Pump Tuning Selects Photon-Pair Bins in Lithium Niobate

Frequency-bin quantum photonics needs photon-pair sources that are bright, spectrally addressable, and compatible with chip-scale routing. Cavity-enhanced SPDC can raise pair generation, but triply resonant designs make the pump, signal, and idler resonances hard to align at once. This paper takes a different route: it keeps the pump in a single-pass configuration while resonantly enhancing the down-converted photons in a periodically poled lithium niobate photonic-crystal Fabry–Perot microresonator.

Core Contribution

The central result is a quantum frequency comb whose correlated bin pairing is selected by the pump wavelength rather than fixed by a single cavity-pump resonance. The device generates cavity-defined signal-idler pairs across 1495–1570 nm, with 461 measured bins, while preserving pairwise correlations on the fixed comb grid. That matters because frequency-bin systems need both multiplexing and control: a bright comb is less useful if the usable output channels are set only by fabrication tolerances.

The other useful contribution is the paper’s separation of intrinsic generation from extracted photon-pair flux. The authors show that the bin with the highest internal cavity enhancement is not necessarily the bin with the highest usable output brightness. Extraction is treated as part of the source performance, not as a secondary packaging loss.

Technical Approach

Figure 1 shows the integrated device: a periodically poled lithium niobate photonic-crystal Fabry–Perot microresonator designed for SPDC quantum frequency comb generation. The photonic-crystal cavity supplies narrow resonances for the signal and idler fields, while periodic poling supplies the χ² quasi-phase matching. The pump is not forced to be resonant with the cavity, which relaxes the usual triply resonant alignment problem.

The paper models the intrinsic pair-generation rate for a conjugate resonance pair [i, −i] as a cavity-enhanced SPDC process. In the authors’ notation, the intrinsic rate scales with the nonlinear coupling strength, the effective single-pass interaction length, the pump photon flux, and the inverse of the combined signal-idler decay rates. The loaded rate then multiplies the intrinsic rate by the extraction efficiencies of both photons:

with ηext,i tied to the loaded and coupling quality factors through QL/Qc. This is the paper’s practical design point: high intrinsic Q increases internal generation, but the output flux depends on how efficiently the resonator is loaded into the channel.

Results and Analysis

The headline numbers support the device claim. The authors report intrinsic and loaded brightnesses of 69.9 MHz/µW and 1.88 MHz/µW, respectively, a maximum coincidence-to-accidental ratio of 16,000, and loaded spectral brightness approaching 4.29 × 10^9 pairs/(s·mW·nm). Frequency-resolved measurements identify 461 cavity-defined bins across 1495–1570 nm, and the selected high-CAR bin pair reaches heralded gH²(0) as low as 4.0 × 10^-4.

Figure 2 is the main calibration bridge between nonlinear gain and photon-pair brightness. It supports the claim that the measured SPDC rates are not just raw counts, but are tied back to a calibrated nonlinear response of the same device platform.

The loading analysis is especially useful. For six representative conjugate bin pairs, the paper reports intrinsic brightness dropping from 124 kHz/µW for the [2, −2] pair to 39.8 kHz/µW for [120, −120], then to 5.38 kHz/µW for [179, −179] and 3.63 kHz/µW for [228, −228] outside the photonic bandgap. Loaded brightness follows a different trend: it rises from 2.31 kHz/µW for [2, −2] to 11.1 kHz/µW for [87, −87] even while intrinsic brightness declines. The interpretation is straightforward: narrowing resonances improves internal generation, but stronger output coupling can make a lower-intrinsic-gain bin more useful at the chip output.

Figure 3 shows the fixed cavity comb grid and frequency-bin correlations. The important point is not only the number of bins, but that pump tuning deterministically selects which signal-idler bins are paired within that grid.

The single-photon measurements add a stricter test. At 89.3 µW on-chip pump power, the undercoupled [2, −2] pair gives gH²(0) = 0.087 ± 0.016, the near-critical [120, −120] pair gives 0.024 ± 0.0036, and the overcoupled [228, −228] pair gives 0.0004 ± 0.00011, with corresponding intrinsic pair-generation rates of 8.62 MHz, 4.66 MHz, and 332 kHz. That is a clear trade-off rather than a simple monotonic win: the cleanest heralded statistics appear in a bin with much lower intrinsic generation.

Caveats in Practice

The evidence is experimental and quantitatively detailed, but it is still a single-device demonstration. The paper maps extraction-aware loading within one PhC-FP platform rather than showing programmable network operation with active switching, full quantum-state protocols, or system-level loss budgets. For source designers, the value is concrete: the device shows how pump selection and wavelength-dependent loading can be used together, but it also shows why reporting only intrinsic brightness or only CAR can hide the usable photon flux.