Hi so i am trying to prove that the function is not analytic but i am having trouble. I am supposed to use the definition $\frac{f(z+h)-f(z)}{h}$. I tried using the fact that $|z|=\sqrt{z\overline{z}}$ but I cannot see anything obvious. I know my aim is to show that if you approach $0$ at different angles you get different limits but to do so i have to somehow make the equation simpler.
How to prove $f(z)=|z|$ is nowhere differentiable
-
0Are you familiar with the Cauchy-Riemann equations? – 2017-02-21
-
0@carmichael561 yes but i am supposed to use the definition – 2017-02-21
-
1@Sorfosh Have you tried different directions for the limit? (Is h a complex or real variable)? – 2017-02-21
-
0The Cauchy-Riemann equations result straight away from the definition of differentiability, once we require that differentiability agree from the complex and real directions. – 2017-02-21
-
0The idea is to directly use the definition – 2017-02-21
-
0@Sorfosh Since the squareroot function is differentiable it is sufficient to analyze the differentiability of f(z)=z (z complement) . That is f(x,y)=(x,y)(x,-y)=P(x,y) (you fill in the polynomial P) . Now tht you just have a polynomial you should be able to figure it out. – 2017-02-21
-
0@Alephnull I am supposed to use only the definition of the derivative – 2017-02-21
-
1HINT: from the definition of derivative at a point $z$ approach with different sequences: in one sequence fix the real part, in other sequence fix the imaginary part, in other sequence fix the argument, and in other sequence fix the modulus. – 2017-02-21
3 Answers
For this, you'd probably be better served using polar form for the variable: $z=re^{i\theta}$ as then $|z|=r$. Applying the definition with this yields: $$\lim_{z\to z_0}\left[\frac{f(z)-f(z_0)}{z-z_0}\right]=\lim_{r\to r_0}\lim_{\theta\to\theta_0}\left[\frac{r-r_0}{re^{i\theta}-r_0e^{i\theta_0}}\right] $$
Then to take different paths, you fix one of the component variables, and then take the limit of the other, letting it vary. For instance, let $\theta\rightarrow\theta_0$, then you're just left with $$\lim_{r\to r_0}\left[\frac{r-r_0}{re^{i\theta_0}-r_0e^{i\theta_0}}\right]=\lim_{r\to r_0}\left[\frac{r-r_0}{e^{i\theta_0}(r-r_0)}\right]=\frac{1}{e^{i\theta_0}} $$
The idea would then be to hold $r$ constant and take the limit for $\theta$. Or to break it up into real and imaginary parts: $z = x+iy$, with the modulus ${\vert{z}\vert}={\sqrt{x^2 + y^2}}$. Then the definition is:$$\lim_{x\to x_0}\lim_{y\to y_0}\left[\frac{\sqrt{x^2 + y^2}-\sqrt{x_0^2+y_0^2}}{(x-x_0)+i(y-y_0)}\right]$$
In order for a function to be differentiable at a point, the limit must be the same from any path, so just repeat the same procedure as above, fixing one variable and letting the other vary until you end up getting two limits that'll be different for any point, at which juncture you'll of proved that the function is nowhere differentiable.
Try using the complex form of the Cauchy-Riemann equations: $\dfrac{\partial f}{\partial \bar{z}}=0$.
If one is required to apply the limit definition of the derivative, then one can proceed as follows.
Analysis for $\displaystyle z\ne 0$
First, we assume that $z\ne 0$. Then, we can write
$$\begin{align} \frac{f(z+h)-f(z)}{h}&=\frac{|z+h|-|z|}{h}\\\\ &=\left(\frac{|z+h|-|z|}{h}\right)\left(\frac{|z+h|+|z|}{|z+h|+|z|}\right)\\\\ &=\frac{|z+h|^2-|z|^2}{h(|z+h|+|z|)}\\\\ &=\frac{2\text{Re}(\bar zh)+|h|^2}{h(|z+h|+|z|)} \end{align}$$
Next, we simply take the limits as $h$ approaches $0$ along the real and imaginary axes, respectively, and show that the limits are unequal.
Using $(1)$ it is easy to see that if $h\to 0$ along the real axis then
$$\lim_{h\to 0}=\text{Re}(z)/|z| \tag 2$$
while if $h\to 0$ along the imaginary axis then
$$\lim_{h\to 0}=\text{Im}(z)/|z| \tag 3$$
Since the limits in $(2)$ and $(3)$ are not equal, then $f'(z)$ fails to exist for $z\ne 0$.
Analysis for $\displaystyle z= 0$ If $z=0$, then we have
$$\frac{f(h)-f(0)}{h}=\frac{|h|}{h}$$
which obviously fails to have a limit as $h\to 0$.
Hence, $f'(z)$ fails to exist for all $z$.
-
0Just one quick question. I proved that $|h|/h$ does not have a limit by changing h into polar coordinates. Is there a simpler way to do so? – 2017-02-21
-
0Yes. Let $h$ be purely real. Then, if the limit is taken from the left, we get $-1$, while if we take the limit from the right, we get $+1$. – 2017-02-21
-
0That seemed a bit informal for me to say that when it comes from left it is -1 and +1 from right that is why i avoided it. I guess i was too cautious. Thank You for the solution. – 2017-02-21
-
0You're welcome. My pleasure. -Mark – 2017-02-21