A mathematical preprint argues that measurements made on a torus could contain enough information to reconstruct the full, all-genus structure of a broad class of Abelian topological phases. The paper’s result is not an experimental finding: it is a set of formal constructions and proofs for Abelian Chern–Simons and extended topological quantum field theories.
The central claim is that normalized torus data can recover a finite quadratic module, written (G,q), and that this reconstructed object can determine the corresponding all-genus extended theory up to symmetric monoidal natural isomorphism. In plainer terms, the study says genus-one information can specify how the theory is organized on surfaces of higher genus within the pointed Abelian setting, without selecting one unique K-matrix presentation.
The study also gives a tolerance for imperfect inputs. Its nearest-row decoding procedure recovers the Abelian fusion algebra when errors in the normalized character rows are below 21.96%, under the paper’s stated assumptions about relabeling, projective normalization, a fixed topological basis and the chosen error model.
Turning torus changes into algebra
The paper’s proposed route is called Berry tomography. It identifies the projective Berry holonomy produced by metric deformations with the representation supplied by the real-polarized extended theory. Berry holonomy here refers to the change accumulated by quantum states as the underlying geometric parameters are varied; the identification provides the bridge from torus observations to the algebraic object the paper wants to reconstruct.
The target is a class of Abelian fractional quantum Hall and spin-liquid phases described by even-lattice Chern–Simons theories. The supplied analysis describes the objects under study as theoretical models and illustrative examples, not as participants, laboratory measurements or a collected empirical dataset.
The reconstruction focuses on normalized S and T data from the torus. The paper treats those data as sufficient, within its model class, to recover the finite quadratic module (G,q). That module is then used as the intrinsic description of the topological order, rather than treating a particular lattice or K-matrix presentation as the unique answer.
The broader conclusion depends on the stated classification of extended Abelian theories by finite quadratic modules. Under that classification, the recovered genus-one object fixes the all-genus extended Abelian theory up to the specified notion of isomorphism. The result is therefore a classification statement inside a defined mathematical framework, not a claim that one set of torus data settles every kind of topological order.
A rule for noisy fusion data
The paper does not assume that measured rows are exact. For noisy data, it takes pointwise products of the measured normalized rows and projects each product onto the nearest row in the finite measured set. This nearest-row step is the proposed way to decode the fusion product—the algebraic rule describing how the theory’s sectors combine.
The formal guarantee is uniform: if the normalized character-row error is less than 21.96%, the decoding recovers the Abelian fusion algebra independently of the number of anyons. The threshold applies after the data have been projectively normalized and a topological basis has been fixed; the paper does not present it as a bound on raw, unprocessed S-matrix entries.
The same error analysis supplies consistency diagnostics for noisy reconstructions. If normalized rows and reduced twists are δ-close to exact pointed data below the stated threshold, the paper bounds the S-consistency defect by ΔS ≤ 3δ + δ² and the T-consistency defect by ΔT ≤ 4δ + 3δ² + δ³. These expressions quantify how far the reconstructed data can depart from the exact algebraic relations under the paper’s assumptions.
The argument uses algebraic lemmas, formal theorems and triangle-inequality bounds rather than conventional statistical estimation or hypothesis testing. The supplied document also includes a dedicated supplement proofs section for the formal arguments.
Examples show the reconstruction in two forms
The paper illustrates the procedure with a cyclic example that reconstructs the group Z3 and its quadratic refinement from normalized rows. The example is meant to show how the algebraic recovery works in a simple case; it is not an independent experimental validation of the method.
A second illustration moves beyond a cyclic group. It recovers the non-cyclic fusion group Z2 ⊕ Z4, showing that the proposed reconstruction is not limited to examples built from one cyclic factor. This, too, is an algebraic example within the theoretical model class rather than a test on measured data.
The paper also links the reconstructed data to the signature class through the Gauss–Milgram relation, according to the verified analysis. That result is presented as part of the theoretical structure recovered from the torus information.
What remains outside the result
The strongest boundary is the scope of the classification. The completeness result is restricted to pointed Abelian order. The paper states that general non-Abelian categories need not be determined by genus-one modular data, so the argument does not establish an all-genus reconstruction theorem for non-Abelian topological orders.
The robustness claim is also conditional. The error model starts after projective normalization, assumes that a topological basis has already been fixed, and does not directly resolve ambiguity in that basis. The supplied analysis says no empirical or experimental validation is reported, so the 21.96% figure remains a formal guarantee rather than a demonstrated performance level in a finite-size or laboratory dataset.
The reconstruction does not produce every possible microscopic description. In particular, it does not recover a unique K-matrix or lattice presentation, and the projective modular representation does not independently reconstruct invertible E8 stacking data. The paper also does not assert that every microscopic Hamiltonian in an Abelian phase is adiabatically connected to the fixed-point theory.
Those limits leave several next tests open in the supplied analysis: applying the decoding procedure to finite-size simulations and experimental Berry matrices while accounting for normalization and basis uncertainty; clarifying its use for microscopic Hamiltonians not shown to connect adiabatically to the fixed point; extending the completeness question to non-Abelian orders; and determining whether additional information can recover E8 stacking data beyond the projective modular representation.
A formal result, not a clinical or laboratory study
The document is an arXiv preprint, labeled arXiv:2608.20330v1 and dated 20 Aug 2026. It reports no data created or analyzed, and the supplied record does not report funding. Its acknowledgments thank Xiao-Gang Wen for encouragement.
Taken on its own terms, the work offers an algorithmic route from normalized torus Berry data to an Abelian anyon theory, with a formal stability result for fusion recovery. Its conclusions should be read as conditional mathematical guarantees for the specified pointed Abelian Chern–Simons and extended-TQFT setting, not as evidence from humans, animals or another biological population.
Paper data and sources
Original title: Torus Berry Data Determine All-Genus Abelian Topological Orders
Authors: Daniel Galviz
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text