Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6-sol)audited 2026-09-30
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.

Quasi-coherence of pushforward for qcqs morphisms

Statement

Assume the Axiom of Choice, inherited through the associated-sheaf and affine equivalence machinery (The Axiom of Choice). Let f:X→S be a morphism of schemes that is quasi-compact and quasi-separated (Quasi-compact and quasi-separated morphisms), and let F be a quasi-coherent OX-module (Quasi-coherent module on a scheme). Then the direct image f∗F (Direct image of a sheaf along a continuous map) is a quasi-coherent OS-module.

The claim includes the empty source and the empty target, the zero module and the identity morphism. Only quasi-compactness and quasi-separatedness of f and quasi-coherence of F are used; no separatedness, Noetherian, reducedness, flatness or finiteness hypothesis is imposed.

Facts & Assumptions

Given: A quasi-compact and quasi-separated morphism f:X→S of schemes and a quasi-coherent OX-module F; in the proof an affine open U=Spec⁡R⊆S is fixed and XU=f−1(U) is written for its inverse image.

[F1]

Direct image (Direct image of a sheaf along a continuous map, Direct image preserves sheaves and objectwise algebraic structure, Restriction of a sheaf to an open subspace): the direct image is defined by (f∗F)(V)=F(f−1V) for open V⊆S with restrictions induced by those of F; if F is a sheaf of modules then so is f∗F; and for open W⊆U⊆S one has ((f∗F)∣U)(W)=F(f−1W).

[F2]

Morphism and scheme quasi-compactness (Quasi-compact and quasi-separated morphisms, Quasi-compact and quasi-separated schemes, Open cover, subcover, and compact topological space; a compact subset is a subspace that is compact in its own right, Every affine scheme is quasi-compact, Schemes, Affine open subschemes): f is quasi-compact when f−1(V) is quasi-compact for every quasi-compact open V⊆S, and quasi-separated when affine opens U1,U2⊆X lying over a common affine open of S have quasi-compact intersection; a scheme is quasi-compact when its underlying space has the finite-subcover property for open covers, so an open cover of a quasi-compact open subscheme has a finite subcover; every affine scheme is quasi-compact, and the affine open subschemes of a scheme form a basis of its topology.

[F3]

Quasi-coherent modules (Quasi-coherent module on a scheme, Module sheaf on an affine scheme): F is quasi-coherent when every point of X has an affine open neighbourhood U=Spec⁡A with F∣U≅M~ for an A-module M; the condition is local on X, invariant under isomorphism and inherited by restrictions to open subschemes, and on an affine scheme each associated sheaf M~ is quasi-coherent.

[F4]

Affine equivalence and distinguished-open sections (Affine quasi-coherent sheaves are modules, Sections of the associated sheaf on basic opens): on an affine scheme V=Spec⁡A every quasi-coherent sheaf G is canonically Γ(V,G)~ and the functors (−)~ and Γ(V,−) are quasi-inverse equivalences; for an A-module M there are canonical identifications Γ(D(a),M~)=Ma, natural in a and M, with restriction D(b)⊆D(a) the localisation map Ma→Mb, and both sides vanish for a=0.

[F5]

Localisation (Localisation of a module at a multiplicative subset, Principal localisation Rf={1,f,f2,…}−1R, A localised module fraction is zero exactly when one denominator kills its numerator, Universal property of localisation: maps that invert S factor uniquely through S−1R): the localisation S−1M consists of fractions m/s with m/s=n/t precisely when u(tm−sn)=0 for some u∈S, and m/s=0 precisely when um=0 for some u∈S; one writes Mf for the localisation at the powers of f; and for a ring map R→C the composite R→C→Cψ(f) sends f to a unit, hence extends uniquely over Rf, so that Cψ(f), and with it every Cψ(f)-module, carries an Rf-module structure.

[F6]

Sheaf axiom (A sheaf on a topological space): compatible sections on an open cover of a sheaf glue uniquely, and two sections are equal once they agree on an open cover.

[F7]

Kernel sheaf (Kernel sheaves are objectwise, while cokernels and images are sheafified): the kernel of a morphism φ:A→B of sheaves of modules is the objectwise kernel subsheaf, ker⁡(φ)(W)=ker⁡(φW).

[F8]

Quasi-coherence is closed under kernels and finite direct sums (Kernels and cokernels of quasi-coherent modules, Abelian subcategory and exact embedding): QCoh⁡(X) is an abelian subcategory of Mod⁡(OX), hence kernels of morphisms of quasi-coherent modules and finite biproducts of quasi-coherent modules are quasi-coherent.

[F9]

Affine charts (Affine schemes are contravariantly equivalent to commutative rings, The map of affine spectra induced by a ring homomorphism, The underlying space of an affine spectrum): a morphism of affine schemes g:Spec⁡C→Spec⁡R is Spec⁡(ψ) for a unique ring map ψ:R→C, given on points by contraction q↦ψ−1(q); consequently g−1D(r)=D(ψ(r)) for every r∈R, and the distinguished opens form a basis.

[F10]

The Axiom of Choice, inherited from the associated-sheaf existence theorem, the affine equivalence and the gluing machinery (The Axiom of Choice).

Proof technique: direct; reduce to an affine target, model every affine chart inside the source by its module of global sections, and express the direct image as the kernel of a morphism between finite sums of such models.

Proof

1.1F1F4F9

Affine model of one chart: let g:Spec⁡C→Spec⁡R be a morphism of affine schemes with corresponding ring map ψ:R→C, let G be a quasi-coherent OSpec⁡C-module and put N=Γ(Spec⁡C,G), regarded as an R-module through ψ. Then G≅N~ as C-modules, and for every r∈R the inverse image g−1D(r) is the distinguished open D(ψ(r)), so the sections of the direct image are Γ(D(r),g∗G)=G(g−1D(r))=G(D(ψ(r)))=Nψ(r), the localisation of the C-module N at ψ(r).

1.2F2F9choose

Covering data inside a fixed affine target chart: fix an affine open U=Spec⁡R⊆S; then U is quasi-compact, so XU=f−1(U) is quasi-compact because f is quasi-compact, and the family of all affine open subschemes of XU, which covers XU, has a finite subcover U1,…,Un (with n=0 meaning XU=∅); for each pair i≤j the intersection Ui∩Uj is quasi-compact by quasi-separatedness, so it has a finite affine cover Uij1,…,Uijmij, and one may take Uii1=Ui. All these are affine opens of X lying over U, their corresponding ring maps are ψi:R→Γ(Ui,OX) and ψijk:R→Γ(Uijk,OX), and we write Ni=Γ(Ui,F) and Nijk=Γ(Uijk,F); only finitely many objects are chosen.

2.1F5

Affine model of the comparison: in the situation of step 1.1 regard each Nψ(r) as an Rr-module through the canonical ring map Rr→Cψ(r); then for every r∈R the map χr:Nr→Nψ(r), n/rk↦n/ψ(r)k in fractions, is well defined, Rr-linear and bijective: if n/rk=n′/rl in Nr then rm(rln−rkn′)=0 for some m, and applying ψ gives ψ(r)m(ψ(r)ln−ψ(r)kn′)=0, so the images agree in Nψ(r); every element of Nψ(r) is a fraction n/ψ(r)k, so χr is surjective; and χr(n/rk)=0 means ψ(r)mn=0 for some m, hence rmn=ψ(r)mn=0 and n/rk=0 in Nr, so χr is injective; linearity is immediate from the fraction formulas.

3.1F3F4F5F6step 1.1step 2.1

The affine model is an associated sheaf: in the situation of steps 1.1 and 2.1 the maps χr are compatible with the restriction maps of NR~ and g∗G, because for D(s)⊆D(r) both composites Nr→Nψ(s) are the canonical localisation maps induced by the ring maps Rr→Rs and Rr→Cψ(r)→Cψ(s); since the distinguished opens form a basis and both sides are sheaves, these compatible isomorphisms on a basis assemble into a unique isomorphism of OSpec⁡R-modules NR~→g∗G whose component on D(r) is χr, by restricting a section over an open set to the distinguished opens it contains, mapping each restriction and gluing in the target; the inverses χr−1 assemble into an inverse in the same way, so g∗G≅NR~ is quasi-coherent.

4.1F3step 1.2step 3.1

Application to the covering charts: by step 3.1 applied to g=f∣Ui:Ui→U with the quasi-coherent module F∣Ui one has (f∣Ui)∗(F∣Ui)≅(Ni)R~, a quasi-coherent sheaf on U; the same argument applies to each chart f∣Uijk:Uijk→U with module Nijk, so every summand occurring below is quasi-coherent.

5.1F1F6F7F8step 1.2step 4.1

The kernel description: let E=⨁i=1n(f∣Ui)∗(F∣Ui) and E′=⨁i≤j⨁k=1mij(f∣Uijk)∗(F∣Uijk); restrictions define an OU-linear morphism d:E→E′ whose component on an open W⊆U is dW((si)i)(i,j,k)=si∣Uijk∩f−1W−sj∣Uijk∩f−1W, and E,E′ are quasi-coherent by [F8] and step 4.1. For every open W⊆U a family (si) lies in ker⁡(dW) exactly when si and sj agree on Ui∩Uj∩f−1W for all i,j, because the opens Uijk∩f−1W cover that intersection and F is separated [F6]; such compatible families are exactly the restrictions to the cover (f−1W∩Ui)i of a section in F(f−1W), by gluing and locality in the sheaf F [F6], so there is an objectwise bijection ker⁡(d)(W)≅F(f−1W)=((f∗F)∣U)(W) which is natural in W and O(W)-linear.

6.1F3F7F8step 1.2step 4.1step 5.1

Conclusion: by step 5.1 the restriction (f∗F)∣U≅ker⁡d is the kernel of a morphism of quasi-coherent OU-modules, hence is quasi-coherent by [F8]; since the affine open U⊆S was arbitrary and affine opens cover S, locality of quasi-coherence [F3] shows that f∗F itself is quasi-coherent.

7.1F2F3F5F10step 2.1step 1.2step 5.1step 6.1∎

Choice and edge cases: the only selections are finite subcovers of the fixed covers by affine opens, the finite covers of the quasi-compact intersections, and finitely many distinguished opens, so no infinite simultaneous choice or localisation at infinitely many primes is made, and the only Axiom of Choice is the inherited one recorded in [F10]. The empty and degenerate cases are covered by the same steps: for XU=∅ one takes n=0 and then E=E′=0 and (f∗F)∣U=0; for r=0 both sides of step 2.1 are the zero module; and for S=∅ or C=0 the relevant spaces and modules are zero as well.

Depends on

Used by

Dependency tree · two levels

76 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources