How does center ordering govern decoding in non-abelian mixed-state symmetry breaking?

Symmetry breaking recast as a decoding task

The paper treats strong-to-weak spontaneous symmetry breaking (sw-SSB) of a non-abelian group G as an operational problem: distinguish two strongly symmetric mixed states that differ by a pair of inserted irreps, after both have been corrupted by strongly symmetric decoherence. Distinguishability is measured by the Uhlmann fidelity between the two states, which behaves like a two-point correlation function of the charge insertions. When that fidelity correlation stops decaying at long distance, the two states can no longer be sharply distinguished, which is the signature of sw-SSB.

For finite G, the same task is dual to distinguishing logical states of the quantum double D(G) under pure-charge noise. That duality is what lets the authors build optimal decoders and read their failure as a phase transition in a statistical mechanics model, rather than as an abstract statement about density matrices.

Non-abelian fusion breaks the classical shortcut

Earlier sw-SSB results relied on abelian symmetries, where decoherence injects neighboring charge pairs that leave the density matrix diagonal in the local charge basis. The problem then collapses to classical stochastic dynamics, and the Uhlmann fidelity reduces to a classical fidelity between probability distributions. That shortcut fails here. Quantum channels with strong non-abelian symmetry necessarily create entanglement: pairs of multi-dimensional irreps must be created in an entangled state so that they fuse to the trivial representation, as when a strongly SU(2)-symmetric channel injects neighboring spin-1/2 pairs in singlet Bell states.

The resulting density matrix is only block-diagonal in the local irrep basis, so the quantum fidelity between two different initial symmetric states generally cannot be computed exactly. The authors instead construct measurement protocols tailored to non-abelian symmetry and express the syndrome probabilities as partition functions of disordered statistical mechanics models.

Local irreps leave the global fusion ambiguous

For abelian groups every irrep is one-dimensional and fusion is deterministic, so the constraint that all irreps fuse to the vacuum reduces to a single additive condition. All on-site irreps connect directly to one global constraint, with no intermediate degrees of freedom. For non-abelian groups such as S3 or SU(2), the singlet projection has non-trivial matrix structure, and the syndrome probability involves a sum over all consistent binary fusion trees: on-site irreps are fused into intermediate irreps, then into coarser ones, terminating at the trivial irrep at the root.

Each internal edge carries an irrep that is not uniquely determined by the on-site data. This ambiguity is the hallmark of non-abelian fusion and the basis of the measurement hierarchy. It also means the syndrome weight is not symmetric: it breaks the strong G×G symmetry down to G_diag × Z(G), where Z(G) is the center of G and G_diag is the weak symmetry that the channel cannot break.

Only the center orders at nonzero error rate

Because the non-abelian syndrome weights act as local symmetry-breaking fields, a decoder restricted to local syndrome measurements has a vanishing threshold for sw-SSB of the full non-abelian symmetry. Yet such a decoder can still see a transition, and that transition involves spontaneous symmetry breaking of the center Z(G). The mechanism is a hiding effect: for any non-abelian group, any irrep of dimension greater than one can hide another one, so at any finite error rate errors can hide the irreps that are trivially charged under the center.

The authors state that this conclusion holds at any finite level of the hierarchy for non-nilpotent groups. The strong symmetry therefore breaks down to Z(G) times the weak symmetry, and the only transition at nonzero error rate is an ordering of the center.

Climbing the fusion tree can only raise the threshold

A natural way to improve on local syndromes is to also measure the fusion of pairs of syndromes, and more generally to measure fusion outcomes on ascending levels of a fusion tree. The authors generalize the partition function to decoders based on measurements reaching up to level N of that tree. These remain disordered statistical mechanics models, but their disorder distributions become increasingly non-local with N.

The data-processing inequality ensures that using measurements from higher levels can only decrease the fidelity correlations, so the observed decoding thresholds can only increase with N. The corresponding distinguishability thresholds form a non-decreasing sequence of lower bounds on the intrinsic sw-SSB threshold. The authors also note that even at finite level the generalized syndrome weight still breaks the strong symmetry down to G × Z(G); restoring the full G×G symmetry would require grouping syndromes into clusters that each fuse to the identity or a one-dimensional irrep.

S3 and SU(2) test two different centers

The two worked examples differ precisely in their centers. S3 has a trivial center, so the framework predicts no transition at any nonzero error rate, and the fidelity correlations of all irreps provably decay to non-zero values. For simple decoherence channels the authors obtain lower bounds on the long-distance saturation values of these correlations, based on the probabilities that insertion of each irrep is locally masked by the noise.

SU(2) has center Z2, so the prediction is richer. Integer-spin irreps have trivial central character and can show long-range fidelity correlations for arbitrarily weak noise, while spin-1/2 fidelity correlations should become long ranged only at a Z2 ordering transition with a non-vanishing threshold. The Hohenberg–Mermin–Wagner question does not arise here because the local syndromes effectively break the continuous strong symmetry down to the discrete center.

Tensor-network results confirm the center picture

The authors formulate the partition functions for decoding with N levels of the fusion tree as two-dimensional tensor networks and contract them with the boundary matrix product state method, which provides a scalable way to sample syndrome configurations and compute fidelity correlations. For S3 the model shows no transition, and the entanglement entropy stays close to zero for all error rates and does not grow with system size, consistent with the absence of a symmetry-breaking transition.

For SU(2) the model exhibits a Z2 transition consistent with the Nishimori-Ising universality class, where the fidelity correlations of spin-1/2 irreps become long-range. The threshold rises from p_c^(1) ≈ 0.174 at level 1, based on local syndromes, to p_c ≈ 0.202 at level 2 of the fusion tree, an increase of 16%. The critical scaling of the entanglement entropy has a slope consistent with the Nishimori-Ising central charge.

The hierarchy bounds the threshold but may not reach it

The central open question is whether the extra information from fusion measurements at higher levels provides tighter lower bounds on p_c, and in particular whether the sequence p_c^(n) saturates below the intrinsic transition or reaches it when the complete fusion tree is measured. For SU(2) and other continuous non-abelian groups, even the existence of a finite threshold remains an open question.

A related issue is how the fusion tree becomes populated as the error rate increases. The percolation argument relates nontrivial fusion channels at large scales to large physical error clusters, but the height of an occupied fusion channel depends on the choice of tree: a fixed partition into spatial regions can split a nearby entangled pair at successive levels, leaving nontrivial irreps high in the tree even at small error rates. Nontrivial irreps high in a fixed tree therefore do not by themselves signal large error clusters, and the relevant construction should optimize the tree around the locations of measured nontrivial local irreps.

Read the threshold as a center-ordering threshold

The decision-relevant point is that a non-abelian sw-SSB threshold should be read as a threshold for ordering of Z(G), not for the full group. For S3, whose center is trivial, no finite-error-rate transition is expected at any level, so a search for a threshold there is misdirected. For SU(2), the meaningful threshold is the Z2 transition detected by half-integer spins, and the level-1 value p_c^(1) ≈ 0.174 is a lower bound that improves to p_c ≈ 0.202 once level-2 fusion outcomes are included.

The 16% rise from level 1 to level 2 is the concrete gain the hierarchy delivers in this paper, and it comes with an explicit caveat: the sequence of thresholds is non-decreasing but its limit relative to the intrinsic transition is unresolved, and for continuous non-abelian groups even a finite threshold is not established. The practical reading is that adding fusion-outcome measurements can only help the decoder, while the quantity being protected is the center.

Sources used in this answer