Distance Preserving Maps on Matrices. Chi-Kwong Li Department of Mathematics College of William and Mary Williamsburg, Virginia

Similar documents
Recent study in preserver problems. Chi-Kwong Li Department of Mathematics College of William and Mary Williamsburg, VA 23185

Linear Preserver Problems

Spectral inequalities and equalities involving products of matrices

for all A M m and B M n Key words. Complex matrix, linear preserver, spectral norm, Ky Fan k-norm, Schatten p-norm, tensor product.

A survey on Linear Preservers of Numerical Ranges and Radii. Chi-Kwong Li

arxiv: v3 [math.ra] 22 Aug 2014

Isometries Between Matrix Algebras

PRESERVERS OF UNITARY SIMILARITY FUNCTIONS ON LIE PRODUCTS OF MATRICES

Contents. Preface for the Instructor. Preface for the Student. xvii. Acknowledgments. 1 Vector Spaces 1 1.A R n and C n 2

Linear and Multilinear Algebra. Linear maps preserving rank of tensor products of matrices

LINEAR PRESERVER PROBLEMS: generalized inverse

A geometric inequality on the positive cone and an application

Compression, Matrix Range and Completely Positive Map

Schur Multiplicative Maps on Matrices

Linear Operators Preserving the Numerical Range (Radius) on Triangular Matrices

LINEAR MAPS PRESERVING NUMERICAL RADIUS OF TENSOR PRODUCTS OF MATRICES

Maps which preserve equality of distance. Must they be linear?

AFFINE MAPPINGS OF INVERTIBLE OPERATORS. Lawrence A. Harris and Richard V. Kadison

Decomposable Numerical Ranges on Orthonormal Tensors

Maps on matrix spaces

Maps preserving the nilpotency of products of operators

Convexity of the Joint Numerical Range

Review problems for MA 54, Fall 2004.

LinGloss. A glossary of linear algebra

CHI-KWONG LI AND YIU-TUNG POON. Dedicated to Professor John Conway on the occasion of his retirement.

Linear Algebra. Workbook

PSEUDOSPECTRA OF SPECIAL OPERATORS AND PSEUDOSECTRUM PRESERVERS

PSEUDOSPECTRA OF SPECIAL OPERATORS AND PSEUDOSECTRUM PRESERVERS

Lecture notes on Quantum Computing. Chapter 1 Mathematical Background

Lecture 5. Ch. 5, Norms for vectors and matrices. Norms for vectors and matrices Why?

Uniqueness of the Solutions of Some Completion Problems

The optimal version of Hua s fundamental theorem of geometry of rectangular matrices

Numerical radius, entropy and preservers on quantum states

CS 766/QIC 820 Theory of Quantum Information (Fall 2011)

A matrix problem and a geometry problem

The optimal version of Hua s fundamental theorem of geometry of matrices

DAVIS WIELANDT SHELLS OF NORMAL OPERATORS

Elementary linear algebra

The automorphism group of separable states in quantum information theory. University of Illinois at Chicago, Chicago, Illinois ,

Abstract. In this article, several matrix norm inequalities are proved by making use of the Hiroshima 2003 result on majorization relations.

Preservers of matrix pairs with a fixed inner product value

Higher rank numerical ranges of rectangular matrix polynomials

Review of linear algebra

Induced Norms, States, and Numerical Ranges

MA441: Algebraic Structures I. Lecture 14

Summer School on Quantum Information Science Taiyuan University of Technology

Ir O D = D = ( ) Section 2.6 Example 1. (Bottom of page 119) dim(v ) = dim(l(v, W )) = dim(v ) dim(f ) = dim(v )

Matrix Lie groups. and their Lie algebras. Mahmood Alaghmandan. A project in fulfillment of the requirement for the Lie algebra course

Lecture 2 Some Sources of Lie Algebras

ALGEBRA QUALIFYING EXAM PROBLEMS LINEAR ALGEBRA

NUMERICAL RANGE OF LIE PRODUCT OF OPERATORS

arxiv:math/ v1 [math.fa] 4 Jan 2007

Multiplicative maps preserving the higher rank numerical ranges and radii

ON CONNES AMENABILITY OF UPPER TRIANGULAR MATRIX ALGEBRAS

1: Lie groups Matix groups, Lie algebras

Quantum computing and mathematical research. Chi-Kwong Li The College of William and Mary

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2017 LECTURE 5

THE SINGULAR VALUE DECOMPOSITION AND LOW RANK APPROXIMATION

Banach Journal of Mathematical Analysis ISSN: (electronic)

Singular Value Decomposition and Polar Form

Duke University, Department of Electrical and Computer Engineering Optimization for Scientists and Engineers c Alex Bronstein, 2014

arxiv:math/ v1 [math.fa] 4 Nov 2000

Fréchet differentiability of the norm of L p -spaces associated with arbitrary von Neumann algebras

arxiv:math/ v2 [math.fa] 29 Mar 2007

1 Linear Algebra Problems

Notes on matrix arithmetic geometric mean inequalities

Operators with numerical range in a closed halfplane

RANKS OF QUANTUM STATES WITH PRESCRIBED REDUCED STATES

NUMERICAL RADIUS PRESERVING LINEAR MAPS ON BANACH ALGEBRAS. Sahand University of Technology New City of Sahand, Tabriz, 51335/1996, IRAN

BERNARD RUSSO. University of California, Irvine

A Note on Eigenvalues of Perturbed Hermitian Matrices

Some inequalities for unitarily invariant norms of matrices

The C-Numerical Range in Infinite Dimensions

DERIVATIONS. Introduction to non-associative algebra. Playing havoc with the product rule? BERNARD RUSSO University of California, Irvine

PRODUCT OF OPERATORS AND NUMERICAL RANGE

Throughout these notes we assume V, W are finite dimensional inner product spaces over C.

Numerical Ranges and Dilations. Dedicated to Professor Yik-Hoi Au-Yeung on the occasion of his retirement

THE ARITHMETIC, GEOMETRIC AND HARMONIC MEANS IN OPERATOR ALGEBRAS AND TRANSFORMATIONS AMONG THEM

Preservers of eigenvalue inclusion sets

Compound matrices and some classical inequalities

Introduction to Numerical Linear Algebra II

BERNARD RUSSO. University of California, Irvine

PROOF OF TWO MATRIX THEOREMS VIA TRIANGULAR FACTORIZATIONS ROY MATHIAS

Quantum Computing Lecture 2. Review of Linear Algebra

Review of Some Concepts from Linear Algebra: Part 2

Lecture 21 - Jordan Algebras and Projective Spaces

Lecture II: Linear Algebra Revisited

CLASSIFICATION OF COMPLETELY POSITIVE MAPS 1. INTRODUCTION

Matrix Support Functional and its Applications

MAT 445/ INTRODUCTION TO REPRESENTATION THEORY

Extremal Characterizations of the Schur Complement and Resulting Inequalities

Matrix Mathematics. Theory, Facts, and Formulas with Application to Linear Systems Theory. Dennis S. Bernstein

GENERALIZED NUMERICAL RANGES AND QUANTUM ERROR CORRECTION

Lecture 23: Determinants (1)

Lecture Notes in Linear Algebra

Linear Algebra Review. Vectors

Positive and doubly stochastic maps, and majorization in Euclidean Jordan algebras

Some inequalities for sum and product of positive semide nite matrices

2. Linear algebra. matrices and vectors. linear equations. range and nullspace of matrices. function of vectors, gradient and Hessian

CANONICAL FORMS, HIGHER RANK NUMERICAL RANGES, TOTALLY ISOTROPIC SUBSPACES, AND MATRIX EQUATIONS

Transcription:

Distance Preserving Maps on Matrices Chi-Kwong Li Department of Mathematics College of William and Mary Williamsburg, Virginia 23187-8795

General problem Suppose is a norm on a matrix space M. Characterize φ : M M such that (a) φ is linear/additive/multiplicative and φ(a) = A for all A M. (b) φ satisfies φ(a) φ(b) = A B for all A, B M, where A B is one of the following operations: A + B, A B, AB, ABA, (AB + BA)/2, etc. (c) If you are not interested in norm, consider other functions F and study φ such that F (φ(a) φ(b)) = F (A B) for all A, B M.

Unitarily invariant norms A norm on M m,n is unitarily invariant (UI) if UAV = A for all A M m,n and unitary matrices U M m, V M n. Evidently, 1. UI norms depend only on the singular valus of A: s 1 (A) s 2 (A). 2. Let U and V be unitary, and define φ : M m,n M m,n by (a) A UAV, or (b) m = n and A UA t V. Then φ is a linear isometry for. Theorem [Li and Tsing, 1990] Let be a UI norm on M m,n not equal to a multiple of the Frobenius norm. A linear isometry of has the form (a) or (b).

Special cases 1. [Kadison, 1951] The operator norm A op = s 1 (A). 2. [Russo, 1969] The trace norm A tr = s 1 (A)+ +s n (A). 3. [Arazy, 1975] For p 1, the Schatten p-norm S p (A) = (s 1 (A) p + + s n (A) p ) 1/p. When p = 2, it reduces to the Frobenius norm. 4. [Grone & Marcus, 1976; Grone, 1977] For 1 k n, the Ky Fan k-norm F k (A) = s 1 (A) + + s k (A). 5. [Li & Tsing, 1988] For 1 k min{m, n} and p 1, the (p, k)-norm A (p,k) = (s 1 (A) p + + s k (A) p ) 1/p. 6. [Li & Tsing, 1988] Given c = (c 1,..., c n ) with c 1 c n, the c-norm A c = c 1 s 1 (A) + + c n s n (A).

Some proof techniques Geometrical approach Let be a norm on M n. Denote the unit norm ball by B = B = {A M n : A 1}, and the set of its extreme points by E = E. A linear map φ : M n M n is an isometry for if and only if φ(e) = E. Example The set of extreme points of the unit ball of the operator norm is the group U n of unitary matrices. Sometimes, special (algebraic/differential) geometrical features of boundary points were used.

Duality technique Let (A, B) = tr (AB ) = a ij bij be the inner product on M m,n. Define the dual norm of on M n by A = max{ (A, B) : B 1}, and the dual transformation of a linear map φ : M n M n to be the unique linear map φ : M n M n such that (φ(a), B) = (A, φ (B)) for all A, B M n. Facts (a) If φ is an isometry for, then φ is an isometry for. (b) If φ has the standard form, then so is φ. (c) The operator norm and the trace norm are dual to each other.

A group theory approach The set of linear isometries for a norm form a group. Let U m U n be the group of operators of the form A UAV for a pair of unitary matrices U, V. Theorem [Djokovic & Li, 1994] Let G be the isometry group of a given UI norm on m n complex matrices. Then one of the following holds. (a) U mn. (b) m n and G = U m U n. (c) m = n and G = U m U n, τ, where τ(a) = A t.

The real case Theorem [Li & Tsing, 1990], [Djokovic & Li, 1994]. Let G be the isometry group of a UI norm on m n real matrices. Then one of the following holds. (a) G = U mn. (b) m n and G = U m U n. (c) m = n and G = U m U n, τ. (d) m = n = 4 and G = U 4 U 4, τ, ψ, ψ(a) = (A + B 1 AC 1 + B 2 AC 2 + B 3 AC 3 )/2 ( ) ( ) ( ) 1 0 0 1 1 0 B 1 =, C 0 1 1 0 1 = 0 1 B 2 = B 3 = ( 0 1 1 0 ( 0 1 1 0 ) ) ( ) 1 0, C 0 1 2 = ( ) 0 1, C 1 0 3 = with ), ( 0 1 1 0 ( ) ( ) 0 1 1 0, 1 0 0 1 ( ) ( ) 0 1 0 1. 1 0 1 0

Facts about ψ [Johnson, Laffey, Li, 1988], [Djokovic, 1990], [Li & Tsing, 1990], [Chang & Li, 1991], [Djokovic & Li, 1994]. 1. {I, B 1, B 2, B 3 } and {I, C 1, C 2, C 3 } are two basis for real quaternions in 4 4 matrices. 2. There is a 4-dimensional subspace W of M 4 such that ψ(a 1 + A 2 ) = A 1 A 2 A 1 W, A 2 W. 3. If A has singular values s 1 s 2 s 2 s 4, then ψ(a) has singular values (s 1 + s 2 + s 3 + δ A s 4 )/2, (s 1 + s 2 s 3 δ A s 4 )/2, (s 1 s 2 + s 3 δ A s 4 )/2, (s 1 s 2 s 3 δ A s 4 )/2, where δ A is the sign of det(a). So, ψ preserves the Ky Fan 2-norm.

4. The set of extreme point of the unit norm ball of the Fy Fan 2-norm on n n real matrices have three connected components: 1 2 O+ n = 1 2 {P : P t P = I n, det(p ) = 1}, 1 2 O n = 1 2 {P : P t P = I n, det(p ) = 1}, R = {xy t : x, y IR n, l 2 (x) = l 2 (y) = 1}. If n 4, φ(r) = R; if n = 4, the special map ψ permutes these components.

General distance preserving maps Characterize mappings φ : M m,n M m,n such that φ(a) φ(b) = A B for all A, B M m,n. By a result of Charzyński [1953], such a map is real affine. Theorem [Chan, Li, Sze, 2005] In the complex case, φ has the form A UAV + R, A UAV + R, or when m = n A UA t V + R, or A UA V + R.

A more natural question Characterize mappings φ : M m,n M m,n such that φ(a) + φ(b) = A + B for all A, B M m,n. Note that φ(a) + φ( A) = A A = 0. So, φ( A) = φ(a) and φ(a) φ(b) = φ(a) + φ( B) = A B. Consequently, φ is real linear.

Multiplicative maps and additional results [Cheung, Fallat, Li, 2002; Guralnick, Li, Rodman, 2003] Theorem Let be a USI norm, and let V = M n (k), SL n, or GL n. A mapping φ : V M n is multiplicative and satisfies if and only if φ has the form φ(a) = A for all A M n A f(det(a))u AU or A f(det(a))u AU. Key step Apply results on multiplicative maps on M (k) n, SL n, GL n by [Jodeit and Lam, 1969] and [Borel and Tits, 1978].

Theorem [Chan, Li, Sze, 2005] Let be a USI norm. A mapping φ : M (k) n has the form where M n satisfies AB = φ(a)φ(b) for all A, B M n A U h(a)u or A U h(a)u, h(x) = U X X = XV X for some unitary U X, V X. Suppose xy t is not always a constant multiple of s 1 (xy t ). Then φ has the form A µ A U AU or A µ A U AU, µ A = 1. Key step Determine mappings such that φ(a)φ(b) = 0 whenever AB = 0.

Further research Characterize φ : V V such that φ(a) φ(b) = A B for all A, B V, where is a certain operation on matrices such as A ± B, AB, AB 1, AB, the Jordan product (AB + BA)/2, the Lie product AB BA, the Jordan triple product ABA, etc. More generally, let F be a norm or other functions on V such as the spectrum, spectral radius, rank, numerical range, etc. Characterize φ : V V such that F (φ(a) φ(b)) = F (A B) for all A, B V.

What about other domains V and co-domains V? * [Li, Šemrl, Sourour, 2003] surjective isometries for Ky Fan k-norm between nest algebras in M n. * [Cheung, Li, Poon, 2004] (non-surjective) isometries for the operator norm between rectangular matrix spaces. * [Li, Poon, Sze, 2005] (non-surjective) isometries for the Ky Fan norms between rectangular matrix spaces. * [Sourour, 1981] surjective isometries for UI norms on symmetrically normed ideals. * [Anoussis and Katavolos, 1995] surjective isometries for Schatten p-norms on nest algebras. * [Arazy, Solel, Moore, Trent, etc. 1980-present] surjective isometries for the operator norm on non-self-adjoint algebras, nest algebras, reflexive operator algebras, CSL algebras, etc.

Comments, answers, are more questions are welcome! Thank you for your attention!