1
$\begingroup$

This is quick question, hopefully. How can I evaluate $f(x) = \sum_{k=1}^{\infty}x^{k^2}$? Some power series can easily be reduced to a geometric series, taylor series etc. via term wise integration/differentiation. I want to find an expression for $f(x)$ not involving series, to be able to calculate the exact value for the sum for different $x\in (-1,1)$. I've already shown that the radius of convergence is 1, and the series looks kind of like the regular geometric series. I've tried to do some term wise integration/differentiation, which however turned out to not work very well. Perhaps this is easy, but it has been a while since I was doing these kind of problems.

Cheers!

$\endgroup$
3
$\begingroup$

$\displaystyle \sum_{k = 1}^{\infty}x^{k^{2}} = -\,{1 \over 2} + {1 \over 2}\sum_{k = -\infty}^{\infty}x^{k^{2}} = \bbox[8px,border:1px groove navy]{{\vartheta_{3}\left(0,x\right) - 1 \over 2}}$ where $\displaystyle\vartheta_{\nu}$ is a Jacobi Theta Function.

$\endgroup$
1
$\begingroup$

As Felix Marin pointed out, your sum is very close to a Jacobi Theta Function.

There are an astounding number of identities for these functions.

For computing it, this identity relates small and large arguments:

If $g(x) = \sum_{k = 1}^{\infty}e^{-k^{2}\pi x} $ then $\dfrac{1+2g(x)}{1+2g(1/x)} =\dfrac1{\sqrt{x}} $.

Since $f(x) = \sum_{k=1}^{\infty}x^{k^2} $, $f(e^{-\pi x}) =g(x) $ or $f(x) = g(-\ln(x)/\pi) $.

$\endgroup$

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.