- Research
- Open access
- Published:
Reverses of Ando’s and Hölder–McCarty’s inequalities
Journal of Inequalities and Applications volume 2018, Article number: 164 (2018)
Abstract
In this paper, we give some reverse-types of Ando’s and Hölder–McCarthy’s inequalities for positive linear maps, and positive invertible operators. For this purpose, we use a recently improved Young inequality and its reverse.
1 Introduction and preliminaries
Let \(\mathcal{B}(\mathcal{H})\) be the \(C^{*}\)-algebra of all bounded linear operators on a complex Hilbert space \(\mathcal{H}\) with the operator norm \(\Vert \cdot \Vert \) and the identity I; also \(\mathcal{M} _{n}(\mathcal{C})\) denotes the space of all \(n\times n\) complex matrices. For an operator \(A \in \mathcal{B}(\mathcal{H})\), we write \(A\geq 0\) if A is positive, and \(A>0\) if A is positive invertible. For \(A, B \in \mathcal{B}(\mathcal{H})\), we say \(A\geq B\) if \(A-B\geq 0\). The Gelfand map \(f(t)\mapsto f(A)\) is an isometrical ∗-isomorphism between the \(C^{*}\)-algebra \(C(\operatorname {sp}(A))\) of continuous functions on the spectrum \(\operatorname {sp}(A)\) of a selfadjoint operator A and the \(C^{*}\)-algebra generated by A and I. A linear map Φ on \(\mathcal{B}(\mathcal{H})\) is positive if \(\Phi (A) \geq 0\) whenever \(A\geq 0\). It is said to be unital if \(\Phi (I)=I\). A continuous function \(f:J\rightarrow \mathcal{R}\) is operator concave if
for all selfadjoint operators \(A, B\in \mathcal{B}(\mathcal{H})\) with spectra in J and all \(\alpha \in [0,1]\).
The well-known Young inequality says that, for positive real numbers a, b and \(0\leq t \leq 1\), we have \(a^{t}b^{1-t}\leq t a+(1-t)b\). Refinements and reverses of this inequality are proven in [2, 9, 14–16] and the references therein. Also Kittaneh et al. in [10] obtained the following improvement of the Young inequality for any positive definite matrices \(A,B\in \mathcal{M}_{n}(\mathcal{C})\):
where \(t\in [0,1]\), \(r=\min \{t,1-t\}\) and \(R=\max \{t,1-t\}\).
Zhao et al. [19] obtained a refinement of Young’s inequality and its reverse as follows:
-
(i)
for \(0< t\leq \frac{1}{2}\),
$$\begin{aligned} &r_{0}\bigl(\sqrt[4]{ab}-\sqrt{a}\bigr)^{2}+r( \sqrt{a}-\sqrt{b})^{2} +a ^{1-t}b^{t} \\ &\quad \leq (1-t)a+tb \\ &\quad \leq R(\sqrt{a}-\sqrt{b})^{2}-r_{0}\bigl(\sqrt[4]{ab}- \sqrt{b}\bigr)^{2}+a ^{1-t}b^{t}; \end{aligned}$$(2) -
(ii)
for \(\frac{1}{2}< t<1\),
$$\begin{aligned} &r_{0}\bigl(\sqrt[4]{ab}-\sqrt{b}\bigr)^{2}+R(\sqrt{a}- \sqrt{b})^{2} +a ^{1-t}b^{t}\\ &\quad \leq (1-t)a+tb \\ &\quad \leq r(\sqrt{a}-\sqrt{b})^{2}-r_{0}\bigl(\sqrt[4]{ab}- \sqrt{a}\bigr)^{2}+a ^{1-t}b^{t}, \end{aligned}$$
where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\) and \(r_{0}=\min \{2r,1-2r \}\).
Sababheh et al. [15, 16] established some refinements and reverses of Young’s inequality as follows:
-
(i)
for \(0\leq t\leq \frac{1}{2}\),
$$\begin{aligned} S_{N}(t;a,b)\leq t a+(1-t)b-a^{t} b^{1-t}\leq (1-t) (\sqrt{a}- \sqrt{b})^{2}-S_{N}(2t; \sqrt{ab},a); \end{aligned}$$(3) -
(ii)
for \(\frac{1}{2}\leq t\leq 1\),
$$\begin{aligned} S_{N}(t;a,b)\leq t a+(1-t)b-a^{t} b^{1-t}\leq t( \sqrt{a}-\sqrt{b})^{2}-S _{N}(2-2t;\sqrt{ab},b), \end{aligned}$$
where
\(s_{j}(t)= ((-1)^{r_{j}}2^{j-1}t+(-1)^{r_{j}+1}[ \frac{r_{j}+1}{2}] )\), \(r_{j}=[2^{j}t]\) and \(k_{j}=[2^{j-1}t]\). Here \([x]\) is the greatest integer less than or equal to x.
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive. The operator t-weighted arithmetic, geometric, and harmonic means of operators A, B are defined by \(A\nabla_{t}B=(1-t)A+tB\), \(A\sharp_{t}B=A^{ \frac{1}{2}}(A^{-\frac{1}{2}}B A^{-\frac{1}{2}})^{t}A^{\frac{1}{2}}\) and \(A!_{t}B=((1-t)A^{-1}+tB^{-1})^{-1}\), respectively. In particular, for \(t =\frac{1}{2}\) we get the usual operator arithmetic mean ∇, the geometric mean ♯ and the harmonic mean !.
2 Results and discussion
For positive real numbers \(a_{i}\) and \(b_{i} \) (\(i=1,2,\ldots,n\)) the Hölder inequality states that
for \(p,q>1\) such that \(\frac{1}{p}+\frac{1}{q}=1\). If \(p=q=2\) in (4), then we get the Cauchy–Schwarz inequality. The Hölder inequality for positive operators \(A_{i}\) and \(B_{i} \) (\(i=1,2,\ldots,n\)) is
where \(0\leq t\leq 1\). In the case \(t=\frac{1}{2}\), we get the operator Cauchy–Schwarz inequality. For further information as regards the Hölder and Cauchy–Schwarz inequalities we refer the reader to [3–5, 11, 12, 20] and the references therein. Ando [1] proved that if Φ is a positive linear map, then for positive operators \(A, B\in \mathcal{B}(\mathcal{H})\) and \(t\in [0,1]\), we have
Recently, some authors presented several reverse-types of Ando’s inequality (see [13, 17]).
The Hölder–McCarthy’s inequality says that for any positive operator A and any unit vector \(x\in \mathcal{H}\), we have
Furuta [8] showed that this inequality is equivalent to Young’s inequality.
3 Conclusions
In this paper, we establish a reverse of Ando’s inequality for positive (non-unital) linear maps and positive definite matrices by using an inequality due to Sababheh. We obtain some reverses of the matrix Hölder and Cauchy–Schwarz inequalities and a reverse of inequality (5) for \(t\in (0,\frac{1}{2}]\) as follows:
5 Main results
To prove our first result, we need the following lemmas.
Lemma 1
([16])
Let \(A, B\in \mathcal{M}_{n}(\mathcal{C})\) be positive definite matrices and \(t\in [0,1]\). Then
where \(\alpha_{j}(t)=\frac{k_{j}(t)}{2^{j-1}}\).
For \(N=2\), we have the following lemma, which is shown in [19] for positive invertible operators.
Lemma 2
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible operators and \(t\in [0,1]\).
-
(i)
If \(0< t\leq \frac{1}{2}\), then
$$\begin{aligned} r_{0}(A\sharp B-2A\sharp_{\frac{1}{4}}B+A)+2t(A\nabla B-A\sharp B)+A \sharp_{t}B\leq A\nabla_{t}B. \end{aligned}$$(7) -
(ii)
If \(\frac{1}{2}< t<1\), then
$$\begin{aligned} r_{0}(A\sharp B-2A\sharp_{\frac{3}{4}}B+B)+2(1-t) (A \nabla B-A\sharp B)+A \sharp_{t}B\leq A\nabla_{t}B, \end{aligned}$$(8)where \(r=\min \{\nu ,1-\nu \}\) and \(r_{0}=\min \{2r,1-2r\}\).
Lemma 3
([16])
Let \(A, B\in \mathcal{M}_{n}(\mathcal{C})\) be positive definite matrices and \(t\in [0,1]\).
-
(i)
If \(0\leq t\leq \frac{1}{2}\), then
$$\begin{aligned} A\nabla_{t}B&\leq A\sharp_{t}B+2(1-t) (A \nabla B-A\sharp B) \\ &\quad {} -\sum_{j=1}^{N}s_{j}(2t) ( A \sharp_{1-\beta_{j}(t)}B+A \sharp_{1+2^{-j}-\beta_{j}(t)}B-2A\sharp_{1-2^{-j-1}-\beta_{j}(t)}B ) . \end{aligned}$$(9) -
(ii)
If \(\frac{1}{2}\leq t\leq 1\), then
$$\begin{aligned} A\nabla_{t}B &\leq A\sharp_{t}B+2t(A\nabla B-A\sharp B) \\ & \quad {} -\sum_{j=1}^{N}s_{j}(2-2t) ( A \sharp_{\gamma_{j}(t)}B+A \sharp_{\gamma_{j}(t)+2^{1-j}}B-2A\sharp_{\gamma_{j}(t)+2^{-j}}B ) , \end{aligned}$$where \(\beta_{j}(t)=2^{-j}k_{j}(2t)\) and \(\gamma_{j}(t)=2^{1-j}k _{j}(2-2t)\).
Remark 4
By using functional calculus and numerical inequalities in [10, 16], we can extend inequality (1), Lemmas 1 and 3 for positive invertible operators.
For \(N=2\), we have the following lemma, which is shown in [19] for positive invertible operators.
Lemma 5
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible operators and \(t\in [0,1]\).
-
(i)
If \(0< t\leq \frac{1}{2}\), then
$$\begin{aligned} A\nabla_{t}B \leq A\sharp_{t}B+2(1-t) (A\nabla B-A\sharp B)-r_{0}(A \sharp B-2A\sharp_{\frac{3}{4}}B+B). \end{aligned}$$(10) -
(ii)
If \(\frac{1}{2}< t<1\), then
$$\begin{aligned} A\nabla_{t}B \leq A\sharp_{t}B+2t(A\nabla B-A \sharp B)-r_{0}(A\sharp B-2A\sharp_{\frac{1}{4}}B+A), \end{aligned}$$(11)where \(r=\min \{\nu ,1-\nu \}\) and \(r_{0}=\min \{2r,1-2r\}\).
Now, we obtain a reverse of Ando’s inequality for positive invertible operators as follows.
Theorem 6
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible operators, Φ be a positive linear map and \(t\in [0,1]\).
-
(i)
If \(0\leq t\leq \frac{1}{2}\), then
$$\begin{aligned} & \Phi (A) \sharp_{t}\Phi (B)-\Phi (A \sharp_{t}B) \\ &\quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+ \frac{1}{2}\bigl(\Phi (A)+\Phi (B)-2\Phi (A) \sharp \Phi (B)\bigr)\biggr) \\ &\quad \quad {} -\sum_{j=1}^{N}s_{j}(2t) \bigl( \Phi (A\sharp_{1-\beta_{j}(t)}B)+ \Phi (A\sharp_{1+2^{-j}-\beta_{j}(t)}B)- 2\Phi (A \sharp_{1-2^{-j-1}-\beta_{j}(t)}B) \bigr) \\ &\quad \quad {} -\sum_{j=1}^{N}s_{j}(t) \bigl( \Phi (A)\sharp_{\alpha_{j}(t)}\Phi (B)+ \Phi (A)\sharp_{2^{1-j}+\alpha_{j}(t)}\Phi (B) \\ &\quad \quad {}-2\Phi (A) \sharp_{2^{-j}+\alpha_{j}(t)}\Phi (B) \bigr) . \end{aligned}$$(12) -
(ii)
If \(\frac{1}{2}\leq t\leq 1\), then
$$\begin{aligned} &\Phi (A) \sharp_{t}\Phi (B)-\Phi (A \sharp_{t}B) \\ &\quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+ \frac{1}{2}\bigl(\Phi (A)+\Phi (B)-2\Phi (A) \sharp \Phi (B)\bigr)\biggr) \\ &\quad \quad {} -\sum_{j=1}^{N}s_{j}(2-2t) \bigl( \Phi (A\sharp_{\gamma_{j}(t)}B)+ \Phi (A\sharp_{\gamma_{j}(t)+2^{1-j}}B)-2\Phi (A \sharp_{\gamma_{j}(t)+2^{-j}}B) \bigr) \\ &\quad \quad {} -\sum_{j=1}^{N}s_{j}(t) \bigl( \Phi (A)\sharp_{\alpha_{j}(t)}\Phi (B)+ \Phi (A)\sharp_{2^{1-j}+\alpha_{j}(t)}\Phi (B) \\ &\quad \quad {}-2\Phi (A) \sharp_{2^{-j}+\alpha_{j}(t)}\Phi (B) \bigr), \end{aligned}$$(13)where \(\alpha_{j}(t)=\frac{k_{j}(t)}{2^{j-1}}\), \(\beta_{j}(t)=2^{-j}k _{j}(2t)\), \(\gamma_{j}(t)=2^{1-j}k_{j}(2-2t)\) and \(R=\max \{t,1-t\}\).
Proof
The proof of inequality (13) is similar to the proof of inequality (12). Thus, we only prove inequality (12).
Let \(0\leq t\leq \frac{1}{2}\). Applying inequalities (10) and (9), we have
Now, using the positive linear map Φ on (14), we get
Moreover, if we replace A and B by \(\Phi (A)\) and \(\Phi (B)\) in inequality (14), respectively, then
From the first inequality of (16) and the second inequality of (15), we have
which implies that
Therefore, applying inequality (1), we get
□
Similarly for \(N=2\) by applying Lemma 2 and Lemma 5, we can obtain a reverse of Ando’s inequality for positive invertible operators.
Corollary 7
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible operators, Φ be a positive linear map and \(t\in [0,1]\).
-
(i)
If \(0< t\leq \frac{1}{2}\), then
$$\begin{aligned} & \Phi (A)\sharp_{t}\Phi (B)-\Phi (A\sharp_{t}B) \\ &\quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+\frac{1}{2}\bigl( \Phi (A)+\Phi (B)-2\Phi (A) \sharp \Phi (B)\bigr)\biggr) \\ &\quad \quad {} -r_{0}\bigl(\Phi (A\sharp B)+\Phi (B)-2\Phi (A\sharp_{\frac{3}{4}}B) \bigr) \\ &\quad \quad {} -r_{0}\bigl(\Phi (A)\sharp \Phi (B)+\Phi (A)-2\bigl(\Phi (A) \sharp_{ \frac{1}{4}}\Phi (B)\bigr)\bigr) \\ & \quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+\frac{1}{2}\bigl( \Phi (A)+ \Phi (B)-2\Phi (A)\sharp \Phi (B)\bigr)\biggr); \end{aligned}$$(17) -
(ii)
if \(\frac{1}{2}< t< 1\), then
$$\begin{aligned} &\Phi (A)\sharp_{t}\Phi (B)-\Phi (A\sharp_{t}B) \\ &\quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+\frac{1}{2}\bigl( \Phi (A)+\Phi (B)-2\Phi (A) \sharp \Phi (B)\bigr)\biggr) \\ &\quad \quad {} -r_{0}\bigl(\Phi (A)\sharp \Phi (B)+\Phi (B)-2\bigl(\Phi (A) \sharp_{ \frac{3}{4}}\Phi (B)\bigr)\bigr) \\ &\quad \quad {} -r_{0}\bigl(\Phi (A\sharp B)+\Phi (A)-2\Phi (A\sharp_{\frac{1}{4}} B)\bigr) \\ &\quad \leq 2R\biggl(\Phi (A)\sharp \Phi (B)-\Phi (A\sharp B)+\frac{1}{2}\bigl( \Phi (A)+ \Phi (B)-2\Phi (A)\sharp \Phi (B)\bigr)\biggr), \end{aligned}$$(18)where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\) and \(r_{0}=\min \{2r,1-2r \}\).
We want to establish some inequalities for positive invertible operators.
Theorem 8
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible. If \(t\in [0,1]\) and Φ, Ψ are two unital positive linear maps, then for any unit vector \(x\in \mathcal{H}\)
-
(i)
for \(0< t\leq \frac{1}{2}\),
$$\begin{aligned} &2r \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle \nabla \bigl\langle \Psi (B)x,x\bigr\rangle -\bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x \bigr\rangle \bigl\langle \Psi \bigl(B^{1/2}\bigr)x,x \bigr\rangle \bigr) \\ &\quad\quad {} + r_{0} \bigl( \bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x \bigr\rangle \bigr)\bigl\langle \Psi \bigl(B^{\frac{1}{2}}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{\frac{3}{4}} \bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{\frac{1}{4}}\bigr)x,x \bigr\rangle \bigr) \\ &\quad \leq (1-t)\bigl\langle \Phi (A)x,x\bigr\rangle +t\bigl\langle \Psi (B)x,x\bigr\rangle - \bigl\langle \Psi \bigl(B^{t}\bigr)x,x\bigr\rangle \bigl\langle \Phi \bigl(A^{1-t}\bigr)x,x\bigr\rangle \\ &\quad \leq 2R \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle \nabla \bigl\langle \Psi (B)x,x \bigr\rangle -\bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{ \frac{1}{2}}\bigr)x,x\bigr\rangle \bigr) \\ &\quad\quad {} -r_{0} \bigl( \bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B ^{\frac{1}{2}}\bigr)x,x\bigr\rangle +\bigl\langle \Psi (B)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{\frac{1}{4}} \bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{\frac{3}{4}}\bigr)x,x \bigr\rangle \bigr) ; \end{aligned}$$(19) -
(ii)
for \(\frac{1}{2}< t<1\),
$$\begin{aligned} &R \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle +\bigl\langle \Psi (B)x,x\bigr\rangle -2 \bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{1/2}\bigr)x,x \bigr\rangle \bigr) \\ & \quad\quad {} + r_{0} \bigl( \bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x \bigr\rangle \bigr)\bigl\langle \Psi \bigl(B^{\frac{1}{2}}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{\frac{1}{4}} \bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{\frac{3}{4}}\bigr)x,x \bigr\rangle \bigr) \\ & \quad \leq (1-t)\bigl\langle \Phi (A)x,x\bigr\rangle +t\bigl\langle \Psi (B)x,x\bigr\rangle - \bigl\langle \Psi \bigl(B^{t}\bigr)x,x\bigr\rangle \bigl\langle \Phi \bigl(A^{1-t}\bigr)x,x\bigr\rangle \\ & \quad \leq r \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle +\bigl\langle \Psi (B)x,x \bigr\rangle -2\bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{ \frac{1}{2}}\bigr)x,x\bigr\rangle \bigr) \\ &\quad\quad {} -r_{0} \bigl( \bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B ^{\frac{1}{2}}\bigr)x,x\bigr\rangle +\bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{\frac{3}{4}} \bigr)x,x\bigr\rangle \bigl\langle \Psi \bigl(B^{\frac{1}{4}}\bigr)x,x \bigr\rangle \bigr) , \end{aligned}$$
where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\), \(r_{0}=\min \{2r,1-2r\}\).
Proof
The proof of part (ii) is similar to the proof of part (i). Thus we only prove (i).
Applying inequality (2) for any positive real numbers k, s, we have
Fix s in (20). Then applying functional calculus to the operator A, we have
If we apply the positive linear map Φ and inner product for \(x\in {\mathcal{H}}\) with \(\Vert x \Vert =1\) in inequality (21), we have
Now, using the functional calculus to the operator B, we have
Taking the positive linear map Ψ and the inner product for \(y\in {\mathcal{H}}\) with \(\Vert y \Vert =1\), we get
Now, if we put \(x=y\), then we get the desired result. □
Theorem 9
Let \(A, B\in \mathcal{B}(\mathcal{H})\) be positive invertible. If \(t\in [0,1]\) and Φ, Ψ are two unital positive linear maps, then for any unit vector \(x\in \mathcal{H}\)
-
(i)
for \(0< t\leq \frac{1}{2}\),
$$\begin{aligned} &2r \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle \nabla \bigl\langle \Psi (B)x,x \bigr\rangle -\bigl\langle \Phi \bigl(A^{\frac{1}{2}}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/2} \bigr) \\ & \quad\quad {} + r_{0} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x \bigr\rangle \bigl\langle \Psi (B)x,x \bigr\rangle ^{1/2}+\bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{3/4}\bigr)x,x \bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/4} \bigr) \\ &\quad \leq (1-t)\bigl\langle \Phi (A)x,x\bigr\rangle +t\bigl\langle \Psi (B)x,x\bigr\rangle - \bigl\langle \Psi (B)x,x\bigr\rangle ^{t}\bigl\langle \Phi \bigl(A^{1-t}\bigr)x,x\bigr\rangle \\ & \quad \leq 2R \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle \nabla \bigl\langle \Psi (B)x,x \bigr\rangle -\bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/2} \bigr) \\ & \quad\quad {} -r_{0} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x \bigr\rangle ^{1/2}+\bigl\langle \Psi (B)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{1/4}\bigr)x,x \bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{3/4} \bigr) ; \end{aligned}$$ -
(ii)
for \(\frac{1}{2}< t<1\),
$$\begin{aligned} &R \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle +\bigl\langle \Psi (B)x,x\bigr\rangle -2 \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/2} \bigr) \\ &\quad\quad {} + r_{0} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x \bigr\rangle \bigr)\bigl\langle \Psi (B)x,x \bigr\rangle ^{1/2}+\bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{1/4} \bigr)x,x \bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{3/4} \bigr) \\ & \quad \leq (1-t)\bigl\langle \Phi (A)x,x\bigr\rangle +t\bigl\langle \Psi (B)x,x\bigr\rangle - \bigl\langle \Psi (B)x,x\bigr\rangle ^{t}\bigl\langle \Phi \bigl(A^{1-t}\bigr)x,x\bigr\rangle \\ & \quad \leq r \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle +\bigl\langle \Psi (B)x,x \bigr\rangle -2\bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/2} \bigr) \\ &\quad\quad {} -r_{0} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle \bigl\langle \Psi (B)x,x \bigr\rangle ^{1/2}+\bigl\langle \Phi (A)x,x\bigr\rangle -2\bigl\langle \Phi \bigl(A^{3/4}\bigr)x,x \bigr\rangle \bigl\langle \Psi (B)x,x\bigr\rangle ^{1/4} \bigr) , \end{aligned}$$
where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\), \(r_{0}=\min \{2r,1-2r \}\).
Proof
The proof of part (ii) is similar to the proof of part (i). Thus we just prove (i). For any positive real number k and any unit vector \(x\in \mathcal{H}\), we have
Applying inequality (22) and the functional calculus for the operator A, we have
Now, using the unital positive operator Φ and the inner product for \(y\in \mathcal{H}\) with \(\Vert y \Vert =1\) in inequality (23), we get
Now, putting \(y=x\), we get the desired result. □
Corollary 10
Let \(A\in \mathcal{B}(\mathcal{H})\) be positive, Φ be a unital positive linear map and \(t\in [0,1]\). Then for any unit vector \(x\in \mathcal{H}\)
-
(i)
for \(0< t\leq \frac{1}{2}\),
$$\begin{aligned} &2r\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi (A)x,x \bigr\rangle ^{\frac{1}{2}}-\bigl\langle \Phi \bigl(A^{\frac{1}{2}} \bigr)x,x\bigr\rangle \bigr) \\ & \quad\quad {} + r_{0}\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/2} \\ & \quad\quad {}-2\bigl\langle \Phi \bigl(A^{3/4} \bigr)x,x\bigr\rangle \bigl\langle \Phi (A)x,x\bigr\rangle ^{-1/4} \bigr) \\ & \quad \leq \bigl\langle \Phi (A)x,x\bigr\rangle ^{t}-\bigl\langle \Phi \bigl(A^{t}\bigr)x,x\bigr\rangle \\ & \quad \leq 2R\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle ^{\frac{1}{2}}-\bigl\langle \Phi \bigl(A^{\frac{1}{2}} \bigr)x,x \bigr\rangle \bigr) \\ & \quad\quad {} - r_{0}\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/2} \\ & \quad\quad {}-2\bigl\langle \Phi \bigl(A^{1/4} \bigr)x,x\bigr\rangle \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/4} \bigr) ; \end{aligned}$$ -
(ii)
for \(\frac{1}{2}< t<1\),
$$\begin{aligned} &2R\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi (A)x,x \bigr\rangle ^{\frac{1}{2}}-\bigl\langle \Phi \bigl(A^{\frac{1}{2}} \bigr)x,x\bigr\rangle \bigr) \\ & \quad\quad {} + r_{0}\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/2} \\ & \quad\quad {}-2\bigl\langle \Phi \bigl(A^{1/4} \bigr)x,x\bigr\rangle \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/4} \bigr) \\ & \quad \leq \bigl\langle \Phi (A)x,x\bigr\rangle ^{t}-\bigl\langle \Phi \bigl(A^{t}\bigr)x,x\bigr\rangle \\ & \quad \leq 2r\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi (A)x,x\bigr\rangle ^{\frac{1}{2}}-\bigl\langle \Phi \bigl(A^{\frac{1}{2}} \bigr)x,x \bigr\rangle \bigr) \\ & \quad\quad {} - r_{0}\bigl\langle \Phi (A)x,x\bigr\rangle ^{t-\frac{1}{2}} \bigl( \bigl\langle \Phi \bigl(A^{1/2}\bigr)x,x\bigr\rangle + \bigl\langle \Phi (A)x,x\bigr\rangle ^{1/2}\\ & \quad\quad {}-2\bigl\langle \Phi \bigl(A^{3/4} \bigr)x,x\bigr\rangle \bigl\langle \Phi (A)x,x\bigr\rangle ^{-1/4} \bigr) , \end{aligned}$$
where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\), \(r_{0}=\min \{2r,1-2r \}\).
Proof
Letting \(\Psi =\Phi \) and \(B=A\) in Theorem 9, we get the desired inequalities. □
In the next result, we obtain a refinement of inequality (5) for \(t\in (0,\frac{1}{2}]\).
Corollary 11
Let \(T\in {\mathcal{B}}(\mathcal{H})\) be positive operator and \(x\in {\mathcal{H}}\) be a unit vector. Then, for \(0< t\leq \frac{1}{2}\), we have
where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\), \(r_{0}=\min \{2r,1-2r \}\).
Proof
If we replace \(\Phi (A)=A\), \(A\in \mathcal{B}(\mathcal{H})\) and t with \(1-t\) in Corollary 10, then we get the desired result. □
6 Some new results
In this section, we prove some difference reverse-types of the Hölder and Cauchy–Schwarz inequalities.
Theorem 12
Let \(A_{i}, B_{i}\in \mathcal{B}(\mathcal{H})\) (\(1\leq i \leq n\)) be positive invertible and \(t\in [0,1]\).
-
(i)
If \(0< t\leq \frac{1}{2}\), then
$$\begin{aligned} & \Biggl( \sum_{i=1}^{n}A_{i} \Biggr) \sharp_{t} \Biggl( \sum_{i=1}^{n}B_{i} \Biggr) - \Biggl( \sum_{i=1}^{n}A_{i} \sharp_{t}B_{i} \Biggr) \\ &\quad \leq R \Biggl( \sum_{i=1}^{n}A_{i}+ \sum_{i=1}^{n}B_{i}-2\sum _{i=1}^{n}(A _{i}\sharp B_{i}) \Biggr) \\ &\quad\quad {}-r_{0} \Biggl( \sum _{i=1}^{n}(A_{i}\sharp B_{i})+ \sum_{i=1}^{n}B_{i}-2\sum _{i=1}^{n}(A_{i} \sharp_{\frac{3}{4}}B_{i}) \Biggr) \\ &\quad\quad {} -r_{0} \Biggl( \sum_{i=1}^{n}A_{i} \sharp \sum_{i=1}^{n}B_{i}+\sum _{i=1} ^{n}A_{i}-2 \Biggl( \sum _{i=1}^{n}A_{i} \sharp_{\frac{1}{4}}\sum_{i=1} ^{n}B_{i} \Biggr) \Biggr) . \end{aligned}$$ -
(ii)
If \(\frac{1}{2}< t< 1\), then
$$\begin{aligned} & \Biggl( \sum_{i=1}^{n}A_{i} \Biggr) \sharp_{t} \Biggl( \sum_{i=1}^{n}B_{i} \Biggr) - \Biggl(\sum_{i=1}^{n}A_{i} \sharp_{t}B_{i} \Biggr)\\ &\quad \leq R \Biggl( \sum _{i=1} ^{n}A_{i}+\sum _{i=1}^{n}B_{i}-2 \Biggl( \sum _{i=1}^{n}A_{i}\sharp B_{i} \Biggr) \Biggr) \\ &\quad \quad {} -r_{0} \Biggl( \Biggl(\sum_{i=1}^{n}A_{i} \Biggr)\sharp \Biggl(\sum_{i=1}^{n}B_{i} \Biggr)+\sum_{i=1}^{n}B_{i}-2 \Biggl( \Biggl(\sum_{i=1}^{n}A_{i} \Biggr)\sharp_{\frac{3}{4}}\Biggl( \sum_{i=1}^{n}B_{i} \Biggr) \Biggr) \Biggr) \\ & \quad \quad {} -r_{0} \Biggl( \sum_{i=1}^{n}(A_{i} \sharp B_{i})+\sum_{i=1}^{n}A_{i}-2 \sum_{i=1}^{n}(A_{i} \sharp_{\frac{1}{4}} B_{i}) \Biggr) , \end{aligned}$$where \(r=\min \{t,1-t\}\), \(R=\max \{t,1-t\}\) and \(r_{0}=\min \{2r,1-2r \}\).
Proof
Taking \(A=\operatorname {diag}(A_{1},\ldots,A_{n})\), \(B=\operatorname {diag}(B_{1},\ldots,B_{n})\) and \(\Phi ( [C_{ij}]_{1\leq i,j\leq n} ) = \sum_{i=1}^{n}C_{ii}\) in equalities (17) and (18), we get the desired inequality. □
Since the function \(f(x)=x^{t}\) (\(t\in [0,1]\)) is an operator concave function, \(\sum_{i=1}^{n}w_{i}T_{i}^{t}\leq (\sum_{i=1}^{n}w_{i}T _{i} )^{t}\) for positive operators \(T_{i}\) and positive real numbers \(w_{i}\) such that \(\sum_{i=1}^{n}w_{i}=1\). Now, Theorem 12 yields a reverse of this inequality as follows.
Example 13
If for positive operators \(T_{i}\) (\(1\leq i \leq n\)), we take \(A_{i}=w_{i}I\) and \(B_{i}=w_{i}T_{i}\) (\(1\leq i \leq n\)), in Theorem 12, where \(w_{i}\)’s are positive real numbers such that \(\sum_{i=1}^{n}w_{i}=1\), we obtain the following inequalities:
-
(i)
If \(0\leq t\leq \frac{1}{2}\), then
$$\begin{aligned} \Biggl(\sum_{i=1}^{n}w_{i}T_{i} \Biggr)^{t}-\sum_{i=1}^{n}w_{i}T_{i}^{t} &\leq R \Biggl(I+\sum_{i=1}^{n}w_{i}T_{i}-2 \sum_{i=1}^{n}w_{i}T_{i}^{1/2} \Biggr) \\ & \quad {} -r_{0} \Biggl(\sum_{i=1}^{n}w_{i}T_{i}^{1/2}+ \sum_{i=1}^{n}w_{i}T_{i}-2 \sum_{i=1}^{n}w_{i}T_{i}^{3/4} \Biggr) \\ & \quad {} -r_{0} \Biggl(\Biggl(\sum_{i=1}^{n}w_{i}T_{i} \Biggr)^{1/2}+I-2\Biggl(\sum_{i=1}^{n}w_{i}T _{i}\Biggr)^{1/4} \Biggr). \end{aligned}$$ -
(ii)
If \(\frac{1}{2}< t\leq 1\), then
$$\begin{aligned} \Biggl(\sum_{i=1}^{n}w_{i}T_{i} \Biggr)^{t}-\sum_{i=1}^{n}w_{i}T_{i}^{t} &\leq R \Biggl(I+\sum_{i=1}^{n}w_{i}T_{i}-2 \sum_{i=1}^{n}w_{i}T_{i}^{1/2} \Biggr) \\ &\quad {} -r_{0} \Biggl(\Biggl(\sum_{i=1}^{n}w_{i}T_{i} \Biggr)^{1/2}+ \sum_{i=1}^{n}w_{i}T _{i}-2\Biggl(\sum_{i=1}^{n}w_{i}T_{i} \Biggr)^{3/4} \Biggr) \\ & \quad {} -r_{0} \Biggl(\sum_{i=1}^{n}w_{i}T_{i}^{1/2}+I -2\sum_{i=1}^{n}w_{i}T _{i}^{1/4} \Biggr). \end{aligned}$$
In [18], the Tsallis relative operator entropy \(T_{t}(A|B)\) for positive invertible operators A, B and \(0< t\leq 1\) is defined as follows:
For further information as regards the Tsallis relative operator entropy see [6] and the references therein. In [7, Proposition 2.3], it is shown that for any unital positive linear map Φ the following inequality holds:
In (25), by similar techniques of Theorem 12, for positive operators \(A_{i}\), \(B_{i}\) (\(i=1,2,\ldots,n\)), we have
In the next theorem, we show a reverse of inequality (26).
Theorem 14
Let \(A_{i}, B_{i}\in \mathcal{B}(\mathcal{H})\) (\(1\leq i\leq n\)) be positive invertible and \(t\in (0,1)\).
-
(i)
If \(0< t\leq \frac{1}{2}\), then
$$\begin{aligned} &T_{t} \Biggl( \sum_{i=1}^{n}A_{i}\Big| \sum_{i=1}^{n}B_{i} \Biggr) -\sum _{i=1} ^{n}\bigl(T_{t}(A_{i}|B_{i}) \bigr) \\ & \quad \leq \frac{1}{t} \Biggl[R \Biggl(\sum_{i=1}^{n}A_{i}+ \sum_{i=1}^{n}B_{i}-2 \sum _{i=1}^{n}(A_{i}\sharp B_{i}) \Biggr) \\ &\quad\quad {}-r_{0} \Biggl(\sum _{i=1}^{n}(A_{i} \sharp B_{i})+ \sum_{i=1}^{n}B_{i}-2\sum _{i=1}^{n}(A_{i} \sharp_{\frac{3}{4}}B_{i}) \Biggr) \\ & \quad\quad {} -r_{0} \Biggl(\sum_{i=1}^{n}A_{i} \sharp \sum_{i=1}^{n}B_{i}+\sum _{i=1} ^{n}A_{i}-2\Biggl(\sum _{i=1}^{n}A_{i} \sharp_{\frac{1}{4}}\sum_{i=1}^{n}B _{i}\Biggr) \Biggr) \Biggr]. \end{aligned}$$ -
(ii)
If \(\frac{1}{2}< t<1\), then
$$\begin{aligned} &T_{t} \Biggl( \sum_{i=1}^{n}A_{i}\Big| \sum_{i=1}^{n}B_{i} \Biggr) - \sum_{i=1} ^{n}\bigl(T_{t}(A_{i}|B_{i}) \bigr)\\ &\quad \leq \frac{1}{t} \Biggl[R \Biggl(\sum_{i=1}^{n}A _{i}+\sum_{i=1}^{n}B_{i}-2 \Biggl(\sum_{i=1}^{n}A_{i}\sharp B_{i}\Biggr) \Biggr) \\ & \quad \quad {} -r_{0} \Biggl(\Biggl(\sum_{i=1}^{n}A_{i} \Biggr)\sharp \Biggl(\sum_{i=1}^{n}B_{i} \Biggr)+\sum_{i=1}^{n}B_{i}-2 \Biggl(\sum_{i=1}^{n}A_{i} \sharp_{\frac{3}{4}}\sum_{i=1} ^{n}B_{i} \Biggr) \Biggr) \\ &\quad \quad {} -r_{0} \Biggl(\sum_{i=1}^{n}(A_{i} \sharp B_{i})+\sum_{i=1}^{n}A_{i}-2 \sum_{i=1}^{n}(A_{i} \sharp_{\frac{1}{4}} B_{i}) \Biggr) \Biggr]. \end{aligned}$$
Proof
Applying Theorem 12 for \(0< t\leq \frac{1}{2}\), we have
whence we get the first inequality. The proof of the second inequality is similar. □
Remark 15
We can present our results for non-invertible operators; see [6]. It is a direct consequence of the definition of the mean in the sense of Kubo–Ando [11] that \(A\sharp_{t} (B+ \varepsilon )\) is a monotone increasing net. Let B be a non-invertible operator and \(\epsilon >0\). It follows from the set \(\{A\sharp_{t}(B+ \epsilon ): \epsilon >0\}\) being bounded above for \(0 < \varepsilon <1\) that the limit
exists in the strong operator topology. So by (27), \(A\sharp _{t}B\) exists.
References
Ando, T.: Concavity of certain maps on positive definite matrices and applications to Hadamard products. Linear Algebra Appl. 27, 203–241 (1979)
Bakherad, M., Krnić, M., Moslehian, M.S.: Reverse Young-type inequalities for matrices and operators. Rocky Mt. J. Math. 46(4), 1089–1105 (2016)
Bakherad, M., Lashkaripour, R., Hajmohamadi, M.: Extensions of interpolation between the arithmetic–geometric mean inequality for matrices. J. Inequal. Appl. 2017, Article ID 209 (2017)
Bhatia, R.: Positive Definite Matrices. Princeton University Press, Princeton (2007)
Fujii, M., Lee, E.-Y., Seo, Y.: A defference counterpart to a matrix Hölder inequality. Linear Algebra Appl. 432, 2565–2571 (2010)
Fujii, M., Nakamoto, R.: Refinements of Hölder–McCarthy inequalities and Young inequality. Adv. Oper. Theory 2, 184–188 (2016)
Furuichi, S., Yanagi, K., Kuriyama, K.: A note on operator inequalities of Tsallis relative operator entropy. Linear Algebra Appl. 407, 19–31 (2005)
Furuta, T.: The Hölder–McCarthy and the Young inequalities are equivalent for Hilbert space operators. Am. Math. Mon. 108, 68–69 (2001)
Hajmohamadi, M., Lashkaripour, R., Bakherad, M.: Some extensions of the Young and Heinz inequalities for matrices. Bull. Iran. Math. Soc. (2018). https://doi.org/10.1007/s41980-018-0064-3
Kittaneh, F., Manasrah, Y.: Reverse Young and Heinz inequalities for matrices. Linear Multilinear Algebra 59(9), 1031–1037 (2011)
Kubo, F., Ando, T.: Means of positive linear operators. Math. Ann. 246, 205–224 (1980)
Lee, E.-Y.: A matrix reverse Cauchy–Schwarz inequality. Linear Algebra Appl. 430, 805–810 (2009)
Mićić, R., Pečarić, J., Perić, I.: Reverses of Ando and Davis–Choi inequality. J. Math. Inequal. 9(2), 615–630 (2015)
Sababheh, M.: Means refinements via convexity. Mediterr. J. Math. 14(3), 125 (2017)
Sababheh, M., Choi, D.: A complete refinement of Young’s inequality. J. Math. Anal. Appl. 440, 379–393 (2016)
Sababheh, M., Moslehian, M.S.: Advanced refinements of Young and Heinz inequalities. J. Number Theory 172, 178–199 (2017)
Seo, Y.: Reverse of Ando’s inequality for positive linear maps. Math. Inequal. Appl. 14(4), 905–910 (2011)
Yanagi, K., Kuriyama, K., Furuichi, S.: Generalized Shannon inequality based on Tsallis relative operator entropy. Linear Algebra Appl. 394(9), 109–118 (2005)
Zhao, J., Wu, J.: Operator inequalities involving improved Young and its reverse inequalities. J. Math. Anal. Appl. 421, 1779–1789 (2015)
Zhao, J., Wu, J.: Some operator inequalities for unitarily invariant norms. Ann. Funct. Anal. 8(2), 240–247 (2017)
Acknowledgements
The third author would like to thank the Tusi Mathematical Research Group (TMRG).
Funding
The authors declare that there is no source of funding for this research.
Author information
Authors and Affiliations
Contributions
The authors contributed equally to the manuscript and read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Hajmohamadi, M., Lashkaripour, R. & Bakherad, M. Reverses of Ando’s and Hölder–McCarty’s inequalities. J Inequal Appl 2018, 164 (2018). https://doi.org/10.1186/s13660-018-1758-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13660-018-1758-z