An ultra-bright, highly-scalable, squeezed light source for hybrid quantum photonics

AI summary

Overview

Research area: Experimental quantum optics and photonics — specifically sources of squeezed light (single-mode squeezed vacuum, SMSV) for hybrid continuous-variable/discrete-variable (CV/DV) quantum photonics, built on periodically poled potassium titanyl phosphate (KTP) waveguides.

Technical level: Intermediate for the experimental results; Advanced for the theoretical framework (multimode spatial Langevin equations, correlation-matrix decomposition, time-ordering and loss during state generation).

Scope: The paper demonstrates and theoretically models a single-pass, periodically poled Type-II KTP waveguide that generates bright, near-single-mode picosecond squeezed vacuum at 1546 nm, and uses a new multimode theory including time-ordering and loss to characterize it via click detection.

What This Paper Is About

Photon-counting-based hybrid quantum photonics (e.g., Gaussian boson sampling) needs squeezed-light sources whose output occupies one single, well-defined spatial and spectro-temporal mode, because photon counting cannot resolve mode structure and therefore integrates over every mode the detector sees. Existing platforms typically satisfy some but not all of the requirements — scalability, single-mode emission, strong and pure squeezing, and compatibility with fiber networks and superconducting nanowire single-photon detectors (SNSPDs). The authors build and validate a KTP waveguide source designed so that the polarization-nondegenerate Type-II signal and idler are near-perfectly indistinguishable, allowing them to be interfered into true single-mode squeezed states, and they supply the theory needed to characterize such a source when driven to high gain.

Key Contributions

1. A source design meeting all four requirements simultaneously. A 20 mm-long periodically poled Type-II KTP waveguide (AdvR Inc.) with a positive phase-matching angle of approximately 57 degrees, engineered for near-perfect spatial and spectral indistinguishability between signal and idler around 1546 nm, producing picosecond-duration pulses compatible with SNSPD photon-number resolution and fiber networks.

2. A complete multimode theoretical framework. Spatial master equations for the correlation matrices D_a, D_b and C_ab that include both propagation loss and nonlinear coupling, valid in the high-gain regime where the usual biphoton/Schmidt and Bloch-Messiah approaches fail, plus a model of realistic g^(2) measurement including detector saturation and imperfect signal–idler separation at the PBS.

3. Comprehensive experimental characterization across gain regimes. Brightness, spectral broadening, joint spectral intensity (JSI), second-order correlation, and Hong-Ou-Mandel interference measured with click detectors and an optical spectrum analyzer, showing strong agreement with theory up to mean photon numbers of 40,000.

4. Identification of failure modes of standard metrology. The paper shows that Schmidt decomposition of the JSI systematically underestimates the true mode number, and that the simplified single-mode loss model breaks down at high gain — establishing when the rigorous approach is mandatory.

Main Findings

• Brightness and efficiency: The source generates mean photon numbers greater than 40,000 at the highest available pump pulse energy of 480 pJ. In the linear (low-gain) regime the mean photon number per pulse of the signal is 20 photons/nJ, corresponding to 20 pairs/nJ.
• Single-mode performance: The measured effective mode number is K = 1.24 ± 0.3. The idealized model predicts a low-gain g^(2) of 1.81, corresponding to an effective mode number of 1.23.
• Indistinguishability: A polarization-based Hong-Ou-Mandel measurement at approximately 0.1 mean photons per mode yields a visibility of (92.6 ± 0.8)%, confirming signal–idler indistinguishability.
• High-gain dynamics are real and measurable: Both the measured JSI and the measured spectral full-width-at-half-maximum broaden with increasing mean photon number; phase-matching side lobes diminish and a single dominant mode emerges in the JSI, in agreement with theory.
• JSI-based mode counting is unreliable: The mode number inferred from Schmidt decomposition of the JSI is significantly lower than the true mode number μ_a predicted by the correlation-matrix decomposition, because JSI analysis assumes a flat phase profile.
• Simplified theory fails at high gain: The conventional sinh²(r) fit deviates from the data at high pump pulse energies, while the rigorous model with time-ordering and loss remains in good agreement.
• Projected squeezing: Simulations indicate the platform can produce squeezing levels approaching −20 dB, and show that the detailed theoretical treatment is required once squeezing exceeds −8 dB.
• Source losses: Propagation loss is 0.25 ± 0.03 dB/cm, plus approximately 8% Fresnel reflection at the uncoated exit facet; total system efficiency for SNSPD measurements was 45% ± 3% (signal) and 44% ± 2% (idler) by the Klyshko method.

Methodology in Plain English

A frequency-comb laser at 80 MHz is pulse-picked down to 1 MHz with an electro-optic modulator to avoid detector latching, then frequency-doubled to pump pulses centered at 772.5 nm (maximum bandwidth roughly 1 nm, maximum average power roughly 900 mW). A programmable spatial light modulator in a folded 4f pulse shaper allows fine control of the pump bandwidth and phase profile. The pump is injected into the 20 mm periodically poled KTP waveguide, where it produces paired photons (signal and idler) that are orthogonally polarized but deliberately made spectrally indistinguishable.

After the waveguide, a coated silicon filter gives about 100 dB of pump suppression and a 12 nm bandpass filter removes background photons (it is far broader than the few-nanometer process bandwidth, so it does not shape the state). A delay compensation crystal corrects the group delay from the waveguide. A half-wave plate plus polarizing beam splitter switch between two configurations: at 0° the PBS separates signal and idler deterministically for JSI and g^(2) measurements in a Hanbury Brown–Twiss setup; at 45° the PBS acts as a 50:50 beamsplitter, producing two independent squeezed states at its outputs.

Detection uses SNSPDs (Single Quantum) with about 90% efficiency and 20 ps jitter, read out by a 2 ps resolution time tagger (Time Tagger X, Swabian Instruments), with calibrated variable attenuators preventing saturation. Spectra are measured directly with an optical spectrum analyzer (Anritsu MS9740B) at high photon numbers, where nanowatt-scale average powers are available. JSI is measured by dispersive-fiber (time-of-flight) spectroscopy using fibers with about −430 ps/nm dispersion at 1550 nm, wavelength-calibrated with a tunable CW laser (EXFO T200S), with transmissions of roughly 45% and 50% for signal and idler.

On the theory side, the authors solve the spatial master equations for the correlation matrices of the signal and idler fields in a discrete frequency basis, including frequency-independent loss matrices and the z-dependent coupling matrix built from the pump spectrum and phase mismatch. Mode structure is extracted by diagonalizing the correlation matrices (Mercer–Wolf modes) rather than by a biphoton Schmidt decomposition, which is what allows the treatment to remain valid at high gain and with loss.

Why This Matters

The paper sets a benchmark for what a squeezed-light source for large-scale hybrid quantum photonic networks must deliver, and shows one platform that delivers all of it at once: high efficiency, high brightness, near-single-mode operation, telecom wavelength, and picosecond pulses compatible with SNSPD photon-number resolution. Equally important, it exposes how commonly used characterization methods — high-gain sinh²(r) fits, JSI-based Schmidt mode counting, and naive click-detection g^(2) — can give misleading answers, which matters for anyone specifying or benchmarking components for a larger machine.

Applications:

• Gaussian boson sampling: multiple SMSV states interfered in a passive linear-optical network followed by photon-number-resolving detection; reduced mode purity is known to reduce the computational complexity of the sampling task.
• Continuous-variable quantum computing and error correction: squeezing bounds set by quantum error correction schemes motivate the projected approach to −20 dB.
• Entanglement generation and hybrid CV/DV protocols: photon subtraction and photon counting as non-Gaussian resources layered on Gaussian squeezed states.
• Fiber-based quantum networks: operation at 1546 nm with picosecond pulses limits fiber dispersion while remaining compatible with existing detection infrastructure.

Industry relevance: The work speaks directly to suppliers and integrators of SNSPDs, time taggers, periodically poled nonlinear waveguides, and pulsed telecom laser systems, and to groups building photonic quantum processors that must specify squeezing, loss, and mode-purity budgets for many parallel sources.

Future Directions

• Reducing loss during state generation, which the paper identifies as a key factor limiting current performance alongside high-gain time-ordering effects.
• Closing the gap to the projected −20 dB squeezing experimentally, now that theory predicts the platform's ceiling and indicates the rigorous model is needed above −8 dB.
• Replacing JSI-based mode estimation with measurement techniques that do not assume a flat phase profile, since the paper demonstrates the Schmidt decomposition of the JSI underestimates the true mode number.
• Scaling to multiplexed architectures, which requires the mode structure to be consistent across multiple sources and robust to experimental parameter variations.

Target Audience

Experimental and theoretical quantum photonics researchers working on squeezed-light sources, Gaussian boson sampling, and CV quantum information; waveguide and nonlinear-optics device engineers using ppKTP; and metrology-minded readers interested in how high-gain and loss-induced multimode dynamics invalidate standard characterization assumptions. The experimental sections are accessible to graduate students with a background in quantum optics, while the theory section is aimed at specialists.

Authors’ abstract

Hybrid quantum photonics seeks to combine the complementary advantages of continuous- and discrete-variable quantum optics. This typically entails photon-counting measurements on entangled states generated by interfering many single-mode squeezed-vacuum (SMSV) states. However, because conventional photon-counting schemes are mode-insensitive, it is critical that the SMSV states occupy a single, well-defined mode. Achieving this requires careful engineering of the process, which determines both the spatial and spectro-temporal properties of the generated state. In addition, the ideal source must be massively scalable, capable of efficiently generating strong squeezing, and remain compatible with existing detection schemes and fiber networks. Although many platforms address one or more of these requirements, satisfying all of them simultaneously remains challenging. Here, we present a source that meets all of these requirements: a single-pass, periodically poled, Type-II potassium titanyl phosphate (KTP) waveguide optimized for scalable hybrid quantum-photonic architectures. The SMSV state produced by the source has a measured effective mode number of 1.24. Furthermore, the source is extremely bright (producing up to 40 000 photons per pulse) and operates at a central wavelength of 1546nm, optimized for fiber-network compatibility and which, in combination with picosecond duration, also enables intrinsic photon-number resolution in superconducting nanowire single-photon detectors. Although this source constitutes an ideal source in a simplified picture, the ultimate limitations of any source will be governed by complex dynamics that arise when the system is driven at high-gain or due to unavoidable loss during state generation. We have therefore developed a complete theoretical framework that enables a comprehensive photon-counting-based characterization of the source.