Answers are grounded in this paper and cite their sources.
A structural analysis of the ABJ chiral anomaly in real (Minkowski) time within the Henstock–Kurzweil (HK) framework. The entire α-dependence of the truncated fermionic partition function is shown to be localised in a single factor exp(2iα tr_M γ₅), which is exactly 1 at every finite mode level; the universal regulator factor arising when the truncation is removed equals 1 for every admissible profile. The geometric coefficient c_d(A) (the Seeley–DeWitt coefficient) enters through an explicit hypothesis and is not derived. The construction is carried out with no Wick rotation, as a Berezin integral over the truncated Grassmann algebra of the modes of a finite mode lattice. Four statements are proved in full and machine-checked in Lean 4 without `sorry`, without `admit` and without new axioms: (A) the truncated real-time Berezin partition function is well defined, equals the determinant of the truncated Dirac operator up to the real-time phase iN_M, is holomorphic in the retained modes of the background field and obeys an explicit modulus bound; (B) the chiral transformation acts on the truncated Grassmann algebra as an invertible linear map whose Jacobian is exp(2iα tr_M γ₅), and tr_M γ₅ = 0 exactly at every finite truncation level, so the Jacobian is exactly 1; (C) consequently the classical chiral Ward identity is exact at every finite truncation level; (D) the anomaly appears only when the truncation is removed after a regulator has been inserted into the mode sum: the regularised difference of two shifted mode sums has a nonzero limit, and the coefficient extracted from it is ∫₀^∞ (−f′(s)) ds = f(0) = 1 in d = 2 and ∫₀^∞ s f″(s) ds = f(0) = 1 in d = 4; (E) this coefficient is the same for every admissible regulator profile, which is the Adler–Bardeen statement in the present framework. The geometric (Dirac-trace) prefactors q²/2π and q²/16π² enter through an explicitly displayed hypothesis and are not derived here; the topological interpretation (F) is stated as a discussion, and only the uniqueness of the topological charge is proved. A "honest report" section records exactly where each proof stops. Lean 4 formalisation: eight modules, 42 registered declarations, all depending only on [propext, Classical.choice, Quot.sound]; no sorry, no admit, no new axioms. Numerical verification script is deterministic (dense linear algebra and Gauss–Legendre quadrature, no Monte Carlo).
No takes yet. Share an insight, caveat, or question.
Yuri N. Berdinsky (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: