⚛️ quantum physics

Detecting entanglement of non-Gaussian continuous-variable states from single-copy homodyne measurements

This paper introduces a single-copy homodyne measurement protocol that utilizes unbiased U-statistic estimators to detect non-Gaussian continuous-variable entanglement via the $p_3$ -PPT criterion, demonstrating that current experimental setups can effectively identify such entanglement with a feasible number of measurements.

Original authors: Moritz Straeter, Michael Tsesmelis, Leong-Chuan Kwek

Published 2026-06-30

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Moritz Straeter, Michael Tsesmelis, Leong-Chuan Kwek

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to figure out if two strangers are secretly communicating with each other (a state physicists call "entanglement"). In the world of quantum physics, these "strangers" are particles of light, and their "conversation" is a deep, invisible connection that allows them to act as a single unit.

For a long time, scientists had a simple rulebook for checking if light particles were talking. If the light behaved in a smooth, predictable, bell-curve way (called "Gaussian"), they could just check the average speed and direction to know if they were connected. But many of the most useful, powerful quantum states are "non-Gaussian"—they are jagged, weird, and unpredictable. The old rulebook was blind to them; it couldn't see the connection even when it was there.

This paper introduces a new, clever detective tool that can spot these hidden connections using only standard, everyday equipment.

The Old Problem: The "Multi-Copy" Bottleneck

Previously, to catch these tricky non-Gaussian connections, scientists had to use a very difficult method. They had to:

Create three identical copies of the light state at the exact same time.
Run them through a complex machine (an interferometer) to mix them together.
Use special, expensive detectors that can count individual photons.

This was like trying to solve a mystery by forcing three identical suspects to stand in a room together and talk. It's hard to get three identical suspects to show up at the exact same second, and the equipment is rare and finicky.

The New Solution: The "Shadow" Detective

The authors of this paper developed a new protocol that works like a shadow puppet show.

Instead of needing three copies of the state at once, they only need one copy at a time. Here is how their "single-copy" method works:

The Randomized Flashlight: Imagine shining a flashlight at an object from a completely random angle every time you look at it. In the lab, they measure the light particles using "homodyne detection," which is essentially measuring the light's wave from a random angle (phase) chosen by a computer.
The Shadow Snapshots: Each time they measure, they get a "snapshot" or a "shadow" of the quantum state. Individually, a single snapshot looks like random noise. It doesn't tell you much.
The Pattern Recognition: The magic happens when they take thousands of these random snapshots and crunch the numbers together using a specific mathematical recipe (called a "U-statistic").
- Think of it like trying to figure out the shape of a hidden object in a dark room by throwing thousands of tennis balls at it from random angles. One ball tells you nothing. But if you map where thousands of balls bounce off, you can reconstruct the object's shape perfectly.
- In this case, the "shape" they are reconstructing isn't the object itself, but a specific mathematical property called the "partial transpose." If this property has a negative value, it proves the two particles are entangled.

Why This is a Big Deal

The paper claims three major advantages that make this accessible to almost any physics lab:

No Multi-Copy Hassle: You don't need to synchronize three copies of the state. You just measure one, record the data, and move to the next. This removes the "triple-coincidence" bottleneck that slowed down previous experiments.
Standard Equipment: You don't need those rare, expensive photon-counting detectors. You only need standard "homodyne" detectors, which are found in almost every quantum optics lab.
Robustness: The method is surprisingly tough. If your equipment isn't perfect (e.g., some light is lost in the fiber optic cable, or the detector isn't 100% efficient), the method still works. It won't give you a "false alarm" saying particles are connected when they aren't. It just might need a few more measurements to be sure.

The Results: Catching the Elusive

The authors tested their method on six different types of quantum states, including some that are notoriously difficult to detect (like "photon-subtracted" states).

The Speed: For states with a moderate amount of energy (about 2 photons on average), they could detect entanglement with 95% certainty using only 1,000 to 10,000 measurements.
The Feasibility: Since modern experiments can take these measurements thousands of times per second, this entire process could be done in a matter of seconds or minutes.

The Bottom Line

This paper proves that you don't need exotic, multi-copy machines to detect complex quantum connections. By taking many simple, random "snapshots" of a single quantum state and using clever math to find the hidden patterns, you can confirm entanglement with standard lab equipment. It turns a difficult, high-tech puzzle into a routine experiment that many more scientists can now perform.

Technical Summary: Detecting Entanglement of Non-Gaussian Continuous-Variable States from Single-Copy Homodyne Measurements

Problem Statement
Certifying the entanglement of non-Gaussian continuous-variable (CV) states remains a significant experimental challenge. While the entanglement of Gaussian states is fully determined by second moments (covariance matrices), standard second-moment criteria (e.g., Duan, Simon) are blind to the entanglement of many non-Gaussian states, such as photon-subtracted states, NOON states, and specific cat state regimes. Although the Shchukin–Vogel hierarchy can theoretically detect any negative partial transpose (NPT) state, the estimation cost scales rapidly with matrix dimension. Recent approaches, such as the $p_3$ -PPT criterion extended to CV systems by Deside et al., utilize low-order moments of the partially transposed density matrix but require multi-copy interferometry and photon-number-resolving (PNR) detectors. These requirements are experimentally demanding, particularly for heralded non-Gaussian states where triple-coincidence rates are severely suppressed.

Methodology
The authors propose a single-copy protocol that estimates the partial-transpose (PT) moments $p_2$ and $p_3$ directly from randomized homodyne measurements, bypassing the need for multi-copy interferometry or PNR detectors.

Data Acquisition: The protocol involves measuring both modes of a bipartite state $\rho$ via independent homodyne detection with uniformly random local oscillator phases $\theta_A, \theta_B$ .
Shadow Tomography: Each measurement outcome $(x, \theta)$ is mapped to a "shadow snapshot" $\rho^{(i)}_N$ in a truncated Fock space of dimension $N+1$ using homodyne pattern functions. These snapshots are unbiased estimators of the truncated state $\rho_N$ .
U-Statistic Estimators: To estimate the multi-copy quantities $p_2 = \text{Tr}[(\rho^{TB})^2]$ $p_{2} = Tr [(ρ^{T B})^{2}]$ and $p_3 = \text{Tr}[(\rho^{TB})^3]$ $p_{3} = Tr [(ρ^{T B})^{3}]$ from single-copy data, the authors construct unbiased U-statistic estimators.
- The second moment estimator $\hat{p}_2$ is the average of a symmetric kernel over all distinct pairs of snapshots.
- The third moment estimator $\hat{p}_3$ utilizes a cyclic-shift kernel over distinct triples of snapshots, where the reverse ordering on one subsystem implements the partial transpose.
- These estimators are computed efficiently in $O(T N^4)$ time using accumulated matrix sums, avoiding combinatorial enumeration.
Entanglement Witnesses: The estimated moments are used to evaluate two $p_3$ $p_{3}$ -PPT witnesses:
- Linear Witness: $W_{lin} = p_3 - (3p_2 - 1)/2$ . A negative value indicates entanglement.
- Quadratic Witness: $W_{quad} = p_3 - p_2^2$ . A negative value indicates entanglement and provides a quantitative bound on negativity.
Truncation and Robustness: The protocol includes a self-certification guarantee: if the truncated witness is negative, the full infinite-dimensional state is entangled, ensuring no false positives from Fock space truncation. Furthermore, local optical loss cannot create false positives, removing the need for detector-efficiency calibration for detection validity.

Key Contributions

Single-Copy Protocol: The work introduces a method to detect bipartite CV entanglement using only single-copy homodyne data, eliminating the triple-coincidence bottleneck and the need for PNR detectors.
Unbiased Estimators: The construction of unbiased U-statistic estimators for PT moments $p_2$ and $p_3$ directly from homodyne shadow data, avoiding the downward-biased eigenvalue estimates associated with full density matrix reconstruction.
Dimension-Free Quantification: The derivation of closed-form, dimension-free lower bounds on entanglement negativity based solely on $p_2$ and $p_3$ .
Robustness: Demonstration that the protocol is robust against local optical loss (no calibration needed for validity) and phase jitter, with computational costs scaling polynomially with the Fock cutoff.

Results
The protocol was benchmarked on six families of entangled states, including Gaussian (two-mode squeezed vacuum) and non-Gaussian states (photon-subtracted/added, NOON, cat, and Walschaers states).

Detection Efficiency: For states with a mean photon number $\bar{n} \approx 2$ , the protocol achieves a 95% one-sided empirical detection probability with approximately $10^3$ to $10^4$ homodyne measurements.
Specific Cases:
- NOON states: Highly efficient detection ( $T \approx 500$ for $n=2$ ) due to small Fock support.
- Walschaers state: Successfully detected entanglement in a state designed to evade all second-moment criteria, requiring $\sim 7,500$ snapshots at $\bar{n} \approx 2$ .
- Mixed States: The protocol detected entanglement in noisy Walschaers states up to a vacuum admixture of $\lambda \approx 0.53$ .
Negativity Bounds: The quadratic witness provided tight negativity bounds for NOON and cat states (capturing $\sim 100\%$ of exact negativity) but was looser for TMSV-type states (18–50%), reflecting the spectral distribution of negative eigenvalues.
Loss Tolerance: The required sample count increases smoothly with optical loss (e.g., from $\sim 3,000$ to $\sim 5,000$ for photon-subtracted TMSV at 10% loss), remaining within the reach of current homodyne experiments operating at kHz–MHz repetition rates.

Significance
The paper establishes that non-Gaussian entanglement certification does not strictly require non-Gaussian measurements. By leveraging randomized homodyne data and classical post-processing, the protocol makes the detection of challenging non-Gaussian entanglement accessible to standard laboratory setups lacking multi-copy interferometry or PNR capabilities. The authors note that while the sample complexity scales polynomially with the Fock cutoff (limiting practical application to $\bar{n} \lesssim 3-4$ ), the method covers many experimentally relevant non-Gaussian states and complements existing multi-copy approaches, particularly for laboratories without specialized hardware. The protocol places non-Gaussian entanglement detection within reach of current experimental capabilities, requiring only standard homodyne setups and classical computation.