The extrinsic information transfer (EXIT) and generalized EXIT (GEXIT) area theorems provide fundamental conservation laws that connect the rate of classical codes to the performance of maximum-a-posteriori decoding on binary memoryless channels. In this work, we generalize these area theorems to quantum error correction. We first develop a coset-entropy formalism for nested linear codes, where the message specifies a coset and encoder randomization selects a codeword within that coset. The area under the resulting coset (G)EXIT function equals the normalized dimension of the quotient. For stabilizer codes, the measured syndrome restricts the Pauli error but does not always determine its logical effect, and the remaining uncertainty is an entropy over logical correction classes. We show that this entropy equals a classical coset entropy associated with the stabilizer group and the Pauli operators that commute with it. The resulting quantum GEXIT function has area equal to the number of encoded qubits per channel use for any stabilizer code, and this applies to the depolarizing channel through its joint four-valued Pauli-error distribution. For CSS codes on the quantum erasure channel and under independent bit- and phase-flip noise, the logical correction problem separates into two binary coset-decoding problems. On the quantum erasure channel, the duality between the stabilizer group and the Pauli operators that commute with it gives a folding identity at complementary erasure probabilities. Using this and the area theorem, we derive an upper bound on the rate of code families with vanishing logical error probability that matches the capacity of the quantum erasure channel. For the depolarizing channel, we also derive a rate-dependent converse, using a connection between correction-class entropy and coherent information to establish an inequality involving the complementary channel.
Comments
This work will be presented in part at ITW 2026 in Tempe, Arizona
The extrinsic information transfer (EXIT) and generalized EXIT (GEXIT) area theorems provide fundamental conservation laws that connect the rate of classical codes to the performance of maximum-a-posteriori decoding on binary memoryless channels. In this work, we generalize these area theorems to quantum error correction. We first develop a coset-entropy formalism for nested linear codes, where the message specifies a coset and encoder randomization selects a codeword within that coset. The area under the resulting coset (G)EXIT function equals the normalized dimension of the quotient. For stabilizer codes, the measured syndrome restricts the Pauli error but does not always determine its logical effect, and the remaining uncertainty is an entropy over logical correction classes. We show that this entropy equals a classical coset entropy associated with the stabilizer group and the Pauli operators that commute with it. The resulting quantum GEXIT function has area equal to the number of encoded qubits per channel use for any stabilizer code, and this applies to the depolarizing channel through its joint four-valued Pauli-error distribution. For CSS codes on the quantum erasure channel and under independent bit- and phase-flip noise, the logical correction problem separates into two binary coset-decoding problems. On the quantum erasure channel, the duality between the stabilizer group and the Pauli operators that commute with it gives a folding identity at complementary erasure probabilities. Using this and the area theorem, we derive an upper bound on the rate of code families with vanishing logical error probability that matches the capacity of the quantum erasure channel. For the depolarizing channel, we also derive a rate-dependent converse, using a connection between correction-class entropy and coherent information to establish an inequality involving the complementary channel.
Quantum error correction protects logical qubits from decoherence but requires repeated syndrome extraction whose fundamental thermodynamic cost has not been characterized exactly. We prove five theorems and two corollaries that supply this characterization for stabilizer codes. First, a no-go result establishes that every error-correction cycle requires strictly positive work, ruling out passive error correction. Second, a duality identity shows that the combined measurement and erasure cost per syndrome-extraction round is determined solely by the number of stabilizer generators; only the split between measurement and erasure depends on the error rate. Third, a coherence dissipation theorem quantifies the additional irreversible work generated when projective stabilizer measurements destroy quantum coherence in the data register. CSS codes eliminate this cost for generators aligned with the noise channel’s eigenbasis, yielding a thermodynamic criterion for code selection under non-Pauli noise. Fourth, a Crooks fluctuation theorem for the full syndrome cycle yields a Jarzynski equality and exponential tail bounds on rare work fluctuations. Fifth, a variance formula captures the effect of syndrome correlations among stabilizers that share physical qubits. We evaluate all results across eleven codes, from a three-qubit repetition code to a 144-qubit quantum low-density parity-check code, establishing a Pareto trade-off between thermodynamic cost and error-suppression capability. Comparison with four experimental platforms—two superconducting, one trapped-ion, and one neutral-atom—shows that current hardware operates five to eleven orders of magnitude above the Landauer limit, with control and readout hardware, not syndrome extraction, as the dominant source of dissipation.
Rényi coherent information, a computable proxy for the von Neumann coherent information, is widely used to study mixed-state phases of matter and decodability transitions in noisy quantum error-correcting codes. However, being a difference of two Rényi entropies, it need not be monotonic in the Rényi index, and lacks the operational meaning of its von Neumann counterpart. Here we address both issues for stabilizer codes. First, for Pauli noise generated by independent Bernoulli events, we prove that the Rényi-n coherent information is nondecreasing in n∈Z+. This follows from a general theorem: if independent random bits are mapped linearly to a fine label T and a coarse label C, then the Rényi entropy difference Hn(C)−Hn(T) is nondecreasing in n. For stabilizer codes, T is the joint syndrome--logical class and C is the syndrome, and the difference is the Rényi-n coherent information up to a constant. The same theorem covers classical linear codes and independent detector error models. Second, for arbitrary stochastic Pauli noise, we give the Rényi-n coherent information an operational meaning via postselection on matching syndromes between one data block and n−1 auxiliary blocks. We determine when this defines a quantum channel and show that saturation of the Rényi-n coherent information is equivalent to asymptotically perfect recovery of the postselected channel. Moreover, the Rényi-n coherent information also upper-bounds the ordinary coherent information achievable after any syndrome-conditioned recovery.
Quantum error correction draws many of its principles and constructions from classical coding theory, adapted to the unique aspects of quantum mechanics. Here we extend this correspondence to a block of logical qubits, each prepared in the same quantum code, by coupling the block through a classical error correcting code: syndrome qubits then carry error information from several logical qubits at once, and collective inference recovers the errors. Algebraically, the construction defines subgroups of the product stabilizer group corresponding to duals of classical codes, and we prove that a lookup table decoder corrects every error pattern respecting the correction radii of the two constituent codes. For classical algebraic code families, the number of syndrome qubits serving L logical qubits scales as O(log2(L+1)) at the code-capacity level, though at the price of stabilizer weight growing with the block length. An entropy bound shows that check weights need not grow with L. Under a phenomenological noise model, the same lookup table identifies measurement errors by nearest-neighbor post-processing, and classical algebraic decoders locate exactly the logical qubits carrying errors even with noisy syndromes. More broadly, we argue that quantum error correction, reduced to its functional core, is source compression in the sense of Shannon, whose source and channel coding theorems bound the overhead rates of quantum post-selection tasks.
As quantum computers remain susceptible to noise, QEC is essential for preserving logical information during computations. However, the performance of QEC codes breaks down beyond certain noise thresholds, revealing fundamental limits on their ability to protect quantum information. These limits can be characterized using information-theoretic measures such as the coherent information, which quantifies the maximum rate at which logical information can be reliably transmitted through a noisy quantum channel. In this work, we establish a direct connection between the CI and the binary cross-entropy loss used when training neural network decoders. Specifically, we show that the CI constitutes a sharp lower bound on the achievable loss for decoders that track logical operators across noisy channels. To this end, we develop a transformer-based neural network model based on maximum likelihood decoding. We train this network to estimate the CI and evaluate its performance on the surface code under three noise models: code capacity, phenomenological, and circuit-level noise. Our results demonstrate that the network accurately predicts CI and yields threshold estimates that closely match known theoretical limits. When used as a decoder, the network significantly outperforms the minimum weight perfect matching decoder in terms of logical error rate. We also introduce a novel soft post-selection scheme that independently treats uncertainty in both logical operators and relies on confidence-based filtering of the network's output. We prove that such post-selection strategies, based on the MLD cosets, are optimal, and demonstrate their scalability in terms of both logical error rate and abort probability. These findings establish transformer-based architectures as powerful tools for QEC and provide the first numerical evidence supporting the optimality and scalability of MLD-based post-selection.
Quantum error correction (QEC) codes protect quantum information from errors due to decoherence. Many of them also serve as prototypical models for exotic topological quantum matters. Investigating the behavior of the QEC codes under decoherence sheds light on not only the codes' robustness against errors but also new out-of-equilibrium quantum phases driven by decoherence. The phase transitions, including the error threshold, of the decohered QEC codes can be probed by the systems' R\'enyi entropies SR with different R\'enyi indices R. In this paper, we study the general construction of the statistical models that characterize the R\'enyi entropies of QEC codes decohered by Pauli noise. We show that these statistical models can be organized into a ``tapestry'' woven by rich duality relations among them. For Calderbank-Shor-Steane (CSS) codes with bit-flip and phase-flip errors, we show that each R\'enyi entropy is captured by a pair of dual statistical models with randomness. For $R=2,3,\ensuremath{\infty}$, there are additional dualities that map between the two error types, relating the critical bit-flip and phase-flip error rates of the decoherence-induced phase transitions in the CSS codes. For CSS codes with an ``em symmetry'' between the X-type and the Z-type stabilizers, the dualities with $R=2,3,\ensuremath{\infty}$ become self-dualities with super-universal self-dual error rates. These self-dualities strongly constrain the phase transitions of the code signaled by ${S}_{R=2,3,\ensuremath{\infty}}$. For general stabilizer codes decohered by generic Pauli noise, we also construct the statistical models that characterize the systems' entropies and obtain general duality relations between Pauli noise with different error rates.
We consider the probability of decoding error, equivalently the fidelity, of finite length quantum codes assuming quantum information transmission via qubit erasure channel. First, we obtain an analytical expression on this probability via combinatorial invariants of quantum codes. Next, we derive a generalization of the MacWilliams identities for those invariants and use them to formulate a linear programming optimization problem, which allows us to obtain lower bounds on the number of unrecoverable sets of i erasures and the probability of decoding error. Our example shows that the obtain bounds can be very tight. Next, we derive upper (achievability) bounds on the probability of decoding error for finite length stabilizer and CSS codes. CSS codes have many advantages for practical applications. Our results show that CSS codes of small lengths, like 100 qubits, visibly lose in performance to generic stabilizer codes.