For Tutors Only - Not for Distribution

1. Let R be a ring. We say that an element x ∈ R is nilpotent if there is some n ∈ N such that
xn = 0. Let N = {x ∈ R : x is nilpotent}.
(a) [8 marks] Show that if R is a commutative ring, then N is an ideal in R. Is N necessarily
an ideal if R is not commutative?
(b) [9 marks] Define what it means for x ∈ R to be a unit. Show that if R is commutative
then the set 1 + N = {1 + x : x ∈ N } is a subgroup of R× the group of units of R.
(c) [8 marks] Let R = Z/nZ. Describe explicitly the elements of the group R× and the ideal
N of nilpotent elements in terms of the prime factors of n. Calculate the order of N .
Show that there are infinitely many n for which 1 + N = R× .

Solution: Part a): [S] Clearly if x ∈ N and r ∈ R then if xn = 0 it follows (rx)n = rn .xn = 0.
If xn = y m = 0, then
X m + n
n+m
(x + y) = xr y s = 0,
r+s=m+n
r

since if r + s = m + n we cannot have both r < n and s < m hence one of the factors xr or y s
is zero, and hence each term in the sum vanishes. If R is not commutative then N is not an
ideal: if R = Mat2 (C) for example, then E12 and E21 are both nilpotent, but A = E12 + E21
is a unit since A2 = I.
Part b): [S] An element x ∈ R is a unit if there is a y ∈ R such that xy = yx = 1. The
units in R form a group R× under multiplication. If x, y ∈ N and R is commutative, then
(1 + x)(1 + y) = 1 + (x + y + xy), and since N is an ideal x + y + xy ∈ N if x, y ∈ N , so that
1 + N is closed under multiplication. Moreover, if xn = 0 we have
n−1
X
(1 + x)(1 + (−x)i ) = 1 + (−1)n−1 xn = 1,
i=1
Pn−1
so that 1 + x is a unit, and since N is an ideal again, i=1 (−x)i ∈ N so that (1 + x)−1 is in
1 + N . Thus 1 + N is a subgroup of R× as claimed.
Part c): [S] An element k + nZ is a unit in Z/nZ if and only if there exists a, b ∈ Z such that
a.k + bn = 1, and hence if and only if h.c.f.(k, n) = 1. In terms of the prime factors of n,
k + nZ is a unit if and only if every prime p dividing n does not divide k.
[N] Next note that by the Chinese Remainder Theorem, if n = ri=1 pai i is the prime factor-
Q
ization of n (where the ai > 0 and pi are distinct primes) then k m = 0 mod n if and only if
k m = 0 mod pai i for each i (1 6 i 6 r). Clearly such an m exists if and only if pi divides k for
each i (1 6 i 6 r). It follows that N = {k(p1 . . . pr ) + nZ : k ∈ Z}.
Alternative: Suppose that k m = 0 mod n. Then if p is a prime dividing n it follows p divides
k m , and hence p divides k (by the defining property of prime elements). Thus if pi (1 6 i 6 r)
are the distinct primes dividing n, then we must have pi | k for each everyQ i (1 6 i 6 k).
Conversely if pi | k for all i, then if a = max{ai : 1 6 i 6 r} (where n = ri=1 pai i ), clearly n
divides k a , and so k mod n is nilpotent.
[N] The order of N is thus |N | = si=1 piai −1 . If n = 2k , then there are 2k−1 odd residues
Q
modulo n, and hence 1 + N = R× for any such n. (In fact these are the only integers for which
1 + N = R× ).

Page 1 of 5 Turn Over

 For Tutors Only - Not for Distribution

2. Let n ∈ Z be any integer.

√
(a) [5 marks] Show that Rn = {a + b n : a, b ∈ Z} is a subring of the complex numbers C.
(b) [8 marks] Let Fn = {r/s : r, s ∈ Rn , s 6= 0} ⊆ C. Then Fn is a field containing Q.
Calculate the degree dn of the field extension Fn /Q.
(c) [7 marks] Assume now that Rn is a Euclidean domain. Prove that if x ∈ Fn satisfies
xm + c1 xm−1 + . . . + cm−1 x + cm = 0, where ci ∈ Rn , (1 6 i 6 m − 1) then x ∈ Rn .
√
(d) [5 marks] Is Z[ −3] a Euclidean domain? Justify your answer carefully.
[You may use any standard properties of a Euclidean domain provided you state them clearly.]

√ √
Solution: Part a):[B] Firstly Rn clearly contains 1 and if we have a1 + b1 n, a2 + b2 n ∈ Rn ,
then √ √ √
(a1 + b1 n)(a2 + b2 n) = (a1 a2 + nb1 b2 ) + (a1 b2 + a2 b1 ) n ∈ Rn ,
Similarly: √ √ √
(a1 + b1 n) − (a2 + b2 n) = (a1 − a2 ) + (b1 − b2 ) n ∈ Rn ,
so that by the subring test, Rn is a subring of C.
Part b):[S] There are two cases: If n is a square, say n = m2 , then clearly Rn = Z and so
√
Fn = Q and dn = 1. Now suppose that n is not a square. We claim that 1 and n are
√
linearly independent over Q. Indeed if a + b n = 0 where a, b ∈ Q, then multiplying by a
√
suitable common denominator we would obtain c + d n = 0, where c, d ∈ Z, and by cancelling
common factors we can assume g.c.d.{c, d} = 1. But then we find c2 = nd2 , and so since Z
is a unique factorisation domain, every prime occuring in n occurs to an even power, so that
√
n is a square, contradicting our assumption. Thus {1, n} are Q-linearly independent. Let
√
Qn = Q-span{1, n}. We claim that Fn = Qn and hence dn = 2 for n a non-square. Indeed
since the formulae in part a) show that Qn is clearly closed under addition and multiplication,
√ √
it is enough to check that if r = a + b n ∈ Rn \{0} then 1/r ∈ Qn . But if we set s = a − b n,
√
then rs = a2 − nb2 6= 0 (since r, s 6= 0 as {1, n} are Q-linearly independent and r 6= 0) and
so r.(s/(a2 − nb2 )) = 1 and s/(a2 − nb2 ) ∈ Qn as required.
√
Alternative: There is a unique ring homomorphism φ : Q[t] → C such that φ(t) = n. Its
√
image, im(φ) is a subring of C which, since it consists of linear combinations of powers of n
√
is clearly the subring generated of Q generated by n. Since C is an integral domain (it is
a field) im(φ) is an integral domain, and so ker(φ) is a prime ideal of Q[t]. Moreover, since
t2 − n ∈ ker(φ), the kernel is nonzero, and so since Q[t] is a PID it follows that ker(φ) is
maximal, and hence im(φ) ∼ = Q[t]/ker(φ) is a field, and is therefore clearly the subfield Fn of
√
C generated by n. Now since Q[t] is a PID, ker(φ) is a principal ideal hf i for a unique monic
irreducible polynomial f (since irreducibles are prime in a PID) and hence f | t2 − n. Now by
Gauss’s Lemma, t2 − n is reducible over Q[t] if and only if it is reducible in Z[t], which since
it is monic, is possible if and only if it has an integer root, that is, if and only if n is a square.
If n is a square, clearly Rn = Z and Fn = Q. Otherwise we see that t2 − n is irreducible and
so is equal to f and ker(φ) = ht2 − ni. It follows Fn ∼ = Q[t]/ht2 − ni is of degree two over Q.
Part c): [S] If Rn is an ED, then we may write x = a/b where a, b ∈ Rn are coprime (indeed
this can done in any integral domain provided highest common factors exist, which they do in
any ED) then we find that
bm + c1 am−1 b + . . . + am = 0.
But then it follows b divides am , and so in particular a and b cannot be coprime unless b is a
unit, in which case x ∈ Rn as required.

Page 2 of 5
 For Tutors Only - Not for Distribution

√
Part d): [N] Note that x = 1− 2 −3 satisfies x2 − x + 1 = 0, but x ∈
/ R−3 , whence by part c) it
cannot be a Euclidean domain.
Alternative: Note
√ that the restriction of the map z 7→ z z̄ is a multiplicative map N : R−3 → Z>0
sending a + b −3 to a + 3b2 . Moreover, if z ∈ R−3 is a unit with inverse w, it follows
2

1 = N (1) = N (zw) = N (z).N (w), so that N (z) = 1. Since a2 + 3b2 = 1 if and only if
×
a = ±1 √ and b = 0 we see that R−3 = {±1}. Next note that N takes the value 4 on each of
2, 1 ± −3, and since the equation N (z) = 2 has no solutions it follows that these elements
are irreducible,
√ and moreover
√ since R× = {±1} they are not associates. But then the equation
2.2 = (1 + −3)(1 − −3) shows that irreducibles in R−3 are not prime, and so R−3 is not a
PID and hence not an ED. (This second solution is close to a problem sheet question, and for
that reason I would expect full justification along the lines given, so it is longer to write out
than the first solution.)

Page 3 of 5 Turn Over

 For Tutors Only - Not for Distribution

3. Let R be a commutative ring.

(a) [6 marks] (i) Let M be an R-module and let X ⊆ M be any subset. Define what it
means for X to be linearly independent, what it means for X to span M , and what
it means for M to be a free module.
(ii) Show how any abelian group is naturally a Z-module.
[You need only describe the Z-module structure, not prove that it satisfies the axioms.]
(b) [6 marks] Let M be a free module over a commutative ring R. Give a proof or a counter-
example to the following:
(i) If X is a spanning set for M then X necessarily contains a basis of M .
(ii) If Y is a linearly independent set, then there is a basis X of M containing Y .
(c) [9 marks] Now let R = Z, M = Z3 and X = {(2, 4, 6), (2, 6, 4), (4, 6, 2)}. Let N be
the submodule spanned by X. Find a basis of M adapted to N , that is, find a basis
{e1 , . . . , en } for M and elements r1 , . . . , rm ∈ R such that {r1 e1 , . . . , rm em } is a basis of
N where m ∈ N and m 6 n.
(d) [4 marks] State the theorem on the canonical form for a finitely generated module over
a Euclidean domain R. Applying the theorem in the case R = Z or otherwise, find with
proof how many isomorphism classes of abelian groups there are of order 675.

Solution: Part a)i):[B] AP subset X is linearly independent if for any n ∈ N and x1 , . . . , xn ∈ X,

r1 , . . . , rn ∈ R whenever ni=1 ri xi = 0 then ri = 0 for all i (1 6 i 6 n). A subset X spans M
if M is the only submodule of M which contains X is the entire module M itself. A module
is free if it has a basis, that is, a set B which is linearly independent and which spans M .
Part a)ii):[B] If M is an abelian group, then for m ∈ M define 0.m = m, and inductively
(n + 1).m = n.m + m for n ∈ Z>0 . If n < 0 then set n.m = −((−n).m) the additive inverse
(in M ) of (−n).m.
Part b)i):[S] If we let R = Z and M = Z, then X = {2, 3} spans M (because 1 = 3 − 2) but no
subset of X spans M , so a spanning set need not contain a basis.
Part b)ii): [S] If we take R = M = Z again, then {2} is a linearly independent set, but it
cannot be extended to a basis of M (as the only bases of Z are {1} and {−1}).
Part c): [S] Using row operations on the matrix with rows given by the three vectors in X,
we reduce to an upper triangular matrix with rows {(2, 4, 6), (0, 2, −2)(0, 0, 6)} (and thus these
rows are linearly independent). Thus if we let F = {(1, 2, 3), (0, 1, −1), (0, 0, 1)} then F is a
basis for M since

(a, b, c) = a.(1, 2, 3) + (b − 2a).(0, 1, −1) + (c − a + b)(0, 0, 1)

and thus {2.(1, 2, 3), 2.(0, 1, −1), 12.(0, 0, 1)} is a basis for N (or note that the change of basis
matrix between this basis and the standard basis is invertible since it has determinant 1 ∈ Z× ).
Alternative: Let {f1 , f2 , f3 } be the standard basis of Z3 and let M1 = Span{e1 }, M2 =
Span{f1 , f2 } and M3 = Z3 . We build a basis of M adapted to N by considering Ni = Mi ∩ N .
We have

N = {n(a, b, c) = (2a + 2b + 4c, 4a + 6b + 6c, 6a + 4b + 2c) : a, b, c ∈ Z}

so that n(a, b, c) lies in N1 if 4a + 6b + 6c = 6a + 4b + 2c = 0. The general solution to these

two equations is (a, b, c) = (3k, −7k, 5k) (k ∈ Z), and so N1 = {(12k, 0, 0) : k ∈ Z}. It follows
that if we set e1 = (1, 0, 0) then {e1 } is a basis of M1 and {12e1 } is a basis of N1 . Next
N2 = {n(a, b, c) : 6a + 4b + 2c = 0}, and so c = −3a − 2b, that is N2 = {(10a + 6b, 14a + 6b, 0) :

Page 4 of 5
 For Tutors Only - Not for Distribution

a, b ∈ Z}. Then N2 /N1 ∼ = {(14a + 6b : a, b ∈ Z}, and since h.c.f.{14, 6} is 2, where 14 − 2.6 = 2,
it follows that if e2 = (−1, 1, 0) then 2e2 ∈ N2 and 2e2 +N1 is a basis of N2 /N1 . Thus {e1 , e2 } is
a basis of M2 and {12e1 , 2e2 } is a basis of N2 . Finally, N/N2 ∼ = {6a+4b+2c : a, b, c ∈ Z} = 2Z,
and 2 = 2c lifts to (4, 6, 2) ∈ N , thus {(12, 0, 0), (−2, 2, 0), (4, 6, 2)} is a basis of N . It follows
that if we set e3 = (2, 3, 1) and r1 = 12, r2 = 2, r3 = 2 then {e1 , e2 , e3 } is a basis of M and
{r1 e1 , r2 e2 , r3 e3 } is a basis of N as required.
Part d): [B for statement of theorem, application is N] The canonical form theorem states that
if M is a finitely generated module over a Euclidean domain then there are non-zero non-unit
elements d1 , d2 , . . . , dk ∈ R (where k ∈ Z>0 ) unique up to units, and a unique integer s ∈ Z>0
such that d1 | d2 | . . . | dk and
k
∼
M
s
M =R ⊕ R/di R,
i=1

An abelian
Qk group of order 675 must be of the form Z/c1 Z ⊕ . . . ⊕ Z/ck Z, where 1 < c1 |c2 | . . . |ck
and i=1 ci = 675. Since 675 = 52 .33 we see k 6 max{2, 3}. If k = 1 then the only possibility
is (675). If k = 2 the possibilities are (5, 135),(3, 225),(15, 75), since c1 can only be 5i 3j where
i 6 1 and j 6 1, with (i, j) 6= (0, 0), while if k = 3, each ci is divisible by 3, and the only
possibilities are (3, 3, 75),(3, 15, 15), thus there are 6 isomorphism classes.
Alternative: The primary decomposition for modules over R a PID says that any finitely
generated R-module is isomorphic to a module of the form
M
Rs ⊕ R/pni i R,
i∈I

where I is a finite set, ni ∈ Z>0 and the pi a prime in R, and moreover the pairs (pi R, ni ) are
unique. Applying this theorem, if M is an abelian group of order 675 = 52 .33 , we see that
s = 0, the primes pi must be 3 or 5. Moreover the integers ni attached to 3 must sum to 3
while those attached to 5 must sum to 2. It follows the integers must be (3), (2, 1), (1, 1, 1) for
5 and (2), (1, 1) for 3, thus there are 3.2 = 6 possibilities for the primary decomposition of M .

Page 5 of 5 End of Last Page