Glossary of scheme theory

This is a glossary of scheme theory. For an introduction to the theory of schemes in algebraic geometry, see affine scheme, projective space, sheaf and scheme. The concern here is to list the fundamental technical definitions and properties of scheme theory. See also list of algebraic geometry topics.

Points

A scheme S is a locally ringed space, so a fortiori a topological space, but the meanings of point of S are threefold:

1. a point P of the underlying topological space;
2. a T-valued point of S is a morphism from T to S, for any scheme T;
3. a geometric point, where S is defined over (is equipped with a morphism to) Spec(K), where K is a field, is a morphism from $\textrm{Spec} (\overline{K})$ to S where $\overline{K}$ is an algebraic closure of K.

Geometric points are what in the most classical cases, for example algebraic varieties that are complex manifolds, would be the ordinary-sense points. The points P of the underlying space include analogues of the generic points (in the sense of Zariski, not that of André Weil), which specialise to ordinary-sense points. The T-valued points are thought of, via Yoneda's lemma, as a way of identifying S with the representable functor h_S it sets up. Historically there was a process by which projective geometry added more points (e.g. complex points, line at infinity) to simplify the geometry by refining the basic objects. The T-valued points were a massive further step.

As part of the predominating Grothendieck approach, there are three corresponding notions of fiber of a morphism: the first being the simple inverse image of a point. The other two are formed by creating fiber products of two morphisms. For example, a geometric fiber of a morphism $S^{\prime} \to S$ is thought of as

$S^{\prime} \times_{S} \textrm{Spec}(\overline{K})$.

This makes the extension from affine schemes, where it is just the tensor product of R-algebras, to all schemes of the fiber product operation a significant (if technically anodyne) result.

Properties of schemes

Most important properties of schemes are local in nature, i.e. a scheme X has a certain property P if and only if for any cover of X by open subschemes X_i, i.e. X=$\cup$ X_i, every X_i has the property P. It is usually the case that is enough to check one cover, not all possible ones. One also says that a certain property is Zariski-local, if one needs to distinguish between the Zariski topology and other possible topologies, like the étale topology.

Consider a scheme X and a cover by affine open subschemes Spec A_i. Using the dictionary between (commutative) rings and affine schemes local properties are thus properties of the rings A_i. A property P is local in the above sense, iff the corresponding property of rings is stable under localization.

For example, we can speak of locally Noetherian schemes, namely those which are covered by the spectra of Noetherian rings. The fact that localizations of a Noetherian ring are still noetherian then means that the property of a scheme of being locally Noetherian is local in the above sense (whence the name). Another example: if a ring is reduced (i.e., has no non-zero nilpotent elements), then so are its localizations.

An example for a non-local property is separatedness (see below for the definition). Any affine scheme is separated, therefore any scheme is locally separated. However, the affine pieces may glue together pathologically to yield a non-separated scheme.

The following is a (non-exhaustive) list of local properties of rings, which are applied to schemes. Let X = $\cup$ Spec A_i be a covering of a scheme by open affine subschemes. For definiteness, let k denote a field in the following. Most of the examples also work with the integers Z as a base, though, or even more general bases.

notion	definition	example	non-example
related to scheme structure
connected	The scheme is connected as a topological space. Since the connected components refine the irreducible components any irreducible scheme is connected but not vice versa. An affine scheme Spec(R) is connected iff the ring R possesses no idempotents other than 0 and 1; such a ring is also called a connected ring.	affine space, projective space	Spec(k[x]×k[x])
irreducible	A scheme X is said to be irreducible when (as a topological space) it is not the union of two closed subsets except if one is equal to X. Using the correspondence of prime ideals and points in an affine scheme, this means X is irreducible iff X is connected and the rings Ai all have exactly one minimal prime ideal. (Rings possessing exactly one minimal prime ideal are therefore also called irreducible.) Any noetherian scheme can be written uniquely as the union of finitely many maximal irreducible non-empty closed subsets, called its irreducible components.	affine space, projective space	Spec k[x,y]/(xy) =
reduced	The Ai are reduced rings. Equivalently, none of its rings of sections $\mathcal O_X(U)$ (U any open subset of X) has any nonzero nilpotent element. Allowing non-reduced schemes is one of the major generalizations from varieties to schemes.	varieties (by definition)	Spec k[x]/(x2)
integral	A scheme that is both reduced and irreducible is called integral. Equivalently, a connected scheme that is covered by the spectra of integral domains. (Strictly speaking, this is not a local property, because the disjoint union of two integral schemes is not integral. However, for irreducible schemes, it is a local property.)	Spec k[t]/f, f irreducible polynomial	Spec A ⊕ B. (A, B ≠ 0)
normal	An integral scheme is called normal, if the Ai are integrally closed domains.	regular schemes	singular curves
related to regularity
regular	The Ai are regular.	smooth varieties over a field	Spec k[x,y]/(x2+x3-y2)=
Cohen-Macaulay	All local rings are Cohen-Macaulay.	regular schemes, Spec k[x,y]/(xy)
related to "size"
locally noetherian	The Ai are Noetherian rings. If in addition a finite number of such affine spectra covers X, the scheme is called noetherian. While it is true that the spectrum of a noetherian ring is a noetherian topological space, the converse is false.	(Virtually everything in algebraic geometry).	$GL_\infty = \cup GL_n$
dimension	The dimension, by definition the maximal length of a chain of irreducible closed subschemes, is a global property. It can be seen locally if a scheme is irreducible. It depends only on the topology, not on the structure sheaf. See also Global dimension.	equidimensional schemes in dimension 0: Artinian schemes, 1: algebraic curves, 2: algebraic surfaces.
catenary	A scheme is catenary, if all chains between two irreducible closed subschemes have the same length.	(Virtually everything, e.g. varieties over a field)

Properties of scheme morphisms

One of Grothendieck's fundamental ideas is to emphasize relative notions, i.e. conditions on morphisms rather than conditions on schemes themselves. The category of schemes has a final object, the spectrum of the ring $\mathbb{Z}$ of integers; so that any scheme S is over $\textrm{Spec} (\mathbb{Z})$, and in a unique way.

For the following definitions, we take as standard notation

$f: Y \to X$

to be a morphism of schemes. Parallel to the properties of schemes above, the following properties of morphisms are also of local nature, i.e. if there is an open covering of X by some open subschemes U_i, such that the restriction of f to f ^{− 1}(U_i) has the property, then f has it, as well.

Notions related to the topological structure

A morphism of schemes is called open (closed) , if the underlying map of topological spaces is open (closed, respectively), i.e. if open subschemes of Y are mapped to open subschemes of X (and similarly for closed). For example, finitely presented flat morphisms are open and proper maps are closed.

A morphism is called dominant, if the image f(Y) is dense. A morphism of affine schemes Spec A → Spec B is dense if and only if the kernel of the corresponding map B → A is contained in the nilradical of A.

A morphism is called quasi-compact, if for some (equivalently: every) open affine cover of X by some U_i = Spec B_i, the preimages f⁻¹(U_i) are quasi-compact.

Open and closed subschemes and immersions

An open subscheme of a scheme X is an open subset U with structure sheaf $\mathcal{O}_X|_U$.^[1]

Closed subschemes of a scheme X are defined to be those occurring in the following construction. Let J be a quasi-coherent sheaf of $\mathcal{O}_X$-ideals. The support of the quotient sheaf $\mathcal{O}_X/J$ is a closed subset Z of X and $(Z,(\mathcal{O}_X/J)|_Z)$ is a scheme called the closed subscheme defined by the quasi-coherent sheaf of ideals J.^[2] The reason the definition of closed subschemes relies on such a construction is that, unlike open subsets, a closed subset of a scheme does not have a unique structure as a subscheme.

A subscheme, without qualifier, of X is a closed subscheme of an open subscheme of X.

Immersions f : Y → X are maps that factor through isomorphisms with subschemes. Specifically, an open immersion factors through an isomorphism with an open subscheme and a closed immersion factors through an isomorphism with a closed subscheme.^[3] Equivalently, f is a closed immersion if, and only if, it induces a homeomorphism from the underlying topological space of Y to a closed subset of the underlying topological space of X, and if the morphism $f^\sharp: \mathcal{O}_X \to f_* \mathcal{O}_Y$ is surjective.^[1] A composition of immersions is again an immersion.^[4]

Some authors, such as Hartshorne in his book Algebraic Geometry and Q. Liu in his book Algebraic Geometry and Arithmetic Curves, define immersions as the composite of an open immersion followed by a closed immersion. These immersions are immersions in the sense above, but the converse is false. Furthermore, under this definition, the composite of two immersions is not necessarily an immersion. However, the two definitions are equivalent when f is quasi-compact.^[5]

Note that an open immersion is completely described by its image in the sense of topological spaces, while a closed immersion is not: $\operatorname{Spec} A/I$ and $\operatorname{Spec} A/J$ may be homeomorphic but not isomorphic. This happens, for example, if I is the radical of J but J is not a radical ideal. When specifying a closed subset of a scheme without mentioning the scheme structure, usually the so-called reduced scheme structure is meant, that is, the scheme structure corresponding to the unique radical ideal consisting of all functions vanishing on that closed subset.

Affine and projective morphisms

A morphism is called affine if the preimage of any open affine subset is again affine. In more fancy terms, affine morphisms are defined by the global Spec construction for sheaves of O_X-Algebras, defined by analogy with the spectrum of a ring. Important affine morphisms are vector bundles, and finite morphisms.

Projective morphisms are defined similarly, but in practice they turn out to be more important than affine morphisms: f is called projective if it factors as a closed immersion followed by the projection of a projective space $\mathbb{P}^{n}_X := \mathbb{P}^n \times X$ to X. Again, one may say, that f is projective if it is given by the global Proj construction on graded commutative O_X-Algebras.

Separated and proper morphisms

A separated morphism is a morphism f such that the fiber product of f with itself along f has its diagonal as a closed subscheme — in other words, the diagonal map is a closed immersion.

As a consequence, a scheme X is separated when the diagonal of X within the scheme product of X with itself is a closed immersion. Emphasizing the relative point of view, one might equivalently define a scheme to be separated if the unique morphism $X \rightarrow \textrm{Spec} (\mathbb{Z})$ is separated.

Notice that for a topological space Y is Hausdorff iff the diagonal embedding

$Y \stackrel{\Delta}{\longrightarrow} Y \times Y$

is closed. In algebraic geometry, the above formulation is used because a scheme is a Hausdorff space if and only if it is zero-dimensional. The difference between the topological and algebro-geometric context comes from the topological structure of the fiber product (in the category of schemes) $X \times_{\textrm{Spec} (\mathbb{Z})} X$, which is different from the product of topological spaces.

Any affine scheme Spec A is separated, because the diagonal corresponds to the surjective map of rings (hence is a closed immersion of schemes):

$A \otimes_{\mathbb Z} A \rightarrow A, a \otimes a' \mapsto a \cdot a'$.

A morphism f : X → Y is called quasi-separated or (X is quasi-separated over Y) if the diagonal morphism X → X ×_YX is quasi-compact. A scheme X is called quasi-separated if X is quasi-separated over Spec(Z). ^[6]

While the separatedness is of rather technical nature, properness has deep geometrical meaning.

A morphism is proper if it is separated, universally closed (i.e. such that fiber products with it preserve closed immersions), and of finite type. Projective morphisms are proper; but the converse is not in general true. See also complete variety. A deep property of proper morphisms is the existence of a Stein factorization, namely the existence of an intermediate scheme such that a morphism can be expressed as one with connected fibres, followed by a finite morphism.

Finite, quasi-finite, finite type, and finite presentation morphisms

A morphism $f: Y \to X$ is finite if X may be covered by affine open sets Spec B such that each f ^{− 1}(Spec B) is affine—say of the form Spec A -- and furthermore A is finitely generated as a B-module. See finite morphism.

The morphism f is locally of finite type if X may be covered by affine open sets Spec B such that each inverse image f ^{− 1}(Spec B) is covered by affine open sets Spec A where each A is finitely generated as a B-algebra.

The morphism f is finite type if X may be covered by affine open sets Spec B such that each inverse image f ^{− 1}(Spec B) is covered by finitely many affine open sets Spec A where each A is finitely generated as a B-algebra.

The morphism f has finite fibers if the fiber over each point $x \in X$ is a finite set. A morphism is quasi-finite if it is of finite type and has finite fibers.

Finite morphisms are quasi-finite, but not all morphisms having finite fibers are quasi-finite, and morphisms of finite type are usually not quasi-finite.

If x is a point of X, then the morphism f is of finite presentation at x (or finitely presented at x) if there is an open affine subset V of Y and an open affine neighbourhood U of x such that f(U) ⊆ V and $\mathcal{O}_X(U)$ is a finitely presented algebra over $\mathcal{O}_Y(V)$. The morphism f is locally of finite presentation if it is finitely presented at all points of X. If Y is locally Noetherian, then f is locally of finite presentation if, and only if, it is locally of finite type.^[7]

The morphism f is of finite presentation (or X is finitely presented over Y) if it is locally of finite presentation, quasi-compact, and quasi-separated. If Y is locally Noetherian, then f is of finite presentation if, and only if, it is of finite type.^[8]

Flat morphisms

A morphism f is flat if it gives rise to a flat map on stalks. When viewing a morphism as a family of schemes parametrized by the points of X, the geometric meaning of flatness could roughly be described by saying that the fibers f ^{− 1}(x) do not vary too wildly.

Unramified and étale morphisms

For a point y in Y, consider the corresponding morphism of local rings

$f^\# \colon \mathcal{O}_{X, f(y)} \to \mathcal{O}_{Y, y}.$

Let $\mathfrak{m}$ be the maximal ideal of $\mathcal{O}_{X,f(y)}$, and let

$\mathfrak{n} = f^\#(\mathfrak{m}) \mathcal{O}_{Y,y}$

be the ideal generated by the image of $\mathfrak{m}$ in $\mathcal{O}_{Y,y}$. The morphism f is unramified if it is locally of finite presentation and if for all y in Y, $\mathfrak{n}$ is the maximal ideal of $\mathcal{O}_{Y,y}$ and the induced map

$\mathcal{O}_{X,f(y)}/\mathfrak{m} \to \mathcal{O}_{Y,y}/\mathfrak{n}$

is a finite, separable field extension. This is the geometric version (and generalization) of an unramified field extension in algebraic number theory.

A morphism f is étale if it is flat and unramified. There are several other equivalent definitions. In the case of smooth varieties X and Y over an algebraically closed field, étale morphisms are precisely those inducing an isomorphism of tangent spaces $df: T_{x} X \rightarrow T_{f(x)} Y$, which coincides with the usual notion of étale map in differential geometry.

Étale morphisms form a very important class of morphisms; they are used to build the so-called étale topology and consequently the étale cohomology, which is nowadays one of the cornerstones of algebraic geometry.

Smooth morphisms

The higher-dimensional analog of étale morphisms are smooth morphisms. There are many different characterisations of smoothness. The following are equivalent definitions of smoothness:

1) for any y ∈ Y, there are open affine neighborhoods V and U of y, x=f(y), respectively, such that the restriction of f to V factors as an étale morphism followed by the projection of affine n-space over U.

2) f is flat, locally of finite presentation, and for every geometric point $\bar{y}$ of Y (a morphism from the spectrum of an algebraically closed field $k(\bar{y})$ to Y), the geometric fiber $X_{\bar{y}}:=X\times_Y \mathrm{Spec} (k(\bar{y}))$ is a smooth n-dimensional variety over $k(\bar{y})$ in the sense of classical algebraic geometry.

Constructions

Scheme-theoretic image

If f : Y → X is any morphism of schemes, the scheme-theoretic image of f is the unique closed subscheme i : Z → X which satisfies the following universal property:

1. f factors through i,
2. if j : Z′ → X is any closed subscheme of X such that f factors through j, then i also factors through j.^{[9] [10]}

This notion is distinct for that of the usual set-theoretic image of f, f(Y). For example, the underlying space of Z always contain (but not necessarily equals) the Zariski closure of f(Y) in X, so if Y is any open (and not closed) subscheme of X and f is the inclusion map, then Z is different from f(Y). When Y is reduced, then Z is the Zariski closure of f(Y) endowed with the structure of reduced closed subscheme. But in general, unless f is quasi-compact, the construction of Z is not local on X.

Notes

1. ^ ^{a b} Hartshorne 1977, §II.3
2. ^ Grothendieck & Dieudonné 1960, 4.1.2 and 4.1.3
3. ^ Grothendieck & Dieudonné 1960, 4.2.1
4. ^ Grothendieck & Dieudonné 1960, 4.2.5
5. ^ Q. Liu, Algebraic Geometry and Arithmetic Curves, exercise 2.3
6. ^ Grothendieck & Dieudonné 1964, 1.2.1
7. ^ Grothendieck & Dieudonné 1964, §1.4
8. ^ Grothendieck & Dieudonné 1964, §1.6
9. ^ Hartshorne 1977, Exercise II.3.11(d)
10. ^ The Stacks Project, Chapter 21, §4.

References

• Grothendieck, Alexandre; Dieudonné, Jean (1960). "Éléments de géométrie algébrique (rédigés avec la collaboration de Jean Dieudonné) : I. Le langage des schémas". Publications Mathématiques de l'IHÉS 4. MR0217083. http://www.numdam.org:80/numdam-bin/feuilleter?id=PMIHES_1960__4_. 
• Grothendieck, Alexandre; Dieudonné, Jean (1964). "Éléments de géométrie algébrique (rédigés avec la collaboration de Jean Dieudonné) : IV. Étude locale des schémas et des morphismes de schémas, Première partie". Publications Mathématiques de l'IHÉS 20. MR0173675. http://www.numdam.org:80/numdam-bin/feuilleter?id=PMIHES_1964__20_. 
• Hartshorne, Robin (1977), Algebraic Geometry, Graduate Texts in Mathematics, 52, New York: Springer-Verlag, ISBN 978-0-387-90244-9, MR0463157