Access the full text.

Sign up today, get DeepDyve free for 14 days.

Mathematics
, Volume 2020 (1612) – Dec 5, 2016

/lp/arxiv-cornell-university/the-threshold-for-subgroup-profiles-to-agree-is-omega-log-n-mOgjQNpLVP

- ISSN
- 1557-2862
- eISSN
- ARCH-3343
- DOI
- 10.4086/toc.2019.v015a019
- Publisher site
- See Article on Publisher Site

THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). JAMES B. WILSON e−3 Abstract. For primes p, e > 2 there are at least p /e groups of 2e+2 order p that have equal multisets of isomorphism types of proper subgroups and proper quotient groups, isomorphic character tables, and power maps. This obstructs recent speculation concerning a path to- wards eﬃcient isomorphism tests for general ﬁnite groups. These groups have a special purpose polylogarithmic-time isomorphism test. 1. Introduction A recent breakthrough result by Babai has pushed the complexity of isomorphism testing of ﬁnite graphs on n vertices to an upper bound of O((log n) )) n for some c ≥ 1 [1]. This brings the complexity of graph iso- morphism within range of the present complexity for isomorphism testing O(µ (n)) of groups of order n. That complexity is bounded by n where given 1 s the prime factorization n = p ··· p , e e 1 s µ (p ··· p ) = max{e , . . . , e }. 1 s 1 s Pultr and Hedrl´ın [15] constructed reductions that imply that group isomor- phism reduces to graph isomorphism in time polynomial in n; see also [14]. When µ (n) is bounded then group isomorphism is in polynomial time in n. For each c > 1, as n → ∞, the number of integers n for which µ (n) ≤ c tends to 1/ζ(c), e.g. 60% of integers are square-free and 99% have µ (n) ≤ 8. Yet ℓ O(log n) for n = p the group isomorphism problem has the complexity of n which makes it an obstacle to the improvement of graph isomorphism. In that vein recent speculation by Gowers [10] and Babai [1, p. 81] has revisited the idea of using a portion of the subgroups of a ﬁnite group to determine isomorphism types of ﬁnite groups. Algorithms for testing isomor- phism have been successfully using such ideas as heuristics for some time; cf. ﬁngerprinting in [8]. Yet, proving eﬃciency based on these heuristics has been obstructed by knowledge of examples of Rottla¨nder, and others, which show that lattices are not enough to characterize isomorphism [16]. Circumventing existing counter-examples, Gowers introduced a threshold criterion. In [10] he asked if as the number d(G) = min{d : G = hx , . . . , x i} 1 d Date : October 17, 2018. Key words and phrases. group isomorphism, proﬁles, p-groups. This research was supported in part by NSF grant DMS 1620454. arXiv:1612.01444v2 [math.GR] 20 Dec 2016 2 JAMES B. WILSON grows toward the upper bound of log |G| (actually µ (|G|) + 1 [11]), is the isomorphism type of G determined by the subgroups that are k-generated, for a k much smaller than d(|G|)? If true it would improve isomorphism test- 2k ing to O(|G| ) steps. Glauberman-Grabowski [9] gave examples G where k ≥ 2 log |G| − 5/2. We will give examples of groups G of odd order |G| = p for which we need k = ℓ−2. So in general we need k ≥ log |G|−2. Fix primes p, e > 2. The Heisenberg group over a ﬁeld F e of order p is: 1 α γ e 1 β e H = H(F ) = : α, β, γ ∈ F . p p By the subgroup proﬁle of a group G we mean the partition of the proper subgroups of G into isomorphism classes. Likewise deﬁne the quotient-group proﬁle. We prove: ′ 2e+2 Theorem 1.1. The groups N ≤ H of index p in H are normal in H and each H/N has the same subgroup and quotient-group proﬁle and has d(H/N) = 2e. Yet, ′ 2e+2 G = {H/N : N ≤ H ,|H : N| = p } p,e e−3 has at least p /e isomorphism classes. 1.1. Further invariants. More can be said about the similarities in the groups of Theorem 1.1. Brauer had asked if non-isomorphic groups could have isomorphic character tables together with exponent structure. Dade oﬀered the ﬁrst counter-examples [7]. The groups in G also have iso- p,e morphic character tables and together with p-th power maps [13]. Indeed, these examples have the largest possible character tables with that property, n n speciﬁcally of size × (the largest any character table can be is n × n). 2 2 p p All noncentral conjugacy classes in H/N have the same size. The groups are both directly and centrally indecomposable and with the same algebraic type of indecomposability (an invariant introduced in [17, Theorem 4.41; 18, Theorem 8] that links indecomposability to isomorphism types of local commutative rings and local Jordan algebras). Barnes-Wall [3] show that the lattice of a nilpotent group of class 2 and exponent p determines the isomorphism type of the group (which corrects an errant remark of the author). The groups in G have maximum sized p,e Θ(log |G|) lattices with |G| subgroups, chains of length log |G| and antichains Θ(log |G|) of length |G| . We have no tools to compare such large lattices. Despite similarities, isomorphism in G is easy to test. p,e Theorem 1.2 ([6,13]). (a) There is a deterministic algorithm that, given a black-box group G, determines if G = H/N for some N < H and if so returns a surjection H → G. The timing is polynomial in e + p. (b) There is a deterministic algorithm that given groups G , G ∈ G , de- 1 2 p,e cides if G G . The timing is polynomial in e + p. 1 2 THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 3 We leave discussion of computational models for groups to the references just cited, which we note involve algorithms that apply to a broader class of problems. Narrowed to our speciﬁc setting where p and e are prime, the ω 2 precise complexity is O(p + e log p) where 2 ≤ ω < 3 is the exponent of feasible matrix multiplication. The leading p can be replaced by log p at the cost of a Las Vegas polynomial-time algorithm. It takes Ω((e log p) ) bits to input the groups we consider by any of the standard methods including matrices, presentations, or permutations. Some of the steps in the proof of Theorem 1.1 can be extracted as special cases of results in [5, 6, 13]. However, there is an increased need to provide an easier introduction into the methods represented in those works. We opt to make this exposition largely self-contained and we rely in as much as possible on proofs based in linear algebra. Preliminaries. We assume all groups in this note are ﬁnite. The Frattini subgroup Φ(G) is the intersection of the maximal subgroups of G. The exponent of a group G is the least positive integer m such that for every m ′ g ∈ G, g = 1. The commutator subgroup H is the smallest normal subgroup whose quotient is abelian, equivalently the subgroup generated by −1 y −1 −1 p commutators [x, y] = x x = x y xy, and G is the subgroup generated by p-th powers. The genus of G is d(Φ(G)). We require the following: ′ p Theorem 1.3 (Burnside Basis Theorem). For a p-group G, Φ(G) = G G d(G) and G/Φ(G) Z . = p 2. A formula for subgroup profiles We prove a formula that, under some hypotheses, calculates the subgroup proﬁles in p-groups. This allows us to construct groups that produce the same proﬁle without need to directly compare the groups. Theorem 2.1. Let G be a p-group in which Aut(G) acts transitively on maximal subgroups and such that for a maximal subgroup M of G, d(G) = 1 + d(M). Then for every J < G, the size of J (J) = {K < G : K = J} is d(M) 1+d(M) X ∼ p − 1 K J, K ≤ M : . p − 1 |M : KΦ(M)| = p f=0 In particular the proﬁle map J 7→ |J (J)| depends only the isomorphism type of a maximal subgroup of G. Lemma 2.2. In a p-group G with a maximal subgroup M having d(G) = 1 + d(M), it follows that Φ(G) = Φ(M). Proof. Using the Burnside Basis Theorem on G and on M we calculate: d(G) |G : Φ(G)| · |Φ(G)| p |Φ(G)| |Φ(G)| 1 = = = . 1+d(M) |G : M| · |M : Φ(M)| · |Φ(M)| |Φ(M)| p |Φ(M)| ′ p ′ p As Φ(M) = M M ≤ G G = Φ(G), we ﬁnd that Φ(M) = Φ(G). 4 JAMES B. WILSON Proof of Theorem 2.1. Fix J < G. We use an Aut(G)-invariant partition: d(G) J (J) = J (J, f), J (J, f) = {K ∈ J (J) : |G : KΦ(G)| = p }. f=1 Let M be the set of maximal subgroups of G. Fix f and deﬁne a bipartite graph between the two sets J (J, f) and M, such that (K, X) ∈ J (J, f) × M is an edge if, and only if, K ≤ X. The action of Aut(G) on this graph permutes the vertices of M transitively. In particular, the degree of every vertex X ∈ M the same as the degree of M. Apply Lemma 2.2 to conclude that Φ(G) = Φ(M). Thus, for every K ≤ M, KΦ(G) = KΦ(M) and so deg M = |{K ≤ M : K J,|G : KΦ(G)| = p }| f−1 = |{K ≤ M : K J, |M : KΦ(M)| = p }|. Next we compute the degree of K ∈ J (J, f), i.e. the size of the set: {X ∈ M : K ≤ X} = {X ∈ M : KΦ(G) ≤ X}. d(G) Since G/Φ(G) = Z and |G : KΦ(G)| = p it follows that: (G/Φ(G))/(KΦ(G)/Φ(G)) Z . In particular the number maximal subgroups of G containing K equals the number of hyperplanes in an f-dimensional Z -vector space. At this point we count the number of edges in our graph in two ways. d(G) f X X p − 1 p − 1 deg M = deg X = deg K = |J (J, f)| . p − 1 p − 1 X∈M K∈J (J,f) 1+d(M) p −1 Thus |J (J, f)| = deg M. The claim follows. p −1 3. Making p-groups with matrices We are interested in quotients of groups of (3 × 3)-matrices, but it will be easier to discuss properties of a larger class of groups. For that we use a general constructions of p-groups that has roots in studies of Brahana and Baer [2, 4]. Fix a set {L , . . . , L } of (r × s)-matrices and deﬁne the 1 t following group of matrices. Here and throughout empty blocks in matrices are presumed to be 0. 1 a c t t r I L b ··· L b a ∈ Z r 1 t p (3.1) B(L , . . . , L ) = : b ∈ Z . 1 t c ∈ Z THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 5 The dimension of hL , . . . , L i is the genus g of the group. If r, s, g ≈ n/3 1 t 3 2 n /27+Θ(n ) then this construction already deﬁnes p isomorphism types of groups of order p , which is approximately a square-root of all the possible groups of order p . So despite humble appearance, this family is extremely complex. Our most important examples will be the Heisenberg groups. As our ﬁelds F are ﬁnite, there exists an ω ∈ F such that F = Z (ω). q q q p e−1 In particular, {1, ω, . . . , ω } is a basis for F as a Z -vector space. Deﬁne q p (k) m(ω) ∈ Z as the constants such that: ij e−1 (k) i j k ω · ω = m(ω) ω . ij k=0 (k) (k) (k) Also let M(ω) ∈ M (Z ) be such that [M(ω) ] = m(ω) . e p (i+1)(j+1) ij (0) (e−1) Example 3.2. If F e = Z (ω) then H(F ) B(M(ω) , . . . , M(ω) ). p p q In the following section we will use the groups B(L , L ) to give an alter- 1 2 native description of the groups H/N ∈ G . We will prove: p,e Theorem 3.3. The groups in G are isomorphic to the groups B(L , L ) p,e 1 2 where {L , L } is a linearly independent set of (e × e)-matrices with L 1 2 1 −1 invertible and L L has an irreducible minimum polynomial of degree e. 3.1. Subgroups by row, column, and matrix elimination. One way to explore the subgroups of the groups B(L , . . . , L ) is to restrict the range 1 g of values of a or b in the formula given in (3.1). For instance, suppose we restrict the coordinate a = 0. The result is that the values in the i-th row of each matrix L , . . . , L can be ignored within that subgroup. Hence 1 g the subgroup we get is isomorphic to the group we obtain by ﬁrst removing the i-th row of each matrix in {L , . . . , L } and then using the construction 1 g of (3.1) to create a group on these smaller matrices. Removing one row produces a maximal subgroup, two rows a subgroup of index p , and so on. The similar idea applies to columns. Reversing the process and inserting rows or columns creates subgroup embeddings. Restricting values of c may result in a subset that is not closed to multi- plication. An easy way to avoid that concern is to eliminate entries c only once the corresponding matrix L = 0. Example 3.4. Using row, column, and matrix insertion, we embed H(Z ) into H(F ). In this example we let F = Z [x]/(x + 1). We partition the 9 9 3 matrices to help identify the row or column insertions. H(Z ) = B([1]) ֒→ B([1], [0]) ֒→ B([1|0], [0|1]) 1 0 0 1 ֒→ B , = H(F 2). 0 −1 1 0 6 JAMES B. WILSON We emphasize that this approach is not guaranteed to explore every sub- group, but it is nevertheless a good place to begin. Next we can construct a family of groups each having a maximal subgroup of a ﬁxed isomorphism type. As in [12, p. 70], for a polynomial a(t) = 0 e−1 e a t + ··· + a t + t ∈ Z [t], the companion matrix will be: 0 e−1 p 0 1 . . . . . . C(a(t)) = . 0 1 a ··· a 0 e−1 Lemma 3.5. For a polynomial a(t) of degree e, the group G = B(I , C(a(t))) has a maximal subgroup M whose isomorphism type depends only on p and e and d(G) = 1 + d(M). Proof. We delete the last row of I and C(a(t)) to obtain: e e z }| { z }| { 1 0 0 1 . . . . . . . . M = B , ֒→ . . . . 1 0 0 1 1 0 0 1 . . . . . . . . . . . . B , = G. 1 0 0 1 0 ··· 1 a ··· a 0 e−1 2e−1 d(G)−1 Evidently |M : Φ(M)| = p = p . 3.2. Quotient groups by linear combinations. Next we will want to explore some of quotient groups of B(L , . . . , L ). One can see in (3.1) 1 g that for each subset {i , . . . , i } ⊆ {1, . . . , g}, there is a natural surjection 1 s B(L , . . . , L ) → B(L , . . . , L ) and that is indeed a group homomorphism. 1 g i i 1 s This is an analogue to the way we created subgroups in the previous section. Likewise, ﬁx scalars (a , . . . , a ). Then there is a surjective homomor- 1 g phism B(L , . . . , L ) → B(a L + ··· + a L ). More generally given a 1 g 1 1 g g (g × g)-matrix A, there is a surjective homomorphism: g g X X B(L , . . . , L ) 7→ B A L , . . . , A ′ L . 1 g 1j j g j j j=1 j=1 3.3. Notable isomorphisms. There are also direct ways to create groups isomorphic to B(L , . . . , L ). For example, for invertible matrices X ∈ 1 g M (Z ) and Y ∈ M (Z ), n p m p t t B(L , . . . , L ) = B(XL Y , . . . , XL Y ). 1 g 1 g THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 7 We may also permute the order of the matrices. In particular we can always insist the ﬁrst matrix have largest rank and that it be expressed in the I 0 form through Gaussian elimination. Thus the groups B(L ) can be 0 0 classiﬁed up to isomorphism by the rank of L . More generally, for each i, j ∈ {1, . . . , g}, and s ∈ Z , ∼ ∼ B(L , . . . , L ) = B(L , . . . , L + sL , . . . , L ) = B(L , . . . , sL , . . . , L ). 1 g 1 i j g 1 i g ′ ′ Thus, if {L , . . . , L } is another basis for hL , . . . , L i, then B(L , . . . , L ) = g 1 g 1 g ′ ′ B(L , . . . , L ). There can be further isomorphisms between these groups, 1 g but these will suﬃce for our present discussion. Using these observations we can make even more complex embeddings. Lemma 3.6. For every a(t), b(t) ∈ Z [t] with deg b(t) =: f and less than e := deg a(t), there is an embedding B(I , C(b(t))) into B(I , C(a(t))). We emphasize that b(t) has no relation to a(t) other than having lower degree. So there is no algebraic reason to guess at the possible embed- ding of B(I , C(b(t))) into B(I , C(a(t))). Yet with matrices it is an easy f e calculation. Proof. Fix b , . . . , b ∈ Z . 0 e−2 p 1 0 1 0 . . . . . . . . . (3.7) = , . . . . 1 0 1 0 b . . . b 1 0 e−2 0 1 0 1 . . . . . . . . . (3.8) = . . . . . 0 1 b . . . b 1 0 e−2 b . . . b 1 0 e−2 0 e−2 e−1 Thus, setting b(t) = b t + ··· + b t + t , we obtain the following 0 e−2 t t embedding. For the isomorphism we are using the identity B(L Y , L Y ) 1 2 B(L , L ) following the calculation of (3.7) and (3.8). 1 2 1 0 0 0 1 0 . . . . . . . . . . . . B(I , C(b(t))) ֒→ B , e−1 . . . . . . 1 0 b . . . b 1 0 e−2 1 0 0 1 . . . . . . . . = B , . . . . 1 0 0 1 ֒→ B(I , C(a(t))). In light of Theorem 3.3, Lemma 3.6 shows that the groups H/N ∈ G p,e Ω(e) each have p isomorphism types of proper subgroups. 8 JAMES B. WILSON Proposition 3.9. Given (e × e)-matrices (L , . . . , L ) with L invertible, 1 g 1 ˜ ˜ there are polynomials a (t)|··· |a (t) and matrices L , . . . , L such that 1 m 3 g ˜ ˜ B(L , L ) = B(I , C(a (t)) ⊕ ··· ⊕ C(a (t)), L , . . . , L ). 1 2 e 1 m 3 g Proof. Use the Frobenius Normal Form [12, p. 93] to ﬁnd a divisor chain a (t)|··· |a (t) and an invertible matrix X such that 1 m −1 −1 X (L L )X = C(a (t)) ⊕ ··· ⊕ C(a (t)). 2 1 m Hence, −1 −1 B(L , L , . . . , L ) = B(I , L L , . . . , L L ) 1 2 g e 2 g 1 1 −1 −1 B(I , C(a (t)) ⊕ ··· ⊕ C(a (t)), . . . , X L L X). e 1 m g −1 −1 So for 3 ≤ i ≤ g, set L = X L L X. i 2 4. Isomorphisms between quotients of Heisenberg groups. We have so far created many groups and demonstrated the ease to which we can control the construction of interesting subgroups and quotients. Our eﬀort now shifts back to Heisenberg groups and in particular we will tackle the question of isomorphisms and automorphisms within G . Our main p,e results in this section are proofs of Theorem 3.3 and: Theorem 4.1. Every isomorphism between nonabelian quotients of H of genus g > 1 lifts to an automorphism of H. This is a special case of [13, Theorem 4.4]. Here we provide a self- contained and largely matrix-based proof. 4.1. The role of commutation. The ﬁrst principle in nilpotent group the- −1 y ory is to treat groups like rings by invoking commutation [x, y] = x x = −1 −1 x y xy as a skew-commutative multiplication. This very nearly dis- tributes over the usual product, in the following way. y z (4.2) [xy, z] = [x, z] [y, z], [x, yz] = [x, z][x, y] . With q = p and H = H(F ), commutation takes the following form. ′ ′ ′ ′ 1 α γ 1 α γ 1 0 αβ − α β (4.3) 1 β , 1 β = 1 0 . 1 1 1 This shows the following two groups are abelian. 1 0 γ ′ ′ m H = [H, H] = 1 0 : γ ∈ F , H/H = {(α, β) : α, β ∈ F }. ′ 2 ′ Evidently there are isomorphisms ι : H/H → hF , +i and ˆι : H → hF , +i. None of these isomorphisms is natural in the category of groups. In partic- ′ ′ ular neither H/H nor H is an obvious F -vector space as scalar multipli- cation is not part of the operations of a group. THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 9 Normal subgroups are now easily described. ′ ′ Lemma 4.4. For h ∈ H − H , [h, H] = H ; thus, if M is normal in H then ′ ′ ′ ′ either H ≤ N or N ≤ H . In either case, (H/N) = H N/N. 4.2. Quotients of H. To inspect the quotients of H we use a method to “linearize” a nilpotent group which is in some sense the reversal of the constructions we gave in Section 3. Early versions of this approach were described by Brahana and Baer [2,4]. Since elements in H = [H, H] commute with the whole group, the iden- ′ ′ ′ tities (4.2) imply that commutation factors through H/H × H/H → H 2 2 and thereby aﬀords a biadditive map [, ] : hF , +i × hF , +i → hF , +i: + q q q ′ ′ ′ ′ (4.5) [(α, β), (α , β )] = αβ − α β. To distinguish between the various roles of [, ] we let [, ] denote group com- mutation and [, ] the biadditive mapping that commutation produces. Remark 4.6. The expression in (4.5) is obviously F -bilinear. However, the ′ ′ ′ relationship of [, ] to the commutation map [, ] : H/H ×H/H → H is only as abelian groups, made explicitly through the (unnatural) choice of (ι, ˆι) above. So geometric information about F -bilinear maps cannot be directly applied in our situation. ′ ′ ′ Now Lemma 4.4 shows that for N < H , (H/N) = H /N. So the com- mutation of the quotient H/N will accordingly aﬀord a new biadditive map H/N π 2 2 g [, ] : hF , +i × hF , +i → hF , +i → Z + q q p ′ ′ ∼ ∼ where π is given as the homomorphism hF , +i = H → H /N = Z . The q p genus of H/N is the value g. Let us look closely at the case of genus g = 1. Fix π : hF , +i → Z . q p Choose a basis {α , . . . , α } for hF , +i as a Z -vector space and such that 1 e q p π(α ) = 1 if i = 1 and 0 otherwise. Deﬁne L = π([(α , 0), (0, α )]) = π(α α ). ij i j i j Regarded as a map of Z -vector spaces we see: 0 L α H/N ′ ′ [(α, β), (α , β )] = α β . t ′ −L 0 β As we vary H/N amongst groups of arbitrary genus 1 ≤ g ≤ e we de- H/N scribe [, ] = [, ] by a linearly independent set of invertible matrices L , . . . , L ∈ M (Z ) such that 1 g e p ′ ′ 0 L α 0 L α ′ ′ 1 g [(α, β), (α , β )] = α β , . . . , α β . + t ′ t ′ −L 0 β −L 0 β 1 g This demonstrates the following correspondence. Theorem 4.7 (Brahana correspondence). A group H/N whose commuta- tion is described by matrices (L , . . . , L ) has an isomorphism to the group 1 g B(L , . . . , L ). In particular all quotients H/N of genus 1 are isomorphic. 1 g 10 JAMES B. WILSON Proof. If g = 1 we can assume L = I and so H/N = B(I ). The required 1 e e isomorphism B(L , . . . , L ) → H/N is as follows. 1 g (4.8) 1 a c t t −1 −1 I L b ··· L b 1 ι (a) ˆι (c) e 1 g −1 7→ 1 ι (b) mod N. Corollary 4.9. The groups in G have the equal quotient group proﬁles. p,e ′ 2e+2 2e+2 ′ ′ Proof. Fix N ≤ H with |H : N| = p . As p = |H : H | · |H : N| = 2e ′ ′ 2 p |H : N| we ﬁnd |H : N| = p , and so H/N has genus 2. As in Lemma 4.4, ′ ′ if N < K ≤ H and K/N is normal in H/N, then K < H or H ≤ K. If ′ f ∼ ∼ H ≤ K then (H/K)/(K/N) H/K Z where p = |H : K|. This does = = not depend on the choice of N. The number of choices for K is the number of 2e subgroups in Z of index f, which again does not depend on N. Otherwise ′ ′ N < K < H and so |H : K| = p. Thus (H/N)/(K/N) = H/K has genus 1. So by Theorem 4.7 its isomorphism type is ﬁxed and independent ′ 2 of N. Finally, H /N = Z so there are exactly p + 1 choices of K with N < K < H . This is independent of N. 4.3. Distributive products. To prove Theorem 4.1 we need a brief detour to discuss distributive products. Take A ⊂ M (Z )×M (Z ). It follows that r p s p M (Z ) decomposes into subspaces as follows. r×s p X ∈ M (Z ) r×s p t ∗ ∗ t ∗ F X − XF : ⊕ {X : ∀(F, F ) ∈ A, F X = XF }. (F, F ) ∈ A r s We write Z ⊗ Z for the right-hand subspace. The projection π from A A p p r s M (Z ) onto Z ⊗ Z allows us to deﬁne a distributive tensor product r×s p A p p r s r s t ⊗ = ⊗ : Z × Z → Z ⊗ Z u ⊗ v = π (u v). A A A p p p p ∗ t ∗ Notice for (F, F ) ∈ A, π (F X) = π (XF ) and so we ﬁnd: A A t t t ∗ ∗ uF ⊗ v = π (F u v) = π (u vF ) = u ⊗ (vF ). A A Consider an example with α β δ −β (4.10) A = , : α, β, γ, δ ∈ F . γ δ −γ α THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 11 For (α, β), (γ, δ) ∈ hF , +i, α β γ δ (α, β) ⊗ (γ, δ) = (1, 0) ⊗ (1, 0) 0 0 0 0 γ δ 0 −β = (1, 0) ⊗ (1, 0) = (1, 0) ⊗ (αδ − βγ, 0). 0 0 0 α Therefore (α, β) ⊗ (γ, δ) 7→ αδ − βγ deﬁnes an isomorphism 2 2 hF , +i ⊗ hF , +i = hF , +i A q q q 2 2 2 2 and furthermore ⊗ : hF , +i × hF , +i → hF , +i ⊗ hF , +i is equivalent A A q q q q to [, ] . That the commutation of the Heisenberg group is a tensor product over a matrix ring is at the core of how Theorem 1.1 is possible. r s t ∗ In general for a distributive product ∗ : Z × Z → Z a pair (F, F ) ∈ p p p r s M (Z ) × M (Z ) is an adjoint if it satisﬁes, for all u ∈ Z and v ∈ Z , r p s p p p (uF)◦v = u◦(vF ). (This is the same notion of adjoints we ﬁnd in texts on linear algebra, cf. [12, p. 143], but we use it on arbitrary products not just inner products.) The adjoint identity is linear and so it deﬁnes a subspace: ∗ r s ∗ Adj(◦) = {(F, F ) : ∀u ∈ Z ,∀v ∈ Z , (uF) ◦ v = u ◦ (vF )}. p p ∗ ∗ ∗ ∗ Under the product (F, F )(G, G ) = (FG, G F ) this makes Adj(◦) into a ring. In fact Adj(◦) is the largest ring A over which the product ◦ factors through the tensor ⊗ , more precisely: Theorem 4.11 (Adjoint-tensor Galois correspondence [5, Theorem 2.11]). r s t Fix a distributive product ◦ : Z ×Z → Z and A ⊂ M (Z )×M (Z ). Then r p s p p p p r s t A ⊂ Adj(◦) if, and only if, there is a homomorphism ˆ◦ : Z ⊗ Z → Z p p p such that u ◦ v = ˆ◦(u ⊗ v). Now we refocus on the goal of Theorem 4.1. e e Lemma 4.12. If F ⊆ A ⊆ M (Z ) and e prime, then A = F or M (Z ). p e p p e p Proof. Let V = Z be an A-module. As F is contained in A, V is also an F -vector space, and it is 1-dimensional. Thus, as an A-module V is simple. Now A is also faithfully represented on V . Thus by Jacobson’s ∼ ∼ Density Theorem [12, p. 262], A is End (V ) = M (Z), where Z = F Z f p is the center of A. Furthermore, e = fs. As e is prime either f = 1 and A = F , or else f = e and s = 1 which makes A = M (Z ). p e p H/N Lemma 4.13. If H/N has genus g > 1 then Adj([, ] ) = Adj([, ] ) = + + M (F ). 2 q Proof. We start by observing some necessary adjoints. The adjoint-tensor H/N Galois correspondence shows Adj([, ] ) ⊂ Adj([, ] ). In our example we + + found M (F ) Adj([, ] ). 2 q + 12 JAMES B. WILSON Next we know that the commutation in H/N is given by a set {L , . . . , L } 1 g of linearly independent invertible matrices. So the linear equations to solve H/N to describe Adj([, ] ) are the following. For each 1 ≤ i ≤ g, ∗ ∗ F F 0 L 0 L F F 11 12 i i 11 12 t t ∗ ∗ F F −L 0 −L 0 F F 21 22 i i 21 22 −t −t −t ∗ t ∗ t ∗ t For i = 1 we get F = L F L , F = −L F L , F = −L F L , 1 1 1 11 22 1 12 12 1 21 21 1 −t −1 ∗ t and F = L F L . Now L adds the further constraint that F L L = 1 2 ij 2 22 11 1 1 −1 L L F . 2 ij Now consider the algebra −1 −1 A = {F ∈ M (Z ) : FL L = L L F}. e p 2 2 1 1 By the previous inclusion we know that F ⊆ A ⊆ M (Z ). If A = M (Z ) q e p e p then L commutes with every matrix and thus L is a scalar matrix. How- 2 2 ever, L and L = I are linearly independent. So L cannot be scalar. As 2 1 e 2 a result A 6= M (Z ). By Lemma 4.12, A = F . That is, e p q H/N α β δ −β Adj([, ] ) ⊆ , : α, β, γ, δ ∈ F . γ δ −γ α H/N So indeed Adj([, ] ) = Adj([, ] ). + + Proof Theorem 3.3. Fix a group B(L , L ). If B(L , L ) is a quotient of H 1 2 1 2 then so is B(L ). By Corollary 4.9, B(L ) B(I ). Therefore we may 1 1 e assume L = I and let a(t) be the minimum polynomial of L . As {L , L } 1 e 2 1 2 are linearly independent we know that L cannot be a scalar matrix and so a(t) has degree at least 2. Now let C(L ) = {F ∈ M (Z ) : FL = L F}. Following the calculation 2 e p 2 2 of the adjoint ring above we know that t t F F F −F H/N 11 12 22 12 Adj([, ] ) = , : F L = L F t t ij 2 2 ij F F −F F 21 22 21 22 = M (C(L )). 2 2 ∼ ∼ Thus, if B(L , L ) is a quotient of H then C(L ) F . Since Z [L ] = = 1 2 2 q p 2 Z [t]/(a(t)) ⊂ C(L ) it follows that Z [L ] is a subﬁeld of F . As a(t) p 2 p 2 q has degree greater than 1 and F has no intermediate ﬁelds, it follows that Z [L ] = F . Thus a(t) is an irreducible polynomial of degree e. p 2 q H/N ∼ ∼ Conversely if L is conjugate to C(a(t)) then Adj([, ] ) M (F ) = = 2 2 q Adj([, ] ). By Adjoint-Tensor Galois correspondence, the commutation in B(I , L ) factors through the tensor product over Adj([, ] ) which is the e 2 commutation of H. Therefore B(I , L ) is a quotient of H. e 2 4.4. Automorphisms of Heisenberg groups. Now we need to consider the automorphisms of H, assuming p > 2. Each automorphism is described by three constituents: (1) a homomorphism τ : hF , +i → hF , +i, q THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 13 α β (2) an invertible matrix over F , and γ δ (3) a ﬁeld automorphism α 7→ α¯ of F . The corresponding automorphism is as follows. ′ ′ ′ ′ ′ ′ ′ 1 α γ 1 α α + β γ (αδ − βγ)γ + τ(α , β ) ′ ′ (4.14) 1 β 7→ . 1 α β + β δ Remark 4.15. Classic knowledge of automorphisms of Heisenberg groups over Z , Z, R and C is largely inapplicable here. For R and C the automor- phisms are presumed to be smooth, ours have no such restrictions. As we 2 2 cautioned in Remark 4.6, in the case of F , [, ] : hF , +i×hF , +i → hF , +i q + q q q ′ ′ ′ is F -bilinear but [, ] : H/H × H/H → H is only biadditive. So only the cases of Z and Z are immediate by standard geometric methods. We have just seen that the commutation of Heisenberg groups is actually a special type of distributive product, a tensor product. This means instead of acting on a biadditive map we can act on a ring Adj([, ] ) = M (F ). + 2 q Theorem 4.16 (Skolem-Noether [12, p. 237]). The ring automorphisms of −1 M (F ) are X 7→ T XT where T is an invertible 2 × 2 matrix and α → α¯ 2 q is a ﬁeld automorphism of F applied to each entry of X. Proof. First the automorphism φ will send αI 7→ α¯I which gives us the ﬁeld 2 2 −1 automorphism σ. Replacing φ with φ(X ) we now have an F -linear au- α β tomorphism. Therefore it maps the minimal right ideal : α, β ∈ F 0 0 0 0 to another minimal right ideal : γ, δ ∈ F , or for some ν ∈ F , q q γ δ nh i o γ δ : γ, δ ∈ F . Each of these is a 2-dimensional vector spaces over F q q νγ νδ so that transformation can be given by an invertible square matrix T . Lemma 4.17. There is an epimorphism Aut(H) → Aut(M (F )). The 2 q kernel consists of those automorphisms that are the identity on H/H . Proof. Let φ : H → H be an automorphism. Since φ([h, k]) = [φ(h), φ(k)], 2e ′ ′ 2e ∼ ∼ φ factors through Z H/H → H/H Z . So we let T be the matrix = = p p representing that transformation. Also we let T be the matrix describing e ′ ′ e ∼ ∼ ˆ the restriction of φ to Z H → H Z . Notice (T, T ) satisfy = = p p ′ ′ ′ ′ [(α, β)T, (α , β )T ] = [(α, β), (α , β )] T. + + Now take (F, F ) ∈ Adj([, ] ). It follows that −1 ′ ′ −1 ′ ′ −1 [(α, β)T FT, (α , β )] = [(α, β)T F, (α , β )T ] T + + ′ ′ −1 ∗ = [(α, β), (α , β )T F T ] . In this way Aut(H) acts on Adj([, ] ) = M (F ). We saw that commuta- + 2 q tion in Aut(H) is the same as the tensor product with Adj([, ] ), so every automorphism of Adj([, ] ) determines an automorphism of H. + 14 JAMES B. WILSON Proof of Theorem 4.1. In the Brahana correspondence we saw that every nonabelian quotient H/N is determined up to isomorphism by the matri- H/N ces (L , . . . , L ) which also deﬁne [, ] . Fix an isomorphism φ : H/N → 1 g 1 ′ ′ 2 ∼ ∼ H/N . Since (H/N )/(H/N ) = H/H = hF , +i, we see φ determines a ma- 2 i i A B trix T = ∈ M (Z ). Using the fact that hF , +i = hF , +i ⊗ 2e p q q M (F ) 2 q C D hF , +i, we deﬁne Γ : H → H as follows. P P 1 a (a , b ) ⊗ (x , y ) 1 aA + bC (a , b )T ⊗ (x , y )T i i i i i i i i i i 1 b 7→ 1 aB + bD . 1 1 From our proof of Lemma 4.17 we notice Γ is an automorphism of H if, and −1 only if, T Adj([, ] )T = Adj([, ] ). Since φ is an isomorphism H/N → + + 1 H/N we know that H/N H/N 1 2 −1 T Adj [, ] T = Adj [, ] . + + H/N By Lemma 4.13 we know Adj ([, ] ) = Adj [, ] . e−3 Corollary 4.18. The set G has at least p /e isomorphism types. p,e ′ 2 Proof. The number of subgroups N < H of index p is the number of ′ e subspaces of codimension 2 in a vector space H = Z . That number is e e (p −1)(p −p) ′ e . Meanwhile the action by Aut(H) on H has size e(p − 1); see (p −1)(p−1) e−3 (4.14). So the number of orbits is at least p /e. 5. Proof of Theorem 1.1 Lemma 5.1. For every N ≤ H , Aut(H/N) acts transitively on the maxi- mal subgroups of H/N. 2e Proof. The group SL(2,F ) acts transitively on hyperplanes of Z and those coincide with the maximal subgroups of H. Following (4.14), this action lifts to Aut(H) and is furthermore the identity on H . Thus for N < H , this action transfers to H/N. Lastly, observe that Φ(H/N) = H /N = Φ(H)/N, so the maximal subgroups of H/N are the groups X/N where X is maximal in H. Proof of Theorem 1.1. Corollary 4.18 & 4.9 establish that the set G has p,e e−3 at least p /e isomorphism types and that these groups all have the same quotient group proﬁle. Finally, Lemmas 5.1 & 3.5 allow us to invoke Theo- rem 2.1 to conclude the proof. Acknowledgment. Thanks to Laci Babai & Gene Luks who asked me about proﬁles. Thanks to Bill Kantor for extensive feedback on this article. THE THRESHOLD FOR SUBGROUP PROFILES TO AGREE IS Ω(log n). 15 References [1] L´ aszl´ o Babai, Graph Isomorphism in Quasipolynomial Time. arXiv:1512.03547. [2] Reinhold Baer, Groups with abelian central quotient group, Trans. Amer. Math. Soc. 44 (1938), no. 3, 357–386. MR1501972 [3] D. W. Barnes and G. E. Wall, On normaliser preserving lattice isomorphisms between nilpotent groups, J. Austral. Math. Soc. 4 (1964), 454–469. MR0173712 [4] H. R. Brahana, Metabelian groups and trilinear forms, Duke Math. J. 1 (1935), no. 2, 185–197. MR1545875 [5] Peter A. Brooksbank and James B. Wilson, Groups acting on tensor products, J. Pure Appl. Algebra 218 (2014), no. 3, 405–416. MR3124207 [6] P. A. Brooksbank, J. Maglione, and J. B. Wilson, A fast isomorphism test for groups whose Lie algebra has genus 2, J. Algebra (2016). (in press) (arXiv:1508.03033). [7] E. C. Dade, Answer to a question of R. Brauer, J. Algebra 1 (1964), 1–4. MR0170957 [8] Bettina Eick, C. R. Leedham-Green, and E. A. O’Brien, Constructing automorphism groups of p-groups, Comm. Algebra 30 (2002), no. 5, 2271–2295. MR1904637 [9] George Glauberman and L ukasz Grabowski, Groups with identical k-proﬁles, Theory Comput. 11 (2015), 395–401. MR3437134 [10] Tim Gowers, Comment on Dick Lipton’s blog entry: The Group isomorphism Problem: A Possible Polymath Problem?. Blog en- try started November 7, 2011. Comment cited: November 12, 2011. http://rjlipton.wordpress.com/2011/11/07/the-group-isomorphism-problem-a-possiblepolymath-problem/. [11] Robert M. Guralnick, On the number of generators of a ﬁnite group, Arch. Math. (Basel) 53 (1989), no. 6, 521–523. MR1023965 [12] Nathan Jacobson, Lectures in abstract algebra. Vol. II. Linear algebra, D. Van Nos- trand Co., Inc., Toronto-New York-London, 1953. MR0053905 [13] Mark L. Lewis and James B. Wilson, Isomorphism in expanding families of indistin- guishable groups, Groups Complex. Cryptol. 4 (2012), no. 1, 73–110. MR2921156 [14] Gary L. Miller, Graph isomorphism, general remarks, J. Comput. System Sci. 18 (1979), no. 2, 128–142. MR532172 [15] Z. Hedrl´ ın and A. Pultr, On full embeddings of categories of algebras, Illinois J. Math. 10 (1966), 392–406. MR0191858 [16] Ada Rottlaender, Nachweis der Existenz nicht-isomorpher Gruppen von gleicher Sit- uation der Untergruppen, Math. Z. 28 (1928), no. 1, 641–653. MR1544982 [17] James B. Wilson, Decomposing p-groups via Jordan algebras, J. Algebra 322 (2009), no. 8, 2642–2679. MR2559855 [18] , Existence, algorithms, and asymptotics of direct product decompositions, I, Groups Complex. Cryptol. 4 (2012), no. 1, 33–72. MR2921155 Department of Mathematics, Colorado State University, Fort Collins, CO 80523, E-mail address : James.Wilson@ColoState.Edu

Mathematics – arXiv (Cornell University)

**Published: ** Dec 5, 2016

Loading...

You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!

Read and print from thousands of top scholarly journals.

System error. Please try again!

Already have an account? Log in

Bookmark this article. You can see your Bookmarks on your DeepDyve Library.

To save an article, **log in** first, or **sign up** for a DeepDyve account if you don’t already have one.

Copy and paste the desired citation format or use the link below to download a file formatted for EndNote

Access the full text.

Sign up today, get DeepDyve free for 14 days.

All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.