4
$\begingroup$

Given a non-negative matrix $A$, we call $A$ primitive if $A^k$ has all strictly positive entries with some $k>0$. Given primitive $A$, is there relation between smallest $k$ such that $A^k>0$ and $rank(A)$?

If $A\in\Bbb Z_{\geq0}^{n\times n}$ with $\max_{i,j}A_{ij}\leq M$, is $log(k(A))\leq(log(rk(A)))^{c_M}$ with some $c_M\geq1$?

Is there a geometric meaning to smallest $k$?

$\endgroup$
0

1 Answer 1

5
$\begingroup$

Denote this smallest $k$ by $k(A)$. At least, one has (obvious) $$({\rm rk}(A)=1)\Longrightarrow(k(A)=1).$$ On the other hand, Wielandt's Theorem says that for every primitive matrix $A\ge0_n$, one has $$k(A)\le n^2-2n+2.$$ This bound is optimal: equality holds for the matrix $A=P+E_{n2}$, where $P$ is the matrix of the permutation $i\mapsto i+1$, modulo $n$, and $E_{rs}$ is the matrix with entries $\delta_i^r\delta_j^s$ at positions $(i,j)$.

Do you expect the general formula $$k(A)\le{\rm rk}(A)^2-2{\rm rk}(A)+2\quad?$$ Should be nice.

$\endgroup$
1
  • $\begingroup$ I do not expect such formula. It will be miraculous. Infact I do not expect anything polynomial or exponential. Something superpolynomial is more likelier than even subexponential. $\endgroup$
    – Turbo
    Commented Feb 18, 2015 at 22:10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .