7

IMO 2021, Problem 2.

Let $ n $ be an integer $ \ge 2$ and $x_1,x_2,...,x_n $ be $ n$ reals. prove that

$$\sum_{i=1}^n\sum_{j=1}^n\sqrt{|x_i-x_j|}\le \sum_{i=1}^n\sum_{j=1}^n\sqrt{|x_i+x_j|}$$

I wrote the left sum as $$2\sum_{i=2}^n\sum_{j=1}^{i-1}\sqrt{|x_i-x_j|}$$ and the right one as $$2\sum_{i=2}^n\sum_{j=1}^{i-1}\sqrt{|x_i+x_j|}+\sum_{i=1}^n\sqrt{2|x_i|}$$

but, it seems this is not a good starting point.

  • 2
    Thanks a lot for editing the Title. – hamam_Abdallah Jul 25 '21 at 22:48
  • You're welcome, @hamam_Abdallah. – Shaun Jul 25 '21 at 22:49
  • 1
    This question is very related, although the answer there might use more advanced techniques that what is expected to solve this problem. Also, is this 2021 IMO Problem 2? If so, that should be mentioned in this question. – JimmyK4542 Jul 25 '21 at 22:53
  • 1
    There’s bound to be an AOPS solution. This is Problem 2 of this year’s IMO. – Aphelli Jul 25 '21 at 22:54
  • IMO inequalities tend to require training and knowledge – and sometimes a healthy dose of luck. For a P2, this attempt is just not enough to lead you anywhere. Have you studied small cases? – Aphelli Jul 25 '21 at 23:06
  • @Mindlack Yes, at first i tried induction begining with $ n=2$ but i went nowhere. – hamam_Abdallah Jul 25 '21 at 23:10
  • 1
    If you’re interested in an elementary answer, you should be able to find sketches there: https://artofproblemsolving.com/community/c6h2625850p22697952 . But again – this is an IMO P2. This isn’t supposed to be easy to find. The Italian team was ranked 7th – and none of them got the solution in 4h30. Only one of the Chinese contestants did – and that team was ranked first! – Aphelli Jul 25 '21 at 23:20
  • @Mindlack Thank you very much. i appreciate. thanks again. – hamam_Abdallah Jul 25 '21 at 23:22
  • I can't say they are duplicates since this asks for a particular approach, but the same problem. – VIVID Jul 26 '21 at 10:14

1 Answers1

12

This is certainly not an answer expected in IMO, but the proof works for far more general cases, see this posting, for instance.

Note that

$$ \frac{1}{c} \int_{0}^{\infty} \frac{1-\cos (as)}{s^{3/2}} \, \mathrm{d}s = |a|^{1/2} \quad \text{for any} \quad a \in \mathbb{R}, $$

where $c = \int_{0}^{\infty} \frac{1-\cos t}{t^{3/2}} \, \mathrm{d}t \in (0, \infty)$. This is easily proved by substituting $t = |a|s$. Then

\begin{align*} &\sum_{i,j} |x_i + x_j|^{1/2} - \sum_{i,j} |x_i - x_j|^{1/2} \\ &= \frac{1}{c} \int_{0}^{\infty} \frac{1}{s^{3/2}}\biggl( \sum_{i,j} \cos((x_i- x_j)s) - \cos((x_i + x_j)s) \biggr) \, \mathrm{d}s \\ &= \frac{1}{c} \int_{0}^{\infty} \frac{1}{s^{3/2}}\biggl( \sum_{i,j} 2\sin(x_i s)\sin(x_j s) \biggr) \, \mathrm{d}s \\ &= \frac{1}{c} \int_{0}^{\infty} \frac{2}{s^{3/2}}\biggl( \sum_{i} \sin(x_i s) \biggr)^2 \, \mathrm{d}s \\ &\geq 0. \end{align*}

Of course, it would be fun to have a more elementary solution.

Clement C.
  • 67,323
Sangchul Lee
  • 167,468