On exact Pleijel's constant for some domains (Q1656225)

A classical theorem of Courant states that any eigenfunction \(\varphi_n\) associated with the \(n\)th eigenvalue \(\lambda_n\) of the Dirichlet Laplacian on a bounded domain \(\Omega \subset \mathbb{R}^N\), \(N\geq 2\), has at most \(n\) nodal domains (that is, connected components of the set \(\{x \in \Omega: \varphi_n (x) \neq 0 \}\)). Denote by \(\mu(\varphi_n)\) the number of nodal domains of \(\varphi_n\). Then by another well-known theorem, this time due to \textit{Å. Pleijel} [Commun. Pure Appl. Math. 9, 543--550 (1956; Zbl 0070.32604)], for any bounded domain \(\Omega\) and any choice of \(\varphi_n\), we have \[ Pl(\Omega):=\limsup_{n\to\infty} \frac{\mu(\varphi_n)}{n} < 1. \] This is in marked contrast to the one-dimensional case, where \(\mu(\varphi_n)=n\) for all \(n\geq 1\), by a classical theorem from Sturm-Liouville theory. Very little is known about \(Pl(\Omega)\) for \(\Omega\) in two or more dimensions; even the maximal possible value of \(Pl(\Omega)\), i.e., \(\sup_{\Omega} Pl(\Omega)\), is unknown, and the best known upper bounds are not optimal (see Remark 2.2 of \textit{I. Polterovich} [Proc. Am. Math. Soc. 137, No. 3, 1021--1024 (2009; Zbl 1162.35005)] or \textit{S. Steinerberger} [Ann. Henri Poincaré 15, No. 12, 2299--2319 (2014; Zbl 1319.35132)]. In the current paper, the author computes \(Pl(\Omega)\) explicitly for a number of domains; previously, \(Pl(\Omega)\) seems only to have been known for certain rectangles (see, e.g., Proposition 5.1 of \textit{B. Helffer} and \textit{T. Hoffmann-Ostenhof} [Contemp. Math. 640, 39--57 (2015; Zbl 1346.35132)]). Here, most notably, the value on the unit disk \(B\) in \(\mathbb{R}^2\) is computed via a fine analysis of Bessel functions and their zeros: \[ Pl(B) = 8\sup_{x>0} \left\{ x(\cos(\theta(x))^2\right\} \simeq 0.46, \] where \(\theta = \theta(x)\) is the unique solution of the transcendental equation \[ \tan \theta - \theta = \pi x \] in the interval \(\theta \in (0,\frac{\pi}{2})\). Other domains covered include circular sectors (under a simplicity assumption on the eigenvalues known to hold in many concrete cases), certain planar annuli, and \(N\)-dimensional hyperrectangles under an incommensurability assumption on the side lengths; the latter generalizes the known two-dimensional results.