Peking University's Lin, Chen and Guo Carry Timing and 100 Mbit/s Data on Two Clock Combs, With Two Sampling Combs and CV-QKD Over 100 km Fiber, arXiv 2611.16349, 12 November 2026: TRL 4
Quentir Defense Monitor
Evidence-based insights for quantum defense and security. Published by Quentir Systems LLC · November 13, 2026.

Thirteen researchers from Peking University, the Beijing Institute of Technology, the Beijing University of Posts and Telecommunications and the Chinese Academy of Sciences have run precision timing, a classical data link and continuous-variable quantum key distribution over one 100 kilometer fiber at the same time, with a pair of transmitted clock combs, two additional local-oscillator sampling combs and an independently prepared quantum carrier doing the work that three separate systems usually do. Yiting Lin and twelve colleagues, with Ziyang Chen and Hong Guo of Peking University as corresponding authors, posted the result to arXiv on 12 November 2026, and the headline numbers are 100 megabits per second of data, a timing stability of 0.85 femtoseconds at one second of averaging and a quantum channel that gave up 6.7 percent of its secret key rate for sharing the glass.
A defense reader should care because a distributed sensor array is only as good as its common clock. Radars that net their returns, receivers that locate an emitter from the time difference of arrival, and any future pair of quantum processors or sensors that must act together all need the same second to a precision that satellite timing cannot provide when it is jammed. This post reads the Beijing paper as an optical frequency comb infrastructure in the networking pillar, network time synchronization sub-branch, dual-use in posture, and sets out how far a laboratory spool stands from a cable a program office could specify.
What the paper of 12 November 2026 establishes: two combs, two-way transfer and data modulated onto the clock pulses
An optical frequency comb is a laser that emits a train of very short pulses at a rate fixed by a reference clock, so the pulses themselves are the ticks. The Beijing setup places a clock comb at each end of the fiber, Comb A at site A and Comb B at site B, both locked to a repetition rate of 100 megahertz through a feedback loop that steers each comb to a clock-referenced synthesizer. A second comb at each site, local-oscillator combs X₁ and X₂, runs at 100.001 megahertz, one kilohertz off, and sampling the arriving pulses against this local oscillator comb turns the femtosecond arrival times into a signal electronics can read. The apparatus therefore uses four combs in total. The two clock-comb signals travel in opposite directions through the same fiber, and because the path delay is the same both ways it drops out when the two measurements are combined. The paper states this as the reciprocal propagation delay and its fluctuations being rejected as common-mode terms.
The first new step is to make those same pulses carry data. An arbitrary waveform generator drives an intensity modulator on the output of Comb A, writing non-return-to-zero bits onto the pulse train at 100 megabits per second while leaving the pulse timing untouched. At site B an erbium-doped fiber amplifier, a photodetector and an analog-to-digital converter recover the bits. The receiver reached a bit error rate of one in a million at a received power of minus 46 dBm without forward error correction. The sentence the authors put in their introduction is that they simultaneously transmit 100 megabits per second of classical data and achieve a timing stability of 0.85 femtoseconds at one second over 100 kilometers. That figure is a time deviation, read from the curves in their Figure 2, and the text states that the fused system remains consistent with the standalone timing system, so the data did not cost the clock anything the authors could measure.
The quantum channel: four wavelength plans compared, 0.0037 shot-noise units of added noise and a 6.7 percent key-rate penalty
The second step is to run quantum key distribution down the same fiber. The group used a Gaussian-modulated coherent-state protocol, the continuous-variable family that encodes keys in the amplitude and phase of weak coherent light, at a 100 megahertz symbol rate on the C34 channel of the telecom grid at 1550.12 nanometers, with heterodyne detection against an independent local oscillator digitized at one gigasample per second. The combs and the quantum signal share the fiber through dense wavelength-division multiplexing.
Classical light at milliwatt power beside a quantum channel at a fraction of a photon per symbol raises the excess noise the receiver sees, and excess noise is what sets a continuous-variable key rate. The authors compared four plans for where to put the two comb directions on the grid, C30 and C32, C31 and C33, C32 and C30, and C37 and C35, each swept across comb power, and found that excess noise grows at an accelerating rate with mean comb power in every plan. The plan with the lowest penalty sends the forward comb on C32 and the backward comb on C30 at 0.35 milliwatts. In that configuration the quantum channel saw 0.0037 shot-noise units of added excess noise and delivered 29.67 kilobits per second of secret key over the 100 kilometer link, 6.7 percent below the same system running alone.
The keys then close the loop on the clock. Once per second a pulse triggers a 76-bit packet, a 12-bit preamble and a 64-bit payload, which is encrypted with a one-time pad by exclusive-or against quantum-generated key and sent on the comb carrier. Each timestamp occupies 760 nanoseconds, the average overhead is 76 bits per second and the temporal occupancy is 7.6 parts in ten million. With the encrypted timestamps running, a common frequency reference and Kalman-filtered residual time offsets, the measurement yielded residual timing stability, expressed as a time deviation, of 0.26 femtoseconds at one-second averaging, which the paper says matches the standalone timing system. This measures stability, not absolute accuracy or synchronization performance between independent clocks.
Quantum pillar: networking (network time synchronization). Use posture: dual-use. Technology readiness: TRL 4 of 9. For the reader this means a complete timing, data and key system was assembled and measured in a laboratory over a 100 kilometer spool of fiber, with no deployed cable, no outdoor temperature swing and no second site more than a bench away.
Why the panel reads networking (network time synchronization), dual-use and TRL 4
The pillar is networking because the quantum content of the work is the distribution of keys and of a shared time between two sites over a link, and the sub-branch is the one Quentir's capability map reserves for keeping separated nodes on one clock. The quantum key channel is real, and it is the part the authors use to make their timestamps confidential, so the post could have read quantum key distribution links instead; the choice falls on time synchronization because the paper's own framing, and its best numbers, are about timing, and the keys serve the clock rather than the other way round. The posture is dual-use. A telecom operator or a quantum network builder gains a single fiber that carries its clock, its control plane and its keys, which is a cost argument. A force gains the same cable for a distributed sensor net, and the affiliations on the paper, including the Beijing Institute of Technology's State Key Laboratory of Environment Characteristics and Effects for Near-space and the Chinese Academy of Sciences' Technology and Engineering Center for Space Utilization, say where the authors' institutions expect to use such links. The readiness level is 4. The shared ladder places bench assembly and laboratory testing at level 4, and this is a full system of combs, modulator, quantum transmitter and receiver put together and characterized in one laboratory over spooled fiber. It is above level 3 because the pieces were integrated and measured as a whole, and below level 5 because nothing was tested against a realistic deployed link.
What it would let a force do: one cable for the clock, the orders and the keys of a distributed sensor array
The need for femtosecond-class time over a link is on the public record in the United States. The Defense Advanced Research Projects Agency's Robust Optical Clock Network program, run by Mukund Vengalattore in the Defense Sciences Office, states its goal as a significant increase in the precision and holdover of Department of War timing capability in devices of low size, weight and power. Its first technical area asks for small portable clocks with the precision needed for distributed coherent sensing on airborne or spaceborne platforms, and its second for transportable clocks that hold GPS-quality time for a month without a satellite signal. The program page adds that optical timing signals are harder to jam or spoof than the microwave signals a force uses today. The Beijing paper addresses the link between such clocks, which is the other half of a network: a clock that holds time is of limited use to a sensor array unless the sites can compare that time to a precision finer than the sensing task demands.
How much finer is possible over free space was shown by the National Institute of Standards and Technology in 2013, when Giorgetta, Swann, Sinclair, Baumann, Coddington and Newbury transferred time between combs across a two kilometer open-air path in Boulder at a timing deviation of one femtosecond, with residual instability below one part in a billion billion at a thousand seconds, despite frequent fading from turbulence and obstructions. The Beijing figure of 0.85 femtoseconds over 100 kilometers of fiber sits in the same class, with the fiber trading the turbulence problem for a distance 50 times as long. For a reader following fiber links between national laboratories, the NPL and PTB comparison of two ytterbium-ion optical clocks over a 2220 kilometer fiber link is the deployed-cable counterpart: that work shows what a real route does to a frequency comparison, and the Beijing work shows how much else one route could carry.
For a force, the capability on offer is a distributed sensor array whose clock, control data and key material travel in one fiber, with the clock protected by two-way cancellation of the path and the timestamps hidden under a one-time pad. The operational reading is that a netted radar pair, a passive geolocation baseline or a pair of entangled-photon sources on separate sites could be synchronized through infrastructure a telecom operator already owns, without a dedicated timing network and without a satellite reference, provided the fiber behaves like the spool.
What stands between a 100 km spool and a program office: real cable, authentication, comb size and a network protocol
Six items separate the paper from a requirement. The first is the fiber itself. The authors state that sites A and B are connected by a 100 kilometer spooled fiber, and the text as read this run says nothing about temperature control or drift; a buried or aerial route sees daily temperature cycles and mechanical disturbance that a spool in a laboratory does not, and two-way cancellation only removes what both directions share at the same instant. The second is authentication. The timestamps are encrypted, and the paper describes no mechanism for a receiver to confirm who sent them. The United Kingdom's NCSC, in its publication "Quantum networking technologies", states that quantum key distribution does not provide authentication, that it will not support the use of QKD for government or military applications and that it regards post-quantum cryptography as the primary mitigation to the threat quantum computers pose to cryptography. A timing link a force relies on must resist a false timestamp as much as a read one, and that layer is undocumented in the demonstration. The third is the comb. The authors list more compact and robust comb generation, through integrated comb sources and integrated photonics, as the path forward, which is an admission that the bench combs used here are not field hardware. The fourth is scale. The demonstration is one point-to-point link, and the discussion calls for network-level synchronization protocols, channel management and long-distance quantum communication technology before the approach becomes a network. The fifth is nonlinear noise. Raman scattering, four-wave mixing and cross-phase modulation from the classical light are what drove the excess noise to accelerate with comb power, and the authors say these must be suppressed further for metropolitan and intercity reach. The sixth is independent measurement. Data and code are available on reasonable request, independent replication has not been established by the reviewed sources, and the only stability number stated in the text is the one-second value, so the long-averaging behavior a clock network depends on is read from a figure rather than from a stated result.
What the sources support is this. A Peking University-led team has assembled and measured, in one laboratory over 100 kilometers of spooled fiber, a system in which a pair of transmitted clock combs and two additional local-oscillator sampling combs deliver two-way time transfer at a time deviation of 0.85 femtoseconds at one second while carrying 100 megabits per second of data on the clock pulses, and in which an independently prepared, co-propagating continuous-variable quantum key channel, placed on the grid to minimize noise, keeps 29.67 kilobits per second of secret key at a 6.7 percent penalty and encrypts the timestamps. With a common frequency reference and Kalman-filtered residual time offsets, the encrypted-timestamp measurement yields residual timing stability of 0.26 femtoseconds at one-second averaging, not absolute accuracy or synchronization performance between independent clocks. DARPA has stated the military need for precise timing in distributed coherent sensing, NIST showed the femtosecond free-space counterpart in 2013 and the NCSC has set out why a key channel without authentication does not clear a government or military bar. A test on a deployed route with its temperature history, an authenticated timestamp and a stated multi-hour stability figure are the next three pieces of evidence to wait for.
Sources
Primary source: Yiting Lin, Zishuo Ren, Yunkun Zhao, Jiayu Ma, Zimeng Li, Yang Sun, Yufei Zhang, Yanhao Sun, Lei Yang, Xiangyu Wang, Kemi Xu, Ziyang Chen and Hong Guo (Peking University; Beijing Institute of Technology; Beijing University of Posts and Telecommunications; Technology and Engineering Center for Space Utilization, Chinese Academy of Sciences), "An optical-frequency-comb-enabled unified infrastructure for quantum networks," arXiv 2611.16349, submitted 12 November 2026. Other material: the DARPA program page for the Robust Optical Clock Network, read on 13 November 2026; the NIST publication page for Giorgetta, Swann, Sinclair, Baumann, Coddington and Newbury, "Optical two-way time and frequency transfer over free space," Nature Photonics 7(6), 2013; the NCSC white paper "Quantum security technologies," version of 5 August 2025. The capability reading, the posture argument, the readiness placement and the buyer items are Quentir analysis.