Sums of element orders in groups of odd order
Marcel Herzog*, Patrizia Longobardi** and Mercede Maj**
*School of Mathematical Sciences
Tel-Aviv University
Ramat-Aviv, Tel-Aviv, Israel
**Dipartimento di Matematica
UniversitΓ di Salerno
via Giovanni Paolo II, 132, 84084 Fisciano (Salerno), Italy
Denote by G a finite group and by Ο(G) the sum of
element orders in G. If t is a positive integer, denote by Ctβ
the cyclic group of order t and write Ο(t)=Ο(Ctβ).
In this paper we proved the following Theorem A: Let G be a
non-cyclic group of odd order n=qm, where q is the smallest
prime divisor of n and (m,q)=1. Then the following statements hold.
(1) If q=3, then
Ο(β£Gβ£)Ο(G)ββ€30185β, and equality holds if
and only if n=3β
7β
m1β with (m1β,42)=1 and
G=(C7ββC3β)ΓCm1ββ, with C7ββC3β non-abelian.
(2) If q>3, then
Ο(β£Gβ£)Ο(G)ββ€p5+1p4+p3βp2+1β,
where p is the smallest prime bigger than q and equality holds if
and only if n=qp2m1β with (m1β,p!)=1 and
G=CqβΓCpβΓCpβΓCm1ββ.
KEYWORDS: Group element orders; Finite groups
MSC[2010] 20D60; 20E34
β β support: This work was supported by the National Group for Algebraic and
Geometric Structures, and their Applications (GNSAGA - INDAM), Italy.
The first author is grateful to the Department of
Mathematics of the University of Salerno
for its hospitality and
support, while this investigation was carried out.
1. Introduction
The problem of detecting structural properties of periodic groups by looking at
the orders of their elements has been considered by various authors, from many
different points of view. In [1], H. Amiri, S.M. Jafarian Amiri and I.M. Isaacs
introduced the function
[TABLE]
where o(x) denotes the order of the element x of the finite group G. They
proved the following theorem:
Theorem 1
Let G be a group of order n, and denote by Cnβ the
cyclic group of order n. Then Ο(G)β€Ο(Cnβ), with equality if and only if
G=Cnβ.
Thus Cnβ is the unique group of order n with the largest value of Ο(G) for
groups of that order. Following this publication, in several recent papers
it was proved
that certain classes of finite groups G can be characterized using the function
Ο(G) (see [2],[3],[5],[6],[8],[9],[10],[12],[14],[15] and [18]).
The problem of determining the second largest value
of Ο for groups of order n and finding
groups satisfying that condition was discussed in several papers, for example [4],[7],
[14]
and [17]. Recently S.M. Jafarian Amiri and M. Amiri in [7] (see also [4]) and R. Shen,
G. Chen and C. Wu in [17] studied finite groups G of order n with the second largest
value of Ο(G), and obtained information about the structure of G if
n=p1Ξ±1βββ―ptΞ±tββ, where piβ are primes satisfying
p1β<β―<ptβ, Ξ±iβ are positive integers and Ξ±1β>1.
Our aim is to continue this investigation in the case when Ξ±1β=1.
In our paper [11] we obtained the following result which improves Theorem 1:
Theorem 2
If G is a non-cyclic group of order n, then Ο(G)β€117βΟ(Cnβ). Moreover, this upper bound is best possible.
Recently we determined all groups G of order n satisfying Ο(G)=117βΟ(Cnβ). In fact, we proved in [14] the following theorem:
Theorem 3
Let G be a non-cyclic group of order n. Then Ο(G)=117βΟ(Cnβ) if and only if n=4m with m an odd integer and G=C2βΓC2βΓCmβ.
In the paper [13], we studied groups of order n=p1Ξ±1βββ―ptΞ±tββ
with piβ and Ξ±iβ as stated above,
where p1β=2 and Ξ±1β=1. We proved the following two theorems.
Theorem 4
Let G be a non-cyclic group of order n=2m, where m is
an odd integer. Then Ο(G)β€2113βΟ(Cnβ). Moreover, Ο(G)=2113βΟ(Cnβ) if and only if G=S3βΓCn/6β, where n=6m1β
with (m1β,6)=1 and S3β is the symmetric group on three letters.
Thus the groups G=S3βΓCn/6β, where n=6m1β and (m1β,6)=1, are the groups
with the second largest value of Ο for groups of order 2m, with m odd.
The second result in [13] is the following theorem.
Theorem 5
Let Ξnβ be the set of non-cyclic groups of the fixed
order n=2m, where m is an odd integer, and suppose that m=p1Ξ±1βββ―ptΞ±tββ, where piβ are distint primes and Ξ±iβ are positive integers
for all i. If GβΞnβ, then
[TABLE]
where l=min(piΞ±iβββ£iβ{1,β¦,t}) and Ο(l)=Ο(Clβ).
Moreover, GβΞnβ satisfies
Ο(G)=(31β+3Ο(l)2lβ)Ο(Cnβ) if and only if
G=D2lβΓCn/2lβ, where D2lβ denotes the dihedral group of order 2l.
Thus if n=2m, with m odd, then D2lβΓCn/2lβ is the unique group of
order n with the second largest value of Ο for groups of that order. In
particular, if l=3, then D2lβ=S3β
and 31β+3Ο(l)2lβ=31β+3β
76β=2113β, in agreement with Theorem 4.
Using Theorem 2, Theorem 3, Theorem 4 and results from [7] and [17], we could
obtain the following result:
Theorem 6
Let G be a non-cyclic group of even order
n=2Ξ±m, with m an odd integer. Then the following statements hold.
where Q8β is the quaternion group of order 8 and G1β=β¨a,bβ£a2Ξ±=b3=1,ba=bβ1β©.
In particular, these upper bounds are best possible.
In the paper [14], we started the investigation in the odd case. By Theorem 3
in [11] we have
Theorem 7
If G is a non-cyclic group of order n and if q
denotes the smallest prime divisor of n, then
Ο(Cnβ)Ο(G)β<qβ11β.
Theorem 7 implies the following corollary.
Corollary 8
If G is a non-cyclic group of odd order n, then
Ο(Cnβ)Ο(G)β<21β.
In [14] we proved the following theorem, which yields a better result
in the odd case.
Theorem 9
If G is a non-cyclic group of order n and if q
denotes the smallest prime divisor of n, then
[TABLE]
and the equality holds if and only if n=q2k with (k,q!)=1 and
G=CqβΓCqβΓCkβ.
As shown in Lemma 2.1 of Section 2, the
function f(x)=x5+1x4+x3βx2+1β is strongly decreasing for xβ₯2.
Notice that f(2)=117β and f(3)=6125β.
Thus Theorem 9 implies the following result.
Theorem 10
If G is a non-cyclic group of order n,
then the following statements hold.
Demonstration Proof
Let q denote the smallest prime divisor of n. If n is even, then
q=2 and by Theorem 9 Ο(Cnβ)Ο(G)ββ€f(2), with equality as
required. If n is odd, then qβ₯3 and by Theorem 9
Ο(Cnβ)Ο(G)ββ€f(q). It follows by Lemma 2.1 that f(q)β€f(3),
so Ο(Cnβ)Ο(G)ββ€f(3)=6125β, as required. The condition
for equality is again as required.
β
Hence if G is a non-cyclic group of odd order, then
Ο(Cnβ)Ο(G)ββ€6125β<21β and the upper bound
6125β is best possible.
In this paper we are looking for a result similar to Theorem 5 for groups of odd order
n. Denote by q the smallest prime divisor of n. As in the case when q=2,
we start with groups of order n=qm, where q>2 and (q,m)=1. We denote by
C7ββC3β the non-abelian group of order 21. We prove the following theorem:
Theorem A
Let G be a
non-cyclic group of odd order n=qm, where q is the smallest
prime divisor of n and (m,q)=1. Then the following statements hold.
Theorem A implies the following corollary.
Corollary B
Let G be a non-cyclic group of odd order n=qm,
where q is the smallest
prime divisor of n and (m,q)=1. Then the following statements hold.
Finally, Theorem A and results of [7], [14] and [17] imply the following corollary.
Corollary C
Let G be a non-cyclic group of odd order n=mqΞ±,
where q is the smallest prime divisor of n, Ξ± is a positive integer and
(m,q)=1. Moreover, let p be the smallest prime bigger than q. Then the following
statements hold.
These upper bounds are best possible.
The proof of Theorem A will be achieved in two steps. The first step is
Proposition D
Let
G be a non-cyclic group of odd order n=qm, where q is the smallest
prime divisor of n and (m,q)=1. Then the following statements hold.
The second step is
Proposition E
Let
β£Gβ£=n=qm, where q is the smallest
prime divisor of n and (m,q)=1. Then the following statements hold.
It is clear that Theorem A follows from Propositions D and E. These
propositions will be proved in Sections 3 and 4, respectively. Section
2 will be devoted to preliminary results and Corollary B will be proved in
Section 5.
Our notation will be the usual one (see for example [16]). In particular,
Ο(n) denotes the Eulerβs function on n. If k is a positive integer,
we write Ο(k)=Ο(Ckβ), and, following [17], Ξ»(k)=kΟ(k)β.
Finally, if X is a subset of G, we shall define Ο(X)=Ξ£xβXβo(x).
2. Preliminary results
First we define two functions of real variables which will play important
roles in our proofs.
Definition Definition
Let q>1 be a positive integer. Then
[TABLE]
Our first result is the following lemma.
Lemma 2.1
The functions f and gqβ are strongly decreasing for
xβ₯2.
Demonstration Proof
We have gqβ²β(x)=(q2βq+1)(x2βx+1)2(q2βq)(1βx2)β<0 for xβ₯2. Furthermore, if xβ₯2, then
f(x)=x4βx3+x2βx+1x3βx+1β and
fβ²(x)=(x4βx3+x2βx+1)2βxh(x)β,
where h(x)=x5β4x3+8x2β7x+2>0. Hence also fβ²(x)<0 for xβ₯2.
β
In sequel, we shall often use without reference the fact that if m and n
are integers
satisfying m>nβ₯2, then f(m)<f(n) and gqβ(m)<gqβ(n).
Our second result is the
following important proposition.
Proposition 2.2
Let qβ₯3 be a prime. Denote by p
the smallest prime bigger than q and by
q1β the smallest
prime congruent to 1Β (modΒ q).
Then the following statements hold.
Demonstration Proof
(1) If q=3, then p=5 and q1β=7. Thus
g3β(q1β)=(72β7+1)(32β3+1)72β7+1+7(32β3)β=30185β and
f(p)=55+154+53β52+1β=521121β<30185β, as
required.
(2) By our assumption q>3.
Clearly q1ββ₯2q+1 and by the Bertrandβs conjecture
pβ€2qβ3. As shown in Lemma 2.1, the functions gqβ(x)
and f(x) are both decreasing for xβ₯2, which implies that
[TABLE]
Therefore it suffices to prove that gqβ(2q+1)<f(2qβ3). So we need to prove that
[TABLE]
Since b+1a+1β>baβ if b>a, it suffices to prove that
[TABLE]
or
[TABLE]
This inequality is equivalent to the inequality
[TABLE]
or
[TABLE]
which is certainly true if q>3.
The proof of the proposition is now complete.
β
In the next lemma, we collect some basic information
about the function Ο.
Lemma 2.3
Let G denote a finite group, p,piβ denote primes
and n,m,Ξ±iβ denote positive integers. Then the following statements hold.
We conclude this section with the following two lemmas which follow easily from
previous results in [13].
Lemma 2.4
Let G=(β¨aβ©Γβ¨bβ©)ββ¨yβ©, where p is an odd prime number and o(a)=pΞ±β1 for some
integer Ξ±>1, o(b)=p, (o(y),p)=1, ay=a and by=br for some integer r
not congruent to 1 modulo p. Then
[TABLE]
Demonstration Proof
By Theorem 3.2 of [13] we have Ο(G)=(pβ1)2Ο(Z)+pΟ(β¨aβ©)Ο(β¨yβ©), where Z=Cβ¨yβ©β(β¨bβ©).
On the other hand,
[TABLE]
Thus Ο(β¨aβ©Γβ¨bβ©)Ο(β¨yβ©)=((pβ1)2+pΟ(β¨aβ©))Ο(β¨yβ©) and
Zββ¨yβ© implies that Ο(Z)<Ο(β¨yβ©).
Hence
[TABLE]
as required.
β
Lemma 2.5
Let G be a non-cyclic group and G=Pββ¨yβ©,
where P is a non-trivial cyclic p-subgroup of G for some prime p and
(β£Pβ£,β£β¨yβ©β£)=1. If q is the smallest prime divisor of
β£β¨yβ©β£, then
[TABLE]
Demonstration Proof
Write F=β¨yβ© and let Z=CFβ(P). Then, by Lemma 2.3(6),
we have Ο(G)=Ο(P)Ο(Z)+β£Pβ£(Ο(F)βΟ(Z)). Thus
Ο(β£Gβ£)Ο(G)β=Ο(P)Ο(F)Ο(G)β=Ο(F)Ο(Z)β+Ο(β£Pβ£)β£Pβ£β(1βΟ(F)Ο(Z)β).
Write β£Pβ£=pΞ±, with Ξ± a positive integer, and if k is a positive
integer, let Ξ»(k)=kΟ(k)β. By Lemma 2.8 of [13] we have
Ξ»(p)β€Ξ»(pΞ±), so Ξ»(pΞ±)1ββ€Ξ»(p)1β. Thus we have
[TABLE]
Arguing as in the proof of Proposition 6 of [14], we have
Ο(F)Ο(Z)ββ€q2βq+11β. Hence
[TABLE]
as required.
β
3. Proof of Proposition D
Demonstration Proof
We assume that G is a non-cyclic group of
odd order β£Gβ£=n=qm with (m,q)=1. This implies
that G=MβC, where CβCqβ and M is its normal
complement.
First assume that q=3. Then β£Gβ£=3m with (3,m)=1 and
suppose that Ο(β£Gβ£)Ο(G)β>30185β.
Our aim is to reach a contradiction. We argue by induction on β£Gβ£.
If G=MΓC, then
Ο(G)=Ο(M)Ο(C), where M is non-cyclic and 3 does not divide
β£Mβ£.
By Theorem 4 of [14] and Lemma 2.1 we have
Ο(M)β€55+154+53β52+1βΟ(β£Mβ£)=521121βΟ(β£Mβ£). Hence
Ο(G)β€521121βΟ(β£Gβ£)<30185βΟ(β£Gβ£), a contradiction. Therefore C is not normal in G.
Let p2β be the largest prime divisor of n. Clearly p2ββ₯5.
By Lemma 2.3(4), we have
Ο(β£Gβ£)β₯p2β+13βn2,
so Ο(G)>30185βΟ(β£Gβ£)β₯301(p2β+1)255βn2.
Hence there exists xβG such that
o(x)>301(p2β+1)255βn, which implies that
[G:β¨xβ©]<255301β(p2β+1)<2p2β.
Suppose, first, that p2β does not divide [G:β¨xβ©]. Then β¨xβ©
contains a cyclic Sylow p2β-subgroup P of G.
Since β¨xβ©β€NGβ(P), it
follows that [G:NGβ(P)]<2p2β. But [G:NGβ(P)]=1+kp2β for some non-negative
integer k,
so 1+kp2β<2p2β and since G is of odd order, [G:NGβ(P)]=1+p2β is
impossible.
Hence
[G:NGβ(P)]=1 and P is a normal cyclic Sylow p2β-subgroup of G.
Now Lemma
2.3(5) and our assumptions
imply that Ο(P)Ο(G/P)β₯Ο(G)>30185βΟ(P)Ο(β£G/Pβ£).
Thus Ο(G/P)>30185βΟ(β£G/Pβ£). Since 3 divides β£G/Pβ£,
it follows by the induction
hypothesis
that G/P is cyclic.
Write G=PβF, where F is isomorphic to
G/P.
Then F is cyclic and 3 divides β£Fβ£. Since C is not normal in G, 3 does
not divide
CFβ(P) and since P is cyclic, it follows by Lemma 2.3(7)
that p2β is congruent to 1 (mod
3). Hence
p2ββ₯7 and g3β(p2β)β€g3β(7). Now
Lemma 2.5 implies that
[TABLE]
a contradiction.
Now suppose that p2β does divide [G:β¨xβ©]. Then [G:β¨xβ©]<2p2β
implies that [G:β¨xβ©]=p2β. By Theorem 3.1 of [13], G=PβF,
where
P is a normal Sylow p2β-subgroup of G,
F is a cyclic subgroup of G, and either
P is cyclic, or G is nilpotent, or
G=(β¨aβ©Γβ¨bβ©)ββ¨yβ©, where
o(a)=p2Ξ±β1β for some integer Ξ±>1, o(b)=p2β, (o(y),p2β)=1,
ay=a and by=br
for some integer r not congruent to 1 (mod p2β). If P is cyclic, then we get
a contradiction arguing as in the previous paragraph. If G is nilpotent,
then C is normal in G,
a contradiction. Finally, in the last case, it follows by Lemma 2.4 that
Ο(G)<Ο(β¨aβ©Γβ¨bβ©)Ο(β¨yβ©).
Applying
Theorem 9 to β¨aβ©Γβ¨bβ©
we obtain Ο(G)<p25β+1p24β+p23ββp22β+1βΟ(β£β¨aβ©Γβ¨bβ©β£)Ο(β¨yβ©). Now
Ο(β£β¨aβ©Γβ¨bβ©β£)Ο(β¨yβ©)=Ο(β£Gβ£)
and since p2ββ₯5, Lemma 2.1 implies that
p25β+1p24β+p23ββp22β+1ββ€55+154+53β52+1β=521121β. Thus
Ο(G)<521121βΟ(β£Gβ£)<30185βΟ(β£Gβ£), a contradiction.
The proof in the case when q=3 is now complete.
We assume, now, that q>3.
Then β£Gβ£=qm with (q,m)=1 and
suppose that Ο(β£Gβ£)Ο(G)β>p5+1p4+p3βp2+1β, where
p is the smallest prime larger than q. Denote by q1β the smallest prime
congruent to 1 modulo q. Our aim is to reach a contradiction.
We argue by induction on β£Gβ£.
If G=MΓC, then Ο(G)=Ο(M)Ο(C), where M is non-cyclic
and q does not divide β£Mβ£. Therefore
Ο(β£Gβ£)Ο(G)β=Ο(β£Mβ£)Ο(M)βΟ(β£Cβ£)Ο(C)β=Ο(β£Mβ£)Ο(M)β and by Theorem 9
Ο(β£Mβ£)Ο(M)ββ€q25β+1q24β+q23ββq22β+1β=f(q2β),
where q2β is the smallest prime divisor of β£Mβ£. Obviously q2ββ₯p, so
f(q2β)β€f(p) and
Ο(β£Gβ£)Ο(G)β=Ο(β£Mβ£)Ο(M)ββ€p5+1p4+p3βp2+1β, a contradiction. Therefore C is not normal in G.
Let p2β be the largest prime divisor of n. Then, by Lemma 2.3(4), we have
Ο(β£Gβ£)β₯p2β+1qβn2>p2βqβ1βn2,
so Ο(G)>p5+1p4+p3βp2+1βΟ(β£Gβ£)>(p5+1)p2β(p4+p3βp2+1)(qβ1)βn2. Hence there exists xβG such that
o(x)>(p5+1)p2β(p4+p3βp2+1)(qβ1)βn, which implies that
[G:β¨xβ©]<(p4+p3βp2+1)(qβ1)(p5+1)p2ββ. By Bertrandβs
conjecture pβ€2qβ2, so qβ1β₯2pβ and
[TABLE]
First, suppose that p2β does not divide [G:β¨xβ©]. Then β¨xβ©
contains a cyclic Sylow p2β-subgroup P of G
and it follows, as in the q=3 case, that P is a normal subgroup of G.
Now Lemma
2.3(5) and our assumptions
imply that Ο(P)Ο(G/P)β₯Ο(G)>p5+1p4+p3βp2+1βΟ(P)Ο(β£G/Pβ£).
Thus Ο(G/P)>p5+1p4+p3βp2+1βΟ(β£G/Pβ£). Since
q divides β£G/Pβ£, it follows by the inductive
hypothesis
that G/P is cyclic.
Write G=PβF, where F is isomorphic to G/P. Then F is
a cyclic group
and q divides β£Fβ£.
As shown in the q=3 case, p2β is congruent to 1 (mod
q), implying that p2ββ₯q1β and gqβ(p2β)β€gqβ(q1β).
Since by Proposition 2.2(1) gqβ(q1β)<f(p), applying Lemma 2.5 we
obtain
[TABLE]
a contradiction.
Now suppose that p2β does divide [G:β¨xβ©]. Since [G:β¨xβ©]<2p2β,
it follows that [G:β¨xβ©]=p2β. Then, by Theorem 3.1 of [13],
G=PβF, where
P is a normal Sylow p2β subgroup of G, F is a cyclic subgroup of G, and
either
P is cyclic, or G is nilpotent, or
G=(β¨aβ©Γβ¨bβ©)ββ¨yβ©, where
o(a)=p2Ξ±β1β for some integer Ξ±>1, o(b)=p2β, (o(y),p2β)=1,
ay=a and by=br
for some integer r not congruent to 1 (mod p2β). If P is cyclic, then we get
a contradiction arguing as in the previous paragraph. If G is nilpotent, then C is
normal in G,
a contradiction. Finally, in the last case, it follows by Lemma 2.4 that
Ο(G)<Ο(β¨aβ©Γβ¨bβ©)Ο(β¨yβ©).
Applying
Theorem 9 to β¨aβ©Γβ¨bβ©
we obtain Ο(G)<p25β+1p24β+p23ββp22β+1βΟ(β£β¨aβ©Γβ¨bβ©β£)Ο(β¨yβ©). Now
Ο(β£β¨aβ©Γβ¨bβ©β£)Ο(β¨yβ©)=Ο(β£G))
and p2ββ₯p implies that
p25β+1p24β+p23ββp22β+1ββ€p5+1p4+p3βp2+1β. It follows
that
Ο(G)<p5+1p4+p3βp2+1βΟ(β£Gβ£), a contradiction.
The proof of Proposition D is now complete.
3. Proof of Proposition E
Demonstration Proof
Let q1β be the smallest prime congruent to 1 (mod q).
Recall that p is the smallest prime larger than q,
f(x)=x5+1x4+x3βx2+1β and
gqβ(x)=(x2βx+1)(q2βq+1)x2βx+1+x(q2βq)β.
First assume that q=3.
Then q1β=7 and p=5. First we show that if G=CkβΓ(C7ββC3β)
where (k,42)=1, then Ο(β£Gβ£)Ο(G)β=g3β(7)=30185β. Indeed,
Ο(G)=Ο(Ckβ)Ο(C7ββC3β)=Ο(Ckβ)(72β7+1+14β
3)=Ο(Ckβ)85,
so Ο(β£Gβ£)Ο(G)β=Ο(Ckβ)Ο(C7β)Ο(C3β)Ο(Ckβ)85β=30185β, as claimed.
Now we turn to the βonly ifβ part.
So suppose that q=3, β£Gβ£=n=3m, m is odd, (m,3)=1 and
Ο(β£Gβ£)Ο(G)β=g3β(7)=30185β.
Then G is non-cyclic and G=MβC, where CβC3β
and 3 does not divide β£Mβ£. In particular,
the least prime divisor r of β£Mβ£ satisfies rβ₯5.
Our aim is to show that G=CkβΓ(C7ββC3β)
where (k,42)=1. We argue by induction on β£Gβ£.
If G=MΓC, then M is non-cyclic and Ο(G)=Ο(M)Ο(C).
Moreover, since rβ₯5, Theorem 9 and Lemma 2.1 imply that
Ο(M)β€55+1(53β5+1)(5+1)βΟ(β£Mβ£)=521121βΟ(β£Mβ£).
Hence Ο(β£Gβ£)Ο(G)β=Ο(β£Mβ£)Ο(C)Ο(M)Ο(C)ββ€521121β<30185β, a contradiction.
Therefore C is not normal in G.
Let p2β be the largest prime divisor of β£Gβ£ and let P be a Sylow
p2β-subgroup of G. Then, by Lemma 2.3(4), we have
Ο(β£Gβ£)β₯p2β+13βn2 and our assumptions imply that
Ο(G)=30185βΟ(β£Gβ£)β₯301(p2β+1)255βn2.
Hence there exists xβG such that
o(x)>301(p2β+1)255βn, which implies that
[G:β¨xβ©]<255301β(p2β+1)<2p2β.
First, suppose that p2β does not divide [G:β¨xβ©]. Then
β¨xβ©
contains a cyclic Sylow p2β-subgroup P of G and as shown in the proof
of Proposition D, P is a normal cyclic Sylow subgroup of G.
Now Lemma 2.3(5) and our assumption imply that
Ο(P)Ο(G/P)β₯Ο(G)=30185βΟ(P)Ο(β£G/Pβ£),
with equality if and only if P is central in G.
Hence Ο(G/P)β₯30185βΟ(β£G/Pβ£). Notice that β£G/Pβ£=3m1β,
with m1β odd and (m1β,3)=1.
If G/P is non-cyclic, then
by Proposition D the above inequality implies that
Ο(G/P)=30185βΟ(β£G/Pβ£). Hence
Ο(P)Ο(G/P)=Ο(G) and by Lemma 2.3(5) P is central in G.
Thus G=PΓF, where FβG/P. Moreover, by induction
G/P=CkβΓ(C7ββC3β), where (k,42)=1, and since
FβG/P, we may also write F=CkβΓ(C7ββC3β),
where (k,42)=1. Since (β£Pβ£,β£G/Pβ£)=1, it follows that (β£Pβ£,42k)=1
and G=Cβ£Pβ£βΓF=Ckβ£Pβ£βΓ(C7ββC3β),
where (kβ£Pβ£,42)=1, as required.
So assume that G/P is cyclic.
Write G=PβF, where F is isomorphic to G/P.
Then F is cyclic and 3 divides β£Fβ£. Since C is not normal in G, 3
does not divide
CFβ(P). Therefore, by Lemma 2.3(7), C3β acts fixed-point-freely on P,
implying that p2β is congruent to 1 (mod 3).
Hence p2ββ₯q1β and g3β(p2β)β€g3β(q1β).
Applying Lemma 2.5 we
obtain
[TABLE]
Hence p2β=q1β=7 and
Ο(β£Gβ£)Ο(G)β=7(p22ββp2β+1)p22ββp2β+1+6p2ββ. The second equality implies,
by the proof of Lemma 2.5,
that β£Pβ£=p2β=7. Since p2β is the largest prime divisor of β£Gβ£,
it follows that G=C7ββC3βC5cβ for some non-negative integer c.
Since no 5-element can act fixed -point-freely on C7β, it follows by
Lemma 2.3(7) that
C5cβ acts trivially on C7β. Since 3
does not divide CFβ(P), it follows that G=C5cβΓ(C7ββC3β),
as required.
Now suppose that p2β does divide [G:β¨xβ©]. Since [G:β¨xβ©]<2p2β,
it follows that [G:β¨xβ©]=p2β. Then, by Theorem 3.1 of [13],
G=PβF, where
P is a normal Sylow p2β-subgroup of G, F is a cyclic subgroup of G,
and either
P is cyclic, or G is nilpotent, or
G=(β¨aβ©Γβ¨bβ©)ββ¨yβ©, where
o(a)=p2Ξ±β1β for some integer Ξ±>1, o(b)=p2β,
(o(y),p2β)=1,
ay=a and by=br
for some integer r not congruent to 1 (mod p2β).
Since G/P is cyclic, the previous arguments imply
that if P is cyclic, then G=C5cβΓ(C7ββC3β), where c is
a non-negative integer,
as required. Since C is not normal in G, G is not nilpotent.
Finally, since p2ββ₯5, in the last case we reach
the following contradiction, using
Lemma 2.4, Theorem 9 and Lemma 2.1:
[TABLE]
The proof in the case q=3 is now complete.
Now assume that q>3.
If G=CkβΓCqβΓCpβΓCpβ with (k,p!)=1, then
Ο(G)=Ο(Ckβ)Ο(Cqβ)Ο(CpβΓCpβ)=Ο(Ckβ)Ο(Cqβ)(p+1p4+p3βp2+1β)
and Ο(β£Gβ£)Ο(G)β=(p+1p4+p3βp2+1β)Ο(Cp2β)1β=p5+1p4+p3βp2+1β, as claimed.
Now we turn to the βonly ifβ part.
So suppose that q>3, β£Gβ£=n=qm, m is odd, (m,q)=1 and
Ο(β£Gβ£)Ο(G)β=p5+1p4+p3βp2+1β=f(p).
Then G is non-cyclic and G=MβC, where CβCqβ
and q does not divide β£Mβ£. Denote by r the least prime divisor
of β£Mβ£. Clearly rβ₯p.
Our aim is to show that G=CkβΓCqβΓCpβΓCpβ with (k,p!)=1.
We argue by induction on β£Gβ£.
If G=MΓC, then M is non-cyclic and Ο(G)=Ο(M)Ο(C).
Moreover, since rβ₯p, Theorem 9 and Lemma 2.1 imply that
Ο(M)β€f(r)Ο(β£Mβ£)β€f(p)Ο(β£Mβ£).
Hence
Ο(β£Gβ£)Ο(G)β=Ο(β£Mβ£)Ο(C)Ο(M)Ο(C)ββ€f(p). But by our
assumptions
Ο(β£Gβ£)Ο(G)β=f(p), so r=p and
Ο(β£Mβ£)Ο(M)β=f(p)=f(r). Therefore, by Theorem 9,
M=CpβΓCpβΓCkβ, with (k,p!)=1 and hence
G=CqβΓCpβΓCpβΓCkβ with (k,p!)=1, as required.
Suppose, now, that C is not normal in G.
Let p2β be the largest prime divisor of β£Gβ£ and let P be a Sylow
p2β-subgroup of G. Then, by Lemma 2.3(4), we have
Ο(β£Gβ£)β₯p2β+1qβn2 and our assumptions imply that
Ο(G)=f(p)Ο(β£Gβ£)β₯p2β+1f(p)qβn2.
Hence there exists xβG such that
o(x)>p2β+1f(p)qβn, which implies that
[G:β¨xβ©]<f(p)qp2β+1β.
Now by the Bertrandβs conjecture we have pβ€2qβ3, so
qβ₯2p+3β. Since
f(p)=p5+1p4+p3βp2+1β>p1β, it follows that
f(p)q>2pp+3β and as p2ββ₯p, we get
[TABLE]
First, suppose that p2β does not divide [G:β¨xβ©]. Then
β¨xβ©
contains a cyclic Sylow p2β-subgroup P of G. As shown in the proof
of Proposition D, P is a normal cyclic Sylow subgroup of G.
Now Lemma 2.3(5) and our assumptions imply that
Ο(P)Ο(G/P)β₯Ο(G)=f(p)Ο(P)Ο(β£G/Pβ£).
Hence Ο(G/P)β₯f(p)Ο(β£G/Pβ£). Notice that β£G/Pβ£=qm1β,
with m1β odd and (m1β,q)=1.
Suppose, first, that G/P is non-cyclic.
Then, by Proposition D, the above inequality implies that
Ο(G/P)=f(p)Ο(β£G/Pβ£). Hence
Ο(P)Ο(G/P)=Ο(G) and by Lemma 2.3(5) P is central in G.
Thus G=PΓF, where FβG/P. Moreover, by induction,
G/P=CqβΓCpβΓCpβΓCk1ββ with (k1β,p!)=1 and since
FβG/P, we may also write F=CqβΓCpβΓCpβΓCk1ββ
with (k1β,p!)=1. As (β£Pβ£,β£G/Pβ£)=1, we have (β£Pβ£,k1βpq)=1.
Hence (β£Pβ£,k1β)=1 and as p2β>p, also (β£Pβ£,p!)=1.
Thus G=PΓF=Ck1ββ£Pβ£βΓCqβΓCpβΓCpβ
with (k1ββ£Pβ£,p!)=1, as required.
Suppose, next, that G/P is cyclic.
Write G=PβF, where F is isomorphic to G/P.
Then F is cyclic and q divides β£Fβ£.
Since C is not normal in G, q
does not divide
CFβ(P). Therefore, by Lemma 2.3(7), Cqβ acts fixed-point-freely on P,
implying that p2β is congruent to 1 (mod q).
Hence p2ββ₯p1β and gqβ(p2β)β€gqβ(p1β).
Applying Lemma 2.5 we
obtain
[TABLE]
By Proposition 2.2(1) gqβ(p1β)<f(p), so
Ο(β£Gβ£)Ο(G)β<f(p), in contradiction to our assumptions.
Now suppose that p2β does divide [G:β¨xβ©]. Since [G:β¨xβ©]<2p2β,
it follows that [G:β¨xβ©]=p2β. Then, by Theorem 3.1 of [13],
G=PβF, where
P is a normal Sylow p2β-subgroup of G, F
is a cyclic subgroup of G,
and either
P is cyclic, or G is nilpotent, or
G=(β¨aβ©Γβ¨bβ©)ββ¨yβ©, where
o(a)=p2Ξ±β1β for some integer Ξ±>1, o(b)=p2β,
(o(y),p2β)=1,
ay=a and by=br
for some integer r not congruent to 1 (mod p2β).
If P is cyclic,
then G=PβF, where both P and F are cyclic, and we reach a
contradiction as above.
Since C is not normal in G, G is not nilpotent.
Finally, consider the last case. Since p2ββ₯p, we have
f(p2β)β€f(p). Now, applying
Lemma 2.4 and Theorem 9, we reach the following
contradiction:
[TABLE]
The proof of Proposition E is now complete.
β
5. Proof of Corollary B
Demonstration Proof
Recall that f(x)=x5+1x4+x3βx2+1β.
Let G be a non-cyclic group of odd order n=qm,
where q is the minimum
prime divisor of n and (m,q)=1.
Suppose, first, that q>3. Then, by Theorem A,
[TABLE]
where p is the smallest prime bigger than q.
Since q>3, it follows that qβ₯5, pβ₯7 and by Lemma 2.1
f(p)β€f(7)=2101337β.
Hence if q>3, then
[TABLE]
and by Proposition E equality holds if and only if n=5β
72β
m1β
with
(m1β,7!)=1 and G=C5βΓC7βΓC7βΓCm1ββ,
as required.
Suppose, now, that q=3. Then by Theorem A
[TABLE]
with equality if and only if n=3β
7β
m1β with (m1β,42)=1
and G=(C7ββC3β)ΓCm1ββ.
Since 2101337β<30185β, the result concerning groups with qβ₯3 follows immediately.
β