Subproper splitting for rectangular matrices

In this paper iterative schemes for approximating a solution to a rectangular but consistent linear system Ax = b are studied. Let A?C^{m × n}_r. The splitting A = M ? N is called subproper if R(A) ? R(M) and $\text{R(A}{}^1\text{) ?R(M}{}^1\text{)}$. Consider the iteration $\text{x}{}_{\mathrm{i}}\text{= M}{}^2\text{Nx}{}_{\mathrm{i}}{}_{\mathrm{?1}}\text{+ M}{}^2\text{b}$. We characterize the convergence of this scheme to a solution of the linear system. When A?R^m×ⁿ_r, monotonicity and the concept of subproper regular splitting are used to determine a necessary and a sufficient condition for the scheme to converge to a solution.     

About some problems of spectral synthesis

For a given pair $\text{(A,b)∈}\text{R}{}^{\mathrm{n\times n}}\text{×}\text{R}{}^{\mathrm{n\times 1}}$ such that A is cyclic and b is a cyclic generator (with respect to A) of $\text{R}{}^{\mathrm{n\times 1}}$, it is shown that for every nonnegative integer m we can find a nonnegative integer t and a sequence $\text{{f}{}_{\mathrm{j\}}}{}^{\mathrm{t}}{}_{\mathrm{j=0}}\text{,f}{}_{\mathrm{j}}\text{∈}\text{R}{}^{\mathrm{1\times n}}$,so that a the zeros of the rational function det P(z), where $\text{P(z) = zI ? A ? ∑}{}^{\mathrm{t}}{}_{\mathrm{j=0}}\text{z}{}^{\mathrm{-(m+j)}}\text{b?}$_f, lie in the open unit disc in the complex plane. The result is directly applicable to a stabilizability problem for linear systems with a time delay in control action.     

The asymptotic eigenvalue-distribution for a certain class of random matrices

For an n × n Hermitean matrix A with eigenvalues λ₁, …, λ_n the eigenvalue-distribution is defined by $\text{G(x, A) :=}\text{1}\text{n}$ · number {λ_i: λ_i ? x} for all real x. Let A_n for n = 1, 2, … be an n × n matrix, whose entries a_ik are for i, k = 1, …, n independent complex random variables on a probability space (Ω, $\text{R}$, p) with the same distribution F_a. Suppose that all moments $\text{E}$ | a | ^k, k = 1, 2, … are finite, $\text{E}$a=0 and $\text{E}$ | a | ². Let

$\text{M(A)=}\text{∑}\text{σ=1}\text{s}\text{θ}{}_{\mathrm{\sigma }}\text{P}{}_{\mathrm{\sigma }}\text{(A,A}{}^1\text{)}$

with complex numbers θ_σ and finite products P_σ of factors A and $\text{A}{}^1$ (= Hermitean conjugate) be a function which assigns to each matrix A an Hermitean matrix M(A). The following limit theorem is proved: There exists a distribution function G₀(x) = G₁x) + G₂(x), where G₁ is a step function and G₂ is absolutely continuous, such that with probability $\text{1 G(x, M(}\text{A}{}_{\mathrm{n}}\text{n}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}\text{))}$ converges to G₀(x) as n → ∞ for all continuity points x of G₀. The density g of G₂ vanishes outside a finite interval. There are only finitely many jumps of G₁. Both, G₁ and G₂, can explicitly be expressed by means of a certain algebraic function f, which is determined by equations, which can easily be derived from the special form of M(A). This result is analogous to Wigner's semicircle theorem for symmetric random matrices (E. P. Wigner, Random matrices in physics, SIAM Review9 (1967), 1–23). The examples $\text{A}{}_{\mathrm{r}}\text{A}{}^{\mathrm{1r}}$, $\text{A}{}^{\mathrm{r}}\text{+ A}{}^{\mathrm{1r}}$, $\text{A}{}^{\mathrm{r}}\text{A}{}^{\mathrm{1r}}\text{± A}{}^{\mathrm{1r}}\text{A}{}^{\mathrm{r}}$, r = 1, 2, …, are discussed in more detail. Some inequalities for random matrices are derived. It turns out that with probability 1 the sharpened form

$\text{lim sup}\text{n→∞}\text{∑}\text{i=1}{}^{\mathrm{n}}\text{|γ}{}_{\mathrm{i}}{}^{\mathrm{(n)}}\text{|}{}^2\text{?6A}{}_{\mathrm{n}}\text{6}{}^2\text{? 0.8228…}$

of Schur's inequality for the eigenvalues λ_i⁽ⁿ⁾ of A_n holds. Consequently random matrices do not tend to be normal matrices for large n.     

Linear operators preserving the (p,q)-numerical range

Let $\text{C}{}_{\mathrm{n\times n}}$ and $\text{H}{}_{\mathrm{n}}$ denote respectively the space of n×n complex matrices and the real space of n×n hermitian matrices. Let p,q,n be positive integers such that p?q?n. For $\text{A?}\text{C}{}_{\mathrm{n\times n}}$, the (p,q)-numerical range of A is the set

$\text{W}{}_{\mathrm{p,q}}\text{(A)={trC}{}_{\mathrm{p}}\text{(J}{}_{\mathrm{q}}\text{UAU}{}^1\text{):U}\text{unitary}\text{}}$

, where C_p(X) is the pth compound matrix of X, and Jq is the matrix I_q?O_n-q. Let $\text{L}$ denote $\text{H}$_n or $\text{C}{}_{\mathrm{n\times n}}$. The problem of determining all linear operators T: $\text{L}$→$\text{L}$ such that

$\text{W}{}_{\mathrm{p,q}}\text{(T(A))=W}{}_{\mathrm{p,q}}\text{(A)}\text{for all}\text{A?}\text{L}$

is treated in this paper.     

Duality in hypercomplex function theory

Let $\text{A}$ be the Clifford algebra constructed over a quadratic n-dimensional real vector space with orthogonal basis {e₁,…, e_n}, and e₀ be the identity of $\text{A}$. Furthermore, let M_k(Ω;$\text{A}$) be the set of $\text{A}$-valued functions defined in an open subset Ω of R^m+1 (1 ? m ? n) which satisfy D^kf = 0 in Ω, where D is the generalized Cauchy-Riemann operator $\text{D = ∑}{}_{\mathrm{i\; =\; 0}}{}^{\mathrm{m}}\text{e}{}_{\mathrm{i}}\text{(}\text{?}\text{?x}{}_{\mathrm{i}}\text{)}$ and k? N. The aim of this paper is to characterize the dual and bidual of M_k(Ω;$\text{A}$). It is proved that, if M_k(Ω;$\text{A}$) is provided with the topology of uniform compact convergence, then its strong dual is topologically isomorphic to an inductive limit space of Fréchet modules, which in its turn admits M_k(Ω;$\text{A}$) as its dual. In this way, classical results about the spaces of holomorphic functions and analytic functionals are generalized.     

Duality between some linear preserver problems. II. Isometries with respect to c-special norms and matrices with fixed singular values

Let $\text{F}{}_{\mathrm{m\times n}}$ (m?n) denote the linear space of all m × n complex or real matrices according as $\text{F}$=$\text{C}$ or $\text{R}$. Let c=(c₁,…,c_m)≠0 be such that c₁???c_m?0. The c-spectral norm of a matrix A?$\text{F}$_m×n is the quantity

$\text{6A6}{}_{\mathrm{c}}\text{∑}\text{i=I}\text{m}\text{c}{}_{\mathrm{i}}\text{σ}{}_{\mathrm{i}}\text{(A)}$

. where σ₁(A)???σ_m(A) are the singular values of A. Let d=(d₁,…,d_m)≠0, where d₁???d_m?0. We consider the linear isometries between the normed spaces $\text{(}\text{F}{}_{\mathrm{m\times }}{}_{\mathrm{n}}\text{,∥·∥}{}_{\mathrm{<mtext>c</mtext>}}\text{)}$ and $\text{(}\text{F}{}_{\mathrm{m\times }}{}_{\mathrm{n}}\text{,∥·∥}{}_{\mathrm{d}}\text{)}$, and prove that they are dual transformations of the linear operators which map $\text{L}$(d) onto $\text{L}$(c), where

$\text{L}\text{(c)= {X?}\text{F}{}_{\mathrm{m\times n}}\text{:X}\text{has singular values}\text{c}{}_1\text{,…,c}{}_{\mathrm{m}}\text{}}$

.     

Matricial logarithmic derivatives

If θ is a norm on Cⁿ, then the mapping $\text{A→}\text{lim}{}_{\mathrm{h\downarrow 0}}\text{6I+hA6}{}_{\mathrm{\theta }}\text{?1}\text{/h}$ from M_n(C) (=C^{n × n}) into R is called the logarithmic derivative induced by the vector norm θ. In this paper we generalize this concept to a mapping γ from M_n(C) into M_k(R), where k ? n. Denoting by α(B) the spectral abscissa of a square matrix B (the largest of the real parts of the eigenvalues), we show, in particular, that α(A) ?α(γ(A)). As a byproduct we obtain simple sufficient conditions for the stability of a matrix.     

Explicit zero-free regions for Dirichlet L-functions

Let $\text{L}$_k(S) be the product of the ?(k) Dirichlet L-functions formed with characters modulo k. We prove the existence of explicit numerical zero-free regions for $\text{L}$_k(S). The first result is that $\text{L}$_k(S) has at most a single zero in the region {$\text{s: σ > 1 ?}\text{1}\text{(R}\text{log}\text{M)}\text{}}$, where R = 9.645908801 and M = max {k, k |t|, 10}. The only possible zero in this region is a simple real zero arising from an L-function formed with a real non-principal character. The second result is that if χ₁ and χ₂ are distinct real primitive characters modulo k₁ and k₂, respectively, and if β₁ is a zero of L(s, χ_i), i = 1, 2, then min $\text{{β}{}_1\text{, β}{}_2\text{} < 1 ?}\text{1}\text{(R}{}_1\text{log}\text{M}{}_1\text{)}$, where $\text{R}{}_1\text{=}\text{(5 ? √5)}\text{(15 ? 10√2)}$, and $\text{M}{}_1\text{=}\text{max}\text{{}\text{k}{}_1\text{k}{}_2\text{17}\text{, 13}}$.     

Three problems on the lengths of increasing runs

Let U₁, U₂,… be a sequence of independent, uniform (0, 1) r.v.'s and let R₁, R₂,… be the lengths of increasing runs of {U_i}, i.e., X₁=R₁=inf{i:U_i+1<U_i},…, X_n=R₁+R₂+?+R_n=inf{i:i>X_n?1,U_i+1<U_i}. The first theorem states that the sequence $\text{(}\text{3}\text{2}\text{n)}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}\text{(X}{}_{\mathrm{n}}\text{?2n)}$ can be approximated by a Wiener process in strong sense.Let τ(n) be the largest integer for which R₁+R₂+?+R_τ(n)?n, $\text{R}{}^1{}_{\mathrm{n}}\text{=n?(R}{}_1\text{+R}{}_2\text{+?+R}{}_{\mathrm{\tau (n)}}\text{)}$ and $\text{M}{}_{\mathrm{n}}\text{=}\text{max}\text{{R}{}_1\text{,R}{}_2\text{,…,R}{}_{\mathrm{\tau (n)}}\text{,R}{}^1{}_{\mathrm{n}}\text{}}$. Here M_n is the length of the longest increasing block. A strong theorem is given to characterize the limit behaviour of M_n.The limit distribution of the lengths of increasing runs is our third problem.     

A combinatorial perspective on the non-Radon partitions

Let E be a finite set of points in $\text{R}$^d. Then {A, E ? A} is a non-Radon partition of E iff there is a hyperplane H separating A strictly from E?A. Or equivalently iff ${}_{\mathrm{<mtext>A</mtext>}}\text{O}$ is an acyclic reorientation of (M_Aff(E), O), the oriented matroid canonically determined by E. If (M(E), O) is an oriented matroid without loops then the set $\text{NR(E, O) = {(A, E ? A):}{}_{\mathrm{<mtext>A</mtext>}}\text{O}\text{is acyclic}\text{}}$ determines (M(E), O). In particular the matroidal properties of a finite set of points in $\text{R}$^d are precisely the properties which can be formulated in non-Radon partitions terms. The Möbius function of the poset $\text{A = {A: A ? E,}{}_{\mathrm{<mtext>A</mtext>}}\text{O}\text{is acyclic}\text{}}$ and in a special case its homotopy type are computed. This paper generalizes recent results of P. Edelman (A partial order on the regions of $\text{R}$ⁿ dissected by hyperplanes     

Best pseudo-isolated Gerschgorin disks for Eigenvalues

Let A be an arbitrary n×n matrix, partitioned so that if A=[A_ij], then all submatrices A_ii are square. If x is a positive vector, it is well-known that $\text{G(}\text{x}\text{) =∪}{}^{\mathrm{N}}{}_{\mathrm{i=1}}\text{G}{}_{\mathrm{i}}\text{(}\text{x}\text{)}$, where

$\text{G}{}_{\mathrm{i}}\text{(}\text{x}\text{) =}\text{z}\text{6(zI ? A}{}_{\mathrm{ii}}\text{)}{}^{\mathrm{?1}}\text{6}{}^{\mathrm{?1}}\text{?}\text{1}\text{x}{}_{\mathrm{i}}\text{∑}\text{j = 1}\text{j ≠ i}\text{N}\text{`6A}{}_{\mathrm{ij}}\text{6x}{}_{\mathrm{j}}$

, contains all the eigenvalues of A. The purpose of this paper is to give a new definition of the concept of an isolated subregion of G(x). An algorithm is given for obtaining the best such isolated subregion in a certain sense, and examples are given to show that tighter bounds for some eigenvalues of A may be obtained than with previous algorithms. For ease of computation, each subregion G_i(x) is replaced by the union of circular disks centered at the eigenvalues of A_ii.     

A group of integral points in a matrix parallelepiped

Let A be an n×n integral matrix with determinant D>0, and let P(A) be the n-parallelepiped determined by the columns {A_i}ⁿ_i=1 of A,

$\text{P(A)=}\text{∑}\text{i=1}\text{n}\text{x}{}_{\mathrm{i}}\text{A}{}_{\mathrm{i}}\text{0<x}{}_{\mathrm{i}}\text{<1}$

Let L be the set of integral vectors in P(A), and let G(A) be the subset of L consisting of vectors whose coefficients x_i satisfy 0?x_i<1. We show that G(A), equipped with addition modulo 1 on the coefficients x_i, is an Abelian group of order D, whose invariant factors are the invariant factors of the integral matrix A. We give a formula for |L|, and show that |L| is not a similarity invariant.     

Note on lower bounds for the rank of a matrix

The following results are proved: Let A = (a_ij) be an n × n complex matrix, n ? 2, and let k be a fixed integer, 1 ? k ? n ? 1.(1) If there exists a monotonic G-function f = (f₁,…,f_n) such that for every subset of S of {1,…,n} consisting of k + 1 elements we have

$\text{Π}\text{i∈S}\text{f}{}_{\mathrm{i}}\text{(A)<}\text{Π}\text{i∈S}\text{|a}{}_{\mathrm{ii}}\text{|,}$

then the rank of A is ? n ? k + 1. (2) If A is irreducible and if there exists a G-function f = (f₁,…,f_n) such that for every subset of S of {1,…,n} consisting of k + 1 elements we have

$\text{Π}\text{i∈S}\text{f}{}_{\mathrm{i}}\text{(A)<}\text{Π}\text{i∈S}\text{|a}{}_{\mathrm{ii}}\text{|,}$

then the rank of A is ? n ? k + 1 if k ? 2, n ? 3; it is ? n ? 1 if k = 1.     

Generalized Schur complements

Let A be an n×n complex matrix. For a suitable subspace $\text{M}$ of Cⁿ the Schur compression A _$\text{M}$ and the (generalized) Schur complement A_/$\text{M}$ are defined. If A is written in the form

$\text{A=}\text{B}\text{C}\text{S}\text{T}$

according to the decomposition $\text{C}{}^{\mathrm{n}}\text{=}\text{M}\text{⊕}\text{M}{}^{\mathrm{\perp }}$ and if B is invertible, then

$\text{A}{}_{\mathrm{<mtext>M</mtext>}}\text{=}\text{B}\text{C}\text{S}\text{SB}{}^{\mathrm{?1}}\text{C}$

and

$\text{A}{}_{\mathrm{<mtext>/</mtext><mtext>M</mtext>}}\text{=}\text{0}\text{0}\text{0}\text{T?SB}{}^{\mathrm{?1}}\text{C}\text{·}$

The commutativity rule for Schur complements is proved:

$\text{(A}{}_{\mathrm{<mtext>/</mtext><mtext>M</mtext>}}\text{)}{}_{\mathrm{<mtext>/</mtext><mtext>N</mtext>}}\text{=(A)}{}_{\mathrm{<mtext>/</mtext><mtext>N</mtext>}}\text{)}{}_{\mathrm{<mtext>/</mtext><mtext>M</mtext>}}\text{·}$

This unifies Crabtree and Haynsworth's quotient formula for (classical) Schur complements and Anderson's commutativity rule for shorted operators. Further, the absorption rule for Schur compressions is proved:

$\text{(A}{}_{\mathrm{<mtext>/</mtext><mtext>M</mtext>}}\text{)}{}_{\mathrm{<mtext>N</mtext>}}\text{=(A}{}_{\mathrm{<mtext>N</mtext>}}\text{)}{}_{\mathrm{<mtext>M</mtext>}}\text{=A}{}_{\mathrm{<mtext>M</mtext>}}\text{whenever}\text{M}\text{?}\text{N}$

.     

A multiplication of distributions

If Ω denotes an open subset of $\text{R}$ⁿ (n = 1, 2,…), we define an algebra $\text{g}$ (Ω) which contains the space $\text{D}$′(Ω) of all distributions on Ω and such that $\text{C}{}^{\mathrm{\infty }}\text{(Ω)}$ is a subalgebra of $\text{G}$ (Ω). The elements of $\text{G}$ (Ω) may be considered as “generalized functions” on Ω and they admit partial derivatives at any order that generalize exactly the derivation of distributions. The multiplication in $\text{G}$(Ω) gives therefore a natural meaning to any product of distributions, and we explain how these results agree with remarks of Schwartz on difficulties concerning a multiplication of distributions. More generally if q = 1, 2,…, and $\text{?∈O}{}_{\mathrm{M}}\text{(}\text{R}{}^{\mathrm{2q}}\text{)}$—a classical Schwartz notation—for any G₁,…,G_q∈G(σ), we define naturally an element $\text{?G}{}_1\text{,…,G}{}_{\mathrm{q}}\text{∈G(σ)}$. These results are applied to some differential equations and extended to the vector valued case, which allows the multiplication of vector valued distributions of physics.     

Limiting distributions of two random sequences

For fixed p (0 ≤ p ≤ 1), let {L₀, R₀} = {0, 1} and X₁ be a uniform random variable over {L₀, R₀}. With probability p let {L₁, R₁} = {L₀, X₁} or = {X₁, R₀} according as $\text{X}{}_1\text{≥}\text{1}\text{2}\text{(L}{}_0\text{+ R}{}_0\text{)}\text{or}\text{<}\text{1}\text{2}\text{(L}{}_0\text{+ R}{}_0\text{)}$; with probability 1 ? p let {L₁, R₁} = {X₁, R₀} or = {L₀, X₁} according as $\text{X}{}_1\text{≥}\text{1}\text{2}\text{(L}{}_0\text{+ R}{}_0\text{)}\text{or}\text{<}\text{1}\text{2}\text{(L}{}_0\text{+ R}{}_0\text{)}$, and let X₂ be a uniform random variable over {L₁, R₁}. For n ≥ 2, with probability p let {L_n, R_n} = {L_{n ? 1}, X_n} or = {X_n, R_{n ? 1}} according as $\text{X}{}_{\mathrm{n}}\text{≥}\text{1}\text{2}\text{(L}{}_{\mathrm{n\; ?\; 1}}\text{+ R}{}_{\mathrm{n\; ?\; 1}}\text{)}\text{or}\text{<}\text{1}\text{2}\text{(L}{}_{\mathrm{n\; ?\; 1}}\text{+ R}{}_{\mathrm{n\; ?\; 1}}\text{)}$, with probability 1 ? p let {L_n, R_n} = {X_n, R_{n ? 1}} or = {L_{n ? 1}, X_n} according as $\text{X}{}_{\mathrm{n}}\text{≥}\text{1}\text{2}\text{(L}{}_{\mathrm{n\; ?\; 1}}\text{+ R}{}_{\mathrm{n\; ?\; 1}}\text{)}\text{or}\text{<}\text{1}\text{2}\text{(L}{}_{\mathrm{n\; ?\; 1}}\text{+ R}{}_{\mathrm{n\; ?\; 1}}\text{)}$, and let X_{n + 1} be a uniform random variable over {L_n, R_n}. By this iterated procedure, a random sequence {X_n}_{n ≥ 1} is constructed, and it is easy to see that X_n converges to a random variable Y_p (say) almost surely as n → ∞. Then what is the distribution of Y_p? It is shown that the Beta, (2, 2) distribution is the distribution of Y₁; that is, the probability density function of Y₁ is g(y) = 6y(1 ? y) I_0,1(y). It is also shown that the distribution of Y₀ is not a known distribution but has some interesting properties (convexity and differentiability).     

Injectivity and operator spaces

Let B(H) be the bounded operators on a Hilbert space H. A linear subspace R ? B(H) is said to be an operator system if 1 ?R and R is self-adjoint. Consider the category $\text{b}$ of operator systems and completely positive linear maps. R ∈ $\text{C}$ is said to be injective if given A ? B, A, B ∈ $\text{C}$, each map A → R extends to B. Then each injective operator system is isomorphic to a conditionally complete C¹-algebra. Injective von Neumann algebras R are characterized by any one of the following: (1) a relative interpolation property, (2) a finite “projectivity” property, (3) letting M_m = B(C^m), each map R → N ? M_m has approximate factorizations R → M_n → N, (4) letting K be the orthogonal complement of an operator system N ? M_m, each map $\text{M}{}_{\mathrm{m}}\text{K}\text{→ R}$ has approximate factorizations $\text{M}{}_{\mathrm{m}}\text{K}\text{→ M}{}_{\mathrm{n}}\text{→ R}$. Analogous characterizations are found for certain classes of C¹-algebras.     

Convex polyhedra of doubly stochastic matrices: II. Graph of Ωn

Properties of the graph $\text{G(Ω}{}_{\mathrm{n}}\text{)}$ of the polytope $\text{Ω}{}_{\mathrm{n}}$ of all n × n nonnegative doubly stochastic matrices are studied. If $\text{F}$ is a face of $\text{Ω}{}_{\mathrm{n}}$ which is not a k-dimensional rectangular parallelotope for k ≥ 2, then G($\text{F}$) is Hamilton connected. Prime factor decompositions of the graphs of faces of $\text{Ω}{}_{\mathrm{n}}$ relative to Cartesian product are investigated. In particular, if $\text{F}$ is a face of $\text{Ω}{}_{\mathrm{n}}$, then the number of prime graphs in any prime factor decomposition of G($\text{F}$) equals the number of connected components of the neighborhood of any vertex of G($\text{F}$). Distance properties of the graphs of faces of $\text{Ω}{}_{\mathrm{n}}$ are obtained. Faces $\text{F}$ of $\text{Ω}{}_{\mathrm{n}}$ for which G($\text{F}$) is a clique of $\text{G(Ω}{}_{\mathrm{n}}\text{)}$ are investigated.