Repository navigation
Roadmap Projective Families
Chris Dare edited this page Nov 15, 2026
·
1 revision
schema: roadmap/1
slug: projective-families
project: derived-alg-geo-lean
title: "Projective families — meet arXiv:2607.28411 in the middle"
status: active
phase: sequenced
brief: |
This track maps arXiv:2607.28411v1 onto the mathematics already present in
DerivedAlgGeo and starts only the missing, non-overlapping roots. The paper
is not a blank-slate moduli project: Appendix A, relative HN interfaces,
relative-perfect objects, a groupoid-valued moduli problem, Big-Zariski
stack data, boundedness witnesses, Quot presentations, and much of the
absolute stability foundation already exist.
The track has two trunks. The paper-critical trunk is classical:
relative-perfect and universally-gluable complexes lead to a groupoid-valued
moduli problem, open bounded semistable substacks, semistable reduction, and
good moduli. The dg-enhancement trunk is parallel, not a logical gate on the
paper. It meets the classical trunk at perfect complexes and negative Ext:
the future derived moduli object is expected to have a universally-gluable
one-truncated locus whose classical truncation is the existing moduli stack.
Every milestone follows `docs/architecture/abstraction-tree.md`. Shared
capabilities move upward before paper-specific leaves are added. No item
may introduce `KLinearCategory`, a sibling Grothendieck presentation, a
stability-owned generic stack, or a conclusion carried as a structure field.
goal:
hmw: "How might we complete the prerequisites of arXiv:2607.28411 by joining its top-down construction to the repository's existing bottom-up roots, without producing a second category, K-group, derived category, or moduli hierarchy?"
objective: "Build the canonical relative numerical K-group, categorical transfer and semiorthogonal roots, neutral scheme-derived substrate, and classical moduli spine needed for Theorems 9.1 and 10.3; then assemble the absolute and relative projective-family constructions while the dg-derived enhancement proceeds independently."
key_results:
- "Definition 8.17 is represented by an inhabited geometric interface whose four conditions are theorems or explicit hypotheses at the correct layer."
- "The relative numerical K-group of Definition 8.7 is a saturated quotient of the canonical numerical K-theory spine and has the functoriality of Proposition 8.10."
- "Theorem 9.1 consumes one generic adjoint/exact transfer spine rather than separate absolute and relative copies."
- "The algebraic `(P¹)^n` seed, transfer to `P^n`, and pullback to projective `X` culminate in Theorems 6.2, 6.8, and 10.3."
- "The existing relative-perfect moduli work is completed through standard descent, algebraicity, semistable reduction, and good-moduli properness without overstating the current Big-Zariski presentation layer."
- "The dg enhancement has a proved classical derived seam, but no paper-critical milestone waits for derived moduli foundations."
architecture_contract:
policy: "docs/architecture/abstraction-tree.md"
canonical_roots:
ordinary_linear: "Mathlib CategoryTheory.Preadditive and CategoryTheory.Linear"
dg_linear: "CategoryTheory.DGCategory plus DGLinear as a refinement"
grothendieck: "CategoryTheory.GrothendieckPresentation with K₀Ab, K₀, and K₀dg specializations"
stacks: "generic CategoryTheory.Sites/descent roots; AlgebraicGeometry supplies scheme-site leaves"
derived_geometry: "neutral AlgebraicGeometry derived-category modules; stability families are consumers"
required_review:
- "Name the canonical root and its neutral owning module."
- "Name two consumers or the statement-layer exception."
- "Provide a projection, instance, abbreviation, or proved comparison."
- "Prove agreement when the target already has an instance."
- "Record and retain counterexamples to false generalizations."
prohibited:
- "paper-specific roots imported by generic subjects"
- "parallel Grothendieck presentations or K-linear category hierarchies"
- "algebraicity, preservation, descent, boundedness, or properness represented by unproved conclusion fields"
source:
work: "Li--Liu--Liu--Macrì--Perry--Stellari--Zhao, Stability conditions and moduli spaces on projective families"
arxiv: "2607.28411v1"
coverage_map: "registry/coverage-2607.28411.json"
pinned_pdf_sha256: "f8770154235fe2c82698513b4633b3ee509fa11f722190a4c9f573fca589a98c"
evidence:
- "arXiv:2607.28411v1 §§2--11 and Appendices A--B."
- "arXiv:1902.08184v4 for relative stability, HN structures, boundedness, semistable reduction, and good-moduli inputs."
- "Mathlib pin and the measured repository surfaces recorded by the stability-families and dg-enhancements roadmaps."
- "ADR-0013 owns reusable moduli infrastructure in this monorepo."
- "The 2026-08-27 abstraction audit and numerical-k-theory roadmap record the Grothendieck-group failure and its repair."
wont:
- "No claim that dg enhancements are a prerequisite of the paper as written."
- "No derived Artin-stack or Toën--Vaquié moduli construction before the required homotopical algebra exists."
- "No projectivity claim for good moduli spaces; §1.7 leaves it open."
- "No promotion of a coverage coordinate without the evidence and owner acceptance required by the coverage-map validator."
- "No forced unification of dimension-specific Chern-character formulas; the numerical roadmap already falsified that generalization."
generated_by: { agent: codex, at: "2026-08-28" }
generations:
- { rev: 1, at: "2026-08-28", note: "Track created from a read of the pinned v1 paper, the live main tree, open PRs, worktrees, and all existing roadmaps. PF1 materialized as milestone #33 with issues #850--#852. DG3 materialized separately as #853--#855 because it is a parallel trunk." }
- { rev: 2, at: "2026-08-28", note: "Implementation started in fresh non-overlapping worktrees for #850 (generic stack ownership) and #856 (Appendix B almost-disconnected morphisms). The relative numerical K-group #851 and semiorthogonal root #852 remain open research lanes." }
- { rev: 3, at: "2026-08-28", note: "The merged stack-root lane closed #850. Issue #851 then started with a neutral saturated-quotient root and the algebraic core of Definition 8.7: fiber sum, connected-family relation quotient, universal property, induced-map laws, and connected-family/one-fiber tests. The geometric fiber overlattices and eta specialization classes remain the next slice." }
- { rev: 4, at: "2026-08-28", note: "Issue #851 now packages admissible families as statement-layer comparison data over the same relative quotient root. Point-independence and its torsion-free universal property are inherited from the saturated quotient; the next geometric slice remains the fiber overlattice and eta specialization construction." }
- { rev: 5, at: "2026-08-28", note: "Issue #852 started at the generic triangulated root. The pinned Mathlib reconnaissance selects ObjectProperty orthogonals and the canonical triangulated envelope; the implementation separates the semiorthogonal carrier, triangulated components, classical fullness, strong fullness, and mapped-component orthogonality, with one-block and product-factor inhabitants before geometry." }
- { rev: 6, at: "2026-08-28", note: "Issue #852 merged its generic semiorthogonal root, including independent triangulated/fullness predicates and two compiled inhabitants. Work returns to #851 for the geometric fiber overlattice and eta-specialization adapter on the existing relative quotient." }
- { rev: 7, at: "2026-08-28", note: "Issue #851 now models each geometric fiber group as the image of its eta base-change homomorphism in a common ambient group, records finite relative index as a predicate, and transports admissible-family classes through this image into the existing saturated quotient. Proposition 8.10 functoriality is the next slice." }
- { rev: 8, at: "2026-09-02", note: "Epic e7 registered as #1033 and started with its categorical half: transfer of pre-stability and stability conditions along a phase-detecting functor (Definitions 3.1/3.6, Remarks 3.2/3.7, Lemmas 3.5/3.9) under Phase/Transfer/{PreStability,LocallyFinite}.lean and Support/Transfer.lean, in parallel with the Enriques lane. Conservativity is derived from Slicing.PreimageData rather than assumed. The geometric Dᵇ(Coh) realizations, Propositions 3.3/3.8, Lemma 3.10, and §2.2 remain open under the same epic; PF3 still has no GitHub milestone." }
- { rev: 9, at: "2026-09-02", note: "Second slice of #1033 stacked on the first: Slicing.PreimageData.mapEquiv completes Lemma 3.5(1)/3.9(1) by transporting the lifting witness along compatible twists; AutPair.preimage and AutPair.preimage_act give Lemma 3.2 of arXiv:2601.22994 with charges; HasBayerProperty.preimage / BayerProperty.preimage give the Bayer transfer of Lemma 3.25; stabilityMass_preimage and stabilityDist_preimage_le extend Lemma 3.5(3) by the mass coordinate; AlgebraicGeometry/DerivedCategory/Stability/BoundedCoherentPushforward.lean realizes f_♯σ on Dᵇ(Coh)." }
- { rev: 10, at: "2026-09-02", note: "Third slice of #1033, stacked on the second: Phase/Transfer/Adjoint.lean proves the adjoint transposition of Proposition 3.8, Slicing.MapsSemistableGE.of_mapsSemistableLE, turning the paper's condition (3.2) on the comonad f^* f_* into the Corollary A.23 hypothesis on the monad f^* f_!; the Foundation gains leProp_of_postnikovTower (upper phase cut closed under towers) and geProp_of_forall_hom_eq_zero (orthogonality characterizes the lower cut). Verifying (3.2) geometrically and feeding the conclusion to the Dqc Appendix-A machinery remain the open steps of Proposition 3.8." }
- { rev: 11, at: "2026-09-02", note: "Fourth slice of #1033: Phase/Transfer/IndExtensions.lean and Inducing.lean replace the empty InducingBoundary.lean with the categorical Corollary A.23. Slicing.IndExtensions packages Definition A.22 (Lemma A.14 data at every dual t-structure of the slicing), monad_isRightTExact turns assumption (v') on semistable objects into the right t-exactness Theorem A.17 needs, nonempty_inducedTStructures applies Polishchuk.induce at every phase and reads (A.3)/(A.4) as (A.8), preimageData is the corollary, and mapsSemistableGE_iff ties (v') to the third slice's MapsSemistableGE. Polishchuk.induce gains its first consumer. Companions: the dual t-structure dictionary and boundedness in the Foundation, preimageLift instances and inverseImageLift beside preimageLift. Still open: realizing IndExtensions for Dᵇ(Coh) inside Dqc and the adjunction f_! ⊣ f^* on Dqc." }
- { rev: 12, at: "2026-09-02", note: "Fifth slice of #1033: Proposition 3.8 end to end, categorically. Adjunction.restrictInverseImage and restrictInverseImageRight (TStructure/Restriction.lean) restrict f_! ⊣ f^* and f^* ⊣ f_* to the bounded subcategories when the monad and comonad preserve them; Slicing.IndExtensions.preimageData_of_mapsSemistableLE chains condition (3.2) on the restricted comonad through the adjoint transposition (slice 3) and Corollary A.23 (slice 4) to the preimage slicing. The geometric inputs are now exactly the hypotheses of that one theorem." }
- { rev: 13, at: "2026-09-03", note: "Sixth slice of #1033, the first geometric one: the direct image on Dᵇ(Coh). Coh.pushforward is f_* on coherent sheaves along a closed immersion into a locally Noetherian scheme, exact and coherent by the owned #572 step 2; SchemeBaseChange.HasCoherentPushforward (Families/CoherentPushforward.lean) is the exact-coherent-pushforward contract with closed immersions as its instance, and boundedCoherentDerivedPushforward is Definition 3.1's f_* : Dᵇ(Coh X) ⥤ Dᵇ(Coh Y) through Mathlib's mapDerivedCategory, triangulated and bounded by theorems rather than fields. Stability/BoundedCoherentPullback.lean realizes f^♯σ (Definition 3.1(3); arXiv:2601.22994 Definition 3.1, pullback half) from an explicit witness or from the A.17 output. Still open: a finite-morphism instance of the contract, identity and composition laws for the direct image, Proposition 3.3 itself, and the remaining Dqc inputs of Proposition 3.8." }
- { rev: 14, at: "2026-09-03", note: "Seventh slice of #1033: finite morphisms inhabit the direct-image contract. Coh.pushforward now takes any finite morphism into a locally Noetherian scheme (Modules/Coherent/Pushforward/Finite.lean): coherence generalizes the closed-immersion argument through IsFinite.finite_app and the finite-ring-map affine theorem, and exactness on coherent sheaves comes from a new neutral fact, pushforward along an affine morphism preserves epimorphisms of quasi-coherent sheaves (Modules/Pushforward/Affine.lean), itself from surjectivity of epimorphisms of quasi-coherent sheaves on affine sections (Modules/Affine/Epi.lean, via Mathlib's tildeEquiv). The three coherence theorems are generalised in place under their names (a closed immersion is finite in Mathlib), Coh.pushforward moves to Finite.lean, and hasCoherentPushforwardOfIsFinite replaces the closed-immersion instance and f^♯σ is now available along every finite morphism, the generality of Propositions 3.3/3.24 and of arXiv:2601.22994. Still open: identity and composition laws for the direct image, Proposition 3.3 itself, and the Dqc inputs of Proposition 3.8." }
- { rev: 15, at: "2026-09-03", note: "Eighth slice of #1033: the direct image is pseudofunctorial. Functor.mapDerivedCategory gains its transport, identity, and composition isomorphisms at Mathlib's site (Algebra/Homology/DerivedCategory/ExactFunctor.lean, by Localization.liftNatIso); Families/CoherentPushforwardCoherence.lean derives identity and composition isomorphisms for the coherent, derived, and bounded direct images from the contract's comparison isomorphism alone, so they hold for every instance rather than being postulated as the pullback side's GeometricDerivedPullbackIdentity/Composition classes are; BoundedCoherentPushforwardPreimageData gains identity, comp, preimage_identity, preimage_comp, and the inducing data its identity model. Still open: Proposition 3.3 and the Dqc inputs of Proposition 3.8; a pullback-side inhabitant of HasCoherentPullback."}
- { rev: 16, at: "2026-09-03", note: "Ninth slice of #1033: exact pullback inhabits HasCoherentPullback. Pullback preserves finite presentation along every morphism (Modules/Coherent/Pullback.lean: a presentation on a chart is transported by Mathlib's Presentation.map along pullback on slices, packaged in Modules/Pullback/Restriction.lean as the left adjoint pullbackOverFunctor so that colimit preservation is stated where the transport looks for it, with the slice equivalences matching the structure sheaves in Modules/Restriction/StructureSheaf.lean and finality of inverse image on opens at its topology site), so Coh.pullback exists for every morphism and is exact when module-sheaf pullback is; Families/CoherentPullback.lean discharges the contract for IsExactPullback, hence for flat morphisms, through mapDerivedCategory. The pushforward f_♯σ of stability conditions is thereby available geometrically along flat morphisms for the first time. Still open: PreservesPerfectPullback (locally free sheaves under pullback), Proposition 3.3/3.8 and the Dqc inputs." }
- { rev: 17, at: "2026-09-03", note: "Tenth slice of #1033: coherent pullback preserves perfect complexes. Module-sheaf pullback preserves locally free sheaves along every morphism (Modules/Pullback/LocallyFree.lean: pullback on slices sends free sheaves to free sheaves, and a chart basis is transported through pullbackOverFunctor; the site-level reconstruction LocalGeneratorsData.ofFreeIsos in Algebra/Category/ModuleCat/Sheaf/LocallyFree.lean is the rank-n form of IsInvertible.of_trivializations); fixed-rank atlases pull back with their rank (Divisors/LocallyFreePullback.lean); and Families/PerfectPullback.lean discharges PreservesPerfectPullback for every HasCoherentPullback instance, through the contract comparison and a single-functor isomorphism built from derivedFactors. perfectDerivedPullback now exists wherever the contract does. Still open: retiring PreservesPerfectPullback as a class together with its eleven binders, the GeometricDerivedPullbackIdentity/Composition inhabitants, Proposition 3.3/3.8 and the Dqc inputs." }
- { rev: 18, at: "2026-09-03", note: "Eleventh slice of #1033: PreservesPerfectPullback retired as a class. It was a theorem after the tenth slice, so coherentDerivedPullback_preservesPerfect, perfectDerivedPullback, and the GeometricDerivedPullbackIdentity/Composition contracts now take only HasCoherentPullback, and they live in Families/PerfectPullback.lean after the generator theorem; Families/BoundedGeometry.lean keeps the bounded contract only; the nine [PreservesPerfectPullback _] binders across five declarations in Stability/BoundedCoherentBaseChange.lean are gone (sixteen in all with the seven in Families/BoundedGeometry.lean, together with the class and its global instance). No new mathematics. Still open: the GeometricDerivedPullbackIdentity/Composition inhabitants, Proposition 3.3/3.8 and the Dqc inputs." }
- { rev: 19, at: "2026-09-03", note: "Twelfth slice of #1033: identity and composition laws for coherent derived pullback, for every HasCoherentPullback instance, and the two classes that postulated them retired. Families/CoherentPullbackCoherence.lean mirrors CoherentPushforwardCoherence.lean: sheafPullbackId/Comp from the comparison isomorphism and modulePullbackId/Comp through Coh.ι; coherentDerivedPullbackId/Comp by the universal property of the localization at the contract factorization (DerivedCategory.isoOfFactors, idFactors, compFactors in Algebra/Homology/DerivedCategory/ExactFunctor.lean, since the contract derived pullback is data rather than mapDerivedCategory); bounded and perfect versions through the inclusions. GeometricDerivedPullbackIdentity/Composition are gone with their five binders in Stability/BoundedCoherentBaseChange.lean, which now consumes the theorems. Every pullback-side contract of the epic is now either inhabited for exact pullback or a theorem. Still open: Proposition 3.3/3.8 and the Dqc inputs." }
- { rev: 20, at: "2026-09-03", note: "Thirteenth slice of #1033, opening the Dqc foundations lane: the categorical core of assumption (v') for base change along a field extension. Slicing.IndExtensions.mapsSemistableAisle_of_coproduct proves assumption (v') of Corollary A.23 for any monad sending a semistable object to a coproduct of copies of itself (the mechanism of Theorem 2.8(1): the monad of X_l to X is tensoring with l over k), preimageData_of_coproduct is Corollary A.23 under it, isLE_zero_of_semistable puts semistables in their aisle, and Slicing.preimage_square (Basic.lean) is the shape of Lemma 3.10(1.b)/(2.b): two-step preimage slicings around a commuting square agree. No geometry yet. Still open for Theorem 2.8(1): the Dqc inputs (coproducts and compact generation of Dqc, the bounded coherent locus, Rf_* and Lf^* on Dqc), and, categorically, Corollary A.23 at a source subcategory other than F^{-1}Q: for infinite l/k no nonzero coherent object of X_l has coherent direct image, so the preimage of D^b(Coh X) under pi_* meets D^b(Coh X_l) only in zero and the built-in P = F^{-1}Q of Polishchuk.induce does not reach the paper's slicing." }
- { rev: 21, at: "2026-09-03", note: "Fourteenth slice of #1033, the first Dqc foundation: coproducts in a localization (CategoryTheory/Localization/Coproducts.lean, Mathlib's path). For L a localization at W with a right calculus of fractions, C with coproducts of shape iota and W stable under them, L (coproduct X) is a coproduct of the L (X i) (Localization.isColimitCofan, from right fractions with a common denominator Sigma.map and the Ore condition summand by summand), so D has coproducts of that shape (Localization.hasCoproductsOfShape) and L preserves them (preservesCoproductsOfShape). This is the argument that D(A) has coproducts for A with exact coproducts, at the level of the localization Qh from the homotopy category; the specialization to D(X.Modules) needs coproducts in the homotopy category, AB4 for X.Modules, and homology on D(X.Modules) preserving the coproducts isColimitCofan produces (the hpres input of sigma_mem), the next slices, after which the Dqc locus has coproducts by the existing sigma_mem." }
- { rev: 22, at: "2026-09-03", note: "Fifteenth slice of #1033, the second Dqc foundation: coproducts in the homotopy category. Homotopies on the summands of a coproduct of complexes assemble to a homotopy on the coproduct (Homotopy.sigma in Algebra/Homology/Homotopy/Sigma.lean, through the degreewise description of a coproduct of complexes, HomologicalComplex.sigmaXIso and sigmaXDesc in Algebra/Homology/HomologicalComplexLimits.lean), so the quotient of a coproduct is a coproduct in the homotopy category (HomotopyCategory.isColimitCofan in HomotopyCategory/Coproducts.lean), the homotopy category has coproducts of every shape the underlying category has, and the quotient functor preserves them, both as instances. With the fourteenth slice at Qh this leaves one input for coproducts in D(A): stability of quasi-isomorphisms under coproducts, which follows from exactness of coproducts (AB4) in A because homology commutes with exact colimits." }
- { rev: 23, at: "2026-09-03", note: "Sixteenth slice of #1033, the third Dqc foundation: coproducts in the derived category. Homology of short complexes commutes with exact colimits (ShortComplex.homologyFunctor_preservesColimitsOfShape in Algebra/Homology/ShortComplex/Limits.lean, from HasExactColimitsOfShape through mapHomologyIso at colim and the colimit cocone of a short complex of diagrams), hence homology of complexes does (HomologicalComplex.homologyFunctor_preservesColimitsOfShape in HomologicalComplexLimits.lean, through homologyFunctorIso) and homology on the homotopy category preserves coproducts (HomotopyCategory.homologyFunctor_preservesCoproductsOfShape, through the essentially surjective quotient by preservesCoproductsOfShape_of_essSurj in CategoryTheory/Limits/Preserves/Shapes/Products.lean), so quasi-isomorphisms are stable under coproducts (HomotopyCategory.quasiIso_isStableUnderCoproductsOfShape in Algebra/Homology/Localization.lean) and the fourteenth slice at Qh gives DerivedCategory.hasCoproductsOfShape with Q, Qh and every homology functor preserving them (DerivedCategory/Coproducts.lean), all instances under HasCoproductsOfShape κ C and HasExactColimitsOfShape (Discrete κ) C. Remaining for the Dqc locus: AB4 for X.Modules, after which sigma_mem closes Dqc X under coproducts with its hpres input discharged by homologyFunctor_preservesCoproductsOfShape." }
- { rev: 24, at: "2026-09-03", note: "Seventeenth slice of #1033, the last Dqc coproduct input: coproducts in X.Modules are exact, so Dqc(X) is closed under coproducts with no hypothesis. Colimits of presheaves of modules of a shape are exact when they are in abelian groups (PresheafOfModules.hasExactColimitsOfShape, Algebra/Category/ModuleCat/Presheaf/AB.lean, through the forgetful functor to presheaves of abelian groups and Mathlib's HasExactColimitsOfShape.domain_of_functor), hence so are colimits of sheaves of modules (SheafOfModules.hasExactColimitsOfShape, Sheaf/AB.lean, through the sheafification adjunction and Adjunction.hasExactColimitsOfShape), hence colimits in X.Modules of every such shape (Scheme.Modules.hasExactColimitsOfShape with the classes ab4 and ab5, AlgebraicGeometry/Modules/AB.lean, transported by inferInstanceAs as Mathlib transports limits and colimits; the site-level files register AB4OfSize and AB5OfSize as well). With the sixteenth slice, SchemeDerivedCategory X has those coproducts and every cohomology functor preserves them by instance search, so Dqc.SchemeQuasicoherentDerivedCategory.sigma_mem now takes only the membership of the summands, Dqc(X) is closed under coproducts as an object property, has them, and its inclusion creates and preserves them (isClosedUnderColimitsOfShape_discrete, ι_preservesCoproductsOfShape, through DerivedCategory.cohomologyIn_isClosedUnderColimitsOfShape_discrete): #721's third bullet is closed, and the Dqc locus is closed under coproducts, the first of the closure properties Example A.10 and the Ind-extension of Definition A.22 need. Still open on the Dqc side: compact generation by perfect complexes (RΓ commuting with coproducts) and the bounded-coherent identification, both Mathlib-scale." }
- { rev: 25, at: "2026-09-03", note: "Eighteenth slice of #1033: Theorem A.17 and Corollary A.23 beyond hypothesis (iv). ObjectProperty.liftOfLE (TStructure/Restriction.lean) restricts F to a triangulated P ≤ F⁻¹ Q; formulas (A.3)/(A.4) hold against the large target t-structure with no target subcategory (isLE_iff_isLE_map, isGE_iff_isGE_map) with zero reflection only on P, given that P is stable under the truncations of the source t-structure, and boundedness follows from P ≤ tD.bounded.inverseImage F (tStructure_isBounded_of_le_bounded_inverseImage); Polishchuk.induceLarge is Theorem A.17 in that shape, with truncation stability (Definition A.7) and the bounded-image containment as hypotheses in place of (iv), Polishchuk.induceOfLE its restatement inside a bounded Q for P ≤ F⁻¹ Q through InducedTStructureDataLarge.toInducedTStructureData and tStructure_isBounded_iff_le_bounded, and Polishchuk.induce the case P = F⁻¹ Q where hasInducedTStructure_of_preimage proves the stability; Slicing.IndExtensions.nonempty_inducedTStructures_of_le, preimageData_of_le and preimageData_of_coproduct_of_le (Phase/Transfer/Inducing.lean) are Corollary A.23 on P ≤ F⁻¹ Q. The P ≤ F⁻¹ Q shape is the one Slicing.InducedTStructures consumes, with the slicing on Q; no geometric instance with P ≠ F⁻¹ Q is claimed (f_* along a proper non-finite morphism satisfies the containment but not conservativity on Dᵇ(Coh X): RΓ(ℙ¹, 𝒪(−1)) = 0). Given a triangulated D, induceLarge is induceOfLE at Q = tD.bounded, so the two constructors are interderivable and the large one is primitive. It is not the shape Theorem 2.8(1) needs for an infinite extension, which rev 20 asked for: there π_* sends no nonzero coherent object into Dᵇ(Coh X), so P ≤ F⁻¹ Q fails for Q = Dᵇ(Coh X), and the paper recognizes 𝒫_ℓ against the Ind-extensions on Dqc(X) directly; induceLarge is the Theorem A.17 half of that, and the large-target twin of Slicing.InducedTStructures / preimageData for Corollary A.23 is the open categorical item. Still geometric: the Ind-extensions of the standard slicing inside Dqc(X), the compactly generated source t-structures on Dqc(X_ℓ), stability of Dᵇ(Coh X_ℓ) under their truncations, conservativity of π_* on Dᵇ(Coh X_ℓ), and τ̂-boundedness of π_* F for coherent F (descent to a finite subextension plus closure of compactly generated coaisles under coproducts, Remark A.12(3))." }
- { rev: 26, at: "2026-09-03", note: "Nineteenth slice of #1033, the categorical Theorem 2.8(3): semistability is detected by base change. The coaisle of a compactly generated t-structure is closed under coproducts (TStructure.IsCompactlyGeneratedBy.isGE_one_coproduct, CompactlyGenerated/Coaisle.lean: a map from a compact generator into a coproduct is a finite sum of maps into the summands, and right orthogonality passes to the coproduct-and-extension closure), an object is semistable of phase φ exactly when its HN phases are at least φ and below every ψ > φ (Slicing.semistable_iff_geProp_ltProp, Slicing/CoreConsequences.lean, the first file downstream of all of phiPlus/phiMinus, geProp_of_semistable and semistable_of_phiPlus_eq_phiMinus), so with the Ind-extensions, whose aisles are closed under coproducts by Lemma A.14(i), E is semistable of phase φ iff a nonempty coproduct of copies of it lies in the aisle of τ̂'_φ and the coaisles of τ̂'_ψ for ψ > φ (Slicing.IndExtensions.semistable_iff_isLE_zero_and_isGE_one_of_iso_coproduct, Phase/Transfer/BaseChange.lean, forward by closure under coproducts and backward by closure under retracts), which constrains the coproduct only as an object of D, and when the coproduct itself lies in Q the two conditions reassemble into semistability there (semistable_iff_semistable_of_iso_coproduct), hence for the preimage slicing along F on F⁻¹ Q, E is semistable iff L E is when F L E is a nonempty coproduct of copies of E (semistable_iff_preimage_of_coproduct; no adjunction is used, only the isomorphism, and semistability on the right is that of F L E by definition of the preimage slicing): the categorical Theorem 2.8(3) with F = π_*, L = π^*, π_* π^* E = E ⊗_k ℓ, for a FINITE extension. For an infinite one E ⊗_k ℓ is not coherent, so hQL cannot hold at E and both Q-level statements are vacuous; what the still-open large-target statement of rev 25 will consume is the D-level reduction together with isLE_zero_coproduct and isGE_one_coproduct, none of which mention Q. Coproducts are indexed by Type, as IsCompactObject.exists_finite_sum is. Next in §3.3: Lemma 3.10(1.a), the equivalence of (3.3) and (3.4) under base change, which combines this with Theorem 2.8(1) at Ind-extension level." }
- { rev: 27, at: "2026-09-03", note: "Twentieth slice of #1033, the last purely categorical piece of Theorem 2.8(1) at the level of the phase-indexed t-structures: Corollary A.23 with recognition against the large target. Slicing.IndExtensions.largePhase (Phase/Transfer/IndExtensions.lean) is the phase collection of the Ind-slicing in the ambient category, the aisle of tau_phi cut against the coaisles of tau_psi for psi > phi; largePhase_iff_semistable identifies it with s.P on Q (both halves Lemma A.14(iii), reassembled by semistable_iff_geProp_ltProp of rev 26). Note the primed/unprimed distinction: Theorem 2.8(1) states its recognition on the cuts of the unprimed tau_phi, while IndExtensions carries the primed family that Corollary A.23 uses, and largePhase is the intersection of the primed cuts. Slicing.IndExtensions.InducedTStructuresLarge (Phase/Transfer/Inducing.lean) is the large-target twin of Slicing.InducedTStructures: formulas (A.8) against tau_phi with no target subcategory, so it is statable where s.geProp and s.ltProp are not, which for an arbitrary field extension is everywhere outside Q. Unlike the bounded twin it also carries Step 4 boundedness, which the bounded twin can drop because it never truncates on the source. Its hom_vanishing, stated on the pulled-back collection largePreimagePhase, is shorter than the bounded twin's because largePhase already carries coaisle membership at every psi > phi, so no target HN filtration is needed; ofIso transports it along a natural isomorphism, and toInducedTStructures / ofInducedTStructures make it interderivable with the bounded shape whenever the detecting functor factors through Q (the converse direction must be handed boundedness), so it extends rather than competes. nonempty_inducedTStructuresLarge is Corollary A.23 in that shape, on Polishchuk.induceLarge of rev 25 at every phase. Its hbdd (F sends P into the tau_phi-bounded objects) is the one hypothesis this shape adds over nonempty_inducedTStructures_of_le, which bought it from hle; for F = pi_* it is the descent to a finite subextension of Remark A.12(3), and it JOINS rather than replaces hzero, hG/htC and htrunc on the list of geometric inputs still open for pi. Assembling the output into a Slicing is the accounting of Slicing.PreimageData with one entry moved: closure under isomorphisms and zero membership are still formal, hom_vanishing is now supplied, and two are open -- the shift law, since there is no slicing on D to borrow it from and IndExtensions relates two phases only by the aisle inclusion le_zero_anti (never by the shift), so Phat(phi) shifted by one equals Phat(phi+1) needs a coprodClosure shift equality the repository does not have; and HN existence for largePreimagePhase, which cannot be got the bounded twin's way (Slicing.InducedTStructures.hn_exists lifts a TARGET HN filtration and the ambient category carries no slicing) and needs the finite phase-truncation argument run on the source. Also in this slice: rev 26's semistable_iff_isLE_zero_and_isGE_one_of_iso_coproduct is renamed semistable_iff_largePhase_of_iso_coproduct and restated through largePhase, the normal form, which did not exist when rev 26 landed; there is one name for that statement, not two. Next: the shift law and the HN half, after which Theorem 2.8(1) closes categorically; then Lemma 3.10(1.a)." }
- { rev: 28, at: "2026-09-03", note: "Twenty-first slice of #1033, the phase shift for the Ind-slicing, and with it the slicing P_l itself modulo Harder-Narasimhan existence. ObjectProperty.coprodClosure_le_shift_of_le_shift (Triangulated/CompactlyGenerated.lean) transports the coproduct-and-extension closure into a shift of a DIFFERENT property, which is what distinguishes it from the homogeneous coprodClosure_le_shift of CompactlyGenerated/Existence.lean; the aisles it compares sit at two phases. coprodClosure_map_obj cannot do it, for a library reason rather than a mathematical one: Mathlib installs no CommShift instance on shiftFunctor C n, three rotations of a distinguished triangle returning the shift with all three morphisms negated, so only a sign-corrected commutation isomorphism makes the shift triangulated and none is registered. The signs land on the morphisms, and coprodClosure constrains only the OBJECTS of the triangle in its extension constructor, so Pretriangulated.Triangle.shift_distinguished suffices and the induction goes through. With Slicing.geProp_shift and closure of Q under the shift that gives boundedAisle_shift, hence isLE_zero_shift: the aisles of the Ind-extensions shift with the phase. Definition A.22 asserts nothing relating two phases (rev 27 recorded le_zero_anti, the aisle inclusion, as the only relation), so this is a theorem and not an axiom, resting on clause (i) of Lemma A.14 pinning each aisle to Coprod(P(>= phi)). isGE_one_shift follows without a second argument, the coaisle being the right orthogonal of the aisle (isGE_iff_orthogonal) and the shift fully faithful. largePhase_shift and largePhase_shift_iff are the phase shift Phat(phi) shifted by one equals Phat(phi+1); with largePhase_isClosedUnderIsomorphisms and largePhase_of_isZero these are the three formal Slicing axioms for largePhase. Slicing.IndExtensions.largePreimage (Phase/Transfer/Inducing.lean) then assembles the slicing: five of Slicing's six fields are discharged -- the three formal ones, the shift law transported through F.commShiftIso, and hom_vanishing from rev 27 -- and only hn_exists remains, as a hypothesis. This is Corollary A.23's output in the shape Theorem 2.8(1) needs; identifying it with P_l still wants the primed/unprimed cut comparison (2.8(1) pins P_l by the cuts Phat(> phi) and Phat(<= phi+1) of the UNPRIMED family, while largePhase is the intersection of the primed cuts), and the geometric inputs hzero, hG/htC, htrunc and hbdd remain open for pi. Categorically only hn_exists is left, and it cannot be got the bounded twin's way: Slicing.InducedTStructures.hn_exists lifts a TARGET HN filtration and the ambient category carries no slicing, so it needs the finite phase-truncation argument run on the source, which is what rev 27's Step 4 boundedness field was kept for. Next: that argument, after which Theorem 2.8(1) closes categorically; then Lemma 3.10(1.a)." }
- { rev: 29, at: "2026-09-12", note: "PF5 materialized as milestone #48 for the mass-Hom lane. e14 rewritten as the bound itself (#1177) with six deliverables #1178--#1183 in dependency order: root predicate and HLR v2 pin, Lemma 2.8 reductions, Corollary 2.14(1)/Remark 7.3 from the existing stabilityDist mass comparison, Lemma 7.4 transfer on e7's preimage carrier, Example 2.9 base case, Theorem 7.5. Two consumer epics filed on the same milestone: e19 (#1184, Cheng arXiv:2608.14540: mass-Hom bound implies the full support property) and e20 (#1185, Theorem 8.23 = HLR v2 Theorem 2.35, with inequality (8.4) as the ownable slice and the HLR/BLMNPS chain as named hypotheses on e10). Tilt comparisons split off as e21, unfiled. Source facts pinned: the bound is HLR arXiv:2501.00710v2 Definition 2.7 (v1 Theorem 2.31 is v2 Theorem 2.35; v2 moved its appendix to arXiv:2606.00396, which does not use the bound); the paper cites v2 numbering." }
- { rev: 30, at: "2026-09-12", note: "Dependency update from SF11.1 (#1060): the K-flat base-change category layer now includes perfect, quasicoherent, and bounded semiorthogonal sequences plus componentwise and sequence-level pullback, pushforward, tensor, adjunction, and projection-formula interfaces. This removes the missing categorical shape recorded by e7, but it does not yet supply the geometric semiorthogonality/fullness, projection amplitude and cocontinuity, bounded-preservation, or general Lemma 3.18 witnesses. Consequently #1033 remains gated on SF11's geometric discharge and the later S-local/noetherian-heart work; no HN-existence claim is inferred from the new interfaces." }
- { rev: 31, at: "2026-09-28", note: "Stacks-and-moduli re-scope: #1185 (e20) gains the native blocker #1713 (GM1.5), whose good-moduli definition replaces the ProperGoodModuli : Prop parameter; #1203 moves to PF5 (#48) as task projective-families-t1203. The PF2 e4/e6 and PF5 e13 dispositions stay with MO2.6 (#1629)." }
items:
- id: projective-families-m1
gh_milestone: 33
kind: milestone
title: "PF1 — Shared roots for projective families"
status: in_progress
lane: now
acceptance:
- "Generic stack-in-groupoids declarations live above scheme geometry, and a sublayer gate prevents moduli roots from importing stability-family leaves."
- "The relative numerical K-group is a quotient of the existing K-theory/numerical spine, with connected-family and one-fiber tests."
- "Semiorthogonal decomposition and exceptional-collection vocabulary has a generic categorical root with two inhabitants before the projective-family leaf consumes it."
- "Appendix B's almost-disconnected morphism and filtration property are geometric roots independent of the stability leaf."
- "DG3 proceeds on its own milestone; PF1 records the seam but does not block on it."
- "Zero sorries; relevant gates and runner verification are green."
- id: projective-families-e1
gh_issue: 850
gh_sub_issues: []
kind: epic
parent: projective-families-m1
title: "Move generic stack roots above scheme geometry"
status: complete
priority: must
size: L
summary: "Relocate the generic contents of AlgebraicGeometry/Stacks/Basic to a CategoryTheory/Sites owner, preserve scheme-site adapters, and add a shrinking-allowlist gate for the current Moduli-to-StabilityCondition/Families reverse imports. This is an ownership refactor, not a new stack theorem."
tags: [architecture, stacks, layering]
labels: [area:geometry, arch:layering, type:refactor, source:2607.28411]
- id: projective-families-e2
gh_issue: 851
gh_sub_issues: []
kind: epic
parent: projective-families-m1
title: "Relative numerical K-group on the canonical K-theory spine"
status: in_progress
priority: must
size: XL
summary: "The algebraic core of Definition 8.7 now supports the fiber direct sum, a statement-layer family-relation system, quotient universal property, induced-map laws, point-independence, and the one-fiber test. The geometric slice presents each fiber overlattice as the image of its eta base-change map, keeps finite relative index as a predicate, and factors admissible-family classes into that image before reusing the same quotient. Proposition 8.10 functoriality is next. The uniformly numerical group and §10 image lattice remain further quotients/images, never sibling K-groups."
tags: [k-theory, numerical, relative, prerequisite]
labels: [area:geometry, area:moduli, prerequisite, source:2607.28411]
- id: projective-families-e3
gh_issue: 852
gh_sub_issues: []
kind: epic
parent: projective-families-m1
title: "Semiorthogonal decompositions and exceptional-collection root"
status: complete
priority: must
size: XL
summary: "The merged generic root packages ordered ObjectProperty components over Mathlib orthogonals while keeping triangulated closure, classical fullness, strong fullness, and mapped-component orthogonality as independent predicates or capabilities. One-block and product-factor inhabitants compile before geometry; exceptionality and mutation remain deferred until a theorem consumes them."
tags: [semiorthogonal, exceptional-collection, research]
labels: [area:derived, area:stability, prerequisite, source:2607.28411]
- id: projective-families-e18
gh_issue: 856
gh_sub_issues: []
kind: epic
parent: projective-families-m1
title: "Almost-disconnected morphisms and the filtration property"
status: in_progress
priority: must
size: L
summary: "Formalize Appendix B in the neutral scheme-morphism hierarchy, including flat base change and the `(P¹)^n → P^n` quotient, then expose only an adapter to the stability-transfer leaf."
tags: [almost-disconnected, filtration, morphism-property]
labels: [area:geometry, prerequisite, source:2607.28411]
- id: projective-families-m2
gh_milestone: null
kind: milestone
title: "PF2 — Neutral derived geometry and classical moduli roots"
status: planned
lane: next
acceptance:
- "Dqc, bounded-coherent, perfect, relative-perfect, and derived-pullback modules have neutral AlgebraicGeometry owners; stability-family modules are adapters."
- "Replete subprestacks, boundedness witnesses, open substacks, and presentation data are generic roots consumed by perfect-complex moduli."
- "Arbitrary pullback preserves the universally-gluable relative-perfect locus with proved coherence and a nonflat example."
- "The active SF8/SF9 issues are consumed rather than duplicated."
- id: projective-families-e4
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m2
title: "Extract generic replete-subprestack and boundedness roots"
status: planned
priority: must
size: L
summary: "Generalize the reusable content now tied to RelativePerfectModuliSubproblem, FiniteTypeBoundednessWitness, and open Zariski presentations; perfect-complex names become thin specializations."
tags: [moduli, abstraction, boundedness]
labels: [area:moduli, arch:layering, type:refactor, source:2607.28411]
- id: projective-families-e5
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m2
title: "Neutral scheme-derived category and pullback spine"
status: blocked
priority: must
size: XL
summary: "Move the general Dqc/SchemeDerived/pullback substrate out of StabilityCondition/Families after coordinating with #528, #721--#723, and #554; retain compatibility reexports and do not disrupt active proofs."
tags: [dqc, derived-pullback, layering]
labels: [area:derived, area:geometry, blocked, source:2607.28411]
links:
issues: [528, 554, 721, 722, 723]
roadmaps: [stability-families]
- id: projective-families-e6
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m2
title: "Classical relative-perfect moduli stack to standard descent"
status: blocked
priority: must
size: XL
summary: "Complete arbitrary pullback, fppf or étale descent, the full M_pug diagonal and atlas, and honest open/bounded semistable substacks. Big-Zariski presentation data remains conservatively named until these theorems exist."
tags: [relative-perfect, algebraic-stack, descent]
labels: [area:moduli, blocked, source:2607.28411]
links:
issues: [516, 517, 522, 554]
roadmaps: [stability-families]
- id: projective-families-m3
gh_milestone: null
kind: milestone
title: "PF3 — Absolute construction on E^n, (P¹)^n, P^n, and X"
status: planned
lane: next
acceptance:
- "Theorems 4.2, 5.4, 5.6, 6.2, and 6.8 are connected by proved transfer maps."
- "The algebraic seed uses the generic exceptional-collection root and an actual finite-dimensional algebra realization."
- "Embedding independence is an agreement theorem over the common numerical root."
- id: projective-families-e7
gh_issue: 1033
gh_sub_issues: []
kind: epic
parent: projective-families-m3
title: "Absolute scalar base change and stability transfer"
status: in_progress
priority: must
size: XL
summary: "The categorical half of §3 landed first: Definitions 3.1(2)--(3)/3.6(2)--(3) as one transfer operation on PreStabilityCondition.WithClassMap and StabilityCondition.WithClassMap with class map v ∘ K₀(F), conservativity derived from Slicing.PreimageData, support-property and local-finiteness transfer (Remarks 3.2/3.7), extreme phases and the slicing distance under transfer (Lemmas 3.5/3.9), and the two-witness slicing-order transfer (arXiv:2601.22994 Lemma 3.3). The second slice adds the twist/Bayer/metric layer (Lemma 3.5(1)/3.9(1) in full with the transported lifting witness, the Bayer property under transfer as used by Lemma 3.25, the mass and three-coordinate stability distance under transfer) and the Dᵇ(Coh) realization of f_♯σ along bounded coherent pullback, from an explicit witness or from the A.17 output. The third and fourth slices add the categorical Proposition 3.8: the adjoint transposition of condition (3.2) into the Corollary A.23 hypothesis, and Corollary A.23 itself as the joint between the owned Theorem A.17 (Polishchuk.induce) and the finite phase-truncation argument, from Definition A.22 data. The sixth and seventh slices are geometric: the direct image f_* on Dᵇ(Coh) along a finite morphism (Coh.pushforward, exact by affine-pushforward exactness on quasi-coherent sheaves; the HasCoherentPushforward contract inhabited by every finite morphism; boundedCoherentDerivedPushforward through mapDerivedCategory) and the realization of f^♯σ along it from an explicit witness or the A.17 output. Still open: the geometric inputs of Propositions 3.3/3.8 (Ind-extensions of the standard slicings inside Dqc, the adjunctions f_! ⊣ f^* ⊣ f_* on Dqc), Lemma 3.10 base-change compatibility, and the §2.2 scalar extension."
tags: [stability, base-change, transfer]
labels: [area:lean, area:stability, type:feature, source:2607.28411, source:2601.22994]
- id: projective-families-e8
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m3
title: "The algebraic `(P¹)^n` seed and transfer to projective space"
status: blocked
priority: must
size: XL
summary: "Build the §4 family on powers of an elliptic curve, identify the algebraic seed through the full strong exceptional collection, and push it through the finite quotients to `P^n`."
tags: [elliptic-curve, projective-space, exceptional-collection]
labels: [area:stability, blocked, source:2607.28411]
- id: projective-families-e9
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m3
title: "Pull the projective-space construction to projective schemes"
status: blocked
priority: must
size: XL
summary: "Use a closed projective presentation, finite-Tor pullback, and the canonical numerical image lattice to prove Theorems 6.2 and 6.8 without bundling a second projective variety or numerical K-group."
tags: [projective-scheme, closed-immersion, distinguished-component]
labels: [area:geometry, area:stability, blocked, source:2607.28411]
links:
issues: [572, 586]
- id: projective-families-m4
gh_milestone: null
kind: milestone
title: "PF4 — Relative stability and the main theorem"
status: planned
lane: later
acceptance:
- "Definition 8.17 and Theorem 8.19 consume the existing uniform-support, openness, relative-HN, and boundedness roots."
- "Theorem 9.1 is proved for the paper's finite/faithfully-flat hypotheses."
- "Propositions 10.1--10.2 and Theorem 10.3 assemble the projective-family construction over Setup 8.1 bases."
- "Theorem 1.1 is obtained from Theorems 6.2, 6.8, and 10.3 without an unproved global switch."
- id: projective-families-e10
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m4
title: "Relative stability over a base and deformation"
status: blocked
priority: must
size: XL
summary: "Assemble Definition 8.17, HN structures over Dedekind bases, the stable/semistable moduli subfunctors, and the relative deformation theorem on the canonical family interfaces."
tags: [relative-stability, hn, deformation]
labels: [area:families, blocked, source:2607.28411]
- id: projective-families-e11
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m4
title: "Relative transfer and the `(P¹_S)^n`/`P^n_S` construction"
status: blocked
priority: must
size: XL
summary: "Prove Theorem 9.1 and Propositions 10.1--10.2 from the absolute seed, relative K-theory, HN/moduli inputs, and the generic transfer spine."
tags: [relative-transfer, projective-family]
labels: [area:families, area:stability, blocked, source:2607.28411]
- id: projective-families-e12
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m4
title: "Theorem 10.3 and the distinguished projective-family component"
status: blocked
priority: must
size: XL
summary: "Pull from `P^n_S` to arbitrary flat projective `X/S`, construct the image lattice, prove embedding independence, and assemble Theorem 1.1 and Corollary 10.5."
tags: [main-theorem, projective-family, moduli]
labels: [area:families, area:geometry, blocked, source:2607.28411]
- id: projective-families-m5
gh_milestone: 48
kind: milestone
title: "PF5 — Moduli consequences and complementary branches"
status: in_progress
lane: now
acceptance:
- "Mass-Hom bounds, tilt comparisons, supported total-space categories, and Bondal--Orlov reconstruction are separate consumers of the main spine."
- "Proper good moduli is proved only with the exact characteristic-zero and valuative hypotheses."
- "No projectivity claim is introduced."
- id: projective-families-e13
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m5
title: "Semistable reduction, quasi-properness, and good moduli"
status: blocked
priority: must
size: XL
summary: "Consume SF9's algebraic-stack result, relative HN/semistable reduction, and AHLH hypotheses to construct proper good moduli spaces in the supported characteristic-zero scope."
tags: [semistable-reduction, good-moduli, properness]
labels: [area:moduli, blocked, source:2607.28411]
links:
issues: [522, 524, 525]
# ----- Mass-Hom bounds (§7.1, §8.5; HLR arXiv:2501.00710v2 §2.2/§2.5) -----
# e14 is the bound itself; e19 and e20 are its two literature consumers and
# are separate epics so that PF4-blocked moduli work never holds e14 open.
# Propositions 7.6--7.7 (tilt comparisons) moved to e21, unfiled.
- id: projective-families-e14
gh_issue: 1177
gh_sub_issues: [1178, 1179, 1180, 1181, 1182, 1183]
kind: epic
parent: projective-families-m5
title: "Mass-Hom bounds: HLR Definition 2.7, the categorical calculus, and Theorem 7.5"
status: done
priority: must
size: XL
summary: "Definition 7.1 (= HLR v2 Definition 2.7) as one predicate on a k-linear triangulated category with a test class (⊤ for HLR, schemePerfect for the paper), its invariances, the generator and stable-object reductions (Lemma 2.8 / Remark 7.2), invariance on a connected component (Corollary 2.14(1) / Remark 7.3, from the existing stabilityDist mass comparison), transfer along pullback and pushforward (Lemma 7.4, on the Phase/Transfer preimage carrier), the algebraic base case (Example 2.9), and Theorem 7.5 with Remark 1.4 on top of PF3's distinguished component. Evidence for HLR's Boundedness Conjecture; the augmented-compactification half of HLR is out of scope."
tags: [mass-hom, boundedness-conjecture]
labels: [area:stability, area:lean, type:feature, source:2607.28411, source:2501.00710]
- id: projective-families-e14-1
gh_issue: 1178
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
title: "The mass-Hom bound predicate, its invariances, and the HLR v2 source pin"
status: done
summary: "HasMassHomBound σ T on PreStabilityCondition.WithClassMap with a test class T : ObjectProperty C, finrank of Hom over stabilityMass; invariance under isomorphism, shift, autoequivalence transport (Remark 1.4), and the GL̃⁺(2,ℝ) action; the semistable normal form dim Hom(A,F) ≤ C‖Z(F)‖; the arXiv:2501.00710v2 source pin, label, and a §2.2/§2.5 coverage stub (owner-gated)."
priority: must
size: M
tags: [mass-hom, root, source-pin]
- id: projective-families-e14-2
gh_issue: 1179
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
blocked_by: [1178]
title: "Generator and stable-object reductions (HLR Lemma 2.8, Remark 7.2)"
status: done
summary: "The class of A with a constant is thick (long exact Hom sequence, mass shift-invariance), so a classical generating set of T suffices; and stable F suffices via HN additivity of mass and Jordan–Hölder in the finite-length 𝒫(φ). The Jordan–Hölder input at the Mathlib pin is a named research item, proved or isolated as one hypothesis."
priority: must
size: L
tags: [mass-hom, generators, jordan-holder]
- id: projective-families-e14-3
gh_issue: 1180
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
blocked_by: [1178]
title: "Finite distance and connected components (HLR Corollary 2.14(1), Remark 7.3)"
status: done
summary: "stabilityDist σ τ < ⊤ transfers the bound with constant e^d (stabilityMass_toReal_lt_exp_mul_of_stabilityDist already exists); finite-distance classes are open, so the bound is a union of connected components. Corollary 2.14(2) is not in scope."
priority: must
size: S
tags: [mass-hom, distance, components]
- id: projective-families-e14-4
gh_issue: 1181
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
blocked_by: [1178, 1179]
title: "Transfer of mass-Hom bounds along pullback and pushforward (Lemma 7.4)"
status: done
summary: "One categorical theorem on the preimage-slicing carrier of e7 (L ⊣ G, G conservative, stabilityMass_preimage), with generation transported by L as a named hypothesis; geometric instances f^♯ along finite morphisms (HasCoherentPushforward, perfectDerivedPullback) and f_♯ along faithfully flat morphisms with perfect dualizing complex (blocked on the Dqc adjunction in #1033). Neeman–Ravenel and classical generators of Perf link to #723."
priority: must
size: L
tags: [mass-hom, transfer, generators]
- id: projective-families-e14-5
gh_issue: 1182
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
blocked_by: [1178, 1179]
title: "Algebraic stability conditions have mass-Hom bounds (HLR Example 2.9)"
status: done
summary: "Finite-length heart with finitely many simples: Hom from a simple is bounded by length, and finitely many charges in ℍ ∪ ℝ_{<0} admit a common direction, so length ≤ |Z(F)|/c. Abstract theorem plus one compiled witness (finite-heart lane or Dᵇ(FGModuleCat k) per #543)."
priority: must
size: M
tags: [mass-hom, algebraic, finite-heart]
- id: projective-families-e14-6
gh_issue: 1183
gh_sub_issues: []
kind: deliverable
parent: projective-families-e14
blocked_by: [1180, 1181, 1182]
title: "Theorem 7.5 and Remark 1.4 on the distinguished component"
status: done
summary: "Remark 7.3 plus the two transfers plus the base case, with PF3's distinguished-component inhabitant (Definition 6.4, Theorems 5.4/5.6/6.2) as the one named hypothesis until e8/e9 land; Remark 1.4 from autoequivalence transport; §7 coverage promotion proposed to the owner."
priority: must
size: M
tags: [mass-hom, theorem-7-5]
- id: projective-families-e19
gh_issue: 1184
gh_sub_issues: []
kind: epic
parent: projective-families-m5
blocked_by: [1178]
title: "Mass-Hom bounds imply the full support property (Cheng, arXiv:2608.14540)"
status: done
priority: should
size: L
summary: "Cheng Theorem 2.1 (proper X, Remark 2.2): a mass-Hom bound gives the support property for the full K_num(X), via finitely many perfect A_i separating K_num, the finite Ext-window on semistables (phase vanishing plus Serre duality with the dualizing complex), and the KS quadratic form C²|Z|² − ‖·‖². Also Definition 3.1 / Theorem 3.2 (uniform relative version, blocked on e10) and Theorem 4.2 (canonical component, blocked on e14-6 and PF3). Answers §1.7 of the paper for Stab†."
tags: [mass-hom, support-property, consumer]
labels: [area:stability, area:lean, type:feature, source:2608.14540, source:2607.28411]
- id: projective-families-e20
gh_issue: 1185
gh_sub_issues: []
kind: epic
parent: projective-families-m5
blocked_by: [1033, 1178, 1713]
title: "Theorem 8.23: the mass-Hom bound as the input to HLR's good-moduli theorem"
status: blocked
priority: should
size: L
summary: "Theorem 8.23 (HLR v2 Theorem 2.35 = v1 Theorem 2.31): on smooth projective X a mass-Hom bound makes σ a stability condition over k, hence proper good moduli in characteristic 0. Ownable now: inequality (8.4), the bound uniform over finite-type field extensions (Zariski's lemma, adjunction, π_* preserves HN filtrations from e7's Theorem 2.8 lane). The HLR/BLMNPS chain (Theorem 2.35(1)⇒(2), Propositions 2.38/2.41, Lemma 8.22, [7, 21.24]) is recorded as named hypotheses on e10's Definition 8.17 interface, never as conclusion fields. Selvaggi arXiv:2603.29053 Theorem 2.42 is the relative literature pointer."
tags: [mass-hom, good-moduli, consumer]
labels: [area:stability, area:moduli, area:lean, type:feature, source:2501.00710, source:2607.28411]
- id: projective-families-t1203
gh_issue: 1203
kind: task
parent: projective-families-m5
title: "Derive phase-slice Jordan–Hölder filtrations from local finiteness"
status: planned
priority: should
summary: "Follow-up to #1179, moved to PF5 by the stacks-and-moduli re-scope; its stale process paragraph is replaced by a Definition of done (#1629)."
labels: [research, area:lean, area:stability, type:feature]
- id: projective-families-e21
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m5
title: "Tilt-stability comparisons (Propositions 7.6--7.7)"
status: planned
priority: could
size: L
summary: "Surfaces over ℂ: Stab†_num(Dᵇ(X)) contains all geometric stability conditions, hence all have mass-Hom bounds (Proposition 7.6); threefolds satisfying the BMT conjecture and the lattice condition (i): the tilt-constructed open set 𝔓̃ lies in Stab†_{H_X} (Proposition 7.7). Consumers of e14-6; not filed."
tags: [tilt-stability, mass-hom]
labels: [area:stability, area:geometry, source:2607.28411]
- id: projective-families-e15
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m5
title: "Supported derived categories of vector-bundle total spaces"
status: planned
priority: should
size: XL
summary: "Build the generic supported-derived-category and inducing-slicing inputs for Theorems 7.8--7.9 before specializing to total spaces and local Calabi--Yau examples."
tags: [supported-derived, vector-bundle, local-calabi-yau]
labels: [area:derived, area:geometry, source:2607.28411]
- id: projective-families-e16
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m5
title: "Skyscraper stability and Bondal--Orlov reconstruction"
status: planned
priority: should
size: XL
summary: "Construct the skyscraper-sheaf moduli leaf and prove Proposition 11.6 and Theorem 11.7 on the existing Fourier--Mukai/autoequivalence roots."
tags: [skyscraper, bondal-orlov, reconstruction]
labels: [area:geometry, area:stability, source:2607.28411]
- id: projective-families-m6
gh_milestone: null
kind: milestone
title: "PF6 — DG/classical moduli convergence"
status: parked
lane: later
acceptance:
- "This milestone never blocks PF1--PF5."
- "The derived-moduli foundations exist honestly before a derived stack is defined."
- "The universally-gluable locus is proved one-truncated and its classical truncation is compared with the classical relative-perfect stack."
- id: projective-families-e17
gh_issue: null
gh_sub_issues: []
kind: epic
parent: projective-families-m6
title: "Compare the dg-derived object moduli with the classical universally-gluable stack"
status: parked
priority: could
size: XL
summary: "Long-term Toën--Vaquié convergence: after spaces, derived affines, mapping stacks, and tangent complexes exist, prove negative Ext kills higher homotopy and compare the one-truncation with the Lieblich-style classical moduli stack."
tags: [dg-moduli, derived-stack, parked]
labels: [area:dg, area:moduli, parked]
links:
roadmaps: [dg-enhancements, stability-families]Exported from .claude/roadmap/projective-families.yaml at the commit that removed it from the tree.
Architecture
- Canonical roots and specialization trees
- Structural cutover ledger
- Generalization backlog
- Dependency direction
- Loop-engineering friction log
- Mathematical ownership and abstraction boundarie
- Declaration placement
- When a proposition-valued class is justified
- Enhanced spherical twists: conventions and remai
- Verification history
CI and runners
- Bounded audit batching probe, 2026-09-27
- CI1.10: audit and exception record contention, 2
- Proposed hosted-only CI cutover for CI1.07
- Mac VM runners
- Owner CI lane: self-hosted runners in a private
- CI performance baseline, 2026-09-27
- CI1.07 exact-tree Ubuntu routing sample, 2026-09
- Ubuntu workstation runners (historical topology)
Decision records
- ADR-0001 — The seam is cold: files, not runtime (superseded)
- ADR-0002 — Identity is minted in the topic repo, (superseded)
- ADR-0003 — Elaboration and axiom evidence is pro (superseded)
- ADR-0004 — The statement registry lives in the t (superseded)
- ADR-0005 — Seven axes, four values, no aggregate (superseded)
- ADR-0006 — All contract issues are filed on `bri (superseded)
- ADR-0007 — Where the contract package lives (superseded)
-
ADR-0008 —
@[cites]ships as a shared Lake dep (superseded) - ADR-0009 — The contract package lives with the e (superseded)
- ADR-0010 — How a dg category is encoded, and whe
- ADR-0011 — Hom-complexes are complexes of abelia
- ADR-0012 — One Mathlib-style source root
- ADR-0013 — Own reusable moduli infrastructure in
- ADR-0014 — Stacks, algebraic spaces and moduli:
Roadmaps