Hostname: page-component-cd9895bd7-p9bg8 Total loading time: 0 Render date: 2024-12-26T04:40:32.750Z Has data issue: false hasContentIssue false

ENUMERATION OF GROUPS IN SOME SPECIAL VARIETIES OF A-GROUPS

Published online by Cambridge University Press:  27 August 2024

ARUSHI
Affiliation:
School of Liberal Studies (Mathematics), Dr. B. R. Ambedkar University Delhi, Delhi 110006, India e-mail: arushi.18@stu.aud.ac.in, arushi.garvita@gmail.com
GEETHA VENKATARAMAN*
Affiliation:
School of Liberal Studies (Mathematics), Dr. B. R. Ambedkar University Delhi, Delhi 110006, India
Rights & Permissions [Opens in a new window]

Abstract

We find an upper bound for the number of groups of order n up to isomorphism in the variety ${\mathfrak {S}}={\mathfrak {A}_p}{\mathfrak {A}_q}{\mathfrak {A}_r}$, where p, q and r are distinct primes. We also find a bound on the orders and on the number of conjugacy classes of subgroups that are maximal amongst the subgroups of the general linear group that are also in the variety $\mathfrak {A}_q\mathfrak {A}_r$.

Type
Research Article
Copyright
© The Author(s), 2024. Published by Cambridge University Press on behalf of Australian Mathematical Publishing Association Inc.

1. Introduction

A group is an A-group if its nilpotent subgroups are abelian. For any class of groups ${\mathfrak {B}}$ , we denote the number of groups of order n up to isomorphism by $f_{\mathfrak {B}}(n)$ . Computing $f(n)$ becomes harder as n gets bigger. Thus, in the area of group enumerations, we attempt to approximate $f(n)$ . When counting is restricted to the class of abelian groups, A-groups or groups in general, the asymptotic behaviour of $f(n)$ varies significantly. Let $f_{A,sol}(n)$ be the number of isomorphism classes of soluble A-groups of order n. Dickenson [Reference Dickenson2] showed that $f_{A,sol}(n) \leq n^{c \log {n}}$ for some constant c. McIver and Neumann [Reference McIver and Neumann7] showed that the number of nonisomorphic A-groups of order n is at most $n^{{\lambda }+1}$ , where $\lambda $ is the number of prime divisors of n including multiplicities. In the same paper, they stated the following conjecture based on a result of Higman [Reference Higman4] and Sims [Reference Sims12] on p-group enumerations.

Conjecture 1.1. Let $f(n)$ be the number of (isomorphism classes of groups of) order n. Then $ f(n) \leq n^{({2}/{27} + \epsilon ){\lambda }^2}$ , where $\epsilon \to 0$ as $\lambda \to \infty $ .

In 1993, Pyber [Reference Pyber9] proved a powerful version of Conjecture 1.1: the number of groups of order n with specified Sylow subgroups is at most $n^{75 \mu + 16}$ , where $\mu $ is the largest integer such that $p^{\mu }$ divides n for some prime p. From the results of Higman and Sims, and Pyber, $ f(n) \leq n^{{2}{\mu }^2/27 + O(\mu ^{5/3})}$ . In [Reference Venkataraman13], it was shown that $f_{A,sol}(n) \leq n^{7\mu + 6}$ .

The variety ${\mathfrak {A}_u}{\mathfrak {A}_v}$ consists of all groups G with an abelian normal subgroup N of exponent dividing u such that $G/N$ is abelian of exponent dividing v. (For more on varieties, see [Reference Neumann8].) Let $p, q$ and r be distinct primes. In this paper, we find a bound for $f_{\mathfrak {S}}(n)$ , where $\mathfrak {S}={\mathfrak {A}_p}{\mathfrak {A}_q}{\mathfrak {A}_r}$ and $f_{\mathfrak {S}}(n)$ counts the groups in $\mathfrak {S}$ of order n up to isomorphism. The idea behind studying the variety $\mathfrak {S}$ is that enumerating within the varieties of A-groups might yield a better upper bound for the enumeration function for A-groups. The ‘best’ bounds for A-groups, or even soluble A-groups, still lack the correct leading term. It is believed that a correct leading term for the upper bound of A-groups would lead to the right error term for the enumeration of groups in general.

A few smaller varieties of A-groups have already been studied in [Reference Blackburn, Neumann and Venkataraman1, Ch. 18]. The class of A-groups for which the ‘best’ bounds exist was obtained by enumerating in such small varieties of A-groups, but this did not narrow the difference between the upper and lower bounds for $f_{A,sol}(n)$ because these groups did not contribute a large enough collection of A-groups. Hence, a good lower bound could not be reached. To reduce the difference, we enumerate in the larger variety $\mathfrak {S}$ of A-groups.

Throughout the paper, p, q, r and t are distinct primes. We assume that s is a power of t. We take logarithms to the base $2$ , unless stated otherwise, and follow the convention that $0 \in \mathbb {N}$ . We use $C_m$ to denote a cyclic group of order m for any positive integer m. Let $O_{p'}{(G)}$ denote the largest normal $p'$ -subgroup of G. The techniques we use are similar to those in [Reference Blackburn, Neumann and Venkataraman1, Reference Pyber9, Reference Venkataraman13].

The main result proved in this paper is the following theorem.

Theorem 1.2. Let $n=p^{\alpha }q^{\beta }r^{\gamma }$ , where $\alpha , \beta , \gamma \in \mathbb { N}$ . Then,

$$ \begin{align*}f_{\mathfrak{S}}(n) \leq p^{6\alpha^{2}} 2^{\alpha - 1 + (23/6) \alpha \log\alpha + \alpha \log6} (6^{1/2})^{{({\alpha + \gamma})\beta} + {({\alpha + \beta})\gamma} +{\alpha(\alpha-1)/2}} n^{\beta + \gamma}.\end{align*} $$

To prove Theorem 1.2, we prove a bound on the number of conjugacy classes of subgroups that are maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ and that are in the variety $\mathfrak {A}_q \mathfrak {A}_r$ or $\mathfrak {A}_r$ . We also prove results about the order of primitive subgroups of $S_n$ that are in the variety $\mathfrak {A}_q \mathfrak {A}_r$ and show that they form a single conjugacy class. These results are stated below.

Theorem 1.3. Let q and r be distinct primes. Let G be a primitive subgroup of $S_n$ that is in $\mathfrak {A}_q \mathfrak {A}_r$ and let $|G| = q^\beta r^\gamma $ , where $\beta , \gamma \in \mathbb {N}$ . Let M be a minimal normal subgroup of G.

  1. (i) If $\beta = 0$ , then $|M|$ is a power of r and $|G| = n = r$ with $G \cong C_r$ .

  2. (ii) If $\beta \geq 1$ , then $|M| = q^{\beta } = n$ with $ \beta = {\mathrm{order}} \ q \bmod r$ . Further, $G \cong M \rtimes C_r$ and $|G| = nr < n^{2}$ .

  3. (iii) If $\gamma = 0$ , then $|M|$ is a power of q and $|G| = n = q$ with $G\cong C_q$ .

Theorem 1.4. The primitive subgroups of $S_n$ that are in $\mathfrak {A}_q \mathfrak {A}_r$ and of order $q^{\beta }r^{\gamma }$ , where $\beta , \gamma \in \mathbb { N}$ , form a single conjugacy class.

Theorem 1.5. There exist constants b and c such that the number of conjugacy classes of subgroups that are maximal amongst the subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*} 2^{(b+c) ({\alpha}^2/\sqrt{\log \alpha}) + (5/6) \alpha \log \alpha + \alpha (1+ \log6)} s^{(3+c){\alpha}^2}, \end{align*} $$

where $t, q$ and r are distinct primes, s is a power of t, and $\alpha> 1$ .

Section 2 investigates primitive subgroups of $S_{n}$ that are in $\mathfrak {A}_r$ or $\mathfrak {A}_q\mathfrak {A}_r$ . Sections 3 and 4 deal with subgroups of the general linear group. Theorem 1.2 is proved in Section 5.

2. Primitive subgroups of $S_{n}$ that are in $\mathfrak {A}_r$ or $\mathfrak {A}_q\mathfrak {A}_r$

In this section, we prove results that give us the structure of the primitive subgroups of $S_{n}$ that are in $\mathfrak {A}_r$ or $\mathfrak {A}_q\mathfrak {A}_r$ . We also show that such subgroups form a single conjugacy class. Both Theorems 1.3 and 1.4 are proved in this section.

Theorem 1.3 provides the order of a primitive subgroup of $S_n$ that is in the variety $\mathfrak {A}_q\mathfrak {A}_r$ . By [Reference Venkataraman13, Proposition 2.1], if G is a soluble A-subgroup of $S_n$ , then $|G| \leq (6^{1/2})^{n-1}$ . Indeed, this bound is determined primarily by considering primitive soluble A-subgroups of $S_n$ . This bound would clearly hold for any subgroup of $S_n$ that is in the variety $\mathfrak {A}_q\mathfrak {A}_r$ . However, we show that when the subgroup is primitive and in the variety $\mathfrak {A}_q\mathfrak {A}_r$ , we can do better.

Lemma 2.1. $S_n$ has a primitive subgroup in $\mathfrak {A}_r$ if and only if $n=r$ . In this case, any primitive subgroup G that is in $\mathfrak {A}_r$ will be cyclic of order r. All primitive subgroups of $S_n$ that are in $\mathfrak {A}_r$ form a single conjugacy class.

Proof. Let G be a primitive subgroup of $S_n$ that is in $\mathfrak {A}_r$ . Since G is soluble, M is an elementary abelian r-subgroup. By the O’Nan–Scott theorem [Reference Scott, Cooperstein and Mason10], $|M| = n = |G|$ , so $G = M \cong C_{r}$ and $n=r$ . Conversely, any transitive subgroup G of $S_{r}$ is primitive [Reference Wielandt15, Theorem 8.3]. Since n is prime, any subgroup of order n in $S_n$ will be generated by an n-cycle. Further, any two n-cycles are conjugate in $S_n$ . Thus, the primitive subgroups of $S_n$ that are also in $\mathfrak {A}_r$ form a single conjugacy class.

Proof of Theorem 1.3.

Let G be a subgroup of $S_{\Omega }$ , where $|\Omega |=n$ , and let ${G \in \mathfrak {A}_q\mathfrak {A}_r}$ . Then $G = Q \rtimes R$ , where Q is an elementary abelian Sylow q-subgroup, R is an elementary abelian Sylow r-subgroup and $|G| = q^\beta r^\gamma $ , with $\beta $ , $\gamma \in \mathbb {N}$ . Let M be a minimal normal subgroup of G. Then M is an elementary abelian u-group. Clearly, $|M|=u^{k}$ for some $k> 1$ and for some prime $u \in \{q,r\}$ .

Now $F(G)$ , the Fitting subgroup of G, is an abelian normal subgroup of G and so, by the O’Nan–Scott theorem, $n=|M|=|F(G)|$ . However, $M \leq F(G)$ , therefore, ${M=F(G)}$ and $n=u^k$ . If $\beta \geq 1$ , then $Q \leq F(G)$ and we have $n= q^{\beta } = u^k$ and $M=F(G) = Q$ . Let $H=G_{\alpha }$ be the stabiliser of an $\alpha \in \Omega $ . By [Reference Blackburn, Neumann and Venkataraman1, Proposition 6.13], G is a semidirect product of M by H and H acts faithfully by conjugation on M. By Maschke’s theorem, M is completely reducible. However, M is a minimal normal subgroup of G, so M is a nontrivial irreducible $\mathbb {F}_{q}H$ -module and H is an abelian group acting faithfully on M. By [Reference Venkataraman14, Corollary 4.1], $H\cong C_{r}$ and $ \beta = \dim M = \text {order } q \bmod r$ and the result follows. If $\gamma = 0$ or $\beta = 0$ , then $|G|$ is a power of u, where $u \in \{q, r\}$ . Thus, G is a primitive subgroup that is also in $\mathfrak {A}_u$ and the result follows by Lemma 2.1.

It is clear from these results that if $S_n$ has a primitive subgroup G of order $q^{\beta }r^{\gamma }$ in $\mathfrak {A}_q\mathfrak {A}_r$ , then n must be r or q and G is cyclic with $|G|=n$ , or $n=q^{\beta }$ and G is a semi-direct product of an elementary abelian q-group of order $q^{\beta }$ by a cyclic group of order r. The limits imposed on n and the structure of such primitive subgroups gives the next result.

Proof of Theorem 1.4.

Let G be a primitive subgroup of $S_{\Omega }$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ , where $|\Omega | = n$ , and let $|G|= q^{\beta }r^{\gamma }$ . Let M be a minimal normal subgroup of G. As seen in the proof of Theorem 1.3, $M = F(G)$ and $n=|M|$ is either a power of q or r. If $\gamma = 0$ or $\beta = 0$ , then $|G|$ is a power of u, where $u \in \{q, r\}$ . Thus, G is a primitive subgroup that is also in $\mathfrak {A}_u$ and the result follows by Lemma 2.1.

We know the structure of G when $\beta \geq 1$ from the proof of Theorem 1.3. Hence, H can be regarded as a soluble r-subgroup of $\mathrm{GL}(\beta ,q)$ and it is not difficult to show that the conjugacy class of G in $S_{n}$ is determined by the conjugacy class of H in $\mathrm{GL}(\beta ,q)$ . Let S be a Singer subgroup of $\mathrm{GL}(\beta ,q)$ , so that $|S| = q^{\beta }-1$ . Now, $|H| =r$ and r divides $|S|$ . Further, $\gcd (|\mathrm{GL}(\beta ,q)|/|S|, r)=1$ as $\beta $ is the least positive integer such that ${r \mid q^{\beta }-1}$ . From [Reference Hestenes3, Theorem 2.11], $H^{x} \leq S$ for some $x \in \mathrm{GL}(\beta ,q)$ . Since all Singer subgroups are conjugate in $\mathrm{GL}(\beta ,q)$ , the result follows.

3. Subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_r$

In this section, we prove results that give us a bound on the number of conjugacy classes of the subgroups that are maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_r$ . The limits on the structure of such groups ensures that if they exist, they form a single conjugacy class.

Lemma 3.1. The number of conjugacy classes of irreducible subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_r$ is at most $1$ .

Proof. Let G be a nontrivial irreducible subgroup of $\mathrm{GL}(\alpha , s)$ that is also in $\mathfrak {A}_r$ . Then G is an elementary abelian r-group of order $r^{\gamma }$ , say, where $\gamma \in \mathbb {N}$ . Since G is a faithful abelian irreducible subgroup of $\mathrm{GL}(\alpha , s)$ whose order is coprime to s, it follows that G is cyclic [Reference Venkataraman14, Lemma 4.2]. Thus, $|G|=r$ and $\alpha = d$ , where $d= \text {order } s \bmod r$ . From [Reference Short11, Theorem 2.3.3], the irreducible cyclic subgroups of order r in $\mathrm{GL}(\alpha , s)$ lie in a single conjugacy class.

Proposition 3.2. The number of conjugacy classes of subgroups that are maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_r$ is at most $1$ .

Proof. Let G be maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_r$ . Since $\text {char}(\mathbb {F}_{p}) = t \nmid |G|$ , by Maschke’s theorem, we can find groups $G_i$ such that $G \leq G_{1} \times G_{2} \times \cdots \times G_{k} = \hat {G} \leq \mathrm{GL}(\alpha , s)$ , where for each i, the group $G_i$ is a (maximal) irreducible subgroup of $\mathrm{GL}(\alpha _i, s)$ that is also in $\mathfrak {A}_r$ . Further, $\alpha = \alpha _1 + \cdots + \alpha _k$ . Clearly, $G_i \cong C_r$ and $\alpha _i = d = \text {order } s \bmod r$ for each i. Thus, we must have $\alpha = dk$ and by the maximality of G, we have $G = \hat {G}$ . Further, the conjugacy classes of $G_i$ in $\mathrm{GL}(\alpha _i, s)$ determine the conjugacy class of G in $\mathrm{GL}(\alpha , s)$ .

So if d does not divide $\alpha $ , then $\mathrm{GL}(\alpha , s)$ cannot have an elementary abelian r-subgroup. If $d \mid \alpha $ , then any G that is maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_r$ must have order $r^{k}$ , where $k = \alpha /d$ . By Lemma 3.1, all such groups form a single conjugacy class.

4. Subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$

We prove results that give a bound on the order of subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ and also a bound for the number of conjugacy classes of subgroups that are maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ . Theorem 1.5 is proved in this section.

Proposition 4.1. Let G be a subgroup of $\mathrm{GL}(\alpha , s)$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ .

  1. (i) Let $m = |F(G)|$ . If G is primitive, then $|G| \leq cm$ , where $c = {\mathrm{order}} \ s \bmod m$ and $c \mid \alpha $ . Further, m is either r or q or $qr$ .

  2. (ii) $|G| \leq {(6^{1/2})}^{\alpha - 1} {d}^{\alpha }$ , where $d = \min \{ qr, s \}$ .

Proof. Let $V = (\mathbb {F}_{s})^{\alpha }$ . Let G be a primitive subgroup of $\mathrm{GL}(\alpha , s)$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ and let $|G| = q^{\beta } r^{\gamma }$ , where $\beta $ and $\gamma $ are natural numbers. If $\beta = 0$ or $\gamma = 0$ , then the result follows from Lemma 3.1. Assume that $\beta $ and $\gamma $ are at least 1. Let $F = F(G)$ be the Fitting subgroup of G. Since $G \in \mathfrak {A}_q\mathfrak {A}_r$ , it follows that F is abelian and $|F|=q^{\beta }r^{\gamma _1} =m$ , where $\gamma _1 \leq \gamma $ . By Clifford’s theorem, since G is primitive, $V = X_{1} \oplus X_{2} \oplus \cdots \oplus X_{a}$ as an F-module, where the $X_{i}$ are conjugates of X, an irreducible $\mathbb {F}_{s}F$ -submodule of V. Note that F acts faithfully on X.

Let E be the subalgebra generated by F in $\text {End}(V)$ . The $X_{i}$ are conjugates of X, so E acts faithfully and irreducibly on X and E is commutative. By [Reference Blackburn, Neumann and Venkataraman1, Proposition 8.2 and Theorem 8.3], E is a field. Thus, $E \cong \mathbb {F}_{s^{c}}$ as an $\mathbb {F}_{s}F$ -module, where $c = \dim (X)$ and $\alpha = ac$ . Note that F is an abelian group of order m acting faithfully and irreducibly on X. Consequently, F is cyclic and c is the least positive integer such that $m \mid s^{c} -1 $ . Clearly, $m=q$ or $m=qr$ and so $\beta =1$ . It is not difficult to show that G acts on E by conjugation. Hence, there exists a homomorphism from G to $\text {Gal}_{\mathbb {F}_{s}}(E)$ . Let N be the kernel of this map. Then $N = C_{G}(E) \leq C_{G}(F) \leq F$ . However, $F \leq N$ . Hence, $F=N$ . So ${G}/{F} \leq \text {Gal}_{\mathbb {F}_{s}}(E) \cong C_{c}$ and $|G| \leq cm$ .

Let G be an irreducible imprimitive subgroup of $\mathrm{GL}(\alpha , s)$ that is also in $\mathfrak {A}_q\mathfrak {A}_r$ . Then $G \leq G_{1}\, \mathrm{wr} \,G_{2} \leq \mathrm{GL}(\alpha , s)$ , where $G_{1}$ is a primitive subgroup of $\mathrm{GL}(\alpha _1, s)$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ , and the group $G_{2}$ can be regarded as a transitive subgroup of $S_{k}$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ . Further, $\alpha = \alpha _{1} k$ . By the previous part, $|G_1| \leq c'm'$ , where $c' = \text {order}\ s \bmod m' $ and $m' = |F(G_1)|$ is either r or q or $qr$ . Also $c' \mid \alpha _1$ . By [Reference Venkataraman13, Proposition 2.1], $|G_2| \leq {(6^{1/2})}^{k-1}$ . Using $c' \leq 2^{c'-1} \leq {(6^{1/2})}^{c' - 1}$ , we see that $|G| \leq {(6^{1/2})}^{\alpha - 1} \, {(m')}^{k}$ . Since $m' \mid p^{c'} -1$ , we have ${(m')}^k \leq d^{\alpha }$ , where $d = \min \{ qr, s \}$ .

Since t does not divide q or r, by Maschke’s theorem, any subgroup G of $\mathrm{GL}(\alpha , s)$ that is in $\mathfrak {A}_q\mathfrak {A}_r$ will be completely reducible. Thus, $G \leq G_{1} \times \cdots \times G_{k} \leq \mathrm{GL}(\alpha , s)$ , where the $G_{i}$ are irreducible subgroups of $\mathrm{GL}(\alpha _i, s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ and $\alpha = \alpha _{1} + \cdots + \alpha _{k}$ . Hence, $|G| \leq {(6^{1/2})}^{\alpha - 1} \, {d}^{\alpha }$ , where $d = \min \{ qr, s \}$ .

Proposition 4.2. There exist constants b and c such that the number of conjugacy classes of subgroups that are maximal amongst irreducible subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ is at most $ 2^{(b+c) ({\alpha }^2/\sqrt {\log \alpha }) + (5/6) \log \alpha + \log 6} s^{(3+c){\alpha }^2}$ provided $\alpha> 1$ .

Proof. Let G be a subgroup of $\mathrm{GL}(\alpha , s)$ that is maximal amongst irreducible subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ . Let $|G| = q^{\beta } r^{\gamma }$ , where $\beta $ and $\gamma $ are natural numbers. If $\beta = 0$ or $\gamma = 0$ , then the result follows from Lemma 3.1. Assume that $\beta $ and $\gamma $ are at least 1. Let $V = (\mathbb {F}_{s})^{\alpha }$ and $F = F(G)$ , the Fitting subgroup of G. Then $F=Q \times R_1$ , where Q is the unique Sylow q-subgroup of G and $R_1 \leq R$ , where R is a Sylow r-subgroup of G. So F is abelian and $|F|=q^{\beta }r^{\gamma _1} =m$ , where $\gamma _1 \leq \gamma $ .

From Clifford’s theorem, regarding V as an $\mathbb {F}_{s}F$ -module, $V = Y_{1} \oplus Y_{2} \oplus \cdots \oplus Y_{l}$ , where $Y_{i} = k X_{i}$ for all i, and $X_{1}, \ldots , X_{l}$ are irreducible $\mathbb {F}_{s}F$ -submodules of V. Further, for each $i,j$ , there exists $g_{ij} \in G$ such that $g_{ij}X_{i} = X_{j}$ and, for $i=1, \ldots , l$ , the $X_{i}$ form a maximal set of pairwise nonisomorphic conjugates. Also, the action of G on the $Y_{i}$ is transitive. It is not difficult to check that $ C_{F}(Y_{i}) = C_{F}(X_{i}) = K_i$ , say. Thus, $F/K_i$ acts faithfully on $Y_{i}$ and when its action is restricted to $X_{i}$ , it acts faithfully and irreducibly on $X_{i}$ . Since $X_{i}$ is a nontrivial irreducible faithful $\mathbb {F}_{s}F/K_i$ -module, and t is coprime to q and r, it follows that $F/K_i$ is cyclic and $\dim _{\mathbb {F}_{s}}(X_i) = d_i$ , where $d_i$ is the least positive integer such that $m_i$ divides $s^{d_i} -1$ , and where $m_i$ is the order of $F/K_i$ . Since the $X_i$ are conjugate, $\dim _{\mathbb {F}_{s}}(X_i) = d_i = d$ for all i.

Let $E_{i}$ be the subalgebra generated by $F/K_i$ in $\text {End}_{{\mathbb{F}}_{s}}(Y_{i})$ . Note that $E_{i}$ is commutative as $F/K_i$ is abelian. Further, $X_{i}$ is a faithful irreducible $E_{i}$ -module. So $E_{i}$ is simple and becomes a field such that $E_{i} \cong \mathbb{F}_{s{^d}}$ . We also observe that $\alpha = k l d$ .

Let $k, l, d$ be fixed such that $\alpha = kld$ . Next we find the number of choices for F up to conjugacy in $\mathrm{GL}(V)$ . Clearly,

$$ \begin{align*} F &\leq F/K_1 \times F/K_2 \times \cdots \times F/K_l \leq {E_1}^{*} \times {E_2}^{*} \times \cdots \times {E_l}^{*} \\ &\leq \mathrm{GL}(Y_{1}) \times \mathrm{GL}(Y_{2}) \times \cdots \times \mathrm{GL}(Y_{l}) \leq \mathrm{GL}(V), \end{align*} $$

where ${E_i}^*$ denotes the multiplicative group of the field $E_i$ . Let $E = {E_1}^{*} \times {E_2}^{*} \times \cdots \times {E_l}^{*}$ . Then $|E|= (s^{d} - 1)^{l}$ . Regarding V as an $\mathbb{F}_{s}E$ -module, $V = kX_{1} \oplus kX_{2} \oplus \cdots \oplus kX_{l}$ , where $E_{i}^{*}$ acts faithfully and irreducibly on $X_{i}$ and $\dim _{E_{i}}(X_{i}) = 1$ for all i. Further, for $i \neq j$ , $E_{i}^{*}$ acts trivially on $X_{j}$ . It is not difficult to show that there is only one conjugacy class of subgroups of type E in $\mathrm{GL}(V)$ .

So once $k,l$ and d are chosen such that $\alpha = k l d$ , up to conjugacy, there is only one choice for E. Since E is a direct product of l isomorphic cyclic groups, any subgroup of E can be generated by l elements. In particular, F can be generated by l elements. So the number of choices for F as a subgroup of E is at most $|E|^{l}= (s^{d} - 1)^{{l}^{2}}$ .

Since, G acts transitively on $\{Y_{1},\ldots , Y_{l}\}$ , there exists a homomorphism $\phi $ from G into $S_{l}$ . Let ${N} = \ker (\phi ) = \{g \in G \mid gY_{i} = Y_{i}\text { for all } i \}$ . Clearly, $F \leq {N}$ and $G/{N}$ is a transitive subgroup of $S_{l}$ that is in $\mathfrak {A}_r$ . If $g \in {N}$ , then $gE_{i}g^{-1}=E_{i}$ . Thus, there exists a homomorphism $\psi _{i}: {N} \rightarrow \text {Gal}_{\mathbb {F}_{s}}(E_{i})$ . This induces a homomorphism $\psi $ from N to $\text {Gal}_{\mathbb {F}_{s}}(E_{1}) \times \text {Gal}_{\mathbb {F}_{s}}(E_{2}) \times \cdots \times \text {Gal}_{\mathbb {F}_{s}}(E_{l})$ such that $\ker (\psi ) = \bigcap _{i=1}^{l} N_{i} = F$ , where $N_{i} = \ker (\psi _i) = C_{N}(E_{i})$ . So ${N}/F$ is isomorphic to a subgroup of $\text {Gal}_{\mathbb {F}_{s}}(E_{1}) \times \text {Gal}_{\mathbb {F}_{s}}(E_{2}) \times \cdots \times \text {Gal}_{\mathbb {F}_{s}}(E_{l})$ . Since $\text {Gal}_{\mathbb {F}_{s}}(E_{i}) \cong C_{d}$ for every i, it follows that ${N}/F$ can be generated by l elements.

Let $T=\mathrm{GL}(\alpha , s)$ . Let $\hat {N}=\{x \in N_{T}{(F)} \mid xY_{i}=Y_{i} \ \text {for all}\ i \}$ . Then $F \leq N \leq \hat {N} \leq N_{T}{(F)}$ . We will find the number of choices for N as a subgroup of $\hat {N}$ , given that F has been chosen. The group $\hat {N}$ acts by conjugation on $E_{i}$ and fixes the elements of $\mathbb {F}_s$ . So we have a homomorphism $\rho _{i}: \hat {N} \rightarrow \text {Gal}_{\mathbb {F}_{s}}(E_{i})$ with kernel $C_{\hat {N}}{(E_{i})}$ . Define $C = \bigcap _{i=1}^{l} C_{\hat {N}}{(E_{i})}$ . Note that $N \cap C = F$ . Also, $\hat {N}/C$ is isomorphic to a subgroup of $\text {Gal}_{\mathbb {F}_{s}}(E_{1}) \times \text {Gal}_{\mathbb {F}_{s}}(E_{2}) \times \cdots \times \text {Gal}_{\mathbb {F}_{s}}(E_{l})$ , where each $\text {Gal}_{\mathbb {F}_{s}}(E_{i}) $ is isomorphic to $C_{d}$ . So $|\hat {N}/C| \leq d^{l}$ . Clearly, C centralises $E_{i}$ for each i. Therefore, there exists a homomorphism from C into $\mathrm{GL}_{E_{i}}{(Y_{i})}$ for each i. Hence, C is isomorphic to a subgroup of $\mathrm{GL}_{E_{1}}(Y_{1}) \times \mathrm{GL}_{E_{2}}(Y_{2}) \times \cdots \times \mathrm{GL}_{E_{l}}(Y_{l})$ . As $\dim _{\mathbb {E}_{i}}(Y_i) = k$ and $E_{i} \cong \mathbb{F}_{s^{d}}$ for all i, it follows that $|C| \leq s^{dk^{2}l}$ . Hence, $|\hat {N}| \leq d^{l}s^{dk^{2}l}$ .

Now $NC/C \cong N/(N \cap C) = N/F$ . So $NC/C$ can be generated by l elements since $N/F$ can be generated by l elements. However, $|\hat {N}/C| \leq d^{l}$ , therefore, there are at most $d^{l^{2}}$ choices for $NC/C$ as a subgroup of $\hat {N}/C$ . Once we make a choice for $NC/C$ as a subgroup of $\hat {N}/C$ , we choose a set of l generators for $NC/C$ . As $N \cap C = F$ , we see that N is determined as a subgroup of $\hat {N}$ by F and l other elements that map to the chosen generating set of $NC/C$ . We have $|C|$ choices for an element of $\hat {N}$ that maps to any fixed element of $\hat {N}/C$ . Thus, there are at most $|C|^{l}$ choices for N as a subgroup of $\hat {N}$ once $NC/C$ has been chosen. So we have at most $d^{l^{2}}(s^{{dk^{2}l}})^{l} = d^{l^{2}}s^{dk^{2}l^{2}}$ choices for N as a subgroup of $\hat {N}$ , once F is fixed.

Next we find the number of choices for G given that F and N are fixed as subgroups of T and $\hat {N} \leq T$ , respectively. Let $\hat {Y}=\{y \in N_{T}{(F)} \mid y \text { permutes the } Y_i \}$ . Then $F \leq G \leq \hat {Y} \leq N_{T}{(F)} \leq \mathrm{GL}(V)$ . Also there exists a homomorphism from $\hat {Y}$ to $S_{l}$ with kernel ${\{y \in \hat {Y} \mid yY_{i} = Y_{i} \text { for all } i\} = \hat {N}}$ . Thus, $\hat {Y}/\hat {N}$ may be regarded as a subgroup of $S_{l}$ . However, $G \cap \hat {N} = N$ . Thus, $G/N = G/(G \cap \hat {N}) \cong G\hat {N}/\hat {N}$ . So ${G/N \cong G\hat {N}/\hat {N} \leq \hat {Y}/\hat {N} \leq S_{l}}$ . Note that $G/N$ is a transitive subgroup of $S_{l}$ that is in $ {\mathfrak {A}_{\mathrm{r}}}$ . By [Reference Lucchini5, Theorem 1], there exists a constant b such that $S_{l}$ has at most $ 2^{bl^{2} / {\sqrt {\log l}}}$ transitive subgroups for $l>1$ . Hence, the number of choices for $G\hat {N}/\hat {N}$ as a subgroup of $\hat {Y}/\hat {N}$ is at most $ 2^{bl^{2} / {\sqrt {\log l}}}$ .

By [Reference Lucchini, Menegazzo and Morigi6, Theorem 2], there exists a constant c such that any transitive permutation group of finite degree greater than $1$ can be generated by $\lfloor cl/\sqrt {\log l}\rfloor $ generators. Thus, $G\hat {N}/\hat {N}$ can be generated by $\lfloor cl/\sqrt {\log l}\rfloor $ generators for $l> 1$ . Once a choice for $G\hat {N}/\hat {N}$ is made as a subgroup of $\hat {Y}/\hat {N}$ and $\lfloor cl/\sqrt {\log l}\rfloor $ generators are chosen for $G\hat {N}/\hat {N}$ in $\hat {Y}/\hat {N}$ , then G is determined as a subgroup of $\hat {Y}$ by $\hat {N}$ and the elements of $\hat {Y}$ that map to the $\lfloor cl/\sqrt {\log l}\rfloor $ generators chosen for $G\hat {N}/\hat {N}$ . So we have at most $|\hat {N}|^{\lfloor cl/\sqrt {\log l}\rfloor }$ choices for G as a subgroup of $\hat {Y}$ once a choice of $G\hat {N}/\hat {N}$ in $\hat {Y}/\hat {N}$ is fixed. Hence, there are

$$ \begin{align*}2^{bl^{2} / {\sqrt{\log l}}}\, (d^{l}s^{dk^{2}l})^{\lfloor cl/\sqrt{\log l}\rfloor} \leq 2^{bl^{2} / {\sqrt{\log l}}} d^{cl^{2}/ \sqrt{\log l}} s^{{cdk^{2}l^{2}}/\sqrt{\log(l)}}\end{align*} $$

choices for G as a subgroup of $\hat {Y}$ assuming that choices for F and N have been made. Putting together all these estimates, the number of conjugacy classes of subgroups that are maximal amongst irreducible subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*} \sum_{(k,l,d)}^{} {(s^{d} - 1)^{{l}^{2}} d^{l^{2}} s^{dk^{2}l^{2}} {2^{bl^{2} / {\sqrt{\log l}}}} {d^{cl^{2}/ \sqrt{\log l}}} {s^{{cdk^{2}l^{2}}/\sqrt{\log l}}}}, \end{align*} $$

where $(k,l,d)$ ranges over ordered triples of natural numbers which satisfy $\alpha = k l d$ and $l>1$ . We simplify the above expression as follows. Writing $\alpha = k l d$ ,

$$ \begin{align*} {(s^{d} - 1)^{{l}^{2}}d^{l^{2}}s^{dk^{2}l^{2}}{s^{{cdk^{2}l^{2}}/\sqrt{\log l}}}} \leq s^{(3 +c){\alpha}^2}. \end{align*} $$

Since $x/\sqrt {\log x}$ is increasing for $x> e^{1/2}$ , we have $l/\sqrt {\log l} \leq \alpha /\sqrt {\log \alpha }$ for $l \geq 2$ . Thus, $2^{bl^{2} / {\sqrt {\log l}}} d^{cl^{2}/ \sqrt {\log l}} \leq 2^{(b + c) {\alpha }^2/\sqrt {\log \alpha }}$ .

There are at most $2^{({5}/{6})\log \alpha +\log 6}$ choices for $(k,l,d)$ . Thus, there exist constants b and c such that the number of conjugacy classes of subgroups that are maximal amongst irreducible subgroups of $\mathrm{GL}(\alpha , s)$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*} 2^{(b+c) ({\alpha}^2/\sqrt{\log \alpha}) + (5/6) \log \alpha + \log6} \, s^{(3+c){\alpha}^2}\end{align*} $$

provided $\alpha> 1$ .

Theorem 1.5 follows as a corollary to Proposition 4.2.

Proof of Theorem 1.5.

Let G be maximal amongst subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ . As the characteristic of $\mathbb {F}_{s} = t$ and $t \nmid |G|$ , by Maschke’s theorem, $G \leq \hat {G_{1}}\times \cdots \times \hat {G_{k}} \leq \mathrm{GL}(\alpha , s)$ , where the $\hat {G_{i}}$ are maximal among irreducible subgroups of $\mathrm{GL}(\alpha _{i}, p)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ , and where $\alpha = \alpha _{1} + \cdots + \alpha _{k}$ . By the maximality of G, we have $G = \hat {G_{1}}\times \cdots \times \hat {G_{k}}$ .

The conjugacy classes of $\hat {G_{i}} \in \mathrm{GL}(\alpha _{i}, s)$ determine the conjugacy class of ${G \in \mathrm{GL}(\alpha , s)}$ . So by Proposition 4.2, the number of conjugacy classes of subgroups that are maximal amongst the subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*} \sum_{(\alpha)} \prod_{i=1}^{k} 2^{(b+c) ({\alpha_i}^2/\sqrt{\log \alpha_i}) + (5/6) \log \alpha_i + \log6} s^{(3+c){\alpha_i}^2},\end{align*} $$

where the sum is over all unordered partitions $\alpha _{1}, \ldots , \alpha _{k}$ of $\alpha $ . We assume that if $\alpha _i =1$ for some i, then the part of the expression corresponding to it in the product is 1. Since $x/\sqrt {\log x}$ is increasing for $x> e^{1/2}$ and $\alpha = \alpha _{1} + \cdots + \alpha _{k}$ ,

$$ \begin{align*} \prod_{i=1}^{k} 2^{(b+c) ({\alpha_i}^2/\sqrt{\log \alpha_i}) + (5/6) \log \alpha_i + \log6} \leq 2^{(b+c) ({\alpha}^2/\sqrt{\log \alpha}) + (5/6) \alpha \log \alpha + \alpha \log6}. \end{align*} $$

It is not difficult to show that the number of unordered partitions of $\alpha $ is at most $2^{\alpha -1}$ . So the number of conjugacy classes of subgroups that are maximal amongst the subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*} 2^{(b+c) ({\alpha}^2/\sqrt{\log \alpha}) + (5/6) \alpha \log \alpha + \alpha (1+ \log 6))} s^{(3+c){\alpha}^2} \end{align*} $$

provided $\alpha> 1$ .

We end this section with the following remark that provides an alternative bound.

Remark 4.3. We do not have an estimate for the constants b and c occurring in Theorem 1.5. If we use a weaker fact that any subgroup of $S_n$ can be generated by $\lfloor n/2 \rfloor $ elements for all $n \geq 3$ , then we get a weaker result that the number of transitive subgroups of $S_n$ that are in $\mathfrak {A}_q\mathfrak {A}_r$ is at most $6^{n(n-1)/4} 2^{(n+2) \log n}$ . Using this in the proof of Theorem 1.5 shows that the number of conjugacy classes of subgroups that are maximal amongst the subgroups of $\mathrm{GL}(\alpha , s)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ is at most

$$ \begin{align*}s^{5\alpha^{2}} 6^{\alpha(\alpha-1)/4} 2^{\alpha - 1 + (23/6) \alpha \log\alpha + \alpha \log6},\end{align*} $$

where $t, q$ and r are distinct primes, s is a power of t, and $\alpha \in \mathbb {N}$ .

5. Enumeration of groups in $\mathfrak {A}_p\mathfrak {A}_q \mathfrak {A}_r$

In this section, we prove Theorem 1.2, namely,

$$ \begin{align*}f_{\mathfrak{S}}(n) \leq p^{6\alpha^{2}} 2^{\alpha - 1 + (23/6) \alpha \log\alpha + \alpha \log6} (6^{1/2})^{{({\alpha + \gamma})\beta} + {({\alpha + \beta})\gamma} +{\alpha(\alpha-1)/2}} n^{\beta + \gamma},\end{align*} $$

where $n=p^{\alpha }q^{\beta }r^{\gamma }$ and $\alpha , \beta , \gamma \in \mathbb { N}$ . We use techniques adapted from [Reference Pyber9, Reference Venkataraman13, Reference Venkataraman14].

Proof of Theorem 1.2.

Let G be a group of order $n = p^{\alpha }q^{\beta }r^{\gamma }$ in $\mathfrak {A}_p\mathfrak {A}_q\mathfrak {A}_r$ . Then $G= P \rtimes H$ , where P is the unique Sylow p-subgroup of G and $H \in \mathfrak {A}_q\mathfrak {A}_r$ . So we can write $H =Q \rtimes R$ , where $|Q|=q^{\beta }$ and $|R|=r^{\gamma }$ . Let $G_{1}={G}/{O_{p'}{(G)}}$ , $G_{2}={G}/{O_{q'}{(G)}}$ and $G_{3}={G}/{O_{r'}{(G)}}$ . Clearly, each $G_{i}$ is a soluble A-group and $G \leq G_{1} \times G_{2} \times G_{3}$ as a subdirect product. Further, ${O_{p'}{(G_{1})}}=1={O_{q'}{(G_{2})}}={O_{r'}{(G_{3})}}$ .

Since $G_{1}={G}/{O_{p'}{(G)}}$ , we see that $G_{1} \in \mathfrak {A}_p\mathfrak {A}_q\mathfrak {A}_r$ and if $P_1$ is the Sylow p-subgroup of $G_1$ , then $P_{1} \cong P$ . Thus, $|G_1|=p^{\alpha }q^{\beta _{1}}r^{\gamma _{1}}$ and we can write $G_{1}= P_{1} \rtimes H_{1}$ , where ${H_{1} \in \mathfrak {A}_q\mathfrak {A}_r}$ . So $H_{1}= Q_{1} \rtimes R_{1}$ , where $Q_{1}\in \mathfrak {A}_q$ and $|Q_1|=q^{\beta _{1}}$ , $R_{1}\in \mathfrak {A}_r$ and $|R_1|=r^{\gamma _{1}}$ . Further, $H_{1}$ acts faithfully on $P_{1}$ . Hence, we can regard $H_1 \leq \mathrm{Aut}(P_{1}) \cong \mathrm{GL}(\alpha , p)$ . Let $M_{1}$ be a subgroup that is maximal amongst $p'$ -A-subgroups of $\mathrm{GL}(\alpha , p)$ that are also in $\mathfrak {A}_q\mathfrak {A}_r$ and such that $H_{1} \leq M_{1}$ . Let $\hat {G_{1}}=P_{1}M_{1}$ . The number of conjugacy classes of the $M_{1}$ in $\mathrm{GL}(\alpha , p)$ is at most $p^{5\alpha ^{2}} 6^{\alpha (\alpha -1)/4} 2^{\alpha - 1 + (23/6) \alpha \log \alpha + \alpha \log 6}$ by Remark 4.3.

Since $G_{2}={G}/{O_{q'}{(G)}}$ , we see that $G_{2} \in \mathfrak {A}_q\mathfrak {A}_r$ and if $Q_2$ is the Sylow q-subgroup of  $G_2$ , then $Q_{2} \cong Q$ . Thus, $|G_2|=q^{\beta }r^{\gamma _{2}}$ and we can write $G_{2}=Q_{2} \rtimes H_{2}$ , where $H_{2} \in \mathfrak {A}_r$ . So $|H_2|=r^{\gamma _{2}}$ . Also, $H_2 \leq \mathrm{Aut}(Q_{2}) \cong \mathrm{GL}(\beta , q)$ . Let $M_{2}$ be a subgroup that is maximal amongst $q'$ -A-subgroups of $\mathrm{GL}(\beta , q)$ that are also in $\mathfrak {A}_r$ and such that $H_{2} \leq M_{2}$ . Let $\hat {G_{2}} = Q_{2}M_{2}$ . The number of conjugacy classes of $M_{2}$ in $\mathrm{GL}(\beta , q)$ is at most $1$ by Proposition 3.2.

Since $G_{3}={G}/{O_{r'}{(G)}}$ , we see that $G_{3} \in \mathfrak {A}_r\mathfrak {A}_q$ and if $R_3$ is the Sylow r-subgroup of $G_3$ , then $R_{3} \cong R$ . Thus, $|G_3|=q^{\beta _3}r^{\gamma }$ and we can write $G_{3}=R_{3} \rtimes H_{3}$ , where $H_{3} \in \mathfrak {A}_r$ . So $|H_3|=q^{\beta _{3}}$ . Also, $H_3 \leq \mathrm{Aut}(R_{3}) \cong \mathrm{GL}(\gamma , r)$ . Let $M_{3}$ be a subgroup that is maximal amongst $r'$ -A-subgroups of $\mathrm{GL}(\gamma , r)$ that are also in $\mathfrak {A}_q$ and such that $H_{3} \leq M_{3}$ . Let $\hat {G_{3}} = R_{3}M_{3}$ . The number of conjugacy classes of the $M_{3}$ in $\mathrm{GL}(\gamma , r)$ is at most $1$ by Proposition 3.2.

Let $\hat {G} = \hat {G_{1}} \times \hat {G_{2}} \times \hat {G_{3}}$ . Then $G \leq \hat {G}$ . The choices for $P_{1}, Q_{2}$ and $R_{3}$ are unique, up to isomorphism. We enumerate the possibilities for $\hat {G}$ up to isomorphism and then find the number of subgroups of $\hat {G}$ of order n up to isomorphism. For the former, we count the number of $\hat {G_{i}}$ up to isomorphism which depends on the conjugacy class of the $M_{i}$ . Hence, the number of choices for $\hat {G}$ up to isomorphism is $\prod _{i=1}^{3} \{\text {number of choices for } \hat {G_{i}} \text { up to isomorphism}\}$ . Now we estimate the choices for G as a subgroup of $\hat {G}$ using a method of ‘Sylow systems’ introduced by Pyber in [Reference Pyber9] .

Let $\hat {G}$ be fixed. We count the number of choices for G as a subgroup of $\hat {G}$ . Let ${\cal S} = \{S_{1}, S_{2}, S_{3}\}$ be a Sylow system for G, where $S_{1}$ is the Sylow p-subgroup of G, $S_{2}$ is a Sylow q-subgroup of G and $S_{3}$ is a Sylow r-subgroup of G such that $S_{i}S_{j} = S_{j}S_{i}$ for all $i, j=1,2,3$ . Then $G = S_{1} S_{2} S_{3}$ . By [Reference Blackburn, Neumann and Venkataraman1, Theorem 6.2, page 49], there exists ${{\cal B} = \{B_{1}, B_{2}, B_{3}\}}$ , a Sylow system for $\hat {G}$ such that $S_{i} \leq B_{i}$ , where $B_{1}$ is the Sylow p-subgroup of $\hat {G}$ , $B_{2}$ is a Sylow q-subgroup of $\hat {G}$ and $B_{3}$ is a Sylow r-subgroup of $\hat {G}$ . Note that $|B_1|=p^{\alpha }$ . Further, any two Sylow systems for $\hat {G}$ are conjugate. Hence, the number of choices for G as a subgroup of $\hat {G}$ and up to conjugacy is at most

$$ \begin{align*}\mid \!\{S_1, S_2, S_3 \mid S_{i} \leq B_{i}, |S_1|=p^{\alpha}, |S_2|=q^{\beta}, |S_3|=r^{\gamma}\! \} \mid \,\leq \, |B_1|^{\alpha} |B_2|^{\beta} |B_3|^{\gamma}. \end{align*} $$

We observe that $B_2= T_{21} \times T_{22} \times T_{23}$ , where $T_{2i}$ are Sylow q-subgroups of $\hat {G_{i}}$ for $i=1,2,3$ . From [Reference Venkataraman13, Proposition 3.1], $|T_{21}| \leq |M_1| \leq (6^{1/2})^{{\alpha }-1} p^{\alpha }$ and $|T_{23}| = |M_3| \leq (6^{1/2})^{{\gamma }-1} r^{\gamma }$ . Further, $|T_{22}|=|Q_2|=q^{\beta }$ . Hence, $|B_2| \leq (6^{1/2})^{{\alpha + \gamma }-2} p^{\alpha } q^{\beta } r^{\gamma } \leq (6^{1/2})^{{\alpha + \gamma }} n$ and so $|B_2|^{\beta } \leq (6^{1/2})^{({\alpha + \gamma })\beta } n^{\beta }$ . Similarly, we can show that $|B_3| \leq (6^{1/2})^{{\alpha + \beta }-2} p^{\alpha } q^{\beta } r^{\gamma }$ . So $|B_3|^{\gamma } \leq (6^{1/2})^{({\alpha + \beta })\gamma } n^{\gamma }$ . Putting all the estimates together, the number of choices for G as a subgroup of $\hat {G}$ up to conjugacy is at most $|B_1|^{\alpha } |B_2|^{\beta } |B_3|^{\gamma }$ , which is at most

$$ \begin{align*}p^{{\alpha}^2}\, (6^{1/2})^{({\alpha + \gamma})\beta} n^{\beta} (6^{1/2})^{({\alpha + \beta})\gamma} n^{\gamma} \leq p^{{\alpha}^2}\, (6^{1/2})^{{({\alpha + \gamma})\beta} + {({\alpha + \beta})\gamma}}n^{\beta + \gamma}.\end{align*} $$

Therefore, the number of groups of order $p^{\alpha }q^{\beta }r^{\gamma }$ in $\mathfrak {A}_p\mathfrak {A}_q\mathfrak {A}_r$ up to isomorphism is at most

$$ \begin{align*} & p^{5\alpha^{2}} \, 6^{\alpha(\alpha-1)/4} \,2^{\alpha - 1 + (23/6) \alpha \log\alpha + \alpha \log6}\, p^{{\alpha}^2}\, (6^{1/2})^{{({\alpha + \gamma})\beta} + {({\alpha + \beta})\gamma}}\, n^{\beta + \gamma} \\ &\quad= p^{6\alpha^{2}} \,2^{\alpha - 1 + (23/6) \alpha \log\alpha + \alpha \log6} (6^{1/2})^{{({\alpha + \gamma})\beta} + {({\alpha + \beta})\gamma} +{\alpha(\alpha-1)/2}}\, n^{\beta + \gamma}.\\[-36pt] \end{align*} $$

References

Blackburn, S. R., Neumann, P. M. and Venkataraman, G., Enumeration of Finite Groups (Cambridge University Press, Cambridge, 2007).CrossRefGoogle Scholar
Dickenson, G., ‘On the enumeration of certain classes of soluble groups’, Q. J. Math. 20(1) (1969), 383394.CrossRefGoogle Scholar
Hestenes, M. D., ‘Singer groups’, Canad. J. Math. 22(3) (1970), 492513.CrossRefGoogle Scholar
Higman, G., ‘Enumerating $p$ -groups. I: Inequalities’, Proc. Lond. Math. Soc. (3) 10(3) (1960), 2430.CrossRefGoogle Scholar
Lucchini, A., ‘Enumerating transitive finite permutation groups’, Bull. Lond. Math. Soc. 30(6) (1998), 569577.CrossRefGoogle Scholar
Lucchini, A., Menegazzo, F. and Morigi, M., ‘Asymptotic results for transitive permutation groups’, Bull. Lond. Math. Soc. 32(2) (2000), 191195.CrossRefGoogle Scholar
McIver, A. and Neumann, P. M., ‘Enumerating finite groups’, Q. J. Math. 38(2) (1987), 473488.CrossRefGoogle Scholar
Neumann, H., Varieties of Groups (Springer-Verlag, Berlin–Heidelberg, 1967).CrossRefGoogle Scholar
Pyber, L., ‘Enumerating finite groups of given order’, Ann. of Math. (2) 137(1) (1993), 203220.CrossRefGoogle Scholar
Scott, L., ‘Representations in characteristic $p$ ’, in: The Santa Cruz Conference on Finite Groups, Proceedings of Symposia in Pure Mathematics, 37 (eds. Cooperstein, B. and Mason, G.) (American Mathematical Society, Providence, RI, 1981).Google Scholar
Short, M. W., The Primitive Soluble Permutation Groups of Degree Less than 256 (Springer, Berlin–Heidelberg, 1992).CrossRefGoogle Scholar
Sims, C. C., ‘Enumerating $p$ -groups’, Proc. Lond. Math. Soc. (3) 15(3) (1965), 151166.CrossRefGoogle Scholar
Venkataraman, G., ‘Enumeration of finite soluble groups with Abelian Sylow subgroups’, Q. J. Math. 48(1) (1997), 107125.CrossRefGoogle Scholar
Venkataraman, G., Enumeration of Finite Soluble Groups in Small Varieties of A-groups and Associated Topics, Tech. Report, Centre for Mathematical Sciences, St. Stephen’s College (University of Delhi, Delhi, 1999).Google Scholar
Wielandt, H., Finite Permutation Groups (Academic Press, London, 1964), translated from German by R. Bercov.Google Scholar