Quantum Clock Synchronization (QCS)

• Quantum Clock Synchronization is the use of quantum states, correlations, and interference to synchronize distributed clocks, achieving precision beyond classical limits.
• It employs protocols such as entanglement-assisted time transfer, HOM interference, and two-way time transfer to mitigate propagation delays and environmental disturbances.
• Experimental implementations span fiber, satellite, and integrated networks while addressing challenges like asymmetry attacks and phase inconsistencies.

Quantum Clock Synchronization (QCS) is the use of quantum states, quantum correlations, and quantum measurement to establish a shared notion of time between spatially separated nodes. In the current literature, QCS includes entanglement-assisted time transfer, Hong–Ou–Mandel (HOM) interference-based synchronization, time-of-arrival correlation methods, ticking-qubit and quantum slow-clock-transport schemes, conveyor-belt protocols, and quantum-enhanced two-way time transfer (Khalid et al., 6 Apr 2026). Across these variants, the central task is to estimate and correct clock offsets and drifts with precision beyond classical synchronization limits while improving resilience against propagation-delay uncertainty, atmospheric disturbance, oscillator drift, and adversarial delay manipulation (Khalid et al., 6 Apr 2026). Satellite-ground proposals extend the same objective to hybrid space-terrestrial networks intended for picosecond-accuracy synchronization and future quantum communication infrastructure (Troupe et al., 2022).

1. Foundations and physical principles

QCS differs from classical synchronization because timing information is encoded directly in quantum phase evolution, entanglement correlations, photon-arrival correlations, or interference structure rather than only in deterministic timing waveforms. In entanglement-based and SPDC-based schemes, the relevant resource is often the tight birth-time correlation of photon pairs; one proposal for deep-space and satellite-ground QCS states that SPDC pair-production times are correlated at the $10$–$100$ femtosecond level and are Poisson distributed, so the detected pairs function as an effectively random shared timing code (Troupe et al., 2022). The survey literature correspondingly distinguishes quantum-enhanced synchronization, which improves a mostly classical protocol using quantum resources, from fully quantum synchronization, which uses entanglement and quantum correlations as the basis of the protocol itself (Khalid et al., 6 Apr 2026).

A recurring physical assumption is reciprocity. In two-way and round-trip QCS, the propagation time from Alice to Bob is assumed to equal the propagation time from Bob to Alice, so common delays cancel when forward and backward correlations are combined. This assumption appears in field two-way synchronization over deployed fiber, in round-trip multi-user photonic networks, and in satellite-ground architectures (Quan et al., 2021). A second recurring requirement is phase consistency. Entanglement-based QCS long appeared to contain a hidden assumption that distant parties already shared synchronized clocks or a common phase convention. Asynchronous entanglement purification was later shown to remove this unknown systematic phase even when Alice and Bob have different basis conventions, an overall time offset in the purification sequence, and a noisy channel, thereby closing a major loophole in long-distance photonic QCS (Ilo-Okeke et al., 2017).

The field therefore rests on three linked ideas. First, quantum correlations can define simultaneity more sharply than ordinary pulse exchange. Second, classical postprocessing remains indispensable: almost every protocol still infers the offset from correlation peaks, fringe phases, or timing histograms. Third, the quantum resource does not eliminate calibration questions; it shifts them toward reciprocity, detector response, dispersion, and phase-reference consistency (Khalid et al., 6 Apr 2026).

2. Protocol architectures

The survey categorizes the main protocol families as ticking-qubit handshake QCS, entanglement-based QCS, time-of-arrival QCS, conveyor-belt synchronization, HOM-QCS, time-offset correlation QCS, and quantum-enhanced two-way time transfer (Khalid et al., 6 Apr 2026). These families differ mainly in which observable carries the timing information.

In HOM-based synchronization, two remote optical paths are balanced until the HOM coincidence dip is minimized. After balancing, each clock records local photon-arrival times, and the offset is extracted from the cross-correlation of the event trains. In one proof-of-principle formulation, the event streams are written as

$$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$

and the offset is obtained from

$$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$

The dominant contribution comes from matched photon pairs, so the correlation peak identifies the clock offset once the HOM interferometer has defined simultaneous events in its own reference frame (Quan et al., 2016).

In two-way entangled-photon QCS, the offset is inferred from nonlocal coincidence times in both directions. For the field implementation over a $7$ km deployed fiber, the central estimator is

$t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$

which cancels symmetric path delays under bidirectional reciprocity (Quan et al., 2021). A related round-trip network formulation uses a one-way delay $\tau_{s-u}$ and a round-trip delay $\tau_{s-u-s}$, with

$\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$

to extract the server–user offset while suppressing unknown channel delay (Li et al., 2024).

Time-offset correlation methods focus less on interferometric balancing and more on timestamp postprocessing. For energy-time entangled photons with simple crystal oscillators, one implementation first compensates clock-frequency skew, then extracts the timing offset from a coincidence peak, and finally tracks residual drift during operation. Its jitter model includes the uncompensated-skew term

$\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$

the drift term

$100$0

and the combined live-tracking relation

$100$1

The same work uses

$100$2

for the finite-sample uncertainty of the peak position (Spiess et al., 2021).

Pulsed single-photon or weak-coherent-pulse sources add an explicit temporal grid. In a pulsed-source QCS add-on for a decoy-state BB84-style link, detection times $100$3 are folded into one source period by

$100$4

and the arrival-time histogram

$100$5

yields the relative timing offset inside one repetition period. The frequency-difference estimate for live tracking is then

$100$6

Absolute timing ambiguity by whole periods is resolved once during initialization by shifting the received QKD sequence until the quantum bit error rate is minimized (Spiess et al., 2022).

A more recent practical direction replaces entangled-pair sources with weak coherent pulses (WCPs) and bidirectional HOM interference. In the bidirectional WCP model, the final clock-offset estimator is

$100$7

where $100$8 and $100$9 are pulse-index offsets extracted from HOM coincidence minima and $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$0 are local delay settings. Under the simulated operating point of $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$1, $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$2 Gaussian pulses, $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$3 repetition rate, $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$4, $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$5 detector efficiency, $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$6 detector timing jitter, and $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$7 loss, the estimated offset was $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$8 for a true offset of $$f_A(t)=\frac{1}{\sqrt N}\sum_{i=1}^{N}\delta(t-\tau_i^A), \qquad f_B(t)=\frac{1}{\sqrt N}\sum_{j=1}^{N}\delta(t-\tau_j^B),$$9, corresponding to $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$0 ps accuracy and $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$1 ps standard error (Crum et al., 30 Sep 2025).

3. Multipartite QCS and scaling laws

Multipartite QCS asks how one reference clock can synchronize many remote clocks without the timing signal collapsing as the number of parties grows. Early multiparty work based on $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$2-states led to the widely cited conclusion that synchronization accuracy decreases monotonically with the number of parties. That conclusion was later shown to be resource-specific rather than universal. For symmetric entangled Zen-states $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$3, the oscillation amplitude is

$$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$4

and the synchronization signal becomes

$$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$5

The amplitude is maximized at

$$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$6

so the $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$7 degradation is a property of the $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$8 $$g(\tau)=\int_{-\infty}^{+\infty} f_A(t)f_B(t-\tau)\,dt.$$9-state choice, not a universal multiparty limit (Ben-Av et al., 2011).

This correction was experimentally supported in a four-qubit NMR implementation of both the Krco–Paul $7$0-state protocol and the Ben-Av–Exman $7$1-state protocol. For the four-qubit case, the reported probabilities were

$7$2

for the $7$3-state protocol and

$7$4

for the optimized $7$5-state protocol, with the experimental data agreeing with theory and the $7$6-state protocol yielding lower standard deviations for the inferred offset (Kong et al., 2017).

A separate multipartite line uses GHZ-type maximal multipartite entanglement. In that framework, the synchronization target is the average clock time

$7$7

and each party measures

$7$8

For a given grouping of flipped and unflipped qubits, the collective correlation is

$7$9

The reported uncertainty is

$t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$0

which approaches $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$1 for large $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$2. In the resource-efficiency comparison of that work, the GHZ-type protocol is about twice as qubit-efficient as parallel Bell-pair synchronization and asymptotically four times as accurate per qubit as the symmetric Dicke-state alternative (Ren et al., 2012).

Operation-triggered QCS modifies the trigger itself. Instead of starting evolution by measurement, each node applies a local unitary $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$3 at its local time, preserves entanglement, and then returns all qubits to a center node for a joint measurement. The final state takes the form

$t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$4

so synchronization becomes a multiple phase-estimation problem. For the two-party NOON-state case, the work reports Heisenberg scaling,

$t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$5

and for estimating the average time of $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$6 remote clocks it reports an $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$7 advantage over the best possible strategy using measurement-triggered QCS, again at Heisenberg scale (Yue et al., 2014).

A more recent multiparty construction addresses the phase-convention problem directly through supersinglet purification. In that protocol, Alice’s $t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$8-basis measurement starts the synchronization interval, and each remote node sees a local timing signal

$t_0=\frac{(t_2-t_1)-(t_4-t_3)}{2},$9

For the supersinglet, the amplitudes are

$\tau_{s-u}$0

so the signal remains $\tau_{s-u}$1 with increasing $\tau_{s-u}$2 rather than collapsing as $\tau_{s-u}$3. The same work uses LOCC purification to remove the Preskill phase problem and to prepare the multipartite resource in a phase-assumption-free manner (Oujaa et al., 19 Oct 2025).

A distinct fault-tolerant direction reduces synchronization to detectable Byzantine agreement and distributes the required correlations with a single sequentially transmitted qudit. The initial state

$\tau_{s-u}$4

is acted on by local phase encodings and measured only at the final node, yielding a nonrecursive protocol for clock synchronization in multiprocess networks with arbitrary many faulty processes (Tavakoli et al., 2015).

4. Experimental realizations and performance

Experimental QCS spans proof-of-principle interferometric demonstrations, software-defined timestamp correlation with drifting quartz oscillators, field two-way fiber deployment, postprocessing add-ons to QKD-style pulsed links, and integrated multi-user silicon photonic networks. These implementations differ in physical resource and estimator design, but they all evaluate synchronization through timing jitter, coincidence width, or Allan-type stability metrics.

Platform	Configuration	Representative reported performance
Frequency-entangled photons with HOM balancing	$\tau_{s-u}$5 km fiber link	$\tau_{s-u}$6 ps timing stability at $\tau_{s-u}$7 s; $\tau_{s-u}$8 ps absolute time accuracy (Quan et al., 2016)
Correlated energy-time entangled photons with crystal oscillators	High-loss/turbulence-emulated link	stable synchronization jitter $\tau_{s-u}$9 ps with as few as $\tau_{s-u-s}$0 correlated detection events per $\tau_{s-u-s}$1-ms data package (Spiess et al., 2021)
Two-way entangled-photon field test	$\tau_{s-u-s}$2 km deployed fiber, H-maser to Rb clock	$\tau_{s-u-s}$3 ps at $\tau_{s-u-s}$4 s; improved to $\tau_{s-u-s}$5 ps at $\tau_{s-u-s}$6 s with fiber-optic microwave frequency transfer (Quan et al., 2021)
Pulsed single-photon QKD add-on	$\tau_{s-u-s}$7 MHz weak coherent pulses, $\tau_{s-u-s}$8 m fiber, free-running quartz oscillators	average RMS synchronization jitter $\tau_{s-u-s}$9 ps; TDEV $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$0 ps at $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$1 s, $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$2 ps at $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$3 s, $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$4 ps at $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$5 s (Spiess et al., 2022)
Silicon-chip dual-pumped SFWM QCS network	server with Alice and Charlie at $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$6 km, Bob at $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$7 km, $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$8 h run	lowest TDEV $\Delta t_1=\tau_{s-u}-\frac{\tau_{s-u-s}}{2},$9 ps, $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$0 ps, and $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$1 ps for Alice, Bob, and Charlie, respectively (Li et al., 2024)

These results cover several distinct implementation regimes. The HOM experiment demonstrates how second-order quantum coherence can define simultaneous events once the interferometer is locked to the coincidence minimum. The correlation-only approach with crystal oscillators shows that QCS can be implemented without an external timing reference and with only computer-aided postprocessing. The field two-way experiment demonstrates deployed-fiber operation with heterogeneous clocks and explicitly shows that reference-frequency mismatch can dominate short-term stability. The pulsed-source add-on establishes that a running QKD photon stream can serve as both data carrier and timing carrier with no extra source/receiver hardware and no modification of the quantum source software. The silicon-chip network shows that dense wavelength-division multiplexing and dual-pumped SFWM can turn one source into a multi-user round-trip QCS hub (Quan et al., 2016).

5. Satellite, space-based, and global QCS networks

Space-based QCS is motivated by the limits of fiber transmission at global scale and by the timing requirements of future quantum communication, navigation, sensing, and deep-space experiments. One proposal describes a hybrid space-terrestrial quantum network with satellites carrying compact entangled-photon sources and low-SWaP atomic clocks, ground stations with higher-performance reference clocks, and a hierarchy analogous to NTP in which lower-stratum client clocks are served by large constellations of cheaper LEO satellites while higher-stratum server clocks are serviced by smaller MEO or GEO constellations (Troupe et al., 2022).

Near-term feasibility studies use much more specific hardware models. In one satellite-ground scheme, each nanosatellite carries an SPDC source, APDs, and a chip-scale atomic clock (CSAC), with a simulated source rate

$\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$2

Timestamp streams are cross-correlated via

$\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$3

with peaks at $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$4 and $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$5, giving

$\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$6

Under realistic detector jitter of $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$7 ps and $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$8 ps timestamp resolution, that study identifies a practical threshold of approximately $\sigma_{\text{sync}}=\frac{1}{2}\Delta u T_a,$9 ebits/s for $100$00 ns synchronization and argues that sub-nanosecond synchronization, and tens of picoseconds in favorable settings, are feasible with modest constellations (Haldar et al., 2022).

A more network-level formulation is the quantum-assisted “master clock in the sky.” There the cross-correlations

$100$01

reveal peaks at $100$02 and $100$03, so that

$100$04

The same work uses a motion-induced precision bound

$100$05

to analyze moving links. Its headline result is that a constellation of $100$06 satellites distributed among $100$07 polar orbits at $100$08 km altitude can synchronize clocks spread across the globe at sub-nanosecond precision; for the tested city pairs, the simulations indicate synchronization for $100$09 of the day at $100$10 ns precision (Ducoing et al., 2023).

Relativistic corrections become non-negligible at the intended precision. The deep-space proposal lists Shapiro delay on the order of $100$11 ps for satellite-to-ground transit, higher-order spacetime-curvature effects of a few picoseconds, and orbital, nonspherical-Earth, and tidal corrections ranging from tens of picoseconds down to tenths of a picosecond over a day (Troupe et al., 2022). A dedicated relativistic QCS analysis, based on frequency-entangled pulses and a HOM interferometer in Earth’s Schwarzschild spacetime, finds that gravity distorts the wave-packet overlap and hence the coincidence curve. For a $100$12 km LEO satellite it reports

$100$13

and for GEO it reports

$100$14

with synchronization precision sensitive to both source parameters and satellite altitude (Wang et al., 2015).

These space-based results suggest two distinct roles for QCS. One is infrastructural: a global entanglement-enabled timing layer for distributed quantum networks. The other is metrological: once picosecond or better synchronization is available, the timing network itself becomes sensitive to gravitational and relativistic structure that is negligible in ordinary timing systems (Troupe et al., 2022).

6. Security assumptions, vulnerabilities, and technical constraints

Security claims in QCS are conditional rather than absolute. Two-way QCS is explicitly claimed to provide security against symmetric delay attacks because the offset is inferred from the difference of forward and backward nonlocal coincidence times, and symmetric path perturbations cancel in the ideal formulation (Quan et al., 2021). The same logic underlies other reciprocity-based protocols: security is tied to the assumption that the two directions experience identical propagation delay.

That assumption is now known to be a genuine attack surface. A tunable asymmetric delay attack (T-ADA) was experimentally demonstrated on a $100$15 km round-trip QCS system. The round-trip estimator

$100$16

is biased whenever an attacker introduces direction-dependent delay. In the reported experiment, baseline stability without attack was

$100$17

Under a jump attack, long-term stability degraded to $100$18 ps at $100$19 s; under a spike attack, short-term stability reached $100$20 ps at $100$21 s before decreasing to $100$22 ps at $100$23 s; and under gradual attacks the TDEV rose to $100$24 ps or $100$25 ps at $100$26 s depending on trajectory (Han et al., 24 Oct 2025). The paper states directly that QCS is not secure merely because it uses entangled photons.

A second conceptual vulnerability is phase convention. Entanglement alone does not guarantee that remote laboratories share consistent definitions of $100$27 and $100$28. Asynchronous entanglement purification was introduced precisely to remove the unknown systematic phase in bipartite QCS without prior synchronized clocks (Ilo-Okeke et al., 2017), and supersinglet purification extends the same concern to scalable multiparty broadcasting (Oujaa et al., 19 Oct 2025). A third misconception concerns scaling: monotonic precision loss with participant number is not universal, but specific to certain resource states such as the $100$29-state (Ben-Av et al., 2011).

Experimental limitations remain substantial. The field two-way deployment identifies Rb-clock instability, low event-timer sampling rate of about $100$30 kHz per port, residual dispersion, environmental temperature fluctuations, and event-timer port inhomogeneity as dominant constraints (Quan et al., 2021). The silicon-chip network attributes residual offset and non-femtosecond TDEV to chip coupling efficiency instability, DWDM filtering performance, SNSPD detection efficiency, TDC jitter, count-rate-dependent SNSPD time drift, and fiber and component asymmetry (Li et al., 2024). Network proposals add further constraints: lower-jitter detectors, faster time-tagging electronics, low dark-count rates, and practical calibration of system delays are repeatedly identified as prerequisites for single-digit-picosecond performance (Troupe et al., 2022).

The literature therefore supports a restrained interpretation of QCS. Quantum correlations, HOM interference, and multipartite entanglement can improve precision and, in some regimes, security. But the realized performance of a QCS system is governed jointly by the quantum resource and by reciprocity monitoring, detector and timing hardware, oscillator stability, delay asymmetry, and relativistic calibration.