13
$\begingroup$

Consider a compact connected surface $M$ of some genus $\gamma \geq 2$. A particular case of a famous result of Colin de Verdière (see Construction de laplaciens dont une partie finie du spectre est donnée) says that if we fix $\gamma$ and select a finite sequence $\lambda_0 = 0 < \lambda_1 < \dotsb <\lambda_n$ (note the strict inequality, which is meant to avoid potential constraints on multiplicity in dimension $2$), then we can find a smooth metric $g$ on $M$ such that the given $\lambda_i$'s are the first $n$ eigenvalues of the Laplacian associated to $g$.

Now, my question (which could be trivial, but unfortunately I have no way of knowing as I am totally ignorant of French and nearly as ignorant of the graph-theoretic methods that de Verdière uses) is, if we do not fix $\gamma$, could we also arrange that the metric $g$ be hyperbolic, that is, constant negative curvature? If not, could we at least say that we can pick a $g$ whose curvature is arbitrarily close to being constant?

The reason I even remotely hope for a positive answer here is due to a result of Lohkamp (see Curvature $h$-principles), which says that in compact connected manifolds of higher dimensions, one can improve de Verdière's result in that one can additionally constrain the metric $g$ to have an arbitrary negative upper bound on the Ricci curvature.

Any insights are highly appreciated, thanks!

Edit after Igor Rivin's comment: I should explicitly mention again that I do not want to fix the genus of the surface. In other words, given a finite increasing sequence (starting from $0$) as above, can we find a compact hyperbolic (or close to hyperbolic) surface $(M, g)$ of some genus $\gamma \geq 2$ such that the first $n$ eigenvalues of $(M, g)$ agree with the given sequence?

$\endgroup$
4
  • 2
    $\begingroup$ There are restrictions on the spectrum of a genus $g$ hyperbolic surface: projecteuclid.org/euclid.dmj/1253020546 $\endgroup$
    – Ian Agol
    Oct 20, 2015 at 23:55
  • 1
    $\begingroup$ A remark on the first question. The moduli space of hyperbolic structures on a surface is finite dimensional (6g-6). So it cannot easily (for instance differentiably) map onto a space with more than 6g-6 parameters. $\endgroup$
    – Thomas
    Oct 21, 2015 at 7:35
  • 2
    $\begingroup$ Actually, the question is much more interesting if you read it as not restricting the genus. Then both the comments become irrelevant (or at least less relevant) $\endgroup$
    – Igor Rivin
    Oct 21, 2015 at 10:17
  • $\begingroup$ @IgorRivin Thanks, that is actually what I wanted to ask, but got a bit confused and did not word it properly. Now I have edited it. $\endgroup$
    – SMS
    Oct 21, 2015 at 10:47

1 Answer 1

3
$\begingroup$

Vedrin Šahović in his unpublished thesis [Approximations of Riemannian manifolds with linear curvature constraints, 2009] proved that any compact metric space can be appoximated by hyperbolic surfaces $M_n$ with constant curvature $\kappa_n$. In this case $\kappa_n\to -\infty$ and $\gamma_n\to \infty$.

His construction is quite flexible — it is possible to get in addition convergence of the first $n$ eigenvalues (but please, do not ask me how to do it). So at least one can get a positive answer to approximate version of your question — we can get a hyperbolic manifold with first $n$ eigenvalues arbitrarily close to the given sequence.

By the way, the same holds in all dimensions.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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