6
$\begingroup$

The following question is quite natural, but I am not aware of a reference dealing with it: let $M$ be a compact (smooth) manifold (posssibly with boundary) and $E$ a vector bundle on $M$ with an Hermitian or Euclidean metric, and let $g_u,\: u\in[0,1]$ be a smooth family of Riemannian metrics on $M$. For each $p=0,\ldots,{\rm dim}\: M$ we get a map ${\rm sp}(\Delta^p):\: [0,1] \to {\mathbb R}^{\mathbb N}$ which associates tu $u$ the ordered spectrum (with multiplicities) of the Laplace operator on $p$-forms on $(M,g_u)$ with coefficients in $E$. The question is: is this map continuous (with respect to the product topology on ${\mathbb R}^{\mathbb N}$)?

I believe that this is true, and that you can prove it using the min-max characterization of the spectrum, but as it seems a rather basic question I would think that somebody already answered some version of it.

$\endgroup$
2
  • $\begingroup$ To define your operators, you must also specify a connection on the vector bundle $E$. $\endgroup$
    – BS.
    Jun 10, 2013 at 8:15
  • $\begingroup$ Indeed; I should have added that the bundle $E$ is supposed to be flat (otherwise I guess you can add an arbitrary connection to the data). $\endgroup$ Jun 10, 2013 at 8:39

2 Answers 2

8
$\begingroup$

However, for continuity of individual eigenvalues (with appropriate provisos in the presence of multiplicity), the min-max characterization works fine. More refined statements (e.g. about the behavior of the spectrum for smooth or analytic one-parameter families of metrics) are contained in Kato's classic book `Perturbation theory for Linear Operators'.

$\endgroup$
1
  • $\begingroup$ Thanks for the reference; it seems the theorems on analytic perturbation of operators in Chapter VII can be applied to the problem. $\endgroup$ Jun 11, 2013 at 11:06
2
$\begingroup$

Maybe you find this this helpful. It answers the analogous question for the Dirac operator.

$\endgroup$

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.