On Ozeki's inequality (Q1295947): Difference between revisions
From MaRDI portal
Set profile property. |
Set OpenAlex properties. |
||
Property / OpenAlex ID | |||
Property / OpenAlex ID: W2120848549 / rank | |||
Normal rank |
Latest revision as of 08:47, 30 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On Ozeki's inequality |
scientific article |
Statements
On Ozeki's inequality (English)
0 references
13 June 2000
0 references
The main result of this paper is the following additive converse to Cauchy's inequality. Let \(A\), \(B\) be two commuting selfadjoint operators (on a Hilbert space \(H\)) satisfying \(0\leq m_1\leq A\leq M_1\), \(0\leq m_2\leq B\leq M_2\) and \(M_1M_2\neq 0\). Put \(\alpha= m_1/M_1\) and \(\beta= m_2/M_2\). Then for every unit vector \(x\in H\), i) if \((1+\alpha)(1+ \beta)\leq 2\), then \[ \langle A^2x, x\rangle\langle B^2x,x\rangle- \langle ABx,x\rangle^2\leq {(M_1M_2- m_1m_2)^2\over \{4- (1+\alpha)(1+ \beta)\}(1+\alpha)(1+ \beta)}; \] ii) if \((1+\alpha)(1+\beta)\geq 2\), then \[ \langle A^2x, x\rangle\langle B^2x,x\rangle- \langle ABx,x\rangle^2\leq \textstyle{{1\over 4}}(M_1M_2- m_1m_2)^2. \] Corollary: Let \((a_1,\dots, a_n)\) and \((b_1,\dots, b_n)\) be two \(n\)-tuples of real numbers such that \(0\leq m_1\leq a_k\leq M_1\) and \(0\leq m_2\leq b_k\leq M_2\) \((k= 1,\dots, n)\). Then \[ (\sum a^2_k)(\sum b^2_k)- (\sum a_kb_k)^2\leq \Biggl[{n^2\over 3}\Biggr](M_1M_2- m_1m_2)^2. \] The results above admit a partial extension to the general case of two strictly positive operators, by replacing the product \(AB\) by the Kubo-Ando geometric mean of \(A^2\) and \(B^2\): Let \(A\) and \(B\) be two positive operators on \(H\) satisfying \(0< m_1\leq A\leq M_1\) and \(0< m_2\leq B\leq M_2\) respectively. Then for every unit vector \(x\in H\), \[ \langle A^2x,x\rangle\langle B^2x, x\rangle- \langle A^2\#_{1/2} B^2x,x\rangle^2\leq {1\over 4\gamma^2} (M_1M_2- m_1m_2)^2, \] where \(A^2\#_{1/2} B^2= A(A^{-1} B^2A^{-1})^{1/2}A\) and \(\gamma= \max\{m_1/M_1, m_2/M_2\}\).
0 references
covariance
0 references
additive converse to Cauchy's inequality
0 references
strictly positive operators
0 references
Kubo-Ando geometric mean
0 references