3. Finitely generated algebras over a field
1. The normalization lemma
In this no. and the following, k denotes a commutative field.
Theorem 1 (Normalization Lemma)
Let A be a finitely generated k-algebra and let be an increasing finite sequence of ideals of A such that and . There exists a finite sequence of elements of A which are algebraically independent over k (Chapter III, § 1, no. 1) and such that: (a) A is integral over the ring . (b) For all j such that , there exists an increasing sequence of integers such that for all j the ideal of B is generated .
Note first that it is sufficient to prove the theorem when A is a polynomial algebra . For in the general case, A is isomorphic to a quotient of such an algebra A' by an ideal ; let denote the inverse image of in A' and let () be elements of A' satisfying the conditions of the statement for the ring A' and the increasing sequence of ideals . Then the images of the in A for satisfy the desired conditions; this is obvious for condition (b) and for condition (a) this follows from § 1, no. 1, Proposition 2; finally, if the were not algebraically independent over k, there would be a non-zero polynomial
where the are polynomials in ; as is invertible in , (4) is certainly an equation of integral dependence with coefficients in B, whence our assertion.
The field of fractions of A is therefore algebraic over the field of fractions of B, which proves (Algebra, Chapter V, § 5, no. 3, Theorem 4) that the () are algebraically independent. Moreover, ; for every element may be written where ; but since is integrally closed ( 1, no. 3, Corollary 2 to Proposition 13); therefore , which completes the proof of properties (a) and (b) in this case.
(A2) General case (). We argue by induction on , the case being trivial. We may obviously suppose that (otherwise we may take , for and ). Let be a non-zero element of ; by (A1) there exist such that , are algebraically independent over , is integral over and . By the induction hypothesis there exist elements of and an integer such that is integral over , are algebraically independent over and the ideal is generated by . Then is integral over ( 1, no. 1, Corollary to Proposition 5 ) and hence so is ( 1, no. 1, Proposition 6 ); the same argument as in the case (A1) shows that are algebraically independent over ; finally, as , and and, as is generated (in ) by is generated (in ) by .
(B) Passage from top. Let , be elements of satisfying the conditions of the theorem for the increasing sequence of ideals and let us write . By (A2) there exist elements of and an integer s such that
Corollary 1
Let be an integral domain and a finitely generated -algebra containing as a subring. Then there exist an element of and a subalgebra of isomorphic to a polynomial algebra such that (Chapter II, § 2, no. 1) is integral over .
We write and let the field of fractions of ; clearly is a finitely generated -algebra and, as it contains by hypothesis (Chapter II, § 2, no. 4, Theorem 1), it is not reduced to 0. By Theorem 1 (applied top = 1 and ) there exists therefore a finite sequence of elements of which are algebraically independent over and such that is integral over . Let be a system of generators of the -algebra ; in each of the satisfies an equation of integral dependence
where the are polynomials in the with coefficients in . There exists an element of such that we may write whereby, for and all the coefficients of the are of the form where ; finally, replacing if need be, by a product of elements of , we may assume that in
where the are polynomials in with coefficients in ; if we write , it is seen, multiplying (6) by , that is integral over . We show that the are algebraically independent over ; if there is a relation of the form where for all , we deduce that in , where in ; by hypothesis therefore for all , whence for all . Moreover, in the ring each of the is integral over (§ 1, no. 1, Proposition 2) and, as in , it is seen that the are integral over , which completes the proof (§ 1, no. 1, Proposition 4).
Corollary 2
Let be a field, a subring of and the field of fractions of . If is a finitely generated -algebra, is finite and there exists in such that .
It follows from Corollary 1 that there exist elements of and an element of such that are algebraically independent over (and therefore over L) and that K is integral over the subring . Then it follows from §2, no.1, Lemma 2 that is a field. But the only invertible elements of a polynomial ring over an integral domain C are the invertible elements of C; applying this remark to , it is seen that necessarily and that is a field equal to L by definition of the latter. As K is integral over L and is a finitely generated L-algebra, the degree [K:L] is finite (§ 1, no. 1, Proposition 4).
Corollary 3
Let A be an integral domain, B a finitely generated A-algebra and b an element of B such that for all in A and every integer . Let be the canonical homomorphism; there exists in A such that, for every homomorphism f from A to an algebraically closed field L such that , there exists a homomorphism g from B to L such that and .
The hypothesis on b implies that, if h is the canonical homomorphism of B to , the homomorphism of A to is injective. By Corollary 1 there therefore exist an element of A and a subring of such that is integral over and is isomorphic to a polynomial algebra . Let f be a homomorphism from A to an algebraically closed field L such that ; there exists a homomorphism from to L extending f and hence there exists a homomorphism from to L extending f. As in L, there exists a homomorphism from to L such that
for all and all (Chapter II, §2, no. 1, Proposition 1). Finally, as is integral over , there exists a homomorphism from to L extending (§2, no. 1, Corollary 4 to Theorem 1). If is the canonical homomorphism from B to , solves the problem for is invertible in and hence in L.
Note that, if B is assumed to be an integral domain and in Corollary 3, the hypothesis on b is equivalent to .
2. The integral closure of a finitely generated algebra over a field
Theorem 2
Let A be a finitely generated integral k-algebra, K its field of fractions and the integral closure of A in a field K' which is a finite algebraic extension of K. Then is a finitely generated A-module and a finitely generated k-algebra.
By Theorem 1 there exists a subalgebra C of A isomorphic to a polynomial algebra and such that A is integral over C; is obviously the integral closure of C in K' (§ 1, no. 1, Proposition 6); we may therefore confine our attention to the case where . Let be the quasi-Galois extension of (in an algebraic closure of ) generated by , which is a finite algebraic extension of (Algebra, Chapter V, § 6, no. 3, Corollary 1 to Proposition 9). It will suffice to prove that the integral closure of in is a finitely generated -module, for is a sub--module of and is a Noetherian ring (Chapter III, § 2, no. 10, Corollary 2 to Theorem 2). We may therefore confine our attention to the case where is a quasi-Galois extension of . Then we know (Algebra, Chapter V, § 10, no. 9, Proposition 14) that is a (finite) Galois extension of a (finite) radicial extension of . If is the integral closure of in , is the integral closure of in and it will suffice to prove that is a finitely generated -module and is a finitely generated -module. Now, if it has been proved that is a finitely generated -module, it is a Noetherian domain, integrally closed by definition; the fact that is a finitely generated -module will follow from § 1, no. 6, Corollary 1 to Proposition 18.
We see therefore that we may confine our attention to the case where and where is a finite radicial extension of . Then is generated by a finite family of elements and there exists a power of the characteristic exponent of such that . Let () be the coefficients of the numerators and denominators of the rational functions in , equal to (). Then is contained in the extension , where (we are in an algebraic closure of ) and is contained in the algebraic closure of in . Now, is algebraic over and hence is integral over (, no. 1, Proposition 5); as is integrally closed (, no. 3, Corollary 2 to Proposition 13), it is seen that this ring is the integral closure of in and hence also that of (, no. 1, Proposition 6), in other words . Now clearly is a finitely generated -module (, no. 1, Proposition 4) and, as is a finite extension of , is a finitely generated -module and hence is a finitely generated -module; since is Noetherian and , is a finitely generated -module.
3. The nullstellensatz
Proposition 1
Let be a finitely generated algebra over a field and the algebraic closure of .
(i) If , there exists a -homomorphism from to .
(ii) Let be two -homomorphisms from to . For and to have the same kernel, it is necessary and sufficient that there exist a -automorphism of such that .
(iii) Let be an ideal of . For to be maximal, it is necessary and sufficient that it be the kernel of a k-homomorphism from to .
(iv) For an element of to be such that for every k-homomorphism from to , it is necessary and sufficient that be nilpotent.
Assertion (i) follows from no. 1, Corollary 3 to Theorem 1 applied replacing by , by , by the unit element of and by the canonical injection of into .
If is a k-homomorphism from to , is a subring of containing ; as is an algebraic extension of , is a field (Algebra, Chapter V, § 3, no. 2, Proposition 3) and, if is the kernel off, , isomorphic to , is therefore a field, which proves that is maximal. Conversely, if is a maximal ideal of , it follows from (i) that there exists a k-homomorphism from to and hence a k-homomorphism of to whose kernel contains ; but as is maximal, ; this proves (iii).
We now prove (ii). If is a k-automorphism of such that , clearly and have the same kernel. Conversely, suppose that and have the same kernel; then there exists a k-isomorphism of the field onto the field such that ; but by Algebra, Chapter V, § 6, no. 3, Proposition 7, extends to a k-automorphism of and hence .
Finally, if is such that , for every k-homomorphism from to , and hence since is a field. Conversely, suppose that is not nilpotent; then is a finitely generated -algebra (and therefore a finitely generated k-algebra) not reduced to 0 (Chapter 11, § 2, no. 1, Remark 3) and hence there exists a k-homomorphism from to by (i). If is the canonical homomorphism, is a k-homomorphism from to and , whence .
Let be a field and an extension field of ; an element of is called a zero in of an ideal of the polynomial ring if
for all .
Lemma 1
Let be a finitely generated algebra over a field , a system of generators of this algebra and the ideal of algebraic relations between the with coefficients in (Algebra, Chapter IV, § 2, no. 1). For every extension field of , the mapping is a bijection of the set of k-homomorphisms from to onto the set of zeros of in .
There exists a unique k-algebra homomorphism of onto such that , for and by definition is the kernel of . The mapping f his a bijection of the set of k-homomorphisms from to onto the set of k-homomorphisms from to which are zero on c. For every polynomial and every element we write ; then the mapping is a bijection of onto the set of k-homomorphisms from to (such a homomorphism being determined by its values at the () ); to say that is zero on c means that is a zero of c in , whence the lemma.
If Proposition 1 is applied to the algebra , where c is an ideal of distinct from the whole ring, we obtain by Lemma 1 the following statement:
Proposition 2 (Hilbert’s Nullstellensatz)
Let k be a field and L an algebraic closure of k.
(i) Every ideal not containing 1 admits at least one zero in .
(ii) Let be two elements of ; for the set of polynomials of zero at x to be identical with the set of polynomials of zero at y, it is necessary and sufficient that there exists a k-automorphism s of L such that for .
(iii) For an ideal a of to be maximal, it is necessary and sufficient that there exist an x in such that a is the set of polynomials of zero at x.
(iv) For a polynomial Q of to be zero on the set of zeros in of an ideal c of , it is necessary and sufficient that there exist an integer m > 0 such that .
4. Jacobson rings
Definition 1
A ring A is called a Jacobson ring if every prime ideal of A is the intersection of a family of maximal ideals.
Examples
(1) Every field is a Jacobson ring. (2) The ring is a Jacobson ring, the unique prime ideal which is not maximal (0) being the intersection of the maximal ideals of , where runs through the set of prime numbers (cf. Proposition 4). (3) Let A be a Jacobson ring and let a be an ideal of A. Then is a Jacobson ring, for the ideals of are of the form , where b is an ideal of A containing a and is prime (resp. maximal) if and only if b is.
Proposition 3
For a ring A to be a Jacobson ring, it is necessary and sufficient that, for every ideal a of A, the Jacobson radical of be equal to its nilradical (Chapter II, § 2, no. 6).
The Jacobson radical (resp. nilradical) of is the intersection of the maximal (resp. prime) ideals of (Algebra, Chapter VIII, § 6, no. 3, Definition 3 and Commutative Algebra, Chapter II, § 2, no. 6, Proposition 13). The stated condition means therefore that for every ideal of the intersection of the prime ideals containing is equal to the intersection of the maximal ideals containing . This condition obviously holds for every ideal of if is a Jacobson ring; conversely, if it holds for every prime ideal of , is a Jacobson ring by definition.
Corollary
Let be a Jacobson ring; for every ideal of , the radical of is the intersection of the maximal ideals of containing .
It is sufficient to note that is a Jacobson ring.
Proposition 4
Let be a principal ideal domain and a representative system of extremal elements of (Algebra, Chapter VII, § 1, no. 3, Definition 2). For to be a Jacobson ring, it is necessary and sufficient that be infinite.
The maximal ideals of are the (loc. cit., no. 2, Proposition 2). If is finite, their intersection is the ideal , where (ibid.) and hence different from ; on the other hand, if is infinite, the intersection of the is , every element of being divisible by only a finite number of extremal elements (loc. cit., no. 3, Theorem 2). The proposition then follows from the fact that is the only prime ideal which is not maximal in (Algebra, Chapter VI, § 1, no. 13, Proposition 14 (DIV)).
Proposition 5
Let be a ring and an -algebra integral over . If is a Jacobson ring, so is .
Replacing by its canonical image in , we may assume that . Let be a prime ideal of and let . There exists by hypothesis a family of maximal ideals of whose intersection is equal top. For all there exists a maximal ideal of lying above and containing (§ 2, no. 1, Proposition 1 and Corollary 2 to Theorem 1). If we write , then and , whence (, no. 1, Corollary 1 to Proposition 1).
Theorem 3
Let be a Jacobson ring, a finitely generated -algebra and the canonical homomorphism. Then:* (i) is a Jacobson ring. (ii) *For every maximal ideal of , is a maximal ideal of and is a finite algebraic extension of .
Let be a prime ideal of and . Let be an element of .
As is a finitely generated integral (Alp)-algebra and the canonical homomorphism is injective, there exists an element of such that, for every homomorphism from to an algebraically closed field whose kernel does not contain , there exists a homomorphism from to whose kernel does not contain and for which (no. 1, Corollary 3 to Theorem 1). Since is a Jacobson ring, there exists a maximal ideal of containing and such that . We take to be an algebraic closure of and to be the canonical homomorphism ; let
be a homomorphism such that and . Then
hence is a subfield of (Algebra, Chapter V, § 3, no. 2, Proposition 3) and the kernel of is therefore a maximal ideal of not containing . Thus it is seen that the intersection of the maximal ideals of is reduced to 0, which proves that is a Jacobson ring. Moreover, if is maximal, is necessarily injective and hence is maximal; finally is then a finitely generated algebra over the field and hence is a finite extension of (no. 1, Corollary 2 to Theorem 1).
Corollary 1
Every finitely generated algebra over is a Jacobson ring; for a prime ideal to be maximal, it is necessary and sufficient that the ring be finite.
If the integral domain is finite, it is a field, as, for every in , the mapping of to itself is injective and hence bijective since is finite. Conversely, for every maximal ideal of , the inverse image of in is a maximal ideal () and is finite over the prime field by Theorem 3.
Corollary 2
Let be a family of polynomials in and let be a polynomial in such that, for every system of elements belonging to a finite field and for which for all , also . Then, if is the ideal of generated by the , there exists an integer such that . Moreover, for every reduced ring and every system of elements of such that for all , also .
The second assertion follows from the first since the ideal of consisting of the polynomials such that contains . To show the first assertion, it suffices to note that, for every maximal ideal of containing , is a finite field (Corollary 1) and the hypothesis implies that the canonical image of in is zero; then belongs to the intersection of the maximal ideals of containing , which is the radical of (Corollary to Proposition 3).
Corollary 3
Let be a Jacobson ring. If there exists a finitely generated -algebra containing and which is a field, then is a field and is an algebraic extension of .
It suffices to apply Theorem 3 (ii) with .
Exercises
for every prime ideal of there exists a prime ideal of lying above . Argue by reductio ad absurdum: let and let , which is an ideal distinct from . Let be a prime ideal associated with ; the transporter is then distinct from (Chapter IV, § 1, no. 4, Proposition 9) and there therefore exists such that and
93
See the exercises for § 3.