\documentclass[11pt,reqno]{amsart} \usepackage[T1]{fontenc} \usepackage{lmodern} \usepackage{amsmath,amssymb,amsthm,mathtools} \usepackage[hmargin=2.9cm,vmargin=2.6cm]{geometry} \usepackage{booktabs} \usepackage{array} \usepackage{tikz} \usepackage{xcolor} \usepackage[font=small,labelfont=sc,labelsep=period]{caption} \usepackage{hyperref} \usepackage{microtype} \captionsetup[table]{skip=6pt} \captionsetup[figure]{skip=8pt} \definecolor{navylink}{RGB}{22,56,112} \definecolor{citecrimson}{RGB}{115,32,38} \hypersetup{ pdftitle={The Square Maximizes the First Nonzero Steklov Eigenvalue Among Convex Quadrilaterals}, pdfauthor={L\"uwei Dong, Sirui Yang, and Jintao Pan}, colorlinks=true, linkcolor=navylink, citecolor=citecrimson, urlcolor=navylink } \def\UrlFont{\small\ttfamily} \allowdisplaybreaks[2] \renewcommand{\arraystretch}{1.15} \makeatletter \g@addto@macro\UrlBreaks{\do\_\do\-\do\/\do\.} \renewcommand\subsection{\@startsection{subsection}{2}% \z@{.85\linespacing\@plus.5\linespacing}{.35\linespacing}% {\normalfont\normalsize\bfseries}} \makeatother \newtheorem{theorem}{Theorem} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{proposition}[theorem]{Proposition} \theoremstyle{remark} \newtheorem*{remark}{Remark} \DeclareMathOperator{\Var}{Var} \DeclareMathOperator{\conv}{conv} \renewcommand{\Re}{\operatorname{Re}} \renewcommand{\Im}{\operatorname{Im}} \title[Steklov eigenvalues of convex quadrilaterals]{The Square Maximizes the First Nonzero Steklov Eigenvalue Among Convex Quadrilaterals} \author[L.~Dong]{L\"uwei Dong} \address{Soochow University, Suzhou, China} \author[S.~Yang]{Sirui Yang} \address{Nanjing University of Information Science and Technology, Nanjing, China} \author[J.~Pan]{Jintao Pan} \address{Xiamen University, Xiamen, China} \subjclass[2020]{Primary 35P15; Secondary 49R05, 52A10, 52A40, 65G20} \keywords{Steklov eigenvalue problem, convex quadrilaterals, Weinstock inequality, polygonal shape optimization, interval arithmetic} \begin{document} \begin{abstract} We show that among all convex planar polygons of prescribed perimeter with at most four sides, the square uniquely maximizes the first positive Steklov eigenvalue. If $k \in (0, \pi/2)$ denotes the unique root of $\tan k \tanh k = 1$, then $L\sigma_1(\Omega) \le 8k\tanh k$, with equality if and only if $\Omega$ is a square. By the quadrilateral isoperimetric inequality, the square also uniquely maximizes the area-normalized eigenvalue $\sqrt{|\Omega|}\,\sigma_1(\Omega)$ in the same class. Our proof splits the four-dimensional moduli space of convex quadrilaterals into a neighborhood of the square and its complement. Near the square, a weighted combination of the two first Steklov eigenfunctions of the square produces a negative definite Rayleigh defect; its sign on an explicit parameter box is established via an exact degree-$22$ Taylor expansion in ball arithmetic together with an analytic Cauchy remainder estimate. Far from the square, a boundary-variance inequality along a diagonal reduces the remaining shapes to a compact region, which we cover by $940$ closed subboxes using affine and harmonic quadratic trial functions checked in outward-rounded integer interval arithmetic. \end{abstract} \maketitle \section{Introduction and main results}\label{sec:intro} Let $\Omega \subset \mathbb{R}^2$ be a bounded convex polygon of positive area $|\Omega|$ and perimeter $L = |\partial\Omega|$. The Steklov eigenvalue problem on $\Omega$ asks for nonzero functions $u \in H^1(\Omega)$ and real numbers $\sigma$ satisfying \[ \Delta u = 0 \quad \text{in } \Omega, \qquad \partial_\nu u = \sigma u \quad \text{on } \partial\Omega \] in the weak sense, where $\partial_\nu$ is the outward normal derivative along $\partial\Omega$. Because $\Omega$ is a bounded Lipschitz domain, the trace operator $H^1(\Omega) \to L^2(\partial\Omega)$ is compact, and the Steklov spectrum forms an unbounded sequence \[ 0 = \sigma_0(\Omega) < \sigma_1(\Omega) \le \sigma_2(\Omega) \le \cdots \to +\infty, \] counted with multiplicity; we refer to the surveys~\cite{GP2017,CGGS2024} for background on the spectral geometry of this problem. Writing $d\mu = ds/L$ for normalized arclength measure on $\partial\Omega$ and $E_f = \int_\Omega |\nabla f|^2$ for the Dirichlet energy of $f \in H^1(\Omega)$, the scale-invariant functional $\Sigma(\Omega) := L\sigma_1(\Omega)$ admits the variational characterization \begin{equation}\label{eq:variational} \Sigma(\Omega) := L\sigma_1(\Omega) = \inf_{\substack{f \in H^1(\Omega) \\ \Var_\mu(f) > 0}} \frac{E_f}{\Var_\mu(f)}, \qquad \Var_\mu(f) = \int_{\partial\Omega} f^2\,d\mu - \left(\int_{\partial\Omega} f\,d\mu\right)^2, \end{equation} and is invariant under Euclidean similarities. By Weinstock's classical inequality~\cite{Weinstock1954}, every simply connected planar Lipschitz domain satisfies $\Sigma(\Omega) \le 2\pi$, with equality only for disks. In higher dimensions, the perimeter-normalized bound extends to convex domains~\cite{BFNT2021}, whereas the volume-normalized inequality is due to Brock~\cite{Brock2001} and holds for all bounded Lipschitz domains; in dimension two, the area-normalized bound follows directly from Weinstock's theorem and the isoperimetric inequality. A natural polygonal counterpart, analogous to the P\'olya--Szeg\H{o} conjecture for the first Dirichlet eigenvalue and Laugesen's problem for the first positive Neumann eigenvalue, asks whether the regular $N$-gon maximizes $\Sigma(\Omega)$ among convex polygons of given perimeter with at most $N$ sides. Unlike the Dirichlet problem for triangles and quadrilaterals, where Steiner symmetrization settles the sharp bounds, symmetrization does not preserve the boundary $L^2$-norm and perimeter in a way compatible with \eqref{eq:variational}. The polygonal Steklov problem was raised by Oudet in the 2018 AIM Steklov problem list~\cite{AIM2018} and formulated explicitly for convex polygons with at most $N$ sides by Cheng, Gui, Hu, Li, and Yao~\cite[Conjecture~1.1]{CGHLY2026}, who also established that $\Sigma$ increases monotonically along the sequence of regular $N$-gons~\cite[Theorem~1.1]{CGHLY2026}. Using separation of variables and careful estimates for the resulting transcendental equations, Girouard, Lagac\'e, Polterovich, and Savo~\cite{GLPS2019} proved that the square maximizes $\sigma_1$ among rectangles of fixed perimeter or fixed area. For $N = 3$, a \href{https://github.com/jasondavies/steklov-triangles/blob/9bdb5c5f9e03c02b363058e72eb95d19b5853e36/paper/steklov-triangles.pdf}{preprint by Davies} gives a computer-assisted proof that the equilateral triangle uniquely maximizes $\Sigma$ among all triangles. Here we prove the case $N = 4$. For each triangular competitor $T$, Lemma~\ref{lem:diag} gives the independent bound $\Sigma(T) \le 3\sqrt{3}$, strictly below the value for the square. \begin{theorem}\label{thm:main} Let $\Omega \subset \mathbb{R}^2$ be a convex polygon of positive area with at most four sides, and let $k \in (0, \pi/2)$ be the unique root of $\tan k \tanh k = 1$. Then \[ \Sigma(\Omega) \le 8k\tanh k = 5.50602\dots, \] with equality if and only if $\Omega$ is a square. \end{theorem} Combining Theorem~\ref{thm:main} with the quadrilateral isoperimetric inequality gives the corresponding bound at fixed area. \begin{corollary}\label{cor:area} Among all convex planar polygons of positive area with at most four sides, \[ \sigma_1(\Omega)\sqrt{|\Omega|} \le 2k\tanh k, \] with equality if and only if $\Omega$ is a square. \end{corollary} \begin{proof} Let $\ell_1, \ell_2, \ell_3, \ell_4$ be the successive side lengths of $\Omega$, inserting a fourth collinear vertex if $\Omega$ is a triangle. Splitting $\Omega$ along each diagonal in turn yields $2|\Omega| \le \ell_1 \ell_2 + \ell_3 \ell_4$ and $2|\Omega| \le \ell_2 \ell_3 + \ell_4 \ell_1$. Adding these inequalities gives \[ 4|\Omega| \le (\ell_1 + \ell_3)(\ell_2 + \ell_4) \le \frac{(\ell_1 + \ell_2 + \ell_3 + \ell_4)^2}{4} = \frac{L^2}{4}, \] so $\sqrt{|\Omega|}/L \le 1/4$, with equality precisely for squares. Multiplying Theorem~\ref{thm:main} by $\sqrt{|\Omega|}/L \le 1/4$ and checking equality on a square completes the proof. \end{proof} Two features distinguish the general quadrilateral problem from the rectangular and triangular cases. First, the space of similarity classes of convex quadrilaterals is four-dimensional: beyond the two affine shape parameters governing parallelograms, two nonaffine parameters control trapezoidal and kite-like distortions. Second, the first positive Steklov eigenvalue of the square has multiplicity two, so $\Sigma$ is not smooth at the square and a single trial function cannot detect all perturbation directions simultaneously. Our argument proceeds in three stages. In Section~\ref{sec:coords}, we parametrize convex quadrilaterals up to similarity as bilinear images of the reference square $S = [-1,1]^2$ in four coordinates $(a, c, p, q)$, and we prove an elementary diagonal variance bound (Lemma~\ref{lem:diag}) that excludes all triangles together with all quadrilaterals satisfying $a \ge 0.21$ or $c \ge 0.045$. In Section~\ref{sec:local}, we form a weighted combination of the two first Steklov eigenfunctions of $S$ whose Rayleigh defect $T(a,c,p,q)$ is invariant under the symmetries of the square and has a negative definite Hessian at the origin; an exact degree-$22$ Taylor expansion in ball arithmetic combined with an analytic Cauchy tail bound shows that $T < 0$ throughout the central box $|a| \le 0.21$, $|c| \le 0.045$, $|p| \le 0.20$, $|q| \le 0.28$. In Section~\ref{sec:exterior}, we treat the remaining compact region by partitioning it into $940$ closed subboxes on which explicit affine or harmonic quadratic trial functions yield $\Sigma(\Omega) < 11/2 < 8k\tanh k$, completing the proof of Theorem~\ref{thm:main}. Section~\ref{sec:remarks} gathers several concluding remarks on quantitative stability, possible extensions, and computational details. \section{Shape coordinates and a diagonal estimate}\label{sec:coords} Let $S = [-1,1]^2$. Bilinear interpolation of the four positively ordered vertices of a convex quadrilateral $\Omega$ represents $\Omega$ as the image of $S$ under \begin{equation}\label{eq:bilinear} \mathbf{Z}(x, y) = \bigl(X(x, y), Y(x, y)\bigr) = \mathbf{e}x + \mathbf{f}y + \mathbf{h}xy, \end{equation} where, after applying a suitable Euclidean similarity, \[ \mathbf{e} = (1 + a,\, c), \qquad \mathbf{f} = (c,\, 1 - a), \qquad \mathbf{h} = (p,\, q). \] To reach this normalization, translate the arithmetic mean of the four vertices to the origin. The columns $\mathbf{e}, \mathbf{f}$ of the linear part of $\mathbf{Z}$ are then two consecutive semidiagonals of the Varignon midpoint parallelogram (Figure~\ref{fig:chart}, left), and positive vertex ordering ensures $\det(\mathbf{e}, \mathbf{f}) > 0$. Applying the orthogonal factor from the polar decomposition of $(\mathbf{e}\ \mathbf{f})$ rotates the linear part into a symmetric positive definite matrix, and a dilation rescales its trace to $2$. \begin{figure}[htbp] \centering \begin{tikzpicture}[scale=1.05, line cap=round, line join=round] % Left panel: bilinear chart and midpoint parallelogram \begin{scope}[shift={(0,0)}] \coordinate (V1) at (1.35, 1.05); \coordinate (V2) at (-1.05, 0.95); \coordinate (V3) at (-1.15, -0.85); \coordinate (V4) at (0.85, -1.15); \coordinate (M1) at (0.15, 1.00); \coordinate (M2) at (-1.10, 0.05); \coordinate (M3) at (-0.15, -1.00); \coordinate (M4) at (1.10, -0.05); \coordinate (O) at (0, 0); \fill[gray!14] (V1) -- (V2) -- (V3) -- (V4) -- cycle; \draw[thick] (V1) -- (V2) -- (V3) -- (V4) -- cycle; \draw[dashed] (M1) -- (M2) -- (M3) -- (M4) -- cycle; \draw[->, thick] (O) -- (M4) node[below right=-1pt] {\scriptsize $\mathbf{e}$}; \draw[->, thick] (O) -- (M1) node[above left=-1pt] {\scriptsize $\mathbf{f}$}; \foreach \pt in {V1,V2,V3,V4} \fill (\pt) circle (1.5pt); \foreach \pt in {M1,M2,M3,M4} \fill[gray!60!black] (\pt) circle (1.2pt); \fill (O) circle (1.3pt); \node[above right] at (V1) {\scriptsize $\mathbf{Z}(1,1)$}; \node[above left] at (V2) {\scriptsize $\mathbf{Z}(-1,1)$}; \node[below left] at (V3) {\scriptsize $\mathbf{Z}(-1,-1)$}; \node[below right] at (V4) {\scriptsize $\mathbf{Z}(1,-1)$}; \end{scope} % Right panel: diagonal projection in Lemma 3 \begin{scope}[shift={(6.1,0)}] \coordinate (L) at (-1.65, 0); \coordinate (R) at (1.65, 0); \coordinate (T) at (0.45, 1.15); \coordinate (B) at (-0.30, -0.95); \coordinate (P) at (0.45, 0); \fill[gray!14] (L) -- (B) -- (R) -- (T) -- cycle; \draw[thick] (L) -- (B) -- (R) -- (T) -- cycle; \draw[dashed] (L) -- (R); \draw[dotted, thick] (T) -- (P); \foreach \pt in {L,R,T,B} \fill (\pt) circle (1.5pt); \node[left] at (L) {\scriptsize $(-1,0)$}; \node[right] at (R) {\scriptsize $(1,0)$}; \node[above] at (T) {\scriptsize $(x_0, y_0)$}; \node[below] at (P) {\scriptsize $x_0$}; \node[above left] at (-0.62, 0.56) {\scriptsize $L_1$}; \node[above right] at (1.06, 0.56) {\scriptsize $L_2$}; \end{scope} \end{tikzpicture} \caption{Left: the bilinear chart $\mathbf{Z} \colon [-1,1]^2 \to \Omega$ and the inscribed midpoint parallelogram (dashed) with semidiagonals $\mathbf{e}, \mathbf{f}$. Right: the boundary arc decomposition along a normalized diagonal in Lemma~\ref{lem:diag}.} \label{fig:chart} \end{figure} Setting \begin{equation}\label{eq:jacobian} \begin{aligned} \Delta &= 1 - a^2 - c^2 > 0, & j_x &= (1 + a)q - cp, & j_y &= (1 - a)p - cq, \\ J(x, y) &= \det D\mathbf{Z}(x, y) = \Delta + j_x x + j_y y, & |\Omega| &= 4\Delta, \end{aligned} \end{equation} we see that for a quadrilateral with four strictly convex corners, consecutive edges make positive turns, so $J$ is positive at all four corners of $S$ and hence throughout $S$ by affinity. The map $\mathbf{Z}$ has topological degree $1$ at every interior point of $\Omega$ and degree $0$ outside $\overline{\Omega}$; since every interior preimage has positive Jacobian, $\mathbf{Z}$ maps $(-1,1)^2$ diffeomorphically onto the interior of $\Omega$. In particular, nonnegativity of $J$ at the corners of $S$ reads \[ |j_x| + |j_y| \le \Delta. \] Since $\mathbf{h} = (j_y \mathbf{e} + j_x \mathbf{f})/\Delta$, this convexity condition says precisely that $\mathbf{h} \in \conv(\pm \mathbf{e}, \pm \mathbf{f})$. The remaining discrete symmetries of the square act on $(a, c, p, q)$ as follows: exchanging the two coordinate axes sends $(a, c, p, q) \mapsto (-a, c, q, p)$; reflecting across one axis sends $(a, c, p, q) \mapsto (a, -c, p, -q)$; and inversion through the origin sends $(a, c, p, q) \mapsto (a, c, -p, -q)$. Composing these symmetries allows us to assume without loss of generality that $a \ge 0$, $c \ge 0$, and $p \ge 0$. Whenever $0 \le a \le 0.21$ and $0 \le c \le 0.045$, the inclusion $\mathbf{h} \in \conv(\pm \mathbf{e}, \pm \mathbf{f})$ bounds the components of $\mathbf{h}$ by \[ 0 \le p \le \max(1 + a,\, c) \le 1.21, \qquad |q| \le \max(c,\, 1 - a) \le 1. \] \begin{lemma}\label{lem:diag} Let $D$ be the length of either diagonal of a convex quadrilateral $\Omega$. Then \[ \Sigma(\Omega) \le \frac{12|\Omega|}{D^2}. \] The same inequality holds when $\Omega$ is a triangle and $D$ is the length of any of its sides. \end{lemma} \begin{proof} Align the chosen diagonal with the horizontal axis and rescale so that its endpoints lie at $(-1, 0)$ and $(1, 0)$ (Figure~\ref{fig:chart}, right). Each of the two polygonal arcs of $\partial\Omega$ joining these endpoints consists of two line segments meeting at an intermediate vertex $(x_0, y_0)$, whose lengths $L_1, L_2$ satisfy $L_1^2 - L_2^2 = (x_0 + 1)^2 - (x_0 - 1)^2 = 4x_0$. Under normalized arclength on this arc, set $\delta = (L_1 - L_2)/(L_1 + L_2)$, so the two segments carry weights $(1 + \delta)/2$ and $(1 - \delta)/2$. Conditional on each segment, the horizontal coordinate is uniform on $[-1, x_0]$ or $[x_0, 1]$, with conditional variances $(x_0 + 1)^2/12$ and $(x_0 - 1)^2/12$ and conditional means $(x_0 - 1)/2$ and $(x_0 + 1)/2$ differing by $1$. By the law of total variance, the variance of the horizontal coordinate along this arc equals \[ \frac{1 + \delta}{2}\frac{(x_0 + 1)^2}{12} + \frac{1 - \delta}{2}\frac{(x_0 - 1)^2}{12} + \frac{1 - \delta^2}{4} = \frac{1}{3} + \frac{(x_0 - \delta)(x_0 + 3\delta)}{12}. \] Because $L_1 + L_2 \ge 2$ by the triangle inequality, the identity $\delta = 4x_0/(L_1 + L_2)^2$ implies that $\delta$ has the same sign as $x_0$ and $|\delta| \le |x_0|$. Thus $(x_0 - \delta)(x_0 + 3\delta) \ge 0$, so the variance on each arc is at least $1/3$. Since the variance of a mixture of two probability distributions is at least the average of their component variances, the boundary variance over all of $\partial\Omega$ is also at least $1/3$. Restoring the length scale gives $\Var_\mu(f) \ge D^2/12$ for the unit-slope linear projection $f$, whose Dirichlet energy is $E_f = |\Omega|$. Inserting $f$ into \eqref{eq:variational} establishes the claim. For a triangle, subdividing the chosen side at its midpoint yields two arcs of the same two-segment type. \end{proof} For a triangle $\Omega$, choosing $D$ to be its longest side forces the third vertex to lie in the intersection of the two disks of radius $D$ centered at the endpoints of that side, so its altitude is at most $\sqrt{3}\,D/2$ and $|\Omega| \le \sqrt{3}\,D^2/4$. Lemma~\ref{lem:diag} then gives \begin{equation}\label{eq:triangle} \Sigma(\Omega) \le 3\sqrt{3} < \frac{11}{2}. \end{equation} For a convex quadrilateral in the chart \eqref{eq:bilinear} with $c \ge 0$, the longer diagonal has length $D = 2|\mathbf{e} + \mathbf{f}|$, where $|\mathbf{e} + \mathbf{f}|^2 = 2(1 + a^2 + c^2 + 2c)$. Applying Lemma~\ref{lem:diag} yields \begin{equation}\label{eq:diag-quad} \Sigma(\Omega) \le \frac{6\Delta}{1 + a^2 + c^2 + 2c} < \frac{11}{2} \qquad \text{whenever } 23(a^2 + c^2) + 22c > 1. \end{equation} In particular, \eqref{eq:diag-quad} holds whenever $a \ge 0.21$ or $c \ge 0.045$. \section{Local analysis near the square}\label{sec:local} On the reference square $S = [-1,1]^2$, let $k \in (0, \pi/2)$ be the unique solution of $\tan k \tanh k = 1$, and put $s = k\tanh k$. By~\cite[Theorem~1.6 and Proposition~4.2]{GLPS2019}, the first positive Steklov eigenvalue of $S$ is $\sigma_1(S) = s$, with multiplicity $2$, and the corresponding eigenspace is spanned by the harmonic functions \[ u(X, Y) = \sin(kX)\cosh(kY), \qquad v(X, Y) = \cosh(kX)\sin(kY). \] In the coordinates \eqref{eq:bilinear}, write $L$ for the perimeter of $\Omega$, put $l = L/8$, and define for $f \in H^1(\Omega)$ \[ E_f = \int_\Omega |\nabla f|^2, \qquad B_f = \int_{\partial\Omega} f^2, \qquad m_f = \int_{\partial\Omega} f, \qquad H_f = l^2 E_f - s l B_f + \frac{s}{8}m_f^2. \] Whenever $H_f < 0$, testing the centered function $f - m_f/L$ in \eqref{eq:variational} yields $L\sigma_1(\Omega) < 8s$. To probe both modes of the double eigenspace at once, we form the weighted Rayleigh defect \[ T(a, c, p, q) = (1 + a)H_u + (1 - a)H_v, \] whose weights $1 \pm a$ remain positive throughout the region considered below. The three discrete symmetries of the square act on $(a, c, p, q)$ by \[ (a, c, p, q) \longmapsto (a, -c, p, -q),\quad (a, c, -p, -q),\quad (-a, c, q, p), \] leaving $T$ invariant. At the origin $(0,0,0,0)$ we have $T(0) = 0$, since $u$ and $v$ are zero-mean Steklov eigenfunctions on $S$ and $l = 1$. Inversion eliminates every monomial of odd total degree in $(p, q)$ (in particular $p, q, ap, aq, cp, cq$ in degrees $1$ and $2$); reflection across one axis eliminates $c, ac, pq$; and coordinate exchange removes the linear term in $a$ while equating the coefficients of $p^2$ and $q^2$. Consequently, the quadratic Taylor polynomial of $T$ at the origin takes the diagonal form $\gamma_a a^2 + \gamma_c c^2 + \gamma_h(p^2 + q^2)$. Evaluating these second-order coefficients via the exact formulas of Section~\ref{sec:taylor-exact} gives \[ -5.874 < \gamma_a < -5.873, \qquad -28.171 < \gamma_c < -28.170, \qquad -1.863 < \gamma_h < -1.862. \] Thus the square is a strict local maximizer of $\Sigma$ across all four shape coordinates, including both nonaffine directions $(p, q)$. We now turn this local negativity into an explicit neighborhood of the origin. \subsection{Exact Taylor coefficients}\label{sec:taylor-exact} For $\varepsilon = \pm 1$, the constant boundary speeds along the horizontal edges $y = \varepsilon$ and vertical edges $x = \varepsilon$ of $S$ are \[ w_{0,\varepsilon} = \sqrt{(1 + a + \varepsilon p)^2 + (c + \varepsilon q)^2}, \qquad w_{\varepsilon,0} = \sqrt{(1 - a + \varepsilon q)^2 + (c + \varepsilon p)^2}, \] and the normalized perimeter is $4l = \sum_{\varepsilon=\pm 1}(w_{0,\varepsilon} + w_{\varepsilon,0})$. The area Jacobian \eqref{eq:jacobian} expands as \[ J(x, y) = 1 - a^2 - c^2 + (q + aq - cp)x + (p - ap - cq)y. \] For $u(X, Y) = \sin(kX)\cosh(kY)$, trigonometric and hyperbolic identities give \[ |\nabla u|^2 = \frac{1}{2}k^2\bigl[\cos(2kX) + \cosh(2kY)\bigr], \qquad u^2 = \frac{1}{4}\bigl[1 - \cos(2kX)\bigr]\bigl[1 + \cosh(2kY)\bigr], \] while the corresponding formulas for $v$ follow by replacing $(a, c, p, q)$ with $(-a, c, q, p)$. Every area or boundary integral entering $E_u, B_u, m_u$ is a linear combination of integrals of exponentials $\exp\bigl(k(i\alpha X + \beta Y)\bigr)$ with $\alpha, \beta \in \{0, \pm 1, \pm 2\}$, where $i = \sqrt{-1}$. Writing \[ i\alpha X + \beta Y = (i\alpha x + \beta y) + a(i\alpha x - \beta y) + c(i\alpha y + \beta x) + p(i\alpha xy) + q(\beta xy), \] the coefficient of $a^{\nu_1} c^{\nu_2} p^{\nu_3} q^{\nu_4}$ in the Taylor series of $\exp\bigl(k(i\alpha X + \beta Y)\bigr)$ is \[ e^{k(i\alpha x + \beta y)}\frac{k^{\nu_1+\nu_2+\nu_3+\nu_4}}{\nu_1!\,\nu_2!\,\nu_3!\,\nu_4!} (i\alpha x - \beta y)^{\nu_1} (i\alpha y + \beta x)^{\nu_2} (i\alpha xy)^{\nu_3} (\beta xy)^{\nu_4}. \] Each resulting one-dimensional integral of $x^n e^{\lambda x}$ reduces either to an endpoint evaluation (along an edge) or to the elementary integral $I_n(\lambda) = \int_{-1}^1 x^n e^{\lambda x}\,dx$, which obeys the recurrence \[ I_0(\lambda) = \frac{e^\lambda - e^{-\lambda}}{\lambda}, \qquad I_n(\lambda) = \frac{e^\lambda - (-1)^n e^{-\lambda} - n I_{n-1}(\lambda)}{\lambda} \qquad (\lambda \neq 0), \] with $I_n(0) = 2/(n+1)$ for even $n$ and $I_n(0) = 0$ for odd $n$. Combining these exact integrals with the finite binomial expansions of the four edge speeds $w_{0,\varepsilon}, w_{\varepsilon,0}$ through degree $N_0$ determines every Taylor coefficient of $T$ algebraically, without numerical quadrature. Isolating $k$ by bisection on $\tan k \tanh k = 1$ and evaluating the recurrences and series products in outward-rounded Arb ball arithmetic~\cite{Johansson2017} yields rigorous enclosures for all coefficients. \subsection{Cauchy remainder and the local box}\label{sec:whole-box} Write $z = (a, c, p, q)$, let $R_1, R_2, R_3, R_4 > 0$ be prescribed box half-widths (also denoted $R_a, R_c, R_p, R_q$), and define the positive definite quadratic form \[ Q(z) = 5.8 a^2 + 28 c^2 + 1.8 p^2 + 1.8 q^2 = \sum_{i=1}^4 \kappa_i z_i^2. \] From the enclosures on $\gamma_a, \gamma_c, \gamma_h$ above, the degree-$2$ part of $T(z)$ is bounded above by $-Q(z)$. For any monomial $z^\nu$ of total degree $n \ge 3$, if every exponent $\nu_i$ is even and its Taylor coefficient is nonpositive, the term is nonpositive throughout the box and may be dropped; otherwise, factoring out either $z_i^2$ (when $\nu_i \ge 2$) or $|z_i z_j|$ (when $\nu_i, \nu_j \ge 1$) and estimating the remaining factors by $|z_r| \le R_r$ gives \[ |z^\nu| \le Q(z)\,R^\nu \min\left( \left\{\frac{1}{\kappa_i R_i^2} : \nu_i \ge 2\right\} \cup \left\{\frac{1}{2\sqrt{\kappa_i \kappa_j}\,R_i R_j} : i < j,\; \nu_i \nu_j > 0\right\} \right). \] To control the Taylor tail of degree $\ge N_0 + 1$, write $z = t\zeta$ with $|\zeta_i| \le R_i$ and $|\zeta_i| = R_i$ for at least one index $i$, so that $0 \le t \le 1$ and $Q(z) \ge t^2 \min_i \kappa_i R_i^2$. Set $R_h = \max(R_p, R_q)$ and choose a radius $\rho > 1$ satisfying \[ \rho^2\bigl[(R_a + R_p)^2 + (R_c + R_q)^2\bigr] < 1, \qquad \rho^2\bigl[(R_a + R_q)^2 + (R_c + R_p)^2\bigr] < 1. \] Each squared edge speed factors as $\bigl(1 + t(\xi + i\eta)\bigr)\bigl(1 + t(\xi - i\eta)\bigr)$ with real $\xi, \eta$ obeying $\rho^2(\xi^2 + \eta^2) < 1$, so the principal square roots (positive at $t = 0$) extend holomorphically to the closed disk $|t| \le \rho$. Every other term in $T(t\zeta)$ is entire in $t$. On the circle $|t| = \rho$, define \begin{align*} \theta_0 &= \rho(R_a + R_c + R_h), & U &= \cosh(\theta_0)\cosh(1 + \theta_0), \\ G &= \frac{1}{2}\bigl[\cosh(2\theta_0) + \cosh(2 + 2\theta_0)\bigr], & w_*^2 &= \bigl[1 + \rho(R_a + R_h)\bigr]^2 + \rho^2(R_c + R_h)^2, \end{align*} and \begin{align*} J_* &= 1 + \rho(R_p + R_q) + \rho^2\bigl[R_a^2 + R_c^2 + (R_a + R_c)(R_p + R_q)\bigr], \\ M &= 2(1 + \rho R_a)w_*^2\bigl[4J_* G + 16 s U^2\bigr]. \end{align*} For every $(x, y) \in S$ and $|t| = \rho$, the complexified coordinates satisfy $|\Im X|, |\Im Y| \le \theta_0$ and $|\Re X|, |\Re Y| \le 1 + \theta_0$. Since $0 < k < 1$, the bounds $|\sin(kX)| \le \cosh(\theta_0)$ and $|\cosh(kY)| \le \cosh(1 + \theta_0)$ give $|u|, |v| \le U$, and the analytically continued energy densities have moduli at most $G$. Likewise, $|J| \le J_*$ on $S$ and each continued edge speed has modulus at most $w_*$. Therefore, \[ |E_f| \le 4J_* G, \qquad |B_f| \le 8w_* U^2, \qquad |m_f| \le 8w_* U, \qquad |l| \le w_*, \] whence $|T(t\zeta)| \le M$ on $|t| = \rho$. By Cauchy's estimate, the Taylor tail of degree $\ge N_0 + 1$ divided by $Q(z)$ is bounded uniformly by \[ \frac{M \rho^{-(N_0+1)}}{(1 - 1/\rho)\min_i \kappa_i R_i^2}. \] Whenever the sum of the relative bounds from degrees $3$ through $N_0$ and this Cauchy remainder is less than $1$, it follows that $T(z) < 0$ for every nonzero $z$ in the box. We choose the parameters \[ (R_a, R_c, R_p, R_q) = (0.21,\, 0.045,\, 0.20,\, 0.28), \qquad N_0 = 22, \qquad \rho = 1.82. \] Evaluating the Taylor coefficients and remainder estimate in Arb ball arithmetic shows that terms of degrees $3$ through $22$ contribute at most $0.5553\,Q$, that $M < 5{,}240$, and that the Cauchy tail contributes at most $0.2140\,Q$. Their sum lies below $0.7693\,Q$, yielding \begin{equation}\label{eq:local-box} T(a, c, p, q) \le -0.23\,Q(a, c, p, q) < 0 \end{equation} at every nonzero point of the central box. The same enclosure for $k$ verifies that $8s > 11/2$. \section{Exterior region: affine and harmonic quadratic trial functions}\label{sec:exterior} Retaining $s = k\tanh k$ and combining the symmetry reduction of Section~\ref{sec:coords} with \eqref{eq:diag-quad}, it remains to consider convex quadrilaterals in the parameter rectangle \[ 0 \le a \le 0.21, \qquad 0 \le c \le 0.045, \qquad 0 \le p \le 1.21, \qquad -1 \le q \le 1 \] lying outside the central box $p \le 0.20$, $|q| \le 0.28$. On this complement, it suffices to exhibit a trial function $f \in H^1(\Omega)$ satisfying \begin{equation}\label{eq:ext-goal} 11\Var_\mu(f) - 2E_f > 0, \end{equation} for then \eqref{eq:variational} implies $\Sigma(\Omega) < 11/2 < 8s$. Under the bilinear map \eqref{eq:bilinear}, the four edges of $\Omega$ are parametrized by $\mathbf{M}_i + \mathbf{V}_i\tau$ for $-1 \le \tau \le 1$ ($i = 1, 2, 3, 4$), where the pairs $(\mathbf{M}_i, \mathbf{V}_i)$ are (up to orientation) \[ (\mathbf{f},\, \mathbf{e} + \mathbf{h}), \qquad (-\mathbf{f},\, \mathbf{e} - \mathbf{h}), \qquad (\mathbf{e},\, \mathbf{f} + \mathbf{h}), \qquad (-\mathbf{e},\, \mathbf{f} - \mathbf{h}). \] Set $w_i = |\mathbf{V}_i|$ and $W = \sum_{i=1}^4 w_i$, so that $L = 2W$. \subsection{Affine trial functions} For a constant vector $\mathbf{d} \in \mathbb{R}^2$, the linear function $f(\mathbf{Z}) = \mathbf{d}\cdot\mathbf{Z}$ has Dirichlet energy $E_f = 4\Delta|\mathbf{d}|^2$. Given a scalar shift $b_0 \in \mathbb{R}$, put $A_i = \mathbf{d}\cdot\mathbf{M}_i - b_0$ and $B_i = \mathbf{d}\cdot\mathbf{V}_i$. Since $\frac{1}{2}\int_{-1}^1 \tau^2\,d\tau = 1/3$, condition \eqref{eq:ext-goal} is equivalent to \begin{equation}\label{eq:affine-test} F_1 := \sum_{i=1}^4 w_i\bigl(33 A_i^2 + 11 B_i^2 - 24\Delta|\mathbf{d}|^2\bigr), \qquad W F_1 > 33\left(\sum_{i=1}^4 w_i A_i\right)^2, \end{equation} because the difference between the left- and right-hand sides equals $3W^2\bigl(11\Var_\mu(f) - 2E_f\bigr)$. \subsection{Harmonic quadratic trial functions} Let $\mathbf{d} \in \mathbb{R}^2$ and let $K = \bigl(\begin{smallmatrix} r_1 & r_2 \\ r_2 & -r_1 \end{smallmatrix}\bigr)$ be a traceless symmetric $2 \times 2$ matrix, and consider the harmonic quadratic polynomial \[ f(\mathbf{Z}) = \mathbf{d}\cdot\mathbf{Z} + \mathbf{Z}^T K \mathbf{Z}. \] Its gradient is $\nabla f(\mathbf{Z}) = \mathbf{d} + 2K\mathbf{Z}$, and $K^2 = (r_1^2 + r_2^2)I$. Integrating \eqref{eq:jacobian} over the symmetric square $S$ yields the three physical area moments \begin{align*} \int_\Omega 1 &= 4\Delta, \qquad \int_\Omega \mathbf{Z} = \frac{4}{3}(j_x \mathbf{e} + j_y \mathbf{f}), \\ 9\int_\Omega |\mathbf{Z}|^2 &= 4\Delta\bigl(3|\mathbf{e}|^2 + 3|\mathbf{f}|^2 + |\mathbf{h}|^2\bigr) + 8\bigl(j_x \mathbf{f}\cdot\mathbf{h} + j_y \mathbf{e}\cdot\mathbf{h}\bigr), \end{align*} from which the Dirichlet energy follows explicitly as \[ E_f = 4\Delta|\mathbf{d}|^2 + 4(K\mathbf{d})\cdot\int_\Omega \mathbf{Z} + 4(r_1^2 + r_2^2)\int_\Omega |\mathbf{Z}|^2. \] Along edge $i$, write $f(\mathbf{M}_i + \mathbf{V}_i\tau) - b_0 = A_i + B_i\tau + C_i\tau^2$, where \[ A_i = \mathbf{d}\cdot\mathbf{M}_i + \mathbf{M}_i^T K \mathbf{M}_i - b_0, \qquad B_i = \mathbf{d}\cdot\mathbf{V}_i + 2\mathbf{M}_i^T K \mathbf{V}_i, \qquad C_i = \mathbf{V}_i^T K \mathbf{V}_i. \] Put $\phi_i = 3A_i + C_i$ and $\psi_i = 5\phi_i^2 + 15B_i^2 + 4C_i^2$. Direct integration over $\tau \in [-1,1]$ shows that the edge averages of $f - b_0$ and $(f - b_0)^2$ equal $\phi_i/3$ and $\psi_i/45$, respectively. Clearing denominators in \eqref{eq:ext-goal} yields the equivalent polynomial inequality \begin{equation}\label{eq:quad-test} F_2 := \sum_{i=1}^4 w_i(11 \psi_i - 90 E_f), \qquad W F_2 > 55\left(\sum_{i=1}^4 w_i \phi_i\right)^2, \end{equation} whose left-hand side minus its right-hand side equals $45 W^2\bigl(11\Var_\mu(f) - 2E_f\bigr)$. \subsection{Subdivision of the exterior region}\label{sec:subdiv} We decompose the exterior of the central box inside $[0, 0.21] \times [0, 0.045] \times [0, 1.21] \times [-1, 1]$ into six closed root boxes sharing the same $(a, c)$ intervals (Table~\ref{tab:exterior}): the first three cover the complement of $[0, 0.40] \times [-0.60, 0.60]$ in the $(p, q)$ rectangle, and the remaining three cover the portion of $[0, 0.40] \times [-0.60, 0.60]$ outside $[0, 0.20] \times [-0.28, 0.28]$. Because every box in the partition is closed, all shared boundary faces are accounted for. \begin{table}[htbp] \centering \caption{Subboxes in the exterior partition, classified by elimination criterion. All six root boxes share $a \in [0, 0.21]$ and $c \in [0, 0.045]$.} \label{tab:exterior} {\small\begin{tabular}{llrrrrr} \toprule $p$ & $q$ & Affine & Diagonal & Infeasible & Quadratic & Total \\ \midrule $[0.40, 1.21]$ & $[-1, 1]$ & $65$ & $0$ & $8$ & $0$ & $73$ \\ $[0, 0.40]$ & $[0.60, 1]$ & $60$ & $9$ & $0$ & $0$ & $69$ \\ $[0, 0.40]$ & $[-1, -0.60]$ & $187$ & $36$ & $1$ & $0$ & $224$ \\ $[0.20, 0.40]$ & $[-0.60, 0.60]$ & $37$ & $0$ & $0$ & $180$ & $217$ \\ $[0, 0.20]$ & $[0.28, 0.60]$ & $19$ & $1$ & $0$ & $107$ & $127$ \\ $[0, 0.20]$ & $[-0.60, -0.28]$ & $19$ & $3$ & $0$ & $208$ & $230$ \\ \midrule Total & & $387$ & $49$ & $9$ & $495$ & $940$ \\ \bottomrule \end{tabular}} \end{table} A subbox is eliminated if outward-rounded interval enclosures establish either $|j_x| + |j_y| > \Delta$ throughout the subbox (so that it contains no admissible convex quadrilateral), or one of the spectral inequalities \eqref{eq:diag-quad}, \eqref{eq:affine-test}, or \eqref{eq:quad-test}. All interval endpoints are integer multiples of $2^{-32}$, with rational root endpoints rounded outward; interval products use integer floor and ceiling, and square roots use integer square roots rounded outward. On each subbox, a midpoint covariance computation or generalized eigenproblem in floating point suggests candidate affine or harmonic quadratic coefficients, which are rounded to integers (with $b_0$ rounded to $2^{-32}\mathbb{Z}$). Once these explicit trial parameters are fixed, the verification of $F_1 > 0$ and \eqref{eq:affine-test}, or $F_2 > 0$ and \eqref{eq:quad-test}, is carried out entirely in integer interval arithmetic and does not rely on the accuracy of the floating-point step. When none of the four criteria succeeds on a subbox, we bisect the coordinate of largest weighted width using weights $(2, 2, 1, 1)$ on $(a, c, p, q)$, keeping the midpoint in both closed child subboxes. Across the six root boxes, $934$ bisections suffice, yielding a closed cover of $940$ subboxes (Table~\ref{tab:exterior}) on every one of which an elimination criterion holds. \begin{proof}[Completion of the proof of Theorem~\ref{thm:main}] Every triangle satisfies \eqref{eq:triangle}, and $11/2 < 8s$. For a convex quadrilateral $\Omega$, pass to the normalized chart \eqref{eq:bilinear} with $a, c, p \ge 0$. If $a \ge 0.21$ or $c \ge 0.045$, inequality \eqref{eq:diag-quad} yields $\Sigma(\Omega) < 11/2 < 8s$. Otherwise, convexity restricts $(p, q)$ to $[0, 1.21] \times [-1, 1]$. On the central box $p \le 0.20$, $|q| \le 0.28$, inequality \eqref{eq:local-box} gives $T < 0$ and hence $\Sigma(\Omega) < 8s$ whenever $(a, c, p, q) \neq (0, 0, 0, 0)$. Every remaining admissible tuple $(a, c, p, q)$ lies in at least one of the $940$ closed subboxes of Table~\ref{tab:exterior}, where again $\Sigma(\Omega) < 11/2 < 8s$. At $(a, c, p, q) = (0, 0, 0, 0)$, the domain is the reference square $S = [-1,1]^2$ with $\Sigma(S) = 8s$. Undoing the similarity normalization shows that equality holds precisely for squares. \end{proof} \section{Concluding remarks}\label{sec:remarks} We close with three observations concerning stability, possible extensions, and the computational implementation. \medskip \noindent\textbf{Quantitative stability.} Although Theorem~\ref{thm:main} is stated as a sharp inequality with equality characterization, the proof yields a quantitative stability estimate near the square. Indeed, for any $f \in H^1(\Omega)$ we have \[ H_f = l^2\bigl(E_f - 8s\Var_\mu(f)\bigr) \ge l^2\Var_\mu(f)\bigl(\Sigma(\Omega) - 8s\bigr), \] so \eqref{eq:local-box} implies that \[ 8s - \Sigma(\Omega) \ge \frac{0.23\,Q(a, c, p, q)}{l^2\bigl[(1+a)\Var_\mu(u) + (1-a)\Var_\mu(v)\bigr]} \] throughout the central box. Because the denominator is continuous and bounded above on the compact central box, while $8s - \Sigma(\Omega) > 8s - 11/2 > 0$ outside it and $a^2 + c^2 + p^2 + q^2$ is uniformly bounded across all convex quadrilaterals (since $a^2 + c^2 < 1$ and $\mathbf{h} \in \conv(\pm\mathbf{e}, \pm\mathbf{f})$), the Steklov deficit $8k\tanh k - \Sigma(\Omega)$ controls $a^2 + c^2 + p^2 + q^2$ in the normalized chart \eqref{eq:bilinear} up to an explicit positive constant. \medskip \noindent\textbf{Nonconvex quadrilaterals and higher polygons.} Convexity enters our proof in two places: the bilinear diffeomorphism \eqref{eq:bilinear} and the interior-diagonal decomposition of Lemma~\ref{lem:diag}. It is natural to ask whether Theorem~\ref{thm:main} remains valid for nonconvex simple quadrilaterals. Note that reflecting the two edges incident to a reflex vertex across the line joining its neighbors strictly increases the area while preserving the perimeter, so the quadrilateral isoperimetric bound $\sqrt{|\Omega|}/L < 1/4$ still holds for all nonconvex simple quadrilaterals; consequently, any extension of Theorem~\ref{thm:main} to nonconvex quadrilaterals would automatically carry Corollary~\ref{cor:area} along with it. For convex $N$-gons with $N \ge 5$, two new obstacles arise: the dimension $2N - 4$ of the moduli space grows with $N$, and explicit closed-form Steklov eigenfunctions are no longer available for regular $N$-gons when $N \ge 5$, so a local analysis near the regular $N$-gon would require rigorous approximations to the first Steklov eigenfunctions together with a lower bound on the spectral gap $\sigma_3 - \sigma_1$. \medskip \noindent\textbf{Computational details.} The finite calculations of Sections~\ref{sec:local} and~\ref{sec:exterior} are performed by compact, self-contained Python programs. For the local box, the recurrences of Section~\ref{sec:taylor-exact} and the Cauchy tail estimate of Section~\ref{sec:whole-box} are evaluated using Arb ball arithmetic~\cite{Johansson2017} at $200$-bit working precision, isolating $k$ to an interval of width at most $2^{-150}$. For the exterior region, the subdivision of Section~\ref{sec:subdiv} uses exact integer interval arithmetic on $2^{-32}\mathbb{Z}$. All analytic steps connecting these finite inequalities to the Steklov spectrum are contained in Sections~\ref{sec:coords}--\ref{sec:exterior}. \begin{thebibliography}{10} \bibitem{AIM2018} American Institute of Mathematics, \emph{Steklov eigenproblems}, 2018 workshop problem list, Section~4, Problem~4.6 (\'E.~Oudet). \bibitem{Brock2001} F.~Brock, An isoperimetric inequality for eigenvalues of the Stekloff problem, \emph{Z. Angew. Math. Mech.} \textbf{81} (2001), no.~1, 69--71. \bibitem{BFNT2021} D.~Bucur, V.~Ferone, C.~Nitsch, and C.~Trombetti, Weinstock inequality in higher dimensions, \emph{J. Differential Geom.} \textbf{118} (2021), no.~1, 1--21. \bibitem{CGHLY2026} Z.~Cheng, C.~Gui, Y.~Hu, Q.~Li, and R.~Yao, Global strict monotonicity and asymptotics of the first Steklov eigenvalue for regular polygons, preprint (2026), \href{https://arxiv.org/abs/2603.25116v2}{arXiv:2603.25116v2}. \bibitem{CGGS2024} B.~Colbois, A.~Girouard, C.~Gordon, and D.~Sher, Some recent developments on the Steklov eigenvalue problem, \emph{Rev. Mat. Complut.} \textbf{37} (2024), no.~1, 1--161. \bibitem{GLPS2019} A.~Girouard, J.~Lagac\'e, I.~Polterovich, and A.~Savo, The Steklov spectrum of cuboids, \emph{Mathematika} \textbf{65} (2019), no.~2, 272--310. \href{https://arxiv.org/abs/1711.03075}{arXiv:1711.03075}. \bibitem{GP2017} A.~Girouard and I.~Polterovich, Spectral geometry of the Steklov problem, \emph{J. Spectr. Theory} \textbf{7} (2017), no.~2, 321--359. \bibitem{Johansson2017} F.~Johansson, Arb: efficient arbitrary-precision midpoint-radius interval arithmetic, \emph{IEEE Trans. Comput.} \textbf{66} (2017), no.~8, 1281--1292. \href{https://doi.org/10.1109/TC.2017.2690633}{doi:10.1109/TC.2017.2690633}. \bibitem{Weinstock1954} R.~Weinstock, Inequalities for a classical eigenvalue problem, \emph{J. Rational Mech. Anal.} \textbf{3} (1954), 745--753. \end{thebibliography} \end{document}