Let $C_R$ be the semi-circle of radius $R$ in the upper half plane, centered at the origin (oriented counter-clockwise). I would like to prove that $$ \lim_{R\to\infty} \int_{C_R} \frac{e^{iz}}{z} dz = 0 $$ I am aware that the above limit is a corollary of Jordan's Lemma. Is there a more direct, elegant way of showing that the limit is $0$?

I have one idea using integration by parts. We have $$ \int_{C_R} \frac{e^{iz}}{z}dz = \frac{e^{iz}}{iz}\left.\right|_{-R}^{R}+\int_{C_R}\frac{e^{iz}}{iz^{2}} dz $$ Using basic estimates we get that $$ \left|\int_{C_R} \frac{e^{iz}}{z}dz \right| \le \frac{1}{R}+\frac{1}{R}+\frac{K\pi}{R}\to 0 $$ as $R\to\infty$. Here $K$ is some constant. I used: $$ \left|\int_{C_R}\frac{e^{iz}}{iz^{2}} dz\right| \le \underbrace{(\pi R)}_{\textrm{length}}\cdot\left|\sup_{y\in(-R, R)}e^{-y}\right|\frac{1}{R^2}\le \frac{K\pi}{R} $$ So my questions are:

1) Is my solution actually correct?

2) Is there a possibly better approach for this problem?

  • $\begingroup$ $K$ isn't a constant it depends on $R$! $\endgroup$ – azarel Sep 18 '13 at 14:48
  • 1
    $\begingroup$ You don't have $\sup\limits_{y \in (-R,R)} e^{-y}$, you have $\sup\limits_{z \in C_R} \left\lvert e^{iz}\right\rvert$, and that is $1$, since $\operatorname{Im} z \geqslant 0$ on $C_R$. $\endgroup$ – Daniel Fischer Sep 18 '13 at 14:55
  • $\begingroup$ @DanielFischer: Oh my! I am an idiot... So my solution actually works with your modification, right? $\endgroup$ – Prism Sep 18 '13 at 15:01
  • $\begingroup$ I agree with the part after the "...". Not sure about the part before. $\endgroup$ – Daniel Fischer Sep 18 '13 at 15:02
  • $\begingroup$ @DanielFischer: Haha! Thanks Daniel. Okay, since I like $x, y$ notation to keep track of what's happening, I guess this could also be written as $\sup\limits_{z \in C_R} \left\lvert e^{iz}\right\rvert=\sup\limits_{x^2+y^2=R^2, y\ge 0}\left\lvert e^{-y+ix}\right\rvert=\sup\limits_{y\in(0, R)}\left\lvert e^{-y}\right\rvert = 1$ $\endgroup$ – Prism Sep 18 '13 at 15:09

$$z=Re^{it}=R\cos t+iR\sin t\;,\;\;0< t< \pi\implies \left|\;e^{iz}\;\right|=e^{-R\sin t}$$

Note that since $\,\sin t> 0\;$ , we get

$$\lim_{R\to\infty}e^{-R\sin t}=0\;,\;\;\forall\,t\in (0,\pi)$$

so applying the estimation lemma directly we get

$$\left|\;\int\limits_{C_R}\frac{e^{iz}}zdz\;\right|\leq\max_{0\leq t\leq \pi}\frac{e^{-R\sin t}}RR\pi\xrightarrow[R\to\infty]{}0$$

  • $\begingroup$ For $t \in\{0,\pi\}$, you have the exponential factor with modulus $1$, so the estimation by $\sup \lvert f(z)\rvert \cdot \operatorname{length}\gamma$ doesn't converge to $0$. $\endgroup$ – Daniel Fischer Sep 18 '13 at 15:01
  • $\begingroup$ I don't need $\;t=0\,,\,\pi\;$ as these two points on the semicircle $\,C_R\;$ are covered by the real interval $\;[-R,R]\;$ ... $\endgroup$ – DonAntonio Sep 18 '13 at 15:03
  • $\begingroup$ Still, $\sup \left\lvert \frac{e^{iz}}{z}\right\rvert = \frac{1}{R}$. You can show that the integral converges to $0$ in many ways, but it's not as easy as that. $\endgroup$ – Daniel Fischer Sep 18 '13 at 15:05
  • $\begingroup$ I think now I see your point: my thought was going into $\;f(Re^{it})\;$ and etc. but this is in fact almost the same as proving, and using, J.L. ... $\endgroup$ – DonAntonio Sep 18 '13 at 15:10

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.