arXiv:2610.21909
Looking Beyond Visibility: Resolving the Temporal Structure of Continuous-Wave Four-Photon Hong-Ou-Mandel Interference
Sabine Häussler, Meritxell Cabrejo-Ponce, Ana Maria Torres, Baghdasar Baghdasaryan, Karen Lozano-Mendéz, Fabian Steinlechner
Abstract
Hong-Ou-Mandel interference underpins many photonic quantum information protocols and is commonly characterized by a single visibility parameter. This quantity is typically obtained by averaging over a fixed coincidence window, thereby discarding details of the underlying temporal structure that are critical in continuous-wave processes. Four-photon interference with continuous-wave sources furthermore requires photon pairs with long coherence times, which we generate in microring resonators and measure with ultrafast time-resolved detection. Here, we model and resolve their interference and temporal structure by recording all four photon arrival times and mapping their relative delays onto a three-dimensional coincidence histogram relative to a trigger detection. The resulting structure preserves relative arrival-time correlations that are usually discarded and enables post-processed optimization of coincidence windows to trade off visibility against count rate. This method provides a general benchmark for continuous-wave sources in interference-based quantum network protocols.
AI Summary
Continuous-wave (CW) Hong–Ou–Mandel measurements can yield different visibilities depending on temporal postselection. This paper examines how to retain the four-photon timing structure, define the distinguishable reference rate, and quantify the visibility–count-rate trade-off for asynchronously emitted photon pairs. The experiment interferes heralded photons from two silicon-nitride microring sources, recording all four arrival times and constructing a three-dimensional histogram of relative delays. The sources have reported coherence times of 433 and 445 ps; the two-fold timing-jitter budget is approximately 34 ps. Transmitted–transmitted and reflected–reflected coincidence pathways form intersecting branches, with interference suppressing coincidences near their intersection. A model combines their interfering amplitudes with leading accidental-coincidence contributions, using measured singles rates and two-fold histograms as inputs. Two reconstruction procedures are compared. Scanning only one heralding-channel delay requires broad output-channel windows to recover a baseline. Jointly scanning a herald and a corresponding output channel follows each correlation branch, allowing narrow windows; the two distant branch rates are summed to define $R_\infty$. With equal 600 ps windows, the reported reference four-fold rate is 387 counts/hour at 84.6% visibility; 300 ps windows give 83 counts/hour at 93.2%, or 96.8% after accidental subtraction. The model reproduces reference-rate scaling but underestimates coincidences at the interference minimum. It omits higher-order emission, imperfect source purity, polarization mismatch, spectral drift, and detector jitter, and is treated as an ideal-performance upper bound. Accidental-corrected visibility is intended for characterization, not protocol-performance estimates.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.