Rearrangement inequality

In mathematics, the rearrangement inequality states that

:$x\_ny\_1\; +\; cdots\; +\; x\_1y\_nle\; x\_\{sigma\; (1)\}y\_1\; +\; cdots\; +\; x\_\{sigma\; (n)\}y\_nle\; x\_1y\_1\; +\; cdots\; +\; x\_ny\_n$

for every choice of real numbers

:$x\_1lecdotsle\; x\_nquad\; ext\{and\}quad\; y\_1lecdotsle\; y\_n$

and every permutation

:$x\_\{sigma(1)\},dots,x\_\{sigma(n)\}$

of "x"₁, . . ., "x_n". If the numbers are different, meaning that

:

then the lower bound is attained only for the permutation which reverses the order, i.e. σ("i") = "n" − "i" + 1 for all "i" = 1, ..., "n", and the upper bound is attained only for the identity, i.e. σ("i") = "i" for all "i" = 1, ..., "n".

Note that the rearrangement inequality makes no assumptions on the signs of the real numbers.

General rearrangement inequality

For an integer "n" ≤ 4 and any two sets of real numbers

: $x\_1lecdotsle\; x\_n\; ext\{\; and\; \}y\_1lecdotsle\; y\_n,,$

we have

:$x\_\{sigma\; (1)\}y\_1\; +\; cdots\; +\; x\_\{sigma\; (n)\}y\_n\; leq\; x\_\{pi(1)\}y\_1\; +\; cdots\; +\; x\_\{pi(n)\}y\_n$

as soon as the permutation "π" has smaller number of inversions (i.e., such pairs of indices $(i,j)$ that

Original publication [cite journal |author=Alan Wayne |title=Inequalities and inversions of order |journal=Scripta Mathematica |year=1946 |volume=12 |issue=2 |pages=164–169] of the general rearrangement inequality claimed its correctness for all integer "n" ≥ 1, however for "n" = 5 and up there exist counterexamples.

Applications

Many famous inequalities can be proved by the rearrangement inequality, such as the arithmetic mean – geometric mean inequality, the Cauchy–Schwarz inequality, and Chebyshev's sum inequality.

Proof

The lower bound follows by applying the upper bound to

:$-x\_nlecdotsle-x\_1.$

Therefore, it suffices to prove the upper bound. Since there are only finitely many permutations, there exists at least one for which

:$x\_\{sigma\; (1)\}y\_1\; +\; cdots\; +\; x\_\{sigma\; (n)\}y\_n$

is maximal. In case there are several permutations with this property, let σ denote one with the highest number of fixed points.

We will now prove by contradiction, that σ has to be the identity (then we are done). Assume that σ is not the identity. Then there exists a "j" in {1, ..., "n" − 1} such that σ("j") ≠ "j" and σ("i") = "i" for all "i" in {1, ..., "j" − 1}. Hence σ("j") > "j" and there exists "k" in {"j" + 1, ..., "n"} with σ("k") = "j". Now

:

:$0le(x\_\{sigma(j)\}-x\_j)(y\_k-y\_j).\; quad(2)$

Expanding this product and rearranging gives

:$x\_\{sigma(j)\}y\_j+x\_jy\_kle\; x\_jy\_j+x\_\{sigma(j)\}y\_k,,\; quad(3)$

hence the permutation

:

which arises from σ by exchanging the values σ("j") and σ("k"), has at least one additional fixed point compared to σ, namely at "j", and also attains the maximum. This contradicts the choice of σ.

If

:

then we have strict inequalities at (1), (2), and (3), hence the maximum can only be attained by the identity, any other permutation σ cannot be optimal.

References