Separable neighbourhood of identity in C$^{\ast}$-algebras

This paper investigates the existence and size of separable neighborhoods around the identity in bipartite C*-algebras by reducing the problem to estimating completely bounded norms of contractive positive maps, thereby characterizing these neighborhoods via the algebra's rank and resolving a conjecture by Musat and Rørdam.

The Gist

The Big Picture: Entanglement vs. Separation

Imagine you are a quantum physicist. In your world, you have two systems (let's call them Alice and Bob). Sometimes, they are completely independent; you can describe Alice's state and Bob's state separately. This is called being separable.

Other times, they are entangled. This is a spooky connection where you cannot describe Alice without describing Bob. They are inextricably linked.

The central question of this paper is: How "strong" does a connection have to be to become entangled?

Specifically, the authors look at the "Identity Element." In quantum mechanics, the identity represents a state of perfect balance or "maximum mixedness" (like a perfectly shuffled deck of cards where nothing is special). It is the safest, most neutral starting point.

The paper asks: If we nudge this neutral starting point just a tiny bit, does it immediately become entangled? Or is there a "safe zone" (a neighborhood) around the center where everything stays separable?

---

The Analogy: The "Safe Zone" in a Forest

Think of the Identity Element as a campfire in the middle of a forest.

• Separable elements are like dry, safe wood around the fire.
• Entangled elements are like a sudden, wild forest fire that spreads uncontrollably.

In the world of small, finite systems (like standard computer chips or simple quantum bits), we already knew there was a Safe Zone. If you stand close enough to the campfire (the identity), you are safe. You have to walk a certain distance away before you hit the "fire" (entanglement).

The size of this safe zone depends on the complexity of the forest. In a small forest (finite dimensions), the safe zone is a nice, big circle.

The Problem: What happens in an infinite forest? What if the forest goes on forever (infinite-dimensional C∗-algebras)? Does the safe zone disappear? Is the fire so close that even standing right next to the campfire is dangerous?

---

The Discovery: The "Rank" of the System

The authors, Mizanur Rahaman and Mateusz Wasilewski, discovered that the size of this safe zone depends entirely on a property they call Rank.

Think of Rank as the "maximum complexity" or the "number of distinct colors" a system can display.

• Finite Rank: The system is like a limited palette of colors (e.g., only 3 colors). It's manageable.
• Infinite Rank: The system has an infinite number of colors. It's chaotic and unbounded.

The Main Result

The paper proves a simple rule for the size of the Safe Zone (let's call it $r$):

$$r = \frac{1}{\text{Minimum Rank of the two systems}}$$

1. If both systems have Finite Rank: There is a safe zone!

   • If Alice has 3 colors and Bob has 5, the safe zone radius is $1/3$. You can wiggle a little bit, and you are still safe.
   • The smaller the rank, the bigger the safe zone.

2. If at least one system has Infinite Rank: The safe zone vanishes.

   • If Alice is infinite (like an infinite-dimensional Hilbert space), the radius becomes $1/\infty = 0$.
   • Meaning: In an infinite system, there is no safe zone. Even the tiniest, tiniest nudge away from the identity creates entanglement. The "fire" is touching the campfire.

The Secret Weapon: The "Magic Mirror" (Positive Maps)

How did they figure this out? They didn't just look at the wood; they looked at the rules governing the wood.

They used a mathematical tool called Positive Maps. Imagine a "Magic Mirror" that takes an object and reflects it.

• If the mirror is "Positive," it keeps things safe (doesn't turn safe wood into fire).
• If the mirror is "Completely Bounded," it has a limit on how much it can distort things.

The authors realized that the size of the safe zone is directly linked to the strength of these Magic Mirrors.

• They proved that calculating the size of the safe zone is exactly the same as calculating the maximum "distortion power" (the completely bounded norm) of these mirrors.
• They found that the distortion power is limited by the Rank. If the Rank is infinite, the mirrors can distort things infinitely, meaning the safe zone collapses to zero.

Why This Matters

1. Solving a Mystery: This answers a long-standing question about whether entanglement is "dense" (everywhere) in infinite systems. The answer is yes. In infinite systems, entanglement is everywhere; you can't find a safe spot near the center.
2. Settling a Bet: The paper resolves a recent conjecture by mathematicians Musat and Rørdam. They guessed that the "complexity" (Rank) of the system dictates the behavior of entanglement. This paper proves them right and gives the exact formula.
3. Practical Implication: If you are building a quantum computer using infinite-dimensional systems (which is theoretically possible), you have to be incredibly careful. You cannot rely on a "buffer zone" of safety. The moment you deviate from the perfect identity, you are entangled.

Summary in One Sentence

The paper proves that in the quantum world, the size of the "safe zone" where things aren't entangled is determined by the system's complexity (Rank); if the system is infinitely complex, that safe zone shrinks to nothing, meaning entanglement is unavoidable even for the slightest change.

Technical Summary

1. Problem Statement

The paper addresses the structural properties of separable elements within bipartite $C^*$-algebras. Specifically, it investigates the existence and size of a separable neighbourhood around the identity element ($1_A \otimes 1_B$).

• Context: In finite-dimensional quantum systems (matrix algebras), it is well-established that sufficiently small perturbations of the maximally mixed state (the identity) remain separable. This implies a "ball" of separable states exists around the identity.
• The Gap: While this phenomenon is understood for matrix algebras $M_n(\mathbb{C})$, its extension to general (potentially infinite-dimensional) $C^*$-algebras was an open problem.
• Core Question: Given two unital $C^*$-algebras $A$ and $B$, what is the radius $\gamma(A, B)$ of the largest ball centered at $1_A \otimes 1_B$ (in the operator norm) such that every element $1_A \otimes 1_B - x$ with $\|x\| \leq \gamma(A, B)$ is separable?
• Definitions:
  • Separable Element: An element in the closure of the convex cone generated by positive tensors $a \otimes b$ where $a \in A_+, b \in B_+$.
  • Rank of a $C^*$-algebra: Defined as the supremum of the dimensions of its irreducible representations. An algebra is subhomogeneous if this rank is finite; otherwise, it has infinite rank.

2. Methodology

The authors employ a multi-faceted approach combining operator space theory, the theory of completely positive maps, and von Neumann algebra techniques.

• Reduction to Completely Bounded (cb) Norms:
  The central insight is that the size of the separable neighbourhood is inversely related to the supremum of the cb-norms of contractive positive maps between the algebras.

  • Let $\eta(A, B) = \sup \{ \|\Phi\|_{cb} : \Phi: A \to B \text{ is a contractive positive map} \}$.
  • The authors establish the duality: $\gamma(A, B) = 1/\eta(A, B)$.

• Generalization of the Horodecki Criterion:
  The paper extends the famous Horodecki criterion (which characterizes separability via positive maps) from finite dimensions to general $C^*$-algebras.

  • They prove that for a nuclear $C^*$-algebra $A$, an element $x \in A \otimes B$ is separable if and only if $(\text{Id} \otimes \phi)(x) \geq 0$ for all finite-rank positive maps $\phi: B \to A$.
  • This is achieved by lifting the problem to the biduals ($A^{**}$ and $B^{**}$), which are von Neumann algebras, and utilizing the injectivity of the bidual of a nuclear algebra.

• Structural Analysis via Biduals:
  The authors analyze the structure of subhomogeneous $C^*$-algebras. They show that if $A$ is subhomogeneous of rank $n$, its bidual $A^{**}$ decomposes into a direct sum of algebras of the form $M_k \otimes L^\infty(X)$. This allows the reduction of infinite-dimensional problems to matrix-algebra problems combined with commutative algebras.

• Approximation Techniques:
  The proofs utilize Kaplansky's density theorem to approximate elements in the bidual by elements in the original algebra and employ local reflexivity properties of nuclear algebras to approximate maps.

3. Key Contributions and Results

A. The Main Theorem

The paper provides a complete characterization of the separable neighbourhood radius $\gamma(A, B)$ in terms of the rank of the algebras.

Theorem: For unital $C^*$-algebras $A$ and $B$:
$$\gamma(A, B) = \frac{1}{\min(\text{rank}(A), \text{rank}(B))}$$
(Note: If $\min(\text{rank}(A), \text{rank}(B)) = \infty$, then $\gamma(A, B) = 0$.)

Equivalently, the supremum of the cb-norms of contractive positive maps is:
$$\eta(A, B) = \min(\text{rank}(A), \text{rank}(B))$$

B. Specific Findings

1. Infinite Rank Case: If either $A$ or $B$ has infinite rank (e.g., $B(H)$ for a separable infinite-dimensional Hilbert space), then $\gamma(A, B) = 0$. This means there is no non-trivial separable neighbourhood around the identity in such systems. This contrasts with the finite-dimensional case and aligns with results showing entangled states are dense in certain topologies for infinite-dimensional systems.
2. Subhomogeneous Case: If both algebras are subhomogeneous, a non-trivial separable ball exists. The radius is determined strictly by the smaller of the two ranks.
3. Matrix Algebra Recovery: The result recovers the known finite-dimensional result by Gurvits and Barnum: for $M_n \otimes M_m$, the radius is $1/\min(n, m)$.
4. Extension to General Maps: The authors also derive bounds for general linear maps (not just positive ones), showing $\|\Psi\|_{cb} \leq \min(n, m)\|\Psi\|$ for maps between matrix algebras, extending previous work by Ando.

C. Resolution of the Musat-Rørdam Conjecture

The paper resolves a conjecture from Musat and Rørdam (2025) regarding the quantity $\kappa(A, B)$, defined as the supremum of norms of functionals positive on separable elements.

• Result: $\kappa(A, B) = \min(\text{rank}(A), \text{rank}(B))$.
• This confirms that the "size" of the set of separable-positive functionals is directly governed by the algebraic rank, providing a unified view of separability and entanglement in $C^*$-algebras.

4. Significance

• Bridging Finite and Infinite Dimensions: The paper successfully generalizes a fundamental concept of quantum information theory (separability around the identity) from finite-dimensional matrix algebras to the broad class of $C^*$-algebras.
• Structural Insight: It establishes a profound link between the geometric property of separability (existence of a ball) and the algebraic property of the rank (dimension of irreducible representations).
• Operator Space Theory: The work deepens the understanding of completely bounded norms of positive maps, showing that the "worst-case" cb-norm is dictated by the largest matrix block present in the algebra's structure.
• Implications for Quantum Information:
  • It clarifies that in infinite-dimensional systems (like those involving $B(H)$), the identity element is a "singular" point regarding separability; any perturbation, no matter how small, can generate entanglement.
  • It provides a precise criterion for when a quantum system is "robust" against entanglement generation near the maximally mixed state: the system must be subhomogeneous.

In summary, the paper provides a definitive answer to the separable neighbourhood problem in $C^*$-algebras, proving that the existence and size of such a neighbourhood are entirely determined by the finite rank of the underlying algebras, and resolving a recent conjecture in the field.