4

Let $a,b,c>0$ such that $b<\sqrt{ac}$, $c<\frac{2ab}{a+b}$. Show that $b^2+c^2-a^2\leq bc$.

I tried to construct a triangle with $a,b,c$ and to apply The cosine rule, but I am not sure that it's possible to construct it and also I have no idea how to prove that an angle it's greater than $60^{\circ}$.

ale
  • 1,744

4 Answers4

4

Both conditions and the hypothesis are homogenous, so we can scale the variables until $a=1$. Then the problem is equivalent to the implication: $$b<\sqrt{c}\land c<\frac{2b}{1+b}\Rightarrow b^2+c^2\le1+bc.$$ New conditions are more convenient and can be rewritten to: $$b^2<c<\frac{2b}{1+b}.$$ Let's devide the condition by $b$ and the hypothesis by $b^2$, the problem is equivalen to: $$b<\frac{c}{b}<\frac{2}{1+b}\Rightarrow1+\Big(\frac{c}{b}\Big)^2\le\frac{c}{b}+\frac{1}{b^2}.$$ Let's introduce $r:=\frac{c}{b}$. The problem is equivalent to: $$b<r<\frac{2}{1+b}\Rightarrow1-r+r^2\le\frac{1}{b^2}.$$ The condition implies: $$b<\frac{2}{1+b}\Rightarrow b+b^2<2\Rightarrow\Big(b+\frac{1}{2}\Big)^2<\frac{9}{4}\Rightarrow b+\frac{1}{2}<\frac{3}{2}\Rightarrow b<1.$$ Returning the limitation $b>0$ together with the $\frac{2}{1+b}$ being decreasing with the supremum at $b=0$ we get: $$0<b<1\land0<r<2.$$ The case $r\le1$ is easy: $$1-r+r^2=\frac{3}{4}+\Big(\frac{1}{2}-r\Big)^2\le\frac{3}{4}+\max_{0<r\le1}\Big|\frac{1}{2}-r\Big|^2=\frac{3}{4}+\Big(\frac{1}{2}\Big)^2=1=\frac{1}{1}<\frac{1}{b^2}.$$ For $r>1$ we have $$\frac{d}{dr}(1-r+r^2)=2r-1>1>0,$$ so the expression $1-r+r^2$ increases with $r$ and we have $$1-r+r^2<1-\frac{2}{1+b}+\Big(\frac{2}{1+b}\Big)^2=\frac{3+b^2}{(1+b)^2}=\frac{1}{b^2}\frac{(3+b^2)b^2}{(1+b)^2}=$$ $$\frac{1}{b^2}\frac{b^2+2b^2+b^4}{(1+b)^2}<\frac{1}{b^2}\frac{1+2b+b^2}{(1+b)^2}=\frac{1}{b^2}.$$

donaastor
  • 1,677
4

Another solution.

First, we find $c<\frac{2ab}{a+b}\leq \frac{2\left(\frac{a+b}{2}\right)^2}{a+b}=\frac{a+b}{2}$ by applying AM-GM to $ab$. We also find $b<\sqrt{ac}\leq \frac{a+c}{2}$.

Now if $(b+c)\leq a$, this implies $(b+c)^2\leq a^2 \Rightarrow b^2+c^2-a^2\leq -2bc<bc$, and we are done.

If not, then $b+c>a$ we'll have a triangle $\triangle ABC$. By the law of cosine,

$$a^2 = b^2+c^2-2bc\cos{A} \Rightarrow b^2+c^2-a^2=2bc\cos{A}$$ we only need to consider cases where $\angle A<60^{\circ}$ which would make our statement false.

Now $\angle A<60^{\circ}$ implies $a$ can't be the longest side, and the condition $b<\sqrt{ac}$ implies $b$ can't be the longest side. Therefore $c$ must be the longest side. However $c<\frac{a+b}{2}$, so that contradicts the fact we have a triangle.

Giant Ray
  • 156
  • @amWhy The fact that its more elementary and relates to the triangle? – Giant Ray Dec 13 '22 at 20:54
  • 2
    @amWhy Do users in the stack exchange all need to touch grass or what? I provided my own solution without knowledge and the other two and all I get is getting repeatedly questioned about the value of my proof, about whether it has any merits or what. if you hate seeing "repeat" solutions, just downvote my answer and be done, will you? I come here to solve problems and have fun, can't care less about some imaginary internet points. – Giant Ray Dec 13 '22 at 21:10
  • @amWhy I was really annoyed because I had most of it by noon. I needed to get some other stuff out of the way. And I seemed to be the one getting repeatedly questioned. You said "provide instead another proof", because I didn't post mine first. And no, I've been lurking for a few years. If I see my approach being taken in existing answers, I will refrain from posting. We can disagree on whether my proof is sufficiently different than the other two to merit being posted. Have a wonderful rest of your day. – Giant Ray Dec 13 '22 at 21:30
  • No offense intended. I can related to "lurking" for a time, before becoming active. I will delete my initial comments. Is that enough" I thank you for the time you spent observing. Again, I appreciate your answer. – amWhy Dec 13 '22 at 21:32
  • @amWhy sorry about the moment of anger. Was really annoyed for a moment. No offense intended from my side as well. :) – Giant Ray Dec 13 '22 at 21:37
  • :) I look forward to your continued participation!! – amWhy Dec 13 '22 at 21:38
  • 1
    I am puzzled by solutions like this. I never understood how people come up with them. In my answer you can clearly see how I'm trying to "clear" the problem from extra details and dig up the simplest equivalent of the exercise. Can you say something about how you construct a solution like this? Like, what are your mid-step half-solutions and what were your targets in these stages? – donaastor Dec 13 '22 at 22:13
4

Using $ac - b^2 > 0$ and $2ab - (a + b)c > 0$, we have \begin{align*} &bc - (b^2 + c^2 - a^2)\\ >\,&bc - (b^2 + c^2 - a^2) - (ac - b^2) - [2ab - (a + b)c]\\ =\,& a^2 - 2ab + 2bc - c^2\\ =\,&(a - c)(a + c - 2b)\\ >\,& 0 \end{align*} where we have used $a > c$ and $a + c > 2b$.
(Explanations: By AM-GM, we have $2b < 2\sqrt{ac} \le a+c$. Also, by AM-GM, we have $$c < \frac{2ab}{a + b} \le \frac{2ab}{2\sqrt{ab}} = \sqrt{ab} < \sqrt{a \sqrt{ac}}$$ which results in $c < a$.)

We are done.

River Li
  • 37,323
3

Let $a<\sqrt{b^2-bc+c^2}.$

Thus, $$b^2<ac<c\sqrt{b^2-bc+c^2},$$ which gives $$(b-c)(b^2(b+c)+c^3)<0$$ or $$b<c.$$ Also, since $$a(2b-c)>bc,$$ we obtain: $$2b-c>0$$ and $$\sqrt{b^2-bc+c^2}>a>\frac{bc}{2b-c},$$ which gives $$(b-c)(4b^3-4b^2c+4bc^2-c^3)>0$$ or $$4b^3-4b^2c+4bc^2-c^3<0.$$ But $$4b^3-4b^2c+4bc^2-c^3=b(2b-c)^2+c^2(2b-c)+c^2b>0,$$ which is a contradiction.