2
$\begingroup$

Given a bounded polygonal domain $D$ in $\mathbb{R}^2$, the Neumann eigenfunctions have continuous version on $\overline{D}$. The eigenfunctions also have critical points at vertices of $D$ (I have been taught the simple proof by Prof. Mateusz Kwaśnicki at this page).

However, the modulus of continuity of the eigenfunctions at vertices seem not obvious. For example, when $D$ is an equilateral triangle and $\phi$ is a Neumann eigenfunction on $D$, can we describe the modulus of continuity of $\phi$ at a vertex $p$ ? It holds that $|(\nabla \phi)(p)|=0$. Can we expect that $|\phi(x)-\phi(p)|=o(|x-p|^\alpha)$ as $x \to p$ for some $\alpha>1$.

If so, what is the optimal exponent?

$\endgroup$
5
  • 4
    $\begingroup$ Grisvard's book is a standard reference for this. $\endgroup$ Commented Feb 16, 2020 at 12:35
  • 1
    $\begingroup$ @MichaelRenardy Thank you very much for your comment. Is the title of the book "Elliptic Problems in Nonsmooth Domains"? I'll check it. $\endgroup$
    – sharpe
    Commented Feb 16, 2020 at 13:35
  • 1
    $\begingroup$ To complete Michael's comment, the order of the singularity at a corner is explicit and depends only upon the angle of this corner. Its magnitude can be expressed as an integral, but this is not so much explicit. $\endgroup$ Commented Feb 19, 2020 at 16:55
  • $\begingroup$ @DenisSerre Thank you very much for your comment. For example, can you tell me the order for Neumann eigenfunctions on equilateral triangle? $\endgroup$
    – sharpe
    Commented Feb 19, 2020 at 17:14
  • $\begingroup$ I suspect that the order is greater than or equal to $2$. $\endgroup$
    – sharpe
    Commented Feb 19, 2020 at 18:01

1 Answer 1

1
$\begingroup$

Let me develop a little my comment. Consider a corner of the domain, with tip at the origin and one edge horizontal. Denote $\alpha$ its angle of aperture. The leading singularity of a solution of $\Delta u=f$ with a smooth $f$, and Neumann condition, is $A\Re(z^\kappa)$. With $z=e^{i\theta}$, its normal derivative is $r^{-1}\partial_\theta$. Whence the condition $\sin(\kappa\alpha)=0$, where we must take the smallest root $\kappa$, in absence of better information: $$\kappa=\frac\pi\alpha.$$ For $\alpha=\frac\pi3$, this gives $\kappa=3$, which is not so bad as the eigenfunction is $C^{3-\epsilon}$. But the larger $\alpha$, the less regular are the eigenfunctions.

Edit. When $f=\lambda u$, which is in $H^1(\Omega)$, it has enough "smoothness" that the above applies.

$\endgroup$
1
  • $\begingroup$ Thank you for your kind reply. I have a question on it. In the proof, you assume that $f$ is smooth, right? In your definition, "smooth" is $C^1(\overline{D})$? I am also considering the regularity of eigenfunctions of the Neumann Laplacian and the Neumann semigroup. Hence, it is inconvenient to assume that f is smooth in advance. $\endgroup$
    – sharpe
    Commented Feb 21, 2020 at 9:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .