22.06.2013 Views

Polyharmonic boundary value problems

Polyharmonic boundary value problems

Polyharmonic boundary value problems

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

1.3 The first eigen<strong>value</strong> 15<br />

also [244, Proposition 4.4]. Therefore, for a complete proof of (1.24), “only” Item<br />

2 is missing!<br />

1.3.3 A Steklov eigen<strong>value</strong> problem<br />

Usually, eigen<strong>value</strong> <strong>problems</strong> arise when one studies oscillation modes in the respective<br />

time dependent problem in order to have a physically well motivated theory<br />

and representation of solutions.<br />

However, in what follows, a most natural motivation for considering a further<br />

eigen<strong>value</strong> problem comes from a seemingly quite different mathematical question.<br />

We explain how L2-estimates for the Dirichlet problem for harmonic functions link<br />

with the Steklov eigen<strong>value</strong> problem for biharmonic functions.<br />

Let Ω ⊂ Rn be a bounded smooth domain and consider the problem<br />

<br />

∆u = 0 in Ω,<br />

(1.25)<br />

u = g on ∂Ω,<br />

where g ∈ L 2 (∂Ω). It is well-known that (1.25) admits a unique solution u ∈<br />

H 1/2 (Ω) ⊂ L 2 (Ω), see e.g. [275, Remarque 7.2, p. 202] and also [237, 238] for<br />

an extension to nonsmooth domains. One is then interested in a priori estimates,<br />

namely in determining the sharp constant CΩ such that<br />

u L 2 (Ω) ≤ CΩ g L 2 (∂Ω) .<br />

By Fichera’s principle of duality [170] (see also Section 3.3.2) one sees that CΩ<br />

coincides with the inverse of the first Steklov eigen<strong>value</strong> δ1 = δ1(Ω), namely the<br />

smallest constant a such that the problem<br />

<br />

∆ 2u = 0 in Ω,<br />

(1.26)<br />

u = ∆u − auν = 0 on ∂Ω,<br />

admits a nontrivial solution. Notice that the “true” eigen<strong>value</strong> problem for the hinged<br />

plate equation should include the curvature in the second <strong>boundary</strong> condition, see<br />

(1.8). The map Ω ↦→ δ1(Ω) has several surprising properties which we establish<br />

in Section 3.3.2. By rescaling, one sees that δ1(kΩ) = k −1 δ1(Ω) for any bounded<br />

domain Ω and any k > 0 so that δ1(kΩ) → 0 as k → ∞. One is then led to seek<br />

domains which minimise δ1 under suitable constraints, the most natural one being<br />

the volume constraint. Smith [373] stated that, analogously to the Faber-Krahn result<br />

[162, 253, 254], the minimiser for δ1 should exist and be a ball, at least for planar<br />

domains. But, as noticed by Kuttler and Sigillito, the argument in [373] contains a<br />

gap. In the “Note added in proof” in [374, p. 111], Smith writes:<br />

Although the result is probably true, a correct proof has not yet been found.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!