Leap Seconds Explained: Why Clocks Occasionally Gain a Second

Twenty-seven times since 1972, the world's clocks have been held back by exactly one second, and a minute has officially contained sixty-one of them. Leap seconds exist because the two things we call time do not agree: atomic clocks tick with extraordinary regularity, while the Earth's rotation, which defines the day, is slightly irregular and slowly slowing. Keeping the two in step requires an occasional correction, and that correction has caused more trouble in computing than almost any other timekeeping decision.

This article explains where the discrepancy comes from, who decides when a correction is applied, how large systems avoid a minute of sixty-one seconds altogether, and why an international vote has now set an end date for the practice. The current instant in Coordinated Universal Time, and your own offset from it, are shown on the local date and time page.

What Is a Leap Second?

A leap second is an extra second inserted into Coordinated Universal Time to keep it aligned with the Earth's actual rotation. When one is applied, the last minute of the chosen day runs 23:59:58, 23:59:59, 23:59:60, and only then 00:00:00 of the next day.

That 23:59:60 reading is entirely legal in ISO 8601, which explicitly permits a value of 60 in the seconds field for exactly this purpose. Very little software accepts it. The insertion is scheduled at the end of a UTC day, with 30 June and 31 December preferred, and 31 March and 30 September available as secondary options that have never yet been used.

Why Do We Need Leap Seconds?

Because the second is defined by atomic physics while the day is defined by the planet, and the planet is not a precise timekeeper. Tidal friction from the Moon gradually slows the Earth, and shorter-term variations come from the movement of the atmosphere, oceans and molten core.

Two timescales result. International Atomic Time counts seconds without regard to the sky. UT1 tracks the Earth's actual angular position, and drifts against atomic time by a fraction of a second each year. Coordinated Universal Time is the compromise: it ticks with atomic seconds, so it is suitable for engineering, but it is nudged by whole seconds so it never diverges from UT1 by more than 0.9 seconds. Without that nudging, civil noon and solar noon would drift apart, though at the observed rate the difference would take many centuries to become noticeable to anyone outside astronomy. The cumulative effect of the corrections so far is that atomic time now runs 37 seconds ahead of UTC.

Who Decides When One Happens?

The International Earth Rotation and Reference Systems Service, known as the IERS, monitors the planet's rotation and announces each correction. The announcement is published in a document called Bulletin C, usually about six months ahead of the date.

Six months is not much notice for a global infrastructure change, which is part of the complaint against the practice. The IERS cannot give more, because the Earth's rotation is not predictable far in advance. Every leap second so far has been positive, adding a second rather than removing one, and the most recent was inserted at the end of 2016. There has been no correction since, because the planet has recently been rotating slightly faster than it did through most of the twentieth century.

How Do Computers Cope With Leap Seconds?

Mostly by pretending they do not happen. POSIX time, the count of seconds behind almost every operating system, defines every day as exactly 86,400 seconds long, so there is no value available to represent the extra second at all.

• Repeat the second: the traditional approach, in which the clock reports the same value twice. Simple, but it makes time briefly non-monotonic, which breaks anything assuming timestamps only increase.
• Step the clock: jump back one second. This has caused live incidents, most memorably in 2012 when a kernel bug sent Linux servers worldwide into a spin.
• Smear the second: spread the adjustment across many hours by running the clock fractionally slow. Major cloud providers do this, typically over a window centred on the event.
• Ignore it entirely: what most consumer devices do, resynchronising with a time server afterwards and quietly correcting.

Leap smearing is now the dominant approach because it never produces an invalid timestamp and never moves the clock backwards. Its cost is that a smearing machine disagrees with an unsmeared one by up to a second during the window, which is why mixing the two within one cluster is unwise. The underlying representation is described in Unix epoch time explained, and any value can be decoded with the Unix timestamp converter.

Why Leap Seconds Are Being Retired

In November 2022 the General Conference on Weights and Measures resolved to stop inserting them by 2035 at the latest. The plan is to let Coordinated Universal Time and the Earth's rotation drift apart by a much larger tolerance, with the details of any future correction left to be agreed before then.

The argument that won was practical. Every insertion requires coordinated action across satellite navigation, telecommunications, finance and computing, the notice period is short, and the benefit is a correction that no ordinary activity depends on. Astronomers, who do depend on the Earth's orientation, can work directly with UT1. There is a further wrinkle: because the planet has lately been spinning faster, the first correction ever needed might be a negative leap second, removing 23:59:59 from a day. No software has been tested against that, and the prospect of it hastened the decision to abandon the whole mechanism.

Does Any of This Affect You?

Almost certainly not, unless you run distributed systems where sub-second ordering matters. A one-second discrepancy is invisible in a calendar appointment and irrelevant to a wall clock, and no leap second has been applied for years.

Where it does matter, the discipline is the same as for every other timekeeping hazard: keep one authoritative source of the current instant, do not assume timestamps are strictly increasing, and store instants rather than local readings, for the reasons set out in why you should store dates in UTC. If your concern is the more mundane matter of which offset your own machine reports, the UTC offset detector answers that directly, and the distinction between an offset and a zone is covered in what is my UTC offset.

Conclusion

Leap seconds are the seam between a timescale built on atomic physics and a day defined by a planet that will not keep to schedule. Twenty-seven of them have been inserted since 1972, each announced by the IERS a few months ahead, and each handled by modern infrastructure through smearing rather than a jump. With the practice due to end by 2035, the leap second is on its way to becoming a historical curiosity. See the current moment in Coordinated Universal Time and in your own zone on the local date and time page, or explore the other date and time tools on localdatetime.now.