The protocol in question is phase-matching quantum key distribution, or PM-QKD, a variant of the twin-field family that has attracted considerable attention since its introduction. Twin-field schemes are special because they beat the fundamental rate-loss bound that constrains ordinary repeaterless quantum communication, the linear scaling of secret key rate with channel transmittance first formalized by Pirandola and colleagues in 2017. Instead of one party sending photons directly to the other, both Alice and Bob send weak optical pulses to an untrusted intermediate node where single-photon interference takes place. The phase relationships between the two arms carry the key information, and the security analysis converts interference outcomes into estimates of bit and phase error rates. PM-QKD simplifies this picture considerably: both parties prepare states in only a single basis, which serves simultaneously for key generation and for parameter estimation, streamlining the protocol and easing its experimental demands.

Real experiments, however, need detectors, and the detector of choice for cost-conscious, field-deployable systems is the single-photon avalanche diode, or SPAD. These compact semiconductor devices are cheap, robust and small, which makes them far more practical than the superconducting nanowire detectors that dominate laboratory record-setting runs. But SPADs carry a hidden flaw: afterpulsing. When an avalanche is triggered by a genuine photon, charge carriers can become trapped in defects within the semiconductor’s depletion region. If a trapped carrier is released after the detector has been re-armed, it can itself initiate a new avalanche, a spurious click that has nothing to do with any arriving photon. These ghost avalanches are correlated with the detector’s own history, a non-Markovian property that has been documented in earlier studies of InGaAs/InP devices, and they contaminate the very statistics that quantum key distribution relies upon for its security proofs.

The trouble is that modern security analyses of PM-QKD are typically built on idealized detector models. When experimenters run numerical simulations to predict the secret key rate and the maximum tolerable transmission loss of a system, they feed in observables such as the quantum bit error rate and the phase error rate. If afterpulses inflate or distort those observables relative to what the idealized model assumes, the simulation diverges from the actual experiment. A system designed on paper to be secure and efficient may, in the laboratory, deliver fewer secret bits than expected, or worse, its security parameters may be misjudged because the error rates used in the proof no longer describe the true quantum channel. The new work addresses this gap directly by constructing an afterpulse-compatible model for PM-QKD, one that folds the statistics of trapped-charge-induced avalanches into the decoy-state estimation framework that underpins practical protocol security.

The decoy-state method, introduced in its practical form by Ma, Qi, Zhao and Lo in 2005, is the standard defense against photon-number-splitting attacks. Alice randomly chooses among several intensity settings, so-called decoy states, when preparing her weak coherent pulses, and the observed yields and error rates of each setting allow the parties to tightly bound the behavior of the single-photon contributions that actually carry the key. In the afterpulse-compatible model developed by the Nanjing-based team, the afterpulse probability enters these yield and error estimates, so that the bounds on the bit error rate and the phase error rate reflect the detector’s real, memory-laden behavior. The phase error rate, which cannot be measured directly and must be estimated, is particularly sensitive to any discrepancy between modeled and observed statistics, which is why afterpulses matter so much for twin-field type protocols where phase information is central.

What do the simulations reveal? The headline finding is sobering but unsurprising in direction: afterpulses degrade both the secret key rate and the maximum transmission loss that PM-QKD can tolerate. Every spurious avalanche adds noise to the interference statistics, raising the apparent error rates and shrinking the fraction of sifted data that can be certified as secret. At long distances, where the genuine photon signal arriving at the intermediate node is already vanishingly faint, the afterpulse background becomes proportionally more damaging, eating into the very regime where twin-field protocols earn their advantage over the rate-loss bound. For system designers, the message is clear: ignoring afterpulses in the planning stage of a PM-QKD deployment will produce optimistic predictions that the hardware cannot honor.

Fortunately, the study does not stop at diagnosis. To counteract the degradation, the authors apply advantage distillation, a classical post-processing technique that sits between parameter estimation and the final key extraction. The procedure works as follows. Alice and Bob each divide their raw bit strings into blocks of b bits. Alice selects a random bit c and sends Bob the message consisting of each of her block bits XORed with c. Bob XORs the received block with his own corresponding block. If the result is either all ones or all zeros, meaning the two blocks are perfectly correlated or perfectly anti-correlated, the pair keeps the first bit of the block as raw key material; otherwise they discard the block entirely. Because an eavesdropper knows the transmitted message, only one bit per successful block can be retained, but the surviving bits are drawn from a dramatically cleaner sub-ensemble with a much lower effective error rate.

The mathematics of the advantage-distillation-enhanced key rate, laid out in the paper’s appendix, shows how the error rate after distillation becomes a weighted combination of the probabilities that a b-bit block is all errors or all correct, and how the bounds on the eavesdropper’s channel parameters are transformed through the same block-wise filtering. The optimization is performed over the block size b and over the decomposition of the channel into error components constrained by the decoy-state estimates. When b equals one, every raw key bit passes the filter and the formula reduces to the standard PM-QKD key rate, which confirms the internal consistency of the treatment. For larger blocks, the trade-off is explicit: fewer bits survive, but each survivor is more trustworthy, and the net effect in the afterpulse-degraded regime is a higher secret key rate and a larger tolerable transmission loss than the protocol achieves without distillation.

This is not the first time advantage distillation has been deployed to rescue practical quantum key distribution. Previous work by overlapping groups has shown its benefits for decoy-state protocols, for reference-frame-independent schemes, for measurement-device-independent systems, and for phase-matching QKD itself in the absence of afterpulse modeling. The novelty here lies in the combination: an afterpulse-aware security model for PM-QKD, coupled with a distillation layer that specifically claws back the performance lost to detector imperfections. The approach complements hardware-level remedies that other researchers have pursued, such as ultra-narrowband interference circuits that suppress noise in InGaAs/InP avalanche photodiodes and gating schemes with widely tunable repetition rates, by treating the residual afterpulse contribution at the level of information processing rather than detector engineering.

The broader significance is about closing the gap between record-breaking demonstrations and deployable infrastructure. Twin-field quantum key distribution has now been demonstrated over 830 kilometers of fiber and, with finite-key analysis, beyond 1000 kilometers, but those landmark experiments leaned on superconducting detectors and elaborate phase stabilization. A practical national-scale quantum network will need nodes that are inexpensive, compact and rugged, which is exactly the case for SPAD-based receivers. By quantifying what afterpulses cost PM-QKD and showing that advantage distillation recovers a meaningful share of the loss, the study gives engineers a realistic design envelope and a concrete mitigation strategy. As quantum networks edge from laboratory benches toward metropolitan backbones, such careful accounting of every spurious avalanche is what turns a beautiful physical principle into a working, trustworthy communication system.

Subject of Research: Modeling afterpulse effects in single-photon avalanche detectors and mitigating them with advantage distillation in practical phase-matching quantum key distribution

Article Title: Afterpulse analysis for practical phase-matching quantum key distribution

Article References: Shu, C.-H., Zhao, Y.-H., Wen, Z., Sun, M.-R., Zhu, J.-R., Zhang, C.-M., & Li, H.-W. (2026). Afterpulse analysis for practical phase-matching quantum key distribution. Quantum Information Processing, 25(10), Article 336. https://doi.org/10.1007/s11128-026-05364-1

Image Credits: AI Generated

DOI: 10.1007/s11128-026-05364-1

Keywords: quantum key distribution, phase-matching QKD, twin-field QKD, afterpulse, single-photon avalanche detector, decoy-state method, advantage distillation, quantum cryptography, secret key rate, photon detectors, quantum communication, rate-loss bound