A five-vertex negative answer to Johnson and Paparella's Question 9.1 on Hadamard-closed spectral polyhedra
Question 9.1 of Johnson and Paparella, Perron similarities and the nonnegative inverse eigenvalue problem, arXiv:2409.07682v4, Section 9, p. 27, has a negative answer for both polyhedral cones and polytopes. For every integer m >= 1 we construct a rational polygon with m+4 edges whose lift to x_0=1 and conical hull are closed under coordinatewise multiplication and contain the all-ones vector. Every vector in the polygonal lift is the spectrum, in a fixed coordinate order, of a symmetric stochastic matrix with a common eigenbasis. Neither set is the complete spectratope or spectracone of any complex 3-by-3 matrix: the diagonal entries and trace obstruct the required facets away from the identity. The five-vertex case already fails these constraints. Cartesian products give full-dimensional counterexamples in every coordinate size n >= 3. With boundary in the source's complex Euclidean space or its first-coordinate affine hyperplane, Conjecture 9.2 also fails: (1,0,...,0) is a boundary stochastic spectrum but is not radially extremal.
Authors
- Wenlin Zhang (ORCID: https://orcid.org/0009-0005-2414-7968)
- Haobo Ma
Institutions
- National University of Singapore (SG)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24410090
- Primary Topic
- Matrix Theory and Algorithms
- Type
- preprint