ScalingStacks

2. The Monge-Ampère operator [01D2]

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

2. The Monge-Ampère operator

In the complex case, the Monge-Ampère operator is a second order differential operator: we set MA⁡(ϕ)=(d​dc​ϕ)n\operatorname{MA}(\phi)=(dd^{c}\phi)^{n} for a smooth metric ϕ\phi. It is a nonlinear operator if n>1n>1. When ϕ\phi is semipositive, MA⁡(ϕ)\operatorname{MA}(\phi) is a smooth positive measure on Xan{X^{\mathrm{an}}} of mass (Ln)(L^{n}). It is a volume form, that is, equivalent to Lebesgue measure, if ϕ\phi is positive.

Next we turn to the non-Archimedean setting. From now on we assume that KK is discretely valued. Pick a uniformizer tt of the maximal ideal in the valuation ring RR of KK.

Consider a model metric ϕℒ\phi_{\mathcal{L}}, associated to a model (𝒳,ℒ)({\mathcal{X}},{\mathcal{L}}) of (X,L)(X,L) over Spec⁡R\operatorname{Spec}R. Write the special fiber as 𝒳0=div⁡t=∑i∈Ibi​Ei{\mathcal{X}}_{0}=\operatorname{div}t=\sum_{i\in I}b_{i}E_{i}, where EiE_{i} are the irreducible components of 𝒳0{\mathcal{X}}_{0} and bi∈𝐙>0b_{i}\in{\mathbf{Z}}_{>0}. To each EiE_{i} is associated a unique (divisorial) point xi∈Xanx_{i}\in{X^{\mathrm{an}}}. We then define

MA⁡(ϕ):=∑i∈Ibi​(ℒ|Ei)n​δxi.\operatorname{MA}(\phi):=\sum_{i\in I}b_{i}({\mathcal{L}}|_{E_{i}})^{n}\delta_{x_{i}}.

If ϕℒ\phi_{\mathcal{L}} is semipositive, ℒ|Ei{\mathcal{L}}|_{E_{i}} is nef; hence (ℒ|Ei)n≥0({\mathcal{L}}|_{E_{i}})^{n}\geq 0 and MA⁡(ϕℒ)\operatorname{MA}(\phi_{\mathcal{L}}) is a positive measure. Its total mass is

∫Xan1⋅MA⁡(ϕℒ)=∑i∈Ibi​(ℒ|Ei)n=(ℒn⋅𝒳0)=(ℒn⋅𝒳η)=(Ln).\int_{{X^{\mathrm{an}}}}1\cdot\operatorname{MA}(\phi_{\mathcal{L}})=\sum_{i\in I}b_{i}({\mathcal{L}}|_{E_{i}})^{n}=({\mathcal{L}}^{n}\cdot{\mathcal{X}}_{0})=({\mathcal{L}}^{n}\cdot{\mathcal{X}}_{\eta})=(L^{n}).

Here the second to last equality follows from the flatness of 𝒳{\mathcal{X}} over Spec⁡R\operatorname{Spec}R, and the last equality from 𝒳η≃X{\mathcal{X}}_{\eta}\simeq X, ℒη≃L{\mathcal{L}}_{\eta}\simeq L.

From now on assume that LL is ample, that is, we have a polarized pair (X,L)(X,L). In both the complex and non-Archimedean case we define MA⁡(ϕ)\operatorname{MA}(\phi) for a continuous semipositive metric by MA⁡(ϕ):=limm→∞MA⁡(ϕm)\operatorname{MA}(\phi):=\lim_{m\to\infty}\operatorname{MA}(\phi_{m}) for any sequence (ϕm)1∞(\phi_{m})_{1}^{\infty} converging uniformly to ϕ\phi. Of course, it is not obvious that the limit exists or independent of the sequence (ϕm)1∞(\phi_{m})_{1}^{\infty}. In the complex case this is a very special case of the Bedford-Taylor theory developed in [BT82, BT87]. The analogous analysis in the non-Archimedean case is due to Chambert-Loir [CL06].

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