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

This paper demonstrates the generation of 12.1 ± 0.2 dB squeezed light from a broadband waveguide optical parametric amplifier by employing a machine-learning-optimized spatial light modulator in the local oscillator path to minimize spatial mode mismatch loss.

The Gist

🌟 The Big Idea: Making Light "Quieter"

Imagine you are trying to hear a whisper in a very noisy room. To hear it better, you need to turn down the background noise. In the world of quantum physics, scientists use special "quiet" light called squeezed light.

Normally, light has a bit of "fuzziness" or noise (like static on an old TV). Squeezed light is a special trick where scientists reduce the noise in one part of the light wave, making it incredibly precise. This is a superpower for future quantum computers and ultra-sensitive sensors (like the ones that detect gravitational waves).

The goal of this paper was to make this light even "quieter" (more squeezed) than ever before.

🚧 The Problem: The "Spilled Water" Effect

The team was using a device called a waveguide (think of it as a tiny, super-efficient highway for light) to create this squeezed light. They had managed to get a pretty good result before (about 10 "decibels" of squeezing).

But they hit a wall. Why?

Imagine you have a fancy, uniquely shaped water bottle (the squeezed light) and you are trying to pour it into a standard round cup (the measuring device, called the Local Oscillator). Even if you try your best, the shapes don't match perfectly. Some water spills out. In physics, this "spilled water" is called loss.

In previous experiments, the light coming out of the waveguide didn't match the shape of the measuring beam perfectly. About 4% of the signal was lost just because the shapes didn't line up. This kept them stuck at the 10 dB limit.

🛠️ The Solution: The "Smart Funhouse Mirror"

To fix the shape mismatch, the team used a device called a Spatial Light Modulator (SLM).

• Analogy: Think of the SLM as a digital funhouse mirror. It can change its surface shape instantly to bend light exactly how you want it.

However, figuring out the perfect shape for the mirror is incredibly hard. There are billions of possible shapes. If you tried to guess them one by one, it would take forever.

Enter Machine Learning.
Instead of a human trying to tweak the mirror, they let a computer algorithm do the work.

• Analogy: Imagine a chef trying to perfect a soup recipe. Instead of guessing, the chef tastes the soup, adds a pinch of salt, tastes it again, adds a pinch of pepper, and keeps going until it tastes perfect.
• The Difference: In the past, scientists used a "proxy" (like checking the signal strength) to guess if the soup was good. In this experiment, the computer tasted the actual result (the squeezing level) directly. It knew exactly when it hit the jackpot.

🔄 The Secret Sauce: The Double Bounce

They also added a clever trick to the setup. They made the light beam bounce off the "funhouse mirror" twice before measuring it.

• Analogy: If you try to fix a wobbly table leg by turning one screw, it might still be shaky. But if you have two screws to adjust, you have much more control.
• Result: This "double-reflection" gave the computer twice as many knobs to turn, allowing it to fix much more complex shape problems.

🏆 The Result: Breaking the Record

By combining the waveguide, the smart mirror, and the machine learning chef, they achieved something new:

• Old Record: ~10 dB of squeezing.
• New Record: 12.1 dB of squeezing.

They reduced the "spilled water" (loss) from 4% down to almost nothing (0.4% mismatch loss). The total system loss was only 4.4%.

🚀 Why Should You Care?

You might be thinking, "Who cares about 2 extra decibels of light noise?"

Here is why it matters:

1. Faster Quantum Computers: Quantum computers need this "quiet" light to do calculations. The more squeezing you have, the fewer errors the computer makes. This moves us closer to building a quantum computer that can solve problems normal computers can't touch.
2. Better Sensors: This technology helps build sensors that are sensitive enough to detect ripples in space-time (gravitational waves) or tiny changes in the human body.
3. Speed: Unlike older methods that were slow, this light is incredibly fast (THz bandwidth). It's like upgrading from a dial-up internet connection to 5G.

📝 Summary

The scientists built a "smart mirror" controlled by a computer that learned how to shape light perfectly. By fixing the mismatch between the light being made and the light being measured, they reduced waste and created the "quietest" light ever recorded in this type of setup. It’s a major step toward building the super-computers of the future.

Technical Summary

Here is a detailed technical summary of the paper "Generation of 12 dB squeezed light from a waveguide optical parametric amplifier using a machine-learning-controlled spatial light modulator."

1. Problem Statement

Squeezed light is a critical resource for quantum metrology and teleportation-based quantum computing, where high squeezing levels and broad bandwidths are required to improve computational accuracy and clock frequency. While Optical Parametric Oscillators (OPOs) can generate high squeezing levels, their bandwidth is typically limited to the MHz–GHz range by the resonator geometry. In contrast, single-pass waveguide Optical Parametric Amplifiers (OPAs) offer THz-order bandwidths but have historically been limited to lower squeezing levels (previously capped at ~10 dB).

The primary bottleneck for waveguide OPAs is spatial mode mismatch loss between the generated squeezed light and the Local Oscillator (LO) during homodyne detection. Previous attempts to minimize this loss used the visibility between a "probe beam" and the LO as a proxy for mode matching. However, this indirect measure proved misleading; high visibility did not guarantee low loss due to complex spatial aberrations. Consequently, mode mismatch accounted for a significant portion of total system loss, preventing the realization of squeezing levels above 10 dB in waveguide systems.

2. Methodology

To overcome the limitations of previous proxy-based optimization, the authors implemented a direct optimization approach using a machine-learning-controlled Spatial Light Modulator (SLM).

• Experimental Setup: The system utilizes a Periodically Poled Lithium Niobate (PPLN) waveguide OPA to generate squeezed vacuum states. The seed laser operates at 1545.32 nm (fundamental) with a second harmonic pump at 772.66 nm.
• SLM Configuration: An SLM is placed in the path of the LO. Crucially, the LO undergoes a double-reflection configuration off the SLM surface. This doubles the spatial degrees of freedom compared to single-reflection setups, allowing for more complex wavefront correction.
• Optimization Algorithm: The authors employed Bayesian Optimization (BO) to find the optimal phase mask for the SLM.
  • Objective Function: Unlike prior works that optimized probe beam visibility, this algorithm directly uses the measured squeezing level (dB) as the objective function. This ensures the optimization targets the final figure of merit.
  • Parameterization: The phase mask is parameterized as a combination of Zernike polynomials (for wavefront correction) and cylindrical lenses (for astigmatism and mode matching). The SLM is split into two halves (Left/Right) to accommodate the double-reflection geometry, totaling 40 optimized parameters per run.
  • Measurement Protocol: To ensure robustness against power fluctuations, a differential measurement protocol was used. The squeezing level is calculated as the difference between the shot-noise power and the squeezed noise power measured by an Electrical Spectrum Analyzer (ESA).
• Process: The BO algorithm iteratively adjusts the SLM phase mask (200–400 iterations). After finding the optimal mask, the LO is manually realigned to correct for thermal drifts, and the process repeats with the new mask as a starting point.

3. Key Contributions

• Direct Squeezing Optimization: The paper introduces a novel method where the squeezing level itself serves as the objective function for machine learning, bypassing the inaccuracies of indirect visibility proxies.
• Double-Reflection SLM Architecture: By reflecting the LO twice off the SLM, the system significantly increases the spatial degrees of freedom, enabling the correction of complex aberrations that single-reflection setups cannot address.
• Mode-Selection Capability: The authors demonstrate that the SLM does not merely match a pure Gaussian mode but effectively performs spatial mode selection. It molds the LO phase front to overlap with the specific spatial mode (or superposition of modes) emitted by the OPA that exhibits the highest degree of squeezing, accommodating fabrication imperfections in the waveguide.
• Record-Breaking Squeezing: The work achieves the highest squeezing level reported to date for a waveguide OPA system.

4. Results

• Squeezing Level: The system successfully generated 12.1 ± 0.2 dB of squeezed light. This surpasses the previous record of 10.1 ± 0.2 dB for waveguide OPAs.
• Loss Analysis: The total system loss was reduced to 4.4%.
  • Waveguide internal/propagation loss: ~2%
  • Optical element propagation loss: ~1%
  • Photodiode quantum efficiency loss: ~1%
  • Spatial mode mismatch loss: Reduced to approximately 0.4% (down from ~4% in previous work).
• Stability: Measurements followed a Gaussian distribution with a standard deviation of 0.07 dB, indicating high stability.
• Bandwidth: While the homodyne detector limited the measurement bandwidth to ~35 MHz, the source itself maintains a THz-order bandwidth (consistent with previous PPLN waveguide characterizations).
• Noise Performance: The circuit noise clearance was greater than 28 dB at a 3 MHz sideband frequency.

5. Significance

This achievement represents a major milestone for optical quantum computing, specifically for teleportation-based architectures that rely on cluster states generated from squeezed light.

• Speed and Accuracy: By combining THz-order bandwidth (from the waveguide OPA) with high squeezing levels (12 dB), the system enables high-speed, low-noise information processing. This addresses the fundamental clock frequency limitations of optical quantum computers.
• Scalability: The method demonstrates that high-quality squeezing can be achieved without complex cavity resonators, simplifying experimental implementations.
• Future Outlook: The authors suggest that further reductions in waveguide internal loss and improved phase-locking stability could push squeezing levels even higher. Additionally, the double-reflection SLM configuration could be adapted for signal beams to enhance detection efficiency in broadband real-time quadrature measurements, facilitating the generation and tomography of ultrafast non-Gaussian states.