The Setting
Take two positive unital linear maps sending real symmetric matrices to real symmetric matrices of positive orders, together with a linear map from to rectangular matrices linked by a Schur-complement equivalence. A unit vector in acts through the difference between the identity matrix and its outer product to cut out orthogonal kernel projections in the two target spaces. For two unit vectors in and a parameter in , the problem relates a biquadratic form to the Euclidean operator norm of the projected rectangular matrix and determines when that projected block vanishes. The Euclidean operator norm of a positive semidefinite matrix equals its maximum eigenvalue, and inverse square roots along affine rays of positive semidefinite matrices converge to the orthogonal projection onto the kernel.
Definitions
For a nonnegative integer , real -dimensional Euclidean space is the space of real -tuples with the standard Euclidean inner product✦, and the space of real symmetric matrices of order is the subspace of real square matrices equal to their transpose.✦ Write when a real matrix is positive semidefinite and when is positive definite, and let denote the identity matrix. Over a commutative ring , write for the -tuples with entries in and for the matrices over (with the square matrices).
Let be a commutative ring, with indices in taken cyclically. For , the quadratic matrix is the matrix over with entry (1)✦ For , the Choi-Lam biquadratic form is the polynomial defined in terms of (1) by (2)✦
Given over a commutative ring , the matrix map acts on a matrix over written in block form with , , and by (3)✦
A normalized pencil with integer dimensions consists of linear maps , , and such that and are positive ( and for every with ) and unital ( and ), and the Schur-complement equivalence holds: for all with and all , with given by (3), (4)✦
Given a nonnegative integer and a vector , the outer product is the symmetric matrix with entry .✦
For any real square matrix (with ), its kernel projection is the matrix representing the orthogonal projection of onto .✦
For a normalized pencil with dimensions and vectors , the left kernel projection and the right kernel projection are (5)
The inverse square root of a symmetric matrix (for a nonnegative integer ), given an orthogonal matrix of eigenvectors of with eigenvalues , is the matrix (6)✦ where is the diagonal matrix with -th diagonal entry .
The Euclidean operator norm of a matrix (for nonnegative integers and ) is the norm of the linear map defined by .✦
For vectors and in with and , the wedge coordinates of and form the vector .✦
The Claim
Given an integer and a positive semidefinite matrix , write for its real spectrum; then (7)✦
Take a nonnegative integer and a positive semidefinite matrix . As in , the family defined by (6) satisfies (8)✦ in , and (9)✦
Suppose is a normalized pencil with dimensions , matrices satisfy and with and , and ; then (10)✦
Theorem 1. Let be a normalized pencil with dimensions . For every pair of unit vectors and in and every , (11)✦ For every unit vector and every , (12)✦
In the paper, the result reads: if , and are positive unital linear maps, and is linear such that for every , , and , where and stand for the paper’s block map and Choi-Lam form, and and stand for the paper’s kernel projections and , then for every pair of unit vectors and in and every , .
What Is Checked
The library states (7) over integers and positive semidefinite matrices ; (8) and (9) over nonnegative integers and positive semidefinite matrices ; (10) over a normalized pencil , matrices with , , , and , and ; (11) over a normalized pencil , unit vectors , and ; and (12) over a normalized pencil , a unit vector , and .