Minimum Counterexamples in Group Theory

Author(s): Desmond MacHale

Source: Mathematics Magazine, Vol. 54, No. 1 (Jan., 1981), pp. 23-28
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2689377
Accessed: 19-06-2016 01:58 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access
to Mathematics Magazine

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
 However, even when a conjecture about finite groups turns out to be false, the question
naturally arises as to which group furnishes us with a counterexample of smallest possible order.
The search for counterexamples is an important aspect of the teaching of modem algebra, and
the finding of minimum counterexamples, or "least criminals" as they are sometimes called,
gives the student an excellent opportunity of becoming familiar with the groups of "small"
order. Indeed, it is remarkable just how many conjectures can be refuted with a knowledge of
the structures of the groups of order 12 or less.
In this paper we give a minimum counterexample for each of a number of not implausible
conjectures about finite groups. Proofs can be found in standard references on group theory, but
we have included a few arguments to illustrate the search technique. We also suggest many
further problems, solved and perhaps unsolved, in this area. We work with groups given by
generators and defining relations, or with groups given by their faithful permutation representa-
tions. We note that a minimum counterexample need not be unique since nonisomorphic
minimum counterexamples (of the same order) may possibly exist for a given conjecture (as in
Conjecture 1 below).
For handy reference throughout this note we list in TABLE 1 the notation which we shall use,
and in TABLE 2 the groups of order less than 12.

G a finite group Order Distinct nonisomorphic groups

Z(G) the center of G I Cl

G' the commutator subgroup of G

IGI the order of G

C. the cyclic group of order n 3 C3
Dn the dihedral group of order 2n 4 C4, C2 x C2
AAn the alternating group of order n !/2 5 C5
Q-- the quaternion group of order 8 6 C6 C2 X C3, S3 -D3
AXB the direct product of A and B 7 C7

F the set {X2 1 x E G) of all squares in G 8 C8, C4 X C2, C2 X C2 X C2, D4, Q

G-H the groups G and H are isomorphic
H<JG H is a normal subgroup of G 9 C9, C3 x C3
Aut G the group of automorphisms of G 10 CI0 _C2 X C5, D5
<)(G) the Frattini subgroup of G (see [21) [ 1 C1l

TABLE 1 TABLE 2

We will use standard notation for generators and relations for a group. For example,
S3 =<a, bIa3 =b2 = 1, bab=a2> denotes the fact that the symmetric group S3 can be described
as the group whose elements are products of two elements a, b subject to the conditions that a is
of order 3, b is of order 2, and the triple product bab collapses to a2. This "generators and
relations" notation allows us to avoid the tedious use of multiplication tables.

CONJECTURE 1. In any group G, the set F of all squares of elements of G is a subgroup of G.

If G is Abelian and x2, y2 are elements of F, then x2y2 =(Xy)2, so F is closed and clearly
nonempty and thus a subgroup of G. This fact rules out as counterexamples all groups of order
less than 12 with the possible exceptions of S3, D4, Q and D5.
Now S3 D3 =<a, bIa3 =b2 = 1, bab =a2 > and direct calculation shows that F= { 1, a, a2) is
a subgroup of S3. Next, for D4 =<a, bla4 =b2 -1, bab=a3>, F= {1, a2}, which is a subgroup of
D4. Similarly, Q=<a, bIa44 =1, a2 =b2, b 'ab=a3> giving F= {1, a2), which is a subgroup of
Q. Finally, D5 =<a, bIa5 =b2 = 1, bab=a4> giving F= {1, a, a2, a3, a4), a subgroup of D5.

24 MATHEMATICS MAGAZINE

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
 FIGURE 1. The rotation group of the regu-
.zz a's \ttt lar tetrahedron permutes the vertices, and
hence can be identified with A 4. This group
X 2 is a minimum counterexample to several
conjectures.

Thus a minimum counterexample has order at least 12, and we will now show that A4, of
order 12, is a minimum counterexample. Representing A4 as a permutation group, we have

A4 = {e,(12)(34),(13)(24),(14)(23),(123),(132),(124),(142),(134),(143),(234),(243))
(see FIGURE 1) and

F= {e,(123),(132),(124),(142), (134),(143),(234),(243)}.

So, since I Fl =9, and 9 is not a divisor of IA4 1, F is not a subgroup of A4. Note that the dicyclic
group of order 12 given by <a,b Ia6 = 1, b2 = a3, b'-ab = a-1> is also a minimum counterexam-
ple.

CONJECTURE 2. The converse of Lagrange's theorem is true; i.e., if n divides I GI, then G has a
subgroup of order n.

Since it is known that the converse of Lagrange's theorem is true for all finite Abelian groups,
we can rule out each group of order less than 12 as a minimum counterexample except possibly
S3, D4, Q and D5. For each of these groups we can produce subgroups of appropriate orders, in
terms of the generators already given:

S3: (1), {l,b}, {1,a,a2}, S3.

D4:{l), {1,a2), {l,a,a2,a3}, D4.
Q:{l}, {1,a2), {l,a,a2 a3) Q.
D5:{l), {l,b), {1,a,a2,a3,a4), D5.
Hence none of these groups is a counterexample. However, A4 is a minimum counterexample,
since we can show that it has no subgroup of order 6. Direct calculation shows that the
conjugacy classes in A4 are

{e}, {(12)(34),(13)(24),(14)(23)}, ((123), (134), (243), (142))

and

((132), (143), (234), (124)).

If H were a subgroup of order 6 in A4, then, since [A4: H] =2, H< A4. Therefore H must
consist of complete conjugacy classes of A4 and of course eE-H. The five nonidentity elements
of H must be made up by taking complete classes with either 3 or 4 elements, an impossibility.
Thus H cannot exist and so A4 is a minimum counterexample.

CONJECTURE 3. If A and B are subgroups of G such that BK< A and AK< G, then BK< G.

Since every subgroup of an Abelian group is normal, the only group of order less than 8 to be
examined is S3. Now {1)< {1, a, a2KS3 is the only relevant normal chain in S3, and since
{11 < S3, S3 does not produce a counterexample. However, D4 of order 8 is a minimum
counterexample, as we now prove. Since D4=<a,bIa4=b2=1, bab=a3>, we take B={1,b),

VOL. 54, NO. 1, JANUARY 1981 25

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
A = {1, a2, b, a2b}. Then B <2A, since A is Abelian, and A K D4 since [1D4: A]=2, but clearly B is
not normal in D4 since a -'ba B.

CONJECTURE 4. All groups of odd order are Abelian.

This conjecture might be optimistically made on the strength of Feit and Thompson's remarka-
ble result that all groups of odd order are soluble. We have the following well-known results:

(a) Groups of order p or p2 are Abelian, where p is a prime.

(b) If p and q are distinct primes with p > q and if q does not divide p-1, then there is a unique
group of order pq and this group is the Abelian group CPq Cp X Cq.

These two results eliminate all groups of odd order less than 21 as counterexamples. However,
there is a group G of order 21, given by G = <a, b I a7= b3 = 1, b l'ab = a2>, and G is clearly
nonAbelian, so G is the required minimum counterexample.

CONJECTURE 5. Z(G) is a fully invariant subgroup of G, i.e., Z(G) is mapped into Z(G) by
every endowtorphism of G.

If G is Abelian or Z(G) = ( 1), then Z(G) is fully invariant, so all groups of order less than 12
are ruled out as counterexamples, except possibly D4 and Q. In both of these groups it is easy to
show that Z(G)=G', so Z(G) is fully invariant because it is known that G' is fully invariant.
However, D6 = <a, b I a6 = 1=b2, bab=a5> of order 12 is a minimum counterexample. In D6 the
mapping a->b, b->b, induces an endomorphism 0 of D6 such that (a3)0=b, where a3 EZ(D6)
and b Z(D6).

CONJECTURE 6. If N< G, then G contains a subgroup isomorphic to the factor group G/N.

This very natural conjecture is in fact true for finite Abelian groups, so S3 is the only group of
order less than 8 which needs to be considered. Now {1}, (1, a, a2), and S3 are the normal
subgroups of S3, and the corresponding factor groups are isomorphic to S3, C2, and C1.
However, S3 has subgroups isomorphic to each of these groups. But the quaternion group Q is a
minimum counterexample. Let Q=<a, bIa4 = 1, a2 =b2, b 'ab=a3>. Then <a2> =A< Q and
Q/Az=C2 x C2. However, Q has no subgroup isomorphic to C2 x C2, since Q has only one
element of order 2.

We close with a number of conjectures, all of which are false. The reader is challenged to
produce a minimum counterexample in each case. At the time of writing, minimum counterex-
amples to those conjectures marked with an asterisk were unknown to the author. Some of these
are likely to present considerable difficulty, and the author would welcome comments or
solutions.

CONJECTURE 7. In any group G, the set of all commutators forms a subgroup.

CONJECTURE 8. If every subgroup of G is normal, then G is Abelian.

CONJECTURE 9. If every proper subgroup of G is cyclic, then G is cyclic.

CONJECTURE 10. If every proper subgroup of G is cyclic, then G is Abelian.

CONJECTURE 1 1. If every proper subgroup of G is Abelian, then G is Abelian.

CONJECTURE 12. Every normal subgroup of G is characteristic in G.

CONJECTURE 13. Every characteristic subgroup is fully invariant in G.

CONJECTURE 14. Given a group G, there exists a group H such that G=-H'.

CONJECTURE 15. Given a group G, there exists a group H such that G=Aut H.

26 MATHEMATICS MAGAZINE

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
 CONJECTURE 16. Given a nonAbelian group G, there exists a finite group H such that G = Aut H.

CONJECTURE 17. Given a nonAbelian group G, there exists a finite group H such that G=
H/Z(H).

CONJECTURE 18. Given a group G, there exists a finite group H such that G ??(H).

CONJECTURE 19. Every group G has a subgroup of prime index.

CONJECTURE 20. If G is a simple group, then G is Abelian.

CONJECTURE 21. Every insoluble group G is simple.

CONJECTURE 22. If G is a group with trivial center, then G = G'.

CONJECTURE 23. If G is a group with G = G', then G has trivial center. (See [3], page 56.)

CONJECTURE 24. If every group of order I G I is cyclic, then I G I must be a prime number (ignore
IGI= 1).
CONJECTURE 25*. If an automorphism a of G sends every conjugacy class of G onto itself, then a
must be inner. (See [2], page 23.)

CONJECTURE 26*. If G has a fixed-point-free automorphism, then G is nilpotent. (See [3], page
336.)

CONJECTURE 27*. If G is nonAbelian, then Aut G is nonAbelian.

CONJECTURE 28*. If G is nonAbelian, then lAut G I is even, i.e., every nonAbelian group has an
automorphism of order 2.

CONJECTURE 29. If G is nonAbelian, then G is a 2-generator group.

CONJECTURE 30. If G is a nonAbelian p-group, then Aut G cannot also be a p-group.

CONJECTURE 31*. If G is nonAbelian and IGI is odd, then G has an outer automorphism.

CONJECTURE 32. If H is a proper subgroup of G, then lAut G I > lAut HI. (See [2], page 24.)

CONJECTURE 33. If I G =n> 1, then there are less than n distinct isomorphism classes of groups
of order n.

CONJECTURE 34. If G and H are groups such that G and H have exactly the same number of
elements of each order, then G-H.

CONJECTURE 35. If H is a normal nilpotent subgroup of G such that G/H is nilpotent, then G is
nilpotent.

CONJECTURE 36*. If G is a nonAbelian p-group, where p is odd, then Aut G cannot also be a
p-group.

CONJECTURE 37. The kernel of a Frobenius group G is Abelian.

CONJECTURE 38. If I G I = I HI and Aut G-Aut H, then G-H.

CONJECTURE 39. If H is a subgroup of G, then lb(H) 5??(G). (See [2], page 50.)

CONJECTURE 40. If I G I = I HI, and G and H have the same character table, then G-H.

CONJECTURE 41. In any permutation group G, the product of transpositions, no two of which are
equal, cannot be the identity element of G.

CONJECTURE 42. If A and B are subgroups of G such that A C B c G, where A is characteristic in

G, then A is characteristic in B.

VOL. 54, NO. 1, JANUARY 1981 27

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms
 CONJECTURE 43. If G is a noncyclic group with I G l = n, then G can be faithfully embedded in Sm
for some m < n.

CONJECTURE 44. If G is. a nonAbelian group with I G = n, then G can be faithfully embedded in
Sm for some m < n.

CONJECTURE 45. Any element of G' is the product of at most two elements of F.

CONJECTURE 46. If A and B are normal subgroups of a group G such that A B, then
G/A zG/B.

CONJECTURE 47. If A and B are normal subgroups of a group G such that G/A G/B, then
A -B.

References

[1] H. F. Blichfeldt, L. E. Dickson, and G. A. Miller, Theory and Applications of Finite Groups, Dover
Reprint, New York, 1961.
[2] J. D. Dixon, Problems in Group Theory, Blaisdell, Waltham, Mass., 1967.
[3] D. Gorenstein, Finite Groups, Harper and Row, New York, 1968.

Regions of Convergence
for a Generalized Lambert Series

DONALD H. TRAHAN
Naval Postgraduate School
Monterey, California 93940

In this note, we discuss infinite series of the form

Inz
G(z)=l :lj'
n

where the coefficients an and cn are complex numbers and Cn Z - 1. We will call this type of
infinite series a G-series. A G-series is a power series if, for all n, cn = 0 and a Lambert series if,
for all n, cn = - 1.
In the literature a G-series is usually considered as a generalized Lambert series. For our
investigation it might be better to think of a Lambert series as a generalized power series since
the questions we will consider can be readily understood in terms of what is known about power
series. For example, we inquire-does a G-series have a radius of convergence? Are there
convergence criteria for G-series similar to well-known criteria of power series?
We first consider Lambert series. In general, a Lambert series is analytic at the origin and
therefore has a power series expansion at the origin. Some of these power series expansions are
very interesting. For example, J. H. Lambert found that for I z I < 1
00 n 00

1zn nZ = z + 2Z2 + 2z3 + 3z4 +2z'+ z6 +

n=l n=l

where Tn is the number of divisors of n. More generally, provided r is a real number and I z I < 1,

28 MATHEMATICS MAGAZINE

This content downloaded from 148.235.65.253 on Sun, 19 Jun 2016 01:58:58 UTC
All use subject to http://about.jstor.org/terms