I was investigating the following corollary to Liouville's Theorem in Complex Analysis: if $f(z)$ is entire and $\lim_{z\rightarrow \infty}z^{-n}f(z)=0$, then $f(z)$ is a polynomial in $z$ of degree at most $n-1$.

My proof uses the complex L'Hospital Rule and Induction on $n$. The base case $n=0$ is obvious. To illustrate the idea: for the case $n=2$ Suppose that $f(z)$ is not bounded, so that $$\lim_{z\rightarrow \infty}{f(z)}=\infty.$$ Then by using L'Hospital, we have $$\lim_{z\rightarrow \infty}{\frac{f(z)}{z^2}}=\lim_{z\rightarrow \infty}{\frac{f'(z)}{2z}}.$$ If $f'(z)$ is not bounded, then we have further $$\lim_{z\rightarrow \infty}{\frac{f(z)}{z^2}}=\lim_{z\rightarrow \infty}{\frac{f'(z)}{2z}}=\lim_{z\rightarrow \infty}{\frac{f''(z)}{2}}=0.$$ Thus, by Liouville's Theorem, we know that $f''(z)$ is equal to some constant, and since it approaches 0 at infinity, is simply 0. Then $f'(z)=A$ (technically this actually leads to a contradiction though, since then $f'(z)$ is bounded, contrary to assumption) for some complex number $A$, and thus $f(z)=Az+B$ for complex numbers $A$ and $B$. To complete the case analysis, if either $f(z)$ or $f'(z)$ were bounded, we could have applied the Liouville's Theorem to see that they were both constants, and thus $f(z)$ would be a polynomial in $z$ of degree at most 1.

Now suppose for the induction hypothesis that $\lim_{z\rightarrow \infty}{\frac{f(z)}{z^n}}=0$, we know that $f(z)$ is a polynomial in $z$ of degree at most $n-1$, and consider the case $n+1$. Then if $f(z)$ is not bounded, we know that $$\lim_{z\rightarrow \infty}{\frac{f(z)}{z^{n+1}}}=\lim_{z\rightarrow \infty}{\frac{f'(z)}{nz^n}}=0$$ so that by our induction hypothesis we know that $f'(z)$ is a polynomial in $z$ of degree at most $n-1$, and thus $f(z)$ is a polynomial in $z$ of degree at most $n$.

Is this proof correct? My main concerns are the use of L'Hospital's Rule, for which there seems to be little information about its validity for complex valued functions. Searching around I could find the following link, but it deals with complex-valued functions of a real variable, rather than the more general complex-valued functions of a complex variable.

  • 1
    $\begingroup$ The main problem with your appoach is that even though $f$ is not bounded, you can't conclude that $\lim_{z\to\infty} f(z) = \infty$. (In fact, that is only true for polynomials. Note for example that $f(z) = e^z$ is not bounded, but $\lim_{z\to\infty} e^z$ doesn't exist; not even as $\infty$.) $\endgroup$ – mrf Mar 7 '14 at 18:39
  • $\begingroup$ @mrf, ah, I didn't realize how different the limits at infinity were of complex functions versus real functions. $\endgroup$ – Hayden Mar 7 '14 at 19:51

This exercise is solved using Cauchy's Integral Formula: $$ f^{(m)}(0)=\frac{m!}{2\pi i}\int_{|z-z_0|}\frac{f(z)\,dz}{z^{m+1}} $$ and hence, for $m\ge n$: \begin{align} \lvert f^{(m)}(0)\rvert&=\frac{m!}{2\pi}\left|\int_{|z-z_0|=R}\frac{f(z)\,dz}{z^{m+1}}\right| \le \frac{m!}{2\pi}\cdot \frac{2\pi R}{R^{m+1}}\max_{|z|=R}\lvert f(z)\rvert =m!\frac{\max_{|z|=R}\lvert f(z)\rvert}{R^m}\\&= \frac{m!}{R^{m-n}}\frac{\max_{|z|=R}\lvert f(z)\rvert}{R^n}. \end{align} By hypothesis $$ \lim_{R\to\infty}\frac{\max_{|z|=R}\lvert f(z)\rvert}{R^n}=0, $$ and thus $f^{(m)}(0)=0$, for all $m\ge 0$, which means that the power series of $f$ at $z=0$ is a finite sum missing all the terms higher than the $(n\!-\!1)$-th. Hence $f$ is polynomial of degree at most $n-1$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.