← All problems

Ehrhart positivity conjecture for Berenstein–Zelevinsky polytopes

For every Berenstein–Zelevinsky polytope PP associated with a semisimple Lie algebra, the Ehrhart quasipolynomial Ehr⁡P(t)=∣tP∩Zd∣\operatorname{Ehr}_P(t)=\lvert tP\cap\mathbb{Z}^d\rvert has nonnegative coefficients: if, for some period qq, Ehr⁡P(t)=∑j=0dcr,jtj\operatorname{Ehr}_P(t)=\sum_{j=0}^{d}c_{r,j}t^j whenever t≡r(modq)t\equiv r\pmod q, then cr,j≥0c_{r,j}\ge 0 for every residue class rr and every jj.

References

Source: arXiv

Progress summary

Refreshed
Open · solution reported

A 2026 preprint claims a concrete counterexample in one exceptional case, so the conjecture fails there while related positivity questions remain open.

The conjecture extends Ehrhart-coefficient positivity from classical root systems to Berenstein–Zelevinsky polytopes. Berenstein and Zelevinsky introduced the polyhedral model in work culminating in 2001; later work proposed positivity for classical types.

Known results

  • Berenstein–Zelevinsky introduced the polyhedral model, with the foundational work culminating in 2001.
  • Derksen and Weyman proved polynomiality of stretched Clebsch–Gordan coefficients in type AnA_n.
  • King, Tollu, and Toumazet conjectured nonnegative coefficients in type AnA_n; Magnuson, McAllister, and De Loera extended the conjectural framework to types BnB_n, CnC_n, and DnD_n.

October 2026 claimed type-G2G_2 counterexample

Magnuson and McAllister give explicit type-G2G_2 inequalities, prove that vertex denominators lie in {1,2,3,4,5,6,9}\{1,2,3,4,5,6,9\} and hence divide 180180, and exhibit sharpness. For λ=(1,0)\lambda=(1,0) and μ=ν=(0,1)\mu=\nu=(0,1), the residue class t≡1(mod6)t \equiv 1 \pmod 6 has stretched multiplicity 16t−16\frac{1}{6}t-\frac{1}{6}, whose constant coefficient is negative; this claims failure of positivity for G2G_2.

Current status (as of November 2026): The conjecture is claimed refuted for type G2G_2, while positivity for classical types and other root systems remains open; the counterexample is not independently verified.

Sources

Solutions 0

No solutions have been posted yet.