Lemma 4.12. On overlapping charts of ,
Verified tagged author-source HTML · 1912.02360v1 · cited publication edition alignment unverified.
4.4 Locally convex function
In section 4.2 we produced a collection of local average functions on corresponding to various choices of and with . But the local coordinates are naturally interpreted also as coordinates on (cf. section 3.2), so can be alternatively viewed as a collection of convex functions on the charts of . (Notice these local functions are defined without the need to shrink the domain to ).
The intuition is that up to -small error, the differences of these local functions agree with the cocycle , or equivalently, up to some -small fuzziness glue to a locally convex function on in the sense of Definition 3.22. The more precise statement is
Proof. Since we know the local -oscillation estimate holds in every local region, in a log scale in , not necessarily in the shrinked region ,
Since is convex, a local -bound implies a local -bound in a slightly shrinked region, so in the log scale,
Likewise for . By definition the local potentials differ by
Notice that for a given point on , the log scales on and around have a nontrivial percentage of overlapping measure. Thus
∎
Remark 4.13. The tropical version of is in general larger than ; it typically contains also some subset stretching to infinity along the -direction. If we regard as local functions on instead of , then there is a delicate issue. The Lemma above does not imply that for various choices of glue approximately on overlapping regions far from . The problem is that such overlapping regions have too small measure, which breaks down the proof.