ConceptComplete

Schemes

A scheme is a space that locally looks like the spectrum of a ring. Just as a manifold is built by gluing open subsets of Rn\mathbb{R}^n, a scheme is built by gluing affine schemes Spec⁑A\operatorname{Spec} A. This definition, due to Grothendieck, unifies algebraic geometry over arbitrary base rings and allows nilpotent elements, non-closed-field points, and arithmetic geometry.


From affine schemes to general schemes

Definition2.10Scheme

A scheme is a locally ringed space (X,OX)(X, \mathcal{O}_X) that is locally affine: every point x∈Xx \in X has an open neighborhood UU such that (U,OX∣U)(U, \mathcal{O}_X|_U) is isomorphic (as a locally ringed space) to Spec⁑A\operatorname{Spec} A for some ring AA.

An affine scheme is one of the form Spec⁑A\operatorname{Spec} A. A general scheme is obtained by gluing affine schemes along open subsets.

ExampleThe affine line 𝔸¹

Ak1=Spec⁑k[x]\mathbb{A}^1_k = \operatorname{Spec} k[x] is affine. For k=kΛ‰k = \bar{k}, the closed points correspond to elements a∈ka \in k (via the maximal ideal (xβˆ’a)(x - a)), plus the generic point (0)(0).

Over Z\mathbb{Z}: AZ1=Spec⁑Z[x]\mathbb{A}^1_\mathbb{Z} = \operatorname{Spec} \mathbb{Z}[x]. This is a 22-dimensional scheme: the "arithmetic surface." Its points include:

  • Closed points (p,f(x))(p, f(x)) where pp is prime and ff is irreducible mod pp β€” points of AFp1\mathbb{A}^1_{\mathbb{F}_p}.
  • Generic points of "horizontal curves" (g(x))(g(x)) where g∈Z[x]g \in \mathbb{Z}[x] is irreducible.
  • Generic points of "vertical fibers" (p)(p).
  • The generic point (0)(0).
ExampleThe projective line β„™ΒΉ

Pk1\mathbb{P}^1_k is the simplest non-affine scheme. It is obtained by gluing two copies of A1\mathbb{A}^1:

U0=Spec⁑k[t]andU1=Spec⁑k[s]U_0 = \operatorname{Spec} k[t] \quad \text{and} \quad U_1 = \operatorname{Spec} k[s]

along D(t)β‰…D(s)D(t) \cong D(s) via s=1/ts = 1/t. Concretely:

  • U0∩U1=Spec⁑k[t,1/t]U_0 \cap U_1 = \operatorname{Spec} k[t, 1/t],
  • the gluing isomorphism is k[t,1/t]β‰…k[s,1/s]k[t, 1/t] \cong k[s, 1/s] via t↔1/st \leftrightarrow 1/s.

P1\mathbb{P}^1 is not affine: O(P1)=k\mathcal{O}(\mathbb{P}^1) = k (global sections are constant), but P1\mathbb{P}^1 is not a point. This shows that the functor Γ:Sch→CRingop\Gamma : \mathbf{Sch} \to \mathbf{CRing}^{\mathrm{op}} is not an equivalence on non-affine schemes.

ExampleProjective n-space ℙⁿ

Pkn=Proj⁑k[x0,…,xn]\mathbb{P}^n_k = \operatorname{Proj} k[x_0, \ldots, x_n] is covered by n+1n+1 affine charts:

Ui=Spec⁑k[x0xi,…,xixi^,…,xnxi]β‰…Akn,i=0,…,n.U_i = \operatorname{Spec} k\left[\frac{x_0}{x_i}, \ldots, \widehat{\frac{x_i}{x_i}}, \ldots, \frac{x_n}{x_i}\right] \cong \mathbb{A}^n_k, \quad i = 0, \ldots, n.

Key properties:

  • O(Pn)=k\mathcal{O}(\mathbb{P}^n) = k (only constant global functions).
  • dim⁑Pn=n\dim \mathbb{P}^n = n.
  • Pn\mathbb{P}^n is proper over kk (the algebraic analogue of compact).
  • Pic⁑(Pn)=Z\operatorname{Pic}(\mathbb{P}^n) = \mathbb{Z}, generated by O(1)\mathcal{O}(1).

Proj construction

Definition2.11Proj

Let S=⨁dβ‰₯0SdS = \bigoplus_{d \geq 0} S_d be a graded ring. The Proj construction defines:

Proj⁑S={p∈Spec⁑S∣pΒ isΒ homogeneousΒ andΒ pβŠ‡ΜΈS+}\operatorname{Proj} S = \{\mathfrak{p} \in \operatorname{Spec} S \mid \mathfrak{p} \text{ is homogeneous and } \mathfrak{p} \not\supseteq S_+\}

where S+=⨁d>0SdS_+ = \bigoplus_{d > 0} S_d is the irrelevant ideal. The topology has closed sets V+(I)={p∈Proj⁑S∣pβŠ‡I}V_+(I) = \{\mathfrak{p} \in \operatorname{Proj} S \mid \mathfrak{p} \supseteq I\} for homogeneous ideals II.

For f∈Sdf \in S_d (d>0d > 0), the distinguished open D+(f)=Spec⁑S(f)D_+(f) = \operatorname{Spec} S_{(f)} is affine, where S(f)=(Sf)0S_{(f)} = (S_f)_0 is the degree-00 part of the localization.

ExampleProj of a polynomial ring

Proj⁑k[x0,x1]=Pk1\operatorname{Proj} k[x_0, x_1] = \mathbb{P}^1_k. The two standard opens:

  • D+(x0)=Spec⁑k[x1/x0]=Spec⁑k[t]=A1D_+(x_0) = \operatorname{Spec} k[x_1/x_0] = \operatorname{Spec} k[t] = \mathbb{A}^1.
  • D+(x1)=Spec⁑k[x0/x1]=Spec⁑k[s]=A1D_+(x_1) = \operatorname{Spec} k[x_0/x_1] = \operatorname{Spec} k[s] = \mathbb{A}^1.

More generally, Proj⁑k[x0,…,xn]=Pkn\operatorname{Proj} k[x_0, \ldots, x_n] = \mathbb{P}^n_k.

ExampleProjective hypersurfaces via Proj

A projective variety V(F)βŠ†PnV(F) \subseteq \mathbb{P}^n for homogeneous FF of degree dd is:

V(F)=Proj⁑k[x0,…,xn]/(F).V(F) = \operatorname{Proj} k[x_0, \ldots, x_n]/(F).

For example, the elliptic curve E=V(Y2Zβˆ’X3+XZ2)βŠ†P2E = V(Y^2Z - X^3 + XZ^2) \subseteq \mathbb{P}^2 is:

E=Proj⁑k[X,Y,Z]/(Y2Zβˆ’X3+XZ2).E = \operatorname{Proj} k[X, Y, Z]/(Y^2Z - X^3 + XZ^2).

In the affine chart Zβ‰ 0Z \neq 0: Spec⁑k[x,y]/(y2βˆ’x3+x)\operatorname{Spec} k[x,y]/(y^2 - x^3 + x), where x=X/Zx = X/Z, y=Y/Zy = Y/Z.

ExampleVeronese subrings and Proj

The dd-th Veronese subring of S=k[x0,…,xn]S = k[x_0, \ldots, x_n] is S(d)=⨁kSkdS^{(d)} = \bigoplus_k S_{kd}. Then Proj⁑S(d)β‰…Proj⁑S=Pn\operatorname{Proj} S^{(d)} \cong \operatorname{Proj} S = \mathbb{P}^n (same scheme, different embedding). This is the scheme-theoretic version of the Veronese embedding.

ExampleBlowup as Proj

The blowup of A2\mathbb{A}^2 at the origin is:

Bl⁑0A2=Proj⁑k[x,y][s,t]/(xtβˆ’ys)βŠ†A2Γ—P1\operatorname{Bl}_0 \mathbb{A}^2 = \operatorname{Proj} k[x,y][s,t]/(xt - ys) \subseteq \mathbb{A}^2 \times \mathbb{P}^1

where [s:t][s:t] are the homogeneous coordinates on P1\mathbb{P}^1. This is covered by two affine charts:

  • sβ‰ 0s \neq 0: Spec⁑k[x,t/s]=Spec⁑k[x,u]\operatorname{Spec} k[x, t/s] = \operatorname{Spec} k[x, u] where y=xuy = xu β€” the chart where u=y/xu = y/x.
  • tβ‰ 0t \neq 0: Spec⁑k[y,s/t]=Spec⁑k[y,v]\operatorname{Spec} k[y, s/t] = \operatorname{Spec} k[y, v] where x=yvx = yv β€” the chart where v=x/yv = x/y.

The exceptional divisor E≅P1E \cong \mathbb{P}^1 is the preimage of the origin.


Properties of schemes

Definition2.12Scheme properties

A scheme XX is:

  • Reduced if OX,x\mathcal{O}_{X,x} has no nilpotents for all xx (equivalently, OX(U)\mathcal{O}_X(U) is reduced for all UU).
  • Integral (= reduced + irreducible) if OX(U)\mathcal{O}_X(U) is a domain for all nonempty affine UU.
  • Noetherian if it has a finite cover by spectra of Noetherian rings.
  • Locally Noetherian if every point has a Noetherian open neighborhood.
  • Connected if the underlying topological space is connected.
  • Irreducible if the underlying space is irreducible (every open set is dense).
ExampleNon-reduced schemes
  • Spec⁑k[x]/(x2)\operatorname{Spec} k[x]/(x^2): the double point. One point, but O=k[Ξ΅]\mathcal{O} = k[\varepsilon] with Ξ΅2=0\varepsilon^2 = 0.
  • Spec⁑k[x,y]/(x2,xy)\operatorname{Spec} k[x,y]/(x^2, xy): a "fuzzy point" at the origin, non-reduced in the xx-direction.
  • The scheme-theoretic intersection V(y)∩V(yβˆ’x2)=Spec⁑k[x,y]/(y,yβˆ’x2)=Spec⁑k[x]/(x2)V(y) \cap V(y - x^2) = \operatorname{Spec} k[x,y]/(y, y-x^2) = \operatorname{Spec} k[x]/(x^2): a double point recording the tangency.

Non-reduced schemes arise naturally from:

  • Intersections with multiplicity.
  • Fibers of morphisms at critical values.
  • Deformation theory (flat limits can acquire nilpotents).
ExampleNon-connected scheme

Spec⁑(kΓ—k)=Spec⁑kβŠ”Spec⁑k\operatorname{Spec}(k \times k) = \operatorname{Spec} k \sqcup \operatorname{Spec} k: two disjoint points.

Spec⁑k[x]/(x2βˆ’1)=Spec⁑k[x]/((xβˆ’1)(x+1))β‰…Spec⁑(kΓ—k)\operatorname{Spec} k[x]/(x^2 - 1) = \operatorname{Spec} k[x]/((x-1)(x+1)) \cong \operatorname{Spec}(k \times k): two points x=1x = 1 and x=βˆ’1x = -1. But over F2\mathbb{F}_2: x2βˆ’1=(xβˆ’1)2x^2 - 1 = (x-1)^2, so Spec⁑F2[x]/(xβˆ’1)2\operatorname{Spec} \mathbb{F}_2[x]/(x-1)^2 is a single (double) point β€” connected but non-reduced!

ExampleNon-Noetherian scheme

Spec⁑k[x1,x2,x3,…]\operatorname{Spec} k[x_1, x_2, x_3, \ldots] (polynomial ring in infinitely many variables) is an affine scheme that is not Noetherian: the chain of ideals (x1)βŠ‚(x1,x2)βŠ‚β‹―(x_1) \subset (x_1, x_2) \subset \cdots does not stabilize. In practice, most schemes in algebraic geometry are locally Noetherian.


Subschemes

Definition2.13Open and closed subschemes

An open subscheme of (X,OX)(X, \mathcal{O}_X) is (U,OX∣U)(U, \mathcal{O}_X|_U) for an open UβŠ†XU \subseteq X.

A closed subscheme is determined by a quasi-coherent sheaf of ideals IβŠ†OX\mathcal{I} \subseteq \mathcal{O}_X. The underlying space is Supp⁑(OX/I)\operatorname{Supp}(\mathcal{O}_X/\mathcal{I}) and the structure sheaf is OX/I\mathcal{O}_X/\mathcal{I}.

For an affine scheme Spec⁑A\operatorname{Spec} A:

  • Open subschemes: D(f)=Spec⁑AfD(f) = \operatorname{Spec} A_f and unions thereof.
  • Closed subschemes: V(I)=Spec⁑A/IV(I) = \operatorname{Spec} A/I for ideals IβŠ†AI \subseteq A.
ExampleClosed subschemes with the same support

The following are all closed subschemes of A1=Spec⁑k[x]\mathbb{A}^1 = \operatorname{Spec} k[x] supported at the origin:

| Subscheme | Ideal | Ring | "Thickness" | |---|---|---|---| | Spec⁑k\operatorname{Spec} k | (x)(x) | kk | reduced point | | Spec⁑k[x]/(x2)\operatorname{Spec} k[x]/(x^2) | (x2)(x^2) | k[Ρ]k[\varepsilon] | double point | | Spec⁑k[x]/(x3)\operatorname{Spec} k[x]/(x^3) | (x3)(x^3) | k[x]/(x3)k[x]/(x^3) | triple point | | Spec⁑k[x]/(xn)\operatorname{Spec} k[x]/(x^n) | (xn)(x^n) | k[x]/(xn)k[x]/(x^n) | nn-fold point |

As nβ†’βˆžn \to \infty, the limit is Spec⁑k[[x]]\operatorname{Spec} k[[x]], the formal neighborhood of the origin.

ExampleScheme-theoretic intersection vs. set-theoretic

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

  • Set-theoretic intersection: the origin {(0,0)}\{(0,0)\}.
  • Scheme-theoretic intersection: Spec⁑k[x,y]/(y,yβˆ’x2)=Spec⁑k[x]/(x2)\operatorname{Spec} k[x,y]/(y, y-x^2) = \operatorname{Spec} k[x]/(x^2) β€” a double point.

The curves y=xy = x and y=0y = 0:

  • Scheme-theoretic intersection: Spec⁑k[x,y]/(y,yβˆ’x)=Spec⁑k[x]/(x)=Spec⁑k\operatorname{Spec} k[x,y]/(y, y-x) = \operatorname{Spec} k[x]/(x) = \operatorname{Spec} k β€” a simple point.

Scheme-theoretic intersection detects tangency. This is the correct notion for BΓ©zout's theorem.


Schemes over a base

Definition2.14S-scheme

A scheme over SS (or SS-scheme) is a scheme XX equipped with a morphism X→SX \to S (the structure morphism). The category of SS-schemes is the slice category Sch/S\mathbf{Sch}/S.

Most commonly:

  • S=Spec⁑kS = \operatorname{Spec} k: schemes over a field (kk-schemes). All of classical algebraic geometry lives here.
  • S=Spec⁑ZS = \operatorname{Spec} \mathbb{Z}: all schemes are Z\mathbb{Z}-schemes (since Z\mathbb{Z} is initial in CRing\mathbf{CRing}).
  • S=Spec⁑RS = \operatorname{Spec} R for a ring of integers RR: arithmetic geometry.
ExampleSchemes over Spec β„€ β€” arithmetic families

The scheme E=Proj⁑Z[X,Y,Z]/(Y2Zβˆ’X3+XZ2)E = \operatorname{Proj} \mathbb{Z}[X,Y,Z]/(Y^2Z - X^3 + XZ^2) is an elliptic curve over Spec⁑Z\operatorname{Spec} \mathbb{Z}. Its fibers:

  • Generic fiber EQE_\mathbb{Q}: the elliptic curve y2=x3βˆ’xy^2 = x^3 - x over Q\mathbb{Q}.
  • Fiber at (p)(p): EFp=EΓ—Spec⁑ZSpec⁑FpE_{\mathbb{F}_p} = E \times_{\operatorname{Spec}\mathbb{Z}} \operatorname{Spec} \mathbb{F}_p β€” the reduction mod pp.

The number of Fp\mathbb{F}_p-points #E(Fp)=p+1βˆ’ap\#E(\mathbb{F}_p) = p + 1 - a_p where ∣apβˆ£β‰€2p|a_p| \leq 2\sqrt{p} (Hasse bound). The sequence {ap}\{a_p\} encodes the LL-function of EE, connecting geometry to number theory (Birch–Swinnerton-Dyer, modularity theorem, ...).

ExampleBase change / extension of scalars

For a kk-scheme XX and a field extension K/kK/k, the base change is:

XK=XΓ—Spec⁑kSpec⁑K.X_K = X \times_{\operatorname{Spec} k} \operatorname{Spec} K.

For X=V(x2+y2+1)βŠ†AR2X = V(x^2 + y^2 + 1) \subseteq \mathbb{A}^2_\mathbb{R}:

  • Over R\mathbb{R}: X(R)=βˆ…X(\mathbb{R}) = \emptyset (no real points).
  • Over C\mathbb{C}: XC=V(x2+y2+1)βŠ†AC2X_\mathbb{C} = V(x^2 + y^2 + 1) \subseteq \mathbb{A}^2_\mathbb{C} is a smooth affine curve with many points.

The scheme XX itself doesn't change β€” we just see more points after base change.


Important classes of schemes

RemarkTaxonomy of schemes
ClassDefinitionExample
AffineSpec⁑A\operatorname{Spec} AAn\mathbb{A}^n, Spec⁑Z\operatorname{Spec} \mathbb{Z}
ProjectiveClosed in PSn\mathbb{P}^n_SPn\mathbb{P}^n, elliptic curves, Grassmannians
Quasi-projectiveOpen in projectiveAn\mathbb{A}^n, Pnβˆ–V(f)\mathbb{P}^n \setminus V(f)
ProperSeparated, finite type, universally closedPn\mathbb{P}^n, complete varieties
Smooth / RegularAll local rings regularAn\mathbb{A}^n, Pn\mathbb{P}^n, smooth curves
NormalAll local rings integrally closedSmooth schemes, Spec⁑Z\operatorname{Spec} \mathbb{Z}
Cohen–MacaulayAll local rings CMNormal + equidimensional
GorensteinCM + canonical module freeComplete intersections

The hierarchy: Smooth β‡’\Rightarrow Normal β‡’\Rightarrow Cohen–Macaulay β‡’\Rightarrow Gorenstein (for local rings). All implications are strict.

ExampleGroup schemes

A group scheme over SS is a scheme G→SG \to S with morphisms μ:G×SG→G\mu : G \times_S G \to G (multiplication), e:S→Ge : S \to G (identity), ι:G→G\iota : G \to G (inverse) satisfying the group axioms.

Examples:

  • Gm=Spec⁑Z[t,tβˆ’1]\mathbb{G}_m = \operatorname{Spec} \mathbb{Z}[t, t^{-1}]: the multiplicative group. Gm(R)=Rβˆ—\mathbb{G}_m(R) = R^* for any ring RR.
  • Ga=Spec⁑Z[t]\mathbb{G}_a = \operatorname{Spec} \mathbb{Z}[t]: the additive group. Ga(R)=(R,+)\mathbb{G}_a(R) = (R, +).
  • GLn=Spec⁑Z[xij,1/det⁑]GL_n = \operatorname{Spec} \mathbb{Z}[x_{ij}, 1/\det]: the general linear group. GLn(R)=GLn(R)GL_n(R) = GL_n(R).
  • ΞΌn=Spec⁑Z[t]/(tnβˆ’1)\mu_n = \operatorname{Spec} \mathbb{Z}[t]/(t^n - 1): the nn-th roots of unity. Over a field of char p∣np \mid n, this is non-reduced (e.g., ΞΌp=Spec⁑Fp[t]/(tβˆ’1)p\mu_p = \operatorname{Spec} \mathbb{F}_p[t]/(t-1)^p).
  • Elliptic curves with their group law.
ExampleArithmetic surfaces

Spec⁑Z[x]\operatorname{Spec} \mathbb{Z}[x] is a 22-dimensional scheme, an "arithmetic surface." Its closed points correspond to pairs (p,Ξ±)(p, \alpha) where pp is prime and α∈Fpβ€Ύ\alpha \in \overline{\mathbb{F}_p} (up to Galois conjugacy).

The scheme Spec⁑Z[i]=Spec⁑Z[x]/(x2+1)\operatorname{Spec} \mathbb{Z}[i] = \operatorname{Spec} \mathbb{Z}[x]/(x^2 + 1) is a closed subscheme: the "arithmetic curve" of the Gaussian integers. Its points:

  • (1+i)(1 + i) above p=2p = 2 (ramified: 2=βˆ’i(1+i)22 = -i(1+i)^2).
  • (a+bi)(a + bi) and (aβˆ’bi)(a - bi) above p≑1(mod4)p \equiv 1 \pmod{4} (split).
  • (p)(p) above p≑3(mod4)p \equiv 3 \pmod{4} (inert).

This decomposition pattern is governed by quadratic reciprocity!


Why schemes?

RemarkAdvantages of scheme theory

Grothendieck's scheme theory extends classical algebraic geometry in essential ways:

  1. Arbitrary base rings: We can do geometry over Z\mathbb{Z}, Fp\mathbb{F}_p, pp-adic integers Zp\mathbb{Z}_p, or any ring. This unifies algebraic geometry and number theory.

  2. Nilpotents: Scheme-theoretic intersections, fibers, and limits naturally produce nilpotent structure. Ignoring it loses information (intersection multiplicities, infinitesimal deformations).

  3. Generic points: The generic point of an irreducible scheme captures "general behavior." Properties at the generic point extend to a dense open set.

  4. Relative geometry: Schemes over a base SS allow families, deformations, moduli problems. The fiber product XΓ—SYX \times_S Y is the correct notion of "product relative to SS."

  5. Representability: Many moduli problems are representable by schemes (Hilbert schemes, Picard schemes, ...) or by algebraic stacks. This gives a rigorous foundation for "spaces of geometric objects."

As Mumford wrote: "The ΰΈ„ΰΈ³ΰΈ•ΰΈ­ΰΈšto the ΰΈ„ΰΈ³ΰΈ–ΰΈ²ΰΈ‘ 'what is algebraic geometry?' is: it is the study of schemes."