ROGER PENROSE appears baffled by my tardiness, eyeing my sweat-beaded brow with what I can only assume is pity. “I found it very easily,” says the University of Oxford mathematician. “It’s a straight walk up from South Kensington tube.” The irony is not lost on either of us. I had not only arrived late to a meeting about time measurement; I had got lost on the streets of west London trying to find the Royal Institute of Navigation.
That was last year, at the first gathering of a committee to oversee the world’s . It was to be Penrose‘s only appearance: after half an hour he pulled out, suggesting the committee was skirting the real problem with time.
The latest leap second will occur at the end of Tuesday, 30 June. It is the first time since 2012 that the world’s time lords have decreed the clocks must stand still for a second, to ensure that time measured by Earth’s spin and time measured by atomic clocks do not get out of sync. The UK government’s consultation was intended to inform an international debate, due to come to a head later this year, about whether it should be the last time. Proponents of leap seconds argue that they represent a historical connection between astronomical motion and the passage of time that should not be broken. Opponents think the leap-second system is hopelessly muddled and not fit for purpose in our fast-paced, networked world. Not, as I was to find out, that there is much agreement on what should replace it.
Advertisement
Leap seconds are a response to the capricious way Earth spins. A day is the time Earth takes to rotate once on its axis, and the original definition of the second was 1/86,400 of that day. But a day is not always precisely the same length. The gravitational pull of moon and sun are gradually slowing down Earth’s spin, so 100 years ago, the day was a little shorter. Instabilities due to atmospheric effects and Earth’s roiling molten iron core also mean the planet’s spin can slow down or speed up in unpredictable ways.
Such inconsistencies are partly why, in the 1960s, the world embraced a new definition of the second. It was based on atomic physics: specifically, the duration of 9,192,631,770 cycles of one particular resonance of a caesium-133 atom. But that number of cycles was still chosen to correspond to the length of a second in an average sort of day in the 1950s (see “Every second counts“). Soon, as the days imperceptibly lengthened, corrections were needed to stop atomic time and Earth time drifting apart. So in 1972, the leap second system was born.
And it remains how time is made today. The astronomers of the (IERS) track Earth’s rotation using the most stringently fixed reference point they can find – not the sun these days, but light from active galaxies known as quasars billions of light years away. When a variation in Earth’s spin threatens to send this Universal Time (UT1) more than 0.9 seconds astray from the ticking of atomic time, the IERS issues a call, in a document known as “Bulletin C”, to add or subtract a leap second. So far, the order has always been to add.
The result is Coordinated Universal Time, or UTC, the standard time administered by the . And so on 30 June this year, the clocks will briefly (at least officially) . Depending on how much your local time is ahead of or behind UTC, the actual event will seem to happen earlier or later, of course. In the UK, which is currently on UTC+1 summer time, it will happen at approaching 1 am on 1 July; in Australia, it will be some time mid-morning.
This is a highly contentious fudge. Felicitas Arias heads the group at the (BIPM) in Paris that combines astronomical and atomic time to make UTC, and then passes it to the ITU for dissemination. “The present UTC system with leap second adjustments is outmoded,” she says. “But more than that, it creates difficulties on systems needing of a very precise time synchronisation.”
Certainly, technology companies are among the system’s most vocal critics. Attempts to tweak systems to deal with past leap seconds have caused a few isolated cellphone blackouts. In the first seconds of July 2012, the reservations system used by the airline Qantas that had just been added.
A glitch in time
As software becomes ever more complex and interdependent, dire warnings are multiplying of air-traffic control computers malfunctioning or military missiles misfiring. A submission to the UK Leap Seconds consultation, made in a personal capacity by , head of time services at the US Naval Research Laboratory, argued that the effects have been underplayed. “No one can tell you how widespread the issue is because most businesses and institutions do not volunteer information about mistakes,” he wrote, continuing that leap seconds have disabled GPS receivers and other similar systems. A fatal consequence of leap seconds is “unlikely but conceivable”, he concludes.
Not that this gloomy view is unchallenged – and some foresee problems if we abandon leap seconds. The Russian satellite navigation system GLONASS, for instance, takes leap seconds as a given, and would be difficult to reprogram if they were ended. And can leap seconds really pose that much of a challenge to 21st-century tech whizzes? “Leap seconds present some interesting but ultimately simple-to-solve problems,” shrugs Anthony Flavin, who is responsible for making sure the systems of the British telecoms company BT incorporate them without a hitch.
, a computer scientist at the University of Cambridge, is similarly bullish. “The case for abandoning the leap second is overstated,” he says. Plenty of countries fiddle about with time, for example by adding in daylight saving. “Computer systems and smartphones receive those revisions as part of normal security updates,” he says. “The operating system industry has long figured out how to update these things routinely.”
The first gripes about leap seconds were heard in 1999, after IERS started a discussion about their use. They have continued ever since. In 2012, the ITU leap-second group suggested getting rid of them, but couldn’t reach consensus. The group then suggested a compromise: insert a leap hour at some point in the future. That idea seems to have made things worse. “It really annoyed me,” says Kuhn, who was in on the discussions when the US first floated the idea of a leap hour in 2003.
For a start, there’s the question of who would be responsible. “The first one would be about 600 years off, which is unrealistic enough, but it would also be 3600 times more disruptive than a leap second,” he says. “The problem of inserting a UTC leap hour is probably vastly more complicated than the Gregorian calendar reform.” The US proposal, he thinks, was just an attempt to get rid of leap seconds by the back door. “It was somewhat dishonest and not carefully thought through,” he says.
In the end, ITU member countries chose to postpone a decision until this November, when the meets in Geneva, Switzerland. Dennis McCarthy, who once revelled in the titles of science adviser to the director of the Directorate of Time and head of the Earth Orientation Department at the US Naval Observatory, is not confident of a resolution. “I would hope that it will be laid to rest this year, but I suspect that it could also drag on for a while,” he says. Arias is more optimistic, but doesn’t know how the pendulum will swing. “It is difficult to make a prediction,” she says.
Should the world finally tick to the tock of only the atomic clock? (Image: Andrew Brookes, National Physical Laboratory/SPL)
At present, the US, France, Italy and Germany are in favour of ending leap seconds. The UK, Canada and Russia want to keep them. But things change: the UK used to be alone in loving the leap second; Germany and China have changed their position at least twice.
Since the ITU insists it needs consensus, we remain in leap-second limbo. In 2012, at the most recent ITU meeting, David Willetts, then the UK minister responsible for science – and therefore, at the time, time – was challenged to justify his position. He said the British public wanted to keep the traditional link between astronomy and time. He was asked what evidence he had for this. Hence the public consultation, and why both Roger Penrose and I found ourselves in South Kensington last year.
At first glance, it is hard to see why any member of the British – or any other – public should care. If we were to scrap leap seconds, by 2100 we’d be 2 to 3 minutes out of sync with the sun, and about 30 minutes out in 2700. On the first day of 2014, the sun rose on my house at just after 8 in the morning. Without any adjustment, it will eventually get closer to 9 – but only in about a millennium’s time. Is it even possible to have an opinion about this?
Yes, says Kuhn. Leap seconds were accepted in the 1970s because of a cultural assumption of a close connection between human time and Earth’s orientation in space. Nothing has changed, he says, and time’s traditional link with the sun and stars shouldn’t be dismissed just because we are told we are living in a high-tech age. “Timekeeping is an astronomical affair,” he says.
“Time’s traditional link to the sun and the stars should not be dismissed”
, a philosopher of science at the University of Bristol, UK, agrees. “There is something culturally important about recognising that the day isn’t an arbitrary unit of time,” he says. “We have this historic connection between timekeeping and astronomy and the behaviour of the Earth, and we should subordinate our timekeeping to those natural phenomena. There’s something symbolic about saying there’s a determinant of things in the universe that isn’t just the power of money. What’s the real cost – a little bit of hassle for Google?”
In fact, Google says it isn’t a hassle. The company, which was involved in the UK public consultation, said it has worked out how to adjust its servers’ clocks with minimum disruption. But every company currently decides on its own how best to do that. Google smears out the extra second over the previous few hours, whereas BT sticks with adding in an extra second. “We deal with the leap second by including it correctly,” says Flavin. “The last minute of the relevant day will have 61 seconds.” BT’s beloved dial-up speaking clock will add in an extra second’s pause before its third “pip” to take account of it.
So leading up to the 30 June leap second, different companies’ clocks will be telling very slightly different times. That’s a potential problem when it comes to time-stamping legal documents or financial transactions, and could in theory allow someone to game the markets, say. “I’d like to see a standardised way for computers to move through a leap second,” Kuhn says.
He advocates another reform, however: running two time systems in parallel. International Atomic Time (TAI) is the caesium-based time system that has been ticking away without adjustment since the late 1950s. It is an average of the oscillations of over 400 atomic clocks around the world, and thanks to leap seconds is now running 35 seconds ahead of UTC – after 30 June, make that 36 seconds. A sensible step, Kuhn thinks, would be to allow the use of TAI in systems that only need to measure the passage of time, not what the time is. Some systems already do this – some power companies, for example, use TAI to coordinate the output from their generators. But at the moment, that’s not easy to do. Although TAI feeds into the calculation of UTC, it is not itself widely disseminated.
For Penrose, such fiddling doesn’t go far enough. He thinks we need to stop time and start again. He wants us to define a universally agreed zero of time, then adopt a time system he has named “T” – simply the number of seconds after that zero. “I’m proposing that both TAI and UTC be abandoned as universal standards,” he says. “Surely just using the number T is much simpler?”
He advocates using the metronomic bursts of radiation that reach us from fast-spinning neutron stars called pulsars to mark out T, so that we retain a link to astronomical time. Each country would then decide how it wants to define time for its citizens’ everyday use – and whether to shift its clock with leap seconds. There won’t be any confusion because it will simply publish its chosen public time relative to T, just as it does with daylight saving times relative to UTC. Computer-based clocks, meanwhile, can work simply from T without interruption, whether or not a leap second comes along. “I would hope that the UK adopts a leap second into its system, but I wouldn’t get too worked up if it didn’t,” says Penrose.
Going by the , published in September last year, the UK is likely to still be defending the leap second. In general, technology companies followed Google’s line and said losing the leap second would make their lives easier, but that they can – and do – cope with it. A standardised technique, particularly if accredited by the ITU, would be helpful. Religious groups shrugged their shoulders; losing the leap second wouldn’t affect their prayer times or festival dates. As far as the general public were concerned, the report noted that they “were divided between those who were indifferent to the issue and those who wanted to maintain clocks in time with the sun”.
Indifference seems to be the international reaction, too: no other country has expressed any interest in the UK’s public dialogue. In the meantime, in the inscrutable way of government, responsibility for the UK’s time has passed from the minister for science to the minister for business.
Does that signal a turnaround in official attitudes – and if so, does that mean the leap second’s time is up? Only time will tell.
Every second counts
The concept of dividing the day into smaller units goes back to the Bronze Age at least, but the invention of the modern second is usually attributed to Persian polymath Al-Biruni around the end of the first millennium AD. He took the hour – a unit of 1/24 of the day already used by the Greeks and Sumerians among others – and subdivided it twice by 60. Latin scholars later christened the second of these units “secunda”, corresponding to 1/86,400 of a day.
By the mid-20th century, it was clear that fluctuations in Earth’s rotation meant day length varied too much to define the second consistently. Even the “mean solar day”, the average duration of a day over a year, wasn’t accurate enough. And so time itself was reinvented. “It’s an extraordinary story, but also extraordinarily complex,” says of the Royal Observatory Greenwich in London.
Wobbly time
In 1960, after years of debate, the International Committee for Weights and Measures agreed on a starting from the length of a “tropical” year – the time it takes for the sun to return to exactly the same position in the sky, say from one summer solstice to the next. (Owing to a slow shift in the angle at which Earth spins, this is not quite the same as the time it takes the Earth to orbit the sun.)
Unfortunately, Earth’s spin axis also wobbles as it shifts, leading the tropical year’s length to vary, too. The committee nailed it down to a specific year starting at noon GMT on 31 December 1899 (for their purposes “0 January 1900”). According to meticulous calculations made back then by mathematician Simon Newcomb, this lasted just shy of 365 days, 5 hours and 49 minutes – 31,556,925.9747 seconds, in fact.
But hang on a second: to make that last conversion, Newcomb had to assume the second had a set length. And where did he get that from? In the old-fashioned way, by subdividing the mean solar day in 1900 into 86,400.
This complex, somewhat circular reasoning created an unvarying unit, but for practical purposes it was not much use; there was no easy way to get a physical clock ticking to exactly this tock. So in 1967 the committee jumped again and “, defined by a specific number of atomic vibrations: to be precise, “9,192,631,770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium-133 atom”.
But that time corresponded to the length of 1/86,400 of a good old solar day, derived from observations made between 1952 and 1958. Nonetheless, 1967 is seen as a decisive break in the historical link between astronomy and time. “It’s the critical decision that moves from the astronomical to the atomic timescale,” says McEvoy.
Though the atomic second is being measured with ever greater accuracy, that definition still stands. So the fundamental unit of time remains tenuously linked to the duration of a day – albeit an average day in the mid-1950s. Graham Lawton
This article appeared in print under the headline “Stop all the clocks…and then start them again”
Article amended on 26 December 2014
Since this article was first published, the figures in “The world in one second: Distance” have been corrected.
