Understanding the Interplay between Ideals and Quotient Rings in Commutative Algebra - Pro, Exams of Algebra

The relationship between ideals and quotient rings in commutative algebra. The author discusses the importance of assuming the height one prime ideal q of a ring r contracts to a prime ideal generated by an element x in a such that a/xa is regular. The document also covers the strong direct summand conjecture and its connection to the vanishing theorem. The text reduces the problem to the case where i = 1 and m is a cyclic module, and explains how to carry through the reductions to the case where a → r → s are complete local domains and a → s is surjective.

Typology: Exams

Pre 2010

Uploaded on 09/17/2009

koofers-user-b5x
koofers-user-b5x 🇺🇸

9 documents

1 / 3

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Math 711: Lecture of December 2, 2005
Example. The following example shows that one needs to assume that the height one prime
Qof Rcontracts to a prime ideal of Agenerated by an element xAsuch that A/XxA
is regular. A=K[[u, v]] = K[[s2, t3]] K[[s, t]]. Q=st. The contraction is u3v2.
Claim: A(u3v2)(st)Rdoes not split. This is, up to isomorphism, the inclusion
of ARthat sends 1 to (s6t)/(st) = s5+s4t+s3t2+s2t3+st4+t5, which is in
(s2, t3)R= (u, v)R.
One other reduction may seem odd, but is important here: we may replace Rby A+Q
R. This is a ring trapped between Aand R, and so is a module-finite extension of A,Qhas
the same contraction to Aas before, and the issue of whether PQsplits over Acertainly
has not changed. Note that once Ais regular local, P=xA will be principal, and the
hypothesis that A/xA is regular says that xis a regular parameter. The statement that xA
splits from Qimplies that xA splits from xR Q, and hence that Ais a direct summand of
Ras an A-module. So the strong direct summand conjecture implies the direct summand
conjecture. We shall further explain the connection to the vanishing theorem later.
We next want to reduce to the case where i= 1. The point is simply that if i > 1 and
we take a short exact sequence
0M0GM0,
where Gis a finitely generated free module, so that M0is a first module of syzygies of M,
we have a commutative diagram:
TorA
i1(M0, R) TorA
i1(M0, S)
x
x
TorA
i(M, R) TorA
i(M, S)
.
This uses the functoriality of the long exact sequence for Tor. Thus, by repeatedly changing
M, we eventually reduce to the case where i= 1.
We next want to reduce to the case where M
=A/I is a cyclic module. In this reduction
we change the rings, and we may lose that the rings are complete local. But we can get
back to that case afterward. Again, we map a finitely generated free A-module Gonto
M. There is an induced map of symmetric algebras A0=SA(G)SA(M). Replace R
by R0=SR(RAG) and Sby S0=SS(SAG). These are all polynomial rings in the
same variables over the original A,R, and S, and we still have A0R0S0. Suppose
we know that TorA0
i(SA(M), R0)TorA0
i(SA(M), S0) is 0. On A0-modulules, the functors
A0R0and AR, may be identified, and we find that TorA0
(, R0) may be identified
with TorA
(, R) on A0-modules. Similarly, on A0-modules the functors A0S0and
1
pf3

Partial preview of the text

Download Understanding the Interplay between Ideals and Quotient Rings in Commutative Algebra - Pro and more Exams Algebra in PDF only on Docsity!

Math 711: Lecture of December 2, 2005

Example. The following example shows that one needs to assume that the height one prime Q of R contracts to a prime ideal of A generated by an element x ∈ A such that A/XxA is regular. A = K[[u, v]] = K[[s^2 , t^3 ]] ⊆ K[[s, t]]. Q = s − t. The contraction is u^3 − v^2. Claim: A(u^3 − v^2 ) ⊆ (s − t)R does not split. This is, up to isomorphism, the inclusion of A → R that sends 1 to (s^6 − t)/(s − t) = s^5 + s^4 t + s^3 t^2 + s^2 t^3 + st^4 + t^5 , which is in (s^2 , t^3 )R = (u, v)R.

One other reduction may seem odd, but is important here: we may replace R by A+Q ⊆ R. This is a ring trapped between A and R, and so is a module-finite extension of A, Q has the same contraction to A as before, and the issue of whether P ⊆ Q splits over A certainly has not changed. Note that once A is regular local, P = xA will be principal, and the hypothesis that A/xA is regular says that x is a regular parameter. The statement that xA splits from Q implies that xA splits from xR ⊆ Q, and hence that A is a direct summand of R as an A-module. So the strong direct summand conjecture implies the direct summand conjecture. We shall further explain the connection to the vanishing theorem later.

We next want to reduce to the case where i = 1. The point is simply that if i > 1 and we take a short exact sequence

0 → M ′^ → G → M → 0 ,

where G is a finitely generated free module, so that M ′^ is a first module of syzygies of M , we have a commutative diagram:

TorAi− 1 (M ′, R) −−−−→ TorAi− 1 (M ′, S) x  

x  

TorAi (M, R) −−−−→ TorAi (M, S)

This uses the functoriality of the long exact sequence for Tor. Thus, by repeatedly changing M , we eventually reduce to the case where i = 1.

We next want to reduce to the case where M ∼= A/I is a cyclic module. In this reduction we change the rings, and we may lose that the rings are complete local. But we can get back to that case afterward. Again, we map a finitely generated free A-module G onto M. There is an induced map of symmetric algebras A′^ = SA(G)  SA(M ). Replace R by R′^ = SR(R ⊗A G) and S by S′^ = SS (S ⊗A G). These are all polynomial rings in the same variables over the original A, R, and S, and we still have A′^ → R′^ → S′. Suppose

we know that TorA

′ i (SA(M^ ), R ′) → TorA′ i (SA(M^ ), S ′) is 0. On A′-modulules, the functors ⊗A′ R′^ and ⊗A R, may be identified, and we find that TorA

′ ( , R′) may be identified

with TorA • ( , R) on A′-modules. Similarly, on A′-modules the functors ⊗A′^ S′^ and 1

2

⊗AS may be identified, and we likewise obtain that TorA

  • (^ , S ′) and TorA - (^ , S) may be identified on A′-modules. We therefore get that the map

TorAi (SA(M ), R) → TorAi (SA(M ), S)

is 0. Since SA(M ) is a direct sum of A-modules (where the summands are indexed by degree) whose degree one component is M , we find that

TorAi (M, R) → TorAi (M, S)

is 0. We have therefore reduce to the case where M is a cyclic module. We may now carry through the reductions to the case where A → R → S are complete local domains again: these reductions do not affect the fact that M is cyclic.

Moreover, we can reduce to the case where A → S is surjective. To do so, pick a coeffi- cient ring V for A. We have V → A → S. We focus on the case of mixed characteristic: other cases are easier. Map a complete unramified regular local ring T  S. One can lift V → S to a map V → T. Cf. [L. Avramov, H.-B. Foxby, B. Herzog, Structure of local homomorphisms J. Algebra 164 (1994), pp. 124–145]. We shall replace A by A′^ = A⊗̂ V T : it maps onto T , and is regular: if one kills a regular system of parameters in A, one gets T /pT , where p is the generator of the maximal ideal in V. This ring is regular, and so A′ is regular. Let R′^ = R ⊗̂ V T. Then R′^ is module-finite over A′, and embeds in a finitely generated free module over A′. Let M ′^ = A′^ ⊗A M. We have maps A′^ → R′^ and R′^ → S. Even the map A′^ → S is surjective, so that R′^ → S is surjective as well.

Then A′^ is faithfully flat over A, and

A′^ ⊗A TorA • (M, R) ∼= TorA

  • (M^

′, R′)

while

A′^ ⊗A TorA • (M, S) ∼= TorA

  • (M^

′, S),

since, on A-modules, (A′^ ⊗A ) ⊗A′ S ∼= ⊗A S. Since S is regular, S = A/(x 1 ,... , xk) where the xj are part of a regular system of parameters.

Therefore, we may assume that A is complete regular local, that R is a domain module- finite over A, and that S is complete and is A/P where P = (x 1 ,... , xh)A is generated by part of a regular system of parameters. We may also assume that the map R → S is surjective. Let Q be the kernel of this map. Since R/Q ∼= A/P , we must have that R = A + Q.

Therefore, we need to understand what it means for the map from Tor^1 (A/I, R), where R = A + Q a domain, to Tor^1 (A/I, R/Q), where R/Q = A/P is regular, to vanish. We shall see that this turns out to mean that I ∩ IQ = IP for every ideal I of A. (Note that because I is an ideal of A and Q is an ideal of R that does not contain 1, I and IQ are ordinarily incomparable.) We give this result as follows: