Quantum Clocks: Metrology and Relativity

• Quantum clocks are physical systems that operationally encode time via mechanisms such as discrete tick registers, continuous estimation, and entanglement-enhanced atomic setups.
• They demonstrate near-quadratic precision advantages over classical clocks through methods like autonomous ticking, phase stability, and feedback-enhanced tick processing.
• Quantum clocks underpin advances in precision metrology, relativistic time dilation studies, dark matter searches, and cooperative quantum networks for distributed timekeeping.

Quantum clocks are physical systems whose states, outputs, or correlations encode temporal information, but the term does not denote a single model. In the literature it refers to at least four distinct but overlapping constructions: finite-dimensional autonomous devices that emit ticks, continuous quantum systems analyzed through estimation theory, optical atomic clocks whose performance is enhanced by entanglement, and relativistic systems whose internal degrees of freedom register proper time. Across these settings, time is treated operationally: as a record in tick registers, as an estimator inferred from measurement statistics, as a network-wide stabilized signal, or as a covariant observable tied to internal dynamics rather than to an external classical parameter alone (Ranković et al., 2015, Ramezani et al., 2022, Yang et al., 7 May 2025, Smith et al., 2019).

1. Operational meanings and formal definitions

A central operational definition models a quantum clock as a pair

$(\rho_C^0,\mathcal M_{C\to CT}),$

where $C$ is the clockwork, $T$ is a tick register, $\rho_C^0$ is the initial clockwork state, and $\mathcal M_{C\to CT}$ is a CPTP map that couples the clockwork to a fresh register. The generated time scale is the sequence of reduced states on successive tick registers. In this framework, clock time is not an external observable but the output of repeated interactions between a dynamical system and a stream of registers (Ranković et al., 2015).

A related information-theoretic formulation treats a clock as a system that autonomously emits information about time in the form of ticks, with the absence of a tick also carrying temporal information. The size of such a clock is measured by the dimension $d$ of its state space, equivalently by the number of bits $\log_2 d$ that can be stored in it. This framing makes precision a resource-theoretic question: how well can time information be generated at fixed finite dimension (Woods et al., 2018).

A different line of work defines a continuous quantum clock as a quantum system whose dynamical variable evolves continuously in time and whose time reading is inferred from the probability distribution of measurement outcomes. There the primary figures of merit are the estimator variance

$$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$

and the recurrence time, namely the maximum useful interval before the state returns too close to an earlier one and time becomes ambiguous (Ramezani et al., 2022).

In precision-metrology settings, “quantum clock” often means an optical atomic clock whose performance exceeds the Standard Quantum Limit through entanglement. In relativistic settings, it means a system whose internal Hamiltonian generates a physically readable phase that tracks proper time; one formulation uses a covariant time observable, another uses superpositions of internal-energy eigenstates as the clock degree of freedom (Yang et al., 7 May 2025, Smith et al., 2019, Castro-Ruiz et al., 2015). A separate foundational construction introduces the ideal quantum clock as a closed system with mutually orthogonal click states at equidistant times, enabling the definition of a symmetric time operator on a restricted domain (Gessner, 2013). This multiplicity of definitions reflects different tasks rather than a merely terminological divergence.

2. Autonomous ticking clocks, synchronization, and finite-dimensional bounds

For autonomous finite-dimensional clocks, continuity and synchronization are central constraints. An $\epsilon$-continuous clock satisfies

$$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$

so each interaction with a fresh tick register perturbs the clockwork only weakly. Clock quality can then be quantified operationally by the Alternate Ticks Game, in which two clocks send tick registers to a referee and are judged by how long they can maintain the alternating pattern $C$0 without communication after initialization. The success probabilities $C$1 and the expected number of alternating ticks $C$2 are the synchronization metrics in that setting (Ranković et al., 2015).

For reset clocks, the basic precision metric is

$C$3

with $C$4 the mean tick time and $C$5 its variance. Resetting after each tick makes the tick intervals i.i.d., yielding $C$6 for the $C$7-th tick. Classical $C$8-dimensional clocks satisfy the linear bound

$C$9

and the Ladder Clock saturates it. By contrast, the Quasi-Ideal quantum clock attains

$T$0

for every fixed $T$1, establishing a near-quadratic quantum-over-classical precision advantage at fixed dimension (Woods et al., 2018).

The clock-processing problem sharpens this contrast. A finite-dimensional quantum enhancing clock can take a low-accuracy i.i.d. input tick signal and produce an output with improved inaccuracy. The relevant inaccuracy measure is defined from confidence intervals of the $T$2-th tick, and for i.i.d. clocks the alternative precision measure obeys the usual scaling $T$3. The enhancing clock has a no-tick mode with unitary periodic evolution and a tick mode with autonomous tick emission, subject to a stability criterion requiring the next tick time to be approximately independent of the detector switching phase. With no feedback to the input source, the enhancement is temporary; for a Quasi-Ideal Clock of dimension $T$4, the output inaccuracy scales as

$T$5

for large $T$6. With feedback that resets the input clock after each output tick, the improved accuracy can be maintained indefinitely, with

$T$7

The same work shows that finite-dimensional classical clocks modeled as continuous-time Markov chains cannot satisfy the required phase-stability criterion in a nontrivial way, and that classical tick-processing strategies based on bunching improve only with $T$8-type scaling rather than nearly $T$9 (Yang et al., 2019).

3. Continuous clocks, estimation theory, and athermal autonomous designs

In the continuous-clock framework, the clock is a parameter-estimation device. Classical estimation bounds are expressed through Fisher information,

$\rho_C^0$0

and quantum limits through the quantum Fisher information $\rho_C^0$1. For a one-qubit clock with Hamiltonian $\rho_C^0$2, optimal projective readout gives $\rho_C^0$3 and the shot-noise bound

$\rho_C^0$4

Its limitation is structural: improving precision by increasing $\rho_C^0$5 shortens the recurrence time, which scales inversely with the same frequency (Ramezani et al., 2022).

A two-qubit clock separates these roles. With a Hamiltonian containing two frequencies $\rho_C^0$6 and $\rho_C^0$7, the Fisher information becomes

$\rho_C^0$8

so precision depends on both frequencies while the recurrence time is controlled by the slower one,

$\rho_C^0$9

This permits better precision and longer recurrence time than a collection of $\mathcal M_{C\to CT}$0 one-qubit clocks using the same nominal resource count. Entangled clocks improve precision further, reaching Heisenberg-like scaling

$\mathcal M_{C\to CT}$1

but the gain is accompanied by worsened recurrence time because the effective phase-accumulation frequency grows with $\mathcal M_{C\to CT}$2 (Ramezani et al., 2022).

A different autonomous architecture uses athermal resources rather than thermal gradients. In that model, unobserved continuous measurements act as engineered reservoirs, and two-level or three-level systems transduce measurement-induced noise into tick events. The average dynamics is Lindbladian with a Hermitian double-commutator measurement term, the ticking rate is maximized when the measured observable maximally non-commutes with the clock Hamiltonian, and large-deviation methods characterize the full counting statistics of ticks. The tick statistics can be sub-Poissonian, quantified by Mandel’s $\mathcal M_{C\to CT}$3 parameter, with the explicit two-level result

$\mathcal M_{C\to CT}$4

and minimum $\mathcal M_{C\to CT}$5. Hybrid variants combine measurement-driven and thermal resources within the same autonomous clock model (Manikandan, 2022).

4. Entanglement-enhanced optical atomic clocks

In optical atomic clocks, the dominant benchmark is the Standard Quantum Limit, which stems from the uncorrelated projection noise of each atom. For a coherent spin state, the metrological squeezing parameter is defined as

$\mathcal M_{C\to CT}$6

where $\mathcal M_{C\to CT}$7 is the spin-noise reduction and $\mathcal M_{C\to CT}$8 the effective Ramsey contrast; $\mathcal M_{C\to CT}$9 indicates an entangled and metrologically useful state. A 2025 experiment realized this regime in a strontium optical lattice clock by using cavity-based quantum nondemolition measurements to prepare two spin-squeezed ensembles of about $d$0 $d$1Sr atoms each in a two-dimensional optical lattice (Yang et al., 7 May 2025).

That experiment used synchronous Ramsey spectroscopy with interrogation time $d$2 ms. After correcting for state-preparation and measurement errors, the synchronous comparison achieved a metrological gain of $d$3 dB beyond the SQL. Before correction, the optimal QND strength produced a maximum spin-noise reduction of $d$4 dB and a metrological gain of $d$5 dB. In clock-operation terms, the instability was

$d$6

and after $d$7 minutes the single-clock fractional frequency uncertainty reached

$d$8

The work identifies this as the most precise entanglement-enhanced clock to date and explicitly connects such performance to relativity, geodesy, tests of constant stability, and the interplay of gravity and quantum entanglement (Yang et al., 7 May 2025).

Remote entanglement has also been demonstrated directly between optical clocks. A two-node network of $d$9Sr$\log_2 d$0 ions separated by $\log_2 d$1 m used a photonic link to generate heralded Bell pairs with fidelity $\log_2 d$2 in about $\log_2 d$3–$\log_2 d$4 ms. The entangled comparison reduces the single-shot uncertainty in the clock-frequency difference by a factor close to $\log_2 d$5, as expected for the two-particle Heisenberg limit, thereby halving the number of measurements needed for a given precision. In the laser-dephasing-limited regime, the same protocol yields a factor $\log_2 d$6 reduction in required measurements relative to conventional unentangled correlation spectroscopy. A proof-of-principle AC Stark shift applied to one ion was measured with the entangled network as a direct demonstration of enhanced remote differential spectroscopy (Nichol et al., 2021).

5. Cooperative networks and sensing applications

Quantum networking extends clock enhancement from single devices to distributed architectures. One proposal considers $\log_2 d$7 geographically remote optical clocks that share non-local GHZ-type states spanning all nodes. The network runs in cycles of entangled-state preparation, interrogation by local oscillators, and feedback. The essential observable is the center-of-mass frequency,

$\log_2 d$8

which the global GHZ state probes directly rather than estimating each local phase independently. A cascade of GHZ states of increasing size overcomes phase-wrapping and laser-noise limitations, enabling a phase-estimation scaling

$\log_2 d$9

The corresponding Allan-deviation analysis shows a quantum-enhanced $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$0 regime at short averaging times and approach to the fundamental decoherence-limited bound at long times, with the cooperative network reaching that limit faster than classical synchronization. The same proposal embeds security features against internal sabotage and external eavesdropping through test rounds, quantum key distribution, and controlled distribution of stabilized signals (Kómár et al., 2013).

A conceptually different cooperative protocol refines a broadcast time signal without distributed entanglement. Each network node receives a central low-accuracy signal and uses a local finite-dimensional quantum clock as an enhancing clock. The refined local output becomes more accurate than either the broadcast signal or the local clock alone, and the combined inaccuracy can be reduced essentially to the product of the two individual inaccuracies. This suggests a primitive for clock signal processing in quantum networks, where local clocks are not only standards to be compared but active processors of time information (Yang et al., 2019).

Clock networks also function as sensors for new physics. In dark-matter searches, atomic and proposed nuclear clocks are used as probes of oscillations in fundamental constants induced by ultralight scalar dark matter. The relevant sensitivity coefficient $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$1 spans $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$2 to $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$3 for most optical lattice and trapped-ion clocks, reaches $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$4 in Hg$$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$5 quadrupole and $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$6 in Yb$$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$7 octupole clocks, and may be around $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$8 in proposed $$\Delta t=\sqrt{\langle t^2\rangle-\langle t\rangle^2}$$9Th nuclear clocks. Three interrogation strategies have been compared: differential spectroscopy, narrowband dynamical decoupling, and a new broadband dynamical decoupling algorithm based on randomly timed $\epsilon$0 pulses. In simulations with realistic noise and dark-matter decoherence, broadband dynamical decoupling achieved sensitivity comparable to narrowband dynamical decoupling while maintaining broad frequency coverage and reaching the $\epsilon$1 dark-matter mass range without significant loss of sensitivity. The same work also proposed electron-bridge excitation pathways at $\epsilon$2 nm and about $\epsilon$3 nm as an alternative to direct vacuum-ultraviolet excitation of the thorium transition (Zaheer et al., 2023).

6. Relativistic and gravitational quantum clocks

When clocks are treated as physical quantum systems, internal energy contributes to inertia through the operator-valued mass

$\epsilon$4

This makes the manner in which a clock is set into motion physically consequential. For a momentum boost, all internal branches receive the same momentum but not the same velocity, because their masses differ. The resulting internal phase shift does not reproduce a single classical Lorentz factor. A velocity boost built from the operator-valued mass instead gives all branches the same velocity and recovers the standard classical time-dilation factor in the appropriate limit. The distinction yields a small additional frequency shift in ion-trap atomic clocks and produces non-ideal ticking in the Salecker–Wigner–Peres clock model when only state-independent forces are applied (Paige et al., 2018).

A complementary operational treatment defines a relativistic quantum clock by its internal degrees of freedom and a covariant time observable. For two such clocks, one can compute the probability that clock $\epsilon$5 reads $\epsilon$6 conditioned on clock $\epsilon$7 reading $\epsilon$8. When the center-of-mass states are localized Gaussian momentum wave packets, the conditional mean proper times satisfy the classical time-dilation relation. When one clock is in a coherent superposition of localized momentum packets, an additional interference term appears in the mean proper time, producing a genuinely quantum correction to time dilation. The same framework derives a proper-time–energy uncertainty relation and, through $\epsilon$9, a proper-time–mass uncertainty relation from the Helstrom–Holevo bound (Smith et al., 2019).

Gravitational interaction between quantum clocks produces a further departure from the ideal independent-clock picture. For two two-level clocks separated by distance $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$0, the effective weak-field interaction Hamiltonian

$$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$1

makes the proper time of each clock depend on the internal-energy state of the other. If the clocks begin in superpositions of energy eigenstates, they become entangled through gravitational time dilation. For equal gaps $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$2, the characteristic mixing time is

$$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$3

and for $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$4 interacting clocks the reduced visibility of one clock decays with a decoherence time

$$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$5

In a related synchronization model with different gaps $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$6 and $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$7, the measurement probabilities and quantum Fisher information for estimating the gravity-induced time difference depend sensitively on the energy gaps and the inter-clock distance; improved metrological precision is interpreted there as an indicator of gravity-generated entanglement (Castro-Ruiz et al., 2015, Wang et al., 2019).

These phenomena motivate generalized relativistic principles for delocalized clocks. A quantum generalization of Einstein’s Equivalence Principle has been formulated for clocks in superpositions of positions and velocities, with validity shown to be equivalent to the possibility of transforming to the perspective of an arbitrary quantum reference frame associated with the clock. Entangled clocks in an interferometer in Earth’s gravitational field are proposed as a verification platform, where violations would manifest as modified laboratory-frame detector probabilities. Independently of these interference effects, gravity also sets a stability bound on localized clocks through the competition between gravitational redshift from uncertain height and special-relativistic time dilation from quantum momentum spread. For a trapped clock of mass $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$8 in a field $$\| \operatorname{tr}_{T}\circ \mathcal M_{C\to CT}-\mathcal I_C\|_\diamond \le \epsilon,$$9, the minimum fractional instability is

$C$00

with the concrete estimate that a single-ion hydrogen maser clock on Earth cannot achieve a stability better than about one part in $C$01 (Cepollaro et al., 2021, Sinha et al., 2014).

7. Time operators, stationary clocks, and stochastic clock time

The ideal quantum clock construction addresses the longstanding time-operator problem within standard Schrödinger-picture quantum theory. An ideal clock is a closed system with equidistant times $C$02 and orthogonal click states $C$03 satisfying

$C$04

On an invariant dense subspace $C$05, a projection-like map $C$06 attaches the time labels $C$07 to clock states, and the symmetrized time operator is defined by

$C$08

On that restricted domain, $C$09 is symmetric and canonically conjugate to the Hamiltonian,

$C$10

so the time-energy uncertainty relation

$C$11

follows. Pauli’s objection is evaded because $C$12 is symmetric rather than self-adjoint and is defined only on the physically motivated clock domain (Gessner, 2013).

A more radical proposal shows that elapsed time can be inferred even when the clock is always in a stationary “off” branch. The construction uses counterfactual measurements: an initial off state $C$13 is coherently mixed with an on state, the on branch evolves, and a final unitary plus postselection arranges destructive interference so that successful outcomes reveal the elapsed time while formally having zero probability of the clock having been on during the interval. For the simplest one-tick case, one ancilla is necessary, each counterfactual outcome occurs with probability $C$14, and the total counterfactual probability is $C$15. The authors interpret this as constructive support for a substantival view of time, though that interpretation is explicitly tied to the counterfactual reading and to the exclusion of noncontextual ontic models (Strelchuk et al., 2021).

A different foundational extension replaces the Newtonian time parameter in quantum dynamics by the time shown by a stochastic quantum clock $C$16. The clock is required to be non-decreasing, to tick at random with random tick sizes, and to satisfy $C$17. Ordinary unitary evolution becomes

$C$18

and averaging over the clock process yields a generalized master equation whose leading term is always the von Neumann equation. For gamma-distributed clock increments, the coefficients are

$C$19

the leading correction is Lindblad-like with the Hamiltonian as generator, and higher-order terms extend both von Neumann and Lindblad dynamics. Using the $C$20 hyperfine transition, the condition that the induced decoherence rate remain below $C$21 implies

$C$22

This line of work uses “quantum clock” not as a sensor or a ticking device, but as a stochastic replacement for background Newtonian time itself (Brody et al., 2 Feb 2026).

Taken together, these strands show that quantum clocks are simultaneously metrological devices, autonomous information emitters, relativistic probes, network resources, and candidates for temporal reference systems internal to quantum theory. The common thread is operational: clock performance is determined by how temporal information is generated, encoded, transmitted, stabilized, or conditioned by quantum dynamics, rather than by appeal to an idealized external time variable alone.