The Setting
Take a finite index set , two complex Hermitian matrices and indexed by , and a real function . When satisfies a Lipschitz bound with constant on the eigenvalues or spectra of and , the question is how the difference under the functional calculus is bounded in terms of in the Hilbert-Schmidt norm and in the norm on the matrix space.
Norms and Matrix Flattening
Let denote the space of complex matrices equipped with the norm , and write for the conjugate transpose of , given by . A Hermitian matrix is a matrix in that satisfies .
The squared Hilbert-Schmidt norm of a matrix is the real number given by (1): (1)✦ The Hilbert-Schmidt norm of a matrix is the nonnegative square root of .
For a Hermitian matrix , write for its -indexed family of real eigenvalues and for its real spectrum. For a Hermitian matrix and a real function , denote by the functional calculus of at .
Let be the complex Euclidean space of arrays equipped with the norm. The matrix flattening map is the continuous -linear map assigning to each the array .✦ The matrix unflattening map is the continuous -linear map assigning to each the matrix .✦ Write and for the operator norms of and .
The Claim
Theorem 1. Let be Hermitian matrices, let be a function, and let be a real number. If for all , then the squared Hilbert-Schmidt norm satisfies (2): (2)✦ If for all and , then the norm on satisfies (3): (3)✦
In the paper, the result reads: if a real function (written in the paper) is -Lipschitz on an interval containing the spectra of two Hermitian matrices and , then .
What Is Checked
The theorem is stated over Hermitian matrices indexed by a finite set , a real function and a constant satisfying the pointwise Lipschitz condition on pairs of eigenvalues or points of the real spectra, the functional calculus on , the squared Hilbert-Schmidt norm, and the norm on together with the continuous -linear flattening and unflattening maps and .
Writing for the trace on , the library also states the trace identity (4) for every : (4)✦ For every and all unitary matrices , the squared Hilbert-Schmidt norm satisfies unitary invariance (5): (5)✦