KEI FUNANO
3 min
We establish two universal inequalities for Dirichlet eigenvalues of the Laplacian on a Euclidean convex domain.
This paper by Kei Funano establishes universal inequalities—geometry-independent bounds on ratios of Dirichlet eigenvalues of the Laplacian—for bounded convex domains in ℝⁿ. These results control how eigenvalues λ_k(Ω) scale with their index k relative to lower eigenvalues λ_l(Ω), motivated by Weyl's law asymptotics λ_k ∼ (4π² k / (ω_n vol(Ω)))^{2/n}.
The Dirichlet Laplacian on a bounded domain Ω solves -Δu = λu in Ω with u=0 on ∂Ω, yielding a discrete spectrum 0 < λ₁(Ω) ≤ λ₂(Ω) ≤ ⋯ → ∞. These eigenvalues measure the 'vibrational frequencies' of the domain with fixed boundaries, analogous to a drumhead. Convex domains (bounded convex sets in ℝⁿ) are the focus, as their geometry allows sharp approximations via Hatcher's lemma and John's theorem.
Weyl's law predicts the large-k asymptotic, suggesting ratios λ_k/λ_l should behave like (k/l)^{2/n}. Universal inequalities make this precise with explicit constants: for any convex Ω and k ≥ l,
λ_k(Ω) ≤ c_n (k/l)^{2/n} λ_l(Ω),
where c_n = 12n³ (j_{n/2 -1,1})² and j_{ν,1} is the first positive zero of the Bessel function J_ν. This upper bound holds universally, independent of Ω's specific shape.
A conditional lower bound holds when λ_l(Ω) > α_n λ_1(Ω) (α_n = 4/(2π² (j_{n/2 -1,1})² n³)):
β_n (k/l)^{2/n} λ_l(Ω) ≤ λ_k(Ω),
with β_n = 1/(12n³ (j_{n/2 -1,1})²). Without this condition, counterexamples like thin rectangles [0,k]×[0,1] fail it. As a corollary, eigenvalue multiplicities m_k(Ω) are bounded: m_k(Ω) ≤ 12^{n/2} n^{3/2^n} (j_{n/2 -1,1})^{-n} k under the same condition, controlling degeneracy.
Key tools: Domain monotonicity—if Ω₁ ⊆ Ω₂ then λ_k(Ω₂) ≤ λ_k(Ω₁), so bounds on smaller domains imply larger ones. Hatcher's lemma approximates any convex Ω by inscribed/exscribed orthotopes (rectangular boxes), reducing proofs to explicit rectangle eigenvalues (products of 1D π²/a_i²). This geometric sandwich yields sharp constants via Bessel function zeros from 1D separation-of-variables solutions.
These results advance spectral geometry by providing the first explicit lower bounds for arbitrary eigenvalue ratios on convex domains, with applications to eigenvalue multiplicity and asymptotic control beyond consecutive ratios.
Sam: It unifies past inequalities and could help design resonators or quantum devices by matching target tones to convex shapes. Thanks for listening to ResearchPod.