⚛️ quantum physics

Dynamical codes for hardware with noisy readouts

This paper optimizes the measurement schedules of dynamically condensed colour codes for hardware with noisy readouts by introducing the "teraquop volume" metric, demonstrating that strategically repeating measurements improves performance under measurement-biased noise while highlighting the critical role of correlated error decoding via belief matching.

Original authors: Peter-Jan H. S. Derks, Alex Townsend-Teague, Jens Eisert, Markus S. Kesselring, Oscar Higgott, Benjamin J. Brown

Published 2026-07-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Peter-Jan H. S. Derks, Alex Townsend-Teague, Jens Eisert, Markus S. Kesselring, Oscar Higgott, Benjamin J. Brown

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 are trying to send a precious, fragile message across a stormy ocean. The message is your "quantum computer," and the storm is "noise" (errors) that constantly tries to scramble your data. To survive, you wrap your message in a protective bubble called a Quantum Error Correction Code.

This paper is about designing the best possible protective bubble for a specific type of quantum computer called a "lattice surgery" machine. The authors are trying to figure out the most efficient way to build this bubble so it uses the least amount of space and time while still keeping the message safe.

Here is a breakdown of their work using simple analogies:

1. The Problem: The Noisy Weather

In the real world, quantum computers are messy. The tools used to check if the message is safe (called "measurements") are often the most broken part of the system.

The Analogy: Imagine you are a lighthouse keeper checking a ship's hull. Your eyes (the measurements) are shaky and often tell you there's a crack when there isn't one, or miss a real crack. The ship itself (the data qubits) is relatively stable, but your shaky eyes are the main source of trouble.
The Goal: The authors wanted to see if they could change how and when the lighthouse keeper checks the ship to handle these shaky eyes better.

2. The Solution: Dynamical Codes (The Flexible Schedule)

Traditionally, error correction codes are like a rigid checklist: "Check the front, then the back, then the left, then the right."

The Innovation: This paper looks at Dynamical Codes. Think of these as a flexible schedule. Instead of a rigid checklist, you can decide to check the front three times in a row, or skip the back for a while, depending on the weather.
The Specific Codes: They tested two main types of schedules:
- The "XYZ" Schedule: Checks three different types of things (X, Y, and Z) in a cycle.
- The "XZ" Schedule: Checks only two types (X and Z), skipping the Y.

3. The Metric: The "Teraquop Volume"

To decide which schedule is best, they needed a scorecard. They invented a metric called the Teraquop Volume.

The Analogy: Imagine you are packing a suitcase for a trip. You have two constraints: how much space the suitcase takes up (number of qubits) and how long the trip takes (number of measurement rounds).
- A "Footprint" only measures how big the suitcase is.
- The "Volume" measures the suitcase size multiplied by the trip duration.
Why it matters: A code might use a tiny suitcase but take a million years to finish the trip. Another might be huge but finish in a second. The "Volume" tells you the true total cost of the trip. The authors found that the time it takes to finish the trip was usually the biggest factor, not the size of the suitcase.

4. The Big Discovery: The Decoder Matters Most

The authors tested two different "interpreters" (decoders) to read the lighthouse keeper's shaky notes:

MWPM (The Simple Matcher): A basic algorithm that just connects the dots in the simplest way.
Belief Matching (The Smart Detective): A sophisticated algorithm that looks at the whole picture, considers probabilities, and uses "hunches" to figure out what really happened.

The Result:

When using the Simple Matcher, the "XZ" schedule (checking fewer things) was better. It was like having a simpler checklist that was less likely to confuse the simple algorithm.
When using the Smart Detective, the "XYZ" schedule (checking everything) became the winner. The Smart Detective could handle the extra information and use it to correct errors much better.
The Twist: In some cases, using the Smart Detective turned the worst-performing code into the best-performing one. It's like giving a novice driver a GPS vs. a professional driver with a GPS; the professional makes the car perform miracles.

5. The "Repeat" Strategy: Checking Twice?

The authors also tested a strategy called repeating measurements. If your eyes are shaky, maybe you should check the same spot twice to be sure?

The Intuition: They thought repeating checks would always help, especially when the "eyes" (measurements) were the main problem.
The Reality: It only helped in one specific scenario: when the measurement errors were extremely dominant (like a blinding fog).
The Surprising Finding: In most other cases (even when measurements were noisy), repeating the checks actually made things worse or didn't help at all.
- Why? By spending time repeating the same check, you leave the ship unguarded for longer periods. It's like the lighthouse keeper staring at the front of the ship for an hour to be sure, while a storm hits the back of the ship unseen. The "time cost" of repeating checks outweighed the benefit of the extra certainty.

6. The Takeaway

Tailor to the Noise: There is no "one size fits all." If your hardware has shaky measurements, you need a different schedule than if your hardware has shaky data.
Use the Right Brain: The choice of decoder (Simple vs. Smart) changes which code is best. A "Smart" decoder can unlock the potential of more complex codes.
Don't Over-Check: Repeating measurements is a trap. It usually wastes time and leaves the system vulnerable to other types of errors, unless the measurement noise is overwhelmingly bad.
Measure the Whole Trip: To truly understand how efficient a quantum computer is, you must look at both the space (qubits) and the time (rounds) together. The "Volume" metric is a better ruler than just looking at the size of the machine.

In short, the paper teaches us that to build a reliable quantum computer, we shouldn't just build a bigger shield; we need to build a smarter, more flexible shield that matches the specific type of storm we are facing, and we need a smart enough interpreter to read the shield's signals.

Technical Summary: Dynamical Codes for Hardware with Noisy Readouts

Problem Statement

Quantum error correction (QEC) is essential for fault-tolerant quantum computing, yet the resource overhead remains a significant bottleneck. While dynamical stabilizer codes, specifically Dynamically Condensed Colour Codes (DCCCs), offer a flexible alternative to static surface codes by defining error correction through a schedule of measurements, their performance under realistic hardware constraints requires further optimization.

A critical gap exists in tailoring these codes to hardware platforms where measurement errors are dominant (i.e., measurements are noisier than unitary or idling operations). This noise model is relevant to superconducting qubit proposals. Previous work has focused on unbiased noise or noise biased toward specific Pauli errors (e.g., $Z$ -bias), but fewer studies have addressed measurement-biased noise. Furthermore, standard performance metrics often focus on spatial footprint (number of qubits) rather than the full spacetime cost, potentially obscuring trade-offs between circuit depth and qubit count.

Methodology

The authors investigate the resource requirements of lattice surgery quantum computers using DCCCs under various noise models and decoding strategies.

1. Code Family and Schedules

The study focuses on a large class of DCCCs defined by their measurement schedules, denoted as $X^a Y^b Z^c$ codes. These codes involve consecutively measuring sets of $XX$, $YY$, and $ZZ$ operators on a honeycomb lattice.

Repetition: The authors introduce the technique of repeating measurements (Schedule-Induced Gauge Fixing), where a specific set of weight-two measurements is repeated $r$ times within the schedule. This allows for the construction of smaller detectors (edge detectors) formed from consecutive identical measurements.
Asymmetry: They explore asymmetric schedules where different operators are repeated different numbers of times to tailor the code to specific noise biases.

2. Noise Models

The simulations cover two primary categories of noise:

Direct Parity Measurement Noise: Assumes native two-qubit parity measurements. Errors include single-qubit Pauli errors ( $p_X, p_Y, p_Z$ ) and measurement errors ( $p_m$ ). The study varies the measurement bias ( $\eta_m = p_m / (p_X+p_Y+p_Z)$ ) and $Z$ -error bias ( $\eta_Z = 2p_Z / (p_X+p_Y)$ ).
Auxiliary Qubit Circuit Noise: Decomposes multi-qubit measurements into elementary gates using auxiliary qubits. Three specific models are simulated:
- Standard Depolarizing (SD).
- Superconducting-Inspired (SI), modeling real hardware error rates.
- Entangling Measurement (EM), where multi-qubit measurements are implemented directly.

3. Decoding Strategies

Two decoders are compared:

Minimum-Weight Perfect Matching (MWPM): Requires decomposing hyperedges (errors violating >2 detectors) into ordinary edges, potentially losing correlation information.
Belief Matching: Uses belief propagation (BP) on the full hypergraph to estimate marginal error probabilities before applying MWPM. This exploits correlations lost during decomposition.

4. Performance Metric: Teraquop Volume

The authors generalize the "teraquop footprint" to the teraquop volume.

Definition: The product of the number of qubits ( $N$ ) and the number of measurement rounds ( $T$ ) required to achieve a logical error rate of $10^{-12}$ .
Rationale: This metric captures spacetime trade-offs. A code with high qubit overhead but low depth might be preferable in time-constrained scenarios, while a low-qubit, high-depth code might suit qubit-constrained scenarios. The volume metric treats these equally, whereas a footprint metric would unfairly penalize one over the other.

Key Contributions

1. Optimization of Measurement Schedules

The paper systematically optimizes the measurement schedule of DCCCs for different noise biases. It demonstrates that the optimal code structure (e.g., $X^a Z^b$ vs. $X^a Y^b Z^c$ ) and the utility of repeating measurements depend heavily on the specific noise model and the decoder used.

2. The Role of Decoders and Correlated Errors

A major finding is the critical importance of using decoders that account for correlated errors (hyperedges).

Under MWPM, codes with fewer measurements per detector (like $X^a Z^b$ ) generally perform better because they generate fewer violated detectors.
Under Belief Matching, codes with more complex schedules (like $X^a Y^b Z^c$ ) often outperform simpler codes. The belief propagation step recovers information about hyperedge errors that MWPM loses during decomposition, effectively turning a "bug" (complex detectors) into a "feature."

3. Impact of Repeating Measurements

The study analyzes the effects of repeating measurements:

Measurement-Biased Noise: Repeating measurements significantly improves performance (reduces teraquop volume) when noise is heavily biased toward measurement errors. This is because repeating measurements creates "edge detectors" that quickly catch measurement errors, increasing the spacelike distance against pure measurement noise.
Unbiased or $Z$ -Biased Noise: Repeating measurements offers little to no improvement. In fact, it can degrade performance against timelike logical errors caused by data qubit errors, as the effective timelike distance decreases.

4. Teraquop Volume Analysis

The analysis reveals that differences in performance between codes are primarily driven by the number of measurement rounds required (temporal overhead) rather than the number of qubits (spatial footprint). This underscores the necessity of using the teraquop volume rather than just the footprint for resource estimation.

Key Results

Decoder Dependence: For direct parity measurement noise, $X^a Y^b Z^c$ codes are the worst performers under MWPM but the best under Belief Matching. Conversely, $X^a Z^b$ codes are superior under MWPM. This trend holds across all auxiliary qubit circuit noise models.
Noise Bias Effects:
- Measurement Bias: Repeating measurements is highly beneficial. For strong measurement bias, the best codes involve repeating measurements.
- $Z$ -Bias: Repeating measurements does not significantly improve performance for $Z$ -biased noise within the simulated range ( $\eta_Z \le 16$ ). The intuition that asymmetric schedules (measuring $X/Y$ more often) would help was not borne out by simulations in this regime.
Auxiliary Qubit Models:
- Under Entangling Measurement (EM) noise, the $X^2 Z^2$ code performs exceptionally well with MWPM, outperforming other DCCCs.
- Under Superconducting-Inspired (SI) noise, repeating measurements showed negligible benefit, contrary to the expectation that the moderate measurement bias in SI models would favor repetition.
Metric Sensitivity: The dramatic performance differences between decoders and code types are almost entirely visible in the teraquop height (number of rounds), not the footprint. Using only the footprint would obscure these critical distinctions.

Significance and Claims

The paper claims that dynamical codes offer a practical route to large-scale quantum computation by allowing flexibility in construction to match hardware noise characteristics.

Tailoring to Hardware: The work provides a concrete methodology for optimizing DCCCs for hardware where measurements are the dominant source of noise, a scenario relevant to superconducting qubits.
Decoder Selection: It highlights that the choice of decoder (MWPM vs. Belief Matching) can fundamentally change which code is optimal, emphasizing the importance of utilizing correlated error information.
Resource Estimation: The introduction and application of the teraquop volume is presented as a crucial advancement for resource estimation, arguing that spacetime trade-offs must be evaluated holistically rather than focusing solely on qubit counts.
Modest Scope: The authors are modest about the impact of repeating measurements, noting that while it helps under strong measurement bias, it offers little benefit for unbiased or $Z$ -biased noise in the simulated regimes. They do not claim universal superiority of repeated schedules but rather a conditional benefit dependent on the noise profile.

The work concludes by suggesting that combining tailored schedules with other techniques (like Clifford conjugation or time vortex defects) could further reduce overhead, and calls for experimental demonstrations of these tailored DCCCs.