HAL Id: hal-03171502
https://hal.archives-ouvertes.fr/hal-03171502
Preprint submitted on 17 Mar 2021
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Some inequalities for some positive block matrices
Antoine Mhanna
To cite this version:
Antoine Mhanna. Some inequalities for some positive block matrices. 2021. �hal-03171502�
Some inequalities for some positive block matrices
Antoine Mhanna 1
Kfardebian, Lebanon
Abstract
We review Lin’s inequality from [2] and (re)prove that if M = [X i,j ] n i,j=1 is positive semi-definite then L n
i=1 (X i,i − P
j6=i
X j,j ) ≤ M τ ≤ I n ⊗
n
P
i=1
X i,i where M τ is the partial transpose of M. In particular for such M ∈ M + n ( M m ) we prove that kM k ≤ min(m, n)k P n
i=1
X i,i k for all symmetric norms. Some classical results are also discussed in terms of permutations.
Keywords: positive block matrix; partial transpose; symmetric norm;
permutations.
2010 MSC: 15A60; 15A42; 15B99; 05A18.
1. Introduction and preliminaries
Let M n (M m ) denote the space of n × n complex matrices with entries in M m ( C ). We write M H n ( M m ) respectively M + n ( M m ) for the set of Hermitian block matrices respectively positive semi-definite matrices in M n ( M m ). Let Im(X) := X − X ∗
2i resp. Re(X) := X + X ∗
2 be the imaginary part resp.
the real part of a matrix X, if W (X) denotes the field of values of X then W (Re(X)) = <(W (X)) and W (Im(X)) = =(W (X)) see [3]. A positive semi-definite matrix A is denoted by A ≥ 0 and a norm k.k over the space of matrices is a symmetric norm if kU AV k = kAk for all A and all unitaries U and V.
For a matrix M = [X i,j ] ∈ M n ( M m ) we denote M τ the matrix [X i,j ∗ ] where each block is replaced by its adjoint. We define Tr 1 (M ) := P n
i=1 X i,i , Tr 2 (M ) := [Tr(X i,j )] for Tr 1 (M ) resp. Tr 2 (M ) is an m × m matrix resp. an n × n complex matrix. M is positive partial transpose (P.P.T.) if M ≥ 0
1
Email: tmhanat@yahoo.com
and M τ ≥ 0. I m (or I) is the identity matrix of dimension m and a 0 entry is an all zero submatrix. We finally denote λ i (M) resp. σ i (M ), i = 1, . . . , n the eigenvalues resp. the singular values of M in decreasing order.
We need the following notations used latter in the article:
For X ∈ M n , the vertical width ω v (X) respectively the width ω(X) of W (X) is the smallest possible real number such that W (X) is contained in a vertical strip of width ω v (X) respectively in a strip of width ω(X) with ω(X) ≤ ω v (X).
If S = {M k , 1 ≤ k ≤ i, M k ∈ M m } of cardinality i, {S} denotes the 2 i−1 set of matrices {M 1 ± . . . ± M i } and an element in {S} is denoted by S [σ] , we write ω v ({S}) to mean the 2 i−1 tuple ω v (S [σ] ). Notice that if the width of W (A) is ω, then one can find a θ ∈ [0, π] such that
rI m ≤ Re(e iθ A) ≤ (r + ω)I m for some r ∈ R and ω = ω v (e iθ A).
Finally if M := [X i,j ] n i,j=1 ∈ M n ( M m ) with blocks X i,j we write M = [
M[x k,l ] i,j ] where
M[x k,l ] i,j is the complex number belonging to block X i,j on its line k and column l for k, l: 1 . . . m and i, j: 1 . . . n one has
M T
[x k,l ] i,j =
M[x l,k ] j,i ;
M τ[x k,l ] i,j =
M[x l,k ] i,j . The matrix ˜ M defined in M m ( M n ) by
˜M