⚛️ quantum physics

Nonlocality distillation can outperform entanglement distillation

The paper demonstrates that for a small number of shared state copies, nonlocality distillation can achieve a higher CHSH violation and greater resource efficiency than entanglement distillation, even when the latter utilizes communication.

Original authors: Peter Høyer, Jibran Rashid, Razeen ud Din

Published 2026-03-03

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Peter Høyer, Jibran Rashid, Razeen ud Din

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 you have a box of slightly damaged, noisy playing cards. You want to turn them into a perfect, high-quality deck to play a special game where you need to win against the laws of physics (specifically, the game of "CHSH," which tests if your cards are magically connected).

This paper is about two different strategies to fix your noisy cards: Entanglement Distillation and Nonlocality Distillation. The authors, Peter Høyer and his team, discovered that sometimes, the second strategy is actually better, even though it sounds more complicated.

Here is the breakdown using simple analogies:

1. The Two Resources: "The Bond" vs. "The Magic Trick"

In the quantum world, there are two related but different superpowers:

Entanglement: Think of this as a strong, invisible rope tying two people together. If one person moves, the other moves instantly, no matter how far apart they are.
Nonlocality: Think of this as a magic trick where two people guess each other's cards perfectly without talking. It's the proof that the rope exists.

Usually, we think: "If I fix the rope (Entanglement), the magic trick (Nonlocality) will automatically get better." The paper asks: Is that always true?

2. The Problem: Too Many Noisy Copies

Imagine you have a pile of 2 or 3 broken cards (copies of a noisy quantum state). You want to make them perfect.

The Old Way (Entanglement Distillation): This is like trying to repair the rope. To do this, Alice and Bob (the players) have to call each other on the phone, compare notes, and perform complex surgery on the cards to weave a perfect rope. It requires a lot of effort and communication.
The New Way (Nonlocality Distillation): This is like trying to perform the magic trick directly. Instead of fixing the rope first, you just try to coordinate your guesses to win the game immediately.

3. The Big Surprise

The authors found a counter-intuitive result:
When you only have a few noisy cards (2 or 3 copies), the "Magic Trick" strategy (Nonlocality Distillation) wins.

The Result: Even though the "Rope Fixing" strategy (Entanglement) is allowed to use the phone to talk, it still produces a worse magic trick score than the strategy that doesn't talk as much.
The Analogy: It's like trying to fix a broken radio. The "Entanglement" team spends hours soldering wires and talking to the factory to get a perfect signal. The "Nonlocality" team just tweaks the antenna and listens. Surprisingly, for a short time, the antenna tweak gives a clearer signal than the full repair job.

4. Why Does This Happen?

The "All-or-Nothing" Trap: The "Rope Fixing" strategy is designed to create a perfect Bell state (a perfect rope). If you don't have enough noisy cards, the repair job often fails completely, leaving you with nothing.
The "Good Enough" Approach: The "Magic Trick" strategy doesn't aim for perfection immediately. It aims to squeeze out the maximum possible "weirdness" (violation of classical physics) from the messy cards you have right now. It's more efficient at harvesting value from a small, messy pile.

5. The Cost of Doing Business (Resources)

The paper also looked at how much "fuel" (computing power, time, and hardware) each method needs.

Entanglement Distillation is like building a massive factory. It requires many more "workers" (qubits), takes much longer, and burns more energy.
Nonlocality Distillation is like a lean startup. It uses fewer resources, runs faster, and is much cheaper to run on current, imperfect quantum computers (called NISQ devices).

6. The Takeaway

For a huge pile of cards: If you have thousands of copies, the "Rope Fixing" (Entanglement) method eventually wins because you can make a perfect rope.
For a small pile of cards (2 or 3): The "Magic Trick" (Nonlocality) method is superior. It gets a better score, requires less communication, and uses less computing power.

In summary: The paper proves that you don't always need to "fix the relationship" (entanglement) to get the best "communication results" (nonlocality). Sometimes, just working with the messy reality you have is more efficient than trying to force a perfect solution. This is a huge deal for building real-world quantum computers, where resources are scarce and we often only have a few noisy qubits to work with.

1. Problem Statement

In quantum information theory, entanglement and nonlocality are distinct but related resources. While entanglement is a necessary condition for nonlocality, not all entangled states exhibit nonlocality (e.g., some admit a local hidden variable model).

Entanglement Distillation: Protocols that convert multiple copies of noisy entangled states into fewer, high-fidelity Bell states. These are well-studied, often requiring classical communication between parties (LOCC).
Nonlocality Distillation: Protocols that aim to maximize the violation of the CHSH inequality (a measure of nonlocality) from noisy states, potentially without generating a pure Bell state.

The Core Question: When only a small number of copies ( $n$ ) of a noisy state are available (insufficient to deterministically generate a perfect Bell state), does the optimal entanglement distillation protocol also yield the optimal nonlocality distillation (maximum CHSH violation)? The authors investigate whether directly distilling nonlocality can outperform the indirect route of distilling entanglement first.

2. Methodology

The authors compare two frameworks for $n$ copies of a noisy bipartite state:

A. Entanglement Distillation (ED)

Protocol: They utilize the optimal one-way communication protocol for pure states established by Lo and Popescu.
Mechanism: The protocol attempts to probabilistically distill a Bell state ( $|\psi\rangle$ ). If successful, the CHSH value is $2\sqrt{2}$ . If it fails, the resulting state is a product state with a CHSH value of 2.
Calculation: The expected CHSH value ( $V_{ED}$ ) is calculated as a weighted average based on the success probability ( $p_{succ}$ ) derived from the Schmidt coefficients of the input state.

B. Nonlocality Distillation (ND)

Protocol: They employ collective measurements on $n$ copies of the state, following bounds derived by Liang and Doherty.
Mechanism: Instead of targeting a Bell state, the protocol optimizes observables ( $A_i, B_j$ ) to maximize the CHSH correlator directly.
Calculation: For pure states, they use the Liang-Doherty bound formula involving Schmidt coefficients. For mixed states, they construct a specific operator $N$ and bound its norm to derive the maximum achievable CHSH value.

C. Resource Estimation

The authors simulate the quantum circuits required for both protocols using the Azure Quantum Resource Estimator.
They analyze Logical Qubits, Circuit Depth, T-States (a proxy for fault-tolerant overhead), and Runtime.
They specifically compare the "Distillation" phase (Phase 2), noting that ED requires complex multi-qubit operations and ancilla management, whereas ND relies on direct measurements.

3. Key Contributions & Theorems

Theorem 1: Pure States

For bipartite pure states, the authors prove that nonlocality distillation achieves a higher CHSH value than optimal entanglement distillation for small numbers of copies:

Case $n=2$ : ND outperforms ED for noise parameter $p \in [0.5, 0.85]$ .
Case $n=3$ : ND outperforms ED for $p \in [0.5, 0.746]$ .
Case $n=4$ : The advantage vanishes; ED becomes superior or equal.

Theorem 2: Mixed States

For mixed states (specifically a mixture of Bell states), the authors prove:

Case $n=2$ : The optimal ND protocol achieves the same CHSH value as the pure state protocol, which is strictly higher than the optimal ED protocol.
Case $n=3$ : They derive a specific upper bound for the CHSH value:
$V \leq 2\sqrt{1 + (1 - 24p^2 + 80p^3 - 120p^4 + 96p^5 - 32p^6)}$
This bound demonstrates that ND can still outperform ED for mixed states with $n=2$ .

4. Results

Performance Comparison (CHSH Violation)

Small $n$ Regime: When the number of copies is small ( $n=2, 3$ ) and the noise is significant, Nonlocality Distillation yields a higher CHSH violation than Entanglement Distillation.
Large $n$ Regime: As $n$ increases (e.g., $n=4$ ), the advantage of ND disappears, and ED becomes the superior strategy because it can deterministically generate perfect Bell states.
Communication: Crucially, the authors note that ED requires classical communication to succeed, yet ND still outperforms it in the low-copy regime. This reinforces that entanglement and nonlocality are distinct resources.

Resource Efficiency

Using the Azure Quantum Resource Estimator for $n=2$ copies:

Nonlocality Distillation: Requires 15 Logical Qubits, 63 Depth, 150 T-States, and 0.00025s runtime.
Entanglement Distillation: Requires 18 Logical Qubits, 512 Depth, 500 T-States, and 0.00225s runtime.
Conclusion: ND is significantly more resource-efficient, requiring roughly 200x less circuit depth and 3x fewer T-States than ED. This is attributed to the fact that ED requires complex ancilla-based operations to "purify" the state, whereas ND performs direct collective measurements.

5. Significance and Implications

Resource Efficiency in NISQ Era: For Near-Term Intermediate Scale Quantum (NISQ) devices with limited coherence times and qubit counts, Nonlocality Distillation offers a more viable path to generating strong quantum correlations than Entanglement Distillation.
Distinctness of Resources: The results provide strong evidence that entanglement and nonlocality are not interchangeable. Maximizing one does not necessarily maximize the other, especially under resource constraints.
Quantum Key Distribution (QKD): The authors propose that ND could serve as a superior reconciliation technique for Device-Independent QKD (DIQKD). Since ND can extract stronger nonlocal correlations from noisy states without the heavy overhead of distilling Bell states, it may enable higher key rates or tolerate higher noise levels in DIQKD protocols.
Sparse State Preparation: The paper highlights the efficiency of sparse state preparation algorithms in reducing the overhead of these distillation protocols.

Summary

The paper demonstrates that for a limited number of noisy quantum state copies, directly distilling nonlocality is superior to distilling entanglement. It achieves higher CHSH violations with significantly lower quantum resource overhead (fewer qubits, gates, and time). This challenges the assumption that entanglement distillation is the universal optimal strategy for generating quantum correlations and suggests new avenues for efficient quantum communication and cryptography.