5

I need to show that if $f$ is an entire function it has a pole at infinity if and only if it is a polynomial. If I start with a polynomial, it is easy to show that it has a pole at infinity, but the other implication is harder.

That is, I want to prove that if $f$ has a pole at infinity it has to be a polynomial:

$f$ entire, pole at infinity $\rightarrow$ $f$ is polynomial

I do know that if $f$ is a polynomial than around zero we have that:

$f(1/z)=z^{-n}\cdot h(z)$, where $h$ is holomorphic and non-vanishing around zero. But how does this help?

I could try contrapositive. That is, if $f$ is not a polynomial, then showing that it can't have a pole at infinity. That is:

$f$ is not polynomial $\rightarrow$ $f(1/z)$ does not have a pole at zero

But this also seems hard, any tips?

ett
  • 1,404
user119615
  • 10,176
  • 5
  • 44
  • 112
  • 3
    Isn't this very straightforward? If $f$ has a pole at $\infty$ then $f(\frac{1}{z})$ has a pole at $0$, so the latter's power series at $0$ has finitely many negative terms. This means $f$'s power series at $0$ had finitely many positive terms; and it can't have negative terms or it wouldn't be entire. Thus it's a polynomial. – juan arroyo Jan 11 '17 at 00:31
  • 3
    @juanarroyo Not straightforward to me. – user5826 Nov 05 '19 at 02:06

2 Answers2

11

Suppose $f$ has a pole at $\infty$ of order $n$. Use the integral formula for the $n$th derivative of $f$, which is as follows: $f^{(n)}(a) = \frac{n!}{2 \pi i} \int_C \frac{f(z)}{(z-a)^{n+1}}dz$, where $C$ is a positively oriented circle enclosing $a$ with radius.

To employ the formula here, suppose $|a|< R$ and let $C$ be the circle of radius $2R$ centered at zero. Then $|f^{(n+1)}(a)| \le \frac{(n+1)!}{2\pi} \int_C |\frac{f(z)}{(z-a)^{n+2}}||dz|$. Note that $|z-a| > R$ on the domain of integration, so we can upper bound further by $\frac{(n+1)!}{2\pi R^{n+2}} \int_C |f(z)||dz|$. The fact that $f$ has a pole at $\infty$ of order $n$ means that there exists a positive constant $c_1$ such that $ |f(z)| < |c_1z^n|$ for $|z|$ sufficiently large. Therefore as $R \to \infty$, $|f^{(n+1)}(a)| \le \frac{c_1 (n+1)! (2R)^n}{2 \pi R^{n+2}} \int_C |dz| = \frac{c_1 (n+1)! 2^n R^{n+1}}{R^{n+2}}$, which clearly approaches zero as $R \to \infty$. It follows that $f^{(n+1)}(a) = 0$. Since $f$ has a derivative which is identically zero, it is a polynomial.

The Count
  • 3,620
Vik78
  • 3,877
7

Since $f$ has a pole at $\infty$ there exists an $R > 0$ such that $|f(z)| \geq 1$ for all $|z| > R$. Now let $z_1, ..., z_k$ be the zeros of $f$, counted with multiplicity, in $\overline{D(0,R)}$. There are only finitely many inside $\{ z: |z| \leq R \}$ or else the zeros would have an accumulation point in this compact set and then $f$ would be identically zero by the uniqueness theorem which is not possible since $f$ has a pole at $\infty$.

Consider the function $h(z) = \frac{(z-z_1)...(z-z_k)}{f(z)}$ then $h$ is an entire function. On the set $\{z: |z| > R \}$, $|h(z)| \leq C R^k$ by the Cauchy estimates since $h$ is an entire function of polynomial growth and hence it must be a polynomial. On the compact set $\overline{D(0,R)}$, we have $h(z) \leq M$ for some $M>0$ because $h$ is a continuous function on the compact set an hence it attains its maximum there. So $h$ is an entire function that is bounded since $|h(z)| \leq \max \{CR^k, M\}$ on all of $\mathbb{C}$, hence $h$ is a constant function.

But then, $f(z) = h(z)(z-z_1)..(z-z_k)=C(z-z_1)...(z-z_k)$ a polynomial.

Mustafa Said
  • 4,137
  • How do you know $f$ has only finitely many zeros? If $f$ has infinitely many zeros but no accumulation points in its zero set, it may be non-constant. – The Substitute Feb 14 '15 at 12:51
  • 1
    Since $|f(z)| \geq 1$ for $|z| > R$, all the zeroes lie in a compact set, so if there are infinitely many of them, there is such an accumulation point. – D_S Jul 15 '15 at 02:55
  • But my question is, how are you concluding that $h$ is bounded? – D_S Jul 15 '15 at 03:03
  • @D_S I am concluding that $h$ is bounded because there exists a fixed $R > 0$ and fixed $K > 0$ such that $|f(z)| \leq \max{(CR^k, K)}$. – Mustafa Said Jul 16 '15 at 21:06
  • @MustafaSaid : Your estimate for $h(z)$ on ${z : |z| > R}$ doesn't hold true. – M.G May 07 '16 at 00:07
  • @M.G is right: on ${ |z| > R}$ you have a lower bound on $|f|$ and also on $|z|$, so you have no bound on $|h|$. Edit: I'm sorry, didn't realize this was such an old question. Seems it was recently rejuvenated... – amcerbu Jan 10 '17 at 14:49
  • Can't you just bypass using $h$ at all? The fact that $f$ has a pole at infinity means, by definition, that $f$ is an entire function of polynomial growth, so you can just apply Cauchy's estimate to $f$ directly like I did. – Vik78 Jan 10 '17 at 18:18
  • @MustafaSaid, do we actually need to invoke Cauchy's estimates here? Since $f>1$ outside of $D(0,R)$, the function $h$ is bounded above by the polynomial $\prod (z-z_j)$ and is thus itself a polynomial. $h$ a polynomial implies $f$ rational, which implies f is a polynomial since it has no finite poles. I guess I'm just not seeing where we used Cuachy's estimates. – Van Latimer Dec 10 '17 at 01:22