Ring theory / 10

Ring-to-space synthesis

The complete picture is a controlled reconstruction: from a ring one extracts prime points, Zariski topology, local rings of functions, affine schemes, and coherent sheaves of finite algebraic data.

This article synthesizes the “from rings to spaces” narrative around the mature dictionary developed in the previous articles.

The central reconstruction problem

Start with a commutative ring \(A\). We want to interpret \(A\) as functions on some space. For familiar rings this is easy: \(\mathbb C[x]\) is the ring of polynomial functions on \(\mathbb C\). But arbitrary rings may have nilpotents, zero divisors, arithmetic primes, and no visible set of classical points.

The solution is:

\[A \quad\rightsquigarrow\quad \operatorname{Spec}A \quad\rightsquigarrow\quad (\operatorname{Spec}A,\mathcal O_{\operatorname{Spec}A}).\]

The first arrow constructs the point set and topology. The second adds local rings of functions.

Global functions and residue-field values

The most concrete way to see a ring element as a function is through residue fields. For every prime \(\mathfrak p\in\operatorname{Spec}A\), define

\[\kappa(\mathfrak p)=\operatorname{Frac}(A/\mathfrak p).\]

Then an element \(a\in A\) gives a pointwise value

\[a(\mathfrak p)\in\kappa(\mathfrak p), \qquad a(\mathfrak p)=a\bmod\mathfrak p.\]

Equivalently, there is a set-theoretic section of the projection from the disjoint union of residue fields to \(\operatorname{Spec}A\):

\[\operatorname{Spec}A \to \coprod_{\mathfrak p\in\operatorname{Spec}A}\kappa(\mathfrak p), \qquad \mathfrak p\mapsto a(\mathfrak p)\text{ in the }\kappa(\mathfrak p)\text{-component}.\]

On \(\operatorname{mSpec}\mathbb C[x]\), this is ordinary evaluation \((x-\alpha)\mapsto f(\alpha)\). On \(\operatorname{mSpec}\mathbb Z\), it is simultaneous reduction modulo every prime \((p)\mapsto n\bmod p\). Passing from \(\operatorname{mSpec}\) to \(\operatorname{Spec}\) adds generic values: \(\mathbb C(x)\) at \((0)\subset\mathbb C[x]\), and \(\mathbb Q\) at \((0)\subset\mathbb Z\).

This residue-field viewpoint is useful, but it is not the whole scheme. The structure sheaf upgrades pointwise values to local rings, local fractions, restrictions, and gluing. It also remembers nilpotent functions whose value in every residue field is zero.

Why prime ideals are the points

Maximal ideals recover closed points in many classical situations, but they fail the functorial test: inverse images of maximal ideals need not be maximal. Prime ideals do satisfy this test.

If \(A\to B\) and \(\mathfrak q\in\operatorname{Spec}B\), then \(\varphi^{-1}(\mathfrak q)\in\operatorname{Spec}A\). Thus every ring homomorphism gives a continuous map in the opposite direction:

\[\operatorname{Spec}B\to\operatorname{Spec}A.\]

Prime ideals also include generic points. These are not optional technicalities. They encode irreducible components and make the topology sober enough for algebraic geometry.

Why the Zariski topology is the right topology

Closed sets are defined by simultaneous vanishing:

\[V(I)=\{\mathfrak p:I\subseteq\mathfrak p\}.\]

The identities

\[V(I)\cup V(J)=V(IJ),\qquad \bigcap_\alpha V(I_\alpha)=V\left(\sum_\alpha I_\alpha\right)\]

show that these sets are closed under finite unions and arbitrary intersections. Together with \(V(0)=\operatorname{Spec}A\) and \(V(A)=\emptyset\), they form the closed sets of a topology.

The basic opens are

\[D(f)=\{\mathfrak p:f\notin\mathfrak p\}.\]

They represent the condition that a function does not vanish. Their intersections satisfy \(D(f)\cap D(g)=D(fg)\), matching multiplication of functions.

Why localization is restriction

On the open set \(D(f)\), the function \(f\) is nowhere zero. Algebraically, “nowhere zero” means \(f\) should be invertible. Therefore the ring of functions on \(D(f)\) is

\[A_f.\]

This is the conceptual core of the structure sheaf:

\[\mathcal O_{\operatorname{Spec}A}(D(f))=A_f.\]

The stalk at \(\mathfrak p\) is obtained by allowing all functions not vanishing at \(\mathfrak p\) to become denominators:

\[\mathcal O_{\operatorname{Spec}A,\mathfrak p}=A_{\mathfrak p}.\]

This local ring has a unique maximal ideal, consisting of germs whose image in the residue field \(\kappa(\mathfrak p)\) is zero.

Why quotient rings are closed subspaces

An ideal \(I\) is a system of equations. Passing from \(A\) to \(A/I\) imposes those equations. Therefore, as topological spaces,

\[\operatorname{Spec}(A/I)\cong V(I)\]

and as a scheme \(\operatorname{Spec}(A/I)\) is the closed subscheme cut out by \(I\).

This distinction matters. The closed subset \(V(I)\) depends only on \(\sqrt I\), but the closed subscheme defined by \(A/I\) retains nilpotent and multiplicity information.

Why modules become sheaves

If \(A\) is a ring of functions, then an \(A\)-module \(M\) is a family of algebraic data controlled by those functions. The corresponding sheaf is

\[\widetilde M(D(f))=M_f.\]

Its stalks are

\[(\widetilde M)_{\mathfrak p}=M_{\mathfrak p}.\]

This is the module version of the structure sheaf. It says local behavior of a module is obtained by localizing the module.

Why coherent sheaves are finite geometry

When \(A\) is Noetherian, finitely generated modules form a robust finiteness class. Their associated sheaves are coherent:

\[M\text{ finite over }A \quad\leadsto\quad \widetilde M\text{ coherent on }\operatorname{Spec}A.\]

Coherent sheaves include:

vector bundles, when the module is finite locally free;
ideal sheaves, from ideals \(I\subset A\);
structure sheaves of closed subschemes, from \(A/I\);
torsion sheaves supported on closed subsets;
relation modules among equations.

They are the correct category for finite algebraic objects on a scheme.

A complete running example

Let

\[A=k[x,y]/(xy),\qquad X=\operatorname{Spec}A.\]

The minimal primes \((x)\) and \((y)\) are the generic points of two irreducible components. The maximal ideal \((x,y)\) is their intersection point.

The basic open \(D(x)\) has coordinate ring \(A_x\). Since \(xy=0\) and \(x\) is invertible in \(A_x\), one gets \(y=0\). Thus \(D(x)\) lies on the component where \(y=0\). Similarly \(D(y)\) lies on the other component.

The quotient \(A/(x)\) defines the closed subscheme \(V(x)\), the \(y\)-axis component. The sheaf \(\widetilde{A/(x)}\) is coherent and supported on that component. The ideal sheaf \(\widetilde{(x)}\) records functions locally divisible by \(x\). As a module, however, the ideal \((x)\) is annihilated by \(y\), so the support of \(\widetilde{(x)}\) is \(V(y)\). This illustrates why an ideal sheaf, the quotient it defines, and its module-theoretic support should be kept distinct.

This single example displays the whole dictionary: primes as points, quotients as closed subspaces, localization as open restriction, modules as sheaves, and coherent sheaves as finite geometry.

Final dictionary

Ring-theoretic operation	Geometric meaning
\(A\)	Ring of global functions on an affine scheme
\(\mathfrak p\subset A\) prime	Point of \(\operatorname{Spec}A\)
\(\mathfrak m\subset A\) maximal	Closed point
\(V(I)\)	Closed locus where \(I\) vanishes
\(D(f)\)	Open locus where \(f\) is invertible
\(A_f\)	Functions on \(D(f)\)
\(A_{\mathfrak p}\)	Local ring at \(\mathfrak p\)
\(A/I\)	Closed subscheme cut out by \(I\)
\(M_f\)	Sections of \(\widetilde M\) on \(D(f)\)
\(M_{\mathfrak p}\)	Stalk of \(\widetilde M\) at \(\mathfrak p\)
finite \(A\)-module over Noetherian \(A\)	coherent sheaf on \(\operatorname{Spec}A\)

The five recurring pitfalls

Prime ideals are the points, not just maximal ideals. Sheaves include restriction and gluing data, not only values on opens. Quasi-coherent does not imply coherent. Localization is open restriction, not quotienting. Global algebra and local algebra are related by sheaves and stalks, but neither can be ignored.