ConceptComplete

Spectral Theory of the Laplacian

The spectrum of the Hodge Laplacian encodes geometric and topological information about the manifold. Spectral geometry asks: "Can one hear the shape of a drum?"


The Eigenvalue Problem

Definition10.5Spectrum of the Laplacian

On a compact Riemannian manifold (M,g)(M, g) (without boundary), the eigenvalue problem for the Hodge Laplacian on kk-forms is

Δω=λω,ωΩk(M).\Delta \omega = \lambda \omega, \quad \omega \in \Omega^k(M).

The spectrum Speck(M,g)\mathrm{Spec}_k(M, g) is the set of eigenvalues 0λ0λ1λ20 \leq \lambda_0 \leq \lambda_1 \leq \lambda_2 \leq \cdots \to \infty, counted with multiplicity.

Theorem10.4Spectral theorem for the Hodge Laplacian

On a compact Riemannian manifold:

  1. The eigenvalues 0λ0λ10 \leq \lambda_0 \leq \lambda_1 \leq \cdots are non-negative and tend to ++\infty.
  2. The eigenspaces Eλ={ω:Δω=λω}E_\lambda = \{\omega : \Delta\omega = \lambda\omega\} are finite-dimensional.
  3. The eigenforms form a complete orthonormal basis for L2Ωk(M)L^2\Omega^k(M).
  4. The multiplicity of λ=0\lambda = 0 equals the kk-th Betti number bkb_k.

Spectral Geometry

ExampleSpectra of model spaces
  1. Circle S1S^1 of length 2π2\pi: On functions, Δf=f\Delta f = -f'', eigenvalues λn=n2\lambda_n = n^2 (n0n \geq 0), with eigenfunctions cos(nθ)\cos(n\theta), sin(nθ)\sin(n\theta).

  2. Sphere SnS^n: On functions, λk=k(k+n1)\lambda_k = k(k+n-1) with multiplicity (k+nn)(k+n2n)\binom{k+n}{n} - \binom{k+n-2}{n} (spherical harmonics of degree kk).

  3. Flat torus Tn=Rn/ΛT^n = \mathbb{R}^n/\Lambda: On functions, λv=4π2v2\lambda_v = 4\pi^2|v|^2 for vv in the dual lattice Λ\Lambda^*.


Weyl's Law and the Heat Kernel

Theorem10.5Weyl's asymptotic law

Let Nk(λ)=#{j:λjλ}N_k(\lambda) = \#\{j : \lambda_j \leq \lambda\} be the eigenvalue counting function for the Laplacian on functions on a compact nn-manifold. Then

N0(λ)ωnVol(M)(2π)nλn/2as λ,N_0(\lambda) \sim \frac{\omega_n \mathrm{Vol}(M)}{(2\pi)^n} \lambda^{n/2} \quad \text{as } \lambda \to \infty,

where ωn\omega_n is the volume of the unit ball in Rn\mathbb{R}^n. This determines dim(M)\dim(M) and Vol(M)\mathrm{Vol}(M) from the spectrum.

Definition10.6Heat kernel

The heat kernel K(t,x,y)K(t, x, y) is the fundamental solution of the heat equation (t+Δx)K=0(\partial_t + \Delta_x)K = 0 with K(0,x,y)=δy(x)K(0, x, y) = \delta_y(x). It has the spectral expansion

K(t,x,y)=jeλjtϕj(x)ϕj(y),K(t, x, y) = \sum_j e^{-\lambda_j t} \phi_j(x) \phi_j(y),

where {ϕj}\{\phi_j\} are orthonormal eigenfunctions. The heat trace is Z(t)=MK(t,x,x)dvol=jeλjtZ(t) = \int_M K(t, x, x) \, \mathrm{dvol} = \sum_j e^{-\lambda_j t}.

RemarkCan you hear the shape of a drum?

Milnor (1964) showed that two 16-dimensional flat tori can have the same Laplacian spectrum but not be isometric, answering Kac's question negatively. However, the spectrum determines many geometric invariants: dimension, volume, total scalar curvature (via the heat trace asymptotics Z(t)(4πt)n/2k=0aktkZ(t) \sim (4\pi t)^{-n/2}\sum_{k=0}^\infty a_k t^k, where a0=Vol(M)a_0 = \mathrm{Vol}(M) and a1=16MSdvola_1 = \frac{1}{6}\int_M S \, \mathrm{dvol}).

RemarkThe Atiyah-Singer index theorem

The index of an elliptic operator DD (the difference dimkerDdimkerD\dim\ker D - \dim\ker D^*) can be computed from topological data via the heat kernel: ind(D)=tr(etDD)tr(etDD)\mathrm{ind}(D) = \mathrm{tr}(e^{-tD^*D}) - \mathrm{tr}(e^{-tDD^*}) for any t>0t > 0. This underlies the Atiyah-Singer index theorem, one of the deepest results connecting analysis, geometry, and topology.