Skip to content

Architecture Spherical Twist Roadmap

Chris Dare edited this page Nov 15, 2026 · 1 revision

Enhanced spherical twists: conventions and remaining seams

This note records the mathematical convention implemented by the dg-cone and Fourier--Mukai layers. It is a dependency and claim ledger, not an assertion that the repository already constructs a spherical twist for a geometric Enriques surface.

Primary references and convention

  • Seidel--Thomas, Braid group actions on derived categories of coherent sheaves, arXiv:math/0001043. For an object E, the twist is the functorial cone of the evaluation RHom(E,-) ⊗ E ⟶ id. In the smooth-projective Fourier--Mukai realization, its kernel is Cone(E^∨ ⊠ E ⟶ O_Δ) (Lemma 3.2).
  • Anno--Logvinenko, Spherical DG-functors, arXiv:1309.5035v2. For a functor S with left and right adjoints L and R, their convention uses the four functorial triangles SR ⟶ id_B ⟶ T, T' ⟶ id_B ⟶ SL, F ⟶ id_A ⟶ RS, and LS ⟶ id_A ⟶ F'. Full sphericality includes twist/cotwist equivalences and two maps identifying the adjoints. Their main theorem derives all four conditions from any two in a Morita-enhanced dg setting; it is not merely a theorem of triangulated categories.

The repository therefore calls Cone(SR ⟶ id_B) a twist candidate, not a spherical twist, until the remaining equivalence and comparison conditions have been discharged.

Implemented dependency tree

DGFunctor.HomogeneousNatTrans
├─ all integer degrees and Koszul naturality
├─ pointwise differential and vertical composition
├─ h0Comparison: H⁰(DGFunctor C D) ⥤ (H⁰ C ⥤ H⁰ D), no fullness or faithfulness
│  └─ h0Iso: a Z⁰ dg-functor isomorphism descends canonically, with unit/composition laws
├─ whiskerLeft / whiskerRight (both sign-free) and additive-hom packaging
│  ├─ interchange: the Godement sign (-1)^(m n) is naturality of the second
│  └─ hcomp: the Godement product, graded Leibniz, strictly associative
└─ DGCategory (DGFunctor A B)
   └─ IsClosed

IsConeOf
└─ homogeneousLift (one owner of the cone lift)
   ├─ arbitrary-degree squares with homotopy: differential law
   ├─ additivity, identity, composition laws for strict squares
   ├─ projection and inclusion formulas
   ├─ lift := homogeneousLift 0, HomotopySquare := degree-zero square + closedness
   │  ├─ HomotopySquare.strict: a commuting square with zero homotopy
   │  └─ IsConeOf.Morphism with the shift-free `fst` square
   │     └─ ConePresentation category, no pretriangulated instance needed
   ├─ isoOfStrictSquare: endpoint isomorphisms in a strict square induce a
   │  canonical Z⁰ isomorphism of arbitrary chosen cones
   └─ HomogeneousNatTrans.ConeData
         ├─ objectwise cone choices
         ├─ assembled cone DGFunctor
         ├─ canonical inr/inl transformations and boundary equation
         ├─ fst / snd: the cone projections, graded-natural in degrees 1 and 0
         ├─ isConeOf: a cone in the dg category DGFunctor C D itself, so the
         │  twist candidate is a cone OF FUNCTORS, the form in which
         │  Anno--Logvinenko state SR ⟶ Id_B ⟶ T
         ├─ isoOfStrictSquare: the generic cone isomorphism in Z⁰(DGFunctor C D)
         ├─ triangleFunctor: H⁰ C ⥤ Triangle (H⁰ D), every value distinguished;
            the third square is the connecting map, carried by IsConeOf.Morphism
            ├─ triangleNatTrans: natural in a STRICT square of closed
            │  degree-zero transformations; the cone lifts are natural on the
            │  nose.  The homotopy-coherent case needs uniqueness of the lift
            │  up to homotopy, which the repository does not have.
            ├─ triangleIsoOfStrictSquare: endpoint isomorphisms upgrade the
            │  comparison to a natural isomorphism of triangle functors
            └─ compareIso: two ConeData for one transformation give canonically
               isomorphic triangle functors, so the cone choices do not matter
         └─ inverse-rotation/source-regrading coherence
            ├─ inverseRotateFirstH0: the inverse-rotated first maps as a NatTrans
            ├─ shiftedFstH0: H⁰ of the regraded closed degree-one projection
            └─ shiftedFstH0_eq: equality through shiftedFunctorH0Iso and the
               packaged `[1][-1]` cancellation

DGAdjunction
├─ closed unit and counit
├─ dg triangle identities
├─ h0 : an ordinary Mathlib adjunction between the H⁰ functors
├─ CounitConeData
│  ├─ twist candidate Cone(LR ⟶ id)
│  └─ twistTriangleFunctor: H⁰ D ⥤ Triangle (H⁰ D), every value distinguished,
│     first two maps the counit and the canonical inclusion
├─ UnitConeData
│  └─ unshifted unit-cone candidate Cone(id ⟶ RL)
└─ a compatible left/right adjunction pair canonically supplies
   ├─ twistAdjointComparison: L T[-1] ⟶ R
   └─ cotwistAdjointComparison: R ⟶ C L
      ├─ cotwistAdjointComparisonShiftedH0: R ⟶ (C[-1] L)[1]
      └─ EnhancedAdjunctionComparison records IsIso on their H⁰ maps and
         exposes the conventional shifted-target cotwist isomorphism

The cone constructions and comparison maps do not say that either cone is an
autoequivalence, do not call the adjunction spherical, and do not derive the
comparison conditions from the equivalence conditions.

HasCopower K X / HasCopowers C
└─ copowerData: a noncomputably selected CopowerData witness

HasEvaluationData E
├─ chosenEvaluationData: a noncomputably selected EvaluationData witness
└─ supplied automatically by HasCopowers C

HasLinearEvaluationData k E
├─ chosenLinearEvaluationData: a noncomputably selected LinearEvaluationData witness
└─ supplied automatically by HasLinearCopowers k C

IsLinearCopowerOf k K X ZK
├─ homComplexIso: the representing equivalence in every degree
├─ coefficientMap: strict homogeneous action of coefficient cochains
└─ homotopyEquivIso: homotopy-equivalent coefficient complexes give
   isomorphic witnessed copowers in H⁰

CochainComplex.homologyModel (any abelian category)
├─ generic zero-differential model, with no formality assertion
└─ ModuleCat/division-ring homotopyEquivHomologyModel:
   every unbounded coefficient complex is noncanonically homotopy equivalent
   to its zero-differential homology model;
   no finite support, finite dimension, naturality, or quasi-isomorphism claim

HasLinearCopowers k C
├─ linearCopowerFunctor k X: C^dg(ModuleCat k) ⟶ C, a k-linear dg functor
├─ DGLinear.homFunctor k E: C ⟶ C^dg(ModuleCat k), a k-linear dg functor
├─ linearCopowerAdjunction k E: (- ⊗ E) ⊣ Hom(E,-)
│  ├─ unit: the selected universal copower chain map
│  ├─ counit: exactly the selected scalar-linear evaluation transformation
│  └─ counit cone = selected scalar-linear object-twist cone definitionally
└─ C = Cdg(ModuleCat k), same universe: Mathlib's total tensor product
   supplies the chosen object, tensor--Hom cochain equivalence, and instance

HomotopyCategory/DGEnhancement/LinearCopower
└─ H⁰ sends homotopy-equivalent coefficient complexes to isomorphic selected
   objects through the existing H⁰(C^dg) ≃ HomotopyCategory seam

LinearEvaluationData k E
├─ functor = Hom(E,-) ⊗ E, a k-linear dg functor
├─ evaluation : functor ⟶ id, closed in degree zero
├─ compareIso: canonical Z⁰ isomorphism between choices, strictly compatible
│  with evaluation
├─ IsEulerCopower: shared rank-one K₀ formula, obtained from supplied finite
│  cohomology presentations when all linear copowers exist
├─ TwistConeData: thin specialization of the generic evaluation cone
│  ├─ twistTriangleFunctor: H⁰ C ⥤ Triangle (H⁰ C), all values distinguished
│  ├─ coherent comparison across evaluation and cone choices
│  ├─ adjunctionTwistIso / adjunctionTwistTriangleIso:
│  │  canonical DG and full-triangle identifications with the copower--Hom counit twist
│  ├─ K₀ action = identity minus evaluation
│  └─ numerical twistK₀ via IsEulerCopower; dg exactness is automatic
└─ no additive EvaluationData adapter or autoequivalence; the standard
   module-complex dg category now has a concrete same-universe instance

EvaluationData E
├─ functor = RHom(E,-) ⊗ E, evaluation : functor ⟶ id, closed in degree zero
├─ compareIso: canonical Z⁰ isomorphism between any two choices, strictly
│  compatible with evaluation
└─ TwistConeData: the object twist T_E = Cone(evaluation)
   ├─ compareIso: canonical coherent Z⁰ isomorphism across both evaluation and
   │  cone choices, strictly compatible with id ⟶ T_E
   ├─ twistTriangleIsoOfEvaluation: coherent natural isomorphism of the full triangles
   │  across both evaluation and cone choices
   ├─ triangle functor H⁰ C ⥤ Triangle (H⁰ C), every value distinguished
   ├─ H⁰(T_E) triangulated automatically from the dg functor
   └─ no adjunction, no autoequivalence, no sphericality

EnhancedAdjunctionCones
├─ twist and dual-cotwist cones
├─ all four STORED cone functors are exact on H⁰ automatically; the retained
│  3-by-3 constructor computes the witness from its endpoints and is block
│  diagonal in the four coordinates of the two splittings.
├─ unshifted cones underlying dual twist and cotwist
├─ dualTwistFunctor and cotwistFunctor name their conventional `[-1]` shifts;
│  shiftedFunctorH0Iso compares their H⁰ functors with pointwise shift, and
│  the generic signed interface transports their exactness
├─ the four triangles as functors on H⁰, every value distinguished, each first
│  map the corresponding unit or counit.  The dual twist and the cotwist are
│  the INVERSE ROTATIONS of their unshifted cone triangles: invRotate applies
│  the ⟦-1⟧ shift and reorders in one step, and the first vertex is the value
│  of the dg shifted cone functor on the nose (DGFunctor.shiftedFunctor_h0_obj).
│  The unshifted forms are kept, since TwistCotwistEquivalenceConditions is
│  stated against the unshifted cone functors.  No sphericality, and no relation
│  among the four.
├─ K₀ action of each conventional functor is identity minus its adjunction
│  composite, first on object classes and then as a homomorphism under exactness
├─ canonical closed dg adjunction comparisons
│  ├─ `L T[-1] ⟶ R`, from the regraded twist-cone projection and left counit
│  └─ `R ⟶ C L`, from the left unit and cotwist-cone inclusion
│     └─ both descend to named H⁰ maps
├─ generic inverse-rotation/source-regrading coherence
│  ├─ `inverseRotateFirstH0`: first maps of inverse-rotated cone triangles
│  ├─ `shiftedFstH0`: H⁰ of the regraded closed degree-one projection
│  └─ equality through `shiftedFunctorH0Iso`, using packaged `[1][-1]`
│     cancellation and no new cone-specific sign choice
├─ AdjointComparisonConditions
│  └─ Mathlib IsIso on those canonical H⁰ maps, with asIso accessors
└─ TwistCotwistEquivalenceConditions
   ├─ cotwistH0Equivalence spends the unshifted condition on the actual cotwist
   └─ twist/cotwist exact equivalences use Mathlib's canonical package

Triangle.FirstMapNormalizationData (T : J ⥤ Triangle C)
├─ supplied raw-to-named isomorphisms on the first two projections
├─ supplied square identifying the transported first map with a named map
└─ normalized triangle family naturally isomorphic to T, preserving
   pointwise distinguishedness and leaving the third projection unchanged

Enhancement W (kernel category W ≃ H⁰ of a pretriangulated dg category)
├─ liftedCocycle / conePresentation: noncanonical enhanced lift of any ordinary map
└─ coneTriangleFunctor: dg cones read in W through the comparison equivalence
   └─ FourierMukai.KernelCone: pointwise and source-natural transform triangles
      ├─ KernelConeNormalizationData: FM adapter to Triangle.FirstMapNormalizationData
      │  └─ shiftedConeKernel: enhanced shift whose transform is the shifted cone transform
      └─ KernelTransformationData: a kernel arrow realizes any named transformation
         ├─ KernelTransformationConeData: enhanced representative and chosen dg cone
         │  └─ normalizationData: feeds the generic endpoint transport
         ├─ FourierMukai.CounitKernelData: adjunction-counit specialization
         │  ├─ ordinary kernel arrow conv Q P ⟶ O_Δ whose transform is the counit
         │  │  (conv Q P represents Φ_Q ⋙ Φ_P diagrammatically; Huybrechts writes P ∘ Q)
         │  └─ toConeData: choose a closed representative and dg cone in any enhancement
         │     └─ FourierMukai.CounitKernelConeData
         │        ├─ twistKernel presents the selected enhanced cone
         │        └─ exact kernel evaluation gives a source-natural counit triangle
         │           Φ_Q ⋙ Φ_P ⟶ 𝟭_Y ⟶ twist, pointwise distinguished
         ├─ AdjunctionUnitKernelData: definitional right-adjunction-unit specialization
         │  ├─ ordinary kernel arrow O_Δ ⟶ conv P Q whose transform is the unit
         │  └─ AdjunctionUnitKernelConeData
         │     ├─ exact kernel evaluation gives the unshifted unit triangle
         │     │  𝟭_X ⟶ Φ_P ⋙ Φ_Q ⟶ cotwistCone ⟶ (𝟭_X)⟦1⟧,
         │     │  pointwise distinguished
         │     ├─ cotwist = cotwistCone⟦-1⟧, pointwise functor-category shift
         │     ├─ cotwistKernel: shifted enhanced cone presenting cotwist
         │     └─ inverse rotation gives
         │        cotwist ⟶ 𝟭_X ⟶ Φ_P ⋙ Φ_Q ⟶ cotwist⟦1⟧,
         │        pointwise distinguished
         ├─ DualTwistKernelData: left-adjunction unit specialization after swapping C and C'
         │  └─ DualTwistKernelConeData reuses the unit/cotwist construction
         │     ├─ dualTwistKernel presents the pointwise shifted left-unit cone
         │     └─ dualTwist ⟶ 𝟭_Y ⟶ Φ_Q ⋙ Φ_P ⟶ dualTwist⟦1⟧,
         │        pointwise distinguished
         └─ DualCotwistKernelData: left-adjunction counit specialization after swapping C and C'
            └─ DualCotwistKernelConeData reuses the counit/twist construction
               ├─ dualCotwistKernel presents the selected left-counit cone
               └─ Φ_P ⋙ Φ_Q ⟶ 𝟭_X ⟶ dualCotwist ⟶ (Φ_P ⋙ Φ_Q)⟦1⟧,
                  pointwise distinguished

The kernel category is never required to be an H⁰: the comparison equivalence of an Enhancement carries a geometric kernel category such as Dᵇ(Coh(Y × Y)) to the dg side, and the shift and exactness compatibility of that comparison are instance hypotheses to be discharged by the realization.

The ordinary-category interface

SphericalTwistData k C E (SphericalTwist/Definition.lean) is the functor T_E on an arbitrary k-linear pretriangulated C, stated once and supplied rather than constructed:

SphericalTwistData k C E
├─ T with T.Additive, T.CommShift ℤ, T.IsTriangulated as instance fields
├─ copower : C ⥤ C           (F ↦ Hom^•(E,F) ⊗ E, supplied; no tensor is defined here)
├─ ev : copower ⟶ 𝟭, π : 𝟭 ⟶ T, δ : T ⟶ copower ⋙ ⟦1⟧, and the distinguished triangle at each F
├─ copower_class : K₀.IsRankOne copower (chiRight k C E) [E]      (second supplied input)
├─ class_T       [T F] = [F] - [copower F]       proved from K₀.of_triangle
└─ map_eq_twistK₀  K₀.map T = twistK₀ k C E     proved, not a field

The two dg object twists above realize it (SphericalTwist/ObjectTwistData.lean): additive and scalar-linear, given the Euler copower formula, with T's three instances proved by DGFunctor.h0CommShift and DGFunctor.h0IsTriangulated. The Seidel--Thomas autoequivalence and braid theorems are supplied statements, AutoequivalenceStatement and BraidStatement, nowhere asserted; the first projects to TwistShaped and so to the action on Stab, the second only to the K₀ braid identity, which is far weaker.

Deliberately open seams

  1. IsPretriangulated (DGFunctor C D) is registered whenever D is pretriangulated (DGFunctor.isPretriangulated_dgFunctor), so a cone of dg functors is now a cone in a pretriangulated dg category rather than a pointwise construction. The shift is DGFunctor.shiftedFunctor, whose action on a degree-p morphism carries the sign (-1)^(n * p); the Leibniz rule forces it (IsShiftBy.shiftMap_d) and no constant sign works.

    The comparison data is now there too. IsShiftBy.shiftMap_compare makes the objectwise comparison of two chosen shifts natural in every degree, and IsShiftBy.comp'_shiftMap says transport across a composite shift is the two transports in turn. On top of those, DGFunctor.shiftedFunctorAdd and DGFunctor.shiftedFunctorZero are the two coherences -- shifting by n then m agrees with shifting by n + m, and shifting by 0 changes nothing -- each closed and invertible, so each is an isomorphism in the dg category of dg functors rather than merely a map.

    This family is now assembled into Mathlib's shift interface at the first ordinary categorical boundary where that statement makes sense: DGFunctor.z0HasShift is a HasShift (Z0 (DGFunctor C D)) ℤ instance. Its functors are DGFunctor.z0ShiftFunctor; they act on closed degree-zero transformations by HomogeneousNatTrans.shiftedDegreeZero, whose identity and composition laws are explicit. Its zero and addition isomorphisms are z0ShiftFunctorZeroIso and z0ShiftFunctorAddIso. The latter uses shiftedFunctorAddInv as its hom, because Mathlib points from the total shift to the iterated shift, and shiftedFunctorAdd as its inv.

    DGFunctor.z0ShiftMkCore proves all three standard coherences. Its associativity law reuses DGFunctor.shiftedFunctorAdd_assoc: the inverse paths of the two categorical composites are exactly shiftedFunctorAddAssocLeft and shiftedFunctorAddAssocRight. The two unit laws identify addition by zero with, respectively, the shifted inverse zero comparison and the inverse zero comparison on the shifted functor. Nothing parallel is installed on H0 (DGFunctor C D): that category already receives the generic H0.hasShift from its pretriangulated dg structure, so a second instance would duplicate an existing abstraction. The dg-functor shift packaging seam is therefore closed.

    Source regrading is now packaged at the same root rather than rebuilt by cone consumers. IsShiftBy.precompEquiv says that precomposition with a selected shift inverse identifies Hom^p(X,W) with Hom^(n+p)(X[n],W), and HomogeneousNatTrans.sourceShiftEquiv applies it in the dg category of dg functors. It preserves and reflects closedness. Thus a closed degree-one cone projection can be read canonically as a closed degree-zero map out of the selected [-1] functor shift; no new shift instance or sign convention is introduced.

    Inverse rotation now consumes that interface generically. H0.shiftFunctorCompIsoId_hom_app exposes the dg-witness representative of Mathlib's packaged cancellation, and ConeData.inverseRotateFirstH0_app_eq_shiftedFstH0 proves that the first map of the inverse-rotated cone triangle is the source-regraded projection. The minus sign in coneTriangle.mor₃ cancels the minus sign in invRotate, and ConeData.shiftedFstH0_eq states the result as an equality of natural transformations through shiftedFunctorH0Iso. This is a cone-generic coherence theorem, not another adjunction comparison.

    At the H⁰ boundary, shiftedFunctorH0Iso_commShift additionally proves that the package constructed directly from F[n] agrees with the composite package on H⁰(F) ⋙ [n]; the proof reduces the two shift orders with the associativity comparison and cancels the two Koszul signs. The transported form composes this with endpoint equivalences, using the generic Pretriangulated.commShiftIso_commShift rather than another isolated comparison formula.

  2. The repository has strict dg functors, not the Morita quasi-functor and bimodule framework used by the spherical-functor theorem. Consequently it does not claim that the two recorded equivalence conditions imply full sphericality. The two strict adjunction comparisons themselves are now present at the generic dg-cone root. CounitConeData.twistAdjointComparison pastes the regraded closed projection T[-1] ⟶ S R with the left counit to obtain L T[-1] ⟶ R; UnitConeData.cotwistAdjointComparison pastes the left unit with the cotwist-cone inclusion to obtain R ⟶ C L. Both are closed degree-zero transformations with explicit component formulas and named H⁰ maps. The conventional target in Anno--Logvinenko's notation is (F L)[1] for F = C[-1]. DGFunctor.shiftedFunctorCompIsoIdH0 descends Mathlib's shiftFunctorCompIsoId from the HasShift category of closed dg functors, and cotwistAdjointComparisonShiftedH0 composes its inverse with the raw map to obtain R ⟶ (C[-1] L)[1]. Thus the cancellation uses the packaged add, zero, unit, and associativity laws instead of a paper-specific shift comparison. AdjointComparisonConditions records Mathlib IsIso witnesses for these canonical maps and exposes their asIsos; cotwistShiftedIso is the conventional shifted-target form of the same cotwist condition. It does not accept unrelated natural isomorphisms. No IsIso conclusion follows from construction or from TwistCotwistEquivalenceConditions.

    That comparison seam is now closed at the reusable H⁰ functoriality root. The identity components of DGFunctor.h0IdIso and DGFunctor.h0CompIso have named computation lemmas; strict dg vertical composition and both dg whiskerings descend to ordinary natural transformations, with whiskering passing through the canonical h0CompIso compositors. Those compositors now satisfy the explicit associativity and two unit equations expected of Mathlib's pseudofunctor interface. They remain equations rather than a Pseudofunctor instance: the repository has not bundled its universe-polymorphic dg categories and quotient two-cells into the bicategory that such an instance requires.

    DGAdjunction.h0_whiskerLeft_counit uses this boundary to identify the descended left-counit whisker with ordinary left whiskering of h0Counit, through the canonical compositor, associator, and right unitor. Thus twistAdjointComparisonH0_eq_inverseRotateFirstH0_comp_h0Counit advances the existing inverse-rotation factorization all the way to the ordinary adjunction counit. It introduces neither a second comparison map nor an invertibility or sphericality conclusion.

    The formally dual law DGAdjunction.h0_whiskerRight_unit identifies the descended right-whiskered dg unit with ordinary right whiskering of h0Unit, through the left unitor, associator, and canonical compositors. UnitConeData.cotwistAdjointComparisonH0_eq_h0Unit_comp_inrH0 combines it with the generic descent law for the left-whiskered cone inclusion, factoring the unshifted comparison R ⟶ C L through the ordinary unit and descended inr. The conventional cotwist remains F = C[-1]; its existing [-1][1] cancellation stays a separate shift-coherence step, so this factorization claims no new shifted comparison.

    DGAdjunction.h0 now compares a dg adjunction with an ordinary one on H⁰. DGAdjunction.H0Presentation further conjugates that adjunction through supplied equivalences from the two homotopy categories and identifies its endpoints with named ordinary functors. It reuses Mathlib's adjunction composition and natural-isomorphism transport, and exposes the resulting unit and counit formulas for later cone and Fourier--Mukai consumers. It constructs neither the endpoint comparisons nor a kernel realization. When the named endpoints are transforms, RightAdjointKernelData.ofH0Presentation and its left-adjoint mirror feed the resulting adjunction into the existing kernel-adjunction interfaces; their correspondence-swap comparisons are definitional. They still construct no kernel or kernel arrow. The endpoint-normalization operation is the generic Triangle.FirstMapNormalizationData: it replaces the first two projections of any triangle-valued functor and makes a compatible first map literal. DGAdjunction.H0Presentation.counitFirstMapNormalizationData now supplies that interface for the transported dg counit triangle. Its normalized family has literal vertices G ⋙ F, 𝟭 Y, and the transported dg twist, with literal first map (P.toAdjunction A).counit; the remaining two maps are named natural transformations, and the raw and presented families are naturally isomorphic. Pointwise distinguishedness uses only the target equivalence's triangulatedness. Thus the generic ordinary dg-counit-cone bridge is closed. It does not identify this transported dg twist with an independently chosen Fourier--Mukai cone, make the twist exact, or assert autoequivalence or sphericality. The source-side mirror unitFirstMapNormalizationData likewise presents the transported dg unit triangle with literal vertices 𝟭 X, F ⋙ G, and the transported unshifted unit cone, with first map (P.toAdjunction A).unit; it names the remaining maps and raw comparison and uses only the source equivalence's triangulatedness for pointwise distinguishedness. Its inverse rotation is now the generic presentedCotwistTriangle, whose first vertex is the pointwise [-1] shift transportedCotwist and whose second map is literally the presented adjunction unit. The reusable DGFunctor.transportedShiftedFunctorH0Iso and its specialization transportedCotwistH0Iso compare that first vertex with the transport of H⁰ of the actual shifted dg unit cone; equality after transport is not claimed. transportedCotwistH0Iso_commShift also specializes the canonical sign-correct compatibility of this comparison. DGFunctor.transportedH0 now packages the shift and exactness capabilities of an arbitrary H⁰ dg functor after ordinary equivalence transport. Its cotwist specialization proves exactness from the supplied triangulated source equivalence and packages an ordinary exact autoequivalence only under an explicit equivalence hypothesis on the unshifted cone. Thus the generic ordinary adjunction-cone presentation, conventional cotwist rotation, and transported-capability seams are closed. When the named functors are Fourier--Mukai transforms and the kernel unit is presented by RightAdjointKernelData.ofH0Presentation, presentedUnitTriangleObjIso now uses Mathlib's triangle-isomorphism completion theorem at each source object to compare this triangle with an independently chosen Fourier--Mukai unit cone. Inverse rotation gives the corresponding objectwise cotwist comparison, including the transport of the actual shifted dg cone. The chosen third component is not proved natural in the source object. On the counit side, presentedCounitTriangleObjIso directly compares the presented dg twist triangle with the independently chosen Fourier--Mukai counit cone at every target object, again with no naturality for the chosen third component. These automatic comparisons alone yield no functor isomorphism. Triangle.FirstMapNormalizationData.ComparisonData now packages the extra natural third-vertex isomorphism and two remaining triangle squares. PresentedUnitComparisonData and PresentedCounitComparisonData specialize it and derive natural twist/cotwist and triangle-family isomorphisms when that data is supplied. Their nested ShiftCompatibility refinements can additionally select Fourier--Mukai twist/cotwist CommShift structures and require Mathlib compatibility of the projected natural isomorphisms; this transfers IsTriangulated from the dg presentations. On the cotwist side, the comparison from the actual shifted dg cone is then compatible by composition, with no additional field. Neither the bare comparison data nor the refinements are constructed here. Given the separate explicit H⁰ equivalence hypotheses on the dg twist and unshifted unit cone, however, the natural comparisons now package the selected twist and cotwist kernels as KernelAutoequivalences; ShiftCompatibility together with the supplied triangulated endpoint equivalence makes those ordinary equivalences exact. This still implies no dg quasi-equivalence, choice independence, target shift-structure uniqueness, inverse-kernel formula, or sphericality. The dg notion is still the strict one: Anno--Logvinenko work with homotopy adjunctions of bimodules, and no comparison with those exists.

    The left-adjunction dual twist now reaches this same interface without a parallel abstraction. DGAdjunctionDualTwistComparison swaps the endpoint categories and correspondences and reuses PresentedUnitComparisonData through the definitional conversion from the presented left-adjoint kernel to the corresponding right-adjoint datum. Its semantic PresentedDualTwistComparisonData alias exposes the objectwise and supplied natural comparisons, conventional and actual-dg shift compatibility, exactness, and conditional KernelAutoequivalence packaging. The conventional dual twist receives exactly the existing [-1] shift; the facade does not construct comparison data, infer equivalence, or add a second CommShift structure. Symmetrically, DGAdjunctionDualCotwistComparison reuses the counit/twist comparison for the left-adjunction counit. Its dual cotwist is the unshifted dg counit cone, so this facade adds no inverse rotation or [-1] shift. It exposes the same objectwise, supplied-natural, shift-compatible, exact, and conditional kernel-autoequivalence layers under dual-cotwist names. Thus both left-adjunction facades are closed over the two existing comparison roots; neither creates new comparison or shift data.

    The canonical TwistCotwistEquivalenceConditions now also feeds the two right-adjunction Fourier--Mukai kernel packages directly. Its twist and unshifted-cotwist-cone quasi-equivalences imply exactly the two H⁰ equivalence witnesses required by the existing comparison constructors; the cotwist constructor remains solely responsible for the conventional [-1] shift. The adapters package the selected twist and cotwist kernels independently, since their endocorrespondences and enhancement categories need not agree. Their exactness still requires the separately supplied comparison ShiftCompatibility and triangulated endpoint equivalence. These conditions provide no equivalence for either left-adjunction dual cone, no relation between the two selected kernels, and no sphericality.

    The conditions are no longer inert, though. DGFunctor.h0Equivalence (dg-enhancements-e10) turns a quasi-equivalence into an equivalence on H⁰, so twistH0Equivalence and cotwistConeH0Equivalence make the twist and the unshifted cotwist cone autoequivalences of H⁰. The conventional dg functors are now named dualTwistFunctor and cotwistFunctor. The functor equality DGFunctor.shiftedFunctor_h0_eq and its natural-isomorphism wrapper compare H⁰(F[n]) with H⁰(F) ⋙ [n]; the reusable shiftedFunctorH0Equivalence transports ordinary equivalences through that comparison. Consequently cotwistH0Equivalence spends the recorded unshifted condition on the actual [-1] cotwist. It does not infer a dg quasi-equivalence. This is the first categorical invertibility statement about a twist here; everything earlier was numerical, on K₀, or a construction with no invertibility attached. Exactness is separate from equivalence. Both dg-level ingredients are now automatic for every dg functor. DGFunctor.preservesShifts maps the invertible homogeneous shift element, while DGFunctor.preservesChosenCones maps the strong split cone witness; the images of the extracted fst and snd projections give an explicit inverse to the target splitting map. DGFunctor.h0CommShift and h0IsTriangulated package the resulting coherent, non-instance conclusion on H⁰. The older cone-functor constructors retaining endpoint witnesses remain useful structured proofs, but those arguments are no longer mathematical prerequisites. Together with twistH0Equivalence, automatic exactness makes the twist an exact autoequivalence. The conventional dual twist and cotwist are the [-1] shifts of two stored cones. The reusable Pretriangulated.shiftFunctorCommShift now supplies the required Koszul sign and shiftFunctorIsTriangulated proves exactness of every integral shift without installing a global instance. DGFunctor.shiftedFunctorH0CommShift and shiftedFunctorH0IsTriangulated compose that signed package with the canonical comparison on H⁰(F) and transport across shiftedFunctorH0Iso; the latter derives cone preservation automatically while retaining the shift witness that selects its comparison. The new shiftedFunctorH0Iso_commShift identifies that transported package with the direct package on H⁰(F[n]), and transportedShiftedFunctorH0Iso_commShift preserves the identification after ordinary endpoint transport. Thus downstream cotwist consumers can compose compatibility proofs without manufacturing a source CommShift with Functor.CommShift.ofIso. Consequently dualTwistH0IsTriangulated and cotwistH0IsTriangulated close exactness for the two conventional [-1] functors without endpoint cone-preservation arguments. Finally twistH0EquivalenceIsTriangulated and cotwistH0EquivalenceIsTriangulated combine the ordinary equivalence and exactness halves with Mathlib's canonical Equivalence.IsTriangulated interface, which derives compatible inverse shift data and exactness. The four distinguished adjunction triangles now also compute their conventional functors' K₀ actions as the identity minus the corresponding adjunction composite. For an object twist, the evaluation triangle now similarly gives identity minus the evaluation functor on K₀. EvaluationData.IsEulerCopower names the explicit, choice-invariant realization capability that its class is chiRight k C E X • [E]; under that input the object classes match the existing numerical twistK₀ formula, and automatic dg-functor exactness makes the induced maps match. None of these results is sphericality, which still needs all four Anno--Logvinenko conditions and the Morita framework above.

  3. CategoryTheory/Shift/FunctorCategory.lean now supplies the pointwise shift on a functor category, and Functor.ExactFamily is now built on that canonical structure: it extends Mathlib's F.CommShift ℤ and adds only pointwise triangulatedness. The former FamilyCommShift record, its separately stored evaluated comparisons, and its manual naturality field have been removed.

    The bifunctor projection is coherent at the data level, not only at the level of comparison maps. Functor.CommShift₂.firstFamilyCommShift and secondFamilyCommShift assemble the two functor-valued CommShift structures explicitly from Mathlib's CommShift₂; they are not global instances. After composition with the strict evaluation comparison, firstFamilyEvaluationCommShift_eq and secondFamilyEvaluationCommShift_eq identify them with the partial structures already selected by CommShift₂. Thus ExactBifunctor still retains the Koszul compatibility between its two variables, while its two ExactFamily projections transport triangulatedness across an equality of the complete CommShift data. Evaluation of ExactFamily.mapTriangle is consequently proved against Mathlib's ordinary mapTriangle; no parallel triangle-map interface remains. This seam is closed.

  4. EvaluationData.functor is the generic RHom(E,-) ⊗ E dg functor and EvaluationData.evaluation its degree-zero transformation to the identity, both in Algebra/Homology/DGCategory/Copower.lean. They are built on IsCopowerOf, which states the tensoring the way IsShiftBy states the shift -- by its universal property, as data plus a bijectivity condition -- because HomologicalComplex.HasTensor does not synthesize for the ℤ-indexed shape at the pin, so there is no tensor product of complexes to build the object with. HasCopower K X and HasCopowers C now package mere existence in Mathlib's HasLimit style: the classes contain no preferred object, while copowerData makes a noncomputable selection for consumers. The narrower HasEvaluationData E is exactly the capability needed here, and follows automatically from HasCopowers C.

    This choice is coherent rather than merely available. The canonical comparison between two copowers is closed and composes strictly, so any two EvaluationData E choices determine a canonical isomorphism EvaluationData.compareIso in Z⁰ (DGFunctor C C). The theorem compare_comp_evaluation says that this isomorphism commutes strictly with their evaluation transformations. Thus later consumers may depend on the existence capability without treating a selected evaluation family as mathematically significant. Moreover, DGFunctor.PreservesChosenCones.ofIso transports strong cone preservation across any such dg-functor isomorphism. This remains useful compatibility data even though exactness no longer requires a supplied evaluation-specific witness.

    The functor is not inert. evaluation_isClosed gives the transformation objectwise cones, so EvaluationData.TwistConeData is the Seidel--Thomas twist T_E = Cone(RHom(E,-) ⊗ E ⟶ id) of an object, and its twistTriangleFunctor is a functor H⁰ C ⥤ Triangle (H⁰ C) with every value distinguished. This is a second twist, not a duplicate of the adjunction one: CounitConeData.twist is the cone of an adjunction counit, this one is attached to a single object with no adjunction in sight. The two agree when RHom(E,-) and - ⊗ E are the adjoint pair of a spherical functor out of Perf(k), which is exactly what cannot be stated here.

    H⁰(T_E) is a triangulated functor without an evaluation-specific exactness hypothesis. DGFunctor.preservesShifts holds for every dg functor because a shift element is a closed two-sided invertible element (IsShiftBy.inv, hom_inv, inv_hom). Cone preservation is automatic for a different reason: IsConeOf.bijective extracts fst and snd, whose five matrix identities survive any additive dg functor and supply the inverse to the mapped cone splitting for every target object. The earlier mapping-out/mapping-in objection from the copower interface was therefore a red herring: the proof uses the cone's projections, not the copower universal property. No pretriangulated assumption is needed to map an individual witness; it is needed only to choose cones globally.

    There is a second, independent boundary at scalars. IsCopowerOf is AddCommGrpCat-valued and represents all additive cochains, so it models an additive copower over ℤ; the coefficient in chiRight is a dimension over k. The former therefore cannot imply the latter at general k. DGCategory.Linear now exposes the genuinely k-linear Hom-complex, and DGCategory.LinearCopower supplies the copower root over a commutative ring. Its universal family is a chain map in ModuleCat k; composition is a linear equivalence with Mathlib's HomComplex.Cochain, not with all additive cochains. Thus there is no forgetful projection to IsCopowerOf, because retaining its additive surjectivity would be inconsistent with the intended tensor product. DGCategory.LinearEvaluation now assembles that root into LinearEvaluationData k E: a k-linear dg functor, a closed evaluation transformation, and coherent comparison isomorphisms between choices. DGLinear.homFunctor now assembles fixed-source right composition into the k-linear dg functor Hom(E,-), and linearCopowerAdjunction packages the selected copower functor as its strict dg left adjoint. The unit is the universal copower chain map, the counit is definitionally the selected scalar-linear evaluation transformation, and both triangle identities are the existing representing equations. Thus scalar-linear evaluation is no longer merely adjunction-shaped. Its selected evaluation cones are definitionally the adjunction's counit cones. Arbitrary evaluation choices compare canonically at the dg-functor and full H⁰ triangle-functor levels, strictly over the identity inclusion and coherently under further changes of choice. The source is the dg category of all module complexes, not Perf(k); this supplies no additive evaluation adapter, adjoint on the other side, quasi-equivalence, or sphericality statement. DGCategory.LinearCopowerFunctor now packages the homogeneous coefficient action as a k-linear dg functor out of C^dg(ModuleCat k). Its Hom-complex comparison commutes with differentials, chain-homotopic maps agree in H⁰, and homotopy equivalences yield isomorphic witnessed copowers. The selected wrapper through the existing Cdg.h0Functor seam lives in HomotopyCategory.DGEnhancement.LinearCopower. This is not quasi-isomorphism invariance. CochainComplex.FiniteCohomologyPresentation now records an explicit homotopy equivalence to a finite biproduct of single homology objects, and normalizes that model to degree-zero singles shifted by -i using Mathlib's SingleFunctors.shiftIso. The DG-facing FiniteCohomologyPresentation.linearCopowerIso transports this data through linearCopowerFunctor, using ordinary additivity to preserve the finite biproduct and the established coherent shift comparison. Thus the shifted finite-sum presentation seam is closed for supplied formality data. The scalar unit is now computed directly from the linear representing property, and its strict comparison is packaged as an isomorphism in Z⁰ C before descending to H⁰ C. DGEnhancement.LinearCopowerFiniteFree owns the reusable next leaf: a chosen finite basis expands a degree-zero coefficient copower into copies of the original object, using categorical reindexing for an arbitrary finite index universe; Module.finBasis supplies the finite-free finrank specialization. A supplied finite cohomology presentation with finite free homology therefore yields a nested finite biproduct of shifts of X, with multiplicity the corresponding finrank. These isomorphisms are noncanonical and make no basis-independence claim. The scalar-linear Hom-cohomology seam is now closed as well. IsShiftBy.linearHomIso upgrades the defining degreewise bijections of a chosen shift to an isomorphism of ModuleCat k-valued Hom-complexes, and H0.homologyShiftLinearEquiv combines it with Mathlib's homology and shift APIs to identify Hⁿ(DGLinear.homComplex k X Y) linearly with Hom_{H⁰ C}(X, Y⟦n⟧). The same-sign +n shift on the target is intentional: right composition with the degree--n shift element lands in the (-n)-shifted Hom-complex, whose degree n homology is degree zero. This comparison has representative laws computing the degree-zero map as H0.homMk and the explicit-shift map as right composition by the shift element. It is pointwise and non-numerical; it does not assert finiteness, formality, or a dimension formula. The first generic K₀ prerequisite is also closed: K₀.of_biproduct computes the class of any finite biproduct as the sum of the classes of its factors, with a constant-family multiplicity corollary. This lives at the triangulated Grothendieck-group root and is intentionally independent of copowers. The supplied-presentation scalar-copower Euler class is now closed as well: FiniteCohomologyPresentation.linearCopowerK₀Of combines linearCopowerFinrankIso, K₀.of_biproduct, and K₀.of_shift_int to compute the selected copower class as Mathlib's HomologicalComplex.homologyEulerChar times [X]. Its explicit finite presentation and finite-free hypotheses are data, not inferred formality. The shared endofunctor-level numerical seam is now closed. K₀.IsRankOne records the objectwise formula before exactness is known, is invariant under natural isomorphism, and determines K₀.map only when the usual additive, shift-compatible, triangulated hypotheses are supplied. H0.homComplex_homologyEulerChar_eq_chiHom identifies the two existing junk-total Euler sums, and LinearEvaluationData.IsEulerCopower.ofFiniteCohomologyPresentations is owned by the homotopy-category DG-enhancement layer, where it consumes a supplied presentation family to prove the direct scalar-linear evaluation formula. The generic H⁰ umbrella exports only the predicate and its choice-independence theorem. The existing additive EvaluationData.IsEulerCopower is a second specialization of the same interface, not a consequence of the linear one. Any passage between the two evaluation packages remains explicit. The coefficient-side formality foundation is now closed over a division ring: CochainComplex.homotopyEquivHomologyModel constructs a noncanonical homotopy equivalence from every unbounded ModuleCat k-valued cochain complex to its zero-differential homology model. This uses projective splittings and imposes no finite-dimensionality or support hypothesis. The finite-support comparison is now closed: CochainComplex.homologyModelIsoFiniteCohomologyModel identifies the full model with finiteCohomologyModel under an explicit vanishing witness, and FiniteCohomologyPresentation.ofFiniteSupport composes it with formality. The automatic H⁰ Euler consumer is now closed as well. H0.homComplexFiniteCohomologyPresentation selects the finite support from HomFiniteBounded, and LinearEvaluationData.IsEulerCopower.ofHomFiniteBounded feeds those presentations into the existing finite-presentation interface. The direct numerical twist corollaries are now closed too: twistK₀Of_eq_twistK₀_ofHomFiniteBounded and twistK₀Map_eq_twistK₀_ofHomFiniteBounded no longer ask callers to pass the automatic IsEulerCopower witness or a chosen-cone witness. The generic, additive object-twist, scalar-linear object-twist, and enhanced-adjunction exactness and K₀ consumers now all use automatic dg-functor cone preservation internally. No naturality or quasi-isomorphism invariance is inferred from formality. The direct scalar-linear cone consumer is now closed as well. LinearEvaluationData.TwistConeData is a thin name for the generic cone of scalar-linear evaluation; its choice comparison, full H⁰ triangle comparison, and retained structured cone-preservation constructor delegate to the existing strict-square, cone-triangle, and 3-by-3 interfaces, while exactness uses the automatic dg-functor package. For the selected copowers this cone is definitionally the counit cone of linearCopowerAdjunction; arbitrary choices are canonically identified with that presentation in Z⁰ and as full triangle functors. The generic H⁰ leaf computes its K₀ action as identity minus evaluation, and SphericalTwist.LinearObjectTwistK0 combines that formula with LinearEvaluationData.IsEulerCopower to obtain the existing numerical twistK₀, without an additive-evaluation adapter. Exactness is automatic for its dg functor, but this proves no autoequivalence.

    Concrete scalar-linear existence is now closed for the standard dg category Cdg (ModuleCat k) in the same universe. DGCategory.Model.LinearCopower uses Mathlib's total tensor product, coproduct eliminator, and module tensor universal property to construct the degreewise tensor--Hom linear equivalence; the Koszul formula for the total differential proves that the curried identity is the universal chain map. This supplies HasLinearCopowers k (Cdg (ModuleCat k)) without a parallel tensor construction. The same-universe restriction is Mathlib's current monoidal ModuleCat restriction, not an abstract requirement. Concrete additive HasCopowers remains open, as do scalar-linear instances for other dg categories. The generic existence/choice interfaces, including linear evaluation, and their comparison coherences are closed, as is choice-independence of the additive cone-preservation capability. The comparison seam is closed at both dg-functor and full H⁰-triangle levels. IsConeOf.isoOfStrictSquare lifts endpoint isomorphisms in a strict square, ConeData.triangleIsoOfStrictSquare carries the result to a natural isomorphism of triangle functors, and EvaluationData.TwistConeData.twistTriangleIsoOfEvaluation applies it to the evaluation square while preserving the original same-evaluation twistTriangleIso wrapper. These comparisons commute with all three triangle components, are identities on one choice, and compose coherently. No theorem relates the object twist to a spherical object: that comparison needs Perf(k) as a dg category, which the repository does not have, so nothing here calls E spherical or claims T_E is an autoequivalence.

  5. KernelTransformationData now owns the reusable statement that an ordinary kernel arrow realizes a named natural transformation between functors in supplied endpoint presentations. KernelTransformationConeData owns its noncanonical enhanced representative and cone, while KernelConeNormalizationData feeds its raw transform family to the generic Triangle.FirstMapNormalizationData endpoint/first-map transport. CounitKernelData and CounitKernelConeData retain their public contracts as equivalent adjunction-counit presentations, with inverse adapters and simp round trips to those interfaces. In particular, CounitKernelData.arrow and its exact transform equation are supplied. Once they are available, Z0.toH0_full and pretriangulated cone existence choose a noncanonical closed representative and dg cone in any enhancement; forgetting those choices recovers the original ordinary datum. The strong hypothesis E.kernelTransform.Full is one sufficient abstract constructor, not a theorem proved here. Producing the ordinary arrow geometrically still needs convolution, the diagonal unit kernel, the adjunction trace, and proof that its transform is the counit. Given those supplied data, CounitKernelConeData.counitTriangleInSource specializes the generic normalization interface: Mathlib's triangle transport identifies the raw enhanced-cone triangle with a source-natural triangle whose first two vertices are literally Φ_Q ⋙ Φ_P and 𝟭 Y, and whose first map is literally the adjunction counit. Every value is distinguished under the existing exactness hypotheses. This is not a distinguished triangle in the functor category, does not make the twist exact or an autoequivalence, and does not make the selected enhancement representative or cone canonical. CommShift and IsTriangulated for the enhancement comparison remain supplied where these pointwise distinguished triangles are consumed. The right-adjunction unit kernel is now the second generic consumer: AdjunctionUnitKernelData realizes O_Δ ⟶ P ⋆ Q, and its enhanced form normalizes the unshifted triangle 𝟭_X ⟶ Φ_P ⋙ Φ_Q ⟶ cotwistCone ⟶ (𝟭_X)⟦1⟧. AdjunctionUnitKernelConeData.cotwist then uses the pointwise functor-category shift and cotwistTriangleInSource inverse-rotates the family to cotwist ⟶ 𝟭_X ⟶ Φ_P ⋙ Φ_Q ⟶ cotwist⟦1⟧. This is still only a choice-dependent ordinary functor and a pointwise distinguished family. The Fourier--Mukai normalization root also transports a shifted enhanced cone through the enhancement and kernel-family CommShift comparisons. Specializing at -1 gives cotwistKernel, an explicit kernel whose transform is naturally isomorphic to the cotwist, and hence proves only that the cotwist is a kernel functor. For the right-adjunction unit, the independently chosen dg and kernel cone triangles are now noncanonically isomorphic at every source object, and so are their conventional cotwist objects. This pointwise comparison does not identify the two cotwist functors naturally. The additional realization data and remaining triangle-map squares are now explicitly packaged by PresentedUnitComparisonData, which derives the natural comparison but is not constructed here. Its ShiftCompatibility refinement records compatibility of the conventional cotwist isomorphism with a selected Fourier--Mukai CommShift and transfers triangulatedness. Compatibility of the intermediate actual shifted dg cotwist follows from the canonical transported comparison and is exposed by transportedDGCotwistIso_commShift. The same supplied natural comparison transports kernel-functor status to both versions of the dg cotwist and, under the explicit equivalence hypothesis on the unshifted dg unit cone, packages cotwistKernelAutoequivalence; the existing shift refinement and supplied triangulated source equivalence make that equivalence exact. The left-adjunction unit is now a third consumer: DualTwistKernelData swaps the two correspondences and reuses LeftAdjointKernelData.toRightAdjointKernelData, so its enhanced form names the kernel-presented dual twist and the pointwise distinguished family dualTwist ⟶ 𝟭_Y ⟶ Φ_Q ⋙ Φ_P ⟶ dualTwist⟦1⟧ without a second cone or normalization construction. The corresponding left-adjunction counit is a fourth consumer: DualCotwistKernelData performs the same swap through the generic counit/twist interface, names the selected cone kernel, and exposes Φ_P ⋙ Φ_Q ⟶ 𝟭_X ⟶ dualCotwist ⟶ (Φ_P ⋙ Φ_Q)⟦1⟧ pointwise, again without new cone machinery. Pointwise distinguishedness cannot supply CommShift or IsTriangulated for either shifted cone functor. The corresponding dg-adjunction counit triangle has an ordinary presentation through DGAdjunction.H0Presentation; it is now noncanonically isomorphic at every target object to the independently selected Fourier--Mukai counit triangle, so the transported dg twist and kernel twist are objectwise isomorphic. PresentedCounitComparisonData now packages the additional natural twist isomorphism and remaining triangle-map squares and derives the natural triangle comparison, while its ShiftCompatibility refinement transfers triangulatedness to a selected compatible Fourier--Mukai shift structure. Under an explicit equivalence hypothesis on H⁰ of the dg twist it also packages the selected cone kernel as twistKernelAutoequivalence, with exactness supplied jointly by that refinement and the triangulated target equivalence. No geometric constructor supplies either record or either equivalence hypothesis.

  6. No theorem currently identifies a categorical spherical object with a spherical functor from Perf(k), or derives the Seidel--Thomas autoequivalence from SerreFunctor.IsSphericalObject.

  7. The Enriques classification layer supplies categorical spherical objects and paper-level Fourier--Mukai conclusions through explicit seams. It does not yet provide the dg evaluation functor, counit kernel arrow, or a concrete Enriques surface needed to instantiate this tree.

These seams should remain below any Enriques-specific adapter. In particular, an Enriques leaf must consume a generic evaluation/counit-kernel theorem; it must not add a second paper-specific cone or spherical-twist structure.


Exported from docs/architecture/spherical-twist-roadmap.md at the commit that removed it from the tree.

Home

Architecture

CI and runners

Decision records

Roadmaps

Clone this wiki locally