Two kinds of time
There are two ways to measure a day:
- Astronomical time follows the Earth’s rotation. The scale used for this is called UT1. It is what the sun and stars actually do.
- Atomic time follows the vibrations of atoms in atomic clocks. International Atomic Time (TAI) is computed from hundreds of clocks around the world and is extremely steady.
The SI second has been defined by the cesium atom since 1967. The Earth’s rotation, however, is not perfectly regular. It speeds up and slows down slightly because of tides, movements in the Earth’s core, the atmosphere and oceans, and other effects. Over time, atomic time and the Earth drift apart.
What a leap second does
Coordinated Universal Time (UTC), the time standard the world’s clocks are based on, runs at the atomic rate but is kept within 0.9 seconds of UT1. When the gap approaches that limit, one extra second is added to UTC. That extra second is the leap second.
A positive leap second is added as 23:59:60 UTC at the end of the chosen day:
| UTC time | Note |
|---|---|
| 23:59:58 | normal |
| 23:59:59 | normal |
| 23:59:60 | the leap second |
| 00:00:00 | the next day begins |
In theory a negative leap second is also possible, in which 23:59:59 would be skipped, but one has never been needed.
Who decides
The International Earth Rotation and Reference Systems Service (IERS) monitors the Earth’s rotation. It publishes a notice called Bulletin C twice a year, announcing whether a leap second will be added at the end of the next 30 June or 31 December. Those two dates are the preferred options, which gives about six months’ warning.
A short history
UTC with leap seconds began in 1972. At that point UTC was set 10 seconds behind TAI. Between 1972 and 2016, 27 leap seconds were added, some years having two and many years having none. The 27th was added at the end of 31 December 2016, bringing the difference between TAI and UTC to 37 seconds.
Leap seconds were added fairly often in the 1970s, then less frequently over time. In recent years the Earth’s rotation has been slightly faster on average, which reduced the need for new ones.
Why leap seconds cause problems
Most computer systems assume every minute has 60 seconds. A 61-second minute can confuse software that sorts events by time, calculates durations or expects time never to repeat. Leap seconds have caused outages in the past, including problems on some websites and Linux servers after the leap second of 30 June 2012.
To avoid this, some large technology companies use a leap smear: instead of adding one second at once, their time servers slow their clocks very slightly over many hours, so no clock ever shows 23:59:60. The result is that, during the smear, different systems may disagree by a fraction of a second.
The end of leap seconds
Because of these problems, the General Conference on Weights and Measures decided in 2022 that the tolerance between UTC and UT1 should be increased by or before 2035. In practice this means leap seconds will be phased out, and UTC will be allowed to drift further from the Earth’s rotation. How a much larger correction might eventually be handled, perhaps decades or centuries from now, has not yet been decided.
Does it affect you?
For everyday life, no. One second is far too small to notice on a wall clock, and phones get the correct time automatically. Leap seconds matter for navigation satellites, telecommunications, financial trading, astronomy and any system that records precise timestamps.
To see how UTC relates to the time in London, read UTC vs GMT. For the calendar version of the same idea, see what is a leap year. You can check the current UTC time on the GMT page.