On the Structure of Involutions and Symmetric Spaces of Dihedral Groups

ON THE STRUCTURE OF INVOLUTIONS AND SYMMETRIC SPACES OF DIHEDRAL GROUPS arXiv:1205.3207v1 [math.GR] 14 May 2012 K. K. A. CUNNINGHAM, T. J. EDGAR, A. G. HELMINCK, B. F. JONES, H. OH, R. SCHWELL, J. F. VASQUEZ Abstract. We initiate the study of analogues of symmetric spaces for the family of finite dihedral groups. In particular, we investigate the structure of the automorphism group, characterize the involutions of the automorphism group, and determine the fixed-group and symmetric space of each automorphism. Introduction Let G be a group and θ ∈ Aut(G) such that θm = id. We then define the following two sets: H = Gθ = {g ∈ G | θ(g) = g} Q = {g ∈ G | g = xθ(x)−1 for some x ∈ G}. The set H is the fixed-point subgroup of θ and Q is known as a Generalized Symmetric Space. If θ is an involution and G is a real reductive Lie group, then the set Q is a reductive symmetric space. If G is a reductive algebraic group defined over an algebraically closed field k, then Q is also known as a symmetric variety and if G is defined over a non-algebraically closed field k, then the set Qk := {xθ(x)−1 | x ∈ Gk } is called a symmetric k-variety. Here Gk denotes the set of k-rational points of G. Reductive symmetric spaces and symmetric kvarieties are well known for their role in many areas of mathematics. They are probably best known for their fundamental role in representation theory [5]. The generalized symmetric spaces as defined above are of importance in a number of areas as well, including group theory, number theory, and representation theory. One of the first questions that arises in the study of these generalized symmetric spaces is the classification of the automorphisms up to isomorphy, where isomorphy is given by either conjugation by inner automorphisms, outer automorphisms, or both depending on which makes the most sense. To analyze the structure of these generalized symmetric spaces one looks at orbits of both the group itself and the fixed point group. Both G and H act on Q by θ-twisted conjugation which we denote by ∗. For g ∈ G and q ∈ Q we have g ∗ q = gqθ(g)−1 . Since H is the set of fixed points, this action is simply conjugation when restricted to H. Also, if θ is an involution, then Q ∼ = G/H under the map τ : G → Q given by τ (g) = gθ(g)−1 . We are interested in classifying G and H orbits in Q and describing how the G-orbits decompose into H-orbits. In particular, if θ is an involution it is known that H\Q ∼ = H\G/H. These orbits and double cosets play an important role in representation theory. The authors thank the American Institute of Mathematics in Palo Alto, CA for their generous support. 1 This paper focuses on understanding the structures described above when G is the dihedral group of order 2n, that is G = Dn . In particular, we determine the sets H and Q explicitly, provide a procedure for enumerating the involutions, and give a closed formula for counting the equivalence classes of involutions of Dn . The paper is organized as follows. In Section 1, we introduce the necessary preliminaries including our choices for notation and the well known description of the automorphism group of Dn as well as some relevant examples. In Section 2, we investigate and describe all the automorphisms of Dn of a fixed order. We recall the notion of equivalence of automorphisms and give simple conditions for two automorphisms of Dn to be equivalent. Furthermore, we provide a formula for computing the total number of equivalence classes of automorphisms of a fixed order. In Section 3, we give full descriptions of the sets H and Q as well as the orbits of Q under the action by H and by G. In Section 4, we find stronger results for the involutions in the automorphism group. Our main technical result is to partition the set of involutions in such a way that there are at most two distinct equivalence classes of involutions in each piece of the partition (cf. Theorem 4.5). We also show that in the involution case, Q is always a subgroup of G, and we describe the subgroup structure of Q. Furthermore, we introduce the set of twisted involutions, R, and give a characterization of this set and its relation to the generalized symmetric space. Using our results on involutions of Dn , we provide a counterexample to a standard theorem on equivalence of involutions which holds for algebraic groups. Finally, in Section 5, we complete the discussion by applying our methods to the infinite dihedral group in order to understand the automorphisms of finite order, as well as H and Q in that context. 1. Preliminaries 1.1. The Dihedral Group and its Automorphism Group. Throughout the paper we denote by Dn the group of symmetries of the regular n-gon. More specifically, Dn is a finitely generated group given by the presentation Dn = hr, s | rn = s2 = 1, sr = r−1 si. We use this presentation, instead of the presentation as a Coxeter group (see [2, Example 1.2.7], [6, §1.1, §4.2], [3, §1.2]), because it is convenient for describing the automorphism group of Dn . From the presentation it is clear that Dn = {1, r, r2 , . . . , rn−1 , s, rs, r2 s, . . . , rn−1 s} and we say that an element of Dn is presented in normal form if it is written as rk sm for some integers 0 ≤ k < n and s ∈ {0, 1}. Throughout the paper, Zn denotes the additive group of integers modulo n and Un denotes the multiplicative group of units of Zn . Recall that a ∈ Un if and only if (ȧ, n) = 1 for any representative integer ȧ of a ∈ Zn . We also find it useful to have terminology for the k-th roots of unity in Zn (equivalently in Un ), which can be described by Rkn := {a ∈ Un | ak = 1}. The automorphism group of Dn is well known (see [9, Theorem A]) and is described in the following lemma. 2 Lemma 1.1. The automorphism group of Dn is isomorphic to the group of affine linear transformations of Zn : Aut(Dn ) ∼ = Aff(Zn ) = {ax + b : Zn → Zn | a ∈ Un , b ∈ Zn } and the action of ax + b on elements of Dn in normal form is given by: (ax + b).(rk sm ) = rak+bm sm . Proof. Observe that θ(r) must have order n since r does. The elements of Dn of order n are precisely the {ra | a ∈ Un }. So let θ(r) = ra for some a relatively prime to n. Similarly, θ(s) = rb s for some 0 ≤ b < n since s has order 2 and the only elements of Dn of order 2 are of the form rb s. Putting these together, we see that if θ ∈ Aut(Dn ), then for any rk sm ∈ Dn we must have θ(rk sm ) = rak+bm sm . There is now a clear map from Aut(Dn ) → Aff(Zn ) and it is easily checked to be an isomorphism.  Throughout the paper we abuse notation and write θ = ax + b for elements θ ∈ Aut(Dn ) according to Lemma 1.1. Example 1.2. Consider θ = conj(g), conjugation by an element g ∈ Dn . If θ = ax + b from Lemma 1.1, and if g = rk , then it is easy to see that a = 1 and b = 2k (in Un and Zn respectively). Also, if g = rk s, then we have a = n − 1 and b = 2k. In particular, if θ = ax + b is an inner automorphism, then (abusing notation) we have a ≡ ±1 (mod n). Also, we see that when n is even, there are n distinct inner automorphisms (corresponding to a ≡ ±1 (mod n) and b ∈ h2i ≤ Zn , and when n is odd, there are 2n distinct inner automorphisms corresponding to a ≡ ±1 (mod n) and b ∈ h2i = Zn . Finally, note that the identity automorphism of Dn corresponds to x ∈ Aff(Zn ). Example 1.3. It is well known from the theory of Coxeter groups that when n is even, there is an outer automorphism of Dn corresponding to the interchanging of the two conjugacy classes of reflections, also known as the diagram automorphism (cf. [3, §4.2]). This automorphism is described by θ = ax + b where a = n − 1 and b = n − 1. The previous example shows that this automorphism is inner when n is odd, and outer when n is even. 2. Automorphisms of Dn In this section we describe the action of the automorphism group on Dn , we characterize the automorphisms of fixed order, and we discuss the notion of equivalent automorphisms. For any c ∈ Zn , we define ZDiv(c) = {y ∈ Zn | cy ≡ 0 mod n}. It is trivial to check that ZDiv(c) is a subgroup of Zn . In particular, ZDiv(c) = ker πc where πc : Zn → Zn n given by πc (x) = cx. Note that im(πc ) = hci and so | im(πc )| = gcd(c,n) . Since im(πc ) ∼ = ∼ Zn / ker(πc ) = Zn /ZDiv(c), it follows that |ZDiv(c)| = gcd(c, n). We use this notation to describe the automorphisms of finite order dividing k. Proposition 2.1. Let k ≥ 1 be an integer and θ ∈ Aut(Dn ) with θ = ax + b. Then θk = id if and only if a ∈ Rkn and b ∈ ZDiv(ak−1 + ak−2 + · · · + a + 1). Proof. Straightforward computation in Aut(Dn ) gives us that θk = ak x + (ak−1 + ak−2 + · · · + a + 1)b 3 Since we have identified the identity automorphism with x ∈ Aff(Zn ) the following two equations hold: (2.1) ak ≡ 1 (mod n) (ak−1 + ak−2 + · · · + 1)b ≡ 0 (mod n). In the notation described above, these are evidently equivalent to a ∈ Rkn and b ∈ ZDiv(ak−1 + ak−2 + · · · + a + 1).  In what follows, we let Autk (Dn ) = {θ ∈ Aut(Dn ) | θk = id} ⊆ Aut(Dn ). Proposition 2.2. For any n and k ≥ 1, we have X | Autk (Dn )| = gcd(ak−1 + ak−2 + · · · + a + 1, n). a∈Rkn Proof. This follows from Proposition 2.1: for any a ∈ Rkn there are |ZDiv(ak−1 + ak−2 + · · · + 1)| = gcd(ak−1 + ak−2 + · · · + 1, n) elements b such that (ak−1 + ak−2 + · · · + 1)b ≡ 0 (mod n), and every automorphism θ ∈ Autk (Dn ) must be of this form.  Definition 2.3. Let θ1 , θ2 ∈ Aut(Dn ). We say that θ1 is equivalent to θ2 and write θ1 ∼ θ2 if and only if they are conjugate to each other, i.e. if there is σ ∈ Aut(Dn ) with σθ1 σ −1 = θ2 . For any θ ∈ Aut(Dn ) we let θ = {σ ∈ Aut(Dn ) | θ ∼ σ} be the equivalence class of θ. Remark 2.4. This definition of equivalence is broad (since we allow conjugation by any automorphism, not just the inner ones), but still useful. In particular, it simplifies the statement of the following proposition and it allows us not to worry about the parity of n in several places. Proposition 2.5. Let θ1 = ax + b ∈ Aut(Dn ) and θ2 = cx + d ∈ Aut(Dn ). Then θ1 ∼ θ2 if and only if a = c and f b − d ∈ ha − 1i ≤ Zn for some f ∈ Un . Proof. Suppose that σ = f x + g ∈ Aut(Dn ) (so that f ∈ Un and g ∈ Zn ). It is easily checked that σ −1 = f −1 x − f −1 g where f −1 ∈ Zn since f ∈ Un . Then σθ1 σ −1 = τ , where τ = f (a(f −1 x − f −1 g) + b) + g = ax + f b − g(a − 1). Thus, we can conclude that σθ1 σ −1 = θ2 if and only if a = c and f b − g(a − 1) = d. The second equation can be written as f b − d = g(a − 1), and thus f b − d ∈ ha − 1i.  The previous proposition allows us to describe the conjugacy classes of automorphisms with order dividing k. Proposition 2.6. Suppose n is fixed and let k ≥ 1. (1) For any a ∈ Rkn , ha − 1i ≤ ZDiv(ak−1 + ak−2 + · · · + a + 1). (2) For any a ∈ Rkn , Un acts on the cosets ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i. (3) The set Autk (Dn ) is partitioned into equivalence classes indexed by pairs (a, B) where a ∈ Rkn and B is an orbit of Un on ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i. Proof. Part (1) is trivial since (a − 1)(ak−1 + ak−2 + · · · + a + 1) = ak − 1 ≡ 0. For part (2), we simply recognize that Un = Aut(Zn ) acts on Zn by multiplication and since every subgroup is cyclic, Un must stabilize the subgroups of Zn . Finally, part (3) is simply Proposition 2.5 in terms of the Un -action on ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i.  4 Proposition 2.5 and Proposition 2.6 help us determine the total number of equivalence classes of finite order automorphisms. For fixed a ∈ Rkn , we want to compute Na := |{θ | θ = ax + b ∈ Autk (Dn )}|. (2.2) According to Proposition 2.6, given a ∈ Rkn , computing Na amounts to counting the number of orbits of the Un -action on ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i. Theorem 2.7. Let n be fixed and let a ∈ Rkn . Then, the number of orbits of Un on ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i is equal to the number of divisors of gcd(a − 1, n) gcd(ak−1 + ak−2 + · · · + a + 1, n) . n Proof. The Un -orbits on Zn are indexed by the subgroups of Zn . Thus, the Un -orbits on ZDiv(ak−1 + ak−2 + · · · + a + 1)/ha − 1i are indexed by subgroups L ≤ Zn such that ha − 1i ≤ L ≤ ZDiv(ak−1 + ak−2 + · · · + a + 1). It is well known that the subgroup lattice of Zn is isomorphic to the divisor lattice of n. Under the previous lattice isomorphism, the subgroup ha − 1i corresponds to the divisor gcd(a − 1, n) and the subgroup ZDiv(ak−1 + ak−2 + · · · + a + 1) corresponds to the divisor n , k−1 gcd(a + · · · + a + 1, n) and the subgroups between these groups correspond to the divisors of n between gcd(a−1, n) n and gcd(ak−1 +···+a+a,n) in the divisor lattice. Finally, it is known that this sub-lattice of the divisors of n is isomorphic to the divisor lattice of gcd(a − 1, n) n gcd(ak−1 +···+a+1,n) = gcd(a − 1, n) gcd(ak−1 + ak−2 + · · · + a + 1, n) . n  Given a specific n, Theorem 2.7 and Proposition 2.6 allow us to compute all the equivalence classes of automorphisms of order k using any standard computer algebra package. In section 4, we investigate Na when a ∈ R2n (i.e. θ is an involution). 3. Fixed Groups and Symmetric Spaces of Automorphisms Recall from the introduction that we are interested in two different subsets of a group G. Namely, given an automorphism, θ, we want to compute Hθ = Gθ = {x ∈ G | θ(x) = x} and Qθ = {g ∈ G | g = xθ(x)−1 for some x ∈ G}. When θ is understood to be fixed, we drop the subscript from our notation. The following theorem characterizes these spaces in the case of dihedral groups. Theorem 3.1. Let G = Dn and θ = ax + b ∈ Aut(Dn ) be of finite order, and let H and Q be as defined above. Then H = {rk | k(a − 1) ≡ 0 (mod n)} ∪ {rk s | k(a − 1) ≡ −b (mod n)} and Q = {rk | k ∈ ha − 1i ∪ (−b + ha − 1i)}. 5 Proof. Recall that if θ = ax+b ∈ Aut(Dn ), then the formula θ(rk sm ) = rak+bm sm provides the full description of the action of θ. All of the results of the theorem arise from the definitions of H and Q using the action described above by θ. We demonstrate the computation for H in what follows, and Q is similar. Suppose θ(rk ) = rk , then rak = rk and so ak − k ≡ 0 (mod n), i.e. k(a − 1) ≡ 0 (mod n). On the other hand, if θ(rk s) = rk s, then rak+b s = rk s and so ak − k + b ≡ 0 (mod n), i.e. k(a − 1) ≡ −b (mod n).  Using the descriptions of H and Q, we obtain further results as well. Corollary 3.2. Let θ = ax + b ∈ Aut(Dn ) be of finite order. Then (1) If b 6∈ ha − 1i, then H ∼ = ZDiv(a − 1) is cyclic. (2) If b ∈ ha − 1i, then H ∼ = ZDiv(a − 1) o Z2 is dihedral. Corollary 3.3. Let n, k, and θ = ax + b ∈ Autk (Dn ) be fixed. If b ∈ ha − 1i then Q is a subgroup of Dn and Q ∼ = ha − 1i. Proof. Since b ∈ ha − 1i, then it follows from Theorem 3.1 that Q = {rk | k ∈ ha − 1i}.  We will show in the next section that when θ is an involution, Q is always a subgroup of Dn (see Corollary 4.8). Corollary 3.4. Let θ = ax + b ∈ Aut(Dn ) be a fixed automorphism. If b 6∈ ha − 1i, then HQ 6= Dn . If b ∈ ha − 1i, then the following are equivalent (1) HQ = Dn (2) H ∩ Q = {1} in Dn , and n (3) gcd(a − 1, n) is relatively prime to gcd(a−1,n) . Proof. From Corollary 3.2 and the fact that Q ⊆ hri, we see that if b 6∈ ha − 1i then HQ ⊆ hri 6= Dn . If b ∈ ha − 1i then H contains a reflection and so HQ = Dn if and only if r ∈ HQ. From Corollaries 3.2 and 3.3, we see that r ∈ HQ if and only if the subgroups ZDiv(a − 1) and ha − 1i in Zn have relatively prime generators. This last condition is clearly equivalent to both (2) and (3).  Example 3.5. Let G = D36 . We illustrate Theorem 3.1 and its corollaries for three different automorphisms of D36 . First, consider the automorphism θ1 = 19x+18. Note that θ1 is an involution. Applying Theorem 3.1 yields that for θ1 , H = H1 = {1, r2 , r4 , . . . , r34 , r3 s, r5 s, . . . , r35 s} and Q = Q1 = {1, r18 }, which gives that H1 ∩ Q1 = {1, r18 } = Q1 . Observe that the statement of Corollary 3.3 holds in this case: H1 Q1 6= D36 . This follows from the fact that 18 ∈ h18i, but gcd(18, 36) = 18 is not relatively prime to 36/ gcd(18, 36) = 2. Now consider θ2 = 5x + 2 which is an automorphism of order 6. In this case, H2 = {1, r9 , r18 , r27 } and Q2 = {1, r2 , r4 , . . . , r34 }. Thus, H2 ∩ Q2 = {1, r18 } and H2 Q2 = D36 . However, in this case it is because 2 6∈ h4i. Finally, consider the automorphism θ3 = 5x+4 of D36 . This is also an automorphism of order 6. In this case, H3 = {1, r9 , r18 , r27 , r8 s, r17 s, r26 s, r35 s} and Q3 = {1, r4 , r8 , r12 , . . . , r32 }. Thus, H3 ∩ Q3 = {1} and this agrees with Corollary 3.3 since r9 r28 = r and so H3 Q3 = D36 . From Theorem 3.1, we see that H is the disjoint union of {rk | k(a − 1) ≡ 0 (mod n)} and {rk s | k(a − 1) ≡ −b (mod n)}. Notice that the second set may be empty if there is no solution, i, to the equation i(a − 1) ≡ −b (mod n) for fixed a and b (i.e. if b 6∈ ha − 1i). In fact, the two possibilities for the H-orbits on Q are determined by the existence of such a solution. 6 Proposition 3.6. Let G = Dn and θ = ax + b ∈ Autk (Dn ) for some k. (1) If b 6∈ ha − 1i, then the H-orbits on Q are:  H\Q = {rj } | j ∈ ha − 1i ∪ (−b + ha − 1i) . (2) If b ∈ ha − 1i, then the H-orbits on Q are:  H\Q = {rj , r−j } | j ∈ ha − 1i . In either situation G\Q = {Q}, i.e. there is a single G-orbit on Q. Proof. Since H is fixed by θ, the action of H on Q is simply by conjugation. We note that Q ⊆ hri ≤ Dn and so we only need to describe the action of H on a general element, rj . Then, let ri ∈ {rk | k(a−1) ≡ 0 (mod n)} ⊆ H. We see ri rj r−i = rj so hri fixes Q pointwise. Now suppose ri s ∈ {rk s | k(a − 1) ≡ −b (mod n)} ⊆ H. We have ri srj (ri s)−1 = r−j and so ri s takes rj to r−j . The result now follows since we will have {rj , r−j } as an orbit if and only if there is ri s ∈ H, which is true if and only if b ∈ ha − 1i. Finally, we demonstrate that every element of Q is in the G-orbit of 1 ∈ Q. Notice that every element of Q is either of the form q1 = rp(a−1) ∈ Q or q2 = r−b+p(a−1) ∈ Q. We have r−p ∈ G and r−p 1θ(r−p )−1 = rp(a−1) = q1 . Additionally, we have r−p s ∈ G and r−p s1θ(r−p s)−1 = r−p rap−b = r−b+p(a−1) = q2 . Therefore, for any q ∈ Q, we can find g ∈ G such that g1θ(g)−1 = q and so G ∗ 1 = Q. Hence G\Q = {Q}.  Example 3.7. Revisiting θ1 , θ2 , and θ3 of Example 3.5, we apply Proposition 3.6 to obtain that for θ1 , since 18 ∈ h18i and r18 = r−18 ∈ D36 , H1 \Q1 = {{1}, {r18 }}; for θ2 , since 2 6∈ h4i, H2 \Q2 = {{1}, {r2 }, {r4 }, . . . , {r34 }}; lastly, for θ3 , since 4 ∈ h4i, H3 \Q3 = {{1}, {r4 , r−4 }, {r8 , r−8 }, . . . , {r32 , r−32 }}. All of the results in this section hold for automorphism of arbitrary finite order k, but in the next section we demonstrate that we can say more when we restrict our attention to k = 2, the involutions. In Section 5, we describe which of these results hold in the infinite dihedral group. 4. Involutions in Aut(Dn ) In this section, we utilize the results from sections 2 and 3 and expand on them in the situation when θ is an involution in Aut(Dn ). For this entire section, we assume that θ = ax + b ∈ Aut(Dn ) and θ2 = id. In particular, Proposition 2.1 forces a ∈ R2n and b ∈ ZDiv(a + 1), i.e. (a+1)b ≡ 0 (mod n). The discussion prior to Proposition 2.1 describes how to determine the total number of involutions in Aut(Dn ). Recall that Aut2 (Dn ) := {θ ∈ Aut(Dn ) | θ2 = id} is the collection of involutions along with the identity automorphism. For our discussion of the involutions in Aut(Dn ), it is necessary to understand the structure of R2n , the square roots of unity in Zn . The following results can be found in an elementary number theory text (e.g. [7, Example 3.18]) but we include them here for completeness and for the usefulness of the proof. 7 600 168 =23 ·3 ·7 500 120 =23 ·3 ·5 400 300 200 100 24 =23 ·3 50 100 150 200 Figure 1. The number of involutions in Aut(Dn ) for n ≤ 200. The dashed lines represent the lines with slope 1, 2, and 3. Theorem 4.1. Suppose that n ≥ 1 and n = primes. Then,  k  2 |R2n | = 2k+1  2k+2 2m pr11 · · · prkk where the pi are distinct odd if m = 0, 1 if m = 2 if m ≥ 3. Remark 4.2. We include the proof of Theorem 4.1 in Appendix A because it demonstrates exactly how to construct the square roots of unity in Zn provided we have the prime factorization of n. The construction only involves the Euclidean Algorithm and a map provided by the Chinese Remainder Theorem. One can use the procedure described in the proof to effectively compute the square roots of unity in Zn for any n. The construction of the square roots of unity provided in the appendix also provides the understanding necessary for Corollary 4.6 below. The following corollary is the statement of Proposition 2.2 for involutions. Corollary 4.3. For any n ≥ 1, | Aut2 (Dn )| = X gcd(a + 1, n). a∈R2n Remark 4.4. Corollary 4.3 together with the proof of Theorem 4.1 (see Remark 4.2) gives an easy way to compute the total number of involutions in Aut(Dn ). Figure 1 provides a graph of the number of involutions in Dn for n ≤ 200. This plot was generated using Sage [8]. Now we proceed to determine the equivalence classes of involutions. Let θ1 = ax + b and θ2 = cx + d. We know from Proposition 2.5 that if θ1 ∼ θ2 , then a = c. Thus, to 8 determine the equivalence classes of involutions it suffices to fix an a ∈ R2n and determine when ax+b ∼ ax+d. Furthermore, according to Proposition 2.6, with a fixed, we simply need to describe the action of Un on ZDiv(a + 1)/ha − 1i in order to describe all the equivalence classes. Recall that we denote the equivalence class of an involution θ by θ and the number of equivalence classes with fixed leading coefficient a by:  Na := θ | θ = ax + b ∈ Aut2 (Dn ) . Theorem 4.5. Let a ∈ R2n . The following hold. (1) ha − 1i ≤ ZDiv(a + 1) and |ZDiv(a + 1)/ha − 1i| ≤ 2. (2) Na = |ZDiv(a + 1)/ha − 1i| ≤ 2. (3) Na = 2 if and only if n = 2m · k where k ≥ 1 is odd, m > 0, and a ≡ ±1 (mod 2m ). Proof. The first part of (1) is simply a restatement of Proposition 2.6 part (1). Now, suppose n and we have that |ZDiv(a + 1)/ha − 1i| = j. It is well known that |ha − 1i| = gcd(a−1,n) already observed that |ZDiv(a + 1)| = gcd(a + 1, n). Thus we have that gcd(a+1,n) n gcd(a−1,n) = j, or gcd(a + 1, n) gcd(a − 1, n) = jn. Now, suppose that n = 2m pr11 · · · prkk with m ≥ 0, the pi distinct odd primes, and ri ≥ 1. Then, we must have that 0 gcd(a + 1, n) = 2m ps11 · · · pskk with m0 ≤ m and 0 ≤ si ≤ ri for all i, and 00 gcd(a − 1, n) = 2m pt11 · · · ptkk with m00 ≤ m and 0 ≤ ti ≤ ri for all i. Now, we note that for all i ∈ {1, ..., k} either si = 0 or ti = 0. Indeed, if si > 0 and ti > 0 then pi divides both a − 1 and a + 1 which is impossible (since pi > 2). Similarly, either 0 00 m0 ≤ 1 or m00 ≤ 1; otherwise, 2min{m ,m } divides a − 1 and a + 1 which is impossible. Since gcd(a + 1, n) gcd(a − 1, n) = jn, we have that 0 00 2m +m p1s1 +t1 · · · pskk +tk = j2m pr11 · · · prkk . Now, since for all i either ti = 0 or si = 0, we necessarily have si + ti = ri . Dividing out, 0 00 we get 2m +m = j2m . We also know that m0 ≤ m and m00 ≤ m and since either m0 ≤ 1 0 00 or m00 ≤ 1, we have m ≤ m0 + m00 ≤ m + 1. Dividing out we have j = 2m +m −m and so 1 ≤ j ≤ 2 as required. Now, according to Theorem 2.7, Na is the number of divisors of gcd(a−1,n)ngcd(a+1,n) , and thus Na is the number of divisors of j computed above. Therefore, Na = 1 if j = 1 and Na = 2 if j = 2. In either case, Na = |ZDiv(a + 1)/ha − 1i| ≤ 2 proving Claim (2). Finally, according to the proof of (1), we have j = 2 if and only if m0 = 1 and m00 = m or m0 = m and m00 = 1. In the first case, we have that 2m divides a − 1, and in the second case we have that 2m divides a + 1. Therefore, a ≡ ±1 (mod 2m ).  Theorem 4.5, Theorem 4.1 and Remark 4.2 combined together allow us to compute the number of distinct equivalence classes of involutions in Aut2 (Dn ) for fixed n if we have the prime factorization of n. In fact, we get closed formula for the number of equivalence classes elements in Aut2 (Dn ) as follows. 9 Cn 12 10 m =0 m =1 m =2 m =3 m =4 m =5 8 6 4 2 5 10 15 20 25 30 35 40 45 50 n Figure 2. Number of equivalence class of involutions in Aut(Dn ) vs. n for n ≤ 50. Corollary 4.6. Suppose that n ≥ 1 and n = 2m pr11 · · · prkk where the pi are distinct odd primes. Then, the number of equivalence classes, Cn , of Aut2 (Dn ) is given by  2k if m = 0    2k+1 if m = 1 Cn = k+2  2 if m = 2    k+3 k+1 2 −2 if m ≥ 3. Proof. According to Theorem 4.5, we know Na for every a ∈ R2n . When m = 0, n is odd and we know that Na = 1 for all a. Thus, there is one equivalence class for every element of R2n proving the first formula. If m = 1, 2, then every a ∈ R2n is odd and so satisfies a ≡ ±1 (mod m). Thus for every element a ∈ R2n , we get two equivalence classes for a total of 2 · |R2n | equivalence classes. Finally, when m ≥ 3, using the construction of the elements of R2n found in the proof in the appendix, we see that exactly half of these elements satisfy the condition a ≡ ±1 (mod m). Therefore, for half the elements in R2n we have Na = 1 and for the other half we have Na = 2 giving us a total of 2k+2 + (2k+2 − 21 · 2k+2 ) = 2k+3 − 2k+1 equivalence classes, proving the result.  Remark 4.7. According to the Online Encyclopedia of Integer Sequences, the sequence described in Corollary 4.6 also counts the number (up to isomorphism) of groups of order 2n 10 that have a subgroup isomorphic to Zn [1]. To be more precise, we include a brief discussion of the natural bijection between these two objects. Suppose θ = ax+b is an involution. Then wa : Z2 → Aut(Zn ) = Un given by wa (0) = 1 and wa (1) = −a is an injective homomorphism and induces a Z2 -action on Zn . Next, we define f : Z2 × Z2 → Zn by f (c, d) = 0 if c = 0 or d = 0 and f (1, 1) = b; it is easily shown that f is a 2-cocycle if and only if b ∈ ZDiv(a + 1). Finally, it is well known (see [10, Chapter 6.6])) that 2-cocyles give rise to group extensions and two different 2-cocycles give isomorphic group extensions if they are cohomologous. A simple calculation shows that if θ = ax + b and θ0 = ax + b0 , then the corresponding 2cocycles are cohomologous if and only if b − b0 ∈ ha − 1i, which matches the characterization of automorphisms being equivalent above. We now use our previous results from Section 3 to fully describe the sets Hθ and Qθ when θ is an involution; moreover, in this situation, we are also interested in the set of twisted involutions R = Rθ = {x ∈ G | θ(x) = x−1 }. If θ = ax + b is an involution, a quick calculation gives (4.1) R = {rk | k ∈ ZDiv(a + 1)} ∪ {rk s | k(a − 1) ≡ −b (mod n)}. Following the results from Theorem 3.1 we describe Q and R \ Q. Corollary 4.8. If θ = ax + b is an involution, then Q is a subgroup of hri and under the ∼ = natural isomorphism ψ : hri −→ Zn we have ( ha − 1i, if b ∈ ha − 1i ψ(Q) = ZDiv(a + 1), otherwise. Furthermore, R \ Q = {rk | k ∈ ZDiv(a + 1) \ ha − 1i} ∪ {rk s | k(a − 1) ≡ −b (mod n)} Proof. We proceed using Theorem 3.1 and applying Theorem 4.5. Indeed, by Theorem 3.1 ψ(Q) consists of the union of the two cosets ha−1i and −b+ha−1i. These cosets are the same if −b ∈ ha−1i (which means b ∈ ha−1i) and then ψ(Q) = ha−1i ≤ ZDiv(a + 1), or they are distinct and by Theorem 4.5 part (1) there are only 2 cosets total so ha−1i∪(−b+ha−1i) = ZDiv(a + 1). In either case, Q is a subgroup of hri and ψ(Q) is as described in the statement of the corollary. Finally, the result about R \ Q is immediate from equation (4.1).  Corollary 4.9. If b 6∈ ha − 1i then R = Q. Remark 4.10. We note that due to Theorem 4.5, ψ(Q) given in Corollary 4.8 is almost always ZDiv(a + 1) for involutions. The only instance where ψ(Q) 6= ZDiv(a + 1) occurs when |ZDiv(a + 1)/ha − 1i| = 2 and b ∈ ha − 1i. We revisit θ1 = 19x + 18 from Example 3.5, which is an involution, noting that this is in fact a case where b ∈ ha−1i. Recalling that Q1 = {1, r18 }, we see that ψ(Q1 ) is indeed isomorphic to ha − 1i = h18i < Z36 . We also have that R = R1 = {1, r, r3 , r5 , . . . , r17 , r18 , r19 , r21 , . . . , r35 }, and thus R1 \ Q1 = {r, r3 , r5 , . . . , r35 }. Corollary 4.8 is interesting because in the setting of algebraic groups, the symmetric space Q is almost never a subgroup; whereas here Q is always a subgroup. In the setting of algebraic groups, it is known that involutions θ1 ∼ θ2 if and only if Hθ1 ∼ = Hθ2 (see [4]). Now we show that this result does not hold for finite groups. Corollary 4.11. There exists n and involutions θ1 , θ2 ∈ Aut(Dn ) such that Hθ1 = Hθ2 but θ1 6∼ θ2 . 11 Proof. Let n = 8 and θ1 = 7x and θ2 = 3x. According to Proposition 2.5, θ1 6∼ θ2 because 7 6= 3 (in U8 ). However, according to Theorem 3.1, Hθ1 = {1, r4 , s, r4 s} and Hθ2 = {1, r4 , s, r4 s}.  5. The Infinite Dihedral Group D∞ So far we have discussed the finite dihedral groups Dn . However, it turns out that there are similar results for the infinite dihedral group D∞ = hr, s | s2 = 1, rs = sr−1 i. In this case, the automorphisms are the affine linear transformations of Z, so are of the form ax + b where b ∈ Z and a ∈ {±1}. Then, the congruences given in equation (2.1) become equations over the integers. In particular, it easy to show that the only automorphisms of finite order, besides the identity, are the involutions and they have the form −x + b where b ∈ Z. Proposition 5.1. The following hold: (1) If θ ∈ Aut(D∞ ) has finite order, then θ = −x + b for some integer b. (2) −x + b ∼ −x + d if and only if b ≡ d (mod 2). Proof. Part (1) follows from the equations (2.1) given in the proof of Proposition 2.1. For part (2), observe that Proposition 2.5 holds for D∞ and says that −x + b ∼ −x + d if and only if either b − d ∈ 2Z or −b − d ∈ 2Z, equivalently: b ≡ d (mod 2).  We see that in D∞ the situation is simple: the only automorphisms of finite order are involutions and there are only two distinct equivalence classes of involutions represented by χ0 = −x = conj(s) (inner) and χ1 = −x + 1 (outer). Note that χ1 represents the class of the diagram automorphism discussed in Example 1.3. The fixed group and symmetric space of an involution θ = χi is similarly easy to compute. (5.1) Hχ0 = {1, s}, Qχ0 = hr2 i, Rχ0 = hr2 i ∪ {s} Hχ1 = {1}, Qχ1 = hri, Rχ1 = hri. We note that the descriptions in equation (5.1) show that Corollary 4.8 holds in the infinite case as well. However, in this case, the description is simpler because the description of Q depends only on whether b is even or odd. We also note that despite the fact that the two cases are different when viewing Q as a subgroup of D∞ , we always have Q ∼ = Z. Appendix A. Proof of Theorem 4.1 Proof. Our goal is to determine the size of |R2n | for all n. First, suppose that n = pr where p is an odd prime and r ≥ 1. Then, suppose a ∈ Zpr and that a2 − 1 ≡ 0 (mod pr ). Then we have that (a + 1)(a − 1) ≡ 0 (mod pr ) and so (a + 1)(a − 1) = kpr for some k ∈ Z. Since p is an odd prime, it is clear that both a + 1 and a − 1 cannot divide pd for any d ≥ 1 and so we have that a+1 = pr or a−1 = pr . This implies that a = 1 or a = n−1, thus |R2pr | = 2. Next, it is clear that |R22 | = 1 and |R24 | = 2. Suppose that n = 2m with m ≥ 3. Next, suppose that a ∈ Zn \ {1, n − 1} with a2 ≡ 1 (mod n). Since n is even, we must have that a is odd, so assume that a = 2k + 1 for some 0 ≤ k ≤ 2m−1 . Then 1 ≡ a2 ≡ (2k + 1)2 ≡ 4k 2 + 4k + 1 (mod n) so that 4k(k + 1) = l · 2m for some l ≥ 1. In particular, since m ≥ 3, we have k(k + 1) = l · 2m−2 for some m. Either k or k + 1 is odd and thus cannot divide 2m−2 . 12 Case 1: k is even. So k = h · 2m−2 for some 1 ≤ h ≤ 2 (since k ≤ 2m−1 ). Then, if h = 1, a = 2m−1 + 1. If h = 2, then a ≡ 1 (mod n). Case 2: k is odd. So k + 1 is even and k + 1 = h · 2m−2 with 1 ≤ h ≤ 2. Then, if h = 1, a = 2m−1 − 1. If h = 2, then a = 2m − 1 ≡ n − 1 ≡ −1 (mod n). So, there are four possibilities for a: 1, −1, 2m−1 + 1, 2m−1 − 1. It is easy to check that these are all in fact square roots of 1, showing that |R22m | = 4 if m ≥ 3. Finally, the result holds due to the fact that the Chinese Remainder Theorem guarantees that if n = 2m pr11 · · · prkk then Zn ∼ = Z2m × Z r1 × · · · × Z rk . p1 pk Then, it is clear that any square root of (1, ..., 1) (the identity from the right hand side) must be of the form (a0 , a1 , ..., ak ) where a0 ∈ R22m and ai ∈ Rnpri for i ∈ {1, ..., k}. We i have counted the possibilities for ai above and so we can multiply these together to get all the possible choices for a ∈ Zn with a2 ≡ 1 (mod n). In particular, we get |R2n | = 2k if  m ∈ {0, 1}, |R2n | = 2 · 2k if m = 2 and |R2n | = 22 · 2k if m ≥ 3 as required. References 1. OEIS Foundation Inc. (2012), The On-Line Encyclopedia of Integer Sequences, http://oeis.org/A147848. 2. Anders Björner and Francesco Brenti, Combinatorics of Coxeter groups, Graduate Texts in Mathematics, vol. 231, Springer, New York, 2005. 3. Nicolas Bourbaki, Lie groups and Lie algebras. Chapters 4–6, Elements of Mathematics (Berlin), Springer-Verlag, Berlin, 2002, Translated from the 1968 French original by Andrew Pressley. 4. A. G. Helminck and S. P. Wang, On rationality properties of involutions of reductive groups, Adv. in Math. 99 (1993), 26–96. 5. A.G. Helminck, Symmetric k-varieties, Proc. Sympos. Pure Math. 56 (1994), no. 1, 233–279. 6. James E. Humphreys, Reflection groups and Coxeter groups, Cambridge Studies in Advanced Mathematics, vol. 29, Cambridge University Press, Cambridge, 1990. 7. Gareth A. Jones and J. Mary Jones, Elementary number theory, Springer Undergraduate Mathematics Series, Springer-Verlag London Ltd., London, 1998. 8. W. A. Stein et al., Sage Mathematics Software (Version 4.8), The Sage Development Team, 2012, http://www.sagemath.org. 9. Gary L. Walls, Automorphism groups, Amer. Math. Monthly 93 (1986), no. 6, 459–462. 10. Charles A. Weibel, An introduction to homological algebra, Cambridge Studies in Advanced Mathematics, vol. 38, Cambridge University Press, Cambridge, 1994. MR 1269324 (95f:18001) 13 View publication stats