Paperlane

no. 16 · Tuesday 17 November 2026

Spectra of stochastic matrices with a Toeplitz interior

Brecht Verbeken and Vincent Ginis

math.PR · v2 · 46 pp

Abstract

For stochastic matrices of order $n$, the Karpelevič region is the set of all possible eigenvalues. We study matrices whose first and last $m$ rows are arbitrary probability vectors, while the remaining rows have a Toeplitz pattern. As $m$ increases, their eigenvalue regions expand, first reaching the full Karpelevič region at $m=\lceil(n-1)/4\rceil$ for $n\ge3$. Reproducing every complete stochastic spectrum, with multiplicities, requires $m=\lfloor n/2\rfloor$. For a boundary arc whose endpoints have orders $q<s$ as roots of unity, the least $m$ needed to attain the whole arc is $0$ for $q=1$ and $\left\lceil\frac12\left\lfloor n/q\right\rfloor\right\rceil$ for $q\ge2$. For a Karpelevič boundary arc whose nonreal points between the endpoints first occur as stochastic eigenvalues at order $n\ge4$, fixing any one of these eigenvalues fixes the characteristic polynomial of every stochastic matrix of order $n$ with that eigenvalue. For these arcs and $m\ge1$, our family attains either the whole arc or only its endpoints. Further results describe the smaller eigenvalue regions and the angles at which their boundaries approach roots of unity.

comments
46 pages, 4 figures, 16 references; revised abstract and expanded exposition, including clarifications of constructions and proofs; added funding acknowledgement
fields
Probability, Rings and Algebras
pdf ↗ page arXiv:2610.42072 ↗