Documentation

DerivedAlgGeo.AlgebraicGeometry.DerivedCategory.Families.BaseChangeBoundedAmplitude

Finite amplitude and the bounded-coherent restriction after base change #

BaseChangeDecomposition.lean leaves the bounded-coherent restriction of a base-change decomposition as the pair of obligations DecompositionData.BoundedData records: the chosen quasicoherent projectors preserve the intrinsic bounded-coherent locus, and the restricted components generate it. This file splits the first obligation along the decomposition of that locus and discharges the half that is formal.

The locus is schemeBoundedQuasicoherent ⊓ schemeFinitePresentationCohomology (Dqc/BoundedCoherent.lean). Preservation of the first conjunct follows from a cohomological amplitude bound on the projectors and from nothing else, by RightProjectionData.HasFiniteCohomologicalAmplitude.preserves_bounded. Preservation of the second does not follow from any amplitude bound: finite presentation of the cohomology sheaves of a projected object is a genuine coherence statement about the geometry, so PreservesCoherentCohomology stays an input, as does generation of the restricted locus.

Amplitude is measured against a t-structure on Dqc(X_T). The repository has no such t-structure -- boundedness in Dqc(X_T) is measured through the inclusion by the canonical t-structure of the ambient all-sheaf derived category -- so TargetBoundedTStructure carries one together with the only property of it the argument uses: that its bounded objects are the ambiently bounded ones. Supplying that structure is a geometric obligation, and it is deliberately not the base-changed t-structure of a source t-structure, which belongs to a separate lane.

Nothing here asserts finite amplitude for any projector, or constructs a t-structure on Dqc(X_T).

A t-structure on Dqc(X_T) that presents the ambient bounded locus.

bounded_iff is the whole interface: it says that the t-bounded objects of this t-structure are exactly those whose underlying complex is bounded for the canonical t-structure of the all-sheaf derived category, which is how Dqc.schemeBoundedQuasicoherent is defined. No compatibility with pullback, pushforward, or a source t-structure is asserted or used.

The pair is what keeps this non-vacuous. Amplitude against an arbitrary t-structure would say nothing about the locus BoundedData cares about, and a t-structure so coarse that every object is bounded would make amplitude cheap -- but then bounded_iff would assert that every object of Dqc(X_T) has bounded cohomology, which is false. What a consumer gains by choosing the t-structure is only the freedom to prove amplitude in whichever presentation is convenient.

Instances For

    The chosen quasicoherent projectors have finite cohomological amplitude for a presenting t-structure. The interval may depend on the component.

    Equations
    Instances For

      The chosen quasicoherent projectors preserve finite presentation of every cohomology sheaf.

      This is the coherence half of the bounded restriction and is left as an explicit input: it is a finiteness statement about projected objects, and no amplitude bound implies it.

      Equations
      Instances For

        Assemble the bounded-coherent restriction from finite amplitude, coherence preservation, and generation of the restricted locus.

        The bounded-coherent base-change sequence is strong as soon as the quasicoherent projectors have finite amplitude and preserve coherence.

        The bounded-coherent restriction of the Proposition 3.15 decomposition whose projectors were constructed from compact-generator approximations.

        Everything categorical has already been discharged: semiorthogonality, generation of the perfect and quasicoherent sequences, the projectors themselves, and now preservation of boundedness. What is still supplied here is exactly the geometry that no formal step produces -- a presenting t-structure, finite amplitude against it, coherence preservation, and generation of the bounded locus.