a timestamp,
• syncing with a calendar,
• getting a stock trade priced

It all traces back to a few hundred atomic clocks averaging each other in real time.

What comes next

Cesium won’t be the way we measure the second forever. It’s about to be deposed.

Optical lattice clocks, using strontium or ytterbium atoms instead of cesium, operating at visible-light frequencies instead of microwave, are now about a hundred times more precise. They lose one second per about 30 billion years, which is roughly twice the age of the universe.

The BIPM is planning to redefine the SI second around an optical transition, possibly by 2030.

When that happens:

1. Mean solar second = 1/86,400 of an Earth day
2. Ephemeris second chosen to match the mean solar second
3. Cesium second chosen to match the ephemeris second
4. Optical second chosen to match the cesium second ← the new one

Three layers of backwards-compatibility will become four.

Tomorrow we’ll look at those optical clocks, what they are, how they work, and how they let us measure what cesium can’t.

Sources

• Louis Essen — Wikipedia
• William Markowitz — Wikipedia
• Second — Wikipedia
• Ephemeris time — Wikipedia
• Atomic clock — Wikipedia
• International Atomic Time — Wikipedia
• NIST-F2 — Wikipedia
• Optical clock — Wikipedia

I’d appreciate a follow. You can subscribe with your email below. The emails go out once a week, or you can find me on Mastodon at @[email protected].