High-Visibility Franson Interference Enabled by Passive Photonic Integrated Interferometers at Telecom Wavelengths

Abstract

High-visibility Franson interference at telecom C-band wavelengths is achieved using a cascaded periodically poled lithium niobate (PPLN) waveguide photon-pair source combined with fully passive, path-imbalanced Mach-Zehnder interferometers implemented on photonic integrated circuits (PICs). The interferometers require neither on-chip phase shifters nor active stabilization; instead, the phase is scanned via thermal tuning of the chip. By employing a narrow-linewidth continuous-wave (CW) pump and dense wavelength-division multiplexing (DWDM) filtering, energy-time entangled photon pairs with high spectral indistinguishability are generated. We achieve a 4.8% heralding efficiency and a two-photon interference visibility of 97.1% from sinusoidal fringe fitting (raw visibility 95.2% and background-corrected visibility 95.6%), alongside a coincidence-to-accidental ratio (CAR) exceeding 1000 at only 1.7 mW of pump power. These results represent one of the highest Franson-interference visibilities reported for a PIC-based analyzer within a compact, fiber-integrated platform.

Figures (5)

• Figure 1: Experimental setup. A CW laser at 1560 nm pumps two cascaded periodically poled lithium niobate (PPLN) waveguides (second-harmonic generation, SHG, followed by spontaneous parametric down-conversion, SPDC). A DWDM separates the signal (CH22) and idler (CH20), each entering a matched unbalanced Mach--Zehnder interferometer (uMZI) with a path difference of approximately $\Delta t \approx 0.8$ ns after polarization cleaning. Detection is performed using superconducting nanowire single-photon detectors (SNSPDs) and a time-to-digital converter (TDC) for coincidence counting and visibility analysis.
• Figure 2: (a) Temperature tuning of the PPLN waveguide measured via SHG. Open circles: measured SHG power; solid curve: $\mathrm{sinc}^{2}$ fit. The extracted FWHM of $\approx 3.53^{\circ}\mathrm{C}$ gives the temperature-acceptance bandwidth of the QPM structure. The peak temperature sets the operating point for SPDC. (b) Measured coincidence histogram (red) together with a fit to the two-photon correlation model of Eq. \ref{['eq:franson_hist_model']} for the constructive interference setting $\phi = 0$ (blue dashed). The three Gaussian peaks correspond to the LS ($\tau \approx -\Delta t$), SS/LL ($\tau \approx 0$), and SL ($\tau \approx +\Delta t$) path contributions of the Franson interferometer, with only the central SS/LL peak exhibiting phase-dependent modulation.
• Figure 3: (a) Signal ($S'_s$) and idler ($S'_i$) singles rates together with the inferred pair-generation rate $R_{\mathrm{pair}}$ as a function of the squared pump power $P_{1560}^{2}$. The linear dependence confirms the expected quadratic scaling of the SPDC pair rate with the fundamental pump power in the cascaded SHG$\rightarrow$SPDC architecture. (b) Heralding efficiencies $\eta_s$ and $\eta_i$ as a function of pump power $P_{1560}$. The efficiencies peak near 2 mW and decrease with increasing pump power due to rising multi-pair emission probability and mild detector saturation. The close agreement between $\eta_s$ and $\eta_i$ confirms balanced optical throughput in the signal and idler arms.
• Figure 4: (a) Coincidence-to-accidental ratio (CAR) as a function of the fundamental pump power $P_{1560}$. The data are plotted on a logarithmic scale and follow the expected $\mathrm{CAR}\propto 1/P_\omega^2$ dependence (solid line) derived for the cascaded SHG–SPDC process, corresponding to a slope of approximately $-0.2$ decades per dB. The CAR decreases from $\sim 10^{3}$ at the lowest pump powers to $\sim 10^{2}$ at $P_{1560}=8~\mathrm{dBm}$, indicating the increasing impact of multi-pair events at higher brightness. (b) Coincidence histograms measured for constructive ($\phi=0$) and destructive ($\phi=\pi$) two-photon interference. The central peak (SS/LL) exhibits the expected phase-dependent modulation, increasing for $\phi=0$ and decreasing for $\phi=\pi$, while the side peaks (SL and LS) remain constant under phase variation, as they correspond to distinguishable path combinations. This invariance of the SL/LS peaks provides a direct baseline for identifying genuine energy--time interference in the central SS/LL contribution.
• Figure 5: (a) Central-peak coincidence rate versus phase; the sinusoidal fit $C(\phi)=B[1+V\cos(\alpha\phi+\phi_0)]$ yields $V=97.1\%$. Raw visibility from (a): $95.2\%$; background-corrected: $95.6\%$. (b) Raw two-photon interference visibility as a function of the fundamental pump power $P_{1560}$. As the pump power increases, the visibility decreases from about $95\%$ to $89\%$, consistent with the growing contribution of multi-pair emission and accidental coincidences at higher source brightness.