I'm usually not a proponent of the mentality “here is a tool, what results can we prove with it?” (as I prefer to start at the other end with a well-motivated problem), but this famously entertaining MO post about awfully sophisticated proofs of simple results got me wondering whether there are any applications of de Branges's theorem (a.k.a. the Bieberbach conjecture) to get estimates in analytic combinatorics. Namely, the hypothetical result would provide a (non-asymptotic) upper bound on a counting sequence by using (at least as a key step) de Branges's Theorem after checking that the associated generating function is related to a univalent mapping from the unit disk.

I'd be interested to see any such nontrivial examples (including entertaining ones that give an awfully sophisticated proof of an estimate that already has a more elementary known proof).

ADDED (09/20/2023): Just to illustrate the idea, here is one application that gives a rough estimate. The function $\tan z$ is univalent in the disk of radius $\pi/2$, so $\frac{2}{\pi} \tan \left( \frac{\pi}{2} z\right)$ is in class $S$ (univalent in the unit disk and normalized to vanish at the origin and have unit derivative at the origin). Applying de Branges's Theorem gives a coefficient bound, and using that $\tan z$ is the exponential generating function for the alternating permutations, this leads to the upper bound $A_n \leq n \left(\frac{2}{\pi}\right)^{n-1} n!$ for the number of alternating permutations $A_n$ of odd size $n$. This estimate is off by an asymptotically linear factor. Since $\tan z$ is odd, we can get a better estimate using the Robertson conjecture (also proved by de Branges). In search of an application with an asymptotically sharp estimate (or at least at the correct order of growth), perhaps an example with a pole of order two should be considered in light of the form taken by the extremal--the Koebe function $z/(1-z)^2$.

  • 1
    $\begingroup$ You might warm up to the "here's a tool, what results can we prove with it?" mentality a bit if you think those folks might just be starting at the end of a well-motivated (possibly conjectured) tool. If you think about it, it's not entirely different than "here's a deep theorem/conjecture, what theories can we develop around it?" Methodology is a theory about how-to's after all. I totally understand where you're coming from, though. Sorry for being off-topic! $\endgroup$ Sep 19 at 4:58
  • $\begingroup$ A difficult (and intriguing) question seems to be, how to relate properties of a combinatorial family, to functional properties of its generating series, such as injectivity on the unit disk. $\endgroup$ Sep 19 at 9:32
  • $\begingroup$ @YuichiroFujiwara I agree that occasionally the tool itself is so attractive that it invites exploring possible applications even for the problem-oriented folks. $\endgroup$ Sep 19 at 17:17
  • $\begingroup$ @PietroMajer Yes, that sums up an essential difficulty, especially if one is trying to treat a large class of examples. But starting small and just looking for an application to one particular counting sequence, with generating function say involving some composition of elementary functions, it seems feasible that univalence can be verified directly. ...then the difficulty becomes finding an example where the final estimate we get is nontrivial. $\endgroup$ Sep 19 at 17:41

1 Answer 1


In these lecture slides, Applications of de Branges–Rovnyak decomposition to graph theory Michio Seto reports on work with Sho Suda and Tetsuji Taniguchi on applications of de Branges–Rovnyak decomposition to graph theory. I think the applications use the method of de Branges' original proof, rather than his actual theorem. The author remarks that there are proofs of the graph theory results that don't use the de Branges approach.


Your Answer

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