Atomic clocks lock an oscillator to a fixed atomic or nuclear transition frequency; the best optical versions now agree to 3.2 parts in 10^18, meeting a stated milestone criterion for redefining the SI second, while a parallel effort tries to shrink the same physics onto photonic chips and into deployed networks.
Summary
An atomic clock disciplines a local oscillator to a transition in an atom, ion or nucleus, whose frequency is set by nature rather than by a manufactured artefact. Caesium microwave standards define the current SI second. Optical clocks interrogate transitions some five orders of magnitude higher in frequency, which divides the same fractional error over far more cycles and buys stability. The frontier platforms in the sources are trapped ions (27Al+), neutral atoms in optical lattices (171Yb, 87Sr) and atoms in optical tweezers (88Sr).
The performance-limiting parameters are not the atoms. They are the local oscillator, the environment and the readout. A clock laser must be quiet enough to interrogate a hertz-wide line, which today means a cryogenic reference cavity; Brownian noise in the thin-film mirror coatings of those cavities is the dominant fundamental limit. Laser frequency noise at high Fourier frequencies is not fixed by cavity locking alone, and one recent demonstration combined a cryogenic silicon cavity with an integrated Brillouin laser to suppress noise across more than seven decades, giving sub-hertz phase-integrated linewidth and sub-hertz Rabi spectroscopy on a three-dimensional 87Sr lattice clock. On the readout side, decoherence of the local oscillator produces phase slip errors that cap useful interrogation time, and both weak-measurement protocols and entangled multi-level probes are being proposed and tested to push past that.
A third path bypasses electron-shell transitions entirely. Thorium-229 has a nuclear transition at 148 nm that can be excited by laser light in a doped crystal, promising a solid-state clock that is more robust against external perturbations and more sensitive to variations in fundamental constants.
Evidence base
- 2026-07-20: Frequency ratios between 27Al+, 171Yb and 87Sr clocks measured with total fractional uncertainty at or below 3.2 x 10^-18, meeting a milestone criterion for redefinition of the SI second, using a common ultrastable reference over 3.6 km of phase-stabilised fibre.
- 2026-07-20: The same work reports discrepancies against its own earlier measurements of about 1 x 10^-16 for 87Sr ratios and 1.6 x 10^-17 for Al+/Yb.
- 2026-06-04: A thorium-229 nuclear clock was closed as a feedback loop by stabilising a continuous-wave laser to the 148 nm nuclear transition in a millimetre-sized room-temperature CaF2 crystal, compared against a Yb+ single-ion clock.
- 2026-05-29: A cryogenic silicon cavity combined with an integrated Brillouin laser suppressed frequency noise over more than seven decades, giving under 1 Hz phase-integrated linewidth and 0.2 Hz^2/Hz above 10 MHz, verified by sub-hertz Rabi spectroscopy on a 3D 87Sr lattice clock.
- 2026-06-15: Independently running miniature rubidium clocks plus post-processing synchronised three nodes on deployed metropolitan telecom fibre for eight continuous hours, giving Hong-Ou-Mandel visibility above 70 per cent, statistically equivalent to a White Rabbit link.
- 2026-05-05: A thin-film lithium tantalate UV electro-optic modulator achieved VpiL of 85 mV.cm at 375 nm, 22.7 dB extinction ratio and 1.3 dB insertion loss, up to four orders of magnitude better in bandwidth per VpiL than bulk technologies, targeted at portable atomic clocks.
- 2026-08-03: A coating-free monolithic fused silica total-internal-reflection resonator reached finesse 1225 with a 50 mm^3 mode volume, removing the Brownian coating noise that limits conventional reference cavities.
Open questions
Assessment drafted 2026-08-31 from up to 14 KB sources using the technology-scorecard framework; scores are a draft read pending review.