Recently I had use for the following inequality, which was proven in a rather inelegant way:

For all $y \in [0,1]$ and all integers $0\leq i \leq s$, the following inequality holds: \begin{align*}\left(2+2y^i-4y^s+y^{2s-i}\right)^s \leq \left(2-y^s\right)^{2s-i}. \end{align*}

I now wonder:
1. Is this a special case of some known inequality?
2. Does anyone know some nice proof for this and similar inequalities in one or more variables?


Notice that at the endpoint cases $i=0, i=s$ we have equality. Therefore to prove the inequality it is enough to show that the map $$ i \mapsto \ln(2+2y^{i}-4y^{s}+y^{2s-i}), \quad 0\leq i \leq s $$ is convex. We have $$ \frac{d^{2}}{di^{2}} \ln(2+2y^{i}-4y^{s}+y^{2s-i})=\frac{2y^{2i}(\ln(y))^{2}(4y^{2(s-i)}-4y^{s-i}-2y^{3(s-i)}+y^{-i}(2+y^{2(s-i)}))}{(2+2y^{i}-4y^{s}+2y^{2s-i})^{2}} $$ If you denote $y^{s-i}=a$ where $0\leq a\leq 1$ then $$ 4y^{2(s-i)}-4y^{s-i}-2y^{3(s-i)}+y^{-i}(2+y^{2(s-i)}) \geq 4a^{2}-4a-2a^{3}+2+a^{2} $$ and the minimal value of the latter expression when $a \in [0,1]$ is $\frac{26}{27}> 0$

  • $\begingroup$ I'm confused. The LHS is convex in $i$, but so is the the RHS... So couldn't the RHS be more convex and thus the inequality fail? $\endgroup$ – Yaakov Baruch Aug 28 '17 at 20:04
  • 2
    $\begingroup$ I got it. after taking $\text{ln}()$ the RHS is linear... Nice proof! $\endgroup$ – Yaakov Baruch Aug 28 '17 at 20:16

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.