Startseite On a conjecture by Haipeng Qu
Artikel Öffentlich zugänglich

On a conjecture by Haipeng Qu

  • Marius Tărnăuceanu EMAIL logo
Veröffentlicht/Copyright: 20. Januar 2019

Abstract

In this note, we prove that D8×C2n-3 is the non-elementary abelian 2-group of order 2n, n3, whose number of subgroups of possible orders is maximal. This solves a conjecture by Haipeng Qu. A formula for counting the subgroups of an (almost) extraspecial 2-group is also presented.

1 Introduction

Let G be a finite p-group of order pn. For k=0,1,,n, we denote by sk(G) the number of subgroups of order pk of G. The starting point for our discussion is given by [7, Theorem 1.4] which proves that if p is odd and G is non-elementary abelian, then

(1.1)sk(G)sk(Mp×Cpn-3)for all 0kn,

where

Mp=a,bap=bp=cp=1,[a,b]=c,[c,a]=[c,b]=1

is the non-abelian group of order p3 and exponent p. Moreover, in [7] it is conjectured that the inequalities in (1.1) also hold for p=2. Note that in this case there is no direct analogue of Mp, as a group of exponent 2 is abelian. But the dihedral group of order 8 is close, as it is at least generated by elements of order 2.

In the current note we prove the inequalities in (1.1) for p=2 by replacing Mp with D8. This completes the work of Haipeng Qu. We also give a formula for the number of subgroups of an (almost) extraspecial 2-group depending on the number of elementary abelian subgroups of possible orders. Our main result is the following.

Theorem 1.1.

Let G be a finite non-elementary abelian 2-group of order 2n, where n3. Then

(1.2)sk(G)sk(D8×C2n-3)for all 0kn.

Throughout this paper, we will use A*B to denote the (amalgamated) central product of the groups A and B having isomorphic centres, and X*r to denote the central product of r copies of the group X. Also, we will denote by (nk)p the number of subgroups of order pk in an elementary abelian p-group of order pn. Other notation is standard and can be found in books such as [1, 4].

We recall several basic definitions and results on 2-groups that will be useful to us. A finite 2-group is called:

  1. extraspecial if Z(G)=G=Φ(G) has order 2,

  2. almost extraspecial if G=Φ(G) has order 2 and Z(G)C4,

  3. generalized extraspecial if G=Φ(G) has order 2 and GZ(G).

The structure of these groups is well known (see, for example, [2, Theorem 2.3] and [8, Lemma 3.2]):

Lemma 1.2.

Let G be a finite 2-group.

  1. If G is extraspecial, then |G|=22r+1 for some positive integer r and either GD8*r or GQ8*D8*(r-1).

  2. If G is almost extraspecial, then |G|=22r+2 for some positive integer r and GD8*r*C4.

  3. If G is generalized extraspecial, then either GE×A or G(E*C4)×A, where E is an extraspecial 2 -group and A is an elementary abelian 2 -group.

The key result of [7] is Theorem 1.3. For p=2, it states the following.

Lemma 1.3.

Let G be a finite 2-group of order 2n and let M be a normal subgroup of order 2 of G. Then

sk(G)sk(G/M×C2)for all 0kn.

We also present a result that follows from [3, Corollary 2].

Lemma 1.4.

Let G be a finite non-elementary abelian 2-group of order 2n, n3, and let L1(G) be the set of cyclic subgroups of G. Then

|L1(G)|72n-3=|L1(D8×C2n-3)|.

2 Proofs of the main results

First of all, we recall the well-known Goursat’s lemma (see e.g. [9, (4.19)]) that will be used intensively in our proofs.

Theorem 2.1.

Let A and B be two finite groups. Then every subgroup H of the direct product A×B is completely determined by a quintuple (A1,A2,B1,B2,φ), where A1A2A, B1B2B and φ:A2/A1B2/B1 is an isomorphism, more precisely

H={(a,b)A2×B2φ(aA1)=bB1}.

Moreover, we have |H|=|A1||B2|=|A2||B1|.

Next we remark that Lemma 1.3 can easily be generalized in the following way.

Lemma 2.2.

Let G be a finite 2-group of order 2n and let M be a normal subgroup of order 2r of G. Then

sk(G)sk(G/M×C2r)for all 0kn.

Proof.

Take a chief series of G containing M and use induction on r. ∎

The following lemma shows that the inequalities in (1.2) hold for finite abelian 2-groups.

Lemma 2.3.

Let G be a finite abelian 2-group of order 2n, n3. If G is non-elementary abelian, then

sk(G)sk(D8×C2n-3)for all 0kn.

Proof.

Since G is not elementary abelian, we infer that it has a subgroup M of order 2n-2 such that G/MC4. Then Lemma 2.2 leads to

sk(G)sk(C4×C2n-2)for all 0kn,

and so it suffices to prove that

(2.1)sk(G1×C2n-3)sk(D8×C2n-3)for all 0kn,

where G1=C4×C2. By Theorem 2.1, we know that a subgroup of order 2k of G1×C2n-3 is completely determined by a quintuple (A1,A2,B1,B2,φ), where A1A2G1, B1B2C2n-3, φ:A2/A1B2/B1 is an isomorphism, and |A2||B1|=2k. Note that φ can be chosen in a unique way for the elementary abelian sections of orders 1 and 2 of G1, and in 6=|Aut(C22)| ways for the elementary abelian sections of order 4 of G1. We distinguish the following cases:

Case (a): A2=1. Then A1=1 and B1=B2 is one of the (n-3k)2 subgroups of order 2k of C2n-3. Clearly, these determine (n-3k)2 distinct subgroups of the group G1×C2n-3.

Case (b): A2 is one of the three subgroups of order 2 of G1. Then B1 is of order 2k-1 and can be chosen in (n-3k-1)2 ways. If A1=A2, then we have B2=B1, while if A1=1, then B2 is one of the 2n-k-2-1 subgroups of order 2k of C2n-3 containing B1. So, in this case we have

3(n-3k-1)2+3(2n-k-2-1)(n-3k-1)2=32n-k-2(n-3k-1)2

distinct subgroups of G1×C2n-3.

Case (c): A2 is one of the two cyclic subgroups of order 4 of G1. Then B1 is of order 2k-2 and can be chosen in (n-3k-2)2 ways. If A1=A2, then we have B2=B1, while if |A1|=2, then B2 is one of the 2n-k-1-1 subgroups of order 2k-1 of C2n-3 containing B1. So, in this case we have

2(n-3k-2)2+2(2n-k-1-1)(n-3k-1)2=2n-k(n-3k-2)2

distinct subgroups of G1×C2n-3.

Case (d): A2 is the unique subgroup isomorphic to C22 of G1. Again, B1 is of order 2k-2 and can be chosen in (n-3k-2)2 ways. If A1=A2, then we have B2=B1; if |A1|=2, then B2 is one of the 2n-k-1-1 subgroups of order 2k-1 of C2n-3 containing B1; if A1=1, then B2 is one of the (n-k-12)2 subgroups of order 2k of C2n-3 containing B1. One obtains

(n-3k-2)2+3(2n-k-1-1)(n-3k-2)2+6(n-k-12)2(n-3k-2)2=22n-2k-2(n-3k-2)2

distinct subgroups of G1×C2n-3.

Case (e): A2=G1. In this case B1 is of order 2k-3 and can be chosen in (n-3k-3)2 ways. If A1=A2, then B2=B1; if |A1|=4, then B2 is one of the 2n-k-1 subgroups of order 2k-2 of C2n-3 containing B1; if A1 is the unique subgroup of order 2 of G1 such that A2/A1C22, then B2 is one of the (n-k2)2 subgroups of order 2k-1 of C2n-3 containing B1. One obtains

(n-3k-3)2+3(2n-k-1)(n-3k-3)2+6(n-k2)2(n-3k-3)2=22n-2k(n-3k-3)2

distinct subgroups of G1×C2n-3.

By summing all the above quantities, we get

sk(G1×C2n-3)=(n-3k)2+32n-k-2(n-3k-1)2+2n-k(2n-k-2+1)(n-3k-2)2
+22n-2k(n-3k-3)2.

A similar computation leads to

sk(D8×C2n-3)=(n-3k)2+52n-k-2(n-3k-1)2+2n-k-1(2n-k+1)(n-3k-2)2
+22n-2k(n-3k-3)2.

It is now clear that the inequalities in (2.1) are true, completing the proof. ∎

In what follows we will focus on describing the subgroup lattice of an (almost) extraspecial 2-group G. This can be easily made by using [2, Lemma 2.6].

Lemma 2.4.

If G is an (almost) extraspecial 2-group, then L(G) consists of:

  1. the trivial subgroup.

  2. all subgroups containing Φ(G); moreover, these are the normal non-trivial subgroups of G.

  3. all complements of Φ(G) in the elementary abelian subgroups of order 4 containing Φ(G); moreover,

    1. Φ(G) has 2i complements in an elementary abelian subgroup of order 2i+1 containing Φ(G),

    2. two non-normal subgroups H and K of G are conjugate if and only if HΦ(G)=KΦ(G),

    3. given two non-normal subgroups H and K of G, if HxK for some x[G/NG(H)], then x is the unique element of [G/NG(H)] with this property.

It is well known that the order of maximal elementary abelian subgroups of an extraspecial 2-group of order 22r+1 or of an almost extraspecial 2-group of order 22r+2 is 2r+1. Let us denote by ei(G) the number of elementary abelian subgroups of order 2i containing Φ(G), i=2,3,,r+1. Under this notation, a formula for the number of subgroups of G can be inferred from Lemma 2.4.

Corollary 2.5.

If G is an extraspecial 2-group of order 22r+1, then

(2.2)|L(G)|=1+i=02r(2ri)2+i=1rei+1(G)2i,

while if G is an almost extraspecial 2-group of order 22r+2, then

(2.3)|L(G)|=1+i=02r+1(2r+1i)2+i=1rei+1(G)2i.

In the particular cases r=2 and r=1, equalities (2.2) and (2.3), respectively, lead to some known results (see e.g. [5, Section 3.3] and [6, Example 4.5]).

Examples.

(a) For G=D8*D8 we obtain e2(G)=9 and e3(G)=6, implying that

|L(G)|=1+i=04(4i)2+29+46=110.

(b) For G=Q8*D8 we obtain e2(G)=5 and e3(G)=0, implying that

|L(G)|=1+i=04(4i)2+25=78.

(c) For G=D8*C4 we obtain e2(G)=3, and so

|L(G)|=1+i=03(3i)2+23=23.

Remark.

As we have seen above, the subgroup structure of an (almost) extraspecial 2-group G depends on its elementary abelian subgroups containing the subgroup Φ(G)=x. We remark that these are the totally singular subspaces of G/Φ(G) with respect to the quadratic form q:G/Φ(G)𝔽2, where q(v¯) is the element a𝔽2 such that v2=xa for all v¯G/Φ(G).

Next we will compare the subgroup lattices of an (almost) extraspecial 2-group G of order 2n, n3, and of D8×C2n-3. Since Φ(D8×C2n-3) is of order 2, it follows that D8×C2n-3/Φ(D8×C2n-3) and G/Φ(G) have the same dimension over 𝔽2, namely n-1. Also, we note that the order of maximal elementary abelian subgroups of D8×C2n-3 is 2n-1, which is greater than or equal to the order of maximal elementary abelian subgroups of G. We infer that

|L(D8×C2n-3)|=1+i=0n-1(n-1i)2+i=1n-2ei+1(D8×C2n-3)2i.

The following lemma will be crucial in our proof.

Lemma 2.6.

Under the above notation, we have e2(G)e2(D8×C2n-3).

Proof.

By Lemma 1.4, we know that

|L1(G)||L1(D8×C2n-3)|.

Let ci(G) be the number of cyclic subgroups of order 2i of G. Since both G and D8×C2n-3 are of exponent 4, the above inequality can be rewritten as

1+c2(G)+c4(G)1+c2(D8×C2n-3)+c4(D8×C2n-3).

On the other hand, we have

2n=1+c2(G)+2c4(G)=1+c2(D8×C2n-3)+2c4(D8×C2n-3)

and consequently

2n-c4(G)2n-c4(D8×C2n-3),

i.e.

(2.4)c4(D8×C2n-3)c4(G).

It is clear that the cyclic subgroups of order 4 of these groups contain the Frattini subgroup. Thus (2.4) leads to e2(G)e2(D8×C2n-3), as desired. ∎

Since any elementary abelian subgroup of G containing Φ(G) is a direct sum of elementary abelian subgroups of order 4, from Lemma 2.6 we infer that

ei(G)ei(D8×C2n-3)for all i.

An immediate consequence of this fact is that the inequalities in (1.2) also hold for (almost) extraspecial 2-groups.

Corollary 2.7.

If G is an (almost) extraspecial 2-group of order 2n, n3, then

sk(G)sk(D8×C2n-3)for all 0kn.

A similar thing can be said about elementary abelian sections of the groups G and D8×C2n-3.

Corollary 2.8.

Given a 2-group G, we denote by S(α,β)(G) the set of all elementary abelian sections H2/H1C2α of G with |H1|=2β. If G is (almost) extraspecial of order 2n, then

|𝒮(α,β)(G)||𝒮(α,β)(D8×C2n-3)|for all α and β.

Proof.

Write

𝒮(α,β)(G)=𝒮(α,β)1(G)𝒮(α,β)2(G)𝒮(α,β)3(G)𝒮(α,β)4(G),

where

𝒮(α,β)1(G)={H2/H1𝒮(α,β)(G)Φ(G)H1},
𝒮(α,β)2(G)={H2/H1𝒮(α,β)(G)H1=1},
𝒮(α,β)3(G)={H2/H1𝒮(α,β)(G)Φ(G)H2,H11},
𝒮(α,β)4(G)={H2/H1𝒮(α,β)(G)Φ(G)H2,Φ(G)H1,H11}.

Since G/Φ(G)D8×C2n-3/Φ(D8×C2n-3), we have

(2.5)|𝒮(α,β)1(G)|=|𝒮(α,β)1(D8×C2n-3)|.

Clearly,

𝒮(α,β)2(G)=𝒮(α,0)(G)

is the number of elementary abelian subgroups of order 2α of G, and so

|𝒮(α,β)2(G)|=eα(G)+eα+1(G)2α,

implying that

(2.6)|𝒮(α,β)2(G)||𝒮(α,β)2(D8×C2n-3)|.

Furthermore, we observe that every section H2/H1𝒮(α,β)3(G) determines a section H2Φ(G)/H1Φ(G)𝒮(α,β+1)1(G) with

H2Φ(G)C2α+β+1andH1Φ(G)C2β+1.

Conversely, every section A/B𝒮(α,β+1)1(G) with

AC2α+β+1andBC2β+1

determines 2β sections H2/H1𝒮(α,β)3(G). As there are eα+β+1(G)(α+β+1β+1)2 such sections A/B𝒮(α,β+1)1(G), we infer that

(2.7)|𝒮(α,β)3(G)|=eα+β+1(G)(α+β+1β+1)22β|𝒮(α,β)3(D8×C2n-3)|.

Let H2/H1𝒮(α,β)4(G). Then Φ(H2)H1. On the other hand, we have

Φ(H2)Φ(G)

because H2 is normal in G. Thus Φ(H2)=1, i.e. H2 is elementary abelian, and it can be chosen in eα+β(G) ways. Also, H1 is a complement of Φ(G) in one of the (α+ββ+1)2 subgroups of order 2β+1 of H2, and it can be chosen in (α+ββ+1)22β ways. Hence

(2.8)|𝒮(α,β)4(G)|=eα+β(G)(α+ββ+1)22β|𝒮(α,β)4(D8×C2n-3)|.

Obviously, relations (2.5)–(2.8) lead to

|𝒮(α,β)(G)||𝒮(α,β)(D8×C2n-3),

as desired. ∎

We are now able to prove our main result.

Proof of Theorem 1.1.

Let |G|=2m. If m=0, then G is abelian and the conclusion follows from Lemma 2.3. Assume that m1.

If G/G is not elementary abelian, then

sk(G)sk(G/G×C2m)sk(D8×C2n-3)for all 0kn,

where the first inequality is obtained by Lemma 2.2, while the second one by Lemma 2.3.

If G/G is elementary abelian, then G=Φ(G). Let M be a normal subgroup of G such that MG and [G:M]=2. Then

sk(G)sk(G1×C2m-1)for all 0kn,

where G1=G/M satisfies the conditions

G1=Φ(G1),|G1|=2andG1Z(G1),

i.e. it is a generalized extraspecial 2-group. Then either

G1E×AorG1(E*C4)×A,

where E is an extraspecial 2-group and A is an elementary abelian 2-group, by Lemma 1.2. In other words, G1 is a direct product of an (almost) extraspecial 2-group and an elementary abelian 2-group. So, it suffices to prove that if G2 is an (almost) extraspecial 2-group of order 2q, then

(2.9)sk(G2×C2n-q)sk((D8×C2q-3)×C2n-q)for all 0kn.

Let A be one of the groups G2 or D8×C2q-3. From Theorem 2.1 it follows that a subgroup HA×C2n-q of order 2k is completely determined by a quintuple (A1,A2,B1,B2,φ), where A1A2A, B1B2C2n-q, φ:A2/A1B2/B1 is an isomorphism, and |A2||B1|=2k. By fixing the section B2/B1 of C2n-q and φAut(B2/B1), we infer that H depends only on the choice of the section A2/A1𝒮(α,β)(A), where (α,β){(i,j)i=0,1,,k,j=0,1,,k-i}. So, to prove the inequalities in (2.9), it suffices to compare the numbers of elementary abelian sections A2/A1 of the two groups G2 and D8×C2q-3, where |A1| and |A2| are arbitrary. It is now clear that the conclusion follows from Corollary 2.8, completing the proof. ∎

Finally, we mention that a result similar to Lemma 1.4 can be obtained from Theorem 1.1.

Corollary 2.9.

Let G be a finite non-elementary abelian 2-group of order 2n, n3. Then

|L(G)||L(D8×C2n-3)|.

Communicated by Andrea Lucchini


Acknowledgements

The author is grateful to the reviewer for their remarks which improve the previous version of the paper.

References

[1] Y. Berkovich, Groups of Prime Power Order. Vol. 1, De Gruyter, Berlin, 2008. 10.1515/9783110208221Suche in Google Scholar

[2] S. Bouc and N. Mazza, The Dade group of (almost) extraspecial p-groups, J. Pure Appl. Algebra 192 (2004), 21–51. 10.1016/j.jpaa.2004.02.008Suche in Google Scholar

[3] M. Garonzi and I. Lima, On the number of cyclic subgroups of a finite group, Bull. Brazil. Math. Soc. 49 (2018), 515–530. 10.1007/s00574-018-0068-xSuche in Google Scholar

[4] I. M. Isaacs, Finite Group Theory, American Mathematical Society, Providence, 2008. 10.1090/gsm/092Suche in Google Scholar

[5] D. Lewis, Containment of subgroups in a direct product of groups, Ph.D. thesis, Binghamton University, 2011. Suche in Google Scholar

[6] D. Lewis, A. Almousa and E. Elert, Embedding properties in central products, preprint (2014), https://arxiv.org/abs/1408.0076. Suche in Google Scholar

[7] H. Qu, Finite non-elementary abelian p-groups whose number of subgroups is maximal, Israel J. Math. 195 (2013), 773–781. 10.1007/s11856-012-0114-0Suche in Google Scholar

[8] R. Stancu, Almost all generalized extraspecial p-groups are resistant, J. Algebra 249 (2002), 120–126. 10.1006/jabr.2001.9069Suche in Google Scholar

[9] M. Suzuki, Group Theory. I, Springer, Berlin, 1982. 10.1007/978-3-642-61804-8Suche in Google Scholar

[10] M. Suzuki, Group Theory. II, Springer, Berlin, 1986. 10.1007/978-3-642-86885-6Suche in Google Scholar

Received: 2018-10-18
Revised: 2018-11-19
Published Online: 2019-01-20
Published in Print: 2019-05-01

© 2019 Walter de Gruyter GmbH, Berlin/Boston

Heruntergeladen am 7.10.2025 von https://www.degruyterbrill.com/document/doi/10.1515/jgth-2018-0202/html?licenseType=limitedFree
Button zum nach oben scrollen