10.36 Finite and integral ring extensions
Trivial lemmas concerning finite and integral ring maps. We recall the definition.
Definition 10.36.1. Let \varphi : R \to S be a ring map.
An element s \in S is integral over R if there exists a monic polynomial P(x) \in R[x] such that P^\varphi (s) = 0, where P^\varphi (x) \in S[x] is the image of P under \varphi : R[x] \to S[x].
The ring map \varphi is integral if every s \in S is integral over R.
Lemma 10.36.2. Let \varphi : R \to S be a ring map. Let y \in S. If there exists a finite R-submodule M of S such that 1 \in M and yM \subset M, then y is integral over R.
Proof.
Consider the map \varphi : M \to M, x \mapsto y \cdot x. By Lemma 10.16.2 there exists a monic polynomial P \in R[T] with P(\varphi ) = 0. In the ring S we get P(y) = P(y) \cdot 1 = P(\varphi )(1) = 0.
\square
Lemma 10.36.3. A finite ring map is integral.
Proof.
Let R \to S be finite. Let y \in S. Apply Lemma 10.36.2 to M = S to see that y is integral over R.
\square
Lemma 10.36.4. Let \varphi : R \to S be a ring map. Let s_1, \ldots , s_ n be a finite set of elements of S. In this case s_ i is integral over R for all i = 1, \ldots , n if and only if there exists an R-subalgebra S' \subset S finite over R containing all of the s_ i.
Proof.
If each s_ i is integral, then the subalgebra generated by \varphi (R) and the s_ i is finite over R. Namely, if s_ i satisfies a monic equation of degree d_ i over R, then this subalgebra is generated as an R-module by the elements s_1^{e_1} \ldots s_ n^{e_ n} with 0 \leq e_ i \leq d_ i - 1. Conversely, suppose given a finite R-subalgebra S' containing all the s_ i. Then all of the s_ i are integral by Lemma 10.36.3.
\square
Lemma 10.36.5. Let R \to S be a ring map. The following are equivalent
R \to S is finite,
R \to S is integral and of finite type, and
there exist x_1, \ldots , x_ n \in S which generate S as an algebra over R such that each x_ i is integral over R.
Proof.
Clear from Lemma 10.36.4.
\square
Lemma 10.36.6.slogan Suppose that R \to S and S \to T are integral ring maps. Then R \to T is integral.
Proof.
Let t \in T. Let P(x) \in S[x] be a monic polynomial such that P(t) = 0. Apply Lemma 10.36.4 to the finite set of coefficients of P. Hence t is integral over some subalgebra S' \subset S finite over R. Apply Lemma 10.36.4 again to find a subalgebra T' \subset T finite over S' and containing t. Lemma 10.7.3 applied to R \to S' \to T' shows that T' is finite over R. The integrality of t over R now follows from Lemma 10.36.3.
\square
Lemma 10.36.7. Let R \to S be a ring homomorphism. The set
S' = \{ s \in S \mid s\text{ is integral over }R\}
is an R-subalgebra of S.
Proof.
This is clear from Lemmas 10.36.4 and 10.36.3.
\square
Lemma 10.36.8. Let R_ i\to S_ i be ring maps i = 1, \ldots , n. Let R and S denote the product of the R_ i and S_ i respectively. Then an element s = (s_1, \ldots , s_ n) \in S is integral over R if and only if each s_ i is integral over R_ i.
Proof.
Omitted.
\square
Definition 10.36.9. Let R \to S be a ring map. The ring S' \subset S of elements integral over R, see Lemma 10.36.7, is called the integral closure of R in S. If R \subset S we say that R is integrally closed in S if R = S'.
In particular, we see that R \to S is integral if and only if the integral closure of R in S is all of S.
Lemma 10.36.10. Let R_ i\to S_ i be ring maps i = 1, \ldots , n. Denote the integral closure of R_ i in S_ i by S'_ i. Further let R and S denote the product of the R_ i and S_ i respectively. Then the integral closure of R in S is the product of the S'_ i. In particular R \to S is integrally closed if and only if each R_ i \to S_ i is integrally closed.
Proof.
This follows immediately from Lemma 10.36.8.
\square
Lemma 10.36.11. Integral closure commutes with localization: If A \to B is a ring map, and S \subset A is a multiplicative subset, then the integral closure of S^{-1}A in S^{-1}B is S^{-1}B', where B' \subset B is the integral closure of A in B.
Proof.
Since localization is exact we see that S^{-1}B' \subset S^{-1}B. Suppose x \in B' and f \in S. Then x^ d + \sum _{i = 1, \ldots , d} a_ i x^{d - i} = 0 in B for some a_ i \in A. Hence also
(x/f)^ d + \sum \nolimits _{i = 1, \ldots , d} a_ i/f^ i (x/f)^{d - i} = 0
in S^{-1}B. In this way we see that S^{-1}B' is contained in the integral closure of S^{-1}A in S^{-1}B. Conversely, suppose that x/f \in S^{-1}B is integral over S^{-1}A. Then we have
(x/f)^ d + \sum \nolimits _{i = 1, \ldots , d} (a_ i/f_ i) (x/f)^{d - i} = 0
in S^{-1}B for some a_ i \in A and f_ i \in S. This means that
(f'f_1 \ldots f_ d x)^ d + \sum \nolimits _{i = 1, \ldots , d} f^ i(f')^ if_1^ i \ldots f_ i^{i - 1} \ldots f_ d^ i a_ i (f'f_1 \ldots f_ dx)^{d - i} = 0
for a suitable f' \in S. Hence f'f_1\ldots f_ dx \in B' and thus x/f \in S^{-1}B' as desired.
\square
Lemma 10.36.12.slogan Let \varphi : R \to S be a ring map. Let x \in S. The following are equivalent:
x is integral over R, and
for every prime ideal \mathfrak p \subset R the element x \in S_{\mathfrak p} is integral over R_{\mathfrak p}.
Proof.
It is clear that (1) implies (2). Assume (2). Consider the R-algebra S' \subset S generated by \varphi (R) and x. Let \mathfrak p be a prime ideal of R. Then we know that x^ d + \sum _{i = 1, \ldots , d} \varphi (a_ i) x^{d - i} = 0 in S_{\mathfrak p} for some a_ i \in R_{\mathfrak p}. Hence we see, by looking at which denominators occur, that for some f \in R, f \not\in \mathfrak p we have a_ i \in R_ f and x^ d + \sum _{i = 1, \ldots , d} \varphi (a_ i) x^{d - i} = 0 in S_ f. This implies that S'_ f is finite over R_ f. Since \mathfrak p was arbitrary and \mathop{\mathrm{Spec}}(R) is quasi-compact (Lemma 10.17.8) we can find finitely many elements f_1, \ldots , f_ n \in R which generate the unit ideal of R such that S'_{f_ i} is finite over R_{f_ i}. Hence we conclude from Lemma 10.23.2 that S' is finite over R. Hence x is integral over R by Lemma 10.36.4.
\square
Lemma 10.36.13.slogan Let R \to S and R \to R' be ring maps. Set S' = R' \otimes _ R S.
If R \to S is integral so is R' \to S'.
If R \to S is finite so is R' \to S'.
Proof.
We prove (1). Let s_ i \in S be generators for S over R. Each of these satisfies a monic polynomial equation P_ i over R. Hence the elements 1 \otimes s_ i \in S' generate S' over R' and satisfy the corresponding polynomial P_ i' over R'. Since these elements generate S' over R' we see that S' is integral over R'. Proof of (2) omitted.
\square
Lemma 10.36.14. Let R \to S be a ring map. Let f_1, \ldots , f_ n \in R generate the unit ideal.
If each R_{f_ i} \to S_{f_ i} is integral, so is R \to S.
If each R_{f_ i} \to S_{f_ i} is finite, so is R \to S.
Proof.
Proof of (1). Let s \in S. Consider the ideal I \subset R[x] of polynomials P such that P(s) = 0. Let J \subset R denote the ideal (!) of leading coefficients of elements of I. By assumption and clearing denominators we see that f_ i^{n_ i} \in J for all i and certain n_ i \geq 0. Hence J contains 1 and we see s is integral over R. Proof of (2) omitted.
\square
Lemma 10.36.15. Let A \to B \to C be ring maps.
If A \to C is integral so is B \to C.
If A \to C is finite so is B \to C.
Proof.
Omitted.
\square
Lemma 10.36.16. Let A \to B \to C be ring maps. Let B' be the integral closure of A in B, let C' be the integral closure of B' in C. Then C' is the integral closure of A in C.
Proof.
Omitted.
\square
Lemma 10.36.17. Suppose that R \to S is an integral ring extension with R \subset S. Then \varphi : \mathop{\mathrm{Spec}}(S) \to \mathop{\mathrm{Spec}}(R) is surjective.
Proof.
Let \mathfrak p \subset R be a prime ideal. We have to show \mathfrak pS_{\mathfrak p} \not= S_{\mathfrak p}, see Lemma 10.18.6. The localization R_{\mathfrak p} \to S_{\mathfrak p} is injective (as localization is exact) and integral by Lemma 10.36.11 or 10.36.13. Hence we may replace R, S by R_{\mathfrak p}, S_{\mathfrak p} and we may assume R is local with maximal ideal \mathfrak m and it suffices to show that \mathfrak mS \not= S. Suppose 1 = \sum f_ i s_ i with f_ i \in \mathfrak m and s_ i \in S in order to get a contradiction. Let R \subset S' \subset S be such that R \to S' is finite and s_ i \in S', see Lemma 10.36.4. The equation 1 = \sum f_ i s_ i implies that the finite R-module S' satisfies S' = \mathfrak m S'. Hence by Nakayama's Lemma 10.20.1 we see S' = 0. Contradiction.
\square
Lemma 10.36.18. Let R be a ring. Let K be a field. If R \subset K and K is integral over R, then R is a field and K is an algebraic extension. If R \subset K and K is finite over R, then R is a field and K is a finite algebraic extension.
Proof.
Assume that R \subset K is integral. By Lemma 10.36.17 we see that \mathop{\mathrm{Spec}}(R) has 1 point. Since clearly R is a domain we see that R = R_{(0)} is a field (Lemma 10.25.1). The other assertions are immediate from this.
\square
Lemma 10.36.19. Let k be a field. Let S be a k-algebra over k.
If S is a domain and finite dimensional over k, then S is a field.
If S is integral over k and a domain, then S is a field.
If S is integral over k then every prime of S is a maximal ideal (see Lemma 10.26.5 for more consequences).
Proof.
The statement on primes follows from the statement “integral + domain \Rightarrow field”. Let S integral over k and assume S is a domain, Take s \in S. By Lemma 10.36.4 we may find a finite dimensional k-subalgebra k \subset S' \subset S containing s. Hence S is a field if we can prove the first statement. Assume S finite dimensional over k and a domain. Pick s\in S. Since S is a domain the multiplication map s : S \to S is surjective by dimension reasons. Hence there exists an element s_1 \in S such that ss_1 = 1. So S is a field.
\square
Lemma 10.36.20. Suppose R \to S is integral. Let \mathfrak q, \mathfrak q' \in \mathop{\mathrm{Spec}}(S) be distinct primes having the same image in \mathop{\mathrm{Spec}}(R). Then neither \mathfrak q \subset \mathfrak q' nor \mathfrak q' \subset \mathfrak q.
Proof.
Let \mathfrak p \subset R be the image. By Remark 10.18.5 the primes \mathfrak q, \mathfrak q' correspond to ideals in S \otimes _ R \kappa (\mathfrak p). Thus the lemma follows from Lemma 10.36.19.
\square
Lemma 10.36.21. Suppose R \to S is finite. Then the fibres of \mathop{\mathrm{Spec}}(S) \to \mathop{\mathrm{Spec}}(R) are finite.
Proof.
By the discussion in Remark 10.18.5 the fibres are the spectra of the rings S \otimes _ R \kappa (\mathfrak p). As R \to S is finite, these fibre rings are finite over \kappa (\mathfrak p) hence Noetherian by Lemma 10.31.1. By Lemma 10.36.20 every prime of S \otimes _ R \kappa (\mathfrak p) is a minimal prime. Hence by Lemma 10.31.6 there are at most finitely many.
\square
Lemma 10.36.22. Let R \to S be a ring map such that S is integral over R. Let \mathfrak p \subset \mathfrak p' \subset R be primes. Let \mathfrak q be a prime of S mapping to \mathfrak p. Then there exists a prime \mathfrak q' with \mathfrak q \subset \mathfrak q' mapping to \mathfrak p'.
Proof.
We may replace R by R/\mathfrak p and S by S/\mathfrak q. This reduces us to the situation of having an integral extension of domains R \subset S and a prime \mathfrak p' \subset R. By Lemma 10.36.17 we win.
\square
The property expressed in the lemma above is called the “going up property” for the ring map R \to S, see Definition 10.41.1.
Lemma 10.36.23. Let R \to S be a finite and finitely presented ring map. Let M be an S-module. Then M is finitely presented as an R-module if and only if M is finitely presented as an S-module.
Proof.
One of the implications follows from Lemma 10.6.4. To see the other assume that M is finitely presented as an S-module. Pick a presentation
S^{\oplus m} \longrightarrow S^{\oplus n} \longrightarrow M \longrightarrow 0
As S is finite as an R-module, the kernel of S^{\oplus n} \to M is a finite R-module. Thus from Lemma 10.5.3 we see that it suffices to prove that S is finitely presented as an R-module.
Pick y_1, \ldots , y_ n \in S such that y_1, \ldots , y_ n generate S as an R-module. By Lemma 10.36.2 each y_ i is integral over R. Choose monic polynomials P_ i(x) \in R[x] with P_ i(y_ i) = 0. Consider the ring
S' = R[x_1, \ldots , x_ n]/(P_1(x_1), \ldots , P_ n(x_ n))
Then we see that S is of finite presentation as an S'-algebra by Lemma 10.6.2. Since S' \to S is surjective, the kernel J = \mathop{\mathrm{Ker}}(S' \to S) is finitely generated as an ideal by Lemma 10.6.3. Hence J is a finite S'-module (immediate from the definitions). Thus S = \mathop{\mathrm{Coker}}(J \to S') is of finite presentation as an S'-module by Lemma 10.5.3. Hence, arguing as in the first paragraph, it suffices to show that S' is of finite presentation as an R-module. Actually, S' is free as an R-module with basis the monomials x_1^{e_1} \ldots x_ n^{e_ n} for 0 \leq e_ i < \deg (P_ i). Namely, write R \to S' as the composition
R \to R[x_1]/(P_1(x_1)) \to R[x_1, x_2]/(P_1(x_1), P_2(x_2)) \to \ldots \to S'
This shows that the ith ring in this sequence is free as a module over the (i - 1)st one with basis 1, x_ i, \ldots , x_ i^{\deg (P_ i) - 1}. The result follows easily from this by induction. Some details omitted.
\square
Lemma 10.36.24. Let R be a ring. Let x, y \in R be nonzerodivisors. Let R[x/y] \subset R_{xy} be the R-subalgebra generated by x/y, and similarly for the subalgebras R[y/x] and R[x/y, y/x]. If R is integrally closed in R_ x or R_ y, then the sequence
0 \to R \xrightarrow {(-1, 1)} R[x/y] \oplus R[y/x] \xrightarrow {(1, 1)} R[x/y, y/x] \to 0
is a short exact sequence of R-modules.
Proof.
Since x/y \cdot y/x = 1 it is clear that the map R[x/y] \oplus R[y/x] \to R[x/y, y/x] is surjective. Let \alpha \in R[x/y] \cap R[y/x]. To show exactness in the middle we have to prove that \alpha \in R. By assumption we may write
\alpha = a_0 + a_1 x/y + \ldots + a_ n (x/y)^ n = b_0 + b_1 y/x + \ldots + b_ m(y/x)^ m
for some n, m \geq 0 and a_ i, b_ j \in R. Pick some N > \max (n, m). Consider the finite R-submodule M of R_{xy} generated by the elements
(x/y)^ N, (x/y)^{N - 1}, \ldots , x/y, 1, y/x, \ldots , (y/x)^{N - 1}, (y/x)^ N
We claim that \alpha M \subset M. Namely, it is clear that (x/y)^ i (b_0 + b_1 y/x + \ldots + b_ m(y/x)^ m) \in M for 0 \leq i \leq N and that (y/x)^ i (a_0 + a_1 x/y + \ldots + a_ n(x/y)^ n) \in M for 0 \leq i \leq N. Hence \alpha is integral over R by Lemma 10.36.2. Note that \alpha \in R_ x, so if R is integrally closed in R_ x then \alpha \in R as desired.
\square
Comments (2)
Comment #8406 by Francisco Gallardo on
Comment #9019 by Stacks project on