Flags of closed submanifolds and their ideal sheaves #
Ideal-sheaf bookkeeping along a flag S ⊂ Y ⊂ M of closed submanifolds, as in Kollár's reduction
of the functoriality for closed embeddings to the case of a hypersurface [Kol07, 108]: locally
there is a chain of smooth subvarieties Y = Y_0 ⊂ Y_1 ⊂ ⋯ ⊂ Y_c = X such that "each is a
hypersurface in the next one".
IsClosedSubmanifold.idealSheaf_le_of_subset: a closed submanifold contained in another has the larger reduced ideal sheaf (both have the vanishing stalks, which are antitone in the set);IsClosedSubmanifold.idealSheaf_preimage_of_isLocalDiffeomorph: the ideal sheaf of the preimage of a closed submanifold under a local analytic isomorphism is the pull-back of its ideal sheaf (comap_idealSheaf_of_isLocalDiffeomorphread backwards; the pull-back of a blow-up centre along a smooth morphism, [Kol07, 30.1]);IsAdaptedChart.exists_flag: from one adapted chart ofSof codimensions + 1whose source is the whole manifold, the zero setYof the firstsadapted coordinates is a closed submanifold of codimensionscontainingS, andSis a hypersurface of the bundledY(IsClosedSubmanifold.preimage_val_of_subsetat codimension(s + 1) − s);IsClosedSubmanifold.pullback_idealSheaf_inclusionMap: the pull-back to the bundledYof the reduced ideal sheaf ofS ⊆ Yis the reduced ideal sheaf ofSinY, for any closed-submanifold structure onSinY(the hypothesisS ⊆ Yis needed: a hypersurface tangent toYis a counterexample);IdealSheaf.isNonzeroEverywhere_pullback_inclusionMap_of_le: an ideal sheafI ⊇ 𝓘_Swhose pull-back toSis nonzero everywhere pulls back toY ⊇ Snonzero everywhere — offSits stalk is the unit ideal; at a point ofS, a zero pull-back toYmeans every germ ofIlies in the kernel of the germ map ofY, the stalk of𝓘_Y ≤ 𝓘_S, so the pull-back toSis zero too (the hypothesis𝓘_S ≤ Iis needed: anIequal to𝓘_Yalong a component ofYmissingSis a counterexample).
Everything is assembled from the reduced ideal sheaf of a closed submanifold (vanishing stalks,
stalkIdeal_idealSheaf_eq_vanishingStalk), the germ map of the inclusion (onto, with kernel the
stalk of the reduced ideal, germMap_inclusionMap) and the relative closed-submanifold structure
(IsClosedSubmanifold.preimage_val_of_subset, idealSheaf_preimageVal_eq_pullback). These lemmas
are used for the commutation of the resolution with closed embeddings in general codimension
(Hironaka/Resolution/Analytic/Functor/ChainClosedEmbeddingChain.lean).
A closed submanifold contained in another has the larger reduced ideal sheaf (𝓘_Y ⊆ 𝓘_S
along the flag of [Kol07, 108]): both have the vanishing stalks, which are antitone in the set.
The ideal sheaf of the preimage of a closed submanifold under a local analytic isomorphism is
the pull-back of its ideal sheaf ([Kol07, 30.1]): comap_idealSheaf_of_isLocalDiffeomorph read
backwards (comap is the pull-back by definition).
The flag S ⊂ Y ⊂ M from one adapted chart ([Kol07, 108], read from the bottom): from an
adapted chart of S (codimension s + 1) whose source is the whole manifold, the zero set Y of
the first s adapted coordinates is a closed submanifold of codimension s containing S, and
S is a hypersurface of the bundled Y (IsClosedSubmanifold.preimage_val_of_subset at
codimension (s + 1) - s).
The pull-back to the bundled Y of the reduced ideal sheaf of S ⊆ Y is the reduced ideal
sheaf of S as a closed submanifold of Y, for any closed-submanifold structure on S in Y
(idealSheaf_preimageVal_eq_pullback; the two structures on the same set have the same reduced
ideal sheaf, both with the vanishing stalks). The hypothesis S ⊆ Y is needed: a hypersurface
tangent to Y is a counterexample.
The kernel of the germ map of the inclusion of a closed submanifold at p is the stalk of its
reduced ideal sheaf at p (germMap_inclusionMap and stalkIdeal_idealSheaf_of_mem, at a
general model).
An ideal sheaf containing 𝓘_S whose pull-back to S is nonzero everywhere pulls back to
Y ⊇ S nonzero everywhere: off S its stalk is the unit ideal; at a point of S, a zero
pull-back to Y means every germ of I lies in the kernel of the germ map of Y, the stalk of
𝓘_Y ≤ 𝓘_S, so the pull-back to S is zero too. The hypothesis 𝓘_S ≤ I is needed: an I
equal to 𝓘_Y along a component of Y missing S is a counterexample.