# Discrete majorization type inequalities for convex functions on rectangles

## Abstract

In this paper, we present several discrete majorization type inequalities for the convex functions defined on rectangles.

## Introduction and preliminaries

We start this section by giving some brief introduction about convex function and related results.

### Definition 1.1

Let $$\textbf{I}\subseteq \mathbb{R}^{n}$$ be a convex set. Then the function $$\phi :\textbf{I}\mapsto \mathbb{R}$$ is said to be convex if the inequality

$$\phi \bigl(\zeta \textbf{x}+(1-\zeta )\textbf{y}\bigr)\leq \zeta \phi ( \textbf{x})+(1-\zeta )\phi (\textbf{y})$$

holds for all $$\textbf{x},\textbf{y}\in \textbf{I}$$ and $$\zeta \in [0,1]$$.

It is well-known that a convex function may not be differentiable. If the function ϕ is convex, then the support line inequality

$$\phi (\textbf{x})-\phi (\textbf{y})\geq \nabla _{+}\phi (\textbf{y}) ( \textbf{x}-\textbf{y})$$

holds for all $$\textbf{x}, \textbf{y}\in \textbf{I}$$, where

$$\nabla _{+}\phi (\textbf{y}) (\textbf{x}-\textbf{y})= \biggl\langle \frac{ \partial \phi _{+}(\textbf{y})}{\partial \textbf{y}}, (\textbf{x}- \textbf{y}) \biggr\rangle ,\qquad \frac{\partial \phi _{+}(\textbf{y})}{ \partial \textbf{y}}= \biggl(\frac{\partial \phi _{+}(\textbf{y})}{ \partial y_{1}},\frac{\partial \phi _{+}(\textbf{y})}{\partial y_{2}}, \ldots , \frac{\partial \phi _{+}(\textbf{y})}{\partial y_{n}} \biggr)$$

for $$\textbf{x}=(x_{1},x_{2},\dots ,x_{n})$$, $$\textbf{y}=(y_{1},y_{2}, \dots ,y_{n})\in \textbf{I}$$ and $$\langle \cdot ,\cdot \rangle$$ is the ordinary inner product in $$\mathbb{R}^{n}$$.

Convex functions have many important applications in mathematics, physics, statistics and engineering [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]. Currently, many refinements, variants, generalizations and extensions for the convexity can be found in the literature [21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50].

In , Dragomir introduced the definition of the coordinate convex functions as follows:

### Definition 1.2

(See )

Let $$[a_{1},a_{2}]$$ and $$[b_{1},b_{2}]$$ be two intervals in $$\mathbb{R}$$ and $$S=[a_{1},a_{2}]\times [b_{1},b_{2}]$$. A function $$\phi :S\mapsto \mathbb{R}$$ is said to be coordinate convex on S if the partial functions $$\phi _{y}: [a_{1},a_{2}]\mapsto \mathbb{R}$$ and $$\phi _{x}:[b_{1},b_{2}]\mapsto \mathbb{R}$$ defined by

$$\phi _{y}(u)=\phi (u,y), \qquad \phi _{x}(v)=\phi (x,v)$$

are convex.

### Lemma 1.3

(See )

Every convex function defined on a rectangle is coordinate convex, but the converse is not true, in general.

In the remaining part of this section, we give a comprehensive introduction about majorization theory.

Let $$n\geq 2$$, $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b_{2},\dots ,b_{n})$$ be two n-tuples of real numbers, and

$$a_{}\geq a_{}\geq \cdots \geq a_{[n]}, \qquad b_{}\geq b_{} \geq \cdots \geq b_{[n]}$$

be their ordered arrangement.

### Definition 1.4

The n-tuple b is said to be majorized by the n-tuple a, or a majorizes b, in symbols $$\textbf{a}\succ \textbf{b}$$, if

\begin{aligned}& \sum_{i=1}^{k}a_{[i]}\geq \sum _{i=1}^{k}b_{[i]} \quad (k=1,2,\dots ,n-1), \\& \sum_{i=1}^{n}b_{i}=\sum _{i=1}^{n}a_{i}. \end{aligned}

Let $$\textbf{a}, \textbf{b}\in \mathbb{R}^{n}$$ be two vectors such that a majorizes b. Then from the basic knowledge of linear algebra we clearly see that there exist a set of probabilities $$(q_{1},q_{2},\dots ,q_{n})$$ with $$\sum_{i=1}^{n}q_{i}=1$$ and a set of permutations $$(P_{1},P_{2},\dots ,P_{n})$$ such that $$\textbf{a}=\sum_{i=1}^{n}P_{i}q_{i}\textbf{b}$$. Alternatively, it can be shown that there exists a doubly stochastic matrix D such that $$\textbf{a}=D \textbf{b}$$. In fact, the latter characterization of majorization relation implies that the set of vectors a that satisfy $$\textbf{a}\succ \textbf{b}$$ is the convex hull spanned by the n! points formed from the permutations of the elements of b.

Let S and T be two Hermitian operators. Then we say that the Hermitian operator S majorizes the Hermitian operator T if the set of eigenvalues of S majorizes the set of eigenvalues values of T.

Majorization is a partial order relation between the vectors, which precisely defines the vague notion that the components of one vector are “less spread out” or “more nearly equal” than the components of another vector. And the functions that preserve the majorization order are called Schur convex functions. Many problems arising in signal processing and communications involve comparing vector-valued strategies or solving optimization problems with vector- or matrix-valued variables. Majorization theory is a key tool that allows us to solve or simplify these problems.

The following Theorem 1.5 is well-known in the literature as the majorization theorem and for its proof we refer to Marshall and Olkin . This result is due to Hardy, Littlewood and Pólya  and it can also be found in .

### Theorem 1.5

Let I be an interval in $$\mathbb{R}$$, and $$\textbf{a}=(a _{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b_{2},\dots ,b_{n})$$ be two n-tuples such that $$a_{i},b_{i}\in \mathbf{{I}}$$ ($$i=1,2, \dots ,n$$). Then the inequality

$$\sum_{i=1}^{n}\phi (a_{i})\geq \sum_{i=1}^{n}\phi (b_{i})$$

holds for every continuous convex function $$\phi :\mathbf {I}\mapsto \mathbb{R}$$ if and only if $$\textbf{a}\succ \textbf{b}$$.

The following Theorem 1.6 is a weighted version of Theorem 1.5 and is given by Fuchs .

### Theorem 1.6

Let $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b _{2},\dots , b_{n})$$ be two decreasing n-tuples such that $$a_{i}, b_{i}\in \mathbf{{I}}$$ ($$i=1,2,\dots ,n$$), and $$\textbf{p}=(p _{1},p_{2},\dots ,p_{n})$$ be a real n-tuple with

\begin{aligned}& \sum_{i=1}^{k}p_{i}a_{i} \geq \sum_{i=1}^{k}p_{i}b_{i} \quad (k=1,2, \dots ,n-1), \\& \sum_{i=1}^{n}p_{i}a_{i}= \sum_{i=1}^{n}p_{i}b_{i}. \end{aligned}

Then the inequality

$$\sum_{i=1}^{n}p_{i}\phi (a_{i})\geq \sum_{i=1}^{n}p_{i} \phi (b_{i})$$
(1.1)

holds for each continuous convex function $$\phi :\mathbf {I}\mapsto \mathbb{R}$$.

Another result similar to that above with some relaxed conditions on a, b and stricter condition on function ϕ was obtained by Bullen, Vasić and Stanković .

### Theorem 1.7

Let $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b _{2},\dots ,b_{n})$$ be two decreasing n-tuples, and $$\textbf{p}=(p _{1},p_{2},\dots , p_{n})$$ be a real n-tuple. If the inequality

$$\sum_{i=1}^{k}p_{i}a_{i} \geq \sum_{i=1}^{k}p_{i}b_{i}$$
(1.2)

holds for $$k=1,2,\dots ,n$$, then inequality (1.1) holds for each continuous increasing convex function $$\phi :\mathbf {I}\mapsto \mathbb{R}$$. If a and b are increasing n-tuples and the reverse inequality in (1.2) holds for $$k=1,2,\dots ,n$$, then inequality (1.1) holds for each continuous decreasing convex function $$\phi :\mathbf {I}\mapsto \mathbb{R}$$.

Dragomir  presented another majorization result, which has been obtained by using support line and Chebyshev’s inequalities.

### Theorem 1.8

Let I be an interval in $$\mathbb{R}$$, $$\phi :\mathbf {I}\mapsto \mathbb{R}$$ be a convex function, $$\textbf{a}=(a_{1}, a_{2},\dots , a _{n})$$ and $$\textbf{b}=(b_{1}, b_{2},\dots , b_{n})$$ be two real n-tuples such that $$a_{i}, b_{i}\in \mathbf{{I}}$$ ($$i=1,2,\dots ,n$$), and $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ be a non-negative real n-tuple with $$P_{n}=\sum_{i=1}^{n}p_{i}>0$$. If b and $$\textbf{a}-\textbf{b}$$ are monotonic in the same sense with $$\sum_{i=1}^{n}p_{i}a_{i}=\sum_{i=1}^{n}p_{i}ab_{i}$$, then one has

$$\sum_{i=1}^{n}p_{i} \phi (a_{i})\geq \sum_{i=1}^{n}p_{i} \phi (b_{i}).$$
(1.3)

If ϕ is strictly convex on I and $$p_{i}>0$$ ($$i=1,2, \dots ,n$$), then equality holds in (1.3) if and only if $$a_{i}=b_{i}$$ for all $$i=1,2,\dots ,n$$.

In this paper, our focus is on the majorization type results for the convex functions defined on rectangles. We shall extend classical majorization inequality for majorized tuples and establish weighted versions of majorization inequalities for certain tuples, for example, monotonic tuples in the same sense, monotonic tuples in mean, etc. For obtaining these results, we use Chebyshev’s inequality, Abel transformation, support line inequality of convex function and the fact that every convex function defined on rectangles is coordinate convex. At the end of the paper, we provide Favard’s type inequalities by using the generalized majorization results.

## Main results

We start by giving a majorization inequality for the convex functions defined on rectangles by using majorized tuples.

### Theorem 2.1

Let $$\mathbf {I}_{1}$$ and $$\mathbf {I}_{2}$$ be any two intervals in $$\mathbb{R}$$, $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b _{1},b_{2},\dots ,b_{n})$$ be two n-tuples such that $$a_{i},b_{i} \in \mathbf {I}_{1}$$ ($$i=1,2,\dots ,n$$), and $$\textbf{c}=(c_{1},c_{2}, \dots ,c_{m})$$ and $$\textbf{d}=(d_{1},d_{2},\dots ,d_{m})$$ be two m-tuples such that $$c_{j},d_{j}\in \mathbf {I}_{2}$$ ($$j=1,2,\dots ,m$$). If $$\textbf{a}\succ \textbf{b}$$ and $$\textbf{c}\succ \textbf{d}$$, then the inequality

$$\sum_{i=1}^{n}\sum _{j=1}^{m}\phi (a_{i},c_{j})\geq \sum_{i=1}^{n} \sum _{j=1}^{m}\phi (b_{i},d_{j})$$
(2.1)

holds for each convex function $$\phi :\textbf{I}_{1}\times \textbf{I} _{2}\mapsto \mathbb{R}$$.

### Proof

Without loss of generality, we assume that the tuples a, b, c and d are in decreasing order and $$a_{i}\neq b_{i}$$, $$c_{j}\neq d_{j}$$ for all $$i\in \{1,2,\dots ,n\}$$ and $$j\in \{1,2,\dots ,m\}$$. Let

\begin{aligned}& A_{k}=\sum_{i=1}^{k}a_{i},\qquad B_{k}=\sum_{i=1}^{k}b_{i} \quad (k=1,2, \dots ,n), \\& C_{l}=\sum_{j=1}^{l}c_{j},\qquad D_{l}=\sum_{j=1}^{l}d_{j} \quad (l=1,2, \dots ,m), \\& A_{0}=B_{0}=C_{0}=D_{0}=0. \end{aligned}

Then it follows from the definition of majorization that

$$A_{n}=B_{n},\qquad C_{m}=D_{m}.$$

Let $$t_{i,j}$$ and $$s_{i,j}$$ be defined by

\begin{aligned}& t_{i,j}:=\nabla \phi (a_{i},b_{i};c_{j})= \frac{\phi (a_{i},c_{j})- \phi (b_{i},c_{j})}{a_{i}-b_{i}}, \\& s_{i,j}:=\nabla \phi (b_{i};c_{j},d_{j})= \frac{\phi (b_{i},c_{j})- \phi (b_{i},d_{j})}{c_{j}-d_{j}}. \end{aligned}

Then we clearly see that

\begin{aligned} \phi (a_{i},c_{j})-\phi (b_{i},d_{j}) =& \phi (a_{i},c_{j})-\phi (b_{i},c _{j})+ \phi (b_{i},c_{j})-\phi (b_{i},d_{j}) \\ =&\frac{\phi (a_{i},c_{j})-\phi (b_{i},c_{j})}{a_{i}-b_{i}}(a_{i}-b _{i})+\frac{\phi (b_{i},c_{j})-\phi (b_{i},d_{j})}{c_{j}-d_{j}}(c_{j}-d _{j}) \\ =&t_{i,j} (A_{i}-A_{i-1}-B_{i}+B_{i-1} )+s_{i,j} (C_{j}-C _{j-1}-D_{j}+D_{j-1} ). \end{aligned}

Summing over all i and j gives

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}\phi (a_{i},c_{j})-\sum _{i=1}^{n}\sum_{j=1} ^{m}\phi (b_{i},d_{j}) \\& \quad =\sum_{i=1}^{n}\sum _{j=1}^{m}t_{i,j} (A_{i}-A_{i-1}-B_{i}+B_{i-1} ) \\& \quad \quad {}+\sum_{i=1}^{n}\sum _{j=1}^{m}s_{i,j} (C_{j}-C_{j-1}-D_{j}+D_{j-1} ) \\& \quad =\sum_{j=1}^{m} \Biggl[\sum _{i=1}^{n}t_{i,j} (A_{i}-B_{i} )- \sum_{i=1}^{n}t_{i,j} (A_{i-1}-B_{i-1} ) \Biggr] \\& \qquad {}+\sum_{i=1}^{n} \Biggl[\sum _{j=1}^{m}s_{i,j} (C_{j}-D_{j} )- \sum_{j=1}^{m}s_{i,j} (C_{j-1}-D_{j-1} ) \Biggr] \\& \quad =\sum_{j=1}^{m} \Biggl[\sum _{i=1}^{n-1}t_{i,j} (A_{i}-B_{i} )- \sum_{i=2}^{n}t_{i,j} (A_{i-1}-B_{i-1} ) \Biggr] \\& \qquad {}+\sum_{i=1}^{n} \Biggl[\sum _{j=1}^{m-1}s_{i,j} (C_{j}-D_{j} )- \sum_{j=2}^{m}s_{i,j} (C_{j-1}-D_{j-1} ) \Biggr] \\& \quad =\sum_{j=1}^{m} \Biggl[\sum _{i=1}^{n-1}t_{i,j} (A_{i}-B_{i} )- \sum_{i=1}^{n-1}t_{i+1,j} (A_{i}-B_{i} ) \Biggr] \\& \qquad {}+\sum_{i=1}^{n} \Biggl[\sum _{j=1}^{m-1}s_{i,j} (C_{j}-D_{j} )- \sum_{j=1}^{m-1}s_{i,j+1} (C_{j}-D_{j} ) \Biggr] \\& \quad =\sum_{j=1}^{m} \Biggl[\sum _{i=1}^{n-1} (t_{i,j}-t_{i+1,j} ) (A_{i}-B_{i} ) \Biggr] \\& \qquad {}+\sum_{i=1}^{n} \Biggl[\sum _{j=1}^{m-1} (s_{i,j}-s_{i,j+1} ) (C_{j}-D_{j} ) \Biggr]. \end{aligned}
(2.2)

Since ϕ is a convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, ϕ is a coordinate convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$. Thus, $$t_{i,j}$$ is decreasing with respect to i for each fixed j and $$s_{i,j}$$ is decreasing with respect to j for each fixed i. Hence $$t_{i,j}-t_{i+1,j}\geq 0$$ for all $$i\in \{1,2,\dots ,n-1\}$$ and $$s_{i,j}-s_{i,j+1}\geq 0$$ for all $$j\in \{1,2,\dots ,m-1\}$$. From the definition of majorization we get $$A_{i}-B_{i}\geq 0$$ for all $$i\in \{1,2,\dots ,n-1\}$$ and $$C_{j}-D _{j}\geq 0$$ for all $$j\in \{1,2,\dots ,m-1\}$$. Therefore, the right-hand side of (2.2) is non-negative, and hence we have

$$\sum_{i=1}^{n}\sum _{j=1}^{m}\phi (a_{i},c_{j})-\sum _{i=1}^{n}\sum_{j=1} ^{m}\phi (b_{i},d_{j})\geq 0,$$

which is equivalent to (2.1). □

In the following Theorem 2.2, we prove a general inequality for the convex functions defined on rectangles, which implies majorization inequality for certain tuples.

### Theorem 2.2

Let $$\mathbf {I}_{1}$$ and $$\mathbf {I}_{2}$$ be any two intervals in $$\mathbb{R}$$, $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b _{1},b_{2},\dots ,b_{n})$$ be two n-tuples such that $$a_{i}, b_{i} \in \mathbf {I}_{1}$$ ($$i=1,2,\dots ,n$$), $$\textbf{c}=(c_{1},c_{2},\dots ,c _{m})$$ and $$\textbf{d}=(d_{1},d_{2},\dots ,d_{m})$$ be two m-tuples such that $$c_{j},d_{j}\in \mathbf {I}_{2}$$ ($$j=1,2,\dots ,m$$) and $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ and $$\textbf{w}=(w_{1},w_{2}, \dots ,w_{m})$$ be any positive real n- and m-tuples, respectively. If $$\phi :\mathbf {I}_{1}\times \mathbf {I}_{2}\mapsto \mathbb{R}$$ is a convex function, then one has

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})-\sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}) \\& \quad \geq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i})+ \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}), \end{aligned}
(2.3)

where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$) and $$s_{j}$$ is the positive partial derivative of ϕ with respect to the second variable at $$d_{j}$$ ($$j=1,2,\dots ,m$$).

### Proof

It follows from the convexity of the function $$\phi :\mathbf {I}_{1} \times \mathbf {I}_{2}\mapsto \mathbb{R}$$ that

$$\phi (x,y)-\phi (w,z)\geq \bigl\langle \nabla \phi (w,z),(x-w,y-z)\bigr\rangle$$

for all $$(x,y),(w,z)\in \mathbf {I}_{1}\times \mathbf {I}_{2}$$. That is,

$$\phi (x,y)-\phi (w,z)\geq \frac{\partial \phi }{\partial w}(w,z) (x-w)+\frac{ \partial \phi }{\partial z}(w,z) (y-z).$$
(2.4)

Now, applying (2.4) and by choosing $$x\rightarrow a_{i}$$, $$y\rightarrow c_{i}$$, $$w\rightarrow b_{i}$$ and $$z\rightarrow d_{j}$$, we get

$$\phi (a_{i},c_{j})-\phi (b_{i},d_{j}) \geq t_{i}(a_{i}-b_{i})+s_{j}(c _{j}-d_{j}).$$
(2.5)

Multiplying both sides of (2.5) by $$p_{i}w_{j}$$ and summing over the indices, we obtain

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})-\sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}) \\& \quad \geq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i})+ \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}). \end{aligned}

□

If $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ and $$\textbf{w}=(w_{1},w _{2},\dots ,w_{m})$$ are two tuples, then throughout this paper $$P_{k}$$ and $$W_{j}$$ are defined by $$P_{k}=\sum_{i=1}^{k}p_{i}$$ and $$W_{j}=\sum_{i=1}^{j}w_{i}$$, $$k=1,2,\dots ,n$$ and $$j=1,2,\dots ,m$$.

Some majorization type results, which are obtained from the above Theorem 2.2, are given in the form of the following Propositions 2.3 and 2.5.

### Proposition 2.3

Assume that all the hypotheses of Theorem 2.2 hold. Additionally, b and $$\textbf{a}-\textbf{b}$$ are monotonic in the same sense, and c and $$\textbf{c}-\textbf{d}$$ are monotonic in the same sense. If

$$\sum_{i=1}^{n}a_{i}p_{i}= \sum_{i=1}^{n}b_{i}p_{i}$$
(2.6)

and

$$\sum_{j=1}^{m}c_{j}w_{j}= \sum_{j=1}^{m}d_{j}w_{j},$$
(2.7)

then

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})\geq \sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}).$$
(2.8)

### Proof

Since ϕ is a convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, ϕ is a coordinate convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$. If b is an increasing n-tuple, then $$(t_{1},t_{2},\dots ,t_{n})$$ is an increasing n-tuple, where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$). If b and $$\textbf{a}-\textbf{b}$$ are increasing n-tuples, then, applying Chebyshev’s inequality to the first term on right-hand side of (2.3) and using (2.6), we have

\begin{aligned} \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i}) =& \sum_{j=1} ^{m}w_{j} \Biggl[\sum _{i=1}^{n}p_{i}t_{i}(a_{i}-b_{i}) \Biggr] \\ \geq& \sum_{j=1}^{m}w_{j} \Biggl[ \frac{1}{P_{n}}\sum_{i=1}^{n}p_{i}t _{i}\sum_{i=1}^{n}p_{i}(a_{i}-b_{i}) \Biggr]=0. \end{aligned}
(2.9)

Similarly, since ϕ is a coordinate convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, if d is an increasing m-tuple, then $$(s_{1},s_{2},\dots ,s_{m})$$ is an increasing m-tuple, where $$s_{j}$$ is the positive partial derivative of ϕ with respect to the second variable at $$d_{j}$$ ($$j=1,2,\dots ,m$$). If d and $$\textbf{c}-\textbf{d}$$ are increasing m-tuples, then, applying Chebyshev’s inequality to the second term on right-hand side of (2.3) and using (2.7), we have

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}) \\& \quad = \sum_{i=1} ^{n}p_{i} \Biggl[\sum _{j=1}^{m}w_{j}s_{j}(c_{j}-d_{j}) \Biggr] \\& \quad \geq \sum_{i=1}^{n}p_{i} \Biggl[ \frac{1}{W_{m}}\sum_{j=1}^{m}w_{j}s _{j}\sum_{j=1}^{m}w_{j}(c_{j}-d_{j}) \Biggr]=0. \end{aligned}
(2.10)

Using (2.9) and (2.10) in (2.3), we get

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(a_{i}-b_{i})- \sum_{i=1} ^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}) \geq 0,$$

which is equivalent to (2.8).

Similarly, we can prove inequality (2.8) in the remaining cases. □

### Remark 2.4

In what follows, a convex function is said to be monotonic increasing if it is monotonic increasing with respect to each of its variables.

### Proposition 2.5

Let all the assumptions of Theorem 2.2 hold. If $$\phi :\mathbf {I}_{1} \times \mathbf {I}_{2}\mapsto \mathbb{R}$$ is an increasing convex function, b and $$\textbf{a}-\textbf{b}$$ are monotonic in the same sense, d and $$\textbf{c}-\textbf{d}$$ are monotonic in the same sense, and

$$\sum_{i=1}^{n}a_{i}p_{i} \geq \sum_{i=1}^{n}b_{i}p_{i}$$
(2.11)

and

$$\sum_{j=1}^{m}c_{j}w_{j} \geq \sum_{j=1}^{m}d_{j}w_{j},$$

then inequality (2.8) holds.

### Proof

Since ϕ is an increasing function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, we have $$t_{i}\geq 0$$ ($$i=1,2,\dots ,n$$), where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$), thus

$$\sum_{i=1}^{n}p_{i}t_{i} \geq 0.$$
(2.12)

Using (2.11) and (2.12) in the right-hand side of (2.9), we have

\begin{aligned} \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i}) \geq 0. \end{aligned}
(2.13)

Similarly, we have

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}) \geq 0.$$
(2.14)

Using (2.13) and (2.14) in (2.3), we get (2.8).

Similarly, we can prove inequality (2.8) in the remaining cases. □

The following Theorem 2.6 is another weighted discrete version of majorization theorem.

### Theorem 2.6

Let $$\textbf{I}_{1}$$, $$\textbf{I}_{2}$$ be two intervals in $$\mathbb{R}$$, $$\phi :\mathbf {I}_{1}\times \mathbf {I}_{2}\mapsto \mathbb{R}$$ be a convex function, $$\textbf{a}=(a_{1},a_{2},\dots , a_{n})$$ and $$\textbf{b}=(b_{1},b_{2},\dots , b_{n})$$ be two n-tuples such that $$a_{i},b_{i}\in \mathbf {I}_{1}$$ ($$i=1,2,\dots ,n$$), $$\textbf{c}=(c_{1},c _{2},\dots ,c_{m})$$ and $$\textbf{d}=(d_{1},d_{2},\dots ,d_{m})$$ be two m-tuples such that $$c_{j},d_{j}\in \mathbf {I}_{2}$$ ($$j=1,2,\dots ,m$$), $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ and $$\textbf{w}=(w_{1},w_{2}, \dots ,w_{n})$$ be two any positive real n- and m-tuples, respectively, and

\begin{aligned}& \sum_{i=1}^{k}b_{i}p_{i} \leq \sum_{i=1}^{k}a_{i}p_{i} \quad (k=1,2, \dots ,n-1), \end{aligned}
(2.15)
\begin{aligned}& \sum_{i=1}^{k}d_{j}w_{j} \leq \sum_{i=1}^{k}c_{j}w_{j} \quad (k=1,2, \dots ,m-1), \end{aligned}
(2.16)
\begin{aligned}& \sum_{i=1}^{n}b_{i}p_{i}= \sum_{i=1}^{n}a_{i}p_{i}, \end{aligned}
(2.17)
\begin{aligned}& \sum_{i=1}^{m}d_{j}w_{j}= \sum_{i=1}^{m}c_{j}w_{j}. \end{aligned}
(2.18)

Then the following statements are true:

1. (i)

If b and d are decreasing n- and m-tuples, respectively, then

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (b_{i},d_{j})\leq \sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (a_{i},c_{j}).$$
(2.19)
2. (ii)

If a and c are increasing n- and m-tuples, respectively, then

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})\leq \sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}).$$
(2.20)

### Proof

For part (i), we use Abel’s transformation to prove part (i). Let

$$A_{0}=B_{0}=0,\qquad A_{k}=\sum _{i=1}^{k}p_{i}a_{i},\qquad B_{k}= \sum_{i=1}^{k}p_{i}b_{i} \quad (k=1,2,\dots ,n)$$

and

$$C_{0}=D_{0}=0, \qquad C_{k}=\sum _{j=1}^{k}w_{j}c_{j},\qquad D_{k}= \sum_{j=1}^{k}w_{j}d_{j} \quad (k=1,2,\dots ,m).$$

Then from (2.17) and (2.18) we have

$$A_{n}=B_{n},\qquad C_{m}=D_{m}.$$

Since ϕ is a convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, ϕ is a coordinate convex function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$. If b and d are decreasing n- and m-tuples, respectively, then $$(t_{1},t_{2}, \dots ,t_{n})$$ and $$(s_{1},s_{2},\dots ,s_{m})$$ are decreasing n- and m-tuples, respectively, where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$) and $$s_{j}$$ is the positive partial derivative of ϕ with respect to the second variable at $$d_{j}$$ ($$j=1,2, \dots ,m$$). It follows from (2.3) that

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})-\sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}) \\& \quad \geq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i})+ \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}) \\& \quad =\sum_{j=1}^{m}w_{j} \Biggl[\sum _{i=1}^{n}t_{i}(p_{i}a_{i}-p_{i}b_{i}) \Biggr]+ \sum_{i=1}^{n}p_{i} \Biggl[\sum_{j=1}^{m}s_{j}(w_{j}c_{j}-w_{j}d_{j}) \Biggr] \\& \quad =\sum_{j=1}^{m}w_{j} \Biggl[\sum _{i=1}^{n}t_{i}(A_{i}-A_{i-1}-B_{i}+B _{i-1}) \Biggr] \\& \qquad {}+\sum_{i=1}^{n}p_{i} \Biggl[\sum _{j=1}^{m}s_{j}(C_{i}-C_{i-1}-D_{i}+D _{i-1}) \Biggr] \\& \quad =\sum_{j=1}^{m}w_{j} \Biggl[\sum _{i=1}^{n}t_{i}(A_{i}-B_{i})- \sum_{i=1} ^{n}t_{i}(A_{i-1}-B_{i-1}) \Biggr] \\& \qquad {}+\sum_{i=1}^{n}p_{i} \Biggl[\sum _{j=1}^{m}s_{j}(C_{i}-D_{i})- \sum_{j=1} ^{m}s_{j}(C_{i-1}-D_{i-1}) \Biggr] \\& \quad =\sum_{j=1}^{m}w_{j} \Biggl[\sum _{i=1}^{n-1}(t_{i}-t_{i+1}) (A_{i}-B _{i}) \Biggr] +\sum_{i=1}^{n}p_{i} \Biggl[\sum _{j=1}^{m-1}(s_{j}-s_{j+1}) (C_{i}-D _{i}) \Biggr]. \end{aligned}
(2.21)

Since $$(t_{1},t_{2},\dots ,t_{n})$$ and $$(s_{1},s_{2},\dots ,s_{m})$$ are decreasing n- and m-tuples, respectively, $$t_{i}- t_{i+1}\geq 0$$ ($$i=1,2,\dots ,n-1$$) and $$s_{j}-s_{j+1}\geq 0$$ ($$j=1,2,\dots ,m-1$$). Also from the assumptions (2.15) and (2.16) we know that $$A_{i}-B_{i} \geq 0$$ ($$i=1,2,\dots ,n-1$$) and $$C_{j}-D_{j}\geq 0$$ ($$j=1,2,\dots , m-1$$). Thus

$$\sum_{j=1}^{m}w_{j} \Biggl[\sum _{i=1}^{n-1}(t_{i}-t_{i+1}) (A_{i}-B_{i}) \Biggr]+ \sum_{i=1}^{n}p_{i} \Biggl[\sum_{j=1}^{m-1}(s_{j}-s_{j+1}) (C_{i}-D_{i}) \Biggr] \geq 0.$$
(2.22)

Using (2.22) in (2.21), we get

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})-\sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j})\geq 0,$$

which is equivalent to (2.19).

Similarly, we can prove inequality (2.20) for the remaining cases. □

### Definition 2.7

Let $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ be a positive real n-tuple. Then the real n-tuple $$\textbf{a}=(a_{1},a_{2},\dots ,a _{n})$$ is said to be monotonic increasing in mean relative to $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ if

$$\frac{1}{P_{k}}\sum_{i=1}^{k}p_{i}a_{i} \leq \frac{1}{P_{k+1}}\sum_{i=1} ^{k+1}a_{i}p_{i} \quad (k=1,2,\dots ,n-1),$$

and decreasing in mean relative to $$\textbf{p}=(p_{1},p_{2},\dots ,p _{n})$$ if

$$\frac{1}{P_{k}}\sum_{i=1}^{k}p_{i}a_{i} \geq \frac{1}{P_{k+1}}\sum_{i=1} ^{k+1}a_{i}p_{i} \quad (k=1,2,\dots ,n-1).$$

The following Lemma 2.8 is due to Biernacki  (for a generalization, see Burkill and Mirsky ).

### Lemma 2.8

Let $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b _{2},\dots ,b_{n})$$ be any two real n-tuples, which are monotonic in mean relative to positive real n-tuple $$\textbf{p}=(p_{1},p_{2}, \dots ,p_{n})$$ in the same sense, that is,

$$\frac{1}{P_{k}}\sum_{i=1}^{k}p_{i}a_{i} \lesseqqgtr \frac{1}{P_{k+1}} \sum_{i=1}^{k+1}a_{i}p_{i} \quad (k=1,2,\dots ,n-1)$$

and

$$\frac{1}{P_{k}}\sum_{i=1}^{k}p_{i}b_{i} \lesseqqgtr \frac{1}{P_{k+1}} \sum_{i=1}^{k+1}b_{i}p_{i} \quad (k=1,2,\dots ,n-1).$$

Then

$$\frac{1}{P_{n}}\sum_{i=1}^{n}p_{i}a_{i}b_{i} \geq \frac{1}{P_{n}}\sum_{i=1}^{n}a_{i}p_{i} \frac{1}{P_{n}}\sum_{i=1}^{n}b_{i}p_{i}.$$
(2.23)

If one tuple is decreasing in mean and the other one is increasing in mean, then the reverse inequality holds in (2.23).

Now, we state another result for convex functions and for arbitrary monotonic tuples in mean.

### Theorem 2.9

Let all the assumptions of Theorem 2.2 hold. Additionally, if b and $$\textbf{a}-\textbf{b}$$ are monotonic n-tuples in mean relative to $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ in the same sense, and c and $$\textbf{c}-\textbf{d}$$ are monotonic m-tuples in mean relative to $$\textbf{w}=(w_{1},w_{2},\dots ,w_{m})$$ in the same sense, then

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})-\sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}) \\& \quad \geq \frac{1}{P_{n}}\sum_{j=1}^{m}\sum _{i=1}^{n}w_{j}p_{i}t_{i} \sum_{i=1}^{n}p_{i}(a_{i}-b_{i}) \\& \qquad {}+\frac{1}{W_{m}}\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j} \sum_{j=1} ^{m}w_{j}(c_{j}-d_{j}), \end{aligned}
(2.24)

where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$) and $$s_{j}$$ is the partial positive derivative of ϕ with respect to the second variable at $$d_{j}$$ ($$j=1,2,\dots ,m$$).

### Proof

It follows from the proof of Proposition 2.3 that $$(t_{1},t_{2}, \dots ,t_{n})$$ is an increasing n-tuple. Now if b and $$\textbf{a}-\textbf{b}$$ are monotonic increasing in mean relative to p, then, applying Chebyshev’s inequality to first term on the right-hand side of (2.3), we have

\begin{aligned} \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}t_{i}(a_{i}-b_{i}) =& \sum_{j=1} ^{m}w_{j} \Biggl[\sum _{i=1}^{n}p_{i}t_{i}(a_{i}-b_{i}) \Biggr] \\ \geq& \sum_{j=1}^{m}w_{j} \Biggl[ \frac{1}{P_{n}}\sum_{i=1}^{n}p_{i}t _{i}\sum_{i=1}^{n}p_{i}(a_{i}-b_{i}) \Biggr] \\ =&\frac{1}{P_{n}}\sum_{j=1}^{m}\sum _{i=1}^{n}w_{j}p_{i}t_{i} \sum_{i=1} ^{n}p_{i}(a_{i}-b_{i}). \end{aligned}
(2.25)

Similarly, we have

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j}(c_{j}-d_{j}) \geq \frac{1}{W _{m}}\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}s_{j} \sum_{j=1}^{m}w_{j}(c _{j}-d_{j}).$$
(2.26)

Using (2.25) and (2.26) in (2.3), we get (2.24).

Similarly, we can prove inequality (2.24) in the remaining cases. □

### Corollary 2.10

Assume that all the hypotheses of Theorem 2.9 hold. Additionally, if

$$\sum_{i=1}^{n}a_{i}p_{i}= \sum_{i=1}^{n}b_{i}p_{i}$$
(2.27)

and

$$\sum_{j=1}^{m}c_{j}w_{j}= \sum_{j=1}^{m}d_{j}w_{j},$$
(2.28)

then

$$\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi (a_{i},c_{j})\geq \sum_{i=1} ^{n}\sum_{j=1}^{m}p_{i}w_{j} \phi (b_{i},d_{j}).$$
(2.29)

### Proof

Using (2.27) and (2.28) on the right-hand side of (2.24), we get (2.29). □

In the following Corollary 2.11, we obtain a majorization inequality by using an increasing convex function.

### Corollary 2.11

Let all the assumptions of Theorem 2.9 hold. If $$\phi :\mathbf {I}_{1} \times \mathbf {I}_{2}\mapsto \mathbb{R}$$ is an increasing convex function, and

$$\sum_{i=1}^{n}a_{i}p_{i} \geq \sum_{i=1}^{n}b_{i}p_{i}$$
(2.30)

and

$$\sum_{j=1}^{m}c_{j}w_{j} \geq \sum_{j=1}^{m}d_{j}w_{j},$$
(2.31)

then inequality (2.29) holds.

### Proof

Since ϕ is an increasing function on $$\textbf{I}_{1}\times \textbf{I}_{2}$$, we get that $$t_{i}\geq 0$$, $$s_{j}\geq 0$$ ($$i=1,2,\dots ,n$$, $$j=1,2,\dots ,m$$), where $$t_{i}$$ is the positive partial derivative of ϕ with respect to the first variable at $$b_{i}$$ ($$i=1,2,\dots ,n$$) and $$s_{j}$$ is the partial positive derivative of ϕ with respect to the second variable at $$d_{j}$$ ($$j=1,2,\dots ,m$$), thus

\begin{aligned}& \sum_{i=1}^{n}p_{i}t_{i} \geq 0, \end{aligned}
(2.32)
\begin{aligned}& \sum_{j=1}^{m}w_{j}s_{j} \geq 0. \end{aligned}
(2.33)

Hence using (2.30), (2.31), (2.32) and (2.33) on the right-hand side of (2.24), we obtain inequality (2.29). □

The following Lemma 2.12 can be found in the literature .

### Lemma 2.12

Let $$\textbf{v}=(v_{1}, v_{2}, \dots , v_{n})$$ be a positive real n-tuple and $$\textbf{x}=(x_{1}, x_{2}, \dots , x_{n})$$ be an increasing real n-tuple. Then the inequality

$$\sum_{i=1}^{k}x_{i}v_{i} \sum_{i=1}^{n}v_{i}\leq \sum _{i=1}^{n}x_{i}v _{i} \sum_{i=1}^{k}v_{i}$$
(2.34)

holds for $$k=1,2,\dots ,n$$. If x is a decreasing real n-tuple, then the reverse inequality holds in (2.32).

If $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b _{2},\dots ,b_{n})$$ with $$b_{i}\neq 0$$ for $$i=1, 2, \dots , n$$, then $$\frac{\textbf{a}}{\textbf{b}}=(\frac{a_{1}}{b_{1}},\frac{a_{2}}{b _{2}},\dots ,\frac{a_{n}}{b_{n}})$$.

### Theorem 2.13

Let $$\mathbf {I}_{1}$$ and $$\mathbf {I}_{2}$$ be any two intervals in $$\mathbb{R}$$, $$\phi :\mathbf {I}_{1}\times \mathbf {I}_{2}\mapsto \mathbb{R}$$ be a convex function, $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ and $$\textbf{b}=(b_{1},b_{2},\dots ,b_{n})$$ be two positive n-tuples such that $$a_{i},b_{i}\in \mathbf {I}_{1}$$ ($$i=1,2,\dots ,n$$), $$\textbf{c}=(c _{1},c_{2},\dots ,c_{m})$$ and $$\textbf{d}=(d_{1},d_{2},\dots ,d_{m})$$ be two positive m-tuples such that $$c_{j},d_{j}\in \mathbf {I}_{2}$$ ($$j=1,2,\dots ,m$$), $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ and $$\textbf{w}=(w_{1},w_{2},\dots ,w_{m})$$ be any positive real n- and m-tuples, respectively, and $$\textbf{a}/\textbf{b}$$ and $$\textbf{c}/ \textbf{d}$$ are decreasing n- and m-tuples, respectively. Then the following statements are true:

1. (i)

If a is an increasing n-tuple and c is an increasing m-tuple, then

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr) \\& \quad \leq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{b_{i}}{ \sum_{i=1}^{n}b_{i}p_{i}},\frac{d_{j}}{\sum_{j=1}^{m}w_{j}d_{j}} \biggr). \end{aligned}
(2.35)
2. (ii)

If b is a decreasing n-tuple and d is a decreasing m-tuple, then the reverse inequality holds in (2.33).

If $$\textbf{a}/\textbf{b}$$ and $$\textbf{c}/\textbf{d}$$ are increasing n and m-tuples, respectively, then we have the following statements:

3. (iii)

If b is an increasing n-tuple and d is an increasing m-tuple, then

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{b_{i}}{\sum_{i=1}^{n}b_{i}p_{i}},\frac{d_{j}}{\sum_{j=1}^{m}w_{j}d_{j}} \biggr) \\& \quad \leq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{ \sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr). \end{aligned}
(2.36)
4. (iv)

If a is a decreasing n-tuple and c is a decreasing m-tuple, then the reverse inequality holds in (2.34).

### Proof

(i) Let $$\textbf{a}/\textbf{b}$$ and $$\textbf{c}/\textbf{d}$$ are decreasing n- and m-tuples, respectively. Then using Lemma 2.12 with $$\textbf{x}=\textbf{a}/\textbf{b}$$ and $$\textbf{v}=\textbf{p} \textbf{b}$$, we obtain

$$\sum_{i=1}^{n}p_{i}a_{i} \sum_{i=1}^{k}p_{i}b_{i} \leq \sum_{i=1}^{k}a _{i}p_{i} \sum_{i=1}^{n}p_{i}b_{i} \quad (k=1,2,\dots ,n).$$

That is,

$$\sum_{i=1}^{k}p_{i} \biggl( \frac{b_{i}}{\sum_{i=1}^{n}p_{i}b_{i}} \biggr) \leq \sum_{i=1}^{k}p_{i} \biggl(\frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}} \biggr) \quad (k=1,2,\dots ,n).$$
(2.37)

Similarly, using Lemma 2.12 with $$\textbf{x}=\textbf{c}/\textbf{d}$$ and $$\textbf{v}=\textbf{d}\textbf{w}$$, we get

$$\sum_{i=1}^{n}w_{j}c_{j} \sum_{i=1}^{k}w_{j}d_{j} \leq \sum_{i=1}^{k}c _{j}w_{j} \sum_{i=1}^{n}d_{j}w_{j} \quad (k=1,2,\dots ,m).$$

That is,

$$\sum_{j=1}^{k}w_{j} \biggl( \frac{d_{j}}{\sum_{j=1}^{m}w_{j}d_{j}} \biggr) \leq \sum_{j=1}^{k}w_{j} \biggl(\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr)\quad (k=1,2,\dots ,m).$$
(2.38)

Also, it is obvious that

$$\sum_{i=1}^{n}p_{i} \biggl( \frac{b_{i}}{\sum_{i=1}^{n}p_{i}b_{i}} \biggr)= \sum_{i=1}^{n}p_{i} \biggl(\frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}} \biggr)$$
(2.39)

and

$$\sum_{j=1}^{n}w_{j} \biggl( \frac{d_{j}}{\sum_{j=1}^{m}w_{j}d_{j}} \biggr)= \sum_{j=1}^{n}w_{j} \biggl(\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr).$$
(2.40)

If a and c are increasing, then using Theorem 2.6(ii) and (2.35)–(2.38), we have

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr) \\& \quad \leq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{b_{i}}{ \sum_{i=1}^{n}b_{i}p_{i}},\frac{d_{j}}{\sum_{j=1}^{m}w_{j}d_{j}} \biggr). \end{aligned}

Similarly, we can prove the remaining cases. □

### Definition 2.14

(See )

An n-tuple $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ is said to be concave if

$$2a_{k}\geq a_{k+1}+a_{k-1}$$

for all $$k=2, 3, \dots ,n-1$$.

### Definition 2.15

(See )

An n-tuple $$\textbf{a}=(a_{1},a_{2},\dots ,a_{n})$$ is said to be convex if

$$2a_{k}\leq a_{k+1}+a_{k-1}$$

for all $$k=2, 3, \dots ,n-1$$.

### Corollary 2.16

Let $$\phi :[0,\infty )\times [0,\infty )\longrightarrow \mathbb{R}$$ be a convex function, $$\textbf{a}=(a_{1},a_{2},\dots , a_{n})$$ be a positive n-tuple, $$\textbf{c}=(c_{1},c_{2},\dots ,c_{m})$$ be a positive m-tuple, and $$\textbf{p}=(p_{1},p_{2},\dots ,p_{n})$$ and $$\textbf{w}=(w_{1},w_{2},\dots ,w_{m})$$ be any positive real n- and m-tuples, respectively. Then the following statements are true:

1. (i)

If a is an increasing concave n-tuple and c is an increasing concave m-tuple, then

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{i-1}{\sum_{i=1} ^{n}p_{i}({i-1})},\frac{j-1}{\sum_{j=1}^{m}w_{j}({j-1)}} \biggr) \\& \quad \geq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{ \sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr). \end{aligned}
(2.41)
2. (ii)

If a is an increasing convex n-tuple with $$a_{1}=0$$ and c is an increasing convex m-tuple with $$c_{1}=0$$, then the reverse inequality holds in (2.39).

3. (iii)

If a is a decreasing concave n-tuple and c is a decreasing concave m-tuple, then

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{n-i}{\sum_{i=1} ^{n}p_{i}({n-i})},\frac{m-j}{\sum_{j=1}^{m}w_{j}({m-j)}} \biggr) \\& \quad \geq \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{ \sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr). \end{aligned}
(2.42)
4. (iv)

If a is a decreasing convex n-tuple with $$a_{n}=0$$ and c is a decreasing convex m-tuple with $$c_{m}=0$$, then the reverse inequality holds in (2.40).

### Proof

(i) Let $$\textbf{b}=(b_{1}, b_{2}, \dots , b_{n})$$ and $$\textbf{d}=(d _{1}, d_{2}, \dots , d_{m})$$ be respectively the n- and m-tuples such that $$b_{1}=\epsilon < a_{1}/a_{2}$$, $$d_{1}=\delta < c_{1}/c_{2}$$, $$b_{i}=i-1$$ for $$i=2, 3, \dots , n$$, and $$d_{j}=j-1$$ for $$j=2, 3, \dots , m$$. Then $$\textbf{a}/\textbf{b}$$ and $$\textbf{c}/\textbf{d}$$ are decreasing n- and m-tuples, respectively. It follows from Theorem 2.13 that

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr) \\& \quad \leq p_{1}w_{1}\phi \biggl(\frac{\epsilon }{\epsilon p_{1}+\sum_{i=2} ^{n}(i-1)p_{i}}, \frac{\delta }{w_{1}\delta +\sum_{j=2}^{m}(j-1)w_{j}} \biggr) \\& \qquad {}+\sum_{i=2}^{n}\sum _{j=2}^{m}p_{i}w_{j}\phi \biggl( \frac{i-1}{\epsilon p_{1}+\sum_{i=2}^{n}(i-1)p_{i}},\frac{j-1}{w_{1}\delta +\sum_{j=2} ^{m}(j-1)w_{j}} \biggr). \end{aligned}

Taking $$\epsilon \rightarrow 0$$ and $$\delta \rightarrow 0$$, we obtain

\begin{aligned}& \sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{a_{i}}{\sum_{i=1}^{n}p_{i}a_{i}},\frac{c_{j}}{\sum_{j=1}^{m}w_{j}c_{j}} \biggr) \\& \quad \leq p_{1}w_{1}\phi (0,0)+\sum _{i=2}^{n} \sum_{j=2}^{m}p_{i}w_{j} \phi \biggl(\frac{i-1}{\epsilon p_{1}+\sum_{i=2}^{n}(i-1)p_{i}},\frac{j-1}{w _{1}\delta +\sum_{j=2}^{m}(j-1)w_{j}} \biggr) \\& \quad =\sum_{i=1}^{n}\sum _{j=1}^{m}p_{i}w_{j}\phi \biggl( \frac{i-1}{\epsilon p_{1}+\sum_{i=2}^{n}(i-1)p_{i}},\frac{j-1}{w_{1}\delta +\sum_{j=2} ^{m}(j-1)w_{j}} \biggr). \end{aligned}

This proves (2.39).

Similarly, we can use other parts of Theorem 2.13 to prove the required results for the remaining cases. □

### Remark 2.17

For some related integral majorization inequalities for the convex functions defined on rectangles, see ; and for some other recent results related to majorization, see [63,64,65].

## Results and discussion

In the article, we have generalized the classical majorization inequality for majorized tuples, established several weighted version of majorization inequalities for certain tuples and provided Favard’s type inequalities by the use of Chebyshev’s inequality, Abel transformation and support line inequality.

## Conclusions

In this paper, we have extended some discrete majorization type inequalities of convex functions from intervals to rectangles. The given results are generalizations of the previously known results. Our approach may have further applications in the theory of majorization.

## References

1. Roberts, A.W., Varberg, D.E.: Convex Functions. Academic Press, New York (1973)

2. Pečarić, J.E., Proschan, F., Tong, T.Y.: Convex Functions, Partial Orderings, and Statistical Applications. Academic Press, Boston (1992)

3. Udrişte, C.: Convex Functions and Optimization Methods on Riemannian Manifolds. Kluwer Academic, Dordrecht (1994)

4. Zhang, X.-M., Chu, Y.-M.: Convexity of the integral arithmetic mean of a convex function. Rocky Mt. J. Math. 40(3), 1061–1068 (2010)

5. Zhang, X.-M., Chu, Y.-M., Zhang, X.-H.: The Hermite–Hadamard type inequality of GA-convex functions and its applications. J. Inequal. Appl. 2010, Article ID 507560 (2010)

6. Wang, M.-K., Chu, Y.-M., Qiu, S.-L., Jiang, Y.-P.: Convexity of the complete elliptic integrals of the first kind with respect to Hölder means. J. Math. Anal. Appl. 388(2), 1141–1146 (2012)

7. Yang, Z.-H., Chu, Y.-M., Wang, M.-K.: Monotonicity criterion for the quotient of power series with applications. J. Math. Anal. Appl. 428(1), 587–604 (2015)

8. Yang, Z.-H., Qian, W.-M., Chu, Y.-M., Zhang, W.: Monotonicity rule for the quotient of two functions and its applications. J. Inequal. Appl. 2017, Article ID 106 (2017)

9. Yang, Z.-H., Chu, Y.-M.: A monotonicity property involving the generalized elliptic integral of the first kind. Math. Inequal. Appl. 20(3), 729–735 (2017)

10. Yang, Z.-H., Qian, W.-M., Chu, Y.-M., Zhang, W.: On rational bounds for the gamma function. J. Inequal. Appl. 2017, Article ID 210 (2017)

11. Wang, M.-K., Li, Y.-M., Chu, Y.-M.: Inequalities and infinite product formula for Ramanujan generalized modular equation function. Ramanujan J. 46(1), 189–200 (2018)

12. Yang, Z.-H., Qian, W.-M., Chu, Y.-M., Zhang, W.: On approximating the arithmetic–geometric mean and complete elliptic integral of the first kind. J. Math. Anal. Appl. 462(2), 1714–1726 (2018)

13. Yang, Z.-H., Qian, W.-M., Chu, Y.-M., Zhang, W.: On approximating the error function. Math. Inequal. Appl. 21(2), 469–479 (2018)

14. Wang, M.-K., Chu, Y.-M.: Landen inequalities for a class of hypergeometric functions with applications. Math. Inequal. Appl. 21(2), 521–537 (2018)

15. Niculescu, C.P., Persson, L.-E.: Convex Functions and Their Applications. Springer, Cham (2018)

16. Wang, M.-K., Qiu, S.-L., Chu, Y.-M.: Infinite series formula for Hübner upper bound function with applications to Hersch–Pfluger distortion function. Math. Inequal. Appl. 21(3), 629–648 (2018)

17. Zhao, T.-H., Wang, M.-K., Zhang, W., Chu, Y.-M.: Quadratic transformation inequalities for Gaussian hypergeometric function. J. Inequal. Appl. 2018, Article ID 251 (2018)

18. Yang, Z.-H., Qian, W.-M., Chu, Y.-M.: Monotonicity properties and bounds involving the complete elliptic integral of the first kind. Math. Inequal. Appl. 21(4), 1185–1199 (2018)

19. Yang, Z.-H., Chu, Y.-M., Zhang, W.: High accuracy asymptotic bounds for the complete elliptic integral of the second kind. Appl. Math. Comput. 348, 552–564 (2019)

20. Wang, M.-K., Chu, Y.-M., Zhang, W.: The precise estimates for the solution of Ramanujan’s generalized modular equation. Ramanujan J. https://doi.org/10.1007/s11139-018-0130-8

21. Chu, Y.-M., Lv, Y.-P.: The Schur harmonic convexity of the Hamy symmetric function and its applications. J. Inequal. Appl. 2009, Article ID 838529 (2009)

22. Chu, Y.-M., Xia, W.-F., Zhao, T.-H.: Schur convexity for a class of symmetric functions. Sci. China Math. 53(2), 465–474 (2010)

23. Chu, Y.-M., Sun, T.-C.: The Schur harmonic convexity for a class of symmetric functions. Acta Math. Sci. 30B(5), 1501–1506 (2010)

24. Chu, Y.-M., Wang, G.-D., Zhang, X.-H.: The Schur multiplicative and harmonic convexities of the complete symmetric function. Math. Nachr. 284(5–6), 653–663 (2011)

25. Chu, Y.-M., Wang, M.-K., Qiu, S.-L.: Optimal combinations bounds of root-square and arithmetic means for Toader mean. Proc. Indian Acad. Sci. Math. Sci. 122(1), 41–51 (2012)

26. Chu, Y.-M., Wang, M.-K.: Optimal Lehmer mean bounds for the Toader mean. Results Math. 61(3–4), 223–229 (2012)

27. Chu, Y.-M., Xia, W.-F., Zhang, X.-H.: The Schur concavity, Schur multiplicative and harmonic convexities of the second dual form of the Hamy symmetric function with applications. J. Multivar. Anal. 105, 412–421 (2012)

28. Adil Khan, M., Zheer Ullah, S., Chu, Y.-M.: The concept of coordinate strongly convex functions and related inequalities. Rev. R. Acad. Cienc. Exactas Fís. Nat., Ser. A Mat. (2018). https://doi.org/10.1007/s13398-018-0615-8

29. Chu, Y.-M., Zhao, T.-H.: Concavity of the error function with respect to Hölder means. Math. Inequal. Appl. 19(2), 589–595 (2016)

30. Wang, M.-K., Chu, Y.-M., Jiang, Y.-P.: Ramanujan’s cubic transformation inequalities for zero-balanced hypergeometric functions. Rocky Mt. J. Math. 46(2), 679–691 (2016)

31. Yang, Y.-Y., Qian, W.-M., Chu, Y.-M.: Refinements of bounds for Neuman means with applications. J. Nonlinear Sci. Appl. 9(4), 1529–1540 (2016)

32. Wang, M.-K., Chu, Y.-M., Song, Y.-Q.: Asymptotical formulas for Gaussian and generalized hypergeometric functions. Appl. Math. Comput. 276, 44–60 (2016)

33. Adil Khan, M., Chu, Y.-M., Khan, T.U., Khan, J.: Some new inequalities of Hermite–Hadamard type for s-convex functions with applications. Open Math. 15, 1414–1430 (2017)

34. İşcan, İ., Turhan, S., Maden, S.: Hermite–Hadamard and Simpson-like type inequalities for differentiable p-quasi-convex functions. Filomat 31(19), 5945–5953 (2017)

35. Latif, M.A., Dragomir, S.S., Momoniat, E.: Some estimates on the Hermite–Hadamard inequality through geometrically quasi-convex functions. Miskolc Math. Notes 18(2), 933–946 (2017)

36. Khurshid, Y., Adil Khan, M., Chu, Y.-M., Khan, Z.A.: Hermite–Hadamard–Fejér inequalities for conformable fractional integrals via preinvex functions. J. Funct. Spaces 2019, Article ID 3146210 (2019)

37. Olbryś, A.: On the $$\mathbb{K}$$-Riemann integral and Hermite–Hadamard inequalities for $$\mathbb{K}$$-convex functions. Aequ. Math. 91(3), 429–444 (2017)

38. Qian, W.-M., Chu, Y.-M.: Sharp bounds for a special quasi-arithmetic mean in terms of arithmetic and geometric means with two parameters. J. Inequal. Appl. 2017, Article ID 274 (2017)

39. Khurshid, Y., Adil Khan, M., Chu, Y.-M.: Conformable integral inequalities of the Hermite–Hadamard type in terms of GG- and GA-convexities. J. Funct. Spaces 2019, Article ID 6926107 (2019)

40. Wang, M.-K., Chu, Y.-M.: Refinements of transformation inequalities for zero-balanced hypergeoemtrci functions. Acta Math. Sci. 37B(3), 607–622 (2017)

41. Yang, Z.-H., Zhang, W., Chu, Y.-M.: Sharp Gautschi inequality for parameter $$0< p<1$$ with applications. Math. Inequal. Appl. 20(4), 1107–1120 (2017)

42. Qian, W.-M., Zhang, X.-H., Chu, Y.-M.: Sharp bounds for the Toader–Qi mean in terms of harmonic and geometric means. J. Math. Inequal. 11(1), 121–127 (2017)

43. Latif, M.A., Dragomir, S.S., Momoniat, E.: Inequalities of Hermite–Hadamard type for n-times differentiable $$(\alpha , m)$$-logarithmically convex functions. J. Comput. Anal. Appl. 25(4), 751–759 (2018)

44. Wang, S.-H., Wu, S.-H.: Some new Hermite–Hadamard type inequalities for operator m-convex and $$(\alpha , m)$$-convex functions on the co-ordinates. J. Comput. Anal. Appl. 25(3), 474–487 (2018)

45. Xu, H.-Z., Chu, Y.-M., Qian, W.-M.: Sharp bounds for the Sándor–Yang means in terms of arithmetic and contra-harmonic means. J. Inequal. Appl. 2018, Article ID 127 (2018)

46. Song, Y.-Q., Adil Khan, M., Zaheer Ullah, S., Chu, Y.-M.: Integral inequalities involving strongly convex functions. J. Funct. Spaces 2018, Article ID 6595921 (2018)

47. Adil Khan, M., Begum, S., Khurshid, Y., Chu, Y.-M.: Ostrowski type inequalities involving conformable fractional integrals. J. Inequal. Appl. 2018, Article ID 70 (2018)

48. Adil Khan, M., Chu, Y.-M., Kashuri, A., Liko, R., Ali, G.: Conformable fractional integrals versions of Hermite–Hadamard inequalities and their generalizations. J. Funct. Spaces 2018, Article ID 6928130 (2018)

49. Huang, T.-R., Han, B.-W., Ma, X.-Y., Chu, Y.-M.: Optimal bounds for the generalized Euler–Mascheroni constant. J. Inequal. Appl. 2018, Article ID 118 (2018)

50. Huang, T.-R., Tan, S.-Y., Ma, X.-Y., Chu, Y.-M.: Monotonicity properties and bounds for the complete p-elliptic integrals. J. Inequal. Appl. 2018, Article ID 239 (2018)

51. Dragomir, S.S.: On the Hadamard’s inequality for convex functions on the co-ordinates in a rectangle from the plane. Taiwan. J. Math. 5(4), 775–788 (2001)

52. Marshall, A.W., Olkin, I.: Inequalities: Theory of Majorization and Its Applications. Academic Press, New York (1979)

53. Hardy, G.H., Littlewood, J.E., Pólya, G.: Inequalities. Cambridge University Press, Cambridge (1952)

54. Karamata, J.: Sur une inégalité relative aux fonctions convexes. Publ. Math. Univ. Belgrade 1, 145–148 (1932)

55. Fuchs, L.: A new proof of an inequality of Hardy–Littlewood–Pólya. Mat. Tidsskr., B 1947, 53–54 (1947)

56. Bullen, P.S., Vasić, P.M., Stanković, L.: A problem of A. Oppenheim. Publ. Elektroteh. Fak. Univ. Beogr., Ser. Mat. Fiz. 412(460), 21–30 (1973)

57. Dragomir, S.S.: Some majorisation type discrete inequalities for convex functions. Math. Inequal. Appl. 7(2), 207–216 (2004)

58. Biernacki, M.: Sur une inégalité entre les intégrales due à Tchébyscheff. Ann. Univ. Mariae Curie-Skłodowska 5, 23–29 (1951)

59. Burkill, H., Mirsky, L.: Comments on Chebycheff’s inequality. Period. Math. Hung. 6, 3–16 (1975)

60. Latif, N., Pečarić, J.E., Perić, I.: On discrete Favard’s and Berwald’s inequalities. Commun. Math. Anal. 12(2), 34–57 (2012)

61. Mitrinović, D.S., Vasić, P.M.: Analytic Inequalities. Springer, New York (1970)

62. Wu, S.-H., Adil Khan, M., Basir, A., Saadati, R.: Some majorization integral inequalities for functions defined on rectangles. J. Inequal. Appl. 2018, Article ID 146 (2018)

63. Adil Khan, M., Latif, N., Pečarić, J.E.: Generalization of majorization theorem. J. Math. Inequal. 9(3), 847–872 (2015)

64. Adil Khan, M., Latif, N., Pečarić, J.E.: Generalization of majorization theorem via Abel–Gontscharoff polynomial. Rad Hrvat. Akad. Znan. Umjet. Mat. Znan. 19(523), 91–116 (2015)

65. Adil Khan, M., Khan, J., Pečarić, J.E.: Generalization of Jensen’s and Jensen–Steffensen’s inequalities by generalized majorization theorem. J. Math. Inequal. 11(4), 1049–1074 (2017)

## Funding

This work was supported by the Natural Science Foundation of China (Grant Nos. 11701176, 11626101, 11601485) and the Natural Science Foundation of Fujian Province (Grant No. 2016J01023).

## Author information

Authors

### Contributions

All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.

### Corresponding author

Correspondence to Yu-Ming Chu.

## Ethics declarations

### Competing interests

The authors declare that they have no competing interests. 