The remarkable accuracy of atomic clocks makes them excellent instruments for timekeeping and other precision measurements. Writing in Nature, the Boulder Atomic Clock Optical Network (BACON) Collaboration1 reports extremely accurate comparisons of three world-leading clocks in Boulder, Colorado, housed at the National Institute of Standards and Technology (NIST) and the JILA research institute. The authors show how their clock comparisons provide insights into fundamental physics and represent substantial progress towards redefining the second in the International System of Units (SI).
Atomic clocks ‘tick’ at a rate determined by the frequency of light that is emitted or absorbed when an atom changes from one energy state to another. Clocks based on different atoms run at different rates, and the term ‘optical clock’ refers to one that runs at an optical frequency. Three of the world’s best optical clocks are the aluminium-ion and ytterbium clocks at NIST and the strontium clock at JILA. The measured frequencies of all three clocks are estimated to be correct to within a fractional uncertainty of 2 parts in 1018 or better2–4. This level of uncertainty could, in principle, allow the clocks to keep time so accurately that they would gain or lose no more than one second over the age of the Universe. Such optical clocks would be 100 times more accurate than caesium clocks5.
There is therefore a desire to redefine the SI second in terms of an optical-clock frequency and to move away from the current definition based on caesium. But before such a redefinition is possible, scientists must build confidence in the reproducibility of optical clocks through a series of clock comparisons. The target accuracy for these comparisons is at the level of parts in 1018 to clearly demonstrate the superiority of optical clocks over caesium clocks5.
Clock comparisons are carried out by measuring ratios of the optical-clock frequencies using instruments called femtosecond-frequency combs. Until now, the best comparisons between optical clocks based on different atoms6–11 have been at the level of parts in 1017. The BACON Collaboration presents measurements of optical-frequency ratios reaching uncertainties at the level of parts in 1018, bringing the redefinition of the SI second a step closer.
Such frequency-ratio measurements are no mean feat, and are equivalent to determining the distance from Earth to the Moon to within a few nanometres. Therefore, exceptional care is required to control any sources of frequency offset. The authors compared the aluminium-ion and ytterbium clocks at NIST and the strontium clock at JILA over several months to check reproducibility and reduce statistical uncertainties. They found that the day-to-day variation in the ratios was slightly larger than expected after accounting for all known effects. This observation suggests the existence of unknown effects, which are intrinsically hard to quantify. Nevertheless, the authors developed a statistical model to account for these uncertainties and to enable a rigorous assessment of the total measurement uncertainty.
The BACON Collaboration used redundant elements throughout the network of optical clocks to check for any sources of bias. In particular, the authors linked the clocks at NIST and JILA in two ways (Fig. 1). They used a 3.6-kilometre optical-fibre link — a tried-and-tested method for connecting signals between optical clocks at the required level of uncertainty. And, in what they believe to be a world first, they compared the clocks using a 1.5-km ‘free-space’ link by sending laser pulses through the air between NIST and JILA along a straight line joining the two institutes. They found that the two types of connection provided a similar level of uncertainty, apart from when data could not be obtained from the free-space link during a snowstorm.