Stochastic matrices realising the boundary of the Karpelevič region (Q2062816)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Stochastic matrices realising the boundary of the Karpelevič region
scientific article

    Statements

    Stochastic matrices realising the boundary of the Karpelevič region (English)
    0 references
    0 references
    0 references
    3 January 2022
    0 references
    An entrywise nonnegative matrix \(A\) is called \textit{stochastic} if every row sums to the unity. \textit{A. Kolmogoroff} [Bull. Univ. Etat Moscou, Ser. Int., Sect. A, Math. et Mecan. 1, Fasc. 3, 1--16 (1937; Zbl 0018.41302)] posed the problem of characterizing the region \[ \Theta_n := \{ z \in \mathbb{C} \mid z \in \sigma(A), A \in \textsf{M}_n (\mathbb{R}), A \text{ stochastic} \}, \] now called the \textit{Karpelevich region}, which is known to be compact, star-convex (with star-centers at \(0\) and \(1\)), and a subset of the unit disc in \(\mathbb{C}\). \textit{F. I. Karpelevich} [Izv. Akad. Nauk SSSR, Ser. Mat. 15, 361--383 (1951; Zbl 0043.01603)] solved this problem by characterizing the set of extremal values: \[ E_n := \{ z \in \Theta_n \mid \alpha z \not \in \Theta_n, \forall \alpha > 1 \} \subseteq \partial \Theta_n. \] In the formulation by \textit{H. Ito} [Linear Algebra Appl. 267, 241--246 (1997; Zbl 0888.15006)] of the Karpelevich theorem, the points in \(\partial \Theta_n\) between the consecutive roots of unity \(\exp(2 \pi ip/q)\) and \(\exp(2 \pi ir/s)\), with \(q < s\), satisfy the polynomial equation \[ t^s(t^q - \beta)^{\lfloor \frac{n}{q} \rfloor} = \alpha^{\lfloor \frac{n}{q} \rfloor} t^{q\lfloor \frac{n}{q} \rfloor} \] where \(\alpha \in [0,1]\) and \(\beta := 1 - \alpha\). This area of research remained dormant until the recent work of \textit{C. R. Johnson} and \textit{P. Paparella} [Linear Algebra Appl. 520, 1--15 (2017; Zbl 1359.15029)], who provided, for every \(\lambda = \lambda(\alpha) \in \partial \Theta_n\), a parametric stochastic matrix \(M = M(\alpha)\) with \(\lambda\) as an eigenvalue. This result resolved a seventy-year old problem posed by \textit{N. Dmitriev} and \textit{E. Dynkin} [Izv. Akad. Nauk SSSR, Ser. Mat. 10, 167--184 (1946; Zbl 0060.03305)]. Furthermore, several novel conjectures were made about the nature of the Karpelevich region. More recently, \textit{B. Kim} and \textit{J. Kim} [Linear Algebra Appl. 595, 13--23 (2020; Zbl 1434.15020)] resolved two of the conjectures in the affirmative posed by \textit{C. R. Johnson} and \textit{P. Paparella}. Continuing in this vein, the authors investigate the extent to which the realizing matrices of points on the boundary are unique (see [\textit{C. R. Johnson} and \textit{P. Paparella}, Linear Algebra Appl. 520, 1--15 (2017; Zbl 1359.15029), Problem 6.1]).
    0 references
    0 references
    0 references
    0 references
    0 references
    stochastic matrix
    0 references
    eigenvalue
    0 references
    Markov chain
    0 references
    Karpelevič region
    0 references
    0 references
    0 references