On an inequality

Abstract

Authors

Keywords

?

Paper coordinates

T. Popoviciu, Sur une inegalité, Mathematica, 23 (1947-1948), pp. 127-128 (in French).

PDF

About this paper

Journal

Mathematica

Publisher Name
DOI
Print ISSN
Online ISSN

[MR0027024]

google scholar link

??

Paper (preprint) in HTML form

1948 b -Popoviciu- Mathematica - On an inequality
Original text
Rate this translation
Your feedback will be used to help improve Google Translate

ON AN INEQUALITY

BY

TIBERIU POPOVICIU

Requested on April 27, 1948
In the "New Annals of Mathematics" ( 1 ) ( 1 ) ^((1)){ }^{(1)}(1)LA Le Cointe demonstrates that if has 1 , has 2 , , has 2 m has 1 , has 2 , , has 2 m a_(1)^('),a_(2)^('),dots,a_(2m)^(')a_{1}^{\prime}, a_{2}^{\prime}, \ldots, a_{2 m}^{\prime}has1,has2,,has2mare the 2 m 2 m 2m2 m2mpositive numbers has 1 , has 2 , , has 2 m has 1 , has 2 , , has 2 m a_(1),a_(2),dots,a_(2m)a_{1}, a_{2}, \ldots, a_{2 m}has1,has2,,has2marranged in non-descending (or non-ascending) order, we have
has 1 has 2 has m + has m + 1 has m + 2 has 2 m has 1 has 2 has m + has m + 1 has m + 2 has 2 m has 1 has 2 has m + has m + 1 has m + 2 has 2 m has 1 has 2 has m + has m + 1 has m + 2 has 2 m a_(1)^(')a_(2)^(')dotsa_(m)^(')+a_(m+1)^(')a_(m+2)^(')dotsa_(2m)^(') >= a_(1)a_(2)dotsa_(m)+a_(m+1)a_(m+2)dotsa_(2m)a_{1}^{\prime} a_{2}^{\prime} \ldots a_{m}^{\prime}+a_{m+1}^{\prime} a_{m+2}^{\prime} \ldots a_{2 m}^{\prime} \geqq a_{1} a_{2} \ldots a_{m}+a_{m+1} a_{m+2} \ldots a_{2m}has1has2hasm+hasm+1hasm+2has2mhas1has2hasm+hasm+1hasm+2has2m
We will give a generalization of this property by demonstrating the
Theorem. - Let φ ( x ) , ψ ( x ) φ ( x ) , ψ ( x ) varphi(x),psi(x)\varphi(x), \psi(x)φ(x),ψ(x)two continuous and increasing functions in the interval. ( has , b ) ( has , b ) (a,b)(a, b)(has,b)And φ 1 , ψ 1 φ 1 , ψ 1 varphi^(-1),psi^(-1)\varphi^{-1}, \psi^{-1}φ1,ψ1their inverse functions. Let us set
M ( x 1 , x 2 , , x n ) = φ 1 ( i = 1 n φ ( x i ) n ) N ( x 1 , x 2 , , x m ) = ψ 1 ( i = 1 m ψ ( x i ) m ) M x 1 , x 2 , , x n = φ 1 i = 1 n φ x i n N x 1 , x 2 , , x m = ψ 1 i = 1 m ψ x i m {:[M(x_(1),x_(2),dots,x_(n))=varphi^(-1)((sum_(i=1)^(n)varphi(x_(i)))/(n))],[N(x_(1),x_(2),dots,x_(m))=psi^(-1)((sum_(i=1)^(m)psi(x_(i)))/(m))]:}\begin{aligned} & \mathrm{M}\left(x_{1}, x_{2}, \ldots, x_{n}\right)=\varphi^{-1}\left(\frac{\sum_{i=1}^{n} \varphi\left(x_{i}\right)}{n}\right) \\ & \mathrm{N}\left(x_{1}, x_{2}, \ldots, x_{m}\right)=\psi^{-1}\left(\frac{\sum_{i=1}^{m} \psi\left(x_{i}\right)}{m}\right) \end{aligned}M(x1,x2,,xn)=φ1(i=1nφ(xi)n)N(x1,x2,,xm)=ψ1(i=1mψ(xi)m)
Let's still be has 1 , has 2 , , has n m n m has 1 , has 2 , , has n m n m a_(1),a_(2),dots,a_(nm)nma_{1}, a_{2}, \ldots, a_{nm} nmhas1,has2,,hasnmnmnumbers belonging to the interval ( has , b ) ( has , b ) (a,b)(a, b)(has,b)And has 1 , has 2 , , has n m has 1 , has 2 , , has n m a_(1)^('),a_(2)^('),dots,a_(nm)^(')a_{1}^{\prime}, a_{2}^{\prime}, \ldots, a_{nm}^{\prime}has1,has2,,hasnmthese same numbers arranged in non-increasing (or non-descending) order.
Then we have inequality
(1)
M ( N ( has 1 , has 2 , , has m ) , N ( has m + 1 , has m + 2 , , has 2 m ) , N ( has ( n 1 ) m + 1 , has ( n 1 ) m + 2 , , has n m ) ) resp . M ( N ( has 1 , has 2 , , has m ) , N ( has m + 1 , has m + 2 , , has 2 m ) , , N ( has ( n 1 ) m + 1 , has ( n 1 ) m + 2 , , has n m ) ) M N has 1 , has 2 , , has m , N has m + 1 , has m + 2 , , has 2 m , N has ( n 1 ) m + 1 , has ( n 1 ) m + 2 , , has n m resp . M N has 1 , has 2 , , has m , N has m + 1 , has m + 2 , , has 2 m , , N has ( n 1 ) m + 1 , has ( n 1 ) m + 2 , , has n m {:[M((N)(a_(1)^('),a_(2)^('),dots,a_(m)^(')),N(a_(m+1)^('),a_(m+2)^('),dots,a_(2m)^(')),dots:}],[{:N(a_((n-1)m+1)^('),a_((n-1)m+2)^('),dots,a_(nm)^('))) >= resp. <= ],[M((N)(a_(1),a_(2),dots,a_(m)),N(a_(m+1),a_(m+2),dots,a_(2m)),dots,:}],[{:N(a_((n-1)m+1),a_((n-1)m+2),dots,a_(nm)))]:}\begin{gathered} \mathrm{M}\left(\mathrm{~N}\left(a_{1}^{\prime}, a_{2}^{\prime}, \ldots, a_{m}^{\prime}\right), \mathrm{N}\left(a_{m+1}^{\prime}, a_{m+2}^{\prime}, \ldots, a_{2 m}^{\prime}\right), \ldots\right. \\ \left.\mathrm{N}\left(a_{(n-1) m+1}^{\prime}, a_{(n-1) m+2}^{\prime}, \ldots, a_{n m}^{\prime}\right)\right) \geqq \mathrm{resp} . \leqq \\ \mathrm{M}\left(\mathrm{~N}\left(a_{1}, a_{2}, \ldots, a_{m}\right), \mathrm{N}\left(a_{m+1}, a_{m+2}, \ldots, a_{2 m}\right), \ldots,\right. \\ \left.\mathrm{N}\left(a_{(n-1) m+1}, a_{(n-1) m+2}, \ldots, a_{n m}\right)\right) \end{gathered}M( N(has1,has2,,hasm),N(hasm+1,hasm+2,,has2m),N(has(n1)m+1,has(n1)m+2,,hasnm))resp.M( N(has1,has2,,hasm),N(hasm+1,hasm+2,,has2m),,N(has(n1)m+1,has(n1)m+2,,hasnm))
depending on the function φ ( ψ 1 ) φ ψ 1 varphi(psi^(-1))\varphi\left(\psi^{-1}\right)φ(ψ1)is non-concave resp. non-convex.
( 1 1 ^(1){ }^{1}1) T. 2. 372-374 (1843).
Let us designate by f f fffthe function φ ( ψ 1 ) φ ψ 1 varphi(psi^(-1))\varphi\left(\psi^{-1}\right)φ(ψ1)and let us suppose, to fix the ideas, that this function is non-concave. The functions φ 1 , ψ 1 φ 1 , ψ 1 varphi^(-1),psi^(-1)\varphi^{-1}, \psi^{-1}φ1,ψ1being also increasing, the inequality to be demonstrated amounts to
i = 1 n f ( j = 1 m ψ ( a ( i 1 ) m + j ) m m ) i = 1 n f ( j = 1 m ψ ( a ( i 1 ) m + j ) m m ) i = 1 n f j = 1 m ψ a ( i 1 ) m + j m m i = 1 n f j = 1 m ψ a ( i 1 ) m + j m m sum_(i=1)^(n)f((sum_(j=1)^(m)(psi(a_((i-1)m+j)^(')))/(m))/(m)) >= sum_(i=1)^(n)f((sum_(j=1)^(m)(psi(a_((i-1)m+j)))/(m))/(m))\sum_{i=1}^{n} f\left(\frac{\sum_{j=1}^{m} \frac{\psi\left(a_{(i-1) m+j}^{\prime}\right)}{m}}{m}\right) \geqq \sum_{i=1}^{n} f\left(\frac{\sum_{j=1}^{m} \frac{\psi\left(a_{(i-1) m+j}\right)}{m}}{m}\right)i=1nf(I=1mψ(has(i1)m+I)mm)i=1nf(I=1mψ(has(i1)m+I)mm)
which is a consequence of the following property of Messrs. GH Hardy, JE Littlewood and G. Polya ( 2 ) 2 (^(2))\left({ }^{2}\right)(2).
If f ( x ) f ( x ) f(x)f(x)f(x)is a non-concave function and if ( x i x i + 1 , y i y i + 1 , i = 1 , 2 , , n 1 ) x i x i + 1 , y i y i + 1 , i = 1 , 2 , , n 1 ) (x_(i) >= x_(i+1),y_(i) >= y_(i+1),i=:}1,2,dots,n-1)\left(x_{i} \geq x_{i+1}, y_{i} \geqq y_{i+1}, i=\right. 1,2, \ldots, n-1)(xixi+1,yiyi+1,i=1,2,,n1)
x 1 + x 2 + + x i y 1 + y 2 + + y i , i = 1 , 2 , , n 1 , x 1 + x 2 + + x n = y 1 + y 2 + + y n x 1 + x 2 + + x i y 1 + y 2 + + y i , i = 1 , 2 , , n 1 , x 1 + x 2 + + x n = y 1 + y 2 + + y n {:[x_(1)+x_(2)+cdots+x_(i) >= y_(1)+y_(2)+cdots+y_(i)","quad i=1","2","dots","n-1","],[x_(1)+x_(2)+cdots+x_(n)=y_(1)+y_(2)+cdots+y_(n)]:}\begin{aligned} & x_{1}+x_{2}+\cdots+x_{i} \geqq y_{1}+y_{2}+\cdots+y_{i}, \quad i=1,2, \ldots, n-1, \\ & x_{1}+x_{2}+\cdots+x_{n}=y_{1}+y_{2}+\cdots+y_{n} \end{aligned}x1+x2++xiy1+y2++yi,i=1,2,,n1,x1+x2++xn=y1+y2++yn
we have inequality
i = 1 n f ( x i ) i = 1 n f ( y i ) i = 1 n f x i i = 1 n f y i sum_(i=1)^(n)f(x_(i)) >= sum_(i=1)^(n)f(y_(i))\sum_{i=1}^{n} f\left(x_{i}\right) \geqq \sum_{i=1}^{n} f\left(y_{i}\right)i=1nf(xi)i=1nf(yi)
In the case where the function φ ( ψ 1 ) φ ψ 1 varphi(psi^(-1))\varphi\left(\psi^{-1}\right)φ(ψ1)is convex (resp. concave) it is easy to find the cases where the equality is valid in (1).
The property of Le Cointe corresponds to the case where n = 2 , φ = x , ψ = log x n = 2 , φ = x , ψ = log x n=2,varphi=x,psi=log xn=2, \varphi=x, \psi=\log xn=2,φ=x,ψ=logx.
By particularizing the functions φ , ψ φ , ψ varphi,psi\varphi, \psiφ,ψwe find various specific statements.

  1. ( 2 ) ( 2 ) ^((2)){ }^{(2)}(2)„Some simple inequalities satisfied by convex function", Messenger of Math., 58, 145-152 (1929). See also "Inequalities", p. 89.
1948

Related Posts