Vous êtes sur la page 1sur 20

110BH - RING/MODULE THEORY TOPICS (WINTER 2014)

GEUNHO GIM

Contents
1. An example of a rng with no maximal ideal 1
2. Rngs satisfying xn = x for all x ∈ R 2
3. Sums of two squares 3
4. An example of a PID which is not a Euclidean domain 4
5. Polynomial Functions 6
6. Classes of rings 7
7. Examples of diagram chasing 8
8. ACC and DCC 9
M∞ Y∞
9. Z and Z 10
i=0 i=0
10. Selected problems from Elman’s note 12
11. Some Examples 19
References 20

1. An example of a rng with no maximal ideal


Zorn’s lemma shows that a ring has a maximal ideal. But it is possible for a rng (ring without
unity) to have no maximal ideal.
Definition 1.1. An abelian group G is called divisible if for all x ∈ G, n ∈ N, there is y ∈ G
such that x = ny = y + y + · · · + y (or x = y n if we use multiplicative notation).
| {z }
n times

Example 1.2. (Q, +), (R, +), (C, +) and any field of characteristic 0 (considered as an additive
group) are divisible. Q/Z, Zp∞ and any vector space over a field of characteristic 0 are also
divisible. But (Z/pZ, +), (Q× , ·) and (R× , ·) are not divisible.
Proposition 1.3. Let G be a divisible abelian group. Then G does not have a maximal subgroup.
(i.e., for any proper subgroup H of G, there is a proper subgroup H 0 of G properly containing
H.)
Proof. Suppose H ( G is a maximal subgroup. By Correspondence Principle, G/H is simple
and abelian, thus isomorphic to Z/pZ for some prime p. Choose g ∈ G \ H, then there is g 0 ∈ G
such that pg 0 = g. But pg 0 ∈ H since (G : H) = p. Contradiction! 
Proposition 1.4. Let G be a divisible abelian group. Define x · y = 0 for all x, y ∈ G. Then G
is a rng without a maximal ideal.
Updated : Mar 14, 2014
1
Proof. Checking that G is a ring with multiplication · is straightforward. Note that we have
a one-to-one correspondence between ideals of G and additive subgroups of G. Thus maximal
ideals of G correspond to maximal additive subgroups of G, which don’t exist by the previous
proposition. 

2. Rngs satisfying xn = x for all x ∈ R


Theorem 2.1. Let R be a ring satisfying x2 = x for all x ∈ R. Then R is commutative.
Proof. For any x ∈ R, we have 0 = (x + 1R )2 − (x + 1R ) = 2x. For any x, y ∈ R, we have
0 = (x + y)2 − (x + y) = xy + yx = xy − yx
since x2 = x, y 2 = y and 2yx = 0. 

From now on, we let R be a rng (R doesn’t have to have 1R ). Also let Z(R) be the center of
R. Note that Z(R) = {a ∈ R | ar = ra for all r ∈ R} is a subring of R.
Lemma 2.2. Suppose xn = x for all x ∈ R for some n ≥ 2. Then,
(1) if ab = 0 for a, b ∈ R, then ba = 0.
(2) if a2 = ka for a ∈ R, k ∈ Z, then ka ∈ Z(R).
(3) an−1 ∈ Z(R) for all a ∈ R.
Proof. (1) ba = (ba)n = b(ab)n−1 a = 0.
(2) For any x ∈ R, 0 = (ka)x − a2 x = a(kx − ax) and 0 = x(ka) − xa2 = (kx − xa)a. By (1),
we have
x(ka) − axa = (kx − ax)a = 0 = a(kx − xa) = (ka)x − axa
thus x(ka) = (ka)x.
(3) (an−1 )2 = a2n−2 = an · an−2 = a · an−2 = an−1 . By (2), an−1 ∈ Z(R). 
Theorem 2.3. If x3 = x for all x ∈ R, then R is commutative.
Proof. Note that a2 ∈ Z(R) for all a ∈ R by 2.2. Thus we have
xy = (xy)3 = x(yx)2 y = (yx)2 xy = yxyx2 y = yx3 y 2 = y 3 x3 = yx
for any x, y ∈ R. 
Theorem 2.4. If x4 = x for all x ∈ R, then R is commutative.
Proof. For any a ∈ R, we have −a = (−a)4 = a4 = a. Since (a2 + a)2 = a4 + 2a3 + a2 = a2 + a,
a2 + a ∈ Z(R) for all a ∈ R by 2.2(2). Also we have
ab + ba = ((a + b)2 + (a + b)) − (a2 + a) − (b2 + b) ∈ Z(R)
for any a, b ∈ R. For x, y ∈ R, we have
xy = x4 y = x4 y + x2 yx2 − x2 yx2 = x2 (x2 y + yx2 ) − x2 yx2 = (x2 y + yx2 )x2 − x2 yx2 = yx4 = yx
since x2 y + yx2 ∈ Z(R). 
Theorem 2.5. If x5 = x for all x ∈ R, then R is commutative.
2
Proof. Note that (a4 + a2 )2 = a8 + 2a6 + a4 = 2(a4 + a2 ), thus 2(a4 + a2 ) ∈ Z(R) for all a ∈ R
by 2.2(2). Also by 2.2(3), 2a2 = 2(a4 + a2 ) − 2a4 ∈ Z(R) for all a ∈ R. Furthermore,

4a3 = 2(a2 + a)2 − 2a4 − 2a2 ∈ Z(R)


2a3 + a2 + 5a = (a2 + a)5 − (a2 + a) − 10a4 − 2(4a3 ) − 2(2a2 ) ∈ Z(R)

for all a ∈ R. Also,

2(a2 + a)3 + (a2 + a)2 + 5(a2 + a) = 7a4 + 4a3 + 8a2 + 11a ∈ Z(R)

for all a ∈ R. Finally, we have

a = (7a4 + 4a3 + 8a2 + 11a) − 7a4 − 3(2a2 ) − 2(2a3 + a2 + 5a) ∈ Z(R)

for all a ∈ R. Thus, R is commutative. 

Theorem 2.6. If x6 = x for all x ∈ R, then R is commutative.

Proof. For any a ∈ R, we have −a = (−a)6 = a6 = a. Also,

0 = (a2 + a)6 − (a2 + a) = 6a11 + 15a10 + 20a9 + 15a8 + 6a7 = a5 + a3

for all a ∈ R. Thus, a = a6 = a · a5 = a · a3 = a4 for all a ∈ R. By 2.4, R is commutative. 

Question 2.7. What if x7 = x for all x ∈ R?

In general, the following theorem holds.

Theorem 2.8 (Jacobson). Let R be a rng. Assume that for each x ∈ R, there is n(x) ∈ Z≥2
such that xn(x) = x (n(x) is not fixed!). Then R is commutative.

Proof. See [4]. 

3. Sums of two squares


Theorem 3.1. Z[i] is a Euclidean domain.

Proof. We will show that the norm function N : Z[i] → Z≥0 is a Euclidean function on Z[i].
Given x = a + bi, y = c + di ∈ Z[i], let xy = r + si for r, s ∈ Q. Choose m, n ∈ Z so that
|r − m| ≤ 12 and |s − n| ≤ 12 . Let q = m + ni and r = y − qx, then
y  1
N (r) = N (x)N − q = N (x)N ((r − m) + (s − n)i) ≤ N (x) < N (x)
x 2


Since every Euclidean domain is a PID, Z[i] is a PID (and also is a UFD). We will use this
to classify sums of two squares in Z.

Theorem 3.2. Let n ∈ N and write n = 2e0 pe11 · · · perr q1f1 · · · qsfs ∈ Z where pi ≡ 1(mod 4),
qj ≡ 3(mod 4) are odd primes. Then, n = x2 + y 2 for some x, y ∈ Z if and only if fj ’s are
even.
3
Proof. Step 1 : If p ≡ 1(mod 4) is a prime, then p | x2 + 1 for some x ∈ Z.

p−1 p+1
−1 ≡ (p − 1)! ≡ 1 · 2 · · · · · · · (p − 2) · (p − 1)
2  2 
p−1 p−1
≡ 1 · 2··· − · · · (−2) · (−1)
2 2
p−1 2
 
p−1
≡ (−1) 2 1 · 2···
2
 2
p−1
≡ 1 · 2··· (mod p)
2
Step 2 : If p ≡ 1(mod 4) is a prime, then p is a sum of two squares.
By Step 1, p | x2 + 1 = (x + i)(x − i) in Z[i] for some x ∈ Z. Since p - x + i, x − i in Z[i], p is not
prime. Since Z[i] is a PID, p is not irreducible. Thus, we can write p = αβ for some nonunits
α, β ∈ Z[i]. We have p2 = N (p) = N (α)N (β). Because N (r) = 1 if and only if r ∈ Z[i]× , we
have N (α) = N (β) = p. If α = u + vi for u, v ∈ Z, then p = N (α) = u2 + v 2 .
Step 3 : ”if” part of the theorem.
By Step 2, pi ’s are sums of two squares, and so is 2 = 12 + 12 . The equality (x2 + y 2 )(z 2 + w2 ) =
(xz + yw)2 + (xw − yz)2 shows that the product of sums of two squares is also a sum of two
squares. We can write 2e0 pe11 · · · perr = α2 + β 2 for some α, β ∈ N. If fj ’s are even, then
f /2 f /2
n = (αq11 · · · qsfs /2 )2 + (βq11 · · · qsfs /2 )2
Step 4 : ”only if” part of the theorem.
Suppose n = x2 + y 2 and d = (x, y), x = dx0 , y = dy0 , then n = d2 (x20 + y02 ) and (x0 , y0 ) = 1.
If p | x20 + y02 for some odd prime p, then p - x0 , y0 . (otherwise, p will divide both of them.) In
Z/pZ, we have x0 2 = −y0 2 .
 p−1  2 ! p−1 2
x0 x0 p−1
1= = = (−1) 2
y0 y0
in Z/pZ shows that p ≡ 1 mod 4. This shows that qj ’s cannot divide x20 + y02 , thus should divide
d2 . So we have 2 | fj for all j. 
The followings are generalizations of this type of classification.
Theorem 3.3. n ∈ N is a sum of three squares if and only if n 6= 4m (8k + 7) for m, k ∈ Z≥0 .
Theorem 3.4. Every positive integer is a sum of four(thus, or more) squares.
Theorem 3.5. For each n ∈ N, there is an integer g(n) ∈ N such that every positive integer
is a sum of g(n) nth powers.

4. An example of a PID which is not a Euclidean domain



Let θ = −1−2 −19 and R = Z[θ] = {a + bθ | a, b ∈ Z} ⊆ C. We will show that R is a PID, but
is not a Euclidean domain.
Lemma 4.1. (1) R× = {±1}
(2) 2, 3 ∈ R are irreducible.
4
Proof. (1) Consider the norm function N : R → R defined by N (r) = rr̄ where r̄ is the complex
conjugate of r. For a + bθ ∈ R, we have

(2a − b)2 + 19b2
  
b b −19
N (a + bθ) = N a− − = = a2 − ab + 5b2 ∈ Z≥0
2 2 4
We can see that N is a multiplicative homomorphism, thus for x, y ∈ R, N (x) | N (y) in Z if
x | y in R. If N (α) = 1 for some α ∈ R, then α ∈ R× since αᾱ = 1 and ᾱ ∈ R. Conversely, if
α ∈ R× , then there N (α) | N (1) = 1 in Z, thus we have N (α) = ±1.
Write α = a + bθ, then α ∈ R× if and only if N (α) = 1 if and only if (2a − b)2 + 19b2 = 4 if
and only if a = ±1, b = 0 if and only if α = ±1.
(2) Suppose 2 = αβ for α, β ∈ R. We have 4 = N (2) = N (α)N (β). Note that no element in R
has norm 2 because (2a − b)2 + 19b2 = 8 has no integer solution. Thus, N (α) = 1 or N (β) = 1,
i.e., α ∈ R× or β ∈ R× . Therefore, 2 is irreducible. Similarly, so is 3. 
Theorem 4.2. R is not a Euclidean domain.
Proof. Suppose R has a Euclidean function δ : R \ {0} → Z≥0 satisfying the division algorithm.
Choose m ∈ R \ (R× ∪ {0}) with minimal δ(m).
There are q, r ∈ R such that 2 = mq + r with r = 0 or δ(r) < δ(m). By the choice of m, we
have either r = 0 or r ∈ R× = {±1}. If r = 0, then we have 2 = mq. Since 2 is irreducible and
m is not a unit, we have q = ±1 and m = ∓2. If r = −1, then we have 3 = mq. By the similar
argument, we have q = ±1 and m = ∓3. If r = 1, then 1 = mq, which cannot happen since m
is not a unit. Therefore, m ∈ {±2, ±3}.
Now we apply the division algorithm for θ. There are q0 , r0 ∈ R such that θ = mq0 + r0 ,
and similarly, r = 0, −1, 1. We have mq = θ + r. The norm of the left-hand side is N (mq) =
N (m)N (q), thus divisible either by 4 = N (±2) or by 9 = N (±3). On the other hand, we have
N (θ) = N (θ + 1) = 5 and N (θ − 1) = 7 on the right-hand side. Contradiction! 
Theorem 4.3. R is a PID.
Proof. Let a be an ideal of R and √choose 0 6= a ∈ a with minimal N (a). Choose b ∈ a. We √will
show that b ∈ (a). Since Im f = − 219 , there is an integer m ∈ Z such that |Im( ab + mθ)| ≤ 419 .

Case 1 : |Im( ab + mθ)| < 23
There is an integer n ∈ Z such that |Re( ab + mθ + n)| ≤ 21 . Thus N ( ab + mθ + n) < 1,
and N (b + a(mθ + n)) < N (a). Since b + a(mθ + n) ∈ a and N (a) is minimal, we have
b = −a(mθ√
+ n) ∈ (a). √
Case 2 : 2 ≤ Im( ab + mθ) ≤ 419
3

We have
√ √ √
3 √ 3 3 √
   
19 b
− < 3− < 3− ≤ Im 2 + mθ + θ ≤ 0
2 2 2 a
There is n ∈ Z such that |Re( 2b 1
a + 2mθ + θ + n)| ≤ 2 . By the same argument with Case 1, we
have N (2b + a(2mθ + θ + n)) < N (a) and −b − amθ = a(θ+n) 2 ∈ a. If n is even, then aθ
2 ∈ a
and 2 = ( 2 )θ̄ − 2a ∈ a, which is a contradiction since 0 < N ( a2 ) < N (a). If n is odd, then
a aθ
a(θ+1)
2 ∈ a and a2 = ( a(θ+1)
2 )θ + 1 − 2a ∈ a, which is also a contradiction.
√ √
Case 3 : − 419 ≤ Im( ab + mθ) ≤ − 2
3

Apply Case 2 for −( ab + mθ). 

5
5. Polynomial Functions
Let R be a ring and Func(R, R) = {f : R → R | f is a function} (we don’t assume any
property of f ) be the ring of functions from R to R. The ring structure of Func(R, R) is given
pointwise: (f + g)(a) = f (a) + g(a), (f g)(a) = f (a)g(a). We consider the map
Φ = ΦR : R[t] → Func(R, R)
f (t) 7 → (Φ(f ) : a 7→ f (a))
Definition 5.1. A function f ∈ Func(R, R) is called a polynomial function if f ∈ Im Φ.
Example 5.2. Φ is not surjective in general. Let R = R and consider exp ∈ Func(R, R) where
et
exp(a) = ea . For any polynomial f (t) ∈ R[t], we have lim = ∞, thus exp ∈
/ Im Φ.
t → ∞ f (t)

Example 5.3. Φ is not injective in general. Let R = Z/pZ, then we have xp = x for all x ∈ Z/pZ.
Therefore, Φ(tp ) = Φ(t).
Example 5.4. Φ is in general not a ring homomorphism. Suppose that R is not commutative,
and ab 6= ba for a, b ∈ R. Consider f (t) = t, g(t) = a ∈ R[t]. Then, we have (Φ(f )Φ(g))(b) =
(Φ(f )(b))(Φ(g)(b)) = ba 6= ab = Φ(f g)(b).
If R is an infinite domain or a finite field, then we can say something more about the function
Φ.
Theorem 5.5. If R is an infinite domain, then Φ is injective.
Proof. Suppose f (t), g(t) ∈ R[t] and Φ(f ) = Φ(g). This means we have (f − g)(a) = 0 for all
a ∈ R. If f (t) − g(t) 6= 0, then f − g can have at most deg(f − g) distinct roots. Since R is
infinite, we have f (t) = g(t). 
Theorem 5.6. Let R = Fq be a (the) finite field of order q, then we have ker Φ = (tq − t) and
Φ is surjective. Therefore, we have Fq [t]/(tq − t) ∼
= Func(Fq , Fq ).
Proof. (In 110C, we will see that q must be a power of prime, and there are only one finite field
of order q up to isomorphism. But we won’t need these facts for the proof of this theorem.)
Note that |F× ×
q | is a multiplicative group of order q − 1. Thus for any a ∈ Fq , we have a
q−1 = 1.
q q
This means t − t ∈ Fq [t] has all the elements of Fq as its roots, so (t − t) ⊆ ker Φ. Suppose
g(t) ∈ ker Φ. By euclidean algorithm, there are q(t), r(t) ∈ Fq [t] such that g(t) = (tq −t)q(t)+r(t)
and r = 0 or deg r < deg(tq − t) = q. Since g(a) = 0 = aq − a for all a ∈ Fq , we get r(a) = 0 for
all a ∈ Fq by evaluating t = a to the above equation. If r 6= 0, then r can only have deg r(< q)
distinct roots in Fq . Thus we have r = 0 and g ∈ (tq − t).
By the first isomorphism theorem, we have
Fq [t]/(tq − t) ∼
= Im Φ ⊆ Func(Fq , Fq )
Now we can compare two sides by counting the number of elements. Note that for any f (t) ∈
Fq [t], we can find r(t) ∈ Fq [t] such that f (t) = r(t) in Fq [t]/(tq − t) and deg r < q by euclidean
algorithm. Also any r1 (t) 6= r2 (t) ∈ Fq [t] with deg r1 < q and deg r2 < q, we have r1 (t) 6= r2 (t)
in Fq [t]/(tq − t) because the difference r1 − r2 has degree less than q, thus cannot be in (tq − t).
|Fq [t]/(tq − t)| = # of polynomials of degree less than q over Fq
= |{a0 + a1 t + · · · + aq−1 tq−1 | ai ∈ Fq }| = q q
On the other hand, we have
|Fq [t]/(tq − t)| = |Im Φ| ≤ |Func(Fq , Fq )| = q q = |Fq [t]/(tq − t)|
6
This shows the surjectivity of Φ. 
Corollary 5.7. Any function f : Z/pZ → Z/pZ is a polynomial function.
The above corollary is interesting, but doesn’t tell you which polynomial determines the
function f . There’s a useful formula to find a polynomial defining a function f .
Let F be a field (not necessarily finite) and consider the pairs (x0 , y0 ), (x1 , y1 ), · · · , (xk , yk ) ∈
F × F where xi 6= xj if i 6= j. Then we can find a polynomial L(t) ∈ F [t] such that L(xi ) = yi
for all i = 0, · · · , k with deg L ≤ k. We first define
Y t − xj
li (t) = ∈ F [t]
xi − xj
j6=i
0≤j≤k

Note that the denominator is not zero by assumption. Then we get


(
1 if j = i
li (xj ) =
0 if j 6= i
k
X
Now we define L(t) = yi li (t), then L(xi ) = yi is satisfied for all i. This is called Lagrange
i=0
interpolation.
Example 5.8. Consider k + 1 points on the xy-plane whose x-coordinates are distinct. Then
there exists a polynomial in R[t] with degree at most k whose graph passes through those k + 1
points.
Example 5.9. Consider the function f : Z/5Z → Z/5Z defined by f (0) = 3, f (1) = 0, f (2) =
1, f (3) = 1, f (4) = 0. We apply the method above to find a polynomial in Lf (t) ∈ Z/5Z[t]
defining f . Firstly, we have
(t − 1)(t − 2)(t − 3)(t − 4)
l0 (t) = = −t4 + 1
(0 − 1)(0 − 2)(0 − 3)(0 − 4)
Similarly, we can get l2 (t) = −(t3 − t)(t + 2) and l3 (t) = −(t3 − t)(t − 2). Therefore,
Lf (t) = 3l0 (t) + l2 (t) + l3 (t) = 2t2 + 3
We can easily check that we have indeed Lf (a) = f (a) for all a ∈ Z/5Z.
A more general thing also holds, and the proof is similar (count the number of elements!).
Theorem 5.10. Define Φn : Fq [t1 , · · · , tn ] → Func(Fnq , Fq ) similarly. Then Φn is surjective
and ker Φn = (tq1 − t1 , tq2 − t2 , · · · , tqn − tn ). Thus we have
Fq [t1 , t2 , · · · , tn ]/(tq − t1 , tq − t2 , · · · , tq − tn ) ∼
1 2 n q= Func(Fn , Fq )

6. Classes of rings
Consider the following classes of rings.
(1) Field ( (2) Euclidean domain ( (3) PID (
(4a) Noetherian domain, (4b) UFD
( (5) Integral domain ( (6) Commutative ring ( (7) All rings

7
(1) ⇒ (2) ⇒ (3) ⇒ (4a) or (4b) ⇒ (5) ⇒ (6) ⇒ (7) Clear
(2) ; (1) Z √
(3) ; (2) Z[ −1−2 19 ] by the previous section.
(4a) ; (3), (4b) ; (3) Since Z is a UFD and a Noetherian domain, Z[t] is also a UFD ([2] Ch
30) and a Noetherian
√ domain (Hilbert basis theorem).
√ But (2, t) ⊆ Z[t] is not principal.
(4a) ; (4b) Z[ −5] is Noetherian because Z[ −5] ∼
= Z[t]/(t 2 + 5) and Z[t] is Noetherian, but
√ √ √
UFD because 1 + −5 ∈ Z[ −5] is irreducible, but not prime. (1 + −5 | 6 = 2 · 3
is not a √
but 1 + −5 - 2, 3)
(4b) ; (4a) Z[t1 , t2 , · · · , tn · · · ] is a UFD because any polynomial has only finitely many vari-
ables and Z[t1 , t2 , · · · , tn ] is a UFD for all n by induction. But it is not Noetherian because we
have a strictly increasing sequence of ideals (t1 ) ( (t1 , t2 ) ( · · · .
(5) ; (4a), (5) ; (4b) The examples above work.
(6) ; (5) Z/4Z
(7) ; (6) M2 (R)

7. Examples of diagram chasing


Let R be a ring. All arrows in this section are R-module homomorphisms.
Theorem 7.1 (Five lemma). Suppose we have the following commutative diagram of R-modules
with exact rows.
f g h i
A −−−−→ B −−−−→ C −−−−→ D −−−−→ E
    
α α α α α
y 1 y 2 y 3 y 4 y 5
f0 g0 h0 i0
A0 −−−−→ B 0 −−−−→ C 0 −−−−→ D0 −−−−→ E 0
Then the following holds.
(1) If α2 , α4 are surjective and α5 is injective, then α3 is surjective.
(2) If α2 , α4 are injective and α1 is surjective, then α3 is injective.
(3) If α2 , α4 are bijective, α5 is injective and α1 is surjective, then α3 is bijective.
Proof. Diagram chasing! 
Theorem 7.2 (Nine lemma). Suppose we have the following commutative diagram of R-
modules.
0 0 0
  
  
y y y
0 −−−−→ A1 −−−−→ B1 −−−−→ C1 −−−−→ 0
  
  
y y y
0 −−−−→ A2 −−−−→ B2 −−−−→ C2 −−−−→ 0
  
  
y y y
0 −−−−→ A3 −−−−→ B3 −−−−→ C3 −−−−→ 0
  
  
y y y
0 0 0
8
If all columns and two bottom rows are exact, then so is the top row. If all columns and two
top rows are exact, then so is the bottom row.
Proof. Another diagram chasing! 
Theorem 7.3 (Snake lemma). Suppose we have the following commutative diagram of R-
modules with exact rows.
f g
A −−−−→ B −−−−→ C −−−−→ 0
  
α β γ
y y y
f0 g0
0 −−−−→ A0 −−−−→ B 0 −−−−→ C 0
Then, we have the following exact sequence of R-modules.
f˜ g̃ δ f¯0 g¯0 id
0 → ker f ,→ ker α → ker β → ker γ → coker α → coker β → coker γ →C coker g 0 → 0
The maps are defined as follows. f˜ = f |ker α , g̃ = g|ker β and f¯0 , g¯0 , idC are the quotient maps of
f 0 , g 0 , idC . Given c ∈ ker γ, choose b ∈ B such that g(b) = c. By commutativity of the diagram,
we have g 0 β(b) = γg(b) = γ(c) = 0. Thus, β(b) ∈ ker g 0 = Im f 0 . Define δ(c) = ā0 ∈ A0 / im α so
that β(b) = f 0 (a0 ).
Proof. Well-definedness of the maps above and the exactness of the sequence can be checked
by diagram chasing!
For example, let’s check the exactness at ker γ. Firstly, let c = g̃(b) for b ∈ ker β. By definition
of the map δ, δ(c) = a0 where f 0 (a0 ) = β(b) = 0. We have a0 = 0 since f 0 is injective, thus
c ∈ ker δ.
On the other hand, suppose d(c) = 0̄. By definition of δ, there are a0 ∈ im α and b ∈ B such that
g(b) = c and f 0 (a0 ) = β(b). Choose a ∈ A satisfying α(a) = a0 . Since β(b) = f 0 α(a) = βf (a), we
have b − f (a) ∈ ker β and g(b − f (a)) = g(b) − gf (a) = g(b) ∈ C. Thus, c = g(b) = g̃(b) ∈ im g̃.
Other parts of the proof are similar. 
Theorem 7.4 (Kernel-cokernel lemma). Suppose that we have R-module homomorphisms
f g
A → B → C (not necessarily exact). Then we have the following exact sequence.
0 → ker f → ker g ◦ f → ker g → coker f → coker g ◦ f → coker g → 0
Proof. Apply the snake lemma to the following commutative diagram.
f
A −−−−→ B −−−−→ coker f −−−−→ 0
  
gf g 
y y y
1
0 −−−−→ C −−−C−→ C −−−−→ 0


8. ACC and DCC


Theorem 8.1. Let M be a nonzero R-module for a ring R. Then, M has a composition series
(a chain of submodules with simple factors) if and only if M satiesfies both the ascending
(Noetherian) and descending (Artinian) chain condition.
9
Proof. Suppose M has a composition series of lengh n. If either chain condition fails to hold,
then one can find a chain of submodules with length n + 1. The two series have an equivalent
refinement which has length at least n + 1, but the composition series does not have a proper
refinement. Contradiction!
Conversely, suppose that M satisfies both ACC and DCC. Since M is Noetherian, the set of
proper submodules of M has a maximal element, say M1 . Also the set of proper submodules
of M1 has a maximal element M2 . In this way, we can construct a strictly decreasing chain of
submodules of M
M ) M1 ) M2 ) · · ·
with simple factors. Since M also satiesfies DCC, this chain has to stabilize, which gives a finite
composition series of M . 


M ∞
Y
9. Z and Z
i=0 i=0

Definition 9.1. We define the following Z-modules



Y
Z = {(a0 , a1 , a2 , · · · ) | ai ∈ Z}
i=0

M ∞
Y
Z = {(a0 , a1 , a2 , · · · ) | ai ∈ Z, ai = 0 for all but finitely many i} ⊆ Z
i=0 i=0
∞ ∞
Z∼ Z∼
M Y
ai ti as Z-modules (not as rings!).
P
Note that we have = Z[t], and = Z[[t]] by (ai ) 7→
i=0 i=0

M
Remark 9.2. Z is Z-free with a basis {1}. Thus, Z is Z-free with the standard basis {ei }∞
i=0
i=0

Y
where ei = (0, 0, · · · , 0, |{z}
1 , 0, · · · ). But {ei } does not span Z because (1, 1, 1, · · · ) ∈
/ h{ei }i.
ith i=0


Y ∞
M
Lemma 9.3. Let f : Z → Z be a Z-module homomorphism. If Z ⊆ ker f , then f = 0.
i=0 i=0

Z. Since gcd(2i , 3i ) = 1 for all i, there are ai , bi ∈ Z such that


Q
Proof. Let x = (xi ) ∈
ai 2i + bi 3i = xi for all i.
 
Xk−1 X
f ((ai 2i )∞
i=0 ) = f
 ai 2i ei + ai 2i ei 
i=0 i≥k
   
k−1
X X X
= ai 2i f (ei ) + 2k f  ai 2i−k ei  = 2k f  ai 2i−k ei 
i=0 i≥k i≥k

since ei ∈ ker f for all i. So 2k | f ((ai 2i )i ) ∈ Z for all k, which means f ((ai 2i )i ) = 0. Similarly
we have f ((bi 3i )i ) = 0 and f (x) = f ((ai 2i )i ) + f ((bi 3i )i ) = 0. 
10
P P
Remark 9.4. We cannot prove that f ( xi ) = f (xi ) for infinite sums! Note that we have
this for finite sums by repeatedly using the fact that f is an additive homomorphism.

Y
Theorem 9.5. Z is not a free Z-module.
i=0
Q Q Q
Proof. Suppose that Z is free with a basis B. Note that Z is uncountable ( Z contains
Q L
{±1}, which has cardinality ℵ1 ), thus B is also uncountable. Since L each element in Z
can be written as a finite Z-linear combinationL of elements in B, and Z is countable, we
have a countable subset
Q B 0 ( B such that Z ⊆ hBQ 0 i. Choose b ∈ B \ B0 and consider
the projection πb : Z → Z which maps an elementL of Z to its coefficient of b. Then πb is
a well-defined Z-module homomorphism, and Z ⊆ hB0 i ⊆ ker πb . By 9.3, we should have
πb = 0, but we have πb (b) = 1. Contradiction! 
Q
Remark 9.6. Z is an example of a torsion-free Z-module which is not free. This doesn’t
happen in the category of finitely generated modules over Z.

!
Z , then Rm ∼
Y
Theorem 9.7. Let R = EndZ = Rn as free R-modules for all m, n ∈ N.
i=0
(this says that the rank of a free R-module is not well-defined!)
Proof. Consider the following elements in R.
f1 (a0 , a1 , a2 , · · · ) = (a0 , a2 , a4 , · · · )
f2 (a0 , a1 , a2 , · · · ) = (a1 , a3 , a5 , · · · )
g1 (a0 , a1 , a2 , · · · ) = (a0 , 0, a1 , 0, a2 , 0, · · · )
g2 (a0 , a1 , a2 , · · · ) = (0, a0 , 0, a1 , 0, a2 , · · · )
Note that we have f1 g1 = f2 g2 = 1R , f1 g2 = f2 g1 = 0 and g1 f1 + g2 f2 = 1R .
For any h ∈ R,
h = h ◦ 1R = h(g1 f1 + g2 f2 ) = (hg1 )f1 + (hg2 )f2
This shows {f1 , f2 } spans R. If we have h1 f1 + h2 f2 = 0 for some h1 , h2 ∈ R, then
0 = 0 ◦ g1 = (h1 f1 + h2 f2 )g1 = h1 (f1 g1 ) + h2 (f2 g1 ) = h1

L can show2that h2 = 0, thus {f1 , f2 } is linearly independent. Therefore


Similarly we we have
R = Rf1 Rf2 ∼ = R as free R-modules. By induction, we can show that R ∼ = Rn , thus
Rm ∼
= Rn for any m, n ∈ N as free R−modules. 
Definition 9.8. Let M be an R-module. We define the dual R-module of M by
M ∗ = HomR (M, R) = {f : M → R | f is a R-homomorphism}
with (f + g)(m) = f (m) + g(m), (rf )(m) = rf (m) for f, g ∈ M ∗ , r ∈ R.

!∗ ∞ ∞
!∗ ∞
∼ ∼
M Y Y M
Theorem 9.9. We have (1) Z = Z and (2) Z = Z.
i=0 i=0 i=0 i=0

Proof. (1) We define


M  Y
φ : HomZ Z, Z → Z
f 7→ (f (ei ))∞
i=0
We can see that φ is a well-defined Z-module homomorphism. L
If φ(f ) = 0, then we have f (ei ) = 0 for all i. Since {ei } is a basis of Z, we have f = 0.
11
Therefore φ is injective.
For any ((ni )∞
Q L P P
i=0 ) ∈ Z, we define g : Z → Z by g( i ai eL
i) = i ai ni . Then g is well-
defined (only finitely many ai ’s are nonzero) and g ∈ HomZ ( Z, Z). Also we can see that
φ(g) = ((g(ei ))i ) = ((ni )i ), thus φ is surjective.
(2) We define
Y  M
ψ : HomZ Z, Z → Z
f (f (ei ))∞
7→ i=0
Q
We first check well-definedness of ψ. Given f ∈ HomZ ( Z, Z), we choose integers 0 < n1 <
k−1
X
2ni |f (ei )| < 2nk −1 for each k. Consider x = ((2ni )i ) ∈ Z, then
Q
n2 < · · · satisfying
i=0
 
k−1
X X
f (x) = f  2ni ei + 2ni ei 
i=0 i≥k
 
k−1
X X
= 2ni f (ei ) + 2nk f  2ni −nk ei 
i=0 i≥k
| {z }
call this bk

for each k. Because



Xk−1
|f (x)| ≥ |2nk bk | − 2ni f (ei )


i=0
nk nk −1
> 2 |bk | − 2
for all k, and f (x) ∈ Z, we have bk = 0 for sufficiently large k. Since we have 2nk bk =
2nk f (ek ) + 2nk+1 bk+1 , f (ek ) = 0 for sufficiently large k. Thus, f is well-defined.
We can easily check that f is a Z-module homomorphism L from definition.
If ψ(f ) = 0, then we have f (ei ) = 0 for all i. Thus, Z ⊆ ker f , and f = 0 by 9.3. Therefore
ψ is injective. L Q P P
For any ((mi )i ) ∈ Z, we define g : Z → Z by g( ai eiQ ) = ai mi . Then g is well-
defined (only finitely many mi ’s are nonzero) and g ∈ HomZ ( Z, Z). Also we can see that
ψ(g) = (g(ei )i ) = ((mi )i ), thus ψ is surjective. 

10. Selected problems from Elman’s note


Problem (21.20.3). Show if R is a commutative ring, then for any ideal B in Mn (R), there
exists an ideal A in R satisfying B = Mn (A). In particular, if R is simple, so is Mn (R). Show
that Mn (R) is never a division ring if n > 1.
Proof. Let Eij ∈ Mn (R) be the matrix with 1 on (i, j)-th component and 0 on the other
components. Then we have (
Eil if j = k
Eij Ekl =
0 otherwise
Let A be the ideal generated by the entries of elements in B. Clearly, we have B ⊆ Mn (A).
Pt
Conversely, let A = (aij ) ∈ Mn (A). Suppose a11 = s=1 rs bs where rs ∈ R and bs is the
12
(is , js )-th component of Bs ∈ B. Then, we have
t
X t
X
a11 E11 = rs bs E11 = rs E1,is Bs Ejs ,1 ∈ B
s=1 s=1
P
Thus A = aij Eij ∈ B by a similar argument.
If R is simple, then the only ideals of Mn (R) are Mn ({0}) = {0n } and Mn (R), thus Mn (R) is
simple. P
For n ≥ 2, consider E11 ∈ Mn (R). Suppose E11 A = In . If A = i,j aij Eij , then
n
X n
X
Eii = In = E11 A = a1j E1j
i=1 j=1

which is not possible because E22 does not appear in the right-hand side. Therefore, E11 is not
invertible, and Mn (R) is not a division ring. 
Problem (21.20.13). Let a1 , · · · , an be ideals in R, at least n − 2 of which are prime. Let
S ⊂ R be a subrng (it does not have to have 1) contained in a1 ∪ a2 ∪ · · · ∪ an . Then one of
the aj ’s contains S. In particular, if p1 , · · · , pn are prime ideals in R and b is an ideal properly
contained in S satisfying S \ b ⊂ p1 ∪ · · · ∪ pn , then S lies in one of the pi ’s.
Proof. Let m(≤ n) be the minimal number of ai ’s whose union
[ contains S. By reindexing if
necessary, we can assume that S ⊆ a1 ∪ · · · ∪ am and S * aj if |J| < m. Note that
j∈J⊆{1,2,··· ,n}
m
[ [ [
S= (S ∩ ai ) and ∅ 6= S ∩ ai * aj by the choice of m. Choose ai ∈ (S ∩ ai ) \ aj .
i=1 j6=i j6=i
1≤j≤m 1≤j≤m
If m = 2, then we have a1 + a2 ∈ S ⊆ a1 + a2 . If a1 + a2 ∈ a1 , then a2 = (a1 + a2 ) − a1 ∈ a1 ,
which is a contradiction. If a1 + a2 ∈ a2 , we also get a contradiction.
If m ≥ 3, then at least one of ai ’s, say a1 , is prime. Consider the element a = a1 + a2 a3 · · · am ∈
S ⊆ a1 ∪ · · · ∪ am . If a ∈ a1 , then a2 a3 · · · am = a − a1 ∈ a1 . Since a1 is prime, we have
some aj ∈ a1 for j ≥ 2, which is a contradiction. If a ∈ aj for some j ≥ 2, then a1 =
a − (a2 · · · aj−1 aj+1 · · · am )aj ∈ aj , thus a contradiction.
This shows that m = 1, i.e., S ⊆ aj for some j. 

Problem (25.21.2). Produce elements a and b in the domain R = {x + 2y −1 | x, y ∈ Z}
having no gcd. Prove your elements do not have a gcd.
Proof. Consider a = 4 + 4i and b = 8 in R. We can see that 2 and 2 + 2i are common divisors of
√(x) | N (y) in Z if x | y in R, and there is no element having N (α) = 2 in
a and b. Note that N
R because N (x + 2y −1) = x2 + 4y 2 = 2 has no solution in Z. Suppose there is a gcd g = (a, b)
in R.
2, 2 + 2i | g | a, b
gives us
4, 8 | N (g) | N (a) = 32, N (b) = 64
in Z. If N (g) = 8, then N ( g2 ) g
= 2. If N (g) = 16, then we have N ( 2+2i ) = 2. If N (g) = 32, then
8
N ( g ) = 2. In all cases we end up getting elements in R with norms equal to 2, which is not
possible. Therefore, a and b do not have a gcd in R. 
Problem (25.21.14). Prove that a domain in which every prime ideal is principal is a PID.
13
Proof. Let R be a domain satisfying the assumption,
S = {a ⊆ R | a is an ideal which is not principal}
[
and assume that S is nonempty. For any chain of ideals a1 ⊆ a2 ⊆ · · · in S, the union ai is
S i
not principal. (If i ai = (a), then a ∈ aj for some j and aj = (a).) This means any chain in
S has an upper bound in S, so we can apply Zorn’s lemma to get a maximal element m in S.
Note that m is not prime by assumption, thus we have some a, b ∈ R such that a, b ∈ / m and
ab ∈ m. Since m + (a) is an ideal properly containing m, m + (a) = (c) for some c ∈ R. Define
the ideal quotient (m : (a)) = {r ∈ R | ra ∈ m}, then we can easily check that (m : (a)) is an
ideal of R containing m + (b). Let (m : (a)) = (d) for d ∈ R. Since da ∈ m, we have
(cd) ⊆ (m + (a))(d) ⊆ m
On the other hand, for any x ∈ m ⊆ m + (a) = (c), x = cy for some y ∈ R. Because a ∈ (c), we
have ya ∈ (cy) ⊆ m, which means y ∈ (m : (a)) = (d). Thus x ∈ (cd) and m ⊆ (cd). Therefore
m = (cd) and this gives us a contradiction. So S is empty, and R is a PID. 
Problem (29.19.1). Let R be a commutative ring. Show that a polynomial f = a0 + a1 t +
· · · + an tn in R[t] is a unit in R[t] if and only if a0 is a unit in R and ai is nilpotent for every
i > 0.
Proof. (⇐) Suppose a0 is a unit and ai is nilpotent for every i > 0. Consider g = a−1 0 f =
1 + aa01 t + · · · + aan0 tn . Suppose we have ami
i
= 0, and let M = m 1 + · · · + mn . Note that g − 1 is
nilpotent in R[t] by
n
!M
X a i
X
(g − 1)M = ti = a−M
0 br1 r2 ···rn ari i tiri = 0 (br1 r2 ···rn ∈ Z)
a0
i=1 r1 +r2 +···+rn =M

because ri ≥ mi for some i. Thus, g = 1 + (g − 1) ∈ R[t]× (if u ∈ R[t] is nilpotent, then


1 + u ∈ R[t]× : see [2], 23.16.4) and f = a0 g ∈ R[t]× .
(⇒) We give two solutions.
(Solution 1) We use induction on deg f . Suppose f (t)g(t) = 1 for some g(t) = b0 + b1 t + · · · +
bm tm ∈ R[t]. By comparing the coefficients on both sides, we get
an bm = 0
an bm−1 + an−1 bm = 0
an bm−2 + an−1 bm−1 + an−2 bm = 0
..
.
a0 b0 = 1
The last equality a0 b0 = 1 shows that a0 is a unit. Thus if deg f = 0, then we’re done. Now we
assume deg f = n ≥ 1. We multiply an on the second equality, then we get a2n bm−1 = 0. If we
multiply a2n on the third equality, we get a3n bm−2 = 0. In this way, we finally get am+1
n b0 = 0.
Since am+1
n = am+1 b a = 0, a is nilpotent. Now we have g(t)(f (t) − a tn ) = 1 − g(t)a tn ∈
n 0 0 n n n
R[t]× because g(t)an tn is nilpotent. Thus f (t) − an tn is also a unit with deg(f (t) − an tn ) < n.
By the induction hypothesis, ai is nilpotent for all 1 ≤ i ≤ n − 1.
(Solution 2) Let p be a prime ideal of R and consider the projection R[t] → (R/p)[t]. Note that
R/p is a domain, and so is (R/p)[t]. Thus if f ∈ R[t]× , then f¯ = f mod p ∈ (R/p)[t]× = (R/p)× .
14
This implies ai ∈ p for all i > 0 and all prime ideals p. Therefore
\
ai ∈ p = nil(R)
p, prime

for all i > 0. (See [2], Cor 23.15) 


Problem (29.19.2). Let R be a nontrivial commutative ring and f a zero divisor in R[t].
Show that there exists a nonzero element b in R so that bf = 0.
Proof. Write f (t) = a0 + a1 t + · · · + an tn . Let g(t) = b0 + b1 t + · · · + bm tm be a nonzero
polynomial of minimal degree satisfying gf = 0. Suppose m > 0. If gai = 0 for all 0 ≤ i ≤ n,
then bm ai = 0 for all i and we get bm f = 0 which is a contradiction. Thus we can assume that
there is a maximal j ≥ 0 such that gaj 6= 0. Now we have
0 = gf = (b0 + b1 t + · · · + bm tm )(a0 + a1 t + · · · + aj tj + aj+1 tj+1 + · · · + an tn )
= (b0 + b1 t + · · · + bm tm )(a0 + a1 t + · · · + aj tj )
This means aj bm = 0 and (aj g)f = aj (gf ) = 0. But deg aj g < deg g contradicts the minimality
of deg g. Therefore we have m = 0. 
Problem (38.23.6). Let f : Zn → Zm be a Z−module homomorphism. Let Sl be the standard
basis for Zl . Prove that f is monic if and only if the rank of [f ]SSm
n
is n and f is epic if and
only if a gcd of the mth ordered minors of [f ]SSm
n
is 1.
Proof. (Solution 1) We first prove this by using the following theorem.

(Theorem) Let R be a PID and A ∈ Mm,n (R). Then there are matrices P ∈ GLm (R), Q ∈
GLn (R) such that P AQ = diag(a1 , a2 , · · · , ar , 0, · · · , 0) (called a Smith Normal Form of A) for
ai 6= 0, a1 | a2 | · · · | ar . Furthermore, the ideals (a1 ) ⊃ (a2 ) ⊃ · · · ⊃ (ar ) completely determine
a Smith Normal Form of A and al ∼ ∆l /∆l−1 where ∆l = ∆l,A is a gcd of all l × l minors of A.

Let A = [f ]SSm
n
. Then by the theorem above, we can find P ∈ GLm (Z), Q ∈ GLn (Z) such that
P AQ = diag(a1 , a2 , · · · , ar , 0, · · · , 0). Clearly we have r ≤ min(m, n). We write P = [idZm ]βSm
and Q = [idZn ]Sαn where α = {v1 , v2 · · · , vn } and β = {w1 , w2 , · · · , wm } are basis of Zn and Zm
respectively. Now we have
P AQ = [idZm ]βSm [f ]SSm
n
[idZn ]Sαn = [idZm ◦ f ◦ idZn ]βα = [f ]βα = diag(a1 , · · · , ar , 0, · · · , 0)
We can easily see that (
Zvr+1 + · · · + Zvn if n > r
ker f =
0 if n = r
and im f = Za1 w1 + · · · + Zar wr from [f ]βα . Now,
f is monic ⇐⇒ ker f = 0 ⇐⇒ r = n ⇐⇒ rank im f = n ⇐⇒ rank [f ]SSm
n
=n

f is epic ⇐⇒ im f = Za1 w1 + · · · + Zar wr = Zm ⇐⇒ r = m and ai = ±1 for all i


⇐⇒ ∆m,[f ]β = 1
α

⇐⇒ ∆m,[f ]Sm = 1
Sn

because ∆m = a1 a2 · · · am and ∆l of similar matrices are associates for all l by Lemma 37.6.

15
(Solution 2) Now we prove this without using a Smith Normal Form.
For the first part, let Sn = {e1 , e2 , · · · , en } and Sm = {f1 , f2 , · · · , fm }.
n
!
X
f is monic ⇐⇒ ker f = 0 ⇐⇒ f ni ei = 0 implies ni = 0 for all i
i=1
n
X
⇐⇒ ni f (ei ) = 0 implies ni = 0 for all i
i=1
⇐⇒ {f (ei )}ni=1 is a basis of im f
⇐⇒ rank[f ]SSm
n
=n
For the second part, we first prove that if f is epic, then ∆m,[f ]Sm = 1. Let A = [f ]SSm
n
. Since f
  Sn
vi1
 vi2 
is surjective, we can find vi =  ..  such that Avi = fi for each i. Let V be the n × m matrix
 
 . 
vin
whose ith column is vi , then we have AV = Im . We denote the ith column(resp. row) of A by
[A]i (resp. [A]i ). Using the fact that the determinant function is multilinear on columns, we can
show that
1 = det(Im ) = det(v11 [A]1 + · · · + v1n [A]n , · · · , vm1 [A]1 + · · · + vmn [A]n )
= v11 v21 · · · vn1 det([A]1 , [A]1 , · · · , [A]1 ) + · · · + v1n v2n · · · vmn det([A]n , [A]n , · · · , [A]n )
X
= ui1 i2 ···im det([A]i1 , · · · , [A]im )
1≤i1 <i2 <···<im ≤n

for some ui1 i2 ···im ∈ Z. We used the facts that the determinant of a matrix with two identical
columns is zero, and that changing the order of columns only affects the sign of the determinant.
The m × m minors of A look like ([A]i1 · · · [A]im ), thus the above equation says that 1 is a Z-
linear combination of the determinants of m × m minors of A, i.e., ∆m = 1.
Finally, we prove that if ∆m = 1, then f is epic. Since a gcd of the determinants of m × m
minors of A is 1, we can find integers ri1 i2 ···im ∈ Z such that
X
1= ri1 i2 ···im det([A]i1 , [A]i2 , · · · , [A]im )
1≤i1 <i2 <···<im ≤n

Let Bi1 i2 ···im = adj([A]i1 [A]i2 · · · [A]im ) ∈ Mm,m (Z) and define B^ i1 i2 ···im ∈ Mn,m (Z) by
(
[Bi1 i2 ···im ]k if j = ik
[B^ i1 i2 ···im ]j =
0 otherwise
i1 im i1 im
i1 i2 ···im = ([A] · · · [A] )Bi1 i2 ···im = det([A] · · · [A] )Im . Because
Then we have AB^
X  X 
i1 im
A( ri1 i2 ···im B^
i1 i2 ···im ) = ri1 i2 ···im det([A] · · · [A] ) Im = Im

is surjective, so is A = [f ]SSm
n
. 
Problem (38.23.7). Let R be a commutative ring. Let En (R) be the subgroup of GLn (R)
generated by all matrices of the form I + λ where λ is a matrix with precisely one non-zero
entry and this entry does not occur on the diagonal. Suppose that R is a euclidean ring. Show
that SLn (R) = En (R).
16
Proof. We use induction on n.
When n = 1, this is trivial because we have SL1 (R) = {(1R )} = E1 (R).
Suppose n > 1. Note that every generator of En (R) has the form I + λEij for some λ ∈ R and
i 6= j, thus has determinant 1. So it is clear that En (R) ⊆ SLn (R).
Conversely, suppose A ∈ SLn (R). For matrices M, N ∈ Mn (R), we write


M ∼ N if N = P M Q for some P, Q ∈ En (R)

∗ ∗
(∼ is an equivalence relation!) We will show that A ∼ T for some T ∈ En (R), which implies
that A ∈ En (R). We can also easily see that multiplying I + λEij on the left is equivalent to
adding λ times j-th row to i-th row, and multiplying I + λEij on the right is equivalent to
adding λ times i-th column to j-th column.
Since A = (aij ) 6= 0, we can find a nonzero component aij 6= 0. We want to have aij | aik , alj
for all 1 ≤ k, l ≤ n. If this is not the case, pick aik (or alj ) which is not divisible by aij . Since
R is a euclidean ring, we can find elements q, r ∈ R such that aik (or alj )= qaij + r with
r 6= 0, δ(r) < δ(aij ) where δ is a euclidean function on R. Now A(I − qEjk )(or (I − qEli )A) has
r as one of its entries. If r divides all entries on the same row/column, we’re good. Otherwise
repeat the same process. This process has to stop after finite number of steps because we will
have strictly decreasing values of δ, which is defined on Z≥0 .

Therefore we can assume that A ∼ B = (bij ) for some B ∈ SLn (R) satisfying bij | bik , blj for
some i, j and 1 ≤ k, l ≤ n. Let bik = ek bij , blj = fl bij (ek , fl ∈ R) and
   
 Y   Y 
C=
 (I − fl Eli )B 
  (I − ek Ejk )

l6=i k6=j
1≤l≤n 1≤k≤n

then

0
 
..
.
 
 
 .. 
∗ . ∗
 
 
∗ 
0

A∼C= 
 0 ··· ··· 0 bij 0 ··· 0 
 

 0 

 .. 
 ∗ . ∗ 
0

Also note that bij | det C = det A = 1, thus bij ∈ R× . Let c11 be the (1, 1)-th component of C,
and


 (I + E1i )C(I + 1−cbij Ej1 )
11
if i 6= 1, j 6= 1

C(I + 1−c11 E )

if i = 1, j 6= 1
b1j j1
D= 1−c11


 (I + bi1 E1i )C if i 6= 1, j = 1

C if i = j = 1
17
then
 
1 ∗    
∗  ∗
 Y Y
A ∼ D = (dij ) =   ∼  (I − dl1 El1 ) D  (I − d1k E1k )
 ∗ ∗ 
2≤l≤n 2≤k≤n
 
1 0
 
= 
 0

A1 

for some A1 ∈ SLn−1 (R). By inductive hypothesis, A1 ∈ En−1 (R), i.e., there are matrices
U1 , U2 , · · · , Ur such that Uk = In−1 + λk Eik jk for some λk ∈ R, 1 ≤ ik , jk ≤ n − 1 and
A1 = U1 U2 · · · Ur . Define
 
1 0
 
Vk =   = In + λk Ei +1,j +1 ∈ En (R)
 0 k k
Uk 

then  
1 0
∗  
A∼  = V1 V2 · · · Vr ∈ En (R)
 0 A1 


Problem (38.23.8). Let A be a finite abelian group and let
 = {χ : A → C× | χ a group homomorphism}
It is easily checked that  is a group via χ1 χ2 (x) = χ1 (x)χ2 (x). Show that
(1) A and  have the same order and, in fact, are isomorphic.
P
(2) If χ is not the identity element of Â, then a∈A χ(a) = 0.
Proof. (1) Suppose first A is cyclic and A ∼
= Z/nZ. Define
φ:A → Â = Hom(A, C× )
a 7→ (χa : b 7→ ζnab )
2πi
where ζn = e n . φ is a well-defined (χa ∈ Â for all a ∈ A) group homomorphism. (χa+b = χa χb )
If χa = 1A , then 1 = χa (1) = ζna , i.e., a = 0. Thus, φ is injective.
For any χ ∈ Â, 1 = χ(0) = χ(1)n . So χ(1) = ζnk for some k, and χ = χk because χ(1) completely
determines χ. Thus, φ is surjective.
For a general finite abelian group A, we know that A ∼
L
= i Ai where Ai ’s are cyclic groups.
Therefore, we have
\ ∼Y
 ∼ Âi ∼ Ai ∼
M Y
= Ai = = =A
i i i
fact that HomR ( i∈I Ai , B) ∼
L
Note that the second isomorphism comes from a more generalL
Q =
i∈I Hom R (A i , B) (defined by f →
7 (f ◦ ι i ) where ι i : A i ,→ A i ) for R-modules A i , B and
18
any (possible infinite) index set I. (Check this!)
(2) If χ 6= 1A , then we can find b ∈ A such that χ(b) 6= 1. We have
X X X
(χ(b) − 1) χ(a) = χ(b)χ(a) − χ(a)
a∈A a∈A a∈A
X X
= χ(a + b) − χ(a) = 0
a∈A a∈A
P
Thus, a∈A χ(a) = 0. 

11. Some Examples


Example 11.1. Let R = Z/6Z and S = {0̄, 3̄} ⊆ R. S itself is a ring, but is not a subring of R
/ R× .
because 1S = 3̄ 6= 1̄ = 1R . Note that 1S = 3̄ ∈
Example 11.2. Let X be a set and P(X) be the power set(the set of subsets) of X. We define
two operations +, · on P(X) by A+B = (A−B)∪(B −A) = (A∪B)\(A∩B) and A·B = A∩B
for A, B ∈ P(X). We can check that (P(X), +, ·) is a ring with 0P(X) = ∅, 1P(X) = X. Note
that −A = A for A ∈ P(X) and P(X)× = {X}.
Example 11.3. One good thing about Noetherian rings is that we don’t need Zorn’s lemma
to have a maximal element in a set of ideals. However, when we prove that the ascending
chain condition implies maximal principle, we still need Axiom of Choice(AC). In the proof,
we construct a strictly ascending chain of ideals by assuming maximal principle doesn’t hold.
Here we need Axiom of Dependent Choice(DC) to get an infinite sequence of ideals, and it is
known that DC is not provable in ZF without AC. Induction is not enough because it only
gives a finite chain.
Example 11.4. Let F be a field and consider an irreducible polynomial f ∈ F [t]. If F is perfect
(i.e., char F = 0 or char F = p and F = F p for some prime p), then f 0 6= 0.
Suppose f 0 (t) = 0. If char F = 0, then f 0 6= 0 P
because otherwise f is a constant, which is a unit
in F [t]. If char F = p, then there is g(t) = ni=0 bi ti ∈ F [t] such that f (t) = g(tPp ). Because
p
F p
Pn= F , iwe canPfind ai ∈ F such that ai = bi for all i. Then, we have f (t) = ni=0 bi tip =
p n i p
i=0 (ai t ) = ( i=0 ai t ) (binomial theorem!), and this contradicts the assumption that f is
irreducible.
However, this can happen in general. let F = Fp (y) where Fp is a finite field of order p and y
is an indeterminate. Consider f (t) = tp − y ∈ F [t]. Note that y is prime in Fp [y], thus tp − y
is irreducible in (Fp [y])[t] by applying Eisenstein criterion with a prime y ∈ Fp [y]. By Gauss
lemma, it is also irreducible in (Frac(Fp [y]))[t] = (Fp (y))[t] where Frac(R) for a domain R
means the quotient field of R. But we have f 0 (t) = ptp−1 = 0 because char Fp (y) = p.
Example 11.5. We have the following implications of properties of modules.
free =⇒ projective =⇒ flat =⇒ torsion-free
Z/6Z = Z/2Z ⊕ Z/3Z is a free Z/6Z-module, thus Z/2Z is projective but not free over Z/6Z.
Let R = Z ⊕ (⊕∞ i=1 Z/2Z), then the principal ideal (2, 0)R is flat but not projective over R.
Let k be a field and R = k[X, Y ], I = (X, Y ). Then I is torsion-free but not flat over R.
A projective module over a local ring or a PID is free. A flat module over a perfect ring is
projective. A torsion-free module over a Dedekind domain is flat.
19
References
[1] K. Conrad, Expository Papers, http://www.math.uconn.edu/˜kconrad/blurbs/
[2] R. Elman, Lectures in Algebra Course Note, http://www.math.ucla.edu/˜rse/
[3] T. Hungerford, Algebra, Springer, 1974
[4] N. Jacobson, Structure theory for algebraic algebras of bounded degree, Annals of Mathematics, Vol. 46 (1945),
pp. 695-707
[5] J. Milne, Class Field Theory, http://www.jmilne.org/math/CourseNotes/CFT.pdf
[6] R. Wilson, An example of a PID which is not a Euclidean domain, http://www.maths.qmul.ac.uk/
˜raw/MTH5100/PIDnotED.pdf

20

Vous aimerez peut-être aussi