Every Rank-Two Entangled State is Projectively Steerable

This paper proves that every rank-two bipartite entangled state is projectively steerable in at least one direction (and two-way when effective local dimensions are equal), demonstrating that the bifurcation between entanglement and steering does not occur even for the first genuinely mixed rank under projective measurements.

The Gist

The Big Picture: A Game of Remote Control

Imagine two people, Alice and Bob, who share a mysterious, linked object (a quantum state). They are far apart.

• Entanglement means their objects are linked in a way that defies normal logic.
• Steering is a specific game Alice plays: she measures her part of the object, and based on her result, she can "steer" Bob's object into a specific state. If she can do this in a way that Bob cannot explain using a pre-agreed secret plan (a "hidden variable"), she has successfully "steered" him.

For a long time, physicists knew that if the objects were in a perfectly pure state (like a single, crisp note), Alice could always steer Bob. But what about mixed states? These are "messier" states, like a chord with some noise in it.

The big question this paper answers is: Is there a "messy" state that is still linked (entangled) but impossible to steer?

The authors prove that for the first level of messiness (called "Rank Two"), the answer is NO. If the state is linked, Alice can always steer Bob, provided she uses the right kind of measurement.

---

The Core Analogy: The "Flat Spot" on a Hill

To understand the proof, imagine the world of quantum states as a giant landscape.

• The Valley (The Safe Zone): This represents states that are not linked (separable).
• The Hills: These represent linked (entangled) states.
• The Boundary: The edge where the safe valley meets the hills.

The authors discovered a rule about how these hills touch the boundary.

1. The "Pure Contact" (Finding the Edge)

The paper starts by showing that if you have a "Rank Two" state (the first level of messiness), you can always find a specific measurement Alice can make that pushes Bob's state right to the very edge of the boundary.

• Analogy: Imagine rolling a ball (Alice's measurement) down a hill. The authors prove that for this specific type of hill, the ball must roll all the way to the very edge of the cliff (a "pure contact"). You can't stop it halfway up the slope.

2. The "Wobble" (The Proof of Steering)

Once the ball is at the edge, the authors look at what happens if Alice wiggles her measurement just a tiny bit.

• The Physics: If the state is truly linked, that tiny wiggle causes Bob's state to jump sideways (linearly) along the edge.
• The Trap: If Bob were just following a pre-agreed secret plan (a "Local Hidden State"), his state would only be able to move inward or stay put (quadratically). It cannot jump sideways instantly.
• The Result: Because Bob's state jumps sideways, it proves he wasn't following a secret plan. Alice has successfully "steered" him.

3. What if the Ball Doesn't Wiggle? (The "Degenerate" Case)

The authors had to consider a tricky scenario: What if the ball hits the edge, but wiggling it doesn't make it jump sideways? (This is called a "degenerate" contact).

• The Twist: They proved that for "Rank Two" states, if this happens, the state is actually not linked at all (it's separable).
• The Logic: If the state is linked, the "wobble" must happen. If the wobble doesn't happen, the state was never linked to begin with. Therefore, for every actual linked state, the wobble exists, and steering is possible.

---

The "One-Way" vs. "Two-Way" Rule

The paper also clarifies who can steer whom, depending on the size of their "rooms" (dimensions).

• The Rule: If Alice is in a bigger room than Bob, she can definitely steer him. If they are in rooms of the same size, they can steer each other (two-way steering).
• Analogy: Think of it like a spotlight. If Alice has a huge spotlight (high dimension) and Bob has a small target (low dimension), Alice can easily hit the target. If they both have the same size spotlights, they can both hit each other.

Why This Matters (According to the Paper)

1. No Exceptions: Before this, scientists wondered if there was a "hidden" type of messy, linked state that couldn't be steered. This paper says: No. At the very first level of messiness (Rank Two), if it's linked, it's steerable.
2. No Complex Math Needed: Usually, proving steering requires complex calculations or "inequalities" (like checking a long list of rules). This paper shows you can tell if steering is possible just by looking at the shape of the state's "support" (where it exists) and its "kernel" (where it is zero).
3. A Simple Certificate: If you have a messy linked state, you don't need to run a supercomputer to find a steering strategy. You just need to find that "pure contact" point and check if the "wobble" exists. If it does, you have your proof.

Summary in One Sentence

The authors proved that for the simplest kind of "messy" linked quantum states, entanglement automatically guarantees steering, because the very geometry of these states forces a "wobble" that a secret plan could never mimic.

Technical Summary

Technical Summary: Every Rank-Two Entangled State is Projectively Steerable

Problem Statement
The paper addresses the structural relationship between quantum entanglement and Einstein–Podolsky–Rosen (EPR) steering within the hierarchy of quantum correlations. While pure entangled states are known to be steerable via projective measurements within their Schmidt supports, the behavior of mixed states remains more complex. Specifically, rank-two represents the first genuinely mixed rank where a separation (bifurcation) between entanglement and steering could theoretically occur. The central question is whether there exist rank-two bipartite entangled states that admit a Local Hidden State (LHS) model under projective measurements, thereby failing to be steerable. Previous work has established that entanglement does not guarantee steering in general (e.g., Werner states), but the status of the lowest-rank mixed states remained an open structural question.

Methodology
The authors employ a geometric and algebraic approach rooted in the support-kernel structure of density operators, rather than relying on steering-inequality optimization or reduction to two-qubit systems. The methodology proceeds through three main logical steps:

1. Rank-Forced Pure Contacts: The authors analyze the contraction map of a rank-$r$ state $\rho$ on $C^m \otimes C^n$. By viewing the unnormalized conditional states as an $m$-dimensional linear subspace $L$ within the space of $n \times r$ matrices, they apply the projective dimension theorem. They demonstrate that if the effective local dimension $m$ of the untrusted party satisfies $m > (n-1)(r-1)$, the subspace $L$ must intersect the Segre variety of rank-one matrices. This forces the existence of a "pure contact"—a projective outcome on Alice that prepares a pure (rank-one) conditional state on Bob.
2. Boundary-Contact Certificates: Upon establishing a pure contact, the authors examine the local geometry at the boundary of the trusted state cone. They utilize a "support-kernel" condition: if a neighboring projective outcome produces a first-order tangent block connecting the support of the pure state to its kernel (a nonzero matrix element $\langle \phi | B | \beta \rangle$ where $\phi$ is in the support and $\beta$ in the kernel), this creates a linear scaling in the off-diagonal terms versus a quadratic scaling in the diagonal terms. This geometric mismatch prevents any fixed finite hidden-state measure from reproducing the statistics, proving steering. Furthermore, this same block constitutes a Negative Partial Transpose (NPT) minor.
3. Degenerate-Contact Reduction: The proof addresses the case where the initial pure contact is "degenerate" (i.e., the support-kernel block vanishes). For rank-two states, the authors prove that such degeneracy forces the state to decompose into a product component and a residual component. If the original state is entangled, this residual component must be an entangled pure state. This residual state necessarily possesses a nondegenerate pure contact in the same direction, thereby closing the logical gap.

Key Contributions

• Theorem 1 (Complete Rank-Two Theorem): The paper proves that every rank-two bipartite entangled state is projectively steerable in at least one direction. Specifically, if the effective local dimensions are $m$ and $n$, steering from the larger dimension to the smaller ($m \ge n \implies A \to B$) is guaranteed. If $m=n$, the state is two-way projectively steerable.
• Support-Kernel Geometry as a Certificate: The authors establish that the support and kernel of a low-rank state provide a direct, basis-independent certificate for steering. This certificate requires only the identification of a contact vector, a tangent vector, and a kernel direction, without the need for constructing auxiliary ensembles or optimizing steering inequalities.
• Equivalence in the Rank-Two Sector: The work demonstrates that within the rank-two sector, entanglement is equivalent to NPT, which is in turn equivalent to projective steering (given the appropriate directional constraints). There are no "exceptional" rank-two entangled families hidden behind a projective LHS model.

Results
The primary result is the classification of the entire rank-two sector as a "complete projective-steering sector."

• Directionality: In rectangular dimensions ($m \neq n$), steering is guaranteed from the larger effective dimension to the smaller. In equal dimensions, it is two-way.
• No Bifurcation: The potential bifurcation between entanglement and steering does not occur at rank two; the gap closes completely for this stratum.
• Certificate Efficiency: The paper provides a practical, low-rank certificate. If a rank-two state is entangled, one can verify steering by checking for a nondegenerate support-kernel block in the tangent space of a forced pure contact. This is computationally more robust than full Semidefinite Programming (SDP) searches or inequality optimization.

Significance
The paper claims to close the first genuinely mixed-state test of the entanglement-steering hierarchy. By proving that rank-two entangled states are always steerable under projective measurements, the authors push the first possible separation between entanglement and steering beyond rank two. The significance lies in shifting the paradigm from inequality-based verification to structural verification: the density operator's own support and kernel geometry contain the steering certificate. This implies that for low-rank states, structural data (often available from source constraints or reconstruction) is sufficient to determine steering status, making the verification process more efficient and robust against noise or model assumptions. The work ties the EPR steering hierarchy to determinantal algebraic geometry and low-rank PPT separability, offering a unified geometric perspective on quantum correlations.