Groups containing locally maximal product-free sets of size 4

Every locally maximal product-free set S in a finite group G satisfies G = S ∪ SS ∪ S⁻¹S ∪ SS⁻¹ ∪ √S, where SS = {xy | x, y ∈ S}, S⁻¹S = {x⁻¹y | x, y ∈ S}, SS⁻¹ = {xy⁻¹ | x, y ∈ S} and √S = {x ∈ G | x² ∈ S}. To better understand locally maximal product-free sets, Bertram asked whether every locally...

Full description

Saved in:
Bibliographic Details
Published in:Algebra and Discrete Mathematics
Date:2021
Main Author: Anabanti, C.S.
Format: Article
Language:English
Published: Інститут прикладної математики і механіки НАН України 2021
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/188705
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Groups containing locally maximal product-free sets of size 4 / C.S. Anabanti // Algebra and Discrete Mathematics. — 2021. — Vol. 31, № 2. — С. 167–194. — Бібліогр.: 12 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-188705
record_format dspace
spelling Anabanti, C.S.
2023-03-11T15:51:12Z
2023-03-11T15:51:12Z
2021
Groups containing locally maximal product-free sets of size 4 / C.S. Anabanti // Algebra and Discrete Mathematics. — 2021. — Vol. 31, № 2. — С. 167–194. — Бібліогр.: 12 назв. — англ.
1726-3255
DOI:10.12958/adm1347
2020 MSC: 20D60, 05E15, 11B75
https://nasplib.isofts.kiev.ua/handle/123456789/188705
Every locally maximal product-free set S in a finite group G satisfies G = S ∪ SS ∪ S⁻¹S ∪ SS⁻¹ ∪ √S, where SS = {xy | x, y ∈ S}, S⁻¹S = {x⁻¹y | x, y ∈ S}, SS⁻¹ = {xy⁻¹ | x, y ∈ S} and √S = {x ∈ G | x² ∈ S}. To better understand locally maximal product-free sets, Bertram asked whether every locally maximal product-free set S in a finite abelian group satisfy |√S| ≤ 2|S|. This question was recently answered in the negation by the current author. Here, we improve some results on the structures and sizes of finite groups in terms of their locally maximal product-free sets. A consequence of our results is the classification of abelian groups that contain locally maximal product-free sets of size 4, continuing the work of Street, Whitehead, Giudici and Hart on the classification of groups containing locally maximal product-free sets of small sizes. We also obtain partial results on arbitrary groups containing locally maximal product-free sets of size 4, and conclude with a conjecture on the size 4 problem as well as an open problem on the general case.
en
Інститут прикладної математики і механіки НАН України
Algebra and Discrete Mathematics
Groups containing locally maximal product-free sets of size 4
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title Groups containing locally maximal product-free sets of size 4
spellingShingle Groups containing locally maximal product-free sets of size 4
Anabanti, C.S.
title_short Groups containing locally maximal product-free sets of size 4
title_full Groups containing locally maximal product-free sets of size 4
title_fullStr Groups containing locally maximal product-free sets of size 4
title_full_unstemmed Groups containing locally maximal product-free sets of size 4
title_sort groups containing locally maximal product-free sets of size 4
author Anabanti, C.S.
author_facet Anabanti, C.S.
publishDate 2021
language English
container_title Algebra and Discrete Mathematics
publisher Інститут прикладної математики і механіки НАН України
format Article
description Every locally maximal product-free set S in a finite group G satisfies G = S ∪ SS ∪ S⁻¹S ∪ SS⁻¹ ∪ √S, where SS = {xy | x, y ∈ S}, S⁻¹S = {x⁻¹y | x, y ∈ S}, SS⁻¹ = {xy⁻¹ | x, y ∈ S} and √S = {x ∈ G | x² ∈ S}. To better understand locally maximal product-free sets, Bertram asked whether every locally maximal product-free set S in a finite abelian group satisfy |√S| ≤ 2|S|. This question was recently answered in the negation by the current author. Here, we improve some results on the structures and sizes of finite groups in terms of their locally maximal product-free sets. A consequence of our results is the classification of abelian groups that contain locally maximal product-free sets of size 4, continuing the work of Street, Whitehead, Giudici and Hart on the classification of groups containing locally maximal product-free sets of small sizes. We also obtain partial results on arbitrary groups containing locally maximal product-free sets of size 4, and conclude with a conjecture on the size 4 problem as well as an open problem on the general case.
issn 1726-3255
url https://nasplib.isofts.kiev.ua/handle/123456789/188705
citation_txt Groups containing locally maximal product-free sets of size 4 / C.S. Anabanti // Algebra and Discrete Mathematics. — 2021. — Vol. 31, № 2. — С. 167–194. — Бібліогр.: 12 назв. — англ.
work_keys_str_mv AT anabantics groupscontaininglocallymaximalproductfreesetsofsize4
first_indexed 2025-11-26T21:44:28Z
last_indexed 2025-11-26T21:44:28Z
_version_ 1850773956941840384
fulltext “adm-n2” — 2021/7/19 — 10:26 — page 167 — #3 © Algebra and Discrete Mathematics RESEARCH ARTICLE Volume 31 (2021). Number 2, pp. 167–194 DOI:10.12958/adm1347 Groups containing locally maximal product-free sets of size 4 C. S. Anabanti Communicated by I. Ya. Subbotin Abstract. Every locally maximal product-free set S in a finite group G satisfies G = S ∪ SS ∪ S−1S ∪ SS−1 ∪ √ S, where SS = {xy | x, y ∈ S}, S−1S = {x−1y | x, y ∈ S}, SS−1 = {xy−1 | x, y ∈ S} and √ S = {x ∈ G | x2 ∈ S}. To better understand locally maximal product-free sets, Bertram asked whether every locally maximal product-free set S in a finite abelian group satisfy | √ S| 6 2|S|. This question was recently answered in the negation by the current author. Here, we improve some results on the structures and sizes of finite groups in terms of their locally maximal product- free sets. A consequence of our results is the classification of abelian groups that contain locally maximal product-free sets of size 4, continuing the work of Street, Whitehead, Giudici and Hart on the classification of groups containing locally maximal product-free sets of small sizes. We also obtain partial results on arbitrary groups containing locally maximal product-free sets of size 4, and conclude with a conjecture on the size 4 problem as well as an open problem on the general case. 1. Introduction Let S be a non-empty subset of a finite group G. Then S is product- free in G if there is no solution to the equation xy = z for x, y, z ∈ S; equivalently, if S ∩ SS = ∅, where SS = {xy | x, y ∈ S}. For a finite group G, a locally maximal product-free set in G is a product-free subset 2020 MSC: 20D60, 05E15, 11B75. Key words and phrases: product-free sets, locally maximal, maximal, groups. “adm-n2” — 2021/7/19 — 10:26 — page 168 — #4 168 Groups containing LMPFS of size 4 S of G such that given any other product-free set T in G with S ⊆ T , then S = T . Since every product-free set in a finite group G is contained in a locally maximal product-free set in G, we can gain information about product-free sets in a group by studying its locally maximal product-free sets. In connection with Group Ramsey Theory, Street and Whitehead [11] noted that every partition of a finite group G (or in fact, of G∗ = G \ {1}) into product-free sets can be embedded into a covering by locally maximal product-free sets, and hence to find such partitions, it is useful to understand locally maximal product-free sets. Among other results, they calculated locally maximal product-free sets in groups of small orders, up to 16 in [11,12] as well as a few higher sizes. Giudici and Hart [9] started the classification of finite groups containing small locally maximal product-free sets. They classified finite groups containing locally maximal product-free sets of sizes 1 and 2, as well as some of size 3. The size 3 problem was resolved in [5]. The reader may see [1, 4, 6, 11] for a concept ‘filled groups’ studied for locally maximal product-free sets. A locally maximal product- free set in a group G can be characterised as a product-free subset S of G satisfying G = S ∪ SS ∪ S−1S ∪ SS−1 ∪ √ S, (1) where SS = {xy | x, y ∈ S}, S−1S = {x−1y | x, y ∈ S}, SS−1 = {xy−1 | x, y ∈ S} and √ S = {x ∈ G | x2 ∈ S} (see [9, Lemma 3.1]). Each (locally maximal product-free) set S in a finite group of odd order satisfies | √ S| = |S|. No such result is known for finite groups of even order in general. To better understand locally maximal product-free sets (LMPFS for short), Bertram [7, p. 41] asked the question: does every locally maximal product-free set S in a finite abelian group satisfy | √ S| 6 2|S|? This question was answered in the negation in [3, pp. 2–3]. The answer shows that we can’t rely on | √ S| 6 2|S| to obtain a reasonable bound on the order of an arbitrary finite abelian group of even order containing a locally maximal product-free set S. The set S = {x2} consisting of the unique involution in the Quaternion group Q8 is locally maximal product-free in Q8 and satisfies | √ S| = 6|S| = 3 4 |Q8|. This shows that relying on | √ S| 6 2|S| to obtain a bound on all non-abelian finite groups containing locally maximal product-free sets will be disastrous too. We note that √ S cannot always be removable from equation 1 as seen in Remark 3.4(a) of this paper, with the locally maximal product-free subset S = {x, x6} of G = 〈x | x7 = 1〉 ∼= C7. Though not every locally maximal product-free set is a maximal (by cardinality) product-free set, the subset S of C7 given here is clearly a maximal by cardinality product-free set. In particular, “adm-n2” — 2021/7/19 — 10:26 — page 169 — #5 C. S. Anabanti 169 |S ∪ SS ∪ SS−1 ∪ S−1S| = |S ∪ SS| = 5 and | √ S| = 2. Unfortunately, this example shows that the proof of Theorem 3 of [10] is not correct, as the author assumed that every element of a finite group G which is not an element of a maximal product-free subset S of G is either an element of SS, SS−1 or S−1S. The sizes of SS, SS−1 and S−1S were optimised in [10, Theorem 3] that it is difficult to get a counter example of the theorem itself, even though the absence of √ S made the proof wrong. We devote this paper to obtaining structures and sizes of finite groups in terms of their locally maximal product-free sets S, without necessarily relying on | √ S| 6 2|S|. Throughout this discussion, G is an arbitrary finite group, except where otherwise stated. 2. Preliminaries Let S and V be subsets of G. We define the following: SV = {sv | s ∈ S, v ∈ V }; S−1 = {s−1 | s ∈ S}; T (S) = S ∪ SS ∪ SS−1 ∪ S−1S; √ S = {x ∈ G | x2 ∈ S}. Lemma 2.1 ([9, Lemma 3.1]). Let S be a product-free set in a group G. Then S is locally maximal product-free if and only if G = T (S) ∪ √ S. Proposition 2.2 ([9, Proposition 3.2]). Let S be a LMPFS in G. Then 〈S〉 is a normal subgroup of G. Furthermore, G/〈S〉 is either trivial or an elementary abelian 2-group. Theorem 2.3 ([9, Theorem 3.4]). If S is a LMPFS in G, then |G| 6 2|T (S)| · |〈S〉|. Notation 2.4. Let S ⊆ G. We define Ŝ := {s ∈ S | √ {s} 6⊂ 〈S〉}. Proposition 2.5 ([9, Proposition 3.6]). Suppose S is a LMPFS in G and that 〈S〉 is not an elementary abelian 2-group. If |Ŝ| = 1, then |G| = 2|〈S〉|. Proposition 2.6 ([9, Proposition 3.7]). Suppose S is locally maximal product-free in G. Then every element s of Ŝ has even order. Moreover all odd powers of s lie in S. Proposition 2.7 ([9, Proposition 3.8]). Let S be a LMPFS in G. If there exists s ∈ S and integers m1, . . . ,mt such that Ŝ = {s, sm1 , . . . , smt}, then |G| divides 4|〈S〉|. “adm-n2” — 2021/7/19 — 10:26 — page 170 — #6 170 Groups containing LMPFS of size 4 Lemma 2.8 ([9, Lemma 3.9]). Suppose S is a locally maximal product-free set in a group G. If S ∩ S−1 = ∅, then G = T (S) ∪ T (S)−1. Corollary 2.9 ([9, Corollary 3.10]). If S is a LMPFS in G such that S ∩ S−1 = ∅, then |G| 6 4|S|2 + 1. We write D2n = 〈x, y | xn = y2 = (xy)2 = 1〉 for the finite dihedral group of order 2n. Lemma 2.10 ([6, Lemma 3.10]). There is no locally maximal product-free set of size 4 consisting of at most one involution in a finite dihedral group. Theorem 2.11 ([6, Theorem 3.11]). Suppose S is a LMPFS of size 4 in a finite dihedral group G. Then up to automorphisms of G, the possible choices are given as follows: |G| S 8 {y, xy, x2y, x3y}, {x, x3, y, x2y} 10 {x2, x3, y, x4y} 12 {x3, y, xy, x2y}, {x2, x3, y, x5y}, {x, x5, y, x3y}, {x, x4, y, x3y} 14 {x2, x3, y, x6y} 16 {x2, x3, y, x7y}, {x, x6, y, x4y} 18 {x2, x5, y, x8y} 20 {x, x8, y, x5y} 3. Main results Let S be a locally maximal product-free set in a finite group G. If the exponent of G is 2, then S−1S = SS = SS−1 and √ S = ∅; so G = S ∪ SS. If the exponent of G is 3, then for x ∈ √ S, we have that x2 ∈ S, so x4 = x ∈ SS, and we conclude that √ S ⊆ SS. In the light of Equation (1) therefore G = S ∪ SS ∪ S−1S ∪ SS−1; whence |G| = |S ∪ SS ∪ SS−1 ∪ S−1S| 6 3|S|2 − |S|+ 1 since |SS| 6 |S|2 and |SS−1∪S−1S| 6 2|S|2−2|S|+1. Now, suppose the exponent of G is 4. If S ∩ S−1 = ∅, then S consists of elements of order 4 only, and as the square roots of elements of order 4 have order 8, we have that √ S = ∅; thus G = S∪SS∪S−1S∪SS−1. Again, |G| 6 3|S|2−|S|+1. We begin this study by examining locally maximal product-free sets S with the property that S ∩ S−1 = ∅. Lemma 2.8 is that if S is a locally maximal product-free set in a finite group G such that S ∩ S−1 = ∅, then G = S ∪ SS ∪ S−1S ∪ SS−1 ∪ S−1 ∪ (SS)−1. Corollary 2.9 is that “adm-n2” — 2021/7/19 — 10:26 — page 171 — #7 C. S. Anabanti 171 if S is a locally maximal product-free set in a finite group G such that S ∩ S−1 = ∅, then |G| 6 4|S|2 + 1. The first result here (Theorem 3.2 below) improves Lemma 2.8 and Corollary 2.9. For finite groups of odd order, Proposition 3.3 below gives a much tighter and broader result for Corollary 2.9. Notation 3.1. For a subset S of a group G, we write I(S) for the set of all involutions in S. Theorem 3.2. Suppose S is a LMPFS in a finite group G such that S ∩ S−1 = ∅. Then G = SS ∪ S−1S ∪ SS−1 ∪ (SS)−1. Moreover, |G| 6 4|S|2 − 2|S| − |I(G)|+ 1. Proof. First, S−1 ∩ SS−1 = ∅; for if x−1 = yz−1 where x, y, z ∈ S, then z = xy, a contradiction. Similarly,S−1∩S−1S = ∅. In the light of Equation (1) therefore S−1 ⊆ SS ∪ √ S. Let x−1 ∈ S−1 be arbitrary. Suppose x−1 ∈ √ S. Then x−2 ∈ S. As x ∈ S, we have that (x)(x−2) = x−1 ∈ SS. So, S−1 ⊆ SS; whence S ⊆ (SS)−1, and G = SS∪S−1S∪SS−1∪ (SS)−1 follows from Lemma 2.8. Clearly, |G| 6 4|S|2 − 2|S|+ 1. However, each involution used in obtaining the bound of |G| given as “ |G| 6 4|S|2 − 2|S|+1" is counted at least twice. Now, x ∈ SS if and only if x ∈ (SS)−1; so all such involutions are counted twice. If x ∈ S−1S, then x = y−1z for some y, z ∈ S. Thus, x = z−1y as well. So x is counted at least twice in S−1S. The same applies to x ∈ SS−1. By removing the second count on involutions of G, we obtain that |G| 6 4|S|2 − 2|S| − |I(G)|+ 1. Proposition 3.3. If S is a locally maximal product-free set of size k in a finite group G of odd order and 0 6 |S ∩ S−1| = l 6 k, then |G| 6 3k2 − l(2k − 1) + 1. Proof. Suppose S is a LMPFS of size k in a finite group G of odd order and 0 6 |S ∩ S−1| = l 6 k. As each element of S has a unique square root, we have that | √ S| = k. If l = 0, then |SS−1 ∪ S−1S| 6 2k2 − 2k+ 1 and |SS| 6 k2; so the result follows from Equation (1). Now, suppose 1 6 |S ∩ S−1| = l 6 k. Let S = {x1, · · · , xl, xl+1, · · · , xk} and S−1 = {x1, · · · , xl, x−1 l+1, · · · , x−1 k }. Then |SS| 6 k2 − l + 1. (SS ∪ SS−1) \ SS ⊆ x1{x−1 l+1, · · · , x−1 k } ∪ x2{x−1 l+1, · · · , x−1 k } ∪ · · · ∪ xl{x−1 l+1, · · · , x−1 k } ∪ xl+1{x−1 l+2, x −1 l+3, · · · , x−1 k } ∪ xl+2{x−1 l+1, x −1 l+3, x −1 l+4, · · · , x−1 k } ∪ · · · ∪ xk{x−1 l+1, x −1 l+2, · · · , x−1 k−1}. “adm-n2” — 2021/7/19 — 10:26 — page 172 — #8 172 Groups containing LMPFS of size 4 Therefore |(SS ∪ SS−1) \ SS| 6 l(k − l) + (k − l)(k − l − 1) = (k − l)(k − 1). (2) Similarly, |(SS ∪ S−1S) \ SS| 6 (k − l)(k − 1). By Equation (1) therefore |G| 6 3k2 − l(2k − 1) + 1. Remark 3.4. (a) The bound of |G| in Proposition 3.3 is tight as it is attained with (k, l) = (2, 2) as S = {x, x−1} ⊂ G = 〈x | x7 = 1〉 ∼= C7 meets it. (b) Theorem 3.2 and Proposition 3.3 point at the need for a bound on the sizes of finite groups of even order containing locally maximal product-free sets S satisfying S∩S−1 6= ∅. Such universal bound is hard to obtain (see Question 3.29 at the end of the paper) when G is an arbitrary finite group of even order due to the difficulty in bounding | √ S| as we saw an example of a locally maximal product-free subset S of Q8 satisfying | √ S| = 3 4 |Q8|; an interested reader may see [2, Proposition 2.1] for a family of groups G for which there exists a locally maximal product-free set T such that | √ T | is comparable to |G|. However, in the finite abelian case, progress is possible. In the light of Equation (1), a finite abelian group G containing a locally maximal product-free set S can be characterised as: G = S ∪ SS ∪ SS−1 ∪ √ S. (3) If |G| is odd, then as each element of G has exactly one square root, | √ S| = |S|. Using |SS| 6 |S|(|S|+ 1) 2 and |SS−1| 6 |S|2 − |S|+ 1 (4) in equation (3), we obtain that |G| 6 3|S|2 + 3|S|+ 2 2 . (5) On the other hand, if |G| is even, then using (4) together with | √ S| 6 |G| 2 yield |G| 6 3|S|2 + |S|+ 2. (6) We note here that | √ S| 6 |G| 2 follows from the fact that √ S is product- free in a finite abelian group G whenever S is product-free, and that a product-free set in G has size at most |G| 2 . “adm-n2” — 2021/7/19 — 10:26 — page 173 — #9 C. S. Anabanti 173 Observation 3.5. Let S be a set of size k in a finite abelian group G such that 1 6 |S ∩ S−1| = l 6 k. We define A(l) to be a non-negative integer which is less than or equal to the maximal number of identity 1’s in {x1x1, · · · , x1xk} ∪ {x2x2, · · · , x2xk} ∪ · · · ∪ {xk−1xk−1, xk−1xk} ∪ {xkxk}. So |SS| 6 k(k + 1) 2 −A(l) + 1. (7) Suppose |G| is even. Then A(l) := { l+1 2 if l is odd; l 2 if l is even. Whence |SS| 6 k(k + 1)− (l − 1) 2 or k(k + 1)− (l − 2) 2 (8) according to whether l is odd or even. Now, suppose |G| is odd. Each element of G has a unique inverse; so l > 2 and even. Thus, A(l) = l 2 , and we conclude from inequality (7) that |SS| 6 k(k + 1)− l + 2 2 . Inequality (7) can be used to obtain a better bound for |G| when S∩S−1 6= ∅. For instance, if S is a locally maximal product-free set of size k in a finite abelian group G of odd order such that 1 6 |S ∩ S−1| = l 6 k, then |G| 6 3k2−l(2k−1)+(3k+2) 2 , and this result improves inequality (5) for S ∩ S−1 6= ∅. Lemma 3.6 below gives a bound on |G| when S ∩ S−1 = ∅. Lemma 3.6. Suppose S is a locally maximal product-free set in a finite abelian group G such that S ∩ S−1 = ∅. Then G = SS ∪ SS−1 ∪ (SS)−1. Moreover, |G| 6 2|S|2 − |I(G)|+ 1. Also, if I(G) ⊆ I(SS−1), then |G| 6 2|S|2 − 2|I(G)|+ 1. Proof. As SS−1 = S−1S, the part “G = SS∪SS−1∪(SS)−1" follows from G = SS ∪ SS−1 ∪ S−1S ∪ (SS)−1 in Theorem 3.2. As |SS| = |(SS)−1| 6 |S|(|S|+1) 2 and |SS−1| 6 |S|2 − |S|+ 1, we conclude that |G| 6 2|S|2 + 1. However, each involution that gives rise to the bound on |G| is counted at least twice. Suppose x is an involution of G. Now, x ∈ SS if and only if x ∈ (SS)−1; so all such involutions are counted twice. If x ∈ SS−1, then x = yz−1 for some y, z ∈ S. Thus, x = zy−1 as well. So x is counted at least twice in SS−1. By removing the second count on involutions of G, “adm-n2” — 2021/7/19 — 10:26 — page 174 — #10 174 Groups containing LMPFS of size 4 we obtain that |G| 6 2|S|2 − |I(G)|+ 1. Now, suppose I(G) ⊆ I(SS−1). Let x ∈ SS−1 be an involution. Then there exist y, z ∈ S (for y 6= z) such that x = yz−1. Now, 1 = x2 = y2z−2; so y2 = z2. This shows that given each involution yz−1 ∈ SS−1, we can find two elements y2, z2 ∈ SS such that y2 = z2. Thus, we can find a pair of repeated elements (y2, z2) for y2, z2 ∈ SS such that y2 = z2. By discarding one element in each pair of the repeated elements of SS, we obtain that |G| 6 2|S|2−2|I(G)|+1. The first bound on |G| in Lemma 3.6 is tight as S = {x, y3, x3y} ⊂ G = 〈x, y | x4 = 1 = y4, xy = yx〉 ∼= C4 × C4 meets it. Remark 3.7. An observation of [3, p. 2] is that if a finite abelian group of order less than or equal to 52 contains a LMPFS S of size 6 or less, then | √ S| 6 2|S|. This observation, together with Equation (3), Inequalities (2), (5) and (8) and Lemma 3.6 imply that if a finite abelian group G contains a locally maximal product-free set of size 4, then |G| 6 32. The locally maximal product-free sets of size 4 in abelian groups of order up to 32 were checked in GAP [8]. We also used the SmallGroup library in [8] to restrict our search to the abelian groups. If at least two LMPFS of size 4 are found in a certain abelian group G, we check whether there is an automorphism of G that takes one to another, and if there is, we display only one such set. In other words, we display only one locally maximal product-free set in each orbit of the action of the automorphism groups of G. Our computational results are summarised in Table 1 below. For notations in Table 1, n4 is the number of locally maximal product-free sets of size 4 in G while M4 shows the corresponding sizes of each orbit of the displayed locally maximal product-free sets under the action of automorphism groups of G. We take this opportunity to correct a mistake of [11, Table 1], where the authors mentioned that every locally maximal product-free set of size 4 in C14 is mapped by an automorphism of C14 to either S1 = {x2, x5, x9, x13} or S2 = {x4, x5, x6, x7}. This is not true as Table 1 shows that there are five LMPFS up to automorphisms of C14. In particular, as Aut(Cn) is isomorphic to the unit group C× n whose order is φ(n) (where φ(n) is Euler’s totient function), the automorphisms of C14 are maps Φi : x 7→ xi for x ∈ C14 and i ∈ {1, 3, 5, 9, 11, 13}. Therefore, the LMPFS {x, x3, x8, x13} and {x, x4, x7, x12} are mapped by Φ9 and Φ5 respectively into S1 and S2, and the other three product-free sets in Table 1 (viz. {x, x3, x8, x10}, {x, x4, x6, x13} and {x, x6, x8, x13}) are mapped to neither S1 nor S2. “adm-n2” — 2021/7/19 — 10:26 — page 175 — #11 C. S. Anabanti 175 Table 1. Finite abelian groups containing LMPFS of size 4. G n4 LMPFS S of size 4 in G M4 C8 = 〈x | x8 = 1〉 1 {x, x3, x5, x7} 1 C4 × C2 = 〈x1, x2 | x4 1 = 1 = x2 2, x1x2 = x2x1〉 3 {x1, x 3 1, x2, x 2 1x2}, {x1, x 3 1, x1x2, x 3 1x2} 2, 1 C3 2 = 〈x1, x2, x3 | x2 i = 1, xixj = xjxi for 1 6 i, j 6 3〉 7 {x1, x2, x3, x1x2x3} 7 C10 = 〈x | x10 = 1〉 2 {x, x4, x6, x9} 2 C11 = 〈x | x11 = 1〉 5 {x, x3, x8, x10} 5 C12 = 〈x | x12 = 1〉 9 {x, x4, x6, x11}, {x, x4, x7, x10}, {x2, x3, x8, x9}, {x2, x3, x9, x10} 4, 2, 2, 1 C6 × C2 = 〈x1, x2 | x6 1 = 1 = x2 2, x1x2 = x2x1〉 11 {x1, x 5 1, x2, x 3 1x2}, {x1, x2, x 4 1, x 3 1x2}, {x1, x1x2, x 4 1, x 4 1x2} 3, 6, 2 C13 = 〈x | x13 = 1〉 21 {x, x3, x5, x12}, {x, x3, x10, x12}, {x, x5, x8, x12} 12, 6, 3 C14 = 〈x | x14 = 1〉 27 {x, x3, x8, x10}, {x, x3, x8, x13}, {x, x4, x6, x13}, {x, x4, x7, x12}, {x, x6, x8, x13} 6, 6, 6, 6, 3 C15 = 〈x | x15 = 1〉 16 {x, x3, x5, x7}, {x, x3, x7, x12} 8, 8 C16 = 〈x | x16 = 1〉 37 {x, x3, x10, x12}, {x, x4, x6, x9}, {x, x4, x6, x15}, {x, x4, x9, x14}, {x, x6, x9, x14}, {x, x6, x10, x14}, {x2, x6, x10, x14} 8, 4, 8, 4, 4, 8, 1 C4 × C4 = 〈x1, x2 | x4 1 = 1 = x4 2, x1x2 = x2x1〉 6 {x1, x 3 1, x 2 2, x 2 1x 2 2} 6 “adm-n2” — 2021/7/19 — 10:26 — page 176 — #12 176 Groups containing LMPFS of size 4 Table 1, continued G n4 LMPFS S of size 4 in G M4 C8 × C2 = 〈x1, x2 | x8 1 = 1 = x2 2, x1x2 = x2x1〉 42 {x1, x2, x 6 1, x 5 1x2}, {x1, x2, x 6 1, x 4 1x2}, {x1, x2, x 2 1x2, x 5 1x2}, {x1, x1x2, x 6 1, x 6 1x2}, {x1, x 6 1, x 2 1x2, x 6 1x2}, {x2, x 2 1, x 6 1, x 4 1x2}, {x2 1, x 6 1, x 2 1x2, x 6 1x2} 8, 8, 8, 8, 8, 1, 1 C4 × C2 × C2 = 〈x1, x2, x3 | x4 1 = 1 = x2 2, x 2 3 = 1, xixj = xjxi for 1 6 i, j 6 3〉 4 {x2, x3, x 2 1, x 2 1x2x3} 4 C17 = 〈x | x17 = 1〉 48 {x, x3, x8, x13}, {x, x3, x8, x14}, {x, x3, x11, x13} 16, 16, 16 C18 = 〈x | x18 = 1〉 54 {x, x3, x5, x12}, {x, x3, x8, x14}, {x, x3, x9, x14}, {x, x3, x12, x14}, {x, x4, x9, x16}, {x, x4, x10, x17}, {x, x5, x8, x12}, {x, x5, x8, x17}, {x, x6, x9, x16} 6, 6, 6, 6, 6, 6, 6, 6, 6 C6 × C3 = 〈x1, x2 | x6 1 = 1 = x3 2, x1x2 = x2x1〉 48 {x1, x 5 1, x2, x 3 1x2}, {x1, x2, x 5 1x 2 2, x 3 1} 24, 24 C19 = 〈x | x19 = 1〉 36 {x, x3, x5, x13}, {x, x4, x6, x9} 18, 18 C10×C2 = 〈x1, x2 | x10 1 = 1 = x2 2, x1x2 = x2x1〉 28 {x1, x2, x 5 1x2, x 8 1}, {x1, x1x2, x 4 1, x 8 1x2}, {x1, x1x2, x 8 1, x 6 1x2} 12, 12, 4 C20 = 〈x | x20 = 1〉 36 {x, x3, x10, x16}, {x, x3, x14, x16}, {x, x4, x11, x18}, {x, x5, x14, x18}, {x, x6, x8, x11}, {x2, x5, x15, x16} 8, 8, 4, 8, 4, 4 “adm-n2” — 2021/7/19 — 10:26 — page 177 — #13 C. S. Anabanti 177 Table 1, continued G n4 LMPFS S of size 4 in G M4 C21 = 〈x | x21 = 1〉 34 {x, x3, x5, x15}, {x, x4, x10, x17}, {x, x4, x14, x16}, {x, x8, x12, x18} 12, 12, 4, 6 C22 = 〈x | x22 = 1〉 10 {x, x4, x10, x17} 10 C24 = 〈x | x24 = 1〉 4 {x, x6, x17, x21} 4 C9 × C3 = 〈x1, x2 | x9 1 = 1 = x2 3, x1x2 = x2x1〉 36 {x1, x1x2, x 3 1, x 7 1x 2 2}, {x1, x1x2, x 6 1, x 7 1x 2 2} 18, 18 C3 3 = 〈x1, x2, x3 | x3 i = 1, xixj = xjxi for 1 6 i, j 6 3〉 468 {x1, x2, x3, x 2 1x 2 2x 2 3} 468 Suppose a finite nonabelian group G contains a locally maximal product-free set S of size 4. If |G| is odd, then Proposition 3.3 tells us that |G| 6 49. Now, suppose |G| is even. If S ∩ S−1 = ∅, then The- orem 3.2 tells us that |G| 6 56. So the only case left is to bound the size of a finite group G of even order which contains a locally maximal product-free set of size 4 such that |S ∩ S−1| > 1. We use GAP [8] to check all locally maximal product-free sets of size 4 in nonabelian groups of order at most 56. The result shows that 45 nonabelian groups contains locally maximal product-free sets of size 4, and over 15% of them are dihedral groups. More importantly, the largest size of such group is 40. We shall study a special case for the generating set of the locally maximal product-free sets as we aim to prove the following: Theorem 3.8. If G is a finite group containing a locally maximal product- free set of size 4 such that every 2-element subset of S generates 〈S〉, then |G| 6 40. We develop preliminary results that we shall put together to prove Theorem 3.8. In particular, we aim to prove Theorem 3.8 by considering each of the following three cases: (a) S contains at least two involutions; (b) S contains no involution; (c) S contains only one involution. Before we proceed, we state the following result (Lemma 3.9 below) for a finite group G which we shall employ whenever necessary, without necessarily quoting the result. Lemma 3.9. If S is a LMPFS in a group G, then S is locally maximal product-free in 〈S〉. “adm-n2” — 2021/7/19 — 10:26 — page 178 — #14 178 Groups containing LMPFS of size 4 Proof. Suppose S is a LMPFS in a finite group G. To show that S is locally maximal product-free in 〈S〉 suffices to show that S is product-free in 〈S〉 and T (S) ∪ {g ∈ 〈S〉 : g2 ∈ S} = 〈S〉. The first is clear since 〈S〉 ⊆ G and S is product-free in G. For the latter, we first recall that T (S) ⊆ 〈S〉. Let g ∈ 〈S〉 \ T (S) be arbitrary. As g ∈ G and S is locally maximal product-free in G, we have that g2 ∈ S. Therefore for all g ∈ 〈S〉, either g ∈ T (S) or g2 ∈ S; whence S is locally maximal product-free in 〈S〉. As a consequence to Theorem 2.11, we give the following result: Corollary 3.10. No finite group contains a LMPFS S of size 4 such that every two element subset of S generates 〈S〉, and S contains at least two involutions. Proof. Suppose a finite group G contains a LMPFS S of size 4 such that every two element subset of S generates 〈S〉, and S contains at least two involutions. Then 〈S〉 is dihedral. In the light of Lemma 3.9 and Theorem 2.11, 〈S〉 is one of D8, D10, D12, D14, D16, D18 or D20. Suppose 〈S〉 = D8. Then 〈S〉 = 〈y, x2y〉 ∼= C2 × C2; a contradiction. Suppose 〈S〉 is one of D10, D14, D16, D18 or D20. Then S contains two rotations; so the group generated by S is also the group generated by such two rotations, which is abelian (in particular, not dihedral); a contradiction. Finally, suppose 〈S〉 = D12. If S = {x3, y, xy, x2y}, then 〈S〉 = 〈x3, y〉 ∼= C2 × C2; a contradiction. If S is any of the other three locally maximal product-free sets in D12, then the group generated by any two rotations in such S is not dihedral; a contradiction. Therefore, no such G (respectively S) exists. Before we proceed, we give the following result. Table 2. Nonabelian groups (of order up to 40) that contain a LMPFS of size 4. G n4 LMPFS S of size 4 in G M4 D8 = 〈x, y | x4 = 1 = y2, xy = yx−1〉 3 {y, xy, x2y, x3y}, {x, x3, y, x2y} 1, 2 Q8 = 〈x, y | x4 = 1, x2 = y2, xy = yx−1〉 3 {x, x3, y, x2y} 3 D10 = 〈x, y | x5 = 1 = y2, xy = yx−1〉 10 {x2, x3, y, x4y} 10 Q12 = 〈x, y | x6 = 1, x3 = y2, xy = yx−1〉 9 {x, x5, y, x3y}, {x, x4, y, x3y} 3, 6 “adm-n2” — 2021/7/19 — 10:26 — page 179 — #15 C. S. Anabanti 179 Table 2, continued G n4 LMPFS S of size 4 in G M4 A4 = 〈x, y | x3 = y2 = (xy)3 = 1〉 2 {x, yx, x2yx, xy} 2 D12 = 〈x, y | x6 = 1 = y2, xy = yx−1〉 27 {x3, y, xy, x5y}, {x2, x3, y, x5y}, {x, x5, y, x3y}, {x, x4, y, x3y} 6, 12, 3, 6 D14 = 〈x, y | x7 = 1 = y2, xy = yx−1〉 42 {x2, x3, y, x6y} 42 (C4 × C2)⋊α C2 = 〈x, y | x4 = y2 = (xyx)2 = (yx−1)4 = (yxyx−1)2 = 1〉 2 {x2, y, (xy)2, x3yx} 2 C4 ⋊ C4 = 〈x, y | x4 = y4 = x−1yxy = x2y−1x2y = (y−1x−2y−1)2 = 1〉 2 {x2, y, x3yx, x2y2} 2 M16 = C8 ⋊ C2 = 〈x, y | x8 = 1 = y2, xy = yx5〉 26 {x, x4, x7, y}, {x, x6, y, x4y}, {x2, x6, y, x4y}, {x, x6, x2y, x6y}, {x2, x6, x2y, x6y} 8, 8, 1, 8, 1 D16 = 〈x, y | x8 = y2 = (xy)2 = 1〉 48 {x2, x3, y, x7y}, {x, x6, y, x4y} 32, 16 QD16 = 〈x, y | x8 = 1 = y2, xy = yx3〉 16 {x, x6, y, x4y}, {x, x6, x3y, x7y} 8, 8 Q16 = 〈x, y | x8 = 1, x4 = y2, xy = yx−1〉 16 {x, x6, y, x4y} 16 D8 ∗ C4 = 〈x, y, z | x2 = y2 = z4 = 1 = z−1xzx = z−1yzy = yz2xyx〉 8 {x, y, xyz, (xy)2} 8 D18 = 〈x, y | x9 = 1 = y2, xy = yx−1〉 54 {x2, x5, y, x8y} 54 C3 × S3 = 〈x, y, z | x2 = y3 = z3 = (xz)2 = y−1xyx = z−1y−1zy = 1〉 72 {x, y, xz, yzx}, {x, y, z, xyz}, {x, y, z, yzx}, {x, y, z, xy2zx}, {x, y, xyz, yzx}, {x, z, xyz, xy2z}, {x, z, yzx, y2zx}, {y, xy, z, xy2zx} 6, 12, 12, 12, 6, 6, 6, 12 (C3 × C3)⋊ C2 = 〈x, y, z | x2 = y3 = z3 = (xy)2 = (xz)2 = z−1y−1zy = 1〉 144 {y, x, xz, xyxzx}, {y, x, xz, yzx} 72, 72 “adm-n2” — 2021/7/19 — 10:26 — page 180 — #16 180 Groups containing LMPFS of size 4 Table 2, continued G n4 LMPFS S of size 4 in G M4 Q20 = 〈x, y | x10 = 1, x5 = y2, xy = yx−1〉 20 {x, x8, y, x5y} 20 Suz(2) = 〈x, y | x5 = 1 = y4, xy = yx2〉 (the only non-simple Suzuki group). 40 {x, y2, x2y, xy3}, {x3, y2, x2y, xy3} 20, 20 D20 = 〈x, y | x10 = 1 = y2, xy = yx−1〉 20 {x, x8, y, x5y} 20 C7 ⋊ C3 = 〈x, y | x3 = y7 = y−1x−1yxy−1 = 1〉 280 {x, y, (xy)2x, xyx}, {x, y, (xy)2x, x2yx2}, {x, y, (yx)2, x(xy)2}, {x, y, xyx, x(xy)2}, {x, y, x2yx2, (xy)2}, {x, y, yx2y, (xy)2}, {x, x2y, x(xy)2x, yx}, {x, x2y, xyx, yx2}, {x2, y, (xy)2x, xyx} 42, 21, 42, 42, 42, 42, 14, 14, 21 C3 ⋊ C8 = 〈a, x | x8 = a3 = x−1axa = 1〉 39 {x, a, x6, x4a}, {x, xa, x6, x4ax}, {a, x2, x6, x4a}, {x2, x6, x4a, x3ax} 24, 12, 2, 1 SL(2, 3) = 〈x, y | x3 = y4 = 1 = y−1xyxy−1x = x−1y−1(x−1y)2 = (x−1y−1)3〉 72 {x, x2y, y3, yx2}, {x, x2y, y3, x(xy)2y}, {x, y, x2y3, x2yx}, {x2y, y, x2yx, xy2} 24, 24, 12, 12 Q24 = 〈x, y | x12 = 1, x6 = y2, xy = yx−1〉 4 {x, x6, x8, x11} 4 Q12 × C2 = 〈x, y, a | x4 = y2 = a3 = x−1axa = yx−1yx = a−1yay = 1〉 6 {y, a, x2, x2ya}, {y, x2, x2ya, xyax} 4, 2 B(2, 3) = 〈x, y | x3 = y3 = 1 = (x−1y−1)3 = (y−1x)3〉 252 {y, x, (xy)2, x2y2xy}, {y, x, (xy)2y, x2y2x} 144, 108 C9 ⋊ C3 = 〈x, y | x9 = 1 = y3, xy = yx4〉 144 {x6, x8, y, x4y2}, {x3, y2, x4y2, x8y2}, {x, x3, xy, x4y2}, {x, x6, xy, x4y2} 54, 54, 18, 18 C7 ⋊ C4 = 〈x, z | x4 = z7 = x−1zxz = (x−1z)2x−2 = 1〉 6 {z, x2, x3z2x, x2z3} 6 (C4 × C2)⋊ C4 = 〈x, y | x4 = y4 = y2x−1y−2x = xyx2y−1x = y2x−1y−2x = (yx)2(y−1x)2 = (yxy−1x−1)2 = (xyx−1y)2 = 1〉 1 {x2, y2, (xy)2, x2(xy)2y2} 1 “adm-n2” — 2021/7/19 — 10:26 — page 181 — #17 C. S. Anabanti 181 Table 2, continued G n4 LMPFS S of size 4 in G M4 C4 ⋊ C8 = 〈x, y | x8 = y4 = x−1yxy = y−1x−1y2xy−1 = 1〉 1 {x6, x2, x4y2, y2} 1 C8 ⋊Quasidih type C4 = 〈x, y| x8 = 1 = y4, xy = yx3〉 9 {x, x6, y2, x4y2}, {x2, x6, y2, x4y2} 8, 1 C8 ⋊Dih type C4 = 〈x, y | x8 = 1 = y4, xy = yx−1〉 17 {x, x6, y2, x5y2}, {x, x6, y2, x4y2}, {x, y2, x2y2, x5y2}, {x2, x6, y2, x4y2} 4, 8, 4, 1 C4 ⋊D8 = 〈x, y | xy−2x−1y−2 = x(xy)2x = x2(xy−1)2 = y−1x−1y3x = 1〉 9 {y, x2, x4y2, x6}, {x5yx, x2, x4y2, x6}, {x2, y2, x4y2, x6} 4, 4, 1 (C4 ⋊ C4)× C2 = 〈x, y, z | z2 = y4 = x4 = x−1yxy = zx−1zx = 1 = zy−1zy = x2y−1x2y = (y−1x−2y−1)2〉 2 {x2y2z, z, x2, y2} 2 C4 ×Q8 = 〈x, y, z | x4 = y4 = z4 = 1 = x−1yxy = z−1x−1zx = z−1y−1zy = y2zx−2z = x−1zx−1y−2z = x2y−1x2y〉 2 {x2, y2, z, x2y2z} 2 (C2 ×Q8)⋊ C2 = 〈x, y, z | x4 = z2 = xyxy−1 = y2x2 = yxyx−1 = zy−1zy = (xzx)2 = (zx−1)4 = (zxzx−1)2 = 1〉 2 {x2, z, (xz)2, x3zx} 2 C2 2 ⋊ C3 2 = 〈x, y, z | x4 = y4 = z2x2 = y−1x−1yx−1 = z−1y−1zy = y2x−1y2x = zxy2zx = (x−1y−2x−1)2 = 1〉 2 {x2, y2, yz, x2y3z} 2 C4 ⋊Q8 = 〈x, y, z | y4 = z4 = x2y2 = yxyx−1 = x−1zxz = z−1y−1zy = 1〉 4 {x2, z, x3zx, x2z2} 4 C2 × C2 ×Q8 = 〈x, y, z, a | x4 = a2 = z2 = y2x2 = yxyx−1 = ax−1ax = ay−1ay = zx−1zx = (az)2 = zy−1zy = 1〉 4 {x2, z, a, x2za} 4 C9 ⋊ C4 = 〈x, z | x4 = z9 = x−1zxz = 1〉 18 {x2, z, x2z2, xz3x}, {x2, z, x3z2x, x2z4}, {x2, z, x3z3x, x2z4} 6, 6, 6 (C3 × C3)⋊ C4 = 〈x, z, a | x4 = z3 = a3 = x−1zxz = x−1axa = a−1z−1az = (x−1z)2x−2 = z−1ax−1az−1x = 1〉 24 {z, x2, x2a, xzax} 24 “adm-n2” — 2021/7/19 — 10:26 — page 182 — #18 182 Groups containing LMPFS of size 4 Table 2, continued G n4 LMPFS S of size 4 in G M4 C3 ×A4 = 〈x, y, z | x3 = y3 = z2 = 1 = y−1x−1yx = zy−1zy = (zx)3 = (x−1z)3〉 144 {x2, yzx, x2y2, x2zx2} 144 C5 ⋊inverse map C8 = 〈x, y | x8 = 1 = y5, xyx−1 = y−1〉 4 {x2, x−2, y, x4y2} 4 Q40 = 〈x, y | x20 = 1, x10 = y2, xy = yx−1〉 8 {x10, x11, x12, x17} 8 Q20 × C2 = 〈x, y, z | x10 = 1 = z2, x5 = y2, xy = yx−1, yz = zy, xz = zx〉 8 {x−2, y2, z, xz} 8 For notation in Table 2, n4 is the number of locally maximal product- free sets of size 4 in G while M4 shows the corresponding sizes of each orbit of the displayed locally maximal product-free sets under the action of automorphism groups of G. Theorem 3.11. Let S be a locally maximal product-free set of size 4 in a finite group G such that |G| 6 57. Then the possibilities for S and G are given in Tables 1 and 2. Proof. We checked for groups of order from 8 up to 57 that contains locally maximal product-free sets of size 4 in GAP [8]. Then listed all such locally maximal product-free sets S of size 4 up to automorphisms of each such group G in Tables 1 and 2. Corollary 3.12. If S is a LMPFS of size 4 in a finite group G of odd order, then both S and G are contained in Tables 1 and 2. Proof. Follows from Proposition 3.3 and Theorem 3.11. Lemma 3.13. Let S be a LMPFS of size 4 in a finite group G such that every 2-element subset of S generates 〈S〉. If S contains no involution, then either |G| 6 40 or 〈S〉 is cyclic. Proof. If S∩S−1 = ∅, then by Theorem 3.2, |G| 6 57. Theorem 3.11 tells us that (G,S) is one of the possibilities in Tables 1 and 2. In particular, |G| 6 40. Suppose S ∩ S−1 6= ∅. Then S contains two elements a and b such that b = a−1. As every 2-element subset of S generates 〈S〉, we have that 〈S〉 = 〈a, a−1〉 = 〈a〉. So 〈S〉 is cyclic. Proposition 3.14. Suppose S is a LMPFS of size 4 in a group G. If 〈S〉 is cyclic, then |G| 6 40. “adm-n2” — 2021/7/19 — 10:26 — page 183 — #19 C. S. Anabanti 183 Proof. As 〈S〉 is cyclic, in the light of Lemma 3.9 and Remark 3.7, |〈S〉| 6 24. Table 1 shows various possibilities for G and S. Proposition 2.6 tells us that each element s of Ŝ has even order; whence if 〈S〉 is any of the cyclic groups of odd order, then Ŝ = ∅ and we conclude that G = 〈S〉. In the light of Table 1 therefore |G| 6 21 < 40. Suppose 〈S〉 is cyclic of even order. Consider 〈S〉 = C8 = 〈x | x8 = 1〉 and S = {x, x3, x5, x7}. If any of x, x3, x5 or x7 is contained in Ŝ, then Ŝ consists of power of a single element; by Proposition 2.7, |G| divides 32. If none of x, x3, x5 or x7 is con- tained in Ŝ, then Ŝ = ∅; so G = 〈S〉. Consider 〈S〉 = C10 = 〈x | x10 = 1〉 and S = {x, x4, x6, x9}. As x4 and x6 have odd order, by Proposition 2.6, x4, x6 6∈ Ŝ. Clearly, x, x9 6∈ Ŝ since x3, (x9)3 6∈ S. Therefore Ŝ = ∅; so G = 〈S〉. Consider 〈S〉 = C12 = 〈x | x12 = 1〉. By Table 1, there are four such LMPFS up to automorphisms of C12. By Proposition 2.6, x4 6∈ Ŝ because it has odd order. Suppose S = {x, x4, x6, x11}. By Proposition 2.6, x, x11 6∈ Ŝ because x3, (x11)3 6∈ S. So Ŝ 6 1 and we conclude by Propo- sition 3.6 that |G| 6 24. If S = {x, x4, x7, x10}, then by Proposition 2.6, Ŝ = ∅ since (x)3, (x7)3, (x10)3 6∈ S; so G = 〈S〉. Now, suppose S is any of {x2, x3, x8, x9} or {x2, x3, x9, x10}. In the light of Proposition 2.6, in the first case, x2, x8 6∈ Ŝ, and in the latter case, x2, x10 6∈ Ŝ. If none of x3 or x9 is an element of Ŝ, then Ŝ = ∅, and we conclude that G = 〈S〉. If any of x3 or x9 is contained in Ŝ, then both are contained in Ŝ; by Proposition 2.7 therefore |G| divides 48. Suppose |G| = 48. As √ S has only elements of or- der at least 3, we note that the number of elements of order at least 3 in G is 46. Among all the 47 nonabelian groups of order 48, only the groups whose GAP ID are [48, 1], [48, 8], [48, 18], [48, 27] and [48, 28] have 46 elements of order at least 3. We checked each of them for a LMPFS of size 4, and could not find such. Therefore, if S is any of {x2, x3, x8, x9} or {x2, x3, x9, x10}, then |G| 6 24. Now, consider 〈S〉 = C14 = 〈x | x14 = 1〉. Up to au- tomorphisms of C14, the LMPFS of size 4 in C14 are {x, x3, x8, x10}, {x, x3, x8, x13}, {x, x4, x6, x13}, {x, x6, x8, x13} and {x, x4, x7, x12}. In the light of Proposition 2.6, all elements of order 7 and 14 in the respective sets S do not lie in Ŝ. This means that in the first four cases, G = 〈S〉. For the latter case, only the involution is a possible element of Ŝ; thus, |Ŝ| 6 1, and we conclude by Proposition 2.5 that |G| 6 28. Consider 〈S〉 = C16 = 〈x | x16 = 1〉. Up to automorphisms of C16, the LMPFS S of size 4 are {x, x3, x10, x12}, {x, x4, x6, x9}, {x, x4, x6, x15}, {x, x4, x9, x14}, {x, x6, x9, x14}, {x, x6, x10, x14} and {x2, x6, x10, x14}. For the first six cases, Ŝ = ∅; so G = 〈S〉. For the last case, S = S−1; so 〈S〉 ∼= C8; a contradiction as 〈S〉 ∼= C16. Consider 〈S〉 = C18 = 〈x | x18 = 1〉. By Table 1, there are 9 such LMPFS up to automorphisms of C18. In the “adm-n2” — 2021/7/19 — 10:26 — page 184 — #20 184 Groups containing LMPFS of size 4 light of Proposition 2.6, any of the 9 locally maximal product-free sets S which does not contain the unique involution x9 gives rise to Ŝ = ∅; so G = 〈S〉. For the LMPFS which contains the unique involution, we have that |Ŝ| 6 1; whence by Proposition 2.5 therefore |G| 6 36. Consider 〈S〉 = C20 = 〈x | x20 = 1〉. Here, there are six such LMPFS up to automor- phisms of C20. They are {x, x3, x14, x16}, {x, x4, x11, x18}, {x, x5, x14, x18}, {x, x6, x8, x11}, {x, x3, x10, x16} and {x2, x5, x15, x16}. The first four cases can be handled with Proposition 2.6 to give that Ŝ = ∅; so G = 〈S〉. For S = {x, x3, x10, x16}, we have that |Ŝ| 6 1; so |G| 6 40. Finally, let S = {x2, x5, x15, x16}. By Proposition 2.6, the only possible element of Ŝ are x5 and x15. If none of them are in Ŝ, then Ŝ = ∅ and G = 〈S〉. Suppose at least one of them is in Ŝ, then as all odd powers of such element lies in S, both elements must belong to Ŝ, and Proposition 2.7 tells us that |G| divides 80. Suppose |G| = 80. As √ S has only elements of order at least 4, we note that the number of elements of order at least 4 in G is 78. Among all the 47 nonabelian groups of order 80, only the groups whose GAP ID are [80, 1], [80, 3], [80, 8], [80, 18] and [80, 27] contain 78 elements of order at least 4. We checked each of them for a LMPFS of size 4, and couldn’t find such. Therefore, if S = {x2, x5, x15, x16}, then |G| 6 40. For 〈S〉 = C22 or C24, there is only one such LMPFS up to automorphisms of the respective groups and a direct check using Proposition 2.6 tells us that Ŝ = ∅; so G = 〈S〉. Corollary 3.15. If S is a LMPFS of size 4 in a finite group G such that every two element subset of S generates 〈S〉 and S contains no involution, then |G| 6 40. Proof. Follows from Lemma 3.13 and Proposition 3.14. Proposition 3.16. Suppose S is a LMPFS of size 4 in a finite group G such that every two element subset of S generates 〈S〉 and S contains only one involution. Then either |G| 6 40 or S = {a, b, c, d}, where c is the unique involution in S and either a, b and d have order 3, or that a has order greater than 3 together with a−1 = bd and none of b and d is an involution. Proof. Suppose S = {a, b, c, d}, where c is an involution, and each of a, b and d has order at least 3. Consider a−1. Recall that G = T (S) ∪ √ S. Suppose a−1 ∈ √ S. Then a−2 ∈ S. This implies that either a has order 3 (by a−2 = a) or 〈S〉 is cyclic ( by a−2 ∈ {b, c, d}; for instance a−2 = b “adm-n2” — 2021/7/19 — 10:26 — page 185 — #21 C. S. Anabanti 185 implies that 〈a, b〉 = 〈a〉 ) . In the latter case, Proposition 3.14 tells us that |G| 6 40. Suppose a−1 ∈ T (S). Then T (S) ⊆      1, a, b, c, d, a2, b2, d2, ab, ba, ac, ca, ad, da, bc, cb, bd, db, cd, dc, ab−1, ba−1, ca−1, a−1c, cb−1, b−1c, a−1b, b−1a, ad−1, d−1a, da−1, a−1d, bd−1, d−1b, cd−1, d−1c, db−1, b−1d      (9) If a−1 ∈ {b, b2, d, d2, ab, ba, ad, da, ab−1, ba−1, ad−1, da−1, a−1b, b−1a, a−1d, d−1a}, then 〈S〉 is cyclic, generated by either a, b or d. In the light of Proposition 3.14 therefore |G| 6 40. If a−1 ∈ {c, ac, ca, a−1c, ca−1}, then 〈S〉 is cyclic; again |G| 6 40. Since a has order at least 3, we have that a−1 6∈ {1, a}. Note that a−1 6∈ {bc, cb, dc, cd, b−1c, cb−1, d−1c, cd−1, b−1d, db−1, d−1b, bd−1}; otherwise S is not product-free. The only remaining possibilities is that a−1 ∈ {a2, bd, db}. We also perform a similiar analysis with b−1 and d−1. Our conclusion is that either 〈S〉 is cyclic (which by Proposition 3.14 implies that |G| 6 40) or either a has order 3 with either a−1 = bd or a−1 = db with similar statement for b and d. In the latter case, we can assume without loss of generality that either all of a, b and d have order 3, or that a has order greater than 3 together with a−1 = bd and none of b and d is an involution. In the latter part of Proposition 3.16, our goal is to show that |G| 6 40. We resolve the first case of the latter part of Proposition 3.16 in Lemma 3.20 below, and the second case in Corollary 3.24, Lemma 3.25 and Remark 3.26. We will be considering the following special case: S is a LMPFS of size 4 in a group G such that every two-element subset of S generates 〈S〉. Furthermore S = {a, b, c, d} where c is an involution but none of a, b and d is an involution. We shall impose an additional condition that ‘〈S〉 is neither abelian nor dihedral’ (see Assumption 3.18 below). To do so, we first clear the air with the following: Lemma 3.17. Let S be a LMPFS of size 4 in a finite group G such that S contains exactly one involution and every two element subset of S generates 〈S〉. Suppose 〈S〉 is either abelian or dihedral. Then 〈S〉 must be abelian and |G| 6 40. “adm-n2” — 2021/7/19 — 10:26 — page 186 — #22 186 Groups containing LMPFS of size 4 Proof. In the light of Lemmas 3.9 and 2.10, 〈S〉 cannot be dihedral. So, it must be that 〈S〉 is abelian. If 〈S〉 is cyclic, then Proposition 3.14 tells us that |G| 6 40. Now, suppose 〈S〉 is a non-cyclic abelian group. By Remark 3.7 and Table 1, 〈S〉 is one of C4 × C2, C 3 2 , C6 × C2, C4 × C4, C8 × C2, C4×C2×C2, C6×C3 and C10×C2. The LMPFS in the groups C4×C2, C 3 2 , C6×C2, C4×C4, C4×C2×C2 and C10×C2 do not meet the requirement of our defined S in terms of the orders of its elements. In fact, the only possibilities (up to automorphisms of respective group) that satisfy the condition that S has only one element of order 2 and other elements have order at least 3 is that (〈S〉, S) ∈ {(C8 × C2, {x1, x2, x61, x51x2}), (C8 × C2, {x1, x2, x21x2, x51x2}), (C6 × C3, {x1, x2, x51x22, x31})}. Proposition 2.6 tells us that elements of Ŝ have even order, and if s ∈ Ŝ, then all odd pow- ers of s lies in S. In the listed representatives of S, we see immediately that if a non-involution x ∈ S ⊂ C8 ×C2, then x3 /∈ S, and if a non-involution y ∈ S ⊂ C6×C3, then y5 /∈ S. So in all cases |Ŝ| 6 1. In the light of Propo- sition 2.5 therefore |G| 6 2|S| in each of the possibilities, from where we de- duce that |G| 6 32 or 36 according as 〈S〉 = C8×C2 or 〈S〉 = C6×C3. How- ever, the only possibility is ( 〈S〉, S, |G| ) = ( C8×C2, {x1, x2, x51x22, x31}, 32 ) since 〈S〉 is cyclic in each of the other two cases; for instance if G = C6×C3, then 〈S〉 = 〈x1, x31〉 = 〈x1 | x61 = 1〉 ∼= C6. Assumption 3.18. Suppose S is a locally maximal product-free set of size 4 in a finite group G such that every two-element subset of S generates 〈S〉. Furthermore S = {a, b, c, d} where c is an involution but none of a, b or d is an involution, and 〈S〉 is neither abelian nor dihedral. Lemma 3.19. Suppose Assumption 3.18 holds. Then I(G) ⊆ {c, a2, b2, d2, ab, ba, ad, da, bd, db, ab−1, b−1a, ad−1, d−1a, bd−1, d−1b}. Proof. The set I(G) consisting of all involutions in G is a subset of T (S), because G = T (S) ∪ √ S and no element of √ S can be an involu- tion. Clearly 1, a, b, d are not involutions. Suppose x ∈ {a±1, b±1, d±1}, and cx is an involution. Now 〈S〉 = 〈c, x〉 because 〈S〉 is generated by every 2-element subset of S. But 〈c, x〉 = 〈c, cx〉, which is dihe- dral; a contradiction. The case where xc is an involution is the same. So we can eliminate ac, bc, dc, a−1c, b−1c, d−1c, ca, cb, cd, ca−1, cb−1 and cd−1. This means I(G) ⊆ {c, a2, b2, d2, ab, ba, ad, da, bd, db, ab−1, ba−1, a−1b, b−1a, ad−1, da−1, a−1d, d−1a, bd−1, db−1, b−1d, d−1b}. If an element “adm-n2” — 2021/7/19 — 10:26 — page 187 — #23 C. S. Anabanti 187 g is an involution, then g−1 = g, so we only need to include one repre- sentative from {g, g−1} in the list of possible involutions. This means we need only include one of ab−1 and ba−1, for example. There are six such pairs, allowing us to remove ba−1, a−1b, da−1, a−1d, db−1 and b−1d from the list as if they are involutions then they will equal an element that is on the list. Thus I(G) ⊆ {c, a2, b2, d2, ab, ba, ad, da, bd, db, ab−1, b−1a, ad−1, d−1a, bd−1, d−1b}. Lemma 3.20. Suppose Assumption 3.18 holds. If a, b and d all have order 3, then 〈S〉 ∼= A4 and |G| 6 24. Proof. By Lemma 3.19, and the fact that a2, b2 and d2 are not involutions, we have I(G) ⊆ {c, ab, ba, ad, da, bd, db, ab−1, b−1a, ad−1, d−1a, bd−1, d−1b}. If c is the only involution in G, then c is central, and hence commutes with a. But 〈S〉 = 〈a, c〉, which implies 〈S〉 is abelian, contrary to Assumption 3.18. Therefore there are elements x, y ∈ {a±1, b±1, d±1} with y 6= x±1, such that xy is an involution. This implies 〈S〉 = 〈x, y : x3 = y3 = (xy)2 = 1〉, which is a presentation of A4. By Proposition 2.6, any element of Ŝ must have even order; so Ŝ ⊆ {c}. In the light of Proposition 2.5 therefore |G| 6 24. Lemma 3.21. Suppose S is a LMPFS of size 4 in a finite group G such that every two-element subset of S generates 〈S〉. Furthermore suppose S = {a, b, c, d} where c is an involution, a−1 = bd has order at least 4 and none of b or d is an involution. Then G has either 1, 3, 5, 7 or 9 involutions, and I(G) ⊆ {c, a2, b2, d2, ab−1, b−1a, ad−1, d−1a, bd−1, d−1b}. Proof. Since a−1 = bd, we get b−1 = da and d−1 = ab. None of these elements can be involutions. Now o(bd) = o(db) and o(da) = o(ad) and o(ab) = o(ba). Hence these elements can’t be involutions either. The result now follows immediately from Lemma 3.19 and the fact that any group containing involutions has an odd number of them. Proposition 3.22. Suppose Assumption 3.18 holds. Let K be the cen- tralizer CG(c) of c in G, and let J = K ∩ T (S). Then |K| 6 4|J |, and hence |G| 6 4|J | · |cG|. Proof. Let K be the centralizer in G of c. If x ∈ K and x2 = a, then that would imply 〈S〉 is abelian, because 〈S〉 = 〈a, c〉 = 〈x2, c〉 and x “adm-n2” — 2021/7/19 — 10:26 — page 188 — #24 188 Groups containing LMPFS of size 4 commutes with c. Similarly x cannot be a square root of b or d. Hence K ⊆ √ c ∪ J . Suppose there exist x, y ∈ √ c ∩ K with xy /∈ J . Let z ∈ √ c ∩ K. Now xy, xz and yz are elements of K because K is a subgroup. Suppose xz /∈ J and yz /∈ J . Then we have (yz)2 = c, which implies zy = cyz. Similarly yx = cxy and zx = cxz. But now (xyz)2 = xyzxyz = xycxzyz = cxyxzyz = cxcxyzyz = c2x2(yz)2 = 1. Therefore xyz /∈ √ c. Thus xyz ∈ J . So either xz ∈ J , yz ∈ J or xyz ∈ J . Hence√ c∩K ⊆ x−1J ∪ y−1J ∪ (xy)−1J . Remembering that K = J ∪ ( √ c∩K) we immediately derive |K| 6 4|J |. The remaining possibility is that there do not exist x, y ∈ √ c∩K with xy /∈ J . This means either K = J (because√ c ∩K = ∅), or that there is some x ∈ √ c ∩K, but for all y ∈ √ c ∩K we have xy ∈ J . Hence √ c ∩K ⊆ x−1J . Either way, |K| 6 2|J |. Hence |K| 6 4|J | and so |G| 6 4|J | · |cG|. Lemma 3.23. Suppose Assumption 3.18 holds. Let K be the centralizer CG(c) of c in G, and let J = K ∩ T (S). Then J ⊆ { 1, c, a2, b2, d2, ba, ad, db, ab−1, ba−1, a−1b, b−1a, ad−1, da−1, a−1d, d−1a, bd−1, db−1, b−1d, d−1b } . In particular, |J | 6 20. Proof. Since 〈S〉 is not abelian, J doesn’t contain a, b or d. Similarly no ele- ment of the form xc or cx can be contained in J , where x ∈ {a±1, b±1, d±1} as this would imply the presence in J of either a, b or d. Hence we remove these elements from our original list for T (S). The other observation is that since a−1 = bd, b−1 = da and d−1 = ab, these three elements cannot be contained in J , because again this would imply the presence in J of a, b or d. Corollary 3.24. Suppose S is a LMPFS of size 4 in a finite group G such that every two-element subset of S generates 〈S〉. In particular, take S = {a, b, c, d} where c is an involution, a−1 = bd has order at least 4 and none of b or d is an involution. Furthermore, suppose 〈S〉 is not dihedral or abelian, then |G| 6 720. Moreover, G has 1, 3, 5, 7 or 9 involutions and I(G) ⊆ {c, a2, b2, d2, ab−1, b−1a, ad−1, d−1a, bd−1, d−1b}. Proof. As a−1 = bd, by Lemma 3.21 we know |cG| 6 9 and by Lemma 3.23 we have |J | 6 20. Hence by Proposition 3.22, |G| 6 4× 20× 9 = 720. The latter fact ( that G has 1, 3, 5, 7 or 9 involutions and possible elements of I(G) ) follows from Lemma 3.21. “adm-n2” — 2021/7/19 — 10:26 — page 189 — #25 C. S. Anabanti 189 Lemma 3.25. In Corollary 3.24, no LMPFS S exists if G contains either 1 or 9 involutions. Proof. Let G and S be as defined in Corollary 3.24. If G contains only one involution, then c is central; so 〈S〉 = 〈a, c〉 is abelian, contradicting the hypothesis. Suppose G contains exactly 9 involutions. So nine out of the ten likely elements of I(G) listed in Corollary 3.24 are involutions. Note that for any g, h ∈ G, o(gh) = o(hg). So ab−1 is an involution if and only if b−1a is, and so on. Since only one of the above list of ten things is not an involution, it must be the case that all of ab−1, b−1a, ad−1, d−1a, bd−1 and d−1b are involutions, as is c, and exactly two of a2, b2 and d2 are involutions. Without loss of generality, suppose a2 and b2 are involutions. Using a−1 = bd, we get ad−1 = a2b and so on; so the nine involutions of G as c, a2, b2, ab−1, b−1a, a2b, aba, bab and ab2. From (b−1a)2 = (aba)2, we obtain b−1a = aba2b2; whence ab−1a = a2ba2b2 = b. Now ab−1a−1 = ba−2 = ba2; a contradiction since o(b) = 4, however (ba2)2 = 1 = (a2b)2. Thus no such locally maximal product-free set S exists. Remark 3.26. The aim of this remark is to assert that in Corollary 3.24, |G| < 40 if G contains either 3, 5 or 7 involutions. If G has ex- actly 7 involutions, then at least two of the three pairs ab−1, b−1a, ad−1, d−1a, bd−1, d−1b are involutions; so it means we can remove at least 4 elements from the count of J . Hence |J | 6 16 and |cG| 6 7. By Propo- sition 3.22 therefore |G| 6 4 × 16 × 7 = 448. We show that this case is not easily resolvable like the case where the number of involutions in G is 9 as seen in Lemma 3.25. For instance, if the two pairs ab−1, b−1a, bd−1 and d−1b are involutions, together three other involutions c, (ab)2 and a2, then |〈S〉| yields different values. As d−1 = ab, we write 〈S〉 = 〈a, b | (ab−1)2 = (ab2)2 = (bab)2 = a4 = (ab)4 = bn = 1〉, where n is the order of b; for example if n = 3, 4, 5 or 6, then 〈S〉 ∼= S4, D8, D10 or C2 × S4 respectively. So we can’t make further deductions from 〈S〉, with- out knowing at least a dihedral subgroup of 〈S〉. Now, suppose G contains exactly 5 involutions. Then we can remove at least 2 elements from the count of J . So |J | 6 18 and |cG| 6 5. Hence |G| 6 4×18×5 = 360. Finally, if G contains exactly 3 involutions, then we can’t remove anything from J necessarily; so |J | 6 20 and |cG| 6 3. Therefore |G| 6 4× 20× 3 = 240. In each of these cases, S = {c, b, d, (bd)−1}, where b and d are arbitrary elements of order at least three. We checked in GAP for existence of a LMPFS S in the respective groups of even orders. Our result is summarised below. “adm-n2” — 2021/7/19 — 10:26 — page 190 — #26 190 Groups containing LMPFS of size 4 |I(G)| |NAG|I(G)|| GAP I.D of groups G contain- ing required S 8 6 n 6 40 3, 5, 7 47 [20, 3] and [32, 14] 42 6 n 6 240 3, 5, 7 1665 ———————— 242 6 n 6 360 5, 7 4525 ———————— 362 6 n 6 448 7 3036 ———————— In the table above, n gives the range of orders of the nonabelian groups of even order tested. We performed four checks in GAP. The first line of result is the outcome of our first check. Our first check was for nonabelian groups of orders from 8 up to 40 that contain either 3, 5 or 7 involutions. Only 47 nonabelian groups of even order n ∈ [8, 40] contain either 3, 5 or 7 involutions. Among these groups, only in the groups whose GAP IDs are [20, 3] and [32, 14] that we found our required LMPFS. The group of order 20 mentioned is mainly referred to as the only non-simple Suzuki group; it is denoted by Suz(2) = 〈x, y | x5 = 1 = y4, xy = yx2〉. There are 40 LMPFS of size 4 in this group ( examples are {x, y2, x2y, xy3} and {x3, y2, x2y, xy3} ) ; each of the 40 LMPFS is of our required form, and under automorphisms of the group, is one of the two mentioned. On the other hand, the group of order 32 mentioned is called the semidihedral group of C8 and C4 of dihedral type. It has a presentation as C8 ⋊Dih type C4 = 〈x, y | x8 = 1 = y4, xy = yx−1〉. This group has 17 LMPFS of size 4; only 8 of them are of our required form, and up to automorphisms of the group, any LMPFS of our kind is either {x, x6, y2, x5y2} or {x, y2, x2y2, x5y2} (4 belonging to each class). Our second check was for nonabelian groups of even orders from order 42 up till 240 that contain either 3, 5 or 7 involutions. Only 1665 nonabelian groups in that range contain either 3, 5 or 7 involutions, and our search shows that no such group contains our desired LMPFS of size 4. The table is now easily understood for the third and fourth check, which are precisely the last two rows of the table. We now turn back to give a proof of Theorem 3.8. Proof of Theorem 3.8. Suppose G is a finite group containing a LMPFS of size 4 such that every 2-element subset of S generates 〈S〉. Corollary 3.10 tells us that no such S exists if S were to contain at least two involutions. If S contains no involution, then Corollary 3.15 tells us that |G| 6 40. Finally, if S contains exactly one involution, then Proposition 3.16, Lemma 3.20, Lemma 3.17, Corollary 3.24, Lemma 3.25 and Remark 3.26 yield |G| 6 40. “adm-n2” — 2021/7/19 — 10:26 — page 191 — #27 C. S. Anabanti 191 A deduction from the classification of finite groups containing locally maximal product-free sets of size 4 studied in this paper is the following: Corollary 3.27. If a finite group G contains a LMPFS S of size 4, then either |G| 6 40 or G is nonabelian, |G| is even, S ∩ S−1 6= ∅ and not every 2-element subset of S generates 〈S〉. Proof. Follows from Theorems 3.2, 3.8, 3.11, Remark 3.7 and Corollary 3.12 . We close the discussion on finite groups containing LMPFS of size 4 with the following: Conjecture 3.28. If a finite group G contains a LMPFS of size 4, then |G| 6 40. In the light of Theorem 3.2 and Proposition 3.3 as well as some computational investigations, we ask the following question: Question 3.29. Does there exists a finite group G containing a locally maximal product-free set of size k (for k > 4) such that |G| > 2k(2k− 1)? We write nk (resp. mk) for the maximal size of a finite abelian group of even (resp. odd) order containing a LMPFS of size k. Some experimental results on finite abelian groups G of even order nk (resp. odd order mk) containing LMPFS of size k are reported below. k nk G 1 4 C4 2 8 C8 and C4 × C2 3 16 C4 × C4 4 24 C24 5 36 C36, C18 × C2, C12 × C3, C6 × C6 6 48 C48, C12 × C4, C12 × C2 2 , C6 × C3 2 7 64 C3 4 , C8 × C4 × C2 k mk G 1 3 C3 2 7 C7 3 15 C15 4 27 C3 3 , C9 × C3 5 35 C35 6 45 C45, C15 × C3 7 61 C61 One may be moved by the above result to conjecture that ‘if a finite abelian group G of even order contains a LMPFS of size k, then |G| 6 (k + 1)2’. However, such conjectural statement cannot hold in general as C84 contains some locally maximal product-free sets of size 8. On another remark, let C2n = 〈x | x2n = 1〉 be the finite cyclic group of order 2n, and suppose S is a locally maximal product-free set in a finite cyclic group of even order containing the unique involution in the group. Then S is “adm-n2” — 2021/7/19 — 10:26 — page 192 — #28 192 Groups containing LMPFS of size 4 also a locally maximal product-free subset of the finite dicyclic group Q4n = 〈x, y | x2n = 1, xn = y2, xy = yx−1〉 of order 4n; the reason is because {xiy | 0 6 i 6 2n− 1} ⊂ √ xn ⊆ √ S, and already we know that {xi | 0 6 i 6 2n−1} ⊆ S∪SS∪SS−1∪ √ S. Thus, one may obtain a lower bound on the maximal size of a finite group containing a locally maximal product-free set of size k by using the bound from the dicyclic group counterpart. For instance, the maximal size of a finite cyclic group of even order containing a locally maximal product-free set of size k which contains the unique involution is 4, 6, 12, 20, 30 and 40 for k = 1, 2, 3, 4, 5 and 6 respectively. An example of a locally maximal product-free set containing the unique involution in C4, C6, C12, C20, C30 and C40 is given respectively as {x2}, {x2, x3}, {x2, x6, x10}, {x4, x7, x9, x10}, {x2, x6, x15, x22, x27} and {x3, x8, x20, x29, x33, x39}. This tells us that there are locally maximal product-free set(s) of sizes 1, 2, 3, 4, 5 and 6 in Q8, Q12, Q24, Q40, Q60 and Q80 respectively. So if a finite group G contains a locally maximal product-free set of size 1, 2, 3, 4, 5 or 6, then |G| > 8, 12, 24, 40, 60 or 80 respectively. Experimental results as well as results of [5, 9] suggest that the largest size of a finite group containing a LMPFS of size 1, 2, 3, 4, 5 or 6 is 8, 16, 24, 40, 64 or 96 respectively. We take this opportunity to list all finite groups G of expected highest possible size which contain locally maximal product-free sets of size k for k ∈ {1, 2, 3, 4, 5}. k G 1 G8 := 〈x, y | x4 = 1, x2 = y2, xy = yx−1〉 ∼= Q8 2 G16A := 〈x, y | x4 = 1 = y4, xy = y−1x〉 G16B := 〈x, y, z | x4 = 1 = z2, x2 = y2, xy = yx−1, yz = zy, xz = zx〉 ∼= Q8 × C2 3 G24A := 〈x, y | x12 = 1, x6 = y2, xy = yx−1〉 ∼= Q24 G24B := 〈x, y, z | x4 = 1 = z3, x2 = y2, xy = yx−1, yz = zy, xz = zx〉 ∼= Q8 × C3 4 G40A := 〈x, y | x8 = 1 = y5, xy = y−1x〉 G40B := 〈x, y | x20 = 1, x10 = y2, xy = yx−1〉 ∼= Q40 G40C := 〈x, y, z | x10 = 1 = z2, x5 = y2, xy = yx−1, yz = zy, xz = zx〉 ∼= Q20 × C2 5 G64A := 〈a, b | a8 = b4 = b−2a−1b−2a = ab−1a2ba = a−1b−1ab−1a−1b−1ab = 1〉 “adm-n2” — 2021/7/19 — 10:26 — page 193 — #29 C. S. Anabanti 193 G64B := 〈a, b | a4 = b8 = a2b−1a2b = a−1b2ab2 = (aba−1b)2 = 1〉 G64C := 〈a, b, c | a4 = b4 = c2 = cb−1cb = ca−1ca = a2b−1a−2b = ba−1b2ab = b−1a−1bab−1aba−1 = (a−1b−2a−1)2 = a−1(ba)2ba−1b = 1〉 G64D := 〈a, b, c | a4 = b4 = c2 = cb−1cb = a−1bab = bab−2a−1b = (aca)2 = a2b−1a2b = (ca−1)4 = (b−1a−2b−1)2 = (caca−1)2 = 1〉 G64E := 〈a, b, c | a4 = b4 = c2 = ba−1ba = b2cb−2c = a2ba2b−1 = ca−1b−2ca−1 = (cb−1)2(cb)2 = (a−1b2a−1)2 = (cbacba−1)2 = 1〉 G64F := 〈a, b, c | a4 = b8 = c2 = cbcb−1 = ca−1ca = a−1bab = a2b−1a2b = 1〉 G64G := 〈x, y, z | x8 = 1 = z4, x4 = y2, xy = yx−1, yz = zy, xz = zx〉 ∼= Q16 × C4 G64H := 〈a, b, c | a4 = c−1b−1cb = c4a2 = c2ac2a−1 = b4a2 = b2ab2a−1 = b−1ac2ba−1 = c−1a−1b−1cb−1a−1 = 1〉 G64I := 〈a, b, c | a4 = c4 = a−1bab = a−1cac = c−1b−1cb = a−1b−4a−1 = a2c−1a2c = (c−1a−2c−1)2 = 1〉 G64J := 〈a, b, c | b4 = c8 = ab−1a−1b−1 = c−1b−1cb = a−1cac = b−1a−1ba−1 = 1〉 G64K := 〈a, b, c, d | a4 = b4 = c2 = d2 = da−1da = a−1bab = ca−1ca = db−1db = (dc)2 = cb−1cb = a2b−1a2b = (b−1a−2b−1)2 = 1〉 ∼= G16A × C2 2 G64L := 〈a, b, c, d | a4 = c2 = d2 = aba−1b = db−1db = (dc)2 = cb−1cb = ba2b = da−1da = b−1a−1ba−1 = a2ca−2c = (ca)4 = cacdaca−1cda−1 = 1〉 G64M := 〈a, b, c, d | a4 = b2 = d2 = c2a2 = caca−1 = c2a−2 = (db)2 = dc−1dc = (aba)2 = a−1ba−1cbc = (a2d)2 = da−1bdab = (ba−1)4 = c−1bcbaba−1b = 1〉 G64N := 〈a, b, c, d, g | a4 = c2 = d2 = g2 = b2a2 = baba−1 = gb−1gb = ca−1ca = ga−1ga = (gc)2 = cb−1cb = da−1da = (gd)2 = (dc)2 = db−1db = 1〉 ∼= Q8 × C3 2 . Acknowledgements The author is grateful to Professor Sarah Hart for useful discussion during the writing of this paper. He is also thankful to Birkbeck college for the financial support provided. “adm-n2” — 2021/7/19 — 10:26 — page 194 — #30 194 Groups containing LMPFS of size 4 References [1] C. S. Anabanti, On filled soluble groups, Communications in Algebra, 46(11) (2018), 4914–4917. [2] C. S. Anabanti, On the three questions of Bertram on locally maximal sum-free sets, Quaestiones Mathematicae, 44(3) (2021), 301–305. [3] C. S. Anabanti, Three questions of Bertram on locally maximal sum-free sets, Applicable Algebra in Engineering, Communication and Computing, 30(2) (2019), 127–134. [4] C. S. Anabanti, G. Erskine and S. B. Hart, Groups whose locally maximal product- free sets are complete, The Australasian Journal of Combinatorics, 71(3) (2018), 544–563. [5] C. S. Anabanti and S. B. Hart, Groups containing small locally maximal product- free sets, International Journal of Combinatorics, vol. 2016, Article ID 8939182 (2016), 5pp. [6] C. S. Anabanti and S. B. Hart, On a conjecture of Street and Whitehead on locally maximal product-free sets, The Australasian Journal of Combinatorics, 63(3) (2015), 385–398. [7] E. A. Bertram, Some applications of Graph Theory to Finite Groups, Discrete Mathematics, 44 (1983), 31–43. [8] The GAP Group, GAP – Groups, Algorithms, and Programming, Version 4.8.6 ; 2016, (http://www.gap-system.org). [9] M. Giudici and S. Hart, Small maximal sum-free sets, The Electronic Journal of Combinatorics, 16 (2009), 17pp. [10] K. S. Kedlaya, Product-free subsets of groups, American Mathematical Monthly, 105 (1998), 900–906. [11] A. P. Street and E. G. Whitehead Jr., Group Ramsey Theory, Journal of Combi- natorial Theory Series A, 17 (1974), 219–226. [12] A. P. Street and E. G. Whitehead, Jr., Sum-free sets, difference sets and cyclotomy, Comb. Math., Lect. notes in Mathematics, Springer-Verlag, 403 (1974), 109–124. Contact information Chimere Stanley Anabanti Department of Mathematics and Applied Mathematics, University of Pretoria, South Africa E-Mail(s): chimere.anabanti@up.ac.za Web-page(s): https://www.up.ac.za/ mathematics-and-applied-mathematics/ article/2953786/ dr-cs-chimere-anabanti Received by the editors: 05.03.2019.