Mass--Hom bounds on bounded coherent derived categories #
This file only identifies the geometric test class. It does not manufacture Hom-finiteness: a consumer must supply the appropriate geometric finiteness input before forming a mass--Hom bound.
Perfect objects viewed inside Dᵇ(Coh X). The property is the inverse
image of schemePerfect X along the fully faithful bounded inclusion.
Equations
Instances For
The mass--Hom predicate specialized to perfect test objects in
Dᵇ(Coh X). Hom-finiteness and k-linearity stay explicit inputs.
Equations
Instances For
Remark 7.2: stable-target bounds on a classically generating collection
give the perfect mass--Hom bound. Classical generation is kept as the
explicit equality G.triangEnvelope = boundedSchemePerfect X; this abstract
adapter does not manufacture a geometric generating theorem.
Finite pullback transfer #
The finite-morphism pullback half of Lemma 7.4, with its two unfinished geometric inputs explicit.
For a finite morphism, boundedCoherentDerivedPushforward f is the actual
exact direct image used to construct f^♯σ. Any linear left adjoint pull
transfers the mass--Hom bound provided its perfect images classically generate
the source perfect test class.
The repository does not yet construct the required bounded coherent derived
pullback for an arbitrary finite morphism: perfectDerivedPullback is defined
only under the exact coherent-pullback contract and has domain Perf, whereas
the adjunction here is against all of Dᵇ(Coh). Accordingly pull, adj, and
the generation containment are parameters rather than manufactured instances.
The future geometric discharges are tracked by #1033 and #723.
Flat pushforward transfer #
The transfer core of the faithfully-flat half of Lemma 7.4, with the lower shriek and its generation input explicit.
Flatness is the hypothesis this statement uses: it supplies the exact bounded
coherent pullback f^* that detects f_♯σ, through isExactPullbackOfFlat
and hasCoherentPullbackOfIsExactPullback. Given a linear left adjoint
lowerShriek ⊣ f^*, the abstract preimage theorem transfers a perfect
mass--Hom bound once the lowerShriek images of perfect tests generate the
target perfect class.
Why surjectivity is not a hypothesis here. Lemma 7.4(2) assumes f
faithfully flat with relative dualizing complex ω_f^• in D_perf, and those
two extra assumptions do exactly two jobs. The dualizing complex builds
f_! = f_*(- ⊗ ω_f^•) together with its adjunction against f^*; surjectivity
makes f^* conservative on Dqc, which is what the Neeman--Ravenel criterion
turns into the statement that f_! carries a classical generator of Perf T
to one of Perf U. Both jobs are discharged into explicit parameters here --
lowerShriek and adj for the first, hgen for the second -- because the
repository owns neither construction. A Surjective f.left binder would
therefore be inert: it is the hypothesis of results this statement assumes
rather than proves, and carrying it would claim a geometric input the proof
never consumes. Those two constructions are the precise remaining #1033 and
#723 interfaces; discharging them is what earns back the faithfully flat
hypothesis, and the name, on a downstream theorem.