4
$\begingroup$

Let $q = q(z)\,dz^2$ be a quadratic differential on the unit disk $D(1)$, normalized so that $\int_{D(1)}|q| = 1$. I have a fairly convoluted argument that for any smaller disk $D(r)$ with $r < 1$, there is an upper bound on the area: $$ \int_{D_r} |q| \le r^2. $$ (The optimal quadratic differential is just $dz^2$.) This result feels quite elementary, and there ought to be a better proof; does anyone have one?

(This fits into a larger context of more general surfaces and subsurfaces. I can prove similar statements there with non-explicit bounds, but this would be a nice warm-up example if there's a short proof.)

$\endgroup$

2 Answers 2

2
$\begingroup$

The family of averages of $|q|$ on $D(r)$ is indeed non-decreasing. To see this, let $A(r)$ be the average of $|q|$ on $D(r)$. Let $r<\rho$ Then $A(r)=v(r)$, where $$v(z)=\frac{1}{\pi}\int_{0}^{2\pi} \int_{0}^{1}|q(zse^{it})|s \, ds \, dt.$$ Now, such a $v$ is subharmonic on $D(\rho)$ (see Ransford's book "Potential theory in the complex plane", theorem 2.4.8.). Clearly $v$ is also radial. By the maximum principle, we get $$A(r)=v(r) \leq \sup_{\partial D(\rho)} v = v(\rho)=A(\rho).$$

This also works if $|q|$ is replaced by any subharmonic function.

$\endgroup$
6
  • $\begingroup$ I don't follow the last equality. I like the observation that $|q| \le h$. $\endgroup$ Dec 12, 2014 at 0:10
  • $\begingroup$ Yes, I made a silly mistake. I corrected my answer, hopefully it works now. $\endgroup$ Dec 12, 2014 at 0:32
  • $\begingroup$ That's great! Just what I was looking for. $\endgroup$ Dec 12, 2014 at 3:08
  • $\begingroup$ In particular, I had been thinking about subharmonic functions, but somehow didn't notice that the statement could be generalized. $\endgroup$ Dec 12, 2014 at 3:09
  • $\begingroup$ @DylanThurston : I'm glad I could help. $\endgroup$ Dec 12, 2014 at 3:42
5
$\begingroup$

Am I right that in fact, your question can be re-stated in the following way: if $q(z)$ is a holomorphic function in $D(1)$ such that the average of $|q|$ over this disc is equal to 1, then the average of $|q|$ over $D(r)$ is at most 1? If yes, the first idea that comes to my mind would be to use that $|q|$ is subharmonic. Then you can proceed in two ways:

1) For a subharmonic function, if you average it w.r.t. a measure and then diffuse this measure, the average changes non-decreasingly. So -- you can launch a Brownian motion with the initial condition chosen w.r.t. the uniform measure on $D(r)$, and stop it in a (Markovian) stopping time chosen in such a way that the new distribution is uniform on $D(1)$ (a Skorokhod-like problem). There is a way of doing so, but it could be overcomplicated to write down.

2) The previous part brings into attention that the family of averages on $D(r)$ should be non-decreasing in $r$. And to show this, it suffices to show that an average on the radius $r$ circle is non-decreasing in $r$. And to compare these two averages, the previous arguments work quite well: you launch a BM from any point of an inner circle, and let it stop once it intersects the outer one. The rotational symmetry implies that the exit points are also distributed uniformly; and whatever you gain during this random walk (that is, the integral of $\Delta |q|\ge 0$ w.r.t. the expectation of the occupation measure) is the difference between the two.

$\endgroup$
3
  • $\begingroup$ In (2), I don't think it's immediate that (average on radius $r$ circle non-decreasing) $\implies$ (average on $D(r)$ non-decreasing); I believe there's an extra derivative in there. $\endgroup$ Dec 12, 2014 at 0:13
  • 1
    $\begingroup$ @DylanThurston: Integration rather then derivative, so we're on the safe side here. An average on $D(R)$ is a weighted average of radius $r$ averages by a measure $\mu_R=2r dr /R^2$. To compare averages over $D(R_1)$ and over $D(R_2)$, you notice that the map $r\to (R_2/R_1) r$ sends $\mu_{R_1}$ to $\mu_{R_2}$, and compare individual circle averages over radius $r$ and over radius $(R_2/R_1) r$. $\endgroup$ Dec 12, 2014 at 7:58
  • $\begingroup$ Thanks for expanding that out for me. I still find the other argument a little more elementary, but this is nice too. $\endgroup$ Dec 12, 2014 at 13:35

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.