Birkhoff's Theorem for Panstochastic Matrices

From MaRDI portal



Abstract: The panstochastic analogue of Birkhoff's Theorem on doubly-stochastic matrices is proved in the case n=5. It is shown that this analogue fails when n>1, ne5.


An \(n \times n\) real matrix with nonnegative entries is doubly stochastic if the sum of the entries along any of its rows or columns is equal to \(1\). A doubly stochastic matrix is panstochastic if the sum of the entries along any downward diagonal or upward diagonal, either broken or unbroken, is equal to \(1\). A linear combination is called convex if the coefficients are nonnegative and their sum is equal to \(1\). \textit{G. Birkhoff} [Rev. Univ. Nac. Tucumán, Ser. A 5, 147-151 (1946; Zbl 0060.07906)] showed that every doubly stochastic matrix can be expressed as a convex combination of permutation matrices. Birkhoff's theorem has been generalized in various ways e. g. for matrices with entries in lattice-ordered abelian groups [\textit{H. Schneider}, Acta Math. Acad. Sci. Hung. 30, 91-94 (1977; Zbl 0368.06015)]. In this paper Birkhoff's theorem is generalized on the case of panstochastic matrices. Results show that Birkhoff's theorem holds for the case of \(n=5\). The theorem that a \(5 \times 5\) real matrix is panstochastic if and only if it is a convex combination of panstochastic permutation matrices is proved. It is also shown that if \(n > 1\) and \(n \neq 5\), then there is some \(n \times n\) panstochastic matrix that is not a convex combination of panstochastic permutation matrices.











This page was built for publication: Birkhoff's Theorem for Panstochastic Matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3146536)