Bourbaki

4. Spectra of rings and supports of modules

1. Irreducible spaces

Definition 1

A topological space XX is called irreducible if every finite intersection of non-empty open sets of XX is non-empty.

By considering the empty family of open sets of XX it is seen that an irreducible space is non-empty; for a topological space XX to be irreducible, it is necessary and sufficient that it be non-empty and that the intersection of two non-empty open sets of XX be always non-empty (or, what amounts to the same thing, that the union of two closed sets distinct from XX be always distinct from XX).

Proposition 1

Let XX be a non-empty topological space. The following conditions are equivalent: (a) XX is irreducible; (b) every non-empty open set of XX is dense in XX; (c) every open set of XX is connected.

By definition, a set which is dense in XX is a set which meets every non-empty open set, hence (a) and (b) are equivalent. It is immediate that (c) implies (a), for if U1U_1 and U2U_2 are disjoint non-empty open sets, U1∪U2U_1 \cup U_2 is a disconnected open set. Finally let us show that (a) implies (c): if UU is a disconnected open set, it is the union of two disjoint non-empty sets U′,U′′U', U'' which are open in UU and hence also open in XX, which implies that XX is not irreducible.

A Hausdorff space is irreducible only if it consists of a single point.

A subset EE of a topological space XX is called an irreducible set if the subspace EE of XX is irreducible. For this to be so, it is necessary and sufficient that, for every pair of sets U,VU, V which are open in XX and meet EE, U∩VU \cap V also meet EE, or (what amounts to the same thing) that, for every pair of sets F,GF, G which are closed in XX and satisfy E⊂F∪GE \subset F \cup G, E⊂FE \subset F or E⊂GE \subset G. By induction on nn, we deduce that, if (Fi)1≤i≤n(F_i)_{1 \leq i \leq n} is a finite family of closed sets of XX such that E⊂⋃i=1nFiE \subset \bigcup_{i=1}^n F_i, there exists an index ii such that E⊂FiE \subset F_i.

Proposition 2

In a topological space XX, for a set EE to be irreducible, it is necessary and sufficient that its closure E‾\overline{E} be so.

For an open set of XX to meet EE, it is necessary and sufficient that it meet E‾\overline{E}; then the proposition follows from the above remarks.

Proposition 3

(i) *If XX is an irreducible space, every non-empty open subset of XX is irreducible.

(ii) Let (Uα)α∈A(U_\alpha)_{\alpha \in A} be a non-empty covering of a topological space XX consisting of open sets such that Uα∩Uβ=∅U_\alpha \cap U_\beta = \emptyset for every pair of indices (α,β)(\alpha, \beta). If the sets UαU_\alpha are irreducible, the space XX is irreducible.

(i) If XX is irreducible, U⊂XU \subset X non-empty and open in XX and V⊂UV \subset U non-empty and open in UU, VV is also open in XX, therefore dense in XX and a fortiori dense in UU. Then UU is irreducible (Proposition 1).

(ii) Let us show that, for every non-empty open set VV in XX, V∩Uα≠∅V \cap U_\alpha \neq \emptyset for all α∈A\alpha \in A: it follows that V∩UαV \cap U_\alpha is dense in UαU_\alpha, by hypothesis, hence that VV is dense in XX and this proves that XX is irreducible (Proposition 1). Now there exists at least one index yy such that V∩Uy≠∅V \cap U_y \neq \emptyset; as Uα∩Uy≠∅U_\alpha \cap U_y \neq \emptyset for all α\alpha and V∩UyV \cap U_y is dense in UyU_y, Uy∩V≠∅U_y \cap V \neq \emptyset and a fortiori Uy∩V≠∅U_y \cap V \neq \emptyset, which completes the proof of (ii).

Proposition 4

Let XX and YY be two topological spaces and ff a continuous mapping from XX to YY. For every irreducible subset EE of XX, f(E)f(E) is an irreducible subset of YY.

If U,VU, V are two open sets of YY which meet f(E)f(E), f−1(U)f^{-1}(U) and f−1(V)f^{-1}(V) are open sets of XX which meet EE. Consequently, f−1(U)∩f−1(V)=f−1(U∩V)f^{-1}(U) \cap f^{-1}(V) = f^{-1}(U \cap V) meets EE, which implies that U∩VU \cap V meets f(E)f(E) and proves the proposition.

Definition 2

Every maximal irreducible subset of a topological space XX is called an irreducible component of XX.

It follows from Proposition 2 that every irreducible component of XX is closed in XX.

Proposition 5

Let XX be a topological space. Every irreducible subset of XX is contained in an irreducible component of XX and XX is the union of its irreducible components.

To prove the first assertion, it is sufficient, by virtue of Zorn's Lemma, to prove that the set 3 of irreducible subsets of XX is inductive. Let G\mathscr{G} be a subset of S\mathfrak{S} totally ordered by inclusion; we show that the union EE of the sets F∈GF \in \mathscr{G} is irreducible. Let U,VU, V be two open sets of XX which meet EE; as G\mathscr{G} is totally ordered, there exists a set F∈GF \in \mathscr{G} meeting UU and VV; as FF is irreducible, U∩VU \cap V meets FF and hence also EE, which proves that EE is irreducible and hence that 3 is inductive. The second assertion follows from the first, for every subset of XX consisting of a single point is irreducible.

Corollary

Every connected component of a topological space XX is a union of irreducible components of XX.

Every irreducible subspace of XX is connected by Proposition 1 and hence is contained in a connected component of XX.

Note that two distinct irreducible components of XX may have points in common (Exercise 11).

Proposition 6

Let XX be a topological space and (Pi)1≤i≤n(P_i)_{1 \leq i \leq n} ajinite covering of XX consisting & irreducible closed sets. Then the irreducible components of XX are the maximal elements (with respect to inclusion) of the set of PiP_i.

We may restrict ourselves to the case where no two PiP_i are comparable. If EE is an irreducible subset of XX, then E⊂⋃i=1nPiE \subset \bigcup_{i=1}^n P_i, and hence EE is contained in one of the closed sets PiP_i; this proves that the PiP_i are the only maximal irreducible subsets of XX.

Corollary

Let XX be a topological space and EE a subspace with only ajinite number of distinct irreducible components QiQ_i (1≤i≤n1 \leq i \leq n); then the irreducible components of the closure E‾\overline{E} in XX are the closures Q‾i\overline{Q}_i of the QiQ_i (1≤i≤n1 \leq i \leq n) and Q‾i#Q‾j\overline{Q}_i \# \overline{Q}_j for i≠ji \neq j.

E‾\overline{E} is the union of the Q‾i\overline{Q}_i, which are irreducible (Proposition 2); as QiQ_i is closed in EE, Q‾i∩E=Qi\overline{Q}_i \cap E = Q_i; as Qi∉QjQ_i \notin Q_j for i≠ji \neq j, Q‾i∉Q‾j\overline{Q}_i \notin \overline{Q}_j, whence the corollary by virtue of Proposition 6.

Remark

Suppose that XX has only ajinite number of distinct irreducible components XiX_i (1≤i≤n1 \leq i \leq n); then U+=C(⋃j≠iXj)U_+ = \mathbf{C}(\bigcup_{j \neq i} X_j) is open in XX and dense in XiX_i since Xi∉⋃j≠iXjX_i \notin \bigcup_{j \neq i} X_j; the UiU_i (1≤i≤n1 \leq i \leq n) are therefore non-empty open sets of XX which are irreducible (Proposition 2), pairwise disjoint and with their union dense in XX.

Proposition 7

Let UU be an open subset of a topological space XX. The mapping V↦V‾V \mapsto \overline{V} (closure in XX) is a bijection of the set & irreducible subsets of UU which are closed in UU onto the set & irreducible subsets & XX which are closed in XX and meet UU; the inverse bijection is Z↦Z∩UZ \mapsto Z \cap U. In particular, this bijection maps the set & irreducible components & UU onto the set & irreducible components & XX which meet UU.

If VV is closed in UU and irreducible, V‾\overline{V} is irreducible (Proposition 2) and V=V‾∩UV = \overline{V} \cap U. Conversely, if ZZ is irreducible and closed in XX and meets UU, Z∩UZ \cap U is a non-empty open subset of ZZ, hence irreducible (Proposition 3) and dense in ZZ and, as ZZ is closed, Z=Z‾∩UZ = \overline{Z} \cap U. This proves the proposition.

2. Noetherian topological spaces

Definition 3

A topological space XX is called Noetherian if every non-empty set of closed subsets of XX, ordered by inclusion, has a minimal element.

It amounts to the same to say that every non-empty set of open subsets of XX, ordered by inclusion, has a maximal element, or that every decreasing (resp. increasing) sequence of closed (resp. open) sets is stationary (Set Theory, Chapter 111, § 6, no. 5, Proposition 6).

Proposition 8

(i) Every subspace of a Noetherian space is Noetherian. (ii) Let (Ai)i∈I(A_i)_{i \in I} be a finite covering of a topological space XX. If the subspaces AiA_i of XX are Noetherian, XX is Noetherian.

(i) Let XX be a Noetherian space, AA a subspace of XX and (Fi)(F_i) a decreasing sequence of subsets of AA which are closed in AA; then Fi=Fn‾∩AF_i = \overline{F_n} \cap A and the closures Fn‾\overline{F_n} of the FiF_i in XX form a decreasing sequence of closed subsets of XX. As this sequence is stationary, so is the sequence (Fi)(F_i).

(ii) Let (Gn)n≥0(G_n)_{n \geq 0} be a decreasing sequence of closed subsets of XX; by hypothesis, each of the sequences (Gn∩Ai)n≥0(G_n \cap A_i)_{n \geq 0} is stationary. As II is finite, there is an integer n0n_0 such that, for n≥n0n \geq n_0, Gn∩Ai=Gn0∩AiG_n \cap A_i = G_{n_0} \cap A_i, for all i∈Ii \in I. But

Gn=⋃i∈I(Gn∩Ai) G_n = \bigcup_{i \in I} (G_n \cap A_i)
and hence the sequence (Gn)(G_n) is stationary and XX is Noetherian.

Proposition 9

For a topological space XX to be Noetherian, it is necessary and sufficient that every open set in XX be quasi-compact.

To show that the condition is necessary, it is sufficient, by virtue of Proposition 8, to prove that every Noetherian space XX is quasi-compact. Let (Ui)i∈I(U_i)_{i \in I} be an open covering of XX; the set of finite unions of sets UiU_i is non-empty and hence admits a maximal element V=⋃i∈HUiV = \bigcup_{i \in H} U_i, where HH is a finite subset of II. By definition, V∪Ui=VV \cup U_i = V for all i∈Ii \in I and hence V=XV = X.

Conversely, suppose that every open set of XX is quasi-compact and let (Ui)(U_i) be an increasing sequence of open subsets of XX. The union VV of the UiU_i is open and hence quasi-compact; as (Ui)(U_i) is an open covering of VV, there is a finite subfamily of (Ui)(U_i) which is a covering of VV and hence V=UnV = U_n for some index nn, which proves that the sequence (Ui)(U_i) is stationary.

Lemma 1 ("Principle of Noetherian induction"). Let E be an ordered set every non-empty subset of which admits a minimal element. Let F be a subset of E with the following property : if a∈Ea \in E is such that the relation x<ax < a implies x∈Fx \in F, then a∈Fa \in F. Then F=EF = E.

Suppose F≠EF \neq E; then CFCF would have a minimal element bb. By definition, x∈Fx \in F for all x<bx < b, which implies that b∈Fb \in F, which is a contradiction.

Proposition 10

If X is a Noetherian space, the set of irreducible components of X (and a fortiori the set of connected components of X) is finite.

It is sufficient to prove that X is a finite union of irreducible closed subsets (no. 1, Proposition 6). Let us show that the principle of Noetherian induction can be applied taking E to be the set of closed subsets of X, ordered by inclusion, and F to be the set of finite unions of irreducible closed subsets. Let Y be a closed subset of X such that every closed subset ≠Y\neq Y of Y belongs to F. If Y is irreducible, then Y∈FY \in F by definition; otherwise, Y is the union of two closed subsets Y1, Y2Y_1,\ Y_2 which are distinct from Y. Then Y1∈FY_1 \in F and Y2∈FY_2 \in F by hypothesis, whence Y∈FY \in F by definition of F.

In particular it follows that a Hausdorff Noetherian space is necessarily finite.

3. The prime spectrum of a ring

Let A be a ring and X the set of prime ideals of A. For every subset M of A, we denote by V(M)V(M) the set of prime ideals of A containing M; clearly, if aa is the ideal of A generated by M, V(M)=V(a)V(M) = V(a); if M consists of a single element f, we write V(f)V(f) instead of V({f})V(\{f\}) and we have V(f)=V(Af)V(f) = V(Af). The mapping M↦V(M)M \mapsto V(M) is decreasing with respect to inclusion in A and X. Moreover, the following formulae hold :

(1)

V(0)=X,V(1)=O; V(0) = X, \quad V(1) = O;

(2)

V(⋃i∈IMi)=V(∑i∈IMi)=⋂i∈IV(Mi) V\left(\bigcup_{i \in I} M_i\right) = V\left(\sum_{i \in I} M_i\right) = \bigcap_{i \in I} V(M_i)
for every family (Mi)i∈I(M_i)_{i \in I} of subsets of A;

(3)

V(a∩a′)=V(aa′)=V(a)∪V(a′) V(a \cap a') = V(aa') = V(a) \cup V(a')
for every pair of ideals a, a′a,\ a' in A. Formulae (1) and (2) are obvious; on the other hand, formula (3) means that, for a prime ideal pp of A to contain one of the ideals aa or a′a', it is necessary and sufficient that it contain aa′aa' or that it contain a∩a′a \cap a'; then it is a consequence of § 1, no. 1, Proposition 1. The second formula (1) has the following converse: if aa is an ideal of AA such that V(a)=∅V(a) = \varnothing, then a=Aa = A, for there is no maximal ideal of AA containing aa. Finally, if aa is an ideal of AA and r(a)r(a) is its radical (§2§ 2, no. 6, Definition 4), then

V(a)=V(r(a)) V(a) = V(r(a))

as follows from §2§ 2, no. 6, Corollary 1 to Proposition 13.

Formulae (1) to (3) show that the subsets V(M)V(M) of XX satisfy the closed set axioms of a topology (General Topology, Chapter I, §1§ 1, no. 4).

Definition 4

Let AA be a ring. The set XX of prime ideals of AA, with the topology whose closed sets are the sets V(M)V(M), where MM runs through P(A)\mathfrak{P}(A), is called the prime spectrum of AA and denoted by Spec(A)\mathrm{Spec}(A). The topology thus defined is called the spectral or Zariski topology on XX.

Clearly the relation Spec(A)=∅\mathrm{Spec}(A) = \varnothing is equivalent to A={0}A = \{0\}.

Let XX be the prime spectrum of a ring AA; for all f∈Af \in A, let us denote by XfX_f the set of prime ideals of AA not containing ff; then Xf=X−V(f)X_f = X - V(f) and XfX_f is therefore an open set. By (2), every closed subset of XX is an intersection of closed sets of the form V(f)V(f) and hence the XX, form a base of the spectral topology on XX. Moreover, it follows immediately from the definitions that

X0=O,X1=X, X_0 = O,\quad X_1 = X,

and more generally Xf−1=XX_{f^{-1}} = X for every invertible element ff of AA;

Xg−1=X,nX,for all f,g in A. X_{g^{-1}} = X, nX, \quad \text{for all } f, g \text{ in } A.

For every subset YY of XX, let us denote by J(Y)\mathfrak{J}(Y) the intersection of the prime ideals of AA which belong to YY. Clearly J(Y)\mathfrak{J}(Y) is an ideal of AA and the mapping Y↦J(Y)Y \mapsto \mathfrak{J}(Y) is decreasing with respect to inclusion in XX and AA. Clearly the relations

J(∅)=A \mathfrak{J}(\varnothing) = A
J(⋃λ∈LYλ)=⋂λ∈LJ(Yλ) \mathfrak{J}\left( \bigcup_{\lambda \in L} Y_\lambda \right) = \bigcap_{\lambda \in L} \mathfrak{J}(Y_\lambda)

hold for every family (Yλ)λ∈L(Y_\lambda)_{\lambda \in L} of subsets of XX. Moreover:

Proposition 11

Let AA be a ring, aa an ideal of AA and YY a subset of X=Spec(A)X = \mathrm{Spec}(A). (i) V(a)V(a) is closed in XX and J(Y)\mathfrak{J}(Y) is an ideal of AA which is equal to its radical. (ii) J(V(a))\mathfrak{J}(V(a)) is the radical of aa and V(J(Y))V(\mathfrak{J}(Y)) is the closure of YY in XX.

(iii) The mappings 3 and V define decreasing bijections, one of which is the inverse of the other, between the set of closed subsets of X and the set of ideals of A which are equal to their radicals.

Assertion (i) and the first assertion of (ii) follow from the definitions and § 2, no. 6, Corollary 1 to Proposition 13. If a closed set V(M) (for some M ⊂ A) contains Y, then M ⊂ p for every prime ideal p ∈ Y, whence M ⊂ J(Y) and consequently V(M) ⊃ V(J(Y)); as Y ⊂ V(J(Y)), V(J(Y)) is the smallest closed set of X containing Y, which completes the proof of (ii). Finally, it follows from (ii) that, if a is a prime ideal equal to its radical, then J(V(a)) = a and that, if Y is closed in X, then V(J(Y)) = Y; this proves (iii).

It follows immediately from Proposition 11 that, if M is any subset of A and Y is any subset of X, then V(M) = V(J(V(M))) and J(Y) = J(V(J(Y))).

Corollary 1

For every family (Y_λ)_{λ ∈ L} of closed subsets of X, J(∩_{λ ∈ L} Y_λ) is the radical of the sum of the ideals J(Y_λ).

It follows from Proposition 11 (iii) that J(∩_{λ ∈ L} Y_λ) is the smallest ideal which is equal to its radical and contains all the J(Y_λ); this ideal then contains ∑_{λ ∈ L} J(Y_λ) and therefore also the radical of ∑_{λ ∈ L} J(Y_λ) (§ 2, no. 6, Corollary 2 to Proposition 13), whence the corollary.

Corollary 2

Let r(a) denote the radical of an ideal a of A; if a and b are two ideals of A, the relation V(a) ⊂ V(b) is equivalent to b ⊂ r(a) and r(b) ⊂ r(a).

It is immediate that the relations b ⊂ r(a) and r(b) ⊂ r(a) are equivalent and, as V(a) = V(r(a)), the corollary follows immediately from Proposition 11, (iii).

Corollary 3

Let (f_λ)_{λ ∈ L} be a family of elements of A. For an element g ∈ A to satisfy X_g ⊂ ∪_{λ ∈ L} X_{f_λ}, it is necessary and sufficient that there exist an integer n > 0 such that g^n belongs to the ideal generated by the f_λ.

The relation X_g ⊂ ∪_{λ ∈ L} X_{f_λ} is equivalent to V(g) ⊃ ∩_{λ ∈ L} V(f_λ) and it is sufficient to apply Corollary 2.

COROLLARY 4, For two elements f, g of A to satisfy X_f = X_g, it is necessary and sufficient that there exist two integers m > 0, n > 0 such that f^m ∈ Ag and g^n ∈ Af.

Corollary 5

For f∈Af \in A to satisfy Xf=∅X_f = \varnothing, it is necessary and sufficient that ff be nilpotent.

This follows immediately from Corollary 4.

Corollary 6

The closure of a set consisting of a point p∈X=Spec(A)p \in X = \mathrm{Spec}(A) is the set V(p)V(p) of prime ideals containing pp. For the set {p}\{p\} to be closed in XX (or, as we shall also say by an abuse of language, for pp to be a closedpoint of XX), it is necessary and sufficient that pp be maximal.

Corollary 7

If AA is a Noetherian ring, X=Spec(A)X = \mathrm{Spec}(A) is a Noetherian space.

Proposition 12

For all f∈Af \in A, the open set XfX_f in X=Spec(A)X = \mathrm{Spec}(A) is quasi-compact; in particular, the space XX is quasi-compact.

As the XgX_g form a base of the topology, it is sufficient to prove that, if (gλ)λ∈L(g_\lambda)_{\lambda \in L} is a family of elements of AA such that Xf⊂⋃λ∈LXgλX_f \subset \bigcup_{\lambda \in L} X_{g_\lambda}, then there exists a finite subfamily (gλi)i∈I(g_{\lambda_i})_{i \in I} such that Xf=⋃i∈IXgλiX_f = \bigcup_{i \in I} X_{g_{\lambda_i}}. But the relation Xf=⋃λ∈LXgλX_f = \bigcup_{\lambda \in L} X_{g_\lambda} means that there exists an integer n>0n > 0 and a finite subfamily (gλi)i∈I(g_{\lambda_i})_{i \in I} such that "f" belongs to the ideal generated by that subfamily (Corollary 3 to Proposition 11); whence the proposition.

Proposition 13

Let A,A′A, A' be two rings, X=Spec(A),X′=Spec(A′)X = \mathrm{Spec}(A), X' = \mathrm{Spec}(A') and hh a homomorphism from AA to A′A'; the mapping ah:p′↦h‾(p′)^a h : p' \mapsto \overline{h}(p') from X′X' to XX is continuous.

For M⊂AM \subset A, the set (ah)−1(V(M))(^a h)^{-1}(V(M)) is the set of prime ideals p′p' of A′A' such that M⊂h‾(p′)M \subset \overline{h}(p'), which is equivalent to h(M)⊂p′h(M) \subset p'; this set is then equal to V(h(M))V(h(M)) and is therefore closed.

We call ah^a h the mapping associated with the homomorphism hh.

Remark

If hh is surjective and aa is its kernel, it follows from the definition of the spectral topology that ah^a h is a homeomorphism of X′X' onto the closed subspace V(a)V(a) of XX; for a prime ideal p′p' of A′A' to contain an ideal b′b' of A′A', it is necessary and sufficient that h‾(p′)\overline{h}(p') contain h‾(b′)\overline{h}(b'); we see first that ah^a h is injective by taking b′b' to be prime; moreover, for every ideal b′b' of A′A', the image under ah^a h of V(b′)V(b') is V(h‾(b′))V(\overline{h}(b')), whence our assertion, the ideals of the form h‾(b′)\overline{h}(b') all being ideals of AA which contain aa.

Corollary

Let SS be a multiplicative subset of AA, A′=S−1AA' = S^{-1}A and hh the canonical homomorphism iASi_A^S; then ah^a h is a homeomorphism of X′=Spec(A′)X' = \mathrm{Spec}(A') onto the subspace of X=Spec(A)X = \mathrm{Spec}(A) consisting of the prime ideals of AA which do not meet SS.

Let f′=f/sf' = f/s, where f∈A,s∈Sf \in A, s \in S; then Xf′′=Xf/1′X'_{f'} = X'_{f/1} since s/1s/1 is invertible in A′A'. We know already that ah^a h is injective and that, for all p′∈X′p' \in X', the relations f/1∈p′f/1 \in p' and f∈h−1(p′)=ah(p′)f \in h^{-1}(p') = ^a h(p') are equivalent and hence the conditions p′∈Xf/1′p' \in X'_{f/1} and ah(p′)∈Xf^a h(p') \in X_f are equivalent; this shows that ah(Xf′′)^a h(X'_{f'}) is equal to Xf∩ah(X′)X_f \cap ^a h(X'), whence the first assertion, since the X,(resp. Xf′′)X, (resp.\ X'_{f'}) form a base of the topology of XX (resp. XX). The second assertion follows from , no. 5, Proposition 11 (ii)

Proposition 14

Let AA be a ring. For a subset YY of X=Spec(A)X = \mathrm{Spec}(A) to be irreducible, it is necessary and sufficient that the ideal J(Y)\mathfrak{J}(Y) be prime.

Writing p=J(Y)p = \mathfrak{J}(Y), we note that, for an element f∈Af \in A, the relation f∈pf \in p is equivalent to Y⊂V(f)Y \subset V(f). Suppose that YY is irreducible and let f,gf, g be elements of AA such that fg∈pfg \in p. Then

Y⊂V(fg)=V(f)∪V(g); Y \subset V(fg) = V(f) \cup V(g);
as YY is irreducible and V(f)V(f) and V(g)V(g) are closed, Y⊂V(f)Y \subset V(f) or Y⊂V(g)Y \subset V(g), hence f∈pf \in p or g∈pg \in p, which proves that pp is prime.

Suppose now that pp is prime; then Y‾=V(p)\overline{Y} = V(p) (Proposition 11 (ii)) and, as pp is prime, p=J({p})p = \mathfrak{J}(\{p\}), whence Y‾=V(J({p}))={p}\overline{Y} = V(\mathfrak{J}(\{p\})) = \{p\} (Proposition 11 (ii)). As a set consisting of a single point is irreducible, YY is irreducible (no. 1, Proposition 2).

Corollary 1

For a ring AA to be such that X=Spec(A)X = \mathrm{Spec}(A) is irreducible, it is necessary and sufficient that the quotient of AA by its nilradical N\mathfrak{N} be an integral domain.

By Proposition 11 (i), J(X)\mathfrak{J}(X) is the radical of the ideal (0)(0), that is N\mathfrak{N}.

Corollary 2

The mapping p↦V(p)p \mapsto V(p) is a bijection of X=Spec(A)X = \mathrm{Spec}(A) onto the set of irreducible closed subsets of XX; in particular the irreducible components of a closed subset YY of XX are the sets V(p)V(p), where pp runs through the set of minimal elements of the set of prime ideals of AA which contain J(Y)\mathfrak{J}(Y).

As J(V(p))=p\mathfrak{J}(V(p)) = p for every prime ideal pp of AA and Y=V(J(Y))Y = V(\mathfrak{J}(Y)) for every closed subset YY of XX, the first assertion follows from Proposition 14; on the other hand, for Y⊃V(p)Y \supset V(p), it is necessary and sufficient that

p=J(V(p))⊃J(Y) p = \mathfrak{J}(V(p)) \supset \mathfrak{J}(Y)
(Proposition 11), whence the second assertion.

Corollary 3

The set of minimal prime ideals of a Noetherian ring AA is finite.

X=Spec(A)X = \mathrm{Spec}(A) has only a finite number of irreducible components (Corollary 7 to Proposition 11 and no. 2, Proposition 10) and the corollary follows from Corollary 2 above.

Proposition 15

Let AA be a ring, II a finite set and EE the set of orthogonal families (ei)i∈I(e_i)_{i \in I} of idempotents ei≠0e_i \neq 0 of AA such that ∑i∈Iei=1\sum_{i \in I} e_i = 1. For all (ei)i∈I∈E(e_i)_{i \in I} \in E, we set ω‾((ei)i∈I)=(V(A(1−ei)))i∈I\overline{\omega}((e_i)_{i \in I}) = (\mathrm{V}(A(1 - e_i)))_{i \in I}, σ((ei)i∈I)=(Aei)i∈I\sigma((e_i)_{i \in I}) = (Ae_i)_{i \in I}. Then ω‾\overline{\omega} is a bijection of EE onto the set PP of partitions (Ui)i∈I(U_i)_{i \in I} of X=Spec(A)X = \mathrm{Spec}(A) into open sets and σ\sigma is a bijection of EE onto the set SS of families (ai)i∈I(a_i)_{i \in I} of ideals ≠0\neq 0 of AA such that AA is the direct sum of the aia_i.

Let (ei)i∈I(e_i)_{i \in I} be an element of EE and set Yi=V(A(1−ei))Y_i = \mathrm{V}(A(1 - e_i)); if i≠ji \neq j, then 1=1−ei+ei(1−ej)∈A(1−ei)+A(1−ej)1 = 1 - e_i + e_i(1 - e_j) \in A(1 - e_i) + A(1 - e_j), whence Yi∩Yj=∅Y_i \cap Y_j = \varnothing (formulae (1) and (2)). On the other hand,

⋃i∈IYi=V(∏i∈IA(1−ei)) \bigcup_{i \in I} Y_i = \mathrm{V}\left( \prod_{i \in I} A(1 - e_i) \right)

(formula (3));

by hypothesis ∏i∈I(1−ei)=1−∑i∈Iei=0\prod_{i \in I} (1 - e_i) = 1 - \sum_{i \in I} e_i = 0, whence ⋃i∈IYi=X\bigcup_{i \in I} Y_i = X (formula (I)). As the YiY_i are closed, they are also open, whence ω‾(E)⊂P\overline{\omega}(E) \subset P. Also, obviously A=∑i∈IAeiA = \sum_{i \in I} Ae_i; if 0=∑i∈Iaiei0 = \sum_{i \in I} a_i e_i where ai∈Aa_i \in A, we obtain, by multiplying by eie_i, O=aiei2=aieiO = a_i e_i^2 = a_i e_i for all ii; this proves that σ(E)⊂S\sigma(E) \subset S.

Lemma 2

If* e,fe, f are two idempotents of AA such that AeAe and AfAf have the same radical then e=fe = f.

There exists by hypothesis integers m≥0,n≥0m \geq 0, n \geq 0 such that e=em∈Afe = e^m \in Af and f=fn∈Aef = f^n \in Ae; let x,yx, y be elements of AA such that e=xf,f=yee = xf, f = ye; then ef=x2=x=eef = x^2 = x = e and similarly ef=ye2=ye=fef = ye^2 = ye = f, whence e=fe = f.

Lemma 2 and Corollary 2 to Proposition 11 show that the mappings ω‾\overline{\omega} and σ\sigma are injective.

Let us show that σ\sigma is surjective. If (ai)i∈I(a_i)_{i \in I} is an element of SS, there are elements ei∈aie_i \in a_i such that 1=∑i∈Iei1 = \sum_{i \in I} e_i; if i≠ji \neq j, then eiej∈ai∩aj={0}e_i e_j \in a_i \cap a_j = \{0\}, whence

ei=∑j∈Ieiej=ei2; e_i = \sum_{j \in I} e_i e_j = e_i^2;

finally, Aei⊂aiAe_i \subset a_i, for all i∈Ii \in I and ∑i∈IAei=A\sum_{i \in I} Ae_i = A, whence Aei=aiAe_i = a_i.

It remains to prove that ω‾\overline{\omega} is surjective. Let (Ui)i∈I(U_i)_{i \in I} be an element of PP and set Zi=⋃j≠iUjZ_i = \bigcup_{j \neq i} U_j; as UiU_i and ZiZ_i are closed, there exist ideals ai,bia_i, b_i of AA such that Ui=V(ai),Zi=V(bi)U_i = \mathrm{V}(a_i), Z_i = \mathrm{V}(b_i). We now show that it is possible to suppose further that ai∩bi=0a_i \cap b_i = 0. Now Ui∩Zi=∅U_i \cap Z_i = \varnothing, whence ai+bi=Aa_i + b_i = A; let ai,bi∈ai,bia_i, b_i \in a_i, b_i be such that ai+bi=1a_i + b_i = 1. Then X=Ui∪Zi=V(aibi)X = U_i \cup Z_i = \mathrm{V}(a_i b_i) (formula (3)); every element of aibia_i b_i is therefore nilpotent (Corollary 2 to Proposition 11); let pp be an integer such that aipbip=0a_i^p b_i^p = 0. Then Ui⊂V(Aai)=V(Aaip)U_i \subset \mathrm{V}(Aa_i) = \mathrm{V}(Aa_i^p),

Zi⊂V(Abi)=V(Abip) Z_i \subset \mathrm{V}(Ab_i) = \mathrm{V}(Ab_i^p)

and V(Aai)∩V(Abi)=V(Aai+Abi)=∅V(Aa_i) \cap V(Ab_i) = V(Aa_i + Ab_i) = \varnothing, hence Ui=V(Aaip)U_i = V(Aa_i^p) and Zi=V(Abip)Z_i = V(Ab_i^p), which establishes our assertion by replacing aia_i by AaipAa_i^p and bib_i by AbipAb_i^p. The ideals aia_i and bib_i thus chosen, it follows from the fact that σ\sigma is bijective that there exist two idempotents fi∈ai,ei∈bif_i \in a_i, e_i \in b_i such that 1=ei+fi,eifi=0,ai=Afi,bi=Aei1 = e_i + f_i, e_i f_i = 0, a_i = Af_i, b_i = Ae_i. If i≠ji \neq j, then X=Z,∪Zj=V(Aeiej)X = Z, \cup Z_j = V(Ae_i e_j), and as eieje_i e_j is idempotent, Lemma 2 shows that eiej=0e_i e_j = 0. Finally e=∑i∈Ieie = \sum_{i \in I} e_i is idempotent and ei∈Aee_i \in Ae for all i∈Ii \in I, whence V(Ae)⊂ZV(Ae) \subset Z, for all ii; it follows that

V(Ae)=∅=V(A.1) V(Ae) = \varnothing = V(A.1)

and Lemma 2 shows also that e=1e = 1.

Corollary 1

Let AA be a ring, rr a nil ideal of AA and h:A→A/rh : A \to A/r the canonical homomorphism. For every finite orthogonal family (ei′)i∈I(e'_i)_{i \in I} of idempotents of A/rA/r such that ∑i∈Iei′=1\sum_{i \in I} e'_i = 1, there exists a finite orthogonal family (ei)i∈I(e_i)_{i \in I} of idempotents of AA such that ∑i∈Iei=1\sum_{i \in I} e_i = 1 and h(ei)=ei′h(e_i) = e'_i for all i∈Ii \in I.

We write A′=A/rA' = A/r. We know (Remark following Proposition 13) that

ah:Spec⁡(A′)→Spec⁡(A) ah : \operatorname{Spec}(A') \to \operatorname{Spec}(A)

is a homeomorphism, every prime ideal of AA containing rr by hypothesis. Proposition 15 shows that there exists in AA a finite orthogonal family (ei)i∈I(e_i)_{i \in I}, of idempotents such that ∑i∈Iei=1\sum_{i \in I} e_i = 1 and that the image under ahah of V(A′(1−ei′))V(A'(1 - e'_i)) is V(A(1−ei))V(A(1 - e_i)). But clearly V(A(1−ei))V(A(1 - e_i)) is also the image under ahah of

V(A′(1−h(ei′))); V(A'(1 - h(e'_i)));

as 1−ei′1 - e'_i and 1−h(ei)1 - h(e_i) are idempotent, Lemma 2 shows that ei′=h(ei)e'_i = h(e_i), whence the corollary.

Corollary 2

For the prime spectrum X=Spec⁡(A)X = \operatorname{Spec}(A) of a ring AA to be connected, it is necessary and sufficient that AA contain no idempotents other than 00 and 11.

To say that XX is not connected means that there exists in XX a set which is open and closed and distinct from ∅\varnothing and XX.

4. The support of a module

Definition 5

Let AA be a ring and MM an AA-module. The set of prime ideals pp of AA such that Mp≠0M_p \neq 0 is called the support of MM and is denoted by Supp⁡(M)\operatorname{Supp}(M).

As every maximal ideal of AA is prime, it follows immediately from § 3, no. 3, Corollary 2 to Theorem 1, that for AA-module MM to be equal to 00, it is necessary and sufficient that Supp⁡(M)=∅\operatorname{Supp}(M) = \varnothing.

Example

Let aa be an ideal of AA; in the notation of no. 3, we have

V(a)=Supp⁡(A/a). V(a) = \operatorname{Supp}(A/a).
If pp is a prime of AA such that a∉pa \notin p, then (A/a)p=0(A/a)_p = 0 (§ 3, no. 1, Remark 3); if on the other hand a⊂pa \subset p, aApaA_p is contained in the maximal ideal pAppA_p of AA, and (A/a)p(A/a)_p is isomorphic to Ap/aApA_p/aA_p and hence is non-zero (§ 3, no. 1, Proposition 3); whence our assertion. In particular, Supp⁡(A)=Spec⁡(A)\operatorname{Supp}(A) = \operatorname{Spec}(A).

Proposition 16

Let AA be a ring and MM an AA-module. (i) If NN is a submodule of MM, then

Supp⁡(M)=Supp⁡(N)∪Supp⁡(M/N). \operatorname{Supp}(M) = \operatorname{Supp}(N) \cup \operatorname{Supp}(M/N).
(ii) If MM is the sum of a family (Ni)i∈I(N_i)_{i \in I} of submodules, then
Supp⁡(M)=⋃i∈ISupp⁡(Ni). \operatorname{Supp}(M) = \bigcup_{i \in I} \operatorname{Supp}(N_i).
(i) From the exact sequence 0→N→M→M/N→00 \to N \to M \to M/N \to 0, we derive, for every prime ideal pp of AA, the exact sequence
0→Np→Mp→(M/N)p→0 0 \to N_p \to M_p \to (M/N)_p \to 0
(§ 2, no. 4, Theorem 1). For MM, to be reduced to 0, it is necessary and sufficient that NpN_p and (M/N)p(M/N)_p be so. In other words, the relation p∉Supp⁡(M)p \notin \operatorname{Supp}(M) is equivalent to "p ∉Supp⁡(N)\notin \operatorname{Supp}(N) and p∉Supp⁡(M/N)p \notin \operatorname{Supp}(M/N)", which proves (i). (ii) For every prime ideal pp of AA, MM, is the sum of the family of submodules (Ni)p(N_i)_p (§ 2, no. 4). To say that M,≠0M, \neq 0 means that there exists i∈Ii \in I such that (Ni)p≠0(N_i)_p \neq 0, whence (ii).

Corollary

Let AA be a ring, MM an AA-module, (mi)i∈I(m_i)_{i \in I} a system of generators of MM and aa, the annihilator of mim_i. Then*

Supp⁡(M)=⋃i∈IV(ai). \operatorname{Supp}(M) = \bigcup_{i \in I} V(a_i).
Supp⁡(M)=⋃i∈ISupp⁡(Ami)\operatorname{Supp}(M) = \bigcup_{i \in I} \operatorname{Supp}(Am_i) by Proposition 16 (ii). On the other hand, AmiAm_i is isomorphic to the AA-module A/aiA/a_i and we have seen that
Supp⁡(A/ai)=V(ai) \operatorname{Supp}(A/a_i) = V(a_i)
(Example above).

Proposition 17

Let AA be a ring, MM an AA-module and aa its annihilator; if MM is finitely generated, then* Supp⁡(M)=V(a)\operatorname{Supp}(M) = V(a) and Supp⁡(M)\operatorname{Supp}(M) is therefore closed in Spec⁡(A)\operatorname{Spec}(A). Let (mi)1≤i≤n(m_i)_{1 \leq i \leq n} be a system of generators of MM and let aa, be the annihilator of mim_i; then a=⋂i=1naia = \bigcap_{i=1}^n a_i, hence V(a)=⋃i=1nV(ai)V(a) = \bigcup_{i=1}^n V(a_i) (no. 3, equation (3)) and the Proposition follows from the Corollary to Proposition 16.

Corollary 1

Let AA be a ring, MM a finitely generated AA-module and aa an element of AA. For aa to belong to every prime ideal of the support of MM, it is necessary and sufficient that the homothety of MM with ratio aa be nilpotent.

It follows from Proposition 17 that the intersection of the prime ideals belonging to Supp⁡(M)\operatorname{Supp}(M) is the radical of the annihilator aa of MM (no. 3, Proposition 11 (ii)). To say that aa belongs to this radical is equivalent to saying that there exist a power ak∈aa^k \in a and hence that akM=0a^k M = 0.

Corollary 2

Let AA be a Noetherian ring, MM a finitely generated AA-module and aa an ideal of AA. For Supp⁡(M)⊂V(a)\operatorname{Supp}(M) \subset V(a), it is necessary and sufficient that there exist an integer kk such that akM=0a^k M = 0.

If bb is the annihilator of MM, the relation Supp⁡(M)⊂V(a)\operatorname{Supp}(M) \subset V(a) is equivalent to V(b)⊂V(a)V(b) \subset V(a) by Proposition 17 and hence to a⊂r(b)a \subset r(b), where r(b)r(b) is the radical of bb (no. 3, Corollary 2 to Proposition 11). Since AA is Noetherian, this condition is also equivalent to the existence of an integer k>0k > 0 such that ak⊂ba^k \subset b (§ 2, no. 6, Proposition 15).

Proposition 18

Let M,M′M, M' be two finitely generated modules over a ring AA; then

Supp⁡(M⊗AM′)=Supp⁡(M)∩Supp⁡(M′). \operatorname{Supp}(M \otimes_A M') = \operatorname{Supp}(M) \cap \operatorname{Supp}(M').

We need to prove that, if pp is a prime ideal of AA, the relations (M⊗AM′)p≠0(M \otimes_A M')_p \neq 0 and "Mp≠0M_p \neq 0 and Mp′≠0M'_p \neq 0" are equivalent. As the ApA_p-modules M,⊗ApMp′M, \otimes_{A_p} M'_p and (M⊗AM′)p(M \otimes_A M')_p are isomorphic (§ 2, no. 7, Proposition 18), our assertion follows from the following lemma:

Lemma 3

Let BB be a local ring and EE and E′E' two finitely generated BB-modules. If E≠0E \neq 0 and E′≠0E' \neq 0, then E⊗BE′≠0E \otimes_B E' \neq 0.

Let kk be the residue field of BB. By virtue of § 3, no. 2, Proposition 4, k⊗BE≠0k \otimes_B E \neq 0 and k⊗BE′≠0k \otimes_B E' \neq 0; then we deduce that

(k⊗BE)⊗k(k⊗BE′)≠0 (k \otimes_B E) \otimes_k (k \otimes_B E') \neq 0

(Algebra, Chapter II, § 3, no. 7). But, since the tensor product is associative (loc. cit., § 3, no. 8), this tensor product is isomorphic to

E⊗B((k⊗kk)⊗BE′)=E⊗B(k⊗BE′) E \otimes_B ((k \otimes_k k) \otimes_B E') = E \otimes_B (k \otimes_B E')

and therefore to k⊗B(E⊗BE′)k \otimes_B (E \otimes_B E'), whence the lemma.

Corollary

Let MM be a finitely generated AA-module and nn its annihilator. For every ideal aa of AA, Supp⁡(M/aM)=V(a)∩V(n)=V(a+n)\operatorname{Supp}(M/aM) = V(a) \cap V(n) = V(a + n).

M/aM=M⊗A(A/a)M/aM = M \otimes_A (A/a) and A/aA/a is finitely generated.

Proposition 19

Let A,BA, B be two rings, ϕ:A→B\phi : A \to B a homomorphism and

aϕ:Spec⁡(B)→Spec⁡(A) ^{a}\phi : \operatorname{Spec}(B) \to \operatorname{Spec}(A)
the continuous mapping associated with ϕ\phi (Proposition 13). For every AA-module MM,
Supp⁡(M(B))⊂aϕ(Supp⁡(M));if also M is finitely generated, then \operatorname{Supp}(M_{(B)}) \subset ^{a}\phi(\operatorname{Supp}(M)) ; \text{if also } M \text{ is finitely generated, then}
Supp⁡(M(B))=aϕ−1(Supp⁡(M)). \operatorname{Supp}(M_{(B)}) = ^{a}\phi^{-1}(\operatorname{Supp}(M)).
Let qq be a prime ideal of BB and p=ϕ−1(q)p = \phi^{-1}(q). Suppose that qq belongs to Supp⁡(M(B))\operatorname{Supp}(M_{(B)}); then M(B)⊗BBq=(M⊗AB)⊗BBq=M⊗ABq=(M⊗AAp)⊗ABqM_{(B)} \otimes_B B_q = (M \otimes_A B) \otimes_B B_q = M \otimes_A B_q = (M \otimes_A A_p) \otimes_A B_q, since the homomorphism A→B→BqA \to B \to B_q factors into A→Ap→BqA \to A_p \to B_q (§ 2, no. 1, Proposition 2); the hypothesis M(B)⊗BBq≠0M_{(B)} \otimes_B B_q \neq 0 implies therefore M⊗AAq≠0M \otimes_A A_q \neq 0, whence the first assertion. As the homomorphism ϕq:Ap→Bq\phi_q : A_p \to B_q is local, the second assertion follows from the following lemma:

Lemma 4

Let A,BA, B be two local rings, p:A→Bp : A \to B a local homomorphism and EE a finitely generated AA-module. If E≠0E \neq 0, then E(B)≠0E_{(B)} \neq 0.

Let mm be the maximal ideal of AA and k=A/mk = A/m the residue field; the hypothesis implies that B⊗Ak=B/mB≠0B \otimes_A k = B/mB \neq 0; since the tensor product is associative, (E⊗AB)⊗Ak(E \otimes_A B) \otimes_A k is isomorphic to E⊗A(B⊗Ak)E \otimes_A (B \otimes_A k), hence also to E⊗A(k⊗k(B⊗Ak))E \otimes_A (k \otimes_k (B \otimes_A k)) and finally to (E⊗Ak)⊗k(B⊗Ak)(E \otimes_A k) \otimes_k (B \otimes_A k); by § 3, no. 2, Proposition 4, E⊗Ak≠0E \otimes_A k \neq 0, hence (E⊗AB)⊗Ak≠0(E \otimes_A B) \otimes_A k \neq 0 (Algebra, Chapter II, § 3, no. 7) and a fortiori E⊗AB≠0E \otimes_A B \neq 0.

Proposition 20

Let AA be a ring and MM a finitely generated AA-module. For every prime ideal p∈Supp⁡(M)p \in \operatorname{Supp}(M), there exists a non-zero AA-homomorphism w:M→A/pw : M \to A/p.

Let p∈Supp⁡(M)p \in \operatorname{Supp}(M). As MM is finitely generated and Mp≠0M_p \neq 0,

Mp/pMp=Mp⊗A(Ap/pAp)≠0 M_p/pM_p = M_p \otimes_A (A_p/pA_p) \neq 0
(§ 3, no. 2, Proposition 4). Let K=Ap/pApK = A_p/pA_p be the field of fractions of the integral domain A/pA/p; since Mp/pMpM_p/pM_p is a vector space over KK, which is not reduced to 0, there exists a non-zero linear form u:Mp/pMp→Ku : M_p/pM_p \to K. If (xi)1<n∗(x_i)_1 <_n^* is a system of generators of MM, xˉi\bar{x}_i the image of xix_i, in the (A/p)(A/p)-module Mp/pMpM_p/pM_p, there exists an element α≠0\alpha \neq 0 of A/pA/p such that the αu(xˉi)\alpha u(\bar{x}_i) belong to A/pA/p for 1≤i≤n1 \leq i \leq n; hence v=αuv = \alpha u is a non-zero (A/p)(A/p)-linear mapping from Mp/pMpM_p/pM_p to A/pA/p. The composition
w:M⟶M,⟶Mp/pMp⟶vA/p w : M \longrightarrow M, \longrightarrow M_p/pM_p \overset{v}{\longrightarrow} A/p
is therefore the required homomorphism.

Exercises

(b) Let n=dim⁡(M)n = \dim(M). The direct factors of M of dimension 1 (resp. n−1n - 1) are called lines (resp. hyperplanes). An automorphism u∈GL(M)u \in \mathbf{GL}(M) distinct from the identity is called a transvection if there exists a hyperplane H of M all of whose elements are invariant under uu; then u(x)=x+aϕ(x)u(x) = x + a\phi(x), where ϕ\phi is a linear form on M such that H=Ker(ϕ)H = \mathrm{Ker}(\phi) and a∈Ha \in H; obtain the converse. If A is commutative, show that every automorphism u∈GL(M)u \in \mathbf{GL}(M) of determinant 1 is a product of transvections (observe that, in the matrix of uu with respect to any basis of M, every column contains at least one invertible element of A and a matrix of the form I+Eij (i≠j)I + E_{ij} \ (i \neq j) is the matrix of a transvection).

(c) Give an example of an automorphism u∈GL(M)u \in \mathbf{GL}(M) such that the kernel of 1−u1 - u is not a direct factor of M (take A to be the local ring K[[X]]K[[X]] of formal power series in one indeterminate over a field K and M=AM = A).

(d) Give an example of direct factors N, P of M (necessarily free) such that N+PN + P and N∩PN \cap P are not direct factors of M. (Take A=K[X]/(X2)A = K[X]/(X^2), where K is a field and M=A2M = A^2.)

¶ 15. Let A be a (commutative) ring and M an A-module.

(a) For a submodule M' of M to be pure (Chapter I, § 2, Exercise 24), it is necessary and sufficient that, for every maximal ideal mm of A, Mm′M'_m be a pure sub-&-module of MmM_m (use Theorem 1 of no. 3).

(b) Suppose that A is a local ring with maximal ideal mm and M a finitely generated free A-module. For a finitely generated submodule M' of M to be pure, it is necessary and sufficient that it be a direct factor of M. (Using Corollary 1 to Proposition 5 of no. 2, reduce it to proving that, if M′⊂mMM' \subset mM, M′M' can only be a pure submodule of M if M′=0M' = 0.)

16. Let (Aλ,fμλ)(A_\lambda, f_{\mu\lambda}) be a direct system of local rings, such that the fμλf_{\mu\lambda} are local homomorphisms; let mλm_\lambda be the maximal ideal of AλA_\lambda and Kλ=Aλ/mλK_\lambda = A_\lambda/m_\lambda. Then A=lim←⁡AλA = \varprojlim A_\lambda is a local ring whose maximal ideal is m=lim←⁡mλm = \varprojlim m_\lambda and residue field is Kˉ=lim←⁡Kλ\bar{K} = \varprojlim K_\lambda. Moreover, if mμ=Aμmλm_\mu = A_\mu m_\lambda for λ<μ\lambda < \mu, then m=Amm = Am, for all λ\lambda.

94

See the exercises for § 4.

Transcribed from the printed volume. Not translated.