1 Introduction and preliminaries

An automorphism \(\alpha \) of a group G is called \(\mathrm{IA}\) -automorphism if \( x^{-1}\alpha (x) \in G'\), for all \( x\in G \). This concept was introduced by Bachmuch [1] in 1965. We remark the letters \(\mathrm{I}\) and \(\mathrm{A}\) as to remind the reader that are those automorphisms which induce identity on the abelianized group, \(G/G'\).

Also, if \( x^{-1}\alpha (x) \in Z(G) \) for all \( x\in G \), then we say that \(\alpha \) is a central automorphism, and if \(\alpha \) preserves all conjugacy classes of G, then it is called a class preserving automorphism. The set of all such automorphisms are denoted by \(\mathrm{IA}(G)\), \(\mathrm{Aut_Z}(G)\) and \(\mathrm{Aut_C}(G)\), respectively. These concepts were introduced and studied by Curran [2] in 2001 and Yadav [15, 16] in 2009 and 2013. Clearly, the set of all \(\mathrm{IA}\)-automorphisms, which fix the centre element-wise, forms a normal subgroup of \(\mathrm{IA}(G) \) and is denoted by \(\mathrm{IA_Z}(G) \) (see [11, 12] for more information).

For any element x of a group G and automorphism \(\alpha \) of G, the autocommutator of x and \(\alpha \) is defined as follows:

$$\begin{aligned} {}[x, \alpha ]=x^{-1}x^{\alpha }=x^{-1}\alpha (x). \end{aligned}$$

Now, using the above notation, we have the following:

$$\begin{aligned} \mathrm{Aut_Z}(G)= & {} \lbrace \alpha \in \mathrm{Aut}(G) \mid [x, \alpha ] \in Z(G),\quad \forall x\in G\rbrace , \\ \mathrm{Aut_C}(G)= & {} \lbrace \alpha \in \mathrm{Aut}(G) \mid x^\alpha \in x^{G},\quad \forall x\in G\rbrace , \\ \mathrm{IA_Z}(G)= & {} \lbrace \alpha \in \mathrm{Aut}(G) \mid [x, \alpha ]\in G',~\alpha (z)=z,\quad \forall x\in G~and~z\in Z(G)\rbrace . \end{aligned}$$

One can easily check that any class preserving automorphism is an \(\mathrm{IA}\)-automorphism, which fixes the centre element-wise and hence

$$\begin{aligned} \mathrm{Inn}(G)\le & {} \mathrm{Aut_C}(G) \le \mathrm{IA_Z}(G) \le \mathrm{IA}(G) \le \mathrm{Aut}(G) , \\ Z(\mathrm{Inn}(G))\le & {} \mathrm{Aut_Z}(G) \le \mathrm{Aut}(G). \end{aligned}$$

The following example shows that every \(\mathrm{IA_Z}\)-automorphism is not necessarily inner automorphism.

Example 1.1

Consider the group

$$\begin{aligned} G=\langle a,b,x\mid [a,x]=[b,x]=1, [a,b]=x^{s},~s\ne 1 \rangle . \end{aligned}$$

Clearly, \(G'=\langle x^{s}\rangle ,~~Z(G)=\langle x\rangle \) and \(G/Z(G)=\langle \bar{a},\bar{b}\rangle \cong \mathrm{Inn}(G)\). The \(\mathrm{IA_Z}\)-automorphism \(\alpha \) defined by \(\alpha (a)=ax^{s},~~\alpha (b)=bx^{s},~~\alpha (x)=x\) is a non-inner automorphism.

Note that, \(\mathrm{Aut_Z}(G)\) fixes the derived subgroup \(G'\) element-wise and using this property, we have the following.

Lemma 1.2

For any group G, the central automorphisms commute with \(\mathrm{IA_Z}\)-automorphisms of G.

Proof

Assume \(\alpha \) is a central automorphism of G, then \(\alpha (x)=xz\) for any \(x\in G\) and some element \(z\in Z(G)\). Clearly, every central automorphism \(\alpha \) fixes \(G^{\prime }\) element-wise, and hence for any \(\beta \in \mathrm{IA_Z}(G)\), we have

$$\begin{aligned} \alpha \beta (x)=\alpha (xx^{-1}\beta (x))=\alpha (x) x^{-1}\beta (x)=xzx^{-1}\beta (x)=\beta (\alpha (x))=\beta \alpha (x),\end{aligned}$$

which gives the result. \(\square \)

2 \(\mathrm{IA_Z}\)-automorphisms of a group

Considering the converse of Schur’s Theorem, Niroomand [10] proved that if the derived subgroup \(G'\) of a given group G is finite and its central factor group, G / Z(G), is generated by d elements, then \(|G/Z(G)|\leqslant |G'|^{d}\). Also, Sury [14] generalized this result as follows.

If \(G'\) is finite and G / Z(G) is generated by d elements, then \(|\mathrm{Inn}(G)|\leqslant |K(G)| ^{d}\), where \(K(G)=\langle [x, \alpha ] \mid x\in G, \alpha \in \mathrm{Aut}(G)\rangle \) is the autocommutator subgroup of the group G, see also [4] and [8].

Here, we give a further generalized version of the above result, which improves [10].

Theorem 2.1

Let G be any group with finite derived subgroup. If d is the minimal number of generators of the central factor group of G, then \(|\mathrm{IA_Z}(G)|\leqslant |G^{\prime }|^{d}\).

Proof

Assume that G / Z(G) has a minimal set of generators \(x_{1}Z(G), \dots , x_{d}Z(G)\) and \(\alpha \in \mathrm{IA_Z}(G)\), which fixes Z(G) element-wise. Consider the following map:

$$\begin{aligned} \psi :\mathrm{IA_Z}(G) \longrightarrow \underbrace{G' \times G' \times \cdots \times G'}_{d-times}\ , \end{aligned}$$

given by \(\psi (\alpha )=([x_{1}, \alpha ], [x_{2}, \alpha ], \cdots ,[x_{d}, \alpha ])\), for all \(\alpha \in \mathrm{IA_Z}(G)\).

One can easily check that \(\psi \) is injective, as \(\psi (\alpha )=\psi (\beta )\) implies that \([x_i, \alpha ]=[x_i, \beta ]\), for all \(1\leqslant i\leqslant d\) and any \(\alpha , \beta \in \mathrm{IA_Z}(G)\). Hence, \(\alpha = \beta \) and so \(\vert \mathrm{IA_Z}(G) \vert \leqslant \vert G' \vert ^d\). \(\square \)

Remark 2.2

One notes that the above theorem improves the result in [10], as \(\mathrm{Inn}(G)\le \mathrm{Aut_C}(G)\le \mathrm{IA_Z}(G)\).

Clearly, \(\mathrm{Aut_Z}(G)\cap \mathrm{Inn}(G)=Z(\mathrm{Inn}(G))\), for any group G. Now, we use this property to obtain the following.

Lemma 2.3

Let G be a group such that \(Z(G)\le G^{\prime }\). Then \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\) if and only if \(\mathrm{Inn}(G)=Z(\mathrm{Inn}(G))\).

Proof

Assume \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\), then \(\mathrm{Inn}(G)=\mathrm{IA_Z}(G) \cap \mathrm{Inn}(G)=Z(\mathrm{Inn}(G))\).

Conversely, let \(\mathrm{Inn}(G)=Z(\mathrm{Inn}(G))\). Then \(G' \le Z(G)\) and so the definition of \(\mathrm{IA_Z}(G)\) implies that \(\mathrm{IA_Z}(G) \le \mathrm{Aut_Z}(G)\). On the other hand, by the assumption \(\mathrm{Aut_Z}(G) \le \mathrm{IA_Z}(G)\). Therefore, \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\). \(\square \)

The next result shows that the subgroup \(Z(\mathrm{IA_Z}(G))\) is between \(Z(\mathrm{Inn}(G))\) and \(\mathrm{Aut_Z}(G)\).

Proposition 2.4

For any group G,

$$\begin{aligned} \mathrm{Aut_Z}(G) \cap \mathrm{IA_Z}(G) = Z(\mathrm{IA_Z}(G)). \end{aligned}$$

Proof

Suppose that \(\alpha \in \mathrm{Aut_Z}(G) \cap \mathrm{IA_Z}(G)\) and \(\beta \in \mathrm{IA_Z}(G)\). Lemma 1.2 implies that \([\alpha ,\beta ]=id\)., as \(\alpha \in \mathrm{Aut_Z}(G)\) and hence \(\alpha \in Z(\mathrm{IA_Z}(G))\).

Conversely, for any \(\alpha \in Z(\mathrm{IA_Z}(G))\) and \(\beta \in \mathrm{Inn}(G)\le \mathrm{IA_Z}(G)\), we have \([\alpha ,\beta ]=id\). This implies that \(\alpha \in C_{\mathrm{Aut}(G)}(\mathrm{Inn}(G))=\mathrm{Aut_Z}(G)\), and so \(\alpha \in \mathrm{Aut_Z}(G) \cap \mathrm{IA_Z}(G)\). \(\square \)

In 1940, Hall [3] introduced the concept of isoclinism between two groups and it was extended to n-isoclinism, which is an equivalent relation among all groups and it is weaker than isomorphism. In 1976, Leedham-Green and McKay [7] extended this concept to isologism with respect to a given variety of groups. There have been extensive studies in this area of mathematics (see [5, 6], for more information).

Definition 2.5

Let G and H be arbitrary groups and assume \(\alpha :G/Z(G)\rightarrow H/Z(H)\) and \(\beta :G' \rightarrow H'\) be isomorphisms such that the following diagram is commutative:

where \(f_1(g_1Z(G), g_2Z(G))=[g_1, g_2]\) and \(f_2(h_1Z(H), h_2Z(H))=[h_1, h_2]\), for each \(h_i\in \alpha (g_iZ(G))\), \(i=1, 2\), and \(\beta \) is an isomorphism induced by \(\alpha \). Then \((\alpha ,\beta )\) is said to be isoclinism, so that G and H are called isoclinic and denoted by \(G{\sim }H\).

Here, we characterize all finite p-groups, in which their \(\mathrm{IA_Z}\)-automorphisms are equal to central automorphisms. Yadav [16] proved that if two finite groups G and H are isoclinic, then \(\mathrm{Aut_C}(G) \cong \mathrm{Aut_C}(H)\). With the same assumption, Pradeep [11] showed that \(\mathrm{IA_Z}(G) \cong \mathrm{IA_Z}(H)\) and he applied his result to prove the following.

Theorem 2.6

([11], Theorem B(1)) Let G be a finite p-group. Then \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\) if and only if \(Z(G)=G'\).

The following results can be deduced using the above theorem.

Proposition 2.7

Let G be a d-generating finite p-group and \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\). Then \(|\mathrm{IA_Z}(G)|=|G'| ^{d}\).

Proof

The assumption and Theorem 2.6 imply that G / Z(G) is abelian. Hence, assume \(G/Z(G)=\langle \bar{x_{1}}\rangle \times \langle \bar{x_{2}}\rangle \times \cdots \times \langle \bar{x_{d}}\rangle \), where \(\bar{x_{i}}=x_{i}Z(G)\) is of order \(p^{n_{i}}\), for some positive integer \(n_i\), \(i=1, \ldots , d\). Since \(Z(G)=G' \le \Phi (G)\), by [16] Lemma 3.5, \(\lbrace x_{1},\ldots ,x_{d}\rbrace \) is the minimal generating set for G. On the other hand, \(\mathrm{Hom}(\langle \bar{x_{i}}\rangle ,G^{\prime })=G^{\prime }\), and hence Corollary 2.2 [11] implies that

$$\begin{aligned} \vert \mathrm{IA_Z}(G)\vert =\vert \mathrm{Hom}(G/Z(G) ,G')\vert =\Pi _{i=1}^{d}\vert \mathrm{Hom}(\langle \bar{x_{i}}\rangle ,G')\vert =\Pi _{i=1}^{d}\vert G' \vert =\vert G' \vert ^{d}. \end{aligned}$$

\(\square \)

Theorem 2.8

Let G and H be isoclinic finite p-groups, and \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\). Then \(\mathrm{Aut_Z}(H)=\mathrm{IA_Z}(H)\) if and only if |G| = |H|.

Proof

Assume that \(\mathrm{Aut_Z}(H)=\mathrm{IA_Z}(H)\). The assumption and Theorem 2.6 imply that \(G'=Z(G)\). The isoclinic property of G and H implies that

$$\begin{aligned} |Z(G)| =|G'|=|H'|=|Z(H)|. \end{aligned}$$

Therefore,

$$\begin{aligned} |H| =|H/Z(H)| |Z(H)| =|G/Z(G)||Z(G)| =|G|. \end{aligned}$$

Conversely, if \(|G| =|H|\) then \( |Z(G)|=|Z(H)|\), as G and H are isoclinic. On the other hand, \(Z(G)=G'\cong H'\) and hence \(\vert H' \vert = \vert Z(H) \vert \). Clearly, G is nilpotent of class 2 and isoclinic with H, which imply that H is also nilpotent of class 2. Then \(H'\le Z(H)\) and so \(H'=Z(H)\). Now, Theorem 2.6 implies that \(\mathrm{Aut_Z}(H)=\mathrm{IA_Z}(H)\). \(\square \)

The following example gives a class of isoclinic groups, in which their \(\mathrm{IA_Z}\)-automorphisms are the same as central automorphisms.

Example 2.9

Consider the group

$$\begin{aligned}&G_{i}=\left\langle a, a_{1}, a_{2}, \cdots , a_{2s} \mid a^{p^{r+t}}=1, a_{1}^{p^{r}} = \cdots = a_{2s}^{p^{r}} = a^{p^{i}}, \right. \\&\qquad \qquad \qquad \qquad \qquad \qquad \left. [a_{1}, a_{2}]=[a_{2}, a_{3}]= \cdots =[a_{2s-1}, a_{2s}]=a^{p^{t}} \right\rangle , \end{aligned}$$

where \(r, s, t \geqslant 1\) and \(1\leqslant i\leqslant r\).

Clearly, the group \(G_{i}\) is nilpotent of class 2 and it is of order \(p^{r(2s+1)+t}\). Also one can easily see that

$$\begin{aligned} Z(G_{i})\cong C_{p^{r+t}}, G'_{i}\cong C_{p^{r}}, \frac{G_{i}}{Z(G_{i})}\cong (C_{p^{r}})^{2s}. \end{aligned}$$

Now, Proposition 3.2 of [13] and Lemma 2.3 imply that

$$\begin{aligned} \mathrm{Aut_Z}(G_{i})=\mathrm{IA_z}(G_{i})=Z(\mathrm{Inn}(G_{i}))=\mathrm{Inn}(G_{i})\cong (C_{p^{r}})^{2s}, \end{aligned}$$

as the derived subgroup of the group \(G_{i}\) is cyclic, for all \(1\leqslant i\leqslant r\).

The following lemma of Morigi [9] is useful for our further investigation.

Lemma 2.10

([9], Lemma 0.4) Let G be a finite nilpotent group of class 2. Then \(\exp (G^{\prime })=\exp (G/Z(G))\) and in the decomposition of G / Z(G) into direct product of cyclic groups at least two factors of maximal orders must occur.

We recall that the centre of any non-abelian p-group of order \(p^{n}\) lies between \(p^{2}\) and \(p^{n-2}\).

Theorem 2.11

Let G be a non-abelian p-group with \(\mathrm{Aut_Z}(G)=\mathrm{IA_Z}(G)\). If \(p^{3}\le |G|\le p^{7}\), then the derived subgroups of such groups are either cyclic or elementary abelian p-groups.

Proof

By the assumption and Theorem 2.6, \(Z(G)=G^{\prime }\). It is sufficient to show that Z(G) is cyclic or elementary abelian. The result is obvious, when G is of order \(p^{3}\) or \(p^{4}\). Now assume G is of order \(p^{5}\), then \(|Z(G)|=p^{2}\) or \(p^{3}\). If \(|Z(G)|=p^{3}\), then \(|G/Z(G)|=p^{2}\). By Morigi’s Lemma, we have

$$\begin{aligned} \exp (Z(G))=\exp (G^{\prime })=\exp (G/Z(G))=p. \end{aligned}$$

Hence, Z(G) is elementary abelian. Suppose that \(|G|=p^{6}\). If \(|Z(G)|=p^{2}\) or \(p^{4}\), then Morigi’s Lemma gives the result. Assume \(|Z(G)|=p^{3}\) and \(Z(G)\cong C_{p^{2}}\times C_{p}\). Then

$$\begin{aligned} \exp (Z(G))=\exp (G^{\prime })=\exp (G/Z(G))=p^2. \end{aligned}$$

Therefore,

$$\begin{aligned} G/Z(G)\cong C_{p^{2}}\times C_{p}, \end{aligned}$$

which contradicts Morigi’s Lemma and so \(Z(G)\cong C_{p^{3}}\) or \(C_{p}\times C_{p}\times C_{p}\).

Finally, assume \(|G|=p^{7}\). If \(|Z(G)|=p^{2}\) or \(p^{5}\), then again the result is obtained, by Morigi’s Lemma. Assume Z(G) is of order \(p^{3}\) and \(Z(G)\cong C_{p^{2}}\times C_{p}\), then

$$\begin{aligned} G/Z(G)\cong C_{p^{2}}\times C_{p^{2}}. \end{aligned}$$

So G has two non-commutative generators and hence \(G^{\prime }\) is cyclic, which contradicts the property that \(Z(G)=G^{\prime }\).

If \(|Z(G)|=p^{4}\) and

$$\begin{aligned} Z(G)\cong C_{p^{2}}\times C_{p}\times C_{p},~ C_{p^{3}}\times C_{p} \quad \mathrm{or} \quad C_{p^{2}}\times C_{p^{2}}, \end{aligned}$$

then

$$\begin{aligned} \exp (Z(G))=\exp (G^{\prime })=\exp (G/Z(G))=p^{2}\quad \mathrm{or} \quad p^{3} \end{aligned}$$

and \(|G/Z(G)|=p^{3}\), which contradicts Morigi’s Lemma and so,

$$\begin{aligned} Z(G)\cong C_{p^{4}} \quad \mathrm{or} \quad C_{p}\times C_{p}\times C_{p}\times C_{p}. \end{aligned}$$

\(\square \)