ProofComplete

Proof of the Hodge Decomposition via Elliptic Theory

The Hodge decomposition rests on the theory of elliptic operators on compact manifolds. We present the key analytical ingredients and show how they combine to yield the decomposition.


Elliptic Operators on Compact Manifolds

Definition10.7Elliptic operator

A linear differential operator P:Γ(E)Γ(F)P: \Gamma(E) \to \Gamma(F) of order mm between vector bundles is elliptic if its principal symbol σm(P)(x,ξ):ExFx\sigma_m(P)(x, \xi): E_x \to F_x is an isomorphism for all xMx \in M and ξTxM{0}\xi \in T^*_xM \setminus \{0\}.

The Hodge Laplacian Δ=dδ+δd\Delta = d\delta + \delta d has principal symbol σ2(Δ)(x,ξ)=ξ2Id\sigma_2(\Delta)(x, \xi) = |\xi|^2 \mathrm{Id}, which is positive definite for ξ0\xi \neq 0. Hence Δ\Delta is elliptic.


Key Analytical Results

Theorem10.9Elliptic regularity

If PP is an elliptic operator of order mm on a compact manifold and Pu=fPu = f with fHsf \in H^s (Sobolev space), then uHs+mu \in H^{s+m}. In particular, if fCf \in C^\infty, then uCu \in C^\infty. Applied to Δ\Delta: any distributional solution of Δω=f\Delta\omega = f with ff smooth is itself smooth.

Theorem10.10Fredholm property

An elliptic operator P:Hs+m(E)Hs(F)P: H^{s+m}(E) \to H^s(F) on a compact manifold is Fredholm: it has finite-dimensional kernel, closed range, and finite-dimensional cokernel. The index ind(P)=dimkerPdimcokerP\mathrm{ind}(P) = \dim\ker P - \dim\mathrm{coker}\, P is a topological invariant.


The Hodge Decomposition Proof

Proof

We prove the decomposition Ωk=HkdΩk1δΩk+1\Omega^k = \mathcal{H}^k \oplus d\Omega^{k-1} \oplus \delta\Omega^{k+1} in detail.

Step 1: The orthogonal complement of kerΔ\ker\Delta. Since Δ\Delta is self-adjoint and elliptic on a compact manifold:

  • kerΔ\ker\Delta is finite-dimensional (by the Fredholm property).
  • L2Ωk=kerΔim(Δ)L^2\Omega^k = \ker\Delta \oplus \overline{\mathrm{im}(\Delta)} (since Δ\Delta is self-adjoint, (kerΔ)=im(Δ)(\ker\Delta)^\perp = \overline{\mathrm{im}(\Delta)}).
  • The image is closed (Fredholm property), so im(Δ)=im(Δ)\overline{\mathrm{im}(\Delta)} = \mathrm{im}(\Delta).

Thus L2Ωk=kerΔim(Δ)L^2\Omega^k = \ker\Delta \oplus \mathrm{im}(\Delta), and by elliptic regularity (smooth inputs give smooth outputs), this holds at the smooth level: Ωk=HkΔΩk\Omega^k = \mathcal{H}^k \oplus \Delta\Omega^k.

Step 2: Splitting im(Δ)\mathrm{im}(\Delta). We claim ΔΩk=dΩk1δΩk+1\Delta\Omega^k = d\Omega^{k-1} \oplus \delta\Omega^{k+1} (orthogonal).

Orthogonality: dα,δβ=d2α,β=0\langle d\alpha, \delta\beta \rangle = \langle d^2\alpha, \beta \rangle = 0 since d2=0d^2 = 0.

Inclusion \supset: dα=d(δd+dδ)Gαd\alpha = d(\delta d + d\delta)G\alpha' (using the Green's operator), and similarly for δ\delta. In fact, for ωHk\omega \perp \mathcal{H}^k, we have ω=ΔGω=dδGω+δdGω\omega = \Delta G\omega = d\delta G\omega + \delta d G\omega, where dδGωdΩk1d\delta G\omega \in d\Omega^{k-1} and δdGωδΩk+1\delta d G\omega \in \delta\Omega^{k+1}.

Inclusion \subset: Δω=(dδ+δd)ω=d(δω)+δ(dω)dΩk1+δΩk+1\Delta\omega = (d\delta + \delta d)\omega = d(\delta\omega) + \delta(d\omega) \in d\Omega^{k-1} + \delta\Omega^{k+1}.

Step 3: Assembling the decomposition. From Steps 1 and 2:

Ωk=HkdΩk1δΩk+1,\Omega^k = \mathcal{H}^k \oplus d\Omega^{k-1} \oplus \delta\Omega^{k+1},

with pairwise orthogonality. Every ωΩk\omega \in \Omega^k decomposes uniquely as ω=Hω+dα+δβ\omega = H\omega + d\alpha + \delta\beta, where HH is the harmonic projection, α=δGω\alpha = \delta G\omega, and β=dGω\beta = dG\omega. \blacksquare


Remarks

RemarkThe Green's operator

The Green's operator G:ΩkΩkG: \Omega^k \to \Omega^k is defined by GHk=0G|_{\mathcal{H}^k} = 0 and G=Δ1G = \Delta^{-1} on (Hk)(\mathcal{H}^k)^\perp. It satisfies ΔG+H=Id=GΔ+H\Delta G + H = \mathrm{Id} = G\Delta + H, where HH is the harmonic projection. The Green's operator is a compact, self-adjoint operator on L2ΩkL^2\Omega^k that smoothes by two derivatives.

RemarkManifolds with boundary

For manifolds with boundary, the Hodge decomposition requires boundary conditions. The two natural choices are absolute boundary conditions (ινωM=0\iota_\nu \omega|_{\partial M} = 0 and ινdωM=0\iota_\nu d\omega|_{\partial M} = 0) and relative boundary conditions (ωM=0\omega|_{\partial M} = 0 and δωM=0\delta\omega|_{\partial M} = 0). Each gives a Hodge decomposition, with the harmonic forms corresponding to absolute (resp. relative) cohomology.