Abstract
Let \(\alpha \) be an automorphism of a finite group \(G\) and assume that \(G = \big \{\,[g , \alpha ] : g \in G\,\big \}\cdot C_G (\alpha )\). We prove that the order of the subgroup \([G, \alpha ]\) is bounded above by \(n^{\log _2 (n+1)}\) where \(n\) is the index of \(C_G (\alpha )\) in \(G\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(A\) be a finite group that acts on the finite group \(G.\) In the case where \((|G|,|A|)=1\), there are several very useful relations between the groups \(G\) and \(A\), some of which are as follows: (i) \(G=[G,A]\cdot C_G(A)\), (ii) \([G,A,A]=[G,A]\) and (iii) \(C_{G/N} (A) = C_G (A)N/N\) for any \(A\)-invariant normal subgroup \(N\) of \(G\). Almost all of the research papers studying this kind of action concerned with the situations where the fixed point subgroup \(C_G(A)\) has a restricted structure. However, Parker and Quick [1] considered a dual situation by assuming that the index of \(C_G(A)\) is bounded. As this assumption clearly gives no restriction to \(C_{G}(A)\), they focused their attention on the group \([G,A]\) and proved that \(|[G,A]|\leqslant n^{\log _2 (n+1)}\) if \(|G:C_G(A)|\leqslant n.\)
We consider here a special noncoprime action in view of [1]:
Let \(\alpha \) be an automorphism of the finite group \(G\) such that for every \(x\in G, x=[g,\alpha ]\cdot z\) for some \(g\in G\) and \(z\in C_G(\alpha )\).
In the literature a finite group \(G\) admitting such an automorphism \(\alpha \) is called an \(\alpha \) -CCP group where the acronym CCP stands for “centralizer commutator product”. Lemma 2.1 below shows that nice relations indicated above which are valid in the case of a coprime action also survive in the setting of \(\alpha \)-CCP groups. The study of \(\alpha \)-CCP groups was started by Stein [2] who proved that the subgroup \([G, \alpha ]\) is solvable. The goal of the present paper is to give an upper bound for the order of \([G, \alpha ]\) in terms of the index of \(C_G (\alpha )\) in \(G\). Namely, we prove the following:
Theorem A
Let \(G\) be an \(\alpha \)-CCP group such that \(| G : C_G (\alpha )|\leqslant n\). Then \(|[G,\alpha ] | \leqslant n^{\log _2 (n+1)}\).
An internal reformulation of Theorem A can be stated as
Theorem B
Let \(H\) be a finite group containing an element x such that \(H=\big \{[h,x] : h\in H\big \}\cdot C_H(x)\). If \(|H:C_H(x)|\leqslant n\) then \(|[H,x]|\leqslant n^{log_{2}(n+1)}\).
Theorem A is the \(\alpha \)-CCP analogue of [1, TheoremA]. The key lemma in our proof is Lemma 3.1 which we obtain as the \(\alpha \)-CCP analogue of [1, Lemma 2.1]. The rest of the paper contains the proof of Theorem A and some technical results pertaining to the proof of Theorem A; all of which are proven in a similar fashion as in the proofs of [1, Proposition 2.2], [1, Corollary 2.3] and [1, TheoremA] with obvious changes, namely using Lemma 3.1 instead of [1, Lemma 2.1]. For the sake of completeness we present a proof here for each of them.
In Sect. 2 we state and prove some preliminary facts about \(\alpha \)-CCP groups. Section 3 is concerned with our key lemma, namely Lemma 3.1, and its consequences. We prove our main result Theorem A and its equivalent Theorem B in Sect. 4.
All groups are assumed to be finite. The notation and terminology are standard.
2 Preliminaries on \(\alpha \)-CCP groups
Lemma 2.1
The following hold for any \(\alpha \)-CCP group \(G\).
-
(i)
\(G=[G,\alpha ]\cdot C_G(\alpha )\) and \([G,\alpha , \alpha ]=[G,\alpha ]\). Furthermore \(G=[G,\alpha ]\times C_G(\alpha )\) whenever \(G\) is abelian.
-
(ii)
Every \(\alpha \)-invariant subgroup \(S\) of \(G\) is also an \(\alpha \)-CCP group and we have \(\big \{[x,\alpha ] : x\in S\big \}=\big \{[g,\alpha ] : g\in G\big \}\cap S.\)
-
(iii)
\(G/N\) is an \(\alpha \)-CCP group for any \(\alpha \)-invariant normal subgroup \(N\) of \(G\).
-
(iv)
If \([g, \alpha ]^f \in C_G (\alpha )\) for some \(f\) and \(g\) in G, then \(g\in C_G (\alpha )\).
-
(v)
\(C_{G/N} (\alpha ) = C_G (\alpha ) N / N\) for any \(\alpha \)-invariant normal subgroup \(N\) of \(G\).
-
(vi)
\(\big \{\,[g, \alpha ] : g\in G \,\big \}\) is a transversal to \(C_G (\alpha )\). Furthermore \(\alpha ^G\) is a transversal to \(C_H (\alpha ^{a})\) for any \(a\in G\) in the semidirect product \(H = G \langle \alpha \rangle \).
Proof
This lemma gives almost the same information as in [2, Proposition 2.2] on an \(\alpha \)-CCP group \(G\). We need only to show that \(G=[G,\alpha ]\times C_G(\alpha )\) when \(G\) is abelian: Notice that \([G,\alpha ]=\big \{\,[g,\alpha ] : g\in G \,\big \}\) when \(G\) is abelian and also observe that for any \([g,\alpha ]\in C_G(\alpha ) \), we have \([g,\alpha ]=1\) by Lemma 2.1(iv). \(\square \)
The following lemma is crucial in proving our key lemma Lemma 3.1.
Lemma 2.2
Let \(G\) be an \(\alpha \)-CCP group and set \(H = G \langle \alpha \rangle \). Then
-
(i)
the map \(f_{\alpha ^{a}} : \alpha ^G \longrightarrow \alpha ^G\) defined by \(f_{\alpha ^{a}} (\alpha ^g) = (\alpha ^{a})^{\alpha ^g}\) is a bijection for any \(a \in G\),
-
(ii)
for any \(X\leqslant H\) with \(X \cap \alpha ^G \ne \phi \) and for any \(\alpha ^{a} \in X\) we have
$$\begin{aligned} (\alpha ^{a})^X = (\alpha ^{a})^{X \cap \alpha ^G}= X \cap \alpha ^G. \end{aligned}$$
Proof
\(\alpha ^G\) is a transversal to \(C_H (\alpha ^{a})\) by Lemma 2.1(vi). If \(g\) and \(h\) are elements of \(G\) such that \((\alpha ^{a})^{\alpha ^g } = (\alpha ^{a})^{\alpha ^h}\), then \(\alpha ^g (\alpha ^h)^{-1} \in C_H (\alpha ^{a})\) and so \(\alpha ^g = \alpha ^h\). This proves \((i)\) since \(\alpha ^G\) is finite.
It is straightforward to verify that \((\alpha ^{a})^{X \cap \alpha ^G} \subseteq (\alpha ^{a})^X \subseteq X \cap \alpha ^G\). If \(\alpha ^y \in X \cap \alpha ^G\), then \(\alpha ^y = (\alpha ^{a})^{\alpha ^h}\) for some \(h \in G\) by part (i). This yields \(\alpha ^y \in (\alpha ^{a})^{\alpha ^G}\). Notice that \(f_{\alpha ^{a}} (X \cap \alpha ^G) \subseteq X \cap \alpha ^G\) as \(\alpha ^{a} \in X\), and so \(f_{\alpha ^{a}} (X \cap \alpha ^G) = X \cap \alpha ^G\) since \(f_{\alpha ^{a}}\) is a bijection. Then \(\alpha ^h \in X\) and hence \(X \cap \alpha ^G \subseteq (\alpha ^{a})^{X \cap \alpha ^G}\) which establishes the claim (ii). \(\square \)
3 Some technical lemmas pertaining to the proof of Theorem A
The following results are modifications of Lemma 2.1, Proposition 2.2 and Corollary 2.3 in [1] for \(\alpha \)-CCP groups.
Lemma 3.1
Let \(G\) be an \(\alpha \)-CCP group and let \({\mathcal O} = {\alpha }^G\). If \(I \subseteq {\mathcal O}\) and \(\Theta \) is an orbit of \(\langle I \rangle \) on \({\mathcal O}\), then \(\langle I \rangle \leqslant \langle \Theta \rangle \). Furthermore if some member of \(\Theta \) is not contained in \(\langle I \rangle \), then \( \langle I \rangle < \langle \Theta \rangle \).
Proof
To ease the notation set \(K = \langle I \rangle \) and let \(\Theta = ({\alpha }^x)^K\). It should be noted that \(K \langle \Theta \rangle \) is a subgroup of \(G\) because \(K\) normalizes \(\langle \Theta \rangle \). Set now \(L = K \langle \Theta \rangle \). Since \(L \cap {\alpha }^G \ne \phi \) and \({\alpha }^x \in L\), we have
by Lemma 2.2(ii). Then, for any generator \(\alpha ^y\) of \(K\), we have
This completes the proof. \(\square \)
Lemma 3.2
When \(G\) is an \(\alpha \)-CCP group the group \(\langle {\alpha }^G\rangle \) can be generated by \(\log _2 \Big ( \frac{2 (n+p-1)}{p}\Big )\) conjugates of \(\alpha \) where \(p\) is the smallest positive divisor of the order of \(\alpha \) and \(|G:C_G(\alpha )|\leqslant n\).
Proof
We let \({\mathcal O} = {\alpha }^G\) and consider the action of \(\langle {\alpha }\rangle \) on \({\mathcal O}\) by conjugation. Suppose first that \(\langle \alpha \rangle \) has a fixed point \({\alpha }^x\) which is different from \(\alpha \). Then \([\alpha , x] \in C_G (\alpha )\) and hence \([\alpha , x]= 1\) by Lemma 2.1(iv). This contradiction shows that \(\alpha \) is the only fixed point of \(\langle \alpha \rangle \) in its action on \({\mathcal O}\).
Define \(K_0 = 1\), \(K_1 = \langle \alpha \rangle \) and for \(j > 1\), \(K_j = \langle K_{j-1}, {\alpha }_j \rangle \) where at each stage \({\alpha }_j \in {\mathcal O}\) is chosen to maximize the order of \(K_j\). Since \(G\) is finite, there exists \(k\) such that \(K_k = \langle {\alpha }^G\rangle \) and \(K_{k-1} \ne \langle {\alpha }^G\rangle \). Now \(\langle {\alpha }^G\rangle = \langle \alpha _1, \alpha _2, \ldots , \alpha _k\rangle \) where \(\alpha _1=\alpha \). Fix \(j \in \{1,\ldots ,k\}\) and let \(I=\{\alpha _1, \alpha _2, \ldots , \alpha _j\}\). Now \(K_j=\langle I\rangle .\) Choose an orbit \(\Theta \) of \(K_j \) with representative \(B\) where \(B\nleq K_j\). Then \(K_j< \langle \Theta \rangle \) by Lemma 3.1. If \(\Theta \) were also an orbit of \(K_{j-1}\), then we would have
contradicting the choice of \(\alpha _j.\) Therefore \(\Theta \) is a union of at least two orbits of \(K_{j-1}\) on \(\mathcal {O}.\) Notice also that \(B\nleq K_i\) for each \(i=1, \ldots ,j\). Thus \(\Theta \) is a union of at least \(2^{j-1}\) orbits of \(\langle \alpha \rangle \) on \(\mathcal {O}\), each of which has length at least \(p.\) Since \(\alpha _j\leqslant K_j\) for \(i\leqslant j\) we see that the set \(\Omega = \{\alpha _1\}\cup \{{\alpha _2}^{K_1}\}\cup \ldots \cup \{{\alpha _i}^{K_{i-1}}\}\) is contained in \(K_j.\) Therefore \(\Omega \cap {\alpha _{i+1}}^{K_i}\) is empty as \(\alpha _{i+1}\nleq K_i.\) Then \(\mathcal {O}\supseteq \{\alpha _1\}\cup \{{\alpha _2}^{K_1}\}\cup \ldots \cup \{{\alpha _k}^{K_{k-1}}\}\) and the right hand side is a disjoint union. So
Consequently we have \(k-1 \leqslant \log _2 \big ( \frac{n+p-1}{p}\big )\) as claimed. \(\square \)
Lemma 3.3
Let \(G\) be an \(\alpha \)-CCP group. Suppose that \(G\) is a \(p\)-group for some prime \(p\) with \(|G : C_G (\alpha )|\leqslant p^m \). Then \(|[G, \alpha ]| \leqslant p^{\frac{m^2 + m}{2}}\).
Proof
Firstly we handle the case where \(G\) is of class at most two by induction on the order of \(G.\) By Lemma 2.1(i) we have \([G,\alpha ]=[G,\alpha ,\alpha ]\) and \(G/G'=[G/G',\alpha ]\times C_{G/G'}(\alpha )\). Then \(G=[G,\alpha ]\) by induction and hence \(C_{G/G'}(\alpha )=1\), that is \(C_G(\alpha )\leqslant G'\). Thus \(|G:G'|\leqslant p^m.\) In this case the proof is in a similar fashion as in the proof of [1, Proposition 3.1]. For the sake of completeness we present it here. Let the abelian group \(\bar{G}=G/G'\) be the direct product of nontrivial cyclic subgroups \(\langle \bar{x_i}\rangle \) for \(i=1,\ldots ,d\) where \(|\bar{x_i}|=p^{m_i}.\) We have \(G=\langle x_1,\ldots ,x_d\rangle \) since \(G'\leqslant \Phi (G).\) It is straightforward now to verify that \(G'=\langle [x_j,x_i] : 1\leqslant i<j\leqslant d \rangle \) since \(G'\leqslant Z(G).\) Set \(H_i=\langle x_{i+1},\ldots ,x_d, G'\rangle \). Then \(G'=\prod _{i=1}^{d-1}[H_i,x_i]\) for \(i=1,\ldots ,d-1.\) We have \(|[H_i,x_i]|\leqslant |H_i/G'|=p^{m_{i+1}+\cdots +m_d}\) due to the fact that \(h\longmapsto [h,x_i] \) defines a homomorphism from \(H_i/G'\) onto \([H_i,x_i]\). Thus \(|G|\leqslant \prod _{i=1}^{d}p^{m_i}\prod _{i=1}^{d-1}|[H_i,x_i]|\leqslant p^M\) where \(M=\sum _{i=1}^{d}im_i.\) It can be proven by induction on \(d\) that \(M\leqslant (m^2+m)/2\). This completes the proof when \(G\) is of class at most two.
Suppose now that \(G\) has class \(c\) with \(c\geqslant 3.\) Again assume \(|G|\) minimal, therefore \(G=[G,\alpha ].\) The proof in this case is in a very similar fashion as in the proof of [1, Theorem B]. Note that \(\gamma _{c-1}(G)\) is abelian. We also observe that \([\gamma _{c-1}(G),\alpha ]\ne 1\), because otherwise \([\gamma _{c-1}(G),\alpha , G]=1=[G,\gamma _{c-1}(G),\alpha ]\) and hence \(\gamma _{c-1}(G)\leqslant Z(G)\) by the Three Subgroup Lemma. Let now \(N\) be of minimal order among all normal \(\alpha \)-invariant subgroups of \(G\) contained in \(\gamma _{c-1}(G)\) and are not centralized by \(\alpha \). Let \(|G/N:C_{G/N}(\alpha )|=p^r\). As \(C_{G/N}(\alpha )=C_G(\alpha )N/N\) by Lemma 2.1(v) we have \(|G:C_G(\alpha )N|=p^r\). Note that \(G/N\) and \(N\) are both \(\alpha \)-CCP groups by Lemma 2.1(i). It follows by induction that \(|[G/N,\alpha ]|\leqslant p^{\frac{r^2 + r}{2}}.\) As \([G/N,\alpha ]=[G,\alpha ]N/N=G/N\) we have \(|G/N|\leqslant p^{\frac{r^2 + r}{2}} \). Let now \(|N:C_N(\alpha )|=p^s.\) Since \(N\) is abelian we have \(N=[N,\alpha ]\times C_N(\alpha )\) and so \(|[N,\alpha ]|=p^s\). It remains to bound \(|N/[N,\alpha ]|\) suitably. As \(N\) is contained in \(\gamma _{c-1}(G)\) we have \([N,G]\leqslant \gamma _{c}(G)\leqslant Z(G).\) Hence for \(g\in G\) the map \(x\mapsto [x,g]\) for \(x\in N\), is a homomorphism with kernel \(C_N(g)\), in particular \([N,G]\) lies in the kernel and \(|N:C_N(g)|=|[N,g]|\). Set now \(H=[N,\alpha ][N,G]\). Observe that \(1\ne [N,\alpha ]=[N,\alpha ,\alpha ]\leqslant [H,\alpha ]\) by Lemma 2.1(i). It follows by minimality of \(N\) that \(H=N\). Thus
We also observe that \([N,G']=1\) by the three subgroup Lemma as \([N,G,G]=1=[G,N,G]\). This gives that \(NC_G(\alpha )\leqslant G'\leqslant C_G(N)\). As \(N\leqslant \gamma _{c-1}(G)\leqslant G'\) we get \(NC_G(\alpha )\leqslant G'\leqslant C_G(N).\) Therefore \(|G:C_G(N)|\leqslant p^r.\) Let \(Y\) be a minimal generating set for \(G\) modulo \(C_G(N).\) Then \(|Y|\leqslant r.\) Since \([N,G]\leqslant Z(G) \) we also see that \([N,G]=\prod _{y\in Y}[N,y]\). Thus \(|[N,G]|\leqslant |[N,\alpha ]|^{|Y|}\leqslant p^{sr} \). So \(|N|=|[N,G][N,\alpha ]|\leqslant p^{s(r+1)} \) whence \(|G|=|G/N|\cdot |N|\leqslant p^{(r^2+r)+s(r+1)}\). This establishes the claim as
\(\square \)
4 Proof of the main results
In this section we present a proof of Theorem A and deduce Theorem B.
Proof of Theorem A
Let \(G\) be a minimal counterexample to the theorem. Then \(G = [G,\alpha ]\) by induction as \([G,\alpha ] = [G,\alpha ,\alpha ]\) by Lemma 2.1(i). As a consequence \(C_G (\alpha ) \leqslant G'\), and \(G\) is nonabelian. The main result of [2] gives that the group \(G\) is solvable and hence \(F (G)\ne 1\). If \(\left[ F(G),\alpha \right] = 1\), then \(G \leqslant C_G \left( F (G)\right) = Z\left( F (G)\right) \) by the Three Subgroup Lemma, which is a contradiction as \(G\) is nonabelian. Thus \(\left[ F (G),\alpha \right] \ne 1\) and hence there is a prime \(p\) dividing \(|F(G)|\) such that \(\left[ O_p(G),\alpha \right] \ne 1.\) Notice that if \(\left[ Z_2(O_p(G)),\alpha \right] =1\), then \(Z_2\left( O_p(G)\right) \leqslant Z(G)\) by the Three Subgroup Lemma as \([G,\alpha ]=G\). This forces \(O_p(G)=Z_2(O_p(G))=Z(O_p(G))\) which contradicts the fact that \(\left[ O_p(G),\alpha \right] \ne 1.\) Let \(Q\) be minimal element of the set \(\{S : S \) is a normal \(\alpha \)-invariant subgroup of \(G\) which is contained in \(Z_2\left( O_p(G)\right) \) such that \([S,\alpha ]\ne 1\}\). Clearly \([Q',\alpha ]=1\) by the minimality of \(Q\) and so \(Q'\leqslant Z(G)\) by the Three Subgroup Lemma. Set now \(Q_0= \langle [Q,\alpha ]^G \rangle \). Note that both \(Q\) and \(|G/Q|\) are \(\alpha \)-CCP groups. So we have \( [Q,\alpha ]=[Q,\alpha ,\alpha ]\) by Lemma 2.1 \((i)\). Thus \(1\ne \big [[Q,\alpha ]\big ]\leqslant [Q_0,\alpha ]\) and hence \(Q=Q_0\) by the minimality of \(Q\). Now \(|QC_G(\alpha ):C_G(\alpha )|=|Q:C_Q(\alpha )|=p^m\) for some \(m.\) Let \(|G:QC_G(\alpha )|=r\). Then \(r\leqslant {\frac{n}{p^m}}.\) We observe by Lemma 3.3 that \(|[Q, \alpha ]|\leqslant p^{\frac{m^2 + m}{2}}.\) Set \(R=C_{[Q,\alpha ]}(\alpha )\). Then \(R\leqslant [Q,\alpha ]'\leqslant Q'\) and hence \(R\leqslant Z(G)\). Now
So \(|R|={\frac{|[Q,\alpha ]|}{p^m}}\leqslant p^{\frac{m^2 - m}{2}}\). It remains to bound \(|G/R|\) suitably.
Set \(\bar{G}=G/R.\) The group \(\bar{Q}\) is the product of at most \(log_2(r+1)\) of the conjugates of \([\bar{Q},\alpha ]\) in \(\bar{G}\): To see this let \(H=G\rtimes \langle \alpha \rangle .\) Note that \(Q\rtimes H\) and \(C_G(\alpha )\langle \alpha \rangle Q\leqslant N_H(Q\langle \alpha \rangle )\). Set \(\tilde{H}=H/Q\). Now \(|\tilde{H}:N_{\tilde{H}}(\langle \tilde{\alpha } \rangle )|\leqslant |H:Q\langle \alpha \rangle C_G(\alpha )|=r.\) By Lemma 3.2 \(\langle {\langle \tilde{\alpha } \rangle }^{\tilde{H}}\rangle \) can be generated by at most \(k=log_2 (r+1)\) conjugates of \(\langle \tilde{\alpha } \rangle \). That is \(\langle {(\langle \tilde{\alpha } \rangle )}^{\tilde{H}}\rangle =\langle \tilde{\alpha _1},\ldots ,\tilde{\alpha _k}\rangle \) where each \(\alpha _i\) is a conjugate of \(\alpha \) and \(\alpha _1=\alpha .\) Note that \(H=[G,\alpha ]\langle \alpha \rangle ={\langle {\alpha }^H \rangle }C_G(\alpha )=MQC_G(\alpha )\) where \(M=\langle \alpha _1,\ldots ,\alpha _k\rangle C_G(\alpha )\). Therefore
We are now ready to complete the proof of Theorem A. By the above paragraph we have \(|\bar{Q}|=|\langle [\bar{Q},\alpha ]^{\bar{G}}\rangle |\leqslant |[\bar{Q},\alpha ]|^k=p^{mk}\) and so \(|Q|\leqslant p^{mk+(\frac{m^2 - m}{2})}\). Notice that \(|G/Q|\leqslant r^k\) by induction. Thus
This contradiction completes the proof of Theorem A. \(\square \)
Remark 4.1
As indicated in the introduction one can reformulate Theorem A as Theorem B. Their equivalence can be easily seen as follows:
Suppose that Theorem A is true. Set \(H=G\rtimes \langle \alpha \rangle \) and \(x=\alpha \) in \(H.\) Then \([G,\alpha ]=[H,x]\) and \(\big \{[g,\alpha ] : g\in G\big \}=\big \{[h,x] : h\in H\big \}\) and \(|G:C_G(\alpha )|=|H:C_H(x)|=n.\) Therefore \(|[G,\alpha ]|=|[H,x]\leqslant n^{log_2 (n+1)}\) by Theorem A. Conversely suppose that Theorem B is true and let \(H\) be a finite group containing an element x such that \(H=\big \{[h,x] : h\in H\big \}C_H(x)\) holds. Set \(G=H\) and let \(\alpha \) denote the inner automorphism of \(G\) induced by \(x\). Then by applying Theorem B we have \(|[H,x]|\leqslant n^{log_2 (n+1)}\) as desired.
References
Parker, C., Quick, M.: Coprime automorphisms and their commutators. J. Algebra 244, 260–272 (2001)
Stein, A.: A conjugacy class as a transversal in a finite group. J. Algebra 239, 365–390 (2001)
Acknowledgments
The author thanks the referee for his/her careful reading and some corrections.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Ercan, G. Finite groups having centralizer commutator product property. Rend. Circ. Mat. Palermo 64, 341–346 (2015). https://doi.org/10.1007/s12215-015-0192-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12215-015-0192-z