Professional Documents
Culture Documents
A Generalization of The Rearrangement Inequality: 1 Main Result
A Generalization of The Rearrangement Inequality: 1 Main Result
Jan Holstermann
Abstract
In this article we present a generalization of the rearrangement inequal-
ity and show its applications by solving a USAJMO problem. The classi-
cal rearrangement inequality deals with two sequences ai ,P bi , i = 1, 2, ..., n,
n
which
Pn are both sorted
Pn in ascending order. It states that i=1 bi an−i+1 6
i=1 bi aσ(i) 6 i=1 bi ai for any permutation σ of the numbers 1, 2, ..., n.
In the generalization the bi aj terms are replaced by fi (aj ).
1 Main Result
To start off we will prove the version with only two summands on each side.
Lemma 1. Let a1 6 a2 be real numbers and fi : [a1 , a2 ] → R, i = 1, 2 be
functions, such that f10 (x) 6 f20 (x) for all x ∈ [a1 , a2 ]. We have
Proof. We have
Proof. Let di (j) = fi (aj ) − fi−1 (aj ). Our lemma translates to di (j) 6 di (k) for
any j 6 k. Thus, we have
n
X n
X
fi (aσ(i) ) = f1 (aσ(i) ) + d2 (σ(i)) + d3 (σ(i)) + ... + di (σ(i))
i=1 i=1
Xn n
X
= f1 (aσ(i) ) + dj (σ(j)) + dj (σ(j + 1)) + ... + dj (σ(n))
i=1 j=2
Xn Xn
6 f1 (aπ(i) ) + dj (π(j)) + dj (π(j + 1)) + ... + dj (π(n))
i=1 j=2
Xn
= fi (aπ(i) ).
i=1
Proof. For the upper bound let π(x) = x. For the lower bound let σ(x) =
n − x + 1.
Remark. Setting fi (x) = bi x we get the classical Rearrangement inequality.