0
$\begingroup$

Let $\Omega = \{ z + z^2/2 : |z| < 1\}$ be the interior of a shape in the complex plane, and let $f(z)$ be an analytic function. Show the following formula:

$$ \int_\Omega f \, dA = \frac{3\pi}{2} \, f(0) + \frac{\pi}{2} \, f'(0)$$

One strategy is to notice that both sides are linear in $f$ so we can assume a power series expansion, $f(x)=\sum a_n \, z^n$. This is not very geometric. Stokes theorem is another possibility.

It's slightly odd to integrate complex function with area? Perhaps I'm a bit rusty, I calculated the area $dA$ should be:

$$(dx + idy)\,(dx - idy) = 2i\, dx \,dy $$

I plotted an image of the interior region. It's simply connected but it's not quite a circle.

As stated in the comments, the map $z \mapsto z + z^2/2$ seems to be injective on the unit disk $\mathbb{D} = \{ |z| < 1 \}$:

enter image description here

Originally I had posted $z \mapsto z/2 + z^2$ which has another interesting shape and analogous question could be posed.

enter image description here

$\endgroup$
  • $\begingroup$ $dA$ should be $dx\wedge dy = \frac{1}{2i} d\overline{z} \wedge dz$. Since $g \colon z \mapsto z + \frac{z^2}{2}$ is injective on the unit disk, a change-of-variables suggests itself. $\endgroup$ – Daniel Fischer Dec 6 '17 at 14:20
  • $\begingroup$ @DanielFischer No it's not, the image curve is clearly overlapping. $\endgroup$ – cactus314 Dec 6 '17 at 14:24
  • $\begingroup$ We have $\operatorname{Re} g'(z) = \operatorname{Re} (1+z) > 0$ on the unit disk, so $g$ is injective there. I'm not sure what you plotted, I agree that that doesn't look injective. $\endgroup$ – Daniel Fischer Dec 6 '17 at 14:31
  • $\begingroup$ @DanielFischer I have the corret picture now. Could you explain where the $f'(0)$ term happens? $\endgroup$ – cactus314 Dec 6 '17 at 14:41
  • $\begingroup$ From the change of variables, you get $$\frac{1}{2i}\int_{\mathbb{D}} f(g(z)) (1+z)(1+\overline{z})\,d\overline{z}\wedge dz.$$ Change to polar coordinates, and all terms containing an $e^{ik\varphi}$ with $k\neq 0$ integrate to $0$. From $\overline{g'(z)}$ you have a term $e^{-i\varphi}$, and that gets you the $f'(0)$ term by the grouping $f'(0)\cdot re^{i\varphi}\cdot (1 + \dotsc)(\dotsc + e^{-i\varphi})\, rd\varphi\,dr$. $\endgroup$ – Daniel Fischer Dec 6 '17 at 14:48
2
$\begingroup$

We must of course assume that $f$ is integrable over $\Omega$. Then, since $g$ is injective, we have

\begin{align} \int_{\Omega} f\,dA &= \frac{1}{2i} \int_{\Omega} f(w) \, d\overline{w} \wedge dw \\ &= \frac{1}{2i} \int_{\mathbb{D}} f(g(z))\, d\overline{g(z)} \wedge dg(z) \\ &= \frac{1}{2i} \int_{\mathbb{D}} d\bigl(\overline{g(z)}\cdot f(g(z))\,dg(z)\bigr). \end{align}

If $f\circ g$ is continuous on $\overline{\mathbb{D}}$, we can directly apply Stokes' theorem to that last integral, otherwise we apply it to the integral over $r\cdot \mathbb{D}$ for $0 < r < 1$ and let $r\to 1$. Let's do that anyway. Then Stokes' theorem yields

\begin{align} \frac{1}{2i} \int_{r\mathbb{D}}&\, d\bigl(\overline{g(z)}\cdot f(g(z))\,dg(z)\bigr) \\ &= \frac{1}{2i} \int_{\lvert z\rvert = r} \overline{g(z)}\bigl(f(g(z))g'(z)\bigr)\,dz \\ &= \frac{1}{2i} \int_{\lvert z\rvert = r} \biggl(\overline{z} + \frac{\overline{z}^2}{2}\biggr) f(g(z))(1+z)\,dz \\ &= \frac{1}{2i} \int_{\lvert z\rvert = r} \biggl(\frac{r^2}{z} + \frac{r^4}{2z^2}\biggr)f(g(z))(1+z)\,dz \tag{$\overline{z} = r^2/z$} \\ &= \pi r^2 \frac{1}{2\pi i} \int_{\lvert z\rvert = r} \frac{f(g(z))(1+z)}{z}\,dz + \frac{\pi r^4}{2}\cdot \frac{1}{2\pi i} \int_{\lvert z\rvert = r} \frac{f(g(z))(1+z)}{z^2}\,dz\\ &= \pi r^2 f(g(0))(1+0) + \frac{\pi r^4}{2} \frac{d}{dz}\biggr\rvert_{z = 0}\bigl(f(g(z))(1+z)\bigr) \\ &= \pi r^2 f(0) + \frac{\pi r^4}{2}\bigl(f(0) + f'(0)\cdot g'(0)^2\bigr) \\ &= \pi f(0)\biggl(r^2 + \frac{r^4}{2}\biggr) + \frac{\pi r^4}{2} f'(0)\,. \end{align}

Taking the limit $r \to 1$ then gets us to

$$\int_{\Omega} f\,dA = \frac{3\pi}{2} f(0) + \frac{\pi}{2} f'(0).$$

$\endgroup$
  • $\begingroup$ Right. It doesn't just "fall out" of the change of variables. I think the $f'(0)$ term has to do with the cusp of the shape at $g(-1)=-1/2$. Because of that one point, the interior of the cardioid is not diffeomorphic to the disk. $\endgroup$ – cactus314 Dec 6 '17 at 16:11

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.