Abstract
We study finite groups G having a subgroup H and \(D \subset G {\setminus } H\) such that (i) the multiset \(\{ xy^{-1}:x,y \in D\}\) has every element that is not in H occur the same number of times (such a D is called a relative difference set); (ii) \(G=D\cup D^{(-1)} \cup H\); (iii) \(D \cap D^{(-1)} =\emptyset \). We show that \(|H|=2\), that H is central and that G is a group with a single involution. We also show that G cannot be abelian. We give infinitely many examples of such groups, including certain dicyclic groups, by using results of Schmidt and Ito.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Here G will always be a finite group. We identify \(X \subseteq G\) with the element \(\sum _{x \in X} x \in {\mathbb {Q}} G\), and let \(X^{(-1)}=\{x^{-1}:x \in X\}\). We write \({\mathcal {C}}_n\) for the cyclic group of order n. Let \(H\le G\) and \(h=|H|\). Then a \((v,k,\lambda )\)-relative difference set (relative to H) is a subset \(D \subset G{\setminus } H, |D|=k, v=|G|\), such that \(DD^{(-1)}=\lambda (G-H)+k\), so that \( g \in G{\setminus } H\) occurs \(\lambda \) times in the multiset \(\{xy^{-1}:x,y \in D\}\).
We now further assume
-
(1)
\(D \cap D^{(-1)}=\emptyset \);
-
(2)
\(G = D \cup D^{(-1)} \cup H\) (disjoint union).
A group having a difference set of the above type will be called a \((v,k,\lambda )\)-skew relative Hadamard difference set group (with difference set D and subgroup H); or a \((v,k,\lambda )\)-SRHDS group. Recall the following related concept: a group G is a skew Hadamard difference set if it has a difference set D where \(G=D\cup D^{(-1)} \cup \{1\}\) and \(D \cap D^{(-1)}=\emptyset \). Such groups have been studied in [1,2,3,4,5,6,7,8].
In this paper we find infinitely many examples of such SRHDS groups. We also find groups that cannot be SRHDS groups, but which satisfy certain properties of a SRHDS group, as given in:
Theorem 1.1
For a \((v,k,\lambda )\) SRHDS group G with difference set D and subgroup H we have:
-
(i)
\(|H|=2\);
-
(ii)
\(H \triangleleft G\);
-
(iii)
G is a group having a single involution;
-
(iv)
\(v \equiv 0 {\text {mod}}8\);
-
(v)
G is not abelian.
-
(vi)
A Sylow 2-subgroup is a generalized quaternion group.
For part (vi), suppose that G is a finite group with a unique involution. Then a Sylow 2-subgroup of G also has a unique involution. Now 2-groups with unique involution were determined by Burnside (see [9, Theorems 6.11, 6.12] and [10, 11]); they are cyclic or generalized quaternion groups. Corollary 4.5 shows they cannot be cyclic.
Groups with a single involution are studied in [12,13,14]. Dicyclic groups \(\textrm{Dic}_{ {v}}\) are examples of such groups and we show that each \(\textrm{Dic}_{8p}, 1\le p<9\) is an SRHDS group. However, we show that \(\textrm{Dic}_{72}\) has no SRHDS (Proposition 8.1).
We now establish a connection between SRHDS groups and Hadamard groups. Recall that a Hadamard group is a group G containing \(H\le Z(G)\) of order 2 such that there is an H-transversal \(D, |D|=v/2,\) that is a relative difference set relative to H (so that \(DD^{(-1)}=\lambda (G-H)+|D|\) and \(HD=G\)).
We show that if \(D \subset G\) is a SRHDS, then G is also a Hadamard group (where \(E=D+1\) is the relative difference set); see Proposition 2.5. Thus it is natural to try to obtain results for SRHDS groups that are similar to the results of Schmidt and Ito [15, 16] from the Hadamard group situation. For example Schmidt and Ito show that if \(4p-1\) or \(2p-1\) is a prime power, then the groups \(\textrm{Dic}_{8p}\) or \(\textrm{Dic}_{4p}\) (respectively) are Hadamard groups. For dicyclic SRHDS groups we show:
Theorem 1.2
If \(p \in {\mathbb {N}}\) and \(4p-1\) is a prime power, then \(\textrm{Dic}_{8p}\) is a SRHDS group.
There is no analogous result when \(2p-1\) is prime. Now Ito [16] determines a ‘doubling process’ that takes a Hadamard difference set for \(\textrm{Dic}_{ {v}}\) and produces a Hadamard difference set for \(\textrm{Dic}_{2 {v}}\). For us this doubling process gives:
Theorem 1.3
If \(p \in {\mathbb {N}}\) and \(4p-1\) is a prime power, then \(\textrm{Dic}_{16p}\) is a SRHDS group.
We note that this doubling process does not work in general in the context of a SRHDS, however in our next paper we will show that it does work for a SRHDS under an additional hypothesis that we call doubly symmetric that is satisfied in the situation of Theorem 1.2, so that in this case we obtain an SRHDS in \(\textrm{Dic}_{16p}\). This will allow us to prove, in the next paper, among other things:
Theorem 1.4
Let \(G=\mathrm Dic_{8\cdot 2^u}\) be a generalized quaternion group for some \(u\in {\mathbb {Z}}_{\ge 0}\). Then G contains a doubly symmetric SRHDS if and only if \(2^{u+1}-1\) is either prime or 1.
Lastly, the following is a consequence of Proposition 8.2.
Theorem 1.5
Let \(G = C_p \times \textrm{Dic}_{8n}\) with \(p>2\) prime and n odd. Then G is not a SRHDS group.
2 \(|H|=2\) and Normality of H
Recall that for \(p \ge 2\) the dicyclic group of order 4p is
A generalized quaternion group, \(Q_{2^a}\), is the dicyclic group \(\textrm{Dic}_{2^a}\), \(a\ge 3\).
Proposition 2.1
Let G be a SRHDS group with subgroup H. Then G has a single involution t, and \(H=\langle t\rangle \). In particular \(h=2, H\le Z(G)\) and \(H \triangleleft G\).
Proof
Let \(D\subset G\) be a SRHDS. Now D has no involutions since \(D \cap D^{(-1)}=\emptyset \). Since \(G-(D+D^{(-1)})=H\) all involutions are contained in H.
If \(d_i \in D, h_i \in H,i=1,2\), with \(h_1d_1=h_2d_2\in Hd_1 \cap Hd_2,\) then \(h_2^{-1}h_1=d_2d_1^{-1} \in H\), so that \(h_2^{-1}h_1=d_2d_1^{-1} =1\) (since \(DD^{(-1)}=\lambda (G-H)+k\) implies that the only element of H of the form \(d_2d_1^{-1} \) is 1). Thus \(d_1=d_2\) and \(h_1=h_2\).
Thus the cosets \(Hd, d \in D,\) are disjoint and so \(|\cup _{d \in D} Hd|=|H|\cdot |D|=hk\). Since \(Hd \subset G - H\) for \(d \in D\), we see that \(hk=|\cup _{d \in D} Hd| \le |G{\setminus } H|= |D+D^{(-1)}|=2k\). Thus \(h\le 2\) and so \(h=2\) as \(h>1\). The rest of the result follows. \(\square \)
This proves (i), (ii) and (iii) of Theorem 1.1. In what follows we will let \(H=\langle t\rangle \), where \(t \in Z(G)\) has order 2. Then:
These equations give: \(v=2k+2, k^2=k+\lambda (v-2),\) and solving gives (i) of
Lemma 2.2
-
(i)
\(v=2k+2, \quad \lambda =(k-1)/2=(v-4)/4\) and 4|v.
-
(ii)
\(DH=HD=D^{(-1)}H=HD^{(-1)}=G-H. \)
-
(iii)
\(G,D,D^{(-1)},H\) all commute.
Proof
From \(D \subset G-H\) we have \(DH \cap H = \emptyset \), and \(DH \subset G-H\); but \(|G-H|=2k=|DH|\), so that
giving (ii).
Since \(D^{(-1)}=G-D-H\) and \(H \le Z(G)\) it now follows that D and \(D^{(-1)}\) commute. This shows that \(G,D,D^{(-1)},H\) all commute. \(\square \)
Lemma 2.3
Let G be a SRHDS group with difference set D and subgroup \(H=\langle t\rangle .\) Then \(D^{(-1)}=tD\).
Proof
We have \(D+Dt=(1+t)D=HD=G-H=D+D^{(-1)}\). \(\square \)
We now define Schur rings [17,18,19,20]. A subring \({\mathfrak {S}}\) of \({\mathbb {Z}} G\) is a Schur ring (or S-ring) if there is a partition \({\mathcal {K}}=\{C_i\}_{i=1}^r\) of G such that:
-
1.
\(\{1_G\}\in {\mathcal {K}}\);
-
2.
for each \(C\in {\mathcal {K}}\), \(C^{(-1)}\in {\mathcal {K}}\);
-
3.
\({C_i}\cdot {C_j}=\sum _k \lambda _{i,j,k} {C_k}\); for all \(i,j \le r\).
The \(C_i\) are called the principal sets of \({\mathfrak {S}}\). Then we have:
Lemma 2.4
\(\{1\}, \{t\},D,D^{(-1)}\) are the principal sets of a commutative Schur ring.
Proof
Now \(\{1\}, \{t\},D,D^{(-1)}\) partition G and \(D^{(-1)}=tD,tD^{(-1)}=D,t^2=1,D^{(-1)}D=DD^{(-1)}=\lambda (G-H)+k= \lambda (D+D^{(-1)})+k, D^2=tDD^{(-1)}=t( \lambda (D+D^{(-1)})+k).\) This concludes the proof. \(\square \)
Proposition 2.5
If \(D \subset G\) is a SRHDS, then G is a Hadamard group.
Proof
Now \(DD^{(-1)}=\lambda (G-H)+k\). Let \(E=D+1\), so that \(EE^{(-1)}=DD^{(-1)}+D+D^{(-1)}+1= \lambda (G-H)+k+(G-H) =(\lambda +1)(G-H)+k+1,\) as required. \(\square \)
3 Intersection Numbers
Let \(N\triangleleft G\) and let \(g_1,g_2,\dots ,g_r\) be coset representatives for G/N. Then for each \(1\le i \le r\) there is \(1\le i'\le r\) such that \(g_ig_{i'} \in N\) i.e. \(Ng_i\cdot Ng_{i'}=N\) in G/N. If G is a SRHDS group with difference set D, then the numbers \(n_i=|D \cap Ng_i|\) are called the intersection numbers. Standard techniques give (see Section 7.1 of [21]):
Lemma 3.1
Let \(D \subset G\) be a SRHDS with subgroup \(H=\langle t\rangle , t^2=1\). Let \(N\triangleleft G\) have order s and index r in G. Let \(g_1=1,g_2,\dots ,g_r\) be coset representatives for G/N and let \(n_i=|D \cap Ng_i|, 1\le i\le r\). Then
Lemma 3.2
Let \(N \triangleleft G\) where \(D \subset G\) is a SRHDS with subgroup H and \(H \cap N=\{1\}\). Let \(Ng_3,\cdots ,Ng_r\) be the cosets that don’t meet H, and let \(n_i=|D \cap Ng_i|\). Suppose that we have distinct \(i,i'>2\) where \(g_ig_{i'} \in N\). Then \(n_i+n_{i'}=|N|\).
Proof
We have \( n_i=|D \cap Ng_i|=|D^{(-1)} \cap Ng_i^{-1}|=|D^{(-1)} \cap Ng_{i'}|. \) If \(i\ge 3\), then \(Ng_{i'} \subset G {\setminus } H=D+D^{(-1)}\), so that
\(\square \)
The next result concerns intersection numbers for subgroups that are not necessarily normal:
Proposition 3.3
Let G be a SRHDS group with difference set D and subgroup H. Let \(K\le G\) be any subgroup where \(t \in K\). Let \(b=|G:K|\) and let \(g_0=1,g_1,\ldots ,g_{b-1}\) be coset representatives for \(K \le G\). Let \(k_i=|D \cap Kg_i|, 0\le i<b\). Then \( k_0=|K|/2-1 \text { and } k_i=|K|/2, \,\,\, 0<i<b. \)
Let \(D_i=D \cap Kg_i,i=0,\ldots ,b-1\). Then \( \sum _{i=0}^{b-1}D_iD_i^{(-1)}=\lambda (K-H)+k. \)
Proof
We have \(D^{(-1)}=tD\). Let \(D_i=D \cap Kg_i\); then \( tD_i=t(D \cap Kg_i)=(tD) \cap tKg_i=D^{(-1)} \cap Kg_i, \) so that \(D \cap tD=\emptyset \) and \(i>0\) gives
Taking cardinalities, again using \(D \cap tD=\emptyset \), gives \(2k_i=|K|,\) for \(i>0\). Then \(\sum _{i=0}^{b-1}k_i=k\) now gives
but \(v=b \cdot |K|, \) from which we obtain \(k_0=|K|/2-1\).
Now from \(DD^{(-1)}=\lambda (G-H)+k\) and \(D=\sum _{i=0}^{b-1} D_ig_i\) we get \(\sum _{i=0}^{b-1}D_iD_i^{(-1)}+\cdots =\lambda (G-H)+k\), so that \(\sum _{i=0}^{b-1}D_iD_i^{(-1)}\subseteq \lambda (K-H)+k\). The last part will follow if we can show that both sides of this equation have the same size.
From \(b=v/|K|\) and the first part, the size of the left hand side is
and (since \(H \subset K\)) the number of elements of the right hand side is \( \lambda (|K|-2)+k=2p|K|-|K|+1, \) and we are done. \(\square \)
4 Direct Products and G is not Abelian
Let \(\zeta _n=\exp {2\pi i/n}, n \in {\mathbb {N}}.\) We first show
Theorem 4.1
Suppose that \(N \unlhd G \), \(G/N \cong {\mathcal {C}}_{2^a}, a\ge 2\), and \(t \notin N\). Assume that \(k=|G|/2-1\) is not a perfect square. Then G is not a SRHDS group.
Proof
Note that \(a\ge 2\) means that k is odd. Now assume that G is a SRHDS group and that \(G/N=\langle rN \rangle \cong {\mathcal {C}}_{2^a}, r \in G\). For \(g \in G\) we have \(g=r^ib, 0\le i<2^a, b \in N\). Then there is a linear character \(\chi ':G/N\rightarrow {\mathbb {C}}^\times , \chi '(rN)=\zeta _{2^a} \) that induces \(\chi :G\rightarrow {\mathbb {C}}^\times , \chi (r^ib)=\chi '(r^iN). \) Here \(N =\ker \chi \). Then we can write
Since \(t \notin N\) we have \(\chi (t)=-1\) and so \(\chi (H)=0.\) We certainly have \(\chi (G)=0\). From \(G=D+D^{(-1)}+H\) we get \(\chi (D)+\chi (D^{(-1)})=0\), and from \(DD^{(-1)}=\lambda (G-H)+k\) we get \( \chi (D)\chi (D^{(-1)})=k. \) These give \(\chi (D)^2=-k, \text{ and } \text{ so } \chi (D)=\pm \sqrt{-k}. \) But
which gives \(\sqrt{k} \in {\mathbb {Q}}(i,\zeta _{2^a}) = \mathbb Q(\zeta _{2^a}),\) since \(a\ge 2\). But the Galois group of \( \mathbb Q(\zeta _{2^a})/{\mathbb {Q}}\) is \({\mathcal {C}}_2 \times \mathcal C_{2^{a-2}}\). These groups have at most three subgroups of index 2. The Galois correspondence tells us that \( \mathbb Q(\zeta _{2^a})\) contains at most three quadratic extensions, the only possibilities being \({\mathbb {Q}}(i)/{\mathbb {Q}}, {\mathbb {Q}}(\sqrt{2})/{\mathbb {Q}}\) and \({\mathbb {Q}}(\sqrt{-2})/{\mathbb {Q}}\). But the hypothesis says that k is not a perfect integer square, so that \(\sqrt{k} \notin {\mathbb {Z}}\). Now \(k>1\) is also odd, and so \(\sqrt{k} \notin {\mathbb {Q}}(i), {\mathbb {Q}}(\sqrt{2}), {\mathbb {Q}}(\sqrt{-2}).\) This contradiction gives Theorem 4.1. \(\square \)
Corollary 4.2
Suppose that \(N \unlhd G \), \(G/N \cong {\mathcal {C}}_{2^a}, a\ge 3\), and \(t \notin N\). Then G is not a SRHDS group.
Proof
Since \(2^a\ge 8\) we see that \(k=(|G|-2)/2\) satisfies \(k \equiv 3 {\text {mod}}4\), and so the result follows from Theorem 4.1. \(\square \)
Corollary 4.3
If G is abelian with \(|G|\equiv 0 {\text {mod}}8\), then G is not a SRHDS group.
Proof
Let G be an abelian SRHDS group, and write \(G=A \times N\) where A is a Sylow 2-subgroup, and N is a subgroup of odd order. Since G has a single involution, we see that A is cyclic, say of order \(2^a\). The results now follow from Corollary 4.2. \(\square \)
Corollary 4.4
If G is a SRHDS group, then \(v=|G| \equiv 0 {\text {mod}}8.\)
Proof
Assume that G is a SRHDS group with subgroup \(H=\langle t\rangle \) and difference set D. Then we know that 4|v by Lemma 2.2, so suppose that \(|G|=4n\) where n is odd. Then a Sylow 2-subgroup of G must be \({\mathcal {C}}_4=\langle r\rangle \) and \(t=r^2\). Burnside’s theorem [9, Theorem 5.13] shows that \(\langle r\rangle \) has a complement \(N \triangleleft G, |N|=n, G=N \rtimes \langle r\rangle \). So we can write \(D=D_0+D_1r+D_2r^2+D_3r^3, D_i \subset N\). Now \(D+D^{(-1)}=G-H=N+Nr+Nr^2+Nr^3-H\) then gives
Next, \(D^{(-1)}=tD\) gives
Using \(D_1+\big (D_3^{(-1)}\big )^{r^3}=N\) and \(\big (D_3^{(-1)}\big )^{r^3}=tD_1\) we get \(D_1(1+t)=N\). However \(D_1(1+t)\) has an even number of elements (counting multiplicities), while |N| is odd. This contradiction gives the result. \(\square \)
Corollaries 4.3 and 4.4 now prove Theorem 1.1 (iv) and (v).
Corollary 4.5
If G is a SRHDS group, then a Sylow 2-subgroup of G is not cyclic.
Proof
Assume G is a SRHDS group with cyclic Sylow 2-subgroup \(\langle r \rangle \). By Corollary 4.4, \(|\langle r \rangle |\ge 8\). Again, Burnside’s theorem [9, Theorem 5.13] shows that \(\langle r\rangle \) has a complement \(N \triangleleft G, G=N \rtimes \langle r\rangle \). This now contradicts Corollary 4.2. \(\square \)
This concludes the proof of Theorem 1.1.
5 Construction of Some SRHDS Groups
We need the following set-up: For prime power \(q=4p-1, p \in \mathbb N, \) we let \({\mathbb {F}}_{q^n}\) be the finite field of order \(q^n\). Let \(tr: {\mathbb {F}}_{q^2}\rightarrow {\mathbb {F}}_q, \beta \mapsto \beta ^q\) be the trace function. Let \(\alpha \in {\mathbb {F}}_{q^2}\) satisfy \(tr(\alpha )=0\). Let \({\mathbb {F}}_{q^2}^*=\langle z\rangle \). Let \(Q=\{u^2: u \in {\mathbb {F}}_q, u\ne 0\}\). Then \(-1\notin Q\) since \(q\equiv 3 {\text {mod}}4\). Now choose \(D \in {\mathbb {F}}_q {\setminus } (Q \cup \{0\})\). Then any \(\beta \in {\mathbb {F}}_{q^2}\) has the form \(\beta =a+b\sqrt{D},\) for some \( c,d \in {\mathbb {F}}_q\) and \(tr(c+ d\sqrt{D})=c-d\sqrt{D}.\) Write \(\alpha =a+b\sqrt{D}\). Then \(tr(\alpha )=0\) if and only if \(a=0\), so we can choose \(\alpha =\sqrt{D}.\)
Let \(U\le {\mathbb {F}}_{q^2}^*\) be the subgroup of order \((q-1)/2\), and let \(\pi : {\mathbb {F}}_{q^2}^* \rightarrow W:={\mathbb {F}}_{q^2}^*/U\) be the natural map.
Theorem 5.1
Suppose that \(4p-1\) is a prime power. Then \(\textrm{Dic}_{8p}\) contains a SRHDS.
Proof
We follow [15, Theorem 3.3].
Let \(q=4p-1\) and assume the above set-up. Let \(g:=\pi (z)\) be a generator for W and note that \(|W|=2(q+1)=8p\). Let \( R=\{\pi (x): x \in {\mathbb {F}}_{q^2}^*, tr(\alpha x) \in Q \}. \) Then by [22, Thm 2.2.12], R is a relative \((q+1,2,q,(q-1)/2)\) difference set in W relative to the subgroup \(H:=\langle g^{4p}\rangle \) of order 2.
Define \(R_1,R_2 \subset W_2:=\langle g^2\rangle \) by \( R=R_1+R_2g. \) Since R is a relative \((q+1,2,q,(q-1)/2)\) difference set, \(RR^{(-1)}=\frac{q-1}{2} (W-H)+q\) from which we get
If \(d \in {\mathbb {F}}_{q^2}^*\) has order dividing \(q+1\), then \(d^q=d^{-1}\) and so
Thus if tr\((\alpha d) \in Q\), then tr\((\alpha d^{-1}) \in -Q.\) But \(q \equiv 3 {\text {mod}}4\) tells us that \(g^{4p}=-1 \notin Q\), so that tr\((\alpha g^{4p} d^{-1})\in Q\). Thus \(g^{4p}d^{-1} \in R_1\). Now the order of \(g^{4p}d^{-1}\) is a divisor of \(2(q+1)=|W|\). This gives a bijection, \(Ud \leftrightarrow U g^{4p}d^{-1},\) between the elements of \(R_1 \subset W_2\), which then gives \(R_1^{(-1)}=g^{4p}R_1\).
Now let \(G=\textrm{Dic}_{8p}\) \(=\langle a,b|a^{2p}=b^2,b^4=1,a^b=a^{-1}\rangle \) and identify \(\langle a\rangle \) with \(W_2\), so that \(a \leftrightarrow g^2\). From \(R_1^{(-1)}=g^{4p}R_1\) we see that if \(\gamma \in R_1 \cap R_1^{(-1)}\), then \(g^{4p} \in R_1R_1^{(-1)}\), a contradiction to R being a relative difference set relative to H. It follows that \(R_1 \cap R_1^{(-1)} =\emptyset .\) Now \(1,-1=g^{4p} \notin R_1\) as tr\((\alpha 1)=0 \notin Q\), and so
Then (5.1) and \(R_1^{(-1)}=g^{4p}R_1\) gives
so that we have the first part of
Lemma 5.2
-
(i)
\(R_1+1\) is a transversal for \(W_2/H\).
-
(ii)
\(R_2\) is a transversal for \(W_2/H\).
Proof
(ii) We first show that R+1 is a transversal for W/H.
If \(u \in W\), then tr\((\alpha u) \in Q\), and it follows that tr\((\alpha g^{4p} u) =- tr(\alpha u) \notin Q\). This sets up a bijection \(u \leftrightarrow g^{4p} u\) of \(W-H\) where the orbits of this bijection are the non-trivial H-cosets and a transversal corresponds to the elements of Q.
Since R+1 is a transversal for W/H and \(R_1+1\) is a transversal for \(W_2/H\) it follows that \(R_2\) is a transversal for \(W_2/H\). This concludes the proof. \(\square \)
Now if \(\alpha =\sqrt{D}, \beta =a+b\sqrt{D}\), then tr\((\alpha \beta )=2bD \in Q\) if and only if \(2b \in {\mathbb {F}}_q^* {\setminus } Q\).
Define \( S:=a^{2p} R_1+R_2b. \) First we show that \(SS^{(-1)}=\lambda \left( G -H\right) +k\) where \(k=(v-2)/2, \lambda =(k-1)/2\):
as desired. Next we need
Lemma 5.3
For S as above we have \(S \cap S^{(-1)} =\emptyset .\)
Proof
So assume that \(r \in S \cap S^{(-1)}, S=a^{2p} R_1+R_2b\). Then there are two cases.
-
(a)
First assume that \(r \in \langle a\rangle .\) Then there are \(x^i, x^j \in R_1\) where \(r=a^{2p}a^i=a^{2p}a^{-j}\) so we have \(i=-j\). Since a corresponds to \(g^2\) the elements \(g^{2i},g^{-2j}\) satisfy tr\((\alpha g^{2i}), \) tr\((\alpha g^{-2j}) \in Q\). Let \(g^{i}=c+b \sqrt{D}\). Then tr\((\alpha g^{2i}), \) tr\((\alpha g^{-2j}) \in Q\) (respectively) gives \( 4bcD \in Q,-\frac{4bcD}{(c^2-b^2D)^2} \in Q\) (respectively), which in turn gives \(-1 \in Q\), a contradiction.
-
(b)
Next assume that \(r \in \langle a\rangle b.\) Then there are i, j such that \(r=a^ib=(a^jb)^{-1}=a^{j+2p}b,\) where \(a^i, a^j \in R_2\). Thus \(i=j+2p\). As in the first case this gives tr\((\alpha g^{2i+1}),\) tr\((\alpha g^{2j+1})=tr(\alpha g^{2i-4p+1})\in Q\). Since tr\((\alpha g^{2i-4p+1})=-\)tr\((\alpha g^{2i+1})\), this gives \(-1 \in Q\), a contradiction.
From \(S \cap S^{(-1)} =\emptyset =S \cap H\) we get \(G=S+S^{(-1)}+H\) and so Eq. (5.2) shows that S is a SRHDS, giving Theorem 5.1. \(\square \)
We next wish to show that we can double these examples (see Sect. 6 for the definition of this doubling process), and we will need the following symmetry results:
Symmetry proof for \(R_1\) . Now \(S=a^{2p}R_1+R_2b\) and if \(a^i \in a^{2p}R_1\), then \(i=2p+j\) where \(tr(\alpha z^{2j}) \in Q\). We note that z, the generator of \({\mathbb {F}}_{q^2}^*\), has order \(q^2-1\), and so \((z^{q})^q=z\), showing that the non-trivial Galois automorphism is given by \(z \mapsto z^q\).
So from \(tr(\alpha z^{2j}) \in Q\) we get \(tr(\alpha ^q z^{2jq}) \in Q\). But \(\alpha ^q=-\alpha =\alpha z^{(q^2-1)/2}\). Thus
This if \(j'=(jq+(q^2-1)/4)\), then \(a^{2p+j'} \in a^{2p}R_1\), and so \(j \mapsto j'\) determines a function \(R_1 \rightarrow R_1\) that one can show is an involution.
One can then check that \(j=p+r\) is sent to \(j'=p-r\) (recalling that j is defined mod 4p). This gives a ‘reflective’ symmetry for \(R_1\).
Symmetry proof for \(R_2\) . We now do a similar thing for \(R_2\). So let \(a^ib \in R_2b\), so that \(tr(\alpha z^{2i+1}) \in Q\). Then acting by the Galois automorphism we get
This similarly gives the involutive map
\(\square \)
6 The Doubling Process
Lemma 6.1
Let \(D\subset G=\textrm{Dic}_{ {v}}=\langle x,y\rangle , v=4n, k=2n-1, \lambda =n-1\). Let \(K=\langle x\rangle \), \(k_1=n-1,k_2=n\) and let \(D=D_1+D_2y, D_i \subset K, k_i=|D_i|.\) Then the requirement that \(D=D_1+D_2y\) is a SRHDS is equivalent to (a)–(d):
Proof
One checks that \(D=D_1+D_2y\) is a SRHDS is equivalent to the conditions
-
(i)
\(D_1 \cup \{1\}\) and \(D_2\) are transversals for K/H (this comes from looking at \(G-H=D+D^{(-1)}=D_1+D_2y+(D_1^{(-1)}+(D_2y)^{(-1)})\)).
-
(ii)
\( \lambda D_1H+k=D_1D_1^{(-1)}+D_2D_2^{(-1)}\);
-
(iii)
\( \lambda Ky=D_2D_1y+D_1D_2y^{-1} \) (from \(DD^{(-1)}=\lambda (G-H)+k\));
-
(iv)
\(D_1^{(-1)}=tD_1\) and \(D_i^y=D_i^{(-1)}.\)
Now (iii) is equivalent to \(D_1D_2(1+t)=\lambda K\) or \(D_1 K=\lambda K\). But \(D_1 K=\lambda K\) follows directly from \(D_i \subset K, \) and \(|D_1|=\lambda \). Thus (ii) and (iii) are equivalent to \(\lambda D_1H+k=D_1D_1^{(-1)}+D_2D_2^{(-1)}.\) \(\square \)
Write \(D = D_0 + D_1 y\). We construct the set \(E\subseteq \textrm{Dic}_{16p}\) as
We show that if \(D_1\) satisfies the symmetry: \(x^{2i} \in D_1\) implies \(x^{4p-2i-2} \in D_1\), then E is a \((v_2, k_2, \lambda _2)\)-SRHDS with \(v_2 = 16p\), \(k_2 = 8p - 1\), and \(\lambda _2 = 4p - 1\).
Theorem 6.2
Let \(\textrm{Dic}_{16p}=\langle x,y\) | \(x^{4p}=y^2,y^4=1, x^y=x^{-1}\rangle \), \(t=y^2\). We let \(\textrm{Dic}_{8p}=\langle x^2,y \rangle \le \textrm{Dic}_{16p}.\) Let D be a \((v_1, k_1, \lambda _1)\)-SRHDS in \(\textrm{Dic}_{8p}\), with \(v_1 = 8p\), \(k_1 = 4p - 1\), and \(\lambda _1 = 2p - 1\). Then the unique involution t in \(\textrm{Dic}_{16p}\) is the same as the unique involution in \(\textrm{Dic}_{8p}\).
Write \(D = D_0 + D_1 y, D_i \subset \langle x^2\rangle \), and let \(E = E_0 + E_1 y \ \subseteq \textrm{Dic}_{16p}\) where:
Assume that \(D_1\) satisfies the symmetry: \(x^{2i} \in D_1\) implies \(x^{4p-2i-2} \in D_1\). Then E is a \((v_2, k_2, \lambda _2)\)-SRHDS with \(v_2 = 16p\), \(k_2 = 8p - 1\), and \(\lambda _2 = 4p - 1\).
Proof
We note that \(D^{(-1)}=tD\) implies that \(E^{(-1)}=tE\). We also observe that the map \(x^{2i}\rightarrow x^{4p-2i-2}\) is an involution. Using Lemma 6.1, to show E is a SRHDS it suffices to show that E satisfies
This is sufficient because conditions (1) and (2) along with \(E^{(-1)}=tE\) imply conditions (a) and (c) of Lemma 6.1. First we note that E does not contain t or the identity, as this would imply that \(D_0\) contains these. We now show (2), which will imply (1). We split condition (2) into cases by considering the intersection of E with each coset of \(\langle x^2 \rangle \), all of which cosets are their own inverses. There are four such cosets: \(\langle x^2 \rangle , \langle x^2 \rangle x,\langle x^2 \rangle y,\) and \(\langle x^2 \rangle xy\).
\(\langle x^2\rangle :\) For \( E \cap \langle x^2 \rangle =D_0\), we know that \(x^{2i} \in D_0\) implies \(x^{-2i} \not \in D_0\) since \(D_0\cap D_0^{(-1)}=\emptyset \).
\(\langle x^2\rangle x:\) We have \(E \cap \langle x^2 \rangle x = D_1 x\). We show \(D_1 x\cap (D_1x)^{(-1)}=\emptyset \).
Here we used the symmetry and the fact that \((D_1y)\cap (D_1y)^{(-1)}=\emptyset \) where \((D_1y)^{(-1)}=tD_1y\).
\(\langle x^2\rangle y:\) Here we have \(E \cap \langle x^2 \rangle y = D_0^{(-1)} y + y\). First we check that \(D_0^{(-1)} y\) doesn’t contain any of its inverses:
We also check the additional y doesn’t have an inverse in \(D_0^{(-1)} y\):
\(\langle x^2\rangle xy:\) Here we have \(E \cap \langle x^2 \rangle x y = D_1^{(-1)} x^{-1} t y\), and
Thus \(E \cap E^{(-1)} = \emptyset \). This concludes (2) and implies (1), since both E and \(E^{(-1)}\) don’t intersect \(\langle t \rangle \) and \(|E|=k_2=8p-1.\)
Now we prove (3): we have
For E to be a SRHDS we need (6.2) to be equal to \(\lambda _2(\langle x \rangle - \langle t \rangle )+k_2\). Looking at just the even powers of x, we need
to be equal to \(\lambda _2(\langle x^2 \rangle - \langle t \rangle ) + k_2\). We note that \(D_0 + D_0^{(-1)} = \langle x^2 \rangle - \langle t \rangle \), and \(D_0 D_0^{(-1)} + D_1 D_1^{(-1)}=\lambda _1(\langle x^2 \rangle -\langle t \rangle )+k_1\) since D is a SRHDS for \(\langle x^2,y \rangle \). Since \(\frac{k_2 - 1}{2} = \lambda _2\), we have
as desired. We now look at the odd powers of x in (6.2), which must equal \(\lambda _2 \langle x^2 \rangle x\). We see that
Looking at the first two terms of (6.3), \(D_0 + 1\) is a transversal of \(\langle t \rangle \) in \(\langle x^2 \rangle \), so \((1 + t)\left( D_0 + 1 \right) = \langle x^2 \rangle \) and \((1 + t)\left( D_0 + 1 \right) ^{(-1)} = \langle x^2 \rangle \). So we can reduce (6.3) to
To evaluate the last two terms of (6.3), we note that (6.1) gives us: if \(x^{2i} \in D_1\), then \(x^{-2i-2} \not \in D_1\). Thus \(D_1\) and \(\left( D_1 x^2 \right) ^{(-1)}\) are disjoint, so their sum is \(\langle x^2 \rangle \) since \(|D_1|=4p\). Thus \( \left( D_1 x\right) ^{(-1)} + D_1 x = \left( \left( D_1 x^{-2}\right) ^{(-1)} + D_1\right) x = \langle x^2 \rangle x. \) So the sum of the odd powered terms is
as desired. Therefore we have shown (3), and E is a SRHDS. \(\square \)
Corollary 6.3
The set \(E=E_0+E_1y\) as defined above is an SRHDS in \(\textrm{Dic}_{16p}\) if \(D = D_0 + D_1 y\) is an SRHDS in \(\textrm{Dic}_{8p}\) and \(x^{2i} \in D_1 \) implies \(x^{- 2i - 2} \in D_1\).
Proof
This follows by applying the automorphism \(\varphi (x)=x, \varphi (y)=x^{2p}y\) to \(\textrm{Dic}_{16p}\) in the preceding theorem. We have that D is a SRHDS for \(\textrm{Dic}_{8p}\) if and only if \(\varphi (D)\) is, and similarly E is a SRHDS for \(\textrm{Dic}_{16p}\) if and only if \(\varphi (E)\) is. The condition \(x^{2i} \in \varphi (D_1)\) implies \(x^{- 2i - 2} \in \varphi (D_1)\) is equivalent to the condition \(x^{2i} \in D_1\) implies \(x^{4p-2i-2} \in D_1\). \(\square \)
Many other equivalent symmetries can be obtained by using a different automorphism that fixes \(\langle x \rangle \). The one we have used is that obtained at the end of Theorem 5.1. In the SRHDS \(S=a^{2p}R_1+R_2b\) of \(\textrm{Dic}_{8p}\) from Theorem 5.1, we showed that \(a^i\in R_2\) implies \(a^{-i-1}\in R_2.\) See (5.3). As a subgroup of \(\textrm{Dic}_{16p}\), this is the necessary symmetry condition for Corollary 6.3 to apply. Thus \(\textrm{Dic}_{16p}\) is a SRHDS group when \(4p-1\) is a prime power. This proves Theorem 1.3. \(\square \)
7 D and Cosets of \(Q_8\)
Let G be a SRHDS group with subgroup H and difference set D. Suppose that \(Q\le G\) has even order and that \(g_0=1,\ldots ,g_{p-1}\) is a transversal for \(Q \le G\). Then we can write
Lemma 7.1
Let \(Q\le G\) be as above. For all subsets \(F \subseteq Q\) of size greater than |Q|/2, the multiplicity of t in \(FF^{(-1)}\) is greater than zero. \(\square \)
Proof
Now \(t \in Q\), so \(H \le Q\) and if \(|F| >|Q|/2\), then some coset of \(H \le Q\) meets F in two elements and so \(t \in FF^{(-1)}\). \(\square \)
Now \(DD^{(-1)}=\lambda (G-H)+k\) and a part of the left hand side is \( \sum _{i=0}^{p-1} F_iF_i^{(-1)}\). Thus \(|F_i|\le |Q|/2\) when D is written as in Eq. (7.1).
Now let \(f_i=|F_i|,0\le i<p-1,\) so that
Since \(f_i \le |Q|/2\) we must have \(f_i=|Q|/2\) for all \(0\le i\le p-1\) except one. To see that \(f_0=|Q|/{2}-1\) we just note that \(Q-H\) has \(|Q|-2\) elements that come in inverse pairs. Thus \(f_0=|Q|/{2}-1\).
Next note that \(DD^{(-1)}=\lambda (G-H)+k\) and \( F_iF_i^{(-1)} \subseteq Q\). We want to show
Now, \(v=8p, k=\frac{|Q|}{2}p-1,\lambda =\frac{|Q|}{4}p-1\) and so \(\lambda (Q-H)+k\) has \((\frac{|Q|}{4}p-1)(|Q|-2)+(\frac{|Q|}{2}p-1)=\frac{|Q|^2}{4}p-|Q|+1\) elements, while \(\sum _{i=0}^{p-1} F_iF_i^{(-1)}\) has \( \left( \frac{|Q|}{2}-1\right) ^2+(p-1)\left( \frac{|Q|}{2}\right) ^2=\frac{|Q|^2}{4}p-|Q|+1\) elements, so we must have Eq. (7.2).
For \(Q=Q_8,\) considering those \(F_i\) of size \(|Q|/2=4\) a Magma [12] calculation gives the following result by finding all those subsets \(F \subset Q_8\) such that \(FF^{(-1)}\) does not contain t:
Lemma 7.2
Suppose that \(Q=Q_8\le G.\) Then each \(F_i\) of size 4 is one of the following 16 sets:
Each of these is a relative difference set for \(Q_8\). Thus each \(F_i, i>0,\) is a relative difference set for \(Q_8\). It follows then from Eq. (7.2) that \(F_0\) is a SRHDS for \(Q_8\). Thus \(F_0\) is determined by
Lemma 7.3
The following sets are equal:
-
(i)
The set of all SRHDS for \(Q_8=\langle i,j,k \rangle \).
-
(ii)
The set of all conjugate (by elements of \(Q_8\))-translates (by elements of H) of \(\{i,j,k\}\).
-
(iii)
The set of all \(\{a,b,c\} \subset Q_8 {\setminus } H\) where \(|\{a,b,c\}|=3\) and \(t \notin \{uv^{-1}:u,v \in \{a,b,c\}\}.\) \(\square \)
Call this common set \({\mathcal {S}}\) and note that \(|{\mathcal {S}}|=8\).
Now any \(F_0\) must satisfy (iii), so \(F_0 \in {\mathcal {S}}\). Further, we can choose \(F_0\) to be any element of \({\mathcal {S}}\) by applying the operations in (ii) to D, which still result in a SRHDS.
Assume that \(G=\textrm{Dic}_{8p}\) so that a transversal of \(Q_8\le G\) is \(1,x,\ldots , x^{p-1}\). Now we can write \(D=F_0+F_1x+F_2x^2+\cdots +F_{p-1}x^{p-1}\) where \(F_i \subset Q_8\) and \(F_0\in {\mathcal {S}}\).
Here each \(F_i, i>0,\) is one of the 16 subsets of \(Q_8\) in Lemma 7.2 and \( F_i=(1+x^p)(a+by)=a+by+x^pa+x^pby, \text{ where } a,b \in \langle x^p\rangle . \)
Now \(D^{(-1)}t=D\) and so if \(F_ix^i \subset D\), then \(t(F_ix^i)^{(-1)}=tx^{-i}F_i^{(-1)} \subset D\). Here \( F_i^{(-1)}=a^{-1}+bty+x^{-p}a^{-1}+x^pbty, \) and so
Thus \(F_i\) and \(t(F_ix^i)^{(-1)}\) have \( byx^{i} + x^pbyx^{i}\) in common and so
We denote this by \(J_i(a,b)\), so that D is a union of \(D_0\) and some of the \(J_i(a,b)\).
Now \(J_i(a,b)\) has four elements in \(Q_8x^i\) and has two elements in \(Q_8x^{-i}\). Since we know that each non-trivial coset of \(Q_8\) has to contain four elements of D we know that D has to contain some \(J_{-i}(c,d)\) so that
This is true if and only if we have \( a+x^pa=b^{-1}t+x^{-p}b^{-1}t\) and \( (a^{-1}+x^{-p}a^{-1})t=b+x^pb\). However these equations are equivalent and we note that for any choice of \(a \in \langle x^p\rangle \) there is a \(b \in \langle x^p\rangle \) that solves the first equation.
Thus we now obtain eight element sets by taking the union of these two \(J's\). We denote these by \(L_i(a,b,c)\):
We note that \(L_i(a,b,c)=L_j(a',b',c')\) if and only if \(i=j, a=a', b=b', c=c'.\) For \(1\le i\le p-1\) let \(\mathcal L_i=\{L_i(a,b,c):a,b,c \in \langle x^p\rangle \}\). Then \(|\mathcal L_i|=64\).
8 Groups that are not SRHDS Groups
Proposition 8.1
The dicyclic group \(\textrm{Dic}_{72}\) is not a SRHDS group.
Proof
Suppose it is and that D is the SRHDS. Let \(G=\textrm{Dic}_{72}\) = \(\langle x,y|x^{36}=1,y^2=x^{18}, x^y=x^{-1}\rangle .\) Then by the above section there are \(D_i \in {\mathcal {L}}_i, 1\le i\le 4,\) such that \(D=D_0+\sum _{i=1}^4 D_i\). There are \(64=|L_i|\) choices for each \(D_i, 1\le i\le 4\). Using the standard irreducible representation \(\rho : \textrm{Dic}_{72}\rightarrow \textrm{GL}(2,{\mathbb {C}})\) given by \(\rho (x)=\begin{bmatrix} \zeta _{36}&{}0\\ 0&{}\zeta _{36}^{-1}\end{bmatrix}, \rho (y)=\begin{bmatrix} 0&{}-1\\ 1&{}0\end{bmatrix},\zeta _{36}=e^{2\pi i/36},\) we have \(\rho (G)=\rho (H)=0\). From \(D+D^{(-1)}=G-H\) we then have \(\rho (D)+\rho (D^{(-1)})=0\). By \(DD^{(-1)}=\lambda (G-H)+k\) we have \(\rho (D)\rho (D^{(-1)})=kI_2=35I_2\). Therefore, \(35I_2=\rho (D)\rho (D^{(-1)})= -\rho (D)^2\). A Magma calculation determines that of the \(64^4\) possibilites for D, only 648 have \(\rho (D)^2=-35I_2\). Another Magma [23] calculation verifies that none of these 648 give a SRHDS, completing the proof. \(\square \)
Proposition 8.2
Let G be a group where \(Q_8 \le G\). Suppose that there is an epimorphism \(\pi :G\rightarrow {\mathcal {C}}_p \times Q_8\) for p prime where \(\pi (Q_8)=\{1\} \times Q_8\) and \(|\ker \pi |\) is odd. Then G is not a SRHDS group.
Proof
So suppose that G is a SRHDS group with difference set D and subgroup \(H=\langle t\rangle \). Let \(Q_8=\langle x,y|x^4,x^2=y^2,x^y=x^{-1}\rangle \le G\), so that \(t=x^2, \pi (x)=x, \pi (y)=y\). First note that p must be odd since G has a unique involution. Let \(N=\ker \pi \). Put \({\mathcal {C}}_p =\langle \pi ( r) \rangle , r \in G,\) so that we can write
We note that \(|D_{i,j,k}|\le |N|\).
Let \(p_2=(p-1)/2.\) We can also write \(D =\sum _{i=0}^{p-1} r^iD_i, D_i \subset \langle x,y,N\rangle \) so that
From \(D^{(-1)}=tD\) we get \(D_i^{(-1)}r^{-i}=tr^{p-i}D_{p-i}, 0\le i<p,\) so that \(D_{p-i} = tr^{-p}(D_{i}^{(-1)})^{r^{-i}}.\) Thus \(D=D_0+\sum _{i=1}^{p_2} r^iD_i + r^{-i} t(D_{i}^{(-1)})^{r^{-i}}. \)
Now let \(\rho :Q_8 \rightarrow \textrm{GL}(2,{\mathbb {Q}}( {i})), \textit{i}\) = \(\sqrt{-1},\) be an irreducible faithful unitary representation of \(Q_8\) where \(\rho (x)=\begin{bmatrix} i&{}0\\ 0&{}-i\end{bmatrix}, \rho (y)=\begin{bmatrix} 0&{}-1\\ 1&{}0\end{bmatrix}.\) Then the \(\mathbb Q\)-span of the image of \(\rho \) has basis
since \(\rho (x^2)=-B_1\). We note from Lemma 7.3 that we may assume \(D_0=\{x,y,xy\},\) so \(\rho (D_0)=\begin{bmatrix} i&{}-i-1\\ 1-i&{}-i\end{bmatrix}=B_2+B_3+B_4\).
Let \(\omega =\exp {2 \pi i/p}\). Then \(\pi ,\) \(\rho \) and \(r \mapsto \omega I_2\) determine an irreducible unitary representation of G that we also call \(\rho \). Then \(\rho (r^iD_i)=\omega ^i \sum _{j=1}^4 a_{ij}B_j\), where \(a_{ij} \in {\mathbb {Z}}\), so that
Here \(B_1^*=B_1, B_2^*=-B_2, B_3^*=-B_3, B_4^*= -B_4\).
This gives
We can write this as
From \(DD^{(-1)}=\lambda (G-H)+k\) and \(D^{(-1)}=tD\) we get \(D^2=\lambda (G-H)+kt\). Now if \(\rho (D)^2=(e_{ij})\), then from \((e_{ij})=\rho (D^2)=\rho (\lambda (G-H)+tk)=-kI_2\) and Eq. (8.2) we get
Solving, we must have either
Now we find \(a_1,\cdots ,a_4\) in terms of the \(a_{ij}\). From (8.1) and (8.2) we have
From this we get
Now if we have (i) \(a_1=0\), then \(p>2\) is a prime means that the \(\omega ^i-\omega ^{-i},i=1,2,\cdots ,p_2\) are linearly independent over \({\mathbb {Q}}\), so that we must than have \(a_{i1}=0\) for all i.
Observe from previous definitions that \(a_{i1}=|D_{0,i,0}|-|D_{0,i,2}|.\) From \(D^{(-1)}=tD\) and \(D \cup D^{(-1)}=G-\langle t \rangle \) we have \(|D_{0,i,0}|+|D_{0,i,2}|=|N|\). So \(|D_{0,i,0}|=|D_{0,i,2}|=|N|/2\). Thus |N| is even, which contradicts our assumption on \(\ker \pi \).
So now assume (ii), so that
But \(-\rho (D^2)=\rho (DD^{(-1)})=kI_2\) then gives \(a_1^2=-k.\) Here \(a_1 \in {\mathbb {Q}}[\omega ]\). Recall that \(\omega =e^\frac{2\pi i}{p}\), so the Galois group of \([{\mathbb {Q}}(\omega ):{\mathbb {Q}}]\) is cyclic of even order \(p-1\). By the Galois correspondence, \(\mathbb Q(\omega )\) has a unique quadratic subfield. In particular, we can verify that the subfield is exactly \({\mathbb {Q}}(\sqrt{p})\) if \(p\equiv 1 \pmod {4}\), and \({\mathbb {Q}}(\sqrt{-p})\) if \(p\equiv 3 \pmod {4}\). This follows from the Gauss sum:
Note that \(k \equiv 3 \pmod {4}\) so k is not an integer square. Therefore \(a_1^2=-k\) implies \(k=px^2\) for some \(x \in {\mathbb {Z}}\). However, \(k=4p|N|-1\) so we have a contradiction, as k must be congruent to both 0 and \(-1 \pmod {p}\). \(\square \)
9 Groups of Order Less Than or Equal to 72
Here are the non-dicyclic groups (using magma notation) of order at most 72 that meet the following requirements: (i) they are not abelian; (ii) their Sylow 2-subgroups are generalized quaternion groups; (iii) they have a single involution.
We note that all of the dicyclic groups of order less than 72 and divisible by 8 are SRHDS groups by Theorems 1.2 and 1.3, while \(\textrm{Dic}_{72}\) is not by Proposition 8.1.
We will determine whether the remaining groups have a SRHDS. If they have a SRHDS then we give a SRHDS explicitly. If not, then we give a proof that the group is not a SRHDS group.
In the cases of \(G_{72, 3},\,\,\, G_{72, 11}, \,\,\,G_{72, 24},\,\,\, G_{72,25},\) and \(G_{72,31},\) we use the following process to show they are not SRHDS groups: Given one of the four groups G, we take a right transversal \(g_0=1, \dots , g_8\) for \(Q_8\le G\). Assuming there is an SRHDS D, we write D as in (7.1). We can assume \(F_0=\{x,y,xy\}\) by Lemma 7.3. By Lemma 7.2, there are 16 possibilities for each \(F_i\), and a Magma [23] calculation verifies that none of these combinations give a SRHDS.
-
(1)
\(G_{24,3}=\textrm{SL}(2,3)=\langle a,b,c,d| a^3=1, b^2 = d, c^2 = d, d^2, b^a = c, c^a = b c, c^b = c d\rangle .\) Here \(D=\{a^2 c d, a b c d, a c d, c d, a^2 b d, a^2 d, a^2 b c, a, b c, a b, b \}.\)
-
(2)
\(G_{24,11}={\mathcal {C}}_3 \times Q_8\). This is not a SRHDS group by Proposition 8.2.
-
(3)
\(G_{40,11}={\mathcal {C}}_5 \times Q_8\). This is not a SRHDS group by Proposition 8.2.
-
(4)
\(G_{48,18}={\mathcal {C}}_3 \rtimes \textrm{Dic}_{16}\) \(= \langle a,b,c,d,e|d^2=e^3=1, a^2=b^2=c^2=d, b^a=bc, c^a=c^b=cd, d^a=d^b=d^c=d, e^a=e^2, e^b=e^c=e^d=e\rangle \) and let D be
$$\begin{aligned}&\{ a d e^2, d e^2, a e, e, a b c e^2, a b c, b c e^2, a b d e^2, b d e^2, b c e, a c d, a c d e^2, a b d,\\ {}&c d e^2, c d, a c d e, c d e, b d e, b c d, a, a b c d e, b, a b e\}. \end{aligned}$$ -
(5)
\(G_{48,27}={\mathcal {C}}_3 \times \textrm{Dic}_{16}\). We show \(G_{48,27}\) is not a SRHDS group. Let \({\mathcal {C}}_3=\langle r \rangle \). Then \(D=D_0+D_1r+D_2r^2, D_i \subset \textrm{Dic}_{16}\). Now \(D^{(-1)}=tD\) gives \(D_0^{(-1)}=tD_0\) and \(D_2=tD_1^{(-1)}\). Also Lemma 3.1 shows that the sizes of \(D_0,D_1,D_2\) are 7, 8, 8 (in some order). By replacing D by \(r^iD\) if necessary we may assume that \(|D_0|=7\) and that \(D_0+1,D_1,D_2\) are transversals for G/H. Using \(D_0^{(-1)}=tD_0\) one sees that there are 64 possible \(D_0\)s and 256 possible \(D_1\)s. Further, \(D_2\) is determined by \(D_2=tD_1^{(-1)}\). There are thus \(64 \cdot 256\) possibilities for D and one checks that none of these give a SRHDS.
-
(6)
Let \(G_{48,28} = \langle a,b,c,d,e|b^3=e^2=1, a^2=c^2=d^2=e, b^a=b^2, c^a=d, c^b=de, d^a=c, d^b=cd, d^c=de, e^a=e^b=e^c=e^d=e\rangle \). Here one D is
$$\begin{aligned}&\{ a b^2 d e, a b^2 c d e, b^2 c d e, c e, a b c, b^2 c, b c, d, a d e, a b^2 c e, a c, a b^2, a c d, c d, \\ {}&b^2 d, b^2 e, a b d e, b d e, b c d, a, a b, a b c d e, b\}. \end{aligned}$$ -
(7)
\(G_{72,3} =Q_8 \rtimes {\mathcal {C}}_9=\langle i,j,b|i^4=j^4=b^9=1, i^j=i^{-1}, i^2=j^2, i^b=j, j^b=ij\rangle \). The Magma search described at the beginning of this section shows this is not an SRHDS group.
-
(8)
\(G_{72,11} ={\mathcal {C}}_9 \times Q_8\). The Magma search described at the beginning of this section shows this is not an SRHDS group.
-
(9)
\(G_{72,24} ={\mathcal {C}}_3^2 \rtimes Q_8=\langle a,b,i,j|a^3=b^3=i^4=j^4=1, ab=ba, i^j=i^{-1}, i^2=j^2, a^i=a, b^i=b^2,a^j=a^2, b^j=b\rangle \). The Magma search described at the beginning of this section shows this is not an SRHDS group.
-
(10)
\(G_{72,25} ={\mathcal {C}}_3 \times SL(2,3)\). The Magma search described at the beginning of this section shows this is not an SRHDS group.
-
(11)
\(G_{72,26} ={\mathcal {C}}_3 \times Dic_{24}\). This is not an SRHDS group by Proposition 8.2.
-
(12)
\(G_{72,31} ={\mathcal {C}}_3^2 \rtimes Q_8=\langle a,b,i,j|a^3=b^3=i^4=j^4=1, ab=ba, i^j=i^{-1}, i^2=j^2, a^i=a^2, b^i=b^2,a^j=a, b^j=b\rangle \). The Magma search described at the beginning of this section shows this is not an SRHDS group.
-
(13)
\(G_{72,38} ={\mathcal {C}}_3^2 \times Q_8\). This is not an SRHDS group by Proposition 8.2.
Data Availibility
All data generated or analysed during this study are included in this published article.
References
Chen, Y.Q., Feng, T.: Abelian and non-abelian Paley type group schemes. Preprint
Cohen, H.: A Course in Computational Algebraic Number Theory, GTM, vol. 138. Springer, Berlin (1996)
Ding, C., Yuan, J.: A family of skew Hadamard difference sets. J. Combin. Theory Ser. A 113, 1526–1535 (2006)
Ding, C., Wang, Z., Xiang, Q.: Skew Hadamard difference sets from the Ree-Tits slice symplectic spreads in PG(3,32h+1). J. Combin. Theory Ser. A 114, 867–887 (2007)
Evans, R.J.: Nonexistence of twentieth power residue difference sets. Acta Arith. 84, 397–402 (1999)
Feng, T., Xiang, Q.: Strongly regular graphs from union of cyclotomic classes. arXiv:1010.4107v2. MR2927417
Ikuta, T., Munemasa, A.: Pseudocyclic association schemes and strongly regular graphs. Eur. J. Combin. 31, 1513–1519 (2010)
Coulter, R.S., Gutekunst, T.: Special subsets of difference sets with particular emphasis on skew Hadamard difference sets. Des. Codes Cryptogr. 53(1), 1–12 (2009)
Isaacs, I.: Martin finite group theory. In: Graduate Studies in Mathematics, vol. 92. American Mathematical Society, Providence, pp. xii+350 (2008)
Babai, L., Cameron, P.J.: Automorphisms and enumeration of switching classes of tournaments. Electron. J. Combin. 7, Research Paper 38 (2000)
https://cameroncounts.wordpress.com/2011/06/22/groups-with-unique-involution
Malzan, J.: On groups with a single involution. Pac. J. Math. 57(2), 481–489 (1975)
Malzan, J.: Corrections to: “On groups with a single involution” (Pacific J. Math. 57 (1975), no. 2, 481–489). Pac. J. Math. 67(2), 555 (1976)
Isaacs, I.M.: Real representations of groups with a single involution. Pac. J. Math. 71(2), 463–464 (1977)
Schmidt, B.: Williamson matrices and a conjecture of Ito’s. Des. Codes Cryptogr. 17(1–3), 61–68 (1999)
Ito, N.: On Hadamard groups. III. Kyushu J. Math. 51(2), 369–379 (1997)
Muzychuk, M., Ponomarenko, I.: Schur rings. Eur. J. Combin. 30(6), 1526–1539 (2009)
Schur, I.: Zur Theorie der einfach transitiven Permutationsgruppen, pp. 598–623. Sitz. Preuss. Akad. Wiss, Berlin, Phys-math Klasse (1933)
Wielandt, H.: Finite Permutation Groups. Academic Press, New York–London, pp. x+114 (1964)
Wielandt, H.: Zur theorie der einfach transitiven permutationsgruppen II. Math. Z. 52, 384–393 (1949)
Moore, E.H., Pollatsek, H.S.: Difference sets. Connecting algebra, combinatorics, and geometry. In: Student Mathematical Library, vol. 67, pp. xiv+298. American Mathematical Society, Providence (2013)
Pott, A.: Finite geometry and character theory. In: Lecture Notes in Mathematics, vol. 1601. Springer, Berlin (1995)
Bosma, W., Cannon, J.: MAGMA. University of Sydney, Sydney (1994)
Acknowledgements
All computations made in the preparation of this paper were accomplished using Magma [23]. The first, second, third, and fifth authors thank Brigham Young University Department of Mathematics for funding during the writing of this paper. We are also grateful for useful suggestions from a referee.
Funding
The first, second, third, and fifth authors thank Brigham Young University Department of Mathematics for funding during the writing of this paper.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Anderson, G., Haviland, A., Holmes, M. et al. Difference Sets Disjoint from a Subgroup III: The Skew Relative Cases. Graphs and Combinatorics 39, 67 (2023). https://doi.org/10.1007/s00373-023-02662-8
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00373-023-02662-8