Fppf topology¶
A Grothendieck topology on schemes whose covering families are jointly surjective flat morphisms locally of finite presentation.
Core Idea¶
The fppf topology specifies which families of scheme morphisms count as covers. A family \(\{f_i:T_i\to T\}\) is an fppf covering when every \(f_i\) is flat and locally of finite presentation and the images of the \(T_i\) together cover \(T\). Flatness and presentation are tests on each map; surjectivity is a test on the family. An individual open immersion can therefore be one member of a cover without being surjective.[1]
These families satisfy the covering rules needed for a Grothendieck site: isomorphisms cover, fppf covering families compose, and pulling a covering family back along a scheme morphism gives another fppf covering family. When speaking of a site, this entry uses the size-controlled big site \((\mathit{Sch}/S)_{fppf}\) over a specified base scheme \(S\). The covering predicate and the chosen site's size convention should be stated separately.[1]
This named topology is the rule for admissible covers. Quasi-coherent-sheaf descent and fppf cohomology are uses of the rule under further theorems and coefficient choices; their conclusions do not define what an fppf cover is.[2][3]
How would you explain it like I'm…
Rules for What Counts as Covering
Flat Finitely Presented Covers
Structural Signature¶
Sig role-phrases: schemes and morphisms → flat locally finitely presented maps → joint image coverage → base-change-stable covering site.
- Scheme carrier. A target scheme \(T\) and a family of schemes \(T_i\) mapping to it supply the typed objects. The topology is applied in a category of schemes over a base, not to arbitrary subsets of one topological space.[1]
- Per-map condition. Each \(f_i\) must be flat and locally of finite presentation. Dropping the presentation condition changes the cover test. Equivalently, the disjoint-union map from all family sources is flat, locally of finite presentation, and surjective.[1]
- Family coverage. The union of the images is all of \(T\). Each map may cover only part of the target; a collection of valid per-map morphisms that misses a point is not an fppf cover.[1]
- Stability and site. Isomorphisms, composition of covering families, and base change preserve the fppf-covering rule. This permits the stated big fppf site. For an affine target, a cover has a finite affine refinement; in the standard affine basis each map is flat and of finite presentation. The basis description is a special construction, not a replacement for the general locally finite-presentation definition.[1]
What It Is Not¶
Fppf is not an alternative spelling of fpqc. Every fppf cover is fpqc, but the fpqc definition uses a distinct affine-local finite covering condition and does not impose the fppf local finite-presentation condition on each map. A flat, surjective map alone does not certify the fppf predicate.[4][1]
Nor is the topology a theorem that arbitrary objects descend effectively. Descent asks whether objects and compatible identifications over a cover come from an object over the base; the answer depends on the object class. Quasi-coherent sheaves satisfy a specific fpqc descent theorem, which applies to fppf covers because they are fpqc covers. That consequence is separate from the choice of covering families.[2][4]
Scope of Application¶
Use the fppf covering test for scheme-theoretic sites and sheaves when flat, locally finitely presented families are the local morphisms permitted. The familiar Zariski, étale, and smooth coverings satisfy the fppf test, while fppf also admits some non-étale covers. For an affine target, standard finite affine covers are a convenient basis, with the stated source and morphism restrictions.[1]
Applications need their own hypotheses. For example, the cited descent theorem concerns quasi-coherent sheaves with descent data over fpqc coverings. The cited Kummer statement identifies \(H^1_{fppf}(S,\mu_n)\) with specified line-bundle/root pairs and separately gives an étale identification when \(n\) is invertible on \(S\). Neither source licenses an unrestricted assertion about all data, all cohomology degrees, or all group schemes.[2][3]
Clarity¶
Three questions prevent most mistaken identifications. Are all family maps flat? Are they each locally of finite presentation? Do their images together cover the target? A “yes” to all three identifies an fppf covering under the stated convention. The French abbreviation is a naming aid, but “faithfully flat” should not be applied to every individual member of a jointly covering family: an open chart embedding may not be surjective.[1]
The distinction between a general fppf cover and a standard affine fppf cover also matters. The latter has finitely many affine sources and morphisms of finite presentation over an affine target. A proof can use that finite basis after refinement without quietly declaring every original cover finite or every original map globally finitely presented.[1]
Manages Complexity¶
The rule packages many possible local presentations of one scheme into a checkable predicate. Instead of separately listing every acceptable family, one verifies two morphism properties and joint coverage. Stability under composition and base change then lets a cover be replaced by another fppf covering family during local arguments without changing the topology.[1]
The affine refinement result supplies a second simplification: over an affine target, an arbitrary fppf covering can be refined by finitely many affine sources. The proof may pass to that basis for computations, but an object-specific descent or cohomology claim still needs its own theorem. Topological admissibility does not establish effectiveness of arbitrary gluing data.[1][2]
Abstract Reasoning¶
Given a proposed family, type the target and each source as schemes over the chosen base. Check flatness and local finite presentation for every arrow, then test whether the images jointly cover the target. If the family is constructed by base change or by composing known fppf covering families, the stability lemma can discharge those tests; if an affine calculation is easier, seek the standard affine refinement rather than changing the general definition.[1]
After coverage is established, ask a separate question about what is being glued or measured. A quasi-coherent sheaf with descent data falls under a specified fpqc theorem; a \(\mu_n\) cohomology calculation has its coefficient and degree assumptions. An assertion about an arbitrary object class cannot be inferred from the fppf label alone.[2][3]
Knowledge Transfer¶
The same fppf test can be applied to unlike scheme families. The two affine charts of a projective line give a Zariski open-immersion family; a characteristic-\(p\) root torsor gives one finite flat cover that is not étale-locally trivial. Their geometry and uses differ, but the operative roles—flat locally finitely presented maps and joint coverage—remain the same.[5][1][6]
This transfer stays inside algebraic geometry. An ordinary open cover shares the idea of locality but does not by itself express the scheme-morphism and flatness conditions. A move from fppf to fpqc also changes the exact predicate, so a proof using the former's finite-presentation hypothesis must be checked again under the latter's rules.[4]
Examples¶
Two affine charts of a projective line¶
For a field \(k\), the projective line \(\mathbf P^1_k\) is the union of two standard affine opens. Their open immersions into \(\mathbf P^1_k\) form a Zariski covering and hence an fppf covering. Mapped back: the carrier is \(\mathbf P^1_k\); each chart immersion supplies a flat locally finitely presented map; the pair jointly covers the target, although neither chart alone does; within the stated big fppf site this family is a cover, and the general stability under further fppf composition and base change comes from the covering lemma, not from this example alone.[5][1]
A characteristic-\(p\) root torsor¶
Let \(K=\mathbf F_p(t)\). The \(\mu_p\)-torsor \(\operatorname{Spec}(K[x]/(x^p-t))\to\operatorname{Spec}(K)\) is fppf-locally trivial but not étale-locally trivial. The quotient is finite free of rank \(p\) over \(K\), so the single map is flat and finitely presented; it is surjective. Mapped back: the carrier is a field scheme, the one map supplies the per-map properties and the whole target image, and its non-étale behavior demonstrates that fppf is broader than an étale-only cover test. Within the stated big fppf site this singleton family is a cover; the general stability under further fppf composition and base change comes from the covering lemma, not from this example alone. Torsor classification is an application, not a fourth condition in the fppf definition.[6][1]
Structural Tensions¶
The reviewed sources do not establish an inherent optimization tradeoff as part of the fppf covering identity. Its boundary is exact: local finite presentation must accompany flatness and joint coverage. The larger fpqc class permits different flat coverings, but treating “more covers versus fewer covers” as a goal to optimize would add a purpose that the definition itself does not supply. For a proposed use, state which cover class the theorem actually assumes.[1][4]
Structural–Framed Character¶
The entry is structural within algebraic geometry, with a named, domain-bound mathematical test. Five aspects keep that description precise:
- Evaluative weight: low. An fppf covering is determined by provable morphism properties and image coverage, not an aesthetic, ethical or policy ranking.[1]
- Human-practice dependence: low for the validity of a cover, but present in the decision to use this site for a particular mathematical question. A researcher can choose a topology; the chosen fppf predicate then has fixed proof obligations.[1]
- Institutional origin and conventions: historical French naming and the selection of a size-controlled big site reflect mathematical practice. Such conventions determine which formal site is being discussed, while the criterion for a cover is fixed once that convention is stated. This entry does not infer a historical priority claim from the name.[1]
- Vocabulary travel: “cover,” “local,” and “glue” travel widely; flatness, local finite presentation and scheme morphisms do not carry their fppf meaning to an unrelated setting.[1]
- Import versus recognition: a new scheme family literally instantiates fppf only after the three cover tests are proved. Recognizing a similar local-to-global shape outside algebraic geometry does not import this topology.[1]
Its character: a proof-governed covering structure with limited portability, because its generic locality skeleton can travel while its flat and presentation tests remain specific to scheme theory.
Structural Core vs. Domain Accent¶
The broadly portable skeleton is a locality structure that designates covers and supports reasoning from local data. Whether this categorical-cover skeleton warrants a future Prime is an open Prime-review question; the live Topology Prime instead defines an open-set structure on a space and emphasizes deformation invariants, so it is not silently assigned this genus. Within mathematics, a general Grothendieck topology gives the nearer categorical genus of covering families stable under pullback and composition. The fppf differentia is the scheme carrier and the conjunction of flatness, local finite presentation, and joint image coverage. If those conditions are removed, one has a different topology or merely a general covering idea.[1]
This entry remains domain-specific because its recognition and failure tests require scheme-theoretic flatness and presentation. Zariski and non-étale torsor examples show breadth within the domain, not a cross-domain Prime. The notation, base scheme, and particular sheaf or torsor may change; the fppf covering predicate cannot.[1][6]
Instantiates / Related Primes¶
This entry is a kind of Grothendieck topology.
The live Topology Prime is a comparison neighbor with a different open-set carrier and deformation-centered identity, while the live Grothendieck topology draft names the nearer categorical covering genus. The strict parent relation to that genus rests on a scheme-specific cover test: every fppf topology supplies categorical covering structure, while a general Grothendieck topology need not use flat locally finitely presented scheme maps. The paired parent repair removes the Grothendieck entry’s unsupported strict link to the open-set Topology Prime. The fppf edge uses the repaired categorical parent and does not inherit that Prime ancestry.[1]
The live Descent (Mathematics) entry is related by use rather than by a proved kind-of relation: fppf covers can be hypotheses of descent statements, while descent itself concerns compatible local objects and effectivity. The live V-topology is another neighboring cover choice with a different test. Neither thematic association is an edge proposal.[2]
Relationships to Other Abstractions¶
Current abstraction Fppf topology Domain-specific
Parents (1) — outgoing typed relations
-
Fppf topology is a kind of Grothendieck topology Domain-specific
The fppf covering rule is a Grothendieck topology on schemes, with flat locally finitely presented jointly covering families as its differentia.The fppf topology supplies a Grothendieck covering structure on schemes over a base. Each map in an fppf covering family is flat and locally of finite presentation, and their images jointly cover the target (Stacks 34.7.1); identity, composition and base-change stability hold (Stacks 34.7.3). A chosen size-controlled big site represents arbitrary fppf covers by selected covers or refinements under Stacks 34.7.6-34.7.8; not every arbitrary family is literally a selected site covering. General Grothendieck covering axioms do not impose this scheme-specific predicate, so the child adds a stable differentia. The paired parent repair removes its unsupported strict prime:topology edge, so the child does not inherit an open-set Topology Prime ancestry.
Hierarchy path (1) — routes to 1 parentless root
- Fppf topology → is a kind of Grothendieck topology
Neighborhood in Abstraction Space¶
Fppf topology sits in a crowded region of the domain-specific corpus (37th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.
Family — Graph Structures & Combinatorial Objects (48 abstractions)
Nearest neighbors
- Sheaf — 0.88
- A-paracompact Space — 0.88
- Closed Immersion — 0.88
- Algebraic stack — 0.87
- Descent (Mathematics) — 0.87
Computed from structural-signature embeddings · 2026-10-31
Not to Be Confused With¶
Fpqc topology: a distinct affine-local flat covering rule; fppf covers qualify as fpqc, but the reverse cannot be assumed. Étale topology: its covers are fppf covers, yet a characteristic-\(p\) \(\mu_p\) torsor can be fppf-locally but not étale-locally trivial. Faithfully flat morphism: a property of a map, whereas an fppf cover can be a jointly covering family whose individual members are not surjective. Flat cohomology: an application built from a site and coefficients, not the covering predicate itself. Descent: an object-specific recovery question rather than an automatic consequence for every object.[1][4][6][3][2]
References¶
[1] The Stacks Project. “The Stacks Project”. Section 34.7, “The fppf topology,” tag 021L. Primary collaborative mathematical reference, full web text inspected. Definition 34.7.1 gives the family test; Lemmas 34.7.2–34.7.4 give inclusions, stability and affine refinement; Definitions 34.7.5, 34.7.6 and 34.7.8 give standard affine covers and the size-controlled big site. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z
[2] The Stacks Project. “The Stacks Project”. Section 35.5, “Fpqc descent of quasi-coherent sheaves,” tag 023R. Primary collaborative mathematical reference, full web text inspected. Proposition 35.5.2 states the effective-descent theorem for the specified object class. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g
[3] The Stacks Project. “The Stacks Project”. Lemma 59.28.5, tag 040Q. Primary collaborative mathematical reference, full web text inspected. It identifies \(H^1_{fppf}(S,\mu_n)\) with specified line-bundle/root pairs and states an étale identification when \(n\) is invertible on \(S\). registry ↩a ↩b ↩c ↩d
[4] The Stacks Project. “The Stacks Project”. Section 34.9, “The fpqc topology,” tag 022A. Primary collaborative mathematical reference, full web text inspected. Definition 34.9.1 states the affine-local fpqc test; Lemma 34.9.7 proves every fppf cover is fpqc. registry ↩a ↩b ↩c ↩d ↩e
[5] The Stacks Project. “The Stacks Project”. Example 26.21.8, tag 01KQ. Primary collaborative mathematical reference, full web text inspected. The projective line over a field has two affine-open charts. registry ↩a ↩b
[6] The Stacks Project. “The Stacks Project”. Section 110.55, “A non-algebraic classifying stack,” tag 077C. Primary collaborative mathematical reference, full web text inspected. The paragraph beginning “The first objection” supplies the characteristic-\(p\) \(\mu_p\) torsor and its fppf versus étale local-triviality contrast. registry ↩a ↩b ↩c ↩d