My question is primarily conceptual: Consider a function $f(z)$ which has a branch cut from $z=0$ to $z=\infty$ along the positive Re(z) axis. If I wish to integrate it along a small, clockwise circle around $z=0$, what care must I take when integrating over the branch cut? I've been trying to find references and failing, so if you have a link to some notes or a book reference, I'd appreciate it. If instead, you could give me a hint on the following example, I would appreciate that, too.

In particular, I am trying to compute the contour integral $\oint_{\varepsilon} \frac{\log(z)}{(z+a)^2} dz$ where $a \in \mathbb{C}$, and $\varepsilon$ is the small, counter-clockwise circle around $z=0$. I have tried the following:

  1. Parametrizing: $z=\rho e^{i \theta}$, $dz = iz d\theta$, so that$^{\dagger}$ $\oint_{\varepsilon} \frac{\log(z)}{(z-a)} dz = i \int_0^{2 \pi} \frac{\log \rho + i\theta}{(\rho e^{i\theta}+a)^2}\rho e^{i \theta} d\theta=g(\theta)|_{0}^{2\pi}$

Ultimately, I want to take the limit $\rho \rightarrow 0$, but this causes $\log \rho$ to diverge.

  1. Would this work?: $\int \frac{\log(z)}{(z+a)^2} dz = \frac{z \log(z) - (a+z)\log(a+z)}{a(a+z)}+c$, so is the following true? $\oint_{\varepsilon} \frac{\log(z)}{(z+a)^2} dz = \left[ \frac{z \log(z) - (a+z)\log(a+z)}{a(a+z)} \right]_{\theta=0}^{\theta=2 \pi} = 2 \pi i \frac{\rho}{a(a+\rho)}$

$\dagger$ : $g(\theta)|_0^{2\pi} = \frac{i \rho}{a} \left[ \log \rho \left( i \log\left( 1 + \frac{z}{a} \right) + \theta - \frac{i}{\left( 1 + \frac{z}{a} \right)} \right) + i \left(Li_2\left(-\frac{z}{a} \right) + \frac{\theta}{2} \left(i\theta-2+\frac{2}{\left( 1 + \frac{z}{a} \right)}\right) - (i+\theta)\log\left( 1 + \frac{z}{a} \right) \right) \right]_0^{2\pi}$

$g(\theta)|_0^{2\pi}= \frac{2 \pi i \rho}{a}\left[ \log \rho + \frac{1}{2}\left(2 \pi + 2i - \frac{2i}{ 1 + \frac{\rho}{a} } \right) + i \log \left( 1 + \frac{\rho}{a} \right) \right]$

where $z(\theta) = \rho e^{i \theta}$ and $Li_n(z)$ is the polylogarithm.


I think what you did is fine, up to the nonelementary integral you got.

When you set up such an integral you have to state clearly what you mean by $\log z$ when $z\in\Omega:={\mathbb C}\setminus{\mathbb R}_{\geq0}$. It seems that you have the following in mind: When $z=\rho e^{i\theta}$ with $0<\theta<2\pi$ then $\log z:=\log\rho +i\theta$.

(i) Parametrizing $\gamma:=\partial D_\rho$ you obtain $$\int_\gamma{\log z\over(a+z)^2}\>dz=i\int_0^{2\pi}{\log\rho +i\theta \over(a+\rho e^{i\theta})^2}\rho e^{i\theta}\>d\theta\ .$$ This integral only seems to be nonelementary (see below), but in the limit $\rho\to0+$ you obtain $0$ anyway, since $\lim_{\rho\to0+}\rho\log\rho=0$.

(ii) The given integrand ${\log z\over(a+z)^2}$ has a primitive in $\Omega$, namely the function $$f(z):={1\over a}\left({z\over a+z}\log z-\log(a+z)\right)\ .$$ Taking limits at the "ends" of $\gamma$ caused by the cut we therefore get $$\int_\gamma{\log z\over(a+z)^2}\>dz=\lim_{\theta\to 2\pi-} f\bigl(\rho e^{i\theta}\bigr)-\lim_{\theta\to 0+} f\bigl(\rho e^{i\theta}\bigr)={2\pi i\>\rho\over a(a+\rho)}\ .$$ Note that the $\log(a+z)$ term gives no contribution as long as $\rho<|a|$.

  • $\begingroup$ Thanks Christian Blatter and @Dr. MV for your helpful responses. So there's nothing particularly interesting about doing a contour integral in which the contour crosses a branch cut of the integrand, other than the integrand (and in this case, the antiderivative) jumping in value across the cut? There's no rule against integrating along a contour over the cut and/or using the either side of the cut as the start and end points of a closed contour? $\endgroup$ – R.C. Oct 20 '15 at 19:40
  • $\begingroup$ You are not allowed to integrate across the cut using the FTC in some way, because the integrand is not the derivative of a function which is smooth in a neighborhood of the cut. $\endgroup$ – Christian Blatter Oct 21 '15 at 8:16

You were on the right track. The integral around the circle of radius $\rho$ and centered at $z=0$ can be expressed as

$$\oint_{|z|=\rho}\frac{\log z}{(z+a)^2}\,dz=\int_0^{2\pi}\frac{\log \rho +i\theta}{(\rho e^{i\theta}+a)^2}\,\rho d\theta$$

Now, recall that we have $\lim_{\rho \to 0}\rho \log \rho=0$. Therefore, for $a\ne0$, the limit of the integral of interest is zero.

  • $\begingroup$ Please let me know how I can improve my answer. I really want to give you the best answer I can. If the answer was not useful to warrant an up vote, I am happy to delete. $\endgroup$ – Mark Viola Nov 11 '15 at 18:47

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.