On central automorphisms fixing the center element-wise
Manoj K. Yadav
School of Mathematics, Harish-Chandra Research Institute
Chhatnag Road, Jhunsi, Allahabad - 211 019, INDIA
[email protected]
Abstract.
Let G be a finite p-group of nilpotency class 2. We find necessary and
sufficient conditions on G such that each central automorphism of G fixes
the center of G element-wise.
2000 Mathematics Subject Classification. 20D45, 20D15
Research supported by DST (SERC Division), the Govt. of INDIA
1. Introduction
Let G be a finite group. By Z(G), γ2(G), Aut(G) and Inn(G),
we denote the center, the commutator subgroup,
the group of all automorphisms and the group of all inner automorphisms
of G respectively. For x∈G, [G,x] denotes the set {[g,x]=g−1x−1gx∣g∈G}.
To say that some H is a subgroup (proper subgroup) of G we write
H≤G (H<G). For any group H and an abelian group K,
Hom(H,K) denotes the group of all homomorphisms from H to K.
An automorphism α of G is called central if
x−1α(x)∈Z(G) for all x∈G. The set of all central
automorphisms of G, which is here denoted by Autcent(G), is a normal
subgroup of Aut(G). Notice that Autcent(G)=CAut(G)(Inn(G)), the
centralizer of the subgroup Inn(G) in the group Aut(G).
Let M and N be two normal subgroups of G. By AutN(G)
we mean the subgroup of Aut(G) consisting of all the
automorphisms which centralize G/N and by AutM(G)
we mean the subgroup of Aut(G) consisting of all the
automorphisms which centralize M.
We denote AutN(G)∩AutM(G) by AutMN(G).
Throughout the paper p always denotes a prime number.
In a very recent article of Attar [2], it is proved for a finite
p-group G that
AutZ(G)Z(G)(G)=Inn(G) if and only if G is abelian or G is
nilpotent of class 2 and Z(G) is cyclic. So it may be interesting to
study finite p-groups G such that AutZ(G)Z(G)(G) contains
Inn(G) and coincides with Autcent(G).
Notice that if Inn(G)≤Autcent(G), then the nilpotency class of G is
at most 2. Moreover, if G is abelian with
Autcent(G)=AutZ(G)Z(G)(G), then
it follows that G=1. So we restrict our attention to finite p-groups
of class 2.
Let G be a finite p-group of class 2.
Then G/Z(G) and γ2(G) have equal exponent pc (say). Let
[TABLE]
where Cpai is a cyclic group of order pai, 1≤i≤r, and
a1≥a2≥⋯≥ar>0. Let k be the largest
integer between 1 and r such that a1=a2=⋯=ak=c. Notice that k≥2. Set Mˉ=M/Z(G)=Cpa1×⋯×Cpak. Let
[TABLE]
where b1≥b2≥⋯≥bs>0,
be a cyclic decomposition of G/γ2(G) such that Mˉ is
isomorphic to a subgroup of Nˉ=N/γ2(G)
:=Cpb1×Cpb2×⋯×Cpbk.
Using the above terminology we state our result in
Theorem**.**
Let G be a finite p-group of nilpotency class 2. Then Autcent(G)=AutZ(G)Z(G)(G) if and only if r=s, (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ and the exponents of Z(G) and γ2(G) are
equal.
A generalization of the main theorem of Attar [2] is given in
Proposition 2.3.
As a consequence of these results, we derive the main theorem of Curran and
McCaughan [4] in Corollary 2.6.
2. Proofs
Let G be a finite group and M be a central subgroup of G. Let α∈AutM(G). Then we can define a homomorphism fα from
G to M
such that fα(x)=x−1α(x). On the other hand, given a
homomorphism
f from G to M, we can always define an endomorphism αf of G
such that αf(x)=xf(x). But αf is an automorphism of G if
and only if for every non-trivial element m∈M, f(m)=m−1 .
The following lemma is an easy exercise.
Lemma 2.1**.**
Let G be a finite group and M be a central subgroup of G. Let M≤∩Ker(f), where f runs over all elements in Hom(G,M).
Then the correspondence f→αf defined in the
preceeding paragraph is a one-to-one mapping from Hom(G,M)
onto AutM(G). Moreover, this correspondence gives
rise to a natural isomorphism between AutZ(G)M(G) and
Hom(G/Z(G),M).
A finite group G is said to be purely non-abelian if it does not
have a non-trivial abelian direct factor. The following lemma is
from [1].
Lemma 2.2**.**
Let G be a purely non-abelian finite group. Then the correspondence α→fα defined above with M replaced by Z(G) is a
one-to-one mapping of Autcent(G) onto Hom(G/γ2(G),Z(G)).
The following proposition generalizes the main result of Attar [2].
Proposition 2.3**.**
Let G be a non-abelian finite p-group and M be a central subgroup of
G. Then AutZ(G)M(G)=Inn(G) if and only if the nilpotency
class of G is 2, γ2(G)≤M and M is cyclic.
Proof.
Since M≤Z(G), f(m)=1 for all f∈Hom(G/Z(G),M) and
m∈M. Thus it follows from Lemma 2.1 that there is a
natural isomorphism between AutZ(G)M(G) and Hom(G/Z(G),M).
Now suppose that AutZ(G)M(G) =Inn(G). Since
Inn(G)=AutZ(G)M(G)≤Autcent(G), it follows that the
nilpotency class of G is at most 2. As
G is non-abelian, it must be of class 2. Thus the exponents of G/Z(G)
and γ2(G) are equal. Let {x1,x2,⋯,xd} be a minimal
generating set for G. Let ij denote the inner automorphism induced by
xj, i.e., ij(g)=xj−1gxj for all g∈G,
where 1≤j≤d. So g−1ij(g)∈M for all g∈G and 1≤j≤d, since Inn(G)=AutZ(G)M(G). But g−1ij(g)=g−1xj−1gxj=[g,xj]. So
{g−1ij(g)∣g∈G}=[G,xj]≤M. Since
γ2(G)=⟨[G,xj]∣1≤j≤d⟩, it follows that γ2(G)≤M. Therefore the exponent of M is greater than or equal to the
exponent of γ2(G). Let the exponent of M be pe.
If possible, suppose that M is not cyclic. Then M=Cpe×N,
where Cpe is a
cyclic subgroup of order pe and N is some non-trivial
proper subgroup of M. Now
[TABLE]
Thus ∣AutZ(G)M(G)∣=∣Hom(G/Z(G),M)∣>∣Inn(G)∣.
This contradicts the given hypothesis. Hence M must be cyclic.
Conversely suppose that the nilpotency class of G is 2, γ2(G)≤M and M is cyclic. Then
[TABLE]
since the exponent of M is at least equal to the exponent of G/Z(G).
This completes the proof of the
proposition.
□
∎
The following lemma can be proved by a counting argument using Corollary 3.3 of
[3]. But here we prove it by a direct argument.
Lemma 2.4**.**
Let G be a finite p-group such that
Autcent(G)=AutZ(G)Z(G)(G). Then G is purely non-abelian.
Proof.
Assume contrarily that G is not purely non-abelian. Then G=H×A, where H is purely non-abelian and A is non-trivial abelian
subgroup of G.
Let {x1,x2,⋯,xr} and {y1,y2, ⋯,ys} be
minimal generating sets for
H and A respectively. Then S:={x1,x2, ⋯,xr,y1,y2,⋯,ys} is a minimal generating set for G.
Obviously Z(H)∩Φ(G)=1, where Φ(G) denotes the Frattini
subgroup of G. So we
can always choose a non-trivial element z∈Z(H)∩Φ(G) such that
zp=1. Define a map f from G to G by f(w)=wz for all w∈S.
Now it follows from the lemma (there is only one lemma in the paper)
of [5] that f is an automorphism of G.
Notice that f∈Autcent(G), but f∈AutZ(G)Z(G)(G). For,
f(yi)=yiz=yi, however yi∈Z(G) for all i such that 1≤i≤s. This contradicts the given hypothesis that Autcent(G)=AutZ(G)Z(G)(G). Hence G must be purely non-abelian. This completes
the proof of the lemma. □
∎
Lemma 2.5**.**
Let A and B be two finite abelian p-groups such that
A=Cpa1×Cpa2 ×⋯×Cpas,
where a1≥a2≥⋯≥as>0, and
B=Cpb1×Cpb2×⋯×Cpbs, where b1≥b2≥⋯≥bs>0.
Let bj≥aj for all j,
1≤j≤s, and bj>aj for some such j. Let t be the smallest
integer between 1 and s such that aj=bj for all j such that
t+1≤j≤s.
Then, for any finite abelian p-group C, ∣Hom(A,C)∣<∣Hom(B,C)∣
if and only if the exponent of C is at least pat+1.
Proof.
Let H=Cpa1×⋯×Cpat, K=Cpb1×⋯×Cpbt and D≅Cpat+1×⋯×Cpas≅Cpbt+1×⋯×Cpbs, since bj=aj for all j such that t+1≤j≤s. Then
[TABLE]
and
[TABLE]
Thus ∣Hom(A,C)∣<∣Hom(B,C)∣ if and only if
∣Hom(H,C)∣<∣Hom(K,C)∣. Since bj≥aj, we have
∣Hom(H,C)∣<∣Hom(K,C)∣ if and only if ∣Hom(H,Cpc)∣<∣Hom(K,Cpc)∣ for, at least, one cyclic group Cpc
which appears
in the cyclic decomposition of C. Notice that ∣Hom(H,Cpc)∣=∣Hom(K,Cpc)∣ if c≤at. Hence ∣Hom(H,Cpc)∣<∣Hom(K,Cpc)∣ if and only if c>at. Since the exponent of
C is at least c, this completes the proof of the lemma. □
∎
Now we are ready to prove our theorem which is stated in Section 1. In the
proof, we use the same terminology as we used in the statement of the theorem.
Proof of the theorem.
We have G/Z(G)=Cpa1×Cpa2×⋯×Cpar and G/γ2(G)=Cpb1×Cpb2×⋯×Cpbs. Since G/Z(G) is a quotient group of
G/γ2(G), we get r≤s and bj≥aj for all j
such that 1≤j≤r.
Suppose that Autcent(G)=AutZ(G)Z(G)(G).
Then G is purely non-abelian by Lemma 2.4. Thus by Lemma
2.2 there is a one-to-one correspondence between
Autcent(G) and Hom(G/γ2(G),Z(G)). Also by Lemma 2.1
(with M=Z(G)) there is a natural
isomorphism between AutZ(G)Z(G)(G) and Hom(G/Z(G),Z(G)).
First we claim that r=s. Suppose, if possible, that r<s.
Since bj≥aj>0 for all j such that 1≤j≤r and bj>0 for
r+1≤j≤s, we have
[TABLE]
This proves that AutZ(G)Z(G)(G)<Autcent(G). But this contradicts
our supposition. Hence r=s and our claim is true. This obviously gives the
inequalities bj≥aj for all j such that 1≤j≤s.
Now we claim that (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ.
Let us suppose contrary, i.e,
(G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ. Then bj>aj for
some j such that k+1≤j≤s. For, if bj=aj for all such j,
then A:=Cpak+1×⋯×Cpas
is isomorphic to B:=Cpbk+1×⋯×Cpbs
and therefore
(G/Z(G))/Mˉ≅A≅B≅(G/γ2(G))/Nˉ, which
we are not taking.
Let t be the smallest integer between k+1 and s such that bj=aj
for all j satisfying t+1≤j≤s. From our choice of k we have
pc>pak+1≥pat, where pc is the exponent of both G/Z(G)
as well as γ2(G). Since the exponent of Z(G) is greater than
or equal to pc, it now follows that the
exponent of Z(G) is at least pat+1. Now applying Lemma
2.5 with C=Z(G), we get
∣Hom(A,Z(G))∣<∣Hom(B,Z(G))∣. So
[TABLE]
since ∣Hom(Mˉ,Z(G))∣≤∣Hom(Nˉ,Z(G))∣.
Thus AutZ(G)Z(G)(G)<Autcent(G). This again contradicts
our supposition. Hence (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ.
Finally suppose that the exponent of Z(G) is not equal to the exponent of
γ2(G). Thus the exponent of Z(G) is at least pc+1, since the
exponent of γ2(G) is pc. Also ∣Z(G)∣>∣γ2(G)∣ and
therefore ∣G/Z(G)∣<∣G/γ2(G)∣.
Since (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ,
it follows that ∣Nˉ∣>∣Mˉ∣. Since Nˉ=Cpb1×⋯×Cpbk and Mˉ=Cpa1×⋯×Cpak, where bj≥aj and aj=c for all j
such that 1≤j≤k, we conclude that bj>aj for some such j.
Let t be the smallest integer between 1 and k such
that bj=aj=c for all j satisfying t+1≤j≤k.
We know that the exponent of Z(G) is at least pc+1=pat+1.
So applying Lemma 2.5 with A=Mˉ, B=Nˉ
and C=Z(G), we get ∣Hom(Mˉ,Z(G))∣<∣Hom(Nˉ,Z(G))∣.
Let D≅Cpak+1×⋯×Cpas≅Cpbk+1×⋯×Cpbs. So
[TABLE]
Thus AutZ(G)Z(G)(G)<Autcent(G), which is again a contradiction.
Hence the exponents of Z(G) and γ2(G) are
equal. This completes the necessary part of the theorem.
Conversely suppose that r=s, (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ and the exponents of Z(G) and γ2(G)
are equal.
Since r=s, it follows that G is purely non-abelian. For, suppose that
G is not purely non-abelian. Then G=H×A, where A is
non-trivial abelian and H is purely non-abelian subgroup of G.
So γ2(G)=γ2(H) and
Z(G)=Z(H)×A. This implies that
[TABLE]
and
[TABLE]
These equations imply that the rank of G/γ2(G) is strictly greater
than the rank of H/γ2(H) and the rank of G/Z(G) is equal to the
rank of H/Z(H), where the rank of a finite abelian p-group X is
defined to be the number of non-trivial cyclic factors in a cyclic
decomposition of X.
Since the nilpotency class of H is 2, we have γ2(H)⊆Z(H). So H/Z(H) is a quotient group of H/γ2(H). Thus
it follows that the rank of H/Z(H) is less than or equal to the rank of
H/γ2(H). Hence the rank of G/γ2(G) is strictly greater
than the rank of G/Z(G). This implies that s>r, because s is the
rank of G/γ2(G) and r is the rank of G/Z(G). This contradition
proves that G is purely non-abelian.
Since r=s and (G/Z(G))/Mˉ≅(G/γ2(G))/Nˉ, we have
G/Z(G)=Cpa1×Cpa2
×⋯×Cpas
and G/γ2(G)=Cpb1×Cpb2×⋯×Cpbs such that Cpak+1×⋯×Cpas
≅Cpbk+1×⋯×Cpbs≅D (say). So
the smallest t, 1≤t≤s such that aj=bj for all j satisfying
t+1≤j≤s, is at the most k.
Now suppose that AutZ(G)Z(G)(G)<Autcent(G).
Then by Lemma 2.1 (with M=Z(G)) and Lemma 2.2,
we have
[TABLE]
since G is purely non-abelian.
So it follows that bj>aj for some j such that 1≤j≤s (to be
more precise, it happens for some j such that 1≤j≤k). For, if bj=aj for all j, then
∣Hom(G/Z(G),Z(G))∣=∣Hom(G/γ2(G),Z(G))∣, which we are not
taking. Now applying
Lemma 2.5 with A=G/Z(G), B=G/γ2(G) and C=Z(G),
we conclude that the exponent of Z(G) must be at least pat+1,
which is
strictly bigger than pak=pc - the exponent of γ2(G). This
contradicts our supposition that Z(G) and γ2(G) have equal
exponents. Hence AutZ(G)Z(G)(G)=Autcent(G),
which completes the proof of the theorem.
□
It is fairly easy to deduce the following corollary from Proposition
2.3 and the theorem.
Corollary 2.6**.**
Let G be a non-abelian finite p-group. Then Autcent(G)=Inn(G) if and
only if Z(G)=γ2(G) and Z(G) is cyclic.
Remark 2.7**.**
It is of interest to find necessary and sufficient conditions for a
finite p-group G of arbitrary nilpotency class such that Autcent(G)=AutZ(G)Z(G)(G).
Acknowledgements. I thank Prof. E. C. Dade for some useful
discussion and the referee for his/her useful comments and suggestions.