We are asked to compute: Res$[\frac{z^n+1}{z^n-1},e^{2\pi ki/n}]$, where I am assuming $k\in\mathbb Z$. The only tools I am aware of to compute residues comes from relating the function to power series or simple and double poles... This function has a simple pole at the indicated point... but the limit seems like it would be messy. Is there any other way to go about this besides taking the limit of $$\lim_{z\rightarrow e^{2\pi ki/n}} \frac{z^n+1}{z^n-1} (z-e^{2\pi ki/n})$$?


The residue of $p(z)/q(z)$ is $p(z_0)/q'(z_0)$. Easy! - but note the conditions in the next paragraph.

This method is well worth knowing. The full details: if $p,q$ are analytic at $z_0$, and $p(z_0)\ne0$, and $q(z_0)=0$, and $q'(z_0)\ne0$, then $p/q$ has a simple pole at $z_0$ and the residue is $p(z_0)/q'(z_0)$.

You can prove the residue formula by noting that under the stated conditions $$\lim_{z\to z_0}\frac{q(z)}{z-z_0}=\lim_{z\to z_0}\frac{q(z)-q(z_0)}{z-z_0}=q'(z_0)$$ and so $${\rm Res}\Bigl(\frac{p}{q},z_0\Bigr)=\lim_{z\to z_0}(z-z_0)\frac{p(z)}{q(z)} =\frac{p(z_0)}{q'(z_0)}\ .$$

In this specific case we have $$\frac{p(z)}{q'(z)}=\frac{z^n+1}{nz^{n-1}}$$ and the residue is $$\frac{e^{2k\pi i}+1}{ne^{2k\pi i(n-1)/n}}=\frac{2}{n}e^{2k\pi i/n}\ .$$

  • $\begingroup$ I believe this is only the case for a simple pole, IIRC. (regarding first line) $\endgroup$ – apnorton May 2 '14 at 2:29
  • $\begingroup$ @anorton, true, but I hope I have made that sufficiently clear in what follows - happy to change it if not. $\endgroup$ – David May 2 '14 at 2:54
  • 2
    $\begingroup$ You have, I just wanted to make sure to draw attention to it in case someone just skimmed the answer. (+1, though--you made a good writeup) $\endgroup$ – apnorton May 2 '14 at 3:00
  • $\begingroup$ @anorton fair enough, thanks! - have added a warning just in case. $\endgroup$ – David May 2 '14 at 3:01
  • 1
    $\begingroup$ Doesn't the proof of L'Hopital's Rule rely on the Mean Value Theorem, i.e., can we use it in complex analysis? $\endgroup$ – user2154420 May 2 '14 at 10:40

If you have a meromorphic function $\frac{f(z)}{g(z)}$ and the denominator $g$ has a simple zero in $z_0$, then you have

$$\operatorname{Res} \left(\frac{f(z)}{g(z)}; z_0\right) = \frac{f(z_0)}{g'(z_0)}.$$

You can derive that from the Taylor expansions of $f$ and $g$:

$$f(z) = f(z_0) + (z-z_0)f'(z_0) + O((z-z_0)^2); \quad g(z) = (z-z_0)g'(z_0) + O((z-z_0)^2),$$


$$\begin{align} \frac{f(z)}{g(z)} &= \frac{f(z_0) + (z-z_0)\tilde{f}(z)}{(z-z_0)\left(g'(z_0) + (z-z_0)\tilde{g}(z)\right)}\\ &= \frac{f(z_0)}{(z-z_0)g'(z_0)}\left(1 - (z-z_0)\frac{\tilde{g}(z)}{g'(z_0)} + O((z-z_0)^2)\right) + \frac{\tilde{f}(z)}{g'(z_0) + (z-z_0)\tilde{g}(z)}\\ &= \frac{f(z_0)/g'(z_0)}{z-z_0} + O(1). \end{align}$$


Well you know that the $z^n + 1$ part evaluates to $2$... For the rest you get a polynomial $z^{n-1} + (*)\cdot z^{n-2} + ... (*)^{n-1}$, where $(*)$ is $e^{2\pi\cdot k\cdot i/n}$. So the limit evalutes to $\frac{2}{n\cdot e^{-2\pi\cdot k\cdot i/n}} = \frac{2e^{2\pi\cdot k\cdot i/n}}{n}$. Now, I could easily have made a mistake... so check this with the formula derived by Fischer. This evaluates to $\frac{2}{n\zeta^{n-1}} = \frac{2}{n\zeta^{-1}} = \frac{2\zeta}{n}$, where we are evaluating the limit as $\zeta = e^{2\pi\cdot k\cdot i/n}$.


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.