Bourbaki

3. Local rings. Passage from the local to the global

1. Local rings

Proposition 1

Let A\mathbf{A} be a ring and II the set of non-invertible elements of A\mathbf{A}. The set II is the union of the ideals of A\mathbf{A} which are distinct from A\mathbf{A}. Moreover, the following conditions are equivalent: (a) II is an ideal. (b) The set of ideals of A\mathbf{A} distinct from A\mathbf{A} has a greatest element. (c) A\mathbf{A} has a unique maximal ideal.

Cited by 2

Cited by Definition 1, Definition 3

The relation x∈Ix \in I is equivalent to 1∉xA1 \notin xA and hence xA≠AxA \neq \mathbf{A}. If aa is an ideal of A\mathbf{A} distinct from A\mathbf{A} and x∈ax \in a, then xA⊂axA \subset a, hence xA≠AxA \neq \mathbf{A} and x∈Ix \in I. Hence every ideal of A\mathbf{A} distinct from A\mathbf{A} is contained in II and every element x∈Ix \in I belongs to a principal ideal xA≠AxA \neq \mathbf{A}. This proves the first assertion, which immediately implies the equivalence of (a), (b) and (c).

Remark (1). Note that, if (c) holds, II is the Jacobson radical of the ring A\mathbf{A} (Algebra, Chapter VIII, §6, no. 3, Definition 3).

Definition 1

A ring A\mathbf{A} is called a local ring if it satisfies the equivalent conditions (a), (b) and (c) of Proposition 1. The quotient of A\mathbf{A} by its Jacobson radical (which is then the unique maximal ideal of A\mathbf{A}) is called the residue field of A\mathbf{A}.

Cites 1

Cites Proposition 1

Definition 2

Let A,B\mathbf{A}, \mathbf{B} be two local rings and m,nm, n their respective maximal ideals. A homomorphism u:A→Bu : \mathbf{A} \to \mathbf{B} is called local if u(m)⊂nu(m) \subset n.

Cites 2

Cites Proposition 4, A IV § 5, TG I § 6. Filters

This amounts to saying that uˉ−1(n)=m\bar{u}^{-1}(n) = m, for uˉ−1(n)\bar{u}^{-1}(n) is then an ideal containing mm and not containing 1 and hence equal to mm. Taking quotients, we then derive canonically from uu an injective homomorphism A/m→B/n\mathbf{A}/m \to \mathbf{B}/n from the residue field of A\mathbf{A} to that of B\mathbf{B}.

Examples (1) A field is a local ring. A ring which is reduced to 0 is not a local ring. (2) Let A\mathbf{A} be a local ring and kk its residue field. The ring of formal power series B=A[[X1,…,Xn]]B = \mathbf{A}[[X_1, \ldots, X_n]] is a local ring, for the non-invertible elements of BB are the formal power series whose constant terms are not invertible in A\mathbf{A} (Algebra, Chapter IV, §5, no. 6, Proposition 4). The canonical injection of A\mathbf{A} into BB is a local homomorphism and the corresponding injection of residue fields is an isomorphism. (3) Let bb be an ideal of a ring A\mathbf{A} which is only contained in a single maximal ideal mm; then A/bA/b is a local ring with maximal ideal m/bm/b and residue field canonically isomorphic to A/mA/m. This applies in particular to the case b=mkb = m^k, mm being any maximal ideal of AA (§1§ 1, no. 1, Corollary to Proposition 1). If AA itself is a local ring with maximal ideal mm, then for every ideal b≠Ab \neq A of AA, A/bA/b is a local ring, the canonical homomorphism A→A/bA \to A/b a local homomorphism and the corresponding homomorphism of residue fields an isomorphism.

(4) Let XX be a topological space, x0x_0 a point of XX and AA the ring of germs at the point x0x_0 of real-valued functions continuous in a neighbourhood of x0x_0 (General Topology, Chapter I, § 6, no. 10). Clearly, for the germ at x0x_0 of a continuous function ff to be invertible in AA, it is necessary and sufficient that f(x0)≠0f(x_0) \neq 0, as that implies that f(x)≠0f(x) \neq 0 in a neighbourhood of x0x_0. The ring AA is therefore a local ring whose maximal ideal mm is the set of germs of functions which are zero at x0x_0; taking quotients, the mapping g↦g(x0)g \mapsto g(x_0) of AA to R\mathbf{R} gives an isomorphism of the residue field A/mA/m onto R\mathbf{R}.

Proposition 2

Let AA be a ring and pp a prime ideal of AA. The ring AA, is local; its maximal ideal is the ideal pAp=pvpA_p = p_v, generated by the canonical image of pp in AA; its residue field is canonically isomorphic to the field of fractions of A/pA/p.

Cites 2 · Cited by 2

Cites Proposition 10, Proposition 11

Cited by Proposition 3, Theorem 1

Let S=A−pS = A - p and j:A→Avj : A \to A_v, be the canonical homomorphism; the hypothesis that pp is prime implies that pp is saturated with respect to SS, hence j−1(pAp)=pj^{-1}(pA_p) = p (§2§ 2, no. 4, Proposition 10) and, as the ideals of AA not meeting SS are those contained in pp, the first two assertions are special cases of §2§ 2, no. 5, Proposition 11 (ii). Moreover, iff is the canonical homomorphism A→A/pA \to A/p, f(S)f(S) is the set of elements ≠0\neq 0 of the integral domain A/pA/p and hence the last assertion is a special case of §2§ 2, no. 5, Proposition 11 (i).

Definition 3

Let AA be a ring and pp a prime ideal of AA. The ring AA, is called the local ring of AA at pp, or the local ring of pp, when there is no ambiguity.

Cites 1

Cites Proposition 1

Remark (2). If AA is a local ring and mm its maximal ideal, the elements of A−mA - m are invertible (Proposition 1) and hence Am′A_{m'} is canonically identified with AA (§2§ 2, no. 1, Remark 5).

Examples (5) Let pp be a prime number. The local ring Z(p)\mathbf{Z}_{(p)} is the set of rational numbers a/ba/b, where a,ba, b are rational integers with bb prime top; the residue field of Z(p)\mathbf{Z}_{(p)} is isomorphic to the prime field Fp=Z/(p)\mathbf{F}_p = \mathbf{Z}/(p).

(6) Let VV be an affine algebraic variety, AA the ring of regular functions on VV, WW an irreducible subvariety of VV and pp the (necessarily prime) ideal of AA consisting of the functions which are zero at every point of WW. The ring AA, is called the local ring of WW on VV. \*

Proposition 3

Let AA be a ring, pp a prime ideal of AA and S=A−pS = A - p. For every ideal b′b' of ApA_p distinct from ApA_p, let bb be the ideal (iAS)−1(b′)(i_A^S)^{-1}(b') of AA so that b′=bApb' = bA_p.

Cites 2 · Cited by 4

Cites Proposition 2, Proposition 11

Cited by Corollary, Example, II § 4, Corollary 1, VII § 1, Theorem 3, VII § 1

(i) Let ff be the canonical homomorphism A→A/bA \to A/b. The homomorphism from AA, to (A/b)p/b(A/b)_{p/b} canonically associated with ff (§2§ 2, no. 1, Proposition 2) is surjective and its kernel is b′b', which defines, by taking quotients, a canonical isomorphism of Ap/b′A_p/b' onto (A/b)p/b(A/b)_{p/b}.

(ii) The mapping b′→b=(iAS)−1(b′)b' \to b = (i_A^S)^{-1}(b'), restricted to the set of prime ideals of ApA_p, is an isomorphism (with respect to inclusion) of this set onto the set of prime ideals of AA contained in pp. If b′b' is prime in ApA_p, there exists an isomorphism of the ring AbA_b onto the ring (Ap)b′(A_p)_{b'}, which maps a/sa/s to (a/1)/(s/1)(a/1)/(s/1) for all a∈A,s∈A−ba \in A, s \in A - b.

This is just a particular case of §2§ 2, no. 5, Proposition 11.

Remarks (3) If aa is an ideal of AA not contained in pp, then aAp=AaA_p = A, and (A/a)p=0(A/a)_p = 0 (§2§ 2, no. 5, Remark). (4) Let A,BA, B be two rings, ρ:A→B\rho : A \to B a homomorphism, qq a prime ideal of BB and pp the prime ideal ρ−1(q)\rho^{-1}(q) of AA. As ρ(A−p)⊂B−q\rho(A - p) \subset B - q, a canonical homomorphism ρq:A→Bq\rho_q : A \to B_q is derived from ρ\rho (§2§ 2, no. 1, Proposition 2) and it is immediate that ρq(pAp)⊂qBq\rho_q(pA_p) \subset qB_q, hence ρ\rho is a local homomorphism.

2. Modules over a local ring

Proposition 4

Let AA be a ring which is not necessarily commutative, mm a right ideal of AA contained in the Jacobson radical of AA and MM a left AA-module. Suppose that one of the following conditions holds : (i) MM is finitely generated; (ii) mm is nilpotent. Then the relation (Ad/m)⊗AM=0(A_d/m) \otimes_A M = 0 implies M=0M = 0.

Cited by 6

Cited by Proposition 11, Corollary 1, Proposition 5, Lemma 3, II § 4, Lemma 4, II § 4, Proposition 20, II § 4

The assertion with respect to hypothesis (i) is precisely Corollary 3 to Proposition 6 of Algebra, Chapter VIII, §6§ 6, no. 3. On the other hand, the relation (Ad/m)⊗AM=0(A_d/m) \otimes_A M = 0 is equivalent to M=mMM = mM and hence implies M=mnMM = m^nM for every integer n>0n > 0; whence the assertion with respect to hypothesis (ii).

Corollary 1

Let AA be a ring which is not necessarily commutative, mm a right ideal of AA contained in the Jacobson radical of AA, MM and NN two left AA-modules and u:M→Nu : M \to N an AA-linear mapping. If NN is finitely generated or mm is nilpotent and

1⊗u:(Ad/m)⊗AM→(Ad/m)⊗AN 1 \otimes u : (A_d/m) \otimes_A M \to (A_d/m) \otimes_A N
is surjective, then uu is surjective.

Cites 2 · Cited by 5

Cites Proposition 4, Proposition 6

Cited by Exercise 4, § 3, Corollary 2, Proposition 6, Proposition 13, III § 2, Theorem 2, VI § 8

(Ad/m)⊗A(N/u(M))(A_d/m) \otimes_A (N/u(M)) is canonically isomorphic to

((Ad/m)⊗AN)/Im⁡(1⊗u) ((A_d/m) \otimes_A N)/\operatorname{Im}(1 \otimes u)

(Algebra, Chapter II, , no. 6, Corollary 1 to Proposition 6); then the hypothesis implies (Ad/m)⊗A(N/u(M))=0(A_d/m) \otimes_A (N/u(M)) = 0, hence N/u(M)=0N/u(M) = 0 by Proposition 4

Corollary 2

Let AA be a ring which is not necessarily commutative, mm a two-sided ideal of AA contained in the Jacobson radical of AA, MM a left AA-module and (xi)i∈I(x_i)_{i \in I} a family of elements of MM. If MM is finitely generated or mm is nilpotent and the elements 1⊗xi1 \otimes x_i (i∈Ii \in I) generate the left (A/m)(A/m)-module (A/m)⊗AM(A/m) \otimes_A M, the xix_i generate MM.

Cites 1 · Cited by 5

Cites Corollary 1

Cited by Exercise 13, § 3, Exercise 3, § 3, Proposition 5, Proposition 7, Exercise 18, § 2

Let (ei)i∈I(e_i)_{i \in I} be the canonical basis of the left AA-module As(I)A_s^{(I)}: it is sufficient to apply Corollary 1 to the AA-linear mapping u:As(I)→Mu : A_s^{(I)} \to M such that u(ei)=xiu(e_i) = x_i for all i∈Ii \in I.

Proposition 5

Let AA be a ring which is not necessarily commutative, mm a two-sided ideal of AA contained in the Jacobson radical of AA and MM a left AA-module. Suppose that one of the following conditions holds: (i) MM is finitely presented; (ii) mm is nilpotent. Then, if (A/m)⊗AM=M/mM(A/m) \otimes_A M = M/mM is a free left (A/m)(A/m)-module and the canonical homomorphism from m⊗AMm \otimes_A M to MM is injective, MM is a free AA-module. To be precise, if (xi)i∈I(x_i)_{i \in I} is a family of elements of MM such that (1⊗xi)(1 \otimes x_i) is a basis of the (A/m)(A/m)-module M/mMM/mM, (xi)(x_i) is a basis of MM.

Cites 4 · Cited by 7

Cites Lemma 9, I § 2, Proposition 4, Corollary 2, Proposition 2, I § 1

Cited by Corollary 1, Corollary 2, Corollary 5, Proposition 5, II § 5, Corollary 2, III § 3, Exercise 2, § 4, Exercise 9, § 3

If a∈Aa \in A, x∈Mx \in M and aˉ\bar{a} is the class of aa in A/mA/m, then aˉ⊗x=1⊗(ax)\bar{a} \otimes x = 1 \otimes (ax) and hence the hypothesis implies that there exists a family (xi)i∈I(x_i)_{i \in I} of elements of MM such that (1⊗xi)(1 \otimes x_i) is a basis of the (A/m)(A/m)-module (A/m)⊗AM(A/m) \otimes_A M. We already know that the xix_i generate MM (Corollary 2 to Proposition 4); we shall see that they are linearly independent over AA. To this end, let us consider the free AA-module L=As(I)L = A_s^{(I)}; let (ei)(e_i) be its canonical basis and u:As(I)→Mu : A_s^{(I)} \to M the AA-linear mapping such that u(ei)=xiu(e_i) = x_i for all i∈Ii \in I; if RR is the kernel of uu, we shall prove that R=0R = 0. Under hypothesis (i), (A/m)⊗AM(A/m) \otimes_A M is a finitely generated (A/m)(A/m)-module, hence II is necessarily finite and RR is a finitely generated AA-module by Chapter I, § 2, no. 8, Lemma 9. Then by Proposition 4 it will be sufficient to prove (under either hypothesis) that R=mRR = mR.

Let jj be the canonical injection R→LR \to L; then there is a commutative diagram

m⊗R→1⊗jm⊗L→1⊗um⊗M↓a↓c↓cR→jL→uM \begin{array}{ccccc} m \otimes R & \xrightarrow{1 \otimes j} & m \otimes L & \xrightarrow{1 \otimes u} & m \otimes M \\ \downarrow^a & & \downarrow^c & & \downarrow^c \\ R & \xrightarrow{j} & L & \xrightarrow{u} & M \end{array}

in Which the two rows are exact, jj is injective and 1⊗u1 \otimes u is surjective (Chapter

0→dCoker(a)⟶Coker(b)→vCoker(c) 0 \xrightarrow{d} \mathrm{Coker}(a) \longrightarrow \mathrm{Coker}(b) \xrightarrow{v} \mathrm{Coker}(c)

(Chapter I, § 1, no. 4, Proposition 2); it is sufficient to verify that v is bijective, for then we deduce that Coker(a) = 0, in other words that a is surjective and consequently R = mR. Now, Coker(b) = (A/m) \otimes_A L and

Coker(c)=(A/m)⊗AM \mathrm{Coker}(c) = (A/m) \otimes_A M

and by definition v(1 \otimes e_i) = 1 \otimes x_i; as (1 \otimes e_i) is a basis of (A/m) \otimes_A L, the definition of the x_i shows that v is bijective.

Corollary 1

Let A be a not necessarily commutative ring, m the Jacobson radical of A and M a left A-module. Suppose that A/m is a field, that the canonical homomorphism from m \otimes_A M to M is injective and that one of conditions (i), (ii) of Proposition 5 is satisfied. For a family (y_\lambda) of elements of M to be a basis of a direct factor of M, it is necessary and sufficient that the family (1 \otimes y_\lambda) be free in M/mM.

Cites 2 · Cited by 3

Cites Proposition 5, Theorem 2, A II § 7

Cited by Exercise 13, § 3, Exercise 4, § 5, Lemma 7, VII § 4

If this condition holds, it can be assumed that (y_\lambda) is a subfamily of a family (x_i) of elements of M such that (1 \otimes x_i) is a basis of M/mM (Algebra, Chapter II, § 7, no. 1, Theorem 2) and Proposition 5 then proves that (x_i) is a basis of M.

Corollary 2

Let A be a not necessarily commutative ring, m the Jacobson radical of A and M a left A-module. Suppose that A/m is a field and that one of the following conditions holds: (i) M is finitely presented; (ii) m is nilpotent. Then the following properties are equivalent: (a) M is free; (b) M is projective; (c) M is flat; (d) the canonical homomorphism m \otimes_A M \to M is injective; (e) Tor_1^A(A/m, M) = 0.

Cites 2 · Cited by 3

Cites Proposition 5, Exercise 3, § 3

Cited by Corollary, Proposition 6, Corollary 2, II § 5

The implications (a) \Rightarrow (b) \Rightarrow (c) \Rightarrow (d) are immediate. As A/m is a field, (A/m) \otimes_A M is a free (A/m)-module and Proposition 5 shows that (d) implies (a).

Finally, we know that Tor_1^A(A, M) = 0 and from the exact sequence 0 \to m \to A \to A/m \to 0 we therefore derive the exact sequence

0→Tor1A(A/m,M)→m⊗AM→M; 0 \to \mathrm{Tor}_1^A(A/m, M) \to m \otimes_A M \to M;

this proves that Tor1A(A/m,M)_1^A(A/m, M) is isomorphic to the kernel of the canonical homomorphism m⊗AM→Mm \otimes_A M \to M; whence the equivalence of (d) and (e). \*

It can be shown that, for every ring AA with Jacobson radical mm such that A/mA/m is a field, every projective AA-module is free (Exercise 3).

Proposition 6

Let AA be a ring which is not necessarily commutative and mm its Jacobson radical; suppose that A/mA/m is a field. Let MM and NN be two finitely generated free AA-modules and u:M→Nu: M \to N a homomorphism. The following properties are equivalent*:

Cites 5 · Cited by 5

Cites Lemma 8, I § 2, Corollary 1, Corollary 2, Proposition 21, A II § 1, Proposition 8, A II § 5

Cited by Exercise 4, § 3, Corollary, Corollary, Corollary 2, Corollary 1

(a) uu is an isomorphism of MM onto a direct factor of NN; (b) 1⊗u:(A/m)⊗AM→(A/m)⊗AN1 \otimes u: (A/m) \otimes_A M \to (A/m) \otimes_A N is injective; (c) uu is injective and Coker⁡(u)\operatorname{Coker}(u) is a free AA-module; (d) the transpose homomorphism tu:N∗→M∗{}^t u: N^* \to M^* is surjective.

We know (Algebra, Chapter II, § 1, no. 11, Proposition 21) that, if N/u(M)N/u(M) is free, u(M)u(M) is a direct factor of NN, hence (c) implies (a); conversely, (a) implies that Coker⁡(u)\operatorname{Coker}(u), isomorphic to a complement of u(M)u(M) in NN, is a finitely generated projective AA-module and a fortiori finitely presented (Chapter I, § 2, no. 8, Lemma 8); hence this module is free by Corollary 2 to Proposition 5 and (a) implies (c). On the other hand, (a) obviously implies (b). For simplicity we write M′=(A/m)⊗AMM' = (A/m) \otimes_A M, N′=(A/m)⊗ANN' = (A/m) \otimes_A N; as MM and NN are finitely generated, the duals M′∗{M'}^* and N′∗{N'}^* of the (A/m)(A/m)-modules M′M' and N′N' are canonically identified with M∗⊗A(A/m)M^* \otimes_A (A/m) and N∗⊗A(A/m)N^* \otimes_A (A/m) and t(1⊗u){}^t(1 \otimes u) with (tu)⊗1({}^t u) \otimes 1 (Algebra, Chapter II, § 5, no. 4, Proposition 8); as M′M' and N′N' are vector spaces over the field A/mA/m, the hypothesis that 1⊗u1 \otimes u is injective implies that t(1⊗u){}^t(1 \otimes u) is surjective (Algebra, Chapter 11, § 7, no. 5, Proposition 10); Corollary 1 to Proposition 4 then shows that tu{}^t u is surjective and we have thus proved that (b) implies (d). Finally we show that (d) implies (a). Suppose that tu{}^t u is surjective; as M∗M^* is free, there exists a homomorphism ff from M∗M^* to N∗N^* such that 1M∗=tu∘f1_{M^*} = {}^t u \circ f (Algebra, Chapter II, § 1, no. 11, Proposition 21); as MM and NN are free and finitely generated, there exists a homomorphism gg from NN to MM such that f=tgf = {}^t g; hence t1M=1M∗=tu∘tg=t(g∘u){}^t 1_M = 1_{M^*} = {}^t u \circ {}^t g = {}^t(g \circ u), whence 1M=g∘u1_M = g \circ u; this proves that uu is an isomorphism of MM onto a submodule which is a direct factor of NN (Algebra, Chapter 11, § 1, no. 9, Corollary 2 to Proposition 15).

Corollary

Under the hypotheses of Proposition 6 the following properties are equivalent*:

Cites 4 · Cited by 1

Cites Proposition 6, A II § 7. Vector spaces, Corollary, Corollary 2

Cited by Theorem 1, II § 5

(a) uu is an isomorphism of MM onto NN; (b) MM and NN have the same rank (Algebra, Chapter II, § 7, no. 2) and uu is surjective; (c) 1⊗u:M/mM→N/mN1 \otimes u: M/mM \to N/mN is bijective.

Clearly (a) implies (b); (b) implies that 1⊗u1 \otimes u is surjective; moreover the hypothesis that M and N have the same rank implies that so do the vector spaces (A/m)⊗AM(A/m) \otimes_A M and (A/m)⊗AN(A/m) \otimes_A N over A/mA/m, hence 1⊗u1 \otimes u is bijective (Algebra, Chapter II, , no. 4, Corollary to Proposition 9) and (b) implies (c). Finally, condition (c) implies, by Proposition 6, that N is the direct sum of u(M)u(M) and a free submodule P and uu is an isomorphism of M onto u(M)u(M); if P≠0P \neq 0, then (A/m)⊗AP#0(A/m) \otimes_A P \# 0 and 1⊗u1 \otimes u would not be surjective; hence (c) implies (a)

The propositions proved above in this no. will usually be applied when AA is a local ring and mm its maximal ideal. Corollary 2 to Proposition 5 is then completed by

Proposition 7

Let AA be a reduced local ring, mm its maximal ideal, (pi)i∈I(p_i)_{i \in I} the family of minimal prime ideals of AA, KiK_i the field of fractions of A/piA/p_i and MM a finitely generated AA-module. For MM to be free it is necessary and sufficient that

Cites 3 · Cited by 1

Cites Corollary 2, A II § 1. Modules, Proposition 13

Cited by Exercise 5, § 3

(1)[(A/m)⊗AM:(A/m)]=[K,⊗AM:K]for all i∈I.(1)\qquad [(A/m) \otimes_A M : (A/m)] = [K, \otimes_A M : K]\qquad \text{for all } i \in I.

If M is free, clearly the two sides of (1) are equal to the rank of M for all i∈Ii \in I. Suppose now that the condition is satisfied and denote by nn the common value of the two sides of (1); by Corollary 2 to Proposition 4 M has a system of nn generators xjx_j (1≤j≤n1 \leq j \leq n). Suppose first that AA is an integral domain, in which case pi=0p_i = 0 for all i∈Ii \in I. The elements 1⊗xj1 \otimes x_j (1≤j≤n1 \leq j \leq n) generate the vector space K⊗MK \otimes M over the field of fractions K of AA; but as by hypothesis this space is of dimension nn over K, the elements 1⊗xj1 \otimes x_j are linearly independent over K. It follows (Algebra, Chapter II, § 1, no. 13, Remark 1) that the xjx_j are linearly independent over AA and hence form a basis of M.

Passing to the general case, there exists a surjective homomorphism vv from L=A′′L = A'' onto M. Consider the commutative diagram

L→vM∣u∣↓↓u′∏i((A/pi)⊗L)→v′∏i((A/pi)⊗M) \begin{array}{ccc} L & \xrightarrow{v} & M \\ |u| \downarrow & & \downarrow u' \\ \prod_i ((A/p_i) \otimes L) & \xrightarrow{v'} & \prod_i ((A/p_i) \otimes M) \end{array}

where uu (resp. u′u') is the mapping x↦(ϕi(x))x \mapsto (\phi_i(x)) (resp. y↦(ψi(y))y \mapsto (\psi_i(y))),

ϕi:L→(A/pi)⊗L \phi_i : L \to (A/p_i) \otimes L

(resp. ψi:M→(A/pi)⊗M\psi_i : M \to (A/p_i) \otimes M) being the canonical mapping, and v′v' is the product of the 1A/pi@ve1_{A/p_i} @ v_e. Then (A/p)/(m/pi)⊗A/pi((A/pi)⊗AM)=(A/m)⊗AM(A/p)/(m/p_i) \otimes_{A/p_i} ((A/p_i) \otimes_A M) = (A/m) \otimes_A M and, as A/piA/p_i is a local integral domain, it follows from the first part of the argument that each of the 1A/pi⊗v1_{A/p_i} \otimes v is an isomorphism; then so is v′v'. On the other hand, as AA is reduced, ⋂ipi=(0)\bigcap_i p_i = (0) (§2§ 2, no. 6, Proposition 13), whence ⋂i(piL)=0\bigcap_i (p_i L) = 0 since LL is free (Algebra, Chapter II, §3§ 3, no. 7, Remark); as piLp_i L is the kernel of ϕi\phi_i, this shows that uu is injective. It follows that v′∘u=u′∘vv' \circ u = u' \circ v is injective, hence vv is injective and, as vv is surjective by definition, this shows that MM is free.

3. Passage from the local to the global

Proposition 8

Let AA be a ring, mm a maximal ideal of AA and MM an AA-module. If there exists an ideal aa of AA such that mm is the only maximal ideal of AA containing aa and aM=0aM = 0, then the canonical homomorphism M→MM \to M, is bijective.

Cited by 2

Cited by Proposition 9, Proposition 5, VI § 8

A/aA/a is then a local ring with maximal ideal m/am/a; MM can be considered as an (A/a)(A/a)-module; for all s∈A−ms \in A - m the canonical image of ss in A/aA/a is invertible, hence the homothety x↦sxx \mapsto sx of MM is bijective from the definition of MM, as the solution of a universal problem (§2§ 2, no. 2); whence the proposition.

In particular, if there exists k⩾0k \geqslant 0 such that mkM=0m^k M = 0, the homomorphism M→MmM \to M_m is bijective (§1§ 1, no. 1, Corollary to Proposition 1).

Proposition 9

Let AA be a ring, mm a maximal ideal of AA, MM an AA-module and kk an integer ⩾0\geqslant 0. The canonical homomorphism M→Mm/mkMmM \to M_m/m^k M_m is surjective, has kernel mkMm^k M and defines an isomorphism of M/mkMM/m^k M onto Mm/mkMmM_m/m^k M_m.

Cites 3 · Cited by 5

Cites Corollary, II § 2, Theorem 1, II § 2, Proposition 8

Cited by Corollary, Corollary, Corollary 2, Proposition 18, III § 2, Proposition 8, III § 3

Since the case k=0k = 0 is trivial, suppose that k⩾1k \geqslant 1. It follows from Proposition 8 that the canonical homomorphism M/mkM→(M/mkM)mM/m^k M \to (M/m^k M)_m is bijective. On the other hand (M/mkM)m(M/m^k M)_m is canonically identified with Mm/(mkM)mM_m/(m^k M)_m (§ 2, no. 4, Theorem 1) and hence (mkM)m=mkMm(m^k M)_m = m^k M_m (§ 2, no. 7, Corollary to Proposition 18), whence there is an isomorphism of M/mkMM/m^k M onto Mm/mkMmM_m/m^k M_m which maps the class of an element x∈Mx \in M to the class of x/1x/1.

Corollary

Let AA be a ring, m1,m2,…,mnm_1, m_2, \ldots, m_n distinct maximal ideals of AA, MM an AA-module and k1,k2,…,knk_1, k_2, \ldots, k_n integers ⩾0\geqslant 0. The canonical homomorphism from MM to ∏nMm/mi1k1Mm\prod^n M_m/m_{i_1}^{k_1} M_m is surjective and its kernel is (⋂nmi1k1)M\left( \bigcap^n m_{i_1}^{k_1} \right) M.

Cites 3 · Cited by 1

Cites Proposition 3, Proposition 6, Proposition 9

Cited by Corollary

This follows easily from Proposition 9 and §1§ 1, no. 2, Proposition 6, the mi1k1m_{i_1}^{k_1} being relatively prime in pairs (§1§ 1, no. 2, Proposition 3).

In the rest of this no. AA will denote a ring and Ω(A)\Omega(A) (or Ω\Omega) the set of maximal ideals of AA.

Proposition 10

The A-module* ⨁m∈ΩAm\bigoplus_{m \in \Omega} \mathbf{A}_m, the direct sum of the Am\mathbf{A}_m where m∈Ωm \in \Omega, is faithfully flat.

Cites 3 · Cited by 2

Cites Proposition 2, I § 2, Proposition 1, I § 3, Theorem 1, II § 2

Cited by Proposition 2, Theorem 1

Each of the Am\mathbf{A}_m is a flat A-module (§ 2, no. 4, Theorem 1), hence E=⨁m∈ΩAmE = \bigoplus_{m \in \Omega} \mathbf{A}_m is flat (Chapter I, § 2, no. 3, Proposition 2). Moreover, for every maximal ideal mm of AA, mAmmA_m is the unique maximal ideal of Am\mathbf{A}_m, hence mAm≠AmmA_m \neq A_m, whence it follows that mE≠EmE \neq E and consequently EE is faithfully flat (Chapter I, § 3, no. 1, Proposition 1 (d)).

Theorem 1

Let* M,NM, N be two A-modules, u:M→Nu : M \to N an A-homomorphism and, for all m∈Ωm \in \Omega, let um:M,→Nmu_m : M, \to N_m be the corresponding AmA_m-*homomorphism* (§ 2, no. 2, Remark 5). For uu to be injective (resp. surjective, bijective, zero), it is necessary and sufficient that, for all m∈Ωm \in \Omega, umu_m be injective (resp. surjective, bijective, zero).

Cites 3 · Cited by 11

Cites Proposition 1, I § 3, Proposition 10, Proposition 2

Cited by Exercise 7, § 3, Proposition 11, Proposition 13, Proposition 15, Proposition 17, Corollary 1, Corollary 5, Exercise 17, § 5, Exercise 9, § 5, Theorem 1, II § 5, Corollary 5, III § 3

To say that, for all m∈Ωm \in \Omega, umu_m is injective (resp. surjective, bijective, zero) is equivalent to saying that the homomorphism ⨁mum:⨁M,→⨁N\bigoplus_m u_m : \bigoplus M, \to \bigoplus N, has the same property. But ⨁mMm=M⊗AE,⨁mNm=N⊗AE\bigoplus_m M_m = M \otimes_A E, \bigoplus_m N_m = N \otimes_A E and ⨁mum=u⊗1\bigoplus_m u_m = u \otimes 1, where E=⨁mAmE = \bigoplus_m \mathbf{A}_m; as EE is faithfully flat (Proposition 10), the theorem follows from Chapter I, § 3, no. 1, Proposition 1 (c) and Proposition 2.

Corollary 1

Let* MM be an A-module, N a submodule of MM and xx an element of MM. For x∈Nx \in N, it is necessary and sufficient that, for all m∈Rm \in R, the canonical image of xx in MM, belong to NN.

Cites 2 · Cited by 5

Cites Theorem 1, II § 2, Theorem 1

Cited by Exercise 11, § 3, Proposition 12, Corollary 2, Corollary 3, Corollary 4

Let xˉ\bar{x} be the class of xx in M/NM/N; to say that x∈Nx \in N means that the A-linear mapping u:a↦axˉu : a \mapsto a\bar{x} from AA to M/NM/N is zero. Now, (M/N)m(M/N)_m is identified with Mm/NmM_m/N_m (§ 2, no. 4, Theorem 1) and um:Am→Mm/Nmu_m : A_m \to M_m/N_m with the mapping A↦λxˉmA \mapsto \lambda \bar{x}_m, where xˉm\bar{x}_m is the class mod. NN, of the canonical image of xx in MM. As the relation u=0u = 0 is equivalent to um=0u_m = 0 for all mm by Theorem 1, this proves the corollary.

Corollary 2

Let* MM be an A-module and, for all m∈Ωm \in \Omega, let fmf_m be the canonical mapping M→MmM \to M_m. The homomorphism x↦(fm(x))x \mapsto (f_m(x)) of MM to ∏m∈ΩMm\prod_{m \in \Omega} M_m is injective.

Cites 1 · Cited by 3

Cites Corollary 1

Cited by Definition 5, II § 4, Definition 2, II § 5, Corollary 3, V § 2

Applying Corollary 1 in the case N=0N = 0, it is seen that the relation x=0x = 0 is equivalent to fm(x)=0f_m(x) = 0 for all m∈Ωm \in \Omega.

Corollary 3

(i) Let bb be an ideal of AA and aa an element of AA. For a∈ba \in b, it is necessary and sufficient that, for all m∈Ωm \in \Omega, the canonical image of aa in AmA_m belong to bAmbA_m.

Cites 2

Cites Corollary, II § 2, Corollary 1

(ii) In particular, let b and c be two elements of A. For c to be a multiple of b, it is necessary and sufficient that, for all m∈Ωm \in \Omega, the canonical image of c in A, be a multiple of that of b.

As bAm=bmbA_m = b_m (§ 2, no. 7, Corollary to Proposition 18), (i) is a special case of Corollary 1; (ii) follows from (i) applied to the ideal AbAb.

Corollary 4

Let A be an integral domain, K its field of fractions and M a torsion-free A-module such that M is canonically identified with a sub-A-module of K⊗AMK \otimes_A M. Then, for all m∈Ωm \in \Omega, MmM_m is canonically identified with a sub-A-module of K⊗AMK \otimes_A M and M=⋂m∈ΩMmM = \bigcap_{m \in \Omega} M_m.

Cites 2

Cites Theorem 1, II § 2, Corollary 1

As M is identified with a submodule of K⊗AMK \otimes_A M, M, is identified with a sub-\&-module of (K⊗AM)m=Km⊗AM(K \otimes_A M)_m = K_m \otimes_A M (§ 2, no. 4, Theorem 1); as Km=KK_m = K, we see straightway that MmM_m is torsion-free; moreover, the commutativity of the diagram

M⟶K⊗AM↓↑Mm⟶(K⊗AM)m \begin{array}{ccc} M & \longrightarrow & K \otimes_A M \\ \downarrow & & \uparrow \\ M_m & \longrightarrow & (K \otimes_A M)_m \end{array}

proves that the canonical mapping M→MmM \to M_m is injective. The corollary then follows from Corollary 1 applied to the A-module K⊗AMK \otimes_A M and its submodule M.

In particular, for every integral domain A,

(2)

A=⋂m∈ΩAm. A = \bigcap_{m \in \Omega} A_m.

Corollary 5

Let A be a ring. Every system of generators of the A-module AnA^n with n elements is a basis of AnA^n.

Cites 2

Cites Proposition 5, Theorem 1

Let (ei)1≤i≤n(e_i)_{1 \leq i \leq n} be the canonical basis of A′′A'', (xi)1≤i≤n(x_i)_{1 \leq i \leq n} a system of generators of AnA^n with n elements and u:An→A′′u : A^n \to A'' the A-linear mapping such that u(ei)=xiu(e_i) = x_i for 1≤i≤n1 \leq i \leq n. By hypothesis uu is surjective and it is necessary to show that uu is injective. By virtue of Theorem 1 this can immediately be reduced to the case where A is a local ring; if mm is the maximal ideal of A, the elements 1⊗xi1 \otimes x_i (1≤i≤n1 \leq i \leq n) in (A/m)n(A/m)^n then form a system of generators of the free (A/m)(A/m)-module (A/m)n(A/m)^n; as A/mA/m is a field, this system is a basis of (A/m)n(A/m)^n; as AnA^n is a free A-module, we deduce from Proposition 5 that (xi)(x_i) is a basis of AnA^n.

Proposition 11

Let M be an A-module, N a finitely generated A-module and u:M→Nu : M \to N a homomorphism. For u to be surjective, it is necessary and sufficient that, for all m∈S2m \in S_2, the homomorphism M/mM→N/mNM/mM \to N/mN derived from uu by taking quotients be surjective.

Cites 2 · Cited by 2

Cites Proposition 4, Theorem 1

Cited by Proposition 2, Proposition 3

It follows from Theorem 1 that, for uu to be surjective, it is necessary and sufficient that um:Mm→Nmu_m : M_m \to N_m be surjective for all m∈Ωm \in \Omega. As AA is a local ring and NmN_m is a finitely generated AmA_m-module, this amounts to saying that the homomorphism um′:Mm/mMm→Nm/mNmu'_m : M_m/mM_m \to N_m/mN_m, obtained by taking quotients, is surjective (no. 2, Corollary 1 to Proposition 4); but Mm/mMmM_m/mM_m (resp. Nm/mNmN_m/mN_m) is identified with M/mMM/mM (resp. N/mNN/mN) (Proposition9), whence the proposition.

Proposition 12

Let E, F, G be three A-modules and v:G→Fv : G \to F, u:E→Fu : E \to F homomorphisms. Suppose that E is finitely presented. For there to exist a homomorphism w:E→Gw : E \to G such that uu factors into E→wG→vFE \xrightarrow{w} G \xrightarrow{v} F, it is necessary and sufficient that, for all m∈S2m \in S_2, there exist a homomorphism wm:Em→Gw^m : E_m \to G, such that um:Em→Fu_m : E_m \to F, factors into E→wmGm→vmFE \xrightarrow{w^m} G_m \xrightarrow{v_m} F,

Cites 2 · Cited by 3

Cites Proposition 19, II § 2, Corollary 1

Cited by Corollary 1, Corollary 2, Proposition 15, III § 2

The existence of ww satisfying the above statement is equivalent to the following property: uu belongs to the image P of the mapping

r=Hom⁡(1E,v):Hom⁡A(E,G)→Hom⁡A(E,F). r = \operatorname{Hom}(1_E, v) : \operatorname{Hom}_A(E, G) \to \operatorname{Hom}_A(E, F).

Now, (Hom⁡A(E,F))m(\operatorname{Hom}_A(E, F))_m (resp. (Hom⁡A(E,G))m(\operatorname{Hom}_A(E, G))_m) is canonically identified with Hom⁡Am(Em,Fm)\operatorname{Hom}_{A_m}(E_m, F_m) (resp. Hom⁡Am(Em,Gm)\operatorname{Hom}_{A_m}(E_m, G_m)) (§ 2, no. 7, Proposition 19 (i)), the canonical image of uu in (Hom⁡A(E,F))m(\operatorname{Hom}_A(E, F))_m is identified with umu_m, rmr_m is identified with Hom⁡Am(1Em,v)\operatorname{Hom}_{A_m}(1_{E_m}, v) and PmP_m with the image of rmr_m. The proposition then follows from Corollary 1 to Theorem 1 applied to Hom⁡A(E,F)\operatorname{Hom}_A(E, F) and its submodule P.

Corollary 1

Let M be an A-module and N a submodule of M such that M/N is finitely presented. For N to be a direct factor of M, it is necessary and sufficient that, for all m∈S2m \in S_2, N, be a direct factor of M.

Cites 2 · Cited by 1

Cites Proposition 12, Proposition 14, A II § 1

Cited by Corollary 2

To say that N is a direct factor of M means that the identity homomorphism on M/N factors into M/N→wM→ϕM/NM/N \xrightarrow{w} M \xrightarrow{\phi} M/N where ϕ\phi is the canonical homomorphism and ww a homomorphism (Algebra, Chapter II, § 1, no. 9, Proposition 14); as (M/N)m=Mm/Nm(M/N)_m = M_m/N_m and ϕm\phi_m is the canonical homomorphism Mm→Mm/NmM_m \to M_m/N_m, the corollary follows easily from Proposition 12.

Corollary 2

Let M be a finitely generated free A-module and N a submodule of M which is a finitely generated free A-module. For N to be a direct factor of M, it is necessary and sufficient that, for all m∈Ωm \in \Omega, mN=N∩(mM)mN = N \cap (mM).

Cites 5

Cites Proposition 6, Proposition 9, Lemma 8, I § 2, Proposition 12, Corollary 1

By definition M/N is finitely presented; on the other hand, N, and M, are finitely generated free AnA_n-modules. For NN to be a direct factor of MM, it is necessary and sufficient that the canonical mapping Nm/mNm→Mm/mMmN_m / mN_m \to M_m / mM_m be injective (no. 2, Proposition 6); this is the same as saying that the canonical mapping N/mN→M/mMN / mN \to M / mM must be injective (Proposition 9), and as its kernel is (N∩mM)/mN(N \cap mM) / mN, this proves the corollary.

Proposition 12 (resp. its Corollary 1) will be applied in particular when AA is Noetherian and EE (resp. M/NM/N) a finitely generated AA-module (Chapter I, § 2, no. 8, Lemma 8).

4. Localization of flatness

Proposition 13

Let SS be a multiplicative subset of a ring AA and MM an AA-module. If MM is jflat (resp. faithfully flat), S−1MS^{-1}M is ajflat (resp. faithfully flat) S−1AS^{-1}A-module and a flat AA-module.

Cites 4 · Cited by 4

Cites Corollary 2, I § 2, Corollary 3, I § 2, Proposition 5, I § 3, Theorem 1

Cited by Corollary, Proposition 15, Proposition 7, Corollary 2, II § 5

As S−1M=M⊗AS−1AS^{-1}M = M \otimes_A S^{-1}A, the first assertion follows from Chapter I, § 2, no. 7, Corollary 2 to Proposition 8 (resp. Chapter I, § 3, no. 3, Proposition 5); moreover, S−1AS^{-1}A is a flat AA-module (§2§ 2, no. 4, Theorem 1); hence if MM is a flat AA-module, so is S−1MS^{-1}M by virtue of Chapter I, § 2, no. 7, Corollary 3 to Proposition 8.

Remark

If NN is an S−1AS^{-1}A-module, S−1NS^{-1}N is identified with NN and this is consequently equivalent to saying that NN is a flat S−1AS^{-1}A-module or a flat AA-module.

Proposition 14

Let AA be a ring, BB a commutative AA-algebra and TT a multiplicative subset of BB. If NN is a BB-module which is flat as an AA-module, T−1NT^{-1}N is a flat AA-module.

Cites 1 · Cited by 1

Cites Proposition 8, I § 2

Cited by Proposition 15

T−1N=T−1B⊗BNT^{-1}N = T^{-1}B \otimes_B N; the proposition then follows from Chapter I, § 2, no. 7, Proposition 8, applied with AA replaced by BB, BB by AA, EE by T−1BT^{-1}B and FF by NN.

Proposition 15

Let A,BA, B be two rings, ϕ:A→B\phi : A \to B a homomorphism and NN a BB-module. The following properties are equivalent: (a) NN is a flat AA-module. (b) For every maximal ideal nn of BB, NnN_n is a flat AA-module. (c) *For every maximal ideal nn of BB, if we write m=ϕ(n)m = \phi(n), NnN_n is a flat AmA_m-module.

Cites 3 · Cited by 1

Cites Proposition 13, Proposition 14, Theorem 1

Cited by Corollary

For all a∉ma \notin m, the homothety of NnN_n induced by aa is bijective, hence NnN_n is canonically identified with (Nn)m(N_n)_m and the equivalence of (b) and (c) follows from the

Remark following Proposition 13; the fact that (a) implies (b) is a special case of Proposition 14. It remains to prove that (b) implies (a), that is, that, if (b) holds, for every injective A-module homomorphism u:M→M′u : M \to M', the homomorphism v=1@u:N⊗AM→N@∗M′v = 1 @ u : N \otimes_A M \to N @^* M' is injective. Now, vv is also a B-module homomorphism and, for it to be injective, it is necessary and sufficient that vn:(N⊗AM)n→(N⊗AM′)nv_n : (N \otimes_A M)_n \to (N \otimes_A M')_n be so for every maximal ideal nn of BB (no. 3, Theorem 1). As

(N⊗AM)n=Bn⊗B(N@AM)=Nn⊗AM, (N \otimes_A M)_n = B_n \otimes_B (N @_A M) = N_n \otimes_A M,
vnv_n is just the homomorphism 1⊗u:Nn⊗AM→Nn⊗AM′1 \otimes u : N_n \otimes_A M \to N_n \otimes_A M', which is injective since NnN_n is a flat A-module by hypothesis.

Corollary

For an A-module M to be flat (resp. faithfully flat), it is necessary and sufficient that, for every maximal ideal m of A, M, be a flat (resp. faithfully flat) A_m-module.

Cites 4 · Cited by 2

Cites Proposition 1, I § 3, Proposition 13, Proposition 15, Proposition 9

Cited by Exercise 9, § 3, Exercise 19, § 2

The necessity of the conditions follows from Proposition 13. Conversely, if M,M, is a flat AmA_m-module for every maximal ideal mm of AA, MM is a flat A-module by virtue of Proposition 15 applied to the case where ϕ\phi is the identity. Finally, if M,M, is a faithfully flat AmA_m-module for all mm, then mMm=mAmMm≠M,mM_m = mA_mM_m \neq M, hence mM≠MmM \neq M for all mm (no. 3, Proposition 9), which proves that MM is a faithfully flat A-module (Chapter I, § 3, no. 1, Proposition 1 (d)).

5. Semi-local rings

Proposition 16

Let A be a ring. The following properties are equivalent: (a) the set of maximal ideals of A is finite; (b) the quotient of A by its Jacobson radical is the direct composition of a finite number of fields.

Cites 2 · Cited by 1

Cites Definition 2, A VIII § 6, Proposition 5, II § 1

Cited by Definition 4

Suppose that the quotient of A by its Jacobson radical R\mathfrak{R} is a direct composition of a finite number of fields. Then A/RA/\mathfrak{R} possesses only a finite number of ideals and a fortiori only a finite number of maximal ideals. As every maximal ideal contains R\mathfrak{R} (Algebra, Chapter VIII, § 6, no. 2, Definition 2), the maximal ideals of A are the inverse images of the maximal ideals of A/RA/\mathfrak{R} under the canonical homomorphism A→A/RA \to A/\mathfrak{R}; hence they are finite in number.

Conversely, suppose that A has only a finite number of distinct maximal ideals m1,…,mnm_1, \ldots, m_n. The A/miA/m_i are fields and it follows from § 1, no. 2, Proposition 5 that the canonical mapping A→∏i=1nA/miA \to \prod_{i=1}^n A/m_i is surjective; as its kernel ⋂i=1nmi\bigcap_{i=1}^n m_i is the Jacobson radical R\mathfrak{R} (Algebra, Chapter VIII, § 6, no. 2, Definition 2), A/RA/\mathfrak{R} is isomorphic to ∏i=1nA/mi\prod_{i=1}^n A/m_i.

Definition 4

A ring is called semi-local if it satisfies the equivalent conditions (a), (b) of Proposition 16.

Cites 1

Cites Proposition 16

Example

Every local ring is semi-local. Every quotient of a semi-local ring is semi-local. Every finite product of semi-local rings is semi-local.

If A is a Noetherian semi-local ring and B is an A-algebra which is a finitely generated A-module, then B is semi-local (Chapter IV, § 2, no. 5, Corollary 3 to Proposition 9).

Another example, generalizing the construction of the local rings ApA_{\mathfrak{p}}, is provided by the following proposition:

Cites 1

Cites Corollary 3, IV § 2

Proposition 17

Let A be a ring and p1,…,pn\mathfrak{p}_1, \ldots, \mathfrak{p}_n prime ideals of A. We write

S=⋂1n(A−pi)=A−⋃1npi. S = \bigcap_1^n (A - \mathfrak{p}_i) = A - \bigcup_1^n \mathfrak{p}_i.

Cites 4 · Cited by 2

Cites Proposition 2, II § 1, Proposition 11, II § 2, Theorem 1, III § 2. General results on filtered rings and modules

Cited by Corollary, Remark, II § 5

(a) The ring S−1AS^{-1}A is semi-local; if q1,…,qrq_1, \ldots, q_r are the distinct maximal elements (with respect to inclusion) of the set of pi\mathfrak{p}_i, the maximal ideals of S−1AS^{-1}A are the S−1qjS^{-1}q_j (1≤j≤r1 \leq j \leq r) and these ideals are distinct.

(b) The ring ApiA_{\mathfrak{p}_i} is canonically isomorphic to (S−1A)S−1pi(S^{-1}A)_{S^{-1}\mathfrak{p}_i} for 1≤i≤n1 \leq i \leq n.

(c) If A is an integral domain, then S−1A=⋂1nApiS^{-1}A = \bigcap_1^n A_{\mathfrak{p}_i}, in the field of fractions of A.

(a) The ideals of A not meeting S are the ideals contained in the union of the pi\mathfrak{p}_i and hence in at least one of the pi\mathfrak{p}_i (§ 1, no. 1, Proposition 2); the qjq_j are therefore the maximal elements of the set of ideals not meeting S; consequently, the S−1qjS^{-1}q_j are the maximal ideals of S−1AS^{-1}A by § 2, no. 5, Proposition 11 (ii).

(b) is a special case of § 2, no. 5, Proposition 11 (iii).

(c) Suppose that A is an integral domain. If pi⊂pk\mathfrak{p}_i \subset \mathfrak{p}_k, then Api⊃ApkA_{\mathfrak{p}_i} \supset A_{\mathfrak{p}_k}; to prove (c), we may therefore suppose that no two pi\mathfrak{p}_i are comparable. Then it follows from (a) and no. 3, Corollary 4 to Theorem 1 that S−1A=⋂1n(S−1A)S−1piS^{-1}A = \bigcap_1^n (S^{-1}A)_{S^{-1}\mathfrak{p}_i}; whence (c) by virtue of (b).

If A is an integral domain, so is S−1AS^{-1}A and Proposition 17 then provides an example of a semi-local ring which is not a direct composition of local rings (cf. Chapter III, § 2, no. 13).

Corollary

Let A be an integral domain and p1,…,pn\mathfrak{p}_1, \ldots, \mathfrak{p}_n prime ideals of A, no two of which are comparable with respect to inclusion. If A=⋂1nApiA = \bigcap_1^n A_{\mathfrak{p}_i} in the field of fractions of A, the maximal ideals of A are p1,…,pn\mathfrak{p}_1, \ldots, \mathfrak{p}_n.

Cites 1 · Cited by 1

Cites Proposition 17

Cited by Proposition 2, VI § 7

Setting S=⋂i=1n(A−pi)S = \bigcap_{i=1}^n (\mathbf{A} - p_i), S−1A=AS^{-1}\mathbf{A} = \mathbf{A} by Proposition 17 (c); hence the elements of SS are invertible in A\mathbf{A} and S−1pi=piS^{-1}p_i = p_i for all ii. Our assertion then follows by virtue of Proposition 17 (a).

Exercises

See the exercises for § 3.

Transcribed from the printed volume. Not translated.