Existence of metrics maximizing the first Laplace eigenvalue on closed surfaces
Mikhail Karpukhin, Romain Petrides, Daniel Stern
TL;DR
The paper resolves the global existence of metrics maximizing the normalized first Laplace eigenvalue on all closed surfaces by proving strict monotonicity of the conformal supremum under both cross-cap and handle attachments. It introduces refined extension operators that distinguish even/odd symmetries on the Möbius band and cylinder to obtain first-order estimates at attachment points, forcing vanishing derivatives of sphere-valued eigenmaps and leading to contradictions when maximizing would fail. Consequently, it proves $\Lambda_1(M\#\mathbb{RP}^2) > \Lambda_1(M)$ and $\Lambda_1(M\#\mathbb{T}^2) > \Lambda_1(M)$ for every closed $M$, and deduces the existence (smooth away from finitely many conical points) of $\bar{\lambda}_1$-maximizing metrics on all topologies. The results bridge orientable and nonorientable cases and connect extremal metrics to minimal immersion theory, providing a unified framework for this class of spectral optimization problems.
Abstract
Building on seminal work of Nadirashvili and previous work of the authors, we prove the existence of metrics maximizing the area-normalized first eigenvalue of the Laplacian on every closed nonorientable surface, and give a simple new proof of existence in the orientable case complementing that of [Pet24b], thus resolving the long-standing existence problem for $λ_1$-maximizing metrics on closed surfaces of any topology. Namely, we prove by contradiction that the supremum $Λ_1(M)$ of the normalized first eigenvalue over all metrics on $M$ obeys the strict monotonicity $Λ_1(M\#\mathbb{RP}^2)>Λ_1(M)$ and $Λ_1(M\#\mathbb{T}^2)>Λ_1(M)$ under the attachment of cross-caps and handles, via a substantial refinement of techniques introduced in [KKMS24].
