2
$\begingroup$

A basic principle in complex function theory is that one can split off zeros of holomorphic functions in a similar way as for polynomials: If $f$ is holomorphic near $0$ and $f(0) = 0$, then $f(z) = zg(z)$ for some holomorphic function $g$. A crucial property of this type of factorization is that it preserves boundedness: If, say, $f \in H^\infty(\mathbb D)$, then also $g \in H^\infty(\mathbb D)$. I am interested in generalizations to several complex variables:

Thus, let $f$ be a holomorphic function on the $2$-disc $\mathbb D \times \mathbb D$ with the following properties: $f$ is antisymmetric (i.e. $f(z,w) = -f(w,z)$), bounded and $f(z,0) = f(0,z) = 0$ for all $z \in \mathbb D$. In particular, $f$ vanishes along the diagonal $\{w=z\}$. By an easy reduction to the one-dimensional case (using a linear change of variables), we see that $f$ can be factorized as $f(z,w) = (z-w)g(z,w)$ for some holomorphic function $g$ on the $2$-disc.

My first question is, whether $g$ is necessarily bounded. I suspect that the answer is negative, but a concrete counterexample would be nice. In this case, my second question would be: What can be said about the boundary behaviour of $g$?

(If I am not mistaken, then the radial limit function $\partial g$ of $g$ on the $2$-torus exists almost everywhere and is bounded on the complement of any neighbourhood of the diagonal circle. The question is, what happens near the diagonal. If $g$ is not bounded, then $\partial g$ will blow up near the diagonal - the question is, how badly?)

$\endgroup$

2 Answers 2

2
$\begingroup$

No, $g$ is not necessarily bounded. Let $h(z,w)$ be (a branch of) $log (z+w-2)$ and let $f(z,w)$ be $(z-w)g(z,w)=(z-w)zwh(z,w)$.

Then on the one hand $zwh(z,w)$ is unbounded as $z$ and $w$ (in $\mathbb D$) both approach $1$, but on the other hand $f(z,w)$ remains bounded: the inequality $|z-w|^2+|z+w|^2=2|z|^2+2|w|^2\le 4$ shows that $|z-w|$ is going to $0$ faster than $|\log (z+w-2)\ |$ is going to $+\infty$.

$\endgroup$
4
  • $\begingroup$ Ah, yes that is what I was looking for. I was playing around with logarithms myself, but I had the problem of finding a branch defined on all of $\mathbb D^2$. Of course, since the image of $\mathbb D^2$ under $z+w-2$ avoids $0$ and is simply-connected there is no problem here. Still, this answers my first question (so I will accept it), but not my second one. In this example we have that $(z-w)^\epsilon\cdot h(z,w)$ is bounded, so the unboundedness is rather mild. Can we modify the example to get something worse? $\endgroup$ Jan 8, 2013 at 17:01
  • $\begingroup$ You can use a power of $z+w-2$ instead of log. $\endgroup$ Jan 9, 2013 at 2:30
  • $\begingroup$ Sure, but it seems I still get a rate, i.e. I can find $\epsilon$ such that $f(z,w)/|z-w|^\epsilon$ is bounded. $\endgroup$ Jan 9, 2013 at 9:20
  • $\begingroup$ Well, you could choose something smaller than $(z+w-2)^{-1}$ but bigger than $(z+w-2)^{-1+\epsilon}$. But that's only a little worse. $\endgroup$ Jan 9, 2013 at 13:37
0
$\begingroup$

Under you assumptions, $f$ vanishes on $(\lbrace 0\rbrace\times\mathbb C)\cup (\mathbb C\times \lbrace 0\rbrace)\cup \lbrace (z,z): z\in \mathbb C\rbrace$ intersected with the domain. Thus $f(z,w)=z.w.(z-w).h(z,w)$ and $h$ is symmetric. Since you just removed zero divisors, any polar divisor that was there in $f$ is still there in $h$.

So if $f$ was bounded on an open domain, $h$ is still bounded there.

Edit: Tom's answer woke me up. The last conclusion is wrong. If the polar divisor touches $D\times D$ in the boundary, you get a counterexample. Take $h(z,w)=1/(z+w-2)$.

$\endgroup$
1
  • $\begingroup$ It seems that I am missing a point here: I understand that by splitting of the zeros one does not introduce any new poles, i.e. h is holomorphic in the domain. However, I don't understand how that affects boundedness, which concerns only the behaviour of f and h near the boundary. I guess there is a basic fact about holomorphic functions here that I miss? $\endgroup$ Jan 8, 2013 at 13:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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