Published October 26, 2026 | Version v2
Preprint Restricted

Sign-optimised Viro patchwork data for real loci of Calabi-Yau hypersurfaces in the Kreuzer-Skarke list

Description

CORRECTION NOTICE (added October 2026). The title of this record has been changed. It previously read "Systematic Survey of G2 Betti Numbers from Joyce-Karigiannis Constructions on Kreuzer-Skarke Polytopes"; external indexes may still carry that form. The DOI is unchanged and continues to resolve here.

That former title, and the original description, presented the deposited data as G2 Betti numbers achievable via Joyce-Karigiannis constructions. That presentation is withdrawn. What the data contain is the topology of Viro patchworks on dual polytopes from the Kreuzer-Skarke list, computed by sign optimisation. They are not Betti numbers of G2 manifolds.

The Joyce-Karigiannis theorem (J. Differential Geom. 117 (2021) 255-343, Theorem 1.1) requires, in addition to the data computed here, a compact torsion-free G2 orbifold whose singular locus is associative, together with a closed, coclosed, nowhere-vanishing 1-form on that locus - a condition its authors describe as very strong. That condition is verified for no entry in these tables. The tabulated pairs are combinatorially achievable, not realised.

Two further corrections. (1) The claim of a "unique (b2, b3) = (27, 451) construction" is withdrawn: uniqueness was not established, and the pair is not known to be realised by any G2 manifold. (2) The invariants reported for the flagship entry, n_L = 14, sum b_1 = 18 and chi(L) = -8, are mutually inconsistent. Every closed odd-dimensional manifold has Euler characteristic zero, and fourteen closed orientable three-manifolds with sum b_1 = 18 give chi = 14 - 18 + 18 - 14 = 0, not -8. At least one of the three values is in error.

A third correction, identified in October 2026. The lattice obstruction theorem of the manuscript is stated for elements of order n >= 5, and is false as stated at n = 6, where the character value is 4 cos(2 pi / 6) = 2, an integer; the manuscript's own corollary lists Z_6 among the admissible cyclic factors, contradicting the theorem one page later. The correct hypothesis is n outside {1, 2, 3, 4, 6}, the complement of the crystallographic set. The repair costs nothing downstream: the corollary excluding (b2, b3) = (27, 451) from flat orbifold resolutions needs only the exclusion of Z_5, and n = 5 remains outside that set. The manuscript deposited in this version carries the repair; the April 2026 version does not. The repair does not affect the withdrawal above; the G2 framing of these data remains withdrawn.

The manuscript based on these data has been withdrawn from peer review. Readers should treat the deposited tables as real-locus patchwork data only.

The computational pipeline itself - sign optimisation over dual-polytope triangulations - is sound; a more rigorous treatment of the same objects is given by Renaudineau and Shaw, Ann. Sci. Ec. Norm. Super. 56 (2023) 945-980, and by Matessi and Renaudineau, J. Inst. Math. Jussieu (2025).

Files in this version. The April 2026 manuscript has been removed and replaced by the corrected one; the earlier version of this record retains it, and its DOI continues to resolve there. The three data files - the CSV table, the JSON metadata and the Python source - are carried over unchanged and keep their original checksums.

--- Original description, retained for the record: ---

First systematic computational survey of G₂ Betti numbers 
achievable via Joyce–Karigiannis constructions on 998 toric 
Calabi–Yau threefolds from the complete Kreuzer–Skarke database 
(h¹¹ = 1 to 491). Contains: complete CSV data tables (27,000 
entries), combined JSON metadata, exact sign distribution for 
the unique (b₂, b₃) = (27, 451) construction on CY₃(13, 433), 
and all Python source code for reproducibility.

Files

Restricted

The record is publicly accessible, but files are restricted. Log in to check if you have access.