3

If $f(z)$ is entire, and $f(z)$ is real iff $z$ is real, prove $f'(z) \ne 0$ for all real $z$.

Edit: Sorry, somehow my preliminary efforts didn't make it in here:

I know that $f$ cannot be constant, that's a contradiction, and that if $f$ is entire, then it is analytic and holomorphic everywhere.

I have seen (but not quite understood) proofs that $f$ can have at most one zero. I have not done "winding numbers" and it seems most proofs of this employ the use of these.

Because $f$ is entire, it is continuous and so all $z$ in the upper half plane have the same sign of their imaginary part.

My first thought was by contradiction. Assume $f'(z)=0$ for some $z_0 \in C$. Then $a_1 = 0$ in the taylor series of $f$, so then $f(z)= f(z_0) + a_2*(z-z_0)^2 + O((z-z_0)^2)$. Then I'm not sure where to go from here.

I was also thinking of setting $f = u(x,y) + iv(x,y)$, and noting that $v(x,0)=0$, but I can't see how this would get me anywhere with derivatives.

mucciolo
  • 2,950
K. Ann
  • 59
  • 4
    What are your thoughts on this problem? We are not a homework answering service. We require you to show some effort or share your own thoughts. – Cameron Williams May 04 '17 at 16:37
  • 1
    Welcome to the Mathematics SE K,Ann.It is very helpful to know what trials and errors you have made. This way we can better assess your capabilities and act accordingly. Ergo "What have you tried?"If you don't do so, most probably questions get closed here, I.e. no new answers can be added. Since you are new user, I am not voting to close this question. But be sure to add your effort unless this question will be closed later. – Jaideep Khare May 04 '17 at 16:39
  • Following on from @JaideepKhare, including such information also helps us understand your mathematical capabilities and give a relevant answer, which allows you to gain the most from such a question. – mdave16 May 04 '17 at 16:42
  • 3
  • Man, I really liked @mdave16's comment. So true, yet gentle enough for even the most delicate of sensibilities. Refined, but not pretentious. Forceful, but not blunt. I can't wait to be good at math and still be a nice person, like him! – rocksNwaves May 04 '17 at 16:54

2 Answers2

1

As you stated $f(z)$ can't be constant, otherwise, from $f(z) \in \mathbb{R} \Leftrightarrow z \in \mathbb{R}$, $f(z)$ must be a real constant which happens to be real constant for non real $z$ too.

Now, by contradiction, let's assume $$f'(z)=0, \forall z \in \mathbb{R} \tag{1}$$ which also means $$f'(0)=0 \tag{2}$$ Then, using Taylor series:

$$f(z) = a_0+\sum_{n=1}a_nz^n \tag{3}$$ $$f'(z) = a_1+\sum_{n=2}na_nz^{n-1} \overset{(2)}{\Rightarrow} \color{red}{a_1=0}$$ But then $$f'(z)= \sum_{n=2}na_nz^{n-1}= z\left(\sum_{n=2}na_nz^{n-2}\right)=z\cdot g_1(z) \overset{(1)}{\Rightarrow} g_1(z)=0, \forall z \in \mathbb{R}, z \ne 0$$ $g_1(z)$ is also entire and continuous. This means that for example $0=g_1\left(\frac{1}{n}\right)\rightarrow g_1(0)=0$, as $n \rightarrow \infty, n \in \mathbb{N}$, so $g_1(0)=0 \Rightarrow \color{red}{a_2=0}$.

But then $$f'(z)=z\left(\sum_{n=2}na_nz^{n-2}\right)=z^2\left(\sum_{n=3}na_nz^{n-3}\right)=z^2\cdot g_2(z) \overset{(1)}{\Rightarrow} g_2(z)=0, \forall z \in \mathbb{R},z \ne 0$$ similarly, $g_2(z)$ is also entire (and continuous), thus $g_2(0)=0 \Rightarrow \color{red}{a_3=0}$.

Continuing like this, by induction, we conclude $\color{red}{a_n=0, \forall n>0}$. But then, from $(3)$, $f(z)=a_0$ which is a constant, contradiction with the statement at the beginning.

rtybase
  • 16,907
0

Step 1. Clearly $\mathrm{Im}\,f(x+iy)\ne 0$, whenever $y\ne 0$. In particular, $\mathrm{Im}\,f(z)$ maintains sign in the upper half plane and the lower half plane, and it has different signs in each half plane. Without loss of generality, $\mathrm{Im}\,f(x+iy)> 0$, if $y>0$, and $\mathrm{Im}\,f(x+iy)< 0$, if $y<0$.

Step 2. Assume that $f'(x)=0$, for some $x\in\mathbb R$. Without loss of generality, assume that $x=0$. Otherwise, replace $f$ by $f-f(0)$. So $f'(0)=0$, and $f\not\equiv 0$, implies the existence of a $k\in\mathbb N$, such that $f(z)=z^kg(z)$, where $g$ entire and $g(0)\ne 0$. Thus, there exists an $a\in\mathbb R$, such that, if $f=u+iv$, then $$ 0\ne a=f^{(k)}(0)=\partial_x^kf(0)=\partial_x^ku(0)=\partial_x^ku(0)=\partial_y^kv(0). $$ Clearly $k$ is odd, otherwise $v$ would have a local minimum, which is impossible, since $v<0$, in the lower half plane. In fact, $a>0$, since $v$ is increasing in $y$ at $y=0$. Hence $f(x)>0$ and $f$ strictly increasing for $x>0$, and $x$ sufficiently small and $f(x)<0$ and $f$ also strictly increasing, for $x<0$ and $|x|$ sufficiently small. But this means, that the equation $$ f(x+iy)=\eta $$ would have $k$ solution for $|\eta|$ is a small disk centred at $0$. But all the $k$ roots have to be negative, which means that $k=1$, since $f$ is strictly increasing. Hence, $f'(0)\ne 0$.

In you need to show that $f$ is linear, then, observe that

a. $f$ has to be strictly increasing, in $x\in\mathbb R$, and by Picard's Little Theorem, $f$ is a polynomial, since, the neighbourhoods of infinity do not contain absolutely small real numbers. But the only root in real, and there is only one real real root and single, and hence $f$ is a linear function of the form $$ f(z)=az+b $$ where $a,b\in\mathbb R$ and $a\ne 0$.