4
$\begingroup$

Do you know proofs or references for the following inequality:

There exists a positive constant $C>0$ such that for any complex numbers $a_1,\ldots,a_n$

$$ |a_1|+\cdots+|a_n| \leq C\sup_{z_1^3=1,\ldots,z_n^3=1 } |a_1z_1+\cdots + a_n z_n| $$ where the supremum is taken over the complex numbers $z_1,\ldots,z_n$ such that $z_1^3=1,\ldots,z_n^3=1$?

$\endgroup$
5
$\begingroup$

The strategy is to rotate the $a$ values so that they are roughly pointing in the same direction. More formally, for each complex $a$ choose a third root of unity $z$ so that the argument $\theta $ of $az$ lies between plus and minus $\pi/3$.

Then \begin{eqnarray*} |a_1z_1+\cdots +a_nz_n|&\geq&|\mbox{Re}(a_1z_1+\cdots +a_nz_n)| \cr &=&|a_1z_1|\cos(\theta_1)+\cdots +|a_nz_n|\cos(\theta_n) \cr &\geq &(|a_1|+\cdots +|a_n|)\cos(\pi/3)\cr &=&{1\over 2}\ (|a_1|+\cdots +|a_n|)\cr \end{eqnarray*}

This gives your result with $C=2$.

$\endgroup$
  • $\begingroup$ It seems to me that we must take the the "furthest away from the argument" and not the "closest argument". What do you think about this? $\endgroup$ – Zouba Nov 11 '11 at 15:46
  • $\begingroup$ You are right, I didn't describe it well. I will modify my answer. $\endgroup$ – user940 Nov 11 '11 at 15:51
4
$\begingroup$

Notice that the mapping $\|\cdot\|_* \colon \mathbb{C}^n \to \mathbb{R}$ given by $$\|(a_1,\dots,a_n)\|_* = \sup_{z_1^3=z_2^3=\dots=1} |a_1z_1 + \dots + a_n z_n|$$ defines a norm. The result then follows by the equivalence of norms in finite dimensional vector spaces.

The triangle inequality is easily seen to be true: Just use the ordinary triangle inequality. To prove that if $\|(a_1,\dots,a_n)\|_* = 0$ then $(a_1,\dots,a_n) = 0$ notice that if one of the coordinates (let's say $a_1$) wasn't $0$, then there would be three different numbers: $a_1 z_1 + a_2 + \dots + a_n$, $a_1 z_2 + a_2 + \dots + a_n$ and $a_1 z_3 + a_2 + \dots + a_n$ one of which is non-zero and thus the sup we get would be non-zero also.

$\endgroup$
  • $\begingroup$ Does the op want a constant independent of $n$, though? $\endgroup$ – David Mitra Nov 11 '11 at 15:14
  • $\begingroup$ Yes, the constant $C$ must be independent of $n$. $\endgroup$ – Zouba Nov 11 '11 at 15:21
  • $\begingroup$ @David Mitra: A good point. $\endgroup$ – J. J. Nov 11 '11 at 15:23

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.