Hermite–Hadamard inequality

From HandWiki

In mathematics, the Hermite–Hadamard inequality, named after Charles Hermite and Jacques Hadamard and sometimes also called Hadamard's inequality, states that if a function ƒ : [ab] → R is convex, then the following chain of inequalities hold:

[math]\displaystyle{ f\left( \frac{a+b}{2}\right) \le \frac{1}{b - a}\int_a^b f(x)\,dx \le \frac{f(a) + f(b)}{2}. }[/math]

The inequality has been generalized to higher dimensions: if [math]\displaystyle{ \Omega \subset \mathbb{R}^n }[/math] is a bounded, convex domain and [math]\displaystyle{ f:\Omega \rightarrow \mathbb{R} }[/math] is a positive convex function, then

[math]\displaystyle{ \frac{1}{|\Omega|} \int_\Omega f(x) \, dx \leq \frac{c_n}{|\partial \Omega|} \int_{\partial \Omega} f(y) \, d\sigma(y) }[/math]

where [math]\displaystyle{ c_n }[/math] is a constant depending only on the dimension.

A corollary on Vandermonde-type integrals

Suppose that −∞ < a < b < ∞, and choose n distinct values {xj}nj=1 from (a, b). Let f:[a, b] → be convex, and let I denote the "integral starting at a" operator; that is,

[math]\displaystyle{ (If)(x)=\int_a^x{f(t)\,dt} }[/math].

Then

[math]\displaystyle{ \sum_{i=1}^n \frac {(I^{n-1}F)(x_i)}{\prod_{j\neq i}{(x_i-x_j)}}\leq \frac {1}{n!} \sum_{i=1}^n f(x_i) }[/math]

Equality holds for all {xj}nj=1 iff f is linear, and for all f iff {xj}nj=1 is constant, in the sense that

[math]\displaystyle{ \lim_{\{x_j\}_j\to\alpha}{\sum_{i=1}^n \frac {(I^{n-1}F)(x_i)}{\prod_{j\neq i}{(x_i-x_j)}}}=\frac{f(\alpha)}{(n-1)!} }[/math]

The result follows from induction on n.

References