ScalingStacks

Hamiltonian variations [04GF]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

Hamiltonian variations

If the Lagrangian angle of the minimizer satisfies −π/2+ϵ<infLθ≤supLθ<π/2−ϵ-\pi/2+\epsilon<\inf_{L}\theta\leq\sup_{L}\theta<\pi/2-\epsilon, then we have a more elliptic argument. Given any compactly supported global C∞C^{\infty} Hamiltonian function HH on XX, we can associate a 1-parameter family of symplectomorphisms ϕt\phi_{t} by exponentiating the Hamiltonian vector field. Since d​ϕtd\phi_{t} only moves the tangent planes by O⁡(|t|)O(|t|) for small |t|≪1|t|\ll 1, the Lagrangian angle of ϕt​(L)\phi_{t}(L) is still within (−π/2+ϵ,π/2−ϵ)(-\pi/2+\epsilon,\pi/2-\epsilon), namely the quantitatively almost calibrated condition is preserved.

Under global Hamiltonian deformations, the first variation of the Solomon functional is

δ​S​(H)=dd​t​𝒮​(ϕt​(L))|t=0=∫LH​Im​(e−i​θ^​Ω).\delta S(H)=\frac{d}{dt}\mathcal{S}(\phi_{t}(L))|_{t=0}=\int_{L}H\text{Im}(e^{-i\hat{\theta}}\Omega).

We need another ingredient which is expected to hold once the Floer theory is sufficiently developed in the weak regularity setting:

  • •

    The class of unobstructed exact Lagrangian objects is preserved by Hamiltonian isotopies. As such ϕt​(L)\phi_{t}(L) should remain inside the class ℒ\mathcal{L}.

These would imply that the minimizer LL satisfies

∫LH​Im​(e−i​θ^​Ω)=0.\int_{L}H\text{Im}(e^{-i\hat{\theta}}\Omega)=0.

for any compactly supported C∞C^{\infty} function on XX. This means Im​(e−i​θ^​Ω)=0\text{Im}(e^{-i\hat{\theta}}\Omega)=0 as currents, which is equivalent to θ=θ^\theta=\hat{\theta} under the almost calibrated setting.

Remark 5.29.

The assumption that −π/2+ϵ<infLθ≤supLθ<π/2−ϵ-\pi/2+\epsilon<\inf_{L}\theta\leq\sup_{L}\theta<\pi/2-\epsilon for the minimizer is not innocent, but represents a principal gap in our program to find special Lagrangian currents. The problem is that if on the minimizer supLθL=π2−ϵ\sup_{L}\theta_{L}=\frac{\pi}{2}-\epsilon, and a priori LL has no regularity assumption (eg. the Lagrangian angle may a priori be highly oscillatory), then we lack techniques to construct Lagrangian competitors which remain quantitatively almost calibrated.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.