1
$\begingroup$

The series $$f(n) = \sum_{n=0}^\infty \frac{1}{\sqrt{n}},$$

doesn't converge as it is a p-series with $p = 1/2 <1$. If I add an oscillating term to the series such that

$$f(n) = \sum_{n=0}^\infty e^{in} \frac{1}{\sqrt{n}},$$

what can I say about this series? Does it now converge? If so, is it slow to converge? That is, would I need to take a large number of terms to get an accurate numerical approximation?

$\endgroup$
3
  • 1
    $\begingroup$ you mean $\sum_{n=1}^\infty$ $\endgroup$ – reuns Oct 17 '16 at 12:23
  • 2
    $\begingroup$ The series converges by Dirichlet's test. $\endgroup$ – achille hui Oct 17 '16 at 12:37
  • 1
    $\begingroup$ @achillehui it is the same thing as the summation by parts $\endgroup$ – reuns Oct 17 '16 at 12:48
2
$\begingroup$

Use the summation by parts :$$\sum_{n=1}^N z^n n^{-s} = N^{-s} \sum_{n=1}^N z^n + \sum_{n=1}^{N-1} (n^{-s}-(n+1)^{-s}) \sum_{k=1}^n z^k$$ $$ = N^{-s} \frac{z^{N+1}-1}{z-1}+ \sum_{n=1}^{N-1} (n^{-s}-(n+1)^{-s}) \frac{z^{n+1}-1}{z-1}$$ and the conclusion is that $n^{-s}-(n+1)^{-s} = \int_n^{n+1} s x^{-s-1}dx = sn^{-s-1}+ \mathcal{O}(n^{-s-2})$

The function $Li_s(z) = \sum_{n=1}^\infty z^n n^{-s}$ is the polylogarithm, and your series is $Li_{1/2}(e^i)$.

$\endgroup$
2
$\begingroup$

A somewhat lower-powered calculation uses that $$ \frac{1}{\sqrt{n}}-\frac1{\sqrt{n+1}} =\frac1{\sqrt{n}\sqrt{n+1}(\sqrt{n}+\sqrt{n+1})} \sim\frac1{2n^{3/2}} $$ so that the series on the right of $$ (1-z)\sum_{n=1}^\infty\frac{z^n}{\sqrt{n}}=z-\sum_{n=2}^\infty\left(\frac{1}{\sqrt{n-1}}-\frac1{\sqrt{n}}\right)z^n $$ converges absolutely for all $z:|z|\le 1$.

$\endgroup$
2
  • $\begingroup$ But that is not the series I asked about. I don't see how they relate..you have several more terms, a different starting index and no exponential function how does this relate to the series in my question? $\endgroup$ – ManUtdBloke Jan 6 '17 at 22:14
  • $\begingroup$ The same semi-telescoping trick works on the partial sums. As $z=e^i\ne 1$, the additional factor $(1-z)$ does not change the convergence behavior. $\endgroup$ – Lutz Lehmann Jan 7 '17 at 1:18
1
$\begingroup$

This is not an answer but it is too long for a comment.

Starting from @user1952009's answer $$\sum_{n=1}^\infty \frac{e^{i n}}{\sqrt{n}}=\text{Li}_{\frac{1}{2}}\left(e^i\right)$$ we can expand the result in terms of the zeta function and get $$\sum_{n=1}^\infty \frac{e^{i n}}{\sqrt{n}}=\frac{\zeta \left(\frac{1}{2},\frac{1}{2 \pi }\right)+\zeta \left(\frac{1}{2},1-\frac{1}{2 \pi }\right) } 2+i\frac{\zeta \left(\frac{1}{2},\frac{1}{2 \pi }\right)-\zeta \left(\frac{1}{2},1-\frac{1}{2 \pi }\right)} 2$$

$\endgroup$

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.