Prove that if $f$ is holomorphic in the unit disc, bounded and not identically zero, and $z_1, z_2, z_3, \dotsc, z_n, \dotsc$ are its zeros ($\vert z_k \vert$ $\lt1$ ),then $$\sum_{k=1}^\infty (1-\vert z_k \vert) \lt \infty$$ [Hint:Use Jensen's formula.]

Since Jensen's formula can be used when $f$ vanishes nowhere on the circle $C_R$. I notice that there exist an increasing sequence $r_n$ for $\lim_{n\to \infty} r_n = 1$, and $f$ vanishes nowhere on each $C_{r_n}$.
Suppose $f(0) \neq 0$, then use Jensen's formula on each circle $C_r$ and get $$ \sum_{k=1}^{n_r} \log \vert z_k \vert = \log \vert f(0) \vert + n_r \cdot \log r - \frac{1}{2\pi} \int_{0}^{2\pi} \log \vert f(re^{i\theta}) \vert \,\mathrm{d}\theta, $$
where $n_r$ denotes the numbers of zeros inside the disc $C_r$. But I don't know how to estimate the limit of $n_r \log r$ as $r$ tends to $1$.

  • $\begingroup$ Maybe my complex analysis is weak here, but it appears that $f$ is zero if it has infinitely many zeroes. If $f$ is bounded, it can be continued on an open set $U$, such that $B_1(0) \Subset U$, i.e. $\bar{B_1(0)} \subset U$. But the zero-set of $f$ has an accumulation point in $\bar{B_1(0)}$. Therefore $f \equiv 0$ in $B_1(0)$ by the identity theorem. Did I do a mistake here? $\endgroup$ – red_trumpet Aug 25 '18 at 13:31
  • $\begingroup$ If $f$ is holomorphic in an open disc that vanishes on a sequence of distinct points with a limit point in the disc.Then $f$ is identically $0$. However, for this question, the limit point might on the unti circle. $\endgroup$ – J.Guo Aug 25 '18 at 13:40
  • $\begingroup$ Yeah, that's why I wanted to expand $f$ on an open which contains the closure of the unit circle. But actually I'm not sure anymore if this is possible. $\endgroup$ – red_trumpet Aug 25 '18 at 13:42
  • 1
    $\begingroup$ Consider function $f(z)=\sum_{n=0}^{\infty}z^{2^n}$ for $z$ in the unit disc. This function can extends continuously to the unit circle, but cannot be analytically continued past the unit circle.(For nowhere differentiable in its real part). $\endgroup$ – J.Guo Aug 25 '18 at 13:53
  • $\begingroup$ Sorry,I made a mistake here, function might be $f(z)=\sum_{n=0}^{\infty}2^{-n\alpha}z^{2^n}$ for any $\alpha$ is real and positive.And you can check the condition when z is real. $\endgroup$ – J.Guo Aug 25 '18 at 14:01

Of course this is a theorem instead of an exercise in many complex books, so we may as well add MSE to the list of places one can look it up...

Don't pull out the $r$ from $\log(|z_k|/r)$. Instead look at it this way: Define $$\log^+(t)=\begin{cases}\log(t),&(t>1), \\0,&(0<t\le1).\end{cases}$$

Note that $$\sum_k\log^+(r/|z_k|)=-\sum_{|z_k|<r}\log(|z_k|/r).$$ So (assuming wlog that $f(0)\ne0$) Jensen implies that $$\lim_{r\to1}\sum_k\log^+(r/|z_k|)<\infty.$$

Applying the Monotone Convergence Theorem to that sum shows that $$\sum_k\log(1/|z_k|)<\infty.$$

Or a more elementary version of the same argument: Say $|f|\le c$. Fix $N$. If $r$ is close enough to $1$ that $|z_k|<r$ for $k=1,\dots, N$ then Jensen shows that $$\sum_{k=1}^N\log(|z_k|/r) \ge\sum_{|z_k|<r}\log(|z_k|/r)\ge\log|f(0)|-\log(c).$$Since $N$ is fixed we can let $r\to1$: $$\sum_{k=1}^N\log|z_k|\ge\log|f(0)|-\log(c).$$So $\sum_{k=1}^\infty\log|z_k|>-\infty$.


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.