The geometric Step 3 does not depend on the enumeration of the components #
The input family ofDivisorFamily E a e σ of
Hironaka/Resolution/Algebraic/Monomial/Geometric/Pieces.lean enumerates the irreducible components
of the members of E by σ : Fin k ≃ Components E. Kollár's Step 3 ([Kol07, 111, Step 3]) does not
depend on that enumeration: its choices are made on faces compared by their label tuples and sums,
never by the numbering of the pieces. Two enumerations of the same components give the same input
family up to a permutation of the pieces, and the permutation is a refinement along the identity in
the sense of Hironaka/Resolution/Algebraic/Monomial/Geometric/Refinement.lean
(refinesAlong_id_ofDivisorFamily): the same label count, labels and exponents through the
relabelling σ₁⁻¹ ∘ σ₂, every piece inside (indeed equal to) its parent, every parent the union of
its unique child, and no two children of one parent. The functoriality of the geometric Step 3 under
smooth surjections (realize_pullback_of_surjective of
Hironaka/Resolution/Algebraic/Monomial/Geometric/PullbackRun.lean) at h = 𝟙 X then identifies
the two runs (realize_ofDivisorFamily_indep; the pull-back along the identity is the sequence
itself). Also here: two exponent functions agreeing at the generic points of the members give the
same input family (ofDivisorFamily_congr: ofDivisorFamily reads its exponents there only). Not
in the sources. Used in Hironaka/Resolution/Algebraic/MarkedOrderReduction/Step3Functorial.lean to
show that the third step of BMO_{n,m} is well defined.
Two exponent functions agreeing at the generic points of the members give the same input family.
Two exponent functions agreeing at the generic points of the members give the same input
family: ofDivisorFamily reads its exponents there only. The structure morphism f is present
because the statements about the input family of a triple over k are all made over its structure
morphism; the proof does not need it, the equality being one of definitions on any scheme
(ofDivisorFamily_congr_of_forall).
The relabelling of the pieces between two enumerations of the components: the piece c of the
second enumeration is the piece σ₁⁻¹ (σ₂ c) of the first (and 0 off the range).
Equations
Instances For
The piece of the relabelled index is the piece itself.
The input family of one enumeration of the components refines the input family of another
along 𝟙 X through the relabelling σ₁⁻¹ ∘ σ₂: the same label count, labels and exponents
through the relabelling, each piece equal to its parent, each parent the union of its unique
child, and no two children of one parent.
The geometric Step 3 does not depend on the enumeration of the components: for two
enumerations of the components of E (the same label isomorphism) the realised runs coincide,
by realize_pullback_of_surjective at h = 𝟙 X with the identity refinement
refinesAlong_id_ofDivisorFamily and pullback_id.