Nuclear Clock vs Atomic Clock: How Thorium-229 Keeps Time

A nuclear clock uses a transition within an atomic nucleus as its frequency reference. Conventional atomic clocks use transitions involving electronic structure. Both compare an oscillator with a reproducible quantum resonance and correct its frequency.

Thorium-229 is unusual because a low-energy nuclear transition can be addressed with vacuum-ultraviolet laser light. Two papers published in Nature on 7 October 2026 demonstrate operating clocks based on that transition. ^{[1], [2]}

The advance is a working nuclear reference with feedback. It does not establish universal superiority over optical atomic clocks. Stability within a run, reproducibility between runs and evaluated systematic uncertainty must be assessed separately.

What actually ticks in a quantum clock?

A clock combines an oscillator with a counting system. In a quantum clock, atoms or nuclei provide a frequency reference that helps keep the oscillator on target.

The apparatus probes the transition and determines whether the oscillator is above or below the desired resonance. Feedback adjusts it, and the corrected oscillation can be compared or counted using suitable electronics and optical techniques.

The atom is not a miniature pendulum, and an operating clock need not wait for individual spontaneous emissions at regular intervals. Its reference is the transition frequency. NIST’s explanation connects this resonance condition to the oscillations used for timekeeping. ^[3]

The physical reference and the working instrument are therefore different things: a useful transition must be integrated with interrogation, readout, control and calibration.

What changes when the reference is nuclear?

An atom contains a nucleus surrounded by electrons. Electronic and nuclear transitions involve different internal degrees of freedom and respond differently to their environment.

Moving the frequency reference into the nucleus can reduce some environmental sensitivities. It does not remove every electric, magnetic or temperature-dependent shift.

A nucleus embedded in a crystal experiences its local surroundings, including electric-field gradients. Those fields can split or shift its levels. The response also depends on the transition chosen and the host material.

The useful question is consequently quantitative: which shifts occur, how large are they, how stable are they and how well can they be corrected? A nuclear clock still needs an uncertainty budget.

Why thorium-229 is exceptional

Most nuclear transitions require much more energetic radiation than typical electronic transitions. Thorium-229 has a comparatively low-lying nuclear isomer near 8.4 electronvolts, corresponding to light at approximately 148 nanometres.

This makes direct laser interrogation possible. The radiation is in the vacuum-ultraviolet region, outside human vision and subject to demanding optical and propagation requirements.

Direct laser excitation of the thorium transition was reported in 2024. That established an essential spectroscopic step, but excitation alone is not the same as operating a clock. ^[4]

A complete device must convert the measured response into a frequency error signal and use that signal to steer an oscillator. It must then show that the resulting output behaves as claimed.

Why radioactive decay is not the clock’s tick

Thorium-229 is radioactive, but the clock does not measure time by counting a steady sequence of radioactive decays.

Individual decay times are unpredictable. A half-life describes how a population changes statistically. A resonance frequency instead reflects a particular energy difference between quantum states.

The relationship is simple: transition energy equals Planck’s constant multiplied by frequency. The clock repeatedly interrogates that frequency and adjusts its oscillator.

The lifetime of the excited state matters because it is related to the transition’s natural linewidth. However, the linewidth observed in a crystal can be broadened by additional effects, and a long lifetime does not require the electronics to wait that long between every correction.

Absorption readout can reveal detuning without waiting for the slow spontaneous fluorescence decay.

How feedback turns spectroscopy into a clock

Consider a laser close to a resonance. A suitable interrogation method produces an error signal with opposite signs on opposite sides of the line centre.

The controller interprets that sign and magnitude, changes the laser frequency and repeats. Between corrections, the oscillator’s own stability still matters.

This general explanation does not require every clock to use identical hardware or simply alternate between two static probe frequencies. Frequency modulation and demodulated absorption can provide the required discriminator.

Recent continuous-wave thorium absorption spectroscopy made fast readout available for this purpose. ^[5]

A strong absorption dip is not by itself a running clock. The experiment must close the loop and show how accurately and stably the output follows the reference. This is where quantised energy levels become an operating instrument.

What the two October demonstrations show

Toscani De Col and colleagues locked a laser to thorium in calcium fluoride and compared it with a ytterbium-ion clock. The main text reports fractional instability in the low 10⁻¹⁴ range within a continuous run, but reproducibility between days limited to about 5 × 10⁻¹³. It associates changes between runs with the sampled crystal region after optical realignment. These qualified performance statements are more useful than quoting the abstract’s optimistic wording alone. ^[1]

Huang and colleagues reported a separate laser-locked thorium clock. Their reported in-loop instability scales as 5 × 10⁻¹³ divided by the square root of averaging time in seconds over the characterised regime. Transition frequencies in independently fabricated crystals agreed at roughly the 10⁻¹³ level. ^[2]

These metrics describe different checks. Neither should be relabelled as a complete evaluation of absolute accuracy.

Stability, reproducibility and accuracy are different

Stability describes fluctuations as a function of averaging time. Reproducibility asks how closely the result returns when relevant conditions, samples or measurement runs change. Accuracy is assessed through the measurement’s systematic corrections and uncertainty relative to its intended reference.

A stable oscillator can be consistently offset. Two repeated runs can disagree even when each run looks smooth. Two devices can agree while sharing an unrecognised systematic effect.

An ordinary thermometer illustrates the distinction. Repeating the same reading establishes repeatability, but does not show that its zero point is correct. Comparing it with another thermometer helps only if their errors are sufficiently independent and understood.

Clock comparisons face the same logic at much smaller fractional scales. Reporting a low instability does not justify an age-of-the-universe time-loss claim without further assumptions about systematic offsets and long-term behaviour.

What averaging can and cannot improve

Suppose a measurement is limited by independent random fluctuations whose uncertainty falls as the inverse square root of averaging time. Increasing the averaging time by a factor of one hundred reduces that contribution by a factor of ten.

This scaling applies only while the assumed noise regime holds. A persistent offset remains, and slow drift can eventually make further averaging less helpful.

There is also a distinction between in-loop and independent verification. A control system can report small residual errors in the signal it is correcting while another measurement reveals additional noise or bias.

An in-loop estimate is useful evidence about operation, but it should be labelled. A comparison with a separate reference and a systematic-error analysis answer additional questions.

Why a crystal helps and complicates the measurement

A solid host can contain many interrogated nuclei in a small volume. A larger signal can make it easier to estimate the resonance frequency and could support compact devices.

The same host creates variation in local environments. Strain, defects, temperature changes and different occupied sites can broaden or shift the response. A measurement can also change if the optical beam samples a different part of the crystal.

This creates a practical trade-off between convenient signal collection and environmental control. A stronger signal does not remove an unknown shift in its centre.

Good clock design therefore includes reproducible optical alignment and careful material characterisation alongside improvements to the light source and feedback electronics.

How clock comparisons can constrain dark matter

Some ultralight-dark-matter models predict tiny changes in quantities that determine atomic and nuclear transition energies. Different transitions would respond differently, allowing researchers to search for variation in a frequency ratio.

The Toscani De Col study used its comparison to set model-dependent constraints, not to report a detection. ^[1]

A hypothetical signal must be distinguished from environmental drift, periodic laboratory disturbances and calibration effects. A null result constrains only the tested model parameters and the frequencies to which the measurement is sensitive.

This complements particle and field searches for dark matter. It does not replace them, because each technique probes a different interaction and parameter range.

A stronger new-physics sensor need not be the best timekeeper

A transition can be especially responsive to a hypothetical change in a physical parameter. That response can compensate for a measurement noise level worse than another clock’s when the goal is to constrain that particular parameter.

Conversely, a highly stable reference may have limited sensitivity to the effect being sought. Comparisons therefore need both an uncertainty estimate and a justified response model.

The nuclear-clock programme combines several goals: better frequency references, reproducible solid-state instruments and distinctive tests of fundamental physics. Progress in one does not automatically establish progress in all three.

The next meaningful comparisons will pair measured performance with explicit operating conditions and independently evaluated systematic effects.