Suppose I have a compact surface (possibly with boundary), and consider the eigenfunctions of the Laplacian, normalized so that their $L^2$ norms are $1.$ Is there some general result or conjecture on how the $L^p$ norm of $f_\lambda$ (where $\lambda$ is the eigenvalue) behaves as function of $\lambda?$ Presumably this depends on the geometry of the surface (and, of course, nothing about this question is special to two dimensions, except that this case might be easier).

  • 2
    $\begingroup$ Search for Sogge's results on $L^p$ bounds. I believe this is an active area of study, especially when one restricts to manifolds with interesting geometric properties. $\endgroup$ Sep 5, 2014 at 1:22
  • $\begingroup$ I haven't read this paper myself, but I believe the work of Sogge (J. Funct. Analysis 1988) may have results of the kind you want? $\endgroup$
    – Lucia
    Sep 5, 2014 at 1:24
  • $\begingroup$ For "interesting" cases, much better bounds are known (and expected) than the general theory by Sogge. See for example the result of Iwaniec-Sarnak and all the recent work on Berry's random wave model. $\endgroup$
    – Asaf
    Sep 5, 2014 at 7:42
  • $\begingroup$ @Asaf Any particular paper you recommend (on the Berry random wave thing)? $\endgroup$
    – Igor Rivin
    Sep 5, 2014 at 13:29
  • $\begingroup$ This arxiv.org/pdf/0903.3420v1.pdf very nice survey article of Zelditch contains quite a bit of discussion about this very topic, including an overview of some of the work of Sogge. $\endgroup$ Sep 5, 2014 at 14:08

1 Answer 1


As pointed out already by the comments, Sogge has indeed made a lot of contributions in this area. Consider a 2-dimensional compact Riemannian manifold without boundary, then the $L^2$ normalized eigenfunction of Laplacian $e_{\lambda}$ which sastisfy $$ -\Delta e_{\lambda}=\lambda^2e_{\lambda} $$ has the following estimate $$ \|e_{\lambda}\|_{L^p}\leq C\lambda^{\sigma(p)},~~\lambda\ge 1, $$ here $\sigma(p)=\frac{1}{2}(\frac{1}{2}-\frac{1}{p})$, if $2\leq p\leq 6$, and $\sigma(p)=2(\frac{1}{2}-\frac{1}{p})-1$, if $6\leq p\leq \infty$. For dimension $n>2$, there are similar results.

However, for manifold with boundary, the problem becomes more difficult, for dimension 2, see also the Acta paper by Smith and Sogge http://arxiv.org/pdf/math/0605682.pdf.

It's also interesting to improve the bound above. The above estimates is sharp in general. The $L^{\infty}$ norm is saturated by the Zonal function (which is concentrated near the pole) on the round sphere. For lower $p$(below the critical point, which is 6 here), it's saturated by the highest weight hamonics (which is concentrated near the equator ). However, for manifold with nonpositive curvature, on can get some improvement($\log \lambda$), and for flat torus $\mathbb{T}^n$, one can get a further improvement ($\lambda^{\epsilon(n)})$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.