18-dB on-chip vacuum squeezing in an adaptively poled lithium niobate waveguide

Technical Summary: 18-dB on-chip vacuum squeezing in an adaptively poled lithium niobate waveguide

Problem Statement
Quantum squeezed states of light offer the potential to surpass classical measurement limits and enable advanced quantum information processing. However, generating these states in a scalable, low-loss format remains a significant challenge. While bulk optical parametric oscillators (OPOs) have achieved squeezing levels up to 15 dB, they lack the size, weight, and power (SWaP) advantages of photonic integrated circuits (PICs). Thin-film lithium niobate (TFLN) is a leading platform for integrated squeezing due to its tight modal confinement and large second-order nonlinear coefficients. However, previous TFLN implementations have been limited by high propagation losses, difficulties in maintaining quasi-phase-matching (QPM) over long distances due to film thickness variations, and the stringent temperature stabilization required for resonant approaches. Furthermore, accurately inferring true on-chip squeezing performance is complicated by the difficulty of distinguishing between propagation losses, facet coupling losses, and phase noise, particularly in high-gain regimes where vacuum noise mixing becomes significant.

Methodology
The authors demonstrate continuous-wave (CW) quantum squeezing using a 1.6-cm-long, adaptively poled, periodically poled thin-film lithium niobate (PP-TFLN) traveling-wave waveguide. The device is fabricated on a 600-nm-thick MgO-doped X-cut TFLN wafer. To address the challenge of nanometer-scale thickness variations in the TFLN film, which disrupt phase matching, the team mapped the wafer thickness prior to poling and tailored the local poling period to maintain a constant QPM wavelength across the entire device length. This adaptive poling strategy enables a high second-harmonic generation (SHG) efficiency of 2740 % W⁻¹.

To rigorously quantify the on-chip performance, the authors developed a distributed loss-and-squeezing model. Unlike previous approaches that approximate losses as lumped elements, this model accounts for the continuous generation of squeezing and the simultaneous mixing of vacuum noise due to propagation loss along the waveguide. The methodology involves:

1. Bidirectional SHG Measurements: Measuring SHG efficiency in both forward and backward propagation directions to characterize facet coupling losses at both telecom (fundamental) and visible (second-harmonic) wavelengths.
2. Non-linear Solver: Using a Levenberg-Marquardt algorithm to solve a system of non-linear equations that relate measured transmission, bidirectional SHG efficiencies, and theoretical on-chip squeezing models. This allows for the independent determination of facet losses, phase noise, and nonlinear interaction strength without assuming symmetric losses.
3. Monte Carlo Simulations: Embedding Monte Carlo simulations within the solver to propagate experimental uncertainties and establish confidence intervals for the inferred parameters.

Key Results

• Measured Performance: At a pump power of 205 mW, the device produced a measured squeezing of -0.66 ± 0.35 dB and anti-squeezing of +13.22 ± 0.44 dB. The measured squeezing was limited primarily by detection losses (including facet coupling) and phase noise (estimated at ~37 mrad).
• Inferred On-Chip Performance: Using the distributed model, the authors inferred a record on-chip squeezing level of -18.06 ± 0.43 dB and anti-squeezing of +20.64 ± 0.35 dB.
• Statistical Validation: The 95% confidence interval for the on-chip squeezing is calculated to be [-18.96, -17.25] dB.
• Device Characteristics: The waveguide exhibits a low propagation loss of 0.1 dB/cm (intrinsic quality factors of ~3.2 million at 1570 nm). The adaptive poling design resulted in a single dominant SHG peak with minimal deviation from the ideal sinc-squared response.

Significance and Claims
The paper claims this work represents the highest squeezing reported for any integrated photonic platform to date. It establishes thin-film lithium niobate as a high-performance, scalable platform for continuous-variable quantum sensing, communications, and computing.

Key contributions highlighted by the authors include:

1. First Assumption-Free Statistical Validation: This is the first work to employ a distributed and bidirectional model to infer on-chip squeezing with such high accuracy, avoiding the errors associated with lumped-loss approximations.
2. Scalable CW Squeezing: The device achieves high squeezing levels using continuous-wave pumping at an order of magnitude lower peak power than pulsed schemes, while avoiding the complexity of cavity locking and Pound–Drever–Hall stabilization required by resonant systems.
3. Engineering Insights: The study demonstrates that at high squeezing levels, even sub-percent optical losses and phase noise on the order of tens of milliradians play a non-negligible role, underscoring the critical importance of distributed loss and noise engineering in traveling-wave systems.

The authors note that while the current device is limited by facet losses and poling depth, future improvements in waveguide length, propagation loss (approaching 1.3 dB/m), and SHG efficiency could theoretically push on-chip squeezing toward -32 dB.