Alec Kriebel

A type-D4 counterexample to the canonical-potential exponent converse

A list of linear inequalities can describe the faces of a cone. This note gives an explicit type-D₄ example in which every listed inequality is needed, yet the canonical potential has a coordinate exponent outside the proposed bound. The counterexample concerns a particular historical formulation and its designated canonical coordinates.

Open the original PDF · Download PDF · Source and verification archive · BibTeX · DOI: 10.5281/zenodo.23439776

Earlier computation and the note’s contribution

Denis Mironov’s earlier crystal dataset, Case 267, already contains the same reduced word and its complete string-coordinate support. This note credits that input. It extracts the consequence for the historical canonical-potential assertion, reconstructs the canonical potential and supplies explicit integer facet certificates. It makes no exclusive-priority or current-openness claim.

The historical converse appears in Schumann’s 2021 Oberwolfach contribution. The seed, potential and chamber-map conventions come from Koshevoy and Schumann and Genz, Koshevoy and Schumann. Koshevoy and Mironov’s later computational article cites the earlier repository.

The explicit example

For Spin₈(C), label node 2 as the central node of the D₄ diagram and take the reduced word

i = (1, 2, 3, 2, 4, 2, 1, 2, 3, 2, 4, 2).

In its specified canonical reduced-word cluster-X coordinates, the potential has 22 distinct monomials, each with coefficient one. Their 22 tropical inequalities define distinct facets of a full-dimensional cone in R¹². Nevertheless, two monomials have X₆ exponent −2.

This disproves the necessary direction of the historical canonical-W converse: irredundancy of these inequalities does not force every exponent in those canonical coordinates to have absolute value at most one.

Why the coordinates matter

The chamber-ansatz change to string coordinates preserves the inequality presentation and its facets, but it can change the exponent of an individual coordinate. In this example every exponent after that pullback lies in {−1, 0, 1}.

The later string-coordinate converse is not refuted. Neither is its separate coefficient criterion, nor the conclusion of the cited KS sufficient-criterion theorem. The note identifies a false equivalence of the two coordinate bounds within a printed proof; it does not supply a general replacement proof for that step. Its facet argument is direct and does not use that assertion.

Proof and verification materials

The manuscript displays the canonical potential and integer certificates separating each facet inequality from all the others. The unchanged verification archive includes the manuscript source, certificate data and exact-arithmetic programs. These materials support inspection of this explicit example; they do not establish a general classification or a later conjecture.

Original manuscript abstract

The exponent-bound converse stated for canonical GHKK potentials in Schumann's 2021 Oberwolfach report is false. For an explicit reduced word in type D₄, the canonical potential has 22 combined monomials, all of coefficient one, and all 22 inequalities define distinct facets, although two monomials have exponent −2 in one canonical coordinate. The reduced word and its complete string-coordinate support occur in earlier public computations; we extract the consequence for the historical formulation and supply an independent canonical calculation and exact integer certificates. The chamber-ansatz pullback has all exponents in {−1,0,1}, so the later string-coordinate converse is not refuted. No first-priority or current-openness claim is made.

Original files and citation

The native files are listed in their original record order and retain their original names and bytes.

  1. Source and verification archive — Canonical_W_D4_Verification_v1.0.zip · Zenodo original
  2. Manuscript PDF — Canonical_W_D4_Note_v1.0.pdf · Zenodo original

Kriebel, Alec (2026). A type-D4 counterexample to the canonical-potential exponent converse. Version 1.0. Zenodo. 10.5281/zenodo.23439776. Download BibTeX · Zenodo record.

The original manuscript, certificates and review text are licensed under CC BY 4.0; the original Python verification code is MIT licensed. Third-party sources retain their own rights. The earlier unlicensed dataset and downloaded primary papers are linked as sources and are not redistributed here.

Read the unchanged manuscript

Open PDF · Download PDF · Original PDF on Zenodo

Review status and contributions

AI tools were used extensively in developing, solving, reconstructing, verifying and reviewing the arguments, searching the literature, and preparing the note and code. This work is unrefereed and has not undergone conventional external human peer review. Internal AI reviews are not a formal proof-assistant certificate. Alec Kriebel is responsible for the contents.

Substantial mathematical or expository contributions can support collaboration. Authorship and acknowledgments would reflect the work actually contributed.