Regular maps of order 2-powers
Dong-Dong Houa, Yan-Quan Fenga, Young Soo Kwonb
aDepartment of Mathematics, Beijing Jiaotong University, Beijing,
100044, P.R. China
bMathematics, Yeungnam University, Kyongsan 712-749, Republic of Korea
Abstract
In this paper, we consider the possible types of regular maps of order 2n, where the order of a regular map is the order of automorphism group of the map.
For n≤11, M. Conder classified all regular maps of order 2n. It is easy to classify regular maps of order 2n whose valency or covalency is 2 or 2n−1.
So we assume that n≥12 and 2≤s,t≤n−2 with s≤t to consider regular maps of order 2n with type {2s,2t}. We show that for s+t≤n or for s+t>n with s=t, there exists a regular map of order 2n with type {2s,2t}, and furthermore, we classify regular maps of order 2n with types {2n−2,2n−2} and {2n−3,2n−3}.
We conjecture that, if s+t>n with s<t, then there is no regular map of order 2n with type {2s,2t}, and we confirm the conjecture for t=n−2 and n−3.
Keywords: Regular map, 2-group, automorphism group.
2010 Mathematics Subject Classification: 20B25, 05C10.
††footnotetext:
E-mails: holderhandsome@bjtu.edu.cn, yqfeng@bjtu.edu.cn, ysookwon@ynu.ac.kr
1 Introduction
A map M is a 2-cell embedding of a connected graph into a
closed surface. The embedded graph is called the underlying
graph of the map. A map is called orientable or nonorientable
according to whether the supporting surface is orientable or
non-orientable. A map automorphism is a permutation of flags
(mutually incident vertex-edge-face triples) that preserves their
relations of having a vertex, edge or face in common, namely, it
induces an automorphism of its underlying graph which extends to a
self-homeomorphism of its supporting surface. The automorphism group
of a map always acts freely on flags. If this action is transitive
as well, the map is said to be regular. Obviously, if M is
regular, then all vertices of M have the same valency, say, l,
and all face boundary walks have the same length, say m. In this
case, M is said to have type {m,l}. If the order of automorphism group of a regular map is n,
we say that the regular map has order n, namely, the order of regular map equals to the number of flags in the map.
The study of regular
maps has a long and rich history and was progressed substantially by
Brahana [5], and Bryant and Singerman [6]. For
more development of the theory of maps, we refer the readers to
[1, 9, 16].
There are three different approaches to classify regular maps on surfaces: Classification of regular maps by underlying graphs [15],
by map automorphism groups [21, 27], and by supporting surfaces [7, 11, 13].
In this paper, we concentrate on classifications of regular maps by automorphism groups.
Malnič, Nedela and Škoviera [25] proved that each
regular map with a nilpotent automorphism group (which is called a
nilpotent regular map) can be uniquely decomposed into a direct
product of two regular maps: the automorphism group of one is a
2-group and the other map is a semistar of odd valency. This
implies that the automorphism group of a nilpotent regular map with
a simple underlying graph is a 2-group.
So the classification of regular maps whose automorphism groups are
2-groups is important to understand nilpotent regular maps. A regular map is
called a regular 2-map if its automorphism group is a
2-group. In [25], Malnič et al gave a
complete classification of nilpotent regular maps of nilpotency class
2. It is proved in [10] that given the class, there are
finitely many simple regular 2-maps. However, for the regular
2-maps with multiple edges and given class, it is possible to list
it by a computer and the classification of those seems to be
difficult. In [2], regular 2-maps of class 3 has been
classified. Furthermore, Hu et al. [19] classified
regular 2-maps for maximal class by using the classification of
2-groups with a cyclic maximal subgroup.
There are many papers [12, 17, 23, 28, 29] showing that for any given positive integers m and l satisfying 1/m+1/l≤1/2, there exist infinitely many regular maps of type {m,l} by constructive or non-constructive way. But, those papers deal with only existence of regular maps of given type without considering the order of automorphism group. In this paper, we consider the possible types of regular maps whose automorphism groups are 2-groups of order 2n for a given integer n.
For n≤11, all such regular maps are listed in
[8]. It is easy to classify regular maps of order 2n whose valency or covalency is 2 or 2n−1.
So we assume that n≥12 and 2≤s,t≤n−2 with s≤t to consider regular maps of order 2n with type {2s,2t}. We prove that for s+t≤n, or s+t>n and s=t, there exists a regular map of order 2n with type {2s,2t}, and for s+t>n, there is no regular map of order 2n with type {2s,2t} when t=n−2 or n−3. Furthermore, we classify regular maps of order 2n with types {2n−2,2n−2} and {2n−3,2n−3}.
This paper is organized as follows. In Section 2,
we give some background and properties of regular maps. In Section 3 we consider the existence of a regular map M of order 2n with given types, and in Section 4, classifications of all regular maps of order 2n with types {2n−2,2n−2} and {2n−3,2n−3} are given.
2 Preliminaries
In this section we present basic facts about regular maps. We
introduce regular maps by starting from their automorphism groups
that are known to be quotients of extended triangle
groups [6]. In all the forthcoming group presentations we
will assume that the listed exponents are the true orders of the
corresponding elements.
A finite regular map M can in this way be identified with a
(partial) three-generator presentation of a finite group G,
isomorphic to the automorphism group Aut(M) of M, of the form
[TABLE]
where dots indicate possible presence of additional relations.
In [24], Li and Širáň considered regular maps
whose automorphism groups correspond to the case that at least one
of ρ0,ρ1,ρ2 is the identity. So we just consider the case
that none of ρ0,ρ1,ρ2 is equal to 1. We will construct a
regular map M such that Aut(M)≅G.
Take topological triangles as many as the order of G. These
triangles are the flags of our map to be constructed. Each
flag is labeled by an element of G, whereby distinct flags have
distinct labels. By this way, the set of flags is identified with
the set of elements of G. Moreover, the three sides of each flag
are displayed in blue, black and red lines. In each flag, the
blue, black and red sides are labeled by ρ0, ρ1 and
ρ2, respectively.
Next, for each g∈G and each w∈{ρ0,ρ1,ρ2}, we
take the flags g and gw and identify their two sides whose
labels are w. Applying this identification procedure with all
flags, one can form a closed surface S. The union of all
red segments determines the underlying graph and its 2-cell
embedding on S constitutes our map M. By M=M(G;ρ0,ρ1,ρ2) we denote the map constructed in this way and
call M the map associated with the presentation (∗) of G.
Considering this construction of map related to the group G, one
may identify vertices, edges and faces of the map M with the left
cosets of the subgroups ⟨ρ1,ρ2⟩, ⟨ρ0,ρ2⟩ and
⟨ρ0,ρ1⟩, respectively, and their mutual incidence is
determined by non-empty intersection.
The group G has two natural actions on the flag set of M (that
is, on itself), namely, by left and right multiplication. The right
multiplication by the generators ρ0,ρ1, and ρ2 applied to
any flag is called the longitudinal reflection, the
corner reflection, and the transversal reflection of
the flag. Regarding the action of G by left multiplication, note
that if two flags h1,h2∈G are related by one of the above
reflections, then for any g∈G the flags gh1 and gh2 are
related by the same reflection. In this sense the left
multiplication corresponds to an automorphism of M. This enables
us to identify the (full) automorphism group Aut(M) of the map
M=M(G;ρ0,ρ1,ρ2) with the group G and its left action
on itself. In particular, regular maps of type {m,l}
(that is, of face length m and vertex valence l) can be
identified with presentations of finite 3-generator groups as in
(∗). This allows one to translate the entire theory of regular
maps into a purely group-theoretical language. It is
known [6, 24] that the supporting surface of M is
orientable if and only if the rotation subgroup ⟨ρ0ρ1,ρ1ρ2⟩ is a subgroup
of index 2 in G. The following proposition shows when two given
regular maps M(1)=Map(G(1),{ρ0(1),ρ1(1),ρ2(1)}) and M(2)=Map(G(2),{ρ0(2),ρ1(2),ρ2(2)}) are isomorphic.
Proposition 2.1
M(1)* and M(2) are isomorphic if and only if there is a group isomorphism f:G(1)↦G(2) such that f(ρ0(1))=ρ0(2), f(ρ1(1))=ρ1(2) and f(ρ2(1))=ρ2(2).*
For a group G with x,y∈G, denote by [x,y] the commutators x−1y−1xy of x and y, and the derived group G′ of G is the subgroup generated by all commutators [x,y] for any x,y∈G.
The following proposition is a basic property of commutators.
Proposition 2.2
Let G be a group. Then, for any x,y,z∈G, [xy,z]=[x,z]y[y,z] and [x,yz]=[x,z][x,y]z.
With Proposition 2.2, it is easy to prove that if G is generated by a subset M, then G′ is generated by all conjugates in G of elements [xi,xj] with xi,xj∈M (see [20, Hilfsatz III.1.11]). Then it is easy to obtain the following result.
Proposition 2.3
[18, Lemma 4.2]*
Let G=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ2)2⟩. Then G′=⟨[ρ0,ρ1],[ρ1,ρ2],[ρ0,ρ1]ρ2⟩.*
The Frattini subgroup, denoted by Φ(G), of a finite group G is defined to be the intersection of all maximal subgroups of G, and Φ(G) can be characterised as the set of non-generators of G.
Proposition 2.4
[14, Theorem 9.2(a)]*
Let G be a finite group and S⊂G. Then G=⟨S⟩ if and only if G=⟨S,Φ(G)⟩.*
Let G be a finite p-group for a prime p, and set ℧1(G)=⟨gp ∣ g∈G⟩. The following theorem is the well-known Burnside Basis Theorem.
Proposition 2.5
[3, Theorem 1.12]*
Let G be a p-group and ∣G:Φ(G)∣=pd.*
G/Φ(G)≅Zpd. Moreover, if N⊲G and G/N is elementary abelian, then Φ(G)≤N.
Every minimal generating set of G contains exactly d elements.
Φ(G)=G′℧1(G). In particular, if p=2, then Φ(G)=℧1(G).
The following is a classification of 2-groups having a maximal cyclic subgroup.
Proposition 2.6
[3, Theorem 1.2]*
Let G be a group of order 2n. If G has a cyclic
subgroup of order 2n−1, then G lies in one of the following six classes:*
G=⟨a ∣ a2n=1⟩, n≥1, the cyclic group.
G=⟨a,b ∣ a2n−1=1, b2=1, ab=a⟩, n≥2, the abelian group.
G=⟨a,b ∣ a2n−1=1, b2=1, ab=a−1⟩, n≥3, the dihedral group. All elements in G\⟨a⟩ are involutions.
G=⟨a,b ∣ a2n−1=1, b2=a2n−2, ab=a−1⟩, n≥3, the generalized
quaternion group. The group G contains exactly one involution, and all elements in
G\⟨a⟩ have the same order 4.
G=⟨a,b ∣ a2n−1=1,b2=1, ab=a1+2n−2⟩, n≥4.**
G=⟨a,b ∣ a2n−1=1,b2=1, ab=a−1+2n−2⟩, n≥4, the
semidihedral group.
For a group G and g∈G, we denote by o(g) the order of g in G. From Proposition 2.6, we have the following result.
Corollary 2.7
Let n≥4 and let G be a group of order 2n with a cyclic subgroup of order 2n−1. Then uv=u or u1+2n−2 for any u,v∈G with o(u)=o(v)=2n−1.
Proof. Let G(i) be the groups defined in Proposition 2.6 (i) for each i∈{1,2,3,4,5,6}. Note that G(1) and G(2) are abelian. Since n≥4, it is easy to see that G(3) and G(4) has a unique cyclic subgroup of order 2n−1. For G(6), we have (ajb)2=aj(aj)b=aj(ab)j=aj⋅2n−2 for j∈Z2n−1, that is, o(ajb)=2 or 4, and hence G(6) also has a unique cyclic subgroup of order 2n−1. Thus ⟨u⟩=⟨v⟩ for G(3), G(4) and G(6). It follows uv=u for G=G(i) with i=1,2,3,4 or 6.
Let G=G(5). It is easy to see G′=⟨a2n−2⟩≅Z2. Assume uv=u. Then ⟨u⟩=⟨v⟩ and [u,v]=a2n−2. Since o(u)=o(v)=2n−1, we have
G=⟨u,v⟩, which implies ⟨u⟩∩⟨v⟩=⟨u2⟩=⟨v2⟩. In particular, ⟨u2⟩=⟨v2⟩=⟨a2⟩, forcing a2n−2=u2n−2. It follows uv=u[u,v]=ua2n−2=u1+2n−2, as required.
The following proposition gives the automorphism group of some regular maps of order 2n for each n≥10.
Proposition 2.8
[18, Theorem 1.2 and Theorem 1.3(1)]*
Let n≥10, s,t≥2 and n−s−t≥1. Set R(ρ0,ρ1,ρ2)={ρ02,ρ12,ρ22,(ρ0ρ1)2s,(ρ1ρ2)2t,(ρ0ρ2)2, [(ρ0ρ1)4,ρ2],[ρ0,(ρ1ρ2)4]} and define*
[TABLE]
and L=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)22,(ρ1ρ2)2n−3,(ρ0ρ2)2,[(ρ0ρ1)2,ρ2](ρ1ρ2)2n−4⟩.
Then ∣H∣=∣L∣=2n and
the listed exponents are the true orders of the corresponding elements.
3 Existence of regular maps of order 2n
Let M be a regular map of order 2n with type {2s,2t} and s≤t. For n≤11, all such regular maps M are listed in [8]. Let n≥12. It is easy to see that if s=1
then Aut(M)≅C2×D2n−1 or D2n, implying t=n−2 or n−1, respectively. If t=n−1 then Aut(M)≅D2n and s is 2 or n−1. Let 2≤s,t≤n−2. In this section we prove that if either s+t≤n, or s+t>n and s=t, then there exists a regular map of order 2n with type {2s,2t} in Theorem 3.2, and if s+t>n, s=t and t=n−2 or n−3 then there is no regular map of order 2n with type {2s,2t} in Theorem 3.3.
We first prove a lemma which will be used frequently in the paper.
Lemma 3.1
Let G=⟨ρ0,ρ1,ρ2⟩ such that ρ02=ρ12=ρ22=(ρ0ρ2)2=1. Then
[(ρ0ρ1)2,ρ2]=[ρ0,(ρ1ρ2)2]ρ2ρ1* and (ρ0ρ2ρ1)2=((ρ2ρ1)2(ρ1ρ0)2)ρ0;*
If [(ρ0ρ1)2,ρ2]=1, then [ρ0,(ρ1ρ2)2]=[(ρ0ρ1)4,ρ2]=[ρ0,(ρ1ρ2)4]=1.
In particular, ⟨(ρ0ρ1)2⟩⊲G, ⟨(ρ1ρ2)2⟩⊲G, [(ρ0ρ1)2,(ρ1ρ2)2]=1 and (ρ0ρ2ρ1)2i=(ρ2ρ1)2i(ρ1ρ0)2i for any positive integer i.
Proof. Since [ρ0,ρ2]=1, we have [(ρ0ρ1)2,ρ2]=[ρ0,ρ2]ρ1ρ0ρ1[ρ1ρ0ρ1,ρ2]=[ρ1ρ0ρ1,ρ2]=[ρ0,ρ1ρ2ρ1]ρ1=[ρ0,ρ1ρ2ρ1ρ2ρ2]ρ1=[ρ0,ρ2]ρ1[ρ0,(ρ1ρ2)2]ρ2ρ1=[ρ0,(ρ1ρ2)2]ρ2ρ1 by Proposition 2.2, and (ρ0ρ2ρ1)2=ρ0ρ2ρ1ρ2ρ0ρ1=ρ0(ρ2ρ1)2ρ1ρ0ρ1=((ρ2ρ1)2(ρ1ρ0)2)ρ0.
To prove (2), let [(ρ0ρ1)2,ρ2]=1. By (1), [ρ0,(ρ1ρ2)2]=1.
Since the relation [(ρ0ρ1)2,ρ2]=1 implies that (ρ0ρ1)2 and ρ2 commute each other, we have
[(ρ0ρ1)4,ρ2]=1, and similarly, [ρ0,(ρ1ρ2)4]=1.
Since (ρ0ρ1)ρ1=ρ1ρ0=(ρ0ρ1)−1 and (ρ0ρ1)ρ0=ρ1ρ0=(ρ0ρ1)−1, we have ⟨(ρ0ρ1)2⟩⊲G as [(ρ0ρ1)2,ρ2]=1, and since [ρ0,(ρ1ρ2)2]=1, we have ⟨(ρ1ρ2)2⟩⊲G. Furthermore, ((ρ0ρ1)2)(ρ1ρ2)2=(ρ0ρ1)2, that is, [(ρ0ρ1)2,(ρ1ρ2)2]=1. By (1), (ρ0ρ2ρ1)2=((ρ2ρ1)2(ρ1ρ0)2)ρ0=(ρ2ρ1)2(ρ1ρ0)2, and it follows (ρ0ρ2ρ1)2i=(ρ2ρ1)2i(ρ1ρ0)2i for any positive integer i.
Theorem 3.2
Let n≥12 and 2≤s,t≤n−2 such that either s+t≤n or s+t>n and s=t. Then there exists a regular map M of order 2n and {2s,2t} with Aut(M)=G:
for s+t≤n−1, G=\left\{\begin{array}[]{ll}\langle\rho_{0},\rho_{1},\rho_{2}\ |\ R_{1},[(\rho_{0}\rho_{1})^{2},\rho_{2}]^{2^{\frac{n-s-t-1}{2}}}\rangle,&n-s-t\mbox{ odd }\\
\langle\rho_{0},\rho_{1},\rho_{2}\ |\ R_{1},[(\rho_{0}\rho_{1})^{2},(\rho_{1}\rho_{2})^{2}]^{2^{\frac{n-s-t-2}{2}}}\rangle,&n-s-t\mbox{ even, }\end{array}\right.
for s+t=n, G=⟨ρ0,ρ1,ρ2 ∣ R2,[(ρ0ρ1)2,ρ2],(ρ0ρ1)2s−1⋅(ρ1ρ2)2t−1⟩,
for s+t>n and s=t, G=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)2n−t−1⋅(ρ1ρ2)2n−t−1,[(ρ0ρ1)2,ρ2](ρ2ρ1)4⟩,
where R1={ρ02,ρ12,ρ22,(ρ0ρ1)2s,(ρ1ρ2)2t,(ρ0ρ2)2,[(ρ0ρ1)4,ρ2],[ρ0,(ρ1ρ2)4]} and
R2={ρ02, ρ12, ρ22,
(ρ0ρ2)2, (ρ0ρ1)2s,(ρ1ρ2)2t}.
Proof. The part (1) is true by Proposition 2.8. Now we consider parts (2) and (3), that is, the cases for s+t=n and s+t>n with s=t. To finish the proof, it suffices to show that ∣G∣=2n, o(ρ0ρ1)=2s and o(ρ1ρ2)=2t.
Case 1: s+t=n.
In this case, we have G=⟨ρ0,ρ1,ρ2 ∣ R2,[(ρ0ρ1)2,ρ2],(ρ0ρ1)2s−1(ρ1ρ2)2t−1⟩, where R2={ρ02,ρ12,ρ22,(ρ0ρ2)2,(ρ0ρ1)2s,(ρ1ρ2)2t}. Set
[TABLE]
Note that s+t<n+1 and (n+1)−s−t=1. Replacing n by n+1 in part (1), we have ∣H∣=2n+1, o(ρ0ρ1)=2s and o(ρ1ρ2)=2t in H. By Lemma 3.1(2), the relations [(ρ0ρ1)4,ρ2] and [ρ0,(ρ1ρ2)4] in H follow from other relations in H, and thus
[TABLE]
By Lemma 3.1(2), ⟨(ρ0ρ1)2⟩⊲H and ⟨(ρ1ρ2)2⟩⊲H, and since
o((ρ0ρ1)2s−1)=2 and o((ρ1ρ2)2t−1)=2, we have (ρ0ρ1)2s−1,(ρ1ρ2)2t−1∈Z(H), the center of H.
Suppose (ρ0ρ1)2s−1=(ρ1ρ2)2t−1. Then H/⟨(ρ0ρ1)2s−1⟩≅H1 with ∣H1∣=2n, where
[TABLE]
By Lemma 3.1(2), [(ρ0ρ1)4,ρ2]=1 and [ρ0,(ρ1ρ2)4]=1 in H1, and since (s−1)+(t−1)<n−1 and (n−1)−(s−1)−(t−1)=1, part (1) implies ∣H1∣=2n−1, a contradiction.
Thus (ρ0ρ1)2s−1=(ρ1ρ2)2t−1, that is, o((ρ0ρ1)2s−1⋅(ρ1ρ2)2t−1)=2 in H. Let K=⟨(ρ0ρ1)2s−1⋅(ρ1ρ2)2t−1⟩. Then o(ρ0ρ1K)=2s and o(ρ1ρ2K)=2t in H/K. Clearly, H/K≅G, and hence ∣G∣=2n, o(ρ0ρ1)=2s and o(ρ1ρ2)=2t, as required.
Case 2: s+t>n and s=t.
In this case, G=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)2n−t−1(ρ1ρ2)2n−t−1,[(ρ0ρ1)2,ρ2](ρ2ρ1)4⟩, where R2={ρ02,ρ12,ρ22,(ρ0ρ2)2,(ρ0ρ1)2t,(ρ1ρ2)2t}. Set
[TABLE]
Note that (n−t−1)+t<n and n−(n−t−1)−t=1. By part (1), we have ∣H∣=2n, o(ρ0ρ1)=2n−t−1 and o(ρ1ρ2)=2t in H. By Lemma 3.1(2), the relations [(ρ0ρ1)4,ρ2] and [ρ0,(ρ1ρ2)4] follows from other relations in H, and thus
[TABLE]
Note that (ρ1ρ0)2n−t−1=1 and (ρ2ρ1)2t=1 in H. Since s+t>n and s=t, we have 1≤n−t−1<t, and by Lemma 3.1(2), (ρ0ρ2ρ1)2t=(ρ2ρ1)2t(ρ1ρ0)2t=1 and (ρ0ρ2ρ1)2n−t−1=(ρ2ρ1)2n−t−1(ρ1ρ0)2n−t−1=(ρ2ρ1)2n−t−1, which implies o(ρ0ρ2ρ1)=o(ρ2ρ1)=2t in H and (ρ0ρ2ρ1)2n−t−1⋅(ρ1ρ2)2n−t−1=1.
Moreover, [(ρ0ρ2ρ1)2,ρ2]=[(ρ2ρ1)2(ρ1ρ0)2,ρ2]=[(ρ2ρ1)2,ρ2](ρ1ρ0)2[(ρ1ρ0)2,ρ2]=((ρ1ρ2)4)(ρ1ρ0)2=(ρ1ρ2)4, that is, [(ρ0ρ2ρ1)2,ρ2]⋅(ρ2ρ1)4=1. Clearly, H=⟨ρ0ρ2,ρ1,ρ2⟩. Now it is easy to see that the generators ρ0ρ2,ρ1,ρ2 in H satisfy the same relations as ρ0,ρ1,ρ2 do in G. Thus, there is an epimorphism ϕ:G↦H such that ρ0ϕ=ρ0ρ2, ρ1ϕ=ρ1 and ρ2ϕ=ρ2.
On the other hand, since [(ρ0ρ1)2,ρ2](ρ2ρ1)4=1 in G, Lemma 3.1(1) implies (ρ1ρ2)4=[(ρ0ρ1)2,ρ2]=[ρ0,(ρ1ρ2)2]ρ2ρ1, that is, [ρ0,(ρ1ρ2)2]=(ρ1ρ2)4 in G.
Thus, [ρ0ρ1,(ρ1ρ2)2]=[ρ0,(ρ1ρ2)2]ρ1[ρ1,(ρ1ρ2)2]=[(ρ1ρ2)4]ρ1(ρ1ρ2)4=1 and [(ρ0ρ1)2,(ρ1ρ2)2]=1.
Note that (ρ0ρ2ρ1)2=ρ2(ρ0ρ1)2ρ2(ρ2ρ1)2=(ρ0ρ1)2[(ρ0ρ1)2,ρ2](ρ2ρ1)2=(ρ0ρ1)2(ρ1ρ2)4(ρ2ρ1)2=(ρ0ρ1)2(ρ1ρ2)2. This implies (ρ0ρ2ρ1)2n−t−1=1 because (ρ0ρ1)2n−t−1⋅(ρ1ρ2)2n−t−1=1 in G and n−t−1≥1.
Furthermore, [(ρ0ρ2ρ1)2,ρ2]=[(ρ0ρ1)2(ρ1ρ2)2,ρ2]=[(ρ0ρ1)2,ρ2](ρ1ρ2)2
[(ρ1ρ2)2,ρ2]=(ρ1ρ2)4(ρ2ρ1)4=1. Thus, the generators ρ0ρ2,ρ1,ρ2 in G satisfy the same relations as ρ0,ρ1,ρ2 do in H, which implies that there is an epimorphism φ:H↦G such that ρ0φ=ρ0ρ2, ρ1φ=ρ1 and ρ2φ=ρ2.
Since ∣H∣=2n, both ϕ and φ are isomorphisms, and hence ∣G∣=2n. Since o(ρ1ρ2)=2t in H and (ρ1ρ2)φ=ρ1ρ2, we have o(ρ1ρ2)=2t in G, and since o(ρ0ρ2ρ1)=2t in H and (ρ0ρ2ρ1)φ=ρ0ρ1, we have o(ρ0ρ1)=2t in G, as required.
Theorem 3.3
Let n≥12 and 2≤s,t≤n−2 such that s+t>n and s<t. Then there is no regular map of order 2n with type {2s,2t} for t=n−2 or n−3.
Proof. Suppose to the contrary that M is a regular map of order 2n with type {2s,2t} for t=n−2 or n−3. Let G=Aut(M). Then ∣G∣=2n and Aut(M) is generated by three involutions, say ρ0,ρ1 and ρ2, such that ρ0ρ2=ρ2ρ0, o(ρ0ρ1)=2s and o(ρ1ρ2)=2t. Since ∣G∣=2n, G cannot be generated by any two of ρ0, ρ1 and ρ2, that is, {ρ0,ρ1,ρ2} is a minimal generating set of G. In particular, ∣⟨ρ1,ρ2⟩∣=2t+1 and ∣⟨ρ0,ρ1⟩∣=2s+1. Note that [ρ0,ρ1]=[ρ0,ρ2ρ2ρ1]=[ρ0,ρ2ρ1]=(ρ1ρ2)ρ0ρ2ρ1, that is, (ρ1ρ2)ρ0=(ρ0ρ1)2ρ1ρ2.
Let t=n−2. Then o(ρ1ρ2)=2n−2 and ∣⟨ρ1,ρ2⟩∣=2n−1. Since ∣G∣=2n, we have ⟨ρ1,ρ2⟩⊴G, and hence ⟨ρ1ρ2⟩⊴G, because ⟨ρ1ρ2⟩ is characteristic in ⟨ρ1,ρ2⟩. It follows (ρ1ρ2)ρ0=(ρ1ρ2)i for some i∈Z2n−2. Since o(ρ0)=2 and o(ρ1ρ2)=2n−2, we have i2=1 in Z2n−2∗, where Z2n−2∗ is the multiplicative group of Z2n−2 consisting of numbers of Z2n−2 coprime to 2.
It is well known that Z2n−2∗≅Z2n−4×Z2, and so the equation i2=1 has four solutions in Z2n−2∗, that is, i=1, −1, 2n−3+1 or 2n−3−1. It follows (ρ1ρ2)ρ0=ρ1ρ2,ρ2ρ1,ρ1ρ2(ρ1ρ2)2n−3 or ρ2ρ1(ρ1ρ2)2n−3, and hence (ρ0ρ1)2=[ρ0,ρ1]=(ρ1ρ2)ρ0ρ2ρ1=1,(ρ2ρ1)2,(ρ1ρ2)2n−3 or (ρ2ρ1)2(ρ1ρ2)2n−3, which implies s=1,2 or s=t=2n−2, but this is impossible because s+t>n and s<t.
Let t=n−3. Then o(ρ1ρ2)=2n−3 and ∣⟨ρ1,ρ2⟩∣=2n−2. Let A=⟨ρ0ρ1,ρ1ρ2⟩ and B=⟨(ρ0ρ1)2,ρ1ρ2⟩. Then A⊴G and ∣G:A∣≤2. Since {ρ0,ρ1,ρ2} is a minimal generating set, by Proposition 2.5, G has rank 3, that is, d(G)=3. This implies that d(A)=2 and ∣A∣=2n−1.
Clearly, (ρ1ρ2)ρ0ρ1=(ρ1ρ0)2(ρ2ρ1)∈B, and hence B⊴A. Thus ∣A/B∣≤2. By Proposition 2.5, (ρ0ρ1)2∈℧1(A)≤Φ(A).
If B=A, then A/Φ(A)=B/Φ(A)=⟨ρ1ρ2Φ(A)⟩, and by Proposition 2.4, A=⟨ρ1ρ2,Φ(A)⟩=⟨ρ1ρ2⟩, contradicting d(A)=2. Thus B=A and ∣B∣=2n−2.
Recall (ρ1ρ2)ρ0=(ρ0ρ1)2ρ1ρ2∈B.
Since ∣B∣=2n−2 and o(ρ1ρ2)=o((ρ1ρ2)ρ0)=2n−3, we have ⟨(ρ1ρ2)2⟩=⟨((ρ1ρ2)ρ0)2⟩, and since o((ρ1ρ2)2)=2n−4, we have ((ρ1ρ2)ρ0)2=(ρ1ρ2)2i for some i∈Z2n−4. Now (ρ1ρ2)2=((ρ1ρ2)2)ρ02=((ρ1ρ2)2i)ρ0=(((ρ1ρ2)ρ0)2)i=((ρ1ρ2)2)i2,
which implies
i2−1=0 in Z2n−4∗. It follows i=±1 or 2n−5±1.
Since ρ1ρ2 and (ρ1ρ2)ρ0 are elements of order 2n−3 in B and ∣B∣=2n−2, Corollary 2.7 implies (ρ1ρ2)(ρ1ρ2)ρ0=(ρ1ρ2) or (ρ1ρ2)1+2n−4, and since (ρ0ρ1)2=(ρ1ρ2)ρ0ρ2ρ1, we have (ρ1ρ2)(ρ0ρ1)2=(ρ1ρ2) or (ρ1ρ2)1+2n−4. It follows
((ρ1ρ2)ρ0)2=((ρ0ρ1)2ρ1ρ2)((ρ0ρ1)2ρ1ρ2)=(ρ0ρ1)4(ρ1ρ2)(ρ0ρ1)2ρ1ρ2=(ρ0ρ1)4(ρ1ρ2)2 or (ρ0ρ1)4(ρ1ρ2)2+2n−4.
Let ((ρ1ρ2)ρ0)2=(ρ0ρ1)4(ρ1ρ2)2. Then (ρ1ρ2)2i=(ρ0ρ1)4(ρ1ρ2)2, and since i=±1 or 2n−5±1, we have (ρ0ρ1)4=(ρ1ρ2)2i−2=1, (ρ2ρ1)4, (ρ1ρ2)2n−4 or (ρ2ρ1)4(ρ1ρ2)2n−4. It follows s=2,3 or s=t=2n−3, each of which is impossible because s+t>n and s<t.
Let ((ρ1ρ2)ρ0)2=(ρ0ρ1)4(ρ1ρ2)2+2n−4.
Since o(ρ1ρ2)=2n−3, we have (ρ0ρ1)4=(ρ1ρ2)2n−4+2i−2=(ρ1ρ2)2n−4, (ρ2ρ1)4(ρ1ρ2)2n−4, 1 or (ρ2ρ1)4. This means that s=2,3 or s=t=2n−3, each of which is impossible because s+t>n and s<t.
Based on Theorem 3.3, we would like to propose the following conjecture.
Conjecture 3.4
For any positive integers n,s,t such that n≥12, 2≤s,t≤n−2, s+t>n and s=t, there is no regular map of order 2n with type {2s,2t}.
A computation with Magma shows that the conjecture is true for n≤12.
4 Regular maps of order 2n with certain types
In this section, we classify regular maps of order 2n with types {2n−2,2n−2}
and {2n−3,2n−3} in Theorem 4.3. It appears that these regular maps have the following automorphism groups Gi for 1≤i≤6. Let n≥12 and let
[TABLE]
We define six groups as following:
=⟨ρ0,ρ1,ρ2 ∣ R1,(ρ0ρ1)2n−2,(ρ1ρ2)2n−2,(ρ0ρ1)2(ρ1ρ2)2⟩,
=⟨ρ0,ρ1,ρ2 ∣ R1,(ρ0ρ1)2n−2,(ρ1ρ2)2n−2,(ρ0ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−3⟩,
=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)4(ρ1ρ2)4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4⟩,
=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)4(ρ1ρ2)4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4(ρ1ρ2)2n−4⟩,
=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4(ρ1ρ2)2n−4⟩,
=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4⟩.
We first prove that the groups Gi for 1≤i≤5 have order 2n. This is also true for G6, but the proof is quite different with that given in Theorem 4.3.
Lemma 4.1
The groups Gi for 1≤i≤5 have order 2n, and
the listed exponents are the true orders of the corresponding elements in each Gi. Furthermore, ∣G6∣≤2n.
Proof. Note that (n−2)+(n−2)>n and (n−3)+(n−3)>n. By taking (s,t,n)=(n−2,n−2,n),(n−3,n−3,n) in Theorem 3.2(3), we know that both groups
=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)2n−2,(ρ1ρ2)2n−2,(ρ0ρ2)2,(ρ0ρ1)2(ρ1ρ2)2,[(ρ0ρ1)2,ρ2](ρ2ρ1)4⟩,
=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)2n−3,(ρ1ρ2)2n−3,(ρ0ρ2)2,(ρ0ρ1)4(ρ1ρ2)4,[(ρ0ρ1)2,ρ2](ρ2ρ1)4⟩,
have order 2n and
the listed exponents are the true orders of the corresponding elements in H1 and H3. For H1, we have (ρ0ρ1)2=(ρ1ρ2)−2, and since [(ρ0ρ1)2,ρ2]=(ρ0ρ1)−2ρ2(ρ0ρ1)2ρ2=(ρ1ρ2)2ρ2(ρ1ρ2)−2ρ2=(ρ1ρ2)4, the relation [(ρ0ρ1)2,ρ2](ρ2ρ1)4 in H1 is redundant and so G1=H1.
For H3, we have (ρ0ρ1)4=(ρ2ρ1)4, and hence [(ρ0ρ1)2,ρ2](ρ2ρ1)4=1 if and only if [(ρ0ρ1)2,ρ2](ρ0ρ1)4=1. It follows G3=H3.
For G2, we have (ρ0ρ1)2=(ρ2ρ1)2+2n−3 and ((ρ0ρ1)2)ρ2=(ρ1ρ2)2+2n−3. By Proposition 2.3,
G2′=⟨[ρ0,ρ1],[ρ1,ρ2],[ρ0,ρ1]ρ2⟩=⟨(ρ0ρ1)2,(ρ1ρ2)2,((ρ0ρ1)2)ρ2⟩=⟨(ρ1ρ2)2⟩, and so ∣G2′∣=o((ρ1ρ2)2)≤2n−3.
Note that G2/G2′ is abelian and is generated by three involutions. Thus ∣G2/G2′∣≤23 and hence ∣G2∣=∣G2/G2′∣⋅∣G2′∣≤2n.
By taking (s,t,n)=(2,n−2,n) in Theorem 3.2(2), we know that the group
[TABLE]
has order 2n and
the listed exponents are the true orders of the corresponding elements in H2. Since [(ρ0ρ1)2,ρ2]=1 in H2, Lemma 3.1(2) implies that
(ρ0ρ2ρ1)2=(ρ2ρ1)2(ρ1ρ0)2=(ρ2ρ1)2(ρ2ρ1)2n−3 in H2. It follows o(ρ0ρ2ρ1)=2n−2 and (ρ0ρ2ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−3=1 in H2. So the generators ρ0ρ2,ρ1,ρ2 in H2 satisfy the same relations as ρ0,ρ1,ρ2 do in G2, and hence there is an epimorphism ϕ:G2↦H2 such that ρ0ϕ=ρ0ρ2, ρ1ϕ=ρ1 and ρ2ϕ=ρ2. Since ∣H2∣=2n and ∣G2∣≤2n, we have ∣G2∣=2n, and since (ρ0ρ1)ϕ=ρ0ρ2ρ1, we have o(ρ0ρ1)=2n−2 in G2.
Let G=G4,G5 or G6. Then (ρ0ρ1)4=(ρ2ρ1)4 or (ρ2ρ1)4(ρ2ρ1)2n−4, and since (ρ1ρ2)2n−3=1, we have (ρ0ρ1)2n−4=((ρ0ρ1)4)2n−8=(ρ2ρ1)2n−4. Let K=⟨(ρ1ρ2)2n−4⟩. Then ∣K∣≤2 and K⊲G, because (ρ0ρ1)ρ0=(ρ0ρ1)−1, (ρ0ρ1)ρ1=(ρ0ρ1)−1 and (ρ1ρ2)ρ2=(ρ1ρ2)−1. The three generators ρ0K,ρ1K,ρ2K in G/K satisfy the same relations as ρ0,ρ1,ρ2 in G3, where n is replaced by n−1, and hence ∣G/K∣≤2n−1 (here, we need check ∣G3∣=211 for n=11 and this can be done by Magma). It follows ∣G∣≤2n−1⋅2=2n.
Set H4=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)4,(ρ1ρ2)2n−3,(ρ0ρ2)2,[(ρ0ρ1)2,ρ2](ρ2ρ1)2n−4⟩. Note that (ρ2ρ1)2n−4=(ρ1ρ2)2n−4 in H4. Now H4=L with L given in Proposition 2.8, and hence ∣H4∣=2n and the listed exponents are the true orders of the corresponding elements in H4.
Since [(ρ0ρ1)2,ρ2]=(ρ1ρ2)2n−4 and (ρ0ρ1)4=1 in H4, Proposition 2.2 implies [(ρ0ρ1)2,ρ1ρ2]=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ1]ρ2=(ρ1ρ2)2n−4 and hence [(ρ0ρ1)2,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ1ρ2][(ρ0ρ1)2,ρ1ρ2]ρ1ρ2=(ρ1ρ2)2n−3=1 in H4. By Lemma 3.1(1), [ρ0,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2]ρ1ρ2=(ρ1ρ2)2n−4 and hence [ρ0,(ρ1ρ2)4]=[ρ0,(ρ1ρ2)2][ρ0,(ρ1ρ2)2](ρ1ρ2)2=(ρ1ρ2)2n−4(ρ1ρ2)2n−4=1.
Again by Lemma 3.1(1), (ρ0ρ2ρ1)4=((ρ2ρ1)4(ρ1ρ0)4)ρ0=((ρ2ρ1)4)ρ0=(ρ2ρ1)4 in H4, which implies o(ρ0ρ2ρ1)=2n−3 and (ρ0ρ2ρ1)4(ρ1ρ2)4=1.
By Lemma 3.1(1), (ρ0ρ2ρ1)2=((ρ2ρ1)2(ρ1ρ0)2)ρ0,
and since [(ρ2ρ1)2(ρ1ρ0)2,ρ2])=[(ρ2ρ1)2,ρ2](ρ1ρ0)2[(ρ1ρ0)2,ρ2]=((ρ1ρ2)4)(ρ1ρ0)2(ρ1ρ2)2n−4=(ρ1ρ2)4(ρ1ρ2)2n−4, one may have [(ρ0ρ2ρ1)2,ρ2]=[(ρ2ρ1)2(ρ1ρ0)2,ρ2])ρ0=(ρ1ρ2)4(ρ1ρ2)2n−4.
Since (ρ0ρ2ρ1)4(ρ1ρ2)4=1, we have [(ρ0ρ2ρ1)2,ρ2](ρ0ρ2ρ1)4(ρ1ρ2)2n−4=1. So the generators ρ0ρ2,ρ1,ρ2 in H4 satisfy the same relations as ρ0,ρ1,ρ2 do in G4, and hence there is an epimorphism ϕ:G4↦H4 such that ρ0ϕ=ρ0ρ2, ρ1ϕ=ρ1 and ρ2ϕ=ρ2. Since ∣H4∣=2n and ∣G4∣≤2n, we have ∣G4∣=2n, and since (ρ0ρ1)ϕ=ρ0ρ2ρ1, we have o(ρ0ρ1)=o(ρ1ρ2)=2n−3 in G4.
By taking (s,t,n)=(3,n−3,n) in Theorem 3.2(2), we know that the group
[TABLE]
has order 2n and the listed exponents are the true orders of the corresponding elements in H5. Since [(ρ0ρ1)2,ρ2]=1 and (ρ0ρ1)4=(ρ1ρ0)4 in H5, by Lemma 3.1(2) we have (ρ0ρ2ρ1)4=(ρ2ρ1)4(ρ0ρ1)4=(ρ2ρ1)4(ρ2ρ1)2n−4 and (ρ0ρ2ρ1)8=(ρ2ρ1)8(ρ0ρ1)8=(ρ2ρ1)8, implying o(ρ0ρ2ρ1)=o(ρ1ρ2)=2n−3 and (ρ0ρ2ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−4=1 in H5. Again by Lemma 3.1(2), [ρ0,(ρ1ρ2)2]=1 and hence
[(ρ0ρ2ρ1)2,ρ2]=[(ρ2ρ1)2(ρ0ρ1)2,ρ2]=[(ρ2ρ1)2,ρ2](ρ0ρ1)2[(ρ0ρ1)2,ρ2]=(ρ1ρ2)4 in H5, which implies [(ρ0ρ2ρ1)2,ρ2](ρ0ρ2ρ1)4(ρ1ρ2)2n−4
=1.
Thus the generators ρ0ρ2,ρ1,ρ2 in H5 satisfy the same relations as ρ0,ρ1,ρ2 do in G5, and so there is an epimorphism ϕ:G5↦H5 such that ρ0ϕ=ρ0ρ2, ρ1ϕ=ρ1 and ρ2ϕ=ρ2. Since ∣H5∣=2n and ∣G∣≤2n, we have ∣G5∣=2n, and since (ρ0ρ1)ϕ=ρ0ρ2ρ1, we have o(ρ0ρ1)=2n−3 and o(ρ1ρ2)=2n−3 in G5.
The following concerns a quotient of a regular map of order a 2-power.
Lemma 4.2
Let n≥12 and s+t>n with 2≤s,t≤n−2. Let G=⟨ρ0,ρ1,ρ2⟩ such that ∣G∣=2n, o(ρ0)=o(ρ1)=o(ρ2)=o(ρ0ρ2)=2, o(ρ0ρ1)=2s and o(ρ1ρ2)=2t. Then N=⟨(ρ1ρ2)2t−1⟩⊴G and G/N=⟨ρ0N,ρ1N,ρ2N⟩. Furthermore, G/N=2n−1, o(ρ0N)=o(ρ1N)=o(ρ2N)=o(ρ0ρ2N)=2, o(ρ0ρ1N)=2s−1 and o(ρ1ρ2N)=2t−1.
Proof. Since ∣⟨ρ0ρ1⟩⟨ρ1ρ2⟩∣=∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣∣⟨ρ0ρ1⟩∣∣⟨ρ1ρ2⟩∣=∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣2s+t>∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣2n=∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣∣G∣, we have ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩=1. Note that ρ0 and ρ1 normalize ⟨ρ0ρ1⟩, and ρ1 and ρ2 normalize ⟨ρ1ρ2⟩. Since G=⟨ρ0,ρ1,ρ2⟩, we have ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩⊴G, and since every subgroup of ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩ is characteristic in the cyclic group ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩, it is normal in G. In particular, N=⟨(ρ0ρ1)2s−1⟩=⟨(ρ1ρ2)2t−1⟩≅Z2 and N⊴G. It follows ∣G/N∣=2n−1, G/N=⟨ρ0N,ρ1N,ρ2N⟩, o(ρ0ρ1N)=2s−1 and o(ρ1ρ2N)=2t−1.
Note that ⟨ρ0,ρ1⟩≅D2s+1, ⟨ρ1,ρ2⟩≅D2t+1 and ⟨ρ0,ρ2⟩≅Z2×Z2. Since s,t≤n−2 and ∣G∣=2n, {ρ0,ρ1,ρ2} is a minimal generating set of G. It follows ρ0,ρ0ρ2,ρ2∈N, and hence o(ρ0N)=o(ρ2N)=o(ρ0ρ2N)=2. If ρ1∈N, then ⟨ρ1⟩⊴G and hence ∣G∣≤∣⟨ρ0,ρ2⟩∣∣⟨ρ1⟩∣≤23, a contradiction. Thus ρ1∈N and o(ρ1N)=2.
Theorem 4.3
Let n≥12 and let M be a regular map of order 2n. Then
- (1)
M* has type {2n−2,2n−2} if and only if Aut(M)≅G1 or G2;*
2. (2)
M* has type {2n−3,2n−3} if and only if Aut(M)≅G3,G4,G5 or G6.*
Proof. By Lemma 4.1, G1 and G2 are automorphism groups of regular maps of order 2n and type {2n−2,2n−2}. For the necessity part in (1), let G=⟨ρ0,ρ1,ρ2⟩ be the automorphism group of a regular map of order 2n with type {2n−2,2n−2}. Then o(ρ0)=o(ρ1)=o(ρ2)=o(ρ0ρ2)=2 and o(ρ0ρ1)=o(ρ1ρ2)=2n−2. To finish the proof of part (1), we only need to show G=G1 or G2, that is, to show (ρ0ρ1)2(ρ1ρ2)2=1 or (ρ0ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−3=1 in G. This is true for n=12 by Magma. Let us begin by induction on n
Assume n≥13. Take N=⟨(ρ1ρ2)2n−3⟩. By Lemma 4.2, N⊴G and G/N is the automorphism group of some regular map of order 2n−1 with type {2n−3,2n−3}. Write G=G/N and x=xN for any x∈G. Since ∣G∣=2n−1, the induction hypothesis implies that we may assume G=G1 or G2, where
=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)2n−3,(ρ1ρ2)2n−3,(ρ0ρ2)2,(ρ0ρ1)2(ρ1ρ2)2⟩,
=⟨ρ0,ρ1,ρ2 ∣ ρ02,ρ12,ρ22,(ρ0ρ1)2n−3,(ρ1ρ2)2n−3,(ρ0ρ2)2,(ρ0ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−4⟩.
Suppose G=G2. Since ⟨(ρ1ρ2)2n−3⟩≅Z2, we have (ρ0ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−4=1 or (ρ1ρ2)2n−3, that is, (ρ0ρ1)2(ρ1ρ2)2=(ρ1ρ2)δ⋅2n−4 with δ=1 or −1. It follows (ρ0ρ1)2=(ρ1ρ2)δ⋅2n−4−2∈⟨(ρ1ρ2)2⟩, and hence [(ρ0ρ1)2,(ρ1ρ2)2]=1 and ((ρ0ρ1)2)ρ2=(ρ0ρ1)−2. Thus (ρ0ρ1)4(ρ1ρ2)4=((ρ0ρ1)2(ρ1ρ2)2)2=(ρ1ρ2)2n−3 and (ρ0ρ1)8(ρ1ρ2)8=1, which implies (ρ1ρ2)δ⋅2n−4=(ρ1ρ0)δ⋅2n−4.
Since (ρ1ρ2)2=(ρ1ρ0)2(ρ1ρ2)δ⋅2n−4=(ρ1ρ0)δ⋅2n−4+2, we have ((ρ2ρ1)2)ρ0=(ρ2ρ1)−2.
Note that [(ρ0ρ1)2,ρ2]=(ρ1ρ0)2((ρ0ρ1)2)ρ2=(ρ1ρ0)4
and [ρ0,(ρ1ρ2)2]=((ρ2ρ1)2)ρ0(ρ1ρ2)2
=(ρ1ρ2)4. By Lemma 3.1(1), [ρ0,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2]ρ1ρ2 and so (ρ1ρ0)4=(ρ1ρ2)4. It follows (ρ1ρ2)2n−3=(ρ0ρ1)4(ρ1ρ2)4=1, contrary to o(ρ1ρ2)=2n−2.
Now we have G=G1. Since N=⟨(ρ1ρ2)2n−3⟩≅Z2, we have (ρ0ρ1)2(ρ1ρ2)2=1 or (ρ1ρ2)2n−3. For the latter, (ρ0ρ1)2(ρ1ρ2)2(ρ1ρ2)2n−3=1. Thus G≅G1 or G2.
For the sufficiency part in (2), by Lemma 4.1 we only need to show that G6 has order 2n and the listed exponents are the true orders of the corresponding elements. Let
[TABLE]
Note that ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩⊴H6 because ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩ is normalized by ρ0,ρ1 and ρ2. Since (ρ0ρ1)4(ρ1ρ2)2n−4=1, we have ⟨(ρ0ρ1)4⟩⊴H6 and H6/⟨(ρ0ρ1)4⟩≅L6, where
[TABLE]
Note that [(ρ0ρ1)2,ρ2]=1 in L6. By Lemma 3.1(2), [(ρ0ρ1)4,ρ2]=1 and [ρ0,(ρ0ρ1)4]=1 in L6. Since n−1−2−(n−4)=1, we have ∣L6∣=2n−1 by taking (n,s,t)=(n−1,2,n−4) for H in Propositions 2.8. It follows ∣H6∣=∣L6∣⋅∣⟨(ρ0ρ1)4⟩∣≤2n−1⋅2=2n.
Now we claim ∣H6∣=2n and the listed exponents are the true orders of the corresponding elements. To do that, we will construct a permutation group A of order at least 2n that is an epimorphic image of H6.
Set t=2n−4 and write
ci=8t−i−1 for 0≤i≤8t−1,
ijtk=jt+8i+k and cijtk=jt+8ci+k for 0≤j≤3 and 1≤k≤8. Note that 0≤i≤8t−1 if and only if 0≤ci≤8t−1. Clearly, 1≤ijtk,cijtk≤4t. Let A=⟨a,b,c⟩, where a,b,c are permutations on the set {1,2,⋯,2n−2}, defined as
\begin{array}[]{rl}a=&\prod_{i=0}^{\frac{t}{8}-1}(i_{2t}^{1},ci_{3t}^{8})(i_{2t}^{8},ci_{3t}^{1})(i_{0}^{2},ci_{2t}^{7})(i_{t}^{2},i_{2t}^{2})(i_{3t}^{2},ci_{t}^{7})(i_{0}^{7},i_{3t}^{7})(i_{0}^{3},ci_{2t}^{6})(i_{t}^{3},i_{2t}^{3})\\
&(i_{3t}^{3},ci_{t}^{6})(i_{0}^{6},i_{3t}^{6})(i_{0}^{4},ci_{t}^{5})(i_{0}^{5},ci_{t}^{4}),\\
b=&\prod_{j=0}^{3}\prod_{i=0}^{\frac{t}{8}-1}(i_{jt}^{1},i_{jt}^{2})(i_{jt}^{3},i_{jt}^{4})(i_{jt}^{5},i_{jt}^{6})(i_{jt}^{7},i_{jt}^{8}),\\
c=&\prod_{j=0}^{3}[\prod_{i=0}^{\frac{t}{8}-1}(i_{jt}^{2},i_{jt}^{3})(i_{jt}^{4},i_{jt}^{5})(i_{jt}^{6},i_{jt}^{7})\cdot\prod_{i=0}^{\frac{t}{8}-2}(i_{jt}^{8},(i+1)_{jt}^{1})]\\
&\prod_{i=0}^{1}[(0_{0+2ti}^{1})((\frac{t}{8}-1)_{t+2ti}^{8})((\frac{t}{8}-1)_{0+2ti}^{8},0_{t+2ti}^{1})].\\
\end{array}
Here, (i+1)jt1=jt+8(i+1)+1 for 0≤i≤8t−2.
It is easy to see that a is fixed under conjugacy of c, that is, ac=a. It follows (ac)2=1.
Let α=a,b, or c. Then α is an involution. Recall that ci=8t−i−1. Since 0≤i≤8t−1 if and only if 0≤ci≤8t−1, it is easy to see that if α interchanges ij1tk1 and ij2tk2 then α also interchanges cij1tk1 and cij2tk2, and if α interchanges ij1tk1 and cij2tk2 then α also interchanges cij1tk1 and ij2tk2. These facts are very helpful for the following computations.
\begin{array}[]{lcl}ab&=&\prod_{i=0}^{\frac{t}{8}-1}(i_{0}^{1},i_{0}^{2},ci_{2t}^{8},i_{3t}^{2},ci_{t}^{8},ci_{t}^{7},i_{3t}^{1},ci_{2t}^{7})(i_{0}^{3},ci_{2t}^{5},ci_{2t}^{6},i_{0}^{4},ci_{t}^{6},i_{3t}^{4},i_{3t}^{3},ci_{t}^{5})\\
&&(i_{0}^{5},ci_{t}^{3},ci_{2t}^{4},ci_{2t}^{3},ci_{t}^{4},i_{0}^{6},i_{3t}^{5},i_{3t}^{6})(i_{0}^{7},i_{3t}^{8},ci_{2t}^{2},ci_{t}^{1},ci_{t}^{2},ci_{2t}^{1},i_{3t}^{7},i_{0}^{8}),\\
bc&=&\prod_{j=0}^{1}(1+2tj,3+2tj,\cdots,2t-1+2tj,2t+2tj,2t-2+2tj,\cdots,2+2tj),\\
(ab)^{2}&=&\prod_{i=0}^{\frac{t}{8}-1}(i_{0}^{1},ci_{2t}^{8},ci_{t}^{8},i_{3t}^{1})(i_{0}^{8},i_{3t}^{8},ci_{t}^{1},ci_{2t}^{1})(i_{0}^{2},i_{3t}^{2},ci_{t}^{7},ci_{2t}^{7})(i_{0}^{7},ci_{2t}^{2},ci_{t}^{2},i_{3t}^{7})\\
&&(i_{0}^{3},ci_{2t}^{6},ci_{t}^{6},i_{3t}^{3})(i_{0}^{6},i_{3t}^{6},ci_{t}^{3},ci_{2t}^{3})(i_{0}^{4},i_{3t}^{4},ci_{t}^{5},ci_{2t}^{5})(i_{0}^{5},ci_{2t}^{4},ci_{t}^{4},i_{3t}^{5}),\\
(ab)^{4}&=&\prod_{i=0}^{\frac{t}{8}-1}(i_{0}^{1},ci_{t}^{8},)(i_{t}^{1},ci_{0}^{8})(i_{2t}^{1},ci_{3t}^{8})(i_{3t}^{1},ci_{2t}^{8})(i_{0}^{2},ci_{t}^{7})(i_{t}^{2},ci_{0}^{7})(i_{2t}^{2},ci_{3t}^{7})(i_{3t}^{2},ci_{2t}^{7})\\
&&(i_{0}^{3},ci_{t}^{6})(i_{t}^{3},ci_{0}^{6})(i_{2t}^{6},ci_{3t}^{3})(i_{3t}^{6},ci_{2t}^{3})(i_{0}^{4},ci_{t}^{5})(i_{t}^{4},ci_{0}^{5})(i_{2t}^{4},ci_{3t}^{5})(i_{3t}^{4},ci_{2t}^{5}),\\
((ab)^{2})^{c}&=&\prod_{i=0}^{\frac{t}{8}-1}(i_{0}^{1},i_{3t}^{1},ci_{t}^{8},ci_{2t}^{8})(i_{0}^{8},ci_{2t}^{1},ci_{t}^{1},i_{3t}^{8})(i_{0}^{2},ci_{2t}^{7},ci_{t}^{7},i_{3t}^{2})(i_{0}^{7},i_{3t}^{7},ci_{t}^{2},ci_{2t}^{2})\\
&&(i_{0}^{3},i_{3t}^{3},ci_{t}^{6},ci_{2t}^{6})(i_{0}^{6},ci_{2t}^{3},ci_{t}^{3},i_{3t}^{6})(i_{0}^{4},ci_{2t}^{5},ci_{t}^{5},i_{3t}^{4})(i_{0}^{5},i_{3t}^{5},ci_{t}^{4},ci_{2t}^{4}).\\
\end{array}
The above computations imply (ab)8=1 and (bc)2n−3=1. Furthermore, (ab)−2=c(ab)2c, that is, [(ab)2,c]=(ab)−4=(ab)4. It is clear that (bc)2n−4 interchanges i0k and cit9−k as i0k+cit9−k=2t+1 (note that 1≤i0k≤t and t+1≤cit9−k≤2t), and similarly (bc)2n−4 interchanges i2tk and ci3t9−k.
It is easy to check (bc)2n−4=(ab)4=[(ab)2,c].
So the generators a,b,c of A satisfy the same relations as ρ0,ρ1,ρ2 do in H6, and hence there is an epimorphism α from H6 to A such that ρ0α=a, ρ1α=b and ρ2α=c. Clearly, A is transitive on {1,2,⋯,2n−2} and a,c∈A1, the stabilizer of 1 in A. It follows that ∣A∣≥2n. Since ∣H6∣≤2n, α is an isomorphism and ∣H6∣=∣A∣=2n. Therefore the listed exponents are the true orders of the corresponding elements, as claimed.
Recall that G6=⟨ρ0,ρ1,ρ2 ∣ R2,(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4⟩ with
R2={ρ02,ρ12,ρ22,(ρ0ρ2)2,(ρ0ρ1)2n−3,(ρ1ρ2)2n−3}. Note that in H6, (ρ1ρ0)2((ρ0ρ1)2)ρ2=[(ρ0ρ1)2,ρ2]=(ρ1ρ0)4=(ρ0ρ1)4=(ρ1ρ2)2n−4. This implies ((ρ0ρ1)2)ρ2=(ρ0ρ1)−2 and ((ρ0ρ1)2)ρ1ρ2=(ρ0ρ1)2, that is, [(ρ0ρ1)2,ρ1ρ2]=1. In particular, [(ρ0ρ1)2,(ρ1ρ2)2]=1. By Lemma 3.1(1), [ρ0,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2]ρ1ρ2=((ρ1ρ2)2n−4)ρ1ρ2=(ρ1ρ2)2n−4, that is, ((ρ2ρ1)2)ρ0=(ρ1ρ2)2n−4−2.
Now we have (ρ0ρ2ρ1)2=((ρ2ρ1)2)ρ0(ρ0ρ1)2=(ρ1ρ2)2n−4−2(ρ0ρ1)2 in H6, and since [(ρ0ρ1)2,(ρ1ρ2)2]=1, it follows that (ρ0ρ2ρ1)4=(ρ1ρ2)−4(ρ0ρ1)4 and (ρ0ρ2ρ1)8=(ρ1ρ2)−8
(ρ0ρ1)8=(ρ2ρ1)8 , which implies o(ρ0ρ2ρ1)=o(ρ1ρ2)=2n−3 in H6.
Thus, (ρ0ρ2ρ1)4(ρ1ρ2)4
(ρ1ρ2)2n−4=(ρ1ρ2)−4(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−4=1.
Since ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩⊴H6, we have ⟨(ρ1ρ2)2n−4⟩⊴H6, and since ⟨(ρ1ρ2)2n−4⟩≅Z2, we have (ρ1ρ2)2n−4∈Z(H6), the center of H6. It follows
[(ρ0ρ2ρ1)2,ρ2](ρ0ρ2ρ1)4=[(ρ2ρ1)2(ρ0ρ1)2,ρ2](ρ0ρ2ρ1)4=[(ρ2ρ1)2,ρ2](ρ0ρ1)2
[(ρ0ρ1)2,ρ2](ρ0ρ2ρ1)4=(ρ1ρ2)4(ρ0ρ1)4(ρ1ρ2)−4(ρ0ρ1)4=1.
So the generators ρ0ρ2,ρ1,ρ2 in H6 satisfy the same relations as ρ0,ρ1,ρ2 do in G6, and there is an epimorphism ϕ:G6↦H6 such that ρ0ϕ=ρ0ρ2, ρ1ϕ=ρ1 and ρ2ϕ=ρ2. Since ∣H6∣=2n and ∣G6∣≤2n, ϕ is an isomorphism. It follows ∣G6∣=2n, and since (ρ0ρ1)ϕ=ρ0ρ2ρ1, we have o(ρ0ρ1)=2n−3 and o(ρ0ρ1)=2n−3; furthermore the listed exponents in G6 are the true orders of the corresponding elements. This finishes the proof of sufficiency part in (2).
To prove the necessity part in (2), let G be the automorphism group of a regular map of order 2n and type {2n−3,2n−3}. Then o(ρ0)=o(ρ1)=o(ρ2)=o(ρ0ρ2)=2 and o(ρ0ρ1)=o(ρ1ρ2)=2n−3. We only need to show G=G3,G4,G5 or G6, and it will be done by induction on n. This is true for n=10 by Magma.
Assume n≥11. Take N=⟨(ρ1ρ2)2n−4⟩. By Lemma 4.2, N⊴G and G=G/N is the automorphism group of a regular map of order 2n−1 with type {2n−4,2n−4}. Since ∣G∣=2n−1, by induction hypothesis we may assume G=G3,G4,G5 or G6 with R={ρ02,ρ12,ρ22,(ρ0ρ1)2n−4,(ρ1ρ2)2n−4,(ρ0ρ2)2}, where
=⟨ρ0,ρ1,ρ2 ∣ R,(ρ0ρ1)4(ρ1ρ2)4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4⟩,
=⟨ρ0,ρ1,ρ2 ∣ R,(ρ0ρ1)4(ρ1ρ2)4,[(ρ0ρ1)2,ρ2](ρ0ρ1)4(ρ1ρ2)2n−5⟩,
=⟨ρ0,ρ1,ρ2 ∣ R,(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−5,[(ρ0ρ1)2,ρ2](ρ0ρ1)4(ρ1ρ2)2n−5⟩,
=⟨ρ0,ρ1,ρ2 ∣ R,(ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−5,[(ρ0ρ1)2,ρ2](ρ0ρ1)4⟩.
Let H=⟨ρ0ρ1,ρ1ρ2⟩ be the rotation subgroup of G. Then ∣G:H∣≤2. Note that ∣⟨ρ0,ρ1⟩∣=∣⟨ρ1,ρ2⟩∣=2n−2, ∣⟨ρ0,ρ2⟩∣=4 and ∣G∣=2n. This implies that {ρ0,ρ1,ρ2} is a minimal generating set of G. By Proposition 2.5 (2), G has rank 3, and since H is generated by two elements, we have ∣G:H∣=2, that is, ∣H∣=2n−1. Since ∣⟨ρ0ρ1⟩⟨ρ1ρ2⟩∣=∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣∣⟨ρ0ρ1⟩∣∣⟨ρ1ρ2⟩∣=∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣22n−6≤∣H∣=2n−1, we have ∣⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩∣≥2n−5. Then ⟨ρ0ρ1⟩∩⟨ρ1ρ2⟩ has a subgroup of order 2n−5, which is the unique subgroup of order 2n−5 in ⟨ρ0ρ1⟩ and ⟨ρ1ρ2⟩ respectively, that is, ⟨(ρ0ρ1)4⟩ and ⟨(ρ1ρ2)4⟩. It follows ⟨(ρ0ρ1)4⟩=⟨(ρ1ρ2)4⟩.
Suppose G=G4 or G5. Then [(ρ0ρ1)2,ρ2](ρ0ρ1)4(ρ1ρ2)2n−5=1 or (ρ1ρ2)2n−4
because ⟨(ρ1ρ2)2n−4⟩≅Z2. It follows [(ρ0ρ1)2,ρ2](ρ0ρ1)4=(ρ1ρ2)δ⋅2n−5 with δ=1 or −1. Since ⟨(ρ0ρ1)4⟩=⟨(ρ1ρ2)4⟩ and [(ρ0ρ1)2,ρ2]∈⟨(ρ0ρ1)4⟩, we have [(ρ0ρ1)2,ρ2]ρ0=[(ρ0ρ1)2,ρ2]ρ1=[(ρ0ρ1)2,ρ2]ρ2=[(ρ0ρ1)2,ρ2]−1.
By Proposition 2.2, [(ρ0ρ1)2,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ1ρ2ρ1]ρ2=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ2]ρ0ρ1ρ2=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ2]−1=1. On the other hand, [(ρ0ρ1)2,ρ1]=(ρ1ρ0)4∈⟨(ρ1ρ2)4⟩, and hence [(ρ0ρ1)2,ρ1]ρ2=(ρ1ρ0)−4 and
[(ρ0ρ1)2,ρ1ρ2]=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ1]ρ2=(ρ1ρ2)δ⋅2n−5(ρ1ρ0)4(ρ1ρ0)−4=(ρ1ρ2)δ⋅2n−5. It follows 1=[(ρ0ρ1)2,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ1ρ2][(ρ0ρ1)2,ρ1ρ2]ρ1ρ2=(ρ1ρ2)2n−4, which is impossible because o(ρ1ρ2)=2n−3.
Suppose G=G6. Since ⟨(ρ1ρ2)2n−4⟩≅Z2, we have (ρ0ρ1)4(ρ1ρ2)4(ρ1ρ2)2n−5=1 or (ρ1ρ2)2n−4, and [(ρ0ρ1)2,ρ2](ρ0ρ1)4=1 or (ρ1ρ2)2n−4,
which implies (ρ0ρ1)4(ρ1ρ2)4=(ρ1ρ2)δ⋅2n−5 with δ=1 or −1 and [(ρ0ρ1)2,ρ2]=(ρ1ρ2)γ⋅2n−4(ρ1ρ0)4 with γ=0 or 1. Since ⟨(ρ0ρ1)4⟩=⟨(ρ1ρ2)4⟩, we have [(ρ0ρ1)2,ρ2]∈⟨(ρ1ρ2)4⟩. By Lemma 3.1(1), we have [ρ0,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2]ρ1ρ2=[(ρ0ρ1)2,ρ2].
It follows, by
Proposition 2.2, that [(ρ0ρ1)2,(ρ1ρ2)2]=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ2]ρ0ρ1ρ2=[(ρ0ρ1)2,ρ2][(ρ0ρ1)2,ρ2]−1=1, and also [ρ0,(ρ1ρ2)4]=[ρ0,(ρ1ρ2)2][ρ0,(ρ1ρ2)2](ρ1ρ2)2=[(ρ0ρ1)2,ρ2]2=(ρ1ρ0)8. Since [ρ0,(ρ1ρ2)4]=((ρ2ρ1)4)ρ0(ρ1ρ2)4=(ρ2ρ1)−4(ρ1ρ2)4=(ρ1ρ2)8 as ⟨(ρ0ρ1)4⟩=⟨(ρ1ρ2)4⟩, we have 1=(ρ0ρ1)8(ρ1ρ2)8=((ρ0ρ1)4(ρ1ρ2)4)2=((ρ2ρ1)δ⋅2n−5)2=(ρ2ρ1)2n−4, which is
impossible because o(ρ1ρ2)=2n−3.
Thus, G=G3. Since ⟨(ρ1ρ2)2n−4⟩≅Z2, we have [(ρ0ρ1)2,ρ2](ρ0ρ1)4=1 or (ρ1ρ2)2n−4 and (ρ0ρ1)4(ρ1ρ2)4=1 or (ρ1ρ2)2n−4, and hence G=G3,G4,G5 or G6.
Acknowledgements: This work was partially supported by the National Natural Science Foundation of China (11571035, 11731002), the 111 Project of China (B16002), and the third author was supported by Basic Science Research
Program through the National Research Foundation of Korea
funded by the Ministry of Education (2015R1D1A1A09059016).