2
$\begingroup$

Are there sources that treat questions like the following ones?

  1. Suppose that $f\colon\mathbb{C}\to\mathbb{C}$ is an entire function such that $f(x)$ is real for all real $x$ and $f(x)\sim1/x$ as $x\to\infty$. Does it then follow that $f^{(k)}(x)=O(1/x^{k+1})$ for $x>x_0$, where $x_0$ is a nonnegative real number?

  2. Suppose that $f\colon\mathbb{C}\to\overline{\mathbb{C}}$ is a meromorphic function such that $f(x)$ is real for all real $x$ and $f(x)\sim x$ as $x\to\infty$. Does it then follow that $f^{(k)}(x)=O(x^{1-k})$ for $x>x_0$, where $x_0$ is a nonnegative real number?

In both cases, it would be good to also have more or less explicit bounds on the constants in $O(\cdot)$.

If the answer to such a question is negative in general, what additional conditions are needed to ensure the desired asymptotics?

Improve this question
$\endgroup$
4
  • 2
    $\begingroup$ Consider what happens, for instance, if you change $f(z)$ to $f(z)+z^{-2}\cos({\pi/2}(e^z-z))$. $\endgroup$
    – Michael Renardy
    Commented Nov 5, 2015 at 23:00
  • $\begingroup$ Thank you Michael for the nice point. I admit that my question was not sufficiently well thought out; sorry. Pietro's reference to Carleman's result is quite educational to me. Still, is there a way to answer the more general question posed above: What additional conditions are needed to ensure the desired asymptotics? $\endgroup$
    – Iosif Pinelis
    Commented Nov 6, 2015 at 4:47
  • 1
    $\begingroup$ If the asymptotics applies in a sector rather than just the real axis, you can use Cauchy's formula for derivatives. $\endgroup$
    – Michael Renardy
    Commented Nov 6, 2015 at 7:50
  • $\begingroup$ Thank you Michael for another nice comment. I had indeed tried to bound $f^{(k)}(x)$ for "my" particular function $f$ of interest by showing that $|f(z)|=O(|z|+1)$ for $z$ in an open sector containing $(0,\infty)\subset\mathbb{R}$, but have succeeded only now in that. $\endgroup$
    – Iosif Pinelis
    Commented Nov 8, 2015 at 2:25

2 Answers 2

Reset to default
3
$\begingroup$

I think the answer to both is negative. By a result by Carleman, entire real functions are dense in $C(\mathbb{R},\mathbb{R})$ with the Withney topology, so there is an entire $f$ such that $$1/x + \sin(e^x) /x^2 < f(x) < 1/x + \sin(e^x)/x^2 +1/x^3$$ for all $x>1$; this clearly forces $f'(x)$ to be unbounded for $x\to+\infty$. Analogously for the other question.

Improve this answer
$\endgroup$
3
  • $\begingroup$ References for Carleson's results in this answer: mathoverflow.net/questions/26243/… $\endgroup$
    – Pietro Majer
    Commented Nov 5, 2015 at 23:17
  • $\begingroup$ You need to modify this slightly: your inequalities force $f(x) \to \infty$ as $x \to 0+$, which is not good for an entire function. Maybe make it $x > 1$. $\endgroup$
    – Robert Israel
    Commented Nov 6, 2015 at 0:30
  • $\begingroup$ Also, that's Carleman, not Carleson. $\endgroup$
    – Robert Israel
    Commented Nov 6, 2015 at 0:33
0
$\begingroup$

The answer to the second question is negative: Consider $$ f(z) = \frac{z}{1+z^2}(1+\frac12 \sin (z^2)) $$ Here the first derivative is asymptotic to 1, though the funtion goes to $\frac{1}{x}$.

Improve this answer
$\endgroup$
1
  • $\begingroup$ The first derivative oscillates forever in $(-1,1)$. $\endgroup$
    – Brendan McKay
    Commented Nov 6, 2015 at 8:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .