The Hidden Science of Timekeeping: How We Measure a Second (and Why It Keeps Changing)
A second feels obvious—until you ask what it is. Inside atomic physics, global coordination, and the leap-second compromise that keeps clocks friendly to the sky.
By TheMurrow Editorial
February 2, 2026
Key Points
- 1Replace the sky with physics: the SI second is fixed by caesium-133, not Earth’s irregular rotation.
- 2Separate definitions from policies: the SI second stays stable, while UTC and leap seconds remain contested civil-time rules.
- 3Rely on global coordination: BIPM combines ~85 labs and ~450 clocks, using GNSS/TWSTFT links to build UTC and TAI.
A second feels simple—until you try to define it
A second feels like the simplest thing in the world—an unremarkable tick on a microwave clock, a heartbeat, the pause before you answer a hard question. Yet the second is also one of humanity’s most carefully engineered agreements: a unit of time defined not by the sky above us, but by the physics inside an atom.
That choice wasn’t aesthetic. It was a concession to reality. Earth does not rotate like a perfectly machined wheel. It wobbles, speeds up, slows down—subtly, unpredictably, and for reasons ranging from ocean tides to earthquakes. Civilizations once built “time” on the Sun’s rhythm; modern technology can’t afford the drift.
So when people ask why “the second” keeps changing, they’re often mixing up two different stories: the stable scientific definition of the SI second, and the messier politics of civil timekeeping—UTC, leap seconds, and what it means to keep noon near the middle of the day.
That choice wasn’t aesthetic. It was a concession to reality. Earth does not rotate like a perfectly machined wheel. It wobbles, speeds up, slows down—subtly, unpredictably, and for reasons ranging from ocean tides to earthquakes. Civilizations once built “time” on the Sun’s rhythm; modern technology can’t afford the drift.
So when people ask why “the second” keeps changing, they’re often mixing up two different stories: the stable scientific definition of the SI second, and the messier politics of civil timekeeping—UTC, leap seconds, and what it means to keep noon near the middle of the day.
“The SI second is designed to be the same everywhere. Civil time is designed to stay friendly to the sky.”
— — TheMurrow Editorial
The second isn’t “1/86,400 of a day” anymore—and that’s the point
For centuries, timekeeping treated a day as fundamental: divide the mean solar day into 86,400 parts and you have a second. The Bureau International des Poids et Mesures (BIPM) notes that before 1960, the second was defined as 1/86,400 of the mean solar day—a definition rooted in astronomy and everyday experience.
Earth, unfortunately, refuses to behave like a metrologist’s dream. Its rotation is irregular, influenced by tides, atmospheric circulation, earthquakes, and interactions between the core and mantle. Those changes are tiny but consequential. Precision science, navigation, telecommunications, and financial systems don’t just want “roughly right.” They want reproducible time—time that can be rebuilt in any lab, in any country, without asking the planet how it feels today.
Earth, unfortunately, refuses to behave like a metrologist’s dream. Its rotation is irregular, influenced by tides, atmospheric circulation, earthquakes, and interactions between the core and mantle. Those changes are tiny but consequential. Precision science, navigation, telecommunications, and financial systems don’t just want “roughly right.” They want reproducible time—time that can be rebuilt in any lab, in any country, without asking the planet how it feels today.
The 1960 stopgap: the ephemeris second
In 1960, the 11th CGPM adopted an “ephemeris” definition of the second based on the tropical year 1900—an astronomical workaround that leaned on orbital mechanics rather than Earth’s variable rotation. BIPM’s history of the second describes it as a response to the inadequacy of Earth-rotation time for precision measurement.
It didn’t last long. Atomic standards proved more reproducible and far more practical for building clocks that could keep pace with modern needs.
It didn’t last long. Atomic standards proved more reproducible and far more practical for building clocks that could keep pace with modern needs.
What it means for readers
A time unit tied to an atom, not the sky, sounds abstract. Yet the implication is concrete: the second is intended to be a universal reference. Whether you measure it in Paris, Tokyo, or on a spacecraft, it should be the same unit—because it is defined by an invariant property of nature rather than a moving planet.
1/86,400
Before 1960, the second was historically defined as 1/86,400 of the mean solar day—until Earth’s irregular rotation made that unsuitable for precision.
The modern SI second: 9,192,631,770 counts of caesium
The modern second is not a fraction of a day. It’s a count.
In 1967–1968, the 13th CGPM defined the second as the duration of 9,192,631,770 periods of radiation corresponding to the transition between two hyperfine levels of the ground state of caesium-133. That number is not a poetic flourish; it’s a specification precise enough to anchor the world’s measurement system.
In 1967–1968, the 13th CGPM defined the second as the duration of 9,192,631,770 periods of radiation corresponding to the transition between two hyperfine levels of the ground state of caesium-133. That number is not a poetic flourish; it’s a specification precise enough to anchor the world’s measurement system.
“A second is not a slice of the day. It’s a specified number of oscillations in caesium-133.”
— — TheMurrow Editorial
The 2019 reframing: constants, not artifacts
The SI underwent a conceptual modernization in 2018, taking effect 20 May 2019. Instead of defining units by physical artifacts or procedural descriptions, the 26th CGPM reframed them around fixed numerical values of defining constants.
For time, the substance stayed the same but the style changed: the second is defined by fixing the value of the caesium hyperfine transition frequency at ΔνCs = 9,192,631,770 Hz. The definition reads like physics because it is physics.
For time, the substance stayed the same but the style changed: the second is defined by fixing the value of the caesium hyperfine transition frequency at ΔνCs = 9,192,631,770 Hz. The definition reads like physics because it is physics.
9,192,631,770
The SI second is anchored to 9,192,631,770 periods of radiation from the caesium-133 hyperfine transition—a reproducible definition independent of Earth’s rotation.
The ideal caesium atom doesn’t exist in a lab
Precision comes with an asterisk. In 1997, the CIPM clarified that the definition refers to a caesium atom at rest at 0 K—an idealized “unperturbed” condition. Real clocks operate in the real world, where temperature, motion, electromagnetic fields, and engineering constraints shift the measured frequency.
Metrology labs therefore build clocks that approximate the ideal and apply corrections for known perturbations. That is not a weakness of the definition; it’s the work of making a definition real.
Metrology labs therefore build clocks that approximate the ideal and apply corrections for known perturbations. That is not a weakness of the definition; it’s the work of making a definition real.
Key Insight
Atomic time definitions aim for universality; real-world clocks aim for realization—approximating the ideal atom and correcting for disturbances like temperature, motion, and fields.
The confusion: the SI second is stable, civil time is contentious
Search engines are filled with anxious questions: Is the second changing? Are we redefining time again? Most of the time, the worry is misdirected.
Two different “changings” get tangled together:
- The definition of the SI second (rarely updated, and currently stable).
- Civil timekeeping rules (UTC, leap seconds, how close civil time stays to Earth rotation), which are frequently debated and operationally difficult.
The SI second is a metrology choice: build a unit that is reproducible and invariant. Civil time is a societal choice: decide how to label moments so that clocks remain aligned with the rotating Earth—at least loosely.
Two different “changings” get tangled together:
- The definition of the SI second (rarely updated, and currently stable).
- Civil timekeeping rules (UTC, leap seconds, how close civil time stays to Earth rotation), which are frequently debated and operationally difficult.
The SI second is a metrology choice: build a unit that is reproducible and invariant. Civil time is a societal choice: decide how to label moments so that clocks remain aligned with the rotating Earth—at least loosely.
Why that distinction matters
If you run a network, build GPS-dependent systems, manage high-frequency trading infrastructure, or schedule satellites, “time” means more than “what your phone says.” It means a technical timescale with defined behavior.
The SI second provides the stable “tick.” Civil time decides how to group those ticks into calendar time that humans recognize. Confusing the two leads to needless panic—and bad decisions.
The SI second provides the stable “tick.” Civil time decides how to group those ticks into calendar time that humans recognize. Confusing the two leads to needless panic—and bad decisions.
“Precision timekeeping isn’t a single clock. It’s a treaty between physics and society.”
— — TheMurrow Editorial
SI second vs. civil timekeeping
Before
- SI second (physics-defined
- invariant)
- reproducible anywhere
- rarely updated
After
- Civil time (UTC rules)
- leap seconds
- debated policies to keep clocks aligned with Earth rotation
How the world builds time: TAI, UTC, and the quiet authority of BIPM
Even with a clean definition of the second, someone has to realize it—turn it into working time you can use. That work is coordinated by the BIPM, which describes a global system drawing on contributions from roughly 85 laboratories maintaining around 450 atomic clocks worldwide for UTC computation.
That scale matters. No single clock is perfect. The modern approach is statistical and cooperative: combine many clocks, weight their performance, and compute a timescale more stable than any individual device.
That scale matters. No single clock is perfect. The modern approach is statistical and cooperative: combine many clocks, weight their performance, and compute a timescale more stable than any individual device.
85 laboratories
BIPM coordinates UTC computation using contributions from roughly 85 labs worldwide—an international ensemble rather than a single “master clock.”
450 atomic clocks
UTC computation draws on around 450 atomic clocks globally, combined statistically to produce a timescale more stable than any one device.
TAI: the continuous atomic backbone
International Atomic Time (TAI) is the continuous timescale built from those atomic clocks. TAI does not stop for leap seconds; it just runs. For engineers and scientists, that continuity is a feature, not a bug.
UTC: civil time, tied (loosely) to Earth
Coordinated Universal Time (UTC) runs at the same rate as TAI but differs by an integer number of seconds because of leap seconds. Leap seconds exist to keep UTC close to UT1, a time derived from Earth’s rotation.
BIPM’s description is crisp: UTC is the global civil reference time, while TAI is continuous atomic time. Their difference is not fractional—UTC is offset from TAI by whole seconds.
BIPM’s description is crisp: UTC is the global civil reference time, while TAI is continuous atomic time. Their difference is not fractional—UTC is offset from TAI by whole seconds.
UTC(k): what your country actually maintains
Most nations do not broadcast “UTC” directly. They maintain their own real-time realization: UTC(k), where k identifies the lab (for example, UTC(NIST) in the United States). BIPM then publishes the offset UTC – UTC(k) at 5-day intervals in Circular T.
For readers outside time metrology, Circular T is a reminder that “the time” is not a single object sitting in a vault. It’s a continuously maintained international alignment problem.
For readers outside time metrology, Circular T is a reminder that “the time” is not a single object sitting in a vault. It’s a continuously maintained international alignment problem.
The global timekeeping stack (in practice)
- ✓Define the second via caesium-133 (SI)
- ✓Realize it with national and metrology-lab atomic clocks
- ✓Combine clock data into TAI (continuous)
- ✓Derive UTC from TAI plus leap seconds (civil reference)
- ✓Maintain UTC(k) locally and publish UTC–UTC(k) in Circular T
Time transfer: comparing clocks across continents without moving them
A global timescale requires global comparison. You cannot fly a clock from Paris to Tokyo every day and hope to keep modern networks synchronized. Instead, labs compare clocks using time transfer—techniques that relate distant clocks through signals and careful calibration.
BIPM lists several key methods used in UTC computation:
- GNSS-based methods such as GPS common-view, carrier-phase, and all-in-view
- Two-way satellite time and frequency transfer (TWSTFT)
- Increasing use of new systems and links, including Galileo
BIPM lists several key methods used in UTC computation:
- GNSS-based methods such as GPS common-view, carrier-phase, and all-in-view
- Two-way satellite time and frequency transfer (TWSTFT)
- Increasing use of new systems and links, including Galileo
Galileo joins the official toolkit
In a detail that signals where the field is heading, BIPM noted the first official use of Galileo in UTC computation in Circular T no. 425 (June 2023). The expansion beyond GPS is not a branding exercise; it’s resilience and precision.
More signals, more paths, and better measurement techniques improve the reliability of comparisons—especially as clocks themselves become more accurate and demand better transfer infrastructure.
More signals, more paths, and better measurement techniques improve the reliability of comparisons—especially as clocks themselves become more accurate and demand better transfer infrastructure.
Practical implication: the bottleneck moves
As atomic and future optical clocks improve, the limiting factor often becomes comparison: how well labs can transfer time between one another. Better clocks stress the pipelines that stitch the world together.
For anyone building systems that depend on precise timing—telecom synchronization, power grids, satellite operations—the message is clear: time is only as good as your ability to distribute it.
For anyone building systems that depend on precise timing—telecom synchronization, power grids, satellite operations—the message is clear: time is only as good as your ability to distribute it.
Key takeaway: precision is a network property
Even with world-class clocks, global time depends on comparison and distribution. As clocks improve, time-transfer links increasingly determine real-world timing performance.
Leap seconds: the one-second fix that can cause outsized trouble
Leap seconds are small in magnitude and large in consequence. UTC is kept within ±0.9 seconds of UT1 (Earth rotation time). When Earth’s rotation drifts far enough, timekeepers add a leap second to UTC to pull it back toward UT1.
Operationally, that means a minute with 61 seconds—a deeply unfriendly concept for computers and systems designed around uniform time steps.
Operationally, that means a minute with 61 seconds—a deeply unfriendly concept for computers and systems designed around uniform time steps.
Why leap seconds exist
Leap seconds are a compromise between two values:
- Keep civil time aligned with the sky (so noon remains near when the Sun is highest).
- Maintain a uniform time scale for technology.
Without leap seconds, UTC would gradually drift away from UT1. With leap seconds, UTC remains tied to Earth’s rotation, but systems must cope with discontinuities.
- Keep civil time aligned with the sky (so noon remains near when the Sun is highest).
- Maintain a uniform time scale for technology.
Without leap seconds, UTC would gradually drift away from UT1. With leap seconds, UTC remains tied to Earth’s rotation, but systems must cope with discontinuities.
Why engineers dislike them
Leap seconds can break assumptions in software, databases, distributed systems, and logging infrastructure. Many systems are designed to assume time monotonically increases in uniform steps; a repeated second is an edge case that must be handled carefully.
Real-world case studies exist across the industry, but the core point doesn’t require brand-name anecdotes: the leap second is the rare event that tests every assumption you didn’t know your code was making.
Real-world case studies exist across the industry, but the core point doesn’t require brand-name anecdotes: the leap second is the rare event that tests every assumption you didn’t know your code was making.
Multiple perspectives, same underlying tension
Astronomers and those concerned with Earth orientation value keeping civil time close to UT1. Many technologists argue that discontinuities impose risk disproportionate to their benefit.
Both sides have a coherent position because they optimize for different things: the sky’s meaning versus the network’s stability.
Both sides have a coherent position because they optimize for different things: the sky’s meaning versus the network’s stability.
Leap seconds in civil timekeeping
Pros
- +Keeps UTC close to UT1
- +preserves a sky-linked meaning of noon
- +supports Earth-orientation needs
Cons
- -Introduces discontinuities
- -breaks software assumptions
- -complicates distributed systems and logging
±0.9 seconds
UTC is maintained within ±0.9 seconds of UT1; leap seconds are added when Earth’s rotation drifts far enough to threaten that bound.
What changes next: not the second (yet), but how we manage global time
Readers often sense that “something is changing” because timekeeping is under active discussion. The more accurate statement is that civil time rules face pressure, and the metrological community is also considering future directions.
The SI second is stable—optical clocks are the long-term question
The SI second’s caesium definition is not updated frequently. The research record points to a likely future debate: whether to redefine the second using optical transitions rather than microwave caesium. Optical clocks promise higher frequencies and potentially better precision.
No date in the provided record commits the world to an optical redefinition. The more responsible framing is that the existing definition remains the backbone, while technology evolves toward clocks whose performance will eventually challenge how the SI second is realized and compared.
No date in the provided record commits the world to an optical redefinition. The more responsible framing is that the existing definition remains the backbone, while technology evolves toward clocks whose performance will eventually challenge how the SI second is realized and compared.
Timekeeping infrastructure will keep evolving
What will change sooner, and more visibly, is the infrastructure: how UTC is computed, which satellite systems contribute to time transfer, and how national labs maintain UTC(k). The inclusion of Galileo in official UTC computation in 2023 is a concrete example of that evolution already underway.
For readers, the practical takeaway is not that your wristwatch will break. The takeaway is that the global systems you rely on—navigation, communications, finance—depend on a continuously maintained international process that must keep adapting as accuracy improves.
For readers, the practical takeaway is not that your wristwatch will break. The takeaway is that the global systems you rely on—navigation, communications, finance—depend on a continuously maintained international process that must keep adapting as accuracy improves.
A second is physics; “the time” is coordination
The second has become a statement of intent: time defined by physics rather than a fickle planet. The real drama now plays out elsewhere—in the coordination of hundreds of clocks, in the satellite links that compare them, and in the uneasy compromise between atomic regularity and an Earth that refuses to spin on schedule.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering science.
Frequently Asked Questions
Is a second still 1/86,400 of a day?
No. Before 1960, that was the historical approach: define the second as 1/86,400 of the mean solar day. Earth’s rotation turned out to be irregular, making that definition unsuitable for precision. The SI second is now defined using a fixed atomic frequency tied to caesium-133, making it reproducible anywhere.
What exactly is the SI definition of the second today?
The SI second is defined by fixing the value of the caesium hyperfine transition frequency at ΔνCs = 9,192,631,770 Hz (formalized in the 2018 SI revision, effective 20 May 2019). Historically, the definition is expressed as 9,192,631,770 periods of radiation for that transition.
If the second is defined by caesium, why do we need many clocks?
No real clock is perfectly isolated from temperature, motion, or electromagnetic effects. Metrology uses ensembles of clocks because combining many independent sources produces a more stable timescale than any single device. BIPM describes UTC computation drawing on about 85 laboratories and around 450 atomic clocks worldwide.
What’s the difference between TAI and UTC?
TAI (International Atomic Time) is a continuous atomic timescale. UTC (Coordinated Universal Time) runs at the same rate as TAI but differs by a whole number of seconds due to leap seconds, which are added to keep UTC close to Earth-rotation time (UT1). TAI is continuous; UTC is civil time with occasional discontinuities.
What is UTC(k), and why does it matter?
UTC(k) is a country or lab’s real-time realization of UTC (for example, UTC(NIST)). Because UTC itself is computed and published with a lag, national labs maintain UTC(k) locally and compare it to UTC. BIPM publishes offsets UTC – UTC(k) in Circular T at 5-day intervals, allowing global alignment.
How do laboratories compare clocks across long distances?
Labs use time transfer methods, including GNSS techniques (GPS common-view, carrier-phase, all-in-view) and two-way satellite time and frequency transfer (TWSTFT). BIPM also reports expanding use of other satellite systems; notably, Galileo was first used officially in UTC computation in Circular T no. 425 (June 2023).
