The Setting
Take the hyperbolicity cone in the space of triples consisting of two real symmetric matrices and a vector in . The question is whether is spectrahedral, or whether a normalized matrix pencil can represent its Schur-complement inequality. Along a unit vector with , wedge coordinates map the orthogonal complement of to , raising the question of whether this map is a linear isomorphism and whether a biquadratic form can match the squared Euclidean operator norm of a linear matrix family on that orthogonal complement.
The Hyperbolicity Cone and Normalized Pencils
For a nonnegative integer and a set , write for the square matrices over , and for nonnegative integers and , write for the real matrices. Denote by the -dimensional Euclidean space and by the subspace of real symmetric matrices, writing when is positive semidefinite, when it is positive definite, and for the identity matrix. The ambient space is the 23-dimensional product of triples , with base point .
Over a commutative ring and for , the quadratic matrix is the matrix over defined with cyclic indices by (1): (1) Given a commutative ring and , the Choi-Lam biquadratic form is the polynomial defined with cyclic indices by (2): (2)
Given a commutative ring and , the matrix map is the linear map sending with , , and to (3): (3)✦
Over a commutative ring , the determinantal formula for and is (4): (4)✦ The polynomial in the coordinates of is the real polynomial defined by applying (4) to the symmetric matrices determined by upper-triangular indeterminates for and for , together with indeterminates for .
At each point , the univariate restriction of along is the real polynomial .✦ The closed hyperbolicity cone of at is the subset of defined by (5): (5)
A normalized pencil consists of positive integers together with real linear maps , , and such that and are positive ( implies and for all ) and unital ( and ), and such that for all with and all , the Schur-complement equivalence (6) holds: (6)
Projections, Operator Norm, and Tangent Coordinates
For a nonnegative integer and a square matrix , the kernel projection is the matrix of the orthogonal projection of onto .✦ For a vector , the outer product is the symmetric matrix with -entry . For a normalized pencil with dimensions and , the left kernel projection is the matrix in defined by (7): (7)✦
The Euclidean operator norm of a real matrix (for nonnegative integers and ) is the operator norm of the linear map determined by .✦
Write and for the Euclidean inner product and norm on . At , the tangent space is the linear subspace of vectors orthogonal to , given by (8): (8)
For vectors and in with and , the wedge coordinates of and form the vector .✦ For any , the tangent coordinate map is the real linear map from to defined by (9): (9)✦
The Claim
Theorem 1. Let be a unit vector with . Then there exists a real linear isomorphism satisfying (10): (10)✦
Theorem 2. Let and be nonnegative integers, let be a real matrix, and let and . Then the entries of satisfy (11): (11)✦
Theorem 3. Let be a normalized pencil with dimensions , let be a unit vector with , and let be a real linear map. Then there exist and satisfying (12): (12)✦
Theorem 4. No normalized pencil exists (13): (13)✦
Theorem 5. The hyperbolicity cone is not spectrahedral: for every positive integer and every real linear map , (14): (14)✦
In the paper, the result reads: for with , (with standing for ) is a linear isomorphism, so (where and is its differential at ) has bilinear entries satisfying , and Lemma 2.1 forces and hence , contradicting for the first standard basis vector .
What Is Checked
The tangent coordinate map (9) and theorems (10) through (14) are stated over the tangent space and wedge coordinates in (assuming and in (10) and (12)), the Euclidean operator norm on real matrices in (11), normalized pencils with left kernel projection , Choi-Lam form , and real linear maps in (12) and (13), and the hyperbolicity cone of at together with real linear maps for positive integers in (14).