Let $a_i$ be real positive numbers, $i=0, \dots, N-1$, with $N$ even.

Define sums $S_k = \sum_{i=0}^{N-1} a_i a_{i+k}$, for $k=0, \dots, N-1$, where the indices are understood to be cyclic, i.e. $a_{N+j} = a_j$.

Prove or disprove: For all $k$: $S_k \ge S_{N/2}$.

Ideas: one approach to this might be the Rearrangement Inequality (RI). Without loss of generality, we can order the $\{a_i\}$ in such a way that they are ascending. Then amongst the $\{a_{i+k}\}$, only $\{a_{i+0}\}$ is ascending and for all other $k$, the $\{a_{i+k}\}$ are not ascending (as a consequence of the cyclic nature). So by the RI, the largest sum is $S_0$.

In the same vein, if one finds that $S_{N/2}$ is "most disturbing" to the ascending order, then using the RI would show the claim. I didn't manage to do that.


The conjecture is false.

Let $N=4$

and $a_0=1,a_1=10,a_2=11,a_3=1000$

Then $S_1=S_3=12120<S_2=20022$

  • 1
    $\begingroup$ Nice counterexample! Thanks! $\endgroup$ – Andreas Sep 27 '17 at 9:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.