The dimension of a triple #
Kollár's functors BO_m, BMO_m of [Kol07, Theorems 68 and 69] are assembled from the stages of
the induction of [Kol07, 70], one stage per dimension bound. A triple is equidimensional by
[Kol07, Notation 64 (1)] — its structure morphism is smooth of ONE relative dimension
(Triple.smoothOfRelativeDimension : ∃ n, SmoothOfRelativeDimension n (X ↘ Spec k)) — so "dim X",
the maximum of the local dimensions, is that relative dimension: Triple.dim T chooses it. Every
triple then lies in the stage dim X (HasDimLE T T.dim), and tower base T.dim is the canonical
stage for X (the coherence of Hironaka.Resolution.Algebraic.Stage.Coherence makes every higher
stage agree with it). On a nonempty X the relative dimension is unique
(SmoothOfRelativeDimension.eq_of_nonempty), so HasDim n ↔ dim = n and HasDimLE n ↔ dim ≤ n
(HironakaExamples.Stage.BaseCases); on the empty scheme every n is a relative dimension and
dim is the chosen one. No cotangent-rank computation is needed: equidimensional triples carry
their dimension as data.
The dimension of a triple (Kollár's dim X = n of [Kol07, Theorem 103], for the
equidimensional X of [Kol07, Notation 64 (1)]): the relative dimension of its structure morphism,
chosen from Triple.smoothOfRelativeDimension; unique when X is nonempty
(SmoothOfRelativeDimension.eq_of_nonempty).
Equations
- T.dim = Classical.choose ⋯