2
$\begingroup$

Given $a>0$ and $ac-b^2>0$ show

$cy^2+a[(x+\frac{by}{a})^2-(\frac{by}{a})^2] > 0$

I'm completely confused about this, I've tried a few approaches. I end up getting stuck saying that I know $cy^2>0$ using the 2nd of the given inequalities, but I can't show the $a[...]$ part is >0 since all I know is x and y are non-zero.

Any guidance? Comes from a larger question about showing a symmetric matrix [a, b, b, c] is positive definite if that helps.

Thanks for any nudges :)

  • 3
    If you are workking with $\begin{pmatrix}a&b\\b&c\end{pmatrix}$ and the corresponding quadratic form $ax^2+2bxy+cy^2$, you should have there $-\frac{b^2y^2}{a^2}$ instead of $-\frac{by^2}{a}$ -- just as Myself writes in the answer bellow.2011-03-29
  • 0
    @Martin: good remark, I actually checked the latexcode to see what exactly was meant by the last square :-) Since it says \frac{by}{a}^2, I interpreted it as I did2011-03-29
  • 0
    Sorry about that - fixed the question now.2011-03-29

1 Answers 1

4

Rewrite it as $cy^2 + a\left((x+\tfrac{by}{a})^2 - \frac{b^2y^2}{a^2}\right) = a(x+\tfrac{by}{a})^2 + y^2 (\frac{ac}{a} - \frac{b^2}{a}) $.

Now use what you know and use that squares are always positive.

  • 0
    Brilliant. If you don't mind, how did you get to that so fast? I spent well over an hour and didn't see that. Is it possible to learn with practice or is it genuinely just something you looked at and saw? Edit: annoyingly, it does seem quite "obvious" now.2011-03-29
  • 0
    @Tim Green: It is definitely possible to learn with practice, even though I don't think there's much to learn in this particular case, you just fill in the missing hole between what you know and what you want to proof...2011-03-29
  • 2
    Completing the square is a standard way of proving the quadratic formula or finding where a quadratic is $0$. By the way, you need to exclude $x=0,y=0$ if you want a strict inequality.2011-03-29
  • 0
    Yep - one of the other features given was the vector x is non-zero so I can infer x and y cannot be zero at the same time.2011-03-29