Stability of Quantum LDPC States under Local Pauli Noise

arXiv:2610.22357 · quant-ph, cond-mat.dis-nn, cond-mat.stat-mech · Submitted 2026-10-15 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Stability of Quantum LDPC States under Local Pauli Noise".

Mira: The gist The exponential decay of noise-induced conditional mutual information (CMI) across a separating buffer in quantum low-density parity-check (qLDPC) code states provides a general criterion for mixed-state phase stability…

Kai: First, who's behind it and why it matters.

Title and authors: Kai: So we’re looking at this paper now, "Stability of Quantum LDPC States under Local Pauli Noise." It tackles the stability of these quantum low-density parity-check codes when they're hit by local noise.

Mira: It’s a big topic because these qLDPC codes are central to both quantum error correction and condensed matter physics, so understanding how stable their mixed states are is really important for identifying phases of matter that aren't just pure states.

Lev: For me, the main thing we need to focus on is how this relates to what we can actually build. We’re not just talking about theory here; we need to know if these stability criteria translate into something practical for running real hardware.

Kai: Exactly. So what does this paper actually show us about that stability? It suggests that the spatial decay of conditional mutual information, CMI, is a key diagnostic for whether the mixed state is stable under local Pauli noise.

Mira: And it shows that this CMI decays exponentially with the width of a separating buffer B in two different noise regimes, and importantly, this decay rate doesn't depend on how big the whole system gets.

Lev: That independence from system size is huge for us because scaling up hardware makes tracking these correlations much harder if they aren't decaying properly.

Kai: It sets up two complementary noise regimes where this exponential decay holds: one at weak noise, and another at strong noise, and each regime has its own mechanism driving the decay.

Mira: At weak noise, it uses the error-cluster bound which relies on how sparse errors are: correlations between A and C require an error cluster to cross the buffer B, which is exponentially unlikely.

Lev: So when we’re in that weak noise setting, we are looking at situations where errors have to travel a long way across the buffer before they can connect two parts of the system and cause trouble.

Kai: Then there's the strong noise regime which uses an information-percolation bound, which relies on the idea that the buffer itself retains very little information.

Mira: In that strong noise setting, correlations between A and C can only persist if they go through a connected chain of informative qudits in the buffer B, and even that probability is exponentially suppressed.

Lev: That sounds like we’re looking at how information leaks or stays trapped inside the buffer structure when it’s heavily corrupted by noise.

Title and authors: Kai: The paper also gives us some specific examples where these bounds apply, for instance, for the Z2 toric code under X dephasing, where the bounds apply for error rates p less than zero point zero two nine and p greater than zero point one four six <ref:2610.22357#pg3,for the Z2 toric code under X dephasing>.

Mira: Those thresholds bracket the Nishimori critical point which is about pc equal to zero point one zero nine, which is a really interesting spot between those two regimes we’re talking about <ref:2610.22357#pg3>.

Lev: That critical point tells us where the physics fundamentally changes its behavior regarding this information decay, and it’s important for mapping out our noise landscape.

Kai: Beyond just the Z2 toric code, the paper extends these stability results to topological invariants like Levin–Wen topological entropy, showing they are robust under noise.

Mira: That robustness is shown through a spatial domain-wall construction which controls every local move, and it shows that changes in the Levin–Wen CMI are bounded by an exponentially small correction related to the buffer width r.

Lev: So if we’re trying to monitor topological order in hardware, this means we can expect very little change unless the buffer gets incredibly wide.

Kai: And on a more practical side, that exponential decay in the weak noise regime implies that there exists a constant number of layers of quasi-local Petz recovery maps that can approximately reverse the noise.

Mira: That leads to a result where for a finite lattice, there’s a constant depth, range R equal to theta of log Lambda over epsilon channel DR such that the trace-distance error is less than epsilon.

Lev: That means we can actually implement some sort of parallel correction scheme using these Petz maps on our real hardware setup, which is a lot for us to get right.

Kai: It’s a sophisticated way to handle noise, and it gives us a concrete operational criterion based on this CMI decay instead of just waiting for logical errors to vanish.

Mira: The paper also shows that this stability criterion can be extended to geometrically k-local Pauli noise by using the information-percolation bound.

Lev: That’s important because real noise isn't always perfectly local, so having a framework for geometric locality helps us predict how much we can actually recover.

Kai: The paper also points out some limitations, like a counterexample showing that CMI stability is not guaranteed for arbitrary tripartitions where products of local checks can cancel their support outside ABC.

Mira: That means just having the qLDPC conditions isn't enough for every possible way we partition the system; we need those extra space-uniformity assumptions.

Title and authors: Lev: So, when we’re designing an error correction protocol, we have to be careful about how we structure those checks to make sure they don't fall into those problematic tripartitions.

Kai: Overall, this paper provides a general information-theoretic stability criterion that is more powerful than just looking for vanishing logical errors.

Mira: It gives us these two distinct noise regimes and the corresponding bounds, which lets us select the right analytical tool for diagnosing stability based on how noisy our system is.

Lev: For implementation, it points toward a constant-depth recovery map construction that works on finite lattices, which is a tangible result for error mitigation overhead.

Kai: It also gives us a way to monitor topological order using the Levin–Wen CMI bound, which helps us quantify how stable those invariants are in our physical systems.

Mira: So we’re looking at an information-theoretic stability criterion that links noise structure directly to observable correlation decay across a buffer.

Lev: Before we wrap up this paper, I just want to say that for running this on real hardware, the overhead of setting up those parallel Petz maps is significant, but the potential for robust decoding is there.

Kai: And from my side with the superconducting qubits, it’s helpful because it gives us a way to measure if our state stays stable when we introduce noise in a way that might look fine locally.

Mira: So in summary of this paper, "Stability of Quantum LDPC States under Local Pauli Noise," we get two complementary noise regimes defined by the error-cluster bound and the information-percolation bound, which together give us exponential decay for the CMI across a buffer B.

Lev: It’s an analytic explanation for numerical observations, but be careful, the paper itself notes that these bounds require an additional local-normalizer completeness condition to hold in the strong noise regime.

Kai: And it does suggest that if we look at geometrically k-local noise, we can use the information-percolation bound to get a bound on CMI that depends on ∂BA and ∂BC, which is useful when our physical errors are constrained spatially.

Mira: This paper moves beyond just saying a state is stable or unstable; it gives us a measurable way—the exponential decay rate of the CMI—to diagnose the stability of mixed states under Pauli noise, which is what we need for understanding quantum phases.

Lev: It’s a solid foundation for using information theory to guide our experimental design and error mitigation strategies in the next generation of quantum hardware.

The paper's summary: Kai: So this paper, "Stability of Quantum LDPC States under Local Pauli Noise," it’s basically saying we don't just need to look for logical errors vanishing; we can use the spatial decay of conditional mutual information, CMI, as a way to check if a mixed state is actually stable when local Pauli noise hits it.

Mira: Right. It boils down to this idea that you measure how fast that CMI drops across some buffer B separating parts of the code. The main result is that this drop happens exponentially with the width of that buffer, and it holds true in two different noise scenarios, weak and strong.

Lev: And what I find interesting for hardware is that this exponential decay rate doesn't depend on how big the whole system gets, which makes it a really nice scaling property to have when we think about large-scale systems.

Kai: So the paper breaks it down into two noise regimes. In the weak noise regime, they use something called the Error-Cluster Bound, which hinges on error clusters needing to cross that buffer B in a way that’s exponentially unlikely.

Mira: That makes sense for weak noise; you need a significant error region to bridge A and C across the buffer B, and that probability is suppressed.

Lev: Then there’s the strong noise regime, where they use an Information-Percolation Bound, which says correlations can only survive if they travel along a connected chain of qudits in that buffer B, and even that connection probability is exponentially suppressed.

Kai: That’s what I mean by it being a diagnostic; you pick the right bound based on whether your noise looks more like sparse errors or more like pervasive corruption.

Mira: And for someone just listening to the show, this means we have two different mathematical tools to check stability depending on what kind of noise we're dealing with.

Lev: The paper also touches on how this applies to topological invariants, specifically the Levin–Wen topological entropy, showing that these are pretty robust against noise, with only an exponentially small correction related to the buffer width r.

Kai: And because of that decay in the weak regime, they show there’s a constant number of layers of quasi-local Petz recovery maps you can use to roughly reverse the noise on a finite lattice.

Mira: That leads to a specific result where you get a channel DR with range R equal to theta of log lambda over epsilon, which means we can actually build some parallel correction scheme using these maps for noisy states.

Lev: It’s tangible, I like it. It shows that this isn't just abstract theory; it gives us a path for noise mitigation overhead on real hardware setups.

Kai: And they also show how you can extend this stability criterion to geometrically k-local Pauli noise using the information-percolation bound, which is useful because real errors often have geometric constraints.

Mira: But there’s a caveat too; the paper points out a counterexample showing that CMI stability isn't guaranteed for every arbitrary tripartition you might choose, suggesting we need more assumptions on how we structure those checks.

Lev: So while the decay criterion is powerful, it’s not a universal pass for any layout; you still have to be careful about the specific way you partition your system.

Kai: Overall, this work gives us a way to diagnose stability by measuring CMI decay rather than just waiting for logical errors to disappear, and it points toward practical recovery methods.

Mira: It really shifts the focus from just "is it stable?" to "how fast is the information leaking out?" so we get a more measurable metric for quantum phases.

Lev: It’s a solid foundation for guiding experimental design and error mitigation strategies in the next generation of quantum hardware.

The paper's improvements: Kai: So, if we’re talking about improvements suggested by this paper, it’s really about moving from just checking stability to actually fixing things when noise gets bad enough.

Mira: Right. The main suggestion is this constant-depth recovery map construction that works on finite lattices, which means you can approximate the noiseless state within a trace-distance error epsilon.

Lev: That constant depth, range R equal to theta of log lambda over epsilon thing they mentioned is what makes it runnable for real hardware; it gives us a specific budget for how many layers of parallel maps we need.

Kai: So this isn't just theory about stability; there’s a concrete recipe for noise mitigation that you can actually implement on your superconducting qubits.

Mira: Exactly. And they also show that the Levin–Wen topological entropy is robust, meaning if you are monitoring those properties, you only need to worry about an exponentially small correction related to the buffer width r.

Lev: That spatial domain-wall construction proof they used for that stability actually helps us understand how much noise we can tolerate before we have to start worrying about losing the topological order entirely.

Kai: It’s a bit like having a safety margin on your physical system; you know the theoretical stability holds even with some of that buffer width error built in.

Mira: And they also give us this way to handle geometrically k-local Pauli noise using that information-percolation bound, which is useful when we know our errors are constrained spatially in the chip.

Lev: That geometric locality constraint helps narrow down the analysis, making the resulting bounds more specific to how a physical error might actually manifest on a lattice.

Kai: It’s helpful because it grounds these abstract information bounds into something that matches how noise behaves physically across the hardware structure.

Mira: So, what this paper really changes for us is giving us actionable steps—a recovery map construction and a way to monitor topological properties—instead of just saying the state might be stable or unstable in theory.

Lev: It’s about shifting from a static check to an operational framework where we can actively fight the noise with these parallel operations.

Kai: And it also provides that measurable metric, the CMI decay rate, which lets us quantify exactly how robust our phase of matter is under local noise conditions.

Conclusion: Kai: So to wrap up this discussion on "Stability of Quantum LDPC States under Local Pauli Noise," we’ve seen how they use conditional mutual information decay to diagnose whether a mixed state holds up against local Pauli noise across different noise regimes.

Mira: It really boils down to using the spatial drop in CMI as a measurable way to check stability, and it works even when the noise is strong or weak because of those two different bounds they developed.

Lev: And for me, what this means for error correction is that we get a recovery map construction with constant depth and range R equal to theta of log lambda over epsilon, which is something we can actually design a protocol around.

Kai: So it’s not just theory about stability; it’s a concrete recipe for noise mitigation that you can implement on your superconducting qubits.

Mira: Right. And because they showed how the Levin–Wen topological entropy stays stable with only an exponentially small correction related to the buffer width r, we have a better way to monitor those physical invariants in our systems.

Lev: I think the most important part is that this gives us a measurable metric for robustness based on information theory, which guides our experimental design and error mitigation strategies.

Kai: So we’ve seen how they classify the noise regimes into weak and strong, and how each regime requires a specific tool—the error-cluster bound or the information-percolation bound.

Mira: It's a solid foundation for using information theory to guide our experimental design in the next generation of quantum hardware.

Lev: And it’s definitely something we can use to predict how much overhead those parallel maps will land on real hardware setups.

Department of Physics and Institute for Condensed Matter Theory, University of Illinois Urbana-Champaign · Korea Institute for Advanced Study

quant-ph, cond-mat.dis-nn, cond-mat.stat-mech

Submitted: 2026-10-15

Updated: 2026-10-15

Comments: 29 pages, 5 figures, including appendices

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 80/100

The gist: The gist The exponential decay of noise-induced conditional mutual information (CMI) across a separating buffer in quantum low-density parity-check (qLDPC) code states provides a general criterion

Key concepts

Conditional Mutual Information (CMI): CMI measures the statistical dependency between two sets of variables (A and C) given a third set (B). In this context, it acts as a diagnostic tool; its exponential decay across the buffer indicates whether the quantum state remains stable under noise.
Error-Cluster Bound: This bound applies in weak noise regimes. It states that for correlations between two parts of the code to persist across a buffer, an error cluster must cross it, which is exponentially unlikely. This relies on the sparsity of errors in the system.
Information-Percolation Bound: This bound applies in strong noise regimes. It treats information flow as a percolation process where buffer qudits are independently open with a certain probability. Stability depends on whether informative correlations can persist along rare, connected chains of these open buffer qudits.

Terminology

Conditional Mutual Information (CMI) — 1× in this paper · explained in 3 episodes
Mutual Information — 1× in this paper · explained in 2 episodes
Robustness — 1× in this paper · explained in 5 episodes
Spectral Gap — 1× in this paper · explained in 5 episodes
Trace — 1× in this paper · explained in 2 episodes

Summary

The gist The exponential decay of noise-induced conditional mutual information (CMI) across a separating buffer in quantum low-density parity-check (qLDPC) code states provides a general criterion for mixed-state phase stability under local Pauli noise.

Stability Criterion via CMI Decay

The study investigates the stability of the normalized code-projector state of a Zd qLDPC stabilizer code under local Pauli noise by analyzing the spatial decay of conditional mutual information (CMI) I(A: CB)ρ, which serves as a diagnostic for mixed-state phase stability. The core finding is that this CMI decays exponentially with the width of the buffer B in two complementary noise regimes, both independent of system size.

(a) Weak Noise Regime:

At weak noise, exponential decay holds for full-system tripartitions where correlations between the two sides require an error cluster to cross the buffer, which is an exponentially unlikely event. This regime is governed by the Error-Cluster Bound (Theorem 1), where correlations between A and C require an error cluster to cross the buffer, which is exponentially unlikely. The bound states that I(A: CB)ρ ≤ f(∂BA, ∂BC)e−c0 dist(A,C)B.

(b) Strong Noise Regime:

At strong noise, exponential decay holds whenever the buffer supports no logical operator, since correlations can persist only along rare chains of buffer qudits that remain informative, and this is governed by the Information-Percolation Bound (Theorem 2). This bound relies on the complementary effect that the buffer retains very little information, as correlations between A and C can then persist only through a connected chain of buffer qudits that remain informative, whose probability is again exponentially suppressed.

Formal Bounds and Regimes

The paper establishes two complementary mechanisms governing the decay of CMI under different noise regimes. The weak noise regime is characterized by the Error-Cluster Bound, which uses the sparsity of errors, requiring an error cluster to cross the buffer, which is exponentially unlikely. The strong noise regime relies on the Information-Percolation Bound, which utilizes a percolation process where each buffer qudit is independently open with probability ηi.

(c) Application to Toric Code:

For the Z2 toric code under X-dephasing, the bounds apply for p 0.146 (information-percolation bound), bracketing the Nishimori critical point pc ≃ 0.109.

Robustness of Topological Invariants

The stability results extend to topological invariants, specifically the Levin–Wen topological entropy, which is shown to be robust under noise. Corollary 2.2 shows that the Levin–Wen CMI is stable up to an exponentially small correction in buffer width r, as γLW(ρ) − γLW(ρ0) ≤ Cdef δ(r). This stability is proven using a spatial domain-wall construction that controls every local move, including those crossing the spatial noise boundary.

Local Recovery and Decoding

The exponential decay of CMI in the weak-noise regime implies the existence of a constant number of layers of quasi-local Petz recovery maps that approximately reverse the noise. Corollary 2.3 establishes that for a finite lattice, there exists a constant-depth, range R = Θ(log(Λ/ε)) channel DR such that∥DR(ρ) − ρ0∥1 ≤ ε. This recovery scheme is constructed using a block construction where recovery regions Qs,j are chosen to be pairwise disjoint within each layer, allowing for parallel application of Petz maps.

Generalization to Geometrically Local Noise

The stability criterion can be extended to geometrically k-local Pauli noise using the Information-Percolation Bound (Theorem 2). Corollary 3.1 shows that if the noise is geometrically k-local, there exist system-size-independent constants c0, K > 0 such that I(A: CB)N(ρ0) ≤ Kec0 ∂BA + ∂BC 2e−c0 dist(A,C)B.

Counterexamples and Limitations

The paper provides a counterexample for arbitrary tripartitions of the form in Fig. 1(b), demonstrating that CMI stability is not guaranteed for general partitions where products of local checks can cancel their support outside ABC, leaving stabilizers in S(AB) and S(BC) that overlap on a small region Q ⊆ B. This suggests that qLDPC conditions alone are insufficient for arbitrary tripartitions, necessitating additional assumptions for stability.

Conclusion and Operational Criterion

The exponential decay of noise-induced CMI provides a more general information-theoretic stability criterion than traditional methods like vanishing logical error, with local reversibility as an operational consequence. The decay constant α depends on both stabilizer weight and qudit degree.

Appendix Details

The proof of Theorem 1 utilizes a coupling argument to bound the total variation distance between the conditional distribution and the product of its marginals, which is then related to the probability that a connected component of disagreement patterns intersects both interfaces. The strong noise analysis involves relating local information loss to contraction coefficients ηi, leading to an information-percolation inequality where the CMI is bounded by log d times the expected number of interface generators connected to the opposite side in an auxiliary percolation process. The final bound derived is I(A: C B)ρ ≤ 3D2(log d)s2e−c0t, where t = dist(A, C)B − 1.

Specific Noise Analysis

For the toric code under X-dephasing, the analysis separates the classical record into independent X and Z sectors, showing that only the noisy X sector contributes to the CMI decay. The threshold for error-cluster argument is p pIP ≈ 0.146.

Spatial Deformation Proof

The proof of Corollary 2.2 uses a spatial domain-wall construction to show that the Levin–Wen CMI is stable up to an exponentially small correction in r, by bounding the accumulated error by 2mδ(r). The final result for the Levin–Wen CMI is γLW(ρ; G0) − γLW(ρ0; G0) ≤ 2mδ(r).

Recovery Map Construction

The block construction and reverse-order decoder illustrate the method for achieving the recovery bound, where the total number of blocks is O(nRD), and each Petz map acts on Qs,j = O(R D) = O(logD nε) qudits. The depth of the quasi-local block is at most χ = O(1).

Specific Noise Thresholds

The weak-noise threshold comes from convergence of a weighted path sum, and the strong-noise threshold comes from the exact critical density 1/2 of square-lattice bond percolation. The Nishimori point pc ≃ 0.109 lies between these thresholds.

Final Bound Derivation

The final bound derived in Eq.

Improvements for AI systems

Bold header: Exponential CMI Decay Criterion for Mixed-State Stability

The system can utilize exponential decay of conditional mutual information (CMI) as a general criterion for mixed-state phase stability under local Pauli noise, replacing reliance on vanishing logical errors alone. This allows the AI to rigorously diagnose whether a mixed quantum state remains stable as a phase of matter by measuring the spatial decay of CMI across buffers.

Bold header: Constant-Depth Quasi-Local Petz Recovery

AI systems can implement a constant-depth, range R = Θ(log(Λ/ε)) channel DR that recovers the noiseless state within trace-distance error ε, as established by Corollary 2.3 and Appendix I. This enables robust noise mitigation for quantum circuits by applying parallel layers of quasi-local Petz maps.

Bold header: Robust Topological Entropy Monitoring

The AI can monitor the stability of Levin–Wen topological entropy, utilizing the space-uniformity bound (Corollary 2.2) to ensure that changes in CMI are bounded by an exponentially small correction related to buffer width r, providing a metric for topological order robustness.

Bold header: Noise Regime Classification

The AI can classify the noise regime into two complementary regimes—weak noise (governed by the error-cluster bound, Theorem 1) and strong noise (governed by the information-percolation bound, Theorem 2)—to select the appropriate analytical tool for stability assessment.

Bold header: Geometrically k-Local Noise Handling

For geometrically k-local Pauli noise, the AI can employ Corollary 3.1, which establishes an exponential CMI decay bound (Eq. E7) by incorporating geometric locality constraints into the information-percolation framework.