ConceptComplete

The Spectrum of a Ring

The spectrum Spec⁡A\operatorname{Spec} A is the fundamental construction bridging commutative algebra and geometry. It turns any commutative ring into a topological space with a structure sheaf, making it the basic building block of scheme theory.


Definition and points

Definition2.5Spectrum of a ring

Let AA be a commutative ring with 11. The spectrum of AA is

Spec⁡A={p⊆A∣p is a prime ideal}\operatorname{Spec} A = \{\mathfrak{p} \subseteq A \mid \mathfrak{p} \text{ is a prime ideal}\}

as a set. Each prime ideal p\mathfrak{p} is a point of Spec⁡A\operatorname{Spec} A. The residue field at p\mathfrak{p} is κ(p)=Frac⁡(A/p)\kappa(\mathfrak{p}) = \operatorname{Frac}(A/\mathfrak{p}).

ExampleSpec ℤ

Spec⁡Z={(0),(2),(3),(5),(7),(11),…}.\operatorname{Spec} \mathbb{Z} = \{(0), (2), (3), (5), (7), (11), \ldots\}.

  • The closed points are (p)(p) for primes pp, with residue field κ((p))=Fp\kappa((p)) = \mathbb{F}_p.
  • The generic point (0)(0) has residue field κ((0))=Q\kappa((0)) = \mathbb{Q}.

The closure {(0)}‾=Spec⁡Z\overline{\{(0)\}} = \operatorname{Spec} \mathbb{Z} (the generic point is dense). Think of Spec⁡Z\operatorname{Spec} \mathbb{Z} as a "curve" whose points are the primes, with a fat generic point spread over the whole space.

ExampleSpec k (a field)

If kk is a field, then Spec⁡k={(0)}\operatorname{Spec} k = \{(0)\}: a single point. This is the terminal object in the category of affine schemes (over kk, Spec⁡k\operatorname{Spec} k plays the role of a point).

ExampleSpec k[x] — the affine line

Spec⁡k[x]\operatorname{Spec} k[x] for kk algebraically closed:

  • Closed points: (x−a)(x - a) for each a∈ka \in k, with κ=k\kappa = k. These correspond to the "classical" points of A1\mathbb{A}^1.
  • Generic point: (0)(0), with κ=k(x)\kappa = k(x) (the function field).

So Spec⁡k[x]=A1∪{generic point}\operatorname{Spec} k[x] = \mathbb{A}^1 \cup \{\text{generic point}\}. The generic point is dense; its closure is the entire space.

For k=Rk = \mathbb{R} (not algebraically closed): Spec⁡R[x]\operatorname{Spec} \mathbb{R}[x] has additional closed points like (x2+1)(x^2 + 1) with residue field C\mathbb{C}. These "non-rational points" correspond to conjugate pairs ±i\pm i.

ExampleSpec k[x, y] — the affine plane

Spec⁡k[x,y]\operatorname{Spec} k[x, y] for k=kˉk = \bar{k} has three types of points:

| Type | Prime ideal | Residue field | Dimension | |---|---|---|---| | Closed points | (x−a,y−b)(x - a, y - b) | kk | 00 | | Generic points of curves | (f(x,y))(f(x,y)), ff irred. | k(x,y)/(f)=k(C)k(x,y)/(f) = k(C) | 11 | | Generic point | (0)(0) | k(x,y)k(x,y) | 22 |

The point (y−x2)(y - x^2) is the "generic point of the parabola" — its closure {(y−x2)}‾=V(y−x2)\overline{\{(y-x^2)\}} = V(y - x^2) is the parabola. Every irreducible subvariety has a unique generic point.

ExampleSpec k[ε]/(ε²) — the dual numbers

Spec⁡k[ε]/(ε2)\operatorname{Spec} k[\varepsilon]/(\varepsilon^2) has a single point (ε)({\varepsilon}), but the ring is not a field: it has a nilpotent element ε≠0\varepsilon \neq 0 with ε2=0\varepsilon^2 = 0.

This is a "fat point" or "point with tangent direction." A morphism Spec⁡k[ε]/(ε2)→X\operatorname{Spec} k[\varepsilon]/(\varepsilon^2) \to X corresponds to a point p∈Xp \in X together with a tangent vector at pp. This is why scheme theory is more flexible than classical algebraic geometry: nilpotents carry infinitesimal information.

ExampleSpec of a local ring

For a local ring (R,m)(R, \mathfrak{m}), Spec⁡R\operatorname{Spec} R has a unique closed point m\mathfrak{m}. Examples:

  • Spec⁡k[[x]]\operatorname{Spec} k[[x]] has two points: the closed point (x)(x) and the generic point (0)(0). Think of it as a "formal neighborhood of a point on a curve."
  • Spec⁡Z(p)\operatorname{Spec} \mathbb{Z}_{(p)} has two points: (p)(p) (closed) and (0)(0) (generic). This is the "local picture of Spec⁡Z\operatorname{Spec} \mathbb{Z} at the prime pp."
  • Spec⁡k\operatorname{Spec} k is the special case where m=(0)\mathfrak{m} = (0).
ExampleSpec of a product ring

Spec⁡(A×B)≅Spec⁡A⊔Spec⁡B\operatorname{Spec}(A \times B) \cong \operatorname{Spec} A \sqcup \operatorname{Spec} B (disjoint union). For example:

Spec⁡(k×k)={(1,0),(0,1)}=two points.\operatorname{Spec}(k \times k) = \{(1,0), (0,1)\} = \text{two points}.

More interesting: by CRT, Spec⁡Z/6Z≅Spec⁡(Z/2×Z/3)={(2),(3)}\operatorname{Spec} \mathbb{Z}/6\mathbb{Z} \cong \operatorname{Spec}(\mathbb{Z}/2 \times \mathbb{Z}/3) = \{(2), (3)\}, two points.

But Spec⁡Z/4Z\operatorname{Spec} \mathbb{Z}/4\mathbb{Z} is a single point (2)(2) with nilpotent structure (2ˉ2=0ˉ\bar{2}^2 = \bar{0}). Topologically it's the same as Spec⁡F2\operatorname{Spec} \mathbb{F}_2, but the scheme structure remembers the multiplicity.


The Zariski topology

Definition2.6Zariski topology on Spec A

For an ideal I⊆AI \subseteq A, define the vanishing set

V(I)={p∈Spec⁡A∣p⊇I}.V(I) = \{\mathfrak{p} \in \operatorname{Spec} A \mid \mathfrak{p} \supseteq I\}.

The sets V(I)V(I) are the closed sets of the Zariski topology on Spec⁡A\operatorname{Spec} A. Properties:

  • V(A)=∅V(A) = \emptyset, V(0)=Spec⁡AV(0) = \operatorname{Spec} A.
  • V(I)∪V(J)=V(IJ)=V(I∩J)V(I) \cup V(J) = V(IJ) = V(I \cap J).
  • ⋂αV(Iα)=V(∑αIα)\bigcap_\alpha V(I_\alpha) = V(\sum_\alpha I_\alpha).

The distinguished (basic) open sets D(f)=Spec⁡A∖V(f)={p∣f∉p}D(f) = \operatorname{Spec} A \setminus V(f) = \{\mathfrak{p} \mid f \notin \mathfrak{p}\} for f∈Af \in A form a basis for the topology.

ExampleZariski topology on 𝔸¹

Spec⁡k[x]\operatorname{Spec} k[x]: the closed sets are V(f(x))V(f(x)) for f∈k[x]f \in k[x]. Since k[x]k[x] is a PID, every closed set is either all of Spec⁡k[x]\operatorname{Spec} k[x] or a finite set of closed points. So the topology is the cofinite topology on the closed points, plus the generic point whose closure is everything.

This is highly non-Hausdorff: any two nonempty open sets intersect.

ExampleTopology of Spec ℤ

The closed sets of Spec⁡Z\operatorname{Spec} \mathbb{Z} are: ∅\emptyset, Spec⁡Z\operatorname{Spec} \mathbb{Z}, and finite sets of closed points {(p1),…,(pn)}=V(p1⋯pn)\{(p_1), \ldots, (p_n)\} = V(p_1 \cdots p_n). Same cofinite topology on the closed points.

The open set D(6)=Spec⁡Z∖{(2),(3)}D(6) = \operatorname{Spec} \mathbb{Z} \setminus \{(2), (3)\} consists of primes p≠2,3p \neq 2, 3 plus the generic point. As a scheme: D(6)≅Spec⁡Z[1/6]D(6) \cong \operatorname{Spec} \mathbb{Z}[1/6].

ExampleZariski topology on 𝔸²

Spec⁡k[x,y]\operatorname{Spec} k[x,y]: closed sets include V(f)V(f) for irreducible ff (curves), V(x−a,y−b)V(x-a, y-b) (points), and intersections/unions. The topology is much richer than in dimension 1:

  • V(y−x2)V(y - x^2) is a closed parabola.
  • V(xy)V(xy) is the union of the two axes.
  • V(x−a,y−b)={(a,b)}V(x - a, y - b) = \{(a,b)\} is a closed point.

The generic point of V(y−x2)V(y - x^2) is the prime (y−x2)(y - x^2); its closure is the parabola. The generic point (0)(0) is dense in everything.


The structure sheaf

Definition2.7Structure sheaf

On Spec⁡A\operatorname{Spec} A, the structure sheaf OSpec⁡A\mathcal{O}_{\operatorname{Spec} A} is defined on the basis of distinguished opens by:

O(D(f))=Af(localization of A at f).\mathcal{O}(D(f)) = A_f \quad (\text{localization of } A \text{ at } f).

In particular, O(Spec⁡A)=A\mathcal{O}(\operatorname{Spec} A) = A (global sections recover the ring).

The stalk at a prime p\mathfrak{p} is the local ring:

OSpec⁡A,p=Ap={as∣a∈A, s∉p}.\mathcal{O}_{\operatorname{Spec} A, \mathfrak{p}} = A_\mathfrak{p} = \left\{\frac{a}{s} \mid a \in A, \ s \notin \mathfrak{p}\right\}.

The pair (Spec⁡A,OSpec⁡A)(\operatorname{Spec} A, \mathcal{O}_{\operatorname{Spec} A}) is an affine scheme.

ExampleStructure sheaf of Spec ℤ
  • O(Spec⁡Z)=Z\mathcal{O}(\operatorname{Spec} \mathbb{Z}) = \mathbb{Z} (global sections).
  • O(D(p))=Z[1/p]\mathcal{O}(D(p)) = \mathbb{Z}[1/p] (integers with pp inverted).
  • O(D(6))=Z[1/6]=Z[1/2,1/3]\mathcal{O}(D(6)) = \mathbb{Z}[1/6] = \mathbb{Z}[1/2, 1/3].
  • Stalk at (p)(p): OSpec⁡Z,(p)=Z(p)\mathcal{O}_{\operatorname{Spec}\mathbb{Z}, (p)} = \mathbb{Z}_{(p)} (integers localized at pp).
  • Stalk at (0)(0): OSpec⁡Z,(0)=Q\mathcal{O}_{\operatorname{Spec}\mathbb{Z}, (0)} = \mathbb{Q} (the function field).
ExampleStructure sheaf of 𝔸¹

For k[x]k[x]:

  • O(A1)=k[x]\mathcal{O}(\mathbb{A}^1) = k[x].
  • O(D(x))=k[x,1/x]\mathcal{O}(D(x)) = k[x, 1/x] — polynomials in xx and 1/x1/x.
  • O(D(x(x−1)))=k[x,1/(x(x−1))]\mathcal{O}(D(x(x-1))) = k[x, 1/(x(x-1))] — rational functions regular away from 00 and 11.
  • OA1,(x−a)=k[x](x−a)\mathcal{O}_{\mathbb{A}^1, (x-a)} = k[x]_{(x-a)} — rational functions regular at aa.
ExampleStructure sheaf of a nodal curve

Let A=k[x,y]/(y2−x2(x+1))A = k[x,y]/(y^2 - x^2(x+1)) (nodal cubic). Then Spec⁡A\operatorname{Spec} A is the affine nodal curve. At the node m=(x,y)\mathfrak{m} = (x, y):

OSpec⁡A,m=Am\mathcal{O}_{\operatorname{Spec} A, \mathfrak{m}} = A_\mathfrak{m}

is a local ring that is not a domain (the completion A^m≅k[[u,v]]/(uv)\hat{A}_\mathfrak{m} \cong k[[u,v]]/(uv), two branches). The local ring detects the singularity: dim⁡km/m2=2>dim⁡=1\dim_k \mathfrak{m}/\mathfrak{m}^2 = 2 > \dim = 1, so the point is singular.


Functoriality: ring homomorphisms become continuous maps

Definition2.8Induced map on spectra

A ring homomorphism φ:A→B\varphi : A \to B induces a continuous map

φ∗=aφ:Spec⁡B→Spec⁡A,q↦φ−1(q).\varphi^* = {}^a\varphi : \operatorname{Spec} B \to \operatorname{Spec} A, \quad \mathfrak{q} \mapsto \varphi^{-1}(\mathfrak{q}).

(The preimage of a prime ideal is prime.) This gives a contravariant functor Spec⁡:CRingop→Top\operatorname{Spec} : \mathbf{CRing}^{\mathrm{op}} \to \mathbf{Top}.

Moreover, φ\varphi induces a morphism of structure sheaves φ#:OSpec⁡A→(φ∗)∗OSpec⁡B\varphi^\# : \mathcal{O}_{\operatorname{Spec} A} \to (\varphi^*)_*\mathcal{O}_{\operatorname{Spec} B}, making (φ∗,φ#)(\varphi^*, \varphi^\#) a morphism of locally ringed spaces.

ExampleSurjection → closed immersion

The surjection k[x,y]↠k[x,y]/(y−x2)≅k[x]k[x,y] \twoheadrightarrow k[x,y]/(y - x^2) \cong k[x] induces

Spec⁡k[x]↪Spec⁡k[x,y]=A2\operatorname{Spec} k[x] \hookrightarrow \operatorname{Spec} k[x,y] = \mathbb{A}^2

which is the inclusion of the parabola y=x2y = x^2 into the plane. In general, surjections A↠A/IA \twoheadrightarrow A/I correspond to closed immersions V(I)↪Spec⁡AV(I) \hookrightarrow \operatorname{Spec} A.

ExampleLocalization → open immersion

The localization k[x]→k[x,1/x]k[x] \to k[x, 1/x] induces

Spec⁡k[x,1/x]↪Spec⁡k[x]=A1\operatorname{Spec} k[x, 1/x] \hookrightarrow \operatorname{Spec} k[x] = \mathbb{A}^1

which is the inclusion of D(x)=A1∖{0}D(x) = \mathbb{A}^1 \setminus \{0\}. Localization maps A→S−1AA \to S^{-1}A correspond to open immersions into distinguished opens.

ExampleFrobenius in characteristic p

In characteristic p>0p > 0, the Frobenius φ:A→A\varphi : A \to A, a↦apa \mapsto a^p induces Spec⁡A→Spec⁡A\operatorname{Spec} A \to \operatorname{Spec} A. On A1=Spec⁡Fp[x]\mathbb{A}^1 = \operatorname{Spec} \mathbb{F}_p[x]:

φ∗:Spec⁡Fp[x]→Spec⁡Fp[x],(x−a)↦(x−ap)=(x−a)\varphi^* : \operatorname{Spec} \mathbb{F}_p[x] \to \operatorname{Spec} \mathbb{F}_p[x], \quad (x - a) \mapsto (x - a^p) = (x - a)

since ap=aa^p = a in Fp\mathbb{F}_p (Fermat's little theorem). So on Fp\mathbb{F}_p-points, the Frobenius is the identity! But on the scheme level it's a nontrivial degree-pp morphism. The fixed points of Frobenius on Fp‾\overline{\mathbb{F}_p}-points are exactly the Fp\mathbb{F}_p-rational points — this is the foundation of the Weil conjectures.

ExampleSpec 𝔽ₚ → Spec ℤ — a fiber of the arithmetic curve

The quotient Z↠Fp\mathbb{Z} \twoheadrightarrow \mathbb{F}_p induces Spec⁡Fp↪Spec⁡Z\operatorname{Spec} \mathbb{F}_p \hookrightarrow \operatorname{Spec} \mathbb{Z}, the inclusion of the closed point (p)(p). Think of this as the fiber of the "arithmetic curve" Spec⁡Z\operatorname{Spec} \mathbb{Z} over the "prime pp." The inclusion Z↪Q\mathbb{Z} \hookrightarrow \mathbb{Q} gives Spec⁡Q↪Spec⁡Z\operatorname{Spec} \mathbb{Q} \hookrightarrow \operatorname{Spec} \mathbb{Z}: the generic fiber.

An equation like y2=x3−xy^2 = x^3 - x over Z\mathbb{Z} gives a scheme over Spec⁡Z\operatorname{Spec} \mathbb{Z} whose fibers are:

  • Over (0)(0): the elliptic curve E/QE/\mathbb{Q}.
  • Over (p)(p): the reduction E/FpE/\mathbb{F}_p (smooth for p≠2p \neq 2, singular at p=2p = 2).

This "spreading out" perspective is central to arithmetic geometry.


Generic points and specialization

Definition2.9Generic point

A point η∈X\eta \in X is a generic point of an irreducible closed subset ZZ if {η}‾=Z\overline{\{\eta\}} = Z. In Spec⁡A\operatorname{Spec} A, every irreducible closed subset V(p)V(\mathfrak{p}) has a unique generic point, namely p\mathfrak{p} itself:

{p}‾=V(p).\overline{\{\mathfrak{p}\}} = V(\mathfrak{p}).

We say p\mathfrak{p} specializes to q\mathfrak{q} (written p⇝q\mathfrak{p} \leadsto \mathfrak{q}) if q∈{p}‾\mathfrak{q} \in \overline{\{\mathfrak{p}\}}, i.e., p⊆q\mathfrak{p} \subseteq \mathfrak{q}.

ExampleSpecialization in 𝔸²

In Spec⁡k[x,y]\operatorname{Spec} k[x,y]:

(0)⇝(y−x2)⇝(x−1,y−1)(0) \leadsto (y - x^2) \leadsto (x - 1, y - 1)

The generic point (0)(0) specializes to the generic point of the parabola, which specializes to the point (1,1)(1,1) on the parabola. This chain corresponds to the inclusion of primes:

(0)⊂(y−x2)⊂(x−1,y−1).(0) \subset (y - x^2) \subset (x-1, y-1).

The chain length is 2=dim⁡A22 = \dim \mathbb{A}^2, illustrating that the Krull dimension equals the maximal chain length.

ExampleGeometric meaning of the generic point

The generic point η=(0)∈Spec⁡k[x]\eta = (0) \in \operatorname{Spec} k[x] is the "point in general position." Evaluating a polynomial f(x)f(x) at η\eta gives ff itself in k(x)k(x). So:

  • f(η)=0f(\eta) = 0 in κ(η)=k(x)\kappa(\eta) = k(x)   ⟺  \iff f=0f = 0 as a polynomial   ⟺  \iff ff vanishes everywhere.
  • f(η)≠0f(\eta) \neq 0   ⟺  \iff f≠0f \neq 0   ⟺  \iff ff vanishes on only finitely many closed points.

A property holds "at the generic point" iff it holds "on a dense open set." This is the scheme-theoretic formulation of "a general polynomial of degree dd has dd distinct roots."


Spec and classical varieties

RemarkFrom varieties to schemes

For kk algebraically closed, the closed points of Spec⁡k[x1,…,xn]/I\operatorname{Spec} k[x_1, \ldots, x_n]/I are in bijection with V(I)⊆AnV(I) \subseteq \mathbb{A}^n (by the Nullstellensatz: maximal ideals =(x1−a1,…,xn−an)= (x_1 - a_1, \ldots, x_n - a_n) with (a1,…,an)∈V(I)(a_1, \ldots, a_n) \in V(I)).

So Spec⁡\operatorname{Spec} extends classical algebraic geometry by adding:

  1. Generic points of irreducible subvarieties (one for each prime ideal).
  2. Non-closed-field points (primes with residue field ≠k\neq k, relevant over non-algebraically-closed fields).
  3. Nilpotent structure (e.g., Spec⁡k[x]/(x2)\operatorname{Spec} k[x]/(x^2) vs. Spec⁡k\operatorname{Spec} k).
ExampleSpec of a non-reduced ring: double point

Spec⁡k[x]/(x2)\operatorname{Spec} k[x]/(x^2) and Spec⁡k[x]/(x)\operatorname{Spec} k[x]/(x) have the same underlying topological space (a single point (x)(x)). But as schemes they differ:

  • Spec⁡k\operatorname{Spec} k has ring kk (reduced).
  • Spec⁡k[x]/(x2)\operatorname{Spec} k[x]/(x^2) has ring k[ε]k[\varepsilon] with ε2=0\varepsilon^2 = 0 (non-reduced, "double point").

The scheme V(x2)⊆A1V(x^2) \subseteq \mathbb{A}^1 remembers that the origin has multiplicity 2. Similarly, V(xn)=Spec⁡k[x]/(xn)V(x^n) = \operatorname{Spec} k[x]/(x^n) is the origin with multiplicity nn.

ExampleIntersection multiplicity via Spec

In A2\mathbb{A}^2, consider the intersection of y=0y = 0 and y=x2y = x^2:

Spec⁡k[x,y]/(y,y−x2)=Spec⁡k[x]/(x2).\operatorname{Spec} k[x,y]/(y, y - x^2) = \operatorname{Spec} k[x]/(x^2).

This is a double point — the intersection has multiplicity 22 at the origin because the parabola is tangent to the xx-axis. Compare with y=0∩y=xy = 0 \cap y = x:

Spec⁡k[x,y]/(y,y−x)=Spec⁡k[x]/(x)=Spec⁡k.\operatorname{Spec} k[x,y]/(y, y - x) = \operatorname{Spec} k[x]/(x) = \operatorname{Spec} k.

A simple (reduced) point — transverse intersection has multiplicity 11. Scheme theory naturally computes intersection multiplicities via the tensor product of rings.


Key properties of Spec

RemarkProperties

Spec⁡A\operatorname{Spec} A satisfies:

  1. Quasi-compact: Every open cover has a finite subcover (Spec⁡A=⋃D(fi)  ⟹  (f1,…,fn)=A\operatorname{Spec} A = \bigcup D(f_i) \implies (f_1, \ldots, f_n) = A).
  2. Sober: Every irreducible closed subset has a unique generic point.
  3. T0T_0 but not T1T_1: Points can be distinguished by open sets, but not every singleton is closed.
  4. Connected   ⟺  \iff AA has no nontrivial idempotents.
  5. Irreducible   ⟺  \iff nil⁡(A)\operatorname{nil}(A) is prime (  ⟺  \iff AredA_{\mathrm{red}} is a domain).
  6. Noetherian   ⟺  \iff AA is Noetherian (= every open is quasi-compact).

The functor Spec⁡:CRingop→AffSch\operatorname{Spec} : \mathbf{CRing}^{\mathrm{op}} \to \mathbf{AffSch} is an equivalence of categories (with quasi-inverse Γ\Gamma: global sections). This is the affine analogue of the equivalence between affine varieties and finitely generated reduced kk-algebras — but now with no restrictions on the ring.


Summary: dictionary Algebra ↔ Geometry

RemarkThe algebra-geometry dictionary for Spec
Algebra (AA)Geometry (Spec⁡A\operatorname{Spec} A)
Prime ideal p\mathfrak{p}Point of Spec⁡A\operatorname{Spec} A
Maximal ideal m\mathfrak{m}Closed point
Minimal primeGeneric point of irreducible component
p⊆q\mathfrak{p} \subseteq \mathfrak{q}p\mathfrak{p} specializes to q\mathfrak{q}
Quotient A↠A/IA \twoheadrightarrow A/IClosed subscheme V(I)↪Spec⁡AV(I) \hookrightarrow \operatorname{Spec} A
Localization A→AfA \to A_fOpen immersion D(f)↪Spec⁡AD(f) \hookrightarrow \operatorname{Spec} A
AA reduced (nil⁡(A)=0\operatorname{nil}(A) = 0)No embedded nilpotent structure
AA domainSpec⁡A\operatorname{Spec} A irreducible
AA localUnique closed point
AA fieldSingle point (reduced)
AA NoetherianNoetherian topological space
Krull dim⁡A\operatorname{Krull\,dim} Adim⁡Spec⁡A\dim \operatorname{Spec} A
A⊗kBA \otimes_k BSpec⁡A×kSpec⁡B\operatorname{Spec} A \times_k \operatorname{Spec} B (fiber product)
ApA_\mathfrak{p} regular localSpec⁡A\operatorname{Spec} A smooth at p\mathfrak{p}