3
$\begingroup$

Has anyone seen any function of the following type?

$$ g(x):=\sum_{n=0}^\infty \frac{x^n}{n!}\exp\left(-\frac{a^n}{x}\right),\quad a>1,x\ge 0. $$

The question is whether for some constant $c>0$,

$$ \lim_{x\rightarrow\infty}\frac{1}{x}\log g(x) \ge c. $$

Thanks a lot for any hints!

$\endgroup$
  • 3
    $\begingroup$ Have you optimized by $n$? What is the largest term of this series for $a$ and $x$ given? $\endgroup$ – Seva Nov 26 '14 at 18:27
  • $\begingroup$ Thanks Seva, could you please be more precise? $\endgroup$ – Anand Nov 26 '14 at 18:30
  • $\begingroup$ For $a$ and $x$ given, what is the maximum of $x^n\exp(-a^n/x)/n!$ over all $n$? $\endgroup$ – Seva Nov 26 '14 at 19:45
  • $\begingroup$ Or, at least, for what $n$ isn't the second factor just nuking the first? $\endgroup$ – fedja Nov 27 '14 at 1:24
4
$\begingroup$

This function doesn't have exponential increase. Take $N=N(x)\sim \ln x$, with a big implied constant. Then $C^n\exp(-a^n/x)$ is small for $n\ge N$, so this part of the series cannot have the same size as an exponential function.

However, for $n\le N$, the terms of the exponential series are still increasing in $n$, so we can estimate this part as follows: $$ \sum_{n=0}^N \frac{x^n}{n!}\exp(-a^n/x) \lesssim N \frac{x^N}{N!} \sim \sqrt{\ln x} \left( \frac{ex}{\ln x} \right)^{\ln x} , $$ by Stirling's formula for the last step, and this is way smaller than $e^{cx}$. Thus $g(x)e^{-cx}\to 0$ for all $c>0$. (These estimates can certainly be done more carefully, but it does answer your question in this form.)

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Dear Christian Remling, thank you very much for your nice answer! :) $\endgroup$ – Anand Nov 27 '14 at 4:48

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.