Detecting bipartite entanglement with PnCP maps and non-negative polynomials

This paper presents a numerically robust implementation of an algorithm for generating Positive non-Completely Positive (PnCP) maps via non-Sum-of-Squares polynomials, demonstrating their theoretical uniqueness and superior ability to detect PPT entangled states compared to existing criteria.

The Gist

The Big Picture: Finding the "Invisible" Glue

Imagine you have two boxes. Sometimes, the contents of these boxes are just two separate things sitting next to each other (like a sock in one box and a shoe in another). We call this separable. But sometimes, the contents are magically linked in a way that defies normal physics; what happens to the sock instantly affects the shoe, no matter how far apart they are. This is entanglement.

The problem scientists face is: How do we tell the difference?
For simple boxes, we have easy tests. But for more complex boxes (specifically 3x3 dimensions), there are "tricky" states that look separable to our standard tests but are actually entangled. These are like "ghosts" that hide in plain sight.

The Old Tools vs. The New Tool

For a long time, scientists used a tool called the Partial Transpose (think of it as a specific type of mirror). If you look at the state in the mirror and it looks "broken" (negative), you know it's entangled. But if it looks "okay" (positive), the mirror says, "I don't know, it might be separable."

However, the "ghost" states mentioned above pass the mirror test. They look positive, so the mirror fails to catch them.

The authors of this paper introduce a new, more sensitive tool based on Positive Non-Completely Positive (PnCP) maps.

• The Analogy: Imagine you have a sieve (a filter) that catches big rocks but lets sand through. The old mirror test is a sieve with big holes; it catches the obvious entangled states but lets the "ghost" states slip through.
• The new PnCP maps are like a sieve with much finer mesh. They are mathematical tools designed specifically to catch those "ghost" states that the old mirror misses.

How They Built the New Tool

The authors didn't just guess how to build this new sieve. They used a clever connection between two different worlds: Quantum Physics and Polynomials (math equations with variables like $x$ and $y$).

1. The Math Trick: They looked at a specific type of math equation (a polynomial) that is always positive (never goes below zero) but cannot be built by simply adding up squares of other equations. In math, these are rare and special "non-sum-of-squares" polynomials.
2. The Translation: They used a mathematical "translator" (an isomorphism) to turn these special, tricky polynomials into the quantum "sieves" (PnCP maps) needed to catch the entangled states.
3. The Code: They wrote a computer program (available on GitHub) to automatically generate these polynomials and turn them into working quantum detectors. They added a special "safety check" to make sure the computer didn't make tiny calculation errors that would ruin the results.

What They Found

The authors tested their new detectors against a library of 2,000 tricky "ghost" states (PPT entangled states). Here is what happened:

• The Old Guards Failed: When they ran these states through standard, well-known tests (like the "Realignment" criterion or the "Covariance Matrix" test), 98.3% of the time, the tests said, "These are safe/separable." The tests missed the entanglement.
• The New Tool Succeeded: Their new PnCP maps successfully detected the entanglement in these states.
• The "Ghost" Nature: The authors found that these new maps are very sensitive. They sit right on the edge of the mathematical "cone" of valid detectors. This means they are great at finding the specific "ghost" states, but they are fragile. If you add a little bit of noise (like static on a radio), they might stop working. They are precise, not robust.

The "Family" of Detectors

The paper also discovered something interesting about how these tools work.

• Usually, one map creates one specific detector (like a single flashlight beam).
• The authors showed that you can actually create a whole family of detectors from that single map by slightly changing the angle of the beam.
• By testing many different angles (using different "Schmidt rank" states), they could find a better angle that caught the entangled states even more clearly than the standard "Choi" detector.

What They Did NOT Claim

It is important to note what the paper does not say:

• They did not claim this is a practical, everyday tool for engineers yet. The math is complex, and the detectors are fragile against noise.
• They did not claim this solves the problem of finding entanglement in all cases instantly. The paper admits that finding these states is computationally hard (NP-hard).
• They did not suggest using machine learning to "train" these maps on specific states. They analyzed the algorithm and found that the random choices made during the process don't change smoothly, meaning a simple "learning" approach wouldn't work easily.

Summary

In short, the authors built a new, highly specialized mathematical "net" to catch a specific type of quantum entanglement that has been hiding from our best existing nets. They proved mathematically that this net works, showed it catches states others miss, and made the code public so others can try it. However, the net is delicate and sits right on the edge of the mathematical rules, meaning it's a powerful theoretical discovery rather than a rugged, ready-to-use industrial tool.

Technical Summary

Technical Summary: Detecting Bipartite Entanglement with PnCP Maps and Non-Negative Polynomials

1. Problem Statement

The characterization of entangled states, known as the separability problem, is a central challenge in quantum information theory. While the Positive Partial Transpose (PPT) criterion provides a necessary and sufficient condition for separability in low-dimensional systems (specifically $2 \times 2$ and $2 \times 3$), it fails to detect "bound entangled" states (PPT entangled states) in higher dimensions. Although complete hierarchies of criteria exist (e.g., the Doherty-Parrilo-Spedalieri or DPS hierarchy based on Semi-Definite Programs), their computational cost scales exponentially, rendering them impractical for many applications.

Positive but not Completely Positive (PnCP) maps serve as a dual tool to entanglement detection: a state $\rho$ is entangled if and only if there exists a PnCP map $\Lambda$ such that $(\mathbb{I} \otimes \Lambda)(\rho)$ is not positive semi-definite. However, constructing new, effective PnCP maps is difficult. Recent theoretical work established a link between PnCP maps and positive polynomials that are not Sum-of-Squares (non-SOS). While an algorithm (KSMZ) was proposed to generate such maps, previous implementations suffered from numerical instabilities, leading to false positives in entanglement detection.

2. Methodology

2.1 Theoretical Framework

The paper relies on the isomorphism between:

1. Positive Maps and Polynomials: The KSMZ isomorphism maps linear maps $\Phi: S_n(\mathbb{R}) \to S_m(\mathbb{R})$ to bihomogeneous polynomials of bidegree $(2,2)$. A map is Positive if and only if the associated polynomial is non-negative on product vectors.
2. Complete Positivity and SOS: A map is Completely Positive (CP) if and only if its associated polynomial admits a Sum-of-Squares (SOS) decomposition.
3. Indecomposability and Non-RSOS: A map is indecomposable (capable of detecting PPT entangled states) if and only if its associated Hermitian polynomial is not a Real Sum-of-Squares (RSOS).

The authors utilize the duality between the DPS hierarchy (for states) and the SOS hierarchy (for polynomials) to classify the generated maps.

2.2 Algorithm Implementation

The authors implemented the KSMZ algorithm (from [1]) in MATLAB, making it available on GitHub. The algorithm constructs positive non-SOS polynomials in two steps:

1. Quotient Ring Construction: It works in the quotient ring $\mathbb{R}[z]/I_{n,m}$, where $I_{n,m}$ encodes the rank-one constraint (the Segre variety). It selects random points on the Segre variety and constructs a base SOS term $H(z)$ vanishing at these points.
2. Perturbation: It adds a perturbation term $f(z)$ that vanishes to second order at the prescribed points but lies outside the span of the products of the linear forms defining $H$. This ensures the resulting polynomial $q_\delta = H + \delta f$ is non-negative on the Segre variety but not SOS in the quotient ring.
3. Pullback: The polynomial is pulled back to a biquadratic form $p(x,y)$, which corresponds to a PnCP map.

2.3 Numerical Robustness

A critical contribution of the implementation is a new numerical test to guarantee positivity. Previous implementations (e.g., [29]) suffered from numerical imprecisions where maps generated were not strictly positive. The authors:

• Use Lasserre's hierarchy to compute a global lower bound on the polynomial values.
• Leverage the property that for non-constant bihomogeneous polynomials, the infimum is either $0$ or $-\infty$.
• Retain only polynomials where the computed lower bound is non-negative, ensuring the generated maps are strictly positive.

2.4 Entanglement Witnesses (EWs)

To apply these maps in optimization (SDP), the authors convert PnCP maps into Entanglement Witnesses via the Choi-Jamiołkowski isomorphism. They further propose a method to recover the full detection power of a map by considering a family of witnesses $\{W_{\Phi, \eta}\}$ derived from the adjoint map $\Phi^\dagger$ acting on vectors $|\eta\rangle$ with maximal Schmidt rank, rather than relying solely on the standard Choi witness.

3. Key Contributions

1. Robust Implementation: A numerically stable implementation of the KSMZ algorithm that filters out false positives, providing a reliable library of PnCP maps.
2. Theoretical Characterization:
   • Indecomposability: The authors prove analytically that the maps generated by the algorithm are indecomposable. This is demonstrated by showing the associated Hermitian polynomials are not RSOS, implying they can detect PPT entangled states.
   • Boundary Location: They prove these maps lie on the boundary of the cone of positive maps, explaining their sensitivity to noise.
   • Inequivalence: The maps are shown to be inequivalent to known families such as the generalized Choi maps and Breuer-Hall maps. Specifically, KSMZ maps lack a skew-symmetric part in their real symmetric core, a structural difference that prevents them from being equivalent to these other maps.
3. Witness Optimization: The paper demonstrates that a single PnCP map generates a family of witnesses. By optimizing over the choice of the vector $|\eta\rangle$ (specifically those with maximal Schmidt rank), one can improve the detection capability (violation magnitude) compared to the standard Choi witness.

4. Results

• Detection of PPT States: Numerical experiments on $3 \times 3$ systems show that the KSMZ maps can detect PPT entangled states that are missed by standard criteria (PPT, realignment, covariance matrix).
• Comparison with QETLAB: When tested against the IsSeparable function in the QETLAB package (which implements a hierarchy of 19 criteria), the KSMZ witnesses detected entanglement in states that IsSeparable failed to classify (specifically, 98.3% of the maximally violating states in their dataset were undetected by standard criteria).
• Magnitude of Violation: The detection of PPT states is "weak" in terms of magnitude. The PPT margin (the negative expectation value on PPT states) is typically on the order of $10^{-6}$ to $10^{-8}$. This indicates the witnesses lie very close to the decomposable cone, consistent with their theoretical location on the boundary of the positive cone.
• Inequivalence Verification: The authors verified that the KSMZ maps detect states that the Partial Transpose criterion misses, and conversely, that the Partial Transpose detects NPT states that the KSMZ maps (in their specific configuration) might not, confirming they are distinct tools.
• Optimization Impact: Using the family of witnesses (optimizing over invertible matrices $A$) improved the violation magnitude for some maps (e.g., from $-5.02 \times 10^{-8}$ to $-1.37 \times 10^{-7}$), though the improvement was not uniform across all maps.

5. Significance and Claims

The paper claims to provide a new, robust criteria for entanglement detection based on the constructive generation of PnCP maps from positive non-SOS polynomials.

• Theoretical Insight: It clarifies the geometric and algebraic relationship between polynomials, maps, and operators, specifically establishing that the KSMZ construction yields indecomposable maps that are structurally distinct from the famous Choi and Breuer-Hall maps.
• Practical Utility: The authors argue that while the detection of PPT states by these maps is numerically fragile (lying close to the decomposable boundary), they offer a unique capability to detect specific PPT entangled states that other standard criteria miss.
• Limitations: The authors are modest about the scalability and robustness. They note that the detection power is limited by the maps' proximity to the decomposable cone, making them sensitive to noise. Furthermore, they explicitly state that randomly generating and applying these maps does not constitute a polynomial-time solution to the separability problem for unknown states, as the generation process itself is not continuous with respect to the target state, hindering machine learning approaches.
• Future Work: The paper suggests that the constructed library of maps could be integrated into existing tools (like IsSeparable) as an additional, inequivalent test. They also propose future work on adding skew-symmetric parts to KSMZ maps to potentially create new classes of PnCP maps.

In summary, the work bridges polynomial optimization and quantum information to provide a numerically verified method for generating a new class of entanglement witnesses, expanding the toolkit for detecting bound entanglement in dimensions where standard criteria fail.