Hyperentanglement-enabled Quantum Privacy Amplification for Classical-Quantum Coexistence in Fiber Networks

Abstract

Integrating quantum communication into existing fiber infrastructures requires coexisting with high-power classical traffic, which generates noise in the quantum channel. Spontaneous Raman Scattering (SpRS) is the dominant noise source, generating a broadband photon background that degrades quantum states and, in Quantum Key Distribution (QKD), can entirely suppress secure key generation. In this work, we implement single-copy entanglement distillation utilizing energy-time and frequency hyperentanglement, and we exploit it to perform Quantum Privacy Amplification (QPA) directly at a physical layer. Strict frequency anti-correlations of spontaneous parametric down-conversion are used to selectively discard uncorrelated Raman photons while preserving genuine entangled pairs, suppressing accidental coincidences by a factor $1/N$, where $N$ is the number of frequency channel pairs. Using passive borosilicate-glass photonic integrated circuits and standard DWDM demultiplexers over a 20~km single-mode fiber testbed, we demonstrate a 4~dB improvement in tolerance to classical laser launch power $P_C$ when QPA is performed, extending the secure key threshold from $P_C=-13.3$ dBm to $P_C=-9.3$ dBm. The result shows that off-the-shelf telecom components can be repurposed as physical-layer quantum privacy amplifiers.

The paper studies whether frequency correlations can suppress Raman-induced accidental coincidences in entanglement-based QKD sharing fiber with classical light. It implements frequency-selective postselection, framed as single-copy entanglement distillation and physical-layer quantum privacy amplification, using energy–time/frequency hyperentangled SPDC photon pairs. One photon traverses 20 km of standard single-mode fiber alongside a co-propagating telecom laser. Passive integrated Franson interferometers analyze energy–time correlations, while 100-GHz DWDM filters select two conjugate frequency channels. Accepting only frequency-anticorrelated coincidences retains genuine pairs while rejecting unmatched noise. For $N$ equally populated frequency channels, the model predicts an accidental-coincidence reduction of $1/N$ relative to combining all channels; the experiment uses $N=2$. Measured interference visibility starts near $98\%$ without classical light. Frequency-selected coincidences remain above the stated $78\%$ visibility threshold up to a classical launch power of $-9.3$ dBm, compared with $-13.3$ dBm for the combined 200-GHz reference: a 4-dB increase in power tolerance. Key rates are inferred from measured fringe visibilities using a finite-key calculation with $10^6$ sifted bits, error-correction efficiency $1.15$, and secrecy and correctness parameters of $10^{-12}$. A Raman-noise model reproduces the reported trends. The study considers spontaneous Raman noise and explicitly neglects phase-related effects such as cross-phase modulation. The classical channel is represented by a laser rather than demonstrated data traffic. Source brightness was chosen to keep multipair accidentals small, not to maximize key rate; scaling beyond two frequency channels is proposed rather than demonstrated.