I've got stuck at these problem :

Prove that for any $x \geq 1$ and any $y \geq 1$ the following inequality is true: $$ {x}{\sqrt{y - 1}} + {y}{\sqrt{x - 1}} \leq xy. $$

The first thing that came into my mind was the AM-GM inequality (the extended version - I don't know if it has a specific name, other than mean inequality): $$ HM \leq GM \leq AM \leq SM $$ where $HM$, $GM$, $AM$, and $SM$ refer to the harmonic, geometric, arithmetic, and square mean, respectively.

I assume that's what I need to use, but I didn't get it right.

I would apreciate some hints.


  • $\begingroup$ The extended version of AM-GM is called the power mean inequality i think. $\endgroup$ – Yiyuan Lee Oct 3 '15 at 9:08
  • $\begingroup$ You mean $x,y>1$. $\endgroup$ – user236182 Oct 3 '15 at 9:10
  • $\begingroup$ The inequality is false. Let $x=y=1.25$. $\endgroup$ – user236182 Oct 3 '15 at 9:13
  • 3
    $\begingroup$ Let $a^2 = x-1, b^2 = y-1$, then you want to show $$a+\frac1a+b+\frac1b \le (a^2+1)(b^2+1)$$ which will not hold for any $a, b \in (0, 1)$ as the LHS is clearly larger than $4$, while the RHS is lesser than $4$. $\endgroup$ – Macavity Oct 3 '15 at 9:35
  • $\begingroup$ @Macavity someone edited the post but changhed the inequality. There aren't any fractions line. I've just edited the post to the correct form. Please check it out. $\endgroup$ – scummy Oct 3 '15 at 10:28

With the current version of the problem again let $a^2 = x-1, b^2 = y-1$, then you want to show $$(a^2+1)b + (b^2+1)a \le (a^2+1)(b^2+1)$$ $$\iff (b^2+1-b)a^2-(b^2+1) \cdot a+(b^2+1-b) \ge 0$$ $$\iff a^2-\frac{b^2+1}{b^2+1-b} \cdot a+1 \ge 0$$ $$\iff \left(a-\frac{b^2+1}{2(b^2-b+1)} \right)^2+\frac{(b-1)^2(3b^2-2b+3)}{4(b^2-b+1)^2} \ge 0$$ which is obvious as $3b^2-2b+3 > 0$. Equality is possible iff $a=b=1 \implies x=y=2$.

  • $\begingroup$ I have a question : what if $x,y ∉ Q$? Then, can I still say that $a^2=x−1$ and $b^2=y−1$? What kind of numbers are $a,b$? $\endgroup$ – scummy Oct 5 '15 at 15:49
  • $\begingroup$ @scummy Never assumed $x,y$ are rational! this works for all Real numbers $x,y\ge 1$. Then $a,b$ are non-negative reals. $\endgroup$ – Macavity Oct 5 '15 at 15:53
  • $\begingroup$ Thanks, I've cleared up! Silly question. $\endgroup$ – scummy Oct 5 '15 at 15:57

As Macavity correctly points out, the transformation $x=a^2+1, y=b^2+1$ reveals that the inequality does not hold for $x,y \in (1,2)$ (the LHS being $>4$, while the RHS being $<4$), and the same reasoning proves the inequality for $x,y \in (2,\infty)$.

For $x \in (1,2)$ and $y \in (2 ,\infty)$, the valitity depends on the concrete values of $x,y$.

  • $\begingroup$ Someone edited the problem and changed something. There aren't any fractions. I've just edited the post (again). Please check it out. $\endgroup$ – scummy Oct 3 '15 at 10:24

The given inequality is equivalent to:

$$ x,y\geq 1\Longrightarrow \frac{1}{x\sqrt{x-1}}+\frac{1}{y\sqrt{y-1}}\leq 1$$ but that cannot hold if $x$ or $y$ is close to one, since the function $f(x)=\frac{1}{x\sqrt{x-1}}$ is unbounded in a right neighbourhood of $1$.


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.