10
$\begingroup$

I've been thinking how to prove that an analytic function $f$ is a constant if the absolute value of $f$ is a constant, but I haven't figured it out yet.

What I was thinking is to use Cauchy-Riemann equations, but it didn't work well...

If this is not true, I would like to know the counterexample...

Here is what I tried:

$$|f|=|u+iv|=\sqrt {u^2+v^2}$$

Thus $u^2+v^2$ is a constant.

(1) $\displaystyle u\frac {\delta u}{\delta x}+v\frac {\delta v}{\delta x}=0 $

(2) $\displaystyle u\frac {\delta u}{\delta y}+v\frac {\delta v}{\delta y}=0 $

Plug Cauchy Riemann into (2).

$$\displaystyle -u\frac {\delta v}{\delta x}+v\frac {\delta u}{\delta x}=0 $$

and I'm stuck here...

$\endgroup$
8
  • 4
    $\begingroup$ Cauchy-Riemann equations DO work extremely well here... You might wish to expand on your try. $\endgroup$
    – Did
    Nov 25, 2012 at 15:56
  • 2
    $\begingroup$ C-R in disguise: if $|f|=1$ then $f(z)^{-1} = \overline{f(z)}$. The left hand side is holomorphic and the right hand side can only be holomorphic if $f' = 0$. $\endgroup$
    – WimC
    Nov 25, 2012 at 16:06
  • $\begingroup$ @did I added what I did to my question, so could you point out what's wrong with that or how to proceed from that? $\endgroup$
    – Tengu
    Nov 25, 2012 at 16:08
  • 2
    $\begingroup$ Next: post your own complete answer. $\endgroup$
    – Did
    Nov 25, 2012 at 16:37
  • 1
    $\begingroup$ I love to solve this using this simple trick: try $e^{f(x)}$ and then use the Maximum principle :) $\endgroup$
    – b00n heT
    May 8, 2014 at 20:01

4 Answers 4

8
$\begingroup$

This method uses the fact that if $f$ and $\bar{f}$ are both analytic then $f$ is constant. If $|f|=0$ then $f$ is always zero. If $c=|f|>0$ we have $c^2=f\bar{f}$ then $\bar{f}=c^2/f$. Since $f\neq 0$ it follows that $\bar{f}$ is analytic, and hence $f$ is constant.

$\endgroup$
7
$\begingroup$

Since there has been no posted/accepted answer, I'll post my own solution.

Let $f = u + iv$, so $|f| = |u + iv| = \sqrt{u^2 + v^2}$.

This implies $u^2 + v^2 = k$ for some constant $k$. If $k = 0$ then we are done, so consider $ k \ne 0$. Now taking partial derivatives we find

$$uu_x + vv_x = 0$$ $$uu_y + vv_y = 0$$

Using Cauchy-Riemann equations

$$uv_y + vv_x = 0$$ $$-uv_x + vv_y = 0$$

Equating both sides gives $ v_x(v+u) + v_y(u-v) = 0$ and the result follows immediately.

$\endgroup$
2
  • 4
    $\begingroup$ How the result follows immediately? $\endgroup$
    – zabop
    Jan 7, 2020 at 14:49
  • 3
    $\begingroup$ To me, it doesn't, but let's work with where we left off once we applied Cauchy Riemann equations: $$uv_y + vv_x = 0 \ \ -uv_x + vv_y = 0$$ Solving this system is the same as solving the matrix equation $$\begin{bmatrix} v & u \\ -u & v \end{bmatrix} \begin{bmatrix} v_x \\ v_y \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \end{bmatrix}$$ If you assume the constant $k = u^2 + v^2$ is nonzero, then the determinant of the first matrix is nonzero, hence it's invertible. Multiplying our matrix equation by the inverse tells us $v_x = v_y = 0$. Cauchy Riemann then will tell us $u_x = u_y = 0$. $\endgroup$
    – Michael
    Feb 4, 2021 at 1:47
3
$\begingroup$

You can also deduce this from the open mapping theorem: a nonconstant holomorphic function is an open map. If $|f|$ is constant, then $f(\mathbf C)$ is contained in the circle of radius $|f|$, which has empty interior. Hence $f$ is constant.

$\endgroup$
0
$\begingroup$

If $f$ is analytic on all of $\mathbb C$ then $f$ is constant by Liouville's theorem. However, Chris' argument works in greater generality.

$\endgroup$
1
  • 2
    $\begingroup$ You probably want to be careful you're not using this result or one very much like it to deduce Liouville... $\endgroup$ Sep 5, 2013 at 3:05

You must log in to answer this question.

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