ConceptComplete

Fiber Product and Base Change

The fiber product is one of the most fundamental constructions in scheme theory, providing a categorical framework for understanding intersections, fibers of morphisms, base change, and many other geometric phenomena. This concept makes precise the intuitive notion of "pulling back" geometric objects along morphisms.

Universal Property of Fiber Products

DefinitionFiber Product

Let f:XSf: X \to S and g:YSg: Y \to S be morphisms of schemes. The fiber product of XX and YY over SS, denoted X×SYX \times_S Y, is a scheme together with projection morphisms p1:X×SYXp_1: X \times_S Y \to X and p2:X×SYYp_2: X \times_S Y \to Y satisfying:

  1. Commutativity: fp1=gp2f \circ p_1 = g \circ p_2
  2. Universal property: For any scheme TT with morphisms α:TX\alpha: T \to X and β:TY\beta: T \to Y such that fα=gβf \circ \alpha = g \circ \beta, there exists a unique morphism γ:TX×SY\gamma: T \to X \times_S Y making the diagram commute:
TβYγgX×SYp2Yp1gXfS\begin{array}{ccc} T & \xrightarrow{\beta} & Y \\ \downarrow{\gamma} & & \downarrow{g} \\ X \times_S Y & \xrightarrow{p_2} & Y \\ \downarrow{p_1} & & \downarrow{g} \\ X & \xrightarrow{f} & S \end{array}

That is, p1γ=αp_1 \circ \gamma = \alpha and p2γ=βp_2 \circ \gamma = \beta.

Remark

The fiber product is characterized by its universal property, which guarantees uniqueness up to unique isomorphism. When it exists, we say that the category of schemes has fiber products. The fundamental theorem is that fiber products always exist in the category of schemes.

Existence Theorem

TheoremExistence of Fiber Products

Let f:XSf: X \to S and g:YSg: Y \to S be morphisms of schemes. Then the fiber product X×SYX \times_S Y exists in the category of schemes.

Moreover, if S=Spec(R)S = \mathrm{Spec}(R), X=Spec(A)X = \mathrm{Spec}(A), and Y=Spec(B)Y = \mathrm{Spec}(B) are affine schemes, then:

X×SY=Spec(ARB)X \times_S Y = \mathrm{Spec}(A \otimes_R B)
Remark

The general construction proceeds by covering XX, YY, and SS with affine open sets and gluing the affine fiber products. The affine case is fundamental and relies on the tensor product of rings.

The Affine Case

The affine case provides the foundation for understanding fiber products through commutative algebra.

ExampleBasic Affine Fiber Product

Let S=Spec(Z)S = \mathrm{Spec}(\mathbb{Z}), X=Spec(Z[x])X = \mathrm{Spec}(\mathbb{Z}[x]), and Y=Spec(Z[y])Y = \mathrm{Spec}(\mathbb{Z}[y]). Then:

X×SY=Spec(Z[x]ZZ[y])=Spec(Z[x,y])X \times_S Y = \mathrm{Spec}(\mathbb{Z}[x] \otimes_\mathbb{Z} \mathbb{Z}[y]) = \mathrm{Spec}(\mathbb{Z}[x,y])

This is the affine plane over Z\mathbb{Z}, which we can think of as the product of the affine line with itself.

ExampleTensor Product Computation

Consider S=Spec(k)S = \mathrm{Spec}(k) for a field kk, X=Spec(k[x]/(x2))X = \mathrm{Spec}(k[x]/(x^2)), and Y=Spec(k[y]/(y3))Y = \mathrm{Spec}(k[y]/(y^3)). Then:

X×SY=Spec(k[x](x2)kk[y](y3))=Spec(k[x,y](x2,y3))X \times_S Y = \mathrm{Spec}\left(\frac{k[x]}{(x^2)} \otimes_k \frac{k[y]}{(y^3)}\right) = \mathrm{Spec}\left(\frac{k[x,y]}{(x^2, y^3)}\right)

This represents a "thickened point" with specific nilpotent structure in two directions.

ExampleField Extensions and Tensor Products

Let kk be a field and K=k[t]/(f(t))K = k[t]/(f(t)) a field extension where ff is irreducible of degree nn. Consider:

X×Spec(k)Y=Spec(KkK)X \times_{\mathrm{Spec}(k)} Y = \mathrm{Spec}(K \otimes_k K)

If K/kK/k is separable, then KkKK \otimes_k K has nn distinct KK-algebra homomorphisms to KK (corresponding to embeddings of KK into an algebraic closure), and:

KkKK××K(n copies)K \otimes_k K \cong K \times \cdots \times K \quad (n \text{ copies})

If K/kK/k is inseparable, the structure is more subtle, involving nilpotents.

ExampleIntersection as Fiber Product

Let X=Spec(k[x,y])X = \mathrm{Spec}(k[x,y]) be the affine plane, and consider two closed subschemes:

  • V1=V(yx2)XV_1 = V(y - x^2) \subseteq X (a parabola)
  • V2=V(y1)XV_2 = V(y - 1) \subseteq X (a horizontal line)

Both naturally map to XX, so we can form:

V1×XV2=Spec(k[x,y](yx2)k[x,y]k[x,y](y1))V_1 \times_X V_2 = \mathrm{Spec}\left(\frac{k[x,y]}{(y-x^2)} \otimes_{k[x,y]} \frac{k[x,y]}{(y-1)}\right)

The tensor product is:

k[x,y](yx2,y1)=k[x](x21)=k[x]/(x1)(x+1)\frac{k[x,y]}{(y-x^2, y-1)} = \frac{k[x]}{(x^2-1)} = k[x]/(x-1)(x+1)

Over an algebraically closed field, this gives two points: the intersections at (1,1)(1,1) and (1,1)(-1,1).

Proof

To compute the tensor product, note that:

k[x,y](yx2)k[x,y]k[x,y](y1)k[x,y](yx2)k[x,y]/(ideals)k[x,y]/(y1)\frac{k[x,y]}{(y-x^2)} \otimes_{k[x,y]} \frac{k[x,y]}{(y-1)} \cong \frac{k[x,y]}{(y-x^2)} \otimes_{k[x,y]/(\mathrm{ideals})} k[x,y]/(y-1)

In the left factor, y=x2y = x^2, and in the right factor, y=1y = 1. These conditions must both hold in the tensor product, giving x2=1x^2 = 1.

Fibers of Morphisms

One of the most important applications of fiber products is describing the fibers of a morphism.

DefinitionFiber of a Morphism

Let f:XYf: X \to Y be a morphism of schemes and let yYy \in Y be a point. The fiber of ff over yy is the fiber product:

Xy:=X×YSpec(κ(y))X_y := X \times_Y \mathrm{Spec}(\kappa(y))

where κ(y)=OY,y/my\kappa(y) = \mathcal{O}_{Y,y}/\mathfrak{m}_y is the residue field at yy.

ExampleFiber of an Affine Morphism

Consider f:Spec(C[x,y])Spec(C[t])f: \mathrm{Spec}(\mathbb{C}[x,y]) \to \mathrm{Spec}(\mathbb{C}[t]) induced by tx2+y2t \mapsto x^2 + y^2. For a point aCa \in \mathbb{C}, the fiber over aa is:

Xa=Spec(C[x,y]C[t]C[t]/(ta))=Spec(C[x,y](x2+y2a))X_a = \mathrm{Spec}\left(\mathbb{C}[x,y] \otimes_{\mathbb{C}[t]} \mathbb{C}[t]/(t-a)\right) = \mathrm{Spec}\left(\frac{\mathbb{C}[x,y]}{(x^2+y^2-a)}\right)

For a0a \neq 0, this is a circle. For a=0a = 0, it's a single point (the origin). The fiber over the generic point is Spec(C(t)[x,y]/(x2+y2t))\mathrm{Spec}(\mathbb{C}(t)[x,y]/(x^2+y^2-t)), a curve over the function field.

ExampleFiber of a Projection

Consider the projection π:Ak2Ak1\pi: \mathbb{A}^2_k \to \mathbb{A}^1_k given by (x,y)x(x,y) \mapsto x. This corresponds to the ring homomorphism k[t]k[x,y]k[t] \to k[x,y] sending txt \mapsto x.

For a closed point aka \in k, the fiber is:

π1(a)=Spec(k[x,y]/(xa))Spec(k[y])=Ak1\pi^{-1}(a) = \mathrm{Spec}(k[x,y]/(x-a)) \cong \mathrm{Spec}(k[y]) = \mathbb{A}^1_k

Each fiber is an affine line. The fiber over the generic point η\eta is:

π1(η)=Spec(k(x)[y])=Ak(x)1\pi^{-1}(\eta) = \mathrm{Spec}(k(x)[y]) = \mathbb{A}^1_{k(x)}

an affine line over the function field k(x)k(x).

ExampleFiber of a Finite Morphism

Consider f:Spec(C[t])Spec(C[s])f: \mathrm{Spec}(\mathbb{C}[t]) \to \mathrm{Spec}(\mathbb{C}[s]) induced by stns \mapsto t^n (an nn-fold cover). The fiber over aCa \in \mathbb{C} is:

Xa=Spec(C[t](tna))X_a = \mathrm{Spec}\left(\frac{\mathbb{C}[t]}{(t^n - a)}\right)

If a0a \neq 0, this splits as nn distinct points (the nn-th roots of aa). If a=0a = 0, we get:

X0=Spec(C[t](tn))X_0 = \mathrm{Spec}\left(\frac{\mathbb{C}[t]}{(t^n)}\right)

a non-reduced scheme (a "fat point" of length nn).

Base Change and Extension of Scalars

Base change is the process of "changing the base scheme" for a morphism or scheme.

DefinitionBase Change

Let f:XSf: X \to S be an SS-scheme and g:TSg: T \to S a morphism. The base change of XX from SS to TT is:

XT:=X×STX_T := X \times_S T

This comes with a natural morphism fT:XTTf_T: X_T \to T making XTX_T a TT-scheme.

ExampleExtension of Scalars for Varieties

Let XX be a variety over a field kk, say X=Spec(A)X = \mathrm{Spec}(A) where AA is a finitely generated kk-algebra. For a field extension K/kK/k, the base change is:

XK=X×Spec(k)Spec(K)=Spec(AkK)X_K = X \times_{\mathrm{Spec}(k)} \mathrm{Spec}(K) = \mathrm{Spec}(A \otimes_k K)

For example, if X=V(x2+y2+1)AR2X = V(x^2 + y^2 + 1) \subseteq \mathbb{A}^2_\mathbb{R}, then over R\mathbb{R} this has no real points. But:

XC=Spec(R[x,y](x2+y2+1)RC)=Spec(C[x,y](x2+y2+1))X_\mathbb{C} = \mathrm{Spec}\left(\frac{\mathbb{R}[x,y]}{(x^2+y^2+1)} \otimes_\mathbb{R} \mathbb{C}\right) = \mathrm{Spec}\left(\frac{\mathbb{C}[x,y]}{(x^2+y^2+1)}\right)

which does have complex points, e.g., (i,0)(i, 0) and (0,i)(0, i).

ExampleReduction Modulo p

Let X=Spec(Z[x,y]/(f(x,y)))X = \mathrm{Spec}(\mathbb{Z}[x,y]/(f(x,y))) for some polynomial fZ[x,y]f \in \mathbb{Z}[x,y]. The reduction modulo pp is:

XFp=X×Spec(Z)Spec(Fp)=Spec(Z[x,y](f)ZFp)X_{\mathbb{F}_p} = X \times_{\mathrm{Spec}(\mathbb{Z})} \mathrm{Spec}(\mathbb{F}_p) = \mathrm{Spec}\left(\frac{\mathbb{Z}[x,y]}{(f)} \otimes_\mathbb{Z} \mathbb{F}_p\right)

This equals Spec(Fp[x,y]/(fˉ))\mathrm{Spec}(\mathbb{F}_p[x,y]/(\bar{f})) where fˉ\bar{f} is ff reduced modulo pp.

For instance, if X=Spec(Z[x]/(x22))X = \mathrm{Spec}(\mathbb{Z}[x]/(x^2 - 2)), then:

  • XF3=Spec(F3[x]/(x22))X_{\mathbb{F}_3} = \mathrm{Spec}(\mathbb{F}_3[x]/(x^2-2)) has two points since x22x^2 - 2 splits in F3\mathbb{F}_3
  • XF5=Spec(F5[x]/(x22))X_{\mathbb{F}_5} = \mathrm{Spec}(\mathbb{F}_5[x]/(x^2-2)) is a field extension of degree 2
ExampleBase Change for Families

Consider a family of elliptic curves over Ak1\mathbb{A}^1_k given by:

EAk1\mathcal{E} \to \mathbb{A}^1_k

For any kk-scheme TT, we can pull back the family to get:

ET=E×Ak1TT\mathcal{E}_T = \mathcal{E} \times_{\mathbb{A}^1_k} T \to T

Specifically, if T=Spec(k)T = \mathrm{Spec}(k) corresponding to a point t0Ak1t_0 \in \mathbb{A}^1_k, we recover the fiber Et0\mathcal{E}_{t_0}, a single elliptic curve.

Products of Schemes

When S=Spec(k)S = \mathrm{Spec}(k) for a field kk, the fiber product becomes the ordinary product.

ExampleProduct of Affine Spaces

Over a field kk:

Akm×kAkn=Spec(k[x1,,xm])×Spec(k)Spec(k[y1,,yn])\mathbb{A}^m_k \times_k \mathbb{A}^n_k = \mathrm{Spec}(k[x_1,\ldots,x_m]) \times_{\mathrm{Spec}(k)} \mathrm{Spec}(k[y_1,\ldots,y_n])=Spec(k[x1,,xm]kk[y1,,yn])=Spec(k[x1,,xm,y1,,yn])=Akm+n= \mathrm{Spec}(k[x_1,\ldots,x_m] \otimes_k k[y_1,\ldots,y_n]) = \mathrm{Spec}(k[x_1,\ldots,x_m,y_1,\ldots,y_n]) = \mathbb{A}^{m+n}_k
ExampleSegre Embedding

The product Pkm×kPkn\mathbb{P}^m_k \times_k \mathbb{P}^n_k exists as a scheme, and the Segre embedding shows it can be embedded in Pk(m+1)(n+1)1=Pkmn+m+n\mathbb{P}^{(m+1)(n+1)-1}_k = \mathbb{P}^{mn+m+n}_k via:

([x0::xm],[y0::yn])[xiyj]i,j([x_0:\cdots:x_m], [y_0:\cdots:y_n]) \mapsto [x_i y_j]_{i,j}

On affine charts, if we take Spec(k[x1/x0,,xm/x0])\mathrm{Spec}(k[x_1/x_0, \ldots, x_m/x_0]) and Spec(k[y1/y0,,yn/yn])\mathrm{Spec}(k[y_1/y_0, \ldots, y_n/y_n]), their product is:

Spec(k[x1x0,,xmx0]kk[y1y0,,yny0])=Spec(k[xiyjx0y0])\mathrm{Spec}\left(k\left[\frac{x_1}{x_0}, \ldots, \frac{x_m}{x_0}\right] \otimes_k k\left[\frac{y_1}{y_0}, \ldots, \frac{y_n}{y_0}\right]\right) = \mathrm{Spec}\left(k\left[\frac{x_i y_j}{x_0 y_0}\right]\right)

The Diagonal Morphism

The diagonal morphism is a fundamental tool for studying separatedness and other properties.

DefinitionDiagonal Morphism

Let f:XSf: X \to S be an SS-scheme. The diagonal morphism is:

ΔX/S:XX×SX\Delta_{X/S}: X \to X \times_S X

defined by the universal property applied to the pair (idX,idX):XX(id_X, id_X): X \rightrightarrows X.

Remark

On points, the diagonal morphism sends xXx \in X to (x,x)X×SX(x,x) \in X \times_S X. This is well-defined because f(x)=f(x)f(x) = f(x) in SS.

ExampleDiagonal of the Affine Line

For X=Ak1=Spec(k[t])X = \mathbb{A}^1_k = \mathrm{Spec}(k[t]) over S=Spec(k)S = \mathrm{Spec}(k):

X×SX=Spec(k[t]kk[t])=Spec(k[s,t])X \times_S X = \mathrm{Spec}(k[t] \otimes_k k[t]) = \mathrm{Spec}(k[s,t])

The diagonal Δ:XX×SX\Delta: X \to X \times_S X corresponds to the ring homomorphism:

k[s,t]k[t],st,ttk[s,t] \to k[t], \quad s \mapsto t, \quad t \mapsto t

The image of the diagonal is V(st)Ak2V(s-t) \subseteq \mathbb{A}^2_k, the line s=ts = t.

ExampleDiagonal as Closed Embedding

For any separated scheme XX over SS, the diagonal Δ:XX×SX\Delta: X \to X \times_S X is a closed embedding.

For instance, if X=AknX = \mathbb{A}^n_k over kk, then:

Δ:AknAkn×kAkn=Ak2n\Delta: \mathbb{A}^n_k \to \mathbb{A}^n_k \times_k \mathbb{A}^n_k = \mathbb{A}^{2n}_k

is the closed embedding corresponding to the ideal (x1y1,,xnyn)(x_1 - y_1, \ldots, x_n - y_n).

TheoremSeparatedness via Diagonal

A morphism f:XSf: X \to S is separated if and only if the diagonal morphism ΔX/S:XX×SX\Delta_{X/S}: X \to X \times_S X is a closed embedding.

Graph of a Morphism

DefinitionGraph of a Morphism

Let f:XYf: X \to Y be a morphism of SS-schemes. The graph of ff is the morphism:

Γf:XX×SY\Gamma_f: X \to X \times_S Y

defined by Γf=(idX,f)\Gamma_f = (id_X, f) using the universal property.

ExampleGraph of a Polynomial

Consider f:Ak1Ak1f: \mathbb{A}^1_k \to \mathbb{A}^1_k given by a polynomial p(x)k[x]p(x) \in k[x]. The graph is:

Γf:Ak1Ak1×kAk1=Ak2\Gamma_f: \mathbb{A}^1_k \to \mathbb{A}^1_k \times_k \mathbb{A}^1_k = \mathbb{A}^2_k

This corresponds to the ring homomorphism:

k[x,y]k[x],xx,yp(x)k[x,y] \to k[x], \quad x \mapsto x, \quad y \mapsto p(x)

The image is the closed subscheme V(yp(x))Ak2V(y - p(x)) \subseteq \mathbb{A}^2_k.

ExampleGraph of Frobenius

Let k=Fpk = \mathbb{F}_p and consider the Frobenius morphism F:Ak1Ak1F: \mathbb{A}^1_k \to \mathbb{A}^1_k given by xxpx \mapsto x^p. The graph is:

ΓF:Ak1Ak2\Gamma_F: \mathbb{A}^1_k \to \mathbb{A}^2_k

with image V(yxp)V(y - x^p), a curve in the plane defined over Fp\mathbb{F}_p.

TheoremGraph as Closed Embedding

If f:XYf: X \to Y is a morphism of SS-schemes and XX is separated over SS, then the graph Γf:XX×SY\Gamma_f: X \to X \times_S Y is a closed embedding.

Concrete Intersection Computations

ExampleTwisted Cubic Intersecting a Line

Consider the twisted cubic CP3C \subseteq \mathbb{P}^3 parametrized by [s3:s2t:st2:t3][s^3:s^2t:st^2:t^3] and a line LL given by x0=x1=0x_0 = x_1 = 0. To find their intersection scheme-theoretically, we form:

CL=C×P3LC \cap L = C \times_{\mathbb{P}^3} L

In affine coordinates where t0t \neq 0, we have C:(x,y,z)=(s3,s2,s)C: (x,y,z) = (s^3, s^2, s) and L:x=y=0L: x = y = 0. The intersection conditions are s3=0s^3 = 0 and s2=0s^2 = 0, giving s=0s = 0 with multiplicity 3. Thus:

CL=Spec(k[s](s3))C \cap L = \mathrm{Spec}\left(\frac{k[s]}{(s^3)}\right)

a non-reduced point of length 3, reflecting that the cubic is tangent to the line to third order.

ExampleIntersection of Plane Curves with Multiplicity

Let C1=V(yx2)C_1 = V(y - x^2) and C2=V(y)C_2 = V(y) in Ak2\mathbb{A}^2_k. Their intersection is:

C1C2=Spec(k[x,y](yx2,y))=Spec(k[x](x2))C_1 \cap C_2 = \mathrm{Spec}\left(\frac{k[x,y]}{(y-x^2, y)}\right) = \mathrm{Spec}\left(\frac{k[x]}{(x^2)}\right)

This is a non-reduced point at the origin with multiplicity 2, correctly capturing the tangency of the parabola to the xx-axis.

Base Change and Preservation of Properties

Many important properties of morphisms are preserved under base change.

TheoremProperties Preserved Under Base Change

Let f:XSf: X \to S be a morphism and TST \to S a base change. The following properties of ff are preserved by base change to fT:XTTf_T: X_T \to T:

  1. Open/closed embedding
  2. Affine/finite/quasi-finite
  3. Flat
  4. Smooth/étale
  5. Proper/separated
  6. Surjective (if TST \to S is surjective)
ExampleFlat Base Change

Consider the family X=Spec(k[t][x]/(x2t))Spec(k[t])=S\mathcal{X} = \mathrm{Spec}(k[t][x]/(x^2 - t)) \to \mathrm{Spec}(k[t]) = S. This is a flat family: the fiber over t=a0t = a \neq 0 is two points, while the fiber over t=0t = 0 is a double point.

For any morphism TST \to S, the base change XTT\mathcal{X}_T \to T remains flat. For instance, if T=Spec(k[s])T = \mathrm{Spec}(k[s]) and TST \to S is given by ts2t \mapsto s^2, then:

XT=Spec(k[s][x]k[t]k[t]/(x2t))=Spec(k[s][x](x2s2))\mathcal{X}_T = \mathrm{Spec}\left(k[s][x] \otimes_{k[t]} k[t]/(x^2-t)\right) = \mathrm{Spec}\left(\frac{k[s][x]}{(x^2-s^2)}\right)

This is still flat over TT.

ExampleBase Change Theorem for Proper Morphisms

Let f:XSf: X \to S be proper and TST \to S any morphism. Then fT:XTTf_T: X_T \to T is proper.

For example, if X=PknX = \mathbb{P}^n_k over S=Spec(k)S = \mathrm{Spec}(k), then for any extension K/kK/k:

(Pkn)K=PKn(\mathbb{P}^n_k)_K = \mathbb{P}^n_K

is proper over Spec(K)\mathrm{Spec}(K).

Geometric Fibers vs Scheme-Theoretic Fibers

DefinitionGeometric Fiber

Let f:XYf: X \to Y be a morphism and yYy \in Y a point. The geometric fiber over yy is:

Xyˉ:=X×YSpec(κ(y))X_{\bar{y}} := X \times_Y \mathrm{Spec}(\overline{\kappa(y)})

where κ(y)\overline{\kappa(y)} is an algebraic closure of the residue field.

ExampleComparing Geometric and Scheme-Theoretic Fibers

Consider X=Spec(Q[x]/(x22))Spec(Q)X = \mathrm{Spec}(\mathbb{Q}[x]/(x^2-2)) \to \mathrm{Spec}(\mathbb{Q}).

The scheme-theoretic fiber over the unique point is XX itself, which is connected (it's the spectrum of a field).

The geometric fiber is:

XQ=Spec(Q[x](x22)QQ)=Spec(Q×Q)X_{\overline{\mathbb{Q}}} = \mathrm{Spec}\left(\frac{\mathbb{Q}[x]}{(x^2-2)} \otimes_\mathbb{Q} \overline{\mathbb{Q}}\right) = \mathrm{Spec}\left(\overline{\mathbb{Q}} \times \overline{\mathbb{Q}}\right)

This consists of two points (corresponding to x=±2x = \pm\sqrt{2}), which are the geometric realizations of the scheme-theoretic point.

ExampleInseparable Extension

Let k=Fp(t)k = \mathbb{F}_p(t) and K=k[x]/(xpt)K = k[x]/(x^p - t). Consider X=Spec(K)Spec(k)X = \mathrm{Spec}(K) \to \mathrm{Spec}(k).

The scheme-theoretic fiber is XX itself. However:

Xkˉ=Spec(k[x](xpt)kkˉ)X_{\bar{k}} = \mathrm{Spec}\left(\frac{k[x]}{(x^p-t)} \otimes_k \bar{k}\right)

Since tt has a pp-th root α\alpha in kˉ\bar{k}, we have xpt=(xα)px^p - t = (x - \alpha)^p in kˉ[x]\bar{k}[x], so:

Xkˉ=Spec(kˉ[x](xα)p)X_{\bar{k}} = \mathrm{Spec}\left(\frac{\bar{k}[x]}{(x-\alpha)^p}\right)

This is a non-reduced point of length pp, reflecting the inseparability of the extension.

ExampleArithmetic Fibers

Consider the scheme X=Spec(Z[i])Spec(Z)X = \mathrm{Spec}(\mathbb{Z}[i]) \to \mathrm{Spec}(\mathbb{Z}).

Over the generic point η=(0)\eta = (0), the fiber is:

Xη=Spec(Q(i))X_\eta = \mathrm{Spec}(\mathbb{Q}(i))

Over a prime pp, the fiber is:

Xp=Spec(Z[i]ZFp)=Spec(Fp[x](x2+1))X_p = \mathrm{Spec}\left(\mathbb{Z}[i] \otimes_\mathbb{Z} \mathbb{F}_p\right) = \mathrm{Spec}\left(\frac{\mathbb{F}_p[x]}{(x^2+1)}\right)

The structure depends on pp:

  • If p1(mod4)p \equiv 1 \pmod{4}, then x2+1x^2 + 1 splits: XpSpec(Fp×Fp)X_p \cong \mathrm{Spec}(\mathbb{F}_p \times \mathbb{F}_p) (two points)
  • If p3(mod4)p \equiv 3 \pmod{4}, then x2+1x^2 + 1 is irreducible: Xp=Spec(Fp2)X_p = \mathrm{Spec}(\mathbb{F}_{p^2}) (one point)
  • If p=2p = 2, then x2+1=(x+1)2x^2 + 1 = (x+1)^2: X2=Spec(F2[x]/(x+1)2)X_2 = \mathrm{Spec}(\mathbb{F}_2[x]/(x+1)^2) (a double point)
ExampleBase Change Commutes with Further Base Change

Given morphisms XSX \to S, TST \to S, and TTT' \to T, there is a canonical isomorphism:

(XT)TXT(X_T)_{T'} \cong X_{T'}

For example, if XX is a variety over Q\mathbb{Q}, T=Spec(R)T = \mathrm{Spec}(\mathbb{R}), and T=Spec(C)T' = \mathrm{Spec}(\mathbb{C}), then:

(XR)CXC(X_\mathbb{R})_\mathbb{C} \cong X_\mathbb{C}

This reflects the fact that we can extend scalars in one step or in stages.

Advanced Examples

ExampleFiber Product of Non-Affine Schemes

Consider Pk1×kPk1\mathbb{P}^1_k \times_k \mathbb{P}^1_k. This cannot be computed directly as a spectrum, but we can cover it with affine opens.

If Pk1=U0U1\mathbb{P}^1_k = U_0 \cup U_1 where U0=Spec(k[x])U_0 = \mathrm{Spec}(k[x]) and U1=Spec(k[x1])U_1 = \mathrm{Spec}(k[x^{-1}]), then:

Pk1×kPk1=i,jUi×kUj\mathbb{P}^1_k \times_k \mathbb{P}^1_k = \bigcup_{i,j} U_i \times_k U_j

For instance:

U0×kU0=Spec(k[x]kk[y])=Spec(k[x,y])=Ak2U_0 \times_k U_0 = \mathrm{Spec}(k[x] \otimes_k k[y]) = \mathrm{Spec}(k[x,y]) = \mathbb{A}^2_k

The result is a surface with four affine charts, which is the product P1×P1\mathbb{P}^1 \times \mathbb{P}^1, also known as a quadric surface in P3\mathbb{P}^3 via the Segre embedding.

ExampleFamily of Nodal Cubics

Consider the family of plane cubics:

y2=x3+tx2y^2 = x^3 + tx^2

over Spec(k[t])\mathrm{Spec}(k[t]). This gives a morphism CAk1\mathcal{C} \to \mathbb{A}^1_k.

For t0t \neq 0, the fiber is a nodal cubic (with a node at the origin). For t=0t = 0, the fiber is:

C0=V(y2x3)Ak2\mathcal{C}_0 = V(y^2 - x^3) \subseteq \mathbb{A}^2_k

a cuspidal cubic. The fiber product describes how the geometry varies in the family.

ExampleExceptional Divisor via Fiber Product

The blowup of Ak2\mathbb{A}^2_k at the origin can be described using fiber products. Consider:

Bl0(Ak2)Ak2×kPk1\mathrm{Bl}_0(\mathbb{A}^2_k) \subseteq \mathbb{A}^2_k \times_k \mathbb{P}^1_k

defined by the equation xu1yu0=0xu_1 - yu_0 = 0 where (x,y)(x,y) are coordinates on Ak2\mathbb{A}^2_k and [u0:u1][u_0:u_1] on Pk1\mathbb{P}^1_k.

The exceptional divisor is the fiber over the origin:

E=Bl0(Ak2)×Ak2Spec(k[x,y]/(x,y))Pk1E = \mathrm{Bl}_0(\mathbb{A}^2_k) \times_{\mathbb{A}^2_k} \mathrm{Spec}(k[x,y]/(x,y)) \cong \mathbb{P}^1_k

The fiber product construction is foundational in scheme theory, providing a unified framework for:

  • Understanding intersections scheme-theoretically (with multiplicities)
  • Describing fibers of morphisms (both scheme-theoretic and geometric)
  • Performing base change and extension of scalars
  • Studying families and their specializations
  • Defining separatedness via the diagonal
  • Constructing products, graphs, and many other geometric objects

The interplay between the categorical universal property and concrete computations via tensor products makes fiber products both conceptually elegant and computationally tractable.