A Family of Mappings Associated with Hadamard's Inequality on a Hypercube
FuLi Wang1
Academic Editor: V. Kravchenko, S. Pilipovic
Received24 Dec 2010
Accepted16 Jan 2011
Published16 Feb 2011
Abstract
This paper provides a method to generalize and refine the Hadamard's inequality on a hypercube in -dimensional Euclidean space by establishing a family of mappings.
1. Introduction
For a convex function defined on an open interval of real numbers and with , then the following inequality:
is known as the classical Hadamardβs inequality for the convex function.
In [1], Dragomir established the following mapping:
then the mapping is convex, increasing on , and for all , we have
which is a refinement to the left side of (1.1).
In [2], Yang and Hong established the following mapping:
then the mapping is convex, increasing on , and for all , we have
which is a refinement to the right side of (1.1).
In this paper, we first establish a mapping in Section 2, defined by
then the mapping is convex, increasing on , and for all we have
Furthermore, there is a point so that . Thus, (1.1) has been proved based on the properties of , and the inequality (1.7) is a refinement to the both sides of (1.1).
In Section 3, we also use an analogue of (1.6) to obtain the following Hadamardβs type inequality on a square:
which is a generalization of (1.1) to two-dimensional Euclidean space.
In Section 4, by establishing another analogue of (1.6) on a hypercube, the following Hadamardβs inequality is also pointed out:
which is a generalization of (1.1) to -dimensional Euclidean space.
All the above Hadamardβs inequalities (1.1), (1.8), and (1.9) have a uniform format
Thus, by establishing a family of the mappings, we have provided a method to get Hadamardβs inequality of the convex function on a hypercube in -dimensional Euclidean space. The motivation for the present work is from many generalizations and refinements of the classical inequality (1.1) in [1β8].
2. Properties of the Mapping
The main properties of the mapping defined by (1.6) are embodied in the following theorem.
Theorem 2.1. If is a convex function defined on an open interval of real numbers, with , and the mapping is defined by (1.6), then, one has (i)the mapping is convex on ,(ii)one has the bounds:
(iii)the mapping is increasing on ,(iv)there is a point so that
(v)one obtains, the classical inequality (1.1) and has a refinement of (1.1):
Proof. (i) Let and with , then we have
which proves the convexity of on . (ii) By the convexity of on the interval , we have
the bounds (2.1) hold. (iii) Let . By the convexity of the mapping , we have
Since we have proved that for all in (ii), the monotonicity of is established. (iv) We define the function on by
then
According to the Rolleβs mean-value theorem, there is a point so that
Thus, (2.2) holds. (v) Considering the properties (ii), (iii), and (iv) above, we have proved the Hadamardβs inequality (1.1) and the inequality (2.3).
3. The Hadamardβs Inequality on a Square
We consider a convex function with two variables defined on a square in two-dimensional Euclidean space.
Theorem 3.1. If is a convex function defined on an open region , the square
and the mapping A2 is defined by
which is a parameter curvilinear integral with respect to arc length, then one has (i)the mapping is convex on ,(ii)one has the bounds
(iii)the mapping is increasing on ,(iv)there is a point so that
(v)one has the Hadamardβs type inequality (1.8) and a refinement
Proof. For a fixed , let if , and if , we have
which indicates the mapping is continuous on . (i) Let and with . By the convexity of , we have
Similarly, we have
Thus,
which proves the convexity of on . (ii) By the convexity of and (3.6), we have
the last inequality holds owing to (1.1); as the convexity of we have
and note that
the bounds (3.3) hold. (iii) The monotonicity of follows as in the proof of Theorem 2.1, and we omit the detail. (iv) We define the function on by
then
Let and , we have
According to the Rolleβs mean-value theorem, there is a point so that
thus, (3.4) holds. (v) Considering the properties (ii), (iii), and (iv) above, we have proved the Hadamardβs type inequality (1.8) and the inequality (3.5).
4. Remark
In recent years, several authors have discussed the problem of extending the classical inequality (1.1) to a convex function on a general convex body of -dimensional space [6β9]. For the left side of (1.1) (i.e., Jensenβs inequality), the following inequality was showed in [6, 9] (also see [7]):
where is the barycenter of , and denotes -dimensional Lebesgueβs measure of .
It is a much more delicate problem to extend the right side of (1.1) on a general convex body of -dimensional space [7]. On a closed hyperball of the -dimensional space, de la Cal and CΓ‘rcamo [7] obtained the inequality
where denotes the volumes of , and denotes the areas of the hypersphere , for which was early obtained by Dragomir [4, 5] based on Calculus.
When the convex function was supposed to be differentiable on a closed convex body , the following inequality was proved in [9] based on Stokesβ Formula
where denotes the piecewisely smooth boundary of , denotes the barycenter of , and denotes the exterior unit normal vector in the point .
For a closed hyperball , the inequality (4.2) is able to be deduced from (4.3); and for a hypercube , the inequality (1.9) is also able to be deduced from (4.3); however, these are on the condition that is differentiable.
Now, let be a convex function defined on an open region , the hypercube , the mapping is defined by
which is a parameter surface integral with respect to area of dimension, and let
in the same way as the proof of Theorem 3.1, by the induction we will be able to obtain the inequality (1.9).
Some mappings associated to Hadamardβs inequality have been established in [1β5], which refine the classical Hadamardβs inequality. These mappings are based on Hadamardβs inequality. However, our method in this paper to establish mappings is different from theirs: we first proved (1.1) by the properties of and subsequently established and proved (1.8) based on (1.1). Thus, using the induction, we will be able to establish the mapping and prove Hadamardβs inequality on a hypercube.
References
S. S. Dragomir, βTwo mappings in connection to Hadamard's inequalities,β Journal of Mathematical Analysis and Applications, vol. 167, no. 1, pp. 49β56, 1992.
S. S. Dragomir and R. P. Agarwal, βTwo new mappings associated with Hadamard's inequalities for convex functions,β Applied Mathematics Letters, vol. 11, no. 3, pp. 33β38, 1998.
S. S. Dragomir, βOn Hadamard's inequality on a disk,β Journal of Inequalities in Pure and Applied Mathematics, vol. 1, no. 1, article 2, pp. 1β11, 2000.
S. S. Dragomir, βOn Hadamard's inequality for the convex mappings defined on a ball in the space and applications,β Mathematical Inequalities & Applications, vol. 3, no. 2, pp. 177β187, 2000.
J. de la Cal and J. Cárcamo, βMultidimensional Hermite-Hadamard inequalities and the convex order,β Journal of Mathematical Analysis and Applications, vol. 324, no. 1, pp. 248β261, 2006.
F.-L. Wang, βThe generalizations to several-dimensions of the classical Hadamard’s inequality,β Mathematics in Practice and Theory, vol. 36, no. 9, pp. 370β373, 2006 (Chinese).