0
$\begingroup$

Imagine we have an analytic function on $\mathbb{C}$. Then if the image of the unit circle under the function is a closed curve, say a circle of radius $r$, then is it immediately true that every point in the open unit disk is mapped to one point in the circle?

Put more mathematically: Let $C$ denote the unit circle and $\mathbb{D}$ denote the open unit disk. Then if $$f(C)=\{z\in\mathbb{C}: |z|= r \}$$ then $$f(\mathbb{D})=\{z\in\mathbb{C}: |z|< r\}$$

I saw a very convoluted example of this (well at least I think so, hence the question), which then stated that since the image of the contour of the open set (in the case above the unit disk) is a closed curve, the winding number of every point inside the closed curve is $1$ and the winding number for every point outside the closed curve is $0$. This I understand.

However it then states immediately that as a result the function takes every value in the region bounded by the closed curve exactly once and no value outside of the region (in our case the circle of radius $r$). The example mentioned the argument principle which I have researched and cannot establish a link to this example. If anyone could shed light on this, I'd be grateful.

$\endgroup$
2
  • 1
    $\begingroup$ FYI: The image of a differentiable curve is a differentiable curve if $f$ is analytic, so the image of a circle can't be a semi-circle. $\endgroup$ Mar 24, 2016 at 2:03
  • $\begingroup$ Ah, I was just using it for an example. Is there a better one to use? Maybe just change the semi-circle to a disk of radius $r$. $\endgroup$
    – George1811
    Mar 24, 2016 at 2:05

2 Answers 2

1
$\begingroup$

It's not true that the winding number is 1 or 0. It is an integer. For example, $z^2$ maps the unit circle to a curve that wraps around the origin twice, which gives winding number of $2$. This corresponds to a double $0$ at $z=0$. More generally, suppose $f$ does not take on the value $z$ for any $|w| = 1$, but that it does take on the value of $z$ at $w_1,w_2,\cdots,w_n$ inside this circle. Then there are positive integers $r_1,r_2,\cdots,r_n$ such that $$ f(w)-z = (w-w_1)^{r_1}(w-w_2)^{r_2}\cdots(w-w_n)^{r_n}g(w) $$ where $g$ is non-vanishing in the closed unit disk $\overline{D}$. The curve $\gamma(\theta) = e^{i\theta}$ maps $[0,2\pi]$ to the closed unit disk, and is positively oriented. The image of this curve $\gamma$ under $f$ is a closed curve with winding number about $z$ that is given by the following (and is reduced to something manageable by applying the logarithmic derivative): \begin{align} w_{f\circ \gamma}(z) & =\frac{1}{2\pi i}\int_{0}^{2\pi}\frac{d_{\theta}f(e^{i\theta})}{f(e^{i\theta})-z} \\ & = \frac{1}{2\pi i}\int_{\gamma}\frac{\frac{d}{dw}(f(w)-z)}{f(w)-z}dw \\ & = \frac{1}{2\pi i}\int_{\gamma}\left[\frac{r_1}{w-w_1}+\cdots+\frac{r_n}{w-w_n}+\frac{g'(w)}{g(w)}\right]dw \\ & = r_1+r_2+\cdots+r_n. \end{align} The integral of $g'/g$ vanishes because $g$ is non-vanishing on $\overline{D}$, which makes it holomorphic. So the winding number $\omega_{f\circ\gamma}(z)$ of $f\circ\gamma$ around $z$ is the sum of the orders of the zeros of $h(w)=f(w)-z$ for $w\in D$, counted according to multiplicity. If $f$ does not take on the value of $z$ in $\overline{D}$, then the winding number of $f\circ \gamma$ around $z$ is $0$ because all you get for the integral is the $g'/g$ term.

As an example, the function $z^2$ takes on every value in the open disk $D$ twice in the $D$, corresponding to the fact that the image of the unit circle under $z^2$ winds around everything in the open unit disk twice, and winds around everything outside the closed disk $0$ times.

$\endgroup$
3
  • $\begingroup$ Wow, very good answer, very clear. For your example for $z^2$ does this extend to any function? i.e. if the image of a contour winds round everything in a given region $n$ times, then the image of the interior of the contour under the function takes every value in the given region $n$ times, and nothing outside of that? i.e. does this work for any function? $\endgroup$
    – George1811
    Mar 24, 2016 at 3:32
  • $\begingroup$ @George1811 : Yes, if the image of the curve does not wind around a point, then that point is not in the image (assuming the image curve does not pass through the point.) The number of times it wraps around is the multiplicity. If the winding number is 4, then you can have 4 places where it takes on that value as a single root, or 2 places with a double root, or one place with a quadruple root... it's total multiplicity. $\endgroup$ Mar 24, 2016 at 5:27
  • $\begingroup$ @George1811 : For $z^2$, $0$ is a double root. But $1/2$ has two roots in $|z| < 1$--namely $1/\sqrt{2}$, $-1/\sqrt{2}$, and each of these roots is of order $1$. That's representative of the type of thing that can happen. The image of the unit circle under $z^{n}$ circles $n$ times on the unit circle. Nothing outside the unit circle is in the imagine under $z^n$ of the unit disk, and everything inside is taken on as value with total multiplicity $n$. $\endgroup$ Mar 24, 2016 at 6:21
0
$\begingroup$

Let $|\zeta| < r$ and assume that $f(z)\neq\zeta$ for all $z\in\mathbb D$. Then $\log(f(z)-\zeta)$ is defined on $\mathbb D$ and is holomorphic. Hence, $$ 0 = \int_C \log(f(z)-\zeta)f'(z)\,dz = \int_0^{2\pi}\log(f(e^{it})-\zeta)ie^{it}f'(e^{it})\,dt = \int_\gamma \log(w-\zeta)\,dw, $$ where $\gamma$ is $\{z : |z|=r\}$ (possibly run through several times). But this is obviously a contradiction.

$\endgroup$
2
  • $\begingroup$ I'm not sure I understand the consequence of the contradiction... $\endgroup$
    – George1811
    Mar 24, 2016 at 3:43
  • $\begingroup$ That zero must be in the image of $f$. $\endgroup$ Mar 24, 2016 at 3:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .