Answers are grounded in this paper and cite their sources.
The chiral charge of the overlap Dirac operator is an integer, and on a torus with a unit background flux it equals +1 by construction. What homotopy invariance does not constrain is the following question: when the link phases are random, what fraction of realisations still report +1? We answer it on a two-dimensional torus at L ≤ 20. For gauge-noise amplitudes ε ≤ 0.225π, 100 of 100 realisations return Q = +1 exactly; the pinned fraction falls to 0.75 at ε = 0.325π and to 0.49 at ε = 0.35π, and the unpinning onset does not move between L = 8 and L = 20. Realisations that leave the sector reach Q ∈ [−5, +6] with a standard deviation of 2.21 at ε = 0.5π, yet Q is never non-integer. Interpolating continuously between a pinned and an unpinned realisation, Q changes only where min|λ(H_W)| vanishes; three closings occur on one such path, with sampled minima reaching 7.7 × 10⁻⁶ after mesh refinement, against a median gap of 0.12 elsewhere on the same path. In the non-Hermitian direction the global index stays at +1 to θ = 0.45 and drops only at θ = 0.50, where the gap closes, even though H_W is Hermitian to machine precision for every θ. Both the naive and the biorthogonal single-vector readouts collapse in a single step at θ ≈ 0.17, where cond(V) = 59, and the traced state sits Δ = 0.0022 away from a chirality-zero partner. The failure is therefore in the selection of the state, not in the inner product, and the threshold is set by the partner spacing rather than by the gap or by conditioning. The two failures are separated in θ: the single-mode projection collapses between θ = 0.15 and 0.20, where the global index is still +1 and the gap of H_W remains open, while the index itself moves only at θ = 0.50, where the gap closes. The projection fails first, at a scale set by the separation between the traced state and its nearest chiral neighbour, not by the gap.
No takes yet. Share an insight, caveat, or question.
GenMin Li (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: