Comparison of norms \(|||f(A)-f(B)|||\) and \(|||f(|A-B|)|||\) (Q1822318)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Comparison of norms \(|||f(A)-f(B)|||\) and \(|||f(|A-B|)|||\) |
scientific article |
Statements
Comparison of norms \(|||f(A)-f(B)|||\) and \(|||f(|A-B|)|||\) (English)
0 references
1988
0 references
Let A, B be positive semi-definite matrices, and f(\(\lambda)\) a continuous real-valued function on \([0,\infty)\). Our concern is a comparison of \(\|| f(A)-f(B)\||\) with \(\|| f(| A- B|)\||\) for a unitarily invariant norm \(\|| \cdot \||\). Trace norm, Hilbert-Schmidt norm and operator norm are among typical examples of unitarily invariant norm. A function \(f(\lambda)\) is said to be operator monotone if \(A\geq B\geq 0\) always implies \(f(A)\geq f(B)\). Here \(\geq\) refers to the order relation, induced by the cone of positive semi- definite matrices. Our main result is that if f(\(\lambda)\) is non- negative operator monotone then \[ \|| f(A)-f(B)\|| \leq \|| f(| A-B|)\|| \] and that if the inverse function of g(\(\lambda)\) is operator monotone then \[ \|| g(A)-g(B)\|| \geq \|| g(| A-B|)\||. \] Specialization yields more practical inequalities; \[ (1)\quad \|| A^ p-B^ p\|| \leq | | A- B|^ p\|| \text{ for } 0<p\leq 1, \] \[ (2)\quad \|| A^ p- B^ p\|| \geq \|| | A-B|^ p\|| \text{ for } p\geq 1, \] \[ (3)\quad \|| A^ p\cdot \log (A+I)-B^ p\cdot \log (B+I)\|| \geq \|| | A-B|^ p\cdot \log (| A-B| +I)\|| \text{ for } p\geq 1, \] \[ (4)\quad \|| \log (A+I)-\log (B+I)\|| \leq \|| \log (| A-B| +I)\|| \] \[ (5)\quad \|| e^ A-e^ B\|| \geq \|| e^{| A-B|}-I\||. \]
0 references
norm inequalities
0 references
perturbation
0 references
operator monotone function
0 references
functional calculus
0 references
positive semi-definite matrices
0 references
unitarily invariant norm
0 references
Trace norm
0 references
Hilbert-Schmidt norm
0 references
operator norm
0 references