7
$\begingroup$

Let $q\in \mathbb R$ with $|q|<1$. I am looking for an upper bound of the converging series $$ \sum_{n=1}^\infty q^{(n^2)} $$ with respect to $q$. Using the inequality $q^{(n^2)} \le q^{2n-1}$, one gets $$ \sum_{n=1}^\infty q^{(n^2)} \le q^{-1} \sum_{n=1}^\infty q^{2n} = q\sum_{n=0}^\infty q^{2n} = \frac{q}{1-q^2} $$ Numerical experiments say that this estimate does not reflect the behavior of the sum for $q\to1$. They suggests an upper bound in the order of $\frac1{\sqrt{1-q^2}}$. Is it possible to prove this? Preferably with elementary arguments?

$\endgroup$
  • 1
    $\begingroup$ For your information, $\,\theta_3(0,q)=\sum_{n=-\infty}^\infty q^{n^2}\,$ where $\,\theta_3\,$ is a Theta function. $\endgroup$ – Somos Jul 10 '18 at 10:44
11
$\begingroup$

Hint. Note that as $q\to 1^-$, we have that $\ln(q)<0$ and $$\sum_{n=1}^\infty q^{n^2}=\sum_{n=1}^{\infty}e^{\ln(q)n^2}\leq \int_{x=0}^{\infty}e^{\ln(q)x^2}\,dx=\frac{\sqrt{\pi}}{2\sqrt{-\ln(q)}}\sim \frac{\sqrt{\pi}}{2\sqrt{1-q}}.$$

$\endgroup$
  • $\begingroup$ Thanks for the answer! The second-to-last term seems to be a pretty good approximation numerically for $q\to1$. $\endgroup$ – daw Jul 10 '18 at 11:19

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.