3
$\begingroup$

This may seem like an elementary question, but bear with me; you'll find that it is actually quite hard. Consider the function $$f(x)=\frac{a_nx^n}{\sum_{i=0}^n a_ix^i}$$ with all $a_i\geq 0$ and $a_0=1$. Can you prove that the maximum derivative over $x\in (0,\infty)$, if it exists, is unique?

The usual elementary root counting techniques do not work since $f''(x)$ may have many sign changes. It is hard to control the number of zeros. Perhaps arguing that the numerator and denominator of $f(x)$ and $f'(x)$ are strictly monotone increasing may help.

$\endgroup$
3
  • $\begingroup$ What makes you think this is true? Take $f(x)=\frac{x^2}{x^2 + ax + 1}$ for $a$ between $1$ and $2$; the derivative has two local maxima, one positive and one negative. For $a=1.3$, the negative maximum is smaller; for $a=1.5$ the negative maximum is greater. $\endgroup$ Oct 22, 2014 at 7:03
  • $\begingroup$ I am taking the maximum derivative over $x\in (0,\infty)$. $\endgroup$
    – user60751
    Oct 22, 2014 at 17:24
  • $\begingroup$ oops, my mistake. $\endgroup$ Oct 22, 2014 at 23:55

1 Answer 1

3
$\begingroup$

This appears to be false. Consider $$ f(x)=\frac{\frac{x^{10}}{100}}{1+b+\frac{x^8}{2}+\frac{x^{10}}{100}} $$ for $b\in [5,10]$.

In this range, $f'(x)$ has two local maxima in $(0,\infty)$. One occurs at an $x$-value less than $2$ and the other occurs at an $x$-value between $3$ and $5$. At $b=5$ the second local maximum has a greater value; at $b=10$ the first local maximum has a greater value. So for some $b$ in that range the two maxima have equal value.

$\endgroup$
1
  • $\begingroup$ I'd be interested if there is any reason other than looking at a graph or two that this would be a reasonable thing to think, or if there's an obvious reason that dramatic counterexamples are hard to come by (I guess we think that as $f(0)=0$ and $f(x)<1$ for positive $x$ but approaches $1$ relatively quickly, there isn't much room for dramatic maxima in $f'(x)$). $\endgroup$ Mar 23, 2015 at 1:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.