Scheme-theoretic fibre ($X_y=X{\times }_Y\mathrm{Spec}\kappa (y)$)

The fibre of a morphism of schemes over $y\in Y$ is its base change to the residue field $\kappa (y)$. For $\mathrm{Spec}B\to \mathrm{Spec}A$, with $y$ corresponding to $\mathfrak{p}$, its fibre ring is $B{\otimes }_A\kappa (\mathfrak{p})$. Its structure sheaf records nilpotents and multiplicities that the set-theoretic inverse image cannot see.

For a morphism of schemes $X\to S$ with $S$ an integral scheme and generic point $\eta$, the generic fiber is its scheme-theoretic fiber over $\eta$. On affine schemes $S=\mathrm{Spec}A$ and $X=\mathrm{Spec}B$, it is $\mathrm{Spec}(B{\otimes }_A\mathrm{Frac}A)$. It can be empty, and it need not be a reduced scheme or an irreducible scheme without further assumptions.

Blow up $(0,1,0)$ in ${\mathbb{A}}_{x,y,z}^3$ and compose with $xy:{\mathbb{A}}^3\to {\mathbb{A}}^1$. Nonzero fibres are ${\mathbb{G}}_m\times {\mathbb{A}}^1$. The zero fibre consists of the strict transforms of the two coordinate planes and the exceptional ${\mathbb{P}}^2$, all with multiplicity one. Its plane crossings make it singular; its closed ${\mathbb{P}}^2$ makes it nonaffine. The centre is deliberately on just one component of $xy=0$, keeping the fibre reduced.

The scheme-theoretic fibres over positive real points, negative real points, zero, and nonreal closed points have real coordinate algebras ${\mathbb{R}}^2\times \mathbb{C}$, ${\mathbb{C}}^2$, $\mathbb{R}[x]/(x^4)$, and ${\mathbb{C}}^4$, respectively. Their ordinary irreducible component counts are $3,2,1,4$, so there are four real-scheme isomorphism classes. The last case pulls back an irreducible quadratic to a squarefree degree-eight polynomial, with four conjugate pairs of nonreal roots. The zero fibre retains its nilpotents and is not a reduced point.

A morphism of schemes between integral schemes can have a scheme-theoretic fibre that is neither reduced nor irreducible. For the map of affine lines given by $t=x^2(x-1{)}^2$, the fibre at $t=0$ has ring $k[x]/(x^2(x-1{)}^2)$. The Chinese remainder theorem identifies it with the product of two dual number rings, so it consists of two nonreduced double points. Its nilpotent structure is invisible to the set-theoretic fibre.

For an injective ring homomorphism between nonzero integral domains $A\to B$, the scheme-theoretic fibre over the generic point of $\mathrm{Spec}A$ is $\mathrm{Spec}(S^{-1}B)$ with $S=A\setminus \{0\}$. Localization at these nonzero elements is a nonzero integral domain, so the generic fibre is a nonempty integral scheme. A surjective morphism between these affine schemes necessarily gives an injective ring map: a prime above $(0)$ contains its kernel.