Exactness of the tensor product on ind-objects, (co)equalizer half #
Deligne's 2.2, finite-(co)limit half: for a small rigid abelian
monoidal C with preadditive tensor, the transported tensor product
of Ind C preserves finite colimits and finite limits in each
variable. This closes the gap documented in
RS.Classical.Deligne.IndTensorExact, whose additivity results
cover the finite-(co)product halves; what remains, and is proved
here, is preservation of coequalizers and of equalizers.
The route is a two-stage descent from the exactness of the tensor of
C itself (RS.Classical.Deligne.TensorExact):
- Embedded stage (
RS.preservesColimit_parallelPair_tensorLeft_indOfand its limit and right-hand twins): a parallel pair ofInd Cis presented as a filtered colimit of parallel pairs ofC(Mathlib'sIndParallelPairPresentation), i.e. the pair is isomorphic toparallelPair φ ψ ⋙ Ind.lim I. Tensoring by an embedded object commutes withInd.limup to the natural isomorphismRS.indLimCompTensorLeftIso, built fromRS.indOfTensorIsoand its naturality; on the functor category side the whiskeredtensorLeft apreserves (co)equalizers pointwise becausetensorLeft apreserves all (co)limits in the rigidC, andInd.limpreserves finite limits and colimits. Preservation transports along the presentation. - Descent stage (
RS.isIso_post_parallelPair_tensorLeft,RS.isIso_limitPost_parallelPair_tensorLeftand twins): for a fixed parallel pair, the (co)limit comparison oftensorLeft Ais natural in the tensoring objectA(RS.whiskerLeftCoeqComparison/RS.whiskerLeftEqComparison), between endofunctor-composites that preserve filtered colimits. The filtered-descent lemmaRS.isIso_app_of_isIso_indOfofIndTensorExactthen propagates invertibility from the embedded stage to everyA : Ind C. For coequalizers the required filtered preservation ofcolim-composites is formal; for equalizers it is the exactness of filtered colimits inInd C, derived here asRS.preservesColimitsOfShape_parallelPairLim_indfrom the type-level commutation of filtered colimits with finite limits through the interchange transposeRS.preservesColimitsOfShape_limand the inclusion into presheaves.
Deliverables, for every A : Ind C:
RS.tensorLeft_ind_preservesCoequalizers/RS.tensorRight_ind_preservesCoequalizersandRS.tensorLeft_ind_preservesEqualizers/RS.tensorRight_ind_preservesEqualizers— preservation ofWalkingParallelPaircolimits and limits for both tensoring functors;RS.tensorLeft_ind_preservesFiniteColimits/RS.tensorRight_ind_preservesFiniteColimits— the tensor ofInd Cis right exact in each variable;RS.tensorLeft_ind_preservesFiniteLimits/RS.tensorRight_ind_preservesFiniteLimits— and left exact, combining with the additivity ofIndTensorExact.
The acceptance tests confirm that the cokernel and kernel comparison isomorphisms of both tensoring functors synthesize.
Preservation of the colimit of a composed diagram K ⋙ F by
T, given that F preserves the colimit of K and the composite
functor F ⋙ T does: the colimit cocone of K ⋙ F may be taken to
be the image under F of the colimit cocone of K.
Dual of RS.preservesColimit_comp_diagram.
Transpose of the interchange property, flipped-diagram form: if
K-indexed colimits preserve J-limits, the J-limit functor
preserves the colimit of G.flip for any bifunctor G. The
colimit comparison is identified with the canonical interchange
isomorphism Limits.colimitLimitIso.
Transpose of the interchange property: if K-indexed colimits
preserve J-limits, then the J-limit functor preserves
K-colimits.
The embedding-tensor comparison as a natural isomorphism, left
version: composing the embedding with left tensoring by an embedded
object is left tensoring in C followed by the embedding.
Equations
- RS.indOfCompTensorLeftIso a = CategoryTheory.NatIso.ofComponents (fun (x : C) => RS.indOfTensorIso a x) ⋯
Instances For
The embedding-tensor comparison as a natural isomorphism, right version.
Equations
- RS.indOfCompTensorRightIso a = CategoryTheory.NatIso.ofComponents (fun (x : C) => RS.indOfTensorIso x a) ⋯
Instances For
Left tensoring by an embedded object commutes with Ind.lim:
the composite Ind.lim I ⋙ tensorLeft (indOf.obj a) is the
whiskering of tensorLeft a followed by Ind.lim I. The chain
mirrors Mathlib's Ind.limCompInclusion, with the embedded-tensor
isomorphism in place of Ind.yonedaCompInclusion.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Right tensoring by an embedded object commutes with
Ind.lim.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Embedded stage, left version: left tensoring by an embedded
object preserves the colimit of every parallel pair of Ind C. The
pair is presented as a filtered colimit of parallel pairs of C;
tensoring commutes with the presentation by
RS.indLimCompTensorLeftIso, and on the functor-category side the
whiskered tensorLeft a preserves coequalizers pointwise since the
tensor of the rigid C is exact.
Embedded stage for limits, left version: left tensoring by an
embedded object preserves the limit of every parallel pair of
Ind C — the dual of the coequalizer argument, using that
Ind.lim preserves finite limits and that the tensor of the rigid
C is left exact.
Embedded stage for limits, right version.
Embedded stage, right version.
The parallel pair (A ◁ f, A ◁ g) as a functor of the tensoring
object A. The @[reducible] marking is deliberate: the object
field must reduce at instance transparency for the show-retyped
colimit proofs below to be stateable.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The parallel pair (f ▷ A, g ▷ A) as a functor of the tensoring
object A.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Evaluating RS.whiskerLeftPairFunctor at a stage of the walking
parallel pair gives right tensoring by the corresponding object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Evaluating RS.whiskerRightPairFunctor at a stage gives left
tensoring by the corresponding object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The pair functor preserves filtered colimits, pointwise.
The right-hand pair functor preserves filtered colimits.
The coequalizer comparison for left tensoring, naturally in the
tensoring object: at A it is the canonical morphism from the
colimit of the pair (A ◁ f, A ◁ g) to A ⊗ coeq (f, g).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The coequalizer comparison for right tensoring, naturally in the tensoring object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Descent stage, left version: the colimit comparison of any parallel pair under left tensoring is invertible, for every tensoring ind-object. Filtered descent from the embedded stage.
Descent stage, right version.
Left tensoring on Ind C preserves coequalizers.
Right tensoring on Ind C preserves coequalizers.
Filtered colimits commute with WalkingParallelPair-limits in
Ind C: the parallel-pair limit functor preserves filtered
colimits. The property holds for types, lifts to presheaves
pointwise, transposes through RS.preservesColimitsOfShape_lim,
and transports to Ind C along the inclusion, which creates the
limits and preserves the filtered colimits involved.
The equalizer comparison for left tensoring, naturally in the
tensoring object: at A it is the canonical morphism from
A ⊗ lim (f, g) to the limit of the pair (A ◁ f, A ◁ g).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The equalizer comparison for right tensoring, naturally in the tensoring object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Descent stage for limits, left version: the limit comparison of any parallel pair under left tensoring is invertible, for every tensoring ind-object.
Descent stage for limits, right version.
Left tensoring on Ind C preserves equalizers.
Right tensoring on Ind C preserves equalizers.
Deligne 2.2, right-exactness, left version: left tensoring
on Ind C preserves finite colimits — coequalizers by the descent
above, finite coproducts by the additivity of
RS.Classical.Deligne.IndTensorExact.
Deligne 2.2, right-exactness, right version.
Left-exactness, left version: left tensoring on Ind C
preserves finite limits — equalizers by the dual descent, finite
products by additivity. With the right-exactness above, the tensor
of Ind C is exact in each variable.
Left-exactness, right version.