Mathematics 517 Homework 7

Edgar A. Bering IV
March 18, 2013

Edgar A. Bering IV

VIII.2.16

Let R be a commutative ring and I, J R be ideals. Then we claim that TorR

1 (R/I, R/J) = (I J)/IJ.
(
R

Also, for i > 1 we have Tori R/I, R/J) = Tori1 (I, R/J)
Consider the short exact sequence 0 I R R/I 0. Tensoring (over R) with R/J we get a long
exact sequence of Tor that looks like this when i > 1
TorR
i (R, R/J) = 0

TorR
i (R/I, R/J)

TorR
i1 (I, R/J)

TorR
i1 (R, R/J) = 0

This establishes our second claim. For the first, since TorR
1 (R, R/J) = 0 the result is this exact sequence
that isnt quite short
0TorR
i (R/I, R/J)

I R/J

R R/J

R/I R/J

but it is enough to tell us that TorR

i (R/I, R/J) is the kernel of : I R/J R R/J. The map
(a b) 7 a b on pure tensors (extended linearly) is dandy, but also R R/J
= R/J, and the map
realizing this isomorphism is just multiplication, so we really want to find out the kernel of (a b) = a b
for a I, b R/J. So, suppose that
k
X
aibi = 0
i=1

Pk

Then we have that, picking representatives,

Also, since I is an ideal, we have that
i=1 ai bi J.
Pk

i=1 ai bi I. Finally, I R/J = I/IJ, since the representatives of the kernel are in I J, we con

clude that TorR

1 (R/I, R/J) = ker = (I J)/IJ.

VIII.2.17
Let M, N be finitely generated modules over a PID (Aluffi lets us assume finitely generated for simplicity).
Then TorR
i (M, N ) = 0 for i 2.
Since R is a PID we have a free resolution of M by 0 Rm1 Rm0 M 0, applying R N we
get a long exact sequence of Tor, and for i 2 this looks like
m0
TorR
, N) = 0
i (R

TorR
i (M, N )

m1
TorR
, N) = 0
i1 (R

whence our claim.

VIII.2.19
A criteria for flatness.
First, an R-module M is flat if and only if every monomorphism of R modules A , B induces a
monomorphism of R-modules A R M , B M . Note that every monomorphism gives us a short exact
sequence
0 A B coker 0
so if M is flat then it induces a monomorphism A R M , B R M . Conversely, given any short exact
sequence
0ABP 0

Edgar A. Bering IV

tensoring with M produces an exact sequence with tail

AM BM P M 0
but the map A B was a monomorphism, so tensoring with M induces a monomorphism, so we can add a
zero to the left and the sequence remains exact. Hence M is flat.
P
Next, we claim it sufficies to verify this condition for only finitely generated B. Suppose i ai mi
maps to zero in B R M and M satisfies our flatness criteria for finitely generated B. This only happens if
i (ai , mi ) satisfies the relations of bilinearity inside F R (B M ), but the relations only involve finitely many
(bi , mi ), so we can restrict our attention to ha1 , . . . , ak i , B 0 where B 0 is generated by the bs involved in
the relations. But the restriction of A , B to A0 , B 0 remains a monomorphism, so A0 R M , B 0 R M
is aPmonomorphism (since we have supposed M satisfied the criteria when the target is finitely generated),
so i ai mi = 0, and the map we care about is then also a monomorphism. So it does indeed suffice to
check this criteria for finitely generated targets B.
Finally, we claim it suffices to verify this condition when B = R and A = I. First, we can assume that
B is finitely generated, by our previous paragraph. Since B is finitely generated we can find a sequence of
submodules A = B0 B1 Br = B such that each Bi+1 /Bi is cyclic (take a generating set and
take A hb1 , . . . bj i as you go up through the generators). The map A , B is then a composition of maps
Bi , Bi+1 , so it suffices to verify that each intermediate induced map Bi M Bi+1 M remains a
monomorphism. Since each Bi+1 /Bi is cyclic we have reduced to the case where B is finitely generated and
B/A
= R/I for some ideal I (since B/A is cyclic). So we have a short exact sequence
0

B/A

so that A M B M is a monomorphism if and only if TorR

1 (R/I, M ) = 0, which is a monomorphism if
and only if the condition is satisfied when B = R and A = I.
So, an R module M is flat if and only if the natural map I M M is a monomorphism, that is
I M
= IM for every ideal I R. We also get out of this that M is flat if and only if TorR
1 (R/I, M ) = 0
for every ideal I.

VIII.2.20
Let R be a PID. Then an R-module M is flat if and only if it is torsion-free. By the previous M is
R
flat if and only if TorR
1 (R/I, M ) = 0 for all ideals I, since R is a PID this is Tor1 (R/(a), M ) = 0 for
R
all ideals I, as discussed in lecture Tor1 (R/I, M ) = (0 :M I). In particular, for every a R, we have
TorR
1 (R/(a), M ) = {m M : a.m = 0} = 0, so M is torsion free. Conversely, if M is torsion free, since R is
a PID, to check that TorR
1 (R/I, M ) = 0 for each I, we need only check that {m M : a.m = 0} = 0 for each
a R, since every ideal is principal, and this is just the statement that M is torsion free. So we conclude.

VIII.2.21
We prove that flatness localizes. That is, M is flat if and only if MP is flat for each prime if and only if Mm
is flat for each maximal ideal.
First, suppose M is flat. Let A B be RP modules. Then the inclusion can also be regarded as an
inclusion of R modules, so we get an induced inclusion A R M , B R M , then by exercise 2.6 on the
previous assignment, A R M
= B RP MP , so MP is flat for all primes P and
= A RP MP and B R M
in particular the maximal ideals.

Edgar A. Bering IV

Now suppose MP is flat for each prime (well see that we can get away with just the maximals in the
proof). Let A B be a monomorphism of modules and let K be the kernel of A R M B R M . Then
K = 0 if and only if KP = 0 for all primes if and only if Km = 0 for each maximal ideal (by exercise V.4.12).
So our real goal is to show that KP = ker(AP RP MP BP RP MP ) for each prime, that will do it. So,
let S = R \ P , then since kernels are universal and localization is exact we can start calculating, applying
the exercise 2.6 isomorphisms will give us
KP = S 1 (K) = ker(S 1 (A R M ) S 1 (B R M ))

= ker(A R MP B R MP )

= ker(AP RP MP BP RP MP )

which is exactly what we needed.

VIII.2.22

Let M, N be R-modules and let S be a multiplicative subset of R. We claim that S 1 TorR

i (M, N ) =
S 1 R
(S 1 M, S 1 N ). In exercise 1.25 last assignment we showed that localization preserves homology, so
Tori
R
1
S Tori (M, N ) = S 1 Hi (M R N ) = Hi (S 1 (M R N )) where M is a free resolution of M . Using the
isomorphisms of exercise 2.6 S 1 (Rmi R N )
= (S 1 Rmi ) S 1 R S 1 N , whence Hi (S 1 (M R N ))
=
S 1 R
1
1
1
(S M, S 1 N ), since the exactness of localization means that S 1 M
Hi ((S M ) S 1 R S N ) = Tori
is a free resolution of S 1 M .

VIII.2.23
Let
0

be an exact sequence of R-modules and suppose P is flat. Then we claim that M is flat if and only if N is
R

flat. To do this we will actually show that TorR

i (Q, M ) = Tori (Q, N ) for each i > 0 for any R-module Q.
This is immediate, apply Q R to the short exact sequence to get a long exact sequence of Tor in which,
for i > 0, we get
TorR
i+1 (Q, P ) = 0

TorR
i (Q, M )

TorR
i (Q, N )

TorR
i (Q, P ) = 0

Since P is flat TorR

i (Q, P ) = 0 for i > 0. Then the isomorphism of Tori (Q, M ) = Tori (Q, N ) specializes to
Tori (Q, M ) = 0 for all Q and i > 0 if and only if that is the case for N , which is exactly M is flat if and
only if N is flat. Moreover, looking at the beginning of the long exact sequence in Tor, since P is flat
TorR
1 (P, Q) = 0
so the induced sequence is always exact.

M R Q

N R Q

P R Q

Edgar A. Bering IV

VIII.2.24
Let (R, m) be a commutative Noetherian local ring and M a finitely generated flat R module.
So M/mM is a vector space, and we can chose a basis m1 + mM, . . . , mr + mM over R/m. Then these
generate M by a corollary to Nakayamas lemma (VI.3.10).
We then get an exact sequence
0

Rr

Since R is Noetherian, Rr is a Noetherian module, and so N is finitely generated. Next, we tensor this
with R/m, since M is flat we get an induced short exact sequence
0

N/mN

(R/m)r

M/m

But the second induced map is an isomorphism by our choice of generators of M , so N/mN = 0, so N = mN ,
so by Nakayamas lemma N = 0 (since R is local m is the Jacobson radical). Since N is zero we conclude
M
= Rr , that is M is free. Since free things are flat, over a local ring we have free if and only if flat.
Were then asked to compare this with exercise VI.5.5 where we show that projective modules over a
local ring are free. The gist Im getting here is that everything is really nice in the local case. Impossibly
nice in the local Noetherian case.

VIII.2.25
Let R be a commutative Noetherian ring and let M be a finitely generated R module. We claim M is flat
if and only if TorR
1 (M, R/m) = 0 for every maximal ideal m of R. First, M is flat if and only if Mm is flat,
by Exercise 2.21, and Tor localizes, so it suffices to prove the claim when (R, m) is a local Noetherian ring.
When R is local Noetherian, by the previous question if M is flat then it is free so TorR
1 (M, R/m) = 0.
(M,
R/m)
=
0.
As
in
the
previous
question
we
can
find
elements
m
Conversely, suppose TorR
1 , . . . , mr M
1
so that their cosets are a basis of M/mM and get an exact sequence
0

Rr

where N is finitely generated. Since TorR

i (M, R/m) = 0, tensoring this with R/m gives us a short exact
sequence
0

N/mN

(R/m)r

M/m

but our mi were a basis, so as in the previous N/mN = 0, so by Nakayamas lemma N = 0 and M is free
and therefore flat.

VIII.3.3
Let R, S be commutative rings, M an R-Module, N an (R, S)-bimod, and P an S-mod. We claim that
M R (N S P )
= (M R N ) S P . First we apply a series of functor adjunction relations, for any S module

Edgar A. Bering IV

Q,
HomS (M R (N S P ), Q)
= HomR (M, HomS (N S P, Q))

= HomR (M, HomS (N, Hom(P, Q)))

= HomS (M R N, HomS (P, Q))

= HomS ((M R N ) S P, Q)

So the functors HomS (M R (N S P ), ) and HomS ((M R N ) S P, ) are naturally isomorphic. The
Yoneda lemma we proved on the last assignment gave a covariant embedding into the contravariant functor
category. A dual proof will give a contravariant embedding into the covariant functor category, but the
important point is that this embedding is fully faithful, whence our claim.

VIII.3.10