§ 3. Depth and homological dimension
1. Projective dimension, injective dimension, homological dimension
Let A be a ring, M an A-module. Recall (A, X, p. 134, Def. 1) that the projective dimension of M, denoted by dp_A(M), is the greatest lower bound (in ) of the set of lengths of projective resolutions of M. One has dp_A(0) = -∞ and dp_A(M) ≥ 0 if M is nonzero. In order that M be projective, it is necessary and sufficient that one have dp_A(M) ≤ 0.
Example
Let J be an ideal of A generated by an A-regular sequence . One has dp_A(A/J) ≤ r. In fact, this is clear if A is zero; in the contrary case, the Koszul complex is a free resolution of A/J of length r (loc. cit., p. 159, Remark 3). Moreover, for every A-module N, the A-modules Ext_A^r(A/J, N) and N/JN are isomorphic (loc. cit.); consequently, in order that one have dp_A(A/J) = r, it is necessary and sufficient that J be distinct from A (loc. cit., p. 134, Prop. 1).
One defines analogously the injective dimension of M, denoted by di_A(M), as the greatest lower bound of the set of lengths of injective resolutions of M. One has di_A(0) = -∞, and di_A(M) ≥ 0 if M ≠ 0. In order that M be injective, it is necessary and sufficient that one have di_A(M) ≤ 0.
Proposition 1
Let A be a ring, M an A-module and n an integer ≥ 0. The following conditions are equivalent:
(i) one has di_A(M) ≤ n (in other words, M possesses an injective resolution of length ≤ n) ; (ii) for every A-module N and every integer i > n, one has Ext_A^i(N, M) = 0 ; (iii) for every ideal a of A, one has Ext_A^{n+1}(A/a, M) = 0 ; (iv) for every exact sequence of A-modules
where the are injective, the A-module Q is injective.
(i) ⇒ (ii) : this follows from A, X, p. 100, Th. 1. (ii) ⇒ (iii) : this is clear. (iii) ⇒ (iv) : in the situation of (iv), for every A-module N one has an isomorphism of Ext_A^1(N, Q) onto Ext_A^{n+1}(N, M) (A, X, p. 128, Cor. 4) ; under assumption (iii), the A-module Ext_A^1(A/a, Q) is zero for every ideal a of A, and Q is injective (A, X, p. 93, Prop. 11). (iv) ⇒ (i) : consider the exact sequence (A, X, p. 52)
if condition (iv) is satisfied, the A-module is injective, whence (i).
Recall (A, X, p. 138, Def. 2) that the homological dimension of the ring , denoted by , is the least upper bound in of the set of integers for which there exist two -modules and such that is nonzero. It is therefore also the least upper bound of the set of projective dimensions (or injective dimensions) of all the -modules; when is noetherian, it is enough to restrict oneself to finitely generated -modules (A, X, p. 139, Cor.).
2. Localization of homological dimension
Proposition 2
Let be a ring, and -modules, an integer and a multiplicative subset of . One has a canonical isomorphism of -modules
If the ring is noetherian and the -module finitely generated, one has a canonical isomorphism of -modules
Since the -module is flat, this follows from A, X, p. 110, Prop. 9 and p. 111, Prop. 10.
Corollary
Let be a ring, and -modules, an integer.
a) The support of is contained in , and the same is true of the support of if is noetherian and finitely generated.
b) Suppose that is noetherian, and that the modules and are finitely generated; if the -module is of finite length, the same is true of and of .
If is a prime ideal of not belonging to , one of the modules or is zero, which implies a) in view of Prop. 2.
In order that a finitely generated module over a noetherian ring be of finite length, it is necessary and sufficient that its support consist of maximal ideals (IV, § 2, No. 5, Prop. 7). Under assumption b), the -modules and are finitely generated (A, X, p. 108, Cor.) ; the assertion b) therefore follows from a).
Proposition 3
Let be a noetherian ring, a finitely generated -module, and an -module.
a) One has
where ranges over the set of prime ideals (resp. maximal ideals) of .
b) The mapping of into is upper semi-continuous.
Let us prove a). Let be an integer . Suppose that one has . For every prime ideal of and every -module , the -module is isomorphic to (Prop. 2), hence is zero; one deduces the inequality
Let us prove b). Let p be a prime ideal of A and n = dp_{A_p}(M_p). Let us show that there exists a neighbourhood U of p in Spec(A) such that one has dp_{A_q}(M_q) \leq n for every q \in U. This is clear if n = +\infty ; if n = -\infty , this follows from the fact that the support of M is closed. Suppose now that n is finite and choose an exact sequence of A-modules
Corollary 1
For every multiplicative subset S of A , one has
Corollary 2
If one has dp_{A_m}(M_m) < +\infty for every maximal ideal m of Supp(M) , one has dp_A(M) < +\infty .
In fact, the subspace X of Supp(M) formed by the maximal ideals is quasi-compact (II, § 4, No. 2, Prop. 8 and 9 and No. 3, Cor. 7 of Prop. 11) ; the mapping m \mapsto dp_{A_m}(M_m) of X into \overline{\mathbf{R}} is upper semi-continuous (Prop. 3), hence bounded (TG, IV, p. 30, Cor. of Th. 3).
Remark
Let A be a regular noetherian ring of infinite dimension (VIII, § 5, Exerc. 6 c)). We shall see below (No. 7, Th. 2 and § 4, No. 1, Prop. 1) that one has di_{A_m}(A_m) = dh(A_m) < +\infty for every maximal ideal m of A ; Prop. 3 therefore implies di_A(A) = dh(A) = +\infty . Consequently, the mappings p \mapsto di_{A_p}(N_p) and p \mapsto dh(A_p) are not in general upper semi-continuous.
3. Homological dimension of noetherian rings
Let A be a noetherian local ring, M a finitely generated A-module. Recall (A, X, § 3, No. 6) that a resolution
(1)
(A, X, p. 103, Example 3). Every finitely generated A-module admits such a resolution (A, X, p. 56, Prop. 10).
Proposition 4
Let A be a noetherian local ring, M a finitely generated A-module, and n an integer . The following conditions are equivalent* :
(i) one has ; (ii) one has ; (iii) one has ; (iv) every minimal projective resolution of M has length .
Assertions (i) (ii) and (i) (iii) are immediate (A, X, p. 100, Th. 1). Let L be a minimal projective resolution of M ; if (ii) or (iii) is satisfied, one has by (1). Since every minimal projective resolution of M is isomorphic to L (A, X, p. 54, Prop. 8), one deduces (iv). Implication (iv) (i) is trivial.
Corollary 1
Let A be a noetherian local ring and n an integer . The following conditions are equivalent* :
(i) one has ; (ii) one has and for every pair of A-modules and every integer ; (iii) one has ; (iv) one has ; (v) one has .
It is clear that (i) implies (ii) and that (ii) implies (iii) and (iv). By Prop. 4 applied to the A-module , each of conditions (iii) and (iv) implies (v). Let us prove that (v) implies (i) : if , one has for every A-module M ; hence every finitely generated A-module is of projective dimension (Prop. 4), which implies (A, X, p. 138, Prop. 4).
Corollary 2
One has* .
Remark 1
Let A be a local ring. The A-module is isomorphic to (A, X, p. 72, Example). Hence when A is noetherian, in order that be zero, it is necessary and sufficient that be zero, that is to say, that A be a field. Cor. 1 therefore implies that a noetherian local ring of homological dimension 0 is a field.
Remark 2
Let A be a noetherian local ring, M a finitely generated A-module of finite projective dimension , N a nonzero finitely generated A-module. The A-module is nonzero: let in fact L be a minimal projective resolution of M, and d its differential. We have an exact sequence
Since is zero, one deduces from this, by tensor product with , an isomorphism , whence, taking account of formula (1) above,
which is nonzero by Prop. 4 and Nakayama's lemma. Consequently, the projective dimension of M is the largest integer such that is nonzero.
Remark 3
Let be a noetherian ring, a finitely generated -module of finite projective dimension, a finitely generated -module whose support is equal to . According to the preceding remark and Props. 2 and 3 of No. 2, the projective dimension of is the largest integer such that is nonzero; the support of the -module is the set of elements of such that .
There may exist finitely generated -modules of projective dimension , satisfying for sufficiently large: this is the case for example of the -module when is a Gorenstein local ring which is not regular.
Proposition 5
Let be a noetherian ring, a finitely generated -module and an integer . The following conditions are equivalent :
(i) we have ; (ii) for every maximal ideal of , we have (resp. we have ) ; (iii) for every maximal ideal of , we have (resp. we have ).
(i) (ii) : this is clear. (ii) (iii) : this follows from Prop. 2 of No. 2. (iii) (i) : by Prop. 4, condition (iii) implies the inequality for every maximal ideal of , and one concludes by Prop. 3.
Remark 4
Let be a noetherian ring and an integer . If and are two distinct maximal ideals of , the -modules and are annihilated by , hence are zero. By a proof analogous to that of Corollary 1 of Proposition 4, one deduces from Proposition 5 the equivalence of the following conditions:
(i) one has ; (ii) one has and for every pair of -*modules and every integer* ; (iii) one has for every maximal ideal of ; (iv) one has for every maximal ideal of ; (v) one has for every maximal ideal of .
In particular, one has , where ranges over the set of maximal ideals of .
Proposition 6
Let be a noetherian ring, an -module, an integer . The following conditions are equivalent:
(i) one has ; (ii) for every prime ideal of , one has ; (iii) for every prime ideal of , one has .
If moreover the -module is of finite type, these conditions are equivalent to:
(iv) for every maximal ideal of , one has for .
Observe that condition (iii) is equivalent to
(iii') for every prime ideal of , one has .
In fact, since is annihilated by , the -module is isomorphic to , hence to (No. 2, Proposition 2).
The implications (i) (ii) and (i) (iv) result from Proposition 1 of No. 1, and the implication (ii) (iii') is clear.
(iii) (i): for every -module , let us put . Suppose that (i) is not satisfied. Then there exists (No. 1, Proposition 1) an ideal of such that . Let be an ideal of maximal among the ideals having this property; let us prove that is prime. By IV, § 1, No. 4, Theorem 1 and Theorem 2, there exists a composition sequence of such that each quotient is isomorphic to a module , where () is a prime ideal containing . If were zero for every , one would deduce by induction on from the exact sequences
that is zero, which is not so. There therefore exists an index such that . In view of the maximal character of , one has , so that is prime.
Let be an element of ; from the exact sequence
one deduces an exact sequence
where (A, X, p. 89, Proposition 6). Because of the maximal character of , one has , so that the homomorphism is injective. The nonzero -module is therefore torsion-free. This implies that is nonzero (A, II, p. 117, Corollary 1), which contradicts (iii').
Suppose that the -module is of finite type, and let us prove that (iv) implies (iii). Let be a prime ideal of and a maximal ideal of containing . There exists a saturated chain of prime ideals of with end terms and . The length of this chain is less than ; under assumption (iv), one therefore has
Remark 5
Let be a finitely generated -module; the condition for every maximal ideal of does not necessarily imply . If, for example, the ring is local and is not a Gorenstein ring (No. 7, Def. 1), one has for but .
4. Quotient by a cancellable element
In this No., denotes a ring and an element of cancellable.
Let be an -module and a projective resolution of . By A, X, p. 101, cor. to th. 1, is identified with , and therefore is isomorphic to if , to if , and is zero otherwise (loc. cit., p. 102, Example 1). Consider the complexes of -modules and such that
Similarly, let be an injective resolution of ; let denote the complex . By A, X, p. 101, cor. to th. 1, is identified with , and therefore is isomorphic to if , to if , and is zero otherwise (loc. cit., p. 102, Example 1). From one deduces complexes of -modules and such that
Let be an -module, and an injective resolution of . From exact sequence (2) one deduces an exact sequence of complexes of -modules
Consider the exact homology sequence associated with this exact sequence. By A, X, p. 100, th. 1, for every integer one has isomorphisms
whence a long exact sequence of -modules
(4)
Analogously, let be a projective resolution of the -module . From the exact sequence (3) one deduces an exact sequence of complexes of -modules
.
In view of the isomorphism , the associated homology exact sequence may be written
(5)
Finally, consider the exact sequence of complexes of -modules
deduced from the exact sequence (2); in view of the isomorphism , the associated homology exact sequence may be written
(6)
Proposition 7
Let be a ring, a cancellable element of , an -module such that the homothety is injective, an -module annihilated by , an integer. The canonical homomorphisms of -modules
deduced from the exact sequences (4), (5) and (6) are isomorphisms.
Corollary
Let be a noetherian local ring, a finitely generated -module and an ideal of generated by a sequence of elements of which is both -regular and -regular. Then and .
It is enough to treat the case . Put . The -modules and are isomorphic for every integer (prop. 7); the equality follows from prop. 4 of No. 3. Analogously, the -modules and are isomorphic for every integer , and the equality follows from prop. 6 of No. 3.
5. Depth and projective dimension
Theorem 1 (Auslander-Buchsbaum)
Let be a noetherian local ring and a finitely generated -module of finite projective dimension. Then one has the equality
Let us reason by induction on .
a) If is zero, is a nonzero finitely generated free -module; its depth is equal to (§ 1, No. 1, remark 4).
b) Suppose and choose a minimal projective resolution
of (No. 3, prop. 4, (iv)). The -modules and are nonzero finitely generated free modules, hence of depth (§ 1, No. 1, remark 4). The mapping is zero, which implies that belongs to . By remark 5 of § 1, No. 1, one has .
c) Suppose . Choose an exact sequence
where is a finitely generated free -module. One then has (§ 1, No. 1, remark 4), (A, X, p. 135, cor. 2 c)), whence (induction hypothesis), and in particular . By prop. 1 of § 1, No. 1, one then has , which completes the proof.
Remark
In view of cor. 2 of prop. 4 (No. 3), theorem 1 applied to the -module implies that one is in one or the other of the following cases:
(i) one has ; (ii) one has .
We shall see later (§ 4, No. 2) that (ii) characterizes regular local rings.
Corollary 1
Retain the hypotheses of theorem 1.
a) One has . For there to be equality, it is necessary and sufficient that the maximal ideal be associated to .
b) One has . For there to be equality, it is necessary and sufficient that be free.
a) In fact, “ ” is equivalent to “ ” (, No. 1, remark 2).
b) In fact, “ ” is equivalent to “ is free ”.
COROLLARY 2.: Retain the hypotheses of theorem 1 and suppose moreover that is a Macaulay ring. Then is the sum of the two positive integers and .
In particular, is then greater than , and there is equality if and only if is Cohen-Macaulay.
Corollary 3
Let be a noetherian ring, a finitely generated -module of finite projective dimension, an integer , a finitely generated -module, and the support of the -module (resp. ). Then .
In fact, let . We have (resp. ) by prop. 2 of No. 2, hence (No. 2, prop. 3). Theorem 1 implies . Consequently (, No. 5, prop. 8)
With the terminology of , No. 5, remark 4, the conclusion of cor. 3 means that the modules and are of grade . It implies that the codimension of their support in is (, No. 7, prop. 12).
Corollary 4
Let be a noetherian Macaulay ring and a finitely generated -module of finite projective dimension.
a) Let ; let us denote by the set of irreducible components of containing . We have
b) The mapping of into is upper semi-continuous.
c) The set of prime ideals of such that the -module is Macaulay is open and dense in . Its intersection with is dense in .
a) We may suppose that . Put . By cor. 2 above, we have
Since is a Macaulay ring, we have
and consequently
b) Let , and let be the union of the irreducible components of which do not contain . For every element of , we have , whence by the formula above. Consequently the function is lower semi-continuous; assertion b) then follows from prop. 3 of No. 2.
c) Let be the set of elements of such that is Cohen-Macaulay. The condition is equivalent to , so that is open by b). Since contains , it is enough to prove that is dense in . For every minimal prime ideal of , the -module is of finite length (IV, § 2, No. 5, Corollary 2 of Proposition 7 and § 1, No. 3, Corollary 1 of Proposition 7), hence Cohen-Macaulay; consequently meets all the irreducible components of . One concludes by means of Proposition 1 of II, § 4, No. 1.
6. Depth and injective dimension
Proposition 8
Let be a noetherian ring, a finitely generated -module. Then .
Let be a positive integer less than . There exists a saturated chain of prime ideals such that is a minimal element of the support of ; then the -module is of finite length, so that , hence (§ 1, No. 7, Lemma 3), which implies (No. 3, Proposition 6).
Proposition 9
Let be a noetherian local ring and a nonzero finitely generated -module of finite injective dimension. Then .
Put . We have for , hence by Proposition 6 of No. 3, (iv) (i). Let be the depth of and let be an -regular sequence of elements of (§ 1, No. 4, Theorem 2); put . By the example of No. 1, we have and , hence . But is of depth 0 (§ 1, No. 4, Proposition 7), hence there exists an exact sequence of -modules
From this one deduces an exact sequence of extension modules
since and , one obtains , whence . In the end, we have , which completes the proof.
7. Gorenstein rings
Definition 1
A ring is said to be a Gorenstein ring if it is noetherian and if the -module is of finite injective dimension for every maximal ideal of .
For a noetherian local ring to be a Gorenstein ring, it is necessary and sufficient that be finite; for a noetherian ring to be a Gorenstein ring, it is necessary and sufficient that this be so for for every maximal ideal of .
Proposition 10
Let be a Gorenstein ring; then is a Macaulay ring, and satisfies .
For every maximal ideal of , one has
Thus the noetherian rings such that is finite are the finite-dimensional Gorenstein rings (Prop. 3 of No. 2), and the noetherian rings such that the -module is injective are the artinian Gorenstein rings.
Example 1
For every multiplicative subset of a Gorenstein ring , the ring of fractions is a Gorenstein ring: in fact, let be a maximal ideal of ; it is of the form , where is a prime ideal of not meeting . Let be a maximal ideal of containing ; then the ring is isomorphic to (II, § 2, No. 5, Prop. 11), hence to a ring of fractions of , and consequently satisfies (No. 2, Cor. 1 of Prop. 3).
Example 2
Let be a Gorenstein ring and let be an ideal of , generated by an -regular sequence . The quotient ring is a Gorenstein ring: in fact, for every maximal ideal of containing , the image in of the sequence is -regular and generates the ideal , so that is a Gorenstein ring by Cor. of Prop. 7 (No. 4).
Example 3
Let be a noetherian local ring, an ideal of generated by an -regular sequence of elements of . If is a Gorenstein ring, then the same is true of (No. 4, Cor. of Prop. 7).
Example 4
Let be a regular local noetherian ring; then is a Gorenstein ring. In fact, let be a system of coordinates of (VIII, § 5, No. 1, Def. 1). The sequence is -regular (loc. cit., No. 2, Theorem 1) and generates the ideal ; one can therefore apply Example 3.
Example 5
Every quotient ring of a principal ideal domain is a Gorenstein ring (Example 2). In particular, every monogenic algebra over a field is a Gorenstein ring.
Lemma 1
Let be an artinian local ring. The following conditions are equivalent:
(i) is a Gorenstein ring; (ii) the -module is injective; (iii) the vector space over is of dimension 1.
Let us recall (A, VIII, p. 3.5) that has a minimal nonzero ideal; such an ideal is a simple module, hence isomorphic to . The vector space over is therefore nonzero; saying that it is of dimension 1 means that contains only one minimal nonzero ideal, which is therefore the socle of (A, VIII, § 4, No. 6).
The equivalence of (i) and (ii) was proved after Prop. 10. Suppose that the -module is injective. Let and be two nonzero elements of annihilated by . There exists a unique -linear mapping such that ; since is injective, it extends to an endomorphism of , which implies that belongs to , whence (iii).
Conversely, suppose that is of dimension 1. Let be a finitely generated -module; it is of finite length, hence has a composition series whose quotients are isomorphic to . One deduces by induction on the length of the inequality . In the exact sequence of extension modules
one therefore has . Consequently is surjective, is zero and the -module is injective (No. 3, Prop. 4).
Lemma 2
Let be a local noetherian ring such that . Then for every integer .
Let us reason by induction on . When , is the unique prime ideal of and the assertion follows from Prop. 6 of No. 3. Suppose therefore that and let be a prime ideal of distinct from ; then . If , the induction hypothesis implies for every integer ; by Lemma 3 of § 1, No. 7, this implies for every integer , and in particular for (VIII, § 1, No. 3, Prop. 8, b)).
It remains for us to treat the case where is infinite but where the injective dimension of is finite for every prime ideal of distinct from . For such an ideal, in this case, by Prop. 10, , hence for . Since is infinite, Prop. 6 of No. 3 then implies for every integer .
Theorem 2 (Bass)
Let be a noetherian local ring; put . Let be a maximal secant sequence of elements of , and the ideal which it generates. The following conditions are equivalent:
(i) is a Gorenstein ring; (ii) one has ; (iii) there exists an integer such that ; (iv) one has for and the -vector space is of dimension 1; (v) the ring is Macaulay and the -vector space is of dimension 1; (vi) the sequence is -regular and the -vector space is of dimension 1.
The equivalence of (i), (ii) and (iii) follows from Lemma 2 and Prop. 10. If is zero for every integer , the ring is Macaulay (§ 2, No. 3, Prop. 3); if the ring is Macaulay, the sequence is -regular (loc. cit., Th. 1); if the sequence is -regular, one has for and the -vector spaces and are isomorphic (A, X, p. 166, Prop. 9). This proves the equivalence of conditions (iv), (v) and (vi).
Let us prove that (i) implies (v): if is a Gorenstein ring, it is a Macaulay ring (prop. 10). The sequence is then -regular (§ 2, No. 3, prop. 4), hence is an artinian Gorenstein ring (example 2), so that the -vector space is of dimension 1.
Lastly let us prove that (vi) implies (i): under assumption (vi), the ring is a Gorenstein ring (lemma 1), hence is a Gorenstein ring (example 3).
Corollary
Let be a noetherian local ring of dimension d. The -module is not zero.
This follows from th. 2 if is a Gorenstein ring and from lemma 2 otherwise.
Proposition 11
Let be a noetherian ring. The following conditions are equivalent:
(i) is a Gorenstein ring ; (ii) for every prime ideal of , the -vector space is zero for and of dimension 1 for . (iii) for every maximal ideal of , there exists an integer such that .
(i) (ii) : this follows from th. 2 applied to the local Gorenstein ring (example 1). (ii) (iii) : this is trivial. (iii) (i) : under assumption (iii), is a Gorenstein ring for every maximal ideal of (th. 2), and is Gorenstein.
8. Gorenstein Rings and Flat Algebras
Proposition 12
Let be a local homomorphism of noetherian local rings, making into a flat -module. The following conditions are equivalent:
(i) is a Gorenstein ring ; (ii) and are Gorenstein rings.
Let us first treat the case where the rings and are artinian. Let denote the local ring ; its residue field identifies with . Since is flat over , the -module is isomorphic to (I, § 2, No. 10, prop. 11), hence to . We deduce a sequence of isomorphisms
In particular, one has and the proposition then follows from lemma 1 of No. 7.
Let us pass to the general case. If in the statement one replaces the word "Gorenstein" by the word "Macaulay", the proposition is a particular case of Prop. 10 of § 2, No. 7. We may therefore suppose that the rings , and are Macaulay. The -module is Macaulay (§ 2, No. 1, Example 4). Put , . There exists an -regular sequence of elements of and a sequence of elements of which is regular for the -module (§ 2, No. 3, Prop. 3); let denote the ideal of and the ideal of which they generate respectively. The sequence is -regular (§ 1, No. 6, Prop. 11), and the -module is flat (loc. cit., Prop. 10). The homomorphism of into deduced from by passing to quotients therefore makes into a flat -module, and the ring is isomorphic to . The proposition thus follows from the first part of the proof, taking account of Example 3 of No. 7.
Corollary 1
Let be a homomorphism of noetherian rings making into a flat -module. The following conditions are equivalent:
(i) is a Gorenstein ring; (ii) (resp. (iii)) for every prime ideal (resp. maximal ideal) of , the rings and are Gorenstein.
If moreover is a faithfully flat -module, these conditions imply that is a Gorenstein ring.
Let be a prime ideal of ; put . The ring , isomorphic to a ring of fractions of , is flat over . In order that be a Gorenstein ring, it is necessary and sufficient that the rings and should be so (Prop. 12).
(i) (ii): let be a prime ideal of and . If is a Gorenstein ring, the same is true of , hence of and by what precedes. By the remark of § 2, No. 6, the local ring of at any prime ideal whatever is then a Gorenstein ring, which implies that is a Gorenstein ring, whence (ii).
(ii) (iii): this is clear.
(iii) (i): for every maximal ideal of , it follows from the beginning of the proof (applied with ) that is a Gorenstein ring, whence (i).
If is a faithfully flat -module, the mapping is surjective (II, § 2, No. 5, cor. 4 of prop. 11), whence the last assertion.
Corollary 2
Let be a noetherian ring, an ideal of , the separated completion of for the -adic topology. Consider the following conditions:
(i) is a Gorenstein ring; (ii) is a Gorenstein ring; (iii) for every maximal ideal of containing , is a Gorenstein ring; (iv) for every prime ideal of such that , and are Gorenstein rings.
Conditions (ii) to (iv) are equivalent, and are implied by (i). When the ideal is contained in the radical of , conditions (i) to (iv) are equivalent.
This corollary is deduced from cor. 1 in the same way as corollary 4 of prop. 9 of § 2, No. 7, was deduced: it is enough in the proof to replace the word "Macaulay" by the word "Gorenstein".
Corollary 3
Let be a Gorenstein ring and a finite family of indeterminates. The rings and are Gorenstein rings.
For every field , the ring is Gorenstein (No. 7, example 5); moreover the ring is noetherian. It is therefore a Gorenstein ring (cor. 1). It follows by induction on that is a Gorenstein ring; then cor. 2 is applied.
Exercises
See the exercises for § 3.