04.01.03 · algebraic-geometry / sheaves

Sheafification

shipped3 tiersLean: partial

Anchor (Master): Hartshorne §II; Vakil; Godement *Topologie algébrique et théorie des faisceaux* (1958); Cartan séminaire 1948–55

Intuition [Beginner]

A presheaf is a system of local data: an assignment of a set (or group, or ring) to each open set, with restriction maps. A sheaf is a presheaf where the local data glues consistently — sections agreeing on overlaps assemble to a unique global section. Many natural constructions (kernels, images, quotients) produce a presheaf that is not a sheaf — the local data is correct, but the gluing axiom fails.

Sheafification is the universal way to fix this: it takes a presheaf and produces the closest sheaf. The result is a sheaf $F^{+}$ together with a presheaf morphism $F \to F^{+}$ , satisfying the universal property that any presheaf morphism $F \to G$ to a sheaf $G$ factors uniquely through $F^{+}$ .

The key fact: sheafification preserves stalks. The germs of $F^{+}$ at any point equal the germs of $F$ at that point. So sheafification is not about adding new local data — it is purely about reorganising the global structure to satisfy the gluing axiom. This makes sheafification the central technical tool for working with presheaves: kernels, cokernels, images, tensor products, and pullbacks of sheaves all involve sheafifying an intermediate presheaf.

Visual [Beginner]

A presheaf with sections that almost glue, and the sheafification process producing the canonical sheaf with the same stalks but proper global gluing.

Worked example [Beginner]

The constant presheaf $F_{A}$ on a topological space $X$ assigns the abelian group $A$ to every nonempty open set, with restriction maps the identity:

F_{A} (U) = A (for nonempty U); F_{A} (\emptyset) = 0.

This is a presheaf, but it is not a sheaf in general. Consider $X = {p, q}$ with the discrete topology. The opens are $\emptyset, {p}, {q}, {p, q}$ . Sections of the constant presheaf are: $A, A, A, A$ over the four opens.

The sheaf gluing axiom requires that for the cover ${p, q} = {p} \cup {q}$ , sections on ${p}$ and ${q}$ that "agree on the overlap" $\emptyset$ glue to a unique section on ${p, q}$ . Two elements of $A$ on ${p}$ and ${q}$ vacuously agree on $\emptyset$ (which has section group 0), so the sheaf would need $F_{A} ({p, q}) = A \times A$ , not $A$ .

The constant sheaf $\underline{A}$ is the sheafification: $\underline{A} (U) = A^{π_{0} (U)}$ , one copy of $A$ per connected component of $U$ . On the discrete two-point space, $\underline{A} ({p, q}) = A^{2}$ , matching the gluing requirement.

The constant presheaf and constant sheaf agree on every connected open and on every stalk (each stalk equals $A$ ); they differ on disconnected opens like ${p, q}$ . Sheafification fixes the disconnected case while preserving the local (stalk) structure.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a topological space and $F$ a presheaf on $X$ (of sets, abelian groups, rings, modules, etc.). The sheafification of $F$ , denoted $F^{+}$ , is the sheaf characterised by the following universal property: there is a presheaf morphism $η : F \to F^{+}$ such that for any sheaf $G$ and presheaf morphism $φ : F \to G$ , there exists a unique sheaf morphism $φ : F^{+} \to G$ with $φ \circ η = φ$ .

In categorical language: sheafification is the left adjoint to the inclusion functor $ι : Sh (X) ↪ PSh (X)$ ,

(-)^{+} ⊣ ι, Hom_{Sh} (F^{+}, G) ≅ Hom_{PSh} (F, G) .

Construction via étale space. The most concrete construction: form the étale space

\overset{ˊ}{E} t (F) := x \in X ⨆ F_{x},

with the topology making continuous all sections $s : U \to \overset{ˊ}{E} t (F)$ , $x \mapsto s_{x}$ , induced by sections $s \in F (U)$ . Then $F^{+}$ is the sheaf of continuous local sections of the projection $π : \overset{ˊ}{E} t (F) \to X$ :

F^{+} (U) := {σ : U \to \overset{ˊ}{E} t (F) ∣ σ continuous, π \circ σ = id_{U}} .

The natural map $η : F \to F^{+}$ sends $s \in F (U)$ to the section $x \mapsto s_{x}$ .

Construction via plus-construction (Grothendieck). For a general site, sheafification is performed twice, by the plus construction $F_{0}^{+} (U) := lim_{U cover of U} \overset{ˇ}{H}^{0} (U; F)$ , then $F^{+} := F_{0}^{+} \circ F_{0}^{+}$ . On topological spaces, the étale-space construction gives the same result more directly.

Properties.

(SF1) Sheafification preserves stalks. For every $x \in X$ , $F_{x} = (F^{+})_{x}$ as the colimit construction commutes with the locality completion.

(SF2) Sheafification is the identity on sheaves. If $F$ is already a sheaf, then $η : F \to F^{+}$ is an isomorphism.

(SF3) Sheafification is exact. As a left adjoint, $(-)^{+}$ preserves colimits — including coequalisers, hence cokernels in the abelian-group case. As a left exact functor (filtered colimits commute with finite limits), $(-)^{+}$ is also left exact. Hence sheafification is an exact functor between abelian categories.

(SF4) Functoriality. A morphism $φ : F \to G$ of presheaves induces a morphism $φ^{+} : F^{+} \to G^{+}$ of sheaves. The plus construction is a functor.

Use cases.

Cokernels of sheaf morphisms. The presheaf cokernel $U \mapsto coker (φ (U))$ is generally not a sheaf; the sheafification produces the cokernel in $Sh (X; Ab)$ .
Image sheaves. The presheaf image $U \mapsto im (φ (U))$ is generally not a sheaf; sheafification produces the image sheaf, used in the sheaf-version of the first isomorphism theorem.
Tensor products. For sheaves $F, G$ of $O_{X}$ -modules, the presheaf tensor $U \mapsto F (U) \otimes_{O_{X} (U)} G (U)$ is generally not a sheaf; sheafification gives the sheaf tensor product $F \otimes_{O_{X}} G$ .
Pullback / inverse image. For $f : X \to Y$ continuous and $G$ a sheaf on $Y$ , the inverse-image presheaf $U \mapsto lim_{V \supseteq f (U)} G (V)$ is generally not a sheaf; sheafification gives $f^{- 1} G$ .
Constant sheaf. The constant presheaf with value $A$ sheafifies to the constant sheaf $\underline{A}$ , with $\underline{A} (U) = A^{π_{0} (U)}$ .

In each of these constructions, the presheaf version is the natural pointwise definition, but the resulting object fails the gluing axiom; sheafification is what produces the right sheaf.

Sheafification on sites. For a Grothendieck site $(C, T)$ , the sheafification functor still exists as the left adjoint to the inclusion $Sh (C, T) ↪ PSh (C)$ . Construction via plus-plus or Verdier's stalkwise approximation. This generality is essential in étale cohomology and Grothendieck topologies.

Key theorem with proof [Intermediate+]

Theorem (existence and stalk-preservation of sheafification). For any presheaf $F$ on a topological space $X$ , there exists a sheaf $F^{+}$ and a presheaf morphism $η : F \to F^{+}$ satisfying the universal property: any presheaf morphism $F \to G$ to a sheaf factors uniquely through $η$ . Moreover, $η$ induces an isomorphism on every stalk: $F_{x} ≅ (F^{+})_{x}$ for every $x \in X$ .

Proof. Construction. Define $F^{+} (U)$ as the set of "matching families of germs": functions $σ : U \to ⨆_{x \in U} F_{x}$ with $σ (x) \in F_{x}$ for every $x$ , locally lifted to a section: every $x \in U$ has an open neighbourhood $V \subseteq U$ and a section $s \in F (V)$ with $σ (y) = s_{y}$ for all $y \in V$ .

Equivalently: $F^{+}$ is the sheaf of continuous sections of the étale space $π : \overset{ˊ}{E} t (F) \to X$ , where $\overset{ˊ}{E} t (F) = ⨆_{x} F_{x}$ has the topology generated by section-images.

Restriction maps: $σ ∣_{V} (x) := σ (x)$ for $x \in V$ — this is a function $V \to \overset{ˊ}{E} t (F)$ inheriting the locality property.

Sheaf axioms. (Locality) Two elements of $F^{+} (U)$ agreeing on every member of an open cover agree pointwise (germs match), hence are equal. (Gluing) A locally given matching family of germs glues to a global matching family, since the locality condition is local — if it holds on each member of a cover, it holds on the union.

Universal property. The natural map $η : F \to F^{+}$ sends $s \in F (U)$ to $σ_{s} (x) := s_{x}$ — the matching family of germs of $s$ .

For a sheaf $G$ and a presheaf map $φ : F \to G$ , define $φ (U) : F^{+} (U) \to G (U)$ as follows: given $σ \in F^{+} (U)$ , the matching-family condition gives an open cover ${V_{i}}$ of $U$ and sections $s_{i} \in F (V_{i})$ with $σ (x) = (s_{i})_{x}$ for $x \in V_{i}$ . Apply $φ$ to get $φ (s_{i}) \in G (V_{i})$ . The germs of $φ (s_{i})$ at any $x \in V_{i} \cap V_{j}$ agree (both equal $φ_{x} (σ (x))$ via the pointwise compatibility), so by locality of $G$ , $φ (s_{i}) ∣_{V_{i} \cap V_{j}} = φ (s_{j}) ∣_{V_{i} \cap V_{j}}$ . By gluing on $G$ , the $φ (s_{i})$ patch to a unique $φ (σ) \in G (U)$ .

This $φ$ is well-defined, satisfies $φ \circ η = φ$ , and is unique by the gluing-uniqueness on $G$ .

Stalk preservation. For $x \in X$ , both $F_{x}$ and $(F^{+})_{x}$ are computed as colimits over open neighbourhoods. The map $η_{x} : F_{x} \to (F^{+})_{x}$ sends a germ $s_{x}$ to the germ $σ_{x}$ where $σ$ is the matching-family-of-germs section.

Surjectivity: a germ in $(F^{+})_{x}$ is represented by a matching family $σ$ on some neighbourhood $U$ of $x$ . By the locality condition, on a smaller $V \subseteq U$ containing $x$ , $σ$ agrees with the germs of a section $s \in F (V)$ . So $σ_{x} = (η (s))_{x} = η_{x} (s_{x})$ .

Injectivity: if $η_{x} (s_{x}) = η_{x} (t_{x})$ , then on some neighbourhood $V ∋ x$ , the matching families $σ_{s}$ and $σ_{t}$ agree pointwise. So $s_{y} = t_{y}$ for every $y \in V$ , hence by the colimit definition $s_{x} = t_{x}$ in $F_{x}$ .

So $η_{x} : F_{x} \to (F^{+})_{x}$ is a bijection. $□$

This theorem articulates why sheafification is the right notion: it takes a presheaf to the unique sheaf with the same stalks, and it is universal among such constructions. The stalk-preservation property is what makes sheafification harmless from the local viewpoint while fixing the global structure.

Bridge. The construction here builds toward 04.01.04 (direct and inverse image of sheaves), where the same data is developed in the next layer of the strand. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 1 (easy, proof).

Show that if $F$ is already a sheaf, then $η : F \to F^{+}$ is an isomorphism.

Hint

The universal property gives $F^{+} ≅ F$ via the identity factorisation.

Answer

If $F$ is a sheaf, the identity $id_{F} : F \to F$ is a presheaf morphism to a sheaf. By the universal property of sheafification, it factors uniquely through $η$ : there is a unique sheaf morphism $ρ : F^{+} \to F$ with $ρ \circ η = id_{F}$ .

Conversely, $η \circ ρ : F^{+} \to F^{+}$ and the identity $id_{F^{+}}$ both satisfy: post-composition with $η$ gives $η$ . By the universal property's uniqueness clause, $η \circ ρ = id_{F^{+}}$ .

So $η$ and $ρ$ are mutually inverse, $η$ is an isomorphism. $□$

This is an instance of the general fact: a left adjoint, restricted to the image of the right adjoint, is the identity on isomorphism. The category of sheaves is a reflective subcategory of presheaves, with sheafification the reflector.

Exercise 3 (medium, proof).

Show that sheafification is an exact functor: if $0 \to F^{'} \to F \to F^{''} \to 0$ is an exact sequence of presheaves, then $0 \to F^{' +} \to F^{+} \to F^{'' +} \to 0$ is an exact sequence of sheaves.

Hint

Use stalk preservation and stalkwise exactness for sheaves.

Answer

Exactness in $Sh (X; Ab)$ is detected stalkwise: a sequence is exact iff every stalk sequence is exact (this is the abelian-category property of sheaves).

Exactness in $PSh (X; Ab)$ is also detected stalkwise: stalks are filtered colimits of presheaf-section-data, and filtered colimits are exact in $Ab$ .

Sheafification preserves stalks: $(F)_{x} = (F^{+})_{x}$ for every $x$ .

Combining these: if $0 \to F^{'} \to F \to F^{''} \to 0$ is exact in $PSh$ , every stalk sequence $0 \to F_{x}^{'} \to F_{x} \to F_{x}^{''} \to 0$ is exact. By stalk preservation, the corresponding sheaf sequence has the same stalks, hence the same stalk sequences. So $0 \to F^{' +} \to F^{+} \to F^{'' +} \to 0$ is stalkwise exact, hence exact in $Sh (X; Ab)$ . $□$

This exactness is the key technical property of sheafification: it makes sheafification well-behaved on long exact sequences and homological constructions. As a left adjoint, sheafification is automatically right-exact; the surprise is that it is also left-exact. The combination gives full exactness, making the category of sheaves of abelian groups behave well.

Exercise 4 (medium, numeric).

For the morphism $φ : O \to O$ , $f \mapsto f^{2}$ , on the structure sheaf of $C^{\times}$ (nonzero complex numbers, with the analytic topology), compute the cokernel sheaf.

Hint

Stalkwise, $φ_{z}$ has cokernel $Z /2$ — but the cokernel sheaf is more subtle.

Answer

The map $φ : O_{C^{\times}} \to O_{C^{\times}}$ , $f \mapsto f^{2}$ , on multiplicative units (so we should write this on $O^{\times}$ ):

Stalkwise, on $O_{z}^{\times}$ (units of holomorphic germs at $z$ ): every nonzero germ is a square locally (using a local logarithm), so $φ_{z}$ is surjective and the cokernel stalk is 0.

The presheaf cokernel $U \mapsto O^{\times} (U) / O^{\times} (U)^{2}$ is generally nonzero (for example $U = C^{\times}$ itself: the function $z$ is not globally a square on $C^{\times}$ , so $z mod (O^{\times})^{2}$ is a nontrivial element of the presheaf cokernel).

Sheafification kills these nontrivial presheaf-cokernel elements: since the stalks are zero, the sheaf cokernel is the zero sheaf.

The "obstruction" to global squaring on $C^{\times}$ is detected not by the sheaf cokernel but by higher cohomology: $H^{1} (C^{\times}; \underline{Z /2}) = Z /2$ , the Brauer-like obstruction class of $z$ . This is the prototype example of how sheafification kills local data while leaving the global obstruction in higher cohomology.

Exercise 5 (medium, conceptual).

Why does the cokernel of a sheaf morphism need to be sheafified?

Hint

The presheaf cokernel may fail the gluing axiom even when both inputs are sheaves.

Answer

Let $φ : F \to G$ be a morphism of sheaves. The presheaf cokernel is $U \mapsto coker (φ (U)) = G (U) / im (φ (U))$ .

This presheaf can fail the gluing axiom. Concrete example on $C^{\times}$ : $φ : O^{\times} \to O^{\times}$ , $f \mapsto f^{2}$ . Locally on simply connected opens, $φ (U)$ is surjective (every unit has a local square root), so the presheaf cokernel is 0 on these opens. But $φ (C^{\times})$ is not surjective: $z$ is not a global square. So the presheaf cokernel on $C^{\times}$ is nonzero, while on every simply connected piece of an open cover, the presheaf cokernel is 0. This violates gluing: nonzero on the union, zero on the cover pieces.

Sheafification fixes this: it forces the sheaf cokernel to match the (stalkwise) zero data on simply connected opens, killing the non-locally-detectable global obstruction. The sheaf cokernel of $f \mapsto f^{2}$ on $O_{C^{\times}}^{\times}$ is the zero sheaf, even though the presheaf cokernel is nonzero on $C^{\times}$ .

The "lost" global obstruction reappears in higher cohomology: the long exact sequence

0 \to F \to G \to coker φ \to 0

induces a long exact sequence in cohomology, and the global element of the presheaf cokernel that vanishes in the sheaf cokernel manifests as a connecting-homomorphism class in $H^{1}$ . This is the formal mechanism by which sheafification, sheaf cohomology, and obstruction theory fit together.

Exercise 6 (hard, proof).

Show that sheafification commutes with finite limits but generally not with infinite limits.

Hint

Sheafification is left adjoint, so commutes with colimits; finite limits commute with filtered colimits in $Ab$ .

Answer

Sheafification commutes with colimits. As a left adjoint, $(-)^{+}$ commutes with all (small) colimits.

Sheafification commutes with finite limits. Stalks commute with finite limits (filtered colimits and finite limits commute in $Ab$ and similar locally finitely presentable categories). Stalks commute with sheafification (stalk preservation). Therefore: a finite limit of presheaves has the same stalks as the corresponding finite limit of sheafifications, hence the sheafifications agree.

More precisely: for a finite diagram of presheaves $F_{i}$ with limit $L = lim F_{i}$ :

Each stalk $L_{x} = lim (F_{i})_{x}$ (filtered colimit over $U ∋ x$ commutes with finite limit).
$L^{+}$ has stalks $L_{x} = lim (F_{i})_{x}$ .
$lim F_{i}^{+}$ has stalks $lim (F_{i}^{+})_{x} = lim (F_{i})_{x}$ .
By stalkwise determination of sheaves, $L^{+} ≅ lim F_{i}^{+}$ .

Sheafification does not commute with infinite limits. Filtered colimits do not commute with arbitrary infinite limits in $Ab$ . Concrete failure: an infinite product of sheaves' stalks is generally not the stalk of an infinite product of presheaves' stalks (the issue is whether infinite products commute with filtered colimits — they do not).

Example: take a discrete infinite topological space $X = N$ and the constant sheaves $\underline{Z}_{n}$ for each $n \in N$ . The infinite product $\prod_{n} \underline{Z}_{n}$ in $Sh (X)$ has sections $\prod_{n} Z^{∣ U ∣}$ , but stalkwise behaviour differs from the product of presheaf stalks.

This finite-limit / infinite-limit asymmetry is one of the subtle but important properties of sheafification. It is the categorical reason behind many failures of "naive" stalkwise constructions for infinite operations. $□$

Exercise 7 (hard, conceptual).

Explain how sheafification on a Grothendieck site differs from sheafification on a topological space.

Hint

The plus construction must be applied twice on a general site; once suffices on a topological space.

Answer

On a topological space, the étale-space construction gives the sheafification $F^{+}$ in one step. The étale space topology makes continuous local sections automatically satisfy gluing.

On a general Grothendieck site $(C, T)$ , there is no topological-space backing for $C$ — covers are abstract families of morphisms — and the étale-space construction does not apply directly.

The plus construction on a site is

(F_{0}^{+}) (U) := U \in Cov (U) lim \overset{ˇ}{H}^{0} (U; F),

where the colimit is over covers $U$ of $U$ in the Grothendieck topology, and $\overset{ˇ}{H}^{0}$ is the equaliser of the two restriction maps $\prod_{i} F (U_{i}) ⇉ \prod_{i, j} F (U_{i} \times_{U} U_{j})$ .

This $F_{0}^{+}$ is separated (the locality axiom holds: equal-on-cover implies equal) but not generally a sheaf. Applying the plus construction again, $F^{+} := F_{0}^{+} \circ F_{0}^{+}$ , gives a sheaf — Grothendieck's two-step procedure.

For a topological-space site (covers = open covers), the two-step procedure simplifies: one application of plus is already a sheaf, recovering the étale-space construction. This is because topological covers are "well-behaved" in a way that general site covers may not be (e.g., étale topology, fppf topology, fpqc topology).

The general site-theoretic sheafification is foundational for:

Étale cohomology (Grothendieck-Artin-Verdier 1960s): $Sh (X_{\overset{e}{ˊ} t})$ is the abelian category for étale cohomology of schemes.
Faithfully flat descent (Grothendieck): the fppf and fpqc topoi have their own sheafifications.
Higher topos theory (Lurie): $\infty$ -categorical sheaves on sites generalise classical sheaves; sheafification becomes an $\infty$ -categorical localisation functor.
Crystalline cohomology (Berthelot, Ogus): sheaves on the crystalline site of a scheme require general site-theoretic sheafification.

So sheafification is the universal tool for completing local data to satisfy any prescribed gluing axiom — its full power emerges in the abstract site-theoretic setting, where the topological intuition is replaced by categorical formal data.

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has TopCat.Presheaf.sheafify and the categorical sheafification adjunction in CategoryTheory.Sites.Sheafification.

[object Promise]

Advanced results [Master]

Sheafification on Grothendieck sites. For a Grothendieck site $(C, T)$ , the sheafification functor exists as the left adjoint to $Sh (C, T) ↪ PSh (C)$ . Construction via the plus-plus procedure: $F^{+} := F^{+_{0} +_{0}}$ , where $F^{+_{0}} (U) = lim_{U} \overset{ˇ}{H}^{0} (U; F)$ over covers $U$ of $U$ . The two-fold application is necessary on a general site (one application separates, two applications produce a sheaf).

Reflective localisation. The category of sheaves is a reflective localisation of presheaves: it is obtained by formally inverting the morphisms $F \to F^{+}$ for all presheaves $F$ . This is the Bousfield localisation of $PSh$ at the class of sheafification maps.

Sheafification in derived algebraic geometry. Lurie's formulation: in the $\infty$ -category of presheaves of spaces (or spectra) on a site, sheaves are the local objects with respect to the Grothendieck topology, and sheafification is an $\infty$ -categorical localisation. The construction generalises to hypercomplete sheaves and presents the foundation for $\infty$ -categorical algebraic geometry (Lurie's Higher Topos Theory and Spectral Algebraic Geometry).

Topos-theoretic sheafification. A topos is a category equivalent to $Sh (C, T)$ for some site. The sheafification functor exhibits every topos as the localisation of the presheaf topos $PSh (C)$ at the local-isomorphism class. Subtoposes correspond to Lawvere-Tierney topologies, which are the inner-language characterisations of Grothendieck topologies.

Constructive aspects. Sheafification is constructive in the sense of topos theory: the axiom of choice is needed only to choose representatives in the colimit, and the construction is uniform across the topos. This makes sheafification a tool not only for algebraic geometry but for type theory (homotopy type theory, with sheaves as univalent groupoids) and constructive analysis.

Verdier's perspective. Verdier (1965) gave a homotopical reformulation: the derived category of sheaves on a topological space is a Bousfield localisation of the derived category of presheaves. This connects sheafification to Quillen model structures and modern stable homotopy theory.

Sheafification in $p$ -adic geometry. The pro-étale, pro-finite, and perfectoid topologies (Scholze 2012-) come with their own sheafification procedures, often requiring careful handling of size and the overconvergent or perfectoid conditions. Sheafification on these sites is the foundation of $p$ -adic Hodge theory and the recent geometrisation of the local Langlands programme.

Synthesis. This construction generalises the pattern fixed in 04.01.01 (sheaf), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

Detailed proofs of: existence of sheafification on a Grothendieck site (via plus-plus); exactness of sheafification; commutativity with finite limits but not arbitrary limits; the topos-theoretic characterisation as reflective localisation; Verdier's homotopical perspective — these are deferred to companion units in the topos-theory and derived-categories strands. The basic stalk-preservation and existence theorems are proved in the formal-definition and key-theorem sections.

Connections [Master]

Sheaf 04.01.01 — sheafification is the universal completion of presheaves into sheaves.
Stalk of a sheaf 04.01.02 — sheafification preserves stalks; this is the key structural property.
Direct and inverse image 04.01.04 — the inverse-image presheaf is generally not a sheaf; $f^{- 1} G$ is its sheafification.
Sheaf cohomology 04.03.01 — exactness of sheafification controls long exact sequences in sheaf cohomology.
Coherent sheaf 04.06.02 — many constructions on coherent sheaves (image, cokernel, tensor) require sheafification of an intermediate presheaf.
Quasi-coherent sheaf 04.06.01 — the sheafification of an $R$ -module on $Spec (R)$ is the corresponding quasi-coherent sheaf.
Scheme 04.02.01 — gluing schemes from local data formally requires sheafifying the underlying ringed-space data.

Historical & philosophical context [Master]

The construction of sheafification emerged in the late 1940s and early 1950s through the seminars of Henri Cartan at the École Normale Supérieure. Cartan's exposés (1948–54) developed the precise definitions of sheaves and the étale-space construction, with Jean-Pierre Serre, Armand Borel, and Alexander Grothendieck among the participants. The étale-space description — sheaves as local homeomorphisms over the base space — gave a topological meaning to sheaves and provided the most direct construction of sheafification. The seminars were a key locus for the diffusion of sheaf-theoretic methods through French and international mathematics.

Roger Godement's 1958 monograph Topologie algébrique et théorie des faisceaux gave the first systematic textbook treatment of sheaves and sheafification. Godement's approach was foundational: the étale-space (espace étalé) construction, the comparison of presheaves and sheaves, the plus construction on a topological space, and the consequences for cohomology. Godement's book remains a model of clarity in the foundational period of sheaf theory and is still consulted today.

Alexander Grothendieck's transformation in the late 1950s and 1960s extended sheafification far beyond the topological case. In Sur quelques points d'algèbre homologique (Tôhoku, 1957) he showed that the abelian category of sheaves of abelian groups on a topological space has enough injectives — making derived-functor sheaf cohomology well-defined. The 1960s SGA seminars (especially SGA 4 with Artin and Verdier) introduced Grothendieck topologies and sites, with sheafification as a fundamental construction on any site. The plus-plus procedure for general sites was the technical core, and it enabled étale cohomology, faithfully flat descent, and the rich array of cohomology theories now standard in arithmetic geometry. Grothendieck's perspective: sheafification is not a topological-space-specific notion but a categorical / topos-theoretic operation, the universal completion of any presheaf-style local data into the sheaf satisfying the prescribed gluing.

In contemporary research, sheafification appears in many forms: as a key step in derived algebraic geometry (Lurie's $\infty$ -categorical sheaves), as the foundation of perfectoid spaces (Scholze, requiring sheafification on the pro-étale and pro-finite sites), as a tool in homotopy type theory (univalent sheaves and Bousfield localisations), and as a method in microlocal sheaf theory (Kashiwara-Schapira). Cartan's 1948–55 exposés and Godement's 1958 monograph remain the foundational sources, with Grothendieck-Artin-Verdier SGA 4 the canonical reference for the general site-theoretic version.

Bibliography [Master]

Hartshorne, Algebraic Geometry — §II.1 covers sheafification on topological spaces.
Vakil, The Rising Sea: Foundations of Algebraic Geometry — §2.4, sheafification.
Godement, Topologie algébrique et théorie des faisceaux (1958) — the systematic mid-century reference.
Cartan, Séminaire Cartan (1948–1955) — exposés on sheaves, étale spaces, and sheafification.
Grothendieck, Sur quelques points d'algèbre homologique (Tôhoku, 1957) — sheaves of abelian groups and derived-functor cohomology.
Grothendieck-Artin-Verdier, SGA 4 — sites and the general theory of sheafification.
MacLane-Moerdijk, Sheaves in Geometry and Logic — topos-theoretic introduction.
Bredon, Sheaf Theory — Ch. I, foundations of sheafification.
Iversen, Cohomology of Sheaves — Ch. III, sheafification and homological algebra.
Lurie, Higher Topos Theory — $\infty$ -categorical sheafification.
Kashiwara-Schapira, Sheaves on Manifolds — microlocal perspective and applications.

Prerequisites

04.01.01
04.01.02

Used in

04.01.04

Tier anchors

beginner: Strogatz analogous level for completing local data into a sheaf
intermediate: Hartshorne §II.1; Vakil §2.4; Bredon Ch. I
master: Hartshorne §II; Vakil; Godement *Topologie algébrique et théorie des faisceaux* (1958); Cartan séminaire 1948–55

References

TODO_REF
Hartshorne — Algebraic Geometry · §II.1, sheafification
TODO_REF
Vakil — The Rising Sea · §2.4, sheafification of presheaves
TODO_REF
Godement — Topologie algébrique et théorie des faisceaux · 1958, the systematic introduction of sheafification
TODO_REF
Cartan — Séminaire 1948–1955 · exposés on sheaves, étale spaces, and sheafification
TODO_REF
Bredon — Sheaf Theory · Ch. I, foundations including sheafification

Lean module

Codex.AlgebraicGeometry.Sheaves.Sheafification

Reviewer

TBD

Estimated time

beginner: 10m
intermediate: 30m
master: 50m