Page 107 - Matrices theory and applications
P. 107

5. Nonnegative Matrices
                              90
                              we denote by O(x) the set of indices i such that x i =0, and by o(x)
                              its cardinality, in such a way that ∂K n comprises those points satisfying
                              o(x) ≥ 1. One always has m ij =0 for (i, j) ∈ O(Mx) × O(x) ,where I
                              denotes the complement of I in {1,... ,n}.
                              Proposition 5.5.1 Let x ∈ K n and M ∈ ∆ n be given. Then one has
                                                        o(Mx) ≤ o(x).               c        c
                                                                                           c
                              Moreover, if o(Mx)= o(x), one has m ij =0 for every (i, j) ∈ O(Mx) ×
                              O(x).
                              Proof
                                Let us compute
                                                n                 n

                                o(x) − o(Mx)=         m ij −        m ij =          m ij ≥ 0.
                                                                               c
                                               i=1 O(x)    O(Mx) j=1      O(Mx) ×O(x)
                              The case of equality is immediate.
                                We could have obtained the first part of the proposition by applying
                              Theorem 5.5.2.
                              Corollary 5.5.2 Let I and J be two subsets of {1,... ,n} and let M ∈ ∆ n
                                                                                c
                              be a matrix satisfying m ij =0 for every (i, j) ∈ I × J . Then one has
                                                                                         c
                              |J|≥ |I|. If, moreover, |I| = |J|,then m ij also vanishes for (i, j) ∈ I × J.
                              Proof
                                                                   c
                                                           n
                                It is sufficient to choose x ∈ K with J = O(x)if J is nonempty. If J
                              is empty, the statement is obvious.
                                We shall denote by S∆ n (S for strict) the set of doubly stochastic matri-
                              ces M for which the conditions |I| = |J| and m ij =0 for every (i, j) ∈ I×J c
                              imply either I = ∅ or I = {1,... ,n}. These are also the matrices for which
                              x ∈ ∂K n implies o(Mx) <o(x). This set does not contain permutation
                              matrices P, since these satisfy o(Px)= o(x) for every x ∈ K n .
                                Let M ∈ ∆ n be given. A decomposition of M consists of two partitions
                              I 1 ∪ ··· ∪ I r and J 1 ∪ ··· ∪ J r of the set {1,... ,n} such that
                                              (i ∈ I l ,j ∈ J m ,l  = m)=⇒ m ij =0.
                              From Corollary 5.5.2, we have |I l | = |J l | for every l. Eliminating empty
                              parts if necessary, we can always assume that none of the I l ’s or J l ’s is
                              empty. A decomposition of M furnishes a block structure, in which each
                              row-block has only one nonzero block, and the same for the column-blocks.
                              The blocks of indices I l × J l are themselves stochastic matrices. A matrix
                              of S∆ n admits only the trivial decomposition r =1, I 1 = J 1 = {1,... ,n}.
                                If M admits two decompositions, one with the sets I l ,J l ,1 ≤ l ≤ r,the


                              other one with I ,J ,1 ≤ l ≤ s, let us form the partitions ∪ l,m I      and
                                             l  l                                       lm
   102   103   104   105   106   107   108   109   110   111   112