Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 8 results · all verified · 7 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 1 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Analytic Derived Six Operations And Duality

1 · Prerequisites

2 · Summary

Objective. This page builds the ordinary six-operation calculus for sheaves of vector spaces on analytic spaces of complex algebraic varieties, over an arbitrary field and with the Axiom of Choice isolated exactly where the injective models need it. It begins with the topological notion of a proper continuous map and its basic stability, then records the elementary operations: exact inverse image with its right adjoint direct image, the derived tensor and internal Hom with the tensor–Hom adjunction, and flasque acyclicity. The proper-support functor f! is defined by proper support of sections, and its fibrewise compact-support cohomology is used to construct the exceptional inverse image f! as the right adjoint of Rf! on the bounded constructible specialization, with the support localization, base change, projection and trace calculus. Verdier duality, its compatibility with complex orientation ωX=kX[2d] on smooth complex d-folds, and the coherent six-operation interface follow, and the page closes with the ordinary middle perverse t-structure, its recollement, intersection complexes and the exactness of a∗[d] for a smooth complex map a of pure relative dimension d.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Proper continuous maps of topological spaces

Definition

Let X and Y be topological spaces (Topology on a set, open and closed sets, clopen sets, the closed-set axiomatisation, and the coarser/finer comparison) and let f:X→Y be continuous (Continuity of a map of topological spaces at a point and globally). Then f is proper when for every compact subset K⊆Y (Open cover, subcover, and compact topological space; a compact subset is a subspace that is compact in its own right) its preimage f−1(K) is a compact subset of X. Compactness of subsets is the intrinsic condition of Open cover, subcover, and compact topological space; a compact subset is a subspace that is compact in its own right, namely compactness of the subspace f−1(K) with the subspace topology (Subspace topology: the traces of the open sets, its closed sets and its bases, the continuity of the inclusion, and the characteristic property of a map into a subspace); by A subspace is compact exactly when every family of open subsets of the ambient space covering it, indexed or not, has finitely many members covering it it is equivalent to the ambient reading, so a subset is compact exactly when every cover by open subsets of the ambient space has a finite subcover, and the two readings are used interchangeably below without further comment.

This is the notion of properness used throughout the analytic six-operation development on this page, where the spaces involved are locally compact Hausdorff (Locally compact topological space: every point has a compact neighbourhood; and what this says in a metric space, Hausdorff space: distinct points have disjoint open neighbourhoods; every metrizable space is Hausdorff and the indiscrete topology on two points is not). The stability properties consumed by the later items — closed inclusions are proper, composites of proper maps are proper, properness is preserved by base change along a continuous map into a Hausdorff space, and fibres and closed subspaces of a proper map inherit properness — are proved in the lemma lem-proper-continuous-map-stability on this page. The definition involves no choice principle: it quantifies over compact subsets and preimages only.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Basic stability of proper continuous maps

Statement

Let X, Y and Z be topological spaces (Topology on a set, open and closed sets, clopen sets, the closed-set axiomatisation, and the coarser/finer comparison), with properness as in Proper continuous maps of topological spaces.

  1. If i:Z↪X is the inclusion of a closed subset, then i is proper.
  2. If f:X→Y and g:Y→Z are proper continuous maps, then the composite gf:X→Z is proper.
  3. If f:X→Y is proper, Y is Hausdorff (Hausdorff space: distinct points have disjoint open neighbourhoods; every metrizable space is Hausdorff and the indiscrete topology on two points is not) and u:Y′→Y is continuous, then the pullback f′:X×YY′⟶Y′,X×YY′={(x,y′)∈X×Y′:f(x)=u(y′)}, the subset carrying the subspace topology of the product (The product set ∏i∈IXi of functions choosing a point in each factor, the projections, the box topology, and the product topology as the initial topology of the projections; the empty product is a one-point space, Subspace topology: the traces of the open sets, its closed sets and its bases, the continuity of the inclusion, and the characteristic property of a map into a subspace), is proper.
  4. If f:X→Y is proper, then every fibre f−1(y) is a compact subset of X.
  5. If f:X→Y is proper and S⊆X is closed, then the restriction f∣S:S→Y is proper.
  6. If f:X→Y is proper and Y is locally compact Hausdorff (Locally compact topological space: every point has a compact neighbourhood; and what this says in a metric space), then f is a closed map: the image of every closed subset of X is closed in Y.

These are the stability properties consumed by the closed-support functor f! on this page; no choice principle is used in any of them.

Facts & Assumptions

Given: Topological spaces X, Y, Z and Y′, continuous maps as in the Statement, and the intrinsic convention for compactness of subsets.

[F1]

A continuous f:X→Y is proper exactly when f−1(K) is a compact subset of X for every compact subset K⊆Y (Proper continuous maps of topological spaces).

[F3]

A closed subset of a compact space is a compact subset of it, and finitely many compact subsets have compact union (A closed subspace of a compact space is compact, and a finite union of compact subspaces is compact).

[F4]

A finite product of compact spaces is compact; in particular the binary product of two compact spaces is compact, and a one-point space is compact (A product of finitely many compact spaces is compact in the product topology).

Proof

technique · direct
1.1givenF1F3F6

Closed inclusions are proper. Let i:Z↪X be the inclusion of a closed subset and let K⊆X be compact. Its preimage K∩Z is closed in the compact space K, hence compact by [F3]. The topology on K∩Z induced from Z is the same as that induced from K or X, by transitivity of subspace topology [F6]. Thus i−1(K) is compact in Z, and i is proper by [F1].

1.2givenF1algebra

Composites of proper maps are proper. For K⊆Z compact, (gf)−1(K)=f−1(g−1(K)); the set g−1(K) is compact by [F1] applied to g, and then its f-preimage is compact by [F1] applied to f. Hence gf is proper.

1.3givenF1F4

Fibres of a proper map are compact. A one-point space is compact by [F4], so for each y∈Y the preimage f−1({y}) is compact by [F1].

1.4givenF1F3F6

Restrictions to closed subspaces are proper. Let S⊆X be closed, and give S its subspace topology. For a compact K⊆Y one has (f∣S)−1(K)=f−1(K)∩S. The set f−1(K) is compact by [F1], and f−1(K)∩S is closed in the compact space f−1(K) because S is closed in X; hence f−1(K)∩S is a compact subset of f−1(K) by [F3]. Its topology as a subset of S agrees with its topology as a subset of X by [F6], so (f∣S)−1(K) is a compact subset of S and f∣S is proper by [F1].

1.5givenF1F2F3F4F5F6F7

Properness is preserved by base change along a map into a Hausdorff space. Let K⊆Y′ be compact and put K0=u[K], a compact subset of Y by [F2]. Then f−1(K0) is compact by [F1], so the product f−1(K0)×K is a compact subspace of X×Y′ by [F4]. The pairing (f,u):X×Y′→Y×Y is continuous by [F7], and ΔY is closed in Y×Y by [F5]; hence X×YY′=(f,u)−1(ΔY) is closed in X×Y′, and its intersection with the subspace f−1(K0)×K is closed in that compact space. Moreover this intersection is exactly (f′)−1(K): a pair (x,y′) lies in X×YY′ with y′∈K precisely when f(x)=u(y′) and y′∈K, and then u(y′)∈K0 forces x∈f−1(K0), so (f′)−1(K)=X×YY′∩(f−1(K0)×K). By [F3] this is a compact subset of the product X×Y′, and by [F6] its topology as a subset of X×YY′ agrees with that as a subset of X×Y′; hence f′ is proper by [F1].

1.6givenF1F2F3F8

Proper maps into locally compact Hausdorff spaces are closed. Let f:X→Y be proper with Y locally compact Hausdorff, let F⊆X be closed and let y∈Y∖f[F]. By [F8] there is an open neighbourhood V of y whose closure V‾ is compact. Then f−1(V‾) is compact by [F1] and F∩f−1(V‾) is closed in it, hence compact by [F3]; its image under the continuous map f is f[F]∩V‾, which is compact by [F2] and closed in Y by [F8]. Therefore V∖(f[F]∩V‾) is open in Y, contains y because y∈V and y∉f[F], and is disjoint from f[F]: a point of it lies in V, hence cannot lie in f[F]∖V‾, and it is disjoint from f[F]∩V‾ by construction. Hence f[F] is closed.

2.1givenstep 1.1step 1.2step 1.3step 1.4step 1.5step 1.6∎

The six stability properties are established: claim 1 by step 1.1, claim 2 by step 1.2, claim 4 by step 1.3, claim 5 by step 1.4, claim 3 by step 1.5 and claim 6 by step 1.6. Every argument uses only the preimage form of properness [F1], the preservation of compactness under continuous maps [F2], closed-in-compact compactness [F3], finite products of compact spaces [F4], the closed-diagonal criterion [F5], the intrinsic subspace conventions [F6], the initial product topology [F7] and the locally compact Hausdorff basis of compact neighbourhoods [F8]; none of these selects points or open sets from an infinite family, so no choice principle is used.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Derived operations on analytic field sheaves

Definition

Let k be a field and let X,Y be locally compact Hausdorff spaces (Locally compact topological space: every point has a compact neighbourhood; and what this says in a metric space, Hausdorff space: distinct points have disjoint open neighbourhoods; every metrizable space is Hausdorff and the indiscrete topology on two points is not). Write Sh⁡k(X) for the category of sheaves of k-vector spaces (A sheaf on a topological space, Sheaf cohomology as right derived global sections) viewed as modules over the constant ring sheaf kX, and form derived categories in those sheaves (Derived category of an abelian category). For a continuous map f:X→Y, set f∗=f−1 (Inverse image presheaf and inverse image sheaf) and define f!F on an open V⊆Y to consist of the sections of F on f−1(V) whose closed support (Support of a section of a sheaf, Compactly supported global sections and their right derived functors) maps properly to V, properness being that of Proper continuous maps of topological spaces together with its stability lemma Basic stability of proper continuous maps. Under AC (The Axiom of Choice), Rf∗ and Rf! are computed on bounded-below k-linear injective replacements. Tensor means the signed total tensor over k (Tensor product of sheaves of modules) and internal derived Hom is the injective construction of Derived tensor–Hom and inverse/direct image for field sheaves (The internal Hom sheaf of two module sheaves).

Suppose the uniform all-fibre-open cohomological dimension bound of Bounded fibrewise-soft resolutions from a uniform all-sheaf dimension bound holds, and choose its finite resolution kX→K; for an open U⊆X put KU=jU!jU∗K (Extension by zero for abelian sheaves on an open subspace, c-soft sheaves). For a bounded-below injective complex I on Y, define the exceptional inverse-image candidate to be the sheaf complex with open sections Hom⁡Y∙(f!KU,I). Exceptional inverse image as the adjoint of derived proper-support image ↗ proves that this candidate is a sheaf complex and is independent of the chosen resolution. Set ωX=aX!k and DXA=RHom(A,ωX) when the terminal map satisfies the same finite-dimensional hypotheses. Constructible restriction is a separate geometric theorem; none of these definitions asserts a generic six-functor axiom.

The prescriptions are those of Verdier's exposé: f∗ is the exact inverse image, f! is the sheaf of sections whose closed support is proper over the base — a presheaf by the local nature of properness, whose sheaf property for every sheaf is proved by the exceptional-adjoint theorem recorded in justified_by — Rf∗ is derived from the direct image Direct image of a sheaf along a continuous map on bounded-below injective replacements, and the terms of the chosen finite resolution are fibrewise c-soft, not necessarily c-soft on X. Their acyclicity for f! follows from the fibre formula in its well-definedness theorem; the restricted fibre sheaves are compact-support acyclic by Compact-support acyclicity of c-soft sheaves. The ordinary-operations lemma also supplies the flasque acyclicity (Flasque sheaf) used to compare resolutions. AC supplies the injective replacements and finite fibrewise-soft resolution; the proper-support sections, extension by zero and the exceptional candidate itself are choice-free constructions. Cohomological shifts obey Hi(A[r])=Hi+r(A) in the convention of Derived category of an abelian category.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Exceptional inverse image as the adjoint of derived proper-support image

Statement

Assume AC (The Axiom of Choice) and the uniform all-fibre-open dimension bound d of Bounded fibrewise-soft resolutions from a uniform all-sheaf dimension bound. For a continuous map f:X→Y of locally compact Hausdorff spaces, the proper-support functor f! of Derived operations on analytic field sheaves is left exact, satisfies (Rf!A)y≃RΓc(Xy,A∣Xy) on every fibre, and maps bounded derived categories into bounded derived categories. Its exceptional inverse-image candidate f! represents the natural Hom transposition Hom⁡(Rf!A,B)≃Hom⁡(A,f!B) for bounded source objects A and bounded-below target objects B, and has lower cohomological bound inf⁡(B)−d. When the later geometric theorem proves that f! preserves the bounded constructible subcategory, this becomes the right adjunction Rf!⊣f! there; on that bounded constructible specialization the adjunction unit, counit and resolution comparisons satisfy the triangle, identity and cocycle laws. The proper-support evaluation constructs the counit directly. The proof constructs this adjoint from a finite fibrewise-soft resolution and generator complexes, independently of Verdier biduality; it does not infer an unrestricted D+-source adjunction from the bounded-source generator argument.

Facts & Assumptions

Given: AC; a continuous map f:X→Y of locally compact Hausdorff spaces; a field k; the uniform all-fibre-open cohomological dimension bound d; and the operations, resolution datum kX→K and exceptional candidate of Derived operations on analytic field sheaves.

[F1]

The operations are those of Derived operations on analytic field sheaves: f! consists of sections whose closed support maps properly to the base, KU=jU!jU∗K for the chosen finite resolution kX→K in degrees 0,…,d, HK(I)(U)=Hom⁡Y∙(f!KU,I) for a bounded-below injective complex I on Y, and Rf∗, Rf! are computed on bounded-below k-linear injective replacements (Bounded below complexes admit injective replacements).

[F2]

The ordinary-operations lemma gives exactness of f∗, the tensor–Hom adjunction, flasque acyclicity for sections over every open set, and the fact that bounded-below flasque complexes compute their derived sections: for a bounded-below complex of flasque sheaves the natural map from its ordinary section complex to the derived sections over any open is a quasi-isomorphism (Derived tensor–Hom and inverse/direct image for field sheaves).

[F3]

Sections near a compact set are sections on it: for compact K in a locally compact Hausdorff space the canonical map lim→⁡U⊇KF(U)→Γ(K,i−1F) is an isomorphism, and pairwise disjoint compact pieces of sections can be glued over a neighbourhood of their union with the zero section prescribed on a disjoint compact set (Sections on a compact set extend to a neighbourhood).

[F4]

A k-sheaf is c-soft when its sections over compact subsets extend to global sections; flasque implies c-soft, and c-softness is stable under restriction to open and closed subsets and extension by zero; arbitrary direct sums used below are c-soft because a section over a compact set involves finitely many summands locally, hence only finitely many after a finite compact cover, each of which extends by c-softness (c-soft sheaves).

[F5]

Compact-support acyclicity: injective k-sheaves are flasque by [F2] and hence c-soft by [F4]; c-soft sheaves have Hcq=0 for q>0; sections over compact sets extend to compactly supported global sections with support in a prescribed neighbourhood; and in a short exact sequence with c-soft kernel and middle term the compactly supported sections are short exact (Compact-support acyclicity of c-soft sheaves, Compactly supported global sections and their right derived functors).

[F6]

Γc is left exact and its right derived functors form a cohomological delta functor with long exact sequences; when the space is compact Γc=Γ and Hcq=Hq (Compactly supported global sections and their right derived functors). Under DC, resolutions by acyclic objects compute the derived functors when all syzygies are in the supplied resolution domain (The acyclic-resolution theorem for right derived functors); here that domain is the whole field-sheaf category, and the bounded-complex comparison follows by filtering its finitely many degrees.

[F7]

The uniform all-fibre-open bound d gives every k-sheaf on X a bounded resolution 0→A→K0→⋯→Kd→0 whose terms are c-soft on every fibre, the constant sheaf in particular has such a resolution; the supplier does not assert a bounded-complex compatible resolution, and none is needed below (Bounded fibrewise-soft resolutions from a uniform all-sheaf dimension bound).

[F8]

Closed inclusions are proper; proper maps are stable under composition and Hausdorff base change and have compact fibres. In addition: if f is proper every section's support maps properly to the base, while properness of the support map into a locally compact Hausdorff space forces that support to be closed in the base (Proper continuous maps of topological spaces, Basic stability of proper continuous maps).

[F9]

Extension by zero j! along an open inclusion is exact and satisfies Hom⁡X(j!H,G)≅Hom⁡U(H,j−1G); supports of sections are closed in their domain of definition and restrict compatibly (Extension by zero is left adjoint to restriction and is exact on abelian sheaves, Extension by zero for abelian sheaves on an open subspace, Support of a section of a sheaf).

[F10]

Bounded-above complexes of abelian sheaves admit canonical bounded-above resolutions by flat sheaves with stalk-surjective quasi-isomorphism, and bounded-above flat complexes are K-flat; the construction uses the generators jU!ZU (Flat resolutions of abelian sheaves and K-flatness of bounded-above flat complexes).

[F11]

The derived Hom of Derived hom in the bounded setting is computed by Hom⁡∙(A,IB) for bounded A and bounded-below injective models IB, with the cochain cone convention of Derived category of an abelian category; flasque k-sheaves are acyclic also in the abelian-group sense (Flasque abelian sheaves are Γ-acyclic, Flasque sheaf).

Proof

technique · direct
1.1givenF8F9F12

The proper-support functor is a sheaf, additive and left exact; for proper f it equals f∗, for an open inclusion it is extension by zero, and for a closed inclusion it is direct image. For a sheaf A and open V⊆Y let f!A(V) be the set of s∈A(f−1V) whose support maps properly to V, with restrictions inherited from A; these are well defined because for open W⊆V the support of s∣ is Supp⁡(s)∩f−1W by [F9], and the preimage of a compact subset of W under Supp⁡(s)∩f−1W→W is the corresponding preimage under the proper map Supp⁡(s)→V. Additivity is evident from linearity of restriction and gluing. For the sheaf property let V=⋃Vα and let sα be compatible sections; they glue to s∈A(f−1V) with Supp⁡(s)⊆⋃αSupp⁡(sα) and s∣=sα over f−1Vα. If L⊆V is compact, cover L by finitely many open W1,…,Wn with compact closures Wj‾ contained in some Vα(j), possible by [F12]; then Supp⁡(s)∩f−1(L)⊆⋃j(Supp⁡(sα(j))∩f−1(Wj‾)), a finite union of compact sets because Wj‾⊆Vα(j) is compact and Supp⁡(sα(j))→Vα(j) is proper; and Supp⁡(s)∩f−1(L) is closed in f−1(L) by [F9], so it is a finite union of closed subsets of these compacta, hence compact by [F12]. Thus f!A is a sheaf. Kernels are computed sectionwise, so (f!ker⁡φ)(V)=ker⁡(f!φ)(V) and f! is left exact. If f is proper, every support map is proper by [F8], so f!=f∗. If f=j:U↪Y is an open inclusion, properness of Supp⁡(s)→V is equivalent to closedness of Supp⁡(s) in V by [F8], which is exactly the extension-by-zero condition of [F9]; if f=i:Z↪Y is a closed inclusion then Supp⁡(s) is closed in V∩Z, hence in V, and i!=i∗.

1.2givenF8F9F12

The fibre comparison is injective. Restriction defines (f!A)y→Γc(Xy,A∣Xy): a representative s∈f!A(V) has support S proper over V, so S∩Xy is compact. If its restriction to Xy is zero, every germ sx for x∈Xy is zero, and therefore S∩Xy=∅. The proper map S→V is closed by [F8], so f(S) is closed in V and avoids y. On the open neighbourhood V∖f(S) the section vanishes identically. Its germ in (f!A)y is consequently zero.

1.3givenF3F8F9F12

The comparison map is surjective, hence an isomorphism. Let t∈Γc(Xy,A∣Xy) have compact support Q. By [F3], its restriction near Q is represented by an ambient section s0 on a neighbourhood W of Q. Shrink W so that s0∣W∩Xy=t∣W∩Xy: equality holds near every point of Q and its equality locus is open in the fibre, so an ambient neighbourhood of the compact Q can be chosen inside it. By [F12] choose an open W0 containing Q with compact closure contained in W, and put C=W0‾∖W0. Since C∩Xy is disjoint from Q, the germs of s0 there are zero. Its zero locus is open, so choose an open N⊆W containing C∩Xy on which s0=0. The compact set C∖N is disjoint from Xy; its continuous image is compact and hence closed in the Hausdorff Y, and does not contain y. Let Y0=Y∖f(C∖N), an open neighbourhood of y. Then C∩f−1Y0⊆N. On f−1Y0 glue s0 on W0∩f−1Y0 with zero outside W0‾, using the open vanishing neighbourhood N∩f−1Y0 of the collar to cover and match across it. The resulting section s restricts to t on the fibre and has support contained in W0‾∩f−1Y0. For every compact L⊆Y0, Supp⁡(s)∩f−1L is closed in the compact W0‾, so it is compact. Thus s∈f!A(Y0) represents t. No extension of a section from Q∪(C∩Xy) to all of W0‾ has been assumed. The comparisons are natural under restriction.

2.1givenF5F7step 1.3F6

Fibrewise acyclicity and the fibre formula. If A is injective then A is c-soft by [F5], and restriction of a c-soft sheaf to a closed subspace is c-soft by [F4]; a fibre Xy=f−1({y}) is closed because {y} is closed in Y, so A∣Xy is c-soft and hence Γc-acyclic in positive degrees by [F5]. An injective resolution of A on X restricts to a Γc-acyclic resolution of A∣Xy, so (Rf!A)y≃RΓc(Xy,A∣Xy) by the fibre formula of step 1.3 and exactness of stalks, which evaluates the derived functor on the fibre. The uniform bound d of [F7] makes every fibre Xy have Hcq(Xy,−)=0 for q>d, so Rf! has cohomological dimension at most d fibrewise and maps Db(X) to Db(Y): a bounded complex has bounded fibrewise cohomology, and the fibre formula exhibits Rf! of it as fibrewise compact-support cohomology concentrated in bounded degrees. A fibrewise c-soft sheaf is f!-acyclic by the same stalk computation. The same formula shows that f! commutes with small direct sums: a compactly supported section of a sheaf coproduct involves finitely many summands locally, and a finite cover of its compact support gives one finite set of summands; its coordinate sections are compactly supported. Thus Γc(Xy,⨁Ai∣Xy)=⨁Γc(Xy,Ai∣Xy), and the canonical direct-sum comparison is an isomorphism on stalks.

3.1givenF1F4F5F7F9step 2.1

The complex HK(I) is a termwise flasque sheaf complex with lower bound inf⁡(I)−d. Let I be a bounded-below injective complex on Y and K the finite fibrewise-soft resolution of [F7] in degrees 0,…,d, with KU=jU!jU∗K for open U⊆X. For an open cover U=⋃Ui of an open U, the augmented extension-by-zero Čech complex of KU is exact on stalks: at a point x∈U, the simplex of indices i with x∈Ui has the elementary contracting homotopy that adds a fixed such index, and on U the complex resolves kU columnwise. Each term is a direct sum of sheaves jW!(K∣W), which are fibrewise c-soft by [F4] because K is, and hence f!-acyclic by step 2.1. The cycles of the augmented complex are again f!-acyclic: dimension shifting along its short exact cycle sequences identifies Rqf! of a cycle with Rq+rf! of an earlier cycle for every r, and choosing r>d makes the latter vanish by step 2.1; consequently f! carries the resolution to an exact complex. Since each Iq is injective, Hom⁡Y(−,Iq) preserves exactness, and the first three terms give the sheaf equalizer for the presheaf U↦Hom⁡Y∙(f!KU,I), so HK(I) is a sheaf complex. For V⊆U the map KV→KU is monic and f! is left exact by step 1.1, so f!KV→f!KU is monic; injectivity of Iq then makes the restriction maps of HK(I) surjective in every total degree, because Hom⁡Y(−,Iq) sends a monomorphism to a surjection and finite products over the finite K-degree preserve surjectivity. Thus HK(I) is termwise flasque. Its degree-n term is a finite product of terms Hom⁡Y(f!KU,Ip+n) with p∈[0,d], so it vanishes for n<inf⁡(I)−d.

4.1givenF1F2F9step 2.1step 3.1

The generator comparison is a derived comparison. Put EU=jU!kU. Exact extension by zero [F9] and the resolution kX→K give a quasi-isomorphism EU→KU, whose terms are f!-acyclic by step 2.1; hence f!KU represents Rf!EU. By construction, Hom⁡X∙(EU,HK(I))=Γ(U,HK(I))=Hom⁡Y∙(f!KU,I). The bounded-below flasque complex HK(I) computes derived sections by [F2], so this is the natural identification RHom⁡X(EU,HK(I))≃RHom⁡Y(Rf!EU,I). Its maps are evaluation and restriction; H0 gives the corresponding derived morphism groups.

5.1givenF1F2F4F7F9step 2.1step 3.1step 4.1constructalgebra

A field-linear generator model. For a k-sheaf E put G(E)=⨁(U,s), s∈E(U)EU, mapping the generator 1 to s; this is stalk-surjective. Starting above the upper degree of bounded A, construct M downwards: once Mn+1 and its map α to An+1 exist, take Mn=G(Zn) for Zn={(a,m)∈An⊕Mn+1:dAa=αm, dMm=0}, and use the two projections for αn and dMn. These equations give dM2=0 and dAα=αdM. Every stalk cycle (a,m) in Cone⁡(α) is the boundary of (0,l) for a lift l of (a,−m)∈Zn, so M→A is a quasi-isomorphism. This supplies the k-linear construction rather than interpreting the abelian resolution [F10] as a field-linear resolution. The zigzag A←M→K⊗kM consists of quasi-isomorphisms: the bounded acyclic cone of kX→K is split-contractible at each vector-space stalk, and remains contractible after tensoring with M. The terms of K⊗M are direct sums of KU, hence fibrewise c-soft and f!-acyclic. Also f! commutes with these direct sums: its fibre-stalk formula reduces this to compactly supported sections, which use only finitely many summands on a finite cover of their compact support. Thus evaluation on summands gives Hom⁡X∙(M,HK(I))=Hom⁡Y∙(f!(K⊗M),I).

5.2givenF1F4F8F9step 2.1step 3.1step 4.1algebra

The proper-support counit. For open V⊆Y and closed Q⊆f−1V proper over V, write KQ=iQ∗iQ∗K. The sequence 0→Kf−1V∖Q→Kf−1V→KQ→0 is stalk-exact. Its kernel terms are fibrewise c-soft, so applying f! is exact by step 2.1; applying Hom⁡Y(−,Iq) is exact by injectivity. Taking the kernel of restriction to the complement of Q therefore identifies the complex of Q-supported sections of HK(I) with Hom⁡V∙(f!KQ,I∣V). Since Q→V is proper, f!KQ=f∗KQ. Precomposition with kV→f∗kQ→f∗KQ, the constant-section unit followed by coaugmentation, gives a map to I(V). These maps agree when Q is enlarged and when V is restricted, so they assemble to f!HK(I)→I. Its terms are f!-acyclic: they are flasque by step 3.1, hence globally c-soft and c-soft on the closed fibres by [F4], so step 2.1 applies. This is the derived counit; on each generator its transposition is the evaluation of step 4.1.

6.1givenF2F6F7F11step 2.1step 4.1step 5.1algebra

The generator model computes the required derived comparison. If an acyclic complex has f!-acyclic terms, its cycle sequences give Rqf!Zn+1≅Rq+1f!Zn for q≥1. Iterating backwards more than d times kills these groups by step 2.1. All cycles are therefore f!-acyclic, the cycle sequences remain exact under f!, and the image complex is acyclic. For the bounded-above model K⊗M of bounded A, the same argument in its acyclic low-degree tail shows that a sufficiently low canonical truncation has an f!-acyclic first term. It is then a bounded f!-acyclic model of A, so ordinary acyclic-resolution comparison identifies f!(K⊗M) with Rf!A. Let HK(I)→J be a bounded-below injective replacement. Step 4.1 and flasque acyclicity make Hom⁡X∙(EU,HK(I))→Hom⁡X∙(EU,J) a quasi-isomorphism. Products of these comparisons are exact over k under AC; the Hom bicomplex of bounded-above M and bounded-below targets has finite diagonals, so its total comparison is a quasi-isomorphism. Finally K-injectivity of J identifies Hom⁡X∙(A,J) with Hom⁡X∙(M,J) up to quasi-isomorphism. Together with step 5.1 this proves RHom⁡X(A,HK(I))≃RHom⁡Y(Rf!A,I).

7.1givenF1F2step 3.1step 6.1

The Hom transposition, the adjoint candidate and its lower bound. Taking I=IB a bounded-below injective replacement of a bounded-below complex B with IBn=0 for n<inf⁡(B) by [F1], step 6.1 gives RHom⁡X(A,HK(IB))≃RHom⁡Y(Rf!A,IB), and H0 yields the natural exceptional Hom transposition Hom⁡(Rf!A,B)≃Hom⁡(A,f!B) for bounded A, with f!B represented by HK(IB). By step 3.1 the representative has lower bound inf⁡(IB)−d=inf⁡(B)−d, so f! has lower cohomological bound inf⁡(B)−d. The construction uses the chosen finite resolution and the chosen injective model; any two choices differ by the unique identity-transposition comparisons induced by the resolution comparisons of [F2], so the transposition is well defined.

8.1givenF1F2step 4.1step 6.1step 7.1step 5.2∎

Coherence on the bounded constructible specialization. Step 6.1 gives the Hom transposition and the counit of step 5.2 supplies its evaluation map. When the later geometric theorem proves bounded constructibility of f!, this is a full adjunction Rf!⊣f! on that subcategory. Its unit is the transpose of the identity, and inverse transposition proves both triangle identities. For two choices of K or injective models, the identity transposition gives their comparison; bounded open generators and all their shifts detect its isomorphism through derived sections. Uniqueness of transposition gives the identity and cocycle laws. No unrestricted D+-source adjunction is inferred.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Support localization, base change, projection and trace calculus

Statement

Under the preceding locally compact and finite-dimensional hypotheses, u∗Rf!≃Rf!′v∗ for Cartesian squares; proper maps have Rf!=Rf∗, giving proper ordinary base change. Bounded tensor factors satisfy the projection formula. Proper-support direct images compose canonically. On the bounded constructible specialization proved by the subsequent local duality construction, exceptional adjoints also compose canonically; their units, counits and traces respect composition and Cartesian pasting. For a closed inclusion i:Z→X with complement j, i! is the derived functor of i−1ΓZ, and both localization triangles hold. Locally, RHom(i∗kZ,A) is RΓZA. This lemma does not assume or claim extraordinary dual base change before constructible biduality.

Facts & Assumptions

Given: AC; locally compact Hausdorff X,Y with a continuous map f:X→Y satisfying the uniform all-fibre-open dimension bound d; the operations of Derived operations on analytic field sheaves; a Cartesian square X′→vX over Y′→uY with f′:X′→Y′; and the conventions of Exceptional inverse image as the adjoint of derived proper-support image.

[F1]

f! is a left exact sheaf functor with (Rf!A)y≃RΓc(Xy,A∣Xy), Rf!:Db→Db, and right adjoint candidate f! representing Hom⁡(Rf!A,B)≃Hom⁡(A,f!B) for bounded A and bounded-below B, with lower bound inf⁡(B)−d (Exceptional inverse image as the adjoint of derived proper-support image, Derived operations on analytic field sheaves).

[F2]

The ordinary operations: exactness of f∗, tensor–Hom adjunction and dg identities, flasque acyclicity, and bounded-below flasque complexes computing derived sections (Derived tensor–Hom and inverse/direct image for field sheaves, The internal Hom sheaf of two module sheaves, Tensor product of sheaves of modules).

[F3]

c-soft sheaves are Hcq-acyclic in positive degrees, injectives are c-soft, and c-softness is preserved by restriction to open and closed subsets, extension by zero and finite direct sums; small direct sums have the same property by the finite-coordinate compact-cover argument in the exceptional-adjoint supplier; sections over compact sets extend to compactly supported global sections supported in a prescribed neighbourhood (Compact-support acyclicity of c-soft sheaves, c-soft sheaves, Compactly supported global sections and their right derived functors).

[F4]

Proper f satisfies f!=f∗ and every support maps properly to the base; proper maps are stable under composition and base change; a proper map into a locally compact Hausdorff space is closed and has compact fibres (Proper continuous maps of topological spaces, Basic stability of proper continuous maps).

[F5]

Extension by zero j! is exact and left adjoint to j−1, and there is a short exact sequence 0→j!(F∣U)→F→i∗(F∣Z)→0 for an open U with closed complement Z; a sheaf whose stalks vanish off a closed subset Z is the pushforward of its restriction to Z (Extension by zero is left adjoint to restriction and is exact on abelian sheaves, Extension by zero for abelian sheaves on an open subspace, Extension by zero and the closed complement: a short exact sequence, A sheaf with no stalks off a closed subset is a pushforward).

[F6]

Sections near a compact set glue: for compact K the canonical map lim→⁡U⊇KF(U)→Γ(K,i−1F) is an isomorphism, and sections prescribed on finitely many pairwise disjoint compact sets and the zero section on another disjoint compact set glue over a neighbourhood (Sections on a compact set extend to a neighbourhood).

[F7]

Inverse and direct image: (f−1G)x≅Gf(x); f−1⊣f∗; a morphism is an isomorphism exactly when it is on stalks (Inverse image presheaf and inverse image sheaf, Direct image of a sheaf along a continuous map, Inverse image is left adjoint to direct image on sheaves, A sheaf on a topological space, Sheaf cohomology as right derived global sections).

[F8]

Derived Hom is computed on bounded-below injective models with the cochain convention of Derived hom in the bounded setting and Derived category of an abelian category; flasque sheaves are acyclic (Flasque sheaf).

Proof

technique · direct
1.1givenF2F5F7F8

Closed-support localization. Let i:Z↪X be a closed inclusion with open complement j, and put ΓZF=ker⁡(F→j∗j∗F). Its stalks vanish off Z by [F7], so it is i∗i−1ΓZF by [F5], and every map i∗A→F lands in this kernel because j∗i∗=0; hence i0!F=i−1ΓZF is right adjoint to the exact functor i∗. It preserves injectives, so its derived functor is the right adjoint of i! on the derived level where the adjoint exists. The stalk-exact sequence of [F5] gives the first localization triangle j!j∗A→A→i∗i∗A→j!j∗A[1]. For a flasque injective I the map I→j∗j∗I is surjective on sections over every open set by [F8] and [F5] with kernel ΓZI, and j∗ preserves injectives because j! is exact; hence i∗i!A→A→Rj∗j∗A→i∗i!A[1] is the second localization triangle. Finally the tensor identity H⊗ki∗kZ=i∗i∗H holds by the stalk formula, and tensor–Hom transposition [F2] together with the closed adjunction gives Hom⁡(H,RHom(i∗kZ,A))≅Hom⁡(i∗i∗H,A)≅Hom⁡(H,i∗i!A) locally on each open; hence RHom(i∗kZ,A)≅RΓZA, the actual closed-support object used in the diagonal Thom calculation.

1.2givenF1F3F4

Cartesian base change. In the Cartesian square, inverse image of an injective I on X restricts on every fibre of f′ to the restriction of I on the canonically homeomorphic original fibre, so it is fibrewise c-soft by [F3] and f′-acyclic by [F1]; the section restriction map u∗f!I→f!′v∗I is an isomorphism on stalks by the fibre-stalk computation of [F1]. Applying this to an injective complex proves u∗Rf!≃Rf!′v∗. If f is proper then f′ is proper by [F4], and using f!=f∗ and f!′=f∗′ this is ordinary proper base change u∗Rf∗≃Rf∗′v∗. These are the canonical restriction/exchange maps, not arbitrarily chosen isomorphisms.

1.3givenF3F4F6

Independent verification for proper maps. For proper f the same identification can be checked through ordinary compactly supported cohomology: restriction of injectives to compact fibres is c-soft and Γ-acyclic by [F3], compact-neighbourhood sections are delivered by [F6], and preimage neighbourhoods are cofinal because f is closed by [F4], with V=Y∖f(X∖U) for an open U; this recovers the exchange as the canonical derived adjunction map of ordinary proper base change.

1.4givenF1F2F3

The projection formula. For bounded A choose a bounded-below injective replacement A→I. On each fibre Xy the restriction I∣Xy is c-soft by [F3], and f∗B restricts to the constant complex By; tensoring with each term By is a direct sum of copies of I∣Xy, hence c-soft by [F3], and Γc(E⊗kV)=Γc(E)⊗kV because a compactly supported section has a finite local tensor presentation and a finite compact cover uses only finitely many coefficient vectors. This proves the canonical map Rf!A⊗kB→Rf!(A⊗kf∗B) is an isomorphism on stalks, since both sides compute the fibrewise compact-support cohomology of [F1]; the map is the adjoint of the tensor of the f∗⊣Rf∗ counit with the identity, equivalently the sectionwise map s⊗b↦s⊗f∗b, and the signed totalization of [F2] supplies the bounded-complex version.

1.5givenF1F3F4F6

Composition of proper-support direct images. The support definition gives (gf)!=g!f!: a section whose support is proper over Z has support proper over Y because the inverse image of a compact L⊆Y is a closed subset of the inverse image of the compact g(L) by [F4], and conversely a support locally proper over Y with proper image over Z has compact preimages of compacta in Z. The support of f!s is f(Supp⁡s) by the fibre-stalk identity [F1] and local properness, so the two descriptions are inverse. Direct images of globally c-soft sheaves by f! are c-soft: a section over a compact base set is represented on a neighbourhood by [F6], its total support over that compact set is compact by properness of the representing support, and enclosing it in a relatively compact W, the support-controlled c-soft extension of [F3] supported in W pushes forward and extends the original section over the compact set. Thus f! sends injectives to g!-acyclic sheaves, and applying the two ordinary support functors to an injective resolution gives the canonical derived composition R(gf)!=Rg!Rf!; the same comparison proves associativity for triples, the section support bijection being literally associative.

2.1givenF1step 1.5

Exceptional composition and traces on the bounded constructible specialization. If the represented right adjoints preserve the bounded constructible subcategory, so that the Hom formula of [F1] is a full adjunction there, then exceptional composition (gf)!≃f!g! is the unique adjoint transpose of the composition isomorphism of step 1.5. The traces trf:Rf!f!→id are the counits constructed in [F1], and under the composition identification trgf=trg∘Rg!(trf)g!: both sides have the same transpose, namely the identity of f!g!, as is checked with the two triangle identities. Thus traces compose without a new axiom, and identity units and counits are identity transpositions.

3.1givenF1F2step 1.2step 1.4step 2.1∎

Pasting coherence, and the boundary of the claim. The two composites of base-change maps in a pasted rectangle restrict a representative section to the same fibre section, hence agree before deriving, and resolution comparisons preserve that equality; projection formulas for composites both send pure tensors s⊗b to the same pullback tensor; projection/trace compatibility follows by transposition, both composites transposing to evaluation on s⊗b followed by the same counit. This proves the projection, base-change, unit and trace squares used in pull-push convolution, and the identities hold in the dg resolution model rather than being chosen independently on the triangulated level. The extraordinary proper exchange u!Rf∗≃Rf∗′v! is not claimed here: it is not a right-adjoint mate of the ordinary base-change isomorphism and its proof needs the biduality result of the subsequent local duality construction, which therefore is not assumed.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Constructible Verdier duality and complex orientation

Statement

Assume AC and the actual finite conical/triangulation and derived-incidence-model suppliers listed below. On the analytic space X of a separated finite-type complex algebraic variety, constructible complexes admit the fixed-space finite models required by those suppliers. Then ωX=aX!k and DX=RHom(−,ωX) preserve bounded constructible objects and the natural evaluation A→DXDXA is an isomorphism. On a smooth complex d-variety, the complex orientation normalizes ωX≃kX[2d]. For constructible maps for which the geometric supplier proves the stated bounded outputs, Rf∗DX≃DYRf!, DYRf∗≃Rf!DX, and f!≃DXf∗DY. Proper base change has the coherent extraordinary dual exchange u!Rf∗≃Rf∗′v!. The fixed-space triangulation is not an assumed globally simplicial model of f.

Facts & Assumptions

Given: AC; a separated finite-type complex algebraic variety X with its analytic topology; its terminal map aX:X→pt⁡; a field k; the fixed finite conical/triangulation and derived-incidence models supplied below; and the operations and exceptional adjoint of Derived operations on analytic field sheaves and Exceptional inverse image as the adjoint of derived proper-support image.

[F1]

ωX=aX!k and DX=RHom(−,ωX), with the exceptional adjoint Rf!⊣f! representing Hom⁡(Rf!A,B)≃Hom⁡(A,f!B) for bounded A and bounded-below B (Derived operations on analytic field sheaves, Exceptional inverse image as the adjoint of derived proper-support image).

[F2]

The support-and-trace calculus: closed-support localization, tensor–Hom transposition, RHom(i∗kZ,A)≅RΓZA, the projection formula and canonical composition of proper-support images (Support localization, base change, projection and trace calculus, Derived tensor–Hom and inverse/direct image for field sheaves, The internal Hom sheaf of two module sheaves).

[F3]

Dcb(X;k) is the full subcategory of Db(Sh⁡k(X)) of cohomologically bounded complexes whose cohomology is locally constant of finite rank along a finite smooth algebraic stratification (Algebraically constructible complexes of analytic field sheaves, Locally constant sheaves of finite rank).

[F4]

The finite models: on any separated finite-type complex variety, bounded algebraically constructible objects have bounded finite ordinary and compact-support cohomology, ordinary derived Hom and bounded constructible internal Hom, compatible with coefficient extension, by the actual conical/incidence models locally and finite affine-open descent globally (Complex algebraic constructible complexes have finite conical cochain models); and the derived-incidence model identifies the bounded derived category of finite face-diagrams with the full all-sheaf-derived constructible subcategory under the stated finite triangulation hypothesis, with injective cogenerators the sheaves supported on closed cells (Finite simplicial constructible sheaves have an actual derived incidence model).

[F6]

For a quasi-projective variety, the finite triangulation supplier gives a homeomorphism to ∣K∣∖∣D∣ with K finite and D a subcomplex (Algebraically constructible sheaves have finite compatible stratified and local cochain models). For every open W⊆∣K∣ and every sheaf E on W, Hcq(W,E)=0 for q>dim⁡K, without a constructibility hypothesis (A polyhedral dimension bound for every sheaf). Open extension by zero and closed pushforward preserve compact-support cohomology for arbitrary sheaves (Compact-support calculus along open and closed embeddings).

Proof

technique · direct
1.1givenF1F2F5F6algebra

The dualizing object and its local dual formula. Choose a finite affine-open cover V1,…,Vm of X and [F6]'s finite pairs (Ki,Di) for these quasi-projective pieces, applying the triangulation supplier to the zero constructible complex if necessary. Put d=max⁡idim⁡Ki. Every open of Vi is an open of ∣Ki∣, so [F6] bounds its compact-support cohomology above d for every sheaf. For any open U⊆X and arbitrary sheaf E on U, the finite augmented alternating extension-by-zero complex for the cover (U∩Vi) resolves E: at each stalk it is the augmented simplex complex on the indices containing that point. Its term in degree −p is the sum of the extensions of E from the (p+1)-fold intersections. Each intersection is open in a Vi, and [F6]'s open-support transport bounds the compact-support cohomology of every term above d. Applying derived compact sections to this finite resolution therefore gives only total degrees q−p≤d, proving Hcq(U,E)=0 for q>d. If X is empty take d=0. Moreover, every fibre F=f−1(y) of a continuous map to a Hausdorff space is closed in X. An open W⊆F has the form F∩O for an open O⊆X; the closed inclusion i:W↪O and [F6] give Hcq(W,E)=Hcq(O,i∗E)=0 for q>d. This verifies [F1]'s uniform all-sheaf, all-fibre-open hypothesis for the maps used here, including aX. Thus [F1] defines ωX=aX!k. For bounded A and open U, tensor–Hom transposition gives RΓ(U,DXA)=RHom⁡X(EU⊗kLA,ωX)≃RHom⁡k(RΓc(U,A∣U),k), where EU=jU!kU; the last comparison is the constructed terminal-map adjunction and open support composition [F1, F2]. Likewise RΓ(U,f!ωY)≃RHom⁡Y(Rf!EU,ωY)≃RHom⁡k(RΓc(U,k),k)≃RΓ(U,ωX). These are natural comparisons of global Hom complexes, compatible with restriction to opens. Open generators and their shifts detect cohomology stalks, so they prove dualizing transitivity f!ωY≃ωX without an unrestricted adjunction. Since k is injective over itself, the coefficient Hom is ordinary signed vector-space duality.

1.2givenF1F2

Orientation. On an oriented real n-manifold and a small contractible ball U, the compact-support cochains of k have a single cohomology class in degree n, normalized by the positive orientation generator: on a closed simplex this is the finite relative cellular complex (simplex,boundary), whose quotient is generated by the oriented top cell, and excision identifies a ball with this model. Local models imply ωX≃orX[n], and on a smooth complex d-variety the complex orientation trivializes orX, yielding ωX≃kX[2d]. This fixes the positive complex generator used by the oriented diagonal Thom classes.

1.3givenF1F4

Local systems on smooth strata. For a finite-rank local system L on a smooth stratum the same local model gives DXL≃L∨[n] with n the real dimension, and finite-dimensional vector-space evaluation gives L→∼DXDXL; truncation extends this to bounded locally constant complexes. An induction on strata must not silently assume that DD(j!L)=j!DD(L), since that involves boundary links; the fixed-space cellular construction below supplies that step.

2.1givenF4F3step 1.1

The incidence model locally. Cover the separated finite-type variety X by finitely many affine opens. On each such open V, which is quasi-projective, [F4] supplies a finite regular triangulation with a deleted subcomplex, subordinate to A∣V. This is a model of the fixed space V, without a globally simplicial presentation of any algebraic map. On V the derived-incidence supplier identifies bounded finite face-diagrams with the actual all-sheaf-derived constructible subcategory. Its injective cogenerators Iσ are the constant sheaves on the closed cell intersected with V, pushed forward along that relatively closed subset, and every finite diagram has a finite injective bar resolution by finite sums of them. Restriction of the support-defined duality to V is DV, by step 1.1 and the exact open-inclusion transitivity, so these local resolutions suffice to check the global evaluation map. No single finite deleted-boundary model of an arbitrary separated variety is required.

2.2givenF1F4step 1.2

Duals of the cogenerators. Write Cσ=σ‾∩X, let aσ:int⁡σ↪Cσ be open and iσ:Cσ↪X closed, and put Eσ=iσ∗aσ!k. This specifies the extension by zero for the locally closed cell, whose interior need not be open in X. The compact-support local dual formula of step 1.1 computes DXIσ directly: at an interior point of σ a small neighbourhood in σ‾ is an oriented real dim⁡σ-ball, whose compact-support cochains have the single positive generator in that degree; at a proper face τ<σ a small neighbourhood in σ‾ is a product [0,∞)c×Rr with c>0, whose compact-support cochain complex is acyclic because one half-line has the relative finite cellular model ([0,1],{1}) with boundary an isomorphism and tensoring that contraction with the other factors stays a contraction; at a point outside the closure the complex is zero. The positive orientation restricts to the interior coefficient, so DXIσ=Eσ[dim⁡σ]. Conversely, compact-support cochains of Eσ on a small star of a remaining face τ≤σ are the cochains of an open sector (0,∞)c×Rr≃Rdim⁡σ, with exactly the positive oriented generator in degree dim⁡σ, and restrictions between face-stars preserve it; hence DX(Eσ[dim⁡σ])=Iσ. The evaluation Iσ→DXDXIσ is ordinary finite-dimensional double-dual evaluation at the interior and has the same value at every face by these incidence identifications, so it is an isomorphism. These computations use finite relative cellular complexes and the actual sheaf/cochain comparison, not a guessed stalkwise duality.

2.3givenF4step 1.1step 1.2algebra

Algebraic constructibility of the dualizing object. On a controlled chart about a complex stratum S of dimension e, [F4] supplies a ball in S times a finite normal link-cone model. Applying the local formula of step 1.1 to constant coefficients gives the signed dual of its compact-support complex. The ball factor has its positive oriented class in degree 2e by step 1.2; the normal model is a bounded finite complex by [F4]. The controlled product identifications and their restriction maps therefore give a bounded finite coefficient complex, locally constant along S, with the ball's orientation shift. There are finitely many complex strata and affine charts, so their bounds are uniform. Hence ωX is bounded algebraically constructible. Now [F4]'s internal-Hom theorem applies to both A and ωX, proving bounded algebraic constructibility and coefficient-extension compatibility of DXA. This does not infer complex-stratum constructibility from a partition into real simplices alone.

3.1givenF4step 2.1step 2.2step 2.3

Biduality. On each affine open V of step 2.1, DV is contravariant exact and its double is covariant exact, with evaluation a natural transformation of exact functors. Apply evaluation to the finite local injective incidence resolution of A∣V: it is an isomorphism on every finite sum of cogenerators by step 2.2, and induction through the finitely many resolution cones proves A∣V→DVDV(A∣V) is an isomorphism. These are restrictions of the support-defined global evaluation, so its cone restricts to zero on the affine cover and is zero stalkwise; hence A→DXDXA is an isomorphism. Bounded algebraic constructibility and coefficient-extension compatibility were established in step 2.3. Refinement changes neither the support-defined duality nor evaluation, so the restriction and refinement comparisons have their identity and cocycle laws.

4.1givenF1F2step 1.1step 3.1∎

The duality exchange and the extraordinary proper exchange. After biduality, the internally localized exceptional adjunction with B=ωY gives Rf∗DXA≃DYRf!A: both sides represent the sheaf of derived morphisms from Rf!A to ωY, and the source-side object is RHom(A,f!ωY)=DXA by step 1.1. Applying this identity to DXA and applying DY gives DYRf∗A≃Rf!DXA by the double-dual isomorphism of step 3.1. Next Hom⁡(A,DXf∗DYB)≅Hom⁡(f∗DYB,DXA)≅Hom⁡(DYB,Rf∗DXA)≅Hom⁡(DYRf∗DXA,B)≅Hom⁡(Rf!A,B); these transpositions give a natural comparison f!B→DXf∗DYB, and applying its internally localized chain formula to the generators EU=jU!kU and their shifts shows that the two derived section complexes agree, so the comparison is an isomorphism on cohomology stalks. The right side is bounded constructible by the fixed-space duality and the geometric inverse-image models, proving that f! preserves bounded constructibility; the represented Hom formula is now a genuine adjunction on that category, and inverse transposition proves its triangle identities and fixes all comparison maps. This avoids falsely equating compact-support cohomology downstairs with compact-support cohomology upstairs for a nonproper map. Applying these actual dualities to the ordinary proper base change of the support-and-trace calculus gives the coherent extraordinary exchange u!Rf∗≃Rf∗′v!: the duality isomorphisms and evaluations are natural and respect the sectionwise pasting identities, so the exchange has the mate and trace compatibility needed by convolution.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Coherent six operations on analytic constructible field sheaves

Statement

Assume AC and the actual geometric and derived-incidence suppliers of the Verdier theorem. For analytic algebraic maps on its proved constructible subcategory, the six operations f∗,Rf∗,Rf!,f!,⊗kL,RHom have their three adjunctions, composition and identity comparisons, ordinary proper and extraordinary proper base change, projection formula, support localization, dualizing transitivity and trace. The comparisons commute with Cartesian pasting, composition of maps, evaluation and trace, giving a coherent pull-push product calculus. For a locally trivial product projection with finite-type smooth fibre F, the actual finite product cochain model gives Rpr∗pr∗B≃B⊗kC∗(F;k), with its ordinary unit. Smooth base change for Rf∗ holds whenever a supplied factorization f=p∘j has p proper and j an open inclusion with the conical local models below; proper maps already satisfy it. Their finite constructible models commute with extension of coefficient fields. Geometric output preservation is the exact geometric supplier, not an additional axiom.

Facts & Assumptions

Given: AC; separated finite-type complex algebraic varieties with their analytic topology and the analytic algebraic maps between them; a coefficient field k; the actual finite conical/triangulation and derived-incidence suppliers; and the operations of Derived operations on analytic field sheaves.

[F1]

The support-and-trace calculus: Cartesian base change u∗Rf!≃Rf!′v∗, proper equality Rf!=Rf∗, the projection formula for bounded tensor factors, canonical composition of proper-support images, closed-support localization triangles with RHom(i∗kZ,A)≅RΓZA, composition of traces and coherence under pasting, with no extraordinary dual base change assumed (Support localization, base change, projection and trace calculus).

[F2]

The constructible Verdier theorem: ωX=aX!k and DX are defined, evaluation A→DXDXA is an isomorphism on bounded constructible objects, ωX≃kX[2d] on a smooth complex d-fold, duality exchanges Rf∗, Rf! with D, one has f!≃DXf∗DY, and the coherent extraordinary proper exchange u!Rf∗≃Rf∗′v! holds on the bounded constructible specialization (Constructible Verdier duality and complex orientation).

[F3]

The ordinary operations and their adjunctions: exactness of f∗, f∗⊣Rf∗, the tensor–Hom adjunction and dg identities, flasque acyclicity (Derived tensor–Hom and inverse/direct image for field sheaves).

[F4]

The actual finite conical/incidence models give bounded finite ordinary and compact-support cohomology, ordinary derived Hom and bounded constructible internal Hom on bounded constructible objects, compatible with coefficient extension (Complex algebraic constructible complexes have finite conical cochain models); the derived-incidence model supplies the finite product incidence resolutions and the local star acyclicity and contractible-factor comparisons (Finite simplicial constructible sheaves have an actual derived incidence model), in particular for a product projection the locally constant factor cohomology is computed by the finite complex C∗(F;k) whose sections on contractible factors are acyclic (Locally contractible metric spaces have singular constant sheaf cohomology, Locally constant sheaves of finite rank).

Proof

technique · direct
1.1givenF1F3

The ordinary and support-locality clauses. Apply [F1] to the actual analytic maps. Its proper-support construction and injective models give the ordinary Cartesian exchange, the projection formula, both support localization triangles, canonical composition, and the units and traces of the adjunctions; the equalities are section-level identities carried through the same resolution comparisons, so no unrelated triangulated-category isomorphisms are selected. Together with [F3] this gives the adjunctions f∗⊣Rf∗ and Rf!⊣f! and the tensor–Hom adjunction on the bounded constructible specialization, their identity and composition comparisons, and the ordinary proper base change for proper maps.

1.2givenF1F2F3

The constructible and extraordinary clauses. Apply [F2] to the actual finite conical and incidence models. It supplies duality, orientation and the extraordinary exchange by dualizing the already canonical ordinary proper base-change map, and identifies dualizing transitivity and f!≃DXf∗DY; the tensor–Hom and inverse/direct-image adjunctions are the signed dg comparisons of [F3]. Each duality comparison is evaluation and transposition of the same support model, so functoriality carries the proved pasting and trace squares of [F1] to their dual squares; this yields the coherent six-operation interface on the bounded constructible subcategory whenever its listed geometric supplier proves preservation of that subcategory, which is thus the exact geometric input rather than an additional axiom.

1.3givenF1F4

The product projection. For a locally trivial projection V×F→V with finite-type smooth fibre F, use the finite product incidence resolution of [F4] on quasi-projective charts: tensor the finite base diagram resolution with the constant-diagram resolution of the fibre. For a general separated finite-type fibre, take a finite affine-open cover and its finite affine intersections; apply this construction to each intersection and totalize the finite ordinary Čech derived augmentation of [F4]. This assembles the same comparison with the finite complex C∗(F;k), without assuming a single triangulation of a non-quasi-projective fibre. Each realized term is supported on a product of closed cells minus the deleted faces, and its intersection with a small base star times F has contractible factors, so the independently proved constant-sheaf comparison [F4] makes it section-acyclic; applying sections gives the finite double complex B∣star⊗kC∗(F;k). These are natural restriction complexes, and they identify the actual derived image and its adjunction unit with B⊗kC∗(F;k) and B⊗(k→C∗(F;k)). For smooth base change across an open inclusion j, a local smooth chart is a product with a contractible ball; the boundary-star section model becomes the previous finite cochain model tensored with the ball cochains, whose augmentation is a quasi-isomorphism by an explicit contraction, so the canonical open-inclusion exchange is a stalkwise quasi-isomorphism. If the supplied map has a factorization f=p∘j with p proper and j an open inclusion, combining this exchange with the already constructed proper base-change map for p and its compatibility with composition identifies the composite with the canonical exchange for f; proper maps already satisfy smooth base change through [F1]. No existence of a compactification is inferred: the clause applies only to the stated supplied factorization.

2.1givenF2F4F5∎

Coefficient extension. For a field extension k→K use the geometric supplier's finite local cellular and cochain models [F4]. Scalar extension is exact on vector spaces, commutes with finite sums and cohomology, and commutes with duals of finite-dimensional terms, so ordinary inverse image, finite cochain pushforward, tensor, internal Hom and the support-defined Verdier duality comparisons extend scalars termwise; the exceptional comparison follows from f!=DXf∗DY of [F2], and composition and traces extend because their maps are those same finite evaluation and section comparisons. This argument does not assert that infinite products of injective stalks commute with scalar extension.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Actual analytic constructible sheaves admit coherent six operations

Statement

Assume AC (The Axiom of Choice) and let k be an arbitrary field. On separated finite-type complex varieties, the ordinary analytic sheaf operations restrict to bounded algebraically constructible complexes Dcb(−;k) (Algebraically constructible complexes of analytic field sheaves): f∗⊣Rf∗, Rf!⊣f!, the tensor/internal-Hom adjunction, coherent composition, support localization, ordinary and extraordinary proper base change, the projection formula and trace. Proper f satisfies Rf!=Rf∗. Verdier duality is involutive there and exchanges these adjoints. On a smooth complex d-fold ωX=kX[2d], and for a smooth map of relative dimension d, f!=f∗[2d]. Coefficient extensions commute with these operations on bounded constructible inputs. The bounded constructibility and finite coefficient models are those of Complex algebraic constructible complexes have finite conical cochain models and Finite simplicial constructible sheaves have an actual derived incidence model.

Facts & Assumptions

[F1]

The coherent six-operation theorem gives, on the proved constructible subcategory, the three adjunctions, composition and identity comparisons, ordinary and extraordinary proper base change, the projection formula, support localization, dualizing transitivity and trace, coherent under Cartesian pasting, composition, evaluation and trace, with proper equality Rf!=Rf∗, and with finite constructible models commuting with coefficient extension; geometric output preservation is the exact geometric supplier (Coherent six operations on analytic constructible field sheaves).

[F2]

The bounded constructibility and finite coefficient models are supplied by the finite conical and derived-incidence suppliers, for the stated finite collections, with coefficient-extension compatibility (Complex algebraic constructible complexes have finite conical cochain models, Finite simplicial constructible sheaves have an actual derived incidence model).

[F3]

The exceptional inverse image represents the Hom transposition on bounded sources and bounded-below targets and becomes the right adjoint Rf!⊣f! on the bounded constructible specialization (Exceptional inverse image as the adjoint of derived proper-support image).

[F4]

The constructible Verdier theorem gives involutive duality on bounded constructible objects, exchanging the adjoints, with ωX≃kX[2d] on a smooth complex d-fold and f!≃DXf∗DY, together with the coherent extraordinary proper exchange (Constructible Verdier duality and complex orientation).

[F5]

Dcb(X;k) consists of the bounded complexes whose cohomology is locally constant of finite rank along a finite smooth algebraic stratification (Algebraically constructible complexes of analytic field sheaves); AC is assumed (The Axiom of Choice).

Proof

technique · direct
1.1givenF1F2F3

The operations restrict to Dcb and satisfy the stated adjunctions and diagrams. The compact-support adjoint and the coherent six-operation suppliers [F1] construct the functors together with their adjunction units and counits, not merely unrelated isomorphisms: their composition, Cartesian exchange, projection, trace and localization diagrams are the asserted diagrams, including proper equality and the smooth local-product comparisons. The finite conical and incidence suppliers [F2] prove bounded constructibility and finite coefficient models for the stated finite collection, so the constructed operations restrict to Dcb; and [F3] supplies the right adjoint Rf!⊣f! on that specialization. All vector-space and injective constructions apply over an arbitrary field: the signed cochains use only 0 and the units ±1, so no characteristic hypothesis enters.

1.2givenF1F4F5

Proper equality and involutive duality. Proper f satisfies Rf!=Rf∗ by [F1]. The oriented stratified Verdier supplier [F4] gives the evaluation isomorphism and proves the natural evaluation A→DXDXA is an isomorphism on bounded constructible objects, so duality is involutive there; its adjunction construction gives the exchange of Rf∗, Rf! with the duality functors, and the extraordinary proper base change. All these maps use the same ordinary units, counits and orientation, so their coherence is the coherence of the actual construction and is compatible with the pasting and trace identities of [F1].

1.3givenF1F2F3F4

Smooth normalization and relative pullback. On a smooth complex d-fold the complex orientation identifies ωX≃kX[2d] by [F4]. Let a:X→Y be smooth of pure relative complex dimension d. Smooth analytic local charts identify it, locally on source and target, with pr⁡:W×V→V, where W is an oriented complex d-ball; V need not be smooth. For an open V′⊆V, the generator EW×V′=jW×V′!k has proper-support image EV′⊗kRΓc(W,k)=EV′[−2d], by the finite oriented ball relative-cochain model of [F2] and the support construction of [F1]. For any bounded constructible B on V, the ordinary product comparison of [F2] gives RΓ(W×V′,pr⁡∗B)=RΓ(V′,B) because W is contractible. Consequently RHom⁡(EW×V′,pr⁡∗B[2d])=RHom⁡(Rpr⁡!EW×V′,B). The orientation integration map Rpr⁡!pr⁡∗B[2d]→B transposes under [F3] to a comparison pr⁡∗B[2d]→pr⁡!B, and on these generators its Hom map is exactly the displayed cochain evaluation isomorphism, natural under shrinking product charts; the exceptional adjoint of [F3] identifies the same right-hand complex with RHom⁡(EW×V′,pr⁡!B). Product opens form a neighbourhood basis, so the comparisons identify derived sections and hence cohomology stalks, proving pr⁡!B≃pr⁡∗B[2d]. The positive complex orientation makes the local comparisons agree on overlaps, so they glue to a!≃a∗[2d] on all bounded constructible inputs. The orientation generator maps to 1∈k over every field, including characteristic two. This proves the relative formula on arbitrary constructible complexes rather than only on smooth local systems.

2.1givenF1F2F4∎

Coefficient extension. The finite local cellular and cochain models of [F2] have finite-dimensional terms and k-linear comparison maps, so extending scalars along k→K is exact, commutes with finite sums, cohomology and finite-dimensional duals, and carries the models of [F1] and [F4] to the corresponding models of the extended coefficient field; the exceptional comparison follows from f!=DXf∗DY, and composition and traces extend because their maps are the same finite evaluation and section comparisons. This does not assert that infinite products of injective stalks commute with scalar extension.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Ordinary middle perverse t structure, recollement and intersection complexes

Statement

Assume AC (The Axiom of Choice) and let k be any field. On Dcb(X;k) for a separated finite-type complex variety X, middle perversity is given on smooth complex strata S by the two inequalities Hq(iS∗A)=0 for q>−dim⁡CS and Hq(iS!A)=0 for q<−dim⁡CS. These inequalities define a bounded t structure (Canonical t structure on a derived category) on Dcb(X;k), independent of compatible refinement and gluing across algebraic open/closed decompositions. Verdier duality exchanges its two aisles. The heart is abelian and of finite length, and its simple objects are the intersection complexes j!∗L[dim⁡CZ] of simple finite-rank local systems L (Locally constant sheaves of finite rank) on dense smooth opens of irreducible supports Z. Open restriction is perverse exact. Heart objects have no negative Hom, and the endomorphism ring of a simple object is a division k-algebra. This is the ORDINARY category: no equivariant category, no normalization, no semisimplicity and no noncharacteristic theorem is used or asserted.

Facts & Assumptions

Given: AC; a separated finite-type complex variety X with its analytic topology and the compatible finite smooth algebraic stratifications supplied by the finite conical models; a coefficient field k; the bounded constructible category Dcb(X;k) with its ordinary six operations; the conventions of Derived category of an abelian category and Derived hom in the bounded setting.

[F1]

The six operations on Dcb: the three adjunctions f∗⊣Rf∗, Rf!⊣f! and tensor–Hom, coherent composition, support localization (both localization triangles for an open inclusion j with closed complement i), ordinary and extraordinary proper base change, projection and trace; proper equality; involutive Verdier duality exchanging the adjoints; and coefficient-extension compatibility (Actual analytic constructible sheaves admit coherent six operations).

[F2]

A t-structure on a triangulated category is a pair (T≤0,T≥0) satisfying shift-closure, vanishing of Hom⁡(T≤0,T≥1) and existence of truncation triangles; its heart is T≤0∩T≥0. The ordinary cohomological pair is a t-structure on the bounded derived category (The canonical pair is a t structure, Canonical t structure on a derived category, Derived category of an abelian category).

[F3]

Cone triangles satisfy the octahedral axiom, and octahedral completion relates the cones of composable maps and produces the standard truncation triangles used in gluing t-structures (Cone triangles satisfy the octahedral axiom, The octahedral axiom gives a triangle relating the cones of f, g, and gf).

[F4]

On the smooth strata the finite conical models give the local dimension computations and the bounded constructibility of the operations, per finite compatible stratifications (Complex algebraic constructible complexes have finite conical cochain models).

[F5]

A sheaf is locally constant of finite rank when it is locally isomorphic to a finite power of the constant sheaf, with no global trivialization asserted by that definition (Locally constant sheaves of finite rank); extension by zero, inverse and direct image have their usual exactness and stalk behaviour (Extension by zero for abelian sheaves on an open subspace, Inverse image presheaf and inverse image sheaf, Direct image of a sheaf along a continuous map, A sheaf on a topological space).

Proof

technique · direct
1.1givenF1F2F3F4

The recollement and the glued t-structure. Let U→jX←iZ be an algebraic open/closed decomposition. By [F1] the six-operation supplier gives j!⊣j∗⊣Rj∗ and i∗⊣i∗⊣i!, full faithfulness of i∗ and j!, the vanishing j∗i∗=0 and the two localization triangles j!j∗K→K→i∗i∗K→j!j∗K[1] and i∗i!K→K→Rj∗j∗K→i∗i!K[1]. Given t-structures on the bounded constructible categories of U and Z, define T≤0 by j∗K∈D≤0 and i∗K∈D≤0, and T≥0 by j∗K∈D≥0 and i!K∈D≥0. Orthogonality Hom⁡(T≤0,T≥1)=0 follows from the first localization triangle: the two endpoint Hom groups into B∈T≥1 vanish by adjunction and the two existing t-structures. For truncation, form Y→K→Rj∗τ≥1(j∗K) and then A→Y→i∗τ≥1(i∗Y); octahedral completion [F3] produces A→K→B with a triangle i∗τ≥1(i∗Y)→B→Rj∗τ≥1(j∗K). Applying j∗ shows j∗A=τ≤0j∗K and j∗B=τ≥1j∗K; applying i∗ to A→Y gives i∗A=τ≤0i∗Y, and applying i! to the B-triangle and using i!Rj∗=0 gives i!B=τ≥1i∗Y. Thus A∈T≤0, B∈T≥1 and shift-closure is immediate, so (T≤0,T≥0) is a t-structure in the sense of [F2]. Ordinary truncations preserve finite-rank locally constant cohomology, so [F2] restricts to the stratum-constructible category and shifting by dim⁡CS gives its required t-structure. Gluing a maximal open union of strata to its closed complement and inducting on their finite number proves the inequalities define such a t-structure and that it stays inside Dcb, because the six operations preserve constructibility and boundedness by [F1] and the local dimension data are those of [F4].

2.1givenF1F4step 1.1

Independence of compatible refinement and gluing. Let T⊂S be a refinement of smooth complex strata of codimension c along the same algebraic variety. Ordinary pullback preserves the upper bound −dim⁡CS≤−dim⁡CT, and on locally constant cohomology extraordinary pullback is ordinary pullback shifted by −2c: in local normal coordinates the closed-support complex of a real 2c-ball relative to its puncture has one oriented class in degree 2c, and tensoring with the local coefficient complex computes that shift, so its lower bound becomes −dim⁡CS+2c≥−dim⁡CT; hence both old halves lie in the refined halves, and uniqueness of truncation proves equality on the old category. Taking common refinements gives independence of the chosen compatible stratification, and the same argument applied over each member of an algebraic open/closed decomposition shows the glued structure is independent of the decomposition. This uses the local conical and orientation suppliers of [F4] rather than presupposing stratification independence.

2.2givenF1F5step 1.1

Duality exchanges the aisles. Let S be a smooth complex stratum with e=dim⁡CS. The exchange identities of [F1] give iS!DXA≅DSiS∗A and iS∗DXA≅DSiS!A, and on bounded complexes with finite-rank locally constant cohomology on S, complex orientation and exact finite-dimensional duality give Hj(DSK)≅(H−j−2eK)∨. Thus Ht(iS∗A)=0 for t>−e implies Hj(iS!DXA)=0 for j<−e, and Ht(iS!A)=0 for t<−e implies Hj(iS∗DXA)=0 for j>−e: the two opposite middle-perverse inequalities. Exact finite-dimensional duality holds over every field, and 2e is an integer cochain shift rather than a scalar being inverted. Applying the same argument to DXA and using the involutivity of [F1] gives equality of the exchanged aisles.

2.3givenF2F3step 1.1algebra

The heart is abelian. For a morphism u:A→B of heart objects, its cone C has t-cohomology only in degrees −1,0, by the truncation bounds. Put K=pH−1C and Q=pH0C. Applying octahedral completion to K[1]→C→Q and A→B→C produces one object I with triangles K→A→I→K[1] and I→B→Q→I[1]. The first bounds I in degrees [−1,0] and the second in [0,1], so I is in the heart. Applying Hom from and to heart objects and using orthogonality proves that K is the kernel of u, Q its cokernel, I the cokernel of K→A and the kernel of B→Q. Thus image equals coimage and the additive heart is abelian. For the glued structure j∗ preserves both aisles, hence truncation triangles and these kernels and cokernels, so open restriction is perverse exact.

3.1givenF1step 1.1step 2.3algebra

Intermediate extension and its boundary characterization. For a heart object M on an open j:U↪Z with closed complement i, let J!=pH0j!M and J∗=pH0Rj∗M. Adjunction and orthogonality show that J! is left adjoint and J∗ right adjoint to exact j∗ on hearts; in particular J! is right exact. They restrict to M. Moreover J! has no nonzero quotient supported on the boundary, since Hom⁡(J!,i∗N)=Hom⁡(M,j∗i∗N)=0, and J∗ has no boundary subobject by the dual adjunction. Define j!∗M=im⁡(J!→J∗). It inherits both properties. If P restricts to M and has neither boundary subobjects nor boundary quotients, the adjunction maps J!→P→J∗ have boundary-supported cokernel and kernel, so the first is epic and the second monic; their composite identifies P with j!∗M. A subobject of j!∗M for simple M restricts either to zero or to M; its boundary kernel or the resulting boundary quotient must vanish. Hence j!∗M is simple. Closed pushforward is exact and fully faithful on hearts by recollement, so this construction also defines intermediate extension from a locally closed dense open of a closed support. For a local system on a smooth open of dimension e, its input is M=L[e].

4.1givenF1F4F5step 2.1step 3.1constructalgebra

Simple local systems give simple shifted objects on a smooth connected variety S of complex dimension e. A path transports stalks by composing its local trivializations on a finite subdivision. Transition matrices are locally constant; a common subdivision proves independence of subdivision, and a finite rectangular subdivision of a path homotopy into trivializing opens proves homotopy invariance. Thus local subsystems correspond to invariant subspaces of the transported stalk. For a dense algebraic open U⊆S, every loop in S can be moved off S∖U: triangulate the algebraic subset as in [F4], and perturb a finite polygonal representative in smooth coordinate balls off its finitely many pieces of real codimension at least two. Move its finitely many vertices into U inside the overlapping balls. For a segment with endpoints in U, forbidden intermediate vertices lie on rays from either endpoint through boundary strata. Each ray family is locally parametrized by at most 2e−1 real parameters; covering countably many compact parameter boxes by small cubes bounds the total 2e-dimensional covering volume by an arbitrarily small number. Their union therefore cannot fill a coordinate ball. Choose an intermediate vertex outside both families and join by two straight segments. These replacements avoid the boundary and are homotopies in the balls, fixing the loop base point. Consequently transport on U has the same invariant subspaces as on S, and a simple L remains simple on U. Also L[e] has no perverse subobject or quotient supported on a proper algebraic closed subset: on any smooth boundary stratum of codimension c>0, ordinary restriction has degree −e<−dim⁡T, and extraordinary restriction has degree 2c−e>−dim⁡T, by the relative-ball calculation of step 2.1. These are the strict upper and lower boundary bounds; adjunction and orthogonality kill respectively boundary quotients and subobjects. A putative proper subobject is, on a sufficiently small dense algebraic open, a local subsystem of L[e], since its cohomology there is locally constant and perversity on that smooth open is the shifted ordinary structure. Simplicity after restriction makes it zero or all of L[e] there, and the preceding no-boundary properties then make it zero or all globally. Finite-rank local systems themselves have finite composition series, because every strict sub-local-system increases stalk dimension.

5.1givenF1F4F5step 2.3step 3.1step 4.1algebra

Classification and finite length. For a simple heart object choose a dense smooth open in one irreducible component of its support where it is a nonzero L[e]. A proper sub-local-system would, by the left adjunction in step 3.1, give a nonzero proper image subobject. Thus L is simple. A simple object has neither boundary subobjects nor quotients, so step 3.1 identifies it with the closed pushforward of j!∗L[e]; step 4.1 and step 3.1 prove that every such intersection complex is simple. For finite length, induct on the dimension of an object's support, with the empty support giving zero. Choose a dense smooth open union of its components on which its restriction M is a finite-rank local system with the appropriate shift. Each simple factor Mr has J!Mr↠j!∗Mr with boundary-supported kernel. That kernel has finite length by the dimension induction, and the quotient is simple, so J!Mr has finite length. Right exactness of J! and a finite local-system composition series then show that J!M has finite length. The image of J!M→P has finite length, and its cokernel is supported on the smaller-dimensional boundary, hence also has finite length by induction. Their short exact sequence proves finite length of P. This argument does not assume an arbitrary subobject chain stays constructible for an initially chosen stratification.

6.1givenF1F2F5step 2.3step 5.1∎

Negative Hom and division endomorphism algebras over every field. For m<0 the shift B[m] lies in T≥−m⊆T≥1 and A lies in T≤0, so orthogonality gives Hom⁡(A,B[m])=0: heart objects have no negative Hom. If u:T→T is nonzero with T simple, its kernel and image are subobjects with kernel zero and image T, so u is invertible and the endomorphism ring of T is a division k-algebra. Finite-rank local systems have finite length by the dimension of their underlying vector spaces, and the recollement and kernel/image arguments use only exactness and that finite length, so the ordinary perverse construction uses no characteristic-zero input and no equivariant, normalization, semisimplicity or noncharacteristic theorem.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Shifted smooth complex pullback is exact for the middle perverse t structure

Statement

Assume AC (The Axiom of Choice) and let a:X→Y be a smooth algebraic map of pure relative complex dimension d between separated finite-type complex varieties, over an arbitrary coefficient field k. Then a∗[d] is exact for the ordinary middle-perverse t structures of Ordinary middle perverse t structure, recollement and intersection complexes on Dcb(−;k), and it commutes with their truncation and cohomology functors. Complex orientation gives a!=a∗[2d]. If a is surjective, its normalized pullback is conservative and reflects the perverse support and cosupport inequalities. The compatible finite stratifications used are those of Complex algebraic varieties and maps admit compatible finite stratifications.

Facts & Assumptions

Given: AC; a smooth algebraic map a:X→Y of pure relative complex dimension d between separated finite-type complex varieties; an arbitrary coefficient field k; and the ordinary middle-perverse t-structures of Ordinary middle perverse t structure, recollement and intersection complexes on Dcb(−;k).

[F1]

On Dcb(X;k) the ordinary middle-perverse t-structure is given on smooth complex strata S by the upper inequality Hq(iS∗A)=0 for q>−dim⁡CS and the lower inequality Hq(iS!A)=0 for q<−dim⁡CS; it is a bounded t-structure, Verdier duality exchanges its aisles, and it is independent of the compatible finite stratification (Ordinary middle perverse t structure, recollement and intersection complexes).

[F2]

The six operations on Dcb: exactness of a∗, the adjunctions, coherent composition, base change, projection and trace, involutive duality exchanging the adjoints, and the smooth normalization ωX=kX[2e] on a smooth complex e-fold with a!=a∗[2d] for a smooth map of pure relative dimension d; coefficient extensions commute with these operations (Actual analytic constructible sheaves admit coherent six operations).

[F3]

Smooth algebraic maps admit compatible finite smooth stratifications of source and target: finitely many source and target strata such that each source stratum maps surjectively and smoothly onto one target stratum, with analytic strata smooth complex manifolds and the restrictions smooth submersions; the whole restriction over a target stratum is proper when the map itself is proper (Complex algebraic varieties and maps admit compatible finite stratifications).

[F4]

Inverse image is exact and commutes with shifts; extension by zero, reverse and direct image have their stated exactness and stalk behaviour; the derived Hom conventions are those of Derived category of an abelian category, The internal Hom sheaf of two module sheaves and the definitions of Inverse image presheaf and inverse image sheaf, Direct image of a sheaf along a continuous map, Extension by zero for abelian sheaves on an open subspace, A sheaf on a topological space.

Proof

technique · direct
1.1givenF1F2F3F4constructalgebra

Preservation of the upper aisle. Choose the finite compatible smooth target stratification of [F1] adapted to A, with the compatible refinements of [F3] available on affine charts, and work with the connected components T of the full inverse images a−1(S). Smooth base change makes each nonempty such component smooth of dimension dim⁡CS+d, and smooth local charts make this the product pullback stratification. It has the frontier and Whitney conditions locally because the target stratification does and a smooth chart adds a ball factor; finitely many algebraic components give a finite global stratification. This choice matters: a further arbitrary source refinement need not have the full relative dimension. Write e=dim⁡CS and aT:T→S. Composition of inverse images gives iT∗a∗A=aT∗iS∗A, and exactness gives Hq(iT∗a∗A[d])=aT∗Hq+d(iS∗A). For A in the upper aisle this vanishes whenever q+d>−e, or q>−(e+d), precisely the source support inequality. Independence of compatible refinement [F1] proves upper-aisle preservation on Dcb.

2.1givenF1F2step 1.1

Preservation of the lower aisle. The ordinary six-operation construction [F2] gives DXa∗≅a!DY and the complex-orientation normalization a!≅a∗[2d], where the relative dimension is d. Consequently DX(a∗A[d])≅a∗DYA[d]. The ordinary middle-perverse core of [F1] identifies the lower aisle as the duality image of the upper aisle, and both the duality exchange of [F1] and the pullback comparison of [F2] are used here, so the first step proves preservation of the lower aisle as well. The orientation unit is nonzero over every field, exact finite-dimensional duality is field-general, and [2d] is a cohomological shift rather than a scalar inversion; thus the support-and-duality argument applies to arbitrary k.

3.1givenF1F2F3step 1.1∎

Truncation, cohomology and reflection. Applying the t-exact functor a∗[d] of steps 1.1 and 2.1 to a truncation triangle produces the unique target truncation triangle, proving that it commutes with perverse truncation and perverse cohomology. If a is surjective, inverse image is stalkwise conservative and reflects ordinary cohomological vanishings; applying the same dimension calculation as in step 1.1 to the support and dual support inequalities reflects both perverse inequalities. The dimension count is the same over every field, and no characteristic hypothesis enters, so the stated conservativity and reflection hold for arbitrary k.

5 · Examples, counterexamples and false statements

None yet.

Sources