Practical Entanglement Distillation Protocols with Quadratic Error Suppression

This paper introduces practical, space-optimal entanglement distillation protocols tailored for modular quantum computing that leverage repeated noisy inter-module operations and high-fidelity local gates to achieve quadratic error suppression using only two qubits per module, thereby effectively overcoming noisy links in near-term architectures.

The Gist

Imagine you are trying to send a precious, fragile message between two different rooms in a large, noisy factory.

The Problem: The Noisy Hallway
In the world of quantum computers, these "rooms" are separate chips or refrigerators. Inside each room, the workers (local operations) are very careful, precise, and quiet. They can handle delicate tasks with almost perfect accuracy. However, the hallway connecting the two rooms (the inter-module link) is a disaster zone. It's loud, bumpy, and full of static.

When the workers try to pass a special "entanglement" package (a Bell pair) through this hallway, the package often gets damaged. In the past, scientists had a way to fix these damaged packages, but their methods were like trying to clean a muddy shirt by washing it in a giant, inefficient industrial machine. They required huge amounts of space (many extra workers) and time, and they only cleaned the shirt a little bit.

The Solution: A Smart, Small-Scale Cleanup Crew
The authors of this paper, working at IBM Quantum, have designed a new, highly efficient "cleanup crew" specifically for this noisy hallway.

Think of their new protocol as a two-person detective team stationed in each room.

1. The Setup: Instead of just waiting for a dirty package to arrive and then trying to fix it, these detectives use the noisy hallway itself as a tool. They send a quick, noisy signal back and forth during the cleaning process, not just at the start.
2. The Magic Trick (Quadratic Suppression): Imagine the noise in the hallway is like a 10% chance of a package getting a smudge. Old methods might reduce that smudge chance to 9% or 8% (a linear improvement). This new method is like a magic eraser that reduces the smudge chance from 10% down to 1% (a quadratic improvement). It squares the error rate, making the result exponentially cleaner.
3. Space Efficiency: The most impressive part is how little space they need. Most previous methods required a whole squad of workers (many extra qubits) to do the job. This new protocol works with just two workers per room. It is the smallest possible team that can still catch every type of error.

How They Tested It
The researchers didn't just do this on paper. They took their new "two-person detective" protocol and ran it on real, current IBM quantum computers (specifically the "Pittsburgh" processor).

• They found that the noisy links between chips were indeed the biggest bottleneck.
• When they used their new protocol, the final "packages" (Bell pairs) were significantly cleaner than when they used old methods or no cleaning at all.
• Even though the hallway was still noisy, the new method successfully filtered out the worst of the damage, proving that you don't need a massive machine to fix a small, noisy connection; you just need a smart, compact strategy.

The Bottom Line
This paper introduces a practical, small-scale way to fix noisy connections between different parts of a quantum computer. By using the noisy link itself as part of the solution and requiring only two "workers" (qubits) per side, they can turn a very messy connection into a very clean one. This is a crucial step for building larger quantum computers that are made of many smaller, connected modules.

Technical Summary

Technical Summary: Practical Entanglement Distillation Protocols with Quadratic Error Suppression

Problem Statement
Near-term and early fault-tolerant quantum computing architectures, particularly modular designs, face a critical bottleneck: highly non-uniform error rates. While local operations within a single chip or dilution refrigerator are increasingly reliable, operations connecting different modules (inter-module operations) remain significantly noisier. These inter-module links, often realized via long-range gates or noisy interconnects, can dominate error budgets even when quantum error correction (QEC) is applied. Standard entanglement distillation protocols, typically formulated in a Local Operations and Classical Communication (LOCC) resource model, assume that noisy Bell pairs are generated before the protocol begins and that subsequent processing involves only local operations. This model fails to leverage the specific hardware constraints of modular processors, where the same noisy inter-module gate used to create raw Bell pairs can be reused during the distillation process. Furthermore, existing small-scale protocols often suffer from linear error suppression or require excessive space (qubit count) and time overheads, making them impractical for hardware where resources are constrained.

Methodology
The authors propose a generalized resource model tailored to modular quantum hardware. In this model, two modules possess access to high-fidelity local operations and measurements, alongside the ability to repeatedly use the same noisy inter-module entangling operation during the protocol. The noise is modeled as a stochastic Pauli channel, distinguishing between local errors (assumed negligible for analytical comparison) and inter-module errors (the dominant noise source).

The paper introduces three new protocols designed to minimize space and time overhead while achieving quadratic error suppression (where output error scales as $O(\epsilon^2_{inter})$):

1. Space-Optimal Protocol: Requires only two qubits per module. It utilizes the noisy inter-module gate as an active component within the distillation circuit, rather than solely for initial state preparation. This protocol detects arbitrary single- and two-qubit Pauli errors.
2. Reduced Space-Optimal Protocol: A variant of the above that removes one inter-module gate, achieving quadratic suppression for arbitrary single-qubit errors but only linear suppression for specific two-qubit correlated errors.
3. High-Rate Protocol: A generalization for scenarios with $m$ qubits per module, outputting $m-2$ Bell pairs with quadratic error suppression.

These protocols are compared analytically against existing methods (Recurrence, Maneva-Smolin, and Leung-Shor) based on qubit count, circuit depth (1-D and all-to-all connectivity), duration, and output error rates.

Key Contributions

• Resource Model Shift: The paper shifts the distillation paradigm from a static LOCC model to a dynamic model where noisy inter-module gates are active circuit elements.
• Space-Optimal Quadratic Suppression: The primary contribution is a protocol that achieves quadratic error suppression with the theoretical minimum of two qubits per module (one for the output pair, one for error extraction). This is space-optimal for any protocol detecting inter-module faults.
• Efficiency: The proposed protocols are shown to be among the fastest known methods for quadratic error suppression, significantly reducing time and space overhead compared to existing small-scale protocols.
• Experimental Validation: The protocols were implemented on the ibm_pittsburgh 156-qubit superconducting processor, specifically targeting known noisy links.

Results

• Simulations: Across varying inter-module and local noise strengths, the space-optimal protocol consistently demonstrated the best performance, converging to the lower bounds set by local errors. The reduced protocol showed quadratic suppression only for single-qubit errors, confirming its limitation to specific noise regimes.
• Experiments: On ibm_pittsburgh, the space-optimal protocol consistently yielded the lowest output error rates across all tested noisy links, outperforming single-recurrence, two-round recurrence, and the Leung-Shor protocol.
  • The reduced protocol performed worse than the full space-optimal version, indicating that the physical inter-module errors (CZ gates) contained non-negligible two-qubit components ($\epsilon_2$), not just single-qubit errors.
  • Single-recurrence protocols showed minimal improvement over raw links, as expected due to their linear error suppression.
  • The high-rate protocol demonstrated improved fidelity over the Maneva-Smolin protocol due to quadratic error suppression, though its efficiency was limited by the linear connectivity of the device, which increased duration for larger $m$.

Significance and Claims
The paper claims that inter-module-gate-assisted entanglement distillation is a practical primitive for overcoming noisy links in modular quantum architectures. The significance lies in the ability to trade additional operations and qubits for significantly higher-fidelity entanglement with minimal overhead.

The authors emphasize that their space-optimal protocol is particularly relevant for near-term devices and early fault-tolerant architectures where qubit count and circuit depth are critical constraints. While the advantage is most pronounced when inter-module errors are orders of magnitude larger than local errors, the protocols offer a viable path to improving interconnect fidelities.

The paper maintains a modest scope regarding logical-level applications. It notes that while effective at the physical level, the applicability of these distillation techniques at the logical level depends on the locality of errors induced by inter-module operations. In codes with extensive support (e.g., the Gross code), a single inter-module error could induce correlated errors across multiple logical qubits, potentially rendering distillation ineffective. However, for physical-level interconnects, the technique is presented as broadly applicable and powerful.