betaFeedback

WorksPhysics and Astronomy

An optical-frequency-comb-enabled unified infrastructure for quantum networks

Abstract

Quantum networks require precision timing and classical communication to coordinate distributed quantum operations, yet these functions are typically supported by separate infrastructures, increasing system complexity and limiting scalability. Here we demonstrate an optical-frequency-comb-enabled infrastructure that unifies precision timing, classical communication and quantum communication over a shared fibre link. By encoding classical data onto the intensity of the comb while retaining its pulse timing, we simultaneously transmit data at 100 Mbit/s and achieve a timing stability of 0.85 fs at 1 s over 100 km. We develop a spectral-engineering strategy to suppress the coupling between data modulation and timing extraction. We further establish a power-dependent model of the excess noise induced in a co-propagating continuous-variable quantum key distribution channel, identifying optical power, wavelength allocation and propagation direction as the principal control parameters. After optimization, the system introduces only 0.0037 shot noise unit of additional excess noise and limits the secret-key-rate penalty to 6.7%. Furthermore, quantum-secured timestamp transmission additionally enables absolute time synchronization with a stability of 0.26 fs at 1 s. These results establish optical frequency combs as a common physical layer for integrated quantum-network services.

Quantum networks are expected to enable fundamentally new capabilities beyond classical networks, including secure communication, distributed quantum computing and enhanced quantum sensing [1, 2, 3, 4, 5, 6, 7, 8]. Recent advances in quantum communication, quantum processors and quantum interfaces have demonstrated key building blocks for such networks [9, 10, 11, 12, 13, 14, 15, 16, 17]. However, realizing practical and scalable quantum networks requires more than quantum information processing technologies themselves. Similar to today’s classical Internet, future quantum networks will rely on a complex infrastructure consisting of quantum channels, classical communication networks and high-precision timing systems to coordinate distributed quantum operations [18, 19].

At present, these infrastructures are developed largely independently. Quantum communication systems require classical channels for control and post-processing, while distributed quantum operations rely on precise timing references provided by dedicated synchronization networks [4, 12, 20, 21, 22], as shown in Figure 1a. The coexistence of multiple independent networks introduces substantial challenges in scalability, resource efficiency and system complexity. Moreover, direct integration of these networks at the signal level remains difficult because quantum signals, classical optical signals and precision timing signals have distinct requirements in terms of optical power and noise tolerance [19, 23]. Therefore, a key challenge for future quantum networks is to develop a unified infrastructure that can support timing, communication and quantum applications within a compatible physical framework.

Concept and architecture of an optical-frequency-comb-enabled unified infrastructure for quantum networks

Figure 1. Concept and architecture of an optical-frequency-comb-enabled unified infrastructure for quantum networks. a, Conventional quantum networks rely on separate classical communication and precision timing networks for functions such as control, post-processing and synchronization. b, The optical frequency comb (OFC) carries precision timing through its pulse timing and classical communication through intensity modulation. c, The OFC and weak quantum states are multiplexed and co-transmitted through wavelength-division multiplexing. d, The unified quantum-network infrastructure proposed in this work. WDM, wavelength-division multiplexing; IM, intensity modulator; BS, beam splitter.

Optical frequency combs (OFCs) provide a promising route towards such an integrated infrastructure [24, 25, 26, 27]. Owing to their intrinsically high temporal resolution and multiple degrees of freedom, optical frequency combs have enabled breakthroughs in precision timing and ranging [28, 29, 30, 31, 32, 33, 34, 35], frequency metrology [36, 37, 38, 39] and high-capacity optical communication [40, 41]. Here, we explore a new role of optical frequency combs as a common physical platform for quantum network infrastructure, rather than merely as a precision measurement tool. By encoding classical communication information onto the intensity domain of optical comb signals while preserving their timing information, timing and communication functions can be fused into a shared optical carrier, as illustrated in Figure 1b. Such an approach can reduce the dependence on parallel supporting networks and provide a compatible foundation for quantum tasks. However, integrating communication functions with precision timing introduces fundamental challenges. Optical modulation used for data transmission inevitably generates spectral sidebands, which may overlap with the timing information carried by optical combs and degrade synchronization performance. Furthermore, as illustrated in Figure 1c, introducing high-peak-power classical optical signals into the same fibre infrastructure as weak quantum states may induce excess noise through phase distortion and photon contamination, which arise from Kerr nonlinearities [42, 43], Raman scattering [44] and optical power leakage due to imperfect channel isolation [45, 46]. Therefore, achieving a practical fused infrastructure requires both suppressing timing–communication coupling and ensuring compatibility with quantum channels.

Here we demonstrate an optical-frequency-comb-enabled infrastructure that unifies precision timing, classical communication and quantum communication over a shared fibre link, as illustrated in Figure 1d. By encoding classical data onto the intensity of the comb while retaining its pulse timing, we simultaneously transmit 100 Mbit s−1100\ \mathrm{Mbit}\ \mathrm{s}^{-1} classical data and achieve a timing stability of 0.85 fs0.85\ \mathrm{fs} at 1 s1\ \mathrm{s} over 100 km100\ \mathrm{km}. We further develop a spectral-engineering strategy that

controls the modulation rate, symbol distribution and modulation depth to minimize the coupling between data modulation and timing extraction. To integrate quantum communication, we establish a power-dependent model of the excess noise induced by the broadband, high-peak-power comb and identify optical power, wavelength allocation and propagation direction as the principal control parameters. With these optimizations, the infrastructure coexists with continuous-variable quantum key distribution (CV-QKD) over the same 100-km link, introducing only 0.0037 shot noise unit (SNU) of additional excess noise and limiting the secret-key-rate penalty to 6.7%. Finally, we achieve quantum-secured timestamp transmission enabling absolute time synchronization with a stability of 0.26 fs0.26\ \mathrm{fs} at 1 s1\ \mathrm{s}. These results establish optical frequency combs as a common physical layer for the timing, communication and security functions required by scalable quantum networks.

Results

Optical-frequency-comb-enabled fusion of timing and communication infrastructures

Optical frequency combs naturally provide multiple degrees of freedom for information encoding. The precisely controlled pulse timing enables high-resolution time–frequency transfer, while the optical intensity can be modulated to encode classical communication data. By combining these two degrees of freedom, timing information and communication data can be simultaneously transmitted through the same optical carrier without requiring an additional communication channel (see Methods for the principle of timing synchronization and the timing–communication fusion).

A practical quantum network requires not only quantum information processing nodes, but also supporting infrastructures that provide accurate timing references and classical information exchange. Conventionally, precision timing and classical communication are implemented through independent optical networks, resulting in additional hardware complexity and limited scalability. Here, we demonstrate the fusion of these two essential infrastructures by using an optical frequency comb as a shared physical carrier, as shown in Figure 2a (see Methods for detailed system architecture).

Fusion of precision timing and classical communication enabled by optical frequency combs

Figure 2. Fusion of precision timing and classical communication enabled by optical frequency combs. a, Experimental architecture. The repetition frequency of the optical frequency comb is locked to an atomic clock to transfer accurate time–frequency information, while classical communication data are encoded in the intensity domain through intensity modulation. DDS, direct digital synthesizer; PID, proportional–integral–derivative controller; LOS, linear optical sampling; AWG, arbitrary waveform generator; ADC, analogue-to-digital converter; EDFA, erbium-doped fibre amplifier. b-e, Simulated spectra showing how different modulation strategies reshape the position and power of communication-induced spectral components relative to the timing-sensitive region. The black lines denote the radio-frequency (RF) comb generated by dual-comb beating for time–frequency recovery, while the coloured ones are communication-induced components. The vertical axes give the relative power spectral density (PSD) in b and the relative spectral amplitude in c–e. b, For NRZ and Manchester coding. c, For unmodulated-symbol fraction D=90%D = 90\% (CSR = 9.0) and D=50%D = 50\% (CSR = 1.0). d, For modulation depths m=0.2m = 0.2, 0.5 and 0.9. e, For fm=100f_m = 100 Hz, 1.3 kHz and 50 kHz, where fmf_m denotes the modulation frequency and increases with the bit rate RbR_b. f, Theoretical analysis of the effects of modulation rate and modulation depth on timing stability. g–i, Experimental characterization of the effect of DD, mm and RbR_b on timing stability. From bottom to top, the curves correspond to D=100%D = 100\%, 80% and 50% (Rb=10R_b = 10 Mbit s−1^{-1}, m=1m = 1) in g; m=0m = 0, 0.5 and 1 (Rb=10R_b = 10 Mbit s−1^{-1}, D=50%D = 50\%) in h; and the unmodulated case, 100 and 4 Mbit s−1^{-1} (m=1m = 1, D=50%D = 50\%) in i. Error bars in g–i indicate one standard error of the TDEV estimates. j, Bit error rate (BER) versus received optical power at 100 Mbit s−1^{-1}, D=50%D = 50\% for m=1m = 1 and m=0.5m = 0.5. Shaded regions indicate 95% confidence intervals. The star marks the −46-46 dBm sensitivity at a BER = 10−610^{-6} without forward error correction (FEC).

The key challenge of such a fused architecture is the interaction between communication modulation and timing extraction. Intensity modulation introduces spectral sidebands around comb modes, which may overlap with the spectral components used for precision timing measurement and appear as additional timing noise because timing is recovered from the modulated dual-comb interferogram.

To overcome this limitation, we develop a spectral-engineering strategy that exploits the relationship between modulation and timing characteristics. By optimizing the coding formats, symbol distribution, modulation depth and modulation rate, the communication–induced dominant spectral components can be separated from the timing information and their contribution to timing instability can be minimized. As shown in Figure 2b, different coding formats reshape the modulation spectrum. For example, compared with non-return-to-zero (NRZ), Manchester coding suppresses low-frequency components and redistributes the spectral power towards higher frequencies, reducing the spectral overlap at the expense of communication efficiency. Meanwhile, the energy of communication–induced components can be further controlled through the fraction of unmodulated symbols DD and the modulation depth mm. A higher DD, as in the yellow lines in Figure 2c, increases the carrier-to-sideband power ratio (CSR), and a smaller mm, as in the blue lines in Figure 2d, suppresses the sideband power. These controls therefore provide a trade-off between suppressing timing–communication coupling and maintaining communication performance, such as communication efficiency or sufficiently low bit-error rate (BER). Furthermore, as Figure 2e shows, increasing the modulation frequency fmf_{\mathrm{m}} shifts the sidebands away from the spectral region carrying the time–frequency information, as illustrated by the 50 kHz case, thereby reducing interference.

Theoretical calculations (as shown in Figure 2f; see Supplementary Note 1 for the derivation) quantify the dependence of timing stability on fmf_{\mathrm{m}} and mm, with stability characterized by time deviation (TDEV) at 1 s. As fmf_{\mathrm{m}} increases, the TDEV first rises in the very-low-frequency region and then exhibits an oscillatory, weakly varying regime owing to the finite temporal window WW used in the practical timing-extraction method. Above the characteristic scale set by 1/W1/W, however, the TDEV exhibits an overall decreasing trend with increasing fmf_{\mathrm{m}}. The former two regimes correspond to modulation rates far below those relevant to practical communication. Hence, within the practical communication range, the communication–induced degradation of timing stability diminishes as the modulation rate increases. Meanwhile, reducing mm further shifts the TDEV downward, resulting in a smaller degradation of timing stability. Furthermore, as illustrated in Figure 2g–Figure 2i, experiments reinforce that these controls enable effective decoupling between communication data and timing information for principle validation.

Using this spectral-engineering strategy, we experimentally realize a fused time–frequency communication link based on an optical frequency comb. We use NRZ coding and set D=50%D = 50\% to maintain communication efficiency, and set m=1m = 1 to ensure a low BER. The system simultaneously transfers precision timing signals and classical communication data over a 100-km fibre link, achieving a timing stability of 0.85 fs at 1 s while supporting a communication rate of 100 Mbit s−1^{-1}, as shown in Figure 2i. Meanwhile, it meets the 10−610^{-6} BER criterion at a received optical power of −46 dBm at 100 Mbit s−1^{-1} without forward-error correction, as shown in Figure 2j. It can also be seen that OFCs, as high signal-to-noise ratio light sources, provide a promising approach for low BER communication. These results demonstrate simultaneous high-speed communication and sub-femtosecond timing stability over a shared optical carrier. Integrated optical frequency combs with higher repetition rates could further increase the achievable communication rate by supporting higher symbol rates for intensity encoding.

Integration of quantum tasks through wavelength-division multiplexing

The fusion of timing and classical communication infrastructures provides a unified optical platform for supporting future quantum networks, thereby mitigating resource constraints and improving scalability. However, practical quantum tasks rely on fragile quantum states, which generally require low-noise transmission conditions and cannot directly share a high-power classical carrier.

Unlike conventional continuous-wave (CW) classical channels, the fused carrier here is a broadband, high-peak-power pulse train composed of densely spaced, mutually coherent comb lines. These comb-specific temporal and spectral features make its coexistence with weak quantum states particularly demanding, which has remained largely unexplored. Therefore, a key challenge is to integrate quantum task layers with the fused infrastructure while preserving the performance of quantum operations. Here, we extend the optical-frequency-comb-enabled infrastructure to quantum tasks through wavelength-division multiplexing (WDM), as shown in Figure 3a–Figure 3b (see Methods for the CV-QKD experimental setup). In this architecture, the optical frequency comb serves as the carrier for the fused time-frequency communication infrastructure, while independent weak quantum states occupy dedicated wavelength channels and coexist within the same fibre link. Such a configuration preserves the spectral separation between classical and quantum signals while enabling them to share the same optical transmission infrastructure.

Integration of quantum tasks with the fused time-frequency communication infrastructure

Figure 3. Integration of quantum tasks with the fused time-frequency communication infrastructure. a, Architecture for integrating quantum tasks. b, Wavelength allocation of the classical and quantum channels. c, Noise mechanisms. The broadband, high-peak-power optical frequency comb induces phase distortion through cross-phase modulation, while Raman scattering, four-wave mixing and residual optical leakage cause photon contamination. Together, these effects contribute to the excess noise of the quantum channel. d, Excess noise versus the mean optical-comb power. The labels f and b denote forward and backward comb propagation relative to the CV-QKD signal, respectively, with the quantum channel fixed at C34. Black, yellow, blue and green curves correspond to C30(f), C32(b); C31(f), C33(b); C32(f), C30(b); and C37(f), C35(b), respectively. ξ0\xi_0 is the mean system excess noise measured with the classical channels turned off, and the grey region indicates its measured fluctuation range. e, SKR versus the mean optical-comb power, with the same colour coding as in d. SKR0_0 and the grey region denote the mean and fluctuation range with the classical channels turned off, respectively.

The main limitation of this signal-level integration arises from noise transferred from classical optical channels into quantum channels, as illustrated in Figure 3c. The classical signals can degrade quantum performance through several mechanisms. Cross-phase modulation (XPM), enhanced by the high peak power of the comb, maps classical intensity fluctuations onto the quantum phase and causes phase distortion [42, 43]. Raman scattering and four-wave mixing (FWM) can generate photons within the quantum receiver band, and once in band, these photons cannot be removed by wavelength filtering and contribute to photon contamination [42, 43, 44]. Residual optical leakage due to finite channel isolation also introduces additional photon contamination [43, 45, 46]. For CV-QKD, these effects appear as excess noise and directly reduce the achievable secret key rate (SKR) [43, 45, 47]. Moreover, compared with the conventional CW laser, this high-power and broad-spectral-bandwidth light can further enhance nonlinear interactions, while its highly coherent spectral structure particularly facilitates FWM. In general, appreciable higher-order nonlinear contributions are indicated because of the complex spectral structure of the OFC carrier.

To quantify and optimize this coexistence, we investigate the dependence of quantum-channel performance on classical optical power, wavelength allocation and propagation direction. On this basis, we construct a power-dependent model for the excess noise,

ξtot(Pc)≃ξ0+c1Pc+c2Pc2+c3Pc3(1)\xi_{\mathrm{tot}}(P_{\mathrm{c}}) \simeq\xi_{0} + c_{1}P_{\mathrm{c}} + c_{2}P_{\mathrm{c}}^{2} + c_{3}P_{\mathrm{c}}^{3} \tag*{(1)}

where PcP_{\mathrm{c}} is the mean optical-comb power, ξ0\xi_{0} is the mean system excess noise measured with the classical channels turned off, and c1c_{1}, c2c_{2} and c3c_{3} are configuration-dependent coefficients that predominantly account for Raman scattering and optical leakage, XPM-induced phase distortion and FWM-induced photon contamination, respectively (see Supplementary Note 2 for the model formulation and derivation).

This model quantitatively links the classical-channel conditions to the quantum excess noise and identifies optical power, wavelength allocation and propagation direction as the principal control parameters. As shown in Figure 3d, the model is fitted to the measured excess-noise data for each channel configuration, and the excess noise increases at an accelerating rate with the mean optical-comb power for all channel allocations. Comparison of the fitted curves further reveals the effects of propagation direction and wavelength allocation. Specifically, two-way time–frequency transfer requires bidirectional comb transmission while the contributions of forward and backward Raman noise are different, resulting in the importance of selecting propagation direction. At medium-to-long distances, as in our 100-km link, backward Raman scattering contributes more strongly than forward Raman scattering. By comparing the blue curve and black curve in Figure 3d, we can see that the backward path benefits from a larger spectral separation from the quantum channel. Meanwhile, the comparison between the black curve and the yellow curve shows that increasing the spectral separation from the quantum channel reduces the overall noise contribution. At comparable spectral separations, placing the classical channels on the longer-wavelength side of quantum channel (C34) yields lower excess noise (yellow curve) than placing them on the shorter-wavelength side (green curve), consistent with weaker anti-Stokes than Stokes scattering into the quantum channel. Figure 3e shows the corresponding SKR results, exhibiting the same dependence on wavelength allocation and propagation direction. Hence, increasing the wavelength separation between classical and quantum channels reduces spectral overlap and suppresses noise transfer, while proper selection of propagation direction further minimizes Raman-induced noise. More importantly, there is always an appropriate power range where phase distortion and photon contamination are suppressed, keeping the excess noise low while maintaining the performance of the fused system. In general, by jointly optimizing propagation direction, wavelength allocation and optical power, the performance of the fused architecture can be optimized.

We experimentally demonstrate the compatibility between the fused time-frequency communication infrastructure and a CV-QKD system over a 100-km fibre link, as indicated by the red stars in Figure 3d. By optimizing the classical channel configuration, including optical power (0.35 mW), wavelength allocation and propagation direction (the blue curve: forward channel is C32 and backward channel is C30), the additional excess noise introduced into the quantum channel is suppressed to 0.0037 SNU, yielding an SKR of 29.67 kbps and thus resulting in a secret key rate degradation of 6.7% compared with the standalone quantum communication system (see Methods for the secret-key-rate calculation).

These results demonstrate that the optical-frequency-comb-enabled infrastructure can provide simultaneous support for precision timing, classical communication and quantum tasks, offering a practical pathway towards integrated and scalable quantum network deployment.

Secure absolute time synchronization enabled by the unified quantum network infrastructure

Beyond supporting quantum communication tasks, a practical quantum network requires reliable absolute synchronization among distributed nodes. The encoded sequences serving as timestamps carry absolute timing information, complementing the high-stability relative time transfer provided by the optical comb. Conventional absolute time synchronization relies on dedicated timing infrastructures and communication infrastructures, while security functions are typically implemented independently through cryptographic protocols to encrypt the timestamps. The separation between timing and security limits the ability to provide data–trustworthy synchronization in distributed quantum networks, where both synchronization accuracy and information security are essential.

The unified infrastructure introduced here provides a new approach to secure timing services by combining precision time transfer with quantum-secured information exchange within the same optical network.

To demonstrate this capability, we integrate the precision time transfer system with a CV-QKD channel through WDM. As illustrated in Figure 4a, in this architecture, the timing information is transferred using the optical-comb-based fused infrastructure, while the quantum communication channel provides the information-theoretically secure timestamps by applying one-time-pad encryption through a bitwise exclusive-OR operation using quantum-generated keys. As shown in Figure 4b, the timestamp occupies only a negligible fraction of the transmitted data and therefore has virtually no effect on the stability of the time–frequency information. Hence, the coexistence of these functions enables absolute time synchronization with both high stability and enhanced security with quantum-key protection, compared with conventional synchronization schemes that rely solely on classical encryption strategies.

Figure 4. Secure absolute time synchronization enabled by the unified quantum network infrastructure. a, Unified quantum-network infrastructure for secure timestamp transmission. b, Secure integration mechanism for timestamp transmission. c, Time-difference increments between adjacent acquisition points, showing the short-term timing fluctuations with and without timestamp transmission. d, Experimental results of timing stability. Error bars indicate one standard error of the TDEV estimates. The results obtained with and without timestamp transmission are identical at short time-scales.

Experimentally, we demonstrate secure absolute time synchronization over a 100-km spooled fibre (see Methods for timestamp loading and timing-stability analysis). As shown in Figure 4c, the amplitudes of time-difference increment remain nearly unchanged and are at the sub-femtosecond level, while maintaining secure timestamp exchange enabled by the quantum communication channel. The fused system achieves a TDEV of 0.26 fs at 1 s, as shown in Figure 4d. This performance remains consistent with that of the standalone precision timing system, indicating that the integration of security functions does not compromise timing stability.

This demonstration highlights a new capability enabled by the unified infrastructure: precision timing, classical communication and quantum security can be jointly provided within a common optical network architecture. Such a capability is expected to play an important role in future distributed quantum computing, quantum sensing networks and large-scale quantum communication systems.

Discussion

The realization of large-scale quantum networks requires the coordinated operation of multiple heterogeneous systems, including quantum communication channels, classical control networks and precision timing infrastructures [48]. Existing approaches typically rely on separately deployed systems, which increases hardware complexity and limits scalability. The unified infrastructure demonstrated here provides an alternative paradigm, where optical frequency combs enable the convergence of timing and communication functions while maintaining compatibility with weak quantum states.

A key advantage of this architecture is that it does not require replacing existing quantum technologies or communication infrastructures. Instead, it introduces an additional integration layer that enables quantum tasks to access shared timing and communication resources. Such an approach is compatible with existing fibre-optic networks and can potentially support diverse quantum applications, including distributed quantum computing, quantum sensing networks and large-scale quantum communication systems, as illustrated in Figure 5a.

Figure 5. Towards scalable quantum networks with unified optical infrastructures. a, Prospective functionalities of the fused network. b, Extension from point-to-point links to network-level connectivity.

Several challenges remain for practical deployment. Extending the current demonstration to large-scale quantum networks shown in Figure 5b will require improvements in integrated optical-frequency-comb sources, network-level synchronization protocols and long-distance quantum communication technologies. In particular, compact and robust comb generation, advanced channel management and improved suppression of nonlinear and scattering-induced noise will be important for metropolitan and intercity quantum networks. Future developments in integrated photonics may provide a pathway towards miniaturized and scalable implementations of the proposed architecture.

Beyond individual quantum links, the concept of infrastructure fusion may provide a general framework for future information networks, where precision timing, classical communication and quantum information processing are no longer treated as independent systems. As precise time–frequency dissemination also underpins integrated communication and navigation networks, such fusion naturally extends to these systems, allowing them to be built on the same comb-based physical foundation. By establishing optical frequency combs as a common physical foundation for heterogeneous network functions, this work opens new opportunities for building secure, scalable and integrated quantum network infrastructures.

Methods

Principle of dual-comb two-way time–frequency transfer and timing–communication fusion

An optical frequency comb consists of a train of equally spaced femtosecond pulses in the time domain and a series of equally spaced comb lines in the frequency domain. At each site, the repetition frequencies of the combs are phase-locked to the local clock, while their carrier-envelope offset frequencies are left free-running. The two combs are denoted as the clock comb with repetition frequency frf_r and the local oscillator (LO) comb with repetition frequency fr+Δfrf_r+\Delta f_r, respectively. The clock comb maps the local timescale onto a stable optical pulse train, whose pulse-to-pulse spacing and pulse timing carry frequency and time information, respectively. Before transmission, classical communication data are encoded onto the intensity of the clock comb through intensity modulation while retaining its pulse timing. At the receiver, the same optical-comb signal is used for both data recovery and timing extraction, thereby fusing time–frequency transfer and classical communication on a shared optical carrier.

High-precision timing extraction is performed using dual-comb linear optical sampling (LOS). At each site, the LO comb interferes separately with the local clock comb and the received remote clock comb, generating the reference and target interferograms, respectively. The small repetition-rate offset between the two combs causes the relative pulse delay to scan from pulse to pulse, producing a time-stretched cross-correlation waveform. The single-step delay is

δτs=∣1fr−1fr+Δfr∣≃Δfrfr2,Δfr≪fr.(2)\delta\tau_{\mathrm{s}}=\left|\frac{1}{f_{\mathrm{r}}}-\frac{1}{f_{\mathrm{r}}+\Delta f_{\mathrm{r}}}\right|\simeq\frac{\Delta f_{\mathrm{r}}}{f_{\mathrm{r}}^{2}},\qquad\Delta f_{\mathrm{r}}\ll f_{\mathrm{r}}. \tag*{(2)}

Accordingly, a physical time offset δt\delta t is stretched on the laboratory timescale as

δtlab=αδt,α=frΔfr,(3)\delta t_{\mathrm{lab}}=\alpha\delta t,\qquad\alpha=\frac{f_{\mathrm{r}}}{\Delta f_{\mathrm{r}}}, \tag*{(3)}

where α\alpha is the LOS time-stretching factor [29]. The LOS update period is

TLOS=1Δfr.(4)T_{\mathrm{LOS}}=\frac{1}{\Delta f_{\mathrm{r}}}. \tag*{(4)}

This process maps the pulse cross-correlation onto the laboratory timescale, magnifying both its envelope width and the relative pulse delay by a factor of α\alpha. The resulting LOS interferogram can be directly digitized by electronic acquisition systems for precise timing extraction.

For a reference or target interferogram ii, the envelope Ai(t)A_i(t) is first extracted using a Hilbert transform, and its arrival time is then determined from the envelope centroid [28],

tilab=∫WtAi(t) dt∫WAi(t) dt,i=1,2,3,…,(5)t_i^{\mathrm{lab}}=\frac{\int_W tA_i(t)\,\mathrm{d}t}{\int_W A_i(t)\,\mathrm{d}t},\qquad i=1,2,3,\ldots, \tag*{(5)}

where WW denotes the timing-extraction window. At sites A and B, the respective time differences on the laboratory timescale, tAlabt_A^{\mathrm{lab}} and tBlabt_B^{\mathrm{lab}}, are obtained by comparing the centroid positions of the corresponding reference and target interferograms. The effective time differences are

tA=tAlabα,tB=tBlabα.(6)t_A=\frac{t_A^{\mathrm{lab}}}{\alpha},\qquad t_B=\frac{t_B^{\mathrm{lab}}}{\alpha}. \tag*{(6)}

Let ΔtAB\Delta t_{AB} denote the clock offset at site B relative to site A, and let τAB\tau_{AB} and τBA\tau_{BA} denote the A-to-B and B-to-A propagation delays, respectively. The measured time intervals are then

tA=τBA−ΔtAB,tB=τAB+ΔtAB.(7)t_A=\tau_{BA}-\Delta t_{AB},\qquad t_B=\tau_{AB}+\Delta t_{AB}. \tag*{(7)}

The two-way combination gives

ΔtAB=tB−tA2+tNR,tNR=τBA−τAB2,(8)\Delta t_{AB}=\frac{t_B-t_A}{2}+t_{\mathrm{NR}},\qquad t_{\mathrm{NR}}=\frac{\tau_{BA}-\tau_{AB}}{2}, \tag*{(8)}

where tNRt_{\mathrm{NR}} is the non-reciprocal link delay. Under link reciprocity, that is, τAB≃τBA\tau_{AB}\simeq\tau_{BA},

ΔtAB≃tB−tA2.(9)\Delta t_{AB}\simeq\frac{t_B-t_A}{2}. \tag*{(9)}

Consequently, the reciprocal propagation delay and its fluctuations are rejected as common-mode terms in the two-way measurement, leaving only the inter-site clock offset and residual link non-reciprocity. The fractional frequency offset is further obtained from

yAB(t)=dΔtAB(t)dt.(10)y_{AB}(t)=\frac{\mathrm{d}\Delta t_{AB}(t)}{\mathrm{d}t}. \tag*{(10)}

LOS resolves the relative clock offset within one pulse period at high resolution, while the integer-cycle ambiguity is resolved by the timestamp.

Experimental architecture for timing–communication fusion

The experimental architecture is shown in Figure 2a. Sites A and B are connected by a 100-km spooled fibre. Comb A and Comb B serve as the clock combs at a repetition frequency of 100 MHz, whereas Comb X1_1 and Comb X2_2 serve as the corresponding LO combs at 100.001 MHz. At each site, a clock-referenced direct digital synthesizer (DDS) supplies the reference signals for phase-locking the comb repetition frequencies through proportional–integral–derivative (PID) feedback.

The clock-comb signals propagate in opposite directions over the shared link. At site A, an arbitrary waveform generator (AWG) drives an intensity modulator (IM) to encode classical data onto the transmitted output of Comb A. A local branch is split off before the IM to generate the reference interferogram. At site B, a beam splitter (BS) divides the received modulated signal between the LOS timing-extraction branch and the communication receiver. In the communication branch, the signal is amplified by an erbium-doped fibre amplifier (EDFA), photodetected and digitized by an analogue-to-digital converter (ADC) for data recovery.

At each site, the LO comb interferes separately with the local clock-comb output and the received remote clock-comb signal to generate the reference and target interferograms, respectively. These interferograms are digitized by ADCs, and the time intervals are extracted from their envelope centroids. The time intervals obtained at the two sites are then combined to determine the inter-site time offset using the procedure described in the preceding subsection.

To isolate the system-induced timing instability from the free-running offset between independent remote clocks, we employ a common-clock configuration in which the compared signals share the same frequency reference, so that the measured time difference reflects the residual time offset of the system. The resulting time-difference sequence is then Kalman-filtered to suppress measurement noise. The reported TDEV is used to characterize the stability of this residual time offset.

Experimental setup of CV-QKD

We demonstrate the co-transmission of Gaussian-modulated coherent-state continuous-variable quantum key distribution (GMCS CV-QKD) signals and classical OFC signals over a single fibre link in the C band using dense WDM. Before multiplexing, the quantum signals are prepared with a target modulation variance VAV_{\mathrm{A}}, while the OFC signals are amplified by an erbium-doped fibre amplifier placed in the classical-comb path to compensate for the transmission loss. The quantum and amplified OFC signals are then multiplexed through the WDM system and transmitted over the shared fibre link.

At Alice, a narrow-linewidth CW laser operating at 1550.12 nm, corresponding to the C34 channel, serves as the optical carrier for quantum-state preparation. Coherent states are prepared using an FPGA-driven IQ modulator equipped with an integrated 16-bit digital-to-analog converter, generating GMCS at a repetition rate of 100 MHz. A pilot-sequential scheme is employed to alternately transmit reference pulses and quantum signal pulses, maintaining the quantum signals at the weak-light level while providing the reference required for phase recovery. Finally, a variable optical attenuator is employed to fine-tune the modulation variance of the transmitted GMCS to the optimal value VAV_{\mathrm{A}}.

At Bob, quantum-state measurements are performed via heterodyne detection using an independent local oscillator. The detector output is digitized in real time by an analog-to-digital converter operating at 1 GSa/s. The acquired samples are subsequently subjected to real-time digital signal processing for phase recovery and compensation to recover the quantum signals. The recovered data are then subjected to parameter estimation, information reconciliation, and privacy amplification to extract the final secret key.

Secret key rate calculation of CV-QKD

For a GMCS CV-QKD system with reverse reconciliation, the secret key rate in the asymptotic regime can be expressed as [49, 50]:

K[ξtot(Pc),T]=Reff{βIAB−χBE}+(11)K\left[\xi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right),T\right]=R_{\mathrm{eff}}\left\{\beta I_{\mathrm{AB}}-\chi_{\mathrm{BE}}\right\}_{+} \tag*{(11)}

where ReffR_{\mathrm{eff}} denotes the effective quantum symbol rate, β\beta is the reconciliation efficiency, IABI_{\mathrm{AB}} represents the Shannon mutual information between Alice and Bob, and χBE\chi_{\mathrm{BE}} denotes the maximum information accessible to Eve.

In this experiment, the classical optical signals and CV-QKD signals are co-transmitted over the same fibre link. The classical optical signals affect the CV-QKD system primarily by introducing additional channel noise. Denoting the average classical optical power by PcP_{\mathrm{c}}, the experimentally measured total excess noise, denoted by ξtot(Pc)\xi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right), is substituted into Eq. (11). Accordingly, the total channel noise referred to the channel input can be expressed as

χline(Pc)=1T−1+ξtot(Pc).\chi_{\mathrm{line}}\left(P_{\mathrm{c}}\right)=\frac{1}{T}-1+\xi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right).

For heterodyne detection, the detection-added noise is given by

χhet=1+(1−η)+2νelη.\chi_{\mathrm{het}}=\frac{1+(1-\eta)+2\nu_{\mathrm{el}}}{\eta}.

where η\eta denotes the detection efficiency and νel\nu_{\mathrm{el}} is the normalized electronic noise of the detector. The total noise referred to the channel input can therefore be written as

χtot(Pc)=χline(Pc)+χhetT.\chi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right)=\chi_{\mathrm{line}}\left(P_{\mathrm{c}}\right)+\frac{\chi_{\mathrm{het}}}{T}.

Considering the dependence of the total noise on the average classical optical power PcP_{\mathrm{c}}, the mutual information between Alice and Bob is expressed as

IAB(Pc)=log⁡2[V+χtot(Pc)1+χtot(Pc)].I_{\mathrm{AB}}\left(P_{\mathrm{c}}\right)=\log_{2}\left[\frac{V+\chi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right)}{1+\chi_{\mathrm{tot}}\left(P_{\mathrm{c}}\right)}\right].

where V=VA+1V=V_{\mathrm{A}}+1, and VAV_{\mathrm{A}} denotes the modulation variance.

The average classical optical power PcP_{\mathrm{c}}, effective quantum symbol rate ReffR_{\mathrm{eff}}, reconciliation efficiency β\beta, modulation variance VAV_{\mathrm{A}}, channel transmittance TT, electronic noise νel\nu_{\mathrm{el}}, and detection efficiency η\eta are experimentally measured or predetermined system parameters. The Holevo information between Bob and Eve is given by

χBE(Pc)=G[λ1(Pc)−12]+G[λ2(Pc)−12]−G[λ3(Pc)−12]−G[λ4(Pc)−12]−G[λ5(Pc)−12],\chi_{\mathrm{BE}}(P_{\mathrm{c}})=G\left[\frac{\lambda_{1}(P_{\mathrm{c}})-1}{2}\right]+G\left[\frac{\lambda_{2}(P_{\mathrm{c}})-1}{2}\right]-G\left[\frac{\lambda_{3}(P_{\mathrm{c}})-1}{2}\right]-G\left[\frac{\lambda_{4}(P_{\mathrm{c}})-1}{2}\right]-G\left[\frac{\lambda_{5}(P_{\mathrm{c}})-1}{2}\right],

where G(x)G(x) is defined as

G(x)=(x+1)log⁡2(x+1)−xlog⁡2x.G(x)=(x+1)\log_{2}(x+1)-x\log_{2}x.

We further define

A(Pc)=V2(1−2T)+2T+T2[V+χline(Pc)]2,A(P_{\mathrm{c}})=V^{2}(1-2T)+2T+T^{2}[V+\chi_{\mathrm{line}}(P_{\mathrm{c}})]^{2},
B(Pc)=T2[Vχline(Pc)+1]2,B(P_{\mathrm{c}})=T^{2}[V\chi_{\mathrm{line}}(P_{\mathrm{c}})+1]^{2},
C(Pc)=1T2[V+χtot(Pc)]2{A(Pc)χhet2+B(Pc)+1+2χhet[VB(Pc)+T(V+χline(Pc))]+2T(V2−1)},C(P_{\mathrm{c}})=\frac{1}{T^{2}[V+\chi_{\mathrm{tot}}(P_{\mathrm{c}})]^{2}}\left\{A(P_{\mathrm{c}})\chi_{\mathrm{het}}^{2}+B(P_{\mathrm{c}})+1+2\chi_{\mathrm{het}}\left[V\sqrt{B(P_{\mathrm{c}})}+T(V+\chi_{\mathrm{line}}(P_{\mathrm{c}}))\right]+2T(V^{2}-1)\right\},
D(Pc)=[V+B(Pc)χhetT[V+χtot(Pc)]]2.D(P_{\mathrm{c}})=\left[\frac{V+\sqrt{B(P_{\mathrm{c}})}\chi_{\mathrm{het}}}{T[V+\chi_{\mathrm{tot}}(P_{\mathrm{c}})]}\right]^{2}.

Here, the symplectic eigenvalues can then be expressed as functions of PcP_{\mathrm{c}}:

λ1(Pc)=A(Pc)+A2(Pc)−4B(Pc)2,\lambda_{1}(P_{\mathrm{c}})=\sqrt{\frac{A(P_{\mathrm{c}})+\sqrt{A^{2}(P_{\mathrm{c}})-4B(P_{\mathrm{c}})}}{2}},
λ2(Pc)=A(Pc)−A2(Pc)−4B(Pc)2,\lambda_{2}(P_{\mathrm{c}})=\sqrt{\frac{A(P_{\mathrm{c}})-\sqrt{A^{2}(P_{\mathrm{c}})-4B(P_{\mathrm{c}})}}{2}},
λ3(Pc)=C(Pc)+C2(Pc)−4D(Pc)2,\lambda_{3}(P_{\mathrm{c}})=\sqrt{\frac{C(P_{\mathrm{c}})+\sqrt{C^{2}(P_{\mathrm{c}})-4D(P_{\mathrm{c}})}}{2}},
λ4(Pc)=C(Pc)−C2(Pc)−4D(Pc)2,\lambda_{4}(P_{\mathrm{c}})=\sqrt{\frac{C(P_{\mathrm{c}})-\sqrt{C^{2}(P_{\mathrm{c}})-4D(P_{\mathrm{c}})}}{2}},
λ5(Pc)=1.\lambda_{5}(P_{\mathrm{c}})=1.

By substituting the above expressions into the secret key rate formula, the dependence of the secret key rate KK on the average classical optical power PcP_{\mathrm{c}} can be obtained.

Timestamp loading and timing-stability analysis

In the secure absolute time-synchronization experiment, a 1-PPS signal triggers one 76-bit timestamp packet per second (see Supplementary Note 3 for the timestamp format). The timestamp is protected by quantum keys and loaded onto the optical-comb carrier through intensity modulation at the bit rate of 100 Mbit s−1^{-1}, using NRZ coding with D=50%D=50\% and m=1m=1. Each timestamp has a duration of 760 ns, corresponding to an average timestamp transmission rate of 76 bit s−1^{-1} and a temporal occupancy of only 7.6×10−77.6 \times10^{-7}. This extremely low temporal occupancy makes the impact of timestamp loading on timing stability negligible.

For Figure 4c, the time-difference increments are obtained by differencing adjacent samples and subtracting the global mean of the full increment sequence. The nearly identical fluctuation amplitudes with and without timestamp loading further confirm that timestamp modulation has negligible impact on time–frequency-transfer stability.

Data availability

The data that support the findings of this study are not publicly available. The data are available from the authors upon reasonable request.

Code availability

The codes that support the findings of this study are available on reasonable request from the corresponding authors.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (62201012, 62571006) and the Fundamental Research Funds for the Central Universities (2026CX01008).

Author contributions

Y.L., K.X., Z.C. and H.G. conceived the idea and initiated the project. Y.L., X.W., Z.C. developed the concept of the optical-frequency-comb-enabled unified infrastructure, and Y.L., Z.R., Z.C., K.X., H.G. designed the overall experimental architecture. Y.L., Z.R., Y.F.Z., Z.L., X.W., Z.C. designed and constructed the time–frequency transfer system, and Y.L., Z.R., Y.K.Z., Z.L., Y.F.Z., Y.H.S., and Z.C. performed the experiments. Y.L. and Z.R. developed the theoretical model of modulation-induced timing noise, carried out the numerical simulations and analysed the timing-stability data. Y.L., Y.S., L.Y., and Z.C. designed the classical communication scheme and performed the communication experiments. J.M., Y.H.S., and X.W. designed, constructed and operated the CV-QKD system. Y.L., J.M. and Z.L. performed the CV-QKD-related experiments in the fused system. Y.L. and Z.C. wrote the original manuscript. All authors participated in data analysis and interpretation, reviewed and edited the manuscript, and approved its final version.

Competing interests

The authors declare no competing interests.

Supplementary Information

Supplementary Note 1. Analysis of the effect of intensity modulation on timing stability

To analyze the dependence of linear optical sampling (LOS) timing stability, we consider a single-frequency sinusoidal intensity modulation and evaluate its response under centroid-based timing extraction. Practical modulation waveforms can be decomposed into sinusoidal components, making single-frequency analysis a natural basis for evaluating their impact on timing stability and be able to serve as a proof of principle. Also, we used periodic sequences for principle validation in experiments to ensure a fair comparison among different communication strategies. Let tt denote the time on the laboratory-time axis, and let the signal intensity be

P(t)=P0[1+msin⁡(2πfmt)],0≤m<1.P(t) = P_0 [1 + m \sin(2\pi f_m t)], \qquad0 \le m < 1.

Let the center of the timing-extraction window be tct_c. For the symmetric unmodulated reference, the centroid is

C0(tc)=tc.C_0(t_c) = t_c.

The centroid position of the signal is

C(tc)=∫tc−W/2tc+W/2tP(t) dt∫tc−W/2tc+W/2P(t) dt.C(t_c) = \frac{\int_{t_c-W/2}^{t_c+W/2} tP(t)\,\mathrm{d}t}{\int_{t_c-W/2}^{t_c+W/2} P(t)\,\mathrm{d}t}.

The relative time on the oscilloscope is defined as

u=t−tc,−W2≤u≤W2,u = t - t_c, \qquad-\frac{W}{2} \le u \le\frac{W}{2},

and the modulation-induced time offset is

x(tc)=C(tc)−tc.x(t_c) = C(t_c) - t_c.

Rewriting the denominator in terms of uu and expanding gives

E(tc)=∫−W/2W/2P(tc+u) du=P0∫−W/2W/2[1+msin⁡(2πfmtc)cos⁡(2πfmu)+mcos⁡(2πfmtc)sin⁡(2πfmu)] du=P0W[1+msin⁡(2πfmtc)sin⁡(πfmW)πfmW].\begin{aligned} E(t_c) &= \int_{-W/2}^{W/2} P(t_c+u)\,\mathrm{d}u \\ &= P_0 \int_{-W/2}^{W/2} \left[1 + m\sin(2\pi f_m t_c)\cos(2\pi f_m u) + m\cos(2\pi f_m t_c)\sin(2\pi f_m u)\right]\,\mathrm{d}u \\ &= P_0W\left[1 + m\sin(2\pi f_m t_c)\frac{\sin(\pi f_m W)}{\pi f_m W}\right]. \end{aligned}

The numerator is

M(tc)=∫−W/2W/2(tc+u)P(tc+u) du=tc∫−W/2W/2P(tc+u) du+∫−W/2W/2uP(tc+u) du=tcE(tc)+∫−W/2W/2uP(tc+u) du.\begin{aligned} M(t_c) &= \int_{-W/2}^{W/2} (t_c+u)P(t_c+u)\,\mathrm{d}u \\ &= t_c\int_{-W/2}^{W/2} P(t_c+u)\,\mathrm{d}u + \int_{-W/2}^{W/2} uP(t_c+u)\,\mathrm{d}u \\ &= t_cE(t_c) + \int_{-W/2}^{W/2} uP(t_c+u)\,\mathrm{d}u. \end{aligned}

The centroid shift is therefore

x(tc)=C(tc)−tc=∫−W/2W/2uP(tc+u) du∫−W/2W/2P(tc+u) du.(12)x(t_c) = C(t_c) - t_c = \frac{\int_{-W/2}^{W/2} uP(t_c+u)\,\mathrm{d}u}{\int_{-W/2}^{W/2} P(t_c+u)\,\mathrm{d}u}. \tag*{(12)}

Substituting P(tc+u)P(t_{\mathrm{c}}+u) gives

∫−W/2W/2uP(tc+u) du=P0∫−W/2W/2u[1+msin⁡(2πfmtc)cos⁡(2πfmu)+mcos⁡(2πfmtc)sin⁡(2πfmu)]du=0+P0mcos⁡(2πfmtc)∫−W/2W/2usin⁡(2πfmu) du=−P0Wm2πfm[cos⁡(πfmW)−sin⁡(πfmW)πfmW]cos⁡(2πfmtc).\begin{aligned} \int_{-W/2}^{W/2} uP(t_{\mathrm{c}}+u)\,\mathrm{d}u=P_0\int_{-W/2}^{W/2}u\left[1+m\sin(2\pi f_{\mathrm{m}}t_{\mathrm{c}})\cos(2\pi f_{\mathrm{m}}u)+m\cos(2\pi f_{\mathrm{m}}t_{\mathrm{c}})\sin(2\pi f_{\mathrm{m}}u)\right]\mathrm{d}u \\ &=0+P_0m\cos(2\pi f_{\mathrm{m}}t_{\mathrm{c}})\int_{-W/2}^{W/2}u\sin(2\pi f_{\mathrm{m}}u)\,\mathrm{d}u \\ &=-P_0W\frac{m}{2\pi f_{\mathrm{m}}}\left[\cos(\pi f_{\mathrm{m}}W)-\frac{\sin(\pi f_{\mathrm{m}}W)}{\pi f_{\mathrm{m}}W}\right]\cos(2\pi f_{\mathrm{m}}t_{\mathrm{c}}). \end{aligned}

The exact centroid offset is therefore

xstrict(tc)=−m2πfm[cos⁡(πfmW)−sin⁡(πfmW)πfmW]⋅cos⁡(2πfmtc)1+msin⁡(πfmW)πfmWsin⁡(2πfmtc).(13)x_{\mathrm{strict}}(t_{\mathrm{c}})=-\frac{m}{2\pi f_{\mathrm{m}}}\left[\cos(\pi f_{\mathrm{m}}W)-\frac{\sin(\pi f_{\mathrm{m}}W)}{\pi f_{\mathrm{m}}W}\right]\cdot\frac{\cos(2\pi f_{\mathrm{m}}t_{\mathrm{c}})}{1+m\frac{\sin(\pi f_{\mathrm{m}}W)}{\pi f_{\mathrm{m}}W}\sin(2\pi f_{\mathrm{m}}t_{\mathrm{c}})}. \tag*{(13)}

Let B=−m2πfm[cos⁡(πfmW)−sin⁡(πfmW)πfmW]B=-\frac{m}{2\pi f_{\mathrm{m}}}\left[\cos(\pi f_{\mathrm{m}}W)-\frac{\sin(\pi f_{\mathrm{m}}W)}{\pi f_{\mathrm{m}}W}\right], q=msin⁡(πfmW)πfmWq=m\frac{\sin(\pi f_{\mathrm{m}}W)}{\pi f_{\mathrm{m}}W}, ψ=2πfmtc\psi=2\pi f_{\mathrm{m}}t_{\mathrm{c}}. Equation (S2) then becomes xstrict(ψ)=Bcos⁡ψ1+qsin⁡ψx_{\mathrm{strict}}(\psi)=B\frac{\cos\psi}{1+q\sin\psi}. Further defining θ=ψ−π2\theta=\psi-\frac{\pi}{2} gives

xstrict(θ)=−Bsin⁡θ1+qcos⁡θ.x_{\mathrm{strict}}(\theta)=-B\frac{\sin\theta}{1+q\cos\theta}.

Since ∣q∣<1|q|<1, we define ρ=q1+1−q2\rho=\frac{q}{1+\sqrt{1-q^2}}, such that ∣ρ∣<1|\rho|<1, q=2ρ1+ρ2q=\frac{2\rho}{1+\rho^2}. Using the identity

∑k=1∞rksin⁡(kθ)=rsin⁡θ1−2rcos⁡θ+r2,∣r∣<1,\sum_{k=1}^{\infty}r^k\sin(k\theta)=\frac{r\sin\theta}{1-2r\cos\theta+r^2},\qquad|r|<1,

and taking r=−ρr=-\rho, we obtain ∑k=1∞(−ρ)ksin⁡(kθ)=−ρsin⁡θ1+2ρcos⁡θ+ρ2=−q2sin⁡θ1+qcos⁡θ\sum_{k=1}^{\infty}(-\rho)^k\sin(k\theta)=-\frac{\rho\sin\theta}{1+2\rho\cos\theta+\rho^2}=-\frac{q}{2}\frac{\sin\theta}{1+q\cos\theta}. Hence,

−sin⁡θ1+qcos⁡θ=2q∑k=1∞(−ρ)ksin⁡(kθ).-\frac{\sin\theta}{1+q\cos\theta}=\frac{2}{q}\sum_{k=1}^{\infty}(-\rho)^k\sin(k\theta).

Therefore,

xstrict(θ)=2Bq∑k=1∞(−ρ)ksin⁡(kθ),q≠0,x_{\mathrm{strict}}(\theta)=\frac{2B}{q}\sum_{k=1}^{\infty}(-\rho)^k\sin(k\theta),\qquad q\ne0,

or equivalently,

xstrict(tc)=∑k=1∞Ck(fm,m,W)sin⁡[k(2πfmtc−π2)],x_{\mathrm{strict}}(t_{\mathrm{c}})=\sum_{k=1}^{\infty}C_k(f_{\mathrm{m}},m,W)\sin\left[k\left(2\pi f_{\mathrm{m}}t_{\mathrm{c}}-\frac{\pi}{2}\right)\right],

where Ck(fm,m,W)=2Bq(−ρ)kC_k(f_{\mathrm{m}},m,W)=\frac{2B}{q}(-\rho)^k.

Defining the amplitude of the kkth harmonic as Ak=∣Ck∣A_k=|C_k|, we have

xstrict(tc)=∑k=1∞Ak(fm,m,W)sin⁡(2πkfmtc+ϕk),x_{\mathrm{strict}}(t_{\mathrm{c}})=\sum_{k=1}^{\infty}A_k(f_{\mathrm{m}},m,W)\sin(2\pi kf_{\mathrm{m}}t_{\mathrm{c}}+\phi_k),

where

Ak(fm,m,W)=∣cos⁡X−sinc⁡(X)∣πfm∣sinc⁡(X)∣[∣msinc⁡(X)∣1+1−m2sinc⁡2(X)]k.(14)A_k(f_{\mathrm{m}},m,W)=\frac{|\cos X-\operatorname{sinc}(X)|}{\pi f_{\mathrm{m}}|\operatorname{sinc}(X)|}\left[\frac{|m\operatorname{sinc}(X)|}{1+\sqrt{1-m^2\operatorname{sinc}^2(X)}}\right]^k. \tag*{(14)}

with X=πfmWX=\pi f_mW, sinc⁡(X)=sin⁡XX\operatorname{sinc}(X)=\frac{\sin X}{X}. The initial phase of the kkth harmonic is ϕk=−kπ2+1−sgn⁡(Ck)2π\phi_k=-\frac{k\pi}{2}+\frac{1-\operatorname{sgn}(C_k)}{2}\pi (mod 2π2\pi). For q=0q=0, the corresponding expressions are understood in the continuous limit q→0q\to0.

For a general timing-error process x(t)x(t), the TDEV is obtained from its one-sided power spectral density Sx(f)S_x(f) and the TDEV transfer function HTDEV(f;τ)H_{\mathrm{TDEV}}(f;\tau) as [51]

σx2(τ)=∫0∞Sx(f)∣HTDEV(f;τ)∣2 df,(15)\sigma_x^2(\tau)=\int_0^\infty S_x(f)\lvert H_{\mathrm{TDEV}}(f;\tau)\rvert^2\,\mathrm{d}f, \tag*{(15)}

where ∣HTDEV(f;τ)∣=83∣sin⁡3(πfτ)πfτ∣\lvert H_{\mathrm{TDEV}}(f;\tau)\rvert=\sqrt{\frac{8}{3}}\left\lvert\frac{\sin^3(\pi f\tau)}{\pi f\tau}\right\rvert, and Sx(f)=∑k=1∞Ak2(fm,m,W)2δ(f−kfm)S_x(f)=\sum_{k=1}^{\infty}\frac{A_k^2(f_m,m,W)}{2}\delta(f-kf_m). Substitution into Eq. (15) gives

σx2(τ)=∫0∞∑k=1∞Ak2(fm,m,W)2δ(f−kfm)⋅∣HTDEV(f;τ)∣2 df.\sigma_x^2(\tau)=\int_0^\infty\sum_{k=1}^{\infty}\frac{A_k^2(f_m,m,W)}{2}\delta(f-kf_m)\cdot\lvert H_{\mathrm{TDEV}}(f;\tau)\rvert^2\,\mathrm{d}f.

Using the Dirac-delta identity ∫0∞δ(f−kfm)∣HTDEV(f;τ)∣2 df=∣HTDEV(kfm;τ)∣2\int_0^\infty\delta(f-kf_m)\lvert H_{\mathrm{TDEV}}(f;\tau)\rvert^2\,\mathrm{d}f=\lvert H_{\mathrm{TDEV}}(kf_m;\tau)\rvert^2, the integral reduces to the discrete sum σx2(τ;fm,m,W)=∑k=1∞Ak2(fm,m,W)2∣HTDEV(kfm;τ)∣2\sigma_x^2(\tau;f_m,m,W)=\sum_{k=1}^{\infty}\frac{A_k^2(f_m,m,W)}{2}\lvert H_{\mathrm{TDEV}}(kf_m;\tau)\rvert^2, that is,

σx2(τ;fm,m,W)=43∑k=1∞Ak2(fm,m,W)[∣sin⁡3(πkfmτ)∣πkfmτ]2.\sigma_x^2(\tau;f_m,m,W)=\frac{4}{3}\sum_{k=1}^{\infty}A_k^2(f_m,m,W)\left[\frac{\lvert\sin^3(\pi kf_m\tau)\rvert}{\pi kf_m\tau}\right]^2.

The TDEV is therefore

σx(τ;fm,m,W)={43∑k=1∞Ak2(fm,m,W)[∣sin⁡3(πkfmτ)∣πkfmτ]2}1/2.(16)\sigma_x(\tau;f_m,m,W)=\left\{\frac{4}{3}\sum_{k=1}^{\infty}A_k^2(f_m,m,W)\left[\frac{\lvert\sin^3(\pi kf_m\tau)\rvert}{\pi kf_m\tau}\right]^2\right\}^{1/2}. \tag*{(16)}

Excess-noise model for CV-QKD

After fusion of the communication and time-frequency layers, the impact of the classical channels on the quantum channel can be modeled as

ξtot(Pc)≃ξ0+c1Pc+c2Pc2+c3Pc3,(17)\xi_{\mathrm{tot}}(P_c)\simeq\xi_0+c_1P_c+c_2P_c^2+c_3P_c^3, \tag*{(17)}

where the dominant contributions to c1c_1 are ξleak\xi_{\mathrm{leak}} and ξRam\xi_{\mathrm{Ram}}, while c2c_2 and c3c_3 predominantly arise from ξXPM\xi_{\mathrm{XPM}} and ξFWM\xi_{\mathrm{FWM}}, respectively. Excess noise is introduced through two mechanisms. First, it can be regarded as arising from additional photons detected by the receiver. Since

ξ∝⟨N⟩∝Phν\xi\propto\langle N\rangle\propto\frac{P}{h\nu}

within a given detection window, the problem reduces to determining the dependence of the corresponding noise power on PcP_c. Second, classical light induces changes in the fiber refractive index and thereby degrades the coherence of the quantum light, which is mainly manifested through cross-phase modulation (XPM).

The system excess-noise term ξ0\xi_0 can be measured directly. For the residual optical-leakage noise ξleak\xi_{\mathrm{leak}},

Pleak=∑i10−Ii/10Pi,ξleak=κrxPleak,P_{\mathrm{leak}}=\sum_i10^{-I_i/10}P_i,\qquad\xi_{\mathrm{leak}}=\kappa_{\mathrm{rx}}P_{\mathrm{leak}},

where IiI_i is the isolation between the iith classical channel and the quantum channel, and κrx\kappa_{\mathrm{rx}} is the conversion factor from optical power to shot-noise units (SNU). Therefore,

ξleak∝Pleak∝Pc.\xi_{\mathrm{leak}}\propto P_{\mathrm{leak}}\propto P_c.

For ξRam\xi_{\mathrm{Ram}}, spontaneous Raman scattering is the dominant noise source, whereas stimulated Raman scattering is not dominant. Summarizing the spontaneous Raman scattering noise models in the literature [23, 43, 44, 45], we obtain

ξRam=KRam∑iβσi(fq−fi;fq)[Pi,fwdLe−αL+Pi,bwd1−e−2αL2α],(18)\xi_{\mathrm{Ram}} = K_{\mathrm{Ram}} \sum_i \beta_{\sigma_i}(f_q-f_i;f_q)\left[P_{i,\mathrm{fwd}}Le^{-\alpha L}+P_{i,\mathrm{bwd}}\frac{1-e^{-2\alpha L}}{2\alpha}\right], \tag*{(18)}
σi={AS,fq>fi,S,fq<fi.\sigma_i = \begin{cases} \mathrm{AS}, & f_q>f_i,\\ \mathrm{S}, & f_q<f_i. \end{cases}

Here, KRamK_{\mathrm{Ram}} absorbs the effective linear coefficient and unit conversion and depends on wavelength and transmittance. βσi=βσi(fq−fi;fq)\beta_{\sigma_i}=\beta_{\sigma_i}(f_q-f_i;f_q) is the spontaneous Raman scattering coefficient and depends on the channel position. σi=S\sigma_i=\mathrm{S} or AS\mathrm{AS} indicates whether the quantum channel lies on the Stokes or anti-Stokes side of the corresponding classical channel. In general, βS\beta_{\mathrm{S}} is larger, indicating that Stokes scattering is more probable than anti-Stokes scattering. LL is the transmission distance. The first term in the brackets represents forward Raman-scattering noise, which first increases and then decreases with transmission distance, whereas the second term represents backward Raman-scattering noise, which increases monotonically and gradually saturates with distance. This relation can therefore be used to select the wavelength-channel positions and propagation directions.

For Kerr-nonlinearity-induced noise, the dominant effects are cross-phase modulation (XPM), self-phase modulation (SPM), and four-wave mixing (FWM). Because the quantum signal is extremely weak, SPM can be neglected. XPM induces a phase shift in the quantum channel [43],

ϕXPM=4πn2LeffλcAeff(Pcs−Pcr),(19)\phi_{\mathrm{XPM}}=\frac{4\pi n_2L_{\mathrm{eff}}}{\lambda_cA_{\mathrm{eff}}}\left(P_c^{\mathrm{s}}-P_c^{\mathrm{r}}\right), \tag*{(19)}

where n2n_2 is the nonlinear refractive index, LeffL_{\mathrm{eff}} is the effective length, λc\lambda_c is the classical-channel wavelength, and AeffA_{\mathrm{eff}} is the effective mode area. PcsP_c^{\mathrm{s}} and PcrP_c^{\mathrm{r}} denote the classical optical powers temporally overlapping with the quantum-signal and transmitted reference pulses, respectively. When the random phase perturbation δϕXPM\delta\phi_{\mathrm{XPM}} can be treated as a small quantity,

ξXPM∝VAVar⁡(ϕXPM).\xi_{\mathrm{XPM}}\propto V_A\operatorname{Var}(\phi_{\mathrm{XPM}}).

Considering Pcs−Pcr∝PcP_c^{\mathrm{s}}-P_c^{\mathrm{r}}\propto P_c, we have ϕXPM∝Pc\phi_{\mathrm{XPM}}\propto P_c, and therefore

ξXPM∝Var⁡(ϕXPM)∝Var⁡(Pc)∝Pc2.\xi_{\mathrm{XPM}}\propto\operatorname{Var}(\phi_{\mathrm{XPM}})\propto\operatorname{Var}(P_c)\propto P_c^2.

FWM increases the noise by introducing additional photons [42]. Under the slowly varying envelope approximation, the FWM-generated field amplitude is

Aijk→qFWM∝γAiAjAk∗LeffΦ(Δβ).A_{ijk\to q}^{\mathrm{FWM}}\propto\gamma A_iA_jA_k^*L_{\mathrm{eff}}\Phi(\Delta\beta).

Its squared magnitude gives the corresponding power,

Pijk→qFWM∝γ2PiPjPk∣Φ(Δβ)∣2,P_{ijk\to q}^{\mathrm{FWM}}\propto\gamma^2P_iP_jP_k\left|\Phi(\Delta\beta)\right|^2,

and the corresponding additive noise is ξFWM=∑i,j,kCijkPiPjPk\xi_{\mathrm{FWM}}=\sum_{i,j,k}C_{ijk}P_iP_jP_k. If the powers in different spectral bands vary synchronously with the classical optical power, then

ξFWM≃c3Pc3∝Pc3.\xi_{\mathrm{FWM}}\simeq c_3P_c^3\propto P_c^3.

Timestamp packet format and loading

For the secure absolute time-synchronization experiment, each timestamp packet consists of a fixed 12-bit preamble followed by a 64-bit protected timestamp payload,

bts=[bpre(12)∥bpay(64)],bpre(12)=010101010101.(20)b_{\mathrm{ts}}=\left[b_{\mathrm{pre}}^{(12)}\mathbin{\|}b_{\mathrm{pay}}^{(64)}\right],\qquad b_{\mathrm{pre}}^{(12)}=010101010101. \tag*{(20)}

The alternating preamble provides a distinctive pattern for identifying the packet boundary and aligning the timestamp frame. The 64-bit payload carries the protected timestamp information used to associate the highprecision relative timing with its corresponding absolute-time label.

References

References

  1. [1]G. Wang, D. Song, Z. Pang, F. Wang, H. Dai, W. Li, and B. Wang, Laser interferometry for high-speed railway health inspection using telecom fiber along the line, Nature Communications 16, 4129 (2025).DOI
  2. [2]B. K. Malia, Y. Wu, J. Martínez-Rincón, and M. A. Kasevich, Distributed quantum sensing with mode-entangled spin-squeezed atomic states, Nature 612, 661 (2022).DOI
  3. [3]X. Guo et al., Distributed quantum sensing in a continuous-variable entangled network, Nat. Phys. 16, 281 (2020).DOI
  4. [4]S. Wehner, D. Elkouss, and R. Hanson, Quantum internet: a vision for the road ahead, Science 362, eaam9288 (2018).
  5. [5]G. Marra, C. Clivati, R. Luckett, A. Tampellini, J. Kronjäger, L. Wright, A. Mura, F. Levi, S. Robinson, A. Xuereb, B. Baptie, and D. Calonico, Ultrastable laser interferometry for earthquake detection with terrestrial and submarine cables, Science 361, 486 (2018).DOI
  6. [6]P. Kómár, E. M. Kessler, M. Bishof, L. Jiang, A. S. Sørensen, J. Ye, and M. D. Lukin, A quantum network of clocks, Nat. Phys. 10, 582 (2014).DOI
  7. [7]H. J. Kimble, The quantum internet, Nature 453, 1023 (2008).DOI
  8. [8]J. I. Cirac, A. K. Ekert, S. F. Huelga, and C. Macchiavello, Distributed quantum computation over noisy channels, Phys. Rev. A 59, 4249 (1999).DOI
  9. [9]Y. Huang et al., Quantum fusion of independent networks based on multi-user entanglement swapping, Nat. Photonics 20, 87 (2026).arxiv.org/abs/2312.03279
  10. [10]D. Main et al., Distributed quantum computing across an optical network link, Nature 638, 383 (2025).DOI
  11. [11]M. Pompili et al., Long-distance coherent quantum communications in deployed telecom networks, Nature 640, 911 (2025).
  12. [12]C. M. Knaut et al., Entanglement of nanophotonic quantum memory nodes in a telecom network, Nature 629, 573 (2024).DOI
  13. [13]S. L. N. Hermans et al., Qubit teleportation between non-neighbouring nodes in a quantum network, Nature 605, 663 (2022).DOI
  14. [14]L.-Z. Liu et al., Distributed quantum phase estimation with entangled photons, Nat. Photonics 15, 137 (2021).DOI
  15. [15]M. Pompili et al., Realization of a multinode quantum network of remote solid-state qubits, Science 372, 259 (2021).DOI
  16. [16]Y.-A. Chen et al., An integrated space-to-ground quantum communication network over 4,600 kilometres, Nature 589, 214 (2021).DOI
  17. [17]S. Daiss et al., A quantum-logic gate between distant quantum-network modules, Science 371, 614 (2021).DOI
  18. [18]C. Delle Donne et al., An operating system for executing applications on quantum network nodes, Nature 639, 321 (2025).DOI
  19. [19]I. A. Burenkov et al., Synchronization and coexistence in quantum networks, Opt. Express 31, 11431 (2023).DOI
  20. [20]W. McKenzie et al., Clock synchronization characterization of the Washington DC metropolitan quantum network (DC-QNet), Appl. Phys. Lett. 125, 164004 (2024).DOI
  21. [21]M. Alshowkan, P. G. Evans, B. P. Williams, N. S. V. Rao, C. E. Marvinnney, Y.-Y. Pai, B. J. Lawrie, N. A. Peters, and J. M. Lukens, Advanced architectures for high-performance quantum networking, J. Opt. Commun. Netw. 14, 493 (2022).
  22. [22]D. Cuomo, M. Caleffi, and A. S. Cacciapuoti, Towards a distributed quantum computing ecosystem, IET Quantum Commun. 1, 3 (2020).DOI
  23. [23]M. Iqbal et al., SDN-enabled CV-QKD and classical channels coexistence: key insights for dense wavelength division multiplexing, J. Opt. Commun. Netw. 17, B28 (2025).
  24. [24]Y. Zheng et al., Large-scale quantum communication networks with integrated photonics, Nature 651, 68 (2026).DOI
  25. [25]Z. Chen, Y. Zhang, B. Luo, and H. Guo, Time–frequency transfer over optical fiber, National Science Review 12, nwaf236 (2025).DOI
  26. [26]L. Chang, S. Liu, and J. E. Bowers, Integrated optical frequency comb technologies, Nat. Photonics 16, 95 (2022).DOI
  27. [27]S. A. Diddams, K. Vahala, and T. Udem, Optical frequency combs: coherently uniting the electromagnetic spectrum, Science 369, eaay3676 (2020).DOI
  28. [28]Z. Chen et al., Dual-comb-enhanced microwave clock synchronization over commercial fiber, Optica 11, 1268 (2024).DOI
  29. [29]D. Yu et al., Time interval measurement with linear optical sampling at the femtosecond level, Photon. Res. 11, 2222 (2023).DOI
  30. [30]Q. Shen et al., Free-space dissemination of time and frequency with 10−19 instability over 113 km, Nature 610, 661 (2022).DOI
  31. [31]E. D. Caldwell, L. C. Sinclair, N. R. Newbury, and J.-D. Deschênes, The time-programmable frequency comb and its use in quantum-limited ranging, Nature 610, 667 (2022).DOI
  32. [32]S. Wang, X. Xiang, N. Treps, C. Fabre, T. Liu, S. Zhang, and R. Dong, Sub-shot-noise interferometric timing measurement with a squeezed frequency comb, Phys. Rev. A 98, 053821 (2018).DOI
  33. [33]J.-D. Deschênes, L. C. Sinclair, F. R. Giorgetta, W. C. Swann, E. Baumann, H. Bergeron, M. A. Cermak, and N. R. Newbury, Synchronization of distant optical clocks at the femtosecond level, Phys. Rev. X 6, 021016 (2016).DOI
  34. [34]K. Jung, J. Lim, J. Shin, H. Yang, L.-J. Chen, F. X. Kärtner, H.-S. Kang, C.-K. Min, and J. Kim, Remote laser–microwave synchronization over kilometer-scale fiber link with few-femtosecond drift, J. Lightwave Technol. 32, 3742 (2014).
  35. [35]F. R. Giorgetta, W. C. Swann, L. C. Sinclair, E. Baumann, I. Coddington, and N. R. Newbury, Optical two-way time and frequency transfer over free space, Nat. Photonics 7, 434 (2013).DOI
  36. [36]Y. Yang et al., Unifying frequency metrology across microwave, optical, and free-electron domains, Nat. Commun. 16, 8369 (2025).DOI
  37. [37]D. Yu, Z. Chen, Y. Zhang, Z. Jin, S. Yu, B. Luo, and H. Guo, Microwave frequency transfer over 3000-km fiber based on optical frequency combs and active noise cancellation, Phys. Rev. Res. 6, 023005 (2024).DOI
  38. [38]M. Schioppo et al., Comparing ultrastable lasers at 7 × 10−17 fractional frequency instability through a 2220 km optical fibre network, Nat. Commun. 13, 212 (2022).
  39. [39]K. Jung, J. Shin, J. Kang, S. Hunziker, C.-K. Min, and J. Kim, Frequency comb-based microwave transfer over fiber with 7 × 10−19 instability using fiber-loop optical-microwave phase detectors, Opt. Lett. 39, 1577 (2014).DOI
  40. [40]A. A. Jørgensen et al., Petabit-per-second data transmission using a chip-scale microcomb ring resonator source, Nat. Photonics 16, 798 (2022).
  41. [41]P. Marin-Palomo et al., Microresonator-based solitons for massively parallel coherent optical communications, Nature 546, 274 (2017).
  42. [42]S. Du, Y. Tian, and Y. Li, Impact of four-wave-mixing noise from dense wavelength-division-multiplexing systems on entangled-state continuous-variable quantum key distribution, Phys. Rev. Applied 14, 024013 (2020).DOI
  43. [43]R. Kumar, H. Qin, and R. Alléaume, Coexistence of continuous variable QKD with intense DWDM classical channels, New J. Phys. 17, 043027 (2015).DOI
  44. [44]T. Ferreira da Silva, G. B. Xavier, G. P. Temporão, and J. P. von der Weid, Impact of Raman scattered noise from multiple telecom channels on fiber-optic quantum key distribution systems, J. Lightwave Technol. 32, 2332 (2014).
  45. [45]T. A. Eriksson et al., Wavelength division multiplexing of continuous variable quantum key distribution and 18.3 Tbit/s data channels, Commun. Phys. 2, 9 (2019).DOI
  46. [46]B. Qi, W. Zhu, L. Qian, and H.-K. Lo, Feasibility of quantum key distribution through a dense wavelength division multiplexing network, New J. Phys. 12, 103042 (2010).DOI
  47. [47]F. Kiselev, N. Veselkova, R. Goncharov, and V. Egorov, A theoretical study of subcarrier-wave quantum key distribution system integration with an optical transport network utilizing dense wavelength division multiplexing, Journal of Physics B: Atomic, Molecular and Optical Physics 54, 135502 (2021).DOI
  48. [48]B. P. Dix-Matthews, S. W. Schediwy, D. R. Gozzard, E. Savalle, F.-X. Esnault, T. Lévéque, C. Gravestock, D. D’Mello, S. Karpathakis, M. Tobar, and P. Wolf, Point-to-point stabilized optical frequency transfer with active optics, Nature Communications 12, 515 (2021).
  49. [49]S. Pirandola, U. L. Andersen, L. Banchi, M. Berta, D. Bunandar, R. Colbeck, D. Englund, T. Gehring, C. Lupo, C. Ottaviani, J. L. Pereira, M. Razavi, J. Shamsul Shaari, M. Tomamichel, V. C. Usenko, G. Vallone, P. Villoresi, and P. Wallden, Advances in quantum cryptography, Adv. Opt. Photon. 12, 1012 (2020).
  50. [50]F. Laudenbach, C. Pacher, C.-H. F. Fung, A. Poppe, M. Peev, B. Schrenk, M. Hentschel, P. Walther, and H. Hübel, Continuous-variable quantum key distribution with Gaussian modulation—the theory of practical implementations, Adv. Quantum Technol. 1, 1800001 (2018).DOI
  51. [51]W. J. Riley and D. A. Howe, Handbook of Frequency Stability Analysis, Tech. Rep. NIST Special Publication 1065 (National Institute of Standards and Technology, 2008).DOI

Paper details

vlnc:W_sdrty832
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents