LECTURE 2

1. I-adic completion

Definition 1.1. Let A be a commutative ring. Let I be a partially ordered set. A pair
((Mi )i , {pij }i≤j ) where Mi are A-modules for all i ∈ I, and pij : Mj → Mi are A-linear
for all i ≤ j ∈ I such that

(1) pii = 1Mi .

(2) pij ◦ pjk = pik for all i ≤ j ≤ k.

is called an inverse (or projective) system of A-modules.

Definition 1.2. Let I be a partially order set and ((Mi )i , {pij }i≤j ) an inverse system of
A-modules. A module M = lim Mi together with a family of A-linear maps qi : M → Mi ,
←−
i ∈ I, is called the inverse limit of the system if

(1) pij qj = qi , for all i ≤ j.

(2) for every A-module X and any A-linear maps fi : X → Mi , i ∈ A, such that
pij fj = fi for all j ≤ i, there exists a unique A-linear map F : X → M such that
qi F = fi for all i ∈ I.

Theorem 1.3. Let A be a ring, I a partially ordered set and ((Mi )i , {pji }i≥j ) an inverse
system of A-modules. Then the inverse limit lim Mi exists.
←−
Q
Proof. Consider the A- submodule M of the direct product i Mi defined by {(mi )i :
pij (mj ) = mi , for all i ≤ j ∈ I}. It is routine to check that M = lim Mi . The maps
←−
qi : M → Mi are the canonical projections.

Let A be a ring and I an ideal of A. We can put a topology on A, where the basis of
the topology is given by the sets of the form x + I n , x ∈ A, n ∈ N.

Lemma 1.4. The collection of sets of the form x + I n , x ∈ A, n ∈ N is a topology basis.

1
2 LECTURE 2

Proof. Clearly, the union of these sets is A. Let x + I n and y + I m that intersect non-
trivially, and let z an element in the intersection. We need to show that there is a set of
form u + I k containing z and inside (x + I n ) ∩ (y + I m ).

Let n ≥ m and so I n ⊆ I m . Now, z = x + a = y + b, with a ∈ I n , b ∈ I m . Then

x−y ∈ I m and so x ∈ y+I m . Therefore x+I n ⊆ y+I m and so x+I n ⊆ (x+I n )∩(y+I m ).

Let us endow A with this topology, which will be called the I-adic topology.

We say that a sequence of elements {xn }n is Cauchy in the I-adic topology if for all n
there exists N such that xi − xj ∈ I n for all i, j ≥ N . A sequence {xn }n of elements from
A converges to 0 if for all n there exists N such that xi ∈ I n for all i ≥ N . A sequence
{xn }n converges to an element x ∈ A such that {xn − x}n converges to zero. We say
that A is complete in the I-adic topology if every Cauchy sequence in A converges to an
element in A.

Note that A/I n together with the natural projections pmn : A/I n → A/I m for n ≥ m
form an inverse system. The I-adic completion of A is by definition ÂI := lim A/I n
←−
which is a natural A-algebra. We will generally drop the symbol I from our notation
when the ideal is understood from the context. Note that we have a natural A-algebra
homomorphism i : A → Â with kernel equal to ∩n I n . We say that A is separated in the
I-adic topology if ∩n I n = 0.

Proposition 1.5. The map i : A → Â is a ring isomorphism if and only if A is complete

in the I-adic topology.

Proof. Let {xn }n be a sequence that gives an element of Â. Since i is an isomorphism
we can find an element x ∈ A such that x − xn ∈ I n for all n which implies easily that
{xn }n converges to x.

Let {xn }n be a sequence that gives an element of Â. Therefore for all n, xn+1 − xn ∈ I n
for all n. Hence xi − xj ∈ I n for all i, j ≥ n which gives that {xn } is Cauchy in A. Hence
it is convergent to x an element in A. Therefore, it is enough to show that if {xn }n is a
sequence that converges to zero, then the corresponding element in  is zero as well. By
 LECTURE 2 3

definition, xj − xi ∈ I j for all j ≤ i. For all n there exists N such that xi ∈ I n for all
i ≥ N . But xi − xn ∈ I n for i ≥ n so xn ∈ I n for all n. 

Remark 1.6. (1) A Cauchy sequence in A defines a unique canonical element in Â.
One can check that the difference between a Cauchy sequence and a subsequence
defines a sequence that converges to zero. This can be used to show that that
given a Cauchy sequence {xn }n in A, we can replace it by a sequence {yn }n
that gives the same canonical element in  with the additional property that
yn+1 − yn ∈ I n for all n. This observation is often useful in computations.
(2) Let CI (R) be the collection of Cauchy sequences on R with the I-adic topology.
Let C0 (R) be the collection of Cauchy sequences in R that converge to 0 in the
I-adic topology.
Prove that there is a canonical isomorphism:

CI (R)
' R̂I .
C0 (R)

2. Properties of completion; Artin-Rees lemma

It is helpful to note that any element of ÂI is given by a sequence {xn } such that
xn+1 − xn ∈ I n . Hence we can find an+1 ∈ I n for all n ≥ 0 such that xn = a1 + · · · + an
for all n ≥ 1.

Example 2.1. It can easily be checked that if A = R[x1 , . . . , Xn ] is a polynomial ring

over a ring R, and I = (X1 , . . . , Xn ), then ÂI = R[[X1 , . . . , Xn ]].

Proposition 2.2. Let A, B be two rings and let I be an ideal of A, respectively J be an

ideal of B. Consider f : A → B be a ring homomorphism such that f (I) ⊂ J. Then
there is a canonical ring homomorphism fˆ : ÂI → B̂ J .

Moreover, if f is surjective such that f (I) = J, then fˆ is surjective.

Proof. We have natural maps A/I n → B/J n , so lim A/I n → B/J n for all n which
←−
implies, by applying the universal property of the inverse limit, the first part.

For the second part, consider a sequence of elements in B, say {bn } such that bn+1 ∈ J n ,
yn = b1 + · · · + bn and let {y n }n give an element in B̂ J . But I n maps onto J n via f , so
4 LECTURE 2

we can find a sequence of elements an+1 ∈ I n mappint onto bn+1 . Set xn = a1 + · · · + an ,

for n ≥ 1. Then {xn }n gives an element in ÂI that maps onto the element corresponding
to {y n }n in B̂ J . 

Corollary 2.3. If A is Noetherian and I = (r1 , . . . , rn ) ⊂ A, then ÂI is Noetherian.

A[[X1 ,...,Xn ]]
Remark 2.4. In fact, we have ÂI = (Xi −ri ,...,Xn −rn )
. This fact is left as an exercise.

Proof. Indeed, let I = (r1 , . . . , rn ). Map R[[X1 , . . . , Xn ]] onto A by sending Xi → ri .

This maps (X1 , . . . , Xn ) onto I and hence we obtain ÂI as a quotient of the Noetherian
ring A[[X1 , . . . , Xn ]]. 

Theorem 2.5. Let A be a ring and I and ideal of A. Let π : ÂI → A/I the natural
projection. Then I ÂI ⊆ Ker(π) ⊆ Jac(ÂI ). This implies that there is a one-to-one
correspondence between the maximal ideals in ÂI and the maximal ideal of A/I. In
particular, the completion of a local ring (A, m) at its maximal ideal is a local ring as
well.

Proof. Let {xn } an element x of ÂI that belongs to Ker(π): xn ∈ I for all n. We will
show that 1 + x is invertible in ÂI . Consider yn = n+1 i i
P
i=0 (−1) xn . It is clear that yn define

a Cauchy sequence in A which therefore gives an element y of the completion. But then
zn = 1 − (1 + xn )yn = xn+2
n ∈ I n+2 . This implies that z = {z n }n is 0 in ÂI and then
1 = (1 + x)y in ÂI . 

Definition 2.6. Let A be a ring, I an ideal of A and M an A-module. We say that a

sequence of elements {xn }n in M is Cauchy in the I-adic topology if for all n there exists
N such that xi − xj ∈ I n M for all i, j ≥ N . A sequence {xn }n of elements from M
converges to 0 if for all n there exists N such that xi ∈ I n M for all i ≥ N . A sequence
{xn }n converges to an element x ∈ M such that {xn − x}n converges to zero in M . We
say that M is complete in the I-adic topology if every Cauchy sequence in M converges
to an element in M . We say that M is I-adically separated if ∩∞ n
n=1 I M = 0.

Definition 2.7. The I-adic completion of M is M̂ I : lim M/I n M . It can be checked that
←−
M̂ I is a ÂI -module and there exists a natural A-module homomorphism M → M̂ I with
kernel ∩n I n M .
 LECTURE 2 5

We say that a filtration of submodules of M say {Nn } is cofinal with the filtration
{I n M } if for all n there exists m such that Nm ⊆ I n M and for all t there exists s such
that I s M ⊂ Nt . It can be checked that lim M/Nn ' lim M/I n M (in fact, the filtrations
←− ←−
define the same linear topology on M ).

Moreover, we can see that a Cauchy sequence and a subsequence of it define the same
element in M̂ I , so we assume that every element m ∈ M̂ is defined by a sequence {mn }
such that mn+1 −mn ∈ I n . Therefore there exists zn+1 ∈ I n such that for yn = z0 +. . .+zn
we have that {yn }n gives m.

Proposition 2.8. Let A be a ring, I an ideal of A. Then

(1) Any A-linear map f : M → N of A-modules induces an ÂI -linear map fˆ : M̂ I →

N̂ I . Moreover, f surjective implies that fˆ is surjective.
(2) There exists a natural isomorphism of ÂI -modules M̂ I ⊕ N̂ I ' M̂ I ⊕ N̂ I , for any
two A-modules M, N .
a
(3) The multiplication by an element M → N defines a natural map ÂI -linear map
M̂ I → N̂ I given by the multiplication by the image of a ∈ ÂI .

Proof. The proof of the first part follows the ring case mutatis mutandis. The last two
parts are straightforward 

Let N ⊂ M be a pair of A-modules. In what follows we need to compare the I-adic

topology on N with the topology induced by the I-adic topology on M restricted to N .
In essence we will show that lim N/I n M ∩ N = N̂ I . To prove this we need to develop
←−
some considerations on filtrations of modules and in fact we will be proving a statement
that is more general.

Definition 2.9. Let M be an A-module and I an ideal of A. Let M = {Mn }n be

a filtration of submodules of M , i.e. Mn+1 ⊂ Mn and M0 = M . We say that M is
an I-filtration if IMn ⊂ Mn+1 for all n ≥ 0. The filtration M is called I-stable if
I n Mm = Mn+1 for n  0.

An example of an I-stable filtration is the one given by {I n M }n . The case of M = A

is particularly important because we can associate the following object to the filtration
{I n }n : grI (A) := I n /I n+1 which is an A-module naturally. In fact this object, which is
6 LECTURE 2

called the associated graded ring with respect to the ideal I is a ring with multiplication
defined as follows: ab = ab for any two elements a ∈ I n , b ∈ I m . It can be checked that
this is well-defined and that is extends via distributivity to a multiplication on grI (A).

Now consider an I- filtration M. We can define the following A-module grM (M ) :

⊕n≥0 Mn /Mn+1 . An important feature of it is that this object is in fact an grI (A)-
module. For a ∈ I n /I n+1 and m ∈ Mk /Mk+1 we let am : am ∈ Mn+k /Mn+k+1 . It can be
checked that definition is well-defined. By distributivity, we can extend this to a scalar
multiplication on grM (M ) with elements from grI (A) and we call it the associated graded
module of M with respect to M.

Proposition 2.10. Let A be a ring, I be an ideal of A, M be an A-module, and M be

an I-filtration on M . Then

(1) If A[It] is a finitely generated A-algebra, if I is a finitely generated ideal.

(2) If A[It]/IA[It] ' grI (A) as A-algebras.

Proof. For (1), let I = (a1 , . . . , ar ). Then A[X1 , . . . , xn ] maps onto A[It] under Xi → ai .
For (2), let A[It] → grI (A) that sends at to a ∈ I/I 2 , for any a ∈ I. It can be easily
check that this is an well-defined A-algebra homomorphism with kernel equal to IA[It].

Proposition 2.11. Let A be a ring, I be an ideal of A, M be a finitely generated module

over A, and M be an I-stable filtration on M composed of finitely generated submodules.
Then grM (M ) is a finitely generated module over grI (A).

Proof. Since the filtration is I-stable so I k MN = MN +k for some N ≥ 0 and all k ≥ 0.

I Mn Mn+1
Therefore I2
· Mn+1
= Mn+2
, for all n ≥ N .

This shows that grM (M ) is generated by the union of all the generators of Mn /Mn+1
for n ≤ N . This is a finite set which proves the claim. 

Definition 2.12. Let I be an ideal in A. The A-algebra RI (A) = A[It] ⊂ A[t] is called
the Rees algebra, or the blowup algebra, of A with respect to I. Note that A[It] = ⊕n≥0 I n .
 LECTURE 2 7

Similarly, for an I-filtration M on an A-module M , we can define the Rees module

of M with respect to M by RM (M ) := ⊕n≥0 Mn tn = ⊕n≥0 Mn . Note that RM (M ) is a
module over A[It] in a natural way.

Theorem 2.13. Let A be a ring, I be an ideal of A, M be an A-module with I-filtration

M consisting of finitely generated A-submodules of M . Then the filtration M is I-stable
if and only if RM (M ) is a finitely generated A[It]-module.

Proof. If M is I-stable, then MN +k = I k MN for some N ≥ 0 and for all k ≥ 0. Then

RM (M ) is finitely generated by the union of the generators of Mi , with i ≤ N .

If RM (M ) is finitely generated over A[It], there exists N ≥ 0 such that all generators
belong to the union of Mi , i ≤ N . But ⊕k>0 MN +k is finitely generated as an A[It]-
module (since it is a homomorphic image of RM (M )). Using this and the fact that M
is an I-filtration we derive that I k MN = MN +k , for all k ≥ 0.

Corollary 2.14 (Artin-Rees Lemma). Let M be a finitely generated A-module, where A

is Noetherian. Assume that I is an ideal of A and let N an A-submodule of M .

(1) Let M be an I-stable filtration on M . Then {Mn ∩ N }n is an I-stable filtration

on N .
(2) The filtration {I n M ∩ N }n is I-stable that is there exists c > 0 such that

I n M ∩ N = I n−c (I c M ∩ N ),

for all n ≥ c.

Proof. It suffices to prove (1). Let M0 the filtration with terms Mn ∩ N . Clearly RM0 (N )
is an A[It]-submodule of RM (M ).

Note that A[It] is a Noetherian A-algebra and RM (M ) is a finitely generated A[It]-

module by Theorem 2.13. So, RM0 (N ) is a finitely generated A[It]-module, hence by
Theorem 2.13 we get that M0 is I-stable on N . 

A simple consequence of the Artin-Rees lemma is the Krull Intersection Theorem

8 LECTURE 2

Theorem 2.15. Let A be a Noetherian ring and I an idea and M a finitey generated
A-module. Let N = ∩n I n M . Then exists an element a ∈ A with 1 + a ∈ I that kills N .
In particular, if A is local then N = 0.

Proof. Show that IN = N and then use NAK. 

Theorem 2.16. Let A be a Noetherian ring and I an ideal of A.

(1) If 0 → N → M → P → 0 is a short exact sequence of finitely generated A-modules

then
0 → N̂ I → M̂ I → P̂ I → 0

is a short exact sequence of ÂI -modules.

(2) The universal property of the tensor product implies that there is a natural ÂI -
map ÂI ⊗A M → M̂ I for any A-module M . For every finitely generated A-module
M , ÂI ⊗A M ' M̂ I is an isomorphism.
(3) ÂI is a flat A-algebras which is faithfully flat if (A, m) is a local ring.

Proof. For (1), consider the short exact sequence

0→N →M →P →0

and tensor with A/I n . We get

N/I n N → M/I n M → P/I n P → 0,

and further
0 → N/(I n M ∩ N ) → M/I n M → P/I n P → 0.

This gives a short exact sequence of inverse limits

0 → lim N/(I n M ∩ N ) → M̂ I → N̂ I → 0,
←−
because the maps in the inverse system defined by N/(I n+1 M ∩ N ) → N/(I n M ∩ N )
are surjective. But by the Artin-Rees Lemma the filtrations {I n M ∩ N }n and I n N are
cofinal and hence
lim N/(I n M ∩ N ) ' lim N/I n N = N̂ I .
←− ←−
 LECTURE 2 9

For (2), note that we have the isomorphism

M ⊗A ÂI ' M̂ I ,

if M is finitely generated and free. Since A is Noetherian, we can find two finitely
generated free A-modules such that

E = Am → F = An → M → 0.
to obtain this presentation of M first map a finitely generated free A-module onto M
and then map another free finitely generated A onto the kernel of this first map.

We obtain the following to exact sequences

(ÂI )m → (ÂI )n → M̂ I → 0

and
Am ⊗A ÂI → An ⊗A ÂI → M ⊗A ÂI → 0.

An application of the Five Lemma (or Snake Lemma) gives the result.

Finally for (3) it is enough to show that tensoring with ÂI preserves exactness of
0 → N → M , where N, M are finitely generated. But this follows directly from (2).