Sonderdruck aus

Arch Math , Vol 54, 225-231 (1990) 0003-889X/90/5403-0225 $ 2 90/0

1990 Birkhauser Verlag, Basel

Groups with finitely many non-normal subgroups

By
N S HEKSTER and H W LENSTRA, JR

1. Introduction. In this paper we descnbe all groups that have only fimtely many
non-normal subgroups Examples of such groups are, of course, groups that do not have
any non-normal subgroups at all It is well-known that the latter groups can be complete-
ly descnbed, s follows

Theorem 1. Let G be a group Then all subgroups of G are normal if and only if G
satisfies one oj the followmg two conditions
(i) G ts abehan,
(n) there exist groups A, B such that

(a) G s A Q B, where Q denotes the quatermon group of order 8,

(b) A is an abehan group with the property that every xe A has fimte odd order,
(c) B is an abehan group with x2 = l for all xe B

For the proof, see [l, Theorem 1254] A group G is called hamtoman if it satisfies
cordition (n) of Theorem l
For a pnme number p, denote by Cp a multiphcatively wntten group that is isomor-
phic to the group of complex roots of umty of p-power order Our mam result is s
follows

Theorem 2. Let G be a group Then the number of non-normal subgroups of G is fimte

if and only if G satisfies one of the followmg three conditions
(i) G is abehan or hamtoman,
(u) G is fimte,
(m) there exist a pnme number p and groups A, B such that

(a) G s A B,
(b) A is a fimte group of order not divisible by p, and it is abehan or hamtoman,
(c) B has a normal subgroup C, contamed m the centre of B,jor which B/C is ajimte
abehan p-group and C 5. Cpm

For the proof we refer to Section 4

Archiv der Mathematik 54 15

226 N S HEKSTTR and H W LENSIRA, JR ARCII MATH

We can give a formula for the number of non-normal subgroups of the groups occur-
rmg in Theorem 2 (m) Let G, p, A, B, C be s m Theorem 2 (m) By [,] we denote the map
B/C B/C -> C that is mduced by the map B B -> C sending (g, h) to g'1 h'1 g h If
Jj, J 2 are subgroups of B/C, then we denote by [Jl, J2] the subgroup of C generated by
the image of Jl J2 under [,] Fmally, if D is a finite p-group, we wnte lpD for the number
of factors p in the order of D, so lpD = (log D)/log p

Proposition 3. Let the notation be s just dened, and let k denote the number oj
subgroups oj A Then the number of non-normal subgroups of G equals
k E(lP[B/C,J}~lp[J,J]) #J,

where J ranges over the set of subgroups oj B/C

The proof is given in Section 2
It is edsy to see from Proposition 3 that a group G s m Theorem 2 (m) does not have
non-normal subgroups dt all if and only if B is abehan, this also follows from Theorem l
The sum appearmg m Proposition 3 is clearly divisible by p Assummg that B is not
abelian one can, more precisely, show the followmg If C equdls the centre of B, then the
sum is congruent to p mod p2, and at least p (p + j), and if C is properly contamed in the
centre of B, then the sum is congruent to 0 mod p2, and at least p2 (p + 2) In particular,
dny infinite group that has non-normdl subgroups at all has at least 6 of them, equality
occurs only for the umque non-abehan group contammg C2, s a central subgroup of
mdex 4
Another consequence is the followmg If the number of non-normal subgroups of a
group is a pnme number, or the square or the cube of a pnme number, then the group
is finite
Let G be d group and an automorphism of G If H is a subgroup of G, we say that
fixes H if H = H The followmg result is needed m the proof of Theorem 2

Proposition 4. Lei G be a group Then the followmg two assertwns are eqmvalent
(i) G is an infinite abehan group, and it has an automorphism that Jixes almost all
but not all subgroups oj G,
(n) there exist a pnme number p and groups A, D such that

(b) A is a jimte abehan group oj order not divisible by p,

(c) D is a non-tnvial finite abehan p-group
The same is true if both m (i) and m (n) (b) "abehan" is replaced by "hamiltoman"
The proof is given in Secon 3

2. Proof of Proposition 3. Let the notation be s m Proposition 3

Any subgroup of G equals a subgroup of A times a subgroup of B, so the proof of
Proposition 3 immedidtely reduces to the case that G = B, which we now assume We
wnte G = G/C
Vol 54, 1990 Groups with finitely many non-normal subgroups 227

Let H be a subgroup of G, and H its image in G Clearly we have [H, H] a H n C, and

H is normal in G if and only if [G, H] c H n C
It follows that the number of non-normal subgroups of G equals X n} D, where the
J D
- _
sum ranges over all pairs of subgroups J <= G, D c C for which [ J, J] c D, [G, J] Z>,
and where n3 D is the number of subgroups H of G with H J and H n C = D
For each J, the number of possible D equals f [G, J] - lp [J, J] Now fix J and D Smce
/) is charactenstic m C it is normal m G Hence n_, equals the number of subgroups of
G/D that map isomorphically to J under the natural map G/D -> G/C Smce D contams
[J, J], the inverse image of J m G/D is abelian From C/ ^ Cp it follows that this
mverse image is isomorphic to Cp, J Thus rij D is the number of subgroups of Cp> J
mapping isomorphically to J, and this number equals Hom (J, C p ) = J
We conclude that / z> = * -^ = (lp [G, J] - lp [J, J]) * J, s required This
J D J D J
proves Proposition 3
3. Proof of Proposition 4.

Lemma 5. Let G be an infinite group, and suppose that G is wntten s the umon of afimte
set and a jimte collection of subgroups Then the fimte set can be omitted from this umon
P r o o f This is an immediate consequence of a lemma of B H Neumann, which
asserts the following If a group is wntten s the umon of finitely many cosets of
subgroups, then the cosets occurnng m that umon belongmg to subgroups of infinite
mdex can be omitted For a proof of this lemma, see [?, (4 4), 3, Lemma 417] This proves
Lemma 5

Lemma 6. Let Gbea group and an automorphism oj G that fixes almost all subgroups
of G Then fixes every infinite subgroup of G
P r o o f Let H c G be an infinite subgroup For every e H - H, the subgroup
generated by clearly belongs to the fimte collecUon of subgroups C l G with C C
Smce for every C there are only finitely many e G with C = <x>, it follows that H -
is fimte Lemma 5 now imphes that H n H = H, so H is contamed m (This can also
be seen without Lemma 5 ) Likewise H is co'itamed m ~ 1 H, so H = H, s required
This proves Lemma 6

Lemma 7. Let G be an abehan group that has an element oj infinite order, and an
automorphism of G that jixes almost all subgroups of G Then fixes all subgroups of G
P r o o f Let T be the subgroup of G consisting of all elements of fimte order By
Lemma 6, one has = xl for every G T Smce G/T is not the umon of two
proper subgroups the sign is mdependent of But G generates G, so either is the
identity on G or maps each e G to x " 1 This imphes Lemma 7

Lemma 8. Let G be an abehan group that contams a subgroup of the form Cp C p ,

where p is a pnme number Let be an automorphism oj G that fixes almost all subgroups
of G Then a fixes all subgroups of G

15*
228 N S HEKSTER and H W LENSTRA, JR ARCH MATH

P r o o f By Lemma 6, it suffices to prove that H = H for every fimte subgroup H

of G We may clearly assume that H is cyclic Then it is easy to see that there
exist subgroups Q and C2 of CpxCp,, both isomorphic to Cp, such that
Q => H n (Cp,o x Cp) and C t n C 2 = {1} The infinite subgroups Ci H, C2 H of G are
fixed by , by Lemma 6, so the same is true for (Cl H) n (C2 H) = H This proves
Lemma 8
We now prove Proposition 4. We treat the "abelian" and the "hamiltoman" case
simultaneously
To prove that (n) imphes (i), let G = A x Cp<, D s in Proposition 4 (n) It is clear that
G is an infinite group that is abelian or hamiltoman, s the case may be Let be any
non-trivial homomorphism D -> Cp We prove that the automorphism of G given by
(a, c, d) = (a, (\ d) fixes almost all but not all subgroups of G
If d e D is such that () l, then clearly the subgroup of G generated by (l,l,d)
is not fixed by It remains to prove that fixes almost all subgroups of G Let p" be
the exponent of D Smce Cpm has only fmitely many elements of order at most p2",
almost any subgroup H of G has an element (a, c, d) with order (c) > p2" Takmg the p"-th
power, we see that any such H also contams an element (l, c', 1) with c' D. Then
{1} x {1} <= <(1, c', 1)> ci H, and since acts modulo {1} {1} s the identity
this imphes that fixes H This proves that (n) imphes (i)
To prove that (i) imphes (u), let G be an infinite abelian or hamiltoman group, and let
be an automorphism of G that fixes almost all but not all subgroups of G
For a pnme number /, let G, be the subset of G consistmg of all elements of fimte /-power
order Smce G is abelian or hamiltoman, each G, is a subgroup of G, and it is clearly fixed
by Usmg Lemma 7 we see that G may be identified with the direct sum of all G, For
any set of pnmes, let Gn be the direct sum of all G, with /
Let be a set of pnmes, and ' its complement, so that G = G Gn Each subgroup
of G is the direct sum of a subgroup of G and a subgroup of GK It follows that at least
one of Gn, Gn has a subgroup that is not fixed by , say this is H c G Then HxG is
not fixed by , so Lemma 6 imphes that GK is fimte
This proves that, for any set of pnmes, one of G, G has a subgroup not fixed by
and the other one is fimte
If G, is non-trivial for mfimtely many l, then we can choose such that both and '
contam mfimtely many such /, contradictmg what wejust proved It follows that almost
all G, are trivial Likewise we obtam a contradiction if G, is infinite for two distmct pnmes
/ Hence there exists a umq u e pnme p such that Gp is infinite, and this G has a subgroup
not fixed by For this pnme the group A = Gip} is fimte, it is either abelian 01 hamil-
tonian, and we have G = A Gp
We now first prove that Gp is abelian If this is not the case, then we have p = 2
and Gp^QxB, where B is an abelian group of exponent 2 In this group, two
elements generate the same subgroup if and only if they are conjugate. Hence the
hypothesis that fixes almost all subgroups imphes that Gp is the umon of a fimte set
F and (J {x e Gp = }, with ranging over the inner automorphisms of G. Smce
'
there are only fimtely many \ Lemma 5 imphes that the fimte set F can be omitted
from the umon Hence <x> = <x> for every xeG,soaH = H for every subgroup
Vol. 54, 1990 Groups with finitely many non-norma] subgroups 229

HofGp, which is a contradiction. This proves that Gp is abelian. In particular, A is abelian

if and only if G is abelian, and A is hamiltonian if and only if G is hamiltonian.
For a non-negative integer m, denote by G (m) the subgroup of all elements of Gp of
order dividing pm. We prove that each G (m) is finite. Suppose that this is not the case.
Then we can choose m so large that G (m) is infinite and contains an element x with
<%> ==t <x>. The hypothesis that fixes almost all subgroups of G implies
that G (m) is the union of a finite sei F and (J {x e G (m): = "}, with ranging over
a
the integers mod pm that are not divisible by p. Lemma 5 now implies that the finite set
F can be omitted from the union. Hence <() = <x> for every x e G (m), contradicting
the choice of m. This proves that all G (m) are finite.
For each m, let C (m) be the subgroup (~] G (m + n)"" of G (m). Since G (m) is finite,
nO

we have C (m) = G (m + n)"" for all n exceeding a bound depending on m. This implies
that C(m + 1)" = C(m) for all m, which readily yields that the set C = (J C(m) is a
m
subgroup of Gp that is isomorphic to the direct sum of t copies of C p , for some
non-negative integer t. Because Cpa, is divisible we have Gp ^ C x D for some subgroup
D of G. If n is such that C(l) = G (l + n)"", then D"" = {!}, so D c G (n) and therefore
D is finite. But Gp is infinite, so we must have t > 0. By Lemma 8 we have t < 2. Therefore
t= l, and C Co.
Since Gj, has a subgroup not fixed by , not every subgroup of G is characteristic.
Hence Gp C, and D is non-trivial.
This proves Proposition 4.
4. Proof of Theorem 2. The if-part of Theorem 2 is clear in the cases (i) and (ii), and
in case (iii) it suffices to refer to Proposition 3.
Before we prove the only //'-pari we derive a series of auxiliary results.

Lemma 9. Let Gbz a group with only finitely many non-normal subgroups. Then every
infinite subgroup of G is normal.

P r o o f . This follows from Lemma 6, applied to inner automorphisms of G.

Lemma 10. Let G be a group that has only finitely many non-normal subgroups, and p
a prime number. Suppose that G contains a normal subgroup C isomorphic to Cp for which
G/C is a finite cyclic group. Then G is abelian.
P r o o f . For almost all % in a generating coset of G modulo C the subgroup <x> is
normal in G. Choose such an x. Then the natural map C > G/<x> is surjective, so G/<(x)
is abelian. Therefore the commutator subgroup G' of G is contained in <x>, so G' is finite.
But G' is the homomorphic image Cx"1 of C, so it is divisible s well. Hence G' = {!}.
This proves Lemma 10.

Lemma 11. Let G be a group that has only finitely many non-normal subgroups, and p
a prime number. Suppose that G contains a subgroup C of finite index that is isomorphic
to C , Then C is contained in the centre of G.
p
230 N S HEKSTER and H W LENSTRA, JR ARCH MATH

P r o o f By Lemma 9 the subgroup C is normal Now apply Lemma 10 to subgroups

generated by C and a smgle element of G This proves Lemma l1

Lemma 12. Let G be a group that has only jmitely many non-normal subgroups, and p
a pnme number Suppose that G contams a normal subgroup C that is isomorphe to Cpm for
which G/C ik a jimte p-group Then G/C 11 abehan
P r o o f By Lemma 9 we can choose a positive integer m such that every non-normal
subgroup of G has order less than pm Denote by C (m) the unique subgroup of C of order
pm Then every subgroup of G contammg C(m) is normal, so G/C (m) is abehan or
harmltonian But the Orders of the elements of G/C (m) are exactly all powers of p, so
G/C (m) is not harmltonian Therefore G IC (m] is abelian, and it follows that G/C is abehan
s well This proves Lemma 12

Lemma 13. Let G be an infinite group that has onlyfimtely many non-normal subgroups,
and that is neither abehan nor harmltonian Then G has a normal subgroup F offimte mdex
that is abehan or harmltonian, and that contams a subgroup H that is non-normal m G

P r o o f We construct a sequence of non-normal subgroups Hl, H2, , of G and a

sequence of normal subgroups F1,F2, , of finite mdex m G with Hl <= Fl m the follow-
ing way
Let Hl be any non-normal subgroup of G, and F1 its normahzer in G All conjugates
of H ! are non-normal m G, so they are finite m number Hence F t is offimte mdex m G,
and by Lemma 9 it is normal
Suppose, mductively, that Ht, F, have been constructed If F, is abehan or hamiltoman,
then the construction stops, and F = F H = H, satisfy the conclusion of the lemma If
Ft is not abelian or harmltonian, we let Hl+1 be any non-normal subgroup of F and F H l
its normahzer in F, Then Fl, l is normal of finite mdex in G
To prove that the process stops, it suffices to show that H, , for ; > But Ht is
normal m Ft, whereas H} is not even normal in the subgroup FJ_1 of F, This proves
Lemma 13
We now prove the only i/-part of Theorem 2
Let G be a group that has only fimtely many non-normal subgroups, and that is not
s in (i) or (u) of Theorem 2, i e, G is infinite, and neither abehan nor hamiltoman Let
F, H be chosen s m Lemma 13, and let be an inner automorphism of G with
Then F is an infinite group that is abehan or hamiltoman, and the restnction of to F
is an automorphism of F that fixes almost all but not all subgroups of F Applvmg
Proposition 4 to F we see that F has a subgroup C of finite mdex that is isomorphic to
Cp Then C is of finite mdex in G, so by Lemma 9 each subgroup of G contammg C is
normal Therefore G/C is abehan or harmltonian, and wc can wnte G/C = A x D, where
D is a finite p-group and A is a finite group of order not divisible by p that is abehan or
hamiltoman Also, C is contamed m the centre of G, by Lemma 11
Let E be the unique subgroup of G for which A = E/C, and let be the order oiA Smce
2
C is uniquely divisible by a, the cohomology group H (A, C) vamshes, so E can be
identified with A x C From A = {x e E x" = 1} we see that A is characterisc m E,
Vol 54, 1990 Groups with fimtely many non-normal subgroups 231

and therefore normal m G The subgroup B of G with D = B/C is also normal m G, and
11 follows that G = A B Fmally, applymg Lemma 12 with B m the role of G we see that
B/C is abelian
This completes the proof of Theorem 2

References
[1] M HALL, JR , The theory of groups New York 1959
[2] B H NEUMANN, Groups covered by permutable subsets J London Math Soc 29, 236-248
(1954)
[3] D J S ROBINSON, Finiteness condions and generahzed soluble groups, Part l Ergeb Math
Grenzgeb (2) 62, Berlin-Hcidelberg-New York 1972

Eingegangen am 5 9 1988
Anschnften der Autoren
N S Hekster H W Lenstra, Jr
IBM The Netherlands Department of Mathematics
Paalbergweg 9-11 Umversity of California
1105 AG Amsterdam Berkeley, California 94720
The Netherlands USA