🔢 mathematics

Quantum steering is equivalent to state-preserving conditional expectations

This paper establishes that for pure global states in systems with infinitely many degrees of freedom, the ability to steer any ensemble decomposition of a marginal state via measurements on a commuting subsystem is equivalent to the existence of a state-preserving conditional expectation from the commutant onto the subsystem, thereby linking quantum steering to subfactor theory and the failure of Haag duality.

Original authors: Lauritz van Luijk, Amine Marrakchi, Tobias Osborne, Alexander Stottmeister, Henrik Wilming

Published 2026-08-12

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lauritz van Luijk, Amine Marrakchi, Tobias Osborne, Alexander Stottmeister, Henrik Wilming

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 the universe not as a collection of tiny billiard balls, but as a vast, humming orchestra where every instrument is connected to every other. In the world of quantum physics, this connection is called "entanglement." It's the spooky magic where two particles can be so deeply linked that measuring one instantly tells you something about the other, no matter how far apart they are. Usually, when we study these connections, we imagine them in simple, finite boxes—like two dice that always roll matching numbers. But the real universe is messy and infinite. When physicists try to apply their simple rules to these infinite systems, things get weird. One of the biggest headaches is "purification." Think of it like this: if you have a blurry, incomplete photo of a scene (a "mixed state"), you might wonder if there's a perfect, high-definition original photo (a "pure state") somewhere that explains the blur. In simple quantum systems, there's usually only one way to find that perfect original. But in infinite systems, that rule breaks down. You might have a blurry photo, but there could be a million different high-definition originals that all look the same when you squint. This paper dives into that infinite, messy world to figure out what rules still hold when the simple ones fail.

The authors of this paper are tackling a specific puzzle about how we can "steer" these quantum states. Imagine you and a friend are holding two halves of a quantum puzzle. If you measure your half in a certain way, you can force your friend's half to collapse into a specific pattern. This is called "steering." In the simple, finite world, if you can steer your friend, they can usually steer you back, and everything is neat and tidy. But in these infinite, complex systems, the rules get fuzzy. The paper asks: Under what conditions can one side of a quantum system still steer the other, even when the universe is infinite and the usual rules of "uniqueness" have broken?

The researchers discovered a surprising bridge between quantum steering and a branch of mathematics called "subfactor theory." They found that the ability to steer an infinite quantum system is exactly the same as the existence of a special mathematical tool called a "state-preserving conditional expectation." To use a metaphor, imagine you have a giant, complex library (the whole quantum system) and a small, specific section of it (the subsystem). A "conditional expectation" is like a librarian who can take any book from the whole library and summarize it perfectly for the small section, without losing any of the original story's flavor. The paper proves that if this "perfect summarizer" exists, then you can steer the system. If it doesn't exist, you can't.

This isn't just a guess; the authors provide a rigorous mathematical proof. They show that for a pure quantum state (the high-definition original), the power to steer any possible arrangement of the subsystem is equivalent to having this special mathematical map. They also explore what happens when things get more complicated, like when the system is in a "mixed" or blurry state. They find that even then, the ability to steer depends on whether the system can be broken down into a simple, pure part and a separate, unentangled part.

One of the most interesting things the paper rules out is the idea that "two-way steering" (where both sides can steer each other) is always possible just because the system is "tomographically complete" (meaning we can measure everything). In the finite world, this is true. But in the infinite world, the authors prove that you can have a system where you can steer your friend, but they cannot steer you back, even if you have all the information you need. This happens specifically when a mathematical condition called "Haag duality" fails. It's like having a two-way street where one lane is blocked; you can drive to your friend, but they can't drive back to you.

The paper also connects this to the "index" of the system, a number that measures how much "bigger" the whole system is compared to the part. They show that if you want to steer in both directions using different states, the system's index must be a finite number. But if you want to steer in both directions using the same state, the system must be perfectly symmetric (Haag duality must hold).

In short, this paper doesn't just say "quantum steering is hard in infinite systems." It gives a precise, mathematical "if and only if" condition. It says: "You can steer this infinite system if and only if you have this specific mathematical map." This connects the physical act of measuring and influencing quantum particles to the abstract world of operator algebras, giving physicists a new tool to understand the deep structure of the universe, from the surface codes of quantum computers to the vacuum of space itself. The authors are confident in their proofs, having used advanced mathematical techniques to show that these connections are not just coincidences, but fundamental laws of the quantum world.

Technical Summary: Quantum Steering is Equivalent to State-Preserving Conditional Expectations

Problem Statement
In quantum systems with infinitely many degrees of freedom, modeled by commuting von Neumann algebras $\mathcal{A}$ and $\mathcal{B}$ on a joint Hilbert space $\mathcal{H}$ , fundamental results from finite-dimensional quantum information theory often fail. A primary example is the uniqueness of purifications: even when $\mathcal{A}$ and $\mathcal{B}$ are tomographically complete (i.e., the algebra generated by them is the full algebra of bounded operators, $\mathcal{A} \vee \mathcal{B} = \mathcal{B}(\mathcal{H})$ ), purifications of a state on $\mathcal{A}$ need not be related by unitaries in $\mathcal{B}$ . This failure occurs precisely when Haag duality fails, meaning the commutant $\mathcal{B}'$ is strictly larger than $\mathcal{A}$ .

The central question addressed is which operational entanglement properties survive in such settings where Haag duality fails. Specifically, the authors investigate the conditions under which one subsystem can "steer" the other. Steering refers to the ability of $\mathcal{B}$ to generate any ensemble decomposition of a marginal state on $\mathcal{A}$ via local measurements on $\mathcal{B}$ . While steering is trivial in finite dimensions with a trivial environment, its behavior in infinite-dimensional systems with non-trivial commutants ( $\mathcal{B}' \supsetneq \mathcal{A}$ ) remains to be characterized.

Methodology
The paper employs a rigorous operator-algebraic approach, bridging quantum information theory with subfactor theory and the theory of Jordan algebras.

Generalization to Jordan Algebras: To handle maps that are merely positive rather than completely positive, the authors work within the category of JBW*-algebras (the Jordan analog of von Neumann algebras). They establish a generalized version of the Kadison-Schwarz inequality (the Jordan-Schwarz inequality) for normal unital positive maps.
Ensemble Lifting: The core technical tool is the "ensemble lifting property." The authors prove that for a state-preserving normal unital positive map $\alpha: (N, \phi) \to (M, \omega)$ , the ability to lift every ensemble on $N$ to an ensemble on $M$ is equivalent to the existence of a state-preserving left inverse (a conditional expectation) for $\alpha$ .
Petz Duality: The proofs utilize Petz dual maps and self-polar forms to characterize the lifting property. By analyzing the multiplicative domains of positive maps, they show that the lifting property holds if and only if the map restricts to a $J^*$ -isomorphism on a specific subalgebra that admits a conditional expectation.
Radon-Nikodym Theorem: The connection between steering and ensemble extensions is formalized using the commutant Radon-Nikodym theorem for states on von Neumann algebras.

Key Contributions and Results

Equivalence of Steering and Conditional Expectations (Theorem B): The primary result establishes that for a pure global state $\Omega$ , the ability of $\mathcal{B}$ to steer any ensemble decomposition of the marginal state on $\mathcal{A}$ is equivalent to the existence of a normal conditional expectation $E: \mathcal{B}' \to \mathcal{A}$ that preserves the state $\omega_{\mathcal{B}'}$ (i.e., $\omega_{\mathcal{B}'} = \omega_{\mathcal{A}} \circ E$ ). This provides a direct operational interpretation of state-preserving conditional expectations, linking quantum steering directly to subfactor theory.
Characterization of Two-Way Steering:
- Single State: If $\mathcal{A}$ and $\mathcal{B}$ can steer each other relative to the same normal state $\omega$ , this occurs if and only if Haag duality holds ( $\mathcal{A} = \mathcal{B}'$ ) and $\omega$ is pure.
- Distinct States: If $\mathcal{A}$ and $\mathcal{B}$ can steer each other relative to possibly different states, this is equivalent to the subfactor inclusion $\mathcal{A} \subset \mathcal{B}'$ having a finite Jones-Kosaki-Longo index ( $[\mathcal{B}' : \mathcal{A}] < \infty$ ).
Mixed State Reduction: For irreducible subfactor inclusions ( $\mathcal{A} \vee \mathcal{B} = \mathcal{B}(\mathcal{H})$ ), steering relative to mixed states reduces to steering relative to pure states. Specifically, steering is possible for a mixed state $\omega$ if and only if there exists a tensor product decomposition of the support of $\omega$ such that the state factorizes into a pure state on one factor and an arbitrary state on the other, and the conditional expectation exists on the pure component.
General Lifting Theorem (Theorem D): The paper proves a general theorem for JBW*-algebras stating that ensemble lifting is equivalent to the existence of a unital completely positive left inverse (conditional expectation) for the map, provided the map is state-preserving. This generalizes previous results that were restricted to completely positive maps or von Neumann algebras.

Significance
The paper claims to establish a "direct connection between quantum steering and subfactor theory." By demonstrating that the operational capability of steering is mathematically equivalent to the existence of a state-preserving conditional expectation, the work clarifies the structural requirements for entanglement properties in infinite-dimensional systems.

The authors highlight that this framework resolves the ambiguity regarding which states are preserved by conditional expectations in specific physical models. For instance, in topologically ordered systems (like the surface code), the conditional expectation preserves the ground state. In contrast, for rational conformal field theories, the vacuum state cannot be preserved by such an expectation due to the Reeh-Schlieder property, implying that steering is not possible relative to the vacuum in those specific setups.

The work also addresses the limitations of tomographic completeness; the authors show that even when subsystems are tomographically complete (no environment), steering is not guaranteed, and one-way steering does not necessarily imply two-way steering. This refines the understanding of entanglement in systems where Haag duality fails, a scenario common in topological phases and conformal field theories.

More like this