Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generation of 12 dB squeezed light from a waveguide optical parametric amplifier using a machine-learning-controlled spatial light modulator

Gyeongmin Ha, Kazuki Hirota, Takahiro Kashiwazaki, Takumi Suzuki, Akito Kawasaki, Warit Asavanant, Mamoru Endo, Akira Furusawa

Abstract

We demonstrate the generation of $12.1 \pm 0.2$ dB squeezed light from a periodically poled lithium niobate (PPLN) waveguide optical parametric amplifier (OPA). While single-pass OPAs offer squeezed light with THz-order bandwidths, loss from spatial mode mismatch between the squeezed light and the local oscillator (LO) previously capped the squeezing level at $\sim$10 dB [K. Hirota et al., Opt. Express 34, 7958 (2026)]. In this work, we minimize this loss by introducing a machine-learning-optimized spatial light modulator (SLM) in the path of the LO. Specifically, we employed a double-reflection configuration to increase the spatial degrees of freedom, and directly used the measured squeezing level as the optimization's objective function.

Generation of 12 dB squeezed light from a waveguide optical parametric amplifier using a machine-learning-controlled spatial light modulator

Abstract

We demonstrate the generation of dB squeezed light from a periodically poled lithium niobate (PPLN) waveguide optical parametric amplifier (OPA). While single-pass OPAs offer squeezed light with THz-order bandwidths, loss from spatial mode mismatch between the squeezed light and the local oscillator (LO) previously capped the squeezing level at 10 dB [K. Hirota et al., Opt. Express 34, 7958 (2026)]. In this work, we minimize this loss by introducing a machine-learning-optimized spatial light modulator (SLM) in the path of the LO. Specifically, we employed a double-reflection configuration to increase the spatial degrees of freedom, and directly used the measured squeezing level as the optimization's objective function.
Paper Structure (6 sections, 12 equations, 5 figures)

This paper contains 6 sections, 12 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Experimental Setup
  3. SLM Phase Mask Generation
  4. Results
  5. Discussions
  6. Conclusion

Figures (5)

  • Figure 1: Schematic of the experimental setup used to optimize phase mask of spatial light modulator (SLM) and the squeezing level measurement. Green and red lines represent fundamental and second harmonic beams (1545.32 nm and 772.66 nm) in free space. Dark green is the fundamental beam inside an optical fiber. Gray dotted lines represent the electrical signals used for phase control feedback, and blue solid lines represent control signals for phase mask optimization. The squeezed vacuum states of light (SQZ) were generated in a periodically poled lithium niobate waveguide optical parametric amplifier (squeezer OPA PPLN-WG). A tapping mirror and a separate waveguide (phase-detection OPA PPLN-WG) was used to generate phase-locking signal Hirota10dB:26. A SLM was placed in the path of the local oscillator (LO) to minimize loss from spatial mode mismatch between the squeezed light and the LO. Reflecting the LO twice off the SLM provided additional degrees of freedom for spatial mode control. EDFA: erbium doped fiber amplifier; SHG: second harmonic generator; EOM: electro-optic modulator; AOM: acousto-optic modulator; VOA: variable optical attenuator; DM: dichroic mirror; PD: photodiode; LPF: low pass filter; PID: proportional-integral-derivative controller; ESA: electrical spectrum analyzer.
  • Figure 2: Left: Result of the last optimization run. At every iteration, the squeezing level was automatically measured and used as the objective function for the BO algorithm to optimize. The parameters tested in the first 40 iterations were randomly sampled. The best pattern found was at the 375th iteration (marked in red), initially yielding 12.0 dB. After careful re-alignment, the same pattern generated $12.1 \pm 0.2$ dB. Right: Visualization of this optimal phase mask, where the grayscale gradient represents the 10-bit phase modulation values (black $= 0$, white $= 1023$). The semi-transparent green circles indicate the Gaussian beam size (radii of 1.2 mm) for the two incident beams.
  • Figure 3: Various noise levels measured by balanced homodyne detection using an electrical spectrum analyzer operated in zero-span mode at a sideband frequency of 3 MHz, with a resolution bandwidth of 1 MHz and a video bandwidth of 100 Hz. The pump power was 585 mW, measured immediately after the squeezer OPA. The gray line represents the shot-noise level. The blue and red lines show the squeezed and anti-squeezed noise, respectively, normalized to the shot-noise level. Green line represents the normalized noise level of the squeezed light with the phase of the LO scanned.
  • Figure 4: Left: Various noise levels at sideband frequencies up to 100 MHz. The pump power was 585 mW, measured immediately after the squeezer OPA. The resolution bandwidth was 1 MHz and the video bandwidth was 100 Hz. Right: Noise levels normalized to shot-noise. Light blue is the circuit-noise-corrected squeezed noise.
  • Figure 5: Squeezing and anti-squeezing levels for various pump powers, corrected for circuit noise. The solid curve is a fitting with parameters phase fluctuation of 9 mrad and loss of 4.4%. Dotted line represents the theoretical curve with phase fluctuation and loss from our previous work Hirota10dB:26. Each squeezed and anti-squeezed data point had measurement error of $\pm0.2$ dB.