G. M. Bergman Companion To Lang's Algebra, p.167

G. M. Bergman Companion to Langs Algebra, p.
167
G. M. Bergman Exercises Companion to Langs Algebra, p.168
EXERCISES
Naming of exercises. We will always refer to exercises in Lang by chapter, and then the exercise-
number they have in that chapter, preceded by : L . E.g., exercise 3 at the end of Chapter II of Lang is
here called Exercise II: L 3.
On the other hand, exercises given below are classified by the chapter and section to which they are
related. For instance, exercises related to Langs section VI.2 are numbered VI.2:1, VI.2:2, etc..
(Exercises related to Appendix 2, where Lang gives background on set theory, and to my handout on The
Axiom of Choice, Zorns Lemma, and all that, to be read with that Appendix, are numbered using A2 in
place of the chapter number, and are given at the end. The one exercise based on Langs pp.ix-x, Logical
Prerequisites, is numbered 0:1.)
If an exercise is related to more than one section, I will list it under the last section with which it
assumes familiarity (using the order of sections noted in the Introduction to this Companion when this
differs from Langs). If one exercise assumes the results of another, I indicate this with the symbol [> ...],
meaning that if assigned, it should be given after the other one. In cases where one exercise refers to an
idea in another, or perhaps is motivated by a result in another, but doesnt require one to assume that
result, I use [cf. ...] rather than [> ...].
Exercises marked * or . When I teach the course, I give homework problems, which all students are
expected to do, and a few extra-credit problems, which require more original thinking.
A problem requiring ingenuity can generally be transformed into a straightforward task by adding steps
to lead the student through the solution, so the same problem may be given, with modification, as an
extra-credit problem one year and a regular homework question the next. Below, I have marked with an
asterisk * those exercises that in the form given here require considerable original thinking, and could be
assigned as extra-credit problems.
It sometimes also happens that an exercise seems difficult until one sees some key trick, which makes it
easy. Starting with the 2002 version of these notes, I have begun using another symbol, , to mark
exercises of this sort. In general, such exercises are not good to assign as homework, but are good for
students to think about. Until now, I have avoided including such exercises in these notes, so there will be
very few until I accumulate more.
(Of course, the starred exercises also generally require some insight or key idea; but in such cases the
insight is generally not as straightforward, and still takes some work to apply.)
Notes below on exercises in Lang. Some of the exercises below generalize, clarify, or correct
exercises in Lang; notes on some others of Langs exercises are given at the end of my exercises for each
section. For convenience, this collection of exercises ends with an index to my commentary on Langs
exercises, both in these pages and in the body of this Companion. (But corrections that apply only to
earlier printings of Lang are only noted in the Errata, as usual.)
~ ~ ~
0:1. Consider a diagram of sets and set-maps formed by taking the square diagram at the bottom of p.ix,
putting a fifth set E in the center, and drawing an arrow connecting E to each of the four objects shown
there. (The directions of the arrows will be specified below.)
(a) Show that if the arrow between A and E goes inward (to E ) and the arrow between E and D
goes outward (to D), then regardless of the direction(s) in which the arrows connecting B and E to E
are drawn, commutativity of the four triangles into which those arrows divide the given square implies
commutativity of the square.
(This involves four different cases depending on the directions of those arrows; but with a little thought,
one can deal with these cases simultaneously.)
(b) Show that if all four arrows point in toward E, or if all four arrows point out from E, then
commutativity of those four triangles does not imply commutativity of the square.
I.1:1. Let X be any set, let G be the monoid of all maps X X, where the multiplication is
composition of maps, and let H G denote {g G ! x X, g 1({x}) is countable}. Show that H is a
submonoid of G.
I.1:2. In this exercise, we shall show that there are enormous numbers of isomorphism classes of monoids
of small finite cardinalities. It will be easiest to start by considering isomorphism classes of semigroups,
that is, sets given with an associative composition, not assumed to have a unit. Then, in part (d), we will
adjoin units.
(a) Suppose A and B are disjoint sets, c an element not in A B, and f : A A B any set-map.
Show that the set S = A B {c} becomes a semigroup Sf if we define xy to be f (x, y) if x, y both
belong to A, and to be c otherwise. Assuming that A and B are finite, express the number of distinct
(but not necessarily nonisomorphic) semigroup structures on the set S = A B {c} that may be obtained
by this construction using various maps f, as a function of the cardinalities card(A) and card(B).
(b) Show that given two maps f, g : A A B, any isomorphism between the semigroups Sf and Sg
must carry A into A, B into B, and c to c. Deduce that among the semigroup structures described
in (a), no more than card(A)! card(B)! can belong to any one isomorphism class.
(c) Taking card(A) = card(B) = n, and using the estimate n! n n , obtain a lower bound on the number
of isomorphism classes of semigroups of 2n +1 elements.
(d) Show that if G is any semigroup, and e an element not in G, then G {e} can, in a unique
manner, be made into a monoid having G as a subsemigroup and e as unit. (Note: do not confuse the
uniqueness asked for here with the uniqueness of the identity element of a monoid, proved by Lang on
p.3.) Show that if G and H are nonisomorphic semigroups, then G {e} and H {e} become
nonisomorphic monoids.
(e) Deduce a lower bound on the number of isomorphism classes of monoids of 2n +2 elements.
(f) Find reasonably small integers M and N such that there are at least 100 nonisomorphic monoids of
M elements, and at least 10 100 nonisomorphic monoids of N elements. (You may obtain these bounds
using the estimate of point (e), or you may see whether you can get still smaller M and N by ad hoc
choices of unequal values for |A| and |B|.)
*I.2:1. Let G be a group and S a subset. Let D be the set of all elements x G which have the
property that for every group H and every two homomorphisms p, q: G H agreeing on S, we also
have p(x) = q(x). By the same sort of argument as we gave in the main part of these notes concerning
the statement at the bottom of p.10 of Lang, we can see that D is a subgroup of G containing the
subgroup generated by S. Must it equal the subgroup generated by S ?
I.2:2. (a) Suppose G is a group and H, K are subgroups, and we define HK = {hk ! h H, k K }.
Show that HK is a subgroup of G if and only if it is closed under taking inverses.
(b) Show by example that the corresponding statement is false for 3-fold products HJK of subgroups
of G.
Note on I: L 44 (p.79): The assumption that A and A be abelian is not needed. This exercise can go
with I.2.
I.2:3. (a) Suppose G and H are groups, and a, b, c : G H are set-maps, such that for all x G,
a(x) b(x) = c(x). Show that if any three of the following conditions hold, so does the fourth:
(i) a is a homomorphism.
(ii) b is a homomorphism.
(iii) c is a homomorphism.
(iv) Every element of a(G ) commutes with every element of b(G ).
(b) If one takes H = G, lets a and b be the identity endomorphism of G, and defines c in the
unique way making part (a) applicable, what does part (a) tell us?
(c) If one takes H = G, lets a be the identity endomorphism of G, lets c be the trivial
endomorphism (taking every element to e), and defines b in the unique way making part (a) applicable,
what does part (a) tell us?
I.3:1. = I: L 3 (p.75) but with the last sentence made more precise: Show that G G c is abelian, and that
any homomorphism f of G into an abelian group A can be factored f = hq, where q is the canonical
homomorphism G G G c , and h is a homomorphism of abelian groups, G G c A.
(Recommendation: Before reading carefully the above revision, think about the version in Lang, and try to
guess what it means, since you will frequently encounter such abbreviated statements.)
I.3:2. Do I: L 5 (p.75). Show moreover that in this situation, H is the inverse image of the graph in
question under the natural map G G G N G N .
Deduce that if H is any subgroup of G G, then there exist subgroups K G, K G and
homomorphisms f, f from these respective subgroups into a common group L, such that
H = {(k, k ) K K ! f (k) = f (k )}.
Intuitive idea: Regard H G G as a relation. If two elements r1 , r 2 G are both related via
H to a common element s1 of G, then since H was assumed a group, r1 r 21 is related to e. One
finds that the condition of relating to a common element is an equivalence relation on the group K of
elements of G that relate to elements of G. If one glues together elements of K via this
equivalence relation, and does the analogous gluing job on the group K of elements of G that relate to
at least one element of G, one gets factor groups K N and K N , and the relation given by H
becomes an isomorphism between these, which uniquely determines H.
(Note that for monoids in place of groups, the corresponding arguments fail, because of the lack of
inverses; and there are submonoids of product monoids which do not arise from isomorphisms of
subfactors. An example is the submonoid of Z Z given by {(x, y) ! x y}.)
(Remark on Langs use of the phrase identify ... as in I: L 5 and elsewhere. This usage is not
standard. There is, of course, the standard use of to identify X as Y meaning to determine that X is
Y: He identified the specimen as belonging to an extinct species. And there is the mathematical use of
the phrase to identify X with Y meaning to treat X and Y as the same sometimes by a formal
construction such as dividing out by an equivalence relation; other times by ignoring an irrelevant
distinction, for instance, when one identifies Sn with a subgroup of Sn+1 . It is not clear to me whether
Lang is using identify as in the sense of identify with, or in a blend of the two meanings, e.g., we
determine that N is isomorphic to a normal subgroup of G, and will henceforth not distinguish them.
In any case, I advise you not to adopt this phrase of his.)
I.3:3. (a) Let A 0 A and B 0 B be subgroups of a group G. Prove the following equalities. (The
product-sets named are not assumed to be subgroups.)
(A B 0 ) B = (A B) B 0 , A (A 0 B) = A 0 (A B), (A B 0 ) (A 0 B) = A 0 (A B) B 0 .
(b) Show using the above results (without any further calculations with group elements) that in the
diagram for the Butterfly Lemma on p.21 of Lang, the three groups shown immediately above the
unlabeled node have the property that the intersection of any two of them is the product of the two groups
shown immediately below that node. I.e., show that each of the intersections
(1) (u (U )) (U V ), (U V ) ((u V ) ), ((u V ) ) (u (U ))
is equal to the product
(2) (u V ) (U ).
Hint: The operation on sets is associative and commutative, so the first expression in (1) is
simply the intersection of the sets u (U ), U, and V. An inclusion relation holds between two of
these three sets, allowing one of those two to be dropped from the expression. A similar observation
applies to the second expression in (1). When these redundant terms have been dropped, each of the
expressions in (1) can be transformed using one of the results of part (a).
(c) Verify that the second display on p.21 of Lang is indeed, as he claims, an instance of the third display
(= the second boxed equation on p.17), gotten using for H and N the groups named on the line after
the latter display. Thus, you need to show that H is contained in the normalizer of N, and that the
intersection and product sets in the third display equal the appropriate sets in the second display.
I.3:4. Let G be a group and S a subset of G. We claim there is a least normal subgroup N of G
containing S, i.e., a normal subgroup which contains S, and is contained in every normal subgroup
containing S.
There are two natural ways of constructing this subgroup: as an intersection, and as the closure of S
under appropriate operations.
(a) Prove the existence of such an N in one of the above ways, and describe precisely the other approach
(without going through the proof that it gives the desired group).
(b) Show that if S is closed in G under conjugation (i.e., if for all s S and g G one has
g s g 1 S ), then the above subgroup coincides with the subgroup of G generated by S. Deduce from
this the normality statement of I: L 3.
The group N described above can be called the normal subgroup of G generated by S . (Lang
avoids this wording, perhaps feeling that it would suggest that N is generated by S as a group; e.g., in
the next-to-last sentence on p.68 he simply calls it the smallest normal subgroup containing S .)
(c) Suppose G is a group, H a subgroup, and S a subset of H. Let N1 be the normal subgroup of
H generated by S, and N 2 the normal subgroup of G generated by S. Prove an inclusion that must
hold between N1 and N 2 , but show by example that these subgroups need not be equal, even if H is
normal in G.
Note on I: L 4 (p.75): This result is proved (without the assumption K NH ) as display (c9) of this
Companion. (The assumption K NH has the advantage of guaranteeing that the set HK is a group;
but the result is true without it.)
I.4:1. Show that if m and n are positive integers, there is a nontrivial homomorphism from the cyclic
group of order m to the cyclic group of order n if and only if m and n are not relatively prime.
(Hint: To prove concisely that a homomorphism exists from one cyclic group to another which sends the
given generator to the desired image, note that Zm can be described as the factor-group of Z by the
subgroup mZ, and that the universal property of factor-groups (triangular diagram on p.14 of Lang)
shows just what is required to obtain a homomorphism on a factor-group.)
I.4:2. (a) Show that if a group G is generated by a set S of elements such that every two elements of
S commute, then G is abelian.
(b) Deduce that if G is a group with center C, and G C is cyclic, then G is abelian, i.e., G C is
trivial.
(c) Deduce I: L 7 (p.75).
I.4:3. (a) Show that if a group has a largest proper subgroup (i.e., a proper subgroup that contains all
other proper subgroups), then it is cyclic of finite order.
(b) For which positive integers n does the cyclic group of order n have a largest proper subgroup?
(c) Show that a group having a smallest nonzero subgroup need not be cyclic. (Hint: Look at the groups
described at the bottom of p.9.)
I.4:4. Lets see how far we can generalize the first two parts of the preceding exercise.
(a) Show that if a group G has proper subgroups H1 and H 2 , such that every proper subgroup of G
is contained either in H1 or in H 2 , then G again must be cyclic of finite order.
(b) For which positive integers n does the cyclic group of order n have the property of part (a)?
Clearly, this above pattern cannot continue with the 1 of the preceding exercise and the 2 of
part (a) above replaced by arbitrarily large integers, since that would imply that every finite group was
cyclic! In fact, things start to change at the next step:
*(c) Show that if a group G has proper subgroups H1 , H 2 and H 3 such that every proper subgroup
of G is contained in one of these, then either G is cyclic of finite order, or G has a homomorphism
onto Z 2 Z 2 .
(d) For which positive integers n does the cyclic group of order n have the property assumed in
part (c)? Show that the groups having that property also include Z 2 Z 2 itself, and the two nonabelian
groups of order 8 described at the bottom of p.9 of Lang.
(We will continue the investigation of these groups in Exercise I.6:7.)
I.5:1. Let G be a group. A G-set S is called transitive if it consists of a single orbit.
(a) Let S and T be transitive G-sets, and s an element of S. Show that S and T are isomorphic as
G-sets if and only if there exists t T such that Gs = Gt .
Note that if H is a subgroup of G, the set of left cosets G H can be made a G-set under left
translation: gx = {ga ! a x} (g G, x G H ), and that this G-set will be transitive. Now
(b) Show using (a) that every transitive G-set S is isomorphic to G H for some subgroup H of G.
(c) For S, T, s as in (a), generalize the result of (a) to a criterion for there to exist a homomorphism
S T as G-sets.
(d) Suppose we are given transitive G-sets S and T, and we know the isotropy subgroups H = Gs and
K = Gt of some s S and some t T. Express the criteria of (a) and (c) above in terms of H and K.
(Note that we are concerned with criteria for there to be some isomorphism or homomorphism between the
indicated G-sets; not necessarily one that takes s to t ! So, to express the criterion of (a), you need to
find a group-theoretic condition on H and K which is equivalent to the statement that there exists a
t T with Gt = Gs ; and you need to adapt similarly the condition you got in (c).)
I.5:2. [> I.5:1] (a) Show that if G is a finitely generated group, then for each nonnegative integer n
there are, up to isomorphism, only finitely many n-element G-sets.
(b) Deduce that a finitely generated group has only finitely many subgroups of any finite index n, hence
has only countably many subgroups of finite index.
The next two parts show that two weakened versions of the above statement are false.
(c) If X is a set, and G a subgroup of the additive group of all Z-valued functions on X, then for each
x X, let G (x) = { f G ! f (x) is even}. Show that each of these groups has finite index in G. Give an
example of an uncountable set X and a countable subgroup G of the above sort, such that all the
subgroups G (x) are distinct. (Suggestion: G might be the additive group generated by the characteristic
functions of a countable family of subsets of X which separates points, i.e., has the property that for every
two distinct points x, y X one of these sets contains one of x, y but not the other.) Thus, G is a
countably generated group with uncountably many subgroups of finite index.

(d) Let G be the group of permutations of the integers generated by the translation function t : n n +1,
and the permutation that interchanges 0 and 1 but leaves all other integers fixed. For every set S of
even integers, let G (S ) denote the subgroup of elements of G that fix all members of S. Show that the
subgroups G (S ) are all distinct. Thus, the finitely generated group G has uncountably many subgroups.
Lest the example of (c) lead us to think that all non-finitely-generated groups have uncountably many
subgroups, let us note:
(e) Let p be a prime, and let G be the set of rational numbers of the form a p n for a, n Z. Show
that G forms a subgroup of the additive group of Q, that G is not finitely generated, but that G has
only countably many subgroups.
I.5:3. (a) Let G be a group, H a subgroup, and S a transitive G-set with an element s such that
Gs = H. Let U S be the fixed-set of H, i.e., the set of u S such that ( h H ) hu = u. Show that
if H either has finite order or has finite index in G, then {g G ! gs U } is the normalizer of H
in G.
(b) How can the above set of elements be characterized in the general case (when H is not assumed to
have finite order or index)?
(c) Show by example that in the general case, the above set need not be the normalizer. (Suggestions:
Let G be the permutation group of the integers, and H the subgroup of permutations that fix all negative
integers. Or, let G be the group of maps of the real line into itself of the form x ax +b for real
constants a 0 and b, and H the set of such maps such that a = 1 and b is an integer.)
I.5:4. (a) = I: L 15 (p.76). (Hint: Count the pairs (g, s) such that g fixes s, first as a sum over g, and
then as a sum over s.)
(b) Do I: L 16 (p.76) without using I: L 15. (Hint: G is the union of the left cosets of H, but these are
disjoint. You may ignore the reference to Chapter XIII.)
(c) Show that each of the above results can be deduced easily from the other.
(d) State the lower bounds that your proofs of (a) and (b) give for the number of elements of G having
no fixed points, respectively the number of elements not in the union of all conjugates of H.
I.5:5. If G is a group and N a normal subgroup, let Aut(G, N ) denote the group of those Aut(G)
such that (N ) = N.
(a) Show that every Aut(G, N ) induces a Aut(G N ), and that this construction yields a
homomorphism Aut(G, N ) Aut(G N ).
(b) Show by examples that the above homomorphism need not be one-to-one or onto.
(c) We clearly also have a restriction homomorphism Aut(G, N ) Aut(N ); combining this with the
homomorphism of (a), we get a homomorphism Aut(G, N ) Aut(N ) Aut(G N ). Is this always one-
to-one? Onto?
I.5:6. Let G be a group of permutations of a set S, and n a positive integer. One says that G is
n-transitive (or that it acts n-transitively on S, or that S is an n-transitive G-set) if S has at least n
elements, and if whenever s1 , ... , sn , t1 , ... , tn are elements of S such that s1 , ... , sn are distinct and
t1 , ... , tn are distinct, there exists a g G such that g(s1 ) = t1 , . . . , g(sn ) = tn .
(a) Show that for all n and m with m n, the symmetric group Sn is an m-transitive group of
permutations of {1, ... , n}.
(b) Given a positive integer n, for what values of m is the alternating group An m-transitive as a
group of permutations of {1, ... , n}?
(c) Show that for p a prime, the group of permutations of the set Z pZ consisting of all maps x
ax + b (a, b Z pZ, a 0) is 2-transitive, but not 3-transitive. (This part assumes familiarity with the
definition of the ring Z pZ of integers mod p.)
(d), (e), (f) = I: L 47 (p.80), parts (a), (b), (c) respectively, excluding the last assertion of (c), which we give
separately as:
(g) Show that any 2-transitive group of permutations on a set is primitive, in the sense defined in I: L 46.
(h) Show by example that the converse to (g) is false.
I.5:7. (a) Show by example that Exercise I: L 38(d) (p.79) does not remain true if we delete the condition
that n be prime.
(b) Show, however, that again becomes true if we add the condition s = r +1.
Of course, (a) and (b) will be subsumed if you do (c), which is not as hard as it might seem.
(c) Determine necessary and sufficient conditions on n, r and s for Sn to be generated by [123 ... n]
and [rs].
One may ask, further
(d) Given a transitive subgroup G Sn , must there be a transposition [rs] which, together with G,
generates Sn ?
I.5:8. Let G be a group, S a transitive G-set, s an element of S, and H = Gs . Let Aut(S ) denote
the group of automorphisms of S (invertible homomorphisms S S ) as a G-set. Prove that Aut(S )
=
NH H.
Note on I: L 57 (p.82, based on I.5): Both the Examples at the end have the hypothesis X = GY from the
preceding paragraph. In Example 2, I think discretely just means that there is no assumption that the
operation is continuous in G. (If so, the word is superfluous, since G was assumed discrete.)
I.6:1. [> I.4:2 and I.5:5(a)]. (a) = I: L 24 (p.77). (Suggestions: Get commutativity with the help of
Theorem I.6.5 and I.4:2(b) above. If the group has exponent p, show with the help of Proposition I.2.1
that it is isomorphic to Zp Zp , while if it does not have exponent p, show that it is isomorphic to
Zp 2 .)
(b) Show that if a p-group H acts on a nontrivial p-group G by automorphisms, then its fixed group is
nontrivial. (Hint: You hardly need to use the fact that G is a group.)
Now in the situation of (b), if the fixed subgroup is a normal subgroup N, then I.5:5 becomes
applicable, and allows you to apply (b) above to G N as well. This can help you get the remaining two
parts:
(c) Suppose G is a group isomorphic to Zp Zp , and is an automorphism of G of order p. Show
that G is generated by two elements a, b such that (a) = a and (b) = ab. Show that this is
equivalent to saying that there exists an isomorphism : G
= Zp Zp under which corresponds to the

automorphism of Zp Zp given by (i, j) = (i, j+i). (Indicate the precise sense in which
corresponds to under .)
(d) Show that if G is a group isomorphic to Zp 2 , the only automorphisms of G of exponent p are the
powers of the map a a 1+ p.
I.6:2. [> I.6:1]. Let p be a prime. We shall determine here the structures of all noncommutative groups
of order p 3.
Suppose G is such a group. By Corollary I.6.6, G has a normal subgroup of order p 2; moreover,
by Lemma I.6.7 every subgroup of G of that order is normal. Let us now consider three cases:
(i) G has a normal subgroup N isomorphic to Zp Zp , say generated by elements b and c, and
also has an element a of order p outside of N. Deduce that G is a semidirect product of N and the
subgroup H generated by a. Show that by choosing generators b and c of N appropriately, one can
get c to commute with a, and to satisfy a 1 ba = bc. Assuming the generators are so chosen, obtain a
precise description of the multiplication of G on elements a ib jc k .
Now by I.6:1(a), either (i) above holds or G has a normal subgroup N isomorphic to Zp 2 , say with
cyclic generator b. (Conceivably, both may be true.) Assuming the latter condition, let a be an element
of G not in N. Show that a cannot have order p 3. Thus there are two cases:
(ii) The order of a is p. Deduce that G is again a semidirect product of N and the cyclic group
H generated by a, and that this time, by a change of generator of H, i.e., by replacing a if necessary
by another generator, one can make a 1 ba = b p+1. Again, describe the multiplication explicitly.
(iii) The order of a is p 2. Again show that by modifying the choice of a, one can make
a ba = b p+1. Show also that by modifying the choice of b, one can get b p = a p . Now if we let H
1
denote the subgroup of G generated by a, then the action of H on N by conjugation allows one to
construct a semidirect product group of order p 4. Show that G is the quotient of this group by a certain
normal subgroup of order p. Verify that every element of G can be written uniquely in the form a ib j
with 0 i < p, 0 j < p 2, and describe the multiplication of such elements.
Thus, for every p, we have constructed three nonabelian groups of order p 3, such that every group
of that order has one of these three forms. Show, now, that if p = 2, a certain two of these are
isomorphic, if p 2, a different two are isomorphic, and that in each case, the remaining one is not
isomorphic to the others.
Thus, for each p, there are, up to isomorphism, exactly two nonabelian groups of order p 3.
(From the above facts, it follows that there must be one of constructions (i)-(iii) that we could drop
from our list, and then say that for every prime p, every group of order p 3 is isomorphic to exactly one
of the two groups on the list. However, for p odd, the construction we would be dropping gives a
simpler description of the group in question than the one that is isomorphic to it. So I think this three-
constructions-with-isomorphisms description is more enlightening than the two-groups description would
be.)
I.6:3. [> 1.5:3(a)]. Except where the contrary is stated (in part (d) below), let G be a p-group.
(a) Suppose H is a proper subgroup of G, and S a transitive G-set with an element s such that
Gs = H. Show that the set U defined as in I.5:3(a) has cardinality > 1. (Hint: youve had a result about
G-sets when G is a p-group.)
(b) Deduce that the normalizer of a proper subgroup H of G is strictly larger than H.
(c) Deduce that any maximal proper subgroup of G is normal.
(d) Show that in any group G, a subgroup which is both maximal among all proper subgroups, and
normal, must have prime index.
(e) Let G again be a p-group, and let G p be the subgroup of G generated by all pth powers of
elements of G. Show that G p is normal.
(f) Let G c be the commutator subgroup of G, defined as in Lang, p.20, top paragraph. Show that
G p G c is the intersection of all maximal proper subgroups of G. (This is called the Frattini subgroup
of G.)
(g) Show that if S is a subset of G, the following four conditions are equivalent:
(i) S generates G.
(ii) The image of S generates G G c .
(iii) The image of S generates G G p .
(iv) The image of S generates G G p G c .
(Can you generalize the argument by which you got (g) from (f) to a result about groups which are not
necessarily p-groups?)
I.6:4. [> I.5:1]. We shall determine all nonabelian groups of order 30.
(a) Prove (by a quick application of a result in the material on Structures of groups of order pq) that
every group of order 15 is cyclic (= I: L 6).
Now suppose G is a nonabelian group of order 30.
(b) Show by counting elements that G cannot simultaneously have more than one 3-Sylow subgroup and
more than one 5-Sylow subgroup.
Thus, G must have either a normal 3-Sylow subgroup or a normal 5-Sylow subgroup.
(c) Suppose G has a normal 3-Sylow (respectively, 5-Sylow) subgroup N. Show that in G N , a
5-Sylow (respectively, 3-Sylow) subgroup will be normal, and hence that its inverse image in G, which
we shall call M, will in turn be normal. Determine the structure of M.
(d) Show that in the above situation, G will be a semidirect product of M with a group of order 2.
(e) Express M as a direct product M = M1 M2 , and show that any automorphism of M must carry
M1 into M1 and M2 into M2 . Determine all automorphisms of M of exponent 2. Describe all
possible structures for G (as explicit semidirect products).
(f) Show that of the groups described above, one and only one is not a direct product of two groups of
smaller orders. In the remaining cases, describe the direct product decompositions.
I.6:5. (a) If G is a finite group, show that the following conditions are equivalent: (i) G cannot be
embedded in a direct product of groups of smaller orders. (ii) G has a smallest nontrivial normal
subgroup. (Idea: Consider maps G G N G N . Recall that smallest means contained in all
others.)
(b) [> I.6:4] Deduce that the one group G of I.6:4(f) that was not a direct product of groups of smaller
orders can be written as a subgroup of a direct product of groups of smaller orders. Describe these groups.
*I.6:6. Show that for every positive integer k there exists a positive integer N such that every finite
group of order N has an abelian subgroup of order k. (Suggestion: First do this for p-groups.)
*I.6:7. (a) Suppose as in I.4:4(c) that G is a group having proper subgroups H1 , H 2 and H 3 such
that every proper subgroup of G is contained in one of these. You analyzed in I.4:4(d) the case where G
was cyclic. Now show that if G is finite but not cyclic, then it is a 2-group. (Hint: Will a 2-Sylow
subgroup of G be proper?)
I do not know whether there are any infinite groups satisfying the above assumptions.
*(b) Show that a finite 2-group has the property of the first sentence of (a) if and only if it can be
generated by two, but not by fewer, elements.
(c) Given a group G and a positive integer n, examine the relationships among the three
conditions, (i) G has a family of n (not necessarily distinct) proper subgroups, such that every proper
subgroup of G is contained in at least one of these. (ii) G has n maximal proper subgroups. (iii) G
can be written as a union of n proper subgroups.
In particular, try to determine what implications hold among these three conditions for all n, giving
counterexamples to implications that do not always hold. Examine whether any natural restrictions, such
as finiteness, finite generation, or non-cyclicity, imply any of the implications that do not hold in general.
Some additional exercises that could be given in connection with I.6 are indicated in the paragraph
preceding Exercise I.9:1.
*I.7:1. (a) Show that the group Z of all sequences (a 0 , a1 , ... ) of integers, under componentwise
addition, is not free abelian on any generating set.
(b) Can you describe the group Hom(Z , Z) ?
I.7:2. (a) Show that if M is an abelian group, and we regard it as a commutative monoid and form
K(M ), then the canonical map : M K(M) is an isomorphism.

(b) On the other hand, show that if M is a commutative monoid containing an element z such that
a + z = z for all a M, then K(M) is the 1-element group. (Such an element z is called a zero
element of a monoid, because the element 0 has this property in the monoid formed by any ring under
multiplication, though the notation there is, of course, a0 = 0.)
I.7:3. Let A be a free abelian group, and x an element of A. Show that x is a member of a basis of
A if and only if it cannot be written ny for n >1 and y A.
(Hint for if: Suppose x is written i I ni ei where {ei ! i I } is a basis of A, and the ni are
integers. Consider the subgroup of Z generated by {ni ! i I }.)
I.7:4. Find an example of an abelian group A which is the sum, A = B + C, of two free abelian
subgroups, but which is not itself free abelian.
Of course, this means, necessarily, that this sum is not a direct sum. However, can you do this in such
a way that B and/or C is a direct summand in A ? (The condition that B is a direct summand means
that there exists a subgroup B such that A = B 2 + B ; and similarly for C .) In such a way that A is
torsion-free?
Note: Exercises III: L 5-III: L 8 can go with I.7, assuming a little background knowledge: For III: L 5 one
needs the concept of a basis of n-dimensional real Euclidean space R n, and for III: L 8, the arithmetic of
the rational numbers, and the notation Q* for the multiplicative group of nonzero rationals. Some
corrections to III: L 6, are noted after the exercises going with III.5 below.
I.8:1. Let G be the group of all sequences (a 0 , a1 , a 2 , ... ) with ai Z4 , under componentwise
addition, and let H be the subgroup of G consisting of those elements such that 2a 0 = 0. Show that
each of G and H can be embedded in the other, but that G and H are not isomorphic.
I.8:2. Suppose G is a group such that every finitely generated subgroup of G is cyclic.
(a) Show that if G is torsion-free, then G can be embedded in the additive group Q of rational
numbers.
*(b) Show that if G has torsion, then G can be embedded in the factor-group Q Z .
*I.8:3. Suppose p is a prime, and define Zp = (Q Z )( p), with the ( p) notation as in Lang (p.42,
paragraph before Theorem I.8.1). Show that Aut( Zp ) is uncountable.
I.8:4. [> A2.2:4(a), at end of these exercises].
(a) Show that the subdirectly irreducible finitely generated abelian groups are precisely the cyclic groups
of prime-power order.
(b) Show that the subdirectly irreducible finite p-groups are the p-groups whose centers are cyclic.
(c) [> I.8:2] Assuming the results of both parts of I.8:2, show that the subdirectly irreducible abelian groups
that are not finite are the groups Zp defined in I.8:3 above.
Remark on the next five exercises. We know from results in Lang that all finitely generated torsion
abelian groups are direct sums of finite cyclic groups Zp n , and we may ask whether something like this is
true without the finite generation hypothesis. The next four exercises address this question. Since the
groups Zp defined in Exercise I.8:3 above are not direct sums of proper subgroups, any positive result
along these lines will either have to allow these groups as summands along with the cyclic groups, or have
some hypothesis which excludes them.
Exercise I.8:5 will show that even allowing summands Zp , we do not get such a direct sum
decomposition for arbitrary torsion abelian groups. On the other hand, in Theorem I.8.c2 in this
Companion we got a generalization of the abovementioned result from Lang to the non-finitely-generated
case by strengthening the hypothesis of being torsion to a condition that in the finitely generated case is
equivalent to being torsion. Two other proofs of that same result are indicated in Exercises I.8:6 and I.8:7
below. Exercise I.8:8, on the other hand, will indicate a result for arbitrary torsion abelian groups,
obtained by weakening the conclusion of being a direct sum of cyclic subgroups to one that is equivalent,
in the finitely generated case, to that conclusion.
Exercise I.8:9 shows that yet another result true in the finitely generated case fails for non-finitely-
generated groups.
..
*I.8:5. Let p be a prime, and A the torsion subgroup of I I 1 i < Zp i . Show that A contains no
subgroup isomorphic to Zp , and is not a direct sum of cyclic subgroups.
I.8:6. (Alternative proof of Theorem I.8.c2, this Companion.)
In (a)-(c), let A be an abelian group of prime-power exponent p n . We shall write Ap for {a A !
pa = 0}, and for each i 0 we shall write p i A for the subgroup {p i a ! a A} A.
(a) Show that one can construct, successively, subsets Sn , Sn 1 , ..., S1 of Ap , such that for each i,
Si is a subset of Ap p i 1 A maximal for the condition that no s Si lies in the subgroup of A
generated by Sn Sn1 ... Si {s}.
In (b) and (c) below, S1 , ... , Sn denote sets with this property.
(b) Show that Ap is the direct sum of its cyclic subgroups s for s Si .
(c) Suppose we choose, for each element s of each Si , an element t A such that p i 1 t = s. (Why is
this possible?) Letting T1 , ..., Tn be the sets of elements so obtained, and T = Ti , show that A is
the direct sum of the cyclic subgroups t (t T ).
(d) Deduce that every abelian group of finite exponent n is a direct sum of cyclic subgroups.
*I.8:7. (Yet another proof of Theorem I.8.c2, with most of the work left to you this time.)
Suppose p is a prime, and A an abelian group all of whose elements have exponents a power of p.
Let us call a subgroup B A pure if B pA = pB.
(a) Show that the least integer i such that A has a nonzero pure subgroup of exponent p i is the same
as the least integer i such that A has a maximal cyclic subgroup of exponent p i .
(b) Show that for i as in part (a), any pure subgroup B A of exponent p i is a direct summand.
(Hint: Show that p i A B = {0}, and that p i A can be enlarged to the desired complementary
summand.)
(c) Deduce that every abelian group of finite exponent r is a direct sum of cyclic subgroups.
*I.8:8. [> I.8:7]. Let A be an abelian p-group. Show that, up to isomorphism, one can write
..
+ 1i
2 Ai A I I 1 i Ai ,
where for each i, Ai is a direct sum of copies of Zp i . In the case i = , Zp is defined as in I.8:3.
(Idea: Get decompositions A = A1 2 + B1 , B1 = A 2 2 + B 2 , etc., and show that such a chain of
decompositions leads to the situation of the above display, where A is the intersection of the Bi .)
I.8:9. Let A be an abelian group.
We have seen that if A is finitely generated, then it is the direct sum of its torsion subgroup Ator ,
and a free abelian subgroup B. In the non-finitely-generated case, a torsion-free abelian group need not be
free abelian, but we can still ask whether A will be the direct sum of Ator and a torsion-free subgroup
B. In (b) and (c) below we shall see two examples showing that this is not so in general (with a modified
argument for case (c) given in (d)). The criterion we shall use in both (b) and (c) is proved in (a). In
(e)-(f) we shall examine further properties of these examples.
(a) Show that if C is any subgroup of A, then it is a direct summand in A (i.e., there is a subgroup B
of A such that A = B 2 + C) if and only if the canonical surjection A A C has a right inverse which
is a group homomorphism, and that in this case, we will have B

= A C .
In the two examples below, you will be able to prove that Ator is not a direct summand by showing
that A A tor has properties that no subgroup of A has. To state these properties, let us say that an
element a of an abelian group (written additively) is divisible by an integer n if a = nx for some
group element x. Of the two examples, the first will be easier to state, but the second perhaps gives better
insight into why a direct sum decomposition is impossible.
(b) Let A be the direct product, over all primes p, of the cyclic groups Zp . Show that every element
of A A tor is divisible by all positive integers n, but that no nonzero member of A has this property.
Deduce that Ator is not a direct summand in A.
(c) Let p be any prime, and let Z[p 1] denote the abelian group of all rational numbers that can be
written with denominator a power of p. (This is ring-theoretic notation for the ring generated by
adjoining p 1 to the integers; but here we are using it as shorthand for the underlying additive abelian
group of that ring.) Let us form the factor-group Z[p 1] Z, and for each positive integer n, let
( p n Z) Z denote the cyclic subgroup thereof generated by [ p n ] (the image in this factor-group of
p n Z[p 1]).
The group Z[p 1] Z is what we have previously called Zp ; we are describing it this way because
we will want to use the canonical map Z[p 1] Z[p 1] Z, which we will write a [a]. Indeed, let
us define the subgroup
A Z[p 1] 2 + ( p 1 Z) Z 2
+ ( p 2 Z) Z 2
+ ( p 3 Z) Z 2
+ ... .
to consist of all elements (a, b , b , b , ... ) such that [a] = b + b + b + ... . Show that A A
=
1 2 3 1 2 3 tor
Z[p 1], so that, in particular, every element of this factor-group is divisible by all powers of p, but that
no element of A which maps to a noninteger element of Z[p 1] is divisible by all powers of p. Deduce
that Ator is not a direct summand in A.
(d) In example (c), determine all subgroups B A maximal for the property of being disjoint from Ator ,
and show directly that none of these satisfies A = B 2 + Ator .
We now look at a few related properties of these examples.
(e) In example (b), show that for every prime p, the p-torsion subgroup A( p) A is a direct summand
+ Bp . What is the intersection
in A; and in fact that there is a unique subgroup Bp such that A = A( p) 2
of all the subgroups Bp ?
(f) In example (c), let x 0 A be the element ([1], 0, 0, 0, ... ). Show that x 0 is divisible by p n for all
n, but that there do not exist elements x1 , x 2 , ... such that x 0 = px1 , x1 = px 2 , ... , xi = pxi+1 , ... .
Remark: The divisibility arguments of (b) and (c) above gave relatively simple ways of showing that
A A tor was not embeddable in A. But one can also get examples in which the one-variable equations
a = nx defining divisibility are replaced by systems of several linear equations in several variables. In
this way one can get examples where A A tor has no nonzero elements divisible by infinitely many
integers, but Ator is still not embeddable in A.
The following three exercises, though I am putting them under I.9, could be given any time after I.2,
referring to I.9 only for the definition of bilinear map in the middle of p.48. The first of these exercises
goes naturally with I.9, but the next two are appropriate any time after I.6 has been covered.
I.9:1. If A and B are abelian groups, we note that the set of group homomorphisms A B becomes
an abelian group under the operation defined by ( f +g)(a) = f (a) + g(a). Let us denote this group
Hom(A, B). For three abelian groups A, B and C, the set of bilinear maps A B C may, in the
same way, be made a group Bil(A, B, C). (Take these two facts for granted in your write-up.)
Show that given abelian groups A, B, C there exists an isomorphism : Bil(A, B, C)
Hom(A, Hom(B, C)) characterized by the formula ( )(a)(b) = (a, b).
I.9:2. (a) Suppose A and B are abelian groups (written additively), and : A A B any bilinear
map. Show that the set A B becomes a group G , in general nonabelian, under the operation
(x, y)(x , y) = (x +x , y +y + (x, x)).
(b) For any two elements x, y of a group G, let us write [x, y] for the commutator xyx 1 y 1. (The
commonest symbol for xyx 1 y 1 is (x, y), but since we are working with ordered pairs (x, y) A B, I
felt that the use of that notation for commutators could lead to confusion.) Show that the groups G
constructed in (a) all satisfy the identity [[x, y], z] = e.
(c) Suppose the abelian groups A and B of part (a) both have prime exponent p. Show that if p is
odd, then G will again have exponent p, while if p = 2, G may not have exponent 2, but will have
exponent 4.
(d) Deduce from (a) that if A, B, C are three abelian groups, and : A B C a bilinear map, then
the set A B C becomes a group H under the operation
(x, y, z) (x , y , z) = (x +x , y +y , z +z + (x , y)) (x, x A, y, y B, z, z C).
Deduce that these groups also have the properties stated in (b) and (c).
(e) Show that the group H described in (d) above can be written as a semidirect product A (B C)
and as a semidirect product B (A C). (If your proof is not based on an explicit description of the maps
A Aut(B C) and B Aut(A C) used in these semidirect product constructions, and the bijections
between the set of 3-tuples which form H and the sets of pairs which form these semidirect product
groups, you should determine retrospectively what these maps and bijections are. Since expressing H as
a semidirect product B (A C) involves changing the order of the factors A and B, and elements of
those factors do not commute with each other, the above bijection can be expected to be nontrivial in this
case.)
(f) On the other hand, give an example of a group of the form G as in (a), with A and B both
nontrivial, which is not a semidirect product of proper subgroups.
I.9:3. [> I.9:2(a), and (d) excluding last sentence]. In this exercise we will prove that for large n, there
are many groups of order p n , by a method similar to that used for monoids in Exercise I.1:2, but with
bilinear maps of finite products of Zp in place of set-maps of finite sets. Hence let us begin by counting
bilinear maps.
(The results you are asked to prove below are actually true without the assumption that p is prime; I
have included that assumption because p-groups form a natural interesting class of groups.)
(a) Suppose a, b and c are positive integers, and p a prime. In the abelian group (Zp ) a, let
x1 , ... , xa denote the elements (1, 0, ... , 0), ... , (0, 0, ... , 1); the analogous families of elements of (Zp ) b
and (Zp ) c will be denoted y1 , ... , yb and z1 , ... , zc . Show that every bilinear map : (Zp ) a (Zp ) b
(Zp ) c is determined by the ab elements (xi , yj ) (Zp ) c, and conversely, that each ab-tuple of values
in (Zp ) c can be realized as the images of this family of elements under a bilinear map. Thus, determine
the precise number of bilinear maps (Zp ) a (Zp ) b (Zp ) c. (We dont really need p to be prime here;
I am just assuming this because we will want it later.)
Now by I.9:2(d), each such bilinear map determines a group H with underlying set (Zp ) a
(Zp ) b (Zp ) c (which we may think of as (Zp ) a+b+c ). To complete our argument, we need to show
that not too many of these are isomorphic. It would be nice if we could say that any isomorphism between
two such groups was induced by automorphisms of (Zp ) a, (Zp ) b and (Zp ) c, but this is not so. Rather,
our trick will be to introduce some additional structure, which allows us to distinguish among the groups
determined by any two bilinear maps, then show that taking away this structure does not cause too many
groups to fall together.
Hence, if n is a positive integer, let an n-pointed group mean a system (G, u1 , ... , un ), where G is
a group, and u1 , ... , un are elements of G. A homomorphism of n-pointed groups (G, u1 , ... , un )
(H, 1 , ... , n ) will mean a homomorphism of underlying groups that takes ui to i for i = 1, ... , n.
Now if : (Zp ) a (Zp ) b (Zp ) c is a bilinear map, let H denote the a+b+c-pointed group whose
underlying group is H , defined as in Exercise I.9:2(d), and whose a+b+c distinguished elements are
(xi , 0, 0) (1 i a), (0, yj , 0) (1 j b), and (0, 0, zk ) (1 k c) (or, regarding the underlying set of
H as (Zp ) a+b+c , the a+b+c elements (1, 0, ... , 0), ... , (0, 0, ... , 1)).
(b) Show that if a, b, c are positive integers, and , are bilinear maps (Zp ) a (Zp ) b (Zp ) c,
then there exists a homomorphism H H as a+b+c-pointed groups if and only if = , and that in
this case, the only such homomorphism is the identity. (Warning: One may be tempted to call on the
result at the bottom of p.10 of Lang, and argue that a homomorphism H H which agrees with the
identity map of underlying sets on a generating set for H must agree with the identity map on all
elements. This is not valid, because that statement refers to uniqueness of homomorphisms, and till we
prove what we are trying to prove, we cannot say that the identity map of underlying sets is a
homomorphism H H . So you need to do some calculating.) Deduce a lower bound on the number
of nonisomorphic a+b+c-pointed groups of order p a+b+c, as a function of p, a, b and c.
(c) Determine the number of a+b+c-pointed group structures on a group of order p a+b+c. Comparing
this result with those of (a) and (b), get a lower bound on the number of nonisomorphic groups of order
p a+b+c.
(d) Find reasonably small integers M and N such that the above result implies that the number of
isomorphism classes of group of order p M approaches infinity as p , and that the number of
isomorphism classes of group of order p N is at least p 1000.
I.9:4. (a) Suppose A and B are abelian groups, and A 0 , A1 are subgroups of A and f 0 : A 0 B,
f1 : A1 B homomorphisms, such that for all x A 0 A1 one has f 0 (x) = f1 (x). Show that a
homomorphism f : A 0 + A1 B may be defined by setting f (x 0 + x1 ) = f 0 (x 0 ) + f1 (x1 ).
In parts (b) through (e) below, m will be a fixed positive integer, dual groups will be defined with
respect to a fixed cyclic group Zm , and A and B, where they occur, will denote abelian groups of
exponent m.
(b) Let A be a subgroup of A, and let A^. Show that there exists an element A^ whose
restriction to A is . (Hint: If you have extended to a subgroup A 0 A, and if a is an element
not in A 0 , let A1 = a , and set things up so that part (a) can be applied.)
(c) Show that A 0 A^ 0.
(d) If f : A B is a homomorphism, show that f is surjective if and only if f ^ is injective.
(e) If f : A B is a homomorphism, show that f is injective if and only if f ^ is surjective.
(Remark: If A and B are finite, then in view of Proposition I.9.c1, part (c) above is trivial, either
of (d), (e) is immediate from the other, and even (a) is fairly easy. But for general A and B, these
shortcuts are not available.)
(f) Show by example that the result of (a) fails if A and B are not assumed abelian. (Does it become
true if one or the other of those groups is assumed abelian?)
I.9:5. Show by example that if, in Theorem I.9.4, we delete the assumption that A B is finite, then the
conclusion that (A B )
= (A B )^ can be false.
I.9:6. Let (Ai ) i I be a family (in general infinite) of groups of exponent m.
..
(a) Show that ( 2 + Ai )^ = I I (Ai ^). ..
(b) Show by example that the relation ( I I Ai )^ = 2 + (Ai ^) does not in general hold.
(c) Show that of the two groups named in part (b), one can always be embedded in the other. (In fact,
you should be able to describe a map that always gives such an embedding.)
I.11:1. Let f : A B be a morphism in a category C. Then f is called a monomorphism if for every

object X of C and every pair of morphisms i, j : X A, we have i j fi fj. Dually, f is called
an epimorphism if for every object Y and pair of morphisms i, j : B Y, we have i j if jf.
(a) Show that in Set, the monomorphisms are the one-to-one set maps and the epimorphisms are the
surjective set-maps.
(b) Show that in Monoid, the monomorphisms are the one-to-one homomorphisms, and the class of
epimorphisms includes the surjective homomorphisms.
(c) Show that the inclusion-map of the monoid of nonnegative integers under addition into the monoid of
all integers under addition is an epimorphism in Monoid, though it is not surjective.
(d) Show that in any category C, if p : X Y and q : Y X are morphisms such that pq = id Y ,
then p is an epimorphism and q a monomorphism.
I.11:2. Let A and B be categories. Show that we can define a category B A whose objects are the
functors A B, and whose morphisms are the morphisms of functors (as defined on p.65). More
precisely, show that when we define composition of morphisms and identity morphisms of functors in the
obvious ways, the definition of a category is satisfied except for the axiom CAT1, and indicate briefly how
to fix this deficiency.
I.11:3. [> I.11:2]. Let .. denote the category having two objects, named 0 and 1, and three
morphisms: id0 , id1 and a morphism : 0 1. (You should verify for yourself that these conditions
determine an essentially unique category, but should not hand in the verification.) Thus, following the
notation of the preceding exercise, Group denotes the category whose objects are the functors from
. .
.. to Group, and whose morphisms are the morphisms of functors.
(a) Describe in group-theoretic terms the data that determine an object of Group (without referring to
. .
.., or to the concept of functor), and what constitutes a morphism between two such objects.
(b) Show that the constructions associating to every group homomorphism f : G H the group
Ker( f ) G and the group Im( f ) H can be regarded as functors Group Group.
. .
I.11:4. Suppose C is a category such that any two objects X and Y of C have a product X Y
in C.
(a) Show that any three objects X, Y, Z of C must then have a product, given by (X Y) Z.
(b) In this situation, show that (X Y) Z
= X (Y Z).
I.11:5. In a category C, let us define a retract of an object X to mean an object Y given with
morphisms f : X Y, g : Y X such that fg = id Y ; or, more precisely, a 3-tuple (Y, f, g) such that
Y, f and g satisfy these conditions.
(a) Show that if (Y1 , f1 , g1 ) and (Y2 , f2 , g 2 ) are retracts of the same object X, and if g1 f1 = g 2 f2 ,
then Y1
= Y2 in C.
(b) Show, in fact, that the isomorphism constructed in part (a) will make commuting triangles with f1
and f2 , and with g1 and g 2 .
(c) Show, conversely, that if two retracts (Y1 , f1 , g1 ) and (Y2 , f 2 , g 2 ) of an object X admit
isomorphisms making such commuting triangles, then g1 f1 = g 2 f 2 .
(d) Show that there exists a category R such that for every category C, the set of functors R C is
in natural bijective correspondence with the set of families X, Y, f, g such that (Y, f, g) is a retract of X.
I.11:6. (a) Show that if G is any monoid, then one can construct a category C G by taking for object-
set a singleton {X}, and defining Mor(X, X) = G, with the same composition as in the monoid G.
For the next two parts, define an action of a monoid G on an object Y of any category C to
mean a monoid homomorphism G Mor( Y, Y ).
(b) Verify that if G is a group and C = Set, this is equivalent to Langs definition of an action of G
on Y (I.5).
(c) Show that if G is a monoid and C a category, then an object of C with an action of G on it is
equivalent to a (covariant) functor from C G (defined in (a)) to C.
I.11:7. Let m be a positive integer, and let us write FinAb(m) for the category of finite abelian groups
of exponent m. Show that the duality h Zm : A A^ of I.9, regarded as a contravariant functor
FinAb(m) FinAb(m), takes short exact sequences to short exact sequences. (Cf. Corollary I.9.2.)
(Note our slight abuse of notation: strictly speaking, h Zm should denote a functor to Set; but we are
extending this symbol to mean the functor to FinAb(m) gotten by putting the natural abelian group
structures on these hom-sets.)
I.12:1. Let F = ..., y1 , y 0 , y1 , ... be the free group on a denumerable set of generators {yi ! i Z},
and let x be an infinite cyclic group (the free group on one generator).
(a) Show that there exists a unique homomorphism : x Aut(F ) such that (x)(yi ) = yi+1 for
each i Z.
(b) Show that the semidirect product xF is free on the two elements x, y 0 .

Remark: both parts, (a) and (b) above, can be done via messy calculations, or elegantly, with the help
of the universal property of free groups!
I.12:2. (a) Show that the following four groups are isomorphic. (It is up to you to discover the
isomorphisms. I said in the discussion of semidirect products that the group described in (iii) is called the
infinite dihedral group. Actually, since the groups described below are isomorphic, that term may be
defined in any of these four ways.)
(i) The group of maps Ta, b : Z Z, where Ta, b (n) = an+b, for a {1}, b Z. (This is the set of
all distance-preserving set-maps Z Z, but I dont ask you to prove that.)
(ii) The group with presentation x, y ! x 2 = e, yx = xy 1 .
(iii) The group Z 2 Z, where : Z 2 Aut(Z) sends the nonidentity element of Z 2 to the
automorphism n n of Z.
(iv) The coproduct of two copies of the group Z 2 . (Write these two copies {e, u} and {e, }, for
uniformity of notation.)
(b) Show that in the group (iv) above, every cyclic subgroup generated by an element of the form (u ) c
is normal, and that with three exceptions, every normal subgroup of this group has this form.
I.12:3. Let F be the free group on a denumerable set {x1 , ... , xn , ...}. Given an element w F, a group
G is said to satisfy the identity w = e if every homomorphism F G carries w to e. A class V of
groups is called a variety of groups if there exists some S F such that V is the class of groups
satisfying the identities w = e for all w S.
(a) Verify that the following classes of groups are varieties: (i) the class of all groups, (ii) the class of all
abelian groups, (iii) the class of all groups of exponent 50, (iv) the class of all abelian groups of exponent
50, (v) the class of all trivial groups (i.e., up to isomorphism, the trivial group).
In one of cases (ii)-(iv) above, show explicitly how the definition given above of what it means for a
group to satisfy an identity (via homomorphisms F G), when applied to the identities you give, is
equivalent to the ordinary definition of the class of groups named. In the remaining cases, you may simply
display an S which obviously works in the same way.
(b) Show that for each positive integer n, the class Solv n of all groups G admitting a normal tower of
height n with abelian factors:
G = G0 ... Gn = {e} with Gi G i+1 abelian,
is a variety.
(c) Suppose X is any set, F(X ) the free group on X, and w F (X ). Let us define the condition that a
group G satisfies the identity w = e exactly as in the case X = {x1 , ... , xn , ...} above. Show that for
every w F (X ) there exists some w F({x1 , ... , xn , ...}) such that a group G satisfies the identity
w = e if and only if it satisfies the identity w = e.
(d) Let w F (where F is again the free group on {x1 , ... , xi , ...}). Show that the following conditions
are equivalent: (i) The variety determined by the identity w = e consists (up to isomorphism) of the
trivial group only. (ii) There exists a group homomorphism F Z taking w to 1. (iii) If we denote
by di the exponent-sum of xi in w, i.e., the sum of all exponents to which xi occurs, counting
signs, in the expression for w, then the greatest common divisor of {d1 , ... , di , ...} is 1.
(This shows that if you write down an identity at random, it is likely not to define an interesting
variety.)
I.12:4. [> I.12:3(a)]. (a) Show that for every class C of groups, there exists a least variety V of groups
containing C.
The next four parts will show how to construct the groups in V from those in C.
If C is a class of groups, let us define H(C) to mean the class of homomorphic images of groups in
C, S(C) to mean the class of groups embeddable in groups in C (i.e., groups isomorphic to subgroups
..
of members of C) and P(C) to mean the class of groups isomorphic to products I I I Gi of arbitrary
families (Gi ) i I of groups Gi C (possibly with repeated factors).
(b) Show that every variety V of groups is closed under H, S and P (i.e., that V = H(V) = S(V) =
P(V)).
(c) Show that if C is any class of groups, then the class HSP(C) is closed under H, S and P.
(d) If C is a class of groups, R a member of C, X a set, and f : X R a map, let us say that R is
free on X relative to C if the condition defining a free group (Lang, p.66) is satisfied with group
replaced by member of C throughout. Show that if C is any class of groups closed under S and P,
then the method of proof of Proposition I.12.1 may be generalized to show that for every set X there
exists a group R in C that is free on X relative to C. Do not repeat that proof just indicate what
observations are used in adapting it.
(As noted in the discussion in this Companion regarding p.58 of Lang, there are various ways of talking
about free groups: as pairs (F, f ), as universal maps f alone, or as objects F with the map f
thought of as given. In parts (d)-(f) of this exercise, I am following the last of these.)
(e) Let X be a set, V a variety of groups, F(X ) the free group on X, and FV (X ) the free group on
X relative to V. Show that there is a canonical surjective homomorphism q : F(X ) FV (X ), and show
that for any w F(X ), the identity w = e holds in V if and only if q(w) = e in FV (X ).
(f) (Birkhoffs Theorem, for the case of groups) Show with the help of the preceding results that for any
class C of groups, HSP(C) is the least variety of groups containing C. (Hint: We did not use H
in part (d).) Deduce in turn that a class of groups is a variety if and only if it is closed under H, S
and P.
I.12:5. Show that the free abelian group on generators x1 , ... , xn may be presented as a group by the n
generators x1 , ... , xn , and the n(n 1) 2 relations xi xj = xj xi (1 i < j n).
*I.12:6. [cf. I.12:5.] Show that no presentation of the free abelian group on n generators as a group can
involve fewer than n(n 1) 2 relations.
(Suggestion: First try to show that if F is the free group on generators x1 , ... , xn , then the
commutator subgroup of F cannot be generated as a normal subgroup of F by fewer than n(n 1) 2
elements. Then try to get the above statement from this.)
I.12:7. Lang asserts in the first Example on p.69 that in the group presented by three generators x, y, z
and the three relations
(1) [x, y] = y, [y, z] = z, [z, x] = x,
one has x = y = z = e. Clearly, this is equivalent to saying that in any group with three elements x, y,
z satisfying the above relations, x = y = z = e holds.
Below, we shall prove this assertion, obtaining in the process some results about groups satisfying more
general systems of relations. At one point we will call on a number-theoretic fact that uses an exercise
from II.1, but as we will note at the end, the case of this fact needed for the group described above can be
checked by arithmetic. The exercise after this one outlines an alternative proof of Langs assertion, with
more work left to the reader.
(a) Verify that elements x and y of an arbitrary group satisfy
(2) [x, y] = y
if and only if they satisfy
(3) x y = y 2 x,
equivalently, if and only if they satisfy
(4) x y x 1 = y 2.
(b) Suppose k is a positive integer, and G is a group containing elements x1 , ... , xk satisfying the
cyclic family of relations
(5) [x1 , x 2 ] = x 2 , [x 2 , x 3 ] = x 3 , ..., [xk1 , xk ] = xk , [xk , x1 ] = x1 .
Show that if one of the xi is of finite order, then all of them are of finite order. (Idea: If xim = e,
then conjugating xi+1 by xi m times must leave it fixed; but m applications of (4) gives a different
answer.)
(c) Assuming x1 , ... , xk in (b) all have finite order, let n > 1 be a common exponent for these elements.
By the same method used in (b), show that 2 n 1 is also a common exponent for these elements. But
exercise II.1:2 shows that 2 n 1 is not divisible by n. Deduce that these elements have a common
exponent smaller than n, and argue from this that they are all equal to e.
We can push this further ...
(d) Deduce that if a group contains elements satisfying (5), and a relation xiax bi+1 = e holds with the
integers a and b not both zero, then all xj equal e. (Suggestion: Conjugate the given relation by xi
and compare the result with the original.)
... and still further ...
(e) Deduce that if a group contains elements satisfying (5), and a relation xia x bi+1 x ci+2 = e holds with a
and b both positive, then all xj equal e. (Suggestion: Consider two ways of transforming the given
relation. On the one hand, apply (3), with xi+2 in the role of y, b times, to push the third term to the
left past the second. On the other hand, apply (3), with xi+1 in the role of y, a times, to push the
second term to the left past the first. Then use the general fact that in a group, u = e u = e, to
rearrange the result of the latter calculation so that the unchanged term xia appears on the far left; equate
this with the result of the former calculation, cancel xia , and apply (d).)
We now come to the messy part: Showing that when k = 3, i.e., when (5) has the form (1), we
automatically have a relation of the sort assumed in (e).
(f) Assuming (1), multiply the relation (3) on the left by z 4. Show with the help of (3), adjusted by cyclic
permutations of x, y and z, that in the left-hand side of the resulting equation, you can convert a factor
z 2 x to x 4 z 2, then a factor z 2 y to y z, then a factor x 2 y to y 4 x 2. Likewise show that on the right
you can convert a factor z 4 y 2 to y 2 z, and then a factor z x to x 2 z. Equating the resulting expressions,
deduce that z 2 x 2 y 2 = e, and complete the proof of Langs assertion.
Remarks: When the argument of (b)-(e) is applied to the above relation x 2 y 2 z 2 = e, we find that the
m of the case of (c) that arises is 6, and the only cases of II.1:2 that are needed are gcd(6, 2 6 1) = 3,
and gcd(3, 2 3 1) = 1, which, of course, are verifiable without appealing to II.1:2.
For k > 3 the group presented by (5) can be shown to be nontrivial. By (c), this gives a nontrivial
finitely presented group having no nontrivial finite homomorphic images.
I.12:8. This is an alternative proof of the assertion of Langs proved in the preceding exercise. The part
which we leave to your ingenuity is
*(a) Suppose a group G contains elements x and y satisfying [x, y] = y. Show that x, y is solvable.
For the remainder of the proof, suppose G is generated by three elements x, y, z satisfying
[x, y] = y, [y, z] = z, and [z, x] = x.
(b) Show from these relations that G = [G, G ]. (Group-theorists call a group satisfying this condition
perfect.)
(c) Also show from the given relations that the subgroup x, y contains the elements z x z 1, z 2 x z 2,
and z 2 y z 1.
(d) Verify that the product
(z 2 y z 1) 1 (z 2 x z 2) (z 2 y z 1) (z x z 1) 1 (z 2 y z 1) 1
simplifies formally (without the use of the relations assumed) to z (y 1 x y x 1 y 1)z 2 = z (y 1[x, y])z 2.
Conclude with the help of one of the given relations that z x, y , and deduce that x, y = G. Thus, by
the result of (a) and this result, G is both perfect and solvable. However
(e) Show that the only group that is both perfect and solvable is the trivial group.
II.1:1. (a) State and prove an analog of Goursats Lemma (I: L 5, p.75) for rings.
(b) Deduce a description of all subrings of the ring Z Z. (State your description in a form that is as
simple and concrete as possible for this specific case. Recall that by ring we mean ring with 1, and
by subring we mean subring containing the same 1.)
II.1:2. Show that if n > 1, then n is not a divisor of 2 n1. (Hint: If n is odd, let p be the smallest
prime dividing n, and show that the order of the image of 2 in the multiplicative group of Z pZ is not
a divisor of n. What does this imply about divisibility properties of 2 n1 ?)
Note on II: L 12 (p.116): When Lang says in part (c) Let be the Mobius function such that ..., he
means Let be the Mobius function, i.e., the function defined by the equations .... (I do not
recommend using such that in this way in your writing!)
The function so defined is a member of R, i.e., a map from P to K. In its definition, (1) r
should thus be understood as the identity element 1 K.
Lang asserts in part (c) that the result proved there gives the Mobius inversion formula. That formula
is given as exercise V: L 21 (p.254), which can be done right now, as an application of II: L 12(c). (Lang has
it in Chapter V because he wants to apply it in V: L 22 to finite fields.) In doing it, you may assume that
the abelian group A is the additive group of a ring. (In fact, any abelian group can be embedded in the
additive group of a ring.)
Note on II: L 13-19 (p.116), on Dedekind rings (defined on p.88): In the last line of the page, Lang points
to VII: L 7 for a continuation. There is also related material in III: L 11-13.
II.2:1. Let R be a commutative ring. By a submonoid of R we shall understand a subset containing 1
and closed under multiplication. (Another term sometimes used is multiplicative subset.) We use the
word complement below in its set-theoretic sense.
(a) Show that the following five statements are equivalent.
(i) If I is an ideal of R, and S is maximal among submonoids of R disjoint from I, then the
complement of S in R is an ideal.
(ii) If S is a submonoid of R, and I is maximal among ideals of R disjoint from S, then the
complement of I in R is a submonoid.
(iii) If I and S are an ideal and a submonoid of R, maximal for the property of being disjoint, then
I S = R. (Here maximal means among ordered pairs consisting of a disjoint ideal and submonoid of
R, with respect to the partial ordering on such pairs that makes (I1 , S1 ) (I2 , S 2 ) if and only if
I1 I2 and S1 S 2 .)
(iv) If I and S are an ideal and a submonoid of R respectively, and are disjoint, then there exists
an ideal I I whose complement is a submonoid S S.
(v) If I and S are an ideal and a submonoid of R respectively, and are disjoint, then for every
r R, either the ideal generated by I and r is disjoint from S, or the submonoid generated by S and
r is disjoint from I.
(b) Prove (v) above, and hence all of the statements (i)-(v).
(Statement (ii) is II: L 1, incidentally.)
(c) Show that an ideal of R is prime if and only if its complement is a submonoid of R.
Statement (iv) above, interpreted in the light of (c), is a very useful tool for constructing prime ideals.
Here is an important application:
(d) Show that if I is any ideal of R, then the intersection of all prime ideals containing I is {r R !
( n 0) r n I }. (Hint: For every r not in the above set, you want to find an appropriate prime ideal that
does not contain it.)
II.2:2. [> II.2:1]. Suppose R is a commutative ring, and X R a set having nonempty intersection with
every nonzero prime ideal of R. Let X denote the multiplicative submonoid of R generated by X.
Show that for every nonzero element r R, the set X contains some multiple of r.
II.2:3. [> II.2:1]. We know that an integral domain can be described as a commutative ring R in which
{0} is a prime ideal, and that in this situation, {0} is the least prime ideal of R. If R is a commutative
ring with zero-divisors, we can ask whether it still has a least prime ideal, and if not, whether it has
minimal prime ideals (i.e., prime ideals not containing any smaller prime ideals).
(a) Show by example that a commutative ring need not have a least prime ideal.
(b) Prove that a nonzero commutative ring always has a minimal prime ideal. (Hint: Use I.2:1 to turn the
question into one about the existence of a maximal set of some sort, and use Zorns Lemma to prove the
necessary existence statement.)
(c) Show that a prime ideal P R is minimal (among prime ideals) if and only if for every a P there
exists b R P and an integer n > 0 such that a n b = 0.
II.3:1. (a) = IV: L 8 (p.214). (Though Lang puts this in Chapter IV, it only requires the concepts of II.3.
The assumption that A be entire is not needed in this part, but will be needed in part (c) below.)
(b) Denoting the automorphism of A[X ] that carries X to aX + b by g(a, b), determine the law of
composition of these automorphisms, and show that they form a group, which is a semidirect product of the
additive group of A and the multiplicative group of units of A.
(c) = IV: L 9 (p.214), but adding after automorphism of A[X ] the words inducing the identity on A.
(If you have seen III.1, this means we are looking at automorphisms as an A-algebra.)
(d) Show that if A is a field, then every endomorphism of A[X ] as a ring carries A into itself, and
that every automorphism of A[X ] as a ring carries A onto itself.
(e) Show that if A is a field, the group of all automorphisms of A[X ] as a ring is a semidirect product
of the group of automorphisms of the field A, and the group of A-algebra automorphisms of A[X],
described in (a-c) above. What is the action of the former group on the latter?
II.4:1. Prove the assertion in the second paragraph of p.90 of Lang, that every field of characteristic 0
contains a subfield isomorphic to Q. (You may use the results of II.4, and the fact that the field of
fractions of Z is Q.)
II.4:2. (a) Show that a commutative ring is local if and only if its noninvertible elements form an ideal.
(b) = II: L 3 (p.115).
II.4:3. [> III.2:1-2]. This exercise shows how to define the concept of not necessarily commutative local
ring. Let A be a ring.
(a) Show that the following conditions are equivalent:
(ileft ) A has a unique maximal left ideal.
(iright ) A has a unique maximal right ideal.
(ii) The set of noninvertible elements of A forms a 2-sided ideal.
(Suggestion: Translate (ileft ) into a statement about left annihilators {a A ! ax = 0} of elements x of
simple left A-modules.)
A ring satisfying the above equivalent conditions is said to be local. There are some more equivalent
conditions, which for completeness I give below.
(b) Show that the conditions of part (a) are also equivalent to each of the following:
(iiileft ) The set of non-left-invertible elements of A is closed under addition.
(iiiright ) The set of non-right-invertible elements of A is closed under addition.
(iv) The set of noninvertible elements of A is closed under addition.
(Hint: Show that a ring satisfying (iv) cannot contain any idempotent elements, i.e., elements e satisfying
e 2 = e, other than 0 and 1, but that a ring with right-invertible but not left-invertible elements does
have such idempotents.)
In an earlier version of this Companion, I included another condition which I thought was equivalent to
the above, but turned out not to be:
(c) Show that conditions (i)-(iv) above are not equivalent to:
(v) A has a unique maximal 2-sided ideal m, and A m is a division ring.
(If you havent done all of parts (a) and (b), you can still do this part by obtaining a ring for which the
truth-value of (v) is different from at least one of the other conditions listed.)
II.4:4. In the last paragraph of p.110, Lang gives three equations which say that the operation of
localizing ideals respects the operations of adding, multiplying, and intersecting ideals. More generally,
given any homomorphism of commutative rings, f : A B, let us, as Lang does there, write J (A ) for
the set of all ideals of A, and let us define f* : J (A ) J (B ) to take each ideal a J (A ) to the ideal
B f (a) J (B ), and f * : J (B ) J (A ) to take a J (B ) to f 1(a) J (A ).
(a) Of the three equations Lang proves for the operation S 1(), determine which hold for the operation
f* for every homomorphism f of commutative rings. Prove the equations that hold, and give
counterexamples for those that dont.
(b) Do the same for the operation f *.
(c) Show that the operation S 1() treated by Lang is in fact f* for f the canonical map A S 1 A.
Assuming the results that you proved in (a) above, is the one verification given by Lang at the end of that
paragraph sufficient to establish the three equations he shows?
(d) Again taking for f the canonical map A S 1 A, does the map f * satisfy some or all of the same
three equations?
(e) (Open-ended) What properties can you prove for the composite maps f * f* and f* f *, either in
general, or for the specific cases where f is the canonical map A S 1 A ?
II.5:1. Let k be any field, and k[x] the ring of polynomials over k in one indeterminate x. Let
k[x 2, x 3] denote the subring of k[x] consisting of those polynomials ai x i such that a1 = 0, i.e.,
polynomials with zero linear term. (The symbol k[x 2, x 3] means that this subring is generated over k
by the two elements x 2 and x 3, but I do not ask you to verify this fact.)
(a) Establish necessary and sufficient conditions on nonnegative integers i, j 1 for x i to divide x j in
k[x 2, x 3]. Show by example that this is not equivalent to the condition that x i divide x j in k[x].
(b) Show that the elements x 2 and x 3 have a greatest common divisor in k[x 2, x 3], but not a least
common multiple. (You may take it for granted that any divisor of an element x j is, up to multiplication
by a unit, of the form x i . Remember that least and greatest refer here to divisibility in this ring, not
in k[x].)
(c) Suppose a and b are any nonzero elements of any integral domain R. Show that if u and are
elements of R such that ab = u , then u is a common divisor of a and b if and only if is a
common multiple of a and b. Show moreover that if u and are two other elements such that
ab = u , then u is a multiple of u if and only if is a multiple of . Show, finally, that a least
common multiple of a and b, if it exists, will be a divisor of ab, and that the same is true of a greatest
common divisor of a and b if it exists. (In proving these statements, note that you may not assume that
elements of the ring R have unique factorization, or even admit any factorization into irreducibles.)
(d) The results of (c) appear to imply that two nonzero elements a and b of an integral domain should
have a least common multiple if and only if they have a greatest common divisor. But this contradicts the
example of (b)! Find the fallacy; and verify that one direction of this if and only if is valid.
(e) Find two elements of k[x 2, x 3] that have no greatest common divisor. (Suggestion: two common
multiples of x 2 and x 3.)
II.5:2. Let R be a commutative integral domain. By a prime element of R we will mean a nonzero
element p such that pR is a prime ideal. Show that the following conditions are equivalent:
(i) R is a unique factorization domain.
(ii) Every irreducible element of R is prime, and there is no infinite strictly increasing chain of principal
ideals, Ra1
Ra 2 ... .
(iii) Every nonzero nonunit in R is a product of prime elements.
II.5:3. [> II.2:1, II.5:2] In the context of II.5:2, show that conditions (i)-(iii) of that exercise are also
equivalent to each of:
(iv) Every nonzero prime ideal of R contains a nonzero principal prime ideal.
(v) R has no infinite strictly increasing chain of principal ideals, Ra1
Ra 2 ... , and satisfies the
factorization refinement property: For any relation ab = cd holding among nonzero elements of R,
there exist elements w, x, y, z R such that a = wx, b = yz, c = wz, d = xy.
Suggestions for proving that (iv) implies one of the other conditions: Use II.2:2 with X the set of
nonzero prime elements of R, or II.2:1 with S the monoid generated by these elements. You may also
want to review how Lang proves that a PID (in his language, a principal ring) is a UFD (a factorial
ring), though he doesnt do all parts of that proof in the smoothest possible way.
*II.5:4. Find an example of a commutative ring R with elements x and y such that xR = yR, but such
that there does not exist a unit u R such that y = ux.
II.5:5. Let A be an integral domain and K its field of fractions.
(a) Show that if A is a PID, then every subring R of K containing A is a localization S 1 A for
some multiplicative set S.

(b) Show that if A is a UFD such that every subring R of K containing A is a localization S 1 A for
some multiplicative set S, then A is a PID.
Clarification of II: L 5 (p.115): Change with primes p to where p is a prime, and units u to
where u is a unit.
II.5:6 [> II: L 4, II: L 10]. Let R be the ring of trigonometric polynomials, defined in the last paragraph
before the exercises on p.114, and studied in II.5:10. Let R [i] denote the ring of complex-valued
functions on the real line of the form u + i with u, R.
(a) Show that R [i] is generated over C by the function e ix and its multiplicative inverse.
(b) Show that R [i] is a localization of a polynomial ring, and deduce that it is a principal ideal domain.
For the remainder of this exercise, let I 0 R denote the set of trigonometric polynomials f R that
satisfy f (0) = 0.
(c) Show that I 0 is a non-principal ideal of R.
Thus, R is not a principal ideal domain. However, the remaining parts will show that it has properties
very close to those of a principal ideal domain.
(d) Show that for every ideal J R, there exists a unique ideal J R [i] closed under complex
conjugation and satisfying J R = J. (Remark: This will of course remain true if we delete the
closed under conjugation condition; but then uniqueness can fail. You might look for an example of
this.)
(e) If we apply part (d) to the ideal I 0 , the corresponding ideal of R [i] must be principal. Find a
generator (or better: all possible generators) of this ideal. Why do these elements fail to give generators of
I 0 over R, contradicting (c)?
(f) Let J be a nonzero ideal of R, J the ideal of R [i] described in (d), and f R [i] a generator of
J . Let n be the number of zeroes of f on the interval [0, 2 ), counting multiplicity. Show that the
ideal J is principal if and only if n is even.
(g) Deduce that for every nonzero ideal J R, exactly one of the ideals J and I 0 J is principal.
Remark: From (g), one can deduce that R is a Dedekind domain, a concept briefly mentioned by
Lang under the name Dedekind ring on p.88, and studied further in Exercises II: L 11-19 and other
exercises in later chapters. The idea that principal ideals of R are to all ideals as even integers are to all
integers, suggested by (g), can be made precise by deducing from (g) that the ideal class group of R (also
defined on p.88) is isomorphic to Z 2 .
Historically, Dedekind domains arose in number theory, and the examples that specialists are most
familiar with come from that subject. I think they would consider this one a somewhat exotic example.
Notes on exercises in Lang: II: L 9 belongs under Chapter IV. It is essentially worked in this Companion,
immediately following Theorem IV.1.c2.
Hint to II: L 10(c): If a + b 0( D ) is a unit of R, consider the product of this element with its
complex conjugate.
III.1:1. [> II.4]. As noted in the remarks on this section, we will call a morphism f : A B in a
category an epimorphism if for any object C and any morphisms g, g: B C, one has gf = g f
g = g, rather than using the meaning Lang gives this term.
(a) Show that in the category of left modules over a ring R, the epimorphisms are precisely the surjective
homomorphisms. (Hint: If f is not surjective, take for g and g a zero map and a nonzero map into
an appropriate object.)
(b) Show, on the other hand, that if R is a commutative ring and S a multiplicative system in R, then
the natural map R S 1 R is an epimorphism in the category of commutative rings, but that such maps
are not in general surjective.
III.1:2. In a ring R, the centralizer of a subset X is defined (as in a group or semigroup) to be {r R !
( x X ), rx = xr}.
(a) Show that the centralizer of any subset of a ring is a subring.
(b) Suppose R is a ring, M an R-module, EndZ (M) the ring of abelian group endomorphisms of M,
and End R (M) the ring of R-module endomorphisms of M. Recall that the R-module structure of M is
given by a homomorphism : R EndZ (M). Show that End R (M) is the centralizer of (R) in
EndZ (M).
(If you have read as far as my comment to p.124 of Lang, to the effect that many noncommutative
ring-theorists follow the convention that endomorphisms of right modules are written to the left of module
elements and vice versa, note that we are not following this convention in this exercise since we are here
looking at ring-element actions and module-endomorphisms as members of a common ring EndZ (M)!)
The remaining point is an example of a recurring mathematical pattern, worth noting.
(c) Let R be a ring, and for every subset X R, let us abbreviate the centralizer of X in R by X*.
Show that for every X, we have X ** X and X*** = X*.
(When R is the ring of endomorphisms of an abelian group, and X is a subring thereof, so that we
are in the situation of the preceding parts, then X * is called the commutant of X, and X ** the
bicommutant.)
III.1:3. In the Sketch of Proof of Lemma III.1.c1 in this Companion, the words It is easy to verify
occur three times. Give these three verifications.
Note on III: L 10 (p.168, based on III.1) [> III: L 9, p.167]. Suggested plan of attack: Given a
multiplicative set S, determine what elements of M go to zero under the natural map M S 1 M.
Hence, given a prime ideal p, determine what elements go to zero under M Mp . Then show that for
every nonzero x M, there is a maximal ideal p which does not make x go to zero.
III.1:4. Let A be a ring. In the next-to-last paragraph of p.118, Lang notes that if A is an integral
domain and M an A-module, then the set of elements x M whose annihilator, ann A x = {a A |
rx = 0}, is nonzero, forms a submodule Mtor of M. However, the arguments showing this fail if either
the assumption that A is commutative or the assumption that it has no zero-divisors is deleted.
However, for general A, there may still be one or more ways of describing a submodule of each
A-module consisting of all elements whose annihilators have certain properties.
So determine necessary and sufficient conditions on a set X of left ideals of A for it to be true that
for every left A-module M, the set of elements of M whose annihilators belong to X forms a
submodule MX of M.
III.2:1. A module over a ring R is called simple if it is nonzero, but has no nonzero proper submodules.
Show that if M is a simple R-module, then End R (M ), the ring of endomorphisms of M as an
R-module, is a division ring.
III.2:2. For any element x of a left R-module M, one defines the annihilator of x by
ann R x = {r R ! rx = 0}.
(a) Show that in this situation, ann R x is a left ideal of R, and construct a module isomorphism
between the submodule Rx M generated by x, and the left R-module R ann R x. (Thus, I am asking
you to verify some things I asserted without proof in a comment to p.118 of Lang in this Companion.)
A left module over a ring R is called cyclic if it is generated by a single element. The above result
says that the structure of a cyclic R-module is determined by the annihilator of any generator.
(b) Let Rx and Ry be cyclic R-modules, with ann R x = I and ann R y = J. Clearly, a homomorphism
Rx Ry takes x to an element of the form ay, and is therefore determined by the element a.
Characterize, in terms of the left ideals I and J, those elements a R such that there exists a
homomorphism Rx Ry taking x to ay, and determine when the homomorphisms corresponding to
two such elements a and b are the same. Thus, get a description of the abelian group Hom(Rx, Ry).
(c) Obtain from the above result a description of the endomorphism ring End R (Rx) of a cyclic R-module
Rx in terms of I = ann R x. In this case, you must, of course, determine its multiplication as well as its
additive group structure.
(d) [cf. III.2:1] Show that the simple R-modules are precisely the cyclic R-modules Rx such that ann R x
is a maximal left ideal of R.
(e) In the case where R is commutative, state the forms that the results of (b) and (c) above take, and
with the help of (d) above, describe the division rings that occur as in III.2:1.
III.2:3. [cf. III.2:1]. Part (a) below gives a class of examples showing that the converse to III.2:1 is not
true. Parts (b) and (c) are more difficult questions that this result suggests.
(a) Show that if R is a commutative integral domain which is not a field, and K is its field of fractions,
then the underlying R-module of K is not simple, but its ring of R-module endomorphisms is isomorphic
to the field K.
*(b) Find a finitely generated module M over a ring R such that End R (M ) is a division ring (possibly a
field), but such that M is not simple.
To state the last part of this exercise, let us generalize the construction of part (a) as follows. Suppose
f : A B is any homomorphism of rings, and let us regard B as a left A-module by defining left
multiplication by a A to take x B to f (a)x. Then we see that for every b B, the map x xb
gives an endomorphism of B as left A-module. For convenience, let us write left A-module
endomorphisms of B on the right of the elements they are applied to, and compose them accordingly.
Then the above map is a ring homomorphism B End A ( B). This is clearly one-to-one, but let us ask,
open-endedly,
*(c) For what ring homomorphisms f : A B is the homomorphism B End A ( B) described above an
isomorphism?
III.3:1. Let R be any ring, and 0 L M N 0 a short exact sequence of left R-modules. Show
that the following five conditions are equivalent:
(i) For every left R-module A, the induced sequence 0 Hom(A, L) Hom(A, M)
Hom(A, N) 0 is exact.
(i ) The induced map Hom(N, M) Hom(N, N) carries some element of the former abelian group to
the element id N of the latter abelian group.
(ii) For every left R-module A, the induced sequence 0 Hom(L, A) Hom(M, A)
Hom(N, A) 0 is exact.
(ii ) The induced map Hom(L, L) Hom(M, L) carries some element of the latter abelian group to
the element id L of the former abelian group.
(iii) The original exact sequence of R-modules splits (i.e., satisfies the equivalent conditions of
Proposition III.3.2, p.132. Cf. the discussion of splitting in the context of abelian groups in this
Companion, after the notes on the proof of Lemma I.7.2.)
III.3:2. Let A be a ring.
(a) Show that if an A-module M is the direct sum of submodules M1 , ..., Mn , then its A-module
structure can be extended to a structure of (A ... A )-module (direct product ring with n factors) in such
a way that the element of this ring having a 1 in the ith position and zeros elsewhere acts as the projection
onto the summand Mi ; and conversely, that every (A ... A )-module arises in this way from a direct
sum decomposition.
(b) Show that if I is an infinite set, and we denote by A I the direct product of an I-tuple of copies of
A, then any A-module with a decomposition as a direct sum of an I-tuple of submodules can be given an
A I -module structure in a way analogous to the above, and that the same is true of any A-module with an
expression as a direct product of an I-tuple of modules, but that there are also A I -modules that do not arise
in either of these ways.
III.4:1. Let R be a nonzero ring, M a left R-module, and X a subset of M, and let RX denote the
submodule of M generated by X. Show that X is contained in a basis of M if and only if X is
linearly independent and the factor-module M RX is free.
*III.4:2. Let R be the ring of all continuous real-valued functions on the interval [0,1], and consider the
ideal I = { f R ! ( >0) ( x [0, ]) f (x) = 0}. Show that as an R-module, I is projective.
Note on III: L 5: In giving a proof that follows Langs sketch, you need to explain his we may assume
step; i.e., to say how, given 1 , ..., m , and assuming his inductive hypothesis, you can modify 1 , ...,
m to get a family with the property that he says may be assumed.
III.5:1. We noted in this Companion that if R is a ring, n a positive integer, and A the ring of n n
matrices over R, and we regard the set of column vectors of height n over R as a left A-module P,
then A is isomorphic as a left A-module to the direct sum of n copies of P.
Now suppose that R is a field, or more generally, a division ring, let n and A be as above, and let
M be any left A-module. Let us write eij for the usual matrix units in A.
(a) Show that e11 M, i.e., {e11 x ! x M }, may be regarded as a left R-vector-space.
(b) Let X be any basis of the above vector space. Show that for each x X, the left A-module Ax is
isomorphic to P, and that M is the direct sum over X of these modules.
Thus, every A-module is projective, and is a direct sum of copies of P. (A more general result of this
sort is developed in Exercise XIII.3:1.)
III.5:2. (a) Show that for any positive integer n and infinite field k, the following statements are
equivalent.
(i) An affirmative answer to Open Question III.5.c4 (a).
(ii) An affirmative answer to the same question with the assumption that V is n-dimensional replaced by
the statement that it has dimension at least n, and bases replaced by linearly independent subsets.
(iii) As in (ii), but this time with n-dimensional replaced by n-dimensional, and bases replaced by
spanning subsets.
(Remark: The assumption that k is infinite is convenient here. However, using the fact that any field
k can be embedded in an infinite field, one can show that if the above results are true for all infinite fields
k, they are in fact true for all fields k.)
(b) Show that the above equivalent statements imply that given a set S, and n 2 n-element subsets S11 ,
S12 , ... , Snn , one can choose an element sij from each Sij in such a way that for each i the n
elements sij ( j = 1, ... , n) are distinct, and for each j the n elements Sij (i = 1, ... , n) are distinct.
The assertion of (b) was a longstanding open question, called the Dinitz Conjecture, till it was proved
by Fred Galvin (The list chromatic index of a bipartite multigraph, J. Combin. Theory, Ser. B, 63 (1995)
153-158). The vector-space questions are still open, to the best of my knowledge.
Correction to III: L 6 (p.166): At the start of the next-to-last line, before a Z-basis add a subgroup of
finite index with. (The statement in Lang is false. The above modified statement is almost trivial to
prove, once one sees the method. I dont know any version of the statement that is correct but nontrivial
to prove.)
Note also that in the next-to-last line, the subscript on {ys }, which appears as wes through the first
Springer printing, and as w s in the 4th Springer printing, should be w S.
III.6:1. Let A be a commutative ring, B a commutative A-algebra, and M a B-module. Recall from
Lemma III.1.c3 in this Companion (notes to Langs p.121) that the B-module M may be regarded as an
A-module by letting each a A act on elements of M via the image of a in B. Let us denote by M v
the A-module Hom A (M, A).
(a) Show that M v can be made a B-module by defining, for each b B and f M v, the element
bf M v via the equation (bf )(x) = f (bx).
(b) Suppose A is a field, and b any element of B. Show that multiplication by b is one-to-one on
M v if and only if it is surjective on M, and that it is surjective on M v if and only if it is one-to-one on
M.
(c) Still letting A be a field, let B = A[X ], the polynomial ring over A, and let M = B regarded as a
module over itself. We see from part (b) that M v will contain nonzero elements annihilated by X and
by X +1. Find such elements.
(Remarks: If we had not assumed B commutative, we could still have gotten a result like (a), except
that a left B-module structure on M would have yielded a right B-module structure on M v. More
generally, for any A-module N and left B-module M we get a right B-module structure on
Hom A (M, N ). If, in addition, N is a left B-module, then this induces a left B-module structure on the
same set, such that these two structures together make it a (B, B)-bimodule. Within that bimodule
Hom A (M, N ), one can characterize Hom B (M, N ) as a centralizer, { f ! ( b B) bf = fb}.)
III.6:2. Show that if P is a finitely generated projective left module over a ring R, then the right
R-module P v = Hom(P, R) is also finitely generated and projective, and the natural map P P v v is an
isomorphism.
Note on exercises for III.7: Most of the exercises given earlier in this Companion for I.8 can be used as
exercises for this section, with abelian group everywhere replaced by A-module, where A is a
principal ideal domain. For those exercises that ask for counterexamples, some additional assumptions are
needed (since when A is a field, the statements to which counterexamples are being asked for tend to be
true). In these cases, it might be made part of the exercises to find conditions on a principal ideal domain
under which such counterexamples will exist.
III.6:3. Let k be a field, let f : U V be a linear map of K-vector-spaces, and let f v : V v U v be
the induced homomorphism of dual spaces.
(a) Show that f v is one-to-one if and only if f is onto, and that f v is onto if and only if f is one-to-
one. (This will be easy, using an appropriate result of this section.)
(b) Regarding each of the if-and-only-if assertions of part (b) as two statements, determine which of the
resulting four statements remain true if field K and vector space are replaced by ring A and
A-module.
(Insofar as you find counterexamples, you might see whether you can get them to satisfy to various
additional conditions, such as using modules that are finitely generated, and/or free, and/or with base-ring
A the integers.)
III.7:1. Let F be a free module of finite rank over a principal ideal domain R, and D E F
submodules. Must there exist a basis of F such that both D and E are spanned by multiples of
members of F ? (Cf. Theorem III.7.8.) Give a proof or a counterexample.
III.7:2. Theorem III.7.9 in our text (as corrected in the first Springer printing) has the hypothesis that the
elementary matrices in R generate GLn (R). This leads to the question: For which R is this true?
Show that this condition holds for R = Z.
Suggestion on how to start: Given an element of GLn (R), show that by row operations one can
reduce the sum of the absolute values of the entries in the first column to 1, and finally bring that column
to a form with 1 in the first position and zeroes below. Show that the submatrix below and to the right of
that 1 is then invertible.
We will generalize this result in Exercise IV.1:3.
Note on III: L 14 (p.169, based on III.9): Langs Hint should be considered to apply to the exercise as
a whole, not to part (c) in particular. In fact, it appears to me that part (c) is most easily done by using
parts (a) and (b) to get one case, and a non-snake argument in the remaining two easier cases.
Note on III: L 16 (p.169, based on III.10): I would mark this ; i.e., it is essentially trivial once one
makes an easy observation (at least as it is stated in all printings through the 4th Springer printing: Prove
that the inverse limit of a system of simple groups in which the homomorphisms are surjective is either the
trivial group, or a simple group). Perhaps what Lang intends is, rather, the following. Note that here, no
surjectivity assumption is made.
III.10:1. Prove that the direct limit of a system of simple groups is either the trivial group, or a simple
group.
On the other hand, if we consider inverse limits, but delete the surjectivity assumption, we can pose
*III.10:2. Show by example that an inverse limit of simple groups need not be simple.
IV.1:1. = the handout A principal ideal domain that is not Euclidean. (When I assign this, I generally use
parts (1)-(7) as two thirds of a weekly homework assignment.)
*IV.1:2. Suppose A is a commutative ring, and f A[X ] is a zero-divisor. Show that there exists an
element a A {0} such that af = 0.
IV.1:3. Do Exercise III.7:2 for the more general case where R is a Euclidean domain (defined in these
notes, re p.175).
IV.3:1. Part (a) below generalizes the Eisenstein Irreducibility Criterion (Theorem IV.3.1, p.183). As in
the statement of that result, let A be a unique factorization domain, K its field of fractions, p a prime
element of A, and f (X) = an X n + ... + a 0 A[X ].
(a) Suppose 1 j < m n are such that ai 0 (mod p) for all i < m, not all coefficients of f are
divisible by p, and a 0 / 0 (mod p j+1). Show that when f is factored into irreducibles in K[X ], there
will be a family of at most j of these factors whose product has degree at least m. (Hint: Look at those
factors of f whose constant term is divisible by p. Compare the factorization of f in A [X ] and its
image in (A ( p))[X ].)
(b) Obtain Eisensteins Criterion as a special case.
(c) Deduce, as another case, that if in the statement of Eisensteins Criterion, the condition ai 0
(mod p) is only assumed for i < n 1, then f is either irreducible, or the product of a linear factor and
an irreducible factor of degree n 1.
IV.4:1. If R is a ring, I an ideal of R, and S a subring of R, it is not hard to verify that the set
S + I = {s + i | s S, i I} forms a subring of R. I will not ask you to prove this straightforward fact;
but let us see how this construction can be used to get examples of non-Noetherian rings.
(a) Show that the subring Z + x Q[x] of Q[x] (consisting of all polynomials with rational coefficients,
and constant term an integer) is not Noetherian.
(b) Let k be a field. Show that the subring k + x k [x, y] of k [x, y ] (consisting of all elements in
which the coefficient of each monomial y i (i > 0) is zero) is not Noetherian.

*(c) Show that the ring constructed in (a) above has the property that every finitely generated ideal is
principal, but is not a UFD.
*(d) Show that if k is a field and n a positive integer, then k + x nk[x] is Noetherian, and in fact has
the property that every ideal can be generated by n elements, but contains an ideal which cannot be
generated by < n elements.
IV.6:1. = the handout on Solutions in radicals. (Although this exercise seems naturally to go with VI.7,
it can be assigned any time after IV.6, provided the terms primitive cube roots and fourth roots of
unity are explained and their elementary arithmetic properties noted. I like to assign it early, and then
discuss it again after we read VI.7. As most of a weeks assignment, I typically assign parts (1)-(4) and a
choice of (5) or (6). Part (8) depends on material that is in Hungerford but not Lang, and so should not be
assigned unless this is covered separately. Concerning part (2), the instructor should specify what results
about generating sets for Sn , e.g., I: L 38(a-c) and/or I.5:7, may be assumed.)
Note on IV: L 12(c): The notation Df for the discriminant is as defined on p.193; on p.204 the same
element is written D ( f ).
V.1:1. = the handout on Luroths Theorem. (As a weeks assignment, I typically give parts (1)-(9). Part
(13) is immediate from II.5:1 above, and so should be omitted if that exercise has been assigned.)
In the phrase simple transcendental extension, defined on that handout, the word simple has its
original sense one-fold, in this case generated by a single element. One does not call this a cyclic
extension because in Chapter VI, that term will be used to mean an extension having cyclic Galois group.
V.1:2. Suppose E and F are subfields of a field K, such that every element of E is algebraic over F,
and every element of F is algebraic over E. Must E and F be algebraic over E F ?
Notes on a couple of Langs exercises in Chapter V. One direction of V: L 17 (p.254, based on V.6) is a
result in the text, so the exercise is essentially to prove the converse. Exercise V: L 21 is an immediate
consequence of II: L 12; see note on that exercise at end exercises for II.1. Hint to V: L 28 (p.256, based on
V.2): Reduce to the case where the extension of odd degree has the form k( ), express this as
k[X ] ( f (X )), and thus translate the hypothesis into a statement about polynomials. Expect to use
induction on the degree of f .
Note: Exercises VI.14:1 and VI.14:2 can actually be given at any point after VI.1; cf. comment
preceding those exercises.
VI.1:1. Let K be a finite Galois extension of a field k, with Galois group G. Since G is a group of
permutations of K, we may regard k as a G-set. Let , K.
(a) Show that the orbits G = { | G } and G = { | G } are isomorphic as G-sets by an
isomorphism that carries to if and only if k ( ) = k ( ).
(b) Show that G and G are isomorphic as G-sets by some isomorphism if and only if k ( ) = k ( )
as extensions of k.
(c) In the situations of (a) and (b), what interpretation does the cardinality | G | have in terms of the
extension k ( ) of k ?
(d) Find characterizations analogous to (a) and (b) of the situation where G is the image of G under
a homomorphism of G-sets, with and without the condition that the homomorphism carry to .
(e) Why have I not also asked the analog of question (d) with homomorphic image replaced by
sub-G-set?
The only part of VI.2 that the next exercise assumes is Example 1 on p.269, so it can be given after
covering VI.1, with a reference to that example. I typically give this exercise as a full weeks homework
assignment.
VI.2:1. = VI: L 8 (p.322). However, in parts (a) and (b), assume that you are working over an arbitrary
field k of characteristic 2, rather than Q. Also begin by justifying (briefly) the statement that the roots
of f may be written , .
If you have a printing before the 4th Springer printing, there are errata to part (b) for which you should
see the pages of errata. (There are three errata-listings there. Which one applies depends on the printing
you have.)
In part (c), each subextension other than the base field k itself will be quadratic over an extension just
below it in the lattice, and hence can be obtained by adjoining to that extension the square root of an
element thereof (p.269, Example 1). Hence, try to express each subextension as obtained from k via
successive adjunctions of square roots.
In doing this exercise, you may take for granted that the group that Lang calls D8 is a 2-Sylow
subgroup of S4 , and can be pictured as the group of symmetries of the square by identifying the four
elements on which S4 acts with the vertices of the square. In (b), expect to use repeatedly the
observation that an element lies in k if and only if it is fixed by all members of the Galois group. All the
assertions in that part, including the parenthetical one, are to be proved, even though only one assertion is
preceded by Show that. In (c) you may take for granted that the subgroups of D8 are as shown in the
chart on p.271, for appropriate choices of generators and satisfying the relation displayed a few
lines above the diagram. In the last part of (c), where you are asked for the field associated with each
subgroup of the Galois group, you may find it useful to apply the result of (b) not only over Q, but also
over some of the fields you are trying to determine.
In displaying the diagram of subextensions, you will have to choose between the arrangement that
parallels the diagram of subgroups (with smaller subextensions, corresponding to larger subgroups, at the
top) and the arrangement natural for these subextensions, with larger subextensions at the top. To give the
grader a uniform task, use the latter arrangement. For the same reason, take to be complex conjugation.
The word lattice in the last sentence of (c) means partially ordered set in which every pair of
elements has a least upper bound and a greatest lower bound; but you are not asked to prove anything
about such upper and lower bounds.
VI.2:2. (converse to VI.2:1(a), and related results). Let k be a field not of characteristic 2.
(a) Suppose K is a Galois extension of k whose Galois group G is isomorphic to the group Lang
writes D 8 . Show that K is the splitting field over k of an irreducible polynomial f ( X ) of the form
X 4 + aX 2 + b. (Hint: Find a non-normal subgroup H < D 8 of order 2, let N be the normal subgroup it
generates, and note that K H is generated by the square root of an element K N .)
(b) Show that the above conclusion is also true of any Galois extension K of k having degree 4.
(Suggestion: Consider separately the cases of Galois group Z 4 , and Z 2 Z 2 .)
(c) Show that if an irreducible polynomial f of degree 4 has splitting field of degree not divisible by 3,
then its Galois group is D 8 , Z 4 , or Z 2 Z 2 . Must such an f have the form X 4 + aX 2 + b ?
(d) Show that if f is an irreducible polynomial over any field whose Galois group is the 8-element
quaternion group described by Lang on p.9, then f must have degree 8.
(The study of polynomials having this Galois group is more difficult than the case of D 8 , and will not
be pursued further here.)
Still another variant of the theme of this and the preceding exercise is given in VI.2:6.
VI.2:3. [cf. I.5:6]. Let f be a separable irreducible polynomial of degree n > 1 over a field k, let K
be the splitting field of f , and let 1 , ... , n be the roots of f in K. Show that the following
conditions are equivalent:
(i) f ( X )/( X 1 ) is irreducible in k(1 )[X ].

(ii) G(K k ) is 2-transitive on {1 , ... , n } (as defined in Exercise I.5:6).
VI.2:4. Let k be a field, f a separable irreducible quadratic polynomial over k, and g a separable
irreducible cubic polynomial over k. Let us regard the Galois group G of f g as a group of
permutations of the five roots of this polynomial. Since it permutes the set of roots of f and g
separately, it lies in the obvious copy of S2 S 3 within S 5 .
(a) Show that G need not contain a product of a 2-cycle in S 2 and a 3-cycle in S 3 . (Hint: Find a
subgroup of S2 S 3 with appropriate properties, then note why one can realize this as a Galois group of
the indicated sort.)
(b) Show, on the other hand, that if k is a finite field, then G must indeed contain the product of a
2-cycle and a 3-cycle. (This justifies an argument Lang gives on p.274, below the middle display.)
(c) Abstracting some ideas used above, deduce the following result: If a, b are elements of a finite field
k, then the polynomial (X 3 + aX + b)(X 2 + (4a 2 + 27b 3)) has a root in k.
*VI.2:5. Show that the primeness assumption is needed in VI: L 9 (p.322) by showing that for every
composite positive integer n, the analogous statement about extensions of degree n is false.
VI.2:6. Suppose K is separable quartic extension of k (an extension of degree 4).
(a) Show that if K is not a quadratic extension of a quadratic extension of k, then the normal closure L
of K over k has an element of degree 3.
(Suggestion: Translate the field-theoretic assumptions into conditions on G = G (L k ) and on the
subgroup thereof corresponding to K . Then translate these in turn into conditions on G-sets, using the
result that subgroups H G induce transitive G-sets G H . This will turn the question into one about
subgroups of a finite symmetric group, which is not hard to prove.)
(b) Show that if char(k) 2, then the condition of (a) that K is not a quadratic extension of a
quadratic extension of k is equivalent to saying that K is not generated over k by a root of a
polynomial of the form X 4 + a X 2 + b as in VI.2:1-2.
(c) Show that there exist fields K and k as in part (a).
Note on VI: L 10: The words with n > 2 at the end of the second line belong in the first line, after
degree n. The condition is the symmetric group Sn is ambiguous: it could mean acts as the full
group of permutations on (the subscripts of) the n zeroes of f , or simply is isomorphic, as a group, to
Sn . So for completeness, show that in the situation in question, those two conditions are equivalent,
before showing that they imply the results of (a) and (b).
Note on VI: L 14: I believe one is expected to assume the italicized result stated in Example 7, p.274.
Notes on VI: L 16: Given a Galois extension K k , a normal basis for K k means a set of the form
{ } G for some K, which forms a vector-space basis of K over k. So the hypothesis that
{ } G is a normal basis is just equivalent to saying it is a basis. In VI.13, Lang proves that every
Galois extension has a normal basis. Though we shall not read that section, this exercise, about a property
that holds assuming such a basis given, is not hard to do. Note that since is given, when Lang says
Show that there exists a basis ... consisting of elements of the form S ( ) he means Show that there
exist subsets S1 , ... , Sr of G such that {S1 (), ... , Sr ()} forms a basis ....
VI.3:1. = the handout on Quadratic Reciprocity. Part (1)(i) is immediate from a result in Lang; you may
simply cite that result. (As a weeks homework, I typically assign the main parts, (1)-(8), an application
such as (9), and perhaps one of (11)-(13). Part (11) requires VI.6. (10) is important, but, together with
(1)-(9), would add up to too much for a weeks assignment.)
VI.5:1. Below, you will work through an easy proof of XIII: L 32, a beautiful result that has been
rediscovered many times. The method sketched here is adapted from H. W. Lenstra, Inventiones math. 25,
p.299-325, Proposition 1.3.
Let K, k and V be as in XIII: L 32.
(a) Define a function Tr G : V V analogous to the trace function of VI.5, and show that this function
is k-linear, and takes values in the fixed k-subspace of the action of G on V.
(b) Show that if : V K is a K-linear map, and w V has the property that Tr(Kw) Ker( ),
then w Ker( ). (Hint: Write out the relations involved, and use linear independence of characters.)
(c) Deduce that Tr(V ) lies in no proper K-subspace of V, and conclude that it contains a K-basis of V,
and that this establishes the result of XIII: L 32.
VI.6:1. The relation between the additive and multiplicative forms of Hilberts Theorem 90 suggests that
addition of 1 in a field of characteristic p is analogous to multiplication by a primitive pth root of unity
in a field of characteristic p. Now recall that the discriminant is a polynomial in n indeterminates
which gets multiplied by 1 (a primitive square root of unity) on applying any odd permutation to the
indeterminates. This leads to
(a) Let n be a positive integer. Deduce from results proved in Lang that there exists a rational function
rn in n indeterminates over Z 2 Z with the property that the action of any odd permutation of the
indeterminates sends rn to rn +1. (Recall that rational function means element of the field
(Z 2 Z)(t1 , ... , tn ), though the elements of such a field cannot, strictly speaking, be regarded as
functions.) Show, moreover, that rn (rn +1) is a rational function of the elementary symmetric
polynomials in t1 , ... , tn .
(b) Find an explicit description of such a rational function rn for general n.
(c) For n = 2, 3 obtain formulas expressing rn (rn +1) in terms of the symmetric polynomials in the
given indeterminates.
VI.6:2. Let k be a field, and E the rational function field over k in 2n indeterminates, x1 , ... , xn ,
y1 , ... , yn . Let Sn act on E by simultaneously permuting the two sets of indeterminates, so that (xi ) =
x (i) , (yi ) = y (i) , and let K be the fixed field of this action.
(a) Show that for all i, K(xi ) = K(yi ). Deduce that K(x1 , ... , xn ) = K(y1 , ... , yn ) = E.
*(b) Can you find explicit expressions for the yi in terms of the xi and elements of K ? (I dont know
whether there is an easy solution.)
Note on VI: L 6: In the second sentence of part (b), for Let z be an element read Assume there exists
an element z. I.e., the existence of such an element is not a consequence of the preceding conditions.
(If it were, the exercise would imply that every quadratic extension field was contained in a quartic
extension field, which we know is not so from the case of C R.) On the other hand, in the third line,
Let 2 = means Extend E by adjoining an element satisfying 2 = .
Note on VI: L 29: I suggest doing part (c) before part (b).
Note on VI: L 44: Where Lang has lim , the form Lim
n
_
n
might be clearer.
Note on the next two exercises below: Neither of these actually assumes material in Langs VI.14 or in
my comments on that section; they could be given after VI.1. I put them here because they seemed most
closely related to the ideas of VI.14.
VI.14:1. Do VI: L 26 (p.325), but give the result in as much generality as your proof naturally leads to.
Also see whether you can obtain from your proof any stronger condition on the Galois groups of finite
subextensions of E than that they be cyclic.
*VI.14:2. (generalization of VI: L 25, p.325). Suppose K is an infinite Galois extension of a field k, and
n a positive integer such that for every finite Galois subextension E K, the group Gal(E k ) can be
generated by n elements. Show that k is the fixed field of some n-generator subgroup of Gal(K k ).
VIII.1:1. Let k be a field, K an extension field, and E, F intermediate fields. Recall that EF
denotes the compositum of E and F in K.
(a) Show that tr.deg.(E F k ) + tr.deg.(E F k ) tr.deg.(E k ) + tr.deg.(F k ).
*(b) Can the inequality of (a) be strict?
Note on VIII: L 1 (p.374, based on VIII.1): The last sentence (at least as it appears in all printings through
the 4th Springer printing), Describe all automorphisms and their cardinality, is unreasonable, unless one
allows a very vague sense of Describe. Determine the cardinality of the group of all automorphisms
would be reasonable.
Note on VIII: L 6: I think that k(x) and k(y) in the last line should be ku (x) and ku (y).
X.4:1. Let A be a ring and a A a 2-sided ideal.
(a) Show that if E is a finitely generated projective left A-module, then E aE is a finitely generated
projective left A a-module.
In the remaining parts, let us suppose a is contained in all maximal left ideals, and in (b)-(d), let E,
F be finitely generated projective left A-modules.
(b) Show that a module homomorphism f : E F is an isomorphism if and only if the induced
homomorphism of A a-modules, f : E aE F aF is an isomorphism.
(c) Deduce that E
= F if and only if E aE = F aF.
(d) Show that E is isomorphic to a direct summand in F if and only if E aE is isomorphic to a direct
summand in F aF.
*(e) Show by example that A a may have a finitely generated projective module P such that there is no
finitely generated projective A-module E with E aE
= P.
XIII.1:1. We observed in the comments on XIII.1 that mR n is a left module over mR m and a right
module over nR n , and that each of these rings acts on mR n by endomorphisms as a module over the
other.
Show that the endomorphisms of mR n as a right nR n -module (when written to the left of their
arguments and composed accordingly) form precisely the ring mR m , and likewise that the endomorphisms
of mR n as a left mR m -module (written to the right of their arguments and composed accordingly) form
precisely nR n .
Suppose that by force of habit we write endomorphisms of the left mR m -module mR n to the left of
their arguments, and compose them accordingly. Describe the structure of the endomorphism ring in this
case.
*XIII.2:1. Let R be a field and n a positive integer. Show that any automorphism of the ring nR n
preserves ranks of matrices; i.e., that if A nR n, then rank( (A)) = rank(A). (Hint: Find some
characterization of the rank of A purely in terms of the ring structure of nR n ; e.g., in terms of the left
or right ideal generated by A.)
*XIII.2:2. Let the inner rank of a matrix A mR n over a ring R be defined as in the notes to XIII.2 in
this Companion. Show by example that the inner rank of a matrix A over a commutative integral domain
can be different from the rank of A over the field of fractions of R.
XIII.3:1. [cf. III.4:3 and XIII.1:1]. Let R be any ring, n any positive integer, and L any left
n n
R -module. Let M = {e11 x ! x L }. Essentially as in Exercise III.4:3, this set may be regarded as a left
R-module. (I wont ask you to verify this this time.)

(a) Show that nM = L as left nR n -module. (Here we understand nM to be made an nR n -module, using
the R-module structure on M, as in this section of Lang.)
This shows that every left nR n -module is, up to isomorphism, of the form nM for some left R-module
M. Hence to understand nR n -modules, it remains only to describe their homomorphisms.
(b) Given left R-modules M and N, show that Hom n n ( nM, nN)
R = Hom (M, N) as abelian groups.
R
(It follows that the category of left nR n -modules is equivalent to the category of left R-modules. Note,
however, that under this equivalence, the free left nR n -module of rank r corresponds to the free left
R-modules of rank nr. In general two rings having equivalent module-categories are called Morita
equivalent. There is a general theory describing how this happens; the above is the simplest nontrivial
case.)
*XIII.4:1. This problem concerns a possible converse to XIII: L 26 (p.548).
Suppose n is a positive integer, and
f Z[t11 , t12 , ... , tnn ]
a polynomial in n 2 indeterminates, and suppose f has the property that for every commutative ring A,
every n-generator ideal I = Ax1 + ... + Axn A, and every family of n 2 elements cij A, if we define
yi = cij xj and I = Ay1 + ... + Ayn , then we have
f (c11 , c12 , ... , cnn ) I I.
Must f be divisible (as a polynomial) by det(tij )?
XIII.4:2. Let m and n be positive integers, and let E be a free module of rank m over a
commutative ring R. Show that Lan(E) is a free R-module, and determine its rank. (For n = m you
should, of course, get Corollary XIII.4.12, p.517.)
XIII.4:3. (i) = XIII: L 25 (p.547). Hint: The n = 2 case is easy. To simplify the problem for general n,
apply appropriate invertible transformations to pairs of columns.
*(ii) More generally, suppose we are given an m n matrix A over a principal ideal domain, with m < n.
Can you find necessary and sufficient conditions for this to be completable to an n n matrix of
determinant 1 ?
XIII.5:1. We have been studying the classes of symmetric and alternating bilinear forms on a free module.
This exercise will examine the question of whether there are any other natural classes of bilinear forms.
For simplicity, we will assume our base ring is a field.
Given a field k, let us define a subfunctor X of L 2 to mean a construction associating to every
finite-dimensional k-vector-space V a subspace X(V ) L 2(V, k) in such a way that for every linear map
f : V W among vector spaces, the induced map L 2( f ) : L 2(W ) L 2(V ) carries X(W ) into X(V ).
Let us partially order the subfunctors of L 2 by writing X X if for all V, one has X(V ) X(V ).
Show that the subfunctors of L 2 are precisely L 2 itself, La2, Ls2, and 0, and determine the partial
ordering on this set of four subfunctors. This ordering will depend (to a slight extent) on k.
(Remark: What we are calling L 2(V ) can be naturally identified with Hom k (V 2 k V, k). A more
natural form of the above question would ask for all subfunctors of the functor V V 2 k V. The above
results would follow from the answer to this question with the help of the standard correspondence between
subspaces of a finite-dimensional vector space and subspaces of its dual. But I forewent this elegance to fit
the question to the reading. Note also that one might generalize the original question by replacing
bilinear with n-linear for fixed n 0; equivalently, in the more natural form just mentioned, one
might replace the 2-fold tensor product with the n-fold tensor product. The answer then becomes
considerably more complicated!)
XIII.5:2. Let k be a field, and an infinite cardinal.

(a) Prove that the dual of a -dimensional k-vector-space has dimension at least max(2 , card k).
Suggestion: If card k < 2 , look at the cardinality of the dual space. In the contrary case, show that
the dual has a subspace isomorphic to the space k of all sequences of elements of k, and show that the
exponential sequences (1, c, c 2, ... ) (c k) are linearly independent.
The dual space considered in the above result is clearly isomorphic to the k-vector-space of all k-valued
functions on a set of cardinality . It is interesting that one can get a similar result for a much more
restricted space of functions:
(b) Let S be a set of cardinality , and V the space of k-valued functions on S that assume only
finitely many distinct values. (Equivalently, the space spanned by the characteristic functions of all subsets
of S.) Show that V has dimension 2 .
Suggestion for showing : Let X be a set of cardinality , and note why you can assume without
loss of generality that S is the set of all finite subsets of X. For every subset Y X, let Y V be the
function which takes the value 1 at members of S which are subsets of Y, and 0 at all others. Show
that these 2 members of V are k-linearly independent.
*XIV.2:1. (Cf. Corollary XIV.2.3, p.558.) Let A1 , ... , Ar , B1 , ... , Br be n n matrices over a field k.
Assume there is an invertible matrix C over an extension field k of k such that Bi = C Ai C 1 (i =
1, ... , r). Must there exist an invertible matrix C over k itself with this property? If you cant answer
the question in complete generality, you might see whether you can do so under some additional
assumption.
XIV.3:1. (a) = XIV: L 10 (p.568). (Note that while a polynomial f(t) k[t] can be evaluated at any
element of k, we can merely say that a rational function r(t) = p(t) q(t) k(t) can be evaluated at
almost all elements of k, namely, those that are not roots of q, assuming our fraction written in
lowest terms. Likewise, one can only say that r(A) can be evaluated at most matrices. You should
describe the sense in which one can speak of evaluating r(A), and then state and prove the result asked
for.)
(b) Suppose M is the matrix of an endomorphism f of a finite-dimensional vector-space V, which
again satisfies the hypothesis of Theorem XIV.3.10. We know that f induces endomorphisms L 2( f ),
Ls2( f ) and La2( f ) of the spaces L 2(V), Ls2(V) and La2(V) respectively. Determine the characteristic
polynomials of these maps.
XVI.1:1. (a) Show that if A is a nonzero finitely generated abelian group, then the tensor product
Z A is nonzero.
A2
In contrast:
(b) Let p Z be a prime, and let the group Zp be defined as in I.8:3. Show that Zp 2 Z Zp = {0}.
*(c) Prove the analog of (a) for finitely generated modules over an arbitrary commutative ring. (Whether
this part should be starred really depends on what previous chapters have been covered. But the next
one certainly deserves its star.)
*(d) Does there exist a commutative ring k and a k-module M such that M 2 k M is nonzero, but

M 2 k M 2 k M = {0} ? If you cant answer this for general k, can you answer it for k = Z ?
XVI.1:3. Let E and F be vector-spaces over a field k, and z any element of E 2 k F.
(a) Show that one can write z = ri=1 xi 2 yi where x1 , ... , xr are linearly independent elements of
E, y1 , ... , yr are linearly independent elements of F, and r 0.
(b) Suppose we have another expression for the same element, z = ir=1 xi 2 yi , where the xs and
ys again form linearly independent families. Show that r = r, that the space spanned by the xs is the
same as the space spanned by the xs, and that the space spanned by the ys is the same as the space
spanned by the ys. In fact, show that if we write x for the row vector of elements of E formed by the
xs, x for the row vector formed from the xs, and likewise y and y for the column vectors of
elements of F formed from the ys and the ys, then there exists an invertible r r matrix U over k
such that x = x U, y = U 1y.
XVI.1:4. Let n be a positive integer and k a commutative ring, and let 2 n denote the functor from
k-modules to k-modules that takes every k-module M to the n-fold tensor product M 2 k ... 2
k M, which
we shall write 2 n M. Given any homomorphism h : M N of k-modules, the induced homomorphism
of n-fold tensor products will be written 2 n h: 2nM 2 n N.
By an endomorphism f of the functor 2 n we shall mean any construction associating to each
k-module M an endomorphism fM of the k-module 2 n M, which respects homomorphisms of
k-modules, in the sense that for every such homomorphism h : M N, the following diagram commutes:
nM
2
________
nh
2
2 nN
,, f (M)
,, f (N)
, ,

nM
2
________
2 nh
nN
2
One defines the sum of two endomorphisms of 2 n, and the result of multiplying such an
endomorphism by an element of k, in the obvious ways, and one thus sees that such endomorphisms form
a k-module (at least, if we ignore the question of whether they form a set!).
As an example, it is easy to see that for each element of the symmetric group Sn , there is an
endomorphism - of 2 n whose action on decomposable elements is given by - (x1 2 ... 2
xn ) =
... 2
x 1(1) 2 x 1(n) . We can now state the problem:
Show that every endomorphism f of 2 n is a k-linear combination of the endomorphisms -
( Sn ). (Suggestion: Look at the behavior of f on an appropriate universal element.)
Remark: Clearly one can also compose these endomorphisms, hence they have a natural structure of
k-algebra. It is not hard to deduce from the above result that this algebra is isomorphic to the group
algebra k Sn . But I will spare you the writing out, and myself the grading, of the verification of this fact.
XVI.1:5. [Assumes the material on Tensor products in the noncommutative context, this Companion.]
Let k be a field, and V a 1-dimensional k-vector-space, with basis {x}.
(i) Determine all ways to make V into a (k[t], k[t])-bimodule so that the right and left actions of
elements of k agree with their actions under the given k-vector-space structure.
(ii) For every pair of bimodules E and F of the sort characterized in (i), determine the
k[t] F.
(k[t], k[t])-bimodule structure of E 2
(iii) Show that for appropriate choice of E and F as above, one can on the one hand get E 2 k[t] F

nonisomorphic as a k-vector-space to F 2 k[t] E, and on the other hand, get these objects isomorphic as
k-vector-spaces but nonisomorphic as bimodules.
XVI.2:1. Here are a few examples showing that the homomorphism of Proposition XVI.2.5 (p.610) can
fail to be an isomorphism if E and F are not taken free of finite rank.
(i) Show that if R is a field, and E, F are infinite-dimensional vector spaces over R, then that
homomorphism is one-to-one, but not onto.
(ii) Show that if R = Z, E = Zp (see I.8:3), and F = Q, then that homomorphism is onto but not
one-to-one. (Hint: Show that End(Zp ) is torsion-free as an additive group, and use this to deduce that
the tensor-product of endomorphism rings is nontrivial.)
+ Zp , and F = Zp , then the homomorphism is neither one-to-one
(iii) Show that if R = Z, E = Zp 2 2
nor onto.
On the other hand, there are some nice cases not covered by the proposition in which one can show that
the homomorphism is an isomorphism:
(iv) Show that if E or F is free of finite rank, then that homomorphism is an isomorphism. (Hint: In
F as a direct sum of copies of F, and can use the ideas in the
that case, you get a decomposition of E 2
comment regarding p.505, bottom, in this Companion about using matrices to represent direct
homomorphisms among direct sums of modules.)
(v) Show, still more generally, that the homomorphism is an isomorphism whenever E or F is a finitely
generated projective module.
XVI.3:1. (i) Suppose R is a ring and G a functor from left R-modules to abelian groups, which
respects the group structure on homomorphisms; i.e., such that given f, g : E E, we have G( f +g) =
G( f ) + G( g). Show the equivalence of conditions F1 and F2 (pp.612-613) with the functor G in place
of those pages.
of the functor F 2
(ii) Show by example that these two conditions are not equivalent to F3 in this generality, by looking at
the functor L(F, ) for an appropriate left R-module F. (Suggestion: Figure out to what extent such a
functor must preserve exactness, then look for an example where the case that need not hold fails.)
Remark: Along with functors of the forms F 2 and L(F, ), there is one other familiar class of
functors on categories of modules that respects the group structure on morphisms: the contravariant
functors L(, E). Functors of each of these sorts in general preserve exactness at most, but not all parts of
a short exact sequence. The F or E for which they do preserve exact sequences (in which case they are
called exact functors) are called, respectively, flat modules (discussed in this section), projective modules
(pp.137-139) and injective modules (XX.4).
XVI.3:2. Prove the following partial converse to Proposition XVI.3.2: If R is a unique factorization
domain, and every torsion-free R-module is flat, then R is a principal ideal domain.
(Suggestion: Reduce the problem to showing that if x, y R have g.c.d. 1, then xR + yR = R, and
then somehow get this equality from the fact that xR + yR is flat. One way to use flatness might be to
apply Lemma XVI.3.3 with M F the natural map xR 2 + yR xR + yR, and E = R (xR + yR).)
*XVI.3:3. This exercise will prove the implication (c)(a) of Theorem XVI.3.2 by an argument close to
that of Lazards original paper.
Let E be a right R-module satisfying condition (b) of that theorem. Let H be a free right R-module
on a basis , given with a homomorphism h : H E such that for every x E, infinitely many
elements of are mapped to x. (For instance, we could let F be the free module on the product of the
set of natural numbers and the underlying set of E, and map this to E in the obvious way.)
Now let I be the set of ordered pairs (Hi , Ki ) where Hi is the free submodule of H generated by a
finite subset of , and Ki is a submodule of Hi Ker(h) such that Hi K i is also free. For
i = (Hi , Ki ) I we shall write Fi = Hi K i . Note that for each i I, the map h induces a
homomorphism hi : Fi E. Given i = (Hi , Ki ), j = (Hj , Kj ) I, we shall write i j if Hi Hj and
Ki Kj .
(i) If i j I, describe a natural homomorphism fj,i : Fi Fj over E, and show that if i j k I,
then fk,j fj,i = fk,i .
(ii) Show that if i I, and F is a free module of finite rank given with a homomorphism into E, and
f : Fi F is any homomorphism over E, then there exists j i such that Fj is isomorphic to F as a
module with a homomorphism from Fi into it, and a homomorphism from it into E. (I.e., there is an
isomorphism Fj = F which makes commuting triangles with the maps out of Fi and into E.) This
proof will make use of the assumption that infinitely many elements of map to each element of E, but
the assumption that E satisfies (b) will not yet be needed.
(iii) Show by combining the hypothesis (b) on E and the result of (ii) that the partially ordered set I is
_
directed, and that Lim
i
Fi
= E.
XVI.3:4. Let R be a commutative ring and m a maximal ideal of R. Thus, R m is a field, and
m m 2 can be regarded as a vector-space over that field.
(i) Show that if this vector-space has dimension > 1, then m is not flat as an R-module.
(ii) Show that all the examples of non-flat ideals given in the comment to p.614 (top two paragraphs) in
this Companion are instances of this result.
XVI.3:5. (a) Suppose 0 N M F 0 is a short exact sequence of right modules over a ring R,
and f : F 0 F a module homomorphism. Show how to obtain from this map a module M 0 , a short
exact sequence 0 N M 0 F 0 0, and a map of short exact sequences
0 N M0 F0 0
= f0
0 N M F 0.
(b) Verify that your construction is functorial; that is, that given maps from two free modules F 0 and
F1 into F, and a homomorphism F 0 F1 respecting the maps into F, you get a homomorphism
between your short exact sequences that respect your maps into the given sequences.
(c) Verify that your construction respects direct limits.
(d) Combining these observations with Lazards Theorem, obtain a proof of Lemma XVI.3.3 (p.614) based
on the fact that the desired result holds when F is projective (and so in particular, when it is free).
(e) Similarly, prove the first assertion of Proposition XVI.3.4 (p.615) by first establishing it when F is
free, then applying (a)-(c).
XVI.3:6. Let 0 E E E 0 be a short exact sequence of left modules over a ring R, and F
a right R-module. If F is E-flat and E-flat, must it be E-flat?
XVI.3:7. (Cf. XVI: L 6, p.638.) Suppose E is a flat right R-module and F is an (R, S)-bimodule which
R F is flat as a right S-module.
is flat as a right S-module. Show that E 2
XVI.3:8. Let R be the ring of all continuous functions on the interval [0,1], and E the R-module of
all germs at 0 of such functions; i.e., the R-module R I , where I is the ideal of all functions zero in
the neighborhood of 0. Show that E is a flat R-module. (This same ring and ideal appear in the more
difficult Exercise III.4:2.)
XVI.3:9. Suppose S is a commutative integral domain, K is its field of fractions, and R is a subring of
S having the same field of fractions.
(a) Show that every R-module homomorphism of S into R is given by multiplication by a member of
R, and that the set of r R such that multiplication by r carries S into R forms an ideal of R,
which is the largest ideal of R that is also an ideal of S. This is called the conductor ideal of R in S.
(b) Determine the conductor ideal of k[x] in k[x, x 1 ], and the conductor ideal of k[x 2, x 3] in k[x].
Under what conditions on R and S will 1 belong to the conductor ideal?
(c) Show that if the identity map of S, regarded as a homomorphism of R-modules, factors through a
free R-module, then S = R.
(d) Deduce that if S is finitely presented as an R-module (in particular, if R is Noetherian and S is
finitely generated as an R-module) and is strictly larger than R, then S is not flat as an R-module.
XVI.3:10. Let k be a field, and R = k x, y the free associative k-algebra on indeterminates x and y,
equivalently, the monoid algebra over k on the free monoid on x and y.
(a) Let P denote the right R-module presented by two generators and and the one relation
x y = 0. How that Hom R (P, R) = 0.

(b) Deduce that if f : R S is a ring homomorphism such that either f (x) or f (y) is invertible in S,
then S is not flat as a right R-module (under the induced module structure sr = s f (r)).
In particular, this shows that R cannot be embedded in a division ring that is flat as a right R-module,
and also that the group algebra of the free group on x and y is not flat as a right R-module.
XVI.3:11. Within the division ring of quaternions, let
R = {a + bi + cj + dk ! a, b, c, d Z},
S = {(a + bi + cj + dk) 2 ! a, b, c, d Z, a b c d (mod 2)},
K = {a + bi + cj + dk ! a, b, c, d Q}.
Clearly, R is a ring and K a division ring.
(a) Verify that S is also a ring.
(b) Show that K is flat both as a right R-module and as a right S-module.
(c) Show that S is not flat as a right R-module. (Suggestion: Look for analogs of the ideas of XVI.3:9.)
(d) Show that every right ideal of S is principal, and deduce that every right S-module which is torsion-
free as an abelian group is flat.
Remark: From (c) and (d), we see that S is better behaved than R; e.g., S is a
noncommutative principal right (and left) ideal domain, while R is not. So S is commonly called the
ring of integer quaternions, while there is no special name for R.
*XVI.3:12. (a) Show that if R is a commutative Noetherian ring, and P a finitely presented R-module,
then there exists a homomorphism h of P into a free R-module F of finite rank such that any
homomorphism of P into a free R-module factors through h.
(b) Deduce that the direct product of an (in general, infinite) family of flat R-modules over a commutative
Noetherian ring is flat. (We already know this for direct products of finite families, since these are the
same as direct sums.)
(c) To show that (b) fails for non-Noetherian R, let G be a monoid isomorphic to the monoid [0, )
of nonnegative real numbers, but written multiplicatively as {x r ! r [0, )}, let S be the monoid
algebra kG, and let R be the factor-ring of S by the ideal generated by all x r with r >1. (One might
write this sloppily as k[x [0,1]].) Show that a direct product of countably many copies of R is not flat.
(Hint: The element a = (x, x 2, x 4, ... ) satisfies the relation ax = 0.)
1 1
(d) Prove versions of (a) and (b) for noncommutative R.

XVI.4:1. Let f : R S be a ring homomorphism, let A be a right R-module, and let B be a left
S-module. Establish a natural isomorphism A 2 R B = AS 2 S B as abelian groups. Describe the
isomorphism and its inverse explicitly on generators of these tensor products if your method of obtaining
the isomorphism does not do so.
XVI.5:1. Let R be a commutative ring, and E and F R-modules. It is easy to see that the
F
construction of Langs Corollary XVI.5.5 (p.628) gives a functorial homomorphism E,F : E v 2
Hom(E, F), without the assumption of that corollary that the modules be free of finite rank.
Characterize some of the following classes of modules. (Instructor should specify how many; e.g.,
two.)
(a) Those E such that for all F, E,F is an isomorphism.
(b) Those F such that for all E, E,F is an isomorphism.
(c) Those F such that for all free modules E, E,F is an isomorphism.
(d), (e), ... Similar cases with an isomorphism replaced by one-to-one or onto, and/or with other
classes of modules, so long as you dont hand in two results with slightly varying hypotheses but
essentially the same proofs.

XVI.6:1. Let A be a Z-graded or Z 2 Z-graded algebra, as in the last paragraph on p.631 of Lang.
(i) If the grading monoid is Z, show that what Lang calls the super algebra Asu is actually
isomorphic to the ordinary algebra A.
(Suggestion: Take a polynomial ring R[x], grade it so that x has degree 1, and determine the unique
R-algebra isomorphism R[x] R[x]su carrying x to x. This will involve multiplying homogeneous
elements of each degree by an appropriate coefficient. Verify that these same coefficients give an
isomorphism in the case of any Z-graded algebra.)
(ii) On the other hand, show that if we grade the complex numbers by Z 2 Z, letting C 0 = R and
C1 = i R, then Csu is not isomorphic to C.
*XVI.8:1. For E a module over a commutative ring R and n a positive integer, does S n E = 0
n E = 0?
2
(This question was asked by Moss Sweedler. I succeeded in answering it a few years ago; but it has
potential generalizations in the context of the representation theory of the symmetric group which could
form an interesting research topic.)
XX.4:1. *(a) = XX: L 20 (p.830). (Suggestion [> XX: L 23]: Show that any ideal of S 1 A has the form
S 1 J for an ideal J of A. Use the fact that J can be finitely presented as an A-module to show that
any S 1 A-module homomorphism f : S 1 J S 1 I is of the form s 1- , where s S, and
1 1
- : S J S I is the map induced by an A-module homomorphism : J I. Then get the desired
result using XX: L 23.)
Parts (b) and (c) below refer to the statement of the exercise comprising part (a), but do not depend on
it or one another. Part (d) depends on (a) and (c).
(b) (T.Y.Lam) Show that the conclusion of the above exercise remains true if the assumption A is
Noetherian is replaced by the condition that the module I has no S-torsion; i.e., that no nonzero element
of I is annihilated by an element of S.
*(c) Show that the result in question becomes false if neither the Noetherian condition of part (a) nor the
no S-torsion condition of (b) is assumed. (Suggestion [> XX: L 21(a)]: Let k be an algebraically closed
field, and regard the polynomial ring k[x, y] as a subring of the ring of k-valued functions in two k-valued

variables, k k k. Let A be the subring of k k k spanned by k[x, y] and all functions of finite support,

and let I = k k k regarded as an A-module. XX: L 21(a) reduces the problem of showing I injective to
the same problem for certain much smaller modules; get these cases with the help of the fact that the
element of A having the value 1 at a single point and 0 elsewhere is idempotent, hence induces a
direct product decomposition of A as a ring. Now let S = k[x] {0} k[x, y] A, and get
descriptions of S 1 A, and S 1 I. Show that if S 1 I were injective, then multiplication by y would
map it onto itself, but that it does not.)
*(d) [> III.10] Give a version of the proof of (a) above that proceeds via the following three results: (i) if
A is Noetherian, then a direct limit of injective A-modules is injective; (ii) for any commutative ring A,
any A-module I, and any multiplicative set S, S 1 I can be expressed as a direct limit of copies of I;
(iii) if an S 1 A-module is injective as an A-module, it is injective as an S 1 A-module.
Deduce from (c) that the condition that A be Noetherian cannot be dropped from (i) above.
Notes on exercises connected with Langs Appendix A2: Exercises I.1:1 above assumes that
Appendix A2.1 was read before I.1. If they are read in the reverse order, that can be considered an
exercise for Appendix A2.1.
Exercises A2.2:1-4 below concern Zorns Lemma, developed in Appendix 2.2 and in my handout on
The Axiom of Choice, Zorns Lemma and all that. Since I introduce this material between sections I.6
and I.7 of Lang, most of the exercises below assume the material in the first part of Chapter I.
If Exercise A2.2:1 is not assigned, it should probably be discussed in class.
A2.2:1. Show the following facts. (Negative results should be shown by constructing counterexamples,
and proving relevant properties of these if they are not obvious.)
(a) Every nontrivial finitely generated group G has a maximal proper subgroup. (This means a maximal
element in the class of proper subgroups, partially ordered by inclusion.)
(b) Every nontrivial finitely generated group G has a maximal proper normal subgroup (a maximal
member of the class of proper normal subgroups).
(c) A nontrivial finitely generated group G need not have a minimal nontrivial subgroup. (I.e., there
exists a nontrivial finitely generated group G not having such a subgroup.)
(d) Every nontrivial finitely generated group G has a simple homomorphic image.
(e) A maximal proper subgroup of a group G need not be normal, and a maximal proper normal
subgroup of G need not be a maximal proper subgroup of G.
(f) A nontrivial non-finitely-generated group G need not have a maximal proper subgroup, nor a maximal
proper normal subgroup, nor a simple homomorphic image.
(Remark: The terms maximal proper subgroup, minimal nontrivial subgroup, etc. are often
shortened to maximal subgroup, minimal subgroup, etc.. That is, since the properties of maximality
and minimality in the class of all subgroups of G characterize subgroups that can be named much more
easily, it is taken for granted that when one refers to maximality or minimality, one must be thinking of
proper, respectively nontrivial, subgroups. But there are situations where such abbreviation can lead to
confusion; so be cautious in using them. If you are going to be writing extensively about maximal proper
subgroups and want to use this abbreviated term, you should begin by saying that maximal subgroup
will be understood to mean maximal proper subgroup.)
A2.2:2. [cf. A2.2:1]. (a) Does every nontrivial finitely generated semigroup have a simple homomorphic
image? (A semigroup S is called simple if it is nontrivial, and every homomorphism from S to another
semigroup is either an isomorphism, or is trivial, i.e., has one-element image.)
(b) If G is a group, will every nontrivial finitely generated G-set have a simple homomorphic image?
A2.2:3. Let X be a set. If and are two partial orderings on X, we say that is a refinement
of if
( x, y X ) x y x y,
i.e., if, when we regard the relations and as subsets of X X, the former is contained in the
latter. Clearly, is a refinement of is a partial ordering on the set of partial orderings on X !
(a) Show that every partial ordering on X can be refined to a partial ordering maximal with respect to
refinement.
(b) Show that every such maximal partial ordering is a total ordering.
(c) Is every partial ordering on X the intersection of some set of total orderings?
A2.2:4. For this exercise, we need the following definition and observations.
A subdirect product of a family of groups Gi (i I ) means a subgroup G of the direct product
..
I I I Gi , whose projection to each Gi is surjective. (Precisely: the composite of the inclusion G
.. ..
I I I Gi with each projection I I I Gi Gj is surjective.) Given a group G, we say that G can be
written as a subdirect product of a family of groups Gi if it is isomorphic to such a subgroup of their
product. Note that in this situation, each Gi will be isomorphic to a factor-group G N i , and the
intersection of these Ni will be {e}. Conversely, given G and any system of normal subgroups with
trivial intersection, we get a representation of G as a subdirect product of the factor-groups G N i .
A group G is called subdirectly irreducible if the only representations of G as a subdirect product

are those in which at least one of the maps G Gi is an isomorphism.
(a) Show that G is subdirectly irreducible if and only if it has a smallest nontrivial normal subgroup (a
nontrivial normal subgroup contained in all nontrivial normal subgroups).
(b) Show that if G is a group, x an element of G, and N a normal subgroup of G maximal for the
property x / N, then G N is subdirectly irreducible.
(c) Deduce that every group can be written as a subdirect product of a family of subdirectly irreducible
groups.
(Exercise I.6:5 concerned some restricted cases of the above results. There, finiteness allowed one to
work without Zorns Lemma, and to express the conclusion in terms of embedding in a product of groups
of smaller order.)
A2.2:5. Let G be a group. By a maximal abelian subgroup of G we understand a maximal member
of the class of abelian subgroups; not necessarily maximal as a subgroup of G. Recall also that the
centralizer of a subset X G means {g G ! ( x X ) xg = gx}.
(a) Show that every abelian subgroup of G is contained in a maximal abelian subgroup.
(b) Show that the maximal abelian subgroups of G are precisely those subgroups of G that are their
own centralizers.
(c) Show that for a subgroup (in fact, a subset) A of G, the following three conditions are equival-
ent: (i) A is the intersection of a nonempty set of maximal abelian subgroups of G. (ii) A is the
centralizer of some subset of G containing A. (iii) A is an abelian subgroup of G, and for every
element g G not in A, there is an element h G which commutes with all elements of A, but not
with g. (Note: the next part, (d) is an easy consequence of (c). But if you are having trouble proving one
of the implications needed for (c), you might try to work out a direct proof of (d), then see whether the
argument you arrive at can be modified to complete your proof of (c).)
(d) Show that the center of G is the intersection of all maximal abelian subgroups of G.
A2.4:1. Let A be a partially ordered set. Show that the partial ordering of A can be refined (in the
sense of A2.2:3) to a well-ordering if and only if every nonempty subset of A has an element minimal
with respect to the given ordering. (Hint: Mimic the proof of Theorem A2.4:1.)
Note on A2: L 13 (i.e., Exercise 13 at the end of Appendix 2, p.893). Hint: The set of maps from X to Y
is often denoted Y X . Think laws of exponents, then try to justify the argument precisely.
INDEX TO DISCUSSION OF LANGS EXERCISES IN THIS COMPANION

mostly in the above collection of exercises, but occasionally elsewhere
(excluding errata that apply only to earlier printings, for which see Errata section)
Exercises to Chapter I, pp.75-82: #3 clarified in I.3:1. Note on #4 at end of exercises on I.3.
#5 discussed and generalized in I.3:2. #6 incorporated into I.6:4(a). #7 incorporated into I.5:1. #15 and
#16 incorporated into I.5:4. #24 incorporated into I.6:1. #38(d) generalized in I.5:7. In #44 an
assumption can be deleted, as noted at end of exercises on I.2. #47 incorporated into I.5:6.
#57: Examples at end clarified at end of exercises on I.5.
Exercises to Chapter II, pp.114-116: #1 subsumed by II.2:1. #3 incorporated into II.4:2. #5 clarified at
end of exercises on II.5. #9 belongs in Chapter IV but done in Companion, as noted at end of exercises
on II.1. Clarification of #12 and cross-reference from #13-19 to III: L 11-13 noted at end of exercises on
II.1
Exercises to Chapter III, pp.165-171: Hint to #10 given at end of exercises on III.1. Note that #5-#8 can
go with I.7 after exercises for that section. #6 corrected after exercises on III.5. #14 clarified after
exercises on III.7. #16 discussed (trivial or misstated) after exercises on III.9; related questions
given as III.10:1-2.
Exercises to Chapter IV, pp.213-220: #8 and #9 incorporated into II.3:1. Notation of #12 clarified at end
of exercises on IV.6.
Exercises to Chapter V, pp.253-259: Comments on #17 and #21, and hint to #28, given after exercises for
this chapter. #24 is covered in discussion Examples involving inseparable extensions (this Companion,
note to p.252 of Lang).
Exercises to Chapter VI, pp.320-332: #6 clarified in discussion following exercises to VI.6.
#8 generalized and discussed in VI.2:1. Converse to #9 given as VI.2:5. #10, #14, #16 clarified
following exercises for VI.2. #25 generalized in VI.14:2. Strengthened version of #26 asked for in
VI.14:1. Suggestion on #29 following exercises to VI.6.
Exercises to Chapter VIII, pp.374-375: #1 clarified/criticized after exercises on VIII.1. #6 corrected
after exercises on VIII.1.
Exercises to Chapter XIII, pp.545-552: #25 incorporated into XIII.4:3. Possible converse to #26 given as
XIII.4:1. #32, with different approach, given as VI.5:1; still another approach noted in Further
observations on linear independence of characters (this Companion, note to p.284 of Lang).
Exercises to Chapter XIV, pp.567-570: #10 incorporated, with clarification, in XIV.3:1.
Exercises to Chapter XVI, pp.637-639: #6 generalized to bimodules in XVI.3:7. #13 developed in detail
(with some corrections noted) in Further criteria for flatness (this Companion, note to p.613 of Lang),
and alternative argument to one implication given as XVI.3:3.
Exercises to Chapter XX, pp.826-832: #20 incorporated into XX.4:1.
Exercises to Appendix 2, pp.892-893: Hint to #13 given after exercises to that Appendix.

G. M. Bergman Companion To Lang's Algebra, p.167

Uploaded by

Copyright:

Available Formats

G. M. Bergman Companion To Lang's Algebra, p.167

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

G. M. Bergman Companion To Lang's Algebra, p.167

Uploaded by

Copyright:

Available Formats

G. M. Bergman Companion to Langs Algebra, p.

countably generated group with uncountably many subgroups of finite index.

K(M ), then the canonical map : M K(M) is an isomorphism.

is a group homomorphism, and that in this case, we will have B

I.11:1. Let f : A B be a morphism in a category C. Then f is called a monomorphism if for every

some multiplicative set S.

which the coefficient of each monomial y i (i > 0) is zero) is not Noetherian.

(i) f ( X )/( X 1 ) is irreducible in k(1 )[X ].

R-module. (I wont ask you to verify this this time.)

XIII.5:2. Let k be a field, and an infinite cardinal.

x y = 0. How that Hom R (P, R) = 0.

(d) Prove versions of (a) and (b) for noncommutative R.

essentially the same proofs.

A group G is called subdirectly irreducible if the only representations of G as a subdirect product

INDEX TO DISCUSSION OF LANGS EXERCISES IN THIS COMPANION

You might also like