⚛️ quantum physics

Local Gaussian bounds on the non-destructive discrimination of two-mode squeezed states

This paper investigates the non-destructive discrimination of two-mode squeezed vacuum states using local Gaussian measurements, establishing a tradeoff between discrimination success and state fidelity that can be surpassed by leveraging pre-shared entanglement, thereby extending information-disturbance principles to infinite-dimensional continuous-variable systems.

Original authors: Mi-Jung So, James Moran, Youngrong Lim, Mahn-Soo Choi, Hyukjoon Kwon

Published 2026-09-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mi-Jung So, James Moran, Youngrong Lim, Mahn-Soo Choi, Hyukjoon Kwon

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

In the quantum world, the act of looking at something often changes it. When scientists measure a delicate quantum state, the standard rule is that the measurement destroys the original information, leaving behind a new, altered state. This is a fundamental hurdle for technologies like quantum networks, where information needs to be processed and identified without being wiped out, so that the delicate connections between particles can be reused later. The challenge is to find a way to tell two very similar quantum states apart while keeping the original state as intact as possible. This balance between gaining information and causing disturbance is a central problem in quantum physics, especially when dealing with continuous systems where the variables can take on any value, rather than just discrete steps like in a computer bit.

A team of researchers has tackled this problem by focusing on a specific type of quantum state known as a two-mode squeezed vacuum. These are pairs of light beams that are deeply linked, or entangled, in a way that their properties are correlated. The researchers wanted to know if they could distinguish between two such pairs that differ only in the direction of their squeezing, using only local measurements and classical communication, without destroying the entanglement. They discovered a precise trade-off: the more accurately you identify the state, the more you disturb it. However, by carefully designing a measurement process that involves mixing the incoming light with a local reference beam and then applying a specific correction based on the measurement result, they found an optimal strategy. This strategy, which uses only standard optical tools available in laboratories, achieves a maximum combined score of about 0.627 when the squeezing is strong. This score represents the product of the chance of being right and the quality of the state left behind. The team proved that for their specific class of operations, this is the best possible outcome, establishing a new fundamental limit for this type of quantum task.

The researchers also explored what happens if the two parties performing the measurement are allowed to share an extra pair of entangled beams before they begin. This pre-shared entanglement acts as a resource that can be used to improve the process. They found that if the shared resource is perfectly matched to the state being measured, the task becomes trivial, and the success rate can approach perfection as the squeezing increases. However, in more realistic scenarios where the shared resource is weaker than the state being measured, the situation is more complex. The team showed that even with a weaker resource, it is possible to beat the limit found without any shared entanglement, provided the parties mix the incoming signal with their shared resource in a specific way. They identified a critical point where the strategy must shift from simply measuring everything and rebuilding the state to a more sophisticated mixing process that preserves more of the original information.

The work provides a clear map of how much information can be extracted from these continuous quantum states without destroying them. The researchers simulated various optical circuits to find the best possible settings for their measurements. They found that a relatively simple setup, involving a beam splitter to mix the light, a measurement of the resulting signal, and a local adjustment to the remaining light, is sufficient to reach the theoretical limit for their class of operations. This limit holds true as long as the initial squeezing is strong enough. For weaker states, the best strategy is different, relying more on simple displacement adjustments rather than complex squeezing operations. The study confirms that while there is a hard ceiling on how well one can perform this task without extra help, that ceiling can be raised if the parties are allowed to bring their own entangled resources to the table.

This research extends a known principle from simpler, finite-dimensional systems to the more complex world of continuous variables. In the past, limits on information gain versus disturbance were calculated based on the number of possible states a system could hold. Since continuous systems have an infinite number of possible states, those old formulas did not work. The authors had to build a new framework from the ground up, focusing on the specific properties of the light beams rather than the size of the system. Their results show that even in this infinite-dimensional realm, there are strict, calculable bounds on how much one can learn without causing damage. The findings suggest that while local operations are powerful, they are not limitless, and that the judicious use of pre-shared entanglement is a key factor in pushing those limits further. The study leaves open the question of whether even more complex, multi-stage adaptive measurements could push the score slightly higher, but for the single-stage operations they tested, the optimal path has been clearly charted.

Technical Summary: Local Gaussian Bounds on the Non-Destructive Discrimination of Two-Mode Squeezed States

Problem Statement
Quantum state discrimination is a fundamental task in quantum information, yet standard protocols are typically destructive, irretrievably altering the state upon measurement. In applications such as quantum networks, preserving the quantum state after discrimination is crucial for reusing entanglement. While non-destructive discrimination has been quantified in finite-dimensional systems, extending these results to infinite-dimensional continuous-variable (CV) systems presents a challenge. Existing bounds derived from finite-dimensional Hilbert space dimensions yield trivial results when applied to CV systems. This work addresses the specific problem of non-destructive discrimination between two two-mode squeezed vacuum (TMSV) states, $|+\lambda\rangle$ and $|-\lambda\rangle$ , using only Local Gaussian Operations and Classical Communication (GLOCC). The goal is to establish a fundamental tradeoff between the success probability of discrimination ( $P$ ) and the fidelity ( $F$ ) of the post-measurement state relative to the initial state.

Methodology
The authors formulate the problem within the framework of Gaussian states and operations, utilizing the symplectic structure of phase space. They consider a scenario where two distant parties, Alice and Bob, share a TMSV state with unknown squeezing sign ( $\pm \lambda$ ). The protocol involves:

Local Gaussian Circuits: Each party mixes their signal mode with an ancillary mode (initially vacuum, later potentially a pre-shared TMSV) via a local beamsplitter with angle $\theta$ .
Measurement: Homodyne detection is performed on the ancillary output modes to obtain classical outcomes $M = \{x_3, x_4\}$ .
Adaptive Processing: Based on classical communication of $M$ , the parties apply conditional local Gaussian unitaries ( $\hat{U}_{Gi}(M)$ ) to the surviving signal modes. These unitaries include displacements, phase rotations, and local squeezing.
Optimization: The study focuses on the GLOCC $_1$ regime (at most one adaptive stage). The performance metric is the "score function," defined as the product of the success probability and the average fidelity ( $P \cdot F$ ). The authors analytically derive the optimal parameters for specific circuit configurations and numerically optimize the general circuit to find upper bounds.

Key Contributions and Results

GLOCC $_1$ Bound without Pre-shared Entanglement:
For the case where ancillary modes are vacuum ( $l=0$ ), the authors derive an explicit tradeoff relation. They demonstrate that while simple displacement strategies improve fidelity, the inclusion of local squeezing on the post-measurement modes is critical.
- By optimizing the local squeezing parameter $\gamma$ and the beamsplitter angle $\theta$ , the authors find that for sufficiently large input squeezing ( $\lambda \geq \lambda_c \approx 0.349$ ), the score function saturates at a constant maximum value: $P \cdot F \approx 0.626829$ .
- This maximum is achieved at an optimal success probability $P_{opt} \approx 0.713804$ .
- For weakly squeezed states ( $\lambda < \lambda_c$ ), the optimal strategy coincides with the boundary condition of full reflection ( $\theta = \pi$ ), and the score depends on $\lambda$ .
- Numerical optimization of the fully general local Gaussian circuit confirms that this locally squeezed protocol saturates the GLOCC $_1$ bound, suggesting it is optimal within this class.
Role of Pre-shared Entanglement:
The authors extend the analysis to include pre-shared entanglement between Alice and Bob (ancillary modes prepared in a TMSV with squeezing $l \neq 0$ ).
- Trivial Regime: When the pre-shared entanglement is fully reflective ( $\theta = \pi$ ), the protocol effectively measures the input and re-prepares the pre-shared state. If the pre-shared state matches the input ( $l = \lambda$ ), the score approaches unity as $\lambda \to 1$ . However, for mismatched states ( $l < \lambda$ ), this "trivial" strategy performs poorly for large $\lambda$ , eventually falling below the GLOCC $_1$ bound without entanglement.
- Non-trivial Regime: For mismatched pre-shared states ( $l < \lambda$ ) and large input squeezing, the optimal strategy departs from full reflection. The authors show that mixing the input with the pre-shared entanglement at the beamsplitters, combined with conditional local squeezing, allows the protocol to exceed the GLOCC $_1$ bound derived for the vacuum-ancilla case.
- Crucially, any non-zero amount of pre-shared entanglement allows the score to surpass the limit achievable without it, provided the protocol enters the non-trivial regime where mixing occurs.

Significance and Claims
The paper claims to provide the first infinite-dimensional generalization of the information-gain versus disturbance tradeoff previously established for finite-dimensional systems. By circumventing the reliance on Hilbert space dimension, the authors establish a concrete, non-trivial bound for CV systems.

The work demonstrates that local squeezing is a necessary resource for optimizing the fidelity-success tradeoff in non-destructive discrimination, extending its known utility from teleportation to state discrimination.
It establishes that pre-shared entanglement acts as a resource that can strictly improve the discrimination score beyond the limits of local Gaussian operations without such resources, even when the pre-shared state does not perfectly match the target state.
The authors explicitly note that their results are confined to the GLOCC $_1$ regime and Gaussian operations. They leave open the question of whether fully adaptive protocols (GLOCC $_k$ with $k>1$ ) or non-Gaussian resources could further improve these bounds, noting that non-Gaussian resources are generally required for true quantum advantage in CV systems.

In summary, the paper rigorously defines the limits of non-destructive discrimination for TMSV states under local Gaussian constraints, identifying specific optimal protocols and demonstrating how pre-shared entanglement can be leveraged to overcome these local bounds.

More like this